[CRIU] [PATCH 1/3] mount: Introduce BTRFS engine

[Cyrill Gorcunov]

On Fri, Nov 29, 2013 at 06:47:48PM +0400, Pavel Emelyanov wrote:
 
> > @@ -1803,6 +1807,7 @@ err:
> >  	free_pstree(root_item);
> >  	free_file_locks();
> >  	free_link_remaps();
> > +	btrfs_fini();
> >  
> >  	close_service_fd(CR_PROC_FD_OFF);
> >  
> 
> Not needed, as mount infos are not released either.

I belive we should implement cleaning of mountinfos as well.
But ok, i'll drop it.

> > +static LIST_HEAD(btrfs_mount_head);
> 
> This list is not needed. Put void *private link from mount-info
> to brtfs_root struct and (later, much later) release them altogether.
> 
> > +struct btrfs_subvol_node {
> > +	struct rb_node			rb_node;
> > +	struct list_head		sibling;
> 
> List is used purely for releasing. Can be either removed or
> reworked using rbtree.

If we're not cleaning up it in general -- i'll simply drop this.

> > +
> > +	struct rb_node			rb_dev;
> > +	dev_t				st_dev;
> > +	ino_t				st_ino;
> 
> st_ino it completely unused.

It should be used in debug output, but I forgot to put it there.
Will drop, thanks.

> > +
> > +	char				*name;
> > +	char				*path;
> > +	char				*full;
> > +	int				deleted;
> 
> Plz, comment the fields.

OK.

> > +
> > +	fd = open(m->mountpoint, O_RDONLY);
> > +	if (fd < 0) {
> > +		pr_perror("Can't open mountpoint %s", m->mountpoint);
> > +		return -1;
> > +	}
> > +
> 
> Mountpoint can be unreachable. The open_mountpoint() handles this.

Hmm, open_mountpoint can be used iif the complete mount tree is created,
so I presume you prefer to defer btrfs parsing and do it as a final stage
of mount points tree creation, right?