Re: MADT enumeration for x86_64

[John Marino <dragonflybsd@xxxxxxxxx>]

>Am 24.09.2010 01:10, schrieb Matthew Dillon:
>>     Heh.  My test box froze solid with the MADT stuff.  I had to put
>>     hw.madt_probe_test=1 in my /boot/loader.conf to get the system
>>     to boot.  (Phenom x 6 Gigabyte 880GMA-UD2H MB).
>>
>>     At the moment I'm not sure where the problem is.
>>   
>
>Same on my Phenom x 4 ASUS board. With hw.madt_probe_test=1 it started
>to boot, but the boot was very slow and sluggish. After it completed to
>boot, and I ran a few programs, it somehow deadlocked without giving any
>message on the console.
>
>There are some more patches by Sephe which I'll apply to x86_64.
>
>Regards,
>
>  Michael


This recent patch may explain why the latest x86_64 developmental 
snapshot crashes in a 32-bit version of VirtualBox 3.2.8.  This was 
covered in IRC, but the Sept 27th snapshot crashes very soon after the 
kernel loads in the 32-bit version of VirtualBox. I confirmed this on 
both OpenSolaris and Windows XP. Apparently it does not crash in a 
64-bit version of Vista according to swildner.

It was suggested that I escape to the loader and set 
hw.madt_probe_test=1 before booting the kernel.  I did this, and I was 
able to boot the snapshot. This is definitely a regression because I've 
been running 64-bit DFly 2.6 release on a 32-bit VBox for months without 
issue.

I imagine anyone with access to a 32-bit version of Vbox could reproduce 
what I'm seeing easily.

Regards,
John