net/softnic: add softnic PMD
authorJasvinder Singh <jasvinder.singh@intel.com>
Tue, 10 Oct 2017 10:18:14 +0000 (11:18 +0100)
committerFerruh Yigit <ferruh.yigit@intel.com>
Thu, 12 Oct 2017 00:52:38 +0000 (01:52 +0100)
Add SoftNIC PMD to provide SW fall-back for ethdev APIs.

Signed-off-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
Signed-off-by: Jasvinder Singh <jasvinder.singh@intel.com>
Acked-by: Thomas Monjalon <thomas@monjalon.net>
12 files changed:
MAINTAINERS
config/common_base
doc/api/doxy-api-index.md
doc/api/doxy-api.conf
doc/guides/rel_notes/release_17_11.rst
drivers/net/Makefile
drivers/net/softnic/Makefile [new file with mode: 0644]
drivers/net/softnic/rte_eth_softnic.c [new file with mode: 0644]
drivers/net/softnic/rte_eth_softnic.h [new file with mode: 0644]
drivers/net/softnic/rte_eth_softnic_internals.h [new file with mode: 0644]
drivers/net/softnic/rte_pmd_eth_softnic_version.map [new file with mode: 0644]
mk/rte.app.mk

index 65a931c..b8b5441 100644 (file)
@@ -538,6 +538,11 @@ M: Gaetan Rivet <gaetan.rivet@6wind.com>
 F: drivers/net/failsafe/
 F: doc/guides/nics/fail_safe.rst
 
+Softnic PMD
+M: Jasvinder Singh <jasvinder.singh@intel.com>
+M: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
+F: drivers/net/softnic/
+
 
 Crypto Drivers
 --------------
index 4648943..dc4da9a 100644 (file)
@@ -278,6 +278,11 @@ CONFIG_RTE_LIBRTE_BNXT_PMD=y
 CONFIG_RTE_LIBRTE_SFC_EFX_PMD=y
 CONFIG_RTE_LIBRTE_SFC_EFX_DEBUG=n
 
+#
+# Compile SOFTNIC PMD
+#
+CONFIG_RTE_LIBRTE_PMD_SOFTNIC=y
+
 #
 # Compile software PMD backed by SZEDATA2 device
 #
index 02e92a4..990815f 100644 (file)
@@ -51,6 +51,7 @@ The public API headers are grouped by topics:
   [PCI]                (@ref rte_pci.h)
 
 - **device specific**:
+  [softnic]            (@ref rte_eth_softnic.h),
   [bond]               (@ref rte_eth_bond.h),
   [vhost]              (@ref rte_vhost.h),
   [KNI]                (@ref rte_kni.h),
index 91bb090..9e9fa56 100644 (file)
@@ -35,6 +35,7 @@ INPUT                   = doc/api/doxy-api-index.md \
                           drivers/net/bonding \
                           drivers/net/i40e \
                           drivers/net/ixgbe \
+                          drivers/net/softnic \
                           lib/librte_eal/common/include \
                           lib/librte_eal/common/include/generic \
                           lib/librte_acl \
index 4f5576d..fad1a19 100644 (file)
@@ -59,6 +59,11 @@ New Features
   "Network Interface Controller Drivers" document for more details on this new
   driver.
 
+* **Added SoftNIC PMD.**
+
+  Added new SoftNIC PMD. This virtual device offers applications a software
+  fallback support for traffic management.
+
 * **nfp: Added PF support.**
 
   Previously Netronome's NFP PMD had just support for VFs. PF support is
@@ -288,6 +293,7 @@ The libraries prepended with a plus sign were incremented in this version.
      librte_pmd_i40e.so.2
      librte_pmd_ixgbe.so.2
      librte_pmd_ring.so.2
+   + librte_pmd_softnic.so.1
      librte_pmd_vhost.so.2
      librte_port.so.3
      librte_power.so.1
index aa5edf8..5d2ad2f 100644 (file)
@@ -109,6 +109,11 @@ DIRS-$(CONFIG_RTE_LIBRTE_PMD_KNI) += kni
 endif
 DEPDIRS-kni = $(core-libs) librte_kni
 
+ifeq ($(CONFIG_RTE_LIBRTE_SCHED),y)
+DIRS-$(CONFIG_RTE_LIBRTE_PMD_SOFTNIC) += softnic
+endif # $(CONFIG_RTE_LIBRTE_SCHED)
+DEPDIRS-softnic = $(core-libs) librte_sched
+
 ifeq ($(CONFIG_RTE_LIBRTE_VHOST),y)
 DIRS-$(CONFIG_RTE_LIBRTE_PMD_VHOST) += vhost
 endif # $(CONFIG_RTE_LIBRTE_VHOST)
diff --git a/drivers/net/softnic/Makefile b/drivers/net/softnic/Makefile
new file mode 100644 (file)
index 0000000..c2f42ef
--- /dev/null
@@ -0,0 +1,56 @@
+#   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 $(RTE_SDK)/mk/rte.vars.mk
+
+#
+# library name
+#
+LIB = librte_pmd_softnic.a
+
+CFLAGS += -O3
+CFLAGS += $(WERROR_FLAGS)
+
+EXPORT_MAP := rte_pmd_eth_softnic_version.map
+
+LIBABIVER := 1
+
+#
+# all source are stored in SRCS-y
+#
+SRCS-$(CONFIG_RTE_LIBRTE_PMD_SOFTNIC) += rte_eth_softnic.c
+
+#
+# Export include files
+#
+SYMLINK-y-include += rte_eth_softnic.h
+
+include $(RTE_SDK)/mk/rte.lib.mk
diff --git a/drivers/net/softnic/rte_eth_softnic.c b/drivers/net/softnic/rte_eth_softnic.c
new file mode 100644 (file)
index 0000000..2f92594
--- /dev/null
@@ -0,0 +1,591 @@
+/*-
+ *   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_ethdev.h>
+#include <rte_ethdev_vdev.h>
+#include <rte_malloc.h>
+#include <rte_vdev.h>
+#include <rte_kvargs.h>
+#include <rte_errno.h>
+#include <rte_ring.h>
+
+#include "rte_eth_softnic.h"
+#include "rte_eth_softnic_internals.h"
+
+#define DEV_HARD(p)                                    \
+       (&rte_eth_devices[p->hard.port_id])
+
+#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_HARD_NAME,
+       PMD_PARAM_HARD_TX_QUEUE_ID,
+       NULL
+};
+
+static const struct rte_eth_dev_info pmd_dev_info = {
+       .min_rx_bufsize = 0,
+       .max_rx_pktlen = UINT32_MAX,
+       .max_rx_queues = UINT16_MAX,
+       .max_tx_queues = UINT16_MAX,
+       .rx_desc_lim = {
+               .nb_max = UINT16_MAX,
+               .nb_min = 0,
+               .nb_align = 1,
+       },
+       .tx_desc_lim = {
+               .nb_max = UINT16_MAX,
+               .nb_min = 0,
+               .nb_align = 1,
+       },
+};
+
+static void
+pmd_dev_infos_get(struct rte_eth_dev *dev __rte_unused,
+       struct rte_eth_dev_info *dev_info)
+{
+       memcpy(dev_info, &pmd_dev_info, sizeof(*dev_info));
+}
+
+static int
+pmd_dev_configure(struct rte_eth_dev *dev)
+{
+       struct pmd_internals *p = dev->data->dev_private;
+       struct rte_eth_dev *hard_dev = DEV_HARD(p);
+
+       if (dev->data->nb_rx_queues > hard_dev->data->nb_rx_queues)
+               return -1;
+
+       if (p->params.hard.tx_queue_id >= hard_dev->data->nb_tx_queues)
+               return -1;
+
+       return 0;
+}
+
+static int
+pmd_rx_queue_setup(struct rte_eth_dev *dev,
+       uint16_t rx_queue_id,
+       uint16_t nb_rx_desc __rte_unused,
+       unsigned int socket_id,
+       const struct rte_eth_rxconf *rx_conf __rte_unused,
+       struct rte_mempool *mb_pool __rte_unused)
+{
+       struct pmd_internals *p = dev->data->dev_private;
+
+       if (p->params.soft.intrusive == 0) {
+               struct pmd_rx_queue *rxq;
+
+               rxq = rte_zmalloc_socket(p->params.soft.name,
+                       sizeof(struct pmd_rx_queue), 0, socket_id);
+               if (rxq == NULL)
+                       return -ENOMEM;
+
+               rxq->hard.port_id = p->hard.port_id;
+               rxq->hard.rx_queue_id = rx_queue_id;
+               dev->data->rx_queues[rx_queue_id] = rxq;
+       } else {
+               struct rte_eth_dev *hard_dev = DEV_HARD(p);
+               void *rxq = hard_dev->data->rx_queues[rx_queue_id];
+
+               if (rxq == NULL)
+                       return -1;
+
+               dev->data->rx_queues[rx_queue_id] = rxq;
+       }
+       return 0;
+}
+
+static int
+pmd_tx_queue_setup(struct rte_eth_dev *dev,
+       uint16_t tx_queue_id,
+       uint16_t nb_tx_desc,
+       unsigned int socket_id,
+       const struct rte_eth_txconf *tx_conf __rte_unused)
+{
+       uint32_t size = RTE_ETH_NAME_MAX_LEN + strlen("_txq") + 4;
+       char name[size];
+       struct rte_ring *r;
+
+       snprintf(name, sizeof(name), "%s_txq%04x",
+               dev->data->name, tx_queue_id);
+       r = rte_ring_create(name, nb_tx_desc, socket_id,
+               RING_F_SP_ENQ | RING_F_SC_DEQ);
+       if (r == NULL)
+               return -1;
+
+       dev->data->tx_queues[tx_queue_id] = r;
+       return 0;
+}
+
+static int
+pmd_dev_start(struct rte_eth_dev *dev)
+{
+       struct pmd_internals *p = dev->data->dev_private;
+
+       dev->data->dev_link.link_status = ETH_LINK_UP;
+
+       if (p->params.soft.intrusive) {
+               struct rte_eth_dev *hard_dev = DEV_HARD(p);
+
+               /* The hard_dev->rx_pkt_burst should be stable by now */
+               dev->rx_pkt_burst = hard_dev->rx_pkt_burst;
+       }
+
+       return 0;
+}
+
+static void
+pmd_dev_stop(struct rte_eth_dev *dev)
+{
+       dev->data->dev_link.link_status = ETH_LINK_DOWN;
+}
+
+static void
+pmd_dev_close(struct rte_eth_dev *dev)
+{
+       uint32_t i;
+
+       /* TX queues */
+       for (i = 0; i < dev->data->nb_tx_queues; i++)
+               rte_ring_free((struct rte_ring *)dev->data->tx_queues[i]);
+}
+
+static int
+pmd_link_update(struct rte_eth_dev *dev __rte_unused,
+       int wait_to_complete __rte_unused)
+{
+       return 0;
+}
+
+static const struct eth_dev_ops pmd_ops = {
+       .dev_configure = pmd_dev_configure,
+       .dev_start = pmd_dev_start,
+       .dev_stop = pmd_dev_stop,
+       .dev_close = pmd_dev_close,
+       .link_update = pmd_link_update,
+       .dev_infos_get = pmd_dev_infos_get,
+       .rx_queue_setup = pmd_rx_queue_setup,
+       .tx_queue_setup = pmd_tx_queue_setup,
+       .tm_ops_get = NULL,
+};
+
+static uint16_t
+pmd_rx_pkt_burst(void *rxq,
+       struct rte_mbuf **rx_pkts,
+       uint16_t nb_pkts)
+{
+       struct pmd_rx_queue *rx_queue = rxq;
+
+       return rte_eth_rx_burst(rx_queue->hard.port_id,
+               rx_queue->hard.rx_queue_id,
+               rx_pkts,
+               nb_pkts);
+}
+
+static uint16_t
+pmd_tx_pkt_burst(void *txq,
+       struct rte_mbuf **tx_pkts,
+       uint16_t nb_pkts)
+{
+       return (uint16_t)rte_ring_enqueue_burst(txq,
+               (void **)tx_pkts,
+               nb_pkts,
+               NULL);
+}
+
+static __rte_always_inline int
+run_default(struct rte_eth_dev *dev)
+{
+       struct pmd_internals *p = dev->data->dev_private;
+
+       /* Persistent context: Read Only (update not required) */
+       struct rte_mbuf **pkts = p->soft.def.pkts;
+       uint16_t nb_tx_queues = dev->data->nb_tx_queues;
+
+       /* Persistent context: Read - Write (update required) */
+       uint32_t txq_pos = p->soft.def.txq_pos;
+       uint32_t pkts_len = p->soft.def.pkts_len;
+       uint32_t flush_count = p->soft.def.flush_count;
+
+       /* Not part of the persistent context */
+       uint32_t pos;
+       uint16_t i;
+
+       /* Soft device TXQ read, Hard device TXQ write */
+       for (i = 0; i < nb_tx_queues; i++) {
+               struct rte_ring *txq = dev->data->tx_queues[txq_pos];
+
+               /* Read soft device TXQ burst to packet enqueue buffer */
+               pkts_len += rte_ring_sc_dequeue_burst(txq,
+                       (void **)&pkts[pkts_len],
+                       DEFAULT_BURST_SIZE,
+                       NULL);
+
+               /* Increment soft device TXQ */
+               txq_pos++;
+               if (txq_pos >= nb_tx_queues)
+                       txq_pos = 0;
+
+               /* Hard device TXQ write when complete burst is available */
+               if (pkts_len >= DEFAULT_BURST_SIZE) {
+                       for (pos = 0; pos < pkts_len; )
+                               pos += rte_eth_tx_burst(p->hard.port_id,
+                                       p->params.hard.tx_queue_id,
+                                       &pkts[pos],
+                                       (uint16_t)(pkts_len - pos));
+
+                       pkts_len = 0;
+                       flush_count = 0;
+                       break;
+               }
+       }
+
+       if (flush_count >= FLUSH_COUNT_THRESHOLD) {
+               for (pos = 0; pos < pkts_len; )
+                       pos += rte_eth_tx_burst(p->hard.port_id,
+                               p->params.hard.tx_queue_id,
+                               &pkts[pos],
+                               (uint16_t)(pkts_len - pos));
+
+               pkts_len = 0;
+               flush_count = 0;
+       }
+
+       p->soft.def.txq_pos = txq_pos;
+       p->soft.def.pkts_len = pkts_len;
+       p->soft.def.flush_count = flush_count + 1;
+
+       return 0;
+}
+
+int
+rte_pmd_softnic_run(uint16_t port_id)
+{
+       struct rte_eth_dev *dev = &rte_eth_devices[port_id];
+
+#ifdef RTE_LIBRTE_ETHDEV_DEBUG
+       RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, 0);
+#endif
+
+       return run_default(dev);
+}
+
+static struct ether_addr eth_addr = { .addr_bytes = {0} };
+
+static uint32_t
+eth_dev_speed_max_mbps(uint32_t speed_capa)
+{
+       uint32_t rate_mbps[32] = {
+               ETH_SPEED_NUM_NONE,
+               ETH_SPEED_NUM_10M,
+               ETH_SPEED_NUM_10M,
+               ETH_SPEED_NUM_100M,
+               ETH_SPEED_NUM_100M,
+               ETH_SPEED_NUM_1G,
+               ETH_SPEED_NUM_2_5G,
+               ETH_SPEED_NUM_5G,
+               ETH_SPEED_NUM_10G,
+               ETH_SPEED_NUM_20G,
+               ETH_SPEED_NUM_25G,
+               ETH_SPEED_NUM_40G,
+               ETH_SPEED_NUM_50G,
+               ETH_SPEED_NUM_56G,
+               ETH_SPEED_NUM_100G,
+       };
+
+       uint32_t pos = (speed_capa) ? (31 - __builtin_clz(speed_capa)) : 0;
+       return rate_mbps[pos];
+}
+
+static int
+default_init(struct pmd_internals *p,
+       struct pmd_params *params,
+       int numa_node)
+{
+       p->soft.def.pkts = rte_zmalloc_socket(params->soft.name,
+               2 * DEFAULT_BURST_SIZE * sizeof(struct rte_mbuf *),
+               0,
+               numa_node);
+
+       if (p->soft.def.pkts == NULL)
+               return -ENOMEM;
+
+       return 0;
+}
+
+static void
+default_free(struct pmd_internals *p)
+{
+       rte_free(p->soft.def.pkts);
+}
+
+static void *
+pmd_init(struct pmd_params *params, int numa_node)
+{
+       struct pmd_internals *p;
+       int status;
+
+       p = rte_zmalloc_socket(params->soft.name,
+               sizeof(struct pmd_internals),
+               0,
+               numa_node);
+       if (p == NULL)
+               return NULL;
+
+       memcpy(&p->params, params, sizeof(p->params));
+       rte_eth_dev_get_port_by_name(params->hard.name, &p->hard.port_id);
+
+       /* Default */
+       status = default_init(p, params, numa_node);
+       if (status) {
+               free(p->params.hard.name);
+               rte_free(p);
+               return NULL;
+       }
+
+       return p;
+}
+
+static void
+pmd_free(struct pmd_internals *p)
+{
+       default_free(p);
+
+       free(p->params.hard.name);
+       rte_free(p);
+}
+
+static int
+pmd_ethdev_register(struct rte_vdev_device *vdev,
+       struct pmd_params *params,
+       void *dev_private)
+{
+       struct rte_eth_dev_info hard_info;
+       struct rte_eth_dev *soft_dev;
+       uint32_t hard_speed;
+       int numa_node;
+       uint16_t hard_port_id;
+
+       rte_eth_dev_get_port_by_name(params->hard.name, &hard_port_id);
+       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);
+
+       /* Ethdev entry allocation */
+       soft_dev = rte_eth_dev_allocate(params->soft.name);
+       if (!soft_dev)
+               return -ENOMEM;
+
+       /* dev */
+       soft_dev->rx_pkt_burst = (params->soft.intrusive) ?
+               NULL : /* set up later */
+               pmd_rx_pkt_burst;
+       soft_dev->tx_pkt_burst = pmd_tx_pkt_burst;
+       soft_dev->tx_pkt_prepare = NULL;
+       soft_dev->dev_ops = &pmd_ops;
+       soft_dev->device = &vdev->device;
+
+       /* dev->data */
+       soft_dev->data->dev_private = dev_private;
+       soft_dev->data->dev_link.link_speed = hard_speed;
+       soft_dev->data->dev_link.link_duplex = ETH_LINK_FULL_DUPLEX;
+       soft_dev->data->dev_link.link_autoneg = ETH_LINK_SPEED_FIXED;
+       soft_dev->data->dev_link.link_status = ETH_LINK_DOWN;
+       soft_dev->data->mac_addrs = &eth_addr;
+       soft_dev->data->promiscuous = 1;
+       soft_dev->data->kdrv = RTE_KDRV_NONE;
+       soft_dev->data->numa_node = numa_node;
+       soft_dev->data->dev_flags = RTE_ETH_DEV_DETACHABLE;
+
+       return 0;
+}
+
+static int
+get_string(const char *key __rte_unused, const char *value, void *extra_args)
+{
+       if (!value || !extra_args)
+               return -EINVAL;
+
+       *(char **)extra_args = strdup(value);
+
+       if (!*(char **)extra_args)
+               return -ENOMEM;
+
+       return 0;
+}
+
+static int
+get_uint32(const char *key __rte_unused, const char *value, void *extra_args)
+{
+       if (!value || !extra_args)
+               return -EINVAL;
+
+       *(uint32_t *)extra_args = strtoull(value, NULL, 0);
+
+       return 0;
+}
+
+static int
+pmd_parse_args(struct pmd_params *p, const char *name, const char *params)
+{
+       struct rte_kvargs *kvlist;
+       int ret;
+
+       kvlist = rte_kvargs_parse(params, pmd_valid_args);
+       if (kvlist == NULL)
+               return -EINVAL;
+
+       /* Set default values */
+       memset(p, 0, sizeof(*p));
+       p->soft.name = name;
+       p->soft.intrusive = INTRUSIVE;
+       p->hard.tx_queue_id = SOFTNIC_HARD_TX_QUEUE_ID;
+
+       /* HARD: name (mandatory) */
+       if (rte_kvargs_count(kvlist, PMD_PARAM_HARD_NAME) == 1) {
+               ret = rte_kvargs_process(kvlist, PMD_PARAM_HARD_NAME,
+                       &get_string, &p->hard.name);
+               if (ret < 0)
+                       goto out_free;
+       } else {
+               ret = -EINVAL;
+               goto out_free;
+       }
+
+       /* HARD: tx_queue_id (optional) */
+       if (rte_kvargs_count(kvlist, PMD_PARAM_HARD_TX_QUEUE_ID) == 1) {
+               ret = rte_kvargs_process(kvlist, PMD_PARAM_HARD_TX_QUEUE_ID,
+                       &get_uint32, &p->hard.tx_queue_id);
+               if (ret < 0)
+                       goto out_free;
+       }
+
+out_free:
+       rte_kvargs_free(kvlist);
+       return ret;
+}
+
+static int
+pmd_probe(struct rte_vdev_device *vdev)
+{
+       struct pmd_params p;
+       const char *params;
+       int status;
+
+       struct rte_eth_dev_info hard_info;
+       uint16_t hard_port_id;
+       int numa_node;
+       void *dev_private;
+
+       RTE_LOG(INFO, PMD,
+               "Probing device \"%s\"\n",
+               rte_vdev_device_name(vdev));
+
+       /* Parse input arguments */
+       params = rte_vdev_device_args(vdev);
+       if (!params)
+               return -EINVAL;
+
+       status = pmd_parse_args(&p, rte_vdev_device_name(vdev), params);
+       if (status)
+               return status;
+
+       /* Check input arguments */
+       if (rte_eth_dev_get_port_by_name(p.hard.name, &hard_port_id))
+               return -EINVAL;
+
+       rte_eth_dev_info_get(hard_port_id, &hard_info);
+       numa_node = rte_eth_dev_socket_id(hard_port_id);
+
+       if (p.hard.tx_queue_id >= hard_info.max_tx_queues)
+               return -EINVAL;
+
+       /* Allocate and initialize soft ethdev private data */
+       dev_private = pmd_init(&p, numa_node);
+       if (dev_private == NULL)
+               return -ENOMEM;
+
+       /* Register soft ethdev */
+       RTE_LOG(INFO, PMD,
+               "Creating soft ethdev \"%s\" for hard ethdev \"%s\"\n",
+               p.soft.name, p.hard.name);
+
+       status = pmd_ethdev_register(vdev, &p, dev_private);
+       if (status) {
+               pmd_free(dev_private);
+               return status;
+       }
+
+       return 0;
+}
+
+static int
+pmd_remove(struct rte_vdev_device *vdev)
+{
+       struct rte_eth_dev *dev = NULL;
+       struct pmd_internals *p;
+
+       if (!vdev)
+               return -EINVAL;
+
+       RTE_LOG(INFO, PMD, "Removing device \"%s\"\n",
+               rte_vdev_device_name(vdev));
+
+       /* Find the ethdev entry */
+       dev = rte_eth_dev_allocated(rte_vdev_device_name(vdev));
+       if (dev == NULL)
+               return -ENODEV;
+       p = dev->data->dev_private;
+
+       /* Free device data structures*/
+       pmd_free(p);
+       rte_free(dev->data);
+       rte_eth_dev_release_port(dev);
+
+       return 0;
+}
+
+static struct rte_vdev_driver pmd_softnic_drv = {
+       .probe = pmd_probe,
+       .remove = pmd_remove,
+};
+
+RTE_PMD_REGISTER_VDEV(net_softnic, pmd_softnic_drv);
+RTE_PMD_REGISTER_PARAM_STRING(net_softnic,
+       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
new file mode 100644 (file)
index 0000000..566490a
--- /dev/null
@@ -0,0 +1,67 @@
+/*-
+ *   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.
+ */
+
+#ifndef __INCLUDE_RTE_ETH_SOFTNIC_H__
+#define __INCLUDE_RTE_ETH_SOFTNIC_H__
+
+#include <stdint.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#ifndef SOFTNIC_HARD_TX_QUEUE_ID
+#define SOFTNIC_HARD_TX_QUEUE_ID                       0
+#endif
+
+/**
+ * Run the traffic management function on the softnic device
+ *
+ * This function read the packets from the softnic input queues, insert into
+ * QoS scheduler queues based on mbuf sched field value and transmit the
+ * scheduled packets out through the hard device interface.
+ *
+ * @param portid
+ *    port id of the soft device.
+ * @return
+ *    zero.
+ */
+
+int
+rte_pmd_softnic_run(uint16_t port_id);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* __INCLUDE_RTE_ETH_SOFTNIC_H__ */
diff --git a/drivers/net/softnic/rte_eth_softnic_internals.h b/drivers/net/softnic/rte_eth_softnic_internals.h
new file mode 100644 (file)
index 0000000..08a633f
--- /dev/null
@@ -0,0 +1,114 @@
+/*-
+ *   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.
+ */
+
+#ifndef __INCLUDE_RTE_ETH_SOFTNIC_INTERNALS_H__
+#define __INCLUDE_RTE_ETH_SOFTNIC_INTERNALS_H__
+
+#include <stdint.h>
+
+#include <rte_mbuf.h>
+#include <rte_ethdev.h>
+
+#include "rte_eth_softnic.h"
+
+#ifndef INTRUSIVE
+#define INTRUSIVE                                      0
+#endif
+
+struct pmd_params {
+       /** Parameters for the soft device (to be created) */
+       struct {
+               const char *name; /**< Name */
+               uint32_t flags; /**< Flags */
+
+               /** 0 = Access hard device though API only (potentially slower,
+                *      but safer);
+                *  1 = Access hard device private data structures is allowed
+                *      (potentially faster).
+                */
+               int intrusive;
+       } soft;
+
+       /** Parameters for the hard device (existing) */
+       struct {
+               char *name; /**< Name */
+               uint16_t tx_queue_id; /**< TX queue ID */
+       } hard;
+};
+
+/**
+ * Default Internals
+ */
+
+#ifndef DEFAULT_BURST_SIZE
+#define DEFAULT_BURST_SIZE                             32
+#endif
+
+#ifndef FLUSH_COUNT_THRESHOLD
+#define FLUSH_COUNT_THRESHOLD                  (1 << 17)
+#endif
+
+struct default_internals {
+       struct rte_mbuf **pkts;
+       uint32_t pkts_len;
+       uint32_t txq_pos;
+       uint32_t flush_count;
+};
+
+/**
+ * PMD Internals
+ */
+struct pmd_internals {
+       /** Params */
+       struct pmd_params params;
+
+       /** Soft device */
+       struct {
+               struct default_internals def; /**< Default */
+       } soft;
+
+       /** Hard device */
+       struct {
+               uint16_t port_id;
+       } hard;
+};
+
+struct pmd_rx_queue {
+       /** Hard device */
+       struct {
+               uint16_t port_id;
+               uint16_t rx_queue_id;
+       } hard;
+};
+
+#endif /* __INCLUDE_RTE_ETH_SOFTNIC_INTERNALS_H__ */
diff --git a/drivers/net/softnic/rte_pmd_eth_softnic_version.map b/drivers/net/softnic/rte_pmd_eth_softnic_version.map
new file mode 100644 (file)
index 0000000..fb2cb68
--- /dev/null
@@ -0,0 +1,7 @@
+DPDK_17.11 {
+       global:
+
+       rte_pmd_softnic_run;
+
+       local: *;
+};
index cee21ea..0b8f612 100644 (file)
@@ -68,7 +68,6 @@ _LDLIBS-$(CONFIG_RTE_LIBRTE_IP_FRAG)        += -lrte_ip_frag
 _LDLIBS-$(CONFIG_RTE_LIBRTE_GRO)            += -lrte_gro
 _LDLIBS-$(CONFIG_RTE_LIBRTE_GSO)            += -lrte_gso
 _LDLIBS-$(CONFIG_RTE_LIBRTE_METER)          += -lrte_meter
-_LDLIBS-$(CONFIG_RTE_LIBRTE_SCHED)          += -lrte_sched
 _LDLIBS-$(CONFIG_RTE_LIBRTE_LPM)            += -lrte_lpm
 # librte_acl needs --whole-archive because of weak functions
 _LDLIBS-$(CONFIG_RTE_LIBRTE_ACL)            += --whole-archive
@@ -101,6 +100,7 @@ _LDLIBS-$(CONFIG_RTE_LIBRTE_RING)           += -lrte_ring
 _LDLIBS-$(CONFIG_RTE_LIBRTE_EAL)            += -lrte_eal
 _LDLIBS-$(CONFIG_RTE_LIBRTE_CMDLINE)        += -lrte_cmdline
 _LDLIBS-$(CONFIG_RTE_LIBRTE_REORDER)        += -lrte_reorder
+_LDLIBS-$(CONFIG_RTE_LIBRTE_SCHED)          += -lrte_sched
 
 ifeq ($(CONFIG_RTE_EXEC_ENV_LINUXAPP),y)
 _LDLIBS-$(CONFIG_RTE_LIBRTE_KNI)            += -lrte_kni
@@ -143,6 +143,9 @@ _LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_NULL)       += -lrte_pmd_null
 _LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_PCAP)       += -lrte_pmd_pcap -lpcap
 _LDLIBS-$(CONFIG_RTE_LIBRTE_QEDE_PMD)       += -lrte_pmd_qede
 _LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_RING)       += -lrte_pmd_ring
+ifeq ($(CONFIG_RTE_LIBRTE_SCHED),y)
+_LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_SOFTNIC)      += -lrte_pmd_softnic
+endif
 _LDLIBS-$(CONFIG_RTE_LIBRTE_SFC_EFX_PMD)    += -lrte_pmd_sfc_efx
 _LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_SZEDATA2)   += -lrte_pmd_szedata2 -lsze2
 _LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_TAP)        += -lrte_pmd_tap