Re: [PATCH] ocfs2: validate bg_bits during freefrag scan

[Heming Zhao]

On Thu, Apr 09, 2026 at 02:02:41PM +0800, ZhengYuan Huang wrote:
> [BUG]
> A crafted filesystem can trigger an out-of-bounds bitmap walk when
> OCFS2_IOC_INFO is issued with OCFS2_INFO_FL_NON_COHERENT.
> 
> BUG: KASAN: use-after-free in instrument_atomic_read include/linux/instrumented.h:68 [inline]
> BUG: KASAN: use-after-free in _test_bit include/asm-generic/bitops/instrumented-non-atomic.h:141 [inline]
> BUG: KASAN: use-after-free in test_bit_le include/asm-generic/bitops/le.h:21 [inline]
> BUG: KASAN: use-after-free in ocfs2_info_freefrag_scan_chain fs/ocfs2/ioctl.c:495 [inline]
> BUG: KASAN: use-after-free in ocfs2_info_freefrag_scan_bitmap fs/ocfs2/ioctl.c:588 [inline]
> BUG: KASAN: use-after-free in ocfs2_info_handle_freefrag fs/ocfs2/ioctl.c:662 [inline]
> BUG: KASAN: use-after-free in ocfs2_info_handle_request+0x1c66/0x3370 fs/ocfs2/ioctl.c:754
> Read of size 8 at addr ffff888031bce000 by task syz.0.636/1435
> Call Trace:
>  __dump_stack lib/dump_stack.c:94 [inline]
>  dump_stack_lvl+0xbe/0x130 lib/dump_stack.c:120
>  print_address_description mm/kasan/report.c:378 [inline]
>  print_report+0xd1/0x650 mm/kasan/report.c:482
>  kasan_report+0xfb/0x140 mm/kasan/report.c:595
>  check_region_inline mm/kasan/generic.c:186 [inline]
>  kasan_check_range+0x11c/0x200 mm/kasan/generic.c:200
>  __kasan_check_read+0x11/0x20 mm/kasan/shadow.c:31
>  instrument_atomic_read include/linux/instrumented.h:68 [inline]
>  _test_bit include/asm-generic/bitops/instrumented-non-atomic.h:141 [inline]
>  test_bit_le include/asm-generic/bitops/le.h:21 [inline]
>  ocfs2_info_freefrag_scan_chain fs/ocfs2/ioctl.c:495 [inline]
>  ocfs2_info_freefrag_scan_bitmap fs/ocfs2/ioctl.c:588 [inline]
>  ocfs2_info_handle_freefrag fs/ocfs2/ioctl.c:662 [inline]
>  ocfs2_info_handle_request+0x1c66/0x3370 fs/ocfs2/ioctl.c:754
>  ocfs2_info_handle+0x18d/0x2a0 fs/ocfs2/ioctl.c:828
>  ocfs2_ioctl+0x632/0x6e0 fs/ocfs2/ioctl.c:913
>  vfs_ioctl fs/ioctl.c:51 [inline]
>  __do_sys_ioctl fs/ioctl.c:597 [inline]
>  __se_sys_ioctl fs/ioctl.c:583 [inline]
>  __x64_sys_ioctl+0x197/0x1e0 fs/ioctl.c:583
>  ...
> 
> [CAUSE]
> ocfs2_info_freefrag_scan_chain() uses on-disk bg_bits directly as the
> bitmap scan limit. The coherent path reads group descriptors through
> ocfs2_read_group_descriptor(), which enforces bg_bits <= bg_size * 8.
> The non-coherent path uses ocfs2_read_blocks_sync() instead and skips
> that validation, so an impossible bg_bits value can drive the bitmap
> walk past the end of the block.
> 
> [FIX]
> Check bg_bits against the actual bitmap capacity before scanning the
> group bitmap and fail the ioctl with -EIO if the descriptor is
> impossible. This keeps the coherent path unchanged and gives the
> non-coherent path the same bound needed to avoid walking beyond the
> buffer.
> 
> Signed-off-by: ZhengYuan Huang <gality369@xxxxxxxxx>
> ---
>  fs/ocfs2/ioctl.c | 20 ++++++++++++++++++--
>  1 file changed, 18 insertions(+), 2 deletions(-)
> 
> diff --git a/fs/ocfs2/ioctl.c b/fs/ocfs2/ioctl.c
> index b6864602814c..a341a599e440 100644
> --- a/fs/ocfs2/ioctl.c
> +++ b/fs/ocfs2/ioctl.c
> @@ -441,7 +441,7 @@ static int ocfs2_info_freefrag_scan_chain(struct ocfs2_super *osb,
>  	struct buffer_head *bh = NULL;
>  	struct ocfs2_group_desc *bg = NULL;
>  
> -	unsigned int max_bits, num_clusters;
> +	unsigned int max_bits, max_bitmap_bits, num_clusters;
>  	unsigned int offset = 0, cluster, chunk;
>  	unsigned int chunk_free, last_chunksize = 0;
>  
> @@ -474,11 +474,27 @@ static int ocfs2_info_freefrag_scan_chain(struct ocfs2_super *osb,
>  		}
>  
>  		bg = (struct ocfs2_group_desc *)bh->b_data;
> +		max_bits = le16_to_cpu(bg->bg_bits);
> +		max_bitmap_bits = 8U * le16_to_cpu(bg->bg_size);

The bg->bg_size can also change.
It is better to use ocfs2_group_bitmap_size(osb->sb, 1, osb->s_feature_incompat)*8
to retrieve the max_bitmap_bits value.

> +
> +		/*
> +		 * Non-coherent scans read raw blocks and do not get the
> +		 * bg_bits <= bg_size * 8 validation from
> +		 * ocfs2_read_group_descriptor().
> +		 */
> +		if (max_bits > max_bitmap_bits) {
> +			mlog(ML_ERROR,
> +			     "Group desc #%llu has %u bits but bitmap holds %u\n",
> +			     (unsigned long long)blkno,
> +			     max_bits,
> +			     max_bitmap_bits);
> +			status = -EIO;
> +			goto bail;

Since this function is used to report chain states, it's better to continue
instead of breaking. In my view, reporting the error message and overwriting
max_bits with max_bitmap_bits is sufficient.

Thanks,
Heming
> +		}
>  
>  		if (!le16_to_cpu(bg->bg_free_bits_count))
>  			continue;
>  
> -		max_bits = le16_to_cpu(bg->bg_bits);
>  		offset = 0;
>  
>  		for (chunk = 0; chunk < chunks_in_group; chunk++) {
> -- 
> 2.43.0
> 
>