linux-kernel - Re: [PATCH 2/3] futex: Use RCU-based per-CPU reference counting instead of rcuref

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <103a68e9-9627-42ad-ae86-3b023c3230d7@redhat.com>
Date: Thu, 10 Jul 2025 11:58:50 -0400
From: Waiman Long <llong@...hat.com>
To: Peter Zijlstra <peterz@...radead.org>,
 Sebastian Andrzej Siewior <bigeasy@...utronix.de>
Cc: Waiman Long <llong@...hat.com>, linux-kernel@...r.kernel.org,
 Andr?? Almeida <andrealmeid@...lia.com>, Darren Hart <dvhart@...radead.org>,
 Davidlohr Bueso <dave@...olabs.net>, Ingo Molnar <mingo@...hat.com>,
 Juri Lelli <juri.lelli@...hat.com>, Thomas Gleixner <tglx@...utronix.de>,
 Valentin Schneider <vschneid@...hat.com>
Subject: Re: [PATCH 2/3] futex: Use RCU-based per-CPU reference counting
 instead of rcuref_t

On 7/8/25 3:06 PM, Peter Zijlstra wrote:
> On Tue, Jul 08, 2025 at 03:47:08PM +0200, Sebastian Andrzej Siewior wrote:
>> On 2025-07-08 09:43:44 [-0400], Waiman Long wrote:
>>> This looks somewhat like what the percpu refcount does (see
>>> lib/percpu-refcount.c). Could this be used instead of reinventing the wheel
>>> again?
>>  From the comment:
>>
>>    * futex-ref
>>    *
>>    * Heavily inspired by percpu-rwsem/percpu-refcount; not reusing any of that
>>    * code because it just doesn't fit right.
>>    *
>>    * Dual counter, per-cpu / atomic approach like percpu-refcount, except it
>>    * re-initializes the state automatically, such that the fph swizzle is also a
>>    * transition back to per-cpu.
>>
>> but I leave it up to Peter if he considers merging that.
> Basically what the comment says. Trying to reuse things ended up in a
> mess. It really isn't much code, most of it is comments.
>
I got it now. I am not against adding a variant specific to this code 
giving that we want to fix the performance regression ASAP. Merging it 
to any existing set of helpers may be something we want to do in the future.

Cheers,
Longman