> On Mar 9, 2017, at 8:18 AM, Yigit, Ferruh <ferruh.yi...@intel.com> wrote:
>
> On 3/7/2017 4:31 PM, Pascal Mazon wrote:
>> Tap PMD is flexible, it supports any speed.
>>
>> Signed-off-by: Pascal Mazon <pascal.ma...@6wind.com>
>> ---
>> doc/guides/nics/features/tap.ini | 1 +
>> drivers/net/tap/rte_eth_tap.c | 35 +++++++++++++++++++++++++++++++++++
>> 2 files changed, 36 insertions(+)
>>
>> diff --git a/doc/guides/nics/features/tap.ini
>> b/doc/guides/nics/features/tap.ini
>> index d9b47a003654..dad5a0561087 100644
>> --- a/doc/guides/nics/features/tap.ini
>> +++ b/doc/guides/nics/features/tap.ini
>> @@ -9,6 +9,7 @@ Jumbo frame = Y
>> Promiscuous mode = Y
>> Allmulticast mode = Y
>> Basic stats = Y
>> +Speed capabilities = Y
>> Unicast MAC filter = Y
>> Other kdrv = Y
>> ARMv7 = Y
>> diff --git a/drivers/net/tap/rte_eth_tap.c b/drivers/net/tap/rte_eth_tap.c
>> index 1e46ee36efa2..ef525a3f0826 100644
>> --- a/drivers/net/tap/rte_eth_tap.c
>> +++ b/drivers/net/tap/rte_eth_tap.c
>> @@ -351,6 +351,40 @@ tap_dev_configure(struct rte_eth_dev *dev __rte_unused)
>> return 0;
>> }
>>
>> +static uint32_t
>> +tap_dev_speed_capa(void)
>> +{
>> + uint32_t speed = pmd_link.link_speed;
>
> link_speed is already hardcoded into PMD, so there is nothing to detect
> here. Would it be different if PMD directly return pmd_link.link_speed?
The link speed is passed into the PMD via the command line, which means it can
change per run.
>
>> + uint32_t capa = 0;
>> +
>> + if (speed >= ETH_SPEED_NUM_10M)
>> + capa |= ETH_LINK_SPEED_10M;
>> + if (speed >= ETH_SPEED_NUM_100M)
>> + capa |= ETH_LINK_SPEED_100M;
>> + if (speed >= ETH_SPEED_NUM_1G)
>> + capa |= ETH_LINK_SPEED_1G;
>> + if (speed >= ETH_SPEED_NUM_5G)
>> + capa |= ETH_LINK_SPEED_2_5G;
>> + if (speed >= ETH_SPEED_NUM_5G)
>> + capa |= ETH_LINK_SPEED_5G;
>> + if (speed >= ETH_SPEED_NUM_10G)
>> + capa |= ETH_LINK_SPEED_10G;
>> + if (speed >= ETH_SPEED_NUM_20G)
>> + capa |= ETH_LINK_SPEED_20G;
>> + if (speed >= ETH_SPEED_NUM_25G)
>> + capa |= ETH_LINK_SPEED_25G;
>> + if (speed >= ETH_SPEED_NUM_40G)
>> + capa |= ETH_LINK_SPEED_40G;
>> + if (speed >= ETH_SPEED_NUM_50G)
>> + capa |= ETH_LINK_SPEED_50G;
>> + if (speed >= ETH_SPEED_NUM_56G)
>> + capa |= ETH_LINK_SPEED_56G;
>> + if (speed >= ETH_SPEED_NUM_100G)
>> + capa |= ETH_LINK_SPEED_100G;
>
> I would prefer switch here [1], but functionally both are same, it is
> your call.
>
> [1]
> switch (speed) {
> case ETH_SPEED_NUM_100G:
> capa |= ETH_LINK_SPEED_100G
> /* fallthrough */
> case ETH_SPEED_NUM_56G:
> capa |= ETH_LINK_SPEED_56G
> /* fallthrough */
> ...
> };
>
>
>> +
>> + return capa;
>> +}
>> +
>> static void
>> tap_dev_info(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
>> {
>> @@ -363,6 +397,7 @@ tap_dev_info(struct rte_eth_dev *dev, struct
>> rte_eth_dev_info *dev_info)
>> dev_info->max_tx_queues = internals->nb_queues;
>> dev_info->min_rx_bufsize = 0;
>> dev_info->pci_dev = NULL;
>> + dev_info->speed_capa = tap_dev_speed_capa();
>> }
>>
>> static void
Regards,
Keith
