watchdog: make sure the watchdog thread gets CPU on loaded system
authorMichal Hocko <mhocko@suse.cz>
Fri, 23 Mar 2012 22:01:55 +0000 (15:01 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Fri, 23 Mar 2012 23:58:32 +0000 (16:58 -0700)
commit7a05c0f7bbae91d08b7d0acf016fdb42dbc912ae
tree0fca719baf9c60f5659a28bc977f975acd2ec107
parent397a21f24d455982a8a6f9bc11b5f3326ce3c6ef
watchdog: make sure the watchdog thread gets CPU on loaded system

If the system is loaded while hotplugging a CPU we might end up with a
bogus hardlockup detection.  This has been seen during LTP pounder test
executed in parallel with hotplug test.

The main problem is that enable_watchdog (called when CPU is brought up)
registers perf event which periodically checks per-cpu counter
(hrtimer_interrupts), updated from a hrtimer callback, but the hrtimer
is fired from the kernel thread.

This means that while we already do check for the hard lockup the kernel
thread might be sitting on the runqueue with zillions of tasks so there
is nobody to update the value we rely on and so we KABOOM.

Let's fix this by boosting the watchdog thread priority before we wake
it up rather than when it's already running.  This still doesn't handle
a case where we have the same amount of high prio FIFO tasks but that
doesn't seem to be common.  The current implementation doesn't handle
that case anyway so this is not worse at least.

Unfortunately, we cannot start perf counter from the watchdog thread
because we could miss a real lock up and also we cannot start the
hrtimer watchdog_enable because we there is no way (at least I don't
know any) to start a hrtimer from a different CPU.

[dzickus@redhat.com: fix compile issue with param]
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Reviewed-by: Mandeep Singh Baines <msb@chromium.org>
Signed-off-by: Michal Hocko <mhocko@suse.cz>
Signed-off-by: Don Zickus <dzickus@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
kernel/watchdog.c