[PATCH v2 12/28] net/cnxk: add barrier after meta batch free in scalar

[PATCH v2 12/28] net/cnxk: add barrier after meta batch free in scalar

[Nithin Dabilpuram]

Add barrier after meta batch free in scalar routine when
lmt lines are exactly full to make sure that next LMT line user
in Tx only starts writing the lines only when previous stoerl's
are complete.

Fixes: 4382a7ccf781 ("net/cnxk: support Rx security offload on cn10k")
Cc: stable@dpdk.org

Signed-off-by: Nithin Dabilpuram <ndabilpuram@marvell.com>
---
 drivers/net/cnxk/cn10k_rx.h | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/drivers/net/cnxk/cn10k_rx.h b/drivers/net/cnxk/cn10k_rx.h
index e4f5a55..94c1f1e 100644
--- a/drivers/net/cnxk/cn10k_rx.h
+++ b/drivers/net/cnxk/cn10k_rx.h
@@ -1007,10 +1007,11 @@ cn10k_nix_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts, uint16_t pkts,
 	plt_write64((wdata | nb_pkts), rxq->cq_door);
 
 	/* Free remaining meta buffers if any */
-	if (flags & NIX_RX_OFFLOAD_SECURITY_F && loff) {
+	if (flags & NIX_RX_OFFLOAD_SECURITY_F && loff)
 		nix_sec_flush_meta(laddr, lmt_id + lnum, loff, aura_handle);
-		plt_io_wmb();
-	}
+
+	if (flags & NIX_RX_OFFLOAD_SECURITY_F)
+		rte_io_wmb();
 
 	return nb_pkts;
 }
-- 
2.8.4

[PATCH v2 14/28] net/cnxk: fix roundup size with transport mode

[Nithin Dabilpuram]

For transport mode, roundup needs to be based on L4 data
and shouldn't include L3 length.
By including l3 length, rlen that is calculated and put in
send hdr would cross the final length of the packet in some
scenarios where padding is necessary.
Also when outer and inner checksum offload flags are enabled,
get the l2_len and l3_len from il3ptr and il4ptr.

Fixes: 55bfac717c72 ("net/cnxk: support Tx security offload on cn10k")
Cc: stable@dpdk.org

Signed-off-by: Nithin Dabilpuram <ndabilpuram@marvell.com>
---
 drivers/net/cnxk/cn10k_tx.h | 34 ++++++++++++++++++++++++++--------
 1 file changed, 26 insertions(+), 8 deletions(-)

diff --git a/drivers/net/cnxk/cn10k_tx.h b/drivers/net/cnxk/cn10k_tx.h
index 981bc9b..c25825c 100644
--- a/drivers/net/cnxk/cn10k_tx.h
+++ b/drivers/net/cnxk/cn10k_tx.h
@@ -248,23 +248,29 @@ cn10k_nix_prep_sec_vec(struct rte_mbuf *m, uint64x2_t *cmd0, uint64x2_t *cmd1,
 	uint32_t pkt_len, dlen_adj, rlen;
 	uint8_t l3l4type, chksum;
 	uint64x2_t cmd01, cmd23;
+	uint8_t l2_len, l3_len;
 	uintptr_t dptr, nixtx;
 	uint64_t ucode_cmd[4];
 	uint64_t *laddr;
-	uint8_t l2_len;
 	uint16_t tag;
 	uint64_t sa;
 
 	sess_priv.u64 = *rte_security_dynfield(m);
 
 	if (flags & NIX_TX_NEED_SEND_HDR_W1) {
-		l2_len = vgetq_lane_u8(*cmd0, 8);
 		/* Extract l3l4type either from il3il4type or ol3ol4type */
 		if (flags & NIX_TX_OFFLOAD_L3_L4_CSUM_F &&
-		    flags & NIX_TX_OFFLOAD_OL3_OL4_CSUM_F)
+		    flags & NIX_TX_OFFLOAD_OL3_OL4_CSUM_F) {
+			l2_len = vgetq_lane_u8(*cmd0, 10);
+			/* L4 ptr from send hdr includes l2 and l3 len */
+			l3_len = vgetq_lane_u8(*cmd0, 11) - l2_len;
 			l3l4type = vgetq_lane_u8(*cmd0, 13);
-		else
+		} else {
+			l2_len = vgetq_lane_u8(*cmd0, 8);
+			/* L4 ptr from send hdr includes l2 and l3 len */
+			l3_len = vgetq_lane_u8(*cmd0, 9) - l2_len;
 			l3l4type = vgetq_lane_u8(*cmd0, 12);
+		}
 
 		chksum = (l3l4type & 0x1) << 1 | !!(l3l4type & 0x30);
 		chksum = ~chksum;
@@ -273,6 +279,7 @@ cn10k_nix_prep_sec_vec(struct rte_mbuf *m, uint64x2_t *cmd0, uint64x2_t *cmd1,
 		*cmd0 = vsetq_lane_u16(0, *cmd0, 6);
 	} else {
 		l2_len = m->l2_len;
+		l3_len = m->l3_len;
 	}
 
 	/* Retrieve DPTR */
@@ -281,6 +288,8 @@ cn10k_nix_prep_sec_vec(struct rte_mbuf *m, uint64x2_t *cmd0, uint64x2_t *cmd1,
 
 	/* Calculate dlen adj */
 	dlen_adj = pkt_len - l2_len;
+	/* Exclude l3 len from roundup for transport mode */
+	dlen_adj -= sess_priv.mode ? 0 : l3_len;
 	rlen = (dlen_adj + sess_priv.roundup_len) +
 	       (sess_priv.roundup_byte - 1);
 	rlen &= ~(uint64_t)(sess_priv.roundup_byte - 1);
@@ -360,10 +369,10 @@ cn10k_nix_prep_sec(struct rte_mbuf *m, uint64_t *cmd, uintptr_t *nixtx_addr,
 	uint8_t l3l4type, chksum;
 	uint64x2_t cmd01, cmd23;
 	union nix_send_sg_s *sg;
+	uint8_t l2_len, l3_len;
 	uintptr_t dptr, nixtx;
 	uint64_t ucode_cmd[4];
 	uint64_t *laddr;
-	uint8_t l2_len;
 	uint16_t tag;
 	uint64_t sa;
 
@@ -376,13 +385,19 @@ cn10k_nix_prep_sec(struct rte_mbuf *m, uint64_t *cmd, uintptr_t *nixtx_addr,
 		sg = (union nix_send_sg_s *)&cmd[2];
 
 	if (flags & NIX_TX_NEED_SEND_HDR_W1) {
-		l2_len = cmd[1] & 0xFF;
 		/* Extract l3l4type either from il3il4type or ol3ol4type */
 		if (flags & NIX_TX_OFFLOAD_L3_L4_CSUM_F &&
-		    flags & NIX_TX_OFFLOAD_OL3_OL4_CSUM_F)
+		    flags & NIX_TX_OFFLOAD_OL3_OL4_CSUM_F) {
+			l2_len = (cmd[1] >> 16) & 0xFF;
+			/* L4 ptr from send hdr includes l2 and l3 len */
+			l3_len = ((cmd[1] >> 24) & 0xFF) - l2_len;
 			l3l4type = (cmd[1] >> 40) & 0xFF;
-		else
+		} else {
+			l2_len = cmd[1] & 0xFF;
+			/* L4 ptr from send hdr includes l2 and l3 len */
+			l3_len = ((cmd[1] >> 8) & 0xFF) - l2_len;
 			l3l4type = (cmd[1] >> 32) & 0xFF;
+		}
 
 		chksum = (l3l4type & 0x1) << 1 | !!(l3l4type & 0x30);
 		chksum = ~chksum;
@@ -391,6 +406,7 @@ cn10k_nix_prep_sec(struct rte_mbuf *m, uint64_t *cmd, uintptr_t *nixtx_addr,
 		cmd[1] &= ~(0xFFFFUL << 32);
 	} else {
 		l2_len = m->l2_len;
+		l3_len = m->l3_len;
 	}
 
 	/* Retrieve DPTR */
@@ -399,6 +415,8 @@ cn10k_nix_prep_sec(struct rte_mbuf *m, uint64_t *cmd, uintptr_t *nixtx_addr,
 
 	/* Calculate dlen adj */
 	dlen_adj = pkt_len - l2_len;
+	/* Exclude l3 len from roundup for transport mode */
+	dlen_adj -= sess_priv.mode ? 0 : l3_len;
 	rlen = (dlen_adj + sess_priv.roundup_len) +
 	       (sess_priv.roundup_byte - 1);
 	rlen &= ~(uint64_t)(sess_priv.roundup_byte - 1);
-- 
2.8.4

[PATCH v2 28/28] net/cnxk: fix multi-seg extraction in vwqe path

[Nithin Dabilpuram]

Fix multi-seg extraction in vwqe path to avoid updating mbuf[]
array until it is used via cq0 path.

Fixes: 7fbbc981d54f ("event/cnxk: support vectorized Rx event fast path")
Cc: pbhagavatula@marvell.com
Cc: stable@dpdk.org

Signed-off-by: Nithin Dabilpuram <ndabilpuram@marvell.com>
---
 drivers/net/cnxk/cn10k_rx.h | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/drivers/net/cnxk/cn10k_rx.h b/drivers/net/cnxk/cn10k_rx.h
index 00bec01..5ecb20f 100644
--- a/drivers/net/cnxk/cn10k_rx.h
+++ b/drivers/net/cnxk/cn10k_rx.h
@@ -1673,10 +1673,6 @@ cn10k_nix_recv_pkts_vector(void *args, struct rte_mbuf **mbufs, uint16_t pkts,
 		vst1q_u64((uint64_t *)mbuf2->rearm_data, rearm2);
 		vst1q_u64((uint64_t *)mbuf3->rearm_data, rearm3);
 
-		/* Store the mbufs to rx_pkts */
-		vst1q_u64((uint64_t *)&mbufs[packets], mbuf01);
-		vst1q_u64((uint64_t *)&mbufs[packets + 2], mbuf23);
-
 		if (flags & NIX_RX_MULTI_SEG_F) {
 			/* Multi segment is enable build mseg list for
 			 * individual mbufs in scalar mode.
@@ -1695,6 +1691,10 @@ cn10k_nix_recv_pkts_vector(void *args, struct rte_mbuf **mbufs, uint16_t pkts,
 					    mbuf3, mbuf_initializer, flags);
 		}
 
+		/* Store the mbufs to rx_pkts */
+		vst1q_u64((uint64_t *)&mbufs[packets], mbuf01);
+		vst1q_u64((uint64_t *)&mbufs[packets + 2], mbuf23);
+
 		/* Mark mempool obj as "get" as it is alloc'ed by NIX */
 		RTE_MEMPOOL_CHECK_COOKIES(mbuf0->pool, (void **)&mbuf0, 1, 1);
 		RTE_MEMPOOL_CHECK_COOKIES(mbuf1->pool, (void **)&mbuf1, 1, 1);
-- 
2.8.4

RE: [PATCH v2 28/28] net/cnxk: fix multi-seg extraction in vwqe path

[Pavan Nikhilesh Bhagavatula]

> -----Original Message-----
> From: Nithin Dabilpuram <ndabilpuram@marvell.com>
> Sent: Friday, April 22, 2022 4:17 PM
> To: Jerin Jacob Kollanukkaran <jerinj@marvell.com>; Nithin Kumar
> Dabilpuram <ndabilpuram@marvell.com>; Kiran Kumar Kokkilagadda
> <kirankumark@marvell.com>; Sunil Kumar Kori <skori@marvell.com>; Satha
> Koteswara Rao Kottidi <skoteshwar@marvell.com>
> Cc: dev@dpdk.org; Pavan Nikhilesh Bhagavatula
> <pbhagavatula@marvell.com>; stable@dpdk.org
> Subject: [PATCH v2 28/28] net/cnxk: fix multi-seg extraction in vwqe path
> 
> Fix multi-seg extraction in vwqe path to avoid updating mbuf[]
> array until it is used via cq0 path.
> 
> Fixes: 7fbbc981d54f ("event/cnxk: support vectorized Rx event fast path")
> Cc: pbhagavatula@marvell.com
> Cc: stable@dpdk.org
> 
> Signed-off-by: Nithin Dabilpuram <ndabilpuram@marvell.com>

Acked-by: Pavan Nikhilesh <pbhagavatula@marvell.com>

> ---
>  drivers/net/cnxk/cn10k_rx.h | 8 ++++----
>  1 file changed, 4 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/net/cnxk/cn10k_rx.h b/drivers/net/cnxk/cn10k_rx.h
> index 00bec01..5ecb20f 100644
> --- a/drivers/net/cnxk/cn10k_rx.h
> +++ b/drivers/net/cnxk/cn10k_rx.h
> @@ -1673,10 +1673,6 @@ cn10k_nix_recv_pkts_vector(void *args, struct
> rte_mbuf **mbufs, uint16_t pkts,
>  		vst1q_u64((uint64_t *)mbuf2->rearm_data, rearm2);
>  		vst1q_u64((uint64_t *)mbuf3->rearm_data, rearm3);
> 
> -		/* Store the mbufs to rx_pkts */
> -		vst1q_u64((uint64_t *)&mbufs[packets], mbuf01);
> -		vst1q_u64((uint64_t *)&mbufs[packets + 2], mbuf23);
> -
>  		if (flags & NIX_RX_MULTI_SEG_F) {
>  			/* Multi segment is enable build mseg list for
>  			 * individual mbufs in scalar mode.
> @@ -1695,6 +1691,10 @@ cn10k_nix_recv_pkts_vector(void *args, struct
> rte_mbuf **mbufs, uint16_t pkts,
>  					    mbuf3, mbuf_initializer, flags);
>  		}
> 
> +		/* Store the mbufs to rx_pkts */
> +		vst1q_u64((uint64_t *)&mbufs[packets], mbuf01);
> +		vst1q_u64((uint64_t *)&mbufs[packets + 2], mbuf23);
> +
>  		/* Mark mempool obj as "get" as it is alloc'ed by NIX */
>  		RTE_MEMPOOL_CHECK_COOKIES(mbuf0->pool, (void
> **)&mbuf0, 1, 1);
>  		RTE_MEMPOOL_CHECK_COOKIES(mbuf1->pool, (void
> **)&mbuf1, 1, 1);
> --
> 2.8.4

Re: [PATCH v2 28/28] net/cnxk: fix multi-seg extraction in vwqe path

[Jerin Jacob]

On Fri, Apr 22, 2022 at 4:20 PM Nithin Dabilpuram
<ndabilpuram@marvell.com> wrote:
>
> Fix multi-seg extraction in vwqe path to avoid updating mbuf[]
> array until it is used via cq0 path.
>
> Fixes: 7fbbc981d54f ("event/cnxk: support vectorized Rx event fast path")
> Cc: pbhagavatula@marvell.com
> Cc: stable@dpdk.org
>
> Signed-off-by: Nithin Dabilpuram <ndabilpuram@marvell.com>

Please fix following check-gitlog.sh errors


Wrong headline prefix:
        net/cnxk: add receive channel backpressure for SDP
Is it candidate for Cc: stable@dpdk.org backport?
        common/cnxk: fix SQ flush sequence
        common/cnxk: fix issues in soft expiry disable path
        net/cnxk: optimize Rx fast path for security pkts


> ---
>  drivers/net/cnxk/cn10k_rx.h | 8 ++++----
>  1 file changed, 4 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/net/cnxk/cn10k_rx.h b/drivers/net/cnxk/cn10k_rx.h
> index 00bec01..5ecb20f 100644
> --- a/drivers/net/cnxk/cn10k_rx.h
> +++ b/drivers/net/cnxk/cn10k_rx.h
> @@ -1673,10 +1673,6 @@ cn10k_nix_recv_pkts_vector(void *args, struct rte_mbuf **mbufs, uint16_t pkts,
>                 vst1q_u64((uint64_t *)mbuf2->rearm_data, rearm2);
>                 vst1q_u64((uint64_t *)mbuf3->rearm_data, rearm3);
>
> -               /* Store the mbufs to rx_pkts */
> -               vst1q_u64((uint64_t *)&mbufs[packets], mbuf01);
> -               vst1q_u64((uint64_t *)&mbufs[packets + 2], mbuf23);
> -
>                 if (flags & NIX_RX_MULTI_SEG_F) {
>                         /* Multi segment is enable build mseg list for
>                          * individual mbufs in scalar mode.
> @@ -1695,6 +1691,10 @@ cn10k_nix_recv_pkts_vector(void *args, struct rte_mbuf **mbufs, uint16_t pkts,
>                                             mbuf3, mbuf_initializer, flags);
>                 }
>
> +               /* Store the mbufs to rx_pkts */
> +               vst1q_u64((uint64_t *)&mbufs[packets], mbuf01);
> +               vst1q_u64((uint64_t *)&mbufs[packets + 2], mbuf23);
> +
>                 /* Mark mempool obj as "get" as it is alloc'ed by NIX */
>                 RTE_MEMPOOL_CHECK_COOKIES(mbuf0->pool, (void **)&mbuf0, 1, 1);
>                 RTE_MEMPOOL_CHECK_COOKIES(mbuf1->pool, (void **)&mbuf1, 1, 1);
> --
> 2.8.4
>