Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.33 as permitted sender) client-ip=23.128.96.33;
MIME-Version: 1.0
References: <20231122231202.121277-1-andrey.konovalov@linux.dev>
 <CAB=+i9QFeQqSAhwY_BF-DZvZ9TL_rWz7nMOBhDWhXecamsn=dw@mail.gmail.com> <CA+fCnZdp4+2u8a6mhj_SbdmfQ4dWsXBS8O2W3gygzkctekUivw@mail.gmail.com>
In-Reply-To: <CA+fCnZdp4+2u8a6mhj_SbdmfQ4dWsXBS8O2W3gygzkctekUivw@mail.gmail.com>
From:   Hyeonggon Yoo <42.hyeyoo@gmail.com>
Date:   Thu, 23 Nov 2023 11:58:02 +0900
Message-ID: <CAB=+i9RnOz0jDockOfw3oNageCUF5gmF+nzOzPpoTxtr7eqn7g@mail.gmail.com>
Subject: Re: [PATCH mm] slub, kasan: improve interaction of KASAN and
 slub_debug poisoning
To:     Andrey Konovalov <andreyknvl@gmail.com>
Cc:     andrey.konovalov@linux.dev,
        Andrew Morton <akpm@linux-foundation.org>,
        Marco Elver <elver@google.com>,
        Alexander Potapenko <glider@google.com>,
        Dmitry Vyukov <dvyukov@google.com>,
        Vlastimil Babka <vbabka@suse.cz>, kasan-dev@googlegroups.com,
        Evgenii Stepanov <eugenis@google.com>,
        Oscar Salvador <osalvador@suse.de>,
        Feng Tang <feng.tang@intel.com>, linux-mm@kvack.org,
        linux-kernel@vger.kernel.org,
        Andrey Konovalov <andreyknvl@google.com>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Thu, Nov 23, 2023 at 11:31=E2=80=AFAM Andrey Konovalov <andreyknvl@gmail=
.com> wrote:
>
> On Thu, Nov 23, 2023 at 1:39=E2=80=AFAM Hyeonggon Yoo <42.hyeyoo@gmail.co=
m> wrote:
> >
> > On Thu, Nov 23, 2023 at 8:12=E2=80=AFAM <andrey.konovalov@linux.dev> wr=
ote:
> > >
> > > From: Andrey Konovalov <andreyknvl@google.com>
> > >
> > > When both KASAN and slub_debug are enabled, when a free object is bei=
ng
> > > prepared in setup_object, slub_debug poisons the object data before K=
ASAN
> > > initializes its per-object metadata.
> > >
> > > Right now, in setup_object, KASAN only initializes the alloc metadata=
,
> > > which is always stored outside of the object. slub_debug is aware of
> > > this and it skips poisoning and checking that memory area.
> > >
> > > However, with the following patch in this series, KASAN also starts
> > > initializing its free medata in setup_object. As this metadata might =
be
> > > stored within the object, this initialization might overwrite the
> > > slub_debug poisoning. This leads to slub_debug reports.
> > >
> > > Thus, skip checking slub_debug poisoning of the object data area that
> > > overlaps with the in-object KASAN free metadata.
> > >
> > > Also make slub_debug poisoning of tail kmalloc redzones more precise =
when
> > > KASAN is enabled: slub_debug can still poison and check the tail kmal=
loc
> > > allocation area that comes after the KASAN free metadata.
> > >
> > > Signed-off-by: Andrey Konovalov <andreyknvl@google.com>
> >
> > Thank you for looking at this quickly!
> > Unfortunately the problem isn't fixed yet with the patch.
> >
> > I applied this on top of linux-next and built a kernel with the same co=
nfig,
> > it is still stuck at boot.
>
> Ah, this is caused by a buggy version of "kasan: improve free meta
> storage in Generic KASAN", which made its way into linux-next.
> Reverting that patch should fix the issue. My patch that you bisected
> to exposes the buggy behavior.

1. I reverted the commit "kasan: improve free meta storage in Generic KASAN=
",
    on top of linux-next (next-20231122), and it is still stuck at boot.

2. I reverted the commit "kasan: improve free meta storage in Generic KASAN=
",
    on top of linux-next (next-20231122),
   _and_ applied this patch on top of it, now it boots fine!

--
Hyeonggon