Re: x86: Question regarding the reset value of LINT0

[Paolo Bonzini <pbonzini@redhat.com>]

On 09/04/2015 21:17, Bandan Das wrote:
> > Excluding (1) all of the other issues are related to the VM BIOS. Perhaps
> > KVM should somehow realize which VM BIOS runs? (yes, it sounds just as bad.)
> 
> How about renaming the toggle Avi mentioned above to something more generic
> (KVM_DISABLE_LEGACY_QUIRKS ?) and grouping all the issues together ? Modern userspace
> will always enable it and get the new correct behavior. When more cases are discovered,
> KVM can just add them to the list.

It can be a VM capability (KVM_FIRMWARE_QUIRKS?) that is enabled via
KVM_ENABLE_CAP.

The first argument in struct kvm_enable_cap can be used to add more
quirks in the future.  For now, an argument of zero could be used to:

1) set up LINT0 correctly

2) set up CD and NW correctly in svm_set_cr0

AFAIK the MTRR issue in SeaBIOS was not fixed.  For that, QEMU could
write to MSR_MTRRcap with KVM_SET_MSR.

Setting the "accessed" bit in vmx_set_segment is IMHO harmless, and we
might as well do it even if !enable_unrestricted_guest.

Paolo