sched/fair: Fix enqueue_task_fair() warning some more
authorPhil Auld <pauld@redhat.com>
Tue, 12 May 2020 13:52:22 +0000 (09:52 -0400)
committerPeter Zijlstra <peterz@infradead.org>
Tue, 19 May 2020 18:34:10 +0000 (20:34 +0200)
sched/fair: Fix enqueue_task_fair warning some more

The recent patch, fe61468b2cb (sched/fair: Fix enqueue_task_fair warning)
did not fully resolve the issues with the rq->tmp_alone_branch !=
&rq->leaf_cfs_rq_list warning in enqueue_task_fair. There is a case where
the first for_each_sched_entity loop exits due to on_rq, having incompletely
updated the list.  In this case the second for_each_sched_entity loop can
further modify se. The later code to fix up the list management fails to do
what is needed because se does not point to the sched_entity which broke out
of the first loop. The list is not fixed up because the throttled parent was
already added back to the list by a task enqueue in a parallel child hierarchy.

Address this by calling list_add_leaf_cfs_rq if there are throttled parents
while doing the second for_each_sched_entity loop.

Fixes: fe61468b2cb ("sched/fair: Fix enqueue_task_fair warning")
Suggested-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Phil Auld <pauld@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Dietmar Eggemann <dietmar.eggemann@arm.com>
Reviewed-by: Vincent Guittot <vincent.guittot@linaro.org>
Link: https://lkml.kernel.org/r/20200512135222.GC2201@lorien.usersys.redhat.com
kernel/sched/fair.c

index 02f323b..c6d57c3 100644 (file)
@@ -5479,6 +5479,13 @@ enqueue_task_fair(struct rq *rq, struct task_struct *p, int flags)
                /* end evaluation on encountering a throttled cfs_rq */
                if (cfs_rq_throttled(cfs_rq))
                        goto enqueue_throttle;
+
+               /*
+                * One parent has been throttled and cfs_rq removed from the
+                * list. Add it back to not break the leaf list.
+                */
+               if (throttled_hierarchy(cfs_rq))
+                       list_add_leaf_cfs_rq(cfs_rq);
        }
 
 enqueue_throttle: