[dpdk-dev] [PATCH v6 4/6] ether: Check VMDq RSS mode

Ouyang, Changchun changchun.ouyang at intel.com
Tue Jan 20 02:03:25 CET 2015



> -----Original Message-----
> From: Wodkowski, PawelX
> Sent: Monday, January 19, 2015 6:31 PM
> To: Ouyang, Changchun; dev at dpdk.org
> Cc: Thomas Monjalon; Vlad Zolotarov
> Subject: RE: [dpdk-dev] [PATCH v6 4/6] ether: Check VMDq RSS mode
> 
> > -----Original Message-----
> > From: dev [mailto:dev-bounces at dpdk.org] On Behalf Of Ouyang
> Changchun
> > Sent: Monday, January 12, 2015 6:59 AM
> > To: dev at dpdk.org
> > Subject: [dpdk-dev] [PATCH v6 4/6] ether: Check VMDq RSS mode
> >
> > Check mq mode for VMDq RSS, handle it correctly instead of returning
> > an error; Also remove the limitation of per pool queue number has max
> > value of 1, because the per pool queue number could be 2 or 4 if it is
> > VMDq RSS mode;
> >
> > The number of rxq specified in config will determine the mq mode for
> > VMDq RSS.
> >
> > Signed-off-by: Changchun Ouyang <changchun.ouyang at intel.com>
> >
> > changes in v6:
> >   - More clear error message when queue number is invalid.
> >
> > changes in v5:
> >   - Fix '<' issue, it should be '<=' to test rxq number;
> >   - Extract a function to remove the embeded switch-case statement.
> >
> > ---
> >  lib/librte_ether/rte_ethdev.c | 51
> > ++++++++++++++++++++++++++++++++++++++-----
> >  1 file changed, 46 insertions(+), 5 deletions(-)
> >
> > diff --git a/lib/librte_ether/rte_ethdev.c
> > b/lib/librte_ether/rte_ethdev.c index 95f2ceb..e9e3368 100644
> > --- a/lib/librte_ether/rte_ethdev.c
> > +++ b/lib/librte_ether/rte_ethdev.c
> > @@ -503,6 +503,31 @@ rte_eth_dev_tx_queue_config(struct
> rte_eth_dev
> > *dev, uint16_t nb_queues)  }
> >
> >  static int
> > +rte_eth_dev_check_vf_rss_rxq_num(uint8_t port_id, uint16_t nb_rx_q)
> {
> > +	struct rte_eth_dev *dev = &rte_eth_devices[port_id];
> > +	switch (nb_rx_q) {
> > +	case 1:
> > +	case 2:
> > +		RTE_ETH_DEV_SRIOV(dev).active =
> > +			ETH_64_POOLS;
> > +		break;
> > +	case 4:
> > +		RTE_ETH_DEV_SRIOV(dev).active =
> > +			ETH_32_POOLS;
> > +		break;
> > +	default:
> > +		return -EINVAL;
> > +	}
> > +
> > +	RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool = nb_rx_q;
> > +	RTE_ETH_DEV_SRIOV(dev).def_pool_q_idx =
> > +		dev->pci_dev->max_vfs * nb_rx_q;
> > +
> > +	return 0;
> > +}
> > +
> > +static int
> >  rte_eth_dev_check_mq_mode(uint8_t port_id, uint16_t nb_rx_q,
> uint16_t
> > nb_tx_q,
> >  		      const struct rte_eth_conf *dev_conf)  { @@ -510,8 +535,7
> @@
> > rte_eth_dev_check_mq_mode(uint8_t port_id, uint16_t nb_rx_q,
> uint16_t
> > nb_tx_q,
> >
> >  	if (RTE_ETH_DEV_SRIOV(dev).active != 0) {
> >  		/* check multi-queue mode */
> > -		if ((dev_conf->rxmode.mq_mode == ETH_MQ_RX_RSS) ||
> > -		    (dev_conf->rxmode.mq_mode == ETH_MQ_RX_DCB) ||
> > +		if ((dev_conf->rxmode.mq_mode == ETH_MQ_RX_DCB) ||
> >  		    (dev_conf->rxmode.mq_mode == ETH_MQ_RX_DCB_RSS)
> ||
> >  		    (dev_conf->txmode.mq_mode == ETH_MQ_TX_DCB)) {
> >  			/* SRIOV only works in VMDq enable mode */ @@ -
> 525,7 +549,6 @@
> > rte_eth_dev_check_mq_mode(uint8_t port_id, uint16_t nb_rx_q,
> uint16_t
> > nb_tx_q,
> >  		}
> >
> >  		switch (dev_conf->rxmode.mq_mode) {
> > -		case ETH_MQ_RX_VMDQ_RSS:
> >  		case ETH_MQ_RX_VMDQ_DCB:
> >  		case ETH_MQ_RX_VMDQ_DCB_RSS:
> >  			/* DCB/RSS VMDQ in SRIOV mode, not implement
> yet */ @@ -534,6
> > +557,26 @@ rte_eth_dev_check_mq_mode(uint8_t port_id, uint16_t
> > nb_rx_q, uint16_t nb_tx_q,
> >  					"unsupported VMDQ mq_mode rx
> > %u\n",
> >  					port_id, dev_conf-
> > >rxmode.mq_mode);
> >  			return (-EINVAL);
> > +		case ETH_MQ_RX_RSS:
> > +			PMD_DEBUG_TRACE("ethdev port_id=%" PRIu8
> > +					" SRIOV active, "
> > +					"Rx mq mode is changed from:"
> > +					"mq_mode %u into VMDQ mq_mode
> > %u\n",
> > +					port_id,
> > +					dev_conf->rxmode.mq_mode,
> > +					dev->data-
> > >dev_conf.rxmode.mq_mode);
> > +		case ETH_MQ_RX_VMDQ_RSS:
> > +			dev->data->dev_conf.rxmode.mq_mode =
> > ETH_MQ_RX_VMDQ_RSS;
> > +			if (nb_rx_q <=
> > RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool)
> > +				if
> (rte_eth_dev_check_vf_rss_rxq_num(port_id,
> > nb_rx_q) != 0) {
> > +					PMD_DEBUG_TRACE("ethdev
> > port_id=%d"
> > +						" SRIOV active, invalid queue"
> > +						" number for VMDQ RSS,
> > allowed"
> > +						" value are 1, 2 or 4\n",
> > +						port_id);
> > +					return -EINVAL;
> > +				}
> > +			break;
> >  		default: /* ETH_MQ_RX_VMDQ_ONLY or
> ETH_MQ_RX_NONE */
> >  			/* if nothing mq mode configure, use default scheme
> */
> >  			dev->data->dev_conf.rxmode.mq_mode =
> ETH_MQ_RX_VMDQ_ONLY; @@
> > -553,8 +596,6 @@ rte_eth_dev_check_mq_mode(uint8_t port_id,
> uint16_t
> > nb_rx_q, uint16_t nb_tx_q,
> >  		default: /* ETH_MQ_TX_VMDQ_ONLY or
> ETH_MQ_TX_NONE */
> >  			/* if nothing mq mode configure, use default scheme
> */
> >  			dev->data->dev_conf.txmode.mq_mode =
> ETH_MQ_TX_VMDQ_ONLY;
> > -			if (RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool > 1)
> > -				RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool =
> 1;
> 
> Why you removed this completely?
> 
I have discuss this with Valid in previous email,
The reason for removing it, is that when  rxmode.mq_mode is ETH_MQ_RX_VMDQ_RSS, and txmode.mq_mode is ETH_MQ_TX_NONE,
The logic is incorrect, in such a case the nb_q_per_pool could be 1, or 2, or 4,
Forcedly set it into 1 will always disable the vf rss.

I see you will split nb_q_per_pool into nb_rxq_per_pool and nb_txq_per_pool, that's good way to meet rss, dcb different cases requirement,
So you need consider such case, don't break anything.



More information about the dev mailing list