[dpdk-dev] [PATCH] net/fm10k: fix secondary process crash

[Wang, Xiao W]

Hi Mark,

> -----Original Message-----
> From: Chen, Jing D
> Sent: Friday, March 31, 2017 9:30 AM
> To: Wang, Xiao W <xiao.w.wang at intel.com>
> Cc: dev at dpdk.org; stable at dpdk.org
> Subject: RE: [PATCH] net/fm10k: fix secondary process crash
> 
> 
> 
> > -----Original Message-----
> > From: Wang, Xiao W
> > Sent: Tuesday, March 28, 2017 11:59 AM
> > To: Chen, Jing D <jing.d.chen at intel.com>
> > Cc: dev at dpdk.org; Wang, Xiao W <xiao.w.wang at intel.com>;
> stable at dpdk.org
> > Subject: [PATCH] net/fm10k: fix secondary process crash
> >
> > If the primary process has initialized all the queues to vector pmd mode,
> the
> > secondary process should not use scalar code path, because the per queue
> data
> > structures haven't been prepared for that, e.g. txq->ops is for vector Tx
> rather
> > than scalar Tx.
> >
> > Fixes: a6ce64a97520 ("fm10k: introduce vector driver")
> > Cc: stable at dpdk.org
> >
> > Signed-off-by: Xiao Wang <xiao.w.wang at intel.com>
> > ---
> >  drivers/net/fm10k/fm10k_ethdev.c | 28 ++++++++++++++++++++++++++--
> >  1 file changed, 26 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/net/fm10k/fm10k_ethdev.c
> > b/drivers/net/fm10k/fm10k_ethdev.c
> > index 388f929..680d617 100644
> > --- a/drivers/net/fm10k/fm10k_ethdev.c
> > +++ b/drivers/net/fm10k/fm10k_ethdev.c
> > @@ -2750,6 +2750,21 @@ static void __attribute__((cold))
> >  	int use_sse = 1;
> >  	uint16_t tx_ftag_en = 0;
> >
> > +	if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
> > +		/* primary process has set the ftag flag and txq_flags */
> > +		txq = dev->data->tx_queues[0];
> > +		if (fm10k_tx_vec_condition_check(txq)) {
> > +			dev->tx_pkt_burst = fm10k_xmit_pkts;
> > +			dev->tx_pkt_prepare = fm10k_prep_pkts;
> > +			PMD_INIT_LOG(DEBUG, "Use regular Tx func");
> > +		} else {
> > +			PMD_INIT_LOG(DEBUG, "Use vector Tx func");
> > +			dev->tx_pkt_burst = fm10k_xmit_pkts_vec;
> > +			dev->tx_pkt_prepare = NULL;
> > +		}
> > +		return;
> > +	}
> > +
> 
> Why we need to check process type? What would happen if no changes
> made here?

If no change, then this function will re-set some fields of txq structure.
e.g.       for (i = 0; i < dev->data->nb_tx_queues; i++) {
            		txq = dev->data->tx_queues[i];
            		fm10k_txq_vec_setup(txq);
       	}
Though these fields would be re-set to the same value, it doesn't look good. In secondary, we just read the queues and do not write them.

Best Regards,
Xiao
> 
> >  	if (fm10k_check_ftag(dev->device->devargs))
> >  		tx_ftag_en = 1;
> >
> > @@ -2810,6 +2825,9 @@ static void __attribute__((cold))
> >  	else
> >  		PMD_INIT_LOG(DEBUG, "Use regular Rx func");
> >
> > +	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
> > +		return;
> > +
> >  	for (i = 0; i < dev->data->nb_rx_queues; i++) {
> >  		struct fm10k_rx_queue *rxq = dev->data->rx_queues[i];
> >
> > @@ -2856,9 +2874,15 @@ static void __attribute__((cold))
> >  	dev->tx_pkt_burst = &fm10k_xmit_pkts;
> >  	dev->tx_pkt_prepare = &fm10k_prep_pkts;
> >
> > -	/* only initialize in the primary process */
> > -	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
> > +	/*
> > +	 * Primary process does the whole initialization, for secondary
> > +	 * processes, we just select the same Rx and Tx function as primary.
> > +	 */
> > +	if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
> > +		fm10k_set_rx_function(dev);
> > +		fm10k_set_tx_function(dev);
> >  		return 0;
> > +	}
> >
> >  	rte_eth_copy_pci_info(dev, pdev);
> >  	dev->data->dev_flags |= RTE_ETH_DEV_DETACHABLE;
> > --
> > 1.8.3.1