DPDK patches and discussions
 help / color / mirror / Atom feed
* [dpdk-dev] [PATCH] net/bnx2x: add Rx desc MTU segment limitation
@ 2020-05-05  3:08 Rasesh Mody
  2020-05-10  8:59 ` Jerin Jacob
  0 siblings, 1 reply; 2+ messages in thread
From: Rasesh Mody @ 2020-05-05  3:08 UTC (permalink / raw)
  To: dev, jerinj, ferruh.yigit; +Cc: Rasesh Mody, GR-Everest-DPDK-Dev, Igor Russkikh

Add Rx descriptor limit for number of segments per MTU.
PMD doesn't support Jumbo Rx scatter gather hence set 1 segment per
MTU. Some applciations can adjust mbuf_size based on this value.
For others PMD detects the condition where Rx packet length cannot
be held by configured mbuf size and logs the message.

Signed-off-by: Rasesh Mody <rmody@marvell.com>
Signed-off-by: Igor Russkikh <irusskikh@marvell.com>
---
 doc/guides/nics/bnx2x.rst        | 17 +++++++++++++++++
 drivers/net/bnx2x/bnx2x_ethdev.c |  1 +
 drivers/net/bnx2x/bnx2x_rxtx.c   | 20 ++++++++++++++++++--
 3 files changed, 36 insertions(+), 2 deletions(-)

diff --git a/doc/guides/nics/bnx2x.rst b/doc/guides/nics/bnx2x.rst
index 67d765af8..ab90d8ae5 100644
--- a/doc/guides/nics/bnx2x.rst
+++ b/doc/guides/nics/bnx2x.rst
@@ -108,6 +108,23 @@ Driver compilation and testing
 Refer to the document :ref:`compiling and testing a PMD for a NIC <pmd_build_and_test>`
 for details.
 
+Jumbo: Limitation
+-----------------
+
+Rx descriptor limit for number of segments per MTU is set to 1.
+PMD doesn't support Jumbo Rx scatter gather. Some applciations can
+adjust mbuf_size based on this param and max_pkt_len.
+
+For others, PMD detects the condition where Rx packet length cannot
+be held by configured mbuf size and logs the message.
+
+Example output:
+
+   .. code-block:: console
+
+      [...]
+      [bnx2x_recv_pkts:397(04:00.0:dpdk-port-0)] mbuf size 2048 is not enough to hold Rx packet length more than 2046
+
 SR-IOV: Prerequisites and sample Application Notes
 --------------------------------------------------
 
diff --git a/drivers/net/bnx2x/bnx2x_ethdev.c b/drivers/net/bnx2x/bnx2x_ethdev.c
index 30588b152..adc3690fc 100644
--- a/drivers/net/bnx2x/bnx2x_ethdev.c
+++ b/drivers/net/bnx2x/bnx2x_ethdev.c
@@ -533,6 +533,7 @@ bnx2x_dev_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
 
 	dev_info->rx_desc_lim.nb_max = MAX_RX_AVAIL;
 	dev_info->rx_desc_lim.nb_min = MIN_RX_SIZE_NONTPA;
+	dev_info->rx_desc_lim.nb_mtu_seg_max = 1;
 	dev_info->tx_desc_lim.nb_max = MAX_TX_AVAIL;
 
 	return 0;
diff --git a/drivers/net/bnx2x/bnx2x_rxtx.c b/drivers/net/bnx2x/bnx2x_rxtx.c
index e201b68db..57e2ce504 100644
--- a/drivers/net/bnx2x/bnx2x_rxtx.c
+++ b/drivers/net/bnx2x/bnx2x_rxtx.c
@@ -343,8 +343,9 @@ bnx2x_recv_pkts(void *p_rxq, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
 	struct rte_mbuf *new_mb;
 	uint16_t rx_pref;
 	struct eth_fast_path_rx_cqe *cqe_fp;
-	uint16_t len, pad;
+	uint16_t len, pad, bd_len, buf_len;
 	struct rte_mbuf *rx_mb = NULL;
+	static bool log_once = true;
 
 	rte_spinlock_lock(&(fp)->rx_mtx);
 
@@ -384,6 +385,20 @@ bnx2x_recv_pkts(void *p_rxq, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
 
 		len = cqe_fp->pkt_len_or_gro_seg_len;
 		pad = cqe_fp->placement_offset;
+		bd_len = cqe_fp->len_on_bd;
+		buf_len = rxq->sw_ring[bd_cons]->buf_len;
+
+		/* Check for sufficient buffer length */
+		if (unlikely(buf_len < len + (pad + RTE_PKTMBUF_HEADROOM))) {
+			if (unlikely(log_once)) {
+				PMD_DRV_LOG(ERR, sc, "mbuf size %d is not enough to hold Rx packet length more than %d",
+					    buf_len - RTE_PKTMBUF_HEADROOM,
+					    buf_len -
+					    (pad + RTE_PKTMBUF_HEADROOM));
+				log_once = false;
+			}
+			goto next_rx;
+		}
 
 		new_mb = rte_mbuf_raw_alloc(rxq->mb_pool);
 		if (unlikely(!new_mb)) {
@@ -408,7 +423,8 @@ bnx2x_recv_pkts(void *p_rxq, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
 		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;
+		rx_mb->pkt_len = len;
+		rx_mb->data_len = bd_len;
 		rx_mb->port = rxq->port_id;
 		rte_prefetch1(rte_pktmbuf_mtod(rx_mb, void *));
 
-- 
2.18.0


^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: [dpdk-dev] [PATCH] net/bnx2x: add Rx desc MTU segment limitation
  2020-05-05  3:08 [dpdk-dev] [PATCH] net/bnx2x: add Rx desc MTU segment limitation Rasesh Mody
@ 2020-05-10  8:59 ` Jerin Jacob
  0 siblings, 0 replies; 2+ messages in thread
From: Jerin Jacob @ 2020-05-10  8:59 UTC (permalink / raw)
  To: Rasesh Mody
  Cc: dpdk-dev, Jerin Jacob, Ferruh Yigit, GR-Everest-DPDK-Dev, Igor Russkikh

On Tue, May 5, 2020 at 8:38 AM Rasesh Mody <rmody@marvell.com> wrote:
>
> Add Rx descriptor limit for number of segments per MTU.
> PMD doesn't support Jumbo Rx scatter gather hence set 1 segment per
> MTU. Some applciations can adjust mbuf_size based on this value.
> For others PMD detects the condition where Rx packet length cannot
> be held by configured mbuf size and logs the message.
>
> Signed-off-by: Rasesh Mody <rmody@marvell.com>
> Signed-off-by: Igor Russkikh <irusskikh@marvell.com>

Applied to dpdk-next-net-mrvl/master. Thanks



> ---
>  doc/guides/nics/bnx2x.rst        | 17 +++++++++++++++++
>  drivers/net/bnx2x/bnx2x_ethdev.c |  1 +
>  drivers/net/bnx2x/bnx2x_rxtx.c   | 20 ++++++++++++++++++--
>  3 files changed, 36 insertions(+), 2 deletions(-)
>
> diff --git a/doc/guides/nics/bnx2x.rst b/doc/guides/nics/bnx2x.rst
> index 67d765af8..ab90d8ae5 100644
> --- a/doc/guides/nics/bnx2x.rst
> +++ b/doc/guides/nics/bnx2x.rst
> @@ -108,6 +108,23 @@ Driver compilation and testing
>  Refer to the document :ref:`compiling and testing a PMD for a NIC <pmd_build_and_test>`
>  for details.
>
> +Jumbo: Limitation
> +-----------------
> +
> +Rx descriptor limit for number of segments per MTU is set to 1.
> +PMD doesn't support Jumbo Rx scatter gather. Some applciations can
> +adjust mbuf_size based on this param and max_pkt_len.
> +
> +For others, PMD detects the condition where Rx packet length cannot
> +be held by configured mbuf size and logs the message.
> +
> +Example output:
> +
> +   .. code-block:: console
> +
> +      [...]
> +      [bnx2x_recv_pkts:397(04:00.0:dpdk-port-0)] mbuf size 2048 is not enough to hold Rx packet length more than 2046
> +
>  SR-IOV: Prerequisites and sample Application Notes
>  --------------------------------------------------
>
> diff --git a/drivers/net/bnx2x/bnx2x_ethdev.c b/drivers/net/bnx2x/bnx2x_ethdev.c
> index 30588b152..adc3690fc 100644
> --- a/drivers/net/bnx2x/bnx2x_ethdev.c
> +++ b/drivers/net/bnx2x/bnx2x_ethdev.c
> @@ -533,6 +533,7 @@ bnx2x_dev_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
>
>         dev_info->rx_desc_lim.nb_max = MAX_RX_AVAIL;
>         dev_info->rx_desc_lim.nb_min = MIN_RX_SIZE_NONTPA;
> +       dev_info->rx_desc_lim.nb_mtu_seg_max = 1;
>         dev_info->tx_desc_lim.nb_max = MAX_TX_AVAIL;
>
>         return 0;
> diff --git a/drivers/net/bnx2x/bnx2x_rxtx.c b/drivers/net/bnx2x/bnx2x_rxtx.c
> index e201b68db..57e2ce504 100644
> --- a/drivers/net/bnx2x/bnx2x_rxtx.c
> +++ b/drivers/net/bnx2x/bnx2x_rxtx.c
> @@ -343,8 +343,9 @@ bnx2x_recv_pkts(void *p_rxq, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
>         struct rte_mbuf *new_mb;
>         uint16_t rx_pref;
>         struct eth_fast_path_rx_cqe *cqe_fp;
> -       uint16_t len, pad;
> +       uint16_t len, pad, bd_len, buf_len;
>         struct rte_mbuf *rx_mb = NULL;
> +       static bool log_once = true;
>
>         rte_spinlock_lock(&(fp)->rx_mtx);
>
> @@ -384,6 +385,20 @@ bnx2x_recv_pkts(void *p_rxq, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
>
>                 len = cqe_fp->pkt_len_or_gro_seg_len;
>                 pad = cqe_fp->placement_offset;
> +               bd_len = cqe_fp->len_on_bd;
> +               buf_len = rxq->sw_ring[bd_cons]->buf_len;
> +
> +               /* Check for sufficient buffer length */
> +               if (unlikely(buf_len < len + (pad + RTE_PKTMBUF_HEADROOM))) {
> +                       if (unlikely(log_once)) {
> +                               PMD_DRV_LOG(ERR, sc, "mbuf size %d is not enough to hold Rx packet length more than %d",
> +                                           buf_len - RTE_PKTMBUF_HEADROOM,
> +                                           buf_len -
> +                                           (pad + RTE_PKTMBUF_HEADROOM));
> +                               log_once = false;
> +                       }
> +                       goto next_rx;
> +               }
>
>                 new_mb = rte_mbuf_raw_alloc(rxq->mb_pool);
>                 if (unlikely(!new_mb)) {
> @@ -408,7 +423,8 @@ bnx2x_recv_pkts(void *p_rxq, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
>                 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;
> +               rx_mb->pkt_len = len;
> +               rx_mb->data_len = bd_len;
>                 rx_mb->port = rxq->port_id;
>                 rte_prefetch1(rte_pktmbuf_mtod(rx_mb, void *));
>
> --
> 2.18.0
>

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2020-05-10  9:00 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-05-05  3:08 [dpdk-dev] [PATCH] net/bnx2x: add Rx desc MTU segment limitation Rasesh Mody
2020-05-10  8:59 ` Jerin Jacob

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).