On Fri, Oct 31, 2008 at 01:39:29PM -0400, Chuck Lever wrote:
> I tested a 2.6.28-rc2 kernel on the server side with and without the
> fix. I rebooted the client with "echo b > /proc/sysrq_trigger" for this
> test.
>
> If the server doesn't have the fix, I see the "never saw host x.x.x.x"
> message on the server after the client reboots. If the fix is applied, I
> see the "nlm_host_rebooted (x.x.x.x)" message on the server after the
> client reboots.
>
> So, confirmed that this is a bug in 2.6.28-rc2, and confirmed that the
> below patch addresses the problem. The NSM logic is the same on the
> client and the server, so I didn't test a server reboot.
Excellent, thanks very much!
(Though for the future as a regular regression test from a script, I
think we'd want to test for a lock after reboot with a GETLK rather than
looking for dprintk's in the logs.)
--b.