Routing related crash in -CURRENT, introduced between 5th May and yesterday

[Attila Nagy]

Attila Nagy wrote:
> Hello,
>
> Somewhere between 5th May and yesterday there was a (routing related?) 
> change, which causes this machine crash at boot:
> http://picasaweb.google.com/nagy.attila/20090517Fbsd8Crash#5336859077575768514 
>
> http://picasaweb.google.com/nagy.attila/20090517Fbsd8Crash#5336859069031814370 
>
>
> The machine itself is an HP DL380G4 with bge interfaces and netbooted 
> via PXE.
>
> A build, compiled on 5th May works fine, but this (compiled today, but 
> with a yesterday build this is also the same) isn't.
>
> 7-STABLE also works fine on these kind of machines and this setup.
>
> Another interesting thing is while 7-STABLE (and from 5.x to 7-STABLE 
> as of the start of May (that's the latest build we use, if there were 
> bge related changes MFC-ed since that, I don't know)) can boot on this 
> kind of machines with the default hw.bge.allow_asf=1, -CURRENT can't.
> It stops right after recognizing disk devices, even with verbose boot.
> That is the point, where DHCP (still netbooting) kicks in...
>
> I think these kind of machines are not rare (I admit that not 
> everybody uses netbooting with them, but -CURRENT freezes even when 
> installing from CD, when the installer tries to configure the 
> interfaces), so it would be good to correct (and not MFC what is on 
> HEAD until that) this regression.
>
> If I can do any debugging or give more information, please let me know!
I've found this:
http://lists.freebsd.org/pipermail/svn-src-all/2009-May/008730.html

which seems to be the place where the kernel dies according to the bt. I 
hope qingli will take care of it.

And for the bge stuff, I've just noticed that allow_asf is off in 
7-STABLE, so it's probably not a regression in code, but in behaviour. 
(which can be more easily fixed, but I don't know whether it worths to 
be on)