diff options
Diffstat (limited to 'arch/ia64/kernel/perfmon.c')
| -rw-r--r-- | arch/ia64/kernel/perfmon.c | 752 |
1 files changed, 352 insertions, 400 deletions
diff --git a/arch/ia64/kernel/perfmon.c b/arch/ia64/kernel/perfmon.c index dbb28164b19..5845ffea67c 100644 --- a/arch/ia64/kernel/perfmon.c +++ b/arch/ia64/kernel/perfmon.c @@ -23,7 +23,6 @@ #include <linux/kernel.h> #include <linux/sched.h> #include <linux/interrupt.h> -#include <linux/smp_lock.h> #include <linux/proc_fs.h> #include <linux/seq_file.h> #include <linux/init.h> @@ -41,6 +40,9 @@ #include <linux/capability.h> #include <linux/rcupdate.h> #include <linux/completion.h> +#include <linux/tracehook.h> +#include <linux/slab.h> +#include <linux/cpu.h> #include <asm/errno.h> #include <asm/intrinsics.h> @@ -48,7 +50,6 @@ #include <asm/perfmon.h> #include <asm/processor.h> #include <asm/signal.h> -#include <asm/system.h> #include <asm/uaccess.h> #include <asm/delay.h> @@ -147,7 +148,7 @@ * in UP: * - we need to protect against PMU overflow interrupts (local_irq_disable) * - * spin_lock_irqsave()/spin_lock_irqrestore(): + * spin_lock_irqsave()/spin_unlock_irqrestore(): * in SMP: local_irq_disable + spin_lock * in UP : local_irq_disable * @@ -159,14 +160,14 @@ */ #define PROTECT_CTX(c, f) \ do { \ - DPRINT(("spinlock_irq_save ctx %p by [%d]\n", c, current->pid)); \ + DPRINT(("spinlock_irq_save ctx %p by [%d]\n", c, task_pid_nr(current))); \ spin_lock_irqsave(&(c)->ctx_lock, f); \ - DPRINT(("spinlocked ctx %p by [%d]\n", c, current->pid)); \ + DPRINT(("spinlocked ctx %p by [%d]\n", c, task_pid_nr(current))); \ } while(0) #define UNPROTECT_CTX(c, f) \ do { \ - DPRINT(("spinlock_irq_restore ctx %p by [%d]\n", c, current->pid)); \ + DPRINT(("spinlock_irq_restore ctx %p by [%d]\n", c, task_pid_nr(current))); \ spin_unlock_irqrestore(&(c)->ctx_lock, f); \ } while(0) @@ -228,12 +229,12 @@ #ifdef PFM_DEBUGGING #define DPRINT(a) \ do { \ - if (unlikely(pfm_sysctl.debug >0)) { printk("%s.%d: CPU%d [%d] ", __FUNCTION__, __LINE__, smp_processor_id(), current->pid); printk a; } \ + if (unlikely(pfm_sysctl.debug >0)) { printk("%s.%d: CPU%d [%d] ", __func__, __LINE__, smp_processor_id(), task_pid_nr(current)); printk a; } \ } while (0) #define DPRINT_ovfl(a) \ do { \ - if (unlikely(pfm_sysctl.debug > 0 && pfm_sysctl.debug_ovfl >0)) { printk("%s.%d: CPU%d [%d] ", __FUNCTION__, __LINE__, smp_processor_id(), current->pid); printk a; } \ + if (unlikely(pfm_sysctl.debug > 0 && pfm_sysctl.debug_ovfl >0)) { printk("%s.%d: CPU%d [%d] ", __func__, __LINE__, smp_processor_id(), task_pid_nr(current)); printk a; } \ } while (0) #endif @@ -312,7 +313,7 @@ typedef struct pfm_context { unsigned long th_pmcs[PFM_NUM_PMC_REGS]; /* PMC thread save state */ unsigned long th_pmds[PFM_NUM_PMD_REGS]; /* PMD thread save state */ - u64 ctx_saved_psr_up; /* only contains psr.up value */ + unsigned long ctx_saved_psr_up; /* only contains psr.up value */ unsigned long ctx_last_activation; /* context last activation number for last_cpu */ unsigned int ctx_last_cpu; /* CPU id of current or last CPU used (SMP only) */ @@ -520,20 +521,52 @@ static pmu_config_t *pmu_conf; pfm_sysctl_t pfm_sysctl; EXPORT_SYMBOL(pfm_sysctl); -static ctl_table pfm_ctl_table[]={ - {1, "debug", &pfm_sysctl.debug, sizeof(int), 0666, NULL, &proc_dointvec, NULL,}, - {2, "debug_ovfl", &pfm_sysctl.debug_ovfl, sizeof(int), 0666, NULL, &proc_dointvec, NULL,}, - {3, "fastctxsw", &pfm_sysctl.fastctxsw, sizeof(int), 0600, NULL, &proc_dointvec, NULL,}, - {4, "expert_mode", &pfm_sysctl.expert_mode, sizeof(int), 0600, NULL, &proc_dointvec, NULL,}, - { 0, }, +static struct ctl_table pfm_ctl_table[] = { + { + .procname = "debug", + .data = &pfm_sysctl.debug, + .maxlen = sizeof(int), + .mode = 0666, + .proc_handler = proc_dointvec, + }, + { + .procname = "debug_ovfl", + .data = &pfm_sysctl.debug_ovfl, + .maxlen = sizeof(int), + .mode = 0666, + .proc_handler = proc_dointvec, + }, + { + .procname = "fastctxsw", + .data = &pfm_sysctl.fastctxsw, + .maxlen = sizeof(int), + .mode = 0600, + .proc_handler = proc_dointvec, + }, + { + .procname = "expert_mode", + .data = &pfm_sysctl.expert_mode, + .maxlen = sizeof(int), + .mode = 0600, + .proc_handler = proc_dointvec, + }, + {} }; -static ctl_table pfm_sysctl_dir[] = { - {1, "perfmon", NULL, 0, 0755, pfm_ctl_table, }, - {0,}, +static struct ctl_table pfm_sysctl_dir[] = { + { + .procname = "perfmon", + .mode = 0555, + .child = pfm_ctl_table, + }, + {} }; -static ctl_table pfm_sysctl_root[] = { - {1, "kernel", NULL, 0, 0755, pfm_sysctl_dir, }, - {0,}, +static struct ctl_table pfm_sysctl_root[] = { + { + .procname = "kernel", + .mode = 0555, + .child = pfm_sysctl_dir, + }, + {} }; static struct ctl_table_header *pfm_sysctl_header; @@ -549,21 +582,6 @@ pfm_put_task(struct task_struct *task) } static inline void -pfm_set_task_notify(struct task_struct *task) -{ - struct thread_info *info; - - info = (struct thread_info *) ((char *) task + IA64_TASK_SIZE); - set_bit(TIF_NOTIFY_RESUME, &info->flags); -} - -static inline void -pfm_clear_task_notify(void) -{ - clear_thread_flag(TIF_NOTIFY_RESUME); -} - -static inline void pfm_reserve_page(unsigned long a) { SetPageReserved(vmalloc_to_page((void *)a)); @@ -587,31 +605,22 @@ pfm_unprotect_ctx_ctxsw(pfm_context_t *x, unsigned long f) spin_unlock(&(x)->ctx_lock); } -static inline unsigned int -pfm_do_munmap(struct mm_struct *mm, unsigned long addr, size_t len, int acct) -{ - return do_munmap(mm, addr, len); -} - -static inline unsigned long -pfm_get_unmapped_area(struct file *file, unsigned long addr, unsigned long len, unsigned long pgoff, unsigned long flags, unsigned long exec) -{ - return get_unmapped_area(file, addr, len, pgoff, flags); -} - +/* forward declaration */ +static const struct dentry_operations pfmfs_dentry_operations; -static int -pfmfs_get_sb(struct file_system_type *fs_type, int flags, const char *dev_name, void *data, - struct vfsmount *mnt) +static struct dentry * +pfmfs_mount(struct file_system_type *fs_type, int flags, const char *dev_name, void *data) { - return get_sb_pseudo(fs_type, "pfm:", NULL, PFMFS_MAGIC, mnt); + return mount_pseudo(fs_type, "pfm:", NULL, &pfmfs_dentry_operations, + PFMFS_MAGIC); } static struct file_system_type pfm_fs_type = { .name = "pfmfs", - .get_sb = pfmfs_get_sb, + .mount = pfmfs_mount, .kill_sb = kill_anon_super, }; +MODULE_ALIAS_FS("pfmfs"); DEFINE_PER_CPU(unsigned long, pfm_syst_info); DEFINE_PER_CPU(struct task_struct *, pmu_owner); @@ -621,7 +630,7 @@ EXPORT_PER_CPU_SYMBOL_GPL(pfm_syst_info); /* forward declaration */ -static struct file_operations pfm_file_ops; +static const struct file_operations pfm_file_ops; /* * forward declarations @@ -812,10 +821,9 @@ pfm_rvmalloc(unsigned long size) unsigned long addr; size = PAGE_ALIGN(size); - mem = vmalloc(size); + mem = vzalloc(size); if (mem) { //printk("perfmon: CPU%d pfm_rvmalloc(%ld)=%p\n", smp_processor_id(), size, mem); - memset(mem, 0, size); addr = (unsigned long)mem; while (size > 0) { pfm_reserve_page(addr); @@ -845,7 +853,7 @@ pfm_rvfree(void *mem, unsigned long size) } static pfm_context_t * -pfm_context_alloc(void) +pfm_context_alloc(int ctx_flags) { pfm_context_t *ctx; @@ -856,6 +864,46 @@ pfm_context_alloc(void) ctx = kzalloc(sizeof(pfm_context_t), GFP_KERNEL); if (ctx) { DPRINT(("alloc ctx @%p\n", ctx)); + + /* + * init context protection lock + */ + spin_lock_init(&ctx->ctx_lock); + + /* + * context is unloaded + */ + ctx->ctx_state = PFM_CTX_UNLOADED; + + /* + * initialization of context's flags + */ + ctx->ctx_fl_block = (ctx_flags & PFM_FL_NOTIFY_BLOCK) ? 1 : 0; + ctx->ctx_fl_system = (ctx_flags & PFM_FL_SYSTEM_WIDE) ? 1: 0; + ctx->ctx_fl_no_msg = (ctx_flags & PFM_FL_OVFL_NO_MSG) ? 1: 0; + /* + * will move to set properties + * ctx->ctx_fl_excl_idle = (ctx_flags & PFM_FL_EXCL_IDLE) ? 1: 0; + */ + + /* + * init restart semaphore to locked + */ + init_completion(&ctx->ctx_restart_done); + + /* + * activation is used in SMP only + */ + ctx->ctx_last_activation = PFM_INVALID_ACTIVATION; + SET_LAST_CPU(ctx, -1); + + /* + * initialize notification message queue + */ + ctx->ctx_msgq_head = ctx->ctx_msgq_tail = 0; + init_waitqueue_head(&ctx->ctx_msgq_wait); + init_waitqueue_head(&ctx->ctx_zombieq); + } return ctx; } @@ -876,7 +924,7 @@ pfm_mask_monitoring(struct task_struct *task) unsigned long mask, val, ovfl_mask; int i; - DPRINT_ovfl(("masking monitoring for [%d]\n", task->pid)); + DPRINT_ovfl(("masking monitoring for [%d]\n", task_pid_nr(task))); ovfl_mask = pmu_conf->ovfl_val; /* @@ -955,12 +1003,12 @@ pfm_restore_monitoring(struct task_struct *task) ovfl_mask = pmu_conf->ovfl_val; if (task != current) { - printk(KERN_ERR "perfmon.%d: invalid task[%d] current[%d]\n", __LINE__, task->pid, current->pid); + printk(KERN_ERR "perfmon.%d: invalid task[%d] current[%d]\n", __LINE__, task_pid_nr(task), task_pid_nr(current)); return; } if (ctx->ctx_state != PFM_CTX_MASKED) { printk(KERN_ERR "perfmon.%d: task[%d] current[%d] invalid state=%d\n", __LINE__, - task->pid, current->pid, ctx->ctx_state); + task_pid_nr(task), task_pid_nr(current), ctx->ctx_state); return; } psr = pfm_get_psr(); @@ -1014,7 +1062,8 @@ pfm_restore_monitoring(struct task_struct *task) if ((mask & 0x1) == 0UL) continue; ctx->th_pmcs[i] = ctx->ctx_pmcs[i]; ia64_set_pmc(i, ctx->th_pmcs[i]); - DPRINT(("[%d] pmc[%d]=0x%lx\n", task->pid, i, ctx->th_pmcs[i])); + DPRINT(("[%d] pmc[%d]=0x%lx\n", + task_pid_nr(task), i, ctx->th_pmcs[i])); } ia64_srlz_d(); @@ -1274,14 +1323,12 @@ out: } EXPORT_SYMBOL(pfm_unregister_buffer_fmt); -extern void update_pal_halt_status(int); - static int pfm_reserve_session(struct task_struct *task, int is_syswide, unsigned int cpu) { unsigned long flags; /* - * validy checks on cpu_mask have been done upstream + * validity checks on cpu_mask have been done upstream */ LOCK_PFS(flags); @@ -1323,9 +1370,9 @@ pfm_reserve_session(struct task_struct *task, int is_syswide, unsigned int cpu) cpu)); /* - * disable default_idle() to go to PAL_HALT + * Force idle() into poll mode */ - update_pal_halt_status(0); + cpu_idle_poll_ctrl(true); UNLOCK_PFS(flags); @@ -1333,7 +1380,7 @@ pfm_reserve_session(struct task_struct *task, int is_syswide, unsigned int cpu) error_conflict: DPRINT(("system wide not possible, conflicting session [%d] on CPU%d\n", - pfm_sessions.pfs_sys_session[cpu]->pid, + task_pid_nr(pfm_sessions.pfs_sys_session[cpu]), cpu)); abort: UNLOCK_PFS(flags); @@ -1347,7 +1394,7 @@ pfm_unreserve_session(pfm_context_t *ctx, int is_syswide, unsigned int cpu) { unsigned long flags; /* - * validy checks on cpu_mask have been done upstream + * validity checks on cpu_mask have been done upstream */ LOCK_PFS(flags); @@ -1382,11 +1429,8 @@ pfm_unreserve_session(pfm_context_t *ctx, int is_syswide, unsigned int cpu) is_syswide, cpu)); - /* - * if possible, enable default_idle() to go into PAL_HALT - */ - if (pfm_sessions.pfs_task_sessions == 0 && pfm_sessions.pfs_sys_sessions == 0) - update_pal_halt_status(1); + /* Undo forced polling. Last session reenables pal_halt */ + cpu_idle_poll_ctrl(false); UNLOCK_PFS(flags); @@ -1399,13 +1443,14 @@ pfm_unreserve_session(pfm_context_t *ctx, int is_syswide, unsigned int cpu) * a PROTECT_CTX() section. */ static int -pfm_remove_smpl_mapping(struct task_struct *task, void *vaddr, unsigned long size) +pfm_remove_smpl_mapping(void *vaddr, unsigned long size) { + struct task_struct *task = current; int r; /* sanity checks */ if (task->mm == NULL || size == 0UL || vaddr == NULL) { - printk(KERN_ERR "perfmon: pfm_remove_smpl_mapping [%d] invalid context mm=%p\n", task->pid, task->mm); + printk(KERN_ERR "perfmon: pfm_remove_smpl_mapping [%d] invalid context mm=%p\n", task_pid_nr(task), task->mm); return -EINVAL; } @@ -1414,15 +1459,10 @@ pfm_remove_smpl_mapping(struct task_struct *task, void *vaddr, unsigned long siz /* * does the actual unmapping */ - down_write(&task->mm->mmap_sem); + r = vm_munmap((unsigned long)vaddr, size); - DPRINT(("down_write done smpl_vaddr=%p size=%lu\n", vaddr, size)); - - r = pfm_do_munmap(task->mm, (unsigned long)vaddr, size, 0); - - up_write(&task->mm->mmap_sem); if (r !=0) { - printk(KERN_ERR "perfmon: [%d] unable to unmap sampling buffer @%p size=%lu\n", task->pid, vaddr, size); + printk(KERN_ERR "perfmon: [%d] unable to unmap sampling buffer @%p size=%lu\n", task_pid_nr(task), vaddr, size); } DPRINT(("do_unmap(%p, %lu)=%d\n", vaddr, size, r)); @@ -1464,7 +1504,7 @@ pfm_free_smpl_buffer(pfm_context_t *ctx) return 0; invalid_free: - printk(KERN_ERR "perfmon: pfm_free_smpl_buffer [%d] no buffer\n", current->pid); + printk(KERN_ERR "perfmon: pfm_free_smpl_buffer [%d] no buffer\n", task_pid_nr(current)); return -EINVAL; } #endif @@ -1484,7 +1524,7 @@ pfm_exit_smpl_buffer(pfm_buffer_fmt_t *fmt) * any operations on the root directory. However, we need a non-trivial * d_name - pfm: will go nicely and kill the special-casing in procfs. */ -static struct vfsmount *pfmfs_mnt; +static struct vfsmount *pfmfs_mnt __read_mostly; static int __init init_pfm_fs(void) @@ -1501,13 +1541,6 @@ init_pfm_fs(void) return err; } -static void __exit -exit_pfm_fs(void) -{ - unregister_filesystem(&pfm_fs_type); - mntput(pfmfs_mnt); -} - static ssize_t pfm_read(struct file *filp, char __user *buf, size_t size, loff_t *ppos) { @@ -1517,13 +1550,13 @@ pfm_read(struct file *filp, char __user *buf, size_t size, loff_t *ppos) unsigned long flags; DECLARE_WAITQUEUE(wait, current); if (PFM_IS_FILE(filp) == 0) { - printk(KERN_ERR "perfmon: pfm_poll: bad magic [%d]\n", current->pid); + printk(KERN_ERR "perfmon: pfm_poll: bad magic [%d]\n", task_pid_nr(current)); return -EINVAL; } - ctx = (pfm_context_t *)filp->private_data; + ctx = filp->private_data; if (ctx == NULL) { - printk(KERN_ERR "perfmon: pfm_read: NULL ctx [%d]\n", current->pid); + printk(KERN_ERR "perfmon: pfm_read: NULL ctx [%d]\n", task_pid_nr(current)); return -EINVAL; } @@ -1577,7 +1610,7 @@ pfm_read(struct file *filp, char __user *buf, size_t size, loff_t *ppos) PROTECT_CTX(ctx, flags); } - DPRINT(("[%d] back to running ret=%ld\n", current->pid, ret)); + DPRINT(("[%d] back to running ret=%ld\n", task_pid_nr(current), ret)); set_current_state(TASK_RUNNING); remove_wait_queue(&ctx->ctx_msgq_wait, &wait); @@ -1586,7 +1619,7 @@ pfm_read(struct file *filp, char __user *buf, size_t size, loff_t *ppos) ret = -EINVAL; msg = pfm_get_next_msg(ctx); if (msg == NULL) { - printk(KERN_ERR "perfmon: pfm_read no msg for ctx=%p [%d]\n", ctx, current->pid); + printk(KERN_ERR "perfmon: pfm_read no msg for ctx=%p [%d]\n", ctx, task_pid_nr(current)); goto abort_locked; } @@ -1617,13 +1650,13 @@ pfm_poll(struct file *filp, poll_table * wait) unsigned int mask = 0; if (PFM_IS_FILE(filp) == 0) { - printk(KERN_ERR "perfmon: pfm_poll: bad magic [%d]\n", current->pid); + printk(KERN_ERR "perfmon: pfm_poll: bad magic [%d]\n", task_pid_nr(current)); return 0; } - ctx = (pfm_context_t *)filp->private_data; + ctx = filp->private_data; if (ctx == NULL) { - printk(KERN_ERR "perfmon: pfm_poll: NULL ctx [%d]\n", current->pid); + printk(KERN_ERR "perfmon: pfm_poll: NULL ctx [%d]\n", task_pid_nr(current)); return 0; } @@ -1644,8 +1677,8 @@ pfm_poll(struct file *filp, poll_table * wait) return mask; } -static int -pfm_ioctl(struct inode *inode, struct file *file, unsigned int cmd, unsigned long arg) +static long +pfm_ioctl(struct file *file, unsigned int cmd, unsigned long arg) { DPRINT(("pfm_ioctl called\n")); return -EINVAL; @@ -1662,7 +1695,7 @@ pfm_do_fasync(int fd, struct file *filp, pfm_context_t *ctx, int on) ret = fasync_helper (fd, filp, on, &ctx->ctx_async_queue); DPRINT(("pfm_fasync called by [%d] on ctx_fd=%d on=%d async_queue=%p ret=%d\n", - current->pid, + task_pid_nr(current), fd, on, ctx->ctx_async_queue, ret)); @@ -1677,13 +1710,13 @@ pfm_fasync(int fd, struct file *filp, int on) int ret; if (PFM_IS_FILE(filp) == 0) { - printk(KERN_ERR "perfmon: pfm_fasync bad magic [%d]\n", current->pid); + printk(KERN_ERR "perfmon: pfm_fasync bad magic [%d]\n", task_pid_nr(current)); return -EBADF; } - ctx = (pfm_context_t *)filp->private_data; + ctx = filp->private_data; if (ctx == NULL) { - printk(KERN_ERR "perfmon: pfm_fasync NULL ctx [%d]\n", current->pid); + printk(KERN_ERR "perfmon: pfm_fasync NULL ctx [%d]\n", task_pid_nr(current)); return -EBADF; } /* @@ -1729,7 +1762,7 @@ pfm_syswide_force_stop(void *info) if (owner != ctx->ctx_task) { printk(KERN_ERR "perfmon: pfm_syswide_force_stop CPU%d unexpected owner [%d] instead of [%d]\n", smp_processor_id(), - owner->pid, ctx->ctx_task->pid); + task_pid_nr(owner), task_pid_nr(ctx->ctx_task)); return; } if (GET_PMU_CTX() != ctx) { @@ -1739,7 +1772,7 @@ pfm_syswide_force_stop(void *info) return; } - DPRINT(("on CPU%d forcing system wide stop for [%d]\n", smp_processor_id(), ctx->ctx_task->pid)); + DPRINT(("on CPU%d forcing system wide stop for [%d]\n", smp_processor_id(), task_pid_nr(ctx->ctx_task))); /* * the context is already protected in pfm_close(), we simply * need to mask interrupts to avoid a PMU interrupt race on @@ -1764,7 +1797,7 @@ pfm_syswide_cleanup_other_cpu(pfm_context_t *ctx) int ret; DPRINT(("calling CPU%d for cleanup\n", ctx->ctx_cpu)); - ret = smp_call_function_single(ctx->ctx_cpu, pfm_syswide_force_stop, ctx, 0, 1); + ret = smp_call_function_single(ctx->ctx_cpu, pfm_syswide_force_stop, ctx, 1); DPRINT(("called CPU%d for cleanup ret=%d\n", ctx->ctx_cpu, ret)); } #endif /* CONFIG_SMP */ @@ -1789,16 +1822,16 @@ pfm_flush(struct file *filp, fl_owner_t id) return -EBADF; } - ctx = (pfm_context_t *)filp->private_data; + ctx = filp->private_data; if (ctx == NULL) { - printk(KERN_ERR "perfmon: pfm_flush: NULL ctx [%d]\n", current->pid); + printk(KERN_ERR "perfmon: pfm_flush: NULL ctx [%d]\n", task_pid_nr(current)); return -EBADF; } /* * remove our file from the async queue, if we use this mode. * This can be done without the context being protected. We come - * here when the context has become unreacheable by other tasks. + * here when the context has become unreachable by other tasks. * * We may still have active monitoring at this point and we may * end up in pfm_overflow_handler(). However, fasync_helper() @@ -1808,11 +1841,6 @@ pfm_flush(struct file *filp, fl_owner_t id) * invoked after, it will find an empty queue and no * signal will be sent. In both case, we are safe */ - if (filp->f_flags & FASYNC) { - DPRINT(("cleaning up async_queue=%p\n", ctx->ctx_async_queue)); - pfm_do_fasync (-1, filp, ctx, 0); - } - PROTECT_CTX(ctx, flags); state = ctx->ctx_state; @@ -1898,7 +1926,7 @@ pfm_flush(struct file *filp, fl_owner_t id) * because some VM function reenables interrupts. * */ - if (smpl_buf_vaddr) pfm_remove_smpl_mapping(current, smpl_buf_vaddr, smpl_buf_size); + if (smpl_buf_vaddr) pfm_remove_smpl_mapping(smpl_buf_vaddr, smpl_buf_size); return 0; } @@ -1937,9 +1965,9 @@ pfm_close(struct inode *inode, struct file *filp) return -EBADF; } - ctx = (pfm_context_t *)filp->private_data; + ctx = filp->private_data; if (ctx == NULL) { - printk(KERN_ERR "perfmon: pfm_close: NULL ctx [%d]\n", current->pid); + printk(KERN_ERR "perfmon: pfm_close: NULL ctx [%d]\n", task_pid_nr(current)); return -EBADF; } @@ -2036,7 +2064,7 @@ pfm_close(struct inode *inode, struct file *filp) */ ctx->ctx_state = PFM_CTX_ZOMBIE; - DPRINT(("zombie ctx for [%d]\n", task->pid)); + DPRINT(("zombie ctx for [%d]\n", task_pid_nr(task))); /* * cannot free the context on the spot. deferred until * the task notices the ZOMBIE state @@ -2095,7 +2123,7 @@ doit: filp->private_data = NULL; /* - * if we free on the spot, the context is now completely unreacheable + * if we free on the spot, the context is now completely unreachable * from the callers side. The monitored task side is also cut, so we * can freely cut. * @@ -2126,116 +2154,73 @@ pfm_no_open(struct inode *irrelevant, struct file *dontcare) -static struct file_operations pfm_file_ops = { - .llseek = no_llseek, - .read = pfm_read, - .write = pfm_write, - .poll = pfm_poll, - .ioctl = pfm_ioctl, - .open = pfm_no_open, /* special open code to disallow open via /proc */ - .fasync = pfm_fasync, - .release = pfm_close, - .flush = pfm_flush +static const struct file_operations pfm_file_ops = { + .llseek = no_llseek, + .read = pfm_read, + .write = pfm_write, + .poll = pfm_poll, + .unlocked_ioctl = pfm_ioctl, + .open = pfm_no_open, /* special open code to disallow open via /proc */ + .fasync = pfm_fasync, + .release = pfm_close, + .flush = pfm_flush }; -static int -pfmfs_delete_dentry(struct dentry *dentry) +static char *pfmfs_dname(struct dentry *dentry, char *buffer, int buflen) { - return 1; + return dynamic_dname(dentry, buffer, buflen, "pfm:[%lu]", + dentry->d_inode->i_ino); } -static struct dentry_operations pfmfs_dentry_operations = { - .d_delete = pfmfs_delete_dentry, +static const struct dentry_operations pfmfs_dentry_operations = { + .d_delete = always_delete_dentry, + .d_dname = pfmfs_dname, }; -static int -pfm_alloc_fd(struct file **cfile) +static struct file * +pfm_alloc_file(pfm_context_t *ctx) { - int fd, ret = 0; - struct file *file = NULL; - struct inode * inode; - char name[32]; - struct qstr this; - - fd = get_unused_fd(); - if (fd < 0) return -ENFILE; - - ret = -ENFILE; - - file = get_empty_filp(); - if (!file) goto out; + struct file *file; + struct inode *inode; + struct path path; + struct qstr this = { .name = "" }; /* * allocate a new inode */ inode = new_inode(pfmfs_mnt->mnt_sb); - if (!inode) goto out; + if (!inode) + return ERR_PTR(-ENOMEM); DPRINT(("new inode ino=%ld @%p\n", inode->i_ino, inode)); inode->i_mode = S_IFCHR|S_IRUGO; - inode->i_uid = current->fsuid; - inode->i_gid = current->fsgid; - - sprintf(name, "[%lu]", inode->i_ino); - this.name = name; - this.len = strlen(name); - this.hash = inode->i_ino; - - ret = -ENOMEM; + inode->i_uid = current_fsuid(); + inode->i_gid = current_fsgid(); /* * allocate a new dcache entry */ - file->f_dentry = d_alloc(pfmfs_mnt->mnt_sb->s_root, &this); - if (!file->f_dentry) goto out; + path.dentry = d_alloc(pfmfs_mnt->mnt_root, &this); + if (!path.dentry) { + iput(inode); + return ERR_PTR(-ENOMEM); + } + path.mnt = mntget(pfmfs_mnt); - file->f_dentry->d_op = &pfmfs_dentry_operations; + d_add(path.dentry, inode); - d_add(file->f_dentry, inode); - file->f_vfsmnt = mntget(pfmfs_mnt); - file->f_mapping = inode->i_mapping; + file = alloc_file(&path, FMODE_READ, &pfm_file_ops); + if (IS_ERR(file)) { + path_put(&path); + return file; + } - file->f_op = &pfm_file_ops; - file->f_mode = FMODE_READ; file->f_flags = O_RDONLY; - file->f_pos = 0; + file->private_data = ctx; - /* - * may have to delay until context is attached? - */ - fd_install(fd, file); - - /* - * the file structure we will use - */ - *cfile = file; - - return fd; -out: - if (file) put_filp(file); - put_unused_fd(fd); - return ret; -} - -static void -pfm_free_fd(int fd, struct file *file) -{ - struct files_struct *files = current->files; - struct fdtable *fdt; - - /* - * there ie no fd_uninstall(), so we do it here - */ - spin_lock(&files->file_lock); - fdt = files_fdtable(files); - rcu_assign_pointer(fdt->fd[fd], NULL); - spin_unlock(&files->file_lock); - - if (file) - put_filp(file); - put_unused_fd(fd); + return file; } static int @@ -2261,7 +2246,7 @@ pfm_remap_buffer(struct vm_area_struct *vma, unsigned long buf, unsigned long ad * allocate a sampling buffer and remaps it into the user address space of the task */ static int -pfm_smpl_buffer_alloc(struct task_struct *task, pfm_context_t *ctx, unsigned long rsize, void **user_vaddr) +pfm_smpl_buffer_alloc(struct task_struct *task, struct file *filp, pfm_context_t *ctx, unsigned long rsize, void **user_vaddr) { struct mm_struct *mm = task->mm; struct vm_area_struct *vma = NULL; @@ -2284,7 +2269,7 @@ pfm_smpl_buffer_alloc(struct task_struct *task, pfm_context_t *ctx, unsigned lon * if ((mm->total_vm << PAGE_SHIFT) + len> task->rlim[RLIMIT_AS].rlim_cur) * return -ENOMEM; */ - if (size > task->signal->rlim[RLIMIT_MEMLOCK].rlim_cur) + if (size > task_rlimit(task, RLIMIT_MEMLOCK)) return -ENOMEM; /* @@ -2301,18 +2286,19 @@ pfm_smpl_buffer_alloc(struct task_struct *task, pfm_context_t *ctx, unsigned lon DPRINT(("smpl_buf @%p\n", smpl_buf)); /* allocate vma */ - vma = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL); + vma = kmem_cache_zalloc(vm_area_cachep, GFP_KERNEL); if (!vma) { DPRINT(("Cannot allocate vma\n")); goto error_kmem; } - memset(vma, 0, sizeof(*vma)); + INIT_LIST_HEAD(&vma->anon_vma_chain); /* * partially initialize the vma for the sampling buffer */ vma->vm_mm = mm; - vma->vm_flags = VM_READ| VM_MAYREAD |VM_RESERVED; + vma->vm_file = get_file(filp); + vma->vm_flags = VM_READ|VM_MAYREAD|VM_DONTEXPAND|VM_DONTDUMP; vma->vm_page_prot = PAGE_READONLY; /* XXX may need to change */ /* @@ -2332,8 +2318,8 @@ pfm_smpl_buffer_alloc(struct task_struct *task, pfm_context_t *ctx, unsigned lon down_write(&task->mm->mmap_sem); /* find some free area in address space, must have mmap sem held */ - vma->vm_start = pfm_get_unmapped_area(NULL, 0, size, 0, MAP_PRIVATE|MAP_ANONYMOUS, 0); - if (vma->vm_start == 0UL) { + vma->vm_start = get_unmapped_area(NULL, 0, size, 0, MAP_PRIVATE|MAP_ANONYMOUS); + if (IS_ERR_VALUE(vma->vm_start)) { DPRINT(("Cannot find unmapped area for size %ld\n", size)); up_write(&task->mm->mmap_sem); goto error; @@ -2356,7 +2342,6 @@ pfm_smpl_buffer_alloc(struct task_struct *task, pfm_context_t *ctx, unsigned lon */ insert_vm_struct(mm, vma); - mm->total_vm += size >> PAGE_SHIFT; vm_stat_account(vma->vm_mm, vma->vm_flags, vma->vm_file, vma_pages(vma)); up_write(&task->mm->mmap_sem); @@ -2383,22 +2368,33 @@ error_kmem: static int pfm_bad_permissions(struct task_struct *task) { + const struct cred *tcred; + kuid_t uid = current_uid(); + kgid_t gid = current_gid(); + int ret; + + rcu_read_lock(); + tcred = __task_cred(task); + /* inspired by ptrace_attach() */ DPRINT(("cur: uid=%d gid=%d task: euid=%d suid=%d uid=%d egid=%d sgid=%d\n", - current->uid, - current->gid, - task->euid, - task->suid, - task->uid, - task->egid, - task->sgid)); - - return ((current->uid != task->euid) - || (current->uid != task->suid) - || (current->uid != task->uid) - || (current->gid != task->egid) - || (current->gid != task->sgid) - || (current->gid != task->gid)) && !capable(CAP_SYS_PTRACE); + from_kuid(&init_user_ns, uid), + from_kgid(&init_user_ns, gid), + from_kuid(&init_user_ns, tcred->euid), + from_kuid(&init_user_ns, tcred->suid), + from_kuid(&init_user_ns, tcred->uid), + from_kgid(&init_user_ns, tcred->egid), + from_kgid(&init_user_ns, tcred->sgid))); + + ret = ((!uid_eq(uid, tcred->euid)) + || (!uid_eq(uid, tcred->suid)) + || (!uid_eq(uid, tcred->uid)) + || (!gid_eq(gid, tcred->egid)) + || (!gid_eq(gid, tcred->sgid)) + || (!gid_eq(gid, tcred->gid))) && !capable(CAP_SYS_PTRACE); + + rcu_read_unlock(); + return ret; } static int @@ -2427,7 +2423,7 @@ pfarg_is_sane(struct task_struct *task, pfarg_context_t *pfx) } static int -pfm_setup_buffer_fmt(struct task_struct *task, pfm_context_t *ctx, unsigned int ctx_flags, +pfm_setup_buffer_fmt(struct task_struct *task, struct file *filp, pfm_context_t *ctx, unsigned int ctx_flags, unsigned int cpu, pfarg_context_t *arg) { pfm_buffer_fmt_t *fmt = NULL; @@ -2440,7 +2436,7 @@ pfm_setup_buffer_fmt(struct task_struct *task, pfm_context_t *ctx, unsigned int /* invoke and lock buffer format, if found */ fmt = pfm_find_buffer_fmt(arg->ctx_smpl_buf_id); if (fmt == NULL) { - DPRINT(("[%d] cannot find buffer format\n", task->pid)); + DPRINT(("[%d] cannot find buffer format\n", task_pid_nr(task))); return -EINVAL; } @@ -2451,12 +2447,13 @@ pfm_setup_buffer_fmt(struct task_struct *task, pfm_context_t *ctx, unsigned int ret = pfm_buf_fmt_validate(fmt, task, ctx_flags, cpu, fmt_arg); - DPRINT(("[%d] after validate(0x%x,%d,%p)=%d\n", task->pid, ctx_flags, cpu, fmt_arg, ret)); + DPRINT(("[%d] after validate(0x%x,%d,%p)=%d\n", task_pid_nr(task), ctx_flags, cpu, fmt_arg, ret)); if (ret) goto error; /* link buffer format and context */ ctx->ctx_buf_fmt = fmt; + ctx->ctx_fl_is_sampling = 1; /* assume record() is defined */ /* * check if buffer format wants to use perfmon buffer allocation/mapping service @@ -2468,7 +2465,7 @@ pfm_setup_buffer_fmt(struct task_struct *task, pfm_context_t *ctx, unsigned int /* * buffer is always remapped into the caller's address space */ - ret = pfm_smpl_buffer_alloc(current, ctx, size, &uaddr); + ret = pfm_smpl_buffer_alloc(current, filp, ctx, size, &uaddr); if (ret) goto error; /* keep track of user address of buffer */ @@ -2523,7 +2520,7 @@ pfm_reset_pmu_state(pfm_context_t *ctx) ctx->ctx_all_pmcs[0] = pmu_conf->impl_pmcs[0] & ~0x1; /* - * bitmask of all PMDs that are accesible to this context + * bitmask of all PMDs that are accessible to this context */ ctx->ctx_all_pmds[0] = pmu_conf->impl_pmds[0]; @@ -2573,23 +2570,23 @@ pfm_task_incompatible(pfm_context_t *ctx, struct task_struct *task) * no kernel task or task not owner by caller */ if (task->mm == NULL) { - DPRINT(("task [%d] has not memory context (kernel thread)\n", task->pid)); + DPRINT(("task [%d] has not memory context (kernel thread)\n", task_pid_nr(task))); return -EPERM; } if (pfm_bad_permissions(task)) { - DPRINT(("no permission to attach to [%d]\n", task->pid)); + DPRINT(("no permission to attach to [%d]\n", task_pid_nr(task))); return -EPERM; } /* * cannot block in self-monitoring mode */ if (CTX_OVFL_NOBLOCK(ctx) == 0 && task == current) { - DPRINT(("cannot load a blocking context on self for [%d]\n", task->pid)); + DPRINT(("cannot load a blocking context on self for [%d]\n", task_pid_nr(task))); return -EINVAL; } if (task->exit_state == EXIT_ZOMBIE) { - DPRINT(("cannot attach to zombie task [%d]\n", task->pid)); + DPRINT(("cannot attach to zombie task [%d]\n", task_pid_nr(task))); return -EBUSY; } @@ -2598,14 +2595,14 @@ pfm_task_incompatible(pfm_context_t *ctx, struct task_struct *task) */ if (task == current) return 0; - if ((task->state != TASK_STOPPED) && (task->state != TASK_TRACED)) { - DPRINT(("cannot attach to non-stopped task [%d] state=%ld\n", task->pid, task->state)); + if (!task_is_stopped_or_traced(task)) { + DPRINT(("cannot attach to non-stopped task [%d] state=%ld\n", task_pid_nr(task), task->state)); return -EBUSY; } /* * make sure the task is off any CPU */ - wait_task_inactive(task); + wait_task_inactive(task, 0); /* more to come... */ @@ -2621,11 +2618,11 @@ pfm_get_task(pfm_context_t *ctx, pid_t pid, struct task_struct **task) /* XXX: need to add more checks here */ if (pid < 2) return -EPERM; - if (pid != current->pid) { + if (pid != task_pid_vnr(current)) { read_lock(&tasklist_lock); - p = find_task_by_pid(pid); + p = find_task_by_vpid(pid); /* make sure task cannot go away while we operate on it */ if (p) get_task_struct(p); @@ -2651,79 +2648,46 @@ pfm_context_create(pfm_context_t *ctx, void *arg, int count, struct pt_regs *reg { pfarg_context_t *req = (pfarg_context_t *)arg; struct file *filp; + struct path path; int ctx_flags; + int fd; int ret; /* let's check the arguments first */ ret = pfarg_is_sane(current, req); - if (ret < 0) return ret; + if (ret < 0) + return ret; ctx_flags = req->ctx_flags; ret = -ENOMEM; - ctx = pfm_context_alloc(); - if (!ctx) goto error; + fd = get_unused_fd(); + if (fd < 0) + return fd; - ret = pfm_alloc_fd(&filp); - if (ret < 0) goto error_file; + ctx = pfm_context_alloc(ctx_flags); + if (!ctx) + goto error; - req->ctx_fd = ctx->ctx_fd = ret; + filp = pfm_alloc_file(ctx); + if (IS_ERR(filp)) { + ret = PTR_ERR(filp); + goto error_file; + } - /* - * attach context to file - */ - filp->private_data = ctx; + req->ctx_fd = ctx->ctx_fd = fd; /* * does the user want to sample? */ if (pfm_uuid_cmp(req->ctx_smpl_buf_id, pfm_null_uuid)) { - ret = pfm_setup_buffer_fmt(current, ctx, ctx_flags, 0, req); - if (ret) goto buffer_error; + ret = pfm_setup_buffer_fmt(current, filp, ctx, ctx_flags, 0, req); + if (ret) + goto buffer_error; } - /* - * init context protection lock - */ - spin_lock_init(&ctx->ctx_lock); - - /* - * context is unloaded - */ - ctx->ctx_state = PFM_CTX_UNLOADED; - - /* - * initialization of context's flags - */ - ctx->ctx_fl_block = (ctx_flags & PFM_FL_NOTIFY_BLOCK) ? 1 : 0; - ctx->ctx_fl_system = (ctx_flags & PFM_FL_SYSTEM_WIDE) ? 1: 0; - ctx->ctx_fl_is_sampling = ctx->ctx_buf_fmt ? 1 : 0; /* assume record() is defined */ - ctx->ctx_fl_no_msg = (ctx_flags & PFM_FL_OVFL_NO_MSG) ? 1: 0; - /* - * will move to set properties - * ctx->ctx_fl_excl_idle = (ctx_flags & PFM_FL_EXCL_IDLE) ? 1: 0; - */ - - /* - * init restart semaphore to locked - */ - init_completion(&ctx->ctx_restart_done); - - /* - * activation is used in SMP only - */ - ctx->ctx_last_activation = PFM_INVALID_ACTIVATION; - SET_LAST_CPU(ctx, -1); - - /* - * initialize notification message queue - */ - ctx->ctx_msgq_head = ctx->ctx_msgq_tail = 0; - init_waitqueue_head(&ctx->ctx_msgq_wait); - init_waitqueue_head(&ctx->ctx_zombieq); - - DPRINT(("ctx=%p flags=0x%x system=%d notify_block=%d excl_idle=%d no_msg=%d ctx_fd=%d \n", + DPRINT(("ctx=%p flags=0x%x system=%d notify_block=%d excl_idle=%d no_msg=%d ctx_fd=%d\n", ctx, ctx_flags, ctx->ctx_fl_system, @@ -2737,10 +2701,14 @@ pfm_context_create(pfm_context_t *ctx, void *arg, int count, struct pt_regs *reg */ pfm_reset_pmu_state(ctx); + fd_install(fd, filp); + return 0; buffer_error: - pfm_free_fd(ctx->ctx_fd, filp); + path = filp->f_path; + put_filp(filp); + path_put(&path); if (ctx->ctx_buf_fmt) { pfm_buf_fmt_exit(ctx->ctx_buf_fmt, current, NULL, regs); @@ -2749,6 +2717,7 @@ error_file: pfm_context_free(ctx); error: + put_unused_fd(fd); return ret; } @@ -3356,7 +3325,7 @@ pfm_read_pmds(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs) if (unlikely(!PMD_IS_IMPL(cnum))) goto error; /* * we can only read the register that we use. That includes - * the one we explicitely initialize AND the one we want included + * the one we explicitly initialize AND the one we want included * in the sampling buffer (smpl_regs). * * Having this restriction allows optimization in the ctxsw routine @@ -3480,7 +3449,7 @@ pfm_use_debug_registers(struct task_struct *task) if (pmu_conf->use_rr_dbregs == 0) return 0; - DPRINT(("called for [%d]\n", task->pid)); + DPRINT(("called for [%d]\n", task_pid_nr(task))); /* * do it only once @@ -3511,7 +3480,7 @@ pfm_use_debug_registers(struct task_struct *task) DPRINT(("ptrace_use_dbregs=%u sys_use_dbregs=%u by [%d] ret = %d\n", pfm_sessions.pfs_ptrace_use_dbregs, pfm_sessions.pfs_sys_use_dbregs, - task->pid, ret)); + task_pid_nr(task), ret)); UNLOCK_PFS(flags); @@ -3523,7 +3492,7 @@ pfm_use_debug_registers(struct task_struct *task) * IA64_THREAD_DBG_VALID set. This indicates a task which was * able to use the debug registers for debugging purposes via * ptrace(). Therefore we know it was not using them for - * perfmormance monitoring, so we only decrement the number + * performance monitoring, so we only decrement the number * of "ptraced" debug register users to keep the count up to date */ int @@ -3536,7 +3505,7 @@ pfm_release_debug_registers(struct task_struct *task) LOCK_PFS(flags); if (pfm_sessions.pfs_ptrace_use_dbregs == 0) { - printk(KERN_ERR "perfmon: invalid release for [%d] ptrace_use_dbregs=0\n", task->pid); + printk(KERN_ERR "perfmon: invalid release for [%d] ptrace_use_dbregs=0\n", task_pid_nr(task)); ret = -1; } else { pfm_sessions.pfs_ptrace_use_dbregs--; @@ -3588,7 +3557,7 @@ pfm_restart(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs) /* sanity check */ if (unlikely(task == NULL)) { - printk(KERN_ERR "perfmon: [%d] pfm_restart no task\n", current->pid); + printk(KERN_ERR "perfmon: [%d] pfm_restart no task\n", task_pid_nr(current)); return -EINVAL; } @@ -3597,7 +3566,7 @@ pfm_restart(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs) fmt = ctx->ctx_buf_fmt; DPRINT(("restarting self %d ovfl=0x%lx\n", - task->pid, + task_pid_nr(task), ctx->ctx_ovfl_regs[0])); if (CTX_HAS_SMPL(ctx)) { @@ -3621,11 +3590,11 @@ pfm_restart(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs) pfm_reset_regs(ctx, ctx->ctx_ovfl_regs, PFM_PMD_LONG_RESET); if (rst_ctrl.bits.mask_monitoring == 0) { - DPRINT(("resuming monitoring for [%d]\n", task->pid)); + DPRINT(("resuming monitoring for [%d]\n", task_pid_nr(task))); if (state == PFM_CTX_MASKED) pfm_restore_monitoring(task); } else { - DPRINT(("keeping monitoring stopped for [%d]\n", task->pid)); + DPRINT(("keeping monitoring stopped for [%d]\n", task_pid_nr(task))); // cannot use pfm_stop_monitoring(task, regs); } @@ -3676,22 +3645,22 @@ pfm_restart(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs) * if non-blocking, then we ensure that the task will go into * pfm_handle_work() before returning to user mode. * - * We cannot explicitely reset another task, it MUST always + * We cannot explicitly reset another task, it MUST always * be done by the task itself. This works for system wide because * the tool that is controlling the session is logically doing * "self-monitoring". */ if (CTX_OVFL_NOBLOCK(ctx) == 0 && state == PFM_CTX_MASKED) { - DPRINT(("unblocking [%d] \n", task->pid)); + DPRINT(("unblocking [%d]\n", task_pid_nr(task))); complete(&ctx->ctx_restart_done); } else { - DPRINT(("[%d] armed exit trap\n", task->pid)); + DPRINT(("[%d] armed exit trap\n", task_pid_nr(task))); ctx->ctx_fl_trap_reason = PFM_TRAP_REASON_RESET; PFM_SET_WORK_PENDING(task, 1); - pfm_set_task_notify(task); + set_notify_resume(task); /* * XXX: send reschedule if task runs on another CPU @@ -3773,7 +3742,7 @@ pfm_write_ibr_dbr(int mode, pfm_context_t *ctx, void *arg, int count, struct pt_ * don't bother if we are loaded and task is being debugged */ if (is_loaded && (thread->flags & IA64_THREAD_DBG_VALID) != 0) { - DPRINT(("debug registers already in use for [%d]\n", task->pid)); + DPRINT(("debug registers already in use for [%d]\n", task_pid_nr(task))); return -EBUSY; } @@ -3814,7 +3783,7 @@ pfm_write_ibr_dbr(int mode, pfm_context_t *ctx, void *arg, int count, struct pt_ * is shared by all processes running on it */ if (first_time && can_access_pmu) { - DPRINT(("[%d] clearing ibrs, dbrs\n", task->pid)); + DPRINT(("[%d] clearing ibrs, dbrs\n", task_pid_nr(task))); for (i=0; i < pmu_conf->num_ibrs; i++) { ia64_set_ibr(i, 0UL); ia64_dv_serialize_instruction(); @@ -4003,7 +3972,7 @@ pfm_stop(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs) return -EBUSY; } DPRINT(("task [%d] ctx_state=%d is_system=%d\n", - PFM_CTX_TASK(ctx)->pid, + task_pid_nr(PFM_CTX_TASK(ctx)), state, is_system)); /* @@ -4061,7 +4030,7 @@ pfm_stop(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs) * monitoring disabled in kernel at next reschedule */ ctx->ctx_saved_psr_up = 0; - DPRINT(("task=[%d]\n", task->pid)); + DPRINT(("task=[%d]\n", task_pid_nr(task))); } return 0; } @@ -4186,10 +4155,10 @@ pfm_check_task_exist(pfm_context_t *ctx) do_each_thread (g, t) { if (t->thread.pfm_context == ctx) { ret = 0; - break; + goto out; } } while_each_thread (g, t); - +out: read_unlock(&tasklist_lock); DPRINT(("pfm_check_task_exist: ret=%d ctx=%p\n", ret, ctx)); @@ -4266,11 +4235,12 @@ pfm_context_load(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs) if (is_system) { if (pfm_sessions.pfs_ptrace_use_dbregs) { - DPRINT(("cannot load [%d] dbregs in use\n", task->pid)); + DPRINT(("cannot load [%d] dbregs in use\n", + task_pid_nr(task))); ret = -EBUSY; } else { pfm_sessions.pfs_sys_use_dbregs++; - DPRINT(("load [%d] increased sys_use_dbreg=%u\n", task->pid, pfm_sessions.pfs_sys_use_dbregs)); + DPRINT(("load [%d] increased sys_use_dbreg=%u\n", task_pid_nr(task), pfm_sessions.pfs_sys_use_dbregs)); set_dbregs = 1; } } @@ -4362,7 +4332,7 @@ pfm_context_load(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs) /* allow user level control */ ia64_psr(regs)->sp = 0; - DPRINT(("clearing psr.sp for [%d]\n", task->pid)); + DPRINT(("clearing psr.sp for [%d]\n", task_pid_nr(task))); SET_LAST_CPU(ctx, smp_processor_id()); INC_ACTIVATION(); @@ -4397,7 +4367,7 @@ pfm_context_load(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs) */ SET_PMU_OWNER(task, ctx); - DPRINT(("context loaded on PMU for [%d]\n", task->pid)); + DPRINT(("context loaded on PMU for [%d]\n", task_pid_nr(task))); } else { /* * when not current, task MUST be stopped, so this is safe @@ -4461,7 +4431,7 @@ pfm_context_unload(pfm_context_t *ctx, void *arg, int count, struct pt_regs *reg int prev_state, is_system; int ret; - DPRINT(("ctx_state=%d task [%d]\n", ctx->ctx_state, task ? task->pid : -1)); + DPRINT(("ctx_state=%d task [%d]\n", ctx->ctx_state, task ? task_pid_nr(task) : -1)); prev_state = ctx->ctx_state; is_system = ctx->ctx_fl_system; @@ -4536,7 +4506,7 @@ pfm_context_unload(pfm_context_t *ctx, void *arg, int count, struct pt_regs *reg */ ia64_psr(regs)->sp = 1; - DPRINT(("setting psr.sp for [%d]\n", task->pid)); + DPRINT(("setting psr.sp for [%d]\n", task_pid_nr(task))); } /* * save PMDs to context @@ -4576,7 +4546,7 @@ pfm_context_unload(pfm_context_t *ctx, void *arg, int count, struct pt_regs *reg ctx->ctx_fl_can_restart = 0; ctx->ctx_fl_going_zombie = 0; - DPRINT(("disconnected [%d] from context\n", task->pid)); + DPRINT(("disconnected [%d] from context\n", task_pid_nr(task))); return 0; } @@ -4599,22 +4569,22 @@ pfm_exit_thread(struct task_struct *task) PROTECT_CTX(ctx, flags); - DPRINT(("state=%d task [%d]\n", ctx->ctx_state, task->pid)); + DPRINT(("state=%d task [%d]\n", ctx->ctx_state, task_pid_nr(task))); state = ctx->ctx_state; switch(state) { case PFM_CTX_UNLOADED: /* - * only comes to thios function if pfm_context is not NULL, i.e., cannot + * only comes to this function if pfm_context is not NULL, i.e., cannot * be in unloaded state */ - printk(KERN_ERR "perfmon: pfm_exit_thread [%d] ctx unloaded\n", task->pid); + printk(KERN_ERR "perfmon: pfm_exit_thread [%d] ctx unloaded\n", task_pid_nr(task)); break; case PFM_CTX_LOADED: case PFM_CTX_MASKED: ret = pfm_context_unload(ctx, NULL, 0, regs); if (ret) { - printk(KERN_ERR "perfmon: pfm_exit_thread [%d] state=%d unload failed %d\n", task->pid, state, ret); + printk(KERN_ERR "perfmon: pfm_exit_thread [%d] state=%d unload failed %d\n", task_pid_nr(task), state, ret); } DPRINT(("ctx unloaded for current state was %d\n", state)); @@ -4623,12 +4593,12 @@ pfm_exit_thread(struct task_struct *task) case PFM_CTX_ZOMBIE: ret = pfm_context_unload(ctx, NULL, 0, regs); if (ret) { - printk(KERN_ERR "perfmon: pfm_exit_thread [%d] state=%d unload failed %d\n", task->pid, state, ret); + printk(KERN_ERR "perfmon: pfm_exit_thread [%d] state=%d unload failed %d\n", task_pid_nr(task), state, ret); } free_ok = 1; break; default: - printk(KERN_ERR "perfmon: pfm_exit_thread [%d] unexpected state=%d\n", task->pid, state); + printk(KERN_ERR "perfmon: pfm_exit_thread [%d] unexpected state=%d\n", task_pid_nr(task), state); break; } UNPROTECT_CTX(ctx, flags); @@ -4712,7 +4682,7 @@ recheck: DPRINT(("context %d state=%d [%d] task_state=%ld must_stop=%d\n", ctx->ctx_fd, state, - task->pid, + task_pid_nr(task), task->state, PFM_CMD_STOPPED(cmd))); /* @@ -4758,8 +4728,8 @@ recheck: * the task must be stopped. */ if (PFM_CMD_STOPPED(cmd)) { - if ((task->state != TASK_STOPPED) && (task->state != TASK_TRACED)) { - DPRINT(("[%d] task not in stopped state\n", task->pid)); + if (!task_is_stopped_or_traced(task)) { + DPRINT(("[%d] task not in stopped state\n", task_pid_nr(task))); return -EBUSY; } /* @@ -4780,7 +4750,7 @@ recheck: UNPROTECT_CTX(ctx, flags); - wait_task_inactive(task); + wait_task_inactive(task, 0); PROTECT_CTX(ctx, flags); @@ -4801,7 +4771,7 @@ recheck: asmlinkage long sys_perfmonctl (int fd, int cmd, void __user *arg, int count) { - struct file *file = NULL; + struct fd f = {NULL, 0}; pfm_context_t *ctx = NULL; unsigned long flags = 0UL; void *args_k = NULL; @@ -4852,7 +4822,7 @@ restart_args: * limit abuse to min page size */ if (unlikely(sz > PFM_MAX_ARGSIZE)) { - printk(KERN_ERR "perfmon: [%d] argument too big %lu\n", current->pid, sz); + printk(KERN_ERR "perfmon: [%d] argument too big %lu\n", task_pid_nr(current), sz); return -E2BIG; } @@ -4898,17 +4868,17 @@ restart_args: ret = -EBADF; - file = fget(fd); - if (unlikely(file == NULL)) { + f = fdget(fd); + if (unlikely(f.file == NULL)) { DPRINT(("invalid fd %d\n", fd)); goto error_args; } - if (unlikely(PFM_IS_FILE(file) == 0)) { + if (unlikely(PFM_IS_FILE(f.file) == 0)) { DPRINT(("fd %d not related to perfmon\n", fd)); goto error_args; } - ctx = (pfm_context_t *)file->private_data; + ctx = f.file->private_data; if (unlikely(ctx == NULL)) { DPRINT(("no context for fd %d\n", fd)); goto error_args; @@ -4938,8 +4908,8 @@ abort_locked: if (call_made && PFM_CMD_RW_ARG(cmd) && copy_to_user(arg, args_k, base_sz*count)) ret = -EFAULT; error_args: - if (file) - fput(file); + if (f.file) + fdput(f); kfree(args_k); @@ -4999,11 +4969,11 @@ pfm_context_force_terminate(pfm_context_t *ctx, struct pt_regs *regs) { int ret; - DPRINT(("entering for [%d]\n", current->pid)); + DPRINT(("entering for [%d]\n", task_pid_nr(current))); ret = pfm_context_unload(ctx, NULL, 0, regs); if (ret) { - printk(KERN_ERR "pfm_context_force_terminate: [%d] unloaded failed with %d\n", current->pid, ret); + printk(KERN_ERR "pfm_context_force_terminate: [%d] unloaded failed with %d\n", task_pid_nr(current), ret); } /* @@ -5019,12 +4989,13 @@ pfm_context_force_terminate(pfm_context_t *ctx, struct pt_regs *regs) } static int pfm_ovfl_notify_user(pfm_context_t *ctx, unsigned long ovfl_pmds); + /* * pfm_handle_work() can be called with interrupts enabled * (TIF_NEED_RESCHED) or disabled. The down_interruptible * call may sleep, therefore we must re-enable interrupts * to avoid deadlocks. It is safe to do so because this function - * is called ONLY when returning to user level (PUStk=1), in which case + * is called ONLY when returning to user level (pUStk=1), in which case * there is no risk of kernel stack overflow due to deep * interrupt nesting. */ @@ -5040,7 +5011,8 @@ pfm_handle_work(void) ctx = PFM_GET_CTX(current); if (ctx == NULL) { - printk(KERN_ERR "perfmon: [%d] has no PFM context\n", current->pid); + printk(KERN_ERR "perfmon: [%d] has no PFM context\n", + task_pid_nr(current)); return; } @@ -5048,8 +5020,6 @@ pfm_handle_work(void) PFM_SET_WORK_PENDING(current, 0); - pfm_clear_task_notify(); - regs = task_pt_regs(current); /* @@ -5064,11 +5034,12 @@ pfm_handle_work(void) /* * must be done before we check for simple-reset mode */ - if (ctx->ctx_fl_going_zombie || ctx->ctx_state == PFM_CTX_ZOMBIE) goto do_zombie; - + if (ctx->ctx_fl_going_zombie || ctx->ctx_state == PFM_CTX_ZOMBIE) + goto do_zombie; //if (CTX_OVFL_NOBLOCK(ctx)) goto skip_blocking; - if (reason == PFM_TRAP_REASON_RESET) goto skip_blocking; + if (reason == PFM_TRAP_REASON_RESET) + goto skip_blocking; /* * restore interrupt mask to what it was on entry. @@ -5116,7 +5087,8 @@ do_zombie: /* * in case of interruption of down() we don't restart anything */ - if (ret < 0) goto nothing_to_do; + if (ret < 0) + goto nothing_to_do; skip_blocking: pfm_resume_after_ovfl(ctx, ovfl_regs, regs); @@ -5208,10 +5180,10 @@ pfm_end_notify_user(pfm_context_t *ctx) /* * main overflow processing routine. - * it can be called from the interrupt path or explicitely during the context switch code + * it can be called from the interrupt path or explicitly during the context switch code */ -static void -pfm_overflow_handler(struct task_struct *task, pfm_context_t *ctx, u64 pmc0, struct pt_regs *regs) +static void pfm_overflow_handler(struct task_struct *task, pfm_context_t *ctx, + unsigned long pmc0, struct pt_regs *regs) { pfm_ovfl_arg_t *ovfl_arg; unsigned long mask; @@ -5237,7 +5209,7 @@ pfm_overflow_handler(struct task_struct *task, pfm_context_t *ctx, u64 pmc0, str DPRINT_ovfl(("pmc0=0x%lx pid=%d iip=0x%lx, %s " "used_pmds=0x%lx\n", pmc0, - task ? task->pid: -1, + task ? task_pid_nr(task): -1, (regs ? regs->cr_iip : 0), CTX_OVFL_NOBLOCK(ctx) ? "nonblocking" : "blocking", ctx->ctx_used_pmds[0])); @@ -5416,7 +5388,7 @@ pfm_overflow_handler(struct task_struct *task, pfm_context_t *ctx, u64 pmc0, str * when coming from ctxsw, current still points to the * previous task, therefore we must work with task and not current. */ - pfm_set_task_notify(task); + set_notify_resume(task); } /* * defer until state is changed (shorten spin window). the context is locked @@ -5426,7 +5398,7 @@ pfm_overflow_handler(struct task_struct *task, pfm_context_t *ctx, u64 pmc0, str } DPRINT_ovfl(("owner [%d] pending=%ld reason=%u ovfl_pmds=0x%lx ovfl_notify=0x%lx masked=%d\n", - GET_PMU_OWNER() ? GET_PMU_OWNER()->pid : -1, + GET_PMU_OWNER() ? task_pid_nr(GET_PMU_OWNER()) : -1, PFM_GET_WORK_PENDING(task), ctx->ctx_fl_trap_reason, ovfl_pmds, @@ -5451,7 +5423,7 @@ pfm_overflow_handler(struct task_struct *task, pfm_context_t *ctx, u64 pmc0, str sanity_check: printk(KERN_ERR "perfmon: CPU%d overflow handler [%d] pmc0=0x%lx\n", smp_processor_id(), - task ? task->pid : -1, + task ? task_pid_nr(task) : -1, pmc0); return; @@ -5484,7 +5456,7 @@ stop_monitoring: * * Overall pretty hairy stuff.... */ - DPRINT(("ctx is zombie for [%d], converted to spurious\n", task ? task->pid: -1)); + DPRINT(("ctx is zombie for [%d], converted to spurious\n", task ? task_pid_nr(task): -1)); pfm_clear_psr_up(); ia64_psr(regs)->up = 0; ia64_psr(regs)->sp = 1; @@ -5492,7 +5464,7 @@ stop_monitoring: } static int -pfm_do_interrupt_handler(int irq, void *arg, struct pt_regs *regs) +pfm_do_interrupt_handler(void *arg, struct pt_regs *regs) { struct task_struct *task; pfm_context_t *ctx; @@ -5545,13 +5517,13 @@ pfm_do_interrupt_handler(int irq, void *arg, struct pt_regs *regs) report_spurious1: printk(KERN_INFO "perfmon: spurious overflow interrupt on CPU%d: process %d has no PFM context\n", - this_cpu, task->pid); + this_cpu, task_pid_nr(task)); pfm_unfreeze_pmu(); return -1; report_spurious2: printk(KERN_INFO "perfmon: spurious overflow interrupt on CPU%d: process %d, invalid flag\n", this_cpu, - task->pid); + task_pid_nr(task)); pfm_unfreeze_pmu(); return -1; } @@ -5572,7 +5544,7 @@ pfm_interrupt_handler(int irq, void *arg) start_cycles = ia64_get_itc(); - ret = pfm_do_interrupt_handler(irq, arg, regs); + ret = pfm_do_interrupt_handler(arg, regs); total_cycles = ia64_get_itc(); @@ -5592,7 +5564,7 @@ pfm_interrupt_handler(int irq, void *arg) (*pfm_alt_intr_handler->handler)(irq, arg, regs); } - put_cpu_no_resched(); + put_cpu(); return IRQ_HANDLED; } @@ -5600,7 +5572,7 @@ pfm_interrupt_handler(int irq, void *arg) * /proc/perfmon interface, for debug only */ -#define PFM_PROC_SHOW_HEADER ((void *)NR_CPUS+1) +#define PFM_PROC_SHOW_HEADER ((void *)(long)nr_cpu_ids+1) static void * pfm_proc_start(struct seq_file *m, loff_t *pos) @@ -5609,7 +5581,7 @@ pfm_proc_start(struct seq_file *m, loff_t *pos) return PFM_PROC_SHOW_HEADER; } - while (*pos <= NR_CPUS) { + while (*pos <= nr_cpu_ids) { if (cpu_online(*pos - 1)) { return (void *)*pos; } @@ -5669,24 +5641,8 @@ pfm_proc_show_header(struct seq_file *m) list_for_each(pos, &pfm_buffer_fmt_list) { entry = list_entry(pos, pfm_buffer_fmt_t, fmt_list); - seq_printf(m, "format : %02x-%02x-%02x-%02x-%02x-%02x-%02x-%02x-%02x-%02x-%02x-%02x-%02x-%02x-%02x-%02x %s\n", - entry->fmt_uuid[0], - entry->fmt_uuid[1], - entry->fmt_uuid[2], - entry->fmt_uuid[3], - entry->fmt_uuid[4], - entry->fmt_uuid[5], - entry->fmt_uuid[6], - entry->fmt_uuid[7], - entry->fmt_uuid[8], - entry->fmt_uuid[9], - entry->fmt_uuid[10], - entry->fmt_uuid[11], - entry->fmt_uuid[12], - entry->fmt_uuid[13], - entry->fmt_uuid[14], - entry->fmt_uuid[15], - entry->fmt_name); + seq_printf(m, "format : %16phD %s\n", + entry->fmt_uuid, entry->fmt_name); } spin_unlock(&pfm_buffer_fmt_lock); @@ -5761,7 +5717,7 @@ pfm_proc_show(struct seq_file *m, void *v) return 0; } -struct seq_operations pfm_seq_ops = { +const struct seq_operations pfm_seq_ops = { .start = pfm_proc_start, .next = pfm_proc_next, .stop = pfm_proc_stop, @@ -5838,7 +5794,8 @@ pfm_force_cleanup(pfm_context_t *ctx, struct pt_regs *regs) ia64_psr(regs)->sp = 1; if (GET_PMU_OWNER() == task) { - DPRINT(("cleared ownership for [%d]\n", ctx->ctx_task->pid)); + DPRINT(("cleared ownership for [%d]\n", + task_pid_nr(ctx->ctx_task))); SET_PMU_OWNER(NULL, NULL); } @@ -5850,7 +5807,7 @@ pfm_force_cleanup(pfm_context_t *ctx, struct pt_regs *regs) task->thread.pfm_context = NULL; task->thread.flags &= ~IA64_THREAD_PM_VALID; - DPRINT(("force cleanup for [%d]\n", task->pid)); + DPRINT(("force cleanup for [%d]\n", task_pid_nr(task))); } @@ -6394,7 +6351,7 @@ pfm_flush_pmds(struct task_struct *task, pfm_context_t *ctx) if (PMD_IS_COUNTING(i)) { DPRINT(("[%d] pmd[%d] ctx_pmd=0x%lx hw_pmd=0x%lx\n", - task->pid, + task_pid_nr(task), i, ctx->ctx_pmds[i].val, val & ovfl_val)); @@ -6416,11 +6373,11 @@ pfm_flush_pmds(struct task_struct *task, pfm_context_t *ctx) */ if (pmc0 & (1UL << i)) { val += 1 + ovfl_val; - DPRINT(("[%d] pmd[%d] overflowed\n", task->pid, i)); + DPRINT(("[%d] pmd[%d] overflowed\n", task_pid_nr(task), i)); } } - DPRINT(("[%d] ctx_pmd[%d]=0x%lx pmd_val=0x%lx\n", task->pid, i, val, pmd_val)); + DPRINT(("[%d] ctx_pmd[%d]=0x%lx pmd_val=0x%lx\n", task_pid_nr(task), i, val, pmd_val)); if (is_self) ctx->th_pmds[i] = pmd_val; @@ -6430,7 +6387,6 @@ pfm_flush_pmds(struct task_struct *task, pfm_context_t *ctx) static struct irqaction perfmon_irqaction = { .handler = pfm_interrupt_handler, - .flags = IRQF_DISABLED, .name = "perfmon" }; @@ -6509,7 +6465,7 @@ pfm_install_alt_pmu_interrupt(pfm_intr_handler_desc_t *hdl) } /* save the current system wide pmu states */ - ret = on_each_cpu(pfm_alt_save_pmu_state, NULL, 0, 1); + ret = on_each_cpu(pfm_alt_save_pmu_state, NULL, 1); if (ret) { DPRINT(("on_each_cpu() failed: %d\n", ret)); goto cleanup_reserve; @@ -6554,7 +6510,7 @@ pfm_remove_alt_pmu_interrupt(pfm_intr_handler_desc_t *hdl) pfm_alt_intr_handler = NULL; - ret = on_each_cpu(pfm_alt_restore_pmu_state, NULL, 0, 1); + ret = on_each_cpu(pfm_alt_restore_pmu_state, NULL, 1); if (ret) { DPRINT(("on_each_cpu() failed: %d\n", ret)); } @@ -6597,7 +6553,7 @@ found: return 0; } -static struct file_operations pfm_proc_fops = { +static const struct file_operations pfm_proc_fops = { .open = pfm_proc_open, .read = seq_read, .llseek = seq_lseek, @@ -6675,21 +6631,17 @@ pfm_init(void) /* * create /proc/perfmon (mostly for debugging purposes) */ - perfmon_dir = create_proc_entry("perfmon", S_IRUGO, NULL); + perfmon_dir = proc_create("perfmon", S_IRUGO, NULL, &pfm_proc_fops); if (perfmon_dir == NULL) { printk(KERN_ERR "perfmon: cannot create /proc entry, perfmon disabled\n"); pmu_conf = NULL; return -1; } - /* - * install customized file operations for /proc/perfmon entry - */ - perfmon_dir->proc_fops = &pfm_proc_fops; /* * create /proc/sys/kernel/perfmon (for debugging purposes) */ - pfm_sysctl_header = register_sysctl_table(pfm_sysctl_root, 0); + pfm_sysctl_header = register_sysctl_table(pfm_sysctl_root); /* * initialize all our spinlocks @@ -6761,14 +6713,14 @@ dump_pmu_state(const char *from) printk("CPU%d from %s() current [%d] iip=0x%lx %s\n", this_cpu, from, - current->pid, + task_pid_nr(current), regs->cr_iip, current->comm); task = GET_PMU_OWNER(); ctx = GET_PMU_CTX(); - printk("->CPU%d owner [%d] ctx=%p\n", this_cpu, task ? task->pid : -1, ctx); + printk("->CPU%d owner [%d] ctx=%p\n", this_cpu, task ? task_pid_nr(task) : -1, ctx); psr = pfm_get_psr(); @@ -6816,7 +6768,7 @@ pfm_inherit(struct task_struct *task, struct pt_regs *regs) { struct thread_struct *thread; - DPRINT(("perfmon: pfm_inherit clearing state for [%d]\n", task->pid)); + DPRINT(("perfmon: pfm_inherit clearing state for [%d]\n", task_pid_nr(task))); thread = &task->thread; |