Minki/linux
1
0
Fork 1
mirror of https://github.com/torvalds/linux.git synced 2024-11-12 23:23:03 +00:00
Code Issues Packages Projects Releases Wiki Activity

Btrfs: faster/more efficient insertion of file extent items

Browse Source 
This is an extension to my previous commit titled:

  "Btrfs: faster file extent item replace operations"
  (hash 1acae57b16)

Instead of inserting the new file extent item if we deleted existing
file extent items covering our target file range, also allow to insert
the new file extent item if we didn't find any existing items to delete
and replace_extent != 0, since in this case our caller would do another
tree search to insert the new file extent item anyway, therefore just
combine the two tree searches into a single one, saving cpu time, reducing
lock contention and reducing btree node/leaf COW operations.

This covers the case where applications keep doing tail append writes to
files, which for example is the case of Apache CouchDB (its database and
view index files are always open with O_APPEND).

Signed-off-by: Filipe David Borba Manana <fdmanana@gmail.com>
Signed-off-by: Josef Bacik <jbacik@fb.com>
This commit is contained in:
Filipe David Borba Manana 2014-02-09 23:45:12 +00:00 committed by Josef Bacik
parent 51b98effa4
commit d5f375270a
1 changed files with 29 additions and 21 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

50
fs/btrfs/file.c
View File

@ -720,7 +720,7 @@ int __btrfs_drop_extents(struct btrfs_trans_handle *trans,
if (drop_cache)
btrfs_drop_extent_cache(inode, start, end - 1, 0);
if (start >= BTRFS_I(inode)->disk_i_size)
if (start >= BTRFS_I(inode)->disk_i_size && !replace_extent)
modify_tree = 0;
while (1) {
@ -938,34 +938,42 @@ next_slot:
* Set path->slots[0] to first slot, so that after the delete
* if items are move off from our leaf to its immediate left or
* right neighbor leafs, we end up with a correct and adjusted
* path->slots[0] for our insertion.
* path->slots[0] for our insertion (if replace_extent != 0).
*/
path->slots[0] = del_slot;
ret = btrfs_del_items(trans, root, path, del_slot, del_nr);
if (ret)
btrfs_abort_transaction(trans, root, ret);
}
leaf = path->nodes[0];
/*
* leaf eb has flag EXTENT_BUFFER_STALE if it was deleted (that
* is, its contents got pushed to its neighbors), in which case
* it means path->locks[0] == 0
*/
if (!ret && replace_extent && leafs_visited == 1 &&
path->locks[0] &&
btrfs_leaf_free_space(root, leaf) >=
sizeof(struct btrfs_item) + extent_item_size) {
leaf = path->nodes[0];
/*
* If btrfs_del_items() was called, it might have deleted a leaf, in
* which case it unlocked our path, so check path->locks[0] matches a
* write lock.
*/
if (!ret && replace_extent && leafs_visited == 1 &&
(path->locks[0] == BTRFS_WRITE_LOCK_BLOCKING ||
path->locks[0] == BTRFS_WRITE_LOCK) &&
btrfs_leaf_free_space(root, leaf) >=
sizeof(struct btrfs_item) + extent_item_size) {
key.objectid = ino;
key.type = BTRFS_EXTENT_DATA_KEY;
key.offset = start;
setup_items_for_insert(root, path, &key,
&extent_item_size,
extent_item_size,
sizeof(struct btrfs_item) +
extent_item_size, 1);
*key_inserted = 1;
key.objectid = ino;
key.type = BTRFS_EXTENT_DATA_KEY;
key.offset = start;
if (!del_nr && path->slots[0] < btrfs_header_nritems(leaf)) {
struct btrfs_key slot_key;
btrfs_item_key_to_cpu(leaf, &slot_key, path->slots[0]);
if (btrfs_comp_cpu_keys(&key, &slot_key) > 0)
path->slots[0]++;
}
setup_items_for_insert(root, path, &key,
&extent_item_size,
extent_item_size,
sizeof(struct btrfs_item) +
extent_item_size, 1);
*key_inserted = 1;
}
if (!replace_extent || !(*key_inserted))