DPDK patches and discussions
 help / color / mirror / Atom feed
* [dpdk-dev] [PATCH] net/i40e: fix Tx fn selection when using new ethdev offloads
@ 2018-05-01 13:03 Bruce Richardson
  2018-05-01 13:24 ` Ananyev, Konstantin
  2018-05-01 14:13 ` [dpdk-dev] [PATCH v2] " Bruce Richardson
  0 siblings, 2 replies; 12+ messages in thread
From: Bruce Richardson @ 2018-05-01 13:03 UTC (permalink / raw)
  To: Beilei Xing, Qi Zhang; +Cc: dev, ferruh.yigit, Bruce Richardson

The Tx function selection code in the driver only used the older txq
flags values to check whether the scalar or vector functions should be
used. This caused performance regressions with testpmd io-fwd as the
scalar path rather than the vector one was being used in the default
case. Fix this by changing the code to take account of new offloads and
deleting the defines used for the old ones.

Fixes: 7497d3e2f777 ("net/i40e: convert to new Tx offloads API")

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
---
 drivers/net/i40e/i40e_rxtx.c | 45 +++++++++++++++++++++++---------------------
 1 file changed, 24 insertions(+), 21 deletions(-)

diff --git a/drivers/net/i40e/i40e_rxtx.c b/drivers/net/i40e/i40e_rxtx.c
index ec1ce54ca..c523af575 100644
--- a/drivers/net/i40e/i40e_rxtx.c
+++ b/drivers/net/i40e/i40e_rxtx.c
@@ -40,9 +40,6 @@
 /* Base address of the HW descriptor ring should be 128B aligned. */
 #define I40E_RING_BASE_ALIGN	128
 
-#define I40E_SIMPLE_FLAGS ((uint32_t)ETH_TXQ_FLAGS_NOMULTSEGS | \
-					ETH_TXQ_FLAGS_NOOFFLOADS)
-
 #define I40E_TXD_CMD (I40E_TX_DESC_CMD_EOP | I40E_TX_DESC_CMD_RS)
 
 #ifdef RTE_LIBRTE_IEEE1588
@@ -70,6 +67,12 @@
 #define I40E_TX_OFFLOAD_NOTSUP_MASK \
 		(PKT_TX_OFFLOAD_MASK ^ I40E_TX_OFFLOAD_MASK)
 
+static const uint64_t i40e_simple_ol_mask = (DEV_TX_OFFLOAD_MULTI_SEGS |
+		DEV_TX_OFFLOAD_VLAN_INSERT |
+		DEV_TX_OFFLOAD_SCTP_CKSUM |
+		DEV_TX_OFFLOAD_UDP_CKSUM |
+		DEV_TX_OFFLOAD_TCP_CKSUM);
+
 static inline void
 i40e_rxd_to_vlan_tci(struct rte_mbuf *mb, volatile union i40e_rx_desc *rxdp)
 {
@@ -2108,11 +2111,9 @@ i40e_dev_tx_queue_setup_runtime(struct rte_eth_dev *dev,
 				 dev->data->nb_tx_queues)) {
 		/**
 		 * If it is the first queue to setup,
-		 * set all flags to default and call
+		 * set all flags and call
 		 * i40e_set_tx_function.
 		 */
-		ad->tx_simple_allowed = true;
-		ad->tx_vec_allowed = true;
 		i40e_set_tx_function_flag(dev, txq);
 		i40e_set_tx_function(dev);
 		return 0;
@@ -2128,9 +2129,8 @@ i40e_dev_tx_queue_setup_runtime(struct rte_eth_dev *dev,
 	}
 	/* check simple tx conflict */
 	if (ad->tx_simple_allowed) {
-		if (((txq->txq_flags & I40E_SIMPLE_FLAGS) !=
-		     I40E_SIMPLE_FLAGS) ||
-		    txq->tx_rs_thresh < RTE_PMD_I40E_TX_MAX_BURST) {
+		if ((txq->offloads & i40e_simple_ol_mask) != 0 ||
+				txq->tx_rs_thresh < RTE_PMD_I40E_TX_MAX_BURST) {
 			PMD_DRV_LOG(ERR, "No-simple tx is required.");
 			return -EINVAL;
 		}
@@ -3080,18 +3080,21 @@ i40e_set_tx_function_flag(struct rte_eth_dev *dev, struct i40e_tx_queue *txq)
 		I40E_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 
 	/* Use a simple Tx queue (no offloads, no multi segs) if possible */
-	if (((txq->txq_flags & I40E_SIMPLE_FLAGS) == I40E_SIMPLE_FLAGS)
-			&& (txq->tx_rs_thresh >= RTE_PMD_I40E_TX_MAX_BURST)) {
-		if (txq->tx_rs_thresh <= RTE_I40E_TX_MAX_FREE_BUF_SZ) {
-			PMD_INIT_LOG(DEBUG, "Vector tx"
-				     " can be enabled on this txq.");
-
-		} else {
-			ad->tx_vec_allowed = false;
-		}
-	} else {
-		ad->tx_simple_allowed = false;
-	}
+	ad->tx_simple_allowed = ((txq->offloads & i40e_simple_ol_mask) == 0 &&
+			txq->tx_rs_thresh >= RTE_PMD_I40E_TX_MAX_BURST);
+	ad->tx_vec_allowed = (ad->tx_simple_allowed &&
+			txq->tx_rs_thresh <= RTE_I40E_TX_MAX_FREE_BUF_SZ);
+
+	if (ad->tx_vec_allowed)
+		PMD_INIT_LOG(DEBUG, "Vector Tx can be enabled on Tx queue %u.",
+				txq->queue_id);
+	else if (ad->tx_simple_allowed)
+		PMD_INIT_LOG(DEBUG, "Simple Tx can be enabled on Tx queue %u.",
+				txq->queue_id);
+	else
+		PMD_INIT_LOG(DEBUG,
+				"Neither simple nor vector Tx enabled on Tx queue %u\n",
+				txq->queue_id);
 }
 
 void __attribute__((cold))
-- 
2.14.3

^ permalink raw reply	[flat|nested] 12+ messages in thread

* Re: [dpdk-dev] [PATCH] net/i40e: fix Tx fn selection when using new ethdev offloads
  2018-05-01 13:03 [dpdk-dev] [PATCH] net/i40e: fix Tx fn selection when using new ethdev offloads Bruce Richardson
@ 2018-05-01 13:24 ` Ananyev, Konstantin
  2018-05-01 13:28   ` Bruce Richardson
  2018-05-01 13:52   ` Bruce Richardson
  2018-05-01 14:13 ` [dpdk-dev] [PATCH v2] " Bruce Richardson
  1 sibling, 2 replies; 12+ messages in thread
From: Ananyev, Konstantin @ 2018-05-01 13:24 UTC (permalink / raw)
  To: Richardson, Bruce, Xing, Beilei, Zhang, Qi Z
  Cc: dev, Yigit, Ferruh, Richardson, Bruce

Hi Bruce,

> 
> The Tx function selection code in the driver only used the older txq
> flags values to check whether the scalar or vector functions should be
> used. This caused performance regressions with testpmd io-fwd as the
> scalar path rather than the vector one was being used in the default
> case. Fix this by changing the code to take account of new offloads and
> deleting the defines used for the old ones.
> 
> Fixes: 7497d3e2f777 ("net/i40e: convert to new Tx offloads API")
> 
> Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
> ---
>  drivers/net/i40e/i40e_rxtx.c | 45 +++++++++++++++++++++++---------------------
>  1 file changed, 24 insertions(+), 21 deletions(-)
> 
> diff --git a/drivers/net/i40e/i40e_rxtx.c b/drivers/net/i40e/i40e_rxtx.c
> index ec1ce54ca..c523af575 100644
> --- a/drivers/net/i40e/i40e_rxtx.c
> +++ b/drivers/net/i40e/i40e_rxtx.c
> @@ -40,9 +40,6 @@
>  /* Base address of the HW descriptor ring should be 128B aligned. */
>  #define I40E_RING_BASE_ALIGN	128
> 
> -#define I40E_SIMPLE_FLAGS ((uint32_t)ETH_TXQ_FLAGS_NOMULTSEGS | \
> -					ETH_TXQ_FLAGS_NOOFFLOADS)
> -
>  #define I40E_TXD_CMD (I40E_TX_DESC_CMD_EOP | I40E_TX_DESC_CMD_RS)
> 
>  #ifdef RTE_LIBRTE_IEEE1588
> @@ -70,6 +67,12 @@
>  #define I40E_TX_OFFLOAD_NOTSUP_MASK \
>  		(PKT_TX_OFFLOAD_MASK ^ I40E_TX_OFFLOAD_MASK)
> 
> +static const uint64_t i40e_simple_ol_mask = (DEV_TX_OFFLOAD_MULTI_SEGS |
> +		DEV_TX_OFFLOAD_VLAN_INSERT |
> +		DEV_TX_OFFLOAD_SCTP_CKSUM |
> +		DEV_TX_OFFLOAD_UDP_CKSUM |
> +		DEV_TX_OFFLOAD_TCP_CKSUM);
> +

Seems incomplete.
>From i40e_ethdev.c full-featured tx supports:
dev_info->tx_offload_capa =
                DEV_TX_OFFLOAD_VLAN_INSERT |
                DEV_TX_OFFLOAD_QINQ_INSERT |
                DEV_TX_OFFLOAD_IPV4_CKSUM |
                DEV_TX_OFFLOAD_UDP_CKSUM |
                DEV_TX_OFFLOAD_TCP_CKSUM |
                DEV_TX_OFFLOAD_SCTP_CKSUM |
                DEV_TX_OFFLOAD_OUTER_IPV4_CKSUM |
                DEV_TX_OFFLOAD_TCP_TSO |
                DEV_TX_OFFLOAD_VXLAN_TNL_TSO |
                DEV_TX_OFFLOAD_GRE_TNL_TSO |
                DEV_TX_OFFLOAD_IPIP_TNL_TSO |
                DEV_TX_OFFLOAD_GENEVE_TNL_TSO;

So we probably need the same here plus multiseg.
BTW, it is really strange that we don't have multiseg in tx_offload_capa.
Should be present I think.
Might be worse to create a new define for it, or just use dev_info->tx_offload_capa directly.
Konstantin



>  static inline void
>  i40e_rxd_to_vlan_tci(struct rte_mbuf *mb, volatile union i40e_rx_desc *rxdp)
>  {
> @@ -2108,11 +2111,9 @@ i40e_dev_tx_queue_setup_runtime(struct rte_eth_dev *dev,
>  				 dev->data->nb_tx_queues)) {
>  		/**
>  		 * If it is the first queue to setup,
> -		 * set all flags to default and call
> +		 * set all flags and call
>  		 * i40e_set_tx_function.
>  		 */
> -		ad->tx_simple_allowed = true;
> -		ad->tx_vec_allowed = true;
>  		i40e_set_tx_function_flag(dev, txq);
>  		i40e_set_tx_function(dev);
>  		return 0;
> @@ -2128,9 +2129,8 @@ i40e_dev_tx_queue_setup_runtime(struct rte_eth_dev *dev,
>  	}
>  	/* check simple tx conflict */
>  	if (ad->tx_simple_allowed) {
> -		if (((txq->txq_flags & I40E_SIMPLE_FLAGS) !=
> -		     I40E_SIMPLE_FLAGS) ||
> -		    txq->tx_rs_thresh < RTE_PMD_I40E_TX_MAX_BURST) {
> +		if ((txq->offloads & i40e_simple_ol_mask) != 0 ||
> +				txq->tx_rs_thresh < RTE_PMD_I40E_TX_MAX_BURST) {
>  			PMD_DRV_LOG(ERR, "No-simple tx is required.");
>  			return -EINVAL;
>  		}
> @@ -3080,18 +3080,21 @@ i40e_set_tx_function_flag(struct rte_eth_dev *dev, struct i40e_tx_queue *txq)
>  		I40E_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
> 
>  	/* Use a simple Tx queue (no offloads, no multi segs) if possible */
> -	if (((txq->txq_flags & I40E_SIMPLE_FLAGS) == I40E_SIMPLE_FLAGS)
> -			&& (txq->tx_rs_thresh >= RTE_PMD_I40E_TX_MAX_BURST)) {
> -		if (txq->tx_rs_thresh <= RTE_I40E_TX_MAX_FREE_BUF_SZ) {
> -			PMD_INIT_LOG(DEBUG, "Vector tx"
> -				     " can be enabled on this txq.");
> -
> -		} else {
> -			ad->tx_vec_allowed = false;
> -		}
> -	} else {
> -		ad->tx_simple_allowed = false;
> -	}
> +	ad->tx_simple_allowed = ((txq->offloads & i40e_simple_ol_mask) == 0 &&
> +			txq->tx_rs_thresh >= RTE_PMD_I40E_TX_MAX_BURST);
> +	ad->tx_vec_allowed = (ad->tx_simple_allowed &&
> +			txq->tx_rs_thresh <= RTE_I40E_TX_MAX_FREE_BUF_SZ);
> +
> +	if (ad->tx_vec_allowed)
> +		PMD_INIT_LOG(DEBUG, "Vector Tx can be enabled on Tx queue %u.",
> +				txq->queue_id);
> +	else if (ad->tx_simple_allowed)
> +		PMD_INIT_LOG(DEBUG, "Simple Tx can be enabled on Tx queue %u.",
> +				txq->queue_id);
> +	else
> +		PMD_INIT_LOG(DEBUG,
> +				"Neither simple nor vector Tx enabled on Tx queue %u\n",
> +				txq->queue_id);
>  }
> 
>  void __attribute__((cold))
> --
> 2.14.3

^ permalink raw reply	[flat|nested] 12+ messages in thread

* Re: [dpdk-dev] [PATCH] net/i40e: fix Tx fn selection when using new ethdev offloads
  2018-05-01 13:24 ` Ananyev, Konstantin
@ 2018-05-01 13:28   ` Bruce Richardson
  2018-05-01 13:52   ` Bruce Richardson
  1 sibling, 0 replies; 12+ messages in thread
From: Bruce Richardson @ 2018-05-01 13:28 UTC (permalink / raw)
  To: Ananyev, Konstantin; +Cc: Xing, Beilei, Zhang, Qi Z, dev, Yigit, Ferruh

On Tue, May 01, 2018 at 02:24:39PM +0100, Ananyev, Konstantin wrote:
> Hi Bruce,
> 
> > 
> > The Tx function selection code in the driver only used the older txq
> > flags values to check whether the scalar or vector functions should be
> > used. This caused performance regressions with testpmd io-fwd as the
> > scalar path rather than the vector one was being used in the default
> > case. Fix this by changing the code to take account of new offloads and
> > deleting the defines used for the old ones.
> > 
> > Fixes: 7497d3e2f777 ("net/i40e: convert to new Tx offloads API")
> > 
> > Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
> > ---
> >  drivers/net/i40e/i40e_rxtx.c | 45 +++++++++++++++++++++++---------------------
> >  1 file changed, 24 insertions(+), 21 deletions(-)
> > 
> > diff --git a/drivers/net/i40e/i40e_rxtx.c b/drivers/net/i40e/i40e_rxtx.c
> > index ec1ce54ca..c523af575 100644
> > --- a/drivers/net/i40e/i40e_rxtx.c
> > +++ b/drivers/net/i40e/i40e_rxtx.c
> > @@ -40,9 +40,6 @@
> >  /* Base address of the HW descriptor ring should be 128B aligned. */
> >  #define I40E_RING_BASE_ALIGN	128
> > 
> > -#define I40E_SIMPLE_FLAGS ((uint32_t)ETH_TXQ_FLAGS_NOMULTSEGS | \
> > -					ETH_TXQ_FLAGS_NOOFFLOADS)
> > -
> >  #define I40E_TXD_CMD (I40E_TX_DESC_CMD_EOP | I40E_TX_DESC_CMD_RS)
> > 
> >  #ifdef RTE_LIBRTE_IEEE1588
> > @@ -70,6 +67,12 @@
> >  #define I40E_TX_OFFLOAD_NOTSUP_MASK \
> >  		(PKT_TX_OFFLOAD_MASK ^ I40E_TX_OFFLOAD_MASK)
> > 
> > +static const uint64_t i40e_simple_ol_mask = (DEV_TX_OFFLOAD_MULTI_SEGS |
> > +		DEV_TX_OFFLOAD_VLAN_INSERT |
> > +		DEV_TX_OFFLOAD_SCTP_CKSUM |
> > +		DEV_TX_OFFLOAD_UDP_CKSUM |
> > +		DEV_TX_OFFLOAD_TCP_CKSUM);
> > +
> 
> Seems incomplete.
> From i40e_ethdev.c full-featured tx supports:
> dev_info->tx_offload_capa =
>                 DEV_TX_OFFLOAD_VLAN_INSERT |
>                 DEV_TX_OFFLOAD_QINQ_INSERT |
>                 DEV_TX_OFFLOAD_IPV4_CKSUM |
>                 DEV_TX_OFFLOAD_UDP_CKSUM |
>                 DEV_TX_OFFLOAD_TCP_CKSUM |
>                 DEV_TX_OFFLOAD_SCTP_CKSUM |
>                 DEV_TX_OFFLOAD_OUTER_IPV4_CKSUM |
>                 DEV_TX_OFFLOAD_TCP_TSO |
>                 DEV_TX_OFFLOAD_VXLAN_TNL_TSO |
>                 DEV_TX_OFFLOAD_GRE_TNL_TSO |
>                 DEV_TX_OFFLOAD_IPIP_TNL_TSO |
>                 DEV_TX_OFFLOAD_GENEVE_TNL_TSO;
> 
> So we probably need the same here plus multiseg.
> BTW, it is really strange that we don't have multiseg in tx_offload_capa.
> Should be present I think.
> Might be worse to create a new define for it, or just use dev_info->tx_offload_capa directly.
> Konstantin
> 
Thanks, good point, I never thought to check the advertised capabilities. I
just translated what was being used by the old code txq_flags translation 
function. I'll clean this up and do a V2.

/Bruce

^ permalink raw reply	[flat|nested] 12+ messages in thread

* Re: [dpdk-dev] [PATCH] net/i40e: fix Tx fn selection when using new ethdev offloads
  2018-05-01 13:24 ` Ananyev, Konstantin
  2018-05-01 13:28   ` Bruce Richardson
@ 2018-05-01 13:52   ` Bruce Richardson
  2018-05-01 14:11     ` Ananyev, Konstantin
  1 sibling, 1 reply; 12+ messages in thread
From: Bruce Richardson @ 2018-05-01 13:52 UTC (permalink / raw)
  To: Ananyev, Konstantin; +Cc: Xing, Beilei, Zhang, Qi Z, dev, Yigit, Ferruh

On Tue, May 01, 2018 at 02:24:39PM +0100, Ananyev, Konstantin wrote:
> Hi Bruce,
> 
> > 
> > The Tx function selection code in the driver only used the older txq
> > flags values to check whether the scalar or vector functions should be
> > used. This caused performance regressions with testpmd io-fwd as the
> > scalar path rather than the vector one was being used in the default
> > case. Fix this by changing the code to take account of new offloads and
> > deleting the defines used for the old ones.
> > 
> > Fixes: 7497d3e2f777 ("net/i40e: convert to new Tx offloads API")
> > 
> > Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
> > ---
> >  drivers/net/i40e/i40e_rxtx.c | 45 +++++++++++++++++++++++---------------------
> >  1 file changed, 24 insertions(+), 21 deletions(-)
> > 
> > diff --git a/drivers/net/i40e/i40e_rxtx.c b/drivers/net/i40e/i40e_rxtx.c
> > index ec1ce54ca..c523af575 100644
> > --- a/drivers/net/i40e/i40e_rxtx.c
> > +++ b/drivers/net/i40e/i40e_rxtx.c
> > @@ -40,9 +40,6 @@
> >  /* Base address of the HW descriptor ring should be 128B aligned. */
> >  #define I40E_RING_BASE_ALIGN	128
> > 
> > -#define I40E_SIMPLE_FLAGS ((uint32_t)ETH_TXQ_FLAGS_NOMULTSEGS | \
> > -					ETH_TXQ_FLAGS_NOOFFLOADS)
> > -
> >  #define I40E_TXD_CMD (I40E_TX_DESC_CMD_EOP | I40E_TX_DESC_CMD_RS)
> > 
> >  #ifdef RTE_LIBRTE_IEEE1588
> > @@ -70,6 +67,12 @@
> >  #define I40E_TX_OFFLOAD_NOTSUP_MASK \
> >  		(PKT_TX_OFFLOAD_MASK ^ I40E_TX_OFFLOAD_MASK)
> > 
> > +static const uint64_t i40e_simple_ol_mask = (DEV_TX_OFFLOAD_MULTI_SEGS |
> > +		DEV_TX_OFFLOAD_VLAN_INSERT |
> > +		DEV_TX_OFFLOAD_SCTP_CKSUM |
> > +		DEV_TX_OFFLOAD_UDP_CKSUM |
> > +		DEV_TX_OFFLOAD_TCP_CKSUM);
> > +
> 
> Seems incomplete.
> From i40e_ethdev.c full-featured tx supports:
> dev_info->tx_offload_capa =
>                 DEV_TX_OFFLOAD_VLAN_INSERT |
>                 DEV_TX_OFFLOAD_QINQ_INSERT |
>                 DEV_TX_OFFLOAD_IPV4_CKSUM |
>                 DEV_TX_OFFLOAD_UDP_CKSUM |
>                 DEV_TX_OFFLOAD_TCP_CKSUM |
>                 DEV_TX_OFFLOAD_SCTP_CKSUM |
>                 DEV_TX_OFFLOAD_OUTER_IPV4_CKSUM |
>                 DEV_TX_OFFLOAD_TCP_TSO |
>                 DEV_TX_OFFLOAD_VXLAN_TNL_TSO |
>                 DEV_TX_OFFLOAD_GRE_TNL_TSO |
>                 DEV_TX_OFFLOAD_IPIP_TNL_TSO |
>                 DEV_TX_OFFLOAD_GENEVE_TNL_TSO;
> 
> So we probably need the same here plus multiseg.
> BTW, it is really strange that we don't have multiseg in tx_offload_capa.
> Should be present I think.
> Might be worse to create a new define for it, or just use dev_info->tx_offload_capa directly.
> Konstantin
> 
Thinking about this more, it seems that right now we don't need a masks at
all. Any bits set in the offloads is going to cause us to use the scalar
path or to error out with an invalid offload requested. Yes, it's not
future-proofed in that it will need to be changed if we do end up
supporting some offloads with the vector path in future, but then the same
problem would occur if we just re-use the advertised capabilities too, like
you suggest.

Therefore I think for V2 we'll just check for a non-zero offloads value.

/Bruce

^ permalink raw reply	[flat|nested] 12+ messages in thread

* Re: [dpdk-dev] [PATCH] net/i40e: fix Tx fn selection when using new ethdev offloads
  2018-05-01 13:52   ` Bruce Richardson
@ 2018-05-01 14:11     ` Ananyev, Konstantin
  0 siblings, 0 replies; 12+ messages in thread
From: Ananyev, Konstantin @ 2018-05-01 14:11 UTC (permalink / raw)
  To: Richardson, Bruce; +Cc: Xing, Beilei, Zhang, Qi Z, dev, Yigit, Ferruh



> -----Original Message-----
> From: Richardson, Bruce
> Sent: Tuesday, May 1, 2018 2:53 PM
> To: Ananyev, Konstantin <konstantin.ananyev@intel.com>
> Cc: Xing, Beilei <beilei.xing@intel.com>; Zhang, Qi Z <qi.z.zhang@intel.com>; dev@dpdk.org; Yigit, Ferruh <ferruh.yigit@intel.com>
> Subject: Re: [dpdk-dev] [PATCH] net/i40e: fix Tx fn selection when using new ethdev offloads
> 
> On Tue, May 01, 2018 at 02:24:39PM +0100, Ananyev, Konstantin wrote:
> > Hi Bruce,
> >
> > >
> > > The Tx function selection code in the driver only used the older txq
> > > flags values to check whether the scalar or vector functions should be
> > > used. This caused performance regressions with testpmd io-fwd as the
> > > scalar path rather than the vector one was being used in the default
> > > case. Fix this by changing the code to take account of new offloads and
> > > deleting the defines used for the old ones.
> > >
> > > Fixes: 7497d3e2f777 ("net/i40e: convert to new Tx offloads API")
> > >
> > > Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
> > > ---
> > >  drivers/net/i40e/i40e_rxtx.c | 45 +++++++++++++++++++++++---------------------
> > >  1 file changed, 24 insertions(+), 21 deletions(-)
> > >
> > > diff --git a/drivers/net/i40e/i40e_rxtx.c b/drivers/net/i40e/i40e_rxtx.c
> > > index ec1ce54ca..c523af575 100644
> > > --- a/drivers/net/i40e/i40e_rxtx.c
> > > +++ b/drivers/net/i40e/i40e_rxtx.c
> > > @@ -40,9 +40,6 @@
> > >  /* Base address of the HW descriptor ring should be 128B aligned. */
> > >  #define I40E_RING_BASE_ALIGN	128
> > >
> > > -#define I40E_SIMPLE_FLAGS ((uint32_t)ETH_TXQ_FLAGS_NOMULTSEGS | \
> > > -					ETH_TXQ_FLAGS_NOOFFLOADS)
> > > -
> > >  #define I40E_TXD_CMD (I40E_TX_DESC_CMD_EOP | I40E_TX_DESC_CMD_RS)
> > >
> > >  #ifdef RTE_LIBRTE_IEEE1588
> > > @@ -70,6 +67,12 @@
> > >  #define I40E_TX_OFFLOAD_NOTSUP_MASK \
> > >  		(PKT_TX_OFFLOAD_MASK ^ I40E_TX_OFFLOAD_MASK)
> > >
> > > +static const uint64_t i40e_simple_ol_mask = (DEV_TX_OFFLOAD_MULTI_SEGS |
> > > +		DEV_TX_OFFLOAD_VLAN_INSERT |
> > > +		DEV_TX_OFFLOAD_SCTP_CKSUM |
> > > +		DEV_TX_OFFLOAD_UDP_CKSUM |
> > > +		DEV_TX_OFFLOAD_TCP_CKSUM);
> > > +
> >
> > Seems incomplete.
> > From i40e_ethdev.c full-featured tx supports:
> > dev_info->tx_offload_capa =
> >                 DEV_TX_OFFLOAD_VLAN_INSERT |
> >                 DEV_TX_OFFLOAD_QINQ_INSERT |
> >                 DEV_TX_OFFLOAD_IPV4_CKSUM |
> >                 DEV_TX_OFFLOAD_UDP_CKSUM |
> >                 DEV_TX_OFFLOAD_TCP_CKSUM |
> >                 DEV_TX_OFFLOAD_SCTP_CKSUM |
> >                 DEV_TX_OFFLOAD_OUTER_IPV4_CKSUM |
> >                 DEV_TX_OFFLOAD_TCP_TSO |
> >                 DEV_TX_OFFLOAD_VXLAN_TNL_TSO |
> >                 DEV_TX_OFFLOAD_GRE_TNL_TSO |
> >                 DEV_TX_OFFLOAD_IPIP_TNL_TSO |
> >                 DEV_TX_OFFLOAD_GENEVE_TNL_TSO;
> >
> > So we probably need the same here plus multiseg.
> > BTW, it is really strange that we don't have multiseg in tx_offload_capa.
> > Should be present I think.
> > Might be worse to create a new define for it, or just use dev_info->tx_offload_capa directly.
> > Konstantin
> >
> Thinking about this more, it seems that right now we don't need a masks at
> all. Any bits set in the offloads is going to cause us to use the scalar
> path or to error out with an invalid offload requested. Yes, it's not
> future-proofed in that it will need to be changed if we do end up
> supporting some offloads with the vector path in future, but then the same
> problem would occur if we just re-use the advertised capabilities too, like
> you suggest.
> 
> Therefore I think for V2 we'll just check for a non-zero offloads value.

Ok by me.
Konstantin

^ permalink raw reply	[flat|nested] 12+ messages in thread

* [dpdk-dev] [PATCH v2] net/i40e: fix Tx fn selection when using new ethdev offloads
  2018-05-01 13:03 [dpdk-dev] [PATCH] net/i40e: fix Tx fn selection when using new ethdev offloads Bruce Richardson
  2018-05-01 13:24 ` Ananyev, Konstantin
@ 2018-05-01 14:13 ` Bruce Richardson
  2018-05-01 14:16   ` Bruce Richardson
  2018-05-01 17:52   ` Ananyev, Konstantin
  1 sibling, 2 replies; 12+ messages in thread
From: Bruce Richardson @ 2018-05-01 14:13 UTC (permalink / raw)
  To: Beilei Xing, Qi Zhang
  Cc: dev, ferruh.yigit, konstantin.ananyev, Bruce Richardson

The Tx function selection code in the driver only used the older txq
flags values to check whether the scalar or vector functions should be
used. This caused performance regressions with testpmd io-fwd as the
scalar path rather than the vector one was being used in the default
case. Fix this by changing the code to take account of new offloads and
deleting the defines used for the old ones.

Fixes: 7497d3e2f777 ("net/i40e: convert to new Tx offloads API")

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
---
 drivers/net/i40e/i40e_rxtx.c | 39 ++++++++++++++++++---------------------
 1 file changed, 18 insertions(+), 21 deletions(-)

diff --git a/drivers/net/i40e/i40e_rxtx.c b/drivers/net/i40e/i40e_rxtx.c
index ec1ce54ca..006f5b846 100644
--- a/drivers/net/i40e/i40e_rxtx.c
+++ b/drivers/net/i40e/i40e_rxtx.c
@@ -40,9 +40,6 @@
 /* Base address of the HW descriptor ring should be 128B aligned. */
 #define I40E_RING_BASE_ALIGN	128
 
-#define I40E_SIMPLE_FLAGS ((uint32_t)ETH_TXQ_FLAGS_NOMULTSEGS | \
-					ETH_TXQ_FLAGS_NOOFFLOADS)
-
 #define I40E_TXD_CMD (I40E_TX_DESC_CMD_EOP | I40E_TX_DESC_CMD_RS)
 
 #ifdef RTE_LIBRTE_IEEE1588
@@ -2108,11 +2105,9 @@ i40e_dev_tx_queue_setup_runtime(struct rte_eth_dev *dev,
 				 dev->data->nb_tx_queues)) {
 		/**
 		 * If it is the first queue to setup,
-		 * set all flags to default and call
+		 * set all flags and call
 		 * i40e_set_tx_function.
 		 */
-		ad->tx_simple_allowed = true;
-		ad->tx_vec_allowed = true;
 		i40e_set_tx_function_flag(dev, txq);
 		i40e_set_tx_function(dev);
 		return 0;
@@ -2128,9 +2123,8 @@ i40e_dev_tx_queue_setup_runtime(struct rte_eth_dev *dev,
 	}
 	/* check simple tx conflict */
 	if (ad->tx_simple_allowed) {
-		if (((txq->txq_flags & I40E_SIMPLE_FLAGS) !=
-		     I40E_SIMPLE_FLAGS) ||
-		    txq->tx_rs_thresh < RTE_PMD_I40E_TX_MAX_BURST) {
+		if (txq->offloads != 0 ||
+				txq->tx_rs_thresh < RTE_PMD_I40E_TX_MAX_BURST) {
 			PMD_DRV_LOG(ERR, "No-simple tx is required.");
 			return -EINVAL;
 		}
@@ -3080,18 +3074,21 @@ i40e_set_tx_function_flag(struct rte_eth_dev *dev, struct i40e_tx_queue *txq)
 		I40E_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 
 	/* Use a simple Tx queue (no offloads, no multi segs) if possible */
-	if (((txq->txq_flags & I40E_SIMPLE_FLAGS) == I40E_SIMPLE_FLAGS)
-			&& (txq->tx_rs_thresh >= RTE_PMD_I40E_TX_MAX_BURST)) {
-		if (txq->tx_rs_thresh <= RTE_I40E_TX_MAX_FREE_BUF_SZ) {
-			PMD_INIT_LOG(DEBUG, "Vector tx"
-				     " can be enabled on this txq.");
-
-		} else {
-			ad->tx_vec_allowed = false;
-		}
-	} else {
-		ad->tx_simple_allowed = false;
-	}
+	ad->tx_simple_allowed = (txq->offloads == 0 &&
+			txq->tx_rs_thresh >= RTE_PMD_I40E_TX_MAX_BURST);
+	ad->tx_vec_allowed = (ad->tx_simple_allowed &&
+			txq->tx_rs_thresh <= RTE_I40E_TX_MAX_FREE_BUF_SZ);
+
+	if (ad->tx_vec_allowed)
+		PMD_INIT_LOG(DEBUG, "Vector Tx can be enabled on Tx queue %u.",
+				txq->queue_id);
+	else if (ad->tx_simple_allowed)
+		PMD_INIT_LOG(DEBUG, "Simple Tx can be enabled on Tx queue %u.",
+				txq->queue_id);
+	else
+		PMD_INIT_LOG(DEBUG,
+				"Neither simple nor vector Tx enabled on Tx queue %u\n",
+				txq->queue_id);
 }
 
 void __attribute__((cold))
-- 
2.14.3

^ permalink raw reply	[flat|nested] 12+ messages in thread

* Re: [dpdk-dev] [PATCH v2] net/i40e: fix Tx fn selection when using new ethdev offloads
  2018-05-01 14:13 ` [dpdk-dev] [PATCH v2] " Bruce Richardson
@ 2018-05-01 14:16   ` Bruce Richardson
  2018-05-01 14:37     ` Ferruh Yigit
  2018-05-01 17:52   ` Ananyev, Konstantin
  1 sibling, 1 reply; 12+ messages in thread
From: Bruce Richardson @ 2018-05-01 14:16 UTC (permalink / raw)
  To: Beilei Xing, Qi Zhang; +Cc: dev, ferruh.yigit, konstantin.ananyev

On Tue, May 01, 2018 at 03:13:54PM +0100, Bruce Richardson wrote:
> The Tx function selection code in the driver only used the older txq
> flags values to check whether the scalar or vector functions should be
> used. This caused performance regressions with testpmd io-fwd as the
> scalar path rather than the vector one was being used in the default
> case. Fix this by changing the code to take account of new offloads and
> deleting the defines used for the old ones.
> 
> Fixes: 7497d3e2f777 ("net/i40e: convert to new Tx offloads API")
> 
> Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
> ---
Apologies: forgot to add:

v2: eliminate mask for offload flags, and use vector path only if 
    offloads == 0

^ permalink raw reply	[flat|nested] 12+ messages in thread

* Re: [dpdk-dev] [PATCH v2] net/i40e: fix Tx fn selection when using new ethdev offloads
  2018-05-01 14:16   ` Bruce Richardson
@ 2018-05-01 14:37     ` Ferruh Yigit
  2018-05-01 15:48       ` Ferruh Yigit
  0 siblings, 1 reply; 12+ messages in thread
From: Ferruh Yigit @ 2018-05-01 14:37 UTC (permalink / raw)
  To: Bruce Richardson, Beilei Xing, Qi Zhang; +Cc: dev, konstantin.ananyev

On 5/1/2018 3:16 PM, Bruce Richardson wrote:
> On Tue, May 01, 2018 at 03:13:54PM +0100, Bruce Richardson wrote:
>> The Tx function selection code in the driver only used the older txq
>> flags values to check whether the scalar or vector functions should be
>> used. This caused performance regressions with testpmd io-fwd as the
>> scalar path rather than the vector one was being used in the default
>> case. Fix this by changing the code to take account of new offloads and
>> deleting the defines used for the old ones.
>>
>> Fixes: 7497d3e2f777 ("net/i40e: convert to new Tx offloads API")
>>
>> Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
>> ---
> Apologies: forgot to add:
> 
> v2: eliminate mask for offload flags, and use vector path only if 
>     offloads == 0
> 

Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>

^ permalink raw reply	[flat|nested] 12+ messages in thread

* Re: [dpdk-dev] [PATCH v2] net/i40e: fix Tx fn selection when using new ethdev offloads
  2018-05-01 14:37     ` Ferruh Yigit
@ 2018-05-01 15:48       ` Ferruh Yigit
  0 siblings, 0 replies; 12+ messages in thread
From: Ferruh Yigit @ 2018-05-01 15:48 UTC (permalink / raw)
  To: Bruce Richardson, Beilei Xing, Qi Zhang; +Cc: dev, konstantin.ananyev

On 5/1/2018 3:37 PM, Ferruh Yigit wrote:
> On 5/1/2018 3:16 PM, Bruce Richardson wrote:
>> On Tue, May 01, 2018 at 03:13:54PM +0100, Bruce Richardson wrote:
>>> The Tx function selection code in the driver only used the older txq
>>> flags values to check whether the scalar or vector functions should be
>>> used. This caused performance regressions with testpmd io-fwd as the
>>> scalar path rather than the vector one was being used in the default
>>> case. Fix this by changing the code to take account of new offloads and
>>> deleting the defines used for the old ones.
>>>
>>> Fixes: 7497d3e2f777 ("net/i40e: convert to new Tx offloads API")
>>>
>>> Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
>>> ---
>> Apologies: forgot to add:
>>
>> v2: eliminate mask for offload flags, and use vector path only if 
>>     offloads == 0
>>
> 
> Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>

Applied to dpdk-next-net/master, thanks.

^ permalink raw reply	[flat|nested] 12+ messages in thread

* Re: [dpdk-dev] [PATCH v2] net/i40e: fix Tx fn selection when using new ethdev offloads
  2018-05-01 14:13 ` [dpdk-dev] [PATCH v2] " Bruce Richardson
  2018-05-01 14:16   ` Bruce Richardson
@ 2018-05-01 17:52   ` Ananyev, Konstantin
  2018-05-02  8:24     ` Bruce Richardson
  1 sibling, 1 reply; 12+ messages in thread
From: Ananyev, Konstantin @ 2018-05-01 17:52 UTC (permalink / raw)
  To: Richardson, Bruce, Xing, Beilei, Zhang, Qi Z; +Cc: dev, Yigit, Ferruh



> -----Original Message-----
> From: Richardson, Bruce
> Sent: Tuesday, May 1, 2018 3:14 PM
> To: Xing, Beilei <beilei.xing@intel.com>; Zhang, Qi Z <qi.z.zhang@intel.com>
> Cc: dev@dpdk.org; Yigit, Ferruh <ferruh.yigit@intel.com>; Ananyev, Konstantin <konstantin.ananyev@intel.com>; Richardson,
> Bruce <bruce.richardson@intel.com>
> Subject: [PATCH v2] net/i40e: fix Tx fn selection when using new ethdev offloads
> 
> The Tx function selection code in the driver only used the older txq
> flags values to check whether the scalar or vector functions should be
> used. This caused performance regressions with testpmd io-fwd as the
> scalar path rather than the vector one was being used in the default
> case. Fix this by changing the code to take account of new offloads and
> deleting the defines used for the old ones.
> 
> Fixes: 7497d3e2f777 ("net/i40e: convert to new Tx offloads API")
> 
> Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
> ---
>  drivers/net/i40e/i40e_rxtx.c | 39 ++++++++++++++++++---------------------
>  1 file changed, 18 insertions(+), 21 deletions(-)
> 
> diff --git a/drivers/net/i40e/i40e_rxtx.c b/drivers/net/i40e/i40e_rxtx.c
> index ec1ce54ca..006f5b846 100644
> --- a/drivers/net/i40e/i40e_rxtx.c
> +++ b/drivers/net/i40e/i40e_rxtx.c
> @@ -40,9 +40,6 @@
>  /* Base address of the HW descriptor ring should be 128B aligned. */
>  #define I40E_RING_BASE_ALIGN	128
> 
> -#define I40E_SIMPLE_FLAGS ((uint32_t)ETH_TXQ_FLAGS_NOMULTSEGS | \
> -					ETH_TXQ_FLAGS_NOOFFLOADS)
> -
>  #define I40E_TXD_CMD (I40E_TX_DESC_CMD_EOP | I40E_TX_DESC_CMD_RS)
> 
>  #ifdef RTE_LIBRTE_IEEE1588
> @@ -2108,11 +2105,9 @@ i40e_dev_tx_queue_setup_runtime(struct rte_eth_dev *dev,
>  				 dev->data->nb_tx_queues)) {
>  		/**
>  		 * If it is the first queue to setup,
> -		 * set all flags to default and call
> +		 * set all flags and call
>  		 * i40e_set_tx_function.
>  		 */
> -		ad->tx_simple_allowed = true;
> -		ad->tx_vec_allowed = true;
>  		i40e_set_tx_function_flag(dev, txq);
>  		i40e_set_tx_function(dev);
>  		return 0;
> @@ -2128,9 +2123,8 @@ i40e_dev_tx_queue_setup_runtime(struct rte_eth_dev *dev,
>  	}
>  	/* check simple tx conflict */
>  	if (ad->tx_simple_allowed) {
> -		if (((txq->txq_flags & I40E_SIMPLE_FLAGS) !=
> -		     I40E_SIMPLE_FLAGS) ||
> -		    txq->tx_rs_thresh < RTE_PMD_I40E_TX_MAX_BURST) {
> +		if (txq->offloads != 0 ||
> +				txq->tx_rs_thresh < RTE_PMD_I40E_TX_MAX_BURST) {
>  			PMD_DRV_LOG(ERR, "No-simple tx is required.");
>  			return -EINVAL;
>  		}
> @@ -3080,18 +3074,21 @@ i40e_set_tx_function_flag(struct rte_eth_dev *dev, struct i40e_tx_queue *txq)
>  		I40E_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
> 
>  	/* Use a simple Tx queue (no offloads, no multi segs) if possible */
> -	if (((txq->txq_flags & I40E_SIMPLE_FLAGS) == I40E_SIMPLE_FLAGS)
> -			&& (txq->tx_rs_thresh >= RTE_PMD_I40E_TX_MAX_BURST)) {
> -		if (txq->tx_rs_thresh <= RTE_I40E_TX_MAX_FREE_BUF_SZ) {
> -			PMD_INIT_LOG(DEBUG, "Vector tx"
> -				     " can be enabled on this txq.");
> -
> -		} else {
> -			ad->tx_vec_allowed = false;
> -		}
> -	} else {
> -		ad->tx_simple_allowed = false;
> -	}
> +	ad->tx_simple_allowed = (txq->offloads == 0 &&
> +			txq->tx_rs_thresh >= RTE_PMD_I40E_TX_MAX_BURST);

Actually after another thought - who setup txq->offloads?
I did a quick scan, through i40e code and seems no one does.
So now it seems not possible to enable TX offloads at all.
Konstantin

BTW, seems like rxq->offloads are not properly initialised too.

> +	ad->tx_vec_allowed = (ad->tx_simple_allowed &&
> +			txq->tx_rs_thresh <= RTE_I40E_TX_MAX_FREE_BUF_SZ);
> +
> +	if (ad->tx_vec_allowed)
> +		PMD_INIT_LOG(DEBUG, "Vector Tx can be enabled on Tx queue %u.",
> +				txq->queue_id);
> +	else if (ad->tx_simple_allowed)
> +		PMD_INIT_LOG(DEBUG, "Simple Tx can be enabled on Tx queue %u.",
> +				txq->queue_id);
> +	else
> +		PMD_INIT_LOG(DEBUG,
> +				"Neither simple nor vector Tx enabled on Tx queue %u\n",
> +				txq->queue_id);
>  }
> 
>  void __attribute__((cold))
> --
> 2.14.3

^ permalink raw reply	[flat|nested] 12+ messages in thread

* Re: [dpdk-dev] [PATCH v2] net/i40e: fix Tx fn selection when using new ethdev offloads
  2018-05-01 17:52   ` Ananyev, Konstantin
@ 2018-05-02  8:24     ` Bruce Richardson
  2018-05-02  8:30       ` Zhang, Qi Z
  0 siblings, 1 reply; 12+ messages in thread
From: Bruce Richardson @ 2018-05-02  8:24 UTC (permalink / raw)
  To: Ananyev, Konstantin; +Cc: Xing, Beilei, Zhang, Qi Z, dev, Yigit, Ferruh

On Tue, May 01, 2018 at 06:52:18PM +0100, Ananyev, Konstantin wrote:
> 
> 
> > -----Original Message-----
> > From: Richardson, Bruce
> > Sent: Tuesday, May 1, 2018 3:14 PM
> > To: Xing, Beilei <beilei.xing@intel.com>; Zhang, Qi Z <qi.z.zhang@intel.com>
> > Cc: dev@dpdk.org; Yigit, Ferruh <ferruh.yigit@intel.com>; Ananyev, Konstantin <konstantin.ananyev@intel.com>; Richardson,
> > Bruce <bruce.richardson@intel.com>
> > Subject: [PATCH v2] net/i40e: fix Tx fn selection when using new ethdev offloads
> > 
> > The Tx function selection code in the driver only used the older txq
> > flags values to check whether the scalar or vector functions should be
> > used. This caused performance regressions with testpmd io-fwd as the
> > scalar path rather than the vector one was being used in the default
> > case. Fix this by changing the code to take account of new offloads and
> > deleting the defines used for the old ones.
> > 
> > Fixes: 7497d3e2f777 ("net/i40e: convert to new Tx offloads API")
> > 
> > Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
> > ---
> >  drivers/net/i40e/i40e_rxtx.c | 39 ++++++++++++++++++---------------------
> >  1 file changed, 18 insertions(+), 21 deletions(-)
> > 
> > diff --git a/drivers/net/i40e/i40e_rxtx.c b/drivers/net/i40e/i40e_rxtx.c
> > index ec1ce54ca..006f5b846 100644
> > --- a/drivers/net/i40e/i40e_rxtx.c
> > +++ b/drivers/net/i40e/i40e_rxtx.c
> > @@ -40,9 +40,6 @@
> >  /* Base address of the HW descriptor ring should be 128B aligned. */
> >  #define I40E_RING_BASE_ALIGN	128
> > 
> > -#define I40E_SIMPLE_FLAGS ((uint32_t)ETH_TXQ_FLAGS_NOMULTSEGS | \
> > -					ETH_TXQ_FLAGS_NOOFFLOADS)
> > -
> >  #define I40E_TXD_CMD (I40E_TX_DESC_CMD_EOP | I40E_TX_DESC_CMD_RS)
> > 
> >  #ifdef RTE_LIBRTE_IEEE1588
> > @@ -2108,11 +2105,9 @@ i40e_dev_tx_queue_setup_runtime(struct rte_eth_dev *dev,
> >  				 dev->data->nb_tx_queues)) {
> >  		/**
> >  		 * If it is the first queue to setup,
> > -		 * set all flags to default and call
> > +		 * set all flags and call
> >  		 * i40e_set_tx_function.
> >  		 */
> > -		ad->tx_simple_allowed = true;
> > -		ad->tx_vec_allowed = true;
> >  		i40e_set_tx_function_flag(dev, txq);
> >  		i40e_set_tx_function(dev);
> >  		return 0;
> > @@ -2128,9 +2123,8 @@ i40e_dev_tx_queue_setup_runtime(struct rte_eth_dev *dev,
> >  	}
> >  	/* check simple tx conflict */
> >  	if (ad->tx_simple_allowed) {
> > -		if (((txq->txq_flags & I40E_SIMPLE_FLAGS) !=
> > -		     I40E_SIMPLE_FLAGS) ||
> > -		    txq->tx_rs_thresh < RTE_PMD_I40E_TX_MAX_BURST) {
> > +		if (txq->offloads != 0 ||
> > +				txq->tx_rs_thresh < RTE_PMD_I40E_TX_MAX_BURST) {
> >  			PMD_DRV_LOG(ERR, "No-simple tx is required.");
> >  			return -EINVAL;
> >  		}
> > @@ -3080,18 +3074,21 @@ i40e_set_tx_function_flag(struct rte_eth_dev *dev, struct i40e_tx_queue *txq)
> >  		I40E_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
> > 
> >  	/* Use a simple Tx queue (no offloads, no multi segs) if possible */
> > -	if (((txq->txq_flags & I40E_SIMPLE_FLAGS) == I40E_SIMPLE_FLAGS)
> > -			&& (txq->tx_rs_thresh >= RTE_PMD_I40E_TX_MAX_BURST)) {
> > -		if (txq->tx_rs_thresh <= RTE_I40E_TX_MAX_FREE_BUF_SZ) {
> > -			PMD_INIT_LOG(DEBUG, "Vector tx"
> > -				     " can be enabled on this txq.");
> > -
> > -		} else {
> > -			ad->tx_vec_allowed = false;
> > -		}
> > -	} else {
> > -		ad->tx_simple_allowed = false;
> > -	}
> > +	ad->tx_simple_allowed = (txq->offloads == 0 &&
> > +			txq->tx_rs_thresh >= RTE_PMD_I40E_TX_MAX_BURST);
> 
> Actually after another thought - who setup txq->offloads?
> I did a quick scan, through i40e code and seems no one does.
> So now it seems not possible to enable TX offloads at all.
> Konstantin
> 
> BTW, seems like rxq->offloads are not properly initialised too.
> 
The offloads value should come from the app, no?

^ permalink raw reply	[flat|nested] 12+ messages in thread

* Re: [dpdk-dev] [PATCH v2] net/i40e: fix Tx fn selection when using new ethdev offloads
  2018-05-02  8:24     ` Bruce Richardson
@ 2018-05-02  8:30       ` Zhang, Qi Z
  0 siblings, 0 replies; 12+ messages in thread
From: Zhang, Qi Z @ 2018-05-02  8:30 UTC (permalink / raw)
  To: Richardson, Bruce, Ananyev, Konstantin; +Cc: Xing, Beilei, dev, Yigit, Ferruh



> -----Original Message-----
> From: Richardson, Bruce
> Sent: Wednesday, May 2, 2018 4:25 PM
> To: Ananyev, Konstantin <konstantin.ananyev@intel.com>
> Cc: Xing, Beilei <beilei.xing@intel.com>; Zhang, Qi Z <qi.z.zhang@intel.com>;
> dev@dpdk.org; Yigit, Ferruh <ferruh.yigit@intel.com>
> Subject: Re: [PATCH v2] net/i40e: fix Tx fn selection when using new ethdev
> offloads
> 
> On Tue, May 01, 2018 at 06:52:18PM +0100, Ananyev, Konstantin wrote:
> >
> >
> > > -----Original Message-----
> > > From: Richardson, Bruce
> > > Sent: Tuesday, May 1, 2018 3:14 PM
> > > To: Xing, Beilei <beilei.xing@intel.com>; Zhang, Qi Z
> > > <qi.z.zhang@intel.com>
> > > Cc: dev@dpdk.org; Yigit, Ferruh <ferruh.yigit@intel.com>; Ananyev,
> > > Konstantin <konstantin.ananyev@intel.com>; Richardson, Bruce
> > > <bruce.richardson@intel.com>
> > > Subject: [PATCH v2] net/i40e: fix Tx fn selection when using new
> > > ethdev offloads
> > >
> > > The Tx function selection code in the driver only used the older txq
> > > flags values to check whether the scalar or vector functions should
> > > be used. This caused performance regressions with testpmd io-fwd as
> > > the scalar path rather than the vector one was being used in the
> > > default case. Fix this by changing the code to take account of new
> > > offloads and deleting the defines used for the old ones.
> > >
> > > Fixes: 7497d3e2f777 ("net/i40e: convert to new Tx offloads API")
> > >
> > > Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
> > > ---
> > >  drivers/net/i40e/i40e_rxtx.c | 39
> > > ++++++++++++++++++---------------------
> > >  1 file changed, 18 insertions(+), 21 deletions(-)
> > >
> > > diff --git a/drivers/net/i40e/i40e_rxtx.c
> > > b/drivers/net/i40e/i40e_rxtx.c index ec1ce54ca..006f5b846 100644
> > > --- a/drivers/net/i40e/i40e_rxtx.c
> > > +++ b/drivers/net/i40e/i40e_rxtx.c
> > > @@ -40,9 +40,6 @@
> > >  /* Base address of the HW descriptor ring should be 128B aligned. */
> > >  #define I40E_RING_BASE_ALIGN	128
> > >
> > > -#define I40E_SIMPLE_FLAGS ((uint32_t)ETH_TXQ_FLAGS_NOMULTSEGS
> | \
> > > -					ETH_TXQ_FLAGS_NOOFFLOADS)
> > > -
> > >  #define I40E_TXD_CMD (I40E_TX_DESC_CMD_EOP |
> I40E_TX_DESC_CMD_RS)
> > >
> > >  #ifdef RTE_LIBRTE_IEEE1588
> > > @@ -2108,11 +2105,9 @@ i40e_dev_tx_queue_setup_runtime(struct
> rte_eth_dev *dev,
> > >  				 dev->data->nb_tx_queues)) {
> > >  		/**
> > >  		 * If it is the first queue to setup,
> > > -		 * set all flags to default and call
> > > +		 * set all flags and call
> > >  		 * i40e_set_tx_function.
> > >  		 */
> > > -		ad->tx_simple_allowed = true;
> > > -		ad->tx_vec_allowed = true;
> > >  		i40e_set_tx_function_flag(dev, txq);
> > >  		i40e_set_tx_function(dev);
> > >  		return 0;
> > > @@ -2128,9 +2123,8 @@ i40e_dev_tx_queue_setup_runtime(struct
> rte_eth_dev *dev,
> > >  	}
> > >  	/* check simple tx conflict */
> > >  	if (ad->tx_simple_allowed) {
> > > -		if (((txq->txq_flags & I40E_SIMPLE_FLAGS) !=
> > > -		     I40E_SIMPLE_FLAGS) ||
> > > -		    txq->tx_rs_thresh < RTE_PMD_I40E_TX_MAX_BURST) {
> > > +		if (txq->offloads != 0 ||
> > > +				txq->tx_rs_thresh < RTE_PMD_I40E_TX_MAX_BURST)
> {
> > >  			PMD_DRV_LOG(ERR, "No-simple tx is required.");
> > >  			return -EINVAL;
> > >  		}
> > > @@ -3080,18 +3074,21 @@ i40e_set_tx_function_flag(struct
> rte_eth_dev *dev, struct i40e_tx_queue *txq)
> > >  		I40E_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
> > >
> > >  	/* Use a simple Tx queue (no offloads, no multi segs) if possible */
> > > -	if (((txq->txq_flags & I40E_SIMPLE_FLAGS) == I40E_SIMPLE_FLAGS)
> > > -			&& (txq->tx_rs_thresh >= RTE_PMD_I40E_TX_MAX_BURST)) {
> > > -		if (txq->tx_rs_thresh <= RTE_I40E_TX_MAX_FREE_BUF_SZ) {
> > > -			PMD_INIT_LOG(DEBUG, "Vector tx"
> > > -				     " can be enabled on this txq.");
> > > -
> > > -		} else {
> > > -			ad->tx_vec_allowed = false;
> > > -		}
> > > -	} else {
> > > -		ad->tx_simple_allowed = false;
> > > -	}
> > > +	ad->tx_simple_allowed = (txq->offloads == 0 &&
> > > +			txq->tx_rs_thresh >= RTE_PMD_I40E_TX_MAX_BURST);
> >
> > Actually after another thought - who setup txq->offloads?
> > I did a quick scan, through i40e code and seems no one does.
> > So now it seems not possible to enable TX offloads at all.
> > Konstantin
> >
> > BTW, seems like rxq->offloads are not properly initialised too.
> >
> The offloads value should come from the app, no?

This should be a separate issue, I have submit the fix.

http://dpdk.org/dev/patchwork/patch/39229/

Regard
Qi

^ permalink raw reply	[flat|nested] 12+ messages in thread

end of thread, other threads:[~2018-05-02  8:30 UTC | newest]

Thread overview: 12+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2018-05-01 13:03 [dpdk-dev] [PATCH] net/i40e: fix Tx fn selection when using new ethdev offloads Bruce Richardson
2018-05-01 13:24 ` Ananyev, Konstantin
2018-05-01 13:28   ` Bruce Richardson
2018-05-01 13:52   ` Bruce Richardson
2018-05-01 14:11     ` Ananyev, Konstantin
2018-05-01 14:13 ` [dpdk-dev] [PATCH v2] " Bruce Richardson
2018-05-01 14:16   ` Bruce Richardson
2018-05-01 14:37     ` Ferruh Yigit
2018-05-01 15:48       ` Ferruh Yigit
2018-05-01 17:52   ` Ananyev, Konstantin
2018-05-02  8:24     ` Bruce Richardson
2018-05-02  8:30       ` Zhang, Qi Z

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).