linux-kernel - How can I make cgroup cpu quota and task priority work together without cpuset?

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [day] [month] [year] [list]

Message-ID: <42b40adf.4dfd.18413d75a0e.Coremail.00107082@163.com>
Date:   Wed, 26 Oct 2022 18:29:51 +0800 (CST)
From:   王志强 <00107082@....com>
To:     linux-kernel@...r.kernel.org
Subject: How can I make cgroup cpu quota and task priority work together
 without cpuset?

Hi~!

I am not sure whether this has been raised before:
I have two tasks with different priorities, and I set those two task into
 a cpu cgroup with cpu_quota equal to cpu_period, hence the two tasks 
would only use up 1 cpu core at most,  and the node has multi-cores.
But it turn out cpu cgroup dose not take process priority into 
consideration, what I got is that when the task with higher priority 
need cpu resources, it would only take 50% if the task with lower 
priority is scheduled on another cpu, while the expectation would be >90%.
The problem could be solved if I assign a cpu core statically to those two tasks via cpuset cgroup.....
Is there other solution for this?  

If those two tasks would be scheduled on same cpu anytime, not necessarily
 on the same cpu during each scheduling period, it would be enough.
I tried PR_SCHED_CORE prctl, but no luck.

I think this issue is typical when moving from VM to container,  an API 
in kernel which would bind a group of task together during scheduling 
would be nice, or better if cpu cgroup could take priority into account?

Thanks
David