[PATCH v3 2/4] net/softnic: add traffic management support

[Jasvinder Singh <jasvinder.singh@intel.com>]

Add ethdev Traffic Management API support to SoftNIC PMD.

Signed-off-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
Signed-off-by: Jasvinder Singh <jasvinder.singh@intel.com>
---
v3 changes:
- add more confguration parameters (tm rate, tm queue sizes)

 drivers/net/softnic/Makefile                    |   1 +
 drivers/net/softnic/rte_eth_softnic.c           | 252 +++++++++++++++++++++++-
 drivers/net/softnic/rte_eth_softnic.h           |  16 ++
 drivers/net/softnic/rte_eth_softnic_internals.h | 106 +++++++++-
 drivers/net/softnic/rte_eth_softnic_tm.c        | 180 +++++++++++++++++
 5 files changed, 552 insertions(+), 3 deletions(-)
 create mode 100644 drivers/net/softnic/rte_eth_softnic_tm.c

diff --git a/drivers/net/softnic/Makefile b/drivers/net/softnic/Makefile
index 8d00656..eea76ca 100644
--- a/drivers/net/softnic/Makefile
+++ b/drivers/net/softnic/Makefile
@@ -47,6 +47,7 @@ LIBABIVER := 1
 # all source are stored in SRCS-y
 #
 SRCS-$(CONFIG_RTE_LIBRTE_PMD_SOFTNIC) += rte_eth_softnic.c
+SRCS-$(CONFIG_RTE_LIBRTE_PMD_SOFTNIC) += rte_eth_softnic_tm.c
 
 #
 # Export include files
diff --git a/drivers/net/softnic/rte_eth_softnic.c b/drivers/net/softnic/rte_eth_softnic.c
index 35cb93c..f1906af 100644
--- a/drivers/net/softnic/rte_eth_softnic.c
+++ b/drivers/net/softnic/rte_eth_softnic.c
@@ -42,14 +42,34 @@
 #include <rte_kvargs.h>
 #include <rte_errno.h>
 #include <rte_ring.h>
+#include <rte_sched.h>
 
 #include "rte_eth_softnic.h"
 #include "rte_eth_softnic_internals.h"
 
+#define PMD_PARAM_SOFT_TM					"soft_tm"
+#define PMD_PARAM_SOFT_TM_RATE				"soft_tm_rate"
+#define PMD_PARAM_SOFT_TM_NB_QUEUES			"soft_tm_nb_queues"
+#define PMD_PARAM_SOFT_TM_QSIZE0			"soft_tm_qsize0"
+#define PMD_PARAM_SOFT_TM_QSIZE1			"soft_tm_qsize1"
+#define PMD_PARAM_SOFT_TM_QSIZE2			"soft_tm_qsize2"
+#define PMD_PARAM_SOFT_TM_QSIZE3			"soft_tm_qsize3"
+#define PMD_PARAM_SOFT_TM_ENQ_BSZ			"soft_tm_enq_bsz"
+#define PMD_PARAM_SOFT_TM_DEQ_BSZ			"soft_tm_deq_bsz"
+
 #define PMD_PARAM_HARD_NAME					"hard_name"
 #define PMD_PARAM_HARD_TX_QUEUE_ID			"hard_tx_queue_id"
 
 static const char *pmd_valid_args[] = {
+	PMD_PARAM_SOFT_TM,
+	PMD_PARAM_SOFT_TM_RATE,
+	PMD_PARAM_SOFT_TM_NB_QUEUES,
+	PMD_PARAM_SOFT_TM_QSIZE0,
+	PMD_PARAM_SOFT_TM_QSIZE1,
+	PMD_PARAM_SOFT_TM_QSIZE2,
+	PMD_PARAM_SOFT_TM_QSIZE3,
+	PMD_PARAM_SOFT_TM_ENQ_BSZ,
+	PMD_PARAM_SOFT_TM_DEQ_BSZ,
 	PMD_PARAM_HARD_NAME,
 	PMD_PARAM_HARD_TX_QUEUE_ID,
 	NULL
@@ -157,6 +177,13 @@ pmd_dev_start(struct rte_eth_dev *dev)
 {
 	struct pmd_internals *p = dev->data->dev_private;
 
+	if (tm_used(dev)) {
+		int status = tm_start(p);
+
+		if (status)
+			return status;
+	}
+
 	dev->data->dev_link.link_status = ETH_LINK_UP;
 
 	if (p->params.soft.intrusive) {
@@ -173,7 +200,12 @@ pmd_dev_start(struct rte_eth_dev *dev)
 static void
 pmd_dev_stop(struct rte_eth_dev *dev)
 {
+	struct pmd_internals *p = dev->data->dev_private;
+
 	dev->data->dev_link.link_status = ETH_LINK_DOWN;
+
+	if (tm_used(dev))
+		tm_stop(p);
 }
 
 static void
@@ -294,6 +326,77 @@ rte_pmd_softnic_run_default(struct rte_eth_dev *dev)
 	return 0;
 }
 
+static __rte_always_inline int
+rte_pmd_softnic_run_tm(struct rte_eth_dev *dev)
+{
+	struct pmd_internals *p = dev->data->dev_private;
+
+	/* Persistent context: Read Only (update not required) */
+	struct rte_sched_port *sched = p->soft.tm.sched;
+	struct rte_mbuf **pkts_enq = p->soft.tm.pkts_enq;
+	struct rte_mbuf **pkts_deq = p->soft.tm.pkts_deq;
+	uint32_t enq_bsz = p->params.soft.tm.enq_bsz;
+	uint32_t deq_bsz = p->params.soft.tm.deq_bsz;
+	uint16_t nb_tx_queues = dev->data->nb_tx_queues;
+
+	/* Persistent context: Read - Write (update required) */
+	uint32_t txq_pos = p->soft.tm.txq_pos;
+	uint32_t pkts_enq_len = p->soft.tm.pkts_enq_len;
+	uint32_t flush_count = p->soft.tm.flush_count;
+
+	/* Not part of the persistent context */
+	uint32_t pkts_deq_len, pos;
+	uint16_t i;
+
+	/* Soft device TXQ read, TM enqueue */
+	for (i = 0; i < nb_tx_queues; i++) {
+		struct rte_ring *txq = dev->data->tx_queues[txq_pos];
+
+		/* Read TXQ burst to packet enqueue buffer */
+		pkts_enq_len += rte_ring_sc_dequeue_burst(txq,
+			(void **) &pkts_enq[pkts_enq_len],
+			enq_bsz,
+			NULL);
+
+		/* Increment TXQ */
+		txq_pos++;
+		if (txq_pos >= nb_tx_queues)
+			txq_pos = 0;
+
+		/* TM enqueue when complete burst is available */
+		if (pkts_enq_len >= enq_bsz) {
+			rte_sched_port_enqueue(sched, pkts_enq, pkts_enq_len);
+
+			pkts_enq_len = 0;
+			flush_count = 0;
+			break;
+		}
+	}
+
+	if (flush_count >= FLUSH_COUNT_THRESHOLD) {
+		if (pkts_enq_len)
+			rte_sched_port_enqueue(sched, pkts_enq, pkts_enq_len);
+
+		pkts_enq_len = 0;
+		flush_count = 0;
+	}
+
+	p->soft.tm.txq_pos = txq_pos;
+	p->soft.tm.pkts_enq_len = pkts_enq_len;
+	p->soft.tm.flush_count = flush_count + 1;
+
+	/* TM dequeue, Hard device TXQ write */
+	pkts_deq_len = rte_sched_port_dequeue(sched, pkts_deq, deq_bsz);
+
+	for (pos = 0; pos < pkts_deq_len; )
+		pos += rte_eth_tx_burst(p->hard.port_id,
+			p->params.hard.tx_queue_id,
+			&pkts_deq[pos],
+			(uint16_t) (pkts_deq_len - pos));
+
+	return 0;
+}
+
 int
 rte_pmd_softnic_run(uint8_t port_id)
 {
@@ -303,7 +406,9 @@ rte_pmd_softnic_run(uint8_t port_id)
 	RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, 0);
 #endif
 
-	return rte_pmd_softnic_run_default(dev);
+	return (tm_used(dev)) ?
+		rte_pmd_softnic_run_tm(dev) :
+		rte_pmd_softnic_run_default(dev);
 }
 
 static struct ether_addr eth_addr = { .addr_bytes = {0} };
@@ -378,12 +483,25 @@ pmd_init(struct pmd_params *params, int numa_node)
 		return NULL;
 	}
 
+	/* Traffic Management (TM)*/
+	if (params->soft.flags & PMD_FEATURE_TM) {
+		status = tm_init(p, params, numa_node);
+		if (status) {
+			default_free(p);
+			rte_free(p);
+			return NULL;
+		}
+	}
+
 	return p;
 }
 
 static void
 pmd_free(struct pmd_internals *p)
 {
+	if (p->params.soft.flags & PMD_FEATURE_TM)
+		tm_free(p);
+
 	default_free(p);
 
 	rte_free(p);
@@ -479,7 +597,7 @@ static int
 pmd_parse_args(struct pmd_params *p, const char *name, const char *params)
 {
 	struct rte_kvargs *kvlist;
-	int ret;
+	int i, ret;
 
 	kvlist = rte_kvargs_parse(params, pmd_valid_args);
 	if (kvlist == NULL)
@@ -489,8 +607,120 @@ pmd_parse_args(struct pmd_params *p, const char *name, const char *params)
 	memset(p, 0, sizeof(*p));
 	p->soft.name = name;
 	p->soft.intrusive = INTRUSIVE;
+	p->soft.tm.rate = 0;
+	p->soft.tm.nb_queues = SOFTNIC_SOFT_TM_NB_QUEUES;
+	for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i++)
+		p->soft.tm.qsize[i] = SOFTNIC_SOFT_TM_QUEUE_SIZE;
+	p->soft.tm.enq_bsz = SOFTNIC_SOFT_TM_ENQ_BSZ;
+	p->soft.tm.deq_bsz = SOFTNIC_SOFT_TM_DEQ_BSZ;
 	p->hard.tx_queue_id = SOFTNIC_HARD_TX_QUEUE_ID;
 
+	/* SOFT: TM (optional) */
+	if (rte_kvargs_count(kvlist, PMD_PARAM_SOFT_TM) == 1) {
+		char *s;
+
+		ret = rte_kvargs_process(kvlist, PMD_PARAM_SOFT_TM,
+			&get_string, &s);
+		if (ret < 0)
+			goto out_free;
+
+		if (strcmp(s, "on") == 0)
+			p->soft.flags |= PMD_FEATURE_TM;
+		else if (strcmp(s, "off") == 0)
+			p->soft.flags &= ~PMD_FEATURE_TM;
+		else
+			goto out_free;
+	}
+
+	/* SOFT: TM rate (measured in bytes/second) (optional) */
+	if (rte_kvargs_count(kvlist, PMD_PARAM_SOFT_TM_RATE) == 1) {
+		ret = rte_kvargs_process(kvlist, PMD_PARAM_SOFT_TM_RATE,
+			&get_uint32, &p->soft.tm.rate);
+		if (ret < 0)
+			goto out_free;
+
+		p->soft.flags |= PMD_FEATURE_TM;
+	}
+
+	/* SOFT: TM number of queues (optional) */
+	if (rte_kvargs_count(kvlist, PMD_PARAM_SOFT_TM_NB_QUEUES) == 1) {
+		ret = rte_kvargs_process(kvlist, PMD_PARAM_SOFT_TM_NB_QUEUES,
+			&get_uint32, &p->soft.tm.nb_queues);
+		if (ret < 0)
+			goto out_free;
+
+		p->soft.flags |= PMD_FEATURE_TM;
+	}
+
+	/* SOFT: TM queue size 0 .. 3 (optional) */
+	if (rte_kvargs_count(kvlist, PMD_PARAM_SOFT_TM_QSIZE0) == 1) {
+		uint32_t qsize;
+
+		ret = rte_kvargs_process(kvlist, PMD_PARAM_SOFT_TM_QSIZE0,
+			&get_uint32, &qsize);
+		if (ret < 0)
+			goto out_free;
+
+		p->soft.tm.qsize[0] = (uint16_t) qsize;
+		p->soft.flags |= PMD_FEATURE_TM;
+	}
+
+	if (rte_kvargs_count(kvlist, PMD_PARAM_SOFT_TM_QSIZE1) == 1) {
+		uint32_t qsize;
+
+		ret = rte_kvargs_process(kvlist, PMD_PARAM_SOFT_TM_QSIZE1,
+			&get_uint32, &qsize);
+		if (ret < 0)
+			goto out_free;
+
+		p->soft.tm.qsize[1] = (uint16_t) qsize;
+		p->soft.flags |= PMD_FEATURE_TM;
+	}
+
+	if (rte_kvargs_count(kvlist, PMD_PARAM_SOFT_TM_QSIZE2) == 1) {
+		uint32_t qsize;
+
+		ret = rte_kvargs_process(kvlist, PMD_PARAM_SOFT_TM_QSIZE2,
+			&get_uint32, &qsize);
+		if (ret < 0)
+			goto out_free;
+
+		p->soft.tm.qsize[2] = (uint16_t) qsize;
+		p->soft.flags |= PMD_FEATURE_TM;
+	}
+
+	if (rte_kvargs_count(kvlist, PMD_PARAM_SOFT_TM_QSIZE3) == 1) {
+		uint32_t qsize;
+
+		ret = rte_kvargs_process(kvlist, PMD_PARAM_SOFT_TM_QSIZE3,
+			&get_uint32, &qsize);
+		if (ret < 0)
+			goto out_free;
+
+		p->soft.tm.qsize[3] = (uint16_t) qsize;
+		p->soft.flags |= PMD_FEATURE_TM;
+	}
+
+	/* SOFT: TM enqueue burst size (optional) */
+	if (rte_kvargs_count(kvlist, PMD_PARAM_SOFT_TM_ENQ_BSZ) == 1) {
+		ret = rte_kvargs_process(kvlist, PMD_PARAM_SOFT_TM_ENQ_BSZ,
+			&get_uint32, &p->soft.tm.enq_bsz);
+		if (ret < 0)
+			goto out_free;
+
+		p->soft.flags |= PMD_FEATURE_TM;
+	}
+
+	/* SOFT: TM dequeue burst size (optional) */
+	if (rte_kvargs_count(kvlist, PMD_PARAM_SOFT_TM_DEQ_BSZ) == 1) {
+		ret = rte_kvargs_process(kvlist, PMD_PARAM_SOFT_TM_DEQ_BSZ,
+			&get_uint32, &p->soft.tm.deq_bsz);
+		if (ret < 0)
+			goto out_free;
+
+		p->soft.flags |= PMD_FEATURE_TM;
+	}
+
 	/* HARD: name (mandatory) */
 	if (rte_kvargs_count(kvlist, PMD_PARAM_HARD_NAME) == 1) {
 		ret = rte_kvargs_process(kvlist, PMD_PARAM_HARD_NAME,
@@ -523,6 +753,7 @@ pmd_probe(struct rte_vdev_device *vdev)
 	int status;
 
 	struct rte_eth_dev_info hard_info;
+	uint32_t hard_speed;
 	uint8_t hard_port_id;
 	int numa_node;
 	void *dev_private;
@@ -548,11 +779,19 @@ pmd_probe(struct rte_vdev_device *vdev)
 		return -EINVAL;
 
 	rte_eth_dev_info_get(hard_port_id, &hard_info);
+	hard_speed = eth_dev_speed_max_mbps(hard_info.speed_capa);
 	numa_node = rte_eth_dev_socket_id(hard_port_id);
 
 	if (p.hard.tx_queue_id >= hard_info.max_tx_queues)
 		return -EINVAL;
 
+	if (p.soft.flags & PMD_FEATURE_TM) {
+		status = tm_params_check(&p, hard_speed);
+
+		if (status)
+			return status;
+	}
+
 	/* Allocate and initialize soft ethdev private data */
 	dev_private = pmd_init(&p, numa_node);
 	if (dev_private == NULL)
@@ -605,5 +844,14 @@ static struct rte_vdev_driver pmd_drv = {
 
 RTE_PMD_REGISTER_VDEV(net_softnic, pmd_drv);
 RTE_PMD_REGISTER_PARAM_STRING(net_softnic,
+	PMD_PARAM_SOFT_TM	 "=on|off "
+	PMD_PARAM_SOFT_TM_RATE "=<int> "
+	PMD_PARAM_SOFT_TM_NB_QUEUES "=<int> "
+	PMD_PARAM_SOFT_TM_QSIZE0 "=<int> "
+	PMD_PARAM_SOFT_TM_QSIZE1 "=<int> "
+	PMD_PARAM_SOFT_TM_QSIZE2 "=<int> "
+	PMD_PARAM_SOFT_TM_QSIZE3 "=<int> "
+	PMD_PARAM_SOFT_TM_ENQ_BSZ "=<int> "
+	PMD_PARAM_SOFT_TM_DEQ_BSZ "=<int> "
 	PMD_PARAM_HARD_NAME "=<string> "
 	PMD_PARAM_HARD_TX_QUEUE_ID "=<int>");
diff --git a/drivers/net/softnic/rte_eth_softnic.h b/drivers/net/softnic/rte_eth_softnic.h
index f840345..4388dc5 100644
--- a/drivers/net/softnic/rte_eth_softnic.h
+++ b/drivers/net/softnic/rte_eth_softnic.h
@@ -40,6 +40,22 @@
 extern "C" {
 #endif
 
+#ifndef SOFTNIC_SOFT_TM_NB_QUEUES
+#define SOFTNIC_SOFT_TM_NB_QUEUES			65536
+#endif
+
+#ifndef SOFTNIC_SOFT_TM_QUEUE_SIZE
+#define SOFTNIC_SOFT_TM_QUEUE_SIZE			64
+#endif
+
+#ifndef SOFTNIC_SOFT_TM_ENQ_BSZ
+#define SOFTNIC_SOFT_TM_ENQ_BSZ				32
+#endif
+
+#ifndef SOFTNIC_SOFT_TM_DEQ_BSZ
+#define SOFTNIC_SOFT_TM_DEQ_BSZ				24
+#endif
+
 #ifndef SOFTNIC_HARD_TX_QUEUE_ID
 #define SOFTNIC_HARD_TX_QUEUE_ID			0
 #endif
diff --git a/drivers/net/softnic/rte_eth_softnic_internals.h b/drivers/net/softnic/rte_eth_softnic_internals.h
index dfb7fab..c3c41cc 100644
--- a/drivers/net/softnic/rte_eth_softnic_internals.h
+++ b/drivers/net/softnic/rte_eth_softnic_internals.h
@@ -37,10 +37,19 @@
 #include <stdint.h>
 
 #include <rte_mbuf.h>
+#include <rte_sched.h>
 #include <rte_ethdev.h>
 
 #include "rte_eth_softnic.h"
 
+/**
+ * PMD Parameters
+ */
+
+enum pmd_feature {
+	PMD_FEATURE_TM = 1, /**< Traffic Management (TM) */
+};
+
 #ifndef INTRUSIVE
 #define INTRUSIVE					0
 #endif
@@ -57,6 +66,16 @@ struct pmd_params {
 		 *      (potentially faster).
 		 */
 		int intrusive;
+
+		/** Traffic Management (TM) */
+		struct {
+			uint32_t rate; /**< Rate (bytes/second) */
+			uint32_t nb_queues; /**< Number of queues */
+			uint16_t qsize[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE];
+			/**< Queue size per traffic class */
+			uint32_t enq_bsz; /**< Enqueue burst size */
+			uint32_t deq_bsz; /**< Dequeue burst size */
+		} tm;
 	} soft;
 
 	/** Parameters for the hard device (existing) */
@@ -75,7 +94,7 @@ struct pmd_params {
 #endif
 
 #ifndef FLUSH_COUNT_THRESHOLD
-#define FLUSH_COUNT_THRESHOLD			(1 << 17)
+#define FLUSH_COUNT_THRESHOLD				(1 << 17)
 #endif
 
 struct default_internals {
@@ -86,6 +105,66 @@ struct default_internals {
 };
 
 /**
+ * Traffic Management (TM) Internals
+ */
+
+#ifndef TM_MAX_SUBPORTS
+#define TM_MAX_SUBPORTS					8
+#endif
+
+#ifndef TM_MAX_PIPES_PER_SUBPORT
+#define TM_MAX_PIPES_PER_SUBPORT			4096
+#endif
+
+struct tm_params {
+	struct rte_sched_port_params port_params;
+
+	struct rte_sched_subport_params subport_params[TM_MAX_SUBPORTS];
+
+	struct rte_sched_pipe_params
+		pipe_profiles[RTE_SCHED_PIPE_PROFILES_PER_PORT];
+	uint32_t n_pipe_profiles;
+	uint32_t pipe_to_profile[TM_MAX_SUBPORTS * TM_MAX_PIPES_PER_SUBPORT];
+};
+
+/* TM Levels */
+enum tm_node_level {
+	TM_NODE_LEVEL_PORT = 0,
+	TM_NODE_LEVEL_SUBPORT,
+	TM_NODE_LEVEL_PIPE,
+	TM_NODE_LEVEL_TC,
+	TM_NODE_LEVEL_QUEUE,
+	TM_NODE_LEVEL_MAX,
+};
+
+/* TM Hierarchy Specification */
+struct tm_hierarchy {
+	uint32_t n_tm_nodes[TM_NODE_LEVEL_MAX];
+};
+
+struct tm_internals {
+	/** Hierarchy specification
+	 *
+	 *     -Hierarchy is unfrozen at init and when port is stopped.
+	 *     -Hierarchy is frozen on successful hierarchy commit.
+	 *     -Run-time hierarchy changes are not allowed, therefore it makes
+	 *      sense to keep the hierarchy frozen after the port is started.
+	 */
+	struct tm_hierarchy h;
+
+	/** Blueprints */
+	struct tm_params params;
+
+	/** Run-time */
+	struct rte_sched_port *sched;
+	struct rte_mbuf **pkts_enq;
+	struct rte_mbuf **pkts_deq;
+	uint32_t pkts_enq_len;
+	uint32_t txq_pos;
+	uint32_t flush_count;
+};
+
+/**
  * PMD Internals
  */
 struct pmd_internals {
@@ -95,6 +174,7 @@ struct pmd_internals {
 	/** Soft device */
 	struct {
 		struct default_internals def; /**< Default */
+		struct tm_internals tm; /**< Traffic Management */
 	} soft;
 
 	/** Hard device */
@@ -111,4 +191,28 @@ struct pmd_rx_queue {
 	} hard;
 };
 
+int
+tm_params_check(struct pmd_params *params, uint32_t hard_rate);
+
+int
+tm_init(struct pmd_internals *p, struct pmd_params *params, int numa_node);
+
+void
+tm_free(struct pmd_internals *p);
+
+int
+tm_start(struct pmd_internals *p);
+
+void
+tm_stop(struct pmd_internals *p);
+
+static inline int
+tm_used(struct rte_eth_dev *dev)
+{
+	struct pmd_internals *p = dev->data->dev_private;
+
+	return (p->params.soft.flags & PMD_FEATURE_TM) &&
+		p->soft.tm.h.n_tm_nodes[TM_NODE_LEVEL_PORT];
+}
+
 #endif /* __INCLUDE_RTE_ETH_SOFTNIC_INTERNALS_H__ */
diff --git a/drivers/net/softnic/rte_eth_softnic_tm.c b/drivers/net/softnic/rte_eth_softnic_tm.c
new file mode 100644
index 0000000..7fbea46
--- /dev/null
+++ b/drivers/net/softnic/rte_eth_softnic_tm.c
@@ -0,0 +1,180 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) 2017 Intel Corporation. All rights reserved.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Intel Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <stdint.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include <rte_malloc.h>
+
+#include "rte_eth_softnic_internals.h"
+#include "rte_eth_softnic.h"
+
+#define BYTES_IN_MBPS (1000 * 1000 / 8)
+
+int
+tm_params_check(struct pmd_params *params, uint32_t hard_rate)
+{
+	uint64_t hard_rate_bytes_per_sec = hard_rate * BYTES_IN_MBPS;
+	uint32_t i;
+
+	/* rate */
+	if (params->soft.tm.rate) {
+		if (params->soft.tm.rate > hard_rate_bytes_per_sec)
+			return -EINVAL;
+	} else
+		params->soft.tm.rate =
+			(hard_rate_bytes_per_sec > UINT32_MAX) ?
+				UINT32_MAX : hard_rate_bytes_per_sec;
+
+	/* nb_queues */
+	if (params->soft.tm.nb_queues == 0)
+		return -EINVAL;
+
+	if (params->soft.tm.nb_queues < RTE_SCHED_QUEUES_PER_PIPE)
+		params->soft.tm.nb_queues = RTE_SCHED_QUEUES_PER_PIPE;
+
+	params->soft.tm.nb_queues =
+		rte_align32pow2(params->soft.tm.nb_queues);
+
+	/* qsize */
+	for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i++) {
+		if (params->soft.tm.qsize[i] == 0)
+			return -EINVAL;
+
+		params->soft.tm.qsize[i] =
+			rte_align32pow2(params->soft.tm.qsize[i]);
+	}
+
+	/* enq_bsz, deq_bsz */
+	if ((params->soft.tm.enq_bsz == 0) ||
+		(params->soft.tm.deq_bsz == 0) ||
+		(params->soft.tm.deq_bsz >= params->soft.tm.enq_bsz))
+		return -EINVAL;
+
+	return 0;
+}
+
+int
+tm_init(struct pmd_internals *p,
+	struct pmd_params *params,
+	int numa_node)
+{
+	uint32_t enq_bsz = params->soft.tm.enq_bsz;
+	uint32_t deq_bsz = params->soft.tm.deq_bsz;
+
+	p->soft.tm.pkts_enq = rte_zmalloc_socket(params->soft.name,
+		2 * enq_bsz * sizeof(struct rte_mbuf *),
+		0,
+		numa_node);
+
+	if (p->soft.tm.pkts_enq == NULL)
+		return -ENOMEM;
+
+	p->soft.tm.pkts_deq = rte_zmalloc_socket(params->soft.name,
+		deq_bsz * sizeof(struct rte_mbuf *),
+		0,
+		numa_node);
+
+	if (p->soft.tm.pkts_deq == NULL) {
+		rte_free(p->soft.tm.pkts_enq);
+		return -ENOMEM;
+	}
+
+	return 0;
+}
+
+void
+tm_free(struct pmd_internals *p)
+{
+	rte_free(p->soft.tm.pkts_enq);
+	rte_free(p->soft.tm.pkts_deq);
+}
+
+int
+tm_start(struct pmd_internals *p)
+{
+	struct tm_params *t = &p->soft.tm.params;
+	uint32_t n_subports, subport_id;
+	int status;
+
+	/* Port */
+	p->soft.tm.sched = rte_sched_port_config(&t->port_params);
+	if (p->soft.tm.sched == NULL)
+		return -1;
+
+	/* Subport */
+	n_subports = t->port_params.n_subports_per_port;
+	for (subport_id = 0; subport_id < n_subports; subport_id++) {
+		uint32_t n_pipes_per_subport =
+			t->port_params.n_pipes_per_subport;
+		uint32_t pipe_id;
+
+		status = rte_sched_subport_config(p->soft.tm.sched,
+			subport_id,
+			&t->subport_params[subport_id]);
+		if (status) {
+			rte_sched_port_free(p->soft.tm.sched);
+			return -1;
+		}
+
+		/* Pipe */
+		n_pipes_per_subport = t->port_params.n_pipes_per_subport;
+		for (pipe_id = 0; pipe_id < n_pipes_per_subport; pipe_id++) {
+			int pos = subport_id * TM_MAX_PIPES_PER_SUBPORT +
+				pipe_id;
+			int profile_id = t->pipe_to_profile[pos];
+
+			if (profile_id < 0)
+				continue;
+
+			status = rte_sched_pipe_config(p->soft.tm.sched,
+				subport_id,
+				pipe_id,
+				profile_id);
+			if (status) {
+				rte_sched_port_free(p->soft.tm.sched);
+				return -1;
+			}
+		}
+	}
+
+	return 0;
+}
+
+void
+tm_stop(struct pmd_internals *p)
+{
+	if (p->soft.tm.sched)
+		rte_sched_port_free(p->soft.tm.sched);
+}
-- 
2.9.3