Re: [PATCH 3/5] Don't #VMEXIT(INTR) if we still have event_inj waiting

[Alexander Graf <agraf@suse.de>]

Am 22.09.2009 um 18:39 schrieb Joerg Roedel <joro@8bytes.org>:

> On Fri, Sep 18, 2009 at 03:00:30PM +0200, Alexander Graf wrote:
>> Real hardware would first process the event_inj field and then  
>> notify the
>> host that an interrupt is waiting.
>
> Does it really? I couldn't find this in the SVM spec.

I didn't see it in the spec either, but that's what I saw real  
hardware do which is supported by the exit_info validity check in svm.c

>
>> Let's do the same and just not EXIT_INTR if we have an event  
>> pending for the
>> L2 guest.
>
> Anyway. I think this case is handled good enough with patch 5/5. This
> patch, to be complete must also enable single-steping to exit again
> after the first instruction of the exception handler has ran to inject
> the interrupt. But that would make the whole thing rather compĺicate 
> d.

The NMI injection path has logic for that, but for now it shouldn't  
hurt - we'll get an interrupt sooner or later.

>
>> Signed-off-by: Alexander Graf <agraf@suse.de>
>> ---
>> arch/x86/kvm/svm.c |    4 ++++
>> 1 files changed, 4 insertions(+), 0 deletions(-)
>>
>> diff --git a/arch/x86/kvm/svm.c b/arch/x86/kvm/svm.c
>> index 61efd13..28fcbd0 100644
>> --- a/arch/x86/kvm/svm.c
>> +++ b/arch/x86/kvm/svm.c
>> @@ -1401,6 +1401,10 @@ static inline int nested_svm_intr(struct  
>> vcpu_svm *svm)
>>    if (!(svm->vcpu.arch.hflags & HF_HIF_MASK))
>>        return 0;
>>
>> +    /* We can't EXIT_INTR when we still have an event to inject */
>> +    if (svm->vmcb->control.event_inj)
>> +        return 1;
>> +
>>    svm->vmcb->control.exit_code = SVM_EXIT_INTR;
>>
>>    if (nested_svm_exit_handled(svm)) {
>> -- 
>> 1.6.0.2
>>
>> --
>> To unsubscribe from this list: send the line "unsubscribe kvm" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at  http://vger.kernel.org/majordomo-info.html