Subject: Re: PR 2443: NQNFS problems
To: None <peter@wonderland.org, smd@icp.net>
From: Sean Doran <smd@icp.net>
List: current-users
Date: 05/25/1996 10:51:46
| At som e quiet time I will try it between my two sparcs here. I currently
| do not use the -q flag for NFS, and that way, so far, all is fine.
Uh huh. "Well don't do that then." :-)
NQNFS is substantially faster (it even _feels_ it in some cases),
so I rather like it.
Compiling nfs/nfs_nqlease.c without optimization works just as
well as not using NQNFS in the first place with respect to the
problem with the panic somewhere in nqnfs_clientlease, and
gives the additional usefulness of supporting all the additional
NQNFS features.
I'd love to have someone else reproduce both the panic
when doing ls in a a "mount_nfs -q ..."-ed filesystem
with a standard kernel build on the SPARC, and then
the lack of a panic when doing the same thing if one
goes back after the kernel build and says:
make "CFLAGS=-g -Wall" nfs_nqlease.o
make
I would be just as interested in anybody who doesn't have
the panic in the first place (architecture independent, but
particularly on a SPARC using -current) or who still has
problems with a kernel built with the two extra steps above.
Anyone who wants the ".s" output to stare at (I don't
grok SPARC assembly very well), I'd be happy to send it along.
Sean.