[plug] Re: [NFS] Linux NFS client to Solaris HSM NFS server

[Ian Kent]

One of my colleagues told me that this behavior has appeared in the 2.4
series kernel. He said that it worked as expected in the 2.2 series.
Subsequently I checked 2.2.19 on my Sparc RH 6.2 box and it waited until
the file was available and then continued. There was no message but then
that is not important to us of course, just the behavior.

So has the 2.4 client NFS been broken of is it now fixed?

It would be disappointing to have to regress to a 2.2 kernel and miss out
on the new features and improved performance of 2.4.

Trond Myklebust wrote:

> >>>>> " " == Spencer Shepler <shepler at eng.sun.com> writes:
>
>      > Trond,
>      >   Are you saying that it is difficult to retransmit the request
>      >   when
>      > JUKEBOX is received or user level interaction is difficult?
>
> Neither. I'm just saying that AFAICS you're likely to want to
> implement this at the VFS level above the semaphores etc. that we
> apply in order to protect against local VFS races. If not, you'd be
> likely to cause other processes to hang while your request sleeps.
>
> Given the above premise I can't really see any good reason to do a
> kernel space implementation. You can't protect against races, and
> since the whole idea is to sleep for a fixed period of time, it's
> going to be hard to justify all the effort by quoting significant
> gains in efficiency ;-)...
>
> Cheers,
>   Trond

--
   ,-._|\    Ian Kent
  /      \   Perth, Western Australia
  *_.--._/   E-mail: ian.kent at pobox.com, raven at plug.linux.org.au
        v    Web: http://pobox.com/~ian.kent