[dpdk-stable] [PATCH v1 01/38] net/mvpp2: fix stack corruption

[dpdk-stable] [PATCH v1 01/38] net/mvpp2: fix stack corruption

[lironh]

From: Yuri Chipchev <yuric@marvell.com>

Fixes stack corruption in mrvl_fill_bpool function
in case num > MRVL_PP2_RXD_MAX

Fixes: c3637258d894 ("net/mrvl: fix Rx descriptors number")
Cc: stable@dpdk.org

Signed-off-by: Yuri Chipchev <yuric@marvell.com>
Reviewed-by: Liron Himi <lironh@marvell.com>
---
 drivers/net/mvpp2/mrvl_ethdev.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/net/mvpp2/mrvl_ethdev.c b/drivers/net/mvpp2/mrvl_ethdev.c
index f25cf9e46..93fb30cdb 100644
--- a/drivers/net/mvpp2/mrvl_ethdev.c
+++ b/drivers/net/mvpp2/mrvl_ethdev.c
@@ -1614,8 +1614,8 @@ mrvl_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
 static int
 mrvl_fill_bpool(struct mrvl_rxq *rxq, int num)
 {
-	struct buff_release_entry entries[MRVL_PP2_RXD_MAX];
-	struct rte_mbuf *mbufs[MRVL_PP2_RXD_MAX];
+	struct buff_release_entry entries[num];
+	struct rte_mbuf *mbufs[num];
 	int i, ret;
 	unsigned int core_id;
 	struct pp2_hif *hif;
-- 
2.28.0

[dpdk-stable] [PATCH v1 02/38] net/mvpp2: remove debug log on fast-path

[lironh]

From: Liron Himi <lironh@marvell.com>

in case of non-ip frame the current code reached the 'default'
case which result with function call to log a msg.
those kind of calls should not be performed on fast-path.

The performance for this kind of frames increased by 50%

Fixes: acab7d58c ("net/mvpp2: convert to dynamic logging")
Cc: stable@dpdk.org

Signed-off-by: Liron Himi <lironh@marvell.com>
Reviewed-by: Liron Himi <lironh@marvell.com>
---
 drivers/net/mvpp2/mrvl_ethdev.c | 2 --
 1 file changed, 2 deletions(-)

diff --git a/drivers/net/mvpp2/mrvl_ethdev.c b/drivers/net/mvpp2/mrvl_ethdev.c
index 93fb30cdb..8c3278ec4 100644
--- a/drivers/net/mvpp2/mrvl_ethdev.c
+++ b/drivers/net/mvpp2/mrvl_ethdev.c
@@ -2171,7 +2171,6 @@ mrvl_desc_to_packet_type_and_offset(struct pp2_ppio_desc *desc,
 		*l4_offset = *l3_offset + MRVL_ARP_LENGTH;
 		break;
 	default:
-		MRVL_LOG(DEBUG, "Failed to recognise l3 packet type");
 		break;
 	}
 
@@ -2183,7 +2182,6 @@ mrvl_desc_to_packet_type_and_offset(struct pp2_ppio_desc *desc,
 		packet_type |= RTE_PTYPE_L4_UDP;
 		break;
 	default:
-		MRVL_LOG(DEBUG, "Failed to recognise l4 packet type");
 		break;
 	}
 
-- 
2.28.0

[dpdk-stable] [PATCH v1 03/38] net/mvpp2: fix rx/tx bytes statistics

[lironh]

From: Yuri Chipchev <yuric@marvell.com>

4B of CRC was not included in the bytes statistics

Fixes: bdffe0c70 ("net/mrvl: support basic stats")
Cc: stable@dpdk.org

Signed-off-by: Yuri Chipchev <yuric@marvell.com>
Reviewed-by: Liron Himi <lironh@marvell.com>
---
 drivers/net/mvpp2/mrvl_ethdev.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/net/mvpp2/mrvl_ethdev.c b/drivers/net/mvpp2/mrvl_ethdev.c
index 8c3278ec4..ebd2bb1c8 100644
--- a/drivers/net/mvpp2/mrvl_ethdev.c
+++ b/drivers/net/mvpp2/mrvl_ethdev.c
@@ -1278,7 +1278,6 @@ mrvl_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
 				       rx_stats.drop_fullq +
 				       rx_stats.drop_bm +
 				       rxq->drop_mac;
-		stats->ibytes += rxq->bytes_recv;
 		drop_mac += rxq->drop_mac;
 	}
 
@@ -1306,7 +1305,6 @@ mrvl_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
 
 		stats->q_opackets[idx] = tx_stats.deq_desc;
 		stats->q_obytes[idx] = txq->bytes_sent;
-		stats->obytes += txq->bytes_sent;
 	}
 
 	ret = pp2_ppio_get_statistics(priv->ppio, &ppio_stats, 0);
@@ -1315,6 +1313,8 @@ mrvl_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
 		return ret;
 	}
 
+	stats->ibytes += ppio_stats.rx_bytes;
+	stats->obytes += ppio_stats.tx_bytes;
 	stats->ipackets += ppio_stats.rx_packets - drop_mac;
 	stats->opackets += ppio_stats.tx_packets;
 	stats->imissed += ppio_stats.rx_fullq_dropped +
-- 
2.28.0

[dpdk-stable] [PATCH v1 04/38] net/mvpp2: skip vlan flush

[lironh]

From: Liron Himi <lironh@marvell.com>

vlan-flush in MUSDK is not supported yet.
until it does, the code should be skipped as currently
an redundant  error message is displayed.

Fixes: a8f3d6783 ("net/mrvl: support VLAN filtering")
Cc: stable@dpdk.org

Signed-off-by: Liron Himi <lironh@marvell.com>
Reviewed-by: Liron Himi <lironh@marvell.com>
---
 drivers/net/mvpp2/mrvl_ethdev.c | 18 +++++++++---------
 1 file changed, 9 insertions(+), 9 deletions(-)

diff --git a/drivers/net/mvpp2/mrvl_ethdev.c b/drivers/net/mvpp2/mrvl_ethdev.c
index ebd2bb1c8..0985ccebe 100644
--- a/drivers/net/mvpp2/mrvl_ethdev.c
+++ b/drivers/net/mvpp2/mrvl_ethdev.c
@@ -672,15 +672,15 @@ mrvl_dev_start(struct rte_eth_dev *dev)
 	}
 
 	if (!priv->vlan_flushed) {
-		ret = pp2_ppio_flush_vlan(priv->ppio);
-		if (ret) {
-			MRVL_LOG(ERR, "Failed to flush vlan list");
-			/*
-			 * TODO
-			 * once pp2_ppio_flush_vlan() is supported jump to out
-			 * goto out;
-			 */
-		}
+		/*
+		 * TODO
+		 * once pp2_ppio_flush_vlan() is supported call it
+		 * ret = pp2_ppio_flush_vlan(priv->ppio);
+		 * if (ret) {
+		 *	MRVL_LOG(ERR, "Failed to flush vlan list");
+		 *	goto out;
+		 * }
+		 */
 		priv->vlan_flushed = 1;
 	}
 	ret = mrvl_mtu_set(dev, dev->data->mtu);
-- 
2.28.0

[dpdk-stable] [PATCH v1 05/38] net/mvpp2: remove CRC len from MRU validation

[lironh]

From: Liron Himi <lironh@marvell.com>

CRC is being removed by HW before packet get
write to the buffer, so CRC len should not be
included in MRU validation

Fixes: 79ec62028 ("net/mvpp2: update MTU and MRU related calculations")
Cc: stable@dpdk.org

Signed-off-by: Liron Himi <lironh@marvell.com>
Reviewed-by: Yuri Chipchev <yuric@marvell.com>
---
 drivers/net/mvpp2/mrvl_ethdev.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/net/mvpp2/mrvl_ethdev.c b/drivers/net/mvpp2/mrvl_ethdev.c
index 0985ccebe..dbb193496 100644
--- a/drivers/net/mvpp2/mrvl_ethdev.c
+++ b/drivers/net/mvpp2/mrvl_ethdev.c
@@ -441,8 +441,8 @@ mrvl_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
 	 * when this feature has not been enabled/supported so far
 	 * (TODO check scattered_rx flag here once scattered RX is supported).
 	 */
-	if (mru + MRVL_PKT_OFFS > mbuf_data_size) {
-		mru = mbuf_data_size - MRVL_PKT_OFFS;
+	if (mru - RTE_ETHER_CRC_LEN + MRVL_PKT_OFFS > mbuf_data_size) {
+		mru = mbuf_data_size + RTE_ETHER_CRC_LEN - MRVL_PKT_OFFS;
 		mtu = MRVL_PP2_MRU_TO_MTU(mru);
 		MRVL_LOG(WARNING, "MTU too big, max MTU possible limitted "
 			"by current mbuf size: %u. Set MTU to %u, MRU to %u",
-- 
2.28.0

[dpdk-stable] [PATCH v1 06/38] net/mvpp2: fix frame size checking

[lironh]

From: Liron Himi <lironh@marvell.com>

need to add CRC len to the frame-size to compare against
max_rx_pkt_len which includes it.

Fixes: 79ec62028 ("net/mvpp2: update MTU and MRU related calculations")
Cc: stable@dpdk.org

Signed-off-by: Liron Himi <lironh@marvell.com>
Reviewed-by: Yuri Chipchev <yuric@marvell.com>
---
 drivers/net/mvpp2/mrvl_ethdev.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/net/mvpp2/mrvl_ethdev.c b/drivers/net/mvpp2/mrvl_ethdev.c
index dbb193496..3f05ebe00 100644
--- a/drivers/net/mvpp2/mrvl_ethdev.c
+++ b/drivers/net/mvpp2/mrvl_ethdev.c
@@ -1711,7 +1711,8 @@ mrvl_rx_queue_setup(struct rte_eth_dev *dev, uint16_t idx, uint16_t desc,
 		return -EFAULT;
 	}
 
-	frame_size = buf_size - RTE_PKTMBUF_HEADROOM - MRVL_PKT_EFFEC_OFFS;
+	frame_size = buf_size - RTE_PKTMBUF_HEADROOM -
+		     MRVL_PKT_EFFEC_OFFS + RTE_ETHER_CRC_LEN;
 	if (frame_size < max_rx_pkt_len) {
 		MRVL_LOG(WARNING,
 			"Mbuf size must be increased to %u bytes to hold up "
-- 
2.28.0

[dpdk-stable] [PATCH v1 07/38] net/mvpp2: reduce prints on rx path

[lironh]

From: Liron Himi <lironh@marvell.com>

Fixes: acab7d58c ("net/mvpp2: convert to dynamic logging")
Cc: stable@dpdk.org

Signed-off-by: Liron Himi <lironh@marvell.com>
Reviewed-by: Dana Tearosh <danat@marvell.com>
---
 drivers/net/mvpp2/mrvl_ethdev.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/mvpp2/mrvl_ethdev.c b/drivers/net/mvpp2/mrvl_ethdev.c
index 3f05ebe00..d81b86c02 100644
--- a/drivers/net/mvpp2/mrvl_ethdev.c
+++ b/drivers/net/mvpp2/mrvl_ethdev.c
@@ -2320,7 +2320,7 @@ mrvl_rx_pkt_burst(void *rxq, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
 			     (!rx_done && num < q->priv->bpool_init_size))) {
 			ret = mrvl_fill_bpool(q, MRVL_BURST_SIZE);
 			if (ret)
-				MRVL_LOG(ERR, "Failed to fill bpool");
+				MRVL_LOG(DEBUG, "Failed to fill bpool");
 		} else if (unlikely(num > q->priv->bpool_max_size)) {
 			int i;
 			int pkt_to_remove = num - q->priv->bpool_init_size;
-- 
2.28.0

Re: [dpdk-stable] [dpdk-dev] [PATCH v1 05/38] net/mvpp2: remove CRC len from MRU validation

[Michael Shamis]

Reviewed-by: Michael Shamis <michaelsh@marvell.com>

-----Original Message-----
From: dev <dev-bounces@dpdk.org> On Behalf Of lironh@marvell.com
Sent: Wednesday, December 2, 2020 12:12 PM
To: Jerin Jacob Kollanukkaran <jerinj@marvell.com>
Cc: dev@dpdk.org; Liron Himi <lironh@marvell.com>; stable@dpdk.org; Yuri Chipchev <yuric@marvell.com>
Subject: [dpdk-dev] [PATCH v1 05/38] net/mvpp2: remove CRC len from MRU validation

From: Liron Himi <lironh@marvell.com>

CRC is being removed by HW before packet get write to the buffer, so CRC len should not be included in MRU validation

Fixes: 79ec62028 ("net/mvpp2: update MTU and MRU related calculations")
Cc: stable@dpdk.org

Signed-off-by: Liron Himi <lironh@marvell.com>
Reviewed-by: Yuri Chipchev <yuric@marvell.com>
---
 drivers/net/mvpp2/mrvl_ethdev.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/net/mvpp2/mrvl_ethdev.c b/drivers/net/mvpp2/mrvl_ethdev.c index 0985ccebe..dbb193496 100644
--- a/drivers/net/mvpp2/mrvl_ethdev.c
+++ b/drivers/net/mvpp2/mrvl_ethdev.c
@@ -441,8 +441,8 @@ mrvl_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
 	 * when this feature has not been enabled/supported so far
 	 * (TODO check scattered_rx flag here once scattered RX is supported).
 	 */
-	if (mru + MRVL_PKT_OFFS > mbuf_data_size) {
-		mru = mbuf_data_size - MRVL_PKT_OFFS;
+	if (mru - RTE_ETHER_CRC_LEN + MRVL_PKT_OFFS > mbuf_data_size) {
+		mru = mbuf_data_size + RTE_ETHER_CRC_LEN - MRVL_PKT_OFFS;
 		mtu = MRVL_PP2_MRU_TO_MTU(mru);
 		MRVL_LOG(WARNING, "MTU too big, max MTU possible limitted "
 			"by current mbuf size: %u. Set MTU to %u, MRU to %u",
--
2.28.0

Re: [dpdk-stable] [dpdk-dev] [PATCH v1 06/38] net/mvpp2: fix frame size checking

[Michael Shamis]

Reviewed-by: Michael Shamis <michaelsh@marvell.com>

-----Original Message-----
From: dev <dev-bounces@dpdk.org> On Behalf Of lironh@marvell.com
Sent: Wednesday, December 2, 2020 12:12 PM
To: Jerin Jacob Kollanukkaran <jerinj@marvell.com>
Cc: dev@dpdk.org; Liron Himi <lironh@marvell.com>; stable@dpdk.org; Yuri Chipchev <yuric@marvell.com>
Subject: [dpdk-dev] [PATCH v1 06/38] net/mvpp2: fix frame size checking

From: Liron Himi <lironh@marvell.com>

need to add CRC len to the frame-size to compare against max_rx_pkt_len which includes it.

Fixes: 79ec62028 ("net/mvpp2: update MTU and MRU related calculations")
Cc: stable@dpdk.org

Signed-off-by: Liron Himi <lironh@marvell.com>
Reviewed-by: Yuri Chipchev <yuric@marvell.com>
---
 drivers/net/mvpp2/mrvl_ethdev.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/net/mvpp2/mrvl_ethdev.c b/drivers/net/mvpp2/mrvl_ethdev.c index dbb193496..3f05ebe00 100644
--- a/drivers/net/mvpp2/mrvl_ethdev.c
+++ b/drivers/net/mvpp2/mrvl_ethdev.c
@@ -1711,7 +1711,8 @@ mrvl_rx_queue_setup(struct rte_eth_dev *dev, uint16_t idx, uint16_t desc,
 		return -EFAULT;
 	}
 
-	frame_size = buf_size - RTE_PKTMBUF_HEADROOM - MRVL_PKT_EFFEC_OFFS;
+	frame_size = buf_size - RTE_PKTMBUF_HEADROOM -
+		     MRVL_PKT_EFFEC_OFFS + RTE_ETHER_CRC_LEN;
 	if (frame_size < max_rx_pkt_len) {
 		MRVL_LOG(WARNING,
 			"Mbuf size must be increased to %u bytes to hold up "
--
2.28.0

Re: [dpdk-stable] [dpdk-dev] [PATCH v1 07/38] net/mvpp2: reduce prints on rx path

[Michael Shamis]

Reviewed-by: Michael Shamis <michaelsh@marvell.com>

-----Original Message-----
From: dev <dev-bounces@dpdk.org> On Behalf Of lironh@marvell.com
Sent: Wednesday, December 2, 2020 12:12 PM
To: Jerin Jacob Kollanukkaran <jerinj@marvell.com>
Cc: dev@dpdk.org; Liron Himi <lironh@marvell.com>; stable@dpdk.org; Dana Vardi <danat@marvell.com>
Subject: [dpdk-dev] [PATCH v1 07/38] net/mvpp2: reduce prints on rx path

From: Liron Himi <lironh@marvell.com>

Fixes: acab7d58c ("net/mvpp2: convert to dynamic logging")
Cc: stable@dpdk.org

Signed-off-by: Liron Himi <lironh@marvell.com>
Reviewed-by: Dana Tearosh <danat@marvell.com>
---
 drivers/net/mvpp2/mrvl_ethdev.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/mvpp2/mrvl_ethdev.c b/drivers/net/mvpp2/mrvl_ethdev.c index 3f05ebe00..d81b86c02 100644
--- a/drivers/net/mvpp2/mrvl_ethdev.c
+++ b/drivers/net/mvpp2/mrvl_ethdev.c
@@ -2320,7 +2320,7 @@ mrvl_rx_pkt_burst(void *rxq, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
 			     (!rx_done && num < q->priv->bpool_init_size))) {
 			ret = mrvl_fill_bpool(q, MRVL_BURST_SIZE);
 			if (ret)
-				MRVL_LOG(ERR, "Failed to fill bpool");
+				MRVL_LOG(DEBUG, "Failed to fill bpool");
 		} else if (unlikely(num > q->priv->bpool_max_size)) {
 			int i;
 			int pkt_to_remove = num - q->priv->bpool_init_size;
--
2.28.0

Re: [dpdk-stable] [dpdk-dev] [PATCH v1 04/38] net/mvpp2: skip vlan flush

[Michael Shamis]

Reviewed-by: Michael Shamis <michaelsh@marvell.com>

-----Original Message-----
From: dev <dev-bounces@dpdk.org> On Behalf Of lironh@marvell.com
Sent: Wednesday, December 2, 2020 12:12 PM
To: Jerin Jacob Kollanukkaran <jerinj@marvell.com>
Cc: dev@dpdk.org; Liron Himi <lironh@marvell.com>; stable@dpdk.org
Subject: [dpdk-dev] [PATCH v1 04/38] net/mvpp2: skip vlan flush

From: Liron Himi <lironh@marvell.com>

vlan-flush in MUSDK is not supported yet.
until it does, the code should be skipped as currently an redundant  error message is displayed.

Fixes: a8f3d6783 ("net/mrvl: support VLAN filtering")
Cc: stable@dpdk.org

Signed-off-by: Liron Himi <lironh@marvell.com>
Reviewed-by: Liron Himi <lironh@marvell.com>
---
 drivers/net/mvpp2/mrvl_ethdev.c | 18 +++++++++---------
 1 file changed, 9 insertions(+), 9 deletions(-)

diff --git a/drivers/net/mvpp2/mrvl_ethdev.c b/drivers/net/mvpp2/mrvl_ethdev.c index ebd2bb1c8..0985ccebe 100644
--- a/drivers/net/mvpp2/mrvl_ethdev.c
+++ b/drivers/net/mvpp2/mrvl_ethdev.c
@@ -672,15 +672,15 @@ mrvl_dev_start(struct rte_eth_dev *dev)
 	}
 
 	if (!priv->vlan_flushed) {
-		ret = pp2_ppio_flush_vlan(priv->ppio);
-		if (ret) {
-			MRVL_LOG(ERR, "Failed to flush vlan list");
-			/*
-			 * TODO
-			 * once pp2_ppio_flush_vlan() is supported jump to out
-			 * goto out;
-			 */
-		}
+		/*
+		 * TODO
+		 * once pp2_ppio_flush_vlan() is supported call it
+		 * ret = pp2_ppio_flush_vlan(priv->ppio);
+		 * if (ret) {
+		 *	MRVL_LOG(ERR, "Failed to flush vlan list");
+		 *	goto out;
+		 * }
+		 */
 		priv->vlan_flushed = 1;
 	}
 	ret = mrvl_mtu_set(dev, dev->data->mtu);
--
2.28.0

Re: [dpdk-stable] [dpdk-dev] [PATCH v1 03/38] net/mvpp2: fix rx/tx bytes statistics

[Michael Shamis]

Reviewed-by: Michael Shamis <michaelsh@marvell.com>

-----Original Message-----
From: dev <dev-bounces@dpdk.org> On Behalf Of lironh@marvell.com
Sent: Wednesday, December 2, 2020 12:12 PM
To: Jerin Jacob Kollanukkaran <jerinj@marvell.com>
Cc: dev@dpdk.org; Yuri Chipchev <yuric@marvell.com>; stable@dpdk.org; Liron Himi <lironh@marvell.com>
Subject: [dpdk-dev] [PATCH v1 03/38] net/mvpp2: fix rx/tx bytes statistics

From: Yuri Chipchev <yuric@marvell.com>

4B of CRC was not included in the bytes statistics

Fixes: bdffe0c70 ("net/mrvl: support basic stats")
Cc: stable@dpdk.org

Signed-off-by: Yuri Chipchev <yuric@marvell.com>
Reviewed-by: Liron Himi <lironh@marvell.com>
---
 drivers/net/mvpp2/mrvl_ethdev.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/net/mvpp2/mrvl_ethdev.c b/drivers/net/mvpp2/mrvl_ethdev.c index 8c3278ec4..ebd2bb1c8 100644
--- a/drivers/net/mvpp2/mrvl_ethdev.c
+++ b/drivers/net/mvpp2/mrvl_ethdev.c
@@ -1278,7 +1278,6 @@ mrvl_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
 				       rx_stats.drop_fullq +
 				       rx_stats.drop_bm +
 				       rxq->drop_mac;
-		stats->ibytes += rxq->bytes_recv;
 		drop_mac += rxq->drop_mac;
 	}
 
@@ -1306,7 +1305,6 @@ mrvl_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
 
 		stats->q_opackets[idx] = tx_stats.deq_desc;
 		stats->q_obytes[idx] = txq->bytes_sent;
-		stats->obytes += txq->bytes_sent;
 	}
 
 	ret = pp2_ppio_get_statistics(priv->ppio, &ppio_stats, 0); @@ -1315,6 +1313,8 @@ mrvl_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
 		return ret;
 	}
 
+	stats->ibytes += ppio_stats.rx_bytes;
+	stats->obytes += ppio_stats.tx_bytes;
 	stats->ipackets += ppio_stats.rx_packets - drop_mac;
 	stats->opackets += ppio_stats.tx_packets;
 	stats->imissed += ppio_stats.rx_fullq_dropped +
--
2.28.0

Re: [dpdk-stable] [dpdk-dev] [PATCH v1 01/38] net/mvpp2: fix stack corruption

[Michael Shamis]

Reviewed-by: Michael Shamis <michaelsh@marvell.com>

-----Original Message-----
From: dev <dev-bounces@dpdk.org> On Behalf Of lironh@marvell.com
Sent: Wednesday, December 2, 2020 12:12 PM
To: Jerin Jacob Kollanukkaran <jerinj@marvell.com>
Cc: dev@dpdk.org; Yuri Chipchev <yuric@marvell.com>; stable@dpdk.org; Liron Himi <lironh@marvell.com>
Subject: [dpdk-dev] [PATCH v1 01/38] net/mvpp2: fix stack corruption

From: Yuri Chipchev <yuric@marvell.com>

Fixes stack corruption in mrvl_fill_bpool function in case num > MRVL_PP2_RXD_MAX

Fixes: c3637258d894 ("net/mrvl: fix Rx descriptors number")
Cc: stable@dpdk.org

Signed-off-by: Yuri Chipchev <yuric@marvell.com>
Reviewed-by: Liron Himi <lironh@marvell.com>
---
 drivers/net/mvpp2/mrvl_ethdev.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/net/mvpp2/mrvl_ethdev.c b/drivers/net/mvpp2/mrvl_ethdev.c index f25cf9e46..93fb30cdb 100644
--- a/drivers/net/mvpp2/mrvl_ethdev.c
+++ b/drivers/net/mvpp2/mrvl_ethdev.c
@@ -1614,8 +1614,8 @@ mrvl_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)  static int  mrvl_fill_bpool(struct mrvl_rxq *rxq, int num)  {
-	struct buff_release_entry entries[MRVL_PP2_RXD_MAX];
-	struct rte_mbuf *mbufs[MRVL_PP2_RXD_MAX];
+	struct buff_release_entry entries[num];
+	struct rte_mbuf *mbufs[num];
 	int i, ret;
 	unsigned int core_id;
 	struct pp2_hif *hif;
--
2.28.0

Re: [dpdk-stable] [dpdk-dev] [PATCH v1 02/38] net/mvpp2: remove debug log on fast-path

[Michael Shamis]

Reviewed-by: Michael Shamis <michaelsh@marvell.com>

-----Original Message-----
From: dev <dev-bounces@dpdk.org> On Behalf Of lironh@marvell.com
Sent: Wednesday, December 2, 2020 12:12 PM
To: Jerin Jacob Kollanukkaran <jerinj@marvell.com>
Cc: dev@dpdk.org; Liron Himi <lironh@marvell.com>; stable@dpdk.org
Subject: [dpdk-dev] [PATCH v1 02/38] net/mvpp2: remove debug log on fast-path

From: Liron Himi <lironh@marvell.com>

in case of non-ip frame the current code reached the 'default'
case which result with function call to log a msg.
those kind of calls should not be performed on fast-path.

The performance for this kind of frames increased by 50%

Fixes: acab7d58c ("net/mvpp2: convert to dynamic logging")
Cc: stable@dpdk.org

Signed-off-by: Liron Himi <lironh@marvell.com>
Reviewed-by: Liron Himi <lironh@marvell.com>
---
 drivers/net/mvpp2/mrvl_ethdev.c | 2 --
 1 file changed, 2 deletions(-)

diff --git a/drivers/net/mvpp2/mrvl_ethdev.c b/drivers/net/mvpp2/mrvl_ethdev.c index 93fb30cdb..8c3278ec4 100644
--- a/drivers/net/mvpp2/mrvl_ethdev.c
+++ b/drivers/net/mvpp2/mrvl_ethdev.c
@@ -2171,7 +2171,6 @@ mrvl_desc_to_packet_type_and_offset(struct pp2_ppio_desc *desc,
 		*l4_offset = *l3_offset + MRVL_ARP_LENGTH;
 		break;
 	default:
-		MRVL_LOG(DEBUG, "Failed to recognise l3 packet type");
 		break;
 	}
 
@@ -2183,7 +2182,6 @@ mrvl_desc_to_packet_type_and_offset(struct pp2_ppio_desc *desc,
 		packet_type |= RTE_PTYPE_L4_UDP;
 		break;
 	default:
-		MRVL_LOG(DEBUG, "Failed to recognise l4 packet type");
 		break;
 	}
 
--
2.28.0

Re: [dpdk-stable] [dpdk-dev] [PATCH v1 02/38] net/mvpp2: remove debug log on fast-path

[Jerin Jacob]

On Wed, Dec 2, 2020 at 3:42 PM <lironh@marvell.com> wrote:
>
> From: Liron Himi <lironh@marvell.com>
>
> in case of non-ip frame the current code reached the 'default'

In case of

> case which result with function call to log a msg.
> those kind of calls should not be performed on fast-path.
>
> The performance for this kind of frames increased by 50%
>
> Fixes: acab7d58c ("net/mvpp2: convert to dynamic logging")
> Cc: stable@dpdk.org
>
> Signed-off-by: Liron Himi <lironh@marvell.com>
> Reviewed-by: Liron Himi <lironh@marvell.com>

Please remove the Reviewed-by if you have Signed-off,

> ---
>  drivers/net/mvpp2/mrvl_ethdev.c | 2 --
>  1 file changed, 2 deletions(-)
>
> diff --git a/drivers/net/mvpp2/mrvl_ethdev.c b/drivers/net/mvpp2/mrvl_ethdev.c
> index 93fb30cdb..8c3278ec4 100644
> --- a/drivers/net/mvpp2/mrvl_ethdev.c
> +++ b/drivers/net/mvpp2/mrvl_ethdev.c
> @@ -2171,7 +2171,6 @@ mrvl_desc_to_packet_type_and_offset(struct pp2_ppio_desc *desc,
>                 *l4_offset = *l3_offset + MRVL_ARP_LENGTH;
>                 break;
>         default:
> -               MRVL_LOG(DEBUG, "Failed to recognise l3 packet type");
>                 break;
>         }
>
> @@ -2183,7 +2182,6 @@ mrvl_desc_to_packet_type_and_offset(struct pp2_ppio_desc *desc,
>                 packet_type |= RTE_PTYPE_L4_UDP;
>                 break;
>         default:
> -               MRVL_LOG(DEBUG, "Failed to recognise l4 packet type");

There are other prints in fastpath such as
"MRVL_LOG(ERR, "Failed to receive packets");", please remove those as well.


>                 break;
>         }
>
> --
> 2.28.0
>

Re: [dpdk-stable] [dpdk-dev] [PATCH v1 04/38] net/mvpp2: skip vlan flush

[Jerin Jacob]

On Wed, Dec 23, 2020 at 3:15 PM Michael Shamis <michaelsh@marvell.com> wrote:
>
> Reviewed-by: Michael Shamis <michaelsh@marvell.com>
>
> -----Original Message-----
> From: dev <dev-bounces@dpdk.org> On Behalf Of lironh@marvell.com
> Sent: Wednesday, December 2, 2020 12:12 PM
> To: Jerin Jacob Kollanukkaran <jerinj@marvell.com>
> Cc: dev@dpdk.org; Liron Himi <lironh@marvell.com>; stable@dpdk.org
> Subject: [dpdk-dev] [PATCH v1 04/38] net/mvpp2: skip vlan flush
>
> From: Liron Himi <lironh@marvell.com>
>
> vlan-flush in MUSDK is not supported yet.
> until it does, the code should be skipped as currently an redundant  error message is displayed.
>
> Fixes: a8f3d6783 ("net/mrvl: support VLAN filtering")
> Cc: stable@dpdk.org
>
> Signed-off-by: Liron Himi <lironh@marvell.com>
> Reviewed-by: Liron Himi <lironh@marvell.com>
> ---
>  drivers/net/mvpp2/mrvl_ethdev.c | 18 +++++++++---------
>  1 file changed, 9 insertions(+), 9 deletions(-)
>
> diff --git a/drivers/net/mvpp2/mrvl_ethdev.c b/drivers/net/mvpp2/mrvl_ethdev.c index ebd2bb1c8..0985ccebe 100644
> --- a/drivers/net/mvpp2/mrvl_ethdev.c
> +++ b/drivers/net/mvpp2/mrvl_ethdev.c
> @@ -672,15 +672,15 @@ mrvl_dev_start(struct rte_eth_dev *dev)
>         }
>
>         if (!priv->vlan_flushed) {
> -               ret = pp2_ppio_flush_vlan(priv->ppio);
> -               if (ret) {
> -                       MRVL_LOG(ERR, "Failed to flush vlan list");
> -                       /*
> -                        * TODO
> -                        * once pp2_ppio_flush_vlan() is supported jump to out
> -                        * goto out;
> -                        */
> -               }
> +               /*
> +                * TODO
> +                * once pp2_ppio_flush_vlan() is supported call it
> +                * ret = pp2_ppio_flush_vlan(priv->ppio);
> +                * if (ret) {
> +                *      MRVL_LOG(ERR, "Failed to flush vlan list");
> +                *      goto out;
> +                * }
> +                */

This kind of TODO is no go in upstream code.
I think, MUSDK can be updated to skip the error and have an API detect
the feature
is present or not? so that drivers can make intelligent decisions and
have clean code.



>                 priv->vlan_flushed = 1;
>         }
>         ret = mrvl_mtu_set(dev, dev->data->mtu);
> --
> 2.28.0
>

Re: [dpdk-stable] [dpdk-dev] [PATCH v1 07/38] net/mvpp2: reduce prints on rx path

[Jerin Jacob]

On Wed, Dec 23, 2020 at 3:15 PM Michael Shamis <michaelsh@marvell.com> wrote:
>
> Reviewed-by: Michael Shamis <michaelsh@marvell.com>
>
> -----Original Message-----
> From: dev <dev-bounces@dpdk.org> On Behalf Of lironh@marvell.com
> Sent: Wednesday, December 2, 2020 12:12 PM
> To: Jerin Jacob Kollanukkaran <jerinj@marvell.com>
> Cc: dev@dpdk.org; Liron Himi <lironh@marvell.com>; stable@dpdk.org; Dana Vardi <danat@marvell.com>
> Subject: [dpdk-dev] [PATCH v1 07/38] net/mvpp2: reduce prints on rx path
>
> From: Liron Himi <lironh@marvell.com>

Please have git commit message for every patch.

>
> Fixes: acab7d58c ("net/mvpp2: convert to dynamic logging")
> Cc: stable@dpdk.org
>
> Signed-off-by: Liron Himi <lironh@marvell.com>
> Reviewed-by: Dana Tearosh <danat@marvell.com>
> ---
>  drivers/net/mvpp2/mrvl_ethdev.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/net/mvpp2/mrvl_ethdev.c b/drivers/net/mvpp2/mrvl_ethdev.c index 3f05ebe00..d81b86c02 100644
> --- a/drivers/net/mvpp2/mrvl_ethdev.c
> +++ b/drivers/net/mvpp2/mrvl_ethdev.c
> @@ -2320,7 +2320,7 @@ mrvl_rx_pkt_burst(void *rxq, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
>                              (!rx_done && num < q->priv->bpool_init_size))) {
>                         ret = mrvl_fill_bpool(q, MRVL_BURST_SIZE);
>                         if (ret)
> -                               MRVL_LOG(ERR, "Failed to fill bpool");
> +                               MRVL_LOG(DEBUG, "Failed to fill bpool");

Please squash all log-related patches to a single one.

>                 } else if (unlikely(num > q->priv->bpool_max_size)) {
>                         int i;
>                         int pkt_to_remove = num - q->priv->bpool_init_size;
> --
> 2.28.0
>

[dpdk-stable] [PATCH v2 01/37] net/mvpp2: fix stack corruption

[lironh]

From: Yuri Chipchev <yuric@marvell.com>

Fixes stack corruption in mrvl_fill_bpool function
in case num > MRVL_PP2_RXD_MAX

Fixes: c3637258d894 ("net/mrvl: fix Rx descriptors number")
Cc: stable@dpdk.org

Signed-off-by: Yuri Chipchev <yuric@marvell.com>
Reviewed-by: Liron Himi <lironh@marvell.com>
---
 drivers/net/mvpp2/mrvl_ethdev.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/net/mvpp2/mrvl_ethdev.c b/drivers/net/mvpp2/mrvl_ethdev.c
index f25cf9e46..93fb30cdb 100644
--- a/drivers/net/mvpp2/mrvl_ethdev.c
+++ b/drivers/net/mvpp2/mrvl_ethdev.c
@@ -1614,8 +1614,8 @@ mrvl_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
 static int
 mrvl_fill_bpool(struct mrvl_rxq *rxq, int num)
 {
-	struct buff_release_entry entries[MRVL_PP2_RXD_MAX];
-	struct rte_mbuf *mbufs[MRVL_PP2_RXD_MAX];
+	struct buff_release_entry entries[num];
+	struct rte_mbuf *mbufs[num];
 	int i, ret;
 	unsigned int core_id;
 	struct pp2_hif *hif;
-- 
2.28.0

[dpdk-stable] [PATCH v2 02/37] net/mvpp2: remove debug log on fast-path

[lironh]

From: Liron Himi <lironh@marvell.com>

In case of non-ip frame the current code reached the 'default'
case which result with function call to log a msg.
those kind of calls should not be performed on fast-path.

The performance for this kind of frames increased by 50%

Fixes: acab7d58c ("net/mvpp2: convert to dynamic logging")
Cc: stable@dpdk.org

Signed-off-by: Liron Himi <lironh@marvell.com>
---
 drivers/net/mvpp2/mrvl_ethdev.c | 27 ++++-----------------------
 1 file changed, 4 insertions(+), 23 deletions(-)

diff --git a/drivers/net/mvpp2/mrvl_ethdev.c b/drivers/net/mvpp2/mrvl_ethdev.c
index 93fb30cdb..bfe496ff7 100644
--- a/drivers/net/mvpp2/mrvl_ethdev.c
+++ b/drivers/net/mvpp2/mrvl_ethdev.c
@@ -2171,7 +2171,6 @@ mrvl_desc_to_packet_type_and_offset(struct pp2_ppio_desc *desc,
 		*l4_offset = *l3_offset + MRVL_ARP_LENGTH;
 		break;
 	default:
-		MRVL_LOG(DEBUG, "Failed to recognise l3 packet type");
 		break;
 	}
 
@@ -2183,7 +2182,6 @@ mrvl_desc_to_packet_type_and_offset(struct pp2_ppio_desc *desc,
 		packet_type |= RTE_PTYPE_L4_UDP;
 		break;
 	default:
-		MRVL_LOG(DEBUG, "Failed to recognise l4 packet type");
 		break;
 	}
 
@@ -2253,10 +2251,9 @@ mrvl_rx_pkt_burst(void *rxq, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
 
 	ret = pp2_ppio_recv(q->priv->ppio, q->priv->rxq_map[q->queue_id].tc,
 			    q->priv->rxq_map[q->queue_id].inq, descs, &nb_pkts);
-	if (unlikely(ret < 0)) {
-		MRVL_LOG(ERR, "Failed to receive packets");
+	if (unlikely(ret < 0))
 		return 0;
-	}
+
 	mrvl_port_bpool_size[bpool->pp2_id][bpool->id][core_id] -= nb_pkts;
 
 	for (i = 0; i < nb_pkts; i++) {
@@ -2319,21 +2316,13 @@ mrvl_rx_pkt_burst(void *rxq, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
 
 		if (unlikely(num <= q->priv->bpool_min_size ||
 			     (!rx_done && num < q->priv->bpool_init_size))) {
-			ret = mrvl_fill_bpool(q, MRVL_BURST_SIZE);
-			if (ret)
-				MRVL_LOG(ERR, "Failed to fill bpool");
+			mrvl_fill_bpool(q, MRVL_BURST_SIZE);
 		} else if (unlikely(num > q->priv->bpool_max_size)) {
 			int i;
 			int pkt_to_remove = num - q->priv->bpool_init_size;
 			struct rte_mbuf *mbuf;
 			struct pp2_buff_inf buff;
 
-			MRVL_LOG(DEBUG,
-				"port-%d:%d: bpool %d oversize - remove %d buffers (pool size: %d -> %d)",
-				bpool->pp2_id, q->priv->ppio->port_id,
-				bpool->id, pkt_to_remove, num,
-				q->priv->bpool_init_size);
-
 			for (i = 0; i < pkt_to_remove; i++) {
 				ret = pp2_bpool_get_buff(hif, bpool, &buff);
 				if (ret)
@@ -2526,12 +2515,8 @@ mrvl_tx_pkt_burst(void *txq, struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
 				       sq, q->queue_id, 0);
 
 	sq_free_size = MRVL_PP2_TX_SHADOWQ_SIZE - sq->size - 1;
-	if (unlikely(nb_pkts > sq_free_size)) {
-		MRVL_LOG(DEBUG,
-			"No room in shadow queue for %d packets! %d packets will be sent.",
-			nb_pkts, sq_free_size);
+	if (unlikely(nb_pkts > sq_free_size))
 		nb_pkts = sq_free_size;
-	}
 
 	for (i = 0; i < nb_pkts; i++) {
 		struct rte_mbuf *mbuf = tx_pkts[i];
@@ -2648,10 +2633,6 @@ mrvl_tx_sg_pkt_burst(void *txq, struct rte_mbuf **tx_pkts,
 		 */
 		if (unlikely(total_descs > sq_free_size)) {
 			total_descs -= nb_segs;
-			RTE_LOG(DEBUG, PMD,
-				"No room in shadow queue for %d packets! "
-				"%d packets will be sent.\n",
-				nb_pkts, i);
 			break;
 		}
 
-- 
2.28.0

[dpdk-stable] [PATCH v2 03/37] net/mvpp2: fix Rx/Tx bytes statistics

[lironh]

From: Yuri Chipchev <yuric@marvell.com>

4B of CRC was not included in the bytes statistics

Fixes: bdffe0c70 ("net/mrvl: support basic stats")
Cc: stable@dpdk.org

Signed-off-by: Yuri Chipchev <yuric@marvell.com>
Reviewed-by: Liron Himi <lironh@marvell.com>
---
 drivers/net/mvpp2/mrvl_ethdev.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/net/mvpp2/mrvl_ethdev.c b/drivers/net/mvpp2/mrvl_ethdev.c
index bfe496ff7..a1a3c0e31 100644
--- a/drivers/net/mvpp2/mrvl_ethdev.c
+++ b/drivers/net/mvpp2/mrvl_ethdev.c
@@ -1278,7 +1278,6 @@ mrvl_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
 				       rx_stats.drop_fullq +
 				       rx_stats.drop_bm +
 				       rxq->drop_mac;
-		stats->ibytes += rxq->bytes_recv;
 		drop_mac += rxq->drop_mac;
 	}
 
@@ -1306,7 +1305,6 @@ mrvl_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
 
 		stats->q_opackets[idx] = tx_stats.deq_desc;
 		stats->q_obytes[idx] = txq->bytes_sent;
-		stats->obytes += txq->bytes_sent;
 	}
 
 	ret = pp2_ppio_get_statistics(priv->ppio, &ppio_stats, 0);
@@ -1315,6 +1313,8 @@ mrvl_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
 		return ret;
 	}
 
+	stats->ibytes += ppio_stats.rx_bytes;
+	stats->obytes += ppio_stats.tx_bytes;
 	stats->ipackets += ppio_stats.rx_packets - drop_mac;
 	stats->opackets += ppio_stats.tx_packets;
 	stats->imissed += ppio_stats.rx_fullq_dropped +
-- 
2.28.0

[dpdk-stable] [PATCH v2 04/37] net/mvpp2: remove VLAN flush

[lironh]

From: Liron Himi <lironh@marvell.com>

VLAN-flush in MUSDK is not supported yet.
until it does, the code should be removed as currently
an redundant error message is displayed.

Fixes: a8f3d6783 ("net/mrvl: support VLAN filtering")
Cc: stable@dpdk.org

Signed-off-by: Liron Himi <lironh@marvell.com>
---
 drivers/net/mvpp2/mrvl_ethdev.c | 12 ------------
 drivers/net/mvpp2/mrvl_ethdev.h |  1 -
 2 files changed, 13 deletions(-)

diff --git a/drivers/net/mvpp2/mrvl_ethdev.c b/drivers/net/mvpp2/mrvl_ethdev.c
index a1a3c0e31..559379fcd 100644
--- a/drivers/net/mvpp2/mrvl_ethdev.c
+++ b/drivers/net/mvpp2/mrvl_ethdev.c
@@ -671,18 +671,6 @@ mrvl_dev_start(struct rte_eth_dev *dev)
 		priv->uc_mc_flushed = 1;
 	}
 
-	if (!priv->vlan_flushed) {
-		ret = pp2_ppio_flush_vlan(priv->ppio);
-		if (ret) {
-			MRVL_LOG(ERR, "Failed to flush vlan list");
-			/*
-			 * TODO
-			 * once pp2_ppio_flush_vlan() is supported jump to out
-			 * goto out;
-			 */
-		}
-		priv->vlan_flushed = 1;
-	}
 	ret = mrvl_mtu_set(dev, dev->data->mtu);
 	if (ret)
 		MRVL_LOG(ERR, "Failed to set MTU to %d", dev->data->mtu);
diff --git a/drivers/net/mvpp2/mrvl_ethdev.h b/drivers/net/mvpp2/mrvl_ethdev.h
index db6632f5b..eee5182ce 100644
--- a/drivers/net/mvpp2/mrvl_ethdev.h
+++ b/drivers/net/mvpp2/mrvl_ethdev.h
@@ -186,7 +186,6 @@ struct mrvl_priv {
 	uint8_t bpool_bit;
 	uint8_t rss_hf_tcp;
 	uint8_t uc_mc_flushed;
-	uint8_t vlan_flushed;
 	uint8_t isolated;
 	uint8_t multiseg;
 
-- 
2.28.0

[dpdk-stable] [PATCH v2 05/37] net/mvpp2: remove CRC len from MRU validation

[lironh]

From: Liron Himi <lironh@marvell.com>

CRC is being removed by HW before packet get
write to the buffer, so CRC len should not be
included in MRU validation

Fixes: 79ec62028 ("net/mvpp2: update MTU and MRU related calculations")
Cc: stable@dpdk.org

Signed-off-by: Liron Himi <lironh@marvell.com>
Reviewed-by: Yuri Chipchev <yuric@marvell.com>
---
 drivers/net/mvpp2/mrvl_ethdev.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/net/mvpp2/mrvl_ethdev.c b/drivers/net/mvpp2/mrvl_ethdev.c
index 559379fcd..133006f3e 100644
--- a/drivers/net/mvpp2/mrvl_ethdev.c
+++ b/drivers/net/mvpp2/mrvl_ethdev.c
@@ -441,8 +441,8 @@ mrvl_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
 	 * when this feature has not been enabled/supported so far
 	 * (TODO check scattered_rx flag here once scattered RX is supported).
 	 */
-	if (mru + MRVL_PKT_OFFS > mbuf_data_size) {
-		mru = mbuf_data_size - MRVL_PKT_OFFS;
+	if (mru - RTE_ETHER_CRC_LEN + MRVL_PKT_OFFS > mbuf_data_size) {
+		mru = mbuf_data_size + RTE_ETHER_CRC_LEN - MRVL_PKT_OFFS;
 		mtu = MRVL_PP2_MRU_TO_MTU(mru);
 		MRVL_LOG(WARNING, "MTU too big, max MTU possible limitted "
 			"by current mbuf size: %u. Set MTU to %u, MRU to %u",
-- 
2.28.0

[dpdk-stable] [PATCH v2 06/37] net/mvpp2: fix frame size checking

[lironh]

From: Liron Himi <lironh@marvell.com>

Need to add CRC len to the frame-size to compare against
max_rx_pkt_len which includes it.

Fixes: 79ec62028 ("net/mvpp2: update MTU and MRU related calculations")
Cc: stable@dpdk.org

Signed-off-by: Liron Himi <lironh@marvell.com>
Reviewed-by: Yuri Chipchev <yuric@marvell.com>
---
 drivers/net/mvpp2/mrvl_ethdev.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/net/mvpp2/mrvl_ethdev.c b/drivers/net/mvpp2/mrvl_ethdev.c
index 133006f3e..f7e3f62e3 100644
--- a/drivers/net/mvpp2/mrvl_ethdev.c
+++ b/drivers/net/mvpp2/mrvl_ethdev.c
@@ -1699,7 +1699,8 @@ mrvl_rx_queue_setup(struct rte_eth_dev *dev, uint16_t idx, uint16_t desc,
 		return -EFAULT;
 	}
 
-	frame_size = buf_size - RTE_PKTMBUF_HEADROOM - MRVL_PKT_EFFEC_OFFS;
+	frame_size = buf_size - RTE_PKTMBUF_HEADROOM -
+		     MRVL_PKT_EFFEC_OFFS + RTE_ETHER_CRC_LEN;
 	if (frame_size < max_rx_pkt_len) {
 		MRVL_LOG(WARNING,
 			"Mbuf size must be increased to %u bytes to hold up "
-- 
2.28.0

Re: [dpdk-stable] [PATCH v2 03/37] net/mvpp2: fix Rx/Tx bytes statistics

[Ferruh Yigit]

On 1/22/2021 7:18 PM, lironh@marvell.com wrote:
> From: Yuri Chipchev <yuric@marvell.com>
> 
> 4B of CRC was not included in the bytes statistics
> 
> Fixes: bdffe0c70 ("net/mrvl: support basic stats")
> Cc: stable@dpdk.org
> 
> Signed-off-by: Yuri Chipchev <yuric@marvell.com>
> Reviewed-by: Liron Himi <lironh@marvell.com>
> ---
>   drivers/net/mvpp2/mrvl_ethdev.c | 4 ++--
>   1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/net/mvpp2/mrvl_ethdev.c b/drivers/net/mvpp2/mrvl_ethdev.c
> index bfe496ff7..a1a3c0e31 100644
> --- a/drivers/net/mvpp2/mrvl_ethdev.c
> +++ b/drivers/net/mvpp2/mrvl_ethdev.c
> @@ -1278,7 +1278,6 @@ mrvl_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
>   				       rx_stats.drop_fullq +
>   				       rx_stats.drop_bm +
>   				       rxq->drop_mac;
> -		stats->ibytes += rxq->bytes_recv;
>   		drop_mac += rxq->drop_mac;
>   	}
>   
> @@ -1306,7 +1305,6 @@ mrvl_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
>   
>   		stats->q_opackets[idx] = tx_stats.deq_desc;
>   		stats->q_obytes[idx] = txq->bytes_sent;
> -		stats->obytes += txq->bytes_sent;
>   	}
>   
>   	ret = pp2_ppio_get_statistics(priv->ppio, &ppio_stats, 0);
> @@ -1315,6 +1313,8 @@ mrvl_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
>   		return ret;
>   	}
>   
> +	stats->ibytes += ppio_stats.rx_bytes;
> +	stats->obytes += ppio_stats.tx_bytes;
>   	stats->ipackets += ppio_stats.rx_packets - drop_mac;
>   	stats->opackets += ppio_stats.tx_packets;
>   	stats->imissed += ppio_stats.rx_fullq_dropped +
> 

Is this patch adding CRC to the stats? If so that is wrong, stats should not 
have the CRC.

A recent clarification from Andrew:
https://patches.dpdk.org/patch/86975/

Re: [dpdk-stable] [EXT] Re: [PATCH v2 03/37] net/mvpp2: fix Rx/Tx bytes statistics

[Liron Himi]

-----Original Message-----
From: Ferruh Yigit <ferruh.yigit@intel.com> 
Sent: Tuesday, 26 January 2021 19:03
To: Liron Himi <lironh@marvell.com>; Jerin Jacob Kollanukkaran <jerinj@marvell.com>
Cc: dev@dpdk.org; Yuri Chipchev <yuric@marvell.com>; stable@dpdk.org; Andrew Rybchenko <arybchenko@solarflare.com>
Subject: [EXT] Re: [dpdk-stable] [PATCH v2 03/37] net/mvpp2: fix Rx/Tx bytes statistics

External Email

----------------------------------------------------------------------
On 1/22/2021 7:18 PM, lironh@marvell.com wrote:
> From: Yuri Chipchev <yuric@marvell.com>
> 
> 4B of CRC was not included in the bytes statistics
> 
> Fixes: bdffe0c70 ("net/mrvl: support basic stats")
> Cc: stable@dpdk.org
> 
> Signed-off-by: Yuri Chipchev <yuric@marvell.com>
> Reviewed-by: Liron Himi <lironh@marvell.com>
> ---
>   drivers/net/mvpp2/mrvl_ethdev.c | 4 ++--
>   1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/net/mvpp2/mrvl_ethdev.c 
> b/drivers/net/mvpp2/mrvl_ethdev.c index bfe496ff7..a1a3c0e31 100644
> --- a/drivers/net/mvpp2/mrvl_ethdev.c
> +++ b/drivers/net/mvpp2/mrvl_ethdev.c
> @@ -1278,7 +1278,6 @@ mrvl_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
>   				       rx_stats.drop_fullq +
>   				       rx_stats.drop_bm +
>   				       rxq->drop_mac;
> -		stats->ibytes += rxq->bytes_recv;
>   		drop_mac += rxq->drop_mac;
>   	}
>   
> @@ -1306,7 +1305,6 @@ mrvl_stats_get(struct rte_eth_dev *dev, struct 
> rte_eth_stats *stats)
>   
>   		stats->q_opackets[idx] = tx_stats.deq_desc;
>   		stats->q_obytes[idx] = txq->bytes_sent;
> -		stats->obytes += txq->bytes_sent;
>   	}
>   
>   	ret = pp2_ppio_get_statistics(priv->ppio, &ppio_stats, 0); @@ 
> -1315,6 +1313,8 @@ mrvl_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
>   		return ret;
>   	}
>   
> +	stats->ibytes += ppio_stats.rx_bytes;
> +	stats->obytes += ppio_stats.tx_bytes;
>   	stats->ipackets += ppio_stats.rx_packets - drop_mac;
>   	stats->opackets += ppio_stats.tx_packets;
>   	stats->imissed += ppio_stats.rx_fullq_dropped +
> 

Is this patch adding CRC to the stats? If so that is wrong, stats should not have the CRC.

A recent clarification from Andrew:
https://urldefense.proofpoint.com/v2/url?u=https-3A__patches.dpdk.org_patch_86975_&d=DwIDaQ&c=nKjWec2b6R0mOyPaz7xtfQ&r=M6lEONY3jVnAkBjt0yC6oAsT7Jn2HI7LhNHCQYlnx1Q&m=iYhQSv2FUrmxLNQcxo6cAiLpoeqeTx1mQKadLot0jh0&s=lYaS0RY6n4QOlhzd3Z9PTxtR42lbqxPt1Pm1uODpWJg&e= 

[L.H.] okay, so based on this new clarification, this patch should not be merged to both next and stable branches.
thanks

Re: [dpdk-stable] [EXT] Re: [PATCH v2 03/37] net/mvpp2: fix Rx/Tx bytes statistics

[Ferruh Yigit]

On 1/26/2021 5:25 PM, Liron Himi wrote:
> 
> 
> -----Original Message-----
> From: Ferruh Yigit <ferruh.yigit@intel.com>
> Sent: Tuesday, 26 January 2021 19:03
> To: Liron Himi <lironh@marvell.com>; Jerin Jacob Kollanukkaran <jerinj@marvell.com>
> Cc: dev@dpdk.org; Yuri Chipchev <yuric@marvell.com>; stable@dpdk.org; Andrew Rybchenko <arybchenko@solarflare.com>
> Subject: [EXT] Re: [dpdk-stable] [PATCH v2 03/37] net/mvpp2: fix Rx/Tx bytes statistics
> 
> External Email
> 
> ----------------------------------------------------------------------
> On 1/22/2021 7:18 PM, lironh@marvell.com wrote:
>> From: Yuri Chipchev <yuric@marvell.com>
>>
>> 4B of CRC was not included in the bytes statistics
>>
>> Fixes: bdffe0c70 ("net/mrvl: support basic stats")
>> Cc: stable@dpdk.org
>>
>> Signed-off-by: Yuri Chipchev <yuric@marvell.com>
>> Reviewed-by: Liron Himi <lironh@marvell.com>
>> ---
>>    drivers/net/mvpp2/mrvl_ethdev.c | 4 ++--
>>    1 file changed, 2 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/net/mvpp2/mrvl_ethdev.c
>> b/drivers/net/mvpp2/mrvl_ethdev.c index bfe496ff7..a1a3c0e31 100644
>> --- a/drivers/net/mvpp2/mrvl_ethdev.c
>> +++ b/drivers/net/mvpp2/mrvl_ethdev.c
>> @@ -1278,7 +1278,6 @@ mrvl_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
>>    				       rx_stats.drop_fullq +
>>    				       rx_stats.drop_bm +
>>    				       rxq->drop_mac;
>> -		stats->ibytes += rxq->bytes_recv;
>>    		drop_mac += rxq->drop_mac;
>>    	}
>>    
>> @@ -1306,7 +1305,6 @@ mrvl_stats_get(struct rte_eth_dev *dev, struct
>> rte_eth_stats *stats)
>>    
>>    		stats->q_opackets[idx] = tx_stats.deq_desc;
>>    		stats->q_obytes[idx] = txq->bytes_sent;
>> -		stats->obytes += txq->bytes_sent;
>>    	}
>>    
>>    	ret = pp2_ppio_get_statistics(priv->ppio, &ppio_stats, 0); @@
>> -1315,6 +1313,8 @@ mrvl_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
>>    		return ret;
>>    	}
>>    
>> +	stats->ibytes += ppio_stats.rx_bytes;
>> +	stats->obytes += ppio_stats.tx_bytes;
>>    	stats->ipackets += ppio_stats.rx_packets - drop_mac;
>>    	stats->opackets += ppio_stats.tx_packets;
>>    	stats->imissed += ppio_stats.rx_fullq_dropped +
>>
> 
> Is this patch adding CRC to the stats? If so that is wrong, stats should not have the CRC.
> 
> A recent clarification from Andrew:
> https://urldefense.proofpoint.com/v2/url?u=https-3A__patches.dpdk.org_patch_86975_&d=DwIDaQ&c=nKjWec2b6R0mOyPaz7xtfQ&r=M6lEONY3jVnAkBjt0yC6oAsT7Jn2HI7LhNHCQYlnx1Q&m=iYhQSv2FUrmxLNQcxo6cAiLpoeqeTx1mQKadLot0jh0&s=lYaS0RY6n4QOlhzd3Z9PTxtR42lbqxPt1Pm1uODpWJg&e=
> 
> [L.H.] okay, so based on this new clarification, this patch should not be merged to both next and stable branches.
> thanks
> 

sure, I can drop it in next-net.

[dpdk-stable] [PATCH v3 01/34] net/mvpp2: fix stack corruption

[lironh]

From: Yuri Chipchev <yuric@marvell.com>

Fixes stack corruption in mrvl_fill_bpool function
in case num > MRVL_PP2_RXD_MAX

Fixes: c3637258d894 ("net/mrvl: fix Rx descriptors number")
Cc: stable@dpdk.org

Signed-off-by: Yuri Chipchev <yuric@marvell.com>
Reviewed-by: Liron Himi <lironh@marvell.com>
---
 drivers/net/mvpp2/mrvl_ethdev.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/net/mvpp2/mrvl_ethdev.c b/drivers/net/mvpp2/mrvl_ethdev.c
index f25cf9e46d..93fb30cdb8 100644
--- a/drivers/net/mvpp2/mrvl_ethdev.c
+++ b/drivers/net/mvpp2/mrvl_ethdev.c
@@ -1614,8 +1614,8 @@ mrvl_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
 static int
 mrvl_fill_bpool(struct mrvl_rxq *rxq, int num)
 {
-	struct buff_release_entry entries[MRVL_PP2_RXD_MAX];
-	struct rte_mbuf *mbufs[MRVL_PP2_RXD_MAX];
+	struct buff_release_entry entries[num];
+	struct rte_mbuf *mbufs[num];
 	int i, ret;
 	unsigned int core_id;
 	struct pp2_hif *hif;
-- 
2.28.0

[dpdk-stable] [PATCH v3 02/34] net/mvpp2: remove debug log on fast-path

[lironh]

From: Liron Himi <lironh@marvell.com>

In case of non-ip frame the current code reached the 'default'
case which result with function call to log a msg.
those kind of calls should not be performed on fast-path.

The performance for this kind of frames increased by 50%

Fixes: acab7d58c ("net/mvpp2: convert to dynamic logging")
Cc: stable@dpdk.org

Signed-off-by: Liron Himi <lironh@marvell.com>
---
 drivers/net/mvpp2/mrvl_ethdev.c | 27 ++++-----------------------
 1 file changed, 4 insertions(+), 23 deletions(-)

diff --git a/drivers/net/mvpp2/mrvl_ethdev.c b/drivers/net/mvpp2/mrvl_ethdev.c
index 93fb30cdb8..bfe496ff74 100644
--- a/drivers/net/mvpp2/mrvl_ethdev.c
+++ b/drivers/net/mvpp2/mrvl_ethdev.c
@@ -2171,7 +2171,6 @@ mrvl_desc_to_packet_type_and_offset(struct pp2_ppio_desc *desc,
 		*l4_offset = *l3_offset + MRVL_ARP_LENGTH;
 		break;
 	default:
-		MRVL_LOG(DEBUG, "Failed to recognise l3 packet type");
 		break;
 	}
 
@@ -2183,7 +2182,6 @@ mrvl_desc_to_packet_type_and_offset(struct pp2_ppio_desc *desc,
 		packet_type |= RTE_PTYPE_L4_UDP;
 		break;
 	default:
-		MRVL_LOG(DEBUG, "Failed to recognise l4 packet type");
 		break;
 	}
 
@@ -2253,10 +2251,9 @@ mrvl_rx_pkt_burst(void *rxq, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
 
 	ret = pp2_ppio_recv(q->priv->ppio, q->priv->rxq_map[q->queue_id].tc,
 			    q->priv->rxq_map[q->queue_id].inq, descs, &nb_pkts);
-	if (unlikely(ret < 0)) {
-		MRVL_LOG(ERR, "Failed to receive packets");
+	if (unlikely(ret < 0))
 		return 0;
-	}
+
 	mrvl_port_bpool_size[bpool->pp2_id][bpool->id][core_id] -= nb_pkts;
 
 	for (i = 0; i < nb_pkts; i++) {
@@ -2319,21 +2316,13 @@ mrvl_rx_pkt_burst(void *rxq, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
 
 		if (unlikely(num <= q->priv->bpool_min_size ||
 			     (!rx_done && num < q->priv->bpool_init_size))) {
-			ret = mrvl_fill_bpool(q, MRVL_BURST_SIZE);
-			if (ret)
-				MRVL_LOG(ERR, "Failed to fill bpool");
+			mrvl_fill_bpool(q, MRVL_BURST_SIZE);
 		} else if (unlikely(num > q->priv->bpool_max_size)) {
 			int i;
 			int pkt_to_remove = num - q->priv->bpool_init_size;
 			struct rte_mbuf *mbuf;
 			struct pp2_buff_inf buff;
 
-			MRVL_LOG(DEBUG,
-				"port-%d:%d: bpool %d oversize - remove %d buffers (pool size: %d -> %d)",
-				bpool->pp2_id, q->priv->ppio->port_id,
-				bpool->id, pkt_to_remove, num,
-				q->priv->bpool_init_size);
-
 			for (i = 0; i < pkt_to_remove; i++) {
 				ret = pp2_bpool_get_buff(hif, bpool, &buff);
 				if (ret)
@@ -2526,12 +2515,8 @@ mrvl_tx_pkt_burst(void *txq, struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
 				       sq, q->queue_id, 0);
 
 	sq_free_size = MRVL_PP2_TX_SHADOWQ_SIZE - sq->size - 1;
-	if (unlikely(nb_pkts > sq_free_size)) {
-		MRVL_LOG(DEBUG,
-			"No room in shadow queue for %d packets! %d packets will be sent.",
-			nb_pkts, sq_free_size);
+	if (unlikely(nb_pkts > sq_free_size))
 		nb_pkts = sq_free_size;
-	}
 
 	for (i = 0; i < nb_pkts; i++) {
 		struct rte_mbuf *mbuf = tx_pkts[i];
@@ -2648,10 +2633,6 @@ mrvl_tx_sg_pkt_burst(void *txq, struct rte_mbuf **tx_pkts,
 		 */
 		if (unlikely(total_descs > sq_free_size)) {
 			total_descs -= nb_segs;
-			RTE_LOG(DEBUG, PMD,
-				"No room in shadow queue for %d packets! "
-				"%d packets will be sent.\n",
-				nb_pkts, i);
 			break;
 		}
 
-- 
2.28.0

[dpdk-stable] [PATCH v3 03/34] net/mvpp2: remove VLAN flush

[lironh]

From: Liron Himi <lironh@marvell.com>

VLAN-flush in MUSDK is not supported yet.
until it does, the code should be removed as currently
an redundant error message is displayed.

Fixes: a8f3d6783 ("net/mrvl: support VLAN filtering")
Cc: stable@dpdk.org

Signed-off-by: Liron Himi <lironh@marvell.com>
---
 drivers/net/mvpp2/mrvl_ethdev.c | 12 ------------
 drivers/net/mvpp2/mrvl_ethdev.h |  1 -
 2 files changed, 13 deletions(-)

diff --git a/drivers/net/mvpp2/mrvl_ethdev.c b/drivers/net/mvpp2/mrvl_ethdev.c
index bfe496ff74..0c338c0c00 100644
--- a/drivers/net/mvpp2/mrvl_ethdev.c
+++ b/drivers/net/mvpp2/mrvl_ethdev.c
@@ -671,18 +671,6 @@ mrvl_dev_start(struct rte_eth_dev *dev)
 		priv->uc_mc_flushed = 1;
 	}
 
-	if (!priv->vlan_flushed) {
-		ret = pp2_ppio_flush_vlan(priv->ppio);
-		if (ret) {
-			MRVL_LOG(ERR, "Failed to flush vlan list");
-			/*
-			 * TODO
-			 * once pp2_ppio_flush_vlan() is supported jump to out
-			 * goto out;
-			 */
-		}
-		priv->vlan_flushed = 1;
-	}
 	ret = mrvl_mtu_set(dev, dev->data->mtu);
 	if (ret)
 		MRVL_LOG(ERR, "Failed to set MTU to %d", dev->data->mtu);
diff --git a/drivers/net/mvpp2/mrvl_ethdev.h b/drivers/net/mvpp2/mrvl_ethdev.h
index db6632f5b6..eee5182ce8 100644
--- a/drivers/net/mvpp2/mrvl_ethdev.h
+++ b/drivers/net/mvpp2/mrvl_ethdev.h
@@ -186,7 +186,6 @@ struct mrvl_priv {
 	uint8_t bpool_bit;
 	uint8_t rss_hf_tcp;
 	uint8_t uc_mc_flushed;
-	uint8_t vlan_flushed;
 	uint8_t isolated;
 	uint8_t multiseg;
 
-- 
2.28.0

[dpdk-stable] [PATCH v3 04/34] net/mvpp2: remove CRC len from MRU validation

[lironh]

From: Liron Himi <lironh@marvell.com>

CRC is being removed by HW before packet get
write to the buffer, so CRC len should not be
included in MRU validation

Fixes: 79ec62028 ("net/mvpp2: update MTU and MRU related calculations")
Cc: stable@dpdk.org

Signed-off-by: Liron Himi <lironh@marvell.com>
Reviewed-by: Yuri Chipchev <yuric@marvell.com>
---
 drivers/net/mvpp2/mrvl_ethdev.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/net/mvpp2/mrvl_ethdev.c b/drivers/net/mvpp2/mrvl_ethdev.c
index 0c338c0c00..eafe4f7ac9 100644
--- a/drivers/net/mvpp2/mrvl_ethdev.c
+++ b/drivers/net/mvpp2/mrvl_ethdev.c
@@ -441,8 +441,8 @@ mrvl_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
 	 * when this feature has not been enabled/supported so far
 	 * (TODO check scattered_rx flag here once scattered RX is supported).
 	 */
-	if (mru + MRVL_PKT_OFFS > mbuf_data_size) {
-		mru = mbuf_data_size - MRVL_PKT_OFFS;
+	if (mru - RTE_ETHER_CRC_LEN + MRVL_PKT_OFFS > mbuf_data_size) {
+		mru = mbuf_data_size + RTE_ETHER_CRC_LEN - MRVL_PKT_OFFS;
 		mtu = MRVL_PP2_MRU_TO_MTU(mru);
 		MRVL_LOG(WARNING, "MTU too big, max MTU possible limitted "
 			"by current mbuf size: %u. Set MTU to %u, MRU to %u",
-- 
2.28.0

[dpdk-stable] [PATCH v3 05/34] net/mvpp2: fix frame size checking

[lironh]

From: Liron Himi <lironh@marvell.com>

Need to add CRC len to the frame-size to compare against
max_rx_pkt_len which includes it.

Fixes: 79ec62028 ("net/mvpp2: update MTU and MRU related calculations")
Cc: stable@dpdk.org

Signed-off-by: Liron Himi <lironh@marvell.com>
Reviewed-by: Yuri Chipchev <yuric@marvell.com>
---
 drivers/net/mvpp2/mrvl_ethdev.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/net/mvpp2/mrvl_ethdev.c b/drivers/net/mvpp2/mrvl_ethdev.c
index eafe4f7ac9..6cd5acd337 100644
--- a/drivers/net/mvpp2/mrvl_ethdev.c
+++ b/drivers/net/mvpp2/mrvl_ethdev.c
@@ -1699,7 +1699,8 @@ mrvl_rx_queue_setup(struct rte_eth_dev *dev, uint16_t idx, uint16_t desc,
 		return -EFAULT;
 	}
 
-	frame_size = buf_size - RTE_PKTMBUF_HEADROOM - MRVL_PKT_EFFEC_OFFS;
+	frame_size = buf_size - RTE_PKTMBUF_HEADROOM -
+		     MRVL_PKT_EFFEC_OFFS + RTE_ETHER_CRC_LEN;
 	if (frame_size < max_rx_pkt_len) {
 		MRVL_LOG(WARNING,
 			"Mbuf size must be increased to %u bytes to hold up "
-- 
2.28.0