[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <c6bd487d-1436-66f7-ed2a-8679303ad9b5@redhat.com>
Date: Mon, 6 Feb 2023 11:50:11 -0500
From: Waiman Long <longman@...hat.com>
To: Michal Koutný <mkoutny@...e.com>
Cc: Frederic Weisbecker <frederic@...nel.org>,
LKML <linux-kernel@...r.kernel.org>,
Roman Gushchin <roman.gushchin@...ux.dev>,
Michal Hocko <mhocko@...e.com>,
Marcelo Tosatti <mtosatti@...hat.com>,
Leonardo <leobras@...hat.com>,
Johannes Weiner <hannes@...xchg.org>,
Shakeel Butt <shakeelb@...gle.com>,
Muchun Song <muchun.song@...ux.dev>,
Andrew Morton <akpm@...ux-foundation.org>,
Peter Zijlstra <peterz@...radead.org>,
Thomas Gleixner <tglx@...utronix.de>, cgroups@...r.kernel.org,
linux-mm@...ck.org
Subject: Re: [PATCH 2/2] sched/isolation: Add cpu_is_isolated() API
On 2/6/23 10:47, Michal Koutný wrote:
> Hello.
>
> On Fri, Feb 03, 2023 at 10:53:46PM -0500, Waiman Long <longman@...hat.com> wrote:
>> CPUs in an isolated cpuset partition is similar to HK_TYPE_DOMAIN CPUs as
>> load balancing is disabled. I can add an API to access the cpumask and add
>> to this API. However, that list is dynamic as it can be changed at run time.
>> Will that be a problem?
> I can see a problem already -- as a CPU can be dynamically switched to
> "isolated" mode so should all dependent operations support that (switch)
> too, i.e. the CPUs local PCP caches would have to be drained when the
> CPU enters isolation.
I see the long term goal is to have more isolation capability to be done
dynamically. However, we are not there yet. There is still a lot of work
to do to achieve that.
>
>> Or should that be used separately?
> It'd be nice to have both (cpuset and cmdline flags) eventually unified.
>
> Alas, it only leads me conservatively to:
>
> #ifndef CONFIG_CPUSETS
> // the proposed implementaion
> else
> static inline bool cpu_is_isolated(int cpu) {
> return true;
> }
> #endif
That is too conservative from my point of view. We can have further
discussion when a patch is ready.
Cheers,
Longman
Powered by blists - more mailing lists