> I'm running the kernel nfs server successfully on both alpha and i386
> (latest kernels, 2.1.90 and 2.1.88 respectively). I'll compile a 2.1
> kernel for sparc and see if I can get it going on there as well.
So that looks like either a sparc-specific problem, or a site specific one
(from what I've read of the recent sources, I have the impression that the
i386 and AXP support is sometimes slightly more advanced than the other
ports. Am I right ?)
> Meanwhile, do you get anything in the syslog when you get the failures
> (either from nfsd or mountd)? If you could send them to me and/or post
> the syslog messages perhaps I can figure out what's going on.
definitely yes -- Sorry I've forgotten to mention that.
At first, I got essentially "unauthenticated request from (%x:%d)" (with
the hex-IP address and port of the client(s)), whether I rebooted the
client or not. I added printk() statements in the
fs/nfsd/export.c/exp_export() function to each possible cause of function
exit, to find what was happening. Turns out the code block which begins
with the /* I'm just being paranoid */ comment is executed, because
inode->i_dev != nxp->ex_dev (but inode->i_ino == nxp->ex_ino).
If I understand well how things work, the nxp structure is ultimately
filled by the user-level rpc.mountd (or rpc.nfsd ?) routine, which in Olaf
Kirch's linux-nfs-0.4.22 package is in support/export/nfsctl.c/expsetup()
function. The ex_ino field is filled with the matching field from a
stat() call.
Thanks for you help !
-- Cyrille
------------------------------
Zog Zog
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu