Fix btrfs_wait_ordered_extent_range to properly wait

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

fs/btrfs/file-item.c

@@ -161,7 +161,8 @@ int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode,
 
 	while(bio_index < bio->bi_vcnt) {
 		offset = page_offset(bvec->bv_page) + bvec->bv_offset;
-		if (offset >= ordered->file_offset + ordered->len) {
+		if (offset >= ordered->file_offset + ordered->len ||
+		    offset < ordered->file_offset) {
 			unsigned long bytes_left;
 			sums->len = this_sum_bytes;
 			this_sum_bytes = 0;

fs/btrfs/inode.c

@@ -128,7 +128,9 @@ static int cow_file_range(struct inode *inode, u64 start, u64 end)
 		goto out;
 
 	BUG_ON(num_bytes > btrfs_super_total_bytes(&root->fs_info->super_copy));
+	mutex_lock(&BTRFS_I(inode)->extent_mutex);
 	btrfs_drop_extent_cache(inode, start, start + num_bytes - 1);
+	mutex_unlock(&BTRFS_I(inode)->extent_mutex);
 
 	while(num_bytes > 0) {
 		cur_alloc_size = min(num_bytes, root->fs_info->max_extent);
@@ -144,6 +146,7 @@ static int cow_file_range(struct inode *inode, u64 start, u64 end)
 		em->len = ins.offset;
 		em->block_start = ins.objectid;
 		em->bdev = root->fs_info->fs_devices->latest_bdev;
+		mutex_lock(&BTRFS_I(inode)->extent_mutex);
 		set_bit(EXTENT_FLAG_PINNED, &em->flags);
 		while(1) {
 			spin_lock(&em_tree->lock);
@@ -156,6 +159,7 @@ static int cow_file_range(struct inode *inode, u64 start, u64 end)
 			btrfs_drop_extent_cache(inode, start,
 						start + ins.offset - 1);
 		}
+		mutex_unlock(&BTRFS_I(inode)->extent_mutex);
 
 		cur_alloc_size = ins.offset;
 		ret = btrfs_add_ordered_extent(inode, start, ins.objectid,
@@ -487,6 +491,8 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end)
 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
 	struct extent_map *em;
 	u64 alloc_hint = 0;
+	u64 clear_start;
+	u64 clear_end;
 	struct list_head list;
 	struct btrfs_key ins;
 	int ret;
@@ -509,12 +515,14 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end)
 	ins.objectid = ordered_extent->start;
 	ins.offset = ordered_extent->len;
 	ins.type = BTRFS_EXTENT_ITEM_KEY;
+
 	ret = btrfs_alloc_reserved_extent(trans, root, root->root_key.objectid,
 					  trans->transid, inode->i_ino,
 					  ordered_extent->file_offset, &ins);
 	BUG_ON(ret);
 
 	mutex_lock(&BTRFS_I(inode)->extent_mutex);
+
 	ret = btrfs_drop_extents(trans, root, inode,
 				 ordered_extent->file_offset,
 				 ordered_extent->file_offset +
@@ -528,13 +536,19 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end)
 				       ordered_extent->len, 0);
 	BUG_ON(ret);
 
-
 	spin_lock(&em_tree->lock);
-	em = lookup_extent_mapping(em_tree, ordered_extent->file_offset,
-			       ordered_extent->len);
-	if (em) {
-		clear_bit(EXTENT_FLAG_PINNED, &em->flags);
-		free_extent_map(em);
+	clear_start = ordered_extent->file_offset;
+	clear_end = ordered_extent->file_offset + ordered_extent->len;
+	while(clear_start < clear_end) {
+		em = lookup_extent_mapping(em_tree, clear_start,
+					   clear_end - clear_start);
+		if (em) {
+			clear_bit(EXTENT_FLAG_PINNED, &em->flags);
+			clear_start = em->start + em->len;
+			free_extent_map(em);
+		} else {
+			break;
+		}
 	}
 	spin_unlock(&em_tree->lock);
 

fs/btrfs/ordered-data.c

@@ -324,22 +324,37 @@ void btrfs_start_ordered_extent(struct inode *inode,
 void btrfs_wait_ordered_range(struct inode *inode, u64 start, u64 len)
 {
 	u64 end;
+	u64 orig_end;
+	u64 wait_end;
 	struct btrfs_ordered_extent *ordered;
-	int found;
-	int should_wait = 0;
-
-again:
-	if (start + len < start)
-		end = (u64)-1;
-	else
-		end = start + len - 1;
-	found = 0;
+	u64 mask = BTRFS_I(inode)->root->sectorsize - 1;
+
+	if (start + len < start) {
+		wait_end = (inode->i_size + mask) & ~mask;
+		orig_end = (u64)-1;
+	} else {
+		orig_end = start + len - 1;
+		wait_end = orig_end;
+	}
+
+	/* start IO across the range first to instantiate any delalloc
+	 * extents
+	 */
+#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,22)
+	do_sync_file_range(file, start, wait_end, SYNC_FILE_RANGE_WRITE);
+#else
+	do_sync_mapping_range(inode->i_mapping, start, wait_end,
+			      SYNC_FILE_RANGE_WRITE);
+#endif
+	end = orig_end;
+	wait_on_extent_writeback(&BTRFS_I(inode)->io_tree, start, orig_end);
+
 	while(1) {
 		ordered = btrfs_lookup_first_ordered_extent(inode, end);
 		if (!ordered) {
 			break;
 		}
-		if (ordered->file_offset >= start + len) {
+		if (ordered->file_offset > orig_end) {
 			btrfs_put_ordered_extent(ordered);
 			break;
 		}
@@ -347,21 +362,15 @@ again:
 			btrfs_put_ordered_extent(ordered);
 			break;
 		}
-		btrfs_start_ordered_extent(inode, ordered, should_wait);
-		found++;
+		btrfs_start_ordered_extent(inode, ordered, 1);
 		end = ordered->file_offset;
 		btrfs_put_ordered_extent(ordered);
-		if (end == 0)
+		if (end == 0 || end == start)
 			break;
 		end--;
 	}
-	if (should_wait && found) {
-		should_wait = 0;
-		goto again;
-	}
 }
 
-
 /*
  * find an ordered extent corresponding to file_offset.  return NULL if
  * nothing is found, otherwise take a reference on the extent and return it