[dpdk-dev] [PATCH 08/15] net/virtio: add virtio vDPA op to get features

Maxime Coquelin maxime.coquelin at redhat.com
Tue Sep 3 09:27:38 CEST 2019



On 9/2/19 8:43 AM, Tiwei Bie wrote:
> On Thu, Aug 29, 2019 at 09:59:53AM +0200, Maxime Coquelin wrote:
>> This patch implements the vDPA .get_features() callback.
>>
>> Signed-off-by: Maxime Coquelin <maxime.coquelin at redhat.com>
>> ---
>>  drivers/net/virtio/virtio_vdpa.c | 32 ++++++++++++++++++++++++++++++++
>>  1 file changed, 32 insertions(+)
>>
>> diff --git a/drivers/net/virtio/virtio_vdpa.c b/drivers/net/virtio/virtio_vdpa.c
>> index 07ff1e090..9e2af8313 100644
>> --- a/drivers/net/virtio/virtio_vdpa.c
>> +++ b/drivers/net/virtio/virtio_vdpa.c
>> @@ -173,8 +173,40 @@ virtio_vdpa_get_queue_num(int did, uint32_t *queue_num)
>>  	return 0;
>>  }
>>  
>> +static int
>> +virtio_vdpa_get_features(int did, uint64_t *features)
>> +{
>> +	struct internal_list *list;
>> +	struct virtio_vdpa_device *dev;
>> +
>> +	list = find_internal_resource_by_did(did);
>> +	if (list == NULL) {
>> +		DRV_LOG(ERR, "Invalid device id: %d", did);
>> +		return -1;
>> +	}
>> +
>> +	dev = list->dev;
>> +
>> +	*features = VTPCI_OPS(&dev->hw)->get_features(&dev->hw);
>> +
>> +	if (!(*features & (1ULL << VIRTIO_F_IOMMU_PLATFORM))) {
>> +		DRV_LOG(ERR, "Device does not support IOMMU");
>> +		return -1;
>> +	}
>> +
>> +	if (*features & (1ULL << VIRTIO_NET_F_CTRL_VQ))
>> +		dev->has_ctrl_vq = true;
>> +	else
>> +		dev->has_ctrl_vq = false;
>> +
>> +	*features |= (1ULL << VHOST_USER_F_PROTOCOL_FEATURES);
> 
> We need to drop the VIRTIO_F_IOMMU_PLATFORM bit before
> reporting features to the upper layer as the vDPA backend
> doesn't support the vIOMMU yet.

Indeed, it just work for now because Virtio-user does not advertise
IOMMU feature.

> Regards,
> Tiwei
> 
> 
>> +
>> +	return 0;
>> +}
>> +
>>  static struct rte_vdpa_dev_ops virtio_vdpa_ops = {
>>  	.get_queue_num = virtio_vdpa_get_queue_num,
>> +	.get_features = virtio_vdpa_get_features,
>>  };
>>  
>>  static inline int
>> -- 
>> 2.21.0
>>


More information about the dev mailing list