linux-kernel - Re: [PATCH] kasan: Prevent cpu_quarantine corruption when CPU offline and cache shrink occur at same time

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]

Date:   Fri, 22 Apr 2022 11:27:58 +0200
From:   Dmitry Vyukov <dvyukov@...gle.com>
To:     Andrew Morton <akpm@...ux-foundation.org>
Cc:     Zqiang <qiang1.zhang@...el.com>, ryabinin.a.a@...il.com,
        linux-mm@...ck.org, linux-kernel@...r.kernel.org,
        Alexander Potapenko <glider@...gle.com>,
        Andrey Konovalov <andreyknvl@...il.com>,
        kasan-dev <kasan-dev@...glegroups.com>
Subject: Re: [PATCH] kasan: Prevent cpu_quarantine corruption when CPU offline
 and cache shrink occur at same time

On Fri, 22 Apr 2022 at 00:07, Andrew Morton <akpm@...ux-foundation.org> wrote:
>
> On Thu, 14 Apr 2022 10:59:25 +0800 Zqiang <qiang1.zhang@...el.com> wrote:
>
> > The kasan_quarantine_remove_cache() is called in kmem_cache_shrink()/
> > destroy(), the kasan_quarantine_remove_cache() call is protected by
> > cpuslock in kmem_cache_destroy(), can ensure serialization with
> > kasan_cpu_offline(). however the kasan_quarantine_remove_cache() call
> > is not protected by cpuslock in kmem_cache_shrink(), when CPU going
> > offline and cache shrink occur at same time, the cpu_quarantine may be
> > corrupted by interrupt(per_cpu_remove_cache operation). so add
> > cpu_quarantine offline flags check in per_cpu_remove_cache().
> >
> > ...
> >
>
> Could we please have some reviewer input here?

This is very tricky, I think can follow this:

Reviewed-by: Dmitry Vyukov <dvyukov@...gle.com>

If q->offline is set, then kasan_cpu_offline() will or has already
removed everything from cpu_quarantine and freed, so we can return
early in per_cpu_remove_cache().
If kasan_cpu_offline() hasn't yet removed everything from
cpu_quarantine already, it's actually problematic for the
kmem_cache_destroy() case. But since both kmem_cache_destroy() and
kasan_cpu_offline() are serialized by cpus lock, this case must not
happen.




> > --- a/mm/kasan/quarantine.c
> > +++ b/mm/kasan/quarantine.c
> > @@ -330,6 +330,8 @@ static void per_cpu_remove_cache(void *arg)
> >       struct cpu_shrink_qlist *sq;
> >  #endif
> >       q = this_cpu_ptr(&cpu_quarantine);
> > +     if (READ_ONCE(q->offline))
> > +             return;
> >  #ifndef CONFIG_PREEMPT_RT
> >       qlist_move_cache(q, &to_free, cache);
> >       qlist_free_all(&to_free, cache);
>
> It might be helpful to have a little comment which explains why we're
> doing this?