From cbfeb267cb0ff632dbc8ff02685012bee2e87434 Mon Sep 17 00:00:00 2001 From: John Kacur Date: Thu, 24 Sep 2009 18:01:51 +0200 Subject: perf annotate: Add the cmp_null function and make use of it This function exists in builtin-report.c but not in builtin-annotate.c Functions that use cmp_null are shorter and clearer. Synchronizing functions between these two files will also make it easier to potential share code in the future. Signed-off-by: John Kacur Cc: Peter Zijlstra LKML-Reference: Signed-off-by: Ingo Molnar --- tools/perf/builtin-annotate.c | 30 ++++++++++++++---------------- 1 file changed, 14 insertions(+), 16 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c index 1ec74161581..a33087328bd 100644 --- a/tools/perf/builtin-annotate.c +++ b/tools/perf/builtin-annotate.c @@ -82,6 +82,16 @@ struct sort_entry { size_t (*print)(FILE *fp, struct hist_entry *); }; +static int64_t cmp_null(void *l, void *r) +{ + if (!l && !r) + return 0; + else if (!l) + return -1; + else + return 1; +} + /* --sort pid */ static int64_t @@ -116,14 +126,8 @@ sort__comm_collapse(struct hist_entry *left, struct hist_entry *right) char *comm_l = left->thread->comm; char *comm_r = right->thread->comm; - if (!comm_l || !comm_r) { - if (!comm_l && !comm_r) - return 0; - else if (!comm_l) - return -1; - else - return 1; - } + if (!comm_l || !comm_r) + return cmp_null(comm_l, comm_r); return strcmp(comm_l, comm_r); } @@ -149,14 +153,8 @@ sort__dso_cmp(struct hist_entry *left, struct hist_entry *right) struct dso *dso_l = left->dso; struct dso *dso_r = right->dso; - if (!dso_l || !dso_r) { - if (!dso_l && !dso_r) - return 0; - else if (!dso_l) - return -1; - else - return 1; - } + if (!dso_l || !dso_r) + return cmp_null(dso_l, dso_r); return strcmp(dso_l->name, dso_r->name); } -- cgit v1.2.3-18-g5258 From 8b40f521cf1c9750eab0c04da9075e7484675e9c Mon Sep 17 00:00:00 2001 From: John Kacur Date: Thu, 24 Sep 2009 18:02:18 +0200 Subject: perf tools: Protect header files with a consistent style There was a colorful mix of header guards - standardize them. Signed-off-by: John Kacur LKML-Reference: Signed-off-by: Ingo Molnar --- tools/perf/util/cache.h | 6 +++--- tools/perf/util/callchain.h | 2 +- tools/perf/util/color.h | 6 +++--- tools/perf/util/debug.h | 4 ++++ tools/perf/util/event.h | 3 ++- tools/perf/util/exec_cmd.h | 6 +++--- tools/perf/util/header.h | 6 +++--- tools/perf/util/help.h | 6 +++--- tools/perf/util/levenshtein.h | 6 +++--- tools/perf/util/module.h | 6 +++--- tools/perf/util/parse-events.h | 6 +++--- tools/perf/util/parse-options.h | 6 +++--- tools/perf/util/quote.h | 6 +++--- tools/perf/util/run-command.h | 6 +++--- tools/perf/util/sigchain.h | 6 +++--- tools/perf/util/strbuf.h | 6 +++--- tools/perf/util/string.h | 6 +++--- tools/perf/util/strlist.h | 6 +++--- tools/perf/util/svghelper.h | 6 +++--- tools/perf/util/symbol.h | 6 +++--- tools/perf/util/thread.h | 5 +++++ tools/perf/util/trace-event.h | 6 +++--- tools/perf/util/types.h | 6 +++--- tools/perf/util/values.h | 6 +++--- 24 files changed, 72 insertions(+), 62 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/cache.h b/tools/perf/util/cache.h index 6f8ea9d210b..f26172c0c91 100644 --- a/tools/perf/util/cache.h +++ b/tools/perf/util/cache.h @@ -1,5 +1,5 @@ -#ifndef CACHE_H -#define CACHE_H +#ifndef __PERF_CACHE_H +#define __PERF_CACHE_H #include "util.h" #include "strbuf.h" @@ -117,4 +117,4 @@ extern char *perf_pathdup(const char *fmt, ...) extern size_t strlcpy(char *dest, const char *src, size_t size); -#endif /* CACHE_H */ +#endif /* __PERF_CACHE_H */ diff --git a/tools/perf/util/callchain.h b/tools/perf/util/callchain.h index 43cf3ea9e08..ad4626de4c2 100644 --- a/tools/perf/util/callchain.h +++ b/tools/perf/util/callchain.h @@ -58,4 +58,4 @@ static inline u64 cumul_hits(struct callchain_node *node) int register_callchain_param(struct callchain_param *param); void append_chain(struct callchain_node *root, struct ip_callchain *chain, struct symbol **syms); -#endif +#endif /* __PERF_CALLCHAIN_H */ diff --git a/tools/perf/util/color.h b/tools/perf/util/color.h index 58d597564b9..24e8809210b 100644 --- a/tools/perf/util/color.h +++ b/tools/perf/util/color.h @@ -1,5 +1,5 @@ -#ifndef COLOR_H -#define COLOR_H +#ifndef __PERF_COLOR_H +#define __PERF_COLOR_H /* "\033[1;38;5;2xx;48;5;2xxm\0" is 23 bytes */ #define COLOR_MAXLEN 24 @@ -39,4 +39,4 @@ int color_fwrite_lines(FILE *fp, const char *color, size_t count, const char *bu int percent_color_fprintf(FILE *fp, const char *fmt, double percent); const char *get_percent_color(double percent); -#endif /* COLOR_H */ +#endif /* __PERF_COLOR_H */ diff --git a/tools/perf/util/debug.h b/tools/perf/util/debug.h index 437eea58ce4..02d1fa1c246 100644 --- a/tools/perf/util/debug.h +++ b/tools/perf/util/debug.h @@ -1,4 +1,6 @@ /* For debugging general purposes */ +#ifndef __PERF_DEBUG_H +#define __PERF_DEBUG_H extern int verbose; extern int dump_trace; @@ -6,3 +8,5 @@ extern int dump_trace; int eprintf(const char *fmt, ...) __attribute__((format(printf, 1, 2))); int dump_printf(const char *fmt, ...) __attribute__((format(printf, 1, 2))); void trace_event(event_t *event); + +#endif /* __PERF_DEBUG_H */ diff --git a/tools/perf/util/event.h b/tools/perf/util/event.h index 2c9c26d6ded..c31a5da6458 100644 --- a/tools/perf/util/event.h +++ b/tools/perf/util/event.h @@ -1,5 +1,6 @@ #ifndef __PERF_RECORD_H #define __PERF_RECORD_H + #include "../perf.h" #include "util.h" #include @@ -101,4 +102,4 @@ struct map *map__clone(struct map *self); int map__overlap(struct map *l, struct map *r); size_t map__fprintf(struct map *self, FILE *fp); -#endif +#endif /* __PERF_RECORD_H */ diff --git a/tools/perf/util/exec_cmd.h b/tools/perf/util/exec_cmd.h index effe25eb154..31647ac92ed 100644 --- a/tools/perf/util/exec_cmd.h +++ b/tools/perf/util/exec_cmd.h @@ -1,5 +1,5 @@ -#ifndef PERF_EXEC_CMD_H -#define PERF_EXEC_CMD_H +#ifndef __PERF_EXEC_CMD_H +#define __PERF_EXEC_CMD_H extern void perf_set_argv_exec_path(const char *exec_path); extern const char *perf_extract_argv0_path(const char *path); @@ -10,4 +10,4 @@ extern int execv_perf_cmd(const char **argv); /* NULL terminated */ extern int execl_perf_cmd(const char *cmd, ...); extern const char *system_path(const char *path); -#endif /* PERF_EXEC_CMD_H */ +#endif /* __PERF_EXEC_CMD_H */ diff --git a/tools/perf/util/header.h b/tools/perf/util/header.h index a0761bc7863..a2916b652a1 100644 --- a/tools/perf/util/header.h +++ b/tools/perf/util/header.h @@ -1,5 +1,5 @@ -#ifndef _PERF_HEADER_H -#define _PERF_HEADER_H +#ifndef __PERF_HEADER_H +#define __PERF_HEADER_H #include "../../../include/linux/perf_event.h" #include @@ -44,4 +44,4 @@ perf_header__find_attr(u64 id, struct perf_header *header); struct perf_header *perf_header__new(void); -#endif /* _PERF_HEADER_H */ +#endif /* __PERF_HEADER_H */ diff --git a/tools/perf/util/help.h b/tools/perf/util/help.h index 7128783637b..7f5c6dedd71 100644 --- a/tools/perf/util/help.h +++ b/tools/perf/util/help.h @@ -1,5 +1,5 @@ -#ifndef HELP_H -#define HELP_H +#ifndef __PERF_HELP_H +#define __PERF_HELP_H struct cmdnames { size_t alloc; @@ -26,4 +26,4 @@ int is_in_cmdlist(struct cmdnames *c, const char *s); void list_commands(const char *title, struct cmdnames *main_cmds, struct cmdnames *other_cmds); -#endif /* HELP_H */ +#endif /* __PERF_HELP_H */ diff --git a/tools/perf/util/levenshtein.h b/tools/perf/util/levenshtein.h index 0173abeef52..b0fcb6d8a88 100644 --- a/tools/perf/util/levenshtein.h +++ b/tools/perf/util/levenshtein.h @@ -1,8 +1,8 @@ -#ifndef LEVENSHTEIN_H -#define LEVENSHTEIN_H +#ifndef __PERF_LEVENSHTEIN_H +#define __PERF_LEVENSHTEIN_H int levenshtein(const char *string1, const char *string2, int swap_penalty, int substition_penalty, int insertion_penalty, int deletion_penalty); -#endif +#endif /* __PERF_LEVENSHTEIN_H */ diff --git a/tools/perf/util/module.h b/tools/perf/util/module.h index 8a592ef641c..098e0412bc2 100644 --- a/tools/perf/util/module.h +++ b/tools/perf/util/module.h @@ -1,5 +1,5 @@ -#ifndef _PERF_MODULE_ -#define _PERF_MODULE_ 1 +#ifndef __PERF_MODULE_ +#define __PERF_MODULE_ 1 #include #include "../types.h" @@ -50,4 +50,4 @@ size_t mod_dso__fprintf(struct mod_dso *self, FILE *fp); struct module *mod_dso__find_module(struct mod_dso *self, const char *name); int mod_dso__load_modules(struct mod_dso *dso); -#endif /* _PERF_MODULE_ */ +#endif /* __PERF_MODULE_ */ diff --git a/tools/perf/util/parse-events.h b/tools/perf/util/parse-events.h index 30c60811284..8626a439033 100644 --- a/tools/perf/util/parse-events.h +++ b/tools/perf/util/parse-events.h @@ -1,5 +1,5 @@ -#ifndef _PARSE_EVENTS_H -#define _PARSE_EVENTS_H +#ifndef __PERF_PARSE_EVENTS_H +#define __PERF_PARSE_EVENTS_H /* * Parse symbolic events/counts passed in as options: */ @@ -31,4 +31,4 @@ extern char debugfs_path[]; extern int valid_debugfs_mount(const char *debugfs); -#endif /* _PARSE_EVENTS_H */ +#endif /* __PERF_PARSE_EVENTS_H */ diff --git a/tools/perf/util/parse-options.h b/tools/perf/util/parse-options.h index 2ee248ff27e..948805af43c 100644 --- a/tools/perf/util/parse-options.h +++ b/tools/perf/util/parse-options.h @@ -1,5 +1,5 @@ -#ifndef PARSE_OPTIONS_H -#define PARSE_OPTIONS_H +#ifndef __PERF_PARSE_OPTIONS_H +#define __PERF_PARSE_OPTIONS_H enum parse_opt_type { /* special types */ @@ -174,4 +174,4 @@ extern int parse_opt_verbosity_cb(const struct option *, const char *, int); extern const char *parse_options_fix_filename(const char *prefix, const char *file); -#endif +#endif /* __PERF_PARSE_OPTIONS_H */ diff --git a/tools/perf/util/quote.h b/tools/perf/util/quote.h index a5454a1d1c1..b6a01973391 100644 --- a/tools/perf/util/quote.h +++ b/tools/perf/util/quote.h @@ -1,5 +1,5 @@ -#ifndef QUOTE_H -#define QUOTE_H +#ifndef __PERF_QUOTE_H +#define __PERF_QUOTE_H #include #include @@ -65,4 +65,4 @@ extern void perl_quote_print(FILE *stream, const char *src); extern void python_quote_print(FILE *stream, const char *src); extern void tcl_quote_print(FILE *stream, const char *src); -#endif +#endif /* __PERF_QUOTE_H */ diff --git a/tools/perf/util/run-command.h b/tools/perf/util/run-command.h index cc1837deba8..d79028727ce 100644 --- a/tools/perf/util/run-command.h +++ b/tools/perf/util/run-command.h @@ -1,5 +1,5 @@ -#ifndef RUN_COMMAND_H -#define RUN_COMMAND_H +#ifndef __PERF_RUN_COMMAND_H +#define __PERF_RUN_COMMAND_H enum { ERR_RUN_COMMAND_FORK = 10000, @@ -85,4 +85,4 @@ struct async { int start_async(struct async *async); int finish_async(struct async *async); -#endif +#endif /* __PERF_RUN_COMMAND_H */ diff --git a/tools/perf/util/sigchain.h b/tools/perf/util/sigchain.h index 618083bce0c..1a53c11265f 100644 --- a/tools/perf/util/sigchain.h +++ b/tools/perf/util/sigchain.h @@ -1,5 +1,5 @@ -#ifndef SIGCHAIN_H -#define SIGCHAIN_H +#ifndef __PERF_SIGCHAIN_H +#define __PERF_SIGCHAIN_H typedef void (*sigchain_fun)(int); @@ -8,4 +8,4 @@ int sigchain_pop(int sig); void sigchain_push_common(sigchain_fun f); -#endif /* SIGCHAIN_H */ +#endif /* __PERF_SIGCHAIN_H */ diff --git a/tools/perf/util/strbuf.h b/tools/perf/util/strbuf.h index d2aa86c014c..a3d121d6c83 100644 --- a/tools/perf/util/strbuf.h +++ b/tools/perf/util/strbuf.h @@ -1,5 +1,5 @@ -#ifndef STRBUF_H -#define STRBUF_H +#ifndef __PERF_STRBUF_H +#define __PERF_STRBUF_H /* * Strbuf's can be use in many ways: as a byte array, or to store arbitrary @@ -134,4 +134,4 @@ extern int launch_editor(const char *path, struct strbuf *buffer, const char *co extern int strbuf_branchname(struct strbuf *sb, const char *name); extern int strbuf_check_branch_ref(struct strbuf *sb, const char *name); -#endif /* STRBUF_H */ +#endif /* __PERF_STRBUF_H */ diff --git a/tools/perf/util/string.h b/tools/perf/util/string.h index bf39dfadfd2..15c827475e7 100644 --- a/tools/perf/util/string.h +++ b/tools/perf/util/string.h @@ -1,5 +1,5 @@ -#ifndef _PERF_STRING_H_ -#define _PERF_STRING_H_ +#ifndef __PERF_STRING_H_ +#define __PERF_STRING_H_ #include "types.h" @@ -8,4 +8,4 @@ int hex2u64(const char *ptr, u64 *val); #define _STR(x) #x #define STR(x) _STR(x) -#endif +#endif /* __PERF_STRING_H */ diff --git a/tools/perf/util/strlist.h b/tools/perf/util/strlist.h index 921818e44a5..cb4659306d7 100644 --- a/tools/perf/util/strlist.h +++ b/tools/perf/util/strlist.h @@ -1,5 +1,5 @@ -#ifndef STRLIST_H_ -#define STRLIST_H_ +#ifndef __PERF_STRLIST_H +#define __PERF_STRLIST_H #include #include @@ -36,4 +36,4 @@ static inline unsigned int strlist__nr_entries(const struct strlist *self) } int strlist__parse_list(struct strlist *self, const char *s); -#endif /* STRLIST_H_ */ +#endif /* __PERF_STRLIST_H */ diff --git a/tools/perf/util/svghelper.h b/tools/perf/util/svghelper.h index cd93195aedb..e0781989cc3 100644 --- a/tools/perf/util/svghelper.h +++ b/tools/perf/util/svghelper.h @@ -1,5 +1,5 @@ -#ifndef _INCLUDE_GUARD_SVG_HELPER_ -#define _INCLUDE_GUARD_SVG_HELPER_ +#ifndef __PERF_SVGHELPER_H +#define __PERF_SVGHELPER_H #include "types.h" @@ -25,4 +25,4 @@ extern void svg_close(void); extern int svg_page_width; -#endif +#endif /* __PERF_SVGHELPER_H */ diff --git a/tools/perf/util/symbol.h b/tools/perf/util/symbol.h index 6e849071640..ee164f659ed 100644 --- a/tools/perf/util/symbol.h +++ b/tools/perf/util/symbol.h @@ -1,5 +1,5 @@ -#ifndef _PERF_SYMBOL_ -#define _PERF_SYMBOL_ 1 +#ifndef __PERF_SYMBOL +#define __PERF_SYMBOL 1 #include #include "types.h" @@ -89,4 +89,4 @@ extern struct dso *vdso; extern struct dso *hypervisor_dso; extern const char *vmlinux_name; extern int modules; -#endif /* _PERF_SYMBOL_ */ +#endif /* __PERF_SYMBOL */ diff --git a/tools/perf/util/thread.h b/tools/perf/util/thread.h index 32aea3c1c2a..693ed1ea10b 100644 --- a/tools/perf/util/thread.h +++ b/tools/perf/util/thread.h @@ -1,3 +1,6 @@ +#ifndef __PERF_THREAD_H +#define __PERF_THREAD_H + #include #include #include @@ -20,3 +23,5 @@ void thread__insert_map(struct thread *self, struct map *map); int thread__fork(struct thread *self, struct thread *parent); struct map *thread__find_map(struct thread *self, u64 ip); size_t threads__fprintf(FILE *fp, struct rb_root *threads); + +#endif /* __PERF_THREAD_H */ diff --git a/tools/perf/util/trace-event.h b/tools/perf/util/trace-event.h index 693f815c942..162c3e6deb9 100644 --- a/tools/perf/util/trace-event.h +++ b/tools/perf/util/trace-event.h @@ -1,5 +1,5 @@ -#ifndef _TRACE_EVENTS_H -#define _TRACE_EVENTS_H +#ifndef __PERF_TRACE_EVENTS_H +#define __PERF_TRACE_EVENTS_H #include "parse-events.h" @@ -242,4 +242,4 @@ void *raw_field_ptr(struct event *event, const char *name, void *data); void read_tracing_data(struct perf_event_attr *pattrs, int nb_events); -#endif /* _TRACE_EVENTS_H */ +#endif /* __PERF_TRACE_EVENTS_H */ diff --git a/tools/perf/util/types.h b/tools/perf/util/types.h index 5e75f900594..7d6b8331f89 100644 --- a/tools/perf/util/types.h +++ b/tools/perf/util/types.h @@ -1,5 +1,5 @@ -#ifndef _PERF_TYPES_H -#define _PERF_TYPES_H +#ifndef __PERF_TYPES_H +#define __PERF_TYPES_H /* * We define u64 as unsigned long long for every architecture @@ -14,4 +14,4 @@ typedef signed short s16; typedef unsigned char u8; typedef signed char s8; -#endif /* _PERF_TYPES_H */ +#endif /* __PERF_TYPES_H */ diff --git a/tools/perf/util/values.h b/tools/perf/util/values.h index cadf8cf2a59..2fa967e1a88 100644 --- a/tools/perf/util/values.h +++ b/tools/perf/util/values.h @@ -1,5 +1,5 @@ -#ifndef _PERF_VALUES_H -#define _PERF_VALUES_H +#ifndef __PERF_VALUES_H +#define __PERF_VALUES_H #include "types.h" @@ -24,4 +24,4 @@ void perf_read_values_add_value(struct perf_read_values *values, void perf_read_values_display(FILE *fp, struct perf_read_values *values, int raw); -#endif /* _PERF_VALUES_H */ +#endif /* __PERF_VALUES_H */ -- cgit v1.2.3-18-g5258 From dd68ada2d417e57b848822a1407b5317a54136c5 Mon Sep 17 00:00:00 2001 From: John Kacur Date: Thu, 24 Sep 2009 18:02:49 +0200 Subject: perf tools: Create util/sort.and use it Create util/sort.[ch] and move common functionality for builtin-report.c and builtin-annotate.c there, and make use of it. Signed-off-by: John Kacur LKML-Reference: Signed-off-by: Ingo Molnar --- tools/perf/Makefile | 2 + tools/perf/builtin-annotate.c | 211 +---------------------------- tools/perf/builtin-report.c | 307 +----------------------------------------- tools/perf/util/sort.c | 268 ++++++++++++++++++++++++++++++++++++ tools/perf/util/sort.h | 93 +++++++++++++ 5 files changed, 373 insertions(+), 508 deletions(-) create mode 100644 tools/perf/util/sort.c create mode 100644 tools/perf/util/sort.h (limited to 'tools/perf') diff --git a/tools/perf/Makefile b/tools/perf/Makefile index b5f1953b614..0a9e5aede31 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -339,6 +339,7 @@ LIB_H += util/symbol.h LIB_H += util/module.h LIB_H += util/color.h LIB_H += util/values.h +LIB_H += util/sort.h LIB_OBJS += util/abspath.o LIB_OBJS += util/alias.o @@ -374,6 +375,7 @@ LIB_OBJS += util/trace-event-parse.o LIB_OBJS += util/trace-event-read.o LIB_OBJS += util/trace-event-info.o LIB_OBJS += util/svghelper.o +LIB_OBJS += util/sort.o BUILTIN_OBJS += builtin-annotate.o BUILTIN_OBJS += builtin-help.o diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c index a33087328bd..059c565b31e 100644 --- a/tools/perf/builtin-annotate.c +++ b/tools/perf/builtin-annotate.c @@ -22,12 +22,10 @@ #include "util/parse-options.h" #include "util/parse-events.h" #include "util/thread.h" +#include "util/sort.h" static char const *input_name = "perf.data"; -static char default_sort_order[] = "comm,symbol"; -static char *sort_order = default_sort_order; - static int force; static int input; static int show_mask = SHOW_KERNEL | SHOW_USER | SHOW_HV; @@ -55,207 +53,6 @@ struct sym_ext { static struct rb_root hist; -struct hist_entry { - struct rb_node rb_node; - - struct thread *thread; - struct map *map; - struct dso *dso; - struct symbol *sym; - u64 ip; - char level; - - uint32_t count; -}; - -/* - * configurable sorting bits - */ - -struct sort_entry { - struct list_head list; - - const char *header; - - int64_t (*cmp)(struct hist_entry *, struct hist_entry *); - int64_t (*collapse)(struct hist_entry *, struct hist_entry *); - size_t (*print)(FILE *fp, struct hist_entry *); -}; - -static int64_t cmp_null(void *l, void *r) -{ - if (!l && !r) - return 0; - else if (!l) - return -1; - else - return 1; -} - -/* --sort pid */ - -static int64_t -sort__thread_cmp(struct hist_entry *left, struct hist_entry *right) -{ - return right->thread->pid - left->thread->pid; -} - -static size_t -sort__thread_print(FILE *fp, struct hist_entry *self) -{ - return fprintf(fp, "%16s:%5d", self->thread->comm ?: "", self->thread->pid); -} - -static struct sort_entry sort_thread = { - .header = " Command: Pid", - .cmp = sort__thread_cmp, - .print = sort__thread_print, -}; - -/* --sort comm */ - -static int64_t -sort__comm_cmp(struct hist_entry *left, struct hist_entry *right) -{ - return right->thread->pid - left->thread->pid; -} - -static int64_t -sort__comm_collapse(struct hist_entry *left, struct hist_entry *right) -{ - char *comm_l = left->thread->comm; - char *comm_r = right->thread->comm; - - if (!comm_l || !comm_r) - return cmp_null(comm_l, comm_r); - - return strcmp(comm_l, comm_r); -} - -static size_t -sort__comm_print(FILE *fp, struct hist_entry *self) -{ - return fprintf(fp, "%16s", self->thread->comm); -} - -static struct sort_entry sort_comm = { - .header = " Command", - .cmp = sort__comm_cmp, - .collapse = sort__comm_collapse, - .print = sort__comm_print, -}; - -/* --sort dso */ - -static int64_t -sort__dso_cmp(struct hist_entry *left, struct hist_entry *right) -{ - struct dso *dso_l = left->dso; - struct dso *dso_r = right->dso; - - if (!dso_l || !dso_r) - return cmp_null(dso_l, dso_r); - - return strcmp(dso_l->name, dso_r->name); -} - -static size_t -sort__dso_print(FILE *fp, struct hist_entry *self) -{ - if (self->dso) - return fprintf(fp, "%-25s", self->dso->name); - - return fprintf(fp, "%016llx ", (u64)self->ip); -} - -static struct sort_entry sort_dso = { - .header = "Shared Object ", - .cmp = sort__dso_cmp, - .print = sort__dso_print, -}; - -/* --sort symbol */ - -static int64_t -sort__sym_cmp(struct hist_entry *left, struct hist_entry *right) -{ - u64 ip_l, ip_r; - - if (left->sym == right->sym) - return 0; - - ip_l = left->sym ? left->sym->start : left->ip; - ip_r = right->sym ? right->sym->start : right->ip; - - return (int64_t)(ip_r - ip_l); -} - -static size_t -sort__sym_print(FILE *fp, struct hist_entry *self) -{ - size_t ret = 0; - - if (verbose) - ret += fprintf(fp, "%#018llx ", (u64)self->ip); - - if (self->sym) { - ret += fprintf(fp, "[%c] %s", - self->dso == kernel_dso ? 'k' : '.', self->sym->name); - } else { - ret += fprintf(fp, "%#016llx", (u64)self->ip); - } - - return ret; -} - -static struct sort_entry sort_sym = { - .header = "Symbol", - .cmp = sort__sym_cmp, - .print = sort__sym_print, -}; - -static int sort__need_collapse = 0; - -struct sort_dimension { - const char *name; - struct sort_entry *entry; - int taken; -}; - -static struct sort_dimension sort_dimensions[] = { - { .name = "pid", .entry = &sort_thread, }, - { .name = "comm", .entry = &sort_comm, }, - { .name = "dso", .entry = &sort_dso, }, - { .name = "symbol", .entry = &sort_sym, }, -}; - -static LIST_HEAD(hist_entry__sort_list); - -static int sort_dimension__add(char *tok) -{ - unsigned int i; - - for (i = 0; i < ARRAY_SIZE(sort_dimensions); i++) { - struct sort_dimension *sd = &sort_dimensions[i]; - - if (sd->taken) - continue; - - if (strncasecmp(tok, sd->name, strlen(tok))) - continue; - - if (sd->entry->collapse) - sort__need_collapse = 1; - - list_add_tail(&sd->entry->list, &hist_entry__sort_list); - sd->taken = 1; - - return 0; - } - - return -ESRCH; -} - static int64_t hist_entry__cmp(struct hist_entry *left, struct hist_entry *right) { @@ -1137,5 +934,11 @@ int cmd_annotate(int argc, const char **argv, const char *prefix __used) setup_pager(); + if (field_sep && *field_sep == '.') { + fputs("'.' is the only non valid --field-separator argument\n", + stderr); + exit(129); + } + return __cmd_annotate(); } diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index 19669c20088..7b43504900f 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -27,15 +27,13 @@ #include "util/parse-events.h" #include "util/thread.h" +#include "util/sort.h" static char const *input_name = "perf.data"; -static char default_sort_order[] = "comm,dso,symbol"; -static char *sort_order = default_sort_order; static char *dso_list_str, *comm_list_str, *sym_list_str, *col_width_list_str; static struct strlist *dso_list, *comm_list, *sym_list; -static char *field_sep; static int force; static int input; @@ -53,10 +51,6 @@ static char *pretty_printing_style = default_pretty_printing_style; static unsigned long page_size; static unsigned long mmap_window = 32; -static char default_parent_pattern[] = "^sys_|^do_page_fault"; -static char *parent_pattern = default_parent_pattern; -static regex_t parent_regex; - static int exclude_other = 1; static char callchain_default_opt[] = "fractal,0.5"; @@ -80,304 +74,8 @@ struct callchain_param callchain_param = { static u64 sample_type; -static int repsep_fprintf(FILE *fp, const char *fmt, ...) -{ - int n; - va_list ap; - - va_start(ap, fmt); - if (!field_sep) - n = vfprintf(fp, fmt, ap); - else { - char *bf = NULL; - n = vasprintf(&bf, fmt, ap); - if (n > 0) { - char *sep = bf; - - while (1) { - sep = strchr(sep, *field_sep); - if (sep == NULL) - break; - *sep = '.'; - } - } - fputs(bf, fp); - free(bf); - } - va_end(ap); - return n; -} - -static unsigned int dsos__col_width, - comms__col_width, - threads__col_width; - -/* - * histogram, sorted on item, collects counts - */ - static struct rb_root hist; -struct hist_entry { - struct rb_node rb_node; - - struct thread *thread; - struct map *map; - struct dso *dso; - struct symbol *sym; - struct symbol *parent; - u64 ip; - char level; - struct callchain_node callchain; - struct rb_root sorted_chain; - - u64 count; -}; - -/* - * configurable sorting bits - */ - -struct sort_entry { - struct list_head list; - - const char *header; - - int64_t (*cmp)(struct hist_entry *, struct hist_entry *); - int64_t (*collapse)(struct hist_entry *, struct hist_entry *); - size_t (*print)(FILE *fp, struct hist_entry *, unsigned int width); - unsigned int *width; - bool elide; -}; - -static int64_t cmp_null(void *l, void *r) -{ - if (!l && !r) - return 0; - else if (!l) - return -1; - else - return 1; -} - -/* --sort pid */ - -static int64_t -sort__thread_cmp(struct hist_entry *left, struct hist_entry *right) -{ - return right->thread->pid - left->thread->pid; -} - -static size_t -sort__thread_print(FILE *fp, struct hist_entry *self, unsigned int width) -{ - return repsep_fprintf(fp, "%*s:%5d", width - 6, - self->thread->comm ?: "", self->thread->pid); -} - -static struct sort_entry sort_thread = { - .header = "Command: Pid", - .cmp = sort__thread_cmp, - .print = sort__thread_print, - .width = &threads__col_width, -}; - -/* --sort comm */ - -static int64_t -sort__comm_cmp(struct hist_entry *left, struct hist_entry *right) -{ - return right->thread->pid - left->thread->pid; -} - -static int64_t -sort__comm_collapse(struct hist_entry *left, struct hist_entry *right) -{ - char *comm_l = left->thread->comm; - char *comm_r = right->thread->comm; - - if (!comm_l || !comm_r) - return cmp_null(comm_l, comm_r); - - return strcmp(comm_l, comm_r); -} - -static size_t -sort__comm_print(FILE *fp, struct hist_entry *self, unsigned int width) -{ - return repsep_fprintf(fp, "%*s", width, self->thread->comm); -} - -static struct sort_entry sort_comm = { - .header = "Command", - .cmp = sort__comm_cmp, - .collapse = sort__comm_collapse, - .print = sort__comm_print, - .width = &comms__col_width, -}; - -/* --sort dso */ - -static int64_t -sort__dso_cmp(struct hist_entry *left, struct hist_entry *right) -{ - struct dso *dso_l = left->dso; - struct dso *dso_r = right->dso; - - if (!dso_l || !dso_r) - return cmp_null(dso_l, dso_r); - - return strcmp(dso_l->name, dso_r->name); -} - -static size_t -sort__dso_print(FILE *fp, struct hist_entry *self, unsigned int width) -{ - if (self->dso) - return repsep_fprintf(fp, "%-*s", width, self->dso->name); - - return repsep_fprintf(fp, "%*llx", width, (u64)self->ip); -} - -static struct sort_entry sort_dso = { - .header = "Shared Object", - .cmp = sort__dso_cmp, - .print = sort__dso_print, - .width = &dsos__col_width, -}; - -/* --sort symbol */ - -static int64_t -sort__sym_cmp(struct hist_entry *left, struct hist_entry *right) -{ - u64 ip_l, ip_r; - - if (left->sym == right->sym) - return 0; - - ip_l = left->sym ? left->sym->start : left->ip; - ip_r = right->sym ? right->sym->start : right->ip; - - return (int64_t)(ip_r - ip_l); -} - -static size_t -sort__sym_print(FILE *fp, struct hist_entry *self, unsigned int width __used) -{ - size_t ret = 0; - - if (verbose) - ret += repsep_fprintf(fp, "%#018llx %c ", (u64)self->ip, - dso__symtab_origin(self->dso)); - - ret += repsep_fprintf(fp, "[%c] ", self->level); - if (self->sym) { - ret += repsep_fprintf(fp, "%s", self->sym->name); - - if (self->sym->module) - ret += repsep_fprintf(fp, "\t[%s]", - self->sym->module->name); - } else { - ret += repsep_fprintf(fp, "%#016llx", (u64)self->ip); - } - - return ret; -} - -static struct sort_entry sort_sym = { - .header = "Symbol", - .cmp = sort__sym_cmp, - .print = sort__sym_print, -}; - -/* --sort parent */ - -static int64_t -sort__parent_cmp(struct hist_entry *left, struct hist_entry *right) -{ - struct symbol *sym_l = left->parent; - struct symbol *sym_r = right->parent; - - if (!sym_l || !sym_r) - return cmp_null(sym_l, sym_r); - - return strcmp(sym_l->name, sym_r->name); -} - -static size_t -sort__parent_print(FILE *fp, struct hist_entry *self, unsigned int width) -{ - return repsep_fprintf(fp, "%-*s", width, - self->parent ? self->parent->name : "[other]"); -} - -static unsigned int parent_symbol__col_width; - -static struct sort_entry sort_parent = { - .header = "Parent symbol", - .cmp = sort__parent_cmp, - .print = sort__parent_print, - .width = &parent_symbol__col_width, -}; - -static int sort__need_collapse = 0; -static int sort__has_parent = 0; - -struct sort_dimension { - const char *name; - struct sort_entry *entry; - int taken; -}; - -static struct sort_dimension sort_dimensions[] = { - { .name = "pid", .entry = &sort_thread, }, - { .name = "comm", .entry = &sort_comm, }, - { .name = "dso", .entry = &sort_dso, }, - { .name = "symbol", .entry = &sort_sym, }, - { .name = "parent", .entry = &sort_parent, }, -}; - -static LIST_HEAD(hist_entry__sort_list); - -static int sort_dimension__add(const char *tok) -{ - unsigned int i; - - for (i = 0; i < ARRAY_SIZE(sort_dimensions); i++) { - struct sort_dimension *sd = &sort_dimensions[i]; - - if (sd->taken) - continue; - - if (strncasecmp(tok, sd->name, strlen(tok))) - continue; - - if (sd->entry->collapse) - sort__need_collapse = 1; - - if (sd->entry == &sort_parent) { - int ret = regcomp(&parent_regex, parent_pattern, REG_EXTENDED); - if (ret) { - char err[BUFSIZ]; - - regerror(ret, &parent_regex, err, sizeof(err)); - fprintf(stderr, "Invalid regex: %s\n%s", - parent_pattern, err); - exit(-1); - } - sort__has_parent = 1; - } - - list_add_tail(&sd->entry->list, &hist_entry__sort_list); - sd->taken = 1; - - return 0; - } - - return -ESRCH; -} - static int64_t hist_entry__cmp(struct hist_entry *left, struct hist_entry *right) { @@ -1606,7 +1304,8 @@ setup: return 0; } -static const char * const report_usage[] = { +//static const char * const report_usage[] = { +const char * const report_usage[] = { "perf report [] ", NULL }; diff --git a/tools/perf/util/sort.c b/tools/perf/util/sort.c new file mode 100644 index 00000000000..50e75abb1fd --- /dev/null +++ b/tools/perf/util/sort.c @@ -0,0 +1,268 @@ +#include "sort.h" + +regex_t parent_regex; +char default_parent_pattern[] = "^sys_|^do_page_fault"; +char *parent_pattern = default_parent_pattern; +char default_sort_order[] = "comm,dso,symbol"; +char *sort_order = default_sort_order; +int sort__need_collapse = 0; +int sort__has_parent = 0; + +unsigned int dsos__col_width; +unsigned int comms__col_width; +unsigned int threads__col_width; +static unsigned int parent_symbol__col_width; +char * field_sep; + +LIST_HEAD(hist_entry__sort_list); + +struct sort_entry sort_thread = { + .header = "Command: Pid", + .cmp = sort__thread_cmp, + .print = sort__thread_print, + .width = &threads__col_width, +}; + +struct sort_entry sort_comm = { + .header = "Command", + .cmp = sort__comm_cmp, + .collapse = sort__comm_collapse, + .print = sort__comm_print, + .width = &comms__col_width, +}; + +struct sort_entry sort_dso = { + .header = "Shared Object", + .cmp = sort__dso_cmp, + .print = sort__dso_print, + .width = &dsos__col_width, +}; + +struct sort_entry sort_sym = { + .header = "Symbol", + .cmp = sort__sym_cmp, + .print = sort__sym_print, +}; + +struct sort_entry sort_parent = { + .header = "Parent symbol", + .cmp = sort__parent_cmp, + .print = sort__parent_print, + .width = &parent_symbol__col_width, +}; + +struct sort_dimension { + const char *name; + struct sort_entry *entry; + int taken; +}; + +static struct sort_dimension sort_dimensions[] = { + { .name = "pid", .entry = &sort_thread, }, + { .name = "comm", .entry = &sort_comm, }, + { .name = "dso", .entry = &sort_dso, }, + { .name = "symbol", .entry = &sort_sym, }, + { .name = "parent", .entry = &sort_parent, }, +}; + +int64_t cmp_null(void *l, void *r) +{ + if (!l && !r) + return 0; + else if (!l) + return -1; + else + return 1; +} + +/* --sort pid */ + +int64_t +sort__thread_cmp(struct hist_entry *left, struct hist_entry *right) +{ + return right->thread->pid - left->thread->pid; +} + +int repsep_fprintf(FILE *fp, const char *fmt, ...) +{ + int n; + va_list ap; + + va_start(ap, fmt); + if (!field_sep) + n = vfprintf(fp, fmt, ap); + else { + char *bf = NULL; + n = vasprintf(&bf, fmt, ap); + if (n > 0) { + char *sep = bf; + + while (1) { + sep = strchr(sep, *field_sep); + if (sep == NULL) + break; + *sep = '.'; + } + } + fputs(bf, fp); + free(bf); + } + va_end(ap); + return n; +} + +size_t +sort__thread_print(FILE *fp, struct hist_entry *self, unsigned int width) +{ + return repsep_fprintf(fp, "%*s:%5d", width - 6, + self->thread->comm ?: "", self->thread->pid); +} + +size_t +sort__comm_print(FILE *fp, struct hist_entry *self, unsigned int width) +{ + return repsep_fprintf(fp, "%*s", width, self->thread->comm); +} + +/* --sort dso */ + +int64_t +sort__dso_cmp(struct hist_entry *left, struct hist_entry *right) +{ + struct dso *dso_l = left->dso; + struct dso *dso_r = right->dso; + + if (!dso_l || !dso_r) + return cmp_null(dso_l, dso_r); + + return strcmp(dso_l->name, dso_r->name); +} + +size_t +sort__dso_print(FILE *fp, struct hist_entry *self, unsigned int width) +{ + if (self->dso) + return repsep_fprintf(fp, "%-*s", width, self->dso->name); + + return repsep_fprintf(fp, "%*llx", width, (u64)self->ip); +} + +/* --sort symbol */ + +int64_t +sort__sym_cmp(struct hist_entry *left, struct hist_entry *right) +{ + u64 ip_l, ip_r; + + if (left->sym == right->sym) + return 0; + + ip_l = left->sym ? left->sym->start : left->ip; + ip_r = right->sym ? right->sym->start : right->ip; + + return (int64_t)(ip_r - ip_l); +} + + +size_t +sort__sym_print(FILE *fp, struct hist_entry *self, unsigned int width __used) +{ + size_t ret = 0; + + if (verbose) + ret += repsep_fprintf(fp, "%#018llx %c ", (u64)self->ip, + dso__symtab_origin(self->dso)); + + ret += repsep_fprintf(fp, "[%c] ", self->level); + if (self->sym) { + ret += repsep_fprintf(fp, "%s", self->sym->name); + + if (self->sym->module) + ret += repsep_fprintf(fp, "\t[%s]", + self->sym->module->name); + } else { + ret += repsep_fprintf(fp, "%#016llx", (u64)self->ip); + } + + return ret; +} + +/* --sort comm */ + +int64_t +sort__comm_cmp(struct hist_entry *left, struct hist_entry *right) +{ + return right->thread->pid - left->thread->pid; +} + +int64_t +sort__comm_collapse(struct hist_entry *left, struct hist_entry *right) +{ + char *comm_l = left->thread->comm; + char *comm_r = right->thread->comm; + + if (!comm_l || !comm_r) + return cmp_null(comm_l, comm_r); + + return strcmp(comm_l, comm_r); +} + +/* --sort parent */ + +int64_t +sort__parent_cmp(struct hist_entry *left, struct hist_entry *right) +{ + struct symbol *sym_l = left->parent; + struct symbol *sym_r = right->parent; + + if (!sym_l || !sym_r) + return cmp_null(sym_l, sym_r); + + return strcmp(sym_l->name, sym_r->name); +} + +size_t +sort__parent_print(FILE *fp, struct hist_entry *self, unsigned int width) +{ + return repsep_fprintf(fp, "%-*s", width, + self->parent ? self->parent->name : "[other]"); +} + +int sort_dimension__add(const char *tok) +{ + unsigned int i; + + for (i = 0; i < ARRAY_SIZE(sort_dimensions); i++) { + struct sort_dimension *sd = &sort_dimensions[i]; + + if (sd->taken) + continue; + + if (strncasecmp(tok, sd->name, strlen(tok))) + continue; + + if (sd->entry->collapse) + sort__need_collapse = 1; + + if (sd->entry == &sort_parent) { + int ret = regcomp(&parent_regex, parent_pattern, REG_EXTENDED); + if (ret) { + char err[BUFSIZ]; + + regerror(ret, &parent_regex, err, sizeof(err)); + fprintf(stderr, "Invalid regex: %s\n%s", + parent_pattern, err); + exit(-1); + } + sort__has_parent = 1; + } + + list_add_tail(&sd->entry->list, &hist_entry__sort_list); + sd->taken = 1; + + return 0; + } + + return -ESRCH; +} + diff --git a/tools/perf/util/sort.h b/tools/perf/util/sort.h new file mode 100644 index 00000000000..4684fd6d5c4 --- /dev/null +++ b/tools/perf/util/sort.h @@ -0,0 +1,93 @@ +#ifndef __PERF_SORT_H +#define __PERF_SORT_H +#include "../builtin.h" + +#include "util.h" + +#include "color.h" +#include +#include "cache.h" +#include +#include "symbol.h" +#include "string.h" +#include "callchain.h" +#include "strlist.h" +#include "values.h" + +#include "../perf.h" +#include "debug.h" +#include "header.h" + +#include "parse-options.h" +#include "parse-events.h" + +#include "thread.h" +#include "sort.h" + +extern regex_t parent_regex; +extern char *sort_order; +extern char default_parent_pattern[]; +extern char *parent_pattern; +extern char default_sort_order[]; +extern int sort__need_collapse; +extern int sort__has_parent; +extern char *field_sep; +extern struct sort_entry sort_comm; +extern struct sort_entry sort_dso; +extern struct sort_entry sort_sym; +extern struct sort_entry sort_parent; +extern unsigned int dsos__col_width; +extern unsigned int comms__col_width; +extern unsigned int threads__col_width; + +struct hist_entry { + struct rb_node rb_node; + + struct thread *thread; + struct map *map; + struct dso *dso; + struct symbol *sym; + struct symbol *parent; + u64 ip; + char level; + struct callchain_node callchain; + struct rb_root sorted_chain; + + u64 count; +}; + +/* + * configurable sorting bits + */ + +struct sort_entry { + struct list_head list; + + const char *header; + + int64_t (*cmp)(struct hist_entry *, struct hist_entry *); + int64_t (*collapse)(struct hist_entry *, struct hist_entry *); + size_t (*print)(FILE *fp, struct hist_entry *, unsigned int width); + unsigned int *width; + bool elide; +}; + +extern struct sort_entry sort_thread; +extern struct list_head hist_entry__sort_list; + +extern int repsep_fprintf(FILE *fp, const char *fmt, ...); +extern size_t sort__thread_print(FILE *, struct hist_entry *, unsigned int); +extern size_t sort__comm_print(FILE *, struct hist_entry *, unsigned int); +extern size_t sort__dso_print(FILE *, struct hist_entry *, unsigned int); +extern size_t sort__sym_print(FILE *, struct hist_entry *, unsigned int __used); +extern int64_t cmp_null(void *, void *); +extern int64_t sort__thread_cmp(struct hist_entry *, struct hist_entry *); +extern int64_t sort__comm_cmp(struct hist_entry *, struct hist_entry *); +extern int64_t sort__comm_collapse(struct hist_entry *, struct hist_entry *); +extern int64_t sort__dso_cmp(struct hist_entry *, struct hist_entry *); +extern int64_t sort__sym_cmp(struct hist_entry *, struct hist_entry *); +extern int64_t sort__parent_cmp(struct hist_entry *, struct hist_entry *); +extern size_t sort__parent_print(FILE *, struct hist_entry *, unsigned int); +extern int sort_dimension__add(const char *); + +#endif /* __PERF_SORT_H */ -- cgit v1.2.3-18-g5258 From 3d1d07ecd2009f65cb2091563fa21f9600c36774 Mon Sep 17 00:00:00 2001 From: John Kacur Date: Mon, 28 Sep 2009 15:32:55 +0200 Subject: perf tools: Put common histogram functions in their own file Move histogram related functions into their own files (hist.c and hist.h) and make use of them in builtin-annotate.c and builtin-report.c. Signed-off-by: John Kacur Acked-by: Frederic Weisbecker Cc: Peter Zijlstra LKML-Reference: Signed-off-by: Ingo Molnar --- tools/perf/Makefile | 2 + tools/perf/builtin-annotate.c | 152 +-------------------------------------- tools/perf/builtin-report.c | 164 +----------------------------------------- tools/perf/util/hist.c | 164 ++++++++++++++++++++++++++++++++++++++++++ tools/perf/util/hist.h | 47 ++++++++++++ 5 files changed, 216 insertions(+), 313 deletions(-) create mode 100644 tools/perf/util/hist.c create mode 100644 tools/perf/util/hist.h (limited to 'tools/perf') diff --git a/tools/perf/Makefile b/tools/perf/Makefile index 0a9e5aede31..3a99a9fda64 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -340,6 +340,7 @@ LIB_H += util/module.h LIB_H += util/color.h LIB_H += util/values.h LIB_H += util/sort.h +LIB_H += util/hist.h LIB_OBJS += util/abspath.o LIB_OBJS += util/alias.o @@ -376,6 +377,7 @@ LIB_OBJS += util/trace-event-read.o LIB_OBJS += util/trace-event-info.o LIB_OBJS += util/svghelper.o LIB_OBJS += util/sort.o +LIB_OBJS += util/hist.o BUILTIN_OBJS += builtin-annotate.o BUILTIN_OBJS += builtin-help.o diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c index 059c565b31e..df516dce954 100644 --- a/tools/perf/builtin-annotate.c +++ b/tools/perf/builtin-annotate.c @@ -23,6 +23,7 @@ #include "util/parse-events.h" #include "util/thread.h" #include "util/sort.h" +#include "util/hist.h" static char const *input_name = "perf.data"; @@ -47,45 +48,6 @@ struct sym_ext { char *path; }; -/* - * histogram, sorted on item, collects counts - */ - -static struct rb_root hist; - -static int64_t -hist_entry__cmp(struct hist_entry *left, struct hist_entry *right) -{ - struct sort_entry *se; - int64_t cmp = 0; - - list_for_each_entry(se, &hist_entry__sort_list, list) { - cmp = se->cmp(left, right); - if (cmp) - break; - } - - return cmp; -} - -static int64_t -hist_entry__collapse(struct hist_entry *left, struct hist_entry *right) -{ - struct sort_entry *se; - int64_t cmp = 0; - - list_for_each_entry(se, &hist_entry__sort_list, list) { - int64_t (*f)(struct hist_entry *, struct hist_entry *); - - f = se->collapse ?: se->cmp; - - cmp = f(left, right); - if (cmp) - break; - } - - return cmp; -} /* * collect histogram counts @@ -163,116 +125,6 @@ hist_entry__add(struct thread *thread, struct map *map, struct dso *dso, return 0; } -static void hist_entry__free(struct hist_entry *he) -{ - free(he); -} - -/* - * collapse the histogram - */ - -static struct rb_root collapse_hists; - -static void collapse__insert_entry(struct hist_entry *he) -{ - struct rb_node **p = &collapse_hists.rb_node; - struct rb_node *parent = NULL; - struct hist_entry *iter; - int64_t cmp; - - while (*p != NULL) { - parent = *p; - iter = rb_entry(parent, struct hist_entry, rb_node); - - cmp = hist_entry__collapse(iter, he); - - if (!cmp) { - iter->count += he->count; - hist_entry__free(he); - return; - } - - if (cmp < 0) - p = &(*p)->rb_left; - else - p = &(*p)->rb_right; - } - - rb_link_node(&he->rb_node, parent, p); - rb_insert_color(&he->rb_node, &collapse_hists); -} - -static void collapse__resort(void) -{ - struct rb_node *next; - struct hist_entry *n; - - if (!sort__need_collapse) - return; - - next = rb_first(&hist); - while (next) { - n = rb_entry(next, struct hist_entry, rb_node); - next = rb_next(&n->rb_node); - - rb_erase(&n->rb_node, &hist); - collapse__insert_entry(n); - } -} - -/* - * reverse the map, sort on count. - */ - -static struct rb_root output_hists; - -static void output__insert_entry(struct hist_entry *he) -{ - struct rb_node **p = &output_hists.rb_node; - struct rb_node *parent = NULL; - struct hist_entry *iter; - - while (*p != NULL) { - parent = *p; - iter = rb_entry(parent, struct hist_entry, rb_node); - - if (he->count > iter->count) - p = &(*p)->rb_left; - else - p = &(*p)->rb_right; - } - - rb_link_node(&he->rb_node, parent, p); - rb_insert_color(&he->rb_node, &output_hists); -} - -static void output__resort(void) -{ - struct rb_node *next; - struct hist_entry *n; - struct rb_root *tree = &hist; - - if (sort__need_collapse) - tree = &collapse_hists; - - next = rb_first(tree); - - while (next) { - n = rb_entry(next, struct hist_entry, rb_node); - next = rb_next(&n->rb_node); - - rb_erase(&n->rb_node, tree); - output__insert_entry(n); - } -} - -static unsigned long total = 0, - total_mmap = 0, - total_comm = 0, - total_fork = 0, - total_unknown = 0; - static int process_sample_event(event_t *event, unsigned long offset, unsigned long head) { @@ -861,7 +713,7 @@ more: dsos__fprintf(stdout); collapse__resort(); - output__resort(); + output__resort(total); find_annotations(); diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index 7b43504900f..c1a54fc8527 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -28,6 +28,7 @@ #include "util/thread.h" #include "util/sort.h" +#include "util/hist.h" static char const *input_name = "perf.data"; @@ -55,8 +56,6 @@ static int exclude_other = 1; static char callchain_default_opt[] = "fractal,0.5"; -static int callchain; - static char __cwd[PATH_MAX]; static char *cwd = __cwd; static int cwdlen; @@ -66,50 +65,8 @@ static struct thread *last_match; static struct perf_header *header; -static -struct callchain_param callchain_param = { - .mode = CHAIN_GRAPH_REL, - .min_percent = 0.5 -}; - static u64 sample_type; -static struct rb_root hist; - -static int64_t -hist_entry__cmp(struct hist_entry *left, struct hist_entry *right) -{ - struct sort_entry *se; - int64_t cmp = 0; - - list_for_each_entry(se, &hist_entry__sort_list, list) { - cmp = se->cmp(left, right); - if (cmp) - break; - } - - return cmp; -} - -static int64_t -hist_entry__collapse(struct hist_entry *left, struct hist_entry *right) -{ - struct sort_entry *se; - int64_t cmp = 0; - - list_for_each_entry(se, &hist_entry__sort_list, list) { - int64_t (*f)(struct hist_entry *, struct hist_entry *); - - f = se->collapse ?: se->cmp; - - cmp = f(left, right); - if (cmp) - break; - } - - return cmp; -} - static size_t ipchain__fprintf_graph_line(FILE *fp, int depth, int depth_mask) { int i; @@ -308,7 +265,6 @@ hist_entry_callchain__fprintf(FILE *fp, struct hist_entry *self, return ret; } - static size_t hist_entry__fprintf(FILE *fp, struct hist_entry *self, u64 total_samples) { @@ -573,117 +529,6 @@ hist_entry__add(struct thread *thread, struct map *map, struct dso *dso, return 0; } -static void hist_entry__free(struct hist_entry *he) -{ - free(he); -} - -/* - * collapse the histogram - */ - -static struct rb_root collapse_hists; - -static void collapse__insert_entry(struct hist_entry *he) -{ - struct rb_node **p = &collapse_hists.rb_node; - struct rb_node *parent = NULL; - struct hist_entry *iter; - int64_t cmp; - - while (*p != NULL) { - parent = *p; - iter = rb_entry(parent, struct hist_entry, rb_node); - - cmp = hist_entry__collapse(iter, he); - - if (!cmp) { - iter->count += he->count; - hist_entry__free(he); - return; - } - - if (cmp < 0) - p = &(*p)->rb_left; - else - p = &(*p)->rb_right; - } - - rb_link_node(&he->rb_node, parent, p); - rb_insert_color(&he->rb_node, &collapse_hists); -} - -static void collapse__resort(void) -{ - struct rb_node *next; - struct hist_entry *n; - - if (!sort__need_collapse) - return; - - next = rb_first(&hist); - while (next) { - n = rb_entry(next, struct hist_entry, rb_node); - next = rb_next(&n->rb_node); - - rb_erase(&n->rb_node, &hist); - collapse__insert_entry(n); - } -} - -/* - * reverse the map, sort on count. - */ - -static struct rb_root output_hists; - -static void output__insert_entry(struct hist_entry *he, u64 min_callchain_hits) -{ - struct rb_node **p = &output_hists.rb_node; - struct rb_node *parent = NULL; - struct hist_entry *iter; - - if (callchain) - callchain_param.sort(&he->sorted_chain, &he->callchain, - min_callchain_hits, &callchain_param); - - while (*p != NULL) { - parent = *p; - iter = rb_entry(parent, struct hist_entry, rb_node); - - if (he->count > iter->count) - p = &(*p)->rb_left; - else - p = &(*p)->rb_right; - } - - rb_link_node(&he->rb_node, parent, p); - rb_insert_color(&he->rb_node, &output_hists); -} - -static void output__resort(u64 total_samples) -{ - struct rb_node *next; - struct hist_entry *n; - struct rb_root *tree = &hist; - u64 min_callchain_hits; - - min_callchain_hits = total_samples * (callchain_param.min_percent / 100); - - if (sort__need_collapse) - tree = &collapse_hists; - - next = rb_first(tree); - - while (next) { - n = rb_entry(next, struct hist_entry, rb_node); - next = rb_next(&n->rb_node); - - rb_erase(&n->rb_node, tree); - output__insert_entry(n, min_callchain_hits); - } -} - static size_t output__fprintf(FILE *fp, u64 total_samples) { struct hist_entry *pos; @@ -778,13 +623,6 @@ print_entries: return ret; } -static unsigned long total = 0, - total_mmap = 0, - total_comm = 0, - total_fork = 0, - total_unknown = 0, - total_lost = 0; - static int validate_chain(struct ip_callchain *chain, event_t *event) { unsigned int chain_size; diff --git a/tools/perf/util/hist.c b/tools/perf/util/hist.c new file mode 100644 index 00000000000..82808dc4f8e --- /dev/null +++ b/tools/perf/util/hist.c @@ -0,0 +1,164 @@ +#include "hist.h" + +struct rb_root hist; +struct rb_root collapse_hists; +struct rb_root output_hists; +int callchain; + +struct callchain_param callchain_param = { + .mode = CHAIN_GRAPH_REL, + .min_percent = 0.5 +}; + +unsigned long total; +unsigned long total_mmap; +unsigned long total_comm; +unsigned long total_fork; +unsigned long total_unknown; +unsigned long total_lost; + +/* + * histogram, sorted on item, collects counts + */ + +int64_t +hist_entry__cmp(struct hist_entry *left, struct hist_entry *right) +{ + struct sort_entry *se; + int64_t cmp = 0; + + list_for_each_entry(se, &hist_entry__sort_list, list) { + cmp = se->cmp(left, right); + if (cmp) + break; + } + + return cmp; +} + +int64_t +hist_entry__collapse(struct hist_entry *left, struct hist_entry *right) +{ + struct sort_entry *se; + int64_t cmp = 0; + + list_for_each_entry(se, &hist_entry__sort_list, list) { + int64_t (*f)(struct hist_entry *, struct hist_entry *); + + f = se->collapse ?: se->cmp; + + cmp = f(left, right); + if (cmp) + break; + } + + return cmp; +} + +void hist_entry__free(struct hist_entry *he) +{ + free(he); +} + +/* + * collapse the histogram + */ + +void collapse__insert_entry(struct hist_entry *he) +{ + struct rb_node **p = &collapse_hists.rb_node; + struct rb_node *parent = NULL; + struct hist_entry *iter; + int64_t cmp; + + while (*p != NULL) { + parent = *p; + iter = rb_entry(parent, struct hist_entry, rb_node); + + cmp = hist_entry__collapse(iter, he); + + if (!cmp) { + iter->count += he->count; + hist_entry__free(he); + return; + } + + if (cmp < 0) + p = &(*p)->rb_left; + else + p = &(*p)->rb_right; + } + + rb_link_node(&he->rb_node, parent, p); + rb_insert_color(&he->rb_node, &collapse_hists); +} + +void collapse__resort(void) +{ + struct rb_node *next; + struct hist_entry *n; + + if (!sort__need_collapse) + return; + + next = rb_first(&hist); + while (next) { + n = rb_entry(next, struct hist_entry, rb_node); + next = rb_next(&n->rb_node); + + rb_erase(&n->rb_node, &hist); + collapse__insert_entry(n); + } +} + +/* + * reverse the map, sort on count. + */ + +void output__insert_entry(struct hist_entry *he, u64 min_callchain_hits) +{ + struct rb_node **p = &output_hists.rb_node; + struct rb_node *parent = NULL; + struct hist_entry *iter; + + if (callchain) + callchain_param.sort(&he->sorted_chain, &he->callchain, + min_callchain_hits, &callchain_param); + + while (*p != NULL) { + parent = *p; + iter = rb_entry(parent, struct hist_entry, rb_node); + + if (he->count > iter->count) + p = &(*p)->rb_left; + else + p = &(*p)->rb_right; + } + + rb_link_node(&he->rb_node, parent, p); + rb_insert_color(&he->rb_node, &output_hists); +} + +void output__resort(u64 total_samples) +{ + struct rb_node *next; + struct hist_entry *n; + struct rb_root *tree = &hist; + u64 min_callchain_hits; + + min_callchain_hits = + total_samples * (callchain_param.min_percent / 100); + + if (sort__need_collapse) + tree = &collapse_hists; + + next = rb_first(tree); + + while (next) { + n = rb_entry(next, struct hist_entry, rb_node); + next = rb_next(&n->rb_node); + + rb_erase(&n->rb_node, tree); + output__insert_entry(n, min_callchain_hits); + } +} diff --git a/tools/perf/util/hist.h b/tools/perf/util/hist.h new file mode 100644 index 00000000000..9a8daa12b43 --- /dev/null +++ b/tools/perf/util/hist.h @@ -0,0 +1,47 @@ +#ifndef __PERF_HIST_H +#define __PERF_HIST_H +#include "../builtin.h" + +#include "util.h" + +#include "color.h" +#include +#include "cache.h" +#include +#include "symbol.h" +#include "string.h" +#include "callchain.h" +#include "strlist.h" +#include "values.h" + +#include "../perf.h" +#include "debug.h" +#include "header.h" + +#include "parse-options.h" +#include "parse-events.h" + +#include "thread.h" +#include "sort.h" + +extern struct rb_root hist; +extern struct rb_root collapse_hists; +extern struct rb_root output_hists; +extern int callchain; +extern struct callchain_param callchain_param; +extern unsigned long total; +extern unsigned long total_mmap; +extern unsigned long total_comm; +extern unsigned long total_fork; +extern unsigned long total_unknown; +extern unsigned long total_lost; + +extern int64_t hist_entry__cmp(struct hist_entry *, struct hist_entry *); +extern int64_t hist_entry__collapse(struct hist_entry *, struct hist_entry *); +extern void hist_entry__free(struct hist_entry *); +extern void collapse__insert_entry(struct hist_entry *); +extern void collapse__resort(void); +extern void output__insert_entry(struct hist_entry *, u64); +extern void output__resort(u64); + +#endif /* __PERF_HIST_H */ -- cgit v1.2.3-18-g5258 From 1b46cddfccfec4cc67b187fb53d78198de6a057c Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 28 Sep 2009 14:48:46 -0300 Subject: perf tools: Use rb_tree for maps MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Threads can have many and kernel modules will be represented as a tree of maps as well. Ah, and for a perf.data with 146607 samples: Before: [root@doppio ~]# perf stat -r 5 perf report > /dev/null Performance counter stats for 'perf report' (5 runs): 699.823680 task-clock-msecs # 0.991 CPUs ( +- 0.454% ) 74 context-switches # 0.000 M/sec ( +- 1.709% ) 2 CPU-migrations # 0.000 M/sec ( +- 17.008% ) 23114 page-faults # 0.033 M/sec ( +- 0.000% ) 1381257019 cycles # 1973.721 M/sec ( +- 0.290% ) 1456894438 instructions # 1.055 IPC ( +- 0.007% ) 18779818 cache-references # 26.835 M/sec ( +- 0.380% ) 641799 cache-misses # 0.917 M/sec ( +- 1.200% ) 0.705972729 seconds time elapsed ( +- 0.501% ) [root@doppio ~]# After Performance counter stats for 'perf report' (5 runs): 691.261451 task-clock-msecs # 0.993 CPUs ( +- 0.307% ) 72 context-switches # 0.000 M/sec ( +- 0.829% ) 6 CPU-migrations # 0.000 M/sec ( +- 18.409% ) 23127 page-faults # 0.033 M/sec ( +- 0.000% ) 1366395876 cycles # 1976.670 M/sec ( +- 0.153% ) 1443136016 instructions # 1.056 IPC ( +- 0.012% ) 17956402 cache-references # 25.976 M/sec ( +- 0.325% ) 661924 cache-misses # 0.958 M/sec ( +- 1.335% ) 0.696127275 seconds time elapsed ( +- 0.377% ) I.e. we see some speedup too. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Peter Zijlstra Cc: Mike Galbraith Cc: "H. Peter Anvin" LKML-Reference: <20090928174846.GA3361@ghostprotocols.net> Signed-off-by: Ingo Molnar --- tools/perf/Makefile | 1 + tools/perf/util/event.h | 4 +- tools/perf/util/thread.c | 129 ++++++++++++++++++++++++++++++----------------- tools/perf/util/thread.h | 12 +++-- 4 files changed, 95 insertions(+), 51 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/Makefile b/tools/perf/Makefile index 3a99a9fda64..055290a5b83 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -341,6 +341,7 @@ LIB_H += util/color.h LIB_H += util/values.h LIB_H += util/sort.h LIB_H += util/hist.h +LIB_H += util/thread.h LIB_OBJS += util/abspath.o LIB_OBJS += util/alias.o diff --git a/tools/perf/util/event.h b/tools/perf/util/event.h index c31a5da6458..4c69eb55380 100644 --- a/tools/perf/util/event.h +++ b/tools/perf/util/event.h @@ -3,7 +3,7 @@ #include "../perf.h" #include "util.h" -#include +#include enum { SHOW_KERNEL = 1, @@ -79,7 +79,7 @@ typedef union event_union { } event_t; struct map { - struct list_head node; + struct rb_node rb_node; u64 start; u64 end; u64 pgoff; diff --git a/tools/perf/util/thread.c b/tools/perf/util/thread.c index 45efb5db0d1..9d0945cc66d 100644 --- a/tools/perf/util/thread.c +++ b/tools/perf/util/thread.c @@ -15,7 +15,7 @@ static struct thread *thread__new(pid_t pid) self->comm = malloc(32); if (self->comm) snprintf(self->comm, 32, ":%d", self->pid); - INIT_LIST_HEAD(&self->maps); + self->maps = RB_ROOT; } return self; @@ -31,11 +31,13 @@ int thread__set_comm(struct thread *self, const char *comm) static size_t thread__fprintf(struct thread *self, FILE *fp) { - struct map *pos; + struct rb_node *nd; size_t ret = fprintf(fp, "Thread %d %s\n", self->pid, self->comm); - list_for_each_entry(pos, &self->maps, node) + for (nd = rb_first(&self->maps); nd; nd = rb_next(nd)) { + struct map *pos = rb_entry(nd, struct map, rb_node); ret += map__fprintf(pos, fp); + } return ret; } @@ -93,42 +95,90 @@ register_idle_thread(struct rb_root *threads, struct thread **last_match) return thread; } -void thread__insert_map(struct thread *self, struct map *map) +static void thread__remove_overlappings(struct thread *self, struct map *map) { - struct map *pos, *tmp; - - list_for_each_entry_safe(pos, tmp, &self->maps, node) { - if (map__overlap(pos, map)) { - if (verbose >= 2) { - printf("overlapping maps:\n"); - map__fprintf(map, stdout); - map__fprintf(pos, stdout); - } - - if (map->start <= pos->start && map->end > pos->start) - pos->start = map->end; - - if (map->end >= pos->end && map->start < pos->end) - pos->end = map->start; - - if (verbose >= 2) { - printf("after collision:\n"); - map__fprintf(pos, stdout); - } - - if (pos->start >= pos->end) { - list_del_init(&pos->node); - free(pos); - } + struct rb_node *next = rb_first(&self->maps); + + while (next) { + struct map *pos = rb_entry(next, struct map, rb_node); + next = rb_next(&pos->rb_node); + + if (!map__overlap(pos, map)) + continue; + + if (verbose >= 2) { + printf("overlapping maps:\n"); + map__fprintf(map, stdout); + map__fprintf(pos, stdout); + } + + if (map->start <= pos->start && map->end > pos->start) + pos->start = map->end; + + if (map->end >= pos->end && map->start < pos->end) + pos->end = map->start; + + if (verbose >= 2) { + printf("after collision:\n"); + map__fprintf(pos, stdout); + } + + if (pos->start >= pos->end) { + rb_erase(&pos->rb_node, &self->maps); + free(pos); } } +} + +void maps__insert(struct rb_root *maps, struct map *map) +{ + struct rb_node **p = &maps->rb_node; + struct rb_node *parent = NULL; + const u64 ip = map->start; + struct map *m; + + while (*p != NULL) { + parent = *p; + m = rb_entry(parent, struct map, rb_node); + if (ip < m->start) + p = &(*p)->rb_left; + else + p = &(*p)->rb_right; + } + + rb_link_node(&map->rb_node, parent, p); + rb_insert_color(&map->rb_node, maps); +} + +struct map *maps__find(struct rb_root *maps, u64 ip) +{ + struct rb_node **p = &maps->rb_node; + struct rb_node *parent = NULL; + struct map *m; + + while (*p != NULL) { + parent = *p; + m = rb_entry(parent, struct map, rb_node); + if (ip < m->start) + p = &(*p)->rb_left; + else if (ip > m->end) + p = &(*p)->rb_right; + else + return m; + } + + return NULL; +} - list_add_tail(&map->node, &self->maps); +void thread__insert_map(struct thread *self, struct map *map) +{ + thread__remove_overlappings(self, map); + maps__insert(&self->maps, map); } int thread__fork(struct thread *self, struct thread *parent) { - struct map *map; + struct rb_node *nd; if (self->comm) free(self->comm); @@ -136,7 +186,8 @@ int thread__fork(struct thread *self, struct thread *parent) if (!self->comm) return -ENOMEM; - list_for_each_entry(map, &parent->maps, node) { + for (nd = rb_first(&parent->maps); nd; nd = rb_next(nd)) { + struct map *map = rb_entry(nd, struct map, rb_node); struct map *new = map__clone(map); if (!new) return -ENOMEM; @@ -146,20 +197,6 @@ int thread__fork(struct thread *self, struct thread *parent) return 0; } -struct map *thread__find_map(struct thread *self, u64 ip) -{ - struct map *pos; - - if (self == NULL) - return NULL; - - list_for_each_entry(pos, &self->maps, node) - if (ip >= pos->start && ip <= pos->end) - return pos; - - return NULL; -} - size_t threads__fprintf(FILE *fp, struct rb_root *threads) { size_t ret = 0; diff --git a/tools/perf/util/thread.h b/tools/perf/util/thread.h index 693ed1ea10b..bbb37c1a52e 100644 --- a/tools/perf/util/thread.h +++ b/tools/perf/util/thread.h @@ -2,13 +2,12 @@ #define __PERF_THREAD_H #include -#include #include #include "symbol.h" struct thread { struct rb_node rb_node; - struct list_head maps; + struct rb_root maps; pid_t pid; char shortname[3]; char *comm; @@ -21,7 +20,14 @@ struct thread * register_idle_thread(struct rb_root *threads, struct thread **last_match); void thread__insert_map(struct thread *self, struct map *map); int thread__fork(struct thread *self, struct thread *parent); -struct map *thread__find_map(struct thread *self, u64 ip); size_t threads__fprintf(FILE *fp, struct rb_root *threads); +void maps__insert(struct rb_root *maps, struct map *map); +struct map *maps__find(struct rb_root *maps, u64 ip); + +static inline struct map *thread__find_map(struct thread *self, u64 ip) +{ + return self ? maps__find(&self->maps, ip) : NULL; +} + #endif /* __PERF_THREAD_H */ -- cgit v1.2.3-18-g5258 From a80deb622dba7dfb65d9e27b6b74b7c1963c3635 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 28 Sep 2009 15:23:51 -0300 Subject: perf sched: Remove dead code MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Several variables are not used at all, cut'n'paste leftovers. Also check if the sample_type is RAW earlier, to avoid needless searches. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: "H. Peter Anvin" Cc: Peter Zijlstra Cc: Mike Galbraith Signed-off-by: Ingo Molnar --- tools/perf/builtin-sched.c | 34 ++++------------------------------ 1 file changed, 4 insertions(+), 30 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-sched.c b/tools/perf/builtin-sched.c index ea9c15c0cdf..4470f253570 100644 --- a/tools/perf/builtin-sched.c +++ b/tools/perf/builtin-sched.c @@ -1544,16 +1544,15 @@ process_raw_event(event_t *raw_event __used, void *more_data, static int process_sample_event(event_t *event, unsigned long offset, unsigned long head) { - char level; - int show = 0; - struct dso *dso = NULL; struct thread *thread; u64 ip = event->ip.ip; u64 timestamp = -1; u32 cpu = -1; u64 period = 1; void *more_data = event->ip.__more_data; - int cpumode; + + if (!(sample_type & PERF_SAMPLE_RAW)) + return 0; thread = threads__findnew(event->ip.pid, &threads, &last_match); @@ -1589,32 +1588,7 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) return -1; } - cpumode = event->header.misc & PERF_RECORD_MISC_CPUMODE_MASK; - - if (cpumode == PERF_RECORD_MISC_KERNEL) { - show = SHOW_KERNEL; - level = 'k'; - - dso = kernel_dso; - - dump_printf(" ...... dso: %s\n", dso->name); - - } else if (cpumode == PERF_RECORD_MISC_USER) { - - show = SHOW_USER; - level = '.'; - - } else { - show = SHOW_HV; - level = 'H'; - - dso = hypervisor_dso; - - dump_printf(" ...... dso: [hypervisor]\n"); - } - - if (sample_type & PERF_SAMPLE_RAW) - process_raw_event(event, more_data, cpu, timestamp, thread); + process_raw_event(event, more_data, cpu, timestamp, thread); return 0; } -- cgit v1.2.3-18-g5258 From cad3071424edd7854f63aa80d09473e84f49ed79 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 28 Sep 2009 17:08:18 -0300 Subject: perf trace: Remove dead code MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Several variables are not used at all, cut'n'paste leftovers. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Peter Zijlstra Cc: Mike Galbraith Cc: "H. Peter Anvin" LKML-Reference: <20090928200818.GF3361@ghostprotocols.net> Signed-off-by: Ingo Molnar --- tools/perf/builtin-trace.c | 28 ---------------------------- 1 file changed, 28 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-trace.c index e9d256e2f47..2f938887335 100644 --- a/tools/perf/builtin-trace.c +++ b/tools/perf/builtin-trace.c @@ -53,16 +53,12 @@ process_comm_event(event_t *event, unsigned long offset, unsigned long head) static int process_sample_event(event_t *event, unsigned long offset, unsigned long head) { - char level; - int show = 0; - struct dso *dso = NULL; struct thread *thread; u64 ip = event->ip.ip; u64 timestamp = -1; u32 cpu = -1; u64 period = 1; void *more_data = event->ip.__more_data; - int cpumode; thread = threads__findnew(event->ip.pid, &threads, &last_match); @@ -98,30 +94,6 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) return -1; } - cpumode = event->header.misc & PERF_RECORD_MISC_CPUMODE_MASK; - - if (cpumode == PERF_RECORD_MISC_KERNEL) { - show = SHOW_KERNEL; - level = 'k'; - - dso = kernel_dso; - - dump_printf(" ...... dso: %s\n", dso->name); - - } else if (cpumode == PERF_RECORD_MISC_USER) { - - show = SHOW_USER; - level = '.'; - - } else { - show = SHOW_HV; - level = 'H'; - - dso = hypervisor_dso; - - dump_printf(" ...... dso: [hypervisor]\n"); - } - if (sample_type & PERF_SAMPLE_RAW) { struct { u32 size; -- cgit v1.2.3-18-g5258 From 2ccdc450e658053681202d42ac64b3638f22dc1a Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Thu, 24 Sep 2009 14:24:00 -0700 Subject: perf top: Remove dead {min,max}_ip unused variables MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: "H. Peter Anvin" Cc: Peter Zijlstra Cc: Mike Galbraith LKML-Reference: <20090924212400.GA15321@ghostprotocols.net> Signed-off-by: Ingo Molnar --- tools/perf/builtin-top.c | 13 ------------- 1 file changed, 13 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index 1ca88896eee..bf464ce7e3e 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -96,9 +96,6 @@ static int display_weighted = -1; * Symbols */ -static u64 min_ip; -static u64 max_ip = -1ll; - struct sym_entry { struct rb_node rb_node; struct list_head node; @@ -826,8 +823,6 @@ static int symbol_filter(struct dso *self, struct symbol *sym) static int parse_symbols(void) { - struct rb_node *node; - struct symbol *sym; int use_modules = vmlinux_name ? 1 : 0; kernel_dso = dso__new("[kernel]", sizeof(struct sym_entry)); @@ -837,14 +832,6 @@ static int parse_symbols(void) if (dso__load_kernel(kernel_dso, vmlinux_name, symbol_filter, verbose, use_modules) <= 0) goto out_delete_dso; - node = rb_first(&kernel_dso->syms); - sym = rb_entry(node, struct symbol, rb_node); - min_ip = sym->start; - - node = rb_last(&kernel_dso->syms); - sym = rb_entry(node, struct symbol, rb_node); - max_ip = sym->end; - if (dump_symtab) dso__fprintf(kernel_dso, stderr); -- cgit v1.2.3-18-g5258 From 439d473b4777de510e1322168ac6f2f377ecd5bc Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 2 Oct 2009 03:29:58 -0300 Subject: perf tools: Rewrite and improve support for kernel modules MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: "H. Peter Anvin" Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Avi Kivity Signed-off-by: Ingo Molnar --- tools/perf/Makefile | 2 - tools/perf/builtin-annotate.c | 73 +++--- tools/perf/builtin-report.c | 79 +++--- tools/perf/builtin-top.c | 74 ++---- tools/perf/util/event.h | 6 +- tools/perf/util/module.c | 545 ------------------------------------------ tools/perf/util/module.h | 53 ---- tools/perf/util/sort.c | 38 +-- tools/perf/util/sort.h | 7 +- tools/perf/util/symbol.c | 447 +++++++++++++++++++++++----------- tools/perf/util/symbol.h | 20 +- tools/perf/util/thread.c | 34 +-- tools/perf/util/thread.h | 4 + 13 files changed, 453 insertions(+), 929 deletions(-) delete mode 100644 tools/perf/util/module.c delete mode 100644 tools/perf/util/module.h (limited to 'tools/perf') diff --git a/tools/perf/Makefile b/tools/perf/Makefile index 055290a5b83..8e7509f2d88 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -336,7 +336,6 @@ LIB_H += util/strlist.h LIB_H += util/run-command.h LIB_H += util/sigchain.h LIB_H += util/symbol.h -LIB_H += util/module.h LIB_H += util/color.h LIB_H += util/values.h LIB_H += util/sort.h @@ -364,7 +363,6 @@ LIB_OBJS += util/usage.o LIB_OBJS += util/wrapper.o LIB_OBJS += util/sigchain.o LIB_OBJS += util/symbol.o -LIB_OBJS += util/module.o LIB_OBJS += util/color.o LIB_OBJS += util/pager.o LIB_OBJS += util/header.o diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c index df516dce954..7d5a3b1bcda 100644 --- a/tools/perf/builtin-annotate.c +++ b/tools/perf/builtin-annotate.c @@ -63,6 +63,7 @@ static void hist_hit(struct hist_entry *he, u64 ip) return; sym_size = sym->end - sym->start; + ip = he->map->map_ip(he->map, ip); offset = ip - sym->start; if (offset >= sym_size) @@ -80,7 +81,7 @@ static void hist_hit(struct hist_entry *he, u64 ip) } static int -hist_entry__add(struct thread *thread, struct map *map, struct dso *dso, +hist_entry__add(struct thread *thread, struct map *map, struct symbol *sym, u64 ip, char level) { struct rb_node **p = &hist.rb_node; @@ -89,7 +90,6 @@ hist_entry__add(struct thread *thread, struct map *map, struct dso *dso, struct hist_entry entry = { .thread = thread, .map = map, - .dso = dso, .sym = sym, .ip = ip, .level = level, @@ -130,10 +130,10 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) { char level; int show = 0; - struct dso *dso = NULL; struct thread *thread; u64 ip = event->ip.ip; struct map *map = NULL; + struct symbol *sym = NULL; thread = threads__findnew(event->ip.pid, &threads, &last_match); @@ -155,32 +155,35 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) if (event->header.misc & PERF_RECORD_MISC_KERNEL) { show = SHOW_KERNEL; level = 'k'; - - dso = kernel_dso; - - dump_printf(" ...... dso: %s\n", dso->name); - + sym = kernel_maps__find_symbol(ip, &map); + dump_printf(" ...... dso: %s\n", + map ? map->dso->long_name : ""); } else if (event->header.misc & PERF_RECORD_MISC_USER) { - show = SHOW_USER; level = '.'; - map = thread__find_map(thread, ip); if (map != NULL) { +got_map: ip = map->map_ip(map, ip); - dso = map->dso; + sym = map->dso->find_symbol(map->dso, ip); } else { /* * If this is outside of all known maps, * and is a negative address, try to look it * up in the kernel dso, as it might be a - * vsyscall (which executes in user-mode): + * vsyscall or vdso (which executes in user-mode). + * + * XXX This is nasty, we should have a symbol list in + * the "[vdso]" dso, but for now lets use the old + * trick of looking in the whole kernel symbol list. */ - if ((long long)ip < 0) - dso = kernel_dso; + if ((long long)ip < 0) { + map = kernel_map; + goto got_map; + } } - dump_printf(" ...... dso: %s\n", dso ? dso->name : ""); - + dump_printf(" ...... dso: %s\n", + map ? map->dso->long_name : ""); } else { show = SHOW_HV; level = 'H'; @@ -188,12 +191,7 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) } if (show & show_mask) { - struct symbol *sym = NULL; - - if (dso) - sym = dso->find_symbol(dso, ip); - - if (hist_entry__add(thread, map, dso, sym, ip, level)) { + if (hist_entry__add(thread, map, sym, ip, level)) { fprintf(stderr, "problem incrementing symbol count, skipping event\n"); return -1; @@ -313,7 +311,7 @@ process_event(event_t *event, unsigned long offset, unsigned long head) } static int -parse_line(FILE *file, struct symbol *sym, u64 start, u64 len) +parse_line(FILE *file, struct symbol *sym, u64 len) { char *line = NULL, *tmp, *tmp2; static const char *prev_line; @@ -363,7 +361,7 @@ parse_line(FILE *file, struct symbol *sym, u64 start, u64 len) const char *color; struct sym_ext *sym_ext = sym->priv; - offset = line_ip - start; + offset = line_ip - sym->start; if (offset < len) hits = sym->hist[offset]; @@ -442,7 +440,7 @@ static void free_source_line(struct symbol *sym, int len) /* Get the filename:line for the colored entries */ static void -get_source_line(struct symbol *sym, u64 start, int len, const char *filename) +get_source_line(struct symbol *sym, int len, const char *filename) { int i; char cmd[PATH_MAX * 2]; @@ -467,7 +465,7 @@ get_source_line(struct symbol *sym, u64 start, int len, const char *filename) if (sym_ext[i].percent <= 0.5) continue; - offset = start + i; + offset = sym->start + i; sprintf(cmd, "addr2line -e %s %016llx", filename, offset); fp = popen(cmd, "r"); if (!fp) @@ -519,31 +517,23 @@ static void print_summary(const char *filename) static void annotate_sym(struct dso *dso, struct symbol *sym) { - const char *filename = dso->name, *d_filename; - u64 start, end, len; + const char *filename = dso->long_name, *d_filename; + u64 len; char command[PATH_MAX*2]; FILE *file; if (!filename) return; - if (sym->module) - filename = sym->module->path; - else if (dso == kernel_dso) - filename = vmlinux_name; - - start = sym->obj_start; - if (!start) - start = sym->start; + if (full_paths) d_filename = filename; else d_filename = basename(filename); - end = start + sym->end - sym->start + 1; len = sym->end - sym->start; if (print_line) { - get_source_line(sym, start, len, filename); + get_source_line(sym, len, filename); print_summary(filename); } @@ -552,10 +542,11 @@ static void annotate_sym(struct dso *dso, struct symbol *sym) printf("------------------------------------------------\n"); if (verbose >= 2) - printf("annotating [%p] %30s : [%p] %30s\n", dso, dso->name, sym, sym->name); + printf("annotating [%p] %30s : [%p] %30s\n", + dso, dso->long_name, sym, sym->name); sprintf(command, "objdump --start-address=0x%016Lx --stop-address=0x%016Lx -dS %s|grep -v %s", - (u64)start, (u64)end, filename, filename); + sym->start, sym->end, filename, filename); if (verbose >= 3) printf("doing: %s\n", command); @@ -565,7 +556,7 @@ static void annotate_sym(struct dso *dso, struct symbol *sym) return; while (!feof(file)) { - if (parse_line(file, sym, start, len) < 0) + if (parse_line(file, sym, len) < 0) break; } diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index c1a54fc8527..3ed3baf96ff 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -349,22 +349,17 @@ static int thread__set_comm_adjust(struct thread *self, const char *comm) static struct symbol * -resolve_symbol(struct thread *thread, struct map **mapp, - struct dso **dsop, u64 *ipp) +resolve_symbol(struct thread *thread, struct map **mapp, u64 *ipp) { - struct dso *dso = dsop ? *dsop : NULL; struct map *map = mapp ? *mapp : NULL; u64 ip = *ipp; - if (!thread) - return NULL; - - if (dso) - goto got_dso; - if (map) goto got_map; + if (!thread) + return NULL; + map = thread__find_map(thread, ip); if (map != NULL) { /* @@ -379,29 +374,29 @@ resolve_symbol(struct thread *thread, struct map **mapp, *mapp = map; got_map: ip = map->map_ip(map, ip); - - dso = map->dso; } else { /* * If this is outside of all known maps, * and is a negative address, try to look it * up in the kernel dso, as it might be a - * vsyscall (which executes in user-mode): + * vsyscall or vdso (which executes in user-mode). + * + * XXX This is nasty, we should have a symbol list in + * the "[vdso]" dso, but for now lets use the old + * trick of looking in the whole kernel symbol list. */ - if ((long long)ip < 0) - dso = kernel_dso; + if ((long long)ip < 0) { + map = kernel_map; + if (mapp) + *mapp = map; + } } - dump_printf(" ...... dso: %s\n", dso ? dso->name : ""); + dump_printf(" ...... dso: %s\n", + map ? map->dso->long_name : ""); dump_printf(" ...... map: %Lx -> %Lx\n", *ipp, ip); *ipp = ip; - if (dsop) - *dsop = dso; - - if (!dso) - return NULL; -got_dso: - return dso->find_symbol(dso, ip); + return map ? map->dso->find_symbol(map->dso, ip) : NULL; } static int call__match(struct symbol *sym) @@ -413,7 +408,7 @@ static int call__match(struct symbol *sym) } static struct symbol ** -resolve_callchain(struct thread *thread, struct map *map __used, +resolve_callchain(struct thread *thread, struct map *map, struct ip_callchain *chain, struct hist_entry *entry) { u64 context = PERF_CONTEXT_MAX; @@ -430,8 +425,7 @@ resolve_callchain(struct thread *thread, struct map *map __used, for (i = 0; i < chain->nr; i++) { u64 ip = chain->ips[i]; - struct dso *dso = NULL; - struct symbol *sym; + struct symbol *sym = NULL; if (ip >= PERF_CONTEXT_MAX) { context = ip; @@ -440,17 +434,15 @@ resolve_callchain(struct thread *thread, struct map *map __used, switch (context) { case PERF_CONTEXT_HV: - dso = hypervisor_dso; break; case PERF_CONTEXT_KERNEL: - dso = kernel_dso; + sym = kernel_maps__find_symbol(ip, &map); break; default: + sym = resolve_symbol(thread, &map, &ip); break; } - sym = resolve_symbol(thread, NULL, &dso, &ip); - if (sym) { if (sort__has_parent && call__match(sym) && !entry->parent) @@ -469,7 +461,7 @@ resolve_callchain(struct thread *thread, struct map *map __used, */ static int -hist_entry__add(struct thread *thread, struct map *map, struct dso *dso, +hist_entry__add(struct thread *thread, struct map *map, struct symbol *sym, u64 ip, struct ip_callchain *chain, char level, u64 count) { @@ -480,7 +472,6 @@ hist_entry__add(struct thread *thread, struct map *map, struct dso *dso, struct hist_entry entry = { .thread = thread, .map = map, - .dso = dso, .sym = sym, .ip = ip, .level = level, @@ -641,7 +632,7 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) { char level; int show = 0; - struct dso *dso = NULL; + struct symbol *sym = NULL; struct thread *thread; u64 ip = event->ip.ip; u64 period = 1; @@ -700,35 +691,35 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) show = SHOW_KERNEL; level = 'k'; - dso = kernel_dso; - - dump_printf(" ...... dso: %s\n", dso->name); - + sym = kernel_maps__find_symbol(ip, &map); + dump_printf(" ...... dso: %s\n", + map ? map->dso->long_name : ""); } else if (cpumode == PERF_RECORD_MISC_USER) { show = SHOW_USER; level = '.'; + sym = resolve_symbol(thread, &map, &ip); } else { show = SHOW_HV; level = 'H'; - dso = hypervisor_dso; - dump_printf(" ...... dso: [hypervisor]\n"); } if (show & show_mask) { - struct symbol *sym = resolve_symbol(thread, &map, &dso, &ip); - - if (dso_list && (!dso || !dso->name || - !strlist__has_entry(dso_list, dso->name))) + if (dso_list && + (!map || !map->dso || + !(strlist__has_entry(dso_list, map->dso->short_name) || + (map->dso->short_name != map->dso->long_name && + strlist__has_entry(dso_list, map->dso->long_name))))) return 0; - if (sym_list && (!sym || !strlist__has_entry(sym_list, sym->name))) + if (sym_list && sym && !strlist__has_entry(sym_list, sym->name)) return 0; - if (hist_entry__add(thread, map, dso, sym, ip, chain, level, period)) { + if (hist_entry__add(thread, map, sym, ip, + chain, level, period)) { eprintf("problem incrementing symbol count, skipping event\n"); return -1; } diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index bf464ce7e3e..befef842757 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -22,6 +22,7 @@ #include "util/symbol.h" #include "util/color.h" +#include "util/thread.h" #include "util/util.h" #include #include "util/parse-options.h" @@ -103,6 +104,7 @@ struct sym_entry { unsigned long snap_count; double weight; int skip; + struct map *map; struct source_line *source; struct source_line *lines; struct source_line **lines_tail; @@ -116,12 +118,11 @@ struct sym_entry { static void parse_source(struct sym_entry *syme) { struct symbol *sym; - struct module *module; - struct section *section = NULL; + struct map *map; FILE *file; char command[PATH_MAX*2]; - const char *path = vmlinux_name; - u64 start, end, len; + const char *path; + u64 len; if (!syme) return; @@ -132,27 +133,15 @@ static void parse_source(struct sym_entry *syme) } sym = (struct symbol *)(syme + 1); - module = sym->module; + map = syme->map; + path = map->dso->long_name; - if (module) - path = module->path; - if (!path) - return; - - start = sym->obj_start; - if (!start) - start = sym->start; - - if (module) { - section = module->sections->find_section(module->sections, ".text"); - if (section) - start -= section->vma; - } - - end = start + sym->end - sym->start + 1; len = sym->end - sym->start; - sprintf(command, "objdump --start-address=0x%016Lx --stop-address=0x%016Lx -dS %s", start, end, path); + sprintf(command, + "objdump --start-address=0x%016Lx " + "--stop-address=0x%016Lx -dS %s", + sym->start, sym->end, path); file = popen(command, "r"); if (!file) @@ -184,13 +173,11 @@ static void parse_source(struct sym_entry *syme) if (strlen(src->line)>8 && src->line[8] == ':') { src->eip = strtoull(src->line, NULL, 16); - if (section) - src->eip += section->vma; + src->eip += map->start; } if (strlen(src->line)>8 && src->line[16] == ':') { src->eip = strtoull(src->line, NULL, 16); - if (section) - src->eip += section->vma; + src->eip += map->start; } } pclose(file); @@ -242,16 +229,9 @@ static void lookup_sym_source(struct sym_entry *syme) struct symbol *symbol = (struct symbol *)(syme + 1); struct source_line *line; char pattern[PATH_MAX]; - char *idx; sprintf(pattern, "<%s>:", symbol->name); - if (symbol->module) { - idx = strstr(pattern, "\t"); - if (idx) - *idx = 0; - } - pthread_mutex_lock(&syme->source_lock); for (line = syme->lines; line; line = line->next) { if (strstr(line->line, pattern)) { @@ -513,8 +493,8 @@ static void print_sym_table(void) if (verbose) printf(" - %016llx", sym->start); printf(" : %s", sym->name); - if (sym->module) - printf("\t[%s]", sym->module->name); + if (syme->map->dso->name[0] == '[') + printf(" \t%s", syme->map->dso->name); printf("\n"); } } @@ -784,7 +764,7 @@ static const char *skip_symbols[] = { NULL }; -static int symbol_filter(struct dso *self, struct symbol *sym) +static int symbol_filter(struct map *map, struct symbol *sym) { struct sym_entry *syme; const char *name = sym->name; @@ -806,7 +786,8 @@ static int symbol_filter(struct dso *self, struct symbol *sym) strstr(name, "_text_end")) return 1; - syme = dso__sym_priv(self, sym); + syme = dso__sym_priv(map->dso, sym); + syme->map = map; pthread_mutex_init(&syme->source_lock, NULL); if (!sym_filter_entry && sym_filter && !strcmp(name, sym_filter)) sym_filter_entry = syme; @@ -825,22 +806,14 @@ static int parse_symbols(void) { int use_modules = vmlinux_name ? 1 : 0; - kernel_dso = dso__new("[kernel]", sizeof(struct sym_entry)); - if (kernel_dso == NULL) + if (dsos__load_kernel(vmlinux_name, sizeof(struct sym_entry), + symbol_filter, verbose, use_modules) <= 0) return -1; - if (dso__load_kernel(kernel_dso, vmlinux_name, symbol_filter, verbose, use_modules) <= 0) - goto out_delete_dso; - if (dump_symtab) - dso__fprintf(kernel_dso, stderr); + dsos__fprintf(stderr); return 0; - -out_delete_dso: - dso__delete(kernel_dso); - kernel_dso = NULL; - return -1; } /* @@ -848,10 +821,11 @@ out_delete_dso: */ static void record_ip(u64 ip, int counter) { - struct symbol *sym = dso__find_symbol(kernel_dso, ip); + struct map *map; + struct symbol *sym = kernel_maps__find_symbol(ip, &map); if (sym != NULL) { - struct sym_entry *syme = dso__sym_priv(kernel_dso, sym); + struct sym_entry *syme = dso__sym_priv(map->dso, sym); if (!syme->skip) { syme->count[counter]++; diff --git a/tools/perf/util/event.h b/tools/perf/util/event.h index 4c69eb55380..a39520e6ae8 100644 --- a/tools/perf/util/event.h +++ b/tools/perf/util/event.h @@ -3,6 +3,7 @@ #include "../perf.h" #include "util.h" +#include #include enum { @@ -79,7 +80,10 @@ typedef union event_union { } event_t; struct map { - struct rb_node rb_node; + union { + struct rb_node rb_node; + struct list_head node; + }; u64 start; u64 end; u64 pgoff; diff --git a/tools/perf/util/module.c b/tools/perf/util/module.c deleted file mode 100644 index 0d8c85defcd..00000000000 --- a/tools/perf/util/module.c +++ /dev/null @@ -1,545 +0,0 @@ -#include "util.h" -#include "../perf.h" -#include "string.h" -#include "module.h" - -#include -#include -#include -#include -#include -#include - -static unsigned int crc32(const char *p, unsigned int len) -{ - int i; - unsigned int crc = 0; - - while (len--) { - crc ^= *p++; - for (i = 0; i < 8; i++) - crc = (crc >> 1) ^ ((crc & 1) ? 0xedb88320 : 0); - } - return crc; -} - -/* module section methods */ - -struct sec_dso *sec_dso__new_dso(const char *name) -{ - struct sec_dso *self = malloc(sizeof(*self) + strlen(name) + 1); - - if (self != NULL) { - strcpy(self->name, name); - self->secs = RB_ROOT; - self->find_section = sec_dso__find_section; - } - - return self; -} - -static void sec_dso__delete_section(struct section *self) -{ - free(((void *)self)); -} - -void sec_dso__delete_sections(struct sec_dso *self) -{ - struct section *pos; - struct rb_node *next = rb_first(&self->secs); - - while (next) { - pos = rb_entry(next, struct section, rb_node); - next = rb_next(&pos->rb_node); - rb_erase(&pos->rb_node, &self->secs); - sec_dso__delete_section(pos); - } -} - -void sec_dso__delete_self(struct sec_dso *self) -{ - sec_dso__delete_sections(self); - free(self); -} - -static void sec_dso__insert_section(struct sec_dso *self, struct section *sec) -{ - struct rb_node **p = &self->secs.rb_node; - struct rb_node *parent = NULL; - const u64 hash = sec->hash; - struct section *s; - - while (*p != NULL) { - parent = *p; - s = rb_entry(parent, struct section, rb_node); - if (hash < s->hash) - p = &(*p)->rb_left; - else - p = &(*p)->rb_right; - } - rb_link_node(&sec->rb_node, parent, p); - rb_insert_color(&sec->rb_node, &self->secs); -} - -struct section *sec_dso__find_section(struct sec_dso *self, const char *name) -{ - struct rb_node *n; - u64 hash; - int len; - - if (self == NULL) - return NULL; - - len = strlen(name); - hash = crc32(name, len); - - n = self->secs.rb_node; - - while (n) { - struct section *s = rb_entry(n, struct section, rb_node); - - if (hash < s->hash) - n = n->rb_left; - else if (hash > s->hash) - n = n->rb_right; - else { - if (!strcmp(name, s->name)) - return s; - else - n = rb_next(&s->rb_node); - } - } - - return NULL; -} - -static size_t sec_dso__fprintf_section(struct section *self, FILE *fp) -{ - return fprintf(fp, "name:%s vma:%llx path:%s\n", - self->name, self->vma, self->path); -} - -size_t sec_dso__fprintf(struct sec_dso *self, FILE *fp) -{ - size_t ret = fprintf(fp, "dso: %s\n", self->name); - - struct rb_node *nd; - for (nd = rb_first(&self->secs); nd; nd = rb_next(nd)) { - struct section *pos = rb_entry(nd, struct section, rb_node); - ret += sec_dso__fprintf_section(pos, fp); - } - - return ret; -} - -static struct section *section__new(const char *name, const char *path) -{ - struct section *self = calloc(1, sizeof(*self)); - - if (!self) - goto out_failure; - - self->name = calloc(1, strlen(name) + 1); - if (!self->name) - goto out_failure; - - self->path = calloc(1, strlen(path) + 1); - if (!self->path) - goto out_failure; - - strcpy(self->name, name); - strcpy(self->path, path); - self->hash = crc32(self->name, strlen(name)); - - return self; - -out_failure: - if (self) { - if (self->name) - free(self->name); - if (self->path) - free(self->path); - free(self); - } - - return NULL; -} - -/* module methods */ - -struct mod_dso *mod_dso__new_dso(const char *name) -{ - struct mod_dso *self = malloc(sizeof(*self) + strlen(name) + 1); - - if (self != NULL) { - strcpy(self->name, name); - self->mods = RB_ROOT; - self->find_module = mod_dso__find_module; - } - - return self; -} - -static void mod_dso__delete_module(struct module *self) -{ - free(((void *)self)); -} - -void mod_dso__delete_modules(struct mod_dso *self) -{ - struct module *pos; - struct rb_node *next = rb_first(&self->mods); - - while (next) { - pos = rb_entry(next, struct module, rb_node); - next = rb_next(&pos->rb_node); - rb_erase(&pos->rb_node, &self->mods); - mod_dso__delete_module(pos); - } -} - -void mod_dso__delete_self(struct mod_dso *self) -{ - mod_dso__delete_modules(self); - free(self); -} - -static void mod_dso__insert_module(struct mod_dso *self, struct module *mod) -{ - struct rb_node **p = &self->mods.rb_node; - struct rb_node *parent = NULL; - const u64 hash = mod->hash; - struct module *m; - - while (*p != NULL) { - parent = *p; - m = rb_entry(parent, struct module, rb_node); - if (hash < m->hash) - p = &(*p)->rb_left; - else - p = &(*p)->rb_right; - } - rb_link_node(&mod->rb_node, parent, p); - rb_insert_color(&mod->rb_node, &self->mods); -} - -struct module *mod_dso__find_module(struct mod_dso *self, const char *name) -{ - struct rb_node *n; - u64 hash; - int len; - - if (self == NULL) - return NULL; - - len = strlen(name); - hash = crc32(name, len); - - n = self->mods.rb_node; - - while (n) { - struct module *m = rb_entry(n, struct module, rb_node); - - if (hash < m->hash) - n = n->rb_left; - else if (hash > m->hash) - n = n->rb_right; - else { - if (!strcmp(name, m->name)) - return m; - else - n = rb_next(&m->rb_node); - } - } - - return NULL; -} - -static size_t mod_dso__fprintf_module(struct module *self, FILE *fp) -{ - return fprintf(fp, "name:%s path:%s\n", self->name, self->path); -} - -size_t mod_dso__fprintf(struct mod_dso *self, FILE *fp) -{ - struct rb_node *nd; - size_t ret; - - ret = fprintf(fp, "dso: %s\n", self->name); - - for (nd = rb_first(&self->mods); nd; nd = rb_next(nd)) { - struct module *pos = rb_entry(nd, struct module, rb_node); - - ret += mod_dso__fprintf_module(pos, fp); - } - - return ret; -} - -static struct module *module__new(const char *name, const char *path) -{ - struct module *self = calloc(1, sizeof(*self)); - - if (!self) - goto out_failure; - - self->name = calloc(1, strlen(name) + 1); - if (!self->name) - goto out_failure; - - self->path = calloc(1, strlen(path) + 1); - if (!self->path) - goto out_failure; - - strcpy(self->name, name); - strcpy(self->path, path); - self->hash = crc32(self->name, strlen(name)); - - return self; - -out_failure: - if (self) { - if (self->name) - free(self->name); - if (self->path) - free(self->path); - free(self); - } - - return NULL; -} - -static int mod_dso__load_sections(struct module *mod) -{ - int count = 0, path_len; - struct dirent *entry; - char *line = NULL; - char *dir_path; - DIR *dir; - size_t n; - - path_len = strlen("/sys/module/"); - path_len += strlen(mod->name); - path_len += strlen("/sections/"); - - dir_path = calloc(1, path_len + 1); - if (dir_path == NULL) - goto out_failure; - - strcat(dir_path, "/sys/module/"); - strcat(dir_path, mod->name); - strcat(dir_path, "/sections/"); - - dir = opendir(dir_path); - if (dir == NULL) - goto out_free; - - while ((entry = readdir(dir))) { - struct section *section; - char *path, *vma; - int line_len; - FILE *file; - - if (!strcmp(".", entry->d_name) || !strcmp("..", entry->d_name)) - continue; - - path = calloc(1, path_len + strlen(entry->d_name) + 1); - if (path == NULL) - break; - strcat(path, dir_path); - strcat(path, entry->d_name); - - file = fopen(path, "r"); - if (file == NULL) { - free(path); - break; - } - - line_len = getline(&line, &n, file); - if (line_len < 0) { - free(path); - fclose(file); - break; - } - - if (!line) { - free(path); - fclose(file); - break; - } - - line[--line_len] = '\0'; /* \n */ - - vma = strstr(line, "0x"); - if (!vma) { - free(path); - fclose(file); - break; - } - vma += 2; - - section = section__new(entry->d_name, path); - if (!section) { - fprintf(stderr, "load_sections: allocation error\n"); - free(path); - fclose(file); - break; - } - - hex2u64(vma, §ion->vma); - sec_dso__insert_section(mod->sections, section); - - free(path); - fclose(file); - count++; - } - - closedir(dir); - free(line); - free(dir_path); - - return count; - -out_free: - free(dir_path); - -out_failure: - return count; -} - -static int mod_dso__load_module_paths(struct mod_dso *self) -{ - struct utsname uts; - int count = 0, len, err = -1; - char *line = NULL; - FILE *file; - char *dpath, *dir; - size_t n; - - if (uname(&uts) < 0) - return err; - - len = strlen("/lib/modules/"); - len += strlen(uts.release); - len += strlen("/modules.dep"); - - dpath = calloc(1, len + 1); - if (dpath == NULL) - return err; - - strcat(dpath, "/lib/modules/"); - strcat(dpath, uts.release); - strcat(dpath, "/modules.dep"); - - file = fopen(dpath, "r"); - if (file == NULL) - goto out_failure; - - dir = dirname(dpath); - if (!dir) - goto out_failure; - strcat(dir, "/"); - - while (!feof(file)) { - struct module *module; - char *name, *path, *tmp; - FILE *modfile; - int line_len; - - line_len = getline(&line, &n, file); - if (line_len < 0) - break; - - if (!line) - break; - - line[--line_len] = '\0'; /* \n */ - - path = strchr(line, ':'); - if (!path) - break; - *path = '\0'; - - path = strdup(line); - if (!path) - break; - - if (!strstr(path, dir)) { - if (strncmp(path, "kernel/", 7)) - break; - - free(path); - path = calloc(1, strlen(dir) + strlen(line) + 1); - if (!path) - break; - strcat(path, dir); - strcat(path, line); - } - - modfile = fopen(path, "r"); - if (modfile == NULL) - break; - fclose(modfile); - - name = strdup(path); - if (!name) - break; - - name = strtok(name, "/"); - tmp = name; - - while (tmp) { - tmp = strtok(NULL, "/"); - if (tmp) - name = tmp; - } - - name = strsep(&name, "."); - if (!name) - break; - - /* Quirk: replace '-' with '_' in all modules */ - for (len = strlen(name); len; len--) { - if (*(name+len) == '-') - *(name+len) = '_'; - } - - module = module__new(name, path); - if (!module) - break; - mod_dso__insert_module(self, module); - - module->sections = sec_dso__new_dso("sections"); - if (!module->sections) - break; - - module->active = mod_dso__load_sections(module); - - if (module->active > 0) - count++; - } - - if (feof(file)) - err = count; - else - fprintf(stderr, "load_module_paths: modules.dep parsing failure!\n"); - -out_failure: - if (dpath) - free(dpath); - if (file) - fclose(file); - if (line) - free(line); - - return err; -} - -int mod_dso__load_modules(struct mod_dso *dso) -{ - int err; - - err = mod_dso__load_module_paths(dso); - - return err; -} diff --git a/tools/perf/util/module.h b/tools/perf/util/module.h deleted file mode 100644 index 098e0412bc2..00000000000 --- a/tools/perf/util/module.h +++ /dev/null @@ -1,53 +0,0 @@ -#ifndef __PERF_MODULE_ -#define __PERF_MODULE_ 1 - -#include -#include "../types.h" -#include -#include - -struct section { - struct rb_node rb_node; - u64 hash; - u64 vma; - char *name; - char *path; -}; - -struct sec_dso { - struct list_head node; - struct rb_root secs; - struct section *(*find_section)(struct sec_dso *, const char *name); - char name[0]; -}; - -struct module { - struct rb_node rb_node; - u64 hash; - char *name; - char *path; - struct sec_dso *sections; - int active; -}; - -struct mod_dso { - struct list_head node; - struct rb_root mods; - struct module *(*find_module)(struct mod_dso *, const char *name); - char name[0]; -}; - -struct sec_dso *sec_dso__new_dso(const char *name); -void sec_dso__delete_sections(struct sec_dso *self); -void sec_dso__delete_self(struct sec_dso *self); -size_t sec_dso__fprintf(struct sec_dso *self, FILE *fp); -struct section *sec_dso__find_section(struct sec_dso *self, const char *name); - -struct mod_dso *mod_dso__new_dso(const char *name); -void mod_dso__delete_modules(struct mod_dso *self); -void mod_dso__delete_self(struct mod_dso *self); -size_t mod_dso__fprintf(struct mod_dso *self, FILE *fp); -struct module *mod_dso__find_module(struct mod_dso *self, const char *name); -int mod_dso__load_modules(struct mod_dso *dso); - -#endif /* __PERF_MODULE_ */ diff --git a/tools/perf/util/sort.c b/tools/perf/util/sort.c index 50e75abb1fd..40c9acd41ca 100644 --- a/tools/perf/util/sort.c +++ b/tools/perf/util/sort.c @@ -129,20 +129,32 @@ sort__comm_print(FILE *fp, struct hist_entry *self, unsigned int width) int64_t sort__dso_cmp(struct hist_entry *left, struct hist_entry *right) { - struct dso *dso_l = left->dso; - struct dso *dso_r = right->dso; + struct dso *dso_l = left->map ? left->map->dso : NULL; + struct dso *dso_r = right->map ? right->map->dso : NULL; + const char *dso_name_l, *dso_name_r; if (!dso_l || !dso_r) return cmp_null(dso_l, dso_r); - return strcmp(dso_l->name, dso_r->name); + if (verbose) { + dso_name_l = dso_l->long_name; + dso_name_r = dso_r->long_name; + } else { + dso_name_l = dso_l->short_name; + dso_name_r = dso_r->short_name; + } + + return strcmp(dso_name_l, dso_name_r); } size_t sort__dso_print(FILE *fp, struct hist_entry *self, unsigned int width) { - if (self->dso) - return repsep_fprintf(fp, "%-*s", width, self->dso->name); + if (self->map && self->map->dso) { + const char *dso_name = !verbose ? self->map->dso->short_name : + self->map->dso->long_name; + return repsep_fprintf(fp, "%-*s", width, dso_name); + } return repsep_fprintf(fp, "%*llx", width, (u64)self->ip); } @@ -169,20 +181,16 @@ sort__sym_print(FILE *fp, struct hist_entry *self, unsigned int width __used) { size_t ret = 0; - if (verbose) - ret += repsep_fprintf(fp, "%#018llx %c ", (u64)self->ip, - dso__symtab_origin(self->dso)); + if (verbose) { + char o = self->map ? dso__symtab_origin(self->map->dso) : '!'; + ret += repsep_fprintf(fp, "%#018llx %c ", (u64)self->ip, o); + } ret += repsep_fprintf(fp, "[%c] ", self->level); - if (self->sym) { + if (self->sym) ret += repsep_fprintf(fp, "%s", self->sym->name); - - if (self->sym->module) - ret += repsep_fprintf(fp, "\t[%s]", - self->sym->module->name); - } else { + else ret += repsep_fprintf(fp, "%#016llx", (u64)self->ip); - } return ret; } diff --git a/tools/perf/util/sort.h b/tools/perf/util/sort.h index 4684fd6d5c4..13806d782af 100644 --- a/tools/perf/util/sort.h +++ b/tools/perf/util/sort.h @@ -42,18 +42,15 @@ extern unsigned int threads__col_width; struct hist_entry { struct rb_node rb_node; - + u64 count; struct thread *thread; struct map *map; - struct dso *dso; struct symbol *sym; - struct symbol *parent; u64 ip; char level; + struct symbol *parent; struct callchain_node callchain; struct rb_root sorted_chain; - - u64 count; }; /* diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 559fb06210f..e8829689947 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -2,12 +2,14 @@ #include "../perf.h" #include "string.h" #include "symbol.h" +#include "thread.h" #include "debug.h" #include #include #include +#include const char *sym_hist_filter; @@ -18,12 +20,15 @@ enum dso_origin { DSO__ORIG_UBUNTU, DSO__ORIG_BUILDID, DSO__ORIG_DSO, + DSO__ORIG_KMODULE, DSO__ORIG_NOT_FOUND, }; -static struct symbol *symbol__new(u64 start, u64 len, - const char *name, unsigned int priv_size, - u64 obj_start, int v) +static void dsos__add(struct dso *dso); +static struct dso *dsos__find(const char *name); + +static struct symbol *symbol__new(u64 start, u64 len, const char *name, + unsigned int priv_size, int v) { size_t namelen = strlen(name) + 1; struct symbol *self = calloc(1, priv_size + sizeof(*self) + namelen); @@ -32,10 +37,9 @@ static struct symbol *symbol__new(u64 start, u64 len, return NULL; if (v >= 2) - printf("new symbol: %016Lx [%08lx]: %s, hist: %p, obj_start: %p\n", - (u64)start, (unsigned long)len, name, self->hist, (void *)(unsigned long)obj_start); + printf("new symbol: %016Lx [%08lx]: %s, hist: %p\n", + start, (unsigned long)len, name, self->hist); - self->obj_start= obj_start; self->hist = NULL; self->hist_sum = 0; @@ -60,12 +64,8 @@ static void symbol__delete(struct symbol *self, unsigned int priv_size) static size_t symbol__fprintf(struct symbol *self, FILE *fp) { - if (!self->module) - return fprintf(fp, " %llx-%llx %s\n", + return fprintf(fp, " %llx-%llx %s\n", self->start, self->end, self->name); - else - return fprintf(fp, " %llx-%llx %s \t[%s]\n", - self->start, self->end, self->name, self->module->name); } struct dso *dso__new(const char *name, unsigned int sym_priv_size) @@ -74,6 +74,8 @@ struct dso *dso__new(const char *name, unsigned int sym_priv_size) if (self != NULL) { strcpy(self->name, name); + self->long_name = self->name; + self->short_name = self->name; self->syms = RB_ROOT; self->sym_priv_size = sym_priv_size; self->find_symbol = dso__find_symbol; @@ -100,6 +102,8 @@ static void dso__delete_symbols(struct dso *self) void dso__delete(struct dso *self) { dso__delete_symbols(self); + if (self->long_name != self->name) + free(self->long_name); free(self); } @@ -147,7 +151,7 @@ struct symbol *dso__find_symbol(struct dso *self, u64 ip) size_t dso__fprintf(struct dso *self, FILE *fp) { - size_t ret = fprintf(fp, "dso: %s\n", self->name); + size_t ret = fprintf(fp, "dso: %s\n", self->long_name); struct rb_node *nd; for (nd = rb_first(&self->syms); nd; nd = rb_next(nd)) { @@ -158,7 +162,8 @@ size_t dso__fprintf(struct dso *self, FILE *fp) return ret; } -static int dso__load_kallsyms(struct dso *self, symbol_filter_t filter, int v) +static int dso__load_kallsyms(struct dso *self, struct map *map, + symbol_filter_t filter, int v) { struct rb_node *nd, *prevnd; char *line = NULL; @@ -200,12 +205,12 @@ static int dso__load_kallsyms(struct dso *self, symbol_filter_t filter, int v) * Well fix up the end later, when we have all sorted. */ sym = symbol__new(start, 0xdead, line + len + 2, - self->sym_priv_size, 0, v); + self->sym_priv_size, v); if (sym == NULL) goto out_delete_line; - if (filter && filter(self, sym)) + if (filter && filter(map, sym)) symbol__delete(sym, self->sym_priv_size); else { dso__insert_symbol(self, sym); @@ -241,14 +246,15 @@ out_failure: return -1; } -static int dso__load_perf_map(struct dso *self, symbol_filter_t filter, int v) +static int dso__load_perf_map(struct dso *self, struct map *map, + symbol_filter_t filter, int v) { char *line = NULL; size_t n; FILE *file; int nr_syms = 0; - file = fopen(self->name, "r"); + file = fopen(self->long_name, "r"); if (file == NULL) goto out_failure; @@ -279,12 +285,12 @@ static int dso__load_perf_map(struct dso *self, symbol_filter_t filter, int v) continue; sym = symbol__new(start, size, line + len, - self->sym_priv_size, start, v); + self->sym_priv_size, v); if (sym == NULL) goto out_delete_line; - if (filter && filter(self, sym)) + if (filter && filter(map, sym)) symbol__delete(sym, self->sym_priv_size); else { dso__insert_symbol(self, sym); @@ -410,7 +416,7 @@ static int dso__synthesize_plt_symbols(struct dso *self, int v) Elf *elf; int nr = 0, symidx, fd, err = 0; - fd = open(self->name, O_RDONLY); + fd = open(self->long_name, O_RDONLY); if (fd < 0) goto out; @@ -478,7 +484,7 @@ static int dso__synthesize_plt_symbols(struct dso *self, int v) "%s@plt", elf_sym__name(&sym, symstrs)); f = symbol__new(plt_offset, shdr_plt.sh_entsize, - sympltname, self->sym_priv_size, 0, v); + sympltname, self->sym_priv_size, v); if (!f) goto out_elf_end; @@ -496,7 +502,7 @@ static int dso__synthesize_plt_symbols(struct dso *self, int v) "%s@plt", elf_sym__name(&sym, symstrs)); f = symbol__new(plt_offset, shdr_plt.sh_entsize, - sympltname, self->sym_priv_size, 0, v); + sympltname, self->sym_priv_size, v); if (!f) goto out_elf_end; @@ -515,12 +521,13 @@ out_close: return nr; out: fprintf(stderr, "%s: problems reading %s PLT info.\n", - __func__, self->name); + __func__, self->long_name); return 0; } -static int dso__load_sym(struct dso *self, int fd, const char *name, - symbol_filter_t filter, int v, struct module *mod) +static int dso__load_sym(struct dso *self, struct map *map, const char *name, + int fd, symbol_filter_t filter, int kernel, + int kmodule, int v) { Elf_Data *symstrs, *secstrs; uint32_t nr_syms; @@ -532,7 +539,7 @@ static int dso__load_sym(struct dso *self, int fd, const char *name, GElf_Sym sym; Elf_Scn *sec, *sec_strndx; Elf *elf; - int nr = 0, kernel = !strcmp("[kernel]", self->name); + int nr = 0; elf = elf_begin(fd, ELF_C_READ_MMAP, NULL); if (elf == NULL) { @@ -589,8 +596,6 @@ static int dso__load_sym(struct dso *self, int fd, const char *name, struct symbol *f; const char *elf_name; char *demangled; - u64 obj_start; - struct section *section = NULL; int is_label = elf_sym__is_label(&sym); const char *section_name; @@ -607,7 +612,6 @@ static int dso__load_sym(struct dso *self, int fd, const char *name, continue; section_name = elf_sec__name(&shdr, secstrs); - obj_start = sym.st_value; if (self->adjust_symbols) { if (v >= 2) @@ -615,18 +619,8 @@ static int dso__load_sym(struct dso *self, int fd, const char *name, (u64)sym.st_value, (u64)shdr.sh_addr, (u64)shdr.sh_offset); sym.st_value -= shdr.sh_addr - shdr.sh_offset; - } - - if (mod) { - section = mod->sections->find_section(mod->sections, section_name); - if (section) - sym.st_value += section->vma; - else { - fprintf(stderr, "dso__load_sym() module %s lookup of %s failed\n", - mod->name, section_name); - goto out_elf_end; - } - } + } else if (kmodule) + sym.st_value += shdr.sh_offset; /* * We need to figure out if the object was created from C++ sources * DWARF DW_compile_unit has this, but we don't always have access @@ -638,15 +632,14 @@ static int dso__load_sym(struct dso *self, int fd, const char *name, elf_name = demangled; f = symbol__new(sym.st_value, sym.st_size, elf_name, - self->sym_priv_size, obj_start, v); + self->sym_priv_size, v); free(demangled); if (!f) goto out_elf_end; - if (filter && filter(self, f)) + if (filter && filter(map, f)) symbol__delete(f, self->sym_priv_size); else { - f->module = mod; dso__insert_symbol(self, f); nr++; } @@ -671,7 +664,7 @@ static char *dso__read_build_id(struct dso *self, int v) char *build_id = NULL, *bid; unsigned char *raw; Elf *elf; - int fd = open(self->name, O_RDONLY); + int fd = open(self->long_name, O_RDONLY); if (fd < 0) goto out; @@ -680,7 +673,7 @@ static char *dso__read_build_id(struct dso *self, int v) if (elf == NULL) { if (v) fprintf(stderr, "%s: cannot read %s ELF file.\n", - __func__, self->name); + __func__, self->long_name); goto out_close; } @@ -709,7 +702,7 @@ static char *dso__read_build_id(struct dso *self, int v) bid += 2; } if (v >= 2) - printf("%s(%s): %s\n", __func__, self->name, build_id); + printf("%s(%s): %s\n", __func__, self->long_name, build_id); out_elf_end: elf_end(elf); out_close: @@ -727,6 +720,7 @@ char dso__symtab_origin(const struct dso *self) [DSO__ORIG_UBUNTU] = 'u', [DSO__ORIG_BUILDID] = 'b', [DSO__ORIG_DSO] = 'd', + [DSO__ORIG_KMODULE] = 'K', }; if (self == NULL || self->origin == DSO__ORIG_NOT_FOUND) @@ -734,7 +728,7 @@ char dso__symtab_origin(const struct dso *self) return origin[self->origin]; } -int dso__load(struct dso *self, symbol_filter_t filter, int v) +int dso__load(struct dso *self, struct map *map, symbol_filter_t filter, int v) { int size = PATH_MAX; char *name = malloc(size), *build_id = NULL; @@ -747,7 +741,7 @@ int dso__load(struct dso *self, symbol_filter_t filter, int v) self->adjust_symbols = 0; if (strncmp(self->name, "/tmp/perf-", 10) == 0) { - ret = dso__load_perf_map(self, filter, v); + ret = dso__load_perf_map(self, map, filter, v); self->origin = ret > 0 ? DSO__ORIG_JAVA_JIT : DSO__ORIG_NOT_FOUND; return ret; @@ -760,10 +754,12 @@ more: self->origin++; switch (self->origin) { case DSO__ORIG_FEDORA: - snprintf(name, size, "/usr/lib/debug%s.debug", self->name); + snprintf(name, size, "/usr/lib/debug%s.debug", + self->long_name); break; case DSO__ORIG_UBUNTU: - snprintf(name, size, "/usr/lib/debug%s", self->name); + snprintf(name, size, "/usr/lib/debug%s", + self->long_name); break; case DSO__ORIG_BUILDID: build_id = dso__read_build_id(self, v); @@ -777,7 +773,7 @@ more: self->origin++; /* Fall thru */ case DSO__ORIG_DSO: - snprintf(name, size, "%s", self->name); + snprintf(name, size, "%s", self->long_name); break; default: @@ -787,7 +783,7 @@ more: fd = open(name, O_RDONLY); } while (fd < 0); - ret = dso__load_sym(self, fd, name, filter, v, NULL); + ret = dso__load_sym(self, map, name, fd, filter, 0, 0, v); close(fd); /* @@ -808,89 +804,247 @@ out: return ret; } -static int dso__load_module(struct dso *self, struct mod_dso *mods, const char *name, - symbol_filter_t filter, int v) +static struct rb_root kernel_maps; +struct map *kernel_map; + +static void kernel_maps__insert(struct map *map) { - struct module *mod = mod_dso__find_module(mods, name); - int err = 0, fd; + maps__insert(&kernel_maps, map); +} - if (mod == NULL || !mod->active) - return err; +struct symbol *kernel_maps__find_symbol(u64 ip, struct map **mapp) +{ + /* + * We can't have kernel_map in kernel_maps because it spans an address + * space that includes the modules. The right way to fix this is to + * create several maps, so that we don't have overlapping ranges with + * modules. For now lets look first on the kernel dso. + */ + struct map *map = maps__find(&kernel_maps, ip); + struct symbol *sym; + + if (map) { + ip = map->map_ip(map, ip); + sym = map->dso->find_symbol(map->dso, ip); + } else { + map = kernel_map; + sym = map->dso->find_symbol(map->dso, ip); + } - fd = open(mod->path, O_RDONLY); + if (mapp) + *mapp = map; - if (fd < 0) + return sym; +} + +struct map *kernel_maps__find_by_dso_name(const char *name) +{ + struct rb_node *nd; + + for (nd = rb_first(&kernel_maps); nd; nd = rb_next(nd)) { + struct map *map = rb_entry(nd, struct map, rb_node); + + if (map->dso && strcmp(map->dso->name, name) == 0) + return map; + } + + return NULL; +} + +static int dso__load_module_sym(struct dso *self, struct map *map, + symbol_filter_t filter, int v) +{ + int err = 0, fd = open(self->long_name, O_RDONLY); + + if (fd < 0) { + if (v) + fprintf(stderr, "%s: cannot open %s\n", + __func__, self->long_name); return err; + } - err = dso__load_sym(self, fd, name, filter, v, mod); + err = dso__load_sym(self, map, self->long_name, fd, filter, 0, 1, v); close(fd); return err; } -int dso__load_modules(struct dso *self, symbol_filter_t filter, int v) +static int dsos__load_modules_sym_dir(char *dirname, + symbol_filter_t filter, int v) { - struct mod_dso *mods = mod_dso__new_dso("modules"); - struct module *pos; - struct rb_node *next; - int err, count = 0; + struct dirent *dent; + int nr_symbols = 0, err; + DIR *dir = opendir(dirname); - err = mod_dso__load_modules(mods); + if (!dir) { + if (v) + fprintf(stderr, "%s: cannot open %s dir\n", __func__, + dirname); + return -1; + } - if (err <= 0) - return err; + while ((dent = readdir(dir)) != NULL) { + char path[PATH_MAX]; + + if (dent->d_type == DT_DIR) { + if (!strcmp(dent->d_name, ".") || + !strcmp(dent->d_name, "..")) + continue; + + snprintf(path, sizeof(path), "%s/%s", + dirname, dent->d_name); + err = dsos__load_modules_sym_dir(path, filter, v); + if (err < 0) + goto failure; + } else { + char *dot = strrchr(dent->d_name, '.'), + dso_name[PATH_MAX]; + struct map *map; + struct rb_node *last; + + if (dot == NULL || strcmp(dot, ".ko")) + continue; + snprintf(dso_name, sizeof(dso_name), "[%.*s]", + (int)(dot - dent->d_name), dent->d_name); + + map = kernel_maps__find_by_dso_name(dso_name); + if (map == NULL) + continue; + + snprintf(path, sizeof(path), "%s/%s", + dirname, dent->d_name); + + map->dso->long_name = strdup(path); + if (map->dso->long_name == NULL) + goto failure; + + err = dso__load_module_sym(map->dso, map, filter, v); + if (err < 0) + goto failure; + last = rb_last(&map->dso->syms); + if (last) { + struct symbol *sym; + sym = rb_entry(last, struct symbol, rb_node); + map->end = map->start + sym->end; + } + } + nr_symbols += err; + } - /* - * Iterate over modules, and load active symbols. - */ - next = rb_first(&mods->mods); - while (next) { - pos = rb_entry(next, struct module, rb_node); - err = dso__load_module(self, mods, pos->name, filter, v); + return nr_symbols; +failure: + closedir(dir); + return -1; +} - if (err < 0) - break; +static int dsos__load_modules_sym(symbol_filter_t filter, int v) +{ + struct utsname uts; + char modules_path[PATH_MAX]; - next = rb_next(&pos->rb_node); - count += err; - } + if (uname(&uts) < 0) + return -1; - if (err < 0) { - mod_dso__delete_modules(mods); - mod_dso__delete_self(mods); - return err; - } + snprintf(modules_path, sizeof(modules_path), "/lib/modules/%s/kernel", + uts.release); - return count; + return dsos__load_modules_sym_dir(modules_path, filter, v); } -static inline void dso__fill_symbol_holes(struct dso *self) +/* + * Constructor variant for modules (where we know from /proc/modules where + * they are loaded) and for vmlinux, where only after we load all the + * symbols we'll know where it starts and ends. + */ +static struct map *map__new2(u64 start, struct dso *dso) { - struct symbol *prev = NULL; - struct rb_node *nd; + struct map *self = malloc(sizeof(*self)); - for (nd = rb_last(&self->syms); nd; nd = rb_prev(nd)) { - struct symbol *pos = rb_entry(nd, struct symbol, rb_node); + if (self != NULL) { + self->start = start; + /* + * Will be filled after we load all the symbols + */ + self->end = 0; + + self->pgoff = 0; + self->dso = dso; + self->map_ip = map__map_ip; + RB_CLEAR_NODE(&self->rb_node); + } + return self; +} + +int dsos__load_modules(unsigned int sym_priv_size, + symbol_filter_t filter, int v) +{ + char *line = NULL; + size_t n; + FILE *file = fopen("/proc/modules", "r"); + struct map *map; - if (prev) { - u64 hole = 0; - int alias = pos->start == prev->start; + if (file == NULL) + return -1; - if (!alias) - hole = prev->start - pos->end - 1; + while (!feof(file)) { + char name[PATH_MAX]; + u64 start; + struct dso *dso; + char *sep; + int line_len; - if (hole || alias) { - if (alias) - pos->end = prev->end; - else if (hole) - pos->end = prev->start - 1; - } + line_len = getline(&line, &n, file); + if (line_len < 0) + break; + + if (!line) + goto out_failure; + + line[--line_len] = '\0'; /* \n */ + + sep = strrchr(line, 'x'); + if (sep == NULL) + continue; + + hex2u64(sep + 1, &start); + + sep = strchr(line, ' '); + if (sep == NULL) + continue; + + *sep = '\0'; + + snprintf(name, sizeof(name), "[%s]", line); + dso = dso__new(name, sym_priv_size); + + if (dso == NULL) + goto out_delete_line; + + map = map__new2(start, dso); + if (map == NULL) { + dso__delete(dso); + goto out_delete_line; } - prev = pos; + + dso->origin = DSO__ORIG_KMODULE; + kernel_maps__insert(map); + dsos__add(dso); } + + free(line); + fclose(file); + + v = 1; + return dsos__load_modules_sym(filter, v); + +out_delete_line: + free(line); +out_failure: + return -1; } -static int dso__load_vmlinux(struct dso *self, const char *vmlinux, +static int dso__load_vmlinux(struct dso *self, struct map *map, + const char *vmlinux, symbol_filter_t filter, int v) { int err, fd = open(vmlinux, O_RDONLY); @@ -898,28 +1052,36 @@ static int dso__load_vmlinux(struct dso *self, const char *vmlinux, if (fd < 0) return -1; - err = dso__load_sym(self, fd, vmlinux, filter, v, NULL); - - if (err > 0) - dso__fill_symbol_holes(self); + err = dso__load_sym(self, map, self->long_name, fd, filter, 1, 0, v); close(fd); return err; } -int dso__load_kernel(struct dso *self, const char *vmlinux, - symbol_filter_t filter, int v, int use_modules) +int dsos__load_kernel(const char *vmlinux, unsigned int sym_priv_size, + symbol_filter_t filter, int v, int use_modules) { int err = -1; + struct dso *dso = dso__new(vmlinux, sym_priv_size); + + if (dso == NULL) + return -1; + + dso->short_name = "[kernel]"; + kernel_map = map__new2(0, dso); + if (kernel_map == NULL) + goto out_delete_dso; + + kernel_map->map_ip = vdso__map_ip; if (vmlinux) { - err = dso__load_vmlinux(self, vmlinux, filter, v); + err = dso__load_vmlinux(dso, kernel_map, vmlinux, filter, v); if (err > 0 && use_modules) { - int syms = dso__load_modules(self, filter, v); + int syms = dsos__load_modules(sym_priv_size, filter, v); if (syms < 0) { - fprintf(stderr, "dso__load_modules failed!\n"); + fprintf(stderr, "dsos__load_modules failed!\n"); return syms; } err += syms; @@ -927,18 +1089,34 @@ int dso__load_kernel(struct dso *self, const char *vmlinux, } if (err <= 0) - err = dso__load_kallsyms(self, filter, v); + err = dso__load_kallsyms(dso, kernel_map, filter, v); + + if (err > 0) { + struct rb_node *node = rb_first(&dso->syms); + struct symbol *sym = rb_entry(node, struct symbol, rb_node); - if (err > 0) - self->origin = DSO__ORIG_KERNEL; + kernel_map->start = sym->start; + node = rb_last(&dso->syms); + sym = rb_entry(node, struct symbol, rb_node); + kernel_map->end = sym->end; + + dso->origin = DSO__ORIG_KERNEL; + /* + * XXX See kernel_maps__find_symbol comment + * kernel_maps__insert(kernel_map) + */ + dsos__add(dso); + } return err; + +out_delete_dso: + dso__delete(dso); + return -1; } LIST_HEAD(dsos); -struct dso *kernel_dso; struct dso *vdso; -struct dso *hypervisor_dso; const char *vmlinux_name = "vmlinux"; int modules; @@ -970,7 +1148,7 @@ struct dso *dsos__findnew(const char *name) if (!dso) goto out_delete_dso; - nr = dso__load(dso, NULL, verbose); + nr = dso__load(dso, NULL, NULL, verbose); if (nr < 0) { eprintf("Failed to open: %s\n", name); goto out_delete_dso; @@ -995,43 +1173,20 @@ void dsos__fprintf(FILE *fp) dso__fprintf(pos, fp); } -static struct symbol *vdso__find_symbol(struct dso *dso, u64 ip) -{ - return dso__find_symbol(dso, ip); -} - int load_kernel(void) { - int err; - - kernel_dso = dso__new("[kernel]", 0); - if (!kernel_dso) + if (dsos__load_kernel(vmlinux_name, 0, NULL, verbose, modules) <= 0) return -1; - err = dso__load_kernel(kernel_dso, vmlinux_name, NULL, verbose, modules); - if (err <= 0) { - dso__delete(kernel_dso); - kernel_dso = NULL; - } else - dsos__add(kernel_dso); - vdso = dso__new("[vdso]", 0); if (!vdso) return -1; - vdso->find_symbol = vdso__find_symbol; - dsos__add(vdso); - hypervisor_dso = dso__new("[hypervisor]", 0); - if (!hypervisor_dso) - return -1; - dsos__add(hypervisor_dso); - - return err; + return 0; } - void symbol__init(void) { elf_version(EV_CURRENT); diff --git a/tools/perf/util/symbol.h b/tools/perf/util/symbol.h index ee164f659ed..5339fd82ec9 100644 --- a/tools/perf/util/symbol.h +++ b/tools/perf/util/symbol.h @@ -5,7 +5,6 @@ #include "types.h" #include #include -#include "module.h" #include "event.h" #ifdef HAVE_CPLUS_DEMANGLE @@ -36,10 +35,8 @@ struct symbol { struct rb_node rb_node; u64 start; u64 end; - u64 obj_start; u64 hist_sum; u64 *hist; - struct module *module; void *priv; char name[0]; }; @@ -52,12 +49,14 @@ struct dso { unsigned char adjust_symbols; unsigned char slen_calculated; unsigned char origin; + const char *short_name; + char *long_name; char name[0]; }; extern const char *sym_hist_filter; -typedef int (*symbol_filter_t)(struct dso *self, struct symbol *sym); +typedef int (*symbol_filter_t)(struct map *map, struct symbol *sym); struct dso *dso__new(const char *name, unsigned int sym_priv_size); void dso__delete(struct dso *self); @@ -69,10 +68,12 @@ static inline void *dso__sym_priv(struct dso *self, struct symbol *sym) struct symbol *dso__find_symbol(struct dso *self, u64 ip); -int dso__load_kernel(struct dso *self, const char *vmlinux, - symbol_filter_t filter, int verbose, int modules); -int dso__load_modules(struct dso *self, symbol_filter_t filter, int verbose); -int dso__load(struct dso *self, symbol_filter_t filter, int verbose); +int dsos__load_kernel(const char *vmlinux, unsigned int sym_priv_size, + symbol_filter_t filter, int verbose, int modules); +int dsos__load_modules(unsigned int sym_priv_size, symbol_filter_t filter, + int verbose); +int dso__load(struct dso *self, struct map *map, symbol_filter_t filter, + int verbose); struct dso *dsos__findnew(const char *name); void dsos__fprintf(FILE *fp); @@ -84,9 +85,8 @@ int load_kernel(void); void symbol__init(void); extern struct list_head dsos; -extern struct dso *kernel_dso; +extern struct map *kernel_map; extern struct dso *vdso; -extern struct dso *hypervisor_dso; extern const char *vmlinux_name; extern int modules; #endif /* __PERF_SYMBOL */ diff --git a/tools/perf/util/thread.c b/tools/perf/util/thread.c index 9d0945cc66d..3b56aebb1f4 100644 --- a/tools/perf/util/thread.c +++ b/tools/perf/util/thread.c @@ -16,6 +16,7 @@ static struct thread *thread__new(pid_t pid) if (self->comm) snprintf(self->comm, 32, ":%d", self->pid); self->maps = RB_ROOT; + INIT_LIST_HEAD(&self->removed_maps); } return self; @@ -32,13 +33,20 @@ int thread__set_comm(struct thread *self, const char *comm) static size_t thread__fprintf(struct thread *self, FILE *fp) { struct rb_node *nd; - size_t ret = fprintf(fp, "Thread %d %s\n", self->pid, self->comm); + struct map *pos; + size_t ret = fprintf(fp, "Thread %d %s\nCurrent maps:\n", + self->pid, self->comm); for (nd = rb_first(&self->maps); nd; nd = rb_next(nd)) { - struct map *pos = rb_entry(nd, struct map, rb_node); + pos = rb_entry(nd, struct map, rb_node); ret += map__fprintf(pos, fp); } + ret = fprintf(fp, "Removed maps:\n"); + + list_for_each_entry(pos, &self->removed_maps, node) + ret += map__fprintf(pos, fp); + return ret; } @@ -112,21 +120,13 @@ static void thread__remove_overlappings(struct thread *self, struct map *map) map__fprintf(pos, stdout); } - if (map->start <= pos->start && map->end > pos->start) - pos->start = map->end; - - if (map->end >= pos->end && map->start < pos->end) - pos->end = map->start; - - if (verbose >= 2) { - printf("after collision:\n"); - map__fprintf(pos, stdout); - } - - if (pos->start >= pos->end) { - rb_erase(&pos->rb_node, &self->maps); - free(pos); - } + rb_erase(&pos->rb_node, &self->maps); + /* + * We may have references to this map, for instance in some + * hist_entry instances, so just move them to a separate + * list. + */ + list_add_tail(&pos->node, &self->removed_maps); } } diff --git a/tools/perf/util/thread.h b/tools/perf/util/thread.h index bbb37c1a52e..845d9b62f96 100644 --- a/tools/perf/util/thread.h +++ b/tools/perf/util/thread.h @@ -8,6 +8,7 @@ struct thread { struct rb_node rb_node; struct rb_root maps; + struct list_head removed_maps; pid_t pid; char shortname[3]; char *comm; @@ -25,6 +26,9 @@ size_t threads__fprintf(FILE *fp, struct rb_root *threads); void maps__insert(struct rb_root *maps, struct map *map); struct map *maps__find(struct rb_root *maps, u64 ip); +struct symbol *kernel_maps__find_symbol(const u64 ip, struct map **mapp); +struct map *kernel_maps__find_by_dso_name(const char *name); + static inline struct map *thread__find_map(struct thread *self, u64 ip) { return self ? maps__find(&self->maps, ip) : NULL; -- cgit v1.2.3-18-g5258 From 9735abf11bec48bfbbb1b54772a02deb2ae0c403 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Sat, 3 Oct 2009 10:42:45 -0300 Subject: perf tools: Move hist_entry__add common code to hist.c MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Now perf report and annotate do the callgraph/hit processing in their specialized hist_entry__add functions. Signed-off-by: Arnaldo Carvalho de Melo Acked-by: Frédéric Weisbecker Cc: Peter Zijlstra Cc: Mike Galbraith Signed-off-by: Ingo Molnar --- tools/perf/builtin-annotate.c | 50 +++++++--------------------------- tools/perf/builtin-report.c | 63 +++++++++++-------------------------------- tools/perf/util/hist.c | 46 +++++++++++++++++++++++++++++++ tools/perf/util/hist.h | 3 +++ 4 files changed, 74 insertions(+), 88 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c index 7d5a3b1bcda..855094234f2 100644 --- a/tools/perf/builtin-annotate.c +++ b/tools/perf/builtin-annotate.c @@ -80,48 +80,16 @@ static void hist_hit(struct hist_entry *he, u64 ip) sym->hist[offset]); } -static int -hist_entry__add(struct thread *thread, struct map *map, - struct symbol *sym, u64 ip, char level) +static int hist_entry__add(struct thread *thread, struct map *map, + struct symbol *sym, u64 ip, u64 count, char level) { - struct rb_node **p = &hist.rb_node; - struct rb_node *parent = NULL; - struct hist_entry *he; - struct hist_entry entry = { - .thread = thread, - .map = map, - .sym = sym, - .ip = ip, - .level = level, - .count = 1, - }; - int cmp; - - while (*p != NULL) { - parent = *p; - he = rb_entry(parent, struct hist_entry, rb_node); - - cmp = hist_entry__cmp(&entry, he); - - if (!cmp) { - hist_hit(he, ip); - - return 0; - } - - if (cmp < 0) - p = &(*p)->rb_left; - else - p = &(*p)->rb_right; - } - - he = malloc(sizeof(*he)); - if (!he) + bool hit; + struct hist_entry *he = __hist_entry__add(thread, map, sym, NULL, ip, + count, level, &hit); + if (he == NULL) return -ENOMEM; - *he = entry; - rb_link_node(&he->rb_node, parent, p); - rb_insert_color(&he->rb_node, &hist); - + if (hit) + hist_hit(he, ip); return 0; } @@ -191,7 +159,7 @@ got_map: } if (show & show_mask) { - if (hist_entry__add(thread, map, sym, ip, level)) { + if (hist_entry__add(thread, map, sym, ip, 1, level)) { fprintf(stderr, "problem incrementing symbol count, skipping event\n"); return -1; diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index 3ed3baf96ff..0e83ffcbe55 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -407,9 +407,9 @@ static int call__match(struct symbol *sym) return 0; } -static struct symbol ** -resolve_callchain(struct thread *thread, struct map *map, - struct ip_callchain *chain, struct hist_entry *entry) +static struct symbol **resolve_callchain(struct thread *thread, struct map *map, + struct ip_callchain *chain, + struct symbol **parent) { u64 context = PERF_CONTEXT_MAX; struct symbol **syms = NULL; @@ -444,9 +444,8 @@ resolve_callchain(struct thread *thread, struct map *map, } if (sym) { - if (sort__has_parent && call__match(sym) && - !entry->parent) - entry->parent = sym; + if (sort__has_parent && !*parent && call__match(sym)) + *parent = sym; if (!callchain) break; syms[i] = sym; @@ -465,57 +464,27 @@ hist_entry__add(struct thread *thread, struct map *map, struct symbol *sym, u64 ip, struct ip_callchain *chain, char level, u64 count) { - struct rb_node **p = &hist.rb_node; - struct rb_node *parent = NULL; + struct symbol **syms = NULL, *parent = NULL; + bool hit; struct hist_entry *he; - struct symbol **syms = NULL; - struct hist_entry entry = { - .thread = thread, - .map = map, - .sym = sym, - .ip = ip, - .level = level, - .count = count, - .parent = NULL, - .sorted_chain = RB_ROOT - }; - int cmp; if ((sort__has_parent || callchain) && chain) - syms = resolve_callchain(thread, map, chain, &entry); - - while (*p != NULL) { - parent = *p; - he = rb_entry(parent, struct hist_entry, rb_node); - - cmp = hist_entry__cmp(&entry, he); + syms = resolve_callchain(thread, map, chain, &parent); - if (!cmp) { - he->count += count; - if (callchain) { - append_chain(&he->callchain, chain, syms); - free(syms); - } - return 0; - } + he = __hist_entry__add(thread, map, sym, parent, + ip, count, level, &hit); + if (he == NULL) + return -ENOMEM; - if (cmp < 0) - p = &(*p)->rb_left; - else - p = &(*p)->rb_right; - } + if (hit) + he->count += count; - he = malloc(sizeof(*he)); - if (!he) - return -ENOMEM; - *he = entry; if (callchain) { - callchain_init(&he->callchain); + if (!hit) + callchain_init(&he->callchain); append_chain(&he->callchain, chain, syms); free(syms); } - rb_link_node(&he->rb_node, parent, p); - rb_insert_color(&he->rb_node, &hist); return 0; } diff --git a/tools/perf/util/hist.c b/tools/perf/util/hist.c index 82808dc4f8e..7393a02fd8d 100644 --- a/tools/perf/util/hist.c +++ b/tools/perf/util/hist.c @@ -21,6 +21,52 @@ unsigned long total_lost; * histogram, sorted on item, collects counts */ +struct hist_entry *__hist_entry__add(struct thread *thread, struct map *map, + struct symbol *sym, + struct symbol *sym_parent, + u64 ip, u64 count, char level, bool *hit) +{ + struct rb_node **p = &hist.rb_node; + struct rb_node *parent = NULL; + struct hist_entry *he; + struct hist_entry entry = { + .thread = thread, + .map = map, + .sym = sym, + .ip = ip, + .level = level, + .count = count, + .parent = sym_parent, + }; + int cmp; + + while (*p != NULL) { + parent = *p; + he = rb_entry(parent, struct hist_entry, rb_node); + + cmp = hist_entry__cmp(&entry, he); + + if (!cmp) { + *hit = true; + return he; + } + + if (cmp < 0) + p = &(*p)->rb_left; + else + p = &(*p)->rb_right; + } + + he = malloc(sizeof(*he)); + if (!he) + return NULL; + *he = entry; + rb_link_node(&he->rb_node, parent, p); + rb_insert_color(&he->rb_node, &hist); + *hit = false; + return he; +} + int64_t hist_entry__cmp(struct hist_entry *left, struct hist_entry *right) { diff --git a/tools/perf/util/hist.h b/tools/perf/util/hist.h index 9a8daa12b43..ac2149c559b 100644 --- a/tools/perf/util/hist.h +++ b/tools/perf/util/hist.h @@ -36,6 +36,9 @@ extern unsigned long total_fork; extern unsigned long total_unknown; extern unsigned long total_lost; +struct hist_entry *__hist_entry__add(struct thread *thread, struct map *map, + struct symbol *sym, struct symbol *parent, + u64 ip, u64 count, char level, bool *hit); extern int64_t hist_entry__cmp(struct hist_entry *, struct hist_entry *); extern int64_t hist_entry__collapse(struct hist_entry *, struct hist_entry *); extern void hist_entry__free(struct hist_entry *); -- cgit v1.2.3-18-g5258 From ec218fc4a796a1b584741d59ef22615d96981188 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Sat, 3 Oct 2009 20:30:48 -0300 Subject: perf tools: Remove show_mask bitmask MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit As it was not being exposed via any command line and with --dsos/--comms we can do this and even more, like asking for just kernel + some module: [root@doppio linux-2.6-tip]# perf report --dsos \[kernel\],\[drm\] --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules | head -15 # Samples: 619669 # # Overhead Command Shared Object Symbol # ........ ............... ............. ...... # 7.12% swapper [kernel] [k] read_hpet 6.86% init [kernel] [k] read_hpet 6.22% init [kernel] [k] mwait_idle_with_hints 5.34% swapper [kernel] [k] mwait_idle_with_hints 3.01% firefox [kernel] [.] vread_hpet 2.14% Xorg [drm] [k] drm_clflush_pages 2.09% pidgin [kernel] [.] vread_hpet 1.58% npviewer.bin [kernel] [.] vread_hpet 1.37% swapper [kernel] [k] hpet_next_event 1.23% Xorg [kernel] [k] read_hpet [root@doppio linux-2.6-tip]# Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Peter Zijlstra Cc: Mike Galbraith LKML-Reference: <20091003233048.GA30535@ghostprotocols.net> Signed-off-by: Ingo Molnar --- tools/perf/Makefile | 1 + tools/perf/builtin-annotate.c | 15 ++++----------- tools/perf/builtin-report.c | 35 +++++++++++++---------------------- tools/perf/util/event.h | 6 ------ 4 files changed, 18 insertions(+), 39 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/Makefile b/tools/perf/Makefile index 8e7509f2d88..2c309a5c686 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -323,6 +323,7 @@ LIB_H += ../../include/linux/rbtree.h LIB_H += ../../include/linux/list.h LIB_H += util/include/linux/list.h LIB_H += perf.h +LIB_H += util/event.h LIB_H += util/types.h LIB_H += util/levenshtein.h LIB_H += util/parse-options.h diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c index 855094234f2..35ed97bd0c6 100644 --- a/tools/perf/builtin-annotate.c +++ b/tools/perf/builtin-annotate.c @@ -29,7 +29,6 @@ static char const *input_name = "perf.data"; static int force; static int input; -static int show_mask = SHOW_KERNEL | SHOW_USER | SHOW_HV; static int full_paths; @@ -97,7 +96,6 @@ static int process_sample_event(event_t *event, unsigned long offset, unsigned long head) { char level; - int show = 0; struct thread *thread; u64 ip = event->ip.ip; struct map *map = NULL; @@ -121,13 +119,11 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) } if (event->header.misc & PERF_RECORD_MISC_KERNEL) { - show = SHOW_KERNEL; level = 'k'; sym = kernel_maps__find_symbol(ip, &map); dump_printf(" ...... dso: %s\n", map ? map->dso->long_name : ""); } else if (event->header.misc & PERF_RECORD_MISC_USER) { - show = SHOW_USER; level = '.'; map = thread__find_map(thread, ip); if (map != NULL) { @@ -153,17 +149,14 @@ got_map: dump_printf(" ...... dso: %s\n", map ? map->dso->long_name : ""); } else { - show = SHOW_HV; level = 'H'; dump_printf(" ...... dso: [hypervisor]\n"); } - if (show & show_mask) { - if (hist_entry__add(thread, map, sym, ip, 1, level)) { - fprintf(stderr, - "problem incrementing symbol count, skipping event\n"); - return -1; - } + if (hist_entry__add(thread, map, sym, ip, 1, level)) { + fprintf(stderr, "problem incrementing symbol count, " + "skipping event\n"); + return -1; } total++; diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index 0e83ffcbe55..fe4aadc9630 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -38,7 +38,6 @@ static struct strlist *dso_list, *comm_list, *sym_list; static int force; static int input; -static int show_mask = SHOW_KERNEL | SHOW_USER | SHOW_HV; static int full_paths; static int show_nr_samples; @@ -600,7 +599,6 @@ static int process_sample_event(event_t *event, unsigned long offset, unsigned long head) { char level; - int show = 0; struct symbol *sym = NULL; struct thread *thread; u64 ip = event->ip.ip; @@ -657,42 +655,35 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) cpumode = event->header.misc & PERF_RECORD_MISC_CPUMODE_MASK; if (cpumode == PERF_RECORD_MISC_KERNEL) { - show = SHOW_KERNEL; level = 'k'; - sym = kernel_maps__find_symbol(ip, &map); dump_printf(" ...... dso: %s\n", map ? map->dso->long_name : ""); } else if (cpumode == PERF_RECORD_MISC_USER) { - - show = SHOW_USER; level = '.'; sym = resolve_symbol(thread, &map, &ip); } else { - show = SHOW_HV; level = 'H'; - dump_printf(" ...... dso: [hypervisor]\n"); } - if (show & show_mask) { - if (dso_list && - (!map || !map->dso || - !(strlist__has_entry(dso_list, map->dso->short_name) || - (map->dso->short_name != map->dso->long_name && - strlist__has_entry(dso_list, map->dso->long_name))))) - return 0; + if (dso_list && + (!map || !map->dso || + !(strlist__has_entry(dso_list, map->dso->short_name) || + (map->dso->short_name != map->dso->long_name && + strlist__has_entry(dso_list, map->dso->long_name))))) + return 0; - if (sym_list && sym && !strlist__has_entry(sym_list, sym->name)) - return 0; + if (sym_list && sym && !strlist__has_entry(sym_list, sym->name)) + return 0; - if (hist_entry__add(thread, map, sym, ip, - chain, level, period)) { - eprintf("problem incrementing symbol count, skipping event\n"); - return -1; - } + if (hist_entry__add(thread, map, sym, ip, + chain, level, period)) { + eprintf("problem incrementing symbol count, skipping event\n"); + return -1; } + total += period; return 0; diff --git a/tools/perf/util/event.h b/tools/perf/util/event.h index a39520e6ae8..c2e62be6279 100644 --- a/tools/perf/util/event.h +++ b/tools/perf/util/event.h @@ -6,12 +6,6 @@ #include #include -enum { - SHOW_KERNEL = 1, - SHOW_USER = 2, - SHOW_HV = 4, -}; - /* * PERF_SAMPLE_IP | PERF_SAMPLE_TID | * */ -- cgit v1.2.3-18-g5258 From 5c2068059a0e852f72b7c2608d92170b752d821f Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 5 Oct 2009 14:26:15 -0300 Subject: perf top: Keep the default of asking for kernel module symbols MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Peter Zijlstra Cc: Mike Galbraith LKML-Reference: Signed-off-by: Ingo Molnar --- tools/perf/builtin-top.c | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index befef842757..34d48c1b7a8 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -804,10 +804,8 @@ static int symbol_filter(struct map *map, struct symbol *sym) static int parse_symbols(void) { - int use_modules = vmlinux_name ? 1 : 0; - if (dsos__load_kernel(vmlinux_name, sizeof(struct sym_entry), - symbol_filter, verbose, use_modules) <= 0) + symbol_filter, verbose, 1) <= 0) return -1; if (dump_symtab) -- cgit v1.2.3-18-g5258 From af427bf529c5991be8d1a36f43e2d0141f532f63 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 5 Oct 2009 14:26:17 -0300 Subject: perf tools: Create maps for modules when processing kallsyms MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit So that we get kallsyms processing closer to vmlinux + modules symtabs processing. One change in behaviour is that since when one specifies --vmlinux -m should be used to ask for modules, so it is now for kallsyms as well. Also continue if one manages to load the vmlinux data but module processing fails, so that at least some analisys can be done with part of the needed symbols. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Peter Zijlstra Cc: Mike Galbraith LKML-Reference: Signed-off-by: Ingo Molnar --- tools/perf/util/symbol.c | 161 +++++++++++++++++++++++++++++++++++------------ tools/perf/util/symbol.h | 2 - 2 files changed, 122 insertions(+), 41 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index e8829689947..4dfdefd5ec7 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -27,6 +27,44 @@ enum dso_origin { static void dsos__add(struct dso *dso); static struct dso *dsos__find(const char *name); +static struct rb_root kernel_maps; + +static void dso__set_symbols_end(struct dso *self) +{ + struct rb_node *nd, *prevnd = rb_first(&self->syms); + + if (prevnd == NULL) + return; + + for (nd = rb_next(prevnd); nd; nd = rb_next(nd)) { + struct symbol *prev = rb_entry(prevnd, struct symbol, rb_node), + *curr = rb_entry(nd, struct symbol, rb_node); + + if (prev->end == prev->start) + prev->end = curr->start - 1; + prevnd = nd; + } +} + +static void kernel_maps__fixup_sym_end(void) +{ + struct map *prev, *curr; + struct rb_node *nd, *prevnd = rb_first(&kernel_maps); + + if (prevnd == NULL) + return; + + curr = rb_entry(prevnd, struct map, rb_node); + dso__set_symbols_end(curr->dso); + + for (nd = rb_next(prevnd); nd; nd = rb_next(nd)) { + prev = curr; + curr = rb_entry(nd, struct map, rb_node); + prev->end = curr->start - 1; + dso__set_symbols_end(curr->dso); + } +} + static struct symbol *symbol__new(u64 start, u64 len, const char *name, unsigned int priv_size, int v) { @@ -162,10 +200,9 @@ size_t dso__fprintf(struct dso *self, FILE *fp) return ret; } -static int dso__load_kallsyms(struct dso *self, struct map *map, - symbol_filter_t filter, int v) +static int maps__load_kallsyms(symbol_filter_t filter, int use_modules, int v) { - struct rb_node *nd, *prevnd; + struct map *map = kernel_map; char *line = NULL; size_t n; FILE *file = fopen("/proc/kallsyms", "r"); @@ -179,6 +216,7 @@ static int dso__load_kallsyms(struct dso *self, struct map *map, struct symbol *sym; int line_len, len; char symbol_type; + char *module, *symbol_name; line_len = getline(&line, &n, file); if (line_len < 0) @@ -201,40 +239,50 @@ static int dso__load_kallsyms(struct dso *self, struct map *map, */ if (symbol_type != 'T' && symbol_type != 'W') continue; + + symbol_name = line + len + 2; + module = strchr(symbol_name, '\t'); + if (module) { + char *module_name_end; + + if (!use_modules) + continue; + *module = '\0'; + module = strchr(module + 1, '['); + if (!module) + continue; + module_name_end = strchr(module + 1, ']'); + if (!module_name_end) + continue; + *(module_name_end + 1) = '\0'; + if (strcmp(map->dso->name, module)) { + map = kernel_maps__find_by_dso_name(module); + if (!map) { + fputs("/proc/{kallsyms,modules} " + "inconsistency!\n", stderr); + return -1; + } + } + start = map->map_ip(map, start); + } else + map = kernel_map; /* * Well fix up the end later, when we have all sorted. */ - sym = symbol__new(start, 0xdead, line + len + 2, - self->sym_priv_size, v); + sym = symbol__new(start, 0, symbol_name, + map->dso->sym_priv_size, v); if (sym == NULL) goto out_delete_line; if (filter && filter(map, sym)) - symbol__delete(sym, self->sym_priv_size); + symbol__delete(sym, map->dso->sym_priv_size); else { - dso__insert_symbol(self, sym); + dso__insert_symbol(map->dso, sym); count++; } } - /* - * Now that we have all sorted out, just set the ->end of all - * symbols - */ - prevnd = rb_first(&self->syms); - - if (prevnd == NULL) - goto out_delete_line; - - for (nd = rb_next(prevnd); nd; nd = rb_next(nd)) { - struct symbol *prev = rb_entry(prevnd, struct symbol, rb_node), - *curr = rb_entry(nd, struct symbol, rb_node); - - prev->end = curr->start - 1; - prevnd = nd; - } - free(line); fclose(file); @@ -246,6 +294,24 @@ out_failure: return -1; } +static size_t kernel_maps__fprintf(FILE *fp) +{ + size_t printed = fprintf(stderr, "Kernel maps:\n"); + struct rb_node *nd; + + printed += map__fprintf(kernel_map, fp); + printed += dso__fprintf(kernel_map->dso, fp); + + for (nd = rb_first(&kernel_maps); nd; nd = rb_next(nd)) { + struct map *pos = rb_entry(nd, struct map, rb_node); + + printed += map__fprintf(pos, fp); + printed += dso__fprintf(pos->dso, fp); + } + + return printed + fprintf(stderr, "END kernel maps\n"); +} + static int dso__load_perf_map(struct dso *self, struct map *map, symbol_filter_t filter, int v) { @@ -598,6 +664,7 @@ static int dso__load_sym(struct dso *self, struct map *map, const char *name, char *demangled; int is_label = elf_sym__is_label(&sym); const char *section_name; + u64 sh_offset = 0; if (!is_label && !elf_sym__is_function(&sym)) continue; @@ -613,14 +680,18 @@ static int dso__load_sym(struct dso *self, struct map *map, const char *name, section_name = elf_sec__name(&shdr, secstrs); + if ((kernel || kmodule)) { + if (strstr(section_name, ".init")) + sh_offset = shdr.sh_offset; + } + if (self->adjust_symbols) { if (v >= 2) printf("adjusting symbol: st_value: %Lx sh_addr: %Lx sh_offset: %Lx\n", (u64)sym.st_value, (u64)shdr.sh_addr, (u64)shdr.sh_offset); sym.st_value -= shdr.sh_addr - shdr.sh_offset; - } else if (kmodule) - sym.st_value += shdr.sh_offset; + } /* * We need to figure out if the object was created from C++ sources * DWARF DW_compile_unit has this, but we don't always have access @@ -631,7 +702,7 @@ static int dso__load_sym(struct dso *self, struct map *map, const char *name, if (demangled != NULL) elf_name = demangled; - f = symbol__new(sym.st_value, sym.st_size, elf_name, + f = symbol__new(sym.st_value + sh_offset, sym.st_size, elf_name, self->sym_priv_size, v); free(demangled); if (!f) @@ -804,7 +875,6 @@ out: return ret; } -static struct rb_root kernel_maps; struct map *kernel_map; static void kernel_maps__insert(struct map *map) @@ -975,8 +1045,7 @@ static struct map *map__new2(u64 start, struct dso *dso) return self; } -int dsos__load_modules(unsigned int sym_priv_size, - symbol_filter_t filter, int v) +static int dsos__load_modules(unsigned int sym_priv_size) { char *line = NULL; size_t n; @@ -1034,8 +1103,7 @@ int dsos__load_modules(unsigned int sym_priv_size, free(line); fclose(file); - v = 1; - return dsos__load_modules_sym(filter, v); + return 0; out_delete_line: free(line); @@ -1075,25 +1143,37 @@ int dsos__load_kernel(const char *vmlinux, unsigned int sym_priv_size, kernel_map->map_ip = vdso__map_ip; + if (use_modules && dsos__load_modules(sym_priv_size) < 0) { + fprintf(stderr, "Failed to load list of modules in use! " + "Continuing...\n"); + use_modules = 0; + } + if (vmlinux) { err = dso__load_vmlinux(dso, kernel_map, vmlinux, filter, v); if (err > 0 && use_modules) { - int syms = dsos__load_modules(sym_priv_size, filter, v); + int syms = dsos__load_modules_sym(filter, v); - if (syms < 0) { - fprintf(stderr, "dsos__load_modules failed!\n"); - return syms; - } - err += syms; + if (syms < 0) + fprintf(stderr, "Failed to read module symbols!" + " Continuing...\n"); + else + err += syms; } } if (err <= 0) - err = dso__load_kallsyms(dso, kernel_map, filter, v); + err = maps__load_kallsyms(filter, use_modules, v); if (err > 0) { struct rb_node *node = rb_first(&dso->syms); struct symbol *sym = rb_entry(node, struct symbol, rb_node); + /* + * Now that we have all sorted out, just set the ->end of all + * symbols that still don't have it. + */ + dso__set_symbols_end(dso); + kernel_maps__fixup_sym_end(); kernel_map->start = sym->start; node = rb_last(&dso->syms); @@ -1106,6 +1186,9 @@ int dsos__load_kernel(const char *vmlinux, unsigned int sym_priv_size, * kernel_maps__insert(kernel_map) */ dsos__add(dso); + + if (v > 0) + kernel_maps__fprintf(stderr); } return err; diff --git a/tools/perf/util/symbol.h b/tools/perf/util/symbol.h index 5339fd82ec9..2e4522edeb0 100644 --- a/tools/perf/util/symbol.h +++ b/tools/perf/util/symbol.h @@ -70,8 +70,6 @@ struct symbol *dso__find_symbol(struct dso *self, u64 ip); int dsos__load_kernel(const char *vmlinux, unsigned int sym_priv_size, symbol_filter_t filter, int verbose, int modules); -int dsos__load_modules(unsigned int sym_priv_size, symbol_filter_t filter, - int verbose); int dso__load(struct dso *self, struct map *map, symbol_filter_t filter, int verbose); struct dso *dsos__findnew(const char *name); -- cgit v1.2.3-18-g5258 From a2a99e8e12798706ec1026e5d8fc36f7c86122ce Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 5 Oct 2009 14:26:18 -0300 Subject: perf tools: /proc/modules names don't always match its name MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit $ cut -d' ' -f1 /proc/modules|grep _|wc -l 29 $ cut -d' ' -f1 /proc/modules|grep _|sed 's/$/.ko'/g|while read n;do find /lib/modules/`uname -r` -name $n;done|wc -l 12 For instance: $ grep ^aes_x86 /proc/modules aes_x86_64 9056 2 - Live 0xffffffffa0091000 $ l /lib/modules/2.6.31-tip/kernel/arch/x86/crypto/aes-x86_64.ko -rw-r--r-- 1 root root 136438 2009-09-22 19:05 /lib/modules/2.6.31-tip/kernel/arch/x86/crypto/aes-x86_64.ko Handle that by introducing a strxfrchar routine that replaces dashes with underscores when matching file names to loaded modules. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Peter Zijlstra Cc: Mike Galbraith LKML-Reference: Signed-off-by: Ingo Molnar --- tools/perf/util/string.c | 11 +++++++++++ tools/perf/util/string.h | 1 + tools/perf/util/symbol.c | 3 ++- 3 files changed, 14 insertions(+), 1 deletion(-) (limited to 'tools/perf') diff --git a/tools/perf/util/string.c b/tools/perf/util/string.c index c93eca9a7be..04743d3e903 100644 --- a/tools/perf/util/string.c +++ b/tools/perf/util/string.c @@ -1,3 +1,4 @@ +#include #include "string.h" static int hex(char ch) @@ -32,3 +33,13 @@ int hex2u64(const char *ptr, u64 *long_val) return p - ptr; } + +char *strxfrchar(char *s, char from, char to) +{ + char *p = s; + + while ((p = strchr(p, from)) != NULL) + *p++ = to; + + return s; +} diff --git a/tools/perf/util/string.h b/tools/perf/util/string.h index 15c827475e7..2c84bf65ba0 100644 --- a/tools/perf/util/string.h +++ b/tools/perf/util/string.h @@ -4,6 +4,7 @@ #include "types.h" int hex2u64(const char *ptr, u64 *val); +char *strxfrchar(char *s, char from, char to); #define _STR(x) #x #define STR(x) _STR(x) diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 4dfdefd5ec7..e3eebdd682d 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -189,7 +189,7 @@ struct symbol *dso__find_symbol(struct dso *self, u64 ip) size_t dso__fprintf(struct dso *self, FILE *fp) { - size_t ret = fprintf(fp, "dso: %s\n", self->long_name); + size_t ret = fprintf(fp, "dso: %s\n", self->short_name); struct rb_node *nd; for (nd = rb_first(&self->syms); nd; nd = rb_next(nd)) { @@ -977,6 +977,7 @@ static int dsos__load_modules_sym_dir(char *dirname, snprintf(dso_name, sizeof(dso_name), "[%.*s]", (int)(dot - dent->d_name), dent->d_name); + strxfrchar(dso_name, '-', '_'); map = kernel_maps__find_by_dso_name(dso_name); if (map == NULL) continue; -- cgit v1.2.3-18-g5258 From c3b32fcbc7f4fd9a9b84718b991b175b0fd53f8c Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 5 Oct 2009 14:26:16 -0300 Subject: perf report: Use kernel_maps__find_symbol as fallback to find vdsos, etc MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit In resolve_symbol, as we're moving to breaking the kernel symbols list per address ranges, i.e. kernel linking sections, so that we don't have a big kernel_map that in its range covers what is in the modules. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Peter Zijlstra Cc: Mike Galbraith LKML-Reference: Signed-off-by: Ingo Molnar --- tools/perf/builtin-report.c | 7 ++----- 1 file changed, 2 insertions(+), 5 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index fe4aadc9630..12f8c868fcd 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -384,11 +384,8 @@ got_map: * the "[vdso]" dso, but for now lets use the old * trick of looking in the whole kernel symbol list. */ - if ((long long)ip < 0) { - map = kernel_map; - if (mapp) - *mapp = map; - } + if ((long long)ip < 0) + return kernel_maps__find_symbol(ip, mapp); } dump_printf(" ...... dso: %s\n", map ? map->dso->long_name : ""); -- cgit v1.2.3-18-g5258 From 26a50744b21fff65bd754874072857bee8967f4d Mon Sep 17 00:00:00 2001 From: Tom Zanussi Date: Tue, 6 Oct 2009 01:09:50 -0500 Subject: tracing/events: Add 'signed' field to format files The sign info used for filters in the kernel is also useful to applications that process the trace stream. Add it to the format files and make it available to userspace. Signed-off-by: Tom Zanussi Acked-by: Frederic Weisbecker Cc: rostedt@goodmis.org Cc: lizf@cn.fujitsu.com Cc: hch@infradead.org Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo LKML-Reference: <1254809398-8078-2-git-send-email-tzanussi@gmail.com> Signed-off-by: Ingo Molnar --- tools/perf/util/trace-event-parse.c | 24 ++++++++++++++++++++++++ tools/perf/util/trace-event.h | 1 + 2 files changed, 25 insertions(+) (limited to 'tools/perf') diff --git a/tools/perf/util/trace-event-parse.c b/tools/perf/util/trace-event-parse.c index 55b41b9e383..be8412d699a 100644 --- a/tools/perf/util/trace-event-parse.c +++ b/tools/perf/util/trace-event-parse.c @@ -894,6 +894,21 @@ static int event_read_fields(struct event *event, struct format_field **fields) field->size = strtoul(token, NULL, 0); free_token(token); + if (read_expected(EVENT_OP, (char *)";") < 0) + goto fail_expect; + + if (read_expected(EVENT_ITEM, (char *)"signed") < 0) + goto fail_expect; + + if (read_expected(EVENT_OP, (char *)":") < 0) + goto fail_expect; + + if (read_expect_type(EVENT_ITEM, &token)) + goto fail; + if (strtoul(token, NULL, 0)) + field->flags |= FIELD_IS_SIGNED; + free_token(token); + if (read_expected(EVENT_OP, (char *)";") < 0) goto fail_expect; @@ -2843,6 +2858,15 @@ static void parse_header_field(char *type, return; *size = atoi(token); free_token(token); + if (read_expected(EVENT_OP, (char *)";") < 0) + return; + if (read_expected(EVENT_ITEM, (char *)"signed") < 0) + return; + if (read_expected(EVENT_OP, (char *)":") < 0) + return; + if (read_expect_type(EVENT_ITEM, &token) < 0) + return; + free_token(token); if (read_expected(EVENT_OP, (char *)";") < 0) return; if (read_expect_type(EVENT_NEWLINE, &token) < 0) diff --git a/tools/perf/util/trace-event.h b/tools/perf/util/trace-event.h index 162c3e6deb9..00b440df66d 100644 --- a/tools/perf/util/trace-event.h +++ b/tools/perf/util/trace-event.h @@ -26,6 +26,7 @@ enum { enum format_flags { FIELD_IS_ARRAY = 1, FIELD_IS_POINTER = 2, + FIELD_IS_SIGNED = 4, }; struct format_field { -- cgit v1.2.3-18-g5258 From 2774601811bedd04ee7e38624343ea80b4a62d7e Mon Sep 17 00:00:00 2001 From: Tom Zanussi Date: Tue, 6 Oct 2009 01:09:51 -0500 Subject: perf trace: Add subsystem string to struct event Needed to fully qualify event names for event stream processing. Signed-off-by: Tom Zanussi Acked-by: Frederic Weisbecker Cc: rostedt@goodmis.org Cc: lizf@cn.fujitsu.com Cc: hch@infradead.org Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo LKML-Reference: <1254809398-8078-3-git-send-email-tzanussi@gmail.com> Signed-off-by: Ingo Molnar --- tools/perf/util/trace-event-parse.c | 4 +++- tools/perf/util/trace-event.h | 3 ++- 2 files changed, 5 insertions(+), 2 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/trace-event-parse.c b/tools/perf/util/trace-event-parse.c index be8412d699a..de3fc8bf8bf 100644 --- a/tools/perf/util/trace-event-parse.c +++ b/tools/perf/util/trace-event-parse.c @@ -2950,7 +2950,7 @@ int parse_ftrace_file(char *buf, unsigned long size) return 0; } -int parse_event_file(char *buf, unsigned long size, char *system__unused __unused) +int parse_event_file(char *buf, unsigned long size, char *sys) { struct event *event; int ret; @@ -2977,6 +2977,8 @@ int parse_event_file(char *buf, unsigned long size, char *system__unused __unuse if (ret < 0) die("failed to read event print fmt"); + event->system = strdup(sys); + #define PRINT_ARGS 0 if (PRINT_ARGS && event->print_fmt.args) print_args(event->print_fmt.args); diff --git a/tools/perf/util/trace-event.h b/tools/perf/util/trace-event.h index 00b440df66d..cb92978be30 100644 --- a/tools/perf/util/trace-event.h +++ b/tools/perf/util/trace-event.h @@ -133,6 +133,7 @@ struct event { int flags; struct format format; struct print_fmt print_fmt; + char *system; }; enum { @@ -167,7 +168,7 @@ void print_funcs(void); void print_printk(void); int parse_ftrace_file(char *buf, unsigned long size); -int parse_event_file(char *buf, unsigned long size, char *system); +int parse_event_file(char *buf, unsigned long size, char *sys); void print_event(int cpu, void *data, int size, unsigned long long nsecs, char *comm); -- cgit v1.2.3-18-g5258 From 064739bc4b3d7f424b2f25547e6611bcf0132415 Mon Sep 17 00:00:00 2001 From: Tom Zanussi Date: Tue, 6 Oct 2009 01:09:52 -0500 Subject: perf trace: Add string/dynamic cases to format_flags Needed for distinguishing string fields in event stream processing. Signed-off-by: Tom Zanussi Acked-by: Frederic Weisbecker Cc: rostedt@goodmis.org Cc: lizf@cn.fujitsu.com Cc: hch@infradead.org Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo LKML-Reference: <1254809398-8078-4-git-send-email-tzanussi@gmail.com> Signed-off-by: Ingo Molnar --- tools/perf/util/trace-event-parse.c | 24 ++++++++++++++++++++++++ tools/perf/util/trace-event.h | 2 ++ 2 files changed, 26 insertions(+) (limited to 'tools/perf') diff --git a/tools/perf/util/trace-event-parse.c b/tools/perf/util/trace-event-parse.c index de3fc8bf8bf..6f851f98b5b 100644 --- a/tools/perf/util/trace-event-parse.c +++ b/tools/perf/util/trace-event-parse.c @@ -721,6 +721,24 @@ static int event_read_id(void) return -1; } +static int field_is_string(struct format_field *field) +{ + if ((field->flags & FIELD_IS_ARRAY) && + (!strstr(field->type, "char") || !strstr(field->type, "u8") || + !strstr(field->type, "s8"))) + return 1; + + return 0; +} + +static int field_is_dynamic(struct format_field *field) +{ + if (!strcmp(field->type, "__data_loc")) + return 1; + + return 0; +} + static int event_read_fields(struct event *event, struct format_field **fields) { struct format_field *field = NULL; @@ -865,6 +883,12 @@ static int event_read_fields(struct event *event, struct format_field **fields) free(brackets); } + if (field_is_string(field)) { + field->flags |= FIELD_IS_STRING; + if (field_is_dynamic(field)) + field->flags |= FIELD_IS_DYNAMIC; + } + if (test_type_token(type, token, EVENT_OP, (char *)";")) goto fail; free_token(token); diff --git a/tools/perf/util/trace-event.h b/tools/perf/util/trace-event.h index cb92978be30..5f59a39fb88 100644 --- a/tools/perf/util/trace-event.h +++ b/tools/perf/util/trace-event.h @@ -27,6 +27,8 @@ enum format_flags { FIELD_IS_ARRAY = 1, FIELD_IS_POINTER = 2, FIELD_IS_SIGNED = 4, + FIELD_IS_STRING = 8, + FIELD_IS_DYNAMIC = 16, }; struct format_field { -- cgit v1.2.3-18-g5258 From 42e59d7d19dc4b49feab2a860fd9a8ca3248c833 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Tue, 6 Oct 2009 15:14:21 +0200 Subject: perf tools: Default to 1 KHz auto-sampling freq events Use auto-freq events by default in perf record and perf top. This allows more consistent hardware event sampling, regardless of the intensity of the underlying event. It also keeps us from over-sampling on larger/busier systems. (also make surrounding initializations more consistent) Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker LKML-Reference: Signed-off-by: Ingo Molnar --- tools/perf/builtin-record.c | 52 ++++++++++++++++++++++----------------------- tools/perf/builtin-top.c | 38 ++++++++++++++++----------------- 2 files changed, 45 insertions(+), 45 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 3eeef339c78..494f8c7d752 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -29,43 +29,43 @@ static int fd[MAX_NR_CPUS][MAX_COUNTERS]; static long default_interval = 100000; -static int nr_cpus = 0; +static int nr_cpus = 0; static unsigned int page_size; -static unsigned int mmap_pages = 128; -static int freq = 0; +static unsigned int mmap_pages = 128; +static int freq = 1000; static int output; static const char *output_name = "perf.data"; -static int group = 0; -static unsigned int realtime_prio = 0; -static int raw_samples = 0; -static int system_wide = 0; -static int profile_cpu = -1; -static pid_t target_pid = -1; -static pid_t child_pid = -1; -static int inherit = 1; -static int force = 0; -static int append_file = 0; -static int call_graph = 0; -static int inherit_stat = 0; -static int no_samples = 0; -static int sample_address = 0; -static int multiplex = 0; -static int multiplex_fd = -1; - -static long samples; +static int group = 0; +static unsigned int realtime_prio = 0; +static int raw_samples = 0; +static int system_wide = 0; +static int profile_cpu = -1; +static pid_t target_pid = -1; +static pid_t child_pid = -1; +static int inherit = 1; +static int force = 0; +static int append_file = 0; +static int call_graph = 0; +static int inherit_stat = 0; +static int no_samples = 0; +static int sample_address = 0; +static int multiplex = 0; +static int multiplex_fd = -1; + +static long samples = 0; static struct timeval last_read; static struct timeval this_read; -static u64 bytes_written; +static u64 bytes_written = 0; static struct pollfd event_array[MAX_NR_CPUS * MAX_COUNTERS]; -static int nr_poll; -static int nr_cpu; +static int nr_poll = 0; +static int nr_cpu = 0; -static int file_new = 1; +static int file_new = 1; -struct perf_header *header; +struct perf_header *header = NULL; struct mmap_data { int counter; diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index c574c5b3d0e..d978dc99236 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -55,26 +55,26 @@ static int fd[MAX_NR_CPUS][MAX_COUNTERS]; -static int system_wide = 0; +static int system_wide = 0; static int default_interval = 100000; -static int count_filter = 5; -static int print_entries = 15; +static int count_filter = 5; +static int print_entries = 15; -static int target_pid = -1; -static int inherit = 0; -static int profile_cpu = -1; -static int nr_cpus = 0; -static unsigned int realtime_prio = 0; -static int group = 0; +static int target_pid = -1; +static int inherit = 0; +static int profile_cpu = -1; +static int nr_cpus = 0; +static unsigned int realtime_prio = 0; +static int group = 0; static unsigned int page_size; -static unsigned int mmap_pages = 16; -static int freq = 0; +static unsigned int mmap_pages = 16; +static int freq = 1000; /* 1 KHz */ -static int delay_secs = 2; -static int zero; -static int dump_symtab; +static int delay_secs = 2; +static int zero = 0; +static int dump_symtab = 0; /* * Source @@ -87,11 +87,11 @@ struct source_line { struct source_line *next; }; -static char *sym_filter = NULL; -struct sym_entry *sym_filter_entry = NULL; -static int sym_pcnt_filter = 5; -static int sym_counter = 0; -static int display_weighted = -1; +static char *sym_filter = NULL; +struct sym_entry *sym_filter_entry = NULL; +static int sym_pcnt_filter = 5; +static int sym_counter = 0; +static int display_weighted = -1; /* * Symbols -- cgit v1.2.3-18-g5258 From b209aa1f83964d49a332a7b6b818ebede5cdc6ef Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Tue, 6 Oct 2009 21:21:26 +0200 Subject: perf tools: Start the perf.data mapping at data offset in perf trace Currently, we are mapping perf.data in the beginning of the file and use the data offset as a buffer offset. This may exceed the mapping area if the data offset is upper than page_size * mmap_window and result in a page fault (thing that happen if we merge trace.info in perf.data). Instead, let's start the mapping in the page that matches our data offset. v2: Drop a junk from another patch (trace_report() removal) Signed-off-by: Frederic Weisbecker Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Mike Galbraith Cc: Paul Mackerras Cc: Tom Zanussi LKML-Reference: <1254856886-10348-1-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar --- tools/perf/builtin-trace.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-trace.c index 5d4c84d8637..d573d4ea6c2 100644 --- a/tools/perf/builtin-trace.c +++ b/tools/perf/builtin-trace.c @@ -143,6 +143,7 @@ static int __cmd_trace(void) int ret, rc = EXIT_FAILURE; unsigned long offset = 0; unsigned long head = 0; + unsigned long shift; struct stat perf_stat; event_t *event; uint32_t size; @@ -180,6 +181,10 @@ static int __cmd_trace(void) return EXIT_FAILURE; } + shift = page_size * (head / page_size); + offset += shift; + head -= shift; + remap: buf = (char *)mmap(NULL, page_size * mmap_window, PROT_READ, MAP_SHARED, input, offset); @@ -192,9 +197,9 @@ more: event = (event_t *)(buf + head); if (head + event->header.size >= page_size * mmap_window) { - unsigned long shift = page_size * (head / page_size); int res; + shift = page_size * (head / page_size); res = munmap(buf, page_size * mmap_window); assert(res == 0); -- cgit v1.2.3-18-g5258 From 03456a158d9067d2f657bec170506009db81756d Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Tue, 6 Oct 2009 23:36:47 +0200 Subject: perf tools: Merge trace.info content into perf.data This drops the trace.info file and move its contents into the common perf.data file. This is done by creating a new trace_info section into this file. A user of perf headers needs to call perf_header__set_trace_info() to save the trace meta informations into the perf.data file. A file created by perf after his patch is unsupported by previous version because the size of the headers have increased. That said, it's two new fields that have been added in the end of the headers, and those could be ignored by previous versions if they just handled the dynamic header size and then ignore the unknow part. The offsets guarantee the compatibility. We'll do a -stable fix for that. But current previous versions handle the header size using its static size, not dynamic, then it's not backward compatible with trace records. Signed-off-by: Frederic Weisbecker Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Paul Mackerras Cc: Mike Galbraith Cc: Paul Mackerras LKML-Reference: <20091006213643.GA5343@nowhere> Signed-off-by: Ingo Molnar --- tools/perf/builtin-record.c | 7 +++---- tools/perf/builtin-sched.c | 1 - tools/perf/builtin-trace.c | 1 - tools/perf/util/header.c | 42 ++++++++++++++++++++++++++++++++++++++ tools/perf/util/header.h | 4 +++- tools/perf/util/trace-event-info.c | 6 ++---- tools/perf/util/trace-event-read.c | 7 ++----- tools/perf/util/trace-event.h | 4 ++-- 8 files changed, 54 insertions(+), 18 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 494f8c7d752..59af03d80d0 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -17,7 +17,6 @@ #include "util/header.h" #include "util/event.h" #include "util/debug.h" -#include "util/trace-event.h" #include #include @@ -566,17 +565,17 @@ static int __cmd_record(int argc, const char **argv) else header = perf_header__new(); - if (raw_samples) { - read_tracing_data(attrs, nr_counters); + perf_header__set_trace_info(); } else { for (i = 0; i < nr_counters; i++) { if (attrs[i].sample_type & PERF_SAMPLE_RAW) { - read_tracing_data(attrs, nr_counters); + perf_header__set_trace_info(); break; } } } + atexit(atexit_header); if (!system_wide) { diff --git a/tools/perf/builtin-sched.c b/tools/perf/builtin-sched.c index 4470f253570..18871380b01 100644 --- a/tools/perf/builtin-sched.c +++ b/tools/perf/builtin-sched.c @@ -1634,7 +1634,6 @@ static int read_events(void) uint32_t size; char *buf; - trace_report(); register_idle_thread(&threads, &last_match); input = open(input_name, O_RDONLY); diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-trace.c index d573d4ea6c2..d9abb4ae5f7 100644 --- a/tools/perf/builtin-trace.c +++ b/tools/perf/builtin-trace.c @@ -149,7 +149,6 @@ static int __cmd_trace(void) uint32_t size; char *buf; - trace_report(); register_idle_thread(&threads, &last_match); input = open(input_name, O_RDONLY); diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index e306857b2c2..212fade7ee7 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -5,6 +5,8 @@ #include "util.h" #include "header.h" +#include "../perf.h" +#include "trace-event.h" /* * Create new perf.data header attribute: @@ -62,6 +64,8 @@ struct perf_header *perf_header__new(void) self->data_offset = 0; self->data_size = 0; + self->trace_info_offset = 0; + self->trace_info_size = 0; return self; } @@ -145,8 +149,16 @@ struct perf_file_header { struct perf_file_section attrs; struct perf_file_section data; struct perf_file_section event_types; + struct perf_file_section trace_info; }; +static int trace_info; + +void perf_header__set_trace_info(void) +{ + trace_info = 1; +} + static void do_write(int fd, void *buf, size_t size) { while (size) { @@ -198,6 +210,23 @@ void perf_header__write(struct perf_header *self, int fd) if (events) do_write(fd, events, self->event_size); + if (trace_info) { + static int trace_info_written; + + /* + * Write it only once + */ + if (!trace_info_written) { + self->trace_info_offset = lseek(fd, 0, SEEK_CUR); + read_tracing_data(fd, attrs, nr_counters); + self->trace_info_size = lseek(fd, 0, SEEK_CUR) - + self->trace_info_offset; + trace_info_written = 1; + } else { + lseek(fd, self->trace_info_offset + + self->trace_info_size, SEEK_SET); + } + } self->data_offset = lseek(fd, 0, SEEK_CUR); @@ -217,6 +246,10 @@ void perf_header__write(struct perf_header *self, int fd) .offset = self->event_offset, .size = self->event_size, }, + .trace_info = { + .offset = self->trace_info_offset, + .size = self->trace_info_size, + }, }; lseek(fd, 0, SEEK_SET); @@ -290,6 +323,15 @@ struct perf_header *perf_header__read(int fd) do_read(fd, events, f_header.event_types.size); event_count = f_header.event_types.size / sizeof(struct perf_trace_event_type); } + + self->trace_info_offset = f_header.trace_info.offset; + self->trace_info_size = f_header.trace_info.size; + + if (self->trace_info_size) { + lseek(fd, self->trace_info_offset, SEEK_SET); + trace_report(fd); + } + self->event_offset = f_header.event_types.offset; self->event_size = f_header.event_types.size; diff --git a/tools/perf/util/header.h b/tools/perf/util/header.h index a2916b652a1..30aee5160dc 100644 --- a/tools/perf/util/header.h +++ b/tools/perf/util/header.h @@ -21,6 +21,8 @@ struct perf_header { u64 data_size; u64 event_offset; u64 event_size; + u64 trace_info_offset; + u64 trace_info_size; }; struct perf_header *perf_header__read(int fd); @@ -40,7 +42,7 @@ void perf_header_attr__add_id(struct perf_header_attr *self, u64 id); u64 perf_header__sample_type(struct perf_header *header); struct perf_event_attr * perf_header__find_attr(u64 id, struct perf_header *header); - +void perf_header__set_trace_info(void); struct perf_header *perf_header__new(void); diff --git a/tools/perf/util/trace-event-info.c b/tools/perf/util/trace-event-info.c index af4b0573b37..831052d4b4f 100644 --- a/tools/perf/util/trace-event-info.c +++ b/tools/perf/util/trace-event-info.c @@ -496,14 +496,12 @@ get_tracepoints_path(struct perf_event_attr *pattrs, int nb_events) return path.next; } -void read_tracing_data(struct perf_event_attr *pattrs, int nb_events) +void read_tracing_data(int fd, struct perf_event_attr *pattrs, int nb_events) { char buf[BUFSIZ]; struct tracepoint_path *tps; - output_fd = open(output_file, O_WRONLY | O_CREAT | O_TRUNC | O_LARGEFILE, 0644); - if (output_fd < 0) - die("creating file '%s'", output_file); + output_fd = fd; buf[0] = 23; buf[1] = 8; diff --git a/tools/perf/util/trace-event-read.c b/tools/perf/util/trace-event-read.c index 1b5c847d2c2..44292e06cca 100644 --- a/tools/perf/util/trace-event-read.c +++ b/tools/perf/util/trace-event-read.c @@ -458,9 +458,8 @@ struct record *trace_read_data(int cpu) return data; } -void trace_report(void) +void trace_report(int fd) { - const char *input_file = "trace.info"; char buf[BUFSIZ]; char test[] = { 23, 8, 68 }; char *version; @@ -468,9 +467,7 @@ void trace_report(void) int show_funcs = 0; int show_printk = 0; - input_fd = open(input_file, O_RDONLY); - if (input_fd < 0) - die("opening '%s'\n", input_file); + input_fd = fd; read_or_die(buf, 3); if (memcmp(buf, test, 3) != 0) diff --git a/tools/perf/util/trace-event.h b/tools/perf/util/trace-event.h index 5f59a39fb88..da77e073c86 100644 --- a/tools/perf/util/trace-event.h +++ b/tools/perf/util/trace-event.h @@ -158,7 +158,7 @@ struct record *trace_read_data(int cpu); void parse_set_info(int nr_cpus, int long_sz); -void trace_report(void); +void trace_report(int fd); void *malloc_or_die(unsigned int size); @@ -244,6 +244,6 @@ unsigned long long raw_field_value(struct event *event, const char *name, void *data); void *raw_field_ptr(struct event *event, const char *name, void *data); -void read_tracing_data(struct perf_event_attr *pattrs, int nb_events); +void read_tracing_data(int fd, struct perf_event_attr *pattrs, int nb_events); #endif /* __PERF_TRACE_EVENTS_H */ -- cgit v1.2.3-18-g5258 From 016e92fbc9ef33689cf654f343a94383d43235e7 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Wed, 7 Oct 2009 12:47:31 +0200 Subject: perf tools: Unify perf.data mapping and events handling This librarizes the perf.data file mapping and handling in various perf tools, roughly reducing the amount of code and fixing the places that mmap from beginning of the file whereas we want to mmap from the beginning of the data, leading to page fault because the mmap window is too small since the trace info are written in the file too. TODO: - convert perf timechart too Signed-off-by: Frederic Weisbecker Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arjan van de Ven LKML-Reference: <20091007104729.GD5043@nowhere> Signed-off-by: Ingo Molnar --- tools/perf/Makefile | 2 + tools/perf/builtin-report.c | 211 +++++++---------------------------------- tools/perf/builtin-sched.c | 140 ++++++---------------------- tools/perf/builtin-trace.c | 129 ++++--------------------- tools/perf/util/data_map.c | 222 ++++++++++++++++++++++++++++++++++++++++++++ tools/perf/util/data_map.h | 31 +++++++ 6 files changed, 334 insertions(+), 401 deletions(-) create mode 100644 tools/perf/util/data_map.c create mode 100644 tools/perf/util/data_map.h (limited to 'tools/perf') diff --git a/tools/perf/Makefile b/tools/perf/Makefile index 5a429966c99..495eb6d97fa 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -342,6 +342,7 @@ LIB_H += util/values.h LIB_H += util/sort.h LIB_H += util/hist.h LIB_H += util/thread.h +LIB_H += util/data_map.h LIB_OBJS += util/abspath.o LIB_OBJS += util/alias.o @@ -378,6 +379,7 @@ LIB_OBJS += util/trace-event-info.o LIB_OBJS += util/svghelper.o LIB_OBJS += util/sort.o LIB_OBJS += util/hist.o +LIB_OBJS += util/data_map.o BUILTIN_OBJS += builtin-annotate.o BUILTIN_OBJS += builtin-help.o diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index 12f8c868fcd..87c4582303b 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -26,6 +26,7 @@ #include "util/parse-options.h" #include "util/parse-events.h" +#include "util/data_map.h" #include "util/thread.h" #include "util/sort.h" #include "util/hist.h" @@ -37,7 +38,6 @@ static char *dso_list_str, *comm_list_str, *sym_list_str, static struct strlist *dso_list, *comm_list, *sym_list; static int force; -static int input; static int full_paths; static int show_nr_samples; @@ -48,15 +48,11 @@ static struct perf_read_values show_threads_values; static char default_pretty_printing_style[] = "normal"; static char *pretty_printing_style = default_pretty_printing_style; -static unsigned long page_size; -static unsigned long mmap_window = 32; - static int exclude_other = 1; static char callchain_default_opt[] = "fractal,0.5"; -static char __cwd[PATH_MAX]; -static char *cwd = __cwd; +static char *cwd; static int cwdlen; static struct rb_root threads; @@ -815,208 +811,71 @@ process_read_event(event_t *event, unsigned long offset, unsigned long head) return 0; } -static int -process_event(event_t *event, unsigned long offset, unsigned long head) -{ - trace_event(event); - - switch (event->header.type) { - case PERF_RECORD_SAMPLE: - return process_sample_event(event, offset, head); - - case PERF_RECORD_MMAP: - return process_mmap_event(event, offset, head); - - case PERF_RECORD_COMM: - return process_comm_event(event, offset, head); - - case PERF_RECORD_FORK: - case PERF_RECORD_EXIT: - return process_task_event(event, offset, head); - - case PERF_RECORD_LOST: - return process_lost_event(event, offset, head); - - case PERF_RECORD_READ: - return process_read_event(event, offset, head); - - /* - * We dont process them right now but they are fine: - */ - - case PERF_RECORD_THROTTLE: - case PERF_RECORD_UNTHROTTLE: - return 0; - - default: - return -1; - } - - return 0; -} - -static int __cmd_report(void) +static int sample_type_check(u64 type) { - int ret, rc = EXIT_FAILURE; - unsigned long offset = 0; - unsigned long head, shift; - struct stat input_stat; - struct thread *idle; - event_t *event; - uint32_t size; - char *buf; - - idle = register_idle_thread(&threads, &last_match); - thread__comm_adjust(idle); - - if (show_threads) - perf_read_values_init(&show_threads_values); - - input = open(input_name, O_RDONLY); - if (input < 0) { - fprintf(stderr, " failed to open file: %s", input_name); - if (!strcmp(input_name, "perf.data")) - fprintf(stderr, " (try 'perf record' first)"); - fprintf(stderr, "\n"); - exit(-1); - } - - ret = fstat(input, &input_stat); - if (ret < 0) { - perror("failed to stat file"); - exit(-1); - } - - if (!force && input_stat.st_uid && (input_stat.st_uid != geteuid())) { - fprintf(stderr, "file: %s not owned by current user or root\n", input_name); - exit(-1); - } - - if (!input_stat.st_size) { - fprintf(stderr, "zero-sized file, nothing to do!\n"); - exit(0); - } - - header = perf_header__read(input); - head = header->data_offset; - - sample_type = perf_header__sample_type(header); + sample_type = type; if (!(sample_type & PERF_SAMPLE_CALLCHAIN)) { if (sort__has_parent) { fprintf(stderr, "selected --sort parent, but no" " callchain data. Did you call" " perf record without -g?\n"); - exit(-1); + return -1; } if (callchain) { fprintf(stderr, "selected -g but no callchain data." " Did you call perf record without" " -g?\n"); - exit(-1); + return -1; } } else if (callchain_param.mode != CHAIN_NONE && !callchain) { callchain = 1; if (register_callchain_param(&callchain_param) < 0) { fprintf(stderr, "Can't register callchain" " params\n"); - exit(-1); + return -1; } } - if (load_kernel() < 0) { - perror("failed to load kernel symbols"); - return EXIT_FAILURE; - } - - if (!full_paths) { - if (getcwd(__cwd, sizeof(__cwd)) == NULL) { - perror("failed to get the current directory"); - return EXIT_FAILURE; - } - cwdlen = strlen(cwd); - } else { - cwd = NULL; - cwdlen = 0; - } - - shift = page_size * (head / page_size); - offset += shift; - head -= shift; - -remap: - buf = (char *)mmap(NULL, page_size * mmap_window, PROT_READ, - MAP_SHARED, input, offset); - if (buf == MAP_FAILED) { - perror("failed to mmap file"); - exit(-1); - } - -more: - event = (event_t *)(buf + head); - - size = event->header.size; - if (!size) - size = 8; - - if (head + event->header.size >= page_size * mmap_window) { - int munmap_ret; - - shift = page_size * (head / page_size); - - munmap_ret = munmap(buf, page_size * mmap_window); - assert(munmap_ret == 0); - - offset += shift; - head -= shift; - goto remap; - } - - size = event->header.size; - - dump_printf("\n%p [%p]: event: %d\n", - (void *)(offset + head), - (void *)(long)event->header.size, - event->header.type); - - if (!size || process_event(event, offset, head) < 0) { - - dump_printf("%p [%p]: skipping unknown header type: %d\n", - (void *)(offset + head), - (void *)(long)(event->header.size), - event->header.type); - - total_unknown++; + return 0; +} - /* - * assume we lost track of the stream, check alignment, and - * increment a single u64 in the hope to catch on again 'soon'. - */ +static struct perf_file_handler file_handler = { + .process_sample_event = process_sample_event, + .process_mmap_event = process_mmap_event, + .process_comm_event = process_comm_event, + .process_exit_event = process_task_event, + .process_fork_event = process_task_event, + .process_lost_event = process_lost_event, + .process_read_event = process_read_event, + .sample_type_check = sample_type_check, +}; - if (unlikely(head & 7)) - head &= ~7ULL; - size = 8; - } +static int __cmd_report(void) +{ + struct thread *idle; + int ret; - head += size; + idle = register_idle_thread(&threads, &last_match); + thread__comm_adjust(idle); - if (offset + head >= header->data_offset + header->data_size) - goto done; + if (show_threads) + perf_read_values_init(&show_threads_values); - if (offset + head < (unsigned long)input_stat.st_size) - goto more; + register_perf_file_handler(&file_handler); -done: - rc = EXIT_SUCCESS; - close(input); + ret = mmap_dispatch_perf_file(&header, input_name, force, full_paths, + &cwdlen, &cwd); + if (ret) + return ret; dump_printf(" IP events: %10ld\n", total); dump_printf(" mmap events: %10ld\n", total_mmap); dump_printf(" comm events: %10ld\n", total_comm); dump_printf(" fork events: %10ld\n", total_fork); dump_printf(" lost events: %10ld\n", total_lost); - dump_printf(" unknown events: %10ld\n", total_unknown); + dump_printf(" unknown events: %10ld\n", file_handler.total_unknown); if (dump_trace) return 0; @@ -1034,7 +893,7 @@ done: if (show_threads) perf_read_values_destroy(&show_threads_values); - return rc; + return ret; } static int @@ -1177,8 +1036,6 @@ int cmd_report(int argc, const char **argv, const char *prefix __used) { symbol__init(); - page_size = getpagesize(); - argc = parse_options(argc, argv, options, report_usage, 0); setup_sorting(); diff --git a/tools/perf/builtin-sched.c b/tools/perf/builtin-sched.c index 18871380b01..e1df7055ab8 100644 --- a/tools/perf/builtin-sched.c +++ b/tools/perf/builtin-sched.c @@ -11,6 +11,7 @@ #include "util/trace-event.h" #include "util/debug.h" +#include "util/data_map.h" #include #include @@ -20,9 +21,6 @@ #include static char const *input_name = "perf.data"; -static int input; -static unsigned long page_size; -static unsigned long mmap_window = 32; static unsigned long total_comm = 0; @@ -35,6 +33,9 @@ static u64 sample_type; static char default_sort_order[] = "avg, max, switch, runtime"; static char *sort_order = default_sort_order; +static char *cwd; +static int cwdlen; + #define PR_SET_NAME 15 /* Set process name */ #define MAX_CPUS 4096 @@ -1594,129 +1595,43 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) } static int -process_event(event_t *event, unsigned long offset, unsigned long head) +process_lost_event(event_t *event __used, + unsigned long offset __used, + unsigned long head __used) { - trace_event(event); - - nr_events++; - switch (event->header.type) { - case PERF_RECORD_MMAP: - return 0; - case PERF_RECORD_LOST: - nr_lost_chunks++; - nr_lost_events += event->lost.lost; - return 0; - - case PERF_RECORD_COMM: - return process_comm_event(event, offset, head); + nr_lost_chunks++; + nr_lost_events += event->lost.lost; - case PERF_RECORD_EXIT ... PERF_RECORD_READ: - return 0; + return 0; +} - case PERF_RECORD_SAMPLE: - return process_sample_event(event, offset, head); +static int sample_type_check(u64 type) +{ + sample_type = type; - case PERF_RECORD_MAX: - default: + if (!(sample_type & PERF_SAMPLE_RAW)) { + fprintf(stderr, + "No trace sample to read. Did you call perf record " + "without -R?"); return -1; } return 0; } +static struct perf_file_handler file_handler = { + .process_sample_event = process_sample_event, + .process_comm_event = process_comm_event, + .process_lost_event = process_lost_event, + .sample_type_check = sample_type_check, +}; + static int read_events(void) { - int ret, rc = EXIT_FAILURE; - unsigned long offset = 0; - unsigned long head = 0; - struct stat perf_stat; - event_t *event; - uint32_t size; - char *buf; - register_idle_thread(&threads, &last_match); + register_perf_file_handler(&file_handler); - input = open(input_name, O_RDONLY); - if (input < 0) { - perror("failed to open file"); - exit(-1); - } - - ret = fstat(input, &perf_stat); - if (ret < 0) { - perror("failed to stat file"); - exit(-1); - } - - if (!perf_stat.st_size) { - fprintf(stderr, "zero-sized file, nothing to do!\n"); - exit(0); - } - header = perf_header__read(input); - head = header->data_offset; - sample_type = perf_header__sample_type(header); - - if (!(sample_type & PERF_SAMPLE_RAW)) - die("No trace sample to read. Did you call perf record " - "without -R?"); - - if (load_kernel() < 0) { - perror("failed to load kernel symbols"); - return EXIT_FAILURE; - } - -remap: - buf = (char *)mmap(NULL, page_size * mmap_window, PROT_READ, - MAP_SHARED, input, offset); - if (buf == MAP_FAILED) { - perror("failed to mmap file"); - exit(-1); - } - -more: - event = (event_t *)(buf + head); - - size = event->header.size; - if (!size) - size = 8; - - if (head + event->header.size >= page_size * mmap_window) { - unsigned long shift = page_size * (head / page_size); - int res; - - res = munmap(buf, page_size * mmap_window); - assert(res == 0); - - offset += shift; - head -= shift; - goto remap; - } - - size = event->header.size; - - - if (!size || process_event(event, offset, head) < 0) { - - /* - * assume we lost track of the stream, check alignment, and - * increment a single u64 in the hope to catch on again 'soon'. - */ - - if (unlikely(head & 7)) - head &= ~7ULL; - - size = 8; - } - - head += size; - - if (offset + head < (unsigned long)perf_stat.st_size) - goto more; - - rc = EXIT_SUCCESS; - close(input); - - return rc; + return mmap_dispatch_perf_file(&header, input_name, 0, 0, &cwdlen, &cwd); } static void print_bad_events(void) @@ -1934,7 +1849,6 @@ static int __cmd_record(int argc, const char **argv) int cmd_sched(int argc, const char **argv, const char *prefix __used) { symbol__init(); - page_size = getpagesize(); argc = parse_options(argc, argv, sched_options, sched_usage, PARSE_OPT_STOP_AT_NON_OPTION); diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-trace.c index d9abb4ae5f7..fb3f3c22021 100644 --- a/tools/perf/builtin-trace.c +++ b/tools/perf/builtin-trace.c @@ -12,11 +12,9 @@ #include "util/debug.h" #include "util/trace-event.h" +#include "util/data_map.h" static char const *input_name = "perf.data"; -static int input; -static unsigned long page_size; -static unsigned long mmap_window = 32; static unsigned long total = 0; static unsigned long total_comm = 0; @@ -27,6 +25,9 @@ static struct thread *last_match; static struct perf_header *header; static u64 sample_type; +static char *cwd; +static int cwdlen; + static int process_comm_event(event_t *event, unsigned long offset, unsigned long head) @@ -112,125 +113,32 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) return 0; } -static int -process_event(event_t *event, unsigned long offset, unsigned long head) +static int sample_type_check(u64 type) { - trace_event(event); - - switch (event->header.type) { - case PERF_RECORD_MMAP ... PERF_RECORD_LOST: - return 0; - - case PERF_RECORD_COMM: - return process_comm_event(event, offset, head); - - case PERF_RECORD_EXIT ... PERF_RECORD_READ: - return 0; + sample_type = type; - case PERF_RECORD_SAMPLE: - return process_sample_event(event, offset, head); - - case PERF_RECORD_MAX: - default: + if (!(sample_type & PERF_SAMPLE_RAW)) { + fprintf(stderr, + "No trace sample to read. Did you call perf record " + "without -R?"); return -1; } return 0; } +static struct perf_file_handler file_handler = { + .process_sample_event = process_sample_event, + .process_comm_event = process_comm_event, + .sample_type_check = sample_type_check, +}; + static int __cmd_trace(void) { - int ret, rc = EXIT_FAILURE; - unsigned long offset = 0; - unsigned long head = 0; - unsigned long shift; - struct stat perf_stat; - event_t *event; - uint32_t size; - char *buf; - register_idle_thread(&threads, &last_match); + register_perf_file_handler(&file_handler); - input = open(input_name, O_RDONLY); - if (input < 0) { - perror("failed to open file"); - exit(-1); - } - - ret = fstat(input, &perf_stat); - if (ret < 0) { - perror("failed to stat file"); - exit(-1); - } - - if (!perf_stat.st_size) { - fprintf(stderr, "zero-sized file, nothing to do!\n"); - exit(0); - } - header = perf_header__read(input); - head = header->data_offset; - sample_type = perf_header__sample_type(header); - - if (!(sample_type & PERF_SAMPLE_RAW)) - die("No trace sample to read. Did you call perf record " - "without -R?"); - - if (load_kernel() < 0) { - perror("failed to load kernel symbols"); - return EXIT_FAILURE; - } - - shift = page_size * (head / page_size); - offset += shift; - head -= shift; - -remap: - buf = (char *)mmap(NULL, page_size * mmap_window, PROT_READ, - MAP_SHARED, input, offset); - if (buf == MAP_FAILED) { - perror("failed to mmap file"); - exit(-1); - } - -more: - event = (event_t *)(buf + head); - - if (head + event->header.size >= page_size * mmap_window) { - int res; - - shift = page_size * (head / page_size); - res = munmap(buf, page_size * mmap_window); - assert(res == 0); - - offset += shift; - head -= shift; - goto remap; - } - - size = event->header.size; - - if (!size || process_event(event, offset, head) < 0) { - - /* - * assume we lost track of the stream, check alignment, and - * increment a single u64 in the hope to catch on again 'soon'. - */ - - if (unlikely(head & 7)) - head &= ~7ULL; - - size = 8; - } - - head += size; - - if (offset + head < (unsigned long)perf_stat.st_size) - goto more; - - rc = EXIT_SUCCESS; - close(input); - - return rc; + return mmap_dispatch_perf_file(&header, input_name, 0, 0, &cwdlen, &cwd); } static const char * const annotate_usage[] = { @@ -249,7 +157,6 @@ static const struct option options[] = { int cmd_trace(int argc, const char **argv, const char *prefix __used) { symbol__init(); - page_size = getpagesize(); argc = parse_options(argc, argv, options, annotate_usage, 0); if (argc) { diff --git a/tools/perf/util/data_map.c b/tools/perf/util/data_map.c new file mode 100644 index 00000000000..242b0555ab9 --- /dev/null +++ b/tools/perf/util/data_map.c @@ -0,0 +1,222 @@ +#include "data_map.h" +#include "symbol.h" +#include "util.h" +#include "debug.h" + + +static struct perf_file_handler *curr_handler; +static unsigned long mmap_window = 32; +static char __cwd[PATH_MAX]; + +static int +process_event_stub(event_t *event __used, + unsigned long offset __used, + unsigned long head __used) +{ + return 0; +} + +void register_perf_file_handler(struct perf_file_handler *handler) +{ + if (!handler->process_sample_event) + handler->process_sample_event = process_event_stub; + if (!handler->process_mmap_event) + handler->process_mmap_event = process_event_stub; + if (!handler->process_comm_event) + handler->process_comm_event = process_event_stub; + if (!handler->process_fork_event) + handler->process_fork_event = process_event_stub; + if (!handler->process_exit_event) + handler->process_exit_event = process_event_stub; + if (!handler->process_lost_event) + handler->process_lost_event = process_event_stub; + if (!handler->process_read_event) + handler->process_read_event = process_event_stub; + if (!handler->process_throttle_event) + handler->process_throttle_event = process_event_stub; + if (!handler->process_unthrottle_event) + handler->process_unthrottle_event = process_event_stub; + + curr_handler = handler; +} + +static int +process_event(event_t *event, unsigned long offset, unsigned long head) +{ + trace_event(event); + + switch (event->header.type) { + case PERF_RECORD_SAMPLE: + return curr_handler->process_sample_event(event, offset, head); + case PERF_RECORD_MMAP: + return curr_handler->process_mmap_event(event, offset, head); + case PERF_RECORD_COMM: + return curr_handler->process_comm_event(event, offset, head); + case PERF_RECORD_FORK: + return curr_handler->process_fork_event(event, offset, head); + case PERF_RECORD_EXIT: + return curr_handler->process_exit_event(event, offset, head); + case PERF_RECORD_LOST: + return curr_handler->process_lost_event(event, offset, head); + case PERF_RECORD_READ: + return curr_handler->process_read_event(event, offset, head); + case PERF_RECORD_THROTTLE: + return curr_handler->process_throttle_event(event, offset, head); + case PERF_RECORD_UNTHROTTLE: + return curr_handler->process_unthrottle_event(event, offset, head); + default: + curr_handler->total_unknown++; + return -1; + } +} + +int mmap_dispatch_perf_file(struct perf_header **pheader, + const char *input_name, + int force, + int full_paths, + int *cwdlen, + char **cwd) +{ + int ret, rc = EXIT_FAILURE; + struct perf_header *header; + unsigned long head, shift; + unsigned long offset = 0; + struct stat input_stat; + size_t page_size; + u64 sample_type; + event_t *event; + uint32_t size; + int input; + char *buf; + + if (!curr_handler) + die("Forgot to register perf file handler"); + + page_size = getpagesize(); + + input = open(input_name, O_RDONLY); + if (input < 0) { + fprintf(stderr, " failed to open file: %s", input_name); + if (!strcmp(input_name, "perf.data")) + fprintf(stderr, " (try 'perf record' first)"); + fprintf(stderr, "\n"); + exit(-1); + } + + ret = fstat(input, &input_stat); + if (ret < 0) { + perror("failed to stat file"); + exit(-1); + } + + if (!force && input_stat.st_uid && (input_stat.st_uid != geteuid())) { + fprintf(stderr, "file: %s not owned by current user or root\n", + input_name); + exit(-1); + } + + if (!input_stat.st_size) { + fprintf(stderr, "zero-sized file, nothing to do!\n"); + exit(0); + } + + *pheader = perf_header__read(input); + header = *pheader; + head = header->data_offset; + + sample_type = perf_header__sample_type(header); + + if (curr_handler->sample_type_check) + if (curr_handler->sample_type_check(sample_type) < 0) + exit(-1); + + if (load_kernel() < 0) { + perror("failed to load kernel symbols"); + return EXIT_FAILURE; + } + + if (!full_paths) { + if (getcwd(__cwd, sizeof(__cwd)) == NULL) { + perror("failed to get the current directory"); + return EXIT_FAILURE; + } + *cwd = __cwd; + *cwdlen = strlen(*cwd); + } else { + *cwd = NULL; + *cwdlen = 0; + } + + shift = page_size * (head / page_size); + offset += shift; + head -= shift; + +remap: + buf = (char *)mmap(NULL, page_size * mmap_window, PROT_READ, + MAP_SHARED, input, offset); + if (buf == MAP_FAILED) { + perror("failed to mmap file"); + exit(-1); + } + +more: + event = (event_t *)(buf + head); + + size = event->header.size; + if (!size) + size = 8; + + if (head + event->header.size >= page_size * mmap_window) { + int munmap_ret; + + shift = page_size * (head / page_size); + + munmap_ret = munmap(buf, page_size * mmap_window); + assert(munmap_ret == 0); + + offset += shift; + head -= shift; + goto remap; + } + + size = event->header.size; + + dump_printf("\n%p [%p]: event: %d\n", + (void *)(offset + head), + (void *)(long)event->header.size, + event->header.type); + + if (!size || process_event(event, offset, head) < 0) { + + dump_printf("%p [%p]: skipping unknown header type: %d\n", + (void *)(offset + head), + (void *)(long)(event->header.size), + event->header.type); + + /* + * assume we lost track of the stream, check alignment, and + * increment a single u64 in the hope to catch on again 'soon'. + */ + + if (unlikely(head & 7)) + head &= ~7ULL; + + size = 8; + } + + head += size; + + if (offset + head >= header->data_offset + header->data_size) + goto done; + + if (offset + head < (unsigned long)input_stat.st_size) + goto more; + +done: + rc = EXIT_SUCCESS; + close(input); + + return rc; +} + + diff --git a/tools/perf/util/data_map.h b/tools/perf/util/data_map.h new file mode 100644 index 00000000000..716d1053b07 --- /dev/null +++ b/tools/perf/util/data_map.h @@ -0,0 +1,31 @@ +#ifndef __PERF_DATAMAP_H +#define __PERF_DATAMAP_H + +#include "event.h" +#include "header.h" + +typedef int (*event_type_handler_t)(event_t *, unsigned long, unsigned long); + +struct perf_file_handler { + event_type_handler_t process_sample_event; + event_type_handler_t process_mmap_event; + event_type_handler_t process_comm_event; + event_type_handler_t process_fork_event; + event_type_handler_t process_exit_event; + event_type_handler_t process_lost_event; + event_type_handler_t process_read_event; + event_type_handler_t process_throttle_event; + event_type_handler_t process_unthrottle_event; + int (*sample_type_check)(u64 sample_type); + unsigned long total_unknown; +}; + +void register_perf_file_handler(struct perf_file_handler *handler); +int mmap_dispatch_perf_file(struct perf_header **pheader, + const char *input_name, + int force, + int full_paths, + int *cwdlen, + char **cwd); + +#endif -- cgit v1.2.3-18-g5258 From 9a92b479b2f088ee2d3194243f4c8e59b1b8c9c2 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Thu, 8 Oct 2009 16:37:12 +0200 Subject: perf tools: Improve thread comm resolution in perf sched When we get sched traces that involve a task that was already created before opening the event, we won't have the comm event for it. So if we can't find the comm event for a given thread, we look at the traces that may contain these informations. Before: ata/1:371 | 0.000 ms | 1 | avg: 3988.693 ms | max: 3988.693 ms | kondemand/1:421 | 0.096 ms | 3 | avg: 345.346 ms | max: 1035.989 ms | kondemand/0:420 | 0.025 ms | 3 | avg: 421.332 ms | max: 964.014 ms | :5124:5124 | 0.103 ms | 5 | avg: 74.082 ms | max: 277.194 ms | :6244:6244 | 0.691 ms | 9 | avg: 125.655 ms | max: 271.306 ms | firefox:5080 | 0.924 ms | 5 | avg: 53.833 ms | max: 257.828 ms | npviewer.bin:6225 | 21.871 ms | 53 | avg: 22.462 ms | max: 220.835 ms | :6245:6245 | 9.631 ms | 21 | avg: 41.864 ms | max: 213.349 ms | After: ata/1:371 | 0.000 ms | 1 | avg: 3988.693 ms | max: 3988.693 ms | kondemand/1:421 | 0.096 ms | 3 | avg: 345.346 ms | max: 1035.989 ms | kondemand/0:420 | 0.025 ms | 3 | avg: 421.332 ms | max: 964.014 ms | firefox:5124 | 0.103 ms | 5 | avg: 74.082 ms | max: 277.194 ms | npviewer.bin:6244 | 0.691 ms | 9 | avg: 125.655 ms | max: 271.306 ms | firefox:5080 | 0.924 ms | 5 | avg: 53.833 ms | max: 257.828 ms | npviewer.bin:6225 | 21.871 ms | 53 | avg: 22.462 ms | max: 220.835 ms | npviewer.bin:6245 | 9.631 ms | 21 | avg: 41.864 ms | max: 213.349 ms | Signed-off-by: Frederic Weisbecker Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Mike Galbraith Cc: Paul Mackerras LKML-Reference: <1255012632-7882-1-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar --- tools/perf/builtin-sched.c | 44 +++++++++++++++++++++++++++++++++++++++----- tools/perf/util/thread.c | 32 +++++++++++++++++++++++++------- tools/perf/util/thread.h | 3 +++ 3 files changed, 67 insertions(+), 12 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-sched.c b/tools/perf/builtin-sched.c index e1df7055ab8..25b91e78433 100644 --- a/tools/perf/builtin-sched.c +++ b/tools/perf/builtin-sched.c @@ -1034,6 +1034,36 @@ add_sched_in_event(struct work_atoms *atoms, u64 timestamp) atoms->nb_atoms++; } +static struct thread * +threads__findnew_from_ctx(u32 pid, struct trace_switch_event *switch_event) +{ + struct thread *th; + + th = threads__findnew_nocomm(pid, &threads, &last_match); + if (th->comm) + return th; + + if (pid == switch_event->prev_pid) + thread__set_comm(th, switch_event->prev_comm); + else + thread__set_comm(th, switch_event->next_comm); + return th; +} + +static struct thread * +threads__findnew_from_wakeup(struct trace_wakeup_event *wakeup_event) +{ + struct thread *th; + + th = threads__findnew_nocomm(wakeup_event->pid, &threads, &last_match); + if (th->comm) + return th; + + thread__set_comm(th, wakeup_event->comm); + + return th; +} + static void latency_switch_event(struct trace_switch_event *switch_event, struct event *event __used, @@ -1059,8 +1089,10 @@ latency_switch_event(struct trace_switch_event *switch_event, die("hm, delta: %Ld < 0 ?\n", delta); - sched_out = threads__findnew(switch_event->prev_pid, &threads, &last_match); - sched_in = threads__findnew(switch_event->next_pid, &threads, &last_match); + sched_out = threads__findnew_from_ctx(switch_event->prev_pid, + switch_event); + sched_in = threads__findnew_from_ctx(switch_event->next_pid, + switch_event); out_events = thread_atoms_search(&atom_root, sched_out, &cmp_pid); if (!out_events) { @@ -1126,7 +1158,7 @@ latency_wakeup_event(struct trace_wakeup_event *wakeup_event, if (!wakeup_event->success) return; - wakee = threads__findnew(wakeup_event->pid, &threads, &last_match); + wakee = threads__findnew_from_wakeup(wakeup_event); atoms = thread_atoms_search(&atom_root, wakee, &cmp_pid); if (!atoms) { thread_atoms_insert(wakee); @@ -1386,8 +1418,10 @@ map_switch_event(struct trace_switch_event *switch_event, die("hm, delta: %Ld < 0 ?\n", delta); - sched_out = threads__findnew(switch_event->prev_pid, &threads, &last_match); - sched_in = threads__findnew(switch_event->next_pid, &threads, &last_match); + sched_out = threads__findnew_from_ctx(switch_event->prev_pid, + switch_event); + sched_in = threads__findnew_from_ctx(switch_event->next_pid, + switch_event); curr_thread[this_cpu] = sched_in; diff --git a/tools/perf/util/thread.c b/tools/perf/util/thread.c index 3b56aebb1f4..8bd5ca2d2f2 100644 --- a/tools/perf/util/thread.c +++ b/tools/perf/util/thread.c @@ -6,15 +6,17 @@ #include "util.h" #include "debug.h" -static struct thread *thread__new(pid_t pid) +static struct thread *thread__new(pid_t pid, int set_comm) { struct thread *self = calloc(1, sizeof(*self)); if (self != NULL) { self->pid = pid; - self->comm = malloc(32); - if (self->comm) - snprintf(self->comm, 32, ":%d", self->pid); + if (set_comm) { + self->comm = malloc(32); + if (self->comm) + snprintf(self->comm, 32, ":%d", self->pid); + } self->maps = RB_ROOT; INIT_LIST_HEAD(&self->removed_maps); } @@ -50,8 +52,10 @@ static size_t thread__fprintf(struct thread *self, FILE *fp) return ret; } -struct thread * -threads__findnew(pid_t pid, struct rb_root *threads, struct thread **last_match) +static struct thread * +__threads__findnew(pid_t pid, struct rb_root *threads, + struct thread **last_match, + int set_comm) { struct rb_node **p = &threads->rb_node; struct rb_node *parent = NULL; @@ -80,7 +84,8 @@ threads__findnew(pid_t pid, struct rb_root *threads, struct thread **last_match) p = &(*p)->rb_right; } - th = thread__new(pid); + th = thread__new(pid, set_comm); + if (th != NULL) { rb_link_node(&th->rb_node, parent, p); rb_insert_color(&th->rb_node, threads); @@ -90,6 +95,19 @@ threads__findnew(pid_t pid, struct rb_root *threads, struct thread **last_match) return th; } +struct thread * +threads__findnew(pid_t pid, struct rb_root *threads, struct thread **last_match) +{ + return __threads__findnew(pid, threads, last_match, 1); +} + +struct thread * +threads__findnew_nocomm(pid_t pid, struct rb_root *threads, + struct thread **last_match) +{ + return __threads__findnew(pid, threads, last_match, 0); +} + struct thread * register_idle_thread(struct rb_root *threads, struct thread **last_match) { diff --git a/tools/perf/util/thread.h b/tools/perf/util/thread.h index 845d9b62f96..75bc843950c 100644 --- a/tools/perf/util/thread.h +++ b/tools/perf/util/thread.h @@ -18,6 +18,9 @@ int thread__set_comm(struct thread *self, const char *comm); struct thread * threads__findnew(pid_t pid, struct rb_root *threads, struct thread **last_match); struct thread * +threads__findnew_nocomm(pid_t pid, struct rb_root *threads, + struct thread **last_match); +struct thread * register_idle_thread(struct rb_root *threads, struct thread **last_match); void thread__insert_map(struct thread *self, struct map *map); int thread__fork(struct thread *self, struct thread *parent); -- cgit v1.2.3-18-g5258 From da21d1b547cbaa2c026cf645753651c25d340923 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Wed, 7 Oct 2009 10:49:00 -0300 Subject: perf tools: Up the verbose level for some really verbose stuff Like printing every symbol created. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: Peter Zijlstra Cc: Paul Mackerras Cc: Mike Galbraith LKML-Reference: <1254923340-4870-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar --- tools/perf/builtin-annotate.c | 4 ++-- tools/perf/builtin-report.c | 4 ++-- tools/perf/util/symbol.c | 4 ++-- 3 files changed, 6 insertions(+), 6 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c index 35ed97bd0c6..8c84320ecb0 100644 --- a/tools/perf/builtin-annotate.c +++ b/tools/perf/builtin-annotate.c @@ -658,10 +658,10 @@ more: if (dump_trace) return 0; - if (verbose >= 3) + if (verbose > 3) threads__fprintf(stdout, &threads); - if (verbose >= 2) + if (verbose > 2) dsos__fprintf(stdout); collapse__resort(); diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index 87c4582303b..f57a23b19f3 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -880,10 +880,10 @@ static int __cmd_report(void) if (dump_trace) return 0; - if (verbose >= 3) + if (verbose > 3) threads__fprintf(stdout, &threads); - if (verbose >= 2) + if (verbose > 2) dsos__fprintf(stdout); collapse__resort(); diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 582ce72ca4d..a6887f94dfe 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -74,7 +74,7 @@ static struct symbol *symbol__new(u64 start, u64 len, const char *name, if (!self) return NULL; - if (v >= 2) + if (v > 2) printf("new symbol: %016Lx [%08lx]: %s, hist: %p\n", start, (unsigned long)len, name, self->hist); @@ -685,7 +685,7 @@ static int dso__load_sym(struct dso *self, struct map *map, const char *name, } if (self->adjust_symbols) { - if (v >= 2) + if (v > 2) printf("adjusting symbol: st_value: %Lx sh_addr: %Lx sh_offset: %Lx\n", (u64)sym.st_value, (u64)shdr.sh_addr, (u64)shdr.sh_offset); -- cgit v1.2.3-18-g5258 From 2e538c4a1847291cf01218d4fe7bb4dc60fef7cf Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Wed, 7 Oct 2009 13:48:56 -0300 Subject: perf tools: Improve kernel/modules symbol lookup This removes the ovelapping of vmlinux addresses with modules, using the ELF section name when using --vmlinux and creating a unique DSO name when using /proc/kallsyms ([kernel].N). This is done by creating multiple 'struct map' instances for address ranges backed by DSOs that have just the symbols for that range and a name that is derived from the ELF section name.o Now it is possible to ask for just the symbols in some particular kernel section: $ perf report -m --vmlinux ../build/tip-recvmmsg/vmlinux \ --dsos [kernel].vsyscall_fn | head -15 52.73% Xorg [.] vread_hpet 18.61% firefox [.] vread_hpet 14.50% npviewer.bin [.] vread_hpet 6.83% compiz [.] vread_hpet 5.73% glxgears [.] vread_hpet 0.63% java [.] vread_hpet 0.30% gnome-terminal [.] vread_hpet 0.23% perf [.] vread_hpet 0.18% xchat [.] vread_hpet $ Now we don't have to first lookup the list of modules and then, if it fails, vmlinux symbols, its just a simple lookup for the map then the symbols, just like for threads. Reports generated using /proc/kallsyms and --vmlinux should provide the same results, modulo the DSO name for sections other than ".text". But they don't right now because things like: ffffffff81011c20-ffffffff81012068 system_call ffffffff81011c30-ffffffff81011c9b system_call_after_swapgs ffffffff81011c9c-ffffffff81011cb6 system_call_fastpath ffffffff81011cb7-ffffffff81011cbb ret_from_sys_call I.e. overlapping symbols, again some ASM special case that we have to fixup. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: Peter Zijlstra Cc: Paul Mackerras Cc: Mike Galbraith LKML-Reference: <1254934136-8503-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar --- tools/perf/util/symbol.c | 288 +++++++++++++++++++++++++++++++++-------------- 1 file changed, 201 insertions(+), 87 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index a6887f94dfe..faa84f5d4f5 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -26,27 +26,35 @@ enum dso_origin { static void dsos__add(struct dso *dso); static struct dso *dsos__find(const char *name); +static struct map *map__new2(u64 start, struct dso *dso); +static void kernel_maps__insert(struct map *map); static struct rb_root kernel_maps; -static void dso__set_symbols_end(struct dso *self) +static void dso__fixup_sym_end(struct dso *self) { struct rb_node *nd, *prevnd = rb_first(&self->syms); + struct symbol *curr, *prev; if (prevnd == NULL) return; + curr = rb_entry(prevnd, struct symbol, rb_node); + for (nd = rb_next(prevnd); nd; nd = rb_next(nd)) { - struct symbol *prev = rb_entry(prevnd, struct symbol, rb_node), - *curr = rb_entry(nd, struct symbol, rb_node); + prev = curr; + curr = rb_entry(nd, struct symbol, rb_node); if (prev->end == prev->start) prev->end = curr->start - 1; - prevnd = nd; } + + /* Last entry */ + if (curr->end == curr->start) + curr->end = roundup(curr->start, 4096); } -static void kernel_maps__fixup_sym_end(void) +static void kernel_maps__fixup_end(void) { struct map *prev, *curr; struct rb_node *nd, *prevnd = rb_first(&kernel_maps); @@ -55,13 +63,17 @@ static void kernel_maps__fixup_sym_end(void) return; curr = rb_entry(prevnd, struct map, rb_node); - dso__set_symbols_end(curr->dso); for (nd = rb_next(prevnd); nd; nd = rb_next(nd)) { prev = curr; curr = rb_entry(nd, struct map, rb_node); prev->end = curr->start - 1; - dso__set_symbols_end(curr->dso); + } + + nd = rb_last(&curr->dso->syms); + if (nd) { + struct symbol *sym = rb_entry(nd, struct symbol, rb_node); + curr->end = sym->end; } } @@ -200,13 +212,16 @@ size_t dso__fprintf(struct dso *self, FILE *fp) return ret; } -static int maps__load_kallsyms(symbol_filter_t filter, int use_modules, int v) +/* + * Loads the function entries in /proc/kallsyms into kernel_map->dso, + * so that we can in the next step set the symbol ->end address and then + * call kernel_maps__split_kallsyms. + */ +static int kernel_maps__load_all_kallsyms(int v) { - struct map *map = kernel_map; char *line = NULL; size_t n; FILE *file = fopen("/proc/kallsyms", "r"); - int count = 0; if (file == NULL) goto out_failure; @@ -216,7 +231,7 @@ static int maps__load_kallsyms(symbol_filter_t filter, int use_modules, int v) struct symbol *sym; int line_len, len; char symbol_type; - char *module, *symbol_name; + char *symbol_name; line_len = getline(&line, &n, file); if (line_len < 0) @@ -241,20 +256,55 @@ static int maps__load_kallsyms(symbol_filter_t filter, int use_modules, int v) continue; symbol_name = line + len + 2; - module = strchr(symbol_name, '\t'); - if (module) { - char *module_name_end; + /* + * Will fix up the end later, when we have all symbols sorted. + */ + sym = symbol__new(start, 0, symbol_name, + kernel_map->dso->sym_priv_size, v); + if (sym == NULL) + goto out_delete_line; + + dso__insert_symbol(kernel_map->dso, sym); + } + + free(line); + fclose(file); + + return 0; + +out_delete_line: + free(line); +out_failure: + return -1; +} + +/* + * Split the symbols into maps, making sure there are no overlaps, i.e. the + * kernel range is broken in several maps, named [kernel].N, as we don't have + * the original ELF section names vmlinux have. + */ +static int kernel_maps__split_kallsyms(symbol_filter_t filter, int use_modules) +{ + struct map *map = kernel_map; + struct symbol *pos; + int count = 0; + struct rb_node *next = rb_first(&kernel_map->dso->syms); + int kernel_range = 0; + + while (next) { + char *module; + + pos = rb_entry(next, struct symbol, rb_node); + next = rb_next(&pos->rb_node); + + module = strchr(pos->name, '\t'); + if (module) { if (!use_modules) - continue; - *module = '\0'; - module = strchr(module + 1, '['); - if (!module) - continue; - module_name_end = strchr(module + 1, ']'); - if (!module_name_end) - continue; - *(module_name_end + 1) = '\0'; + goto delete_symbol; + + *module++ = '\0'; + if (strcmp(map->dso->name, module)) { map = kernel_maps__find_by_dso_name(module); if (!map) { @@ -263,50 +313,77 @@ static int maps__load_kallsyms(symbol_filter_t filter, int use_modules, int v) return -1; } } - start = map->map_ip(map, start); - } else - map = kernel_map; - /* - * Well fix up the end later, when we have all sorted. - */ - sym = symbol__new(start, 0, symbol_name, - map->dso->sym_priv_size, v); + /* + * So that we look just like we get from .ko files, + * i.e. not prelinked, relative to map->start. + */ + pos->start = map->map_ip(map, pos->start); + pos->end = map->map_ip(map, pos->end); + } else if (map != kernel_map) { + char dso_name[PATH_MAX]; + struct dso *dso; + + snprintf(dso_name, sizeof(dso_name), "[kernel].%d", + kernel_range++); + + dso = dso__new(dso_name, + kernel_map->dso->sym_priv_size); + if (dso == NULL) + return -1; + + map = map__new2(pos->start, dso); + if (map == NULL) { + dso__delete(dso); + return -1; + } - if (sym == NULL) - goto out_delete_line; + map->map_ip = vdso__map_ip; + kernel_maps__insert(map); + ++kernel_range; + } - if (filter && filter(map, sym)) - symbol__delete(sym, map->dso->sym_priv_size); - else { - dso__insert_symbol(map->dso, sym); + if (filter && filter(map, pos)) { +delete_symbol: + rb_erase(&pos->rb_node, &kernel_map->dso->syms); + symbol__delete(pos, kernel_map->dso->sym_priv_size); + } else { + if (map != kernel_map) { + rb_erase(&pos->rb_node, &kernel_map->dso->syms); + dso__insert_symbol(map->dso, pos); + } count++; } } - free(line); - fclose(file); - return count; +} -out_delete_line: - free(line); -out_failure: - return -1; + +static int kernel_maps__load_kallsyms(symbol_filter_t filter, + int use_modules, int v) +{ + if (kernel_maps__load_all_kallsyms(v)) + return -1; + + dso__fixup_sym_end(kernel_map->dso); + + return kernel_maps__split_kallsyms(filter, use_modules); } -static size_t kernel_maps__fprintf(FILE *fp) +static size_t kernel_maps__fprintf(FILE *fp, int v) { size_t printed = fprintf(stderr, "Kernel maps:\n"); struct rb_node *nd; - printed += map__fprintf(kernel_map, fp); - printed += dso__fprintf(kernel_map->dso, fp); - for (nd = rb_first(&kernel_maps); nd; nd = rb_next(nd)) { struct map *pos = rb_entry(nd, struct map, rb_node); + printed += fprintf(fp, "Map:"); printed += map__fprintf(pos, fp); - printed += dso__fprintf(pos->dso, fp); + if (v > 1) { + printed += dso__fprintf(pos->dso, fp); + printed += fprintf(fp, "--\n"); + } } return printed + fprintf(stderr, "END kernel maps\n"); @@ -594,6 +671,9 @@ static int dso__load_sym(struct dso *self, struct map *map, const char *name, int fd, symbol_filter_t filter, int kernel, int kmodule, int v) { + struct map *curr_map = map; + struct dso *curr_dso = self; + size_t dso_name_len = strlen(self->short_name); Elf_Data *symstrs, *secstrs; uint32_t nr_syms; int err = -1; @@ -660,10 +740,9 @@ static int dso__load_sym(struct dso *self, struct map *map, const char *name, elf_symtab__for_each_symbol(syms, nr_syms, idx, sym) { struct symbol *f; const char *elf_name; - char *demangled; + char *demangled = NULL; int is_label = elf_sym__is_label(&sym); const char *section_name; - u64 sh_offset = 0; if (!is_label && !elf_sym__is_function(&sym)) continue; @@ -677,14 +756,51 @@ static int dso__load_sym(struct dso *self, struct map *map, const char *name, if (is_label && !elf_sec__is_text(&shdr, secstrs)) continue; + elf_name = elf_sym__name(&sym, symstrs); section_name = elf_sec__name(&shdr, secstrs); - if ((kernel || kmodule)) { - if (strstr(section_name, ".init")) - sh_offset = shdr.sh_offset; + if (kernel || kmodule) { + char dso_name[PATH_MAX]; + + if (strcmp(section_name, + curr_dso->short_name + dso_name_len) == 0) + goto new_symbol; + + if (strcmp(section_name, ".text") == 0) { + curr_map = map; + curr_dso = self; + goto new_symbol; + } + + snprintf(dso_name, sizeof(dso_name), + "%s%s", self->short_name, section_name); + + curr_map = kernel_maps__find_by_dso_name(dso_name); + if (curr_map == NULL) { + u64 start = sym.st_value; + + if (kmodule) + start += map->start + shdr.sh_offset; + + curr_dso = dso__new(dso_name, self->sym_priv_size); + if (curr_dso == NULL) + goto out_elf_end; + curr_map = map__new2(start, curr_dso); + if (curr_map == NULL) { + dso__delete(curr_dso); + goto out_elf_end; + } + curr_map->map_ip = vdso__map_ip; + curr_dso->origin = DSO__ORIG_KERNEL; + kernel_maps__insert(curr_map); + dsos__add(curr_dso); + } else + curr_dso = curr_map->dso; + + goto new_symbol; } - if (self->adjust_symbols) { + if (curr_dso->adjust_symbols) { if (v > 2) printf("adjusting symbol: st_value: %Lx sh_addr: %Lx sh_offset: %Lx\n", (u64)sym.st_value, (u64)shdr.sh_addr, (u64)shdr.sh_offset); @@ -696,25 +812,29 @@ static int dso__load_sym(struct dso *self, struct map *map, const char *name, * DWARF DW_compile_unit has this, but we don't always have access * to it... */ - elf_name = elf_sym__name(&sym, symstrs); demangled = bfd_demangle(NULL, elf_name, DMGL_PARAMS | DMGL_ANSI); if (demangled != NULL) elf_name = demangled; - - f = symbol__new(sym.st_value + sh_offset, sym.st_size, elf_name, - self->sym_priv_size, v); +new_symbol: + f = symbol__new(sym.st_value, sym.st_size, elf_name, + curr_dso->sym_priv_size, v); free(demangled); if (!f) goto out_elf_end; - if (filter && filter(map, f)) - symbol__delete(f, self->sym_priv_size); + if (filter && filter(curr_map, f)) + symbol__delete(f, curr_dso->sym_priv_size); else { - dso__insert_symbol(self, f); + dso__insert_symbol(curr_dso, f); nr++; } } + /* + * For misannotated, zeroed, ASM function sizes. + */ + if (nr > 0) + dso__fixup_sym_end(self); err = nr; out_elf_end: elf_end(elf); @@ -883,27 +1003,17 @@ static void kernel_maps__insert(struct map *map) struct symbol *kernel_maps__find_symbol(u64 ip, struct map **mapp) { - /* - * We can't have kernel_map in kernel_maps because it spans an address - * space that includes the modules. The right way to fix this is to - * create several maps, so that we don't have overlapping ranges with - * modules. For now lets look first on the kernel dso. - */ struct map *map = maps__find(&kernel_maps, ip); - struct symbol *sym; + + if (mapp) + *mapp = map; if (map) { ip = map->map_ip(map, ip); - sym = map->dso->find_symbol(map->dso, ip); - } else { - map = kernel_map; - sym = map->dso->find_symbol(map->dso, ip); + return map->dso->find_symbol(map->dso, ip); } - if (mapp) - *mapp = map; - - return sym; + return NULL; } struct map *kernel_maps__find_by_dso_name(const char *name) @@ -994,6 +1104,14 @@ static int dsos__load_modules_sym_dir(char *dirname, last = rb_last(&map->dso->syms); if (last) { struct symbol *sym; + /* + * We do this here as well, even having the + * symbol size found in the symtab because + * misannotated ASM symbols may have the size + * set to zero. + */ + dso__fixup_sym_end(map->dso); + sym = rb_entry(last, struct symbol, rb_node); map->end = map->start + sym->end; } @@ -1163,17 +1281,11 @@ int dsos__load_kernel(const char *vmlinux, unsigned int sym_priv_size, } if (err <= 0) - err = maps__load_kallsyms(filter, use_modules, v); + err = kernel_maps__load_kallsyms(filter, use_modules, v); if (err > 0) { struct rb_node *node = rb_first(&dso->syms); struct symbol *sym = rb_entry(node, struct symbol, rb_node); - /* - * Now that we have all sorted out, just set the ->end of all - * symbols that still don't have it. - */ - dso__set_symbols_end(dso); - kernel_maps__fixup_sym_end(); kernel_map->start = sym->start; node = rb_last(&dso->syms); @@ -1181,14 +1293,16 @@ int dsos__load_kernel(const char *vmlinux, unsigned int sym_priv_size, kernel_map->end = sym->end; dso->origin = DSO__ORIG_KERNEL; + kernel_maps__insert(kernel_map); /* - * XXX See kernel_maps__find_symbol comment - * kernel_maps__insert(kernel_map) + * Now that we have all sorted out, just set the ->end of all + * maps: */ + kernel_maps__fixup_end(); dsos__add(dso); if (v > 0) - kernel_maps__fprintf(stderr); + kernel_maps__fprintf(stderr, v); } return err; -- cgit v1.2.3-18-g5258 From 97ea1a7fa62af0d8d49a0fc12796b0073537c9d8 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Thu, 8 Oct 2009 21:04:17 +0200 Subject: perf tools: Fix thread comm resolution in perf sched This reverts commit 9a92b479b2f088ee2d3194243f4c8e59b1b8c9c2 ("perf tools: Improve thread comm resolution in perf sched") and fixes the real bug. The bug was elsewhere: We are failing to resolve thread names in perf sched because the table of threads we are building, on top of comm events, has a per process granularity. But perf sched, unlike the other perf tools, needs a per thread granularity as we are profiling every tasks individually. So fix it by building our threads table using the tid instead of the pid as the thread identifier. v2: Revert the previous fix - it is not really needed Signed-off-by: Frederic Weisbecker Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Mike Galbraith Cc: Paul Mackerras LKML-Reference: <1255028657-11158-1-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar --- tools/perf/builtin-sched.c | 46 ++++++---------------------------------------- tools/perf/util/thread.c | 32 +++++++------------------------- tools/perf/util/thread.h | 3 --- 3 files changed, 13 insertions(+), 68 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-sched.c b/tools/perf/builtin-sched.c index 25b91e78433..6b00529ce34 100644 --- a/tools/perf/builtin-sched.c +++ b/tools/perf/builtin-sched.c @@ -638,7 +638,7 @@ process_comm_event(event_t *event, unsigned long offset, unsigned long head) { struct thread *thread; - thread = threads__findnew(event->comm.pid, &threads, &last_match); + thread = threads__findnew(event->comm.tid, &threads, &last_match); dump_printf("%p [%p]: perf_event_comm: %s:%d\n", (void *)(offset + head), @@ -1034,36 +1034,6 @@ add_sched_in_event(struct work_atoms *atoms, u64 timestamp) atoms->nb_atoms++; } -static struct thread * -threads__findnew_from_ctx(u32 pid, struct trace_switch_event *switch_event) -{ - struct thread *th; - - th = threads__findnew_nocomm(pid, &threads, &last_match); - if (th->comm) - return th; - - if (pid == switch_event->prev_pid) - thread__set_comm(th, switch_event->prev_comm); - else - thread__set_comm(th, switch_event->next_comm); - return th; -} - -static struct thread * -threads__findnew_from_wakeup(struct trace_wakeup_event *wakeup_event) -{ - struct thread *th; - - th = threads__findnew_nocomm(wakeup_event->pid, &threads, &last_match); - if (th->comm) - return th; - - thread__set_comm(th, wakeup_event->comm); - - return th; -} - static void latency_switch_event(struct trace_switch_event *switch_event, struct event *event __used, @@ -1089,10 +1059,8 @@ latency_switch_event(struct trace_switch_event *switch_event, die("hm, delta: %Ld < 0 ?\n", delta); - sched_out = threads__findnew_from_ctx(switch_event->prev_pid, - switch_event); - sched_in = threads__findnew_from_ctx(switch_event->next_pid, - switch_event); + sched_out = threads__findnew(switch_event->prev_pid, &threads, &last_match); + sched_in = threads__findnew(switch_event->next_pid, &threads, &last_match); out_events = thread_atoms_search(&atom_root, sched_out, &cmp_pid); if (!out_events) { @@ -1158,7 +1126,7 @@ latency_wakeup_event(struct trace_wakeup_event *wakeup_event, if (!wakeup_event->success) return; - wakee = threads__findnew_from_wakeup(wakeup_event); + wakee = threads__findnew(wakeup_event->pid, &threads, &last_match); atoms = thread_atoms_search(&atom_root, wakee, &cmp_pid); if (!atoms) { thread_atoms_insert(wakee); @@ -1418,10 +1386,8 @@ map_switch_event(struct trace_switch_event *switch_event, die("hm, delta: %Ld < 0 ?\n", delta); - sched_out = threads__findnew_from_ctx(switch_event->prev_pid, - switch_event); - sched_in = threads__findnew_from_ctx(switch_event->next_pid, - switch_event); + sched_out = threads__findnew(switch_event->prev_pid, &threads, &last_match); + sched_in = threads__findnew(switch_event->next_pid, &threads, &last_match); curr_thread[this_cpu] = sched_in; diff --git a/tools/perf/util/thread.c b/tools/perf/util/thread.c index 8bd5ca2d2f2..3b56aebb1f4 100644 --- a/tools/perf/util/thread.c +++ b/tools/perf/util/thread.c @@ -6,17 +6,15 @@ #include "util.h" #include "debug.h" -static struct thread *thread__new(pid_t pid, int set_comm) +static struct thread *thread__new(pid_t pid) { struct thread *self = calloc(1, sizeof(*self)); if (self != NULL) { self->pid = pid; - if (set_comm) { - self->comm = malloc(32); - if (self->comm) - snprintf(self->comm, 32, ":%d", self->pid); - } + self->comm = malloc(32); + if (self->comm) + snprintf(self->comm, 32, ":%d", self->pid); self->maps = RB_ROOT; INIT_LIST_HEAD(&self->removed_maps); } @@ -52,10 +50,8 @@ static size_t thread__fprintf(struct thread *self, FILE *fp) return ret; } -static struct thread * -__threads__findnew(pid_t pid, struct rb_root *threads, - struct thread **last_match, - int set_comm) +struct thread * +threads__findnew(pid_t pid, struct rb_root *threads, struct thread **last_match) { struct rb_node **p = &threads->rb_node; struct rb_node *parent = NULL; @@ -84,8 +80,7 @@ __threads__findnew(pid_t pid, struct rb_root *threads, p = &(*p)->rb_right; } - th = thread__new(pid, set_comm); - + th = thread__new(pid); if (th != NULL) { rb_link_node(&th->rb_node, parent, p); rb_insert_color(&th->rb_node, threads); @@ -95,19 +90,6 @@ __threads__findnew(pid_t pid, struct rb_root *threads, return th; } -struct thread * -threads__findnew(pid_t pid, struct rb_root *threads, struct thread **last_match) -{ - return __threads__findnew(pid, threads, last_match, 1); -} - -struct thread * -threads__findnew_nocomm(pid_t pid, struct rb_root *threads, - struct thread **last_match) -{ - return __threads__findnew(pid, threads, last_match, 0); -} - struct thread * register_idle_thread(struct rb_root *threads, struct thread **last_match) { diff --git a/tools/perf/util/thread.h b/tools/perf/util/thread.h index 75bc843950c..845d9b62f96 100644 --- a/tools/perf/util/thread.h +++ b/tools/perf/util/thread.h @@ -18,9 +18,6 @@ int thread__set_comm(struct thread *self, const char *comm); struct thread * threads__findnew(pid_t pid, struct rb_root *threads, struct thread **last_match); struct thread * -threads__findnew_nocomm(pid_t pid, struct rb_root *threads, - struct thread **last_match); -struct thread * register_idle_thread(struct rb_root *threads, struct thread **last_match); void thread__insert_map(struct thread *self, struct map *map); int thread__fork(struct thread *self, struct thread *parent); -- cgit v1.2.3-18-g5258 From 26dd2cb074d9dc41c9e3cddd7bf175fd0a41febc Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Thu, 8 Oct 2009 22:07:29 +0200 Subject: perf tools: Provide backward compatibility with previous perf.data version We have merged the trace.info file into perf.data by adding one section in the perf headers. This makes it incompatible with previous version: the new perf tools can't read the older perf.data. To support the previous format, we check the headers size. If they have the same size than in the previous format, then ignore the trace info section that doesn't exist. Signed-off-by: Frederic Weisbecker Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Mike Galbraith Cc: Paul Mackerras LKML-Reference: <1255032449-12022-1-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar --- tools/perf/util/header.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-) (limited to 'tools/perf') diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index 212fade7ee7..9aae360c0f2 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -287,10 +287,16 @@ struct perf_header *perf_header__read(int fd) do_read(fd, &f_header, sizeof(f_header)); if (f_header.magic != PERF_MAGIC || - f_header.size != sizeof(f_header) || f_header.attr_size != sizeof(f_attr)) die("incompatible file format"); + if (f_header.size != sizeof(f_header)) { + /* Support the previous format */ + if (f_header.size == offsetof(typeof(f_header), trace_info)) + f_header.trace_info.size = 0; + else + die("incompatible file format"); + } nr_attrs = f_header.attrs.size / sizeof(f_attr); lseek(fd, f_header.attrs.offset, SEEK_SET); -- cgit v1.2.3-18-g5258 From 7e4ff9e3e8f88de8a8536f43294cd32b4e7d9123 Mon Sep 17 00:00:00 2001 From: Mike Galbraith Date: Mon, 12 Oct 2009 07:56:03 +0200 Subject: perf tools: Fix counter sample frequency breakage Commit 42e59d7d19dc4b4 switched to a default sample frequency of 1KHz, which overrides any user supplied count, causing sched, top and timechart to miss events due to their discrete events being flagged PERF_SAMPLE_PERIOD. Override default sample frequency when the user profides a period count, and make both record and top honor that user supplied option. Signed-off-by: Mike Galbraith Cc: Peter Zijlstra Cc: Arjan van de Ven LKML-Reference: <1255326963.15107.2.camel@marge.simson.net> Signed-off-by: Ingo Molnar --- tools/perf/builtin-record.c | 14 +++++++++++++- tools/perf/builtin-top.c | 28 +++++++++++++++++++++------- 2 files changed, 34 insertions(+), 8 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 59af03d80d0..4e3a374e7aa 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -26,7 +26,7 @@ static int fd[MAX_NR_CPUS][MAX_COUNTERS]; -static long default_interval = 100000; +static long default_interval = 0; static int nr_cpus = 0; static unsigned int page_size; @@ -730,6 +730,18 @@ int cmd_record(int argc, const char **argv, const char *prefix __used) attrs[0].config = PERF_COUNT_HW_CPU_CYCLES; } + /* + * User specified count overrides default frequency. + */ + if (default_interval) + freq = 0; + else if (freq) { + default_interval = freq; + } else { + fprintf(stderr, "frequency and count are zero, aborting\n"); + exit(EXIT_FAILURE); + } + for (counter = 0; counter < nr_counters; counter++) { if (attrs[counter].sample_period) continue; diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index d978dc99236..c0f69e80b2c 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -57,7 +57,7 @@ static int fd[MAX_NR_CPUS][MAX_COUNTERS]; static int system_wide = 0; -static int default_interval = 100000; +static int default_interval = 0; static int count_filter = 5; static int print_entries = 15; @@ -975,7 +975,13 @@ static void start_counter(int i, int counter) attr = attrs + counter; attr->sample_type = PERF_SAMPLE_IP | PERF_SAMPLE_TID; - attr->freq = freq; + + if (freq) { + attr->sample_type |= PERF_SAMPLE_PERIOD; + attr->freq = 1; + attr->sample_freq = freq; + } + attr->inherit = (cpu < 0) && inherit; try_again: @@ -1130,11 +1136,6 @@ int cmd_top(int argc, const char **argv, const char *prefix __used) if (argc) usage_with_options(top_usage, options); - if (freq) { - default_interval = freq; - freq = 1; - } - /* CPU and PID are mutually exclusive */ if (target_pid != -1 && profile_cpu != -1) { printf("WARNING: PID switch overriding CPU\n"); @@ -1151,6 +1152,19 @@ int cmd_top(int argc, const char **argv, const char *prefix __used) parse_symbols(); parse_source(sym_filter_entry); + + /* + * User specified count overrides default frequency. + */ + if (default_interval) + freq = 0; + else if (freq) { + default_interval = freq; + } else { + fprintf(stderr, "frequency and count are zero, aborting\n"); + exit(EXIT_FAILURE); + } + /* * Fill in the ones not specifically initialized via -c: */ -- cgit v1.2.3-18-g5258 From 55ffb7a6bd45d0083ffb132381cb46964a4afe01 Mon Sep 17 00:00:00 2001 From: Mike Galbraith Date: Sat, 10 Oct 2009 14:46:04 +0200 Subject: perf sched: Add -C option to measure on a specific CPU To refresh, trying to sched record only one CPU results in bogus latencies as below. I fixed^Wmade it stop doing the bad thing today, by following task migration events properly. Before: marge:/root/tmp # taskset -c 1 perf sched record -C 0 -- sleep 10 marge:/root/tmp # perf sched lat ----------------------------------------------------------------------------------------- Task | Runtime ms | Switches | Average delay ms | Maximum delay ms | ----------------------------------------------------------------------------------------- Xorg:4943 | 1.290 ms | 1 | avg: 1670.132 ms | max: 1670.132 ms | hald-addon-stor:3569 | 0.091 ms | 3 | avg: 658.609 ms | max: 1975.797 ms | hald-addon-stor:3573 | 0.209 ms | 4 | avg: 499.138 ms | max: 1990.565 ms | audispd:4270 | 0.012 ms | 1 | avg: 0.015 ms | max: 0.015 ms | .... marge:/root/tmp # perf sched trace|grep 'Xorg:4943' swapper-0 [000] 401.184013288: sched_stat_runtime: task: Xorg:4943 runtime: 1233188 [ns], vruntime: 19105169779 [ns] rt2870TimerQHan-4947 [000] 402.854140127: sched_stat_wait: task: Xorg:4943 wait: 580073 [ns] rt2870TimerQHan-4947 [000] 402.854141770: sched_migrate_task: task Xorg:4943 [140] from: 1 to: 0 rt2870TimerQHan-4947 [000] 402.854143854: sched_stat_wait: task: Xorg:4943 wait: 0 [ns] rt2870TimerQHan-4947 [000] 402.854145397: sched_switch: task rt2870TimerQHan:4947 [140] (D) ==> Xorg:4943 [140] Xorg-4943 [000] 402.854193133: sched_stat_runtime: task: Xorg:4943 runtime: 56546 [ns], vruntime: 11766332500 [ns] Xorg-4943 [000] 402.854196842: sched_switch: task Xorg:4943 [140] (S) ==> swapper:0 [140] After: marge:/root/tmp # taskset -c 1 perf sched record -C 0 -- sleep 10 marge:/root/tmp # perf sched lat ----------------------------------------------------------------------------------------- Task | Runtime ms | Switches | Average delay ms | Maximum delay ms | ----------------------------------------------------------------------------------------- amarokapp:11150 | 271.297 ms | 878 | avg: 0.130 ms | max: 1.057 ms | konsole:5965 | 1.370 ms | 12 | avg: 0.092 ms | max: 0.855 ms | Xorg:4943 | 179.980 ms | 1109 | avg: 0.087 ms | max: 1.206 ms | hald-addon-stor:3574 | 0.212 ms | 9 | avg: 0.040 ms | max: 0.169 ms | hald-addon-stor:3570 | 0.223 ms | 9 | avg: 0.037 ms | max: 0.223 ms | klauncher:5864 | 0.550 ms | 8 | avg: 0.032 ms | max: 0.048 ms | The 'Maximum delay ms' results are now sane. Signed-off-by: Mike Galbraith LKML-Reference: Signed-off-by: Ingo Molnar --- tools/perf/builtin-sched.c | 101 ++++++++++++++++++++++++++++++++++++++++++++- 1 file changed, 100 insertions(+), 1 deletion(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-sched.c b/tools/perf/builtin-sched.c index 6b00529ce34..387a4423436 100644 --- a/tools/perf/builtin-sched.c +++ b/tools/perf/builtin-sched.c @@ -33,6 +33,8 @@ static u64 sample_type; static char default_sort_order[] = "avg, max, switch, runtime"; static char *sort_order = default_sort_order; +static int profile_cpu = -1; + static char *cwd; static int cwdlen; @@ -75,6 +77,7 @@ enum sched_event_type { SCHED_EVENT_RUN, SCHED_EVENT_SLEEP, SCHED_EVENT_WAKEUP, + SCHED_EVENT_MIGRATION, }; struct sched_atom { @@ -399,6 +402,8 @@ process_sched_event(struct task_desc *this_task __used, struct sched_atom *atom) ret = sem_post(atom->wait_sem); BUG_ON(ret); break; + case SCHED_EVENT_MIGRATION: + break; default: BUG_ON(1); } @@ -746,6 +751,22 @@ struct trace_fork_event { u32 child_pid; }; +struct trace_migrate_task_event { + u32 size; + + u16 common_type; + u8 common_flags; + u8 common_preempt_count; + u32 common_pid; + u32 common_tgid; + + char comm[16]; + u32 pid; + + u32 prio; + u32 cpu; +}; + struct trace_sched_handler { void (*switch_event)(struct trace_switch_event *, struct event *, @@ -770,6 +791,12 @@ struct trace_sched_handler { int cpu, u64 timestamp, struct thread *thread); + + void (*migrate_task_event)(struct trace_migrate_task_event *, + struct event *, + int cpu, + u64 timestamp, + struct thread *thread); }; @@ -1140,7 +1167,12 @@ latency_wakeup_event(struct trace_wakeup_event *wakeup_event, atom = list_entry(atoms->work_list.prev, struct work_atom, list); - if (atom->state != THREAD_SLEEPING) + /* + * You WILL be missing events if you've recorded only + * one CPU, or are only looking at only one, so don't + * make useless noise. + */ + if (profile_cpu == -1 && atom->state != THREAD_SLEEPING) nr_state_machine_bugs++; nr_timestamps++; @@ -1153,11 +1185,51 @@ latency_wakeup_event(struct trace_wakeup_event *wakeup_event, atom->wake_up_time = timestamp; } +static void +latency_migrate_task_event(struct trace_migrate_task_event *migrate_task_event, + struct event *__event __used, + int cpu __used, + u64 timestamp, + struct thread *thread __used) +{ + struct work_atoms *atoms; + struct work_atom *atom; + struct thread *migrant; + + /* + * Only need to worry about migration when profiling one CPU. + */ + if (profile_cpu == -1) + return; + + migrant = threads__findnew(migrate_task_event->pid, &threads, &last_match); + atoms = thread_atoms_search(&atom_root, migrant, &cmp_pid); + if (!atoms) { + thread_atoms_insert(migrant); + register_pid(migrant->pid, migrant->comm); + atoms = thread_atoms_search(&atom_root, migrant, &cmp_pid); + if (!atoms) + die("migration-event: Internal tree error"); + add_sched_out_event(atoms, 'R', timestamp); + } + + BUG_ON(list_empty(&atoms->work_list)); + + atom = list_entry(atoms->work_list.prev, struct work_atom, list); + atom->sched_in_time = atom->sched_out_time = atom->wake_up_time = timestamp; + + nr_timestamps++; + + if (atom->sched_out_time > timestamp) + nr_unordered_timestamps++; +} + static struct trace_sched_handler lat_ops = { .wakeup_event = latency_wakeup_event, .switch_event = latency_switch_event, .runtime_event = latency_runtime_event, .fork_event = latency_fork_event, + .migrate_task_event = latency_migrate_task_event, }; static void output_lat_thread(struct work_atoms *work_list) @@ -1517,6 +1589,26 @@ process_sched_exit_event(struct event *event, printf("sched_exit event %p\n", event); } +static void +process_sched_migrate_task_event(struct raw_event_sample *raw, + struct event *event, + int cpu __used, + u64 timestamp __used, + struct thread *thread __used) +{ + struct trace_migrate_task_event migrate_task_event; + + FILL_COMMON_FIELDS(migrate_task_event, event, raw->data); + + FILL_ARRAY(migrate_task_event, comm, event, raw->data); + FILL_FIELD(migrate_task_event, pid, event, raw->data); + FILL_FIELD(migrate_task_event, prio, event, raw->data); + FILL_FIELD(migrate_task_event, cpu, event, raw->data); + + if (trace_handler->migrate_task_event) + trace_handler->migrate_task_event(&migrate_task_event, event, cpu, timestamp, thread); +} + static void process_raw_event(event_t *raw_event __used, void *more_data, int cpu, u64 timestamp, struct thread *thread) @@ -1540,6 +1632,8 @@ process_raw_event(event_t *raw_event __used, void *more_data, process_sched_fork_event(raw, event, cpu, timestamp, thread); if (!strcmp(event->name, "sched_process_exit")) process_sched_exit_event(event, cpu, timestamp, thread); + if (!strcmp(event->name, "sched_migrate_task")) + process_sched_migrate_task_event(raw, event, cpu, timestamp, thread); } static int @@ -1589,6 +1683,9 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) return -1; } + if (profile_cpu != -1 && profile_cpu != (int) cpu) + return 0; + process_raw_event(event, more_data, cpu, timestamp, thread); return 0; @@ -1771,6 +1868,8 @@ static const struct option latency_options[] = { "sort by key(s): runtime, switch, avg, max"), OPT_BOOLEAN('v', "verbose", &verbose, "be more verbose (show symbol address, etc)"), + OPT_INTEGER('C', "CPU", &profile_cpu, + "CPU to profile on"), OPT_BOOLEAN('D', "dump-raw-trace", &dump_trace, "dump raw trace in ASCII"), OPT_END() -- cgit v1.2.3-18-g5258 From 4ea42b181434bfc6a0a18d32214130a242d489bf Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Thu, 8 Oct 2009 17:17:38 -0400 Subject: perf: Add perf probe subcommand, a kprobe-event setup helper Add perf probe subcommand that implements a kprobe-event setup helper to the perf command. This allows user to define kprobe events using C expressions (C line numbers, C function names, and C local variables). Usage ----- perf probe [] -P 'PROBEDEF' [-P 'PROBEDEF' ...] -k, --vmlinux vmlinux/module pathname -P, --probe probe point definition, where p: kprobe probe r: kretprobe probe GRP: Group name (optional) NAME: Event name FUNC: Function name OFFS: Offset from function entry (in byte) SRC: Source code path LINE: Line number ARG: Probe argument (local variable name or kprobe-tracer argument format is supported.) Changes in v4: - Add _GNU_SOURCE macro for strndup(). Changes in v3: - Remove -r option because perf always be used for online kernel. - Check malloc/calloc results. Changes in v2: - Check synthesized string length. - Rename perf kprobe to perf probe. - Use spaces for separator and update usage comment. - Check error paths in parse_probepoint(). - Check optimized-out variables. Signed-off-by: Masami Hiramatsu Cc: Ingo Molnar Cc: Thomas Gleixner Cc: Arnaldo Carvalho de Melo Cc: Steven Rostedt Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Christoph Hellwig Cc: Ananth N Mavinakayanahalli Cc: Jim Keniston Cc: Frank Ch. Eigler LKML-Reference: <20091008211737.29299.14784.stgit@dhcp-100-2-132.bos.redhat.com> Signed-off-by: Frederic Weisbecker --- tools/perf/Makefile | 10 + tools/perf/builtin-probe.c | 358 +++++++++++++++++++++ tools/perf/builtin.h | 1 + tools/perf/perf.c | 3 + tools/perf/util/probe-finder.c | 690 +++++++++++++++++++++++++++++++++++++++++ tools/perf/util/probe-finder.h | 68 ++++ 6 files changed, 1130 insertions(+) create mode 100644 tools/perf/builtin-probe.c create mode 100644 tools/perf/util/probe-finder.c create mode 100644 tools/perf/util/probe-finder.h (limited to 'tools/perf') diff --git a/tools/perf/Makefile b/tools/perf/Makefile index b5f1953b614..6dabcf1a4df 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -419,6 +419,16 @@ ifneq ($(shell sh -c "(echo '\#include '; echo 'int main(void) { Elf * msg := $(error No libelf.h/libelf found, please install libelf-dev/elfutils-libelf-devel); endif +ifneq ($(shell sh -c "(echo '\#include '; echo '\#include '; echo 'int main(void) { Dwarf_Debug dbg; Dwarf_Error err; dwarf_init(0, DW_DLC_READ, 0, 0, &dbg, &err); return (long)dbg; }') | $(CC) -x c - $(ALL_CFLAGS) -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -ldwarf -lelf -o /dev/null $(ALL_LDFLAGS) > /dev/null 2>&1 && echo y"), y) + msg := $(warning No libdwarf.h found, disables probe subcommand. Please install libdwarf-dev/libdwarf-devel); +else + EXTLIBS += -lelf -ldwarf + LIB_H += util/probe-finder.h + LIB_OBJS += util/probe-finder.o + BUILTIN_OBJS += builtin-probe.o + BASIC_CFLAGS += -DSUPPORT_DWARF +endif + ifdef NO_DEMANGLE BASIC_CFLAGS += -DNO_DEMANGLE else diff --git a/tools/perf/builtin-probe.c b/tools/perf/builtin-probe.c new file mode 100644 index 00000000000..24b64b5cefc --- /dev/null +++ b/tools/perf/builtin-probe.c @@ -0,0 +1,358 @@ +/* + * builtin-probe.c + * + * Builtin probe command: Set up probe events by C expression + * + * Written by Masami Hiramatsu + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. + * + */ +#define _GNU_SOURCE +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#undef _GNU_SOURCE +#include "perf.h" +#include "builtin.h" +#include "util/util.h" +#include "util/parse-options.h" +#include "util/parse-events.h" /* For debugfs_path */ +#include "util/probe-finder.h" + +/* Default vmlinux search paths */ +#define NR_SEARCH_PATH 3 +const char *default_search_path[NR_SEARCH_PATH] = { +"/lib/modules/%s/build/vmlinux", /* Custom build kernel */ +"/usr/lib/debug/lib/modules/%s/vmlinux", /* Red Hat debuginfo */ +"/boot/vmlinux-debug-%s", /* Ubuntu */ +}; + +#define MAX_PATH_LEN 256 +#define MAX_PROBES 128 + +/* Session management structure */ +static struct { + char *vmlinux; + char *release; + int nr_probe; + struct probe_point probes[MAX_PROBES]; + char *events[MAX_PROBES]; +} session; + +static void semantic_error(const char *msg) +{ + fprintf(stderr, "Semantic error: %s\n", msg); + exit(1); +} + +static void perror_exit(const char *msg) +{ + perror(msg); + exit(1); +} + +#define MAX_PROBE_ARGS 128 + +static int parse_probepoint(const struct option *opt __used, + const char *str, int unset __used) +{ + char *argv[MAX_PROBE_ARGS + 2]; /* Event + probe + args */ + int argc, i; + char *arg, *ptr; + struct probe_point *pp = &session.probes[session.nr_probe]; + char **event = &session.events[session.nr_probe]; + int retp = 0; + + if (!str) /* The end of probe points */ + return 0; + + debug("Probe-define(%d): %s\n", session.nr_probe, str); + if (++session.nr_probe == MAX_PROBES) + semantic_error("Too many probes"); + + /* Separate arguments, similar to argv_split */ + argc = 0; + do { + /* Skip separators */ + while (isspace(*str)) + str++; + + /* Add an argument */ + if (*str != '\0') { + const char *s = str; + + /* Skip the argument */ + while (!isspace(*str) && *str != '\0') + str++; + + /* Duplicate the argument */ + argv[argc] = strndup(s, str - s); + if (argv[argc] == NULL) + perror_exit("strndup"); + if (++argc == MAX_PROBE_ARGS) + semantic_error("Too many arguments"); + debug("argv[%d]=%s\n", argc, argv[argc - 1]); + } + } while (*str != '\0'); + if (argc < 2) + semantic_error("Need event-name and probe-point at least."); + + /* Parse the event name */ + if (argv[0][0] == 'r') + retp = 1; + else if (argv[0][0] != 'p') + semantic_error("You must specify 'p'(kprobe) or" + " 'r'(kretprobe) first."); + /* TODO: check event name */ + *event = argv[0]; + + /* Parse probe point */ + arg = argv[1]; + if (arg[0] == '@') { + /* Source Line */ + arg++; + ptr = strchr(arg, ':'); + if (!ptr || !isdigit(ptr[1])) + semantic_error("Line number is required."); + *ptr++ = '\0'; + if (strlen(arg) == 0) + semantic_error("No file name."); + pp->file = strdup(arg); + pp->line = atoi(ptr); + if (!pp->file || !pp->line) + semantic_error("Failed to parse line."); + debug("file:%s line:%d\n", pp->file, pp->line); + } else { + /* Function name */ + ptr = strchr(arg, '+'); + if (ptr) { + if (!isdigit(ptr[1])) + semantic_error("Offset is required."); + *ptr++ = '\0'; + pp->offset = atoi(ptr); + } else + ptr = arg; + ptr = strchr(ptr, '@'); + if (ptr) { + *ptr++ = '\0'; + pp->file = strdup(ptr); + } + pp->function = strdup(arg); + debug("symbol:%s file:%s offset:%d\n", + pp->function, pp->file, pp->offset); + } + free(argv[1]); + + /* Copy arguments */ + pp->nr_args = argc - 2; + if (pp->nr_args > 0) { + pp->args = (char **)malloc(sizeof(char *) * pp->nr_args); + if (!pp->args) + perror_exit("malloc"); + memcpy(pp->args, &argv[2], sizeof(char *) * pp->nr_args); + } + + /* Ensure return probe has no C argument */ + if (retp) + for (i = 0; i < pp->nr_args; i++) + if (is_c_varname(pp->args[i])) + semantic_error("You can't specify local" + " variable for kretprobe"); + debug("%d arguments\n", pp->nr_args); + return 0; +} + +static int open_default_vmlinux(void) +{ + struct utsname uts; + char fname[MAX_PATH_LEN]; + int fd, ret, i; + + ret = uname(&uts); + if (ret) { + debug("uname() failed.\n"); + return -errno; + } + session.release = uts.release; + for (i = 0; i < NR_SEARCH_PATH; i++) { + ret = snprintf(fname, MAX_PATH_LEN, + default_search_path[i], session.release); + if (ret >= MAX_PATH_LEN || ret < 0) { + debug("Filename(%d,%s) is too long.\n", i, uts.release); + errno = E2BIG; + return -E2BIG; + } + debug("try to open %s\n", fname); + fd = open(fname, O_RDONLY); + if (fd >= 0) + break; + } + return fd; +} + +static const char * const probe_usage[] = { + "perf probe [] -P 'PROBEDEF' [-P 'PROBEDEF' ...]", + NULL +}; + +static const struct option options[] = { + OPT_STRING('k', "vmlinux", &session.vmlinux, "file", + "vmlinux/module pathname"), + OPT_CALLBACK('P', "probe", NULL, + "p|r:[GRP/]NAME FUNC[+OFFS][@SRC]|@SRC:LINE [ARG ...]", + "probe point definition, where\n" + "\t\tp:\tkprobe probe\n" + "\t\tr:\tkretprobe probe\n" + "\t\tGRP:\tGroup name (optional)\n" + "\t\tNAME:\tEvent name\n" + "\t\tFUNC:\tFunction name\n" + "\t\tOFFS:\tOffset from function entry (in byte)\n" + "\t\tSRC:\tSource code path\n" + "\t\tLINE:\tLine number\n" + "\t\tARG:\tProbe argument (local variable name or\n" + "\t\t\tkprobe-tracer argument format is supported.)\n", + parse_probepoint), + OPT_END() +}; + +static int write_new_event(int fd, const char *buf) +{ + int ret; + + printf("Adding new event: %s\n", buf); + ret = write(fd, buf, strlen(buf)); + if (ret <= 0) + perror("Error: Failed to create event"); + + return ret; +} + +#define MAX_CMDLEN 256 + +static int synthesize_probepoint(struct probe_point *pp) +{ + char *buf; + int i, len, ret; + pp->probes[0] = buf = (char *)calloc(MAX_CMDLEN, sizeof(char)); + if (!buf) + perror_exit("calloc"); + ret = snprintf(buf, MAX_CMDLEN, "%s+%d", pp->function, pp->offset); + if (ret <= 0 || ret >= MAX_CMDLEN) + goto error; + len = ret; + + for (i = 0; i < pp->nr_args; i++) { + ret = snprintf(&buf[len], MAX_CMDLEN - len, " %s", + pp->args[i]); + if (ret <= 0 || ret >= MAX_CMDLEN - len) + goto error; + len += ret; + } + pp->found = 1; + return pp->found; +error: + free(pp->probes[0]); + if (ret > 0) + ret = -E2BIG; + return ret; +} + +int cmd_probe(int argc, const char **argv, const char *prefix __used) +{ + int i, j, fd, ret, need_dwarf = 0; + struct probe_point *pp; + char buf[MAX_CMDLEN]; + + argc = parse_options(argc, argv, options, probe_usage, + PARSE_OPT_STOP_AT_NON_OPTION); + if (argc || session.nr_probe == 0) + usage_with_options(probe_usage, options); + + /* Synthesize return probes */ + for (j = 0; j < session.nr_probe; j++) { + if (session.events[j][0] != 'r') { + need_dwarf = 1; + continue; + } + ret = synthesize_probepoint(&session.probes[j]); + if (ret == -E2BIG) + semantic_error("probe point is too long."); + else if (ret < 0) { + perror("snprintf"); + return -1; + } + } + + if (!need_dwarf) + goto setup_probes; + + if (session.vmlinux) + fd = open(session.vmlinux, O_RDONLY); + else + fd = open_default_vmlinux(); + if (fd < 0) { + perror("vmlinux/module file open"); + return -1; + } + + /* Searching probe points */ + for (j = 0; j < session.nr_probe; j++) { + pp = &session.probes[j]; + if (pp->found) + continue; + + lseek(fd, SEEK_SET, 0); + ret = find_probepoint(fd, pp); + if (ret <= 0) { + fprintf(stderr, "Error: No probe point found.\n"); + return -1; + } + debug("probe event %s found\n", session.events[j]); + } + close(fd); + +setup_probes: + /* Settng up probe points */ + snprintf(buf, MAX_CMDLEN, "%s/../kprobe_events", debugfs_path); + fd = open(buf, O_WRONLY, O_APPEND); + if (fd < 0) { + perror("kprobe_events open"); + return -1; + } + for (j = 0; j < session.nr_probe; j++) { + pp = &session.probes[j]; + if (pp->found == 1) { + snprintf(buf, MAX_CMDLEN, "%s %s\n", + session.events[j], pp->probes[0]); + write_new_event(fd, buf); + } else + for (i = 0; i < pp->found; i++) { + snprintf(buf, MAX_CMDLEN, "%s%d %s\n", + session.events[j], i, pp->probes[i]); + write_new_event(fd, buf); + } + } + close(fd); + return 0; +} + diff --git a/tools/perf/builtin.h b/tools/perf/builtin.h index e11d8d231c3..ad5f0f4c49e 100644 --- a/tools/perf/builtin.h +++ b/tools/perf/builtin.h @@ -25,5 +25,6 @@ extern int cmd_timechart(int argc, const char **argv, const char *prefix); extern int cmd_top(int argc, const char **argv, const char *prefix); extern int cmd_trace(int argc, const char **argv, const char *prefix); extern int cmd_version(int argc, const char **argv, const char *prefix); +extern int cmd_probe(int argc, const char **argv, const char *prefix); #endif diff --git a/tools/perf/perf.c b/tools/perf/perf.c index 19fc7feb9d5..f598120c009 100644 --- a/tools/perf/perf.c +++ b/tools/perf/perf.c @@ -295,6 +295,9 @@ static void handle_internal_command(int argc, const char **argv) { "version", cmd_version, 0 }, { "trace", cmd_trace, 0 }, { "sched", cmd_sched, 0 }, +#ifdef SUPPORT_DWARF + { "probe", cmd_probe, 0 }, +#endif }; unsigned int i; static const char ext[] = STRIP_EXTENSION; diff --git a/tools/perf/util/probe-finder.c b/tools/perf/util/probe-finder.c new file mode 100644 index 00000000000..ec6f53f29e0 --- /dev/null +++ b/tools/perf/util/probe-finder.c @@ -0,0 +1,690 @@ +/* + * probe-finder.c : C expression to kprobe event converter + * + * Written by Masami Hiramatsu + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. + * + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include "probe-finder.h" + + +/* Dwarf_Die Linkage to parent Die */ +struct die_link { + struct die_link *parent; /* Parent die */ + Dwarf_Die die; /* Current die */ +}; + +static Dwarf_Debug __dw_debug; +static Dwarf_Error __dw_error; + +static void msg_exit(int ret, const char *fmt, ...) +{ + va_list ap; + + va_start(ap, fmt); + fprintf(stderr, "Error: "); + vfprintf(stderr, fmt, ap); + va_end(ap); + + fprintf(stderr, "\n"); + exit(ret); +} + + +/* + * Generic dwarf analysis helpers + */ + +#define X86_32_MAX_REGS 8 +const char *x86_32_regs_table[X86_32_MAX_REGS] = { + "%ax", + "%cx", + "%dx", + "%bx", + "$stack", /* Stack address instead of %sp */ + "%bp", + "%si", + "%di", +}; + +#define X86_64_MAX_REGS 16 +const char *x86_64_regs_table[X86_64_MAX_REGS] = { + "%ax", + "%dx", + "%cx", + "%bx", + "%si", + "%di", + "%bp", + "%sp", + "%r8", + "%r9", + "%r10", + "%r11", + "%r12", + "%r13", + "%r14", + "%r15", +}; + +/* TODO: switching by dwarf address size */ +#ifdef __x86_64__ +#define ARCH_MAX_REGS X86_64_MAX_REGS +#define arch_regs_table x86_64_regs_table +#else +#define ARCH_MAX_REGS X86_32_MAX_REGS +#define arch_regs_table x86_32_regs_table +#endif + +/* Return architecture dependent register string (for kprobe-tracer) */ +static const char *get_arch_regstr(unsigned int n) +{ + return (n <= ARCH_MAX_REGS) ? arch_regs_table[n] : NULL; +} + +/* + * Compare the tail of two strings. + * Return 0 if whole of either string is same as another's tail part. + */ +static int strtailcmp(const char *s1, const char *s2) +{ + int i1 = strlen(s1); + int i2 = strlen(s2); + while (--i1 > 0 && --i2 > 0) { + if (s1[i1] != s2[i2]) + return s1[i1] - s2[i2]; + } + return 0; +} + +/* Find the fileno of the target file. */ +static Dwarf_Unsigned die_get_fileno(Dwarf_Die cu_die, const char *fname) +{ + Dwarf_Signed cnt, i; + Dwarf_Unsigned found = 0; + char **srcs; + int ret; + + if (!fname) + return 0; + + ret = dwarf_srcfiles(cu_die, &srcs, &cnt, &__dw_error); + if (ret == DW_DLV_OK) { + for (i = 0; i < cnt && !found; i++) { + if (strtailcmp(srcs[i], fname) == 0) + found = i + 1; + dwarf_dealloc(__dw_debug, srcs[i], DW_DLA_STRING); + } + for (; i < cnt; i++) + dwarf_dealloc(__dw_debug, srcs[i], DW_DLA_STRING); + dwarf_dealloc(__dw_debug, srcs, DW_DLA_LIST); + } + if (found) + debug("found fno: %d\n", (int)found); + return found; +} + +/* Compare diename and tname */ +static int die_compare_name(Dwarf_Die die, const char *tname) +{ + char *name; + int ret; + ret = dwarf_diename(die, &name, &__dw_error); + ERR_IF(ret == DW_DLV_ERROR); + if (ret == DW_DLV_OK) { + ret = strcmp(tname, name); + dwarf_dealloc(__dw_debug, name, DW_DLA_STRING); + } else + ret = -1; + return ret; +} + +/* Check the address is in the subprogram(function). */ +static int die_within_subprogram(Dwarf_Die sp_die, Dwarf_Addr addr, + Dwarf_Signed *offs) +{ + Dwarf_Addr lopc, hipc; + int ret; + + /* TODO: check ranges */ + ret = dwarf_lowpc(sp_die, &lopc, &__dw_error); + ERR_IF(ret == DW_DLV_ERROR); + if (ret == DW_DLV_NO_ENTRY) + return 0; + ret = dwarf_highpc(sp_die, &hipc, &__dw_error); + ERR_IF(ret != DW_DLV_OK); + if (lopc <= addr && addr < hipc) { + *offs = addr - lopc; + return 1; + } else + return 0; +} + +/* Check the die is inlined function */ +static Dwarf_Bool die_inlined_subprogram(Dwarf_Die die) +{ + /* TODO: check strictly */ + Dwarf_Bool inl; + int ret; + + ret = dwarf_hasattr(die, DW_AT_inline, &inl, &__dw_error); + ERR_IF(ret == DW_DLV_ERROR); + return inl; +} + +/* Get the offset of abstruct_origin */ +static Dwarf_Off die_get_abstract_origin(Dwarf_Die die) +{ + Dwarf_Attribute attr; + Dwarf_Off cu_offs; + int ret; + + ret = dwarf_attr(die, DW_AT_abstract_origin, &attr, &__dw_error); + ERR_IF(ret != DW_DLV_OK); + ret = dwarf_formref(attr, &cu_offs, &__dw_error); + ERR_IF(ret != DW_DLV_OK); + dwarf_dealloc(__dw_debug, attr, DW_DLA_ATTR); + return cu_offs; +} + +/* Get entry pc(or low pc, 1st entry of ranges) of the die */ +static Dwarf_Addr die_get_entrypc(Dwarf_Die die) +{ + Dwarf_Attribute attr; + Dwarf_Addr addr; + Dwarf_Off offs; + Dwarf_Ranges *ranges; + Dwarf_Signed cnt; + int ret; + + /* Try to get entry pc */ + ret = dwarf_attr(die, DW_AT_entry_pc, &attr, &__dw_error); + ERR_IF(ret == DW_DLV_ERROR); + if (ret == DW_DLV_OK) { + ret = dwarf_formaddr(attr, &addr, &__dw_error); + ERR_IF(ret != DW_DLV_OK); + dwarf_dealloc(__dw_debug, attr, DW_DLA_ATTR); + return addr; + } + + /* Try to get low pc */ + ret = dwarf_lowpc(die, &addr, &__dw_error); + ERR_IF(ret == DW_DLV_ERROR); + if (ret == DW_DLV_OK) + return addr; + + /* Try to get ranges */ + ret = dwarf_attr(die, DW_AT_ranges, &attr, &__dw_error); + ERR_IF(ret != DW_DLV_OK); + ret = dwarf_formref(attr, &offs, &__dw_error); + ERR_IF(ret != DW_DLV_OK); + ret = dwarf_get_ranges(__dw_debug, offs, &ranges, &cnt, NULL, + &__dw_error); + ERR_IF(ret != DW_DLV_OK); + addr = ranges[0].dwr_addr1; + dwarf_ranges_dealloc(__dw_debug, ranges, cnt); + return addr; +} + +/* + * Search a Die from Die tree. + * Note: cur_link->die should be deallocated in this function. + */ +static int __search_die_tree(struct die_link *cur_link, + int (*die_cb)(struct die_link *, void *), + void *data) +{ + Dwarf_Die new_die; + struct die_link new_link; + int ret; + + if (!die_cb) + return 0; + + /* Check current die */ + while (!(ret = die_cb(cur_link, data))) { + /* Check child die */ + ret = dwarf_child(cur_link->die, &new_die, &__dw_error); + ERR_IF(ret == DW_DLV_ERROR); + if (ret == DW_DLV_OK) { + new_link.parent = cur_link; + new_link.die = new_die; + ret = __search_die_tree(&new_link, die_cb, data); + if (ret) + break; + } + + /* Move to next sibling */ + ret = dwarf_siblingof(__dw_debug, cur_link->die, &new_die, + &__dw_error); + ERR_IF(ret == DW_DLV_ERROR); + dwarf_dealloc(__dw_debug, cur_link->die, DW_DLA_DIE); + cur_link->die = new_die; + if (ret == DW_DLV_NO_ENTRY) + return 0; + } + dwarf_dealloc(__dw_debug, cur_link->die, DW_DLA_DIE); + return ret; +} + +/* Search a die in its children's die tree */ +static int search_die_from_children(Dwarf_Die parent_die, + int (*die_cb)(struct die_link *, void *), + void *data) +{ + struct die_link new_link; + int ret; + + new_link.parent = NULL; + ret = dwarf_child(parent_die, &new_link.die, &__dw_error); + ERR_IF(ret == DW_DLV_ERROR); + if (ret == DW_DLV_OK) + return __search_die_tree(&new_link, die_cb, data); + else + return 0; +} + +/* Find a locdesc corresponding to the address */ +static int attr_get_locdesc(Dwarf_Attribute attr, Dwarf_Locdesc *desc, + Dwarf_Addr addr) +{ + Dwarf_Signed lcnt; + Dwarf_Locdesc **llbuf; + int ret, i; + + ret = dwarf_loclist_n(attr, &llbuf, &lcnt, &__dw_error); + ERR_IF(ret != DW_DLV_OK); + ret = DW_DLV_NO_ENTRY; + for (i = 0; i < lcnt; ++i) { + if (llbuf[i]->ld_lopc <= addr && + llbuf[i]->ld_hipc > addr) { + memcpy(desc, llbuf[i], sizeof(Dwarf_Locdesc)); + desc->ld_s = + malloc(sizeof(Dwarf_Loc) * llbuf[i]->ld_cents); + ERR_IF(desc->ld_s == NULL); + memcpy(desc->ld_s, llbuf[i]->ld_s, + sizeof(Dwarf_Loc) * llbuf[i]->ld_cents); + ret = DW_DLV_OK; + break; + } + dwarf_dealloc(__dw_debug, llbuf[i]->ld_s, DW_DLA_LOC_BLOCK); + dwarf_dealloc(__dw_debug, llbuf[i], DW_DLA_LOCDESC); + } + /* Releasing loop */ + for (; i < lcnt; ++i) { + dwarf_dealloc(__dw_debug, llbuf[i]->ld_s, DW_DLA_LOC_BLOCK); + dwarf_dealloc(__dw_debug, llbuf[i], DW_DLA_LOCDESC); + } + dwarf_dealloc(__dw_debug, llbuf, DW_DLA_LIST); + return ret; +} + +/* + * Probe finder related functions + */ + +/* Show a location */ +static void show_location(Dwarf_Loc *loc, struct probe_finder *pf) +{ + Dwarf_Small op; + Dwarf_Unsigned regn; + Dwarf_Signed offs; + int deref = 0, ret; + const char *regs; + + op = loc->lr_atom; + + /* If this is based on frame buffer, set the offset */ + if (op == DW_OP_fbreg) { + deref = 1; + offs = (Dwarf_Signed)loc->lr_number; + op = pf->fbloc.ld_s[0].lr_atom; + loc = &pf->fbloc.ld_s[0]; + } else + offs = 0; + + if (op >= DW_OP_breg0 && op <= DW_OP_breg31) { + regn = op - DW_OP_breg0; + offs += (Dwarf_Signed)loc->lr_number; + deref = 1; + } else if (op >= DW_OP_reg0 && op <= DW_OP_reg31) { + regn = op - DW_OP_reg0; + } else if (op == DW_OP_bregx) { + regn = loc->lr_number; + offs += (Dwarf_Signed)loc->lr_number2; + deref = 1; + } else if (op == DW_OP_regx) { + regn = loc->lr_number; + } else + msg_exit(-EINVAL, "Dwarf_OP %d is not supported.\n", op); + + regs = get_arch_regstr(regn); + if (!regs) + msg_exit(-EINVAL, "%lld exceeds max register number.\n", regn); + + if (deref) + ret = snprintf(pf->buf, pf->len, + " %s=%+lld(%s)", pf->var, offs, regs); + else + ret = snprintf(pf->buf, pf->len, " %s=%s", pf->var, regs); + ERR_IF(ret < 0); + ERR_IF(ret >= pf->len); +} + +/* Show a variables in kprobe event format */ +static void show_variable(Dwarf_Die vr_die, struct probe_finder *pf) +{ + Dwarf_Attribute attr; + Dwarf_Locdesc ld; + int ret; + + ret = dwarf_attr(vr_die, DW_AT_location, &attr, &__dw_error); + if (ret != DW_DLV_OK) + goto error; + ret = attr_get_locdesc(attr, &ld, (pf->addr - pf->cu_base)); + if (ret != DW_DLV_OK) + goto error; + /* TODO? */ + ERR_IF(ld.ld_cents != 1); + show_location(&ld.ld_s[0], pf); + free(ld.ld_s); + dwarf_dealloc(__dw_debug, attr, DW_DLA_ATTR); + return ; +error: + msg_exit(-1, "Failed to find the location of %s at this address.\n" + " Perhaps, it was optimized out.\n", pf->var); +} + +static int variable_callback(struct die_link *dlink, void *data) +{ + struct probe_finder *pf = (struct probe_finder *)data; + Dwarf_Half tag; + int ret; + + ret = dwarf_tag(dlink->die, &tag, &__dw_error); + ERR_IF(ret == DW_DLV_ERROR); + if ((tag == DW_TAG_formal_parameter || + tag == DW_TAG_variable) && + (die_compare_name(dlink->die, pf->var) == 0)) { + show_variable(dlink->die, pf); + return 1; + } + /* TODO: Support struct members and arrays */ + return 0; +} + +/* Find a variable in a subprogram die */ +static void find_variable(Dwarf_Die sp_die, struct probe_finder *pf) +{ + int ret; + + if (!is_c_varname(pf->var)) { + /* Output raw parameters */ + ret = snprintf(pf->buf, pf->len, " %s", pf->var); + ERR_IF(ret < 0); + ERR_IF(ret >= pf->len); + return ; + } + + debug("Searching '%s' variable in context.\n", pf->var); + /* Search child die for local variables and parameters. */ + ret = search_die_from_children(sp_die, variable_callback, pf); + if (!ret) + msg_exit(-1, "Failed to find '%s' in this function.\n", + pf->var); +} + +/* Get a frame base on the address */ +static void get_current_frame_base(Dwarf_Die sp_die, struct probe_finder *pf) +{ + Dwarf_Attribute attr; + int ret; + + ret = dwarf_attr(sp_die, DW_AT_frame_base, &attr, &__dw_error); + ERR_IF(ret != DW_DLV_OK); + ret = attr_get_locdesc(attr, &pf->fbloc, (pf->addr - pf->cu_base)); + ERR_IF(ret != DW_DLV_OK); + dwarf_dealloc(__dw_debug, attr, DW_DLA_ATTR); +} + +static void free_current_frame_base(struct probe_finder *pf) +{ + free(pf->fbloc.ld_s); + memset(&pf->fbloc, 0, sizeof(Dwarf_Locdesc)); +} + +/* Show a probe point to output buffer */ +static void show_probepoint(Dwarf_Die sp_die, Dwarf_Signed offs, + struct probe_finder *pf) +{ + struct probe_point *pp = pf->pp; + char *name; + char tmp[MAX_PROBE_BUFFER]; + int ret, i, len; + + /* Output name of probe point */ + ret = dwarf_diename(sp_die, &name, &__dw_error); + ERR_IF(ret == DW_DLV_ERROR); + if (ret == DW_DLV_OK) { + ret = snprintf(tmp, MAX_PROBE_BUFFER, "%s+%u", name, + (unsigned int)offs); + dwarf_dealloc(__dw_debug, name, DW_DLA_STRING); + } else { + /* This function has no name. */ + ret = snprintf(tmp, MAX_PROBE_BUFFER, "0x%llx", pf->addr); + } + ERR_IF(ret < 0); + ERR_IF(ret >= MAX_PROBE_BUFFER); + len = ret; + + /* Find each argument */ + get_current_frame_base(sp_die, pf); + for (i = 0; i < pp->nr_args; i++) { + pf->var = pp->args[i]; + pf->buf = &tmp[len]; + pf->len = MAX_PROBE_BUFFER - len; + find_variable(sp_die, pf); + len += strlen(pf->buf); + } + free_current_frame_base(pf); + + pp->probes[pp->found] = strdup(tmp); + pp->found++; +} + +static int probeaddr_callback(struct die_link *dlink, void *data) +{ + struct probe_finder *pf = (struct probe_finder *)data; + Dwarf_Half tag; + Dwarf_Signed offs; + int ret; + + ret = dwarf_tag(dlink->die, &tag, &__dw_error); + ERR_IF(ret == DW_DLV_ERROR); + /* Check the address is in this subprogram */ + if (tag == DW_TAG_subprogram && + die_within_subprogram(dlink->die, pf->addr, &offs)) { + show_probepoint(dlink->die, offs, pf); + return 1; + } + return 0; +} + +/* Find probe point from its line number */ +static void find_by_line(Dwarf_Die cu_die, struct probe_finder *pf) +{ + struct probe_point *pp = pf->pp; + Dwarf_Signed cnt, i; + Dwarf_Line *lines; + Dwarf_Unsigned lineno = 0; + Dwarf_Addr addr; + Dwarf_Unsigned fno; + int ret; + + ret = dwarf_srclines(cu_die, &lines, &cnt, &__dw_error); + ERR_IF(ret != DW_DLV_OK); + + for (i = 0; i < cnt; i++) { + ret = dwarf_line_srcfileno(lines[i], &fno, &__dw_error); + ERR_IF(ret != DW_DLV_OK); + if (fno != pf->fno) + continue; + + ret = dwarf_lineno(lines[i], &lineno, &__dw_error); + ERR_IF(ret != DW_DLV_OK); + if (lineno != (Dwarf_Unsigned)pp->line) + continue; + + ret = dwarf_lineaddr(lines[i], &addr, &__dw_error); + ERR_IF(ret != DW_DLV_OK); + debug("Probe point found: 0x%llx\n", addr); + pf->addr = addr; + /* Search a real subprogram including this line, */ + ret = search_die_from_children(cu_die, probeaddr_callback, pf); + if (ret == 0) + msg_exit(-1, + "Probe point is not found in subprograms.\n"); + /* Continuing, because target line might be inlined. */ + } + dwarf_srclines_dealloc(__dw_debug, lines, cnt); +} + +/* Search function from function name */ +static int probefunc_callback(struct die_link *dlink, void *data) +{ + struct probe_finder *pf = (struct probe_finder *)data; + struct probe_point *pp = pf->pp; + struct die_link *lk; + Dwarf_Signed offs; + Dwarf_Half tag; + int ret; + + ret = dwarf_tag(dlink->die, &tag, &__dw_error); + ERR_IF(ret == DW_DLV_ERROR); + if (tag == DW_TAG_subprogram) { + if (die_compare_name(dlink->die, pp->function) == 0) { + if (die_inlined_subprogram(dlink->die)) { + /* Inlined function, save it. */ + ret = dwarf_die_CU_offset(dlink->die, + &pf->inl_offs, + &__dw_error); + ERR_IF(ret != DW_DLV_OK); + debug("inline definition offset %lld\n", + pf->inl_offs); + return 0; + } + /* Get probe address */ + pf->addr = die_get_entrypc(dlink->die); + pf->addr += pp->offset; + /* TODO: Check the address in this function */ + show_probepoint(dlink->die, pp->offset, pf); + /* Continue to search */ + } + } else if (tag == DW_TAG_inlined_subroutine && pf->inl_offs) { + if (die_get_abstract_origin(dlink->die) == pf->inl_offs) { + /* Get probe address */ + pf->addr = die_get_entrypc(dlink->die); + pf->addr += pp->offset; + debug("found inline addr: 0x%llx\n", pf->addr); + /* Inlined function. Get a real subprogram */ + for (lk = dlink->parent; lk != NULL; lk = lk->parent) { + tag = 0; + dwarf_tag(lk->die, &tag, &__dw_error); + ERR_IF(ret == DW_DLV_ERROR); + if (tag == DW_TAG_subprogram && + !die_inlined_subprogram(lk->die)) + goto found; + } + msg_exit(-1, "Failed to find real subprogram.\n"); +found: + /* Get offset from subprogram */ + ret = die_within_subprogram(lk->die, pf->addr, &offs); + ERR_IF(!ret); + show_probepoint(lk->die, offs, pf); + /* Continue to search */ + } + } + return 0; +} + +static void find_by_func(Dwarf_Die cu_die, struct probe_finder *pf) +{ + search_die_from_children(cu_die, probefunc_callback, pf); +} + +/* Find a probe point */ +int find_probepoint(int fd, struct probe_point *pp) +{ + Dwarf_Half addr_size = 0; + Dwarf_Unsigned next_cuh = 0; + Dwarf_Die cu_die = 0; + int cu_number = 0, ret; + struct probe_finder pf = {.pp = pp}; + + ret = dwarf_init(fd, DW_DLC_READ, 0, 0, &__dw_debug, &__dw_error); + if (ret != DW_DLV_OK) + msg_exit(-1, "Failed to call dwarf_init(). " + "Maybe, not a dwarf file?\n"); + + pp->found = 0; + while (++cu_number) { + /* Search CU (Compilation Unit) */ + ret = dwarf_next_cu_header(__dw_debug, NULL, NULL, NULL, + &addr_size, &next_cuh, &__dw_error); + ERR_IF(ret == DW_DLV_ERROR); + if (ret == DW_DLV_NO_ENTRY) + break; + + /* Get the DIE(Debugging Information Entry) of this CU */ + ret = dwarf_siblingof(__dw_debug, 0, &cu_die, &__dw_error); + ERR_IF(ret != DW_DLV_OK); + + /* Check if target file is included. */ + if (pp->file) + pf.fno = die_get_fileno(cu_die, pp->file); + + if (!pp->file || pf.fno) { + /* Save CU base address (for frame_base) */ + ret = dwarf_lowpc(cu_die, &pf.cu_base, &__dw_error); + ERR_IF(ret == DW_DLV_ERROR); + if (ret == DW_DLV_NO_ENTRY) + pf.cu_base = 0; + if (pp->line) + find_by_line(cu_die, &pf); + if (pp->function) + find_by_func(cu_die, &pf); + } + dwarf_dealloc(__dw_debug, cu_die, DW_DLA_DIE); + } + ret = dwarf_finish(__dw_debug, &__dw_error); + ERR_IF(ret != DW_DLV_OK); + + return pp->found; +} + diff --git a/tools/perf/util/probe-finder.h b/tools/perf/util/probe-finder.h new file mode 100644 index 00000000000..af920de6486 --- /dev/null +++ b/tools/perf/util/probe-finder.h @@ -0,0 +1,68 @@ +#ifndef _PROBE_FINDER_H +#define _PROBE_FINDER_H + +#define _stringify(n) #n +#define stringify(n) _stringify(n) + +#ifdef DEBUG +#define debug(fmt ...) \ + fprintf(stderr, "DBG(" __FILE__ ":" stringify(__LINE__) "): " fmt) +#else +#define debug(fmt ...) do {} while (0) +#endif + +#define ERR_IF(cnd) \ + do { if (cnd) { \ + fprintf(stderr, "Error (" __FILE__ ":" stringify(__LINE__) \ + "): " stringify(cnd) "\n"); \ + exit(1); \ + } } while (0) + +#define MAX_PATH_LEN 256 +#define MAX_PROBE_BUFFER 1024 +#define MAX_PROBES 128 + +static inline int is_c_varname(const char *name) +{ + /* TODO */ + return isalpha(name[0]) || name[0] == '_'; +} + +struct probe_point { + /* Inputs */ + char *file; /* File name */ + int line; /* Line number */ + + char *function; /* Function name */ + int offset; /* Offset bytes */ + + int nr_args; /* Number of arguments */ + char **args; /* Arguments */ + + /* Output */ + int found; /* Number of found probe points */ + char *probes[MAX_PROBES]; /* Output buffers (will be allocated)*/ +}; + +extern int find_probepoint(int fd, struct probe_point *pp); + +#include +#include + +struct probe_finder { + struct probe_point *pp; /* Target probe point */ + + /* For function searching */ + Dwarf_Addr addr; /* Address */ + Dwarf_Unsigned fno; /* File number */ + Dwarf_Off inl_offs; /* Inline offset */ + + /* For variable searching */ + Dwarf_Addr cu_base; /* Current CU base address */ + Dwarf_Locdesc fbloc; /* Location of Current Frame Base */ + const char *var; /* Current variable name */ + char *buf; /* Current output buffer */ + int len; /* Length of output buffer */ +}; + +#endif /*_PROBE_FINDER_H */ -- cgit v1.2.3-18-g5258 From 23e8ec0d1c410f2f1d81050ee155db229abb1707 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Wed, 7 Oct 2009 18:28:30 -0400 Subject: perf probe: Add perf probe command support without libdwarf Enables 'perf probe' even if libdwarf is not installed. If libdwarf is not found, 'perf probe' just disables dwarf support. Users can use 'perf probe' to set up new events by using kprobe_events format. Signed-off-by: Masami Hiramatsu Cc: Ingo Molnar Cc: Thomas Gleixner Cc: Arnaldo Carvalho de Melo Cc: Steven Rostedt Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Christoph Hellwig Cc: Ananth N Mavinakayanahalli Cc: Jim Keniston Cc: Frank Ch. Eigler LKML-Reference: <20091007222830.1684.25665.stgit@dhcp-100-2-132.bos.redhat.com> Signed-off-by: Frederic Weisbecker --- tools/perf/Makefile | 6 +++--- tools/perf/builtin-probe.c | 42 +++++++++++++++++++++++++++++++++++------- tools/perf/perf.c | 2 -- tools/perf/util/probe-finder.h | 2 ++ 4 files changed, 40 insertions(+), 12 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/Makefile b/tools/perf/Makefile index 6dabcf1a4df..52b1f438e71 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -385,6 +385,7 @@ BUILTIN_OBJS += builtin-stat.o BUILTIN_OBJS += builtin-timechart.o BUILTIN_OBJS += builtin-top.o BUILTIN_OBJS += builtin-trace.o +BUILTIN_OBJS += builtin-probe.o PERFLIBS = $(LIB_FILE) @@ -420,13 +421,12 @@ ifneq ($(shell sh -c "(echo '\#include '; echo 'int main(void) { Elf * endif ifneq ($(shell sh -c "(echo '\#include '; echo '\#include '; echo 'int main(void) { Dwarf_Debug dbg; Dwarf_Error err; dwarf_init(0, DW_DLC_READ, 0, 0, &dbg, &err); return (long)dbg; }') | $(CC) -x c - $(ALL_CFLAGS) -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -ldwarf -lelf -o /dev/null $(ALL_LDFLAGS) > /dev/null 2>&1 && echo y"), y) - msg := $(warning No libdwarf.h found, disables probe subcommand. Please install libdwarf-dev/libdwarf-devel); + msg := $(warning No libdwarf.h found, disables dwarf support. Please install libdwarf-dev/libdwarf-devel); + BASIC_CFLAGS += -DNO_LIBDWARF else EXTLIBS += -lelf -ldwarf LIB_H += util/probe-finder.h LIB_OBJS += util/probe-finder.o - BUILTIN_OBJS += builtin-probe.o - BASIC_CFLAGS += -DSUPPORT_DWARF endif ifdef NO_DEMANGLE diff --git a/tools/perf/builtin-probe.c b/tools/perf/builtin-probe.c index 24b64b5cefc..73c883b715c 100644 --- a/tools/perf/builtin-probe.c +++ b/tools/perf/builtin-probe.c @@ -54,6 +54,7 @@ const char *default_search_path[NR_SEARCH_PATH] = { static struct { char *vmlinux; char *release; + int need_dwarf; int nr_probe; struct probe_point probes[MAX_PROBES]; char *events[MAX_PROBES]; @@ -162,6 +163,8 @@ static int parse_probepoint(const struct option *opt __used, pp->function, pp->file, pp->offset); } free(argv[1]); + if (pp->file) + session.need_dwarf = 1; /* Copy arguments */ pp->nr_args = argc - 2; @@ -173,15 +176,19 @@ static int parse_probepoint(const struct option *opt __used, } /* Ensure return probe has no C argument */ - if (retp) - for (i = 0; i < pp->nr_args; i++) - if (is_c_varname(pp->args[i])) + for (i = 0; i < pp->nr_args; i++) + if (is_c_varname(pp->args[i])) { + if (retp) semantic_error("You can't specify local" " variable for kretprobe"); + session.need_dwarf = 1; + } + debug("%d arguments\n", pp->nr_args); return 0; } +#ifndef NO_LIBDWARF static int open_default_vmlinux(void) { struct utsname uts; @@ -209,6 +216,7 @@ static int open_default_vmlinux(void) } return fd; } +#endif static const char * const probe_usage[] = { "perf probe [] -P 'PROBEDEF' [-P 'PROBEDEF' ...]", @@ -216,10 +224,16 @@ static const char * const probe_usage[] = { }; static const struct option options[] = { +#ifndef NO_LIBDWARF OPT_STRING('k', "vmlinux", &session.vmlinux, "file", "vmlinux/module pathname"), +#endif OPT_CALLBACK('P', "probe", NULL, +#ifdef NO_LIBDWARF + "p|r:[GRP/]NAME FUNC[+OFFS] [ARG ...]", +#else "p|r:[GRP/]NAME FUNC[+OFFS][@SRC]|@SRC:LINE [ARG ...]", +#endif "probe point definition, where\n" "\t\tp:\tkprobe probe\n" "\t\tr:\tkretprobe probe\n" @@ -227,9 +241,13 @@ static const struct option options[] = { "\t\tNAME:\tEvent name\n" "\t\tFUNC:\tFunction name\n" "\t\tOFFS:\tOffset from function entry (in byte)\n" +#ifdef NO_LIBDWARF + "\t\tARG:\tProbe argument (only \n" +#else "\t\tSRC:\tSource code path\n" "\t\tLINE:\tLine number\n" "\t\tARG:\tProbe argument (local variable name or\n" +#endif "\t\t\tkprobe-tracer argument format is supported.)\n", parse_probepoint), OPT_END() @@ -279,7 +297,7 @@ error: int cmd_probe(int argc, const char **argv, const char *prefix __used) { - int i, j, fd, ret, need_dwarf = 0; + int i, j, fd, ret; struct probe_point *pp; char buf[MAX_CMDLEN]; @@ -288,12 +306,19 @@ int cmd_probe(int argc, const char **argv, const char *prefix __used) if (argc || session.nr_probe == 0) usage_with_options(probe_usage, options); - /* Synthesize return probes */ +#ifdef NO_LIBDWARF + if (session.need_dwarf) + semantic_error("Dwarf-analysis is not supported"); +#endif + + /* Synthesize probes without dwarf */ for (j = 0; j < session.nr_probe; j++) { +#ifndef NO_LIBDWARF if (session.events[j][0] != 'r') { - need_dwarf = 1; + session.need_dwarf = 1; continue; } +#endif ret = synthesize_probepoint(&session.probes[j]); if (ret == -E2BIG) semantic_error("probe point is too long."); @@ -303,7 +328,8 @@ int cmd_probe(int argc, const char **argv, const char *prefix __used) } } - if (!need_dwarf) +#ifndef NO_LIBDWARF + if (!session.need_dwarf) goto setup_probes; if (session.vmlinux) @@ -332,6 +358,8 @@ int cmd_probe(int argc, const char **argv, const char *prefix __used) close(fd); setup_probes: +#endif /* !NO_LIBDWARF */ + /* Settng up probe points */ snprintf(buf, MAX_CMDLEN, "%s/../kprobe_events", debugfs_path); fd = open(buf, O_WRONLY, O_APPEND); diff --git a/tools/perf/perf.c b/tools/perf/perf.c index f598120c009..c570d177d5c 100644 --- a/tools/perf/perf.c +++ b/tools/perf/perf.c @@ -295,9 +295,7 @@ static void handle_internal_command(int argc, const char **argv) { "version", cmd_version, 0 }, { "trace", cmd_trace, 0 }, { "sched", cmd_sched, 0 }, -#ifdef SUPPORT_DWARF { "probe", cmd_probe, 0 }, -#endif }; unsigned int i; static const char ext[] = STRIP_EXTENSION; diff --git a/tools/perf/util/probe-finder.h b/tools/perf/util/probe-finder.h index af920de6486..306810c32f6 100644 --- a/tools/perf/util/probe-finder.h +++ b/tools/perf/util/probe-finder.h @@ -44,6 +44,7 @@ struct probe_point { char *probes[MAX_PROBES]; /* Output buffers (will be allocated)*/ }; +#ifndef NO_LIBDWARF extern int find_probepoint(int fd, struct probe_point *pp); #include @@ -64,5 +65,6 @@ struct probe_finder { char *buf; /* Current output buffer */ int len; /* Length of output buffer */ }; +#endif /* NO_LIBDWARF */ #endif /*_PROBE_FINDER_H */ -- cgit v1.2.3-18-g5258 From cfed95a693e1ea5d08b9c9019bc30e448437ee2f Mon Sep 17 00:00:00 2001 From: Vincent Legoll Date: Tue, 13 Oct 2009 10:18:16 +0200 Subject: perf tools: Do not manually count string lengths Use strlen & macros instead of manually counting string lengths as this is error prone and may lend to bugs. Signed-off-by: Vincent Legoll Cc: Linus Torvalds LKML-Reference: <4727185d0910130118m5387058dndb02ac9b384af9f0@mail.gmail.com> Signed-off-by: Ingo Molnar --- tools/perf/perf.c | 16 ++++++++-------- tools/perf/util/cache.h | 5 +++++ 2 files changed, 13 insertions(+), 8 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/perf.c b/tools/perf/perf.c index 19fc7feb9d5..624e62d9d1e 100644 --- a/tools/perf/perf.c +++ b/tools/perf/perf.c @@ -89,8 +89,8 @@ static int handle_options(const char*** argv, int* argc, int* envchanged) /* * Check remaining flags. */ - if (!prefixcmp(cmd, "--exec-path")) { - cmd += 11; + if (!prefixcmp(cmd, CMD_EXEC_PATH)) { + cmd += strlen(CMD_EXEC_PATH); if (*cmd == '=') perf_set_argv_exec_path(cmd + 1); else { @@ -117,8 +117,8 @@ static int handle_options(const char*** argv, int* argc, int* envchanged) (*argv)++; (*argc)--; handled++; - } else if (!prefixcmp(cmd, "--perf-dir=")) { - setenv(PERF_DIR_ENVIRONMENT, cmd + 10, 1); + } else if (!prefixcmp(cmd, CMD_PERF_DIR)) { + setenv(PERF_DIR_ENVIRONMENT, cmd + strlen(CMD_PERF_DIR), 1); if (envchanged) *envchanged = 1; } else if (!strcmp(cmd, "--work-tree")) { @@ -131,8 +131,8 @@ static int handle_options(const char*** argv, int* argc, int* envchanged) *envchanged = 1; (*argv)++; (*argc)--; - } else if (!prefixcmp(cmd, "--work-tree=")) { - setenv(PERF_WORK_TREE_ENVIRONMENT, cmd + 12, 1); + } else if (!prefixcmp(cmd, CMD_WORK_TREE)) { + setenv(PERF_WORK_TREE_ENVIRONMENT, cmd + strlen(CMD_WORK_TREE), 1); if (envchanged) *envchanged = 1; } else if (!strcmp(cmd, "--debugfs-dir")) { @@ -146,8 +146,8 @@ static int handle_options(const char*** argv, int* argc, int* envchanged) *envchanged = 1; (*argv)++; (*argc)--; - } else if (!prefixcmp(cmd, "--debugfs-dir=")) { - strncpy(debugfs_mntpt, cmd + 14, MAXPATHLEN); + } else if (!prefixcmp(cmd, CMD_DEBUGFS_DIR)) { + strncpy(debugfs_mntpt, cmd + strlen(CMD_DEBUGFS_DIR), MAXPATHLEN); debugfs_mntpt[MAXPATHLEN - 1] = '\0'; if (envchanged) *envchanged = 1; diff --git a/tools/perf/util/cache.h b/tools/perf/util/cache.h index f26172c0c91..918eb376abe 100644 --- a/tools/perf/util/cache.h +++ b/tools/perf/util/cache.h @@ -5,6 +5,11 @@ #include "strbuf.h" #include "../perf.h" +#define CMD_EXEC_PATH "--exec-path" +#define CMD_PERF_DIR "--perf-dir=" +#define CMD_WORK_TREE "--work-tree=" +#define CMD_DEBUGFS_DIR "--debugfs-dir=" + #define PERF_DIR_ENVIRONMENT "PERF_DIR" #define PERF_WORK_TREE_ENVIRONMENT "PERF_WORK_TREE" #define DEFAULT_PERF_DIR_ENVIRONMENT ".perf" -- cgit v1.2.3-18-g5258 From f4f0b418188cc7995375acbb54e87c80f21861bd Mon Sep 17 00:00:00 2001 From: Mike Galbraith Date: Tue, 13 Oct 2009 14:57:20 +0200 Subject: perf tools: Remove expensive old debug code from perf top Calling gettimeofday() at high frequency is painful for handicapped boxen. The spot calling gettimeofday() is old unneeded debug code, so remove it. Reported-by: Ingo Molnar Signed-off-by: Mike Galbraith Cc: Peter Zijlstra Cc: Peter Zijlstra LKML-Reference: <1255438640.7173.1.camel@marge.simson.net> Signed-off-by: Ingo Molnar --- tools/perf/builtin-top.c | 15 +-------------- 1 file changed, 1 insertion(+), 14 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index c0f69e80b2c..2d8806bac25 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -870,8 +870,6 @@ static unsigned int mmap_read_head(struct mmap_data *md) return head; } -struct timeval last_read, this_read; - static void mmap_read_counter(struct mmap_data *md) { unsigned int head = mmap_read_head(md); @@ -879,8 +877,6 @@ static void mmap_read_counter(struct mmap_data *md) unsigned char *data = md->base + page_size; int diff; - gettimeofday(&this_read, NULL); - /* * If we're further behind than half the buffer, there's a chance * the writer will bite our tail and mess up the samples under us. @@ -891,14 +887,7 @@ static void mmap_read_counter(struct mmap_data *md) */ diff = head - old; if (diff > md->mask / 2 || diff < 0) { - struct timeval iv; - unsigned long msecs; - - timersub(&this_read, &last_read, &iv); - msecs = iv.tv_sec*1000 + iv.tv_usec/1000; - - fprintf(stderr, "WARNING: failed to keep up with mmap data." - " Last read %lu msecs ago.\n", msecs); + fprintf(stderr, "WARNING: failed to keep up with mmap data.\n"); /* * head points to a known good entry, start there. @@ -906,8 +895,6 @@ static void mmap_read_counter(struct mmap_data *md) old = head; } - last_read = this_read; - for (; old != head;) { event_t *event = (event_t *)&data[old & md->mask]; -- cgit v1.2.3-18-g5258 From d5b889f2ecec7849e851ddd31c34bdfb3482b5de Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Tue, 13 Oct 2009 11:16:29 -0300 Subject: perf tools: Move threads & last_match to threads.c This was just being copy'n'pasted all over. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: Peter Zijlstra Cc: Paul Mackerras Cc: Mike Galbraith LKML-Reference: <20091013141629.GD21809@ghostprotocols.net> Signed-off-by: Ingo Molnar --- tools/perf/builtin-annotate.c | 25 +++++++------------------ tools/perf/builtin-report.c | 26 +++++++------------------- tools/perf/builtin-sched.c | 30 +++++++++++------------------- tools/perf/builtin-trace.c | 13 +++---------- tools/perf/util/thread.c | 27 ++++++++++++++------------- tools/perf/util/thread.h | 8 +++----- 6 files changed, 45 insertions(+), 84 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c index 8c84320ecb0..3fe0de03004 100644 --- a/tools/perf/builtin-annotate.c +++ b/tools/perf/builtin-annotate.c @@ -37,10 +37,6 @@ static int print_line; static unsigned long page_size; static unsigned long mmap_window = 32; -static struct rb_root threads; -static struct thread *last_match; - - struct sym_ext { struct rb_node node; double percent; @@ -96,12 +92,10 @@ static int process_sample_event(event_t *event, unsigned long offset, unsigned long head) { char level; - struct thread *thread; u64 ip = event->ip.ip; struct map *map = NULL; struct symbol *sym = NULL; - - thread = threads__findnew(event->ip.pid, &threads, &last_match); + struct thread *thread = threads__findnew(event->ip.pid); dump_printf("%p [%p]: PERF_EVENT (IP, %d): %d: %p\n", (void *)(offset + head), @@ -166,10 +160,8 @@ got_map: static int process_mmap_event(event_t *event, unsigned long offset, unsigned long head) { - struct thread *thread; struct map *map = map__new(&event->mmap, NULL, 0); - - thread = threads__findnew(event->mmap.pid, &threads, &last_match); + struct thread *thread = threads__findnew(event->mmap.pid); dump_printf("%p [%p]: PERF_RECORD_MMAP %d: [%p(%p) @ %p]: %s\n", (void *)(offset + head), @@ -194,9 +186,8 @@ process_mmap_event(event_t *event, unsigned long offset, unsigned long head) static int process_comm_event(event_t *event, unsigned long offset, unsigned long head) { - struct thread *thread; + struct thread *thread = threads__findnew(event->comm.pid); - thread = threads__findnew(event->comm.pid, &threads, &last_match); dump_printf("%p [%p]: PERF_RECORD_COMM: %s:%d\n", (void *)(offset + head), (void *)(long)(event->header.size), @@ -215,11 +206,9 @@ process_comm_event(event_t *event, unsigned long offset, unsigned long head) static int process_fork_event(event_t *event, unsigned long offset, unsigned long head) { - struct thread *thread; - struct thread *parent; + struct thread *thread = threads__findnew(event->fork.pid); + struct thread *parent = threads__findnew(event->fork.ppid); - thread = threads__findnew(event->fork.pid, &threads, &last_match); - parent = threads__findnew(event->fork.ppid, &threads, &last_match); dump_printf("%p [%p]: PERF_RECORD_FORK: %d:%d\n", (void *)(offset + head), (void *)(long)(event->header.size), @@ -558,7 +547,7 @@ static int __cmd_annotate(void) uint32_t size; char *buf; - register_idle_thread(&threads, &last_match); + register_idle_thread(); input = open(input_name, O_RDONLY); if (input < 0) { @@ -659,7 +648,7 @@ more: return 0; if (verbose > 3) - threads__fprintf(stdout, &threads); + threads__fprintf(stdout); if (verbose > 2) dsos__fprintf(stdout); diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index f57a23b19f3..015c7974596 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -55,9 +55,6 @@ static char callchain_default_opt[] = "fractal,0.5"; static char *cwd; static int cwdlen; -static struct rb_root threads; -static struct thread *last_match; - static struct perf_header *header; static u64 sample_type; @@ -593,15 +590,13 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) { char level; struct symbol *sym = NULL; - struct thread *thread; u64 ip = event->ip.ip; u64 period = 1; struct map *map = NULL; void *more_data = event->ip.__more_data; struct ip_callchain *chain = NULL; int cpumode; - - thread = threads__findnew(event->ip.pid, &threads, &last_match); + struct thread *thread = threads__findnew(event->ip.pid); if (sample_type & PERF_SAMPLE_PERIOD) { period = *(u64 *)more_data; @@ -685,10 +680,8 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) static int process_mmap_event(event_t *event, unsigned long offset, unsigned long head) { - struct thread *thread; struct map *map = map__new(&event->mmap, cwd, cwdlen); - - thread = threads__findnew(event->mmap.pid, &threads, &last_match); + struct thread *thread = threads__findnew(event->mmap.pid); dump_printf("%p [%p]: PERF_RECORD_MMAP %d/%d: [%p(%p) @ %p]: %s\n", (void *)(offset + head), @@ -714,9 +707,7 @@ process_mmap_event(event_t *event, unsigned long offset, unsigned long head) static int process_comm_event(event_t *event, unsigned long offset, unsigned long head) { - struct thread *thread; - - thread = threads__findnew(event->comm.pid, &threads, &last_match); + struct thread *thread = threads__findnew(event->comm.pid); dump_printf("%p [%p]: PERF_RECORD_COMM: %s:%d\n", (void *)(offset + head), @@ -736,11 +727,8 @@ process_comm_event(event_t *event, unsigned long offset, unsigned long head) static int process_task_event(event_t *event, unsigned long offset, unsigned long head) { - struct thread *thread; - struct thread *parent; - - thread = threads__findnew(event->fork.pid, &threads, &last_match); - parent = threads__findnew(event->fork.ppid, &threads, &last_match); + struct thread *thread = threads__findnew(event->fork.pid); + struct thread *parent = threads__findnew(event->fork.ppid); dump_printf("%p [%p]: PERF_RECORD_%s: (%d:%d):(%d:%d)\n", (void *)(offset + head), @@ -857,7 +845,7 @@ static int __cmd_report(void) struct thread *idle; int ret; - idle = register_idle_thread(&threads, &last_match); + idle = register_idle_thread(); thread__comm_adjust(idle); if (show_threads) @@ -881,7 +869,7 @@ static int __cmd_report(void) return 0; if (verbose > 3) - threads__fprintf(stdout, &threads); + threads__fprintf(stdout); if (verbose > 2) dsos__fprintf(stdout); diff --git a/tools/perf/builtin-sched.c b/tools/perf/builtin-sched.c index 387a4423436..73bdad02973 100644 --- a/tools/perf/builtin-sched.c +++ b/tools/perf/builtin-sched.c @@ -24,9 +24,6 @@ static char const *input_name = "perf.data"; static unsigned long total_comm = 0; -static struct rb_root threads; -static struct thread *last_match; - static struct perf_header *header; static u64 sample_type; @@ -641,9 +638,7 @@ static void test_calibrations(void) static int process_comm_event(event_t *event, unsigned long offset, unsigned long head) { - struct thread *thread; - - thread = threads__findnew(event->comm.tid, &threads, &last_match); + struct thread *thread = threads__findnew(event->comm.tid); dump_printf("%p [%p]: perf_event_comm: %s:%d\n", (void *)(offset + head), @@ -1086,8 +1081,8 @@ latency_switch_event(struct trace_switch_event *switch_event, die("hm, delta: %Ld < 0 ?\n", delta); - sched_out = threads__findnew(switch_event->prev_pid, &threads, &last_match); - sched_in = threads__findnew(switch_event->next_pid, &threads, &last_match); + sched_out = threads__findnew(switch_event->prev_pid); + sched_in = threads__findnew(switch_event->next_pid); out_events = thread_atoms_search(&atom_root, sched_out, &cmp_pid); if (!out_events) { @@ -1120,13 +1115,10 @@ latency_runtime_event(struct trace_runtime_event *runtime_event, u64 timestamp, struct thread *this_thread __used) { - struct work_atoms *atoms; - struct thread *thread; + struct thread *thread = threads__findnew(runtime_event->pid); + struct work_atoms *atoms = thread_atoms_search(&atom_root, thread, &cmp_pid); BUG_ON(cpu >= MAX_CPUS || cpu < 0); - - thread = threads__findnew(runtime_event->pid, &threads, &last_match); - atoms = thread_atoms_search(&atom_root, thread, &cmp_pid); if (!atoms) { thread_atoms_insert(thread); atoms = thread_atoms_search(&atom_root, thread, &cmp_pid); @@ -1153,7 +1145,7 @@ latency_wakeup_event(struct trace_wakeup_event *wakeup_event, if (!wakeup_event->success) return; - wakee = threads__findnew(wakeup_event->pid, &threads, &last_match); + wakee = threads__findnew(wakeup_event->pid); atoms = thread_atoms_search(&atom_root, wakee, &cmp_pid); if (!atoms) { thread_atoms_insert(wakee); @@ -1202,7 +1194,7 @@ latency_migrate_task_event(struct trace_migrate_task_event *migrate_task_event, if (profile_cpu == -1) return; - migrant = threads__findnew(migrate_task_event->pid, &threads, &last_match); + migrant = threads__findnew(migrate_task_event->pid); atoms = thread_atoms_search(&atom_root, migrant, &cmp_pid); if (!atoms) { thread_atoms_insert(migrant); @@ -1458,8 +1450,8 @@ map_switch_event(struct trace_switch_event *switch_event, die("hm, delta: %Ld < 0 ?\n", delta); - sched_out = threads__findnew(switch_event->prev_pid, &threads, &last_match); - sched_in = threads__findnew(switch_event->next_pid, &threads, &last_match); + sched_out = threads__findnew(switch_event->prev_pid); + sched_in = threads__findnew(switch_event->next_pid); curr_thread[this_cpu] = sched_in; @@ -1649,7 +1641,7 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) if (!(sample_type & PERF_SAMPLE_RAW)) return 0; - thread = threads__findnew(event->ip.pid, &threads, &last_match); + thread = threads__findnew(event->ip.pid); if (sample_type & PERF_SAMPLE_TIME) { timestamp = *(u64 *)more_data; @@ -1725,7 +1717,7 @@ static struct perf_file_handler file_handler = { static int read_events(void) { - register_idle_thread(&threads, &last_match); + register_idle_thread(); register_perf_file_handler(&file_handler); return mmap_dispatch_perf_file(&header, input_name, 0, 0, &cwdlen, &cwd); diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-trace.c index fb3f3c22021..ccf867dbab5 100644 --- a/tools/perf/builtin-trace.c +++ b/tools/perf/builtin-trace.c @@ -19,9 +19,6 @@ static char const *input_name = "perf.data"; static unsigned long total = 0; static unsigned long total_comm = 0; -static struct rb_root threads; -static struct thread *last_match; - static struct perf_header *header; static u64 sample_type; @@ -32,9 +29,7 @@ static int cwdlen; static int process_comm_event(event_t *event, unsigned long offset, unsigned long head) { - struct thread *thread; - - thread = threads__findnew(event->comm.pid, &threads, &last_match); + struct thread *thread = threads__findnew(event->comm.pid); dump_printf("%p [%p]: PERF_RECORD_COMM: %s:%d\n", (void *)(offset + head), @@ -54,14 +49,12 @@ process_comm_event(event_t *event, unsigned long offset, unsigned long head) static int process_sample_event(event_t *event, unsigned long offset, unsigned long head) { - struct thread *thread; u64 ip = event->ip.ip; u64 timestamp = -1; u32 cpu = -1; u64 period = 1; void *more_data = event->ip.__more_data; - - thread = threads__findnew(event->ip.pid, &threads, &last_match); + struct thread *thread = threads__findnew(event->ip.pid); if (sample_type & PERF_SAMPLE_TIME) { timestamp = *(u64 *)more_data; @@ -135,7 +128,7 @@ static struct perf_file_handler file_handler = { static int __cmd_trace(void) { - register_idle_thread(&threads, &last_match); + register_idle_thread(); register_perf_file_handler(&file_handler); return mmap_dispatch_perf_file(&header, input_name, 0, 0, &cwdlen, &cwd); diff --git a/tools/perf/util/thread.c b/tools/perf/util/thread.c index 3b56aebb1f4..f53fad7c0a8 100644 --- a/tools/perf/util/thread.c +++ b/tools/perf/util/thread.c @@ -6,6 +6,9 @@ #include "util.h" #include "debug.h" +static struct rb_root threads; +static struct thread *last_match; + static struct thread *thread__new(pid_t pid) { struct thread *self = calloc(1, sizeof(*self)); @@ -50,10 +53,9 @@ static size_t thread__fprintf(struct thread *self, FILE *fp) return ret; } -struct thread * -threads__findnew(pid_t pid, struct rb_root *threads, struct thread **last_match) +struct thread *threads__findnew(pid_t pid) { - struct rb_node **p = &threads->rb_node; + struct rb_node **p = &threads.rb_node; struct rb_node *parent = NULL; struct thread *th; @@ -62,15 +64,15 @@ threads__findnew(pid_t pid, struct rb_root *threads, struct thread **last_match) * so most of the time we dont have to look up * the full rbtree: */ - if (*last_match && (*last_match)->pid == pid) - return *last_match; + if (last_match && last_match->pid == pid) + return last_match; while (*p != NULL) { parent = *p; th = rb_entry(parent, struct thread, rb_node); if (th->pid == pid) { - *last_match = th; + last_match = th; return th; } @@ -83,17 +85,16 @@ threads__findnew(pid_t pid, struct rb_root *threads, struct thread **last_match) th = thread__new(pid); if (th != NULL) { rb_link_node(&th->rb_node, parent, p); - rb_insert_color(&th->rb_node, threads); - *last_match = th; + rb_insert_color(&th->rb_node, &threads); + last_match = th; } return th; } -struct thread * -register_idle_thread(struct rb_root *threads, struct thread **last_match) +struct thread *register_idle_thread(void) { - struct thread *thread = threads__findnew(0, threads, last_match); + struct thread *thread = threads__findnew(0); if (!thread || thread__set_comm(thread, "swapper")) { fprintf(stderr, "problem inserting idle task.\n"); @@ -197,12 +198,12 @@ int thread__fork(struct thread *self, struct thread *parent) return 0; } -size_t threads__fprintf(FILE *fp, struct rb_root *threads) +size_t threads__fprintf(FILE *fp) { size_t ret = 0; struct rb_node *nd; - for (nd = rb_first(threads); nd; nd = rb_next(nd)) { + for (nd = rb_first(&threads); nd; nd = rb_next(nd)) { struct thread *pos = rb_entry(nd, struct thread, rb_node); ret += thread__fprintf(pos, fp); diff --git a/tools/perf/util/thread.h b/tools/perf/util/thread.h index 845d9b62f96..1abef3b7455 100644 --- a/tools/perf/util/thread.h +++ b/tools/perf/util/thread.h @@ -15,13 +15,11 @@ struct thread { }; int thread__set_comm(struct thread *self, const char *comm); -struct thread * -threads__findnew(pid_t pid, struct rb_root *threads, struct thread **last_match); -struct thread * -register_idle_thread(struct rb_root *threads, struct thread **last_match); +struct thread *threads__findnew(pid_t pid); +struct thread *register_idle_thread(void); void thread__insert_map(struct thread *self, struct map *map); int thread__fork(struct thread *self, struct thread *parent); -size_t threads__fprintf(FILE *fp, struct rb_root *threads); +size_t threads__fprintf(FILE *fp); void maps__insert(struct rb_root *maps, struct map *map); struct map *maps__find(struct rb_root *maps, u64 ip); -- cgit v1.2.3-18-g5258 From 924a79af2cdee26a034b9bdce8c9c76995b5c901 Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Wed, 14 Oct 2009 15:43:32 -0400 Subject: perf tools: Handle print concatenations in event format file kmem_alloc ftrace event format had a string that was broken up by two tokens. "string 1" "string 2". This patch lets the parser be able to handle the concatenation. Signed-off-by: Steven Rostedt Cc: Peter Zijlstra Cc: Frederic Weisbecker Cc: Arnaldo Carvalho de Melo LKML-Reference: <20091014194357.253818714@goodmis.org> Signed-off-by: Ingo Molnar --- tools/perf/util/trace-event-parse.c | 16 ++++++++++++++++ 1 file changed, 16 insertions(+) (limited to 'tools/perf') diff --git a/tools/perf/util/trace-event-parse.c b/tools/perf/util/trace-event-parse.c index eef60df7a5b..a05c7144ade 100644 --- a/tools/perf/util/trace-event-parse.c +++ b/tools/perf/util/trace-event-parse.c @@ -1734,6 +1734,7 @@ static int event_read_print(struct event *event) if (read_expect_type(EVENT_DQUOTE, &token) < 0) goto fail; + concat: event->print_fmt.format = token; event->print_fmt.args = NULL; @@ -1743,6 +1744,21 @@ static int event_read_print(struct event *event) if (type == EVENT_NONE) return 0; + /* Handle concatination of print lines */ + if (type == EVENT_DQUOTE) { + char *cat; + + cat = malloc_or_die(strlen(event->print_fmt.format) + + strlen(token) + 1); + strcpy(cat, event->print_fmt.format); + strcat(cat, token); + free_token(token); + free_token(event->print_fmt.format); + event->print_fmt.format = NULL; + token = cat; + goto concat; + } + if (test_type_token(type, token, EVENT_DELIM, (char *)",")) goto fail; -- cgit v1.2.3-18-g5258 From 91ff2bc191827f0d3f5ad0a433ff7df7d2dd9aee Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Wed, 14 Oct 2009 15:43:33 -0400 Subject: perf tools: Fix backslash processing on trace print formats The handling of backslashes was broken. It would stop parsing when encountering one. Also, '\n', '\t', '\r' and '\\' were not converted. Signed-off-by: Steven Rostedt Cc: Peter Zijlstra Cc: Frederic Weisbecker Cc: Arnaldo Carvalho de Melo LKML-Reference: <20091014194357.521974680@goodmis.org> Signed-off-by: Ingo Molnar --- tools/perf/util/trace-event-parse.c | 27 +++++++++++++++++++++++++-- 1 file changed, 25 insertions(+), 2 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/trace-event-parse.c b/tools/perf/util/trace-event-parse.c index a05c7144ade..2b75ec2f57e 100644 --- a/tools/perf/util/trace-event-parse.c +++ b/tools/perf/util/trace-event-parse.c @@ -522,7 +522,10 @@ static enum event_type __read_token(char **tok) last_ch = ch; ch = __read_char(); buf[i++] = ch; - } while (ch != quote_ch && last_ch != '\\'); + /* the '\' '\' will cancel itself */ + if (ch == '\\' && last_ch == '\\') + last_ch = 0; + } while (ch != quote_ch || last_ch == '\\'); /* remove the last quote */ i--; goto out; @@ -2325,7 +2328,27 @@ static void pretty_print(void *data, int size, struct event *event) for (; *ptr; ptr++) { ls = 0; - if (*ptr == '%') { + if (*ptr == '\\') { + ptr++; + switch (*ptr) { + case 'n': + printf("\n"); + break; + case 't': + printf("\t"); + break; + case 'r': + printf("\r"); + break; + case '\\': + printf("\\"); + break; + default: + printf("%c", *ptr); + break; + } + + } else if (*ptr == '%') { saveptr = ptr; show_func = 0; cont_process: -- cgit v1.2.3-18-g5258 From 298ebc3ef2a6c569b3eb51651f04e26aecbf8a1d Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Wed, 14 Oct 2009 15:43:34 -0400 Subject: perf tools: Handle trace parsing of < and > The code to handle the '<' and '>' ops was all in place, but they were not in the switch statement to consider them as valid ops. Signed-off-by: Steven Rostedt Cc: Peter Zijlstra Cc: Frederic Weisbecker Cc: Arnaldo Carvalho de Melo LKML-Reference: <20091014194357.807434040@goodmis.org> Signed-off-by: Ingo Molnar --- tools/perf/util/trace-event-parse.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'tools/perf') diff --git a/tools/perf/util/trace-event-parse.c b/tools/perf/util/trace-event-parse.c index 2b75ec2f57e..3e643f5da20 100644 --- a/tools/perf/util/trace-event-parse.c +++ b/tools/perf/util/trace-event-parse.c @@ -1170,6 +1170,8 @@ process_op(struct event *event, struct print_arg *arg, char **tok) strcmp(token, "*") == 0 || strcmp(token, "^") == 0 || strcmp(token, "/") == 0 || + strcmp(token, "<") == 0 || + strcmp(token, ">") == 0 || strcmp(token, "==") == 0 || strcmp(token, "!=") == 0) { -- cgit v1.2.3-18-g5258 From 0959b8d65ce26131c2d5ccfa518a7b76529280fa Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Wed, 14 Oct 2009 15:43:35 -0400 Subject: perf tools: Handle arrays in print fields for trace parsing The array used by the ftrace stack events (caller[x]) causes issues with the parser. This adds code to handle the case, but it also assumes that the array is of type long. Note, this is a special case used (currently) only by the ftrace user and kernel stack records. Signed-off-by: Steven Rostedt Cc: Peter Zijlstra Cc: Frederic Weisbecker Cc: Arnaldo Carvalho de Melo LKML-Reference: <20091014194358.124833639@goodmis.org> Signed-off-by: Ingo Molnar --- tools/perf/util/trace-event-parse.c | 62 +++++++++++++++++++++++++++++++++++++ 1 file changed, 62 insertions(+) (limited to 'tools/perf') diff --git a/tools/perf/util/trace-event-parse.c b/tools/perf/util/trace-event-parse.c index 3e643f5da20..7aeedb09ea7 100644 --- a/tools/perf/util/trace-event-parse.c +++ b/tools/perf/util/trace-event-parse.c @@ -1046,6 +1046,35 @@ out_free: return EVENT_ERROR; } +static enum event_type +process_array(struct event *event, struct print_arg *top, char **tok) +{ + struct print_arg *arg; + enum event_type type; + char *token = NULL; + + arg = malloc_or_die(sizeof(*arg)); + memset(arg, 0, sizeof(*arg)); + + *tok = NULL; + type = process_arg(event, arg, &token); + if (test_type_token(type, token, EVENT_OP, (char *)"]")) + goto out_free; + + top->op.right = arg; + + free_token(token); + type = read_token_item(&token); + *tok = token; + + return type; + +out_free: + free_token(*tok); + free_arg(arg); + return EVENT_ERROR; +} + static int get_op_prio(char *op) { if (!op[1]) { @@ -1192,6 +1221,18 @@ process_op(struct event *event, struct print_arg *arg, char **tok) arg->op.right = right; + } else if (strcmp(token, "[") == 0) { + + left = malloc_or_die(sizeof(*left)); + *left = *arg; + + arg->type = PRINT_OP; + arg->op.op = token; + arg->op.left = left; + + arg->op.prio = 0; + type = process_array(event, arg, tok); + } else { die("unknown op '%s'", token); /* the arg is now the left side */ @@ -1931,6 +1972,7 @@ static unsigned long long eval_num_arg(void *data, int size, { unsigned long long val = 0; unsigned long long left, right; + struct print_arg *larg; switch (arg->type) { case PRINT_NULL: @@ -1957,6 +1999,26 @@ static unsigned long long eval_num_arg(void *data, int size, return 0; break; case PRINT_OP: + if (strcmp(arg->op.op, "[") == 0) { + /* + * Arrays are special, since we don't want + * to read the arg as is. + */ + if (arg->op.left->type != PRINT_FIELD) + goto default_op; /* oops, all bets off */ + larg = arg->op.left; + if (!larg->field.field) { + larg->field.field = + find_any_field(event, larg->field.name); + if (!larg->field.field) + die("field %s not found", larg->field.name); + } + right = eval_num_arg(data, size, event, arg->op.right); + val = read_size(data + larg->field.field->offset + + right * long_size, long_size); + break; + } + default_op: left = eval_num_arg(data, size, event, arg->op.left); right = eval_num_arg(data, size, event, arg->op.right); switch (arg->op.op[0]) { -- cgit v1.2.3-18-g5258 From b99af874829cba2b30d212bc6fd31b56275ee4d2 Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Wed, 14 Oct 2009 15:43:36 -0400 Subject: perf tools: Handle * as typecast in trace parsing The '*' is currently only treated as a multiplication, and it needs to be handled as a typecast pointer. This is the version used by trace-cmd. Signed-off-by: Steven Rostedt Cc: Peter Zijlstra Cc: Frederic Weisbecker Cc: Arnaldo Carvalho de Melo LKML-Reference: <20091014194358.409327875@goodmis.org> Signed-off-by: Ingo Molnar --- tools/perf/util/trace-event-parse.c | 50 ++++++++++++++++--------------------- 1 file changed, 22 insertions(+), 28 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/trace-event-parse.c b/tools/perf/util/trace-event-parse.c index 7aeedb09ea7..f73ee55b51e 100644 --- a/tools/perf/util/trace-event-parse.c +++ b/tools/perf/util/trace-event-parse.c @@ -1217,7 +1217,24 @@ process_op(struct event *event, struct print_arg *arg, char **tok) right = malloc_or_die(sizeof(*right)); - type = process_arg(event, right, tok); + type = read_token_item(&token); + *tok = token; + + /* could just be a type pointer */ + if ((strcmp(arg->op.op, "*") == 0) && + type == EVENT_DELIM && (strcmp(token, ")") == 0)) { + if (left->type != PRINT_ATOM) + die("bad pointer type"); + left->atom.atom = realloc(left->atom.atom, + sizeof(left->atom.atom) + 3); + strcat(left->atom.atom, " *"); + *arg = *left; + free(arg); + + return type; + } + + type = process_arg_token(event, right, tok, type); arg->op.right = right; @@ -1548,7 +1565,6 @@ process_paren(struct event *event, struct print_arg *arg, char **tok) { struct print_arg *item_arg; enum event_type type; - int ptr_cast = 0; char *token; type = process_arg(event, arg, &token); @@ -1556,26 +1572,11 @@ process_paren(struct event *event, struct print_arg *arg, char **tok) if (type == EVENT_ERROR) return EVENT_ERROR; - if (type == EVENT_OP) { - /* handle the ptr casts */ - if (!strcmp(token, "*")) { - /* - * FIXME: should we zapp whitespaces before ')' ? - * (may require a peek_token_item()) - */ - if (__peek_char() == ')') { - ptr_cast = 1; - free_token(token); - type = read_token_item(&token); - } - } - if (!ptr_cast) { - type = process_op(event, arg, &token); + if (type == EVENT_OP) + type = process_op(event, arg, &token); - if (type == EVENT_ERROR) - return EVENT_ERROR; - } - } + if (type == EVENT_ERROR) + return EVENT_ERROR; if (test_type_token(type, token, EVENT_DELIM, (char *)")")) { free_token(token); @@ -1601,13 +1602,6 @@ process_paren(struct event *event, struct print_arg *arg, char **tok) item_arg = malloc_or_die(sizeof(*item_arg)); arg->type = PRINT_TYPE; - if (ptr_cast) { - char *old = arg->atom.atom; - - arg->atom.atom = malloc_or_die(strlen(old + 3)); - sprintf(arg->atom.atom, "%s *", old); - free(old); - } arg->typecast.type = arg->atom.atom; arg->typecast.item = item_arg; type = process_arg_token(event, item_arg, &token, type); -- cgit v1.2.3-18-g5258 From f1d1feecf07261d083859ecfef0d4399036f9683 Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Wed, 14 Oct 2009 15:43:37 -0400 Subject: perf tools: Handle newlines in trace parsing better New lines between args in the trace format can break the parsing. This should not be the case. Signed-off-by: Steven Rostedt Cc: Peter Zijlstra Cc: Frederic Weisbecker Cc: Arnaldo Carvalho de Melo LKML-Reference: <20091014194358.637991808@goodmis.org> Signed-off-by: Ingo Molnar --- tools/perf/util/trace-event-parse.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-) (limited to 'tools/perf') diff --git a/tools/perf/util/trace-event-parse.c b/tools/perf/util/trace-event-parse.c index f73ee55b51e..59e4e4db743 100644 --- a/tools/perf/util/trace-event-parse.c +++ b/tools/perf/util/trace-event-parse.c @@ -1716,12 +1716,18 @@ process_arg_token(struct event *event, struct print_arg *arg, static int event_read_print_args(struct event *event, struct print_arg **list) { - enum event_type type; + enum event_type type = EVENT_ERROR; struct print_arg *arg; char *token; int args = 0; do { + if (type == EVENT_NEWLINE) { + free_token(token); + type = read_token_item(&token); + continue; + } + arg = malloc_or_die(sizeof(*arg)); memset(arg, 0, sizeof(*arg)); -- cgit v1.2.3-18-g5258 From 13999e59343b042b0807be2df6ae5895d29782a0 Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Wed, 14 Oct 2009 15:43:38 -0400 Subject: perf tools: Handle the case with and without the "signed" trace field The trace format files now have a "signed" field. But we should still be able to handle the kernels that do not have this field. Signed-off-by: Steven Rostedt Cc: Peter Zijlstra Cc: Frederic Weisbecker Cc: Arnaldo Carvalho de Melo LKML-Reference: <20091014194358.888239553@goodmis.org> Signed-off-by: Ingo Molnar --- tools/perf/util/trace-event-parse.c | 81 ++++++++++++++++++++++++------------- 1 file changed, 52 insertions(+), 29 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/trace-event-parse.c b/tools/perf/util/trace-event-parse.c index 59e4e4db743..0739b12675f 100644 --- a/tools/perf/util/trace-event-parse.c +++ b/tools/perf/util/trace-event-parse.c @@ -924,23 +924,30 @@ static int event_read_fields(struct event *event, struct format_field **fields) if (read_expected(EVENT_OP, (char *)";") < 0) goto fail_expect; - if (read_expected(EVENT_ITEM, (char *)"signed") < 0) - goto fail_expect; + type = read_token(&token); + if (type != EVENT_NEWLINE) { + /* newer versions of the kernel have a "signed" type */ + if (test_type_token(type, token, EVENT_ITEM, (char *)"signed")) + goto fail; - if (read_expected(EVENT_OP, (char *)":") < 0) - goto fail_expect; + free_token(token); - if (read_expect_type(EVENT_ITEM, &token)) - goto fail; - if (strtoul(token, NULL, 0)) - field->flags |= FIELD_IS_SIGNED; - free_token(token); + if (read_expected(EVENT_OP, (char *)":") < 0) + goto fail_expect; - if (read_expected(EVENT_OP, (char *)";") < 0) - goto fail_expect; + if (read_expect_type(EVENT_ITEM, &token)) + goto fail; + + /* add signed type */ + + free_token(token); + if (read_expected(EVENT_OP, (char *)";") < 0) + goto fail_expect; + + if (read_expect_type(EVENT_NEWLINE, &token)) + goto fail; + } - if (read_expect_type(EVENT_NEWLINE, &token) < 0) - goto fail; free_token(token); *fields = field; @@ -2949,21 +2956,23 @@ static void print_args(struct print_arg *args) } } -static void parse_header_field(char *type, +static void parse_header_field(char *field, int *offset, int *size) { char *token; + int type; if (read_expected(EVENT_ITEM, (char *)"field") < 0) return; if (read_expected(EVENT_OP, (char *)":") < 0) return; + /* type */ if (read_expect_type(EVENT_ITEM, &token) < 0) - return; + goto fail; free_token(token); - if (read_expected(EVENT_ITEM, type) < 0) + if (read_expected(EVENT_ITEM, field) < 0) return; if (read_expected(EVENT_OP, (char *)";") < 0) return; @@ -2972,7 +2981,7 @@ static void parse_header_field(char *type, if (read_expected(EVENT_OP, (char *)":") < 0) return; if (read_expect_type(EVENT_ITEM, &token) < 0) - return; + goto fail; *offset = atoi(token); free_token(token); if (read_expected(EVENT_OP, (char *)";") < 0) @@ -2982,22 +2991,36 @@ static void parse_header_field(char *type, if (read_expected(EVENT_OP, (char *)":") < 0) return; if (read_expect_type(EVENT_ITEM, &token) < 0) - return; + goto fail; *size = atoi(token); free_token(token); if (read_expected(EVENT_OP, (char *)";") < 0) return; - if (read_expected(EVENT_ITEM, (char *)"signed") < 0) - return; - if (read_expected(EVENT_OP, (char *)":") < 0) - return; - if (read_expect_type(EVENT_ITEM, &token) < 0) - return; - free_token(token); - if (read_expected(EVENT_OP, (char *)";") < 0) - return; - if (read_expect_type(EVENT_NEWLINE, &token) < 0) - return; + type = read_token(&token); + if (type != EVENT_NEWLINE) { + /* newer versions of the kernel have a "signed" type */ + if (type != EVENT_ITEM) + goto fail; + + if (strcmp(token, (char *)"signed") != 0) + goto fail; + + free_token(token); + + if (read_expected(EVENT_OP, (char *)":") < 0) + return; + + if (read_expect_type(EVENT_ITEM, &token)) + goto fail; + + free_token(token); + if (read_expected(EVENT_OP, (char *)";") < 0) + return; + + if (read_expect_type(EVENT_NEWLINE, &token)) + goto fail; + } + fail: free_token(token); } -- cgit v1.2.3-18-g5258 From 07a4bdddcf2546ccfbfb3c782deab636c371edeb Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Wed, 14 Oct 2009 15:43:39 -0400 Subject: perf tools: Still continue on failed parsing of an event Even though an event may fail to parse, we should not kill the entire report. The trace should still be able to show what it can. If an event fails to parse, a warning is printed, and the output continues. Signed-off-by: Steven Rostedt Cc: Peter Zijlstra Cc: Frederic Weisbecker Cc: Arnaldo Carvalho de Melo LKML-Reference: <20091014194359.190809589@goodmis.org> Signed-off-by: Ingo Molnar --- tools/perf/util/trace-event-parse.c | 38 ++++++++++++++++++++++++++----------- tools/perf/util/trace-event.h | 14 ++++++++------ 2 files changed, 35 insertions(+), 17 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/trace-event-parse.c b/tools/perf/util/trace-event-parse.c index 0739b12675f..eda0a2488c1 100644 --- a/tools/perf/util/trace-event-parse.c +++ b/tools/perf/util/trace-event-parse.c @@ -613,7 +613,7 @@ static enum event_type read_token_item(char **tok) static int test_type(enum event_type type, enum event_type expect) { if (type != expect) { - die("Error: expected type %d but read %d", + warning("Error: expected type %d but read %d", expect, type); return -1; } @@ -624,13 +624,13 @@ static int test_type_token(enum event_type type, char *token, enum event_type expect, const char *expect_tok) { if (type != expect) { - die("Error: expected type %d but read %d", + warning("Error: expected type %d but read %d", expect, type); return -1; } if (strcmp(token, expect_tok) != 0) { - die("Error: expected '%s' but read '%s'", + warning("Error: expected '%s' but read '%s'", expect_tok, token); return -1; } @@ -668,7 +668,7 @@ static int __read_expected(enum event_type expect, const char *str, int newline_ free_token(token); - return 0; + return ret; } static int read_expected(enum event_type expect, const char *str) @@ -1258,12 +1258,12 @@ process_op(struct event *event, struct print_arg *arg, char **tok) type = process_array(event, arg, tok); } else { - die("unknown op '%s'", token); + warning("unknown op '%s'", token); + event->flags |= EVENT_FL_FAILED; /* the arg is now the left side */ return EVENT_NONE; } - if (type == EVENT_OP) { int prio; @@ -2873,7 +2873,7 @@ void print_event(int cpu, void *data, int size, unsigned long long nsecs, event = trace_find_event(type); if (!event) { - printf("ug! no event found for type %d\n", type); + warning("ug! no event found for type %d", type); return; } @@ -2887,6 +2887,12 @@ void print_event(int cpu, void *data, int size, unsigned long long nsecs, comm, pid, cpu, secs, nsecs, event->name); + if (event->flags & EVENT_FL_FAILED) { + printf("EVENT '%s' FAILED TO PARSE\n", + event->name); + return; + } + pretty_print(data, size, event); printf("\n"); } @@ -3120,12 +3126,16 @@ int parse_event_file(char *buf, unsigned long size, char *sys) die("failed to read event id"); ret = event_read_format(event); - if (ret < 0) - die("failed to read event format"); + if (ret < 0) { + warning("failed to read event format for %s", event->name); + goto event_failed; + } ret = event_read_print(event); - if (ret < 0) - die("failed to read event print fmt"); + if (ret < 0) { + warning("failed to read event print fmt for %s", event->name); + goto event_failed; + } event->system = strdup(sys); @@ -3135,6 +3145,12 @@ int parse_event_file(char *buf, unsigned long size, char *sys) add_event(event); return 0; + + event_failed: + event->flags |= EVENT_FL_FAILED; + /* still add it even if it failed */ + add_event(event); + return -1; } void parse_set_info(int nr_cpus, int long_sz) diff --git a/tools/perf/util/trace-event.h b/tools/perf/util/trace-event.h index da77e073c86..29821acc8db 100644 --- a/tools/perf/util/trace-event.h +++ b/tools/perf/util/trace-event.h @@ -139,12 +139,14 @@ struct event { }; enum { - EVENT_FL_ISFTRACE = 1, - EVENT_FL_ISPRINT = 2, - EVENT_FL_ISBPRINT = 4, - EVENT_FL_ISFUNC = 8, - EVENT_FL_ISFUNCENT = 16, - EVENT_FL_ISFUNCRET = 32, + EVENT_FL_ISFTRACE = 0x01, + EVENT_FL_ISPRINT = 0x02, + EVENT_FL_ISBPRINT = 0x04, + EVENT_FL_ISFUNC = 0x08, + EVENT_FL_ISFUNCENT = 0x10, + EVENT_FL_ISFUNCRET = 0x20, + + EVENT_FL_FAILED = 0x80000000 }; struct record { -- cgit v1.2.3-18-g5258 From ffa1895561645103d8f8059b35d9c06e6eeead2e Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Wed, 14 Oct 2009 15:43:40 -0400 Subject: perf tools: Fix bprintk reading in trace output The bprintk parsing was broken in more ways than one. The file parsing was incorrect, and the words used by the arguments are always 4 bytes aligned, even on 64-bit machines. Signed-off-by: Steven Rostedt Cc: Peter Zijlstra Cc: Frederic Weisbecker Cc: Arnaldo Carvalho de Melo LKML-Reference: <20091014194359.520931637@goodmis.org> Signed-off-by: Ingo Molnar --- tools/perf/util/trace-event-parse.c | 15 +++++++-------- 1 file changed, 7 insertions(+), 8 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/trace-event-parse.c b/tools/perf/util/trace-event-parse.c index eda0a2488c1..93a82fead95 100644 --- a/tools/perf/util/trace-event-parse.c +++ b/tools/perf/util/trace-event-parse.c @@ -284,18 +284,16 @@ void parse_ftrace_printk(char *file, unsigned int size __unused) char *line; char *next = NULL; char *addr_str; - int ret; + char *fmt; int i; line = strtok_r(file, "\n", &next); while (line) { item = malloc_or_die(sizeof(*item)); - ret = sscanf(line, "%as : %as", - (float *)(void *)&addr_str, /* workaround gcc warning */ - (float *)(void *)&item->printk); + addr_str = strtok_r(line, ":", &fmt); item->addr = strtoull(addr_str, NULL, 16); - free(addr_str); - + /* fmt still has a space, skip it */ + item->printk = strdup(fmt+1); item->next = list; list = item; line = strtok_r(NULL, "\n", &next); @@ -2274,8 +2272,9 @@ static struct print_arg *make_bprint_args(char *fmt, void *data, int size, struc case 'u': case 'x': case 'i': - bptr = (void *)(((unsigned long)bptr + (long_size - 1)) & - ~(long_size - 1)); + /* the pointers are always 4 bytes aligned */ + bptr = (void *)(((unsigned long)bptr + 3) & + ~3); switch (ls) { case 0: case 1: -- cgit v1.2.3-18-g5258 From 0d1da915c76838c9ee7af7cdefbcb2bae9424161 Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Wed, 14 Oct 2009 15:43:41 -0400 Subject: perf tools: Handle both versions of ftrace output The ftrace output events can have either arguments or no arguments. The parser needs to be able to handle both. Signed-off-by: Steven Rostedt Cc: Peter Zijlstra Cc: Frederic Weisbecker Cc: Arnaldo Carvalho de Melo LKML-Reference: <20091014194359.790221427@goodmis.org> Signed-off-by: Ingo Molnar --- tools/perf/util/trace-event-parse.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'tools/perf') diff --git a/tools/perf/util/trace-event-parse.c b/tools/perf/util/trace-event-parse.c index 93a82fead95..c174765d405 100644 --- a/tools/perf/util/trace-event-parse.c +++ b/tools/perf/util/trace-event-parse.c @@ -1819,7 +1819,7 @@ static int event_read_print(struct event *event) if (ret < 0) return -1; - return 0; + return ret; fail: free_token(token); @@ -3088,6 +3088,9 @@ int parse_ftrace_file(char *buf, unsigned long size) if (ret < 0) die("failed to read ftrace event print fmt"); + /* New ftrace handles args */ + if (ret > 0) + return 0; /* * The arguments for ftrace files are parsed by the fields. * Set up the fields as their arguments. -- cgit v1.2.3-18-g5258 From cda48461c7fb8431a99b7960480f5f42cc1a5324 Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Wed, 14 Oct 2009 15:43:42 -0400 Subject: perf tools: Add latency format to trace output Add the irqs disabled, preemption count, need resched, and other info that is shown in the latency format of ftrace. # perf trace -l perf-16457 2..s2. 53636.260344: kmem_cache_free: call_site=ffffffff811198f perf-16457 2..s2. 53636.264330: kmem_cache_free: call_site=ffffffff811198f perf-16457 2d.s4. 53636.300006: kmem_cache_free: call_site=ffffffff810d889 Signed-off-by: Steven Rostedt Cc: Peter Zijlstra Cc: Frederic Weisbecker Cc: Arnaldo Carvalho de Melo LKML-Reference: <20091014194400.076588953@goodmis.org> Signed-off-by: Ingo Molnar --- tools/perf/builtin-trace.c | 2 + tools/perf/util/trace-event-parse.c | 120 ++++++++++++++++++++++++++++++------ tools/perf/util/trace-event.h | 11 ++++ 3 files changed, 114 insertions(+), 19 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-trace.c index ccf867dbab5..ce8459ac284 100644 --- a/tools/perf/builtin-trace.c +++ b/tools/perf/builtin-trace.c @@ -144,6 +144,8 @@ static const struct option options[] = { "dump raw trace in ASCII"), OPT_BOOLEAN('v', "verbose", &verbose, "be more verbose (show symbol address, etc)"), + OPT_BOOLEAN('l', "latency", &latency_format, + "show latency attributes (irqs/preemption disabled, etc)"), OPT_END() }; diff --git a/tools/perf/util/trace-event-parse.c b/tools/perf/util/trace-event-parse.c index c174765d405..fde1a434d63 100644 --- a/tools/perf/util/trace-event-parse.c +++ b/tools/perf/util/trace-event-parse.c @@ -40,6 +40,8 @@ int header_page_size_size; int header_page_data_offset; int header_page_data_size; +int latency_format; + static char *input_buf; static unsigned long long input_buf_ptr; static unsigned long long input_buf_siz; @@ -1928,37 +1930,67 @@ static int get_common_info(const char *type, int *offset, int *size) return 0; } -int trace_parse_common_type(void *data) +static int __parse_common(void *data, int *size, int *offset, + char *name) { - static int type_offset; - static int type_size; int ret; - if (!type_size) { - ret = get_common_info("common_type", - &type_offset, - &type_size); + if (!*size) { + ret = get_common_info(name, offset, size); if (ret < 0) return ret; } - return read_size(data + type_offset, type_size); + return read_size(data + *offset, *size); +} + +int trace_parse_common_type(void *data) +{ + static int type_offset; + static int type_size; + + return __parse_common(data, &type_size, &type_offset, + (char *)"common_type"); } static int parse_common_pid(void *data) { static int pid_offset; static int pid_size; + + return __parse_common(data, &pid_size, &pid_offset, + (char *)"common_pid"); +} + +static int parse_common_pc(void *data) +{ + static int pc_offset; + static int pc_size; + + return __parse_common(data, &pc_size, &pc_offset, + (char *)"common_preempt_count"); +} + +static int parse_common_flags(void *data) +{ + static int flags_offset; + static int flags_size; + + return __parse_common(data, &flags_size, &flags_offset, + (char *)"common_flags"); +} + +static int parse_common_lock_depth(void *data) +{ + static int ld_offset; + static int ld_size; int ret; - if (!pid_size) { - ret = get_common_info("common_pid", - &pid_offset, - &pid_size); - if (ret < 0) - return ret; - } + ret = __parse_common(data, &ld_size, &ld_offset, + (char *)"common_lock_depth"); + if (ret < 0) + return -1; - return read_size(data + pid_offset, pid_size); + return ret; } struct event *trace_find_event(int id) @@ -2525,6 +2557,41 @@ static inline int log10_cpu(int nb) return 1; } +static void print_lat_fmt(void *data, int size __unused) +{ + unsigned int lat_flags; + unsigned int pc; + int lock_depth; + int hardirq; + int softirq; + + lat_flags = parse_common_flags(data); + pc = parse_common_pc(data); + lock_depth = parse_common_lock_depth(data); + + hardirq = lat_flags & TRACE_FLAG_HARDIRQ; + softirq = lat_flags & TRACE_FLAG_SOFTIRQ; + + printf("%c%c%c", + (lat_flags & TRACE_FLAG_IRQS_OFF) ? 'd' : + (lat_flags & TRACE_FLAG_IRQS_NOSUPPORT) ? + 'X' : '.', + (lat_flags & TRACE_FLAG_NEED_RESCHED) ? + 'N' : '.', + (hardirq && softirq) ? 'H' : + hardirq ? 'h' : softirq ? 's' : '.'); + + if (pc) + printf("%x", pc); + else + printf("."); + + if (lock_depth < 0) + printf("."); + else + printf("%d", lock_depth); +} + /* taken from Linux, written by Frederic Weisbecker */ static void print_graph_cpu(int cpu) { @@ -2768,6 +2835,11 @@ pretty_print_func_ent(void *data, int size, struct event *event, printf(" | "); + if (latency_format) { + print_lat_fmt(data, size); + printf(" | "); + } + field = find_field(event, "func"); if (!field) die("function entry does not have func field"); @@ -2811,6 +2883,11 @@ pretty_print_func_ret(void *data, int size __unused, struct event *event, printf(" | "); + if (latency_format) { + print_lat_fmt(data, size); + printf(" | "); + } + field = find_field(event, "rettime"); if (!field) die("can't find rettime in return graph"); @@ -2882,9 +2959,14 @@ void print_event(int cpu, void *data, int size, unsigned long long nsecs, return pretty_print_func_graph(data, size, event, cpu, pid, comm, secs, usecs); - printf("%16s-%-5d [%03d] %5lu.%09Lu: %s: ", - comm, pid, cpu, - secs, nsecs, event->name); + if (latency_format) { + printf("%8.8s-%-5d %3d", + comm, pid, cpu); + print_lat_fmt(data, size); + } else + printf("%16s-%-5d [%03d]", comm, pid, cpu); + + printf(" %5lu.%06lu: %s: ", secs, usecs, event->name); if (event->flags & EVENT_FL_FAILED) { printf("EVENT '%s' FAILED TO PARSE\n", diff --git a/tools/perf/util/trace-event.h b/tools/perf/util/trace-event.h index 29821acc8db..f6637c2fa1f 100644 --- a/tools/perf/util/trace-event.h +++ b/tools/perf/util/trace-event.h @@ -239,6 +239,8 @@ extern int header_page_size_size; extern int header_page_data_offset; extern int header_page_data_size; +extern int latency_format; + int parse_header_page(char *buf, unsigned long size); int trace_parse_common_type(void *data); struct event *trace_find_event(int id); @@ -248,4 +250,13 @@ void *raw_field_ptr(struct event *event, const char *name, void *data); void read_tracing_data(int fd, struct perf_event_attr *pattrs, int nb_events); +/* taken from kernel/trace/trace.h */ +enum trace_flag_type { + TRACE_FLAG_IRQS_OFF = 0x01, + TRACE_FLAG_IRQS_NOSUPPORT = 0x02, + TRACE_FLAG_NEED_RESCHED = 0x04, + TRACE_FLAG_HARDIRQ = 0x08, + TRACE_FLAG_SOFTIRQ = 0x10, +}; + #endif /* __PERF_TRACE_EVENTS_H */ -- cgit v1.2.3-18-g5258 From afdf1a404eed236d6f762ee44cc0f1dcc97206e0 Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Wed, 14 Oct 2009 15:43:43 -0400 Subject: perf tools: Handle - and + in parsing trace print format The opterators '-' and '+' are not handled in the trace print format. To do: '++' and '--'. Signed-off-by: Steven Rostedt Cc: Peter Zijlstra Cc: Frederic Weisbecker Cc: Arnaldo Carvalho de Melo LKML-Reference: <20091014194400.330843045@goodmis.org> Signed-off-by: Ingo Molnar --- tools/perf/util/trace-event-parse.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'tools/perf') diff --git a/tools/perf/util/trace-event-parse.c b/tools/perf/util/trace-event-parse.c index fde1a434d63..2d424ff50ec 100644 --- a/tools/perf/util/trace-event-parse.c +++ b/tools/perf/util/trace-event-parse.c @@ -2106,6 +2106,12 @@ static unsigned long long eval_num_arg(void *data, int size, die("unknown op '%s'", arg->op.op); val = left == right; break; + case '-': + val = left - right; + break; + case '+': + val = left + right; + break; default: die("unknown op '%s'", arg->op.op); } -- cgit v1.2.3-18-g5258 From c4dc775f53136cd6af8f88bce67cce9b42751768 Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Wed, 14 Oct 2009 15:43:44 -0400 Subject: perf tools: Remove all char * typecasts and use const in prototype The (char *) for all the static strings was a fix for the symptom and not the disease. The real issue was that the function prototypes needed to be declared "const char *". Signed-off-by: Steven Rostedt Cc: Peter Zijlstra Cc: Frederic Weisbecker Cc: Arnaldo Carvalho de Melo LKML-Reference: <20091014194400.635935008@goodmis.org> Signed-off-by: Ingo Molnar --- tools/perf/util/trace-event-parse.c | 122 ++++++++++++++++++------------------ 1 file changed, 61 insertions(+), 61 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/trace-event-parse.c b/tools/perf/util/trace-event-parse.c index 2d424ff50ec..4b61b497040 100644 --- a/tools/perf/util/trace-event-parse.c +++ b/tools/perf/util/trace-event-parse.c @@ -685,10 +685,10 @@ static char *event_read_name(void) { char *token; - if (read_expected(EVENT_ITEM, (char *)"name") < 0) + if (read_expected(EVENT_ITEM, "name") < 0) return NULL; - if (read_expected(EVENT_OP, (char *)":") < 0) + if (read_expected(EVENT_OP, ":") < 0) return NULL; if (read_expect_type(EVENT_ITEM, &token) < 0) @@ -706,10 +706,10 @@ static int event_read_id(void) char *token; int id; - if (read_expected_item(EVENT_ITEM, (char *)"ID") < 0) + if (read_expected_item(EVENT_ITEM, "ID") < 0) return -1; - if (read_expected(EVENT_OP, (char *)":") < 0) + if (read_expected(EVENT_OP, ":") < 0) return -1; if (read_expect_type(EVENT_ITEM, &token) < 0) @@ -759,7 +759,7 @@ static int event_read_fields(struct event *event, struct format_field **fields) count++; - if (test_type_token(type, token, EVENT_ITEM, (char *)"field")) + if (test_type_token(type, token, EVENT_ITEM, "field")) goto fail; free_token(token); @@ -774,7 +774,7 @@ static int event_read_fields(struct event *event, struct format_field **fields) type = read_token(&token); } - if (test_type_token(type, token, EVENT_OP, (char *)":") < 0) + if (test_type_token(type, token, EVENT_OP, ":") < 0) return -1; if (read_expect_type(EVENT_ITEM, &token) < 0) @@ -892,14 +892,14 @@ static int event_read_fields(struct event *event, struct format_field **fields) field->flags |= FIELD_IS_DYNAMIC; } - if (test_type_token(type, token, EVENT_OP, (char *)";")) + if (test_type_token(type, token, EVENT_OP, ";")) goto fail; free_token(token); - if (read_expected(EVENT_ITEM, (char *)"offset") < 0) + if (read_expected(EVENT_ITEM, "offset") < 0) goto fail_expect; - if (read_expected(EVENT_OP, (char *)":") < 0) + if (read_expected(EVENT_OP, ":") < 0) goto fail_expect; if (read_expect_type(EVENT_ITEM, &token)) @@ -907,13 +907,13 @@ static int event_read_fields(struct event *event, struct format_field **fields) field->offset = strtoul(token, NULL, 0); free_token(token); - if (read_expected(EVENT_OP, (char *)";") < 0) + if (read_expected(EVENT_OP, ";") < 0) goto fail_expect; - if (read_expected(EVENT_ITEM, (char *)"size") < 0) + if (read_expected(EVENT_ITEM, "size") < 0) goto fail_expect; - if (read_expected(EVENT_OP, (char *)":") < 0) + if (read_expected(EVENT_OP, ":") < 0) goto fail_expect; if (read_expect_type(EVENT_ITEM, &token)) @@ -921,18 +921,18 @@ static int event_read_fields(struct event *event, struct format_field **fields) field->size = strtoul(token, NULL, 0); free_token(token); - if (read_expected(EVENT_OP, (char *)";") < 0) + if (read_expected(EVENT_OP, ";") < 0) goto fail_expect; type = read_token(&token); if (type != EVENT_NEWLINE) { /* newer versions of the kernel have a "signed" type */ - if (test_type_token(type, token, EVENT_ITEM, (char *)"signed")) + if (test_type_token(type, token, EVENT_ITEM, "signed")) goto fail; free_token(token); - if (read_expected(EVENT_OP, (char *)":") < 0) + if (read_expected(EVENT_OP, ":") < 0) goto fail_expect; if (read_expect_type(EVENT_ITEM, &token)) @@ -941,7 +941,7 @@ static int event_read_fields(struct event *event, struct format_field **fields) /* add signed type */ free_token(token); - if (read_expected(EVENT_OP, (char *)";") < 0) + if (read_expected(EVENT_OP, ";") < 0) goto fail_expect; if (read_expect_type(EVENT_NEWLINE, &token)) @@ -970,10 +970,10 @@ static int event_read_format(struct event *event) char *token; int ret; - if (read_expected_item(EVENT_ITEM, (char *)"format") < 0) + if (read_expected_item(EVENT_ITEM, "format") < 0) return -1; - if (read_expected(EVENT_OP, (char *)":") < 0) + if (read_expected(EVENT_OP, ":") < 0) return -1; if (read_expect_type(EVENT_NEWLINE, &token)) @@ -1033,7 +1033,7 @@ process_cond(struct event *event, struct print_arg *top, char **tok) *tok = NULL; type = process_arg(event, left, &token); - if (test_type_token(type, token, EVENT_OP, (char *)":")) + if (test_type_token(type, token, EVENT_OP, ":")) goto out_free; arg->op.op = token; @@ -1065,7 +1065,7 @@ process_array(struct event *event, struct print_arg *top, char **tok) *tok = NULL; type = process_arg(event, arg, &token); - if (test_type_token(type, token, EVENT_OP, (char *)"]")) + if (test_type_token(type, token, EVENT_OP, "]")) goto out_free; top->op.right = arg; @@ -1287,7 +1287,7 @@ process_entry(struct event *event __unused, struct print_arg *arg, char *field; char *token; - if (read_expected(EVENT_OP, (char *)"->") < 0) + if (read_expected(EVENT_OP, "->") < 0) return EVENT_ERROR; if (read_expect_type(EVENT_ITEM, &token) < 0) @@ -1447,14 +1447,14 @@ process_fields(struct event *event, struct print_flag_sym **list, char **tok) do { free_token(token); type = read_token_item(&token); - if (test_type_token(type, token, EVENT_OP, (char *)"{")) + if (test_type_token(type, token, EVENT_OP, "{")) break; arg = malloc_or_die(sizeof(*arg)); free_token(token); type = process_arg(event, arg, &token); - if (test_type_token(type, token, EVENT_DELIM, (char *)",")) + if (test_type_token(type, token, EVENT_DELIM, ",")) goto out_free; field = malloc_or_die(sizeof(*field)); @@ -1465,7 +1465,7 @@ process_fields(struct event *event, struct print_flag_sym **list, char **tok) free_token(token); type = process_arg(event, arg, &token); - if (test_type_token(type, token, EVENT_OP, (char *)"}")) + if (test_type_token(type, token, EVENT_OP, "}")) goto out_free; value = arg_eval(arg); @@ -1500,13 +1500,13 @@ process_flags(struct event *event, struct print_arg *arg, char **tok) memset(arg, 0, sizeof(*arg)); arg->type = PRINT_FLAGS; - if (read_expected_item(EVENT_DELIM, (char *)"(") < 0) + if (read_expected_item(EVENT_DELIM, "(") < 0) return EVENT_ERROR; field = malloc_or_die(sizeof(*field)); type = process_arg(event, field, &token); - if (test_type_token(type, token, EVENT_DELIM, (char *)",")) + if (test_type_token(type, token, EVENT_DELIM, ",")) goto out_free; arg->flags.field = field; @@ -1517,11 +1517,11 @@ process_flags(struct event *event, struct print_arg *arg, char **tok) type = read_token_item(&token); } - if (test_type_token(type, token, EVENT_DELIM, (char *)",")) + if (test_type_token(type, token, EVENT_DELIM, ",")) goto out_free; type = process_fields(event, &arg->flags.flags, &token); - if (test_type_token(type, token, EVENT_DELIM, (char *)")")) + if (test_type_token(type, token, EVENT_DELIM, ")")) goto out_free; free_token(token); @@ -1543,19 +1543,19 @@ process_symbols(struct event *event, struct print_arg *arg, char **tok) memset(arg, 0, sizeof(*arg)); arg->type = PRINT_SYMBOL; - if (read_expected_item(EVENT_DELIM, (char *)"(") < 0) + if (read_expected_item(EVENT_DELIM, "(") < 0) return EVENT_ERROR; field = malloc_or_die(sizeof(*field)); type = process_arg(event, field, &token); - if (test_type_token(type, token, EVENT_DELIM, (char *)",")) + if (test_type_token(type, token, EVENT_DELIM, ",")) goto out_free; arg->symbol.field = field; type = process_fields(event, &arg->symbol.symbols, &token); - if (test_type_token(type, token, EVENT_DELIM, (char *)")")) + if (test_type_token(type, token, EVENT_DELIM, ")")) goto out_free; free_token(token); @@ -1585,7 +1585,7 @@ process_paren(struct event *event, struct print_arg *arg, char **tok) if (type == EVENT_ERROR) return EVENT_ERROR; - if (test_type_token(type, token, EVENT_DELIM, (char *)")")) { + if (test_type_token(type, token, EVENT_DELIM, ")")) { free_token(token); return EVENT_ERROR; } @@ -1626,7 +1626,7 @@ process_str(struct event *event __unused, struct print_arg *arg, char **tok) enum event_type type; char *token; - if (read_expected(EVENT_DELIM, (char *)"(") < 0) + if (read_expected(EVENT_DELIM, "(") < 0) return EVENT_ERROR; if (read_expect_type(EVENT_ITEM, &token) < 0) @@ -1636,7 +1636,7 @@ process_str(struct event *event __unused, struct print_arg *arg, char **tok) arg->string.string = token; arg->string.offset = -1; - if (read_expected(EVENT_DELIM, (char *)")") < 0) + if (read_expected(EVENT_DELIM, ")") < 0) return EVENT_ERROR; type = read_token(&token); @@ -1775,13 +1775,13 @@ static int event_read_print(struct event *event) char *token; int ret; - if (read_expected_item(EVENT_ITEM, (char *)"print") < 0) + if (read_expected_item(EVENT_ITEM, "print") < 0) return -1; - if (read_expected(EVENT_ITEM, (char *)"fmt") < 0) + if (read_expected(EVENT_ITEM, "fmt") < 0) return -1; - if (read_expected(EVENT_OP, (char *)":") < 0) + if (read_expected(EVENT_OP, ":") < 0) return -1; if (read_expect_type(EVENT_DQUOTE, &token) < 0) @@ -1811,8 +1811,8 @@ static int event_read_print(struct event *event) token = cat; goto concat; } - - if (test_type_token(type, token, EVENT_DELIM, (char *)",")) + + if (test_type_token(type, token, EVENT_DELIM, ",")) goto fail; free_token(token); @@ -1931,7 +1931,7 @@ static int get_common_info(const char *type, int *offset, int *size) } static int __parse_common(void *data, int *size, int *offset, - char *name) + const char *name) { int ret; @@ -1949,7 +1949,7 @@ int trace_parse_common_type(void *data) static int type_size; return __parse_common(data, &type_size, &type_offset, - (char *)"common_type"); + "common_type"); } static int parse_common_pid(void *data) @@ -1958,7 +1958,7 @@ static int parse_common_pid(void *data) static int pid_size; return __parse_common(data, &pid_size, &pid_offset, - (char *)"common_pid"); + "common_pid"); } static int parse_common_pc(void *data) @@ -1967,7 +1967,7 @@ static int parse_common_pc(void *data) static int pc_size; return __parse_common(data, &pc_size, &pc_offset, - (char *)"common_preempt_count"); + "common_preempt_count"); } static int parse_common_flags(void *data) @@ -1976,7 +1976,7 @@ static int parse_common_flags(void *data) static int flags_size; return __parse_common(data, &flags_size, &flags_offset, - (char *)"common_flags"); + "common_flags"); } static int parse_common_lock_depth(void *data) @@ -1986,7 +1986,7 @@ static int parse_common_lock_depth(void *data) int ret; ret = __parse_common(data, &ld_size, &ld_offset, - (char *)"common_lock_depth"); + "common_lock_depth"); if (ret < 0) return -1; @@ -3049,15 +3049,15 @@ static void print_args(struct print_arg *args) } } -static void parse_header_field(char *field, +static void parse_header_field(const char *field, int *offset, int *size) { char *token; int type; - if (read_expected(EVENT_ITEM, (char *)"field") < 0) + if (read_expected(EVENT_ITEM, "field") < 0) return; - if (read_expected(EVENT_OP, (char *)":") < 0) + if (read_expected(EVENT_OP, ":") < 0) return; /* type */ @@ -3067,27 +3067,27 @@ static void parse_header_field(char *field, if (read_expected(EVENT_ITEM, field) < 0) return; - if (read_expected(EVENT_OP, (char *)";") < 0) + if (read_expected(EVENT_OP, ";") < 0) return; - if (read_expected(EVENT_ITEM, (char *)"offset") < 0) + if (read_expected(EVENT_ITEM, "offset") < 0) return; - if (read_expected(EVENT_OP, (char *)":") < 0) + if (read_expected(EVENT_OP, ":") < 0) return; if (read_expect_type(EVENT_ITEM, &token) < 0) goto fail; *offset = atoi(token); free_token(token); - if (read_expected(EVENT_OP, (char *)";") < 0) + if (read_expected(EVENT_OP, ";") < 0) return; - if (read_expected(EVENT_ITEM, (char *)"size") < 0) + if (read_expected(EVENT_ITEM, "size") < 0) return; - if (read_expected(EVENT_OP, (char *)":") < 0) + if (read_expected(EVENT_OP, ":") < 0) return; if (read_expect_type(EVENT_ITEM, &token) < 0) goto fail; *size = atoi(token); free_token(token); - if (read_expected(EVENT_OP, (char *)";") < 0) + if (read_expected(EVENT_OP, ";") < 0) return; type = read_token(&token); if (type != EVENT_NEWLINE) { @@ -3095,19 +3095,19 @@ static void parse_header_field(char *field, if (type != EVENT_ITEM) goto fail; - if (strcmp(token, (char *)"signed") != 0) + if (strcmp(token, "signed") != 0) goto fail; free_token(token); - if (read_expected(EVENT_OP, (char *)":") < 0) + if (read_expected(EVENT_OP, ":") < 0) return; if (read_expect_type(EVENT_ITEM, &token)) goto fail; free_token(token); - if (read_expected(EVENT_OP, (char *)";") < 0) + if (read_expected(EVENT_OP, ";") < 0) return; if (read_expect_type(EVENT_NEWLINE, &token)) @@ -3121,11 +3121,11 @@ int parse_header_page(char *buf, unsigned long size) { init_input_buf(buf, size); - parse_header_field((char *)"timestamp", &header_page_ts_offset, + parse_header_field("timestamp", &header_page_ts_offset, &header_page_ts_size); - parse_header_field((char *)"commit", &header_page_size_offset, + parse_header_field("commit", &header_page_size_offset, &header_page_size_size); - parse_header_field((char *)"data", &header_page_data_offset, + parse_header_field("data", &header_page_data_offset, &header_page_data_size); return 0; -- cgit v1.2.3-18-g5258 From c171b552a7d316c7e1c3ad6f70a30178dd53e14c Mon Sep 17 00:00:00 2001 From: Li Zefan Date: Thu, 15 Oct 2009 11:22:07 +0800 Subject: perf trace: Add filter Suppport Add a new option "--filter " to perf record, and it should be right after "-e trace_point": #./perf record -R -f -e irq:irq_handler_entry --filter irq==18 ^C # ./perf trace perf-4303 ... irq_handler_entry: irq=18 handler=eth0 init-0 ... irq_handler_entry: irq=18 handler=eth0 init-0 ... irq_handler_entry: irq=18 handler=eth0 init-0 ... irq_handler_entry: irq=18 handler=eth0 init-0 ... irq_handler_entry: irq=18 handler=eth0 See Documentation/trace/events.txt for the syntax of filter expressions. Signed-off-by: Li Zefan Acked-by: Peter Zijlstra Acked-by: Frederic Weisbecker Cc: Steven Rostedt Cc: Tom Zanussi LKML-Reference: <4AD6955F.90602@cn.fujitsu.com> Signed-off-by: Ingo Molnar --- tools/perf/builtin-record.c | 15 ++++++++++++++- tools/perf/util/parse-events.c | 26 ++++++++++++++++++++++++-- tools/perf/util/parse-events.h | 2 ++ 3 files changed, 40 insertions(+), 3 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 4e3a374e7aa..8b2c860c49a 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -374,9 +374,11 @@ static struct perf_header_attr *get_header_attr(struct perf_event_attr *a, int n static void create_counter(int counter, int cpu, pid_t pid) { + char *filter = filters[counter]; struct perf_event_attr *attr = attrs + counter; struct perf_header_attr *h_attr; int track = !counter; /* only the first counter needs these */ + int ret; struct { u64 count; u64 time_enabled; @@ -479,7 +481,6 @@ try_again: multiplex_fd = fd[nr_cpu][counter]; if (multiplex && fd[nr_cpu][counter] != multiplex_fd) { - int ret; ret = ioctl(fd[nr_cpu][counter], PERF_EVENT_IOC_SET_OUTPUT, multiplex_fd); assert(ret != -1); @@ -499,6 +500,16 @@ try_again: } } + if (filter != NULL) { + ret = ioctl(fd[nr_cpu][counter], + PERF_EVENT_IOC_SET_FILTER, filter); + if (ret) { + error("failed to set filter with %d (%s)\n", errno, + strerror(errno)); + exit(-1); + } + } + ioctl(fd[nr_cpu][counter], PERF_EVENT_IOC_ENABLE); } @@ -676,6 +687,8 @@ static const struct option options[] = { OPT_CALLBACK('e', "event", NULL, "event", "event selector. use 'perf list' to list available events", parse_events), + OPT_CALLBACK(0, "filter", NULL, "filter", + "event filter", parse_filter), OPT_INTEGER('p', "pid", &target_pid, "record events on existing pid"), OPT_INTEGER('r', "realtime", &realtime_prio, diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c index 8cfb48cbbea..b097570e962 100644 --- a/tools/perf/util/parse-events.c +++ b/tools/perf/util/parse-events.c @@ -8,9 +8,10 @@ #include "cache.h" #include "header.h" -int nr_counters; +int nr_counters; struct perf_event_attr attrs[MAX_COUNTERS]; +char *filters[MAX_COUNTERS]; struct event_symbol { u8 type; @@ -708,7 +709,6 @@ static void store_event_type(const char *orgname) perf_header__push_event(id, orgname); } - int parse_events(const struct option *opt __used, const char *str, int unset __used) { struct perf_event_attr attr; @@ -745,6 +745,28 @@ int parse_events(const struct option *opt __used, const char *str, int unset __u return 0; } +int parse_filter(const struct option *opt __used, const char *str, + int unset __used) +{ + int i = nr_counters - 1; + int len = strlen(str); + + if (i < 0 || attrs[i].type != PERF_TYPE_TRACEPOINT) { + fprintf(stderr, + "-F option should follow a -e tracepoint option\n"); + return -1; + } + + filters[i] = malloc(len + 1); + if (!filters[i]) { + fprintf(stderr, "not enough memory to hold filter string\n"); + return -1; + } + strcpy(filters[i], str); + + return 0; +} + static const char * const event_type_descriptors[] = { "", "Hardware event", diff --git a/tools/perf/util/parse-events.h b/tools/perf/util/parse-events.h index 8626a439033..b8c1f64bc93 100644 --- a/tools/perf/util/parse-events.h +++ b/tools/perf/util/parse-events.h @@ -17,11 +17,13 @@ extern struct tracepoint_path *tracepoint_id_to_path(u64 config); extern int nr_counters; extern struct perf_event_attr attrs[MAX_COUNTERS]; +extern char *filters[MAX_COUNTERS]; extern const char *event_name(int ctr); extern const char *__event_name(int type, u64 config); extern int parse_events(const struct option *opt, const char *str, int unset); +extern int parse_filter(const struct option *opt, const char *str, int unset); #define EVENTS_HELP_MAX (128*1024) -- cgit v1.2.3-18-g5258 From f39cdf25bf77219676ec5360980ac40b1a7e144a Mon Sep 17 00:00:00 2001 From: Julia Lawall Date: Sat, 17 Oct 2009 08:43:17 +0200 Subject: perf tools: Move dereference after NULL test In each case, if the NULL test on thread is needed, then the dereference should be after the NULL test. A simplified version of the semantic match that detects this problem is as follows (http://coccinelle.lip6.fr/): // @match exists@ expression x, E; identifier fld; @@ * x->fld ... when != \(x = E\|&x\) * x == NULL // Signed-off-by: Julia Lawall LKML-Reference: Signed-off-by: Ingo Molnar --- tools/perf/builtin-annotate.c | 4 ++-- tools/perf/builtin-report.c | 4 ++-- tools/perf/builtin-sched.c | 4 ++-- tools/perf/builtin-trace.c | 4 ++-- 4 files changed, 8 insertions(+), 8 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c index 3fe0de03004..56ba71658d7 100644 --- a/tools/perf/builtin-annotate.c +++ b/tools/perf/builtin-annotate.c @@ -104,14 +104,14 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) event->ip.pid, (void *)(long)ip); - dump_printf(" ... thread: %s:%d\n", thread->comm, thread->pid); - if (thread == NULL) { fprintf(stderr, "problem processing %d event, skipping it.\n", event->header.type); return -1; } + dump_printf(" ... thread: %s:%d\n", thread->comm, thread->pid); + if (event->header.misc & PERF_RECORD_MISC_KERNEL) { level = 'k'; sym = kernel_maps__find_symbol(ip, &map); diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index 015c7974596..a4f8cc20915 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -629,14 +629,14 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) } } - dump_printf(" ... thread: %s:%d\n", thread->comm, thread->pid); - if (thread == NULL) { eprintf("problem processing %d event, skipping it.\n", event->header.type); return -1; } + dump_printf(" ... thread: %s:%d\n", thread->comm, thread->pid); + if (comm_list && !strlist__has_entry(comm_list, thread->comm)) return 0; diff --git a/tools/perf/builtin-sched.c b/tools/perf/builtin-sched.c index c9c68563e96..57ad3f458ef 100644 --- a/tools/perf/builtin-sched.c +++ b/tools/perf/builtin-sched.c @@ -1667,14 +1667,14 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) (void *)(long)ip, (long long)period); - dump_printf(" ... thread: %s:%d\n", thread->comm, thread->pid); - if (thread == NULL) { eprintf("problem processing %d event, skipping it.\n", event->header.type); return -1; } + dump_printf(" ... thread: %s:%d\n", thread->comm, thread->pid); + if (profile_cpu != -1 && profile_cpu != (int) cpu) return 0; diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-trace.c index ce8459ac284..4c129ff0bb1 100644 --- a/tools/perf/builtin-trace.c +++ b/tools/perf/builtin-trace.c @@ -80,14 +80,14 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) (void *)(long)ip, (long long)period); - dump_printf(" ... thread: %s:%d\n", thread->comm, thread->pid); - if (thread == NULL) { eprintf("problem processing %d event, skipping it.\n", event->header.type); return -1; } + dump_printf(" ... thread: %s:%d\n", thread->comm, thread->pid); + if (sample_type & PERF_SAMPLE_RAW) { struct { u32 size; -- cgit v1.2.3-18-g5258 From 4c20194c2de151bca14224ae384b47abf7636a95 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Fri, 16 Oct 2009 20:07:52 -0400 Subject: perf: Check libdwarf APIs for perf probe Check libdwarf APIs for perf probe in tools/perf/Makefile. Since dwarf_get_ranges() has been added from libdwarf 20081231 (and it's the newest function used in probe-finder.c), this just checks whether the function is defined. Signed-off-by: Masami Hiramatsu Cc: Frederic Weisbecker Cc: Arnaldo Carvalho de Melo Cc: Steven Rostedt Cc: Paul Mackerras Cc: Peter Zijlstra LKML-Reference: <20091017000752.16556.92051.stgit@dhcp-100-2-132.bos.redhat.com> Signed-off-by: Ingo Molnar --- tools/perf/Makefile | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/Makefile b/tools/perf/Makefile index 52b1f438e71..03c27b9068a 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -420,8 +420,8 @@ ifneq ($(shell sh -c "(echo '\#include '; echo 'int main(void) { Elf * msg := $(error No libelf.h/libelf found, please install libelf-dev/elfutils-libelf-devel); endif -ifneq ($(shell sh -c "(echo '\#include '; echo '\#include '; echo 'int main(void) { Dwarf_Debug dbg; Dwarf_Error err; dwarf_init(0, DW_DLC_READ, 0, 0, &dbg, &err); return (long)dbg; }') | $(CC) -x c - $(ALL_CFLAGS) -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -ldwarf -lelf -o /dev/null $(ALL_LDFLAGS) > /dev/null 2>&1 && echo y"), y) - msg := $(warning No libdwarf.h found, disables dwarf support. Please install libdwarf-dev/libdwarf-devel); +ifneq ($(shell sh -c "(echo '\#include '; echo '\#include '; echo 'int main(void) { Dwarf_Debug dbg; Dwarf_Error err; Dwarf_Ranges *rng; dwarf_init(0, DW_DLC_READ, 0, 0, &dbg, &err); dwarf_get_ranges(dbg, 0, &rng, 0, 0, &err); return (long)dbg; }') | $(CC) -x c - $(ALL_CFLAGS) -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -ldwarf -lelf -o /dev/null $(ALL_LDFLAGS) > /dev/null 2>&1 && echo y"), y) + msg := $(warning No libdwarf.h found or old libdwarf.h found, disables dwarf support. Please install libdwarf-dev/libdwarf-devel >= 20081231); BASIC_CFLAGS += -DNO_LIBDWARF else EXTLIBS += -lelf -ldwarf -- cgit v1.2.3-18-g5258 From 074fc0e4b3f5d24306c2995f2f3b0bd4759e8aeb Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Fri, 16 Oct 2009 20:08:01 -0400 Subject: perf: Use die() for error cases in perf-probe Use die() for exiting perf-probe with errors. This replaces perror_exit(), msg_exit() and fprintf()+exit() with die(), and uses die() in semantic_error(). This also renames 'die' local variables to 'dw_die' for avoiding name confliction. Signed-off-by: Masami Hiramatsu Cc: Frederic Weisbecker Cc: Arnaldo Carvalho de Melo Cc: Steven Rostedt Cc: Paul Mackerras Cc: Peter Zijlstra LKML-Reference: <20091017000801.16556.46866.stgit@dhcp-100-2-132.bos.redhat.com> Signed-off-by: Ingo Molnar --- tools/perf/builtin-probe.c | 47 +++++++++++------------------------- tools/perf/util/probe-finder.c | 55 +++++++++++++++--------------------------- 2 files changed, 34 insertions(+), 68 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-probe.c b/tools/perf/builtin-probe.c index 73c883b715c..a1467d12547 100644 --- a/tools/perf/builtin-probe.c +++ b/tools/perf/builtin-probe.c @@ -49,6 +49,7 @@ const char *default_search_path[NR_SEARCH_PATH] = { #define MAX_PATH_LEN 256 #define MAX_PROBES 128 +#define MAX_PROBE_ARGS 128 /* Session management structure */ static struct { @@ -60,19 +61,7 @@ static struct { char *events[MAX_PROBES]; } session; -static void semantic_error(const char *msg) -{ - fprintf(stderr, "Semantic error: %s\n", msg); - exit(1); -} - -static void perror_exit(const char *msg) -{ - perror(msg); - exit(1); -} - -#define MAX_PROBE_ARGS 128 +#define semantic_error(msg ...) die("Semantic error :" msg) static int parse_probepoint(const struct option *opt __used, const char *str, int unset __used) @@ -109,7 +98,7 @@ static int parse_probepoint(const struct option *opt __used, /* Duplicate the argument */ argv[argc] = strndup(s, str - s); if (argv[argc] == NULL) - perror_exit("strndup"); + die("strndup"); if (++argc == MAX_PROBE_ARGS) semantic_error("Too many arguments"); debug("argv[%d]=%s\n", argc, argv[argc - 1]); @@ -171,7 +160,7 @@ static int parse_probepoint(const struct option *opt __used, if (pp->nr_args > 0) { pp->args = (char **)malloc(sizeof(char *) * pp->nr_args); if (!pp->args) - perror_exit("malloc"); + die("malloc"); memcpy(pp->args, &argv[2], sizeof(char *) * pp->nr_args); } @@ -260,7 +249,7 @@ static int write_new_event(int fd, const char *buf) printf("Adding new event: %s\n", buf); ret = write(fd, buf, strlen(buf)); if (ret <= 0) - perror("Error: Failed to create event"); + die("failed to create event."); return ret; } @@ -273,7 +262,7 @@ static int synthesize_probepoint(struct probe_point *pp) int i, len, ret; pp->probes[0] = buf = (char *)calloc(MAX_CMDLEN, sizeof(char)); if (!buf) - perror_exit("calloc"); + die("calloc"); ret = snprintf(buf, MAX_CMDLEN, "%s+%d", pp->function, pp->offset); if (ret <= 0 || ret >= MAX_CMDLEN) goto error; @@ -322,10 +311,8 @@ int cmd_probe(int argc, const char **argv, const char *prefix __used) ret = synthesize_probepoint(&session.probes[j]); if (ret == -E2BIG) semantic_error("probe point is too long."); - else if (ret < 0) { - perror("snprintf"); - return -1; - } + else if (ret < 0) + die("snprintf"); } #ifndef NO_LIBDWARF @@ -336,10 +323,8 @@ int cmd_probe(int argc, const char **argv, const char *prefix __used) fd = open(session.vmlinux, O_RDONLY); else fd = open_default_vmlinux(); - if (fd < 0) { - perror("vmlinux/module file open"); - return -1; - } + if (fd < 0) + die("vmlinux/module file open"); /* Searching probe points */ for (j = 0; j < session.nr_probe; j++) { @@ -349,10 +334,8 @@ int cmd_probe(int argc, const char **argv, const char *prefix __used) lseek(fd, SEEK_SET, 0); ret = find_probepoint(fd, pp); - if (ret <= 0) { - fprintf(stderr, "Error: No probe point found.\n"); - return -1; - } + if (ret <= 0) + die("No probe point found.\n"); debug("probe event %s found\n", session.events[j]); } close(fd); @@ -363,10 +346,8 @@ setup_probes: /* Settng up probe points */ snprintf(buf, MAX_CMDLEN, "%s/../kprobe_events", debugfs_path); fd = open(buf, O_WRONLY, O_APPEND); - if (fd < 0) { - perror("kprobe_events open"); - return -1; - } + if (fd < 0) + die("kprobe_events open"); for (j = 0; j < session.nr_probe; j++) { pp = &session.probes[j]; if (pp->found == 1) { diff --git a/tools/perf/util/probe-finder.c b/tools/perf/util/probe-finder.c index ec6f53f29e0..338fdb9e093 100644 --- a/tools/perf/util/probe-finder.c +++ b/tools/perf/util/probe-finder.c @@ -31,6 +31,8 @@ #include #include #include + +#include "util.h" #include "probe-finder.h" @@ -43,20 +45,6 @@ struct die_link { static Dwarf_Debug __dw_debug; static Dwarf_Error __dw_error; -static void msg_exit(int ret, const char *fmt, ...) -{ - va_list ap; - - va_start(ap, fmt); - fprintf(stderr, "Error: "); - vfprintf(stderr, fmt, ap); - va_end(ap); - - fprintf(stderr, "\n"); - exit(ret); -} - - /* * Generic dwarf analysis helpers */ @@ -151,11 +139,11 @@ static Dwarf_Unsigned die_get_fileno(Dwarf_Die cu_die, const char *fname) } /* Compare diename and tname */ -static int die_compare_name(Dwarf_Die die, const char *tname) +static int die_compare_name(Dwarf_Die dw_die, const char *tname) { char *name; int ret; - ret = dwarf_diename(die, &name, &__dw_error); + ret = dwarf_diename(dw_die, &name, &__dw_error); ERR_IF(ret == DW_DLV_ERROR); if (ret == DW_DLV_OK) { ret = strcmp(tname, name); @@ -187,25 +175,25 @@ static int die_within_subprogram(Dwarf_Die sp_die, Dwarf_Addr addr, } /* Check the die is inlined function */ -static Dwarf_Bool die_inlined_subprogram(Dwarf_Die die) +static Dwarf_Bool die_inlined_subprogram(Dwarf_Die dw_die) { /* TODO: check strictly */ Dwarf_Bool inl; int ret; - ret = dwarf_hasattr(die, DW_AT_inline, &inl, &__dw_error); + ret = dwarf_hasattr(dw_die, DW_AT_inline, &inl, &__dw_error); ERR_IF(ret == DW_DLV_ERROR); return inl; } /* Get the offset of abstruct_origin */ -static Dwarf_Off die_get_abstract_origin(Dwarf_Die die) +static Dwarf_Off die_get_abstract_origin(Dwarf_Die dw_die) { Dwarf_Attribute attr; Dwarf_Off cu_offs; int ret; - ret = dwarf_attr(die, DW_AT_abstract_origin, &attr, &__dw_error); + ret = dwarf_attr(dw_die, DW_AT_abstract_origin, &attr, &__dw_error); ERR_IF(ret != DW_DLV_OK); ret = dwarf_formref(attr, &cu_offs, &__dw_error); ERR_IF(ret != DW_DLV_OK); @@ -214,7 +202,7 @@ static Dwarf_Off die_get_abstract_origin(Dwarf_Die die) } /* Get entry pc(or low pc, 1st entry of ranges) of the die */ -static Dwarf_Addr die_get_entrypc(Dwarf_Die die) +static Dwarf_Addr die_get_entrypc(Dwarf_Die dw_die) { Dwarf_Attribute attr; Dwarf_Addr addr; @@ -224,7 +212,7 @@ static Dwarf_Addr die_get_entrypc(Dwarf_Die die) int ret; /* Try to get entry pc */ - ret = dwarf_attr(die, DW_AT_entry_pc, &attr, &__dw_error); + ret = dwarf_attr(dw_die, DW_AT_entry_pc, &attr, &__dw_error); ERR_IF(ret == DW_DLV_ERROR); if (ret == DW_DLV_OK) { ret = dwarf_formaddr(attr, &addr, &__dw_error); @@ -234,13 +222,13 @@ static Dwarf_Addr die_get_entrypc(Dwarf_Die die) } /* Try to get low pc */ - ret = dwarf_lowpc(die, &addr, &__dw_error); + ret = dwarf_lowpc(dw_die, &addr, &__dw_error); ERR_IF(ret == DW_DLV_ERROR); if (ret == DW_DLV_OK) return addr; /* Try to get ranges */ - ret = dwarf_attr(die, DW_AT_ranges, &attr, &__dw_error); + ret = dwarf_attr(dw_die, DW_AT_ranges, &attr, &__dw_error); ERR_IF(ret != DW_DLV_OK); ret = dwarf_formref(attr, &offs, &__dw_error); ERR_IF(ret != DW_DLV_OK); @@ -382,11 +370,11 @@ static void show_location(Dwarf_Loc *loc, struct probe_finder *pf) } else if (op == DW_OP_regx) { regn = loc->lr_number; } else - msg_exit(-EINVAL, "Dwarf_OP %d is not supported.\n", op); + die("Dwarf_OP %d is not supported.\n", op); regs = get_arch_regstr(regn); if (!regs) - msg_exit(-EINVAL, "%lld exceeds max register number.\n", regn); + die("%lld exceeds max register number.\n", regn); if (deref) ret = snprintf(pf->buf, pf->len, @@ -417,8 +405,8 @@ static void show_variable(Dwarf_Die vr_die, struct probe_finder *pf) dwarf_dealloc(__dw_debug, attr, DW_DLA_ATTR); return ; error: - msg_exit(-1, "Failed to find the location of %s at this address.\n" - " Perhaps, it was optimized out.\n", pf->var); + die("Failed to find the location of %s at this address.\n" + " Perhaps, it has been optimized out.\n", pf->var); } static int variable_callback(struct die_link *dlink, void *data) @@ -456,8 +444,7 @@ static void find_variable(Dwarf_Die sp_die, struct probe_finder *pf) /* Search child die for local variables and parameters. */ ret = search_die_from_children(sp_die, variable_callback, pf); if (!ret) - msg_exit(-1, "Failed to find '%s' in this function.\n", - pf->var); + die("Failed to find '%s' in this function.\n", pf->var); } /* Get a frame base on the address */ @@ -568,8 +555,7 @@ static void find_by_line(Dwarf_Die cu_die, struct probe_finder *pf) /* Search a real subprogram including this line, */ ret = search_die_from_children(cu_die, probeaddr_callback, pf); if (ret == 0) - msg_exit(-1, - "Probe point is not found in subprograms.\n"); + die("Probe point is not found in subprograms.\n"); /* Continuing, because target line might be inlined. */ } dwarf_srclines_dealloc(__dw_debug, lines, cnt); @@ -621,7 +607,7 @@ static int probefunc_callback(struct die_link *dlink, void *data) !die_inlined_subprogram(lk->die)) goto found; } - msg_exit(-1, "Failed to find real subprogram.\n"); + die("Failed to find real subprogram.\n"); found: /* Get offset from subprogram */ ret = die_within_subprogram(lk->die, pf->addr, &offs); @@ -649,8 +635,7 @@ int find_probepoint(int fd, struct probe_point *pp) ret = dwarf_init(fd, DW_DLC_READ, 0, 0, &__dw_debug, &__dw_error); if (ret != DW_DLV_OK) - msg_exit(-1, "Failed to call dwarf_init(). " - "Maybe, not a dwarf file?\n"); + die("Failed to call dwarf_init(). Maybe, not a dwarf file.\n"); pp->found = 0; while (++cu_number) { -- cgit v1.2.3-18-g5258 From 89c69c0eee7515cdc217f4278de43547284b3458 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Fri, 16 Oct 2009 20:08:10 -0400 Subject: perf: Use eprintf() for debug messages in perf-probe Replace debug() macro with eprintf() and add -v option for showing those messages in perf-probe. Signed-off-by: Masami Hiramatsu Cc: Frederic Weisbecker Cc: Arnaldo Carvalho de Melo Cc: Steven Rostedt Cc: Paul Mackerras Cc: Peter Zijlstra LKML-Reference: <20091017000810.16556.38013.stgit@dhcp-100-2-132.bos.redhat.com> Signed-off-by: Ingo Molnar --- tools/perf/builtin-probe.c | 23 ++++++++++++++--------- tools/perf/util/probe-finder.c | 12 +++++++----- tools/perf/util/probe-finder.h | 7 ------- 3 files changed, 21 insertions(+), 21 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-probe.c b/tools/perf/builtin-probe.c index a1467d12547..b5ad86a265f 100644 --- a/tools/perf/builtin-probe.c +++ b/tools/perf/builtin-probe.c @@ -35,6 +35,8 @@ #include "perf.h" #include "builtin.h" #include "util/util.h" +#include "util/event.h" +#include "util/debug.h" #include "util/parse-options.h" #include "util/parse-events.h" /* For debugfs_path */ #include "util/probe-finder.h" @@ -76,7 +78,7 @@ static int parse_probepoint(const struct option *opt __used, if (!str) /* The end of probe points */ return 0; - debug("Probe-define(%d): %s\n", session.nr_probe, str); + eprintf("probe-definition(%d): %s\n", session.nr_probe, str); if (++session.nr_probe == MAX_PROBES) semantic_error("Too many probes"); @@ -101,7 +103,7 @@ static int parse_probepoint(const struct option *opt __used, die("strndup"); if (++argc == MAX_PROBE_ARGS) semantic_error("Too many arguments"); - debug("argv[%d]=%s\n", argc, argv[argc - 1]); + eprintf("argv[%d]=%s\n", argc, argv[argc - 1]); } } while (*str != '\0'); if (argc < 2) @@ -131,7 +133,7 @@ static int parse_probepoint(const struct option *opt __used, pp->line = atoi(ptr); if (!pp->file || !pp->line) semantic_error("Failed to parse line."); - debug("file:%s line:%d\n", pp->file, pp->line); + eprintf("file:%s line:%d\n", pp->file, pp->line); } else { /* Function name */ ptr = strchr(arg, '+'); @@ -148,7 +150,7 @@ static int parse_probepoint(const struct option *opt __used, pp->file = strdup(ptr); } pp->function = strdup(arg); - debug("symbol:%s file:%s offset:%d\n", + eprintf("symbol:%s file:%s offset:%d\n", pp->function, pp->file, pp->offset); } free(argv[1]); @@ -173,7 +175,7 @@ static int parse_probepoint(const struct option *opt __used, session.need_dwarf = 1; } - debug("%d arguments\n", pp->nr_args); + eprintf("%d arguments\n", pp->nr_args); return 0; } @@ -186,7 +188,7 @@ static int open_default_vmlinux(void) ret = uname(&uts); if (ret) { - debug("uname() failed.\n"); + eprintf("uname() failed.\n"); return -errno; } session.release = uts.release; @@ -194,11 +196,12 @@ static int open_default_vmlinux(void) ret = snprintf(fname, MAX_PATH_LEN, default_search_path[i], session.release); if (ret >= MAX_PATH_LEN || ret < 0) { - debug("Filename(%d,%s) is too long.\n", i, uts.release); + eprintf("Filename(%d,%s) is too long.\n", i, + uts.release); errno = E2BIG; return -E2BIG; } - debug("try to open %s\n", fname); + eprintf("try to open %s\n", fname); fd = open(fname, O_RDONLY); if (fd >= 0) break; @@ -213,6 +216,8 @@ static const char * const probe_usage[] = { }; static const struct option options[] = { + OPT_BOOLEAN('v', "verbose", &verbose, + "be more verbose (show parsed arguments, etc)"), #ifndef NO_LIBDWARF OPT_STRING('k', "vmlinux", &session.vmlinux, "file", "vmlinux/module pathname"), @@ -336,7 +341,7 @@ int cmd_probe(int argc, const char **argv, const char *prefix __used) ret = find_probepoint(fd, pp); if (ret <= 0) die("No probe point found.\n"); - debug("probe event %s found\n", session.events[j]); + eprintf("probe event %s found\n", session.events[j]); } close(fd); diff --git a/tools/perf/util/probe-finder.c b/tools/perf/util/probe-finder.c index 338fdb9e093..db24c913225 100644 --- a/tools/perf/util/probe-finder.c +++ b/tools/perf/util/probe-finder.c @@ -32,6 +32,8 @@ #include #include +#include "event.h" +#include "debug.h" #include "util.h" #include "probe-finder.h" @@ -134,7 +136,7 @@ static Dwarf_Unsigned die_get_fileno(Dwarf_Die cu_die, const char *fname) dwarf_dealloc(__dw_debug, srcs, DW_DLA_LIST); } if (found) - debug("found fno: %d\n", (int)found); + eprintf("found fno: %d\n", (int)found); return found; } @@ -440,7 +442,7 @@ static void find_variable(Dwarf_Die sp_die, struct probe_finder *pf) return ; } - debug("Searching '%s' variable in context.\n", pf->var); + eprintf("Searching '%s' variable in context.\n", pf->var); /* Search child die for local variables and parameters. */ ret = search_die_from_children(sp_die, variable_callback, pf); if (!ret) @@ -550,7 +552,7 @@ static void find_by_line(Dwarf_Die cu_die, struct probe_finder *pf) ret = dwarf_lineaddr(lines[i], &addr, &__dw_error); ERR_IF(ret != DW_DLV_OK); - debug("Probe point found: 0x%llx\n", addr); + eprintf("Probe point found: 0x%llx\n", addr); pf->addr = addr; /* Search a real subprogram including this line, */ ret = search_die_from_children(cu_die, probeaddr_callback, pf); @@ -581,7 +583,7 @@ static int probefunc_callback(struct die_link *dlink, void *data) &pf->inl_offs, &__dw_error); ERR_IF(ret != DW_DLV_OK); - debug("inline definition offset %lld\n", + eprintf("inline definition offset %lld\n", pf->inl_offs); return 0; } @@ -597,7 +599,7 @@ static int probefunc_callback(struct die_link *dlink, void *data) /* Get probe address */ pf->addr = die_get_entrypc(dlink->die); pf->addr += pp->offset; - debug("found inline addr: 0x%llx\n", pf->addr); + eprintf("found inline addr: 0x%llx\n", pf->addr); /* Inlined function. Get a real subprogram */ for (lk = dlink->parent; lk != NULL; lk = lk->parent) { tag = 0; diff --git a/tools/perf/util/probe-finder.h b/tools/perf/util/probe-finder.h index 306810c32f6..6a7cb0c04e9 100644 --- a/tools/perf/util/probe-finder.h +++ b/tools/perf/util/probe-finder.h @@ -4,13 +4,6 @@ #define _stringify(n) #n #define stringify(n) _stringify(n) -#ifdef DEBUG -#define debug(fmt ...) \ - fprintf(stderr, "DBG(" __FILE__ ":" stringify(__LINE__) "): " fmt) -#else -#define debug(fmt ...) do {} while (0) -#endif - #define ERR_IF(cnd) \ do { if (cnd) { \ fprintf(stderr, "Error (" __FILE__ ":" stringify(__LINE__) \ -- cgit v1.2.3-18-g5258 From 9769833b8e4425dc93fc837bf124c6cb02a51abb Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Fri, 16 Oct 2009 20:08:18 -0400 Subject: perf: Add DIE_IF() macro for error checking Add DIE_IF() macro and replace ERR_IF() with it, and use linux/stringify.h. Signed-off-by: Masami Hiramatsu Cc: Frederic Weisbecker Cc: Arnaldo Carvalho de Melo Cc: Steven Rostedt Cc: Paul Mackerras Cc: Peter Zijlstra LKML-Reference: <20091017000818.16556.82452.stgit@dhcp-100-2-132.bos.redhat.com> Signed-off-by: Ingo Molnar --- tools/perf/Makefile | 1 + tools/perf/util/probe-finder.c | 82 +++++++++++++++++++++--------------------- tools/perf/util/probe-finder.h | 10 ------ tools/perf/util/util.h | 9 +++++ 4 files changed, 51 insertions(+), 51 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/Makefile b/tools/perf/Makefile index 03c27b9068a..1abbf9a5ee5 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -321,6 +321,7 @@ LIB_FILE=libperf.a LIB_H += ../../include/linux/perf_event.h LIB_H += ../../include/linux/rbtree.h LIB_H += ../../include/linux/list.h +LIB_H += ../../include/linux/stringify.h LIB_H += util/include/linux/list.h LIB_H += perf.h LIB_H += util/types.h diff --git a/tools/perf/util/probe-finder.c b/tools/perf/util/probe-finder.c index db24c913225..be997abdf5b 100644 --- a/tools/perf/util/probe-finder.c +++ b/tools/perf/util/probe-finder.c @@ -146,7 +146,7 @@ static int die_compare_name(Dwarf_Die dw_die, const char *tname) char *name; int ret; ret = dwarf_diename(dw_die, &name, &__dw_error); - ERR_IF(ret == DW_DLV_ERROR); + DIE_IF(ret == DW_DLV_ERROR); if (ret == DW_DLV_OK) { ret = strcmp(tname, name); dwarf_dealloc(__dw_debug, name, DW_DLA_STRING); @@ -164,11 +164,11 @@ static int die_within_subprogram(Dwarf_Die sp_die, Dwarf_Addr addr, /* TODO: check ranges */ ret = dwarf_lowpc(sp_die, &lopc, &__dw_error); - ERR_IF(ret == DW_DLV_ERROR); + DIE_IF(ret == DW_DLV_ERROR); if (ret == DW_DLV_NO_ENTRY) return 0; ret = dwarf_highpc(sp_die, &hipc, &__dw_error); - ERR_IF(ret != DW_DLV_OK); + DIE_IF(ret != DW_DLV_OK); if (lopc <= addr && addr < hipc) { *offs = addr - lopc; return 1; @@ -184,7 +184,7 @@ static Dwarf_Bool die_inlined_subprogram(Dwarf_Die dw_die) int ret; ret = dwarf_hasattr(dw_die, DW_AT_inline, &inl, &__dw_error); - ERR_IF(ret == DW_DLV_ERROR); + DIE_IF(ret == DW_DLV_ERROR); return inl; } @@ -196,9 +196,9 @@ static Dwarf_Off die_get_abstract_origin(Dwarf_Die dw_die) int ret; ret = dwarf_attr(dw_die, DW_AT_abstract_origin, &attr, &__dw_error); - ERR_IF(ret != DW_DLV_OK); + DIE_IF(ret != DW_DLV_OK); ret = dwarf_formref(attr, &cu_offs, &__dw_error); - ERR_IF(ret != DW_DLV_OK); + DIE_IF(ret != DW_DLV_OK); dwarf_dealloc(__dw_debug, attr, DW_DLA_ATTR); return cu_offs; } @@ -215,28 +215,28 @@ static Dwarf_Addr die_get_entrypc(Dwarf_Die dw_die) /* Try to get entry pc */ ret = dwarf_attr(dw_die, DW_AT_entry_pc, &attr, &__dw_error); - ERR_IF(ret == DW_DLV_ERROR); + DIE_IF(ret == DW_DLV_ERROR); if (ret == DW_DLV_OK) { ret = dwarf_formaddr(attr, &addr, &__dw_error); - ERR_IF(ret != DW_DLV_OK); + DIE_IF(ret != DW_DLV_OK); dwarf_dealloc(__dw_debug, attr, DW_DLA_ATTR); return addr; } /* Try to get low pc */ ret = dwarf_lowpc(dw_die, &addr, &__dw_error); - ERR_IF(ret == DW_DLV_ERROR); + DIE_IF(ret == DW_DLV_ERROR); if (ret == DW_DLV_OK) return addr; /* Try to get ranges */ ret = dwarf_attr(dw_die, DW_AT_ranges, &attr, &__dw_error); - ERR_IF(ret != DW_DLV_OK); + DIE_IF(ret != DW_DLV_OK); ret = dwarf_formref(attr, &offs, &__dw_error); - ERR_IF(ret != DW_DLV_OK); + DIE_IF(ret != DW_DLV_OK); ret = dwarf_get_ranges(__dw_debug, offs, &ranges, &cnt, NULL, &__dw_error); - ERR_IF(ret != DW_DLV_OK); + DIE_IF(ret != DW_DLV_OK); addr = ranges[0].dwr_addr1; dwarf_ranges_dealloc(__dw_debug, ranges, cnt); return addr; @@ -261,7 +261,7 @@ static int __search_die_tree(struct die_link *cur_link, while (!(ret = die_cb(cur_link, data))) { /* Check child die */ ret = dwarf_child(cur_link->die, &new_die, &__dw_error); - ERR_IF(ret == DW_DLV_ERROR); + DIE_IF(ret == DW_DLV_ERROR); if (ret == DW_DLV_OK) { new_link.parent = cur_link; new_link.die = new_die; @@ -273,7 +273,7 @@ static int __search_die_tree(struct die_link *cur_link, /* Move to next sibling */ ret = dwarf_siblingof(__dw_debug, cur_link->die, &new_die, &__dw_error); - ERR_IF(ret == DW_DLV_ERROR); + DIE_IF(ret == DW_DLV_ERROR); dwarf_dealloc(__dw_debug, cur_link->die, DW_DLA_DIE); cur_link->die = new_die; if (ret == DW_DLV_NO_ENTRY) @@ -293,7 +293,7 @@ static int search_die_from_children(Dwarf_Die parent_die, new_link.parent = NULL; ret = dwarf_child(parent_die, &new_link.die, &__dw_error); - ERR_IF(ret == DW_DLV_ERROR); + DIE_IF(ret == DW_DLV_ERROR); if (ret == DW_DLV_OK) return __search_die_tree(&new_link, die_cb, data); else @@ -309,7 +309,7 @@ static int attr_get_locdesc(Dwarf_Attribute attr, Dwarf_Locdesc *desc, int ret, i; ret = dwarf_loclist_n(attr, &llbuf, &lcnt, &__dw_error); - ERR_IF(ret != DW_DLV_OK); + DIE_IF(ret != DW_DLV_OK); ret = DW_DLV_NO_ENTRY; for (i = 0; i < lcnt; ++i) { if (llbuf[i]->ld_lopc <= addr && @@ -317,7 +317,7 @@ static int attr_get_locdesc(Dwarf_Attribute attr, Dwarf_Locdesc *desc, memcpy(desc, llbuf[i], sizeof(Dwarf_Locdesc)); desc->ld_s = malloc(sizeof(Dwarf_Loc) * llbuf[i]->ld_cents); - ERR_IF(desc->ld_s == NULL); + DIE_IF(desc->ld_s == NULL); memcpy(desc->ld_s, llbuf[i]->ld_s, sizeof(Dwarf_Loc) * llbuf[i]->ld_cents); ret = DW_DLV_OK; @@ -383,8 +383,8 @@ static void show_location(Dwarf_Loc *loc, struct probe_finder *pf) " %s=%+lld(%s)", pf->var, offs, regs); else ret = snprintf(pf->buf, pf->len, " %s=%s", pf->var, regs); - ERR_IF(ret < 0); - ERR_IF(ret >= pf->len); + DIE_IF(ret < 0); + DIE_IF(ret >= pf->len); } /* Show a variables in kprobe event format */ @@ -401,7 +401,7 @@ static void show_variable(Dwarf_Die vr_die, struct probe_finder *pf) if (ret != DW_DLV_OK) goto error; /* TODO? */ - ERR_IF(ld.ld_cents != 1); + DIE_IF(ld.ld_cents != 1); show_location(&ld.ld_s[0], pf); free(ld.ld_s); dwarf_dealloc(__dw_debug, attr, DW_DLA_ATTR); @@ -418,7 +418,7 @@ static int variable_callback(struct die_link *dlink, void *data) int ret; ret = dwarf_tag(dlink->die, &tag, &__dw_error); - ERR_IF(ret == DW_DLV_ERROR); + DIE_IF(ret == DW_DLV_ERROR); if ((tag == DW_TAG_formal_parameter || tag == DW_TAG_variable) && (die_compare_name(dlink->die, pf->var) == 0)) { @@ -437,8 +437,8 @@ static void find_variable(Dwarf_Die sp_die, struct probe_finder *pf) if (!is_c_varname(pf->var)) { /* Output raw parameters */ ret = snprintf(pf->buf, pf->len, " %s", pf->var); - ERR_IF(ret < 0); - ERR_IF(ret >= pf->len); + DIE_IF(ret < 0); + DIE_IF(ret >= pf->len); return ; } @@ -456,9 +456,9 @@ static void get_current_frame_base(Dwarf_Die sp_die, struct probe_finder *pf) int ret; ret = dwarf_attr(sp_die, DW_AT_frame_base, &attr, &__dw_error); - ERR_IF(ret != DW_DLV_OK); + DIE_IF(ret != DW_DLV_OK); ret = attr_get_locdesc(attr, &pf->fbloc, (pf->addr - pf->cu_base)); - ERR_IF(ret != DW_DLV_OK); + DIE_IF(ret != DW_DLV_OK); dwarf_dealloc(__dw_debug, attr, DW_DLA_ATTR); } @@ -479,7 +479,7 @@ static void show_probepoint(Dwarf_Die sp_die, Dwarf_Signed offs, /* Output name of probe point */ ret = dwarf_diename(sp_die, &name, &__dw_error); - ERR_IF(ret == DW_DLV_ERROR); + DIE_IF(ret == DW_DLV_ERROR); if (ret == DW_DLV_OK) { ret = snprintf(tmp, MAX_PROBE_BUFFER, "%s+%u", name, (unsigned int)offs); @@ -488,8 +488,8 @@ static void show_probepoint(Dwarf_Die sp_die, Dwarf_Signed offs, /* This function has no name. */ ret = snprintf(tmp, MAX_PROBE_BUFFER, "0x%llx", pf->addr); } - ERR_IF(ret < 0); - ERR_IF(ret >= MAX_PROBE_BUFFER); + DIE_IF(ret < 0); + DIE_IF(ret >= MAX_PROBE_BUFFER); len = ret; /* Find each argument */ @@ -515,7 +515,7 @@ static int probeaddr_callback(struct die_link *dlink, void *data) int ret; ret = dwarf_tag(dlink->die, &tag, &__dw_error); - ERR_IF(ret == DW_DLV_ERROR); + DIE_IF(ret == DW_DLV_ERROR); /* Check the address is in this subprogram */ if (tag == DW_TAG_subprogram && die_within_subprogram(dlink->die, pf->addr, &offs)) { @@ -537,21 +537,21 @@ static void find_by_line(Dwarf_Die cu_die, struct probe_finder *pf) int ret; ret = dwarf_srclines(cu_die, &lines, &cnt, &__dw_error); - ERR_IF(ret != DW_DLV_OK); + DIE_IF(ret != DW_DLV_OK); for (i = 0; i < cnt; i++) { ret = dwarf_line_srcfileno(lines[i], &fno, &__dw_error); - ERR_IF(ret != DW_DLV_OK); + DIE_IF(ret != DW_DLV_OK); if (fno != pf->fno) continue; ret = dwarf_lineno(lines[i], &lineno, &__dw_error); - ERR_IF(ret != DW_DLV_OK); + DIE_IF(ret != DW_DLV_OK); if (lineno != (Dwarf_Unsigned)pp->line) continue; ret = dwarf_lineaddr(lines[i], &addr, &__dw_error); - ERR_IF(ret != DW_DLV_OK); + DIE_IF(ret != DW_DLV_OK); eprintf("Probe point found: 0x%llx\n", addr); pf->addr = addr; /* Search a real subprogram including this line, */ @@ -574,7 +574,7 @@ static int probefunc_callback(struct die_link *dlink, void *data) int ret; ret = dwarf_tag(dlink->die, &tag, &__dw_error); - ERR_IF(ret == DW_DLV_ERROR); + DIE_IF(ret == DW_DLV_ERROR); if (tag == DW_TAG_subprogram) { if (die_compare_name(dlink->die, pp->function) == 0) { if (die_inlined_subprogram(dlink->die)) { @@ -582,7 +582,7 @@ static int probefunc_callback(struct die_link *dlink, void *data) ret = dwarf_die_CU_offset(dlink->die, &pf->inl_offs, &__dw_error); - ERR_IF(ret != DW_DLV_OK); + DIE_IF(ret != DW_DLV_OK); eprintf("inline definition offset %lld\n", pf->inl_offs); return 0; @@ -604,7 +604,7 @@ static int probefunc_callback(struct die_link *dlink, void *data) for (lk = dlink->parent; lk != NULL; lk = lk->parent) { tag = 0; dwarf_tag(lk->die, &tag, &__dw_error); - ERR_IF(ret == DW_DLV_ERROR); + DIE_IF(ret == DW_DLV_ERROR); if (tag == DW_TAG_subprogram && !die_inlined_subprogram(lk->die)) goto found; @@ -613,7 +613,7 @@ static int probefunc_callback(struct die_link *dlink, void *data) found: /* Get offset from subprogram */ ret = die_within_subprogram(lk->die, pf->addr, &offs); - ERR_IF(!ret); + DIE_IF(!ret); show_probepoint(lk->die, offs, pf); /* Continue to search */ } @@ -644,13 +644,13 @@ int find_probepoint(int fd, struct probe_point *pp) /* Search CU (Compilation Unit) */ ret = dwarf_next_cu_header(__dw_debug, NULL, NULL, NULL, &addr_size, &next_cuh, &__dw_error); - ERR_IF(ret == DW_DLV_ERROR); + DIE_IF(ret == DW_DLV_ERROR); if (ret == DW_DLV_NO_ENTRY) break; /* Get the DIE(Debugging Information Entry) of this CU */ ret = dwarf_siblingof(__dw_debug, 0, &cu_die, &__dw_error); - ERR_IF(ret != DW_DLV_OK); + DIE_IF(ret != DW_DLV_OK); /* Check if target file is included. */ if (pp->file) @@ -659,7 +659,7 @@ int find_probepoint(int fd, struct probe_point *pp) if (!pp->file || pf.fno) { /* Save CU base address (for frame_base) */ ret = dwarf_lowpc(cu_die, &pf.cu_base, &__dw_error); - ERR_IF(ret == DW_DLV_ERROR); + DIE_IF(ret == DW_DLV_ERROR); if (ret == DW_DLV_NO_ENTRY) pf.cu_base = 0; if (pp->line) @@ -670,7 +670,7 @@ int find_probepoint(int fd, struct probe_point *pp) dwarf_dealloc(__dw_debug, cu_die, DW_DLA_DIE); } ret = dwarf_finish(__dw_debug, &__dw_error); - ERR_IF(ret != DW_DLV_OK); + DIE_IF(ret != DW_DLV_OK); return pp->found; } diff --git a/tools/perf/util/probe-finder.h b/tools/perf/util/probe-finder.h index 6a7cb0c04e9..d17fafc2135 100644 --- a/tools/perf/util/probe-finder.h +++ b/tools/perf/util/probe-finder.h @@ -1,16 +1,6 @@ #ifndef _PROBE_FINDER_H #define _PROBE_FINDER_H -#define _stringify(n) #n -#define stringify(n) _stringify(n) - -#define ERR_IF(cnd) \ - do { if (cnd) { \ - fprintf(stderr, "Error (" __FILE__ ":" stringify(__LINE__) \ - "): " stringify(cnd) "\n"); \ - exit(1); \ - } } while (0) - #define MAX_PATH_LEN 256 #define MAX_PROBE_BUFFER 1024 #define MAX_PROBES 128 diff --git a/tools/perf/util/util.h b/tools/perf/util/util.h index 9de2329dd44..0daa341734f 100644 --- a/tools/perf/util/util.h +++ b/tools/perf/util/util.h @@ -134,6 +134,15 @@ extern void die(const char *err, ...) NORETURN __attribute__((format (printf, 1, extern int error(const char *err, ...) __attribute__((format (printf, 1, 2))); extern void warning(const char *err, ...) __attribute__((format (printf, 1, 2))); +#include "../../../include/linux/stringify.h" + +#define DIE_IF(cnd) \ + do { if (cnd) \ + die(" at (" __FILE__ ":" __stringify(__LINE__) "): " \ + __stringify(cnd) "\n"); \ + } while (0) + + extern void set_die_routine(void (*routine)(const char *err, va_list params) NORETURN); extern int prefixcmp(const char *str, const char *prefix); -- cgit v1.2.3-18-g5258 From 595c36490deb49381dc51231a3d5e6b66786ed27 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Fri, 16 Oct 2009 20:08:27 -0400 Subject: perf: Add perf-probe document Add perf-probe subcommand document and add it to command-list. Signed-off-by: Masami Hiramatsu Cc: Frederic Weisbecker Cc: Arnaldo Carvalho de Melo Cc: Steven Rostedt Cc: Paul Mackerras Cc: Peter Zijlstra LKML-Reference: <20091017000827.16556.73539.stgit@dhcp-100-2-132.bos.redhat.com> Signed-off-by: Ingo Molnar --- tools/perf/Documentation/perf-probe.txt | 48 +++++++++++++++++++++++++++++++++ tools/perf/command-list.txt | 1 + 2 files changed, 49 insertions(+) create mode 100644 tools/perf/Documentation/perf-probe.txt (limited to 'tools/perf') diff --git a/tools/perf/Documentation/perf-probe.txt b/tools/perf/Documentation/perf-probe.txt new file mode 100644 index 00000000000..6b6c6aecdf1 --- /dev/null +++ b/tools/perf/Documentation/perf-probe.txt @@ -0,0 +1,48 @@ +perf-probe(1) +============= + +NAME +---- +perf-probe - Define new dynamic tracepoints + +SYNOPSIS +-------- +[verse] +'perf probe' [-k ] -P 'PROBE' [-P 'PROBE' ...] + + +DESCRIPTION +----------- +This command defines dynamic tracepoint events, by symbol and registers +without debuginfo, or by C expressions (C line numbers, C function names, +and C local variables) with debuginfo. + + +OPTIONS +------- +-k:: +--vmlinux:: + Specify vmlinux path which has debuginfo (Dwarf binary). + +-v:: +--verbose:: + Be more verbose (show parsed arguments, etc). + +-P:: +--probe:: + Define a probe point (see PROBE SYNTAX for detail) + +PROBE SYNTAX +------------ +Probe points are defined by following syntax. + + "TYPE:[GRP/]NAME FUNC[+OFFS][@SRC]|@SRC:LINE [ARG ...]" + +'TYPE' specifies the type of probe point, you can use either "p" (kprobe) or "r" (kretprobe) for 'TYPE'. 'GRP' specifies the group name of this probe, and 'NAME' specifies the event name. If 'GRP' is omitted, "kprobes" is used for its group name. +'FUNC' and 'OFFS' specifies function and offset (in byte) where probe will be put. In addition, 'SRC' specifies a source file which has that function (this is mainly for inline functions). +You can specify a probe point by the source line number by using '@SRC:LINE' syntax, where 'SRC' is the source file path and 'LINE' is the line number. +'ARG' specifies the arguments of this probe point. You can use the name of local variable, or kprobe-tracer argument format (e.g. $retval, %ax, etc). + +SEE ALSO +-------- +linkperf:perf-trace[1], linkperf:perf-record[1] diff --git a/tools/perf/command-list.txt b/tools/perf/command-list.txt index 00326e230d8..6475db4f194 100644 --- a/tools/perf/command-list.txt +++ b/tools/perf/command-list.txt @@ -11,3 +11,4 @@ perf-stat mainporcelain common perf-timechart mainporcelain common perf-top mainporcelain common perf-trace mainporcelain common +perf-probe mainporcelain common -- cgit v1.2.3-18-g5258 From 11018201b831e19304c0d639f105ad6c27e120b1 Mon Sep 17 00:00:00 2001 From: Anton Blanchard Date: Sun, 18 Oct 2009 22:29:23 +1100 Subject: perf stat: Add branch performance metric When we count both branches and branch-misses it is useful to print out the percentage of branch-misses: # perf stat -e branches -e branch-misses /bin/true Performance counter stats for '/bin/true': 401684 branches # 0.000 M/sec 23301 branch-misses # 5.801 % Signed-off-by: Anton Blanchard Cc: paulus@samba.org Cc: a.p.zijlstra@chello.nl LKML-Reference: <20091018112923.GQ4808@kryten> Signed-off-by: Ingo Molnar --- tools/perf/builtin-stat.c | 11 +++++++++++ 1 file changed, 11 insertions(+) (limited to 'tools/perf') diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c index 3db31e7bf17..c37368343ff 100644 --- a/tools/perf/builtin-stat.c +++ b/tools/perf/builtin-stat.c @@ -125,6 +125,7 @@ struct stats event_res_stats[MAX_COUNTERS][3]; struct stats runtime_nsecs_stats; struct stats walltime_nsecs_stats; struct stats runtime_cycles_stats; +struct stats runtime_branches_stats; #define MATCH_EVENT(t, c, counter) \ (attrs[counter].type == PERF_TYPE_##t && \ @@ -235,6 +236,8 @@ static void read_counter(int counter) update_stats(&runtime_nsecs_stats, count[0]); if (MATCH_EVENT(HARDWARE, HW_CPU_CYCLES, counter)) update_stats(&runtime_cycles_stats, count[0]); + if (MATCH_EVENT(HARDWARE, HW_BRANCH_INSTRUCTIONS, counter)) + update_stats(&runtime_branches_stats, count[0]); } static int run_perf_stat(int argc __used, const char **argv) @@ -352,6 +355,14 @@ static void abs_printout(int counter, double avg) ratio = avg / total; fprintf(stderr, " # %10.3f IPC ", ratio); + } else if (MATCH_EVENT(HARDWARE, HW_BRANCH_MISSES, counter)) { + total = avg_stats(&runtime_branches_stats); + + if (total) + ratio = avg * 100 / total; + + fprintf(stderr, " # %10.3f %% ", ratio); + } else { total = avg_stats(&runtime_nsecs_stats); -- cgit v1.2.3-18-g5258 From 5a116dd2797677cad48fee2f42267e3cb69f5502 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Sat, 17 Oct 2009 17:12:33 +0200 Subject: perf tools: Use kernel bitmap library Use the kernel bitmap library for internal perf tools uses. Signed-off-by: Frederic Weisbecker Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Mike Galbraith Cc: Paul Mackerras Cc: Steven Rostedt LKML-Reference: <1255792354-11304-1-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar --- tools/perf/Makefile | 16 +++++++++ tools/perf/builtin-record.c | 3 -- tools/perf/builtin-sched.c | 2 -- tools/perf/util/include/asm/bitops.h | 6 ++++ tools/perf/util/include/asm/byteorder.h | 2 ++ tools/perf/util/include/asm/swab.h | 1 + tools/perf/util/include/asm/types.h | 17 +++++++++ tools/perf/util/include/asm/uaccess.h | 14 ++++++++ tools/perf/util/include/linux/bitmap.h | 2 ++ tools/perf/util/include/linux/bitops.h | 27 +++++++++++++++ tools/perf/util/include/linux/compiler.h | 10 ++++++ tools/perf/util/include/linux/ctype.h | 1 + tools/perf/util/include/linux/kernel.h | 59 ++++++++++++++++++++++++++++++++ tools/perf/util/include/linux/types.h | 1 + 14 files changed, 156 insertions(+), 5 deletions(-) create mode 100644 tools/perf/util/include/asm/bitops.h create mode 100644 tools/perf/util/include/asm/byteorder.h create mode 100644 tools/perf/util/include/asm/swab.h create mode 100644 tools/perf/util/include/asm/types.h create mode 100644 tools/perf/util/include/asm/uaccess.h create mode 100644 tools/perf/util/include/linux/bitmap.h create mode 100644 tools/perf/util/include/linux/bitops.h create mode 100644 tools/perf/util/include/linux/compiler.h create mode 100644 tools/perf/util/include/linux/ctype.h create mode 100644 tools/perf/util/include/linux/types.h (limited to 'tools/perf') diff --git a/tools/perf/Makefile b/tools/perf/Makefile index 106c15055b5..2400e5068a2 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -363,6 +363,9 @@ LIB_OBJS += util/parse-options.o LIB_OBJS += util/parse-events.o LIB_OBJS += util/path.o LIB_OBJS += util/rbtree.o +LIB_OBJS += util/bitmap.o +LIB_OBJS += util/hweight.o +LIB_OBJS += util/find_next_bit.o LIB_OBJS += util/run-command.o LIB_OBJS += util/quote.o LIB_OBJS += util/strbuf.o @@ -790,6 +793,19 @@ util/config.o: util/config.c PERF-CFLAGS util/rbtree.o: ../../lib/rbtree.c PERF-CFLAGS $(QUIET_CC)$(CC) -o util/rbtree.o -c $(ALL_CFLAGS) -DETC_PERFCONFIG='"$(ETC_PERFCONFIG_SQ)"' $< +# some perf warning policies can't fit to lib/bitmap.c, eg: it warns about variable shadowing +# from that comes from kernel headers wrapping. +KBITMAP_FLAGS=`echo $(ALL_CFLAGS) | sed s/-Wshadow// | sed s/-Wswitch-default// | sed s/-Wextra//` + +util/bitmap.o: ../../lib/bitmap.c PERF-CFLAGS + $(QUIET_CC)$(CC) -o util/bitmap.o -c $(KBITMAP_FLAGS) -DETC_PERFCONFIG='"$(ETC_PERFCONFIG_SQ)"' $< + +util/hweight.o: ../../lib/hweight.c PERF-CFLAGS + $(QUIET_CC)$(CC) -o util/hweight.o -c $(ALL_CFLAGS) -DETC_PERFCONFIG='"$(ETC_PERFCONFIG_SQ)"' $< + +util/find_next_bit.o: ../../lib/find_next_bit.c PERF-CFLAGS + $(QUIET_CC)$(CC) -o util/find_next_bit.o -c $(ALL_CFLAGS) -DETC_PERFCONFIG='"$(ETC_PERFCONFIG_SQ)"' $< + perf-%$X: %.o $(PERFLIBS) $(QUIET_LINK)$(CC) $(ALL_CFLAGS) -o $@ $(ALL_LDFLAGS) $(filter %.o,$^) $(LIBS) diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 8b2c860c49a..fc3709cba13 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -21,9 +21,6 @@ #include #include -#define ALIGN(x, a) __ALIGN_MASK(x, (typeof(x))(a)-1) -#define __ALIGN_MASK(x, mask) (((x)+(mask))&~(mask)) - static int fd[MAX_NR_CPUS][MAX_COUNTERS]; static long default_interval = 0; diff --git a/tools/perf/builtin-sched.c b/tools/perf/builtin-sched.c index 57ad3f458ef..807ca66e7a8 100644 --- a/tools/perf/builtin-sched.c +++ b/tools/perf/builtin-sched.c @@ -38,8 +38,6 @@ static int cwdlen; #define PR_SET_NAME 15 /* Set process name */ #define MAX_CPUS 4096 -#define BUG_ON(x) assert(!(x)) - static u64 run_measurement_overhead; static u64 sleep_measurement_overhead; diff --git a/tools/perf/util/include/asm/bitops.h b/tools/perf/util/include/asm/bitops.h new file mode 100644 index 00000000000..fbe4d921291 --- /dev/null +++ b/tools/perf/util/include/asm/bitops.h @@ -0,0 +1,6 @@ +#include "../../../../include/asm-generic/bitops/__fls.h" +#include "../../../../include/asm-generic/bitops/fls.h" +#include "../../../../include/asm-generic/bitops/fls64.h" +#include "../../../../include/asm-generic/bitops/__ffs.h" +#include "../../../../include/asm-generic/bitops/ffz.h" +#include "../../../../include/asm-generic/bitops/hweight.h" diff --git a/tools/perf/util/include/asm/byteorder.h b/tools/perf/util/include/asm/byteorder.h new file mode 100644 index 00000000000..39f367cfaf5 --- /dev/null +++ b/tools/perf/util/include/asm/byteorder.h @@ -0,0 +1,2 @@ +#include "../asm/types.h" +#include "../../../../include/linux/swab.h" diff --git a/tools/perf/util/include/asm/swab.h b/tools/perf/util/include/asm/swab.h new file mode 100644 index 00000000000..ed538942523 --- /dev/null +++ b/tools/perf/util/include/asm/swab.h @@ -0,0 +1 @@ +/* stub */ diff --git a/tools/perf/util/include/asm/types.h b/tools/perf/util/include/asm/types.h new file mode 100644 index 00000000000..06703c6cd50 --- /dev/null +++ b/tools/perf/util/include/asm/types.h @@ -0,0 +1,17 @@ +#ifndef PERF_ASM_TYPES_H_ +#define PERF_ASM_TYPES_H_ + +#include +#include "../../types.h" +#include + +/* CHECKME: Not sure both always match */ +#define BITS_PER_LONG __WORDSIZE + +typedef u64 __u64; +typedef u32 __u32; +typedef u16 __u16; +typedef u8 __u8; +typedef s64 __s64; + +#endif /* PERF_ASM_TYPES_H_ */ diff --git a/tools/perf/util/include/asm/uaccess.h b/tools/perf/util/include/asm/uaccess.h new file mode 100644 index 00000000000..d0f72b8fcc3 --- /dev/null +++ b/tools/perf/util/include/asm/uaccess.h @@ -0,0 +1,14 @@ +#ifndef _PERF_ASM_UACCESS_H_ +#define _PERF_ASM_UACCESS_H_ + +#define __get_user(src, dest) \ +({ \ + (src) = *dest; \ + 0; \ +}) + +#define get_user __get_user + +#define access_ok(type, addr, size) 1 + +#endif diff --git a/tools/perf/util/include/linux/bitmap.h b/tools/perf/util/include/linux/bitmap.h new file mode 100644 index 00000000000..821c1033bcc --- /dev/null +++ b/tools/perf/util/include/linux/bitmap.h @@ -0,0 +1,2 @@ +#include "../../../../include/linux/bitmap.h" +#include "../../../../include/asm-generic/bitops/find.h" diff --git a/tools/perf/util/include/linux/bitops.h b/tools/perf/util/include/linux/bitops.h new file mode 100644 index 00000000000..ace57c36d1d --- /dev/null +++ b/tools/perf/util/include/linux/bitops.h @@ -0,0 +1,27 @@ +#ifndef _PERF_LINUX_BITOPS_H_ +#define _PERF_LINUX_BITOPS_H_ + +#define __KERNEL__ + +#define CONFIG_GENERIC_FIND_NEXT_BIT +#define CONFIG_GENERIC_FIND_FIRST_BIT +#include "../../../../include/linux/bitops.h" + +static inline void set_bit(int nr, unsigned long *addr) +{ + addr[nr / BITS_PER_LONG] |= 1UL << (nr % BITS_PER_LONG); +} + +static __always_inline int test_bit(unsigned int nr, const unsigned long *addr) +{ + return ((1UL << (nr % BITS_PER_LONG)) & + (((unsigned long *)addr)[nr / BITS_PER_LONG])) != 0; +} + +unsigned long generic_find_next_zero_le_bit(const unsigned long *addr, unsigned + long size, unsigned long offset); + +unsigned long generic_find_next_le_bit(const unsigned long *addr, unsigned + long size, unsigned long offset); + +#endif diff --git a/tools/perf/util/include/linux/compiler.h b/tools/perf/util/include/linux/compiler.h new file mode 100644 index 00000000000..dfb0713ed47 --- /dev/null +++ b/tools/perf/util/include/linux/compiler.h @@ -0,0 +1,10 @@ +#ifndef _PERF_LINUX_COMPILER_H_ +#define _PERF_LINUX_COMPILER_H_ + +#ifndef __always_inline +#define __always_inline inline +#endif +#define __user +#define __attribute_const__ + +#endif diff --git a/tools/perf/util/include/linux/ctype.h b/tools/perf/util/include/linux/ctype.h new file mode 100644 index 00000000000..bae5783282e --- /dev/null +++ b/tools/perf/util/include/linux/ctype.h @@ -0,0 +1 @@ +#include "../../../../include/linux/ctype.h" diff --git a/tools/perf/util/include/linux/kernel.h b/tools/perf/util/include/linux/kernel.h index a6b87390cb5..4b9204d9b26 100644 --- a/tools/perf/util/include/linux/kernel.h +++ b/tools/perf/util/include/linux/kernel.h @@ -1,6 +1,16 @@ #ifndef PERF_LINUX_KERNEL_H_ #define PERF_LINUX_KERNEL_H_ +#include +#include +#include +#include + +#define DIV_ROUND_UP(n,d) (((n) + (d) - 1) / (d)) + +#define ALIGN(x,a) __ALIGN_MASK(x,(typeof(x))(a)-1) +#define __ALIGN_MASK(x,mask) (((x)+(mask))&~(mask)) + #ifndef offsetof #define offsetof(TYPE, MEMBER) ((size_t) &((TYPE *)0)->MEMBER) #endif @@ -26,4 +36,53 @@ _max1 > _max2 ? _max1 : _max2; }) #endif +#ifndef min +#define min(x, y) ({ \ + typeof(x) _min1 = (x); \ + typeof(y) _min2 = (y); \ + (void) (&_min1 == &_min2); \ + _min1 < _min2 ? _min1 : _min2; }) +#endif + +#ifndef BUG_ON +#define BUG_ON(cond) assert(!(cond)) +#endif + +/* + * Both need more care to handle endianness + * (Don't use bitmap_copy_le() for now) + */ +#define cpu_to_le64(x) (x) +#define cpu_to_le32(x) (x) + +static inline int +vscnprintf(char *buf, size_t size, const char *fmt, va_list args) +{ + int i; + ssize_t ssize = size; + + i = vsnprintf(buf, size, fmt, args); + + return (i >= ssize) ? (ssize - 1) : i; +} + +static inline int scnprintf(char * buf, size_t size, const char * fmt, ...) +{ + va_list args; + ssize_t ssize = size; + int i; + + va_start(args, fmt); + i = vsnprintf(buf, size, fmt, args); + va_end(args); + + return (i >= ssize) ? (ssize - 1) : i; +} + +static inline unsigned long +simple_strtoul(const char *nptr, char **endptr, int base) +{ + return strtoul(nptr, endptr, base); +} + #endif diff --git a/tools/perf/util/include/linux/types.h b/tools/perf/util/include/linux/types.h new file mode 100644 index 00000000000..ed538942523 --- /dev/null +++ b/tools/perf/util/include/linux/types.h @@ -0,0 +1 @@ +/* stub */ -- cgit v1.2.3-18-g5258 From 2ba0825075e76236d22a20decd8e2346a99faabe Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Sat, 17 Oct 2009 17:12:34 +0200 Subject: perf tools: Introduce bitmask'ed additional headers This provides a new set of bitmasked headers. A new field is added in the perf headers that implements a bitmap storing optional features present in the perf.data file. The layout can be pictured like this: (Usual perf headers)(Features bitmap)[Feature 0][Feature n][Feature 255] If the bit n is set, then the feature n is used in this file. They are all set in order. This brings a backward and forward compatibility. The trace_info section has moved into such optional features, this is the first and only one for now. This is backward compatible with the .32 file version although it doesn't support the previous separate trace.info file. And finally it doesn't support the current interim development version. Signed-off-by: Frederic Weisbecker Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Mike Galbraith Cc: Paul Mackerras Cc: Steven Rostedt LKML-Reference: <1255792354-11304-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar --- tools/perf/builtin-record.c | 4 +- tools/perf/util/header.c | 100 +++++++++++++++++++++++++------------------- tools/perf/util/header.h | 30 ++++++++----- 3 files changed, 78 insertions(+), 56 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index fc3709cba13..f0467ff0d8a 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -574,11 +574,11 @@ static int __cmd_record(int argc, const char **argv) header = perf_header__new(); if (raw_samples) { - perf_header__set_trace_info(); + perf_header__feat_trace_info(header); } else { for (i = 0; i < nr_counters; i++) { if (attrs[i].sample_type & PERF_SAMPLE_RAW) { - perf_header__set_trace_info(); + perf_header__feat_trace_info(header); break; } } diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index 9aae360c0f2..171d51b6f35 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -8,6 +8,8 @@ #include "../perf.h" #include "trace-event.h" +#include + /* * Create new perf.data header attribute: */ @@ -48,25 +50,17 @@ void perf_header_attr__add_id(struct perf_header_attr *self, u64 id) */ struct perf_header *perf_header__new(void) { - struct perf_header *self = malloc(sizeof(*self)); + struct perf_header *self = calloc(sizeof(*self), 1); if (!self) die("nomem"); - self->frozen = 0; - - self->attrs = 0; self->size = 1; self->attr = malloc(sizeof(void *)); if (!self->attr) die("nomem"); - self->data_offset = 0; - self->data_size = 0; - self->trace_info_offset = 0; - self->trace_info_size = 0; - return self; } @@ -149,14 +143,12 @@ struct perf_file_header { struct perf_file_section attrs; struct perf_file_section data; struct perf_file_section event_types; - struct perf_file_section trace_info; + feat_mask_t adds_features; }; -static int trace_info; - -void perf_header__set_trace_info(void) +void perf_header__feat_trace_info(struct perf_header *header) { - trace_info = 1; + set_bit(HEADER_TRACE_INFO, perf_header__adds_mask(header)); } static void do_write(int fd, void *buf, size_t size) @@ -172,6 +164,32 @@ static void do_write(int fd, void *buf, size_t size) } } +static void perf_header__adds_write(struct perf_header *self, int fd) +{ + struct perf_file_section trace_sec; + u64 cur_offset = lseek(fd, 0, SEEK_CUR); + unsigned long *feat_mask = perf_header__adds_mask(self); + + if (test_bit(HEADER_TRACE_INFO, feat_mask)) { + /* Write trace info */ + trace_sec.offset = lseek(fd, sizeof(trace_sec), SEEK_CUR); + read_tracing_data(fd, attrs, nr_counters); + trace_sec.size = lseek(fd, 0, SEEK_CUR) - trace_sec.offset; + + /* Write trace info headers */ + lseek(fd, cur_offset, SEEK_SET); + do_write(fd, &trace_sec, sizeof(trace_sec)); + + /* + * Update cur_offset. So that other (future) + * features can set their own infos in this place. But if we are + * the only feature, at least that seeks to the place the data + * should begin. + */ + cur_offset = lseek(fd, trace_sec.offset + trace_sec.size, SEEK_SET); + } +}; + void perf_header__write(struct perf_header *self, int fd) { struct perf_file_header f_header; @@ -210,23 +228,7 @@ void perf_header__write(struct perf_header *self, int fd) if (events) do_write(fd, events, self->event_size); - if (trace_info) { - static int trace_info_written; - - /* - * Write it only once - */ - if (!trace_info_written) { - self->trace_info_offset = lseek(fd, 0, SEEK_CUR); - read_tracing_data(fd, attrs, nr_counters); - self->trace_info_size = lseek(fd, 0, SEEK_CUR) - - self->trace_info_offset; - trace_info_written = 1; - } else { - lseek(fd, self->trace_info_offset + - self->trace_info_size, SEEK_SET); - } - } + perf_header__adds_write(self, fd); self->data_offset = lseek(fd, 0, SEEK_CUR); @@ -246,12 +248,10 @@ void perf_header__write(struct perf_header *self, int fd) .offset = self->event_offset, .size = self->event_size, }, - .trace_info = { - .offset = self->trace_info_offset, - .size = self->trace_info_size, - }, }; + memcpy(&f_header.adds_features, &self->adds_features, sizeof(feat_mask_t)); + lseek(fd, 0, SEEK_SET); do_write(fd, &f_header, sizeof(f_header)); lseek(fd, self->data_offset + self->data_size, SEEK_SET); @@ -274,6 +274,20 @@ static void do_read(int fd, void *buf, size_t size) } } +static void perf_header__adds_read(struct perf_header *self, int fd) +{ + const unsigned long *feat_mask = perf_header__adds_mask(self); + + if (test_bit(HEADER_TRACE_INFO, feat_mask)) { + struct perf_file_section trace_sec; + + do_read(fd, &trace_sec, sizeof(trace_sec)); + lseek(fd, trace_sec.offset, SEEK_SET); + trace_report(fd); + lseek(fd, trace_sec.offset + trace_sec.size, SEEK_SET); + } +}; + struct perf_header *perf_header__read(int fd) { struct perf_header *self = perf_header__new(); @@ -292,9 +306,11 @@ struct perf_header *perf_header__read(int fd) if (f_header.size != sizeof(f_header)) { /* Support the previous format */ - if (f_header.size == offsetof(typeof(f_header), trace_info)) - f_header.trace_info.size = 0; - else + if (f_header.size == offsetof(typeof(f_header), adds_features)) { + unsigned long *mask = (unsigned long *)(void *) + &f_header.adds_features; + bitmap_zero(mask, HEADER_FEAT_BITS); + } else die("incompatible file format"); } nr_attrs = f_header.attrs.size / sizeof(f_attr); @@ -330,13 +346,9 @@ struct perf_header *perf_header__read(int fd) event_count = f_header.event_types.size / sizeof(struct perf_trace_event_type); } - self->trace_info_offset = f_header.trace_info.offset; - self->trace_info_size = f_header.trace_info.size; + memcpy(&self->adds_features, &f_header.adds_features, sizeof(feat_mask_t)); - if (self->trace_info_size) { - lseek(fd, self->trace_info_offset, SEEK_SET); - trace_report(fd); - } + perf_header__adds_read(self, fd); self->event_offset = f_header.event_types.offset; self->event_size = f_header.event_types.size; diff --git a/tools/perf/util/header.h b/tools/perf/util/header.h index 30aee5160dc..0eb4a9126b7 100644 --- a/tools/perf/util/header.h +++ b/tools/perf/util/header.h @@ -12,19 +12,29 @@ struct perf_header_attr { off_t id_offset; }; +#define HEADER_TRACE_INFO 1 + +#define HEADER_FEAT_BITS 256 + +typedef typeof(u64[HEADER_FEAT_BITS / 8]) feat_mask_t; + struct perf_header { - int frozen; - int attrs, size; + int frozen; + int attrs, size; struct perf_header_attr **attr; - s64 attr_offset; - u64 data_offset; - u64 data_size; - u64 event_offset; - u64 event_size; - u64 trace_info_offset; - u64 trace_info_size; + s64 attr_offset; + u64 data_offset; + u64 data_size; + u64 event_offset; + u64 event_size; + feat_mask_t adds_features; }; +static inline unsigned long *perf_header__adds_mask(struct perf_header *self) +{ + return (unsigned long *)(void *)&self->adds_features; +} + struct perf_header *perf_header__read(int fd); void perf_header__write(struct perf_header *self, int fd); @@ -42,7 +52,7 @@ void perf_header_attr__add_id(struct perf_header_attr *self, u64 id); u64 perf_header__sample_type(struct perf_header *header); struct perf_event_attr * perf_header__find_attr(u64 id, struct perf_header *header); -void perf_header__set_trace_info(void); +void perf_header__feat_trace_info(struct perf_header *header); struct perf_header *perf_header__new(void); -- cgit v1.2.3-18-g5258 From db9f11e36d0125a5e3e595ea9ef2e4b89f7e8737 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Sat, 17 Oct 2009 17:57:18 +0200 Subject: perf tools: Use DECLARE_BITMAP instead of an open-coded array Use DECLARE_BITMAP instead of an open coded array for our bitmap of featured sections. This makes the array an unsigned long instead of a u64 but since we use a 256 bits bitmap, the array size shouldn't vary between different boxes. Signed-off-by: Frederic Weisbecker Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Mike Galbraith Cc: Paul Mackerras Cc: Steven Rostedt LKML-Reference: <1255795038-13751-1-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar --- tools/perf/util/header.c | 22 +++++++++------------- tools/perf/util/header.h | 11 +++-------- tools/perf/util/include/asm/asm-offsets.h | 1 + tools/perf/util/include/linux/types.h | 8 +++++++- 4 files changed, 20 insertions(+), 22 deletions(-) create mode 100644 tools/perf/util/include/asm/asm-offsets.h (limited to 'tools/perf') diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index 171d51b6f35..622c60e4525 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -8,8 +8,6 @@ #include "../perf.h" #include "trace-event.h" -#include - /* * Create new perf.data header attribute: */ @@ -143,12 +141,12 @@ struct perf_file_header { struct perf_file_section attrs; struct perf_file_section data; struct perf_file_section event_types; - feat_mask_t adds_features; + DECLARE_BITMAP(adds_features, HEADER_FEAT_BITS); }; void perf_header__feat_trace_info(struct perf_header *header) { - set_bit(HEADER_TRACE_INFO, perf_header__adds_mask(header)); + set_bit(HEADER_TRACE_INFO, header->adds_features); } static void do_write(int fd, void *buf, size_t size) @@ -168,7 +166,7 @@ static void perf_header__adds_write(struct perf_header *self, int fd) { struct perf_file_section trace_sec; u64 cur_offset = lseek(fd, 0, SEEK_CUR); - unsigned long *feat_mask = perf_header__adds_mask(self); + unsigned long *feat_mask = self->adds_features; if (test_bit(HEADER_TRACE_INFO, feat_mask)) { /* Write trace info */ @@ -250,7 +248,7 @@ void perf_header__write(struct perf_header *self, int fd) }, }; - memcpy(&f_header.adds_features, &self->adds_features, sizeof(feat_mask_t)); + memcpy(&f_header.adds_features, &self->adds_features, sizeof(self->adds_features)); lseek(fd, 0, SEEK_SET); do_write(fd, &f_header, sizeof(f_header)); @@ -276,7 +274,7 @@ static void do_read(int fd, void *buf, size_t size) static void perf_header__adds_read(struct perf_header *self, int fd) { - const unsigned long *feat_mask = perf_header__adds_mask(self); + const unsigned long *feat_mask = self->adds_features; if (test_bit(HEADER_TRACE_INFO, feat_mask)) { struct perf_file_section trace_sec; @@ -306,11 +304,9 @@ struct perf_header *perf_header__read(int fd) if (f_header.size != sizeof(f_header)) { /* Support the previous format */ - if (f_header.size == offsetof(typeof(f_header), adds_features)) { - unsigned long *mask = (unsigned long *)(void *) - &f_header.adds_features; - bitmap_zero(mask, HEADER_FEAT_BITS); - } else + if (f_header.size == offsetof(typeof(f_header), adds_features)) + bitmap_zero(f_header.adds_features, HEADER_FEAT_BITS); + else die("incompatible file format"); } nr_attrs = f_header.attrs.size / sizeof(f_attr); @@ -346,7 +342,7 @@ struct perf_header *perf_header__read(int fd) event_count = f_header.event_types.size / sizeof(struct perf_trace_event_type); } - memcpy(&self->adds_features, &f_header.adds_features, sizeof(feat_mask_t)); + memcpy(&self->adds_features, &f_header.adds_features, sizeof(f_header.adds_features)); perf_header__adds_read(self, fd); diff --git a/tools/perf/util/header.h b/tools/perf/util/header.h index 0eb4a9126b7..2ea9dfb1236 100644 --- a/tools/perf/util/header.h +++ b/tools/perf/util/header.h @@ -5,6 +5,8 @@ #include #include "types.h" +#include + struct perf_header_attr { struct perf_event_attr attr; int ids, size; @@ -16,8 +18,6 @@ struct perf_header_attr { #define HEADER_FEAT_BITS 256 -typedef typeof(u64[HEADER_FEAT_BITS / 8]) feat_mask_t; - struct perf_header { int frozen; int attrs, size; @@ -27,14 +27,9 @@ struct perf_header { u64 data_size; u64 event_offset; u64 event_size; - feat_mask_t adds_features; + DECLARE_BITMAP(adds_features, HEADER_FEAT_BITS); }; -static inline unsigned long *perf_header__adds_mask(struct perf_header *self) -{ - return (unsigned long *)(void *)&self->adds_features; -} - struct perf_header *perf_header__read(int fd); void perf_header__write(struct perf_header *self, int fd); diff --git a/tools/perf/util/include/asm/asm-offsets.h b/tools/perf/util/include/asm/asm-offsets.h new file mode 100644 index 00000000000..ed538942523 --- /dev/null +++ b/tools/perf/util/include/asm/asm-offsets.h @@ -0,0 +1 @@ +/* stub */ diff --git a/tools/perf/util/include/linux/types.h b/tools/perf/util/include/linux/types.h index ed538942523..858a38d0843 100644 --- a/tools/perf/util/include/linux/types.h +++ b/tools/perf/util/include/linux/types.h @@ -1 +1,7 @@ -/* stub */ +#ifndef _PERF_LINUX_TYPES_H_ +#define _PERF_LINUX_TYPES_H_ + +#define DECLARE_BITMAP(name,bits) \ + unsigned long name[BITS_TO_LONGS(bits)] + +#endif -- cgit v1.2.3-18-g5258 From 1abc7f5500fff8422f34826a006648d8741d83d3 Mon Sep 17 00:00:00 2001 From: Randy Dunlap Date: Sun, 18 Oct 2009 19:20:24 -0700 Subject: perf tools: Display better error messages on missing packages Check for libelf headers and glibc headers separately so that the error message correctly identifies which package installation is missing/needed. Signed-off-by: Randy Dunlap Cc: paulus@samba.org Cc: a.p.zijlstra@chello.nl Cc: efault@gmx.de Cc: fweisbec@gmail.com Cc: Arnaldo Carvalho de Melo LKML-Reference: <4ADBCCE8.3060300@oracle.com> Signed-off-by: Ingo Molnar --- tools/perf/Makefile | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) (limited to 'tools/perf') diff --git a/tools/perf/Makefile b/tools/perf/Makefile index 2400e5068a2..db89a6de9d0 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -431,8 +431,12 @@ ifeq ($(uname_S),Darwin) PTHREAD_LIBS = endif +ifneq ($(shell sh -c "(echo '\#include '; echo 'int main(void) { const char * version = gnu_get_libc_version(); return (long)version; }') | $(CC) -x c - $(ALL_CFLAGS) -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -o /dev/null $(ALL_LDFLAGS) > /dev/null 2>&1 && echo y"), y) + msg := $(error No gnu/libc-version.h found, please install glibc-dev[el]); +endif + ifneq ($(shell sh -c "(echo '\#include '; echo 'int main(void) { Elf * elf = elf_begin(0, ELF_C_READ_MMAP, 0); return (long)elf; }') | $(CC) -x c - $(ALL_CFLAGS) -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -o /dev/null $(ALL_LDFLAGS) > /dev/null 2>&1 && echo y"), y) - msg := $(error No libelf.h/libelf found, please install libelf-dev/elfutils-libelf-devel and glibc-dev[el]); + msg := $(error No libelf.h/libelf found, please install libelf-dev/elfutils-libelf-devel); endif ifdef NO_DEMANGLE -- cgit v1.2.3-18-g5258 From 12133afffcc7140eea915b1572189a2ea0cf7b0e Mon Sep 17 00:00:00 2001 From: Tim Blechmann Date: Mon, 19 Oct 2009 12:03:33 +0200 Subject: perf stat: Add branch performance events to default output Adds performance event information about branches and branch misses to the default output of perf stat. Signed-off-by: Tim Blechmann Cc: Paul Mackerras Cc: Peter Zijlstra LKML-Reference: <4ADC3975.8050109@klingt.org> Signed-off-by: Ingo Molnar --- tools/perf/builtin-stat.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'tools/perf') diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c index c37368343ff..95a55eaf72f 100644 --- a/tools/perf/builtin-stat.c +++ b/tools/perf/builtin-stat.c @@ -59,6 +59,8 @@ static struct perf_event_attr default_attrs[] = { { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_INSTRUCTIONS }, { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CACHE_REFERENCES}, { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CACHE_MISSES }, + { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_BRANCH_INSTRUCTIONS}, + { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_BRANCH_MISSES }, }; -- cgit v1.2.3-18-g5258 From 56aab464ff6232bcc2f53b26576983dc83f75db7 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Mon, 19 Oct 2009 13:27:08 +0200 Subject: perf stat: Re-align the default_attrs[] array Clean up the array definition to be vertically aligned. No functional effects. Cc: Tim Blechmann Cc: Paul Mackerras Cc: Peter Zijlstra LKML-Reference: <4ADC3975.8050109@klingt.org> Signed-off-by: Ingo Molnar --- tools/perf/builtin-stat.c | 2 ++ 1 files changed, 2 insertions(+), 0 deletions(-) diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c index c373683..95a55ea 100644 --- a/tools/perf/builtin-stat.c +++ b/tools/perf/builtin-stat.c @@ -59,6 +59,8 @@ static struct perf_event_attr default_attrs[] = { { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_INSTRUCTIONS }, { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CACHE_REFERENCES}, { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CACHE_MISSES }, + { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_BRANCH_INSTRUCTIONS}, + { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_BRANCH_MISSES }, }; --- tools/perf/builtin-stat.c | 22 +++++++++++----------- 1 file changed, 11 insertions(+), 11 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c index 95a55eaf72f..90e0a268343 100644 --- a/tools/perf/builtin-stat.c +++ b/tools/perf/builtin-stat.c @@ -50,17 +50,17 @@ static struct perf_event_attr default_attrs[] = { - { .type = PERF_TYPE_SOFTWARE, .config = PERF_COUNT_SW_TASK_CLOCK }, - { .type = PERF_TYPE_SOFTWARE, .config = PERF_COUNT_SW_CONTEXT_SWITCHES}, - { .type = PERF_TYPE_SOFTWARE, .config = PERF_COUNT_SW_CPU_MIGRATIONS }, - { .type = PERF_TYPE_SOFTWARE, .config = PERF_COUNT_SW_PAGE_FAULTS }, - - { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CPU_CYCLES }, - { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_INSTRUCTIONS }, - { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CACHE_REFERENCES}, - { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CACHE_MISSES }, - { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_BRANCH_INSTRUCTIONS}, - { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_BRANCH_MISSES }, + { .type = PERF_TYPE_SOFTWARE, .config = PERF_COUNT_SW_TASK_CLOCK }, + { .type = PERF_TYPE_SOFTWARE, .config = PERF_COUNT_SW_CONTEXT_SWITCHES }, + { .type = PERF_TYPE_SOFTWARE, .config = PERF_COUNT_SW_CPU_MIGRATIONS }, + { .type = PERF_TYPE_SOFTWARE, .config = PERF_COUNT_SW_PAGE_FAULTS }, + + { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CPU_CYCLES }, + { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_INSTRUCTIONS }, + { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CACHE_REFERENCES }, + { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CACHE_MISSES }, + { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_BRANCH_INSTRUCTIONS }, + { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_BRANCH_MISSES }, }; -- cgit v1.2.3-18-g5258 From dd86e72abdbc4b436471af5a97927c6145f5298c Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Mon, 19 Oct 2009 13:33:03 +0200 Subject: perf stat: Count branches first Count branches first, cache-misses second. The reason is that on x86 branches are not counted by all counters on all CPUs. Before: Performance counter stats for 'ls': 0.756653 task-clock-msecs # 0.802 CPUs 0 context-switches # 0.000 M/sec 0 CPU-migrations # 0.000 M/sec 250 page-faults # 0.330 M/sec 2375725 cycles # 3139.781 M/sec 1628129 instructions # 0.685 IPC 19643 cache-references # 25.960 M/sec 4608 cache-misses # 6.090 M/sec 342532 branches # 452.694 M/sec branch-misses 0.000943356 seconds time elapsed After: Performance counter stats for 'ls': 1.056734 task-clock-msecs # 0.859 CPUs 0 context-switches # 0.000 M/sec 0 CPU-migrations # 0.000 M/sec 259 page-faults # 0.245 M/sec 3345932 cycles # 3166.295 M/sec 3074090 instructions # 0.919 IPC 616928 branches # 583.806 M/sec 39279 branch-misses # 6.367 % 21312 cache-references # 20.168 M/sec 3661 cache-misses # 3.464 M/sec 0.001230551 seconds time elapsed (also prettify the printout of branch misses, in case it's getting scaled.) Cc: Tim Blechmann Cc: Paul Mackerras Cc: Peter Zijlstra LKML-Reference: <4ADC3975.8050109@klingt.org> Signed-off-by: Ingo Molnar --- tools/perf/builtin-stat.c | 2 ++ 1 files changed, 2 insertions(+), 0 deletions(-) diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c index c373683..95a55ea 100644 --- a/tools/perf/builtin-stat.c +++ b/tools/perf/builtin-stat.c @@ -59,6 +59,8 @@ static struct perf_event_attr default_attrs[] = { { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_INSTRUCTIONS }, { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CACHE_REFERENCES}, { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CACHE_MISSES }, + { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_BRANCH_INSTRUCTIONS}, + { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_BRANCH_MISSES }, }; --- tools/perf/builtin-stat.c | 20 ++++++++++---------- 1 files changed, 10 insertions(+), 10 deletions(-) diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c index 95a55ea..90e0a26 100644 --- a/tools/perf/builtin-stat.c +++ b/tools/perf/builtin-stat.c @@ -50,17 +50,17 @@ static struct perf_event_attr default_attrs[] = { - { .type = PERF_TYPE_SOFTWARE, .config = PERF_COUNT_SW_TASK_CLOCK }, - { .type = PERF_TYPE_SOFTWARE, .config = PERF_COUNT_SW_CONTEXT_SWITCHES}, - { .type = PERF_TYPE_SOFTWARE, .config = PERF_COUNT_SW_CPU_MIGRATIONS }, - { .type = PERF_TYPE_SOFTWARE, .config = PERF_COUNT_SW_PAGE_FAULTS }, - - { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CPU_CYCLES }, - { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_INSTRUCTIONS }, - { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CACHE_REFERENCES}, - { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CACHE_MISSES }, - { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_BRANCH_INSTRUCTIONS}, - { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_BRANCH_MISSES }, + { .type = PERF_TYPE_SOFTWARE, .config = PERF_COUNT_SW_TASK_CLOCK }, + { .type = PERF_TYPE_SOFTWARE, .config = PERF_COUNT_SW_CONTEXT_SWITCHES }, + { .type = PERF_TYPE_SOFTWARE, .config = PERF_COUNT_SW_CPU_MIGRATIONS }, + { .type = PERF_TYPE_SOFTWARE, .config = PERF_COUNT_SW_PAGE_FAULTS }, + + { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CPU_CYCLES }, + { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_INSTRUCTIONS }, + { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CACHE_REFERENCES }, + { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CACHE_MISSES }, + { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_BRANCH_INSTRUCTIONS }, + { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_BRANCH_MISSES }, }; --- tools/perf/builtin-stat.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c index 90e0a268343..c6df3770b87 100644 --- a/tools/perf/builtin-stat.c +++ b/tools/perf/builtin-stat.c @@ -57,10 +57,10 @@ static struct perf_event_attr default_attrs[] = { { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CPU_CYCLES }, { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_INSTRUCTIONS }, - { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CACHE_REFERENCES }, - { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CACHE_MISSES }, { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_BRANCH_INSTRUCTIONS }, { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_BRANCH_MISSES }, + { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CACHE_REFERENCES }, + { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CACHE_MISSES }, }; @@ -363,7 +363,7 @@ static void abs_printout(int counter, double avg) if (total) ratio = avg * 100 / total; - fprintf(stderr, " # %10.3f %% ", ratio); + fprintf(stderr, " # %10.3f %% ", ratio); } else { total = avg_stats(&runtime_nsecs_stats); -- cgit v1.2.3-18-g5258 From 20639c15d2e78f180d398a6b6422880fac3258bb Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 19 Oct 2009 15:11:36 -0200 Subject: perf tools: Add missing tools/perf/util/include/string.h To cure a bunch of: In file included from util/include/linux/bitmap.h:1, from util/header.h:8, from builtin-trace.c:7: util/include/../../../../include/linux/bitmap.h:8:26: error: linux/string.h: No such file or directory make: *** [builtin-trace.o] Error 1 make: *** Waiting for unfinished jobs.... Signed-off-by: Arnaldo Carvalho de Melo Acked-by: Frederic Weisbecker Cc: Steven Rostedt Cc: Peter Zijlstra LKML-Reference: <1255972296-11500-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar --- tools/perf/Makefile | 1 + tools/perf/util/include/linux/string.h | 1 + 2 files changed, 2 insertions(+) create mode 100644 tools/perf/util/include/linux/string.h (limited to 'tools/perf') diff --git a/tools/perf/Makefile b/tools/perf/Makefile index db89a6de9d0..3b154f17d95 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -329,6 +329,7 @@ LIB_H += ../../include/linux/perf_event.h LIB_H += ../../include/linux/rbtree.h LIB_H += ../../include/linux/list.h LIB_H += util/include/linux/list.h +LIB_H += util/include/linux/string.h LIB_H += perf.h LIB_H += util/event.h LIB_H += util/types.h diff --git a/tools/perf/util/include/linux/string.h b/tools/perf/util/include/linux/string.h new file mode 100644 index 00000000000..3b2f5900276 --- /dev/null +++ b/tools/perf/util/include/linux/string.h @@ -0,0 +1 @@ +#include -- cgit v1.2.3-18-g5258 From 79b9ad361be8c6f3eeea97dd3883e8bcfa989333 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 19 Oct 2009 15:31:31 -0200 Subject: perf tools: Add bunch of missing headers to LIB_H Build dependencies were not properly mapped out. Signed-off-by: Arnaldo Carvalho de Melo Cc: Peter Zijlstra Cc: Frederic Weisbecker LKML-Reference: <1255973491-11626-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar --- tools/perf/Makefile | 17 +++++++++++++++++ 1 file changed, 17 insertions(+) (limited to 'tools/perf') diff --git a/tools/perf/Makefile b/tools/perf/Makefile index 3b154f17d95..64c6b7b57d8 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -328,8 +328,25 @@ LIB_FILE=libperf.a LIB_H += ../../include/linux/perf_event.h LIB_H += ../../include/linux/rbtree.h LIB_H += ../../include/linux/list.h +LIB_H += util/include/linux/bitmap.h +LIB_H += util/include/linux/bitops.h +LIB_H += util/include/linux/compiler.h +LIB_H += util/include/linux/ctype.h +LIB_H += util/include/linux/kernel.h LIB_H += util/include/linux/list.h +LIB_H += util/include/linux/module.h +LIB_H += util/include/linux/poison.h +LIB_H += util/include/linux/prefetch.h +LIB_H += util/include/linux/rbtree.h LIB_H += util/include/linux/string.h +LIB_H += util/include/linux/types.h +LIB_H += util/include/asm/asm-offsets.h +LIB_H += util/include/asm/bitops.h +LIB_H += util/include/asm/byteorder.h +LIB_H += util/include/asm/swab.h +LIB_H += util/include/asm/system.h +LIB_H += util/include/asm/types.h +LIB_H += util/include/asm/uaccess.h LIB_H += perf.h LIB_H += util/event.h LIB_H += util/types.h -- cgit v1.2.3-18-g5258 From bbe2987bea26a684ff11d887dfc4cf39b22c27a2 Mon Sep 17 00:00:00 2001 From: Arjan van de Ven Date: Tue, 20 Oct 2009 07:09:39 +0900 Subject: perf timechart: Add a process filter During the Kernel Summit demo of perf/ftrace/timechart, there was a feature request to have a process filter for timechart so that you can zoom into one or a few processes that you are really interested in. This patch adds basic support for this feature, the -p (--process) option now can select a PID or a process name to be shown. Multiple -p options are allowed, and the combined set will be included in the output. Signed-off-by: Arjan van de Ven Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker LKML-Reference: <20091020070939.7d0fb8a7@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/Documentation/perf-timechart.txt | 5 +- tools/perf/builtin-timechart.c | 105 +++++++++++++++++++++++++++- 2 files changed, 106 insertions(+), 4 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/Documentation/perf-timechart.txt b/tools/perf/Documentation/perf-timechart.txt index a7910099d6f..4b1788355ec 100644 --- a/tools/perf/Documentation/perf-timechart.txt +++ b/tools/perf/Documentation/perf-timechart.txt @@ -31,9 +31,12 @@ OPTIONS -w:: --width=:: Select the width of the SVG file (default: 1000) --p:: +-P:: --power-only:: Only output the CPU power section of the diagram +-p:: +--process:: + Select the processes to display, by name or PID SEE ALSO diff --git a/tools/perf/builtin-timechart.c b/tools/perf/builtin-timechart.c index e8a510d935e..34fad57087f 100644 --- a/tools/perf/builtin-timechart.c +++ b/tools/perf/builtin-timechart.c @@ -153,6 +153,17 @@ static struct wake_event *wake_events; struct sample_wrapper *all_samples; + +struct process_filter; +struct process_filter { + char *name; + int pid; + struct process_filter *next; +}; + +static struct process_filter *process_filter; + + static struct per_pid *find_create_pid(int pid) { struct per_pid *cursor = all_data; @@ -763,11 +774,11 @@ static void draw_wakeups(void) c = p->all; while (c) { if (c->Y && c->start_time <= we->time && c->end_time >= we->time) { - if (p->pid == we->waker) { + if (p->pid == we->waker && !from) { from = c->Y; task_from = strdup(c->comm); } - if (p->pid == we->wakee) { + if (p->pid == we->wakee && !to) { to = c->Y; task_to = strdup(c->comm); } @@ -882,12 +893,89 @@ static void draw_process_bars(void) } } +static void add_process_filter(const char *string) +{ + struct process_filter *filt; + int pid; + + pid = strtoull(string, NULL, 10); + filt = malloc(sizeof(struct process_filter)); + if (!filt) + return; + + filt->name = strdup(string); + filt->pid = pid; + filt->next = process_filter; + + process_filter = filt; +} + +static int passes_filter(struct per_pid *p, struct per_pidcomm *c) +{ + struct process_filter *filt; + if (!process_filter) + return 1; + + filt = process_filter; + while (filt) { + if (filt->pid && p->pid == filt->pid) + return 1; + if (strcmp(filt->name, c->comm) == 0) + return 1; + filt = filt->next; + } + return 0; +} + +static int determine_display_tasks_filtered(void) +{ + struct per_pid *p; + struct per_pidcomm *c; + int count = 0; + + p = all_data; + while (p) { + p->display = 0; + if (p->start_time == 1) + p->start_time = first_time; + + /* no exit marker, task kept running to the end */ + if (p->end_time == 0) + p->end_time = last_time; + + c = p->all; + + while (c) { + c->display = 0; + + if (c->start_time == 1) + c->start_time = first_time; + + if (passes_filter(p, c)) { + c->display = 1; + p->display = 1; + count++; + } + + if (c->end_time == 0) + c->end_time = last_time; + + c = c->next; + } + p = p->next; + } + return count; +} + static int determine_display_tasks(u64 threshold) { struct per_pid *p; struct per_pidcomm *c; int count = 0; + if (process_filter) + return determine_display_tasks_filtered(); + p = all_data; while (p) { p->display = 0; @@ -1153,6 +1241,14 @@ static int __cmd_record(int argc, const char **argv) return cmd_record(i, rec_argv, NULL); } +static int +parse_process(const struct option *opt __used, const char *arg, int __used unset) +{ + if (arg) + add_process_filter(arg); + return 0; +} + static const struct option options[] = { OPT_STRING('i', "input", &input_name, "file", "input file name"), @@ -1160,8 +1256,11 @@ static const struct option options[] = { "output file name"), OPT_INTEGER('w', "width", &svg_page_width, "page width"), - OPT_BOOLEAN('p', "power-only", &power_only, + OPT_BOOLEAN('P', "power-only", &power_only, "output power data only"), + OPT_CALLBACK('p', "process", NULL, "process", + "process selector. Pass a pid or process name.", + parse_process), OPT_END() }; -- cgit v1.2.3-18-g5258 From ed52ce2e3c33dc7626a40fa2da766d1a6460e543 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 19 Oct 2009 17:17:57 -0200 Subject: perf tools: Add ->unmap_ip operation to struct map We need this because we get section relative addresses when reading the symtabs, but when a tool like 'perf annotate' needs to match these address to what 'objdump -dS' produces we need the address + section back again. So in annotate now we look at the 'struct hist_entry' instances (that weren't really being used) so that we iterate only over the symbols that had some hit and get the map where that particular hit happened so that we can get the right address to match with annotate. Verified that at least: perf annotate mmap_read_counter # Uses the ~/bin/perf binary perf annotate --vmlinux /home/acme/git/build/perf/vmlinux intel_pmu_enable_all on a 'perf record perf top' session seems to work. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: Peter Zijlstra Cc: Paul Mackerras Cc: Mike Galbraith LKML-Reference: <1255979877-12533-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar --- tools/perf/builtin-annotate.c | 62 +++++++++++++++++++++++++++++-------------- tools/perf/util/event.h | 8 +++++- tools/perf/util/map.c | 6 +++-- tools/perf/util/symbol.c | 8 +++--- 4 files changed, 58 insertions(+), 26 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c index 56ba71658d7..06f10278b28 100644 --- a/tools/perf/builtin-annotate.c +++ b/tools/perf/builtin-annotate.c @@ -58,9 +58,12 @@ static void hist_hit(struct hist_entry *he, u64 ip) return; sym_size = sym->end - sym->start; - ip = he->map->map_ip(he->map, ip); offset = ip - sym->start; + if (verbose) + fprintf(stderr, "%s: ip=%Lx\n", __func__, + he->map->unmap_ip(he->map, ip)); + if (offset >= sym_size) return; @@ -83,8 +86,7 @@ static int hist_entry__add(struct thread *thread, struct map *map, count, level, &hit); if (he == NULL) return -ENOMEM; - if (hit) - hist_hit(he, ip); + hist_hit(he, ip); return 0; } @@ -260,14 +262,15 @@ process_event(event_t *event, unsigned long offset, unsigned long head) return 0; } -static int -parse_line(FILE *file, struct symbol *sym, u64 len) +static int parse_line(FILE *file, struct hist_entry *he, u64 len) { + struct symbol *sym = he->sym; char *line = NULL, *tmp, *tmp2; static const char *prev_line; static const char *prev_color; unsigned int offset; size_t line_len; + u64 start; s64 line_ip; int ret; char *c; @@ -304,6 +307,8 @@ parse_line(FILE *file, struct symbol *sym, u64 len) line_ip = -1; } + start = he->map->unmap_ip(he->map, sym->start); + if (line_ip != -1) { const char *path = NULL; unsigned int hits = 0; @@ -311,7 +316,7 @@ parse_line(FILE *file, struct symbol *sym, u64 len) const char *color; struct sym_ext *sym_ext = sym->priv; - offset = line_ip - sym->start; + offset = line_ip - start; if (offset < len) hits = sym->hist[offset]; @@ -390,8 +395,10 @@ static void free_source_line(struct symbol *sym, int len) /* Get the filename:line for the colored entries */ static void -get_source_line(struct symbol *sym, int len, const char *filename) +get_source_line(struct hist_entry *he, int len, const char *filename) { + struct symbol *sym = he->sym; + u64 start; int i; char cmd[PATH_MAX * 2]; struct sym_ext *sym_ext; @@ -404,6 +411,7 @@ get_source_line(struct symbol *sym, int len, const char *filename) return; sym_ext = sym->priv; + start = he->map->unmap_ip(he->map, sym->start); for (i = 0; i < len; i++) { char *path = NULL; @@ -415,7 +423,7 @@ get_source_line(struct symbol *sym, int len, const char *filename) if (sym_ext[i].percent <= 0.5) continue; - offset = sym->start + i; + offset = start + i; sprintf(cmd, "addr2line -e %s %016llx", filename, offset); fp = popen(cmd, "r"); if (!fp) @@ -465,8 +473,11 @@ static void print_summary(const char *filename) } } -static void annotate_sym(struct dso *dso, struct symbol *sym) +static void annotate_sym(struct hist_entry *he) { + struct map *map = he->map; + struct dso *dso = map->dso; + struct symbol *sym = he->sym; const char *filename = dso->long_name, *d_filename; u64 len; char command[PATH_MAX*2]; @@ -475,6 +486,12 @@ static void annotate_sym(struct dso *dso, struct symbol *sym) if (!filename) return; + if (verbose) + fprintf(stderr, "%s: filename=%s, sym=%s, start=%Lx, end=%Lx\n", + __func__, filename, sym->name, + map->unmap_ip(map, sym->start), + map->unmap_ip(map, sym->end)); + if (full_paths) d_filename = filename; else @@ -483,7 +500,7 @@ static void annotate_sym(struct dso *dso, struct symbol *sym) len = sym->end - sym->start; if (print_line) { - get_source_line(sym, len, filename); + get_source_line(he, len, filename); print_summary(filename); } @@ -496,7 +513,8 @@ static void annotate_sym(struct dso *dso, struct symbol *sym) dso, dso->long_name, sym, sym->name); sprintf(command, "objdump --start-address=0x%016Lx --stop-address=0x%016Lx -dS %s|grep -v %s", - sym->start, sym->end, filename, filename); + map->unmap_ip(map, sym->start), map->unmap_ip(map, sym->end), + filename, filename); if (verbose >= 3) printf("doing: %s\n", command); @@ -506,7 +524,7 @@ static void annotate_sym(struct dso *dso, struct symbol *sym) return; while (!feof(file)) { - if (parse_line(file, sym, len) < 0) + if (parse_line(file, he, len) < 0) break; } @@ -518,18 +536,22 @@ static void annotate_sym(struct dso *dso, struct symbol *sym) static void find_annotations(void) { struct rb_node *nd; - struct dso *dso; int count = 0; - list_for_each_entry(dso, &dsos, node) { + for (nd = rb_first(&output_hists); nd; nd = rb_next(nd)) { + struct hist_entry *he = rb_entry(nd, struct hist_entry, rb_node); - for (nd = rb_first(&dso->syms); nd; nd = rb_next(nd)) { - struct symbol *sym = rb_entry(nd, struct symbol, rb_node); + if (he->sym && he->sym->hist) { + annotate_sym(he); + count++; + /* + * Since we have a hist_entry per IP for the same + * symbol, free he->sym->hist to signal we already + * processed this symbol. + */ + free(he->sym->hist); + he->sym->hist = NULL; - if (sym->hist) { - annotate_sym(dso, sym); - count++; - } } } diff --git a/tools/perf/util/event.h b/tools/perf/util/event.h index c2e62be6279..6b5be56a827 100644 --- a/tools/perf/util/event.h +++ b/tools/perf/util/event.h @@ -82,6 +82,7 @@ struct map { u64 end; u64 pgoff; u64 (*map_ip)(struct map *, u64); + u64 (*unmap_ip)(struct map *, u64); struct dso *dso; }; @@ -90,7 +91,12 @@ static inline u64 map__map_ip(struct map *map, u64 ip) return ip - map->start + map->pgoff; } -static inline u64 vdso__map_ip(struct map *map __used, u64 ip) +static inline u64 map__unmap_ip(struct map *map, u64 ip) +{ + return ip + map->start - map->pgoff; +} + +static inline u64 identity__map_ip(struct map *map __used, u64 ip) { return ip; } diff --git a/tools/perf/util/map.c b/tools/perf/util/map.c index 804e0238273..4e203d144f9 100644 --- a/tools/perf/util/map.c +++ b/tools/perf/util/map.c @@ -54,9 +54,11 @@ static int strcommon(const char *pathname, char *cwd, int cwdlen) goto out_delete; if (self->dso == vdso || anon) - self->map_ip = vdso__map_ip; - else + self->map_ip = self->unmap_ip = identity__map_ip; + else { self->map_ip = map__map_ip; + self->unmap_ip = map__unmap_ip; + } } return self; out_delete: diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index faa84f5d4f5..3350119f690 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -337,7 +337,7 @@ static int kernel_maps__split_kallsyms(symbol_filter_t filter, int use_modules) return -1; } - map->map_ip = vdso__map_ip; + map->map_ip = map->unmap_ip = identity__map_ip; kernel_maps__insert(map); ++kernel_range; } @@ -790,7 +790,8 @@ static int dso__load_sym(struct dso *self, struct map *map, const char *name, dso__delete(curr_dso); goto out_elf_end; } - curr_map->map_ip = vdso__map_ip; + curr_map->map_ip = identity__map_ip; + curr_map->unmap_ip = identity__map_ip; curr_dso->origin = DSO__ORIG_KERNEL; kernel_maps__insert(curr_map); dsos__add(curr_dso); @@ -1158,6 +1159,7 @@ static struct map *map__new2(u64 start, struct dso *dso) self->pgoff = 0; self->dso = dso; self->map_ip = map__map_ip; + self->unmap_ip = map__unmap_ip; RB_CLEAR_NODE(&self->rb_node); } return self; @@ -1259,7 +1261,7 @@ int dsos__load_kernel(const char *vmlinux, unsigned int sym_priv_size, if (kernel_map == NULL) goto out_delete_dso; - kernel_map->map_ip = vdso__map_ip; + kernel_map->map_ip = kernel_map->unmap_ip = identity__map_ip; if (use_modules && dsos__load_modules(sym_priv_size) < 0) { fprintf(stderr, "Failed to load list of modules in use! " -- cgit v1.2.3-18-g5258 From e42049926ebdcae24fdfdc8f0e3ff8f05f24a60b Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Tue, 20 Oct 2009 14:25:40 -0200 Subject: perf annotate: Use the sym_priv_size area for the histogram We have this sym_priv_size mechanism for attaching private areas to struct symbol entries but annotate wasn't using it, adding private areas to struct symbol in addition to a ->priv pointer. Scrap all that and use the sym_priv_size mechanism. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: Peter Zijlstra Cc: Paul Mackerras Cc: Mike Galbraith LKML-Reference: <1256055940-19511-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar --- tools/perf/builtin-annotate.c | 109 ++++++++++++++++++++++++++++++------------ tools/perf/builtin-report.c | 2 +- tools/perf/util/data_map.c | 2 +- tools/perf/util/event.h | 8 +++- tools/perf/util/map.c | 21 +++++++- tools/perf/util/symbol.c | 56 ++++++++-------------- tools/perf/util/symbol.h | 17 +++---- 7 files changed, 132 insertions(+), 83 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c index 06f10278b28..245692530de 100644 --- a/tools/perf/builtin-annotate.c +++ b/tools/perf/builtin-annotate.c @@ -37,12 +37,44 @@ static int print_line; static unsigned long page_size; static unsigned long mmap_window = 32; +struct sym_hist { + u64 sum; + u64 ip[0]; +}; + struct sym_ext { struct rb_node node; double percent; char *path; }; +struct sym_priv { + struct sym_hist *hist; + struct sym_ext *ext; +}; + +static const char *sym_hist_filter; + +static int symbol_filter(struct map *map, struct symbol *sym) +{ + if (strcmp(sym->name, sym_hist_filter) == 0) { + struct sym_priv *priv = dso__sym_priv(map->dso, sym); + const int size = (sizeof(*priv->hist) + + (sym->end - sym->start) * sizeof(u64)); + + priv->hist = malloc(size); + if (priv->hist) + memset(priv->hist, 0, size); + return 0; + } + /* + * FIXME: We should really filter it out, as we don't want to go thru symbols + * we're not interested, and if a DSO ends up with no symbols, delete it too, + * but right now the kernel loading routines in symbol.c bail out if no symbols + * are found, fix it later. + */ + return 0; +} /* * collect histogram counts @@ -51,10 +83,16 @@ static void hist_hit(struct hist_entry *he, u64 ip) { unsigned int sym_size, offset; struct symbol *sym = he->sym; + struct sym_priv *priv; + struct sym_hist *h; he->count++; - if (!sym || !sym->hist) + if (!sym || !he->map) + return; + + priv = dso__sym_priv(he->map->dso, sym); + if (!priv->hist) return; sym_size = sym->end - sym->start; @@ -67,15 +105,16 @@ static void hist_hit(struct hist_entry *he, u64 ip) if (offset >= sym_size) return; - sym->hist_sum++; - sym->hist[offset]++; + h = priv->hist; + h->sum++; + h->ip[offset]++; if (verbose >= 3) printf("%p %s: count++ [ip: %p, %08Lx] => %Ld\n", (void *)(unsigned long)he->sym->start, he->sym->name, (void *)(unsigned long)ip, ip - he->sym->start, - sym->hist[offset]); + h->ip[offset]); } static int hist_entry__add(struct thread *thread, struct map *map, @@ -162,7 +201,9 @@ got_map: static int process_mmap_event(event_t *event, unsigned long offset, unsigned long head) { - struct map *map = map__new(&event->mmap, NULL, 0); + struct map *map = map__new(&event->mmap, NULL, 0, + sizeof(struct sym_priv), symbol_filter, + verbose); struct thread *thread = threads__findnew(event->mmap.pid); dump_printf("%p [%p]: PERF_RECORD_MMAP %d: [%p(%p) @ %p]: %s\n", @@ -314,17 +355,19 @@ static int parse_line(FILE *file, struct hist_entry *he, u64 len) unsigned int hits = 0; double percent = 0.0; const char *color; - struct sym_ext *sym_ext = sym->priv; + struct sym_priv *priv = dso__sym_priv(he->map->dso, sym); + struct sym_ext *sym_ext = priv->ext; + struct sym_hist *h = priv->hist; offset = line_ip - start; if (offset < len) - hits = sym->hist[offset]; + hits = h->ip[offset]; if (offset < len && sym_ext) { path = sym_ext[offset].path; percent = sym_ext[offset].percent; - } else if (sym->hist_sum) - percent = 100.0 * hits / sym->hist_sum; + } else if (h->sum) + percent = 100.0 * hits / h->sum; color = get_percent_color(percent); @@ -377,9 +420,10 @@ static void insert_source_line(struct sym_ext *sym_ext) rb_insert_color(&sym_ext->node, &root_sym_ext); } -static void free_source_line(struct symbol *sym, int len) +static void free_source_line(struct hist_entry *he, int len) { - struct sym_ext *sym_ext = sym->priv; + struct sym_priv *priv = dso__sym_priv(he->map->dso, he->sym); + struct sym_ext *sym_ext = priv->ext; int i; if (!sym_ext) @@ -389,7 +433,7 @@ static void free_source_line(struct symbol *sym, int len) free(sym_ext[i].path); free(sym_ext); - sym->priv = NULL; + priv->ext = NULL; root_sym_ext = RB_ROOT; } @@ -402,15 +446,16 @@ get_source_line(struct hist_entry *he, int len, const char *filename) int i; char cmd[PATH_MAX * 2]; struct sym_ext *sym_ext; + struct sym_priv *priv = dso__sym_priv(he->map->dso, sym); + struct sym_hist *h = priv->hist; - if (!sym->hist_sum) + if (!h->sum) return; - sym->priv = calloc(len, sizeof(struct sym_ext)); - if (!sym->priv) + sym_ext = priv->ext = calloc(len, sizeof(struct sym_ext)); + if (!priv->ext) return; - sym_ext = sym->priv; start = he->map->unmap_ip(he->map, sym->start); for (i = 0; i < len; i++) { @@ -419,7 +464,7 @@ get_source_line(struct hist_entry *he, int len, const char *filename) u64 offset; FILE *fp; - sym_ext[i].percent = 100.0 * sym->hist[i] / sym->hist_sum; + sym_ext[i].percent = 100.0 * h->ip[i] / h->sum; if (sym_ext[i].percent <= 0.5) continue; @@ -530,7 +575,7 @@ static void annotate_sym(struct hist_entry *he) pclose(file); if (print_line) - free_source_line(sym, len); + free_source_line(he, len); } static void find_annotations(void) @@ -540,19 +585,23 @@ static void find_annotations(void) for (nd = rb_first(&output_hists); nd; nd = rb_next(nd)) { struct hist_entry *he = rb_entry(nd, struct hist_entry, rb_node); + struct sym_priv *priv; - if (he->sym && he->sym->hist) { - annotate_sym(he); - count++; - /* - * Since we have a hist_entry per IP for the same - * symbol, free he->sym->hist to signal we already - * processed this symbol. - */ - free(he->sym->hist); - he->sym->hist = NULL; + if (he->sym == NULL) + continue; - } + priv = dso__sym_priv(he->map->dso, he->sym); + if (priv->hist == NULL) + continue; + + annotate_sym(he); + count++; + /* + * Since we have a hist_entry per IP for the same symbol, free + * he->sym->hist to signal we already processed this symbol. + */ + free(priv->hist); + priv->hist = NULL; } if (!count) @@ -593,7 +642,7 @@ static int __cmd_annotate(void) exit(0); } - if (load_kernel() < 0) { + if (load_kernel(sizeof(struct sym_priv), symbol_filter) < 0) { perror("failed to load kernel symbols"); return EXIT_FAILURE; } diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index a4f8cc20915..bee207ce589 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -680,7 +680,7 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) static int process_mmap_event(event_t *event, unsigned long offset, unsigned long head) { - struct map *map = map__new(&event->mmap, cwd, cwdlen); + struct map *map = map__new(&event->mmap, cwd, cwdlen, 0, NULL, verbose); struct thread *thread = threads__findnew(event->mmap.pid); dump_printf("%p [%p]: PERF_RECORD_MMAP %d/%d: [%p(%p) @ %p]: %s\n", diff --git a/tools/perf/util/data_map.c b/tools/perf/util/data_map.c index 242b0555ab9..18accb8fee4 100644 --- a/tools/perf/util/data_map.c +++ b/tools/perf/util/data_map.c @@ -130,7 +130,7 @@ int mmap_dispatch_perf_file(struct perf_header **pheader, if (curr_handler->sample_type_check(sample_type) < 0) exit(-1); - if (load_kernel() < 0) { + if (load_kernel(0, NULL) < 0) { perror("failed to load kernel symbols"); return EXIT_FAILURE; } diff --git a/tools/perf/util/event.h b/tools/perf/util/event.h index 6b5be56a827..db59c8bbe49 100644 --- a/tools/perf/util/event.h +++ b/tools/perf/util/event.h @@ -101,7 +101,13 @@ static inline u64 identity__map_ip(struct map *map __used, u64 ip) return ip; } -struct map *map__new(struct mmap_event *event, char *cwd, int cwdlen); +struct symbol; + +typedef int (*symbol_filter_t)(struct map *map, struct symbol *sym); + +struct map *map__new(struct mmap_event *event, char *cwd, int cwdlen, + unsigned int sym_priv_size, symbol_filter_t filter, + int v); struct map *map__clone(struct map *self); int map__overlap(struct map *l, struct map *r); size_t map__fprintf(struct map *self, FILE *fp); diff --git a/tools/perf/util/map.c b/tools/perf/util/map.c index 4e203d144f9..55079c0200e 100644 --- a/tools/perf/util/map.c +++ b/tools/perf/util/map.c @@ -3,6 +3,7 @@ #include #include #include +#include "debug.h" static inline int is_anon_memory(const char *filename) { @@ -19,7 +20,9 @@ static int strcommon(const char *pathname, char *cwd, int cwdlen) return n; } - struct map *map__new(struct mmap_event *event, char *cwd, int cwdlen) +struct map *map__new(struct mmap_event *event, char *cwd, int cwdlen, + unsigned int sym_priv_size, symbol_filter_t filter, + int v) { struct map *self = malloc(sizeof(*self)); @@ -27,6 +30,7 @@ static int strcommon(const char *pathname, char *cwd, int cwdlen) const char *filename = event->filename; char newfilename[PATH_MAX]; int anon; + bool new_dso; if (cwd) { int n = strcommon(filename, cwd, cwdlen); @@ -49,10 +53,23 @@ static int strcommon(const char *pathname, char *cwd, int cwdlen) self->end = event->start + event->len; self->pgoff = event->pgoff; - self->dso = dsos__findnew(filename); + self->dso = dsos__findnew(filename, sym_priv_size, &new_dso); if (self->dso == NULL) goto out_delete; + if (new_dso) { + int nr = dso__load(self->dso, self, filter, v); + + if (nr < 0) + eprintf("Failed to open %s, continuing " + "without symbols\n", + self->dso->long_name); + else if (nr == 0) + eprintf("No symbols found in %s, maybe " + "install a debug package?\n", + self->dso->long_name); + } + if (self->dso == vdso || anon) self->map_ip = self->unmap_ip = identity__map_ip; else { diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 3350119f690..0a4898480d6 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -11,8 +11,6 @@ #include #include -const char *sym_hist_filter; - enum dso_origin { DSO__ORIG_KERNEL = 0, DSO__ORIG_JAVA_JIT, @@ -86,22 +84,16 @@ static struct symbol *symbol__new(u64 start, u64 len, const char *name, if (!self) return NULL; - if (v > 2) - printf("new symbol: %016Lx [%08lx]: %s, hist: %p\n", - start, (unsigned long)len, name, self->hist); - - self->hist = NULL; - self->hist_sum = 0; - - if (sym_hist_filter && !strcmp(name, sym_hist_filter)) - self->hist = calloc(sizeof(u64), len); - if (priv_size) { memset(self, 0, priv_size); self = ((void *)self) + priv_size; } self->start = start; self->end = len ? start + len - 1 : start; + + if (v > 2) + printf("%s: %s %#Lx-%#Lx\n", __func__, name, start, self->end); + memcpy(self->name, name, namelen); return self; @@ -919,7 +911,8 @@ char dso__symtab_origin(const struct dso *self) return origin[self->origin]; } -int dso__load(struct dso *self, struct map *map, symbol_filter_t filter, int v) +int dso__load(struct dso *self, struct map *map, + symbol_filter_t filter, int v) { int size = PATH_MAX; char *name = malloc(size), *build_id = NULL; @@ -1335,33 +1328,21 @@ static struct dso *dsos__find(const char *name) return NULL; } -struct dso *dsos__findnew(const char *name) +struct dso *dsos__findnew(const char *name, unsigned int sym_priv_size, + bool *is_new) { struct dso *dso = dsos__find(name); - int nr; - - if (dso) - return dso; - - dso = dso__new(name, 0); - if (!dso) - goto out_delete_dso; - nr = dso__load(dso, NULL, NULL, verbose); - if (nr < 0) { - eprintf("Failed to open: %s\n", name); - goto out_delete_dso; - } - if (!nr) - eprintf("No symbols found in: %s, maybe install a debug package?\n", name); - - dsos__add(dso); + if (!dso) { + dso = dso__new(name, sym_priv_size); + if (dso) { + dsos__add(dso); + *is_new = true; + } + } else + *is_new = false; return dso; - -out_delete_dso: - dso__delete(dso); - return NULL; } void dsos__fprintf(FILE *fp) @@ -1372,9 +1353,10 @@ void dsos__fprintf(FILE *fp) dso__fprintf(pos, fp); } -int load_kernel(void) +int load_kernel(unsigned int sym_priv_size, symbol_filter_t filter) { - if (dsos__load_kernel(vmlinux_name, 0, NULL, verbose, modules) <= 0) + if (dsos__load_kernel(vmlinux_name, sym_priv_size, + filter, verbose, modules) <= 0) return -1; vdso = dso__new("[vdso]", 0); diff --git a/tools/perf/util/symbol.h b/tools/perf/util/symbol.h index 2e4522edeb0..c2a777de9b7 100644 --- a/tools/perf/util/symbol.h +++ b/tools/perf/util/symbol.h @@ -2,6 +2,7 @@ #define __PERF_SYMBOL 1 #include +#include #include "types.h" #include #include @@ -35,9 +36,6 @@ struct symbol { struct rb_node rb_node; u64 start; u64 end; - u64 hist_sum; - u64 *hist; - void *priv; char name[0]; }; @@ -54,10 +52,6 @@ struct dso { char name[0]; }; -extern const char *sym_hist_filter; - -typedef int (*symbol_filter_t)(struct map *map, struct symbol *sym); - struct dso *dso__new(const char *name, unsigned int sym_priv_size); void dso__delete(struct dso *self); @@ -70,15 +64,16 @@ struct symbol *dso__find_symbol(struct dso *self, u64 ip); int dsos__load_kernel(const char *vmlinux, unsigned int sym_priv_size, symbol_filter_t filter, int verbose, int modules); -int dso__load(struct dso *self, struct map *map, symbol_filter_t filter, - int verbose); -struct dso *dsos__findnew(const char *name); +struct dso *dsos__findnew(const char *name, unsigned int sym_priv_size, + bool *is_new); +int dso__load(struct dso *self, struct map *map, + symbol_filter_t filter, int v); void dsos__fprintf(FILE *fp); size_t dso__fprintf(struct dso *self, FILE *fp); char dso__symtab_origin(const struct dso *self); -int load_kernel(void); +int load_kernel(unsigned int sym_priv_size, symbol_filter_t filter); void symbol__init(void); -- cgit v1.2.3-18-g5258 From 8f0b037398a909ccf703ad5f5803066db6327f22 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Tue, 20 Oct 2009 15:08:29 -0200 Subject: perf annotate: Remove requirement of passing a symbol name If the user doesn't pass a symbol name to annotate, it will annotate all the symbols that have hits, in order, just like 'perf report -s comm,dso,symbol'. This is a natural followup patch to the one that uses output_hists to find the symbols with hits. The common case is to annotate the first few entries at the top of a perf report, so lets type less characters. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: Peter Zijlstra Cc: Paul Mackerras Cc: Mike Galbraith LKML-Reference: <1256058509-19678-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar --- tools/perf/builtin-annotate.c | 11 ++--------- 1 file changed, 2 insertions(+), 9 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c index 245692530de..99bac6aa72c 100644 --- a/tools/perf/builtin-annotate.c +++ b/tools/perf/builtin-annotate.c @@ -57,7 +57,8 @@ static const char *sym_hist_filter; static int symbol_filter(struct map *map, struct symbol *sym) { - if (strcmp(sym->name, sym_hist_filter) == 0) { + if (sym_hist_filter == NULL || + strcmp(sym->name, sym_hist_filter) == 0) { struct sym_priv *priv = dso__sym_priv(map->dso, sym); const int size = (sizeof(*priv->hist) + (sym->end - sym->start) * sizeof(u64)); @@ -581,7 +582,6 @@ static void annotate_sym(struct hist_entry *he) static void find_annotations(void) { struct rb_node *nd; - int count = 0; for (nd = rb_first(&output_hists); nd; nd = rb_next(nd)) { struct hist_entry *he = rb_entry(nd, struct hist_entry, rb_node); @@ -595,7 +595,6 @@ static void find_annotations(void) continue; annotate_sym(he); - count++; /* * Since we have a hist_entry per IP for the same symbol, free * he->sym->hist to signal we already processed this symbol. @@ -603,9 +602,6 @@ static void find_annotations(void) free(priv->hist); priv->hist = NULL; } - - if (!count) - printf(" Error: symbol '%s' not present amongst the samples.\n", sym_hist_filter); } static int __cmd_annotate(void) @@ -793,9 +789,6 @@ int cmd_annotate(int argc, const char **argv, const char *prefix __used) sym_hist_filter = argv[0]; } - if (!sym_hist_filter) - usage_with_options(annotate_usage, options); - setup_pager(); if (field_sep && *field_sep == '.') { -- cgit v1.2.3-18-g5258 From c88e4bf60de6253a048cf4e6b3b0715e543e0460 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Tue, 20 Oct 2009 15:54:55 -0200 Subject: perf top: Fix symbol annotation We need to use map->unmap_ip() here too to match section relative symbol address to the absolute address needed to match objdump -dS addresses. Reported-by: Mike Galbraith Signed-off-by: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1256061295-19835-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar --- tools/perf/builtin-top.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index cc662863030..fa20345a0ab 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -141,7 +141,8 @@ static void parse_source(struct sym_entry *syme) sprintf(command, "objdump --start-address=0x%016Lx " "--stop-address=0x%016Lx -dS %s", - sym->start, sym->end, path); + map->unmap_ip(map, sym->start), + map->unmap_ip(map, sym->end), path); file = popen(command, "r"); if (!file) @@ -173,11 +174,11 @@ static void parse_source(struct sym_entry *syme) if (strlen(src->line)>8 && src->line[8] == ':') { src->eip = strtoull(src->line, NULL, 16); - src->eip += map->start; + src->eip = map->unmap_ip(map, src->eip); } if (strlen(src->line)>8 && src->line[16] == ':') { src->eip = strtoull(src->line, NULL, 16); - src->eip += map->start; + src->eip = map->unmap_ip(map, src->eip); } } pclose(file); -- cgit v1.2.3-18-g5258 From 60d526f7fa6246b8e32d5b45610d625a5608d988 Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Tue, 20 Oct 2009 19:19:34 -0400 Subject: perf tools: Add 'make DEBUG=1' to remove the -O6 cflag When using gdb to debug perf, it is practically impossible to use when perf is compiled with -O6. For developers, this patch adds the DEBUG feature to the make command line so that a developer can easily remove the optimization flag. LKML-Reference: <1255590330.8392.446.camel@twins> Signed-off-by: Steven Rostedt Cc: Peter Zijlstra Cc: Frederic Weisbecker Cc: Arnaldo Carvalho de Melo LKML-Reference: <20091020232033.984323261@goodmis.org> Signed-off-by: Ingo Molnar --- tools/perf/Makefile | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) (limited to 'tools/perf') diff --git a/tools/perf/Makefile b/tools/perf/Makefile index 64c6b7b57d8..65e6e52fe37 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -201,7 +201,14 @@ EXTRA_WARNINGS := $(EXTRA_WARNINGS) -Wold-style-definition EXTRA_WARNINGS := $(EXTRA_WARNINGS) -Wstrict-prototypes EXTRA_WARNINGS := $(EXTRA_WARNINGS) -Wdeclaration-after-statement -CFLAGS = $(MBITS) -ggdb3 -Wall -Wextra -std=gnu99 -Werror -O6 -fstack-protector-all -D_FORTIFY_SOURCE=2 $(EXTRA_WARNINGS) +ifeq ("$(origin DEBUG)", "command line") + PERF_DEBUG = $(DEBUG) +endif +ifndef PERF_DEBUG + CFLAGS_OPTIMIZE = -O6 +endif + +CFLAGS = $(MBITS) -ggdb3 -Wall -Wextra -std=gnu99 -Werror $(CFLAGS_OPTIMIZE) -fstack-protector-all -D_FORTIFY_SOURCE=2 $(EXTRA_WARNINGS) LDFLAGS = -lpthread -lrt -lelf -lm ALL_CFLAGS = $(CFLAGS) ALL_LDFLAGS = $(LDFLAGS) -- cgit v1.2.3-18-g5258 From 4e3b799d7dbb2a12ca8dca8d3594d32095772973 Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Tue, 20 Oct 2009 19:19:35 -0400 Subject: perf tools: Use strsep() over strtok_r() for parsing single line The second argument in the strtok_r() function is not to be used generically and can have different implementations. Currently the function parsing of the perf trace code uses the second argument to copy data from. This can crash the tool or just have unpredictable results. The correct solution is to use strsep() which has a defined result. I also added a check to see if the result was correct, and will break out of the loop in case it fails to parse as expected. Reported-by: Arnaldo Carvalho de Melo Signed-off-by: Steven Rostedt Cc: Peter Zijlstra Cc: Frederic Weisbecker LKML-Reference: <20091020232034.237814877@goodmis.org> Signed-off-by: Ingo Molnar --- tools/perf/util/trace-event-parse.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/trace-event-parse.c b/tools/perf/util/trace-event-parse.c index 4b61b497040..eae56050308 100644 --- a/tools/perf/util/trace-event-parse.c +++ b/tools/perf/util/trace-event-parse.c @@ -286,16 +286,19 @@ void parse_ftrace_printk(char *file, unsigned int size __unused) char *line; char *next = NULL; char *addr_str; - char *fmt; int i; line = strtok_r(file, "\n", &next); while (line) { + addr_str = strsep(&line, ":"); + if (!line) { + warning("error parsing print strings"); + break; + } item = malloc_or_die(sizeof(*item)); - addr_str = strtok_r(line, ":", &fmt); item->addr = strtoull(addr_str, NULL, 16); /* fmt still has a space, skip it */ - item->printk = strdup(fmt+1); + item->printk = strdup(line+1); item->next = list; list = item; line = strtok_r(NULL, "\n", &next); -- cgit v1.2.3-18-g5258 From af0a6fa46388e1e0c2d1a672aad84f8f6ef0b20b Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Thu, 22 Oct 2009 23:23:22 +0200 Subject: perf tools: Fix missing top level callchain While recursively printing the branches of each callchains, we forget to display the root. It is never printed. Say we have: symbol f1 f2 | -------- f3 | f4 | ---------f5 f6 Actually we never see that, instead it displays: symbol | --------- f3 | f4 | --------- f5 f6 However f1 is always the same than "symbol" and if we are sorting by symbols first then "symbol", f1 and f2 will be well aligned like in the above example, so displaying f1 looks redundant here. But if we are sorting by something else first (dso, comm, etc...), displaying f1 doesn't look redundant but rather necessary because the symbol is not well aligned anymore with its callchain: comm dso symbol f1 f2 | --------- [...] And we want the callchain to be obvious. So we fix the bug by printing the root branch, but we also filter its first entry if we are sorting by symbols first. Reported-by: Anton Blanchard Signed-off-by: Frederic Weisbecker Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Mike Galbraith Cc: Paul Mackerras LKML-Reference: <1256246604-17156-1-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar --- tools/perf/builtin-report.c | 40 +++++++++++++++++++++++++++++++++------- tools/perf/util/sort.c | 10 +++++++--- tools/perf/util/sort.h | 1 + 3 files changed, 41 insertions(+), 10 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index bee207ce589..3d8c52220f1 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -122,8 +122,8 @@ static void init_rem_hits(void) } static size_t -callchain__fprintf_graph(FILE *fp, struct callchain_node *self, - u64 total_samples, int depth, int depth_mask) +__callchain__fprintf_graph(FILE *fp, struct callchain_node *self, + u64 total_samples, int depth, int depth_mask) { struct rb_node *node, *next; struct callchain_node *child; @@ -174,9 +174,9 @@ callchain__fprintf_graph(FILE *fp, struct callchain_node *self, new_total, cumul); } - ret += callchain__fprintf_graph(fp, child, new_total, - depth + 1, - new_depth_mask | (1 << depth)); + ret += __callchain__fprintf_graph(fp, child, new_total, + depth + 1, + new_depth_mask | (1 << depth)); node = next; } @@ -196,6 +196,33 @@ callchain__fprintf_graph(FILE *fp, struct callchain_node *self, return ret; } +static size_t +callchain__fprintf_graph(FILE *fp, struct callchain_node *self, + u64 total_samples) +{ + struct callchain_list *chain; + int i = 0; + int ret = 0; + + list_for_each_entry(chain, &self->val, list) { + if (chain->ip >= PERF_CONTEXT_MAX) + continue; + + if (!i++ && sort_by_sym_first) + continue; + + if (chain->sym) + ret += fprintf(fp, " %s\n", chain->sym->name); + else + ret += fprintf(fp, " %p\n", + (void *)(long)chain->ip); + } + + ret += __callchain__fprintf_graph(fp, self, total_samples, 1, 1); + + return ret; +} + static size_t callchain__fprintf_flat(FILE *fp, struct callchain_node *self, u64 total_samples) @@ -244,8 +271,7 @@ hist_entry_callchain__fprintf(FILE *fp, struct hist_entry *self, break; case CHAIN_GRAPH_ABS: /* Falldown */ case CHAIN_GRAPH_REL: - ret += callchain__fprintf_graph(fp, chain, - total_samples, 1, 1); + ret += callchain__fprintf_graph(fp, chain, total_samples); case CHAIN_NONE: default: break; diff --git a/tools/perf/util/sort.c b/tools/perf/util/sort.c index 40c9acd41ca..60ced707bd6 100644 --- a/tools/perf/util/sort.c +++ b/tools/perf/util/sort.c @@ -5,8 +5,9 @@ char default_parent_pattern[] = "^sys_|^do_page_fault"; char *parent_pattern = default_parent_pattern; char default_sort_order[] = "comm,dso,symbol"; char *sort_order = default_sort_order; -int sort__need_collapse = 0; -int sort__has_parent = 0; +int sort__need_collapse = 0; +int sort__has_parent = 0; +int sort_by_sym_first; unsigned int dsos__col_width; unsigned int comms__col_width; @@ -265,6 +266,10 @@ int sort_dimension__add(const char *tok) sort__has_parent = 1; } + if (list_empty(&hist_entry__sort_list) && + !strcmp(sd->name, "symbol")) + sort_by_sym_first = true; + list_add_tail(&sd->entry->list, &hist_entry__sort_list); sd->taken = 1; @@ -273,4 +278,3 @@ int sort_dimension__add(const char *tok) return -ESRCH; } - diff --git a/tools/perf/util/sort.h b/tools/perf/util/sort.h index 13806d782af..24c2b709f0d 100644 --- a/tools/perf/util/sort.h +++ b/tools/perf/util/sort.h @@ -39,6 +39,7 @@ extern struct sort_entry sort_parent; extern unsigned int dsos__col_width; extern unsigned int comms__col_width; extern unsigned int threads__col_width; +extern int sort_by_sym_first; struct hist_entry { struct rb_node rb_node; -- cgit v1.2.3-18-g5258 From a4fb581b15949cfd10b64c8af37bc106e95307f3 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Thu, 22 Oct 2009 23:23:23 +0200 Subject: perf tools: Bind callchains to the first sort dimension column Currently, the callchains are displayed using a constant left margin. So depending on the current sort dimension configuration, callchains may appear to be well attached to the first sort dimension column field which is mostly the case, except when the first dimension of sorting is done by comm, because these are right aligned. This patch binds the callchain to the first letter in the first column, whatever type of column it is (dso, comm, symbol). Before: 0.80% perf [k] __lock_acquire __lock_acquire lock_acquire | |--58.33%-- _spin_lock | | | |--28.57%-- inotify_should_send_event | | fsnotify | | __fsnotify_parent After: 0.80% perf [k] __lock_acquire __lock_acquire lock_acquire | |--58.33%-- _spin_lock | | | |--28.57%-- inotify_should_send_event | | fsnotify | | __fsnotify_parent Also, for clarity, we don't put anymore the callchain as is but: - If we have a top level ancestor in the callchain, start it with a first ascii hook. Before: 0.80% perf [kernel] [k] __lock_acquire __lock_acquire lock_acquire | |--58.33%-- _spin_lock | | | |--28.57%-- inotify_should_send_event | | fsnotify [..] [..] After: 0.80% perf [kernel] [k] __lock_acquire | --- __lock_acquire lock_acquire | |--58.33%-- _spin_lock | | | |--28.57%-- inotify_should_send_event | | fsnotify [..] [..] - Otherwise, if we have several top level ancestors, then display these like we did before: 1.69% Xorg | |--21.21%-- vread_hpet | 0x7fffd85b46fc | 0x7fffd85b494d | 0x7f4fafb4e54d | |--15.15%-- exaOffscreenAlloc | |--9.09%-- I830WaitLpRing Signed-off-by: Frederic Weisbecker Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Mike Galbraith Cc: Paul Mackerras Cc: Anton Blanchard LKML-Reference: <1256246604-17156-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar --- tools/perf/builtin-report.c | 82 ++++++++++++++++++++++++++++++++++----------- tools/perf/util/sort.c | 18 +++++++--- tools/perf/util/sort.h | 10 +++++- tools/perf/util/thread.c | 11 ++++++ tools/perf/util/thread.h | 2 ++ 5 files changed, 99 insertions(+), 24 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index 3d8c52220f1..72d58421223 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -59,12 +59,28 @@ static struct perf_header *header; static u64 sample_type; -static size_t ipchain__fprintf_graph_line(FILE *fp, int depth, int depth_mask) + +static size_t +callchain__fprintf_left_margin(FILE *fp, int left_margin) +{ + int i; + int ret; + + ret = fprintf(fp, " "); + + for (i = 0; i < left_margin; i++) + ret += fprintf(fp, " "); + + return ret; +} + +static size_t ipchain__fprintf_graph_line(FILE *fp, int depth, int depth_mask, + int left_margin) { int i; size_t ret = 0; - ret += fprintf(fp, "%s", " "); + ret += callchain__fprintf_left_margin(fp, left_margin); for (i = 0; i < depth; i++) if (depth_mask & (1 << i)) @@ -79,12 +95,12 @@ static size_t ipchain__fprintf_graph_line(FILE *fp, int depth, int depth_mask) static size_t ipchain__fprintf_graph(FILE *fp, struct callchain_list *chain, int depth, int depth_mask, int count, u64 total_samples, - int hits) + int hits, int left_margin) { int i; size_t ret = 0; - ret += fprintf(fp, "%s", " "); + ret += callchain__fprintf_left_margin(fp, left_margin); for (i = 0; i < depth; i++) { if (depth_mask & (1 << i)) ret += fprintf(fp, "|"); @@ -123,7 +139,8 @@ static void init_rem_hits(void) static size_t __callchain__fprintf_graph(FILE *fp, struct callchain_node *self, - u64 total_samples, int depth, int depth_mask) + u64 total_samples, int depth, int depth_mask, + int left_margin) { struct rb_node *node, *next; struct callchain_node *child; @@ -164,7 +181,8 @@ __callchain__fprintf_graph(FILE *fp, struct callchain_node *self, * But we keep the older depth mask for the line seperator * to keep the level link until we reach the last child */ - ret += ipchain__fprintf_graph_line(fp, depth, depth_mask); + ret += ipchain__fprintf_graph_line(fp, depth, depth_mask, + left_margin); i = 0; list_for_each_entry(chain, &child->val, list) { if (chain->ip >= PERF_CONTEXT_MAX) @@ -172,11 +190,13 @@ __callchain__fprintf_graph(FILE *fp, struct callchain_node *self, ret += ipchain__fprintf_graph(fp, chain, depth, new_depth_mask, i++, new_total, - cumul); + cumul, + left_margin); } ret += __callchain__fprintf_graph(fp, child, new_total, depth + 1, - new_depth_mask | (1 << depth)); + new_depth_mask | (1 << depth), + left_margin); node = next; } @@ -190,17 +210,19 @@ __callchain__fprintf_graph(FILE *fp, struct callchain_node *self, ret += ipchain__fprintf_graph(fp, &rem_hits, depth, new_depth_mask, 0, new_total, - remaining); + remaining, left_margin); } return ret; } + static size_t callchain__fprintf_graph(FILE *fp, struct callchain_node *self, - u64 total_samples) + u64 total_samples, int left_margin) { struct callchain_list *chain; + bool printed = false; int i = 0; int ret = 0; @@ -208,17 +230,27 @@ callchain__fprintf_graph(FILE *fp, struct callchain_node *self, if (chain->ip >= PERF_CONTEXT_MAX) continue; - if (!i++ && sort_by_sym_first) + if (!i++ && sort__first_dimension == SORT_SYM) continue; + if (!printed) { + ret += callchain__fprintf_left_margin(fp, left_margin); + ret += fprintf(fp, "|\n"); + ret += callchain__fprintf_left_margin(fp, left_margin); + ret += fprintf(fp, "---"); + + left_margin += 3; + printed = true; + } else + ret += callchain__fprintf_left_margin(fp, left_margin); + if (chain->sym) - ret += fprintf(fp, " %s\n", chain->sym->name); + ret += fprintf(fp, " %s\n", chain->sym->name); else - ret += fprintf(fp, " %p\n", - (void *)(long)chain->ip); + ret += fprintf(fp, " %p\n", (void *)(long)chain->ip); } - ret += __callchain__fprintf_graph(fp, self, total_samples, 1, 1); + ret += __callchain__fprintf_graph(fp, self, total_samples, 1, 1, left_margin); return ret; } @@ -251,7 +283,7 @@ callchain__fprintf_flat(FILE *fp, struct callchain_node *self, static size_t hist_entry_callchain__fprintf(FILE *fp, struct hist_entry *self, - u64 total_samples) + u64 total_samples, int left_margin) { struct rb_node *rb_node; struct callchain_node *chain; @@ -271,7 +303,8 @@ hist_entry_callchain__fprintf(FILE *fp, struct hist_entry *self, break; case CHAIN_GRAPH_ABS: /* Falldown */ case CHAIN_GRAPH_REL: - ret += callchain__fprintf_graph(fp, chain, total_samples); + ret += callchain__fprintf_graph(fp, chain, total_samples, + left_margin); case CHAIN_NONE: default: break; @@ -316,8 +349,19 @@ hist_entry__fprintf(FILE *fp, struct hist_entry *self, u64 total_samples) ret += fprintf(fp, "\n"); - if (callchain) - hist_entry_callchain__fprintf(fp, self, total_samples); + if (callchain) { + int left_margin = 0; + + if (sort__first_dimension == SORT_COMM) { + se = list_first_entry(&hist_entry__sort_list, typeof(*se), + list); + left_margin = se->width ? *se->width : 0; + left_margin -= thread__comm_len(self->thread); + } + + hist_entry_callchain__fprintf(fp, self, total_samples, + left_margin); + } return ret; } diff --git a/tools/perf/util/sort.c b/tools/perf/util/sort.c index 60ced707bd6..b490354d1b2 100644 --- a/tools/perf/util/sort.c +++ b/tools/perf/util/sort.c @@ -7,7 +7,8 @@ char default_sort_order[] = "comm,dso,symbol"; char *sort_order = default_sort_order; int sort__need_collapse = 0; int sort__has_parent = 0; -int sort_by_sym_first; + +enum sort_type sort__first_dimension; unsigned int dsos__col_width; unsigned int comms__col_width; @@ -266,9 +267,18 @@ int sort_dimension__add(const char *tok) sort__has_parent = 1; } - if (list_empty(&hist_entry__sort_list) && - !strcmp(sd->name, "symbol")) - sort_by_sym_first = true; + if (list_empty(&hist_entry__sort_list)) { + if (!strcmp(sd->name, "pid")) + sort__first_dimension = SORT_PID; + else if (!strcmp(sd->name, "comm")) + sort__first_dimension = SORT_COMM; + else if (!strcmp(sd->name, "dso")) + sort__first_dimension = SORT_DSO; + else if (!strcmp(sd->name, "symbol")) + sort__first_dimension = SORT_SYM; + else if (!strcmp(sd->name, "parent")) + sort__first_dimension = SORT_PARENT; + } list_add_tail(&sd->entry->list, &hist_entry__sort_list); sd->taken = 1; diff --git a/tools/perf/util/sort.h b/tools/perf/util/sort.h index 24c2b709f0d..333e664ff45 100644 --- a/tools/perf/util/sort.h +++ b/tools/perf/util/sort.h @@ -39,7 +39,7 @@ extern struct sort_entry sort_parent; extern unsigned int dsos__col_width; extern unsigned int comms__col_width; extern unsigned int threads__col_width; -extern int sort_by_sym_first; +extern enum sort_type sort__first_dimension; struct hist_entry { struct rb_node rb_node; @@ -54,6 +54,14 @@ struct hist_entry { struct rb_root sorted_chain; }; +enum sort_type { + SORT_PID, + SORT_COMM, + SORT_DSO, + SORT_SYM, + SORT_PARENT +}; + /* * configurable sorting bits */ diff --git a/tools/perf/util/thread.c b/tools/perf/util/thread.c index f53fad7c0a8..8cb47f1d8a7 100644 --- a/tools/perf/util/thread.c +++ b/tools/perf/util/thread.c @@ -33,6 +33,17 @@ int thread__set_comm(struct thread *self, const char *comm) return self->comm ? 0 : -ENOMEM; } +int thread__comm_len(struct thread *self) +{ + if (!self->comm_len) { + if (!self->comm) + return 0; + self->comm_len = strlen(self->comm); + } + + return self->comm_len; +} + static size_t thread__fprintf(struct thread *self, FILE *fp) { struct rb_node *nd; diff --git a/tools/perf/util/thread.h b/tools/perf/util/thread.h index 1abef3b7455..53addd77ce8 100644 --- a/tools/perf/util/thread.h +++ b/tools/perf/util/thread.h @@ -12,9 +12,11 @@ struct thread { pid_t pid; char shortname[3]; char *comm; + int comm_len; }; int thread__set_comm(struct thread *self, const char *comm); +int thread__comm_len(struct thread *self); struct thread *threads__findnew(pid_t pid); struct thread *register_idle_thread(void); void thread__insert_map(struct thread *self, struct map *map); -- cgit v1.2.3-18-g5258 From 802da5f2289bbe363acef084805195c11f453c48 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Thu, 22 Oct 2009 23:23:24 +0200 Subject: perf tools: Drop asm/types.h wrapper Wrapping the kernel headers is dangerous when it comes to arch headers. Once we wrap asm/types.h, it will also replace the glibc asm/types.h, not only the kernel one. This results in build errors on some machines. Drop this wrapper and do its work from linux/types.h wrapper, also the glibc asm/types.h can already handle most of the type definition it was doing (typedef __u64, __u32, etc...). Todo: Check the others asm/*.h wrappers to prevent from other conflicts. Reported-by: Ingo Molnar Signed-off-by: Frederic Weisbecker Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Mike Galbraith Cc: Paul Mackerras Cc: Anton Blanchard LKML-Reference: <1256246604-17156-3-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar --- tools/perf/Makefile | 1 - tools/perf/util/include/asm/bitops.h | 12 ++++++++++++ tools/perf/util/include/asm/byteorder.h | 2 +- tools/perf/util/include/asm/types.h | 17 ----------------- tools/perf/util/include/linux/types.h | 2 ++ 5 files changed, 15 insertions(+), 19 deletions(-) delete mode 100644 tools/perf/util/include/asm/types.h (limited to 'tools/perf') diff --git a/tools/perf/Makefile b/tools/perf/Makefile index 65e6e52fe37..0a40c29b238 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -352,7 +352,6 @@ LIB_H += util/include/asm/bitops.h LIB_H += util/include/asm/byteorder.h LIB_H += util/include/asm/swab.h LIB_H += util/include/asm/system.h -LIB_H += util/include/asm/types.h LIB_H += util/include/asm/uaccess.h LIB_H += perf.h LIB_H += util/event.h diff --git a/tools/perf/util/include/asm/bitops.h b/tools/perf/util/include/asm/bitops.h index fbe4d921291..58e9817ffae 100644 --- a/tools/perf/util/include/asm/bitops.h +++ b/tools/perf/util/include/asm/bitops.h @@ -1,6 +1,18 @@ +#ifndef _PERF_ASM_BITOPS_H_ +#define _PERF_ASM_BITOPS_H_ + +#include +#include "../../types.h" +#include + +/* CHECKME: Not sure both always match */ +#define BITS_PER_LONG __WORDSIZE + #include "../../../../include/asm-generic/bitops/__fls.h" #include "../../../../include/asm-generic/bitops/fls.h" #include "../../../../include/asm-generic/bitops/fls64.h" #include "../../../../include/asm-generic/bitops/__ffs.h" #include "../../../../include/asm-generic/bitops/ffz.h" #include "../../../../include/asm-generic/bitops/hweight.h" + +#endif diff --git a/tools/perf/util/include/asm/byteorder.h b/tools/perf/util/include/asm/byteorder.h index 39f367cfaf5..b722abe3a62 100644 --- a/tools/perf/util/include/asm/byteorder.h +++ b/tools/perf/util/include/asm/byteorder.h @@ -1,2 +1,2 @@ -#include "../asm/types.h" +#include #include "../../../../include/linux/swab.h" diff --git a/tools/perf/util/include/asm/types.h b/tools/perf/util/include/asm/types.h deleted file mode 100644 index 06703c6cd50..00000000000 --- a/tools/perf/util/include/asm/types.h +++ /dev/null @@ -1,17 +0,0 @@ -#ifndef PERF_ASM_TYPES_H_ -#define PERF_ASM_TYPES_H_ - -#include -#include "../../types.h" -#include - -/* CHECKME: Not sure both always match */ -#define BITS_PER_LONG __WORDSIZE - -typedef u64 __u64; -typedef u32 __u32; -typedef u16 __u16; -typedef u8 __u8; -typedef s64 __s64; - -#endif /* PERF_ASM_TYPES_H_ */ diff --git a/tools/perf/util/include/linux/types.h b/tools/perf/util/include/linux/types.h index 858a38d0843..196862a81a2 100644 --- a/tools/perf/util/include/linux/types.h +++ b/tools/perf/util/include/linux/types.h @@ -1,6 +1,8 @@ #ifndef _PERF_LINUX_TYPES_H_ #define _PERF_LINUX_TYPES_H_ +#include + #define DECLARE_BITMAP(name,bits) \ unsigned long name[BITS_TO_LONGS(bits)] -- cgit v1.2.3-18-g5258 From 6beba7adbe092e63dfe8d09fbd1e3ec140474a13 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Wed, 21 Oct 2009 17:34:06 -0200 Subject: perf tools: Unify debug messages mechanisms We were using eprintf in some places, that looks at a global 'verbose' level, and at other places passing a 'v' parameter to specify the verbosity level, unify it by introducing pr_{err,warning,debug,etc}, just like in the kernel. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: Peter Zijlstra Cc: Paul Mackerras Cc: Mike Galbraith LKML-Reference: <1256153646-10097-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar --- tools/perf/builtin-annotate.c | 3 +- tools/perf/builtin-record.c | 2 +- tools/perf/builtin-report.c | 9 ++- tools/perf/builtin-sched.c | 4 +- tools/perf/builtin-timechart.c | 13 ++-- tools/perf/builtin-top.c | 2 +- tools/perf/builtin-trace.c | 4 +- tools/perf/util/callchain.c | 2 +- tools/perf/util/debug.c | 4 +- tools/perf/util/debug.h | 3 +- tools/perf/util/event.h | 3 +- tools/perf/util/header.c | 2 +- tools/perf/util/include/linux/kernel.h | 17 +++++ tools/perf/util/map.c | 17 ++--- tools/perf/util/symbol.c | 134 +++++++++++++++------------------ tools/perf/util/symbol.h | 5 +- tools/perf/util/thread.c | 6 +- 17 files changed, 114 insertions(+), 116 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c index 99bac6aa72c..6d63c2eea2c 100644 --- a/tools/perf/builtin-annotate.c +++ b/tools/perf/builtin-annotate.c @@ -203,8 +203,7 @@ static int process_mmap_event(event_t *event, unsigned long offset, unsigned long head) { struct map *map = map__new(&event->mmap, NULL, 0, - sizeof(struct sym_priv), symbol_filter, - verbose); + sizeof(struct sym_priv), symbol_filter); struct thread *thread = threads__findnew(event->mmap.pid); dump_printf("%p [%p]: PERF_RECORD_MMAP %d: [%p(%p) @ %p]: %s\n", diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index f0467ff0d8a..ac5ddfff445 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -630,7 +630,7 @@ static int __cmd_record(int argc, const char **argv) param.sched_priority = realtime_prio; if (sched_setscheduler(0, SCHED_FIFO, ¶m)) { - printf("Could not set realtime priority.\n"); + pr_err("Could not set realtime priority.\n"); exit(-1); } } diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index 72d58421223..b3d814b5455 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -689,7 +689,8 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) dump_printf("... chain: nr:%Lu\n", chain->nr); if (validate_chain(chain, event) < 0) { - eprintf("call-chain problem with event, skipping it.\n"); + pr_debug("call-chain problem with event, " + "skipping it.\n"); return 0; } @@ -700,7 +701,7 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) } if (thread == NULL) { - eprintf("problem processing %d event, skipping it.\n", + pr_debug("problem processing %d event, skipping it.\n", event->header.type); return -1; } @@ -738,7 +739,7 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) if (hist_entry__add(thread, map, sym, ip, chain, level, period)) { - eprintf("problem incrementing symbol count, skipping event\n"); + pr_debug("problem incrementing symbol count, skipping event\n"); return -1; } @@ -750,7 +751,7 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) static int process_mmap_event(event_t *event, unsigned long offset, unsigned long head) { - struct map *map = map__new(&event->mmap, cwd, cwdlen, 0, NULL, verbose); + struct map *map = map__new(&event->mmap, cwd, cwdlen, 0, NULL); struct thread *thread = threads__findnew(event->mmap.pid); dump_printf("%p [%p]: PERF_RECORD_MMAP %d/%d: [%p(%p) @ %p]: %s\n", diff --git a/tools/perf/builtin-sched.c b/tools/perf/builtin-sched.c index 807ca66e7a8..9a48d9626be 100644 --- a/tools/perf/builtin-sched.c +++ b/tools/perf/builtin-sched.c @@ -1666,8 +1666,8 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) (long long)period); if (thread == NULL) { - eprintf("problem processing %d event, skipping it.\n", - event->header.type); + pr_debug("problem processing %d event, skipping it.\n", + event->header.type); return -1; } diff --git a/tools/perf/builtin-timechart.c b/tools/perf/builtin-timechart.c index 34fad57087f..0a2f22261c3 100644 --- a/tools/perf/builtin-timechart.c +++ b/tools/perf/builtin-timechart.c @@ -1162,12 +1162,10 @@ more: size = event->header.size; if (!size || process_event(event) < 0) { - - printf("%p [%p]: skipping unknown header type: %d\n", - (void *)(offset + head), - (void *)(long)(event->header.size), - event->header.type); - + pr_warning("%p [%p]: skipping unknown header type: %d\n", + (void *)(offset + head), + (void *)(long)(event->header.size), + event->header.type); /* * assume we lost track of the stream, check alignment, and * increment a single u64 in the hope to catch on again 'soon'. @@ -1200,7 +1198,8 @@ done: write_svg_file(output_name); - printf("Written %2.1f seconds of trace to %s.\n", (last_time - first_time) / 1000000000.0, output_name); + pr_info("Written %2.1f seconds of trace to %s.\n", + (last_time - first_time) / 1000000000.0, output_name); return rc; } diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index fa20345a0ab..4a9fe228be2 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -809,7 +809,7 @@ static int symbol_filter(struct map *map, struct symbol *sym) static int parse_symbols(void) { if (dsos__load_kernel(vmlinux_name, sizeof(struct sym_entry), - symbol_filter, verbose, 1) <= 0) + symbol_filter, 1) <= 0) return -1; if (dump_symtab) diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-trace.c index 4c129ff0bb1..e566bbe3f22 100644 --- a/tools/perf/builtin-trace.c +++ b/tools/perf/builtin-trace.c @@ -81,8 +81,8 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) (long long)period); if (thread == NULL) { - eprintf("problem processing %d event, skipping it.\n", - event->header.type); + pr_debug("problem processing %d event, skipping it.\n", + event->header.type); return -1; } diff --git a/tools/perf/util/callchain.c b/tools/perf/util/callchain.c index 3b8380f1b47..b3b71258272 100644 --- a/tools/perf/util/callchain.c +++ b/tools/perf/util/callchain.c @@ -206,7 +206,7 @@ fill_node(struct callchain_node *node, struct ip_callchain *chain, } node->val_nr = chain->nr - start; if (!node->val_nr) - printf("Warning: empty node in callchain tree\n"); + pr_warning("Warning: empty node in callchain tree\n"); } static void diff --git a/tools/perf/util/debug.c b/tools/perf/util/debug.c index e8ca98fe0bd..28d520d5a1f 100644 --- a/tools/perf/util/debug.c +++ b/tools/perf/util/debug.c @@ -13,12 +13,12 @@ int verbose = 0; int dump_trace = 0; -int eprintf(const char *fmt, ...) +int eprintf(int level, const char *fmt, ...) { va_list args; int ret = 0; - if (verbose) { + if (verbose >= level) { va_start(args, fmt); ret = vfprintf(stderr, fmt, args); va_end(args); diff --git a/tools/perf/util/debug.h b/tools/perf/util/debug.h index 02d1fa1c246..e8b18a1f87a 100644 --- a/tools/perf/util/debug.h +++ b/tools/perf/util/debug.h @@ -5,7 +5,8 @@ extern int verbose; extern int dump_trace; -int eprintf(const char *fmt, ...) __attribute__((format(printf, 1, 2))); +int eprintf(int level, + const char *fmt, ...) __attribute__((format(printf, 2, 3))); int dump_printf(const char *fmt, ...) __attribute__((format(printf, 1, 2))); void trace_event(event_t *event); diff --git a/tools/perf/util/event.h b/tools/perf/util/event.h index db59c8bbe49..d972b4b0d38 100644 --- a/tools/perf/util/event.h +++ b/tools/perf/util/event.h @@ -106,8 +106,7 @@ struct symbol; typedef int (*symbol_filter_t)(struct map *map, struct symbol *sym); struct map *map__new(struct mmap_event *event, char *cwd, int cwdlen, - unsigned int sym_priv_size, symbol_filter_t filter, - int v); + unsigned int sym_priv_size, symbol_filter_t filter); struct map *map__clone(struct map *self); int map__overlap(struct map *l, struct map *r); size_t map__fprintf(struct map *self, FILE *fp); diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index 622c60e4525..7d26659b806 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -93,7 +93,7 @@ static struct perf_trace_event_type *events; void perf_header__push_event(u64 id, const char *name) { if (strlen(name) > MAX_EVENT_NAME) - printf("Event %s will be truncated\n", name); + pr_warning("Event %s will be truncated\n", name); if (!events) { events = malloc(sizeof(struct perf_trace_event_type)); diff --git a/tools/perf/util/include/linux/kernel.h b/tools/perf/util/include/linux/kernel.h index 4b9204d9b26..21c0274c02f 100644 --- a/tools/perf/util/include/linux/kernel.h +++ b/tools/perf/util/include/linux/kernel.h @@ -85,4 +85,21 @@ simple_strtoul(const char *nptr, char **endptr, int base) return strtoul(nptr, endptr, base); } +#ifndef pr_fmt +#define pr_fmt(fmt) fmt +#endif + +#define pr_err(fmt, ...) \ + do { fprintf(stderr, pr_fmt(fmt), ##__VA_ARGS__); } while (0) +#define pr_warning(fmt, ...) \ + do { fprintf(stderr, pr_fmt(fmt), ##__VA_ARGS__); } while (0) +#define pr_info(fmt, ...) \ + do { fprintf(stderr, pr_fmt(fmt), ##__VA_ARGS__); } while (0) +#define pr_debug(fmt, ...) \ + eprintf(1, pr_fmt(fmt), ##__VA_ARGS__) +#define pr_debugN(n, fmt, ...) \ + eprintf(n, pr_fmt(fmt), ##__VA_ARGS__) +#define pr_debug2(fmt, ...) pr_debugN(2, pr_fmt(fmt), ##__VA_ARGS__) +#define pr_debug3(fmt, ...) pr_debugN(3, pr_fmt(fmt), ##__VA_ARGS__) + #endif diff --git a/tools/perf/util/map.c b/tools/perf/util/map.c index 55079c0200e..c1c55682534 100644 --- a/tools/perf/util/map.c +++ b/tools/perf/util/map.c @@ -21,8 +21,7 @@ static int strcommon(const char *pathname, char *cwd, int cwdlen) } struct map *map__new(struct mmap_event *event, char *cwd, int cwdlen, - unsigned int sym_priv_size, symbol_filter_t filter, - int v) + unsigned int sym_priv_size, symbol_filter_t filter) { struct map *self = malloc(sizeof(*self)); @@ -58,16 +57,16 @@ struct map *map__new(struct mmap_event *event, char *cwd, int cwdlen, goto out_delete; if (new_dso) { - int nr = dso__load(self->dso, self, filter, v); + int nr = dso__load(self->dso, self, filter); if (nr < 0) - eprintf("Failed to open %s, continuing " - "without symbols\n", - self->dso->long_name); + pr_warning("Failed to open %s, continuing " + "without symbols\n", + self->dso->long_name); else if (nr == 0) - eprintf("No symbols found in %s, maybe " - "install a debug package?\n", - self->dso->long_name); + pr_warning("No symbols found in %s, maybe " + "install a debug package?\n", + self->dso->long_name); } if (self->dso == vdso || anon) diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 0a4898480d6..8f0208ce237 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -76,7 +76,7 @@ static void kernel_maps__fixup_end(void) } static struct symbol *symbol__new(u64 start, u64 len, const char *name, - unsigned int priv_size, int v) + unsigned int priv_size) { size_t namelen = strlen(name) + 1; struct symbol *self = calloc(1, priv_size + sizeof(*self) + namelen); @@ -91,8 +91,7 @@ static struct symbol *symbol__new(u64 start, u64 len, const char *name, self->start = start; self->end = len ? start + len - 1 : start; - if (v > 2) - printf("%s: %s %#Lx-%#Lx\n", __func__, name, start, self->end); + pr_debug3("%s: %s %#Lx-%#Lx\n", __func__, name, start, self->end); memcpy(self->name, name, namelen); @@ -209,7 +208,7 @@ size_t dso__fprintf(struct dso *self, FILE *fp) * so that we can in the next step set the symbol ->end address and then * call kernel_maps__split_kallsyms. */ -static int kernel_maps__load_all_kallsyms(int v) +static int kernel_maps__load_all_kallsyms(void) { char *line = NULL; size_t n; @@ -252,7 +251,7 @@ static int kernel_maps__load_all_kallsyms(int v) * Will fix up the end later, when we have all symbols sorted. */ sym = symbol__new(start, 0, symbol_name, - kernel_map->dso->sym_priv_size, v); + kernel_map->dso->sym_priv_size); if (sym == NULL) goto out_delete_line; @@ -300,8 +299,8 @@ static int kernel_maps__split_kallsyms(symbol_filter_t filter, int use_modules) if (strcmp(map->dso->name, module)) { map = kernel_maps__find_by_dso_name(module); if (!map) { - fputs("/proc/{kallsyms,modules} " - "inconsistency!\n", stderr); + pr_err("/proc/{kallsyms,modules} " + "inconsistency!\n"); return -1; } } @@ -351,10 +350,9 @@ delete_symbol: } -static int kernel_maps__load_kallsyms(symbol_filter_t filter, - int use_modules, int v) +static int kernel_maps__load_kallsyms(symbol_filter_t filter, int use_modules) { - if (kernel_maps__load_all_kallsyms(v)) + if (kernel_maps__load_all_kallsyms()) return -1; dso__fixup_sym_end(kernel_map->dso); @@ -362,9 +360,9 @@ static int kernel_maps__load_kallsyms(symbol_filter_t filter, return kernel_maps__split_kallsyms(filter, use_modules); } -static size_t kernel_maps__fprintf(FILE *fp, int v) +static size_t kernel_maps__fprintf(FILE *fp) { - size_t printed = fprintf(stderr, "Kernel maps:\n"); + size_t printed = fprintf(fp, "Kernel maps:\n"); struct rb_node *nd; for (nd = rb_first(&kernel_maps); nd; nd = rb_next(nd)) { @@ -372,17 +370,17 @@ static size_t kernel_maps__fprintf(FILE *fp, int v) printed += fprintf(fp, "Map:"); printed += map__fprintf(pos, fp); - if (v > 1) { + if (verbose > 1) { printed += dso__fprintf(pos->dso, fp); printed += fprintf(fp, "--\n"); } } - return printed + fprintf(stderr, "END kernel maps\n"); + return printed + fprintf(fp, "END kernel maps\n"); } static int dso__load_perf_map(struct dso *self, struct map *map, - symbol_filter_t filter, int v) + symbol_filter_t filter) { char *line = NULL; size_t n; @@ -420,7 +418,7 @@ static int dso__load_perf_map(struct dso *self, struct map *map, continue; sym = symbol__new(start, size, line + len, - self->sym_priv_size, v); + self->sym_priv_size); if (sym == NULL) goto out_delete_line; @@ -534,7 +532,7 @@ static Elf_Scn *elf_section_by_name(Elf *elf, GElf_Ehdr *ep, * And always look at the original dso, not at debuginfo packages, that * have the PLT data stripped out (shdr_rel_plt.sh_type == SHT_NOBITS). */ -static int dso__synthesize_plt_symbols(struct dso *self, int v) +static int dso__synthesize_plt_symbols(struct dso *self) { uint32_t nr_rel_entries, idx; GElf_Sym sym; @@ -618,7 +616,7 @@ static int dso__synthesize_plt_symbols(struct dso *self, int v) "%s@plt", elf_sym__name(&sym, symstrs)); f = symbol__new(plt_offset, shdr_plt.sh_entsize, - sympltname, self->sym_priv_size, v); + sympltname, self->sym_priv_size); if (!f) goto out_elf_end; @@ -636,7 +634,7 @@ static int dso__synthesize_plt_symbols(struct dso *self, int v) "%s@plt", elf_sym__name(&sym, symstrs)); f = symbol__new(plt_offset, shdr_plt.sh_entsize, - sympltname, self->sym_priv_size, v); + sympltname, self->sym_priv_size); if (!f) goto out_elf_end; @@ -654,14 +652,14 @@ out_close: if (err == 0) return nr; out: - fprintf(stderr, "%s: problems reading %s PLT info.\n", - __func__, self->long_name); + pr_warning("%s: problems reading %s PLT info.\n", + __func__, self->long_name); return 0; } static int dso__load_sym(struct dso *self, struct map *map, const char *name, int fd, symbol_filter_t filter, int kernel, - int kmodule, int v) + int kmodule) { struct map *curr_map = map; struct dso *curr_dso = self; @@ -680,15 +678,12 @@ static int dso__load_sym(struct dso *self, struct map *map, const char *name, elf = elf_begin(fd, ELF_C_READ_MMAP, NULL); if (elf == NULL) { - if (v) - fprintf(stderr, "%s: cannot read %s ELF file.\n", - __func__, name); + pr_err("%s: cannot read %s ELF file.\n", __func__, name); goto out_close; } if (gelf_getehdr(elf, &ehdr) == NULL) { - if (v) - fprintf(stderr, "%s: cannot get elf header.\n", __func__); + pr_err("%s: cannot get elf header.\n", __func__); goto out_elf_end; } @@ -794,10 +789,9 @@ static int dso__load_sym(struct dso *self, struct map *map, const char *name, } if (curr_dso->adjust_symbols) { - if (v > 2) - printf("adjusting symbol: st_value: %Lx sh_addr: %Lx sh_offset: %Lx\n", - (u64)sym.st_value, (u64)shdr.sh_addr, (u64)shdr.sh_offset); - + pr_debug2("adjusting symbol: st_value: %Lx sh_addr: " + "%Lx sh_offset: %Lx\n", (u64)sym.st_value, + (u64)shdr.sh_addr, (u64)shdr.sh_offset); sym.st_value -= shdr.sh_addr - shdr.sh_offset; } /* @@ -810,7 +804,7 @@ static int dso__load_sym(struct dso *self, struct map *map, const char *name, elf_name = demangled; new_symbol: f = symbol__new(sym.st_value, sym.st_size, elf_name, - curr_dso->sym_priv_size, v); + curr_dso->sym_priv_size); free(demangled); if (!f) goto out_elf_end; @@ -837,7 +831,7 @@ out_close: #define BUILD_ID_SIZE 128 -static char *dso__read_build_id(struct dso *self, int v) +static char *dso__read_build_id(struct dso *self) { int i; GElf_Ehdr ehdr; @@ -854,15 +848,13 @@ static char *dso__read_build_id(struct dso *self, int v) elf = elf_begin(fd, ELF_C_READ_MMAP, NULL); if (elf == NULL) { - if (v) - fprintf(stderr, "%s: cannot read %s ELF file.\n", - __func__, self->long_name); + pr_err("%s: cannot read %s ELF file.\n", __func__, + self->long_name); goto out_close; } if (gelf_getehdr(elf, &ehdr) == NULL) { - if (v) - fprintf(stderr, "%s: cannot get elf header.\n", __func__); + pr_err("%s: cannot get elf header.\n", __func__); goto out_elf_end; } @@ -884,8 +876,7 @@ static char *dso__read_build_id(struct dso *self, int v) ++raw; bid += 2; } - if (v >= 2) - printf("%s(%s): %s\n", __func__, self->long_name, build_id); + pr_debug2("%s(%s): %s\n", __func__, self->long_name, build_id); out_elf_end: elf_end(elf); out_close: @@ -911,8 +902,7 @@ char dso__symtab_origin(const struct dso *self) return origin[self->origin]; } -int dso__load(struct dso *self, struct map *map, - symbol_filter_t filter, int v) +int dso__load(struct dso *self, struct map *map, symbol_filter_t filter) { int size = PATH_MAX; char *name = malloc(size), *build_id = NULL; @@ -925,7 +915,7 @@ int dso__load(struct dso *self, struct map *map, self->adjust_symbols = 0; if (strncmp(self->name, "/tmp/perf-", 10) == 0) { - ret = dso__load_perf_map(self, map, filter, v); + ret = dso__load_perf_map(self, map, filter); self->origin = ret > 0 ? DSO__ORIG_JAVA_JIT : DSO__ORIG_NOT_FOUND; return ret; @@ -946,7 +936,7 @@ more: self->long_name); break; case DSO__ORIG_BUILDID: - build_id = dso__read_build_id(self, v); + build_id = dso__read_build_id(self); if (build_id != NULL) { snprintf(name, size, "/usr/lib/debug/.build-id/%.2s/%s.debug", @@ -967,7 +957,7 @@ more: fd = open(name, O_RDONLY); } while (fd < 0); - ret = dso__load_sym(self, map, name, fd, filter, 0, 0, v); + ret = dso__load_sym(self, map, name, fd, filter, 0, 0); close(fd); /* @@ -977,7 +967,7 @@ more: goto more; if (ret > 0) { - int nr_plt = dso__synthesize_plt_symbols(self, v); + int nr_plt = dso__synthesize_plt_symbols(self); if (nr_plt > 0) ret += nr_plt; } @@ -1025,34 +1015,29 @@ struct map *kernel_maps__find_by_dso_name(const char *name) } static int dso__load_module_sym(struct dso *self, struct map *map, - symbol_filter_t filter, int v) + symbol_filter_t filter) { int err = 0, fd = open(self->long_name, O_RDONLY); if (fd < 0) { - if (v) - fprintf(stderr, "%s: cannot open %s\n", - __func__, self->long_name); + pr_err("%s: cannot open %s\n", __func__, self->long_name); return err; } - err = dso__load_sym(self, map, self->long_name, fd, filter, 0, 1, v); + err = dso__load_sym(self, map, self->long_name, fd, filter, 0, 1); close(fd); return err; } -static int dsos__load_modules_sym_dir(char *dirname, - symbol_filter_t filter, int v) +static int dsos__load_modules_sym_dir(char *dirname, symbol_filter_t filter) { struct dirent *dent; int nr_symbols = 0, err; DIR *dir = opendir(dirname); if (!dir) { - if (v) - fprintf(stderr, "%s: cannot open %s dir\n", __func__, - dirname); + pr_err("%s: cannot open %s dir\n", __func__, dirname); return -1; } @@ -1066,7 +1051,7 @@ static int dsos__load_modules_sym_dir(char *dirname, snprintf(path, sizeof(path), "%s/%s", dirname, dent->d_name); - err = dsos__load_modules_sym_dir(path, filter, v); + err = dsos__load_modules_sym_dir(path, filter); if (err < 0) goto failure; } else { @@ -1092,7 +1077,7 @@ static int dsos__load_modules_sym_dir(char *dirname, if (map->dso->long_name == NULL) goto failure; - err = dso__load_module_sym(map->dso, map, filter, v); + err = dso__load_module_sym(map->dso, map, filter); if (err < 0) goto failure; last = rb_last(&map->dso->syms); @@ -1119,7 +1104,7 @@ failure: return -1; } -static int dsos__load_modules_sym(symbol_filter_t filter, int v) +static int dsos__load_modules_sym(symbol_filter_t filter) { struct utsname uts; char modules_path[PATH_MAX]; @@ -1130,7 +1115,7 @@ static int dsos__load_modules_sym(symbol_filter_t filter, int v) snprintf(modules_path, sizeof(modules_path), "/lib/modules/%s/kernel", uts.release); - return dsos__load_modules_sym_dir(modules_path, filter, v); + return dsos__load_modules_sym_dir(modules_path, filter); } /* @@ -1225,15 +1210,14 @@ out_failure: } static int dso__load_vmlinux(struct dso *self, struct map *map, - const char *vmlinux, - symbol_filter_t filter, int v) + const char *vmlinux, symbol_filter_t filter) { int err, fd = open(vmlinux, O_RDONLY); if (fd < 0) return -1; - err = dso__load_sym(self, map, self->long_name, fd, filter, 1, 0, v); + err = dso__load_sym(self, map, self->long_name, fd, filter, 1, 0); close(fd); @@ -1241,7 +1225,7 @@ static int dso__load_vmlinux(struct dso *self, struct map *map, } int dsos__load_kernel(const char *vmlinux, unsigned int sym_priv_size, - symbol_filter_t filter, int v, int use_modules) + symbol_filter_t filter, int use_modules) { int err = -1; struct dso *dso = dso__new(vmlinux, sym_priv_size); @@ -1257,26 +1241,26 @@ int dsos__load_kernel(const char *vmlinux, unsigned int sym_priv_size, kernel_map->map_ip = kernel_map->unmap_ip = identity__map_ip; if (use_modules && dsos__load_modules(sym_priv_size) < 0) { - fprintf(stderr, "Failed to load list of modules in use! " - "Continuing...\n"); + pr_warning("Failed to load list of modules in use! " + "Continuing...\n"); use_modules = 0; } if (vmlinux) { - err = dso__load_vmlinux(dso, kernel_map, vmlinux, filter, v); + err = dso__load_vmlinux(dso, kernel_map, vmlinux, filter); if (err > 0 && use_modules) { - int syms = dsos__load_modules_sym(filter, v); + int syms = dsos__load_modules_sym(filter); if (syms < 0) - fprintf(stderr, "Failed to read module symbols!" - " Continuing...\n"); + pr_warning("Failed to read module symbols!" + " Continuing...\n"); else err += syms; } } if (err <= 0) - err = kernel_maps__load_kallsyms(filter, use_modules, v); + err = kernel_maps__load_kallsyms(filter, use_modules); if (err > 0) { struct rb_node *node = rb_first(&dso->syms); @@ -1296,8 +1280,8 @@ int dsos__load_kernel(const char *vmlinux, unsigned int sym_priv_size, kernel_maps__fixup_end(); dsos__add(dso); - if (v > 0) - kernel_maps__fprintf(stderr, v); + if (verbose) + kernel_maps__fprintf(stderr); } return err; @@ -1355,8 +1339,8 @@ void dsos__fprintf(FILE *fp) int load_kernel(unsigned int sym_priv_size, symbol_filter_t filter) { - if (dsos__load_kernel(vmlinux_name, sym_priv_size, - filter, verbose, modules) <= 0) + if (dsos__load_kernel(vmlinux_name, sym_priv_size, filter, + modules) <= 0) return -1; vdso = dso__new("[vdso]", 0); diff --git a/tools/perf/util/symbol.h b/tools/perf/util/symbol.h index c2a777de9b7..77b7b3e4241 100644 --- a/tools/perf/util/symbol.h +++ b/tools/perf/util/symbol.h @@ -63,11 +63,10 @@ static inline void *dso__sym_priv(struct dso *self, struct symbol *sym) struct symbol *dso__find_symbol(struct dso *self, u64 ip); int dsos__load_kernel(const char *vmlinux, unsigned int sym_priv_size, - symbol_filter_t filter, int verbose, int modules); + symbol_filter_t filter, int modules); struct dso *dsos__findnew(const char *name, unsigned int sym_priv_size, bool *is_new); -int dso__load(struct dso *self, struct map *map, - symbol_filter_t filter, int v); +int dso__load(struct dso *self, struct map *map, symbol_filter_t filter); void dsos__fprintf(FILE *fp); size_t dso__fprintf(struct dso *self, FILE *fp); diff --git a/tools/perf/util/thread.c b/tools/perf/util/thread.c index 8cb47f1d8a7..0f6d78c9863 100644 --- a/tools/perf/util/thread.c +++ b/tools/perf/util/thread.c @@ -127,9 +127,9 @@ static void thread__remove_overlappings(struct thread *self, struct map *map) continue; if (verbose >= 2) { - printf("overlapping maps:\n"); - map__fprintf(map, stdout); - map__fprintf(pos, stdout); + fputs("overlapping maps:\n", stderr); + map__fprintf(map, stderr); + map__fprintf(pos, stderr); } rb_erase(&pos->rb_node, &self->maps); -- cgit v1.2.3-18-g5258 From b7cb10e790fbd145296e771f789273a875c15719 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Wed, 21 Oct 2009 17:34:06 -0200 Subject: perf probe: Print debug messages using pr_*() Use the new pr_{err,warning,debug,etc} printout methods, just like in the kernel. Signed-off-by: Arnaldo Carvalho de Melo Cc: Masami Hiramatsu Cc: Frederic Weisbecker Cc: Peter Zijlstra Cc: Paul Mackerras Cc: Mike Galbraith LKML-Reference: <1256153646-10097-1-git-send-email-acme@redhat.com> [ Split this patch out, to keep perf/probes separate. ] Signed-off-by: Ingo Molnar --- tools/perf/builtin-probe.c | 20 ++++++++++---------- tools/perf/util/probe-finder.c | 12 ++++++------ 2 files changed, 16 insertions(+), 16 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-probe.c b/tools/perf/builtin-probe.c index b5ad86a265f..dcb406c7f82 100644 --- a/tools/perf/builtin-probe.c +++ b/tools/perf/builtin-probe.c @@ -78,7 +78,7 @@ static int parse_probepoint(const struct option *opt __used, if (!str) /* The end of probe points */ return 0; - eprintf("probe-definition(%d): %s\n", session.nr_probe, str); + pr_debug("probe-definition(%d): %s\n", session.nr_probe, str); if (++session.nr_probe == MAX_PROBES) semantic_error("Too many probes"); @@ -103,7 +103,7 @@ static int parse_probepoint(const struct option *opt __used, die("strndup"); if (++argc == MAX_PROBE_ARGS) semantic_error("Too many arguments"); - eprintf("argv[%d]=%s\n", argc, argv[argc - 1]); + pr_debug("argv[%d]=%s\n", argc, argv[argc - 1]); } } while (*str != '\0'); if (argc < 2) @@ -133,7 +133,7 @@ static int parse_probepoint(const struct option *opt __used, pp->line = atoi(ptr); if (!pp->file || !pp->line) semantic_error("Failed to parse line."); - eprintf("file:%s line:%d\n", pp->file, pp->line); + pr_debug("file:%s line:%d\n", pp->file, pp->line); } else { /* Function name */ ptr = strchr(arg, '+'); @@ -150,8 +150,8 @@ static int parse_probepoint(const struct option *opt __used, pp->file = strdup(ptr); } pp->function = strdup(arg); - eprintf("symbol:%s file:%s offset:%d\n", - pp->function, pp->file, pp->offset); + pr_debug("symbol:%s file:%s offset:%d\n", + pp->function, pp->file, pp->offset); } free(argv[1]); if (pp->file) @@ -175,7 +175,7 @@ static int parse_probepoint(const struct option *opt __used, session.need_dwarf = 1; } - eprintf("%d arguments\n", pp->nr_args); + pr_debug("%d arguments\n", pp->nr_args); return 0; } @@ -188,7 +188,7 @@ static int open_default_vmlinux(void) ret = uname(&uts); if (ret) { - eprintf("uname() failed.\n"); + pr_debug("uname() failed.\n"); return -errno; } session.release = uts.release; @@ -196,12 +196,12 @@ static int open_default_vmlinux(void) ret = snprintf(fname, MAX_PATH_LEN, default_search_path[i], session.release); if (ret >= MAX_PATH_LEN || ret < 0) { - eprintf("Filename(%d,%s) is too long.\n", i, + pr_debug("Filename(%d,%s) is too long.\n", i, uts.release); errno = E2BIG; return -E2BIG; } - eprintf("try to open %s\n", fname); + pr_debug("try to open %s\n", fname); fd = open(fname, O_RDONLY); if (fd >= 0) break; @@ -341,7 +341,7 @@ int cmd_probe(int argc, const char **argv, const char *prefix __used) ret = find_probepoint(fd, pp); if (ret <= 0) die("No probe point found.\n"); - eprintf("probe event %s found\n", session.events[j]); + pr_debug("probe event %s found\n", session.events[j]); } close(fd); diff --git a/tools/perf/util/probe-finder.c b/tools/perf/util/probe-finder.c index be997abdf5b..54e70718530 100644 --- a/tools/perf/util/probe-finder.c +++ b/tools/perf/util/probe-finder.c @@ -136,7 +136,7 @@ static Dwarf_Unsigned die_get_fileno(Dwarf_Die cu_die, const char *fname) dwarf_dealloc(__dw_debug, srcs, DW_DLA_LIST); } if (found) - eprintf("found fno: %d\n", (int)found); + pr_debug("found fno: %d\n", (int)found); return found; } @@ -442,7 +442,7 @@ static void find_variable(Dwarf_Die sp_die, struct probe_finder *pf) return ; } - eprintf("Searching '%s' variable in context.\n", pf->var); + pr_debug("Searching '%s' variable in context.\n", pf->var); /* Search child die for local variables and parameters. */ ret = search_die_from_children(sp_die, variable_callback, pf); if (!ret) @@ -552,7 +552,7 @@ static void find_by_line(Dwarf_Die cu_die, struct probe_finder *pf) ret = dwarf_lineaddr(lines[i], &addr, &__dw_error); DIE_IF(ret != DW_DLV_OK); - eprintf("Probe point found: 0x%llx\n", addr); + pr_debug("Probe point found: 0x%llx\n", addr); pf->addr = addr; /* Search a real subprogram including this line, */ ret = search_die_from_children(cu_die, probeaddr_callback, pf); @@ -583,8 +583,8 @@ static int probefunc_callback(struct die_link *dlink, void *data) &pf->inl_offs, &__dw_error); DIE_IF(ret != DW_DLV_OK); - eprintf("inline definition offset %lld\n", - pf->inl_offs); + pr_debug("inline definition offset %lld\n", + pf->inl_offs); return 0; } /* Get probe address */ @@ -599,7 +599,7 @@ static int probefunc_callback(struct die_link *dlink, void *data) /* Get probe address */ pf->addr = die_get_entrypc(dlink->die); pf->addr += pp->offset; - eprintf("found inline addr: 0x%llx\n", pf->addr); + pr_debug("found inline addr: 0x%llx\n", pf->addr); /* Inlined function. Get a real subprogram */ for (lk = dlink->parent; lk != NULL; lk = lk->parent) { tag = 0; -- cgit v1.2.3-18-g5258 From fcd14b3203b538dca04a2b065c774c0b57863eec Mon Sep 17 00:00:00 2001 From: Michael Cree Date: Mon, 26 Oct 2009 21:32:06 +1300 Subject: perf tools, Alpha: Add Alpha support to perf.h For the perf tool the patch implements an Alpha specific section in the perf.h header file. Signed-off-by: Michael Cree Cc: Richard Henderson Cc: Ivan Kokshaysky Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1256545926-6972-1-git-send-email-mcree@orcon.net.nz> Signed-off-by: Ingo Molnar --- tools/perf/perf.h | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'tools/perf') diff --git a/tools/perf/perf.h b/tools/perf/perf.h index 8cc4623afd6..216bdb223f6 100644 --- a/tools/perf/perf.h +++ b/tools/perf/perf.h @@ -47,6 +47,12 @@ #define cpu_relax() asm volatile("":::"memory") #endif +#ifdef __alpha__ +#include "../../arch/alpha/include/asm/unistd.h" +#define rmb() asm volatile("mb" ::: "memory") +#define cpu_relax() asm volatile("" ::: "memory") +#endif + #include #include #include -- cgit v1.2.3-18-g5258 From 7f3bedcc93f935631d2363f23de1cc80f04fdf3e Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 26 Oct 2009 19:23:17 -0200 Subject: perf record: Fix race where process can disappear while reading its /proc/pid/tasks Signed-off-by: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: Peter Zijlstra Cc: Paul Mackerras Cc: Mike Galbraith LKML-Reference: <1256592199-9608-1-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar --- tools/perf/builtin-record.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'tools/perf') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index ac5ddfff445..9e1638cc19c 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -206,6 +206,7 @@ static pid_t pid_synthesize_comm_event(pid_t pid, int full) fp = fopen(filename, "r"); if (fp == NULL) { +out_race: /* * We raced with a task exiting - just return: */ @@ -247,6 +248,9 @@ static pid_t pid_synthesize_comm_event(pid_t pid, int full) snprintf(filename, sizeof(filename), "/proc/%d/task", pid); tasks = opendir(filename); + if (tasks == NULL) + goto out_race; + while (!readdir_r(tasks, &dirent, &next) && next) { char *end; pid = strtol(dirent.d_name, &end, 10); -- cgit v1.2.3-18-g5258 From 234fbbf508c58c5084292b11b242377553897459 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 26 Oct 2009 19:23:18 -0200 Subject: perf tools: Generalize event synthesizing routines Because we will need it in 'perf top' to support userspace symbols for existing threads. Now we pass a callback that will receive the synthesized event and then write it to the output file in 'perf record' and in the upcoming patch for 'perf top' we will just immediatelly create the in memory representation of threads and maps. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: Peter Zijlstra Cc: Paul Mackerras Cc: Mike Galbraith LKML-Reference: <1256592199-9608-2-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar --- tools/perf/Makefile | 1 + tools/perf/builtin-record.c | 181 +++----------------------------------------- tools/perf/util/event.c | 177 +++++++++++++++++++++++++++++++++++++++++++ tools/perf/util/event.h | 3 + 4 files changed, 191 insertions(+), 171 deletions(-) create mode 100644 tools/perf/util/event.c (limited to 'tools/perf') diff --git a/tools/perf/Makefile b/tools/perf/Makefile index 0a40c29b238..9f4488d6f8e 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -380,6 +380,7 @@ LIB_OBJS += util/alias.o LIB_OBJS += util/config.o LIB_OBJS += util/ctype.o LIB_OBJS += util/environment.o +LIB_OBJS += util/event.o LIB_OBJS += util/exec_cmd.o LIB_OBJS += util/help.o LIB_OBJS += util/levenshtein.o diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 9e1638cc19c..4a73d89ce5d 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -109,6 +109,12 @@ static void write_output(void *buf, size_t size) } } +static int process_synthesized_event(event_t *event) +{ + write_output(event, event->header.size); + return 0; +} + static void mmap_read(struct mmap_data *md) { unsigned int head = mmap_read_head(md); @@ -191,172 +197,6 @@ static void sig_atexit(void) kill(getpid(), signr); } -static pid_t pid_synthesize_comm_event(pid_t pid, int full) -{ - struct comm_event comm_ev; - char filename[PATH_MAX]; - char bf[BUFSIZ]; - FILE *fp; - size_t size = 0; - DIR *tasks; - struct dirent dirent, *next; - pid_t tgid = 0; - - snprintf(filename, sizeof(filename), "/proc/%d/status", pid); - - fp = fopen(filename, "r"); - if (fp == NULL) { -out_race: - /* - * We raced with a task exiting - just return: - */ - if (verbose) - fprintf(stderr, "couldn't open %s\n", filename); - return 0; - } - - memset(&comm_ev, 0, sizeof(comm_ev)); - while (!comm_ev.comm[0] || !comm_ev.pid) { - if (fgets(bf, sizeof(bf), fp) == NULL) - goto out_failure; - - if (memcmp(bf, "Name:", 5) == 0) { - char *name = bf + 5; - while (*name && isspace(*name)) - ++name; - size = strlen(name) - 1; - memcpy(comm_ev.comm, name, size++); - } else if (memcmp(bf, "Tgid:", 5) == 0) { - char *tgids = bf + 5; - while (*tgids && isspace(*tgids)) - ++tgids; - tgid = comm_ev.pid = atoi(tgids); - } - } - - comm_ev.header.type = PERF_RECORD_COMM; - size = ALIGN(size, sizeof(u64)); - comm_ev.header.size = sizeof(comm_ev) - (sizeof(comm_ev.comm) - size); - - if (!full) { - comm_ev.tid = pid; - - write_output(&comm_ev, comm_ev.header.size); - goto out_fclose; - } - - snprintf(filename, sizeof(filename), "/proc/%d/task", pid); - - tasks = opendir(filename); - if (tasks == NULL) - goto out_race; - - while (!readdir_r(tasks, &dirent, &next) && next) { - char *end; - pid = strtol(dirent.d_name, &end, 10); - if (*end) - continue; - - comm_ev.tid = pid; - - write_output(&comm_ev, comm_ev.header.size); - } - closedir(tasks); - -out_fclose: - fclose(fp); - return tgid; - -out_failure: - fprintf(stderr, "couldn't get COMM and pgid, malformed %s\n", - filename); - exit(EXIT_FAILURE); -} - -static void pid_synthesize_mmap_samples(pid_t pid, pid_t tgid) -{ - char filename[PATH_MAX]; - FILE *fp; - - snprintf(filename, sizeof(filename), "/proc/%d/maps", pid); - - fp = fopen(filename, "r"); - if (fp == NULL) { - /* - * We raced with a task exiting - just return: - */ - if (verbose) - fprintf(stderr, "couldn't open %s\n", filename); - return; - } - while (1) { - char bf[BUFSIZ], *pbf = bf; - struct mmap_event mmap_ev = { - .header = { .type = PERF_RECORD_MMAP }, - }; - int n; - size_t size; - if (fgets(bf, sizeof(bf), fp) == NULL) - break; - - /* 00400000-0040c000 r-xp 00000000 fd:01 41038 /bin/cat */ - n = hex2u64(pbf, &mmap_ev.start); - if (n < 0) - continue; - pbf += n + 1; - n = hex2u64(pbf, &mmap_ev.len); - if (n < 0) - continue; - pbf += n + 3; - if (*pbf == 'x') { /* vm_exec */ - char *execname = strchr(bf, '/'); - - /* Catch VDSO */ - if (execname == NULL) - execname = strstr(bf, "[vdso]"); - - if (execname == NULL) - continue; - - size = strlen(execname); - execname[size - 1] = '\0'; /* Remove \n */ - memcpy(mmap_ev.filename, execname, size); - size = ALIGN(size, sizeof(u64)); - mmap_ev.len -= mmap_ev.start; - mmap_ev.header.size = (sizeof(mmap_ev) - - (sizeof(mmap_ev.filename) - size)); - mmap_ev.pid = tgid; - mmap_ev.tid = pid; - - write_output(&mmap_ev, mmap_ev.header.size); - } - } - - fclose(fp); -} - -static void synthesize_all(void) -{ - DIR *proc; - struct dirent dirent, *next; - - proc = opendir("/proc"); - - while (!readdir_r(proc, &dirent, &next) && next) { - char *end; - pid_t pid, tgid; - - pid = strtol(dirent.d_name, &end, 10); - if (*end) /* only interested in proper numerical dirents */ - continue; - - tgid = pid_synthesize_comm_event(pid, 1); - pid_synthesize_mmap_samples(pid, tgid); - } - - closedir(proc); -} - static int group_fd; static struct perf_header_attr *get_header_attr(struct perf_event_attr *a, int nr) @@ -608,11 +448,10 @@ static int __cmd_record(int argc, const char **argv) if (file_new) perf_header__write(header, output); - if (!system_wide) { - pid_t tgid = pid_synthesize_comm_event(pid, 0); - pid_synthesize_mmap_samples(pid, tgid); - } else - synthesize_all(); + if (!system_wide) + event__synthesize_thread(pid, process_synthesized_event); + else + event__synthesize_threads(process_synthesized_event); if (target_pid == -1 && argc) { pid = fork(); diff --git a/tools/perf/util/event.c b/tools/perf/util/event.c new file mode 100644 index 00000000000..1dae7e3b400 --- /dev/null +++ b/tools/perf/util/event.c @@ -0,0 +1,177 @@ +#include +#include "event.h" +#include "debug.h" +#include "string.h" + +static pid_t event__synthesize_comm(pid_t pid, int full, + int (*process)(event_t *event)) +{ + event_t ev; + char filename[PATH_MAX]; + char bf[BUFSIZ]; + FILE *fp; + size_t size = 0; + DIR *tasks; + struct dirent dirent, *next; + pid_t tgid = 0; + + snprintf(filename, sizeof(filename), "/proc/%d/status", pid); + + fp = fopen(filename, "r"); + if (fp == NULL) { +out_race: + /* + * We raced with a task exiting - just return: + */ + pr_debug("couldn't open %s\n", filename); + return 0; + } + + memset(&ev.comm, 0, sizeof(ev.comm)); + while (!ev.comm.comm[0] || !ev.comm.pid) { + if (fgets(bf, sizeof(bf), fp) == NULL) + goto out_failure; + + if (memcmp(bf, "Name:", 5) == 0) { + char *name = bf + 5; + while (*name && isspace(*name)) + ++name; + size = strlen(name) - 1; + memcpy(ev.comm.comm, name, size++); + } else if (memcmp(bf, "Tgid:", 5) == 0) { + char *tgids = bf + 5; + while (*tgids && isspace(*tgids)) + ++tgids; + tgid = ev.comm.pid = atoi(tgids); + } + } + + ev.comm.header.type = PERF_RECORD_COMM; + size = ALIGN(size, sizeof(u64)); + ev.comm.header.size = sizeof(ev.comm) - (sizeof(ev.comm.comm) - size); + + if (!full) { + ev.comm.tid = pid; + + process(&ev); + goto out_fclose; + } + + snprintf(filename, sizeof(filename), "/proc/%d/task", pid); + + tasks = opendir(filename); + if (tasks == NULL) + goto out_race; + + while (!readdir_r(tasks, &dirent, &next) && next) { + char *end; + pid = strtol(dirent.d_name, &end, 10); + if (*end) + continue; + + ev.comm.tid = pid; + + process(&ev); + } + closedir(tasks); + +out_fclose: + fclose(fp); + return tgid; + +out_failure: + pr_warning("couldn't get COMM and pgid, malformed %s\n", filename); + return -1; +} + +static int event__synthesize_mmap_events(pid_t pid, pid_t tgid, + int (*process)(event_t *event)) +{ + char filename[PATH_MAX]; + FILE *fp; + + snprintf(filename, sizeof(filename), "/proc/%d/maps", pid); + + fp = fopen(filename, "r"); + if (fp == NULL) { + /* + * We raced with a task exiting - just return: + */ + pr_debug("couldn't open %s\n", filename); + return -1; + } + + while (1) { + char bf[BUFSIZ], *pbf = bf; + event_t ev = { + .header = { .type = PERF_RECORD_MMAP }, + }; + int n; + size_t size; + if (fgets(bf, sizeof(bf), fp) == NULL) + break; + + /* 00400000-0040c000 r-xp 00000000 fd:01 41038 /bin/cat */ + n = hex2u64(pbf, &ev.mmap.start); + if (n < 0) + continue; + pbf += n + 1; + n = hex2u64(pbf, &ev.mmap.len); + if (n < 0) + continue; + pbf += n + 3; + if (*pbf == 'x') { /* vm_exec */ + char *execname = strchr(bf, '/'); + + /* Catch VDSO */ + if (execname == NULL) + execname = strstr(bf, "[vdso]"); + + if (execname == NULL) + continue; + + size = strlen(execname); + execname[size - 1] = '\0'; /* Remove \n */ + memcpy(ev.mmap.filename, execname, size); + size = ALIGN(size, sizeof(u64)); + ev.mmap.len -= ev.mmap.start; + ev.mmap.header.size = (sizeof(ev.mmap) - + (sizeof(ev.mmap.filename) - size)); + ev.mmap.pid = tgid; + ev.mmap.tid = pid; + + process(&ev); + } + } + + fclose(fp); + return 0; +} + +int event__synthesize_thread(pid_t pid, int (*process)(event_t *event)) +{ + pid_t tgid = event__synthesize_comm(pid, 1, process); + if (tgid == -1) + return -1; + return event__synthesize_mmap_events(pid, tgid, process); +} + +void event__synthesize_threads(int (*process)(event_t *event)) +{ + DIR *proc; + struct dirent dirent, *next; + + proc = opendir("/proc"); + + while (!readdir_r(proc, &dirent, &next) && next) { + char *end; + pid_t pid = strtol(dirent.d_name, &end, 10); + + if (*end) /* only interested in proper numerical dirents */ + continue; + + event__synthesize_thread(pid, process); + } + + closedir(proc); +} diff --git a/tools/perf/util/event.h b/tools/perf/util/event.h index d972b4b0d38..2ae1177be40 100644 --- a/tools/perf/util/event.h +++ b/tools/perf/util/event.h @@ -111,4 +111,7 @@ struct map *map__clone(struct map *self); int map__overlap(struct map *l, struct map *r); size_t map__fprintf(struct map *self, FILE *fp); +int event__synthesize_thread(pid_t pid, int (*process)(event_t *event)); +void event__synthesize_threads(int (*process)(event_t *event)); + #endif /* __PERF_RECORD_H */ -- cgit v1.2.3-18-g5258 From 5b2bb75a0d4b08cd16bc35ecd674f957fc3b0eb7 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 26 Oct 2009 19:23:19 -0200 Subject: perf top: Support userspace symbols too Example: Compiling the kernel with 'make -k 22 allyesconfig' [root@emilia linux-2.6-tip]# perf top -r 90 ------------------------------------------------------------------------------ PerfTop: 3669 irqs/sec kernel:59.9% [1000Hz cycles], (all, 8 CPUs) ------------------------------------------------------------------------------ samples pcnt function DSO _______ _____ ________________________________ ________________ 3062.00 6.5% clear_page_c [kernel] 2233.00 4.8% _int_malloc /lib64/libc-2.5.so 2100.00 4.5% yylex /home/acme/git/build/allyesconfig/scripts/genksyms/genksyms 2029.00 4.3% memset /lib64/libc-2.5.so 1224.00 2.6% page_fault [kernel] 1075.00 2.3% __GI_strlen /lib64/libc-2.5.so 863.00 1.8% sub_preempt_count [kernel] 822.00 1.8% __GI_memcpy /lib64/libc-2.5.so 810.00 1.7% __GI_vfprintf /lib64/libc-2.5.so 786.00 1.7% _int_free /lib64/libc-2.5.so 775.00 1.7% __GI_strcmp /lib64/libc-2.5.so 748.00 1.6% _spin_lock [kernel] 699.00 1.5% main /home/acme/git/build/allyesconfig/scripts/basic/fixdep 659.00 1.4% add_preempt_count [kernel] 649.00 1.4% yyparse /home/acme/git/build/allyesconfig/scripts/genksyms/genksyms 645.00 1.4% preempt_trace [kernel] 635.00 1.4% __GI___libc_free /lib64/libc-2.5.so 597.00 1.3% trace_preempt_on [kernel] 551.00 1.2% __GI___libc_malloc /lib64/libc-2.5.so 516.00 1.1% _spin_lock_irqsave [kernel] 481.00 1.0% copy_user_generic_string [kernel] 479.00 1.0% unmap_vmas [kernel] 429.00 0.9% _IO_file_xsputn_internal /lib64/libc-2.5.so 425.00 0.9% __GI_strncpy /lib64/libc-2.5.so 416.00 0.9% get_page_from_freelist [kernel] 414.00 0.9% malloc_consolidate /lib64/libc-2.5.so 406.00 0.9% get_parent_ip [kernel] 362.00 0.8% __rmqueue [kernel] 347.00 0.7% in_lock_functions [kernel] 316.00 0.7% __d_lookup [kernel] [root@emilia linux-2.6-tip]# More polishing is needed to print just DSO basename when not --verbose, etc. Supporting a 'comm' column requires some more reworking of 'perf top' internals as we will need to use something like the hist entries 'perf report' uses and will be done in another patch. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: Peter Zijlstra Cc: Paul Mackerras Cc: Mike Galbraith LKML-Reference: <1256592199-9608-3-git-send-email-acme@redhat.com> Signed-off-by: Ingo Molnar --- tools/perf/builtin-top.c | 143 +++++++++++++++++++++++++++++++++-------------- 1 file changed, 101 insertions(+), 42 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index 4a9fe228be2..a02fc414601 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -318,7 +318,7 @@ static void show_details(struct sym_entry *syme) } /* - * Symbols will be added here in record_ip and will get out + * Symbols will be added here in event__process_sample and will get out * after decayed. */ static LIST_HEAD(active_symbols); @@ -459,18 +459,18 @@ static void print_sym_table(void) } if (nr_counters == 1) - printf(" samples pcnt"); + printf(" samples pcnt"); else - printf(" weight samples pcnt"); + printf(" weight samples pcnt"); if (verbose) printf(" RIP "); - printf(" kernel function\n"); - printf(" %s _______ _____", + printf(" function DSO\n"); + printf(" %s _______ _____", nr_counters == 1 ? " " : "______"); if (verbose) - printf(" ________________"); - printf(" _______________\n\n"); + printf(" ________________"); + printf(" ________________________________ ________________\n\n"); for (nd = rb_first(&tmp); nd; nd = rb_next(nd)) { struct symbol *sym; @@ -486,16 +486,15 @@ static void print_sym_table(void) sum_ksamples)); if (nr_counters == 1 || !display_weighted) - printf("%20.2f - ", syme->weight); + printf("%20.2f ", syme->weight); else - printf("%9.1f %10ld - ", syme->weight, syme->snap_count); + printf("%9.1f %10ld ", syme->weight, syme->snap_count); percent_color_fprintf(stdout, "%4.1f%%", pcnt); if (verbose) - printf(" - %016llx", sym->start); - printf(" : %s", sym->name); - if (syme->map->dso->name[0] == '[') - printf(" \t%s", syme->map->dso->name); + printf(" %016llx", sym->start); + printf(" %-32s", sym->name); + printf(" %s", syme->map->dso->short_name); printf("\n"); } } @@ -818,41 +817,97 @@ static int parse_symbols(void) return 0; } -/* - * Binary search in the histogram table and record the hit: - */ -static void record_ip(u64 ip, int counter) +static void event__process_sample(const event_t *self, int counter) { + u64 ip = self->ip.ip; struct map *map; - struct symbol *sym = kernel_maps__find_symbol(ip, &map); - - if (sym != NULL) { - struct sym_entry *syme = dso__sym_priv(map->dso, sym); - - if (!syme->skip) { - syme->count[counter]++; - record_precise_ip(syme, counter, ip); - pthread_mutex_lock(&active_symbols_lock); - if (list_empty(&syme->node) || !syme->node.next) - __list_insert_active_sym(syme); - pthread_mutex_unlock(&active_symbols_lock); + struct sym_entry *syme; + struct symbol *sym; + + switch (self->header.misc & PERF_RECORD_MISC_CPUMODE_MASK) { + case PERF_RECORD_MISC_USER: { + struct thread *thread = threads__findnew(self->ip.pid); + + if (thread == NULL) return; + + map = thread__find_map(thread, ip); + if (map != NULL) { + ip = map->map_ip(map, ip); + sym = map->dso->find_symbol(map->dso, ip); + if (sym == NULL) + return; + userspace_samples++; + break; } } + /* + * If this is outside of all known maps, + * and is a negative address, try to look it + * up in the kernel dso, as it might be a + * vsyscall or vdso (which executes in user-mode). + */ + if ((long long)ip >= 0) + return; + /* Fall thru */ + case PERF_RECORD_MISC_KERNEL: + sym = kernel_maps__find_symbol(ip, &map); + if (sym == NULL) + return; + break; + default: + return; + } + + syme = dso__sym_priv(map->dso, sym); - samples--; + if (!syme->skip) { + syme->count[counter]++; + record_precise_ip(syme, counter, ip); + pthread_mutex_lock(&active_symbols_lock); + if (list_empty(&syme->node) || !syme->node.next) + __list_insert_active_sym(syme); + pthread_mutex_unlock(&active_symbols_lock); + ++samples; + return; + } } -static void process_event(u64 ip, int counter, int user) +static void event__process_mmap(event_t *self) { - samples++; + struct thread *thread = threads__findnew(self->mmap.pid); + + if (thread != NULL) { + struct map *map = map__new(&self->mmap, NULL, 0, + sizeof(struct sym_entry), + symbol_filter); + if (map != NULL) + thread__insert_map(thread, map); + } +} - if (user) { - userspace_samples++; - return; +static void event__process_comm(event_t *self) +{ + struct thread *thread = threads__findnew(self->comm.pid); + + if (thread != NULL) + thread__set_comm(thread, self->comm.comm); +} + +static int event__process(event_t *event) +{ + switch (event->header.type) { + case PERF_RECORD_COMM: + event__process_comm(event); + break; + case PERF_RECORD_MMAP: + event__process_mmap(event); + break; + default: + break; } - record_ip(ip, counter); + return 0; } struct mmap_data { @@ -925,13 +980,11 @@ static void mmap_read_counter(struct mmap_data *md) event = &event_copy; } + if (event->header.type == PERF_RECORD_SAMPLE) + event__process_sample(event, md->counter); + else + event__process(event); old += size; - - if (event->header.type == PERF_RECORD_SAMPLE) { - int user = - (event->header.misc & PERF_RECORD_MISC_CPUMODE_MASK) == PERF_RECORD_MISC_USER; - process_event(event->ip.ip, md->counter, user); - } } md->prev = old; @@ -973,6 +1026,7 @@ static void start_counter(int i, int counter) } attr->inherit = (cpu < 0) && inherit; + attr->mmap = 1; try_again: fd[i][counter] = sys_perf_event_open(attr, target_pid, cpu, group_fd, 0); @@ -1031,6 +1085,11 @@ static int __cmd_top(void) int i, counter; int ret; + if (target_pid != -1) + event__synthesize_thread(target_pid, event__process); + else + event__synthesize_threads(event__process); + for (i = 0; i < nr_cpus; i++) { group_fd = -1; for (counter = 0; counter < nr_counters; counter++) -- cgit v1.2.3-18-g5258 From 85df6f683efa457440eb922272fd5a71aa022ad4 Mon Sep 17 00:00:00 2001 From: Marti Raudsepp Date: Tue, 27 Oct 2009 00:33:04 +0000 Subject: perf tools: Notify user when unrecognized event is specified Previously no indication was given about what went wrong. Signed-off-by: Marti Raudsepp Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <03ec9ee96f17cef05424.1256603584@localhost> Signed-off-by: Ingo Molnar --- tools/perf/util/parse-events.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'tools/perf') diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c index b097570e962..e9e6d5c0ae4 100644 --- a/tools/perf/util/parse-events.c +++ b/tools/perf/util/parse-events.c @@ -678,6 +678,8 @@ parse_event_symbols(const char **str, struct perf_event_attr *attr) if (ret != EVT_FAILED) goto modifier; + fprintf(stderr, "invalid or unsupported event: '%s'\n", *str); + fprintf(stderr, "Run 'perf list' for a list of valid events\n"); return EVT_FAILED; modifier: -- cgit v1.2.3-18-g5258 From 689d30187828afe1faedf050b2f7593515b90c76 Mon Sep 17 00:00:00 2001 From: Marti Raudsepp Date: Tue, 27 Oct 2009 00:33:05 +0000 Subject: perf tools: Output 'perf list' to stdout not stderr Writing to stdout is probably the expected behavior because the user explicitly asked for a list. Signed-off-by: Marti Raudsepp Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <4ebb59420ef057972167.1256603585@localhost> Signed-off-by: Ingo Molnar --- tools/perf/util/parse-events.c | 20 ++++++++++---------- 1 file changed, 10 insertions(+), 10 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c index e9e6d5c0ae4..31baa5a6036 100644 --- a/tools/perf/util/parse-events.c +++ b/tools/perf/util/parse-events.c @@ -806,7 +806,7 @@ static void print_tracepoint_events(void) for_each_event(sys_dirent, evt_dir, evt_dirent, evt_next) { snprintf(evt_path, MAXPATHLEN, "%s:%s", sys_dirent.d_name, evt_dirent.d_name); - fprintf(stderr, " %-42s [%s]\n", evt_path, + printf(" %-42s [%s]\n", evt_path, event_type_descriptors[PERF_TYPE_TRACEPOINT+1]); } closedir(evt_dir); @@ -823,8 +823,8 @@ void print_events(void) unsigned int i, type, op, prev_type = -1; char name[40]; - fprintf(stderr, "\n"); - fprintf(stderr, "List of pre-defined events (to be used in -e):\n"); + printf("\n"); + printf("List of pre-defined events (to be used in -e):\n"); for (i = 0; i < ARRAY_SIZE(event_symbols); i++, syms++) { type = syms->type + 1; @@ -832,19 +832,19 @@ void print_events(void) type = 0; if (type != prev_type) - fprintf(stderr, "\n"); + printf("\n"); if (strlen(syms->alias)) sprintf(name, "%s OR %s", syms->symbol, syms->alias); else strcpy(name, syms->symbol); - fprintf(stderr, " %-42s [%s]\n", name, + printf(" %-42s [%s]\n", name, event_type_descriptors[type]); prev_type = type; } - fprintf(stderr, "\n"); + printf("\n"); for (type = 0; type < PERF_COUNT_HW_CACHE_MAX; type++) { for (op = 0; op < PERF_COUNT_HW_CACHE_OP_MAX; op++) { /* skip invalid cache type */ @@ -852,17 +852,17 @@ void print_events(void) continue; for (i = 0; i < PERF_COUNT_HW_CACHE_RESULT_MAX; i++) { - fprintf(stderr, " %-42s [%s]\n", + printf(" %-42s [%s]\n", event_cache_name(type, op, i), event_type_descriptors[4]); } } } - fprintf(stderr, "\n"); - fprintf(stderr, " %-42s [raw hardware event descriptor]\n", + printf("\n"); + printf(" %-42s [raw hardware event descriptor]\n", "rNNN"); - fprintf(stderr, "\n"); + printf("\n"); print_tracepoint_events(); -- cgit v1.2.3-18-g5258 From f7d7986060b2890fc26db6ab5203efbd33aa2497 Mon Sep 17 00:00:00 2001 From: Anton Blanchard Date: Sun, 18 Oct 2009 01:09:29 +0000 Subject: perf_event: Add alignment-faults and emulation-faults software events Add two more software events that are common to many cpus. Alignment faults: When a load or store is not aligned properly. Emulation faults: When an instruction is emulated in software. Both cause a very significant slowdown (100x or worse), so identifying and fixing them is very important. Signed-off-by: Anton Blanchard Signed-off-by: Paul Mackerras --- tools/perf/design.txt | 2 ++ tools/perf/util/parse-events.c | 4 ++++ 2 files changed, 6 insertions(+) (limited to 'tools/perf') diff --git a/tools/perf/design.txt b/tools/perf/design.txt index fdd42a824c9..f000c30877a 100644 --- a/tools/perf/design.txt +++ b/tools/perf/design.txt @@ -137,6 +137,8 @@ enum sw_event_ids { PERF_COUNT_SW_CPU_MIGRATIONS = 4, PERF_COUNT_SW_PAGE_FAULTS_MIN = 5, PERF_COUNT_SW_PAGE_FAULTS_MAJ = 6, + PERF_COUNT_SW_ALIGNMENT_FAULTS = 7, + PERF_COUNT_SW_EMULATION_FAULTS = 8, }; Counters of the type PERF_TYPE_TRACEPOINT are available when the ftrace event diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c index 8cfb48cbbea..34bd8442393 100644 --- a/tools/perf/util/parse-events.c +++ b/tools/perf/util/parse-events.c @@ -46,6 +46,8 @@ static struct event_symbol event_symbols[] = { { CSW(PAGE_FAULTS_MAJ), "major-faults", "" }, { CSW(CONTEXT_SWITCHES), "context-switches", "cs" }, { CSW(CPU_MIGRATIONS), "cpu-migrations", "migrations" }, + { CSW(ALIGNMENT_FAULTS), "alignment-faults", "" }, + { CSW(EMULATION_FAULTS), "emulation-faults", "" }, }; #define __PERF_EVENT_FIELD(config, name) \ @@ -74,6 +76,8 @@ static const char *sw_event_names[] = { "CPU-migrations", "minor-faults", "major-faults", + "alignment-faults", + "emulation-faults", }; #define MAX_ALIASES 8 -- cgit v1.2.3-18-g5258 From 66bd8424cc05e800db384053bf7ab967e4658468 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Wed, 28 Oct 2009 21:51:21 -0200 Subject: perf tools: Delay loading symtabs till we hit a map with it So that we can have a quicker start on perf top and even speedups in the other tools, as we can have maps with no hits, so no need to load its symtabs. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: Peter Zijlstra Cc: Paul Mackerras Cc: Mike Galbraith LKML-Reference: <1256773881-4191-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/builtin-annotate.c | 4 ++-- tools/perf/builtin-report.c | 4 ++-- tools/perf/builtin-top.c | 5 ++--- tools/perf/util/event.h | 3 ++- tools/perf/util/map.c | 38 ++++++++++++++++++++++---------------- tools/perf/util/symbol.c | 16 +++++++++------- tools/perf/util/symbol.h | 4 ++-- 7 files changed, 41 insertions(+), 33 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c index 6d63c2eea2c..8688bfee42a 100644 --- a/tools/perf/builtin-annotate.c +++ b/tools/perf/builtin-annotate.c @@ -165,7 +165,7 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) if (map != NULL) { got_map: ip = map->map_ip(map, ip); - sym = map->dso->find_symbol(map->dso, ip); + sym = map__find_symbol(map, ip, symbol_filter); } else { /* * If this is outside of all known maps, @@ -203,7 +203,7 @@ static int process_mmap_event(event_t *event, unsigned long offset, unsigned long head) { struct map *map = map__new(&event->mmap, NULL, 0, - sizeof(struct sym_priv), symbol_filter); + sizeof(struct sym_priv)); struct thread *thread = threads__findnew(event->mmap.pid); dump_printf("%p [%p]: PERF_RECORD_MMAP %d: [%p(%p) @ %p]: %s\n", diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index b3d814b5455..f1bcd35bd22 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -455,7 +455,7 @@ got_map: dump_printf(" ...... map: %Lx -> %Lx\n", *ipp, ip); *ipp = ip; - return map ? map->dso->find_symbol(map->dso, ip) : NULL; + return map ? map__find_symbol(map, ip, NULL) : NULL; } static int call__match(struct symbol *sym) @@ -751,7 +751,7 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) static int process_mmap_event(event_t *event, unsigned long offset, unsigned long head) { - struct map *map = map__new(&event->mmap, cwd, cwdlen, 0, NULL); + struct map *map = map__new(&event->mmap, cwd, cwdlen, 0); struct thread *thread = threads__findnew(event->mmap.pid); dump_printf("%p [%p]: PERF_RECORD_MMAP %d/%d: [%p(%p) @ %p]: %s\n", diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index a02fc414601..ee87640b335 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -834,7 +834,7 @@ static void event__process_sample(const event_t *self, int counter) map = thread__find_map(thread, ip); if (map != NULL) { ip = map->map_ip(map, ip); - sym = map->dso->find_symbol(map->dso, ip); + sym = map__find_symbol(map, ip, symbol_filter); if (sym == NULL) return; userspace_samples++; @@ -879,8 +879,7 @@ static void event__process_mmap(event_t *self) if (thread != NULL) { struct map *map = map__new(&self->mmap, NULL, 0, - sizeof(struct sym_entry), - symbol_filter); + sizeof(struct sym_entry)); if (map != NULL) thread__insert_map(thread, map); } diff --git a/tools/perf/util/event.h b/tools/perf/util/event.h index 2ae1177be40..3064a05f0f5 100644 --- a/tools/perf/util/event.h +++ b/tools/perf/util/event.h @@ -106,10 +106,11 @@ struct symbol; typedef int (*symbol_filter_t)(struct map *map, struct symbol *sym); struct map *map__new(struct mmap_event *event, char *cwd, int cwdlen, - unsigned int sym_priv_size, symbol_filter_t filter); + unsigned int sym_priv_size); struct map *map__clone(struct map *self); int map__overlap(struct map *l, struct map *r); size_t map__fprintf(struct map *self, FILE *fp); +struct symbol *map__find_symbol(struct map *self, u64 ip, symbol_filter_t filter); int event__synthesize_thread(pid_t pid, int (*process)(event_t *event)); void event__synthesize_threads(int (*process)(event_t *event)); diff --git a/tools/perf/util/map.c b/tools/perf/util/map.c index c1c55682534..d302e513e06 100644 --- a/tools/perf/util/map.c +++ b/tools/perf/util/map.c @@ -21,7 +21,7 @@ static int strcommon(const char *pathname, char *cwd, int cwdlen) } struct map *map__new(struct mmap_event *event, char *cwd, int cwdlen, - unsigned int sym_priv_size, symbol_filter_t filter) + unsigned int sym_priv_size) { struct map *self = malloc(sizeof(*self)); @@ -29,7 +29,6 @@ struct map *map__new(struct mmap_event *event, char *cwd, int cwdlen, const char *filename = event->filename; char newfilename[PATH_MAX]; int anon; - bool new_dso; if (cwd) { int n = strcommon(filename, cwd, cwdlen); @@ -52,23 +51,10 @@ struct map *map__new(struct mmap_event *event, char *cwd, int cwdlen, self->end = event->start + event->len; self->pgoff = event->pgoff; - self->dso = dsos__findnew(filename, sym_priv_size, &new_dso); + self->dso = dsos__findnew(filename, sym_priv_size); if (self->dso == NULL) goto out_delete; - if (new_dso) { - int nr = dso__load(self->dso, self, filter); - - if (nr < 0) - pr_warning("Failed to open %s, continuing " - "without symbols\n", - self->dso->long_name); - else if (nr == 0) - pr_warning("No symbols found in %s, maybe " - "install a debug package?\n", - self->dso->long_name); - } - if (self->dso == vdso || anon) self->map_ip = self->unmap_ip = identity__map_ip; else { @@ -82,6 +68,26 @@ out_delete: return NULL; } +struct symbol * +map__find_symbol(struct map *self, u64 ip, symbol_filter_t filter) +{ + if (!self->dso->loaded) { + int nr = dso__load(self->dso, self, filter); + + if (nr < 0) { + pr_warning("Failed to open %s, continuing without symbols\n", + self->dso->long_name); + return NULL; + } else if (nr == 0) { + pr_warning("No symbols found in %s, maybe install a debug package?\n", + self->dso->long_name); + return NULL; + } + } + + return self->dso->find_symbol(self->dso, ip); +} + struct map *map__clone(struct map *self) { struct map *map = malloc(sizeof(*self)); diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 8f0208ce237..0273d83f728 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -909,6 +909,8 @@ int dso__load(struct dso *self, struct map *map, symbol_filter_t filter) int ret = -1; int fd; + self->loaded = true; + if (!name) return -1; @@ -1019,6 +1021,8 @@ static int dso__load_module_sym(struct dso *self, struct map *map, { int err = 0, fd = open(self->long_name, O_RDONLY); + self->loaded = true; + if (fd < 0) { pr_err("%s: cannot open %s\n", __func__, self->long_name); return err; @@ -1214,6 +1218,8 @@ static int dso__load_vmlinux(struct dso *self, struct map *map, { int err, fd = open(vmlinux, O_RDONLY); + self->loaded = true; + if (fd < 0) return -1; @@ -1312,19 +1318,15 @@ static struct dso *dsos__find(const char *name) return NULL; } -struct dso *dsos__findnew(const char *name, unsigned int sym_priv_size, - bool *is_new) +struct dso *dsos__findnew(const char *name, unsigned int sym_priv_size) { struct dso *dso = dsos__find(name); if (!dso) { dso = dso__new(name, sym_priv_size); - if (dso) { + if (dso != NULL) dsos__add(dso); - *is_new = true; - } - } else - *is_new = false; + } return dso; } diff --git a/tools/perf/util/symbol.h b/tools/perf/util/symbol.h index 77b7b3e4241..432edbca780 100644 --- a/tools/perf/util/symbol.h +++ b/tools/perf/util/symbol.h @@ -46,6 +46,7 @@ struct dso { unsigned int sym_priv_size; unsigned char adjust_symbols; unsigned char slen_calculated; + bool loaded; unsigned char origin; const char *short_name; char *long_name; @@ -64,8 +65,7 @@ struct symbol *dso__find_symbol(struct dso *self, u64 ip); int dsos__load_kernel(const char *vmlinux, unsigned int sym_priv_size, symbol_filter_t filter, int modules); -struct dso *dsos__findnew(const char *name, unsigned int sym_priv_size, - bool *is_new); +struct dso *dsos__findnew(const char *name, unsigned int sym_priv_size); int dso__load(struct dso *self, struct map *map, symbol_filter_t filter); void dsos__fprintf(FILE *fp); -- cgit v1.2.3-18-g5258 From 8030c5f5a57e018fcdeb1f395d7adc123b48ced6 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Tue, 27 Oct 2009 16:42:53 -0400 Subject: perf/probes: Exit searching after finding target function Exit searching after finding real (not-inlined) function, because there should be no same symbol in that CU. Signed-off-by: Masami Hiramatsu Cc: Steven Rostedt Cc: Jim Keniston Cc: Ananth N Mavinakayanahalli Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: Frederic Weisbecker Cc: Jason Baron Cc: K.Prasad Cc: Peter Zijlstra Cc: Srikar Dronamraju LKML-Reference: <20091027204252.30545.19251.stgit@harusame> Signed-off-by: Ingo Molnar --- tools/perf/util/probe-finder.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/probe-finder.c b/tools/perf/util/probe-finder.c index 54e70718530..b98d35ef711 100644 --- a/tools/perf/util/probe-finder.c +++ b/tools/perf/util/probe-finder.c @@ -585,14 +585,14 @@ static int probefunc_callback(struct die_link *dlink, void *data) DIE_IF(ret != DW_DLV_OK); pr_debug("inline definition offset %lld\n", pf->inl_offs); - return 0; + return 0; /* Continue to search */ } /* Get probe address */ pf->addr = die_get_entrypc(dlink->die); pf->addr += pp->offset; /* TODO: Check the address in this function */ show_probepoint(dlink->die, pp->offset, pf); - /* Continue to search */ + return 1; /* Exit; no same symbol in this CU. */ } } else if (tag == DW_TAG_inlined_subroutine && pf->inl_offs) { if (die_get_abstract_origin(dlink->die) == pf->inl_offs) { -- cgit v1.2.3-18-g5258 From 46ab49267d338eb5056d0077e16346509b9e9284 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Tue, 27 Oct 2009 16:43:02 -0400 Subject: perf/probes: Improve command-line option of perf-probe Change command-line option from -P to --add, and accepting probes without --add too. perf probe --add "probe-define" or, just: perf probe "probe-define" Signed-off-by: Masami Hiramatsu Cc: Steven Rostedt Cc: Jim Keniston Cc: Ananth N Mavinakayanahalli Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: Frederic Weisbecker Cc: Jason Baron Cc: K.Prasad Cc: Peter Zijlstra Cc: Srikar Dronamraju LKML-Reference: <20091027204301.30545.48600.stgit@harusame> Signed-off-by: Ingo Molnar --- tools/perf/builtin-probe.c | 28 ++++++++++++++++++---------- 1 file changed, 18 insertions(+), 10 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-probe.c b/tools/perf/builtin-probe.c index dcb406c7f82..3370dabed15 100644 --- a/tools/perf/builtin-probe.c +++ b/tools/perf/builtin-probe.c @@ -65,8 +65,8 @@ static struct { #define semantic_error(msg ...) die("Semantic error :" msg) -static int parse_probepoint(const struct option *opt __used, - const char *str, int unset __used) +/* Parse a probe point. Note that any error must die. */ +static void parse_probepoint(const char *str) { char *argv[MAX_PROBE_ARGS + 2]; /* Event + probe + args */ int argc, i; @@ -75,9 +75,6 @@ static int parse_probepoint(const struct option *opt __used, char **event = &session.events[session.nr_probe]; int retp = 0; - if (!str) /* The end of probe points */ - return 0; - pr_debug("probe-definition(%d): %s\n", session.nr_probe, str); if (++session.nr_probe == MAX_PROBES) semantic_error("Too many probes"); @@ -176,6 +173,13 @@ static int parse_probepoint(const struct option *opt __used, } pr_debug("%d arguments\n", pp->nr_args); +} + +static int opt_add_probepoint(const struct option *opt __used, + const char *str, int unset __used) +{ + if (str) + parse_probepoint(str); return 0; } @@ -211,7 +215,8 @@ static int open_default_vmlinux(void) #endif static const char * const probe_usage[] = { - "perf probe [] -P 'PROBEDEF' [-P 'PROBEDEF' ...]", + "perf probe [] 'PROBEDEF' ['PROBEDEF' ...]", + "perf probe [] --add 'PROBEDEF' [--add 'PROBEDEF' ...]", NULL }; @@ -222,7 +227,7 @@ static const struct option options[] = { OPT_STRING('k', "vmlinux", &session.vmlinux, "file", "vmlinux/module pathname"), #endif - OPT_CALLBACK('P', "probe", NULL, + OPT_CALLBACK('a', "add", NULL, #ifdef NO_LIBDWARF "p|r:[GRP/]NAME FUNC[+OFFS] [ARG ...]", #else @@ -243,7 +248,7 @@ static const struct option options[] = { "\t\tARG:\tProbe argument (local variable name or\n" #endif "\t\t\tkprobe-tracer argument format is supported.)\n", - parse_probepoint), + opt_add_probepoint), OPT_END() }; @@ -296,8 +301,11 @@ int cmd_probe(int argc, const char **argv, const char *prefix __used) char buf[MAX_CMDLEN]; argc = parse_options(argc, argv, options, probe_usage, - PARSE_OPT_STOP_AT_NON_OPTION); - if (argc || session.nr_probe == 0) + PARSE_OPT_STOP_AT_NON_OPTION); + for (i = 0; i < argc; i++) + parse_probe_event(argv[i]); + + if (session.nr_probe == 0) usage_with_options(probe_usage, options); #ifdef NO_LIBDWARF -- cgit v1.2.3-18-g5258 From 253977b0d87fbb793f12b1661a763ae264028ccf Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Tue, 27 Oct 2009 16:43:10 -0400 Subject: perf/probes: Improve probe point syntax of perf-probe This changes probe point syntax of perf-probe as below [:ABS_LN] [ARGS] or [+OFFS|%return][@SRC] [ARGS] And event name and event group name are automatically generated based on probe-symbol and offset as below. perfprobes/SYMBOL_OFFSET[_NUM] Where SYMBOL is the probing symbol and OFFSET is the byte offset from the symbol. Signed-off-by: Masami Hiramatsu Cc: Steven Rostedt Cc: Jim Keniston Cc: Ananth N Mavinakayanahalli Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: Frederic Weisbecker Cc: Jason Baron Cc: K.Prasad Cc: Peter Zijlstra Cc: Srikar Dronamraju LKML-Reference: <20091027204310.30545.84984.stgit@harusame> Signed-off-by: Ingo Molnar --- tools/perf/builtin-probe.c | 181 +++++++++++++++++++++++++---------------- tools/perf/util/probe-finder.c | 10 +++ tools/perf/util/probe-finder.h | 2 + 3 files changed, 123 insertions(+), 70 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-probe.c b/tools/perf/builtin-probe.c index 3370dabed15..92b4c491f23 100644 --- a/tools/perf/builtin-probe.c +++ b/tools/perf/builtin-probe.c @@ -52,6 +52,7 @@ const char *default_search_path[NR_SEARCH_PATH] = { #define MAX_PATH_LEN 256 #define MAX_PROBES 128 #define MAX_PROBE_ARGS 128 +#define PERFPROBE_GROUP "perfprobe" /* Session management structure */ static struct { @@ -60,20 +61,100 @@ static struct { int need_dwarf; int nr_probe; struct probe_point probes[MAX_PROBES]; - char *events[MAX_PROBES]; } session; #define semantic_error(msg ...) die("Semantic error :" msg) -/* Parse a probe point. Note that any error must die. */ -static void parse_probepoint(const char *str) +/* Parse probe point. Return 1 if return probe */ +static void parse_probe_point(char *arg, struct probe_point *pp) +{ + char *ptr, *tmp; + char c, nc; + /* + * + * perf probe SRC:LN + * perf probe FUNC[+OFFS|%return][@SRC] + */ + + ptr = strpbrk(arg, ":+@%"); + if (ptr) { + nc = *ptr; + *ptr++ = '\0'; + } + + /* Check arg is function or file and copy it */ + if (strchr(arg, '.')) /* File */ + pp->file = strdup(arg); + else /* Function */ + pp->function = strdup(arg); + DIE_IF(pp->file == NULL && pp->function == NULL); + + /* Parse other options */ + while (ptr) { + arg = ptr; + c = nc; + ptr = strpbrk(arg, ":+@%"); + if (ptr) { + nc = *ptr; + *ptr++ = '\0'; + } + switch (c) { + case ':': /* Line number */ + pp->line = strtoul(arg, &tmp, 0); + if (*tmp != '\0') + semantic_error("There is non-digit charactor" + " in line number."); + break; + case '+': /* Byte offset from a symbol */ + pp->offset = strtoul(arg, &tmp, 0); + if (*tmp != '\0') + semantic_error("There is non-digit charactor" + " in offset."); + break; + case '@': /* File name */ + if (pp->file) + semantic_error("SRC@SRC is not allowed."); + pp->file = strdup(arg); + DIE_IF(pp->file == NULL); + if (ptr) + semantic_error("@SRC must be the last " + "option."); + break; + case '%': /* Probe places */ + if (strcmp(arg, "return") == 0) { + pp->retprobe = 1; + } else /* Others not supported yet */ + semantic_error("%%%s is not supported.", arg); + break; + default: + DIE_IF("Program has a bug."); + break; + } + } + + /* Exclusion check */ + if (pp->line && pp->function) + semantic_error("Function-relative line number is not" + " supported yet."); + if (!pp->line && pp->file && !pp->function) + semantic_error("File always requires line number."); + if (pp->offset && !pp->function) + semantic_error("Offset requires an entry function."); + if (pp->retprobe && !pp->function) + semantic_error("Return probe requires an entry function."); + if (pp->offset && pp->retprobe) + semantic_error("Offset can't be used with return probe."); + + pr_debug("symbol:%s file:%s line:%d offset:%d, return:%d\n", + pp->function, pp->file, pp->line, pp->offset, pp->retprobe); +} + +/* Parse an event definition. Note that any error must die. */ +static void parse_probe_event(const char *str) { char *argv[MAX_PROBE_ARGS + 2]; /* Event + probe + args */ int argc, i; - char *arg, *ptr; struct probe_point *pp = &session.probes[session.nr_probe]; - char **event = &session.events[session.nr_probe]; - int retp = 0; pr_debug("probe-definition(%d): %s\n", session.nr_probe, str); if (++session.nr_probe == MAX_PROBES) @@ -103,70 +184,28 @@ static void parse_probepoint(const char *str) pr_debug("argv[%d]=%s\n", argc, argv[argc - 1]); } } while (*str != '\0'); - if (argc < 2) - semantic_error("Need event-name and probe-point at least."); - - /* Parse the event name */ - if (argv[0][0] == 'r') - retp = 1; - else if (argv[0][0] != 'p') - semantic_error("You must specify 'p'(kprobe) or" - " 'r'(kretprobe) first."); - /* TODO: check event name */ - *event = argv[0]; + if (!argc) + semantic_error("An empty argument."); /* Parse probe point */ - arg = argv[1]; - if (arg[0] == '@') { - /* Source Line */ - arg++; - ptr = strchr(arg, ':'); - if (!ptr || !isdigit(ptr[1])) - semantic_error("Line number is required."); - *ptr++ = '\0'; - if (strlen(arg) == 0) - semantic_error("No file name."); - pp->file = strdup(arg); - pp->line = atoi(ptr); - if (!pp->file || !pp->line) - semantic_error("Failed to parse line."); - pr_debug("file:%s line:%d\n", pp->file, pp->line); - } else { - /* Function name */ - ptr = strchr(arg, '+'); - if (ptr) { - if (!isdigit(ptr[1])) - semantic_error("Offset is required."); - *ptr++ = '\0'; - pp->offset = atoi(ptr); - } else - ptr = arg; - ptr = strchr(ptr, '@'); - if (ptr) { - *ptr++ = '\0'; - pp->file = strdup(ptr); - } - pp->function = strdup(arg); - pr_debug("symbol:%s file:%s offset:%d\n", - pp->function, pp->file, pp->offset); - } - free(argv[1]); + parse_probe_point(argv[0], pp); + free(argv[0]); if (pp->file) session.need_dwarf = 1; /* Copy arguments */ - pp->nr_args = argc - 2; + pp->nr_args = argc - 1; if (pp->nr_args > 0) { pp->args = (char **)malloc(sizeof(char *) * pp->nr_args); if (!pp->args) die("malloc"); - memcpy(pp->args, &argv[2], sizeof(char *) * pp->nr_args); + memcpy(pp->args, &argv[1], sizeof(char *) * pp->nr_args); } /* Ensure return probe has no C argument */ for (i = 0; i < pp->nr_args; i++) if (is_c_varname(pp->args[i])) { - if (retp) + if (pp->retprobe) semantic_error("You can't specify local" " variable for kretprobe"); session.need_dwarf = 1; @@ -175,11 +214,11 @@ static void parse_probepoint(const char *str) pr_debug("%d arguments\n", pp->nr_args); } -static int opt_add_probepoint(const struct option *opt __used, +static int opt_add_probe_event(const struct option *opt __used, const char *str, int unset __used) { if (str) - parse_probepoint(str); + parse_probe_event(str); return 0; } @@ -229,17 +268,16 @@ static const struct option options[] = { #endif OPT_CALLBACK('a', "add", NULL, #ifdef NO_LIBDWARF - "p|r:[GRP/]NAME FUNC[+OFFS] [ARG ...]", + "FUNC[+OFFS|%return] [ARG ...]", #else - "p|r:[GRP/]NAME FUNC[+OFFS][@SRC]|@SRC:LINE [ARG ...]", + "FUNC[+OFFS|%return][@SRC]|SRC:LINE [ARG ...]", #endif "probe point definition, where\n" - "\t\tp:\tkprobe probe\n" - "\t\tr:\tkretprobe probe\n" "\t\tGRP:\tGroup name (optional)\n" "\t\tNAME:\tEvent name\n" "\t\tFUNC:\tFunction name\n" "\t\tOFFS:\tOffset from function entry (in byte)\n" + "\t\t%return:\tPut the probe at function return\n" #ifdef NO_LIBDWARF "\t\tARG:\tProbe argument (only \n" #else @@ -248,7 +286,7 @@ static const struct option options[] = { "\t\tARG:\tProbe argument (local variable name or\n" #endif "\t\t\tkprobe-tracer argument format is supported.)\n", - opt_add_probepoint), + opt_add_probe_event), OPT_END() }; @@ -266,7 +304,7 @@ static int write_new_event(int fd, const char *buf) #define MAX_CMDLEN 256 -static int synthesize_probepoint(struct probe_point *pp) +static int synthesize_probe_event(struct probe_point *pp) { char *buf; int i, len, ret; @@ -316,12 +354,12 @@ int cmd_probe(int argc, const char **argv, const char *prefix __used) /* Synthesize probes without dwarf */ for (j = 0; j < session.nr_probe; j++) { #ifndef NO_LIBDWARF - if (session.events[j][0] != 'r') { + if (!session.probes[j].retprobe) { session.need_dwarf = 1; continue; } #endif - ret = synthesize_probepoint(&session.probes[j]); + ret = synthesize_probe_event(&session.probes[j]); if (ret == -E2BIG) semantic_error("probe point is too long."); else if (ret < 0) @@ -349,7 +387,6 @@ int cmd_probe(int argc, const char **argv, const char *prefix __used) ret = find_probepoint(fd, pp); if (ret <= 0) die("No probe point found.\n"); - pr_debug("probe event %s found\n", session.events[j]); } close(fd); @@ -364,13 +401,17 @@ setup_probes: for (j = 0; j < session.nr_probe; j++) { pp = &session.probes[j]; if (pp->found == 1) { - snprintf(buf, MAX_CMDLEN, "%s %s\n", - session.events[j], pp->probes[0]); + snprintf(buf, MAX_CMDLEN, "%c:%s/%s_%x %s\n", + pp->retprobe ? 'r' : 'p', PERFPROBE_GROUP, + pp->function, pp->offset, pp->probes[0]); write_new_event(fd, buf); } else for (i = 0; i < pp->found; i++) { - snprintf(buf, MAX_CMDLEN, "%s%d %s\n", - session.events[j], i, pp->probes[i]); + snprintf(buf, MAX_CMDLEN, "%c:%s/%s_%x_%d %s\n", + pp->retprobe ? 'r' : 'p', + PERFPROBE_GROUP, + pp->function, pp->offset, i, + pp->probes[0]); write_new_event(fd, buf); } } diff --git a/tools/perf/util/probe-finder.c b/tools/perf/util/probe-finder.c index b98d35ef711..6d3bac9f947 100644 --- a/tools/perf/util/probe-finder.c +++ b/tools/perf/util/probe-finder.c @@ -483,10 +483,20 @@ static void show_probepoint(Dwarf_Die sp_die, Dwarf_Signed offs, if (ret == DW_DLV_OK) { ret = snprintf(tmp, MAX_PROBE_BUFFER, "%s+%u", name, (unsigned int)offs); + /* Copy the function name if possible */ + if (!pp->function) { + pp->function = strdup(name); + pp->offset = offs; + } dwarf_dealloc(__dw_debug, name, DW_DLA_STRING); } else { /* This function has no name. */ ret = snprintf(tmp, MAX_PROBE_BUFFER, "0x%llx", pf->addr); + if (!pp->function) { + /* TODO: Use _stext */ + pp->function = strdup(""); + pp->offset = (int)pf->addr; + } } DIE_IF(ret < 0); DIE_IF(ret >= MAX_PROBE_BUFFER); diff --git a/tools/perf/util/probe-finder.h b/tools/perf/util/probe-finder.h index d17fafc2135..240d6cb3cc2 100644 --- a/tools/perf/util/probe-finder.h +++ b/tools/perf/util/probe-finder.h @@ -22,6 +22,8 @@ struct probe_point { int nr_args; /* Number of arguments */ char **args; /* Arguments */ + int retprobe; /* Return probe */ + /* Output */ int found; /* Number of found probe points */ char *probes[MAX_PROBES]; /* Output buffers (will be allocated)*/ -- cgit v1.2.3-18-g5258 From b0ef07324310d66f660a311d4a8d669eda74f801 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Tue, 27 Oct 2009 16:43:19 -0400 Subject: perf/probes: Support function entry relative line number Add function-entry relative line number specifying support to perf-probe. This allows users to define probes by line number from entry of the function. e.g. perf probe schedule:16 Signed-off-by: Masami Hiramatsu Cc: Steven Rostedt Cc: Jim Keniston Cc: Ananth N Mavinakayanahalli Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: Frederic Weisbecker Cc: Jason Baron Cc: K.Prasad Cc: Peter Zijlstra Cc: Srikar Dronamraju LKML-Reference: <20091027204319.30545.30678.stgit@harusame> Signed-off-by: Ingo Molnar --- tools/perf/builtin-probe.c | 14 ++++---- tools/perf/util/probe-finder.c | 79 ++++++++++++++++++++++++++++++++---------- tools/perf/util/probe-finder.h | 2 ++ 3 files changed, 70 insertions(+), 25 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-probe.c b/tools/perf/builtin-probe.c index 92b4c491f23..a99a366230a 100644 --- a/tools/perf/builtin-probe.c +++ b/tools/perf/builtin-probe.c @@ -133,17 +133,16 @@ static void parse_probe_point(char *arg, struct probe_point *pp) } /* Exclusion check */ - if (pp->line && pp->function) - semantic_error("Function-relative line number is not" - " supported yet."); + if (pp->line && pp->offset) + semantic_error("Offset can't be used with line number."); if (!pp->line && pp->file && !pp->function) semantic_error("File always requires line number."); if (pp->offset && !pp->function) semantic_error("Offset requires an entry function."); if (pp->retprobe && !pp->function) semantic_error("Return probe requires an entry function."); - if (pp->offset && pp->retprobe) - semantic_error("Offset can't be used with return probe."); + if ((pp->offset || pp->line) && pp->retprobe) + semantic_error("Offset/Line can't be used with return probe."); pr_debug("symbol:%s file:%s line:%d offset:%d, return:%d\n", pp->function, pp->file, pp->line, pp->offset, pp->retprobe); @@ -270,7 +269,7 @@ static const struct option options[] = { #ifdef NO_LIBDWARF "FUNC[+OFFS|%return] [ARG ...]", #else - "FUNC[+OFFS|%return][@SRC]|SRC:LINE [ARG ...]", + "FUNC[+OFFS|%return|:RLN][@SRC]|SRC:ALN [ARG ...]", #endif "probe point definition, where\n" "\t\tGRP:\tGroup name (optional)\n" @@ -282,7 +281,8 @@ static const struct option options[] = { "\t\tARG:\tProbe argument (only \n" #else "\t\tSRC:\tSource code path\n" - "\t\tLINE:\tLine number\n" + "\t\tRLN:\tRelative line number from function entry.\n" + "\t\tALN:\tAbsolute line number in file.\n" "\t\tARG:\tProbe argument (local variable name or\n" #endif "\t\t\tkprobe-tracer argument format is supported.)\n", diff --git a/tools/perf/util/probe-finder.c b/tools/perf/util/probe-finder.c index 6d3bac9f947..db96186e02a 100644 --- a/tools/perf/util/probe-finder.c +++ b/tools/perf/util/probe-finder.c @@ -114,7 +114,7 @@ static int strtailcmp(const char *s1, const char *s2) } /* Find the fileno of the target file. */ -static Dwarf_Unsigned die_get_fileno(Dwarf_Die cu_die, const char *fname) +static Dwarf_Unsigned cu_find_fileno(Dwarf_Die cu_die, const char *fname) { Dwarf_Signed cnt, i; Dwarf_Unsigned found = 0; @@ -335,6 +335,36 @@ static int attr_get_locdesc(Dwarf_Attribute attr, Dwarf_Locdesc *desc, return ret; } +/* Get decl_file attribute value (file number) */ +static Dwarf_Unsigned die_get_decl_file(Dwarf_Die sp_die) +{ + Dwarf_Attribute attr; + Dwarf_Unsigned fno; + int ret; + + ret = dwarf_attr(sp_die, DW_AT_decl_file, &attr, &__dw_error); + DIE_IF(ret != DW_DLV_OK); + dwarf_formudata(attr, &fno, &__dw_error); + DIE_IF(ret != DW_DLV_OK); + dwarf_dealloc(__dw_debug, attr, DW_DLA_ATTR); + return fno; +} + +/* Get decl_line attribute value (line number) */ +static Dwarf_Unsigned die_get_decl_line(Dwarf_Die sp_die) +{ + Dwarf_Attribute attr; + Dwarf_Unsigned lno; + int ret; + + ret = dwarf_attr(sp_die, DW_AT_decl_line, &attr, &__dw_error); + DIE_IF(ret != DW_DLV_OK); + dwarf_formudata(attr, &lno, &__dw_error); + DIE_IF(ret != DW_DLV_OK); + dwarf_dealloc(__dw_debug, attr, DW_DLA_ATTR); + return lno; +} + /* * Probe finder related functions */ @@ -501,6 +531,7 @@ static void show_probepoint(Dwarf_Die sp_die, Dwarf_Signed offs, DIE_IF(ret < 0); DIE_IF(ret >= MAX_PROBE_BUFFER); len = ret; + pr_debug("Probe point found: %s\n", tmp); /* Find each argument */ get_current_frame_base(sp_die, pf); @@ -536,17 +567,16 @@ static int probeaddr_callback(struct die_link *dlink, void *data) } /* Find probe point from its line number */ -static void find_by_line(Dwarf_Die cu_die, struct probe_finder *pf) +static void find_by_line(struct probe_finder *pf) { - struct probe_point *pp = pf->pp; - Dwarf_Signed cnt, i; + Dwarf_Signed cnt, i, clm; Dwarf_Line *lines; Dwarf_Unsigned lineno = 0; Dwarf_Addr addr; Dwarf_Unsigned fno; int ret; - ret = dwarf_srclines(cu_die, &lines, &cnt, &__dw_error); + ret = dwarf_srclines(pf->cu_die, &lines, &cnt, &__dw_error); DIE_IF(ret != DW_DLV_OK); for (i = 0; i < cnt; i++) { @@ -557,15 +587,20 @@ static void find_by_line(Dwarf_Die cu_die, struct probe_finder *pf) ret = dwarf_lineno(lines[i], &lineno, &__dw_error); DIE_IF(ret != DW_DLV_OK); - if (lineno != (Dwarf_Unsigned)pp->line) + if (lineno != pf->lno) continue; + ret = dwarf_lineoff(lines[i], &clm, &__dw_error); + DIE_IF(ret != DW_DLV_OK); + ret = dwarf_lineaddr(lines[i], &addr, &__dw_error); DIE_IF(ret != DW_DLV_OK); - pr_debug("Probe point found: 0x%llx\n", addr); + pr_debug("Probe line found: line[%d]:%u,%d addr:0x%llx\n", + (int)i, (unsigned)lineno, (int)clm, addr); pf->addr = addr; /* Search a real subprogram including this line, */ - ret = search_die_from_children(cu_die, probeaddr_callback, pf); + ret = search_die_from_children(pf->cu_die, + probeaddr_callback, pf); if (ret == 0) die("Probe point is not found in subprograms.\n"); /* Continuing, because target line might be inlined. */ @@ -587,6 +622,13 @@ static int probefunc_callback(struct die_link *dlink, void *data) DIE_IF(ret == DW_DLV_ERROR); if (tag == DW_TAG_subprogram) { if (die_compare_name(dlink->die, pp->function) == 0) { + if (pp->line) { /* Function relative line */ + pf->fno = die_get_decl_file(dlink->die); + pf->lno = die_get_decl_line(dlink->die) + + pp->line; + find_by_line(pf); + return 1; + } if (die_inlined_subprogram(dlink->die)) { /* Inlined function, save it. */ ret = dwarf_die_CU_offset(dlink->die, @@ -631,9 +673,9 @@ found: return 0; } -static void find_by_func(Dwarf_Die cu_die, struct probe_finder *pf) +static void find_by_func(struct probe_finder *pf) { - search_die_from_children(cu_die, probefunc_callback, pf); + search_die_from_children(pf->cu_die, probefunc_callback, pf); } /* Find a probe point */ @@ -641,7 +683,6 @@ int find_probepoint(int fd, struct probe_point *pp) { Dwarf_Half addr_size = 0; Dwarf_Unsigned next_cuh = 0; - Dwarf_Die cu_die = 0; int cu_number = 0, ret; struct probe_finder pf = {.pp = pp}; @@ -659,25 +700,27 @@ int find_probepoint(int fd, struct probe_point *pp) break; /* Get the DIE(Debugging Information Entry) of this CU */ - ret = dwarf_siblingof(__dw_debug, 0, &cu_die, &__dw_error); + ret = dwarf_siblingof(__dw_debug, 0, &pf.cu_die, &__dw_error); DIE_IF(ret != DW_DLV_OK); /* Check if target file is included. */ if (pp->file) - pf.fno = die_get_fileno(cu_die, pp->file); + pf.fno = cu_find_fileno(pf.cu_die, pp->file); if (!pp->file || pf.fno) { /* Save CU base address (for frame_base) */ - ret = dwarf_lowpc(cu_die, &pf.cu_base, &__dw_error); + ret = dwarf_lowpc(pf.cu_die, &pf.cu_base, &__dw_error); DIE_IF(ret == DW_DLV_ERROR); if (ret == DW_DLV_NO_ENTRY) pf.cu_base = 0; - if (pp->line) - find_by_line(cu_die, &pf); if (pp->function) - find_by_func(cu_die, &pf); + find_by_func(&pf); + else { + pf.lno = pp->line; + find_by_line(&pf); + } } - dwarf_dealloc(__dw_debug, cu_die, DW_DLA_DIE); + dwarf_dealloc(__dw_debug, pf.cu_die, DW_DLA_DIE); } ret = dwarf_finish(__dw_debug, &__dw_error); DIE_IF(ret != DW_DLV_OK); diff --git a/tools/perf/util/probe-finder.h b/tools/perf/util/probe-finder.h index 240d6cb3cc2..bdebca6697d 100644 --- a/tools/perf/util/probe-finder.h +++ b/tools/perf/util/probe-finder.h @@ -41,7 +41,9 @@ struct probe_finder { /* For function searching */ Dwarf_Addr addr; /* Address */ Dwarf_Unsigned fno; /* File number */ + Dwarf_Unsigned lno; /* Line number */ Dwarf_Off inl_offs; /* Inline offset */ + Dwarf_Die cu_die; /* Current CU */ /* For variable searching */ Dwarf_Addr cu_base; /* Current CU base address */ -- cgit v1.2.3-18-g5258 From afb7b4f08e274cecd8337f9444affa288a9cd4c1 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 30 Oct 2009 16:28:23 -0200 Subject: perf tools: Factor out the map initialization Signed-off-by: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: Peter Zijlstra Cc: Paul Mackerras Cc: Mike Galbraith LKML-Reference: <1256927305-4628-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/util/event.h | 2 ++ tools/perf/util/map.c | 28 ++++++++++++++++++---------- tools/perf/util/symbol.c | 12 +++--------- 3 files changed, 23 insertions(+), 19 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/event.h b/tools/perf/util/event.h index 3064a05f0f5..4a158a01bb9 100644 --- a/tools/perf/util/event.h +++ b/tools/perf/util/event.h @@ -105,6 +105,8 @@ struct symbol; typedef int (*symbol_filter_t)(struct map *map, struct symbol *sym); +void map__init(struct map *self, u64 start, u64 end, u64 pgoff, + struct dso *dso); struct map *map__new(struct mmap_event *event, char *cwd, int cwdlen, unsigned int sym_priv_size); struct map *map__clone(struct map *self); diff --git a/tools/perf/util/map.c b/tools/perf/util/map.c index d302e513e06..3b7ce1bf9f8 100644 --- a/tools/perf/util/map.c +++ b/tools/perf/util/map.c @@ -20,6 +20,18 @@ static int strcommon(const char *pathname, char *cwd, int cwdlen) return n; } +void map__init(struct map *self, u64 start, u64 end, u64 pgoff, + struct dso *dso) +{ + self->start = start; + self->end = end; + self->pgoff = pgoff; + self->dso = dso; + self->map_ip = map__map_ip; + self->unmap_ip = map__unmap_ip; + RB_CLEAR_NODE(&self->rb_node); +} + struct map *map__new(struct mmap_event *event, char *cwd, int cwdlen, unsigned int sym_priv_size) { @@ -28,6 +40,7 @@ struct map *map__new(struct mmap_event *event, char *cwd, int cwdlen, if (self != NULL) { const char *filename = event->filename; char newfilename[PATH_MAX]; + struct dso *dso; int anon; if (cwd) { @@ -47,20 +60,15 @@ struct map *map__new(struct mmap_event *event, char *cwd, int cwdlen, filename = newfilename; } - self->start = event->start; - self->end = event->start + event->len; - self->pgoff = event->pgoff; - - self->dso = dsos__findnew(filename, sym_priv_size); - if (self->dso == NULL) + dso = dsos__findnew(filename, sym_priv_size); + if (dso == NULL) goto out_delete; + map__init(self, event->start, event->start + event->len, + event->pgoff, dso); + if (self->dso == vdso || anon) self->map_ip = self->unmap_ip = identity__map_ip; - else { - self->map_ip = map__map_ip; - self->unmap_ip = map__unmap_ip; - } } return self; out_delete: diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 0273d83f728..13677b5dbe5 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -1132,18 +1132,12 @@ static struct map *map__new2(u64 start, struct dso *dso) struct map *self = malloc(sizeof(*self)); if (self != NULL) { - self->start = start; /* - * Will be filled after we load all the symbols + * ->end will be filled after we load all the symbols */ - self->end = 0; - - self->pgoff = 0; - self->dso = dso; - self->map_ip = map__map_ip; - self->unmap_ip = map__unmap_ip; - RB_CLEAR_NODE(&self->rb_node); + map__init(self, start, 0, 0, dso); } + return self; } -- cgit v1.2.3-18-g5258 From 00a192b395b0606ad0265243844b3cd68e73420a Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 30 Oct 2009 16:28:24 -0200 Subject: perf tools: Simplify the symbol priv area mechanism Before we were storing this in the DSO, but in fact this is a property of the 'symbol' class, not something that will vary among DSOs, so move it to a global variable and initialize it using the existing symbol__init routine. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: Peter Zijlstra Cc: Paul Mackerras Cc: Mike Galbraith LKML-Reference: <1256927305-4628-2-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/builtin-annotate.c | 21 ++++++------ tools/perf/builtin-report.c | 4 +-- tools/perf/builtin-sched.c | 2 +- tools/perf/builtin-timechart.c | 2 +- tools/perf/builtin-top.c | 12 +++---- tools/perf/builtin-trace.c | 2 +- tools/perf/util/data_map.c | 2 +- tools/perf/util/event.h | 3 +- tools/perf/util/map.c | 5 ++- tools/perf/util/symbol.c | 73 ++++++++++++++++++++---------------------- tools/perf/util/symbol.h | 24 +++++++------- 11 files changed, 70 insertions(+), 80 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c index 8688bfee42a..77d50a6d680 100644 --- a/tools/perf/builtin-annotate.c +++ b/tools/perf/builtin-annotate.c @@ -55,11 +55,11 @@ struct sym_priv { static const char *sym_hist_filter; -static int symbol_filter(struct map *map, struct symbol *sym) +static int symbol_filter(struct map *map __used, struct symbol *sym) { if (sym_hist_filter == NULL || strcmp(sym->name, sym_hist_filter) == 0) { - struct sym_priv *priv = dso__sym_priv(map->dso, sym); + struct sym_priv *priv = symbol__priv(sym); const int size = (sizeof(*priv->hist) + (sym->end - sym->start) * sizeof(u64)); @@ -92,7 +92,7 @@ static void hist_hit(struct hist_entry *he, u64 ip) if (!sym || !he->map) return; - priv = dso__sym_priv(he->map->dso, sym); + priv = symbol__priv(sym); if (!priv->hist) return; @@ -202,8 +202,7 @@ got_map: static int process_mmap_event(event_t *event, unsigned long offset, unsigned long head) { - struct map *map = map__new(&event->mmap, NULL, 0, - sizeof(struct sym_priv)); + struct map *map = map__new(&event->mmap, NULL, 0); struct thread *thread = threads__findnew(event->mmap.pid); dump_printf("%p [%p]: PERF_RECORD_MMAP %d: [%p(%p) @ %p]: %s\n", @@ -355,7 +354,7 @@ static int parse_line(FILE *file, struct hist_entry *he, u64 len) unsigned int hits = 0; double percent = 0.0; const char *color; - struct sym_priv *priv = dso__sym_priv(he->map->dso, sym); + struct sym_priv *priv = symbol__priv(sym); struct sym_ext *sym_ext = priv->ext; struct sym_hist *h = priv->hist; @@ -422,7 +421,7 @@ static void insert_source_line(struct sym_ext *sym_ext) static void free_source_line(struct hist_entry *he, int len) { - struct sym_priv *priv = dso__sym_priv(he->map->dso, he->sym); + struct sym_priv *priv = symbol__priv(he->sym); struct sym_ext *sym_ext = priv->ext; int i; @@ -446,7 +445,7 @@ get_source_line(struct hist_entry *he, int len, const char *filename) int i; char cmd[PATH_MAX * 2]; struct sym_ext *sym_ext; - struct sym_priv *priv = dso__sym_priv(he->map->dso, sym); + struct sym_priv *priv = symbol__priv(sym); struct sym_hist *h = priv->hist; if (!h->sum) @@ -589,7 +588,7 @@ static void find_annotations(void) if (he->sym == NULL) continue; - priv = dso__sym_priv(he->map->dso, he->sym); + priv = symbol__priv(he->sym); if (priv->hist == NULL) continue; @@ -637,7 +636,7 @@ static int __cmd_annotate(void) exit(0); } - if (load_kernel(sizeof(struct sym_priv), symbol_filter) < 0) { + if (load_kernel(symbol_filter) < 0) { perror("failed to load kernel symbols"); return EXIT_FAILURE; } @@ -769,7 +768,7 @@ static void setup_sorting(void) int cmd_annotate(int argc, const char **argv, const char *prefix __used) { - symbol__init(); + symbol__init(sizeof(struct sym_priv)); page_size = getpagesize(); diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index f1bcd35bd22..1a806d5f05c 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -751,7 +751,7 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) static int process_mmap_event(event_t *event, unsigned long offset, unsigned long head) { - struct map *map = map__new(&event->mmap, cwd, cwdlen, 0); + struct map *map = map__new(&event->mmap, cwd, cwdlen); struct thread *thread = threads__findnew(event->mmap.pid); dump_printf("%p [%p]: PERF_RECORD_MMAP %d/%d: [%p(%p) @ %p]: %s\n", @@ -1093,7 +1093,7 @@ static void setup_list(struct strlist **list, const char *list_str, int cmd_report(int argc, const char **argv, const char *prefix __used) { - symbol__init(); + symbol__init(0); argc = parse_options(argc, argv, options, report_usage, 0); diff --git a/tools/perf/builtin-sched.c b/tools/perf/builtin-sched.c index 9a48d9626be..df44b756cec 100644 --- a/tools/perf/builtin-sched.c +++ b/tools/perf/builtin-sched.c @@ -1937,7 +1937,7 @@ static int __cmd_record(int argc, const char **argv) int cmd_sched(int argc, const char **argv, const char *prefix __used) { - symbol__init(); + symbol__init(0); argc = parse_options(argc, argv, sched_options, sched_usage, PARSE_OPT_STOP_AT_NON_OPTION); diff --git a/tools/perf/builtin-timechart.c b/tools/perf/builtin-timechart.c index 0a2f22261c3..665877e4a94 100644 --- a/tools/perf/builtin-timechart.c +++ b/tools/perf/builtin-timechart.c @@ -1266,7 +1266,7 @@ static const struct option options[] = { int cmd_timechart(int argc, const char **argv, const char *prefix __used) { - symbol__init(); + symbol__init(0); page_size = getpagesize(); diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index ee87640b335..2aea913f7eb 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -789,7 +789,7 @@ static int symbol_filter(struct map *map, struct symbol *sym) strstr(name, "_text_end")) return 1; - syme = dso__sym_priv(map->dso, sym); + syme = symbol__priv(sym); syme->map = map; pthread_mutex_init(&syme->source_lock, NULL); if (!sym_filter_entry && sym_filter && !strcmp(name, sym_filter)) @@ -807,8 +807,7 @@ static int symbol_filter(struct map *map, struct symbol *sym) static int parse_symbols(void) { - if (dsos__load_kernel(vmlinux_name, sizeof(struct sym_entry), - symbol_filter, 1) <= 0) + if (dsos__load_kernel(vmlinux_name, symbol_filter, 1) <= 0) return -1; if (dump_symtab) @@ -859,7 +858,7 @@ static void event__process_sample(const event_t *self, int counter) return; } - syme = dso__sym_priv(map->dso, sym); + syme = symbol__priv(sym); if (!syme->skip) { syme->count[counter]++; @@ -878,8 +877,7 @@ static void event__process_mmap(event_t *self) struct thread *thread = threads__findnew(self->mmap.pid); if (thread != NULL) { - struct map *map = map__new(&self->mmap, NULL, 0, - sizeof(struct sym_entry)); + struct map *map = map__new(&self->mmap, NULL, 0); if (map != NULL) thread__insert_map(thread, map); } @@ -1176,7 +1174,7 @@ int cmd_top(int argc, const char **argv, const char *prefix __used) { int counter; - symbol__init(); + symbol__init(sizeof(struct sym_entry)); page_size = sysconf(_SC_PAGE_SIZE); diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-trace.c index e566bbe3f22..d042d656c56 100644 --- a/tools/perf/builtin-trace.c +++ b/tools/perf/builtin-trace.c @@ -151,7 +151,7 @@ static const struct option options[] = { int cmd_trace(int argc, const char **argv, const char *prefix __used) { - symbol__init(); + symbol__init(0); argc = parse_options(argc, argv, options, annotate_usage, 0); if (argc) { diff --git a/tools/perf/util/data_map.c b/tools/perf/util/data_map.c index 18accb8fee4..c458db9ede6 100644 --- a/tools/perf/util/data_map.c +++ b/tools/perf/util/data_map.c @@ -130,7 +130,7 @@ int mmap_dispatch_perf_file(struct perf_header **pheader, if (curr_handler->sample_type_check(sample_type) < 0) exit(-1); - if (load_kernel(0, NULL) < 0) { + if (load_kernel(NULL) < 0) { perror("failed to load kernel symbols"); return EXIT_FAILURE; } diff --git a/tools/perf/util/event.h b/tools/perf/util/event.h index 4a158a01bb9..0a443bea68d 100644 --- a/tools/perf/util/event.h +++ b/tools/perf/util/event.h @@ -107,8 +107,7 @@ typedef int (*symbol_filter_t)(struct map *map, struct symbol *sym); void map__init(struct map *self, u64 start, u64 end, u64 pgoff, struct dso *dso); -struct map *map__new(struct mmap_event *event, char *cwd, int cwdlen, - unsigned int sym_priv_size); +struct map *map__new(struct mmap_event *event, char *cwd, int cwdlen); struct map *map__clone(struct map *self); int map__overlap(struct map *l, struct map *r); size_t map__fprintf(struct map *self, FILE *fp); diff --git a/tools/perf/util/map.c b/tools/perf/util/map.c index 3b7ce1bf9f8..679011c1b6d 100644 --- a/tools/perf/util/map.c +++ b/tools/perf/util/map.c @@ -32,8 +32,7 @@ void map__init(struct map *self, u64 start, u64 end, u64 pgoff, RB_CLEAR_NODE(&self->rb_node); } -struct map *map__new(struct mmap_event *event, char *cwd, int cwdlen, - unsigned int sym_priv_size) +struct map *map__new(struct mmap_event *event, char *cwd, int cwdlen) { struct map *self = malloc(sizeof(*self)); @@ -60,7 +59,7 @@ struct map *map__new(struct mmap_event *event, char *cwd, int cwdlen, filename = newfilename; } - dso = dsos__findnew(filename, sym_priv_size); + dso = dsos__findnew(filename); if (dso == NULL) goto out_delete; diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 13677b5dbe5..cf2c7f77886 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -26,6 +26,7 @@ static void dsos__add(struct dso *dso); static struct dso *dsos__find(const char *name); static struct map *map__new2(u64 start, struct dso *dso); static void kernel_maps__insert(struct map *map); +unsigned int symbol__priv_size; static struct rb_root kernel_maps; @@ -75,18 +76,17 @@ static void kernel_maps__fixup_end(void) } } -static struct symbol *symbol__new(u64 start, u64 len, const char *name, - unsigned int priv_size) +static struct symbol *symbol__new(u64 start, u64 len, const char *name) { size_t namelen = strlen(name) + 1; - struct symbol *self = calloc(1, priv_size + sizeof(*self) + namelen); - + struct symbol *self = calloc(1, (symbol__priv_size + + sizeof(*self) + namelen)); if (!self) return NULL; - if (priv_size) { - memset(self, 0, priv_size); - self = ((void *)self) + priv_size; + if (symbol__priv_size) { + memset(self, 0, symbol__priv_size); + self = ((void *)self) + symbol__priv_size; } self->start = start; self->end = len ? start + len - 1 : start; @@ -98,9 +98,9 @@ static struct symbol *symbol__new(u64 start, u64 len, const char *name, return self; } -static void symbol__delete(struct symbol *self, unsigned int priv_size) +static void symbol__delete(struct symbol *self) { - free(((void *)self) - priv_size); + free(((void *)self) - symbol__priv_size); } static size_t symbol__fprintf(struct symbol *self, FILE *fp) @@ -109,7 +109,7 @@ static size_t symbol__fprintf(struct symbol *self, FILE *fp) self->start, self->end, self->name); } -struct dso *dso__new(const char *name, unsigned int sym_priv_size) +struct dso *dso__new(const char *name) { struct dso *self = malloc(sizeof(*self) + strlen(name) + 1); @@ -118,7 +118,6 @@ struct dso *dso__new(const char *name, unsigned int sym_priv_size) self->long_name = self->name; self->short_name = self->name; self->syms = RB_ROOT; - self->sym_priv_size = sym_priv_size; self->find_symbol = dso__find_symbol; self->slen_calculated = 0; self->origin = DSO__ORIG_NOT_FOUND; @@ -136,7 +135,7 @@ static void dso__delete_symbols(struct dso *self) pos = rb_entry(next, struct symbol, rb_node); next = rb_next(&pos->rb_node); rb_erase(&pos->rb_node, &self->syms); - symbol__delete(pos, self->sym_priv_size); + symbol__delete(pos); } } @@ -250,8 +249,7 @@ static int kernel_maps__load_all_kallsyms(void) /* * Will fix up the end later, when we have all symbols sorted. */ - sym = symbol__new(start, 0, symbol_name, - kernel_map->dso->sym_priv_size); + sym = symbol__new(start, 0, symbol_name); if (sym == NULL) goto out_delete_line; @@ -317,8 +315,7 @@ static int kernel_maps__split_kallsyms(symbol_filter_t filter, int use_modules) snprintf(dso_name, sizeof(dso_name), "[kernel].%d", kernel_range++); - dso = dso__new(dso_name, - kernel_map->dso->sym_priv_size); + dso = dso__new(dso_name); if (dso == NULL) return -1; @@ -336,7 +333,7 @@ static int kernel_maps__split_kallsyms(symbol_filter_t filter, int use_modules) if (filter && filter(map, pos)) { delete_symbol: rb_erase(&pos->rb_node, &kernel_map->dso->syms); - symbol__delete(pos, kernel_map->dso->sym_priv_size); + symbol__delete(pos); } else { if (map != kernel_map) { rb_erase(&pos->rb_node, &kernel_map->dso->syms); @@ -417,14 +414,13 @@ static int dso__load_perf_map(struct dso *self, struct map *map, if (len + 2 >= line_len) continue; - sym = symbol__new(start, size, line + len, - self->sym_priv_size); + sym = symbol__new(start, size, line + len); if (sym == NULL) goto out_delete_line; if (filter && filter(map, sym)) - symbol__delete(sym, self->sym_priv_size); + symbol__delete(sym); else { dso__insert_symbol(self, sym); nr_syms++; @@ -616,7 +612,7 @@ static int dso__synthesize_plt_symbols(struct dso *self) "%s@plt", elf_sym__name(&sym, symstrs)); f = symbol__new(plt_offset, shdr_plt.sh_entsize, - sympltname, self->sym_priv_size); + sympltname); if (!f) goto out_elf_end; @@ -634,7 +630,7 @@ static int dso__synthesize_plt_symbols(struct dso *self) "%s@plt", elf_sym__name(&sym, symstrs)); f = symbol__new(plt_offset, shdr_plt.sh_entsize, - sympltname, self->sym_priv_size); + sympltname); if (!f) goto out_elf_end; @@ -769,7 +765,7 @@ static int dso__load_sym(struct dso *self, struct map *map, const char *name, if (kmodule) start += map->start + shdr.sh_offset; - curr_dso = dso__new(dso_name, self->sym_priv_size); + curr_dso = dso__new(dso_name); if (curr_dso == NULL) goto out_elf_end; curr_map = map__new2(start, curr_dso); @@ -803,14 +799,13 @@ static int dso__load_sym(struct dso *self, struct map *map, const char *name, if (demangled != NULL) elf_name = demangled; new_symbol: - f = symbol__new(sym.st_value, sym.st_size, elf_name, - curr_dso->sym_priv_size); + f = symbol__new(sym.st_value, sym.st_size, elf_name); free(demangled); if (!f) goto out_elf_end; if (filter && filter(curr_map, f)) - symbol__delete(f, curr_dso->sym_priv_size); + symbol__delete(f); else { dso__insert_symbol(curr_dso, f); nr++; @@ -1141,7 +1136,7 @@ static struct map *map__new2(u64 start, struct dso *dso) return self; } -static int dsos__load_modules(unsigned int sym_priv_size) +static int dsos__load_modules(void) { char *line = NULL; size_t n; @@ -1180,7 +1175,7 @@ static int dsos__load_modules(unsigned int sym_priv_size) *sep = '\0'; snprintf(name, sizeof(name), "[%s]", line); - dso = dso__new(name, sym_priv_size); + dso = dso__new(name); if (dso == NULL) goto out_delete_line; @@ -1224,11 +1219,11 @@ static int dso__load_vmlinux(struct dso *self, struct map *map, return err; } -int dsos__load_kernel(const char *vmlinux, unsigned int sym_priv_size, - symbol_filter_t filter, int use_modules) +int dsos__load_kernel(const char *vmlinux, symbol_filter_t filter, + int use_modules) { int err = -1; - struct dso *dso = dso__new(vmlinux, sym_priv_size); + struct dso *dso = dso__new(vmlinux); if (dso == NULL) return -1; @@ -1240,7 +1235,7 @@ int dsos__load_kernel(const char *vmlinux, unsigned int sym_priv_size, kernel_map->map_ip = kernel_map->unmap_ip = identity__map_ip; - if (use_modules && dsos__load_modules(sym_priv_size) < 0) { + if (use_modules && dsos__load_modules() < 0) { pr_warning("Failed to load list of modules in use! " "Continuing...\n"); use_modules = 0; @@ -1312,12 +1307,12 @@ static struct dso *dsos__find(const char *name) return NULL; } -struct dso *dsos__findnew(const char *name, unsigned int sym_priv_size) +struct dso *dsos__findnew(const char *name) { struct dso *dso = dsos__find(name); if (!dso) { - dso = dso__new(name, sym_priv_size); + dso = dso__new(name); if (dso != NULL) dsos__add(dso); } @@ -1333,13 +1328,12 @@ void dsos__fprintf(FILE *fp) dso__fprintf(pos, fp); } -int load_kernel(unsigned int sym_priv_size, symbol_filter_t filter) +int load_kernel(symbol_filter_t filter) { - if (dsos__load_kernel(vmlinux_name, sym_priv_size, filter, - modules) <= 0) + if (dsos__load_kernel(vmlinux_name, filter, modules) <= 0) return -1; - vdso = dso__new("[vdso]", 0); + vdso = dso__new("[vdso]"); if (!vdso) return -1; @@ -1348,7 +1342,8 @@ int load_kernel(unsigned int sym_priv_size, symbol_filter_t filter) return 0; } -void symbol__init(void) +void symbol__init(unsigned int priv_size) { elf_version(EV_CURRENT); + symbol__priv_size = priv_size; } diff --git a/tools/perf/util/symbol.h b/tools/perf/util/symbol.h index 432edbca780..a471a384073 100644 --- a/tools/perf/util/symbol.h +++ b/tools/perf/util/symbol.h @@ -39,11 +39,17 @@ struct symbol { char name[0]; }; +extern unsigned int symbol__priv_size; + +static inline void *symbol__priv(struct symbol *self) +{ + return ((void *)self) - symbol__priv_size; +} + struct dso { struct list_head node; struct rb_root syms; struct symbol *(*find_symbol)(struct dso *, u64 ip); - unsigned int sym_priv_size; unsigned char adjust_symbols; unsigned char slen_calculated; bool loaded; @@ -53,28 +59,22 @@ struct dso { char name[0]; }; -struct dso *dso__new(const char *name, unsigned int sym_priv_size); +struct dso *dso__new(const char *name); void dso__delete(struct dso *self); -static inline void *dso__sym_priv(struct dso *self, struct symbol *sym) -{ - return ((void *)sym) - self->sym_priv_size; -} - struct symbol *dso__find_symbol(struct dso *self, u64 ip); -int dsos__load_kernel(const char *vmlinux, unsigned int sym_priv_size, - symbol_filter_t filter, int modules); -struct dso *dsos__findnew(const char *name, unsigned int sym_priv_size); +int dsos__load_kernel(const char *vmlinux, symbol_filter_t filter, int modules); +struct dso *dsos__findnew(const char *name); int dso__load(struct dso *self, struct map *map, symbol_filter_t filter); void dsos__fprintf(FILE *fp); size_t dso__fprintf(struct dso *self, FILE *fp); char dso__symtab_origin(const struct dso *self); -int load_kernel(unsigned int sym_priv_size, symbol_filter_t filter); +int load_kernel(symbol_filter_t filter); -void symbol__init(void); +void symbol__init(unsigned int priv_size); extern struct list_head dsos; extern struct map *kernel_map; -- cgit v1.2.3-18-g5258 From d70a5402f9c2e2671b809363616b3508b4c5a565 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 30 Oct 2009 16:28:25 -0200 Subject: perf tools: Improve message about missing symtabs for deleted DSOs Instead of: no symbols found in /usr/lib/gstreamer-0.10/libgsttypefindfunctions.so (deleted), maybe install a debug package? no symbols found in /usr/lib/gstreamer-0.10/libgstaudioconvert.so (deleted), maybe install a debug package? We now emit: /usr/lib/gstreamer-0.10/libgsttypefindfunctions.so was updated, restart the long running apps that use it! /usr/lib/gstreamer-0.10/libgstaudioconvert.so was updated, restart the long running apps that use it! Which is far less misleading about what the cause of the symbol mismatch is. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: Peter Zijlstra Cc: Paul Mackerras Cc: Mike Galbraith LKML-Reference: <1256927305-4628-3-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/util/map.c | 16 ++++++++++++++-- 1 file changed, 14 insertions(+), 2 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/map.c b/tools/perf/util/map.c index 679011c1b6d..f1e21695542 100644 --- a/tools/perf/util/map.c +++ b/tools/perf/util/map.c @@ -75,6 +75,8 @@ out_delete: return NULL; } +#define DSO__DELETED "(deleted)" + struct symbol * map__find_symbol(struct map *self, u64 ip, symbol_filter_t filter) { @@ -86,8 +88,18 @@ map__find_symbol(struct map *self, u64 ip, symbol_filter_t filter) self->dso->long_name); return NULL; } else if (nr == 0) { - pr_warning("No symbols found in %s, maybe install a debug package?\n", - self->dso->long_name); + const char *name = self->dso->long_name; + const size_t len = strlen(name); + const size_t real_len = len - sizeof(DSO__DELETED); + + if (len > sizeof(DSO__DELETED) && + strcmp(name + real_len + 1, DSO__DELETED) == 0) + pr_warning("%.*s was updated, restart the " + "long running apps that use it!\n", + real_len, name); + else + pr_warning("no symbols found in %s, maybe " + "install a debug package?\n", name); return NULL; } } -- cgit v1.2.3-18-g5258 From 900b20d5900045fb9b48f2fb3d80cbdbae3f44c0 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Mon, 2 Nov 2009 19:25:25 +0100 Subject: perf tools: Fix missing symtabs printouts MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Fix: util/map.c: In function ‘map__find_symbol’: util/map.c:97: error: field precision should have type ‘int’, but argument 3 has type ‘size_t’ Also clean up some line wrap damage - we dont line-wrap printk messages. Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: Peter Zijlstra Cc: Paul Mackerras Cc: Mike Galbraith LKML-Reference: <1256927305-4628-3-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/util/map.c | 13 ++++++------- 1 file changed, 6 insertions(+), 7 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/map.c b/tools/perf/util/map.c index f1e21695542..33f868420d7 100644 --- a/tools/perf/util/map.c +++ b/tools/perf/util/map.c @@ -93,13 +93,12 @@ map__find_symbol(struct map *self, u64 ip, symbol_filter_t filter) const size_t real_len = len - sizeof(DSO__DELETED); if (len > sizeof(DSO__DELETED) && - strcmp(name + real_len + 1, DSO__DELETED) == 0) - pr_warning("%.*s was updated, restart the " - "long running apps that use it!\n", - real_len, name); - else - pr_warning("no symbols found in %s, maybe " - "install a debug package?\n", name); + strcmp(name + real_len + 1, DSO__DELETED) == 0) { + pr_warning("%.*s was updated, restart the long running apps that use it!\n", + (int)real_len, name); + } else { + pr_warning("no symbols found in %s, maybe install a debug package?\n", name); + } return NULL; } } -- cgit v1.2.3-18-g5258 From 12e4db4790b1bd2b7ec70eb2a1386c00fc683740 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Tue, 3 Nov 2009 11:29:07 -0200 Subject: perf probe: Annotate variable initialization MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Annotate away this false positive warning on older GCCs: cc1: warnings being treated as errors builtin-probe.c: In function ‘parse_probe_event’: builtin-probe.c:72: warning: ‘nc’ is used uninitialized in this function Signed-off-by: Arnaldo Carvalho de Melo Acked-by: Masami Hiramatsu LKML-Reference: <1257254947-16789-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/builtin-probe.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-probe.c b/tools/perf/builtin-probe.c index a99a366230a..81245238e34 100644 --- a/tools/perf/builtin-probe.c +++ b/tools/perf/builtin-probe.c @@ -69,7 +69,7 @@ static struct { static void parse_probe_point(char *arg, struct probe_point *pp) { char *ptr, *tmp; - char c, nc; + char c, nc = 0; /* * * perf probe SRC:LN -- cgit v1.2.3-18-g5258 From c1e530178540df26eb39f10a972d06f96302ceb4 Mon Sep 17 00:00:00 2001 From: Thiago Farina Date: Tue, 3 Nov 2009 08:28:45 -0500 Subject: perf: Clean up trivial style issues in builtin-help.c Pointed out by checkpatch. Signed-off-by: Thiago Farina Cc: a.p.zijlstra@chello.nl Cc: paulus@samba.org Cc: Valdis.Kletnieks@vt.edu LKML-Reference: <1257254925-5423-1-git-send-email-tfransosi@gmail.com> Signed-off-by: Ingo Molnar --- tools/perf/builtin-help.c | 12 +++++------- 1 file changed, 5 insertions(+), 7 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-help.c b/tools/perf/builtin-help.c index 4fb8734a796..768f9c82631 100644 --- a/tools/perf/builtin-help.c +++ b/tools/perf/builtin-help.c @@ -61,8 +61,7 @@ static const char *get_man_viewer_info(const char *name) { struct man_viewer_info_list *viewer; - for (viewer = man_viewer_info_list; viewer; viewer = viewer->next) - { + for (viewer = man_viewer_info_list; viewer; viewer = viewer->next) { if (!strcasecmp(name, viewer->name)) return viewer->info; } @@ -115,7 +114,7 @@ static int check_emacsclient_version(void) return 0; } -static void exec_woman_emacs(const char* path, const char *page) +static void exec_woman_emacs(const char *path, const char *page) { if (!check_emacsclient_version()) { /* This works only with emacsclient version >= 22. */ @@ -129,7 +128,7 @@ static void exec_woman_emacs(const char* path, const char *page) } } -static void exec_man_konqueror(const char* path, const char *page) +static void exec_man_konqueror(const char *path, const char *page) { const char *display = getenv("DISPLAY"); if (display && *display) { @@ -157,7 +156,7 @@ static void exec_man_konqueror(const char* path, const char *page) } } -static void exec_man_man(const char* path, const char *page) +static void exec_man_man(const char *path, const char *page) { if (!path) path = "man"; @@ -364,9 +363,8 @@ static void show_man_page(const char *perf_cmd) setup_man_path(); for (viewer = man_viewer_list; viewer; viewer = viewer->next) - { exec_viewer(viewer->name, page); /* will return when unable */ - } + if (fallback) exec_viewer(fallback, page); exec_viewer("man", page); -- cgit v1.2.3-18-g5258 From 6d7aa9d721c8c640066142fd9534afcdf68d7f9d Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Tue, 3 Nov 2009 15:52:18 -0200 Subject: perf symbols: Initialize dso->loaded Brown paper bag bug introduced in: 66bd8424cc05e800db384053bf7ab967e4658468 ("perf tools: Delay loading symtabs till we hit a map with it") Without this we were not loading any symtabs that happened to be on a DSO for which the allocated memory for ->loaded was !0. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: Peter Zijlstra Cc: Paul Mackerras Cc: Mike Galbraith LKML-Reference: <1257270738-5669-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/util/symbol.c | 1 + 1 file changed, 1 insertion(+) (limited to 'tools/perf') diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index cf2c7f77886..93c49f4685f 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -121,6 +121,7 @@ struct dso *dso__new(const char *name) self->find_symbol = dso__find_symbol; self->slen_calculated = 0; self->origin = DSO__ORIG_NOT_FOUND; + self->loaded = false; } return self; -- cgit v1.2.3-18-g5258 From 2643ce11457a99a85c5bed8dd631e35968e6ca5a Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Tue, 3 Nov 2009 21:46:10 -0200 Subject: perf symbols: Factor out buildid reading routine So that we can run it without having a DSO instance. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: Peter Zijlstra Cc: Paul Mackerras Cc: Mike Galbraith LKML-Reference: <1257291970-8208-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/util/symbol.c | 51 ++++++++++++++++++++++++++++++++---------------- tools/perf/util/symbol.h | 2 ++ 2 files changed, 36 insertions(+), 17 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index ac94d7b94f6..e7c7cdb851c 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -825,27 +825,27 @@ out_close: return err; } -#define BUILD_ID_SIZE 128 +#define BUILD_ID_SIZE 20 -static char *dso__read_build_id(struct dso *self) +int filename__read_build_id(const char *filename, void *bf, size_t size) { - int i; + int fd, err = -1; GElf_Ehdr ehdr; GElf_Shdr shdr; Elf_Data *build_id_data; Elf_Scn *sec; - char *build_id = NULL, *bid; - unsigned char *raw; Elf *elf; - int fd = open(self->long_name, O_RDONLY); + if (size < BUILD_ID_SIZE) + goto out; + + fd = open(filename, O_RDONLY); if (fd < 0) goto out; elf = elf_begin(fd, PERF_ELF_C_READ_MMAP, NULL); if (elf == NULL) { - pr_err("%s: cannot read %s ELF file.\n", __func__, - self->long_name); + pr_err("%s: cannot read %s ELF file.\n", __func__, filename); goto out_close; } @@ -854,29 +854,46 @@ static char *dso__read_build_id(struct dso *self) goto out_elf_end; } - sec = elf_section_by_name(elf, &ehdr, &shdr, ".note.gnu.build-id", NULL); + sec = elf_section_by_name(elf, &ehdr, &shdr, + ".note.gnu.build-id", NULL); if (sec == NULL) goto out_elf_end; build_id_data = elf_getdata(sec, NULL); if (build_id_data == NULL) goto out_elf_end; - build_id = malloc(BUILD_ID_SIZE); + memcpy(bf, build_id_data->d_buf + 16, BUILD_ID_SIZE); + err = BUILD_ID_SIZE; +out_elf_end: + elf_end(elf); +out_close: + close(fd); +out: + return err; +} + +static char *dso__read_build_id(struct dso *self) +{ + int i, len; + char *build_id = NULL, *bid; + unsigned char rawbf[BUILD_ID_SIZE], *raw; + + len = filename__read_build_id(self->long_name, rawbf, sizeof(rawbf)); + if (len < 0) + goto out; + + build_id = malloc(len * 2 + 1); if (build_id == NULL) - goto out_elf_end; - raw = build_id_data->d_buf + 16; + goto out; bid = build_id; - for (i = 0; i < 20; ++i) { + raw = rawbf; + for (i = 0; i < len; ++i) { sprintf(bid, "%02x", *raw); ++raw; bid += 2; } pr_debug2("%s(%s): %s\n", __func__, self->long_name, build_id); -out_elf_end: - elf_end(elf); -out_close: - close(fd); out: return build_id; } diff --git a/tools/perf/util/symbol.h b/tools/perf/util/symbol.h index 088433062dd..e0d4a583f8d 100644 --- a/tools/perf/util/symbol.h +++ b/tools/perf/util/symbol.h @@ -82,6 +82,8 @@ void dsos__fprintf(FILE *fp); size_t dso__fprintf(struct dso *self, FILE *fp); char dso__symtab_origin(const struct dso *self); +int filename__read_build_id(const char *filename, void *bf, size_t size); + int load_kernel(symbol_filter_t filter); void symbol__init(unsigned int priv_size); -- cgit v1.2.3-18-g5258 From c43f9d1e61e265c6bfafdd65c7f07c8d71a7efc3 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Tue, 3 Nov 2009 19:12:13 -0500 Subject: perf/probes: Update Documentation/perf-probe.txt Update Documentation/perf-probe.txt accoding to recent syntax changes. Signed-off-by: Masami Hiramatsu Acked-by: Frederic Weisbecker Cc: Steven Rostedt Cc: Jim Keniston Cc: Ananth N Mavinakayanahalli Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: Jason Baron Cc: K.Prasad Cc: Peter Zijlstra Cc: Srikar Dronamraju LKML-Reference: <20091104001212.3454.19415.stgit@harusame> Signed-off-by: Ingo Molnar --- tools/perf/Documentation/perf-probe.txt | 17 +++++++++-------- 1 file changed, 9 insertions(+), 8 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/Documentation/perf-probe.txt b/tools/perf/Documentation/perf-probe.txt index 6b6c6aecdf1..9270594e6df 100644 --- a/tools/perf/Documentation/perf-probe.txt +++ b/tools/perf/Documentation/perf-probe.txt @@ -8,7 +8,9 @@ perf-probe - Define new dynamic tracepoints SYNOPSIS -------- [verse] -'perf probe' [-k ] -P 'PROBE' [-P 'PROBE' ...] +'perf probe' [options] --add 'PROBE' [--add 'PROBE' ...] +or +'perf probe' [options] 'PROBE' ['PROBE' ...] DESCRIPTION @@ -21,26 +23,25 @@ and C local variables) with debuginfo. OPTIONS ------- -k:: ---vmlinux:: +--vmlinux=PATH:: Specify vmlinux path which has debuginfo (Dwarf binary). -v:: --verbose:: Be more verbose (show parsed arguments, etc). --P:: ---probe:: +-a:: +--add:: Define a probe point (see PROBE SYNTAX for detail) PROBE SYNTAX ------------ Probe points are defined by following syntax. - "TYPE:[GRP/]NAME FUNC[+OFFS][@SRC]|@SRC:LINE [ARG ...]" + "FUNC[+OFFS|:RLN|%return][@SRC]|SRC:ALN [ARG ...]" -'TYPE' specifies the type of probe point, you can use either "p" (kprobe) or "r" (kretprobe) for 'TYPE'. 'GRP' specifies the group name of this probe, and 'NAME' specifies the event name. If 'GRP' is omitted, "kprobes" is used for its group name. -'FUNC' and 'OFFS' specifies function and offset (in byte) where probe will be put. In addition, 'SRC' specifies a source file which has that function (this is mainly for inline functions). -You can specify a probe point by the source line number by using '@SRC:LINE' syntax, where 'SRC' is the source file path and 'LINE' is the line number. +'FUNC' specifies a probed function name, and it may have one of the following options; '+OFFS' is the offset from function entry address in bytes, 'RLN' is the relative-line number from function entry line, and '%return' means that it probes function return. In addition, 'SRC' specifies a source file which has that function. +It is also possible to specify a probe point by the source line number by using 'SRC:ALN' syntax, where 'SRC' is the source file path and 'ALN' is the line number. 'ARG' specifies the arguments of this probe point. You can use the name of local variable, or kprobe-tracer argument format (e.g. $retval, %ax, etc). SEE ALSO -- cgit v1.2.3-18-g5258 From a7f4328b91fb6e71dbe1fa4d46f3597c9555014d Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Tue, 3 Nov 2009 19:12:21 -0500 Subject: perf/probes: Improve error messages Improve error messages in perf-probe so that users can figure out problems easily. Reported-by: Ingo Molnar Signed-off-by: Masami Hiramatsu Acked-by: Frederic Weisbecker Cc: Steven Rostedt Cc: Jim Keniston Cc: Ananth N Mavinakayanahalli Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: Jason Baron Cc: K.Prasad Cc: Peter Zijlstra Cc: Srikar Dronamraju LKML-Reference: <20091104001221.3454.52030.stgit@harusame> Signed-off-by: Ingo Molnar --- tools/perf/builtin-probe.c | 20 +++++++++++++------- tools/perf/util/probe-finder.c | 2 +- 2 files changed, 14 insertions(+), 8 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-probe.c b/tools/perf/builtin-probe.c index 81245238e34..65bcaed0ef4 100644 --- a/tools/perf/builtin-probe.c +++ b/tools/perf/builtin-probe.c @@ -294,10 +294,11 @@ static int write_new_event(int fd, const char *buf) { int ret; - printf("Adding new event: %s\n", buf); ret = write(fd, buf, strlen(buf)); if (ret <= 0) - die("failed to create event."); + die("Failed to create event."); + else + printf("Added new event: %s\n", buf); return ret; } @@ -310,7 +311,7 @@ static int synthesize_probe_event(struct probe_point *pp) int i, len, ret; pp->probes[0] = buf = (char *)calloc(MAX_CMDLEN, sizeof(char)); if (!buf) - die("calloc"); + die("Failed to allocate memory by calloc."); ret = snprintf(buf, MAX_CMDLEN, "%s+%d", pp->function, pp->offset); if (ret <= 0 || ret >= MAX_CMDLEN) goto error; @@ -363,7 +364,7 @@ int cmd_probe(int argc, const char **argv, const char *prefix __used) if (ret == -E2BIG) semantic_error("probe point is too long."); else if (ret < 0) - die("snprintf"); + die("Failed to synthesize a probe point."); } #ifndef NO_LIBDWARF @@ -375,7 +376,7 @@ int cmd_probe(int argc, const char **argv, const char *prefix __used) else fd = open_default_vmlinux(); if (fd < 0) - die("vmlinux/module file open"); + die("Could not open vmlinux/module file."); /* Searching probe points */ for (j = 0; j < session.nr_probe; j++) { @@ -396,8 +397,13 @@ setup_probes: /* Settng up probe points */ snprintf(buf, MAX_CMDLEN, "%s/../kprobe_events", debugfs_path); fd = open(buf, O_WRONLY, O_APPEND); - if (fd < 0) - die("kprobe_events open"); + if (fd < 0) { + if (errno == ENOENT) + die("kprobe_events file does not exist - please rebuild with CONFIG_KPROBE_TRACER."); + else + die("Could not open kprobe_events file: %s", + strerror(errno)); + } for (j = 0; j < session.nr_probe; j++) { pp = &session.probes[j]; if (pp->found == 1) { diff --git a/tools/perf/util/probe-finder.c b/tools/perf/util/probe-finder.c index db96186e02a..35d5a69aaf9 100644 --- a/tools/perf/util/probe-finder.c +++ b/tools/perf/util/probe-finder.c @@ -688,7 +688,7 @@ int find_probepoint(int fd, struct probe_point *pp) ret = dwarf_init(fd, DW_DLC_READ, 0, 0, &__dw_debug, &__dw_error); if (ret != DW_DLV_OK) - die("Failed to call dwarf_init(). Maybe, not a dwarf file.\n"); + die("No dwarf info found in the vmlinux - please rebuild with CONFIG_DEBUG_INFO.\n"); pp->found = 0; while (++cu_number) { -- cgit v1.2.3-18-g5258 From a225a1d911f0e434dc0407df29fd08e4388f3fa4 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Tue, 3 Nov 2009 19:12:30 -0500 Subject: perf/probes: Fall back to non-dwarf if possible Fall back to non-dwarf probe point if the probe definition may not need dwarf analysis, when perf can't find vmlinux/debuginfo. This might skip some inlined code of target function. Signed-off-by: Masami Hiramatsu Acked-by: Frederic Weisbecker Cc: Steven Rostedt Cc: Jim Keniston Cc: Ananth N Mavinakayanahalli Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: Jason Baron Cc: K.Prasad Cc: Peter Zijlstra Cc: Srikar Dronamraju LKML-Reference: <20091104001229.3454.63987.stgit@harusame> Signed-off-by: Ingo Molnar --- tools/perf/builtin-probe.c | 64 ++++++++++++++++++++++++------------------ tools/perf/util/probe-finder.c | 6 ++-- 2 files changed, 40 insertions(+), 30 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-probe.c b/tools/perf/builtin-probe.c index 65bcaed0ef4..d111a93f220 100644 --- a/tools/perf/builtin-probe.c +++ b/tools/perf/builtin-probe.c @@ -189,7 +189,7 @@ static void parse_probe_event(const char *str) /* Parse probe point */ parse_probe_point(argv[0], pp); free(argv[0]); - if (pp->file) + if (pp->file || pp->line) session.need_dwarf = 1; /* Copy arguments */ @@ -347,36 +347,24 @@ int cmd_probe(int argc, const char **argv, const char *prefix __used) if (session.nr_probe == 0) usage_with_options(probe_usage, options); -#ifdef NO_LIBDWARF if (session.need_dwarf) - semantic_error("Dwarf-analysis is not supported"); -#endif - - /* Synthesize probes without dwarf */ - for (j = 0; j < session.nr_probe; j++) { -#ifndef NO_LIBDWARF - if (!session.probes[j].retprobe) { - session.need_dwarf = 1; - continue; - } -#endif - ret = synthesize_probe_event(&session.probes[j]); - if (ret == -E2BIG) - semantic_error("probe point is too long."); - else if (ret < 0) - die("Failed to synthesize a probe point."); - } - -#ifndef NO_LIBDWARF - if (!session.need_dwarf) - goto setup_probes; +#ifdef NO_LIBDWARF + semantic_error("Debuginfo-analysis is not supported"); +#else /* !NO_LIBDWARF */ + pr_info("Some probes require debuginfo.\n"); if (session.vmlinux) fd = open(session.vmlinux, O_RDONLY); else fd = open_default_vmlinux(); - if (fd < 0) - die("Could not open vmlinux/module file."); + if (fd < 0) { + if (session.need_dwarf) + die("Could not open vmlinux/module file."); + + pr_warning("Could not open vmlinux/module file." + " Try to use symbols.\n"); + goto end_dwarf; + } /* Searching probe points */ for (j = 0; j < session.nr_probe; j++) { @@ -386,14 +374,34 @@ int cmd_probe(int argc, const char **argv, const char *prefix __used) lseek(fd, SEEK_SET, 0); ret = find_probepoint(fd, pp); - if (ret <= 0) - die("No probe point found.\n"); + if (ret < 0) { + if (session.need_dwarf) + die("Could not analyze debuginfo."); + + pr_warning("An error occurred in debuginfo analysis. Try to use symbols.\n"); + break; + } + if (ret == 0) /* No error but failed to find probe point. */ + die("No probe point found."); } close(fd); -setup_probes: +end_dwarf: #endif /* !NO_LIBDWARF */ + /* Synthesize probes without dwarf */ + for (j = 0; j < session.nr_probe; j++) { + pp = &session.probes[j]; + if (pp->found) /* This probe is already found. */ + continue; + + ret = synthesize_probe_event(pp); + if (ret == -E2BIG) + semantic_error("probe point is too long."); + else if (ret < 0) + die("Failed to synthesize a probe point."); + } + /* Settng up probe points */ snprintf(buf, MAX_CMDLEN, "%s/../kprobe_events", debugfs_path); fd = open(buf, O_WRONLY, O_APPEND); diff --git a/tools/perf/util/probe-finder.c b/tools/perf/util/probe-finder.c index 35d5a69aaf9..293cdfc1b8c 100644 --- a/tools/perf/util/probe-finder.c +++ b/tools/perf/util/probe-finder.c @@ -687,8 +687,10 @@ int find_probepoint(int fd, struct probe_point *pp) struct probe_finder pf = {.pp = pp}; ret = dwarf_init(fd, DW_DLC_READ, 0, 0, &__dw_debug, &__dw_error); - if (ret != DW_DLV_OK) - die("No dwarf info found in the vmlinux - please rebuild with CONFIG_DEBUG_INFO.\n"); + if (ret != DW_DLV_OK) { + pr_warning("No dwarf info found in the vmlinux - please rebuild with CONFIG_DEBUG_INFO.\n"); + return -ENOENT; + } pp->found = 0; while (++cu_number) { -- cgit v1.2.3-18-g5258 From 91365bbe4f8c39a821f390f785d606304d6dee3c Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Tue, 3 Nov 2009 19:12:38 -0500 Subject: perf/probes: Rename perf probe events group name Rename the group name of perf probe events to 'probe'. Signed-off-by: Masami Hiramatsu Acked-by: Frederic Weisbecker Cc: Steven Rostedt Cc: Jim Keniston Cc: Ananth N Mavinakayanahalli Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: Jason Baron Cc: K.Prasad Cc: Peter Zijlstra Cc: Srikar Dronamraju LKML-Reference: <20091104001238.3454.70508.stgit@harusame> Signed-off-by: Ingo Molnar --- tools/perf/builtin-probe.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-probe.c b/tools/perf/builtin-probe.c index d111a93f220..d78a3d94549 100644 --- a/tools/perf/builtin-probe.c +++ b/tools/perf/builtin-probe.c @@ -52,7 +52,7 @@ const char *default_search_path[NR_SEARCH_PATH] = { #define MAX_PATH_LEN 256 #define MAX_PROBES 128 #define MAX_PROBE_ARGS 128 -#define PERFPROBE_GROUP "perfprobe" +#define PERFPROBE_GROUP "probe" /* Session management structure */ static struct { -- cgit v1.2.3-18-g5258 From c426bba069e65ea438880a04aa4e7c5b880e1728 Mon Sep 17 00:00:00 2001 From: Hitoshi Mitake Date: Thu, 5 Nov 2009 09:31:31 +0900 Subject: perf bench: Add new directory and header for new subcommand 'bench' This patch adds bench/ directory and bench/bench.h. bench/ directory will contain modules for bench subcommand. bench/bench.h is for listing prototypes of module functions. Signed-off-by: Hitoshi Mitake Cc: Rusty Russell Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Arnaldo Carvalho de Melo Cc: fweisbec@gmail.com Cc: Jiri Kosina LKML-Reference: <1257381097-4743-2-git-send-email-mitake@dcl.info.waseda.ac.jp> Signed-off-by: Ingo Molnar --- tools/perf/bench/bench.h | 9 +++++++++ 1 file changed, 9 insertions(+) create mode 100644 tools/perf/bench/bench.h (limited to 'tools/perf') diff --git a/tools/perf/bench/bench.h b/tools/perf/bench/bench.h new file mode 100644 index 00000000000..59adb279cd7 --- /dev/null +++ b/tools/perf/bench/bench.h @@ -0,0 +1,9 @@ +#ifndef BENCH_H +#define BENCH_H + +extern int bench_sched_messaging(int argc, const char **argv, + const char *prefix); +extern int bench_sched_pipe(int argc, const char **argv, + const char *prefix); + +#endif -- cgit v1.2.3-18-g5258 From e27454cc6352c4226ddc76f5e3a5dedd7dff456a Mon Sep 17 00:00:00 2001 From: Hitoshi Mitake Date: Thu, 5 Nov 2009 09:31:32 +0900 Subject: perf bench: Add sched-messaging.c: Benchmark for scheduler and IPC mechanisms based on hackbench This patch adds bench/sched-messaging.c. This benchmark measures performance of scheduler and IPC mechanisms, and is based on hackbench by Rusty Russell. Example of usage: % perf bench sched messaging -g 20 -l 1000 -s 5.432 # in sec % perf bench sched messaging # run with default options (20 sender and receiver processes per group) (10 groups == 400 processes run) Total time:0.308 sec % perf bench sched messaging -t -g 20 # # be multi-thread, with 20 groups (20 sender and receiver threads per group) (20 groups == 800 threads run) Total time:0.582 sec ( Rusty is the original author of hackbench.c and he said the code is and was under the GPLv2 so fine to be merged. ) Signed-off-by: Hitoshi Mitake Acked-by: Rusty Russell Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Arnaldo Carvalho de Melo Cc: fweisbec@gmail.com Cc: Jiri Kosina LKML-Reference: <1257381097-4743-3-git-send-email-mitake@dcl.info.waseda.ac.jp> Signed-off-by: Ingo Molnar --- tools/perf/bench/sched-messaging.c | 332 +++++++++++++++++++++++++++++++++++++ 1 file changed, 332 insertions(+) create mode 100644 tools/perf/bench/sched-messaging.c (limited to 'tools/perf') diff --git a/tools/perf/bench/sched-messaging.c b/tools/perf/bench/sched-messaging.c new file mode 100644 index 00000000000..36b62c549e2 --- /dev/null +++ b/tools/perf/bench/sched-messaging.c @@ -0,0 +1,332 @@ +/* + * + * builtin-bench-messaging.c + * + * messaging: Benchmark for scheduler and IPC mechanisms + * + * Based on hackbench by Rusty Russell + * Ported to perf by Hitoshi Mitake + * + */ + +#include "../perf.h" +#include "../util/util.h" +#include "../util/parse-options.h" +#include "../builtin.h" +#include "bench.h" + +/* Test groups of 20 processes spraying to 20 receivers */ +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#define DATASIZE 100 + +static int use_pipes = 0; +static unsigned int loops = 100; +static unsigned int thread_mode = 0; +static unsigned int num_groups = 10; +static int simple = 0; + +struct sender_context { + unsigned int num_fds; + int ready_out; + int wakefd; + int out_fds[0]; +}; + +struct receiver_context { + unsigned int num_packets; + int in_fds[2]; + int ready_out; + int wakefd; +}; + +static void barf(const char *msg) +{ + fprintf(stderr, "%s (error: %s)\n", msg, strerror(errno)); + exit(1); +} + +static void fdpair(int fds[2]) +{ + if (use_pipes) { + if (pipe(fds) == 0) + return; + } else { + if (socketpair(AF_UNIX, SOCK_STREAM, 0, fds) == 0) + return; + } + + barf(use_pipes ? "pipe()" : "socketpair()"); +} + +/* Block until we're ready to go */ +static void ready(int ready_out, int wakefd) +{ + char dummy; + struct pollfd pollfd = { .fd = wakefd, .events = POLLIN }; + + /* Tell them we're ready. */ + if (write(ready_out, &dummy, 1) != 1) + barf("CLIENT: ready write"); + + /* Wait for "GO" signal */ + if (poll(&pollfd, 1, -1) != 1) + barf("poll"); +} + +/* Sender sprays loops messages down each file descriptor */ +static void *sender(struct sender_context *ctx) +{ + char data[DATASIZE]; + unsigned int i, j; + + ready(ctx->ready_out, ctx->wakefd); + + /* Now pump to every receiver. */ + for (i = 0; i < loops; i++) { + for (j = 0; j < ctx->num_fds; j++) { + int ret, done = 0; + +again: + ret = write(ctx->out_fds[j], data + done, + sizeof(data)-done); + if (ret < 0) + barf("SENDER: write"); + done += ret; + if (done < DATASIZE) + goto again; + } + } + + return NULL; +} + + +/* One receiver per fd */ +static void *receiver(struct receiver_context* ctx) +{ + unsigned int i; + + if (!thread_mode) + close(ctx->in_fds[1]); + + /* Wait for start... */ + ready(ctx->ready_out, ctx->wakefd); + + /* Receive them all */ + for (i = 0; i < ctx->num_packets; i++) { + char data[DATASIZE]; + int ret, done = 0; + +again: + ret = read(ctx->in_fds[0], data + done, DATASIZE - done); + if (ret < 0) + barf("SERVER: read"); + done += ret; + if (done < DATASIZE) + goto again; + } + + return NULL; +} + +static pthread_t create_worker(void *ctx, void *(*func)(void *)) +{ + pthread_attr_t attr; + pthread_t childid; + int err; + + if (!thread_mode) { + /* process mode */ + /* Fork the receiver. */ + switch (fork()) { + case -1: + barf("fork()"); + break; + case 0: + (*func) (ctx); + exit(0); + break; + default: + break; + } + + return (pthread_t)0; + } + + if (pthread_attr_init(&attr) != 0) + barf("pthread_attr_init:"); + +#ifndef __ia64__ + if (pthread_attr_setstacksize(&attr, PTHREAD_STACK_MIN) != 0) + barf("pthread_attr_setstacksize"); +#endif + + err = pthread_create(&childid, &attr, func, ctx); + if (err != 0) { + fprintf(stderr, "pthread_create failed: %s (%d)\n", + strerror(err), err); + exit(-1); + } + return childid; +} + +static void reap_worker(pthread_t id) +{ + int proc_status; + void *thread_status; + + if (!thread_mode) { + /* process mode */ + wait(&proc_status); + if (!WIFEXITED(proc_status)) + exit(1); + } else { + pthread_join(id, &thread_status); + } +} + +/* One group of senders and receivers */ +static unsigned int group(pthread_t *pth, + unsigned int num_fds, + int ready_out, + int wakefd) +{ + unsigned int i; + struct sender_context *snd_ctx = malloc(sizeof(struct sender_context) + + num_fds * sizeof(int)); + + if (!snd_ctx) + barf("malloc()"); + + for (i = 0; i < num_fds; i++) { + int fds[2]; + struct receiver_context *ctx = malloc(sizeof(*ctx)); + + if (!ctx) + barf("malloc()"); + + + /* Create the pipe between client and server */ + fdpair(fds); + + ctx->num_packets = num_fds * loops; + ctx->in_fds[0] = fds[0]; + ctx->in_fds[1] = fds[1]; + ctx->ready_out = ready_out; + ctx->wakefd = wakefd; + + pth[i] = create_worker(ctx, (void *)receiver); + + snd_ctx->out_fds[i] = fds[1]; + if (!thread_mode) + close(fds[0]); + } + + /* Now we have all the fds, fork the senders */ + for (i = 0; i < num_fds; i++) { + snd_ctx->ready_out = ready_out; + snd_ctx->wakefd = wakefd; + snd_ctx->num_fds = num_fds; + + pth[num_fds+i] = create_worker(snd_ctx, (void *)sender); + } + + /* Close the fds we have left */ + if (!thread_mode) + for (i = 0; i < num_fds; i++) + close(snd_ctx->out_fds[i]); + + /* Return number of children to reap */ + return num_fds * 2; +} + +static const struct option options[] = { + OPT_BOOLEAN('p', "pipe", &use_pipes, + "Use pipe() instead of socketpair()"), + OPT_BOOLEAN('t', "thread", &thread_mode, + "Be multi thread instead of multi process"), + OPT_INTEGER('g', "group", &num_groups, + "Specify number of groups"), + OPT_INTEGER('l', "loop", &loops, + "Specify number of loops"), + OPT_BOOLEAN('s', "simple-output", &simple, + "Do simple output (this maybe useful for" + "processing by scripts or graph tools like gnuplot)"), + OPT_END() +}; + +static const char * const bench_sched_message_usage[] = { + "perf bench sched messaging ", + NULL +}; + +int bench_sched_messaging(int argc, const char **argv, + const char *prefix __used) +{ + unsigned int i, total_children; + struct timeval start, stop, diff; + unsigned int num_fds = 20; + int readyfds[2], wakefds[2]; + char dummy; + pthread_t *pth_tab; + + argc = parse_options(argc, argv, options, + bench_sched_message_usage, 0); + + pth_tab = malloc(num_fds * 2 * num_groups * sizeof(pthread_t)); + if (!pth_tab) + barf("main:malloc()"); + + fdpair(readyfds); + fdpair(wakefds); + + total_children = 0; + for (i = 0; i < num_groups; i++) + total_children += group(pth_tab+total_children, num_fds, + readyfds[1], wakefds[0]); + + /* Wait for everyone to be ready */ + for (i = 0; i < total_children; i++) + if (read(readyfds[0], &dummy, 1) != 1) + barf("Reading for readyfds"); + + gettimeofday(&start, NULL); + + /* Kick them off */ + if (write(wakefds[1], &dummy, 1) != 1) + barf("Writing to start them"); + + /* Reap them all */ + for (i = 0; i < total_children; i++) + reap_worker(pth_tab[i]); + + gettimeofday(&stop, NULL); + + timersub(&stop, &start, &diff); + + if (simple) + printf("%lu.%03lu\n", diff.tv_sec, diff.tv_usec/1000); + else { + printf("(%d sender and receiver %s per group)\n", + num_fds, thread_mode ? "threads" : "processes"); + printf("(%d groups == %d %s run)\n\n", + num_groups, num_groups * 2 * num_fds, + thread_mode ? "threads" : "processes"); + printf("\tTotal time:%lu.%03lu sec\n", + diff.tv_sec, diff.tv_usec/1000); + } + + return 0; +} -- cgit v1.2.3-18-g5258 From c7d9300f367f480aee4663a0e3695c5b48859a1a Mon Sep 17 00:00:00 2001 From: Hitoshi Mitake Date: Thu, 5 Nov 2009 09:31:33 +0900 Subject: perf bench: Add sched-pipe.c: Benchmark for pipe() system call This patch adds bench/sched-pipe.c. bench/sched-pipe.c is a benchmark program to measure performance of pipe() system call. This benchmark is based on pipe-test-1m.c by Ingo Molnar: http://people.redhat.com/mingo/cfs-scheduler/tools/pipe-test-1m.c Example of use: % perf bench sched pipe (executing 1000000 pipe operations between two tasks) Total time:4.499 sec 4.499179 usecs/op 222262 ops/sec % perf bench sched pipe -s -l 1000 0.015 Signed-off-by: Hitoshi Mitake Cc: Rusty Russell Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Arnaldo Carvalho de Melo Cc: fweisbec@gmail.com Cc: Jiri Kosina LKML-Reference: <1257381097-4743-4-git-send-email-mitake@dcl.info.waseda.ac.jp> Signed-off-by: Ingo Molnar --- tools/perf/bench/sched-pipe.c | 113 ++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 113 insertions(+) create mode 100644 tools/perf/bench/sched-pipe.c (limited to 'tools/perf') diff --git a/tools/perf/bench/sched-pipe.c b/tools/perf/bench/sched-pipe.c new file mode 100644 index 00000000000..3214ed20b1a --- /dev/null +++ b/tools/perf/bench/sched-pipe.c @@ -0,0 +1,113 @@ +/* + * + * builtin-bench-pipe.c + * + * pipe: Benchmark for pipe() + * + * Based on pipe-test-1m.c by Ingo Molnar + * http://people.redhat.com/mingo/cfs-scheduler/tools/pipe-test-1m.c + * Ported to perf by Hitoshi Mitake + * + */ + +#include "../perf.h" +#include "../util/util.h" +#include "../util/parse-options.h" +#include "../builtin.h" +#include "bench.h" + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#define LOOPS_DEFAULT 1000000 +static int loops = LOOPS_DEFAULT; +static int simple = 0; + +static const struct option options[] = { + OPT_INTEGER('l', "loop", &loops, + "Specify number of loops"), + OPT_BOOLEAN('s', "simple-output", &simple, + "Do simple output (this maybe useful for" + "processing by scripts or graph tools like gnuplot)"), + OPT_END() +}; + +static const char * const bench_sched_pipe_usage[] = { + "perf bench sched pipe ", + NULL +}; + +int bench_sched_pipe(int argc, const char **argv, + const char *prefix __used) +{ + int pipe_1[2], pipe_2[2]; + int m = 0, i; + struct timeval start, stop, diff; + unsigned long long result_usec = 0; + + /* + * why does "ret" exist? + * discarding returned value of read(), write() + * causes error in building environment for perf + */ + int ret; + pid_t pid; + + argc = parse_options(argc, argv, options, + bench_sched_pipe_usage, 0); + + assert(!pipe(pipe_1)); + assert(!pipe(pipe_2)); + + pid = fork(); + assert(pid >= 0); + + gettimeofday(&start, NULL); + + if (!pid) { + for (i = 0; i < loops; i++) { + ret = read(pipe_1[0], &m, sizeof(int)); + ret = write(pipe_2[1], &m, sizeof(int)); + } + } else { + for (i = 0; i < loops; i++) { + ret = write(pipe_1[1], &m, sizeof(int)); + ret = read(pipe_2[0], &m, sizeof(int)); + } + } + + gettimeofday(&stop, NULL); + timersub(&stop, &start, &diff); + + if (pid) + return 0; + + if (simple) + printf("%lu.%03lu\n", + diff.tv_sec, diff.tv_usec / 1000); + else { + printf("(executing %d pipe operations between two tasks)\n\n", + loops); + + result_usec = diff.tv_sec * 1000000; + result_usec += diff.tv_usec; + + printf("\tTotal time:%lu.%03lu sec\n", + diff.tv_sec, diff.tv_usec / 1000); + printf("\t\t%lf usecs/op\n", + (double)result_usec / (double)loops); + printf("\t\t%d ops/sec\n", + (int)((double)loops / + ((double)result_usec / (double)1000000))); + } + + return 0; +} -- cgit v1.2.3-18-g5258 From 629cc356653719c206a05f4dee5c5e242edb6546 Mon Sep 17 00:00:00 2001 From: Hitoshi Mitake Date: Thu, 5 Nov 2009 09:31:34 +0900 Subject: perf bench: Add builtin-bench.c: General framework for benchmark suites This patch adds builtin-bench.c builtin-bench.c is a general framework for benchmark suites. Signed-off-by: Hitoshi Mitake Cc: Rusty Russell Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Arnaldo Carvalho de Melo Cc: fweisbec@gmail.com Cc: Jiri Kosina LKML-Reference: <1257381097-4743-5-git-send-email-mitake@dcl.info.waseda.ac.jp> Signed-off-by: Ingo Molnar --- tools/perf/builtin-bench.c | 128 +++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 128 insertions(+) create mode 100644 tools/perf/builtin-bench.c (limited to 'tools/perf') diff --git a/tools/perf/builtin-bench.c b/tools/perf/builtin-bench.c new file mode 100644 index 00000000000..31f41643b0c --- /dev/null +++ b/tools/perf/builtin-bench.c @@ -0,0 +1,128 @@ +/* + * + * builtin-bench.c + * + * General benchmarking subsystem provided by perf + * + * Copyright (C) 2009, Hitoshi Mitake + * + */ + +/* + * + * Available subsystem list: + * sched ... scheduler and IPC mechanism + * + */ + +#include "perf.h" +#include "util/util.h" +#include "util/parse-options.h" +#include "builtin.h" +#include "bench/bench.h" + +#include +#include +#include + +struct bench_suite { + const char *name; + const char *summary; + int (*fn)(int, const char **, const char *); +}; + +static struct bench_suite sched_suites[] = { + { "messaging", + "Benchmark for scheduler and IPC mechanisms", + bench_sched_messaging }, + { "pipe", + "Flood of communication over pipe() between two processes", + bench_sched_pipe }, + { NULL, + NULL, + NULL } +}; + +struct bench_subsys { + const char *name; + const char *summary; + struct bench_suite *suites; +}; + +static struct bench_subsys subsystems[] = { + { "sched", + "scheduler and IPC mechanism", + sched_suites }, + { NULL, + NULL, + NULL } +}; + +static void dump_suites(int subsys_index) +{ + int i; + + printf("List of available suites for %s...\n\n", + subsystems[subsys_index].name); + + for (i = 0; subsystems[subsys_index].suites[i].name; i++) + printf("\t%s: %s\n", + subsystems[subsys_index].suites[i].name, + subsystems[subsys_index].suites[i].summary); + + printf("\n"); + return; +} + +int cmd_bench(int argc, const char **argv, const char *prefix __used) +{ + int i, j, status = 0; + + if (argc < 2) { + /* No subsystem specified. */ + printf("Usage: perf bench []\n\n"); + printf("List of available subsystems...\n\n"); + + for (i = 0; subsystems[i].name; i++) + printf("\t%s: %s\n", + subsystems[i].name, subsystems[i].summary); + printf("\n"); + + goto end; + } + + for (i = 0; subsystems[i].name; i++) { + if (strcmp(subsystems[i].name, argv[1])) + continue; + + if (argc < 3) { + /* No suite specified. */ + dump_suites(i); + goto end; + } + + for (j = 0; subsystems[i].suites[j].name; j++) { + if (strcmp(subsystems[i].suites[j].name, argv[2])) + continue; + + status = subsystems[i].suites[j].fn(argc - 2, + argv + 2, prefix); + goto end; + } + + if (!strcmp(argv[2], "-h") || !strcmp(argv[2], "--help")) { + dump_suites(i); + goto end; + } + + printf("Unknown suite:%s for %s\n", argv[2], argv[1]); + status = 1; + goto end; + } + + printf("Unknown subsystem:%s\n", argv[1]); + status = 1; + +end: + return status; +} -- cgit v1.2.3-18-g5258 From 11bd341c043348ecb7462d3bd8e1ad6d00f6892a Mon Sep 17 00:00:00 2001 From: Hitoshi Mitake Date: Thu, 5 Nov 2009 09:31:35 +0900 Subject: perf bench: Modify builtin.h for new prototype This patch modifies builtin.h to add prototype of cmd_bench(). Signed-off-by: Hitoshi Mitake Cc: Rusty Russell Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Arnaldo Carvalho de Melo Cc: fweisbec@gmail.com Cc: Jiri Kosina LKML-Reference: <1257381097-4743-6-git-send-email-mitake@dcl.info.waseda.ac.jp> Signed-off-by: Ingo Molnar --- tools/perf/builtin.h | 1 + 1 file changed, 1 insertion(+) (limited to 'tools/perf') diff --git a/tools/perf/builtin.h b/tools/perf/builtin.h index e11d8d231c3..f0cd5b139b7 100644 --- a/tools/perf/builtin.h +++ b/tools/perf/builtin.h @@ -15,6 +15,7 @@ extern int read_line_with_nul(char *buf, int size, FILE *file); extern int check_pager_config(const char *cmd); extern int cmd_annotate(int argc, const char **argv, const char *prefix); +extern int cmd_bench(int argc, const char **argv, const char *prefix); extern int cmd_help(int argc, const char **argv, const char *prefix); extern int cmd_sched(int argc, const char **argv, const char *prefix); extern int cmd_list(int argc, const char **argv, const char *prefix); -- cgit v1.2.3-18-g5258 From dcba8848d3bc83ec9ee0858b9ae6e4f1c1fa7fa3 Mon Sep 17 00:00:00 2001 From: Hitoshi Mitake Date: Thu, 5 Nov 2009 09:31:36 +0900 Subject: perf bench: Add new subcommand 'bench' to perf.c This patch modifies perf.c for invoking 'bench' subcommand. Signed-off-by: Hitoshi Mitake Cc: Rusty Russell Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Arnaldo Carvalho de Melo Cc: fweisbec@gmail.com Cc: Jiri Kosina LKML-Reference: <1257381097-4743-7-git-send-email-mitake@dcl.info.waseda.ac.jp> Signed-off-by: Ingo Molnar --- tools/perf/perf.c | 1 + 1 file changed, 1 insertion(+) (limited to 'tools/perf') diff --git a/tools/perf/perf.c b/tools/perf/perf.c index 624e62d9d1e..f90ca5ec83f 100644 --- a/tools/perf/perf.c +++ b/tools/perf/perf.c @@ -288,6 +288,7 @@ static void handle_internal_command(int argc, const char **argv) { "list", cmd_list, 0 }, { "record", cmd_record, 0 }, { "report", cmd_report, 0 }, + { "bench", cmd_bench, 0 }, { "stat", cmd_stat, 0 }, { "timechart", cmd_timechart, 0 }, { "top", cmd_top, 0 }, -- cgit v1.2.3-18-g5258 From bfde82ef51e3ea6ab8634d0fdbf5adcdd1b429cb Mon Sep 17 00:00:00 2001 From: Hitoshi Mitake Date: Thu, 5 Nov 2009 09:31:37 +0900 Subject: perf bench: Add subcommand 'bench' to the Makefile This patch modifies Makefile for new files related to 'bench' subcommand. The new code is active from this point on. Signed-off-by: Hitoshi Mitake Cc: Rusty Russell Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Arnaldo Carvalho de Melo Cc: fweisbec@gmail.com Cc: Jiri Kosina LKML-Reference: <1257381097-4743-8-git-send-email-mitake@dcl.info.waseda.ac.jp> Signed-off-by: Ingo Molnar --- tools/perf/Makefile | 7 +++++++ 1 file changed, 7 insertions(+) (limited to 'tools/perf') diff --git a/tools/perf/Makefile b/tools/perf/Makefile index 542b29e2e38..0a2542844a6 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -416,6 +416,13 @@ LIB_OBJS += util/hist.o LIB_OBJS += util/data_map.o BUILTIN_OBJS += builtin-annotate.o + +BUILTIN_OBJS += builtin-bench.o + +# Benchmark modules +BUILTIN_OBJS += bench/sched-messaging.o +BUILTIN_OBJS += bench/sched-pipe.o + BUILTIN_OBJS += builtin-help.o BUILTIN_OBJS += builtin-sched.o BUILTIN_OBJS += builtin-list.o -- cgit v1.2.3-18-g5258 From 8d06367fa79c053a4a56a2ce0bb9e840f5da1236 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Wed, 4 Nov 2009 18:50:43 -0200 Subject: perf symbols: Use the buildids if present With this change 'perf record' will intercept PERF_RECORD_MMAP calls, creating a linked list of DSOs, then when the session finishes, it will traverse this list and read the buildids, stashing them at the end of the file and will set up a new feature bit in the header bitmask. 'perf report' will then notice this feature and populate the 'dsos' list and set the build ids. When reading the symtabs it will refuse to load from a file that doesn't have the same build id. This improves the reliability of the profiler output, as symbols and profiling data is more guaranteed to match. Example: [root@doppio ~]# perf report | head /home/acme/bin/perf with build id b1ea544ac3746e7538972548a09aadecc5753868 not found, continuing without symbols # Samples: 2621434559 # # Overhead Command Shared Object Symbol # ........ ............... ............................. ...... # 7.91% init [kernel] [k] read_hpet 7.64% init [kernel] [k] mwait_idle_with_hints 7.60% swapper [kernel] [k] read_hpet 7.60% swapper [kernel] [k] mwait_idle_with_hints 3.65% init [kernel] [k] 0xffffffffa02339d9 [root@doppio ~]# In this case the 'perf' binary was an older one, vanished, so its symbols probably wouldn't match or would cause subtly different (and misleading) output. Next patches will support the kernel as well, reading the build id notes for it and the modules from /sys. Another patch should also introduce a new plumbing command: 'perf list-buildids' that will then be used in porcelain that is distro specific to fetch -debuginfo packages where such buildids are present. This will in turn allow for one to run 'perf record' in one machine and 'perf report' in another. Future work on having the buildid sent directly from the kernel in the PERF_RECORD_MMAP event is needed to close races, as the DSO can be changed during a 'perf record' session, but this patch at least helps with non-corner cases and current/older kernels. Signed-off-by: Arnaldo Carvalho de Melo Cc: Ananth N Mavinakayanahalli Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: Frederic Weisbecker Cc: Jason Baron Cc: Jim Keniston Cc: K. Prasad Cc: Masami Hiramatsu Cc: Peter Zijlstra Cc: Roland McGrath Cc: Srikar Dronamraju Cc: Steven Rostedt LKML-Reference: <1257367843-26224-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/builtin-record.c | 49 ++++++++++++++++++++++++++-- tools/perf/util/data_map.c | 37 +++++++++++++++++++++ tools/perf/util/event.h | 7 ++++ tools/perf/util/header.c | 10 ++++++ tools/perf/util/header.h | 4 +++ tools/perf/util/map.c | 14 ++++++-- tools/perf/util/symbol.c | 78 ++++++++++++++++++++++++++++++++------------- tools/perf/util/symbol.h | 10 ++++-- 8 files changed, 179 insertions(+), 30 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 4a73d89ce5d..ab333812ace 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -17,6 +17,7 @@ #include "util/header.h" #include "util/event.h" #include "util/debug.h" +#include "util/symbol.h" #include #include @@ -109,9 +110,21 @@ static void write_output(void *buf, size_t size) } } +static void write_event(event_t *buf, size_t size) +{ + /* + * Add it to the list of DSOs, so that when we finish this + * record session we can pick the available build-ids. + */ + if (buf->header.type == PERF_RECORD_MMAP) + dsos__findnew(buf->mmap.filename); + + write_output(buf, size); +} + static int process_synthesized_event(event_t *event) { - write_output(event, event->header.size); + write_event(event, event->header.size); return 0; } @@ -163,14 +176,14 @@ static void mmap_read(struct mmap_data *md) size = md->mask + 1 - (old & md->mask); old += size; - write_output(buf, size); + write_event(buf, size); } buf = &data[old & md->mask]; size = head - old; old += size; - write_output(buf, size); + write_event(buf, size); md->prev = old; mmap_write_tail(md, old); @@ -365,10 +378,38 @@ static void open_counters(int cpu, pid_t pid) nr_cpu++; } +static bool write_buildid_table(void) +{ + struct dso *pos; + bool have_buildid = false; + + list_for_each_entry(pos, &dsos, node) { + struct build_id_event b; + size_t len; + + if (filename__read_build_id(pos->long_name, + &b.build_id, + sizeof(b.build_id)) < 0) + continue; + have_buildid = true; + memset(&b.header, 0, sizeof(b.header)); + len = strlen(pos->long_name) + 1; + len = ALIGN(len, 64); + b.header.size = sizeof(b) + len; + write_output(&b, sizeof(b)); + write_output(pos->long_name, len); + } + + return have_buildid; +} + static void atexit_header(void) { header->data_size += bytes_written; + if (write_buildid_table()) + perf_header__set_feat(header, HEADER_BUILD_ID); + perf_header__write(header, output); } @@ -572,6 +613,8 @@ int cmd_record(int argc, const char **argv, const char *prefix __used) { int counter; + symbol__init(0); + argc = parse_options(argc, argv, options, record_usage, PARSE_OPT_STOP_AT_NON_OPTION); if (!argc && target_pid == -1 && !system_wide) diff --git a/tools/perf/util/data_map.c b/tools/perf/util/data_map.c index c458db9ede6..00a9c114c8d 100644 --- a/tools/perf/util/data_map.c +++ b/tools/perf/util/data_map.c @@ -70,6 +70,39 @@ process_event(event_t *event, unsigned long offset, unsigned long head) } } +static int perf_header__read_build_ids(const struct perf_header *self, + int input, off_t file_size) +{ + off_t offset = self->data_offset + self->data_size; + struct build_id_event bev; + char filename[PATH_MAX]; + int err = -1; + + if (lseek(input, offset, SEEK_SET) < 0) + return -1; + + while (offset < file_size) { + struct dso *dso; + ssize_t len; + + if (read(input, &bev, sizeof(bev)) != sizeof(bev)) + goto out; + + len = bev.header.size - sizeof(bev); + if (read(input, filename, len) != len) + goto out; + + dso = dsos__findnew(filename); + if (dso != NULL) + dso__set_build_id(dso, &bev.build_id); + + offset += bev.header.size; + } + err = 0; +out: + return err; +} + int mmap_dispatch_perf_file(struct perf_header **pheader, const char *input_name, int force, @@ -130,6 +163,10 @@ int mmap_dispatch_perf_file(struct perf_header **pheader, if (curr_handler->sample_type_check(sample_type) < 0) exit(-1); + if (perf_header__has_feat(header, HEADER_BUILD_ID) && + perf_header__read_build_ids(header, input, input_stat.st_size)) + pr_debug("failed to read buildids, continuing...\n"); + if (load_kernel(NULL) < 0) { perror("failed to load kernel symbols"); return EXIT_FAILURE; diff --git a/tools/perf/util/event.h b/tools/perf/util/event.h index 0a443bea68d..34c6fcb82d9 100644 --- a/tools/perf/util/event.h +++ b/tools/perf/util/event.h @@ -61,6 +61,13 @@ struct sample_event{ u64 array[]; }; +#define BUILD_ID_SIZE 20 + +struct build_id_event { + struct perf_event_header header; + u8 build_id[ALIGN(BUILD_ID_SIZE, sizeof(u64))]; + char filename[]; +}; typedef union event_union { struct perf_event_header header; diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index 7d26659b806..050f543fd96 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -149,6 +149,16 @@ void perf_header__feat_trace_info(struct perf_header *header) set_bit(HEADER_TRACE_INFO, header->adds_features); } +void perf_header__set_feat(struct perf_header *self, int feat) +{ + set_bit(feat, self->adds_features); +} + +bool perf_header__has_feat(const struct perf_header *self, int feat) +{ + return test_bit(feat, self->adds_features); +} + static void do_write(int fd, void *buf, size_t size) { while (size) { diff --git a/tools/perf/util/header.h b/tools/perf/util/header.h index 2ea9dfb1236..2f233c5db7e 100644 --- a/tools/perf/util/header.h +++ b/tools/perf/util/header.h @@ -3,6 +3,7 @@ #include "../../../include/linux/perf_event.h" #include +#include #include "types.h" #include @@ -15,6 +16,7 @@ struct perf_header_attr { }; #define HEADER_TRACE_INFO 1 +#define HEADER_BUILD_ID 2 #define HEADER_FEAT_BITS 256 @@ -48,6 +50,8 @@ u64 perf_header__sample_type(struct perf_header *header); struct perf_event_attr * perf_header__find_attr(u64 id, struct perf_header *header); void perf_header__feat_trace_info(struct perf_header *header); +void perf_header__set_feat(struct perf_header *self, int feat); +bool perf_header__has_feat(const struct perf_header *self, int feat); struct perf_header *perf_header__new(void); diff --git a/tools/perf/util/map.c b/tools/perf/util/map.c index 33f868420d7..94ca95073c4 100644 --- a/tools/perf/util/map.c +++ b/tools/perf/util/map.c @@ -84,8 +84,18 @@ map__find_symbol(struct map *self, u64 ip, symbol_filter_t filter) int nr = dso__load(self->dso, self, filter); if (nr < 0) { - pr_warning("Failed to open %s, continuing without symbols\n", - self->dso->long_name); + if (self->dso->has_build_id) { + char sbuild_id[BUILD_ID_SIZE * 2 + 1]; + + build_id__sprintf(self->dso->build_id, + sizeof(self->dso->build_id), + sbuild_id); + pr_warning("%s with build id %s not found", + self->dso->long_name, sbuild_id); + } else + pr_warning("Failed to open %s", + self->dso->long_name); + pr_warning(", continuing without symbols\n"); return NULL; } else if (nr == 0) { const char *name = self->dso->long_name; diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index e7c7cdb851c..a2e95ce1f22 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -121,7 +121,8 @@ struct dso *dso__new(const char *name) self->find_symbol = dso__find_symbol; self->slen_calculated = 0; self->origin = DSO__ORIG_NOT_FOUND; - self->loaded = false; + self->loaded = 0; + self->has_build_id = 0; } return self; @@ -148,6 +149,12 @@ void dso__delete(struct dso *self) free(self); } +void dso__set_build_id(struct dso *self, void *build_id) +{ + memcpy(self->build_id, build_id, sizeof(self->build_id)); + self->has_build_id = 1; +} + static void dso__insert_symbol(struct dso *self, struct symbol *sym) { struct rb_node **p = &self->syms.rb_node; @@ -190,11 +197,30 @@ struct symbol *dso__find_symbol(struct dso *self, u64 ip) return NULL; } -size_t dso__fprintf(struct dso *self, FILE *fp) +int build_id__sprintf(u8 *self, int len, char *bf) { - size_t ret = fprintf(fp, "dso: %s\n", self->short_name); + char *bid = bf; + u8 *raw = self; + int i; + for (i = 0; i < len; ++i) { + sprintf(bid, "%02x", *raw); + ++raw; + bid += 2; + } + + return raw - self; +} + +size_t dso__fprintf(struct dso *self, FILE *fp) +{ + char sbuild_id[BUILD_ID_SIZE * 2 + 1]; struct rb_node *nd; + size_t ret; + + build_id__sprintf(self->build_id, sizeof(self->build_id), sbuild_id); + ret = fprintf(fp, "dso: %s (%s)\n", self->short_name, sbuild_id); + for (nd = rb_first(&self->syms); nd; nd = rb_next(nd)) { struct symbol *pos = rb_entry(nd, struct symbol, rb_node); ret += symbol__fprintf(pos, fp); @@ -825,8 +851,6 @@ out_close: return err; } -#define BUILD_ID_SIZE 20 - int filename__read_build_id(const char *filename, void *bf, size_t size) { int fd, err = -1; @@ -845,7 +869,7 @@ int filename__read_build_id(const char *filename, void *bf, size_t size) elf = elf_begin(fd, PERF_ELF_C_READ_MMAP, NULL); if (elf == NULL) { - pr_err("%s: cannot read %s ELF file.\n", __func__, filename); + pr_debug2("%s: cannot read %s ELF file.\n", __func__, filename); goto out_close; } @@ -874,9 +898,9 @@ out: static char *dso__read_build_id(struct dso *self) { - int i, len; - char *build_id = NULL, *bid; - unsigned char rawbf[BUILD_ID_SIZE], *raw; + int len; + char *build_id = NULL; + unsigned char rawbf[BUILD_ID_SIZE]; len = filename__read_build_id(self->long_name, rawbf, sizeof(rawbf)); if (len < 0) @@ -885,15 +909,8 @@ static char *dso__read_build_id(struct dso *self) build_id = malloc(len * 2 + 1); if (build_id == NULL) goto out; - bid = build_id; - raw = rawbf; - for (i = 0; i < len; ++i) { - sprintf(bid, "%02x", *raw); - ++raw; - bid += 2; - } - pr_debug2("%s(%s): %s\n", __func__, self->long_name, build_id); + build_id__sprintf(rawbf, len, build_id); out: return build_id; } @@ -922,7 +939,7 @@ int dso__load(struct dso *self, struct map *map, symbol_filter_t filter) int ret = -1; int fd; - self->loaded = true; + self->loaded = 1; if (!name) return -1; @@ -940,6 +957,8 @@ int dso__load(struct dso *self, struct map *map, symbol_filter_t filter) more: do { + int berr = 0; + self->origin++; switch (self->origin) { case DSO__ORIG_FEDORA: @@ -956,8 +975,7 @@ more: snprintf(name, size, "/usr/lib/debug/.build-id/%.2s/%s.debug", build_id, build_id + 2); - free(build_id); - break; + goto compare_build_id; } self->origin++; /* Fall thru */ @@ -969,6 +987,22 @@ more: goto out; } + if (self->has_build_id) { + bool match; + build_id = malloc(BUILD_ID_SIZE); + if (build_id == NULL) + goto more; + berr = filename__read_build_id(name, build_id, + BUILD_ID_SIZE); +compare_build_id: + match = berr > 0 && memcmp(build_id, self->build_id, + sizeof(self->build_id)) == 0; + free(build_id); + build_id = NULL; + if (!match) + goto more; + } + fd = open(name, O_RDONLY); } while (fd < 0); @@ -1034,7 +1068,7 @@ static int dso__load_module_sym(struct dso *self, struct map *map, { int err = 0, fd = open(self->long_name, O_RDONLY); - self->loaded = true; + self->loaded = 1; if (fd < 0) { pr_err("%s: cannot open %s\n", __func__, self->long_name); @@ -1225,7 +1259,7 @@ static int dso__load_vmlinux(struct dso *self, struct map *map, { int err, fd = open(vmlinux, O_RDONLY); - self->loaded = true; + self->loaded = 1; if (fd < 0) return -1; diff --git a/tools/perf/util/symbol.h b/tools/perf/util/symbol.h index e0d4a583f8d..f8c1899af48 100644 --- a/tools/perf/util/symbol.h +++ b/tools/perf/util/symbol.h @@ -60,10 +60,12 @@ struct dso { struct list_head node; struct rb_root syms; struct symbol *(*find_symbol)(struct dso *, u64 ip); - unsigned char adjust_symbols; - unsigned char slen_calculated; - bool loaded; + u8 adjust_symbols:1; + u8 slen_calculated:1; + u8 loaded:1; + u8 has_build_id:1; unsigned char origin; + u8 build_id[BUILD_ID_SIZE]; const char *short_name; char *long_name; char name[0]; @@ -81,8 +83,10 @@ void dsos__fprintf(FILE *fp); size_t dso__fprintf(struct dso *self, FILE *fp); char dso__symtab_origin(const struct dso *self); +void dso__set_build_id(struct dso *self, void *build_id); int filename__read_build_id(const char *filename, void *bf, size_t size); +int build_id__sprintf(u8 *self, int len, char *bf); int load_kernel(symbol_filter_t filter); -- cgit v1.2.3-18-g5258 From afe61f677866ffc484e69c4ecca2d316d564d78b Mon Sep 17 00:00:00 2001 From: Clark Williams Date: Sun, 8 Nov 2009 09:01:37 -0600 Subject: perf tools: Add debugfs utility routines for perf Add routines to locate the debugfs mount point and to manage the mounting and unmounting of the debugfs. Signed-off-by: Clark Williams Cc: Arnaldo Carvalho de Melo Cc: Peter Zijlstra LKML-Reference: <20091101155621.2b3503ee@torg> Signed-off-by: Ingo Molnar --- tools/perf/Makefile | 2 + tools/perf/util/debugfs.c | 241 ++++++++++++++++++++++++++++++++++++++++++++++ tools/perf/util/debugfs.h | 25 +++++ 3 files changed, 268 insertions(+) create mode 100644 tools/perf/util/debugfs.c create mode 100644 tools/perf/util/debugfs.h (limited to 'tools/perf') diff --git a/tools/perf/Makefile b/tools/perf/Makefile index 542b29e2e38..b9509b1cc32 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -353,6 +353,7 @@ LIB_H += util/include/asm/swab.h LIB_H += util/include/asm/system.h LIB_H += util/include/asm/uaccess.h LIB_H += perf.h +LIB_H += util/debugfs.h LIB_H += util/event.h LIB_H += util/types.h LIB_H += util/levenshtein.h @@ -378,6 +379,7 @@ LIB_OBJS += util/abspath.o LIB_OBJS += util/alias.o LIB_OBJS += util/config.o LIB_OBJS += util/ctype.o +LIB_OBJS += util/debugfs.o LIB_OBJS += util/environment.o LIB_OBJS += util/event.o LIB_OBJS += util/exec_cmd.o diff --git a/tools/perf/util/debugfs.c b/tools/perf/util/debugfs.c new file mode 100644 index 00000000000..06b73ee02c4 --- /dev/null +++ b/tools/perf/util/debugfs.c @@ -0,0 +1,241 @@ +#include "util.h" +#include "debugfs.h" +#include "cache.h" + +static int debugfs_premounted; +static char debugfs_mountpoint[MAX_PATH+1]; + +static const char *debugfs_known_mountpoints[] = { + "/sys/kernel/debug/", + "/debug/", + 0, +}; + +/* use this to force a umount */ +void debugfs_force_cleanup(void) +{ + debugfs_find_mountpoint(); + debugfs_premounted = 0; + debugfs_umount(); +} + +/* construct a full path to a debugfs element */ +int debugfs_make_path(const char *element, char *buffer, int size) +{ + int len; + + if (strlen(debugfs_mountpoint) == 0) { + buffer[0] = '\0'; + return -1; + } + + len = strlen(debugfs_mountpoint) + strlen(element) + 1; + if (len >= size) + return len+1; + + snprintf(buffer, size-1, "%s/%s", debugfs_mountpoint, element); + return 0; +} + +static int debugfs_found; + +/* find the path to the mounted debugfs */ +const char *debugfs_find_mountpoint(void) +{ + const char **ptr; + char type[100]; + FILE *fp; + + if (debugfs_found) + return (const char *) debugfs_mountpoint; + + ptr = debugfs_known_mountpoints; + while (*ptr) { + if (debugfs_valid_mountpoint(*ptr) == 0) { + debugfs_found = 1; + strcpy(debugfs_mountpoint, *ptr); + return debugfs_mountpoint; + } + ptr++; + } + + /* give up and parse /proc/mounts */ + fp = fopen("/proc/mounts", "r"); + if (fp == NULL) + die("Can't open /proc/mounts for read"); + + while (fscanf(fp, "%*s %" + STR(MAX_PATH) + "s %99s %*s %*d %*d\n", + debugfs_mountpoint, type) == 2) { + if (strcmp(type, "debugfs") == 0) + break; + } + fclose(fp); + + if (strcmp(type, "debugfs") != 0) + return NULL; + + debugfs_found = 1; + + return debugfs_mountpoint; +} + +/* verify that a mountpoint is actually a debugfs instance */ + +int debugfs_valid_mountpoint(const char *debugfs) +{ + struct statfs st_fs; + + if (statfs(debugfs, &st_fs) < 0) + return -ENOENT; + else if (st_fs.f_type != (long) DEBUGFS_MAGIC) + return -ENOENT; + + return 0; +} + + +int debugfs_valid_entry(const char *path) +{ + struct stat st; + + if (stat(path, &st)) + return -errno; + + return 0; +} + +/* mount the debugfs somewhere */ + +int debugfs_mount(const char *mountpoint) +{ + char mountcmd[128]; + + /* see if it's already mounted */ + if (debugfs_find_mountpoint()) { + debugfs_premounted = 1; + return 0; + } + + /* if not mounted and no argument */ + if (mountpoint == NULL) { + /* see if environment variable set */ + mountpoint = getenv(PERF_DEBUGFS_ENVIRONMENT); + /* if no environment variable, use default */ + if (mountpoint == NULL) + mountpoint = "/sys/kernel/debug"; + } + + /* save the mountpoint */ + strncpy(debugfs_mountpoint, mountpoint, sizeof(debugfs_mountpoint)); + + /* mount it */ + snprintf(mountcmd, sizeof(mountcmd), + "/bin/mount -t debugfs debugfs %s", mountpoint); + return system(mountcmd); +} + +/* umount the debugfs */ + +int debugfs_umount(void) +{ + char umountcmd[128]; + int ret; + + /* if it was already mounted, leave it */ + if (debugfs_premounted) + return 0; + + /* make sure it's a valid mount point */ + ret = debugfs_valid_mountpoint(debugfs_mountpoint); + if (ret) + return ret; + + snprintf(umountcmd, sizeof(umountcmd), + "/bin/umount %s", debugfs_mountpoint); + return system(umountcmd); +} + +int debugfs_write(const char *entry, const char *value) +{ + char path[MAX_PATH+1]; + int ret, count; + int fd; + + /* construct the path */ + snprintf(path, sizeof(path), "%s/%s", debugfs_mountpoint, entry); + + /* verify that it exists */ + ret = debugfs_valid_entry(path); + if (ret) + return ret; + + /* get how many chars we're going to write */ + count = strlen(value); + + /* open the debugfs entry */ + fd = open(path, O_RDWR); + if (fd < 0) + return -errno; + + while (count > 0) { + /* write it */ + ret = write(fd, value, count); + if (ret <= 0) { + if (ret == EAGAIN) + continue; + close(fd); + return -errno; + } + count -= ret; + } + + /* close it */ + close(fd); + + /* return success */ + return 0; +} + +/* + * read a debugfs entry + * returns the number of chars read or a negative errno + */ +int debugfs_read(const char *entry, char *buffer, size_t size) +{ + char path[MAX_PATH+1]; + int ret; + int fd; + + /* construct the path */ + snprintf(path, sizeof(path), "%s/%s", debugfs_mountpoint, entry); + + /* verify that it exists */ + ret = debugfs_valid_entry(path); + if (ret) + return ret; + + /* open the debugfs entry */ + fd = open(path, O_RDONLY); + if (fd < 0) + return -errno; + + do { + /* read it */ + ret = read(fd, buffer, size); + if (ret == 0) { + close(fd); + return EOF; + } + } while (ret < 0 && errno == EAGAIN); + + /* close it */ + close(fd); + + /* make *sure* there's a null character at the end */ + buffer[ret] = '\0'; + + /* return the number of chars read */ + return ret; +} diff --git a/tools/perf/util/debugfs.h b/tools/perf/util/debugfs.h new file mode 100644 index 00000000000..3cd14f9ae78 --- /dev/null +++ b/tools/perf/util/debugfs.h @@ -0,0 +1,25 @@ +#ifndef __DEBUGFS_H__ +#define __DEBUGFS_H__ + +#include + +#ifndef MAX_PATH +# define MAX_PATH 256 +#endif + +#ifndef STR +# define _STR(x) #x +# define STR(x) _STR(x) +#endif + +extern const char *debugfs_find_mountpoint(void); +extern int debugfs_valid_mountpoint(const char *debugfs); +extern int debugfs_valid_entry(const char *path); +extern int debugfs_mount(const char *mountpoint); +extern int debugfs_umount(void); +extern int debugfs_write(const char *entry, const char *value); +extern int debugfs_read(const char *entry, char *buffer, size_t size); +extern void debugfs_force_cleanup(void); +extern int debugfs_make_path(const char *element, char *buffer, int size); + +#endif /* __DEBUGFS_H__ */ -- cgit v1.2.3-18-g5258 From 549104f22b3cd4761145eb5fba6ee4d59822da61 Mon Sep 17 00:00:00 2001 From: Clark Williams Date: Sun, 8 Nov 2009 09:03:07 -0600 Subject: perf tools: Modify perf routines to use new debugfs routines modify perf.c get_debugfs_mntpnt() to use the util/debugfs.c debugfs_find_mountpoint() modify util/parse-events.c to use debugfs_valid_mountpoint(). Signed-off-by: Clark Williams Cc: Arnaldo Carvalho de Melo Cc: Peter Zijlstra LKML-Reference: <20091101155720.624cc87e@torg> Signed-off-by: Ingo Molnar --- tools/perf/perf.c | 44 ++++++------------------------------------ tools/perf/util/parse-events.c | 17 ++++------------ 2 files changed, 10 insertions(+), 51 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/perf.c b/tools/perf/perf.c index 624e62d9d1e..601f403fbda 100644 --- a/tools/perf/perf.c +++ b/tools/perf/perf.c @@ -14,6 +14,7 @@ #include "util/run-command.h" #include "util/parse-events.h" #include "util/string.h" +#include "util/debugfs.h" const char perf_usage_string[] = "perf [--version] [--help] COMMAND [ARGS]"; @@ -382,45 +383,12 @@ static int run_argv(int *argcp, const char ***argv) /* mini /proc/mounts parser: searching for "^blah /mount/point debugfs" */ static void get_debugfs_mntpt(void) { - FILE *file; - char fs_type[100]; - char debugfs[MAXPATHLEN]; + const char *path = debugfs_find_mountpoint(); - /* - * try the standard location - */ - if (valid_debugfs_mount("/sys/kernel/debug/") == 0) { - strcpy(debugfs_mntpt, "/sys/kernel/debug/"); - return; - } - - /* - * try the sane location - */ - if (valid_debugfs_mount("/debug/") == 0) { - strcpy(debugfs_mntpt, "/debug/"); - return; - } - - /* - * give up and parse /proc/mounts - */ - file = fopen("/proc/mounts", "r"); - if (file == NULL) - return; - - while (fscanf(file, "%*s %" - STR(MAXPATHLEN) - "s %99s %*s %*d %*d\n", - debugfs, fs_type) == 2) { - if (strcmp(fs_type, "debugfs") == 0) - break; - } - fclose(file); - if (strcmp(fs_type, "debugfs") == 0) { - strncpy(debugfs_mntpt, debugfs, MAXPATHLEN); - debugfs_mntpt[MAXPATHLEN - 1] = '\0'; - } + if (path) + strncpy(debugfs_mntpt, path, sizeof(debugfs_mntpt)); + else + debugfs_mntpt[0] = '\0'; } int main(int argc, const char **argv) diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c index 31baa5a6036..097938a96d7 100644 --- a/tools/perf/util/parse-events.c +++ b/tools/perf/util/parse-events.c @@ -7,6 +7,7 @@ #include "string.h" #include "cache.h" #include "header.h" +#include "debugfs.h" int nr_counters; @@ -149,16 +150,6 @@ static int tp_event_has_id(struct dirent *sys_dir, struct dirent *evt_dir) #define MAX_EVENT_LENGTH 512 -int valid_debugfs_mount(const char *debugfs) -{ - struct statfs st_fs; - - if (statfs(debugfs, &st_fs) < 0) - return -ENOENT; - else if (st_fs.f_type != (long) DEBUGFS_MAGIC) - return -ENOENT; - return 0; -} struct tracepoint_path *tracepoint_id_to_path(u64 config) { @@ -171,7 +162,7 @@ struct tracepoint_path *tracepoint_id_to_path(u64 config) char evt_path[MAXPATHLEN]; char dir_path[MAXPATHLEN]; - if (valid_debugfs_mount(debugfs_path)) + if (debugfs_valid_mountpoint(debugfs_path)) return NULL; sys_dir = opendir(debugfs_path); @@ -510,7 +501,7 @@ static enum event_result parse_tracepoint_event(const char **strp, char sys_name[MAX_EVENT_LENGTH]; unsigned int sys_length, evt_length; - if (valid_debugfs_mount(debugfs_path)) + if (debugfs_valid_mountpoint(debugfs_path)) return 0; evt_name = strchr(*strp, ':'); @@ -788,7 +779,7 @@ static void print_tracepoint_events(void) char evt_path[MAXPATHLEN]; char dir_path[MAXPATHLEN]; - if (valid_debugfs_mount(debugfs_path)) + if (debugfs_valid_mountpoint(debugfs_path)) return; sys_dir = opendir(debugfs_path); -- cgit v1.2.3-18-g5258 From 5ff0cfc67f00fe0feaa1da0b2359232ea4aa0ee7 Mon Sep 17 00:00:00 2001 From: Hitoshi Mitake Date: Mon, 9 Nov 2009 12:31:05 +0900 Subject: perf bench: Fix bench/sched-pipe.c to wait for child process Ingo reported this small 'perf bench sched pipe' output problem: | $ ./perf bench sched pipe | (executing 1000000 pipe operations between two tasks) | | Total time:4.898 sec | $ 4.898586 usecs/op | 204140 ops/sec | | the shell prompt came back before the usecs/op and ops/sec line | was printed. Process teardown race, lack of wait() or so? This caused by lack of calling waitpid() by parent process, so I added it. Signed-off-by: Hitoshi Mitake Cc: Rusty Russell Cc: Peter Zijlstra Cc: Paul Mackerras Cc: Mike Galbraith Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: Jiri Kosina LKML-Reference: <1257737465-7546-1-git-send-email-mitake@dcl.info.waseda.ac.jp> Signed-off-by: Ingo Molnar --- tools/perf/bench/sched-pipe.c | 10 +++++++--- 1 file changed, 7 insertions(+), 3 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/bench/sched-pipe.c b/tools/perf/bench/sched-pipe.c index 3214ed20b1a..6a29100e928 100644 --- a/tools/perf/bench/sched-pipe.c +++ b/tools/perf/bench/sched-pipe.c @@ -26,6 +26,7 @@ #include #include #include +#include #define LOOPS_DEFAULT 1000000 static int loops = LOOPS_DEFAULT; @@ -58,8 +59,8 @@ int bench_sched_pipe(int argc, const char **argv, * discarding returned value of read(), write() * causes error in building environment for perf */ - int ret; - pid_t pid; + int ret, wait_stat; + pid_t pid, retpid; argc = parse_options(argc, argv, options, bench_sched_pipe_usage, 0); @@ -87,8 +88,11 @@ int bench_sched_pipe(int argc, const char **argv, gettimeofday(&stop, NULL); timersub(&stop, &start, &diff); - if (pid) + if (pid) { + retpid = waitpid(pid, &wait_stat, 0); + assert((retpid == pid) && WIFEXITED(wait_stat)); return 0; + } if (simple) printf("%lu.%03lu\n", -- cgit v1.2.3-18-g5258 From ca2b900f9af1586b9889ccc4b12e453c13268bd5 Mon Sep 17 00:00:00 2001 From: Zeev Tarantov Date: Mon, 9 Nov 2009 13:26:13 +0200 Subject: perf tools: Fix syntax in documentation Fix trivial syntax in perf-events user-space tools documentation. Signed-off-by: Zeev Tarantov Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker LKML-Reference: <12d7e64c0911081811i7e5b466cu6706ff6ab3e70db4@mail.gmail.com> Signed-off-by: Ingo Molnar --- tools/perf/Documentation/perf-report.txt | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/Documentation/perf-report.txt b/tools/perf/Documentation/perf-report.txt index 59f0b846cd7..9dccb180b7a 100644 --- a/tools/perf/Documentation/perf-report.txt +++ b/tools/perf/Documentation/perf-report.txt @@ -24,11 +24,11 @@ OPTIONS --dsos=:: Only consider symbols in these dsos. CSV that understands file://filename entries. --n ---show-nr-samples +-n:: +--show-nr-samples:: Show the number of samples for each symbol --T ---threads +-T:: +--threads:: Show per-thread event counters -C:: --comms=:: -- cgit v1.2.3-18-g5258 From 242aa14a67f4e19453fc8a51cffc5ac5ee5bcbd1 Mon Sep 17 00:00:00 2001 From: Hitoshi Mitake Date: Tue, 10 Nov 2009 08:19:59 +0900 Subject: perf bench: Add format constants to bench.h for unified output formatting This patch adds some constants and extern declaration to bench.h. These are used for unified output formatting of 'perf bench'. Signed-off-by: Hitoshi Mitake Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1257808802-9420-2-git-send-email-mitake@dcl.info.waseda.ac.jp> Signed-off-by: Ingo Molnar --- tools/perf/bench/bench.h | 9 +++++++++ 1 file changed, 9 insertions(+) (limited to 'tools/perf') diff --git a/tools/perf/bench/bench.h b/tools/perf/bench/bench.h index 59adb279cd7..42167ea4194 100644 --- a/tools/perf/bench/bench.h +++ b/tools/perf/bench/bench.h @@ -6,4 +6,13 @@ extern int bench_sched_messaging(int argc, const char **argv, extern int bench_sched_pipe(int argc, const char **argv, const char *prefix); +#define BENCH_FORMAT_DEFAULT_STR "default" +#define BENCH_FORMAT_DEFAULT 0 +#define BENCH_FORMAT_SIMPLE_STR "simple" +#define BENCH_FORMAT_SIMPLE 1 + +#define BENCH_FORMAT_UNKNOWN -1 + +extern int bench_format; + #endif -- cgit v1.2.3-18-g5258 From 386d7e9e542c2115d5d300747e57f503458a1617 Mon Sep 17 00:00:00 2001 From: Hitoshi Mitake Date: Tue, 10 Nov 2009 08:20:00 +0900 Subject: perf bench: Modify builtin-bench.c for processing common options This patch modifies builtin-bench.c for processing common options. The first option added is "--format". Users of perf bench will be able to specify output style by --format. Usage example: % ./perf bench sched messaging # with no style specify (20 sender and receiver processes per group) (10 groups == 400 processes run) Total time:1.431 sec % ./perf bench --format=simple sched messaging # specified simple 1.431 Signed-off-by: Hitoshi Mitake Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1257808802-9420-3-git-send-email-mitake@dcl.info.waseda.ac.jp> Signed-off-by: Ingo Molnar --- tools/perf/builtin-bench.c | 79 ++++++++++++++++++++++++++++++++++++++-------- 1 file changed, 65 insertions(+), 14 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-bench.c b/tools/perf/builtin-bench.c index 31f41643b0c..c7505eaff84 100644 --- a/tools/perf/builtin-bench.c +++ b/tools/perf/builtin-bench.c @@ -74,53 +74,104 @@ static void dump_suites(int subsys_index) return; } +static char *bench_format_str; +int bench_format = BENCH_FORMAT_DEFAULT; + +static const struct option bench_options[] = { + OPT_STRING('f', "format", &bench_format_str, "default", + "Specify format style"), + OPT_END() +}; + +static const char * const bench_usage[] = { + "perf bench [] []", + NULL +}; + +static void print_usage(void) +{ + int i; + + printf("Usage: \n"); + for (i = 0; bench_usage[i]; i++) + printf("\t%s\n", bench_usage[i]); + printf("\n"); + + printf("List of available subsystems...\n\n"); + + for (i = 0; subsystems[i].name; i++) + printf("\t%s: %s\n", + subsystems[i].name, subsystems[i].summary); + printf("\n"); +} + +static int bench_str2int(char *str) +{ + if (!str) + return BENCH_FORMAT_DEFAULT; + + if (!strcmp(str, BENCH_FORMAT_DEFAULT_STR)) + return BENCH_FORMAT_DEFAULT; + else if (!strcmp(str, BENCH_FORMAT_SIMPLE_STR)) + return BENCH_FORMAT_SIMPLE; + + return BENCH_FORMAT_UNKNOWN; +} + int cmd_bench(int argc, const char **argv, const char *prefix __used) { int i, j, status = 0; if (argc < 2) { /* No subsystem specified. */ - printf("Usage: perf bench []\n\n"); - printf("List of available subsystems...\n\n"); + print_usage(); + goto end; + } - for (i = 0; subsystems[i].name; i++) - printf("\t%s: %s\n", - subsystems[i].name, subsystems[i].summary); - printf("\n"); + argc = parse_options(argc, argv, bench_options, bench_usage, + PARSE_OPT_STOP_AT_NON_OPTION); + + bench_format = bench_str2int(bench_format_str); + if (bench_format == BENCH_FORMAT_UNKNOWN) { + printf("Unknown format descriptor:%s\n", bench_format_str); + goto end; + } + if (argc < 1) { + print_usage(); goto end; } for (i = 0; subsystems[i].name; i++) { - if (strcmp(subsystems[i].name, argv[1])) + if (strcmp(subsystems[i].name, argv[0])) continue; - if (argc < 3) { + if (argc < 2) { /* No suite specified. */ dump_suites(i); goto end; } for (j = 0; subsystems[i].suites[j].name; j++) { - if (strcmp(subsystems[i].suites[j].name, argv[2])) + if (strcmp(subsystems[i].suites[j].name, argv[1])) continue; - status = subsystems[i].suites[j].fn(argc - 2, - argv + 2, prefix); + status = subsystems[i].suites[j].fn(argc - 1, + argv + 1, prefix); goto end; } - if (!strcmp(argv[2], "-h") || !strcmp(argv[2], "--help")) { + if (!strcmp(argv[1], "-h") || !strcmp(argv[1], "--help")) { dump_suites(i); goto end; } - printf("Unknown suite:%s for %s\n", argv[2], argv[1]); + printf("Unknown suite:%s for %s\n", argv[1], argv[0]); status = 1; goto end; } - printf("Unknown subsystem:%s\n", argv[1]); + printf("Unknown subsystem:%s\n", argv[0]); status = 1; end: -- cgit v1.2.3-18-g5258 From cced06c62a9db6bd6d77e3f0a57dbe47a26d881e Mon Sep 17 00:00:00 2001 From: Hitoshi Mitake Date: Tue, 10 Nov 2009 08:20:01 +0900 Subject: perf bench: Modify bench/bench-messaging.c to adopt unified output formatting This patch modifies bench/bench-messaging.c to adopt unified output formatting: --format option. Usage example: % ./perf bench sched messaging # with no style specify (20 sender and receiver processes per group) (10 groups == 400 processes run) Total time:1.431 sec % ./perf bench --format=simple sched messaging # specified simple 1.431 Signed-off-by: Hitoshi Mitake Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1257808802-9420-4-git-send-email-mitake@dcl.info.waseda.ac.jp> Signed-off-by: Ingo Molnar --- tools/perf/bench/sched-messaging.c | 18 +++++++++++------- 1 file changed, 11 insertions(+), 7 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/bench/sched-messaging.c b/tools/perf/bench/sched-messaging.c index 36b62c549e2..2cc5edcca3d 100644 --- a/tools/perf/bench/sched-messaging.c +++ b/tools/perf/bench/sched-messaging.c @@ -35,7 +35,6 @@ static int use_pipes = 0; static unsigned int loops = 100; static unsigned int thread_mode = 0; static unsigned int num_groups = 10; -static int simple = 0; struct sender_context { unsigned int num_fds; @@ -261,9 +260,6 @@ static const struct option options[] = { "Specify number of groups"), OPT_INTEGER('l', "loop", &loops, "Specify number of loops"), - OPT_BOOLEAN('s', "simple-output", &simple, - "Do simple output (this maybe useful for" - "processing by scripts or graph tools like gnuplot)"), OPT_END() }; @@ -316,9 +312,8 @@ int bench_sched_messaging(int argc, const char **argv, timersub(&stop, &start, &diff); - if (simple) - printf("%lu.%03lu\n", diff.tv_sec, diff.tv_usec/1000); - else { + switch (bench_format) { + case BENCH_FORMAT_DEFAULT: printf("(%d sender and receiver %s per group)\n", num_fds, thread_mode ? "threads" : "processes"); printf("(%d groups == %d %s run)\n\n", @@ -326,6 +321,15 @@ int bench_sched_messaging(int argc, const char **argv, thread_mode ? "threads" : "processes"); printf("\tTotal time:%lu.%03lu sec\n", diff.tv_sec, diff.tv_usec/1000); + break; + case BENCH_FORMAT_SIMPLE: + printf("%lu.%03lu\n", diff.tv_sec, diff.tv_usec/1000); + break; + default: + /* reaching here is something disaster */ + fprintf(stderr, "Unknown format:%d\n", bench_format); + exit(1); + break; } return 0; -- cgit v1.2.3-18-g5258 From 158ba827f6deef4102c5247ed4b6a587f0bd6a07 Mon Sep 17 00:00:00 2001 From: Hitoshi Mitake Date: Tue, 10 Nov 2009 08:20:02 +0900 Subject: perf bench: Modify builtin-pipe.c for processing common options This patch modifies builtin-pipe.c for processing common options. The first option added is "--format". Users of perf bench will be able to specify output style by --format. Usage example: % ./perf bench sched pipe # with no style specify (executing 1000000 pipe operations between two tasks) Total time:5.855 sec 5.855061 usecs/op 170792 ops/sec % ./perf bench --format=simple sched pipe # specified simple 5.988 Signed-off-by: Hitoshi Mitake Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1257808802-9420-5-git-send-email-mitake@dcl.info.waseda.ac.jp> Signed-off-by: Ingo Molnar Cc: Peter Zijlstra Cc: Paul Mackerras --- tools/perf/bench/sched-pipe.c | 22 ++++++++++++++-------- 1 file changed, 14 insertions(+), 8 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/bench/sched-pipe.c b/tools/perf/bench/sched-pipe.c index 6a29100e928..a9ac186714b 100644 --- a/tools/perf/bench/sched-pipe.c +++ b/tools/perf/bench/sched-pipe.c @@ -30,14 +30,10 @@ #define LOOPS_DEFAULT 1000000 static int loops = LOOPS_DEFAULT; -static int simple = 0; static const struct option options[] = { OPT_INTEGER('l', "loop", &loops, "Specify number of loops"), - OPT_BOOLEAN('s', "simple-output", &simple, - "Do simple output (this maybe useful for" - "processing by scripts or graph tools like gnuplot)"), OPT_END() }; @@ -94,10 +90,8 @@ int bench_sched_pipe(int argc, const char **argv, return 0; } - if (simple) - printf("%lu.%03lu\n", - diff.tv_sec, diff.tv_usec / 1000); - else { + switch (bench_format) { + case BENCH_FORMAT_DEFAULT: printf("(executing %d pipe operations between two tasks)\n\n", loops); @@ -111,6 +105,18 @@ int bench_sched_pipe(int argc, const char **argv, printf("\t\t%d ops/sec\n", (int)((double)loops / ((double)result_usec / (double)1000000))); + break; + + case BENCH_FORMAT_SIMPLE: + printf("%lu.%03lu\n", + diff.tv_sec, diff.tv_usec / 1000); + break; + + default: + /* reaching here is something disaster */ + fprintf(stderr, "Unknown format:%d\n", bench_format); + exit(1); + break; } return 0; -- cgit v1.2.3-18-g5258 From 606bc1e18d346fc7d7fb333909cc95b06b1ca5b1 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Tue, 10 Nov 2009 20:50:53 +0900 Subject: perf bench: Clean up bench/bench.h Clean up initializers in bench.h: - No need to break the line for function prototypes, they are more readable in a single line. (even if checkpatch complains about it - We try to align definitions / structure fields vertically, to make it all a bit more readable. Signed-off-by: Ingo Molnar Signed-off-by: Hitoshi Mitake Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1257853855-28934-2-git-send-email-mitake@dcl.info.waseda.ac.jp> --- tools/perf/bench/bench.h | 16 +++++++--------- 1 file changed, 7 insertions(+), 9 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/bench/bench.h b/tools/perf/bench/bench.h index 42167ea4194..9fbd8d745fa 100644 --- a/tools/perf/bench/bench.h +++ b/tools/perf/bench/bench.h @@ -1,17 +1,15 @@ #ifndef BENCH_H #define BENCH_H -extern int bench_sched_messaging(int argc, const char **argv, - const char *prefix); -extern int bench_sched_pipe(int argc, const char **argv, - const char *prefix); +extern int bench_sched_messaging(int argc, const char **argv, const char *prefix); +extern int bench_sched_pipe(int argc, const char **argv, const char *prefix); -#define BENCH_FORMAT_DEFAULT_STR "default" -#define BENCH_FORMAT_DEFAULT 0 -#define BENCH_FORMAT_SIMPLE_STR "simple" -#define BENCH_FORMAT_SIMPLE 1 +#define BENCH_FORMAT_DEFAULT_STR "default" +#define BENCH_FORMAT_DEFAULT 0 +#define BENCH_FORMAT_SIMPLE_STR "simple" +#define BENCH_FORMAT_SIMPLE 1 -#define BENCH_FORMAT_UNKNOWN -1 +#define BENCH_FORMAT_UNKNOWN -1 extern int bench_format; -- cgit v1.2.3-18-g5258 From 9fbc04f2493929a69fd9e53b5fb53c127d7950d5 Mon Sep 17 00:00:00 2001 From: Hitoshi Mitake Date: Tue, 10 Nov 2009 20:50:54 +0900 Subject: perf bench: Add new document about perf-bench This patch adds new document about perf-bench. Man page and html will be provided for user. Signed-off-by: Hitoshi Mitake Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1257853855-28934-3-git-send-email-mitake@dcl.info.waseda.ac.jp> Signed-off-by: Ingo Molnar --- tools/perf/Documentation/perf-bench.txt | 120 ++++++++++++++++++++++++++++++++ 1 file changed, 120 insertions(+) create mode 100644 tools/perf/Documentation/perf-bench.txt (limited to 'tools/perf') diff --git a/tools/perf/Documentation/perf-bench.txt b/tools/perf/Documentation/perf-bench.txt new file mode 100644 index 00000000000..ae525ac5a2c --- /dev/null +++ b/tools/perf/Documentation/perf-bench.txt @@ -0,0 +1,120 @@ +perf-bench(1) +============ + +NAME +---- +perf-bench - General framework for benchmark suites + +SYNOPSIS +-------- +[verse] +'perf bench' [] [] + +DESCRIPTION +----------- +This 'perf bench' command is general framework for benchmark suites. + +COMMON OPTIONS +-------------- +-f:: +--format=:: +Specify format style. +Current available format styles are, + +'default':: +Default style. This is mainly for human reading. +--------------------- +% perf bench sched pipe # with no style specify +(executing 1000000 pipe operations between two tasks) + Total time:5.855 sec + 5.855061 usecs/op + 170792 ops/sec +--------------------- + +'simple':: +This simple style is friendly for automated +processing by scripts. +--------------------- +% perf bench --format=simple sched pipe # specified simple +5.988 +--------------------- + +SUBSYSTEM +--------- + +'sched':: + Scheduler and IPC mechanisms. + +SUITES FOR 'sched' +~~~~~~~~~~~~~~~~~~ +*messaging*:: +Suite for evaluating performance of scheduler and IPC mechanisms. +Based on hackbench by Rusty Russell. + +Options of *pipe* +^^^^^^^^^^^^^^^^^ +-p:: +--pipe:: +Use pipe() instead of socketpair() + +-t:: +--thread:: +Be multi thread instead of multi process + +-g:: +--group=:: +Specify number of groups + +-l:: +--loop=:: +Specify number of loops + +Example of *messaging* +^^^^^^^^^^^^^^^^^^^^^^ + +--------------------- +% perf bench sched messaging # run with default +options (20 sender and receiver processes per group) +(10 groups == 400 processes run) + + Total time:0.308 sec + +% perf bench sched messaging -t -g 20 # be multi-thread,with 20 groups +(20 sender and receiver threads per group) +(20 groups == 800 threads run) + + Total time:0.582 sec +--------------------- + +*pipe*:: +Suite for pipe() system call. +Based on pipe-test-1m.c by Ingo Molnar. + +Options of *pipe* +^^^^^^^^^^^^^^^^^ +-l:: +--loop=:: +Specify number of loops. + +Example of *pipe* +^^^^^^^^^^^^^^^^^ + +--------------------- +% perf bench sched pipe +(executing 1000000 pipe operations between two tasks) + + Total time:8.091 sec + 8.091833 usecs/op + 123581 ops/sec + +% perf bench sched pipe -l 1000 # loop 1000 +(executing 1000 pipe operations between two tasks) + + Total time:0.016 sec + 16.948000 usecs/op + 59004 ops/sec +--------------------- + +SEE ALSO +-------- +linkperf:perf[1] -- cgit v1.2.3-18-g5258 From 8d8d61aadb9d8cce07f7dcdb77a4c20a25d36d07 Mon Sep 17 00:00:00 2001 From: Hitoshi Mitake Date: Tue, 10 Nov 2009 20:50:55 +0900 Subject: perf bench: Modify command-list.txt for the entry of perf-bench This patch modifies command-list.txt for the entry of perf-bench. So perf will show 'bench' in command list. Example: % perf usage: perf [--version] [--help] COMMAND [ARGS] The most commonly used perf commands are: annotate Read perf.data (created by perf record) and display annotated code bench General framework for benchmark suites ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ list List all symbolic event types probe Define new dynamic tracepoints record Run a command and record its profile into perf.data report Read perf.data (created by perf record) and display the profile sched Tool to trace/measure scheduler properties (latencies) stat Run a command and gather performance counter statistics timechart Tool to visualize total system behavior during a workload top System profiling tool. trace Read perf.data (created by perf record) and display trace output See 'perf help COMMAND' for more information on a specific command. Signed-off-by: Hitoshi Mitake Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1257853855-28934-4-git-send-email-mitake@dcl.info.waseda.ac.jp> Signed-off-by: Ingo Molnar --- tools/perf/command-list.txt | 1 + 1 file changed, 1 insertion(+) (limited to 'tools/perf') diff --git a/tools/perf/command-list.txt b/tools/perf/command-list.txt index 00326e230d8..981c40b9a5e 100644 --- a/tools/perf/command-list.txt +++ b/tools/perf/command-list.txt @@ -3,6 +3,7 @@ # command name category [deprecated] [common] # perf-annotate mainporcelain common +perf-bench mainporcelain common perf-list mainporcelain common perf-sched mainporcelain common perf-record mainporcelain common -- cgit v1.2.3-18-g5258 From 79e295d4bd0f524257299e7c4e42f643f21abcc2 Mon Sep 17 00:00:00 2001 From: Hitoshi Mitake Date: Wed, 11 Nov 2009 00:04:00 +0900 Subject: perf bench: Improve builtin-bench.c for more friendly output This patch makes output of perf bench more friendly. Current style of putput, keeping user wait and printing everything suddenly when we finish, may confuse users. So I improved it: | % perf bench sched messaging | # Running sched/messaging benchmark... <- printed right after invocation | # 20 sender and receiver processes per group | # 10 groups == 400 processes run | | Total time: 1.476 [sec] Signed-off-by: Hitoshi Mitake Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1257865442-20252-2-git-send-email-mitake@dcl.info.waseda.ac.jp> Signed-off-by: Ingo Molnar --- tools/perf/builtin-bench.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'tools/perf') diff --git a/tools/perf/builtin-bench.c b/tools/perf/builtin-bench.c index c7505eaff84..90c39baae0d 100644 --- a/tools/perf/builtin-bench.c +++ b/tools/perf/builtin-bench.c @@ -156,6 +156,10 @@ int cmd_bench(int argc, const char **argv, const char *prefix __used) if (strcmp(subsystems[i].suites[j].name, argv[1])) continue; + if (bench_format == BENCH_FORMAT_DEFAULT) + printf("# Running %s/%s benchmark...\n", + subsystems[i].name, + subsystems[i].suites[j].name); status = subsystems[i].suites[j].fn(argc - 1, argv + 1, prefix); goto end; -- cgit v1.2.3-18-g5258 From ff676b193a401b23c84a79a7ec06559f3eaae917 Mon Sep 17 00:00:00 2001 From: Hitoshi Mitake Date: Wed, 11 Nov 2009 00:04:01 +0900 Subject: perf bench: Improve sched-pipe.c with more comfortable output This patch improves sched-pipe.c with more comfortable output. Change points are comment style description and formatting numerical values and its units. Example: | % ./perf bench sched pipe | # Running sched/pipe benchmark... | # Extecuted 1000000 pipe operations between two tasks | | Total time:5.822 [sec] | | 5.822553 usecs/op | 171745 ops/sec Signed-off-by: Hitoshi Mitake Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1257865442-20252-3-git-send-email-mitake@dcl.info.waseda.ac.jp> Signed-off-by: Ingo Molnar --- tools/perf/bench/sched-pipe.c | 11 ++++++----- 1 file changed, 6 insertions(+), 5 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/bench/sched-pipe.c b/tools/perf/bench/sched-pipe.c index a9ac186714b..238185f9797 100644 --- a/tools/perf/bench/sched-pipe.c +++ b/tools/perf/bench/sched-pipe.c @@ -92,17 +92,18 @@ int bench_sched_pipe(int argc, const char **argv, switch (bench_format) { case BENCH_FORMAT_DEFAULT: - printf("(executing %d pipe operations between two tasks)\n\n", + printf("# Extecuted %d pipe operations between two tasks\n\n", loops); result_usec = diff.tv_sec * 1000000; result_usec += diff.tv_usec; - printf("\tTotal time:%lu.%03lu sec\n", - diff.tv_sec, diff.tv_usec / 1000); - printf("\t\t%lf usecs/op\n", + printf(" %14s: %lu.%03lu [sec]\n\n", "Total time", + diff.tv_sec, diff.tv_usec/1000); + + printf(" %14lf usecs/op\n", (double)result_usec / (double)loops); - printf("\t\t%d ops/sec\n", + printf(" %14d ops/sec\n", (int)((double)loops / ((double)result_usec / (double)1000000))); break; -- cgit v1.2.3-18-g5258 From c5659b74f052150791750234f92dcfb29d27efa5 Mon Sep 17 00:00:00 2001 From: Hitoshi Mitake Date: Wed, 11 Nov 2009 00:04:02 +0900 Subject: perf bench: Improve sched-message.c with more comfortable output This patch improves sched-message.c with more comfortable output. Change points are comment style description and formatting numerical values and its units. Example: | % perf bench sched messaging | # Running sched/messaging benchmark... | # 20 sender and receiver processes per group | # 10 groups == 400 processes run | | Total time: 1.490 [sec] Signed-off-by: Hitoshi Mitake Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1257865442-20252-4-git-send-email-mitake@dcl.info.waseda.ac.jp> Signed-off-by: Ingo Molnar Cc: Peter Zijlstra Cc: Paul Mackerras --- tools/perf/bench/sched-messaging.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/bench/sched-messaging.c b/tools/perf/bench/sched-messaging.c index 2cc5edcca3d..605a2a959aa 100644 --- a/tools/perf/bench/sched-messaging.c +++ b/tools/perf/bench/sched-messaging.c @@ -314,12 +314,12 @@ int bench_sched_messaging(int argc, const char **argv, switch (bench_format) { case BENCH_FORMAT_DEFAULT: - printf("(%d sender and receiver %s per group)\n", + printf("# %d sender and receiver %s per group\n", num_fds, thread_mode ? "threads" : "processes"); - printf("(%d groups == %d %s run)\n\n", + printf("# %d groups == %d %s run\n\n", num_groups, num_groups * 2 * num_fds, thread_mode ? "threads" : "processes"); - printf("\tTotal time:%lu.%03lu sec\n", + printf(" %14s: %lu.%03lu [sec]\n", "Total time", diff.tv_sec, diff.tv_usec/1000); break; case BENCH_FORMAT_SIMPLE: -- cgit v1.2.3-18-g5258 From de8967214d8ce536161a1ad6538ad1cb82e7428d Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Wed, 11 Nov 2009 04:51:02 +0100 Subject: perf tools: Synthetize the targeted process Don't forget to also synthetize the targeted process from perf record or we'll miss its dso in the events and then we won't be able to deal with its build-id. We are missing it because it is created after the existing synthetized tasks but before the counters are enabled and can send its mapping event. Signed-off-by: Frederic Weisbecker Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Mike Galbraith Cc: Paul Mackerras Cc: Hitoshi Mitake LKML-Reference: <1257911467-28276-1-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar --- tools/perf/builtin-record.c | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index ab333812ace..9f98b86e747 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -497,13 +497,22 @@ static int __cmd_record(int argc, const char **argv) if (target_pid == -1 && argc) { pid = fork(); if (pid < 0) - perror("failed to fork"); + die("failed to fork"); if (!pid) { if (execvp(argv[0], (char **)argv)) { perror(argv[0]); exit(-1); } + } else { + /* + * Wait a bit for the execv'ed child to appear + * and be updated in /proc + * FIXME: Do you know a less heuristical solution? + */ + usleep(1000); + event__synthesize_thread(pid, + process_synthesized_event); } child_pid = pid; -- cgit v1.2.3-18-g5258 From 8671dab9d5b2f0b444b8d09792384dccbfd43d14 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Wed, 11 Nov 2009 04:51:03 +0100 Subject: perf tools: Move the build-id storage operations to headers So that it makes easier to control it. Especially because we plan to give it a feature section. Signed-off-by: Frederic Weisbecker Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Mike Galbraith Cc: Paul Mackerras Cc: Hitoshi Mitake LKML-Reference: <1257911467-28276-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar --- tools/perf/builtin-record.c | 32 ++------------------------------ tools/perf/util/header.c | 41 ++++++++++++++++++++++++++++++++++++++--- tools/perf/util/header.h | 2 +- 3 files changed, 41 insertions(+), 34 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 9f98b86e747..c35e61b3098 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -378,39 +378,11 @@ static void open_counters(int cpu, pid_t pid) nr_cpu++; } -static bool write_buildid_table(void) -{ - struct dso *pos; - bool have_buildid = false; - - list_for_each_entry(pos, &dsos, node) { - struct build_id_event b; - size_t len; - - if (filename__read_build_id(pos->long_name, - &b.build_id, - sizeof(b.build_id)) < 0) - continue; - have_buildid = true; - memset(&b.header, 0, sizeof(b.header)); - len = strlen(pos->long_name) + 1; - len = ALIGN(len, 64); - b.header.size = sizeof(b) + len; - write_output(&b, sizeof(b)); - write_output(pos->long_name, len); - } - - return have_buildid; -} - static void atexit_header(void) { header->data_size += bytes_written; - if (write_buildid_table()) - perf_header__set_feat(header, HEADER_BUILD_ID); - - perf_header__write(header, output); + perf_header__write(header, output, true); } static int __cmd_record(int argc, const char **argv) @@ -487,7 +459,7 @@ static int __cmd_record(int argc, const char **argv) } if (file_new) - perf_header__write(header, output); + perf_header__write(header, output, false); if (!system_wide) event__synthesize_thread(pid, process_synthesized_event); diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index 050f543fd96..a4d0bbef9a4 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -2,11 +2,13 @@ #include #include #include +#include #include "util.h" #include "header.h" #include "../perf.h" #include "trace-event.h" +#include "symbol.h" /* * Create new perf.data header attribute: @@ -172,7 +174,33 @@ static void do_write(int fd, void *buf, size_t size) } } -static void perf_header__adds_write(struct perf_header *self, int fd) +static bool write_buildid_table(int fd) +{ + struct dso *pos; + bool have_buildid = false; + + list_for_each_entry(pos, &dsos, node) { + struct build_id_event b; + size_t len; + + if (filename__read_build_id(pos->long_name, + &b.build_id, + sizeof(b.build_id)) < 0) + continue; + have_buildid = true; + memset(&b.header, 0, sizeof(b.header)); + len = strlen(pos->long_name) + 1; + len = ALIGN(len, 64); + b.header.size = sizeof(b) + len; + do_write(fd, &b, sizeof(b)); + do_write(fd, pos->long_name, len); + } + + return have_buildid; +} + +static void +perf_header__adds_write(struct perf_header *self, int fd, bool at_exit) { struct perf_file_section trace_sec; u64 cur_offset = lseek(fd, 0, SEEK_CUR); @@ -196,9 +224,16 @@ static void perf_header__adds_write(struct perf_header *self, int fd) */ cur_offset = lseek(fd, trace_sec.offset + trace_sec.size, SEEK_SET); } + + if (at_exit) { + lseek(fd, self->data_offset + self->data_size, SEEK_SET); + if (write_buildid_table(fd)) + perf_header__set_feat(self, HEADER_BUILD_ID); + lseek(fd, cur_offset, SEEK_SET); + } }; -void perf_header__write(struct perf_header *self, int fd) +void perf_header__write(struct perf_header *self, int fd, bool at_exit) { struct perf_file_header f_header; struct perf_file_attr f_attr; @@ -236,7 +271,7 @@ void perf_header__write(struct perf_header *self, int fd) if (events) do_write(fd, events, self->event_size); - perf_header__adds_write(self, fd); + perf_header__adds_write(self, fd, at_exit); self->data_offset = lseek(fd, 0, SEEK_CUR); diff --git a/tools/perf/util/header.h b/tools/perf/util/header.h index 2f233c5db7e..77186c9e605 100644 --- a/tools/perf/util/header.h +++ b/tools/perf/util/header.h @@ -33,7 +33,7 @@ struct perf_header { }; struct perf_header *perf_header__read(int fd); -void perf_header__write(struct perf_header *self, int fd); +void perf_header__write(struct perf_header *self, int fd, bool at_exit); void perf_header__add_attr(struct perf_header *self, struct perf_header_attr *attr); -- cgit v1.2.3-18-g5258 From 57f395a7eabb913d3605d7392be5bdb0837c9f3d Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Wed, 11 Nov 2009 04:51:04 +0100 Subject: perf tools: Split up build id saving into fetch and write We are saving the build id once we stop the profiling. And only after doing that we know if we need to set that feature in the header through the feature bitmap. But if we want a proper feature support in the headers, using a rule of offset/size pairs in sections, we need to know in advance how many features we need to set in the headers, so that we can reserve rooms for their section headers. The current state doesn't allow that, as it forces us to first save the build-ids to the file right after the datas instead of planning any structured layout. That's why this splits up the build-ids processing in two parts: one that fetches the build-ids from the Dso objects, and one that saves them into the file. Signed-off-by: Frederic Weisbecker Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Mike Galbraith Cc: Paul Mackerras Cc: Hitoshi Mitake LKML-Reference: <1257911467-28276-3-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar --- tools/perf/util/event.h | 7 +++++++ tools/perf/util/header.c | 41 +++++++++++++++++------------------------ tools/perf/util/symbol.c | 34 ++++++++++++++++++++++++++++++++++ tools/perf/util/symbol.h | 1 + 4 files changed, 59 insertions(+), 24 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/event.h b/tools/perf/util/event.h index 34c6fcb82d9..1f771ce3a95 100644 --- a/tools/perf/util/event.h +++ b/tools/perf/util/event.h @@ -69,6 +69,13 @@ struct build_id_event { char filename[]; }; +struct build_id_list { + struct build_id_event event; + struct list_head list; + const char *dso_name; + int len; +}; + typedef union event_union { struct perf_event_header header; struct ip_event ip; diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index a4d0bbef9a4..2f702c23f71 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -174,29 +174,18 @@ static void do_write(int fd, void *buf, size_t size) } } -static bool write_buildid_table(int fd) +static void write_buildid_table(int fd, struct list_head *id_head) { - struct dso *pos; - bool have_buildid = false; - - list_for_each_entry(pos, &dsos, node) { - struct build_id_event b; - size_t len; - - if (filename__read_build_id(pos->long_name, - &b.build_id, - sizeof(b.build_id)) < 0) - continue; - have_buildid = true; - memset(&b.header, 0, sizeof(b.header)); - len = strlen(pos->long_name) + 1; - len = ALIGN(len, 64); - b.header.size = sizeof(b) + len; - do_write(fd, &b, sizeof(b)); - do_write(fd, pos->long_name, len); - } + struct build_id_list *iter, *next; + + list_for_each_entry_safe(iter, next, id_head, list) { + struct build_id_event *b = &iter->event; - return have_buildid; + do_write(fd, b, sizeof(*b)); + do_write(fd, (void *)iter->dso_name, iter->len); + list_del(&iter->list); + free(iter); + } } static void @@ -226,10 +215,14 @@ perf_header__adds_write(struct perf_header *self, int fd, bool at_exit) } if (at_exit) { - lseek(fd, self->data_offset + self->data_size, SEEK_SET); - if (write_buildid_table(fd)) + LIST_HEAD(id_list); + + if (fetch_build_id_table(&id_list)) { + lseek(fd, self->data_offset + self->data_size, SEEK_SET); perf_header__set_feat(self, HEADER_BUILD_ID); - lseek(fd, cur_offset, SEEK_SET); + write_buildid_table(fd, &id_list); + lseek(fd, cur_offset, SEEK_SET); + } } }; diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index a2e95ce1f22..9c286db6200 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -851,6 +851,40 @@ out_close: return err; } +bool fetch_build_id_table(struct list_head *head) +{ + bool have_buildid = false; + struct dso *pos; + + list_for_each_entry(pos, &dsos, node) { + struct build_id_list *new; + struct build_id_event b; + size_t len; + + if (filename__read_build_id(pos->long_name, + &b.build_id, + sizeof(b.build_id)) < 0) + continue; + have_buildid = true; + memset(&b.header, 0, sizeof(b.header)); + len = strlen(pos->long_name) + 1; + len = ALIGN(len, 64); + b.header.size = sizeof(b) + len; + + new = malloc(sizeof(*new)); + if (!new) + die("No memory\n"); + + memcpy(&new->event, &b, sizeof(b)); + new->dso_name = pos->long_name; + new->len = len; + + list_add_tail(&new->list, head); + } + + return have_buildid; +} + int filename__read_build_id(const char *filename, void *bf, size_t size) { int fd, err = -1; diff --git a/tools/perf/util/symbol.h b/tools/perf/util/symbol.h index f8c1899af48..0a34a5493f1 100644 --- a/tools/perf/util/symbol.h +++ b/tools/perf/util/symbol.h @@ -86,6 +86,7 @@ char dso__symtab_origin(const struct dso *self); void dso__set_build_id(struct dso *self, void *build_id); int filename__read_build_id(const char *filename, void *bf, size_t size); +bool fetch_build_id_table(struct list_head *head); int build_id__sprintf(u8 *self, int len, char *bf); int load_kernel(symbol_filter_t filter); -- cgit v1.2.3-18-g5258 From 4778d2e4f410c6eea32f594cb2be9590bcb28b84 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Wed, 11 Nov 2009 04:51:05 +0100 Subject: perf tools: Read the build-ids from the header layer Keep the build-ids reading implementation in the data mapping but move its call to the headers so that we have a better control on it (offset seeking, size passing, etc..). Signed-off-by: Frederic Weisbecker Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Mike Galbraith Cc: Paul Mackerras Cc: Hitoshi Mitake LKML-Reference: <1257911467-28276-4-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar --- tools/perf/util/data_map.c | 8 ++------ tools/perf/util/data_map.h | 2 ++ tools/perf/util/header.c | 14 ++++++++++++-- 3 files changed, 16 insertions(+), 8 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/data_map.c b/tools/perf/util/data_map.c index 00a9c114c8d..66e58aaecce 100644 --- a/tools/perf/util/data_map.c +++ b/tools/perf/util/data_map.c @@ -70,8 +70,8 @@ process_event(event_t *event, unsigned long offset, unsigned long head) } } -static int perf_header__read_build_ids(const struct perf_header *self, - int input, off_t file_size) +int perf_header__read_build_ids(const struct perf_header *self, + int input, off_t file_size) { off_t offset = self->data_offset + self->data_size; struct build_id_event bev; @@ -163,10 +163,6 @@ int mmap_dispatch_perf_file(struct perf_header **pheader, if (curr_handler->sample_type_check(sample_type) < 0) exit(-1); - if (perf_header__has_feat(header, HEADER_BUILD_ID) && - perf_header__read_build_ids(header, input, input_stat.st_size)) - pr_debug("failed to read buildids, continuing...\n"); - if (load_kernel(NULL) < 0) { perror("failed to load kernel symbols"); return EXIT_FAILURE; diff --git a/tools/perf/util/data_map.h b/tools/perf/util/data_map.h index 716d1053b07..c4122810e48 100644 --- a/tools/perf/util/data_map.h +++ b/tools/perf/util/data_map.h @@ -27,5 +27,7 @@ int mmap_dispatch_perf_file(struct perf_header **pheader, int full_paths, int *cwdlen, char **cwd); +int perf_header__read_build_ids(const struct perf_header *self, + int input, off_t file_size); #endif diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index 2f702c23f71..915b56edbf0 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -9,6 +9,8 @@ #include "../perf.h" #include "trace-event.h" #include "symbol.h" +#include "data_map.h" +#include "debug.h" /* * Create new perf.data header attribute: @@ -322,6 +324,14 @@ static void perf_header__adds_read(struct perf_header *self, int fd) trace_report(fd); lseek(fd, trace_sec.offset + trace_sec.size, SEEK_SET); } + + if (perf_header__has_feat(self, HEADER_BUILD_ID)) { + struct stat input_stat; + + fstat(fd, &input_stat); + if (perf_header__read_build_ids(self, fd, input_stat.st_size)) + pr_debug("failed to read buildids, continuing...\n"); + } }; struct perf_header *perf_header__read(int fd) @@ -382,14 +392,14 @@ struct perf_header *perf_header__read(int fd) memcpy(&self->adds_features, &f_header.adds_features, sizeof(f_header.adds_features)); - perf_header__adds_read(self, fd); - self->event_offset = f_header.event_types.offset; self->event_size = f_header.event_types.size; self->data_offset = f_header.data.offset; self->data_size = f_header.data.size; + perf_header__adds_read(self, fd); + lseek(fd, self->data_offset, SEEK_SET); self->frozen = 1; -- cgit v1.2.3-18-g5258 From 3e13ab2d83b6867a20663c73c184f29c2fde1558 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Wed, 11 Nov 2009 04:51:06 +0100 Subject: perf tools: Use perf_header__set/has_feat whenever possible And drop the alternate checks/sets using set_bit or other kind of helpers. Signed-off-by: Frederic Weisbecker Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Mike Galbraith Cc: Paul Mackerras Cc: Hitoshi Mitake LKML-Reference: <1257911467-28276-5-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar --- tools/perf/builtin-record.c | 4 ++-- tools/perf/util/header.c | 12 ++---------- tools/perf/util/header.h | 1 - 3 files changed, 4 insertions(+), 13 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index c35e61b3098..326e8a79cab 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -431,11 +431,11 @@ static int __cmd_record(int argc, const char **argv) header = perf_header__new(); if (raw_samples) { - perf_header__feat_trace_info(header); + perf_header__set_feat(header, HEADER_TRACE_INFO); } else { for (i = 0; i < nr_counters; i++) { if (attrs[i].sample_type & PERF_SAMPLE_RAW) { - perf_header__feat_trace_info(header); + perf_header__set_feat(header, HEADER_TRACE_INFO); break; } } diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index 915b56edbf0..9709d38113b 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -148,11 +148,6 @@ struct perf_file_header { DECLARE_BITMAP(adds_features, HEADER_FEAT_BITS); }; -void perf_header__feat_trace_info(struct perf_header *header) -{ - set_bit(HEADER_TRACE_INFO, header->adds_features); -} - void perf_header__set_feat(struct perf_header *self, int feat) { set_bit(feat, self->adds_features); @@ -195,9 +190,8 @@ perf_header__adds_write(struct perf_header *self, int fd, bool at_exit) { struct perf_file_section trace_sec; u64 cur_offset = lseek(fd, 0, SEEK_CUR); - unsigned long *feat_mask = self->adds_features; - if (test_bit(HEADER_TRACE_INFO, feat_mask)) { + if (perf_header__has_feat(self, HEADER_TRACE_INFO)) { /* Write trace info */ trace_sec.offset = lseek(fd, sizeof(trace_sec), SEEK_CUR); read_tracing_data(fd, attrs, nr_counters); @@ -314,9 +308,7 @@ static void do_read(int fd, void *buf, size_t size) static void perf_header__adds_read(struct perf_header *self, int fd) { - const unsigned long *feat_mask = self->adds_features; - - if (test_bit(HEADER_TRACE_INFO, feat_mask)) { + if (perf_header__has_feat(self, HEADER_TRACE_INFO)) { struct perf_file_section trace_sec; do_read(fd, &trace_sec, sizeof(trace_sec)); diff --git a/tools/perf/util/header.h b/tools/perf/util/header.h index 77186c9e605..a22d70b0757 100644 --- a/tools/perf/util/header.h +++ b/tools/perf/util/header.h @@ -49,7 +49,6 @@ void perf_header_attr__add_id(struct perf_header_attr *self, u64 id); u64 perf_header__sample_type(struct perf_header *header); struct perf_event_attr * perf_header__find_attr(u64 id, struct perf_header *header); -void perf_header__feat_trace_info(struct perf_header *header); void perf_header__set_feat(struct perf_header *self, int feat); bool perf_header__has_feat(const struct perf_header *self, int feat); -- cgit v1.2.3-18-g5258 From 9e827dd00a94136b944a538bede67c944d0b740a Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Wed, 11 Nov 2009 04:51:07 +0100 Subject: perf tools: Bring linear set of section headers for features Build a set of section headers for features right after the datas. Each implemented feature will have one of such section header that provides the offset and the size of the data manipulated by the feature. The trace informations have moved after the data and are recorded on exit time. The new layout is as follows: ----------------------- ___ [ magic ] | [ header size ] | [ attr size ] | [ attr content offset ] | [ attr content size ] | [ data offset ] File Headers [ data size ] | [ event_types offset ] | [ event_types size ] | [ feature bitmap ] v [ attr section ] [ events section ] ___ [ X ] | [ X ] | [ X ] Datas [ X ] | [ X ] v ___ [ Feature 1 offset ] | [ Feature 1 size ] Features headers [ Feature 2 offset ] | [ Feature 2 size ] v [ Feature 1 content ] [ Feature 2 content ] ----------------------- We have as many feature's section headers as we have features in use for the current file. Say Feat 1 and Feat 3 are used by the file, but not Feat 2. Then the feature headers will be like follows: [ Feature 1 offset ] | [ Feature 1 size ] Features headers [ Feature 3 offset ] | [ Feature 3 size ] v There is no hole to cover Feature 2 that is not in use here. We only need to cover the needed headers in order, from the lowest feature bit to the highest. Currently we have two features: HEADER_TRACE_INFO and HEADER_BUILD_ID. Both have their contents that follow the feature headers. Putting the contents right after the feature headers is not mandatory though. While we keep the feature headers right after the data and in order, their offsets can point everywhere. We have just put the two above feature contents in the end of the file for convenience. The purpose of this layout change is to have a file format that scales while keeping it simple: having such linear feature headers is less error prone wrt forward/backward compatibility as the content of a feature can be put anywhere, its location can even change by the time, it's fine because its headers will tell where it is. And we know how to find these headers, following the above rules. Signed-off-by: Frederic Weisbecker Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Mike Galbraith Cc: Paul Mackerras Cc: Hitoshi Mitake LKML-Reference: <1257911467-28276-6-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar --- tools/perf/util/data_map.c | 11 ++-- tools/perf/util/data_map.h | 3 +- tools/perf/util/header.c | 110 ++++++++++++++++++++++----------- tools/perf/util/include/linux/bitmap.h | 1 + tools/perf/util/include/linux/ctype.h | 2 +- tools/perf/util/util.h | 3 + 6 files changed, 85 insertions(+), 45 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/data_map.c b/tools/perf/util/data_map.c index 66e58aaecce..aacb814a4ef 100644 --- a/tools/perf/util/data_map.c +++ b/tools/perf/util/data_map.c @@ -70,18 +70,15 @@ process_event(event_t *event, unsigned long offset, unsigned long head) } } -int perf_header__read_build_ids(const struct perf_header *self, - int input, off_t file_size) +int perf_header__read_build_ids(int input, off_t size) { - off_t offset = self->data_offset + self->data_size; struct build_id_event bev; char filename[PATH_MAX]; + off_t offset = lseek(input, 0, SEEK_CUR); + off_t limit = offset + size; int err = -1; - if (lseek(input, offset, SEEK_SET) < 0) - return -1; - - while (offset < file_size) { + while (offset < limit) { struct dso *dso; ssize_t len; diff --git a/tools/perf/util/data_map.h b/tools/perf/util/data_map.h index c4122810e48..20b4037a823 100644 --- a/tools/perf/util/data_map.h +++ b/tools/perf/util/data_map.h @@ -27,7 +27,6 @@ int mmap_dispatch_perf_file(struct perf_header **pheader, int full_paths, int *cwdlen, char **cwd); -int perf_header__read_build_ids(const struct perf_header *self, - int input, off_t file_size); +int perf_header__read_build_ids(int input, off_t file_size); #endif diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index 9709d38113b..ebed4f44ed3 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -186,41 +186,58 @@ static void write_buildid_table(int fd, struct list_head *id_head) } static void -perf_header__adds_write(struct perf_header *self, int fd, bool at_exit) +perf_header__adds_write(struct perf_header *self, int fd) { - struct perf_file_section trace_sec; - u64 cur_offset = lseek(fd, 0, SEEK_CUR); + LIST_HEAD(id_list); + int nr_sections; + struct perf_file_section *feat_sec; + int sec_size; + u64 sec_start; + int idx = 0; + + if (fetch_build_id_table(&id_list)) + perf_header__set_feat(self, HEADER_BUILD_ID); + + nr_sections = bitmap_weight(self->adds_features, HEADER_FEAT_BITS); + if (!nr_sections) + return; + + feat_sec = calloc(sizeof(*feat_sec), nr_sections); + if (!feat_sec) + die("No memory"); + + sec_size = sizeof(*feat_sec) * nr_sections; + + sec_start = self->data_offset + self->data_size; + lseek(fd, sec_start + sec_size, SEEK_SET); if (perf_header__has_feat(self, HEADER_TRACE_INFO)) { + struct perf_file_section *trace_sec; + + trace_sec = &feat_sec[idx++]; + /* Write trace info */ - trace_sec.offset = lseek(fd, sizeof(trace_sec), SEEK_CUR); + trace_sec->offset = lseek(fd, 0, SEEK_CUR); read_tracing_data(fd, attrs, nr_counters); - trace_sec.size = lseek(fd, 0, SEEK_CUR) - trace_sec.offset; - - /* Write trace info headers */ - lseek(fd, cur_offset, SEEK_SET); - do_write(fd, &trace_sec, sizeof(trace_sec)); - - /* - * Update cur_offset. So that other (future) - * features can set their own infos in this place. But if we are - * the only feature, at least that seeks to the place the data - * should begin. - */ - cur_offset = lseek(fd, trace_sec.offset + trace_sec.size, SEEK_SET); + trace_sec->size = lseek(fd, 0, SEEK_CUR) - trace_sec->offset; } - if (at_exit) { - LIST_HEAD(id_list); - if (fetch_build_id_table(&id_list)) { - lseek(fd, self->data_offset + self->data_size, SEEK_SET); - perf_header__set_feat(self, HEADER_BUILD_ID); - write_buildid_table(fd, &id_list); - lseek(fd, cur_offset, SEEK_SET); - } + if (perf_header__has_feat(self, HEADER_BUILD_ID)) { + struct perf_file_section *buildid_sec; + + buildid_sec = &feat_sec[idx++]; + + /* Write build-ids */ + buildid_sec->offset = lseek(fd, 0, SEEK_CUR); + write_buildid_table(fd, &id_list); + buildid_sec->size = lseek(fd, 0, SEEK_CUR) - buildid_sec->offset; } -}; + + lseek(fd, sec_start, SEEK_SET); + do_write(fd, feat_sec, sec_size); + free(feat_sec); +} void perf_header__write(struct perf_header *self, int fd, bool at_exit) { @@ -260,10 +277,11 @@ void perf_header__write(struct perf_header *self, int fd, bool at_exit) if (events) do_write(fd, events, self->event_size); - perf_header__adds_write(self, fd, at_exit); - self->data_offset = lseek(fd, 0, SEEK_CUR); + if (at_exit) + perf_header__adds_write(self, fd); + f_header = (struct perf_file_header){ .magic = PERF_MAGIC, .size = sizeof(f_header), @@ -308,22 +326,44 @@ static void do_read(int fd, void *buf, size_t size) static void perf_header__adds_read(struct perf_header *self, int fd) { + struct perf_file_section *feat_sec; + int nr_sections; + int sec_size; + int idx = 0; + + + nr_sections = bitmap_weight(self->adds_features, HEADER_FEAT_BITS); + if (!nr_sections) + return; + + feat_sec = calloc(sizeof(*feat_sec), nr_sections); + if (!feat_sec) + die("No memory"); + + sec_size = sizeof(*feat_sec) * nr_sections; + + lseek(fd, self->data_offset + self->data_size, SEEK_SET); + + do_read(fd, feat_sec, sec_size); + if (perf_header__has_feat(self, HEADER_TRACE_INFO)) { - struct perf_file_section trace_sec; + struct perf_file_section *trace_sec; - do_read(fd, &trace_sec, sizeof(trace_sec)); - lseek(fd, trace_sec.offset, SEEK_SET); + trace_sec = &feat_sec[idx++]; + lseek(fd, trace_sec->offset, SEEK_SET); trace_report(fd); - lseek(fd, trace_sec.offset + trace_sec.size, SEEK_SET); } if (perf_header__has_feat(self, HEADER_BUILD_ID)) { - struct stat input_stat; + struct perf_file_section *buildid_sec; - fstat(fd, &input_stat); - if (perf_header__read_build_ids(self, fd, input_stat.st_size)) + buildid_sec = &feat_sec[idx++]; + lseek(fd, buildid_sec->offset, SEEK_SET); + if (perf_header__read_build_ids(fd, buildid_sec->size)) pr_debug("failed to read buildids, continuing...\n"); } + + free(feat_sec); }; struct perf_header *perf_header__read(int fd) diff --git a/tools/perf/util/include/linux/bitmap.h b/tools/perf/util/include/linux/bitmap.h index 821c1033bcc..94507639a8c 100644 --- a/tools/perf/util/include/linux/bitmap.h +++ b/tools/perf/util/include/linux/bitmap.h @@ -1,2 +1,3 @@ #include "../../../../include/linux/bitmap.h" #include "../../../../include/asm-generic/bitops/find.h" +#include diff --git a/tools/perf/util/include/linux/ctype.h b/tools/perf/util/include/linux/ctype.h index bae5783282e..a53d4ee1e0b 100644 --- a/tools/perf/util/include/linux/ctype.h +++ b/tools/perf/util/include/linux/ctype.h @@ -1 +1 @@ -#include "../../../../include/linux/ctype.h" +#include "../util.h" diff --git a/tools/perf/util/util.h b/tools/perf/util/util.h index 9de2329dd44..7bd5bdaeb23 100644 --- a/tools/perf/util/util.h +++ b/tools/perf/util/util.h @@ -306,6 +306,7 @@ static inline int has_extension(const char *filename, const char *ext) #undef isascii #undef isspace #undef isdigit +#undef isxdigit #undef isalpha #undef isprint #undef isalnum @@ -323,6 +324,8 @@ extern unsigned char sane_ctype[256]; #define isascii(x) (((x) & ~0x7f) == 0) #define isspace(x) sane_istest(x,GIT_SPACE) #define isdigit(x) sane_istest(x,GIT_DIGIT) +#define isxdigit(x) \ + (sane_istest(toupper(x), GIT_ALPHA | GIT_DIGIT) && toupper(x) < 'G') #define isalpha(x) sane_istest(x,GIT_ALPHA) #define isalnum(x) sane_istest(x,GIT_ALPHA | GIT_DIGIT) #define isprint(x) sane_istest(x,GIT_PRINT) -- cgit v1.2.3-18-g5258 From 5d7bdab75cd56d2bdc0986ae5546be3b09fea70a Mon Sep 17 00:00:00 2001 From: Michael Cree Date: Wed, 11 Nov 2009 20:43:03 +1300 Subject: perf tools: Test -fstack-protector-all compiler option for inclusion in CFLAGS Some architectures (e.g. Alpha) do not support the -fstack-protector-all compiler option and the use of the option with -Werror causes the compiler to abort and the build fails. Test that the compiler supports -fstack-protector-all before inclusion in CFLAGS. Signed-off-by: Michael Cree Cc: Richard Henderson Cc: Ivan Kokshaysky Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <20091111074302.GA3728@omega> Signed-off-by: Ingo Molnar --- tools/perf/Makefile | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'tools/perf') diff --git a/tools/perf/Makefile b/tools/perf/Makefile index b9509b1cc32..e6d42728a33 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -207,7 +207,7 @@ ifndef PERF_DEBUG CFLAGS_OPTIMIZE = -O6 endif -CFLAGS = $(MBITS) -ggdb3 -Wall -Wextra -std=gnu99 -Werror $(CFLAGS_OPTIMIZE) -fstack-protector-all -D_FORTIFY_SOURCE=2 $(EXTRA_WARNINGS) +CFLAGS = $(MBITS) -ggdb3 -Wall -Wextra -std=gnu99 -Werror $(CFLAGS_OPTIMIZE) -D_FORTIFY_SOURCE=2 $(EXTRA_WARNINGS) LDFLAGS = -lpthread -lrt -lelf -lm ALL_CFLAGS = $(CFLAGS) ALL_LDFLAGS = $(LDFLAGS) @@ -259,6 +259,9 @@ PTHREAD_LIBS = -lpthread # explicitly what architecture to check for. Fix this up for yours.. SPARSE_FLAGS = -D__BIG_ENDIAN__ -D__powerpc__ +ifeq ($(shell sh -c "echo 'int foo(void) {char X[2]; return 3;}' | $(CC) -x c -c -Werror -fstack-protector-all - -o /dev/null >/dev/null 2>&1 && echo y"), y) + CFLAGS := $(CFLAGS) -fstack-protector-all +endif ### --- END CONFIGURATION SECTION --- -- cgit v1.2.3-18-g5258 From d2fb8b4151a92223da6a84006f8f248ebeb6677d Mon Sep 17 00:00:00 2001 From: Hitoshi Mitake Date: Sun, 15 Nov 2009 20:36:53 +0900 Subject: perf tools: Add new perf_atoll() function to parse string representing size in bytes This patch modifies util/string.[ch] to add new function: perf_atoll() to parse string representing size in bytes. This function parses (\d+)(b|B|kb|KB|mb|MB|gb|GB) (e.g. "256MB") and returns its numeric value. (e.g. 268435456) Signed-off-by: Hitoshi Mitake Cc: Peter Zijlstra Cc: Paul Mackerras Cc: Frederic Weisbecker LKML-Reference: <1258285013-4759-1-git-send-email-mitake@dcl.info.waseda.ac.jp> Signed-off-by: Ingo Molnar --- tools/perf/util/string.c | 84 ++++++++++++++++++++++++++++++++++++++++++++++++ tools/perf/util/string.h | 1 + 2 files changed, 85 insertions(+) (limited to 'tools/perf') diff --git a/tools/perf/util/string.c b/tools/perf/util/string.c index 04743d3e903..227043577e0 100644 --- a/tools/perf/util/string.c +++ b/tools/perf/util/string.c @@ -1,5 +1,7 @@ #include +#include #include "string.h" +#include "util.h" static int hex(char ch) { @@ -43,3 +45,85 @@ char *strxfrchar(char *s, char from, char to) return s; } + +#define K 1024LL +/* + * perf_atoll() + * Parse (\d+)(b|B|kb|KB|mb|MB|gb|GB|tb|TB) (e.g. "256MB") + * and return its numeric value + */ +s64 perf_atoll(const char *str) +{ + unsigned int i; + s64 length = -1, unit = 1; + + if (!isdigit(str[0])) + goto out_err; + + for (i = 1; i < strlen(str); i++) { + switch (str[i]) { + case 'B': + case 'b': + break; + case 'K': + if (str[i + 1] != 'B') + goto out_err; + else + goto kilo; + case 'k': + if (str[i + 1] != 'b') + goto out_err; +kilo: + unit = K; + break; + case 'M': + if (str[i + 1] != 'B') + goto out_err; + else + goto mega; + case 'm': + if (str[i + 1] != 'b') + goto out_err; +mega: + unit = K * K; + break; + case 'G': + if (str[i + 1] != 'B') + goto out_err; + else + goto giga; + case 'g': + if (str[i + 1] != 'b') + goto out_err; +giga: + unit = K * K * K; + break; + case 'T': + if (str[i + 1] != 'B') + goto out_err; + else + goto tera; + case 't': + if (str[i + 1] != 'b') + goto out_err; +tera: + unit = K * K * K * K; + break; + case '\0': /* only specified figures */ + unit = 1; + break; + default: + if (!isdigit(str[i])) + goto out_err; + break; + } + } + + length = atoll(str) * unit; + goto out; + +out_err: + length = -1; +out: + return length; +} diff --git a/tools/perf/util/string.h b/tools/perf/util/string.h index 2c84bf65ba0..e50b07f8082 100644 --- a/tools/perf/util/string.h +++ b/tools/perf/util/string.h @@ -5,6 +5,7 @@ int hex2u64(const char *ptr, u64 *val); char *strxfrchar(char *s, char from, char to); +s64 perf_atoll(const char *str); #define _STR(x) #x #define STR(x) _STR(x) -- cgit v1.2.3-18-g5258 From 7255fe2a42c612f2b8fe4c347f0a5f0c97d85a46 Mon Sep 17 00:00:00 2001 From: Lucas De Marchi Date: Sun, 15 Nov 2009 12:05:08 -0200 Subject: perf stat: Do not print ratio when task-clock event is not counted The ratio between the number of events and the time elapsed makes sense only if task-clock event is counted. Otherwise it will be simply a (confusing) # 0.000 M/sec This patch outputs the ratio only if task-clock event is counted. Some test examples of before and after: Before: [lucas@skywalker linux.trees.git]$ sudo perf stat -e branch-misses -a -- sleep 1 Performance counter stats for 'sleep 1': 1367818 branch-misses # 0.000 M/sec 1.001494325 seconds time elapsed After (without task-clock): [lucas@skywalker perf]$ sudo ./perf stat -e branch-misses -a -- sleep 1 Performance counter stats for 'sleep 1': 1135044 branch-misses 1.001370775 seconds time elapsed After (with task-clock): [lucas@skywalker perf]$ sudo ./perf stat -e branch-misses -e task-clock -a -- sleep 1 Performance counter stats for 'sleep 1': 1070111 branch-misses # 0.534 M/sec 2002.730893 task-clock-msecs # 1.999 CPUs 1.001640292 seconds time elapsed Signed-off-by: Lucas De Marchi Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo LKML-Reference: <20091115140507.GB21561@skywalker.lan> Signed-off-by: Ingo Molnar --- tools/perf/builtin-stat.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c index c6df3770b87..c70d7200355 100644 --- a/tools/perf/builtin-stat.c +++ b/tools/perf/builtin-stat.c @@ -357,7 +357,8 @@ static void abs_printout(int counter, double avg) ratio = avg / total; fprintf(stderr, " # %10.3f IPC ", ratio); - } else if (MATCH_EVENT(HARDWARE, HW_BRANCH_MISSES, counter)) { + } else if (MATCH_EVENT(HARDWARE, HW_BRANCH_MISSES, counter) && + runtime_branches_stats.n != 0) { total = avg_stats(&runtime_branches_stats); if (total) @@ -365,7 +366,7 @@ static void abs_printout(int counter, double avg) fprintf(stderr, " # %10.3f %% ", ratio); - } else { + } else if (runtime_nsecs_stats.n != 0) { total = avg_stats(&runtime_nsecs_stats); if (total) -- cgit v1.2.3-18-g5258 From 82164161679c448f33092945ea97cb547a13683a Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 16 Nov 2009 13:48:11 -0200 Subject: perf symbols: Call the symbol filter in dso__synthesize_plt_symbols() We need to pass the symbol to the filter so that, for instance, 'perf top' can do filtering and also set the private area it manages, setting the ->map pointer, etc. I found this while running 'perf top' on a machine where hits happened on PLT symbols, where ->map wasn't being set up and segfaults thus happened. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra LKML-Reference: <1258386491-20278-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/util/symbol.c | 26 ++++++++++++++++++++------ 1 file changed, 20 insertions(+), 6 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 9c286db6200..93e4b52ccfe 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -281,6 +281,11 @@ static int kernel_maps__load_all_kallsyms(void) if (sym == NULL) goto out_delete_line; + /* + * We will pass the symbols to the filter later, in + * kernel_maps__split_kallsyms, when we have split the + * maps per module + */ dso__insert_symbol(kernel_map->dso, sym); } @@ -555,7 +560,8 @@ static Elf_Scn *elf_section_by_name(Elf *elf, GElf_Ehdr *ep, * And always look at the original dso, not at debuginfo packages, that * have the PLT data stripped out (shdr_rel_plt.sh_type == SHT_NOBITS). */ -static int dso__synthesize_plt_symbols(struct dso *self) +static int dso__synthesize_plt_symbols(struct dso *self, struct map *map, + symbol_filter_t filter) { uint32_t nr_rel_entries, idx; GElf_Sym sym; @@ -643,8 +649,12 @@ static int dso__synthesize_plt_symbols(struct dso *self) if (!f) goto out_elf_end; - dso__insert_symbol(self, f); - ++nr; + if (filter && filter(map, f)) + symbol__delete(f); + else { + dso__insert_symbol(self, f); + ++nr; + } } } else if (shdr_rel_plt.sh_type == SHT_REL) { GElf_Rel pos_mem, *pos; @@ -661,8 +671,12 @@ static int dso__synthesize_plt_symbols(struct dso *self) if (!f) goto out_elf_end; - dso__insert_symbol(self, f); - ++nr; + if (filter && filter(map, f)) + symbol__delete(f); + else { + dso__insert_symbol(self, f); + ++nr; + } } } @@ -1050,7 +1064,7 @@ compare_build_id: goto more; if (ret > 0) { - int nr_plt = dso__synthesize_plt_symbols(self); + int nr_plt = dso__synthesize_plt_symbols(self, map, filter); if (nr_plt > 0) ret += nr_plt; } -- cgit v1.2.3-18-g5258 From 84fe8488ade7922afa9f3aa77c22d2d92beb9660 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 16 Nov 2009 16:32:41 -0200 Subject: perf symbols: Pass the offset to perf_header__read_build_ids() Signed-off-by: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra LKML-Reference: <1258396365-29217-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/util/data_map.c | 3 +-- tools/perf/util/data_map.h | 2 +- tools/perf/util/header.c | 2 +- 3 files changed, 3 insertions(+), 4 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/data_map.c b/tools/perf/util/data_map.c index aacb814a4ef..14cb8465eb0 100644 --- a/tools/perf/util/data_map.c +++ b/tools/perf/util/data_map.c @@ -70,11 +70,10 @@ process_event(event_t *event, unsigned long offset, unsigned long head) } } -int perf_header__read_build_ids(int input, off_t size) +int perf_header__read_build_ids(int input, off_t offset, off_t size) { struct build_id_event bev; char filename[PATH_MAX]; - off_t offset = lseek(input, 0, SEEK_CUR); off_t limit = offset + size; int err = -1; diff --git a/tools/perf/util/data_map.h b/tools/perf/util/data_map.h index 20b4037a823..ae036ecd762 100644 --- a/tools/perf/util/data_map.h +++ b/tools/perf/util/data_map.h @@ -27,6 +27,6 @@ int mmap_dispatch_perf_file(struct perf_header **pheader, int full_paths, int *cwdlen, char **cwd); -int perf_header__read_build_ids(int input, off_t file_size); +int perf_header__read_build_ids(int input, off_t offset, off_t file_size); #endif diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index ebed4f44ed3..ca0d657eefa 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -359,7 +359,7 @@ static void perf_header__adds_read(struct perf_header *self, int fd) buildid_sec = &feat_sec[idx++]; lseek(fd, buildid_sec->offset, SEEK_SET); - if (perf_header__read_build_ids(fd, buildid_sec->size)) + if (perf_header__read_build_ids(fd, buildid_sec->offset, buildid_sec->size)) pr_debug("failed to read buildids, continuing...\n"); } -- cgit v1.2.3-18-g5258 From 8f41146aedf803856fb6477056e3960cb9ba8f9c Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 16 Nov 2009 16:32:42 -0200 Subject: perf tools: Debug.h needs to include event.h for event_t Signed-off-by: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra LKML-Reference: <1258396365-29217-2-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/util/debug.h | 2 ++ 1 file changed, 2 insertions(+) (limited to 'tools/perf') diff --git a/tools/perf/util/debug.h b/tools/perf/util/debug.h index e8b18a1f87a..c6c24c522de 100644 --- a/tools/perf/util/debug.h +++ b/tools/perf/util/debug.h @@ -2,6 +2,8 @@ #ifndef __PERF_DEBUG_H #define __PERF_DEBUG_H +#include "event.h" + extern int verbose; extern int dump_trace; -- cgit v1.2.3-18-g5258 From 37562eac3767c7f07bb1a1329708ff6453e47570 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 16 Nov 2009 16:32:43 -0200 Subject: perf tools: Generalize perf_header__adds_read() Renaming it to perf_header__process_sections() and passing a callback to handle each feature. The next changesets will introduce 'perf buildid-list' that will handle just the HEADER_BUILD_ID table, ignoring all the other features. Signed-off-by: Arnaldo Carvalho de Melo Acked-by: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra LKML-Reference: <1258396365-29217-3-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/util/header.c | 123 ++++++++++++++++++++++++++--------------------- tools/perf/util/header.h | 31 +++++++++++- 2 files changed, 98 insertions(+), 56 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index ca0d657eefa..d8416f01117 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -128,26 +128,11 @@ static const char *__perf_magic = "PERFFILE"; #define PERF_MAGIC (*(u64 *)__perf_magic) -struct perf_file_section { - u64 offset; - u64 size; -}; - struct perf_file_attr { struct perf_event_attr attr; struct perf_file_section ids; }; -struct perf_file_header { - u64 magic; - u64 size; - u64 attr_size; - struct perf_file_section attrs; - struct perf_file_section data; - struct perf_file_section event_types; - DECLARE_BITMAP(adds_features, HEADER_FEAT_BITS); -}; - void perf_header__set_feat(struct perf_header *self, int feat) { set_bit(feat, self->adds_features); @@ -324,21 +309,23 @@ static void do_read(int fd, void *buf, size_t size) } } -static void perf_header__adds_read(struct perf_header *self, int fd) +int perf_header__process_sections(struct perf_header *self, int fd, + int (*process)(struct perf_file_section *self, + int feat, int fd)) { struct perf_file_section *feat_sec; int nr_sections; int sec_size; int idx = 0; - + int err = 0, feat = 1; nr_sections = bitmap_weight(self->adds_features, HEADER_FEAT_BITS); if (!nr_sections) - return; + return 0; feat_sec = calloc(sizeof(*feat_sec), nr_sections); if (!feat_sec) - die("No memory"); + return -1; sec_size = sizeof(*feat_sec) * nr_sections; @@ -346,25 +333,73 @@ static void perf_header__adds_read(struct perf_header *self, int fd) do_read(fd, feat_sec, sec_size); - if (perf_header__has_feat(self, HEADER_TRACE_INFO)) { - struct perf_file_section *trace_sec; + while (idx < nr_sections && feat < HEADER_LAST_FEATURE) { + if (perf_header__has_feat(self, feat)) { + struct perf_file_section *sec = &feat_sec[idx++]; - trace_sec = &feat_sec[idx++]; - lseek(fd, trace_sec->offset, SEEK_SET); - trace_report(fd); + err = process(sec, feat, fd); + if (err < 0) + break; + } + ++feat; } - if (perf_header__has_feat(self, HEADER_BUILD_ID)) { - struct perf_file_section *buildid_sec; + free(feat_sec); + return err; +}; - buildid_sec = &feat_sec[idx++]; - lseek(fd, buildid_sec->offset, SEEK_SET); - if (perf_header__read_build_ids(fd, buildid_sec->offset, buildid_sec->size)) - pr_debug("failed to read buildids, continuing...\n"); +int perf_file_header__read(struct perf_file_header *self, + struct perf_header *ph, int fd) +{ + lseek(fd, 0, SEEK_SET); + do_read(fd, self, sizeof(*self)); + + if (self->magic != PERF_MAGIC || + self->attr_size != sizeof(struct perf_file_attr)) + return -1; + + if (self->size != sizeof(*self)) { + /* Support the previous format */ + if (self->size == offsetof(typeof(*self), adds_features)) + bitmap_zero(self->adds_features, HEADER_FEAT_BITS); + else + return -1; } - free(feat_sec); -}; + memcpy(&ph->adds_features, &self->adds_features, + sizeof(self->adds_features)); + + ph->event_offset = self->event_types.offset; + ph->event_size = self->event_types.size; + ph->data_offset = self->data.offset; + ph->data_size = self->data.size; + return 0; +} + +static int perf_file_section__process(struct perf_file_section *self, + int feat, int fd) +{ + if (lseek(fd, self->offset, SEEK_SET) < 0) { + pr_debug("Failed to lseek to %Ld offset for feature %d, " + "continuing...\n", self->offset, feat); + return 0; + } + + switch (feat) { + case HEADER_TRACE_INFO: + trace_report(fd); + break; + + case HEADER_BUILD_ID: + if (perf_header__read_build_ids(fd, self->offset, self->size)) + pr_debug("Failed to read buildids, continuing...\n"); + break; + default: + pr_debug("unknown feature %d, continuing...\n", feat); + } + + return 0; +} struct perf_header *perf_header__read(int fd) { @@ -372,23 +407,11 @@ struct perf_header *perf_header__read(int fd) struct perf_file_header f_header; struct perf_file_attr f_attr; u64 f_id; - int nr_attrs, nr_ids, i, j; - lseek(fd, 0, SEEK_SET); - do_read(fd, &f_header, sizeof(f_header)); - - if (f_header.magic != PERF_MAGIC || - f_header.attr_size != sizeof(f_attr)) + if (perf_file_header__read(&f_header, self, fd) < 0) die("incompatible file format"); - if (f_header.size != sizeof(f_header)) { - /* Support the previous format */ - if (f_header.size == offsetof(typeof(f_header), adds_features)) - bitmap_zero(f_header.adds_features, HEADER_FEAT_BITS); - else - die("incompatible file format"); - } nr_attrs = f_header.attrs.size / sizeof(f_attr); lseek(fd, f_header.attrs.offset, SEEK_SET); @@ -422,15 +445,7 @@ struct perf_header *perf_header__read(int fd) event_count = f_header.event_types.size / sizeof(struct perf_trace_event_type); } - memcpy(&self->adds_features, &f_header.adds_features, sizeof(f_header.adds_features)); - - self->event_offset = f_header.event_types.offset; - self->event_size = f_header.event_types.size; - - self->data_offset = f_header.data.offset; - self->data_size = f_header.data.size; - - perf_header__adds_read(self, fd); + perf_header__process_sections(self, fd, perf_file_section__process); lseek(fd, self->data_offset, SEEK_SET); diff --git a/tools/perf/util/header.h b/tools/perf/util/header.h index a22d70b0757..f1b3bf7bdfc 100644 --- a/tools/perf/util/header.h +++ b/tools/perf/util/header.h @@ -15,11 +15,34 @@ struct perf_header_attr { off_t id_offset; }; -#define HEADER_TRACE_INFO 1 -#define HEADER_BUILD_ID 2 +enum { + HEADER_TRACE_INFO = 1, + HEADER_BUILD_ID, + HEADER_LAST_FEATURE, +}; #define HEADER_FEAT_BITS 256 +struct perf_file_section { + u64 offset; + u64 size; +}; + +struct perf_file_header { + u64 magic; + u64 size; + u64 attr_size; + struct perf_file_section attrs; + struct perf_file_section data; + struct perf_file_section event_types; + DECLARE_BITMAP(adds_features, HEADER_FEAT_BITS); +}; + +struct perf_header; + +int perf_file_header__read(struct perf_file_header *self, + struct perf_header *ph, int fd); + struct perf_header { int frozen; int attrs, size; @@ -54,4 +77,8 @@ bool perf_header__has_feat(const struct perf_header *self, int feat); struct perf_header *perf_header__new(void); +int perf_header__process_sections(struct perf_header *self, int fd, + int (*process)(struct perf_file_section *self, + int feat, int fd)); + #endif /* __PERF_HEADER_H */ -- cgit v1.2.3-18-g5258 From 9e03eb2d512e7f3a1e562d4b922aa8b1891750b6 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 16 Nov 2009 16:32:44 -0200 Subject: perf tools: Introduce dsos__fprintf_buildid To print the buildids in the list of dsos. Will be used by 'perf buildid-list' Signed-off-by: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra LKML-Reference: <1258396365-29217-4-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/util/symbol.c | 30 ++++++++++++++++++++++++++---- tools/perf/util/symbol.h | 2 ++ 2 files changed, 28 insertions(+), 4 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 93e4b52ccfe..53de9c4488d 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -212,14 +212,21 @@ int build_id__sprintf(u8 *self, int len, char *bf) return raw - self; } -size_t dso__fprintf(struct dso *self, FILE *fp) +size_t dso__fprintf_buildid(struct dso *self, FILE *fp) { char sbuild_id[BUILD_ID_SIZE * 2 + 1]; - struct rb_node *nd; - size_t ret; build_id__sprintf(self->build_id, sizeof(self->build_id), sbuild_id); - ret = fprintf(fp, "dso: %s (%s)\n", self->short_name, sbuild_id); + return fprintf(fp, "%s", sbuild_id); +} + +size_t dso__fprintf(struct dso *self, FILE *fp) +{ + struct rb_node *nd; + size_t ret = fprintf(fp, "dso: %s (", self->short_name); + + ret += dso__fprintf_buildid(self, fp); + ret += fprintf(fp, ")\n"); for (nd = rb_first(&self->syms); nd; nd = rb_next(nd)) { struct symbol *pos = rb_entry(nd, struct symbol, rb_node); @@ -1428,6 +1435,21 @@ void dsos__fprintf(FILE *fp) dso__fprintf(pos, fp); } +size_t dsos__fprintf_buildid(FILE *fp) +{ + struct dso *pos; + size_t ret = 0; + + list_for_each_entry(pos, &dsos, node) { + ret += dso__fprintf_buildid(pos, fp); + if (verbose) + ret += fprintf(fp, " %s\n", pos->long_name); + else + ret += fprintf(fp, "\n"); + } + return ret; +} + int load_kernel(symbol_filter_t filter) { if (dsos__load_kernel(vmlinux_name, filter, modules) <= 0) diff --git a/tools/perf/util/symbol.h b/tools/perf/util/symbol.h index 0a34a5493f1..51c5a4a0813 100644 --- a/tools/perf/util/symbol.h +++ b/tools/perf/util/symbol.h @@ -80,7 +80,9 @@ int dsos__load_kernel(const char *vmlinux, symbol_filter_t filter, int modules); struct dso *dsos__findnew(const char *name); int dso__load(struct dso *self, struct map *map, symbol_filter_t filter); void dsos__fprintf(FILE *fp); +size_t dsos__fprintf_buildid(FILE *fp); +size_t dso__fprintf_buildid(struct dso *self, FILE *fp); size_t dso__fprintf(struct dso *self, FILE *fp); char dso__symtab_origin(const struct dso *self); void dso__set_build_id(struct dso *self, void *build_id); -- cgit v1.2.3-18-g5258 From c34984b2bbc77596c97c333539bffc90d2033178 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 16 Nov 2009 16:32:45 -0200 Subject: perf buildid-list: New plumbing command With this we can list the buildids in a perf.data file so that we can pipe them to other, distro specific tools that from the buildids can figure out separate packages (foo-debuginfo) where we can find the matching symtabs so that perf report can do its job. E.g: [acme@doppio linux-2.6-tip]$ perf buildid-list | head -5 8e08b117e5458ad3f85da16d42d0fc5cd21c5869 520c2387a587cc5acfcf881e27dba1caaeab4b1f ec8dd400904ddfcac8b1c343263a790f977159dc 7caedbca5a6d8ab39a7fe44bd28c07d3e14a3f3f 379bb828fd08859dbea73279f04abefabc95a6a3 [acme@doppio linux-2.6-tip]$ perf buildid-list -v | head -5 8e08b117e5458ad3f85da16d42d0fc5cd21c5869 /sbin/init 520c2387a587cc5acfcf881e27dba1caaeab4b1f /lib64/ld-2.10.1.so ec8dd400904ddfcac8b1c343263a790f977159dc /lib64/libc-2.10.1.so 7caedbca5a6d8ab39a7fe44bd28c07d3e14a3f3f /sbin/udevd 379bb828fd08859dbea73279f04abefabc95a6a3 /lib64/libdl-2.10.1.so [acme@doppio linux-2.6-tip]$ Signed-off-by: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra LKML-Reference: <1258396365-29217-5-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/Documentation/perf-buildid-list.txt | 34 ++++++++ tools/perf/Makefile | 1 + tools/perf/builtin-buildid-list.c | 116 +++++++++++++++++++++++++ tools/perf/builtin.h | 1 + tools/perf/command-list.txt | 1 + tools/perf/perf.c | 1 + 6 files changed, 154 insertions(+) create mode 100644 tools/perf/Documentation/perf-buildid-list.txt create mode 100644 tools/perf/builtin-buildid-list.c (limited to 'tools/perf') diff --git a/tools/perf/Documentation/perf-buildid-list.txt b/tools/perf/Documentation/perf-buildid-list.txt new file mode 100644 index 00000000000..abab34e0557 --- /dev/null +++ b/tools/perf/Documentation/perf-buildid-list.txt @@ -0,0 +1,34 @@ +perf-buildid-list(1) +==================== + +NAME +---- +perf-buildid-list - List the buildids in a perf.data file + +SYNOPSIS +-------- +[verse] +'perf buildid-list ' + +DESCRIPTION +----------- +This command displays the buildids found in a perf.data file, so that other +tools can be used to fetch packages with matching symbol tables for use by +perf report. + +OPTIONS +------- +-i:: +--input=:: + Input file name. (default: perf.data) +-f:: +--force:: + Don't do ownership validation. +-v:: +--verbose:: + Be more verbose, showing the name of the DSOs after the buildids. + +SEE ALSO +-------- +linkperf:perf-record[1], linkperf:perf-top[1], +linkperf:perf-report[1] diff --git a/tools/perf/Makefile b/tools/perf/Makefile index f7cd89622cf..46a58a81c9a 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -430,6 +430,7 @@ BUILTIN_OBJS += bench/sched-pipe.o BUILTIN_OBJS += builtin-help.o BUILTIN_OBJS += builtin-sched.o +BUILTIN_OBJS += builtin-buildid-list.o BUILTIN_OBJS += builtin-list.o BUILTIN_OBJS += builtin-record.o BUILTIN_OBJS += builtin-report.o diff --git a/tools/perf/builtin-buildid-list.c b/tools/perf/builtin-buildid-list.c new file mode 100644 index 00000000000..2e377e1be43 --- /dev/null +++ b/tools/perf/builtin-buildid-list.c @@ -0,0 +1,116 @@ +/* + * builtin-buildid-list.c + * + * Builtin buildid-list command: list buildids in perf.data + * + * Copyright (C) 2009, Red Hat Inc. + * Copyright (C) 2009, Arnaldo Carvalho de Melo + */ +#include "builtin.h" +#include "perf.h" +#include "util/cache.h" +#include "util/data_map.h" +#include "util/debug.h" +#include "util/header.h" +#include "util/parse-options.h" +#include "util/symbol.h" + +static char const *input_name = "perf.data"; +static int force; + +static const char *const buildid_list_usage[] = { + "perf report []", + NULL +}; + +static const struct option options[] = { + OPT_STRING('i', "input", &input_name, "file", + "input file name"), + OPT_BOOLEAN('f', "force", &force, "don't complain, do it"), + OPT_BOOLEAN('v', "verbose", &verbose, + "be more verbose (show counter open errors, etc)"), + OPT_END() +}; + +static int perf_file_section__process_buildids(struct perf_file_section *self, + int feat, int fd) +{ + if (feat != HEADER_BUILD_ID) + return 0; + + if (lseek(fd, self->offset, SEEK_SET) < 0) { + pr_warning("Failed to lseek to %Ld offset for buildids!\n", + self->offset); + return -1; + } + + if (perf_header__read_build_ids(fd, self->offset, self->size)) { + pr_warning("Failed to read buildids!\n"); + return -1; + } + + return 0; +} + +static int __cmd_buildid_list(void) +{ + int err = -1; + struct perf_header *header; + struct perf_file_header f_header; + struct stat input_stat; + int input = open(input_name, O_RDONLY); + + if (input < 0) { + pr_err("failed to open file: %s", input_name); + if (!strcmp(input_name, "perf.data")) + pr_err(" (try 'perf record' first)"); + pr_err("\n"); + goto out; + } + + err = fstat(input, &input_stat); + if (err < 0) { + perror("failed to stat file"); + goto out_close; + } + + if (!force && input_stat.st_uid && (input_stat.st_uid != geteuid())) { + pr_err("file %s not owned by current user or root\n", + input_name); + goto out_close; + } + + if (!input_stat.st_size) { + pr_info("zero-sized file, nothing to do!\n"); + goto out_close; + } + + err = -1; + header = perf_header__new(); + if (header == NULL) + goto out_close; + + if (perf_file_header__read(&f_header, header, input) < 0) { + pr_warning("incompatible file format"); + goto out_close; + } + + err = perf_header__process_sections(header, input, + perf_file_section__process_buildids); + + if (err < 0) + goto out_close; + + dsos__fprintf_buildid(stdout); +out_close: + close(input); +out: + return err; +} + +int cmd_buildid_list(int argc, const char **argv, const char *prefix __used) +{ + argc = parse_options(argc, argv, options, buildid_list_usage, 0); + setup_pager(); + return __cmd_buildid_list(); +} diff --git a/tools/perf/builtin.h b/tools/perf/builtin.h index f0cd5b139b7..e97954a0a3d 100644 --- a/tools/perf/builtin.h +++ b/tools/perf/builtin.h @@ -16,6 +16,7 @@ extern int check_pager_config(const char *cmd); extern int cmd_annotate(int argc, const char **argv, const char *prefix); extern int cmd_bench(int argc, const char **argv, const char *prefix); +extern int cmd_buildid_list(int argc, const char **argv, const char *prefix); extern int cmd_help(int argc, const char **argv, const char *prefix); extern int cmd_sched(int argc, const char **argv, const char *prefix); extern int cmd_list(int argc, const char **argv, const char *prefix); diff --git a/tools/perf/command-list.txt b/tools/perf/command-list.txt index 981c40b9a5e..d37b16cf18f 100644 --- a/tools/perf/command-list.txt +++ b/tools/perf/command-list.txt @@ -4,6 +4,7 @@ # perf-annotate mainporcelain common perf-bench mainporcelain common +perf-buildid-list mainporcelain common perf-list mainporcelain common perf-sched mainporcelain common perf-record mainporcelain common diff --git a/tools/perf/perf.c b/tools/perf/perf.c index 8936786b42e..53359ebb369 100644 --- a/tools/perf/perf.c +++ b/tools/perf/perf.c @@ -287,6 +287,7 @@ static void handle_internal_command(int argc, const char **argv) static struct cmd_struct commands[] = { { "help", cmd_help, 0 }, { "list", cmd_list, 0 }, + { "buildid-list", cmd_buildid_list, 0 }, { "record", cmd_record, 0 }, { "report", cmd_report, 0 }, { "bench", cmd_bench, 0 }, -- cgit v1.2.3-18-g5258 From dc79c0fc08a94b857aed446bfb47cdfde529400c Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 16 Nov 2009 19:30:26 -0200 Subject: perf tools: Don't die in perf_header_attr__new() We really should propagate such kinds of errors so that users of these library functions decide what to do in such cases instead of exiting in random places like now. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra LKML-Reference: <1258407027-384-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/builtin-record.c | 5 ++++- tools/perf/util/header.c | 22 ++++++++++++---------- tools/perf/util/header.h | 4 +--- 3 files changed, 17 insertions(+), 14 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 04f335ef9a8..4c03bb7a4eb 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -220,7 +220,8 @@ static struct perf_header_attr *get_header_attr(struct perf_event_attr *a, int n h_attr = header->attr[nr]; } else { h_attr = perf_header_attr__new(a); - perf_header__add_attr(header, h_attr); + if (h_attr != NULL) + perf_header__add_attr(header, h_attr); } return h_attr; @@ -308,6 +309,8 @@ try_again: } h_attr = get_header_attr(attr, counter); + if (h_attr == NULL) + die("nomem\n"); if (!file_new) { if (memcmp(&h_attr->attr, attr, sizeof(*attr))) { diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index d8416f01117..2f07a238ffd 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -19,16 +19,16 @@ struct perf_header_attr *perf_header_attr__new(struct perf_event_attr *attr) { struct perf_header_attr *self = malloc(sizeof(*self)); - if (!self) - die("nomem"); - - self->attr = *attr; - self->ids = 0; - self->size = 1; - self->id = malloc(sizeof(u64)); - - if (!self->id) - die("nomem"); + if (self != NULL) { + self->attr = *attr; + self->ids = 0; + self->size = 1; + self->id = malloc(sizeof(u64)); + if (self->id == NULL) { + free(self); + self = NULL; + } + } return self; } @@ -423,6 +423,8 @@ struct perf_header *perf_header__read(int fd) tmp = lseek(fd, 0, SEEK_CUR); attr = perf_header_attr__new(&f_attr.attr); + if (attr == NULL) + die("nomem"); nr_ids = f_attr.ids.size / sizeof(u64); lseek(fd, f_attr.ids.offset, SEEK_SET); diff --git a/tools/perf/util/header.h b/tools/perf/util/header.h index f1b3bf7bdfc..0cbd4c9e982 100644 --- a/tools/perf/util/header.h +++ b/tools/perf/util/header.h @@ -64,9 +64,7 @@ void perf_header__add_attr(struct perf_header *self, void perf_header__push_event(u64 id, const char *name); char *perf_header__find_event(u64 id); - -struct perf_header_attr * -perf_header_attr__new(struct perf_event_attr *attr); +struct perf_header_attr *perf_header_attr__new(struct perf_event_attr *attr); void perf_header_attr__add_id(struct perf_header_attr *self, u64 id); u64 perf_header__sample_type(struct perf_header *header); -- cgit v1.2.3-18-g5258 From 3b6ed98895b0fccd8c387f3fc44016fb922c0658 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 16 Nov 2009 19:30:27 -0200 Subject: perf top: Use all the lines in the screen By querying the current number of rows, if the user specifies the number of entries, use that instead. If the user uses the 'e' command to change the number of lines 0 will mean do it automatically, any other number disables the auto resizing. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra LKML-Reference: <1258407027-384-2-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/builtin-top.c | 42 +++++++++++++++++++++++++++++++++++++++++- 1 file changed, 41 insertions(+), 1 deletion(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index 6613f988a33..3af95203208 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -60,7 +60,7 @@ static int system_wide = 0; static int default_interval = 0; static int count_filter = 5; -static int print_entries = 15; +static int print_entries; static int target_pid = -1; static int inherit = 0; @@ -115,6 +115,36 @@ struct sym_entry { * Source functions */ +/* most GUI terminals set LINES (although some don't export it) */ +static int term_rows(void) +{ + char *lines_string = getenv("LINES"); + int n_lines; + + if (lines_string && (n_lines = atoi(lines_string)) > 0) + return n_lines; +#ifdef TIOCGWINSZ + else { + struct winsize ws; + if (!ioctl(1, TIOCGWINSZ, &ws) && ws.ws_row) + return ws.ws_row; + } +#endif + return 25; +} + +static void update_print_entries(void) +{ + print_entries = term_rows(); + if (print_entries > 9) + print_entries -= 9; +} + +static void sig_winch_handler(int sig __used) +{ + update_print_entries(); +} + static void parse_source(struct sym_entry *syme) { struct symbol *sym; @@ -668,6 +698,11 @@ static void handle_keypress(int c) break; case 'e': prompt_integer(&print_entries, "Enter display entries (lines)"); + if (print_entries == 0) { + update_print_entries(); + signal(SIGWINCH, sig_winch_handler); + } else + signal(SIGWINCH, SIG_DFL); break; case 'E': if (nr_counters > 1) { @@ -1228,5 +1263,10 @@ int cmd_top(int argc, const char **argv, const char *prefix __used) if (target_pid != -1 || profile_cpu != -1) nr_cpus = 1; + if (print_entries == 0) { + update_print_entries(); + signal(SIGWINCH, sig_winch_handler); + } + return __cmd_top(); } -- cgit v1.2.3-18-g5258 From 8ffcda17314cfeb698a667567ea63f63362dffbb Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 16 Nov 2009 21:45:24 -0200 Subject: perf top: Introduce --hide_{user,kernel}_symbols Default continues to be showing all symbols. 'K' and 'U' can be used to toggle showing kernel and user symbols. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra LKML-Reference: <1258415125-15019-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/builtin-top.c | 42 ++++++++++++++++++++++++++++++++++++++++-- 1 file changed, 40 insertions(+), 2 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index 3af95203208..89b7f68a179 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -76,6 +76,9 @@ static int delay_secs = 2; static int zero = 0; static int dump_symtab = 0; +static bool hide_kernel_symbols = false; +static bool hide_user_symbols = false; + /* * Source */ @@ -104,6 +107,7 @@ struct sym_entry { unsigned long snap_count; double weight; int skip; + u8 origin; struct map *map; struct source_line *source; struct source_line *lines; @@ -430,6 +434,13 @@ static void print_sym_table(void) list_for_each_entry_safe_from(syme, n, &active_symbols, node) { syme->snap_count = syme->count[snap]; if (syme->snap_count != 0) { + if ((hide_user_symbols && + syme->origin == PERF_RECORD_MISC_USER) || + (hide_kernel_symbols && + syme->origin == PERF_RECORD_MISC_KERNEL)) { + list_remove_active_sym(syme); + continue; + } syme->weight = sym_weight(syme); rb_insert_active_sym(&tmp, syme); sum_ksamples += syme->snap_count; @@ -637,6 +648,12 @@ static void print_mapped_keys(void) if (nr_counters > 1) fprintf(stdout, "\t[w] toggle display weighted/count[E]r. \t(%d)\n", display_weighted ? 1 : 0); + fprintf(stdout, + "\t[K] hide kernel_symbols symbols. \t(%s)\n", + hide_kernel_symbols ? "yes" : "no"); + fprintf(stdout, + "\t[U] hide user symbols. \t(%s)\n", + hide_user_symbols ? "yes" : "no"); fprintf(stdout, "\t[z] toggle sample zeroing. \t(%d)\n", zero ? 1 : 0); fprintf(stdout, "\t[qQ] quit.\n"); } @@ -650,6 +667,8 @@ static int key_mapped(int c) case 'z': case 'q': case 'Q': + case 'K': + case 'U': return 1; case 'E': case 'w': @@ -727,6 +746,9 @@ static void handle_keypress(int c) case 'F': prompt_percent(&sym_pcnt_filter, "Enter details display event filter (percent)"); break; + case 'K': + hide_kernel_symbols = !hide_kernel_symbols; + break; case 'q': case 'Q': printf("exiting.\n"); @@ -746,6 +768,9 @@ static void handle_keypress(int c) pthread_mutex_unlock(&syme->source_lock); } break; + case 'U': + hide_user_symbols = !hide_user_symbols; + break; case 'w': display_weighted = ~display_weighted; break; @@ -857,11 +882,16 @@ static void event__process_sample(const event_t *self, int counter) struct map *map; struct sym_entry *syme; struct symbol *sym; + u8 origin = self->header.misc & PERF_RECORD_MISC_CPUMODE_MASK; - switch (self->header.misc & PERF_RECORD_MISC_CPUMODE_MASK) { + switch (origin) { case PERF_RECORD_MISC_USER: { - struct thread *thread = threads__findnew(self->ip.pid); + struct thread *thread; + if (hide_user_symbols) + return; + + thread = threads__findnew(self->ip.pid); if (thread == NULL) return; @@ -885,6 +915,9 @@ static void event__process_sample(const event_t *self, int counter) return; /* Fall thru */ case PERF_RECORD_MISC_KERNEL: + if (hide_kernel_symbols) + return; + sym = kernel_maps__find_symbol(ip, &map); if (sym == NULL) return; @@ -897,6 +930,7 @@ static void event__process_sample(const event_t *self, int counter) if (!syme->skip) { syme->count[counter]++; + syme->origin = origin; record_precise_ip(syme, counter, ip); pthread_mutex_lock(&active_symbols_lock); if (list_empty(&syme->node) || !syme->node.next) @@ -1178,6 +1212,8 @@ static const struct option options[] = { OPT_INTEGER('C', "CPU", &profile_cpu, "CPU to profile on"), OPT_STRING('k', "vmlinux", &vmlinux_name, "file", "vmlinux pathname"), + OPT_BOOLEAN('K', "hide_kernel_symbols", &hide_kernel_symbols, + "hide kernel symbols"), OPT_INTEGER('m', "mmap-pages", &mmap_pages, "number of mmap data pages"), OPT_INTEGER('r', "realtime", &realtime_prio, @@ -1200,6 +1236,8 @@ static const struct option options[] = { "profile at this frequency"), OPT_INTEGER('E', "entries", &print_entries, "display this many functions"), + OPT_BOOLEAN('U', "hide_user_symbols", &hide_user_symbols, + "hide user symbols"), OPT_BOOLEAN('v', "verbose", &verbose, "be more verbose (show counter open errors, etc)"), OPT_END() -- cgit v1.2.3-18-g5258 From 1124ba73be6a758965340bd997593b2996649d60 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 16 Nov 2009 21:45:25 -0200 Subject: perf buildid-list: Always show the DSO name Porcelain can ignore it, humans can make more sense of it. Suggested-by: Frederic Weisbecker Suggested-by: Ingo Molnar Suggested-by: Peter Zijlstra Signed-off-by: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra LKML-Reference: <1258415125-15019-2-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/Documentation/perf-buildid-list.txt | 2 +- tools/perf/builtin-buildid-list.c | 2 +- tools/perf/util/symbol.c | 5 +---- 3 files changed, 3 insertions(+), 6 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/Documentation/perf-buildid-list.txt b/tools/perf/Documentation/perf-buildid-list.txt index abab34e0557..01b642c0bf8 100644 --- a/tools/perf/Documentation/perf-buildid-list.txt +++ b/tools/perf/Documentation/perf-buildid-list.txt @@ -26,7 +26,7 @@ OPTIONS Don't do ownership validation. -v:: --verbose:: - Be more verbose, showing the name of the DSOs after the buildids. + Be more verbose. SEE ALSO -------- diff --git a/tools/perf/builtin-buildid-list.c b/tools/perf/builtin-buildid-list.c index 2e377e1be43..7dee9d19ab7 100644 --- a/tools/perf/builtin-buildid-list.c +++ b/tools/perf/builtin-buildid-list.c @@ -28,7 +28,7 @@ static const struct option options[] = { "input file name"), OPT_BOOLEAN('f', "force", &force, "don't complain, do it"), OPT_BOOLEAN('v', "verbose", &verbose, - "be more verbose (show counter open errors, etc)"), + "be more verbose"), OPT_END() }; diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 53de9c4488d..1b77e81b38d 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -1442,10 +1442,7 @@ size_t dsos__fprintf_buildid(FILE *fp) list_for_each_entry(pos, &dsos, node) { ret += dso__fprintf_buildid(pos, fp); - if (verbose) - ret += fprintf(fp, " %s\n", pos->long_name); - else - ret += fprintf(fp, "\n"); + ret += fprintf(fp, " %s\n", pos->long_name); } return ret; } -- cgit v1.2.3-18-g5258 From 11deb1f9f6ca6318fa9470e024b9f0634df48b4c Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Tue, 17 Nov 2009 01:18:09 -0200 Subject: perf tools: Don't die() in perf_header__add_attr() Propagate the errors instead, the users are the ones to decide what to do if a library call fails. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra LKML-Reference: <1258427892-16312-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/builtin-record.c | 5 ++++- tools/perf/util/header.c | 30 ++++++++++++++++++++++-------- tools/perf/util/header.h | 6 ++++-- 3 files changed, 30 insertions(+), 11 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 4c03bb7a4eb..5411be4cfd7 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -221,7 +221,10 @@ static struct perf_header_attr *get_header_attr(struct perf_event_attr *a, int n } else { h_attr = perf_header_attr__new(a); if (h_attr != NULL) - perf_header__add_attr(header, h_attr); + if (perf_header__add_attr(header, h_attr) < 0) { + perf_header_attr__delete(h_attr); + h_attr = NULL; + } } return h_attr; diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index 2f07a238ffd..23ccddae0b0 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -33,6 +33,12 @@ struct perf_header_attr *perf_header_attr__new(struct perf_event_attr *attr) return self; } +void perf_header_attr__delete(struct perf_header_attr *self) +{ + free(self->id); + free(self); +} + void perf_header_attr__add_id(struct perf_header_attr *self, u64 id) { int pos = self->ids; @@ -66,22 +72,28 @@ struct perf_header *perf_header__new(void) return self; } -void perf_header__add_attr(struct perf_header *self, - struct perf_header_attr *attr) +int perf_header__add_attr(struct perf_header *self, + struct perf_header_attr *attr) { int pos = self->attrs; if (self->frozen) - die("frozen"); + return -1; self->attrs++; if (self->attrs > self->size) { - self->size *= 2; - self->attr = realloc(self->attr, self->size * sizeof(void *)); - if (!self->attr) - die("nomem"); + int nsize = self->size * 2; + struct perf_header_attr **nattr; + + nattr = realloc(self->attr, nsize * sizeof(void *)); + if (nattr == NULL) + return -1; + + self->size = nsize; + self->attr = nattr; } self->attr[pos] = attr; + return 0; } #define MAX_EVENT_NAME 64 @@ -434,7 +446,9 @@ struct perf_header *perf_header__read(int fd) perf_header_attr__add_id(attr, f_id); } - perf_header__add_attr(self, attr); + if (perf_header__add_attr(self, attr) < 0) + die("nomem"); + lseek(fd, tmp, SEEK_SET); } diff --git a/tools/perf/util/header.h b/tools/perf/util/header.h index 0cbd4c9e982..b0d5cd707a7 100644 --- a/tools/perf/util/header.h +++ b/tools/perf/util/header.h @@ -58,13 +58,15 @@ struct perf_header { struct perf_header *perf_header__read(int fd); void perf_header__write(struct perf_header *self, int fd, bool at_exit); -void perf_header__add_attr(struct perf_header *self, - struct perf_header_attr *attr); +int perf_header__add_attr(struct perf_header *self, + struct perf_header_attr *attr); void perf_header__push_event(u64 id, const char *name); char *perf_header__find_event(u64 id); struct perf_header_attr *perf_header_attr__new(struct perf_event_attr *attr); +void perf_header_attr__delete(struct perf_header_attr *self); + void perf_header_attr__add_id(struct perf_header_attr *self, u64 id); u64 perf_header__sample_type(struct perf_header *header); -- cgit v1.2.3-18-g5258 From 5875412152ce67fb5087157b86ab6597f91d23e8 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Tue, 17 Nov 2009 01:18:10 -0200 Subject: perf tools: Don't die() in perf_header_attr__add_id() Propagate the errors instead, the users are the ones to decide what to do if a library call fails. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra LKML-Reference: <1258427892-16312-2-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/builtin-record.c | 5 ++++- tools/perf/util/header.c | 18 ++++++++++++------ tools/perf/util/header.h | 2 +- 3 files changed, 17 insertions(+), 8 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 5411be4cfd7..2a85205ba01 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -327,7 +327,10 @@ try_again: exit(-1); } - perf_header_attr__add_id(h_attr, read_data.id); + if (perf_header_attr__add_id(h_attr, read_data.id) < 0) { + pr_warning("Not enough memory to add id\n"); + exit(-1); + } assert(fd[nr_cpu][counter] >= 0); fcntl(fd[nr_cpu][counter], F_SETFL, O_NONBLOCK); diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index 23ccddae0b0..dee1ed2f0d1 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -39,18 +39,23 @@ void perf_header_attr__delete(struct perf_header_attr *self) free(self); } -void perf_header_attr__add_id(struct perf_header_attr *self, u64 id) +int perf_header_attr__add_id(struct perf_header_attr *self, u64 id) { int pos = self->ids; self->ids++; if (self->ids > self->size) { - self->size *= 2; - self->id = realloc(self->id, self->size * sizeof(u64)); - if (!self->id) - die("nomem"); + int nsize = self->size * 2; + u64 *nid = realloc(self->id, nsize * sizeof(u64)); + + if (nid == NULL) + return -1; + + self->size = nsize; + self->id = nid; } self->id[pos] = id; + return 0; } /* @@ -444,7 +449,8 @@ struct perf_header *perf_header__read(int fd) for (j = 0; j < nr_ids; j++) { do_read(fd, &f_id, sizeof(f_id)); - perf_header_attr__add_id(attr, f_id); + if (perf_header_attr__add_id(attr, f_id) < 0) + die("nomem"); } if (perf_header__add_attr(self, attr) < 0) die("nomem"); diff --git a/tools/perf/util/header.h b/tools/perf/util/header.h index b0d5cd707a7..f46a94e09ee 100644 --- a/tools/perf/util/header.h +++ b/tools/perf/util/header.h @@ -67,7 +67,7 @@ char *perf_header__find_event(u64 id); struct perf_header_attr *perf_header_attr__new(struct perf_event_attr *attr); void perf_header_attr__delete(struct perf_header_attr *self); -void perf_header_attr__add_id(struct perf_header_attr *self, u64 id); +int perf_header_attr__add_id(struct perf_header_attr *self, u64 id); u64 perf_header__sample_type(struct perf_header *header); struct perf_event_attr * -- cgit v1.2.3-18-g5258 From a9a70bbce7ab0bf3b1cba3ac662c4d502da6305c Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Tue, 17 Nov 2009 01:18:11 -0200 Subject: perf tools: Don't die() in perf_header__new() Propagate the errors instead, the users are the ones to decide what to do if a library call fails. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra LKML-Reference: <1258427892-16312-3-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/builtin-record.c | 5 +++++ tools/perf/util/header.c | 18 +++++++++++------- 2 files changed, 16 insertions(+), 7 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 2a85205ba01..82260c56db3 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -439,6 +439,11 @@ static int __cmd_record(int argc, const char **argv) else header = perf_header__new(); + if (header == NULL) { + pr_err("Not enough memory for reading perf file header\n"); + return -1; + } + if (raw_samples) { perf_header__set_feat(header, HEADER_TRACE_INFO); } else { diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index dee1ed2f0d1..726a0eb5f19 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -65,14 +65,15 @@ struct perf_header *perf_header__new(void) { struct perf_header *self = calloc(sizeof(*self), 1); - if (!self) - die("nomem"); - - self->size = 1; - self->attr = malloc(sizeof(void *)); + if (self != NULL) { + self->size = 1; + self->attr = malloc(sizeof(void *)); - if (!self->attr) - die("nomem"); + if (self->attr == NULL) { + free(self); + self = NULL; + } + } return self; } @@ -426,6 +427,9 @@ struct perf_header *perf_header__read(int fd) u64 f_id; int nr_attrs, nr_ids, i, j; + if (self == NULL) + die("nomem"); + if (perf_file_header__read(&f_header, self, fd) < 0) die("incompatible file format"); -- cgit v1.2.3-18-g5258 From 3726cc75e581c157202da93bb2333cce25c15c98 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Tue, 17 Nov 2009 01:18:12 -0200 Subject: perf tools: Don't die() in do_write() Propagate the errors instead, the users are the ones to decide what to do if a library call fails. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra LKML-Reference: <1258427892-16312-4-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/util/header.c | 33 ++++++++++++++++++++++----------- 1 file changed, 22 insertions(+), 11 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index 726a0eb5f19..b01a9537977 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -161,31 +161,36 @@ bool perf_header__has_feat(const struct perf_header *self, int feat) return test_bit(feat, self->adds_features); } -static void do_write(int fd, void *buf, size_t size) +static int do_write(int fd, const void *buf, size_t size) { while (size) { int ret = write(fd, buf, size); if (ret < 0) - die("failed to write"); + return -1; size -= ret; buf += ret; } + + return 0; } -static void write_buildid_table(int fd, struct list_head *id_head) +static int write_buildid_table(int fd, struct list_head *id_head) { struct build_id_list *iter, *next; list_for_each_entry_safe(iter, next, id_head, list) { struct build_id_event *b = &iter->event; - do_write(fd, b, sizeof(*b)); - do_write(fd, (void *)iter->dso_name, iter->len); + if (do_write(fd, b, sizeof(*b)) < 0 || + do_write(fd, iter->dso_name, iter->len) < 0) + return -1; list_del(&iter->list); free(iter); } + + return 0; } static void @@ -233,12 +238,14 @@ perf_header__adds_write(struct perf_header *self, int fd) /* Write build-ids */ buildid_sec->offset = lseek(fd, 0, SEEK_CUR); - write_buildid_table(fd, &id_list); + if (write_buildid_table(fd, &id_list) < 0) + die("failed to write buildid table"); buildid_sec->size = lseek(fd, 0, SEEK_CUR) - buildid_sec->offset; } lseek(fd, sec_start, SEEK_SET); - do_write(fd, feat_sec, sec_size); + if (do_write(fd, feat_sec, sec_size) < 0) + die("failed to write feature section"); free(feat_sec); } @@ -256,7 +263,8 @@ void perf_header__write(struct perf_header *self, int fd, bool at_exit) attr = self->attr[i]; attr->id_offset = lseek(fd, 0, SEEK_CUR); - do_write(fd, attr->id, attr->ids * sizeof(u64)); + if (do_write(fd, attr->id, attr->ids * sizeof(u64)) < 0) + die("failed to write perf header"); } @@ -272,13 +280,15 @@ void perf_header__write(struct perf_header *self, int fd, bool at_exit) .size = attr->ids * sizeof(u64), } }; - do_write(fd, &f_attr, sizeof(f_attr)); + if (do_write(fd, &f_attr, sizeof(f_attr)) < 0) + die("failed to write perf header attribute"); } self->event_offset = lseek(fd, 0, SEEK_CUR); self->event_size = event_count * sizeof(struct perf_trace_event_type); if (events) - do_write(fd, events, self->event_size); + if (do_write(fd, events, self->event_size) < 0) + die("failed to write perf header events"); self->data_offset = lseek(fd, 0, SEEK_CUR); @@ -306,7 +316,8 @@ void perf_header__write(struct perf_header *self, int fd, bool at_exit) memcpy(&f_header.adds_features, &self->adds_features, sizeof(self->adds_features)); lseek(fd, 0, SEEK_SET); - do_write(fd, &f_header, sizeof(f_header)); + if (do_write(fd, &f_header, sizeof(f_header)) < 0) + die("failed to write perf header"); lseek(fd, self->data_offset + self->data_size, SEEK_SET); self->frozen = 1; -- cgit v1.2.3-18-g5258 From 751386507701010831d72c522171753d2cd903d2 Mon Sep 17 00:00:00 2001 From: "Michael S. Tsirkin" Date: Thu, 29 Oct 2009 17:20:02 +0200 Subject: perf tools: Support static build This makes it possible to build perf statically, by performing: make LDFLAGS=-static Since static libraries are only searched in the order they are specified, move library list from LDFLAGS to EXTLIBS, so that they are put at the end of linker command line. Signed-off-by: Michael S. Tsirkin Cc: Peter Zijlstra Cc: Paul Mackerras Cc: Frederic Weisbecker Cc: Arnaldo Carvalho de Melo LKML-Reference: <20091029152002.GA5406@redhat.com> [ v2: resolved conflicts ] Signed-off-by: Ingo Molnar --- tools/perf/Makefile | 20 +++++++++++--------- 1 file changed, 11 insertions(+), 9 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/Makefile b/tools/perf/Makefile index 3dbb5c5bb8c..5d1a8b0dff8 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -145,6 +145,8 @@ all:: # Define NO_EXTERNAL_GREP if you don't want "perf grep" to ever call # your external grep (e.g., if your system lacks grep, if its grep is # broken, or spawning external process is slower than built-in grep perf has). +# +# Define LDFLAGS=-static to build a static binary. PERF-VERSION-FILE: .FORCE-PERF-VERSION-FILE @$(SHELL_PATH) util/PERF-VERSION-GEN @@ -208,7 +210,7 @@ ifndef PERF_DEBUG endif CFLAGS = $(MBITS) -ggdb3 -Wall -Wextra -std=gnu99 -Werror $(CFLAGS_OPTIMIZE) -D_FORTIFY_SOURCE=2 $(EXTRA_WARNINGS) -LDFLAGS = -lpthread -lrt -lelf -lm +EXTLIBS = -lpthread -lrt -lelf -lm ALL_CFLAGS = $(CFLAGS) ALL_LDFLAGS = $(LDFLAGS) STRIP ?= strip @@ -470,19 +472,19 @@ ifeq ($(uname_S),Darwin) PTHREAD_LIBS = endif -ifeq ($(shell sh -c "(echo '\#include '; echo 'int main(void) { Elf * elf = elf_begin(0, ELF_C_READ, 0); return (long)elf; }') | $(CC) -x c - $(ALL_CFLAGS) -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -o /dev/null $(ALL_LDFLAGS) > /dev/null 2>&1 && echo y"), y) -ifneq ($(shell sh -c "(echo '\#include '; echo 'int main(void) { const char * version = gnu_get_libc_version(); return (long)version; }') | $(CC) -x c - $(ALL_CFLAGS) -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -o /dev/null $(ALL_LDFLAGS) > /dev/null 2>&1 && echo y"), y) +ifeq ($(shell sh -c "(echo '\#include '; echo 'int main(void) { Elf * elf = elf_begin(0, ELF_C_READ, 0); return (long)elf; }') | $(CC) -x c - $(ALL_CFLAGS) -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -o /dev/null $(ALL_LDFLAGS) $(EXTLIBS) > /dev/null 2>&1 && echo y"), y) +ifneq ($(shell sh -c "(echo '\#include '; echo 'int main(void) { const char * version = gnu_get_libc_version(); return (long)version; }') | $(CC) -x c - $(ALL_CFLAGS) -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -o /dev/null $(ALL_LDFLAGS) $(EXTLIBS) > /dev/null 2>&1 && echo y"), y) msg := $(error No gnu/libc-version.h found, please install glibc-dev[el]); endif - ifneq ($(shell sh -c "(echo '\#include '; echo 'int main(void) { Elf * elf = elf_begin(0, ELF_C_READ_MMAP, 0); return (long)elf; }') | $(CC) -x c - $(ALL_CFLAGS) -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -o /dev/null $(ALL_LDFLAGS) > /dev/null 2>&1 && echo y"), y) + ifneq ($(shell sh -c "(echo '\#include '; echo 'int main(void) { Elf * elf = elf_begin(0, ELF_C_READ_MMAP, 0); return (long)elf; }') | $(CC) -x c - $(ALL_CFLAGS) -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -o /dev/null $(ALL_LDFLAGS) $(EXTLIBS) > /dev/null 2>&1 && echo y"), y) BASIC_CFLAGS += -DLIBELF_NO_MMAP endif else msg := $(error No libelf.h/libelf found, please install libelf-dev/elfutils-libelf-devel and glibc-dev[el]); endif -ifneq ($(shell sh -c "(echo '\#include '; echo '\#include '; echo 'int main(void) { Dwarf_Debug dbg; Dwarf_Error err; Dwarf_Ranges *rng; dwarf_init(0, DW_DLC_READ, 0, 0, &dbg, &err); dwarf_get_ranges(dbg, 0, &rng, 0, 0, &err); return (long)dbg; }') | $(CC) -x c - $(ALL_CFLAGS) -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -ldwarf -lelf -o /dev/null $(ALL_LDFLAGS) > /dev/null 2>&1 && echo y"), y) +ifneq ($(shell sh -c "(echo '\#include '; echo '\#include '; echo 'int main(void) { Dwarf_Debug dbg; Dwarf_Error err; Dwarf_Ranges *rng; dwarf_init(0, DW_DLC_READ, 0, 0, &dbg, &err); dwarf_get_ranges(dbg, 0, &rng, 0, 0, &err); return (long)dbg; }') | $(CC) -x c - $(ALL_CFLAGS) -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -ldwarf -lelf -o /dev/null $(ALL_LDFLAGS) $(EXTLIBS) > /dev/null 2>&1 && echo y"), y) msg := $(warning No libdwarf.h found or old libdwarf.h found, disables dwarf support. Please install libdwarf-dev/libdwarf-devel >= 20081231); BASIC_CFLAGS += -DNO_LIBDWARF else @@ -494,20 +496,20 @@ endif ifdef NO_DEMANGLE BASIC_CFLAGS += -DNO_DEMANGLE else - has_bfd := $(shell sh -c "(echo '\#include '; echo 'int main(void) { bfd_demangle(0, 0, 0); return 0; }') | $(CC) -x c - $(ALL_CFLAGS) -o /dev/null $(ALL_LDFLAGS) -lbfd > /dev/null 2>&1 && echo y") + has_bfd := $(shell sh -c "(echo '\#include '; echo 'int main(void) { bfd_demangle(0, 0, 0); return 0; }') | $(CC) -x c - $(ALL_CFLAGS) -o /dev/null $(ALL_LDFLAGS) $(EXTLIBS) -lbfd > /dev/null 2>&1 && echo y") ifeq ($(has_bfd),y) EXTLIBS += -lbfd else - has_bfd_iberty := $(shell sh -c "(echo '\#include '; echo 'int main(void) { bfd_demangle(0, 0, 0); return 0; }') | $(CC) -x c - $(ALL_CFLAGS) -o /dev/null $(ALL_LDFLAGS) -lbfd -liberty > /dev/null 2>&1 && echo y") + has_bfd_iberty := $(shell sh -c "(echo '\#include '; echo 'int main(void) { bfd_demangle(0, 0, 0); return 0; }') | $(CC) -x c - $(ALL_CFLAGS) -o /dev/null $(ALL_LDFLAGS) $(EXTLIBS) -lbfd -liberty > /dev/null 2>&1 && echo y") ifeq ($(has_bfd_iberty),y) EXTLIBS += -lbfd -liberty else - has_bfd_iberty_z := $(shell sh -c "(echo '\#include '; echo 'int main(void) { bfd_demangle(0, 0, 0); return 0; }') | $(CC) -x c - $(ALL_CFLAGS) -o /dev/null $(ALL_LDFLAGS) -lbfd -liberty -lz > /dev/null 2>&1 && echo y") + has_bfd_iberty_z := $(shell sh -c "(echo '\#include '; echo 'int main(void) { bfd_demangle(0, 0, 0); return 0; }') | $(CC) -x c - $(ALL_CFLAGS) -o /dev/null $(ALL_LDFLAGS) $(EXTLIBS) -lbfd -liberty -lz > /dev/null 2>&1 && echo y") ifeq ($(has_bfd_iberty_z),y) EXTLIBS += -lbfd -liberty -lz else - has_cplus_demangle := $(shell sh -c "(echo 'extern char *cplus_demangle(const char *, int);'; echo 'int main(void) { cplus_demangle(0, 0); return 0; }') | $(CC) -x c - $(ALL_CFLAGS) -o /dev/null $(ALL_LDFLAGS) -liberty > /dev/null 2>&1 && echo y") + has_cplus_demangle := $(shell sh -c "(echo 'extern char *cplus_demangle(const char *, int);'; echo 'int main(void) { cplus_demangle(0, 0); return 0; }') | $(CC) -x c - $(ALL_CFLAGS) -o /dev/null $(ALL_LDFLAGS) $(EXTLIBS) -liberty > /dev/null 2>&1 && echo y") ifeq ($(has_cplus_demangle),y) EXTLIBS += -liberty BASIC_CFLAGS += -DHAVE_CPLUS_DEMANGLE -- cgit v1.2.3-18-g5258 From 11ada26c78febe4662a8e848f3bff74e3200c920 Mon Sep 17 00:00:00 2001 From: "Luck, Tony" Date: Tue, 17 Nov 2009 09:05:56 -0800 Subject: perf tools: Add ia64 support for tools/perf/ Compiler on ia64 rejects the "-m64" option. Add arch specific pieces to perf.h Signed-off-by: Tony Luck Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <4b02d7f43514327a@agluck-desktop.sc.intel.com> Signed-off-by: Ingo Molnar --- tools/perf/Makefile | 8 +++++--- tools/perf/perf.h | 6 ++++++ 2 files changed, 11 insertions(+), 3 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/Makefile b/tools/perf/Makefile index 5d1a8b0dff8..3f0666af93d 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -166,10 +166,12 @@ ifdef NO_64BIT MBITS := -m32 else # - # If we're on a 64-bit kernel, use -m64: + # If we're on a 64-bit kernel (except ia64), use -m64: # - ifneq ($(patsubst %64,%,$(uname_M)),$(uname_M)) - MBITS := -m64 + ifneq ($(uname_M),ia64) + ifneq ($(patsubst %64,%,$(uname_M)),$(uname_M)) + MBITS := -m64 + endif endif endif diff --git a/tools/perf/perf.h b/tools/perf/perf.h index 216bdb223f6..454d5d55f32 100644 --- a/tools/perf/perf.h +++ b/tools/perf/perf.h @@ -53,6 +53,12 @@ #define cpu_relax() asm volatile("" ::: "memory") #endif +#ifdef __ia64__ +#include "../../arch/ia64/include/asm/unistd.h" +#define rmb() asm volatile ("mf" ::: "memory") +#define cpu_relax() asm volatile ("hint @pause" ::: "memory") +#endif + #include #include #include -- cgit v1.2.3-18-g5258 From cfc10d3bcc50d70f72c0f43d03eee965c726ccc0 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Tue, 17 Nov 2009 15:40:53 -0200 Subject: perf symbols: Add a long_name_len member to struct dso MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Using a two bytes hole we already had and since we also need to calculate this strlen for fetching the buildids. We'll use it in 'perf top' to auto-adjust the output based on the terminal width. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1258479655-28662-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/util/symbol.c | 26 +++++++++++++++++++++----- tools/perf/util/symbol.h | 1 + 2 files changed, 22 insertions(+), 5 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 1b77e81b38d..5cc96c86861 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -109,13 +109,24 @@ static size_t symbol__fprintf(struct symbol *self, FILE *fp) self->start, self->end, self->name); } +static void dso__set_long_name(struct dso *self, char *name) +{ + self->long_name = name; + self->long_name_len = strlen(name); +} + +static void dso__set_basename(struct dso *self) +{ + self->short_name = basename(self->long_name); +} + struct dso *dso__new(const char *name) { struct dso *self = malloc(sizeof(*self) + strlen(name) + 1); if (self != NULL) { strcpy(self->name, name); - self->long_name = self->name; + dso__set_long_name(self, self->name); self->short_name = self->name; self->syms = RB_ROOT; self->find_symbol = dso__find_symbol; @@ -888,7 +899,7 @@ bool fetch_build_id_table(struct list_head *head) continue; have_buildid = true; memset(&b.header, 0, sizeof(b.header)); - len = strlen(pos->long_name) + 1; + len = pos->long_name_len + 1; len = ALIGN(len, 64); b.header.size = sizeof(b) + len; @@ -1165,6 +1176,7 @@ static int dsos__load_modules_sym_dir(char *dirname, symbol_filter_t filter) dso_name[PATH_MAX]; struct map *map; struct rb_node *last; + char *long_name; if (dot == NULL || strcmp(dot, ".ko")) continue; @@ -1179,9 +1191,11 @@ static int dsos__load_modules_sym_dir(char *dirname, symbol_filter_t filter) snprintf(path, sizeof(path), "%s/%s", dirname, dent->d_name); - map->dso->long_name = strdup(path); - if (map->dso->long_name == NULL) + long_name = strdup(path); + if (long_name == NULL) goto failure; + dso__set_long_name(map->dso, long_name); + dso__set_basename(map->dso); err = dso__load_module_sym(map->dso, map, filter); if (err < 0) @@ -1420,8 +1434,10 @@ struct dso *dsos__findnew(const char *name) if (!dso) { dso = dso__new(name); - if (dso != NULL) + if (dso != NULL) { dsos__add(dso); + dso__set_basename(dso); + } } return dso; diff --git a/tools/perf/util/symbol.h b/tools/perf/util/symbol.h index 51c5a4a0813..5ad1019607d 100644 --- a/tools/perf/util/symbol.h +++ b/tools/perf/util/symbol.h @@ -66,6 +66,7 @@ struct dso { u8 has_build_id:1; unsigned char origin; u8 build_id[BUILD_ID_SIZE]; + u16 long_name_len; const char *short_name; char *long_name; char name[0]; -- cgit v1.2.3-18-g5258 From 13cc5079f235906e60577dbce8da2f9607e67e93 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Tue, 17 Nov 2009 15:40:54 -0200 Subject: perf top: Auto adjust symbol and dso widths MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We pre-calculate the symbol name length, then after we sort the entries to print, calculate the biggest one and use that for the symbol name width justification, then use the dso->long_name->len to justificate the DSO name, deciding whether using the short or long name depending on how much space we have on the terminal. IOW give as much info to the user as the terminal width allows. Suggested-by: Ingo Molnar Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1258479655-28662-2-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/builtin-top.c | 92 +++++++++++++++++++++++++++++++++++------------- 1 file changed, 67 insertions(+), 25 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index 89b7f68a179..a368978d517 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -78,6 +78,14 @@ static int dump_symtab = 0; static bool hide_kernel_symbols = false; static bool hide_user_symbols = false; +static struct winsize winsize; +static const char *graph_line = + "_____________________________________________________________________" + "_____________________________________________________________________"; +static const char *graph_dotted_line = + "---------------------------------------------------------------------" + "---------------------------------------------------------------------" + "---------------------------------------------------------------------"; /* * Source @@ -107,6 +115,7 @@ struct sym_entry { unsigned long snap_count; double weight; int skip; + u16 name_len; u8 origin; struct map *map; struct source_line *source; @@ -119,34 +128,40 @@ struct sym_entry { * Source functions */ -/* most GUI terminals set LINES (although some don't export it) */ -static int term_rows(void) +static void get_term_dimensions(struct winsize *ws) { - char *lines_string = getenv("LINES"); - int n_lines; - - if (lines_string && (n_lines = atoi(lines_string)) > 0) - return n_lines; -#ifdef TIOCGWINSZ - else { - struct winsize ws; - if (!ioctl(1, TIOCGWINSZ, &ws) && ws.ws_row) - return ws.ws_row; + char *s = getenv("LINES"); + + if (s != NULL) { + ws->ws_row = atoi(s); + s = getenv("COLUMNS"); + if (s != NULL) { + ws->ws_col = atoi(s); + if (ws->ws_row && ws->ws_col) + return; + } } +#ifdef TIOCGWINSZ + if (ioctl(1, TIOCGWINSZ, ws) == 0 && + ws->ws_row && ws->ws_col) + return; #endif - return 25; + ws->ws_row = 25; + ws->ws_col = 80; } -static void update_print_entries(void) +static void update_print_entries(struct winsize *ws) { - print_entries = term_rows(); + print_entries = ws->ws_row; + if (print_entries > 9) print_entries -= 9; } static void sig_winch_handler(int sig __used) { - update_print_entries(); + get_term_dimensions(&winsize); + update_print_entries(&winsize); } static void parse_source(struct sym_entry *syme) @@ -423,6 +438,8 @@ static void print_sym_table(void) struct sym_entry *syme, *n; struct rb_root tmp = RB_ROOT; struct rb_node *nd; + int sym_width = 0, dso_width; + const int win_width = winsize.ws_col - 1; samples = userspace_samples = 0; @@ -434,6 +451,7 @@ static void print_sym_table(void) list_for_each_entry_safe_from(syme, n, &active_symbols, node) { syme->snap_count = syme->count[snap]; if (syme->snap_count != 0) { + if ((hide_user_symbols && syme->origin == PERF_RECORD_MISC_USER) || (hide_kernel_symbols && @@ -453,8 +471,7 @@ static void print_sym_table(void) puts(CONSOLE_CLEAR); - printf( -"------------------------------------------------------------------------------\n"); + printf("%-*.*s\n", win_width, win_width, graph_dotted_line); printf( " PerfTop:%8.0f irqs/sec kernel:%4.1f%% [", samples_per_sec, 100.0 - (100.0*((samples_per_sec-ksamples_per_sec)/samples_per_sec))); @@ -492,26 +509,44 @@ static void print_sym_table(void) printf(", %d CPUs)\n", nr_cpus); } - printf("------------------------------------------------------------------------------\n\n"); + printf("%-*.*s\n\n", win_width, win_width, graph_dotted_line); if (sym_filter_entry) { show_details(sym_filter_entry); return; } + /* + * Find the longest symbol name that will be displayed + */ + for (nd = rb_first(&tmp); nd; nd = rb_next(nd)) { + syme = rb_entry(nd, struct sym_entry, rb_node); + if (++printed > print_entries || + (int)syme->snap_count < count_filter) + continue; + + if (syme->name_len > sym_width) + sym_width = syme->name_len; + } + + printed = 0; + if (nr_counters == 1) printf(" samples pcnt"); else printf(" weight samples pcnt"); + dso_width = winsize.ws_col - sym_width - 29; + if (verbose) printf(" RIP "); - printf(" function DSO\n"); + printf(" %-*.*s DSO\n", sym_width, sym_width, "function"); printf(" %s _______ _____", nr_counters == 1 ? " " : "______"); if (verbose) printf(" ________________"); - printf(" ________________________________ ________________\n\n"); + printf(" %-*.*s %-*.*s\n\n", sym_width, sym_width, graph_line, + dso_width, dso_width, graph_line); for (nd = rb_first(&tmp); nd; nd = rb_next(nd)) { struct symbol *sym; @@ -534,8 +569,11 @@ static void print_sym_table(void) percent_color_fprintf(stdout, "%4.1f%%", pcnt); if (verbose) printf(" %016llx", sym->start); - printf(" %-32s", sym->name); - printf(" %s", syme->map->dso->short_name); + printf(" %-*.*s", sym_width, sym_width, sym->name); + printf(" %-*.*s", dso_width, dso_width, + dso_width >= syme->map->dso->long_name_len ? + syme->map->dso->long_name : + syme->map->dso->short_name); printf("\n"); } } @@ -718,7 +756,7 @@ static void handle_keypress(int c) case 'e': prompt_integer(&print_entries, "Enter display entries (lines)"); if (print_entries == 0) { - update_print_entries(); + sig_winch_handler(SIGWINCH); signal(SIGWINCH, sig_winch_handler); } else signal(SIGWINCH, SIG_DFL); @@ -862,6 +900,9 @@ static int symbol_filter(struct map *map, struct symbol *sym) } } + if (!syme->skip) + syme->name_len = strlen(sym->name); + return 0; } @@ -1301,8 +1342,9 @@ int cmd_top(int argc, const char **argv, const char *prefix __used) if (target_pid != -1 || profile_cpu != -1) nr_cpus = 1; + get_term_dimensions(&winsize); if (print_entries == 0) { - update_print_entries(); + update_print_entries(&winsize); signal(SIGWINCH, sig_winch_handler); } -- cgit v1.2.3-18-g5258 From 1a105f743d9fa5f7b8eeeca0afb789951164a361 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Tue, 17 Nov 2009 15:40:55 -0200 Subject: perf top: Suppress DSO column if only one is present MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit E.g. [root@doppio ~]# perf top -U --------------------------------------------------------------------------- PerfTop: 482 irqs/sec kernel:100.0% [1000Hz cycles], (all, 2 CPUs) --------------------------------------------------------------------------- DSO: vmlinux samples pcnt function _______ _____ _________________________ 471.00 47.9% read_hpet 57.00 5.8% acpi_os_read_port 30.00 3.1% hpet_next_event 30.00 3.1% find_busiest_group 22.00 2.2% schedule 18.00 1.8% sched_clock_local 14.00 1.4% _spin_lock_irqsave 14.00 1.4% native_read_tsc 13.00 1.3% trace_hardirqs_off 9.00 0.9% fget_light 9.00 0.9% ioread8 8.00 0.8% do_sys_poll Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1258479655-28662-3-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/builtin-top.c | 43 ++++++++++++++++++++++++++++++++----------- 1 file changed, 32 insertions(+), 11 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index a368978d517..6db0e37ee33 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -438,8 +438,9 @@ static void print_sym_table(void) struct sym_entry *syme, *n; struct rb_root tmp = RB_ROOT; struct rb_node *nd; - int sym_width = 0, dso_width; + int sym_width = 0, dso_width = 0; const int win_width = winsize.ws_col - 1; + struct dso *unique_dso = NULL, *first_dso = NULL; samples = userspace_samples = 0; @@ -509,7 +510,7 @@ static void print_sym_table(void) printf(", %d CPUs)\n", nr_cpus); } - printf("%-*.*s\n\n", win_width, win_width, graph_dotted_line); + printf("%-*.*s\n", win_width, win_width, graph_dotted_line); if (sym_filter_entry) { show_details(sym_filter_entry); @@ -525,28 +526,47 @@ static void print_sym_table(void) (int)syme->snap_count < count_filter) continue; + if (first_dso == NULL) + unique_dso = first_dso = syme->map->dso; + else if (syme->map->dso != first_dso) + unique_dso = NULL; + + if (syme->map->dso->long_name_len > dso_width) + dso_width = syme->map->dso->long_name_len; + if (syme->name_len > sym_width) sym_width = syme->name_len; } printed = 0; + if (unique_dso) + printf("DSO: %s\n", unique_dso->long_name); + else { + int max_dso_width = winsize.ws_col - sym_width - 29; + if (dso_width > max_dso_width) + dso_width = max_dso_width; + putchar('\n'); + } if (nr_counters == 1) printf(" samples pcnt"); else printf(" weight samples pcnt"); - dso_width = winsize.ws_col - sym_width - 29; - if (verbose) printf(" RIP "); - printf(" %-*.*s DSO\n", sym_width, sym_width, "function"); + printf(" %-*.*s", sym_width, sym_width, "function"); + if (!unique_dso) + printf(" DSO"); + putchar('\n'); printf(" %s _______ _____", nr_counters == 1 ? " " : "______"); if (verbose) printf(" ________________"); - printf(" %-*.*s %-*.*s\n\n", sym_width, sym_width, graph_line, - dso_width, dso_width, graph_line); + printf(" %-*.*s", sym_width, sym_width, graph_line); + if (!unique_dso) + printf(" %-*.*s", dso_width, dso_width, graph_line); + puts("\n"); for (nd = rb_first(&tmp); nd; nd = rb_next(nd)) { struct symbol *sym; @@ -570,10 +590,11 @@ static void print_sym_table(void) if (verbose) printf(" %016llx", sym->start); printf(" %-*.*s", sym_width, sym_width, sym->name); - printf(" %-*.*s", dso_width, dso_width, - dso_width >= syme->map->dso->long_name_len ? - syme->map->dso->long_name : - syme->map->dso->short_name); + if (!unique_dso) + printf(" %-*.*s", dso_width, dso_width, + dso_width >= syme->map->dso->long_name_len ? + syme->map->dso->long_name : + syme->map->dso->short_name); printf("\n"); } } -- cgit v1.2.3-18-g5258 From 51a472decb845e920137284a5cfef51fb7d61206 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Tue, 17 Nov 2009 18:38:00 -0200 Subject: perf top: Introduce helper function to access symbol from sym_entry MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1258490282-1821-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/builtin-top.c | 17 +++++++++++------ 1 file changed, 11 insertions(+), 6 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index 6db0e37ee33..0d60c517c0b 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -128,6 +128,11 @@ struct sym_entry { * Source functions */ +static inline struct symbol *sym_entry__symbol(struct sym_entry *self) +{ + return (struct symbol *)(self + 1); +} + static void get_term_dimensions(struct winsize *ws) { char *s = getenv("LINES"); @@ -181,7 +186,7 @@ static void parse_source(struct sym_entry *syme) goto out_assign; } - sym = (struct symbol *)(syme + 1); + sym = sym_entry__symbol(syme); map = syme->map; path = map->dso->long_name; @@ -276,7 +281,7 @@ out_unlock: static void lookup_sym_source(struct sym_entry *syme) { - struct symbol *symbol = (struct symbol *)(syme + 1); + struct symbol *symbol = sym_entry__symbol(syme); struct source_line *line; char pattern[PATH_MAX]; @@ -325,7 +330,7 @@ static void show_details(struct sym_entry *syme) if (!syme->source) return; - symbol = (struct symbol *)(syme + 1); + symbol = sym_entry__symbol(syme); printf("Showing %s for %s\n", event_name(sym_counter), symbol->name); printf(" Events Pcnt (>=%d%%)\n", sym_pcnt_filter); @@ -573,7 +578,7 @@ static void print_sym_table(void) double pcnt; syme = rb_entry(nd, struct sym_entry, rb_node); - sym = (struct symbol *)(syme + 1); + sym = sym_entry__symbol(syme); if (++printed > print_entries || (int)syme->snap_count < count_filter) continue; @@ -661,7 +666,7 @@ static void prompt_symbol(struct sym_entry **target, const char *msg) pthread_mutex_unlock(&active_symbols_lock); list_for_each_entry_safe_from(syme, n, &active_symbols, node) { - struct symbol *sym = (struct symbol *)(syme + 1); + struct symbol *sym = sym_entry__symbol(syme); if (!strcmp(buf, sym->name)) { found = syme; @@ -685,7 +690,7 @@ static void print_mapped_keys(void) char *name = NULL; if (sym_filter_entry) { - struct symbol *sym = (struct symbol *)(sym_filter_entry+1); + struct symbol *sym = sym_entry__symbol(sym_filter_entry); name = sym->name; } -- cgit v1.2.3-18-g5258 From 5a8e5a3065bf04b7673262fd6c46123e4b888d2b Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Tue, 17 Nov 2009 18:38:01 -0200 Subject: perf top: Allocate space only for the number of counters used MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Reducing memory consumption on a typical desktop machine: From: 32710 root 20 0 172m 142m 1056 S 0.0 4.7 0:00.37 perf To: 420 root 20 0 47528 16m 1056 R 0.3 0.5 0:00.24 perf Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1258490282-1821-2-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/builtin-top.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index 0d60c517c0b..49cf87680fe 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -111,7 +111,6 @@ static int display_weighted = -1; struct sym_entry { struct rb_node rb_node; struct list_head node; - unsigned long count[MAX_COUNTERS]; unsigned long snap_count; double weight; int skip; @@ -122,6 +121,7 @@ struct sym_entry { struct source_line *lines; struct source_line **lines_tail; pthread_mutex_t source_lock; + unsigned long count[0]; }; /* @@ -130,7 +130,7 @@ struct sym_entry { static inline struct symbol *sym_entry__symbol(struct sym_entry *self) { - return (struct symbol *)(self + 1); + return ((void *)self) + symbol__priv_size; } static void get_term_dimensions(struct winsize *ws) @@ -1314,8 +1314,6 @@ int cmd_top(int argc, const char **argv, const char *prefix __used) { int counter; - symbol__init(sizeof(struct sym_entry)); - page_size = sysconf(_SC_PAGE_SIZE); argc = parse_options(argc, argv, options, top_usage, 0); @@ -1332,6 +1330,9 @@ int cmd_top(int argc, const char **argv, const char *prefix __used) if (!nr_counters) nr_counters = 1; + symbol__init(sizeof(struct sym_entry) + + (nr_counters + 1) * sizeof(unsigned long)); + if (delay_secs < 1) delay_secs = 1; -- cgit v1.2.3-18-g5258 From b269876c8d57fb8c801bea1fc34b461646c5abd0 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Tue, 17 Nov 2009 18:38:02 -0200 Subject: perf top: Don't allocate the source parsing members upfront MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Defer to parse_source() time allocating it. Now we use about this much memory: 1724 root 20 0 42104 10m 940 S 0.0 0.4 0:00.23 perf Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1258490282-1821-3-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/builtin-top.c | 76 ++++++++++++++++++++++++++++-------------------- 1 file changed, 45 insertions(+), 31 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index 49cf87680fe..07b92c378ae 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -108,6 +108,13 @@ static int display_weighted = -1; * Symbols */ +struct sym_entry_source { + struct source_line *source; + struct source_line *lines; + struct source_line **lines_tail; + pthread_mutex_t lock; +}; + struct sym_entry { struct rb_node rb_node; struct list_head node; @@ -117,10 +124,7 @@ struct sym_entry { u16 name_len; u8 origin; struct map *map; - struct source_line *source; - struct source_line *lines; - struct source_line **lines_tail; - pthread_mutex_t source_lock; + struct sym_entry_source *src; unsigned long count[0]; }; @@ -172,6 +176,7 @@ static void sig_winch_handler(int sig __used) static void parse_source(struct sym_entry *syme) { struct symbol *sym; + struct sym_entry_source *source; struct map *map; FILE *file; char command[PATH_MAX*2]; @@ -181,8 +186,17 @@ static void parse_source(struct sym_entry *syme) if (!syme) return; - if (syme->lines) { - pthread_mutex_lock(&syme->source_lock); + if (syme->src == NULL) { + syme->src = calloc(1, sizeof(*source)); + if (syme->src == NULL) + return; + pthread_mutex_init(&syme->src->lock, NULL); + } + + source = syme->src; + + if (source->lines) { + pthread_mutex_lock(&source->lock); goto out_assign; } @@ -202,8 +216,8 @@ static void parse_source(struct sym_entry *syme) if (!file) return; - pthread_mutex_lock(&syme->source_lock); - syme->lines_tail = &syme->lines; + pthread_mutex_lock(&source->lock); + source->lines_tail = &source->lines; while (!feof(file)) { struct source_line *src; size_t dummy = 0; @@ -223,8 +237,8 @@ static void parse_source(struct sym_entry *syme) *c = 0; src->next = NULL; - *syme->lines_tail = src; - syme->lines_tail = &src->next; + *source->lines_tail = src; + source->lines_tail = &src->next; if (strlen(src->line)>8 && src->line[8] == ':') { src->eip = strtoull(src->line, NULL, 16); @@ -238,7 +252,7 @@ static void parse_source(struct sym_entry *syme) pclose(file); out_assign: sym_filter_entry = syme; - pthread_mutex_unlock(&syme->source_lock); + pthread_mutex_unlock(&source->lock); } static void __zero_source_counters(struct sym_entry *syme) @@ -246,7 +260,7 @@ static void __zero_source_counters(struct sym_entry *syme) int i; struct source_line *line; - line = syme->lines; + line = syme->src->lines; while (line) { for (i = 0; i < nr_counters; i++) line->count[i] = 0; @@ -261,13 +275,13 @@ static void record_precise_ip(struct sym_entry *syme, int counter, u64 ip) if (syme != sym_filter_entry) return; - if (pthread_mutex_trylock(&syme->source_lock)) + if (pthread_mutex_trylock(&syme->src->lock)) return; - if (!syme->source) + if (syme->src == NULL || syme->src->source == NULL) goto out_unlock; - for (line = syme->lines; line; line = line->next) { + for (line = syme->src->lines; line; line = line->next) { if (line->eip == ip) { line->count[counter]++; break; @@ -276,7 +290,7 @@ static void record_precise_ip(struct sym_entry *syme, int counter, u64 ip) break; } out_unlock: - pthread_mutex_unlock(&syme->source_lock); + pthread_mutex_unlock(&syme->src->lock); } static void lookup_sym_source(struct sym_entry *syme) @@ -287,14 +301,14 @@ static void lookup_sym_source(struct sym_entry *syme) sprintf(pattern, "<%s>:", symbol->name); - pthread_mutex_lock(&syme->source_lock); - for (line = syme->lines; line; line = line->next) { + pthread_mutex_lock(&syme->src->lock); + for (line = syme->src->lines; line; line = line->next) { if (strstr(line->line, pattern)) { - syme->source = line; + syme->src->source = line; break; } } - pthread_mutex_unlock(&syme->source_lock); + pthread_mutex_unlock(&syme->src->lock); } static void show_lines(struct source_line *queue, int count, int total) @@ -324,24 +338,24 @@ static void show_details(struct sym_entry *syme) if (!syme) return; - if (!syme->source) + if (!syme->src->source) lookup_sym_source(syme); - if (!syme->source) + if (!syme->src->source) return; symbol = sym_entry__symbol(syme); printf("Showing %s for %s\n", event_name(sym_counter), symbol->name); printf(" Events Pcnt (>=%d%%)\n", sym_pcnt_filter); - pthread_mutex_lock(&syme->source_lock); - line = syme->source; + pthread_mutex_lock(&syme->src->lock); + line = syme->src->source; while (line) { total += line->count[sym_counter]; line = line->next; } - line = syme->source; + line = syme->src->source; while (line) { float pcnt = 0.0; @@ -366,7 +380,7 @@ static void show_details(struct sym_entry *syme) line->count[sym_counter] = zero ? 0 : line->count[sym_counter] * 7 / 8; line = line->next; } - pthread_mutex_unlock(&syme->source_lock); + pthread_mutex_unlock(&syme->src->lock); if (more) printf("%d lines not displayed, maybe increase display entries [e]\n", more); } @@ -647,10 +661,10 @@ static void prompt_symbol(struct sym_entry **target, const char *msg) /* zero counters of active symbol */ if (syme) { - pthread_mutex_lock(&syme->source_lock); + pthread_mutex_lock(&syme->src->lock); __zero_source_counters(syme); *target = NULL; - pthread_mutex_unlock(&syme->source_lock); + pthread_mutex_unlock(&syme->src->lock); } fprintf(stdout, "\n%s: ", msg); @@ -826,10 +840,10 @@ static void handle_keypress(int c) else { struct sym_entry *syme = sym_filter_entry; - pthread_mutex_lock(&syme->source_lock); + pthread_mutex_lock(&syme->src->lock); sym_filter_entry = NULL; __zero_source_counters(syme); - pthread_mutex_unlock(&syme->source_lock); + pthread_mutex_unlock(&syme->src->lock); } break; case 'U': @@ -915,7 +929,7 @@ static int symbol_filter(struct map *map, struct symbol *sym) syme = symbol__priv(sym); syme->map = map; - pthread_mutex_init(&syme->source_lock, NULL); + syme->src = NULL; if (!sym_filter_entry && sym_filter && !strcmp(name, sym_filter)) sym_filter_entry = syme; -- cgit v1.2.3-18-g5258 From 827f3b4974c5db2968d4979fe6a0ae00ab37bdd8 Mon Sep 17 00:00:00 2001 From: Hitoshi Mitake Date: Wed, 18 Nov 2009 00:20:09 +0900 Subject: perf bench: Add memcpy() benchmark 'perf bench mem memcpy' is a benchmark suite for measuring memcpy() performance. Example on a Intel(R) Core(TM)2 Duo CPU E6850 @ 3.00GHz: | % perf bench mem memcpy -l 1GB | # Running mem/memcpy benchmark... | # Copying 1MB Bytes from 0xb7d98008 to 0xb7e99008 ... | | 726.216412 MB/Sec Signed-off-by: Hitoshi Mitake Cc: Peter Zijlstra Cc: Paul Mackerras Cc: Frederic Weisbecker LKML-Reference: <1258471212-30281-1-git-send-email-mitake@dcl.info.waseda.ac.jp> [ v2: updated changelog, clarified history of builtin-bench.c ] Signed-off-by: Ingo Molnar --- tools/perf/Makefile | 1 + tools/perf/bench/bench.h | 1 + tools/perf/bench/mem-memcpy.c | 186 ++++++++++++++++++++++++++++++++++++++++++ tools/perf/builtin-bench.c | 15 +++- 4 files changed, 202 insertions(+), 1 deletion(-) create mode 100644 tools/perf/bench/mem-memcpy.c (limited to 'tools/perf') diff --git a/tools/perf/Makefile b/tools/perf/Makefile index 3f0666af93d..53e663a5fa2 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -432,6 +432,7 @@ BUILTIN_OBJS += builtin-bench.o # Benchmark modules BUILTIN_OBJS += bench/sched-messaging.o BUILTIN_OBJS += bench/sched-pipe.o +BUILTIN_OBJS += bench/mem-memcpy.o BUILTIN_OBJS += builtin-help.o BUILTIN_OBJS += builtin-sched.o diff --git a/tools/perf/bench/bench.h b/tools/perf/bench/bench.h index 9fbd8d745fa..f7781c6267c 100644 --- a/tools/perf/bench/bench.h +++ b/tools/perf/bench/bench.h @@ -3,6 +3,7 @@ extern int bench_sched_messaging(int argc, const char **argv, const char *prefix); extern int bench_sched_pipe(int argc, const char **argv, const char *prefix); +extern int bench_mem_memcpy(int argc, const char **argv, const char *prefix __used); #define BENCH_FORMAT_DEFAULT_STR "default" #define BENCH_FORMAT_DEFAULT 0 diff --git a/tools/perf/bench/mem-memcpy.c b/tools/perf/bench/mem-memcpy.c new file mode 100644 index 00000000000..d4f4f9806ae --- /dev/null +++ b/tools/perf/bench/mem-memcpy.c @@ -0,0 +1,186 @@ +/* + * mem-memcpy.c + * + * memcpy: Simple memory copy in various ways + * + * Written by Hitoshi Mitake + */ +#include + +#include "../perf.h" +#include "../util/util.h" +#include "../util/parse-options.h" +#include "../util/string.h" +#include "../util/header.h" +#include "bench.h" + +#include +#include +#include +#include +#include + +#define K 1024 + +static const char *length_str = "1MB"; +static const char *routine = "default"; +static int use_clock = 0; + +static const struct option options[] = { + OPT_STRING('l', "length", &length_str, "1MB", + "Specify length of memory to copy. " + "available unit: B, MB, GB (upper and lower)"), + OPT_STRING('r', "routine", &routine, "default", + "Specify routine to copy"), + OPT_BOOLEAN('c', "clock", &use_clock, + "Use CPU clock for measuring"), + OPT_END() +}; + +struct routine { + const char *name; + const char *desc; + void * (*fn)(void *dst, const void *src, size_t len); +}; + +struct routine routines[] = { + { "default", + "Default memcpy() provided by glibc", + memcpy }, + { NULL, + NULL, + NULL } +}; + +static const char * const bench_mem_memcpy_usage[] = { + "perf bench mem memcpy ", + NULL +}; + +static int clock_fd; + +static struct perf_event_attr clock_attr = { + .type = PERF_TYPE_HARDWARE, + .config = PERF_COUNT_HW_CPU_CYCLES +}; + +static void init_clock(void) +{ + clock_fd = sys_perf_event_open(&clock_attr, getpid(), -1, -1, 0); + BUG_ON(clock_fd < 0); +} + +static u64 get_clock(void) +{ + int ret; + u64 clk; + + ret = read(clock_fd, &clk, sizeof(u64)); + BUG_ON(ret != sizeof(u64)); + + return clk; +} + +static double timeval2double(struct timeval *ts) +{ + return (double)ts->tv_sec + + (double)ts->tv_usec / (double)1000000; +} + +int bench_mem_memcpy(int argc, const char **argv, + const char *prefix __used) +{ + int i; + void *dst, *src; + size_t length; + double bps = 0.0; + struct timeval tv_start, tv_end, tv_diff; + u64 clock_start, clock_end, clock_diff; + + clock_start = clock_end = clock_diff = 0ULL; + argc = parse_options(argc, argv, options, + bench_mem_memcpy_usage, 0); + + tv_diff.tv_sec = 0; + tv_diff.tv_usec = 0; + length = (size_t)perf_atoll((char *)length_str); + if ((long long int)length <= 0) { + fprintf(stderr, "Invalid length:%s\n", length_str); + return 1; + } + + for (i = 0; routines[i].name; i++) { + if (!strcmp(routines[i].name, routine)) + break; + } + if (!routines[i].name) { + printf("Unknown routine:%s\n", routine); + printf("Available routines...\n"); + for (i = 0; routines[i].name; i++) { + printf("\t%s ... %s\n", + routines[i].name, routines[i].desc); + } + return 1; + } + + dst = calloc(length, sizeof(char)); + assert(dst); + src = calloc(length, sizeof(char)); + assert(src); + + if (bench_format == BENCH_FORMAT_DEFAULT) { + printf("# Copying %s Bytes from %p to %p ...\n\n", + length_str, src, dst); + } + + if (use_clock) { + init_clock(); + clock_start = get_clock(); + } else + BUG_ON(gettimeofday(&tv_start, NULL)); + + routines[i].fn(dst, src, length); + + if (use_clock) { + clock_end = get_clock(); + clock_diff = clock_end - clock_start; + } else { + BUG_ON(gettimeofday(&tv_end, NULL)); + timersub(&tv_end, &tv_start, &tv_diff); + bps = (double)((double)length / timeval2double(&tv_diff)); + } + + switch (bench_format) { + case BENCH_FORMAT_DEFAULT: + if (use_clock) { + printf(" %14lf Clock/Byte\n", + (double)clock_diff / (double)length); + } else { + if (bps < K) + printf(" %14lf B/Sec\n", bps); + else if (bps < K * K) + printf(" %14lfd KB/Sec\n", bps / 1024); + else if (bps < K * K * K) + printf(" %14lf MB/Sec\n", bps / 1024 / 1024); + else { + printf(" %14lf GB/Sec\n", + bps / 1024 / 1024 / 1024); + } + } + break; + case BENCH_FORMAT_SIMPLE: + if (use_clock) { + printf("%14lf\n", + (double)clock_diff / (double)length); + } else + printf("%lf\n", bps); + break; + default: + /* reaching here is something disaster */ + fprintf(stderr, "Unknown format:%d\n", bench_format); + exit(1); + break; + } + + return 0; +} diff --git a/tools/perf/builtin-bench.c b/tools/perf/builtin-bench.c index 90c39baae0d..e043eb83092 100644 --- a/tools/perf/builtin-bench.c +++ b/tools/perf/builtin-bench.c @@ -12,6 +12,7 @@ * * Available subsystem list: * sched ... scheduler and IPC mechanism + * mem ... memory access performance * */ @@ -43,6 +44,15 @@ static struct bench_suite sched_suites[] = { NULL } }; +static struct bench_suite mem_suites[] = { + { "memcpy", + "Simple memory copy in various ways", + bench_mem_memcpy }, + { NULL, + NULL, + NULL } +}; + struct bench_subsys { const char *name; const char *summary; @@ -53,9 +63,12 @@ static struct bench_subsys subsystems[] = { { "sched", "scheduler and IPC mechanism", sched_suites }, + { "mem", + "memory access performance", + mem_suites }, { NULL, NULL, - NULL } + NULL } }; static void dump_suites(int subsys_index) -- cgit v1.2.3-18-g5258 From d3379ab9050e5522da2aac53d413651fc06be562 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Wed, 18 Nov 2009 20:20:50 -0200 Subject: perf symbols: Fix comparision of build_ids MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit When we read the build_id from the DSO name to then index into /usr/lib/debug/.buildid/DSO_BUILD_ID[0:2]/DSO_BUILD_ID[2:], we were jumping directly to the comparision with the buildid we already have in dso->build_id (that came from the perf.data build_id section, collected at perf record time) unconditionally, even if we didn't had recorded it, and furthermore, comparing a formatted buildid with a rawbuildid, yikes. Fix it by deleting the dso__read_build_id() function, that was really misdesigned anyway, and do the necessary checks and correct comparison of raw buildids. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1258582853-8579-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/util/symbol.c | 52 +++++++++++++++--------------------------------- 1 file changed, 16 insertions(+), 36 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 5cc96c86861..594f36a1da8 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -962,25 +962,6 @@ out: return err; } -static char *dso__read_build_id(struct dso *self) -{ - int len; - char *build_id = NULL; - unsigned char rawbf[BUILD_ID_SIZE]; - - len = filename__read_build_id(self->long_name, rawbf, sizeof(rawbf)); - if (len < 0) - goto out; - - build_id = malloc(len * 2 + 1); - if (build_id == NULL) - goto out; - - build_id__sprintf(rawbf, len, build_id); -out: - return build_id; -} - char dso__symtab_origin(const struct dso *self) { static const char origin[] = { @@ -1001,7 +982,8 @@ char dso__symtab_origin(const struct dso *self) int dso__load(struct dso *self, struct map *map, symbol_filter_t filter) { int size = PATH_MAX; - char *name = malloc(size), *build_id = NULL; + char *name = malloc(size); + u8 build_id[BUILD_ID_SIZE]; int ret = -1; int fd; @@ -1023,8 +1005,6 @@ int dso__load(struct dso *self, struct map *map, symbol_filter_t filter) more: do { - int berr = 0; - self->origin++; switch (self->origin) { case DSO__ORIG_FEDORA: @@ -1036,12 +1016,18 @@ more: self->long_name); break; case DSO__ORIG_BUILDID: - build_id = dso__read_build_id(self); - if (build_id != NULL) { + if (filename__read_build_id(self->long_name, build_id, + sizeof(build_id))) { + char build_id_hex[BUILD_ID_SIZE * 2 + 1]; + + build_id__sprintf(build_id, sizeof(build_id), + build_id_hex); snprintf(name, size, "/usr/lib/debug/.build-id/%.2s/%s.debug", - build_id, build_id + 2); - goto compare_build_id; + build_id_hex, build_id_hex + 2); + if (self->has_build_id) + goto compare_build_id; + break; } self->origin++; /* Fall thru */ @@ -1054,18 +1040,12 @@ more: } if (self->has_build_id) { - bool match; - build_id = malloc(BUILD_ID_SIZE); - if (build_id == NULL) + if (filename__read_build_id(name, build_id, + sizeof(build_id)) < 0) goto more; - berr = filename__read_build_id(name, build_id, - BUILD_ID_SIZE); compare_build_id: - match = berr > 0 && memcmp(build_id, self->build_id, - sizeof(self->build_id)) == 0; - free(build_id); - build_id = NULL; - if (!match) + if (memcmp(build_id, self->build_id, + sizeof(self->build_id)) != 0) goto more; } -- cgit v1.2.3-18-g5258 From e30a3d12ddf04add3268bfceb0e57ffe47f254c6 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Wed, 18 Nov 2009 20:20:51 -0200 Subject: perf symbols: Kill struct build_id_list and die() another day MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit No need for this struct and its allocations, we can just use the ->build_id member we already have in struct dso, then ask for it to be read, and later traverse the dsos list, writing the buildid table to the perf.data file. As a bonus, one more die() function got killed. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1258582853-8579-2-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/util/event.h | 7 ------- tools/perf/util/header.c | 31 ++++++++++++++++++------------- tools/perf/util/symbol.c | 37 +++++++++---------------------------- tools/perf/util/symbol.h | 2 +- 4 files changed, 28 insertions(+), 49 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/event.h b/tools/perf/util/event.h index 1f771ce3a95..34c6fcb82d9 100644 --- a/tools/perf/util/event.h +++ b/tools/perf/util/event.h @@ -69,13 +69,6 @@ struct build_id_event { char filename[]; }; -struct build_id_list { - struct build_id_event event; - struct list_head list; - const char *dso_name; - int len; -}; - typedef union event_union { struct perf_event_header header; struct ip_event ip; diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index b01a9537977..31731f1606b 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -176,18 +176,24 @@ static int do_write(int fd, const void *buf, size_t size) return 0; } -static int write_buildid_table(int fd, struct list_head *id_head) +static int dsos__write_buildid_table(int fd) { - struct build_id_list *iter, *next; - - list_for_each_entry_safe(iter, next, id_head, list) { - struct build_id_event *b = &iter->event; - - if (do_write(fd, b, sizeof(*b)) < 0 || - do_write(fd, iter->dso_name, iter->len) < 0) + struct dso *pos; + + list_for_each_entry(pos, &dsos, node) { + struct build_id_event b; + size_t len; + + if (!pos->has_build_id) + continue; + len = pos->long_name_len + 1; + len = ALIGN(len, 64); + memset(&b, 0, sizeof(b)); + memcpy(&b.build_id, pos->build_id, sizeof(pos->build_id)); + b.header.size = sizeof(b) + len; + if (do_write(fd, &b, sizeof(b)) < 0 || + do_write(fd, pos->long_name, len) < 0) return -1; - list_del(&iter->list); - free(iter); } return 0; @@ -196,14 +202,13 @@ static int write_buildid_table(int fd, struct list_head *id_head) static void perf_header__adds_write(struct perf_header *self, int fd) { - LIST_HEAD(id_list); int nr_sections; struct perf_file_section *feat_sec; int sec_size; u64 sec_start; int idx = 0; - if (fetch_build_id_table(&id_list)) + if (dsos__read_build_ids()) perf_header__set_feat(self, HEADER_BUILD_ID); nr_sections = bitmap_weight(self->adds_features, HEADER_FEAT_BITS); @@ -238,7 +243,7 @@ perf_header__adds_write(struct perf_header *self, int fd) /* Write build-ids */ buildid_sec->offset = lseek(fd, 0, SEEK_CUR); - if (write_buildid_table(fd, &id_list) < 0) + if (dsos__write_buildid_table(fd) < 0) die("failed to write buildid table"); buildid_sec->size = lseek(fd, 0, SEEK_CUR) - buildid_sec->offset; } diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 594f36a1da8..946ec319568 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -883,38 +883,19 @@ out_close: return err; } -bool fetch_build_id_table(struct list_head *head) +bool dsos__read_build_ids(void) { - bool have_buildid = false; + bool have_build_id = false; struct dso *pos; - list_for_each_entry(pos, &dsos, node) { - struct build_id_list *new; - struct build_id_event b; - size_t len; - - if (filename__read_build_id(pos->long_name, - &b.build_id, - sizeof(b.build_id)) < 0) - continue; - have_buildid = true; - memset(&b.header, 0, sizeof(b.header)); - len = pos->long_name_len + 1; - len = ALIGN(len, 64); - b.header.size = sizeof(b) + len; - - new = malloc(sizeof(*new)); - if (!new) - die("No memory\n"); - - memcpy(&new->event, &b, sizeof(b)); - new->dso_name = pos->long_name; - new->len = len; - - list_add_tail(&new->list, head); - } + list_for_each_entry(pos, &dsos, node) + if (filename__read_build_id(pos->long_name, pos->build_id, + sizeof(pos->build_id)) > 0) { + have_build_id = true; + pos->has_build_id = true; + } - return have_buildid; + return have_build_id; } int filename__read_build_id(const char *filename, void *bf, size_t size) diff --git a/tools/perf/util/symbol.h b/tools/perf/util/symbol.h index 5ad1019607d..546eb766d81 100644 --- a/tools/perf/util/symbol.h +++ b/tools/perf/util/symbol.h @@ -89,7 +89,7 @@ char dso__symtab_origin(const struct dso *self); void dso__set_build_id(struct dso *self, void *build_id); int filename__read_build_id(const char *filename, void *bf, size_t size); -bool fetch_build_id_table(struct list_head *head); +bool dsos__read_build_ids(void); int build_id__sprintf(u8 *self, int len, char *bf); int load_kernel(symbol_filter_t filter); -- cgit v1.2.3-18-g5258 From f1617b40596cb341ee6602a9d969c5e4cebe9260 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Wed, 18 Nov 2009 20:20:52 -0200 Subject: perf symbols: Record the build_ids of kernel modules too MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit [root@doppio linux-2.6-tip]# perf record -a sleep 2s;perf buildid-list|tail [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.162 MB perf.data (~7078 samples) ] 881588fa57b3c1696bc91e5e804a11304f093535 [cfg80211] 4d47ce1da9d16bad00c962c072451b7c681e82df [snd_page_alloc] 5146377e89a7caac617f9782f1a02e46263d3a31 [rfkill] 2153b937bff0d345fea83b63a2e1d3138569f83d [i915] 4e6fb1bb97362e3ee4d306988b9ad6912d5fb9ae [drm_kms_helper] f56ef2bf853e3a798f0d8d51f797622e5dc4420e [drm] b0d157a3b5c4e017329ffc07c64623cd6ad65e95 [i2c_algo_bit] 8125374b905ef9fa8b65d98e166b008ad952f198 [i2c_core] fc875c6e5a90e7b915e9d445d0efc859e1b2678c [video] 4b43c5006589f977e9762fdfc7ac1a92b72fca52 [output] [root@doppio linux-2.6-tip]# elfutils libdwfl/linux-kernel-modules.c was used as reference, as suggested by Roland McGrath. Signed-off-by: Arnaldo Carvalho de Melo Cc: Roland McGrath Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1258582853-8579-3-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/util/header.c | 5 +++++ tools/perf/util/symbol.c | 53 +++++++++++++++++++++++++++++++++++++++++++++++- tools/perf/util/symbol.h | 2 ++ 3 files changed, 59 insertions(+), 1 deletion(-) (limited to 'tools/perf') diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index 31731f1606b..d3d656f9a62 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -241,6 +241,11 @@ perf_header__adds_write(struct perf_header *self, int fd) buildid_sec = &feat_sec[idx++]; + /* + * Read the list of loaded modules with its build_ids + */ + dsos__load_modules(); + /* Write build-ids */ buildid_sec->offset = lseek(fd, 0, SEEK_CUR); if (dsos__write_buildid_table(fd) < 0) diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 946ec319568..7b4cedeb302 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -9,6 +9,7 @@ #include #include #include +#include #include enum dso_origin { @@ -943,6 +944,50 @@ out: return err; } +int sysfs__read_build_id(const char *filename, void *build_id, size_t size) +{ + int fd, err = -1; + + if (size < BUILD_ID_SIZE) + goto out; + + fd = open(filename, O_RDONLY); + if (fd < 0) + goto out; + + while (1) { + char bf[BUFSIZ]; + GElf_Nhdr nhdr; + int namesz, descsz; + + if (read(fd, &nhdr, sizeof(nhdr)) != sizeof(nhdr)) + break; + + namesz = (nhdr.n_namesz + 3) & -4U; + descsz = (nhdr.n_descsz + 3) & -4U; + if (nhdr.n_type == NT_GNU_BUILD_ID && + nhdr.n_namesz == sizeof("GNU")) { + if (read(fd, bf, namesz) != namesz) + break; + if (memcmp(bf, "GNU", sizeof("GNU")) == 0) { + if (read(fd, build_id, + BUILD_ID_SIZE) == BUILD_ID_SIZE) { + err = 0; + break; + } + } else if (read(fd, bf, descsz) != descsz) + break; + } else { + int n = namesz + descsz; + if (read(fd, bf, n) != n) + break; + } + } + close(fd); +out: + return err; +} + char dso__symtab_origin(const struct dso *self) { static const char origin[] = { @@ -1218,7 +1263,7 @@ static struct map *map__new2(u64 start, struct dso *dso) return self; } -static int dsos__load_modules(void) +int dsos__load_modules(void) { char *line = NULL; size_t n; @@ -1268,6 +1313,12 @@ static int dsos__load_modules(void) goto out_delete_line; } + snprintf(name, sizeof(name), + "/sys/module/%s/notes/.note.gnu.build-id", line); + if (sysfs__read_build_id(name, dso->build_id, + sizeof(dso->build_id)) == 0) + dso->has_build_id = true; + dso->origin = DSO__ORIG_KMODULE; kernel_maps__insert(map); dsos__add(dso); diff --git a/tools/perf/util/symbol.h b/tools/perf/util/symbol.h index 546eb766d81..da7ec1af255 100644 --- a/tools/perf/util/symbol.h +++ b/tools/perf/util/symbol.h @@ -78,6 +78,7 @@ void dso__delete(struct dso *self); struct symbol *dso__find_symbol(struct dso *self, u64 ip); int dsos__load_kernel(const char *vmlinux, symbol_filter_t filter, int modules); +int dsos__load_modules(void); struct dso *dsos__findnew(const char *name); int dso__load(struct dso *self, struct map *map, symbol_filter_t filter); void dsos__fprintf(FILE *fp); @@ -89,6 +90,7 @@ char dso__symtab_origin(const struct dso *self); void dso__set_build_id(struct dso *self, void *build_id); int filename__read_build_id(const char *filename, void *bf, size_t size); +int sysfs__read_build_id(const char *filename, void *bf, size_t size); bool dsos__read_build_ids(void); int build_id__sprintf(u8 *self, int len, char *bf); -- cgit v1.2.3-18-g5258 From 2446042c93bfc6eeebfc89e88fdef2435d2bb5c4 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Wed, 18 Nov 2009 20:20:53 -0200 Subject: perf symbols: Capture the running kernel buildid too MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit [root@doppio linux-2.6-tip]# perf record -a -f sleep 3s ; perf buildid-list | grep vmlinux [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.171 MB perf.data (~7489 samples) ] 18e7cc53db62a7d35e9d6f6c9ddc23017d38ee9a vmlinux [root@doppio linux-2.6-tip]# Several refactorings were needed so that we can have symmetry between dsos__load_modules() and dsos__load_kernel(), i.e. those functions will respectively create and add to the dsos list the loaded modules and kernel, with its buildids, but not load its symbols. That is something the subcomands that need will have to call dso__load_kernel_sym(), just like we do with modules with dsos__load_module_sym()/dso__load_module_sym(). Next csets will actually use this info to stop producing bogus results using mismatched vmlinux and .ko files. Signed-off-by: Arnaldo Carvalho de Melo Cc: Roland McGrath Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1258582853-8579-4-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/builtin-top.c | 7 +++++- tools/perf/util/header.c | 1 + tools/perf/util/symbol.c | 65 +++++++++++++++++++++++++++--------------------- tools/perf/util/symbol.h | 3 ++- 4 files changed, 46 insertions(+), 30 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index 07b92c378ae..6d770ac7be0 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -948,7 +948,12 @@ static int symbol_filter(struct map *map, struct symbol *sym) static int parse_symbols(void) { - if (dsos__load_kernel(vmlinux_name, symbol_filter, 1) <= 0) + struct dso *kernel = dsos__load_kernel(); + + if (kernel == NULL) + return -1; + + if (dso__load_kernel_sym(kernel, symbol_filter, 1) <= 0) return -1; if (dump_symtab) diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index d3d656f9a62..425a29ba01a 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -241,6 +241,7 @@ perf_header__adds_write(struct perf_header *self, int fd) buildid_sec = &feat_sec[idx++]; + dsos__load_kernel(); /* * Read the list of loaded modules with its build_ids */ diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 7b4cedeb302..4d75e745288 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -1352,17 +1352,11 @@ static int dso__load_vmlinux(struct dso *self, struct map *map, return err; } -int dsos__load_kernel(const char *vmlinux, symbol_filter_t filter, - int use_modules) +int dso__load_kernel_sym(struct dso *self, symbol_filter_t filter, int use_modules) { int err = -1; - struct dso *dso = dso__new(vmlinux); - if (dso == NULL) - return -1; - - dso->short_name = "[kernel]"; - kernel_map = map__new2(0, dso); + kernel_map = map__new2(0, self); if (kernel_map == NULL) goto out_delete_dso; @@ -1374,39 +1368,36 @@ int dsos__load_kernel(const char *vmlinux, symbol_filter_t filter, use_modules = 0; } - if (vmlinux) { - err = dso__load_vmlinux(dso, kernel_map, vmlinux, filter); - if (err > 0 && use_modules) { - int syms = dsos__load_modules_sym(filter); + err = dso__load_vmlinux(self, kernel_map, self->name, filter); + if (err > 0 && use_modules) { + int syms = dsos__load_modules_sym(filter); - if (syms < 0) - pr_warning("Failed to read module symbols!" - " Continuing...\n"); - else - err += syms; - } + if (syms < 0) + pr_warning("Failed to read module symbols!" + " Continuing...\n"); + else + err += syms; } if (err <= 0) err = kernel_maps__load_kallsyms(filter, use_modules); if (err > 0) { - struct rb_node *node = rb_first(&dso->syms); + struct rb_node *node = rb_first(&self->syms); struct symbol *sym = rb_entry(node, struct symbol, rb_node); kernel_map->start = sym->start; - node = rb_last(&dso->syms); + node = rb_last(&self->syms); sym = rb_entry(node, struct symbol, rb_node); kernel_map->end = sym->end; - dso->origin = DSO__ORIG_KERNEL; + self->origin = DSO__ORIG_KERNEL; kernel_maps__insert(kernel_map); /* * Now that we have all sorted out, just set the ->end of all * maps: */ kernel_maps__fixup_end(); - dsos__add(dso); if (verbose) kernel_maps__fprintf(stderr); @@ -1415,7 +1406,7 @@ int dsos__load_kernel(const char *vmlinux, symbol_filter_t filter, return err; out_delete_dso: - dso__delete(dso); + dso__delete(self); return -1; } @@ -1475,18 +1466,36 @@ size_t dsos__fprintf_buildid(FILE *fp) return ret; } -int load_kernel(symbol_filter_t filter) +struct dso *dsos__load_kernel(void) { - if (dsos__load_kernel(vmlinux_name, filter, modules) <= 0) - return -1; + struct dso *kernel = dso__new(vmlinux_name); + if (kernel == NULL) + return NULL; + + kernel->short_name = "[kernel]"; vdso = dso__new("[vdso]"); if (!vdso) - return -1; + return NULL; + + if (sysfs__read_build_id("/sys/kernel/notes", kernel->build_id, + sizeof(kernel->build_id)) == 0) + kernel->has_build_id = true; + dsos__add(kernel); dsos__add(vdso); - return 0; + return kernel; +} + +int load_kernel(symbol_filter_t filter) +{ + struct dso *kernel = dsos__load_kernel(); + + if (kernel == NULL) + return -1; + + return dso__load_kernel_sym(kernel, filter, modules); } void symbol__init(unsigned int priv_size) diff --git a/tools/perf/util/symbol.h b/tools/perf/util/symbol.h index da7ec1af255..f0593a649c3 100644 --- a/tools/perf/util/symbol.h +++ b/tools/perf/util/symbol.h @@ -77,10 +77,10 @@ void dso__delete(struct dso *self); struct symbol *dso__find_symbol(struct dso *self, u64 ip); -int dsos__load_kernel(const char *vmlinux, symbol_filter_t filter, int modules); int dsos__load_modules(void); struct dso *dsos__findnew(const char *name); int dso__load(struct dso *self, struct map *map, symbol_filter_t filter); +int dso__load_kernel_sym(struct dso *self, symbol_filter_t filter, int modules); void dsos__fprintf(FILE *fp); size_t dsos__fprintf_buildid(FILE *fp); @@ -94,6 +94,7 @@ int sysfs__read_build_id(const char *filename, void *bf, size_t size); bool dsos__read_build_ids(void); int build_id__sprintf(u8 *self, int len, char *bf); +struct dso *dsos__load_kernel(void); int load_kernel(symbol_filter_t filter); void symbol__init(unsigned int priv_size); -- cgit v1.2.3-18-g5258 From 4dc0a04bb18fe9b80cefa08694f46a3a19ebfe50 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Thu, 19 Nov 2009 14:55:55 -0200 Subject: perf tools: perf_header__read() shouldn't die() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit And also don't call the constructor in it, this way it adheres to the model the other methods follow. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1258649757-17554-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/builtin-record.c | 16 +++++++------ tools/perf/builtin-timechart.c | 17 ++++++++++---- tools/perf/util/data_map.c | 18 +++++++++++---- tools/perf/util/header.c | 52 +++++++++++++++++++++++++----------------- tools/perf/util/header.h | 7 +++--- 5 files changed, 70 insertions(+), 40 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 82260c56db3..c97cb2ca8fa 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -400,7 +400,7 @@ static int __cmd_record(int argc, const char **argv) struct stat st; pid_t pid = 0; int flags; - int ret; + int err; unsigned long waking = 0; page_size = sysconf(_SC_PAGE_SIZE); @@ -434,16 +434,18 @@ static int __cmd_record(int argc, const char **argv) exit(-1); } - if (!file_new) - header = perf_header__read(output); - else - header = perf_header__new(); - + header = perf_header__new(); if (header == NULL) { pr_err("Not enough memory for reading perf file header\n"); return -1; } + if (!file_new) { + err = perf_header__read(header, output); + if (err < 0) + return err; + } + if (raw_samples) { perf_header__set_feat(header, HEADER_TRACE_INFO); } else { @@ -527,7 +529,7 @@ static int __cmd_record(int argc, const char **argv) if (hits == samples) { if (done) break; - ret = poll(event_array, nr_poll, -1); + err = poll(event_array, nr_poll, -1); waking++; } diff --git a/tools/perf/builtin-timechart.c b/tools/perf/builtin-timechart.c index 665877e4a94..dd4d82ac7aa 100644 --- a/tools/perf/builtin-timechart.c +++ b/tools/perf/builtin-timechart.c @@ -1093,7 +1093,7 @@ static void process_samples(void) static int __cmd_timechart(void) { - int ret, rc = EXIT_FAILURE; + int err, rc = EXIT_FAILURE; unsigned long offset = 0; unsigned long head, shift; struct stat statbuf; @@ -1111,8 +1111,8 @@ static int __cmd_timechart(void) exit(-1); } - ret = fstat(input, &statbuf); - if (ret < 0) { + err = fstat(input, &statbuf); + if (err < 0) { perror("failed to stat file"); exit(-1); } @@ -1122,7 +1122,16 @@ static int __cmd_timechart(void) exit(0); } - header = perf_header__read(input); + header = perf_header__new(); + if (header == NULL) + return -ENOMEM; + + err = perf_header__read(header, input); + if (err < 0) { + perf_header__delete(header); + return err; + } + head = header->data_offset; sample_type = perf_header__sample_type(header); diff --git a/tools/perf/util/data_map.c b/tools/perf/util/data_map.c index 14cb8465eb0..b8fc0fa2f63 100644 --- a/tools/perf/util/data_map.c +++ b/tools/perf/util/data_map.c @@ -106,7 +106,7 @@ int mmap_dispatch_perf_file(struct perf_header **pheader, int *cwdlen, char **cwd) { - int ret, rc = EXIT_FAILURE; + int err, rc = EXIT_FAILURE; struct perf_header *header; unsigned long head, shift; unsigned long offset = 0; @@ -132,8 +132,8 @@ int mmap_dispatch_perf_file(struct perf_header **pheader, exit(-1); } - ret = fstat(input, &input_stat); - if (ret < 0) { + err = fstat(input, &input_stat); + if (err < 0) { perror("failed to stat file"); exit(-1); } @@ -149,8 +149,16 @@ int mmap_dispatch_perf_file(struct perf_header **pheader, exit(0); } - *pheader = perf_header__read(input); - header = *pheader; + header = perf_header__new(); + if (header == NULL) + return -ENOMEM; + + err = perf_header__read(header, input); + if (err < 0) { + perf_header__delete(header); + return err; + } + *pheader = header; head = header->data_offset; sample_type = perf_header__sample_type(header); diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index 425a29ba01a..e66c7bd4cc8 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -78,16 +78,24 @@ struct perf_header *perf_header__new(void) return self; } +void perf_header__delete(struct perf_header *self) +{ + int i; + + for (i = 0; i < self->attrs; ++i) + perf_header_attr__delete(self->attr[i]); + + free(self->attr); + free(self); +} + int perf_header__add_attr(struct perf_header *self, struct perf_header_attr *attr) { - int pos = self->attrs; - if (self->frozen) return -1; - self->attrs++; - if (self->attrs > self->size) { + if (self->attrs == self->size) { int nsize = self->size * 2; struct perf_header_attr **nattr; @@ -98,7 +106,8 @@ int perf_header__add_attr(struct perf_header *self, self->size = nsize; self->attr = nattr; } - self->attr[pos] = attr; + + self->attr[self->attrs++] = attr; return 0; } @@ -441,19 +450,17 @@ static int perf_file_section__process(struct perf_file_section *self, return 0; } -struct perf_header *perf_header__read(int fd) +int perf_header__read(struct perf_header *self, int fd) { - struct perf_header *self = perf_header__new(); struct perf_file_header f_header; struct perf_file_attr f_attr; u64 f_id; int nr_attrs, nr_ids, i, j; - if (self == NULL) - die("nomem"); - - if (perf_file_header__read(&f_header, self, fd) < 0) - die("incompatible file format"); + if (perf_file_header__read(&f_header, self, fd) < 0) { + pr_debug("incompatible file format\n"); + return -EINVAL; + } nr_attrs = f_header.attrs.size / sizeof(f_attr); lseek(fd, f_header.attrs.offset, SEEK_SET); @@ -467,7 +474,7 @@ struct perf_header *perf_header__read(int fd) attr = perf_header_attr__new(&f_attr.attr); if (attr == NULL) - die("nomem"); + return -ENOMEM; nr_ids = f_attr.ids.size / sizeof(u64); lseek(fd, f_attr.ids.offset, SEEK_SET); @@ -475,11 +482,15 @@ struct perf_header *perf_header__read(int fd) for (j = 0; j < nr_ids; j++) { do_read(fd, &f_id, sizeof(f_id)); - if (perf_header_attr__add_id(attr, f_id) < 0) - die("nomem"); + if (perf_header_attr__add_id(attr, f_id) < 0) { + perf_header_attr__delete(attr); + return -ENOMEM; + } + } + if (perf_header__add_attr(self, attr) < 0) { + perf_header_attr__delete(attr); + return -ENOMEM; } - if (perf_header__add_attr(self, attr) < 0) - die("nomem"); lseek(fd, tmp, SEEK_SET); } @@ -487,8 +498,8 @@ struct perf_header *perf_header__read(int fd) if (f_header.event_types.size) { lseek(fd, f_header.event_types.offset, SEEK_SET); events = malloc(f_header.event_types.size); - if (!events) - die("nomem"); + if (events == NULL) + return -ENOMEM; do_read(fd, events, f_header.event_types.size); event_count = f_header.event_types.size / sizeof(struct perf_trace_event_type); } @@ -498,8 +509,7 @@ struct perf_header *perf_header__read(int fd) lseek(fd, self->data_offset, SEEK_SET); self->frozen = 1; - - return self; + return 0; } u64 perf_header__sample_type(struct perf_header *header) diff --git a/tools/perf/util/header.h b/tools/perf/util/header.h index f46a94e09ee..dc8fedb066a 100644 --- a/tools/perf/util/header.h +++ b/tools/perf/util/header.h @@ -55,7 +55,10 @@ struct perf_header { DECLARE_BITMAP(adds_features, HEADER_FEAT_BITS); }; -struct perf_header *perf_header__read(int fd); +struct perf_header *perf_header__new(void); +void perf_header__delete(struct perf_header *self); + +int perf_header__read(struct perf_header *self, int fd); void perf_header__write(struct perf_header *self, int fd, bool at_exit); int perf_header__add_attr(struct perf_header *self, @@ -75,8 +78,6 @@ perf_header__find_attr(u64 id, struct perf_header *header); void perf_header__set_feat(struct perf_header *self, int feat); bool perf_header__has_feat(const struct perf_header *self, int feat); -struct perf_header *perf_header__new(void); - int perf_header__process_sections(struct perf_header *self, int fd, int (*process)(struct perf_file_section *self, int feat, int fd)); -- cgit v1.2.3-18-g5258 From d5eed904bb6010b429b82c47e7cdb6a32f0c1343 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Thu, 19 Nov 2009 14:55:56 -0200 Subject: perf tools: Eliminate some more die() uses in library functions MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This time in perf_header__adds_write, propagating the do_write error returns. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1258649757-17554-2-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/builtin-record.c | 7 ++-- tools/perf/util/header.c | 80 ++++++++++++++++++++++++++++++--------------- tools/perf/util/header.h | 2 +- 3 files changed, 59 insertions(+), 30 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index c97cb2ca8fa..87f98fdb051 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -474,8 +474,11 @@ static int __cmd_record(int argc, const char **argv) } } - if (file_new) - perf_header__write(header, output, false); + if (file_new) { + err = perf_header__write(header, output, false); + if (err < 0) + return err; + } if (!system_wide) event__synthesize_thread(pid, process_synthesized_event); diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index e66c7bd4cc8..d5c81ebc0a8 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -176,7 +176,7 @@ static int do_write(int fd, const void *buf, size_t size) int ret = write(fd, buf, size); if (ret < 0) - return -1; + return -errno; size -= ret; buf += ret; @@ -190,6 +190,7 @@ static int dsos__write_buildid_table(int fd) struct dso *pos; list_for_each_entry(pos, &dsos, node) { + int err; struct build_id_event b; size_t len; @@ -200,33 +201,35 @@ static int dsos__write_buildid_table(int fd) memset(&b, 0, sizeof(b)); memcpy(&b.build_id, pos->build_id, sizeof(pos->build_id)); b.header.size = sizeof(b) + len; - if (do_write(fd, &b, sizeof(b)) < 0 || - do_write(fd, pos->long_name, len) < 0) - return -1; + err = do_write(fd, &b, sizeof(b)); + if (err < 0) + return err; + err = do_write(fd, pos->long_name, len); + if (err < 0) + return err; } return 0; } -static void -perf_header__adds_write(struct perf_header *self, int fd) +static int perf_header__adds_write(struct perf_header *self, int fd) { int nr_sections; struct perf_file_section *feat_sec; int sec_size; u64 sec_start; - int idx = 0; + int idx = 0, err; if (dsos__read_build_ids()) perf_header__set_feat(self, HEADER_BUILD_ID); nr_sections = bitmap_weight(self->adds_features, HEADER_FEAT_BITS); if (!nr_sections) - return; + return 0; feat_sec = calloc(sizeof(*feat_sec), nr_sections); - if (!feat_sec) - die("No memory"); + if (feat_sec == NULL) + return -ENOMEM; sec_size = sizeof(*feat_sec) * nr_sections; @@ -258,23 +261,29 @@ perf_header__adds_write(struct perf_header *self, int fd) /* Write build-ids */ buildid_sec->offset = lseek(fd, 0, SEEK_CUR); - if (dsos__write_buildid_table(fd) < 0) - die("failed to write buildid table"); + err = dsos__write_buildid_table(fd); + if (err < 0) { + pr_debug("failed to write buildid table\n"); + goto out_free; + } buildid_sec->size = lseek(fd, 0, SEEK_CUR) - buildid_sec->offset; } lseek(fd, sec_start, SEEK_SET); - if (do_write(fd, feat_sec, sec_size) < 0) - die("failed to write feature section"); + err = do_write(fd, feat_sec, sec_size); + if (err < 0) + pr_debug("failed to write feature section\n"); +out_free: free(feat_sec); + return err; } -void perf_header__write(struct perf_header *self, int fd, bool at_exit) +int perf_header__write(struct perf_header *self, int fd, bool at_exit) { struct perf_file_header f_header; struct perf_file_attr f_attr; struct perf_header_attr *attr; - int i; + int i, err; lseek(fd, sizeof(f_header), SEEK_SET); @@ -283,8 +292,11 @@ void perf_header__write(struct perf_header *self, int fd, bool at_exit) attr = self->attr[i]; attr->id_offset = lseek(fd, 0, SEEK_CUR); - if (do_write(fd, attr->id, attr->ids * sizeof(u64)) < 0) - die("failed to write perf header"); + err = do_write(fd, attr->id, attr->ids * sizeof(u64)); + if (err < 0) { + pr_debug("failed to write perf header\n"); + return err; + } } @@ -300,20 +312,30 @@ void perf_header__write(struct perf_header *self, int fd, bool at_exit) .size = attr->ids * sizeof(u64), } }; - if (do_write(fd, &f_attr, sizeof(f_attr)) < 0) - die("failed to write perf header attribute"); + err = do_write(fd, &f_attr, sizeof(f_attr)); + if (err < 0) { + pr_debug("failed to write perf header attribute\n"); + return err; + } } self->event_offset = lseek(fd, 0, SEEK_CUR); self->event_size = event_count * sizeof(struct perf_trace_event_type); - if (events) - if (do_write(fd, events, self->event_size) < 0) - die("failed to write perf header events"); + if (events) { + err = do_write(fd, events, self->event_size); + if (err < 0) { + pr_debug("failed to write perf header events\n"); + return err; + } + } self->data_offset = lseek(fd, 0, SEEK_CUR); - if (at_exit) - perf_header__adds_write(self, fd); + if (at_exit) { + err = perf_header__adds_write(self, fd); + if (err < 0) + return err; + } f_header = (struct perf_file_header){ .magic = PERF_MAGIC, @@ -336,11 +358,15 @@ void perf_header__write(struct perf_header *self, int fd, bool at_exit) memcpy(&f_header.adds_features, &self->adds_features, sizeof(self->adds_features)); lseek(fd, 0, SEEK_SET); - if (do_write(fd, &f_header, sizeof(f_header)) < 0) - die("failed to write perf header"); + err = do_write(fd, &f_header, sizeof(f_header)); + if (err < 0) { + pr_debug("failed to write perf header\n"); + return err; + } lseek(fd, self->data_offset + self->data_size, SEEK_SET); self->frozen = 1; + return 0; } static void do_read(int fd, void *buf, size_t size) diff --git a/tools/perf/util/header.h b/tools/perf/util/header.h index dc8fedb066a..d1dbe2b79c4 100644 --- a/tools/perf/util/header.h +++ b/tools/perf/util/header.h @@ -59,7 +59,7 @@ struct perf_header *perf_header__new(void); void perf_header__delete(struct perf_header *self); int perf_header__read(struct perf_header *self, int fd); -void perf_header__write(struct perf_header *self, int fd, bool at_exit); +int perf_header__write(struct perf_header *self, int fd, bool at_exit); int perf_header__add_attr(struct perf_header *self, struct perf_header_attr *attr); -- cgit v1.2.3-18-g5258 From 6b0cb5f9f7033c72b19697c33deab83f0dd9848d Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Thu, 19 Nov 2009 14:55:57 -0200 Subject: perf tools: Don't die() in mmap_dispatch_perf_file MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Propagate the error, that, interestingly, are already handled by all callers :-) Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1258649757-17554-3-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/util/data_map.c | 80 +++++++++++++++++++++++++--------------------- 1 file changed, 44 insertions(+), 36 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/data_map.c b/tools/perf/util/data_map.c index b8fc0fa2f63..5543e7d0487 100644 --- a/tools/perf/util/data_map.c +++ b/tools/perf/util/data_map.c @@ -106,7 +106,7 @@ int mmap_dispatch_perf_file(struct perf_header **pheader, int *cwdlen, char **cwd) { - int err, rc = EXIT_FAILURE; + int err; struct perf_header *header; unsigned long head, shift; unsigned long offset = 0; @@ -118,64 +118,69 @@ int mmap_dispatch_perf_file(struct perf_header **pheader, int input; char *buf; - if (!curr_handler) - die("Forgot to register perf file handler"); + if (curr_handler == NULL) { + pr_debug("Forgot to register perf file handler\n"); + return -EINVAL; + } page_size = getpagesize(); input = open(input_name, O_RDONLY); if (input < 0) { - fprintf(stderr, " failed to open file: %s", input_name); + pr_err("Failed to open file: %s", input_name); if (!strcmp(input_name, "perf.data")) - fprintf(stderr, " (try 'perf record' first)"); - fprintf(stderr, "\n"); - exit(-1); + pr_err(" (try 'perf record' first)"); + pr_err("\n"); + return -errno; } - err = fstat(input, &input_stat); - if (err < 0) { - perror("failed to stat file"); - exit(-1); + if (fstat(input, &input_stat) < 0) { + pr_err("failed to stat file"); + err = -errno; + goto out_close; } + err = -EACCES; if (!force && input_stat.st_uid && (input_stat.st_uid != geteuid())) { - fprintf(stderr, "file: %s not owned by current user or root\n", + pr_err("file: %s not owned by current user or root\n", input_name); - exit(-1); + goto out_close; } - if (!input_stat.st_size) { - fprintf(stderr, "zero-sized file, nothing to do!\n"); - exit(0); + if (input_stat.st_size == 0) { + pr_info("zero-sized file, nothing to do!\n"); + goto done; } + err = -ENOMEM; header = perf_header__new(); if (header == NULL) - return -ENOMEM; + goto out_close; err = perf_header__read(header, input); - if (err < 0) { - perf_header__delete(header); - return err; - } + if (err < 0) + goto out_delete; *pheader = header; head = header->data_offset; sample_type = perf_header__sample_type(header); - if (curr_handler->sample_type_check) - if (curr_handler->sample_type_check(sample_type) < 0) - exit(-1); + err = -EINVAL; + if (curr_handler->sample_type_check && + curr_handler->sample_type_check(sample_type) < 0) + goto out_delete; + err = -ENOMEM; if (load_kernel(NULL) < 0) { - perror("failed to load kernel symbols"); - return EXIT_FAILURE; + pr_err("failed to load kernel symbols\n"); + goto out_delete; } if (!full_paths) { if (getcwd(__cwd, sizeof(__cwd)) == NULL) { - perror("failed to get the current directory"); - return EXIT_FAILURE; + pr_err("failed to get the current directory\n"); + err = -errno; + goto out_delete; } *cwd = __cwd; *cwdlen = strlen(*cwd); @@ -189,11 +194,12 @@ int mmap_dispatch_perf_file(struct perf_header **pheader, head -= shift; remap: - buf = (char *)mmap(NULL, page_size * mmap_window, PROT_READ, - MAP_SHARED, input, offset); + buf = mmap(NULL, page_size * mmap_window, PROT_READ, + MAP_SHARED, input, offset); if (buf == MAP_FAILED) { - perror("failed to mmap file"); - exit(-1); + pr_err("failed to mmap file\n"); + err = -errno; + goto out_delete; } more: @@ -250,10 +256,12 @@ more: goto more; done: - rc = EXIT_SUCCESS; + err = 0; +out_close: close(input); - return rc; + return err; +out_delete: + perf_header__delete(header); + goto out_close; } - - -- cgit v1.2.3-18-g5258 From ba77c9e11111a172c9e8687fe16a6a173a61916f Mon Sep 17 00:00:00 2001 From: Li Zefan Date: Fri, 20 Nov 2009 15:53:25 +0800 Subject: perf: Add 'perf kmem' tool This tool is mostly a perf version of kmemtrace-user. The following information is provided by this tool: - the total amount of memory allocated and fragmentation per call-site - the total amount of memory allocated and fragmentation per allocation - total memory allocated and fragmentation in the collected dataset - ... Sample output: # ./perf kmem record ^C # ./perf kmem --stat caller --stat alloc -l 10 ------------------------------------------------------------------------------ Callsite | Total_alloc/Per | Total_req/Per | Hit | Fragmentation ------------------------------------------------------------------------------ 0xc052f37a | 790528/4096 | 790528/4096 | 193 | 0.000% 0xc0541d70 | 524288/4096 | 524288/4096 | 128 | 0.000% 0xc051cc68 | 481600/200 | 481600/200 | 2408 | 0.000% 0xc0572623 | 297444/676 | 297440/676 | 440 | 0.001% 0xc05399f1 | 73476/164 | 73472/164 | 448 | 0.005% 0xc05243bf | 51456/256 | 51456/256 | 201 | 0.000% 0xc0730d0e | 31844/497 | 31808/497 | 64 | 0.113% 0xc0734c4e | 17152/256 | 17152/256 | 67 | 0.000% 0xc0541a6d | 16384/128 | 16384/128 | 128 | 0.000% 0xc059c217 | 13120/40 | 13120/40 | 328 | 0.000% 0xc0501ee6 | 11264/88 | 11264/88 | 128 | 0.000% 0xc04daef0 | 7504/682 | 7128/648 | 11 | 5.011% 0xc04e14a3 | 4216/191 | 4216/191 | 22 | 0.000% 0xc05041ca | 3524/44 | 3520/44 | 80 | 0.114% 0xc0734fa3 | 2104/701 | 1620/540 | 3 | 23.004% 0xc05ec9f1 | 2024/289 | 2016/288 | 7 | 0.395% 0xc06a1999 | 1792/256 | 1792/256 | 7 | 0.000% 0xc0463b9a | 1584/144 | 1584/144 | 11 | 0.000% 0xc0541eb0 | 1024/16 | 1024/16 | 64 | 0.000% 0xc06a19ac | 896/128 | 896/128 | 7 | 0.000% 0xc05721c0 | 772/12 | 768/12 | 64 | 0.518% 0xc054d1e6 | 288/57 | 280/56 | 5 | 2.778% 0xc04b562e | 157/31 | 154/30 | 5 | 1.911% 0xc04b536f | 80/16 | 80/16 | 5 | 0.000% 0xc05855a0 | 64/64 | 36/36 | 1 | 43.750% ------------------------------------------------------------------------------ ------------------------------------------------------------------------------ Alloc Ptr | Total_alloc/Per | Total_req/Per | Hit | Fragmentation ------------------------------------------------------------------------------ 0xda884000 | 1052672/4096 | 1052672/4096 | 257 | 0.000% 0xda886000 | 262144/4096 | 262144/4096 | 64 | 0.000% 0xf60c7c00 | 16512/128 | 16512/128 | 129 | 0.000% 0xf59a4118 | 13120/40 | 13120/40 | 328 | 0.000% 0xdfd4b2c0 | 11264/88 | 11264/88 | 128 | 0.000% 0xf5274600 | 7680/256 | 7680/256 | 30 | 0.000% 0xe8395000 | 5948/594 | 5464/546 | 10 | 8.137% 0xe59c3c00 | 5748/479 | 5712/476 | 12 | 0.626% 0xf4cd1a80 | 3524/44 | 3520/44 | 80 | 0.114% 0xe5bd1600 | 2892/482 | 2856/476 | 6 | 1.245% ... | ... | ... | ... | ... ------------------------------------------------------------------------------ SUMMARY ======= Total bytes requested: 2333626 Total bytes allocated: 2353712 Total bytes wasted on internal fragmentation: 20086 Internal fragmentation: 0.853375% TODO: - show sym+offset in 'callsite' column - show cross node allocation stats - collect more useful stats? - ... Signed-off-by: Li Zefan Acked-by: Pekka Enberg Acked-by: Peter Zijlstra Cc: Frederic Weisbecker Cc: Steven Rostedt Cc: Eduard - Gabriel Munteanu Cc: linux-mm@kvack.org LKML-Reference: <4B064AF5.9060208@cn.fujitsu.com> Signed-off-by: Ingo Molnar --- tools/perf/Makefile | 1 + tools/perf/builtin-kmem.c | 578 ++++++++++++++++++++++++++++++++++++++++++++++ tools/perf/builtin.h | 1 + tools/perf/perf.c | 27 +-- 4 files changed, 594 insertions(+), 13 deletions(-) create mode 100644 tools/perf/builtin-kmem.c (limited to 'tools/perf') diff --git a/tools/perf/Makefile b/tools/perf/Makefile index 3f0666af93d..d7198c54bb6 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -444,6 +444,7 @@ BUILTIN_OBJS += builtin-timechart.o BUILTIN_OBJS += builtin-top.o BUILTIN_OBJS += builtin-trace.o BUILTIN_OBJS += builtin-probe.o +BUILTIN_OBJS += builtin-kmem.o PERFLIBS = $(LIB_FILE) diff --git a/tools/perf/builtin-kmem.c b/tools/perf/builtin-kmem.c new file mode 100644 index 00000000000..f315b052f81 --- /dev/null +++ b/tools/perf/builtin-kmem.c @@ -0,0 +1,578 @@ +#include "builtin.h" +#include "perf.h" + +#include "util/util.h" +#include "util/cache.h" +#include "util/symbol.h" +#include "util/thread.h" +#include "util/header.h" + +#include "util/parse-options.h" +#include "util/trace-event.h" + +#include "util/debug.h" +#include "util/data_map.h" + +#include + +struct alloc_stat; +typedef int (*sort_fn_t)(struct alloc_stat *, struct alloc_stat *); + +static char const *input_name = "perf.data"; + +static struct perf_header *header; +static u64 sample_type; + +static int alloc_flag; +static int caller_flag; + +sort_fn_t alloc_sort_fn; +sort_fn_t caller_sort_fn; + +static int alloc_lines = -1; +static int caller_lines = -1; + +static char *cwd; +static int cwdlen; + +struct alloc_stat { + union { + struct { + char *name; + u64 call_site; + }; + u64 ptr; + }; + u64 bytes_req; + u64 bytes_alloc; + u32 hit; + + struct rb_node node; +}; + +static struct rb_root root_alloc_stat; +static struct rb_root root_alloc_sorted; +static struct rb_root root_caller_stat; +static struct rb_root root_caller_sorted; + +static unsigned long total_requested, total_allocated; + +struct raw_event_sample { + u32 size; + char data[0]; +}; + +static int +process_comm_event(event_t *event, unsigned long offset, unsigned long head) +{ + struct thread *thread = threads__findnew(event->comm.pid); + + dump_printf("%p [%p]: PERF_RECORD_COMM: %s:%d\n", + (void *)(offset + head), + (void *)(long)(event->header.size), + event->comm.comm, event->comm.pid); + + if (thread == NULL || + thread__set_comm(thread, event->comm.comm)) { + dump_printf("problem processing PERF_RECORD_COMM, skipping event.\n"); + return -1; + } + + return 0; +} + +static void insert_alloc_stat(unsigned long ptr, + int bytes_req, int bytes_alloc) +{ + struct rb_node **node = &root_alloc_stat.rb_node; + struct rb_node *parent = NULL; + struct alloc_stat *data = NULL; + + if (!alloc_flag) + return; + + while (*node) { + parent = *node; + data = rb_entry(*node, struct alloc_stat, node); + + if (ptr > data->ptr) + node = &(*node)->rb_right; + else if (ptr < data->ptr) + node = &(*node)->rb_left; + else + break; + } + + if (data && data->ptr == ptr) { + data->hit++; + data->bytes_req += bytes_req; + data->bytes_alloc += bytes_req; + } else { + data = malloc(sizeof(*data)); + data->ptr = ptr; + data->hit = 1; + data->bytes_req = bytes_req; + data->bytes_alloc = bytes_alloc; + + rb_link_node(&data->node, parent, node); + rb_insert_color(&data->node, &root_alloc_stat); + } +} + +static void insert_caller_stat(unsigned long call_site, + int bytes_req, int bytes_alloc) +{ + struct rb_node **node = &root_caller_stat.rb_node; + struct rb_node *parent = NULL; + struct alloc_stat *data = NULL; + + if (!caller_flag) + return; + + while (*node) { + parent = *node; + data = rb_entry(*node, struct alloc_stat, node); + + if (call_site > data->call_site) + node = &(*node)->rb_right; + else if (call_site < data->call_site) + node = &(*node)->rb_left; + else + break; + } + + if (data && data->call_site == call_site) { + data->hit++; + data->bytes_req += bytes_req; + data->bytes_alloc += bytes_req; + } else { + data = malloc(sizeof(*data)); + data->call_site = call_site; + data->hit = 1; + data->bytes_req = bytes_req; + data->bytes_alloc = bytes_alloc; + + rb_link_node(&data->node, parent, node); + rb_insert_color(&data->node, &root_caller_stat); + } +} + +static void process_alloc_event(struct raw_event_sample *raw, + struct event *event, + int cpu __used, + u64 timestamp __used, + struct thread *thread __used, + int node __used) +{ + unsigned long call_site; + unsigned long ptr; + int bytes_req; + int bytes_alloc; + + ptr = raw_field_value(event, "ptr", raw->data); + call_site = raw_field_value(event, "call_site", raw->data); + bytes_req = raw_field_value(event, "bytes_req", raw->data); + bytes_alloc = raw_field_value(event, "bytes_alloc", raw->data); + + insert_alloc_stat(ptr, bytes_req, bytes_alloc); + insert_caller_stat(call_site, bytes_req, bytes_alloc); + + total_requested += bytes_req; + total_allocated += bytes_alloc; +} + +static void process_free_event(struct raw_event_sample *raw __used, + struct event *event __used, + int cpu __used, + u64 timestamp __used, + struct thread *thread __used) +{ +} + +static void +process_raw_event(event_t *raw_event __used, void *more_data, + int cpu, u64 timestamp, struct thread *thread) +{ + struct raw_event_sample *raw = more_data; + struct event *event; + int type; + + type = trace_parse_common_type(raw->data); + event = trace_find_event(type); + + if (!strcmp(event->name, "kmalloc") || + !strcmp(event->name, "kmem_cache_alloc")) { + process_alloc_event(raw, event, cpu, timestamp, thread, 0); + return; + } + + if (!strcmp(event->name, "kmalloc_node") || + !strcmp(event->name, "kmem_cache_alloc_node")) { + process_alloc_event(raw, event, cpu, timestamp, thread, 1); + return; + } + + if (!strcmp(event->name, "kfree") || + !strcmp(event->name, "kmem_cache_free")) { + process_free_event(raw, event, cpu, timestamp, thread); + return; + } +} + +static int +process_sample_event(event_t *event, unsigned long offset, unsigned long head) +{ + u64 ip = event->ip.ip; + u64 timestamp = -1; + u32 cpu = -1; + u64 period = 1; + void *more_data = event->ip.__more_data; + struct thread *thread = threads__findnew(event->ip.pid); + + if (sample_type & PERF_SAMPLE_TIME) { + timestamp = *(u64 *)more_data; + more_data += sizeof(u64); + } + + if (sample_type & PERF_SAMPLE_CPU) { + cpu = *(u32 *)more_data; + more_data += sizeof(u32); + more_data += sizeof(u32); /* reserved */ + } + + if (sample_type & PERF_SAMPLE_PERIOD) { + period = *(u64 *)more_data; + more_data += sizeof(u64); + } + + dump_printf("%p [%p]: PERF_RECORD_SAMPLE (IP, %d): %d/%d: %p period: %Ld\n", + (void *)(offset + head), + (void *)(long)(event->header.size), + event->header.misc, + event->ip.pid, event->ip.tid, + (void *)(long)ip, + (long long)period); + + if (thread == NULL) { + pr_debug("problem processing %d event, skipping it.\n", + event->header.type); + return -1; + } + + dump_printf(" ... thread: %s:%d\n", thread->comm, thread->pid); + + process_raw_event(event, more_data, cpu, timestamp, thread); + + return 0; +} + +static int sample_type_check(u64 type) +{ + sample_type = type; + + if (!(sample_type & PERF_SAMPLE_RAW)) { + fprintf(stderr, + "No trace sample to read. Did you call perf record " + "without -R?"); + return -1; + } + + return 0; +} + +static struct perf_file_handler file_handler = { + .process_sample_event = process_sample_event, + .process_comm_event = process_comm_event, + .sample_type_check = sample_type_check, +}; + +static int read_events(void) +{ + register_idle_thread(); + register_perf_file_handler(&file_handler); + + return mmap_dispatch_perf_file(&header, input_name, 0, 0, + &cwdlen, &cwd); +} + +static double fragmentation(unsigned long n_req, unsigned long n_alloc) +{ + if (n_alloc == 0) + return 0.0; + else + return 100.0 - (100.0 * n_req / n_alloc); +} + +static void __print_result(struct rb_root *root, int n_lines, int is_caller) +{ + struct rb_node *next; + + printf("\n ------------------------------------------------------------------------------\n"); + if (is_caller) + printf(" Callsite |"); + else + printf(" Alloc Ptr |"); + printf(" Total_alloc/Per | Total_req/Per | Hit | Fragmentation\n"); + printf(" ------------------------------------------------------------------------------\n"); + + next = rb_first(root); + + while (next && n_lines--) { + struct alloc_stat *data; + + data = rb_entry(next, struct alloc_stat, node); + + printf(" %-16p | %8llu/%-6lu | %8llu/%-6lu | %6lu | %8.3f%%\n", + is_caller ? (void *)(unsigned long)data->call_site : + (void *)(unsigned long)data->ptr, + (unsigned long long)data->bytes_alloc, + (unsigned long)data->bytes_alloc / data->hit, + (unsigned long long)data->bytes_req, + (unsigned long)data->bytes_req / data->hit, + (unsigned long)data->hit, + fragmentation(data->bytes_req, data->bytes_alloc)); + + next = rb_next(next); + } + + if (n_lines == -1) + printf(" ... | ... | ... | ... | ... \n"); + + printf(" ------------------------------------------------------------------------------\n"); +} + +static void print_summary(void) +{ + printf("\nSUMMARY\n=======\n"); + printf("Total bytes requested: %lu\n", total_requested); + printf("Total bytes allocated: %lu\n", total_allocated); + printf("Total bytes wasted on internal fragmentation: %lu\n", + total_allocated - total_requested); + printf("Internal fragmentation: %f%%\n", + fragmentation(total_requested, total_allocated)); +} + +static void print_result(void) +{ + if (caller_flag) + __print_result(&root_caller_sorted, caller_lines, 1); + if (alloc_flag) + __print_result(&root_alloc_sorted, alloc_lines, 0); + print_summary(); +} + +static void sort_insert(struct rb_root *root, struct alloc_stat *data, + sort_fn_t sort_fn) +{ + struct rb_node **new = &(root->rb_node); + struct rb_node *parent = NULL; + + while (*new) { + struct alloc_stat *this; + int cmp; + + this = rb_entry(*new, struct alloc_stat, node); + parent = *new; + + cmp = sort_fn(data, this); + + if (cmp > 0) + new = &((*new)->rb_left); + else + new = &((*new)->rb_right); + } + + rb_link_node(&data->node, parent, new); + rb_insert_color(&data->node, root); +} + +static void __sort_result(struct rb_root *root, struct rb_root *root_sorted, + sort_fn_t sort_fn) +{ + struct rb_node *node; + struct alloc_stat *data; + + for (;;) { + node = rb_first(root); + if (!node) + break; + + rb_erase(node, root); + data = rb_entry(node, struct alloc_stat, node); + sort_insert(root_sorted, data, sort_fn); + } +} + +static void sort_result(void) +{ + __sort_result(&root_alloc_stat, &root_alloc_sorted, alloc_sort_fn); + __sort_result(&root_caller_stat, &root_caller_sorted, caller_sort_fn); +} + +static int __cmd_kmem(void) +{ + setup_pager(); + read_events(); + sort_result(); + print_result(); + + return 0; +} + +static const char * const kmem_usage[] = { + "perf kmem [] {record}", + NULL +}; + + +static int ptr_cmp(struct alloc_stat *l, struct alloc_stat *r) +{ + if (l->ptr < r->ptr) + return -1; + else if (l->ptr > r->ptr) + return 1; + return 0; +} + +static int callsite_cmp(struct alloc_stat *l, struct alloc_stat *r) +{ + if (l->call_site < r->call_site) + return -1; + else if (l->call_site > r->call_site) + return 1; + return 0; +} + +static int bytes_cmp(struct alloc_stat *l, struct alloc_stat *r) +{ + if (l->bytes_alloc < r->bytes_alloc) + return -1; + else if (l->bytes_alloc > r->bytes_alloc) + return 1; + return 0; +} + +static int parse_sort_opt(const struct option *opt __used, + const char *arg, int unset __used) +{ + sort_fn_t sort_fn; + + if (!arg) + return -1; + + if (strcmp(arg, "ptr") == 0) + sort_fn = ptr_cmp; + else if (strcmp(arg, "call_site") == 0) + sort_fn = callsite_cmp; + else if (strcmp(arg, "bytes") == 0) + sort_fn = bytes_cmp; + else + return -1; + + if (caller_flag > alloc_flag) + caller_sort_fn = sort_fn; + else + alloc_sort_fn = sort_fn; + + return 0; +} + +static int parse_stat_opt(const struct option *opt __used, + const char *arg, int unset __used) +{ + if (!arg) + return -1; + + if (strcmp(arg, "alloc") == 0) + alloc_flag = (caller_flag + 1); + else if (strcmp(arg, "caller") == 0) + caller_flag = (alloc_flag + 1); + else + return -1; + return 0; +} + +static int parse_line_opt(const struct option *opt __used, + const char *arg, int unset __used) +{ + int lines; + + if (!arg) + return -1; + + lines = strtoul(arg, NULL, 10); + + if (caller_flag > alloc_flag) + caller_lines = lines; + else + alloc_lines = lines; + + return 0; +} + +static const struct option kmem_options[] = { + OPT_STRING('i', "input", &input_name, "file", + "input file name"), + OPT_CALLBACK(0, "stat", NULL, "|", + "stat selector, Pass 'alloc' or 'caller'.", + parse_stat_opt), + OPT_CALLBACK('s', "sort", NULL, "key", + "sort by key: ptr, call_site, hit, bytes", + parse_sort_opt), + OPT_CALLBACK('l', "line", NULL, "num", + "show n lins", + parse_line_opt), + OPT_END() +}; + +static const char *record_args[] = { + "record", + "-a", + "-R", + "-M", + "-f", + "-c", "1", + "-e", "kmem:kmalloc", + "-e", "kmem:kmalloc_node", + "-e", "kmem:kfree", + "-e", "kmem:kmem_cache_alloc", + "-e", "kmem:kmem_cache_alloc_node", + "-e", "kmem:kmem_cache_free", +}; + +static int __cmd_record(int argc, const char **argv) +{ + unsigned int rec_argc, i, j; + const char **rec_argv; + + rec_argc = ARRAY_SIZE(record_args) + argc - 1; + rec_argv = calloc(rec_argc + 1, sizeof(char *)); + + for (i = 0; i < ARRAY_SIZE(record_args); i++) + rec_argv[i] = strdup(record_args[i]); + + for (j = 1; j < (unsigned int)argc; j++, i++) + rec_argv[i] = argv[j]; + + return cmd_record(i, rec_argv, NULL); +} + +int cmd_kmem(int argc, const char **argv, const char *prefix __used) +{ + symbol__init(0); + + argc = parse_options(argc, argv, kmem_options, kmem_usage, 0); + + if (argc && !strncmp(argv[0], "rec", 3)) + return __cmd_record(argc, argv); + else if (argc) + usage_with_options(kmem_usage, kmem_options); + + if (!alloc_sort_fn) + alloc_sort_fn = bytes_cmp; + if (!caller_sort_fn) + caller_sort_fn = bytes_cmp; + + return __cmd_kmem(); +} + diff --git a/tools/perf/builtin.h b/tools/perf/builtin.h index 9b02d85091f..a3d8bf65f26 100644 --- a/tools/perf/builtin.h +++ b/tools/perf/builtin.h @@ -28,5 +28,6 @@ extern int cmd_top(int argc, const char **argv, const char *prefix); extern int cmd_trace(int argc, const char **argv, const char *prefix); extern int cmd_version(int argc, const char **argv, const char *prefix); extern int cmd_probe(int argc, const char **argv, const char *prefix); +extern int cmd_kmem(int argc, const char **argv, const char *prefix); #endif diff --git a/tools/perf/perf.c b/tools/perf/perf.c index 89b82acac7d..cf64049bc9b 100644 --- a/tools/perf/perf.c +++ b/tools/perf/perf.c @@ -285,20 +285,21 @@ static void handle_internal_command(int argc, const char **argv) { const char *cmd = argv[0]; static struct cmd_struct commands[] = { - { "help", cmd_help, 0 }, - { "list", cmd_list, 0 }, { "buildid-list", cmd_buildid_list, 0 }, - { "record", cmd_record, 0 }, - { "report", cmd_report, 0 }, - { "bench", cmd_bench, 0 }, - { "stat", cmd_stat, 0 }, - { "timechart", cmd_timechart, 0 }, - { "top", cmd_top, 0 }, - { "annotate", cmd_annotate, 0 }, - { "version", cmd_version, 0 }, - { "trace", cmd_trace, 0 }, - { "sched", cmd_sched, 0 }, - { "probe", cmd_probe, 0 }, + { "help", cmd_help, 0 }, + { "list", cmd_list, 0 }, + { "record", cmd_record, 0 }, + { "report", cmd_report, 0 }, + { "bench", cmd_bench, 0 }, + { "stat", cmd_stat, 0 }, + { "timechart", cmd_timechart, 0 }, + { "top", cmd_top, 0 }, + { "annotate", cmd_annotate, 0 }, + { "version", cmd_version, 0 }, + { "trace", cmd_trace, 0 }, + { "sched", cmd_sched, 0 }, + { "probe", cmd_probe, 0 }, + { "kmem", cmd_kmem, 0 }, }; unsigned int i; static const char ext[] = STRIP_EXTENSION; -- cgit v1.2.3-18-g5258 From 6671cb1674e69e2aba3d610714bdd3e97a7b51ff Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 20 Nov 2009 20:51:24 -0200 Subject: perf symbols: Remove unrelated actions from dso__load_kernel_sym MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit It should just load kernel symbols, not load the list of modules. There are more stuff to move to other routines, but lets do it in several steps. End goal is to be able to defer symbol table loading till we find a hit for that map address range. So that the kernel & modules are handled just like all the other DSOs in the system. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1258757489-5978-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/builtin-annotate.c | 5 +++-- tools/perf/builtin-report.c | 3 ++- tools/perf/builtin-top.c | 10 +++++++++- tools/perf/util/data_map.c | 2 +- tools/perf/util/symbol.c | 46 ++++++++++++++++++------------------------- tools/perf/util/symbol.h | 7 ++++--- 6 files changed, 38 insertions(+), 35 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c index 77d50a6d680..b6da1476ab1 100644 --- a/tools/perf/builtin-annotate.c +++ b/tools/perf/builtin-annotate.c @@ -33,6 +33,7 @@ static int input; static int full_paths; static int print_line; +static bool use_modules; static unsigned long page_size; static unsigned long mmap_window = 32; @@ -636,7 +637,7 @@ static int __cmd_annotate(void) exit(0); } - if (load_kernel(symbol_filter) < 0) { + if (load_kernel(symbol_filter, use_modules) < 0) { perror("failed to load kernel symbols"); return EXIT_FAILURE; } @@ -742,7 +743,7 @@ static const struct option options[] = { OPT_BOOLEAN('D', "dump-raw-trace", &dump_trace, "dump raw trace in ASCII"), OPT_STRING('k', "vmlinux", &vmlinux_name, "file", "vmlinux pathname"), - OPT_BOOLEAN('m', "modules", &modules, + OPT_BOOLEAN('m', "modules", &use_modules, "load module symbols - WARNING: use only with -k and LIVE kernel"), OPT_BOOLEAN('l', "print-line", &print_line, "print matching source lines (may be slow)"), diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index 1a806d5f05c..0af48401f08 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -38,6 +38,7 @@ static char *dso_list_str, *comm_list_str, *sym_list_str, static struct strlist *dso_list, *comm_list, *sym_list; static int force; +static bool use_modules; static int full_paths; static int show_nr_samples; @@ -1023,7 +1024,7 @@ static const struct option options[] = { "dump raw trace in ASCII"), OPT_STRING('k', "vmlinux", &vmlinux_name, "file", "vmlinux pathname"), OPT_BOOLEAN('f', "force", &force, "don't complain, do it"), - OPT_BOOLEAN('m', "modules", &modules, + OPT_BOOLEAN('m', "modules", &use_modules, "load module symbols - WARNING: use only with -k and LIVE kernel"), OPT_BOOLEAN('n', "show-nr-samples", &show_nr_samples, "Show a column with the number of samples"), diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index 6d770ac7be0..48cc1084bc3 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -953,8 +953,16 @@ static int parse_symbols(void) if (kernel == NULL) return -1; + if (dsos__load_modules() < 0) + pr_debug("Couldn't read the complete list of modules, " + "continuing...\n"); + + if (dsos__load_modules_sym(symbol_filter) < 0) + pr_warning("Failed to read module symbols, continuing...\n"); + if (dso__load_kernel_sym(kernel, symbol_filter, 1) <= 0) - return -1; + pr_debug("Couldn't read the complete list of kernel symbols, " + "continuing...\n"); if (dump_symtab) dsos__fprintf(stderr); diff --git a/tools/perf/util/data_map.c b/tools/perf/util/data_map.c index 5543e7d0487..a444a2645c8 100644 --- a/tools/perf/util/data_map.c +++ b/tools/perf/util/data_map.c @@ -171,7 +171,7 @@ int mmap_dispatch_perf_file(struct perf_header **pheader, goto out_delete; err = -ENOMEM; - if (load_kernel(NULL) < 0) { + if (load_kernel(NULL, 1) < 0) { pr_err("failed to load kernel symbols\n"); goto out_delete; } diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 4d75e745288..3b23c18cd36 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -1230,7 +1230,7 @@ failure: return -1; } -static int dsos__load_modules_sym(symbol_filter_t filter) +int dsos__load_modules_sym(symbol_filter_t filter) { struct utsname uts; char modules_path[PATH_MAX]; @@ -1352,33 +1352,18 @@ static int dso__load_vmlinux(struct dso *self, struct map *map, return err; } -int dso__load_kernel_sym(struct dso *self, symbol_filter_t filter, int use_modules) +int dso__load_kernel_sym(struct dso *self, symbol_filter_t filter, + int use_modules) { - int err = -1; + int err; kernel_map = map__new2(0, self); if (kernel_map == NULL) - goto out_delete_dso; + return -1; kernel_map->map_ip = kernel_map->unmap_ip = identity__map_ip; - if (use_modules && dsos__load_modules() < 0) { - pr_warning("Failed to load list of modules in use! " - "Continuing...\n"); - use_modules = 0; - } - err = dso__load_vmlinux(self, kernel_map, self->name, filter); - if (err > 0 && use_modules) { - int syms = dsos__load_modules_sym(filter); - - if (syms < 0) - pr_warning("Failed to read module symbols!" - " Continuing...\n"); - else - err += syms; - } - if (err <= 0) err = kernel_maps__load_kallsyms(filter, use_modules); @@ -1404,17 +1389,12 @@ int dso__load_kernel_sym(struct dso *self, symbol_filter_t filter, int use_modul } return err; - -out_delete_dso: - dso__delete(self); - return -1; } LIST_HEAD(dsos); struct dso *vdso; const char *vmlinux_name = "vmlinux"; -int modules; static void dsos__add(struct dso *dso) { @@ -1488,14 +1468,26 @@ struct dso *dsos__load_kernel(void) return kernel; } -int load_kernel(symbol_filter_t filter) +int load_kernel(symbol_filter_t filter, bool use_modules) { struct dso *kernel = dsos__load_kernel(); if (kernel == NULL) return -1; - return dso__load_kernel_sym(kernel, filter, modules); + if (use_modules) { + if (dsos__load_modules() < 0) + pr_warning("Failed to load list of modules in use, " + "continuing...\n"); + else if (dsos__load_modules_sym(filter) < 0) + pr_warning("Failed to read module symbols, " + "continuing...\n"); + } + + if (dso__load_kernel_sym(kernel, filter, use_modules) < 0) + pr_warning("Failed to read kernel symbols, continuing...\n"); + + return 0; } void symbol__init(unsigned int priv_size) diff --git a/tools/perf/util/symbol.h b/tools/perf/util/symbol.h index f0593a649c3..3d9d346d101 100644 --- a/tools/perf/util/symbol.h +++ b/tools/perf/util/symbol.h @@ -78,9 +78,11 @@ void dso__delete(struct dso *self); struct symbol *dso__find_symbol(struct dso *self, u64 ip); int dsos__load_modules(void); +int dsos__load_modules_sym(symbol_filter_t filter); struct dso *dsos__findnew(const char *name); int dso__load(struct dso *self, struct map *map, symbol_filter_t filter); -int dso__load_kernel_sym(struct dso *self, symbol_filter_t filter, int modules); +int dso__load_kernel_sym(struct dso *self, symbol_filter_t filter, + int use_modules); void dsos__fprintf(FILE *fp); size_t dsos__fprintf_buildid(FILE *fp); @@ -95,7 +97,7 @@ bool dsos__read_build_ids(void); int build_id__sprintf(u8 *self, int len, char *bf); struct dso *dsos__load_kernel(void); -int load_kernel(symbol_filter_t filter); +int load_kernel(symbol_filter_t filter, bool use_modules); void symbol__init(unsigned int priv_size); @@ -103,5 +105,4 @@ extern struct list_head dsos; extern struct map *kernel_map; extern struct dso *vdso; extern const char *vmlinux_name; -extern int modules; #endif /* __PERF_SYMBOL */ -- cgit v1.2.3-18-g5258 From fd7a346ea292074e9f6cdb5232a57c56bf98fdc9 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 20 Nov 2009 20:51:25 -0200 Subject: perf symbols: Filename__read_build_id should look at .notes section too MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit In the kernel we have more than one notes section, so the linker script combines all and puts them into a ".notes" combined section. So we need to look at both sections and also traverse them looking at multiple GElf_Nhdr entries till we find the one we want, with the build_id. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1258757489-5978-2-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/util/symbol.c | 47 ++++++++++++++++++++++++++++++++++++++--------- 1 file changed, 38 insertions(+), 9 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 3b23c18cd36..d22030828c2 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -899,13 +899,19 @@ bool dsos__read_build_ids(void) return have_build_id; } +/* + * Align offset to 4 bytes as needed for note name and descriptor data. + */ +#define NOTE_ALIGN(n) (((n) + 3) & -4U) + int filename__read_build_id(const char *filename, void *bf, size_t size) { int fd, err = -1; GElf_Ehdr ehdr; GElf_Shdr shdr; - Elf_Data *build_id_data; + Elf_Data *data; Elf_Scn *sec; + void *ptr; Elf *elf; if (size < BUILD_ID_SIZE) @@ -928,14 +934,37 @@ int filename__read_build_id(const char *filename, void *bf, size_t size) sec = elf_section_by_name(elf, &ehdr, &shdr, ".note.gnu.build-id", NULL); - if (sec == NULL) - goto out_elf_end; + if (sec == NULL) { + sec = elf_section_by_name(elf, &ehdr, &shdr, + ".notes", NULL); + if (sec == NULL) + goto out_elf_end; + } - build_id_data = elf_getdata(sec, NULL); - if (build_id_data == NULL) + data = elf_getdata(sec, NULL); + if (data == NULL) goto out_elf_end; - memcpy(bf, build_id_data->d_buf + 16, BUILD_ID_SIZE); - err = BUILD_ID_SIZE; + + ptr = data->d_buf; + while (ptr < (data->d_buf + data->d_size)) { + GElf_Nhdr *nhdr = ptr; + int namesz = NOTE_ALIGN(nhdr->n_namesz), + descsz = NOTE_ALIGN(nhdr->n_descsz); + const char *name; + + ptr += sizeof(*nhdr); + name = ptr; + ptr += namesz; + if (nhdr->n_type == NT_GNU_BUILD_ID && + nhdr->n_namesz == sizeof("GNU")) { + if (memcmp(name, "GNU", sizeof("GNU")) == 0) { + memcpy(bf, ptr, BUILD_ID_SIZE); + err = BUILD_ID_SIZE; + break; + } + } + ptr += descsz; + } out_elf_end: elf_end(elf); out_close: @@ -963,8 +992,8 @@ int sysfs__read_build_id(const char *filename, void *build_id, size_t size) if (read(fd, &nhdr, sizeof(nhdr)) != sizeof(nhdr)) break; - namesz = (nhdr.n_namesz + 3) & -4U; - descsz = (nhdr.n_descsz + 3) & -4U; + namesz = NOTE_ALIGN(nhdr.n_namesz); + descsz = NOTE_ALIGN(nhdr.n_descsz); if (nhdr.n_type == NT_GNU_BUILD_ID && nhdr.n_namesz == sizeof("GNU")) { if (read(fd, bf, namesz) != namesz) -- cgit v1.2.3-18-g5258 From 78075caad99dc36ec6ef5826b7a5273ea14295fc Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 20 Nov 2009 20:51:26 -0200 Subject: perf symbols: Introduce dso__build_id_equal MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Will be used in more places. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1258757489-5978-3-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/util/symbol.c | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index d22030828c2..c324bdf8bab 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -884,6 +884,11 @@ out_close: return err; } +static bool dso__build_id_equal(const struct dso *self, u8 *build_id) +{ + return memcmp(self->build_id, build_id, sizeof(self->build_id)) == 0; +} + bool dsos__read_build_ids(void) { bool have_build_id = false; @@ -1099,8 +1104,7 @@ more: sizeof(build_id)) < 0) goto more; compare_build_id: - if (memcmp(build_id, self->build_id, - sizeof(self->build_id)) != 0) + if (!dso__build_id_equal(self, build_id)) goto more; } -- cgit v1.2.3-18-g5258 From c338aee853db197e1855b393e6d6cc667784537f Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 20 Nov 2009 20:51:27 -0200 Subject: perf symbols: Do lazy symtab loading for the kernel & modules too MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Just like we do with the other DSOs. This also simplifies the kernel_maps setup process, now all that the tools need to do is to call kernel_maps__init and the maps for the modules and kernel will be created, then, later, when kernel_maps__find_symbol() is used, it will also call maps__find_symbol that already checks if the symtab was loaded, loading it if needed. Now if one does 'perf top --hide_kernel_symbols' we won't pay the price of loading the (many) symbols in /proc/kallsyms or vmlinux. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1258757489-5978-4-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/builtin-annotate.c | 8 +- tools/perf/builtin-report.c | 6 +- tools/perf/builtin-top.c | 35 ++------- tools/perf/util/data_map.c | 4 +- tools/perf/util/event.h | 3 + tools/perf/util/header.c | 6 +- tools/perf/util/map.c | 23 ++++++ tools/perf/util/symbol.c | 173 ++++++++++++++---------------------------- tools/perf/util/symbol.h | 9 +-- tools/perf/util/thread.h | 3 +- 10 files changed, 109 insertions(+), 161 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c index b6da1476ab1..203152729a6 100644 --- a/tools/perf/builtin-annotate.c +++ b/tools/perf/builtin-annotate.c @@ -157,7 +157,7 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) if (event->header.misc & PERF_RECORD_MISC_KERNEL) { level = 'k'; - sym = kernel_maps__find_symbol(ip, &map); + sym = kernel_maps__find_symbol(ip, &map, symbol_filter); dump_printf(" ...... dso: %s\n", map ? map->dso->long_name : ""); } else if (event->header.misc & PERF_RECORD_MISC_USER) { @@ -637,9 +637,9 @@ static int __cmd_annotate(void) exit(0); } - if (load_kernel(symbol_filter, use_modules) < 0) { - perror("failed to load kernel symbols"); - return EXIT_FAILURE; + if (kernel_maps__init(use_modules) < 0) { + pr_err("failed to create kernel maps for symbol resolution\b"); + return -1; } remap: diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index 0af48401f08..0d39e805be2 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -449,7 +449,7 @@ got_map: * trick of looking in the whole kernel symbol list. */ if ((long long)ip < 0) - return kernel_maps__find_symbol(ip, mapp); + return kernel_maps__find_symbol(ip, mapp, NULL); } dump_printf(" ...... dso: %s\n", map ? map->dso->long_name : ""); @@ -496,7 +496,7 @@ static struct symbol **resolve_callchain(struct thread *thread, struct map *map, case PERF_CONTEXT_HV: break; case PERF_CONTEXT_KERNEL: - sym = kernel_maps__find_symbol(ip, &map); + sym = kernel_maps__find_symbol(ip, &map, NULL); break; default: sym = resolve_symbol(thread, &map, &ip); @@ -716,7 +716,7 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) if (cpumode == PERF_RECORD_MISC_KERNEL) { level = 'k'; - sym = kernel_maps__find_symbol(ip, &map); + sym = kernel_maps__find_symbol(ip, &map, NULL); dump_printf(" ...... dso: %s\n", map ? map->dso->long_name : ""); } else if (cpumode == PERF_RECORD_MISC_USER) { diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index 48cc1084bc3..ea49c2e9dda 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -830,6 +830,8 @@ static void handle_keypress(int c) case 'q': case 'Q': printf("exiting.\n"); + if (dump_symtab) + dsos__fprintf(stderr); exit(0); case 's': prompt_symbol(&sym_filter_entry, "Enter details symbol"); @@ -946,30 +948,6 @@ static int symbol_filter(struct map *map, struct symbol *sym) return 0; } -static int parse_symbols(void) -{ - struct dso *kernel = dsos__load_kernel(); - - if (kernel == NULL) - return -1; - - if (dsos__load_modules() < 0) - pr_debug("Couldn't read the complete list of modules, " - "continuing...\n"); - - if (dsos__load_modules_sym(symbol_filter) < 0) - pr_warning("Failed to read module symbols, continuing...\n"); - - if (dso__load_kernel_sym(kernel, symbol_filter, 1) <= 0) - pr_debug("Couldn't read the complete list of kernel symbols, " - "continuing...\n"); - - if (dump_symtab) - dsos__fprintf(stderr); - - return 0; -} - static void event__process_sample(const event_t *self, int counter) { u64 ip = self->ip.ip; @@ -1012,7 +990,7 @@ static void event__process_sample(const event_t *self, int counter) if (hide_kernel_symbols) return; - sym = kernel_maps__find_symbol(ip, &map); + sym = kernel_maps__find_symbol(ip, &map, symbol_filter); if (sym == NULL) return; break; @@ -1339,7 +1317,7 @@ static const struct option options[] = { int cmd_top(int argc, const char **argv, const char *prefix __used) { - int counter; + int counter, err; page_size = sysconf(_SC_PAGE_SIZE); @@ -1363,10 +1341,11 @@ int cmd_top(int argc, const char **argv, const char *prefix __used) if (delay_secs < 1) delay_secs = 1; - parse_symbols(); + err = kernel_maps__init(true); + if (err < 0) + return err; parse_source(sym_filter_entry); - /* * User specified count overrides default frequency. */ diff --git a/tools/perf/util/data_map.c b/tools/perf/util/data_map.c index a444a2645c8..e7b6c2bea3d 100644 --- a/tools/perf/util/data_map.c +++ b/tools/perf/util/data_map.c @@ -171,8 +171,8 @@ int mmap_dispatch_perf_file(struct perf_header **pheader, goto out_delete; err = -ENOMEM; - if (load_kernel(NULL, 1) < 0) { - pr_err("failed to load kernel symbols\n"); + if (kernel_maps__init(true) < 0) { + pr_err("failed to setup the kernel maps to resolve symbols\n"); goto out_delete; } diff --git a/tools/perf/util/event.h b/tools/perf/util/event.h index 34c6fcb82d9..f1e39261265 100644 --- a/tools/perf/util/event.h +++ b/tools/perf/util/event.h @@ -115,10 +115,13 @@ typedef int (*symbol_filter_t)(struct map *map, struct symbol *sym); void map__init(struct map *self, u64 start, u64 end, u64 pgoff, struct dso *dso); struct map *map__new(struct mmap_event *event, char *cwd, int cwdlen); +void map__delete(struct map *self); struct map *map__clone(struct map *self); int map__overlap(struct map *l, struct map *r); size_t map__fprintf(struct map *self, FILE *fp); struct symbol *map__find_symbol(struct map *self, u64 ip, symbol_filter_t filter); +void map__fixup_start(struct map *self); +void map__fixup_end(struct map *self); int event__synthesize_thread(pid_t pid, int (*process)(event_t *event)); void event__synthesize_threads(int (*process)(event_t *event)); diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index d5c81ebc0a8..ac3410b8e9e 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -253,11 +253,11 @@ static int perf_header__adds_write(struct perf_header *self, int fd) buildid_sec = &feat_sec[idx++]; - dsos__load_kernel(); /* - * Read the list of loaded modules with its build_ids + * Read the kernel buildid nad the list of loaded modules with + * its build_ids: */ - dsos__load_modules(); + kernel_maps__init(true); /* Write build-ids */ buildid_sec->offset = lseek(fd, 0, SEEK_CUR); diff --git a/tools/perf/util/map.c b/tools/perf/util/map.c index 94ca95073c4..09412321a80 100644 --- a/tools/perf/util/map.c +++ b/tools/perf/util/map.c @@ -75,6 +75,29 @@ out_delete: return NULL; } +void map__delete(struct map *self) +{ + free(self); +} + +void map__fixup_start(struct map *self) +{ + struct rb_node *nd = rb_first(&self->dso->syms); + if (nd != NULL) { + struct symbol *sym = rb_entry(nd, struct symbol, rb_node); + self->start = sym->start; + } +} + +void map__fixup_end(struct map *self) +{ + struct rb_node *nd = rb_last(&self->dso->syms); + if (nd != NULL) { + struct symbol *sym = rb_entry(nd, struct symbol, rb_node); + self->end = sym->end; + } +} + #define DSO__DELETED "(deleted)" struct symbol * diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index c324bdf8bab..cb086cbe6b1 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -27,6 +27,8 @@ static void dsos__add(struct dso *dso); static struct dso *dsos__find(const char *name); static struct map *map__new2(u64 start, struct dso *dso); static void kernel_maps__insert(struct map *map); +static int dso__load_kernel_sym(struct dso *self, struct map *map, + symbol_filter_t filter); unsigned int symbol__priv_size; static struct rb_root kernel_maps; @@ -69,12 +71,6 @@ static void kernel_maps__fixup_end(void) curr = rb_entry(nd, struct map, rb_node); prev->end = curr->start - 1; } - - nd = rb_last(&curr->dso->syms); - if (nd) { - struct symbol *sym = rb_entry(nd, struct symbol, rb_node); - curr->end = sym->end; - } } static struct symbol *symbol__new(u64 start, u64 len, const char *name) @@ -324,7 +320,7 @@ out_failure: * kernel range is broken in several maps, named [kernel].N, as we don't have * the original ELF section names vmlinux have. */ -static int kernel_maps__split_kallsyms(symbol_filter_t filter, int use_modules) +static int kernel_maps__split_kallsyms(symbol_filter_t filter) { struct map *map = kernel_map; struct symbol *pos; @@ -340,9 +336,6 @@ static int kernel_maps__split_kallsyms(symbol_filter_t filter, int use_modules) module = strchr(pos->name, '\t'); if (module) { - if (!use_modules) - goto delete_symbol; - *module++ = '\0'; if (strcmp(map->dso->name, module)) { @@ -382,7 +375,6 @@ static int kernel_maps__split_kallsyms(symbol_filter_t filter, int use_modules) } if (filter && filter(map, pos)) { -delete_symbol: rb_erase(&pos->rb_node, &kernel_map->dso->syms); symbol__delete(pos); } else { @@ -398,17 +390,18 @@ delete_symbol: } -static int kernel_maps__load_kallsyms(symbol_filter_t filter, int use_modules) +static int kernel_maps__load_kallsyms(symbol_filter_t filter) { if (kernel_maps__load_all_kallsyms()) return -1; dso__fixup_sym_end(kernel_map->dso); + kernel_map->dso->origin = DSO__ORIG_KERNEL; - return kernel_maps__split_kallsyms(filter, use_modules); + return kernel_maps__split_kallsyms(filter); } -static size_t kernel_maps__fprintf(FILE *fp) +size_t kernel_maps__fprintf(FILE *fp) { size_t printed = fprintf(fp, "Kernel maps:\n"); struct rb_node *nd; @@ -1042,13 +1035,17 @@ char dso__symtab_origin(const struct dso *self) int dso__load(struct dso *self, struct map *map, symbol_filter_t filter) { int size = PATH_MAX; - char *name = malloc(size); + char *name; u8 build_id[BUILD_ID_SIZE]; int ret = -1; int fd; self->loaded = 1; + if (self->kernel) + return dso__load_kernel_sym(self, map, filter); + + name = malloc(size); if (!name) return -1; @@ -1139,7 +1136,8 @@ static void kernel_maps__insert(struct map *map) maps__insert(&kernel_maps, map); } -struct symbol *kernel_maps__find_symbol(u64 ip, struct map **mapp) +struct symbol *kernel_maps__find_symbol(u64 ip, struct map **mapp, + symbol_filter_t filter) { struct map *map = maps__find(&kernel_maps, ip); @@ -1148,7 +1146,7 @@ struct symbol *kernel_maps__find_symbol(u64 ip, struct map **mapp) if (map) { ip = map->map_ip(map, ip); - return map->dso->find_symbol(map->dso, ip); + return map__find_symbol(map, ip, filter); } return NULL; @@ -1168,28 +1166,9 @@ struct map *kernel_maps__find_by_dso_name(const char *name) return NULL; } -static int dso__load_module_sym(struct dso *self, struct map *map, - symbol_filter_t filter) -{ - int err = 0, fd = open(self->long_name, O_RDONLY); - - self->loaded = 1; - - if (fd < 0) { - pr_err("%s: cannot open %s\n", __func__, self->long_name); - return err; - } - - err = dso__load_sym(self, map, self->long_name, fd, filter, 0, 1); - close(fd); - - return err; -} - -static int dsos__load_modules_sym_dir(char *dirname, symbol_filter_t filter) +static int dsos__set_modules_path_dir(char *dirname) { struct dirent *dent; - int nr_symbols = 0, err; DIR *dir = opendir(dirname); if (!dir) { @@ -1207,14 +1186,12 @@ static int dsos__load_modules_sym_dir(char *dirname, symbol_filter_t filter) snprintf(path, sizeof(path), "%s/%s", dirname, dent->d_name); - err = dsos__load_modules_sym_dir(path, filter); - if (err < 0) + if (dsos__set_modules_path_dir(path) < 0) goto failure; } else { char *dot = strrchr(dent->d_name, '.'), dso_name[PATH_MAX]; struct map *map; - struct rb_node *last; char *long_name; if (dot == NULL || strcmp(dot, ".ko")) @@ -1234,36 +1211,16 @@ static int dsos__load_modules_sym_dir(char *dirname, symbol_filter_t filter) if (long_name == NULL) goto failure; dso__set_long_name(map->dso, long_name); - dso__set_basename(map->dso); - - err = dso__load_module_sym(map->dso, map, filter); - if (err < 0) - goto failure; - last = rb_last(&map->dso->syms); - if (last) { - struct symbol *sym; - /* - * We do this here as well, even having the - * symbol size found in the symtab because - * misannotated ASM symbols may have the size - * set to zero. - */ - dso__fixup_sym_end(map->dso); - - sym = rb_entry(last, struct symbol, rb_node); - map->end = map->start + sym->end; - } } - nr_symbols += err; } - return nr_symbols; + return 0; failure: closedir(dir); return -1; } -int dsos__load_modules_sym(symbol_filter_t filter) +static int dsos__set_modules_path(void) { struct utsname uts; char modules_path[PATH_MAX]; @@ -1274,7 +1231,7 @@ int dsos__load_modules_sym(symbol_filter_t filter) snprintf(modules_path, sizeof(modules_path), "/lib/modules/%s/kernel", uts.release); - return dsos__load_modules_sym_dir(modules_path, filter); + return dsos__set_modules_path_dir(modules_path); } /* @@ -1296,7 +1253,7 @@ static struct map *map__new2(u64 start, struct dso *dso) return self; } -int dsos__load_modules(void) +static int kernel_maps__create_module_maps(void) { char *line = NULL; size_t n; @@ -1360,7 +1317,13 @@ int dsos__load_modules(void) free(line); fclose(file); - return 0; + /* + * Now that we have all sorted out, just set the ->end of all + * maps: + */ + kernel_maps__fixup_end(); + + return dsos__set_modules_path(); out_delete_line: free(line); @@ -1385,40 +1348,17 @@ static int dso__load_vmlinux(struct dso *self, struct map *map, return err; } -int dso__load_kernel_sym(struct dso *self, symbol_filter_t filter, - int use_modules) +static int dso__load_kernel_sym(struct dso *self, struct map *map, + symbol_filter_t filter) { - int err; + int err = dso__load_vmlinux(self, map, self->name, filter); - kernel_map = map__new2(0, self); - if (kernel_map == NULL) - return -1; - - kernel_map->map_ip = kernel_map->unmap_ip = identity__map_ip; - - err = dso__load_vmlinux(self, kernel_map, self->name, filter); if (err <= 0) - err = kernel_maps__load_kallsyms(filter, use_modules); + err = kernel_maps__load_kallsyms(filter); if (err > 0) { - struct rb_node *node = rb_first(&self->syms); - struct symbol *sym = rb_entry(node, struct symbol, rb_node); - - kernel_map->start = sym->start; - node = rb_last(&self->syms); - sym = rb_entry(node, struct symbol, rb_node); - kernel_map->end = sym->end; - - self->origin = DSO__ORIG_KERNEL; - kernel_maps__insert(kernel_map); - /* - * Now that we have all sorted out, just set the ->end of all - * maps: - */ - kernel_maps__fixup_end(); - - if (verbose) - kernel_maps__fprintf(stderr); + map__fixup_start(map); + map__fixup_end(map); } return err; @@ -1479,46 +1419,51 @@ size_t dsos__fprintf_buildid(FILE *fp) return ret; } -struct dso *dsos__load_kernel(void) +static int kernel_maps__create_kernel_map(void) { struct dso *kernel = dso__new(vmlinux_name); if (kernel == NULL) - return NULL; + return -1; + + kernel_map = map__new2(0, kernel); + if (kernel_map == NULL) + goto out_delete_kernel_dso; + + kernel_map->map_ip = kernel_map->unmap_ip = identity__map_ip; kernel->short_name = "[kernel]"; + kernel->kernel = 1; vdso = dso__new("[vdso]"); - if (!vdso) - return NULL; + if (vdso == NULL) + goto out_delete_kernel_map; if (sysfs__read_build_id("/sys/kernel/notes", kernel->build_id, sizeof(kernel->build_id)) == 0) kernel->has_build_id = true; + kernel_maps__insert(kernel_map); dsos__add(kernel); dsos__add(vdso); - return kernel; + return 0; + +out_delete_kernel_map: + map__delete(kernel_map); + kernel_map = NULL; +out_delete_kernel_dso: + dso__delete(kernel); + return -1; } -int load_kernel(symbol_filter_t filter, bool use_modules) +int kernel_maps__init(bool use_modules) { - struct dso *kernel = dsos__load_kernel(); - - if (kernel == NULL) + if (kernel_maps__create_kernel_map() < 0) return -1; - if (use_modules) { - if (dsos__load_modules() < 0) - pr_warning("Failed to load list of modules in use, " - "continuing...\n"); - else if (dsos__load_modules_sym(filter) < 0) - pr_warning("Failed to read module symbols, " - "continuing...\n"); - } - - if (dso__load_kernel_sym(kernel, filter, use_modules) < 0) - pr_warning("Failed to read kernel symbols, continuing...\n"); + if (use_modules && kernel_maps__create_module_maps() < 0) + pr_warning("Failed to load list of modules in use, " + "continuing...\n"); return 0; } diff --git a/tools/perf/util/symbol.h b/tools/perf/util/symbol.h index 3d9d346d101..7a129047c47 100644 --- a/tools/perf/util/symbol.h +++ b/tools/perf/util/symbol.h @@ -64,6 +64,7 @@ struct dso { u8 slen_calculated:1; u8 loaded:1; u8 has_build_id:1; + u8 kernel:1; unsigned char origin; u8 build_id[BUILD_ID_SIZE]; u16 long_name_len; @@ -77,12 +78,8 @@ void dso__delete(struct dso *self); struct symbol *dso__find_symbol(struct dso *self, u64 ip); -int dsos__load_modules(void); -int dsos__load_modules_sym(symbol_filter_t filter); struct dso *dsos__findnew(const char *name); int dso__load(struct dso *self, struct map *map, symbol_filter_t filter); -int dso__load_kernel_sym(struct dso *self, symbol_filter_t filter, - int use_modules); void dsos__fprintf(FILE *fp); size_t dsos__fprintf_buildid(FILE *fp); @@ -96,8 +93,8 @@ int sysfs__read_build_id(const char *filename, void *bf, size_t size); bool dsos__read_build_ids(void); int build_id__sprintf(u8 *self, int len, char *bf); -struct dso *dsos__load_kernel(void); -int load_kernel(symbol_filter_t filter, bool use_modules); +int kernel_maps__init(bool use_modules); +size_t kernel_maps__fprintf(FILE *fp); void symbol__init(unsigned int priv_size); diff --git a/tools/perf/util/thread.h b/tools/perf/util/thread.h index 53addd77ce8..e4b8d437725 100644 --- a/tools/perf/util/thread.h +++ b/tools/perf/util/thread.h @@ -26,7 +26,8 @@ size_t threads__fprintf(FILE *fp); void maps__insert(struct rb_root *maps, struct map *map); struct map *maps__find(struct rb_root *maps, u64 ip); -struct symbol *kernel_maps__find_symbol(const u64 ip, struct map **mapp); +struct symbol *kernel_maps__find_symbol(const u64 ip, struct map **mapp, + symbol_filter_t filter); struct map *kernel_maps__find_by_dso_name(const char *name); static inline struct map *thread__find_map(struct thread *self, u64 ip) -- cgit v1.2.3-18-g5258 From fbd733b815a5a57d7eb0d904edc49d18fd12df5c Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 20 Nov 2009 20:51:28 -0200 Subject: perf symbols: Check vmlinux buildid MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit E.g.: [root@doppio linux-2.6-tip]# perf top -v --vmlinux ../build/tip/vmlinux > /dev/null build_id in vmlinux is e96699725a47413a50c231864a8e7a8ced40a31b while expected is 18e7cc53db62a7d35e9d6f6c9ddc23017d38ee9a, ignoring it I.e. perf top was told to use a vmlinux file that is not the one currently running on the machine, it ignores it and falls back to using /proc/kallsyms. This solves many, at first, mysterious results when people have a stale vmlinux file while keeping the default of trying to use the vmlinux file in the current directory in things like 'perf annotate' where the DWARF info is required and thus we can't use just /proc/kallsyms. Modules buildids are already being checked as of the previous changeset in this series, because we are using the default dso__load routine, that will look at a series of places looking for the best file with a matching buildid, starting in the -debuginfo directories. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1258757489-5978-5-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/util/symbol.c | 28 ++++++++++++++++++++++++++-- 1 file changed, 26 insertions(+), 2 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index cb086cbe6b1..9cf6dbcd158 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -1334,13 +1334,37 @@ out_failure: static int dso__load_vmlinux(struct dso *self, struct map *map, const char *vmlinux, symbol_filter_t filter) { - int err, fd = open(vmlinux, O_RDONLY); + int err = -1, fd; - self->loaded = 1; + if (self->has_build_id) { + u8 build_id[BUILD_ID_SIZE]; + + if (filename__read_build_id(vmlinux, build_id, + sizeof(build_id)) < 0) { + pr_debug("No build_id in %s, ignoring it\n", vmlinux); + return -1; + } + if (!dso__build_id_equal(self, build_id)) { + char expected_build_id[BUILD_ID_SIZE * 2 + 1], + vmlinux_build_id[BUILD_ID_SIZE * 2 + 1]; + + build_id__sprintf(self->build_id, + sizeof(self->build_id), + expected_build_id); + build_id__sprintf(build_id, sizeof(build_id), + vmlinux_build_id); + pr_debug("build_id in %s is %s while expected is %s, " + "ignoring it\n", vmlinux, vmlinux_build_id, + expected_build_id); + return -1; + } + } + fd = open(vmlinux, O_RDONLY); if (fd < 0) return -1; + self->loaded = 1; err = dso__load_sym(self, map, self->long_name, fd, filter, 1, 0); close(fd); -- cgit v1.2.3-18-g5258 From ef6ae724253429ac70d81e65d052f6a346d330bd Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 20 Nov 2009 20:51:29 -0200 Subject: perf symbols: Change the kernel DSO name if it comes from kallsyms MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit So that the user have a clearer indication about the source of the symbols, as we only state buildid mismatches in verbose mode, because 'perf top' would overwrite such warning anyway. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1258757489-5978-6-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/util/symbol.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) (limited to 'tools/perf') diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 9cf6dbcd158..48f87f065a2 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -108,6 +108,8 @@ static size_t symbol__fprintf(struct symbol *self, FILE *fp) static void dso__set_long_name(struct dso *self, char *name) { + if (name == NULL) + return; self->long_name = name; self->long_name_len = strlen(name); } @@ -1377,8 +1379,11 @@ static int dso__load_kernel_sym(struct dso *self, struct map *map, { int err = dso__load_vmlinux(self, map, self->name, filter); - if (err <= 0) + if (err <= 0) { err = kernel_maps__load_kallsyms(filter); + if (err > 0) + dso__set_long_name(self, strdup("[kernel.kallsyms]")); + } if (err > 0) { map__fixup_start(map); -- cgit v1.2.3-18-g5258 From 90c83218c32d7c474da810cd3c9973a43ecbcb9b Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Sat, 21 Nov 2009 14:31:24 -0200 Subject: perf symbols: Fixup kernel_maps__fixup_end end map MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We better call this routine after both the kernel and modules are loaded, because as it was if there weren't modules it would not be called, resulting in kernel_map->end remaining at zero, so no map would be found and consequently the kernel symtab wouldn't get loaded, i.e. no kernel symbols would be resolved. Also this fixes another case, that is when we _have_ modules, but the last map would have its ->end address not set before we loaded its symbols, which would never happen because ->end was not set. Reported-by: Ingo Molnar Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1258821086-11521-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/util/symbol.c | 17 ++++++++++------- 1 file changed, 10 insertions(+), 7 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 48f87f065a2..e161a51c9fe 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -71,6 +71,12 @@ static void kernel_maps__fixup_end(void) curr = rb_entry(nd, struct map, rb_node); prev->end = curr->start - 1; } + + /* + * We still haven't the actual symbols, so guess the + * last map final address. + */ + curr->end = ~0UL; } static struct symbol *symbol__new(u64 start, u64 len, const char *name) @@ -1319,12 +1325,6 @@ static int kernel_maps__create_module_maps(void) free(line); fclose(file); - /* - * Now that we have all sorted out, just set the ->end of all - * maps: - */ - kernel_maps__fixup_end(); - return dsos__set_modules_path(); out_delete_line: @@ -1493,7 +1493,10 @@ int kernel_maps__init(bool use_modules) if (use_modules && kernel_maps__create_module_maps() < 0) pr_warning("Failed to load list of modules in use, " "continuing...\n"); - + /* + * Now that we have all the maps created, just set the ->end of them: + */ + kernel_maps__fixup_end(); return 0; } -- cgit v1.2.3-18-g5258 From c12e15e71d4b32da045e798ffd21cbb6197d1c65 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Sat, 21 Nov 2009 14:31:25 -0200 Subject: perf symbols: Old versions of elf.h don't have NT_GNU_BUILD_ID MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1258821086-11521-2-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/util/symbol.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'tools/perf') diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index e161a51c9fe..86ec6c720f0 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -12,6 +12,10 @@ #include #include +#ifndef NT_GNU_BUILD_ID +#define NT_GNU_BUILD_ID 3 +#endif + enum dso_origin { DSO__ORIG_KERNEL = 0, DSO__ORIG_JAVA_JIT, -- cgit v1.2.3-18-g5258 From e25613683bd5c46d3e8c8ae6416dccc9f357dcdc Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Sat, 21 Nov 2009 14:31:26 -0200 Subject: perf trace: Read_tracing_data should die() another day MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit It better propagate errors, also if we do a simple: [root@doppio linux-2.6-tip]# perf record -R -a -f sleep 3s ; perf trace [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.182 MB perf.data (~7972 samples) ] Fatal: not an trace data file [root@doppio linux-2.6-tip]# That is what is expected, right? I.e. as we didn't specify any tracepoint event via -e, it should gracefully bail out and not SEGFAULT. Signed-off-by: Arnaldo Carvalho de Melo Cc: Steven Rostedt Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1258821086-11521-3-git-send-email-acme@infradead.org> [ Fixed the error messages some more ] Signed-off-by: Ingo Molnar --- tools/perf/util/trace-event-info.c | 22 +++++++++++++++------- tools/perf/util/trace-event-read.c | 4 ++-- tools/perf/util/trace-event.h | 2 +- 3 files changed, 18 insertions(+), 10 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/trace-event-info.c b/tools/perf/util/trace-event-info.c index 831052d4b4f..cace3559553 100644 --- a/tools/perf/util/trace-event-info.c +++ b/tools/perf/util/trace-event-info.c @@ -33,11 +33,11 @@ #include #include #include +#include #include "../perf.h" #include "trace-event.h" - #define VERSION "0.5" #define _STR(x) #x @@ -483,23 +483,31 @@ static struct tracepoint_path * get_tracepoints_path(struct perf_event_attr *pattrs, int nb_events) { struct tracepoint_path path, *ppath = &path; - int i; + int i, nr_tracepoints = 0; for (i = 0; i < nb_events; i++) { if (pattrs[i].type != PERF_TYPE_TRACEPOINT) continue; + ++nr_tracepoints; ppath->next = tracepoint_id_to_path(pattrs[i].config); if (!ppath->next) die("%s\n", "No memory to alloc tracepoints list"); ppath = ppath->next; } - return path.next; + return nr_tracepoints > 0 ? path.next : NULL; } -void read_tracing_data(int fd, struct perf_event_attr *pattrs, int nb_events) + +int read_tracing_data(int fd, struct perf_event_attr *pattrs, int nb_events) { char buf[BUFSIZ]; - struct tracepoint_path *tps; + struct tracepoint_path *tps = get_tracepoints_path(pattrs, nb_events); + + /* + * What? No tracepoints? No sense writing anything here, bail out. + */ + if (tps == NULL) + return -1; output_fd = fd; @@ -528,11 +536,11 @@ void read_tracing_data(int fd, struct perf_event_attr *pattrs, int nb_events) page_size = getpagesize(); write_or_die(&page_size, 4); - tps = get_tracepoints_path(pattrs, nb_events); - read_header_files(); read_ftrace_files(tps); read_event_files(tps); read_proc_kallsyms(); read_ftrace_printk(); + + return 0; } diff --git a/tools/perf/util/trace-event-read.c b/tools/perf/util/trace-event-read.c index 44292e06cca..342dfdd43f8 100644 --- a/tools/perf/util/trace-event-read.c +++ b/tools/perf/util/trace-event-read.c @@ -471,11 +471,11 @@ void trace_report(int fd) read_or_die(buf, 3); if (memcmp(buf, test, 3) != 0) - die("not an trace data file"); + die("no trace data in the file"); read_or_die(buf, 7); if (memcmp(buf, "tracing", 7) != 0) - die("not a trace file (missing tracing)"); + die("not a trace file (missing 'tracing' tag)"); version = read_string(); if (show_version) diff --git a/tools/perf/util/trace-event.h b/tools/perf/util/trace-event.h index f6637c2fa1f..dd51c6872a1 100644 --- a/tools/perf/util/trace-event.h +++ b/tools/perf/util/trace-event.h @@ -248,7 +248,7 @@ unsigned long long raw_field_value(struct event *event, const char *name, void *data); void *raw_field_ptr(struct event *event, const char *name, void *data); -void read_tracing_data(int fd, struct perf_event_attr *pattrs, int nb_events); +int read_tracing_data(int fd, struct perf_event_attr *pattrs, int nb_events); /* taken from kernel/trace/trace.h */ enum trace_flag_type { -- cgit v1.2.3-18-g5258 From 12eac0bf0461910ae6dd7f071f156f75461a37cf Mon Sep 17 00:00:00 2001 From: Hitoshi Mitake Date: Fri, 20 Nov 2009 12:37:17 +0900 Subject: perf bench: Make the mem/memcpy tests more user-friendly mem-memcpy.c uses perf event system calls to obtain CPU clocks. And it suddenly dies with BUG_ON() when it running on Linux doesn't support perf event. Also fail at calloc() can occur easily when too large length is passed. Fail of calloc() causes sudden death with assert(). These behaviours are not friendly. So I fixed the treating of errors. Signed-off-by: Hitoshi Mitake Cc: Peter Zijlstra Cc: Paul Mackerras Cc: Frederic Weisbecker LKML-Reference: <1258688237-3797-1-git-send-email-mitake@dcl.info.waseda.ac.jp> [ v2: improved a few small details ] Signed-off-by: Ingo Molnar --- tools/perf/bench/mem-memcpy.c | 37 ++++++++++++++++++++++--------------- 1 file changed, 22 insertions(+), 15 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/bench/mem-memcpy.c b/tools/perf/bench/mem-memcpy.c index d4f4f9806ae..5165fd1d8d2 100644 --- a/tools/perf/bench/mem-memcpy.c +++ b/tools/perf/bench/mem-memcpy.c @@ -22,9 +22,10 @@ #define K 1024 -static const char *length_str = "1MB"; -static const char *routine = "default"; -static int use_clock = 0; +static const char *length_str = "1MB"; +static const char *routine = "default"; +static int use_clock = 0; +static int clock_fd; static const struct option options[] = { OPT_STRING('l', "length", &length_str, "1MB", @@ -57,17 +58,19 @@ static const char * const bench_mem_memcpy_usage[] = { NULL }; -static int clock_fd; - static struct perf_event_attr clock_attr = { - .type = PERF_TYPE_HARDWARE, - .config = PERF_COUNT_HW_CPU_CYCLES + .type = PERF_TYPE_HARDWARE, + .config = PERF_COUNT_HW_CPU_CYCLES }; static void init_clock(void) { clock_fd = sys_perf_event_open(&clock_attr, getpid(), -1, -1, 0); - BUG_ON(clock_fd < 0); + + if (clock_fd < 0 && errno == ENOSYS) + die("No CONFIG_PERF_EVENTS=y kernel support configured?\n"); + else + BUG_ON(clock_fd < 0); } static u64 get_clock(void) @@ -104,7 +107,8 @@ int bench_mem_memcpy(int argc, const char **argv, tv_diff.tv_sec = 0; tv_diff.tv_usec = 0; length = (size_t)perf_atoll((char *)length_str); - if ((long long int)length <= 0) { + + if ((s64)length <= 0) { fprintf(stderr, "Invalid length:%s\n", length_str); return 1; } @@ -124,9 +128,12 @@ int bench_mem_memcpy(int argc, const char **argv, } dst = calloc(length, sizeof(char)); - assert(dst); + if (!dst) + die("memory allocation failed - maybe length is too large?\n"); + src = calloc(length, sizeof(char)); - assert(src); + if (!src) + die("memory allocation failed - maybe length is too large?\n"); if (bench_format == BENCH_FORMAT_DEFAULT) { printf("# Copying %s Bytes from %p to %p ...\n\n", @@ -136,8 +143,9 @@ int bench_mem_memcpy(int argc, const char **argv, if (use_clock) { init_clock(); clock_start = get_clock(); - } else + } else { BUG_ON(gettimeofday(&tv_start, NULL)); + } routines[i].fn(dst, src, length); @@ -176,9 +184,8 @@ int bench_mem_memcpy(int argc, const char **argv, printf("%lf\n", bps); break; default: - /* reaching here is something disaster */ - fprintf(stderr, "Unknown format:%d\n", bench_format); - exit(1); + /* reaching this means there's some disaster: */ + die("unknown format: %d\n", bench_format); break; } -- cgit v1.2.3-18-g5258 From f3ced7cdb24e7968a353d828955fa2daf4167e72 Mon Sep 17 00:00:00 2001 From: Pekka Enberg Date: Sun, 22 Nov 2009 11:58:00 +0200 Subject: perf kmem: Add --sort hit and --sort frag This patch adds support for "--sort hit" and "--sort frag" to the "perf kmem" tool. The former was already mentioned in the help text and the latter is useful for finding call-sites that exhibit worst case behavior for SLAB allocators. Signed-off-by: Pekka Enberg Cc: Li Zefan Cc: Peter Zijlstra Cc: Frederic Weisbecker Cc: Steven Rostedt Cc: Eduard - Gabriel Munteanu Cc: linux-mm@kvack.org LKML-Reference: <1258883880-7149-1-git-send-email-penberg@cs.helsinki.fi> Signed-off-by: Ingo Molnar --- tools/perf/builtin-kmem.c | 29 ++++++++++++++++++++++++++++- 1 file changed, 28 insertions(+), 1 deletion(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-kmem.c b/tools/perf/builtin-kmem.c index f315b052f81..4145049e7bf 100644 --- a/tools/perf/builtin-kmem.c +++ b/tools/perf/builtin-kmem.c @@ -443,6 +443,15 @@ static int callsite_cmp(struct alloc_stat *l, struct alloc_stat *r) return 0; } +static int hit_cmp(struct alloc_stat *l, struct alloc_stat *r) +{ + if (l->hit < r->hit) + return -1; + else if (l->hit > r->hit) + return 1; + return 0; +} + static int bytes_cmp(struct alloc_stat *l, struct alloc_stat *r) { if (l->bytes_alloc < r->bytes_alloc) @@ -452,6 +461,20 @@ static int bytes_cmp(struct alloc_stat *l, struct alloc_stat *r) return 0; } +static int frag_cmp(struct alloc_stat *l, struct alloc_stat *r) +{ + double x, y; + + x = fragmentation(l->bytes_req, l->bytes_alloc); + y = fragmentation(r->bytes_req, r->bytes_alloc); + + if (x < y) + return -1; + else if (x > y) + return 1; + return 0; +} + static int parse_sort_opt(const struct option *opt __used, const char *arg, int unset __used) { @@ -464,8 +487,12 @@ static int parse_sort_opt(const struct option *opt __used, sort_fn = ptr_cmp; else if (strcmp(arg, "call_site") == 0) sort_fn = callsite_cmp; + else if (strcmp(arg, "hit") == 0) + sort_fn = hit_cmp; else if (strcmp(arg, "bytes") == 0) sort_fn = bytes_cmp; + else if (strcmp(arg, "frag") == 0) + sort_fn = frag_cmp; else return -1; @@ -517,7 +544,7 @@ static const struct option kmem_options[] = { "stat selector, Pass 'alloc' or 'caller'.", parse_stat_opt), OPT_CALLBACK('s', "sort", NULL, "key", - "sort by key: ptr, call_site, hit, bytes", + "sort by key: ptr, call_site, hit, bytes, frag", parse_sort_opt), OPT_CALLBACK('l', "line", NULL, "num", "show n lins", -- cgit v1.2.3-18-g5258 From e57cfcdac6badd846a1cd831de54a1359c2d1eea Mon Sep 17 00:00:00 2001 From: Pekka Enberg Date: Sun, 22 Nov 2009 12:29:44 +0200 Subject: perf symbols: Fix ELF header errors during "perf kmem record" The write_event() function in builtin-record.c writes out all mmap()'d DSOs including non-ELF files like GNOME resource files and such. Therefore, check for ELF_K_ELF in filename__read_build_id() before attempting to read the ELF header with gelf_getehdr(). Fixes the following error messages when running "perf kmem record": penberg@penberg-laptop:~/src/linux/tools/perf$ perf kmem record ^C[ perf record: Woken up 2 times to write data ] [ perf record: Captured and wrote 0.753 MB perf.data (~32885 samples) ] filename__read_build_id: cannot get elf header. filename__read_build_id: cannot get elf header. filename__read_build_id: cannot get elf header. filename__read_build_id: cannot get elf header. filename__read_build_id: cannot get elf header. filename__read_build_id: cannot get elf header. filename__read_build_id: cannot get elf header. filename__read_build_id: cannot get elf header. filename__read_build_id: cannot get elf header. Signed-off-by: Pekka Enberg Cc: Arnaldo Carvalho de Melo Cc: Li Zefan Cc: Peter Zijlstra Cc: Frederic Weisbecker Cc: Steven Rostedt LKML-Reference: <1258885784-11709-1-git-send-email-penberg@cs.helsinki.fi> Signed-off-by: Ingo Molnar --- tools/perf/util/symbol.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'tools/perf') diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 86ec6c720f0..f56158fb4dc 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -921,6 +921,7 @@ int filename__read_build_id(const char *filename, void *bf, size_t size) GElf_Shdr shdr; Elf_Data *data; Elf_Scn *sec; + Elf_Kind ek; void *ptr; Elf *elf; @@ -937,6 +938,10 @@ int filename__read_build_id(const char *filename, void *bf, size_t size) goto out_close; } + ek = elf_kind(elf); + if (ek != ELF_K_ELF) + goto out_elf_end; + if (gelf_getehdr(elf, &ehdr) == NULL) { pr_err("%s: cannot get elf header.\n", __func__); goto out_elf_end; -- cgit v1.2.3-18-g5258 From 7baed9af4bf0d7850045e36d19a43a2c76872b62 Mon Sep 17 00:00:00 2001 From: "Michael S. Tsirkin" Date: Sun, 22 Nov 2009 13:27:27 +0200 Subject: perf tools: Add V=2 option to help debug config issues Make standard error show up on console when V=2 is set. Signed-off-by: Michael S. Tsirkin Cc: Peter Zijlstra Cc: Paul Mackerras Cc: Frederic Weisbecker Cc: Arnaldo Carvalho de Melo LKML-Reference: <20091122112726.GC13644@redhat.com> Signed-off-by: Ingo Molnar --- tools/perf/Makefile | 24 +++++++++++++++--------- 1 file changed, 15 insertions(+), 9 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/Makefile b/tools/perf/Makefile index d7198c54bb6..31da6bed46c 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -2,6 +2,7 @@ all:: # Define V=1 to have a more verbose compile. +# Define V=2 to have an even more verbose compile. # # Define SNPRINTF_RETURNS_BOGUS if your are on a system which snprintf() # or vsnprintf() return -1 instead of number of characters which would @@ -263,7 +264,7 @@ PTHREAD_LIBS = -lpthread # explicitly what architecture to check for. Fix this up for yours.. SPARSE_FLAGS = -D__BIG_ENDIAN__ -D__powerpc__ -ifeq ($(shell sh -c "echo 'int foo(void) {char X[2]; return 3;}' | $(CC) -x c -c -Werror -fstack-protector-all - -o /dev/null >/dev/null 2>&1 && echo y"), y) +ifeq ($(shell sh -c "echo 'int foo(void) {char X[2]; return 3;}' | $(CC) -x c -c -Werror -fstack-protector-all - -o /dev/null "$(QUIET_STDERR)" && echo y"), y) CFLAGS := $(CFLAGS) -fstack-protector-all endif @@ -448,6 +449,11 @@ BUILTIN_OBJS += builtin-kmem.o PERFLIBS = $(LIB_FILE) +ifeq ($(V), 2) + QUIET_STDERR = ">/dev/null" +else + QUIET_STDERR = ">/dev/null 2>&1" +endif # # Platform specific tweaks # @@ -475,19 +481,19 @@ ifeq ($(uname_S),Darwin) PTHREAD_LIBS = endif -ifeq ($(shell sh -c "(echo '\#include '; echo 'int main(void) { Elf * elf = elf_begin(0, ELF_C_READ, 0); return (long)elf; }') | $(CC) -x c - $(ALL_CFLAGS) -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -o /dev/null $(ALL_LDFLAGS) $(EXTLIBS) > /dev/null 2>&1 && echo y"), y) -ifneq ($(shell sh -c "(echo '\#include '; echo 'int main(void) { const char * version = gnu_get_libc_version(); return (long)version; }') | $(CC) -x c - $(ALL_CFLAGS) -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -o /dev/null $(ALL_LDFLAGS) $(EXTLIBS) > /dev/null 2>&1 && echo y"), y) +ifeq ($(shell sh -c "(echo '\#include '; echo 'int main(void) { Elf * elf = elf_begin(0, ELF_C_READ, 0); return (long)elf; }') | $(CC) -x c - $(ALL_CFLAGS) -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -o /dev/null $(ALL_LDFLAGS) $(EXTLIBS) "$(QUIET_STDERR)" && echo y"), y) +ifneq ($(shell sh -c "(echo '\#include '; echo 'int main(void) { const char * version = gnu_get_libc_version(); return (long)version; }') | $(CC) -x c - $(ALL_CFLAGS) -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -o /dev/null $(ALL_LDFLAGS) $(EXTLIBS) "$(QUIET_STDERR)" && echo y"), y) msg := $(error No gnu/libc-version.h found, please install glibc-dev[el]); endif - ifneq ($(shell sh -c "(echo '\#include '; echo 'int main(void) { Elf * elf = elf_begin(0, ELF_C_READ_MMAP, 0); return (long)elf; }') | $(CC) -x c - $(ALL_CFLAGS) -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -o /dev/null $(ALL_LDFLAGS) $(EXTLIBS) > /dev/null 2>&1 && echo y"), y) + ifneq ($(shell sh -c "(echo '\#include '; echo 'int main(void) { Elf * elf = elf_begin(0, ELF_C_READ_MMAP, 0); return (long)elf; }') | $(CC) -x c - $(ALL_CFLAGS) -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -o /dev/null $(ALL_LDFLAGS) $(EXTLIBS) "$(QUIET_STDERR)" && echo y"), y) BASIC_CFLAGS += -DLIBELF_NO_MMAP endif else msg := $(error No libelf.h/libelf found, please install libelf-dev/elfutils-libelf-devel and glibc-dev[el]); endif -ifneq ($(shell sh -c "(echo '\#include '; echo '\#include '; echo 'int main(void) { Dwarf_Debug dbg; Dwarf_Error err; Dwarf_Ranges *rng; dwarf_init(0, DW_DLC_READ, 0, 0, &dbg, &err); dwarf_get_ranges(dbg, 0, &rng, 0, 0, &err); return (long)dbg; }') | $(CC) -x c - $(ALL_CFLAGS) -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -ldwarf -lelf -o /dev/null $(ALL_LDFLAGS) $(EXTLIBS) > /dev/null 2>&1 && echo y"), y) +ifneq ($(shell sh -c "(echo '\#include '; echo '\#include '; echo 'int main(void) { Dwarf_Debug dbg; Dwarf_Error err; Dwarf_Ranges *rng; dwarf_init(0, DW_DLC_READ, 0, 0, &dbg, &err); dwarf_get_ranges(dbg, 0, &rng, 0, 0, &err); return (long)dbg; }') | $(CC) -x c - $(ALL_CFLAGS) -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -ldwarf -lelf -o /dev/null $(ALL_LDFLAGS) $(EXTLIBS) "$(QUIET_STDERR)" && echo y"), y) msg := $(warning No libdwarf.h found or old libdwarf.h found, disables dwarf support. Please install libdwarf-dev/libdwarf-devel >= 20081231); BASIC_CFLAGS += -DNO_LIBDWARF else @@ -499,20 +505,20 @@ endif ifdef NO_DEMANGLE BASIC_CFLAGS += -DNO_DEMANGLE else - has_bfd := $(shell sh -c "(echo '\#include '; echo 'int main(void) { bfd_demangle(0, 0, 0); return 0; }') | $(CC) -x c - $(ALL_CFLAGS) -o /dev/null $(ALL_LDFLAGS) $(EXTLIBS) -lbfd > /dev/null 2>&1 && echo y") + has_bfd := $(shell sh -c "(echo '\#include '; echo 'int main(void) { bfd_demangle(0, 0, 0); return 0; }') | $(CC) -x c - $(ALL_CFLAGS) -o /dev/null $(ALL_LDFLAGS) $(EXTLIBS) -lbfd "$(QUIET_STDERR)" && echo y") ifeq ($(has_bfd),y) EXTLIBS += -lbfd else - has_bfd_iberty := $(shell sh -c "(echo '\#include '; echo 'int main(void) { bfd_demangle(0, 0, 0); return 0; }') | $(CC) -x c - $(ALL_CFLAGS) -o /dev/null $(ALL_LDFLAGS) $(EXTLIBS) -lbfd -liberty > /dev/null 2>&1 && echo y") + has_bfd_iberty := $(shell sh -c "(echo '\#include '; echo 'int main(void) { bfd_demangle(0, 0, 0); return 0; }') | $(CC) -x c - $(ALL_CFLAGS) -o /dev/null $(ALL_LDFLAGS) $(EXTLIBS) -lbfd -liberty "$(QUIET_STDERR)" && echo y") ifeq ($(has_bfd_iberty),y) EXTLIBS += -lbfd -liberty else - has_bfd_iberty_z := $(shell sh -c "(echo '\#include '; echo 'int main(void) { bfd_demangle(0, 0, 0); return 0; }') | $(CC) -x c - $(ALL_CFLAGS) -o /dev/null $(ALL_LDFLAGS) $(EXTLIBS) -lbfd -liberty -lz > /dev/null 2>&1 && echo y") + has_bfd_iberty_z := $(shell sh -c "(echo '\#include '; echo 'int main(void) { bfd_demangle(0, 0, 0); return 0; }') | $(CC) -x c - $(ALL_CFLAGS) -o /dev/null $(ALL_LDFLAGS) $(EXTLIBS) -lbfd -liberty -lz "$(QUIET_STDERR)" && echo y") ifeq ($(has_bfd_iberty_z),y) EXTLIBS += -lbfd -liberty -lz else - has_cplus_demangle := $(shell sh -c "(echo 'extern char *cplus_demangle(const char *, int);'; echo 'int main(void) { cplus_demangle(0, 0); return 0; }') | $(CC) -x c - $(ALL_CFLAGS) -o /dev/null $(ALL_LDFLAGS) $(EXTLIBS) -liberty > /dev/null 2>&1 && echo y") + has_cplus_demangle := $(shell sh -c "(echo 'extern char *cplus_demangle(const char *, int);'; echo 'int main(void) { cplus_demangle(0, 0); return 0; }') | $(CC) -x c - $(ALL_CFLAGS) -o /dev/null $(ALL_LDFLAGS) $(EXTLIBS) -liberty "$(QUIET_STDERR)" && echo y") ifeq ($(has_cplus_demangle),y) EXTLIBS += -liberty BASIC_CFLAGS += -DHAVE_CPLUS_DEMANGLE -- cgit v1.2.3-18-g5258 From b197c7ef7169bd5f11fb9d803b322d0daef7e256 Mon Sep 17 00:00:00 2001 From: "Michael S. Tsirkin" Date: Sun, 22 Nov 2009 15:13:11 +0200 Subject: perf tools: Suggest static libraries as well On error, suggest installing static libraries along with shared libraries. Signed-off-by: Michael S. Tsirkin Cc: Peter Zijlstra Cc: Paul Mackerras Cc: Frederic Weisbecker Cc: Arnaldo Carvalho de Melo LKML-Reference: <20091122131311.GA24318@redhat.com> Signed-off-by: Ingo Molnar --- tools/perf/Makefile | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/Makefile b/tools/perf/Makefile index 31da6bed46c..fce4c3f051e 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -483,7 +483,7 @@ endif ifeq ($(shell sh -c "(echo '\#include '; echo 'int main(void) { Elf * elf = elf_begin(0, ELF_C_READ, 0); return (long)elf; }') | $(CC) -x c - $(ALL_CFLAGS) -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -o /dev/null $(ALL_LDFLAGS) $(EXTLIBS) "$(QUIET_STDERR)" && echo y"), y) ifneq ($(shell sh -c "(echo '\#include '; echo 'int main(void) { const char * version = gnu_get_libc_version(); return (long)version; }') | $(CC) -x c - $(ALL_CFLAGS) -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -o /dev/null $(ALL_LDFLAGS) $(EXTLIBS) "$(QUIET_STDERR)" && echo y"), y) - msg := $(error No gnu/libc-version.h found, please install glibc-dev[el]); + msg := $(error No gnu/libc-version.h found, please install glibc-dev[el]/glibc-static); endif ifneq ($(shell sh -c "(echo '\#include '; echo 'int main(void) { Elf * elf = elf_begin(0, ELF_C_READ_MMAP, 0); return (long)elf; }') | $(CC) -x c - $(ALL_CFLAGS) -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -o /dev/null $(ALL_LDFLAGS) $(EXTLIBS) "$(QUIET_STDERR)" && echo y"), y) @@ -523,7 +523,7 @@ else EXTLIBS += -liberty BASIC_CFLAGS += -DHAVE_CPLUS_DEMANGLE else - msg := $(warning No bfd.h/libbfd found, install binutils-dev[el] to gain symbol demangling) + msg := $(warning No bfd.h/libbfd found, install binutils-dev[el]/zlib-static to gain symbol demangling) BASIC_CFLAGS += -DNO_DEMANGLE endif endif -- cgit v1.2.3-18-g5258 From 87f8ea4cd3680ef7f4da4391aed97abb25eae333 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Sun, 22 Nov 2009 13:21:41 -0200 Subject: perf symbols: Show messages about module loading only if verbose >= 1 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Suggested-by: Ingo Molnar Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1258903301-20584-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/util/symbol.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index f56158fb4dc..74b5b8a1695 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -1189,7 +1189,7 @@ static int dsos__set_modules_path_dir(char *dirname) DIR *dir = opendir(dirname); if (!dir) { - pr_err("%s: cannot open %s dir\n", __func__, dirname); + pr_debug("%s: cannot open %s dir\n", __func__, dirname); return -1; } @@ -1500,8 +1500,8 @@ int kernel_maps__init(bool use_modules) return -1; if (use_modules && kernel_maps__create_module_maps() < 0) - pr_warning("Failed to load list of modules in use, " - "continuing...\n"); + pr_debug("Failed to load list of modules in use, " + "continuing...\n"); /* * Now that we have all the maps created, just set the ->end of them: */ -- cgit v1.2.3-18-g5258 From 50e5095afa8c2be0f35e5c0e21d5f7912340e8f2 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Sun, 22 Nov 2009 14:59:22 -0200 Subject: perf report: Do map lookups in resolve_callchain() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Bug introduced in 439d473b4777de510e1322168ac6f2f377ecd5bc, making the initial map be used for all IPs, so that symbols outside this initial map would either be erroneously resolved or not resolve at all. Reported-by: Ingo Molnar Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1258909162-28496-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/builtin-report.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index 0d39e805be2..7e690f73b51 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -467,7 +467,7 @@ static int call__match(struct symbol *sym) return 0; } -static struct symbol **resolve_callchain(struct thread *thread, struct map *map, +static struct symbol **resolve_callchain(struct thread *thread, struct ip_callchain *chain, struct symbol **parent) { @@ -496,10 +496,10 @@ static struct symbol **resolve_callchain(struct thread *thread, struct map *map, case PERF_CONTEXT_HV: break; case PERF_CONTEXT_KERNEL: - sym = kernel_maps__find_symbol(ip, &map, NULL); + sym = kernel_maps__find_symbol(ip, NULL, NULL); break; default: - sym = resolve_symbol(thread, &map, &ip); + sym = resolve_symbol(thread, NULL, &ip); break; } @@ -529,7 +529,7 @@ hist_entry__add(struct thread *thread, struct map *map, struct hist_entry *he; if ((sort__has_parent || callchain) && chain) - syms = resolve_callchain(thread, map, chain, &parent); + syms = resolve_callchain(thread, chain, &parent); he = __hist_entry__add(thread, map, sym, parent, ip, count, level, &hit); -- cgit v1.2.3-18-g5258 From 81516c5fc83a13a1d12f466aa7e14f5fd62a63ce Mon Sep 17 00:00:00 2001 From: "Michael S. Tsirkin" Date: Sun, 22 Nov 2009 14:13:35 +0200 Subject: perf: Use default compiler mode by default gcc with no flags typically is a sane default for systems to use, and looking at the running kernel is probably broken for cross-builds anyway, so let's not do this. Add EXTRA_CFLAGS so that users can override default gcc mode if they want to. Signed-off-by: Michael S. Tsirkin Acked-by: Arjan van de Ven Cc: Peter Zijlstra Cc: Paul Mackerras Cc: Frederic Weisbecker Cc: Arnaldo Carvalho de Melo LKML-Reference: <20091122121335.GA24254@redhat.com> Signed-off-by: Ingo Molnar --- tools/perf/Makefile | 20 +++----------------- 1 file changed, 3 insertions(+), 17 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/Makefile b/tools/perf/Makefile index fce4c3f051e..3ef6621bf6c 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -148,6 +148,8 @@ all:: # broken, or spawning external process is slower than built-in grep perf has). # # Define LDFLAGS=-static to build a static binary. +# +# Define EXTRA_CFLAGS=-m64 or EXTRA_CFLAGS=-m32 as appropriate for cross-builds. PERF-VERSION-FILE: .FORCE-PERF-VERSION-FILE @$(SHELL_PATH) util/PERF-VERSION-GEN @@ -160,22 +162,6 @@ uname_R := $(shell sh -c 'uname -r 2>/dev/null || echo not') uname_P := $(shell sh -c 'uname -p 2>/dev/null || echo not') uname_V := $(shell sh -c 'uname -v 2>/dev/null || echo not') -# -# Add -m32 for cross-builds: -# -ifdef NO_64BIT - MBITS := -m32 -else - # - # If we're on a 64-bit kernel (except ia64), use -m64: - # - ifneq ($(uname_M),ia64) - ifneq ($(patsubst %64,%,$(uname_M)),$(uname_M)) - MBITS := -m64 - endif - endif -endif - # CFLAGS and LDFLAGS are for the users to override from the command line. # @@ -212,7 +198,7 @@ ifndef PERF_DEBUG CFLAGS_OPTIMIZE = -O6 endif -CFLAGS = $(MBITS) -ggdb3 -Wall -Wextra -std=gnu99 -Werror $(CFLAGS_OPTIMIZE) -D_FORTIFY_SOURCE=2 $(EXTRA_WARNINGS) +CFLAGS = -ggdb3 -Wall -Wextra -std=gnu99 -Werror $(CFLAGS_OPTIMIZE) -D_FORTIFY_SOURCE=2 $(EXTRA_WARNINGS) $(EXTRA_CFLAGS) EXTLIBS = -lpthread -lrt -lelf -lm ALL_CFLAGS = $(CFLAGS) ALL_LDFLAGS = $(LDFLAGS) -- cgit v1.2.3-18-g5258 From bfd451184d80301d1ae970b1ebffde1e9c6240f9 Mon Sep 17 00:00:00 2001 From: Simon Kaempflein Date: Mon, 16 Nov 2009 15:25:53 +1000 Subject: perf record, x86: Print more intelligent error message when sampling fails Print more accurate error message when "perf record" fails because there is no APIC support, on x86. Signed-off-by: Ingo Molnar --- tools/perf/builtin-record.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'tools/perf') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 87f98fdb051..0e519c667e3 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -307,6 +307,12 @@ try_again: printf("\n"); error("perfcounter syscall returned with %d (%s)\n", fd[nr_cpu][counter], strerror(err)); + +#if defined(__i386__) || defined(__x86_64__) + if (attr->type == PERF_TYPE_HARDWARE && err == EOPNOTSUPP) + die("No hardware sampling interrupt available. No APIC? If so then you can boot the kernel with the \"lapic\" boot parameter to force-enable it.\n"); +#endif + die("No CONFIG_PERF_EVENTS=y kernel support configured?\n"); exit(-1); } -- cgit v1.2.3-18-g5258 From 1b290d670ffa883b7e062177463a8efd00eaa2c1 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Mon, 23 Nov 2009 15:42:35 +0100 Subject: perf tools: Add support for breakpoint events in perf tools Add the breakpoint events support with this new sysnopsis: mem:addr[:access] Where addr is a raw addr value in the kernel and access can be either [r][w][x] Example to profile tasklist_lock: $ grep tasklist_lock /proc/kallsyms ffffffff8189c000 D tasklist_lock $ perf record -e mem:0xffffffff8189c000:rw -a -f -c 1 $ perf report # Samples: 62 # # Overhead Command Shared Object Symbol # ........ ............... ............. ...... # 29.03% swapper [kernel] [k] _raw_read_trylock 29.03% swapper [kernel] [k] _raw_read_unlock 19.35% init [kernel] [k] _raw_read_trylock 19.35% init [kernel] [k] _raw_read_unlock 1.61% events/0 [kernel] [k] _raw_read_trylock 1.61% events/0 [kernel] [k] _raw_read_unlock Coming soon: - Support for symbols in the event definition. - Default period to 1 for breakpoint events because these are not high frequency events. The same thing is needed for trace events. Signed-off-by: Frederic Weisbecker Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Paul Mackerras Cc: Prasad LKML-Reference: <1258987355-8751-4-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Paul Mackerras Cc: Prasad --- tools/perf/Documentation/perf-record.txt | 16 ++++-- tools/perf/util/parse-events.c | 84 +++++++++++++++++++++++++++++++- 2 files changed, 95 insertions(+), 5 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/Documentation/perf-record.txt b/tools/perf/Documentation/perf-record.txt index 0ff23de9e45..fc46c0b40f6 100644 --- a/tools/perf/Documentation/perf-record.txt +++ b/tools/perf/Documentation/perf-record.txt @@ -26,11 +26,19 @@ OPTIONS -e:: --event=:: - Select the PMU event. Selection can be a symbolic event name - (use 'perf list' to list all events) or a raw PMU - event (eventsel+umask) in the form of rNNN where NNN is a - hexadecimal event descriptor. + Select the PMU event. Selection can be: + - a symbolic event name (use 'perf list' to list all events) + + - a raw PMU event (eventsel+umask) in the form of rNNN where NNN is a + hexadecimal event descriptor. + + - a hardware breakpoint event in the form of '\mem:addr[:access]' + where addr is the address in memory you want to break in. + Access is the memory access type (read, write, execute) it can + be passed as follows: '\mem:addr[:[r][w][x]]'. + If you want to profile read-write accesses in 0x1000, just set + 'mem:0x1000:rw'. -a:: System-wide collection. diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c index 0faf4f2bb5c..07002746927 100644 --- a/tools/perf/util/parse-events.c +++ b/tools/perf/util/parse-events.c @@ -1,4 +1,4 @@ - +#include "../../../include/linux/hw_breakpoint.h" #include "util.h" #include "../perf.h" #include "parse-options.h" @@ -540,6 +540,81 @@ static enum event_result parse_tracepoint_event(const char **strp, attr, strp); } +static enum event_result +parse_breakpoint_type(const char *type, const char **strp, + struct perf_event_attr *attr) +{ + int i; + + for (i = 0; i < 3; i++) { + if (!type[i]) + break; + + switch (type[i]) { + case 'r': + attr->bp_type |= HW_BREAKPOINT_R; + break; + case 'w': + attr->bp_type |= HW_BREAKPOINT_W; + break; + case 'x': + attr->bp_type |= HW_BREAKPOINT_X; + break; + default: + return EVT_FAILED; + } + } + if (!attr->bp_type) /* Default */ + attr->bp_type = HW_BREAKPOINT_R | HW_BREAKPOINT_W; + + *strp = type + i; + + return EVT_HANDLED; +} + +static enum event_result +parse_breakpoint_event(const char **strp, struct perf_event_attr *attr) +{ + const char *target; + const char *type; + char *endaddr; + u64 addr; + enum event_result err; + + target = strchr(*strp, ':'); + if (!target) + return EVT_FAILED; + + if (strncmp(*strp, "mem", target - *strp) != 0) + return EVT_FAILED; + + target++; + + addr = strtoull(target, &endaddr, 0); + if (target == endaddr) + return EVT_FAILED; + + attr->bp_addr = addr; + *strp = endaddr; + + type = strchr(target, ':'); + + /* If no type is defined, just rw as default */ + if (!type) { + attr->bp_type = HW_BREAKPOINT_R | HW_BREAKPOINT_W; + } else { + err = parse_breakpoint_type(++type, strp, attr); + if (err == EVT_FAILED) + return EVT_FAILED; + } + + /* We should find a nice way to override the access type */ + attr->bp_len = HW_BREAKPOINT_LEN_4; + attr->type = PERF_TYPE_BREAKPOINT; + + return EVT_HANDLED; +} + static int check_events(const char *str, unsigned int i) { int n; @@ -673,6 +748,10 @@ parse_event_symbols(const char **str, struct perf_event_attr *attr) if (ret != EVT_FAILED) goto modifier; + ret = parse_breakpoint_event(str, attr); + if (ret != EVT_FAILED) + goto modifier; + fprintf(stderr, "invalid or unsupported event: '%s'\n", *str); fprintf(stderr, "Run 'perf list' for a list of valid events\n"); return EVT_FAILED; @@ -859,6 +938,9 @@ void print_events(void) "rNNN"); printf("\n"); + printf(" %-42s [hardware breakpoint]\n", "mem:[:access]"); + printf("\n"); + print_tracepoint_events(); exit(129); -- cgit v1.2.3-18-g5258 From cc612d8199089413719397c9d92e5823da578eac Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 23 Nov 2009 16:39:10 -0200 Subject: perf symbols: Look for vmlinux in more places MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Now that we can check the buildid to see if it really matches, this can be done safely: vmlinux /boot/vmlinux /boot/vmlinux- /lib/modules//build/vmlinux /usr/lib/debug/lib/modules/%s/vmlinux More can be added - if you know about distros that put the vmlinux somewhere else please let us know. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1259001550-8194-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/builtin-annotate.c | 3 +- tools/perf/builtin-kmem.c | 2 +- tools/perf/builtin-report.c | 6 ++- tools/perf/builtin-sched.c | 3 +- tools/perf/builtin-top.c | 3 +- tools/perf/builtin-trace.c | 3 +- tools/perf/util/data_map.c | 4 +- tools/perf/util/data_map.h | 2 + tools/perf/util/header.c | 2 +- tools/perf/util/symbol.c | 113 +++++++++++++++++++++++++++++++++++++----- tools/perf/util/symbol.h | 4 +- 11 files changed, 122 insertions(+), 23 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c index 203152729a6..6b13a1ecf1e 100644 --- a/tools/perf/builtin-annotate.c +++ b/tools/perf/builtin-annotate.c @@ -37,6 +37,7 @@ static bool use_modules; static unsigned long page_size; static unsigned long mmap_window = 32; +const char *vmlinux_name; struct sym_hist { u64 sum; @@ -637,7 +638,7 @@ static int __cmd_annotate(void) exit(0); } - if (kernel_maps__init(use_modules) < 0) { + if (kernel_maps__init(vmlinux_name, true, use_modules) < 0) { pr_err("failed to create kernel maps for symbol resolution\b"); return -1; } diff --git a/tools/perf/builtin-kmem.c b/tools/perf/builtin-kmem.c index 4145049e7bf..5d8aeae5000 100644 --- a/tools/perf/builtin-kmem.c +++ b/tools/perf/builtin-kmem.c @@ -291,7 +291,7 @@ static int read_events(void) register_idle_thread(); register_perf_file_handler(&file_handler); - return mmap_dispatch_perf_file(&header, input_name, 0, 0, + return mmap_dispatch_perf_file(&header, input_name, NULL, false, 0, 0, &cwdlen, &cwd); } diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index 7e690f73b51..fe474b7f8ad 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -52,6 +52,7 @@ static char *pretty_printing_style = default_pretty_printing_style; static int exclude_other = 1; static char callchain_default_opt[] = "fractal,0.5"; +const char *vmlinux_name; static char *cwd; static int cwdlen; @@ -925,8 +926,9 @@ static int __cmd_report(void) register_perf_file_handler(&file_handler); - ret = mmap_dispatch_perf_file(&header, input_name, force, full_paths, - &cwdlen, &cwd); + ret = mmap_dispatch_perf_file(&header, input_name, vmlinux_name, + !vmlinux_name, force, + full_paths, &cwdlen, &cwd); if (ret) return ret; diff --git a/tools/perf/builtin-sched.c b/tools/perf/builtin-sched.c index df44b756cec..260f57a72ee 100644 --- a/tools/perf/builtin-sched.c +++ b/tools/perf/builtin-sched.c @@ -1718,7 +1718,8 @@ static int read_events(void) register_idle_thread(); register_perf_file_handler(&file_handler); - return mmap_dispatch_perf_file(&header, input_name, 0, 0, &cwdlen, &cwd); + return mmap_dispatch_perf_file(&header, input_name, NULL, false, 0, 0, + &cwdlen, &cwd); } static void print_bad_events(void) diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index ea49c2e9dda..eef9caab6ee 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -79,6 +79,7 @@ static int dump_symtab = 0; static bool hide_kernel_symbols = false; static bool hide_user_symbols = false; static struct winsize winsize; +const char *vmlinux_name; static const char *graph_line = "_____________________________________________________________________" "_____________________________________________________________________"; @@ -1341,7 +1342,7 @@ int cmd_top(int argc, const char **argv, const char *prefix __used) if (delay_secs < 1) delay_secs = 1; - err = kernel_maps__init(true); + err = kernel_maps__init(vmlinux_name, !vmlinux_name, true); if (err < 0) return err; parse_source(sym_filter_entry); diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-trace.c index d042d656c56..b71198e5dc1 100644 --- a/tools/perf/builtin-trace.c +++ b/tools/perf/builtin-trace.c @@ -131,7 +131,8 @@ static int __cmd_trace(void) register_idle_thread(); register_perf_file_handler(&file_handler); - return mmap_dispatch_perf_file(&header, input_name, 0, 0, &cwdlen, &cwd); + return mmap_dispatch_perf_file(&header, input_name, NULL, false, + 0, 0, &cwdlen, &cwd); } static const char * const annotate_usage[] = { diff --git a/tools/perf/util/data_map.c b/tools/perf/util/data_map.c index e7b6c2bea3d..f318d19b256 100644 --- a/tools/perf/util/data_map.c +++ b/tools/perf/util/data_map.c @@ -101,6 +101,8 @@ out: int mmap_dispatch_perf_file(struct perf_header **pheader, const char *input_name, + const char *vmlinux_name, + bool try_vmlinux_path, int force, int full_paths, int *cwdlen, @@ -171,7 +173,7 @@ int mmap_dispatch_perf_file(struct perf_header **pheader, goto out_delete; err = -ENOMEM; - if (kernel_maps__init(true) < 0) { + if (kernel_maps__init(vmlinux_name, try_vmlinux_path, true) < 0) { pr_err("failed to setup the kernel maps to resolve symbols\n"); goto out_delete; } diff --git a/tools/perf/util/data_map.h b/tools/perf/util/data_map.h index ae036ecd762..3f0d21b3819 100644 --- a/tools/perf/util/data_map.h +++ b/tools/perf/util/data_map.h @@ -23,6 +23,8 @@ struct perf_file_handler { void register_perf_file_handler(struct perf_file_handler *handler); int mmap_dispatch_perf_file(struct perf_header **pheader, const char *input_name, + const char *vmlinux_name, + bool try_vmlinux_path, int force, int full_paths, int *cwdlen, diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index ac3410b8e9e..1332f8ec04a 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -257,7 +257,7 @@ static int perf_header__adds_write(struct perf_header *self, int fd) * Read the kernel buildid nad the list of loaded modules with * its build_ids: */ - kernel_maps__init(true); + kernel_maps__init(NULL, false, true); /* Write build-ids */ buildid_sec->offset = lseek(fd, 0, SEEK_CUR); diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 74b5b8a1695..44d81d5ae8c 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -34,6 +34,8 @@ static void kernel_maps__insert(struct map *map); static int dso__load_kernel_sym(struct dso *self, struct map *map, symbol_filter_t filter); unsigned int symbol__priv_size; +static int vmlinux_path__nr_entries; +static char **vmlinux_path; static struct rb_root kernel_maps; @@ -1386,15 +1388,43 @@ static int dso__load_vmlinux(struct dso *self, struct map *map, static int dso__load_kernel_sym(struct dso *self, struct map *map, symbol_filter_t filter) { - int err = dso__load_vmlinux(self, map, self->name, filter); + int err; + bool is_kallsyms; + + if (vmlinux_path != NULL) { + int i; + pr_debug("Looking at the vmlinux_path (%d entries long)\n", + vmlinux_path__nr_entries); + for (i = 0; i < vmlinux_path__nr_entries; ++i) { + err = dso__load_vmlinux(self, map, vmlinux_path[i], + filter); + if (err > 0) { + pr_debug("Using %s for symbols\n", + vmlinux_path[i]); + dso__set_long_name(self, + strdup(vmlinux_path[i])); + goto out_fixup; + } + } + } + + is_kallsyms = self->long_name[0] == '['; + if (is_kallsyms) + goto do_kallsyms; + err = dso__load_vmlinux(self, map, self->long_name, filter); if (err <= 0) { + pr_info("The file %s cannot be used, " + "trying to use /proc/kallsyms...", self->long_name); + sleep(2); +do_kallsyms: err = kernel_maps__load_kallsyms(filter); - if (err > 0) + if (err > 0 && !is_kallsyms) dso__set_long_name(self, strdup("[kernel.kallsyms]")); } if (err > 0) { +out_fixup: map__fixup_start(map); map__fixup_end(map); } @@ -1403,9 +1433,7 @@ static int dso__load_kernel_sym(struct dso *self, struct map *map, } LIST_HEAD(dsos); -struct dso *vdso; - -const char *vmlinux_name = "vmlinux"; +struct dso *vdso; static void dsos__add(struct dso *dso) { @@ -1457,9 +1485,9 @@ size_t dsos__fprintf_buildid(FILE *fp) return ret; } -static int kernel_maps__create_kernel_map(void) +static int kernel_maps__create_kernel_map(const char *vmlinux_name) { - struct dso *kernel = dso__new(vmlinux_name); + struct dso *kernel = dso__new(vmlinux_name ?: "[kernel.kallsyms]"); if (kernel == NULL) return -1; @@ -1468,10 +1496,10 @@ static int kernel_maps__create_kernel_map(void) if (kernel_map == NULL) goto out_delete_kernel_dso; - kernel_map->map_ip = kernel_map->unmap_ip = identity__map_ip; + kernel_map->map_ip = kernel_map->unmap_ip = identity__map_ip; + kernel->short_name = "[kernel]"; + kernel->kernel = 1; - kernel->short_name = "[kernel]"; - kernel->kernel = 1; vdso = dso__new("[vdso]"); if (vdso == NULL) goto out_delete_kernel_map; @@ -1494,11 +1522,72 @@ out_delete_kernel_dso: return -1; } -int kernel_maps__init(bool use_modules) +static void vmlinux_path__exit(void) +{ + while (--vmlinux_path__nr_entries >= 0) { + free(vmlinux_path[vmlinux_path__nr_entries]); + vmlinux_path[vmlinux_path__nr_entries] = NULL; + } + + free(vmlinux_path); + vmlinux_path = NULL; +} + +static int vmlinux_path__init(void) +{ + struct utsname uts; + char bf[PATH_MAX]; + + if (uname(&uts) < 0) + return -1; + + vmlinux_path = malloc(sizeof(char *) * 5); + if (vmlinux_path == NULL) + return -1; + + vmlinux_path[vmlinux_path__nr_entries] = strdup("vmlinux"); + if (vmlinux_path[vmlinux_path__nr_entries] == NULL) + goto out_fail; + ++vmlinux_path__nr_entries; + vmlinux_path[vmlinux_path__nr_entries] = strdup("/boot/vmlinux"); + if (vmlinux_path[vmlinux_path__nr_entries] == NULL) + goto out_fail; + ++vmlinux_path__nr_entries; + snprintf(bf, sizeof(bf), "/boot/vmlinux-%s", uts.release); + vmlinux_path[vmlinux_path__nr_entries] = strdup(bf); + if (vmlinux_path[vmlinux_path__nr_entries] == NULL) + goto out_fail; + ++vmlinux_path__nr_entries; + snprintf(bf, sizeof(bf), "/lib/modules/%s/build/vmlinux", uts.release); + vmlinux_path[vmlinux_path__nr_entries] = strdup(bf); + if (vmlinux_path[vmlinux_path__nr_entries] == NULL) + goto out_fail; + ++vmlinux_path__nr_entries; + snprintf(bf, sizeof(bf), "/usr/lib/debug/lib/modules/%s/vmlinux", + uts.release); + vmlinux_path[vmlinux_path__nr_entries] = strdup(bf); + if (vmlinux_path[vmlinux_path__nr_entries] == NULL) + goto out_fail; + ++vmlinux_path__nr_entries; + + return 0; + +out_fail: + vmlinux_path__exit(); + return -1; +} + +int kernel_maps__init(const char *vmlinux_name, bool try_vmlinux_path, + bool use_modules) { - if (kernel_maps__create_kernel_map() < 0) + if (try_vmlinux_path && vmlinux_path__init() < 0) return -1; + if (kernel_maps__create_kernel_map(vmlinux_name) < 0) { + vmlinux_path__exit(); + return -1; + } + if (use_modules && kernel_maps__create_module_maps() < 0) pr_debug("Failed to load list of modules in use, " "continuing...\n"); diff --git a/tools/perf/util/symbol.h b/tools/perf/util/symbol.h index 7a129047c47..8c4d026e067 100644 --- a/tools/perf/util/symbol.h +++ b/tools/perf/util/symbol.h @@ -93,7 +93,8 @@ int sysfs__read_build_id(const char *filename, void *bf, size_t size); bool dsos__read_build_ids(void); int build_id__sprintf(u8 *self, int len, char *bf); -int kernel_maps__init(bool use_modules); +int kernel_maps__init(const char *vmlinux_name, bool try_vmlinux_path, + bool use_modules); size_t kernel_maps__fprintf(FILE *fp); void symbol__init(unsigned int priv_size); @@ -101,5 +102,4 @@ void symbol__init(unsigned int priv_size); extern struct list_head dsos; extern struct map *kernel_map; extern struct dso *vdso; -extern const char *vmlinux_name; #endif /* __PERF_SYMBOL */ -- cgit v1.2.3-18-g5258 From 2890284bcf5c13c10fae8a0c20ad2f575118a092 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 23 Nov 2009 17:51:08 -0200 Subject: perf tools: Move graph_line and graph_dotted_line from top MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit So that they can be used in other tools. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1259005869-13487-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/builtin-top.c | 7 ------- tools/perf/util/ctype.c | 8 ++++++++ tools/perf/util/util.h | 3 +++ 3 files changed, 11 insertions(+), 7 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index eef9caab6ee..6a5de90e9b8 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -80,13 +80,6 @@ static bool hide_kernel_symbols = false; static bool hide_user_symbols = false; static struct winsize winsize; const char *vmlinux_name; -static const char *graph_line = - "_____________________________________________________________________" - "_____________________________________________________________________"; -static const char *graph_dotted_line = - "---------------------------------------------------------------------" - "---------------------------------------------------------------------" - "---------------------------------------------------------------------"; /* * Source diff --git a/tools/perf/util/ctype.c b/tools/perf/util/ctype.c index 0b791bd346b..35073621e5d 100644 --- a/tools/perf/util/ctype.c +++ b/tools/perf/util/ctype.c @@ -29,3 +29,11 @@ unsigned char sane_ctype[256] = { A, A, A, A, A, A, A, A, A, A, A, R, R, P, P, 0, /* 112..127 */ /* Nothing in the 128.. range */ }; + +const char *graph_line = + "_____________________________________________________________________" + "_____________________________________________________________________"; +const char *graph_dotted_line = + "---------------------------------------------------------------------" + "---------------------------------------------------------------------" + "---------------------------------------------------------------------"; diff --git a/tools/perf/util/util.h b/tools/perf/util/util.h index f2203a0946b..e1c623e0c99 100644 --- a/tools/perf/util/util.h +++ b/tools/perf/util/util.h @@ -84,6 +84,9 @@ #include #endif +extern const char *graph_line; +extern const char *graph_dotted_line; + /* On most systems would have given us this, but * not on some systems (e.g. GNU/Hurd). */ -- cgit v1.2.3-18-g5258 From 1b145ae58035f30353d78d25bea665091df9b438 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 23 Nov 2009 17:51:09 -0200 Subject: perf kmem: Resolve symbols MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit E.g.: [root@doppio linux-2.6-tip]# perf kmem record sleep 3s [ perf record: Woken up 2 times to write data ] [ perf record: Captured and wrote 0.804 MB perf.data (~35105 samples) ] [root@doppio linux-2.6-tip]# perf kmem --stat caller | head -10 ------------------------------------------------------------------------------ Callsite |Total_alloc/Per | Total_req/Per | Hit | Frag ------------------------------------------------------------------------------ getname/40 | 1519616/4096 | 1519616/4096 | 371| 0.000% seq_read/a2 | 987136/4096 | 987136/4096 | 241| 0.000% __netdev_alloc_skb/43 | 260368/1049 | 259968/1048 | 248| 0.154% __alloc_skb/5a | 77312/256 | 77312/256 | 302| 0.000% proc_alloc_inode/33 | 76480/632 | 76472/632 | 121| 0.010% get_empty_filp/8d | 70272/192 | 70272/192 | 366| 0.000% split_vma/8e | 42064/176 | 42064/176 | 239| 0.000% [root@doppio linux-2.6-tip]# Signed-off-by: Arnaldo Carvalho de Melo Acked-by: Pekka Enberg Cc: Eduard - Gabriel Munteanu Cc: Frédéric Weisbecker Cc: linux-mm@kvack.org Cc: Li Zefan Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Steven Rostedt LKML-Reference: <1259005869-13487-2-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/builtin-kmem.c | 37 +++++++++++++++++++++++-------------- 1 file changed, 23 insertions(+), 14 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-kmem.c b/tools/perf/builtin-kmem.c index 5d8aeae5000..256d18fa047 100644 --- a/tools/perf/builtin-kmem.c +++ b/tools/perf/builtin-kmem.c @@ -307,25 +307,34 @@ static void __print_result(struct rb_root *root, int n_lines, int is_caller) { struct rb_node *next; - printf("\n ------------------------------------------------------------------------------\n"); - if (is_caller) - printf(" Callsite |"); - else - printf(" Alloc Ptr |"); - printf(" Total_alloc/Per | Total_req/Per | Hit | Fragmentation\n"); - printf(" ------------------------------------------------------------------------------\n"); + printf("%.78s\n", graph_dotted_line); + printf("%-28s|", is_caller ? "Callsite": "Alloc Ptr"); + printf("Total_alloc/Per | Total_req/Per | Hit | Frag\n"); + printf("%.78s\n", graph_dotted_line); next = rb_first(root); while (next && n_lines--) { - struct alloc_stat *data; - - data = rb_entry(next, struct alloc_stat, node); + struct alloc_stat *data = rb_entry(next, struct alloc_stat, + node); + struct symbol *sym = NULL; + char bf[BUFSIZ]; + u64 addr; + + if (is_caller) { + addr = data->call_site; + sym = kernel_maps__find_symbol(addr, NULL, NULL); + } else + addr = data->ptr; + + if (sym != NULL) + snprintf(bf, sizeof(bf), "%s/%Lx", sym->name, + addr - sym->start); + else + snprintf(bf, sizeof(bf), "%#Lx", addr); - printf(" %-16p | %8llu/%-6lu | %8llu/%-6lu | %6lu | %8.3f%%\n", - is_caller ? (void *)(unsigned long)data->call_site : - (void *)(unsigned long)data->ptr, - (unsigned long long)data->bytes_alloc, + printf("%-28s|%8llu/%-6lu |%8llu/%-6lu|%6lu|%8.3f%%\n", + bf, (unsigned long long)data->bytes_alloc, (unsigned long)data->bytes_alloc / data->hit, (unsigned long long)data->bytes_req, (unsigned long)data->bytes_req / data->hit, -- cgit v1.2.3-18-g5258 From ee3d250446f1c1be4eceab48f3a23794d9a6564c Mon Sep 17 00:00:00 2001 From: Paul Mackerras Date: Tue, 24 Nov 2009 15:19:43 +1100 Subject: perf tools: Fix compilation on powerpc Currently, perf fails to compile on powerpc with this error: CC util/header.o In file included from util/../perf.h:17, from util/header.c:9: util/../../../arch/powerpc/include/asm/unistd.h:360:27: error: linux/linkage.h: No such file or directory make: *** [util/header.o] Error 1 The reason is that we still have a #define __KERNEL__ in effect at the point where gets included, which means we get extra stuff that we don't need or want. This fixes the problem by undefining __KERNEL__ once we have included the file for which we need __KERNEL__ defined. Signed-off-by: Paul Mackerras Cc: Frederic Weisbecker Cc: Arnaldo Carvalho de Melo Cc: Peter Zijlstra LKML-Reference: <19211.24287.453183.78836@cargo.ozlabs.ibm.com> Signed-off-by: Ingo Molnar --- tools/perf/util/include/linux/bitops.h | 2 ++ 1 file changed, 2 insertions(+) (limited to 'tools/perf') diff --git a/tools/perf/util/include/linux/bitops.h b/tools/perf/util/include/linux/bitops.h index ace57c36d1d..8d63116e943 100644 --- a/tools/perf/util/include/linux/bitops.h +++ b/tools/perf/util/include/linux/bitops.h @@ -7,6 +7,8 @@ #define CONFIG_GENERIC_FIND_FIRST_BIT #include "../../../../include/linux/bitops.h" +#undef __KERNEL__ + static inline void set_bit(int nr, unsigned long *addr) { addr[nr / BITS_PER_LONG] |= 1UL << (nr % BITS_PER_LONG); -- cgit v1.2.3-18-g5258 From 7707b6b6f8d9188b612f9fc88c65411264b1ed57 Mon Sep 17 00:00:00 2001 From: Li Zefan Date: Tue, 24 Nov 2009 13:25:48 +0800 Subject: perf kmem: Add new option to show raw ip Add option "--raw-ip" to show raw ip instead of symbols: # ./perf kmem --stat caller --raw-ip ------------------------------------------------------------------------------ Callsite |Total_alloc/Per | Total_req/Per | Hit | Frag ------------------------------------------------------------------------------ 0xc05301aa | 733184/4096 | 733184/4096 | 179| 0.000% 0xc0542ba0 | 483328/4096 | 483328/4096 | 118| 0.000% ... Also show symbols with format sym+offset instead of sym/offset. Signed-off-by: Li Zefan Acked-by: Pekka Enberg Cc: Eduard - Gabriel Munteanu Cc: Peter Zijlstra Cc: Frederic Weisbecker Cc: linux-mm@kvack.org LKML-Reference: <4B0B6E5C.4080900@cn.fujitsu.com> Signed-off-by: Ingo Molnar --- tools/perf/builtin-kmem.c | 18 ++++++++++-------- 1 file changed, 10 insertions(+), 8 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-kmem.c b/tools/perf/builtin-kmem.c index 256d18fa047..1ef43c212d9 100644 --- a/tools/perf/builtin-kmem.c +++ b/tools/perf/builtin-kmem.c @@ -32,15 +32,14 @@ sort_fn_t caller_sort_fn; static int alloc_lines = -1; static int caller_lines = -1; +static bool raw_ip; + static char *cwd; static int cwdlen; struct alloc_stat { union { - struct { - char *name; - u64 call_site; - }; + u64 call_site; u64 ptr; }; u64 bytes_req; @@ -323,12 +322,14 @@ static void __print_result(struct rb_root *root, int n_lines, int is_caller) if (is_caller) { addr = data->call_site; - sym = kernel_maps__find_symbol(addr, NULL, NULL); + if (!raw_ip) + sym = kernel_maps__find_symbol(addr, + NULL, NULL); } else addr = data->ptr; if (sym != NULL) - snprintf(bf, sizeof(bf), "%s/%Lx", sym->name, + snprintf(bf, sizeof(bf), "%s+%Lx", sym->name, addr - sym->start); else snprintf(bf, sizeof(bf), "%#Lx", addr); @@ -345,9 +346,9 @@ static void __print_result(struct rb_root *root, int n_lines, int is_caller) } if (n_lines == -1) - printf(" ... | ... | ... | ... | ... \n"); + printf(" ... | ... | ... | ... | ... \n"); - printf(" ------------------------------------------------------------------------------\n"); + printf("%.78s\n", graph_dotted_line); } static void print_summary(void) @@ -558,6 +559,7 @@ static const struct option kmem_options[] = { OPT_CALLBACK('l', "line", NULL, "num", "show n lins", parse_line_opt), + OPT_BOOLEAN(0, "raw-ip", &raw_ip, "show raw ip instead of symbol"), OPT_END() }; -- cgit v1.2.3-18-g5258 From 29b3e15289eb66788a0bf5ea4903f9fbeb1ec751 Mon Sep 17 00:00:00 2001 From: Li Zefan Date: Tue, 24 Nov 2009 13:26:10 +0800 Subject: perf kmem: Default to sort by fragmentation Make the output sort by fragmentation by default. Also make the usage of "--sort" option consistent with other perf tools. That is, we support multi keys: "--sort key1[,key2]...". # ./perf kmem --stat caller ------------------------------------------------------------------------------ Callsite |Total_alloc/Per | Total_req/Per | Hit | Frag ------------------------------------------------------------------------------ __netdev_alloc_skb+23 | 5048/1682 | 4564/1521 | 3| 9.588% perf_event_alloc.clone.0+0 | 7504/682 | 7128/648 | 11| 5.011% tracepoint_add_probe+32e | 157/31 | 154/30 | 5| 1.911% alloc_buffer_head+16 | 456/57 | 448/56 | 8| 1.754% radix_tree_preload+51 | 584/292 | 576/288 | 2| 1.370% ... TODO: - Extract duplicate code in builtin-kmem.c and builtin-sched.c into util/sort.c. Signed-off-by: Li Zefan Acked-by: Pekka Enberg Cc: Eduard - Gabriel Munteanu Cc: Peter Zijlstra Cc: Frederic Weisbecker Cc: linux-mm@kvack.org LKML-Reference: <4B0B6E72.7010200@cn.fujitsu.com> Signed-off-by: Ingo Molnar --- tools/perf/builtin-kmem.c | 142 +++++++++++++++++++++++++++++++++++----------- 1 file changed, 108 insertions(+), 34 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-kmem.c b/tools/perf/builtin-kmem.c index 1ef43c212d9..dc86f1e64b6 100644 --- a/tools/perf/builtin-kmem.c +++ b/tools/perf/builtin-kmem.c @@ -26,14 +26,13 @@ static u64 sample_type; static int alloc_flag; static int caller_flag; -sort_fn_t alloc_sort_fn; -sort_fn_t caller_sort_fn; - static int alloc_lines = -1; static int caller_lines = -1; static bool raw_ip; +static char default_sort_order[] = "frag,hit,bytes"; + static char *cwd; static int cwdlen; @@ -371,20 +370,34 @@ static void print_result(void) print_summary(); } +struct sort_dimension { + const char name[20]; + sort_fn_t cmp; + struct list_head list; +}; + +static LIST_HEAD(caller_sort); +static LIST_HEAD(alloc_sort); + static void sort_insert(struct rb_root *root, struct alloc_stat *data, - sort_fn_t sort_fn) + struct list_head *sort_list) { struct rb_node **new = &(root->rb_node); struct rb_node *parent = NULL; + struct sort_dimension *sort; while (*new) { struct alloc_stat *this; - int cmp; + int cmp = 0; this = rb_entry(*new, struct alloc_stat, node); parent = *new; - cmp = sort_fn(data, this); + list_for_each_entry(sort, sort_list, list) { + cmp = sort->cmp(data, this); + if (cmp) + break; + } if (cmp > 0) new = &((*new)->rb_left); @@ -397,7 +410,7 @@ static void sort_insert(struct rb_root *root, struct alloc_stat *data, } static void __sort_result(struct rb_root *root, struct rb_root *root_sorted, - sort_fn_t sort_fn) + struct list_head *sort_list) { struct rb_node *node; struct alloc_stat *data; @@ -409,14 +422,14 @@ static void __sort_result(struct rb_root *root, struct rb_root *root_sorted, rb_erase(node, root); data = rb_entry(node, struct alloc_stat, node); - sort_insert(root_sorted, data, sort_fn); + sort_insert(root_sorted, data, sort_list); } } static void sort_result(void) { - __sort_result(&root_alloc_stat, &root_alloc_sorted, alloc_sort_fn); - __sort_result(&root_caller_stat, &root_caller_sorted, caller_sort_fn); + __sort_result(&root_alloc_stat, &root_alloc_sorted, &alloc_sort); + __sort_result(&root_caller_stat, &root_caller_sorted, &caller_sort); } static int __cmd_kmem(void) @@ -434,7 +447,6 @@ static const char * const kmem_usage[] = { NULL }; - static int ptr_cmp(struct alloc_stat *l, struct alloc_stat *r) { if (l->ptr < r->ptr) @@ -444,6 +456,11 @@ static int ptr_cmp(struct alloc_stat *l, struct alloc_stat *r) return 0; } +static struct sort_dimension ptr_sort_dimension = { + .name = "ptr", + .cmp = ptr_cmp, +}; + static int callsite_cmp(struct alloc_stat *l, struct alloc_stat *r) { if (l->call_site < r->call_site) @@ -453,6 +470,11 @@ static int callsite_cmp(struct alloc_stat *l, struct alloc_stat *r) return 0; } +static struct sort_dimension callsite_sort_dimension = { + .name = "callsite", + .cmp = callsite_cmp, +}; + static int hit_cmp(struct alloc_stat *l, struct alloc_stat *r) { if (l->hit < r->hit) @@ -462,6 +484,11 @@ static int hit_cmp(struct alloc_stat *l, struct alloc_stat *r) return 0; } +static struct sort_dimension hit_sort_dimension = { + .name = "hit", + .cmp = hit_cmp, +}; + static int bytes_cmp(struct alloc_stat *l, struct alloc_stat *r) { if (l->bytes_alloc < r->bytes_alloc) @@ -471,6 +498,11 @@ static int bytes_cmp(struct alloc_stat *l, struct alloc_stat *r) return 0; } +static struct sort_dimension bytes_sort_dimension = { + .name = "bytes", + .cmp = bytes_cmp, +}; + static int frag_cmp(struct alloc_stat *l, struct alloc_stat *r) { double x, y; @@ -485,31 +517,73 @@ static int frag_cmp(struct alloc_stat *l, struct alloc_stat *r) return 0; } +static struct sort_dimension frag_sort_dimension = { + .name = "frag", + .cmp = frag_cmp, +}; + +static struct sort_dimension *avail_sorts[] = { + &ptr_sort_dimension, + &callsite_sort_dimension, + &hit_sort_dimension, + &bytes_sort_dimension, + &frag_sort_dimension, +}; + +#define NUM_AVAIL_SORTS \ + (int)(sizeof(avail_sorts) / sizeof(struct sort_dimension *)) + +static int sort_dimension__add(const char *tok, struct list_head *list) +{ + struct sort_dimension *sort; + int i; + + for (i = 0; i < NUM_AVAIL_SORTS; i++) { + if (!strcmp(avail_sorts[i]->name, tok)) { + sort = malloc(sizeof(*sort)); + if (!sort) + die("malloc"); + memcpy(sort, avail_sorts[i], sizeof(*sort)); + list_add_tail(&sort->list, list); + return 0; + } + } + + return -1; +} + +static int setup_sorting(struct list_head *sort_list, const char *arg) +{ + char *tok; + char *str = strdup(arg); + + if (!str) + die("strdup"); + + while (true) { + tok = strsep(&str, ","); + if (!tok) + break; + if (sort_dimension__add(tok, sort_list) < 0) { + error("Unknown --sort key: '%s'", tok); + return -1; + } + } + + free(str); + return 0; +} + static int parse_sort_opt(const struct option *opt __used, const char *arg, int unset __used) { - sort_fn_t sort_fn; - if (!arg) return -1; - if (strcmp(arg, "ptr") == 0) - sort_fn = ptr_cmp; - else if (strcmp(arg, "call_site") == 0) - sort_fn = callsite_cmp; - else if (strcmp(arg, "hit") == 0) - sort_fn = hit_cmp; - else if (strcmp(arg, "bytes") == 0) - sort_fn = bytes_cmp; - else if (strcmp(arg, "frag") == 0) - sort_fn = frag_cmp; - else - return -1; - if (caller_flag > alloc_flag) - caller_sort_fn = sort_fn; + return setup_sorting(&caller_sort, arg); else - alloc_sort_fn = sort_fn; + return setup_sorting(&alloc_sort, arg); return 0; } @@ -553,8 +627,8 @@ static const struct option kmem_options[] = { OPT_CALLBACK(0, "stat", NULL, "|", "stat selector, Pass 'alloc' or 'caller'.", parse_stat_opt), - OPT_CALLBACK('s', "sort", NULL, "key", - "sort by key: ptr, call_site, hit, bytes, frag", + OPT_CALLBACK('s', "sort", NULL, "key[,key2...]", + "sort by key(s): ptr, call_site, bytes, hit, frag", parse_sort_opt), OPT_CALLBACK('l', "line", NULL, "num", "show n lins", @@ -606,10 +680,10 @@ int cmd_kmem(int argc, const char **argv, const char *prefix __used) else if (argc) usage_with_options(kmem_usage, kmem_options); - if (!alloc_sort_fn) - alloc_sort_fn = bytes_cmp; - if (!caller_sort_fn) - caller_sort_fn = bytes_cmp; + if (list_empty(&caller_sort)) + setup_sorting(&caller_sort, default_sort_order); + if (list_empty(&alloc_sort)) + setup_sorting(&alloc_sort, default_sort_order); return __cmd_kmem(); } -- cgit v1.2.3-18-g5258 From 7d0d39459dab20bf60cac30a1a7d50b286c60cc1 Mon Sep 17 00:00:00 2001 From: Li Zefan Date: Tue, 24 Nov 2009 13:26:31 +0800 Subject: perf kmem: Collect cross node allocation statistics Show cross node memory allocations: # ./perf kmem SUMMARY ======= ... Cross node allocations: 0/3633 Signed-off-by: Li Zefan Acked-by: Pekka Enberg Cc: Eduard - Gabriel Munteanu Cc: Peter Zijlstra Cc: Frederic Weisbecker Cc: linux-mm@kvack.org LKML-Reference: <4B0B6E87.10906@cn.fujitsu.com> Signed-off-by: Ingo Molnar --- tools/perf/builtin-kmem.c | 81 +++++++++++++++++++++++++++++++++++++++++++++-- 1 file changed, 79 insertions(+), 2 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-kmem.c b/tools/perf/builtin-kmem.c index dc86f1e64b6..1ecf3f4415c 100644 --- a/tools/perf/builtin-kmem.c +++ b/tools/perf/builtin-kmem.c @@ -36,6 +36,9 @@ static char default_sort_order[] = "frag,hit,bytes"; static char *cwd; static int cwdlen; +static int *cpunode_map; +static int max_cpu_num; + struct alloc_stat { union { u64 call_site; @@ -54,12 +57,74 @@ static struct rb_root root_caller_stat; static struct rb_root root_caller_sorted; static unsigned long total_requested, total_allocated; +static unsigned long nr_allocs, nr_cross_allocs; struct raw_event_sample { u32 size; char data[0]; }; +#define PATH_SYS_NODE "/sys/devices/system/node" + +static void init_cpunode_map(void) +{ + FILE *fp; + int i; + + fp = fopen("/sys/devices/system/cpu/kernel_max", "r"); + if (!fp) { + max_cpu_num = 4096; + return; + } + + if (fscanf(fp, "%d", &max_cpu_num) < 1) + die("Failed to read 'kernel_max' from sysfs"); + max_cpu_num++; + + cpunode_map = calloc(max_cpu_num, sizeof(int)); + if (!cpunode_map) + die("calloc"); + for (i = 0; i < max_cpu_num; i++) + cpunode_map[i] = -1; + fclose(fp); +} + +static void setup_cpunode_map(void) +{ + struct dirent *dent1, *dent2; + DIR *dir1, *dir2; + unsigned int cpu, mem; + char buf[PATH_MAX]; + + init_cpunode_map(); + + dir1 = opendir(PATH_SYS_NODE); + if (!dir1) + return; + + while (true) { + dent1 = readdir(dir1); + if (!dent1) + break; + + if (sscanf(dent1->d_name, "node%u", &mem) < 1) + continue; + + snprintf(buf, PATH_MAX, "%s/%s", PATH_SYS_NODE, dent1->d_name); + dir2 = opendir(buf); + if (!dir2) + continue; + while (true) { + dent2 = readdir(dir2); + if (!dent2) + break; + if (sscanf(dent2->d_name, "cpu%u", &cpu) < 1) + continue; + cpunode_map[cpu] = mem; + } + } +} + static int process_comm_event(event_t *event, unsigned long offset, unsigned long head) { @@ -157,15 +222,16 @@ static void insert_caller_stat(unsigned long call_site, static void process_alloc_event(struct raw_event_sample *raw, struct event *event, - int cpu __used, + int cpu, u64 timestamp __used, struct thread *thread __used, - int node __used) + int node) { unsigned long call_site; unsigned long ptr; int bytes_req; int bytes_alloc; + int node1, node2; ptr = raw_field_value(event, "ptr", raw->data); call_site = raw_field_value(event, "call_site", raw->data); @@ -177,6 +243,14 @@ static void process_alloc_event(struct raw_event_sample *raw, total_requested += bytes_req; total_allocated += bytes_alloc; + + if (node) { + node1 = cpunode_map[cpu]; + node2 = raw_field_value(event, "node", raw->data); + if (node1 != node2) + nr_cross_allocs++; + } + nr_allocs++; } static void process_free_event(struct raw_event_sample *raw __used, @@ -359,6 +433,7 @@ static void print_summary(void) total_allocated - total_requested); printf("Internal fragmentation: %f%%\n", fragmentation(total_requested, total_allocated)); + printf("Cross CPU allocations: %lu/%lu\n", nr_cross_allocs, nr_allocs); } static void print_result(void) @@ -685,6 +760,8 @@ int cmd_kmem(int argc, const char **argv, const char *prefix __used) if (list_empty(&alloc_sort)) setup_sorting(&alloc_sort, default_sort_order); + setup_cpunode_map(); + return __cmd_kmem(); } -- cgit v1.2.3-18-g5258 From 079d3f653134e2f2ac99dae28b08c0cc64268103 Mon Sep 17 00:00:00 2001 From: Li Zefan Date: Tue, 24 Nov 2009 13:26:55 +0800 Subject: perf kmem: Measure kmalloc/kfree CPU ping-pong call-sites Show statistics for allocations and frees on different cpus: ------------------------------------------------------------------------------------------------------ Callsite | Total_alloc/Per | Total_req/Per | Hit | Ping-pong | Frag ------------------------------------------------------------------------------------------------------ perf_event_alloc.clone.0+0 | 7504/682 | 7128/648 | 11 | 0 | 5.011% alloc_buffer_head+16 | 288/57 | 280/56 | 5 | 0 | 2.778% radix_tree_preload+51 | 296/296 | 288/288 | 1 | 0 | 2.703% tracepoint_add_probe+32e | 157/31 | 154/30 | 5 | 0 | 1.911% do_maps_open+0 | 796/12 | 792/12 | 66 | 0 | 0.503% sock_alloc_send_pskb+16e | 23780/495 | 23744/494 | 48 | 38 | 0.151% anon_vma_prepare+9a | 3744/44 | 3740/44 | 85 | 0 | 0.107% d_alloc+21 | 64948/164 | 64944/164 | 396 | 0 | 0.006% proc_alloc_inode+23 | 262292/676 | 262288/676 | 388 | 0 | 0.002% create_object+28 | 459600/200 | 459600/200 | 2298 | 71 | 0.000% journal_start+67 | 14440/40 | 14440/40 | 361 | 0 | 0.000% get_empty_filp+df | 53504/256 | 53504/256 | 209 | 0 | 0.000% getname+2a | 823296/4096 | 823296/4096 | 201 | 0 | 0.000% seq_read+2b0 | 544768/4096 | 544768/4096 | 133 | 0 | 0.000% seq_open+6d | 17024/128 | 17024/128 | 133 | 0 | 0.000% mmap_region+2e6 | 11704/88 | 11704/88 | 133 | 0 | 0.000% single_open+0 | 1072/16 | 1072/16 | 67 | 0 | 0.000% __alloc_skb+2e | 12544/256 | 12544/256 | 49 | 38 | 0.000% __sigqueue_alloc+4a | 1296/144 | 1296/144 | 9 | 8 | 0.000% tracepoint_add_probe+6f | 80/16 | 80/16 | 5 | 0 | 0.000% ------------------------------------------------------------------------------------------------------ ... Signed-off-by: Li Zefan Acked-by: Pekka Enberg Cc: Eduard - Gabriel Munteanu Cc: Peter Zijlstra Cc: Frederic Weisbecker Cc: linux-mm@kvack.org LKML-Reference: <4B0B6E9F.6020309@cn.fujitsu.com> Signed-off-by: Ingo Molnar --- tools/perf/builtin-kmem.c | 122 +++++++++++++++++++++++++++++++++++----------- 1 file changed, 94 insertions(+), 28 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-kmem.c b/tools/perf/builtin-kmem.c index 1ecf3f4415c..173d6db42ec 100644 --- a/tools/perf/builtin-kmem.c +++ b/tools/perf/builtin-kmem.c @@ -40,13 +40,14 @@ static int *cpunode_map; static int max_cpu_num; struct alloc_stat { - union { - u64 call_site; - u64 ptr; - }; + u64 call_site; + u64 ptr; u64 bytes_req; u64 bytes_alloc; u32 hit; + u32 pingpong; + + short alloc_cpu; struct rb_node node; }; @@ -144,16 +145,13 @@ process_comm_event(event_t *event, unsigned long offset, unsigned long head) return 0; } -static void insert_alloc_stat(unsigned long ptr, - int bytes_req, int bytes_alloc) +static void insert_alloc_stat(unsigned long call_site, unsigned long ptr, + int bytes_req, int bytes_alloc, int cpu) { struct rb_node **node = &root_alloc_stat.rb_node; struct rb_node *parent = NULL; struct alloc_stat *data = NULL; - if (!alloc_flag) - return; - while (*node) { parent = *node; data = rb_entry(*node, struct alloc_stat, node); @@ -172,7 +170,10 @@ static void insert_alloc_stat(unsigned long ptr, data->bytes_alloc += bytes_req; } else { data = malloc(sizeof(*data)); + if (!data) + die("malloc"); data->ptr = ptr; + data->pingpong = 0; data->hit = 1; data->bytes_req = bytes_req; data->bytes_alloc = bytes_alloc; @@ -180,6 +181,8 @@ static void insert_alloc_stat(unsigned long ptr, rb_link_node(&data->node, parent, node); rb_insert_color(&data->node, &root_alloc_stat); } + data->call_site = call_site; + data->alloc_cpu = cpu; } static void insert_caller_stat(unsigned long call_site, @@ -189,9 +192,6 @@ static void insert_caller_stat(unsigned long call_site, struct rb_node *parent = NULL; struct alloc_stat *data = NULL; - if (!caller_flag) - return; - while (*node) { parent = *node; data = rb_entry(*node, struct alloc_stat, node); @@ -210,7 +210,10 @@ static void insert_caller_stat(unsigned long call_site, data->bytes_alloc += bytes_req; } else { data = malloc(sizeof(*data)); + if (!data) + die("malloc"); data->call_site = call_site; + data->pingpong = 0; data->hit = 1; data->bytes_req = bytes_req; data->bytes_alloc = bytes_alloc; @@ -238,7 +241,7 @@ static void process_alloc_event(struct raw_event_sample *raw, bytes_req = raw_field_value(event, "bytes_req", raw->data); bytes_alloc = raw_field_value(event, "bytes_alloc", raw->data); - insert_alloc_stat(ptr, bytes_req, bytes_alloc); + insert_alloc_stat(call_site, ptr, bytes_req, bytes_alloc, cpu); insert_caller_stat(call_site, bytes_req, bytes_alloc); total_requested += bytes_req; @@ -253,12 +256,58 @@ static void process_alloc_event(struct raw_event_sample *raw, nr_allocs++; } -static void process_free_event(struct raw_event_sample *raw __used, - struct event *event __used, - int cpu __used, +static int ptr_cmp(struct alloc_stat *, struct alloc_stat *); +static int callsite_cmp(struct alloc_stat *, struct alloc_stat *); + +static struct alloc_stat *search_alloc_stat(unsigned long ptr, + unsigned long call_site, + struct rb_root *root, + sort_fn_t sort_fn) +{ + struct rb_node *node = root->rb_node; + struct alloc_stat key = { .ptr = ptr, .call_site = call_site }; + + while (node) { + struct alloc_stat *data; + int cmp; + + data = rb_entry(node, struct alloc_stat, node); + + cmp = sort_fn(&key, data); + if (cmp < 0) + node = node->rb_left; + else if (cmp > 0) + node = node->rb_right; + else + return data; + } + return NULL; +} + +static void process_free_event(struct raw_event_sample *raw, + struct event *event, + int cpu, u64 timestamp __used, struct thread *thread __used) { + unsigned long ptr; + struct alloc_stat *s_alloc, *s_caller; + + ptr = raw_field_value(event, "ptr", raw->data); + + s_alloc = search_alloc_stat(ptr, 0, &root_alloc_stat, ptr_cmp); + if (!s_alloc) + return; + + if (cpu != s_alloc->alloc_cpu) { + s_alloc->pingpong++; + + s_caller = search_alloc_stat(0, s_alloc->call_site, + &root_caller_stat, callsite_cmp); + assert(s_caller); + s_caller->pingpong++; + } + s_alloc->alloc_cpu = -1; } static void @@ -379,10 +428,10 @@ static void __print_result(struct rb_root *root, int n_lines, int is_caller) { struct rb_node *next; - printf("%.78s\n", graph_dotted_line); - printf("%-28s|", is_caller ? "Callsite": "Alloc Ptr"); - printf("Total_alloc/Per | Total_req/Per | Hit | Frag\n"); - printf("%.78s\n", graph_dotted_line); + printf("%.102s\n", graph_dotted_line); + printf(" %-34s |", is_caller ? "Callsite": "Alloc Ptr"); + printf(" Total_alloc/Per | Total_req/Per | Hit | Ping-pong | Frag\n"); + printf("%.102s\n", graph_dotted_line); next = rb_first(root); @@ -390,7 +439,7 @@ static void __print_result(struct rb_root *root, int n_lines, int is_caller) struct alloc_stat *data = rb_entry(next, struct alloc_stat, node); struct symbol *sym = NULL; - char bf[BUFSIZ]; + char buf[BUFSIZ]; u64 addr; if (is_caller) { @@ -402,26 +451,28 @@ static void __print_result(struct rb_root *root, int n_lines, int is_caller) addr = data->ptr; if (sym != NULL) - snprintf(bf, sizeof(bf), "%s+%Lx", sym->name, + snprintf(buf, sizeof(buf), "%s+%Lx", sym->name, addr - sym->start); else - snprintf(bf, sizeof(bf), "%#Lx", addr); + snprintf(buf, sizeof(buf), "%#Lx", addr); + printf(" %-34s |", buf); - printf("%-28s|%8llu/%-6lu |%8llu/%-6lu|%6lu|%8.3f%%\n", - bf, (unsigned long long)data->bytes_alloc, + printf(" %9llu/%-5lu | %9llu/%-5lu | %6lu | %8lu | %6.3f%%\n", + (unsigned long long)data->bytes_alloc, (unsigned long)data->bytes_alloc / data->hit, (unsigned long long)data->bytes_req, (unsigned long)data->bytes_req / data->hit, (unsigned long)data->hit, + (unsigned long)data->pingpong, fragmentation(data->bytes_req, data->bytes_alloc)); next = rb_next(next); } if (n_lines == -1) - printf(" ... | ... | ... | ... | ... \n"); + printf(" ... | ... | ... | ... | ... | ... \n"); - printf("%.78s\n", graph_dotted_line); + printf("%.102s\n", graph_dotted_line); } static void print_summary(void) @@ -597,12 +648,27 @@ static struct sort_dimension frag_sort_dimension = { .cmp = frag_cmp, }; +static int pingpong_cmp(struct alloc_stat *l, struct alloc_stat *r) +{ + if (l->pingpong < r->pingpong) + return -1; + else if (l->pingpong > r->pingpong) + return 1; + return 0; +} + +static struct sort_dimension pingpong_sort_dimension = { + .name = "pingpong", + .cmp = pingpong_cmp, +}; + static struct sort_dimension *avail_sorts[] = { &ptr_sort_dimension, &callsite_sort_dimension, &hit_sort_dimension, &bytes_sort_dimension, &frag_sort_dimension, + &pingpong_sort_dimension, }; #define NUM_AVAIL_SORTS \ @@ -703,7 +769,7 @@ static const struct option kmem_options[] = { "stat selector, Pass 'alloc' or 'caller'.", parse_stat_opt), OPT_CALLBACK('s', "sort", NULL, "key[,key2...]", - "sort by key(s): ptr, call_site, bytes, hit, frag", + "sort by keys: ptr, call_site, bytes, hit, pingpong, frag", parse_sort_opt), OPT_CALLBACK('l', "line", NULL, "num", "show n lins", -- cgit v1.2.3-18-g5258 From b23d5767a5818caec8547d0bce1588b02bdecd30 Mon Sep 17 00:00:00 2001 From: Li Zefan Date: Tue, 24 Nov 2009 13:27:11 +0800 Subject: perf kmem: Add help file Add Documentation/perf-kmem.txt Signed-off-by: Li Zefan Acked-by: Pekka Enberg Cc: Eduard - Gabriel Munteanu Cc: Peter Zijlstra Cc: Frederic Weisbecker Cc: linux-mm@kvack.org LKML-Reference: <4B0B6EAF.80802@cn.fujitsu.com> Signed-off-by: Ingo Molnar --- tools/perf/Documentation/perf-kmem.txt | 44 ++++++++++++++++++++++++++++++++++ tools/perf/command-list.txt | 1 + 2 files changed, 45 insertions(+) create mode 100644 tools/perf/Documentation/perf-kmem.txt (limited to 'tools/perf') diff --git a/tools/perf/Documentation/perf-kmem.txt b/tools/perf/Documentation/perf-kmem.txt new file mode 100644 index 00000000000..44b0ce35c28 --- /dev/null +++ b/tools/perf/Documentation/perf-kmem.txt @@ -0,0 +1,44 @@ +perf-kmem(1) +============== + +NAME +---- +perf-kmem - Tool to trace/measure kernel memory(slab) properties + +SYNOPSIS +-------- +[verse] +'perf kmem' {record} [] + +DESCRIPTION +----------- +There's two variants of perf kmem: + + 'perf kmem record ' to record the kmem events + of an arbitrary workload. + + 'perf kmem' to report kernel memory statistics. + +OPTIONS +------- +-i :: +--input=:: + Select the input file (default: perf.data) + +--stat=:: + Select per callsite or per allocation statistics + +-s :: +--sort=:: + Sort the output (default: frag,hit,bytes) + +-l :: +--line=:: + Print n lines only + +--raw-ip:: + Print raw ip instead of symbol + +SEE ALSO +-------- +linkperf:perf-record[1] diff --git a/tools/perf/command-list.txt b/tools/perf/command-list.txt index d3a6e18e4a5..02b09ea17a3 100644 --- a/tools/perf/command-list.txt +++ b/tools/perf/command-list.txt @@ -14,3 +14,4 @@ perf-timechart mainporcelain common perf-top mainporcelain common perf-trace mainporcelain common perf-probe mainporcelain common +perf-kmem mainporcelain common -- cgit v1.2.3-18-g5258 From c9c7ccaf3a2686ed3a44d69bb1f8b55eeead8a4e Mon Sep 17 00:00:00 2001 From: John Kacur Date: Tue, 24 Nov 2009 15:35:00 +0100 Subject: perf tools: Add perf.data to .gitignore Signed-off-by: John Kacur Cc: Frederic Weisbecker Cc: Peter Zijlstra Cc: acme@redhat.com LKML-Reference: <1259073301-11506-2-git-send-email-jkacur@redhat.com> Signed-off-by: Ingo Molnar --- tools/perf/.gitignore | 1 + 1 file changed, 1 insertion(+) (limited to 'tools/perf') diff --git a/tools/perf/.gitignore b/tools/perf/.gitignore index 0854f110bf7..fe08660ce0b 100644 --- a/tools/perf/.gitignore +++ b/tools/perf/.gitignore @@ -12,6 +12,7 @@ perf*.1 perf*.xml perf*.html common-cmds.h +perf.data tags TAGS cscope* -- cgit v1.2.3-18-g5258 From e74328d3a17ed75ffdf72b86f289965823a47240 Mon Sep 17 00:00:00 2001 From: John Kacur Date: Tue, 24 Nov 2009 15:35:01 +0100 Subject: perf tools: Use common process_event functions for annotate and report Prevent bit-rot in perf-annotate by using common functions where possible. Here we create process_events.[ch] to hold the common functions. Signed-off-by: John Kacur Cc: Frederic Weisbecker Cc: Peter Zijlstra Cc: acme@redhat.com LKML-Reference: <1259073301-11506-3-git-send-email-jkacur@redhat.com> Signed-off-by: Ingo Molnar --- tools/perf/Makefile | 2 ++ tools/perf/builtin-annotate.c | 56 ++--------------------------------- tools/perf/builtin-report.c | 63 +-------------------------------------- tools/perf/util/process_event.c | 53 +++++++++++++++++++++++++++++++++ tools/perf/util/process_event.h | 29 ++++++++++++++++++ tools/perf/util/process_events.c | 64 ++++++++++++++++++++++++++++++++++++++++ tools/perf/util/process_events.h | 35 ++++++++++++++++++++++ 7 files changed, 186 insertions(+), 116 deletions(-) create mode 100644 tools/perf/util/process_event.c create mode 100644 tools/perf/util/process_event.h create mode 100644 tools/perf/util/process_events.c create mode 100644 tools/perf/util/process_events.h (limited to 'tools/perf') diff --git a/tools/perf/Makefile b/tools/perf/Makefile index f1537a94a05..de37d492e10 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -369,6 +369,7 @@ LIB_H += util/sort.h LIB_H += util/hist.h LIB_H += util/thread.h LIB_H += util/data_map.h +LIB_H += util/process_events.h LIB_OBJS += util/abspath.o LIB_OBJS += util/alias.o @@ -411,6 +412,7 @@ LIB_OBJS += util/svghelper.o LIB_OBJS += util/sort.o LIB_OBJS += util/hist.o LIB_OBJS += util/data_map.o +LIB_OBJS += util/process_events.o BUILTIN_OBJS += builtin-annotate.o diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c index 6b13a1ecf1e..59b6123abec 100644 --- a/tools/perf/builtin-annotate.c +++ b/tools/perf/builtin-annotate.c @@ -24,6 +24,7 @@ #include "util/thread.h" #include "util/sort.h" #include "util/hist.h" +#include "util/process_events.h" static char const *input_name = "perf.data"; @@ -201,32 +202,6 @@ got_map: return 0; } -static int -process_mmap_event(event_t *event, unsigned long offset, unsigned long head) -{ - struct map *map = map__new(&event->mmap, NULL, 0); - struct thread *thread = threads__findnew(event->mmap.pid); - - dump_printf("%p [%p]: PERF_RECORD_MMAP %d: [%p(%p) @ %p]: %s\n", - (void *)(offset + head), - (void *)(long)(event->header.size), - event->mmap.pid, - (void *)(long)event->mmap.start, - (void *)(long)event->mmap.len, - (void *)(long)event->mmap.pgoff, - event->mmap.filename); - - if (thread == NULL || map == NULL) { - dump_printf("problem processing PERF_RECORD_MMAP, skipping event.\n"); - return 0; - } - - thread__insert_map(thread, map); - total_mmap++; - - return 0; -} - static int process_comm_event(event_t *event, unsigned long offset, unsigned long head) { @@ -247,33 +222,6 @@ process_comm_event(event_t *event, unsigned long offset, unsigned long head) return 0; } -static int -process_fork_event(event_t *event, unsigned long offset, unsigned long head) -{ - struct thread *thread = threads__findnew(event->fork.pid); - struct thread *parent = threads__findnew(event->fork.ppid); - - dump_printf("%p [%p]: PERF_RECORD_FORK: %d:%d\n", - (void *)(offset + head), - (void *)(long)(event->header.size), - event->fork.pid, event->fork.ppid); - - /* - * A thread clone will have the same PID for both - * parent and child. - */ - if (thread == parent) - return 0; - - if (!thread || !parent || thread__fork(thread, parent)) { - dump_printf("problem processing PERF_RECORD_FORK, skipping event.\n"); - return -1; - } - total_fork++; - - return 0; -} - static int process_event(event_t *event, unsigned long offset, unsigned long head) { @@ -288,7 +236,7 @@ process_event(event_t *event, unsigned long offset, unsigned long head) return process_comm_event(event, offset, head); case PERF_RECORD_FORK: - return process_fork_event(event, offset, head); + return process_task_event(event, offset, head); /* * We dont process them right now but they are fine: */ diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index fe474b7f8ad..1826be719b5 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -30,6 +30,7 @@ #include "util/thread.h" #include "util/sort.h" #include "util/hist.h" +#include "util/process_events.h" static char const *input_name = "perf.data"; @@ -54,9 +55,6 @@ static int exclude_other = 1; static char callchain_default_opt[] = "fractal,0.5"; const char *vmlinux_name; -static char *cwd; -static int cwdlen; - static struct perf_header *header; static u64 sample_type; @@ -750,33 +748,6 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) return 0; } -static int -process_mmap_event(event_t *event, unsigned long offset, unsigned long head) -{ - struct map *map = map__new(&event->mmap, cwd, cwdlen); - struct thread *thread = threads__findnew(event->mmap.pid); - - dump_printf("%p [%p]: PERF_RECORD_MMAP %d/%d: [%p(%p) @ %p]: %s\n", - (void *)(offset + head), - (void *)(long)(event->header.size), - event->mmap.pid, - event->mmap.tid, - (void *)(long)event->mmap.start, - (void *)(long)event->mmap.len, - (void *)(long)event->mmap.pgoff, - event->mmap.filename); - - if (thread == NULL || map == NULL) { - dump_printf("problem processing PERF_RECORD_MMAP, skipping event.\n"); - return 0; - } - - thread__insert_map(thread, map); - total_mmap++; - - return 0; -} - static int process_comm_event(event_t *event, unsigned long offset, unsigned long head) { @@ -797,38 +768,6 @@ process_comm_event(event_t *event, unsigned long offset, unsigned long head) return 0; } -static int -process_task_event(event_t *event, unsigned long offset, unsigned long head) -{ - struct thread *thread = threads__findnew(event->fork.pid); - struct thread *parent = threads__findnew(event->fork.ppid); - - dump_printf("%p [%p]: PERF_RECORD_%s: (%d:%d):(%d:%d)\n", - (void *)(offset + head), - (void *)(long)(event->header.size), - event->header.type == PERF_RECORD_FORK ? "FORK" : "EXIT", - event->fork.pid, event->fork.tid, - event->fork.ppid, event->fork.ptid); - - /* - * A thread clone will have the same PID for both - * parent and child. - */ - if (thread == parent) - return 0; - - if (event->header.type == PERF_RECORD_EXIT) - return 0; - - if (!thread || !parent || thread__fork(thread, parent)) { - dump_printf("problem processing PERF_RECORD_FORK, skipping event.\n"); - return -1; - } - total_fork++; - - return 0; -} - static int process_lost_event(event_t *event, unsigned long offset, unsigned long head) { diff --git a/tools/perf/util/process_event.c b/tools/perf/util/process_event.c new file mode 100644 index 00000000000..a970789581a --- /dev/null +++ b/tools/perf/util/process_event.c @@ -0,0 +1,53 @@ +#include "process_event.h" + +char *cwd; +int cwdlen; + +int +process_mmap_event(event_t *event, unsigned long offset, unsigned long head) +{ + struct map *map = map__new(&event->mmap, cwd, cwdlen); + struct thread *thread = threads__findnew(event->mmap.pid); + + dump_printf("%p [%p]: PERF_RECORD_MMAP %d/%d: [%p(%p) @ %p]: %s\n", + (void *)(offset + head), + (void *)(long)(event->header.size), + event->mmap.pid, + event->mmap.tid, + (void *)(long)event->mmap.start, + (void *)(long)event->mmap.len, + (void *)(long)event->mmap.pgoff, + event->mmap.filename); + + if (thread == NULL || map == NULL) { + dump_printf("problem processing PERF_RECORD_MMAP, skipping event.\n"); + return 0; + } + + thread__insert_map(thread, map); + total_mmap++; + + return 0; + +} + +int +process_comm_event(event_t *event, unsigned long offset, unsigned long head) +{ + struct thread *thread = threads__findnew(event->comm.pid); + + dump_printf("%p [%p]: PERF_RECORD_COMM: %s:%d\n", + (void *)(offset + head), + (void *)(long)(event->header.size), + event->comm.comm, event->comm.pid); + + if (thread == NULL || + thread__set_comm_adjust(thread, event->comm.comm)) { + dump_printf("problem processing PERF_RECORD_COMM, skipping event.\n"); + return -1; + } + total_comm++; + + return 0; +} + diff --git a/tools/perf/util/process_event.h b/tools/perf/util/process_event.h new file mode 100644 index 00000000000..6f68c69736c --- /dev/null +++ b/tools/perf/util/process_event.h @@ -0,0 +1,29 @@ +#ifndef __PROCESS_EVENT_H +#define __PROCESS_EVENT_H + +#include "../builtin.h" +#include "util.h" + +#include "color.h" +#include +#include "cache.h" +#include +#include "symbol.h" +#include "string.h" + +#include "../perf.h" +#include "debug.h" + +#include "parse-options.h" +#include "parse-events.h" + +#include "thread.h" +#include "sort.h" +#include "hist.h" + +extern char *cwd; +extern int cwdlen; +extern int process_mmap_event(event_t *, unsigned long, unsigned long); +extern int process_comm_event(event_t *, unsigned long , unsigned long); + +#endif /* __PROCESS_H */ diff --git a/tools/perf/util/process_events.c b/tools/perf/util/process_events.c new file mode 100644 index 00000000000..a9204363efd --- /dev/null +++ b/tools/perf/util/process_events.c @@ -0,0 +1,64 @@ +#include "process_events.h" + +char *cwd; +int cwdlen; + +int +process_mmap_event(event_t *event, unsigned long offset, unsigned long head) +{ + struct map *map = map__new(&event->mmap, cwd, cwdlen); + struct thread *thread = threads__findnew(event->mmap.pid); + + dump_printf("%p [%p]: PERF_RECORD_MMAP %d/%d: [%p(%p) @ %p]: %s\n", + (void *)(offset + head), + (void *)(long)(event->header.size), + event->mmap.pid, + event->mmap.tid, + (void *)(long)event->mmap.start, + (void *)(long)event->mmap.len, + (void *)(long)event->mmap.pgoff, + event->mmap.filename); + + if (thread == NULL || map == NULL) { + dump_printf("problem processing PERF_RECORD_MMAP, skipping event.\n"); + return 0; + } + + thread__insert_map(thread, map); + total_mmap++; + + return 0; +} + +int +process_task_event(event_t *event, unsigned long offset, unsigned long head) +{ + struct thread *thread = threads__findnew(event->fork.pid); + struct thread *parent = threads__findnew(event->fork.ppid); + + dump_printf("%p [%p]: PERF_RECORD_%s: (%d:%d):(%d:%d)\n", + (void *)(offset + head), + (void *)(long)(event->header.size), + event->header.type == PERF_RECORD_FORK ? "FORK" : "EXIT", + event->fork.pid, event->fork.tid, + event->fork.ppid, event->fork.ptid); + + /* + * A thread clone will have the same PID for both + * parent and child. + */ + if (thread == parent) + return 0; + + if (event->header.type == PERF_RECORD_EXIT) + return 0; + + if (!thread || !parent || thread__fork(thread, parent)) { + dump_printf("problem processing PERF_RECORD_FORK, skipping event.\n"); + return -1; + } + total_fork++; + + return 0; +} + diff --git a/tools/perf/util/process_events.h b/tools/perf/util/process_events.h new file mode 100644 index 00000000000..73d092f8328 --- /dev/null +++ b/tools/perf/util/process_events.h @@ -0,0 +1,35 @@ +#ifndef __PROCESS_EVENTS_H +#define __PROCESS_EVENTS_H + +#include "../builtin.h" + +#include "util.h" +#include "color.h" +#include +#include "cache.h" +#include +#include "symbol.h" +#include "string.h" +#include "callchain.h" +#include "strlist.h" +#include "values.h" + +#include "../perf.h" +#include "debug.h" +#include "header.h" + +#include "parse-options.h" +#include "parse-events.h" + +#include "data_map.h" +#include "thread.h" +#include "sort.h" +#include "hist.h" + +extern char *cwd; +extern int cwdlen; + +extern int process_mmap_event(event_t *, unsigned long , unsigned long); +extern int process_task_event(event_t *, unsigned long, unsigned long); + +#endif /* __PROCESS_EVENTS_H */ -- cgit v1.2.3-18-g5258 From 7cc017edb9459193d3b581155a14029e4bef0c49 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Tue, 24 Nov 2009 12:05:14 -0200 Subject: perf top: Always show the DSO column, even if its all the same MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Ingo found it confusing, and I agree with that, for 'perf report' its OK because it is static, but for a tool refreshing it the eventual switch from column to summary at the top may seem confusing. Suggested-by: Ingo Molnar Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1259071517-3242-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/builtin-top.c | 38 +++++++++++--------------------------- 1 file changed, 11 insertions(+), 27 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index 6a5de90e9b8..b9a321fd184 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -451,9 +451,8 @@ static void print_sym_table(void) struct sym_entry *syme, *n; struct rb_root tmp = RB_ROOT; struct rb_node *nd; - int sym_width = 0, dso_width = 0; + int sym_width = 0, dso_width = 0, max_dso_width; const int win_width = winsize.ws_col - 1; - struct dso *unique_dso = NULL, *first_dso = NULL; samples = userspace_samples = 0; @@ -539,11 +538,6 @@ static void print_sym_table(void) (int)syme->snap_count < count_filter) continue; - if (first_dso == NULL) - unique_dso = first_dso = syme->map->dso; - else if (syme->map->dso != first_dso) - unique_dso = NULL; - if (syme->map->dso->long_name_len > dso_width) dso_width = syme->map->dso->long_name_len; @@ -553,14 +547,10 @@ static void print_sym_table(void) printed = 0; - if (unique_dso) - printf("DSO: %s\n", unique_dso->long_name); - else { - int max_dso_width = winsize.ws_col - sym_width - 29; - if (dso_width > max_dso_width) - dso_width = max_dso_width; - putchar('\n'); - } + max_dso_width = winsize.ws_col - sym_width - 29; + if (dso_width > max_dso_width) + dso_width = max_dso_width; + putchar('\n'); if (nr_counters == 1) printf(" samples pcnt"); else @@ -568,17 +558,13 @@ static void print_sym_table(void) if (verbose) printf(" RIP "); - printf(" %-*.*s", sym_width, sym_width, "function"); - if (!unique_dso) - printf(" DSO"); - putchar('\n'); + printf(" %-*.*s DSO\n", sym_width, sym_width, "function"); printf(" %s _______ _____", nr_counters == 1 ? " " : "______"); if (verbose) printf(" ________________"); printf(" %-*.*s", sym_width, sym_width, graph_line); - if (!unique_dso) - printf(" %-*.*s", dso_width, dso_width, graph_line); + printf(" %-*.*s", dso_width, dso_width, graph_line); puts("\n"); for (nd = rb_first(&tmp); nd; nd = rb_next(nd)) { @@ -603,12 +589,10 @@ static void print_sym_table(void) if (verbose) printf(" %016llx", sym->start); printf(" %-*.*s", sym_width, sym_width, sym->name); - if (!unique_dso) - printf(" %-*.*s", dso_width, dso_width, - dso_width >= syme->map->dso->long_name_len ? - syme->map->dso->long_name : - syme->map->dso->short_name); - printf("\n"); + printf(" %-*.*s\n", dso_width, dso_width, + dso_width >= syme->map->dso->long_name_len ? + syme->map->dso->long_name : + syme->map->dso->short_name); } } -- cgit v1.2.3-18-g5258 From b32d133aec5dc882cf783a293f393bfb3f4379e1 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Tue, 24 Nov 2009 12:05:15 -0200 Subject: perf symbols: Simplify symbol machinery setup MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit And also express its configuration toggles via a struct. Now all one has to do is to call symbol__init(NULL) if the defaults are OK, or pass a struct symbol_conf pointer with the desired configuration. If a tool uses kernel_maps__find_symbol() to look at the kernel and modules mappings for a symbol but didn't call symbol__init() first, that will generate a one time warning too, alerting the subcommand developer that symbol__init() must be called. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1259071517-3242-2-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/builtin-annotate.c | 20 ++++++++++---------- tools/perf/builtin-kmem.c | 2 +- tools/perf/builtin-report.c | 15 ++++++++------- tools/perf/builtin-sched.c | 2 +- tools/perf/builtin-top.c | 24 +++++++++++++----------- tools/perf/builtin-trace.c | 2 +- tools/perf/util/data_map.c | 8 -------- tools/perf/util/data_map.h | 2 -- tools/perf/util/header.c | 6 ------ tools/perf/util/include/asm/bug.h | 22 ++++++++++++++++++++++ tools/perf/util/symbol.c | 31 +++++++++++++++++++++---------- tools/perf/util/symbol.h | 11 ++++++++--- 12 files changed, 85 insertions(+), 60 deletions(-) create mode 100644 tools/perf/util/include/asm/bug.h (limited to 'tools/perf') diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c index 59b6123abec..cd97c2b1cc3 100644 --- a/tools/perf/builtin-annotate.c +++ b/tools/perf/builtin-annotate.c @@ -34,11 +34,9 @@ static int input; static int full_paths; static int print_line; -static bool use_modules; static unsigned long page_size; static unsigned long mmap_window = 32; -const char *vmlinux_name; struct sym_hist { u64 sum; @@ -56,6 +54,11 @@ struct sym_priv { struct sym_ext *ext; }; +static struct symbol_conf symbol_conf = { + .priv_size = sizeof(struct sym_priv), + .try_vmlinux_path = true, +}; + static const char *sym_hist_filter; static int symbol_filter(struct map *map __used, struct symbol *sym) @@ -586,11 +589,6 @@ static int __cmd_annotate(void) exit(0); } - if (kernel_maps__init(vmlinux_name, true, use_modules) < 0) { - pr_err("failed to create kernel maps for symbol resolution\b"); - return -1; - } - remap: buf = (char *)mmap(NULL, page_size * mmap_window, PROT_READ, MAP_SHARED, input, offset); @@ -691,8 +689,9 @@ static const struct option options[] = { "be more verbose (show symbol address, etc)"), OPT_BOOLEAN('D', "dump-raw-trace", &dump_trace, "dump raw trace in ASCII"), - OPT_STRING('k', "vmlinux", &vmlinux_name, "file", "vmlinux pathname"), - OPT_BOOLEAN('m', "modules", &use_modules, + OPT_STRING('k', "vmlinux", &symbol_conf.vmlinux_name, + "file", "vmlinux pathname"), + OPT_BOOLEAN('m', "modules", &symbol_conf.use_modules, "load module symbols - WARNING: use only with -k and LIVE kernel"), OPT_BOOLEAN('l', "print-line", &print_line, "print matching source lines (may be slow)"), @@ -718,7 +717,8 @@ static void setup_sorting(void) int cmd_annotate(int argc, const char **argv, const char *prefix __used) { - symbol__init(sizeof(struct sym_priv)); + if (symbol__init(&symbol_conf) < 0) + return -1; page_size = getpagesize(); diff --git a/tools/perf/builtin-kmem.c b/tools/perf/builtin-kmem.c index 173d6db42ec..330dbc762f9 100644 --- a/tools/perf/builtin-kmem.c +++ b/tools/perf/builtin-kmem.c @@ -412,7 +412,7 @@ static int read_events(void) register_idle_thread(); register_perf_file_handler(&file_handler); - return mmap_dispatch_perf_file(&header, input_name, NULL, false, 0, 0, + return mmap_dispatch_perf_file(&header, input_name, 0, 0, &cwdlen, &cwd); } diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index 1826be719b5..0ee3d05a040 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -39,7 +39,6 @@ static char *dso_list_str, *comm_list_str, *sym_list_str, static struct strlist *dso_list, *comm_list, *sym_list; static int force; -static bool use_modules; static int full_paths; static int show_nr_samples; @@ -53,12 +52,13 @@ static char *pretty_printing_style = default_pretty_printing_style; static int exclude_other = 1; static char callchain_default_opt[] = "fractal,0.5"; -const char *vmlinux_name; static struct perf_header *header; static u64 sample_type; +struct symbol_conf symbol_conf; + static size_t callchain__fprintf_left_margin(FILE *fp, int left_margin) @@ -865,8 +865,7 @@ static int __cmd_report(void) register_perf_file_handler(&file_handler); - ret = mmap_dispatch_perf_file(&header, input_name, vmlinux_name, - !vmlinux_name, force, + ret = mmap_dispatch_perf_file(&header, input_name, force, full_paths, &cwdlen, &cwd); if (ret) return ret; @@ -963,9 +962,10 @@ static const struct option options[] = { "be more verbose (show symbol address, etc)"), OPT_BOOLEAN('D', "dump-raw-trace", &dump_trace, "dump raw trace in ASCII"), - OPT_STRING('k', "vmlinux", &vmlinux_name, "file", "vmlinux pathname"), + OPT_STRING('k', "vmlinux", &symbol_conf.vmlinux_name, + "file", "vmlinux pathname"), OPT_BOOLEAN('f', "force", &force, "don't complain, do it"), - OPT_BOOLEAN('m', "modules", &use_modules, + OPT_BOOLEAN('m', "modules", &symbol_conf.use_modules, "load module symbols - WARNING: use only with -k and LIVE kernel"), OPT_BOOLEAN('n', "show-nr-samples", &show_nr_samples, "Show a column with the number of samples"), @@ -1035,7 +1035,8 @@ static void setup_list(struct strlist **list, const char *list_str, int cmd_report(int argc, const char **argv, const char *prefix __used) { - symbol__init(0); + if (symbol__init(&symbol_conf) < 0) + return -1; argc = parse_options(argc, argv, options, report_usage, 0); diff --git a/tools/perf/builtin-sched.c b/tools/perf/builtin-sched.c index 260f57a72ee..dbf089b12de 100644 --- a/tools/perf/builtin-sched.c +++ b/tools/perf/builtin-sched.c @@ -1718,7 +1718,7 @@ static int read_events(void) register_idle_thread(); register_perf_file_handler(&file_handler); - return mmap_dispatch_perf_file(&header, input_name, NULL, false, 0, 0, + return mmap_dispatch_perf_file(&header, input_name, 0, 0, &cwdlen, &cwd); } diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index b9a321fd184..a21247543fc 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -79,7 +79,7 @@ static int dump_symtab = 0; static bool hide_kernel_symbols = false; static bool hide_user_symbols = false; static struct winsize winsize; -const char *vmlinux_name; +struct symbol_conf symbol_conf; /* * Source @@ -128,7 +128,7 @@ struct sym_entry { static inline struct symbol *sym_entry__symbol(struct sym_entry *self) { - return ((void *)self) + symbol__priv_size; + return ((void *)self) + symbol_conf.priv_size; } static void get_term_dimensions(struct winsize *ws) @@ -695,7 +695,7 @@ static void print_mapped_keys(void) fprintf(stdout, "\t[f] profile display filter (count). \t(%d)\n", count_filter); - if (vmlinux_name) { + if (symbol_conf.vmlinux_name) { fprintf(stdout, "\t[F] annotate display filter (percent). \t(%d%%)\n", sym_pcnt_filter); fprintf(stdout, "\t[s] annotate symbol. \t(%s)\n", name?: "NULL"); fprintf(stdout, "\t[S] stop annotation.\n"); @@ -732,7 +732,7 @@ static int key_mapped(int c) case 'F': case 's': case 'S': - return vmlinux_name ? 1 : 0; + return symbol_conf.vmlinux_name ? 1 : 0; default: break; } @@ -1261,7 +1261,8 @@ static const struct option options[] = { "system-wide collection from all CPUs"), OPT_INTEGER('C', "CPU", &profile_cpu, "CPU to profile on"), - OPT_STRING('k', "vmlinux", &vmlinux_name, "file", "vmlinux pathname"), + OPT_STRING('k', "vmlinux", &symbol_conf.vmlinux_name, + "file", "vmlinux pathname"), OPT_BOOLEAN('K', "hide_kernel_symbols", &hide_kernel_symbols, "hide kernel symbols"), OPT_INTEGER('m', "mmap-pages", &mmap_pages, @@ -1295,7 +1296,7 @@ static const struct option options[] = { int cmd_top(int argc, const char **argv, const char *prefix __used) { - int counter, err; + int counter; page_size = sysconf(_SC_PAGE_SIZE); @@ -1313,15 +1314,16 @@ int cmd_top(int argc, const char **argv, const char *prefix __used) if (!nr_counters) nr_counters = 1; - symbol__init(sizeof(struct sym_entry) + - (nr_counters + 1) * sizeof(unsigned long)); + symbol_conf.priv_size = (sizeof(struct sym_entry) + + (nr_counters + 1) * sizeof(unsigned long)); + if (symbol_conf.vmlinux_name == NULL) + symbol_conf.try_vmlinux_path = true; + if (symbol__init(&symbol_conf) < 0) + return -1; if (delay_secs < 1) delay_secs = 1; - err = kernel_maps__init(vmlinux_name, !vmlinux_name, true); - if (err < 0) - return err; parse_source(sym_filter_entry); /* diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-trace.c index b71198e5dc1..75972fd073d 100644 --- a/tools/perf/builtin-trace.c +++ b/tools/perf/builtin-trace.c @@ -131,7 +131,7 @@ static int __cmd_trace(void) register_idle_thread(); register_perf_file_handler(&file_handler); - return mmap_dispatch_perf_file(&header, input_name, NULL, false, + return mmap_dispatch_perf_file(&header, input_name, 0, 0, &cwdlen, &cwd); } diff --git a/tools/perf/util/data_map.c b/tools/perf/util/data_map.c index f318d19b256..b238462b898 100644 --- a/tools/perf/util/data_map.c +++ b/tools/perf/util/data_map.c @@ -101,8 +101,6 @@ out: int mmap_dispatch_perf_file(struct perf_header **pheader, const char *input_name, - const char *vmlinux_name, - bool try_vmlinux_path, int force, int full_paths, int *cwdlen, @@ -172,12 +170,6 @@ int mmap_dispatch_perf_file(struct perf_header **pheader, curr_handler->sample_type_check(sample_type) < 0) goto out_delete; - err = -ENOMEM; - if (kernel_maps__init(vmlinux_name, try_vmlinux_path, true) < 0) { - pr_err("failed to setup the kernel maps to resolve symbols\n"); - goto out_delete; - } - if (!full_paths) { if (getcwd(__cwd, sizeof(__cwd)) == NULL) { pr_err("failed to get the current directory\n"); diff --git a/tools/perf/util/data_map.h b/tools/perf/util/data_map.h index 3f0d21b3819..ae036ecd762 100644 --- a/tools/perf/util/data_map.h +++ b/tools/perf/util/data_map.h @@ -23,8 +23,6 @@ struct perf_file_handler { void register_perf_file_handler(struct perf_file_handler *handler); int mmap_dispatch_perf_file(struct perf_header **pheader, const char *input_name, - const char *vmlinux_name, - bool try_vmlinux_path, int force, int full_paths, int *cwdlen, diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index 1332f8ec04a..271a1600e6f 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -253,12 +253,6 @@ static int perf_header__adds_write(struct perf_header *self, int fd) buildid_sec = &feat_sec[idx++]; - /* - * Read the kernel buildid nad the list of loaded modules with - * its build_ids: - */ - kernel_maps__init(NULL, false, true); - /* Write build-ids */ buildid_sec->offset = lseek(fd, 0, SEEK_CUR); err = dsos__write_buildid_table(fd); diff --git a/tools/perf/util/include/asm/bug.h b/tools/perf/util/include/asm/bug.h new file mode 100644 index 00000000000..7fcc6810adc --- /dev/null +++ b/tools/perf/util/include/asm/bug.h @@ -0,0 +1,22 @@ +#ifndef _PERF_ASM_GENERIC_BUG_H +#define _PERF_ASM_GENERIC_BUG_H + +#define __WARN_printf(arg...) do { fprintf(stderr, arg); } while (0) + +#define WARN(condition, format...) ({ \ + int __ret_warn_on = !!(condition); \ + if (unlikely(__ret_warn_on)) \ + __WARN_printf(format); \ + unlikely(__ret_warn_on); \ +}) + +#define WARN_ONCE(condition, format...) ({ \ + static int __warned; \ + int __ret_warn_once = !!(condition); \ + \ + if (unlikely(__ret_warn_once)) \ + if (WARN(!__warned, format)) \ + __warned = 1; \ + unlikely(__ret_warn_once); \ +}) +#endif diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 44d81d5ae8c..c4ca974b36e 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -6,6 +6,7 @@ #include "debug.h" +#include #include #include #include @@ -37,6 +38,11 @@ unsigned int symbol__priv_size; static int vmlinux_path__nr_entries; static char **vmlinux_path; +static struct symbol_conf symbol_conf__defaults = { + .use_modules = true, + .try_vmlinux_path = true, +}; + static struct rb_root kernel_maps; static void dso__fixup_sym_end(struct dso *self) @@ -1166,7 +1172,9 @@ struct symbol *kernel_maps__find_symbol(u64 ip, struct map **mapp, if (map) { ip = map->map_ip(map, ip); return map__find_symbol(map, ip, filter); - } + } else + WARN_ONCE(RB_EMPTY_ROOT(&kernel_maps), + "Empty kernel_maps, was symbol__init() called?\n"); return NULL; } @@ -1485,9 +1493,9 @@ size_t dsos__fprintf_buildid(FILE *fp) return ret; } -static int kernel_maps__create_kernel_map(const char *vmlinux_name) +static int kernel_maps__create_kernel_map(const struct symbol_conf *conf) { - struct dso *kernel = dso__new(vmlinux_name ?: "[kernel.kallsyms]"); + struct dso *kernel = dso__new(conf->vmlinux_name ?: "[kernel.kallsyms]"); if (kernel == NULL) return -1; @@ -1577,18 +1585,21 @@ out_fail: return -1; } -int kernel_maps__init(const char *vmlinux_name, bool try_vmlinux_path, - bool use_modules) +static int kernel_maps__init(const struct symbol_conf *conf) { - if (try_vmlinux_path && vmlinux_path__init() < 0) + const struct symbol_conf *pconf = conf ?: &symbol_conf__defaults; + + symbol__priv_size = pconf->priv_size; + + if (pconf->try_vmlinux_path && vmlinux_path__init() < 0) return -1; - if (kernel_maps__create_kernel_map(vmlinux_name) < 0) { + if (kernel_maps__create_kernel_map(pconf) < 0) { vmlinux_path__exit(); return -1; } - if (use_modules && kernel_maps__create_module_maps() < 0) + if (pconf->use_modules && kernel_maps__create_module_maps() < 0) pr_debug("Failed to load list of modules in use, " "continuing...\n"); /* @@ -1598,8 +1609,8 @@ int kernel_maps__init(const char *vmlinux_name, bool try_vmlinux_path, return 0; } -void symbol__init(unsigned int priv_size) +int symbol__init(struct symbol_conf *conf) { elf_version(EV_CURRENT); - symbol__priv_size = priv_size; + return kernel_maps__init(conf); } diff --git a/tools/perf/util/symbol.h b/tools/perf/util/symbol.h index 8c4d026e067..5538691494a 100644 --- a/tools/perf/util/symbol.h +++ b/tools/perf/util/symbol.h @@ -49,6 +49,13 @@ struct symbol { char name[0]; }; +struct symbol_conf { + unsigned short priv_size; + bool try_vmlinux_path, + use_modules; + const char *vmlinux_name; +}; + extern unsigned int symbol__priv_size; static inline void *symbol__priv(struct symbol *self) @@ -93,11 +100,9 @@ int sysfs__read_build_id(const char *filename, void *bf, size_t size); bool dsos__read_build_ids(void); int build_id__sprintf(u8 *self, int len, char *bf); -int kernel_maps__init(const char *vmlinux_name, bool try_vmlinux_path, - bool use_modules); size_t kernel_maps__fprintf(FILE *fp); -void symbol__init(unsigned int priv_size); +int symbol__init(struct symbol_conf *conf); extern struct list_head dsos; extern struct map *kernel_map; -- cgit v1.2.3-18-g5258 From 364794845cbc49e638b83d7ef739524291e1e961 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Tue, 24 Nov 2009 12:05:16 -0200 Subject: perf tools: Introduce zalloc() for the common calloc(1, N) case MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This way we type less characters and it looks more like the kzalloc kernel counterpart. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1259071517-3242-3-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/bench/mem-memcpy.c | 4 ++-- tools/perf/builtin-help.c | 4 ++-- tools/perf/builtin-probe.c | 4 ++-- tools/perf/builtin-sched.c | 14 +++++--------- tools/perf/builtin-top.c | 2 +- tools/perf/util/header.c | 2 +- tools/perf/util/parse-events.c | 2 +- tools/perf/util/symbol.c | 11 +++++------ tools/perf/util/thread.c | 2 +- tools/perf/util/util.h | 5 +++++ 10 files changed, 25 insertions(+), 25 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/bench/mem-memcpy.c b/tools/perf/bench/mem-memcpy.c index 5165fd1d8d2..89773178e89 100644 --- a/tools/perf/bench/mem-memcpy.c +++ b/tools/perf/bench/mem-memcpy.c @@ -127,11 +127,11 @@ int bench_mem_memcpy(int argc, const char **argv, return 1; } - dst = calloc(length, sizeof(char)); + dst = zalloc(length); if (!dst) die("memory allocation failed - maybe length is too large?\n"); - src = calloc(length, sizeof(char)); + src = zalloc(length); if (!src) die("memory allocation failed - maybe length is too large?\n"); diff --git a/tools/perf/builtin-help.c b/tools/perf/builtin-help.c index 768f9c82631..9f810b17c25 100644 --- a/tools/perf/builtin-help.c +++ b/tools/perf/builtin-help.c @@ -179,7 +179,7 @@ static void add_man_viewer(const char *name) while (*p) p = &((*p)->next); - *p = calloc(1, (sizeof(**p) + len + 1)); + *p = zalloc(sizeof(**p) + len + 1); strncpy((*p)->name, name, len); } @@ -194,7 +194,7 @@ static void do_add_man_viewer_info(const char *name, size_t len, const char *value) { - struct man_viewer_info_list *new = calloc(1, sizeof(*new) + len + 1); + struct man_viewer_info_list *new = zalloc(sizeof(*new) + len + 1); strncpy(new->name, name, len); new->info = strdup(value); diff --git a/tools/perf/builtin-probe.c b/tools/perf/builtin-probe.c index d78a3d94549..a2f6daf01ec 100644 --- a/tools/perf/builtin-probe.c +++ b/tools/perf/builtin-probe.c @@ -309,9 +309,9 @@ static int synthesize_probe_event(struct probe_point *pp) { char *buf; int i, len, ret; - pp->probes[0] = buf = (char *)calloc(MAX_CMDLEN, sizeof(char)); + pp->probes[0] = buf = zalloc(MAX_CMDLEN); if (!buf) - die("Failed to allocate memory by calloc."); + die("Failed to allocate memory by zalloc."); ret = snprintf(buf, MAX_CMDLEN, "%s+%d", pp->function, pp->offset); if (ret <= 0 || ret >= MAX_CMDLEN) goto error; diff --git a/tools/perf/builtin-sched.c b/tools/perf/builtin-sched.c index dbf089b12de..19eb708a706 100644 --- a/tools/perf/builtin-sched.c +++ b/tools/perf/builtin-sched.c @@ -225,7 +225,7 @@ static void calibrate_sleep_measurement_overhead(void) static struct sched_atom * get_new_event(struct task_desc *task, u64 timestamp) { - struct sched_atom *event = calloc(1, sizeof(*event)); + struct sched_atom *event = zalloc(sizeof(*event)); unsigned long idx = task->nr_events; size_t size; @@ -293,7 +293,7 @@ add_sched_event_wakeup(struct task_desc *task, u64 timestamp, return; } - wakee_event->wait_sem = calloc(1, sizeof(*wakee_event->wait_sem)); + wakee_event->wait_sem = zalloc(sizeof(*wakee_event->wait_sem)); sem_init(wakee_event->wait_sem, 0, 0); wakee_event->specific_wait = 1; event->wait_sem = wakee_event->wait_sem; @@ -323,7 +323,7 @@ static struct task_desc *register_pid(unsigned long pid, const char *comm) if (task) return task; - task = calloc(1, sizeof(*task)); + task = zalloc(sizeof(*task)); task->pid = pid; task->nr = nr_tasks; strcpy(task->comm, comm); @@ -962,9 +962,7 @@ __thread_latency_insert(struct rb_root *root, struct work_atoms *data, static void thread_atoms_insert(struct thread *thread) { - struct work_atoms *atoms; - - atoms = calloc(sizeof(*atoms), 1); + struct work_atoms *atoms = zalloc(sizeof(*atoms)); if (!atoms) die("No memory"); @@ -996,9 +994,7 @@ add_sched_out_event(struct work_atoms *atoms, char run_state, u64 timestamp) { - struct work_atom *atom; - - atom = calloc(sizeof(*atom), 1); + struct work_atom *atom = zalloc(sizeof(*atom)); if (!atom) die("Non memory"); diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index a21247543fc..4c8653a86aa 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -181,7 +181,7 @@ static void parse_source(struct sym_entry *syme) return; if (syme->src == NULL) { - syme->src = calloc(1, sizeof(*source)); + syme->src = zalloc(sizeof(*source)); if (syme->src == NULL) return; pthread_mutex_init(&syme->src->lock, NULL); diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index 271a1600e6f..4b586569bb0 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -63,7 +63,7 @@ int perf_header_attr__add_id(struct perf_header_attr *self, u64 id) */ struct perf_header *perf_header__new(void) { - struct perf_header *self = calloc(sizeof(*self), 1); + struct perf_header *self = zalloc(sizeof(*self)); if (self != NULL) { self->size = 1; diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c index 07002746927..9e5dbd66d34 100644 --- a/tools/perf/util/parse-events.c +++ b/tools/perf/util/parse-events.c @@ -197,7 +197,7 @@ struct tracepoint_path *tracepoint_id_to_path(u64 config) if (id == config) { closedir(evt_dir); closedir(sys_dir); - path = calloc(1, sizeof(path)); + path = zalloc(sizeof(path)); path->system = malloc(MAX_EVENT_LENGTH); if (!path->system) { free(path); diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index c4ca974b36e..8db85b4f553 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -94,15 +94,14 @@ static void kernel_maps__fixup_end(void) static struct symbol *symbol__new(u64 start, u64 len, const char *name) { size_t namelen = strlen(name) + 1; - struct symbol *self = calloc(1, (symbol__priv_size + - sizeof(*self) + namelen)); - if (!self) + struct symbol *self = zalloc(symbol__priv_size + + sizeof(*self) + namelen); + if (self == NULL) return NULL; - if (symbol__priv_size) { - memset(self, 0, symbol__priv_size); + if (symbol__priv_size) self = ((void *)self) + symbol__priv_size; - } + self->start = start; self->end = len ? start + len - 1 : start; diff --git a/tools/perf/util/thread.c b/tools/perf/util/thread.c index 0f6d78c9863..1796625f778 100644 --- a/tools/perf/util/thread.c +++ b/tools/perf/util/thread.c @@ -11,7 +11,7 @@ static struct thread *last_match; static struct thread *thread__new(pid_t pid) { - struct thread *self = calloc(1, sizeof(*self)); + struct thread *self = zalloc(sizeof(*self)); if (self != NULL) { self->pid = pid; diff --git a/tools/perf/util/util.h b/tools/perf/util/util.h index e1c623e0c99..30c5517f2f9 100644 --- a/tools/perf/util/util.h +++ b/tools/perf/util/util.h @@ -302,6 +302,11 @@ extern int xdup(int fd); extern FILE *xfdopen(int fd, const char *mode); extern int xmkstemp(char *template); +static inline void *zalloc(size_t size) +{ + return calloc(1, size); +} + static inline size_t xsize_t(off_t len) { return (size_t)len; -- cgit v1.2.3-18-g5258 From 727dad10c17cbaade3cb6a56bd4863a4630f4d13 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Tue, 24 Nov 2009 12:05:17 -0200 Subject: perf tools: Remove unused wrapper routines MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit And also make xrealloc and xmalloc weak symbols so that we don't have this problem: /usr/lib/gcc/x86_64-redhat-linux/4.4.1/../../../../lib64/libiberty.a(xmalloc.o): In function `xrealloc': (.text+0xc0): multiple definition of `xrealloc' libperf.a(wrapper.o):/home/acme_unencrypted/git/linux-2.6-tip/tools/perf/util/wrapper.c:67: first defined here collect2: ld returned 1 exit status Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1259071517-3242-4-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/util/util.h | 11 ++------- tools/perf/util/wrapper.c | 61 ++--------------------------------------------- 2 files changed, 4 insertions(+), 68 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/util.h b/tools/perf/util/util.h index 30c5517f2f9..c673d882588 100644 --- a/tools/perf/util/util.h +++ b/tools/perf/util/util.h @@ -290,17 +290,10 @@ static inline char *gitstrchrnul(const char *s, int c) * Wrappers: */ extern char *xstrdup(const char *str); -extern void *xmalloc(size_t size); +extern void *xmalloc(size_t size) __attribute__((weak)); extern void *xmemdupz(const void *data, size_t len); extern char *xstrndup(const char *str, size_t len); -extern void *xrealloc(void *ptr, size_t size); -extern void *xcalloc(size_t nmemb, size_t size); -extern void *xmmap(void *start, size_t length, int prot, int flags, int fd, off_t offset); -extern ssize_t xread(int fd, void *buf, size_t len); -extern ssize_t xwrite(int fd, const void *buf, size_t len); -extern int xdup(int fd); -extern FILE *xfdopen(int fd, const char *mode); -extern int xmkstemp(char *template); +extern void *xrealloc(void *ptr, size_t size) __attribute__((weak)); static inline void *zalloc(size_t size) { diff --git a/tools/perf/util/wrapper.c b/tools/perf/util/wrapper.c index 4574ac28396..bf44ca85d23 100644 --- a/tools/perf/util/wrapper.c +++ b/tools/perf/util/wrapper.c @@ -79,43 +79,12 @@ void *xrealloc(void *ptr, size_t size) return ret; } -void *xcalloc(size_t nmemb, size_t size) -{ - void *ret = calloc(nmemb, size); - if (!ret && (!nmemb || !size)) - ret = calloc(1, 1); - if (!ret) { - release_pack_memory(nmemb * size, -1); - ret = calloc(nmemb, size); - if (!ret && (!nmemb || !size)) - ret = calloc(1, 1); - if (!ret) - die("Out of memory, calloc failed"); - } - return ret; -} - -void *xmmap(void *start, size_t length, - int prot, int flags, int fd, off_t offset) -{ - void *ret = mmap(start, length, prot, flags, fd, offset); - if (ret == MAP_FAILED) { - if (!length) - return NULL; - release_pack_memory(length, fd); - ret = mmap(start, length, prot, flags, fd, offset); - if (ret == MAP_FAILED) - die("Out of memory? mmap failed: %s", strerror(errno)); - } - return ret; -} - /* * xread() is the same a read(), but it automatically restarts read() * operations with a recoverable error (EAGAIN and EINTR). xread() * DOES NOT GUARANTEE that "len" bytes is read even if the data is available. */ -ssize_t xread(int fd, void *buf, size_t len) +static ssize_t xread(int fd, void *buf, size_t len) { ssize_t nr; while (1) { @@ -131,7 +100,7 @@ ssize_t xread(int fd, void *buf, size_t len) * operations with a recoverable error (EAGAIN and EINTR). xwrite() DOES NOT * GUARANTEE that "len" bytes is written even if the operation is successful. */ -ssize_t xwrite(int fd, const void *buf, size_t len) +static ssize_t xwrite(int fd, const void *buf, size_t len) { ssize_t nr; while (1) { @@ -179,29 +148,3 @@ ssize_t write_in_full(int fd, const void *buf, size_t count) return total; } - -int xdup(int fd) -{ - int ret = dup(fd); - if (ret < 0) - die("dup failed: %s", strerror(errno)); - return ret; -} - -FILE *xfdopen(int fd, const char *mode) -{ - FILE *stream = fdopen(fd, mode); - if (stream == NULL) - die("Out of memory? fdopen failed: %s", strerror(errno)); - return stream; -} - -int xmkstemp(char *template) -{ - int fd; - - fd = mkstemp(template); - if (fd < 0) - die("Unable to create temporary file: %s", strerror(errno)); - return fd; -} -- cgit v1.2.3-18-g5258 From fcf1203a919c3a3d212c0ed01f5240fd592bf5ae Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Tue, 24 Nov 2009 13:01:52 -0200 Subject: perf symbols: Rename find_symbol routines to find_function MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Paving the way for supporting variable in adition to function symbols. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1259074912-5924-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/builtin-annotate.c | 4 +-- tools/perf/builtin-kmem.c | 3 +- tools/perf/builtin-report.c | 8 ++--- tools/perf/builtin-top.c | 4 +-- tools/perf/util/event.h | 7 +++-- tools/perf/util/map.c | 14 ++++----- tools/perf/util/symbol.c | 68 +++++++++++++++++++++++-------------------- tools/perf/util/symbol.h | 6 ++-- tools/perf/util/thread.h | 4 +-- 9 files changed, 62 insertions(+), 56 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c index cd97c2b1cc3..18ac5eaefc3 100644 --- a/tools/perf/builtin-annotate.c +++ b/tools/perf/builtin-annotate.c @@ -162,7 +162,7 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) if (event->header.misc & PERF_RECORD_MISC_KERNEL) { level = 'k'; - sym = kernel_maps__find_symbol(ip, &map, symbol_filter); + sym = kernel_maps__find_function(ip, &map, symbol_filter); dump_printf(" ...... dso: %s\n", map ? map->dso->long_name : ""); } else if (event->header.misc & PERF_RECORD_MISC_USER) { @@ -171,7 +171,7 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) if (map != NULL) { got_map: ip = map->map_ip(map, ip); - sym = map__find_symbol(map, ip, symbol_filter); + sym = map__find_function(map, ip, symbol_filter); } else { /* * If this is outside of all known maps, diff --git a/tools/perf/builtin-kmem.c b/tools/perf/builtin-kmem.c index 330dbc762f9..35722fafc4d 100644 --- a/tools/perf/builtin-kmem.c +++ b/tools/perf/builtin-kmem.c @@ -445,8 +445,7 @@ static void __print_result(struct rb_root *root, int n_lines, int is_caller) if (is_caller) { addr = data->call_site; if (!raw_ip) - sym = kernel_maps__find_symbol(addr, - NULL, NULL); + sym = kernel_maps__find_function(addr, NULL, NULL); } else addr = data->ptr; diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index 0ee3d05a040..e4b1004e76e 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -448,14 +448,14 @@ got_map: * trick of looking in the whole kernel symbol list. */ if ((long long)ip < 0) - return kernel_maps__find_symbol(ip, mapp, NULL); + return kernel_maps__find_function(ip, mapp, NULL); } dump_printf(" ...... dso: %s\n", map ? map->dso->long_name : ""); dump_printf(" ...... map: %Lx -> %Lx\n", *ipp, ip); *ipp = ip; - return map ? map__find_symbol(map, ip, NULL) : NULL; + return map ? map__find_function(map, ip, NULL) : NULL; } static int call__match(struct symbol *sym) @@ -495,7 +495,7 @@ static struct symbol **resolve_callchain(struct thread *thread, case PERF_CONTEXT_HV: break; case PERF_CONTEXT_KERNEL: - sym = kernel_maps__find_symbol(ip, NULL, NULL); + sym = kernel_maps__find_function(ip, NULL, NULL); break; default: sym = resolve_symbol(thread, NULL, &ip); @@ -715,7 +715,7 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) if (cpumode == PERF_RECORD_MISC_KERNEL) { level = 'k'; - sym = kernel_maps__find_symbol(ip, &map, NULL); + sym = kernel_maps__find_function(ip, &map, NULL); dump_printf(" ...... dso: %s\n", map ? map->dso->long_name : ""); } else if (cpumode == PERF_RECORD_MISC_USER) { diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index 4c8653a86aa..ded6cf65ad9 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -948,7 +948,7 @@ static void event__process_sample(const event_t *self, int counter) map = thread__find_map(thread, ip); if (map != NULL) { ip = map->map_ip(map, ip); - sym = map__find_symbol(map, ip, symbol_filter); + sym = map__find_function(map, ip, symbol_filter); if (sym == NULL) return; userspace_samples++; @@ -968,7 +968,7 @@ static void event__process_sample(const event_t *self, int counter) if (hide_kernel_symbols) return; - sym = kernel_maps__find_symbol(ip, &map, symbol_filter); + sym = kernel_maps__find_function(ip, &map, symbol_filter); if (sym == NULL) return; break; diff --git a/tools/perf/util/event.h b/tools/perf/util/event.h index f1e39261265..882a9531db9 100644 --- a/tools/perf/util/event.h +++ b/tools/perf/util/event.h @@ -119,9 +119,10 @@ void map__delete(struct map *self); struct map *map__clone(struct map *self); int map__overlap(struct map *l, struct map *r); size_t map__fprintf(struct map *self, FILE *fp); -struct symbol *map__find_symbol(struct map *self, u64 ip, symbol_filter_t filter); -void map__fixup_start(struct map *self); -void map__fixup_end(struct map *self); +struct symbol *map__find_function(struct map *self, u64 ip, + symbol_filter_t filter); +void map__fixup_start(struct map *self, struct rb_root *symbols); +void map__fixup_end(struct map *self, struct rb_root *symbols); int event__synthesize_thread(pid_t pid, int (*process)(event_t *event)); void event__synthesize_threads(int (*process)(event_t *event)); diff --git a/tools/perf/util/map.c b/tools/perf/util/map.c index 09412321a80..41c5c4a2001 100644 --- a/tools/perf/util/map.c +++ b/tools/perf/util/map.c @@ -80,18 +80,18 @@ void map__delete(struct map *self) free(self); } -void map__fixup_start(struct map *self) +void map__fixup_start(struct map *self, struct rb_root *symbols) { - struct rb_node *nd = rb_first(&self->dso->syms); + struct rb_node *nd = rb_first(symbols); if (nd != NULL) { struct symbol *sym = rb_entry(nd, struct symbol, rb_node); self->start = sym->start; } } -void map__fixup_end(struct map *self) +void map__fixup_end(struct map *self, struct rb_root *symbols) { - struct rb_node *nd = rb_last(&self->dso->syms); + struct rb_node *nd = rb_last(symbols); if (nd != NULL) { struct symbol *sym = rb_entry(nd, struct symbol, rb_node); self->end = sym->end; @@ -100,8 +100,8 @@ void map__fixup_end(struct map *self) #define DSO__DELETED "(deleted)" -struct symbol * -map__find_symbol(struct map *self, u64 ip, symbol_filter_t filter) +struct symbol *map__find_function(struct map *self, u64 ip, + symbol_filter_t filter) { if (!self->dso->loaded) { int nr = dso__load(self->dso, self, filter); @@ -136,7 +136,7 @@ map__find_symbol(struct map *self, u64 ip, symbol_filter_t filter) } } - return self->dso->find_symbol(self->dso, ip); + return self->dso->find_function(self->dso, ip); } struct map *map__clone(struct map *self) diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 8db85b4f553..4ed379b915f 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -45,9 +45,9 @@ static struct symbol_conf symbol_conf__defaults = { static struct rb_root kernel_maps; -static void dso__fixup_sym_end(struct dso *self) +static void symbols__fixup_end(struct rb_root *self) { - struct rb_node *nd, *prevnd = rb_first(&self->syms); + struct rb_node *nd, *prevnd = rb_first(self); struct symbol *curr, *prev; if (prevnd == NULL) @@ -144,8 +144,8 @@ struct dso *dso__new(const char *name) strcpy(self->name, name); dso__set_long_name(self, self->name); self->short_name = self->name; - self->syms = RB_ROOT; - self->find_symbol = dso__find_symbol; + self->functions = RB_ROOT; + self->find_function = dso__find_function; self->slen_calculated = 0; self->origin = DSO__ORIG_NOT_FOUND; self->loaded = 0; @@ -155,22 +155,22 @@ struct dso *dso__new(const char *name) return self; } -static void dso__delete_symbols(struct dso *self) +static void symbols__delete(struct rb_root *self) { struct symbol *pos; - struct rb_node *next = rb_first(&self->syms); + struct rb_node *next = rb_first(self); while (next) { pos = rb_entry(next, struct symbol, rb_node); next = rb_next(&pos->rb_node); - rb_erase(&pos->rb_node, &self->syms); + rb_erase(&pos->rb_node, self); symbol__delete(pos); } } void dso__delete(struct dso *self) { - dso__delete_symbols(self); + symbols__delete(&self->functions); if (self->long_name != self->name) free(self->long_name); free(self); @@ -182,9 +182,9 @@ void dso__set_build_id(struct dso *self, void *build_id) self->has_build_id = 1; } -static void dso__insert_symbol(struct dso *self, struct symbol *sym) +static void symbols__insert(struct rb_root *self, struct symbol *sym) { - struct rb_node **p = &self->syms.rb_node; + struct rb_node **p = &self->rb_node; struct rb_node *parent = NULL; const u64 ip = sym->start; struct symbol *s; @@ -198,17 +198,17 @@ static void dso__insert_symbol(struct dso *self, struct symbol *sym) p = &(*p)->rb_right; } rb_link_node(&sym->rb_node, parent, p); - rb_insert_color(&sym->rb_node, &self->syms); + rb_insert_color(&sym->rb_node, self); } -struct symbol *dso__find_symbol(struct dso *self, u64 ip) +static struct symbol *symbols__find(struct rb_root *self, u64 ip) { struct rb_node *n; if (self == NULL) return NULL; - n = self->syms.rb_node; + n = self->rb_node; while (n) { struct symbol *s = rb_entry(n, struct symbol, rb_node); @@ -224,6 +224,11 @@ struct symbol *dso__find_symbol(struct dso *self, u64 ip) return NULL; } +struct symbol *dso__find_function(struct dso *self, u64 ip) +{ + return symbols__find(&self->functions, ip); +} + int build_id__sprintf(u8 *self, int len, char *bf) { char *bid = bf; @@ -253,9 +258,9 @@ size_t dso__fprintf(struct dso *self, FILE *fp) size_t ret = fprintf(fp, "dso: %s (", self->short_name); ret += dso__fprintf_buildid(self, fp); - ret += fprintf(fp, ")\n"); + ret += fprintf(fp, ")\nFunctions:\n"); - for (nd = rb_first(&self->syms); nd; nd = rb_next(nd)) { + for (nd = rb_first(&self->functions); nd; nd = rb_next(nd)) { struct symbol *pos = rb_entry(nd, struct symbol, rb_node); ret += symbol__fprintf(pos, fp); } @@ -320,7 +325,7 @@ static int kernel_maps__load_all_kallsyms(void) * kernel_maps__split_kallsyms, when we have split the * maps per module */ - dso__insert_symbol(kernel_map->dso, sym); + symbols__insert(&kernel_map->dso->functions, sym); } free(line); @@ -344,7 +349,7 @@ static int kernel_maps__split_kallsyms(symbol_filter_t filter) struct map *map = kernel_map; struct symbol *pos; int count = 0; - struct rb_node *next = rb_first(&kernel_map->dso->syms); + struct rb_node *next = rb_first(&kernel_map->dso->functions); int kernel_range = 0; while (next) { @@ -394,12 +399,13 @@ static int kernel_maps__split_kallsyms(symbol_filter_t filter) } if (filter && filter(map, pos)) { - rb_erase(&pos->rb_node, &kernel_map->dso->syms); + rb_erase(&pos->rb_node, &kernel_map->dso->functions); symbol__delete(pos); } else { if (map != kernel_map) { - rb_erase(&pos->rb_node, &kernel_map->dso->syms); - dso__insert_symbol(map->dso, pos); + rb_erase(&pos->rb_node, + &kernel_map->dso->functions); + symbols__insert(&map->dso->functions, pos); } count++; } @@ -414,7 +420,7 @@ static int kernel_maps__load_kallsyms(symbol_filter_t filter) if (kernel_maps__load_all_kallsyms()) return -1; - dso__fixup_sym_end(kernel_map->dso); + symbols__fixup_end(&kernel_map->dso->functions); kernel_map->dso->origin = DSO__ORIG_KERNEL; return kernel_maps__split_kallsyms(filter); @@ -485,7 +491,7 @@ static int dso__load_perf_map(struct dso *self, struct map *map, if (filter && filter(map, sym)) symbol__delete(sym); else { - dso__insert_symbol(self, sym); + symbols__insert(&self->functions, sym); nr_syms++; } } @@ -683,7 +689,7 @@ static int dso__synthesize_plt_symbols(struct dso *self, struct map *map, if (filter && filter(map, f)) symbol__delete(f); else { - dso__insert_symbol(self, f); + symbols__insert(&self->functions, f); ++nr; } } @@ -705,7 +711,7 @@ static int dso__synthesize_plt_symbols(struct dso *self, struct map *map, if (filter && filter(map, f)) symbol__delete(f); else { - dso__insert_symbol(self, f); + symbols__insert(&self->functions, f); ++nr; } } @@ -879,7 +885,7 @@ new_symbol: if (filter && filter(curr_map, f)) symbol__delete(f); else { - dso__insert_symbol(curr_dso, f); + symbols__insert(&curr_dso->functions, f); nr++; } } @@ -888,7 +894,7 @@ new_symbol: * For misannotated, zeroed, ASM function sizes. */ if (nr > 0) - dso__fixup_sym_end(self); + symbols__fixup_end(&self->functions); err = nr; out_elf_end: elf_end(elf); @@ -1160,8 +1166,8 @@ static void kernel_maps__insert(struct map *map) maps__insert(&kernel_maps, map); } -struct symbol *kernel_maps__find_symbol(u64 ip, struct map **mapp, - symbol_filter_t filter) +struct symbol *kernel_maps__find_function(u64 ip, struct map **mapp, + symbol_filter_t filter) { struct map *map = maps__find(&kernel_maps, ip); @@ -1170,7 +1176,7 @@ struct symbol *kernel_maps__find_symbol(u64 ip, struct map **mapp, if (map) { ip = map->map_ip(map, ip); - return map__find_symbol(map, ip, filter); + return map__find_function(map, ip, filter); } else WARN_ONCE(RB_EMPTY_ROOT(&kernel_maps), "Empty kernel_maps, was symbol__init() called?\n"); @@ -1432,8 +1438,8 @@ do_kallsyms: if (err > 0) { out_fixup: - map__fixup_start(map); - map__fixup_end(map); + map__fixup_start(map, &map->dso->functions); + map__fixup_end(map, &map->dso->functions); } return err; diff --git a/tools/perf/util/symbol.h b/tools/perf/util/symbol.h index 5538691494a..65846d0c5df 100644 --- a/tools/perf/util/symbol.h +++ b/tools/perf/util/symbol.h @@ -65,8 +65,8 @@ static inline void *symbol__priv(struct symbol *self) struct dso { struct list_head node; - struct rb_root syms; - struct symbol *(*find_symbol)(struct dso *, u64 ip); + struct rb_root functions; + struct symbol *(*find_function)(struct dso *, u64 ip); u8 adjust_symbols:1; u8 slen_calculated:1; u8 loaded:1; @@ -83,7 +83,7 @@ struct dso { struct dso *dso__new(const char *name); void dso__delete(struct dso *self); -struct symbol *dso__find_symbol(struct dso *self, u64 ip); +struct symbol *dso__find_function(struct dso *self, u64 ip); struct dso *dsos__findnew(const char *name); int dso__load(struct dso *self, struct map *map, symbol_filter_t filter); diff --git a/tools/perf/util/thread.h b/tools/perf/util/thread.h index e4b8d437725..74cba6487ed 100644 --- a/tools/perf/util/thread.h +++ b/tools/perf/util/thread.h @@ -26,8 +26,8 @@ size_t threads__fprintf(FILE *fp); void maps__insert(struct rb_root *maps, struct map *map); struct map *maps__find(struct rb_root *maps, u64 ip); -struct symbol *kernel_maps__find_symbol(const u64 ip, struct map **mapp, - symbol_filter_t filter); +struct symbol *kernel_maps__find_function(const u64 ip, struct map **mapp, + symbol_filter_t filter); struct map *kernel_maps__find_by_dso_name(const char *name); static inline struct map *thread__find_map(struct thread *self, u64 ip) -- cgit v1.2.3-18-g5258 From 0d0bea5ea4a0e91feff22ac5e32e14ff3a682247 Mon Sep 17 00:00:00 2001 From: Tom Zanussi Date: Wed, 25 Nov 2009 01:14:58 -0600 Subject: perf tools: Add 'signed' flag setting back into trace-event-parse.c Commit 13999e59343b042b0807be2df6ae5895d29782a0 (perf tools: Handle the case with and without the "signed" trace field) removed code to set the FIELD_IS_SIGNED flag that was originally added by commit 26a50744b21fff65bd754874072857bee8967f4d (tracing/events: Add 'signed' field to format files). This adds it back. Signed-off-by: Tom Zanussi Cc: Steven Rostedt Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker LKML-Reference: <1259133299-23594-2-git-send-email-tzanussi@gmail.com> Signed-off-by: Ingo Molnar --- tools/perf/util/trace-event-parse.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'tools/perf') diff --git a/tools/perf/util/trace-event-parse.c b/tools/perf/util/trace-event-parse.c index eae56050308..7021dc1b0ca 100644 --- a/tools/perf/util/trace-event-parse.c +++ b/tools/perf/util/trace-event-parse.c @@ -941,7 +941,8 @@ static int event_read_fields(struct event *event, struct format_field **fields) if (read_expect_type(EVENT_ITEM, &token)) goto fail; - /* add signed type */ + if (strtoul(token, NULL, 0)) + field->flags |= FIELD_IS_SIGNED; free_token(token); if (read_expected(EVENT_OP, ";") < 0) -- cgit v1.2.3-18-g5258 From 3f5ee186f615a720fe78eb33662ae4da57a1eee3 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 27 Nov 2009 16:29:12 -0200 Subject: perf symbols: Avoid annoying message about loading symbols MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This should be properly fixed when we remove the XXX comment in 'perf report', function resolve_symbol. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1259346563-12568-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/util/symbol.c | 1 + 1 file changed, 1 insertion(+) (limited to 'tools/perf') diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 4ed379b915f..8aab89b2248 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -1516,6 +1516,7 @@ static int kernel_maps__create_kernel_map(const struct symbol_conf *conf) vdso = dso__new("[vdso]"); if (vdso == NULL) goto out_delete_kernel_map; + vdso->loaded = 1; if (sysfs__read_build_id("/sys/kernel/notes", kernel->build_id, sizeof(kernel->build_id)) == 0) -- cgit v1.2.3-18-g5258 From 61f37a824d6782503ff66bf653f2e07902b641a1 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 27 Nov 2009 16:29:13 -0200 Subject: perf symbols: Rename kernel_mapto kernel_map[s]__functions MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit As we'll have kernel_map[s]__variables too. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1259346563-12568-2-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/builtin-annotate.c | 2 +- tools/perf/util/symbol.c | 46 +++++++++++++++++++++---------------------- tools/perf/util/symbol.h | 2 +- 3 files changed, 25 insertions(+), 25 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c index 18ac5eaefc3..377cb7c9bdd 100644 --- a/tools/perf/builtin-annotate.c +++ b/tools/perf/builtin-annotate.c @@ -184,7 +184,7 @@ got_map: * trick of looking in the whole kernel symbol list. */ if ((long long)ip < 0) { - map = kernel_map; + map = kernel_map__functions; goto got_map; } } diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 8aab89b2248..687fb7f8cc0 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -43,7 +43,7 @@ static struct symbol_conf symbol_conf__defaults = { .try_vmlinux_path = true, }; -static struct rb_root kernel_maps; +static struct rb_root kernel_maps__functions; static void symbols__fixup_end(struct rb_root *self) { @@ -71,7 +71,7 @@ static void symbols__fixup_end(struct rb_root *self) static void kernel_maps__fixup_end(void) { struct map *prev, *curr; - struct rb_node *nd, *prevnd = rb_first(&kernel_maps); + struct rb_node *nd, *prevnd = rb_first(&kernel_maps__functions); if (prevnd == NULL) return; @@ -325,7 +325,7 @@ static int kernel_maps__load_all_kallsyms(void) * kernel_maps__split_kallsyms, when we have split the * maps per module */ - symbols__insert(&kernel_map->dso->functions, sym); + symbols__insert(&kernel_map__functions->dso->functions, sym); } free(line); @@ -346,10 +346,10 @@ out_failure: */ static int kernel_maps__split_kallsyms(symbol_filter_t filter) { - struct map *map = kernel_map; + struct map *map = kernel_map__functions; struct symbol *pos; int count = 0; - struct rb_node *next = rb_first(&kernel_map->dso->functions); + struct rb_node *next = rb_first(&kernel_map__functions->dso->functions); int kernel_range = 0; while (next) { @@ -376,7 +376,7 @@ static int kernel_maps__split_kallsyms(symbol_filter_t filter) */ pos->start = map->map_ip(map, pos->start); pos->end = map->map_ip(map, pos->end); - } else if (map != kernel_map) { + } else if (map != kernel_map__functions) { char dso_name[PATH_MAX]; struct dso *dso; @@ -399,12 +399,12 @@ static int kernel_maps__split_kallsyms(symbol_filter_t filter) } if (filter && filter(map, pos)) { - rb_erase(&pos->rb_node, &kernel_map->dso->functions); + rb_erase(&pos->rb_node, &kernel_map__functions->dso->functions); symbol__delete(pos); } else { - if (map != kernel_map) { + if (map != kernel_map__functions) { rb_erase(&pos->rb_node, - &kernel_map->dso->functions); + &kernel_map__functions->dso->functions); symbols__insert(&map->dso->functions, pos); } count++; @@ -420,8 +420,8 @@ static int kernel_maps__load_kallsyms(symbol_filter_t filter) if (kernel_maps__load_all_kallsyms()) return -1; - symbols__fixup_end(&kernel_map->dso->functions); - kernel_map->dso->origin = DSO__ORIG_KERNEL; + symbols__fixup_end(&kernel_map__functions->dso->functions); + kernel_map__functions->dso->origin = DSO__ORIG_KERNEL; return kernel_maps__split_kallsyms(filter); } @@ -431,7 +431,7 @@ size_t kernel_maps__fprintf(FILE *fp) size_t printed = fprintf(fp, "Kernel maps:\n"); struct rb_node *nd; - for (nd = rb_first(&kernel_maps); nd; nd = rb_next(nd)) { + for (nd = rb_first(&kernel_maps__functions); nd; nd = rb_next(nd)) { struct map *pos = rb_entry(nd, struct map, rb_node); printed += fprintf(fp, "Map:"); @@ -1159,17 +1159,17 @@ out: return ret; } -struct map *kernel_map; +struct map *kernel_map__functions; static void kernel_maps__insert(struct map *map) { - maps__insert(&kernel_maps, map); + maps__insert(&kernel_maps__functions, map); } struct symbol *kernel_maps__find_function(u64 ip, struct map **mapp, symbol_filter_t filter) { - struct map *map = maps__find(&kernel_maps, ip); + struct map *map = maps__find(&kernel_maps__functions, ip); if (mapp) *mapp = map; @@ -1178,7 +1178,7 @@ struct symbol *kernel_maps__find_function(u64 ip, struct map **mapp, ip = map->map_ip(map, ip); return map__find_function(map, ip, filter); } else - WARN_ONCE(RB_EMPTY_ROOT(&kernel_maps), + WARN_ONCE(RB_EMPTY_ROOT(&kernel_maps__functions), "Empty kernel_maps, was symbol__init() called?\n"); return NULL; @@ -1188,7 +1188,7 @@ struct map *kernel_maps__find_by_dso_name(const char *name) { struct rb_node *nd; - for (nd = rb_first(&kernel_maps); nd; nd = rb_next(nd)) { + for (nd = rb_first(&kernel_maps__functions); nd; nd = rb_next(nd)) { struct map *map = rb_entry(nd, struct map, rb_node); if (map->dso && strcmp(map->dso->name, name) == 0) @@ -1505,11 +1505,11 @@ static int kernel_maps__create_kernel_map(const struct symbol_conf *conf) if (kernel == NULL) return -1; - kernel_map = map__new2(0, kernel); - if (kernel_map == NULL) + kernel_map__functions = map__new2(0, kernel); + if (kernel_map__functions == NULL) goto out_delete_kernel_dso; - kernel_map->map_ip = kernel_map->unmap_ip = identity__map_ip; + kernel_map__functions->map_ip = kernel_map__functions->unmap_ip = identity__map_ip; kernel->short_name = "[kernel]"; kernel->kernel = 1; @@ -1522,15 +1522,15 @@ static int kernel_maps__create_kernel_map(const struct symbol_conf *conf) sizeof(kernel->build_id)) == 0) kernel->has_build_id = true; - kernel_maps__insert(kernel_map); + kernel_maps__insert(kernel_map__functions); dsos__add(kernel); dsos__add(vdso); return 0; out_delete_kernel_map: - map__delete(kernel_map); - kernel_map = NULL; + map__delete(kernel_map__functions); + kernel_map__functions = NULL; out_delete_kernel_dso: dso__delete(kernel); return -1; diff --git a/tools/perf/util/symbol.h b/tools/perf/util/symbol.h index 65846d0c5df..b42d196dd8e 100644 --- a/tools/perf/util/symbol.h +++ b/tools/perf/util/symbol.h @@ -105,6 +105,6 @@ size_t kernel_maps__fprintf(FILE *fp); int symbol__init(struct symbol_conf *conf); extern struct list_head dsos; -extern struct map *kernel_map; +extern struct map *kernel_map__functions; extern struct dso *vdso; #endif /* __PERF_SYMBOL */ -- cgit v1.2.3-18-g5258 From b0da954a4759ac19fb80a959e53b613fe376bc12 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 27 Nov 2009 16:29:14 -0200 Subject: perf symbols: Split the dsos list into kernel and user parts MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We don't need to look at modules in dsos__findnew because the kernel events come only with user DSOs. Also we need a way to list just the module DSOs so that we can create multiple sets of maps, now that we will support maps for the variables in a symtab. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1259346563-12568-3-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/util/header.c | 12 +++++++++-- tools/perf/util/symbol.c | 56 ++++++++++++++++++++++++++++++++---------------- tools/perf/util/symbol.h | 2 +- 3 files changed, 48 insertions(+), 22 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index 4b586569bb0..4805e6dfd23 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -185,11 +185,11 @@ static int do_write(int fd, const void *buf, size_t size) return 0; } -static int dsos__write_buildid_table(int fd) +static int __dsos__write_buildid_table(struct list_head *head, int fd) { struct dso *pos; - list_for_each_entry(pos, &dsos, node) { + list_for_each_entry(pos, head, node) { int err; struct build_id_event b; size_t len; @@ -212,6 +212,14 @@ static int dsos__write_buildid_table(int fd) return 0; } +static int dsos__write_buildid_table(int fd) +{ + int err = __dsos__write_buildid_table(&dsos__kernel, fd); + if (err == 0) + err = __dsos__write_buildid_table(&dsos__user, fd); + return err; +} + static int perf_header__adds_write(struct perf_header *self, int fd) { int nr_sections; diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 687fb7f8cc0..dc25231813e 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -28,8 +28,7 @@ enum dso_origin { DSO__ORIG_NOT_FOUND, }; -static void dsos__add(struct dso *dso); -static struct dso *dsos__find(const char *name); +static void dsos__add(struct list_head *head, struct dso *dso); static struct map *map__new2(u64 start, struct dso *dso); static void kernel_maps__insert(struct map *map); static int dso__load_kernel_sym(struct dso *self, struct map *map, @@ -855,7 +854,7 @@ static int dso__load_sym(struct dso *self, struct map *map, const char *name, curr_map->unmap_ip = identity__map_ip; curr_dso->origin = DSO__ORIG_KERNEL; kernel_maps__insert(curr_map); - dsos__add(curr_dso); + dsos__add(&dsos__kernel, curr_dso); } else curr_dso = curr_map->dso; @@ -907,12 +906,12 @@ static bool dso__build_id_equal(const struct dso *self, u8 *build_id) return memcmp(self->build_id, build_id, sizeof(self->build_id)) == 0; } -bool dsos__read_build_ids(void) +static bool __dsos__read_build_ids(struct list_head *head) { bool have_build_id = false; struct dso *pos; - list_for_each_entry(pos, &dsos, node) + list_for_each_entry(pos, head, node) if (filename__read_build_id(pos->long_name, pos->build_id, sizeof(pos->build_id)) > 0) { have_build_id = true; @@ -922,6 +921,12 @@ bool dsos__read_build_ids(void) return have_build_id; } +bool dsos__read_build_ids(void) +{ + return __dsos__read_build_ids(&dsos__kernel) || + __dsos__read_build_ids(&dsos__user); +} + /* * Align offset to 4 bytes as needed for note name and descriptor data. */ @@ -1343,7 +1348,7 @@ static int kernel_maps__create_module_maps(void) dso->origin = DSO__ORIG_KMODULE; kernel_maps__insert(map); - dsos__add(dso); + dsos__add(&dsos__kernel, dso); } free(line); @@ -1445,19 +1450,20 @@ out_fixup: return err; } -LIST_HEAD(dsos); +LIST_HEAD(dsos__user); +LIST_HEAD(dsos__kernel); struct dso *vdso; -static void dsos__add(struct dso *dso) +static void dsos__add(struct list_head *head, struct dso *dso) { - list_add_tail(&dso->node, &dsos); + list_add_tail(&dso->node, head); } -static struct dso *dsos__find(const char *name) +static struct dso *dsos__find(struct list_head *head, const char *name) { struct dso *pos; - list_for_each_entry(pos, &dsos, node) + list_for_each_entry(pos, head, node) if (strcmp(pos->name, name) == 0) return pos; return NULL; @@ -1465,12 +1471,12 @@ static struct dso *dsos__find(const char *name) struct dso *dsos__findnew(const char *name) { - struct dso *dso = dsos__find(name); + struct dso *dso = dsos__find(&dsos__user, name); if (!dso) { dso = dso__new(name); if (dso != NULL) { - dsos__add(dso); + dsos__add(&dsos__user, dso); dso__set_basename(dso); } } @@ -1478,26 +1484,38 @@ struct dso *dsos__findnew(const char *name) return dso; } -void dsos__fprintf(FILE *fp) +static void __dsos__fprintf(struct list_head *head, FILE *fp) { struct dso *pos; - list_for_each_entry(pos, &dsos, node) + list_for_each_entry(pos, head, node) dso__fprintf(pos, fp); } -size_t dsos__fprintf_buildid(FILE *fp) +void dsos__fprintf(FILE *fp) +{ + __dsos__fprintf(&dsos__kernel, fp); + __dsos__fprintf(&dsos__user, fp); +} + +static size_t __dsos__fprintf_buildid(struct list_head *head, FILE *fp) { struct dso *pos; size_t ret = 0; - list_for_each_entry(pos, &dsos, node) { + list_for_each_entry(pos, head, node) { ret += dso__fprintf_buildid(pos, fp); ret += fprintf(fp, " %s\n", pos->long_name); } return ret; } +size_t dsos__fprintf_buildid(FILE *fp) +{ + return (__dsos__fprintf_buildid(&dsos__kernel, fp) + + __dsos__fprintf_buildid(&dsos__user, fp)); +} + static int kernel_maps__create_kernel_map(const struct symbol_conf *conf) { struct dso *kernel = dso__new(conf->vmlinux_name ?: "[kernel.kallsyms]"); @@ -1523,8 +1541,8 @@ static int kernel_maps__create_kernel_map(const struct symbol_conf *conf) kernel->has_build_id = true; kernel_maps__insert(kernel_map__functions); - dsos__add(kernel); - dsos__add(vdso); + dsos__add(&dsos__kernel, kernel); + dsos__add(&dsos__user, vdso); return 0; diff --git a/tools/perf/util/symbol.h b/tools/perf/util/symbol.h index b42d196dd8e..5d0371fe8a0 100644 --- a/tools/perf/util/symbol.h +++ b/tools/perf/util/symbol.h @@ -104,7 +104,7 @@ size_t kernel_maps__fprintf(FILE *fp); int symbol__init(struct symbol_conf *conf); -extern struct list_head dsos; +extern struct list_head dsos__user, dsos__kernel; extern struct map *kernel_map__functions; extern struct dso *vdso; #endif /* __PERF_SYMBOL */ -- cgit v1.2.3-18-g5258 From 605ca4ba017455d39ac6991c58eb1e80fb8af48d Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 27 Nov 2009 16:29:15 -0200 Subject: perf symbols: Unexport kernel_map__functions MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit perf annotate was the only user, and it doesn't really need it. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1259346563-12568-4-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/builtin-annotate.c | 8 +++----- tools/perf/util/symbol.c | 3 +-- tools/perf/util/symbol.h | 1 - 3 files changed, 4 insertions(+), 8 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c index 377cb7c9bdd..0846c8a155e 100644 --- a/tools/perf/builtin-annotate.c +++ b/tools/perf/builtin-annotate.c @@ -169,7 +169,6 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) level = '.'; map = thread__find_map(thread, ip); if (map != NULL) { -got_map: ip = map->map_ip(map, ip); sym = map__find_function(map, ip, symbol_filter); } else { @@ -183,10 +182,9 @@ got_map: * the "[vdso]" dso, but for now lets use the old * trick of looking in the whole kernel symbol list. */ - if ((long long)ip < 0) { - map = kernel_map__functions; - goto got_map; - } + if ((long long)ip < 0) + sym = kernel_maps__find_function(ip, &map, + symbol_filter); } dump_printf(" ...... dso: %s\n", map ? map->dso->long_name : ""); diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index dc25231813e..0b8a298d41a 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -36,6 +36,7 @@ static int dso__load_kernel_sym(struct dso *self, struct map *map, unsigned int symbol__priv_size; static int vmlinux_path__nr_entries; static char **vmlinux_path; +static struct map *kernel_map__functions; static struct symbol_conf symbol_conf__defaults = { .use_modules = true, @@ -1164,8 +1165,6 @@ out: return ret; } -struct map *kernel_map__functions; - static void kernel_maps__insert(struct map *map) { maps__insert(&kernel_maps__functions, map); diff --git a/tools/perf/util/symbol.h b/tools/perf/util/symbol.h index 5d0371fe8a0..fb0be9e92bf 100644 --- a/tools/perf/util/symbol.h +++ b/tools/perf/util/symbol.h @@ -105,6 +105,5 @@ size_t kernel_maps__fprintf(FILE *fp); int symbol__init(struct symbol_conf *conf); extern struct list_head dsos__user, dsos__kernel; -extern struct map *kernel_map__functions; extern struct dso *vdso; #endif /* __PERF_SYMBOL */ -- cgit v1.2.3-18-g5258 From 3610583c29563e23dd038d2870f59c88438bf7a3 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 27 Nov 2009 16:29:16 -0200 Subject: perf symbols: Add a 'type' field to struct map MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit That way we will be able to check if the right symtab is loaded in the underlying DSO. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1259346563-12568-5-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/builtin-top.c | 2 +- tools/perf/util/event.h | 12 +++++++++--- tools/perf/util/map.c | 12 +++++++----- tools/perf/util/process_events.c | 2 +- tools/perf/util/symbol.c | 31 +++++++++++++++++++++---------- tools/perf/util/symbol.h | 4 +++- 6 files changed, 42 insertions(+), 21 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index ded6cf65ad9..a0168f260d0 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -996,7 +996,7 @@ static void event__process_mmap(event_t *self) struct thread *thread = threads__findnew(self->mmap.pid); if (thread != NULL) { - struct map *map = map__new(&self->mmap, NULL, 0); + struct map *map = map__new(&self->mmap, MAP__FUNCTION, NULL, 0); if (map != NULL) thread__insert_map(thread, map); } diff --git a/tools/perf/util/event.h b/tools/perf/util/event.h index 882a9531db9..29543bd8800 100644 --- a/tools/perf/util/event.h +++ b/tools/perf/util/event.h @@ -80,6 +80,10 @@ typedef union event_union { struct sample_event sample; } event_t; +enum map_type { + MAP__FUNCTION, +}; + struct map { union { struct rb_node rb_node; @@ -87,6 +91,7 @@ struct map { }; u64 start; u64 end; + enum map_type type; u64 pgoff; u64 (*map_ip)(struct map *, u64); u64 (*unmap_ip)(struct map *, u64); @@ -112,9 +117,10 @@ struct symbol; typedef int (*symbol_filter_t)(struct map *map, struct symbol *sym); -void map__init(struct map *self, u64 start, u64 end, u64 pgoff, - struct dso *dso); -struct map *map__new(struct mmap_event *event, char *cwd, int cwdlen); +void map__init(struct map *self, enum map_type type, + u64 start, u64 end, u64 pgoff, struct dso *dso); +struct map *map__new(struct mmap_event *event, enum map_type, + char *cwd, int cwdlen); void map__delete(struct map *self); struct map *map__clone(struct map *self); int map__overlap(struct map *l, struct map *r); diff --git a/tools/perf/util/map.c b/tools/perf/util/map.c index 41c5c4a2001..52bb4c6cf74 100644 --- a/tools/perf/util/map.c +++ b/tools/perf/util/map.c @@ -20,9 +20,10 @@ static int strcommon(const char *pathname, char *cwd, int cwdlen) return n; } -void map__init(struct map *self, u64 start, u64 end, u64 pgoff, - struct dso *dso) +void map__init(struct map *self, enum map_type type, + u64 start, u64 end, u64 pgoff, struct dso *dso) { + self->type = type; self->start = start; self->end = end; self->pgoff = pgoff; @@ -32,7 +33,8 @@ void map__init(struct map *self, u64 start, u64 end, u64 pgoff, RB_CLEAR_NODE(&self->rb_node); } -struct map *map__new(struct mmap_event *event, char *cwd, int cwdlen) +struct map *map__new(struct mmap_event *event, enum map_type type, + char *cwd, int cwdlen) { struct map *self = malloc(sizeof(*self)); @@ -63,7 +65,7 @@ struct map *map__new(struct mmap_event *event, char *cwd, int cwdlen) if (dso == NULL) goto out_delete; - map__init(self, event->start, event->start + event->len, + map__init(self, type, event->start, event->start + event->len, event->pgoff, dso); if (self->dso == vdso || anon) @@ -103,7 +105,7 @@ void map__fixup_end(struct map *self, struct rb_root *symbols) struct symbol *map__find_function(struct map *self, u64 ip, symbol_filter_t filter) { - if (!self->dso->loaded) { + if (!dso__loaded(self->dso, self->type)) { int nr = dso__load(self->dso, self, filter); if (nr < 0) { diff --git a/tools/perf/util/process_events.c b/tools/perf/util/process_events.c index a9204363efd..53778684641 100644 --- a/tools/perf/util/process_events.c +++ b/tools/perf/util/process_events.c @@ -6,7 +6,7 @@ int cwdlen; int process_mmap_event(event_t *event, unsigned long offset, unsigned long head) { - struct map *map = map__new(&event->mmap, cwd, cwdlen); + struct map *map = map__new(&event->mmap, MAP__FUNCTION, cwd, cwdlen); struct thread *thread = threads__findnew(event->mmap.pid); dump_printf("%p [%p]: PERF_RECORD_MMAP %d/%d: [%p(%p) @ %p]: %s\n", diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 0b8a298d41a..45a4a9a7618 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -29,7 +29,7 @@ enum dso_origin { }; static void dsos__add(struct list_head *head, struct dso *dso); -static struct map *map__new2(u64 start, struct dso *dso); +static struct map *map__new2(u64 start, struct dso *dso, enum map_type type); static void kernel_maps__insert(struct map *map); static int dso__load_kernel_sym(struct dso *self, struct map *map, symbol_filter_t filter); @@ -45,6 +45,16 @@ static struct symbol_conf symbol_conf__defaults = { static struct rb_root kernel_maps__functions; +bool dso__loaded(const struct dso *self, enum map_type type) +{ + return self->loaded & (1 << type); +} + +static void dso__set_loaded(struct dso *self, enum map_type type) +{ + self->loaded |= (1 << type); +} + static void symbols__fixup_end(struct rb_root *self) { struct rb_node *nd, *prevnd = rb_first(self); @@ -387,7 +397,7 @@ static int kernel_maps__split_kallsyms(symbol_filter_t filter) if (dso == NULL) return -1; - map = map__new2(pos->start, dso); + map = map__new2(pos->start, dso, MAP__FUNCTION); if (map == NULL) { dso__delete(dso); return -1; @@ -846,7 +856,8 @@ static int dso__load_sym(struct dso *self, struct map *map, const char *name, curr_dso = dso__new(dso_name); if (curr_dso == NULL) goto out_elf_end; - curr_map = map__new2(start, curr_dso); + curr_map = map__new2(start, curr_dso, + MAP__FUNCTION); if (curr_map == NULL) { dso__delete(curr_dso); goto out_elf_end; @@ -1076,7 +1087,7 @@ int dso__load(struct dso *self, struct map *map, symbol_filter_t filter) int ret = -1; int fd; - self->loaded = 1; + dso__set_loaded(self, map->type); if (self->kernel) return dso__load_kernel_sym(self, map, filter); @@ -1275,7 +1286,7 @@ static int dsos__set_modules_path(void) * they are loaded) and for vmlinux, where only after we load all the * symbols we'll know where it starts and ends. */ -static struct map *map__new2(u64 start, struct dso *dso) +static struct map *map__new2(u64 start, struct dso *dso, enum map_type type) { struct map *self = malloc(sizeof(*self)); @@ -1283,7 +1294,7 @@ static struct map *map__new2(u64 start, struct dso *dso) /* * ->end will be filled after we load all the symbols */ - map__init(self, start, 0, 0, dso); + map__init(self, type, start, 0, 0, dso); } return self; @@ -1333,7 +1344,7 @@ static int kernel_maps__create_module_maps(void) if (dso == NULL) goto out_delete_line; - map = map__new2(start, dso); + map = map__new2(start, dso, MAP__FUNCTION); if (map == NULL) { dso__delete(dso); goto out_delete_line; @@ -1394,7 +1405,7 @@ static int dso__load_vmlinux(struct dso *self, struct map *map, if (fd < 0) return -1; - self->loaded = 1; + dso__set_loaded(self, map->type); err = dso__load_sym(self, map, self->long_name, fd, filter, 1, 0); close(fd); @@ -1522,7 +1533,7 @@ static int kernel_maps__create_kernel_map(const struct symbol_conf *conf) if (kernel == NULL) return -1; - kernel_map__functions = map__new2(0, kernel); + kernel_map__functions = map__new2(0, kernel, MAP__FUNCTION); if (kernel_map__functions == NULL) goto out_delete_kernel_dso; @@ -1533,7 +1544,7 @@ static int kernel_maps__create_kernel_map(const struct symbol_conf *conf) vdso = dso__new("[vdso]"); if (vdso == NULL) goto out_delete_kernel_map; - vdso->loaded = 1; + dso__set_loaded(vdso, MAP__FUNCTION); if (sysfs__read_build_id("/sys/kernel/notes", kernel->build_id, sizeof(kernel->build_id)) == 0) diff --git a/tools/perf/util/symbol.h b/tools/perf/util/symbol.h index fb0be9e92bf..11d41952ce0 100644 --- a/tools/perf/util/symbol.h +++ b/tools/perf/util/symbol.h @@ -69,10 +69,10 @@ struct dso { struct symbol *(*find_function)(struct dso *, u64 ip); u8 adjust_symbols:1; u8 slen_calculated:1; - u8 loaded:1; u8 has_build_id:1; u8 kernel:1; unsigned char origin; + u8 loaded; u8 build_id[BUILD_ID_SIZE]; u16 long_name_len; const char *short_name; @@ -85,6 +85,8 @@ void dso__delete(struct dso *self); struct symbol *dso__find_function(struct dso *self, u64 ip); +bool dso__loaded(const struct dso *self, enum map_type type); + struct dso *dsos__findnew(const char *name); int dso__load(struct dso *self, struct map *map, symbol_filter_t filter); void dsos__fprintf(FILE *fp); -- cgit v1.2.3-18-g5258 From 6a4694a433a218c729d336b348a01bfc720da095 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 27 Nov 2009 16:29:17 -0200 Subject: perf symbols: Better support for multiple symbol tables per dso MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit By using an array of rb_roots in struct dso we can, from a struct map instance to get the right symbol rb_tree more easily. This way we can have just one symbol lookup method for struct map instances, map__find_symbol, instead of one per symtab type (functions, variables). Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1259346563-12568-6-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/builtin-annotate.c | 2 +- tools/perf/builtin-report.c | 2 +- tools/perf/builtin-top.c | 2 +- tools/perf/util/event.h | 12 +++++---- tools/perf/util/map.c | 12 +++++---- tools/perf/util/symbol.c | 59 ++++++++++++++++++++++++++----------------- tools/perf/util/symbol.h | 7 +++-- 7 files changed, 56 insertions(+), 40 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c index 0846c8a155e..c32e7609b77 100644 --- a/tools/perf/builtin-annotate.c +++ b/tools/perf/builtin-annotate.c @@ -170,7 +170,7 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) map = thread__find_map(thread, ip); if (map != NULL) { ip = map->map_ip(map, ip); - sym = map__find_function(map, ip, symbol_filter); + sym = map__find_symbol(map, ip, symbol_filter); } else { /* * If this is outside of all known maps, diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index e4b1004e76e..400bef981c6 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -455,7 +455,7 @@ got_map: dump_printf(" ...... map: %Lx -> %Lx\n", *ipp, ip); *ipp = ip; - return map ? map__find_function(map, ip, NULL) : NULL; + return map ? map__find_symbol(map, ip, NULL) : NULL; } static int call__match(struct symbol *sym) diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index a0168f260d0..abe78bbd154 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -948,7 +948,7 @@ static void event__process_sample(const event_t *self, int counter) map = thread__find_map(thread, ip); if (map != NULL) { ip = map->map_ip(map, ip); - sym = map__find_function(map, ip, symbol_filter); + sym = map__find_symbol(map, ip, symbol_filter); if (sym == NULL) return; userspace_samples++; diff --git a/tools/perf/util/event.h b/tools/perf/util/event.h index 29543bd8800..3ae3c964c90 100644 --- a/tools/perf/util/event.h +++ b/tools/perf/util/event.h @@ -81,7 +81,9 @@ typedef union event_union { } event_t; enum map_type { - MAP__FUNCTION, + MAP__FUNCTION = 0, + + MAP__NR_TYPES, }; struct map { @@ -125,10 +127,10 @@ void map__delete(struct map *self); struct map *map__clone(struct map *self); int map__overlap(struct map *l, struct map *r); size_t map__fprintf(struct map *self, FILE *fp); -struct symbol *map__find_function(struct map *self, u64 ip, - symbol_filter_t filter); -void map__fixup_start(struct map *self, struct rb_root *symbols); -void map__fixup_end(struct map *self, struct rb_root *symbols); +struct symbol *map__find_symbol(struct map *self, u64 addr, + symbol_filter_t filter); +void map__fixup_start(struct map *self); +void map__fixup_end(struct map *self); int event__synthesize_thread(pid_t pid, int (*process)(event_t *event)); void event__synthesize_threads(int (*process)(event_t *event)); diff --git a/tools/perf/util/map.c b/tools/perf/util/map.c index 52bb4c6cf74..69f94fe9db2 100644 --- a/tools/perf/util/map.c +++ b/tools/perf/util/map.c @@ -82,8 +82,9 @@ void map__delete(struct map *self) free(self); } -void map__fixup_start(struct map *self, struct rb_root *symbols) +void map__fixup_start(struct map *self) { + struct rb_root *symbols = &self->dso->symbols[self->type]; struct rb_node *nd = rb_first(symbols); if (nd != NULL) { struct symbol *sym = rb_entry(nd, struct symbol, rb_node); @@ -91,8 +92,9 @@ void map__fixup_start(struct map *self, struct rb_root *symbols) } } -void map__fixup_end(struct map *self, struct rb_root *symbols) +void map__fixup_end(struct map *self) { + struct rb_root *symbols = &self->dso->symbols[self->type]; struct rb_node *nd = rb_last(symbols); if (nd != NULL) { struct symbol *sym = rb_entry(nd, struct symbol, rb_node); @@ -102,8 +104,8 @@ void map__fixup_end(struct map *self, struct rb_root *symbols) #define DSO__DELETED "(deleted)" -struct symbol *map__find_function(struct map *self, u64 ip, - symbol_filter_t filter) +struct symbol *map__find_symbol(struct map *self, u64 addr, + symbol_filter_t filter) { if (!dso__loaded(self->dso, self->type)) { int nr = dso__load(self->dso, self, filter); @@ -138,7 +140,7 @@ struct symbol *map__find_function(struct map *self, u64 ip, } } - return self->dso->find_function(self->dso, ip); + return self->dso->find_symbol(self->dso, self->type, addr); } struct map *map__clone(struct map *self) diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 45a4a9a7618..9a2dd819dee 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -31,6 +31,7 @@ enum dso_origin { static void dsos__add(struct list_head *head, struct dso *dso); static struct map *map__new2(u64 start, struct dso *dso, enum map_type type); static void kernel_maps__insert(struct map *map); +struct symbol *dso__find_symbol(struct dso *self, enum map_type type, u64 addr); static int dso__load_kernel_sym(struct dso *self, struct map *map, symbol_filter_t filter); unsigned int symbol__priv_size; @@ -151,11 +152,13 @@ struct dso *dso__new(const char *name) struct dso *self = malloc(sizeof(*self) + strlen(name) + 1); if (self != NULL) { + int i; strcpy(self->name, name); dso__set_long_name(self, self->name); self->short_name = self->name; - self->functions = RB_ROOT; - self->find_function = dso__find_function; + for (i = 0; i < MAP__NR_TYPES; ++i) + self->symbols[i] = RB_ROOT; + self->find_symbol = dso__find_symbol; self->slen_calculated = 0; self->origin = DSO__ORIG_NOT_FOUND; self->loaded = 0; @@ -180,7 +183,9 @@ static void symbols__delete(struct rb_root *self) void dso__delete(struct dso *self) { - symbols__delete(&self->functions); + int i; + for (i = 0; i < MAP__NR_TYPES; ++i) + symbols__delete(&self->symbols[i]); if (self->long_name != self->name) free(self->long_name); free(self); @@ -234,9 +239,9 @@ static struct symbol *symbols__find(struct rb_root *self, u64 ip) return NULL; } -struct symbol *dso__find_function(struct dso *self, u64 ip) +struct symbol *dso__find_symbol(struct dso *self, enum map_type type, u64 addr) { - return symbols__find(&self->functions, ip); + return symbols__find(&self->symbols[type], addr); } int build_id__sprintf(u8 *self, int len, char *bf) @@ -262,17 +267,25 @@ size_t dso__fprintf_buildid(struct dso *self, FILE *fp) return fprintf(fp, "%s", sbuild_id); } +static const char * map_type__name[MAP__NR_TYPES] = { + [MAP__FUNCTION] = "Functions", +}; + size_t dso__fprintf(struct dso *self, FILE *fp) { + int i; struct rb_node *nd; size_t ret = fprintf(fp, "dso: %s (", self->short_name); ret += dso__fprintf_buildid(self, fp); - ret += fprintf(fp, ")\nFunctions:\n"); + ret += fprintf(fp, ")\n"); + for (i = 0; i < MAP__NR_TYPES; ++i) { + ret += fprintf(fp, "%s:\n", map_type__name[i]); - for (nd = rb_first(&self->functions); nd; nd = rb_next(nd)) { - struct symbol *pos = rb_entry(nd, struct symbol, rb_node); - ret += symbol__fprintf(pos, fp); + for (nd = rb_first(&self->symbols[i]); nd; nd = rb_next(nd)) { + struct symbol *pos = rb_entry(nd, struct symbol, rb_node); + ret += symbol__fprintf(pos, fp); + } } return ret; @@ -335,7 +348,7 @@ static int kernel_maps__load_all_kallsyms(void) * kernel_maps__split_kallsyms, when we have split the * maps per module */ - symbols__insert(&kernel_map__functions->dso->functions, sym); + symbols__insert(&kernel_map__functions->dso->symbols[MAP__FUNCTION], sym); } free(line); @@ -359,7 +372,7 @@ static int kernel_maps__split_kallsyms(symbol_filter_t filter) struct map *map = kernel_map__functions; struct symbol *pos; int count = 0; - struct rb_node *next = rb_first(&kernel_map__functions->dso->functions); + struct rb_node *next = rb_first(&kernel_map__functions->dso->symbols[map->type]); int kernel_range = 0; while (next) { @@ -409,13 +422,13 @@ static int kernel_maps__split_kallsyms(symbol_filter_t filter) } if (filter && filter(map, pos)) { - rb_erase(&pos->rb_node, &kernel_map__functions->dso->functions); + rb_erase(&pos->rb_node, &kernel_map__functions->dso->symbols[map->type]); symbol__delete(pos); } else { if (map != kernel_map__functions) { rb_erase(&pos->rb_node, - &kernel_map__functions->dso->functions); - symbols__insert(&map->dso->functions, pos); + &kernel_map__functions->dso->symbols[map->type]); + symbols__insert(&map->dso->symbols[map->type], pos); } count++; } @@ -430,7 +443,7 @@ static int kernel_maps__load_kallsyms(symbol_filter_t filter) if (kernel_maps__load_all_kallsyms()) return -1; - symbols__fixup_end(&kernel_map__functions->dso->functions); + symbols__fixup_end(&kernel_map__functions->dso->symbols[MAP__FUNCTION]); kernel_map__functions->dso->origin = DSO__ORIG_KERNEL; return kernel_maps__split_kallsyms(filter); @@ -501,7 +514,7 @@ static int dso__load_perf_map(struct dso *self, struct map *map, if (filter && filter(map, sym)) symbol__delete(sym); else { - symbols__insert(&self->functions, sym); + symbols__insert(&self->symbols[map->type], sym); nr_syms++; } } @@ -699,7 +712,7 @@ static int dso__synthesize_plt_symbols(struct dso *self, struct map *map, if (filter && filter(map, f)) symbol__delete(f); else { - symbols__insert(&self->functions, f); + symbols__insert(&self->symbols[map->type], f); ++nr; } } @@ -721,7 +734,7 @@ static int dso__synthesize_plt_symbols(struct dso *self, struct map *map, if (filter && filter(map, f)) symbol__delete(f); else { - symbols__insert(&self->functions, f); + symbols__insert(&self->symbols[map->type], f); ++nr; } } @@ -896,7 +909,7 @@ new_symbol: if (filter && filter(curr_map, f)) symbol__delete(f); else { - symbols__insert(&curr_dso->functions, f); + symbols__insert(&curr_dso->symbols[curr_map->type], f); nr++; } } @@ -905,7 +918,7 @@ new_symbol: * For misannotated, zeroed, ASM function sizes. */ if (nr > 0) - symbols__fixup_end(&self->functions); + symbols__fixup_end(&self->symbols[map->type]); err = nr; out_elf_end: elf_end(elf); @@ -1191,7 +1204,7 @@ struct symbol *kernel_maps__find_function(u64 ip, struct map **mapp, if (map) { ip = map->map_ip(map, ip); - return map__find_function(map, ip, filter); + return map__find_symbol(map, ip, filter); } else WARN_ONCE(RB_EMPTY_ROOT(&kernel_maps__functions), "Empty kernel_maps, was symbol__init() called?\n"); @@ -1453,8 +1466,8 @@ do_kallsyms: if (err > 0) { out_fixup: - map__fixup_start(map, &map->dso->functions); - map__fixup_end(map, &map->dso->functions); + map__fixup_start(map); + map__fixup_end(map); } return err; diff --git a/tools/perf/util/symbol.h b/tools/perf/util/symbol.h index 11d41952ce0..8934814d5a6 100644 --- a/tools/perf/util/symbol.h +++ b/tools/perf/util/symbol.h @@ -65,8 +65,9 @@ static inline void *symbol__priv(struct symbol *self) struct dso { struct list_head node; - struct rb_root functions; - struct symbol *(*find_function)(struct dso *, u64 ip); + struct rb_root symbols[MAP__NR_TYPES]; + struct symbol *(*find_symbol)(struct dso *self, + enum map_type type, u64 addr); u8 adjust_symbols:1; u8 slen_calculated:1; u8 has_build_id:1; @@ -83,8 +84,6 @@ struct dso { struct dso *dso__new(const char *name); void dso__delete(struct dso *self); -struct symbol *dso__find_function(struct dso *self, u64 ip); - bool dso__loaded(const struct dso *self, enum map_type type); struct dso *dsos__findnew(const char *name); -- cgit v1.2.3-18-g5258 From 4e06255f5cf2acf6a5abfe7df8c9690463259dea Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 27 Nov 2009 16:29:18 -0200 Subject: perf symbols: Make the kallsyms loading routines part of the dso class MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit So that the kallsyms loading routines are the direct counterpart of the vmlinux loading ones, i.e. dso__load_kallsyms is the counterpart of dso__load_vmlinux. In the process make them also use the symbols rb tree indexed by map->type, paving the way for supporting other types of symtabs, such as the next one to be supported: variables. This also allowed removal of yet another global variable: kernel_map__functions. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1259346563-12568-7-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/util/symbol.c | 77 ++++++++++++++++++++++++------------------------ 1 file changed, 38 insertions(+), 39 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 9a2dd819dee..956656fcaab 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -37,7 +37,6 @@ static int dso__load_kernel_sym(struct dso *self, struct map *map, unsigned int symbol__priv_size; static int vmlinux_path__nr_entries; static char **vmlinux_path; -static struct map *kernel_map__functions; static struct symbol_conf symbol_conf__defaults = { .use_modules = true, @@ -296,10 +295,11 @@ size_t dso__fprintf(struct dso *self, FILE *fp) * so that we can in the next step set the symbol ->end address and then * call kernel_maps__split_kallsyms. */ -static int kernel_maps__load_all_kallsyms(void) +static int dso__load_all_kallsyms(struct dso *self, struct map *map) { char *line = NULL; size_t n; + struct rb_root *root = &self->symbols[map->type]; FILE *file = fopen("/proc/kallsyms", "r"); if (file == NULL) @@ -342,13 +342,11 @@ static int kernel_maps__load_all_kallsyms(void) if (sym == NULL) goto out_delete_line; - /* * We will pass the symbols to the filter later, in - * kernel_maps__split_kallsyms, when we have split the - * maps per module + * map__split_kallsyms, when we have split the maps per module */ - symbols__insert(&kernel_map__functions->dso->symbols[MAP__FUNCTION], sym); + symbols__insert(root, sym); } free(line); @@ -367,12 +365,14 @@ out_failure: * kernel range is broken in several maps, named [kernel].N, as we don't have * the original ELF section names vmlinux have. */ -static int kernel_maps__split_kallsyms(symbol_filter_t filter) +static int dso__split_kallsyms(struct dso *self, struct map *map, + symbol_filter_t filter) { - struct map *map = kernel_map__functions; + struct map *curr_map = map; struct symbol *pos; int count = 0; - struct rb_node *next = rb_first(&kernel_map__functions->dso->symbols[map->type]); + struct rb_root *root = &self->symbols[map->type]; + struct rb_node *next = rb_first(root); int kernel_range = 0; while (next) { @@ -385,9 +385,9 @@ static int kernel_maps__split_kallsyms(symbol_filter_t filter) if (module) { *module++ = '\0'; - if (strcmp(map->dso->name, module)) { - map = kernel_maps__find_by_dso_name(module); - if (!map) { + if (strcmp(self->name, module)) { + curr_map = kernel_maps__find_by_dso_name(module); + if (curr_map == NULL) { pr_err("/proc/{kallsyms,modules} " "inconsistency!\n"); return -1; @@ -397,9 +397,9 @@ static int kernel_maps__split_kallsyms(symbol_filter_t filter) * So that we look just like we get from .ko files, * i.e. not prelinked, relative to map->start. */ - pos->start = map->map_ip(map, pos->start); - pos->end = map->map_ip(map, pos->end); - } else if (map != kernel_map__functions) { + pos->start = curr_map->map_ip(curr_map, pos->start); + pos->end = curr_map->map_ip(curr_map, pos->end); + } else if (curr_map != map) { char dso_name[PATH_MAX]; struct dso *dso; @@ -410,25 +410,24 @@ static int kernel_maps__split_kallsyms(symbol_filter_t filter) if (dso == NULL) return -1; - map = map__new2(pos->start, dso, MAP__FUNCTION); + curr_map = map__new2(pos->start, dso, map->type); if (map == NULL) { dso__delete(dso); return -1; } - map->map_ip = map->unmap_ip = identity__map_ip; - kernel_maps__insert(map); + curr_map->map_ip = curr_map->unmap_ip = identity__map_ip; + kernel_maps__insert(curr_map); ++kernel_range; } - if (filter && filter(map, pos)) { - rb_erase(&pos->rb_node, &kernel_map__functions->dso->symbols[map->type]); + if (filter && filter(curr_map, pos)) { + rb_erase(&pos->rb_node, root); symbol__delete(pos); } else { - if (map != kernel_map__functions) { - rb_erase(&pos->rb_node, - &kernel_map__functions->dso->symbols[map->type]); - symbols__insert(&map->dso->symbols[map->type], pos); + if (curr_map != map) { + rb_erase(&pos->rb_node, root); + symbols__insert(&curr_map->dso->symbols[curr_map->type], pos); } count++; } @@ -438,15 +437,16 @@ static int kernel_maps__split_kallsyms(symbol_filter_t filter) } -static int kernel_maps__load_kallsyms(symbol_filter_t filter) +static int dso__load_kallsyms(struct dso *self, struct map *map, + symbol_filter_t filter) { - if (kernel_maps__load_all_kallsyms()) + if (dso__load_all_kallsyms(self, map) < 0) return -1; - symbols__fixup_end(&kernel_map__functions->dso->symbols[MAP__FUNCTION]); - kernel_map__functions->dso->origin = DSO__ORIG_KERNEL; + symbols__fixup_end(&self->symbols[map->type]); + self->origin = DSO__ORIG_KERNEL; - return kernel_maps__split_kallsyms(filter); + return dso__split_kallsyms(self, map, filter); } size_t kernel_maps__fprintf(FILE *fp) @@ -1457,9 +1457,8 @@ static int dso__load_kernel_sym(struct dso *self, struct map *map, if (err <= 0) { pr_info("The file %s cannot be used, " "trying to use /proc/kallsyms...", self->long_name); - sleep(2); do_kallsyms: - err = kernel_maps__load_kallsyms(filter); + err = dso__load_kallsyms(self, map, filter); if (err > 0 && !is_kallsyms) dso__set_long_name(self, strdup("[kernel.kallsyms]")); } @@ -1541,18 +1540,19 @@ size_t dsos__fprintf_buildid(FILE *fp) static int kernel_maps__create_kernel_map(const struct symbol_conf *conf) { + struct map *kmap; struct dso *kernel = dso__new(conf->vmlinux_name ?: "[kernel.kallsyms]"); if (kernel == NULL) return -1; - kernel_map__functions = map__new2(0, kernel, MAP__FUNCTION); - if (kernel_map__functions == NULL) + kmap = map__new2(0, kernel, MAP__FUNCTION); + if (kmap == NULL) goto out_delete_kernel_dso; - kernel_map__functions->map_ip = kernel_map__functions->unmap_ip = identity__map_ip; - kernel->short_name = "[kernel]"; - kernel->kernel = 1; + kmap->map_ip = kmap->unmap_ip = identity__map_ip; + kernel->short_name = "[kernel]"; + kernel->kernel = 1; vdso = dso__new("[vdso]"); if (vdso == NULL) @@ -1563,15 +1563,14 @@ static int kernel_maps__create_kernel_map(const struct symbol_conf *conf) sizeof(kernel->build_id)) == 0) kernel->has_build_id = true; - kernel_maps__insert(kernel_map__functions); + kernel_maps__insert(kmap); dsos__add(&dsos__kernel, kernel); dsos__add(&dsos__user, vdso); return 0; out_delete_kernel_map: - map__delete(kernel_map__functions); - kernel_map__functions = NULL; + map__delete(kmap); out_delete_kernel_dso: dso__delete(kernel); return -1; -- cgit v1.2.3-18-g5258 From 23ea4a3fadc6b1692dec935397ea15e2affc1cba Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 27 Nov 2009 16:29:19 -0200 Subject: perf symbols: Kernel_maps should be an array of MAP__NR_TYPES entries MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit So that we can support multiple symbol table types. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1259346563-12568-8-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/util/symbol.c | 51 ++++++++++++++++++++++++++++++++++++------------ tools/perf/util/thread.h | 1 - 2 files changed, 38 insertions(+), 14 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 956656fcaab..581db4c4325 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -29,6 +29,7 @@ enum dso_origin { }; static void dsos__add(struct list_head *head, struct dso *dso); +static struct map *kernel_maps__find_by_dso_name(const char *name); static struct map *map__new2(u64 start, struct dso *dso, enum map_type type); static void kernel_maps__insert(struct map *map); struct symbol *dso__find_symbol(struct dso *self, enum map_type type, u64 addr); @@ -43,7 +44,7 @@ static struct symbol_conf symbol_conf__defaults = { .try_vmlinux_path = true, }; -static struct rb_root kernel_maps__functions; +static struct rb_root kernel_maps[MAP__NR_TYPES]; bool dso__loaded(const struct dso *self, enum map_type type) { @@ -78,10 +79,10 @@ static void symbols__fixup_end(struct rb_root *self) curr->end = roundup(curr->start, 4096); } -static void kernel_maps__fixup_end(void) +static void __kernel_maps__fixup_end(struct rb_root *root) { struct map *prev, *curr; - struct rb_node *nd, *prevnd = rb_first(&kernel_maps__functions); + struct rb_node *nd, *prevnd = rb_first(root); if (prevnd == NULL) return; @@ -101,6 +102,13 @@ static void kernel_maps__fixup_end(void) curr->end = ~0UL; } +static void kernel_maps__fixup_end(void) +{ + int i; + for (i = 0; i < MAP__NR_TYPES; ++i) + __kernel_maps__fixup_end(&kernel_maps[i]); +} + static struct symbol *symbol__new(u64 start, u64 len, const char *name) { size_t namelen = strlen(name) + 1; @@ -449,12 +457,12 @@ static int dso__load_kallsyms(struct dso *self, struct map *map, return dso__split_kallsyms(self, map, filter); } -size_t kernel_maps__fprintf(FILE *fp) +static size_t __kernel_maps__fprintf(enum map_type type, FILE *fp) { - size_t printed = fprintf(fp, "Kernel maps:\n"); + size_t printed = fprintf(fp, "%s:\n", map_type__name[type]); struct rb_node *nd; - for (nd = rb_first(&kernel_maps__functions); nd; nd = rb_next(nd)) { + for (nd = rb_first(&kernel_maps[type]); nd; nd = rb_next(nd)) { struct map *pos = rb_entry(nd, struct map, rb_node); printed += fprintf(fp, "Map:"); @@ -465,6 +473,16 @@ size_t kernel_maps__fprintf(FILE *fp) } } + return printed; +} + +size_t kernel_maps__fprintf(FILE *fp) +{ + size_t printed = fprintf(fp, "Kernel maps:\n"); + int i; + for (i = 0; i < MAP__NR_TYPES; ++i) + printed += __kernel_maps__fprintf(i, fp); + return printed + fprintf(fp, "END kernel maps\n"); } @@ -1191,13 +1209,14 @@ out: static void kernel_maps__insert(struct map *map) { - maps__insert(&kernel_maps__functions, map); + maps__insert(&kernel_maps[map->type], map); } -struct symbol *kernel_maps__find_function(u64 ip, struct map **mapp, - symbol_filter_t filter) +static struct symbol *kernel_maps__find_symbol(u64 ip, enum map_type type, + struct map **mapp, + symbol_filter_t filter) { - struct map *map = maps__find(&kernel_maps__functions, ip); + struct map *map = maps__find(&kernel_maps[type], ip); if (mapp) *mapp = map; @@ -1206,17 +1225,23 @@ struct symbol *kernel_maps__find_function(u64 ip, struct map **mapp, ip = map->map_ip(map, ip); return map__find_symbol(map, ip, filter); } else - WARN_ONCE(RB_EMPTY_ROOT(&kernel_maps__functions), + WARN_ONCE(RB_EMPTY_ROOT(&kernel_maps[type]), "Empty kernel_maps, was symbol__init() called?\n"); return NULL; } -struct map *kernel_maps__find_by_dso_name(const char *name) +struct symbol *kernel_maps__find_function(u64 ip, struct map **mapp, + symbol_filter_t filter) +{ + return kernel_maps__find_symbol(ip, MAP__FUNCTION, mapp, filter); +} + +static struct map *kernel_maps__find_by_dso_name(const char *name) { struct rb_node *nd; - for (nd = rb_first(&kernel_maps__functions); nd; nd = rb_next(nd)) { + for (nd = rb_first(&kernel_maps[MAP__FUNCTION]); nd; nd = rb_next(nd)) { struct map *map = rb_entry(nd, struct map, rb_node); if (map->dso && strcmp(map->dso->name, name) == 0) diff --git a/tools/perf/util/thread.h b/tools/perf/util/thread.h index 74cba6487ed..54580bb8000 100644 --- a/tools/perf/util/thread.h +++ b/tools/perf/util/thread.h @@ -28,7 +28,6 @@ struct map *maps__find(struct rb_root *maps, u64 ip); struct symbol *kernel_maps__find_function(const u64 ip, struct map **mapp, symbol_filter_t filter); -struct map *kernel_maps__find_by_dso_name(const char *name); static inline struct map *thread__find_map(struct thread *self, u64 ip) { -- cgit v1.2.3-18-g5258 From 95011c600740837288a3b34b411244a4d9157c4e Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 27 Nov 2009 16:29:20 -0200 Subject: perf symbols: Support multiple symtabs in struct thread MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Making the routines that were so far specific to the kernel maps useful for all threads. This is done by making the kernel maps be contained in a kernel "thread". This gets the kernel specific routines closer to the userspace counterparts, which will help in reducing the boilerplate for resolving a symbol, as will be demonstrated in the next patches. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1259346563-12568-9-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/builtin-annotate.c | 2 +- tools/perf/builtin-report.c | 2 +- tools/perf/builtin-top.c | 2 +- tools/perf/util/symbol.c | 157 ++++++++++++++++-------------------------- tools/perf/util/symbol.h | 2 +- tools/perf/util/thread.c | 115 ++++++++++++++++++++++++------- tools/perf/util/thread.h | 15 ++-- 7 files changed, 162 insertions(+), 133 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c index c32e7609b77..3ebd70b1ef9 100644 --- a/tools/perf/builtin-annotate.c +++ b/tools/perf/builtin-annotate.c @@ -167,7 +167,7 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) map ? map->dso->long_name : ""); } else if (event->header.misc & PERF_RECORD_MISC_USER) { level = '.'; - map = thread__find_map(thread, ip); + map = thread__find_map(thread, MAP__FUNCTION, ip); if (map != NULL) { ip = map->map_ip(map, ip); sym = map__find_symbol(map, ip, symbol_filter); diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index 400bef981c6..9bd20c2ee3d 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -422,7 +422,7 @@ resolve_symbol(struct thread *thread, struct map **mapp, u64 *ipp) if (!thread) return NULL; - map = thread__find_map(thread, ip); + map = thread__find_map(thread, MAP__FUNCTION, ip); if (map != NULL) { /* * We have to do this here as we may have a dso diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index abe78bbd154..bf6730c7603 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -945,7 +945,7 @@ static void event__process_sample(const event_t *self, int counter) if (thread == NULL) return; - map = thread__find_map(thread, ip); + map = thread__find_map(thread, MAP__FUNCTION, ip); if (map != NULL) { ip = map->map_ip(map, ip); sym = map__find_symbol(map, ip, symbol_filter); diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 581db4c4325..b6a2941e778 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -29,12 +29,11 @@ enum dso_origin { }; static void dsos__add(struct list_head *head, struct dso *dso); -static struct map *kernel_maps__find_by_dso_name(const char *name); +static struct map *thread__find_map_by_name(struct thread *self, char *name); static struct map *map__new2(u64 start, struct dso *dso, enum map_type type); -static void kernel_maps__insert(struct map *map); struct symbol *dso__find_symbol(struct dso *self, enum map_type type, u64 addr); static int dso__load_kernel_sym(struct dso *self, struct map *map, - symbol_filter_t filter); + struct thread *thread, symbol_filter_t filter); unsigned int symbol__priv_size; static int vmlinux_path__nr_entries; static char **vmlinux_path; @@ -44,7 +43,7 @@ static struct symbol_conf symbol_conf__defaults = { .try_vmlinux_path = true, }; -static struct rb_root kernel_maps[MAP__NR_TYPES]; +static struct thread kthread_mem, *kthread = &kthread_mem; bool dso__loaded(const struct dso *self, enum map_type type) { @@ -79,10 +78,10 @@ static void symbols__fixup_end(struct rb_root *self) curr->end = roundup(curr->start, 4096); } -static void __kernel_maps__fixup_end(struct rb_root *root) +static void __thread__fixup_maps_end(struct thread *self, enum map_type type) { struct map *prev, *curr; - struct rb_node *nd, *prevnd = rb_first(root); + struct rb_node *nd, *prevnd = rb_first(&self->maps[type]); if (prevnd == NULL) return; @@ -102,11 +101,11 @@ static void __kernel_maps__fixup_end(struct rb_root *root) curr->end = ~0UL; } -static void kernel_maps__fixup_end(void) +static void thread__fixup_maps_end(struct thread *self) { int i; for (i = 0; i < MAP__NR_TYPES; ++i) - __kernel_maps__fixup_end(&kernel_maps[i]); + __thread__fixup_maps_end(self, i); } static struct symbol *symbol__new(u64 start, u64 len, const char *name) @@ -274,25 +273,16 @@ size_t dso__fprintf_buildid(struct dso *self, FILE *fp) return fprintf(fp, "%s", sbuild_id); } -static const char * map_type__name[MAP__NR_TYPES] = { - [MAP__FUNCTION] = "Functions", -}; - -size_t dso__fprintf(struct dso *self, FILE *fp) +size_t dso__fprintf(struct dso *self, enum map_type type, FILE *fp) { - int i; struct rb_node *nd; size_t ret = fprintf(fp, "dso: %s (", self->short_name); ret += dso__fprintf_buildid(self, fp); ret += fprintf(fp, ")\n"); - for (i = 0; i < MAP__NR_TYPES; ++i) { - ret += fprintf(fp, "%s:\n", map_type__name[i]); - - for (nd = rb_first(&self->symbols[i]); nd; nd = rb_next(nd)) { - struct symbol *pos = rb_entry(nd, struct symbol, rb_node); - ret += symbol__fprintf(pos, fp); - } + for (nd = rb_first(&self->symbols[type]); nd; nd = rb_next(nd)) { + struct symbol *pos = rb_entry(nd, struct symbol, rb_node); + ret += symbol__fprintf(pos, fp); } return ret; @@ -373,7 +363,7 @@ out_failure: * kernel range is broken in several maps, named [kernel].N, as we don't have * the original ELF section names vmlinux have. */ -static int dso__split_kallsyms(struct dso *self, struct map *map, +static int dso__split_kallsyms(struct dso *self, struct map *map, struct thread *thread, symbol_filter_t filter) { struct map *curr_map = map; @@ -394,10 +384,10 @@ static int dso__split_kallsyms(struct dso *self, struct map *map, *module++ = '\0'; if (strcmp(self->name, module)) { - curr_map = kernel_maps__find_by_dso_name(module); + curr_map = thread__find_map_by_name(thread, module); if (curr_map == NULL) { - pr_err("/proc/{kallsyms,modules} " - "inconsistency!\n"); + pr_debug("/proc/{kallsyms,modules} " + "inconsistency!\n"); return -1; } } @@ -425,7 +415,7 @@ static int dso__split_kallsyms(struct dso *self, struct map *map, } curr_map->map_ip = curr_map->unmap_ip = identity__map_ip; - kernel_maps__insert(curr_map); + __thread__insert_map(thread, curr_map); ++kernel_range; } @@ -446,7 +436,7 @@ static int dso__split_kallsyms(struct dso *self, struct map *map, static int dso__load_kallsyms(struct dso *self, struct map *map, - symbol_filter_t filter) + struct thread *thread, symbol_filter_t filter) { if (dso__load_all_kallsyms(self, map) < 0) return -1; @@ -454,35 +444,13 @@ static int dso__load_kallsyms(struct dso *self, struct map *map, symbols__fixup_end(&self->symbols[map->type]); self->origin = DSO__ORIG_KERNEL; - return dso__split_kallsyms(self, map, filter); -} - -static size_t __kernel_maps__fprintf(enum map_type type, FILE *fp) -{ - size_t printed = fprintf(fp, "%s:\n", map_type__name[type]); - struct rb_node *nd; - - for (nd = rb_first(&kernel_maps[type]); nd; nd = rb_next(nd)) { - struct map *pos = rb_entry(nd, struct map, rb_node); - - printed += fprintf(fp, "Map:"); - printed += map__fprintf(pos, fp); - if (verbose > 1) { - printed += dso__fprintf(pos->dso, fp); - printed += fprintf(fp, "--\n"); - } - } - - return printed; + return dso__split_kallsyms(self, map, thread, filter); } size_t kernel_maps__fprintf(FILE *fp) { size_t printed = fprintf(fp, "Kernel maps:\n"); - int i; - for (i = 0; i < MAP__NR_TYPES; ++i) - printed += __kernel_maps__fprintf(i, fp); - + printed += thread__fprintf_maps(kthread, fp); return printed + fprintf(fp, "END kernel maps\n"); } @@ -772,9 +740,9 @@ out: return 0; } -static int dso__load_sym(struct dso *self, struct map *map, const char *name, - int fd, symbol_filter_t filter, int kernel, - int kmodule) +static int dso__load_sym(struct dso *self, struct map *map, + struct thread *thread, const char *name, int fd, + symbol_filter_t filter, int kernel, int kmodule) { struct map *curr_map = map; struct dso *curr_dso = self; @@ -877,7 +845,7 @@ static int dso__load_sym(struct dso *self, struct map *map, const char *name, snprintf(dso_name, sizeof(dso_name), "%s%s", self->short_name, section_name); - curr_map = kernel_maps__find_by_dso_name(dso_name); + curr_map = thread__find_map_by_name(thread, dso_name); if (curr_map == NULL) { u64 start = sym.st_value; @@ -896,7 +864,7 @@ static int dso__load_sym(struct dso *self, struct map *map, const char *name, curr_map->map_ip = identity__map_ip; curr_map->unmap_ip = identity__map_ip; curr_dso->origin = DSO__ORIG_KERNEL; - kernel_maps__insert(curr_map); + __thread__insert_map(kthread, curr_map); dsos__add(&dsos__kernel, curr_dso); } else curr_dso = curr_map->dso; @@ -1121,7 +1089,7 @@ int dso__load(struct dso *self, struct map *map, symbol_filter_t filter) dso__set_loaded(self, map->type); if (self->kernel) - return dso__load_kernel_sym(self, map, filter); + return dso__load_kernel_sym(self, map, kthread, filter); name = malloc(size); if (!name) @@ -1186,7 +1154,7 @@ compare_build_id: fd = open(name, O_RDONLY); } while (fd < 0); - ret = dso__load_sym(self, map, name, fd, filter, 0, 0); + ret = dso__load_sym(self, map, NULL, name, fd, filter, 0, 0); close(fd); /* @@ -1207,16 +1175,11 @@ out: return ret; } -static void kernel_maps__insert(struct map *map) -{ - maps__insert(&kernel_maps[map->type], map); -} - -static struct symbol *kernel_maps__find_symbol(u64 ip, enum map_type type, - struct map **mapp, - symbol_filter_t filter) +static struct symbol *thread__find_symbol(struct thread *self, u64 ip, + enum map_type type, struct map **mapp, + symbol_filter_t filter) { - struct map *map = maps__find(&kernel_maps[type], ip); + struct map *map = thread__find_map(self, type, ip); if (mapp) *mapp = map; @@ -1224,9 +1187,7 @@ static struct symbol *kernel_maps__find_symbol(u64 ip, enum map_type type, if (map) { ip = map->map_ip(map, ip); return map__find_symbol(map, ip, filter); - } else - WARN_ONCE(RB_EMPTY_ROOT(&kernel_maps[type]), - "Empty kernel_maps, was symbol__init() called?\n"); + } return NULL; } @@ -1234,14 +1195,14 @@ static struct symbol *kernel_maps__find_symbol(u64 ip, enum map_type type, struct symbol *kernel_maps__find_function(u64 ip, struct map **mapp, symbol_filter_t filter) { - return kernel_maps__find_symbol(ip, MAP__FUNCTION, mapp, filter); + return thread__find_symbol(kthread, ip, MAP__FUNCTION, mapp, filter); } -static struct map *kernel_maps__find_by_dso_name(const char *name) +static struct map *thread__find_map_by_name(struct thread *self, char *name) { struct rb_node *nd; - for (nd = rb_first(&kernel_maps[MAP__FUNCTION]); nd; nd = rb_next(nd)) { + for (nd = rb_first(&self->maps[MAP__FUNCTION]); nd; nd = rb_next(nd)) { struct map *map = rb_entry(nd, struct map, rb_node); if (map->dso && strcmp(map->dso->name, name) == 0) @@ -1285,7 +1246,7 @@ static int dsos__set_modules_path_dir(char *dirname) (int)(dot - dent->d_name), dent->d_name); strxfrchar(dso_name, '-', '_'); - map = kernel_maps__find_by_dso_name(dso_name); + map = thread__find_map_by_name(kthread, dso_name); if (map == NULL) continue; @@ -1338,7 +1299,7 @@ static struct map *map__new2(u64 start, struct dso *dso, enum map_type type) return self; } -static int kernel_maps__create_module_maps(void) +static int thread__create_module_maps(struct thread *self) { char *line = NULL; size_t n; @@ -1395,7 +1356,7 @@ static int kernel_maps__create_module_maps(void) dso->has_build_id = true; dso->origin = DSO__ORIG_KMODULE; - kernel_maps__insert(map); + __thread__insert_map(self, map); dsos__add(&dsos__kernel, dso); } @@ -1410,7 +1371,7 @@ out_failure: return -1; } -static int dso__load_vmlinux(struct dso *self, struct map *map, +static int dso__load_vmlinux(struct dso *self, struct map *map, struct thread *thread, const char *vmlinux, symbol_filter_t filter) { int err = -1, fd; @@ -1444,15 +1405,14 @@ static int dso__load_vmlinux(struct dso *self, struct map *map, return -1; dso__set_loaded(self, map->type); - err = dso__load_sym(self, map, self->long_name, fd, filter, 1, 0); - + err = dso__load_sym(self, map, thread, self->long_name, fd, filter, 1, 0); close(fd); return err; } static int dso__load_kernel_sym(struct dso *self, struct map *map, - symbol_filter_t filter) + struct thread *thread, symbol_filter_t filter) { int err; bool is_kallsyms; @@ -1462,8 +1422,8 @@ static int dso__load_kernel_sym(struct dso *self, struct map *map, pr_debug("Looking at the vmlinux_path (%d entries long)\n", vmlinux_path__nr_entries); for (i = 0; i < vmlinux_path__nr_entries; ++i) { - err = dso__load_vmlinux(self, map, vmlinux_path[i], - filter); + err = dso__load_vmlinux(self, map, thread, + vmlinux_path[i], filter); if (err > 0) { pr_debug("Using %s for symbols\n", vmlinux_path[i]); @@ -1478,12 +1438,12 @@ static int dso__load_kernel_sym(struct dso *self, struct map *map, if (is_kallsyms) goto do_kallsyms; - err = dso__load_vmlinux(self, map, self->long_name, filter); + err = dso__load_vmlinux(self, map, thread, self->long_name, filter); if (err <= 0) { pr_info("The file %s cannot be used, " "trying to use /proc/kallsyms...", self->long_name); do_kallsyms: - err = dso__load_kallsyms(self, map, filter); + err = dso__load_kallsyms(self, map, thread, filter); if (err > 0 && !is_kallsyms) dso__set_long_name(self, strdup("[kernel.kallsyms]")); } @@ -1535,8 +1495,11 @@ static void __dsos__fprintf(struct list_head *head, FILE *fp) { struct dso *pos; - list_for_each_entry(pos, head, node) - dso__fprintf(pos, fp); + list_for_each_entry(pos, head, node) { + int i; + for (i = 0; i < MAP__NR_TYPES; ++i) + dso__fprintf(pos, i, fp); + } } void dsos__fprintf(FILE *fp) @@ -1563,10 +1526,10 @@ size_t dsos__fprintf_buildid(FILE *fp) __dsos__fprintf_buildid(&dsos__user, fp)); } -static int kernel_maps__create_kernel_map(const struct symbol_conf *conf) +static int thread__create_kernel_map(struct thread *self, const char *vmlinux) { struct map *kmap; - struct dso *kernel = dso__new(conf->vmlinux_name ?: "[kernel.kallsyms]"); + struct dso *kernel = dso__new(vmlinux ?: "[kernel.kallsyms]"); if (kernel == NULL) return -1; @@ -1588,7 +1551,7 @@ static int kernel_maps__create_kernel_map(const struct symbol_conf *conf) sizeof(kernel->build_id)) == 0) kernel->has_build_id = true; - kernel_maps__insert(kmap); + __thread__insert_map(self, kmap); dsos__add(&dsos__kernel, kernel); dsos__add(&dsos__user, vdso); @@ -1656,32 +1619,28 @@ out_fail: return -1; } -static int kernel_maps__init(const struct symbol_conf *conf) +int symbol__init(struct symbol_conf *conf) { const struct symbol_conf *pconf = conf ?: &symbol_conf__defaults; + elf_version(EV_CURRENT); symbol__priv_size = pconf->priv_size; + thread__init(kthread, 0); if (pconf->try_vmlinux_path && vmlinux_path__init() < 0) return -1; - if (kernel_maps__create_kernel_map(pconf) < 0) { + if (thread__create_kernel_map(kthread, pconf->vmlinux_name) < 0) { vmlinux_path__exit(); return -1; } - if (pconf->use_modules && kernel_maps__create_module_maps() < 0) + if (pconf->use_modules && thread__create_module_maps(kthread) < 0) pr_debug("Failed to load list of modules in use, " "continuing...\n"); /* * Now that we have all the maps created, just set the ->end of them: */ - kernel_maps__fixup_end(); + thread__fixup_maps_end(kthread); return 0; } - -int symbol__init(struct symbol_conf *conf) -{ - elf_version(EV_CURRENT); - return kernel_maps__init(conf); -} diff --git a/tools/perf/util/symbol.h b/tools/perf/util/symbol.h index 8934814d5a6..3f9e4a4d83d 100644 --- a/tools/perf/util/symbol.h +++ b/tools/perf/util/symbol.h @@ -92,7 +92,7 @@ void dsos__fprintf(FILE *fp); size_t dsos__fprintf_buildid(FILE *fp); size_t dso__fprintf_buildid(struct dso *self, FILE *fp); -size_t dso__fprintf(struct dso *self, FILE *fp); +size_t dso__fprintf(struct dso *self, enum map_type type, FILE *fp); char dso__symtab_origin(const struct dso *self); void dso__set_build_id(struct dso *self, void *build_id); diff --git a/tools/perf/util/thread.c b/tools/perf/util/thread.c index 1796625f778..2229f82cd63 100644 --- a/tools/perf/util/thread.c +++ b/tools/perf/util/thread.c @@ -9,17 +9,26 @@ static struct rb_root threads; static struct thread *last_match; +void thread__init(struct thread *self, pid_t pid) +{ + int i; + self->pid = pid; + self->comm = NULL; + for (i = 0; i < MAP__NR_TYPES; ++i) { + self->maps[i] = RB_ROOT; + INIT_LIST_HEAD(&self->removed_maps[i]); + } +} + static struct thread *thread__new(pid_t pid) { struct thread *self = zalloc(sizeof(*self)); if (self != NULL) { - self->pid = pid; + thread__init(self, pid); self->comm = malloc(32); if (self->comm) snprintf(self->comm, 32, ":%d", self->pid); - self->maps = RB_ROOT; - INIT_LIST_HEAD(&self->removed_maps); } return self; @@ -44,24 +53,68 @@ int thread__comm_len(struct thread *self) return self->comm_len; } -static size_t thread__fprintf(struct thread *self, FILE *fp) +static const char *map_type__name[MAP__NR_TYPES] = { + [MAP__FUNCTION] = "Functions", +}; + +static size_t __thread__fprintf_maps(struct thread *self, + enum map_type type, FILE *fp) { + size_t printed = fprintf(fp, "%s:\n", map_type__name[type]); struct rb_node *nd; - struct map *pos; - size_t ret = fprintf(fp, "Thread %d %s\nCurrent maps:\n", - self->pid, self->comm); - for (nd = rb_first(&self->maps); nd; nd = rb_next(nd)) { - pos = rb_entry(nd, struct map, rb_node); - ret += map__fprintf(pos, fp); + for (nd = rb_first(&self->maps[type]); nd; nd = rb_next(nd)) { + struct map *pos = rb_entry(nd, struct map, rb_node); + printed += fprintf(fp, "Map:"); + printed += map__fprintf(pos, fp); + if (verbose > 1) { + printed += dso__fprintf(pos->dso, type, fp); + printed += fprintf(fp, "--\n"); + } } - ret = fprintf(fp, "Removed maps:\n"); + return printed; +} + +size_t thread__fprintf_maps(struct thread *self, FILE *fp) +{ + size_t printed = 0, i; + for (i = 0; i < MAP__NR_TYPES; ++i) + printed += __thread__fprintf_maps(self, i, fp); + return printed; +} - list_for_each_entry(pos, &self->removed_maps, node) - ret += map__fprintf(pos, fp); +static size_t __thread__fprintf_removed_maps(struct thread *self, + enum map_type type, FILE *fp) +{ + struct map *pos; + size_t printed = 0; + + list_for_each_entry(pos, &self->removed_maps[type], node) { + printed += fprintf(fp, "Map:"); + printed += map__fprintf(pos, fp); + if (verbose > 1) { + printed += dso__fprintf(pos->dso, type, fp); + printed += fprintf(fp, "--\n"); + } + } + return printed; +} - return ret; +static size_t thread__fprintf_removed_maps(struct thread *self, FILE *fp) +{ + size_t printed = 0, i; + for (i = 0; i < MAP__NR_TYPES; ++i) + printed += __thread__fprintf_removed_maps(self, i, fp); + return printed; +} + +static size_t thread__fprintf(struct thread *self, FILE *fp) +{ + size_t printed = fprintf(fp, "Thread %d %s\n", self->pid, self->comm); + printed += thread__fprintf_removed_maps(self, fp); + printed += fprintf(fp, "Removed maps:\n"); + return printed + thread__fprintf_removed_maps(self, fp); } struct thread *threads__findnew(pid_t pid) @@ -117,7 +170,8 @@ struct thread *register_idle_thread(void) static void thread__remove_overlappings(struct thread *self, struct map *map) { - struct rb_node *next = rb_first(&self->maps); + struct rb_root *root = &self->maps[map->type]; + struct rb_node *next = rb_first(root); while (next) { struct map *pos = rb_entry(next, struct map, rb_node); @@ -132,13 +186,13 @@ static void thread__remove_overlappings(struct thread *self, struct map *map) map__fprintf(pos, stderr); } - rb_erase(&pos->rb_node, &self->maps); + rb_erase(&pos->rb_node, root); /* * We may have references to this map, for instance in some * hist_entry instances, so just move them to a separate * list. */ - list_add_tail(&pos->node, &self->removed_maps); + list_add_tail(&pos->node, &self->removed_maps[map->type]); } } @@ -185,12 +239,26 @@ struct map *maps__find(struct rb_root *maps, u64 ip) void thread__insert_map(struct thread *self, struct map *map) { thread__remove_overlappings(self, map); - maps__insert(&self->maps, map); + maps__insert(&self->maps[map->type], map); } -int thread__fork(struct thread *self, struct thread *parent) +static int thread__clone_maps(struct thread *self, struct thread *parent, + enum map_type type) { struct rb_node *nd; + for (nd = rb_first(&parent->maps[type]); nd; nd = rb_next(nd)) { + struct map *map = rb_entry(nd, struct map, rb_node); + struct map *new = map__clone(map); + if (new == NULL) + return -ENOMEM; + thread__insert_map(self, new); + } + return 0; +} + +int thread__fork(struct thread *self, struct thread *parent) +{ + int i; if (self->comm) free(self->comm); @@ -198,14 +266,9 @@ int thread__fork(struct thread *self, struct thread *parent) if (!self->comm) return -ENOMEM; - for (nd = rb_first(&parent->maps); nd; nd = rb_next(nd)) { - struct map *map = rb_entry(nd, struct map, rb_node); - struct map *new = map__clone(map); - if (!new) + for (i = 0; i < MAP__NR_TYPES; ++i) + if (thread__clone_maps(self, parent, i) < 0) return -ENOMEM; - thread__insert_map(self, new); - } - return 0; } diff --git a/tools/perf/util/thread.h b/tools/perf/util/thread.h index 54580bb8000..3bdd9b2276f 100644 --- a/tools/perf/util/thread.h +++ b/tools/perf/util/thread.h @@ -7,20 +7,22 @@ struct thread { struct rb_node rb_node; - struct rb_root maps; - struct list_head removed_maps; + struct rb_root maps[MAP__NR_TYPES]; + struct list_head removed_maps[MAP__NR_TYPES]; pid_t pid; char shortname[3]; char *comm; int comm_len; }; +void thread__init(struct thread *self, pid_t pid); int thread__set_comm(struct thread *self, const char *comm); int thread__comm_len(struct thread *self); struct thread *threads__findnew(pid_t pid); struct thread *register_idle_thread(void); void thread__insert_map(struct thread *self, struct map *map); int thread__fork(struct thread *self, struct thread *parent); +size_t thread__fprintf_maps(struct thread *self, FILE *fp); size_t threads__fprintf(FILE *fp); void maps__insert(struct rb_root *maps, struct map *map); @@ -29,9 +31,14 @@ struct map *maps__find(struct rb_root *maps, u64 ip); struct symbol *kernel_maps__find_function(const u64 ip, struct map **mapp, symbol_filter_t filter); -static inline struct map *thread__find_map(struct thread *self, u64 ip) +static inline struct map *thread__find_map(struct thread *self, + enum map_type type, u64 ip) { - return self ? maps__find(&self->maps, ip) : NULL; + return self ? maps__find(&self->maps[type], ip) : NULL; } +static inline void __thread__insert_map(struct thread *self, struct map *map) +{ + maps__insert(&self->maps[map->type], map); +} #endif /* __PERF_THREAD_H */ -- cgit v1.2.3-18-g5258 From 1de8e24520ffdcf2a90c842eed937f59079a2abd Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 27 Nov 2009 16:29:21 -0200 Subject: perf symbols: When not using modules, discard its symbols MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1259346563-12568-10-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/util/symbol.c | 6 +++++- tools/perf/util/thread.h | 1 + 2 files changed, 6 insertions(+), 1 deletion(-) (limited to 'tools/perf') diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index b6a2941e778..b788c2f5d67 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -381,6 +381,9 @@ static int dso__split_kallsyms(struct dso *self, struct map *map, struct thread module = strchr(pos->name, '\t'); if (module) { + if (!thread->use_modules) + goto discard_symbol; + *module++ = '\0'; if (strcmp(self->name, module)) { @@ -420,7 +423,7 @@ static int dso__split_kallsyms(struct dso *self, struct map *map, struct thread } if (filter && filter(curr_map, pos)) { - rb_erase(&pos->rb_node, root); +discard_symbol: rb_erase(&pos->rb_node, root); symbol__delete(pos); } else { if (curr_map != map) { @@ -1635,6 +1638,7 @@ int symbol__init(struct symbol_conf *conf) return -1; } + kthread->use_modules = pconf->use_modules; if (pconf->use_modules && thread__create_module_maps(kthread) < 0) pr_debug("Failed to load list of modules in use, " "continuing...\n"); diff --git a/tools/perf/util/thread.h b/tools/perf/util/thread.h index 3bdd9b2276f..59b0d9b577d 100644 --- a/tools/perf/util/thread.h +++ b/tools/perf/util/thread.h @@ -10,6 +10,7 @@ struct thread { struct rb_root maps[MAP__NR_TYPES]; struct list_head removed_maps[MAP__NR_TYPES]; pid_t pid; + bool use_modules; char shortname[3]; char *comm; int comm_len; -- cgit v1.2.3-18-g5258 From 62daacb51a2bf8480e6f6b3696b03f102fc15eb0 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 27 Nov 2009 16:29:22 -0200 Subject: perf tools: Reorganize event processing routines, lotsa dups killed MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit While implementing event__preprocess_sample, that will do all of the symbol lookup in one convenient function, I noticed that util/process_event.[ch] were not being used at all, then started looking if there were other functions that could be shared and... All those functions really don't need to receive offset + head, the only thing they did was common to all of them, so do it at one place instead. Stats about number of each type of event processed now is done in a central place. Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: John Kacur Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1259346563-12568-11-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/Makefile | 2 -- tools/perf/builtin-annotate.c | 63 ++++++++-------------------------- tools/perf/builtin-kmem.c | 33 +++--------------- tools/perf/builtin-report.c | 72 +++++++++++--------------------------- tools/perf/builtin-sched.c | 42 +++-------------------- tools/perf/builtin-top.c | 19 ----------- tools/perf/builtin-trace.c | 40 +++------------------- tools/perf/util/data_map.c | 56 +++++++++++++++++++++++------- tools/perf/util/data_map.h | 2 +- tools/perf/util/event.c | 74 ++++++++++++++++++++++++++++++++++++++++ tools/perf/util/event.h | 17 +++++++++ tools/perf/util/hist.c | 7 ---- tools/perf/util/process_event.c | 53 ---------------------------- tools/perf/util/process_event.h | 29 ---------------- tools/perf/util/process_events.c | 64 ---------------------------------- tools/perf/util/process_events.h | 35 ------------------- 16 files changed, 183 insertions(+), 425 deletions(-) delete mode 100644 tools/perf/util/process_event.c delete mode 100644 tools/perf/util/process_event.h delete mode 100644 tools/perf/util/process_events.c delete mode 100644 tools/perf/util/process_events.h (limited to 'tools/perf') diff --git a/tools/perf/Makefile b/tools/perf/Makefile index de37d492e10..f1537a94a05 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -369,7 +369,6 @@ LIB_H += util/sort.h LIB_H += util/hist.h LIB_H += util/thread.h LIB_H += util/data_map.h -LIB_H += util/process_events.h LIB_OBJS += util/abspath.o LIB_OBJS += util/alias.o @@ -412,7 +411,6 @@ LIB_OBJS += util/svghelper.o LIB_OBJS += util/sort.o LIB_OBJS += util/hist.o LIB_OBJS += util/data_map.o -LIB_OBJS += util/process_events.o BUILTIN_OBJS += builtin-annotate.o diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c index 3ebd70b1ef9..7d39bd2b19b 100644 --- a/tools/perf/builtin-annotate.c +++ b/tools/perf/builtin-annotate.c @@ -19,12 +19,12 @@ #include "perf.h" #include "util/debug.h" +#include "util/event.h" #include "util/parse-options.h" #include "util/parse-events.h" #include "util/thread.h" #include "util/sort.h" #include "util/hist.h" -#include "util/process_events.h" static char const *input_name = "perf.data"; @@ -136,8 +136,7 @@ static int hist_entry__add(struct thread *thread, struct map *map, return 0; } -static int -process_sample_event(event_t *event, unsigned long offset, unsigned long head) +static int process_sample_event(event_t *event) { char level; u64 ip = event->ip.ip; @@ -145,12 +144,8 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) struct symbol *sym = NULL; struct thread *thread = threads__findnew(event->ip.pid); - dump_printf("%p [%p]: PERF_EVENT (IP, %d): %d: %p\n", - (void *)(offset + head), - (void *)(long)(event->header.size), - event->header.misc, - event->ip.pid, - (void *)(long)ip); + dump_printf("(IP, %d): %d: %p\n", event->header.misc, + event->ip.pid, (void *)(long)ip); if (thread == NULL) { fprintf(stderr, "problem processing %d event, skipping it.\n", @@ -198,46 +193,24 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) "skipping event\n"); return -1; } - total++; return 0; } -static int -process_comm_event(event_t *event, unsigned long offset, unsigned long head) +static int event__process(event_t *self) { - struct thread *thread = threads__findnew(event->comm.pid); - - dump_printf("%p [%p]: PERF_RECORD_COMM: %s:%d\n", - (void *)(offset + head), - (void *)(long)(event->header.size), - event->comm.comm, event->comm.pid); - - if (thread == NULL || - thread__set_comm(thread, event->comm.comm)) { - dump_printf("problem processing PERF_RECORD_COMM, skipping event.\n"); - return -1; - } - total_comm++; - - return 0; -} - -static int -process_event(event_t *event, unsigned long offset, unsigned long head) -{ - switch (event->header.type) { + switch (self->header.type) { case PERF_RECORD_SAMPLE: - return process_sample_event(event, offset, head); + return process_sample_event(self); case PERF_RECORD_MMAP: - return process_mmap_event(event, offset, head); + return event__process_mmap(self); case PERF_RECORD_COMM: - return process_comm_event(event, offset, head); + return event__process_comm(self); case PERF_RECORD_FORK: - return process_task_event(event, offset, head); + return event__process_task(self); /* * We dont process them right now but they are fine: */ @@ -621,15 +594,12 @@ more: (void *)(long)event->header.size, event->header.type); - if (!size || process_event(event, offset, head) < 0) { + if (!size || event__process(event) < 0) { dump_printf("%p [%p]: skipping unknown header type: %d\n", (void *)(offset + head), (void *)(long)(event->header.size), event->header.type); - - total_unknown++; - /* * assume we lost track of the stream, check alignment, and * increment a single u64 in the hope to catch on again 'soon'. @@ -649,14 +619,11 @@ more: rc = EXIT_SUCCESS; close(input); - dump_printf(" IP events: %10ld\n", total); - dump_printf(" mmap events: %10ld\n", total_mmap); - dump_printf(" comm events: %10ld\n", total_comm); - dump_printf(" fork events: %10ld\n", total_fork); - dump_printf(" unknown events: %10ld\n", total_unknown); - if (dump_trace) + if (dump_trace) { + event__print_totals(); return 0; + } if (verbose > 3) threads__fprintf(stdout); @@ -665,7 +632,7 @@ more: dsos__fprintf(stdout); collapse__resort(); - output__resort(total); + output__resort(event__total[0]); find_annotations(); diff --git a/tools/perf/builtin-kmem.c b/tools/perf/builtin-kmem.c index 35722fafc4d..e7294c8fc62 100644 --- a/tools/perf/builtin-kmem.c +++ b/tools/perf/builtin-kmem.c @@ -33,9 +33,6 @@ static bool raw_ip; static char default_sort_order[] = "frag,hit,bytes"; -static char *cwd; -static int cwdlen; - static int *cpunode_map; static int max_cpu_num; @@ -126,25 +123,6 @@ static void setup_cpunode_map(void) } } -static int -process_comm_event(event_t *event, unsigned long offset, unsigned long head) -{ - struct thread *thread = threads__findnew(event->comm.pid); - - dump_printf("%p [%p]: PERF_RECORD_COMM: %s:%d\n", - (void *)(offset + head), - (void *)(long)(event->header.size), - event->comm.comm, event->comm.pid); - - if (thread == NULL || - thread__set_comm(thread, event->comm.comm)) { - dump_printf("problem processing PERF_RECORD_COMM, skipping event.\n"); - return -1; - } - - return 0; -} - static void insert_alloc_stat(unsigned long call_site, unsigned long ptr, int bytes_req, int bytes_alloc, int cpu) { @@ -340,8 +318,7 @@ process_raw_event(event_t *raw_event __used, void *more_data, } } -static int -process_sample_event(event_t *event, unsigned long offset, unsigned long head) +static int process_sample_event(event_t *event) { u64 ip = event->ip.ip; u64 timestamp = -1; @@ -366,9 +343,7 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) more_data += sizeof(u64); } - dump_printf("%p [%p]: PERF_RECORD_SAMPLE (IP, %d): %d/%d: %p period: %Ld\n", - (void *)(offset + head), - (void *)(long)(event->header.size), + dump_printf("(IP, %d): %d/%d: %p period: %Ld\n", event->header.misc, event->ip.pid, event->ip.tid, (void *)(long)ip, @@ -403,7 +378,7 @@ static int sample_type_check(u64 type) static struct perf_file_handler file_handler = { .process_sample_event = process_sample_event, - .process_comm_event = process_comm_event, + .process_comm_event = event__process_comm, .sample_type_check = sample_type_check, }; @@ -413,7 +388,7 @@ static int read_events(void) register_perf_file_handler(&file_handler); return mmap_dispatch_perf_file(&header, input_name, 0, 0, - &cwdlen, &cwd); + &event__cwdlen, &event__cwd); } static double fragmentation(unsigned long n_req, unsigned long n_alloc) diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index 9bd20c2ee3d..01ef35cac5f 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -30,7 +30,6 @@ #include "util/thread.h" #include "util/sort.h" #include "util/hist.h" -#include "util/process_events.h" static char const *input_name = "perf.data"; @@ -655,8 +654,7 @@ static int validate_chain(struct ip_callchain *chain, event_t *event) return 0; } -static int -process_sample_event(event_t *event, unsigned long offset, unsigned long head) +static int process_sample_event(event_t *event) { char level; struct symbol *sym = NULL; @@ -673,9 +671,7 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) more_data += sizeof(u64); } - dump_printf("%p [%p]: PERF_RECORD_SAMPLE (IP, %d): %d/%d: %p period: %Ld\n", - (void *)(offset + head), - (void *)(long)(event->header.size), + dump_printf("(IP, %d): %d/%d: %p period: %Ld\n", event->header.misc, event->ip.pid, event->ip.tid, (void *)(long)ip, @@ -743,47 +739,27 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) return -1; } - total += period; + event__stats.total += period; return 0; } -static int -process_comm_event(event_t *event, unsigned long offset, unsigned long head) +static int process_comm_event(event_t *event) { struct thread *thread = threads__findnew(event->comm.pid); - dump_printf("%p [%p]: PERF_RECORD_COMM: %s:%d\n", - (void *)(offset + head), - (void *)(long)(event->header.size), - event->comm.comm, event->comm.pid); + dump_printf(": %s:%d\n", event->comm.comm, event->comm.pid); if (thread == NULL || thread__set_comm_adjust(thread, event->comm.comm)) { dump_printf("problem processing PERF_RECORD_COMM, skipping event.\n"); return -1; } - total_comm++; - - return 0; -} - -static int -process_lost_event(event_t *event, unsigned long offset, unsigned long head) -{ - dump_printf("%p [%p]: PERF_RECORD_LOST: id:%Ld: lost:%Ld\n", - (void *)(offset + head), - (void *)(long)(event->header.size), - event->lost.id, - event->lost.lost); - - total_lost += event->lost.lost; return 0; } -static int -process_read_event(event_t *event, unsigned long offset, unsigned long head) +static int process_read_event(event_t *event) { struct perf_event_attr *attr; @@ -799,14 +775,9 @@ process_read_event(event_t *event, unsigned long offset, unsigned long head) event->read.value); } - dump_printf("%p [%p]: PERF_RECORD_READ: %d %d %s %Lu\n", - (void *)(offset + head), - (void *)(long)(event->header.size), - event->read.pid, - event->read.tid, - attr ? __event_name(attr->type, attr->config) - : "FAIL", - event->read.value); + dump_printf(": %d %d %s %Lu\n", event->read.pid, event->read.tid, + attr ? __event_name(attr->type, attr->config) : "FAIL", + event->read.value); return 0; } @@ -842,11 +813,11 @@ static int sample_type_check(u64 type) static struct perf_file_handler file_handler = { .process_sample_event = process_sample_event, - .process_mmap_event = process_mmap_event, + .process_mmap_event = event__process_mmap, .process_comm_event = process_comm_event, - .process_exit_event = process_task_event, - .process_fork_event = process_task_event, - .process_lost_event = process_lost_event, + .process_exit_event = event__process_task, + .process_fork_event = event__process_task, + .process_lost_event = event__process_lost, .process_read_event = process_read_event, .sample_type_check = sample_type_check, }; @@ -866,19 +837,14 @@ static int __cmd_report(void) register_perf_file_handler(&file_handler); ret = mmap_dispatch_perf_file(&header, input_name, force, - full_paths, &cwdlen, &cwd); + full_paths, &event__cwdlen, &event__cwd); if (ret) return ret; - dump_printf(" IP events: %10ld\n", total); - dump_printf(" mmap events: %10ld\n", total_mmap); - dump_printf(" comm events: %10ld\n", total_comm); - dump_printf(" fork events: %10ld\n", total_fork); - dump_printf(" lost events: %10ld\n", total_lost); - dump_printf(" unknown events: %10ld\n", file_handler.total_unknown); - - if (dump_trace) + if (dump_trace) { + event__print_totals(); return 0; + } if (verbose > 3) threads__fprintf(stdout); @@ -887,8 +853,8 @@ static int __cmd_report(void) dsos__fprintf(stdout); collapse__resort(); - output__resort(total); - output__fprintf(stdout, total); + output__resort(event__stats.total); + output__fprintf(stdout, event__stats.total); if (show_threads) perf_read_values_destroy(&show_threads_values); diff --git a/tools/perf/builtin-sched.c b/tools/perf/builtin-sched.c index 19eb708a706..26b782f26ee 100644 --- a/tools/perf/builtin-sched.c +++ b/tools/perf/builtin-sched.c @@ -22,8 +22,6 @@ static char const *input_name = "perf.data"; -static unsigned long total_comm = 0; - static struct perf_header *header; static u64 sample_type; @@ -32,9 +30,6 @@ static char *sort_order = default_sort_order; static int profile_cpu = -1; -static char *cwd; -static int cwdlen; - #define PR_SET_NAME 15 /* Set process name */ #define MAX_CPUS 4096 @@ -633,27 +628,6 @@ static void test_calibrations(void) printf("the sleep test took %Ld nsecs\n", T1-T0); } -static int -process_comm_event(event_t *event, unsigned long offset, unsigned long head) -{ - struct thread *thread = threads__findnew(event->comm.tid); - - dump_printf("%p [%p]: perf_event_comm: %s:%d\n", - (void *)(offset + head), - (void *)(long)(event->header.size), - event->comm.comm, event->comm.pid); - - if (thread == NULL || - thread__set_comm(thread, event->comm.comm)) { - dump_printf("problem processing perf_event_comm, skipping event.\n"); - return -1; - } - total_comm++; - - return 0; -} - - struct raw_event_sample { u32 size; char data[0]; @@ -1622,8 +1596,7 @@ process_raw_event(event_t *raw_event __used, void *more_data, process_sched_migrate_task_event(raw, event, cpu, timestamp, thread); } -static int -process_sample_event(event_t *event, unsigned long offset, unsigned long head) +static int process_sample_event(event_t *event) { struct thread *thread; u64 ip = event->ip.ip; @@ -1653,9 +1626,7 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) more_data += sizeof(u64); } - dump_printf("%p [%p]: PERF_RECORD_SAMPLE (IP, %d): %d/%d: %p period: %Ld\n", - (void *)(offset + head), - (void *)(long)(event->header.size), + dump_printf("(IP, %d): %d/%d: %p period: %Ld\n", event->header.misc, event->ip.pid, event->ip.tid, (void *)(long)ip, @@ -1677,10 +1648,7 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) return 0; } -static int -process_lost_event(event_t *event __used, - unsigned long offset __used, - unsigned long head __used) +static int process_lost_event(event_t *event __used) { nr_lost_chunks++; nr_lost_events += event->lost.lost; @@ -1704,7 +1672,7 @@ static int sample_type_check(u64 type) static struct perf_file_handler file_handler = { .process_sample_event = process_sample_event, - .process_comm_event = process_comm_event, + .process_comm_event = event__process_comm, .process_lost_event = process_lost_event, .sample_type_check = sample_type_check, }; @@ -1715,7 +1683,7 @@ static int read_events(void) register_perf_file_handler(&file_handler); return mmap_dispatch_perf_file(&header, input_name, 0, 0, - &cwdlen, &cwd); + &event__cwdlen, &event__cwd); } static void print_bad_events(void) diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index bf6730c7603..7a3c0c7aad3 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -991,25 +991,6 @@ static void event__process_sample(const event_t *self, int counter) } } -static void event__process_mmap(event_t *self) -{ - struct thread *thread = threads__findnew(self->mmap.pid); - - if (thread != NULL) { - struct map *map = map__new(&self->mmap, MAP__FUNCTION, NULL, 0); - if (map != NULL) - thread__insert_map(thread, map); - } -} - -static void event__process_comm(event_t *self) -{ - struct thread *thread = threads__findnew(self->comm.pid); - - if (thread != NULL) - thread__set_comm(thread, self->comm.comm); -} - static int event__process(event_t *event) { switch (event->header.type) { diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-trace.c index 75972fd073d..a7750256c40 100644 --- a/tools/perf/builtin-trace.c +++ b/tools/perf/builtin-trace.c @@ -16,38 +16,10 @@ static char const *input_name = "perf.data"; -static unsigned long total = 0; -static unsigned long total_comm = 0; - static struct perf_header *header; static u64 sample_type; -static char *cwd; -static int cwdlen; - - -static int -process_comm_event(event_t *event, unsigned long offset, unsigned long head) -{ - struct thread *thread = threads__findnew(event->comm.pid); - - dump_printf("%p [%p]: PERF_RECORD_COMM: %s:%d\n", - (void *)(offset + head), - (void *)(long)(event->header.size), - event->comm.comm, event->comm.pid); - - if (thread == NULL || - thread__set_comm(thread, event->comm.comm)) { - dump_printf("problem processing PERF_RECORD_COMM, skipping event.\n"); - return -1; - } - total_comm++; - - return 0; -} - -static int -process_sample_event(event_t *event, unsigned long offset, unsigned long head) +static int process_sample_event(event_t *event) { u64 ip = event->ip.ip; u64 timestamp = -1; @@ -72,9 +44,7 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) more_data += sizeof(u64); } - dump_printf("%p [%p]: PERF_RECORD_SAMPLE (IP, %d): %d/%d: %p period: %Ld\n", - (void *)(offset + head), - (void *)(long)(event->header.size), + dump_printf("(IP, %d): %d/%d: %p period: %Ld\n", event->header.misc, event->ip.pid, event->ip.tid, (void *)(long)ip, @@ -101,7 +71,7 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) */ print_event(cpu, raw->data, raw->size, timestamp, thread->comm); } - total += period; + event__stats.total += period; return 0; } @@ -122,7 +92,7 @@ static int sample_type_check(u64 type) static struct perf_file_handler file_handler = { .process_sample_event = process_sample_event, - .process_comm_event = process_comm_event, + .process_comm_event = event__process_comm, .sample_type_check = sample_type_check, }; @@ -132,7 +102,7 @@ static int __cmd_trace(void) register_perf_file_handler(&file_handler); return mmap_dispatch_perf_file(&header, input_name, - 0, 0, &cwdlen, &cwd); + 0, 0, &event__cwdlen, &event__cwd); } static const char * const annotate_usage[] = { diff --git a/tools/perf/util/data_map.c b/tools/perf/util/data_map.c index b238462b898..ca0bedf637c 100644 --- a/tools/perf/util/data_map.c +++ b/tools/perf/util/data_map.c @@ -8,11 +8,9 @@ static struct perf_file_handler *curr_handler; static unsigned long mmap_window = 32; static char __cwd[PATH_MAX]; -static int -process_event_stub(event_t *event __used, - unsigned long offset __used, - unsigned long head __used) +static int process_event_stub(event_t *event __used) { + dump_printf(": unhandled!\n"); return 0; } @@ -40,30 +38,62 @@ void register_perf_file_handler(struct perf_file_handler *handler) curr_handler = handler; } +static const char *event__name[] = { + [0] = "TOTAL", + [PERF_RECORD_MMAP] = "MMAP", + [PERF_RECORD_LOST] = "LOST", + [PERF_RECORD_COMM] = "COMM", + [PERF_RECORD_EXIT] = "EXIT", + [PERF_RECORD_THROTTLE] = "THROTTLE", + [PERF_RECORD_UNTHROTTLE] = "UNTHROTTLE", + [PERF_RECORD_FORK] = "FORK", + [PERF_RECORD_READ] = "READ", + [PERF_RECORD_SAMPLE] = "SAMPLE", +}; + +unsigned long event__total[PERF_RECORD_MAX]; + +void event__print_totals(void) +{ + int i; + for (i = 0; i < PERF_RECORD_MAX; ++i) + pr_info("%10s events: %10ld\n", + event__name[i], event__total[i]); +} + static int process_event(event_t *event, unsigned long offset, unsigned long head) { trace_event(event); + if (event->header.type < PERF_RECORD_MAX) { + dump_printf("%p [%p]: PERF_RECORD_%s", + (void *)(offset + head), + (void *)(long)(event->header.size), + event__name[event->header.type]); + ++event__total[0]; + ++event__total[event->header.type]; + } + switch (event->header.type) { case PERF_RECORD_SAMPLE: - return curr_handler->process_sample_event(event, offset, head); + return curr_handler->process_sample_event(event); case PERF_RECORD_MMAP: - return curr_handler->process_mmap_event(event, offset, head); + return curr_handler->process_mmap_event(event); case PERF_RECORD_COMM: - return curr_handler->process_comm_event(event, offset, head); + return curr_handler->process_comm_event(event); case PERF_RECORD_FORK: - return curr_handler->process_fork_event(event, offset, head); + return curr_handler->process_fork_event(event); case PERF_RECORD_EXIT: - return curr_handler->process_exit_event(event, offset, head); + return curr_handler->process_exit_event(event); case PERF_RECORD_LOST: - return curr_handler->process_lost_event(event, offset, head); + return curr_handler->process_lost_event(event); case PERF_RECORD_READ: - return curr_handler->process_read_event(event, offset, head); + return curr_handler->process_read_event(event); case PERF_RECORD_THROTTLE: - return curr_handler->process_throttle_event(event, offset, head); + return curr_handler->process_throttle_event(event); case PERF_RECORD_UNTHROTTLE: - return curr_handler->process_unthrottle_event(event, offset, head); + return curr_handler->process_unthrottle_event(event); default: curr_handler->total_unknown++; return -1; diff --git a/tools/perf/util/data_map.h b/tools/perf/util/data_map.h index ae036ecd762..3180ff7e363 100644 --- a/tools/perf/util/data_map.h +++ b/tools/perf/util/data_map.h @@ -4,7 +4,7 @@ #include "event.h" #include "header.h" -typedef int (*event_type_handler_t)(event_t *, unsigned long, unsigned long); +typedef int (*event_type_handler_t)(event_t *); struct perf_file_handler { event_type_handler_t process_sample_event; diff --git a/tools/perf/util/event.c b/tools/perf/util/event.c index 1dae7e3b400..70b4aa03b47 100644 --- a/tools/perf/util/event.c +++ b/tools/perf/util/event.c @@ -2,6 +2,7 @@ #include "event.h" #include "debug.h" #include "string.h" +#include "thread.h" static pid_t event__synthesize_comm(pid_t pid, int full, int (*process)(event_t *event)) @@ -175,3 +176,76 @@ void event__synthesize_threads(int (*process)(event_t *event)) closedir(proc); } + +char *event__cwd; +int event__cwdlen; + +struct events_stats event__stats; + +int event__process_comm(event_t *self) +{ + struct thread *thread = threads__findnew(self->comm.pid); + + dump_printf("PERF_RECORD_COMM: %s:%d\n", + self->comm.comm, self->comm.pid); + + if (thread == NULL || thread__set_comm(thread, self->comm.comm)) { + dump_printf("problem processing PERF_RECORD_COMM, skipping event.\n"); + return -1; + } + + return 0; +} + +int event__process_lost(event_t *self) +{ + dump_printf(": id:%Ld: lost:%Ld\n", self->lost.id, self->lost.lost); + event__stats.lost += self->lost.lost; + return 0; +} + +int event__process_mmap(event_t *self) +{ + struct thread *thread = threads__findnew(self->mmap.pid); + struct map *map = map__new(&self->mmap, MAP__FUNCTION, + event__cwd, event__cwdlen); + + dump_printf(" %d/%d: [%p(%p) @ %p]: %s\n", + self->mmap.pid, self->mmap.tid, + (void *)(long)self->mmap.start, + (void *)(long)self->mmap.len, + (void *)(long)self->mmap.pgoff, + self->mmap.filename); + + if (thread == NULL || map == NULL) + dump_printf("problem processing PERF_RECORD_MMAP, skipping event.\n"); + else + thread__insert_map(thread, map); + + return 0; +} + +int event__process_task(event_t *self) +{ + struct thread *thread = threads__findnew(self->fork.pid); + struct thread *parent = threads__findnew(self->fork.ppid); + + dump_printf("(%d:%d):(%d:%d)\n", self->fork.pid, self->fork.tid, + self->fork.ppid, self->fork.ptid); + /* + * A thread clone will have the same PID for both parent and child. + */ + if (thread == parent) + return 0; + + if (self->header.type == PERF_RECORD_EXIT) + return 0; + + if (thread == NULL || parent == NULL || + thread__fork(thread, parent) < 0) { + dump_printf("problem processing PERF_RECORD_FORK, skipping event.\n"); + return -1; + } + + return 0; +} diff --git a/tools/perf/util/event.h b/tools/perf/util/event.h index 3ae3c964c90..13c12c75f97 100644 --- a/tools/perf/util/event.h +++ b/tools/perf/util/event.h @@ -80,6 +80,13 @@ typedef union event_union { struct sample_event sample; } event_t; +struct events_stats { + unsigned long total; + unsigned long lost; +}; + +void event__print_totals(void); + enum map_type { MAP__FUNCTION = 0, @@ -135,4 +142,14 @@ void map__fixup_end(struct map *self); int event__synthesize_thread(pid_t pid, int (*process)(event_t *event)); void event__synthesize_threads(int (*process)(event_t *event)); +extern char *event__cwd; +extern int event__cwdlen; +extern struct events_stats event__stats; +extern unsigned long event__total[PERF_RECORD_MAX]; + +int event__process_comm(event_t *self); +int event__process_lost(event_t *self); +int event__process_mmap(event_t *self); +int event__process_task(event_t *self); + #endif /* __PERF_RECORD_H */ diff --git a/tools/perf/util/hist.c b/tools/perf/util/hist.c index 7393a02fd8d..f26cd9ba00f 100644 --- a/tools/perf/util/hist.c +++ b/tools/perf/util/hist.c @@ -10,13 +10,6 @@ struct callchain_param callchain_param = { .min_percent = 0.5 }; -unsigned long total; -unsigned long total_mmap; -unsigned long total_comm; -unsigned long total_fork; -unsigned long total_unknown; -unsigned long total_lost; - /* * histogram, sorted on item, collects counts */ diff --git a/tools/perf/util/process_event.c b/tools/perf/util/process_event.c deleted file mode 100644 index a970789581a..00000000000 --- a/tools/perf/util/process_event.c +++ /dev/null @@ -1,53 +0,0 @@ -#include "process_event.h" - -char *cwd; -int cwdlen; - -int -process_mmap_event(event_t *event, unsigned long offset, unsigned long head) -{ - struct map *map = map__new(&event->mmap, cwd, cwdlen); - struct thread *thread = threads__findnew(event->mmap.pid); - - dump_printf("%p [%p]: PERF_RECORD_MMAP %d/%d: [%p(%p) @ %p]: %s\n", - (void *)(offset + head), - (void *)(long)(event->header.size), - event->mmap.pid, - event->mmap.tid, - (void *)(long)event->mmap.start, - (void *)(long)event->mmap.len, - (void *)(long)event->mmap.pgoff, - event->mmap.filename); - - if (thread == NULL || map == NULL) { - dump_printf("problem processing PERF_RECORD_MMAP, skipping event.\n"); - return 0; - } - - thread__insert_map(thread, map); - total_mmap++; - - return 0; - -} - -int -process_comm_event(event_t *event, unsigned long offset, unsigned long head) -{ - struct thread *thread = threads__findnew(event->comm.pid); - - dump_printf("%p [%p]: PERF_RECORD_COMM: %s:%d\n", - (void *)(offset + head), - (void *)(long)(event->header.size), - event->comm.comm, event->comm.pid); - - if (thread == NULL || - thread__set_comm_adjust(thread, event->comm.comm)) { - dump_printf("problem processing PERF_RECORD_COMM, skipping event.\n"); - return -1; - } - total_comm++; - - return 0; -} - diff --git a/tools/perf/util/process_event.h b/tools/perf/util/process_event.h deleted file mode 100644 index 6f68c69736c..00000000000 --- a/tools/perf/util/process_event.h +++ /dev/null @@ -1,29 +0,0 @@ -#ifndef __PROCESS_EVENT_H -#define __PROCESS_EVENT_H - -#include "../builtin.h" -#include "util.h" - -#include "color.h" -#include -#include "cache.h" -#include -#include "symbol.h" -#include "string.h" - -#include "../perf.h" -#include "debug.h" - -#include "parse-options.h" -#include "parse-events.h" - -#include "thread.h" -#include "sort.h" -#include "hist.h" - -extern char *cwd; -extern int cwdlen; -extern int process_mmap_event(event_t *, unsigned long, unsigned long); -extern int process_comm_event(event_t *, unsigned long , unsigned long); - -#endif /* __PROCESS_H */ diff --git a/tools/perf/util/process_events.c b/tools/perf/util/process_events.c deleted file mode 100644 index 53778684641..00000000000 --- a/tools/perf/util/process_events.c +++ /dev/null @@ -1,64 +0,0 @@ -#include "process_events.h" - -char *cwd; -int cwdlen; - -int -process_mmap_event(event_t *event, unsigned long offset, unsigned long head) -{ - struct map *map = map__new(&event->mmap, MAP__FUNCTION, cwd, cwdlen); - struct thread *thread = threads__findnew(event->mmap.pid); - - dump_printf("%p [%p]: PERF_RECORD_MMAP %d/%d: [%p(%p) @ %p]: %s\n", - (void *)(offset + head), - (void *)(long)(event->header.size), - event->mmap.pid, - event->mmap.tid, - (void *)(long)event->mmap.start, - (void *)(long)event->mmap.len, - (void *)(long)event->mmap.pgoff, - event->mmap.filename); - - if (thread == NULL || map == NULL) { - dump_printf("problem processing PERF_RECORD_MMAP, skipping event.\n"); - return 0; - } - - thread__insert_map(thread, map); - total_mmap++; - - return 0; -} - -int -process_task_event(event_t *event, unsigned long offset, unsigned long head) -{ - struct thread *thread = threads__findnew(event->fork.pid); - struct thread *parent = threads__findnew(event->fork.ppid); - - dump_printf("%p [%p]: PERF_RECORD_%s: (%d:%d):(%d:%d)\n", - (void *)(offset + head), - (void *)(long)(event->header.size), - event->header.type == PERF_RECORD_FORK ? "FORK" : "EXIT", - event->fork.pid, event->fork.tid, - event->fork.ppid, event->fork.ptid); - - /* - * A thread clone will have the same PID for both - * parent and child. - */ - if (thread == parent) - return 0; - - if (event->header.type == PERF_RECORD_EXIT) - return 0; - - if (!thread || !parent || thread__fork(thread, parent)) { - dump_printf("problem processing PERF_RECORD_FORK, skipping event.\n"); - return -1; - } - total_fork++; - - return 0; -} - diff --git a/tools/perf/util/process_events.h b/tools/perf/util/process_events.h deleted file mode 100644 index 73d092f8328..00000000000 --- a/tools/perf/util/process_events.h +++ /dev/null @@ -1,35 +0,0 @@ -#ifndef __PROCESS_EVENTS_H -#define __PROCESS_EVENTS_H - -#include "../builtin.h" - -#include "util.h" -#include "color.h" -#include -#include "cache.h" -#include -#include "symbol.h" -#include "string.h" -#include "callchain.h" -#include "strlist.h" -#include "values.h" - -#include "../perf.h" -#include "debug.h" -#include "header.h" - -#include "parse-options.h" -#include "parse-events.h" - -#include "data_map.h" -#include "thread.h" -#include "sort.h" -#include "hist.h" - -extern char *cwd; -extern int cwdlen; - -extern int process_mmap_event(event_t *, unsigned long , unsigned long); -extern int process_task_event(event_t *, unsigned long, unsigned long); - -#endif /* __PROCESS_EVENTS_H */ -- cgit v1.2.3-18-g5258 From 1ed091c45ae33b2179d387573c3fe3f3b4adf60a Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 27 Nov 2009 16:29:23 -0200 Subject: perf tools: Consolidate symbol resolving across all tools MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Now we have a very high level routine for simple tools to process IP sample events: int event__preprocess_sample(const event_t *self, struct addr_location *al, symbol_filter_t filter) It receives the event itself and will insert new threads in the global threads list and resolve the map and symbol, filling all this info into the new addr_location struct, so that tools like annotate and report can further process the event by creating hist_entries in their specific way (with or without callgraphs, etc). It in turn uses the new next layer function: void thread__find_addr_location(struct thread *self, u8 cpumode, enum map_type type, u64 addr, struct addr_location *al, symbol_filter_t filter) This one will, given a thread (userspace or the kernel kthread one), will find the given type (MAP__FUNCTION now, MAP__VARIABLE too in the near future) at the given cpumode, taking vdsos into account (userspace hit, but kernel symbol) and will fill all these details in the addr_location given. Tools that need a more compact API for plain function resolution, like 'kmem', can use this other one: struct symbol *thread__find_function(struct thread *self, u64 addr, symbol_filter_t filter) So, to resolve a kernel symbol, that is all the 'kmem' tool needs, its just a matter of calling: sym = thread__find_function(kthread, addr, NULL); The 'filter' parameter is needed because we do lazy parsing/loading of ELF symtabs or /proc/kallsyms. With this we remove more code duplication all around, which is always good, huh? :-) Signed-off-by: Arnaldo Carvalho de Melo Cc: Frédéric Weisbecker Cc: John Kacur Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras LKML-Reference: <1259346563-12568-12-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar --- tools/perf/builtin-annotate.c | 55 +++-------------- tools/perf/builtin-kmem.c | 2 +- tools/perf/builtin-report.c | 135 ++++++++++++------------------------------ tools/perf/builtin-top.c | 44 +++----------- tools/perf/util/event.c | 62 +++++++++++++++++++ tools/perf/util/event.h | 4 ++ tools/perf/util/hist.c | 15 +++-- tools/perf/util/hist.h | 6 +- tools/perf/util/symbol.c | 26 +------- tools/perf/util/symbol.h | 10 ++++ tools/perf/util/thread.c | 12 ++++ tools/perf/util/thread.h | 23 +++++-- 12 files changed, 172 insertions(+), 222 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c index 7d39bd2b19b..7f85c6e159a 100644 --- a/tools/perf/builtin-annotate.c +++ b/tools/perf/builtin-annotate.c @@ -124,71 +124,30 @@ static void hist_hit(struct hist_entry *he, u64 ip) h->ip[offset]); } -static int hist_entry__add(struct thread *thread, struct map *map, - struct symbol *sym, u64 ip, u64 count, char level) +static int hist_entry__add(struct addr_location *al, u64 count) { bool hit; - struct hist_entry *he = __hist_entry__add(thread, map, sym, NULL, ip, - count, level, &hit); + struct hist_entry *he = __hist_entry__add(al, NULL, count, &hit); if (he == NULL) return -ENOMEM; - hist_hit(he, ip); + hist_hit(he, al->addr); return 0; } static int process_sample_event(event_t *event) { - char level; - u64 ip = event->ip.ip; - struct map *map = NULL; - struct symbol *sym = NULL; - struct thread *thread = threads__findnew(event->ip.pid); + struct addr_location al; dump_printf("(IP, %d): %d: %p\n", event->header.misc, - event->ip.pid, (void *)(long)ip); + event->ip.pid, (void *)(long)event->ip.ip); - if (thread == NULL) { + if (event__preprocess_sample(event, &al, symbol_filter) < 0) { fprintf(stderr, "problem processing %d event, skipping it.\n", event->header.type); return -1; } - dump_printf(" ... thread: %s:%d\n", thread->comm, thread->pid); - - if (event->header.misc & PERF_RECORD_MISC_KERNEL) { - level = 'k'; - sym = kernel_maps__find_function(ip, &map, symbol_filter); - dump_printf(" ...... dso: %s\n", - map ? map->dso->long_name : ""); - } else if (event->header.misc & PERF_RECORD_MISC_USER) { - level = '.'; - map = thread__find_map(thread, MAP__FUNCTION, ip); - if (map != NULL) { - ip = map->map_ip(map, ip); - sym = map__find_symbol(map, ip, symbol_filter); - } else { - /* - * If this is outside of all known maps, - * and is a negative address, try to look it - * up in the kernel dso, as it might be a - * vsyscall or vdso (which executes in user-mode). - * - * XXX This is nasty, we should have a symbol list in - * the "[vdso]" dso, but for now lets use the old - * trick of looking in the whole kernel symbol list. - */ - if ((long long)ip < 0) - sym = kernel_maps__find_function(ip, &map, - symbol_filter); - } - dump_printf(" ...... dso: %s\n", - map ? map->dso->long_name : ""); - } else { - level = 'H'; - dump_printf(" ...... dso: [hypervisor]\n"); - } - - if (hist_entry__add(thread, map, sym, ip, 1, level)) { + if (hist_entry__add(&al, 1)) { fprintf(stderr, "problem incrementing symbol count, " "skipping event\n"); return -1; diff --git a/tools/perf/builtin-kmem.c b/tools/perf/builtin-kmem.c index e7294c8fc62..047fef74bd5 100644 --- a/tools/perf/builtin-kmem.c +++ b/tools/perf/builtin-kmem.c @@ -420,7 +420,7 @@ static void __print_result(struct rb_root *root, int n_lines, int is_caller) if (is_caller) { addr = data->call_site; if (!raw_ip) - sym = kernel_maps__find_function(addr, NULL, NULL); + sym = thread__find_function(kthread, addr, NULL); } else addr = data->ptr; diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index 01ef35cac5f..383c4ab4f9a 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -408,55 +408,6 @@ static int thread__set_comm_adjust(struct thread *self, const char *comm) return 0; } - -static struct symbol * -resolve_symbol(struct thread *thread, struct map **mapp, u64 *ipp) -{ - struct map *map = mapp ? *mapp : NULL; - u64 ip = *ipp; - - if (map) - goto got_map; - - if (!thread) - return NULL; - - map = thread__find_map(thread, MAP__FUNCTION, ip); - if (map != NULL) { - /* - * We have to do this here as we may have a dso - * with no symbol hit that has a name longer than - * the ones with symbols sampled. - */ - if (!sort_dso.elide && !map->dso->slen_calculated) - dso__calc_col_width(map->dso); - - if (mapp) - *mapp = map; -got_map: - ip = map->map_ip(map, ip); - } else { - /* - * If this is outside of all known maps, - * and is a negative address, try to look it - * up in the kernel dso, as it might be a - * vsyscall or vdso (which executes in user-mode). - * - * XXX This is nasty, we should have a symbol list in - * the "[vdso]" dso, but for now lets use the old - * trick of looking in the whole kernel symbol list. - */ - if ((long long)ip < 0) - return kernel_maps__find_function(ip, mapp, NULL); - } - dump_printf(" ...... dso: %s\n", - map ? map->dso->long_name : ""); - dump_printf(" ...... map: %Lx -> %Lx\n", *ipp, ip); - *ipp = ip; - - return map ? map__find_symbol(map, ip, NULL) : NULL; -} - static int call__match(struct symbol *sym) { if (sym->name && !regexec(&parent_regex, sym->name, 0, NULL, 0)) @@ -469,7 +420,7 @@ static struct symbol **resolve_callchain(struct thread *thread, struct ip_callchain *chain, struct symbol **parent) { - u64 context = PERF_CONTEXT_MAX; + u8 cpumode = PERF_RECORD_MISC_USER; struct symbol **syms = NULL; unsigned int i; @@ -483,30 +434,31 @@ static struct symbol **resolve_callchain(struct thread *thread, for (i = 0; i < chain->nr; i++) { u64 ip = chain->ips[i]; - struct symbol *sym = NULL; + struct addr_location al; if (ip >= PERF_CONTEXT_MAX) { - context = ip; + switch (ip) { + case PERF_CONTEXT_HV: + cpumode = PERF_RECORD_MISC_HYPERVISOR; break; + case PERF_CONTEXT_KERNEL: + cpumode = PERF_RECORD_MISC_KERNEL; break; + case PERF_CONTEXT_USER: + cpumode = PERF_RECORD_MISC_USER; break; + default: + break; + } continue; } - switch (context) { - case PERF_CONTEXT_HV: - break; - case PERF_CONTEXT_KERNEL: - sym = kernel_maps__find_function(ip, NULL, NULL); - break; - default: - sym = resolve_symbol(thread, NULL, &ip); - break; - } - - if (sym) { - if (sort__has_parent && !*parent && call__match(sym)) - *parent = sym; + thread__find_addr_location(thread, cpumode, MAP__FUNCTION, + ip, &al, NULL); + if (al.sym != NULL) { + if (sort__has_parent && !*parent && + call__match(al.sym)) + *parent = al.sym; if (!callchain) break; - syms[i] = sym; + syms[i] = al.sym; } } @@ -517,20 +469,17 @@ static struct symbol **resolve_callchain(struct thread *thread, * collect histogram counts */ -static int -hist_entry__add(struct thread *thread, struct map *map, - struct symbol *sym, u64 ip, struct ip_callchain *chain, - char level, u64 count) +static int hist_entry__add(struct addr_location *al, + struct ip_callchain *chain, u64 count) { struct symbol **syms = NULL, *parent = NULL; bool hit; struct hist_entry *he; if ((sort__has_parent || callchain) && chain) - syms = resolve_callchain(thread, chain, &parent); + syms = resolve_callchain(al->thread, chain, &parent); - he = __hist_entry__add(thread, map, sym, parent, - ip, count, level, &hit); + he = __hist_entry__add(al, parent, count, &hit); if (he == NULL) return -ENOMEM; @@ -656,14 +605,12 @@ static int validate_chain(struct ip_callchain *chain, event_t *event) static int process_sample_event(event_t *event) { - char level; - struct symbol *sym = NULL; u64 ip = event->ip.ip; u64 period = 1; - struct map *map = NULL; void *more_data = event->ip.__more_data; struct ip_callchain *chain = NULL; int cpumode; + struct addr_location al; struct thread *thread = threads__findnew(event->ip.pid); if (sample_type & PERF_SAMPLE_PERIOD) { @@ -709,32 +656,26 @@ static int process_sample_event(event_t *event) cpumode = event->header.misc & PERF_RECORD_MISC_CPUMODE_MASK; - if (cpumode == PERF_RECORD_MISC_KERNEL) { - level = 'k'; - sym = kernel_maps__find_function(ip, &map, NULL); - dump_printf(" ...... dso: %s\n", - map ? map->dso->long_name : ""); - } else if (cpumode == PERF_RECORD_MISC_USER) { - level = '.'; - sym = resolve_symbol(thread, &map, &ip); - - } else { - level = 'H'; - dump_printf(" ...... dso: [hypervisor]\n"); - } + thread__find_addr_location(thread, cpumode, + MAP__FUNCTION, ip, &al, NULL); + /* + * We have to do this here as we may have a dso with no symbol hit that + * has a name longer than the ones with symbols sampled. + */ + if (al.map && !sort_dso.elide && !al.map->dso->slen_calculated) + dso__calc_col_width(al.map->dso); if (dso_list && - (!map || !map->dso || - !(strlist__has_entry(dso_list, map->dso->short_name) || - (map->dso->short_name != map->dso->long_name && - strlist__has_entry(dso_list, map->dso->long_name))))) + (!al.map || !al.map->dso || + !(strlist__has_entry(dso_list, al.map->dso->short_name) || + (al.map->dso->short_name != al.map->dso->long_name && + strlist__has_entry(dso_list, al.map->dso->long_name))))) return 0; - if (sym_list && sym && !strlist__has_entry(sym_list, sym->name)) + if (sym_list && al.sym && !strlist__has_entry(sym_list, al.sym->name)) return 0; - if (hist_entry__add(thread, map, sym, ip, - chain, level, period)) { + if (hist_entry__add(&al, chain, period)) { pr_debug("problem incrementing symbol count, skipping event\n"); return -1; } diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index 7a3c0c7aad3..e0a374d0e43 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -929,55 +929,28 @@ static int symbol_filter(struct map *map, struct symbol *sym) static void event__process_sample(const event_t *self, int counter) { u64 ip = self->ip.ip; - struct map *map; struct sym_entry *syme; - struct symbol *sym; + struct addr_location al; u8 origin = self->header.misc & PERF_RECORD_MISC_CPUMODE_MASK; switch (origin) { - case PERF_RECORD_MISC_USER: { - struct thread *thread; - + case PERF_RECORD_MISC_USER: if (hide_user_symbols) return; - - thread = threads__findnew(self->ip.pid); - if (thread == NULL) - return; - - map = thread__find_map(thread, MAP__FUNCTION, ip); - if (map != NULL) { - ip = map->map_ip(map, ip); - sym = map__find_symbol(map, ip, symbol_filter); - if (sym == NULL) - return; - userspace_samples++; - break; - } - } - /* - * If this is outside of all known maps, - * and is a negative address, try to look it - * up in the kernel dso, as it might be a - * vsyscall or vdso (which executes in user-mode). - */ - if ((long long)ip >= 0) - return; - /* Fall thru */ + break; case PERF_RECORD_MISC_KERNEL: if (hide_kernel_symbols) return; - - sym = kernel_maps__find_function(ip, &map, symbol_filter); - if (sym == NULL) - return; break; default: return; } - syme = symbol__priv(sym); + if (event__preprocess_sample(self, &al, symbol_filter) < 0 || + al.sym == NULL) + return; + syme = symbol__priv(al.sym); if (!syme->skip) { syme->count[counter]++; syme->origin = origin; @@ -986,8 +959,9 @@ static void event__process_sample(const event_t *self, int counter) if (list_empty(&syme->node) || !syme->node.next) __list_insert_active_sym(syme); pthread_mutex_unlock(&active_symbols_lock); + if (origin == PERF_RECORD_MISC_USER) + ++userspace_samples; ++samples; - return; } } diff --git a/tools/perf/util/event.c b/tools/perf/util/event.c index 70b4aa03b47..233d7ad9bd7 100644 --- a/tools/perf/util/event.c +++ b/tools/perf/util/event.c @@ -249,3 +249,65 @@ int event__process_task(event_t *self) return 0; } + +void thread__find_addr_location(struct thread *self, u8 cpumode, + enum map_type type, u64 addr, + struct addr_location *al, + symbol_filter_t filter) +{ + struct thread *thread = al->thread = self; + + al->addr = addr; + + if (cpumode & PERF_RECORD_MISC_KERNEL) { + al->level = 'k'; + thread = kthread; + } else if (cpumode & PERF_RECORD_MISC_USER) + al->level = '.'; + else { + al->level = 'H'; + al->map = NULL; + al->sym = NULL; + return; + } +try_again: + al->map = thread__find_map(thread, type, al->addr); + if (al->map == NULL) { + /* + * If this is outside of all known maps, and is a negative + * address, try to look it up in the kernel dso, as it might be + * a vsyscall or vdso (which executes in user-mode). + * + * XXX This is nasty, we should have a symbol list in the + * "[vdso]" dso, but for now lets use the old trick of looking + * in the whole kernel symbol list. + */ + if ((long long)al->addr < 0 && thread != kthread) { + thread = kthread; + goto try_again; + } + al->sym = NULL; + } else { + al->addr = al->map->map_ip(al->map, al->addr); + al->sym = map__find_symbol(al->map, al->addr, filter); + } +} + +int event__preprocess_sample(const event_t *self, struct addr_location *al, + symbol_filter_t filter) +{ + u8 cpumode = self->header.misc & PERF_RECORD_MISC_CPUMODE_MASK; + struct thread *thread = threads__findnew(self->ip.pid); + + if (thread == NULL) + return -1; + + dump_printf(" ... thread: %s:%d\n", thread->comm, thread->pid); + + thread__find_addr_location(thread, cpumode, MAP__FUNCTION, + self->ip.ip, al, filter); + dump_printf(" ...... dso: %s\n", + al->map ? al->map->dso->long_name : + al->level == 'H' ? "[hypervisor]" : ""); + return 0; +} diff --git a/tools/perf/util/event.h b/tools/perf/util/event.h index 13c12c75f97..a4cc8105cf6 100644 --- a/tools/perf/util/event.h +++ b/tools/perf/util/event.h @@ -152,4 +152,8 @@ int event__process_lost(event_t *self); int event__process_mmap(event_t *self); int event__process_task(event_t *self); +struct addr_location; +int event__preprocess_sample(const event_t *self, struct addr_location *al, + symbol_filter_t filter); + #endif /* __PERF_RECORD_H */ diff --git a/tools/perf/util/hist.c b/tools/perf/util/hist.c index f26cd9ba00f..0ebf6ee16ca 100644 --- a/tools/perf/util/hist.c +++ b/tools/perf/util/hist.c @@ -14,20 +14,19 @@ struct callchain_param callchain_param = { * histogram, sorted on item, collects counts */ -struct hist_entry *__hist_entry__add(struct thread *thread, struct map *map, - struct symbol *sym, +struct hist_entry *__hist_entry__add(struct addr_location *al, struct symbol *sym_parent, - u64 ip, u64 count, char level, bool *hit) + u64 count, bool *hit) { struct rb_node **p = &hist.rb_node; struct rb_node *parent = NULL; struct hist_entry *he; struct hist_entry entry = { - .thread = thread, - .map = map, - .sym = sym, - .ip = ip, - .level = level, + .thread = al->thread, + .map = al->map, + .sym = al->sym, + .ip = al->addr, + .level = al->level, .count = count, .parent = sym_parent, }; diff --git a/tools/perf/util/hist.h b/tools/perf/util/hist.h index ac2149c559b..3020db0c929 100644 --- a/tools/perf/util/hist.h +++ b/tools/perf/util/hist.h @@ -36,9 +36,9 @@ extern unsigned long total_fork; extern unsigned long total_unknown; extern unsigned long total_lost; -struct hist_entry *__hist_entry__add(struct thread *thread, struct map *map, - struct symbol *sym, struct symbol *parent, - u64 ip, u64 count, char level, bool *hit); +struct hist_entry *__hist_entry__add(struct addr_location *al, + struct symbol *parent, + u64 count, bool *hit); extern int64_t hist_entry__cmp(struct hist_entry *, struct hist_entry *); extern int64_t hist_entry__collapse(struct hist_entry *, struct hist_entry *); extern void hist_entry__free(struct hist_entry *); diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index b788c2f5d67..fffcb937cdc 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -43,7 +43,8 @@ static struct symbol_conf symbol_conf__defaults = { .try_vmlinux_path = true, }; -static struct thread kthread_mem, *kthread = &kthread_mem; +static struct thread kthread_mem; +struct thread *kthread = &kthread_mem; bool dso__loaded(const struct dso *self, enum map_type type) { @@ -1178,29 +1179,6 @@ out: return ret; } -static struct symbol *thread__find_symbol(struct thread *self, u64 ip, - enum map_type type, struct map **mapp, - symbol_filter_t filter) -{ - struct map *map = thread__find_map(self, type, ip); - - if (mapp) - *mapp = map; - - if (map) { - ip = map->map_ip(map, ip); - return map__find_symbol(map, ip, filter); - } - - return NULL; -} - -struct symbol *kernel_maps__find_function(u64 ip, struct map **mapp, - symbol_filter_t filter) -{ - return thread__find_symbol(kthread, ip, MAP__FUNCTION, mapp, filter); -} - static struct map *thread__find_map_by_name(struct thread *self, char *name) { struct rb_node *nd; diff --git a/tools/perf/util/symbol.h b/tools/perf/util/symbol.h index 3f9e4a4d83d..17003efa0b3 100644 --- a/tools/perf/util/symbol.h +++ b/tools/perf/util/symbol.h @@ -63,6 +63,14 @@ static inline void *symbol__priv(struct symbol *self) return ((void *)self) - symbol__priv_size; } +struct addr_location { + struct thread *thread; + struct map *map; + struct symbol *sym; + u64 addr; + char level; +}; + struct dso { struct list_head node; struct rb_root symbols[MAP__NR_TYPES]; @@ -105,6 +113,8 @@ size_t kernel_maps__fprintf(FILE *fp); int symbol__init(struct symbol_conf *conf); +struct thread; +struct thread *kthread; extern struct list_head dsos__user, dsos__kernel; extern struct dso *vdso; #endif /* __PERF_SYMBOL */ diff --git a/tools/perf/util/thread.c b/tools/perf/util/thread.c index 2229f82cd63..603f5610861 100644 --- a/tools/perf/util/thread.c +++ b/tools/perf/util/thread.c @@ -285,3 +285,15 @@ size_t threads__fprintf(FILE *fp) return ret; } + +struct symbol *thread__find_symbol(struct thread *self, + enum map_type type, u64 addr, + symbol_filter_t filter) +{ + struct map *map = thread__find_map(self, type, addr); + + if (map != NULL) + return map__find_symbol(map, map->map_ip(map, addr), filter); + + return NULL; +} diff --git a/tools/perf/util/thread.h b/tools/perf/util/thread.h index 59b0d9b577d..686d6e914d9 100644 --- a/tools/perf/util/thread.h +++ b/tools/perf/util/thread.h @@ -27,19 +27,30 @@ size_t thread__fprintf_maps(struct thread *self, FILE *fp); size_t threads__fprintf(FILE *fp); void maps__insert(struct rb_root *maps, struct map *map); -struct map *maps__find(struct rb_root *maps, u64 ip); - -struct symbol *kernel_maps__find_function(const u64 ip, struct map **mapp, - symbol_filter_t filter); +struct map *maps__find(struct rb_root *maps, u64 addr); static inline struct map *thread__find_map(struct thread *self, - enum map_type type, u64 ip) + enum map_type type, u64 addr) { - return self ? maps__find(&self->maps[type], ip) : NULL; + return self ? maps__find(&self->maps[type], addr) : NULL; } static inline void __thread__insert_map(struct thread *self, struct map *map) { maps__insert(&self->maps[map->type], map); } + +void thread__find_addr_location(struct thread *self, u8 cpumode, + enum map_type type, u64 addr, + struct addr_location *al, + symbol_filter_t filter); +struct symbol *thread__find_symbol(struct thread *self, + enum map_type type, u64 addr, + symbol_filter_t filter); + +static inline struct symbol * +thread__find_function(struct thread *self, u64 addr, symbol_filter_t filter) +{ + return thread__find_symbol(self, MAP__FUNCTION, addr, filter); +} #endif /* __PERF_THREAD_H */ -- cgit v1.2.3-18-g5258 From 956ffd027bedc4106b901eb6a50f0a6c6de4113d Mon Sep 17 00:00:00 2001 From: Tom Zanussi Date: Wed, 25 Nov 2009 01:15:46 -0600 Subject: perf trace: Add scripting ops Adds an interface, scripting_ops, that when implemented for a particular scripting language enables built-in support for trace stream processing using that language. The interface is designed to enable full-fledged language interpreters to be embedded inside the perf executable and thereby make the full capabilities of the supported languages available for trace processing. See below for details on the interface. This patch also adds a couple command-line options to 'perf trace': The -s option option is used to specify the script to be run. Script names that can be used with -s take the form: [language spec:]scriptname[.ext] Scripting languages register a set of 'language specs' that can be used to specify scripts for the registered languages. The specs can be used either as prefixes or extensions. If [language spec:] is used, the script is taken as a script of the matching language regardless of any extension it might have. If [language spec:] is not used, [.ext] is used to look up the language it corresponds to. Language specs are case insensitive. e.g. Perl scripts can be specified in the following ways: Perl:scriptname pl:scriptname.py # extension ignored PL:scriptname scriptname.pl scriptname.perl The -g [language spec] option gives users an easy starting point for writing scripts in the specified language. Scripting support for a particular language can implement a generate_script() scripting op that outputs an empty (or near-empty) set of handlers for all the events contained in a given perf.data trace file - this option gives users a direct way to access that. Adding support for a scripting language --------------------------------------- The main thing that needs to be done do add support for a new language is to implement the scripting_ops interface: It consists of the following four functions: start_script() stop_script() process_event() generate_script() start_script() is called before any events are processed, and is meant to give the scripting language support an opportunity to set things up to receive events e.g. create and initialize an instance of a language interpreter. stop_script() is called after all events are processed, and is meant to give the scripting language support an opportunity to clean up e.g. destroy the interpreter instance, etc. process_event() is called once for each event and takes as its main parameter a pointer to the binary trace event record to be processed. The implementation is responsible for picking out the binary fields from the event record and sending them to the script handler function associated with that event e.g. a function derived from the event name it's meant to handle e.g. 'sched::sched_switch()'. The 'format' information for trace events can be used to parse the binary data and map it into a form usable by a given scripting language; see the Perl implemention in subsequent patches for one possible way to leverage the existing trace format parsing code in perf and map that info into specific scripting language types. generate_script() should generate a ready-to-run script for the current set of events in the trace, preferably with bodies that print out every field for each event. Again, look at the Perl implementation for clues as to how that can be done. This is an optional, but very useful op. Support for a given language should also add a language-specific setup function and call it from setup_scripting(). The language-specific setup function associates the the scripting ops for that language with one or more 'language specifiers' (see below) using script_spec_register(). When a script name is specified on the command line, the scripting ops associated with the specified language are used to instantiate and use the appropriate interpreter to process the trace stream. In general, it should be relatively easy to add support for a new language, especially if the language implementation supports an interface allowing an interpreter to be 'embedded' inside another program (in this case the containing program will be 'perf trace'). If so, it should be relatively straightforward to translate trace events into invocations of user-defined script functions where e.g. the function name corresponds to the event type and the function parameters correspond to the event fields. The event and field type information exported by the event tracing infrastructure (via the event 'format' files) should be enough to parse and send any piece of trace data to the user script. The easiest way to see how this can be done would be to look at the Perl implementation contained in perf/util/trace-event-perl.c/.h. There are a couple of other things that aren't covered by the scripting_ops or setup interface and are technically optional, but should be implemented if possible. One of these is support for 'flag' and 'symbolic' fields e.g. being able to use more human-readable values such as 'GFP_KERNEL' or HI/BLOCK_IOPOLL/TASKLET in place of raw flag values. See the Perl implementation to see how this can be done. The other thing is support for 'calling back' into the perf executable to access e.g. uncommon fields not passed by default into handler functions, or any metadata the implementation might want to make available to users via the language interface. Again, see the Perl implementation for examples. Signed-off-by: Tom Zanussi Cc: fweisbec@gmail.com Cc: rostedt@goodmis.org Cc: anton@samba.org Cc: hch@infradead.org LKML-Reference: <1259133352-23685-2-git-send-email-tzanussi@gmail.com> Signed-off-by: Ingo Molnar --- tools/perf/builtin-trace.c | 255 +++++++++++++++++++++++++++++++++++++++++- tools/perf/util/trace-event.h | 11 ++ 2 files changed, 261 insertions(+), 5 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-trace.c index a7750256c40..e96bb534b94 100644 --- a/tools/perf/builtin-trace.c +++ b/tools/perf/builtin-trace.c @@ -6,6 +6,46 @@ #include "util/thread.h" #include "util/header.h" +static char const *script_name; +static char const *generate_script_lang; + +static int default_start_script(const char *script __attribute((unused))) +{ + return 0; +} + +static int default_stop_script(void) +{ + return 0; +} + +static int default_generate_script(const char *outfile __attribute ((unused))) +{ + return 0; +} + +static struct scripting_ops default_scripting_ops = { + .start_script = default_start_script, + .stop_script = default_stop_script, + .process_event = print_event, + .generate_script = default_generate_script, +}; + +static struct scripting_ops *scripting_ops; + +static void setup_scripting(void) +{ + /* make sure PERF_EXEC_PATH is set for scripts */ + perf_set_argv_exec_path(perf_exec_path()); + + scripting_ops = &default_scripting_ops; +} + +static int cleanup_scripting(void) +{ + return scripting_ops->stop_script(); +} + #include "util/parse-options.h" #include "perf.h" @@ -13,11 +53,12 @@ #include "util/trace-event.h" #include "util/data_map.h" +#include "util/exec_cmd.h" -static char const *input_name = "perf.data"; +static char const *input_name = "perf.data"; -static struct perf_header *header; -static u64 sample_type; +static struct perf_header *header; +static u64 sample_type; static int process_sample_event(event_t *event) { @@ -69,7 +110,8 @@ static int process_sample_event(event_t *event) * field, although it should be the same than this perf * event pid */ - print_event(cpu, raw->data, raw->size, timestamp, thread->comm); + scripting_ops->process_event(cpu, raw->data, raw->size, + timestamp, thread->comm); } event__stats.total += period; @@ -105,6 +147,154 @@ static int __cmd_trace(void) 0, 0, &event__cwdlen, &event__cwd); } +struct script_spec { + struct list_head node; + struct scripting_ops *ops; + char spec[0]; +}; + +LIST_HEAD(script_specs); + +static struct script_spec *script_spec__new(const char *spec, + struct scripting_ops *ops) +{ + struct script_spec *s = malloc(sizeof(*s) + strlen(spec) + 1); + + if (s != NULL) { + strcpy(s->spec, spec); + s->ops = ops; + } + + return s; +} + +static void script_spec__delete(struct script_spec *s) +{ + free(s->spec); + free(s); +} + +static void script_spec__add(struct script_spec *s) +{ + list_add_tail(&s->node, &script_specs); +} + +static struct script_spec *script_spec__find(const char *spec) +{ + struct script_spec *s; + + list_for_each_entry(s, &script_specs, node) + if (strcasecmp(s->spec, spec) == 0) + return s; + return NULL; +} + +static struct script_spec *script_spec__findnew(const char *spec, + struct scripting_ops *ops) +{ + struct script_spec *s = script_spec__find(spec); + + if (s) + return s; + + s = script_spec__new(spec, ops); + if (!s) + goto out_delete_spec; + + script_spec__add(s); + + return s; + +out_delete_spec: + script_spec__delete(s); + + return NULL; +} + +int script_spec_register(const char *spec, struct scripting_ops *ops) +{ + struct script_spec *s; + + s = script_spec__find(spec); + if (s) + return -1; + + s = script_spec__findnew(spec, ops); + if (!s) + return -1; + + return 0; +} + +static struct scripting_ops *script_spec__lookup(const char *spec) +{ + struct script_spec *s = script_spec__find(spec); + if (!s) + return NULL; + + return s->ops; +} + +static void list_available_languages(void) +{ + struct script_spec *s; + + fprintf(stderr, "\n"); + fprintf(stderr, "Scripting language extensions (used in " + "perf trace -s [spec:]script.[spec]):\n\n"); + + list_for_each_entry(s, &script_specs, node) + fprintf(stderr, " %-42s [%s]\n", s->spec, s->ops->name); + + fprintf(stderr, "\n"); +} + +static int parse_scriptname(const struct option *opt __used, + const char *str, int unset __used) +{ + char spec[PATH_MAX]; + const char *script, *ext; + int len; + + if (strcmp(str, "list") == 0) { + list_available_languages(); + return 0; + } + + script = strchr(str, ':'); + if (script) { + len = script - str; + if (len >= PATH_MAX) { + fprintf(stderr, "invalid language specifier"); + return -1; + } + strncpy(spec, str, len); + spec[len] = '\0'; + scripting_ops = script_spec__lookup(spec); + if (!scripting_ops) { + fprintf(stderr, "invalid language specifier"); + return -1; + } + script++; + } else { + script = str; + ext = strchr(script, '.'); + if (!ext) { + fprintf(stderr, "invalid script extension"); + return -1; + } + scripting_ops = script_spec__lookup(++ext); + if (!scripting_ops) { + fprintf(stderr, "invalid script extension"); + return -1; + } + } + + script_name = strdup(script); + + return 0; +} + static const char * const annotate_usage[] = { "perf trace [] ", NULL @@ -117,13 +307,23 @@ static const struct option options[] = { "be more verbose (show symbol address, etc)"), OPT_BOOLEAN('l', "latency", &latency_format, "show latency attributes (irqs/preemption disabled, etc)"), + OPT_CALLBACK('s', "script", NULL, "name", + "script file name (lang:script name, script name, or *)", + parse_scriptname), + OPT_STRING('g', "gen-script", &generate_script_lang, "lang", + "generate perf-trace.xx script in specified language"), + OPT_END() }; int cmd_trace(int argc, const char **argv, const char *prefix __used) { + int err; + symbol__init(0); + setup_scripting(); + argc = parse_options(argc, argv, options, annotate_usage, 0); if (argc) { /* @@ -136,5 +336,50 @@ int cmd_trace(int argc, const char **argv, const char *prefix __used) setup_pager(); - return __cmd_trace(); + if (generate_script_lang) { + struct stat perf_stat; + + int input = open(input_name, O_RDONLY); + if (input < 0) { + perror("failed to open file"); + exit(-1); + } + + err = fstat(input, &perf_stat); + if (err < 0) { + perror("failed to stat file"); + exit(-1); + } + + if (!perf_stat.st_size) { + fprintf(stderr, "zero-sized file, nothing to do!\n"); + exit(0); + } + + scripting_ops = script_spec__lookup(generate_script_lang); + if (!scripting_ops) { + fprintf(stderr, "invalid language specifier"); + return -1; + } + + header = perf_header__new(); + if (header == NULL) + return -1; + + perf_header__read(header, input); + err = scripting_ops->generate_script("perf-trace"); + goto out; + } + + if (script_name) { + err = scripting_ops->start_script(script_name); + if (err) + goto out; + } + + err = __cmd_trace(); + + cleanup_scripting(); +out: + return err; } diff --git a/tools/perf/util/trace-event.h b/tools/perf/util/trace-event.h index dd51c6872a1..e7aaf002e66 100644 --- a/tools/perf/util/trace-event.h +++ b/tools/perf/util/trace-event.h @@ -259,4 +259,15 @@ enum trace_flag_type { TRACE_FLAG_SOFTIRQ = 0x10, }; +struct scripting_ops { + const char *name; + int (*start_script) (const char *); + int (*stop_script) (void); + void (*process_event) (int cpu, void *data, int size, + unsigned long long nsecs, char *comm); + int (*generate_script) (const char *outfile); +}; + +int script_spec_register(const char *spec, struct scripting_ops *ops); + #endif /* __PERF_TRACE_EVENTS_H */ -- cgit v1.2.3-18-g5258 From eb9a42caa7a926beb935a22bc59d981b35f0b652 Mon Sep 17 00:00:00 2001 From: Tom Zanussi Date: Wed, 25 Nov 2009 01:15:47 -0600 Subject: perf trace: Add flag/symbolic format_flags It's useful to know whether a field is a flag or symbolic field for e.g. when generating scripts - it allows us to translate those fields specially rather than literally as plain numeric values. Signed-off-by: Tom Zanussi Cc: fweisbec@gmail.com Cc: rostedt@goodmis.org Cc: anton@samba.org Cc: hch@infradead.org LKML-Reference: <1259133352-23685-3-git-send-email-tzanussi@gmail.com> Signed-off-by: Ingo Molnar --- tools/perf/util/trace-event-parse.c | 17 +++++++++++++++++ tools/perf/util/trace-event.h | 2 ++ 2 files changed, 19 insertions(+) (limited to 'tools/perf') diff --git a/tools/perf/util/trace-event-parse.c b/tools/perf/util/trace-event-parse.c index 7021dc1b0ca..85d7163a9fd 100644 --- a/tools/perf/util/trace-event-parse.c +++ b/tools/perf/util/trace-event-parse.c @@ -48,6 +48,11 @@ static unsigned long long input_buf_siz; static int cpus; static int long_size; +static int is_flag_field; +static int is_symbolic_field; + +static struct format_field * +find_any_field(struct event *event, const char *name); static void init_input_buf(char *buf, unsigned long long size) { @@ -1301,6 +1306,16 @@ process_entry(struct event *event __unused, struct print_arg *arg, arg->type = PRINT_FIELD; arg->field.name = field; + if (is_flag_field) { + arg->field.field = find_any_field(event, arg->field.name); + arg->field.field->flags |= FIELD_IS_FLAG; + is_flag_field = 0; + } else if (is_symbolic_field) { + arg->field.field = find_any_field(event, arg->field.name); + arg->field.field->flags |= FIELD_IS_SYMBOLIC; + is_symbolic_field = 0; + } + type = read_token(&token); *tok = token; @@ -1668,9 +1683,11 @@ process_arg_token(struct event *event, struct print_arg *arg, type = process_entry(event, arg, &token); } else if (strcmp(token, "__print_flags") == 0) { free_token(token); + is_flag_field = 1; type = process_flags(event, arg, &token); } else if (strcmp(token, "__print_symbolic") == 0) { free_token(token); + is_symbolic_field = 1; type = process_symbols(event, arg, &token); } else if (strcmp(token, "__get_str") == 0) { free_token(token); diff --git a/tools/perf/util/trace-event.h b/tools/perf/util/trace-event.h index e7aaf002e66..aeb915778ae 100644 --- a/tools/perf/util/trace-event.h +++ b/tools/perf/util/trace-event.h @@ -29,6 +29,8 @@ enum format_flags { FIELD_IS_SIGNED = 4, FIELD_IS_STRING = 8, FIELD_IS_DYNAMIC = 16, + FIELD_IS_FLAG = 32, + FIELD_IS_SYMBOLIC = 64, }; struct format_field { -- cgit v1.2.3-18-g5258 From 16c632de64a74644a46e7636db26b2cfb530ca13 Mon Sep 17 00:00:00 2001 From: Tom Zanussi Date: Wed, 25 Nov 2009 01:15:48 -0600 Subject: perf trace: Add Perl scripting support Implement trace_scripting_ops to make Perl a supported perf trace scripting language. Additionally adds code that allows Perl trace scripts to access the 'flag' and 'symbolic' (__print_flags(), __print_symbolic()) field information parsed from the trace format files. Also adds the Perl implementation of the generate_script() trace_scripting_op, which creates a ready-to-run perf trace Perl script based on existing trace data. Scripts generated by this implementation print out all the fields for each event mentioned in perf.data (and will detect and generate the proper scripting code for 'flag' and 'symbolic' fields), and will additionally generate handlers for the special 'trace_unhandled', 'trace_begin' and 'trace_end' handlers. Script authors can simply remove the printing code to implement their own custom event handling. Signed-off-by: Tom Zanussi Cc: fweisbec@gmail.com Cc: rostedt@goodmis.org Cc: anton@samba.org Cc: hch@infradead.org LKML-Reference: <1259133352-23685-4-git-send-email-tzanussi@gmail.com> Signed-off-by: Ingo Molnar --- tools/perf/Makefile | 13 + tools/perf/builtin-trace.c | 2 + tools/perf/util/trace-event-parse.c | 18 +- tools/perf/util/trace-event-perl.c | 552 ++++++++++++++++++++++++++++++++++++ tools/perf/util/trace-event-perl.h | 42 +++ tools/perf/util/trace-event.h | 7 + 6 files changed, 629 insertions(+), 5 deletions(-) create mode 100644 tools/perf/util/trace-event-perl.c create mode 100644 tools/perf/util/trace-event-perl.h (limited to 'tools/perf') diff --git a/tools/perf/Makefile b/tools/perf/Makefile index f1537a94a05..19e37cd14ae 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -407,6 +407,7 @@ LIB_OBJS += util/thread.o LIB_OBJS += util/trace-event-parse.o LIB_OBJS += util/trace-event-read.o LIB_OBJS += util/trace-event-info.o +LIB_OBJS += util/trace-event-perl.o LIB_OBJS += util/svghelper.o LIB_OBJS += util/sort.o LIB_OBJS += util/hist.o @@ -489,6 +490,15 @@ else LIB_OBJS += util/probe-finder.o endif +PERL_EMBED_LDOPTS = `perl -MExtUtils::Embed -e ldopts` +PERL_EMBED_CCOPTS = `perl -MExtUtils::Embed -e ccopts` + +ifneq ($(shell sh -c "(echo '\#include '; echo '\#include '; echo 'int main(void) { perl_alloc(); return 0; }') | $(CC) -x c - $(PERL_EMBED_CCOPTS) -o /dev/null $(PERL_EMBED_LDOPTS) > /dev/null 2>&1 && echo y"), y) + BASIC_CFLAGS += -DNO_LIBPERL +else + ALL_LDFLAGS += $(PERL_EMBED_LDOPTS) +endif + ifdef NO_DEMANGLE BASIC_CFLAGS += -DNO_DEMANGLE else @@ -860,6 +870,9 @@ util/hweight.o: ../../lib/hweight.c PERF-CFLAGS util/find_next_bit.o: ../../lib/find_next_bit.c PERF-CFLAGS $(QUIET_CC)$(CC) -o util/find_next_bit.o -c $(ALL_CFLAGS) -DETC_PERFCONFIG='"$(ETC_PERFCONFIG_SQ)"' $< +util/trace-event-perl.o: util/trace-event-perl.c PERF-CFLAGS + $(QUIET_CC)$(CC) -o util/trace-event-perl.o -c $(ALL_CFLAGS) $(PERL_EMBED_CCOPTS) -Wno-redundant-decls -Wno-strict-prototypes -Wno-unused-parameter $< + perf-%$X: %.o $(PERFLIBS) $(QUIET_LINK)$(CC) $(ALL_CFLAGS) -o $@ $(ALL_LDFLAGS) $(filter %.o,$^) $(LIBS) diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-trace.c index e96bb534b94..ca8ebf1ec64 100644 --- a/tools/perf/builtin-trace.c +++ b/tools/perf/builtin-trace.c @@ -38,6 +38,8 @@ static void setup_scripting(void) /* make sure PERF_EXEC_PATH is set for scripts */ perf_set_argv_exec_path(perf_exec_path()); + setup_perl_scripting(); + scripting_ops = &default_scripting_ops; } diff --git a/tools/perf/util/trace-event-parse.c b/tools/perf/util/trace-event-parse.c index 85d7163a9fd..1f16495e559 100644 --- a/tools/perf/util/trace-event-parse.c +++ b/tools/perf/util/trace-event-parse.c @@ -1888,7 +1888,7 @@ find_any_field(struct event *event, const char *name) return find_field(event, name); } -static unsigned long long read_size(void *ptr, int size) +unsigned long long read_size(void *ptr, int size) { switch (size) { case 1: @@ -1973,7 +1973,7 @@ int trace_parse_common_type(void *data) "common_type"); } -static int parse_common_pid(void *data) +int trace_parse_common_pid(void *data) { static int pid_offset; static int pid_size; @@ -2025,6 +2025,14 @@ struct event *trace_find_event(int id) return event; } +struct event *trace_find_next_event(struct event *event) +{ + if (!event) + return event_list; + + return event->next; +} + static unsigned long long eval_num_arg(void *data, int size, struct event *event, struct print_arg *arg) { @@ -2164,7 +2172,7 @@ static const struct flag flags[] = { { "HRTIMER_RESTART", 1 }, }; -static unsigned long long eval_flag(const char *flag) +unsigned long long eval_flag(const char *flag) { int i; @@ -2694,7 +2702,7 @@ get_return_for_leaf(int cpu, int cur_pid, unsigned long long cur_func, if (!(event->flags & EVENT_FL_ISFUNCRET)) return NULL; - pid = parse_common_pid(next->data); + pid = trace_parse_common_pid(next->data); field = find_field(event, "func"); if (!field) die("function return does not have field func"); @@ -2980,7 +2988,7 @@ void print_event(int cpu, void *data, int size, unsigned long long nsecs, return; } - pid = parse_common_pid(data); + pid = trace_parse_common_pid(data); if (event->flags & (EVENT_FL_ISFUNCENT | EVENT_FL_ISFUNCRET)) return pretty_print_func_graph(data, size, event, cpu, diff --git a/tools/perf/util/trace-event-perl.c b/tools/perf/util/trace-event-perl.c new file mode 100644 index 00000000000..c56b08d704d --- /dev/null +++ b/tools/perf/util/trace-event-perl.c @@ -0,0 +1,552 @@ +/* + * trace-event-perl. Feed perf trace events to an embedded Perl interpreter. + * + * Copyright (C) 2009 Tom Zanussi + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + * + */ + +#include +#include +#include +#include +#include + +#include "../perf.h" +#include "util.h" +#include "trace-event.h" +#include "trace-event-perl.h" + +INTERP my_perl; + +#define FTRACE_MAX_EVENT \ + ((1 << (sizeof(unsigned short) * 8)) - 1) + +struct event *events[FTRACE_MAX_EVENT]; + +static struct scripting_context *scripting_context; + +static char *cur_field_name; +static int zero_flag_atom; + +static void define_symbolic_value(const char *ev_name, + const char *field_name, + const char *field_value, + const char *field_str) +{ + unsigned long long value; + dSP; + + value = eval_flag(field_value); + + ENTER; + SAVETMPS; + PUSHMARK(SP); + + XPUSHs(sv_2mortal(newSVpv(ev_name, 0))); + XPUSHs(sv_2mortal(newSVpv(field_name, 0))); + XPUSHs(sv_2mortal(newSVuv(value))); + XPUSHs(sv_2mortal(newSVpv(field_str, 0))); + + PUTBACK; + if (get_cv("main::define_symbolic_value", 0)) + call_pv("main::define_symbolic_value", G_SCALAR); + SPAGAIN; + PUTBACK; + FREETMPS; + LEAVE; +} + +static void define_symbolic_values(struct print_flag_sym *field, + const char *ev_name, + const char *field_name) +{ + define_symbolic_value(ev_name, field_name, field->value, field->str); + if (field->next) + define_symbolic_values(field->next, ev_name, field_name); +} + +static void define_symbolic_field(const char *ev_name, + const char *field_name) +{ + dSP; + + ENTER; + SAVETMPS; + PUSHMARK(SP); + + XPUSHs(sv_2mortal(newSVpv(ev_name, 0))); + XPUSHs(sv_2mortal(newSVpv(field_name, 0))); + + PUTBACK; + if (get_cv("main::define_symbolic_field", 0)) + call_pv("main::define_symbolic_field", G_SCALAR); + SPAGAIN; + PUTBACK; + FREETMPS; + LEAVE; +} + +static void define_flag_value(const char *ev_name, + const char *field_name, + const char *field_value, + const char *field_str) +{ + unsigned long long value; + dSP; + + value = eval_flag(field_value); + + ENTER; + SAVETMPS; + PUSHMARK(SP); + + XPUSHs(sv_2mortal(newSVpv(ev_name, 0))); + XPUSHs(sv_2mortal(newSVpv(field_name, 0))); + XPUSHs(sv_2mortal(newSVuv(value))); + XPUSHs(sv_2mortal(newSVpv(field_str, 0))); + + PUTBACK; + if (get_cv("main::define_flag_value", 0)) + call_pv("main::define_flag_value", G_SCALAR); + SPAGAIN; + PUTBACK; + FREETMPS; + LEAVE; +} + +static void define_flag_values(struct print_flag_sym *field, + const char *ev_name, + const char *field_name) +{ + define_flag_value(ev_name, field_name, field->value, field->str); + if (field->next) + define_flag_values(field->next, ev_name, field_name); +} + +static void define_flag_field(const char *ev_name, + const char *field_name, + const char *delim) +{ + dSP; + + ENTER; + SAVETMPS; + PUSHMARK(SP); + + XPUSHs(sv_2mortal(newSVpv(ev_name, 0))); + XPUSHs(sv_2mortal(newSVpv(field_name, 0))); + XPUSHs(sv_2mortal(newSVpv(delim, 0))); + + PUTBACK; + if (get_cv("main::define_flag_field", 0)) + call_pv("main::define_flag_field", G_SCALAR); + SPAGAIN; + PUTBACK; + FREETMPS; + LEAVE; +} + +static void define_event_symbols(struct event *event, + const char *ev_name, + struct print_arg *args) +{ + switch (args->type) { + case PRINT_NULL: + break; + case PRINT_ATOM: + define_flag_value(ev_name, cur_field_name, "0", + args->atom.atom); + zero_flag_atom = 0; + break; + case PRINT_FIELD: + if (cur_field_name) + free(cur_field_name); + cur_field_name = strdup(args->field.name); + break; + case PRINT_FLAGS: + define_event_symbols(event, ev_name, args->flags.field); + define_flag_field(ev_name, cur_field_name, args->flags.delim); + define_flag_values(args->flags.flags, ev_name, cur_field_name); + break; + case PRINT_SYMBOL: + define_event_symbols(event, ev_name, args->symbol.field); + define_symbolic_field(ev_name, cur_field_name); + define_symbolic_values(args->symbol.symbols, ev_name, + cur_field_name); + break; + case PRINT_STRING: + break; + case PRINT_TYPE: + define_event_symbols(event, ev_name, args->typecast.item); + break; + case PRINT_OP: + if (strcmp(args->op.op, ":") == 0) + zero_flag_atom = 1; + define_event_symbols(event, ev_name, args->op.left); + define_event_symbols(event, ev_name, args->op.right); + break; + default: + /* we should warn... */ + return; + } + + if (args->next) + define_event_symbols(event, ev_name, args->next); +} + +static inline struct event *find_cache_event(int type) +{ + static char ev_name[256]; + struct event *event; + + if (events[type]) + return events[type]; + + events[type] = event = trace_find_event(type); + if (!event) + return NULL; + + sprintf(ev_name, "%s::%s", event->system, event->name); + + define_event_symbols(event, ev_name, event->print_fmt.args); + + return event; +} + +static void perl_process_event(int cpu, void *data, + int size __attribute((unused)), + unsigned long long nsecs, char *comm) +{ + struct format_field *field; + static char handler[256]; + unsigned long long val; + unsigned long s, ns; + struct event *event; + int type; + int pid; + + dSP; + + type = trace_parse_common_type(data); + + event = find_cache_event(type); + if (!event) + die("ug! no event found for type %d", type); + + pid = trace_parse_common_pid(data); + + sprintf(handler, "%s::%s", event->system, event->name); + + s = nsecs / NSECS_PER_SEC; + ns = nsecs - s * NSECS_PER_SEC; + + scripting_context->event_data = data; + + ENTER; + SAVETMPS; + PUSHMARK(SP); + + XPUSHs(sv_2mortal(newSVpv(handler, 0))); + XPUSHs(sv_2mortal(newSViv(PTR2IV(scripting_context)))); + XPUSHs(sv_2mortal(newSVuv(cpu))); + XPUSHs(sv_2mortal(newSVuv(s))); + XPUSHs(sv_2mortal(newSVuv(ns))); + XPUSHs(sv_2mortal(newSViv(pid))); + XPUSHs(sv_2mortal(newSVpv(comm, 0))); + + /* common fields other than pid can be accessed via xsub fns */ + + for (field = event->format.fields; field; field = field->next) { + if (field->flags & FIELD_IS_STRING) { + int offset; + if (field->flags & FIELD_IS_DYNAMIC) { + offset = *(int *)(data + field->offset); + offset &= 0xffff; + } else + offset = field->offset; + XPUSHs(sv_2mortal(newSVpv((char *)data + offset, 0))); + } else { /* FIELD_IS_NUMERIC */ + val = read_size(data + field->offset, field->size); + if (field->flags & FIELD_IS_SIGNED) { + XPUSHs(sv_2mortal(newSViv(val))); + } else { + XPUSHs(sv_2mortal(newSVuv(val))); + } + } + } + + PUTBACK; + if (get_cv(handler, 0)) + call_pv(handler, G_SCALAR); + else if (get_cv("main::trace_unhandled", 0)) { + XPUSHs(sv_2mortal(newSVpv(handler, 0))); + XPUSHs(sv_2mortal(newSViv(PTR2IV(scripting_context)))); + XPUSHs(sv_2mortal(newSVuv(cpu))); + XPUSHs(sv_2mortal(newSVuv(nsecs))); + XPUSHs(sv_2mortal(newSViv(pid))); + XPUSHs(sv_2mortal(newSVpv(comm, 0))); + call_pv("main::trace_unhandled", G_SCALAR); + } + SPAGAIN; + PUTBACK; + FREETMPS; + LEAVE; +} + +static void run_start_sub(void) +{ + dSP; /* access to Perl stack */ + PUSHMARK(SP); + + if (get_cv("main::trace_begin", 0)) + call_pv("main::trace_begin", G_DISCARD | G_NOARGS); +} + +/* + * Start trace script + */ +static int perl_start_script(const char *script) +{ + const char *command_line[2] = { "", NULL }; + + command_line[1] = script; + + my_perl = perl_alloc(); + perl_construct(my_perl); + + if (perl_parse(my_perl, NULL, 2, (char **)command_line, (char **)NULL)) + return -1; + + perl_run(my_perl); + if (SvTRUE(ERRSV)) + return -1; + + run_start_sub(); + + fprintf(stderr, "perf trace started with Perl script %s\n\n", script); + + return 0; +} + +/* + * Stop trace script + */ +static int perl_stop_script(void) +{ + dSP; /* access to Perl stack */ + PUSHMARK(SP); + + if (get_cv("main::trace_end", 0)) + call_pv("main::trace_end", G_DISCARD | G_NOARGS); + + perl_destruct(my_perl); + perl_free(my_perl); + + fprintf(stderr, "\nperf trace Perl script stopped\n"); + + return 0; +} + +static int perl_generate_script(const char *outfile) +{ + struct event *event = NULL; + struct format_field *f; + char fname[PATH_MAX]; + int not_first, count; + FILE *ofp; + + sprintf(fname, "%s.pl", outfile); + ofp = fopen(fname, "w"); + if (ofp == NULL) { + fprintf(stderr, "couldn't open %s\n", fname); + return -1; + } + + fprintf(ofp, "# perf trace event handlers, " + "generated by perf trace -g perl\n"); + + fprintf(ofp, "# Licensed under the terms of the GNU GPL" + " License version 2\n\n"); + + fprintf(ofp, "# The common_* event handler fields are the most useful " + "fields common to\n"); + + fprintf(ofp, "# all events. They don't necessarily correspond to " + "the 'common_*' fields\n"); + + fprintf(ofp, "# in the format files. Those fields not available as " + "handler params can\n"); + + fprintf(ofp, "# be retrieved using Perl functions of the form " + "common_*($context).\n"); + + fprintf(ofp, "# See Context.pm for the list of available " + "functions.\n\n"); + + fprintf(ofp, "use lib \"$ENV{'PERF_EXEC_PATH'}/scripts/perl/" + "Perf-Trace-Util/lib\";\n"); + + fprintf(ofp, "use lib \"./Perf-Trace-Util/lib\";\n"); + fprintf(ofp, "use Perf::Trace::Core;\n"); + fprintf(ofp, "use Perf::Trace::Context;\n"); + fprintf(ofp, "use Perf::Trace::Util;\n\n"); + + fprintf(ofp, "sub trace_begin\n{\n\t# optional\n}\n\n"); + fprintf(ofp, "sub trace_end\n{\n\t# optional\n}\n\n"); + + while ((event = trace_find_next_event(event))) { + fprintf(ofp, "sub %s::%s\n{\n", event->system, event->name); + fprintf(ofp, "\tmy ("); + + fprintf(ofp, "$event_name, "); + fprintf(ofp, "$context, "); + fprintf(ofp, "$common_cpu, "); + fprintf(ofp, "$common_secs, "); + fprintf(ofp, "$common_nsecs,\n"); + fprintf(ofp, "\t $common_pid, "); + fprintf(ofp, "$common_comm,\n\t "); + + not_first = 0; + count = 0; + + for (f = event->format.fields; f; f = f->next) { + if (not_first++) + fprintf(ofp, ", "); + if (++count % 5 == 0) + fprintf(ofp, "\n\t "); + + fprintf(ofp, "$%s", f->name); + } + fprintf(ofp, ") = @_;\n\n"); + + fprintf(ofp, "\tprint_header($event_name, $common_cpu, " + "$common_secs, $common_nsecs,\n\t " + "$common_pid, $common_comm);\n\n"); + + fprintf(ofp, "\tprintf(\""); + + not_first = 0; + count = 0; + + for (f = event->format.fields; f; f = f->next) { + if (not_first++) + fprintf(ofp, ", "); + if (count && count % 4 == 0) { + fprintf(ofp, "\".\n\t \""); + } + count++; + + fprintf(ofp, "%s=", f->name); + if (f->flags & FIELD_IS_STRING || + f->flags & FIELD_IS_FLAG || + f->flags & FIELD_IS_SYMBOLIC) + fprintf(ofp, "%%s"); + else if (f->flags & FIELD_IS_SIGNED) + fprintf(ofp, "%%d"); + else + fprintf(ofp, "%%u"); + } + + fprintf(ofp, "\\n\",\n\t "); + + not_first = 0; + count = 0; + + for (f = event->format.fields; f; f = f->next) { + if (not_first++) + fprintf(ofp, ", "); + + if (++count % 5 == 0) + fprintf(ofp, "\n\t "); + + if (f->flags & FIELD_IS_FLAG) { + if ((count - 1) % 5 != 0) { + fprintf(ofp, "\n\t "); + count = 4; + } + fprintf(ofp, "flag_str(\""); + fprintf(ofp, "%s::%s\", ", event->system, + event->name); + fprintf(ofp, "\"%s\", $%s)", f->name, + f->name); + } else if (f->flags & FIELD_IS_SYMBOLIC) { + if ((count - 1) % 5 != 0) { + fprintf(ofp, "\n\t "); + count = 4; + } + fprintf(ofp, "symbol_str(\""); + fprintf(ofp, "%s::%s\", ", event->system, + event->name); + fprintf(ofp, "\"%s\", $%s)", f->name, + f->name); + } else + fprintf(ofp, "$%s", f->name); + } + + fprintf(ofp, ");\n"); + fprintf(ofp, "}\n\n"); + } + + fprintf(ofp, "sub trace_unhandled\n{\n\tmy ($event_name, $context, " + "$common_cpu, $common_secs, $common_nsecs,\n\t " + "$common_pid, $common_comm) = @_;\n\n"); + + fprintf(ofp, "\tprint_header($event_name, $common_cpu, " + "$common_secs, $common_nsecs,\n\t $common_pid, " + "$common_comm);\n}\n\n"); + + fprintf(ofp, "sub print_header\n{\n" + "\tmy ($event_name, $cpu, $secs, $nsecs, $pid, $comm) = @_;\n\n" + "\tprintf(\"%%-20s %%5u %%05u.%%09u %%8u %%-20s \",\n\t " + "$event_name, $cpu, $secs, $nsecs, $pid, $comm);\n}"); + + fclose(ofp); + + fprintf(stderr, "generated Perl script: %s\n", fname); + + return 0; +} + +struct scripting_ops perl_scripting_ops = { + .name = "Perl", + .start_script = perl_start_script, + .stop_script = perl_stop_script, + .process_event = perl_process_event, + .generate_script = perl_generate_script, +}; + +#ifdef NO_LIBPERL +void setup_perl_scripting(void) +{ + fprintf(stderr, "Perl scripting not supported." + " Install libperl-dev[el] and rebuild perf to get it.\n"); +} +#else +void setup_perl_scripting(void) +{ + int err; + err = script_spec_register("Perl", &perl_scripting_ops); + if (err) + die("error registering Perl script extension"); + + err = script_spec_register("pl", &perl_scripting_ops); + if (err) + die("error registering pl script extension"); + + scripting_context = malloc(sizeof(struct scripting_context)); +} +#endif diff --git a/tools/perf/util/trace-event-perl.h b/tools/perf/util/trace-event-perl.h new file mode 100644 index 00000000000..6c94fa93013 --- /dev/null +++ b/tools/perf/util/trace-event-perl.h @@ -0,0 +1,42 @@ +#ifndef __PERF_TRACE_EVENT_PERL_H +#define __PERF_TRACE_EVENT_PERL_H +#ifdef NO_LIBPERL +typedef int INTERP; +#define dSP +#define ENTER +#define SAVETMPS +#define PUTBACK +#define SPAGAIN +#define FREETMPS +#define LEAVE +#define SP +#define ERRSV +#define G_SCALAR (0) +#define G_DISCARD (0) +#define G_NOARGS (0) +#define PUSHMARK(a) +#define SvTRUE(a) (0) +#define XPUSHs(s) +#define sv_2mortal(a) +#define newSVpv(a,b) +#define newSVuv(a) +#define newSViv(a) +#define get_cv(a,b) (0) +#define call_pv(a,b) (0) +#define perl_alloc() (0) +#define perl_construct(a) (0) +#define perl_parse(a,b,c,d,e) (0) +#define perl_run(a) (0) +#define perl_destruct(a) (0) +#define perl_free(a) (0) +#else +#include +#include +typedef PerlInterpreter * INTERP; +#endif + +struct scripting_context { + void *event_data; +}; + +#endif /* __PERF_TRACE_EVENT_PERL_H */ diff --git a/tools/perf/util/trace-event.h b/tools/perf/util/trace-event.h index aeb915778ae..b1e58d3d947 100644 --- a/tools/perf/util/trace-event.h +++ b/tools/perf/util/trace-event.h @@ -245,10 +245,14 @@ extern int latency_format; int parse_header_page(char *buf, unsigned long size); int trace_parse_common_type(void *data); +int trace_parse_common_pid(void *data); struct event *trace_find_event(int id); +struct event *trace_find_next_event(struct event *event); +unsigned long long read_size(void *ptr, int size); unsigned long long raw_field_value(struct event *event, const char *name, void *data); void *raw_field_ptr(struct event *event, const char *name, void *data); +unsigned long long eval_flag(const char *flag); int read_tracing_data(int fd, struct perf_event_attr *pattrs, int nb_events); @@ -272,4 +276,7 @@ struct scripting_ops { int script_spec_register(const char *spec, struct scripting_ops *ops); +extern struct scripting_ops perl_scripting_ops; +void setup_perl_scripting(void); + #endif /* __PERF_TRACE_EVENTS_H */ -- cgit v1.2.3-18-g5258 From bcefe12eff5dca6fdfa94ed85e5bee66380d5cd9 Mon Sep 17 00:00:00 2001 From: Tom Zanussi Date: Wed, 25 Nov 2009 01:15:49 -0600 Subject: perf trace: Add perf trace scripting support modules for Perl Add Perf-Trace-Util Perl module and some scripts that use it. Core.pm contains Perl code to define and access flag and symbolic fields. Util.pm contains general-purpose utility functions. Also adds some makefile bits to install them in libexec/perf-core/scripts/perl (or wherever perfexec_instdir points). Signed-off-by: Tom Zanussi Cc: fweisbec@gmail.com Cc: rostedt@goodmis.org Cc: anton@samba.org Cc: hch@infradead.org LKML-Reference: <1259133352-23685-5-git-send-email-tzanussi@gmail.com> Signed-off-by: Ingo Molnar --- tools/perf/Makefile | 7 + .../perf/scripts/perl/Perf-Trace-Util/Makefile.PL | 12 ++ tools/perf/scripts/perl/Perf-Trace-Util/README | 35 +++++ .../perl/Perf-Trace-Util/lib/Perf/Trace/Core.pm | 157 +++++++++++++++++++ .../perl/Perf-Trace-Util/lib/Perf/Trace/Util.pm | 88 +++++++++++ tools/perf/scripts/perl/rw-by-file.pl | 105 +++++++++++++ tools/perf/scripts/perl/rw-by-pid.pl | 170 +++++++++++++++++++++ tools/perf/scripts/perl/wakeup-latency.pl | 103 +++++++++++++ tools/perf/scripts/perl/workqueue-stats.pl | 129 ++++++++++++++++ 9 files changed, 806 insertions(+) create mode 100644 tools/perf/scripts/perl/Perf-Trace-Util/Makefile.PL create mode 100644 tools/perf/scripts/perl/Perf-Trace-Util/README create mode 100644 tools/perf/scripts/perl/Perf-Trace-Util/lib/Perf/Trace/Core.pm create mode 100644 tools/perf/scripts/perl/Perf-Trace-Util/lib/Perf/Trace/Util.pm create mode 100644 tools/perf/scripts/perl/rw-by-file.pl create mode 100644 tools/perf/scripts/perl/rw-by-pid.pl create mode 100644 tools/perf/scripts/perl/wakeup-latency.pl create mode 100644 tools/perf/scripts/perl/workqueue-stats.pl (limited to 'tools/perf') diff --git a/tools/perf/Makefile b/tools/perf/Makefile index 19e37cd14ae..efbc0e86421 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -980,6 +980,13 @@ export perfexec_instdir install: all $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(bindir_SQ)' $(INSTALL) perf$X '$(DESTDIR_SQ)$(bindir_SQ)' + $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/scripts/perl/Perf-Trace-Util/lib/Perf/Trace' + $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/scripts/perl/bin' + $(INSTALL) scripts/perl/Perf-Trace-Util/lib/Perf/Trace/* -t '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/scripts/perl/Perf-Trace-Util/lib/Perf/Trace' + $(INSTALL) scripts/perl/*.pl -t '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/scripts/perl' + $(INSTALL) scripts/perl/bin/* -t '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/scripts/perl/bin' + $(INSTALL) scripts/perl/Perf-Trace-Util/Makefile.PL -t '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/scripts/perl/Perf-Trace-Util' + $(INSTALL) scripts/perl/Perf-Trace-Util/README -t '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/scripts/perl/Perf-Trace-Util' ifdef BUILT_INS $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)' $(INSTALL) $(BUILT_INS) '$(DESTDIR_SQ)$(perfexec_instdir_SQ)' diff --git a/tools/perf/scripts/perl/Perf-Trace-Util/Makefile.PL b/tools/perf/scripts/perl/Perf-Trace-Util/Makefile.PL new file mode 100644 index 00000000000..b0de02e6950 --- /dev/null +++ b/tools/perf/scripts/perl/Perf-Trace-Util/Makefile.PL @@ -0,0 +1,12 @@ +use 5.010000; +use ExtUtils::MakeMaker; +# See lib/ExtUtils/MakeMaker.pm for details of how to influence +# the contents of the Makefile that is written. +WriteMakefile( + NAME => 'Perf::Trace::Util', + VERSION_FROM => 'lib/Perf/Trace/Util.pm', # finds $VERSION + PREREQ_PM => {}, # e.g., Module::Name => 1.1 + ($] >= 5.005 ? ## Add these new keywords supported since 5.005 + (ABSTRACT_FROM => 'lib/Perf/Trace/Util.pm', # retrieve abstract from module + AUTHOR => 'Tom Zanussi ') : ()), +); diff --git a/tools/perf/scripts/perl/Perf-Trace-Util/README b/tools/perf/scripts/perl/Perf-Trace-Util/README new file mode 100644 index 00000000000..0a58378f083 --- /dev/null +++ b/tools/perf/scripts/perl/Perf-Trace-Util/README @@ -0,0 +1,35 @@ +Perf-Trace-Util version 0.01 +============================ + +This module contains utility functions for use with perf trace. + +INSTALLATION + +Building perf with perf trace Perl scripting should install this +module in the right place. + +You should make sure libperl is installed first e.g. apt-get install +libperl-dev. + +DEPENDENCIES + +This module requires these other modules and libraries: + + blah blah blah + +COPYRIGHT AND LICENCE + +Put the correct copyright and licence information here. + +Copyright (C) 2009 by Tom Zanussi + +This library is free software; you can redistribute it and/or modify +it under the same terms as Perl itself, either Perl version 5.10.0 or, +at your option, any later version of Perl 5 you may have available. + +Alternatively, this software may be distributed under the terms of the +GNU General Public License ("GPL") version 2 as published by the Free +Software Foundation. + + + diff --git a/tools/perf/scripts/perl/Perf-Trace-Util/lib/Perf/Trace/Core.pm b/tools/perf/scripts/perl/Perf-Trace-Util/lib/Perf/Trace/Core.pm new file mode 100644 index 00000000000..fd250fb7be1 --- /dev/null +++ b/tools/perf/scripts/perl/Perf-Trace-Util/lib/Perf/Trace/Core.pm @@ -0,0 +1,157 @@ +package Perf::Trace::Core; + +use 5.010000; +use strict; +use warnings; + +require Exporter; + +our @ISA = qw(Exporter); + +our %EXPORT_TAGS = ( 'all' => [ qw( +) ] ); + +our @EXPORT_OK = ( @{ $EXPORT_TAGS{'all'} } ); + +our @EXPORT = qw( +define_flag_field define_flag_value flag_str dump_flag_fields +define_symbolic_field define_symbolic_value symbol_str dump_symbolic_fields +); + +our $VERSION = '0.01'; + +my %flag_fields; +my %symbolic_fields; + +sub flag_str +{ + my ($event_name, $field_name, $value) = @_; + + my $string; + + if ($flag_fields{$event_name}{$field_name}) { + my $print_delim = 0; + foreach my $idx (sort {$a <=> $b} keys %{$flag_fields{$event_name}{$field_name}{"values"}}) { + if (!$value && !$idx) { + $string .= "$flag_fields{$event_name}{$field_name}{'values'}{$idx}"; + last; + } + if ($idx && ($value & $idx) == $idx) { + if ($print_delim && $flag_fields{$event_name}{$field_name}{'delim'}) { + $string .= " $flag_fields{$event_name}{$field_name}{'delim'} "; + } + $string .= "$flag_fields{$event_name}{$field_name}{'values'}{$idx}"; + $print_delim = 1; + $value &= ~$idx; + } + } + } + + return $string; +} + +sub define_flag_field +{ + my ($event_name, $field_name, $delim) = @_; + + $flag_fields{$event_name}{$field_name}{"delim"} = $delim; +} + +sub define_flag_value +{ + my ($event_name, $field_name, $value, $field_str) = @_; + + $flag_fields{$event_name}{$field_name}{"values"}{$value} = $field_str; +} + +sub dump_flag_fields +{ + for my $event (keys %flag_fields) { + print "event $event:\n"; + for my $field (keys %{$flag_fields{$event}}) { + print " field: $field:\n"; + print " delim: $flag_fields{$event}{$field}{'delim'}\n"; + foreach my $idx (sort {$a <=> $b} keys %{$flag_fields{$event}{$field}{"values"}}) { + print " value $idx: $flag_fields{$event}{$field}{'values'}{$idx}\n"; + } + } + } +} + +sub symbol_str +{ + my ($event_name, $field_name, $value) = @_; + + if ($symbolic_fields{$event_name}{$field_name}) { + foreach my $idx (sort {$a <=> $b} keys %{$symbolic_fields{$event_name}{$field_name}{"values"}}) { + if (!$value && !$idx) { + return "$symbolic_fields{$event_name}{$field_name}{'values'}{$idx}"; + last; + } + if ($value == $idx) { + return "$symbolic_fields{$event_name}{$field_name}{'values'}{$idx}"; + } + } + } + + return undef; +} + +sub define_symbolic_field +{ + my ($event_name, $field_name) = @_; + + # nothing to do, really +} + +sub define_symbolic_value +{ + my ($event_name, $field_name, $value, $field_str) = @_; + + $symbolic_fields{$event_name}{$field_name}{"values"}{$value} = $field_str; +} + +sub dump_symbolic_fields +{ + for my $event (keys %symbolic_fields) { + print "event $event:\n"; + for my $field (keys %{$symbolic_fields{$event}}) { + print " field: $field:\n"; + foreach my $idx (sort {$a <=> $b} keys %{$symbolic_fields{$event}{$field}{"values"}}) { + print " value $idx: $symbolic_fields{$event}{$field}{'values'}{$idx}\n"; + } + } + } +} + +1; +__END__ +=head1 NAME + +Perf::Trace::Core - Perl extension for perf trace + +=head1 SYNOPSIS + + use Perf::Trace::Core + +=head1 SEE ALSO + +Perf (trace) documentation + +=head1 AUTHOR + +Tom Zanussi, Etzanussi@gmail.com + +=head1 COPYRIGHT AND LICENSE + +Copyright (C) 2009 by Tom Zanussi + +This library is free software; you can redistribute it and/or modify +it under the same terms as Perl itself, either Perl version 5.10.0 or, +at your option, any later version of Perl 5 you may have available. + +Alternatively, this software may be distributed under the terms of the +GNU General Public License ("GPL") version 2 as published by the Free +Software Foundation. + +=cut diff --git a/tools/perf/scripts/perl/Perf-Trace-Util/lib/Perf/Trace/Util.pm b/tools/perf/scripts/perl/Perf-Trace-Util/lib/Perf/Trace/Util.pm new file mode 100644 index 00000000000..052f132ced2 --- /dev/null +++ b/tools/perf/scripts/perl/Perf-Trace-Util/lib/Perf/Trace/Util.pm @@ -0,0 +1,88 @@ +package Perf::Trace::Util; + +use 5.010000; +use strict; +use warnings; + +require Exporter; + +our @ISA = qw(Exporter); + +our %EXPORT_TAGS = ( 'all' => [ qw( +) ] ); + +our @EXPORT_OK = ( @{ $EXPORT_TAGS{'all'} } ); + +our @EXPORT = qw( +avg nsecs nsecs_secs nsecs_nsecs nsecs_usecs print_nsecs +); + +our $VERSION = '0.01'; + +sub avg +{ + my ($total, $n) = @_; + + return $total / $n; +} + +my $NSECS_PER_SEC = 1000000000; + +sub nsecs +{ + my ($secs, $nsecs) = @_; + + return $secs * $NSECS_PER_SEC + $nsecs; +} + +sub nsecs_secs { + my ($nsecs) = @_; + + return $nsecs / $NSECS_PER_SEC; +} + +sub nsecs_nsecs { + my ($nsecs) = @_; + + return $nsecs - nsecs_secs($nsecs); +} + +sub nsecs_str { + my ($nsecs) = @_; + + my $str = sprintf("%5u.%09u", nsecs_secs($nsecs), nsecs_nsecs($nsecs)); + + return $str; +} + +1; +__END__ +=head1 NAME + +Perf::Trace::Util - Perl extension for perf trace + +=head1 SYNOPSIS + + use Perf::Trace::Util; + +=head1 SEE ALSO + +Perf (trace) documentation + +=head1 AUTHOR + +Tom Zanussi, Etzanussi@gmail.com + +=head1 COPYRIGHT AND LICENSE + +Copyright (C) 2009 by Tom Zanussi + +This library is free software; you can redistribute it and/or modify +it under the same terms as Perl itself, either Perl version 5.10.0 or, +at your option, any later version of Perl 5 you may have available. + +Alternatively, this software may be distributed under the terms of the +GNU General Public License ("GPL") version 2 as published by the Free +Software Foundation. + +=cut diff --git a/tools/perf/scripts/perl/rw-by-file.pl b/tools/perf/scripts/perl/rw-by-file.pl new file mode 100644 index 00000000000..61f91561d84 --- /dev/null +++ b/tools/perf/scripts/perl/rw-by-file.pl @@ -0,0 +1,105 @@ +#!/usr/bin/perl -w +# (c) 2009, Tom Zanussi +# Licensed under the terms of the GNU GPL License version 2 + +# Display r/w activity for files read/written to for a given program + +# The common_* event handler fields are the most useful fields common to +# all events. They don't necessarily correspond to the 'common_*' fields +# in the status files. Those fields not available as handler params can +# be retrieved via script functions of the form get_common_*(). + +use 5.010000; +use strict; +use warnings; + +use lib "$ENV{'PERF_EXEC_PATH'}/scripts/perl/Perf-Trace-Util/lib"; +use lib "./Perf-Trace-Util/lib"; +use Perf::Trace::Core; +use Perf::Trace::Util; + +# change this to the comm of the program you're interested in +my $for_comm = "perf"; + +my %reads; +my %writes; + +sub syscalls::sys_enter_read +{ + my ($event_name, $context, $common_cpu, $common_secs, $common_nsecs, + $common_pid, $common_comm, $nr, $fd, $buf, $count) = @_; + + if ($common_comm eq $for_comm) { + $reads{$fd}{bytes_requested} += $count; + $reads{$fd}{total_reads}++; + } +} + +sub syscalls::sys_enter_write +{ + my ($event_name, $context, $common_cpu, $common_secs, $common_nsecs, + $common_pid, $common_comm, $nr, $fd, $buf, $count) = @_; + + if ($common_comm eq $for_comm) { + $writes{$fd}{bytes_written} += $count; + $writes{$fd}{total_writes}++; + } +} + +sub trace_end +{ + printf("file read counts for $for_comm:\n\n"); + + printf("%6s %10s %10s\n", "fd", "# reads", "bytes_requested"); + printf("%6s %10s %10s\n", "------", "----------", "-----------"); + + foreach my $fd (sort {$reads{$b}{bytes_requested} <=> + $reads{$a}{bytes_requested}} keys %reads) { + my $total_reads = $reads{$fd}{total_reads}; + my $bytes_requested = $reads{$fd}{bytes_requested}; + printf("%6u %10u %10u\n", $fd, $total_reads, $bytes_requested); + } + + printf("\nfile write counts for $for_comm:\n\n"); + + printf("%6s %10s %10s\n", "fd", "# writes", "bytes_written"); + printf("%6s %10s %10s\n", "------", "----------", "-----------"); + + foreach my $fd (sort {$writes{$b}{bytes_written} <=> + $writes{$a}{bytes_written}} keys %writes) { + my $total_writes = $writes{$fd}{total_writes}; + my $bytes_written = $writes{$fd}{bytes_written}; + printf("%6u %10u %10u\n", $fd, $total_writes, $bytes_written); + } + + print_unhandled(); +} + +my %unhandled; + +sub print_unhandled +{ + if ((scalar keys %unhandled) == 0) { + return; + } + + print "\nunhandled events:\n\n"; + + printf("%-40s %10s\n", "event", "count"); + printf("%-40s %10s\n", "----------------------------------------", + "-----------"); + + foreach my $event_name (keys %unhandled) { + printf("%-40s %10d\n", $event_name, $unhandled{$event_name}); + } +} + +sub trace_unhandled +{ + my ($event_name, $context, $common_cpu, $common_secs, $common_nsecs, + $common_pid, $common_comm) = @_; + + $unhandled{$event_name}++; +} + + diff --git a/tools/perf/scripts/perl/rw-by-pid.pl b/tools/perf/scripts/perl/rw-by-pid.pl new file mode 100644 index 00000000000..da601fae1a0 --- /dev/null +++ b/tools/perf/scripts/perl/rw-by-pid.pl @@ -0,0 +1,170 @@ +#!/usr/bin/perl -w +# (c) 2009, Tom Zanussi +# Licensed under the terms of the GNU GPL License version 2 + +# Display r/w activity for all processes + +# The common_* event handler fields are the most useful fields common to +# all events. They don't necessarily correspond to the 'common_*' fields +# in the status files. Those fields not available as handler params can +# be retrieved via script functions of the form get_common_*(). + +use 5.010000; +use strict; +use warnings; + +use lib "$ENV{'PERF_EXEC_PATH'}/scripts/perl/Perf-Trace-Util/lib"; +use lib "./Perf-Trace-Util/lib"; +use Perf::Trace::Core; +use Perf::Trace::Util; + +my %reads; +my %writes; + +sub syscalls::sys_exit_read +{ + my ($event_name, $context, $common_cpu, $common_secs, $common_nsecs, + $common_pid, $common_comm, + $nr, $ret) = @_; + + if ($ret > 0) { + $reads{$common_pid}{bytes_read} += $ret; + } else { + if (!defined ($reads{$common_pid}{bytes_read})) { + $reads{$common_pid}{bytes_read} = 0; + } + $reads{$common_pid}{errors}{$ret}++; + } +} + +sub syscalls::sys_enter_read +{ + my ($event_name, $context, $common_cpu, $common_secs, $common_nsecs, + $common_pid, $common_comm, + $nr, $fd, $buf, $count) = @_; + + $reads{$common_pid}{bytes_requested} += $count; + $reads{$common_pid}{total_reads}++; + $reads{$common_pid}{comm} = $common_comm; +} + +sub syscalls::sys_exit_write +{ + my ($event_name, $context, $common_cpu, $common_secs, $common_nsecs, + $common_pid, $common_comm, + $nr, $ret) = @_; + + if ($ret <= 0) { + $writes{$common_pid}{errors}{$ret}++; + } +} + +sub syscalls::sys_enter_write +{ + my ($event_name, $context, $common_cpu, $common_secs, $common_nsecs, + $common_pid, $common_comm, + $nr, $fd, $buf, $count) = @_; + + $writes{$common_pid}{bytes_written} += $count; + $writes{$common_pid}{total_writes}++; + $writes{$common_pid}{comm} = $common_comm; +} + +sub trace_end +{ + printf("read counts by pid:\n\n"); + + printf("%6s %20s %10s %10s %10s\n", "pid", "comm", + "# reads", "bytes_requested", "bytes_read"); + printf("%6s %-20s %10s %10s %10s\n", "------", "--------------------", + "-----------", "----------", "----------"); + + foreach my $pid (sort {$reads{$b}{bytes_read} <=> + $reads{$a}{bytes_read}} keys %reads) { + my $comm = $reads{$pid}{comm}; + my $total_reads = $reads{$pid}{total_reads}; + my $bytes_requested = $reads{$pid}{bytes_requested}; + my $bytes_read = $reads{$pid}{bytes_read}; + + printf("%6s %-20s %10s %10s %10s\n", $pid, $comm, + $total_reads, $bytes_requested, $bytes_read); + } + + printf("\nfailed reads by pid:\n\n"); + + printf("%6s %20s %6s %10s\n", "pid", "comm", "error #", "# errors"); + printf("%6s %20s %6s %10s\n", "------", "--------------------", + "------", "----------"); + + foreach my $pid (keys %reads) { + my $comm = $reads{$pid}{comm}; + foreach my $err (sort {$reads{$b}{comm} cmp $reads{$a}{comm}} + keys %{$reads{$pid}{errors}}) { + my $errors = $reads{$pid}{errors}{$err}; + + printf("%6d %-20s %6d %10s\n", $pid, $comm, $err, $errors); + } + } + + printf("\nwrite counts by pid:\n\n"); + + printf("%6s %20s %10s %10s\n", "pid", "comm", + "# writes", "bytes_written"); + printf("%6s %-20s %10s %10s\n", "------", "--------------------", + "-----------", "----------"); + + foreach my $pid (sort {$writes{$b}{bytes_written} <=> + $writes{$a}{bytes_written}} keys %writes) { + my $comm = $writes{$pid}{comm}; + my $total_writes = $writes{$pid}{total_writes}; + my $bytes_written = $writes{$pid}{bytes_written}; + + printf("%6s %-20s %10s %10s\n", $pid, $comm, + $total_writes, $bytes_written); + } + + printf("\nfailed writes by pid:\n\n"); + + printf("%6s %20s %6s %10s\n", "pid", "comm", "error #", "# errors"); + printf("%6s %20s %6s %10s\n", "------", "--------------------", + "------", "----------"); + + foreach my $pid (keys %writes) { + my $comm = $writes{$pid}{comm}; + foreach my $err (sort {$writes{$b}{comm} cmp $writes{$a}{comm}} + keys %{$writes{$pid}{errors}}) { + my $errors = $writes{$pid}{errors}{$err}; + + printf("%6d %-20s %6d %10s\n", $pid, $comm, $err, $errors); + } + } + + print_unhandled(); +} + +my %unhandled; + +sub print_unhandled +{ + if ((scalar keys %unhandled) == 0) { + return; + } + + print "\nunhandled events:\n\n"; + + printf("%-40s %10s\n", "event", "count"); + printf("%-40s %10s\n", "----------------------------------------", + "-----------"); + + foreach my $event_name (keys %unhandled) { + printf("%-40s %10d\n", $event_name, $unhandled{$event_name}); + } +} + +sub trace_unhandled +{ + my ($event_name, $context, $common_cpu, $common_secs, $common_nsecs, + $common_pid, $common_comm) = @_; + + $unhandled{$event_name}++; +} diff --git a/tools/perf/scripts/perl/wakeup-latency.pl b/tools/perf/scripts/perl/wakeup-latency.pl new file mode 100644 index 00000000000..ed58ef284e2 --- /dev/null +++ b/tools/perf/scripts/perl/wakeup-latency.pl @@ -0,0 +1,103 @@ +#!/usr/bin/perl -w +# (c) 2009, Tom Zanussi +# Licensed under the terms of the GNU GPL License version 2 + +# Display avg/min/max wakeup latency + +# The common_* event handler fields are the most useful fields common to +# all events. They don't necessarily correspond to the 'common_*' fields +# in the status files. Those fields not available as handler params can +# be retrieved via script functions of the form get_common_*(). + +use 5.010000; +use strict; +use warnings; + +use lib "$ENV{'PERF_EXEC_PATH'}/scripts/perl/Perf-Trace-Util/lib"; +use lib "./Perf-Trace-Util/lib"; +use Perf::Trace::Core; +use Perf::Trace::Util; + +my %last_wakeup; + +my $max_wakeup_latency; +my $min_wakeup_latency; +my $total_wakeup_latency; +my $total_wakeups; + +sub sched::sched_switch +{ + my ($event_name, $context, $common_cpu, $common_secs, $common_nsecs, + $common_pid, $common_comm, + $prev_comm, $prev_pid, $prev_prio, $prev_state, $next_comm, $next_pid, + $next_prio) = @_; + + my $wakeup_ts = $last_wakeup{$common_cpu}{ts}; + if ($wakeup_ts) { + my $switch_ts = nsecs($common_secs, $common_nsecs); + my $wakeup_latency = $switch_ts - $wakeup_ts; + if ($wakeup_latency > $max_wakeup_latency) { + $max_wakeup_latency = $wakeup_latency; + } + if ($wakeup_latency < $min_wakeup_latency) { + $min_wakeup_latency = $wakeup_latency; + } + $total_wakeup_latency += $wakeup_latency; + $total_wakeups++; + } + $last_wakeup{$common_cpu}{ts} = 0; +} + +sub sched::sched_wakeup +{ + my ($event_name, $context, $common_cpu, $common_secs, $common_nsecs, + $common_pid, $common_comm, + $comm, $pid, $prio, $success, $target_cpu) = @_; + + $last_wakeup{$target_cpu}{ts} = nsecs($common_secs, $common_nsecs); +} + +sub trace_begin +{ + $min_wakeup_latency = 1000000000; + $max_wakeup_latency = 0; +} + +sub trace_end +{ + printf("wakeup_latency stats:\n\n"); + print "total_wakeups: $total_wakeups\n"; + printf("avg_wakeup_latency (ns): %u\n", + avg($total_wakeup_latency, $total_wakeups)); + printf("min_wakeup_latency (ns): %u\n", $min_wakeup_latency); + printf("max_wakeup_latency (ns): %u\n", $max_wakeup_latency); + + print_unhandled(); +} + +my %unhandled; + +sub print_unhandled +{ + if ((scalar keys %unhandled) == 0) { + return; + } + + print "\nunhandled events:\n\n"; + + printf("%-40s %10s\n", "event", "count"); + printf("%-40s %10s\n", "----------------------------------------", + "-----------"); + + foreach my $event_name (keys %unhandled) { + printf("%-40s %10d\n", $event_name, $unhandled{$event_name}); + } +} + +sub trace_unhandled +{ + my ($event_name, $context, $common_cpu, $common_secs, $common_nsecs, + $common_pid, $common_comm) = @_; + + $unhandled{$event_name}++; +} diff --git a/tools/perf/scripts/perl/workqueue-stats.pl b/tools/perf/scripts/perl/workqueue-stats.pl new file mode 100644 index 00000000000..511302c8a49 --- /dev/null +++ b/tools/perf/scripts/perl/workqueue-stats.pl @@ -0,0 +1,129 @@ +#!/usr/bin/perl -w +# (c) 2009, Tom Zanussi +# Licensed under the terms of the GNU GPL License version 2 + +# Displays workqueue stats +# +# Usage: +# +# perf record -c 1 -f -a -R -e workqueue:workqueue_creation -e +# workqueue:workqueue_destruction -e workqueue:workqueue_execution +# -e workqueue:workqueue_insertion +# +# perf trace -p -s tools/perf/scripts/perl/workqueue-stats.pl + +use 5.010000; +use strict; +use warnings; + +use lib "$ENV{'PERF_EXEC_PATH'}/scripts/perl/Perf-Trace-Util/lib"; +use lib "./Perf-Trace-Util/lib"; +use Perf::Trace::Core; +use Perf::Trace::Util; + +my @cpus; + +sub workqueue::workqueue_destruction +{ + my ($event_name, $context, $common_cpu, $common_secs, $common_nsecs, + $common_pid, $common_comm, + $thread_comm, $thread_pid) = @_; + + $cpus[$common_cpu]{$thread_pid}{destroyed}++; + $cpus[$common_cpu]{$thread_pid}{comm} = $thread_comm; +} + +sub workqueue::workqueue_creation +{ + my ($event_name, $context, $common_cpu, $common_secs, $common_nsecs, + $common_pid, $common_comm, + $thread_comm, $thread_pid, $cpu) = @_; + + $cpus[$common_cpu]{$thread_pid}{created}++; + $cpus[$common_cpu]{$thread_pid}{comm} = $thread_comm; +} + +sub workqueue::workqueue_execution +{ + my ($event_name, $context, $common_cpu, $common_secs, $common_nsecs, + $common_pid, $common_comm, + $thread_comm, $thread_pid, $func) = @_; + + $cpus[$common_cpu]{$thread_pid}{executed}++; + $cpus[$common_cpu]{$thread_pid}{comm} = $thread_comm; +} + +sub workqueue::workqueue_insertion +{ + my ($event_name, $context, $common_cpu, $common_secs, $common_nsecs, + $common_pid, $common_comm, + $thread_comm, $thread_pid, $func) = @_; + + $cpus[$common_cpu]{$thread_pid}{inserted}++; + $cpus[$common_cpu]{$thread_pid}{comm} = $thread_comm; +} + +sub trace_end +{ + print "workqueue work stats:\n\n"; + my $cpu = 0; + printf("%3s %6s %6s\t%-20s\n", "cpu", "ins", "exec", "name"); + printf("%3s %6s %6s\t%-20s\n", "---", "---", "----", "----"); + foreach my $pidhash (@cpus) { + while ((my $pid, my $wqhash) = each %$pidhash) { + my $ins = $$wqhash{'inserted'}; + my $exe = $$wqhash{'executed'}; + my $comm = $$wqhash{'comm'}; + if ($ins || $exe) { + printf("%3u %6u %6u\t%-20s\n", $cpu, $ins, $exe, $comm); + } + } + $cpu++; + } + + $cpu = 0; + print "\nworkqueue lifecycle stats:\n\n"; + printf("%3s %6s %6s\t%-20s\n", "cpu", "created", "destroyed", "name"); + printf("%3s %6s %6s\t%-20s\n", "---", "-------", "---------", "----"); + foreach my $pidhash (@cpus) { + while ((my $pid, my $wqhash) = each %$pidhash) { + my $created = $$wqhash{'created'}; + my $destroyed = $$wqhash{'destroyed'}; + my $comm = $$wqhash{'comm'}; + if ($created || $destroyed) { + printf("%3u %6u %6u\t%-20s\n", $cpu, $created, $destroyed, + $comm); + } + } + $cpu++; + } + + print_unhandled(); +} + +my %unhandled; + +sub print_unhandled +{ + if ((scalar keys %unhandled) == 0) { + return; + } + + print "\nunhandled events:\n\n"; + + printf("%-40s %10s\n", "event", "count"); + printf("%-40s %10s\n", "----------------------------------------", + "-----------"); + + foreach my $event_name (keys %unhandled) { + printf("%-40s %10d\n", $event_name, $unhandled{$event_name}); + } +} + +sub trace_unhandled +{ + my ($event_name, $context, $common_cpu, $common_secs, $common_nsecs, + $common_pid, $common_comm) = @_; + + $unhandled{$event_name}++; +} -- cgit v1.2.3-18-g5258 From d1b93772be78486397693fc39d3ddea3fda90105 Mon Sep 17 00:00:00 2001 From: Tom Zanussi Date: Wed, 25 Nov 2009 01:15:50 -0600 Subject: perf trace: Add interface to access perf data from Perl handlers The Perl scripting support for perf trace allows most of a trace event's data to be accessed directly as handler arguments, but not all of it e.g. the less common fields aren't passed in. To give scripts access to the other fields and/or any other data or metadata in the main perf executable that might be useful, a way to access the C data in perf from Perl is needed; this patch uses the Perl XS facility to do it for the common_xxx event fields not passed to handler functions. Context.pm exports three functions to Perl scripts that access fields for the current event by calling back into perf: common_pc(), common_flags() and common_lock_depth(). Support for common_flags() field values was added to Core.pm and a script used to sanity check these and other basic scripting features, check-perf-trace.pl, was also added. Signed-off-by: Tom Zanussi Cc: fweisbec@gmail.com Cc: rostedt@goodmis.org Cc: anton@samba.org Cc: hch@infradead.org LKML-Reference: <1259133352-23685-6-git-send-email-tzanussi@gmail.com> Signed-off-by: Ingo Molnar --- tools/perf/Makefile | 6 +- tools/perf/scripts/perl/Perf-Trace-Util/Context.c | 134 +++++++++++++++++++++ tools/perf/scripts/perl/Perf-Trace-Util/Context.xs | 41 +++++++ .../perf/scripts/perl/Perf-Trace-Util/Makefile.PL | 11 +- tools/perf/scripts/perl/Perf-Trace-Util/README | 34 +++++- .../perl/Perf-Trace-Util/lib/Perf/Trace/Context.pm | 55 +++++++++ .../perl/Perf-Trace-Util/lib/Perf/Trace/Core.pm | 35 ++++++ tools/perf/scripts/perl/Perf-Trace-Util/typemap | 1 + tools/perf/scripts/perl/check-perf-trace.pl | 106 ++++++++++++++++ tools/perf/util/trace-event-parse.c | 6 +- tools/perf/util/trace-event-perl.c | 46 ++++++- tools/perf/util/trace-event-perl.h | 9 ++ tools/perf/util/trace-event.h | 3 + 13 files changed, 474 insertions(+), 13 deletions(-) create mode 100644 tools/perf/scripts/perl/Perf-Trace-Util/Context.c create mode 100644 tools/perf/scripts/perl/Perf-Trace-Util/Context.xs create mode 100644 tools/perf/scripts/perl/Perf-Trace-Util/lib/Perf/Trace/Context.pm create mode 100644 tools/perf/scripts/perl/Perf-Trace-Util/typemap create mode 100644 tools/perf/scripts/perl/check-perf-trace.pl (limited to 'tools/perf') diff --git a/tools/perf/Makefile b/tools/perf/Makefile index efbc0e86421..8ad57b51d64 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -497,6 +497,7 @@ ifneq ($(shell sh -c "(echo '\#include '; echo '\#include '; e BASIC_CFLAGS += -DNO_LIBPERL else ALL_LDFLAGS += $(PERL_EMBED_LDOPTS) + LIB_OBJS += scripts/perl/Perf-Trace-Util/Context.o endif ifdef NO_DEMANGLE @@ -873,6 +874,9 @@ util/find_next_bit.o: ../../lib/find_next_bit.c PERF-CFLAGS util/trace-event-perl.o: util/trace-event-perl.c PERF-CFLAGS $(QUIET_CC)$(CC) -o util/trace-event-perl.o -c $(ALL_CFLAGS) $(PERL_EMBED_CCOPTS) -Wno-redundant-decls -Wno-strict-prototypes -Wno-unused-parameter $< +scripts/perl/Perf-Trace-Util/Context.o: scripts/perl/Perf-Trace-Util/Context.c PERF-CFLAGS + $(QUIET_CC)$(CC) -o scripts/perl/Perf-Trace-Util/Context.o -c $(ALL_CFLAGS) $(PERL_EMBED_CCOPTS) -Wno-redundant-decls -Wno-strict-prototypes -Wno-unused-parameter -Wno-nested-externs $< + perf-%$X: %.o $(PERFLIBS) $(QUIET_LINK)$(CC) $(ALL_CFLAGS) -o $@ $(ALL_LDFLAGS) $(filter %.o,$^) $(LIBS) @@ -1072,7 +1076,7 @@ distclean: clean # $(RM) configure clean: - $(RM) *.o */*.o $(LIB_FILE) + $(RM) *.o */*.o */*/*.o */*/*/*.o $(LIB_FILE) $(RM) $(ALL_PROGRAMS) $(BUILT_INS) perf$X $(RM) $(TEST_PROGRAMS) $(RM) *.spec *.pyc *.pyo */*.pyc */*.pyo common-cmds.h TAGS tags cscope* diff --git a/tools/perf/scripts/perl/Perf-Trace-Util/Context.c b/tools/perf/scripts/perl/Perf-Trace-Util/Context.c new file mode 100644 index 00000000000..3ba3ffc5416 --- /dev/null +++ b/tools/perf/scripts/perl/Perf-Trace-Util/Context.c @@ -0,0 +1,134 @@ +/* + * This file was generated automatically by ExtUtils::ParseXS version 2.18_02 from the + * contents of Context.xs. Do not edit this file, edit Context.xs instead. + * + * ANY CHANGES MADE HERE WILL BE LOST! + * + */ + +#line 1 "Context.xs" +/* + * Context.xs. XS interfaces for perf trace. + * + * Copyright (C) 2009 Tom Zanussi + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + * + */ + +#include "EXTERN.h" +#include "perl.h" +#include "XSUB.h" +#include "../../../util/trace-event-perl.h" + +#ifndef PERL_UNUSED_VAR +# define PERL_UNUSED_VAR(var) if (0) var = var +#endif + +#line 41 "Context.c" + +XS(XS_Perf__Trace__Context_get_common_pc); /* prototype to pass -Wmissing-prototypes */ +XS(XS_Perf__Trace__Context_get_common_pc) +{ +#ifdef dVAR + dVAR; dXSARGS; +#else + dXSARGS; +#endif + if (items != 1) + Perl_croak(aTHX_ "Usage: %s(%s)", "Perf::Trace::Context::get_common_pc", "context"); + PERL_UNUSED_VAR(cv); /* -W */ + { + struct scripting_context * context = INT2PTR(struct scripting_context *,SvIV(ST(0))); + int RETVAL; + dXSTARG; + + RETVAL = get_common_pc(context); + XSprePUSH; PUSHi((IV)RETVAL); + } + XSRETURN(1); +} + + +XS(XS_Perf__Trace__Context_get_common_flags); /* prototype to pass -Wmissing-prototypes */ +XS(XS_Perf__Trace__Context_get_common_flags) +{ +#ifdef dVAR + dVAR; dXSARGS; +#else + dXSARGS; +#endif + if (items != 1) + Perl_croak(aTHX_ "Usage: %s(%s)", "Perf::Trace::Context::get_common_flags", "context"); + PERL_UNUSED_VAR(cv); /* -W */ + { + struct scripting_context * context = INT2PTR(struct scripting_context *,SvIV(ST(0))); + int RETVAL; + dXSTARG; + + RETVAL = get_common_flags(context); + XSprePUSH; PUSHi((IV)RETVAL); + } + XSRETURN(1); +} + + +XS(XS_Perf__Trace__Context_get_common_lock_depth); /* prototype to pass -Wmissing-prototypes */ +XS(XS_Perf__Trace__Context_get_common_lock_depth) +{ +#ifdef dVAR + dVAR; dXSARGS; +#else + dXSARGS; +#endif + if (items != 1) + Perl_croak(aTHX_ "Usage: %s(%s)", "Perf::Trace::Context::get_common_lock_depth", "context"); + PERL_UNUSED_VAR(cv); /* -W */ + { + struct scripting_context * context = INT2PTR(struct scripting_context *,SvIV(ST(0))); + int RETVAL; + dXSTARG; + + RETVAL = get_common_lock_depth(context); + XSprePUSH; PUSHi((IV)RETVAL); + } + XSRETURN(1); +} + +#ifdef __cplusplus +extern "C" +#endif +XS(boot_Perf__Trace__Context); /* prototype to pass -Wmissing-prototypes */ +XS(boot_Perf__Trace__Context) +{ +#ifdef dVAR + dVAR; dXSARGS; +#else + dXSARGS; +#endif + const char* file = __FILE__; + + PERL_UNUSED_VAR(cv); /* -W */ + PERL_UNUSED_VAR(items); /* -W */ + XS_VERSION_BOOTCHECK ; + + newXSproto("Perf::Trace::Context::get_common_pc", XS_Perf__Trace__Context_get_common_pc, file, "$"); + newXSproto("Perf::Trace::Context::get_common_flags", XS_Perf__Trace__Context_get_common_flags, file, "$"); + newXSproto("Perf::Trace::Context::get_common_lock_depth", XS_Perf__Trace__Context_get_common_lock_depth, file, "$"); + if (PL_unitcheckav) + call_list(PL_scopestack_ix, PL_unitcheckav); + XSRETURN_YES; +} + diff --git a/tools/perf/scripts/perl/Perf-Trace-Util/Context.xs b/tools/perf/scripts/perl/Perf-Trace-Util/Context.xs new file mode 100644 index 00000000000..24facb3696d --- /dev/null +++ b/tools/perf/scripts/perl/Perf-Trace-Util/Context.xs @@ -0,0 +1,41 @@ +/* + * Context.xs. XS interfaces for perf trace. + * + * Copyright (C) 2009 Tom Zanussi + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + * + */ + +#include "EXTERN.h" +#include "perl.h" +#include "XSUB.h" +#include "../../../util/trace-event-perl.h" + +MODULE = Perf::Trace::Context PACKAGE = Perf::Trace::Context +PROTOTYPES: ENABLE + +int +get_common_pc(context) + struct scripting_context * context + +int +get_common_flags(context) + struct scripting_context * context + +int +get_common_lock_depth(context) + struct scripting_context * context + diff --git a/tools/perf/scripts/perl/Perf-Trace-Util/Makefile.PL b/tools/perf/scripts/perl/Perf-Trace-Util/Makefile.PL index b0de02e6950..decdeb0f678 100644 --- a/tools/perf/scripts/perl/Perf-Trace-Util/Makefile.PL +++ b/tools/perf/scripts/perl/Perf-Trace-Util/Makefile.PL @@ -3,10 +3,15 @@ use ExtUtils::MakeMaker; # See lib/ExtUtils/MakeMaker.pm for details of how to influence # the contents of the Makefile that is written. WriteMakefile( - NAME => 'Perf::Trace::Util', - VERSION_FROM => 'lib/Perf/Trace/Util.pm', # finds $VERSION + NAME => 'Perf::Trace::Context', + VERSION_FROM => 'lib/Perf/Trace/Context.pm', # finds $VERSION PREREQ_PM => {}, # e.g., Module::Name => 1.1 ($] >= 5.005 ? ## Add these new keywords supported since 5.005 - (ABSTRACT_FROM => 'lib/Perf/Trace/Util.pm', # retrieve abstract from module + (ABSTRACT_FROM => 'lib/Perf/Trace/Context.pm', # retrieve abstract from module AUTHOR => 'Tom Zanussi ') : ()), + LIBS => [''], # e.g., '-lm' + DEFINE => '-I ../..', # e.g., '-DHAVE_SOMETHING' + INC => '-I.', # e.g., '-I. -I/usr/include/other' + # Un-comment this if you add C files to link with later: + OBJECT => 'Context.o', # link all the C files too ); diff --git a/tools/perf/scripts/perl/Perf-Trace-Util/README b/tools/perf/scripts/perl/Perf-Trace-Util/README index 0a58378f083..adb99aa3a7b 100644 --- a/tools/perf/scripts/perl/Perf-Trace-Util/README +++ b/tools/perf/scripts/perl/Perf-Trace-Util/README @@ -3,6 +3,34 @@ Perf-Trace-Util version 0.01 This module contains utility functions for use with perf trace. +Core.pm and Util.pm are pure Perl modules; Core.pm contains routines +that the core perf support for Perl calls on and should always be +'used', while Util.pm contains useful but optional utility functions +that scripts may want to use. Context.pm contains the Perl->C +interface that allows scripts to access data in the embedding perf +executable; scripts wishing to do that should 'use Context.pm'. + +The Perl->C perf interface is completely driven by Context.xs. If you +want to add new Perl functions that end up accessing C data in the +perf executable, you add desciptions of the new functions here. +scripting_context is a pointer to the perf data in the perf executable +that you want to access - it's passed as the second parameter, +$context, to all handler functions. + +After you do that: + + perl Makefile.PL # to create a Makefile for the next step + make # to create Context.c + + edit Context.c to add const to the char* file = __FILE__ line in + XS(boot_Perf__Trace__Context) to silence a warning/error. + + You can delete the Makefile, object files and anything else that was + generated e.g. blib and shared library, etc, except for of course + Context.c + + You should then be able to run the normal perf make as usual. + INSTALLATION Building perf with perf trace Perl scripting should install this @@ -15,12 +43,10 @@ DEPENDENCIES This module requires these other modules and libraries: - blah blah blah + None COPYRIGHT AND LICENCE -Put the correct copyright and licence information here. - Copyright (C) 2009 by Tom Zanussi This library is free software; you can redistribute it and/or modify @@ -31,5 +57,3 @@ Alternatively, this software may be distributed under the terms of the GNU General Public License ("GPL") version 2 as published by the Free Software Foundation. - - diff --git a/tools/perf/scripts/perl/Perf-Trace-Util/lib/Perf/Trace/Context.pm b/tools/perf/scripts/perl/Perf-Trace-Util/lib/Perf/Trace/Context.pm new file mode 100644 index 00000000000..6c7f3659cb1 --- /dev/null +++ b/tools/perf/scripts/perl/Perf-Trace-Util/lib/Perf/Trace/Context.pm @@ -0,0 +1,55 @@ +package Perf::Trace::Context; + +use 5.010000; +use strict; +use warnings; + +require Exporter; + +our @ISA = qw(Exporter); + +our %EXPORT_TAGS = ( 'all' => [ qw( +) ] ); + +our @EXPORT_OK = ( @{ $EXPORT_TAGS{'all'} } ); + +our @EXPORT = qw( + common_pc common_flags common_lock_depth +); + +our $VERSION = '0.01'; + +require XSLoader; +XSLoader::load('Perf::Trace::Context', $VERSION); + +1; +__END__ +=head1 NAME + +Perf::Trace::Context - Perl extension for accessing functions in perf. + +=head1 SYNOPSIS + + use Perf::Trace::Context; + +=head1 SEE ALSO + +Perf (trace) documentation + +=head1 AUTHOR + +Tom Zanussi, Etzanussi@gmail.com + +=head1 COPYRIGHT AND LICENSE + +Copyright (C) 2009 by Tom Zanussi + +This library is free software; you can redistribute it and/or modify +it under the same terms as Perl itself, either Perl version 5.10.0 or, +at your option, any later version of Perl 5 you may have available. + +Alternatively, this software may be distributed under the terms of the +GNU General Public License ("GPL") version 2 as published by the Free +Software Foundation. + +=cut diff --git a/tools/perf/scripts/perl/Perf-Trace-Util/lib/Perf/Trace/Core.pm b/tools/perf/scripts/perl/Perf-Trace-Util/lib/Perf/Trace/Core.pm index fd250fb7be1..9df376a9f62 100644 --- a/tools/perf/scripts/perl/Perf-Trace-Util/lib/Perf/Trace/Core.pm +++ b/tools/perf/scripts/perl/Perf-Trace-Util/lib/Perf/Trace/Core.pm @@ -16,10 +16,45 @@ our @EXPORT_OK = ( @{ $EXPORT_TAGS{'all'} } ); our @EXPORT = qw( define_flag_field define_flag_value flag_str dump_flag_fields define_symbolic_field define_symbolic_value symbol_str dump_symbolic_fields +trace_flag_str ); our $VERSION = '0.01'; +my %trace_flags = (0x00 => "NONE", + 0x01 => "IRQS_OFF", + 0x02 => "IRQS_NOSUPPORT", + 0x04 => "NEED_RESCHED", + 0x08 => "HARDIRQ", + 0x10 => "SOFTIRQ"); + +sub trace_flag_str +{ + my ($value) = @_; + + my $string; + + my $print_delim = 0; + + foreach my $idx (sort {$a <=> $b} keys %trace_flags) { + if (!$value && !$idx) { + $string .= "NONE"; + last; + } + + if ($idx && ($value & $idx) == $idx) { + if ($print_delim) { + $string .= " | "; + } + $string .= "$trace_flags{$idx}"; + $print_delim = 1; + $value &= ~$idx; + } + } + + return $string; +} + my %flag_fields; my %symbolic_fields; diff --git a/tools/perf/scripts/perl/Perf-Trace-Util/typemap b/tools/perf/scripts/perl/Perf-Trace-Util/typemap new file mode 100644 index 00000000000..840836804aa --- /dev/null +++ b/tools/perf/scripts/perl/Perf-Trace-Util/typemap @@ -0,0 +1 @@ +struct scripting_context * T_PTR diff --git a/tools/perf/scripts/perl/check-perf-trace.pl b/tools/perf/scripts/perl/check-perf-trace.pl new file mode 100644 index 00000000000..4e7dc0a407a --- /dev/null +++ b/tools/perf/scripts/perl/check-perf-trace.pl @@ -0,0 +1,106 @@ +# perf trace event handlers, generated by perf trace -g perl +# (c) 2009, Tom Zanussi +# Licensed under the terms of the GNU GPL License version 2 + +# This script tests basic functionality such as flag and symbol +# strings, common_xxx() calls back into perf, begin, end, unhandled +# events, etc. Basically, if this script runs successfully and +# displays expected results, perl scripting support should be ok. + +use lib "$ENV{'PERF_EXEC_PATH'}/scripts/perl/Perf-Trace-Util/lib"; +use lib "./Perf-Trace-Util/lib"; +use Perf::Trace::Core; +use Perf::Trace::Context; +use Perf::Trace::Util; + +sub trace_begin +{ + print "trace_begin\n"; +} + +sub trace_end +{ + print "trace_end\n"; + + print_unhandled(); +} + +sub irq::softirq_entry +{ + my ($event_name, $context, $common_cpu, $common_secs, $common_nsecs, + $common_pid, $common_comm, + $vec) = @_; + + print_header($event_name, $common_cpu, $common_secs, $common_nsecs, + $common_pid, $common_comm); + + print_uncommon($context); + + printf("vec=%s\n", + symbol_str("irq::softirq_entry", "vec", $vec)); +} + +sub kmem::kmalloc +{ + my ($event_name, $context, $common_cpu, $common_secs, $common_nsecs, + $common_pid, $common_comm, + $call_site, $ptr, $bytes_req, $bytes_alloc, + $gfp_flags) = @_; + + print_header($event_name, $common_cpu, $common_secs, $common_nsecs, + $common_pid, $common_comm); + + print_uncommon($context); + + printf("call_site=%p, ptr=%p, bytes_req=%u, bytes_alloc=%u, ". + "gfp_flags=%s\n", + $call_site, $ptr, $bytes_req, $bytes_alloc, + + flag_str("kmem::kmalloc", "gfp_flags", $gfp_flags)); +} + +# print trace fields not included in handler args +sub print_uncommon +{ + my ($context) = @_; + + printf("common_preempt_count=%d, common_flags=%s, common_lock_depth=%d, ", + common_pc($context), trace_flag_str(common_flags($context)), + common_lock_depth($context)); + +} + +my %unhandled; + +sub print_unhandled +{ + if ((scalar keys %unhandled) == 0) { + return; + } + + print "\nunhandled events:\n\n"; + + printf("%-40s %10s\n", "event", "count"); + printf("%-40s %10s\n", "----------------------------------------", + "-----------"); + + foreach my $event_name (keys %unhandled) { + printf("%-40s %10d\n", $event_name, $unhandled{$event_name}); + } +} + +sub trace_unhandled +{ + my ($event_name, $context, $common_cpu, $common_secs, $common_nsecs, + $common_pid, $common_comm) = @_; + + $unhandled{$event_name}++; +} + +sub print_header +{ + my ($event_name, $cpu, $secs, $nsecs, $pid, $comm) = @_; + + printf("%-20s %5u %05u.%09u %8u %-20s ", + $event_name, $cpu, $secs, $nsecs, $pid, $comm); +} diff --git a/tools/perf/util/trace-event-parse.c b/tools/perf/util/trace-event-parse.c index 1f16495e559..0302405aa2c 100644 --- a/tools/perf/util/trace-event-parse.c +++ b/tools/perf/util/trace-event-parse.c @@ -1982,7 +1982,7 @@ int trace_parse_common_pid(void *data) "common_pid"); } -static int parse_common_pc(void *data) +int parse_common_pc(void *data) { static int pc_offset; static int pc_size; @@ -1991,7 +1991,7 @@ static int parse_common_pc(void *data) "common_preempt_count"); } -static int parse_common_flags(void *data) +int parse_common_flags(void *data) { static int flags_offset; static int flags_size; @@ -2000,7 +2000,7 @@ static int parse_common_flags(void *data) "common_flags"); } -static int parse_common_lock_depth(void *data) +int parse_common_lock_depth(void *data) { static int ld_offset; static int ld_size; diff --git a/tools/perf/util/trace-event-perl.c b/tools/perf/util/trace-event-perl.c index c56b08d704d..d179adebc54 100644 --- a/tools/perf/util/trace-event-perl.c +++ b/tools/perf/util/trace-event-perl.c @@ -30,6 +30,21 @@ #include "trace-event.h" #include "trace-event-perl.h" +void xs_init(pTHX); + +void boot_Perf__Trace__Context(pTHX_ CV *cv); +void boot_DynaLoader(pTHX_ CV *cv); + +void xs_init(pTHX) +{ + const char *file = __FILE__; + dXSUB_SYS; + + newXS("Perf::Trace::Context::bootstrap", boot_Perf__Trace__Context, + file); + newXS("DynaLoader::boot_DynaLoader", boot_DynaLoader, file); +} + INTERP my_perl; #define FTRACE_MAX_EVENT \ @@ -227,6 +242,33 @@ static inline struct event *find_cache_event(int type) return event; } +int get_common_pc(struct scripting_context *context) +{ + int pc; + + pc = parse_common_pc(context->event_data); + + return pc; +} + +int get_common_flags(struct scripting_context *context) +{ + int flags; + + flags = parse_common_flags(context->event_data); + + return flags; +} + +int get_common_lock_depth(struct scripting_context *context) +{ + int lock_depth; + + lock_depth = parse_common_lock_depth(context->event_data); + + return lock_depth; +} + static void perl_process_event(int cpu, void *data, int size __attribute((unused)), unsigned long long nsecs, char *comm) @@ -290,6 +332,7 @@ static void perl_process_event(int cpu, void *data, } PUTBACK; + if (get_cv(handler, 0)) call_pv(handler, G_SCALAR); else if (get_cv("main::trace_unhandled", 0)) { @@ -328,7 +371,8 @@ static int perl_start_script(const char *script) my_perl = perl_alloc(); perl_construct(my_perl); - if (perl_parse(my_perl, NULL, 2, (char **)command_line, (char **)NULL)) + if (perl_parse(my_perl, xs_init, 2, (char **)command_line, + (char **)NULL)) return -1; perl_run(my_perl); diff --git a/tools/perf/util/trace-event-perl.h b/tools/perf/util/trace-event-perl.h index 6c94fa93013..666a864f5dd 100644 --- a/tools/perf/util/trace-event-perl.h +++ b/tools/perf/util/trace-event-perl.h @@ -29,6 +29,11 @@ typedef int INTERP; #define perl_run(a) (0) #define perl_destruct(a) (0) #define perl_free(a) (0) +#define pTHX void +#define CV void +#define dXSUB_SYS +#define pTHX_ +static inline void newXS(const char *a, void *b, const char *c) {} #else #include #include @@ -39,4 +44,8 @@ struct scripting_context { void *event_data; }; +int get_common_pc(struct scripting_context *context); +int get_common_flags(struct scripting_context *context); +int get_common_lock_depth(struct scripting_context *context); + #endif /* __PERF_TRACE_EVENT_PERL_H */ diff --git a/tools/perf/util/trace-event.h b/tools/perf/util/trace-event.h index b1e58d3d947..81698d5e650 100644 --- a/tools/perf/util/trace-event.h +++ b/tools/perf/util/trace-event.h @@ -246,6 +246,9 @@ extern int latency_format; int parse_header_page(char *buf, unsigned long size); int trace_parse_common_type(void *data); int trace_parse_common_pid(void *data); +int parse_common_pc(void *data); +int parse_common_flags(void *data); +int parse_common_lock_depth(void *data); struct event *trace_find_event(int id); struct event *trace_find_next_event(struct event *event); unsigned long long read_size(void *ptr, int size); -- cgit v1.2.3-18-g5258 From 89fbf0b8a021cbf60abeacfb6b538e97c83afada Mon Sep 17 00:00:00 2001 From: Tom Zanussi Date: Wed, 25 Nov 2009 01:15:51 -0600 Subject: perf trace: Add Documentation for perf trace Perl support Adds perf-trace-perl Documentation and a link to it from the perf-trace page. Signed-off-by: Tom Zanussi Cc: fweisbec@gmail.com Cc: rostedt@goodmis.org Cc: anton@samba.org Cc: hch@infradead.org LKML-Reference: <1259133352-23685-7-git-send-email-tzanussi@gmail.com> Signed-off-by: Ingo Molnar --- tools/perf/Documentation/perf-trace-perl.txt | 219 +++++++++++++++++++++++++++ tools/perf/Documentation/perf-trace.txt | 11 +- 2 files changed, 229 insertions(+), 1 deletion(-) create mode 100644 tools/perf/Documentation/perf-trace-perl.txt (limited to 'tools/perf') diff --git a/tools/perf/Documentation/perf-trace-perl.txt b/tools/perf/Documentation/perf-trace-perl.txt new file mode 100644 index 00000000000..c5f55f43909 --- /dev/null +++ b/tools/perf/Documentation/perf-trace-perl.txt @@ -0,0 +1,219 @@ +perf-trace-perl(1) +================== + +NAME +---- +perf-trace-perl - Process trace data with a Perl script + +SYNOPSIS +-------- +[verse] +'perf trace' [-s [lang]:script[.ext] ] + +DESCRIPTION +----------- + +This perf trace option is used to process perf trace data using perf's +built-in Perl interpreter. It reads and processes the input file and +displays the results of the trace analysis implemented in the given +Perl script, if any. + +STARTER SCRIPTS +--------------- + +You can avoid reading the rest of this document by running 'perf trace +-g perl' in the same directory as an existing perf.data trace file. +That will generate a starter script containing a handler for each of +the event types in the trace file; it simply prints every available +field for each event in the trace file. + +You can also look at the existing scripts in +~/libexec/perf-core/scripts/perl for typical examples showing how to +do basic things like aggregate event data, print results, etc. Also, +the check-perf-trace.pl script, while not interesting for its results, +attempts to exercise all of the main scripting features. + +EVENT HANDLERS +-------------- + +When perf trace is invoked using a trace script, a user-defined +'handler function' is called for each event in the trace. If there's +no handler function defined for a given event type, the event is +ignored (or passed to a 'trace_handled' function, see below) and the +next event is processed. + +Most of the event's field values are passed as arguments to the +handler function; some of the less common ones aren't - those are +available as calls back into the perf executable (see below). + +As an example, the following perf record command can be used to record +all sched_wakeup events in the system: + + # perf record -c 1 -f -a -M -R -e sched:sched_wakeup + +Traces meant to be processed using a script should be recorded with +the above options: -c 1 says to sample every event, -a to enable +system-wide collection, -M to multiplex the output, and -R to collect +raw samples. + +The format file for the sched_wakep event defines the following fields +(see /sys/kernel/debug/tracing/events/sched/sched_wakeup/format): + +---- + format: + field:unsigned short common_type; + field:unsigned char common_flags; + field:unsigned char common_preempt_count; + field:int common_pid; + field:int common_lock_depth; + + field:char comm[TASK_COMM_LEN]; + field:pid_t pid; + field:int prio; + field:int success; + field:int target_cpu; +---- + +The handler function for this event would be defined as: + +---- +sub sched::sched_wakeup +{ + my ($event_name, $context, $common_cpu, $common_secs, + $common_nsecs, $common_pid, $common_comm, + $comm, $pid, $prio, $success, $target_cpu) = @_; +} +---- + +The handler function takes the form subsystem::event_name. + +The $common_* arguments in the handler's argument list are the set of +arguments passed to all event handlers; some of the fields correspond +to the common_* fields in the format file, but some are synthesized, +and some of the common_* fields aren't common enough to to be passed +to every event as arguments but are available as library functions. + +Here's a brief description of each of the invariant event args: + + $event_name the name of the event as text + $context an opaque 'cookie' used in calls back into perf + $common_cpu the cpu the event occurred on + $common_secs the secs portion of the event timestamp + $common_nsecs the nsecs portion of the event timestamp + $common_pid the pid of the current task + $common_comm the name of the current process + +All of the remaining fields in the event's format file have +counterparts as handler function arguments of the same name, as can be +seen in the example above. + +The above provides the basics needed to directly access every field of +every event in a trace, which covers 90% of what you need to know to +write a useful trace script. The sections below cover the rest. + +SCRIPT LAYOUT +------------- + +Every perf trace Perl script should start by setting up a Perl module +search path and 'use'ing a few support modules (see module +descriptions below): + +---- + use lib "$ENV{'PERF_EXEC_PATH'}/scripts/perl/Perf-Trace-Util/lib"; + use lib "./Perf-Trace-Util/lib"; + use Perf::Trace::Core; + use Perf::Trace::Context; + use Perf::Trace::Util; +---- + +The rest of the script can contain handler functions and support +functions in any order. + +Aside from the event handler functions discussed above, every script +can implement a set of optional functions: + +*trace_begin*, if defined, is called before any event is processed and +gives scripts a chance to do setup tasks: + +---- + sub trace_begin + { + } +---- + +*trace_end*, if defined, is called after all events have been + processed and gives scripts a chance to do end-of-script tasks, such + as display results: + +---- +sub trace_end +{ +} +---- + +*trace_unhandled*, if defined, is called after for any event that + doesn't have a handler explicitly defined for it. The standard set + of common arguments are passed into it: + +---- +sub trace_unhandled +{ + my ($event_name, $context, $common_cpu, $common_secs, + $common_nsecs, $common_pid, $common_comm) = @_; +} +---- + +The remaining sections provide descriptions of each of the available +built-in perf trace Perl modules and their associated functions. + +AVAILABLE MODULES AND FUNCTIONS +------------------------------- + +The following sections describe the functions and variables available +via the various Perf::Trace::* Perl modules. To use the functions and +variables from the given module, add the corresponding 'use +Perf::Trace::XXX' line to your perf trace script. + +Perf::Trace::Core Module +~~~~~~~~~~~~~~~~~~~~~~~~ + +These functions provide some essential functions to user scripts. + +The *flag_str* and *symbol_str* functions provide human-readable +strings for flag and symbolic fields. These correspond to the strings +and values parsed from the 'print fmt' fields of the event format +files: + + flag_str($event_name, $field_name, $field_value) - returns the string represention corresponding to $field_value for the flag field $field_name of event $event_name + symbol_str($event_name, $field_name, $field_value) - returns the string represention corresponding to $field_value for the symbolic field $field_name of event $event_name + +Perf::Trace::Context Module +~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +Some of the 'common' fields in the event format file aren't all that +common, but need to be made accessible to user scripts nonetheless. + +Perf::Trace::Context defines a set of functions that can be used to +access this data in the context of the current event. Each of these +functions expects a $context variable, which is the same as the +$context variable passed into every event handler as the second +argument. + + common_pc($context) - returns common_preempt count for the current event + common_flags($context) - returns common_flags for the current event + common_lock_depth($context) - returns common_lock_depth for the current event + +Perf::Trace::Util Module +~~~~~~~~~~~~~~~~~~~~~~~~ + +Various utility functions for use with perf trace: + + nsecs($secs, $nsecs) - returns total nsecs given secs/nsecs pair + nsecs_secs($nsecs) - returns whole secs portion given nsecs + nsecs_nsecs($nsecs) - returns nsecs remainder given nsecs + nsecs_str($nsecs) - returns printable string in the form secs.nsecs + avg($total, $n) - returns average given a sum and a total number of values + +SEE ALSO +-------- +linkperf:perf-trace[1] diff --git a/tools/perf/Documentation/perf-trace.txt b/tools/perf/Documentation/perf-trace.txt index 41ed75398ca..07065efa60e 100644 --- a/tools/perf/Documentation/perf-trace.txt +++ b/tools/perf/Documentation/perf-trace.txt @@ -20,6 +20,15 @@ OPTIONS --dump-raw-trace=:: Display verbose dump of the trace data. +-s:: +--script=:: + Process trace data with the given script ([lang]:script[.ext]). + +-g:: +--gen-script=:: + Generate perf-trace.[ext] starter script for given language, + using current perf.data. + SEE ALSO -------- -linkperf:perf-record[1] +linkperf:perf-record[1], linkperf:perf-trace-perl[1] -- cgit v1.2.3-18-g5258 From 1ae4a971250c55e473ca53c78011fcf73809885d Mon Sep 17 00:00:00 2001 From: Tom Zanussi Date: Wed, 25 Nov 2009 01:15:52 -0600 Subject: perf trace: Add a scripts/perl/bin for perf trace shell scripts To capture the relevant events for a given Perl script and to avoid having to continually remember and type in long command-lines, add a scripts/perl/bin directory containing two simple shell scripts for each Perl script, one for recording and one for processing/display. For example, to record perf data for the rw-by-pid.pl script, run scripts/perl/bin/rw-by-pid-record and to actually run the script and display the output run scripts/perl/bin/rw-by-pid-report. Signed-off-by: Tom Zanussi Cc: fweisbec@gmail.com Cc: rostedt@goodmis.org Cc: anton@samba.org Cc: hch@infradead.org LKML-Reference: <1259133352-23685-8-git-send-email-tzanussi@gmail.com> Signed-off-by: Ingo Molnar --- tools/perf/scripts/perl/bin/check-perf-trace-record | 7 +++++++ tools/perf/scripts/perl/bin/check-perf-trace-report | 5 +++++ tools/perf/scripts/perl/bin/rw-by-file-record | 2 ++ tools/perf/scripts/perl/bin/rw-by-file-report | 5 +++++ tools/perf/scripts/perl/bin/rw-by-pid-record | 2 ++ tools/perf/scripts/perl/bin/rw-by-pid-report | 5 +++++ tools/perf/scripts/perl/bin/wakeup-latency-record | 6 ++++++ tools/perf/scripts/perl/bin/wakeup-latency-report | 5 +++++ tools/perf/scripts/perl/bin/workqueue-stats-record | 2 ++ tools/perf/scripts/perl/bin/workqueue-stats-report | 6 ++++++ 10 files changed, 45 insertions(+) create mode 100644 tools/perf/scripts/perl/bin/check-perf-trace-record create mode 100644 tools/perf/scripts/perl/bin/check-perf-trace-report create mode 100644 tools/perf/scripts/perl/bin/rw-by-file-record create mode 100644 tools/perf/scripts/perl/bin/rw-by-file-report create mode 100644 tools/perf/scripts/perl/bin/rw-by-pid-record create mode 100644 tools/perf/scripts/perl/bin/rw-by-pid-report create mode 100644 tools/perf/scripts/perl/bin/wakeup-latency-record create mode 100644 tools/perf/scripts/perl/bin/wakeup-latency-report create mode 100644 tools/perf/scripts/perl/bin/workqueue-stats-record create mode 100644 tools/perf/scripts/perl/bin/workqueue-stats-report (limited to 'tools/perf') diff --git a/tools/perf/scripts/perl/bin/check-perf-trace-record b/tools/perf/scripts/perl/bin/check-perf-trace-record new file mode 100644 index 00000000000..c7ec5de2f53 --- /dev/null +++ b/tools/perf/scripts/perl/bin/check-perf-trace-record @@ -0,0 +1,7 @@ +#!/bin/bash +perf record -c 1 -f -a -M -R -e kmem:kmalloc -e irq:softirq_entry + + + + + diff --git a/tools/perf/scripts/perl/bin/check-perf-trace-report b/tools/perf/scripts/perl/bin/check-perf-trace-report new file mode 100644 index 00000000000..89948b01502 --- /dev/null +++ b/tools/perf/scripts/perl/bin/check-perf-trace-report @@ -0,0 +1,5 @@ +#!/bin/bash +perf trace -s ~/libexec/perf-core/scripts/perl/check-perf-trace.pl + + + diff --git a/tools/perf/scripts/perl/bin/rw-by-file-record b/tools/perf/scripts/perl/bin/rw-by-file-record new file mode 100644 index 00000000000..b25056ebf96 --- /dev/null +++ b/tools/perf/scripts/perl/bin/rw-by-file-record @@ -0,0 +1,2 @@ +#!/bin/bash +perf record -c 1 -f -a -M -R -e syscalls:sys_enter_read -e syscalls:sys_enter_write diff --git a/tools/perf/scripts/perl/bin/rw-by-file-report b/tools/perf/scripts/perl/bin/rw-by-file-report new file mode 100644 index 00000000000..f5dcf9cb5bd --- /dev/null +++ b/tools/perf/scripts/perl/bin/rw-by-file-report @@ -0,0 +1,5 @@ +#!/bin/bash +perf trace -s ~/libexec/perf-core/scripts/perl/rw-by-file.pl + + + diff --git a/tools/perf/scripts/perl/bin/rw-by-pid-record b/tools/perf/scripts/perl/bin/rw-by-pid-record new file mode 100644 index 00000000000..8903979c5b6 --- /dev/null +++ b/tools/perf/scripts/perl/bin/rw-by-pid-record @@ -0,0 +1,2 @@ +#!/bin/bash +perf record -c 1 -f -a -M -R -e syscalls:sys_enter_read -e syscalls:sys_exit_read -e syscalls:sys_enter_write -e syscalls:sys_exit_write diff --git a/tools/perf/scripts/perl/bin/rw-by-pid-report b/tools/perf/scripts/perl/bin/rw-by-pid-report new file mode 100644 index 00000000000..cea16f78a3a --- /dev/null +++ b/tools/perf/scripts/perl/bin/rw-by-pid-report @@ -0,0 +1,5 @@ +#!/bin/bash +perf trace -s ~/libexec/perf-core/scripts/perl/rw-by-pid.pl + + + diff --git a/tools/perf/scripts/perl/bin/wakeup-latency-record b/tools/perf/scripts/perl/bin/wakeup-latency-record new file mode 100644 index 00000000000..6abedda911a --- /dev/null +++ b/tools/perf/scripts/perl/bin/wakeup-latency-record @@ -0,0 +1,6 @@ +#!/bin/bash +perf record -c 1 -f -a -M -R -e sched:sched_switch -e sched:sched_wakeup + + + + diff --git a/tools/perf/scripts/perl/bin/wakeup-latency-report b/tools/perf/scripts/perl/bin/wakeup-latency-report new file mode 100644 index 00000000000..85769dc456e --- /dev/null +++ b/tools/perf/scripts/perl/bin/wakeup-latency-report @@ -0,0 +1,5 @@ +#!/bin/bash +perf trace -s ~/libexec/perf-core/scripts/perl/wakeup-latency.pl + + + diff --git a/tools/perf/scripts/perl/bin/workqueue-stats-record b/tools/perf/scripts/perl/bin/workqueue-stats-record new file mode 100644 index 00000000000..fce6637b19b --- /dev/null +++ b/tools/perf/scripts/perl/bin/workqueue-stats-record @@ -0,0 +1,2 @@ +#!/bin/bash +perf record -c 1 -f -a -M -R -e workqueue:workqueue_creation -e workqueue:workqueue_destruction -e workqueue:workqueue_execution -e workqueue:workqueue_insertion diff --git a/tools/perf/scripts/perl/bin/workqueue-stats-report b/tools/perf/scripts/perl/bin/workqueue-stats-report new file mode 100644 index 00000000000..aa68435be92 --- /dev/null +++ b/tools/perf/scripts/perl/bin/workqueue-stats-report @@ -0,0 +1,6 @@ +#!/bin/bash +perf trace -s ~/libexec/perf-core/scripts/perl/workqueue-stats.pl + + + + -- cgit v1.2.3-18-g5258 From cf72344d1ad7b33805ef8d65e758b267e6f4cb8d Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Sat, 28 Nov 2009 10:11:00 +0100 Subject: perf scripting: Fix build Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker LKML-Reference: Signed-off-by: Ingo Molnar --- tools/perf/builtin-trace.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'tools/perf') diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-trace.c index ca8ebf1ec64..abb914aa7be 100644 --- a/tools/perf/builtin-trace.c +++ b/tools/perf/builtin-trace.c @@ -5,6 +5,8 @@ #include "util/symbol.h" #include "util/thread.h" #include "util/header.h" +#include "util/exec_cmd.h" +#include "util/trace-event.h" static char const *script_name; static char const *generate_script_lang; -- cgit v1.2.3-18-g5258 From f8be4231f82ab56a87ce74906671afbe1aa9ec75 Mon Sep 17 00:00:00 2001 From: Tom Zanussi Date: Mon, 30 Nov 2009 01:18:46 -0600 Subject: perf trace/scripting: Silence PERL_EMBED_* backtick errors The backtick shell substitutions for PERL_EMBED_LDOPT/CCOPT make a lot of noise on stderr if Embed.pm isn't installed - this silences them. Signed-off-by: Tom Zanussi Cc: fweisbec@gmail.com Cc: rostedt@goodmis.org Cc: anton@samba.org Cc: hch@infradead.org LKML-Reference: <1259565529-6407-2-git-send-email-tzanussi@gmail.com> Signed-off-by: Ingo Molnar --- tools/perf/Makefile | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/Makefile b/tools/perf/Makefile index 8ad57b51d64..d62a2d7ff4e 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -490,8 +490,8 @@ else LIB_OBJS += util/probe-finder.o endif -PERL_EMBED_LDOPTS = `perl -MExtUtils::Embed -e ldopts` -PERL_EMBED_CCOPTS = `perl -MExtUtils::Embed -e ccopts` +PERL_EMBED_LDOPTS = `perl -MExtUtils::Embed -e ldopts 2>/dev/null` +PERL_EMBED_CCOPTS = `perl -MExtUtils::Embed -e ccopts 2>/dev/null` ifneq ($(shell sh -c "(echo '\#include '; echo '\#include '; echo 'int main(void) { perl_alloc(); return 0; }') | $(CC) -x c - $(PERL_EMBED_CCOPTS) -o /dev/null $(PERL_EMBED_LDOPTS) > /dev/null 2>&1 && echo y"), y) BASIC_CFLAGS += -DNO_LIBPERL -- cgit v1.2.3-18-g5258 From e136323c5a8a7d91d17c5b7b340758bb9dd33739 Mon Sep 17 00:00:00 2001 From: Tom Zanussi Date: Mon, 30 Nov 2009 01:18:47 -0600 Subject: perf trace/scripting: Ignore shadowed variable warning for perf-trace-perl.c The debugging versions of the ENTER and LEAVE internal perl macros, used when embedding perl, define a local block with a my_perl perl variable that shadows a global variable of the same name, which is also the name expected by the embedding API for the embedded interpreter. Since we don't have control over the code generated in this case (it's an externality) and can't get rid of the warning, ignore it. Signed-off-by: Tom Zanussi Cc: fweisbec@gmail.com Cc: rostedt@goodmis.org Cc: anton@samba.org Cc: hch@infradead.org LKML-Reference: <1259565529-6407-3-git-send-email-tzanussi@gmail.com> Signed-off-by: Ingo Molnar --- tools/perf/Makefile | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'tools/perf') diff --git a/tools/perf/Makefile b/tools/perf/Makefile index d62a2d7ff4e..20cd66362d2 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -872,7 +872,7 @@ util/find_next_bit.o: ../../lib/find_next_bit.c PERF-CFLAGS $(QUIET_CC)$(CC) -o util/find_next_bit.o -c $(ALL_CFLAGS) -DETC_PERFCONFIG='"$(ETC_PERFCONFIG_SQ)"' $< util/trace-event-perl.o: util/trace-event-perl.c PERF-CFLAGS - $(QUIET_CC)$(CC) -o util/trace-event-perl.o -c $(ALL_CFLAGS) $(PERL_EMBED_CCOPTS) -Wno-redundant-decls -Wno-strict-prototypes -Wno-unused-parameter $< + $(QUIET_CC)$(CC) -o util/trace-event-perl.o -c $(ALL_CFLAGS) $(PERL_EMBED_CCOPTS) -Wno-redundant-decls -Wno-strict-prototypes -Wno-unused-parameter -Wno-shadow $< scripts/perl/Perf-Trace-Util/Context.o: scripts/perl/Perf-Trace-Util/Context.c PERF-CFLAGS $(QUIET_CC)$(CC) -o scripts/perl/Perf-Trace-Util/Context.o -c $(ALL_CFLAGS) $(PERL_EMBED_CCOPTS) -Wno-redundant-decls -Wno-strict-prototypes -Wno-unused-parameter -Wno-nested-externs $< -- cgit v1.2.3-18-g5258 From 61381de0504181368672a83d2e14c38dbaf3c136 Mon Sep 17 00:00:00 2001 From: Tom Zanussi Date: Mon, 30 Nov 2009 01:18:48 -0600 Subject: perf trace/scripting: Fix Perl common_* access functions The common_* functions (e.g. common_pc(), etc) are exported as common_* but named get_common_*, resulting in unresolved subroutine errors when executing scripts. Make the internal and external names match. Signed-off-by: Tom Zanussi Cc: fweisbec@gmail.com Cc: rostedt@goodmis.org Cc: anton@samba.org Cc: hch@infradead.org LKML-Reference: <1259565529-6407-4-git-send-email-tzanussi@gmail.com> Signed-off-by: Ingo Molnar --- tools/perf/scripts/perl/Perf-Trace-Util/Context.c | 30 +++++++++++----------- tools/perf/scripts/perl/Perf-Trace-Util/Context.xs | 6 ++--- tools/perf/util/trace-event-perl.c | 6 ++--- tools/perf/util/trace-event-perl.h | 6 ++--- 4 files changed, 24 insertions(+), 24 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/scripts/perl/Perf-Trace-Util/Context.c b/tools/perf/scripts/perl/Perf-Trace-Util/Context.c index 3ba3ffc5416..af78d9a52a7 100644 --- a/tools/perf/scripts/perl/Perf-Trace-Util/Context.c +++ b/tools/perf/scripts/perl/Perf-Trace-Util/Context.c @@ -39,8 +39,8 @@ #line 41 "Context.c" -XS(XS_Perf__Trace__Context_get_common_pc); /* prototype to pass -Wmissing-prototypes */ -XS(XS_Perf__Trace__Context_get_common_pc) +XS(XS_Perf__Trace__Context_common_pc); /* prototype to pass -Wmissing-prototypes */ +XS(XS_Perf__Trace__Context_common_pc) { #ifdef dVAR dVAR; dXSARGS; @@ -48,22 +48,22 @@ XS(XS_Perf__Trace__Context_get_common_pc) dXSARGS; #endif if (items != 1) - Perl_croak(aTHX_ "Usage: %s(%s)", "Perf::Trace::Context::get_common_pc", "context"); + Perl_croak(aTHX_ "Usage: %s(%s)", "Perf::Trace::Context::common_pc", "context"); PERL_UNUSED_VAR(cv); /* -W */ { struct scripting_context * context = INT2PTR(struct scripting_context *,SvIV(ST(0))); int RETVAL; dXSTARG; - RETVAL = get_common_pc(context); + RETVAL = common_pc(context); XSprePUSH; PUSHi((IV)RETVAL); } XSRETURN(1); } -XS(XS_Perf__Trace__Context_get_common_flags); /* prototype to pass -Wmissing-prototypes */ -XS(XS_Perf__Trace__Context_get_common_flags) +XS(XS_Perf__Trace__Context_common_flags); /* prototype to pass -Wmissing-prototypes */ +XS(XS_Perf__Trace__Context_common_flags) { #ifdef dVAR dVAR; dXSARGS; @@ -71,22 +71,22 @@ XS(XS_Perf__Trace__Context_get_common_flags) dXSARGS; #endif if (items != 1) - Perl_croak(aTHX_ "Usage: %s(%s)", "Perf::Trace::Context::get_common_flags", "context"); + Perl_croak(aTHX_ "Usage: %s(%s)", "Perf::Trace::Context::common_flags", "context"); PERL_UNUSED_VAR(cv); /* -W */ { struct scripting_context * context = INT2PTR(struct scripting_context *,SvIV(ST(0))); int RETVAL; dXSTARG; - RETVAL = get_common_flags(context); + RETVAL = common_flags(context); XSprePUSH; PUSHi((IV)RETVAL); } XSRETURN(1); } -XS(XS_Perf__Trace__Context_get_common_lock_depth); /* prototype to pass -Wmissing-prototypes */ -XS(XS_Perf__Trace__Context_get_common_lock_depth) +XS(XS_Perf__Trace__Context_common_lock_depth); /* prototype to pass -Wmissing-prototypes */ +XS(XS_Perf__Trace__Context_common_lock_depth) { #ifdef dVAR dVAR; dXSARGS; @@ -94,14 +94,14 @@ XS(XS_Perf__Trace__Context_get_common_lock_depth) dXSARGS; #endif if (items != 1) - Perl_croak(aTHX_ "Usage: %s(%s)", "Perf::Trace::Context::get_common_lock_depth", "context"); + Perl_croak(aTHX_ "Usage: %s(%s)", "Perf::Trace::Context::common_lock_depth", "context"); PERL_UNUSED_VAR(cv); /* -W */ { struct scripting_context * context = INT2PTR(struct scripting_context *,SvIV(ST(0))); int RETVAL; dXSTARG; - RETVAL = get_common_lock_depth(context); + RETVAL = common_lock_depth(context); XSprePUSH; PUSHi((IV)RETVAL); } XSRETURN(1); @@ -124,9 +124,9 @@ XS(boot_Perf__Trace__Context) PERL_UNUSED_VAR(items); /* -W */ XS_VERSION_BOOTCHECK ; - newXSproto("Perf::Trace::Context::get_common_pc", XS_Perf__Trace__Context_get_common_pc, file, "$"); - newXSproto("Perf::Trace::Context::get_common_flags", XS_Perf__Trace__Context_get_common_flags, file, "$"); - newXSproto("Perf::Trace::Context::get_common_lock_depth", XS_Perf__Trace__Context_get_common_lock_depth, file, "$"); + newXSproto("Perf::Trace::Context::common_pc", XS_Perf__Trace__Context_common_pc, file, "$"); + newXSproto("Perf::Trace::Context::common_flags", XS_Perf__Trace__Context_common_flags, file, "$"); + newXSproto("Perf::Trace::Context::common_lock_depth", XS_Perf__Trace__Context_common_lock_depth, file, "$"); if (PL_unitcheckav) call_list(PL_scopestack_ix, PL_unitcheckav); XSRETURN_YES; diff --git a/tools/perf/scripts/perl/Perf-Trace-Util/Context.xs b/tools/perf/scripts/perl/Perf-Trace-Util/Context.xs index 24facb3696d..fb78006c165 100644 --- a/tools/perf/scripts/perl/Perf-Trace-Util/Context.xs +++ b/tools/perf/scripts/perl/Perf-Trace-Util/Context.xs @@ -28,14 +28,14 @@ MODULE = Perf::Trace::Context PACKAGE = Perf::Trace::Context PROTOTYPES: ENABLE int -get_common_pc(context) +common_pc(context) struct scripting_context * context int -get_common_flags(context) +common_flags(context) struct scripting_context * context int -get_common_lock_depth(context) +common_lock_depth(context) struct scripting_context * context diff --git a/tools/perf/util/trace-event-perl.c b/tools/perf/util/trace-event-perl.c index d179adebc54..2e1cc3c11c7 100644 --- a/tools/perf/util/trace-event-perl.c +++ b/tools/perf/util/trace-event-perl.c @@ -242,7 +242,7 @@ static inline struct event *find_cache_event(int type) return event; } -int get_common_pc(struct scripting_context *context) +int common_pc(struct scripting_context *context) { int pc; @@ -251,7 +251,7 @@ int get_common_pc(struct scripting_context *context) return pc; } -int get_common_flags(struct scripting_context *context) +int common_flags(struct scripting_context *context) { int flags; @@ -260,7 +260,7 @@ int get_common_flags(struct scripting_context *context) return flags; } -int get_common_lock_depth(struct scripting_context *context) +int common_lock_depth(struct scripting_context *context) { int lock_depth; diff --git a/tools/perf/util/trace-event-perl.h b/tools/perf/util/trace-event-perl.h index 666a864f5dd..8fe0d866fe1 100644 --- a/tools/perf/util/trace-event-perl.h +++ b/tools/perf/util/trace-event-perl.h @@ -44,8 +44,8 @@ struct scripting_context { void *event_data; }; -int get_common_pc(struct scripting_context *context); -int get_common_flags(struct scripting_context *context); -int get_common_lock_depth(struct scripting_context *context); +int common_pc(struct scripting_context *context); +int common_flags(struct scripting_context *context); +int common_lock_depth(struct scripting_context *context); #endif /* __PERF_TRACE_EVENT_PERL_H */ -- cgit v1.2.3-18-g5258 From 8ea339adc0a48236008e59dd21564d71c37b331c Mon Sep 17 00:00:00 2001 From: Tom Zanussi Date: Mon, 30 Nov 2009 01:18:49 -0600 Subject: perf trace/scripting: Add Fedora libperl install note to doc Fedora needs perl-ExtUtils-Embed for Perl scripting, which also brings along libperl-devel; note this info for the convenience of Fedora users. Signed-off-by: Tom Zanussi Cc: fweisbec@gmail.com Cc: rostedt@goodmis.org Cc: anton@samba.org Cc: hch@infradead.org LKML-Reference: <1259565529-6407-5-git-send-email-tzanussi@gmail.com> Signed-off-by: Ingo Molnar --- tools/perf/scripts/perl/Perf-Trace-Util/README | 4 ++-- tools/perf/util/trace-event-perl.c | 4 +++- 2 files changed, 5 insertions(+), 3 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/scripts/perl/Perf-Trace-Util/README b/tools/perf/scripts/perl/Perf-Trace-Util/README index adb99aa3a7b..9a970763079 100644 --- a/tools/perf/scripts/perl/Perf-Trace-Util/README +++ b/tools/perf/scripts/perl/Perf-Trace-Util/README @@ -36,8 +36,8 @@ INSTALLATION Building perf with perf trace Perl scripting should install this module in the right place. -You should make sure libperl is installed first e.g. apt-get install -libperl-dev. +You should make sure libperl and ExtUtils/Embed.pm are installed first +e.g. apt-get install libperl-dev or yum install perl-ExtUtils-Embed. DEPENDENCIES diff --git a/tools/perf/util/trace-event-perl.c b/tools/perf/util/trace-event-perl.c index 2e1cc3c11c7..51e833fd58c 100644 --- a/tools/perf/util/trace-event-perl.c +++ b/tools/perf/util/trace-event-perl.c @@ -577,7 +577,9 @@ struct scripting_ops perl_scripting_ops = { void setup_perl_scripting(void) { fprintf(stderr, "Perl scripting not supported." - " Install libperl-dev[el] and rebuild perf to get it.\n"); + " Install libperl and rebuild perf to enable it. e.g. " + "apt-get install libperl-dev (ubuntu), yum install " + "perl-ExtUtils-Embed (Fedora), etc.\n"); } #else void setup_perl_scripting(void) -- cgit v1.2.3-18-g5258 From bab81b624e970f1138535a465ad2b26b6bb0dd6c Mon Sep 17 00:00:00 2001 From: Li Zefan Date: Tue, 1 Dec 2009 14:04:49 +0800 Subject: perf annotate: Fix perf data parsing perf-annotate doesn't parse perf.data correctly in that it doesn't read perf header. Fix this by using mmap_dispatch_perf_file(). Before: TOTAL events: 17565 MMAP events: 3221 LOST events: 10 COMM events: 235 EXIT events: 2 THROTTLE events: 1 UNTHROTTLE events: 2 FORK events: 10 READ events: 1 SAMPLE events: 14083 After: TOTAL events: 17290 MMAP events: 3203 LOST events: 0 COMM events: 234 EXIT events: 1 THROTTLE events: 0 UNTHROTTLE events: 0 FORK events: 0 READ events: 0 SAMPLE events: 13852 Signed-off-by: Li Zefan Cc: Frederic Weisbecker Cc: Arnaldo Carvalho de Melo Cc: Peter Zijlstra Cc: Arjan van de Ven LKML-Reference: <4B14B201.9030708@cn.fujitsu.com> Signed-off-by: Ingo Molnar --- tools/perf/builtin-annotate.c | 143 ++++++------------------------------------ tools/perf/util/event.c | 3 +- 2 files changed, 19 insertions(+), 127 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c index 7f85c6e159a..0bf2e8f9af5 100644 --- a/tools/perf/builtin-annotate.c +++ b/tools/perf/builtin-annotate.c @@ -25,19 +25,16 @@ #include "util/thread.h" #include "util/sort.h" #include "util/hist.h" +#include "util/data_map.h" static char const *input_name = "perf.data"; static int force; -static int input; static int full_paths; static int print_line; -static unsigned long page_size; -static unsigned long mmap_window = 32; - struct sym_hist { u64 sum; u64 ip[0]; @@ -156,35 +153,6 @@ static int process_sample_event(event_t *event) return 0; } -static int event__process(event_t *self) -{ - switch (self->header.type) { - case PERF_RECORD_SAMPLE: - return process_sample_event(self); - - case PERF_RECORD_MMAP: - return event__process_mmap(self); - - case PERF_RECORD_COMM: - return event__process_comm(self); - - case PERF_RECORD_FORK: - return event__process_task(self); - /* - * We dont process them right now but they are fine: - */ - - case PERF_RECORD_THROTTLE: - case PERF_RECORD_UNTHROTTLE: - return 0; - - default: - return -1; - } - - return 0; -} - static int parse_line(FILE *file, struct hist_entry *he, u64 len) { struct symbol *sym = he->sym; @@ -485,99 +453,26 @@ static void find_annotations(void) } } +static struct perf_file_handler file_handler = { + .process_sample_event = process_sample_event, + .process_mmap_event = event__process_mmap, + .process_comm_event = event__process_comm, + .process_fork_event = event__process_task, +}; + static int __cmd_annotate(void) { - int ret, rc = EXIT_FAILURE; - unsigned long offset = 0; - unsigned long head = 0; - struct stat input_stat; - event_t *event; - uint32_t size; - char *buf; - - register_idle_thread(); - - input = open(input_name, O_RDONLY); - if (input < 0) { - perror("failed to open file"); - exit(-1); - } - - ret = fstat(input, &input_stat); - if (ret < 0) { - perror("failed to stat file"); - exit(-1); - } - - if (!force && input_stat.st_uid && (input_stat.st_uid != geteuid())) { - fprintf(stderr, "file: %s not owned by current user or root\n", input_name); - exit(-1); - } - - if (!input_stat.st_size) { - fprintf(stderr, "zero-sized file, nothing to do!\n"); - exit(0); - } - -remap: - buf = (char *)mmap(NULL, page_size * mmap_window, PROT_READ, - MAP_SHARED, input, offset); - if (buf == MAP_FAILED) { - perror("failed to mmap file"); - exit(-1); - } - -more: - event = (event_t *)(buf + head); - - size = event->header.size; - if (!size) - size = 8; - - if (head + event->header.size >= page_size * mmap_window) { - unsigned long shift = page_size * (head / page_size); - int munmap_ret; - - munmap_ret = munmap(buf, page_size * mmap_window); - assert(munmap_ret == 0); - - offset += shift; - head -= shift; - goto remap; - } - - size = event->header.size; - - dump_printf("%p [%p]: event: %d\n", - (void *)(offset + head), - (void *)(long)event->header.size, - event->header.type); - - if (!size || event__process(event) < 0) { - - dump_printf("%p [%p]: skipping unknown header type: %d\n", - (void *)(offset + head), - (void *)(long)(event->header.size), - event->header.type); - /* - * assume we lost track of the stream, check alignment, and - * increment a single u64 in the hope to catch on again 'soon'. - */ - - if (unlikely(head & 7)) - head &= ~7ULL; - - size = 8; - } - - head += size; - - if (offset + head < (unsigned long)input_stat.st_size) - goto more; + struct perf_header *header; + struct thread *idle; + int ret; - rc = EXIT_SUCCESS; - close(input); + idle = register_idle_thread(); + register_perf_file_handler(&file_handler); + ret = mmap_dispatch_perf_file(&header, input_name, 0, 0, + &event__cwdlen, &event__cwd); + if (ret) + return ret; if (dump_trace) { event__print_totals(); @@ -595,7 +490,7 @@ more: find_annotations(); - return rc; + return ret; } static const char * const annotate_usage[] = { @@ -644,8 +539,6 @@ int cmd_annotate(int argc, const char **argv, const char *prefix __used) if (symbol__init(&symbol_conf) < 0) return -1; - page_size = getpagesize(); - argc = parse_options(argc, argv, options, annotate_usage, 0); setup_sorting(); diff --git a/tools/perf/util/event.c b/tools/perf/util/event.c index 233d7ad9bd7..414b89d1bde 100644 --- a/tools/perf/util/event.c +++ b/tools/perf/util/event.c @@ -186,8 +186,7 @@ int event__process_comm(event_t *self) { struct thread *thread = threads__findnew(self->comm.pid); - dump_printf("PERF_RECORD_COMM: %s:%d\n", - self->comm.comm, self->comm.pid); + dump_printf(": %s:%d\n", self->comm.comm, self->comm.pid); if (thread == NULL || thread__set_comm(thread, self->comm.comm)) { dump_printf("problem processing PERF_RECORD_COMM, skipping event.\n"); -- cgit v1.2.3-18-g5258 From 5cbd08056142dcb2aea0dca7261afcb810a63c55 Mon Sep 17 00:00:00 2001 From: Li Zefan Date: Tue, 1 Dec 2009 14:05:16 +0800 Subject: perf timechart: Remove open-coded event parsing code Convert builtin-timechart.c to mmap_dispatch_perf_file() + perf_file_handler. Signed-off-by: Li Zefan Acked-by: Arjan van de Ven Cc: Frederic Weisbecker Cc: Arnaldo Carvalho de Melo Cc: Peter Zijlstra LKML-Reference: <4B14B21C.2040406@cn.fujitsu.com> [ v2: cleaned up the printout, fixed a whitespace detail ] Signed-off-by: Ingo Molnar --- tools/perf/builtin-timechart.c | 170 ++++++++--------------------------------- 1 file changed, 30 insertions(+), 140 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-timechart.c b/tools/perf/builtin-timechart.c index dd4d82ac7aa..cb58b6605fc 100644 --- a/tools/perf/builtin-timechart.c +++ b/tools/perf/builtin-timechart.c @@ -29,14 +29,14 @@ #include "util/header.h" #include "util/parse-options.h" #include "util/parse-events.h" +#include "util/event.h" +#include "util/data_map.h" #include "util/svghelper.h" static char const *input_name = "perf.data"; static char const *output_name = "output.svg"; -static unsigned long page_size; -static unsigned long mmap_window = 32; static u64 sample_type; static unsigned int numcpus; @@ -49,8 +49,6 @@ static u64 first_time, last_time; static int power_only; -static struct perf_header *header; - struct per_pid; struct per_pidcomm; @@ -156,9 +154,9 @@ struct sample_wrapper *all_samples; struct process_filter; struct process_filter { - char *name; - int pid; - struct process_filter *next; + char *name; + int pid; + struct process_filter *next; }; static struct process_filter *process_filter; @@ -1045,36 +1043,6 @@ static void write_svg_file(const char *filename) svg_close(); } -static int -process_event(event_t *event) -{ - - switch (event->header.type) { - - case PERF_RECORD_COMM: - return process_comm_event(event); - case PERF_RECORD_FORK: - return process_fork_event(event); - case PERF_RECORD_EXIT: - return process_exit_event(event); - case PERF_RECORD_SAMPLE: - return queue_sample_event(event); - - /* - * We dont process them right now but they are fine: - */ - case PERF_RECORD_MMAP: - case PERF_RECORD_THROTTLE: - case PERF_RECORD_UNTHROTTLE: - return 0; - - default: - return -1; - } - - return 0; -} - static void process_samples(void) { struct sample_wrapper *cursor; @@ -1090,114 +1058,38 @@ static void process_samples(void) } } - -static int __cmd_timechart(void) +static int sample_type_check(u64 type) { - int err, rc = EXIT_FAILURE; - unsigned long offset = 0; - unsigned long head, shift; - struct stat statbuf; - event_t *event; - uint32_t size; - char *buf; - int input; - - input = open(input_name, O_RDONLY); - if (input < 0) { - fprintf(stderr, " failed to open file: %s", input_name); - if (!strcmp(input_name, "perf.data")) - fprintf(stderr, " (try 'perf record' first)"); - fprintf(stderr, "\n"); - exit(-1); - } - - err = fstat(input, &statbuf); - if (err < 0) { - perror("failed to stat file"); - exit(-1); - } - - if (!statbuf.st_size) { - fprintf(stderr, "zero-sized file, nothing to do!\n"); - exit(0); - } + sample_type = type; - header = perf_header__new(); - if (header == NULL) - return -ENOMEM; - - err = perf_header__read(header, input); - if (err < 0) { - perf_header__delete(header); - return err; - } - - head = header->data_offset; - - sample_type = perf_header__sample_type(header); - - shift = page_size * (head / page_size); - offset += shift; - head -= shift; - -remap: - buf = (char *)mmap(NULL, page_size * mmap_window, PROT_READ, - MAP_SHARED, input, offset); - if (buf == MAP_FAILED) { - perror("failed to mmap file"); - exit(-1); - } - -more: - event = (event_t *)(buf + head); - - size = event->header.size; - if (!size) - size = 8; - - if (head + event->header.size >= page_size * mmap_window) { - int ret2; - - shift = page_size * (head / page_size); - - ret2 = munmap(buf, page_size * mmap_window); - assert(ret2 == 0); - - offset += shift; - head -= shift; - goto remap; - } - - size = event->header.size; - - if (!size || process_event(event) < 0) { - pr_warning("%p [%p]: skipping unknown header type: %d\n", - (void *)(offset + head), - (void *)(long)(event->header.size), - event->header.type); - /* - * assume we lost track of the stream, check alignment, and - * increment a single u64 in the hope to catch on again 'soon'. - */ - - if (unlikely(head & 7)) - head &= ~7ULL; - - size = 8; + if (!(sample_type & PERF_SAMPLE_RAW)) { + fprintf(stderr, "No trace samples found in the file.\n" + "Have you used 'perf timechart record' to record it?\n"); + return -1; } - head += size; + return 0; +} - if (offset + head >= header->data_offset + header->data_size) - goto done; +static struct perf_file_handler file_handler = { + .process_comm_event = process_comm_event, + .process_fork_event = process_fork_event, + .process_exit_event = process_exit_event, + .process_sample_event = queue_sample_event, + .sample_type_check = sample_type_check, +}; - if (offset + head < (unsigned long)statbuf.st_size) - goto more; +static int __cmd_timechart(void) +{ + struct perf_header *header; + int ret; -done: - rc = EXIT_SUCCESS; - close(input); + register_perf_file_handler(&file_handler); + ret = mmap_dispatch_perf_file(&header, input_name, 0, 0, + &event__cwdlen, &event__cwd); + if (ret) + return EXIT_FAILURE; process_samples(); @@ -1210,7 +1102,7 @@ done: pr_info("Written %2.1f seconds of trace to %s.\n", (last_time - first_time) / 1000000000.0, output_name); - return rc; + return EXIT_SUCCESS; } static const char * const timechart_usage[] = { @@ -1277,8 +1169,6 @@ int cmd_timechart(int argc, const char **argv, const char *prefix __used) { symbol__init(0); - page_size = getpagesize(); - argc = parse_options(argc, argv, options, timechart_usage, PARSE_OPT_STOP_AT_NON_OPTION); -- cgit v1.2.3-18-g5258 From f41b1e43c41e99c39a2222578a7806032c045c79 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Mon, 30 Nov 2009 19:19:27 -0500 Subject: perf probe: Change a debugging message from pr_info to pr_debug Change annoying debug-info using notice from pr_info() to pr_debug(), since the message always printed when user adds a probe point which requires debug-info. Signed-off-by: Masami Hiramatsu Cc: systemtap Cc: DLE Cc: Steven Rostedt Cc: Jim Keniston Cc: Ananth N Mavinakayanahalli Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: Frederic Weisbecker Cc: Jason Baron Cc: K.Prasad Cc: Peter Zijlstra Cc: Srikar Dronamraju Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker LKML-Reference: <20091201001927.10235.63645.stgit@harusame> Signed-off-by: Ingo Molnar --- tools/perf/builtin-probe.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-probe.c b/tools/perf/builtin-probe.c index a2f6daf01ec..4e418afd670 100644 --- a/tools/perf/builtin-probe.c +++ b/tools/perf/builtin-probe.c @@ -351,7 +351,7 @@ int cmd_probe(int argc, const char **argv, const char *prefix __used) #ifdef NO_LIBDWARF semantic_error("Debuginfo-analysis is not supported"); #else /* !NO_LIBDWARF */ - pr_info("Some probes require debuginfo.\n"); + pr_debug("Some probes require debuginfo.\n"); if (session.vmlinux) fd = open(session.vmlinux, O_RDONLY); -- cgit v1.2.3-18-g5258 From 57d250df7deb3e1742fbf3cc3230119731109552 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Mon, 30 Nov 2009 19:19:34 -0500 Subject: perf probe: Add probe-finder.h without libdwarf Add probe-finder.h as LIB_H without libdwarf, because that header is included even if no libdwarf. Signed-off-by: Masami Hiramatsu Cc: systemtap Cc: DLE Cc: Steven Rostedt Cc: Jim Keniston Cc: Ananth N Mavinakayanahalli Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: Frederic Weisbecker Cc: Jason Baron Cc: K.Prasad Cc: Peter Zijlstra Cc: Srikar Dronamraju Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker LKML-Reference: <20091201001934.10235.44656.stgit@harusame> Signed-off-by: Ingo Molnar --- tools/perf/Makefile | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'tools/perf') diff --git a/tools/perf/Makefile b/tools/perf/Makefile index f1537a94a05..76e4b04d408 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -369,6 +369,7 @@ LIB_H += util/sort.h LIB_H += util/hist.h LIB_H += util/thread.h LIB_H += util/data_map.h +LIB_H += util/probe-finder.h LIB_OBJS += util/abspath.o LIB_OBJS += util/alias.o @@ -485,7 +486,6 @@ ifneq ($(shell sh -c "(echo '\#include '; echo '\#include Date: Mon, 30 Nov 2009 19:19:43 -0500 Subject: perf probe: Fix argv array size in probe parser Since the syntax has been changed, probe definition needs parameters less than MAX_PROBE_ARGS + 1 (probe-point + arguments). Signed-off-by: Masami Hiramatsu Cc: systemtap Cc: DLE Cc: Steven Rostedt Cc: Jim Keniston Cc: Ananth N Mavinakayanahalli Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: Frederic Weisbecker Cc: Jason Baron Cc: K.Prasad Cc: Peter Zijlstra Cc: Srikar Dronamraju Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker LKML-Reference: <20091201001943.10235.80367.stgit@harusame> Signed-off-by: Ingo Molnar --- tools/perf/builtin-probe.c | 11 +++++++---- 1 file changed, 7 insertions(+), 4 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-probe.c b/tools/perf/builtin-probe.c index 4e418afd670..510fdd4e5d3 100644 --- a/tools/perf/builtin-probe.c +++ b/tools/perf/builtin-probe.c @@ -151,7 +151,7 @@ static void parse_probe_point(char *arg, struct probe_point *pp) /* Parse an event definition. Note that any error must die. */ static void parse_probe_event(const char *str) { - char *argv[MAX_PROBE_ARGS + 2]; /* Event + probe + args */ + char *argv[MAX_PROBE_ARGS + 1]; /* probe + args */ int argc, i; struct probe_point *pp = &session.probes[session.nr_probe]; @@ -169,6 +169,9 @@ static void parse_probe_event(const char *str) /* Add an argument */ if (*str != '\0') { const char *s = str; + /* Check the limit number of arguments */ + if (argc == MAX_PROBE_ARGS + 1) + semantic_error("Too many arguments"); /* Skip the argument */ while (!isspace(*str) && *str != '\0') @@ -178,9 +181,9 @@ static void parse_probe_event(const char *str) argv[argc] = strndup(s, str - s); if (argv[argc] == NULL) die("strndup"); - if (++argc == MAX_PROBE_ARGS) - semantic_error("Too many arguments"); - pr_debug("argv[%d]=%s\n", argc, argv[argc - 1]); + pr_debug("argv[%d]=%s\n", argc, argv[argc]); + argc++; + } } while (*str != '\0'); if (!argc) -- cgit v1.2.3-18-g5258 From 934b1f5fd0c9a2ddde5a4487695c126243d9a42b Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Mon, 30 Nov 2009 19:19:51 -0500 Subject: perf probe: Fix probe array index for multiple probe points Fix the index of formatted probe array for multiple probe points, which should be probes[i] instead of probes[0]. Signed-off-by: Masami Hiramatsu Cc: systemtap Cc: DLE Cc: Steven Rostedt Cc: Jim Keniston Cc: Ananth N Mavinakayanahalli Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: Frederic Weisbecker Cc: Jason Baron Cc: K.Prasad Cc: Peter Zijlstra Cc: Srikar Dronamraju Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker LKML-Reference: <20091201001950.10235.54781.stgit@harusame> Signed-off-by: Ingo Molnar --- tools/perf/builtin-probe.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-probe.c b/tools/perf/builtin-probe.c index 510fdd4e5d3..5f47e624e57 100644 --- a/tools/perf/builtin-probe.c +++ b/tools/perf/builtin-probe.c @@ -428,7 +428,7 @@ end_dwarf: pp->retprobe ? 'r' : 'p', PERFPROBE_GROUP, pp->function, pp->offset, i, - pp->probes[0]); + pp->probes[i]); write_new_event(fd, buf); } } -- cgit v1.2.3-18-g5258 From 50656eec82684d03add0f4f4b4875a20bd8f9755 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Mon, 30 Nov 2009 19:19:58 -0500 Subject: perf probe: Move probe event utility functions to probe-event.c Split probe event (kprobe-events and perf probe events) utility functions from builtin-probe.c to probe-event.c. Signed-off-by: Masami Hiramatsu Cc: systemtap Cc: DLE Cc: Steven Rostedt Cc: Jim Keniston Cc: Ananth N Mavinakayanahalli Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: Frederic Weisbecker Cc: Jason Baron Cc: K.Prasad Cc: Peter Zijlstra Cc: Srikar Dronamraju Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker LKML-Reference: <20091201001958.10235.90243.stgit@harusame> Signed-off-by: Ingo Molnar --- tools/perf/Makefile | 2 + tools/perf/builtin-probe.c | 227 ++--------------------------------- tools/perf/util/probe-event.c | 273 ++++++++++++++++++++++++++++++++++++++++++ tools/perf/util/probe-event.h | 10 ++ 4 files changed, 294 insertions(+), 218 deletions(-) create mode 100644 tools/perf/util/probe-event.c create mode 100644 tools/perf/util/probe-event.h (limited to 'tools/perf') diff --git a/tools/perf/Makefile b/tools/perf/Makefile index 76e4b04d408..f8537cf812c 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -370,6 +370,7 @@ LIB_H += util/hist.h LIB_H += util/thread.h LIB_H += util/data_map.h LIB_H += util/probe-finder.h +LIB_H += util/probe-event.h LIB_OBJS += util/abspath.o LIB_OBJS += util/alias.o @@ -412,6 +413,7 @@ LIB_OBJS += util/svghelper.o LIB_OBJS += util/sort.o LIB_OBJS += util/hist.o LIB_OBJS += util/data_map.o +LIB_OBJS += util/probe-event.o BUILTIN_OBJS += builtin-annotate.o diff --git a/tools/perf/builtin-probe.c b/tools/perf/builtin-probe.c index 5f47e624e57..bf20df2e816 100644 --- a/tools/perf/builtin-probe.c +++ b/tools/perf/builtin-probe.c @@ -40,6 +40,7 @@ #include "util/parse-options.h" #include "util/parse-events.h" /* For debugfs_path */ #include "util/probe-finder.h" +#include "util/probe-event.h" /* Default vmlinux search paths */ #define NR_SEARCH_PATH 3 @@ -51,8 +52,6 @@ const char *default_search_path[NR_SEARCH_PATH] = { #define MAX_PATH_LEN 256 #define MAX_PROBES 128 -#define MAX_PROBE_ARGS 128 -#define PERFPROBE_GROUP "probe" /* Session management structure */ static struct { @@ -63,155 +62,17 @@ static struct { struct probe_point probes[MAX_PROBES]; } session; -#define semantic_error(msg ...) die("Semantic error :" msg) - -/* Parse probe point. Return 1 if return probe */ -static void parse_probe_point(char *arg, struct probe_point *pp) -{ - char *ptr, *tmp; - char c, nc = 0; - /* - * - * perf probe SRC:LN - * perf probe FUNC[+OFFS|%return][@SRC] - */ - - ptr = strpbrk(arg, ":+@%"); - if (ptr) { - nc = *ptr; - *ptr++ = '\0'; - } - - /* Check arg is function or file and copy it */ - if (strchr(arg, '.')) /* File */ - pp->file = strdup(arg); - else /* Function */ - pp->function = strdup(arg); - DIE_IF(pp->file == NULL && pp->function == NULL); - - /* Parse other options */ - while (ptr) { - arg = ptr; - c = nc; - ptr = strpbrk(arg, ":+@%"); - if (ptr) { - nc = *ptr; - *ptr++ = '\0'; - } - switch (c) { - case ':': /* Line number */ - pp->line = strtoul(arg, &tmp, 0); - if (*tmp != '\0') - semantic_error("There is non-digit charactor" - " in line number."); - break; - case '+': /* Byte offset from a symbol */ - pp->offset = strtoul(arg, &tmp, 0); - if (*tmp != '\0') - semantic_error("There is non-digit charactor" - " in offset."); - break; - case '@': /* File name */ - if (pp->file) - semantic_error("SRC@SRC is not allowed."); - pp->file = strdup(arg); - DIE_IF(pp->file == NULL); - if (ptr) - semantic_error("@SRC must be the last " - "option."); - break; - case '%': /* Probe places */ - if (strcmp(arg, "return") == 0) { - pp->retprobe = 1; - } else /* Others not supported yet */ - semantic_error("%%%s is not supported.", arg); - break; - default: - DIE_IF("Program has a bug."); - break; - } - } - - /* Exclusion check */ - if (pp->line && pp->offset) - semantic_error("Offset can't be used with line number."); - if (!pp->line && pp->file && !pp->function) - semantic_error("File always requires line number."); - if (pp->offset && !pp->function) - semantic_error("Offset requires an entry function."); - if (pp->retprobe && !pp->function) - semantic_error("Return probe requires an entry function."); - if ((pp->offset || pp->line) && pp->retprobe) - semantic_error("Offset/Line can't be used with return probe."); - - pr_debug("symbol:%s file:%s line:%d offset:%d, return:%d\n", - pp->function, pp->file, pp->line, pp->offset, pp->retprobe); -} - /* Parse an event definition. Note that any error must die. */ static void parse_probe_event(const char *str) { - char *argv[MAX_PROBE_ARGS + 1]; /* probe + args */ - int argc, i; struct probe_point *pp = &session.probes[session.nr_probe]; pr_debug("probe-definition(%d): %s\n", session.nr_probe, str); if (++session.nr_probe == MAX_PROBES) - semantic_error("Too many probes"); - - /* Separate arguments, similar to argv_split */ - argc = 0; - do { - /* Skip separators */ - while (isspace(*str)) - str++; - - /* Add an argument */ - if (*str != '\0') { - const char *s = str; - /* Check the limit number of arguments */ - if (argc == MAX_PROBE_ARGS + 1) - semantic_error("Too many arguments"); - - /* Skip the argument */ - while (!isspace(*str) && *str != '\0') - str++; - - /* Duplicate the argument */ - argv[argc] = strndup(s, str - s); - if (argv[argc] == NULL) - die("strndup"); - pr_debug("argv[%d]=%s\n", argc, argv[argc]); - argc++; - - } - } while (*str != '\0'); - if (!argc) - semantic_error("An empty argument."); - - /* Parse probe point */ - parse_probe_point(argv[0], pp); - free(argv[0]); - if (pp->file || pp->line) - session.need_dwarf = 1; - - /* Copy arguments */ - pp->nr_args = argc - 1; - if (pp->nr_args > 0) { - pp->args = (char **)malloc(sizeof(char *) * pp->nr_args); - if (!pp->args) - die("malloc"); - memcpy(pp->args, &argv[1], sizeof(char *) * pp->nr_args); - } + die("Too many probes (> %d) are specified.", MAX_PROBES); - /* Ensure return probe has no C argument */ - for (i = 0; i < pp->nr_args; i++) - if (is_c_varname(pp->args[i])) { - if (pp->retprobe) - semantic_error("You can't specify local" - " variable for kretprobe"); - session.need_dwarf = 1; - } + /* Parse perf-probe event into probe_point */ + session.need_dwarf = parse_perf_probe_event(str, pp); pr_debug("%d arguments\n", pp->nr_args); } @@ -288,59 +149,15 @@ static const struct option options[] = { "\t\tALN:\tAbsolute line number in file.\n" "\t\tARG:\tProbe argument (local variable name or\n" #endif - "\t\t\tkprobe-tracer argument format is supported.)\n", + "\t\t\tkprobe-tracer argument format.)\n", opt_add_probe_event), OPT_END() }; -static int write_new_event(int fd, const char *buf) -{ - int ret; - - ret = write(fd, buf, strlen(buf)); - if (ret <= 0) - die("Failed to create event."); - else - printf("Added new event: %s\n", buf); - - return ret; -} - -#define MAX_CMDLEN 256 - -static int synthesize_probe_event(struct probe_point *pp) -{ - char *buf; - int i, len, ret; - pp->probes[0] = buf = zalloc(MAX_CMDLEN); - if (!buf) - die("Failed to allocate memory by zalloc."); - ret = snprintf(buf, MAX_CMDLEN, "%s+%d", pp->function, pp->offset); - if (ret <= 0 || ret >= MAX_CMDLEN) - goto error; - len = ret; - - for (i = 0; i < pp->nr_args; i++) { - ret = snprintf(&buf[len], MAX_CMDLEN - len, " %s", - pp->args[i]); - if (ret <= 0 || ret >= MAX_CMDLEN - len) - goto error; - len += ret; - } - pp->found = 1; - return pp->found; -error: - free(pp->probes[0]); - if (ret > 0) - ret = -E2BIG; - return ret; -} - int cmd_probe(int argc, const char **argv, const char *prefix __used) { int i, j, fd, ret; struct probe_point *pp; - char buf[MAX_CMDLEN]; argc = parse_options(argc, argv, options, probe_usage, PARSE_OPT_STOP_AT_NON_OPTION); @@ -352,7 +169,7 @@ int cmd_probe(int argc, const char **argv, const char *prefix __used) if (session.need_dwarf) #ifdef NO_LIBDWARF - semantic_error("Debuginfo-analysis is not supported"); + die("Debuginfo-analysis is not supported"); #else /* !NO_LIBDWARF */ pr_debug("Some probes require debuginfo.\n"); @@ -398,41 +215,15 @@ end_dwarf: if (pp->found) /* This probe is already found. */ continue; - ret = synthesize_probe_event(pp); + ret = synthesize_trace_kprobe_event(pp); if (ret == -E2BIG) - semantic_error("probe point is too long."); + die("probe point definition becomes too long."); else if (ret < 0) die("Failed to synthesize a probe point."); } /* Settng up probe points */ - snprintf(buf, MAX_CMDLEN, "%s/../kprobe_events", debugfs_path); - fd = open(buf, O_WRONLY, O_APPEND); - if (fd < 0) { - if (errno == ENOENT) - die("kprobe_events file does not exist - please rebuild with CONFIG_KPROBE_TRACER."); - else - die("Could not open kprobe_events file: %s", - strerror(errno)); - } - for (j = 0; j < session.nr_probe; j++) { - pp = &session.probes[j]; - if (pp->found == 1) { - snprintf(buf, MAX_CMDLEN, "%c:%s/%s_%x %s\n", - pp->retprobe ? 'r' : 'p', PERFPROBE_GROUP, - pp->function, pp->offset, pp->probes[0]); - write_new_event(fd, buf); - } else - for (i = 0; i < pp->found; i++) { - snprintf(buf, MAX_CMDLEN, "%c:%s/%s_%x_%d %s\n", - pp->retprobe ? 'r' : 'p', - PERFPROBE_GROUP, - pp->function, pp->offset, i, - pp->probes[i]); - write_new_event(fd, buf); - } - } - close(fd); + add_trace_kprobe_events(session.probes, session.nr_probe); return 0; } diff --git a/tools/perf/util/probe-event.c b/tools/perf/util/probe-event.c new file mode 100644 index 00000000000..7335a3b5e49 --- /dev/null +++ b/tools/perf/util/probe-event.c @@ -0,0 +1,273 @@ +/* + * probe-event.c : perf-probe definition to kprobe_events format converter + * + * Written by Masami Hiramatsu + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. + * + */ + +#define _GNU_SOURCE +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#undef _GNU_SOURCE +#include "event.h" +#include "debug.h" +#include "parse-events.h" /* For debugfs_path */ +#include "probe-event.h" + +#define MAX_CMDLEN 256 +#define MAX_PROBE_ARGS 128 +#define PERFPROBE_GROUP "probe" + +#define semantic_error(msg ...) die("Semantic error :" msg) + +/* Parse probepoint definition. */ +static void parse_perf_probe_probepoint(char *arg, struct probe_point *pp) +{ + char *ptr, *tmp; + char c, nc = 0; + /* + * + * perf probe SRC:LN + * perf probe FUNC[+OFFS|%return][@SRC] + */ + + ptr = strpbrk(arg, ":+@%"); + if (ptr) { + nc = *ptr; + *ptr++ = '\0'; + } + + /* Check arg is function or file and copy it */ + if (strchr(arg, '.')) /* File */ + pp->file = strdup(arg); + else /* Function */ + pp->function = strdup(arg); + DIE_IF(pp->file == NULL && pp->function == NULL); + + /* Parse other options */ + while (ptr) { + arg = ptr; + c = nc; + ptr = strpbrk(arg, ":+@%"); + if (ptr) { + nc = *ptr; + *ptr++ = '\0'; + } + switch (c) { + case ':': /* Line number */ + pp->line = strtoul(arg, &tmp, 0); + if (*tmp != '\0') + semantic_error("There is non-digit charactor" + " in line number."); + break; + case '+': /* Byte offset from a symbol */ + pp->offset = strtoul(arg, &tmp, 0); + if (*tmp != '\0') + semantic_error("There is non-digit charactor" + " in offset."); + break; + case '@': /* File name */ + if (pp->file) + semantic_error("SRC@SRC is not allowed."); + pp->file = strdup(arg); + DIE_IF(pp->file == NULL); + if (ptr) + semantic_error("@SRC must be the last " + "option."); + break; + case '%': /* Probe places */ + if (strcmp(arg, "return") == 0) { + pp->retprobe = 1; + } else /* Others not supported yet */ + semantic_error("%%%s is not supported.", arg); + break; + default: + DIE_IF("Program has a bug."); + break; + } + } + + /* Exclusion check */ + if (pp->line && pp->offset) + semantic_error("Offset can't be used with line number."); + + if (!pp->line && pp->file && !pp->function) + semantic_error("File always requires line number."); + + if (pp->offset && !pp->function) + semantic_error("Offset requires an entry function."); + + if (pp->retprobe && !pp->function) + semantic_error("Return probe requires an entry function."); + + if ((pp->offset || pp->line) && pp->retprobe) + semantic_error("Offset/Line can't be used with return probe."); + + pr_debug("symbol:%s file:%s line:%d offset:%d, return:%d\n", + pp->function, pp->file, pp->line, pp->offset, pp->retprobe); +} + +/* Parse perf-probe event definition */ +int parse_perf_probe_event(const char *str, struct probe_point *pp) +{ + char *argv[MAX_PROBE_ARGS + 1]; /* probe + args */ + int argc, i, need_dwarf = 0; + + /* Separate arguments, similar to argv_split */ + argc = 0; + do { + /* Skip separators */ + while (isspace(*str)) + str++; + + /* Add an argument */ + if (*str != '\0') { + const char *s = str; + /* Check the limit number of arguments */ + if (argc == MAX_PROBE_ARGS + 1) + semantic_error("Too many arguments"); + + /* Skip the argument */ + while (!isspace(*str) && *str != '\0') + str++; + + /* Duplicate the argument */ + argv[argc] = strndup(s, str - s); + if (argv[argc] == NULL) + die("strndup"); + pr_debug("argv[%d]=%s\n", argc, argv[argc]); + argc++; + } + } while (*str != '\0'); + if (!argc) + semantic_error("An empty argument."); + + /* Parse probe point */ + parse_perf_probe_probepoint(argv[0], pp); + free(argv[0]); + if (pp->file || pp->line) + need_dwarf = 1; + + /* Copy arguments */ + pp->nr_args = argc - 1; + if (pp->nr_args > 0) { + pp->args = (char **)malloc(sizeof(char *) * pp->nr_args); + if (!pp->args) + die("malloc"); + memcpy(pp->args, &argv[1], sizeof(char *) * pp->nr_args); + } + + /* Ensure return probe has no C argument */ + for (i = 0; i < pp->nr_args; i++) + if (is_c_varname(pp->args[i])) { + if (pp->retprobe) + semantic_error("You can't specify local" + " variable for kretprobe"); + need_dwarf = 1; + } + + return need_dwarf; +} + +int synthesize_trace_kprobe_event(struct probe_point *pp) +{ + char *buf; + int i, len, ret; + + pp->probes[0] = buf = zalloc(MAX_CMDLEN); + if (!buf) + die("Failed to allocate memory by zalloc."); + ret = snprintf(buf, MAX_CMDLEN, "%s+%d", pp->function, pp->offset); + if (ret <= 0 || ret >= MAX_CMDLEN) + goto error; + len = ret; + + for (i = 0; i < pp->nr_args; i++) { + ret = snprintf(&buf[len], MAX_CMDLEN - len, " %s", + pp->args[i]); + if (ret <= 0 || ret >= MAX_CMDLEN - len) + goto error; + len += ret; + } + pp->found = 1; + + return pp->found; +error: + free(pp->probes[0]); + if (ret > 0) + ret = -E2BIG; + + return ret; +} + +static int write_trace_kprobe_event(int fd, const char *buf) +{ + int ret; + + ret = write(fd, buf, strlen(buf)); + if (ret <= 0) + die("Failed to create event."); + else + printf("Added new event: %s\n", buf); + + return ret; +} + +void add_trace_kprobe_events(struct probe_point *probes, int nr_probes) +{ + int i, j, fd; + struct probe_point *pp; + char buf[MAX_CMDLEN]; + + snprintf(buf, MAX_CMDLEN, "%s/../kprobe_events", debugfs_path); + fd = open(buf, O_WRONLY, O_APPEND); + if (fd < 0) { + if (errno == ENOENT) + die("kprobe_events file does not exist -" + " please rebuild with CONFIG_KPROBE_TRACER."); + else + die("Could not open kprobe_events file: %s", + strerror(errno)); + } + + for (j = 0; j < nr_probes; j++) { + pp = probes + j; + if (pp->found == 1) { + snprintf(buf, MAX_CMDLEN, "%c:%s/%s_%x %s\n", + pp->retprobe ? 'r' : 'p', PERFPROBE_GROUP, + pp->function, pp->offset, pp->probes[0]); + write_trace_kprobe_event(fd, buf); + } else + for (i = 0; i < pp->found; i++) { + snprintf(buf, MAX_CMDLEN, "%c:%s/%s_%x_%d %s\n", + pp->retprobe ? 'r' : 'p', + PERFPROBE_GROUP, + pp->function, pp->offset, i, + pp->probes[i]); + write_trace_kprobe_event(fd, buf); + } + } + close(fd); +} diff --git a/tools/perf/util/probe-event.h b/tools/perf/util/probe-event.h new file mode 100644 index 00000000000..0089c455eca --- /dev/null +++ b/tools/perf/util/probe-event.h @@ -0,0 +1,10 @@ +#ifndef _PROBE_EVENT_H +#define _PROBE_EVENT_H + +#include "probe-finder.h" + +extern int parse_perf_probe_event(const char *str, struct probe_point *pp); +extern int synthesize_trace_kprobe_event(struct probe_point *pp); +extern void add_trace_kprobe_events(struct probe_point *probes, int nr_probes); + +#endif /*_PROBE_EVENT_H */ -- cgit v1.2.3-18-g5258 From e1c01d61a98703fcc80d15b8068ec36d5a215f7e Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Mon, 30 Nov 2009 19:20:05 -0500 Subject: perf probe: Add argv_split() from lib/argv_split.c Add argv_split() ported from lib/argv_split.c to string.c and use it in util/probe-event.c. Signed-off-by: Masami Hiramatsu Cc: systemtap Cc: DLE Cc: Steven Rostedt Cc: Jim Keniston Cc: Ananth N Mavinakayanahalli Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: Frederic Weisbecker Cc: Jason Baron Cc: K.Prasad Cc: Peter Zijlstra Cc: Srikar Dronamraju Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker LKML-Reference: <20091201002005.10235.55602.stgit@harusame> Signed-off-by: Ingo Molnar --- tools/perf/util/probe-event.c | 55 +++++++---------------- tools/perf/util/string.c | 101 ++++++++++++++++++++++++++++++++++++++++++ tools/perf/util/string.h | 2 + 3 files changed, 118 insertions(+), 40 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/probe-event.c b/tools/perf/util/probe-event.c index 7335a3b5e49..e3a683ab976 100644 --- a/tools/perf/util/probe-event.c +++ b/tools/perf/util/probe-event.c @@ -32,6 +32,7 @@ #undef _GNU_SOURCE #include "event.h" +#include "string.h" #include "debug.h" #include "parse-events.h" /* For debugfs_path */ #include "probe-event.h" @@ -132,62 +133,36 @@ static void parse_perf_probe_probepoint(char *arg, struct probe_point *pp) /* Parse perf-probe event definition */ int parse_perf_probe_event(const char *str, struct probe_point *pp) { - char *argv[MAX_PROBE_ARGS + 1]; /* probe + args */ + char **argv; int argc, i, need_dwarf = 0; - /* Separate arguments, similar to argv_split */ - argc = 0; - do { - /* Skip separators */ - while (isspace(*str)) - str++; - - /* Add an argument */ - if (*str != '\0') { - const char *s = str; - /* Check the limit number of arguments */ - if (argc == MAX_PROBE_ARGS + 1) - semantic_error("Too many arguments"); - - /* Skip the argument */ - while (!isspace(*str) && *str != '\0') - str++; - - /* Duplicate the argument */ - argv[argc] = strndup(s, str - s); - if (argv[argc] == NULL) - die("strndup"); - pr_debug("argv[%d]=%s\n", argc, argv[argc]); - argc++; - } - } while (*str != '\0'); - if (!argc) - semantic_error("An empty argument."); + argv = argv_split(str, &argc); + if (!argv) + die("argv_split failed."); + if (argc > MAX_PROBE_ARGS + 1) + semantic_error("Too many arguments"); /* Parse probe point */ parse_perf_probe_probepoint(argv[0], pp); - free(argv[0]); if (pp->file || pp->line) need_dwarf = 1; - /* Copy arguments */ + /* Copy arguments and ensure return probe has no C argument */ pp->nr_args = argc - 1; - if (pp->nr_args > 0) { - pp->args = (char **)malloc(sizeof(char *) * pp->nr_args); - if (!pp->args) - die("malloc"); - memcpy(pp->args, &argv[1], sizeof(char *) * pp->nr_args); - } - - /* Ensure return probe has no C argument */ - for (i = 0; i < pp->nr_args; i++) + pp->args = zalloc(sizeof(char *) * pp->nr_args); + for (i = 0; i < pp->nr_args; i++) { + pp->args[i] = strdup(argv[i + 1]); + if (!pp->args[i]) + die("Failed to copy argument."); if (is_c_varname(pp->args[i])) { if (pp->retprobe) semantic_error("You can't specify local" " variable for kretprobe"); need_dwarf = 1; } + } + argv_free(argv); return need_dwarf; } diff --git a/tools/perf/util/string.c b/tools/perf/util/string.c index 227043577e0..0977cf43178 100644 --- a/tools/perf/util/string.c +++ b/tools/perf/util/string.c @@ -127,3 +127,104 @@ out_err: out: return length; } + +/* + * Helper function for splitting a string into an argv-like array. + * originaly copied from lib/argv_split.c + */ +static const char *skip_sep(const char *cp) +{ + while (*cp && isspace(*cp)) + cp++; + + return cp; +} + +static const char *skip_arg(const char *cp) +{ + while (*cp && !isspace(*cp)) + cp++; + + return cp; +} + +static int count_argc(const char *str) +{ + int count = 0; + + while (*str) { + str = skip_sep(str); + if (*str) { + count++; + str = skip_arg(str); + } + } + + return count; +} + +/** + * argv_free - free an argv + * @argv - the argument vector to be freed + * + * Frees an argv and the strings it points to. + */ +void argv_free(char **argv) +{ + char **p; + for (p = argv; *p; p++) + free(*p); + + free(argv); +} + +/** + * argv_split - split a string at whitespace, returning an argv + * @str: the string to be split + * @argcp: returned argument count + * + * Returns an array of pointers to strings which are split out from + * @str. This is performed by strictly splitting on white-space; no + * quote processing is performed. Multiple whitespace characters are + * considered to be a single argument separator. The returned array + * is always NULL-terminated. Returns NULL on memory allocation + * failure. + */ +char **argv_split(const char *str, int *argcp) +{ + int argc = count_argc(str); + char **argv = zalloc(sizeof(*argv) * (argc+1)); + char **argvp; + + if (argv == NULL) + goto out; + + if (argcp) + *argcp = argc; + + argvp = argv; + + while (*str) { + str = skip_sep(str); + + if (*str) { + const char *p = str; + char *t; + + str = skip_arg(str); + + t = strndup(p, str-p); + if (t == NULL) + goto fail; + *argvp++ = t; + } + } + *argvp = NULL; + +out: + return argv; + +fail: + argv_free(argv); + return NULL; +} diff --git a/tools/perf/util/string.h b/tools/perf/util/string.h index e50b07f8082..bfecec265a1 100644 --- a/tools/perf/util/string.h +++ b/tools/perf/util/string.h @@ -6,6 +6,8 @@ int hex2u64(const char *ptr, u64 *val); char *strxfrchar(char *s, char from, char to); s64 perf_atoll(const char *str); +char **argv_split(const char *str, int *argcp); +void argv_free(char **argv); #define _STR(x) #x #define STR(x) _STR(x) -- cgit v1.2.3-18-g5258 From 4de189fe6e5ad8241f6f8709d2e2ba4c3aeae33a Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Mon, 30 Nov 2009 19:20:17 -0500 Subject: perf probe: Add --list option for listing current probe events Add --list option for listing currently defined probe events in the kernel. This shows events in below format; [group:event] for example: [probe:schedule_0] schedule+30 cpu Note that source file/line information is not supported yet. So even if you added a probe by line, it will be shown in . Signed-off-by: Masami Hiramatsu Cc: systemtap Cc: DLE Cc: Steven Rostedt Cc: Jim Keniston Cc: Ananth N Mavinakayanahalli Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: Frederic Weisbecker Cc: Jason Baron Cc: K.Prasad Cc: Peter Zijlstra Cc: Srikar Dronamraju Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker LKML-Reference: <20091201002017.10235.76575.stgit@harusame> Signed-off-by: Ingo Molnar --- tools/perf/builtin-probe.c | 12 ++- tools/perf/util/probe-event.c | 231 ++++++++++++++++++++++++++++++++++++++---- tools/perf/util/probe-event.h | 5 + 3 files changed, 230 insertions(+), 18 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-probe.c b/tools/perf/builtin-probe.c index bf20df2e816..b5d15cf2547 100644 --- a/tools/perf/builtin-probe.c +++ b/tools/perf/builtin-probe.c @@ -62,6 +62,8 @@ static struct { struct probe_point probes[MAX_PROBES]; } session; +static bool listing; + /* Parse an event definition. Note that any error must die. */ static void parse_probe_event(const char *str) { @@ -119,6 +121,7 @@ static int open_default_vmlinux(void) static const char * const probe_usage[] = { "perf probe [] 'PROBEDEF' ['PROBEDEF' ...]", "perf probe [] --add 'PROBEDEF' [--add 'PROBEDEF' ...]", + "perf probe --list", NULL }; @@ -129,6 +132,7 @@ static const struct option options[] = { OPT_STRING('k', "vmlinux", &session.vmlinux, "file", "vmlinux/module pathname"), #endif + OPT_BOOLEAN('l', "list", &listing, "list up current probes"), OPT_CALLBACK('a', "add", NULL, #ifdef NO_LIBDWARF "FUNC[+OFFS|%return] [ARG ...]", @@ -164,9 +168,15 @@ int cmd_probe(int argc, const char **argv, const char *prefix __used) for (i = 0; i < argc; i++) parse_probe_event(argv[i]); - if (session.nr_probe == 0) + if ((session.nr_probe == 0 && !listing) || + (session.nr_probe != 0 && listing)) usage_with_options(probe_usage, options); + if (listing) { + show_perf_probe_events(); + return 0; + } + if (session.need_dwarf) #ifdef NO_LIBDWARF die("Debuginfo-analysis is not supported"); diff --git a/tools/perf/util/probe-event.c b/tools/perf/util/probe-event.c index e3a683ab976..7f4f288c642 100644 --- a/tools/perf/util/probe-event.c +++ b/tools/perf/util/probe-event.c @@ -29,10 +29,13 @@ #include #include #include +#include +#include #undef _GNU_SOURCE #include "event.h" #include "string.h" +#include "strlist.h" #include "debug.h" #include "parse-events.h" /* For debugfs_path */ #include "probe-event.h" @@ -43,6 +46,19 @@ #define semantic_error(msg ...) die("Semantic error :" msg) +/* If there is no space to write, returns -E2BIG. */ +static int e_snprintf(char *str, size_t size, const char *format, ...) +{ + int ret; + va_list ap; + va_start(ap, format); + ret = vsnprintf(str, size, format, ap); + va_end(ap); + if (ret >= (int)size) + ret = -E2BIG; + return ret; +} + /* Parse probepoint definition. */ static void parse_perf_probe_probepoint(char *arg, struct probe_point *pp) { @@ -166,6 +182,103 @@ int parse_perf_probe_event(const char *str, struct probe_point *pp) return need_dwarf; } +/* Parse kprobe_events event into struct probe_point */ +void parse_trace_kprobe_event(const char *str, char **group, char **event, + struct probe_point *pp) +{ + char pr; + char *p; + int ret, i, argc; + char **argv; + + pr_debug("Parsing kprobe_events: %s\n", str); + argv = argv_split(str, &argc); + if (!argv) + die("argv_split failed."); + if (argc < 2) + semantic_error("Too less arguments."); + + /* Scan event and group name. */ + ret = sscanf(argv[0], "%c:%m[^/ \t]/%m[^ \t]", + &pr, group, event); + if (ret != 3) + semantic_error("Failed to parse event name: %s", argv[0]); + pr_debug("Group:%s Event:%s probe:%c\n", *group, *event, pr); + + if (!pp) + goto end; + + pp->retprobe = (pr == 'r'); + + /* Scan function name and offset */ + ret = sscanf(argv[1], "%m[^+]+%d", &pp->function, &pp->offset); + if (ret == 1) + pp->offset = 0; + + /* kprobe_events doesn't have this information */ + pp->line = 0; + pp->file = NULL; + + pp->nr_args = argc - 2; + pp->args = zalloc(sizeof(char *) * pp->nr_args); + for (i = 0; i < pp->nr_args; i++) { + p = strchr(argv[i + 2], '='); + if (p) /* We don't need which register is assigned. */ + *p = '\0'; + pp->args[i] = strdup(argv[i + 2]); + if (!pp->args[i]) + die("Failed to copy argument."); + } + +end: + argv_free(argv); +} + +int synthesize_perf_probe_event(struct probe_point *pp) +{ + char *buf; + char offs[64] = "", line[64] = ""; + int i, len, ret; + + pp->probes[0] = buf = zalloc(MAX_CMDLEN); + if (!buf) + die("Failed to allocate memory by zalloc."); + if (pp->offset) { + ret = e_snprintf(offs, 64, "+%d", pp->offset); + if (ret <= 0) + goto error; + } + if (pp->line) { + ret = e_snprintf(line, 64, ":%d", pp->line); + if (ret <= 0) + goto error; + } + + if (pp->function) + ret = e_snprintf(buf, MAX_CMDLEN, "%s%s%s%s", pp->function, + offs, pp->retprobe ? "%return" : "", line); + else + ret = e_snprintf(buf, MAX_CMDLEN, "%s%s%s%s", pp->file, line); + if (ret <= 0) + goto error; + len = ret; + + for (i = 0; i < pp->nr_args; i++) { + ret = e_snprintf(&buf[len], MAX_CMDLEN - len, " %s", + pp->args[i]); + if (ret <= 0) + goto error; + len += ret; + } + pp->found = 1; + + return pp->found; +error: + free(pp->probes[0]); + + return ret; +} + int synthesize_trace_kprobe_event(struct probe_point *pp) { char *buf; @@ -174,15 +287,15 @@ int synthesize_trace_kprobe_event(struct probe_point *pp) pp->probes[0] = buf = zalloc(MAX_CMDLEN); if (!buf) die("Failed to allocate memory by zalloc."); - ret = snprintf(buf, MAX_CMDLEN, "%s+%d", pp->function, pp->offset); - if (ret <= 0 || ret >= MAX_CMDLEN) + ret = e_snprintf(buf, MAX_CMDLEN, "%s+%d", pp->function, pp->offset); + if (ret <= 0) goto error; len = ret; for (i = 0; i < pp->nr_args; i++) { - ret = snprintf(&buf[len], MAX_CMDLEN - len, " %s", - pp->args[i]); - if (ret <= 0 || ret >= MAX_CMDLEN - len) + ret = e_snprintf(&buf[len], MAX_CMDLEN - len, " %s", + pp->args[i]); + if (ret <= 0) goto error; len += ret; } @@ -191,12 +304,105 @@ int synthesize_trace_kprobe_event(struct probe_point *pp) return pp->found; error: free(pp->probes[0]); - if (ret > 0) - ret = -E2BIG; return ret; } +static int open_kprobe_events(int flags, int mode) +{ + char buf[PATH_MAX]; + int ret; + + ret = e_snprintf(buf, PATH_MAX, "%s/../kprobe_events", debugfs_path); + if (ret < 0) + die("Failed to make kprobe_events path."); + + ret = open(buf, flags, mode); + if (ret < 0) { + if (errno == ENOENT) + die("kprobe_events file does not exist -" + " please rebuild with CONFIG_KPROBE_TRACER."); + else + die("Could not open kprobe_events file: %s", + strerror(errno)); + } + return ret; +} + +/* Get raw string list of current kprobe_events */ +static struct strlist *get_trace_kprobe_event_rawlist(int fd) +{ + int ret, idx; + FILE *fp; + char buf[MAX_CMDLEN]; + char *p; + struct strlist *sl; + + sl = strlist__new(true, NULL); + + fp = fdopen(dup(fd), "r"); + while (!feof(fp)) { + p = fgets(buf, MAX_CMDLEN, fp); + if (!p) + break; + + idx = strlen(p) - 1; + if (p[idx] == '\n') + p[idx] = '\0'; + ret = strlist__add(sl, buf); + if (ret < 0) + die("strlist__add failed: %s", strerror(-ret)); + } + fclose(fp); + + return sl; +} + +/* Free and zero clear probe_point */ +static void clear_probe_point(struct probe_point *pp) +{ + int i; + + if (pp->function) + free(pp->function); + if (pp->file) + free(pp->file); + for (i = 0; i < pp->nr_args; i++) + free(pp->args[i]); + if (pp->args) + free(pp->args); + for (i = 0; i < pp->found; i++) + free(pp->probes[i]); + memset(pp, 0, sizeof(pp)); +} + +/* List up current perf-probe events */ +void show_perf_probe_events(void) +{ + unsigned int i; + int fd; + char *group, *event; + struct probe_point pp; + struct strlist *rawlist; + struct str_node *ent; + + fd = open_kprobe_events(O_RDONLY, 0); + rawlist = get_trace_kprobe_event_rawlist(fd); + close(fd); + + for (i = 0; i < strlist__nr_entries(rawlist); i++) { + ent = strlist__entry(rawlist, i); + parse_trace_kprobe_event(ent->s, &group, &event, &pp); + synthesize_perf_probe_event(&pp); + printf("[%s:%s]\t%s\n", group, event, pp.probes[0]); + free(group); + free(event); + clear_probe_point(&pp); + } + + strlist__delete(rawlist); +} + static int write_trace_kprobe_event(int fd, const char *buf) { int ret; @@ -216,16 +422,7 @@ void add_trace_kprobe_events(struct probe_point *probes, int nr_probes) struct probe_point *pp; char buf[MAX_CMDLEN]; - snprintf(buf, MAX_CMDLEN, "%s/../kprobe_events", debugfs_path); - fd = open(buf, O_WRONLY, O_APPEND); - if (fd < 0) { - if (errno == ENOENT) - die("kprobe_events file does not exist -" - " please rebuild with CONFIG_KPROBE_TRACER."); - else - die("Could not open kprobe_events file: %s", - strerror(errno)); - } + fd = open_kprobe_events(O_WRONLY, O_APPEND); for (j = 0; j < nr_probes; j++) { pp = probes + j; diff --git a/tools/perf/util/probe-event.h b/tools/perf/util/probe-event.h index 0089c455eca..88db7d1a947 100644 --- a/tools/perf/util/probe-event.h +++ b/tools/perf/util/probe-event.h @@ -2,9 +2,14 @@ #define _PROBE_EVENT_H #include "probe-finder.h" +#include "strlist.h" extern int parse_perf_probe_event(const char *str, struct probe_point *pp); +extern int synthesize_perf_probe_event(struct probe_point *pp); +extern void parse_trace_kprobe_event(const char *str, char **group, + char **event, struct probe_point *pp); extern int synthesize_trace_kprobe_event(struct probe_point *pp); extern void add_trace_kprobe_events(struct probe_point *probes, int nr_probes); +extern void show_perf_probe_events(void); #endif /*_PROBE_EVENT_H */ -- cgit v1.2.3-18-g5258 From b498ce1f2753b9724b2fc05d2057f7d1490cfa93 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Mon, 30 Nov 2009 19:20:25 -0500 Subject: perf probe: Simplify event naming Simplify event naming as _. Each event name is globally unique (group name is not checked). So, if there is schedule_0, next probe event on schedule() will be schedule_1. Signed-off-by: Masami Hiramatsu Cc: systemtap Cc: DLE Cc: Steven Rostedt Cc: Jim Keniston Cc: Ananth N Mavinakayanahalli Cc: Christoph Hellwig Cc: Frank Ch. Eigler Cc: Frederic Weisbecker Cc: Jason Baron Cc: K.Prasad Cc: Peter Zijlstra Cc: Srikar Dronamraju Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker LKML-Reference: <20091201002024.10235.2353.stgit@harusame> Signed-off-by: Ingo Molnar --- tools/perf/util/probe-event.c | 67 ++++++++++++++++++++++++++++++++++--------- tools/perf/util/probe-event.h | 3 ++ 2 files changed, 56 insertions(+), 14 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/probe-event.c b/tools/perf/util/probe-event.c index 7f4f288c642..e42f3acc9a7 100644 --- a/tools/perf/util/probe-event.c +++ b/tools/perf/util/probe-event.c @@ -403,6 +403,29 @@ void show_perf_probe_events(void) strlist__delete(rawlist); } +/* Get current perf-probe event names */ +static struct strlist *get_perf_event_names(int fd) +{ + unsigned int i; + char *group, *event; + struct strlist *sl, *rawlist; + struct str_node *ent; + + rawlist = get_trace_kprobe_event_rawlist(fd); + + sl = strlist__new(false, NULL); + for (i = 0; i < strlist__nr_entries(rawlist); i++) { + ent = strlist__entry(rawlist, i); + parse_trace_kprobe_event(ent->s, &group, &event, NULL); + strlist__add(sl, event); + free(group); + } + + strlist__delete(rawlist); + + return sl; +} + static int write_trace_kprobe_event(int fd, const char *buf) { int ret; @@ -416,30 +439,46 @@ static int write_trace_kprobe_event(int fd, const char *buf) return ret; } +static void get_new_event_name(char *buf, size_t len, const char *base, + struct strlist *namelist) +{ + int i, ret; + for (i = 0; i < MAX_EVENT_INDEX; i++) { + ret = e_snprintf(buf, len, "%s_%d", base, i); + if (ret < 0) + die("snprintf() failed: %s", strerror(-ret)); + if (!strlist__has_entry(namelist, buf)) + break; + } + if (i == MAX_EVENT_INDEX) + die("Too many events are on the same function."); +} + void add_trace_kprobe_events(struct probe_point *probes, int nr_probes) { int i, j, fd; struct probe_point *pp; char buf[MAX_CMDLEN]; + char event[64]; + struct strlist *namelist; - fd = open_kprobe_events(O_WRONLY, O_APPEND); + fd = open_kprobe_events(O_RDWR, O_APPEND); + /* Get current event names */ + namelist = get_perf_event_names(fd); for (j = 0; j < nr_probes; j++) { pp = probes + j; - if (pp->found == 1) { - snprintf(buf, MAX_CMDLEN, "%c:%s/%s_%x %s\n", - pp->retprobe ? 'r' : 'p', PERFPROBE_GROUP, - pp->function, pp->offset, pp->probes[0]); + for (i = 0; i < pp->found; i++) { + /* Get an unused new event name */ + get_new_event_name(event, 64, pp->function, namelist); + snprintf(buf, MAX_CMDLEN, "%c:%s/%s %s\n", + pp->retprobe ? 'r' : 'p', + PERFPROBE_GROUP, event, + pp->probes[i]); write_trace_kprobe_event(fd, buf); - } else - for (i = 0; i < pp->found; i++) { - snprintf(buf, MAX_CMDLEN, "%c:%s/%s_%x_%d %s\n", - pp->retprobe ? 'r' : 'p', - PERFPROBE_GROUP, - pp->function, pp->offset, i, - pp->probes[i]); - write_trace_kprobe_event(fd, buf); - } + /* Add added event name to namelist */ + strlist__add(namelist, event); + } } close(fd); } diff --git a/tools/perf/util/probe-event.h b/tools/perf/util/probe-event.h index 88db7d1a947..0c6fe56fe38 100644 --- a/tools/perf/util/probe-event.h +++ b/tools/perf/util/probe-event.h @@ -12,4 +12,7 @@ extern int synthesize_trace_kprobe_event(struct probe_point *pp); extern void add_trace_kprobe_events(struct probe_point *probes, int nr_probes); extern void show_perf_probe_events(void); +/* Maximum index number of event-name postfix */ +#define MAX_EVENT_INDEX 1024 + #endif /*_PROBE_EVENT_H */ -- cgit v1.2.3-18-g5258 From c19e33aa840e9202ef8d4c93056b59f3edc2208d Mon Sep 17 00:00:00 2001 From: Liming Wang Date: Wed, 2 Dec 2009 14:11:46 +0800 Subject: perf tools: Fix _GNU_SOURCE macro related strndup() build error strndup is a GNU extension. So dont include string.h without defining _GNU_SOURCE (it results in a compile error otherwise). Remove these includes as util.h does it already. Signed-off-by: Liming Wang Acked-by: Frederic Weisbecker Acked-by: Xiao Guangrong Cc: peterz@infradead.org Cc: mhiramat@redhat.com LKML-Reference: <1259734306-26323-1-git-send-email-liming.wang@windriver.com> Signed-off-by: Ingo Molnar --- tools/perf/util/string.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/string.c b/tools/perf/util/string.c index 0977cf43178..f24a8cc933d 100644 --- a/tools/perf/util/string.c +++ b/tools/perf/util/string.c @@ -1,5 +1,3 @@ -#include -#include #include "string.h" #include "util.h" -- cgit v1.2.3-18-g5258 From bdad0db7dbdb37d0bb3c7d0f65cd3ff599ea6ecb Mon Sep 17 00:00:00 2001 From: Xiao Guangrong Date: Wed, 2 Dec 2009 16:08:41 +0800 Subject: perf_event: Fix compile error Fix: cc1: warnings being treated as errors builtin-probe.c: In function 'cmd_probe': builtin-probe.c:163: error: unused variable 'fd' Signed-off-by: Xiao Guangrong Cc: Masami Hiramatsu Cc: Peter Zijlstra LKML-Reference: <4B162089.8000907@cn.fujitsu.com> [ v2: use NO_LIBDWARF instead of __used ] Signed-off-by: Ingo Molnar --- tools/perf/builtin-probe.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-probe.c b/tools/perf/builtin-probe.c index b5d15cf2547..a58e11b7ea8 100644 --- a/tools/perf/builtin-probe.c +++ b/tools/perf/builtin-probe.c @@ -160,7 +160,10 @@ static const struct option options[] = { int cmd_probe(int argc, const char **argv, const char *prefix __used) { - int i, j, fd, ret; + int i, j, ret; +#ifndef NO_LIBDWARF + int fd; +#endif struct probe_point *pp; argc = parse_options(argc, argv, options, probe_usage, -- cgit v1.2.3-18-g5258 From 93aaa45a6ad3f983180223601fc663cc551ad499 Mon Sep 17 00:00:00 2001 From: Liming Wang Date: Wed, 2 Dec 2009 16:42:54 +0800 Subject: perf tools: Replace %m with %a in sscanf Not all glibc support %m and it results in a compile error if %m not supported. Replace it with %a and (float *) casts. Signed-off-by: Liming Wang Acked-by: Frederic Weisbecker Cc: peterz@infradead.org Cc: mhiramat@redhat.com LKML-Reference: <1259743374-9950-1-git-send-email-liming.wang@windriver.com> Signed-off-by: Ingo Molnar --- tools/perf/util/probe-event.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/probe-event.c b/tools/perf/util/probe-event.c index e42f3acc9a7..cd7fbda5e2a 100644 --- a/tools/perf/util/probe-event.c +++ b/tools/perf/util/probe-event.c @@ -199,8 +199,8 @@ void parse_trace_kprobe_event(const char *str, char **group, char **event, semantic_error("Too less arguments."); /* Scan event and group name. */ - ret = sscanf(argv[0], "%c:%m[^/ \t]/%m[^ \t]", - &pr, group, event); + ret = sscanf(argv[0], "%c:%a[^/ \t]/%a[^ \t]", + &pr, (float *)(void *)group, (float *)(void *)event); if (ret != 3) semantic_error("Failed to parse event name: %s", argv[0]); pr_debug("Group:%s Event:%s probe:%c\n", *group, *event, pr); @@ -211,7 +211,7 @@ void parse_trace_kprobe_event(const char *str, char **group, char **event, pp->retprobe = (pr == 'r'); /* Scan function name and offset */ - ret = sscanf(argv[1], "%m[^+]+%d", &pp->function, &pp->offset); + ret = sscanf(argv[1], "%a[^+]+%d", (float *)(void *)&pp->function, &pp->offset); if (ret == 1) pp->offset = 0; -- cgit v1.2.3-18-g5258