Re: [PATCH 2/8] vmxnet3: implement reta query and reta update

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

On 5/3/22 07:22, Pankaj Gupta wrote:
> Added reta query and reta update support for VMXNET3
> 
> Tested, using testpmd, for different hardware version on ESXi 7.0 Update 2.
> 
> Signed-off-by: Pankaj Gupta <pagupta@vmware.com>
> ---
>   drivers/net/vmxnet3/vmxnet3_ethdev.c | 67 ++++++++++++++++++++++++++++
>   1 file changed, 67 insertions(+)
> 
> diff --git a/drivers/net/vmxnet3/vmxnet3_ethdev.c b/drivers/net/vmxnet3/vmxnet3_ethdev.c
> index a7e1e5fef5..eb65499cf2 100644
> --- a/drivers/net/vmxnet3/vmxnet3_ethdev.c
> +++ b/drivers/net/vmxnet3/vmxnet3_ethdev.c
> @@ -95,6 +95,14 @@ static int vmxnet3_mac_addr_set(struct rte_eth_dev *dev,
>   				 struct rte_ether_addr *mac_addr);
>   static void vmxnet3_process_events(struct rte_eth_dev *dev);
>   static void vmxnet3_interrupt_handler(void *param);
> +static int
> +vmxnet3_rss_reta_update(struct rte_eth_dev *dev,
> +			struct rte_eth_rss_reta_entry64 *reta_conf,
> +			uint16_t reta_size);
> +static int
> +vmxnet3_rss_reta_query(struct rte_eth_dev *dev,
> +		       struct rte_eth_rss_reta_entry64 *reta_conf,
> +		       uint16_t reta_size);
>   static int vmxnet3_dev_rx_queue_intr_enable(struct rte_eth_dev *dev,
>   						uint16_t queue_id);
>   static int vmxnet3_dev_rx_queue_intr_disable(struct rte_eth_dev *dev,
> @@ -137,6 +145,8 @@ static const struct eth_dev_ops vmxnet3_eth_dev_ops = {
>   	.tx_queue_release     = vmxnet3_dev_tx_queue_release,
>   	.rx_queue_intr_enable = vmxnet3_dev_rx_queue_intr_enable,
>   	.rx_queue_intr_disable = vmxnet3_dev_rx_queue_intr_disable,
> +	.reta_update          = vmxnet3_rss_reta_update,
> +	.reta_query           = vmxnet3_rss_reta_query,
>   };
>   
>   struct vmxnet3_xstats_name_off {
> @@ -1696,3 +1706,60 @@ RTE_PMD_REGISTER_PCI_TABLE(net_vmxnet3, pci_id_vmxnet3_map);
>   RTE_PMD_REGISTER_KMOD_DEP(net_vmxnet3, "* igb_uio | uio_pci_generic | vfio-pci");
>   RTE_LOG_REGISTER_SUFFIX(vmxnet3_logtype_init, init, NOTICE);
>   RTE_LOG_REGISTER_SUFFIX(vmxnet3_logtype_driver, driver, NOTICE);
> +
> +static int
> +vmxnet3_rss_reta_update(struct rte_eth_dev *dev,
> +			struct rte_eth_rss_reta_entry64 *reta_conf,
> +			uint16_t reta_size)
> +{
> +	int i, idx, shift;
> +	struct vmxnet3_hw *hw = dev->data->dev_private;
> +	struct VMXNET3_RSSConf *dev_rss_conf = hw->rss_conf;
> +
> +	if (reta_size != dev_rss_conf->indTableSize) {
> +		PMD_DRV_LOG(ERR,
> +			"The size of hash lookup table configured (%d) doesn't match "
> +			"the supported number (%d)",
> +			reta_size, dev_rss_conf->indTableSize);
> +		return -EINVAL;
> +	}
> +
> +	for (i = 0; i < reta_size; i++) {
> +		idx = i / RTE_ETH_RETA_GROUP_SIZE;
> +		shift = i % RTE_ETH_RETA_GROUP_SIZE;
> +		if (reta_conf[idx].mask & (1ULL << shift))

Please, use RTE_BIT64(shift)

> +			dev_rss_conf->indTable[i] = (uint8_t)reta_conf[idx].reta[shift];

Not a problem, but as I understand explicit type cast is not required.

> +	}
> +
> +	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
> +				VMXNET3_CMD_UPDATE_RSSIDT);
> +
> +	return 0;
> +}
> +
> +static int
> +vmxnet3_rss_reta_query(struct rte_eth_dev *dev,
> +		       struct rte_eth_rss_reta_entry64 *reta_conf,
> +		       uint16_t reta_size)
> +{
> +	int i, idx, shift;
> +	struct vmxnet3_hw *hw = dev->data->dev_private;
> +	struct VMXNET3_RSSConf *dev_rss_conf = hw->rss_conf;
> +
> +	if (reta_size != dev_rss_conf->indTableSize) {
> +		PMD_DRV_LOG(ERR,
> +			"Size of requested hash lookup table (%d) doesn't "
> +			"match the configured size (%d)",
> +			reta_size, dev_rss_conf->indTableSize);
> +		return -EINVAL;
> +	}
> +
> +	for (i = 0; i < reta_size; i++) {
> +		idx = i / RTE_ETH_RETA_GROUP_SIZE;
> +		shift = i % RTE_ETH_RETA_GROUP_SIZE;
> +		if (reta_conf[idx].mask & (1ULL << shift))

Use RTE_BIT64(shift)

> +			reta_conf[idx].reta[shift] = dev_rss_conf->indTable[i];
> +	}
> +
> +	return 0;
> +}