lk/linux-kernel-test
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Btrfs: switch extent_map to a rw lock

Browse Source 
There are two main users of the extent_map tree.  The
first is regular file inodes, where it is evenly spread
between readers and writers.

The second is the chunk allocation tree, which maps blocks from
logical addresses to phyiscal ones, and it is 99.99% reads.

The mapping tree is a point of lock contention during heavy IO
workloads, so this commit switches things to a rw lock.

Signed-off-by: Chris Mason <chris.mason@oracle.com>
This commit is contained in:
Chris Mason
2009-09-02 16:24:52 -04:00
parent 57fd5a5ff8
commit 890871be85
10 changed files with 57 additions and 60 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
fs/btrfs/file.c
View File

@ -188,15 +188,15 @@ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end,
if (!split2)
split2 = alloc_extent_map(GFP_NOFS);
spin_lock(&em_tree->lock);
write_lock(&em_tree->lock);
em = lookup_extent_mapping(em_tree, start, len);
if (!em) {
spin_unlock(&em_tree->lock);
write_unlock(&em_tree->lock);
break;
}
flags = em->flags;
if (skip_pinned && test_bit(EXTENT_FLAG_PINNED, &em->flags)) {
spin_unlock(&em_tree->lock);
write_unlock(&em_tree->lock);
if (em->start <= start &&
(!testend || em->start + em->len >= start + len)) {
free_extent_map(em);
@ -259,7 +259,7 @@ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end,
free_extent_map(split);
split = NULL;
}
spin_unlock(&em_tree->lock);
write_unlock(&em_tree->lock);
/* once for us */
free_extent_map(em);