On Wed, Oct 13, 2021 at 04:31:22PM +1100, Dave Chinner wrote:
> On Tue, Oct 12, 2021 at 04:33:01PM -0700, Darrick J. Wong wrote:
> > From: Darrick J. Wong <djwong@xxxxxxxxxx>
> >
> > Split out the btree level information into a separate struct and put it
> > at the end of the cursor structure as a VLA. The realtime rmap btree
> > (which is rooted in an inode) will require the ability to support many
> > more levels than a per-AG btree cursor, which means that we're going to
> > create two btree cursor caches to conserve memory for the more common
> > case.
> >
> > Signed-off-by: Darrick J. Wong <djwong@xxxxxxxxxx>
> > Reviewed-by: Chandan Babu R <chandan.babu@xxxxxxxxxx>
> > Reviewed-by: Christoph Hellwig <hch@xxxxxx>
> > ---
> > fs/xfs/libxfs/xfs_alloc.c | 6 +-
> > fs/xfs/libxfs/xfs_bmap.c | 10 +--
> > fs/xfs/libxfs/xfs_btree.c | 168 +++++++++++++++++++++++----------------------
> > fs/xfs/libxfs/xfs_btree.h | 28 ++++++--
> > fs/xfs/scrub/bitmap.c | 22 +++---
> > fs/xfs/scrub/bmap.c | 2 -
> > fs/xfs/scrub/btree.c | 47 +++++++------
> > fs/xfs/scrub/trace.c | 7 +-
> > fs/xfs/scrub/trace.h | 10 +--
> > fs/xfs/xfs_super.c | 2 -
> > fs/xfs/xfs_trace.h | 2 -
> > 11 files changed, 164 insertions(+), 140 deletions(-)
>
> Hmmm - subject of the patch doesn't really match the changes being
> made - there's nothing here that makes the btree cursor heights
> dynamic. It's just a structure layout change...
"xfs: prepare xfs_btree_cur for dynamic cursor heights" ?
>
> > @@ -415,9 +415,9 @@ xfs_btree_dup_cursor(
> > * For each level current, re-get the buffer and copy the ptr value.
> > */
> > for (i = 0; i < new->bc_nlevels; i++) {
> > - new->bc_ptrs[i] = cur->bc_ptrs[i];
> > - new->bc_ra[i] = cur->bc_ra[i];
> > - bp = cur->bc_bufs[i];
> > + new->bc_levels[i].ptr = cur->bc_levels[i].ptr;
> > + new->bc_levels[i].ra = cur->bc_levels[i].ra;
> > + bp = cur->bc_levels[i].bp;
> > if (bp) {
> > error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp,
> > xfs_buf_daddr(bp), mp->m_bsize,
> > @@ -429,7 +429,7 @@ xfs_btree_dup_cursor(
> > return error;
> > }
> > }
> > - new->bc_bufs[i] = bp;
> > + new->bc_levels[i].bp = bp;
> > }
> > *ncur = new;
> > return 0;
>
> ObHuh: that dup_cursor code seems like a really obtuse way of doing:
>
> bip = cur->bc_levels[i].bp->b_log_item;
> bip->bli_recur++;
> new->bc_levels[i] = cur->bc_levels[i];
>
> But that's not a problem this patch needs to solve. Just something
> that made me go hmmmm...
Yeah, I noticed that too while I was checking the results of my sed
script.
> > @@ -922,11 +922,11 @@ xfs_btree_readahead(
> > (lev == cur->bc_nlevels - 1))
> > return 0;
> >
> > - if ((cur->bc_ra[lev] | lr) == cur->bc_ra[lev])
> > + if ((cur->bc_levels[lev].ra | lr) == cur->bc_levels[lev].ra)
> > return 0;
>
> That's whacky logic. Surely that's just:
>
> if (cur->bc_levels[lev].ra & lr)
> return 0;
This is an early-exit test, which means the careful check is necessary.
If (some day) someone calls this function with (LEFTRA|RIGHTRA) to
readahead both siblings on a btree level where one sibling has been ra'd
but not the other, we must avoid taking the branch.
> > diff --git a/fs/xfs/libxfs/xfs_btree.h b/fs/xfs/libxfs/xfs_btree.h
> > index 1018bcc43d66..f31f057bec9d 100644
> > --- a/fs/xfs/libxfs/xfs_btree.h
> > +++ b/fs/xfs/libxfs/xfs_btree.h
> > @@ -212,6 +212,19 @@ struct xfs_btree_cur_ino {
> > #define XFS_BTCUR_BMBT_INVALID_OWNER (1 << 1)
> > };
> >
> > +struct xfs_btree_level {
> > + /* buffer pointer */
> > + struct xfs_buf *bp;
> > +
> > + /* key/record number */
> > + uint16_t ptr;
> > +
> > + /* readahead info */
> > +#define XFS_BTCUR_LEFTRA 1 /* left sibling has been read-ahead */
> > +#define XFS_BTCUR_RIGHTRA 2 /* right sibling has been read-ahead */
> > + uint16_t ra;
> > +};
>
> The ra variable is a bit field. Can we define the values obviously
> as bit fields with (1 << 0) and (1 << 1) instead of 1 and 2?
Done.
> > @@ -242,8 +250,17 @@ struct xfs_btree_cur
> > struct xfs_btree_cur_ag bc_ag;
> > struct xfs_btree_cur_ino bc_ino;
> > };
> > +
> > + /* Must be at the end of the struct! */
> > + struct xfs_btree_level bc_levels[];
> > };
> >
> > +static inline size_t
> > +xfs_btree_cur_sizeof(unsigned int nlevels)
> > +{
> > + return struct_size((struct xfs_btree_cur *)NULL, bc_levels, nlevels);
> > +}
>
> Ooooh, yeah, we really need comments explaining how many btree
> levels these VLAs are tracking, because this one doesn't have a "-
> 1" in it like the previous one I commented on....
/*
* Compute the size of a btree cursor that can handle a btree of a given
* height. The bc_levels array handles node and leaf blocks, so its
* size is exactly nlevels.
*/
> > diff --git a/fs/xfs/scrub/trace.c b/fs/xfs/scrub/trace.c
> > index c0ef53fe6611..816dfc8e5a80 100644
> > --- a/fs/xfs/scrub/trace.c
> > +++ b/fs/xfs/scrub/trace.c
> > @@ -21,10 +21,11 @@ xchk_btree_cur_fsbno(
> > struct xfs_btree_cur *cur,
> > int level)
> > {
> > - if (level < cur->bc_nlevels && cur->bc_bufs[level])
> > + if (level < cur->bc_nlevels && cur->bc_levels[level].bp)
> > return XFS_DADDR_TO_FSB(cur->bc_mp,
> > - xfs_buf_daddr(cur->bc_bufs[level]));
> > - if (level == cur->bc_nlevels - 1 && cur->bc_flags & XFS_BTREE_LONG_PTRS)
> > + xfs_buf_daddr(cur->bc_levels[level].bp));
> > + else if (level == cur->bc_nlevels - 1 &&
> > + cur->bc_flags & XFS_BTREE_LONG_PTRS)
>
> No need for an else there as the first if () clause returns.
> Also, needs more () around that "a & b" second line.
TBH I think we check the wrong flag, and that last bit should be:
if (level == cur->bc_nlevels - 1 &&
(cur->bc_flags & XFS_BTREE_ROOT_IN_INODE))
return XFS_INO_TO_FSB(cur->bc_mp, cur->bc_ino.ip->i_ino);
return NULLFSBLOCK;
But for now I'll stick to the straight replacement and tack on another
patch to fix that.
--D
>
> Cheers,
>
> Dave.
> --
> Dave Chinner
> david@xxxxxxxxxxxxx
