From: Darrick J. Wong <darrick.wong@xxxxxxxxxx> In repair phase 5, inject_lost_blocks takes the blocks that we allocated but didn't use for constructing the new AG btrees and puts them back in the filesystem by adding them to the free space. The only btree that can overestimate like that are the free space btrees, but in principle, any of the btrees can do that. If the others did, the rmap record owner for those blocks won't necessarily be OWNER_AG, and if it isn't, repair will fail. Get rid of this logic bomb so that we can use it for /any/ block count overestimation, and then we can use it to clean up after all reconstruction of any btree type. Signed-off-by: Darrick J. Wong <darrick.wong@xxxxxxxxxx> --- repair/phase5.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/repair/phase5.c b/repair/phase5.c index 44a6bda8..75c480fd 100644 --- a/repair/phase5.c +++ b/repair/phase5.c @@ -2516,8 +2516,8 @@ inject_lost_blocks( if (error) goto out_cancel; - error = -libxfs_free_extent(tp, *fsb, 1, &XFS_RMAP_OINFO_AG, - XFS_AG_RESV_NONE); + error = -libxfs_free_extent(tp, *fsb, 1, + &XFS_RMAP_OINFO_ANY_OWNER, XFS_AG_RESV_NONE); if (error) goto out_cancel;
