Received-SPF: pass (google.com: domain of linux-kernel+bounces-179154-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45d1:ec00::1 as permitted sender) client-ip=2604:1380:45d1:ec00::1;
From: Thomas Gleixner <tglx@linutronix.de>
To: Yury Norov <yury.norov@gmail.com>
Cc: linux-kernel@vger.kernel.org, Greg Kroah-Hartman
 <gregkh@linuxfoundation.org>, "Paul E. McKenney" <paulmck@kernel.org>,
 "Rafael J. Wysocki" <rafael@kernel.org>, Anna-Maria Behnsen
 <anna-maria@linutronix.de>, Ben Segall <bsegall@google.com>, Daniel
 Bristot de Oliveira <bristot@redhat.com>, Dietmar Eggemann
 <dietmar.eggemann@arm.com>, Frederic Weisbecker <frederic@kernel.org>,
 Imran Khan <imran.f.khan@oracle.com>, Ingo Molnar <mingo@redhat.com>,
 Johannes Weiner <hannes@cmpxchg.org>, Juri Lelli <juri.lelli@redhat.com>,
 Leonardo Bras <leobras@redhat.com>, Mel Gorman <mgorman@suse.de>, Peter
 Zijlstra <peterz@infradead.org>, Rik van Riel <riel@surriel.com>, Steven
 Rostedt <rostedt@goodmis.org>, Tejun Heo <tj@kernel.org>, Valentin
 Schneider <vschneid@redhat.com>, Vincent Guittot
 <vincent.guittot@linaro.org>, Waiman Long <longman@redhat.com>, Zefan Li
 <lizefan.x@bytedance.com>, cgroups@vger.kernel.org
Subject: Re: [PATCH 6/6] tick/common: optimize cpumask_equal() usage
In-Reply-To: <CAAH8bW-o_zz_C_NFnjL3uP1BXyC4OF-BAR2Dk2Xd-DFDOZpodQ@mail.gmail.com>
References: <20240513220146.1461457-1-yury.norov@gmail.com>
 <20240513220146.1461457-7-yury.norov@gmail.com> <878r0cn6a5.ffs@tglx>
 <874jb0n5rt.ffs@tglx>
 <CAAH8bW-o_zz_C_NFnjL3uP1BXyC4OF-BAR2Dk2Xd-DFDOZpodQ@mail.gmail.com>
Date: Tue, 14 May 2024 22:47:57 +0200
Message-ID: <87msosktmq.ffs@tglx>
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain

On Tue, May 14 2024 at 09:47, Yury Norov wrote:
>> Instead of sprinkling these conditional all over the place, can't you
>> just do the obvious and check for ptr1 == ptr2 in bitmap_copy() and
>> bitmap_equal()?
>
> I proposed this a while (few years) ago, and it has been rejected. On
> bitmaps level we decided not to do that for the reasons memcpy() and
> memcmp() doesn't, and on cpumasks and nodemasks level it hasn't
> been discussed at all.
>
> Now that most of bitmap ops have inline and outline implementation,
> we technically can move this checks in outline code, as inline bitmap
> ops are very lightweight already.
>
> So I see the following options:
>  - Implement these sanity checks in outline bitmap API (lib/bitmap.c);
>  - Implement them on cpumask and nodemask level; or
>  - add a new family of helpers that do this check, like
>   bitmap_copy_if_needed() (better name appreciated).
>
> The argument against #1 and #2 these days was that memcpy() and
> similarly bitmap_copy() with dst == src may be a sign of error, and
> we don't want to add a code that optimizes for it.

That's a fair argument.

> Now, I ran the kernel through the LTP test and in practice all the
> cases that I spot look pretty normal. So I can continue sprinkling
> the checks once a few years, or do something like described above.

I don't see these checks as valuable in most cases and I detest them as
they make the code harder to read.

Except for smp_call_function_many_cond() and to a lesser extent
irq_do_set_affinity() none of them you added really matters.

Though it might be worth to have helper functions which make it obvious
that the src == dst case is intentional.

Thanks,

        tglx