From: Yury Norov <ynorov@nvidia.com>
To: Ingo Molnar <mingo@redhat.com>,
Peter Zijlstra <peterz@infradead.org>,
Juri Lelli <juri.lelli@redhat.com>,
Vincent Guittot <vincent.guittot@linaro.org>,
Dietmar Eggemann <dietmar.eggemann@arm.com>,
Steven Rostedt <rostedt@goodmis.org>,
Ben Segall <bsegall@google.com>, Mel Gorman <mgorman@suse.de>,
Valentin Schneider <vschneid@redhat.com>,
K Prateek Nayak <kprateek.nayak@amd.com>,
linux-kernel@vger.kernel.org
Cc: Yury Norov <yury.norov@gmail.com>
Subject: Re: [PATCH] sched/core: Make fallback CPU selection NUMA-aware
Date: Wed, 18 Nov 2026 16:13:21 -0500 [thread overview]
Message-ID: <av4U8VYR-NGq2F19@yury> (raw)
In-Reply-To: <arFYBf58i6J35SLZ@yury>
Ping again?
On Mon, Sep 21, 2026 at 12:15:01PM -0400, Yury Norov wrote:
> Ping?
>
> On Fri, Sep 04, 2026 at 11:36:56PM -0400, Yury Norov wrote:
> > select_fallback_rq() checks the local node first, but then scans the task's
> > affinity mask in a numerical order. On systems with more than two NUMA
> > nodes, that can select a CPU farther away than necessary.
> >
> > Walk the scheduler's NUMA hop masks and consider each newly reached CPU
> > once, preserving locality across the full fallback search and after
> > affinity relaxation.
> >
> > The NUMA masks can be unavailable or incomplete while topology is rebuilt.
> > Search any online CPUs not covered by the masks before relaxing affinity
> > so fallback selection remains reliable during that window.
> >
> > Signed-off-by: Yury Norov <ynorov@nvidia.com>
> > ---
> > kernel/sched/core.c | 47 ++++++++++++++++++++++++---------------------
> > 1 file changed, 25 insertions(+), 22 deletions(-)
> >
> > diff --git a/kernel/sched/core.c b/kernel/sched/core.c
> > index f78275192036..2155bad0e1ab 100644
> > --- a/kernel/sched/core.c
> > +++ b/kernel/sched/core.c
> > @@ -3554,34 +3554,37 @@ EXPORT_SYMBOL_GPL(kick_process);
> > */
> > static int select_fallback_rq(int cpu, struct task_struct *p)
> > {
> > - int nid = cpu_to_node(cpu);
> > - const struct cpumask *nodemask = NULL;
> > + int nid = IS_ENABLED(CONFIG_NUMA) ? cpu_to_node(cpu) : NUMA_NO_NODE;
> > enum { cpuset, possible, fail } state = cpuset;
> > int dest_cpu;
> >
> > - /*
> > - * If the node that the CPU is on has been offlined, cpu_to_node()
> > - * will return -1. There is no CPU on the node, and we should
> > - * select the CPU on the other node.
> > - */
> > - if (nid != -1) {
> > - nodemask = cpumask_of_node(nid);
> > -
> > - /* Look for allowed, online CPU in same node. */
> > - for_each_cpu(dest_cpu, nodemask) {
> > - if (is_cpu_allowed(p, dest_cpu))
> > - return dest_cpu;
> > - }
> > - }
> > -
> > for (;;) {
> > - /* Any allowed, online CPU? */
> > - for_each_cpu(dest_cpu, p->cpus_ptr) {
> > - if (!is_cpu_allowed(p, dest_cpu))
> > - continue;
> > + const struct cpumask *prev = cpu_none_mask, *cpus;
> > +
> > + /* Look for the closest allowed, online CPU. */
> > + rcu_read_lock();
> > + for_each_numa_hop_mask(cpus, nid) {
> > + for_each_cpu_andnot(dest_cpu, cpus, prev) {
> > + if (is_cpu_allowed(p, dest_cpu)) {
> > + rcu_read_unlock();
> > + goto out;
> > + }
> > + }
> > + prev = cpus;
> > + }
> >
> > - goto out;
> > + /*
> > + * NUMA masks may be unavailable or incomplete while the
> > + * topology is being rebuilt. Search CPUs not covered by them
> > + * before relaxing the task's affinity.
> > + */
> > + for_each_cpu_andnot(dest_cpu, p->cpus_ptr, prev) {
> > + if (is_cpu_allowed(p, dest_cpu)) {
> > + rcu_read_unlock();
> > + goto out;
> > + }
> > }
> > + rcu_read_unlock();
> >
> > /* No more Mr. Nice Guy. */
> > switch (state) {
> > --
> > 2.53.0
prev parent reply other threads:[~2026-11-18 21:13 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-09-05 3:36 [PATCH] sched/core: Make fallback CPU selection NUMA-aware Yury Norov
2026-09-21 16:15 ` Yury Norov
2026-11-18 21:13 ` Yury Norov [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=av4U8VYR-NGq2F19@yury \
--to=ynorov@nvidia.com \
--cc=bsegall@google.com \
--cc=dietmar.eggemann@arm.com \
--cc=juri.lelli@redhat.com \
--cc=kprateek.nayak@amd.com \
--cc=linux-kernel@vger.kernel.org \
--cc=mgorman@suse.de \
--cc=mingo@redhat.com \
--cc=peterz@infradead.org \
--cc=rostedt@goodmis.org \
--cc=vincent.guittot@linaro.org \
--cc=vschneid@redhat.com \
--cc=yury.norov@gmail.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox