Re: [PATCH v3 05/10] IOMMU: add common API for device reserved memory

[Jan Beulich <jbeulich@suse.com>]

On 04.08.2022 16:38, Marek Marczykowski-Górecki wrote:
> On Thu, Aug 04, 2022 at 04:25:38PM +0200, Jan Beulich wrote:
>> On 26.07.2022 05:23, Marek Marczykowski-Górecki wrote:
>>> --- a/xen/drivers/passthrough/iommu.c
>>> +++ b/xen/drivers/passthrough/iommu.c
>>> @@ -651,6 +651,51 @@ bool_t iommu_has_feature(struct domain *d, enum iommu_feature feature)
>>>      return is_iommu_enabled(d) && test_bit(feature, dom_iommu(d)->features);
>>>  }
>>>  
>>> +#define MAX_EXTRA_RESERVED_RANGES 20
>>> +struct extra_reserved_range {
>>> +    unsigned long start;
>>> +    unsigned long nr;
>>> +    uint32_t sbdf;
>>
>> It's not easy to judge why this isn't pci_sbdf_t when no callers
>> exist at this point.
> 
> I'm following here types used in the rest of IOMMU code. Especially,
> this field is later passed to iommu_grdm_t func, which is:
> 
> typedef int iommu_grdm_t(xen_pfn_t start, xen_ulong_t nr, u32 id, void *ctxt);
>                                                           ^^^^
> 
> I can probably use pci_sbdf_t here, but it will be cast to u32 later
> anyway...

No, rather than a cast you'd use the union's sbdf field. And yes, eventually
that function typedef you refer to will want switching to pci_sbdf_t as well.

Jan