OSDN Git Service

sched: Fix UCLAMP_FLAG_IDLE setting
authorQuentin Perret <qperret@google.com>
Thu, 5 Aug 2021 10:21:53 +0000 (11:21 +0100)
committerPeter Zijlstra <peterz@infradead.org>
Fri, 6 Aug 2021 12:25:25 +0000 (14:25 +0200)
commitca4984a7dd863f3e1c0df775ae3e744bff24c303
tree18e9c49b5f93cd2b7b572f6478aec674809763e0
parentb4da13aa28d4fd0071247b7b41c579ee8a86c81a
sched: Fix UCLAMP_FLAG_IDLE setting

The UCLAMP_FLAG_IDLE flag is set on a runqueue when dequeueing the last
uclamp active task (that is, when buckets.tasks reaches 0 for all
buckets) to maintain the last uclamp.max and prevent blocked util from
suddenly becoming visible.

However, there is an asymmetry in how the flag is set and cleared which
can lead to having the flag set whilst there are active tasks on the rq.
Specifically, the flag is cleared in the uclamp_rq_inc() path, which is
called at enqueue time, but set in uclamp_rq_dec_id() which is called
both when dequeueing a task _and_ in the update_uclamp_active() path. As
a result, when both uclamp_rq_{dec,ind}_id() are called from
update_uclamp_active(), the flag ends up being set but not cleared,
hence leaving the runqueue in a broken state.

Fix this by clearing the flag in update_uclamp_active() as well.

Fixes: e496187da710 ("sched/uclamp: Enforce last task's UCLAMP_MAX")
Reported-by: Rick Yiu <rickyiu@google.com>
Signed-off-by: Quentin Perret <qperret@google.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Qais Yousef <qais.yousef@arm.com>
Tested-by: Dietmar Eggemann <dietmar.eggemann@arm.com>
Link: https://lore.kernel.org/r/20210805102154.590709-2-qperret@google.com
kernel/sched/core.c