[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20250626061637.GF17401@unreal>
Date: Thu, 26 Jun 2025 09:16:37 +0300
From: Leon Romanovsky <leon@...nel.org>
To: K Prateek Nayak <kprateek.nayak@....com>
Cc: Ingo Molnar <mingo@...hat.com>, Peter Zijlstra <peterz@...radead.org>,
Juri Lelli <juri.lelli@...hat.com>,
Vincent Guittot <vincent.guittot@...aro.org>,
linux-kernel@...r.kernel.org, Steve Wahl <steve.wahl@....com>,
Dietmar Eggemann <dietmar.eggemann@....com>,
Steven Rostedt <rostedt@...dmis.org>,
Ben Segall <bsegall@...gle.com>, Mel Gorman <mgorman@...e.de>,
Valentin Schneider <vschneid@...hat.com>,
Madadi Vineeth Reddy <vineethr@...ux.ibm.com>
Subject: Re: [PATCH] sched/topology: Correct "sched_domains_curr_level" in
topology_span_sane()
On Tue, Jun 24, 2025 at 04:12:35AM +0000, K Prateek Nayak wrote:
> The updated topology_span_sane() algorithm in commit ce29a7da84cd
> ("sched/topology: Refinement to topology_span_sane speedup") works on
> the "sched_domain_topology_level" hierarchy to detect overlap in
> !SDTL_OVERLAP domains using the tl->mask() as opposed to the
> sched_domain hierarchy (and the sched_domain_span()) in the previous
> approach.
>
> For NODE domain, the cpumask retunred by tl->mask() depends on the
> "sched_domains_curr_level". Unless the "sched_domains_curr_level" is
> reset during topology_span_sane(), it reflects the "numa_level" of the
> last sched_domain created which is incorrect.
>
> Reset the "sched_domains_curr_level" to the "tl->numa_level" during
> topology_span_sane(). Although setting "topology_span_sane" to 0 in
> topology_span_sane() should be enough since all domains with
> numa_level > 0 currently set SDTL_OVERLAP flag, setting it to
> "tl->numa_level" makes it more explicit and future proof against changes
> in the same area.
<...>
> Changes are based on tip:sched/urgent at commit 914873bc7df9 ("Merge tag
> 'x86-build-2025-05-25' of
> git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip")
> ---
> kernel/sched/topology.c | 9 +++++++++
> 1 file changed, 9 insertions(+)
Sorry for the delay, we had internal issues with our verification cloud :(.
Thanks
Powered by blists - more mailing lists