On Fri, Feb 01, 2019 at 02:01:38PM -0500, Brian Foster wrote:
> On Thu, Jan 31, 2019 at 03:18:27PM -0800, Darrick J. Wong wrote:
> > From: Darrick J. Wong <darrick.wong@xxxxxxxxxx>
> >
> > There's a loop that searches an unlinked bucket list to find the inode
> > that points to a given inode. Hoist this into a separate function;
> > later we'll use our iunlink backref cache to bypass the slow list
> > operation. No functional changes.
> >
> > Signed-off-by: Darrick J. Wong <darrick.wong@xxxxxxxxxx>
> > ---
> > fs/xfs/xfs_inode.c | 136 ++++++++++++++++++++++++++++++++++++++--------------
> > 1 file changed, 99 insertions(+), 37 deletions(-)
> >
> >
> > diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c
> > index ea38b66fbc59..d5b3f8fdac7e 100644
> > --- a/fs/xfs/xfs_inode.c
> > +++ b/fs/xfs/xfs_inode.c
> > @@ -2084,6 +2084,100 @@ xfs_iunlink(
> > return error;
> > }
> >
> > +/* Return the imap, dinode pointer, and buffer for an inode. */
> > +STATIC int
> > +xfs_iunlink_map_ino(
> > + struct xfs_trans *tp,
> > + xfs_ino_t next_ino,
>
> Might be good to clean up some of these variable names as this code gets
> factored out. E.g., "next_ino" doesn't mean much more than "ino" in the
> context of this helper.
Ok.
> > + struct xfs_imap *imap,
> > + struct xfs_dinode **dipp,
> > + struct xfs_buf **bpp)
> > +{
> > + struct xfs_mount *mp = tp->t_mountp;
> > + int error;
> > +
> > + imap->im_blkno = 0;
> > + error = xfs_imap(mp, tp, next_ino, imap, 0);
> > + if (error) {
> > + xfs_warn(mp, "%s: xfs_imap returned error %d.",
> > + __func__, error);
> > + return error;
> > + }
> > +
> > + error = xfs_imap_to_bp(mp, tp, imap, dipp, bpp, 0, 0);
> > + if (error) {
> > + xfs_warn(mp, "%s: xfs_imap_to_bp returned error %d.",
> > + __func__, error);
> > + return error;
> > + }
> > +
> > + return 0;
> > +}
> > +
> > +/*
> > + * Walk the unlinked chain from @head_agino until we find the inode that
> > + * points to @target_ino. Return the inode number, map, dinode pointer,
> > + * and inode cluster buffer of that inode as @ino, @imap, @dipp, and @bpp.
> > + *
> > + * Do not call this function if @target_agino is the head of the list.
> > + */
> > +STATIC int
> > +xfs_iunlink_map_prev(
> > + struct xfs_trans *tp,
> > + struct xfs_perag *pag,
> > + xfs_agino_t head_agino,
> > + xfs_agino_t target_agino,
> > + xfs_ino_t *ino,
> > + struct xfs_imap *imap,
> > + struct xfs_dinode **dipp,
> > + struct xfs_buf **bpp)
> > +{
>
> It would also be nice to have some oneline comments next to the params
> of some of these functions.
In addition to the description given above the function?
> > + struct xfs_imap last_imap;
> > + struct xfs_mount *mp = tp->t_mountp;
> > + struct xfs_buf *last_ibp = NULL;
> > + struct xfs_dinode *last_dip;
> > + xfs_ino_t next_ino = NULLFSINO;
> > + xfs_agino_t next_agino;
> > + int error;
> > +
> > + ASSERT(head_agino != target_agino);
> > +
> > + next_agino = head_agino;
> > + while (next_agino != target_agino) {
> > + xfs_agino_t unlinked_agino;
> > +
> > + if (last_ibp)
> > + xfs_trans_brelse(tp, last_ibp);
> > +
> > + next_ino = XFS_AGINO_TO_INO(mp, pag->pag_agno, next_agino);
> > + error = xfs_iunlink_map_ino(tp, next_ino, &last_imap,
> > + &last_dip, &last_ibp);
> > + if (error)
> > + return error;
> > +
> > + unlinked_agino = be32_to_cpu(last_dip->di_next_unlinked);
> > + if (!xfs_verify_agino(mp, pag->pag_agno, next_agino) ||
>
> Should the next_agino above be unlinked_agino? Also is the == check
> below a loop check (comment pls)?
Yes, fixed.
/*
* Make sure this pointer is valid and isn't an obvious
* infinite loop.
*/
> > + next_agino == unlinked_agino) {
> > + XFS_CORRUPTION_ERROR(__func__,
> > + XFS_ERRLEVEL_LOW, mp,
> > + last_dip, sizeof(*last_dip));
> > + error = -EFSCORRUPTED;
> > + return error;
> > + }
> > + next_agino = unlinked_agino;
> > + }
> > +
> > + /* Should never happen, but don't return garbage. */
> > + if (next_ino == NULLFSINO)
> > + return -EFSCORRUPTED;
> > +
> > + *ino = next_ino;
> > + memcpy(imap, &last_imap, sizeof(last_imap));
> > + *dipp = last_dip;
> > + *bpp = last_ibp;
> > + return 0;
> > +}
> > +
> > /*
> > * Pull the on-disk inode from the AGI unlinked list.
> > */
> > @@ -2153,43 +2247,11 @@ xfs_iunlink_remove(
> > } else {
> > struct xfs_imap imap;
> >
> > - /*
> > - * We need to search the list for the inode being freed.
> > - */
> > - last_ibp = NULL;
> > - while (next_agino != agino) {
> > - if (last_ibp)
> > - xfs_trans_brelse(tp, last_ibp);
> > -
> > - imap.im_blkno = 0;
> > - next_ino = XFS_AGINO_TO_INO(mp, agno, next_agino);
> > -
> > - error = xfs_imap(mp, tp, next_ino, &imap, 0);
> > - if (error) {
> > - xfs_warn(mp,
> > - "%s: xfs_imap returned error %d.",
> > - __func__, error);
> > - goto out_unlock;
> > - }
> > -
> > - error = xfs_imap_to_bp(mp, tp, &imap, &last_dip,
> > - &last_ibp, 0, 0);
> > - if (error) {
> > - xfs_warn(mp,
> > - "%s: xfs_imap_to_bp returned error %d.",
> > - __func__, error);
> > - goto out_unlock;
> > - }
> > -
> > - next_agino = be32_to_cpu(last_dip->di_next_unlinked);
> > - if (!xfs_verify_agino(mp, agno, next_agino)) {
> > - XFS_CORRUPTION_ERROR(__func__,
> > - XFS_ERRLEVEL_LOW, mp,
> > - last_dip, sizeof(*last_dip));
> > - error = -EFSCORRUPTED;
> > - goto out_unlock;
> > - }
> > - }
> > + /* We need to search the list for the inode being freed. */
> > + error = xfs_iunlink_map_prev(tp, pag, next_agino, agino,
> > + &next_ino, &imap, &last_dip, &last_ibp);
>
> Another variable nit... next_ino is kind of confusing here, particularly
> where we still use next_agino as well. The former is actually the "prev"
> ino output, right?
Yep, will fix.
--D
> Brian
>
> > + if (error)
> > + goto out_unlock;
> >
> > /*
> > * Now last_ibp points to the buffer previous to us on the
> >
