Re: KVM internal error. Suberror: 1, emulation failure

[Gleb Natapov <gleb@redhat.com>]

On Thu, Jul 18, 2013 at 07:58:31AM +0200, Paolo Bonzini wrote:
> Il 17/07/2013 18:16, Dave Hansen ha scritto:
> > I'm causing qemu to spew these emulation failure messages until I kill
> > it.  The guest kernel being run has been hacked up pretty heavily and is
> > probably either accessing bad physical addresses (above the address
> > ranges in the e820 table) or trying to DMA to bad addresses.
> > 
> > What I'd really like qemu to be doing is trapping back in to the guest
> > kernel to have it handle this issue.  Then I'd have a better chance of
> > dumping out some debugging information to see where I went wrong.
> 
> This is happening because the kernel is executing a PCMPEQB instruction
> on an invalid memory address.  This instruction is not yet emulated by
> KVM.  If you want QEMU to trap back to the guest kernel, you can add
> emulation of the instruction to arch/x86/kvm/emulate.c.
> 
> If you do not really care about the guest doing something sane, you can
> use a stub emulation function that is just "return emulate_ud(ctxt)".
> That alone could be a good starting point to attach a kernel debugger to
> the guest.
> 
This is the behaviour that he currently gets (assuming there is no bug
somewhere, run ftrace to check), see my other reply. Not sure what he does
in his #UD handler that emulation error reappear. Restart offending process?

--
			Gleb.