From 9bc08a45fb117c696e4940cfa1208cb1cc7a2f25 Mon Sep 17 00:00:00 2001 From: Dave Chinner Date: Thu, 2 Sep 2010 15:14:38 +1000 Subject: 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 Reviewed-by: Christoph Hellwig --- fs/xfs/linux-2.6/xfs_buf.c | 8 +------- fs/xfs/linux-2.6/xfs_buf.h | 1 - 2 files changed, 1 insertion(+), 8 deletions(-) (limited to 'fs') 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( ASSERT(btp == bp->b_target); if (bp->b_file_offset == range_base && bp->b_buffer_length == range_length) { - /* - * If we look at something, bring it to the - * front of the list for next time. - */ atomic_inc(&bp->b_hold); - list_move(&bp->b_hash_list, &hash->bh_list); goto found; } } @@ -1443,8 +1438,7 @@ xfs_alloc_bufhash( { unsigned int i; - btp->bt_hashshift = external ? 3 : 8; /* 8 or 256 buckets */ - btp->bt_hashmask = (1 << btp->bt_hashshift) - 1; + btp->bt_hashshift = external ? 3 : 12; /* 8 or 4096 buckets */ btp->bt_hash = kmem_zalloc_large((1 << btp->bt_hashshift) * sizeof(xfs_bufhash_t)); 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 { size_t bt_smask; /* per device buffer hash table */ - uint bt_hashmask; uint bt_hashshift; xfs_bufhash_t *bt_hash; -- cgit v1.2.3 From 72656c46f50b8dfe50e15793692982e636e3df20 Mon Sep 17 00:00:00 2001 From: Dave Chinner Date: Fri, 3 Sep 2010 12:19:33 +1000 Subject: xfs: prevent 32bit overflow in space reservation If we attempt to preallocate more than 2^32 blocks of space in a single syscall, the transaction block reservation will overflow leading to a hangs in the superblock block accounting code. This is trivially reproduced with xfs_io. Fix the problem by capping the allocation reservation to the maximum number of blocks a single xfs_bmapi() call can allocate (2^21 blocks). Signed-off-by: Dave Chinner Reviewed-by: Christoph Hellwig --- fs/xfs/xfs_vnodeops.c | 13 ++++++++++--- 1 file changed, 10 insertions(+), 3 deletions(-) (limited to 'fs') diff --git a/fs/xfs/xfs_vnodeops.c b/fs/xfs/xfs_vnodeops.c index 66d585c6917c..4c7c7bfb2b2f 100644 --- a/fs/xfs/xfs_vnodeops.c +++ b/fs/xfs/xfs_vnodeops.c @@ -2299,15 +2299,22 @@ xfs_alloc_file_space( e = allocatesize_fsb; } + /* + * The transaction reservation is limited to a 32-bit block + * count, hence we need to limit the number of blocks we are + * trying to reserve to avoid an overflow. We can't allocate + * more than @nimaps extents, and an extent is limited on disk + * to MAXEXTLEN (21 bits), so use that to enforce the limit. + */ + resblks = min_t(xfs_fileoff_t, (e - s), (MAXEXTLEN * nimaps)); if (unlikely(rt)) { - resrtextents = qblocks = (uint)(e - s); + resrtextents = qblocks = resblks; resrtextents /= mp->m_sb.sb_rextsize; resblks = XFS_DIOSTRAT_SPACE_RES(mp, 0); quota_flag = XFS_QMOPT_RES_RTBLKS; } else { resrtextents = 0; - resblks = qblocks = \ - XFS_DIOSTRAT_SPACE_RES(mp, (uint)(e - s)); + resblks = qblocks = XFS_DIOSTRAT_SPACE_RES(mp, resblks); quota_flag = XFS_QMOPT_RES_REGBLKS; } -- cgit v1.2.3