after ip=autoconf WAS: Kernel (2.4.X series) not doing the initial boot.

Stephane Enten tuf at
Thu Oct 11 17:00:41 CEST 2001

On Wed, Oct 10, 2001 at 03:32:03PM +0200, Geert Stappers wrote:
 > At 23:19 +0200 10/9/01, Stephane Enten wrote:
 > snip
 > >
 > >Regarding the tests-of-the-day, I found something I'd call a bug:
 > >in my kernels (2.4.10 and 2.4.9), if you just put the nfsroot= kernel
 > >parameter in your "append" LILO configuration, the kernel is *not*
 > >doing the autoconfiguration part.
 > >(While the doc states that omitting the ip= parameter is OK.)
 > >
 > >You have to specify 'ip=autoconf' and then the kernel do his DHCP
 > >or BOOTP or RARP request.
 > >
 > >No I have to find why the hell he isn't happy with the DHCPOFFER
 > >returned by the DHCP server, but that's a huge step passed :)
 > >
 > Once I had an encounter with a 3C509 with IRQ on line 9.
 > The module was inserted without any options, so it assumed IRQ10.
 > According tcpdump it did transmit packets, but it didn't react
 > on the answers-packets due the wrong interrupt line.
 > I suggest you append an other parameter.
 > Such as "ether=9,0,eth0"  which means eth0 at IRQ 9 and autoprobe I/O-base.
 > ( )

I added that, I planed to add it anyway since our machines has 2 NICs and
since their IRQs are reversed they are reversed (ie. port 0 = eth1, port 1 =
eth0 ... annoying :) ).

However, the DHCP code still did not accept the DHCPOFFER.
I explored the net/ipv4/ipconfig.c again and found that there is some 
code to check if ic_myaddr equals INADDR_NONE.

I found that my ic_myaddr was set at at the DHCP request time,
don't know why.

So I added "ip=" in my kernel config, then the
ic_myaddr was at INADDR_NONE ( and the kernel was happy,
did the NFS mount and now I can work on tweaking the installation part :)


More information about the linux-fai mailing list