DPDK patches and discussions
 help / color / mirror / Atom feed
From: Bing Zhao <bingz@nvidia.com>
To: <viacheslavo@nvidia.com>, <matan@nvidia.com>
Cc: <dev@dpdk.org>, <orika@nvidia.com>, <rasland@nvidia.com>,
	<thomas@monjalon.net>, <lizh@nvidia.com>, <shunh@nvidia.com>
Subject: [dpdk-dev] [PATCH v3 7/7] net/mlx5: add meter support for trTCM profiles
Date: Wed, 21 Jul 2021 11:54:21 +0300	[thread overview]
Message-ID: <20210721085421.13111-8-bingz@nvidia.com> (raw)
In-Reply-To: <20210721085421.13111-1-bingz@nvidia.com>

The support of RFC2698 and RFC4115 are added in mlx5 PMD. Only the
ASO metering supports these two profiles.

Signed-off-by: Bing Zhao <bingz@nvidia.com>
---
 doc/guides/nics/mlx5.rst               |   1 +
 doc/guides/rel_notes/release_21_08.rst |   1 +
 drivers/common/mlx5/mlx5_prm.h         |   5 +-
 drivers/net/mlx5/mlx5_flow_aso.c       |  23 ++++-
 drivers/net/mlx5/mlx5_flow_meter.c     | 112 ++++++++++++++++---------
 5 files changed, 98 insertions(+), 44 deletions(-)

diff --git a/doc/guides/nics/mlx5.rst b/doc/guides/nics/mlx5.rst
index 1f5b6fb954..ae267d315f 100644
--- a/doc/guides/nics/mlx5.rst
+++ b/doc/guides/nics/mlx5.rst
@@ -433,6 +433,7 @@ Limitations
      - RED: must be DROP.
   - Policy actions of RSS for green and yellow should have the same configuration except queues.
   - meter profile packet mode is supported.
+  - meter profiles of RFC2697, RFC2698 and RFC4115 are supported.
 
 - Integrity:
 
diff --git a/doc/guides/rel_notes/release_21_08.rst b/doc/guides/rel_notes/release_21_08.rst
index 03d4fd059a..e159615deb 100644
--- a/doc/guides/rel_notes/release_21_08.rst
+++ b/doc/guides/rel_notes/release_21_08.rst
@@ -92,6 +92,7 @@ New Features
   * Added support for matching on VXLAN header last 8-bits reserved field.
   * Optimized multi-thread flow rule insertion rate.
   * Added support for metering policy actions of yellow color.
+  * Added support for metering trTCM RFC2698 and RFC4115.
 
 * **Added Wangxun ngbe PMD.**
 
diff --git a/drivers/common/mlx5/mlx5_prm.h b/drivers/common/mlx5/mlx5_prm.h
index 7950070976..88705be9d6 100644
--- a/drivers/common/mlx5/mlx5_prm.h
+++ b/drivers/common/mlx5/mlx5_prm.h
@@ -3031,11 +3031,12 @@ struct mlx5_aso_mtr_dseg {
 #define ASO_DSEG_VALID_OFFSET 31
 #define ASO_DSEG_BO_OFFSET 30
 #define ASO_DSEG_SC_OFFSET 28
+#define ASO_DSEG_BBOG_OFFSET 27
 #define ASO_DSEG_MTR_MODE 24
 #define ASO_DSEG_CBS_EXP_OFFSET 24
 #define ASO_DSEG_CBS_MAN_OFFSET 16
-#define ASO_DSEG_CIR_EXP_MASK 0x1F
-#define ASO_DSEG_CIR_EXP_OFFSET 8
+#define ASO_DSEG_XIR_EXP_MASK 0x1F
+#define ASO_DSEG_XIR_EXP_OFFSET 8
 #define ASO_DSEG_EBS_EXP_OFFSET 24
 #define ASO_DSEG_EBS_MAN_OFFSET 16
 #define ASO_DSEG_EXP_MASK 0x1F
diff --git a/drivers/net/mlx5/mlx5_flow_aso.c b/drivers/net/mlx5/mlx5_flow_aso.c
index 23e22e560a..e11327a11b 100644
--- a/drivers/net/mlx5/mlx5_flow_aso.c
+++ b/drivers/net/mlx5/mlx5_flow_aso.c
@@ -747,10 +747,27 @@ mlx5_aso_mtr_sq_enqueue_single(struct mlx5_aso_sq *sq,
 		wqe->aso_dseg.mtrs[dseg_idx].v_bo_sc_bbog_mm =
 				RTE_BE32((1 << ASO_DSEG_VALID_OFFSET) |
 				(MLX5_FLOW_COLOR_GREEN << ASO_DSEG_SC_OFFSET));
-	/* Only needed for RFC2697. */
-	if (fm->profile->srtcm_prm.ebs_eir)
+	switch (fmp->profile.alg) {
+	case RTE_MTR_SRTCM_RFC2697:
+		/* Only needed for RFC2697. */
+		if (fm->profile->srtcm_prm.ebs_eir)
+			wqe->aso_dseg.mtrs[dseg_idx].v_bo_sc_bbog_mm |=
+					RTE_BE32(1 << ASO_DSEG_BO_OFFSET);
+		break;
+	case RTE_MTR_TRTCM_RFC2698:
 		wqe->aso_dseg.mtrs[dseg_idx].v_bo_sc_bbog_mm |=
-				RTE_BE32(1 << ASO_DSEG_BO_OFFSET);
+				RTE_BE32(1 << ASO_DSEG_BBOG_OFFSET);
+		break;
+	case RTE_MTR_TRTCM_RFC4115:
+	default:
+		break;
+	}
+	/*
+	 * Note:
+	 * Due to software performance reason, the token fields will not be
+	 * set when posting the WQE to ASO SQ. It will be filled by the HW
+	 * automatically.
+	 */
 	sq->head++;
 	sq->pi += 2;/* Each WQE contains 2 WQEBB's. */
 	rte_io_wmb();
diff --git a/drivers/net/mlx5/mlx5_flow_meter.c b/drivers/net/mlx5/mlx5_flow_meter.c
index 4f57b7e04e..2d91a6fcf0 100644
--- a/drivers/net/mlx5/mlx5_flow_meter.c
+++ b/drivers/net/mlx5/mlx5_flow_meter.c
@@ -55,7 +55,7 @@ mlx5_flow_meter_action_create(struct mlx5_priv *priv,
 	MLX5_SET(flow_meter_parameters, fmp, cbs_exponent, val);
 	val = (cbs_cir >> ASO_DSEG_CBS_MAN_OFFSET) & ASO_DSEG_MAN_MASK;
 	MLX5_SET(flow_meter_parameters, fmp, cbs_mantissa, val);
-	val = (cbs_cir >> ASO_DSEG_CIR_EXP_OFFSET) & ASO_DSEG_EXP_MASK;
+	val = (cbs_cir >> ASO_DSEG_XIR_EXP_OFFSET) & ASO_DSEG_EXP_MASK;
 	MLX5_SET(flow_meter_parameters, fmp, cir_exponent, val);
 	val = (cbs_cir & ASO_DSEG_MAN_MASK);
 	MLX5_SET(flow_meter_parameters, fmp, cir_mantissa, val);
@@ -194,18 +194,18 @@ mlx5_flow_meter_profile_validate(struct rte_eth_dev *dev,
 				  NULL, "Metering algorithm not supported.");
 }
 
-/**
- * Calculate mantissa and exponent for cir.
+/*
+ * Calculate mantissa and exponent for cir / eir.
  *
- * @param[in] cir
+ * @param[in] xir
  *   Value to be calculated.
  * @param[out] man
  *   Pointer to the mantissa.
  * @param[out] exp
  *   Pointer to the exp.
  */
-static void
-mlx5_flow_meter_cir_man_exp_calc(int64_t cir, uint8_t *man, uint8_t *exp)
+static inline void
+mlx5_flow_meter_xir_man_exp_calc(int64_t xir, uint8_t *man, uint8_t *exp)
 {
 	int64_t _cir;
 	int64_t delta = INT64_MAX;
@@ -216,8 +216,8 @@ mlx5_flow_meter_cir_man_exp_calc(int64_t cir, uint8_t *man, uint8_t *exp)
 	for (m = 0; m <= 0xFF; m++) { /* man width 8 bit */
 		for (e = 0; e <= 0x1F; e++) { /* exp width 5bit */
 			_cir = (1000000000ULL * m) >> e;
-			if (llabs(cir - _cir) <= delta) {
-				delta = llabs(cir - _cir);
+			if (llabs(xir - _cir) <= delta) {
+				delta = llabs(xir - _cir);
 				_man = m;
 				_exp = e;
 			}
@@ -227,7 +227,7 @@ mlx5_flow_meter_cir_man_exp_calc(int64_t cir, uint8_t *man, uint8_t *exp)
 	*exp = _exp;
 }
 
-/**
+/*
  * Calculate mantissa and exponent for xbs.
  *
  * @param[in] xbs
@@ -237,7 +237,7 @@ mlx5_flow_meter_cir_man_exp_calc(int64_t cir, uint8_t *man, uint8_t *exp)
  * @param[out] exp
  *   Pointer to the exp.
  */
-static void
+static inline void
 mlx5_flow_meter_xbs_man_exp_calc(uint64_t xbs, uint8_t *man, uint8_t *exp)
 {
 	int _exp;
@@ -275,37 +275,63 @@ mlx5_flow_meter_param_fill(struct mlx5_flow_meter_profile *fmp,
 	struct mlx5_flow_meter_srtcm_rfc2697_prm *srtcm = &fmp->srtcm_prm;
 	uint8_t man, exp;
 	uint32_t cbs_exp, cbs_man, cir_exp, cir_man;
-	uint32_t ebs_exp, ebs_man;
-	uint64_t cir, cbs, ebs;
+	uint32_t eir_exp, eir_man, ebs_exp, ebs_man;
+	uint64_t cir, cbs, eir, ebs;
 
-	if (fmp->profile.alg != RTE_MTR_SRTCM_RFC2697)
-		return -rte_mtr_error_set(error, ENOTSUP,
+	if (!priv->sh->meter_aso_en) {
+		/* Legacy FW metering will only support srTCM. */
+		if (fmp->profile.alg != RTE_MTR_SRTCM_RFC2697)
+			return -rte_mtr_error_set(error, ENOTSUP,
 				RTE_MTR_ERROR_TYPE_METER_PROFILE,
-				NULL, "Metering algorithm not supported.");
-	if (!priv->sh->meter_aso_en && fmp->profile.packet_mode)
-		return -rte_mtr_error_set(error, ENOTSUP,
-			RTE_MTR_ERROR_TYPE_METER_PROFILE,
-			NULL, "Metering algorithm packet mode not supported.");
-	if (priv->sh->meter_aso_en && fmp->profile.packet_mode) {
-		cir = fmp->profile.srtcm_rfc2697.cir <<
-				MLX5_MTRS_PPS_MAP_BPS_SHIFT;
-		cbs = fmp->profile.srtcm_rfc2697.cbs <<
-				MLX5_MTRS_PPS_MAP_BPS_SHIFT;
-		ebs = fmp->profile.srtcm_rfc2697.ebs <<
-				MLX5_MTRS_PPS_MAP_BPS_SHIFT;
-	} else {
+				NULL, "Metering algorithm is not supported.");
+		if (fmp->profile.packet_mode)
+			return -rte_mtr_error_set(error, ENOTSUP,
+				RTE_MTR_ERROR_TYPE_METER_PROFILE, NULL,
+				"Metering algorithm packet mode is not supported.");
+	}
+	switch (fmp->profile.alg) {
+	case RTE_MTR_SRTCM_RFC2697:
 		cir = fmp->profile.srtcm_rfc2697.cir;
 		cbs = fmp->profile.srtcm_rfc2697.cbs;
+		eir = 0;
 		ebs = fmp->profile.srtcm_rfc2697.ebs;
+		break;
+	case RTE_MTR_TRTCM_RFC2698:
+		MLX5_ASSERT(fmp->profile.trtcm_rfc2698.pir >
+			    fmp->profile.trtcm_rfc2698.cir &&
+			    fmp->profile.trtcm_rfc2698.pbs >
+			    fmp->profile.trtcm_rfc2698.cbs);
+		cir = fmp->profile.trtcm_rfc2698.cir;
+		cbs = fmp->profile.trtcm_rfc2698.cbs;
+		/* EIR / EBS are filled with PIR / PBS. */
+		eir = fmp->profile.trtcm_rfc2698.pir;
+		ebs = fmp->profile.trtcm_rfc2698.pbs;
+		break;
+	case RTE_MTR_TRTCM_RFC4115:
+		cir = fmp->profile.trtcm_rfc4115.cir;
+		cbs = fmp->profile.trtcm_rfc4115.cbs;
+		eir = fmp->profile.trtcm_rfc4115.eir;
+		ebs = fmp->profile.trtcm_rfc4115.ebs;
+		break;
+	default:
+		return -rte_mtr_error_set(error, EINVAL,
+				RTE_MTR_ERROR_TYPE_METER_PROFILE, NULL,
+				"Metering algorithm mode is invalid");
+	}
+	/* Adjust the values for PPS mode. */
+	if (fmp->profile.packet_mode) {
+		cir <<= MLX5_MTRS_PPS_MAP_BPS_SHIFT;
+		cbs <<= MLX5_MTRS_PPS_MAP_BPS_SHIFT;
+		eir <<= MLX5_MTRS_PPS_MAP_BPS_SHIFT;
+		ebs <<= MLX5_MTRS_PPS_MAP_BPS_SHIFT;
 	}
 	/* cir = 8G * cir_mantissa * 1/(2^cir_exponent)) Bytes/Sec */
-	mlx5_flow_meter_cir_man_exp_calc(cir, &man, &exp);
+	mlx5_flow_meter_xir_man_exp_calc(cir, &man, &exp);
 	/* Check if cir mantissa is too large. */
-	if (exp > ASO_DSEG_CIR_EXP_MASK)
+	if (exp > ASO_DSEG_XIR_EXP_MASK)
 		return -rte_mtr_error_set(error, ENOTSUP,
 					  RTE_MTR_ERROR_TYPE_MTR_PARAMS, NULL,
-					  "meter profile parameter cir is"
-					  " not supported.");
+					  "meter profile parameter cir is not supported.");
 	cir_man = man;
 	cir_exp = exp;
 	 /* cbs = cbs_mantissa * 2^cbs_exponent */
@@ -314,25 +340,33 @@ mlx5_flow_meter_param_fill(struct mlx5_flow_meter_profile *fmp,
 	if (exp > ASO_DSEG_EXP_MASK)
 		return -rte_mtr_error_set(error, ENOTSUP,
 					  RTE_MTR_ERROR_TYPE_MTR_PARAMS, NULL,
-					  "meter profile parameter cbs is"
-					  " not supported.");
+					  "meter profile parameter cbs is not supported.");
 	cbs_man = man;
 	cbs_exp = exp;
 	srtcm->cbs_cir = rte_cpu_to_be_32(cbs_exp << ASO_DSEG_CBS_EXP_OFFSET |
 					  cbs_man << ASO_DSEG_CBS_MAN_OFFSET |
-					  cir_exp << ASO_DSEG_CIR_EXP_OFFSET |
+					  cir_exp << ASO_DSEG_XIR_EXP_OFFSET |
 					  cir_man);
+	mlx5_flow_meter_xir_man_exp_calc(eir, &man, &exp);
+	/* Check if eir mantissa is too large. */
+	if (exp > ASO_DSEG_XIR_EXP_MASK)
+		return -rte_mtr_error_set(error, ENOTSUP,
+					  RTE_MTR_ERROR_TYPE_MTR_PARAMS, NULL,
+					  "meter profile parameter eir is not supported.");
+	eir_man = man;
+	eir_exp = exp;
 	mlx5_flow_meter_xbs_man_exp_calc(ebs, &man, &exp);
 	/* Check if ebs mantissa is too large. */
 	if (exp > ASO_DSEG_EXP_MASK)
 		return -rte_mtr_error_set(error, ENOTSUP,
 					  RTE_MTR_ERROR_TYPE_MTR_PARAMS, NULL,
-					  "meter profile parameter ebs is"
-					  " not supported.");
+					  "meter profile parameter ebs is not supported.");
 	ebs_man = man;
 	ebs_exp = exp;
 	srtcm->ebs_eir = rte_cpu_to_be_32(ebs_exp << ASO_DSEG_EBS_EXP_OFFSET |
-					  ebs_man << ASO_DSEG_EBS_MAN_OFFSET);
+					  ebs_man << ASO_DSEG_EBS_MAN_OFFSET |
+					  eir_exp << ASO_DSEG_XIR_EXP_OFFSET |
+					  eir_man);
 	if (srtcm->cbs_cir)
 		fmp->g_support = 1;
 	if (srtcm->ebs_eir)
@@ -1008,7 +1042,7 @@ mlx5_flow_meter_action_modify(struct mlx5_priv *priv,
 				cbs_mantissa, val);
 		}
 		if (modify_bits & MLX5_FLOW_METER_OBJ_MODIFY_FIELD_CIR) {
-			val = (cbs_cir >> ASO_DSEG_CIR_EXP_OFFSET) &
+			val = (cbs_cir >> ASO_DSEG_XIR_EXP_OFFSET) &
 				ASO_DSEG_EXP_MASK;
 			MLX5_SET(flow_meter_parameters, attr,
 				cir_exponent, val);
@@ -1389,7 +1423,7 @@ mlx5_flow_meter_modify_state(struct mlx5_priv *priv,
 				&srtcm, modify_bits, 0, 0);
 	else
 		ret = mlx5_flow_meter_action_modify(priv, fm,
-						   &fm->profile->srtcm_prm,
+						    &fm->profile->srtcm_prm,
 						    modify_bits, 0, 1);
 	if (ret)
 		return -rte_mtr_error_set(error, -ret,
-- 
2.27.0


  parent reply	other threads:[~2021-07-21  8:55 UTC|newest]

Thread overview: 24+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-07-05 15:57 [dpdk-dev] [PATCH 0/6] support yellow color policy in mlx5 Bing Zhao
2021-07-05 15:57 ` [dpdk-dev] [PATCH 1/6] net/mlx5: add yellow color default policy Bing Zhao
2021-07-05 15:57 ` [dpdk-dev] [PATCH 2/6] net/mlx5: support yellow in meter policy validation Bing Zhao
2021-07-05 15:57 ` [dpdk-dev] [PATCH 3/6] net/mlx5: enable meter bucket overflow for yellow color Bing Zhao
2021-07-05 15:57 ` [dpdk-dev] [PATCH 4/6] net/mlx5: added support for yellow policy rules Bing Zhao
2021-07-05 15:57 ` [dpdk-dev] [PATCH 5/6] net/mlx5: split policies handling of colors Bing Zhao
2021-07-05 15:57 ` [dpdk-dev] [PATCH 6/6] doc: update mlx5 metering policy part Bing Zhao
2021-07-18 17:18 ` [dpdk-dev] [PATCH v2 0/7] support yellow color policy in mlx5 Bing Zhao
2021-07-18 17:18   ` [dpdk-dev] [PATCH v2 1/7] net/mlx5: handle yellow case in default meter policy Bing Zhao
2021-07-18 17:18   ` [dpdk-dev] [PATCH v2 2/7] net/mlx5: enable meter bucket overflow for yellow color Bing Zhao
2021-07-18 17:18   ` [dpdk-dev] [PATCH v2 3/7] net/mlx5: added support for yellow policy rules Bing Zhao
2021-07-18 17:18   ` [dpdk-dev] [PATCH v2 4/7] net/mlx5: split policies handling of colors Bing Zhao
2021-07-18 17:18   ` [dpdk-dev] [PATCH v2 5/7] net/mlx5: support yellow in meter policy validation Bing Zhao
2021-07-18 17:18   ` [dpdk-dev] [PATCH v2 6/7] net/mlx5: check consistency of meter policy and profile Bing Zhao
2021-07-18 17:18   ` [dpdk-dev] [PATCH v2 7/7] net/mlx5: add meter support for trTCM profiles Bing Zhao
2021-07-21  8:54 ` [dpdk-dev] [PATCH v3 0/7] support yellow color policy in mlx5 Bing Zhao
2021-07-21  8:54   ` [dpdk-dev] [PATCH v3 1/7] net/mlx5: handle yellow case in default meter policy Bing Zhao
2021-07-21  8:54   ` [dpdk-dev] [PATCH v3 2/7] net/mlx5: enable meter bucket overflow for yellow color Bing Zhao
2021-07-21  8:54   ` [dpdk-dev] [PATCH v3 3/7] net/mlx5: added support for yellow policy rules Bing Zhao
2021-07-21  8:54   ` [dpdk-dev] [PATCH v3 4/7] net/mlx5: split policies handling of colors Bing Zhao
2021-07-21  8:54   ` [dpdk-dev] [PATCH v3 5/7] net/mlx5: support yellow in meter policy validation Bing Zhao
2021-07-21  8:54   ` [dpdk-dev] [PATCH v3 6/7] net/mlx5: check consistency of meter policy and profile Bing Zhao
2021-07-21  8:54   ` Bing Zhao [this message]
2021-07-22 11:28   ` [dpdk-dev] [PATCH v3 0/7] support yellow color policy in mlx5 Thomas Monjalon

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20210721085421.13111-8-bingz@nvidia.com \
    --to=bingz@nvidia.com \
    --cc=dev@dpdk.org \
    --cc=lizh@nvidia.com \
    --cc=matan@nvidia.com \
    --cc=orika@nvidia.com \
    --cc=rasland@nvidia.com \
    --cc=shunh@nvidia.com \
    --cc=thomas@monjalon.net \
    --cc=viacheslavo@nvidia.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
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).