[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <D0D163D9-640F-4881-93C4-CCE40B5701F9@linux.alibaba.com>
Date: Tue, 10 Sep 2024 18:05:22 +0800
From: 刘嵩 <liusong@...ux.alibaba.com>
To: Michal Koutný <mkoutny@...e.com>
Cc: tj@...nel.org,
lizefan.x@...edance.com,
hannes@...xchg.org,
cgroups@...r.kernel.org,
linux-kernel@...r.kernel.org
Subject: Re: [RFC PATCH] sched, cgroup: cgroup1 can also take the
non-RUNTIME_INF min
> 2024年9月10日 17:35,Michal Koutný <mkoutny@...e.com> 写道:
>
> Hello.
>
> On Tue, Sep 10, 2024 at 03:48:32PM GMT, Liu Song <liusong@...ux.alibaba.com> wrote:
>> For the handling logic of child_quota, there is no need to distinguish
>> between cgroup1 and cgroup2, so unify the handling logic here.
>
> IIUC, v1 check prevents quota overcommit while v2 doesn't. So this
> isn't user invisible change (i.e. there may be a need to distinguish the
> two).
>
Hi
In cgroup2, child task group quota can set exceed parent task group quota.
However, at the scheduling level, the cpu runqueue (rq) performs checks
on the runtime of the CFS run queue (cfs_rq) in a hierarchical way.
Therefore, even though the child group quota appears to be greater than the
parent group quota, it will not actually receive a runtime that exceeds the
parent's quota. This logic also holds true for cgroup1, so there is no need to
differentiate the handling logic in this case.
Thanks
> Regards,
> Michal
Powered by blists - more mailing lists