[Call For Testing] VirtualBox for FreeBSD! take2

[Juergen Lock]

In article <20090518155613.GT86061 at carrick.bishnet.net> you write:
>On Mon, May 18, 2009 at 08:38:18AM -0500, Doug Poland wrote:
>> On Sun, May 17, 2009 19:25, Greg Byshenk wrote:
>> > On Sun, May 17, 2009 at 08:09:20PM +0200, Martin Wilke wrote:
>> >>
>> >> We rolled a new tarball with the patch from Juergen Lock [1]
>> >> with a posible fix for AMD64 users, tested on 3 machines
>> >> which now works without problems. Many Thanks to him for
>> >> his nice work!
>> >>
>> >> http://people.freebsd.org/~miwi/vbox/virtualbox_2.tgz
>> 
>> Working for me with a fresh cvsup of 7.2-STABLE (17 May 09). 
>> Installing a 7.2-RELEASE amd64 guest as I write this.
>
>Works for me, but I've not been able to get a 64 bit guest to work. I
>get the following error when booting the installer:
>
>	CPU doesn't support long mode
>
>I think I've got all the right options set. Maybe my CPU doesn't support
>the necessary VM stuff to let a 64 bit guest work?
>
Yes, afaik vbox can only run amd64 guests if the host cpu has svm (amd)
or vmx (intel) and you tell vbox to use that in the guest settings.
(`Enable VT-X/AMD-V'.)  And btw, npt/ept (`Enable Nested Paging') is also
worth using if your host cpu has it, it should speed up the vm a little.

 And if you have a guest (especially amd64) that then still hangs/crashes
at boot or otherwise doesnt come up, check if the io-apic is disabled in
the guest settings and enable it...  (I think there are also older 32 bit
guests that need it disabled, otherwise the option wouldn't be there :)
- but usually its the other way around.)

>I'm also still seeing the following on loading vboxdrv:
>
>kldload: unexpected relocation type 10
>kldload: unexpected relocation type 10
>kldload: unexpected relocation type 10
>kldload: unexpected relocation type 10
>kldload: unexpected relocation type 10
>kldload: unexpected relocation type 10
>kldload: unexpected relocation type 10
>kldload: unexpected relocation type 10
>kldload: unexpected relocation type 10
>kldload: unexpected relocation type 10
>kldload: unexpected relocation type 10
>kldload: unexpected relocation type 10
>vboxdrv: fAsync=0 offMin=0x41f offMax=0x53b
>
>Are others still seeing that on amd64?

 These are caused by the in-kernel linker not properly skipping debug
symbols in klds (they are harmless, altho may waste some kernel memory),
a workaround has been committed to vbox svn in the meantime (uncommenting
an objcopy --strip-debug in Config.kmk that was already in there...)
This is btw also the reason why we have seperate foo.symbols files in
our kernel dirs which kgdb will then pick up - tho unfortunately it
doesn't seem to do that for klds in /boot/modules, at least on 7-stable.

 HTH,
	Juergen

PS: I trimmed the Cc'd lists to only -emulation...