On 2/6/2018 11:21 AM, zhouyangchao wrote: Can you please provide more information why this patch is needed?
> Signed-off-by: Yangchao Zhou <zhouya...@gmail.com> > --- > drivers/net/bnx2x/bnx2x_rxtx.c | 8 +++++--- > 1 file changed, 5 insertions(+), 3 deletions(-) > > diff --git a/drivers/net/bnx2x/bnx2x_rxtx.c b/drivers/net/bnx2x/bnx2x_rxtx.c > index a0d4ac9..d8a3225 100644 > --- a/drivers/net/bnx2x/bnx2x_rxtx.c > +++ b/drivers/net/bnx2x/bnx2x_rxtx.c > @@ -140,7 +140,8 @@ bnx2x_dev_rx_queue_setup(struct rte_eth_dev *dev, > return -ENOMEM; > } > rxq->sw_ring[idx] = mbuf; > - rxq->rx_ring[idx] = mbuf->buf_iova; > + rxq->rx_ring[idx] = > + rte_cpu_to_le_64(rte_mbuf_data_iova_default(mbuf)); > } > rxq->pkt_first_seg = NULL; > rxq->pkt_last_seg = NULL; > @@ -400,7 +401,8 @@ bnx2x_recv_pkts(void *p_rxq, struct rte_mbuf **rx_pkts, > uint16_t nb_pkts) > > rx_mb = rxq->sw_ring[bd_cons]; > rxq->sw_ring[bd_cons] = new_mb; > - rxq->rx_ring[bd_prod] = new_mb->buf_iova; > + rxq->rx_ring[bd_prod] = > + rte_cpu_to_le_64(rte_mbuf_data_iova_default(new_mb)); > > rx_pref = NEXT_RX_BD(bd_cons) & MAX_RX_BD(rxq); > rte_prefetch0(rxq->sw_ring[rx_pref]); > @@ -409,7 +411,7 @@ bnx2x_recv_pkts(void *p_rxq, struct rte_mbuf **rx_pkts, > uint16_t nb_pkts) > rte_prefetch0(&rxq->sw_ring[rx_pref]); > } > > - rx_mb->data_off = pad; > + rx_mb->data_off = pad + RTE_PKTMBUF_HEADROOM; > rx_mb->nb_segs = 1; > rx_mb->next = NULL; > rx_mb->pkt_len = rx_mb->data_len = len; >
