Here's the boot log for a normal boot with pci=nomsi.

The initrd should be setup properly. I built the kernel using the latest 
unstable and the normal linux make commands (make linux-2.6-pvops 
config|build|install), which should do the right thing.

The Xen boot was an uncoaxed boot. It usually gets up to the point of loading 
Alsa (but this message doesn't manage to get printed to the serial console, just 
to the screen attached to my test box). After that, it's a matter of waiting and 
pushing keys/buttons to try to get it to keep going. It seems to have issues 
trying to read the harddrives as well.


Patrick


Jeremy Fitzhardinge wrote:
> On 07/01/09 00:32, Patrick Colp wrote:
>> Hmmm, OK. Well, something is still going wrong. I find that adding
>> things like irqpoll and/or acpi=noirq seems to help a bit. By that I
>> mean that it gets further into the boot process before stalling. If I
>> wait a while and press random keys/buttons (like Num Lock or the power
>> button), then that seems to coax it into booting a little bit more,
>> but I've never managed to get it to completely boot. The furthest I
>> got was to where it tries to read the root filesystem.
>>
>> I could give you a dump of the boot-up messages, but aside from the
>> ioapic stuff, there's nothing out of the ordinary.
> 
> Yeah, from those logs it looks OK.  Was that a "coaxed" Xen boot?  It
> certainly looks like lots of interrupts are being received properly
> (drives detected, usb probed, etc).
> 
> Also, for a more direct comparison, what happens when you boot native
> with pci=nomsi?  I assume it will work, but it will make the logs
> directly comparable; or maybe the system just doesn't work without MSI?
> 
>> Write protecting the kernel read-only data: 7936k
>> Loading, please wait...
>> Couldnt get a file descriptor referring to the console
> Is your initrd set up right?
> 
>     J