Re: [syzbot] [xfs?] possible deadlock in xfs_ilock_data_map_shared

From: Dave Chinner
Date: Fri Apr 26 2024 - 17:20:22 EST

[cc linux-mm@xxxxxxxxx]

On Fri, Apr 26, 2024 at 09:32:28AM -0700, Darrick J. Wong wrote:
> On Thu, Apr 25, 2024 at 07:46:28AM -0700, syzbot wrote:
> > Hello,
> >
> > syzbot found the following issue on:
> >
> > HEAD commit: 977b1ef51866 Merge tag 'block-6.9-20240420' of git://git.k..
> > git tree: upstream
> > console output: https://syzkaller.appspot.com/x/log.txt?x=126497cd180000
> > kernel config: https://syzkaller.appspot.com/x/.config?x=d239903bd07761e5
> > dashboard link: https://syzkaller.appspot.com/bug?extid=b7e8d799f0ab724876f9
> > compiler: Debian clang version 15.0.6, GNU ld (GNU Binutils for Debian) 2.40
> >
> > Unfortunately, I don't have any reproducer for this issue yet.
> >
> > Downloadable assets:
> > disk image: https://storage.googleapis.com/syzbot-assets/08d7b6e107aa/disk-977b1ef5.raw.xz
> > vmlinux: https://storage.googleapis.com/syzbot-assets/9c5e543ffdcf/vmlinux-977b1ef5.xz
> > kernel image: https://storage.googleapis.com/syzbot-assets/04a6d79d2f69/bzImage-977b1ef5.xz
> >
> > IMPORTANT: if you fix the issue, please add the following tag to the commit:
> > Reported-by: syzbot+b7e8d799f0ab724876f9@xxxxxxxxxxxxxxxxxxxxxxxxx
> >
> > XFS (loop2): Ending clean mount
> > ======================================================
> > WARNING: possible circular locking dependency detected
> > 6.9.0-rc4-syzkaller-00266-g977b1ef51866 #0 Not tainted
> > ------------------------------------------------------
> > syz-executor.2/7915 is trying to acquire lock:
> > ffffffff8e42a800 (fs_reclaim){+.+.}-{0:0}, at: might_alloc include/linux/sched/mm.h:312 [inline]
> > ffffffff8e42a800 (fs_reclaim){+.+.}-{0:0}, at: slab_pre_alloc_hook mm/slub.c:3746 [inline]
> > ffffffff8e42a800 (fs_reclaim){+.+.}-{0:0}, at: slab_alloc_node mm/slub.c:3827 [inline]
> > ffffffff8e42a800 (fs_reclaim){+.+.}-{0:0}, at: kmalloc_trace+0x47/0x360 mm/slub.c:3992
> >
> > but task is already holding lock:
> > ffff888056da8118 (&xfs_dir_ilock_class){++++}-{3:3}, at: xfs_ilock_data_map_shared+0x4f/0x70 fs/xfs/xfs_inode.c:114
> >
> > which lock already depends on the new lock.
> >
> >
> > the existing dependency chain (in reverse order) is:
> >
> > -> #1 (&xfs_dir_ilock_class){++++}-{3:3}:
> > lock_acquire+0x1ed/0x550 kernel/locking/lockdep.c:5754
> > down_write_nested+0x3d/0x50 kernel/locking/rwsem.c:1695
> > xfs_reclaim_inode fs/xfs/xfs_icache.c:945 [inline]
> > xfs_icwalk_process_inode fs/xfs/xfs_icache.c:1631 [inline]
> > xfs_icwalk_ag+0x120e/0x1ad0 fs/xfs/xfs_icache.c:1713
> > xfs_icwalk fs/xfs/xfs_icache.c:1762 [inline]
> > xfs_reclaim_inodes_nr+0x257/0x360 fs/xfs/xfs_icache.c:1011
> > super_cache_scan+0x411/0x4b0 fs/super.c:227
> > do_shrink_slab+0x707/0x1160 mm/shrinker.c:435
> > shrink_slab+0x1092/0x14d0 mm/shrinker.c:662
> > shrink_one+0x453/0x880 mm/vmscan.c:4774
> > shrink_many mm/vmscan.c:4835 [inline]
> > lru_gen_shrink_node mm/vmscan.c:4935 [inline]
> > shrink_node+0x3b17/0x4310 mm/vmscan.c:5894
> > kswapd_shrink_node mm/vmscan.c:6704 [inline]
> > balance_pgdat mm/vmscan.c:6895 [inline]
> > kswapd+0x1882/0x38a0 mm/vmscan.c:7164
> > kthread+0x2f2/0x390 kernel/kthread.c:388
> > ret_from_fork+0x4d/0x80 arch/x86/kernel/process.c:147
> > ret_from_fork_asm+0x1a/0x30 arch/x86/entry/entry_64.S:244
> >
> > -> #0 (fs_reclaim){+.+.}-{0:0}:
> > check_prev_add kernel/locking/lockdep.c:3134 [inline]
> > check_prevs_add kernel/locking/lockdep.c:3253 [inline]
> > validate_chain+0x18cb/0x58e0 kernel/locking/lockdep.c:3869
> > __lock_acquire+0x1346/0x1fd0 kernel/locking/lockdep.c:5137
> > lock_acquire+0x1ed/0x550 kernel/locking/lockdep.c:5754
> > __fs_reclaim_acquire mm/page_alloc.c:3698 [inline]
> > fs_reclaim_acquire+0x88/0x140 mm/page_alloc.c:3712
> > might_alloc include/linux/sched/mm.h:312 [inline]
> > slab_pre_alloc_hook mm/slub.c:3746 [inline]
> > slab_alloc_node mm/slub.c:3827 [inline]
> > kmalloc_trace+0x47/0x360 mm/slub.c:3992
> > kmalloc include/linux/slab.h:628 [inline]
> > add_stack_record_to_list mm/page_owner.c:177 [inline]
> > inc_stack_record_count mm/page_owner.c:219 [inline]
> > __set_page_owner+0x561/0x810 mm/page_owner.c:334
> > set_page_owner include/linux/page_owner.h:32 [inline]
> > post_alloc_hook+0x1ea/0x210 mm/page_alloc.c:1534
> > prep_new_page mm/page_alloc.c:1541 [inline]
> > get_page_from_freelist+0x3410/0x35b0 mm/page_alloc.c:3317
> > __alloc_pages+0x256/0x6c0 mm/page_alloc.c:4575
> > __alloc_pages_bulk+0x729/0xd40 mm/page_alloc.c:4523
> > alloc_pages_bulk_array include/linux/gfp.h:202 [inline]
> > xfs_buf_alloc_pages+0x1a7/0x860 fs/xfs/xfs_buf.c:398
> > xfs_buf_find_insert+0x19a/0x1540 fs/xfs/xfs_buf.c:650
> > xfs_buf_get_map+0x149c/0x1ae0 fs/xfs/xfs_buf.c:755
> > xfs_buf_read_map+0x111/0xa60 fs/xfs/xfs_buf.c:860
> > xfs_trans_read_buf_map+0x260/0xad0 fs/xfs/xfs_trans_buf.c:289
> > xfs_da_read_buf+0x2b1/0x470 fs/xfs/libxfs/xfs_da_btree.c:2674
> > xfs_dir3_block_read+0x92/0x1a0 fs/xfs/libxfs/xfs_dir2_block.c:145
> > xfs_dir2_block_lookup_int+0x109/0x7d0 fs/xfs/libxfs/xfs_dir2_block.c:700
> > xfs_dir2_block_lookup+0x19a/0x630 fs/xfs/libxfs/xfs_dir2_block.c:650
> > xfs_dir_lookup+0x633/0xaf0 fs/xfs/libxfs/xfs_dir2.c:399
>
> Hm. We've taken an ILOCK in xfs_dir_lookup, and now we're reading a
> directory block. We don't have PF_MEMALLOC_NOFS set, nor do we pass
> GFP_NOFS when allocating the xfs_buf pages.
>
> Nothing in this code path sets PF_MEMALLOC_NOFS explicitly, nor does it
> create a xfs_trans_alloc_empty, which would set that. Prior to the
> removal of kmem_alloc, I think we were much more aggressive about
> GFP_NOFS usage.

This isn't an XFS bug. The XFS code is correct - the callsite in the
buffer cache is using GFP_KERNEL | __GFP_NOLOCKDEP explicitly to
avoid these sorts of false positives.

Please take a closer look at the stack trace - there's a second
memory allocation taking place there way below the XFS memory
allocation inside the page owner tracking code itself:

static void add_stack_record_to_list(struct stack_record *stack_record,
gfp_t gfp_mask)
{
unsigned long flags;
struct stack *stack;

/* Filter gfp_mask the same way stackdepot does, for consistency */
gfp_mask &= ~GFP_ZONEMASK;
gfp_mask &= (GFP_ATOMIC | GFP_KERNEL);
gfp_mask |= __GFP_NOWARN;

set_current_in_page_owner();
stack = kmalloc(sizeof(*stack), gfp_mask);
if (!stack) {
unset_current_in_page_owner();
return;
}
unset_current_in_page_owner();
....

Look familiar? That exactly the same gfp mask filtering that the
stackdepot code was doing that caused this issue with KASAN:

https://lore.kernel.org/linux-xfs/000000000000fbf10e06164f3695@xxxxxxxxxx/

Which was fixed with this patch:

https://lore.kernel.org/linux-xfs/20240418141133.22950-1-ryabinin.a.a@xxxxxxxxx/

Essentially, we're now playing whack-a-mole with internal kernel
debug code that doesn't honor __GFP_NOLOCKDEP....

MM-people: can you please do an audit of all the nested allocations
that occur inside the public high level allocation API and ensure
that they all obey __GFP_NOLOCKDEP so we don't have syzbot keep
tripping over them one at a time?

-Dave.
--
Dave Chinner
david@xxxxxxxxxxxxx