On Mon, Aug 12, 2024 at 07:23:00AM +0200, Christoph Hellwig wrote:
> When xfs_icwalk_ag skips an inode because it was RCU freed from another
> AG, the slot for the inode in the batch array needs to be zeroed. We
> also really shouldn't try to grab the inode in that case (or at very
> least undo the grab), so move the call to xfs_icwalk_ag after this sanity
> check.
>
> Fixes: 1a3e8f3da09c ("xfs: convert inode cache lookups to use RCU locking")
> Signed-off-by: Christoph Hellwig <hch@xxxxxx>
> ---
> fs/xfs/xfs_icache.c | 11 +++++++----
> 1 file changed, 7 insertions(+), 4 deletions(-)
>
> diff --git a/fs/xfs/xfs_icache.c b/fs/xfs/xfs_icache.c
> index ae3c049fd3a216..3ee92d3d1770db 100644
> --- a/fs/xfs/xfs_icache.c
> +++ b/fs/xfs/xfs_icache.c
> @@ -1701,9 +1701,6 @@ xfs_icwalk_ag(
> for (i = 0; i < nr_found; i++) {
> struct xfs_inode *ip = batch[i];
>
> - if (done || !xfs_icwalk_igrab(goal, ip, icw))
> - batch[i] = NULL;
> -
> /*
> * Update the index for the next lookup. Catch
> * overflows into the next AG range which can occur if
> @@ -1716,8 +1713,14 @@ xfs_icwalk_ag(
> * us to see this inode, so another lookup from the
> * same index will not find it again.
> */
> - if (XFS_INO_TO_AGNO(mp, ip->i_ino) != pag->pag_agno)
> + if (XFS_INO_TO_AGNO(mp, ip->i_ino) != pag->pag_agno) {
> + batch[i] = NULL;
> continue;
> + }
> +
> + if (done || !xfs_icwalk_igrab(goal, ip, icw))
> + batch[i] = NULL;
IOWs, if @ip has been freed and reallocated to a different AG, then we
don't want to touch it at all, not even to check IRECLAIMABLE in igrab.
I think that sounds correct so
Reviewed-by: Darrick J. Wong <djwong@xxxxxxxxxx>
--D
> +
> first_index = XFS_INO_TO_AGINO(mp, ip->i_ino + 1);
> if (first_index < XFS_INO_TO_AGINO(mp, ip->i_ino))
> done = true;
> --
> 2.43.0
>
>
