1 files changed, 73 insertions, 16 deletions

diff --git a/fs/btrfs/reflink.c b/fs/btrfs/reflink.c
index b24396cf2f99..f4ec06b53aa0 100644
--- a/fs/btrfs/reflink.c
+++ b/fs/btrfs/reflink.c
@@ -106,12 +106,8 @@ static int copy_inline_to_page(struct btrfs_inode *inode,
 	set_bit(BTRFS_INODE_NO_DELALLOC_FLUSH, &inode->runtime_flags);
 
 	if (comp_type == BTRFS_COMPRESS_NONE) {
-		char *map;
-
-		map = kmap(page);
-		memcpy(map, data_start, datal);
+		memcpy_to_page(page, 0, data_start, datal);
 		flush_dcache_page(page);
-		kunmap(page);
 	} else {
 		ret = btrfs_decompress(comp_type, data_start, page, 0,
 				       inline_size, datal);
@@ -482,9 +478,9 @@ process_slot:
 			clone_info.file_offset = new_key.offset;
 			clone_info.extent_buf = buf;
 			clone_info.is_new_extent = false;
-			ret = btrfs_replace_file_extents(inode, path, drop_start,
-					new_key.offset + datal - 1, &clone_info,
-					&trans);
+			ret = btrfs_replace_file_extents(BTRFS_I(inode), path,
+					drop_start, new_key.offset + datal - 1,
+					&clone_info, &trans);
 			if (ret)
 				goto out;
 		} else if (type == BTRFS_FILE_EXTENT_INLINE) {
@@ -553,8 +549,26 @@ process_slot:
 		 */
 		btrfs_release_path(path);
 
-		ret = btrfs_replace_file_extents(inode, path, last_dest_end,
-				destoff + len - 1, NULL, &trans);
+		/*
+		 * When using NO_HOLES and we are cloning a range that covers
+		 * only a hole (no extents) into a range beyond the current
+		 * i_size, punching a hole in the target range will not create
+		 * an extent map defining a hole, because the range starts at or
+		 * beyond current i_size. If the file previously had an i_size
+		 * greater than the new i_size set by this clone operation, we
+		 * need to make sure the next fsync is a full fsync, so that it
+		 * detects and logs a hole covering a range from the current
+		 * i_size to the new i_size. If the clone range covers extents,
+		 * besides a hole, then we know the full sync flag was already
+		 * set by previous calls to btrfs_replace_file_extents() that
+		 * replaced file extent items.
+		 */
+		if (last_dest_end >= i_size_read(inode))
+			set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
+				&BTRFS_I(inode)->runtime_flags);
+
+		ret = btrfs_replace_file_extents(BTRFS_I(inode), path,
+				last_dest_end, destoff + len - 1, NULL, &trans);
 		if (ret)
 			goto out;
 
@@ -590,6 +604,20 @@ static void btrfs_double_extent_lock(struct inode *inode1, u64 loff1,
 	lock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
 }
 
+static void btrfs_double_mmap_lock(struct inode *inode1, struct inode *inode2)
+{
+	if (inode1 < inode2)
+		swap(inode1, inode2);
+	down_write(&BTRFS_I(inode1)->i_mmap_lock);
+	down_write_nested(&BTRFS_I(inode2)->i_mmap_lock, SINGLE_DEPTH_NESTING);
+}
+
+static void btrfs_double_mmap_unlock(struct inode *inode1, struct inode *inode2)
+{
+	up_write(&BTRFS_I(inode1)->i_mmap_lock);
+	up_write(&BTRFS_I(inode2)->i_mmap_lock);
+}
+
 static int btrfs_extent_same_range(struct inode *src, u64 loff, u64 len,
 				   struct inode *dst, u64 dst_loff)
 {
@@ -806,6 +834,16 @@ static int btrfs_remap_file_range_prep(struct file *file_in, loff_t pos_in,
 					    len, remap_flags);
 }
 
+static bool file_sync_write(const struct file *file)
+{
+	if (file->f_flags & (__O_SYNC | O_DSYNC))
+		return true;
+	if (IS_SYNC(file_inode(file)))
+		return true;
+
+	return false;
+}
+
 loff_t btrfs_remap_file_range(struct file *src_file, loff_t off,
 		struct file *dst_file, loff_t destoff, loff_t len,
 		unsigned int remap_flags)
@@ -818,10 +856,12 @@ loff_t btrfs_remap_file_range(struct file *src_file, loff_t off,
 	if (remap_flags & ~(REMAP_FILE_DEDUP | REMAP_FILE_ADVISORY))
 		return -EINVAL;
 
-	if (same_inode)
-		inode_lock(src_inode);
-	else
+	if (same_inode) {
+		btrfs_inode_lock(src_inode, BTRFS_ILOCK_MMAP);
+	} else {
 		lock_two_nondirectories(src_inode, dst_inode);
+		btrfs_double_mmap_lock(src_inode, dst_inode);
+	}
 
 	ret = btrfs_remap_file_range_prep(src_file, off, dst_file, destoff,
 					  &len, remap_flags);
@@ -834,10 +874,27 @@ loff_t btrfs_remap_file_range(struct file *src_file, loff_t off,
 		ret = btrfs_clone_files(dst_file, src_file, off, len, destoff);
 
 out_unlock:
-	if (same_inode)
-		inode_unlock(src_inode);
-	else
+	if (same_inode) {
+		btrfs_inode_unlock(src_inode, BTRFS_ILOCK_MMAP);
+	} else {
+		btrfs_double_mmap_unlock(src_inode, dst_inode);
 		unlock_two_nondirectories(src_inode, dst_inode);
+	}
+
+	/*
+	 * If either the source or the destination file was opened with O_SYNC,
+	 * O_DSYNC or has the S_SYNC attribute, fsync both the destination and
+	 * source files/ranges, so that after a successful return (0) followed
+	 * by a power failure results in the reflinked data to be readable from
+	 * both files/ranges.
+	 */
+	if (ret == 0 && len > 0 &&
+	    (file_sync_write(src_file) || file_sync_write(dst_file))) {
+		ret = btrfs_sync_file(src_file, off, off + len - 1, 0);
+		if (ret == 0)
+			ret = btrfs_sync_file(dst_file, destoff,
+					      destoff + len - 1, 0);
+	}
 
 	return ret < 0 ? ret : len;
 }