[Prev][Next][Index][Thread]

BUG (?) in oskit lance driver



Mostly for kicks, I was trying to make netboot work under VMWare. I have tried two netboot kernels -- one with just the lance driver from the autobuilder web page, the other built (by me) under RH7.0 using kgcc (per previous mail). I see the same results from both kernels (which I suppose should be encouraging, since my own build isn't crashing altogether :-)
 
The netboot kernel comes up, prompts, and I tell it to proceed. It then probes some very ridculous number of ethernet interfaces (it gets as high as "eth357"), announcing with each one that it has found the AMD PCI Net card. Mind you, there is only one such card in the VMWare emulator!
 
I'm guessing that what is going on here is that the Lance driver is configuring the card -- it's definitely getting all the way to the end of the probe1() rountine -- but is then failing to recognize that the card has already been configured. Does this seem possible?
 
Any clues greatfully accepted. Before I try to debug yet another kernel I thought I'ld ask if anyone has seen this before.
 
Oh. It's VMWare for Linux 2.02.
 
I'll just boot kernels off a hacked boot floppy for now, but I suspect that VMWare is one configuration you guys will want to have work.. ;-)
 
Oh. The kernel sometimes (not always) gets as far as issuing bootp requests. VMware's dhcpd definitely sees them and issues BOOTREPLY packets if one hacks their dhcpd.conf file appropriately. It does not appear that the netboot kernel ever sees the replies, however, as it proceeds to run the bootp() protocol on the next card (which is, of course, the same card as before...).
 
 
Jonathan

Follow-Ups: