tick: broadcast: Check broadcast mode on CPU hotplug
authorStephen Boyd <sboyd@codeaurora.org>
Thu, 11 Jul 2013 14:00:59 +0000 (07:00 -0700)
committerThomas Gleixner <tglx@linutronix.de>
Fri, 12 Jul 2013 10:35:40 +0000 (12:35 +0200)
commita272dcca1802a7e265a56e60b0d0a6715b0a8ac2
tree3bc6e5294be0ff0fed757752dab4b2e4d52dde03
parentf2006e27396f55276f24434f56e208d86e7f9908
tick: broadcast: Check broadcast mode on CPU hotplug

On ARM systems the dummy clockevent is registered with the cpu
hotplug notifier chain before any other per-cpu clockevent. This
has the side-effect of causing the dummy clockevent to be
registered first in every hotplug sequence. Because the dummy is
first, we'll try to turn the broadcast source on but the code in
tick_device_uses_broadcast() assumes the broadcast source is in
periodic mode and calls tick_broadcast_start_periodic()
unconditionally.

On boot this isn't a problem because we typically haven't
switched into oneshot mode yet (if at all). During hotplug, if
the broadcast source isn't in periodic mode we'll replace the
broadcast oneshot handler with the broadcast periodic handler and
start emulating oneshot mode when we shouldn't. Due to the way
the broadcast oneshot handler programs the next_event it's
possible for it to contain KTIME_MAX and cause us to hang the
system when the periodic handler tries to program the next tick.
Fix this by using the appropriate function to start the broadcast
source.

Reported-by: Stephen Warren <swarren@nvidia.com>
Tested-by: Stephen Warren <swarren@nvidia.com>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Cc: Mark Rutland <Mark.Rutland@arm.com>
Cc: Marc Zyngier <marc.zyngier@arm.com>
Cc: ARM kernel mailing list <linux-arm-kernel@lists.infradead.org>
Cc: John Stultz <john.stultz@linaro.org>
Cc: Joseph Lo <josephl@nvidia.com>
Link: http://lkml.kernel.org/r/20130711140059.GA27430@codeaurora.org
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
kernel/time/tick-broadcast.c