]> git.baikalelectronics.ru Git - kernel.git/commit
perf/core: Collapse more IPI loops
authorPeter Zijlstra <peterz@infradead.org>
Thu, 3 Dec 2015 17:35:21 +0000 (18:35 +0100)
committerIngo Molnar <mingo@kernel.org>
Wed, 6 Jan 2016 10:15:29 +0000 (11:15 +0100)
commite5ae5e110046dadb99a7ef0a47f0e4b09d4fb236
tree0536354f45426a6438e94512d6f38dbae8da697a
parent5552443f1a5863595d0a37b7cce1a9faccf9ba7d
perf/core: Collapse more IPI loops

This patch collapses the two 'hard' cases, which are
perf_event_{dis,en}able().

I cannot seem to convince myself the current code is correct.

So starting with perf_event_disable(); we don't strictly need to test
for event->state == ACTIVE, ctx->is_active is enough. If the event is
not scheduled while the ctx is, __perf_event_disable() still does the
right thing.  Its a little less efficient to IPI in that case,
over-all simpler.

For perf_event_enable(); the same goes, but I think that's actually
broken in its current form. The current condition is: ctx->is_active
&& event->state == OFF, that means it doesn't do anything when
!ctx->active && event->state == OFF. This is wrong, it should still
mark the event INACTIVE in that case, otherwise we'll still not try
and schedule the event once the context becomes active again.

This patch implements the two function using the new
event_function_call() and does away with the tricky event->state
tests.

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Alexander Shishkin <alexander.shishkin@intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
kernel/events/core.c