Re: [PATCH 1/3] btrfs: extended inode refs

[Mark Fasheh <mfasheh@suse.de>]

On Fri, Jul 06, 2012 at 04:56:40PM +0200, Jan Schmidt wrote:
> Hi Mark,
> 
> Sorry, I'm a bit late with my feedback.

No problem, thank you for taking the time to read these patches! Reply to
your most substantial comment is below. Everything else you pointed out I'll
have fixed up for an updated patch (don't think I need to reply individually
to those).


> On Mon, May 21, 2012 at 23:46 (+0200), Mark Fasheh wrote:

> > @@ -189,6 +435,19 @@ int btrfs_insert_inode_ref(struct btrfs_trans_handle *trans,
> >  
> >  out:
> >  	btrfs_free_path(path);
> > +
> > +	if (ret == -EMLINK) {
> > +		struct btrfs_super_block *disk_super = root->fs_info->super_copy;
> > +		/* We ran out of space in the ref array. Need to
> > +		 * add an extended ref. */
> > +		if (btrfs_super_incompat_flags(disk_super)
> > +		    & BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF)
> 
> Uhm. Good place to check here. But please help me to find the place where this
> is added to the super block's flags in the first place.

mkfs.btrfs  :)

The kernel should never be adding the flag as things are currently designed.

Thanks again!
	--Mark

--
Mark Fasheh