Re: [PATCH 2/2] Btrfs-progs: use bitfield instead of integer for some variants in fs_info

[David Sterba <dsterba@suse.cz>]

On Wed, Feb 26, 2014 at 10:25:24AM +0800, Wang Shilong wrote:
> On 02/26/2014 12:28 AM, David Sterba wrote:
> >On Tue, Feb 25, 2014 at 07:48:57PM +0800, Wang Shilong wrote:
> >>Signed-off-by: Wang Shilong <wangsl.fnst@cn.fujitsu.com>
> >>---
> >>  ctree.h | 8 +++++---
> >>  1 file changed, 5 insertions(+), 3 deletions(-)
> >>
> >>diff --git a/ctree.h b/ctree.h
> >>index 3cc3477..9b461af 100644
> >>--- a/ctree.h
> >>+++ b/ctree.h
> >>@@ -984,9 +984,11 @@ struct btrfs_fs_info {
> >>  	struct btrfs_fs_devices *fs_devices;
> >>  	struct list_head space_info;
> >>  	int system_allocs;
> >>-	int readonly;
> >>-	int on_restoring;
> >>-	int is_chunk_recover;
> >>+
> >>+	unsigned int readonly:1;
> >>+	unsigned int on_restoring:1;
> >>+	unsigned int is_chunk_recover:1;
> >Well, the integers are wasteful, but there's only one instance of
> >fs_info per fsck run, so it saves like 8 bytes in total. I'm not sure
> >this patch is needed.
> Originally, i wrote this patch, it is because i was going to add a new flag.
> However, that approach has been replaced by a better idea.
> 
> Anyway, this patch will be a good start, however, i don't insist on it.
> We can also change it  when we have to add another flag later.

Ok, let's keep it, as an example of the pattern 'bitfields for flags'.