Re: [PATCH] ext4: fix WARNING in ext4_expand_extra_isize_ea

From: Jan Kara
Date: Thu Dec 01 2022 - 09:22:07 EST


On Thu 01-12-22 21:25:07, yebin (H) wrote:
>
>
> On 2022/12/1 20:19, Jan Kara wrote:
> > Hello!
> >
> > On Thu 01-12-22 16:48:44, Ye Bin wrote:
> > > From: Ye Bin <yebin10@xxxxxxxxxx>
> > >
> > > Syzbot found the following issue:
> > > ------------[ cut here ]------------
> > > WARNING: CPU: 1 PID: 3631 at mm/page_alloc.c:5534 __alloc_pages+0x30a/0x560 mm/page_alloc.c:5534
> > > Modules linked in:
> > > CPU: 1 PID: 3631 Comm: syz-executor261 Not tainted 6.1.0-rc6-syzkaller-00308-g644e9524388a #0
> > > Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 10/26/2022
> > > RIP: 0010:__alloc_pages+0x30a/0x560 mm/page_alloc.c:5534
> > > RSP: 0018:ffffc90003ccf080 EFLAGS: 00010246
> > > RAX: ffffc90003ccf0e0 RBX: 000000000000000c RCX: 0000000000000000
> > > RDX: 0000000000000028 RSI: 0000000000000000 RDI: ffffc90003ccf108
> > > RBP: ffffc90003ccf198 R08: dffffc0000000000 R09: ffffc90003ccf0e0
> > > R10: fffff52000799e21 R11: 1ffff92000799e1c R12: 0000000000040c40
> > > R13: 1ffff92000799e18 R14: dffffc0000000000 R15: 1ffff92000799e14
> > > FS: 0000555555c10300(0000) GS:ffff8880b9900000(0000) knlGS:0000000000000000
> > > CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> > > CR2: 00007ffc36f70000 CR3: 00000000744ad000 CR4: 00000000003506e0
> > > DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
> > > DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
> > > Call Trace:
> > > <TASK>
> > > __alloc_pages_node include/linux/gfp.h:223 [inline]
> > > alloc_pages_node include/linux/gfp.h:246 [inline]
> > > __kmalloc_large_node+0x8a/0x1a0 mm/slab_common.c:1096
> > > __do_kmalloc_node mm/slab_common.c:943 [inline]
> > > __kmalloc+0xfe/0x1a0 mm/slab_common.c:968
> > > kmalloc include/linux/slab.h:558 [inline]
> > > ext4_xattr_move_to_block fs/ext4/xattr.c:2558 [inline]
> > > ext4_xattr_make_inode_space fs/ext4/xattr.c:2673 [inline]
> > > ext4_expand_extra_isize_ea+0xe3f/0x1cd0 fs/ext4/xattr.c:2765
> > > __ext4_expand_extra_isize+0x2b8/0x3f0 fs/ext4/inode.c:5857
> > > ext4_try_to_expand_extra_isize fs/ext4/inode.c:5900 [inline]
> > > __ext4_mark_inode_dirty+0x51a/0x670 fs/ext4/inode.c:5978
> > > ext4_inline_data_truncate+0x548/0xd00 fs/ext4/inline.c:2021
> > > ext4_truncate+0x341/0xeb0 fs/ext4/inode.c:4221
> > > ext4_process_orphan+0x1aa/0x2d0 fs/ext4/orphan.c:339
> > > ext4_orphan_cleanup+0xb60/0x1340 fs/ext4/orphan.c:474
> > > __ext4_fill_super fs/ext4/super.c:5515 [inline]
> > > ext4_fill_super+0x80ed/0x8610 fs/ext4/super.c:5643
> > > get_tree_bdev+0x400/0x620 fs/super.c:1324
> > > vfs_get_tree+0x88/0x270 fs/super.c:1531
> > > do_new_mount+0x289/0xad0 fs/namespace.c:3040
> > > do_mount fs/namespace.c:3383 [inline]
> > > __do_sys_mount fs/namespace.c:3591 [inline]
> > > __se_sys_mount+0x2d3/0x3c0 fs/namespace.c:3568
> > > do_syscall_x64 arch/x86/entry/common.c:50 [inline]
> > > do_syscall_64+0x3d/0xb0 arch/x86/entry/common.c:80
> > > entry_SYSCALL_64_after_hwframe+0x63/0xcd
> > > </TASK>
> > >
> > > Reason is allocate 16M memory by kmalloc, but MAX_ORDER is 11, kmalloc
> > > can allocate maxium size memory is 4M.
> > > XATTR_SIZE_MAX is currently 64k, but EXT4_XATTR_SIZE_MAX is '(1 << 24)',
> > > so 'ext4_xattr_check_entries()' regards this length as legal. Then trigger
> > > warning in 'ext4_xattr_move_to_block()'.
> > > To solve above issue, adjust EXT4_XATTR_SIZE_MAX to '(1 << 22)' which
> > > is kmalloc can allocate maxium size.
> > >
> > > Reported-by: syzbot+4d99a966fd74bdeeec36@xxxxxxxxxxxxxxxxxxxxxxxxx
> > > Fixes: 54dd0e0a1b25 ("ext4: add extra checks to ext4_xattr_block_get()")
> > > Signed-off-by: Ye Bin <yebin10@xxxxxxxxxx>
> > Thanks for the report and the fix but I think it is actually wrong. We
> > cannot just change EXT4_XATTR_SIZE_MAX because there may be already
> > filesystems with this large extended attributes and we'd be suddently
> Firstly, thanks for your advice.
> I have a question. Since the Linux-2.6.12-rc2 version, XATTR can only be set
> to 64K
> at most. I understand that no file's extended attribute exceeds 64K.

You're right that VFS actually limits xattr size to 64k. So the chances
that someone actually has filesystem with larger xattrs are slim. But I
know that Lustre guys run with their modified kernels and they were the
ones implementing ea_inode feature so maybe they'd bumped the VFS limit as
well in their kernels. Dunno. Anyway using kvmalloc() (like the xattr core
does) looks like a better fix.

Honza
--
Jan Kara <jack@xxxxxxxx>
SUSE Labs, CR