Message-ID: <1615803189.26681.2.camel@mtksdccf07>
Date:   Mon, 15 Mar 2021 18:13:09 +0800
From:   Walter Wu <walter-zh.wu@...iatek.com>
To:     Dmitry Vyukov <dvyukov@...gle.com>
CC:     Andrey Ryabinin <ryabinin.a.a@...il.com>,
        Alexander Potapenko <glider@...gle.com>,
        Matthias Brugger <matthias.bgg@...il.com>,
        "Andrey Konovalov" <andreyknvl@...gle.com>,
        Andrew Morton <akpm@...ux-foundation.org>,
        Jens Axboe <axboe@...nel.dk>, Oleg Nesterov <oleg@...hat.com>,
        kasan-dev <kasan-dev@...glegroups.com>,
        Linux-MM <linux-mm@...ck.org>,
        LKML <linux-kernel@...r.kernel.org>,
        Linux ARM <linux-arm-kernel@...ts.infradead.org>,
        wsd_upstream <wsd_upstream@...iatek.com>,
        <linux-mediatek@...ts.infradead.org>
Subject: Re: [PATCH] task_work: kasan: record task_work_add() call stack

On Mon, 2021-03-15 at 11:03 +0100, 'Dmitry Vyukov' via kasan-dev wrote:
> On Mon, Mar 15, 2021 at 10:38 AM Walter Wu <walter-zh.wu@...iatek.com> wrote:
> >
> > On Mon, 2021-03-15 at 07:58 +0100, 'Dmitry Vyukov' via kasan-dev wrote:
> > > On Mon, Mar 15, 2021 at 3:00 AM Walter Wu <walter-zh.wu@...iatek.com> wrote:
> > > >
> > > > Why record task_work_add() call stack?
> > > > Syzbot reports many use-after-free issues for task_work, see [1].
> > > > After see the free stack and the current auxiliary stack, we think
> > > > they are useless, we don't know where register the work, this work
> > > > may be the free call stack, so that we miss the root cause and
> > > > don't solve the use-after-free.
> > > >
> > > > Add task_work_add() call stack into KASAN auxiliary stack in
> > > > order to improve KASAN report. It is useful for programmers
> > > > to solve use-after-free issues.
> > > >
> > > > [1]: https://groups.google.com/g/syzkaller-bugs/search?q=kasan%20use-after-free%20task_work_run
> > > >
> > > > Signed-off-by: Walter Wu <walter-zh.wu@...iatek.com>
> > > > Suggested-by: Dmitry Vyukov <dvyukov@...gle.com>
> > > > Cc: Andrey Ryabinin <ryabinin.a.a@...il.com>
> > > > Cc: Dmitry Vyukov <dvyukov@...gle.com>
> > > > Cc: Andrey Konovalov <andreyknvl@...gle.com>
> > > > Cc: Alexander Potapenko <glider@...gle.com>
> > > > Cc: Andrew Morton <akpm@...ux-foundation.org>
> > > > Cc: Matthias Brugger <matthias.bgg@...il.com>
> > > > Cc: Jens Axboe <axboe@...nel.dk>
> > > > Cc: Oleg Nesterov <oleg@...hat.com>
> > > > ---
> > > >  kernel/task_work.c | 3 +++
> > > >  mm/kasan/kasan.h   | 2 +-
> > > >  2 files changed, 4 insertions(+), 1 deletion(-)
> > > >
> > > > diff --git a/kernel/task_work.c b/kernel/task_work.c
> > > > index 9cde961875c0..f255294377da 100644
> > > > --- a/kernel/task_work.c
> > > > +++ b/kernel/task_work.c
> > > > @@ -55,6 +55,9 @@ int task_work_add(struct task_struct *task, struct callback_head *work,
> > > >                 break;
> > > >         }
> > > >
> > > > +       /* record the work call stack in order to print it in KASAN reports */
> > > > +       kasan_record_aux_stack(work);
> > >
> > > I think this call should be done _before_ we actually queue the work,
> > > because this function may operate on non-current task.
> > > Consider, we queue the work, the other task already executes it and
> > > triggers use-after-free, now only now we record the stack.
> >
> > agree, what do you think below change?
> >
> > --- a/kernel/task_work.c
> > +++ b/kernel/task_work.c
> > @@ -34,6 +34,9 @@ int task_work_add(struct task_struct *task, struct
> > callback_head *work,
> >  {
> >     struct callback_head *head;
> >
> > +   /* record the work call stack in order to print it in KASAN reports
> > */
> > +   kasan_record_aux_stack(work);
> > +
> 
> This looks good to me.
> 
> 
> >     do {
> >         head = READ_ONCE(task->task_works);
> >         if (unlikely(head == &work_exited))
> > @@ -55,9 +58,6 @@ int task_work_add(struct task_struct *task, struct
> > callback_head *work,
> >         break;
> >     }
> >
> > -   /* record the work call stack in order to print it in KASAN reports
> > */
> > -   kasan_record_aux_stack(work);
> > -
> >     return 0;
> >  }
> >
> > > Moreover, I think we can trigger use-after-free here ourselves while
> > > recording the aux stack. We queued the work, and the work can cause
> > > own free, so it's not necessary live by now.
> >
> > Sorry, I don't fully know your meaning, do you mean we should add an
> > abort when detect use-after-free?
> 
> I meant that where we had the kasan_record_aux_stack(work) call in the
> first version of the patch, work can be already freed. We must not
> access work after queueing it.
> 

Got it. Now I must treat urgent issue, I will send v2 patch tomorrow.

Thanks for your review.

> > > >         return 0;
> > > >  }
> > > >
> > > > diff --git a/mm/kasan/kasan.h b/mm/kasan/kasan.h
> > > > index 3436c6bf7c0c..d300fe9415bd 100644
> > > > --- a/mm/kasan/kasan.h
> > > > +++ b/mm/kasan/kasan.h
> > > > @@ -146,7 +146,7 @@ struct kasan_alloc_meta {
> > > >         struct kasan_track alloc_track;
> > > >  #ifdef CONFIG_KASAN_GENERIC
> > > >         /*
> > > > -        * call_rcu() call stack is stored into struct kasan_alloc_meta.
> > > > +        * Auxiliary stack is stored into struct kasan_alloc_meta.
> > > >          * The free stack is stored into struct kasan_free_meta.
> > > >          */
> > > >         depot_stack_handle_t aux_stack[2];
> > > > --
> > > > 2.18.0
> > >
> >
> > --
> > You received this message because you are subscribed to the Google Groups "kasan-dev" group.
> > To unsubscribe from this group and stop receiving emails from it, send an email to kasan-dev+unsubscribe@...glegroups.com.
> > To view this discussion on the web visit https://groups.google.com/d/msgid/kasan-dev/1615801102.24887.4.camel%40mtksdccf07.
> 

Powered by blists - more mailing lists