Re: [PATCH] fs/buffer: fix a NULL pointer dereference in drop_buffers()

[Matthew Wilcox]

On Fri, Nov 18, 2022 at 03:54:54PM +0800, Liu Shixin wrote:
> On 2022/11/18 13:30, Matthew Wilcox wrote:
> > On Wed, Nov 09, 2022 at 05:50:18PM +0800, Liu Shixin wrote:
> >> syzbot found a null-ptr-deref by KASAN:
> >>
> >>  BUG: KASAN: null-ptr-deref in instrument_atomic_read include/linux/instrumented.h:71 [inline]
> >>  BUG: KASAN: null-ptr-deref in atomic_read include/linux/atomic/atomic-instrumented.h:27 [inline]
> >>  BUG: KASAN: null-ptr-deref in buffer_busy fs/buffer.c:2856 [inline]
> >>  BUG: KASAN: null-ptr-deref in drop_buffers+0x61/0x2f0 fs/buffer.c:2868
> >>  Read of size 4 at addr 0000000000000060 by task syz-executor.5/24786
> >>
> >>  CPU: 0 PID: 24786 Comm: syz-executor.5 Not tainted 6.0.0-syzkaller-09589-g55be6084c8e0 #0
> >>  Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 09/22/2022
> >>  Call Trace:
> >>   <TASK>
> >>   __dump_stack lib/dump_stack.c:88 [inline]
> >>   dump_stack_lvl+0x1e3/0x2cb lib/dump_stack.c:106
> >>   print_report+0xf1/0x220 mm/kasan/report.c:436
> >>   kasan_report+0xfb/0x130 mm/kasan/report.c:495
> >>   kasan_check_range+0x2a7/0x2e0 mm/kasan/generic.c:189
> >>   instrument_atomic_read include/linux/instrumented.h:71 [inline]
> >>   atomic_read include/linux/atomic/atomic-instrumented.h:27 [inline]
> >>   buffer_busy fs/buffer.c:2856 [inline]
> >>   drop_buffers+0x61/0x2f0 fs/buffer.c:2868
> >>   try_to_free_buffers+0x2b1/0x640 fs/buffer.c:2898
> >> [...]
> >>
> >> We use folio_has_private() to decide whether call filemap_release_folio(),
> >> which may call try_to_free_buffers() then. folio_has_private() return true
> >> for both PG_private and PG_private_2. We should only call try_to_free_buffers()
> >> for case PG_private. So we should recheck PG_private in try_to_free_buffers().
> >>
> >> Reported-by: syzbot+fbdb4ec578ebdcfb9ed2@xxxxxxxxxxxxxxxxxxxxxxxxx
> >> Fixes: 266cf658efcf ("FS-Cache: Recruit a page flags for cache management")
> > but this can only happen for a filesystem which uses both bufferheads
> > and PG_private_2.  afaik there aren't any of those in the tree.  so
> > this bug can't actually happen.
> >
> > if you have your own filesystem that does, you need to submit it.
> This null-ptr-deref is found by syzbot, not by my own filesystem. I review the related code and
> found no other possible cause. There are lock protection all the place calling try_to_free_buffers().
> So I only thought of this one possibility. I'm also trying to reproduce the problem but haven't
> been successful.
> 
> If this can't actually happen, maybe I'm missing something when review the code. I'll keep trying
> to see if I can reproduce the problem.

perhaps you could include more information, like the rest of the call
stack so we can see what filesystem is involved?