Re: [PATCH v4 08/14] vfio/common: Record DMA mapped IOVA ranges

[Joao Martins]

On 07/03/2023 10:08, Cédric Le Goater wrote:
> On 3/7/23 03:57, Alex Williamson wrote:
>> On Tue,  7 Mar 2023 02:02:52 +0000
>> Joao Martins <joao.m.martins@oracle.com> wrote:
>>> +static void vfio_dirty_tracking_init(VFIOContainer *container,
>>> +                                     VFIODirtyRanges *dirty)
>>> +{
>>> +    memset(dirty, 0, sizeof(*dirty));
>>> +    dirty->ranges.min32 = UINT32_MAX;
>>> +    dirty->ranges.min64 = UINT64_MAX;
>>> +    dirty->listener = vfio_dirty_tracking_listener;
>>> +    dirty->container = container;
>>> +
>>
>> I was actually thinking the caller would just pass
>> VFIODirtyTrackingRange and VFIODirtyRanges would be allocated on the
>> stack here, perhaps both are defined private to this file, but this
>> works and we can refine later if we so decide.  
> 
> It is true that vfio_devices_dma_logging_start() only needs
> a VFIODirtyTrackingRange struct and not the VFIODirtyRanges struct
> which is a temporary structure for the dirty ranges calculation.
> That would be nicer to have if you respin a v5.
> 
I can.

> I would rename VFIODirtyRanges to VFIODirtyRangesListener and
> VFIODirtyTrackingRange to VFIODirtyRanges.
> 
Better naming indeed.

> I am not sure they need to be in include/hw/vfio/vfio-common.h but
> that seems to be the VFIO practice.
> 
I can move as Alex also suggested it. There's already have
vfio_giommu_dirty_notifier and VFIOBitmap as private structures. I don't expect
that this will be used by other files