[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <ze4cjlj36ads7d6gpbpxryqxcnrguh5etokbmlrrzh3uxw4hb4@mcfszlpsozuv>
Date: Sun, 9 Nov 2025 20:36:54 -0500
From: Aaron Tomlin <atomlin@...mlin.com>
To: mingo@...hat.com, peterz@...radead.org, juri.lelli@...hat.com,
vincent.guittot@...aro.org, dietmar.eggemann@....com, rostedt@...dmis.org,
bsegall@...gle.com, mgorman@...e.de, vschneid@...hat.com
Cc: pauld@...hat.com, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] sched/isolation: Enforce at least one housekeeping CPU
per node unless maxcpus limits
On Sat, Oct 25, 2025 at 12:11:13PM -0400, Aaron Tomlin wrote:
> This patch improves housekeeping CPU selection logic by enforcing that
> each online NUMA node has at least one dedicated housekeeping CPU,
> ensuring better NUMA locality for kernel threads and timed work.
>
> Before assigning additional housekeeping CPUs, the patch checks if any
> online NUMA node contains CPUs with logical IDs greater than or equal to
> max_cpus=. If so, per-node NUMA enforcement is skipped and a
> warning is issued, since some nodes would be unserviceable given the CPU
> limit.
>
> If NUMA enforcement is possible, each online node lacking a housekeeping
> CPU will have one present CPU (the lowest logical ID) assigned and
> included in the housekeeping staging mask, with a warning logged for
> visibility. The final guarantee that at least one present housekeeping
> CPU is assigned across the system remains intact.
Hi,
Just following up on this patch submission.
Please let me know if it requires any further revision or if there is any
feedback I should address.
Kind regards,
--
Aaron Tomlin
Powered by blists - more mailing lists