aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorDave Chinner <david@fromorbit.com>2010-09-02 01:14:38 -0400
committerDave Chinner <david@fromorbit.com>2010-09-02 01:14:38 -0400
commit9bc08a45fb117c696e4940cfa1208cb1cc7a2f25 (patch)
tree610e4cb520d62c4ad6ae0f20ddd64cd15520c33a
parent2bfc96a127bc1cc94d26bfaa40159966064f9c8c (diff)
xfs: improve buffer cache hash scalability
When doing large parallel file creates on a 16p machines, large amounts of time is being spent in _xfs_buf_find(). A system wide profile with perf top shows this: 1134740.00 19.3% _xfs_buf_find 733142.00 12.5% __ticket_spin_lock The problem is that the hash contains 45,000 buffers, and the hash table width is only 256 buffers. That means we've got around 200 buffers per chain, and searching it is quite expensive. The hash table size needs to increase. Secondly, every time we do a lookup, we promote the buffer we find to the head of the hash chain. This is causing cachelines to be dirtied and causes invalidation of cachelines across all CPUs that may have walked the hash chain recently. hence every walk of the hash chain is effectively a cold cache walk. Remove the promotion to avoid this invalidation. The results are: 1045043.00 21.2% __ticket_spin_lock 326184.00 6.6% _xfs_buf_find A 70% drop in the CPU usage when looking up buffers. Unfortunately that does not result in an increase in performance underthis workload as contention on the inode_lock soaks up most of the reduction in CPU usage. Signed-off-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de>
-rw-r--r--fs/xfs/linux-2.6/xfs_buf.c8
-rw-r--r--fs/xfs/linux-2.6/xfs_buf.h1
2 files changed, 1 insertions, 8 deletions
diff --git a/fs/xfs/linux-2.6/xfs_buf.c b/fs/xfs/linux-2.6/xfs_buf.c
index ea79072f5210..d72cf2bb054a 100644
--- a/fs/xfs/linux-2.6/xfs_buf.c
+++ b/fs/xfs/linux-2.6/xfs_buf.c
@@ -440,12 +440,7 @@ _xfs_buf_find(
440 ASSERT(btp == bp->b_target); 440 ASSERT(btp == bp->b_target);
441 if (bp->b_file_offset == range_base && 441 if (bp->b_file_offset == range_base &&
442 bp->b_buffer_length == range_length) { 442 bp->b_buffer_length == range_length) {
443 /*
444 * If we look at something, bring it to the
445 * front of the list for next time.
446 */
447 atomic_inc(&bp->b_hold); 443 atomic_inc(&bp->b_hold);
448 list_move(&bp->b_hash_list, &hash->bh_list);
449 goto found; 444 goto found;
450 } 445 }
451 } 446 }
@@ -1443,8 +1438,7 @@ xfs_alloc_bufhash(
1443{ 1438{
1444 unsigned int i; 1439 unsigned int i;
1445 1440
1446 btp->bt_hashshift = external ? 3 : 8; /* 8 or 256 buckets */ 1441 btp->bt_hashshift = external ? 3 : 12; /* 8 or 4096 buckets */
1447 btp->bt_hashmask = (1 << btp->bt_hashshift) - 1;
1448 btp->bt_hash = kmem_zalloc_large((1 << btp->bt_hashshift) * 1442 btp->bt_hash = kmem_zalloc_large((1 << btp->bt_hashshift) *
1449 sizeof(xfs_bufhash_t)); 1443 sizeof(xfs_bufhash_t));
1450 for (i = 0; i < (1 << btp->bt_hashshift); i++) { 1444 for (i = 0; i < (1 << btp->bt_hashshift); i++) {
diff --git a/fs/xfs/linux-2.6/xfs_buf.h b/fs/xfs/linux-2.6/xfs_buf.h
index d072e5ff923b..2a05614f0b92 100644
--- a/fs/xfs/linux-2.6/xfs_buf.h
+++ b/fs/xfs/linux-2.6/xfs_buf.h
@@ -137,7 +137,6 @@ typedef struct xfs_buftarg {
137 size_t bt_smask; 137 size_t bt_smask;
138 138
139 /* per device buffer hash table */ 139 /* per device buffer hash table */
140 uint bt_hashmask;
141 uint bt_hashshift; 140 uint bt_hashshift;
142 xfs_bufhash_t *bt_hash; 141 xfs_bufhash_t *bt_hash;
143 142