At 12:46 +0200 on 26 Jul (1311684389), Jeroen Groenewegen van der Weyden wrote:
Here my input
Thanks. Looks very similar to the bug I thought I fixed with my patch.
VCPUs 1 and 3 are always at 0xa01a1c9d: <0f> 01 c2 eb 03 0f 01 c3
which is probably this fragment of code from KVM's vmx_vcpu_run():
/* Enter guest mode */
"jne .Llaunched \n\t"
__ex(ASM_VMX_VMLAUNCH) "\n\t"
"jmp .Lkvm_vmx_return \n\t"
".Llaunched: " __ex(ASM_VMX_VMRESUME) "\n\t"
".Lkvm_vmx_return: "
So, just like the case I saw, they're trying to VMLAUNCH a VMCS and
failing. That should only fail if the VMCS is already launched.
I think the reason they're _stuck_ is that error paths for VMLAUNCH and
VMRESUME emulation are wrong; I can fix them up a little but I suspect
that won't solve the problem; just change it from a hang to some other
failure mode.
Can you try the attached patch instead of the previous one?
Tim.
_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel
