linux-kernel - Re: [PATCH v1 4/7] sched/isolation: Adjust affinity of managed irqs according to change of housekeeping cpumask

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>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <CADDUTFzSAXm-OUhC8OZ9vL1pYuVybjofUuYa4hWi9y9N0QtCfg@mail.gmail.com>
Date: Tue, 20 Aug 2024 09:21:34 +0300
From: Costa Shulyupin <costa.shul@...hat.com>
To: Thomas Gleixner <tglx@...utronix.de>
Cc: longman@...hat.com, pauld@...hat.com, linux-kernel@...r.kernel.org, 
	Ming Lei <ming.lei@...hat.com>, Christoph Hellwig <hch@....de>, Bart Wensley <bwensley@...hat.com>
Subject: Re: [PATCH v1 4/7] sched/isolation: Adjust affinity of managed irqs
 according to change of housekeeping cpumask

> On Sat, 18 May 2024 at 04:25, Thomas Gleixner <[1]tglx@...utronix.de> wrote:
>
>     Reconfiguring the housekeeping CPUs on a life system is expensive and a
>     slow path operation no matter what.
>
>     So why inflicting all of this nonsense to the kernel instead of
>     cleverly (ab)using CPU hotplug for it in user space.

On Mon, 27 May 2024 at 18:31, Thomas Gleixner <tglx@...utronix.de> wrote:
> That's non-trivial because it's not only the interrupt affinity.
>
> Drivers also have their queues associated accordingly and if you just
> change the interrupt affinity under the hood then all the other
> associations don't get updated and don't work.
>
> That needs to be solved at some other level IMO.

Hello Thomas,

Our telco clients run DPDK in OpenShift/Kubernetes containers.
DPDK requires isolated cpus to run real-time processes.
Kubernetes manages allocation of resources for containers.
Unfortunately Kubernetes doesn't support dynamic CPU offlining/onlining:
https://github.com/kubernetes/kubernetes/issues/67500
and is not planning to support it.
Solving the issue at the application level appears to be even
less straightforward than addressing it at the kernel level.

What would you recommend?

Thanks,
Costa