[PATCH 12/14] repair: switch block usage bitmap to a btree

[Christoph Hellwig]

On Thu, Oct 22, 2009 at 11:22:42AM -0500, Alex Elder wrote:
> I have a few minor things I would like to see changed at some
> point, but nothing looks obviously incorrect so I'll wait and
> post a patch against this once it's committed.
> 
> Here are a couple other general thoughts:
> - One minor concern is that the btree code has cases in which
>   the peek routines don't work (when the keys_valid flag in the
>   btree root is zero) and this code doesn't check for that.
>   I'll just assume for now that never happens here.

It should not happen, but yes - making this explicit would be good.

> - The bitfield code used for the real-time volume map is
>   generally useful and could be separated into its own module.

Not sure.  It's really a crufty leftover.  It should probably also use
a btree instead, but I fear any effort spent on the RT subvolume is
pretty much wasted anyway.