On 15.08.2012, at 02:17, Scott Wood wrote:
> On 08/14/2012 06:04 PM, Alexander Graf wrote:
>> Generic KVM code might want to know whether we are inside guest context
>> or outside. It also wants to be able to push us out of guest context.
>>
>> Add support to the BookE code for the generic vcpu->mode field that describes
>> the above states.
>>
>> Signed-off-by: Alexander Graf <agraf@xxxxxxx>
>> ---
>> arch/powerpc/kvm/booke.c | 11 +++++++++++
>> 1 files changed, 11 insertions(+), 0 deletions(-)
>>
>> diff --git a/arch/powerpc/kvm/booke.c b/arch/powerpc/kvm/booke.c
>> index bcf87fe..70a86c0 100644
>> --- a/arch/powerpc/kvm/booke.c
>> +++ b/arch/powerpc/kvm/booke.c
>> @@ -501,6 +501,15 @@ static int kvmppc_prepare_to_enter(struct kvm_vcpu *vcpu)
>> continue;
>> }
>>
>> + if (vcpu->mode == EXITING_GUEST_MODE) {
>> + r = 1;
>> + break;
>> + }
>> +
>> + /* Going into guest context! Yay! */
>> + vcpu->mode = IN_GUEST_MODE;
>> + smp_wmb();
>> +
>> break;
>> }
>
> Normally on entry to this function mode should be OUTSIDE_GUEST_MODE,
> right? How could it possibly be EXITING_GUEST_MODE then, since that
> only replaces IN_GUEST_MODE?
>
> This doesn't match what x86 does with mode on entry. Mode is supposed
> to be set to IN_GUEST_MODE before requests are checked.
>
> I'm not sure what the point of EXITING_GUEST_MODE is at all, compared to
> just waiting until after interrupts are disabled before setting
> IN_GUEST_MODE (which we do on ppc, but not on x86 even though it seems
> like a trivial change), plus the existing ordering between mode and
> requests.
Well, the only real use case I could find for the mode was the remote vcpu kick. If we're not outside of guest mode, we get an IPI to notify us that requests are outstanding.
So I only get us into OUTSIDE_GUEST_MODE when we really exit __vcpu_run, thus are in user space. That doesn't reflect what x86 does, right, but so doesn't our whole loop concept.
However, since we might do the vcpu_block in this loop, we will never really get into OUTSIDE_GUEST_MODE, right? Hrm. So what would you suggest? Do all the handle_exit in OUTSIDE_GUEST_MODE scope and then reenter IN_GUEST_MODE in prepare_to_enter? We still wouldn't need an EXITING_GUEST_MODE though.
Alex
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
