In reset_nlink(), we believe rec->found_link as accurate number of the valid links. However it only records the number of found DIR_ITEM, so we can't use it as reliable value.
Before this patch, in some case, leaf corruption recovery will believe there is a valid backref but don't add_link() since it can't find any valid one and don't put it into the lost+found dir. So the recovered inode will be considered as an orphan item without orphan item and repair_inode_orphan_item() will add orphan item for it, causing all the filename/filetype we recovered being a waste of time. Signed-off-by: Qu Wenruo <[email protected]> --- cmds-check.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) diff --git a/cmds-check.c b/cmds-check.c index aa47984..1bb95dc 100644 --- a/cmds-check.c +++ b/cmds-check.c @@ -1959,6 +1959,9 @@ static int reset_nlink(struct btrfs_trans_handle *trans, struct btrfs_inode_item *inode_item; int ret = 0; + /* We don't believe this either, reset it and iterate backref */ + rec->found_link = 0; + /* Remove all backref including the valid ones */ list_for_each_entry_safe(backref, tmp, &rec->backrefs, list) { ret = btrfs_unlink(trans, root, rec->ino, backref->dir, @@ -1973,7 +1976,8 @@ static int reset_nlink(struct btrfs_trans_handle *trans, backref->found_inode_ref)) { list_del(&backref->list); free(backref); - } + } else + rec->found_link++; } /* Set nlink to 0 */ -- 2.2.1 -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html
