Re: [Xen-devel] [PATCH] x86: properly handle MSI-X unmask operation from guests

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 11.11.13 at 17:25, "Jan Beulich" <JBeulich@xxxxxxxx> wrote:
>>>> On 11.11.13 at 17:16, "Xu, Dongxiao" <dongxiao.xu@xxxxxxxxx> wrote:
>>>  -----Original Message-----
>>> From: xen-devel-bounces@xxxxxxxxxxxxx 
>>> [mailto:xen-devel-bounces@xxxxxxxxxxxxx] On Behalf Of Jan Beulich
>>> But I this should be done entirely inside the hypervisor - after all
>>> it is the hypervisor that forwards the request to qemu, so upon
>>> completion of the request it could as well do the necessary
>>> unmasking (as long as it doesn't need the interrupt masked for
>>> internal purposes).
>> 
>> Currently, Feng's patch handles the unmask operation when QEMU issues the 
>> MSI-X update hypercall (XEN_DOMCTL_bind_pt_irq).
>> However, if add the unmask operation in Xen's generic I/O emulation code 
>> path, we may need to introduce a new MSI-X specific callback function, which 
>> is supposed to be executed after QEMU I/O operation is done. Is this what 
> you 
>> mean?
> 
> Yes.

But of course that's the more difficult to implement operation.
If there's a way to forward the necessary information from
the write path to the path handling whatever qemu issues as
operations to the hypervisor in response to the writes
(_without_ altering the qemu interface), that might still be the
easier route implementation wise.

Jan


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