On Wed, May 05, 2004 at 10:04:17AM +1000, Nathan Scott wrote:
> Hi Herbert,
> On Tue, May 04, 2004 at 12:56:30PM +0200, Herbert Poetzl wrote:
> > On Tue, May 04, 2004 at 12:32:34PM +1000, Nathan Scott wrote:
> > >
> > > Actually, it turns out that both of these are fine (thought
> > > I tested this!) because the XFS_XFLAG and XFS_DIFLAG values
> > well, never said that it wouldn't work ;)
> > PS: is there some documentation about the XFS
> > quota code, I'd like to extend the quota system
> > by allowing more than one quota hash for each
> > filesystem (vfsmount)
> Not really. The code has a fairly liberal sprinkling of comments
> though, thats probably your best bet. Can you elaborate a bit
> more on what you mean by one quota hash for each fs? (separate
> dquot hash tables per-fs?) Is that based on observed performance
> problems, or to provide some alternate functionality...?
I did a quota hash implementation for 2.4 and 2.6
to remove the strict sb <-> quota requirements
(this is for ext2/ext3) which is called "quota
hash abstractions", and this in turn is used by
the Linux-VServer per context quota patches, which
basically require to have a separate 'quota'
store (files for ext2/ext3) for each context, if
the contexts share the same filesystem ...
this works pretty well, and also allows for
vfsmount type quota or directory based quota,
but the XFS quota is just weird to me atm, so I
didn't implement it for XFS yet ...
if you need more information, just let me know ...