Re: [PATCH v2 00/16] sched/deadline: first bandwidth accounting fixes and selftests
From: Mathieu Desnoyers
Date: Thu Nov 19 2026 - 08:41:58 EST
Hi Juri,
On 2026-11-19 05:20, Juri Lelli wrote:
> Thanks for splitting this out. I ran v2 in 8-CPU guests with lockdep and
> proxy exec on.
[...]
> I also covered two of the "not tested" items:
Thanks for these runs. In v3, I will drop cgroup v1 and sched_ext from
"Not tested", and say that you tested them.
> - cgroup v1 cpusets (two exclusive cpusets, root load balancing off):
> admission, affinity, refused moves and hotplug all behave as on v2.
> On both v1 and v2, moving a task A -> B -> A leaves its bandwidth
> in B, so A is under-counted once the task dies. The leak is on rc2
> too and the full v7 fixes it, so it is not a regression.
OK. The second series keeps that fix.
> Pointed at ext_server, server_params,
> server_rebuilds, server_states and server_reclaim pass. hotplug_server
> still fails there, as scx_simple gets ejected when the CPU goes
> offline (I haven't dug further). Could the test pick the server that
> is active?
Yes. The server cases will use ext_server while a sched_ext scheduler
is loaded. hotplug_server will skip then.
> Please move this into the changelog of patch 10 itself, as the cover
> won't end up in git. IIUC, the owner then sits in the patch 8 bail-out
> with a deadline of "now" until it unlocks. Other deadline tasks keep
> their bandwidth, but lose latency, and RT/fair tasks on that CPU get
> possibly starved.
Yes, that is what the test sees: each refill gives the owner no runtime
and a deadline of about now. I will move the paragraph into patch 10,
with your impact.
> Doesn't your separate "Fix an overflow of the 0-lag time with long
> periods" also fits here, right before patch 10, which adds a caller of
> that computation?
Yes. I will move it before patch 10 in v3.
> Patch 5: yes, keep my Reviewed-by on the version with the cpu_active()
> check.
Thanks.
Mathieu
--
Mathieu Desnoyers
EfficiOS Inc.
https://www.efficios.com