Re: [Xen-devel] [PATCH v4 12/21] xen/passthrough: iommu: Split generic IOMMU code

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 22.04.14 at 20:02, <julien.grall@xxxxxxxxxx> wrote:
> On 04/22/2014 05:59 PM, Jan Beulich wrote:
>>>>> On 22.04.14 at 18:45, <julien.grall@xxxxxxxxxx> wrote:
>>> On 04/22/2014 05:33 PM, Jan Beulich wrote:
>>>>>>> On 22.04.14 at 16:58, <julien.grall@xxxxxxxxxx> wrote:
>>>>>>> +void __hwdom_init arch_iommu_check_hwdom_reqs(struct domain *d)
>>>>>>> +{
>>>>>>> +    if ( !iommu_enabled )
>>>>>>> +        panic("Presently, iommu must be enabled for pvh dom0\n");
>>>>>>> +}
>>>>>>
>>>>>> Message text (containing PVH) and function name (not containing
>>>>>> PVH) don't fit together, nor does the conditional really establish a
>>>>>> connection.
>>>>>
>>>>> Do you prefer a comment, or an explicit check to is_pvh_domain(d)?
>>>>
>>>> That depends on where it would go: If the caller checks for PVH, then
>>>> the function name should change. If the caller doesn't, then I don't
>>>> see how you'd avoid getting here for non-PVH.
>>>
>>> The caller will go there when the DOM0 is auto-translated (i.e PVH as
>>> dom0 can't be an HVM).
>>>
>>> I can remove PVH from the log, but for the user it's not accurate.
>> 
>> In which case the function name should reflect this.
> 
> What about arch_iommu_check_autotranslate_hwdom_reqs?

Getting quite long, but seems okay. Perhaps drop the _reqs?

Jan


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