Btrfs: turbo charge fsync
At least for the vm workload. Currently on fsync we will
1) Truncate all items in the log tree for the given inode if they exist
and
2) Copy all items for a given inode into the log
The problem with this is that for things like VMs you can have lots of
extents from the fragmented writing behavior, and worst yet you may have
only modified a few extents, not the entire thing. This patch fixes this
problem by tracking which transid modified our extent, and then when we do
the tree logging we find all of the extents we've modified in our current
transaction, sort them and commit them. We also only truncate up to the
xattrs of the inode and copy that stuff in normally, and then just drop any
extents in the range we have that exist in the log already. Here are some
numbers of a 50 meg fio job that does random writes and fsync()s after every
write
Original Patched
SATA drive 82KB/s 140KB/s
Fusion drive 431KB/s 2532KB/s
So around 2-6 times faster depending on your hardware. There are a few
corner cases, for example if you truncate at all we have to do it the old
way since there is no way to be sure what is in the log is ok. This
probably could be done smarter, but if you write-fsync-truncate-write-fsync
you deserve what you get. All this work is in RAM of course so if your
inode gets evicted from cache and you read it in and fsync it we'll do it
the slow way if we are still in the same transaction that we last modified
the inode in.
The biggest cool part of this is that it requires no changes to the recovery
code, so if you fsync with this patch and crash and load an old kernel, it
will run the recovery and be a-ok. I have tested this pretty thoroughly
with an fsync tester and everything comes back fine, as well as xfstests.
Thanks,
Signed-off-by: Josef Bacik <jbacik@fusionio.com>
diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c
index 7c97b33..1fe82cf 100644
--- a/fs/btrfs/extent_map.c
+++ b/fs/btrfs/extent_map.c
@@ -35,6 +35,7 @@
void extent_map_tree_init(struct extent_map_tree *tree)
{
tree->map = RB_ROOT;
+ INIT_LIST_HEAD(&tree->modified_extents);
rwlock_init(&tree->lock);
}
@@ -54,7 +55,9 @@
em->in_tree = 0;
em->flags = 0;
em->compress_type = BTRFS_COMPRESS_NONE;
+ em->generation = 0;
atomic_set(&em->refs, 1);
+ INIT_LIST_HEAD(&em->list);
return em;
}
@@ -72,6 +75,7 @@
WARN_ON(atomic_read(&em->refs) == 0);
if (atomic_dec_and_test(&em->refs)) {
WARN_ON(em->in_tree);
+ WARN_ON(!list_empty(&em->list));
kmem_cache_free(extent_map_cache, em);
}
}
@@ -198,6 +202,12 @@
em->block_len += merge->block_len;
em->block_start = merge->block_start;
merge->in_tree = 0;
+ if (merge->generation > em->generation) {
+ em->generation = merge->generation;
+ list_move(&em->list, &tree->modified_extents);
+ }
+
+ list_del_init(&merge->list);
rb_erase(&merge->rb_node, &tree->map);
free_extent_map(merge);
}
@@ -211,11 +221,29 @@
em->block_len += merge->len;
rb_erase(&merge->rb_node, &tree->map);
merge->in_tree = 0;
+ if (merge->generation > em->generation) {
+ em->generation = merge->generation;
+ list_move(&em->list, &tree->modified_extents);
+ }
+ list_del_init(&merge->list);
free_extent_map(merge);
}
}
-int unpin_extent_cache(struct extent_map_tree *tree, u64 start, u64 len)
+/**
+ * unpint_extent_cache - unpin an extent from the cache
+ * @tree: tree to unpin the extent in
+ * @start: logical offset in the file
+ * @len: length of the extent
+ * @gen: generation that this extent has been modified in
+ * @prealloc: if this is set we need to clear the prealloc flag
+ *
+ * Called after an extent has been written to disk properly. Set the generation
+ * to the generation that actually added the file item to the inode so we know
+ * we need to sync this extent when we call fsync().
+ */
+int unpin_extent_cache(struct extent_map_tree *tree, u64 start, u64 len,
+ u64 gen)
{
int ret = 0;
struct extent_map *em;
@@ -228,10 +256,11 @@
if (!em)
goto out;
+ list_move(&em->list, &tree->modified_extents);
+ em->generation = gen;
clear_bit(EXTENT_FLAG_PINNED, &em->flags);
try_merge_map(tree, em);
-
free_extent_map(em);
out:
write_unlock(&tree->lock);
@@ -358,6 +387,7 @@
WARN_ON(test_bit(EXTENT_FLAG_PINNED, &em->flags));
rb_erase(&em->rb_node, &tree->map);
+ list_del_init(&em->list);
em->in_tree = 0;
return ret;
}