Re: [Xen-devel] [PATCH v12 1/7] passthrough: don't migrate pirq when it is delivered through VT-d PI

[Chao Gao <chao.gao@xxxxxxxxx>]

On Fri, Apr 07, 2017 at 05:50:36AM -0600, Jan Beulich wrote:
>>>> On 07.04.17 at 06:07, <chao.gao@xxxxxxxxx> wrote:
>> Cc: kevin
>> 
>> On Fri, Apr 07, 2017 at 04:38:00AM -0600, Jan Beulich wrote:
>>>>>> On 06.04.17 at 02:30, <chao.gao@xxxxxxxxx> wrote:
>>>> --- a/xen/arch/x86/hvm/hvm.c
>>>> +++ b/xen/arch/x86/hvm/hvm.c
>>>> @@ -438,6 +438,9 @@ static int hvm_migrate_pirq(struct domain *d, struct 
>> hvm_pirq_dpci *pirq_dpci,
>>>>      struct vcpu *v = arg;
>>>>  
>>>>      if ( (pirq_dpci->flags & HVM_IRQ_DPCI_MACH_MSI) &&
>>>> +         (pirq_dpci->flags & HVM_IRQ_DPCI_GUEST_MSI) &&
>>>> +         /* Needn't migrate pirq if this pirq is delivered to guest 
>> directly.*/
>>>> +         (!pirq_dpci->gmsi.posted) &&
>>>>           (pirq_dpci->gmsi.dest_vcpu_id == v->vcpu_id) )
>>>>      {
>>>
>>>I don't think I've seen you address Kevin's comment on this for v11,
>>>and like Kevin I can't immediately see why the above addition would
>>>be correct. Do you perhaps mean
>>>
>>>    if ( (pirq_dpci->flags & HVM_IRQ_DPCI_MACH_MSI) &&
>>>         /* Needn't migrate pirq if this pirq is delivered to guest 
>>> directly.*/
>>>         (!pirq_dpci->gmsi.posted ||
>>>          <whatever is appropriate here, if anything>) &&
>>>         (pirq_dpci->gmsi.dest_vcpu_id == v->vcpu_id) )
>> 
>> Sorry to Kevin. And thanks to point it out. 
>> But I thought we had discussed this in 
>> https://lists.xenproject.org/archives/html/xen-devel/2017-03/msg04383.html. 
>> I 
>> did think you agreed with me.
>> gmsi is invalid when pirq_dpci is not GUEST_MSI, is there something I have
>> ignored?
>
>You've been talking about GUEST_PCI there, which I did (and do)
>agree we can't handle here. So for the purposes of your series,
>simply adding the gmsi.posted check would be the right thing imo.
>I don't think I see anything wrong with the ->gmsi accesses here:
>The GUEST_PCI code simply doesn't set them, so dest_vcpu_id
>will still be -1 (from pt_pirq_init()). So I don't see any bug being
>fixed here with the extra other check you add. If you agree, I
>can take that line and the commit message sentence out while
>committing.

Ok. I admit I said it's bug is wrong. feel free to do what you want.

Thanks
Chao

>
>Jan
>

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
https://lists.xen.org/xen-devel