Comment 15 for bug 212485

I'm doing a git (as we speak) between 2.6.24-12.18 and 2.6.24-14-25 on my Ubuntu Hardy system. I'm assuming for the moment that 2.6.24-12.18 is a known "good" version of rpcb_clnt.c. In addition, I'll be installing a 2.6.26 kernel + patches from source on my system and retesting.

In the meantime are we certain that there is not another issue in play? For instance, in my case, I'm running on a Sun x4100 so this means I have a nForce NIC and use the forcedeth driver. Periodically my NIC will go vegetative and instance is denoted with messages like:

  Jul 31 12:53:16 prod-unix-shell04 kernel: [ 596.950188] nv_stop_tx: TransmitterStatus remained busy<6>eth0: link up.

When this happens my system will fall of the network until it is power cycled -- as using ifup, ifconfig, or even a reboot doesn't set the NIC straight. Sometimes this happens when I'm using NFSv4 and sometimes this happens when I'm not. It always happens when I'm doing lots of I/O (apt, bonnie++, SQL queries, dd, etc.).

Given what's been described here earlier (RE: pulling network cable and watching nfsv4 go haywire) I'm curious if the problem is not limited to changes to rpcb_clnt.c.

Does anyone have thoughts on this?