[v3] net/mlx5: fix Tx metadata for multi-segment packet

Message ID 1548830609-29686-1-git-send-email-dekelp@mellanox.com
State Accepted, archived
Delegated to: Shahaf Shuler
Headers show
Series
  • [v3] net/mlx5: fix Tx metadata for multi-segment packet
Related show

Checks

Context Check Description
ci/mellanox-Performance-Testing success Performance Testing PASS
ci/intel-Performance-Testing success Performance Testing PASS
ci/Intel-compilation success Compilation OK
ci/checkpatch success coding style OK

Commit Message

Dekel Peled Jan. 30, 2019, 6:43 a.m.
Original patch implemented the use of match_metadata offload in the
different burst functions.
The concurrent use of match_metadata and multi_segs offloads was
not handled.

This patch updates function txq_scatter_v(), to pass metadata value
from mbuf to wqe, when indicated by offload flags.

Fixes: 6bd7fbd03c62 ("net/mlx5: support metadata as flow rule criteria")
Cc: stable@dpdk.org

Signed-off-by: Dekel Peled <dekelp@mellanox.com>

---
v3: Update title, modify indentation.
v2: Apply code review comments.
---
---
 drivers/net/mlx5/mlx5_rxtx_vec_neon.h | 12 +++++++++---
 drivers/net/mlx5/mlx5_rxtx_vec_sse.h  | 11 ++++++++---
 2 files changed, 17 insertions(+), 6 deletions(-)

Comments

Yongseok Koh Jan. 30, 2019, 7:24 a.m. | #1
> On Jan 30, 2019, at 3:43 PM, Dekel Peled <dekelp@mellanox.com> wrote:
> 
> Original patch implemented the use of match_metadata offload in the
> different burst functions.
> The concurrent use of match_metadata and multi_segs offloads was
> not handled.
> 
> This patch updates function txq_scatter_v(), to pass metadata value
> from mbuf to wqe, when indicated by offload flags.
> 
> Fixes: 6bd7fbd03c62 ("net/mlx5: support metadata as flow rule criteria")
> Cc: stable@dpdk.org
> 
> Signed-off-by: Dekel Peled <dekelp@mellanox.com>
> 
> ---

Acked-by: Yongseok Koh <yskoh@mellanox.com>
 
Thanks

> v3: Update title, modify indentation.
> v2: Apply code review comments.
> ---
> ---
> drivers/net/mlx5/mlx5_rxtx_vec_neon.h | 12 +++++++++---
> drivers/net/mlx5/mlx5_rxtx_vec_sse.h  | 11 ++++++++---
> 2 files changed, 17 insertions(+), 6 deletions(-)
> 
> diff --git a/drivers/net/mlx5/mlx5_rxtx_vec_neon.h b/drivers/net/mlx5/mlx5_rxtx_vec_neon.h
> index 883fe1b..38e915c 100644
> --- a/drivers/net/mlx5/mlx5_rxtx_vec_neon.h
> +++ b/drivers/net/mlx5/mlx5_rxtx_vec_neon.h
> @@ -104,6 +104,8 @@
> 		sizeof(struct mlx5_wqe) / MLX5_WQE_DWORD_SIZE;
> 	unsigned int n;
> 	volatile struct mlx5_wqe *wqe = NULL;
> +	bool metadata_ol =
> +		txq->offloads & DEV_TX_OFFLOAD_MATCH_METADATA ? true : false;
> 
> 	assert(elts_n > pkts_n);
> 	mlx5_tx_complete(txq);
> @@ -127,6 +129,9 @@
> 		uint8x16_t *t_wqe;
> 		uint8_t *dseg;
> 		uint8x16_t ctrl;
> +		rte_be32_t metadata =
> +			metadata_ol && (buf->ol_flags & PKT_TX_METADATA) ?
> +			buf->tx_metadata : 0;
> 
> 		assert(segs_n);
> 		max_elts = elts_n - (elts_head - txq->elts_tail);
> @@ -164,9 +169,10 @@
> 		ctrl = vqtbl1q_u8(ctrl, ctrl_shuf_m);
> 		vst1q_u8((void *)t_wqe, ctrl);
> 		/* Fill ESEG in the header. */
> -		vst1q_u16((void *)(t_wqe + 1),
> -			  ((uint16x8_t) { 0, 0, cs_flags, rte_cpu_to_be_16(len),
> -					  0, 0, 0, 0 }));
> +		vst1q_u32((void *)(t_wqe + 1),
> +			  ((uint32x4_t){ 0,
> +					 cs_flags << 16 | rte_cpu_to_be_16(len),
> +					 metadata, 0 }));
> 		txq->wqe_ci = wqe_ci;
> 	}
> 	if (!n)
> diff --git a/drivers/net/mlx5/mlx5_rxtx_vec_sse.h b/drivers/net/mlx5/mlx5_rxtx_vec_sse.h
> index 14117c4..fb384ef 100644
> --- a/drivers/net/mlx5/mlx5_rxtx_vec_sse.h
> +++ b/drivers/net/mlx5/mlx5_rxtx_vec_sse.h
> @@ -104,6 +104,8 @@
> 		sizeof(struct mlx5_wqe) / MLX5_WQE_DWORD_SIZE;
> 	unsigned int n;
> 	volatile struct mlx5_wqe *wqe = NULL;
> +	bool metadata_ol =
> +		txq->offloads & DEV_TX_OFFLOAD_MATCH_METADATA ? true : false;
> 
> 	assert(elts_n > pkts_n);
> 	mlx5_tx_complete(txq);
> @@ -125,6 +127,9 @@
> 		uint16_t max_wqe;
> 		__m128i *t_wqe, *dseg;
> 		__m128i ctrl;
> +		rte_be32_t metadata =
> +			metadata_ol && (buf->ol_flags & PKT_TX_METADATA) ?
> +			buf->tx_metadata : 0;
> 
> 		assert(segs_n);
> 		max_elts = elts_n - (elts_head - txq->elts_tail);
> @@ -165,9 +170,9 @@
> 		_mm_store_si128(t_wqe, ctrl);
> 		/* Fill ESEG in the header. */
> 		_mm_store_si128(t_wqe + 1,
> -				_mm_set_epi16(0, 0, 0, 0,
> -					      rte_cpu_to_be_16(len), cs_flags,
> -					      0, 0));
> +				_mm_set_epi32(0, metadata,
> +					      (rte_cpu_to_be_16(len) << 16) |
> +					      cs_flags, 0));
> 		txq->wqe_ci = wqe_ci;
> 	}
> 	if (!n)
> -- 
> 1.8.3.1
>
Shahaf Shuler Feb. 10, 2019, 3:10 p.m. | #2
Wednesday, January 30, 2019 9:24 AM, Yongseok Koh:
> Subject: Re: [PATCH v3] net/mlx5: fix Tx metadata for multi-segment packet
> 
> > On Jan 30, 2019, at 3:43 PM, Dekel Peled <dekelp@mellanox.com> wrote:
> >
> > Original patch implemented the use of match_metadata offload in the
> > different burst functions.
> > The concurrent use of match_metadata and multi_segs offloads was not
> > handled.
> >
> > This patch updates function txq_scatter_v(), to pass metadata value
> > from mbuf to wqe, when indicated by offload flags.
> >
> > Fixes: 6bd7fbd03c62 ("net/mlx5: support metadata as flow rule
> > criteria")
> > Cc: stable@dpdk.org
> >
> > Signed-off-by: Dekel Peled <dekelp@mellanox.com>

Applied to next-net-mlx, thanks .

> >
> > ---
> 
> Acked-by: Yongseok Koh <yskoh@mellanox.com>
> 
> Thanks
> 
> > v3: Update title, modify indentation.
> > v2: Apply code review comments.
> > ---
> > ---
> > drivers/net/mlx5/mlx5_rxtx_vec_neon.h | 12 +++++++++---
> > drivers/net/mlx5/mlx5_rxtx_vec_sse.h  | 11 ++++++++---
> > 2 files changed, 17 insertions(+), 6 deletions(-)
> >
> > diff --git a/drivers/net/mlx5/mlx5_rxtx_vec_neon.h
> > b/drivers/net/mlx5/mlx5_rxtx_vec_neon.h
> > index 883fe1b..38e915c 100644
> > --- a/drivers/net/mlx5/mlx5_rxtx_vec_neon.h
> > +++ b/drivers/net/mlx5/mlx5_rxtx_vec_neon.h
> > @@ -104,6 +104,8 @@
> > 		sizeof(struct mlx5_wqe) / MLX5_WQE_DWORD_SIZE;
> > 	unsigned int n;
> > 	volatile struct mlx5_wqe *wqe = NULL;
> > +	bool metadata_ol =
> > +		txq->offloads & DEV_TX_OFFLOAD_MATCH_METADATA ?
> true : false;
> >
> > 	assert(elts_n > pkts_n);
> > 	mlx5_tx_complete(txq);
> > @@ -127,6 +129,9 @@
> > 		uint8x16_t *t_wqe;
> > 		uint8_t *dseg;
> > 		uint8x16_t ctrl;
> > +		rte_be32_t metadata =
> > +			metadata_ol && (buf->ol_flags &
> PKT_TX_METADATA) ?
> > +			buf->tx_metadata : 0;
> >
> > 		assert(segs_n);
> > 		max_elts = elts_n - (elts_head - txq->elts_tail); @@ -164,9
> +169,10
> > @@
> > 		ctrl = vqtbl1q_u8(ctrl, ctrl_shuf_m);
> > 		vst1q_u8((void *)t_wqe, ctrl);
> > 		/* Fill ESEG in the header. */
> > -		vst1q_u16((void *)(t_wqe + 1),
> > -			  ((uint16x8_t) { 0, 0, cs_flags,
> rte_cpu_to_be_16(len),
> > -					  0, 0, 0, 0 }));
> > +		vst1q_u32((void *)(t_wqe + 1),
> > +			  ((uint32x4_t){ 0,
> > +					 cs_flags << 16 |
> rte_cpu_to_be_16(len),
> > +					 metadata, 0 }));
> > 		txq->wqe_ci = wqe_ci;
> > 	}
> > 	if (!n)
> > diff --git a/drivers/net/mlx5/mlx5_rxtx_vec_sse.h
> > b/drivers/net/mlx5/mlx5_rxtx_vec_sse.h
> > index 14117c4..fb384ef 100644
> > --- a/drivers/net/mlx5/mlx5_rxtx_vec_sse.h
> > +++ b/drivers/net/mlx5/mlx5_rxtx_vec_sse.h
> > @@ -104,6 +104,8 @@
> > 		sizeof(struct mlx5_wqe) / MLX5_WQE_DWORD_SIZE;
> > 	unsigned int n;
> > 	volatile struct mlx5_wqe *wqe = NULL;
> > +	bool metadata_ol =
> > +		txq->offloads & DEV_TX_OFFLOAD_MATCH_METADATA ?
> true : false;
> >
> > 	assert(elts_n > pkts_n);
> > 	mlx5_tx_complete(txq);
> > @@ -125,6 +127,9 @@
> > 		uint16_t max_wqe;
> > 		__m128i *t_wqe, *dseg;
> > 		__m128i ctrl;
> > +		rte_be32_t metadata =
> > +			metadata_ol && (buf->ol_flags &
> PKT_TX_METADATA) ?
> > +			buf->tx_metadata : 0;
> >
> > 		assert(segs_n);
> > 		max_elts = elts_n - (elts_head - txq->elts_tail); @@ -165,9
> +170,9
> > @@
> > 		_mm_store_si128(t_wqe, ctrl);
> > 		/* Fill ESEG in the header. */
> > 		_mm_store_si128(t_wqe + 1,
> > -				_mm_set_epi16(0, 0, 0, 0,
> > -					      rte_cpu_to_be_16(len), cs_flags,
> > -					      0, 0));
> > +				_mm_set_epi32(0, metadata,
> > +					      (rte_cpu_to_be_16(len) << 16) |
> > +					      cs_flags, 0));
> > 		txq->wqe_ci = wqe_ci;
> > 	}
> > 	if (!n)
> > --
> > 1.8.3.1
> >

Patch

diff --git a/drivers/net/mlx5/mlx5_rxtx_vec_neon.h b/drivers/net/mlx5/mlx5_rxtx_vec_neon.h
index 883fe1b..38e915c 100644
--- a/drivers/net/mlx5/mlx5_rxtx_vec_neon.h
+++ b/drivers/net/mlx5/mlx5_rxtx_vec_neon.h
@@ -104,6 +104,8 @@ 
 		sizeof(struct mlx5_wqe) / MLX5_WQE_DWORD_SIZE;
 	unsigned int n;
 	volatile struct mlx5_wqe *wqe = NULL;
+	bool metadata_ol =
+		txq->offloads & DEV_TX_OFFLOAD_MATCH_METADATA ? true : false;
 
 	assert(elts_n > pkts_n);
 	mlx5_tx_complete(txq);
@@ -127,6 +129,9 @@ 
 		uint8x16_t *t_wqe;
 		uint8_t *dseg;
 		uint8x16_t ctrl;
+		rte_be32_t metadata =
+			metadata_ol && (buf->ol_flags & PKT_TX_METADATA) ?
+			buf->tx_metadata : 0;
 
 		assert(segs_n);
 		max_elts = elts_n - (elts_head - txq->elts_tail);
@@ -164,9 +169,10 @@ 
 		ctrl = vqtbl1q_u8(ctrl, ctrl_shuf_m);
 		vst1q_u8((void *)t_wqe, ctrl);
 		/* Fill ESEG in the header. */
-		vst1q_u16((void *)(t_wqe + 1),
-			  ((uint16x8_t) { 0, 0, cs_flags, rte_cpu_to_be_16(len),
-					  0, 0, 0, 0 }));
+		vst1q_u32((void *)(t_wqe + 1),
+			  ((uint32x4_t){ 0,
+					 cs_flags << 16 | rte_cpu_to_be_16(len),
+					 metadata, 0 }));
 		txq->wqe_ci = wqe_ci;
 	}
 	if (!n)
diff --git a/drivers/net/mlx5/mlx5_rxtx_vec_sse.h b/drivers/net/mlx5/mlx5_rxtx_vec_sse.h
index 14117c4..fb384ef 100644
--- a/drivers/net/mlx5/mlx5_rxtx_vec_sse.h
+++ b/drivers/net/mlx5/mlx5_rxtx_vec_sse.h
@@ -104,6 +104,8 @@ 
 		sizeof(struct mlx5_wqe) / MLX5_WQE_DWORD_SIZE;
 	unsigned int n;
 	volatile struct mlx5_wqe *wqe = NULL;
+	bool metadata_ol =
+		txq->offloads & DEV_TX_OFFLOAD_MATCH_METADATA ? true : false;
 
 	assert(elts_n > pkts_n);
 	mlx5_tx_complete(txq);
@@ -125,6 +127,9 @@ 
 		uint16_t max_wqe;
 		__m128i *t_wqe, *dseg;
 		__m128i ctrl;
+		rte_be32_t metadata =
+			metadata_ol && (buf->ol_flags & PKT_TX_METADATA) ?
+			buf->tx_metadata : 0;
 
 		assert(segs_n);
 		max_elts = elts_n - (elts_head - txq->elts_tail);
@@ -165,9 +170,9 @@ 
 		_mm_store_si128(t_wqe, ctrl);
 		/* Fill ESEG in the header. */
 		_mm_store_si128(t_wqe + 1,
-				_mm_set_epi16(0, 0, 0, 0,
-					      rte_cpu_to_be_16(len), cs_flags,
-					      0, 0));
+				_mm_set_epi32(0, metadata,
+					      (rte_cpu_to_be_16(len) << 16) |
+					      cs_flags, 0));
 		txq->wqe_ci = wqe_ci;
 	}
 	if (!n)