Re: [PATCH v8 16/33] sched/deadline: Keep the last CPU a task has in its root domain
From: Mathieu Desnoyers
Date: Fri Nov 20 2026 - 08:51:15 EST
> Does cpuset_cpus_allowed_without() leave pmask empty when from reaches
> top_cpuset and @cpu was the last active CPU in top_cpuset.effective_cpus?
Yes, by reading. It needs isolcpus=domain, with every housekeeping CPU
in a partition. When the top cpuset loses its last CPU, v9 will check
every cpuset, the top one included. v9 will not have a test of this case
yet.
> Can this loop walk past an unpopulated valid parent partition root that
> stays valid with an empty effective_cpus mask during hotplug?
Maybe. I have not traced it in full. partition_child_lost() does return
NULL at p_outer, and the walk then goes up to the top cpuset. I have not
checked how compute_effective_cpumask() treats p_outer. v9 will not
change this path. I will look at it in a follow-up, with a test.
> Can an offline that invalidates an isolated partition (PRS_ISOLATED) still
> move a deadline task from def_root_domain to a housekeeping root domain
> without being refused?
Yes. HK_TYPE_DOMAIN_BOOT does not know the CPUs of an isolated
partition. v9 will record with each task whether def_root_domain counts
its bandwidth, in a new field, dl_bw_def. The rebuild will keep the
bandwidth on that side. It will look at the root domain of each CPU, not
at the boot mask.
Your first example will be a new case,
hotplug_isolated_partition_falls_back. It fails without the fix, in
guests booted with isolcpus=domain. With the fix, it passes. Your second
example loses the last CPU of the top cpuset: the fix for your first
point covers it.
Thanks,
Mathieu
--
Mathieu Desnoyers
EfficiOS Inc.
https://www.efficios.com