[grub #24] Grub stop can not load the kernel image via second network card

[hokoat]

Reporter: address@hidden
Summary: Grub stop can not load the kernel image via second network card
Version: 0.92
Type: software bug

Message:
I have a CPCI system with two server blades.
First server blade is a Sparc based server blade with 3 network interface cards 
(2 Intel and 1 Sun Happy Meal) with Suse 7.3 running on it. This is the boot 
server. 

Second server blade is a Intel P3 based Server blade with two network interface 
cards (Intel EtherExpressPro100 82559ER).

The boot architecture should work as a redundant system for high availability 
reasons. Therefore two Hubs are interconnect the boards and building two 
independent subnets (one at the first NIC(ETH1) and the other one at the second 
NIC(ETH2)).

When the remote boot server blade is booting via the first interface (default) 
all works fine. When I disable the interface on the boot server which is 
connected to first interface(ETH1) of remote boot server blade the boot process 
starts via the second interface(ETH2). The pxegrub is loaded and executed by 
the server blade. Now pxegrub should oad the grubconfig file, kernel and the 
ramdisk. But this does not happen because grub is trying to propagate now the 
ETH1 interface and not ETH2 as before. Following output is written to console:


Intel UNDI, PXE-2.0 (build 071)
Copyright (C) 1997,1998,1999  Intel Corporation  

DHCP MAC ADDR: 00 80 42 11 E7 DC
PXE-E51: No DHCP or BOOTP offers received


Intel UNDI, PXE-2.0 (build 071)
Copyright (C) 1997,1998,1999  Intel Corporation 

DHCP MAC ADDR: 00 80 42 11 E7 DD
CLIENT IP: 192.168.0.4  MASK: 255.255.255.224  DHCP IP: 192.168.0.2             
Found Intel EtherExpressPro100 82559ER at 0x1080, ROM address 0x0000            
Probing...[Intel EtherExpressPro100 82559ER]Ethernet addr: 00:80:42:11:E7:DC    

My Question are:
Is there a way to configure grub that grub is loading the kernel via the 
interface which was used to load grub or is this a SW bug?

Exists a patch for that if this is a SW bug.


Thanks in advance

Holger Koch


----
Please send followups to <address@hidden>.