Re: NFS and cached atime

Trond Myklebust (trond.myklebust@fys.uio.no)
30 Apr 1999 14:37:39 +0200


Miquel van Smoorenburg <miquels@cistron.nl> writes:

> According to Trond Myklebust:
> > miquels@cistron.nl (Miquel van Smoorenburg) writes:
> > > Sounds like upon a read, the NFS client should update the atime on
> > > the server even if the file was cached.
> >
> > This is all fairly easy to implement. I'd suggest just comparing the
> > cached value of atime with the current time, and updating it in
> > nfs_file_read if the discrepancy exceeds some timeout value.
>
> I've checked BSDI 4.0 talking to a NetApp NFS server, and it behaves
> the same as Linux; atime doesn't get updated.
>
> > What are people's feelings about this?
>
> I wonder about that too - what do other systems do ?
>

DU4.0 does the same as we do (as of 2.2.7): it sends a 'getattr' that
is used for cache revalidation upon opening of the file, and that's
it...

I haven't made a full investigation (perhaps people with Solaris, or
SGI clients could check this out with a 'tcpdump') but my impression
has been that read-caching implies that atime is not updated. This is
why I've ignored earlier calls to do so.

Cheers,
Trond

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/