linux-kernel - Re: [PATCH] refcount: introduce refcount_is

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]

Date:   Thu, 18 Nov 2021 11:49:39 +0100
From:   Peter Zijlstra <peterz@...radead.org>
To:     yajun.deng@...ux.dev
Cc:     will@...nel.org, boqun.feng@...il.com, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] refcount: introduce refcount_is_one() helper function

On Thu, Nov 18, 2021 at 10:34:44AM +0000, yajun.deng@...ux.dev wrote:
> November 18, 2021 4:44 PM, "Peter Zijlstra" <peterz@...radead.org> wrote:
> 
> > On Thu, Nov 18, 2021 at 08:12:56AM +0000, yajun.deng@...ux.dev wrote:
> > 
> >> November 18, 2021 3:42 PM, "Peter Zijlstra" <peterz@...radead.org> wrote:
> >> 
> >> On Thu, Nov 18, 2021 at 11:53:28AM +0800, Yajun Deng wrote:
> >> 
> >> There are many cases where it is necessary to determine if refcount is one,
> >> introduce refcount_is_one() helper function for these cases.
> >> 
> >> Give me one that is not racy?
> >> 
> >> The following determine refcount is one, 35 count.
> > 
> > Very good, now get me one that isn't broken :-)
> 
> Sorry, I didn't understand what is the 'isn't broken'。

What's the value of refcount_read() given that at any moment a
concurrent refcount_{inc,dec}() can happen?

If you can't know the current value (per the above) then what's the
value of knowing it was one some time ago?

Fundamentally using refcount_read() in control flow is broken, it's a
very bad anti-pattern.