Re: [syzbot] [erofs?] KMSAN: uninit-value in erofs_fc_fill_super

[Gao Xiang via Linux-erofs <linux-erofs@lists.ozlabs.org>]

Hi exfat maintainers,

On Sat, Dec 28, 2024 at 11:45:22PM -0800, syzbot wrote:
> Hello,
> 
> syzbot found the following issue on:
> 
> HEAD commit:    9b2ffa6148b1 Merge tag 'mtd/fixes-for-6.13-rc5' of git://g..
> git tree:       upstream
> console output: https://syzkaller.appspot.com/x/log.txt?x=112374c4580000
> kernel config:  https://syzkaller.appspot.com/x/.config?x=f9048090d7bb0d06
> dashboard link: https://syzkaller.appspot.com/bug?extid=1379ee6b9a14d5dacaf2
> 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/244f25c1a275/disk-9b2ffa61.raw.xz
> vmlinux: https://storage.googleapis.com/syzbot-assets/0d14fc6634fd/vmlinux-9b2ffa61.xz
> kernel image: https://storage.googleapis.com/syzbot-assets/cb152a4c0fd2/bzImage-9b2ffa61.xz
> 
> IMPORTANT: if you fix the issue, please add the following tag to the commit:
> Reported-by: syzbot+1379ee6b9a14d5dacaf2@syzkaller.appspotmail.com
> 
> exFAT-fs (loop7): failed to load upcase table (idx : 0x00010000, chksum : 0x1a9973fb, utbl_chksum : 0xe619d30d)
> =====================================================
> BUG: KMSAN: uninit-value in erofs_read_superblock fs/erofs/super.c:274 [inline]
> BUG: KMSAN: uninit-value in erofs_fc_fill_super+0x66a/0x2520 fs/erofs/super.c:614
>  erofs_read_superblock fs/erofs/super.c:274 [inline]
>  erofs_fc_fill_super+0x66a/0x2520 fs/erofs/super.c:614
>  vfs_get_super fs/super.c:1280 [inline]
>  get_tree_nodev+0x183/0x350 fs/super.c:1299
>  erofs_fc_get_tree+0x34d/0x450 fs/erofs/super.c:721
>  vfs_get_tree+0xb1/0x5a0 fs/super.c:1814
>  do_new_mount+0x71f/0x15e0 fs/namespace.c:3507
>  path_mount+0x742/0x1f10 fs/namespace.c:3834
>  do_mount fs/namespace.c:3847 [inline]
>  __do_sys_mount fs/namespace.c:4057 [inline]
>  __se_sys_mount+0x722/0x810 fs/namespace.c:4034
>  __x64_sys_mount+0xe4/0x150 fs/namespace.c:4034
>  x64_sys_call+0x39bf/0x3c30 arch/x86/include/generated/asm/syscalls_64.h:166
>  do_syscall_x64 arch/x86/entry/common.c:52 [inline]
>  do_syscall_64+0xcd/0x1e0 arch/x86/entry/common.c:83
>  entry_SYSCALL_64_after_hwframe+0x77/0x7f
> 
> Uninit was created at:
>  __alloc_pages_noprof+0x9a7/0xe00 mm/page_alloc.c:4776
>  alloc_pages_mpol_noprof+0x299/0x990 mm/mempolicy.c:2269
>  alloc_pages_noprof mm/mempolicy.c:2348 [inline]
>  folio_alloc_noprof+0x1db/0x310 mm/mempolicy.c:2355
>  filemap_alloc_folio_noprof+0xa6/0x440 mm/filemap.c:1009
>  __filemap_get_folio+0xac4/0x1550 mm/filemap.c:1951
>  block_write_begin+0x6e/0x2b0 fs/buffer.c:2221
>  exfat_write_begin+0xfb/0x400 fs/exfat/inode.c:434
>  exfat_extend_valid_size fs/exfat/file.c:553 [inline]
>  exfat_file_write_iter+0x771/0x12a0 fs/exfat/file.c:598
>  do_iter_readv_writev+0x88a/0xa30
>  vfs_writev+0x56a/0x14f0 fs/read_write.c:1050
>  do_pwritev fs/read_write.c:1146 [inline]
>  __do_sys_pwritev2 fs/read_write.c:1204 [inline]
>  __se_sys_pwritev2+0x262/0x460 fs/read_write.c:1195
>  __x64_sys_pwritev2+0x11f/0x1a0 fs/read_write.c:1195
>  x64_sys_call+0x368c/0x3c30 arch/x86/include/generated/asm/syscalls_64.h:329
>  do_syscall_x64 arch/x86/entry/common.c:52 [inline]
>  do_syscall_64+0xcd/0x1e0 arch/x86/entry/common.c:83
>  entry_SYSCALL_64_after_hwframe+0x77/0x7f
>

Currently, I don't think it's an EROFS issue but since it doesn't have
a valid reproducer so I don't have an exact idea.

This case is out of EROFS file-backed mount, which seems to read EROFS
superblock (erofs_read_superblock -> erofs_read_metabuf -> erofs_bread)
via exfat page cache:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/fs/erofs/super.c?h=v6.13-rc5#n274

But it seems exfat returns an unlocked uptodate page without fully
initialized data.

I'm not sure if it's a post-EOF read for this specific regular inode,
but IMHO, at least mmap read allows post-EOF read within the same page,
so it'd be better to leave the whole page initialized on the exfat side.

I'm not sure if it's related to exfat_extend_valid_size() though.

Thanks,
Gao Xiang