Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Tue, 13 Oct 2020 16:20:42 +0100
From:   Quentin Perret <qperret@google.com>
To:     "zhuguangqing83" <zhuguangqing83@gmail.com>
Cc:     <lukasz.luba@arm.com>, <rjw@rjwysocki.net>, <pavel@ucw.cz>,
        <len.brown@intel.com>, <linux-pm@vger.kernel.org>,
        <linux-kernel@vger.kernel.org>,
        "'zhuguangqing'" <zhuguangqing@xiaomi.com>
Subject: Re: [PATCH] PM / EM: consult something about cpumask in
 em_dev_register_perf_domain
Message-ID: <20201013152042.GA183119@google.com>
References: <098a01d6a112$9cd7f410$d687dc30$@gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <098a01d6a112$9cd7f410$d687dc30$@gmail.com>
Precedence: bulk

On Tuesday 13 Oct 2020 at 11:40:31 (+0800), zhuguangqing83 wrote:
> I did not observe anything wrong for my use-case. But I think it's possible
> in
> theory that cpu_dev maybe NULL. I observe that in the function
> scmi_cpufreq_init(), before calling em_dev_register_perf_domain(),
> 'policy->cpus' can be ensure that all the cpu_dev in CPU mask are not NULL.
> But maybe we can not ensure all the clients do the check.  This could happen
> if the arch did not set up cpu_dev since this CPU is not in cpu_present mask
> and the driver did not send a correct CPU mask during registration.

Admittedly this has only been tested on Arm64 where I think we can
safely assume that all possible CPUs have been registered at once -- see
topology_init().

And for allowing to register CPUs late in a perf domain, I'm not opposed
to it in principle but that has deep implications as the existing EM
users (e.g. EAS) currently hard rely on it being static after
registration. If you have a real need for it and a patch that adds the
feature and fixes all the users I'll be happy to look at it :)

Thanks,
Quentin