Re: [PATCH 6.12.y 0/3] backport missing dependencies of d66792919d4f

[Lukas Beckmann <lbckmnn@mailbox.org>]

On Sun, May 24, 2026 at 08:09:19AM -0400, Sasha Levin wrote:
> > Commit d66792919d4f ("sched/deadline: Use revised wakeup rule for
> > dl_server") in the 6.12.y stable tree (upstream commit 14a857056466)
> > depends on three upstream commits that were not backported:
> >
> >   commit cccb45d7c429 ("sched/deadline: Less agressive dl_server handling")
> >   commit 4ae8d9aa9f9d ("sched/deadline: Fix dl_server getting stuck")
> >   commit a3a70caf7906 ("sched/deadline: Fix dl_server behaviour")
> 
> Thanks for tracking this down. Before I queue this, the series is
> missing two more follow-up fixes that both carry "Fixes: cccb45d7c429"
> and that are needed for the dl_server logic to behave correctly:
> 
>   4717432dfd99 ("sched/deadline: Fix dl_server_stopped()")
>   bb4700adc3ab ("sched/deadline: Always stop dl-server before changing parameters")
> 
> Without 4717432dfd99 the dl_server_stopped() check is inverted (it
> returns the wrong polarity after cccb45d7c429), and without
> bb4700adc3ab the per-rq running_bw accounting can get out of sync when
> dl-server parameters change while it is still active.

I did notice 4717432dfd99 while preparing v1, since it caused the
conflicts in a3a70caf7906. However, since 4ae8d9aa9f9d ("sched/deadline:
Fix dl_server getting stuck") removes dl_server_stopped() entirely, the
end result is the same whether 4717432dfd99 is included or not.

That said, including it is probably better :)

> 
> Could you send a v2 that includes both follow-ups on top of the
> existing three? Then this should be safe to apply as a whole.

Yes, I will do that.

Thanks
Lukas