pxegrub not loading on broadcom undi pxe-2.1 v9.4.5

pxegrub not loading on broadcom undi pxe-2.1 v9.4.5

[Joe McDonagh]

Hi all, I am having a rough time getting pxegrub to boot on a silicon 
mechanics whitebox. I compiled the source from grub-1.96, and tried the 
svn version, both with the same results. I haven't yet tested any other 
systems, however this is what I see from tcpdump on the server (there is 
only one client on this specific network):

[/srv/tftp] > sudo tcpdump -vvvi eth0
tcpdump: listening on eth0, link-type EN10MB (Ethernet), capture size 96 
bytes
19:02:59.971032 arp who-has 192.168.4.52 tell 192.168.4.1
19:03:43.177172 IP (tos 0x0, ttl 20, id 0, offset 0, flags [none], proto 
UDP (17), length 576) 0.0.0.0.bootpc > 255.255.255.255.bootps: 
BOOTP/DHCP, Request from 00:30:48:5e:55:2c (oui Unknown), length 548, 
xid 0x495e552c, secs 4, Flags [Broadcast] (0x8000)
      Client-Ethernet-Address 00:30:48:5e:55:2c (oui Unknown) [|bootp]
19:03:43.177675 IP (tos 0x10, ttl 64, id 0, offset 0, flags [none], 
proto UDP (17), length 328) 192.168.4.1.bootps > 255.255.255.255.bootpc: 
BOOTP/DHCP, Reply, length 300, xid 0x495e552c, secs 4, Flags [Broadcast] 
(0x8000)
      Your-IP 192.168.4.52
      Server-IP 192.168.4.1
      Client-Ethernet-Address 00:30:48:5e:55:2c (oui Unknown) [|bootp]
19:03:47.144381 IP (tos 0x0, ttl 20, id 1, offset 0, flags [none], proto 
UDP (17), length 576) 0.0.0.0.bootpc > 255.255.255.255.bootps: 
BOOTP/DHCP, Request from 00:30:48:5e:55:2c (oui Unknown), length 548, 
xid 0x495e552c, secs 4, Flags [Broadcast] (0x8000)
      Client-Ethernet-Address 00:30:48:5e:55:2c (oui Unknown) [|bootp]
19:03:47.152693 IP (tos 0x10, ttl 64, id 0, offset 0, flags [none], 
proto UDP (17), length 328) 192.168.4.1.bootps > 255.255.255.255.bootpc: 
BOOTP/DHCP, Reply, length 300, xid 0x495e552c, secs 4, Flags [Broadcast] 
(0x8000)
      Your-IP 192.168.4.52
      Server-IP 192.168.4.1
      Client-Ethernet-Address 00:30:48:5e:55:2c (oui Unknown) [|bootp]
19:03:47.153468 arp who-has 192.168.4.1 tell 192.168.4.52
19:03:47.153481 arp reply 192.168.4.1 is-at 00:0c:29:99:f1:93 (oui Unknown)
19:03:47.153623 IP (tos 0x0, ttl 20, id 2, offset 0, flags [none], proto 
UDP (17), length 52) 192.168.4.52.2070 > 192.168.4.1.tftp: [udp sum ok]  
24 RRQ "pxegrub" octet tsize 0
19:03:47.155292 IP (tos 0x0, ttl 64, id 27095, offset 0, flags [DF], 
proto UDP (17), length 41) 192.168.4.1.42402 > 192.168.4.52.2070: [bad 
udp cksum 8098!] UDP, length 13
19:03:47.155437 IP (tos 0x0, ttl 20, id 3, offset 0, flags [none], proto 
UDP (17), length 45) 192.168.4.52.2070 > 192.168.4.1.42402: [udp sum ok] 
UDP, length 17
19:03:47.155639 IP (tos 0x0, ttl 20, id 4, offset 0, flags [none], proto 
UDP (17), length 57) 192.168.4.52.2071 > 192.168.4.1.tftp: [udp sum ok]  
29 RRQ "pxegrub" octet blksize 1456
19:03:47.158053 IP (tos 0x0, ttl 64, id 27095, offset 0, flags [DF], 
proto UDP (17), length 43) 192.168.4.1.47107 > 192.168.4.52.2071: [bad 
udp cksum a629!] UDP, length 15
19:03:47.158284 IP (tos 0x0, ttl 20, id 5, offset 0, flags [none], proto 
UDP (17), length 32) 192.168.4.52.2071 > 192.168.4.1.47107: [udp sum ok] 
UDP, length 4
19:03:47.158511 IP (tos 0x0, ttl 64, id 27096, offset 0, flags [DF], 
proto UDP (17), length 1056) 192.168.4.1.47107 > 192.168.4.52.2071: UDP, 
length 1028
19:03:47.158764 IP (tos 0x0, ttl 20, id 6, offset 0, flags [none], proto 
UDP (17), length 32) 192.168.4.52.2071 > 192.168.4.1.47107: [udp sum ok] 
UDP, length 4
19:03:52.151890 arp who-has 192.168.4.52 tell 192.168.4.1
19:03:53.150933 arp who-has 192.168.4.52 tell 192.168.4.1
19:03:54.151439 arp who-has 192.168.4.52 tell 192.168.4.1


The PXE client at this point is just sitting there doing nothing. Does 
this look familiar to anyone? Not sure if the bad checksums have 
anything to do with it. It might be worth nothing that the PXE server is 
a virtual machine bridged to a NIC on the host.

Re: pxegrub not loading on broadcom undi pxe-2.1 v9.4.5

[Joey Korkames]

Joe McDonagh writes:

> Hi all, I am having a rough time getting pxegrub to boot on a silicon 
> mechanics whitebox. I compiled the source from grub-1.96, and tried the 
> svn version, both with the same results. I haven't yet tested any other 
> systems, however this is what I see from tcpdump on the server (there is 
> only one client on this specific network):
> 
8% snip %8
> 
> 
> The PXE client at this point is just sitting there doing nothing. Does 
> this look familiar to anyone? Not sure if the bad checksums have 
> anything to do with it. It might be worth nothing that the PXE server is 
> a virtual machine bridged to a NIC on the host.

Broadcom & NVIDIA PXE Roms are the worst in my experience - slow and buggy.
I've had to upgrade ROMs on motherboards with these before.

I'd start by making sure you're using the latest tftpd-hpa server. Then test booting 
the latest pxelinux.0 from syslinux.zytor.com.
If that works, then your grub2 build process might need a look-see.

MODULES="pxe pxecmd help reboot serial multiboot pc configfile
normal linux bsd chain echo ls sleep elf aout terminfo handler boot efiemu
efiemu_pnvram acpi mmap xnu vbe vbetest vbeinfo font gfxterm biosdisk ufs"

grub-mkimage --output=core.img --prefix="(pxe)/defaulttftprootdir/grub2" $MODULES
cat /usr/lib/grub/i386-pc/pxeboot.img core.img > pxegrub2