[PATCH 6.12.y v2 3/5] sched/deadline: Always stop dl-server before changing parameters

From: Juri Lelli <[email protected]>

commit bb4700adc3abec34c0a38b64f66258e4e233fc16 upstream.

Commit cccb45d7c4295 ("sched/deadline: Less agressive dl_server
handling") reduced dl-server overhead by delaying disabling servers only
after there are no fair task around for a whole period, which means that
deadline entities are not dequeued right away on a server stop event.
However, the delay opens up a window in which a request for changing
server parameters can break per-runqueue running_bw tracking, as
reported by Yuri.

Close the problematic window by unconditionally calling dl_server_stop()
before applying the new parameters (ensuring deadline entities go
through an actual dequeue).

Fixes: cccb45d7c4295 ("sched/deadline: Less agressive dl_server handling")
Reported-by: Yuri Andriaccio <[email protected]>
Signed-off-by: Juri Lelli <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Reviewed-by: Valentin Schneider <[email protected]>
Link: https://lore.kernel.org/r/20250721-upstream-fix-dlserver-lessaggressive-b4-v1-1-4ebc10c87e40@redhat.com
Signed-off-by: Lukas Beckmann <[email protected]>
---
 kernel/sched/debug.c | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)

diff --git a/kernel/sched/debug.c b/kernel/sched/debug.c
index 7d14e9fa53ac..564ea17ae405 100644
--- a/kernel/sched/debug.c
+++ b/kernel/sched/debug.c
@@ -378,10 +378,8 @@ static ssize_t sched_fair_server_write(struct file *filp, const char __user *ubu
 			return  -EINVAL;
 		}
 
-		if (rq->cfs.h_nr_queued) {
-			update_rq_clock(rq);
-			dl_server_stop(&rq->fair_server);
-		}
+		update_rq_clock(rq);
+		dl_server_stop(&rq->fair_server);
 
 		retval = dl_server_apply_params(&rq->fair_server, runtime, period, 0);
 
-- 
2.54.0