linux-kernel - Re: [PATCH 1/7] kthread: Don't allocate kthread

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>] [day] [month] [year] [list]

Message-ID: <8735hm1iz6.fsf@email.froward.int.ebiederm.org>
Date:   Fri, 06 May 2022 15:53:01 -0500
From:   "Eric W. Biederman" <ebiederm@...ssion.com>
To:     chill <maximkabox13@...il.com>
Cc:     linux-arch@...r.kernel.org, Tejun Heo <tj@...nel.org>,
        Peter Zijlstra <peterz@...radead.org>,
        Vincent Guittot <vincent.guittot@...aro.org>,
        Al Viro <viro@...iv.linux.org.uk>,
        Jens Axboe <axboe@...nel.dk>,
        Thomas Gleixner <tglx@...utronix.de>,
        Linus Torvalds <torvalds@...uxfoundation.org>,
        linux-kernel@...r.kernel.org, stable@...r.kernel.org
Subject: Re: [PATCH 1/7] kthread: Don't allocate kthread_struct for init and
 umh

chill <maximkabox13@...il.com> writes:

> this looks like a real uaf vulnerability and can be executed by the user

The potential to use memory after it has been freed appears completely
real.  As such it is a bug and it should definitely be fixed.  That is
as far as I can see.

What I don't see, and I am very bad at this so I could be missing
something, is what bad thing kthread_is_per_cpu could be tricked into
doing.

I see a window of a single instruction which reads a single bit
that normally will return false.  If that bit instead reads true
it looks like the scheduler will simply decide to not run the
process on another cpu.

So I will put this change in linux-next.  It will be tested and I will
send it to Linus when the merge window for v5.19 opens.  After Linus
merges this I expect after a week or so it will be backported to the
various stable kernels.  Not that it needs to go farther than about
v5.17 where I introduced the bug.

Eric