eventdev: implement the northbound APIs
authorJerin Jacob <jerin.jacob@caviumnetworks.com>
Tue, 6 Dec 2016 01:51:46 +0000 (07:21 +0530)
committerJerin Jacob <jerin.jacob@caviumnetworks.com>
Tue, 4 Apr 2017 17:12:00 +0000 (19:12 +0200)
This patch implements northbound eventdev API interface using
southbond driver interface

Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
config/common_base
lib/Makefile
lib/librte_eal/common/include/rte_log.h
lib/librte_eventdev/Makefile [new file with mode: 0644]
lib/librte_eventdev/rte_eventdev.c [new file with mode: 0644]
lib/librte_eventdev/rte_eventdev.h
lib/librte_eventdev/rte_eventdev_pmd.h
lib/librte_eventdev/rte_eventdev_version.map [new file with mode: 0644]
mk/rte.app.mk

index d0f1a8b..5951aab 100644 (file)
@@ -468,6 +468,14 @@ CONFIG_RTE_LIBRTE_PMD_CRYPTO_SCHEDULER_DEBUG=n
 #
 CONFIG_RTE_LIBRTE_PMD_NULL_CRYPTO=y
 
+#
+# Compile generic event device library
+#
+CONFIG_RTE_LIBRTE_EVENTDEV=y
+CONFIG_RTE_LIBRTE_EVENTDEV_DEBUG=n
+CONFIG_RTE_EVENT_MAX_DEVS=16
+CONFIG_RTE_EVENT_MAX_QUEUES_PER_DEV=64
+
 #
 # Compile librte_ring
 #
index 531b162..689b7b4 100644 (file)
@@ -51,6 +51,8 @@ DEPDIRS-librte_ether += librte_mbuf
 DIRS-$(CONFIG_RTE_LIBRTE_CRYPTODEV) += librte_cryptodev
 DEPDIRS-librte_cryptodev := librte_eal librte_mempool librte_ring librte_mbuf
 DEPDIRS-librte_cryptodev += librte_kvargs
+DIRS-$(CONFIG_RTE_LIBRTE_EVENTDEV) += librte_eventdev
+DEPDIRS-librte_eventdev := librte_eal librte_mbuf
 DIRS-$(CONFIG_RTE_LIBRTE_VHOST) += librte_vhost
 DEPDIRS-librte_vhost := librte_eal librte_mempool librte_mbuf librte_ether
 DIRS-$(CONFIG_RTE_LIBRTE_HASH) += librte_hash
index 954b96c..c50d22a 100644 (file)
@@ -80,6 +80,7 @@ extern struct rte_logs rte_logs;
 #define RTE_LOGTYPE_MBUF    0x00010000 /**< Log related to mbuf. */
 #define RTE_LOGTYPE_CRYPTODEV 0x00020000 /**< Log related to cryptodev. */
 #define RTE_LOGTYPE_EFD     0x00040000 /**< Log related to EFD. */
+#define RTE_LOGTYPE_EVENTDEV 0x00080000 /**< Log related to eventdev. */
 
 /* these log types can be used in an application */
 #define RTE_LOGTYPE_USER1   0x01000000 /**< User-defined log type 1. */
diff --git a/lib/librte_eventdev/Makefile b/lib/librte_eventdev/Makefile
new file mode 100644 (file)
index 0000000..e06346a
--- /dev/null
@@ -0,0 +1,53 @@
+#   BSD LICENSE
+#
+#   Copyright(c) 2016 Cavium networks. 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 Cavium networks 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_eventdev.a
+
+# library version
+LIBABIVER := 1
+
+# build flags
+CFLAGS += -O3
+CFLAGS += $(WERROR_FLAGS)
+
+# library source files
+SRCS-y += rte_eventdev.c
+
+# export include files
+SYMLINK-y-include += rte_eventdev.h
+SYMLINK-y-include += rte_eventdev_pmd.h
+
+# versioning export map
+EXPORT_MAP := rte_eventdev_version.map
+
+include $(RTE_SDK)/mk/rte.lib.mk
diff --git a/lib/librte_eventdev/rte_eventdev.c b/lib/librte_eventdev/rte_eventdev.c
new file mode 100644 (file)
index 0000000..b13eb00
--- /dev/null
@@ -0,0 +1,986 @@
+/*
+ *   BSD LICENSE
+ *
+ *   Copyright(c) 2016 Cavium networks. 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 Cavium networks 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 <ctype.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <stdarg.h>
+#include <errno.h>
+#include <stdint.h>
+#include <inttypes.h>
+#include <sys/types.h>
+#include <sys/queue.h>
+
+#include <rte_byteorder.h>
+#include <rte_log.h>
+#include <rte_debug.h>
+#include <rte_dev.h>
+#include <rte_pci.h>
+#include <rte_memory.h>
+#include <rte_memcpy.h>
+#include <rte_memzone.h>
+#include <rte_eal.h>
+#include <rte_per_lcore.h>
+#include <rte_lcore.h>
+#include <rte_atomic.h>
+#include <rte_branch_prediction.h>
+#include <rte_common.h>
+#include <rte_malloc.h>
+#include <rte_errno.h>
+
+#include "rte_eventdev.h"
+#include "rte_eventdev_pmd.h"
+
+struct rte_eventdev rte_event_devices[RTE_EVENT_MAX_DEVS];
+
+struct rte_eventdev *rte_eventdevs = &rte_event_devices[0];
+
+static struct rte_eventdev_global eventdev_globals = {
+       .nb_devs                = 0
+};
+
+struct rte_eventdev_global *rte_eventdev_globals = &eventdev_globals;
+
+/* Event dev north bound API implementation */
+
+uint8_t
+rte_event_dev_count(void)
+{
+       return rte_eventdev_globals->nb_devs;
+}
+
+int
+rte_event_dev_get_dev_id(const char *name)
+{
+       int i;
+
+       if (!name)
+               return -EINVAL;
+
+       for (i = 0; i < rte_eventdev_globals->nb_devs; i++)
+               if ((strcmp(rte_event_devices[i].data->name, name)
+                               == 0) &&
+                               (rte_event_devices[i].attached ==
+                                               RTE_EVENTDEV_ATTACHED))
+                       return i;
+       return -ENODEV;
+}
+
+int
+rte_event_dev_socket_id(uint8_t dev_id)
+{
+       struct rte_eventdev *dev;
+
+       RTE_EVENTDEV_VALID_DEVID_OR_ERR_RET(dev_id, -EINVAL);
+       dev = &rte_eventdevs[dev_id];
+
+       return dev->data->socket_id;
+}
+
+int
+rte_event_dev_info_get(uint8_t dev_id, struct rte_event_dev_info *dev_info)
+{
+       struct rte_eventdev *dev;
+
+       RTE_EVENTDEV_VALID_DEVID_OR_ERR_RET(dev_id, -EINVAL);
+       dev = &rte_eventdevs[dev_id];
+
+       if (dev_info == NULL)
+               return -EINVAL;
+
+       memset(dev_info, 0, sizeof(struct rte_event_dev_info));
+
+       RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->dev_infos_get, -ENOTSUP);
+       (*dev->dev_ops->dev_infos_get)(dev, dev_info);
+
+       dev_info->dequeue_timeout_ns = dev->data->dev_conf.dequeue_timeout_ns;
+
+       dev_info->pci_dev = dev->pci_dev;
+       return 0;
+}
+
+static inline int
+rte_event_dev_queue_config(struct rte_eventdev *dev, uint8_t nb_queues)
+{
+       uint8_t old_nb_queues = dev->data->nb_queues;
+       uint8_t *queues_prio;
+       unsigned int i;
+
+       RTE_EDEV_LOG_DEBUG("Setup %d queues on device %u", nb_queues,
+                        dev->data->dev_id);
+
+       /* First time configuration */
+       if (dev->data->queues_prio == NULL && nb_queues != 0) {
+               /* Allocate memory to store queue priority */
+               dev->data->queues_prio = rte_zmalloc_socket(
+                               "eventdev->data->queues_prio",
+                               sizeof(dev->data->queues_prio[0]) * nb_queues,
+                               RTE_CACHE_LINE_SIZE, dev->data->socket_id);
+               if (dev->data->queues_prio == NULL) {
+                       dev->data->nb_queues = 0;
+                       RTE_EDEV_LOG_ERR("failed to get mem for queue priority,"
+                                       "nb_queues %u", nb_queues);
+                       return -(ENOMEM);
+               }
+       /* Re-configure */
+       } else if (dev->data->queues_prio != NULL && nb_queues != 0) {
+               RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->queue_release, -ENOTSUP);
+
+               for (i = nb_queues; i < old_nb_queues; i++)
+                       (*dev->dev_ops->queue_release)(dev, i);
+
+               /* Re allocate memory to store queue priority */
+               queues_prio = dev->data->queues_prio;
+               queues_prio = rte_realloc(queues_prio,
+                               sizeof(queues_prio[0]) * nb_queues,
+                               RTE_CACHE_LINE_SIZE);
+               if (queues_prio == NULL) {
+                       RTE_EDEV_LOG_ERR("failed to realloc queue priority,"
+                                               " nb_queues %u", nb_queues);
+                       return -(ENOMEM);
+               }
+               dev->data->queues_prio = queues_prio;
+
+               if (nb_queues > old_nb_queues) {
+                       uint8_t new_qs = nb_queues - old_nb_queues;
+
+                       memset(queues_prio + old_nb_queues, 0,
+                               sizeof(queues_prio[0]) * new_qs);
+               }
+       } else if (dev->data->queues_prio != NULL && nb_queues == 0) {
+               RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->queue_release, -ENOTSUP);
+
+               for (i = nb_queues; i < old_nb_queues; i++)
+                       (*dev->dev_ops->queue_release)(dev, i);
+       }
+
+       dev->data->nb_queues = nb_queues;
+       return 0;
+}
+
+static inline int
+rte_event_dev_port_config(struct rte_eventdev *dev, uint8_t nb_ports)
+{
+       uint8_t old_nb_ports = dev->data->nb_ports;
+       void **ports;
+       uint16_t *links_map;
+       uint8_t *ports_dequeue_depth;
+       uint8_t *ports_enqueue_depth;
+       unsigned int i;
+
+       RTE_EDEV_LOG_DEBUG("Setup %d ports on device %u", nb_ports,
+                        dev->data->dev_id);
+
+       /* First time configuration */
+       if (dev->data->ports == NULL && nb_ports != 0) {
+               dev->data->ports = rte_zmalloc_socket("eventdev->data->ports",
+                               sizeof(dev->data->ports[0]) * nb_ports,
+                               RTE_CACHE_LINE_SIZE, dev->data->socket_id);
+               if (dev->data->ports == NULL) {
+                       dev->data->nb_ports = 0;
+                       RTE_EDEV_LOG_ERR("failed to get mem for port meta data,"
+                                       "nb_ports %u", nb_ports);
+                       return -(ENOMEM);
+               }
+
+               /* Allocate memory to store ports dequeue depth */
+               dev->data->ports_dequeue_depth =
+                       rte_zmalloc_socket("eventdev->ports_dequeue_depth",
+                       sizeof(dev->data->ports_dequeue_depth[0]) * nb_ports,
+                       RTE_CACHE_LINE_SIZE, dev->data->socket_id);
+               if (dev->data->ports_dequeue_depth == NULL) {
+                       dev->data->nb_ports = 0;
+                       RTE_EDEV_LOG_ERR("failed to get mem for port deq meta,"
+                                       "nb_ports %u", nb_ports);
+                       return -(ENOMEM);
+               }
+
+               /* Allocate memory to store ports enqueue depth */
+               dev->data->ports_enqueue_depth =
+                       rte_zmalloc_socket("eventdev->ports_enqueue_depth",
+                       sizeof(dev->data->ports_enqueue_depth[0]) * nb_ports,
+                       RTE_CACHE_LINE_SIZE, dev->data->socket_id);
+               if (dev->data->ports_enqueue_depth == NULL) {
+                       dev->data->nb_ports = 0;
+                       RTE_EDEV_LOG_ERR("failed to get mem for port enq meta,"
+                                       "nb_ports %u", nb_ports);
+                       return -(ENOMEM);
+               }
+
+               /* Allocate memory to store queue to port link connection */
+               dev->data->links_map =
+                       rte_zmalloc_socket("eventdev->links_map",
+                       sizeof(dev->data->links_map[0]) * nb_ports *
+                       RTE_EVENT_MAX_QUEUES_PER_DEV,
+                       RTE_CACHE_LINE_SIZE, dev->data->socket_id);
+               if (dev->data->links_map == NULL) {
+                       dev->data->nb_ports = 0;
+                       RTE_EDEV_LOG_ERR("failed to get mem for port_map area,"
+                                       "nb_ports %u", nb_ports);
+                       return -(ENOMEM);
+               }
+       } else if (dev->data->ports != NULL && nb_ports != 0) {/* re-config */
+               RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->port_release, -ENOTSUP);
+
+               ports = dev->data->ports;
+               ports_dequeue_depth = dev->data->ports_dequeue_depth;
+               ports_enqueue_depth = dev->data->ports_enqueue_depth;
+               links_map = dev->data->links_map;
+
+               for (i = nb_ports; i < old_nb_ports; i++)
+                       (*dev->dev_ops->port_release)(ports[i]);
+
+               /* Realloc memory for ports */
+               ports = rte_realloc(ports, sizeof(ports[0]) * nb_ports,
+                               RTE_CACHE_LINE_SIZE);
+               if (ports == NULL) {
+                       RTE_EDEV_LOG_ERR("failed to realloc port meta data,"
+                                               " nb_ports %u", nb_ports);
+                       return -(ENOMEM);
+               }
+
+               /* Realloc memory for ports_dequeue_depth */
+               ports_dequeue_depth = rte_realloc(ports_dequeue_depth,
+                       sizeof(ports_dequeue_depth[0]) * nb_ports,
+                       RTE_CACHE_LINE_SIZE);
+               if (ports_dequeue_depth == NULL) {
+                       RTE_EDEV_LOG_ERR("failed to realloc port dequeue meta,"
+                                               " nb_ports %u", nb_ports);
+                       return -(ENOMEM);
+               }
+
+               /* Realloc memory for ports_enqueue_depth */
+               ports_enqueue_depth = rte_realloc(ports_enqueue_depth,
+                       sizeof(ports_enqueue_depth[0]) * nb_ports,
+                       RTE_CACHE_LINE_SIZE);
+               if (ports_enqueue_depth == NULL) {
+                       RTE_EDEV_LOG_ERR("failed to realloc port enqueue meta,"
+                                               " nb_ports %u", nb_ports);
+                       return -(ENOMEM);
+               }
+
+               /* Realloc memory to store queue to port link connection */
+               links_map = rte_realloc(links_map,
+                       sizeof(dev->data->links_map[0]) * nb_ports *
+                       RTE_EVENT_MAX_QUEUES_PER_DEV,
+                       RTE_CACHE_LINE_SIZE);
+               if (dev->data->links_map == NULL) {
+                       dev->data->nb_ports = 0;
+                       RTE_EDEV_LOG_ERR("failed to realloc mem for port_map,"
+                                       "nb_ports %u", nb_ports);
+                       return -(ENOMEM);
+               }
+
+               if (nb_ports > old_nb_ports) {
+                       uint8_t new_ps = nb_ports - old_nb_ports;
+
+                       memset(ports + old_nb_ports, 0,
+                               sizeof(ports[0]) * new_ps);
+                       memset(ports_dequeue_depth + old_nb_ports, 0,
+                               sizeof(ports_dequeue_depth[0]) * new_ps);
+                       memset(ports_enqueue_depth + old_nb_ports, 0,
+                               sizeof(ports_enqueue_depth[0]) * new_ps);
+                       memset(links_map +
+                               (old_nb_ports * RTE_EVENT_MAX_QUEUES_PER_DEV),
+                               0, sizeof(ports_enqueue_depth[0]) * new_ps);
+               }
+
+               dev->data->ports = ports;
+               dev->data->ports_dequeue_depth = ports_dequeue_depth;
+               dev->data->ports_enqueue_depth = ports_enqueue_depth;
+               dev->data->links_map = links_map;
+       } else if (dev->data->ports != NULL && nb_ports == 0) {
+               RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->port_release, -ENOTSUP);
+
+               ports = dev->data->ports;
+               for (i = nb_ports; i < old_nb_ports; i++)
+                       (*dev->dev_ops->port_release)(ports[i]);
+       }
+
+       dev->data->nb_ports = nb_ports;
+       return 0;
+}
+
+int
+rte_event_dev_configure(uint8_t dev_id,
+                       const struct rte_event_dev_config *dev_conf)
+{
+       struct rte_eventdev *dev;
+       struct rte_event_dev_info info;
+       int diag;
+
+       RTE_EVENTDEV_VALID_DEVID_OR_ERR_RET(dev_id, -EINVAL);
+       dev = &rte_eventdevs[dev_id];
+
+       RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->dev_infos_get, -ENOTSUP);
+       RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->dev_configure, -ENOTSUP);
+
+       if (dev->data->dev_started) {
+               RTE_EDEV_LOG_ERR(
+                   "device %d must be stopped to allow configuration", dev_id);
+               return -EBUSY;
+       }
+
+       if (dev_conf == NULL)
+               return -EINVAL;
+
+       (*dev->dev_ops->dev_infos_get)(dev, &info);
+
+       /* Check dequeue_timeout_ns value is in limit */
+       if (!dev_conf->event_dev_cfg & RTE_EVENT_DEV_CFG_PER_DEQUEUE_TIMEOUT) {
+               if (dev_conf->dequeue_timeout_ns < info.min_dequeue_timeout_ns
+                       || dev_conf->dequeue_timeout_ns >
+                                info.max_dequeue_timeout_ns) {
+                       RTE_EDEV_LOG_ERR("dev%d invalid dequeue_timeout_ns=%d"
+                       " min_dequeue_timeout_ns=%d max_dequeue_timeout_ns=%d",
+                       dev_id, dev_conf->dequeue_timeout_ns,
+                       info.min_dequeue_timeout_ns,
+                       info.max_dequeue_timeout_ns);
+                       return -EINVAL;
+               }
+       }
+
+       /* Check nb_events_limit is in limit */
+       if (dev_conf->nb_events_limit > info.max_num_events) {
+               RTE_EDEV_LOG_ERR("dev%d nb_events_limit=%d > max_num_events=%d",
+               dev_id, dev_conf->nb_events_limit, info.max_num_events);
+               return -EINVAL;
+       }
+
+       /* Check nb_event_queues is in limit */
+       if (!dev_conf->nb_event_queues) {
+               RTE_EDEV_LOG_ERR("dev%d nb_event_queues cannot be zero",
+                                       dev_id);
+               return -EINVAL;
+       }
+       if (dev_conf->nb_event_queues > info.max_event_queues) {
+               RTE_EDEV_LOG_ERR("%d nb_event_queues=%d > max_event_queues=%d",
+               dev_id, dev_conf->nb_event_queues, info.max_event_queues);
+               return -EINVAL;
+       }
+
+       /* Check nb_event_ports is in limit */
+       if (!dev_conf->nb_event_ports) {
+               RTE_EDEV_LOG_ERR("dev%d nb_event_ports cannot be zero", dev_id);
+               return -EINVAL;
+       }
+       if (dev_conf->nb_event_ports > info.max_event_ports) {
+               RTE_EDEV_LOG_ERR("id%d nb_event_ports=%d > max_event_ports= %d",
+               dev_id, dev_conf->nb_event_ports, info.max_event_ports);
+               return -EINVAL;
+       }
+
+       /* Check nb_event_queue_flows is in limit */
+       if (!dev_conf->nb_event_queue_flows) {
+               RTE_EDEV_LOG_ERR("dev%d nb_flows cannot be zero", dev_id);
+               return -EINVAL;
+       }
+       if (dev_conf->nb_event_queue_flows > info.max_event_queue_flows) {
+               RTE_EDEV_LOG_ERR("dev%d nb_flows=%x > max_flows=%x",
+               dev_id, dev_conf->nb_event_queue_flows,
+               info.max_event_queue_flows);
+               return -EINVAL;
+       }
+
+       /* Check nb_event_port_dequeue_depth is in limit */
+       if (!dev_conf->nb_event_port_dequeue_depth) {
+               RTE_EDEV_LOG_ERR("dev%d nb_dequeue_depth cannot be zero",
+                                       dev_id);
+               return -EINVAL;
+       }
+       if (dev_conf->nb_event_port_dequeue_depth >
+                        info.max_event_port_dequeue_depth) {
+               RTE_EDEV_LOG_ERR("dev%d nb_dq_depth=%d > max_dq_depth=%d",
+               dev_id, dev_conf->nb_event_port_dequeue_depth,
+               info.max_event_port_dequeue_depth);
+               return -EINVAL;
+       }
+
+       /* Check nb_event_port_enqueue_depth is in limit */
+       if (!dev_conf->nb_event_port_enqueue_depth) {
+               RTE_EDEV_LOG_ERR("dev%d nb_enqueue_depth cannot be zero",
+                                       dev_id);
+               return -EINVAL;
+       }
+       if (dev_conf->nb_event_port_enqueue_depth >
+                        info.max_event_port_enqueue_depth) {
+               RTE_EDEV_LOG_ERR("dev%d nb_enq_depth=%d > max_enq_depth=%d",
+               dev_id, dev_conf->nb_event_port_enqueue_depth,
+               info.max_event_port_enqueue_depth);
+               return -EINVAL;
+       }
+
+       /* Copy the dev_conf parameter into the dev structure */
+       memcpy(&dev->data->dev_conf, dev_conf, sizeof(dev->data->dev_conf));
+
+       /* Setup new number of queues and reconfigure device. */
+       diag = rte_event_dev_queue_config(dev, dev_conf->nb_event_queues);
+       if (diag != 0) {
+               RTE_EDEV_LOG_ERR("dev%d rte_event_dev_queue_config = %d",
+                               dev_id, diag);
+               return diag;
+       }
+
+       /* Setup new number of ports and reconfigure device. */
+       diag = rte_event_dev_port_config(dev, dev_conf->nb_event_ports);
+       if (diag != 0) {
+               rte_event_dev_queue_config(dev, 0);
+               RTE_EDEV_LOG_ERR("dev%d rte_event_dev_port_config = %d",
+                               dev_id, diag);
+               return diag;
+       }
+
+       /* Configure the device */
+       diag = (*dev->dev_ops->dev_configure)(dev);
+       if (diag != 0) {
+               RTE_EDEV_LOG_ERR("dev%d dev_configure = %d", dev_id, diag);
+               rte_event_dev_queue_config(dev, 0);
+               rte_event_dev_port_config(dev, 0);
+       }
+
+       dev->data->event_dev_cap = info.event_dev_cap;
+       return diag;
+}
+
+static inline int
+is_valid_queue(struct rte_eventdev *dev, uint8_t queue_id)
+{
+       if (queue_id < dev->data->nb_queues && queue_id <
+                               RTE_EVENT_MAX_QUEUES_PER_DEV)
+               return 1;
+       else
+               return 0;
+}
+
+int
+rte_event_queue_default_conf_get(uint8_t dev_id, uint8_t queue_id,
+                                struct rte_event_queue_conf *queue_conf)
+{
+       struct rte_eventdev *dev;
+
+       RTE_EVENTDEV_VALID_DEVID_OR_ERR_RET(dev_id, -EINVAL);
+       dev = &rte_eventdevs[dev_id];
+
+       if (queue_conf == NULL)
+               return -EINVAL;
+
+       if (!is_valid_queue(dev, queue_id)) {
+               RTE_EDEV_LOG_ERR("Invalid queue_id=%" PRIu8, queue_id);
+               return -EINVAL;
+       }
+
+       RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->queue_def_conf, -ENOTSUP);
+       memset(queue_conf, 0, sizeof(struct rte_event_queue_conf));
+       (*dev->dev_ops->queue_def_conf)(dev, queue_id, queue_conf);
+       return 0;
+}
+
+static inline int
+is_valid_atomic_queue_conf(const struct rte_event_queue_conf *queue_conf)
+{
+       if (queue_conf && (
+               ((queue_conf->event_queue_cfg &
+                       RTE_EVENT_QUEUE_CFG_TYPE_MASK)
+                       == RTE_EVENT_QUEUE_CFG_ALL_TYPES) ||
+               ((queue_conf->event_queue_cfg &
+                       RTE_EVENT_QUEUE_CFG_TYPE_MASK)
+                       == RTE_EVENT_QUEUE_CFG_ATOMIC_ONLY)
+               ))
+               return 1;
+       else
+               return 0;
+}
+
+static inline int
+is_valid_ordered_queue_conf(const struct rte_event_queue_conf *queue_conf)
+{
+       if (queue_conf && (
+               ((queue_conf->event_queue_cfg &
+                       RTE_EVENT_QUEUE_CFG_TYPE_MASK)
+                       == RTE_EVENT_QUEUE_CFG_ALL_TYPES) ||
+               ((queue_conf->event_queue_cfg &
+                       RTE_EVENT_QUEUE_CFG_TYPE_MASK)
+                       == RTE_EVENT_QUEUE_CFG_ORDERED_ONLY)
+               ))
+               return 1;
+       else
+               return 0;
+}
+
+
+int
+rte_event_queue_setup(uint8_t dev_id, uint8_t queue_id,
+                     const struct rte_event_queue_conf *queue_conf)
+{
+       struct rte_eventdev *dev;
+       struct rte_event_queue_conf def_conf;
+
+       RTE_EVENTDEV_VALID_DEVID_OR_ERR_RET(dev_id, -EINVAL);
+       dev = &rte_eventdevs[dev_id];
+
+       if (!is_valid_queue(dev, queue_id)) {
+               RTE_EDEV_LOG_ERR("Invalid queue_id=%" PRIu8, queue_id);
+               return -EINVAL;
+       }
+
+       /* Check nb_atomic_flows limit */
+       if (is_valid_atomic_queue_conf(queue_conf)) {
+               if (queue_conf->nb_atomic_flows == 0 ||
+                   queue_conf->nb_atomic_flows >
+                       dev->data->dev_conf.nb_event_queue_flows) {
+                       RTE_EDEV_LOG_ERR(
+               "dev%d queue%d Invalid nb_atomic_flows=%d max_flows=%d",
+                       dev_id, queue_id, queue_conf->nb_atomic_flows,
+                       dev->data->dev_conf.nb_event_queue_flows);
+                       return -EINVAL;
+               }
+       }
+
+       /* Check nb_atomic_order_sequences limit */
+       if (is_valid_ordered_queue_conf(queue_conf)) {
+               if (queue_conf->nb_atomic_order_sequences == 0 ||
+                   queue_conf->nb_atomic_order_sequences >
+                       dev->data->dev_conf.nb_event_queue_flows) {
+                       RTE_EDEV_LOG_ERR(
+               "dev%d queue%d Invalid nb_atomic_order_seq=%d max_flows=%d",
+                       dev_id, queue_id, queue_conf->nb_atomic_order_sequences,
+                       dev->data->dev_conf.nb_event_queue_flows);
+                       return -EINVAL;
+               }
+       }
+
+       if (dev->data->dev_started) {
+               RTE_EDEV_LOG_ERR(
+                   "device %d must be stopped to allow queue setup", dev_id);
+               return -EBUSY;
+       }
+
+       RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->queue_setup, -ENOTSUP);
+
+       if (queue_conf == NULL) {
+               RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->queue_def_conf,
+                                       -ENOTSUP);
+               (*dev->dev_ops->queue_def_conf)(dev, queue_id, &def_conf);
+               def_conf.event_queue_cfg = RTE_EVENT_QUEUE_CFG_DEFAULT;
+               queue_conf = &def_conf;
+       }
+
+       dev->data->queues_prio[queue_id] = queue_conf->priority;
+       return (*dev->dev_ops->queue_setup)(dev, queue_id, queue_conf);
+}
+
+uint8_t
+rte_event_queue_count(uint8_t dev_id)
+{
+       struct rte_eventdev *dev;
+
+       dev = &rte_eventdevs[dev_id];
+       return dev->data->nb_queues;
+}
+
+uint8_t
+rte_event_queue_priority(uint8_t dev_id, uint8_t queue_id)
+{
+       struct rte_eventdev *dev;
+
+       dev = &rte_eventdevs[dev_id];
+       if (dev->data->event_dev_cap & RTE_EVENT_DEV_CAP_QUEUE_QOS)
+               return dev->data->queues_prio[queue_id];
+       else
+               return RTE_EVENT_DEV_PRIORITY_NORMAL;
+}
+
+static inline int
+is_valid_port(struct rte_eventdev *dev, uint8_t port_id)
+{
+       if (port_id < dev->data->nb_ports)
+               return 1;
+       else
+               return 0;
+}
+
+int
+rte_event_port_default_conf_get(uint8_t dev_id, uint8_t port_id,
+                                struct rte_event_port_conf *port_conf)
+{
+       struct rte_eventdev *dev;
+
+       RTE_EVENTDEV_VALID_DEVID_OR_ERR_RET(dev_id, -EINVAL);
+       dev = &rte_eventdevs[dev_id];
+
+       if (port_conf == NULL)
+               return -EINVAL;
+
+       if (!is_valid_port(dev, port_id)) {
+               RTE_EDEV_LOG_ERR("Invalid port_id=%" PRIu8, port_id);
+               return -EINVAL;
+       }
+
+       RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->port_def_conf, -ENOTSUP);
+       memset(port_conf, 0, sizeof(struct rte_event_port_conf));
+       (*dev->dev_ops->port_def_conf)(dev, port_id, port_conf);
+       return 0;
+}
+
+int
+rte_event_port_setup(uint8_t dev_id, uint8_t port_id,
+                    const struct rte_event_port_conf *port_conf)
+{
+       struct rte_eventdev *dev;
+       struct rte_event_port_conf def_conf;
+       int diag;
+
+       RTE_EVENTDEV_VALID_DEVID_OR_ERR_RET(dev_id, -EINVAL);
+       dev = &rte_eventdevs[dev_id];
+
+       if (!is_valid_port(dev, port_id)) {
+               RTE_EDEV_LOG_ERR("Invalid port_id=%" PRIu8, port_id);
+               return -EINVAL;
+       }
+
+       /* Check new_event_threshold limit */
+       if ((port_conf && !port_conf->new_event_threshold) ||
+                       (port_conf && port_conf->new_event_threshold >
+                                dev->data->dev_conf.nb_events_limit)) {
+               RTE_EDEV_LOG_ERR(
+                  "dev%d port%d Invalid event_threshold=%d nb_events_limit=%d",
+                       dev_id, port_id, port_conf->new_event_threshold,
+                       dev->data->dev_conf.nb_events_limit);
+               return -EINVAL;
+       }
+
+       /* Check dequeue_depth limit */
+       if ((port_conf && !port_conf->dequeue_depth) ||
+                       (port_conf && port_conf->dequeue_depth >
+               dev->data->dev_conf.nb_event_port_dequeue_depth)) {
+               RTE_EDEV_LOG_ERR(
+                  "dev%d port%d Invalid dequeue depth=%d max_dequeue_depth=%d",
+                       dev_id, port_id, port_conf->dequeue_depth,
+                       dev->data->dev_conf.nb_event_port_dequeue_depth);
+               return -EINVAL;
+       }
+
+       /* Check enqueue_depth limit */
+       if ((port_conf && !port_conf->enqueue_depth) ||
+                       (port_conf && port_conf->enqueue_depth >
+               dev->data->dev_conf.nb_event_port_enqueue_depth)) {
+               RTE_EDEV_LOG_ERR(
+                  "dev%d port%d Invalid enqueue depth=%d max_enqueue_depth=%d",
+                       dev_id, port_id, port_conf->enqueue_depth,
+                       dev->data->dev_conf.nb_event_port_enqueue_depth);
+               return -EINVAL;
+       }
+
+       if (dev->data->dev_started) {
+               RTE_EDEV_LOG_ERR(
+                   "device %d must be stopped to allow port setup", dev_id);
+               return -EBUSY;
+       }
+
+       RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->port_setup, -ENOTSUP);
+
+       if (port_conf == NULL) {
+               RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->port_def_conf,
+                                       -ENOTSUP);
+               (*dev->dev_ops->port_def_conf)(dev, port_id, &def_conf);
+               port_conf = &def_conf;
+       }
+
+       dev->data->ports_dequeue_depth[port_id] =
+                       port_conf->dequeue_depth;
+       dev->data->ports_enqueue_depth[port_id] =
+                       port_conf->enqueue_depth;
+
+       diag = (*dev->dev_ops->port_setup)(dev, port_id, port_conf);
+
+       /* Unlink all the queues from this port(default state after setup) */
+       if (!diag)
+               diag = rte_event_port_unlink(dev_id, port_id, NULL, 0);
+
+       if (diag < 0)
+               return diag;
+
+       return 0;
+}
+
+uint8_t
+rte_event_port_dequeue_depth(uint8_t dev_id, uint8_t port_id)
+{
+       struct rte_eventdev *dev;
+
+       dev = &rte_eventdevs[dev_id];
+       return dev->data->ports_dequeue_depth[port_id];
+}
+
+uint8_t
+rte_event_port_enqueue_depth(uint8_t dev_id, uint8_t port_id)
+{
+       struct rte_eventdev *dev;
+
+       dev = &rte_eventdevs[dev_id];
+       return dev->data->ports_enqueue_depth[port_id];
+}
+
+uint8_t
+rte_event_port_count(uint8_t dev_id)
+{
+       struct rte_eventdev *dev;
+
+       dev = &rte_eventdevs[dev_id];
+       return dev->data->nb_ports;
+}
+
+int
+rte_event_port_link(uint8_t dev_id, uint8_t port_id,
+                   const uint8_t queues[], const uint8_t priorities[],
+                   uint16_t nb_links)
+{
+       struct rte_eventdev *dev;
+       uint8_t queues_list[RTE_EVENT_MAX_QUEUES_PER_DEV];
+       uint8_t priorities_list[RTE_EVENT_MAX_QUEUES_PER_DEV];
+       uint16_t *links_map;
+       int i, diag;
+
+       RTE_EVENTDEV_VALID_DEVID_OR_ERR_RET(dev_id, -EINVAL);
+       dev = &rte_eventdevs[dev_id];
+       RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->port_link, -ENOTSUP);
+
+       if (!is_valid_port(dev, port_id)) {
+               RTE_EDEV_LOG_ERR("Invalid port_id=%" PRIu8, port_id);
+               return -EINVAL;
+       }
+
+       if (queues == NULL) {
+               for (i = 0; i < dev->data->nb_queues; i++)
+                       queues_list[i] = i;
+
+               queues = queues_list;
+               nb_links = dev->data->nb_queues;
+       }
+
+       if (priorities == NULL) {
+               for (i = 0; i < nb_links; i++)
+                       priorities_list[i] = RTE_EVENT_DEV_PRIORITY_NORMAL;
+
+               priorities = priorities_list;
+       }
+
+       for (i = 0; i < nb_links; i++)
+               if (queues[i] >= RTE_EVENT_MAX_QUEUES_PER_DEV)
+                       return -EINVAL;
+
+       diag = (*dev->dev_ops->port_link)(dev->data->ports[port_id], queues,
+                                               priorities, nb_links);
+       if (diag < 0)
+               return diag;
+
+       links_map = dev->data->links_map;
+       /* Point links_map to this port specific area */
+       links_map += (port_id * RTE_EVENT_MAX_QUEUES_PER_DEV);
+       for (i = 0; i < diag; i++)
+               links_map[queues[i]] = (uint8_t)priorities[i];
+
+       return diag;
+}
+
+#define EVENT_QUEUE_SERVICE_PRIORITY_INVALID (0xdead)
+
+int
+rte_event_port_unlink(uint8_t dev_id, uint8_t port_id,
+                     uint8_t queues[], uint16_t nb_unlinks)
+{
+       struct rte_eventdev *dev;
+       uint8_t all_queues[RTE_EVENT_MAX_QUEUES_PER_DEV];
+       int i, diag;
+       uint16_t *links_map;
+
+       RTE_EVENTDEV_VALID_DEVID_OR_ERR_RET(dev_id, -EINVAL);
+       dev = &rte_eventdevs[dev_id];
+       RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->port_unlink, -ENOTSUP);
+
+       if (!is_valid_port(dev, port_id)) {
+               RTE_EDEV_LOG_ERR("Invalid port_id=%" PRIu8, port_id);
+               return -EINVAL;
+       }
+
+       if (queues == NULL) {
+               for (i = 0; i < dev->data->nb_queues; i++)
+                       all_queues[i] = i;
+               queues = all_queues;
+               nb_unlinks = dev->data->nb_queues;
+       }
+
+       for (i = 0; i < nb_unlinks; i++)
+               if (queues[i] >= RTE_EVENT_MAX_QUEUES_PER_DEV)
+                       return -EINVAL;
+
+       diag = (*dev->dev_ops->port_unlink)(dev->data->ports[port_id], queues,
+                                       nb_unlinks);
+
+       if (diag < 0)
+               return diag;
+
+       links_map = dev->data->links_map;
+       /* Point links_map to this port specific area */
+       links_map += (port_id * RTE_EVENT_MAX_QUEUES_PER_DEV);
+       for (i = 0; i < diag; i++)
+               links_map[queues[i]] = EVENT_QUEUE_SERVICE_PRIORITY_INVALID;
+
+       return diag;
+}
+
+int
+rte_event_port_links_get(uint8_t dev_id, uint8_t port_id,
+                        uint8_t queues[], uint8_t priorities[])
+{
+       struct rte_eventdev *dev;
+       uint16_t *links_map;
+       int i, count = 0;
+
+       RTE_EVENTDEV_VALID_DEVID_OR_ERR_RET(dev_id, -EINVAL);
+       dev = &rte_eventdevs[dev_id];
+       if (!is_valid_port(dev, port_id)) {
+               RTE_EDEV_LOG_ERR("Invalid port_id=%" PRIu8, port_id);
+               return -EINVAL;
+       }
+
+       links_map = dev->data->links_map;
+       /* Point links_map to this port specific area */
+       links_map += (port_id * RTE_EVENT_MAX_QUEUES_PER_DEV);
+       for (i = 0; i < RTE_EVENT_MAX_QUEUES_PER_DEV; i++) {
+               if (links_map[i] != EVENT_QUEUE_SERVICE_PRIORITY_INVALID) {
+                       queues[count] = i;
+                       priorities[count] = (uint8_t)links_map[i];
+                       ++count;
+               }
+       }
+       return count;
+}
+
+int
+rte_event_dequeue_timeout_ticks(uint8_t dev_id, uint64_t ns,
+                                uint64_t *timeout_ticks)
+{
+       struct rte_eventdev *dev;
+
+       RTE_EVENTDEV_VALID_DEVID_OR_ERR_RET(dev_id, -EINVAL);
+       dev = &rte_eventdevs[dev_id];
+       RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->timeout_ticks, -ENOTSUP);
+
+       if (timeout_ticks == NULL)
+               return -EINVAL;
+
+       (*dev->dev_ops->timeout_ticks)(dev, ns, timeout_ticks);
+       return 0;
+}
+
+int
+rte_event_dev_dump(uint8_t dev_id, FILE *f)
+{
+       struct rte_eventdev *dev;
+
+       RTE_EVENTDEV_VALID_DEVID_OR_ERR_RET(dev_id, -EINVAL);
+       dev = &rte_eventdevs[dev_id];
+       RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->dump, -ENOTSUP);
+
+       (*dev->dev_ops->dump)(dev, f);
+       return 0;
+
+}
+
+int
+rte_event_dev_start(uint8_t dev_id)
+{
+       struct rte_eventdev *dev;
+       int diag;
+
+       RTE_EDEV_LOG_DEBUG("Start dev_id=%" PRIu8, dev_id);
+
+       RTE_EVENTDEV_VALID_DEVID_OR_ERR_RET(dev_id, -EINVAL);
+       dev = &rte_eventdevs[dev_id];
+       RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->dev_start, -ENOTSUP);
+
+       if (dev->data->dev_started != 0) {
+               RTE_EDEV_LOG_ERR("Device with dev_id=%" PRIu8 "already started",
+                       dev_id);
+               return 0;
+       }
+
+       diag = (*dev->dev_ops->dev_start)(dev);
+       if (diag == 0)
+               dev->data->dev_started = 1;
+       else
+               return diag;
+
+       return 0;
+}
+
+void
+rte_event_dev_stop(uint8_t dev_id)
+{
+       struct rte_eventdev *dev;
+
+       RTE_EDEV_LOG_DEBUG("Stop dev_id=%" PRIu8, dev_id);
+
+       RTE_EVENTDEV_VALID_DEVID_OR_RET(dev_id);
+       dev = &rte_eventdevs[dev_id];
+       RTE_FUNC_PTR_OR_RET(*dev->dev_ops->dev_stop);
+
+       if (dev->data->dev_started == 0) {
+               RTE_EDEV_LOG_ERR("Device with dev_id=%" PRIu8 "already stopped",
+                       dev_id);
+               return;
+       }
+
+       dev->data->dev_started = 0;
+       (*dev->dev_ops->dev_stop)(dev);
+}
+
+int
+rte_event_dev_close(uint8_t dev_id)
+{
+       struct rte_eventdev *dev;
+
+       RTE_EVENTDEV_VALID_DEVID_OR_ERR_RET(dev_id, -EINVAL);
+       dev = &rte_eventdevs[dev_id];
+       RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->dev_close, -ENOTSUP);
+
+       /* Device must be stopped before it can be closed */
+       if (dev->data->dev_started == 1) {
+               RTE_EDEV_LOG_ERR("Device %u must be stopped before closing",
+                               dev_id);
+               return -EBUSY;
+       }
+
+       return (*dev->dev_ops->dev_close)(dev);
+}
index 014e1ec..e1bd05f 100644 (file)
@@ -972,6 +972,8 @@ struct rte_event {
        };
 };
 
+
+struct rte_eventdev_driver;
 struct rte_eventdev_ops;
 struct rte_eventdev;
 
@@ -993,6 +995,49 @@ typedef uint16_t (*event_dequeue_burst_t)(void *port, struct rte_event ev[],
                uint16_t nb_events, uint64_t timeout_ticks);
 /**< @internal Dequeue burst of events from port of a device */
 
+#define RTE_EVENTDEV_NAME_MAX_LEN      (64)
+/**< @internal Max length of name of event PMD */
+
+/**
+ * @internal
+ * The data part, with no function pointers, associated with each device.
+ *
+ * This structure is safe to place in shared memory to be common among
+ * different processes in a multi-process configuration.
+ */
+struct rte_eventdev_data {
+       int socket_id;
+       /**< Socket ID where memory is allocated */
+       uint8_t dev_id;
+       /**< Device ID for this instance */
+       uint8_t nb_queues;
+       /**< Number of event queues. */
+       uint8_t nb_ports;
+       /**< Number of event ports. */
+       void **ports;
+       /**< Array of pointers to ports. */
+       uint8_t *ports_dequeue_depth;
+       /**< Array of port dequeue depth. */
+       uint8_t *ports_enqueue_depth;
+       /**< Array of port enqueue depth. */
+       uint8_t *queues_prio;
+       /**< Array of queue priority. */
+       uint16_t *links_map;
+       /**< Memory to store queues to port connections. */
+       void *dev_private;
+       /**< PMD-specific private data */
+       uint32_t event_dev_cap;
+       /**< Event device capabilities(RTE_EVENT_DEV_CAP_)*/
+       struct rte_event_dev_config dev_conf;
+       /**< Configuration applied to device. */
+
+       RTE_STD_C11
+       uint8_t dev_started : 1;
+       /**< Device state: STARTED(1)/STOPPED(0) */
+
+       char name[RTE_EVENTDEV_NAME_MAX_LEN];
+       /**< Unique identifier name */
+} __rte_cache_aligned;
 
 /** @internal The data structure associated with each event device. */
 struct rte_eventdev {
@@ -1007,8 +1052,23 @@ struct rte_eventdev {
        event_dequeue_burst_t dequeue_burst;
        /**< Pointer to PMD dequeue burst function. */
 
+       struct rte_eventdev_data *data;
+       /**< Pointer to device data */
+       const struct rte_eventdev_ops *dev_ops;
+       /**< Functions exported by PMD */
+       struct rte_pci_device *pci_dev;
+       /**< PCI info. supplied by probing */
+       const struct rte_eventdev_driver *driver;
+       /**< Driver for this device */
+
+       RTE_STD_C11
+       uint8_t attached : 1;
+       /**< Flag indicating the device is attached */
 } __rte_cache_aligned;
 
+extern struct rte_eventdev *rte_eventdevs;
+/** @internal The pool of rte_eventdev structures. */
+
 
 /**
  * Schedule one or more events in the event dev.
@@ -1019,8 +1079,13 @@ struct rte_eventdev {
  * @param dev_id
  *   The identifier of the device.
  */
-void
-rte_event_schedule(uint8_t dev_id);
+static inline void
+rte_event_schedule(uint8_t dev_id)
+{
+       struct rte_eventdev *dev = &rte_eventdevs[dev_id];
+       if (*dev->schedule)
+               (*dev->schedule)(dev);
+}
 
 /**
  * Enqueue a burst of events objects or an event object supplied in *rte_event*
@@ -1055,9 +1120,23 @@ rte_event_schedule(uint8_t dev_id);
  *
  * @see rte_event_port_enqueue_depth()
  */
-uint16_t
+static inline uint16_t
 rte_event_enqueue_burst(uint8_t dev_id, uint8_t port_id,
-                       const struct rte_event ev[], uint16_t nb_events);
+                       const struct rte_event ev[], uint16_t nb_events)
+{
+       struct rte_eventdev *dev = &rte_eventdevs[dev_id];
+
+       /*
+        * Allow zero cost non burst mode routine invocation if application
+        * requests nb_events as const one
+        */
+       if (nb_events == 1)
+               return (*dev->enqueue)(
+                       dev->data->ports[port_id], ev);
+       else
+               return (*dev->enqueue_burst)(
+                       dev->data->ports[port_id], ev, nb_events);
+}
 
 /**
  * Converts nanoseconds to *timeout_ticks* value for rte_event_dequeue_burst()
@@ -1149,9 +1228,24 @@ rte_event_dequeue_timeout_ticks(uint8_t dev_id, uint64_t ns,
  *
  * @see rte_event_port_dequeue_depth()
  */
-uint16_t
+static inline uint16_t
 rte_event_dequeue_burst(uint8_t dev_id, uint8_t port_id, struct rte_event ev[],
-                       uint16_t nb_events, uint64_t timeout_ticks);
+                       uint16_t nb_events, uint64_t timeout_ticks)
+{
+       struct rte_eventdev *dev = &rte_eventdevs[dev_id];
+
+       /*
+        * Allow zero cost non burst mode routine invocation if application
+        * requests nb_events as const one
+        */
+       if (nb_events == 1)
+               return (*dev->dequeue)(
+                       dev->data->ports[port_id], ev, timeout_ticks);
+       else
+               return (*dev->dequeue_burst)(
+                       dev->data->ports[port_id], ev, nb_events,
+                               timeout_ticks);
+}
 
 /**
  * Link multiple source event queues supplied in *queues* to the destination
index 40552aa..e60eca9 100644 (file)
 extern "C" {
 #endif
 
+#include <string.h>
+
+#include <rte_dev.h>
+#include <rte_pci.h>
+#include <rte_malloc.h>
+#include <rte_log.h>
+#include <rte_common.h>
+
 #include "rte_eventdev.h"
 
+#ifdef RTE_LIBRTE_EVENTDEV_DEBUG
+#define RTE_PMD_DEBUG_TRACE(...) \
+       rte_pmd_debug_trace(__func__, __VA_ARGS__)
+#else
+#define RTE_PMD_DEBUG_TRACE(...)
+#endif
+
+/* Logging Macros */
+#define RTE_EDEV_LOG_ERR(fmt, args...) \
+       RTE_LOG(ERR, EVENTDEV, "%s() line %u: " fmt "\n",  \
+                       __func__, __LINE__, ## args)
+
+#ifdef RTE_LIBRTE_EVENTDEV_DEBUG
+#define RTE_EDEV_LOG_DEBUG(fmt, args...) \
+       RTE_LOG(DEBUG, EVENTDEV, "%s() line %u: " fmt "\n",  \
+                       __func__, __LINE__, ## args)
+#else
+#define RTE_EDEV_LOG_DEBUG(fmt, args...) (void)0
+#endif
+
+/* Macros to check for valid device */
+#define RTE_EVENTDEV_VALID_DEVID_OR_ERR_RET(dev_id, retval) do { \
+       if (!rte_event_pmd_is_valid_dev((dev_id))) { \
+               RTE_EDEV_LOG_ERR("Invalid dev_id=%d\n", dev_id); \
+               return retval; \
+       } \
+} while (0)
+
+#define RTE_EVENTDEV_VALID_DEVID_OR_RET(dev_id) do { \
+       if (!rte_event_pmd_is_valid_dev((dev_id))) { \
+               RTE_EDEV_LOG_ERR("Invalid dev_id=%d\n", dev_id); \
+               return; \
+       } \
+} while (0)
+
+#define RTE_EVENTDEV_DETACHED  (0)
+#define RTE_EVENTDEV_ATTACHED  (1)
+
+/** Global structure used for maintaining state of allocated event devices */
+struct rte_eventdev_global {
+       uint8_t nb_devs;        /**< Number of devices found */
+       uint8_t max_devs;       /**< Max number of devices */
+};
+
+extern struct rte_eventdev_global *rte_eventdev_globals;
+/** Pointer to global event devices data structure. */
+extern struct rte_eventdev *rte_eventdevs;
+/** The pool of rte_eventdev structures. */
+
+/**
+ * Get the rte_eventdev structure device pointer for the named device.
+ *
+ * @param name
+ *   device name to select the device structure.
+ *
+ * @return
+ *   - The rte_eventdev structure pointer for the given device ID.
+ */
+static inline struct rte_eventdev *
+rte_event_pmd_get_named_dev(const char *name)
+{
+       struct rte_eventdev *dev;
+       unsigned int i;
+
+       if (name == NULL)
+               return NULL;
+
+       for (i = 0, dev = &rte_eventdevs[i];
+                       i < rte_eventdev_globals->max_devs; i++) {
+               if ((dev->attached == RTE_EVENTDEV_ATTACHED) &&
+                               (strcmp(dev->data->name, name) == 0))
+                       return dev;
+       }
+
+       return NULL;
+}
+
+/**
+ * Validate if the event device index is valid attached event device.
+ *
+ * @param dev_id
+ *   Event device index.
+ *
+ * @return
+ *   - If the device index is valid (1) or not (0).
+ */
+static inline unsigned
+rte_event_pmd_is_valid_dev(uint8_t dev_id)
+{
+       struct rte_eventdev *dev;
+
+       if (dev_id >= rte_eventdev_globals->nb_devs)
+               return 0;
+
+       dev = &rte_eventdevs[dev_id];
+       if (dev->attached != RTE_EVENTDEV_ATTACHED)
+               return 0;
+       else
+               return 1;
+}
+
 /**
  * Definitions of all functions exported by a driver through the
  * the generic structure of type *event_dev_ops* supplied in the
diff --git a/lib/librte_eventdev/rte_eventdev_version.map b/lib/librte_eventdev/rte_eventdev_version.map
new file mode 100644 (file)
index 0000000..45072db
--- /dev/null
@@ -0,0 +1,33 @@
+DPDK_17.05 {
+       global:
+
+       rte_eventdevs;
+
+       rte_event_dev_count;
+       rte_event_dev_get_dev_id;
+       rte_event_dev_socket_id;
+       rte_event_dev_info_get;
+       rte_event_dev_configure;
+       rte_event_dev_start;
+       rte_event_dev_stop;
+       rte_event_dev_close;
+       rte_event_dev_dump;
+
+       rte_event_port_default_conf_get;
+       rte_event_port_setup;
+       rte_event_port_dequeue_depth;
+       rte_event_port_enqueue_depth;
+       rte_event_port_count;
+       rte_event_port_link;
+       rte_event_port_unlink;
+       rte_event_port_links_get;
+
+       rte_event_queue_default_conf_get;
+       rte_event_queue_setup;
+       rte_event_queue_count;
+       rte_event_queue_priority;
+
+       rte_event_dequeue_timeout_ticks;
+
+       local: *;
+};
index 9c3a753..97f6abb 100644 (file)
@@ -88,6 +88,7 @@ _LDLIBS-$(CONFIG_RTE_LIBRTE_MBUF)           += -lrte_mbuf
 _LDLIBS-$(CONFIG_RTE_LIBRTE_NET)            += -lrte_net
 _LDLIBS-$(CONFIG_RTE_LIBRTE_ETHER)          += -lrte_ethdev
 _LDLIBS-$(CONFIG_RTE_LIBRTE_CRYPTODEV)      += -lrte_cryptodev
+_LDLIBS-$(CONFIG_RTE_LIBRTE_EVENTDEV)       += -lrte_eventdev
 _LDLIBS-$(CONFIG_RTE_LIBRTE_MEMPOOL)        += -lrte_mempool
 _LDLIBS-$(CONFIG_RTE_LIBRTE_RING)           += -lrte_ring
 _LDLIBS-$(CONFIG_RTE_LIBRTE_EAL)            += -lrte_eal