]> git.karo-electronics.de Git - linux-beck.git/commit
perf: Drop sample rate when sampling is too slow
authorDave Hansen <dave.hansen@linux.intel.com>
Fri, 21 Jun 2013 15:51:36 +0000 (08:51 -0700)
committerIngo Molnar <mingo@kernel.org>
Sun, 23 Jun 2013 09:52:57 +0000 (11:52 +0200)
commit14c63f17b1fde5a575a28e96547a22b451c71fb5
tree781f7327f4341a3d27197e88994b1859e9b51722
parent2ab00456ea8a0d79acb1390659b98416111880b2
perf: Drop sample rate when sampling is too slow

This patch keeps track of how long perf's NMI handler is taking,
and also calculates how many samples perf can take a second.  If
the sample length times the expected max number of samples
exceeds a configurable threshold, it drops the sample rate.

This way, we don't have a runaway sampling process eating up the
CPU.

This patch can tend to drop the sample rate down to level where
perf doesn't work very well.  *BUT* the alternative is that my
system hangs because it spends all of its time handling NMIs.

I'll take a busted performance tool over an entire system that's
busted and undebuggable any day.

BTW, my suspicion is that there's still an underlying bug here.
Using the HPET instead of the TSC is definitely a contributing
factor, but I suspect there are some other things going on.
But, I can't go dig down on a bug like that with my machine
hanging all the time.

Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: paulus@samba.org
Cc: acme@ghostprotocols.net
Cc: Dave Hansen <dave@sr71.net>
[ Prettified it a bit. ]
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Documentation/sysctl/kernel.txt
arch/x86/kernel/cpu/perf_event.c
include/linux/perf_event.h
kernel/events/core.c
kernel/sysctl.c