Re: [dpdk-dev] [PATCH] net/virtio: report max/min/align desc limits in dev info get

[Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>]

On 9/13/21 10:52 PM, Maxime Coquelin wrote:
> 
> 
> On 8/20/21 2:48 PM, Andrew Rybchenko wrote:
>> From: Ivan Ilchenko <ivan.ilchenko@oktetlabs.ru>
>>
>> Report max/min/align descriptors limits in device info get callback.
>> Before calling the callback, rte_eth_dev_info_get() provides
>> default values of nb_min as zero and nb_max as UINT16_MAX that are
>> not correct for the driver, so one can't rely on them.
>>
>> Signed-off-by: Ivan Ilchenko <ivan.ilchenko@oktetlabs.ru>
>> Signed-off-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
>> ---
>>  drivers/net/virtio/virtio_ethdev.c | 25 +++++++++++++++++++++++++
>>  1 file changed, 25 insertions(+)
>>
>> diff --git a/drivers/net/virtio/virtio_ethdev.c b/drivers/net/virtio/virtio_ethdev.c
>> index e58085a2c9..601c03e079 100644
>> --- a/drivers/net/virtio/virtio_ethdev.c
>> +++ b/drivers/net/virtio/virtio_ethdev.c
>> @@ -33,6 +33,7 @@
>>  #include "virtio_logs.h"
>>  #include "virtqueue.h"
>>  #include "virtio_rxtx.h"
>> +#include "virtio_rxtx_simple.h"
>>  #include "virtio_user/virtio_user_dev.h"
>>  
>>  static int  virtio_dev_configure(struct rte_eth_dev *dev);
>> @@ -2536,6 +2537,30 @@ virtio_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
>>  	if ((host_features & tso_mask) == tso_mask)
>>  		dev_info->tx_offload_capa |= DEV_TX_OFFLOAD_TCP_TSO;
>>  
>> +	if (host_features & (1ULL << VIRTIO_F_RING_PACKED)) {
>> +		/*
>> +		 * According to 2.7 Packed Virtqueues,
>> +		 * 2.7.10.1 Structure Size and Alignment:
>> +		 * The Queue Size value does not have to be a power of 2.
>> +		 */
>> +		dev_info->rx_desc_lim.nb_max = UINT16_MAX;
>> +	} else {
>> +		/*
>> +		 * According to 2.6 Split Virtqueues:
>> +		 * Queue Size value is always a power of 2. The maximum Queue
>> +		 * Size value is 32768.
>> +		 */
>> +		dev_info->rx_desc_lim.nb_max = 32768;
>> +	}
>> +	/*
>> +	 * Actual minimum is not the same for virtqueues of different kinds,
>> +	 * but to avoid tangling the code with separate branches, rely on
>> +	 * default thresholds since desc number must be at least of their size.
>> +	 */
>> +	dev_info->rx_desc_lim.nb_min = RTE_MAX(DEFAULT_RX_FREE_THRESH,
>> +					       RTE_VIRTIO_VPMD_RX_REARM_THRESH);
>> +	dev_info->rx_desc_lim.nb_align = 1;
>> +
> 
> It makes sense, but shouldn't we do the same for dev_info->tx_desc_lim?

Yes, you're right. We'll care about it.

I suggest to add "Rx" in the summary and care about Tx in a
subsequent patch.

net/virtio: report max/min/align Rx desc limits in dev info

OK?