Re: KVM on Via Nano (Isaiah) CPUs? <Virus checked>

[Andreas Tanz <andreas.tanz@kvt.de>]

Am 19.03.2009 schrieb Avi Kivity:
> > [101324.097968] vmx->handle_exception 00 : giving some infos
> > [101324.097973] vmx->handle_exception 01 : vect_info: 0x0
> > [101324.097978] vmx->handle_exception 02 : intr_info: 0x80000b0d, is_page_fault()==0
> > [101324.097983] vmx->handle_exception 03 : irq_chip_in_kernel()==1
> > [101324.097988] vmx->handle_exception 04 : is_external_interrupt()==0
> > [101324.097993] vmx->handle_exception 0a : kvm_rip_read(vcpu) returned 0x3154
> > [101324.097998] vmx->handle_exception 0f : vcpu->arch.rmode.active: 0x1
> > [101324.098039] vmx->handle_exception 10 : handle_rmode_exception(vcpu, intr_info & INTR_INFO_VECTOR_MASK, error_code) returned 0x1
> > [101324.098050] returning from kvm_handle_exit, cause 3, retval = 1, exit_reason = 0
> > .... until kvm get's killed
> >   
> 
> The code in question is:
> 
>     3152:       fb                      sti   
>     3153:       55                      push   %bp
> 
> *** fault here ***
> 
>     3154:       89 e5                   mov    %sp,%bp
>     3156:       8b 7e 2a                mov    0x2a(%bp),%di
>     3159:       8b 46 28                mov    0x28(%bp),%ax
>     315c:       8b 4e 08                mov    0x8(%bp),%cx
>     315f:       81 ff 00 f8             cmp    $0xf800,%di
> 
> So it looks like we tried to inject the IDE interrupt (that's the first 
> point where it is possible to inject interrupts), but via fails to 
> inject the interrupt for some reason.
> 
> I'll need to think how to debug this further.
> 

Maybe the reason is elsewhere....
I'm unable to build any GCC 4.* (build crashes in stage 2 with a segfault when creating build/genoutput).
I think there are some Nano specific x86-64-quirks gcc doesn't know about...
Further it's impossible to build glibc in multilib mode (x86-64 together with x68-32).
I'll post this in gentoo's bugzilla....
Everything else compiles fine (with precompiled gcc and glibc)