perf_events: fix bogus context time tracking
PATCH 1/1 perf_events: fix bogus context time tracking
You can only call update_context_time() when the context is active, i.e., the thread it is attached to is still running.
However, perf_event_read() can be called even when the context is inactive, e.g., user read() the counters. The call to update_context_time() must be conditioned on the status of the context, otherwise, bogus time_enabled, time_running may be returned. Here is an example on AMD64. The task program is an example from libpfm4. The -p prints deltas every 1s.
$ task -p -e cpu_clk_unhalted sleep 5
2,266,610 cpu_clk_unhalted (0.00% scaling, ena=2,158,982, run=2,158,982)
0 cpu_clk_unhalted (0.00% scaling, ena=2,158,982, run=2,158,982)
0 cpu_clk_unhalted (0.00% scaling, ena=2,158,982, run=2,158,982)
0 cpu_clk_unhalted (0.00% scaling, ena=2,158,982, run=2,158,982)
0 cpu_clk_unhalted (0.00% scaling, ena=2,158,982, run=2,158,982)
5,242,358,071 cpu_clk_unhalted (99.95% scaling, ena=5,000,359,984, run=2,319,270)Whereas if you don't read deltas, e.g., no call to perf_event_read() until the process terminates:
$ task -e cpu_clk_unhalted sleep 5
2,497,783 cpu_clk_unhalted (0.00% scaling, ena=2,376,899, run=2,376,899)Notice that time_enable, time_running are bogus in the first example causing bogus scaling.
This patch fixes the problem, by conditionally calling update_context_time() in perf_event_read().
---
kernel/perf_event.c +7 -1
@@ -1775,7 +1775,13 @@ static u64 perf_event_read(struct perf_event *event)unsigned long flags;raw_spin_lock_irqsave (&ctx->lock, flags); -update_context_time (ctx); +/* +* may read while context is not active +* (e.g., thread is blocked), in that case +* we cannot update context time +*/ +if (ctx->is_active) +update_context_time (ctx);update_event_times (event);raw_spin_unlock_irqrestore (&ctx->lock, flags); }
Quoted 24 lines · Stephane Eranian
You can only call
update_context_time()when the context is active, i.e., the thread it is attached to is still running.However,
perf_event_read()can be called even when the context is inactive, e.g., userread()the counters. The call toupdate_context_time()must be conditioned on the status of the context, otherwise, bogustime_enabled,time_runningmay be returned. Here is an example on AMD64. The task program is an example from libpfm4. The -p prints deltas every 1s.$ task -p -e cpu_clk_unhalted sleep 5 2,266,610 cpu_clk_unhalted (0.00% scaling, ena=2,158,982, run=2,158,982) 0 cpu_clk_unhalted (0.00% scaling, ena=2,158,982, run=2,158,982) 0 cpu_clk_unhalted (0.00% scaling, ena=2,158,982, run=2,158,982) 0 cpu_clk_unhalted (0.00% scaling, ena=2,158,982, run=2,158,982) 0 cpu_clk_unhalted (0.00% scaling, ena=2,158,982, run=2,158,982) 5,242,358,071 cpu_clk_unhalted (99.95% scaling, ena=5,000,359,984, run=2,319,270)Whereas if you don't read deltas, e.g., no call to
perf_event_read()until the process terminates:$ task -e cpu_clk_unhalted sleep 5 2,497,783 cpu_clk_unhalted (0.00% scaling, ena=2,376,899, run=2,376,899)
Quoted 7 lines
Notice that
time_enable,time_runningare bogus in the first example causing bogus scaling.This patch fixes the problem, by conditionally calling
update_context_time()inperf_event_read().Signed-off-by Stephane Eranian <[email protected]>
Thanks, tagged it for -stable as well.
[tip:perf/core] perf_events: Fix bogus context time tracking
In reply to Stephane EranianCommit-ID: c530ccd9a1864a44a7ff35826681229ce9f2357a Gitweb: http://git.kernel.org/tip/c530ccd9a1864a44a7ff35826681229ce9f2357a Author: Stephane Eranian <[email protected]> AuthorDate: Fri, 15 Oct 2010 15:26:01 +0200 Committer: Ingo Molnar <[email protected]> CommitDate: Mon, 18 Oct 2010 19:58:46 +0200
perf_events: Fix bogus context time tracking
You can only call update_context_time() when the context is active, i.e., the thread it is attached to is still running.
However, perf_event_read() can be called even when the context is inactive, e.g., user read() the counters. The call to update_context_time() must be conditioned on the status of the context, otherwise, bogus time_enabled, time_running may be returned. Here is an example on AMD64. The task program is an example from libpfm4. The -p prints deltas every 1s.
$ task -p -e cpu_clk_unhalted sleep 5
2,266,610 cpu_clk_unhalted (0.00% scaling, ena=2,158,982, run=2,158,982)
0 cpu_clk_unhalted (0.00% scaling, ena=2,158,982, run=2,158,982)
0 cpu_clk_unhalted (0.00% scaling, ena=2,158,982, run=2,158,982)
0 cpu_clk_unhalted (0.00% scaling, ena=2,158,982, run=2,158,982)
0 cpu_clk_unhalted (0.00% scaling, ena=2,158,982, run=2,158,982)
5,242,358,071 cpu_clk_unhalted (99.95% scaling, ena=5,000,359,984, run=2,319,270)Whereas if you don't read deltas, e.g., no call to perf_event_read() until the process terminates:
$ task -e cpu_clk_unhalted sleep 5
2,497,783 cpu_clk_unhalted (0.00% scaling, ena=2,376,899, run=2,376,899)Notice that time_enable, time_running are bogus in the first example causing bogus scaling.
This patch fixes the problem, by conditionally calling update_context_time() in perf_event_read().
LKML-Reference: <[email protected]>
--- kernel/perf_event.c | 8 +++++++- 1 files changed, 7 insertions(+), 1 deletions(-)
kernel/perf_event.c +7 -1
@@ -1780,7 +1780,13 @@ static u64 perf_event_read(struct perf_event *event)unsigned long flags;raw_spin_lock_irqsave (&ctx->lock, flags); -update_context_time (ctx); +/* +* may read while context is not active +* (e.g., thread is blocked), in that case +* we cannot update context time +*/ +if (ctx->is_active) +update_context_time (ctx);update_event_times (event);raw_spin_unlock_irqrestore (&ctx->lock, flags); }