Re: Arla on FreeBSD

Robert Watson <[email protected]> writes:
> On Thu, 15 Feb 2007, Tomas Olsson wrote:
> >> Also, you could look at file handle API, that would save you of path 
> >> lookups after the vnode is looked up first time (look around for vfs_vptofh 
> >> and vfs_fhtovp ops). This API is used by NFS server, so it shall work :).
> >>
> > That's an idea. We used that before in another context, but then we ran into 
> > trouble with too many OSes so we dropped it.  Depends on how much code one 
> > can share.
> 
> It has been ages since I looked at the xfs/nnpfs internals, but I thought it 
> already used file handles to identify cache files to the kernel?  I seem to 
> recall several of our fh*() system calls were added by Assar specifically in 
> support of Arla work.
> 
Sounds likely. I first saw it from arlad's side, so I've always thought of
it as an optimization to avoid lookups in that end without pondering it all
that much.

> Is it the case that one file visible in /afs may now be backed by pages 
> coming from different cache files?
>
Yup. There ware requests for low latency fetching of ID3 tags etc, and
working with files larger than the cache.

> How does this interact with the vnode pager?  I assume you actually end
> up with two pagers -- one for the file as visible in /afs, and then
> satisfy the pager's VOP requests by forwarding them to the right one of
> the cache vnodes?
>
Exactly. As always, just more fine grained.

> I trust you don't mmap the cache vnode, which would lead to cache
> consistency issues between any mappings of the /afs vnode vs the cache
> vnode?
>
Me too :-)

But this is still a suboptimal world where we end up with double pager
layers, and consistency can still be a problem.  In the long term it would
be nice to have arlad open it's cache nodes through nnpfs, be it small
files or just one large VM space.

> Any chance we could attract you to BSDCan to meet in person to discuss this? 
>
Chance, not probable; May is beyond my horizon. And I do expect to have the
arla working properly on fbsd well before that.

/t
_______________________________________________
[email protected] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-fs
To unsubscribe, send any mail to "[email protected]"