Re: [PATCH] xen: move PIRQ EOI map fields to arch_domain

["Roger Pau Monné" <roger.pau@citrix.com>]

On 08/04/14 15:01, Jan Beulich wrote:
>>>> On 08.04.14 at 14:16, <roger.pau@citrix.com> wrote:
>> On 08/04/14 13:33, Jan Beulich wrote:
>>>>>> On 08.04.14 at 12:33, <roger.pau@citrix.com> wrote:
>>>> @@ -354,7 +353,7 @@ ret_t do_physdev_op(int cmd, XEN_GUEST_HANDLE_PARAM(void) 
>> arg)
>>>>          }
>>>>          mfn = page_to_mfn(page);
>>>>  
>>>> -        if ( cmpxchg(&v->domain->arch.pv_domain.pirq_eoi_map_mfn,
>>>> +        if ( cmpxchg(&v->domain->arch.pirq_eoi_map_mfn,
>>>
>>> In order to get here you also need to modify hvm_physdev_op().
>>
>> This is on top of Mukesh PVH Dom0 v8 series (as stated on the 
>> description), which has the following in hvm_physdev_op:
>>
>> static long hvm_physdev_op(int cmd, XEN_GUEST_HANDLE_PARAM(void) arg)
>> {
>>     switch ( cmd )
>>     {
>>     default:
>>         if ( !is_pvh_vcpu(current) || !is_hardware_domain(current->domain) )
>>             return -ENOSYS;
>>         /* fall through */
>>     case PHYSDEVOP_map_pirq:
>>     case PHYSDEVOP_unmap_pirq:
>>     case PHYSDEVOP_eoi:
>>     case PHYSDEVOP_irq_status_query:
>>     case PHYSDEVOP_get_free_pirq:
>>         return do_physdev_op(cmd, arg);
>>     }
>> }
> 
> I don't recall that series to include a revert of a7ca5c40 ("x86/pvh:
> disallow PHYSDEVOP_pirq_eoi_gmfn_v2/v1"), and if it did it would
> be wrong, as then it would have to already include the changes you
> are doing here.

OK, now I see it, Mukesh Dom0 PVH v8 series is based on top of
b3c0519e019de3fcd993913ab48c327daec35115, which is previous to your
commit of a7ca5c40:

https://oss.oracle.com/git/?p=mrathor/xen.git;a=shortlog;h=dom0pvh-v8

Will rebase on top of unstable and post a v2, do you have any other
comments related to the patch that I should fix before posting v2?

Roger.