linux-kernel - Re: Odd-sized kmem_cache_alloc and slub

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: <202010081608.E1401C067@keescook>
Date:   Thu, 8 Oct 2020 16:10:36 -0700
From:   Kees Cook <keescook@...omium.org>
To:     Marco Elver <elver@...gle.com>
Cc:     Pekka Enberg <penberg@...il.com>,
        Alexander Potapenko <glider@...gle.com>,
        Andrew Morton <akpm@...ux-foundation.org>,
        David Rientjes <rientjes@...gle.com>,
        Joonsoo Kim <iamjoonsoo.kim@....com>,
        Christoph Lameter <cl@...ux.com>,
        kasan-dev <kasan-dev@...glegroups.com>,
        LKML <linux-kernel@...r.kernel.org>,
        "linux-mm@...ck.org" <linux-mm@...ck.org>
Subject: Re: Odd-sized kmem_cache_alloc and slub_debug=Z

On Mon, Aug 17, 2020 at 08:31:35PM +0200, Marco Elver wrote:
> On Fri, 7 Aug 2020 at 21:06, Pekka Enberg <penberg@...il.com> wrote:
> ...
> > Yeah, it reproduces with defconfig too, as long as you remember to
> > pass "slub_debug=Z"... :-/
> >
> > The following seems to be the culprit:
> >
> > commit 3202fa62fb43087387c65bfa9c100feffac74aa6
> > Author: Kees Cook <keescook@...omium.org>
> > Date:   Wed Apr 1 21:04:27 2020 -0700
> >
> >     slub: relocate freelist pointer to middle of object
> >
> > Reverting this commit and one of it's follow up fixes from Kees from
> > v5.8 makes the issue go away for me. Btw, please note that caches with
> > size 24 and larger do not trigger this bug, so the issue is that with
> > small enough object size, we're stomping on allocator metadata (I
> > assume part of the freelist).
> 
> Was there a patch to fix this? Checking, just in case I missed it.

Hi! I've finally carved out time to look at this, and I've figured it
out. My prior redzoning fix was actually wrong; I'll send a patch to
fix it harder. In the meantime, I've also discovered that the redzoning
code wasn't safe for sizes smaller than sizeof(void *). This oversight
is, I think, what caused me to incorrectly handle the changed offset the
first time. :P

Anyway, patches incoming...

-- 
Kees Cook