NFSv4.1 status

Rick Macklem rmacklem at uoguelph.ca
Tue Dec 13 02:11:20 UTC 2011


Julian Elischer wrote:
> On 12/11/11 8:37 AM, Rick Macklem wrote:
> > Hi,
> >
> > I thought I'd post to let anyone who is interested know what the
> > status of NFSv4.1 is. First off, the slides here give you an
> > overview of where the vendors are:
> >     http://people.redhat.com/steved/Lisa-11/LISA-11-pNFS-BoF-final.pdf
> >
> > You'll notice that FreeBSD isn't mentioned, which is fair, since no
> > NFSv4.1 support is in FreeBSD9.0. However, I now have the basic
> > NFSv4.1
> > client support written and lightly tested
> but Panassus and Netapp were.. have you talked to them about
> interactions
> between your code and theirs when the time comes to merge?

Well, I was able to do some testing at the NFSv4.1 Bakeathon last
June. These Bakeathons + Connectathon are where the interoperability
testing goes on. Unfortunately, I can't normally attend, since going
to these costs $$$ and, as I suspect you already know, I'm not funded
to do this work. I was able to make it to the one last June, because
it was in Ann Arbor MI and I could drive my little Toyota Echo there
(doesn't burn much gas) and stay in the el-cheapo Motel 6. (Also,
the Bakeathons don't have any registration fee, unlike Connectathon.)

Now, you normally sign an NDA which covers what you learn at these for 1 year.
For some reason, that didn't happen last June, but I am still hesitate
to discuss what I learned, except to say that the code was mostly
written during that week and was working well against the servers that
I could test against there by the end of the week.

I hope to be able to repeat the above, since they are planning another
Ann Arbor Bakeathon in the June 2012 timeframe.

Also, someone at Netapp was attempting to make their server simulator
(runs just like a real Netapp Filer, except very slowly as a Linux task)
available to me, but I haven't heard from him lately, so I suspect that it may never happen.
He's a good guy, but can't do anything the company won't allow.
(He also mentioned that I would probably have to sign an NDA and until
 I see and read the document, I cannot say if I will be willing to do so.)
This person is also trying to convince the testing team at Netapp to do
FreeBSD client to Netapp filer testing, but I have no idea if he`s made
progress on that either.

> > For releng9:
> >    http://people.freebsd.org/~rmacklem/nfsv4.1-client/nfsv4.1-releng9.patch
> >
> > For head/-current, there is an up-to-date source tree at:
> >    http://svnweb.freebsd.org/base/projects/nfsv4.1-client/sys/
> >
> > It does not include any pNFS support at this point, but I'm am just
> > starting
> > to work on that and hopefully will have something for the file
> > layout soon
> > and hope to get to test it against various servers at the NFSv4
> > Bakeathon in June.
> >
> > rick
> > ps: There has been no work done yet on a FreeBSD NFSv4.1 server as
> > far as I know.
> Pannassus (how DO you spell that?) and netapp must have that part
> worked out.

If you are referring to a server, then yes, the slides would indicate
that they are ready to ship soon. I suspect both companies have invested
many man years of engineering (ie. many $$$$) to this effort. They would
definitely consider the results proprietary. (Although Netapp uses some
FreeBSD code, it is my understanding that their NFS code is a separate
world from that of FreeBSD. As noted above, even getting to test against
a pre-release binary is considered proprietary and they haven't, as yet,
agreed to let me do that even if I sign some NDA.)

Btw, both companies have sponsored work related to the Linux client.
For example, Trond Myklebust, who is a/the major Linux NFS client developer,
is a Netapp employee and Panasus has contributed Linux client work that
they have done "in house" for their object layout. (Someone from Panasus
has indicated that I should try to adhere to the Linux client API that
sits between the generic layout handler and the layout specific driver.
I am hoping that implies that I may have access to their layout driver under
a non-GPL`d open source license at some point, but this is just dreaming
on my part at this point.:-)

I have recently received some email related to how to patch a Linux
kernel to get a basic pNFS server going to testing. Hopefully I can
get this going. (I currently use a Fedora15 Linux NFSv4.1 server for
testing, but it doesn`t have pNFS support in it.)

Did this long winded writeup cover what you were asking...rick

> > _______________________________________________
> > freebsd-fs at freebsd.org mailing list
> > http://lists.freebsd.org/mailman/listinfo/freebsd-fs
> > To unsubscribe, send any mail to
> > "freebsd-fs-unsubscribe at freebsd.org"
> >


More information about the freebsd-fs mailing list