RE: [TEST v10] ethdev: support single queue per port

["Morten Brørup" <mb@smartsharesystems.com>]

> From: Stephen Hemminger [mailto:stephen@networkplumber.org]
> Sent: Thursday, 7 November 2024 00.26
T v10] ethdev: support single queue per port
> 
> On Wed,  6 Nov 2024 22:02:16 +0000
> Morten Brørup <mb@smartsharesystems.com> wrote:
> 
> > iff --git a/drivers/net/vmxnet3/vmxnet3_ethdev.c
> b/drivers/net/vmxnet3/vmxnet3_ethdev.c
> > index 78fac63ab6..1752c58069 100644
> > --- a/drivers/net/vmxnet3/vmxnet3_ethdev.c
> > +++ b/drivers/net/vmxnet3/vmxnet3_ethdev.c
> > @@ -1470,42 +1470,52 @@ vmxnet3_dev_stats_get(struct rte_eth_dev
> *dev, struct rte_eth_stats *stats)
> >  	struct vmxnet3_hw *hw = dev->data->dev_private;
> >  	struct UPT1_TxStats txStats;
> >  	struct UPT1_RxStats rxStats;
> > +	uint64_t packets, bytes;
> >
> >  	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
> VMXNET3_CMD_GET_STATS);
> >
> >  	for (i = 0; i < hw->num_tx_queues; i++) {
> >  		vmxnet3_tx_stats_get(hw, i, &txStats);
> >
> > -		stats->q_opackets[i] = txStats.ucastPktsTxOK +
> > +		packets = txStats.ucastPktsTxOK +
> >  			txStats.mcastPktsTxOK +
> >  			txStats.bcastPktsTxOK;
> >
> > -		stats->q_obytes[i] = txStats.ucastBytesTxOK +
> > +		bytes = txStats.ucastBytesTxOK +
> >  			txStats.mcastBytesTxOK +
> >  			txStats.bcastBytesTxOK;
> >
> > -		stats->opackets += stats->q_opackets[i];
> > -		stats->obytes += stats->q_obytes[i];
> > +		stats->opackets += packets;
> > +		stats->obytes += bytes;
> >  		stats->oerrors += txStats.pktsTxError +
> txStats.pktsTxDiscard;
> > +
> > +		if (i < RTE_ETHDEV_QUEUE_STAT_CNTRS) {
> > +			stats->q_opackets[i] = packets;
> > +			stats->q_obytes[i] = bytes;
> > +		}
> >  	}
> >
> >  	for (i = 0; i < hw->num_rx_queues; i++) {
> >  		vmxnet3_rx_stats_get(hw, i, &rxStats);
> >
> > -		stats->q_ipackets[i] = rxStats.ucastPktsRxOK +
> > +		packets = rxStats.ucastPktsRxOK +
> >  			rxStats.mcastPktsRxOK +
> >  			rxStats.bcastPktsRxOK;
> >
> > -		stats->q_ibytes[i] = rxStats.ucastBytesRxOK +
> > +		bytes = rxStats.ucastBytesRxOK +
> >  			rxStats.mcastBytesRxOK +
> >  			rxStats.bcastBytesRxOK;
> >
> > -		stats->ipackets += stats->q_ipackets[i];
> > -		stats->ibytes += stats->q_ibytes[i];
> > -
> > -		stats->q_errors[i] = rxStats.pktsRxError;
> > +		stats->ipackets += packets;
> > +		stats->ibytes += bytes;
> >  		stats->ierrors += rxStats.pktsRxError;
> >  		stats->imissed += rxStats.pktsRxOutOfBuf;
> > +
> > +		if (i < RTE_ETHDEV_QUEUE_STAT_CNTRS) {
> > +			stats->q_ipackets[i] = packets;
> > +			stats->q_ibytes[i] = bytes;
> > +			stats->q_errors[i] = rxStats.pktsRxError;
> > +		}
> >  	}
> >
> >  	return 0;
> 
> This fixes a bug in existing code of RTE_ETHDEV_QUEUE_STAT_CNTRS < num
> queues.
> Probably deserves its own patch with Fixes

Thanks for noticing, Stephen.

Ferruh also noticed, and has already applied such a patch:
https://patchwork.dpdk.org/project/dpdk/patch/20241104105220.1421305-1-mb@smartsharesystems.com/

The turnaround time was unusual quick on that one - it went in and out of Patchwork in a flash. Don't blink! ;-)