RE: [PATCH] net/iavf: fix mismatch between rx_pkt_burst and RX descriptor

["Zhou, YidingX" <yidingx.zhou@intel.com>]

> -----Original Message-----
> From: Zhang, Qi Z <qi.z.zhang@intel.com>
> Sent: Saturday, May 7, 2022 10:09 AM
> To: Zhou, YidingX <yidingx.zhou@intel.com>; Wu, Jingjing
> <jingjing.wu@intel.com>; Xing, Beilei <beilei.xing@intel.com>
> Cc: Yang, Qiming <qiming.yang@intel.com>; stable@dpdk.org; Yeleswarapu,
> Ramamani <ramamani.yeleswarapu@intel.com>
> Subject: RE: [PATCH] net/iavf: fix mismatch between rx_pkt_burst and RX
> descriptor
> 
> 
> 
> > -----Original Message-----
> > From: Zhou, YidingX <yidingx.zhou@intel.com>
> > Sent: Saturday, May 7, 2022 5:34 PM
> > To: Wu, Jingjing <jingjing.wu@intel.com>; Xing, Beilei
> > <beilei.xing@intel.com>
> > Cc: Yang, Qiming <qiming.yang@intel.com>; Zhang, Qi Z
> > <qi.z.zhang@intel.com>; stable@dpdk.org; Yeleswarapu, Ramamani
> > <ramamani.yeleswarapu@intel.com>
> > Subject: [PATCH] net/iavf: fix mismatch between rx_pkt_burst and RX
> > descriptor
> >
> > Some kernel drivers return the capability
> > VIRTCHNL_VF_OFFLOAD_RX_FLEX_DESC when
> IAVF_RXDID_COMMS_OVS_1 is not
> > supported. This causes PMD to use rx_pkt_burst that handles the Flex
> > Receive Descriptor format, but actually configures the RXDID into
> > IAVF_RXDID_LEGACY_1, then the fields of rte_mbuf Will be filled with
> > wrong values in rx_pkt_burst, which will eventually lead to coredump.
> >
> > This patch fixes mismatch between rx_pkt_burst and rx descriptor.
> >
> > Fixes: 12b435bf8f2f ("net/iavf: support flex desc metadata
> > extraction")
> > Cc: stable@dpdk.org
> >
> > Signed-off-by: Yiding Zhou <yidingx.zhou@intel.com>
> > ---
> >  drivers/net/iavf/iavf_rxtx.c | 20 ++++++++++++++------
> >  1 file changed, 14 insertions(+), 6 deletions(-)
> >
> > diff --git a/drivers/net/iavf/iavf_rxtx.c
> > b/drivers/net/iavf/iavf_rxtx.c index 345f6aeebc..69584264de 100644
> > --- a/drivers/net/iavf/iavf_rxtx.c
> > +++ b/drivers/net/iavf/iavf_rxtx.c
> > @@ -2908,6 +2908,18 @@ iavf_set_rx_function(struct rte_eth_dev *dev)
> > bool use_avx512 = false;  bool use_flex = false;
> >
> > +if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RX_FLEX_DESC)
> > +use_flex = true;
> 
> No need this check, we can init use_flex as true;
> 

I'm not sure if use_flex can be init as true. is it possible that vf->vf_res->vf_cap_flags doesn't contain VIRTCHNL_VF_OFFLOAD_RX_FLEX_DESC?

> > +
> > +for (i = 0; i < dev->data->nb_rx_queues; i++) { rxq =
> > +dev->data->rx_queues[i]; if (rxq->rxdid <= IAVF_RXDID_LEGACY_1 ||
> > +!(vf->supported_rxdid & BIT(rxq->rxdid))) {
> 
> Check if rxq->rxdid is in supported list is not necessary, this has been
> guaranteed when we set it.
> 

According to debugging, this is not guaranteed.
rxq->rxdid is set to IAVF_RXDID_COMMS_OVS_1 by default when vf->supported_rxdid == 6 (just contains IAVF_RXDID_LEGACY_1 and IAVF_RXDID_FLEX_NIC).

> Also its better to print some warning message here,  if we saw some rxq-
> >rxdid is flex and some rxq->rxdid is legacy as we have to set rx_burst as
> legacy for all , this is something not be expected by user
> 
> 
Agreed, I will do this in v2. Thanks.