[PATCH 1/2] drm/virtio: Defer the host dumb buffer creation

[Jocelyn Falempe]

On 23/08/2024 09:04, Gerd Hoffmann wrote:
>> +static void virtio_gpu_deferred_create(struct virtio_gpu_object *bo,
>> +				       struct virtio_gpu_device *vgdev,
>> +				       const struct drm_mode_fb_cmd2 *mode_cmd)
>> +{
>> +	struct virtio_gpu_object_params params = { 0 };
>> +
>> +	params.format = virtio_gpu_translate_format(mode_cmd->pixel_format);
>> +	params.width = mode_cmd->width;
>> +	params.height = mode_cmd->height;
>> +	params.size = params.height * params.width * 4;
>> +	params.size = ALIGN(params.size, PAGE_SIZE);
>> +	params.dumb = true;
> 
> I'd suggest to simply store the complete virtio_gpu_object_params struct
> in virtio_gpu_object instead of re-creating it here.

The struct params is much bigger than the struct virtio_gpu_object, so I 
though it would waste too much memory. Using a pointer would add an 
alloc/free pair, and a potential source of memleak. And as we have the 
required parameters in struct drm_mode_fb_cmd2, I found it better this way.

> 
>> diff --git a/drivers/gpu/drm/virtio/virtgpu_drv.h b/drivers/gpu/drm/virtio/virtgpu_drv.h
>> index 64c236169db88..8d1e8dcfa8c15 100644
>> --- a/drivers/gpu/drm/virtio/virtgpu_drv.h
>> +++ b/drivers/gpu/drm/virtio/virtgpu_drv.h
>> @@ -93,6 +93,9 @@ struct virtio_gpu_object {
>>   	bool dumb;
>>   	bool created;
>>   	bool host3d_blob, guest_blob;
>> +	bool deferred;
>> +	struct virtio_gpu_mem_entry *ents;
>> +	unsigned int nents;
>>   	uint32_t blob_mem, blob_flags;
> 
> Put them into a new block separated by newline, add a comment saying
> those are needed for virtio_gpu_deferred_create?
Yes, I will do that in v2

> 
>> @@ -229,9 +231,14 @@ int virtio_gpu_object_create(struct virtio_gpu_device *vgdev,
>>   						  objs, fence);
>>   		virtio_gpu_object_attach(vgdev, bo, ents, nents);
>>   	} else {
>> -		virtio_gpu_cmd_create_resource(vgdev, bo, params,
>> -					       objs, fence);
>> -		virtio_gpu_object_attach(vgdev, bo, ents, nents);
>> +		/* Fence is used only with blob or virgl */
>> +		WARN_ONCE(fence != NULL, "deferred doesn't support fence\n");
> 
> Additionally check for param->dumb to take the deferred code path?  That
> should make sure there is no fence.

I think I can also duplicate virtio_gpu_object_create(), and have one 
version only for deferred dumb buffer. I will see if that doesn't make 
too much code duplication.

> 
> take care,
>    Gerd
> 

Thanks for the review,

-- 

Jocelyn