On Tue, Feb 21, 2012 at 09:15:09AM +1100, Dave Chinner wrote:
> On Sun, Feb 19, 2012 at 09:28:22PM -0500, Christoph Hellwig wrote:
> > Instead of keeping a separate per-filesystem list of dquots we can walk
> > the radix tree for the two places where we need to iterate all quota
> > structures.
> >
> > Signed-off-by: Christoph Hellwig <hch@xxxxxx>
>
> Still a bit of a lookup vs iterate hold-off remaining, but I think
> we can live with the reduction now. If it proves to be a noticable
> problem then we can look towards RCU based lookups, I think.
> Otherwise it looks OK to me.
We only walk the list during:
- mount time for quotacheck
- umount time
- full quotaoff
in the first two cases lookups can't happen at all, and in the latter
they at least shoudn't get this far. So I really don't think it's
a problem. And the full quotaoff even for quota accounting is something
I'd like to remove sooner or later, it just causes a lot of corner cases
for very little gain.
|