Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
sched/fair: Fix the negative lag increase fix
Vincent reported that my rework of his original patch lost a little
something.
Specifically it got the return value wrong; it should not compare
against the old se->vlag, but rather against the current value. Since
the thing that matters is if the effective vruntime of an entity is
affected and the thing needs repositioning or not.
Fixes: 059258b0d424 ("sched/fair: Prevent negative lag increase during delayed dequeue")
Reported-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Tested-by: Vincent Guittot <vincent.guittot@linaro.org>
Link: https://patch.msgid.link/20260423094107.GT3102624%40noisy.programming.kicks-ass.net
+10
-5
+10
-5
kernel/sched/fair.c
+10
-5
kernel/sched/fair.c
···
847
847
* Similarly, check that the entity didn't gain positive lag when DELAY_ZERO
848
848
* is set.
849
849
*
850
-
* Return true if the lag has been adjusted.
850
+
* Return true if the vlag has been modified. Specifically:
851
+
*
852
+
* se->vlag != avg_vruntime() - se->vruntime
853
+
*
854
+
* This can be due to clamping in entity_lag() or clamping due to
855
+
* sched_delayed. Either way, when vlag is modified and the entity is
856
+
* retained, the tree needs to be adjusted.
851
857
*/
852
858
static __always_inline
853
859
bool update_entity_lag(struct cfs_rq *cfs_rq, struct sched_entity *se)
854
860
{
855
-
s64 vlag = entity_lag(cfs_rq, se, avg_vruntime(cfs_rq));
856
-
bool ret;
861
+
u64 avruntime = avg_vruntime(cfs_rq);
862
+
s64 vlag = entity_lag(cfs_rq, se, avruntime);
857
863
858
864
WARN_ON_ONCE(!se->on_rq);
859
865
···
869
863
if (sched_feat(DELAY_ZERO))
870
864
vlag = min(vlag, 0);
871
865
}
872
-
ret = (vlag == se->vlag);
873
866
se->vlag = vlag;
874
867
875
-
return ret;
868
+
return avruntime - vlag != se->vruntime;
876
869
}
877
870
878
871
/*