NFS mount RPC error when loading kernel after PXE boot(was "Re: Vbox/PXE booting works (fwd)")

[Lawrence Stewart]

Hi Danny and all,

This does indeed work as advertised, thanks for pointing me at this 
information.

The new issue I'm dealing with is that the pxe boot loader now happily 
runs, gets all the appropriate DHCP options and then stalls trying to 
mount the NFS share to pull the kernel from. I've verified that the NFS 
configuration works, both using a separate physical machine and by 
firing up the VM I'm trying to PXE boot with the fixit CD and mounting 
the share from the fixit shell i.e. the problem seems to be localised to 
the boot loader.

When pxe booting, it gets to printing the "pxe_open:" lines with correct 
server addr, path and gateway IP. It then stalls for a few mins where 
you would normally see the spinning "|" character, and then prints "NFS 
MOUNT RPC error: 60". It then appears to keep trying to load the kernel 
without success for a substantial period of time (more than 10 mins). It 
eventually gives up and dumps me at the loader prompt i.e.

can't load 'kernel'
Type '?' for a list of commands, 'help' for more detailed help.
OK

Looking at tcpdump on the NFS server, I see a number of UDP 
request/replies between port 1023 on the VM and port 111 on the server. 
After about 5 such exchanges, the Vm tries directing queries from port 
1023 to port 2304 on the server which triggers ICMP port unreachables to 
go back.

I also tested with pxeboot binaries from 7.3, 8-stable and head - all 
have the same problem.

Has anyone seen this behaviour before? What am I doing differently that 
stops me from being able to do this whilst others seem to be able to?

Cheers,
Lawrence