Btrfs: New data=ordered implementation
The old data=ordered code would force commit to wait until
all the data extents from the transaction were fully on disk. This
introduced large latencies into the commit and stalled new writers
in the transaction for a long time.
The new code changes the way data allocations and extents work:
* When delayed allocation is filled, data extents are reserved, and
the extent bit EXTENT_ORDERED is set on the entire range of the extent.
A struct btrfs_ordered_extent is allocated an inserted into a per-inode
rbtree to track the pending extents.
* As each page is written EXTENT_ORDERED is cleared on the bytes corresponding
to that page.
* When all of the bytes corresponding to a single struct btrfs_ordered_extent
are written, The previously reserved extent is inserted into the FS
btree and into the extent allocation trees. The checksums for the file
data are also updated.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
index 40a5f53..3f82a6e 100644
--- a/fs/btrfs/extent_io.c
+++ b/fs/btrfs/extent_io.c
@@ -793,6 +793,13 @@
}
EXPORT_SYMBOL(set_extent_dirty);
+int set_extent_ordered(struct extent_io_tree *tree, u64 start, u64 end,
+ gfp_t mask)
+{
+ return set_extent_bit(tree, start, end, EXTENT_ORDERED, 0, NULL, mask);
+}
+EXPORT_SYMBOL(set_extent_ordered);
+
int set_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
int bits, gfp_t mask)
{
@@ -812,8 +819,8 @@
gfp_t mask)
{
return set_extent_bit(tree, start, end,
- EXTENT_DELALLOC | EXTENT_DIRTY, 0, NULL,
- mask);
+ EXTENT_DELALLOC | EXTENT_DIRTY,
+ 0, NULL, mask);
}
EXPORT_SYMBOL(set_extent_delalloc);
@@ -825,6 +832,13 @@
}
EXPORT_SYMBOL(clear_extent_dirty);
+int clear_extent_ordered(struct extent_io_tree *tree, u64 start, u64 end,
+ gfp_t mask)
+{
+ return clear_extent_bit(tree, start, end, EXTENT_ORDERED, 1, 0, mask);
+}
+EXPORT_SYMBOL(clear_extent_ordered);
+
int set_extent_new(struct extent_io_tree *tree, u64 start, u64 end,
gfp_t mask)
{
@@ -1395,10 +1409,9 @@
if (--bvec >= bio->bi_io_vec)
prefetchw(&bvec->bv_page->flags);
-
if (tree->ops && tree->ops->writepage_end_io_hook) {
ret = tree->ops->writepage_end_io_hook(page, start,
- end, state);
+ end, state, uptodate);
if (ret)
uptodate = 0;
}
@@ -1868,9 +1881,14 @@
unlock_extent(tree, cur, end, GFP_NOFS);
break;
}
-
extent_offset = cur - em->start;
+ if (extent_map_end(em) <= cur) {
+printk("bad mapping em [%Lu %Lu] cur %Lu\n", em->start, extent_map_end(em), cur);
+ }
BUG_ON(extent_map_end(em) <= cur);
+ if (end < cur) {
+printk("2bad mapping end %Lu cur %Lu\n", end, cur);
+ }
BUG_ON(end < cur);
iosize = min(extent_map_end(em) - cur, end - cur + 1);
@@ -1976,6 +1994,7 @@
u64 last_byte = i_size_read(inode);
u64 block_start;
u64 iosize;
+ u64 unlock_start;
sector_t sector;
struct extent_map *em;
struct block_device *bdev;
@@ -1988,7 +2007,6 @@
u64 nr_delalloc;
u64 delalloc_end;
-
WARN_ON(!PageLocked(page));
page_offset = i_size & (PAGE_CACHE_SIZE - 1);
if (page->index > end_index ||
@@ -2030,6 +2048,7 @@
delalloc_start = delalloc_end + 1;
}
lock_extent(tree, start, page_end, GFP_NOFS);
+ unlock_start = start;
end = page_end;
if (test_range_bit(tree, start, page_end, EXTENT_DELALLOC, 0)) {
@@ -2038,6 +2057,11 @@
if (last_byte <= start) {
clear_extent_dirty(tree, start, page_end, GFP_NOFS);
+ unlock_extent(tree, start, page_end, GFP_NOFS);
+ if (tree->ops && tree->ops->writepage_end_io_hook)
+ tree->ops->writepage_end_io_hook(page, start,
+ page_end, NULL, 1);
+ unlock_start = page_end + 1;
goto done;
}
@@ -2047,6 +2071,11 @@
while (cur <= end) {
if (cur >= last_byte) {
clear_extent_dirty(tree, cur, page_end, GFP_NOFS);
+ unlock_extent(tree, unlock_start, page_end, GFP_NOFS);
+ if (tree->ops && tree->ops->writepage_end_io_hook)
+ tree->ops->writepage_end_io_hook(page, cur,
+ page_end, NULL, 1);
+ unlock_start = page_end + 1;
break;
}
em = epd->get_extent(inode, page, page_offset, cur,
@@ -2071,8 +2100,16 @@
block_start == EXTENT_MAP_INLINE) {
clear_extent_dirty(tree, cur,
cur + iosize - 1, GFP_NOFS);
+
+ unlock_extent(tree, unlock_start, cur + iosize -1,
+ GFP_NOFS);
+ if (tree->ops && tree->ops->writepage_end_io_hook)
+ tree->ops->writepage_end_io_hook(page, cur,
+ cur + iosize - 1,
+ NULL, 1);
cur = cur + iosize;
page_offset += iosize;
+ unlock_start = cur;
continue;
}
@@ -2119,7 +2156,8 @@
set_page_writeback(page);
end_page_writeback(page);
}
- unlock_extent(tree, start, page_end, GFP_NOFS);
+ if (unlock_start <= page_end)
+ unlock_extent(tree, unlock_start, page_end, GFP_NOFS);
unlock_page(page);
return 0;
}