diff options
Diffstat (limited to 'tools/power/cpupower')
37 files changed, 1105 insertions, 411 deletions
diff --git a/tools/power/cpupower/.gitignore b/tools/power/cpupower/.gitignore index 8a83dd2ffc1..d42073f1260 100644 --- a/tools/power/cpupower/.gitignore +++ b/tools/power/cpupower/.gitignore @@ -20,3 +20,10 @@ utils/cpufreq-set.o utils/cpufreq-aperf.o cpupower bench/cpufreq-bench +debug/kernel/Module.symvers +debug/i386/centrino-decode +debug/i386/dump_psb +debug/i386/intel_gsic +debug/i386/powernow-k8-decode +debug/x86_64/centrino-decode +debug/x86_64/powernow-k8-decode diff --git a/tools/power/cpupower/Makefile b/tools/power/cpupower/Makefile index e8a03aceceb..2e2ba2efa0d 100644 --- a/tools/power/cpupower/Makefile +++ b/tools/power/cpupower/Makefile @@ -19,6 +19,16 @@ # along with this program; if not, write to the Free Software # Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA # +OUTPUT=./ +ifeq ("$(origin O)", "command line") + OUTPUT := $(O)/ +endif + +ifneq ($(OUTPUT),) +# check that the output directory actually exists +OUTDIR := $(shell cd $(OUTPUT) && /bin/pwd) +$(if $(OUTDIR),, $(error output directory "$(OUTPUT)" does not exist)) +endif # --- CONFIGURATION BEGIN --- @@ -52,7 +62,7 @@ LIB_MAJ= 0.0.0 LIB_MIN= 0 PACKAGE = cpupower -PACKAGE_BUGREPORT = cpufreq@vger.kernel.org +PACKAGE_BUGREPORT = linux-pm@vger.kernel.org LANGUAGES = de fr it cs pt @@ -87,6 +97,7 @@ AR = $(CROSS)ar STRIP = $(CROSS)strip RANLIB = $(CROSS)ranlib HOSTCC = gcc +MKDIR = mkdir # Now we set up the build system @@ -95,12 +106,12 @@ HOSTCC = gcc # set up PWD so that older versions of make will work with our build. PWD = $(shell pwd) -GMO_FILES = ${shell for HLANG in ${LANGUAGES}; do echo po/$$HLANG.gmo; done;} +GMO_FILES = ${shell for HLANG in ${LANGUAGES}; do echo $(OUTPUT)po/$$HLANG.gmo; done;} export CROSS CC AR STRIP RANLIB CFLAGS LDFLAGS LIB_OBJS # check if compiler option is supported -cc-supports = ${shell if $(CC) ${1} -S -o /dev/null -xc /dev/null > /dev/null 2>&1; then echo "$(1)"; fi;} +cc-supports = ${shell if $(CC) ${1} -S -o /dev/null -x c /dev/null > /dev/null 2>&1; then echo "$(1)"; fi;} # use '-Os' optimization if available, else use -O2 OPTIMIZATION := $(call cc-supports,-Os,-O2) @@ -117,20 +128,25 @@ UTIL_OBJS = utils/helpers/amd.o utils/helpers/topology.o utils/helpers/msr.o \ utils/helpers/sysfs.o utils/helpers/misc.o utils/helpers/cpuid.o \ utils/helpers/pci.o utils/helpers/bitmask.o \ utils/idle_monitor/nhm_idle.o utils/idle_monitor/snb_idle.o \ + utils/idle_monitor/hsw_ext_idle.o \ utils/idle_monitor/amd_fam14h_idle.o utils/idle_monitor/cpuidle_sysfs.o \ utils/idle_monitor/mperf_monitor.o utils/idle_monitor/cpupower-monitor.o \ utils/cpupower.o utils/cpufreq-info.o utils/cpufreq-set.o \ - utils/cpupower-set.o utils/cpupower-info.o utils/cpuidle-info.o + utils/cpupower-set.o utils/cpupower-info.o utils/cpuidle-info.o \ + utils/cpuidle-set.o + +UTIL_SRC := $(UTIL_OBJS:.o=.c) + +UTIL_OBJS := $(addprefix $(OUTPUT),$(UTIL_OBJS)) UTIL_HEADERS = utils/helpers/helpers.h utils/idle_monitor/cpupower-monitor.h \ utils/helpers/bitmask.h \ utils/idle_monitor/idle_monitors.h utils/idle_monitor/idle_monitors.def -UTIL_SRC := $(UTIL_OBJS:.o=.c) - LIB_HEADERS = lib/cpufreq.h lib/sysfs.h LIB_SRC = lib/cpufreq.c lib/sysfs.c LIB_OBJS = lib/cpufreq.o lib/sysfs.o +LIB_OBJS := $(addprefix $(OUTPUT),$(LIB_OBJS)) CFLAGS += -pipe @@ -168,88 +184,98 @@ endif # the actual make rules -all: libcpupower cpupower $(COMPILE_NLS) $(COMPILE_BENCH) +all: libcpupower $(OUTPUT)cpupower $(COMPILE_NLS) $(COMPILE_BENCH) -lib/%.o: $(LIB_SRC) $(LIB_HEADERS) +$(OUTPUT)lib/%.o: $(LIB_SRC) $(LIB_HEADERS) $(ECHO) " CC " $@ $(QUIET) $(CC) $(CFLAGS) -fPIC -o $@ -c lib/$*.c -libcpupower.so.$(LIB_MAJ): $(LIB_OBJS) +$(OUTPUT)libcpupower.so.$(LIB_MAJ): $(LIB_OBJS) $(ECHO) " LD " $@ $(QUIET) $(CC) -shared $(CFLAGS) $(LDFLAGS) -o $@ \ -Wl,-soname,libcpupower.so.$(LIB_MIN) $(LIB_OBJS) - @ln -sf $@ libcpupower.so - @ln -sf $@ libcpupower.so.$(LIB_MIN) + @ln -sf $(@F) $(OUTPUT)libcpupower.so + @ln -sf $(@F) $(OUTPUT)libcpupower.so.$(LIB_MIN) -libcpupower: libcpupower.so.$(LIB_MAJ) +libcpupower: $(OUTPUT)libcpupower.so.$(LIB_MAJ) # Let all .o files depend on its .c file and all headers # Might be worth to put this into utils/Makefile at some point of time $(UTIL_OBJS): $(UTIL_HEADERS) -.c.o: +$(OUTPUT)%.o: %.c $(ECHO) " CC " $@ $(QUIET) $(CC) $(CFLAGS) -I./lib -I ./utils -o $@ -c $*.c -cpupower: $(UTIL_OBJS) libcpupower.so.$(LIB_MAJ) +$(OUTPUT)cpupower: $(UTIL_OBJS) $(OUTPUT)libcpupower.so.$(LIB_MAJ) $(ECHO) " CC " $@ - $(QUIET) $(CC) $(CFLAGS) $(LDFLAGS) -lcpupower -lrt -lpci -L. -o $@ $(UTIL_OBJS) + $(QUIET) $(CC) $(CFLAGS) $(LDFLAGS) $(UTIL_OBJS) -lcpupower -lrt -lpci -L$(OUTPUT) -o $@ $(QUIET) $(STRIPCMD) $@ -po/$(PACKAGE).pot: $(UTIL_SRC) +$(OUTPUT)po/$(PACKAGE).pot: $(UTIL_SRC) $(ECHO) " GETTEXT " $@ $(QUIET) xgettext --default-domain=$(PACKAGE) --add-comments \ - --keyword=_ --keyword=N_ $(UTIL_SRC) && \ - test -f $(PACKAGE).po && \ - mv -f $(PACKAGE).po po/$(PACKAGE).pot + --keyword=_ --keyword=N_ $(UTIL_SRC) -p $(@D) -o $(@F) -po/%.gmo: po/%.po +$(OUTPUT)po/%.gmo: po/%.po $(ECHO) " MSGFMT " $@ $(QUIET) msgfmt -o $@ po/$*.po create-gmo: ${GMO_FILES} -update-po: po/$(PACKAGE).pot +update-po: $(OUTPUT)po/$(PACKAGE).pot $(ECHO) " MSGMRG " $@ $(QUIET) @for HLANG in $(LANGUAGES); do \ echo -n "Updating $$HLANG "; \ - if msgmerge po/$$HLANG.po po/$(PACKAGE).pot -o \ - po/$$HLANG.new.po; then \ - mv -f po/$$HLANG.new.po po/$$HLANG.po; \ + if msgmerge po/$$HLANG.po $< -o \ + $(OUTPUT)po/$$HLANG.new.po; then \ + mv -f $(OUTPUT)po/$$HLANG.new.po $(OUTPUT)po/$$HLANG.po; \ else \ echo "msgmerge for $$HLANG failed!"; \ - rm -f po/$$HLANG.new.po; \ + rm -f $(OUTPUT)po/$$HLANG.new.po; \ fi; \ done; -compile-bench: libcpupower.so.$(LIB_MAJ) - @V=$(V) confdir=$(confdir) $(MAKE) -C bench +compile-bench: $(OUTPUT)libcpupower.so.$(LIB_MAJ) + @V=$(V) confdir=$(confdir) $(MAKE) -C bench O=$(OUTPUT) + +# we compile into subdirectories. if the target directory is not the +# source directory, they might not exists. So we depend the various +# files onto their directories. +DIRECTORY_DEPS = $(LIB_OBJS) $(UTIL_OBJS) $(GMO_FILES) +$(DIRECTORY_DEPS): | $(sort $(dir $(DIRECTORY_DEPS))) + +# In the second step, we make a rule to actually create these directories +$(sort $(dir $(DIRECTORY_DEPS))): + $(ECHO) " MKDIR " $@ + $(QUIET) $(MKDIR) -p $@ 2>/dev/null clean: - -find . \( -not -type d \) -and \( -name '*~' -o -name '*.[oas]' \) -type f -print \ + -find $(OUTPUT) \( -not -type d \) -and \( -name '*~' -o -name '*.[oas]' \) -type f -print \ | xargs rm -f - -rm -f $(UTIL_BINS) - -rm -f $(IDLE_OBJS) - -rm -f cpupower - -rm -f libcpupower.so* - -rm -rf po/*.gmo po/*.pot - $(MAKE) -C bench clean + -rm -f $(OUTPUT)cpupower + -rm -f $(OUTPUT)libcpupower.so* + -rm -rf $(OUTPUT)po/*.gmo + -rm -rf $(OUTPUT)po/*.pot + $(MAKE) -C bench O=$(OUTPUT) clean install-lib: $(INSTALL) -d $(DESTDIR)${libdir} - $(CP) libcpupower.so* $(DESTDIR)${libdir}/ + $(CP) $(OUTPUT)libcpupower.so* $(DESTDIR)${libdir}/ $(INSTALL) -d $(DESTDIR)${includedir} $(INSTALL_DATA) lib/cpufreq.h $(DESTDIR)${includedir}/cpufreq.h install-tools: $(INSTALL) -d $(DESTDIR)${bindir} - $(INSTALL_PROGRAM) cpupower $(DESTDIR)${bindir} + $(INSTALL_PROGRAM) $(OUTPUT)cpupower $(DESTDIR)${bindir} install-man: $(INSTALL_DATA) -D man/cpupower.1 $(DESTDIR)${mandir}/man1/cpupower.1 $(INSTALL_DATA) -D man/cpupower-frequency-set.1 $(DESTDIR)${mandir}/man1/cpupower-frequency-set.1 $(INSTALL_DATA) -D man/cpupower-frequency-info.1 $(DESTDIR)${mandir}/man1/cpupower-frequency-info.1 + $(INSTALL_DATA) -D man/cpupower-idle-set.1 $(DESTDIR)${mandir}/man1/cpupower-idle-set.1 + $(INSTALL_DATA) -D man/cpupower-idle-info.1 $(DESTDIR)${mandir}/man1/cpupower-idle-info.1 $(INSTALL_DATA) -D man/cpupower-set.1 $(DESTDIR)${mandir}/man1/cpupower-set.1 $(INSTALL_DATA) -D man/cpupower-info.1 $(DESTDIR)${mandir}/man1/cpupower-info.1 $(INSTALL_DATA) -D man/cpupower-monitor.1 $(DESTDIR)${mandir}/man1/cpupower-monitor.1 @@ -257,13 +283,13 @@ install-man: install-gmo: $(INSTALL) -d $(DESTDIR)${localedir} for HLANG in $(LANGUAGES); do \ - echo '$(INSTALL_DATA) -D po/$$HLANG.gmo $(DESTDIR)${localedir}/$$HLANG/LC_MESSAGES/cpupower.mo'; \ - $(INSTALL_DATA) -D po/$$HLANG.gmo $(DESTDIR)${localedir}/$$HLANG/LC_MESSAGES/cpupower.mo; \ + echo '$(INSTALL_DATA) -D $(OUTPUT)po/$$HLANG.gmo $(DESTDIR)${localedir}/$$HLANG/LC_MESSAGES/cpupower.mo'; \ + $(INSTALL_DATA) -D $(OUTPUT)po/$$HLANG.gmo $(DESTDIR)${localedir}/$$HLANG/LC_MESSAGES/cpupower.mo; \ done; install-bench: @#DESTDIR must be set from outside to survive - @sbindir=$(sbindir) bindir=$(bindir) docdir=$(docdir) confdir=$(confdir) $(MAKE) -C bench install + @sbindir=$(sbindir) bindir=$(bindir) docdir=$(docdir) confdir=$(confdir) $(MAKE) -C bench O=$(OUTPUT) install install: all install-lib install-tools install-man $(INSTALL_NLS) $(INSTALL_BENCH) @@ -271,8 +297,12 @@ uninstall: - rm -f $(DESTDIR)${libdir}/libcpupower.* - rm -f $(DESTDIR)${includedir}/cpufreq.h - rm -f $(DESTDIR)${bindir}/utils/cpupower - - rm -f $(DESTDIR)${mandir}/man1/cpufreq-set.1 - - rm -f $(DESTDIR)${mandir}/man1/cpufreq-info.1 + - rm -f $(DESTDIR)${mandir}/man1/cpupower.1 + - rm -f $(DESTDIR)${mandir}/man1/cpupower-frequency-set.1 + - rm -f $(DESTDIR)${mandir}/man1/cpupower-frequency-info.1 + - rm -f $(DESTDIR)${mandir}/man1/cpupower-set.1 + - rm -f $(DESTDIR)${mandir}/man1/cpupower-info.1 + - rm -f $(DESTDIR)${mandir}/man1/cpupower-monitor.1 - for HLANG in $(LANGUAGES); do \ rm -f $(DESTDIR)${localedir}/$$HLANG/LC_MESSAGES/cpupower.mo; \ done; diff --git a/tools/power/cpupower/README b/tools/power/cpupower/README index fd9d4c0d668..1c68f47663b 100644 --- a/tools/power/cpupower/README +++ b/tools/power/cpupower/README @@ -1,6 +1,4 @@ -The cpufrequtils package (homepage: -http://www.kernel.org/pub/linux/utils/kernel/cpufreq/cpufrequtils.html ) -consists of the following elements: +The cpupower package consists of the following elements: requirements ------------ @@ -11,10 +9,10 @@ providing cpuid.h is needed. For both it's not explicitly checked for (yet). -libcpufreq +libcpupower ---------- -"libcpufreq" is a library which offers a unified access method for userspace +"libcpupower" is a library which offers a unified access method for userspace tools and programs to the cpufreq core and drivers in the Linux kernel. This allows for code reduction in userspace tools, a clean implementation of the interaction to the cpufreq core, and support for both the sysfs and proc @@ -28,22 +26,22 @@ make su make install -should suffice on most systems. It builds default libcpufreq, -cpufreq-set and cpufreq-info files and installs them in /usr/lib and -/usr/bin, respectively. If you want to set up the paths differently and/or -want to configure the package to your specific needs, you need to open -"Makefile" with an editor of your choice and edit the block marked -CONFIGURATION. +should suffice on most systems. It builds libcpupower to put in +/usr/lib; cpupower, cpufreq-bench_plot.sh to put in /usr/bin; and +cpufreq-bench to put in /usr/sbin. If you want to set up the paths +differently and/or want to configure the package to your specific +needs, you need to open "Makefile" with an editor of your choice and +edit the block marked CONFIGURATION. THANKS ------ Many thanks to Mattia Dongili who wrote the autotoolization and -libtoolization, the manpages and the italian language file for cpufrequtils; +libtoolization, the manpages and the italian language file for cpupower; to Dave Jones for his feedback and his dump_psb tool; to Bruno Ducrot for his powernow-k8-decode and intel_gsic tools as well as the french language file; and to various others commenting on the previous (pre-)releases of -cpufrequtils. +cpupower. Dominik Brodowski diff --git a/tools/power/cpupower/ToDo b/tools/power/cpupower/ToDo index 874b78b586e..6e8b89f282e 100644 --- a/tools/power/cpupower/ToDo +++ b/tools/power/cpupower/ToDo @@ -3,7 +3,6 @@ ToDos sorted by priority: - Use bitmask functions to parse CPU topology more robust (current implementation has issues on AMD) - Try to read out boost states and frequencies on Intel -- Adjust README - Somewhere saw the ability to read power consumption of RAM from HW on Intel SandyBridge -> another monitor? - Add another c1e debug idle monitor diff --git a/tools/power/cpupower/bench/Makefile b/tools/power/cpupower/bench/Makefile index 2b67606fc3e..7ec7021a29c 100644 --- a/tools/power/cpupower/bench/Makefile +++ b/tools/power/cpupower/bench/Makefile @@ -1,29 +1,36 @@ -LIBS = -L../ -lm -lcpupower +OUTPUT := ./ +ifeq ("$(origin O)", "command line") +ifneq ($(O),) + OUTPUT := $(O)/ +endif +endif -OBJS = main.o parse.o system.o benchmark.o +LIBS = -L../ -L$(OUTPUT) -lm -lcpupower + +OBJS = $(OUTPUT)main.o $(OUTPUT)parse.o $(OUTPUT)system.o $(OUTPUT)benchmark.o CFLAGS += -D_GNU_SOURCE -I../lib -DDEFAULT_CONFIG_FILE=\"$(confdir)/cpufreq-bench.conf\" -%.o : %.c +$(OUTPUT)%.o : %.c $(ECHO) " CC " $@ $(QUIET) $(CC) -c $(CFLAGS) $< -o $@ -cpufreq-bench: $(OBJS) +$(OUTPUT)cpufreq-bench: $(OBJS) $(ECHO) " CC " $@ $(QUIET) $(CC) -o $@ $(CFLAGS) $(OBJS) $(LIBS) -all: cpufreq-bench +all: $(OUTPUT)cpufreq-bench install: mkdir -p $(DESTDIR)/$(sbindir) mkdir -p $(DESTDIR)/$(bindir) mkdir -p $(DESTDIR)/$(docdir) mkdir -p $(DESTDIR)/$(confdir) - install -m 755 cpufreq-bench $(DESTDIR)/$(sbindir)/cpufreq-bench + install -m 755 $(OUTPUT)cpufreq-bench $(DESTDIR)/$(sbindir)/cpufreq-bench install -m 755 cpufreq-bench_plot.sh $(DESTDIR)/$(bindir)/cpufreq-bench_plot.sh install -m 644 README-BENCH $(DESTDIR)/$(docdir)/README-BENCH install -m 755 cpufreq-bench_script.sh $(DESTDIR)/$(docdir)/cpufreq-bench_script.sh install -m 644 example.cfg $(DESTDIR)/$(confdir)/cpufreq-bench.conf clean: - rm -f *.o - rm -f cpufreq-bench + rm -f $(OUTPUT)*.o + rm -f $(OUTPUT)cpufreq-bench diff --git a/tools/power/cpupower/debug/i386/Makefile b/tools/power/cpupower/debug/i386/Makefile index d08cc1ead9b..c05cc0ac80c 100644 --- a/tools/power/cpupower/debug/i386/Makefile +++ b/tools/power/cpupower/debug/i386/Makefile @@ -1,20 +1,41 @@ +OUTPUT=./ +ifeq ("$(origin O)", "command line") + OUTPUT := $(O)/ +endif + +DESTDIR = +bindir = /usr/bin + +INSTALL = /usr/bin/install + + default: all -centrino-decode: centrino-decode.c - $(CC) $(CFLAGS) -o centrino-decode centrino-decode.c +$(OUTPUT)centrino-decode: centrino-decode.c + $(CC) $(CFLAGS) -o $@ centrino-decode.c -dump_psb: dump_psb.c - $(CC) $(CFLAGS) -o dump_psb dump_psb.c +$(OUTPUT)dump_psb: dump_psb.c + $(CC) $(CFLAGS) -o $@ dump_psb.c -intel_gsic: intel_gsic.c - $(CC) $(CFLAGS) -o intel_gsic -llrmi intel_gsic.c +$(OUTPUT)intel_gsic: intel_gsic.c + $(CC) $(CFLAGS) -o $@ -llrmi intel_gsic.c -powernow-k8-decode: powernow-k8-decode.c - $(CC) $(CFLAGS) -o powernow-k8-decode powernow-k8-decode.c +$(OUTPUT)powernow-k8-decode: powernow-k8-decode.c + $(CC) $(CFLAGS) -o $@ powernow-k8-decode.c -all: centrino-decode dump_psb intel_gsic powernow-k8-decode +all: $(OUTPUT)centrino-decode $(OUTPUT)dump_psb $(OUTPUT)intel_gsic $(OUTPUT)powernow-k8-decode clean: - rm -rf centrino-decode dump_psb intel_gsic powernow-k8-decode + rm -rf $(OUTPUT)centrino-decode + rm -rf $(OUTPUT)dump_psb + rm -rf $(OUTPUT)intel_gsic + rm -rf $(OUTPUT)powernow-k8-decode + +install: + $(INSTALL) -d $(DESTDIR)${bindir} + $(INSTALL) $(OUTPUT)centrino-decode $(DESTDIR)${bindir} + $(INSTALL) $(OUTPUT)powernow-k8-decode $(DESTDIR)${bindir} + $(INSTALL) $(OUTPUT)dump_psb $(DESTDIR)${bindir} + $(INSTALL) $(OUTPUT)intel_gsic $(DESTDIR)${bindir} -.PHONY: all default clean +.PHONY: all default clean install diff --git a/tools/power/cpupower/debug/i386/intel_gsic.c b/tools/power/cpupower/debug/i386/intel_gsic.c index 53f5293c9c9..d032c826d42 100644 --- a/tools/power/cpupower/debug/i386/intel_gsic.c +++ b/tools/power/cpupower/debug/i386/intel_gsic.c @@ -66,7 +66,7 @@ int main (void) printf("ecx = 0x%.8x\n", r.ecx); printf("edx = 0x%.8x\n", r.edx); printf("Note also that some BIOS do not support the initial " - "GSIC call, but the newer\nspeeedstep-smi driver may " + "GSIC call, but the newer\nspeedstep-smi driver may " "work.\nFor this, you need to pass some arguments to " "the speedstep-smi driver:\n"); printf("\tsmi_cmd=0x?? smi_port=0x?? smi_sig=1\n"); diff --git a/tools/power/cpupower/debug/kernel/cpufreq-test_tsc.c b/tools/power/cpupower/debug/kernel/cpufreq-test_tsc.c index 66cace601e5..5224ee5b392 100644 --- a/tools/power/cpupower/debug/kernel/cpufreq-test_tsc.c +++ b/tools/power/cpupower/debug/kernel/cpufreq-test_tsc.c @@ -18,19 +18,16 @@ * 5.) if the third value, "diff_pmtmr", changes between 2. and 4., the * TSC-based delay routine on the Linux kernel does not correctly * handle the cpufreq transition. Please report this to - * cpufreq@vger.kernel.org + * linux-pm@vger.kernel.org */ #include <linux/kernel.h> #include <linux/module.h> #include <linux/init.h> #include <linux/delay.h> - +#include <linux/acpi.h> #include <asm/io.h> -#include <acpi/acpi_bus.h> -#include <acpi/acpi_drivers.h> - static int pm_tmr_ioport = 0; /*helper function to safely read acpi pm timesource*/ diff --git a/tools/power/cpupower/debug/x86_64/Makefile b/tools/power/cpupower/debug/x86_64/Makefile index 3326217dd31..1c521452671 100644 --- a/tools/power/cpupower/debug/x86_64/Makefile +++ b/tools/power/cpupower/debug/x86_64/Makefile @@ -1,14 +1,30 @@ +OUTPUT=./ +ifeq ("$(origin O)", "command line") + OUTPUT := $(O)/ +endif + +DESTDIR = +bindir = /usr/bin + +INSTALL = /usr/bin/install + + default: all -centrino-decode: ../i386/centrino-decode.c +$(OUTPUT)centrino-decode: ../i386/centrino-decode.c $(CC) $(CFLAGS) -o $@ $< -powernow-k8-decode: ../i386/powernow-k8-decode.c +$(OUTPUT)powernow-k8-decode: ../i386/powernow-k8-decode.c $(CC) $(CFLAGS) -o $@ $< -all: centrino-decode powernow-k8-decode +all: $(OUTPUT)centrino-decode $(OUTPUT)powernow-k8-decode clean: - rm -rf centrino-decode powernow-k8-decode + rm -rf $(OUTPUT)centrino-decode $(OUTPUT)powernow-k8-decode + +install: + $(INSTALL) -d $(DESTDIR)${bindir} + $(INSTALL) $(OUTPUT)centrino-decode $(DESTDIR)${bindir} + $(INSTALL) $(OUTPUT)powernow-k8-decode $(DESTDIR)${bindir} -.PHONY: all default clean +.PHONY: all default clean install diff --git a/tools/power/cpupower/man/cpupower-frequency-info.1 b/tools/power/cpupower/man/cpupower-frequency-info.1 index bb60a8d1e45..9c85a382e35 100644 --- a/tools/power/cpupower/man/cpupower-frequency-info.1 +++ b/tools/power/cpupower/man/cpupower-frequency-info.1 @@ -1,4 +1,4 @@ -.TH "cpupower-frequency-info" "1" "0.1" "Mattia Dongili" "" +.TH "CPUPOWER\-FREQUENCY\-INFO" "1" "0.1" "" "cpupower Manual" .SH "NAME" .LP cpupower frequency\-info \- Utility to retrieve cpufreq kernel information @@ -50,8 +50,9 @@ Prints out information like provided by the /proc/cpufreq interface in 2.4. and \fB\-m\fR \fB\-\-human\fR human\-readable output for the \-f, \-w, \-s and \-y parameters. .TP -\fB\-h\fR \fB\-\-help\fR -Prints out the help screen. +\fB\-n\fR \fB\-\-no-rounding\fR +Output frequencies and latencies without rounding off values. +.TP .SH "REMARKS" .LP By default only values of core zero are displayed. How to display settings of diff --git a/tools/power/cpupower/man/cpupower-frequency-set.1 b/tools/power/cpupower/man/cpupower-frequency-set.1 index 685f469093a..3eacc8d03d1 100644 --- a/tools/power/cpupower/man/cpupower-frequency-set.1 +++ b/tools/power/cpupower/man/cpupower-frequency-set.1 @@ -1,4 +1,4 @@ -.TH "cpupower-freqency-set" "1" "0.1" "Mattia Dongili" "" +.TH "CPUPOWER\-FREQUENCY\-SET" "1" "0.1" "" "cpupower Manual" .SH "NAME" .LP cpupower frequency\-set \- A small tool which allows to modify cpufreq settings. @@ -26,8 +26,6 @@ specific frequency to be set. Requires userspace governor to be available and lo \fB\-r\fR \fB\-\-related\fR modify all hardware-related CPUs at the same time .TP -\fB\-h\fR \fB\-\-help\fR -Prints out the help screen. .SH "REMARKS" .LP By default values are applied on all cores. How to modify single core diff --git a/tools/power/cpupower/man/cpupower-idle-info.1 b/tools/power/cpupower/man/cpupower-idle-info.1 new file mode 100644 index 00000000000..7b3646adb92 --- /dev/null +++ b/tools/power/cpupower/man/cpupower-idle-info.1 @@ -0,0 +1,91 @@ +.TH "CPUPOWER-IDLE-INFO" "1" "0.1" "" "cpupower Manual" +.SH "NAME" +.LP +cpupower idle\-info \- Utility to retrieve cpu idle kernel information +.SH "SYNTAX" +.LP +cpupower [ \-c cpulist ] idle\-info [\fIoptions\fP] +.SH "DESCRIPTION" +.LP +A tool which prints out per cpu idle information helpful to developers and interested users. +.SH "OPTIONS" +.LP +.TP +\fB\-f\fR \fB\-\-silent\fR +Only print a summary of all available C-states in the system. +.TP +\fB\-e\fR \fB\-\-proc\fR +deprecated. +Prints out idle information in old /proc/acpi/processor/*/power format. This +interface has been removed from the kernel for quite some time, do not let +further code depend on this option, best do not use it. + +.SH IDLE\-INFO DESCRIPTIONS +CPU sleep state statistics and descriptions are retrieved from sysfs files, +exported by the cpuidle kernel subsystem. The kernel only updates these +statistics when it enters or leaves an idle state, therefore on a very idle or +a very busy system, these statistics may not be accurate. They still provide a +good overview about the usage and availability of processor sleep states on +the platform. + +Be aware that the sleep states as exported by the hardware or BIOS and used by +the Linux kernel may not exactly reflect the capabilities of the +processor. This often is the case on the X86 architecture when the acpi_idle +driver is used. It is also possible that the hardware overrules the kernel +requests, due to internal activity monitors or other reasons. +On recent X86 platforms it is often possible to read out hardware registers +which monitor the duration of sleep states the processor resided in. The +cpupower monitor tool (cpupower\-monitor(1)) can be used to show real sleep +state residencies. Please refer to the architecture specific description +section below. + +.SH IDLE\-INFO ARCHITECTURE SPECIFIC DESCRIPTIONS +.SS "X86" +POLL idle state + +If cpuidle is active, X86 platforms have one special idle state. +The POLL idle state is not a real idle state, it does not save any +power. Instead, a busy\-loop is executed doing nothing for a short period of +time. This state is used if the kernel knows that work has to be processed +very soon and entering any real hardware idle state may result in a slight +performance penalty. + +There exist two different cpuidle drivers on the X86 architecture platform: + +"acpi_idle" cpuidle driver + +The acpi_idle cpuidle driver retrieves available sleep states (C\-states) from +the ACPI BIOS tables (from the _CST ACPI function on recent platforms or from +the FADT BIOS table on older ones). +The C1 state is not retrieved from ACPI tables. If the C1 state is entered, +the kernel will call the hlt instruction (or mwait on Intel). + +"intel_idle" cpuidle driver + +In kernel 2.6.36 the intel_idle driver was introduced. +It only serves recent Intel CPUs (Nehalem, Westmere, Sandybridge, Atoms or +newer). On older Intel CPUs the acpi_idle driver is still used (if the BIOS +provides C\-state ACPI tables). +The intel_idle driver knows the sleep state capabilities of the processor and +ignores ACPI BIOS exported processor sleep states tables. + +.SH "REMARKS" +.LP +By default only values of core zero are displayed. How to display settings of +other cores is described in the cpupower(1) manpage in the \-\-cpu option +section. +.SH REFERENCES +http://www.acpi.info/spec.htm +.SH "FILES" +.nf +\fI/sys/devices/system/cpu/cpu*/cpuidle/state*\fP +\fI/sys/devices/system/cpu/cpuidle/*\fP +.fi +.SH "AUTHORS" +.nf +Thomas Renninger <trenn@suse.de> +.fi +.SH "SEE ALSO" +.LP +cpupower(1), cpupower\-monitor(1), cpupower\-info(1), cpupower\-set(1), +cpupower\-idle\-set(1) diff --git a/tools/power/cpupower/man/cpupower-idle-set.1 b/tools/power/cpupower/man/cpupower-idle-set.1 new file mode 100644 index 00000000000..3e6799d7a79 --- /dev/null +++ b/tools/power/cpupower/man/cpupower-idle-set.1 @@ -0,0 +1,77 @@ +.TH "CPUPOWER-IDLE-SET" "1" "0.1" "" "cpupower Manual" +.SH "NAME" +.LP +cpupower idle\-set \- Utility to set cpu idle state specific kernel options +.SH "SYNTAX" +.LP +cpupower [ \-c cpulist ] idle\-info [\fIoptions\fP] +.SH "DESCRIPTION" +.LP +The cpupower idle\-set subcommand allows to set cpu idle, also called cpu +sleep state, specific options offered by the kernel. One example is disabling +sleep states. This can be handy for power vs performance tuning. +.SH "OPTIONS" +.LP +.TP +\fB\-d\fR \fB\-\-disable\fR <STATE_NO> +Disable a specific processor sleep state. +.TP +\fB\-e\fR \fB\-\-enable\fR <STATE_NO> +Enable a specific processor sleep state. +.TP +\fB\-D\fR \fB\-\-disable-by-latency\fR <LATENCY> +Disable all idle states with a equal or higher latency than <LATENCY> +.TP +\fB\-E\fR \fB\-\-enable-all\fR +Enable all idle states if not enabled already. + +.SH "REMARKS" +.LP +Cpuidle Governors Policy on Disabling Sleep States + +.RS 4 +Depending on the used cpuidle governor, implementing the kernel policy +how to choose sleep states, subsequent sleep states on this core, might get +disabled as well. + +There are two cpuidle governors ladder and menu. While the ladder +governor is always available, if CONFIG_CPU_IDLE is selected, the +menu governor additionally requires CONFIG_NO_HZ. + +The behavior and the effect of the disable variable depends on the +implementation of a particular governor. In the ladder governor, for +example, it is not coherent, i.e. if one is disabling a light state, +then all deeper states are disabled as well. Likewise, if one enables a +deep state but a lighter state still is disabled, then this has no effect. +.RE +.LP +Disabling the Lightest Sleep State may not have any Affect + +.RS 4 +If criteria are not met to enter deeper sleep states and the lightest sleep +state is chosen when idle, the kernel may still enter this sleep state, +irrespective of whether it is disabled or not. This is also reflected in +the usage count of the disabled sleep state when using the cpupower idle-info +command. +.RE +.LP +Selecting specific CPU Cores + +.RS 4 +By default processor sleep states of all CPU cores are set. Please refer +to the cpupower(1) manpage in the \-\-cpu option section how to disable +C-states of specific cores. +.RE +.SH "FILES" +.nf +\fI/sys/devices/system/cpu/cpu*/cpuidle/state*\fP +\fI/sys/devices/system/cpu/cpuidle/*\fP +.fi +.SH "AUTHORS" +.nf +Thomas Renninger <trenn@suse.de> +.fi +.SH "SEE ALSO" +.LP +cpupower(1), cpupower\-monitor(1), cpupower\-info(1), cpupower\-set(1), +cpupower\-idle\-info(1) diff --git a/tools/power/cpupower/man/cpupower-info.1 b/tools/power/cpupower/man/cpupower-info.1 index 58e21196f17..340bcd0be7d 100644 --- a/tools/power/cpupower/man/cpupower-info.1 +++ b/tools/power/cpupower/man/cpupower-info.1 @@ -3,7 +3,7 @@ cpupower\-info \- Shows processor power related kernel or hardware configurations .SH SYNOPSIS .ft B -.B cpupower info [ \-b ] [ \-s ] [ \-m ] +.B cpupower info [ \-b ] .SH DESCRIPTION \fBcpupower info \fP shows kernel configurations or processor hardware diff --git a/tools/power/cpupower/man/cpupower-monitor.1 b/tools/power/cpupower/man/cpupower-monitor.1 index d5cfa265c3d..914cbb9d9cd 100644 --- a/tools/power/cpupower/man/cpupower-monitor.1 +++ b/tools/power/cpupower/man/cpupower-monitor.1 @@ -7,11 +7,11 @@ cpupower\-monitor \- Report processor frequency and idle statistics .RB "\-l" .B cpupower monitor -.RB [ "\-m <mon1>," [ "<mon2>,..." ] ] +.RB [ -c ] [ "\-m <mon1>," [ "<mon2>,..." ] ] .RB [ "\-i seconds" ] .br .B cpupower monitor -.RB [ "\-m <mon1>," [ "<mon2>,..." ] ] +.RB [ -c ][ "\-m <mon1>," [ "<mon2>,..." ] ] .RB command .br .SH DESCRIPTION @@ -64,6 +64,17 @@ Only display specific monitors. Use the monitor string(s) provided by \-l option Measure intervall. .RE .PP +\-c +.RS 4 +Schedule the process on every core before starting and ending measuring. +This could be needed for the Idle_Stats monitor when no other MSR based +monitor (has to be run on the core that is measured) is run in parallel. +This is to wake up the processors from deeper sleep states and let the +kernel re +-account its cpuidle (C-state) information before reading the +cpuidle timings from sysfs. +.RE +.PP command .RS 4 Measure idle and frequency characteristics of an arbitrary command/workload. @@ -99,15 +110,23 @@ May work poorly on Linux-2.6.20 through 2.6.29, as the \fBacpi-cpufreq \fP kernel frequency driver periodically cleared aperf/mperf registers in those kernels. -.SS "Nehalem" "SandyBridge" +.SS "Nehalem" "SandyBridge" "HaswellExtended" Intel Core and Package sleep state counters. Threads (hyperthreaded cores) may not be able to enter deeper core states if its sibling is utilized. Deepest package sleep states may in reality show up as machine/platform wide sleep states and can only be entered if all cores are idle. Look up Intel manuals (some are provided in the References section) for further details. - -.SS "Ontario" "Liano" +The monitors are named after the CPU family where the sleep state capabilities +got introduced and may not match exactly the CPU name of the platform. +For example an IvyBridge processor has sleep state capabilities which got +introduced in Nehalem and SandyBridge processor families. +Thus on an IvyBridge processor one will get Nehalem and SandyBridge sleep +state monitors. +HaswellExtended extra package sleep state capabilities are available only in a +specific Haswell (family 0x45) and probably also other future processors. + +.SS "Fam_12h" "Fam_14h" AMD laptop and desktop processor (family 12h and 14h) sleep state counters. The registers are accessed via PCI and therefore can still be read out while cores have been offlined. diff --git a/tools/power/cpupower/man/cpupower-set.1 b/tools/power/cpupower/man/cpupower-set.1 index c4954a9fe4e..2bcc696f449 100644 --- a/tools/power/cpupower/man/cpupower-set.1 +++ b/tools/power/cpupower/man/cpupower-set.1 @@ -3,7 +3,7 @@ cpupower\-set \- Set processor power related kernel or hardware configurations .SH SYNOPSIS .ft B -.B cpupower set [ \-b VAL ] [ \-s VAL ] [ \-m VAL ] +.B cpupower set [ \-b VAL ] .SH DESCRIPTION @@ -55,44 +55,6 @@ Use \fBcpupower -c all info -b\fP to verify. This options needs the msr kernel driver (CONFIG_X86_MSR) loaded. .RE -.PP -\-\-sched\-mc, \-m [ VAL ] -.RE -\-\-sched\-smt, \-s [ VAL ] -.RS 4 -\-\-sched\-mc utilizes cores in one processor package/socket first before -processes are scheduled to other processor packages/sockets. - -\-\-sched\-smt utilizes thread siblings of one processor core first before -processes are scheduled to other cores. - -The impact on power consumption and performance (positiv or negativ) heavily -depends on processor support for deep sleep states, frequency scaling and -frequency boost modes and their dependencies between other thread siblings -and processor cores. - -Taken over from kernel documentation: - -Adjust the kernel's multi-core scheduler support. - -Possible values are: -.RS 2 -0 - No power saving load balance (default value) - -1 - Fill one thread/core/package first for long running threads - -2 - Also bias task wakeups to semi-idle cpu package for power -savings -.RE - -sched_mc_power_savings is dependent upon SCHED_MC, which is -itself architecture dependent. - -sched_smt_power_savings is dependent upon SCHED_SMT, which -is itself architecture dependent. - -The two files are independent of each other. It is possible -that one file may be present without the other. .SH "SEE ALSO" cpupower-info(1), cpupower-monitor(1), powertop(1) diff --git a/tools/power/cpupower/utils/builtin.h b/tools/power/cpupower/utils/builtin.h index c10496fbe3c..2284c8ea4e2 100644 --- a/tools/power/cpupower/utils/builtin.h +++ b/tools/power/cpupower/utils/builtin.h @@ -5,6 +5,7 @@ extern int cmd_set(int argc, const char **argv); extern int cmd_info(int argc, const char **argv); extern int cmd_freq_set(int argc, const char **argv); extern int cmd_freq_info(int argc, const char **argv); +extern int cmd_idle_set(int argc, const char **argv); extern int cmd_idle_info(int argc, const char **argv); extern int cmd_monitor(int argc, const char **argv); diff --git a/tools/power/cpupower/utils/cpufreq-info.c b/tools/power/cpupower/utils/cpufreq-info.c index 28953c9a7bd..b4b90a97662 100644 --- a/tools/power/cpupower/utils/cpufreq-info.c +++ b/tools/power/cpupower/utils/cpufreq-info.c @@ -82,29 +82,42 @@ static void proc_cpufreq_output(void) } } +static int no_rounding; static void print_speed(unsigned long speed) { unsigned long tmp; - if (speed > 1000000) { - tmp = speed % 10000; - if (tmp >= 5000) - speed += 10000; - printf("%u.%02u GHz", ((unsigned int) speed/1000000), - ((unsigned int) (speed%1000000)/10000)); - } else if (speed > 100000) { - tmp = speed % 1000; - if (tmp >= 500) - speed += 1000; - printf("%u MHz", ((unsigned int) speed / 1000)); - } else if (speed > 1000) { - tmp = speed % 100; - if (tmp >= 50) - speed += 100; - printf("%u.%01u MHz", ((unsigned int) speed/1000), - ((unsigned int) (speed%1000)/100)); - } else - printf("%lu kHz", speed); + if (no_rounding) { + if (speed > 1000000) + printf("%u.%06u GHz", ((unsigned int) speed/1000000), + ((unsigned int) speed%1000000)); + else if (speed > 100000) + printf("%u MHz", (unsigned int) speed); + else if (speed > 1000) + printf("%u.%03u MHz", ((unsigned int) speed/1000), + (unsigned int) (speed%1000)); + else + printf("%lu kHz", speed); + } else { + if (speed > 1000000) { + tmp = speed%10000; + if (tmp >= 5000) + speed += 10000; + printf("%u.%02u GHz", ((unsigned int) speed/1000000), + ((unsigned int) (speed%1000000)/10000)); + } else if (speed > 100000) { + tmp = speed%1000; + if (tmp >= 500) + speed += 1000; + printf("%u MHz", ((unsigned int) speed/1000)); + } else if (speed > 1000) { + tmp = speed%100; + if (tmp >= 50) + speed += 100; + printf("%u.%01u MHz", ((unsigned int) speed/1000), + ((unsigned int) (speed%1000)/100)); + } + } return; } @@ -113,26 +126,38 @@ static void print_duration(unsigned long duration) { unsigned long tmp; - if (duration > 1000000) { - tmp = duration % 10000; - if (tmp >= 5000) - duration += 10000; - printf("%u.%02u ms", ((unsigned int) duration/1000000), - ((unsigned int) (duration%1000000)/10000)); - } else if (duration > 100000) { - tmp = duration % 1000; - if (tmp >= 500) - duration += 1000; - printf("%u us", ((unsigned int) duration / 1000)); - } else if (duration > 1000) { - tmp = duration % 100; - if (tmp >= 50) - duration += 100; - printf("%u.%01u us", ((unsigned int) duration/1000), - ((unsigned int) (duration%1000)/100)); - } else - printf("%lu ns", duration); - + if (no_rounding) { + if (duration > 1000000) + printf("%u.%06u ms", ((unsigned int) duration/1000000), + ((unsigned int) duration%1000000)); + else if (duration > 100000) + printf("%u us", ((unsigned int) duration/1000)); + else if (duration > 1000) + printf("%u.%03u us", ((unsigned int) duration/1000), + ((unsigned int) duration%1000)); + else + printf("%lu ns", duration); + } else { + if (duration > 1000000) { + tmp = duration%10000; + if (tmp >= 5000) + duration += 10000; + printf("%u.%02u ms", ((unsigned int) duration/1000000), + ((unsigned int) (duration%1000000)/10000)); + } else if (duration > 100000) { + tmp = duration%1000; + if (tmp >= 500) + duration += 1000; + printf("%u us", ((unsigned int) duration / 1000)); + } else if (duration > 1000) { + tmp = duration%100; + if (tmp >= 50) + duration += 100; + printf("%u.%01u us", ((unsigned int) duration/1000), + ((unsigned int) (duration%1000)/100)); + } else + printf("%lu ns", duration); + } return; } @@ -525,6 +550,7 @@ static struct option info_opts[] = { { .name = "latency", .has_arg = no_argument, .flag = NULL, .val = 'y'}, { .name = "proc", .has_arg = no_argument, .flag = NULL, .val = 'o'}, { .name = "human", .has_arg = no_argument, .flag = NULL, .val = 'm'}, + { .name = "no-rounding", .has_arg = no_argument, .flag = NULL, .val = 'n'}, { }, }; @@ -538,7 +564,8 @@ int cmd_freq_info(int argc, char **argv) int output_param = 0; do { - ret = getopt_long(argc, argv, "oefwldpgrasmyb", info_opts, NULL); + ret = getopt_long(argc, argv, "oefwldpgrasmybn", info_opts, + NULL); switch (ret) { case '?': output_param = '?'; @@ -575,6 +602,9 @@ int cmd_freq_info(int argc, char **argv) } human = 1; break; + case 'n': + no_rounding = 1; + break; default: fprintf(stderr, "invalid or unknown argument\n"); return EXIT_FAILURE; diff --git a/tools/power/cpupower/utils/cpufreq-set.c b/tools/power/cpupower/utils/cpufreq-set.c index dd1539eb8c6..a416de80c55 100644 --- a/tools/power/cpupower/utils/cpufreq-set.c +++ b/tools/power/cpupower/utils/cpufreq-set.c @@ -257,7 +257,7 @@ int cmd_freq_set(int argc, char **argv) print_unknown_arg(); return -EINVAL; } - if ((sscanf(optarg, "%s", gov)) != 1) { + if ((sscanf(optarg, "%19s", gov)) != 1) { print_unknown_arg(); return -EINVAL; } diff --git a/tools/power/cpupower/utils/cpuidle-info.c b/tools/power/cpupower/utils/cpuidle-info.c index b028267c137..75e66de7e7a 100644 --- a/tools/power/cpupower/utils/cpuidle-info.c +++ b/tools/power/cpupower/utils/cpuidle-info.c @@ -22,7 +22,7 @@ static void cpuidle_cpu_output(unsigned int cpu, int verbose) { - int idlestates, idlestate; + unsigned int idlestates, idlestate; char *tmp; printf(_ ("Analyzing CPU %d:\n"), cpu); @@ -31,21 +31,11 @@ static void cpuidle_cpu_output(unsigned int cpu, int verbose) if (idlestates == 0) { printf(_("CPU %u: No idle states\n"), cpu); return; - } else if (idlestates <= 0) { - printf(_("CPU %u: Can't read idle state info\n"), cpu); - return; - } - tmp = sysfs_get_idlestate_name(cpu, idlestates - 1); - if (!tmp) { - printf(_("Could not determine max idle state %u\n"), - idlestates - 1); - return; } printf(_("Number of idle states: %d\n"), idlestates); - printf(_("Available idle states:")); - for (idlestate = 1; idlestate < idlestates; idlestate++) { + for (idlestate = 0; idlestate < idlestates; idlestate++) { tmp = sysfs_get_idlestate_name(cpu, idlestate); if (!tmp) continue; @@ -57,11 +47,15 @@ static void cpuidle_cpu_output(unsigned int cpu, int verbose) if (!verbose) return; - for (idlestate = 1; idlestate < idlestates; idlestate++) { + for (idlestate = 0; idlestate < idlestates; idlestate++) { + int disabled = sysfs_is_idlestate_disabled(cpu, idlestate); + /* Disabled interface not supported on older kernels */ + if (disabled < 0) + disabled = 0; tmp = sysfs_get_idlestate_name(cpu, idlestate); if (!tmp) continue; - printf("%s:\n", tmp); + printf("%s%s:\n", tmp, (disabled) ? " (DISABLED) " : ""); free(tmp); tmp = sysfs_get_idlestate_desc(cpu, idlestate); @@ -106,21 +100,13 @@ static void cpuidle_general_output(void) static void proc_cpuidle_cpu_output(unsigned int cpu) { long max_allowed_cstate = 2000000000; - int cstates, cstate; + unsigned int cstate, cstates; cstates = sysfs_get_idlestate_count(cpu); if (cstates == 0) { - /* - * Go on and print same useless info as you'd see with - * cat /proc/acpi/processor/../power - * printf(_("CPU %u: No C-states available\n"), cpu); - * return; - */ - } else if (cstates <= 0) { - printf(_("CPU %u: Can't read C-state info\n"), cpu); + printf(_("CPU %u: No C-states info\n"), cpu); return; } - /* printf("Cstates: %d\n", cstates); */ printf(_("active state: C0\n")); printf(_("max_cstate: C%u\n"), cstates-1); diff --git a/tools/power/cpupower/utils/cpuidle-set.c b/tools/power/cpupower/utils/cpuidle-set.c new file mode 100644 index 00000000000..d45d8d775c0 --- /dev/null +++ b/tools/power/cpupower/utils/cpuidle-set.c @@ -0,0 +1,181 @@ +#include <unistd.h> +#include <stdio.h> +#include <errno.h> +#include <stdlib.h> +#include <limits.h> +#include <string.h> +#include <ctype.h> + +#include <getopt.h> + +#include "cpufreq.h" +#include "helpers/helpers.h" +#include "helpers/sysfs.h" + +static struct option info_opts[] = { + { .name = "disable", + .has_arg = required_argument, .flag = NULL, .val = 'd'}, + { .name = "enable", + .has_arg = required_argument, .flag = NULL, .val = 'e'}, + { .name = "disable-by-latency", + .has_arg = required_argument, .flag = NULL, .val = 'D'}, + { .name = "enable-all", + .has_arg = no_argument, .flag = NULL, .val = 'E'}, + { }, +}; + + +int cmd_idle_set(int argc, char **argv) +{ + extern char *optarg; + extern int optind, opterr, optopt; + int ret = 0, cont = 1, param = 0, disabled; + unsigned long long latency = 0, state_latency; + unsigned int cpu = 0, idlestate = 0, idlestates = 0; + char *endptr; + + do { + ret = getopt_long(argc, argv, "d:e:ED:", info_opts, NULL); + if (ret == -1) + break; + switch (ret) { + case '?': + param = '?'; + cont = 0; + break; + case 'd': + if (param) { + param = -1; + cont = 0; + break; + } + param = ret; + idlestate = atoi(optarg); + break; + case 'e': + if (param) { + param = -1; + cont = 0; + break; + } + param = ret; + idlestate = atoi(optarg); + break; + case 'D': + if (param) { + param = -1; + cont = 0; + break; + } + param = ret; + latency = strtoull(optarg, &endptr, 10); + if (*endptr != '\0') { + printf(_("Bad latency value: %s\n"), optarg); + exit(EXIT_FAILURE); + } + break; + case 'E': + if (param) { + param = -1; + cont = 0; + break; + } + param = ret; + break; + case -1: + cont = 0; + break; + } + } while (cont); + + switch (param) { + case -1: + printf(_("You can't specify more than one " + "output-specific argument\n")); + exit(EXIT_FAILURE); + case '?': + printf(_("invalid or unknown argument\n")); + exit(EXIT_FAILURE); + } + + /* Default is: set all CPUs */ + if (bitmask_isallclear(cpus_chosen)) + bitmask_setall(cpus_chosen); + + for (cpu = bitmask_first(cpus_chosen); + cpu <= bitmask_last(cpus_chosen); cpu++) { + + if (!bitmask_isbitset(cpus_chosen, cpu)) + continue; + + if (sysfs_is_cpu_online(cpu) != 1) + continue; + + idlestates = sysfs_get_idlestate_count(cpu); + if (idlestates <= 0) + continue; + + switch (param) { + case 'd': + ret = sysfs_idlestate_disable(cpu, idlestate, 1); + if (ret == 0) + printf(_("Idlestate %u disabled on CPU %u\n"), idlestate, cpu); + else if (ret == -1) + printf(_("Idlestate %u not available on CPU %u\n"), + idlestate, cpu); + else if (ret == -2) + printf(_("Idlestate disabling not supported by kernel\n")); + else + printf(_("Idlestate %u not disabled on CPU %u\n"), + idlestate, cpu); + break; + case 'e': + ret = sysfs_idlestate_disable(cpu, idlestate, 0); + if (ret == 0) + printf(_("Idlestate %u enabled on CPU %u\n"), idlestate, cpu); + else if (ret == -1) + printf(_("Idlestate %u not available on CPU %u\n"), + idlestate, cpu); + else if (ret == -2) + printf(_("Idlestate enabling not supported by kernel\n")); + else + printf(_("Idlestate %u not enabled on CPU %u\n"), + idlestate, cpu); + break; + case 'D': + for (idlestate = 0; idlestate < idlestates; idlestate++) { + disabled = sysfs_is_idlestate_disabled + (cpu, idlestate); + state_latency = sysfs_get_idlestate_latency + (cpu, idlestate); + printf("CPU: %u - idlestate %u - state_latency: %llu - latency: %llu\n", + cpu, idlestate, state_latency, latency); + if (disabled == 1 || latency > state_latency) + continue; + ret = sysfs_idlestate_disable + (cpu, idlestate, 1); + if (ret == 0) + printf(_("Idlestate %u disabled on CPU %u\n"), idlestate, cpu); + } + break; + case 'E': + for (idlestate = 0; idlestate < idlestates; idlestate++) { + disabled = sysfs_is_idlestate_disabled + (cpu, idlestate); + if (disabled == 1) { + ret = sysfs_idlestate_disable + (cpu, idlestate, 0); + if (ret == 0) + printf(_("Idlestate %u enabled on CPU %u\n"), idlestate, cpu); + } + } + break; + default: + /* Not reachable with proper args checking */ + printf(_("Invalid or unknown argument\n")); + exit(EXIT_FAILURE); + break; + } + } + return EXIT_SUCCESS; +} diff --git a/tools/power/cpupower/utils/cpupower-info.c b/tools/power/cpupower/utils/cpupower-info.c index 3f68632c28c..136d979e958 100644 --- a/tools/power/cpupower/utils/cpupower-info.c +++ b/tools/power/cpupower/utils/cpupower-info.c @@ -18,8 +18,6 @@ static struct option set_opts[] = { { .name = "perf-bias", .has_arg = optional_argument, .flag = NULL, .val = 'b'}, - { .name = "sched-mc", .has_arg = optional_argument, .flag = NULL, .val = 'm'}, - { .name = "sched-smt", .has_arg = optional_argument, .flag = NULL, .val = 's'}, { }, }; @@ -37,8 +35,6 @@ int cmd_info(int argc, char **argv) union { struct { - int sched_mc:1; - int sched_smt:1; int perf_bias:1; }; int params; @@ -49,23 +45,13 @@ int cmd_info(int argc, char **argv) textdomain(PACKAGE); /* parameter parsing */ - while ((ret = getopt_long(argc, argv, "msb", set_opts, NULL)) != -1) { + while ((ret = getopt_long(argc, argv, "b", set_opts, NULL)) != -1) { switch (ret) { case 'b': if (params.perf_bias) print_wrong_arg_exit(); params.perf_bias = 1; break; - case 'm': - if (params.sched_mc) - print_wrong_arg_exit(); - params.sched_mc = 1; - break; - case 's': - if (params.sched_smt) - print_wrong_arg_exit(); - params.sched_smt = 1; - break; default: print_wrong_arg_exit(); } @@ -78,25 +64,6 @@ int cmd_info(int argc, char **argv) if (bitmask_isallclear(cpus_chosen)) bitmask_setbit(cpus_chosen, 0); - if (params.sched_mc) { - ret = sysfs_get_sched("mc"); - printf(_("System's multi core scheduler setting: ")); - if (ret < 0) - /* if sysfs file is missing it's: errno == ENOENT */ - printf(_("not supported\n")); - else - printf("%d\n", ret); - } - if (params.sched_smt) { - ret = sysfs_get_sched("smt"); - printf(_("System's thread sibling scheduler setting: ")); - if (ret < 0) - /* if sysfs file is missing it's: errno == ENOENT */ - printf(_("not supported\n")); - else - printf("%d\n", ret); - } - /* Add more per cpu options here */ if (!params.perf_bias) return ret; @@ -125,11 +92,12 @@ int cmd_info(int argc, char **argv) if (params.perf_bias) { ret = msr_intel_get_perf_bias(cpu); if (ret < 0) { - printf(_("Could not read perf-bias value\n")); - break; + fprintf(stderr, + _("Could not read perf-bias value[%d]\n"), ret); + exit(EXIT_FAILURE); } else printf(_("perf-bias: %d\n"), ret); } } - return ret; + return 0; } diff --git a/tools/power/cpupower/utils/cpupower-set.c b/tools/power/cpupower/utils/cpupower-set.c index dc4de376211..573c75f8e3f 100644 --- a/tools/power/cpupower/utils/cpupower-set.c +++ b/tools/power/cpupower/utils/cpupower-set.c @@ -18,9 +18,7 @@ #include "helpers/bitmask.h" static struct option set_opts[] = { - { .name = "perf-bias", .has_arg = optional_argument, .flag = NULL, .val = 'b'}, - { .name = "sched-mc", .has_arg = optional_argument, .flag = NULL, .val = 'm'}, - { .name = "sched-smt", .has_arg = optional_argument, .flag = NULL, .val = 's'}, + { .name = "perf-bias", .has_arg = required_argument, .flag = NULL, .val = 'b'}, { }, }; @@ -38,13 +36,11 @@ int cmd_set(int argc, char **argv) union { struct { - int sched_mc:1; - int sched_smt:1; int perf_bias:1; }; int params; } params; - int sched_mc = 0, sched_smt = 0, perf_bias = 0; + int perf_bias = 0; int ret = 0; setlocale(LC_ALL, ""); @@ -52,7 +48,7 @@ int cmd_set(int argc, char **argv) params.params = 0; /* parameter parsing */ - while ((ret = getopt_long(argc, argv, "m:s:b:", + while ((ret = getopt_long(argc, argv, "b:", set_opts, NULL)) != -1) { switch (ret) { case 'b': @@ -66,28 +62,6 @@ int cmd_set(int argc, char **argv) } params.perf_bias = 1; break; - case 'm': - if (params.sched_mc) - print_wrong_arg_exit(); - sched_mc = atoi(optarg); - if (sched_mc < 0 || sched_mc > 2) { - printf(_("--sched-mc param out " - "of range [0-%d]\n"), 2); - print_wrong_arg_exit(); - } - params.sched_mc = 1; - break; - case 's': - if (params.sched_smt) - print_wrong_arg_exit(); - sched_smt = atoi(optarg); - if (sched_smt < 0 || sched_smt > 2) { - printf(_("--sched-smt param out " - "of range [0-%d]\n"), 2); - print_wrong_arg_exit(); - } - params.sched_smt = 1; - break; default: print_wrong_arg_exit(); } @@ -96,19 +70,6 @@ int cmd_set(int argc, char **argv) if (!params.params) print_wrong_arg_exit(); - if (params.sched_mc) { - ret = sysfs_set_sched("mc", sched_mc); - if (ret) - fprintf(stderr, _("Error setting sched-mc %s\n"), - (ret == -ENODEV) ? "not supported" : ""); - } - if (params.sched_smt) { - ret = sysfs_set_sched("smt", sched_smt); - if (ret) - fprintf(stderr, _("Error setting sched-smt %s\n"), - (ret == -ENODEV) ? "not supported" : ""); - } - /* Default is: set all CPUs */ if (bitmask_isallclear(cpus_chosen)) bitmask_setall(cpus_chosen); diff --git a/tools/power/cpupower/utils/cpupower.c b/tools/power/cpupower/utils/cpupower.c index 52bee591c1c..7cdcf88659c 100644 --- a/tools/power/cpupower/utils/cpupower.c +++ b/tools/power/cpupower/utils/cpupower.c @@ -12,17 +12,14 @@ #include <string.h> #include <unistd.h> #include <errno.h> +#include <sys/types.h> +#include <sys/stat.h> +#include <sys/utsname.h> #include "builtin.h" #include "helpers/helpers.h" #include "helpers/bitmask.h" -struct cmd_struct { - const char *cmd; - int (*main)(int, const char **); - int needs_root; -}; - #define ARRAY_SIZE(x) (sizeof(x)/sizeof(x[0])) static int cmd_help(int argc, const char **argv); @@ -43,10 +40,17 @@ int be_verbose; static void print_help(void); +struct cmd_struct { + const char *cmd; + int (*main)(int, const char **); + int needs_root; +}; + static struct cmd_struct commands[] = { { "frequency-info", cmd_freq_info, 0 }, { "frequency-set", cmd_freq_set, 1 }, { "idle-info", cmd_idle_info, 0 }, + { "idle-set", cmd_idle_set, 1 }, { "set", cmd_set, 1 }, { "info", cmd_info, 0 }, { "monitor", cmd_monitor, 0 }, @@ -168,6 +172,8 @@ int main(int argc, const char *argv[]) { const char *cmd; unsigned int i, ret; + struct stat statbuf; + struct utsname uts; cpus_chosen = bitmask_alloc(sysconf(_SC_NPROCESSORS_CONF)); @@ -194,6 +200,15 @@ int main(int argc, const char *argv[]) get_cpu_info(0, &cpupower_cpu_info); run_as_root = !getuid(); + if (run_as_root) { + ret = uname(&uts); + if (!ret && !strcmp(uts.machine, "x86_64") && + stat("/dev/cpu/0/msr", &statbuf) != 0) { + if (system("modprobe msr") == -1) + fprintf(stderr, _("MSR access not available.\n")); + } + } + for (i = 0; i < ARRAY_SIZE(commands); i++) { struct cmd_struct *p = commands + i; diff --git a/tools/power/cpupower/utils/helpers/amd.c b/tools/power/cpupower/utils/helpers/amd.c index 87d5605bdda..6437ef39aee 100644 --- a/tools/power/cpupower/utils/helpers/amd.c +++ b/tools/power/cpupower/utils/helpers/amd.c @@ -112,14 +112,12 @@ int decode_pstates(unsigned int cpu, unsigned int cpu_family, int amd_pci_get_num_boost_states(int *active, int *states) { struct pci_access *pci_acc; - int vendor_id = 0x1022; - int boost_dev_ids[4] = {0x1204, 0x1604, 0x1704, 0}; struct pci_dev *device; uint8_t val = 0; *active = *states = 0; - device = pci_acc_init(&pci_acc, vendor_id, boost_dev_ids); + device = pci_slot_func_init(&pci_acc, 0x18, 4); if (device == NULL) return -ENODEV; diff --git a/tools/power/cpupower/utils/helpers/cpuid.c b/tools/power/cpupower/utils/helpers/cpuid.c index 906895d21cc..93b0aa74ca0 100644 --- a/tools/power/cpupower/utils/helpers/cpuid.c +++ b/tools/power/cpupower/utils/helpers/cpuid.c @@ -158,6 +158,8 @@ out: cpu_info->caps |= CPUPOWER_CAP_HAS_TURBO_RATIO; case 0x2A: /* SNB */ case 0x2D: /* SNB Xeon */ + case 0x3A: /* IVB */ + case 0x3E: /* IVB Xeon */ cpu_info->caps |= CPUPOWER_CAP_HAS_TURBO_RATIO; cpu_info->caps |= CPUPOWER_CAP_IS_SNB; break; diff --git a/tools/power/cpupower/utils/helpers/helpers.h b/tools/power/cpupower/utils/helpers/helpers.h index 2747e738efb..aa9e95486a2 100644 --- a/tools/power/cpupower/utils/helpers/helpers.h +++ b/tools/power/cpupower/utils/helpers/helpers.h @@ -66,8 +66,8 @@ enum cpupower_cpu_vendor {X86_VENDOR_UNKNOWN = 0, X86_VENDOR_INTEL, #define CPUPOWER_CAP_AMD_CBP 0x00000004 #define CPUPOWER_CAP_PERF_BIAS 0x00000008 #define CPUPOWER_CAP_HAS_TURBO_RATIO 0x00000010 -#define CPUPOWER_CAP_IS_SNB 0x00000011 -#define CPUPOWER_CAP_INTEL_IDA 0x00000012 +#define CPUPOWER_CAP_IS_SNB 0x00000020 +#define CPUPOWER_CAP_INTEL_IDA 0x00000040 #define MAX_HW_PSTATES 10 @@ -92,6 +92,14 @@ extern int get_cpu_info(unsigned int cpu, struct cpupower_cpu_info *cpu_info); extern struct cpupower_cpu_info cpupower_cpu_info; /* cpuid and cpuinfo helpers **************************/ +struct cpuid_core_info { + int pkg; + int core; + int cpu; + + /* flags */ + unsigned int is_online:1; +}; /* CPU topology/hierarchy parsing ******************/ struct cpupower_topology { @@ -101,18 +109,12 @@ struct cpupower_topology { unsigned int threads; /* per core */ /* Array gets mallocated with cores entries, holding per core info */ - struct { - int pkg; - int core; - int cpu; - - /* flags */ - unsigned int is_online:1; - } *core_info; + struct cpuid_core_info *core_info; }; extern int get_cpu_topology(struct cpupower_topology *cpu_top); extern void cpu_topology_release(struct cpupower_topology cpu_top); + /* CPU topology/hierarchy parsing ******************/ /* X86 ONLY ****************************************/ @@ -132,8 +134,11 @@ extern unsigned long long msr_intel_get_turbo_ratio(unsigned int cpu); /* PCI stuff ****************************/ extern int amd_pci_get_num_boost_states(int *active, int *states); -extern struct pci_dev *pci_acc_init(struct pci_access **pacc, int vendor_id, - int *dev_ids); +extern struct pci_dev *pci_acc_init(struct pci_access **pacc, int domain, + int bus, int slot, int func, int vendor, + int dev); +extern struct pci_dev *pci_slot_func_init(struct pci_access **pacc, + int slot, int func); /* PCI stuff ****************************/ diff --git a/tools/power/cpupower/utils/helpers/pci.c b/tools/power/cpupower/utils/helpers/pci.c index cd2eb6fe41c..9690798e644 100644 --- a/tools/power/cpupower/utils/helpers/pci.c +++ b/tools/power/cpupower/utils/helpers/pci.c @@ -10,19 +10,24 @@ * **pacc : if a valid pci_dev is returned * *pacc must be passed to pci_acc_cleanup to free it * - * vendor_id : the pci vendor id matching the pci device to access - * dev_ids : device ids matching the pci device to access + * domain: domain + * bus: bus + * slot: slot + * func: func + * vendor: vendor + * device: device + * Pass -1 for one of the six above to match any * * Returns : * struct pci_dev which can be used with pci_{read,write}_* functions * to access the PCI config space of matching pci devices */ -struct pci_dev *pci_acc_init(struct pci_access **pacc, int vendor_id, - int *dev_ids) +struct pci_dev *pci_acc_init(struct pci_access **pacc, int domain, int bus, + int slot, int func, int vendor, int dev) { - struct pci_filter filter_nb_link = { -1, -1, -1, -1, vendor_id, 0}; + struct pci_filter filter_nb_link = { domain, bus, slot, func, + vendor, dev }; struct pci_dev *device; - unsigned int i; *pacc = pci_alloc(); if (*pacc == NULL) @@ -31,14 +36,20 @@ struct pci_dev *pci_acc_init(struct pci_access **pacc, int vendor_id, pci_init(*pacc); pci_scan_bus(*pacc); - for (i = 0; dev_ids[i] != 0; i++) { - filter_nb_link.device = dev_ids[i]; - for (device = (*pacc)->devices; device; device = device->next) { - if (pci_filter_match(&filter_nb_link, device)) - return device; - } + for (device = (*pacc)->devices; device; device = device->next) { + if (pci_filter_match(&filter_nb_link, device)) + return device; } pci_cleanup(*pacc); return NULL; } + +/* Typically one wants to get a specific slot(device)/func of the root domain + and bus */ +struct pci_dev *pci_slot_func_init(struct pci_access **pacc, int slot, + int func) +{ + return pci_acc_init(pacc, 0, 0, slot, func, -1, -1); +} + #endif /* defined(__i386__) || defined(__x86_64__) */ diff --git a/tools/power/cpupower/utils/helpers/sysfs.c b/tools/power/cpupower/utils/helpers/sysfs.c index c6343024a61..851c7a16ca4 100644 --- a/tools/power/cpupower/utils/helpers/sysfs.c +++ b/tools/power/cpupower/utils/helpers/sysfs.c @@ -37,25 +37,6 @@ unsigned int sysfs_read_file(const char *path, char *buf, size_t buflen) return (unsigned int) numread; } -static unsigned int sysfs_write_file(const char *path, - const char *value, size_t len) -{ - int fd; - ssize_t numwrite; - - fd = open(path, O_WRONLY); - if (fd == -1) - return 0; - - numwrite = write(fd, value, len); - if (numwrite < 1) { - close(fd); - return 0; - } - close(fd); - return (unsigned int) numwrite; -} - /* * Detect whether a CPU is online * @@ -108,6 +89,33 @@ int sysfs_is_cpu_online(unsigned int cpu) /* CPUidle idlestate specific /sys/devices/system/cpu/cpuX/cpuidle/ access */ + +/* CPUidle idlestate specific /sys/devices/system/cpu/cpuX/cpuidle/ access */ + +/* + * helper function to check whether a file under "../cpuX/cpuidle/stateX/" dir + * exists. + * For example the functionality to disable c-states was introduced in later + * kernel versions, this function can be used to explicitly check for this + * feature. + * + * returns 1 if the file exists, 0 otherwise. + */ +unsigned int sysfs_idlestate_file_exists(unsigned int cpu, + unsigned int idlestate, + const char *fname) +{ + char path[SYSFS_PATH_MAX]; + struct stat statbuf; + + + snprintf(path, sizeof(path), PATH_TO_CPU "cpu%u/cpuidle/state%u/%s", + cpu, idlestate, fname); + if (stat(path, &statbuf) != 0) + return 0; + return 1; +} + /* * helper function to read file from /sys into given buffer * fname is a relative path under "cpuX/cpuidle/stateX/" dir @@ -140,6 +148,40 @@ unsigned int sysfs_idlestate_read_file(unsigned int cpu, unsigned int idlestate, return (unsigned int) numread; } +/* + * helper function to write a new value to a /sys file + * fname is a relative path under "../cpuX/cpuidle/cstateY/" dir + * + * Returns the number of bytes written or 0 on error + */ +static +unsigned int sysfs_idlestate_write_file(unsigned int cpu, + unsigned int idlestate, + const char *fname, + const char *value, size_t len) +{ + char path[SYSFS_PATH_MAX]; + int fd; + ssize_t numwrite; + + snprintf(path, sizeof(path), PATH_TO_CPU "cpu%u/cpuidle/state%u/%s", + cpu, idlestate, fname); + + fd = open(path, O_WRONLY); + if (fd == -1) + return 0; + + numwrite = write(fd, value, len); + if (numwrite < 1) { + close(fd); + return 0; + } + + close(fd); + + return (unsigned int) numwrite; +} + /* read access to files which contain one numeric value */ enum idlestate_value { @@ -147,6 +189,7 @@ enum idlestate_value { IDLESTATE_POWER, IDLESTATE_LATENCY, IDLESTATE_TIME, + IDLESTATE_DISABLE, MAX_IDLESTATE_VALUE_FILES }; @@ -155,6 +198,7 @@ static const char *idlestate_value_files[MAX_IDLESTATE_VALUE_FILES] = { [IDLESTATE_POWER] = "power", [IDLESTATE_LATENCY] = "latency", [IDLESTATE_TIME] = "time", + [IDLESTATE_DISABLE] = "disable", }; static unsigned long long sysfs_idlestate_get_one_value(unsigned int cpu, @@ -224,8 +268,59 @@ static char *sysfs_idlestate_get_one_string(unsigned int cpu, return result; } +/* + * Returns: + * 1 if disabled + * 0 if enabled + * -1 if idlestate is not available + * -2 if disabling is not supported by the kernel + */ +int sysfs_is_idlestate_disabled(unsigned int cpu, + unsigned int idlestate) +{ + if (sysfs_get_idlestate_count(cpu) <= idlestate) + return -1; + + if (!sysfs_idlestate_file_exists(cpu, idlestate, + idlestate_value_files[IDLESTATE_DISABLE])) + return -2; + return sysfs_idlestate_get_one_value(cpu, idlestate, IDLESTATE_DISABLE); +} + +/* + * Pass 1 as last argument to disable or 0 to enable the state + * Returns: + * 0 on success + * negative values on error, for example: + * -1 if idlestate is not available + * -2 if disabling is not supported by the kernel + * -3 No write access to disable/enable C-states + */ +int sysfs_idlestate_disable(unsigned int cpu, + unsigned int idlestate, + unsigned int disable) +{ + char value[SYSFS_PATH_MAX]; + int bytes_written; + + if (sysfs_get_idlestate_count(cpu) <= idlestate) + return -1; + + if (!sysfs_idlestate_file_exists(cpu, idlestate, + idlestate_value_files[IDLESTATE_DISABLE])) + return -2; + + snprintf(value, SYSFS_PATH_MAX, "%u", disable); + + bytes_written = sysfs_idlestate_write_file(cpu, idlestate, "disable", + value, sizeof(disable)); + if (bytes_written) + return 0; + return -3; +} + unsigned long sysfs_get_idlestate_latency(unsigned int cpu, - unsigned int idlestate) + unsigned int idlestate) { return sysfs_idlestate_get_one_value(cpu, idlestate, IDLESTATE_LATENCY); } @@ -257,7 +352,7 @@ char *sysfs_get_idlestate_desc(unsigned int cpu, unsigned int idlestate) * Negativ in error case * Zero if cpuidle does not export any C-states */ -int sysfs_get_idlestate_count(unsigned int cpu) +unsigned int sysfs_get_idlestate_count(unsigned int cpu) { char file[SYSFS_PATH_MAX]; struct stat statbuf; @@ -362,22 +457,7 @@ char *sysfs_get_cpuidle_driver(void) */ int sysfs_get_sched(const char *smt_mc) { - unsigned long value; - char linebuf[MAX_LINE_LEN]; - char *endp; - char path[SYSFS_PATH_MAX]; - - if (strcmp("mc", smt_mc) && strcmp("smt", smt_mc)) - return -EINVAL; - - snprintf(path, sizeof(path), - PATH_TO_CPU "sched_%s_power_savings", smt_mc); - if (sysfs_read_file(path, linebuf, MAX_LINE_LEN) == 0) - return -1; - value = strtoul(linebuf, &endp, 0); - if (endp == linebuf || errno == ERANGE) - return -1; - return value; + return -ENODEV; } /* @@ -388,21 +468,5 @@ int sysfs_get_sched(const char *smt_mc) */ int sysfs_set_sched(const char *smt_mc, int val) { - char linebuf[MAX_LINE_LEN]; - char path[SYSFS_PATH_MAX]; - struct stat statbuf; - - if (strcmp("mc", smt_mc) && strcmp("smt", smt_mc)) - return -EINVAL; - - snprintf(path, sizeof(path), - PATH_TO_CPU "sched_%s_power_savings", smt_mc); - sprintf(linebuf, "%d", val); - - if (stat(path, &statbuf) != 0) - return -ENODEV; - - if (sysfs_write_file(path, linebuf, MAX_LINE_LEN) == 0) - return -1; - return 0; + return -ENODEV; } diff --git a/tools/power/cpupower/utils/helpers/sysfs.h b/tools/power/cpupower/utils/helpers/sysfs.h index 8cb797bbceb..d28f11fedbd 100644 --- a/tools/power/cpupower/utils/helpers/sysfs.h +++ b/tools/power/cpupower/utils/helpers/sysfs.h @@ -7,8 +7,16 @@ extern unsigned int sysfs_read_file(const char *path, char *buf, size_t buflen); +extern unsigned int sysfs_idlestate_file_exists(unsigned int cpu, + unsigned int idlestate, + const char *fname); + extern int sysfs_is_cpu_online(unsigned int cpu); +extern int sysfs_is_idlestate_disabled(unsigned int cpu, + unsigned int idlestate); +extern int sysfs_idlestate_disable(unsigned int cpu, unsigned int idlestate, + unsigned int disable); extern unsigned long sysfs_get_idlestate_latency(unsigned int cpu, unsigned int idlestate); extern unsigned long sysfs_get_idlestate_usage(unsigned int cpu, @@ -19,7 +27,7 @@ extern char *sysfs_get_idlestate_name(unsigned int cpu, unsigned int idlestate); extern char *sysfs_get_idlestate_desc(unsigned int cpu, unsigned int idlestate); -extern int sysfs_get_idlestate_count(unsigned int cpu); +extern unsigned int sysfs_get_idlestate_count(unsigned int cpu); extern char *sysfs_get_cpuidle_governor(void); extern char *sysfs_get_cpuidle_driver(void); diff --git a/tools/power/cpupower/utils/helpers/topology.c b/tools/power/cpupower/utils/helpers/topology.c index 4eae2c47ba4..c13120af519 100644 --- a/tools/power/cpupower/utils/helpers/topology.c +++ b/tools/power/cpupower/utils/helpers/topology.c @@ -20,9 +20,8 @@ #include <helpers/sysfs.h> /* returns -1 on failure, 0 on success */ -int sysfs_topology_read_file(unsigned int cpu, const char *fname) +static int sysfs_topology_read_file(unsigned int cpu, const char *fname, int *result) { - unsigned long value; char linebuf[MAX_LINE_LEN]; char *endp; char path[SYSFS_PATH_MAX]; @@ -31,20 +30,12 @@ int sysfs_topology_read_file(unsigned int cpu, const char *fname) cpu, fname); if (sysfs_read_file(path, linebuf, MAX_LINE_LEN) == 0) return -1; - value = strtoul(linebuf, &endp, 0); + *result = strtol(linebuf, &endp, 0); if (endp == linebuf || errno == ERANGE) return -1; - return value; + return 0; } -struct cpuid_core_info { - unsigned int pkg; - unsigned int thread; - unsigned int cpu; - /* flags */ - unsigned int is_online:1; -}; - static int __compare(const void *t1, const void *t2) { struct cpuid_core_info *top1 = (struct cpuid_core_info *)t1; @@ -53,9 +44,9 @@ static int __compare(const void *t1, const void *t2) return -1; else if (top1->pkg > top2->pkg) return 1; - else if (top1->thread < top2->thread) + else if (top1->core < top2->core) return -1; - else if (top1->thread > top2->thread) + else if (top1->core > top2->core) return 1; else if (top1->cpu < top2->cpu) return -1; @@ -73,28 +64,42 @@ static int __compare(const void *t1, const void *t2) */ int get_cpu_topology(struct cpupower_topology *cpu_top) { - int cpu, cpus = sysconf(_SC_NPROCESSORS_CONF); + int cpu, last_pkg, cpus = sysconf(_SC_NPROCESSORS_CONF); - cpu_top->core_info = malloc(sizeof(struct cpupower_topology) * cpus); + cpu_top->core_info = malloc(sizeof(struct cpuid_core_info) * cpus); if (cpu_top->core_info == NULL) return -ENOMEM; cpu_top->pkgs = cpu_top->cores = 0; for (cpu = 0; cpu < cpus; cpu++) { cpu_top->core_info[cpu].cpu = cpu; cpu_top->core_info[cpu].is_online = sysfs_is_cpu_online(cpu); - cpu_top->core_info[cpu].pkg = - sysfs_topology_read_file(cpu, "physical_package_id"); - if ((int)cpu_top->core_info[cpu].pkg != -1 && - cpu_top->core_info[cpu].pkg > cpu_top->pkgs) - cpu_top->pkgs = cpu_top->core_info[cpu].pkg; - cpu_top->core_info[cpu].core = - sysfs_topology_read_file(cpu, "core_id"); + if(sysfs_topology_read_file( + cpu, + "physical_package_id", + &(cpu_top->core_info[cpu].pkg)) < 0) + return -1; + if(sysfs_topology_read_file( + cpu, + "core_id", + &(cpu_top->core_info[cpu].core)) < 0) + return -1; } - cpu_top->pkgs++; qsort(cpu_top->core_info, cpus, sizeof(struct cpuid_core_info), __compare); + /* Count the number of distinct pkgs values. This works + because the primary sort of the core_info struct was just + done by pkg value. */ + last_pkg = cpu_top->core_info[0].pkg; + for(cpu = 1; cpu < cpus; cpu++) { + if(cpu_top->core_info[cpu].pkg != last_pkg) { + last_pkg = cpu_top->core_info[cpu].pkg; + cpu_top->pkgs++; + } + } + cpu_top->pkgs++; + /* Intel's cores count is not consecutively numbered, there may * be a core_id of 3, but none of 2. Assume there always is 0 * Get amount of cores by counting duplicates in a package diff --git a/tools/power/cpupower/utils/idle_monitor/amd_fam14h_idle.c b/tools/power/cpupower/utils/idle_monitor/amd_fam14h_idle.c index 202e555988b..2116df9ad83 100644 --- a/tools/power/cpupower/utils/idle_monitor/amd_fam14h_idle.c +++ b/tools/power/cpupower/utils/idle_monitor/amd_fam14h_idle.c @@ -20,8 +20,6 @@ #include "idle_monitor/cpupower-monitor.h" #include "helpers/helpers.h" -/******** PCI parts could go into own file and get shared ***************/ - #define PCI_NON_PC0_OFFSET 0xb0 #define PCI_PC1_OFFSET 0xb4 #define PCI_PC6_OFFSET 0xb8 @@ -82,10 +80,7 @@ static cstate_t amd_fam14h_cstates[AMD_FAM14H_STATE_NUM] = { }; static struct pci_access *pci_acc; -static int pci_vendor_id = 0x1022; -static int pci_dev_ids[2] = {0x1716, 0}; static struct pci_dev *amd_fam14h_pci_dev; - static int nbp1_entered; struct timespec start_time; @@ -286,13 +281,13 @@ struct cpuidle_monitor *amd_fam14h_register(void) if (cpupower_cpu_info.vendor != X86_VENDOR_AMD) return NULL; - if (cpupower_cpu_info.family == 0x14) { - if (cpu_count <= 0 || cpu_count > 2) { - fprintf(stderr, "AMD fam14h: Invalid cpu count: %d\n", - cpu_count); - return NULL; - } - } else + if (cpupower_cpu_info.family == 0x14) + strncpy(amd_fam14h_monitor.name, "Fam_14h", + MONITOR_NAME_LEN - 1); + else if (cpupower_cpu_info.family == 0x12) + strncpy(amd_fam14h_monitor.name, "Fam_12h", + MONITOR_NAME_LEN - 1); + else return NULL; /* We do not alloc for nbp1 machine wide counter */ @@ -303,7 +298,9 @@ struct cpuidle_monitor *amd_fam14h_register(void) sizeof(unsigned long long)); } - amd_fam14h_pci_dev = pci_acc_init(&pci_acc, pci_vendor_id, pci_dev_ids); + /* We need PCI device: Slot 18, Func 6, compare with BKDG + for fam 12h/14h */ + amd_fam14h_pci_dev = pci_slot_func_init(&pci_acc, 0x18, 6); if (amd_fam14h_pci_dev == NULL || pci_acc == NULL) return NULL; @@ -325,7 +322,7 @@ static void amd_fam14h_unregister(void) } struct cpuidle_monitor amd_fam14h_monitor = { - .name = "Ontario", + .name = "", .hw_states = amd_fam14h_cstates, .hw_states_num = AMD_FAM14H_STATE_NUM, .start = amd_fam14h_start, diff --git a/tools/power/cpupower/utils/idle_monitor/cpupower-monitor.c b/tools/power/cpupower/utils/idle_monitor/cpupower-monitor.c index 0d6571e418d..c4bae9203a6 100644 --- a/tools/power/cpupower/utils/idle_monitor/cpupower-monitor.c +++ b/tools/power/cpupower/utils/idle_monitor/cpupower-monitor.c @@ -39,6 +39,7 @@ static int mode; static int interval = 1; static char *show_monitors_param; static struct cpupower_topology cpu_top; +static unsigned int wake_cpus; /* ToDo: Document this in the manpage */ static char range_abbr[RANGE_MAX] = { 'T', 'C', 'P', 'M', }; @@ -84,7 +85,7 @@ int fill_string_with_spaces(char *s, int n) void print_header(int topology_depth) { int unsigned mon; - int state, need_len, pr_mon_len; + int state, need_len; cstate_t s; char buf[128] = ""; int percent_width = 4; @@ -93,7 +94,6 @@ void print_header(int topology_depth) printf("%s|", buf); for (mon = 0; mon < avail_monitors; mon++) { - pr_mon_len = 0; need_len = monitors[mon]->hw_states_num * (percent_width + 3) - 1; if (mon != 0) { @@ -315,16 +315,28 @@ int fork_it(char **argv) int do_interval_measure(int i) { unsigned int num; + int cpu; + + if (wake_cpus) + for (cpu = 0; cpu < cpu_count; cpu++) + bind_cpu(cpu); for (num = 0; num < avail_monitors; num++) { dprint("HW C-state residency monitor: %s - States: %d\n", monitors[num]->name, monitors[num]->hw_states_num); monitors[num]->start(); } + sleep(i); + + if (wake_cpus) + for (cpu = 0; cpu < cpu_count; cpu++) + bind_cpu(cpu); + for (num = 0; num < avail_monitors; num++) monitors[num]->stop(); + return 0; } @@ -333,7 +345,7 @@ static void cmdline(int argc, char *argv[]) int opt; progname = basename(argv[0]); - while ((opt = getopt(argc, argv, "+li:m:")) != -1) { + while ((opt = getopt(argc, argv, "+lci:m:")) != -1) { switch (opt) { case 'l': if (mode) @@ -352,6 +364,9 @@ static void cmdline(int argc, char *argv[]) mode = show; show_monitors_param = optarg; break; + case 'c': + wake_cpus = 1; + break; default: print_wrong_arg_exit(); } diff --git a/tools/power/cpupower/utils/idle_monitor/cpupower-monitor.h b/tools/power/cpupower/utils/idle_monitor/cpupower-monitor.h index 9312ee1f2db..9e43f3371fb 100644 --- a/tools/power/cpupower/utils/idle_monitor/cpupower-monitor.h +++ b/tools/power/cpupower/utils/idle_monitor/cpupower-monitor.h @@ -65,4 +65,21 @@ extern long long timespec_diff_us(struct timespec start, struct timespec end); "could be inaccurate\n"), mes, ov); \ } + +/* Taken over from x86info project sources -> return 0 on success */ +#include <sched.h> +#include <sys/types.h> +#include <unistd.h> +static inline int bind_cpu(int cpu) +{ + cpu_set_t set; + + if (sched_getaffinity(getpid(), sizeof(set), &set) == 0) { + CPU_ZERO(&set); + CPU_SET(cpu, &set); + return sched_setaffinity(getpid(), sizeof(set), &set); + } + return 1; +} + #endif /* __CPUIDLE_INFO_HW__ */ diff --git a/tools/power/cpupower/utils/idle_monitor/hsw_ext_idle.c b/tools/power/cpupower/utils/idle_monitor/hsw_ext_idle.c new file mode 100644 index 00000000000..ebeaba6571a --- /dev/null +++ b/tools/power/cpupower/utils/idle_monitor/hsw_ext_idle.c @@ -0,0 +1,196 @@ +/* + * (C) 2010,2011 Thomas Renninger <trenn@suse.de>, Novell Inc. + * + * Licensed under the terms of the GNU GPL License version 2. + * + * Based on SandyBridge monitor. Implements the new package C-states + * (PC8, PC9, PC10) coming with a specific Haswell (family 0x45) CPU. + */ + +#if defined(__i386__) || defined(__x86_64__) + +#include <stdio.h> +#include <stdint.h> +#include <stdlib.h> +#include <string.h> + +#include "helpers/helpers.h" +#include "idle_monitor/cpupower-monitor.h" + +#define MSR_PKG_C8_RESIDENCY 0x00000630 +#define MSR_PKG_C9_RESIDENCY 0x00000631 +#define MSR_PKG_C10_RESIDENCY 0x00000632 + +#define MSR_TSC 0x10 + +enum intel_hsw_ext_id { PC8 = 0, PC9, PC10, HSW_EXT_CSTATE_COUNT, + TSC = 0xFFFF }; + +static int hsw_ext_get_count_percent(unsigned int self_id, double *percent, + unsigned int cpu); + +static cstate_t hsw_ext_cstates[HSW_EXT_CSTATE_COUNT] = { + { + .name = "PC8", + .desc = N_("Processor Package C8"), + .id = PC8, + .range = RANGE_PACKAGE, + .get_count_percent = hsw_ext_get_count_percent, + }, + { + .name = "PC9", + .desc = N_("Processor Package C9"), + .desc = N_("Processor Package C2"), + .id = PC9, + .range = RANGE_PACKAGE, + .get_count_percent = hsw_ext_get_count_percent, + }, + { + .name = "PC10", + .desc = N_("Processor Package C10"), + .id = PC10, + .range = RANGE_PACKAGE, + .get_count_percent = hsw_ext_get_count_percent, + }, +}; + +static unsigned long long tsc_at_measure_start; +static unsigned long long tsc_at_measure_end; +static unsigned long long *previous_count[HSW_EXT_CSTATE_COUNT]; +static unsigned long long *current_count[HSW_EXT_CSTATE_COUNT]; +/* valid flag for all CPUs. If a MSR read failed it will be zero */ +static int *is_valid; + +static int hsw_ext_get_count(enum intel_hsw_ext_id id, unsigned long long *val, + unsigned int cpu) +{ + int msr; + + switch (id) { + case PC8: + msr = MSR_PKG_C8_RESIDENCY; + break; + case PC9: + msr = MSR_PKG_C9_RESIDENCY; + break; + case PC10: + msr = MSR_PKG_C10_RESIDENCY; + break; + case TSC: + msr = MSR_TSC; + break; + default: + return -1; + }; + if (read_msr(cpu, msr, val)) + return -1; + return 0; +} + +static int hsw_ext_get_count_percent(unsigned int id, double *percent, + unsigned int cpu) +{ + *percent = 0.0; + + if (!is_valid[cpu]) + return -1; + + *percent = (100.0 * + (current_count[id][cpu] - previous_count[id][cpu])) / + (tsc_at_measure_end - tsc_at_measure_start); + + dprint("%s: previous: %llu - current: %llu - (%u)\n", + hsw_ext_cstates[id].name, previous_count[id][cpu], + current_count[id][cpu], cpu); + + dprint("%s: tsc_diff: %llu - count_diff: %llu - percent: %2.f (%u)\n", + hsw_ext_cstates[id].name, + (unsigned long long) tsc_at_measure_end - tsc_at_measure_start, + current_count[id][cpu] - previous_count[id][cpu], + *percent, cpu); + + return 0; +} + +static int hsw_ext_start(void) +{ + int num, cpu; + unsigned long long val; + + for (num = 0; num < HSW_EXT_CSTATE_COUNT; num++) { + for (cpu = 0; cpu < cpu_count; cpu++) { + hsw_ext_get_count(num, &val, cpu); + previous_count[num][cpu] = val; + } + } + hsw_ext_get_count(TSC, &tsc_at_measure_start, 0); + return 0; +} + +static int hsw_ext_stop(void) +{ + unsigned long long val; + int num, cpu; + + hsw_ext_get_count(TSC, &tsc_at_measure_end, 0); + + for (num = 0; num < HSW_EXT_CSTATE_COUNT; num++) { + for (cpu = 0; cpu < cpu_count; cpu++) { + is_valid[cpu] = !hsw_ext_get_count(num, &val, cpu); + current_count[num][cpu] = val; + } + } + return 0; +} + +struct cpuidle_monitor intel_hsw_ext_monitor; + +static struct cpuidle_monitor *hsw_ext_register(void) +{ + int num; + + if (cpupower_cpu_info.vendor != X86_VENDOR_INTEL + || cpupower_cpu_info.family != 6) + return NULL; + + switch (cpupower_cpu_info.model) { + case 0x45: /* HSW */ + break; + default: + return NULL; + } + + is_valid = calloc(cpu_count, sizeof(int)); + for (num = 0; num < HSW_EXT_CSTATE_COUNT; num++) { + previous_count[num] = calloc(cpu_count, + sizeof(unsigned long long)); + current_count[num] = calloc(cpu_count, + sizeof(unsigned long long)); + } + intel_hsw_ext_monitor.name_len = strlen(intel_hsw_ext_monitor.name); + return &intel_hsw_ext_monitor; +} + +void hsw_ext_unregister(void) +{ + int num; + free(is_valid); + for (num = 0; num < HSW_EXT_CSTATE_COUNT; num++) { + free(previous_count[num]); + free(current_count[num]); + } +} + +struct cpuidle_monitor intel_hsw_ext_monitor = { + .name = "HaswellExtended", + .hw_states = hsw_ext_cstates, + .hw_states_num = HSW_EXT_CSTATE_COUNT, + .start = hsw_ext_start, + .stop = hsw_ext_stop, + .do_register = hsw_ext_register, + .unregister = hsw_ext_unregister, + .needs_root = 1, + .overflow_s = 922000000 /* 922337203 seconds TSC overflow + at 20GHz */ +}; +#endif /* defined(__i386__) || defined(__x86_64__) */ diff --git a/tools/power/cpupower/utils/idle_monitor/idle_monitors.def b/tools/power/cpupower/utils/idle_monitor/idle_monitors.def index e3f8d9b2b18..0d6ba4dbb9c 100644 --- a/tools/power/cpupower/utils/idle_monitor/idle_monitors.def +++ b/tools/power/cpupower/utils/idle_monitor/idle_monitors.def @@ -2,6 +2,7 @@ DEF(amd_fam14h) DEF(intel_nhm) DEF(intel_snb) +DEF(intel_hsw_ext) DEF(mperf) #endif DEF(cpuidle_sysfs) diff --git a/tools/power/cpupower/utils/idle_monitor/snb_idle.c b/tools/power/cpupower/utils/idle_monitor/snb_idle.c index a1bc07cd53e..efc8a69c9ab 100644 --- a/tools/power/cpupower/utils/idle_monitor/snb_idle.c +++ b/tools/power/cpupower/utils/idle_monitor/snb_idle.c @@ -150,9 +150,19 @@ static struct cpuidle_monitor *snb_register(void) || cpupower_cpu_info.family != 6) return NULL; - if (cpupower_cpu_info.model != 0x2A - && cpupower_cpu_info.model != 0x2D) + switch (cpupower_cpu_info.model) { + case 0x2A: /* SNB */ + case 0x2D: /* SNB Xeon */ + case 0x3A: /* IVB */ + case 0x3E: /* IVB Xeon */ + case 0x3C: /* HSW */ + case 0x3F: /* HSW */ + case 0x45: /* HSW */ + case 0x46: /* HSW */ + break; + default: return NULL; + } is_valid = calloc(cpu_count, sizeof(int)); for (num = 0; num < SNB_CSTATE_COUNT; num++) { |