clockevent: Prevent dead lock on clockevents_lock
commitf833bab87fca5c3ce13778421b1365845843b976
authorSuresh Siddha <suresh.b.siddha@intel.com>
Mon, 17 Aug 2009 21:34:59 +0000 (17 14:34 -0700)
committerThomas Gleixner <tglx@linutronix.de>
Wed, 19 Aug 2009 16:15:10 +0000 (19 18:15 +0200)
tree020c3e6a70dba34544c963c34647b605035be8fb
parentde809347aeef0a68c04576c464414d0e4dce59fc
clockevent: Prevent dead lock on clockevents_lock

Currently clockevents_notify() is called with interrupts enabled at
some places and interrupts disabled at some other places.

This results in a deadlock in this scenario.

cpu A holds clockevents_lock in clockevents_notify() with irqs enabled
cpu B waits for clockevents_lock in clockevents_notify() with irqs disabled
cpu C doing set_mtrr() which will try to rendezvous of all the cpus.

This will result in C and A come to the rendezvous point and waiting
for B. B is stuck forever waiting for the spinlock and thus not
reaching the rendezvous point.

Fix the clockevents code so that clockevents_lock is taken with
interrupts disabled and thus avoid the above deadlock.

Also call lapic_timer_propagate_broadcast() on the destination cpu so
that we avoid calling smp_call_function() in the clockevents notifier
chain.

This issue left us wondering if we need to change the MTRR rendezvous
logic to use stop machine logic (instead of smp_call_function) or add
a check in spinlock debug code to see if there are other spinlocks
which gets taken under both interrupts enabled/disabled conditions.

Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
Cc: "Pallipadi Venkatesh" <venkatesh.pallipadi@intel.com>
Cc: "Brown Len" <len.brown@intel.com>
LKML-Reference: <1250544899.2709.210.camel@sbs-t61.sc.intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
arch/x86/kernel/process.c
drivers/acpi/processor_idle.c
kernel/time/clockevents.c
kernel/time/tick-broadcast.c