On Thu, Jan 23, 2020 at 07:51:44AM +0300, Dan Carpenter wrote:
> Hello Darrick J. Wong,
>
> The patch 0bb9d159bd01: "xfs: streamline xfs_attr3_leaf_inactive"
> from Jan 14, 2020, leads to the following static checker warning:
>
> fs/xfs/xfs_attr_inactive.c:122 xfs_attr3_leaf_inactive()
> error: uninitialized symbol 'error'.
>
> fs/xfs/xfs_attr_inactive.c
> 90 struct xfs_attr_leaf_entry *entry;
> 91 struct xfs_attr_leaf_name_remote *name_rmt;
> 92 int error;
> ^^^^^
>
> 93 int i;
> 94
> 95 xfs_attr3_leaf_hdr_from_disk(mp->m_attr_geo, &ichdr, leaf);
> 96
> 97 /*
> 98 * Find the remote value extents for this leaf and invalidate their
> 99 * incore buffers.
> 100 */
> 101 entry = xfs_attr3_leaf_entryp(leaf);
> 102 for (i = 0; i < ichdr.count; entry++, i++) {
>
> Smatch complains that we might not enter this loop or maybe we always
> hit a continue statement.
In theory there should never be an attr leaf block with zero entries,
but let's fix this anyway. Patch soon; thanks for the bug report. :)
--D
> 103 int blkcnt;
> 104
> 105 if (!entry->nameidx || (entry->flags & XFS_ATTR_LOCAL))
> 106 continue;
> 107
> 108 name_rmt = xfs_attr3_leaf_name_remote(leaf, i);
> 109 if (!name_rmt->valueblk)
> 110 continue;
> 111
> 112 blkcnt = xfs_attr3_rmt_blocks(dp->i_mount,
> 113 be32_to_cpu(name_rmt->valuelen));
> 114 error = xfs_attr3_rmt_stale(dp,
> 115 be32_to_cpu(name_rmt->valueblk), blkcnt);
> 116 if (error)
> 117 goto err;
> 118 }
> 119
> 120 xfs_trans_brelse(*trans, bp);
> 121 err:
> 122 return error;
>
> Possibly uninitialized
>
> 123 }
>
> regards,
> dan carpenter
