syzbot reported a oob in gfs2_fill_super. [1]

When setting blocksize fails, the s_blocksize_bits value is incorrect.

[1]
UBSAN: shift-out-of-bounds in fs/gfs2/ops_fstype.c:1172:19
shift exponent 4294967287 is too large for 64-bit type 'unsigned long'
CPU: 1 UID: 0 PID: 6127 Comm: syz.0.42 Not tainted 6.14.0-next-20250404-syzkaller #0 PREEMPT(full)
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 02/12/2025
Call Trace:
 <TASK>
 __dump_stack lib/dump_stack.c:94 [inline]
 dump_stack_lvl+0x241/0x360 lib/dump_stack.c:120
 ubsan_epilogue lib/ubsan.c:231 [inline]
 __ubsan_handle_shift_out_of_bounds+0x3c8/0x420 lib/ubsan.c:492
 gfs2_fill_super+0x255e/0x27b0 fs/gfs2/ops_fstype.c:1172
 get_tree_bdev_flags+0x490/0x5c0 fs/super.c:1636
 gfs2_get_tree+0x54/0x220 fs/gfs2/ops_fstype.c:1330
 vfs_get_tree+0x90/0x2b0 fs/super.c:1759
 do_new_mount+0x2cf/0xb70 fs/namespace.c:3879
 do_mount fs/namespace.c:4219 [inline]
 __do_sys_mount fs/namespace.c:4430 [inline]
 __se_sys_mount+0x38c/0x400 fs/namespace.c:4407
 do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
 do_syscall_64+0xf3/0x230 arch/x86/entry/syscall_64.c:94
 entry_SYSCALL_64_after_hwframe+0x77/0x7f

Reported-by: syzbot+b0018b7468b2af33b4d5@syzkaller.appspotmail.com
Signed-off-by: Edward Adam Davis <eadavis@qq.com>
---
 fs/gfs2/ops_fstype.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/fs/gfs2/ops_fstype.c b/fs/gfs2/ops_fstype.c
index e83d293c3614..d5bc25164b89 100644
--- a/fs/gfs2/ops_fstype.c
+++ b/fs/gfs2/ops_fstype.c
@@ -1167,6 +1167,10 @@ static int gfs2_fill_super(struct super_block *sb, struct fs_context *fc)
 	/* Set up the buffer cache and fill in some fake block size values
 	   to allow us to read-in the on-disk superblock. */
 	sdp->sd_sb.sb_bsize = sb_min_blocksize(sb, 512);
+	if (!sdp->sd_sb.sb_bsize) {
+		error = -EINVAL;
+		goto fail_free;
+	}
 	sdp->sd_sb.sb_bsize_shift = sb->s_blocksize_bits;
 	sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift - 9;
 	sdp->sd_fsb2bb = BIT(sdp->sd_fsb2bb_shift);
-- 
2.43.0

On Sat, Apr 12, 2025 at 7:16 AM Edward Adam Davis <eadavis@qq.com> wrote:
> syzbot reported a oob in gfs2_fill_super. [1]
>
> When setting blocksize fails, the s_blocksize_bits value is incorrect.
>
> [1]
> UBSAN: shift-out-of-bounds in fs/gfs2/ops_fstype.c:1172:19
> shift exponent 4294967287 is too large for 64-bit type 'unsigned long'
> CPU: 1 UID: 0 PID: 6127 Comm: syz.0.42 Not tainted 6.14.0-next-20250404-syzkaller #0 PREEMPT(full)
> Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 02/12/2025
> Call Trace:
>  <TASK>
>  __dump_stack lib/dump_stack.c:94 [inline]
>  dump_stack_lvl+0x241/0x360 lib/dump_stack.c:120
>  ubsan_epilogue lib/ubsan.c:231 [inline]
>  __ubsan_handle_shift_out_of_bounds+0x3c8/0x420 lib/ubsan.c:492
>  gfs2_fill_super+0x255e/0x27b0 fs/gfs2/ops_fstype.c:1172
>  get_tree_bdev_flags+0x490/0x5c0 fs/super.c:1636
>  gfs2_get_tree+0x54/0x220 fs/gfs2/ops_fstype.c:1330
>  vfs_get_tree+0x90/0x2b0 fs/super.c:1759
>  do_new_mount+0x2cf/0xb70 fs/namespace.c:3879
>  do_mount fs/namespace.c:4219 [inline]
>  __do_sys_mount fs/namespace.c:4430 [inline]
>  __se_sys_mount+0x38c/0x400 fs/namespace.c:4407
>  do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
>  do_syscall_64+0xf3/0x230 arch/x86/entry/syscall_64.c:94
>  entry_SYSCALL_64_after_hwframe+0x77/0x7f
>
> Reported-by: syzbot+b0018b7468b2af33b4d5@syzkaller.appspotmail.com
> Signed-off-by: Edward Adam Davis <eadavis@qq.com>
> ---
>  fs/gfs2/ops_fstype.c | 4 ++++
>  1 file changed, 4 insertions(+)
>
> diff --git a/fs/gfs2/ops_fstype.c b/fs/gfs2/ops_fstype.c
> index e83d293c3614..d5bc25164b89 100644
> --- a/fs/gfs2/ops_fstype.c
> +++ b/fs/gfs2/ops_fstype.c
> @@ -1167,6 +1167,10 @@ static int gfs2_fill_super(struct super_block *sb, struct fs_context *fc)
>         /* Set up the buffer cache and fill in some fake block size values
>            to allow us to read-in the on-disk superblock. */
>         sdp->sd_sb.sb_bsize = sb_min_blocksize(sb, 512);
> +       if (!sdp->sd_sb.sb_bsize) {
> +               error = -EINVAL;
> +               goto fail_free;
> +       }
>         sdp->sd_sb.sb_bsize_shift = sb->s_blocksize_bits;
>         sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift - 9;
>         sdp->sd_fsb2bb = BIT(sdp->sd_fsb2bb_shift);
> --
> 2.43.0

Thanks for this patch.

I've made a small coding style adjustment, added a check for the
return value of sb_set_blocksize(), and I've changed the description
to:

    gfs2: check sb_min_blocksize return value

    Check the return value of sb_min_blocksize(): it will be 0 when the
    requested block size is invalid.

    In addition, check the return value of sb_set_blocksize() as well.

Andreas