Diskless boot problem

[Sławek Żak]

Hi,

I have a problem with booting Dell 2850 over network. The machine
reads kernel over net, boots upto mounting / from NFS and then
crashes.

Tcpdump output:

12:15:58.919683 arp who-has 10.158.190.73 tell 10.158.190.74
12:15:58.919702 arp reply 10.158.190.73 is-at 00:11:43:d3:6e:e1
12:15:58.920058 IP 10.158.190.74.475209176 > 10.158.190.73.2049: 92
getattr [|nfs]
12:15:58.920134 IP 10.158.190.73.2049 > 10.158.190.74.475209176: reply
ok 28 getattr ERROR: Stale NFS file handle
12:15:58.920432 arp who-has 10.158.190.73 tell 10.158.190.74
12:15:58.920442 arp reply 10.158.190.73 is-at 00:11:43:d3:6e:e1
12:15:58.920681 IP 10.158.190.74.475209177 > 10.158.190.73.2049: 100
lookup [|nfs]
12:15:58.920707 IP 10.158.190.73.2049 > 10.158.190.74.475209177: reply
ok 28 lookup ERROR: Stale NFS file handle
12:15:58.920932 IP 10.158.190.74.475209178 > 10.158.190.73.2049: 100
lookup [|nfs]
12:15:58.920963 IP 10.158.190.73.2049 > 10.158.190.74.475209178: reply
ok 28 lookup ERROR: Stale NFS file handle
12:15:58.952180 IP 10.158.190.74.475209179 > 10.158.190.73.2049: 100
lookup [|nfs]
12:15:58.952277 IP 10.158.190.73.2049 > 10.158.190.74.475209179: reply
ok 28 lookup ERROR: Stale NFS file handle
12:15:58.984785 IP 10.158.190.74.475209180 > 10.158.190.73.2049: 100
lookup [|nfs]
12:15:58.984866 IP 10.158.190.73.2049 > 10.158.190.74.475209180: reply
ok 28 lookup ERROR: Stale NFS file handle
12:15:59.020500 IP 10.158.190.74.475209181 > 10.158.190.73.2049: 104
lookup [|nfs]
12:15:59.020573 IP 10.158.190.73.2049 > 10.158.190.74.475209181: reply
ok 28 lookup ERROR: Stale NFS file handle
12:15:59.054130 IP 10.158.190.74.475209182 > 10.158.190.73.2049: 104
lookup [|nfs]
12:15:59.054224 IP 10.158.190.73.2049 > 10.158.190.74.475209182: reply
ok 28 lookup ERROR: Stale NFS file handle

I wonder where the `Stale NFS handle'  error comes from, as the client
doesn't seem to have mounted the filesystem over NFS from what I can
see. On the console of the diskless I have this:

NFS ROOT: 10.158.190.73:/var/www/FreeBSD-5.4-x86-PXE
em0: Link is up 100 Mbps Half Duplex
exec /sbin/init: error 70
exec /sbin/oinit: error 70
exec /sbin/init.bak: error 70
exec /rescue/init: error 70
exec /stand/sysinstall: error 70
init: not found in path
/sbin/init:/sbin/oinit:/sbin/init.bak:/rescue/init:/stand/sysinstall
panic: no init
Uptime: 55s
Cannot dump. No dump device defined.
Automatic reboot in 15 seconds - press a key on the console to abort

The speed for em0 is obviously wrong. Setting on the switch is 100
full-duplex. Our network wizards can f***kup autonegotiation on Cisco
Catalyst, so it must stay that way. Intel em-s tend to hang for a
couple of seconds before getting on the net so it might be the
problem. On the other hand kernel loads just fine over TFTP.

Any thoughts?

Thanks, /S
-- 
Sławek Żak / UNIX Systems Administrator