Re: [PATCH] net/bnx2x: reserve enough headroom for mbuf prepend

From: Ferruh Yigit <ferruh.yigit@intel.com>
To: harish.patil@cavium.com, rasesh.mody@cavium.com
Cc: zhouyangchao <zhouyates@gmail.com>, dev@dpdk.org
Subject: Re: [PATCH] net/bnx2x: reserve enough headroom for mbuf prepend
Date: Fri, 20 Apr 2018 11:31:36 +0100	[thread overview]
Message-ID: <c4cb9a60-5233-1364-2ff3-d21265b66542@intel.com> (raw)
In-Reply-To: <CABLiTuxYAtevgtmusOZN7c8+o-rTt9bZ1B_MVx9pxCLLEKrbQw@mail.gmail.com>

On 3/8/2018 5:57 AM, zhouyangchao wrote:
> When allocating a new mbuf for Rx, the value of m->data_off should be 
> reset to its default value (RTE_PKTMBUF_HEADROOM), instead of reusing 
> the previous undefined value, which could cause the packet to have a too 
> small or too high headroom.

Hi Harish, Rasesh,

Reminder of this patch waiting for your review?

> 
> On Mon, Mar 5, 2018 at 11:28 PM Ferruh Yigit <ferruh.yigit@intel.com
> <mailto:ferruh.yigit@intel.com>> wrote:
> 
>     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 <zhouyates@gmail.com
>     <mailto:zhouyates@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;
>     >
> 