btrfs-progs: mkfs: enable NO_HOLES by default

The no-holes feature reduces consumption of metadata by not representing
file holes. Reducing metadata is a good thing in general, this is the
main goal to enable this by default.

There's a drawback, related to the missing information about holes. The
'check' tool cannot use it to cross-reference extent information and in
some cases may not be able to detect a problem.

The no-hole feature can be also enabled by 'btrfstune -n' on an
unmounted filesystem.

Issue: #405
Signed-off-by: David Sterba <dsterba@suse.com>

Documentation/mkfs.btrfs.asciidoc

@@ -241,7 +241,7 @@ are selected
 reduced-size metadata for extent references, saves a few percent of metadata
 
 *no-holes*::
-(kernel support since 3.14)
+(default since btrfs-progs 5.15, kernel support since 3.14)
 +
 improved representation of file extents where holes are not explicitly
 stored as an extent, saves a few percent of metadata if sparse files are used

common/fsfeatures.c

@@ -110,7 +110,7 @@ static const struct btrfs_feature mkfs_features[] = {
 		.sysfs_name	= "no_holes",
 		VERSION_TO_STRING2(compat, 3,14),
 		VERSION_TO_STRING2(safe, 4,0),
-		VERSION_NULL(default),
+		VERSION_TO_STRING2(default, 5,15),
 		.desc		= "no explicit hole extents for files"
 	}, {
 		.name		= "raid1c34",

common/fsfeatures.h

@@ -22,6 +22,7 @@
 #define BTRFS_MKFS_DEFAULT_NODE_SIZE SZ_16K
 #define BTRFS_MKFS_DEFAULT_FEATURES 				\
 		(BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF		\
+		| BTRFS_FEATURE_INCOMPAT_NO_HOLES		\
 		| BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA)
 
 /*