Re: [PATCH 2/4] xfs: abort dir/attr btree operation if btree is obviously weird

[Brian Foster <bfoster@redhat.com>]

On Wed, Oct 25, 2017 at 10:49:30PM -0700, Darrick J. Wong wrote:
> From: Darrick J. Wong <darrick.wong@oracle.com>
> 
> Abort an dir/attr btree operation if the attr btree has obvious problems
> like loops back to the root or pointers don't point down the tree.
> Found by fuzzing btree[0].before to zero in xfs/402, which livelocks on
> the cycle in the attr btree.
> 
> Apply the same checks to xfs_da3_node_lookup_int.
> 
> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
> ---
>  fs/xfs/libxfs/xfs_da_btree.c |   34 +++++++++++++++++++++++++++++++++-
>  fs/xfs/xfs_attr_list.c       |   32 ++++++++++++++++++++++++++++++++
>  2 files changed, 65 insertions(+), 1 deletion(-)
> 
> 
> diff --git a/fs/xfs/libxfs/xfs_da_btree.c b/fs/xfs/libxfs/xfs_da_btree.c
> index 6d43358..3dbeda6 100644
> --- a/fs/xfs/libxfs/xfs_da_btree.c
> +++ b/fs/xfs/libxfs/xfs_da_btree.c
> @@ -1466,6 +1466,7 @@ xfs_da3_node_lookup_int(
>  	int			max;
>  	int			error;
>  	int			retval;
> +	unsigned int		expected_level = -1U;
>  	struct xfs_inode	*dp = state->args->dp;
>  
>  	args = state->args;
> @@ -1474,7 +1475,7 @@ xfs_da3_node_lookup_int(
>  	 * Descend thru the B-tree searching each level for the right
>  	 * node to use, until the right hashval is found.
>  	 */
> -	blkno = (args->whichfork == XFS_DATA_FORK)? args->geo->leafblk : 0;
> +	blkno = args->geo->leafblk;
>  	for (blk = &state->path.blk[0], state->path.active = 1;
>  			 state->path.active <= XFS_DA_NODE_MAXDEPTH;
>  			 blk++, state->path.active++) {
> @@ -1496,6 +1497,8 @@ xfs_da3_node_lookup_int(
>  		    blk->magic == XFS_ATTR3_LEAF_MAGIC) {
>  			blk->magic = XFS_ATTR_LEAF_MAGIC;
>  			blk->hashval = xfs_attr_leaf_lasthash(blk->bp, NULL);
> +			if (expected_level == -1U)
> +				expected_level = 0;

Now that these functions check that expected_level reaches 0, it seems
like we could just initialize it to 0 and eliminate these
checks/assignments..? I think that may also help catch the case a single
node block with an invalid level == 0.

>  			break;
>  		}
>  
> @@ -1504,6 +1507,8 @@ xfs_da3_node_lookup_int(
>  			blk->magic = XFS_DIR2_LEAFN_MAGIC;
>  			blk->hashval = xfs_dir2_leaf_lasthash(args->dp,
>  							      blk->bp, NULL);
> +			if (expected_level == -1U)
> +				expected_level = 0;
>  			break;
>  		}
>  
> @@ -1517,6 +1522,26 @@ xfs_da3_node_lookup_int(
>  		dp->d_ops->node_hdr_from_disk(&nodehdr, node);
>  		btree = dp->d_ops->node_tree_p(node);
>  
> +		/* Tree taller than we can handle; bail out! */
> +		if (nodehdr.level >= XFS_DA_NODE_MAXDEPTH)
> +			return -EFSCORRUPTED;
> +
> +		if (blkno == args->geo->leafblk) {
> +			/*
> +			 * This is the root node, set up for the
> +			 * next level we want to see.
> +			 */
> +			expected_level = nodehdr.level - 1;
> +		} else if (expected_level != nodehdr.level) {
> +			/*
> +			 * Not the level we were expecting, which
> +			 * implies that the tree is bad.
> +			 */
> +			return -EFSCORRUPTED;
> +		} else {
> +			expected_level--;
> +		}
> +

And just an aesthetic nit that these stanzas could probably be made a
little easier to read:

		/*
		 * Track the level from the root node ...
		 */
		if (blk == args->geo->leafblk)
			expected_level = nodehdr.level - 1;
		else if (expected_level != nodehdr.level)
			return -EFSCORRUPTED;
		else
			expected_level--;

Otherwise looks Ok.

Brian

>  		max = nodehdr.count;
>  		blk->hashval = be32_to_cpu(btree[max - 1].hashval);
>  
> @@ -1562,8 +1587,15 @@ xfs_da3_node_lookup_int(
>  			blk->index = probe;
>  			blkno = be32_to_cpu(btree[probe].before);
>  		}
> +
> +		/* We can't point back to the root. */
> +		if (blkno == args->geo->leafblk)
> +			return -EFSCORRUPTED;
>  	}
>  
> +	if (expected_level != 0)
> +		return -EFSCORRUPTED;
> +
>  	/*
>  	 * A leaf block that ends in the hashval that we are interested in
>  	 * (final hashval == search hashval) means that the next block may
> diff --git a/fs/xfs/xfs_attr_list.c b/fs/xfs/xfs_attr_list.c
> index 48423eb..9a8dafc 100644
> --- a/fs/xfs/xfs_attr_list.c
> +++ b/fs/xfs/xfs_attr_list.c
> @@ -222,6 +222,7 @@ xfs_attr_node_list_lookup(
>  	struct xfs_trans		*tp = context->tp;
>  	int				i;
>  	int				error = 0;
> +	unsigned int			expected_level = -1U;
>  	uint16_t			magic;
>  
>  	ASSERT(*pbp == NULL);
> @@ -236,6 +237,8 @@ xfs_attr_node_list_lookup(
>  		switch (magic) {
>  		case XFS_ATTR_LEAF_MAGIC:
>  		case XFS_ATTR3_LEAF_MAGIC:
> +			if (expected_level == -1U)
> +				expected_level = 0;
>  			goto found_leaf;
>  		case XFS_DA_NODE_MAGIC:
>  		case XFS_DA3_NODE_MAGIC:
> @@ -249,6 +252,26 @@ xfs_attr_node_list_lookup(
>  
>  		dp->d_ops->node_hdr_from_disk(&nodehdr, node);
>  
> +		/* Tree taller than we can handle; bail out! */
> +		if (nodehdr.level >= XFS_DA_NODE_MAXDEPTH)
> +			goto out_corruptbuf;
> +
> +		if (cursor->blkno == 0) {
> +			/*
> +			 * This is the root node, set up for the
> +			 * next level we want to see.
> +			 */
> +			expected_level = nodehdr.level - 1;
> +		} else if (expected_level != nodehdr.level) {
> +			/*
> +			 * Not the level we were expecting, which
> +			 * implies that the tree is bad.
> +			 */
> +			goto out_corruptbuf;
> +		} else {
> +			expected_level--;
> +		}
> +
>  		btree = dp->d_ops->node_tree_p(node);
>  		for (i = 0; i < nodehdr.count; btree++, i++) {
>  			if (cursor->hashval <= be32_to_cpu(btree->hashval)) {
> @@ -262,15 +285,24 @@ xfs_attr_node_list_lookup(
>  			goto out_buf;
>  
>  		xfs_trans_brelse(tp, *pbp);
> +
> +		/* We can't point back to the root. */
> +		if (cursor->blkno == 0) {
> +			error = -EFSCORRUPTED;
> +			goto out;
> +		}
>  	}
>  
>  found_leaf:
> +	if (expected_level != 0)
> +		goto out_corruptbuf;
>  	return error;
>  
>  out_corruptbuf:
>  	error = -EFSCORRUPTED;
>  out_buf:
>  	xfs_trans_brelse(tp, *pbp);
> +out:
>  	*pbp = NULL;
>  	return error;
>  }
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-xfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html