PM: runtime: Add safety net to supplier device release
authorRafael J. Wysocki <rafael.j.wysocki@intel.com>
Fri, 10 Dec 2021 16:10:13 +0000 (17:10 +0100)
committerRafael J. Wysocki <rafael.j.wysocki@intel.com>
Fri, 17 Dec 2021 15:25:14 +0000 (16:25 +0100)
commitd1579e61192e0e686faa4208500ef4c3b529b16c
tree0f359554ef6c967b002cbbf214d572271fb96a14
parentc24efa6732788f0be22cdf5d2aedd5e3117e983f
PM: runtime: Add safety net to supplier device release

Because refcount_dec_not_one() returns true if the target refcount
becomes saturated, it is generally unsafe to use its return value as
a loop termination condition, but that is what happens when a device
link's supplier device is released during runtime PM suspend
operations and on device link removal.

To address this, introduce pm_runtime_release_supplier() to be used
in the above cases which will check the supplier device's runtime
PM usage counter in addition to the refcount_dec_not_one() return
value, so the loop can be terminated in case the rpm_active refcount
value becomes invalid, and update the code in question to use it as
appropriate.

This change is not expected to have any visible functional impact.

Reported-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
drivers/base/core.c
drivers/base/power/runtime.c
include/linux/pm_runtime.h