Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <17e804ea-64b5-97e1-d5fa-571157e15746@redhat.com>
Date:   Sat, 25 Mar 2023 18:08:18 -0400
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.7.1
Subject: Re: [PATCH 3/5] cgroup/cpuset: Find another usable CPU if none found
 in current cpuset
Content-Language: en-US
To:     =?UTF-8?Q?Michal_Koutn=c3=bd?= <mkoutny@suse.com>,
        Will Deacon <will@kernel.org>
Cc:     Tejun Heo <tj@kernel.org>, Zefan Li <lizefan.x@bytedance.com>,
        Johannes Weiner <hannes@cmpxchg.org>,
        Shuah Khan <shuah@kernel.org>, cgroups@vger.kernel.org,
        linux-kernel@vger.kernel.org, linux-kselftest@vger.kernel.org,
        Peter Zijlstra <peterz@infradead.org>
References: <20230306200849.376804-1-longman@redhat.com>
 <20230306200849.376804-4-longman@redhat.com>
 <20230314181749.5b4k6selbgdhl3up@blackpad>
 <58a1a878-fa0b-285d-3e43-2b5103d3c770@redhat.com>
 <20230317122708.ax3m2d4zijkfdzjq@blackpad>
 <ca664da8-0f47-06b2-a94c-82b2f9a1c3aa@redhat.com>
 <20230324143247.GA27199@willie-the-truck>
 <20230324181936.5sf6xjc5a4vacuku@blackpad>
From:   Waiman Long <longman@redhat.com>
In-Reply-To: <20230324181936.5sf6xjc5a4vacuku@blackpad>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
Precedence: bulk


On 3/24/23 14:19, Michal Koutný wrote:
> On Fri, Mar 24, 2023 at 02:32:50PM +0000, Will Deacon <will@kernel.org> wrote:
>> So approaches such as killing tasks or rejecting system calls tend not
>> to work as well, since you inevitably get divergent behaviour leading
>> to functional breakage rather than e.g. performance anomalies.
> What about temporary performance drop from 100% to 0% aka freezing the
> tasks for the duration of the mismatching affinity config?

That can be a lot of extra work to freeze it. I will prefer something 
simpler.

Without this patch, I believe it will lead to a cpumask of 0 which will 
cause the scheduler to pick a fallback cpu. It looks like the fallback 
code may be able to pick up the right cpu or it may panic the system 
(less likely).

Cheers,
Longman
>
>
>> Having said that, the behaviour we currently have in mainline seems to
>> be alright, so please don't go out of your way to accomodate these SoCs.
> I see. (Just wondering what you think about the fourth option above.)
>
> Thanks,
> Michal