Re: radix tree implementation for quota ?

[David Holland <dholland-tech%netbsd.org@localhost>]

On Sun, Nov 28, 2010 at 05:53:59PM +0100, Manuel Bouyer wrote:
 > our FFS quota system is pretty outdated and is less and less
 > suitable for modern storage (quotacheck is long and can't be
 > avoided). I'm thinking about coding a new quota system [...]

Unfortunately, before you can change the FFS quota system you have to
change the kernel interface to it to be fs-independent. Currently it
exposes the structure and semantics of the traditional FFS quota
structures, not only to callers above the vfs layer but also to
userland. This means that currently only ufs-based filesystems can
have quota support and they must all do it using the FFS format and
structures.

Because I wanted quotas on LFS to keep working after unplugging LFS
from UFS, my LFS tree contains a start on this problem, but
unfortunately it's not yet good enough to commit. Plus it's entangled
with LFS stuff.

I remember I floated some proposals about fixing the quota API a while
back, but I don't remember what if anything came of the discussion; I
think not much.

(Also, why a radix tree? Radix trees are generally not very efficient.
If you're going to, though, you might want to reuse the direct,
indirect, double indirect, etc. method FFS uses for block mapping.)

-- 
David A. Holland
dholland%netbsd.org@localhost