perf_counter: Provide a way to enable counters on exec
PATCH v2 1/1 perf_counter: Provide a way to enable counters on exec
This provides a way to mark a counter to be enabled on the next exec. This is useful for measuring the total activity of a program without including overhead from the process that launches it.
This also changes the perf stat command to use this new facility.
---
v2: only unclone if we enabled one or more counters.
This differs from Ingo's approach a little in that I don't keep the counter force-disabled until exec, and I reuse the comm hook instead of having a new perf_counter_exec hook (not that a new hook would necessarily be a bad idea). Also, my locking is a bit heavier, it's more or less like the old perf_counter_task_enable().
include/linux/perf_counter.h | 3 +- kernel/perf_counter.c | 50 ++++++++++++++++++++++++++++++++++++++++++ tools/perf/builtin-stat.c | 6 ++-- 3 files changed, 55 insertions(+), 4 deletions(-)
include/linux/perf_counter.h +2 -1
@@ -179,8 +179,9 @@ struct perf_counter_attr { comm : 1, /* include comm data */ freq : 1, /* use freq, not period */ inherit_stat : 1, /* per task counts */ + enable_on_exec : 1, /* next exec enables */ - __reserved_1 : 52; + __reserved_1 : 51;__u32 wakeup_events;/* wakeup every n events */ __u32 __reserved_2;
kernel/perf_counter.c +50 -0
62 lines not on this page · raw message
tools/perf/builtin-stat.c +3 -3
@@ -116,8 +116,9 @@ static void create_perf_stat_counter(int counter, int pid) fd[cpu][counter], strerror(errno)); } }else { - attr->inherit = inherit; - attr->disabled =1 ; + attr->inherit = inherit; + attr->disabled =1 ; + attr->enable_on_exec =1 ; fd[0 ][counter] =sys_perf_counter_open (attr, pid,-1 ,-1 ,0 );if (fd[0 ][counter] <0 && verbose)
@@ -262,7 +263,6 @@ static int run_perf_stat(int argc, const char **argv)* Enable counters and exec the command: */ t0 =rdclock (); -prctl (PR_TASK_PERF_COUNTERS_ENABLE );close (go_pipe[1 ]);wait (&status);
Signature · 1 line
1.6.0.4
Commit-ID: 57e7986ed142417498155ebcd5eaf617ac37136d Gitweb: http://git.kernel.org/tip/57e7986ed142417498155ebcd5eaf617ac37136d Author: Paul Mackerras <[email protected]> AuthorDate: Tue, 30 Jun 2009 16:07:19 +1000 Committer: Ingo Molnar <[email protected]> CommitDate: Tue, 30 Jun 2009 12:00:16 +0200
perf_counter: Provide a way to enable counters on exec
This provides a way to mark a counter to be enabled on the next exec. This is useful for measuring the total activity of a program without including overhead from the process that launches it.
This also changes the perf stat command to use this new facility.
LKML-Reference: <[email protected]>
--- include/linux/perf_counter.h | 3 +- kernel/perf_counter.c | 50 ++++++++++++++++++++++++++++++++++++++++++ tools/perf/builtin-stat.c | 6 ++-- 3 files changed, 55 insertions(+), 4 deletions(-)
include/linux/perf_counter.h +2 -1
@@ -179,8 +179,9 @@ struct perf_counter_attr { comm : 1, /* include comm data */ freq : 1, /* use freq, not period */ inherit_stat : 1, /* per task counts */ + enable_on_exec : 1, /* next exec enables */ - __reserved_1 : 52; + __reserved_1 : 51;__u32 wakeup_events;/* wakeup every n events */ __u32 __reserved_2;
kernel/perf_counter.c +50 -0
62 lines not on this page · raw message
tools/perf/builtin-stat.c +3 -3
@@ -116,8 +116,9 @@ static void create_perf_stat_counter(int counter, int pid) fd[cpu][counter], strerror(errno)); } }else { - attr->inherit = inherit; - attr->disabled =1 ; + attr->inherit = inherit; + attr->disabled =1 ; + attr->enable_on_exec =1 ; fd[0 ][counter] =sys_perf_counter_open (attr, pid,-1 ,-1 ,0 );if (fd[0 ][counter] <0 && verbose)
@@ -262,7 +263,6 @@ static int run_perf_stat(int argc, const char **argv)* Enable counters and exec the command: */ t0 =rdclock (); -prctl (PR_TASK_PERF_COUNTERS_ENABLE );close (go_pipe[1 ]);wait (&status);
Quoted 7 lines · Paul Mackerras
This provides a way to mark a counter to be enabled on the next exec. This is useful for measuring the total activity of a program without including overhead from the process that launches it.
This also changes the perf stat command to use this new facility.
Signed-off-by Paul Mackerras <[email protected]>
Quoted 8 lines
---
v2: only unclone if we enabled one or more counters.This differs from Ingo's approach a little in that I don't keep the counter force-disabled until exec, and I reuse the comm hook instead of having a new
perf_counter_exechook (not that a new hook would necessarily be a bad idea). Also, my locking is a bit heavier, it's more or less like the oldperf_counter_task_enable().
Hmm, we might want to add that second hook since this will also trigger for things like prctl(PR_SET_NAME).
Also, we would probably want this below...
---
Subject: perf_counter: unify unclone context
We have grown multiple unclone sites, stick it in a function.
--- kernel/perf_counter.c | 36 +++++++++++++++++------------------- 1 files changed, 17 insertions(+), 19 deletions(-)
kernel/perf_counter.c +17 -19
@@ -146,6 +146,14 @@ static void put_ctx(struct perf_counter_context *ctx) } } +static void unclone_ctx (struct perf_counter_context *ctx) +{ +if (ctx->parent_ctx) { +put_ctx (ctx->parent_ctx); + ctx->parent_ctx =NULL ; + } +} +/* * Get the perf_counter_context for a task and lock it. * This has to cope with with the fact that until it is locked,
@@ -1463,10 +1471,8 @@ static void perf_counter_enable_on_exec(struct task_struct *task)/* * Unclone this context if we enabled any counter. */ -if (enabled && ctx->parent_ctx) { -put_ctx (ctx->parent_ctx); - ctx->parent_ctx =NULL ; - } +if (enabled) +unclone_ctx (ctx);spin_unlock (&ctx->lock);
@@ -1526,7 +1532,6 @@ __perf_counter_init_context(struct perf_counter_context *ctx,static struct perf_counter_context *find_get_context (pid_t pid,int cpu) { -struct perf_counter_context *parent_ctx;struct perf_counter_context *ctx;struct perf_cpu_context *cpuctx;struct task_struct *task;
@@ -1586,11 +1591,7 @@ static struct perf_counter_context *find_get_context(pid_t pid, int cpu) retry: ctx =perf_lock_task_context (task, &flags);if (ctx) { - parent_ctx = ctx->parent_ctx; -if (parent_ctx) { -put_ctx (parent_ctx); - ctx->parent_ctx =NULL ;/* no longer a clone */ - } +unclone_ctx (ctx);spin_unlock_irqrestore (&ctx->lock, flags); }
@@ -4255,15 +4256,12 @@ void perf_counter_exit_task(struct task_struct *child)*/ spin_lock (&child_ctx->lock); child->perf_counter_ctxp =NULL ; -if (child_ctx->parent_ctx) { -/* -* This context is a clone; unclone it so it can't get -* swapped to another process while we're removing all -* the counters from it. -*/ -put_ctx (child_ctx->parent_ctx); - child_ctx->parent_ctx =NULL ; - } +/* +* If this context is a clone; unclone it so it can't get +* swapped to another process while we're removing all +* the counters from it. +*/ +unclone_ctx (child_ctx);spin_unlock (&child_ctx->lock);local_irq_restore (flags);