linux-kernel - Re: [RFC PATCH] locking/percpu-rwsem: use this_cpu_{inc|dec}() for read

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

Message-ID: <20200915181112.GE2674@hirez.programming.kicks-ass.net>
Date:   Tue, 15 Sep 2020 20:11:12 +0200
From:   peterz@...radead.org
To:     Will Deacon <will@...nel.org>
Cc:     Oleg Nesterov <oleg@...hat.com>, Hou Tao <houtao1@...wei.com>,
        Ingo Molnar <mingo@...hat.com>,
        Dennis Zhou <dennis@...nel.org>, Tejun Heo <tj@...nel.org>,
        Christoph Lameter <cl@...ux.com>, linux-kernel@...r.kernel.org,
        linux-fsdevel@...r.kernel.org, Jan Kara <jack@...e.cz>
Subject: Re: [RFC PATCH] locking/percpu-rwsem: use this_cpu_{inc|dec}() for
 read_count

On Tue, Sep 15, 2020 at 05:11:23PM +0100, Will Deacon wrote:
> On Tue, Sep 15, 2020 at 06:03:44PM +0200, peterz@...radead.org wrote:
> > On Tue, Sep 15, 2020 at 05:51:50PM +0200, peterz@...radead.org wrote:
> > 
> > > Anyway, I'll rewrite the Changelog and stuff it in locking/urgent.
> > 
> > How's this?
> > 
> > ---
> > Subject: locking/percpu-rwsem: Use this_cpu_{inc,dec}() for read_count
> > From: Hou Tao <houtao1@...wei.com>
> > Date: Tue, 15 Sep 2020 22:07:50 +0800
> > 
> > From: Hou Tao <houtao1@...wei.com>
> > 
> > The __this_cpu*() accessors are (in general) IRQ-unsafe which, given
> > that percpu-rwsem is a blocking primitive, should be just fine.
> > 
> > However, file_end_write() is used from IRQ context and will cause
> > load-store issues.
> 
> ... on architectures where the per-cpu accessors are not atomic.

That's not entirely accurate, on x86 for example the per-cpu ops are not
atomic, but they are not susceptible to this problem due to them being a
single instruction from the point of interrupts -- either they wholly
happen or they don't.

So I'd reformulate it like: "... on architectures where the per-cpu
accessors are not natively irq-safe" ?