On Wed, Sep 18, 2019 at 11:56:16AM -0700, Darrick J. Wong wrote:
> On Mon, Sep 16, 2019 at 08:16:28AM -0400, Brian Foster wrote:
> > If the size lookup lands in the last block of the by-size btree, the
> > near mode algorithm scans the entire block for the extent with best
> > available locality. In preparation for similar best available
> > extent tracking across both btrees, extend the allocation cursor
> > with best extent data and lift the associated state from the cntbt
> > last block scan code. No functional changes.
> >
> > Signed-off-by: Brian Foster <bfoster@xxxxxxxxxx>
> > ---
> > fs/xfs/libxfs/xfs_alloc.c | 63 ++++++++++++++++++++-------------------
> > 1 file changed, 33 insertions(+), 30 deletions(-)
> >
> > diff --git a/fs/xfs/libxfs/xfs_alloc.c b/fs/xfs/libxfs/xfs_alloc.c
> > index 5c34d4c41761..ee46989ab723 100644
> > --- a/fs/xfs/libxfs/xfs_alloc.c
> > +++ b/fs/xfs/libxfs/xfs_alloc.c
> > @@ -716,6 +716,11 @@ struct xfs_alloc_cur {
> > struct xfs_btree_cur *cnt; /* btree cursors */
> > struct xfs_btree_cur *bnolt;
> > struct xfs_btree_cur *bnogt;
> > + xfs_agblock_t rec_bno;/* extent startblock */
> > + xfs_extlen_t rec_len;/* extent length */
> > + xfs_agblock_t bno; /* alloc bno */
> > + xfs_extlen_t len; /* alloc len */
> > + xfs_extlen_t diff; /* diff from search bno */
> > unsigned busy_gen;/* busy state */
> > bool busy;
> > };
> > @@ -735,6 +740,11 @@ xfs_alloc_cur_setup(
> >
> > ASSERT(args->alignment == 1 || args->type != XFS_ALLOCTYPE_THIS_BNO);
> >
> > + acur->rec_bno = 0;
> > + acur->rec_len = 0;
> > + acur->bno = 0;
> > + acur->len = 0;
> > + acur->diff = -1;
>
> xfs_extlen_t is a uint32_t; is this going to cause comparison problems?
>
How so? This is the distance (in FSB) between the allocation hint and
the current candidate extent. I didn't think that could exceed a 32-bit
value, but I could be missing something (if so, that sounds like a bug
fix for a separate patch since this should be mostly refactoring).
> Also ... assuming that acur->diff is the successor to the bdiff variable
> below, shouldn't it be initialized to zero like bdiff was?
>
So I think the init to -1 carried over from my initial approach of
hacking on this from scratch. Initializing to a max possible diff seemed
more clear and less fragile than 0, which basically means perfect
locality. Looking at this patch in isolation, it seems it doesn't matter
how diff is initialized because acur->len is zero until we find the
first candidate extent and assign a real value to acur->diff.
If you skip ahead to the next patch, however, you'll see the logic is
split up in xfs_alloc_cur_check() such that we depend on a "proper"
initialization of acur->diff to function correctly. I guess this is kind
of a quirk of how the patches were split up. I'll change it to init to 0
in this patch and switch it over to -1 in the next with an explanation
in the commit log...
Brian
> --D
>
> > acur->busy = false;
> > acur->busy_gen = 0;
> >
> > @@ -1247,10 +1257,7 @@ xfs_alloc_ag_vextent_near(
> > * but we never loop back to the top.
> > */
> > while (xfs_btree_islastblock(acur.cnt, 0)) {
> > - xfs_extlen_t bdiff;
> > - int besti=0;
> > - xfs_extlen_t blen=0;
> > - xfs_agblock_t bnew=0;
> > + xfs_extlen_t diff;
> >
> > #ifdef DEBUG
> > if (dofirst)
> > @@ -1281,8 +1288,8 @@ xfs_alloc_ag_vextent_near(
> > break;
> > }
> > i = acur.cnt->bc_ptrs[0];
> > - for (j = 1, blen = 0, bdiff = 0;
> > - !error && j && (blen < args->maxlen || bdiff > 0);
> > + for (j = 1;
> > + !error && j && (acur.len < args->maxlen || acur.diff > 0);
> > error = xfs_btree_increment(acur.cnt, 0, &j)) {
> > /*
> > * For each entry, decide if it's better than
> > @@ -1301,44 +1308,40 @@ xfs_alloc_ag_vextent_near(
> > args->len = XFS_EXTLEN_MIN(ltlena, args->maxlen);
> > xfs_alloc_fix_len(args);
> > ASSERT(args->len >= args->minlen);
> > - if (args->len < blen)
> > + if (args->len < acur.len)
> > continue;
> > - ltdiff = xfs_alloc_compute_diff(args->agbno, args->len,
> > + diff = xfs_alloc_compute_diff(args->agbno, args->len,
> > args->alignment, args->datatype, ltbnoa,
> > ltlena, <new);
> > if (ltnew != NULLAGBLOCK &&
> > - (args->len > blen || ltdiff < bdiff)) {
> > - bdiff = ltdiff;
> > - bnew = ltnew;
> > - blen = args->len;
> > - besti = acur.cnt->bc_ptrs[0];
> > + (args->len > acur.len || diff < acur.diff)) {
> > + acur.rec_bno = ltbno;
> > + acur.rec_len = ltlen;
> > + acur.diff = diff;
> > + acur.bno = ltnew;
> > + acur.len = args->len;
> > }
> > }
> > /*
> > * It didn't work. We COULD be in a case where
> > * there's a good record somewhere, so try again.
> > */
> > - if (blen == 0)
> > + if (acur.len == 0)
> > break;
> > - /*
> > - * Point at the best entry, and retrieve it again.
> > - */
> > - acur.cnt->bc_ptrs[0] = besti;
> > - error = xfs_alloc_get_rec(acur.cnt, <bno, <len, &i);
> > - if (error)
> > - goto out;
> > - XFS_WANT_CORRUPTED_GOTO(args->mp, i == 1, out);
> > - ASSERT(ltbno + ltlen <= be32_to_cpu(XFS_BUF_TO_AGF(args->agbp)->agf_length));
> > - args->len = blen;
> >
> > /*
> > - * We are allocating starting at bnew for blen blocks.
> > + * Allocate at the bno/len tracked in the cursor.
> > */
> > - args->agbno = bnew;
> > - ASSERT(bnew >= ltbno);
> > - ASSERT(bnew + blen <= ltbno + ltlen);
> > - error = xfs_alloc_fixup_trees(acur.cnt, acur.bnolt, ltbno,
> > - ltlen, bnew, blen, XFSA_FIXUP_CNT_OK);
> > + args->agbno = acur.bno;
> > + args->len = acur.len;
> > + ASSERT(acur.bno >= acur.rec_bno);
> > + ASSERT(acur.bno + acur.len <= acur.rec_bno + acur.rec_len);
> > + ASSERT(acur.rec_bno + acur.rec_len <=
> > + be32_to_cpu(XFS_BUF_TO_AGF(args->agbp)->agf_length));
> > +
> > + error = xfs_alloc_fixup_trees(acur.cnt, acur.bnolt,
> > + acur.rec_bno, acur.rec_len, acur.bno, acur.len,
> > + 0);
> > if (error)
> > goto out;
> > trace_xfs_alloc_near_first(args);
> > --
> > 2.20.1
> >
