Btrfs: Transaction commit: don't use filemap_fdatawait

After writing out all the remaining btree blocks in the transaction,
the commit code would use filemap_fdatawait to make sure it was all
on disk.  This means it would wait for blocks written by other procs
as well.

The new code walks the list of blocks for this transaction again
and waits only for those required by this transaction.

Signed-off-by: Chris Mason <chris.mason@oracle.com>

1 files changed, 31 insertions, 5 deletions

diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index 9d3d08e9f8d..6bcb0876f9b 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -303,12 +303,12 @@ int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans,
 				     struct btrfs_root *root)
 {
 	int ret;
-	int err;
+	int err = 0;
 	int werr = 0;
 	struct extent_io_tree *dirty_pages;
 	struct page *page;
 	struct inode *btree_inode = root->fs_info->btree_inode;
-	u64 start;
+	u64 start = 0;
 	u64 end;
 	unsigned long index;
 
@@ -317,12 +317,15 @@ int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans,
 	}
 	dirty_pages = &trans->transaction->dirty_pages;
 	while(1) {
-		ret = find_first_extent_bit(dirty_pages, 0, &start, &end,
+		ret = find_first_extent_bit(dirty_pages, start, &start, &end,
 					    EXTENT_DIRTY);
 		if (ret)
 			break;
-		clear_extent_dirty(dirty_pages, start, end, GFP_NOFS);
 		while(start <= end) {
+			if (btrfs_congested_async(root->fs_info, 0))
+				congestion_wait(WRITE, HZ/10);
+			cond_resched();
+
 			index = start >> PAGE_CACHE_SHIFT;
 			start = (u64)(index + 1) << PAGE_CACHE_SHIFT;
 			page = find_lock_page(btree_inode->i_mapping, index);
@@ -343,7 +346,30 @@ int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans,
 			page_cache_release(page);
 		}
 	}
-	err = filemap_fdatawait(btree_inode->i_mapping);
+	while(1) {
+		ret = find_first_extent_bit(dirty_pages, 0, &start, &end,
+					    EXTENT_DIRTY);
+		if (ret)
+			break;
+
+		clear_extent_dirty(dirty_pages, start, end, GFP_NOFS);
+		while(start <= end) {
+			index = start >> PAGE_CACHE_SHIFT;
+			start = (u64)(index + 1) << PAGE_CACHE_SHIFT;
+			page = find_get_page(btree_inode->i_mapping, index);
+			if (!page)
+				continue;
+			if (PageDirty(page)) {
+				lock_page(page);
+				err = write_one_page(page, 0);
+				if (err)
+					werr = err;
+			}
+			wait_on_page_writeback(page);
+			page_cache_release(page);
+			cond_resched();
+		}
+	}
 	if (err)
 		werr = err;
 	return werr;