event/dpaa2: add self test
authorHemant Agrawal <hemant.agrawal@nxp.com>
Mon, 30 Sep 2019 08:32:14 +0000 (14:02 +0530)
committerJerin Jacob <jerinj@marvell.com>
Sat, 5 Oct 2019 13:39:31 +0000 (15:39 +0200)
This patch add support for testing dpaa2 eventdev self test
for basic sanity for parallel and atomic queues.

Signed-off-by: Hemant Agrawal <hemant.agrawal@nxp.com>
drivers/event/dpaa2/Makefile
drivers/event/dpaa2/dpaa2_eventdev.c
drivers/event/dpaa2/dpaa2_eventdev.h
drivers/event/dpaa2/dpaa2_eventdev_logs.h
drivers/event/dpaa2/dpaa2_eventdev_selftest.c [new file with mode: 0644]
drivers/event/dpaa2/meson.build

index e0bb527..c6ab326 100644 (file)
@@ -18,9 +18,9 @@ CFLAGS += -I$(RTE_SDK)/drivers/bus/fslmc/portal
 CFLAGS += -I$(RTE_SDK)/drivers/mempool/dpaa2
 CFLAGS += -I$(RTE_SDK)/drivers/event/dpaa2
 LDLIBS += -lrte_eal -lrte_eventdev
-LDLIBS += -lrte_bus_fslmc -lrte_mempool_dpaa2 -lrte_pmd_dpaa2
-LDLIBS += -lrte_bus_vdev
 LDLIBS += -lrte_common_dpaax
+LDLIBS += -lrte_bus_fslmc -lrte_mempool_dpaa2 -lrte_pmd_dpaa2
+LDLIBS += -lrte_bus_vdev -lrte_mempool -lrte_mbuf
 CFLAGS += -I$(RTE_SDK)/drivers/net/dpaa2
 CFLAGS += -I$(RTE_SDK)/drivers/net/dpaa2/mc
 
@@ -40,5 +40,6 @@ CFLAGS += -DALLOW_EXPERIMENTAL_API
 #
 SRCS-$(CONFIG_RTE_LIBRTE_PMD_DPAA2_EVENTDEV) += dpaa2_hw_dpcon.c
 SRCS-$(CONFIG_RTE_LIBRTE_PMD_DPAA2_EVENTDEV) += dpaa2_eventdev.c
+SRCS-$(CONFIG_RTE_LIBRTE_PMD_DPAA2_EVENTDEV) += dpaa2_eventdev_selftest.c
 
 include $(RTE_SDK)/mk/rte.lib.mk
index 834d3cb..5249d2f 100644 (file)
@@ -964,6 +964,7 @@ static struct rte_eventdev_ops dpaa2_eventdev_ops = {
        .port_unlink      = dpaa2_eventdev_port_unlink,
        .timeout_ticks    = dpaa2_eventdev_timeout_ticks,
        .dump             = dpaa2_eventdev_dump,
+       .dev_selftest     = test_eventdev_dpaa2,
        .eth_rx_adapter_caps_get = dpaa2_eventdev_eth_caps_get,
        .eth_rx_adapter_queue_add = dpaa2_eventdev_eth_queue_add,
        .eth_rx_adapter_queue_del = dpaa2_eventdev_eth_queue_del,
index bdac1aa..abc038e 100644 (file)
@@ -98,4 +98,6 @@ struct dpaa2_eventdev {
 struct dpaa2_dpcon_dev *rte_dpaa2_alloc_dpcon_dev(void);
 void rte_dpaa2_free_dpcon_dev(struct dpaa2_dpcon_dev *dpcon);
 
+int test_eventdev_dpaa2(void);
+
 #endif /* __DPAA2_EVENTDEV_H__ */
index 86f2e53..5da85c6 100644 (file)
@@ -1,5 +1,5 @@
 /* SPDX-License-Identifier: BSD-3-Clause
- * Copyright 2018 NXP
+ * Copyright 2018-2019 NXP
  */
 
 #ifndef _DPAA2_EVENTDEV_LOGS_H_
@@ -35,4 +35,10 @@ extern int dpaa2_logtype_event;
 #define DPAA2_EVENTDEV_DP_WARN(fmt, args...) \
        DPAA2_EVENTDEV_DP_LOG(WARNING, fmt, ## args)
 
+#define dpaa2_evdev_info(fmt, ...) DPAA2_EVENTDEV_LOG(INFO, fmt, ##__VA_ARGS__)
+#define dpaa2_evdev_dbg(fmt, ...) DPAA2_EVENTDEV_LOG(DEBUG, fmt, ##__VA_ARGS__)
+#define dpaa2_evdev_err(fmt, ...) DPAA2_EVENTDEV_LOG(ERR, fmt, ##__VA_ARGS__)
+#define dpaa2_evdev__func_trace dpaa2_evdev_dbg
+#define dpaa2_evdev_selftest dpaa2_evdev_info
+
 #endif /* _DPAA2_EVENTDEV_LOGS_H_ */
diff --git a/drivers/event/dpaa2/dpaa2_eventdev_selftest.c b/drivers/event/dpaa2/dpaa2_eventdev_selftest.c
new file mode 100644 (file)
index 0000000..ba4f4bd
--- /dev/null
@@ -0,0 +1,833 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright 2018-2019 NXP
+ */
+
+#include <rte_atomic.h>
+#include <rte_common.h>
+#include <rte_cycles.h>
+#include <rte_debug.h>
+#include <rte_eal.h>
+#include <rte_ethdev.h>
+#include <rte_eventdev.h>
+#include <rte_hexdump.h>
+#include <rte_mbuf.h>
+#include <rte_malloc.h>
+#include <rte_memcpy.h>
+#include <rte_launch.h>
+#include <rte_lcore.h>
+#include <rte_per_lcore.h>
+#include <rte_random.h>
+#include <rte_bus_vdev.h>
+#include <rte_test.h>
+
+#include "dpaa2_eventdev.h"
+#include "dpaa2_eventdev_logs.h"
+
+#define MAX_PORTS 4
+#define NUM_PACKETS (1 << 18)
+#define MAX_EVENTS  8
+#define DPAA2_TEST_RUN(setup, teardown, test) \
+       dpaa2_test_run(setup, teardown, test, #test)
+
+static int total;
+static int passed;
+static int failed;
+static int unsupported;
+
+static int evdev;
+static struct rte_mempool *eventdev_test_mempool;
+
+struct event_attr {
+       uint32_t flow_id;
+       uint8_t event_type;
+       uint8_t sub_event_type;
+       uint8_t sched_type;
+       uint8_t queue;
+       uint8_t port;
+       uint8_t seq;
+};
+
+static uint32_t seqn_list_index;
+static int seqn_list[NUM_PACKETS];
+
+static void
+seqn_list_init(void)
+{
+       RTE_BUILD_BUG_ON(NUM_PACKETS < MAX_EVENTS);
+       memset(seqn_list, 0, sizeof(seqn_list));
+       seqn_list_index = 0;
+}
+
+struct test_core_param {
+       rte_atomic32_t *total_events;
+       uint64_t dequeue_tmo_ticks;
+       uint8_t port;
+       uint8_t sched_type;
+};
+
+static int
+testsuite_setup(void)
+{
+       const char *eventdev_name = "event_dpaa2";
+
+       evdev = rte_event_dev_get_dev_id(eventdev_name);
+       if (evdev < 0) {
+               dpaa2_evdev_dbg("%d: Eventdev %s not found - creating.",
+                               __LINE__, eventdev_name);
+               if (rte_vdev_init(eventdev_name, NULL) < 0) {
+                       dpaa2_evdev_err("Error creating eventdev %s",
+                                       eventdev_name);
+                       return -1;
+               }
+               evdev = rte_event_dev_get_dev_id(eventdev_name);
+               if (evdev < 0) {
+                       dpaa2_evdev_err("Error finding newly created eventdev");
+                       return -1;
+               }
+       }
+
+       return 0;
+}
+
+static void
+testsuite_teardown(void)
+{
+       rte_event_dev_close(evdev);
+}
+
+static void
+devconf_set_default_sane_values(struct rte_event_dev_config *dev_conf,
+                       struct rte_event_dev_info *info)
+{
+       memset(dev_conf, 0, sizeof(struct rte_event_dev_config));
+       dev_conf->dequeue_timeout_ns = info->min_dequeue_timeout_ns;
+       dev_conf->nb_event_ports = info->max_event_ports;
+       dev_conf->nb_event_queues = info->max_event_queues;
+       dev_conf->nb_event_queue_flows = info->max_event_queue_flows;
+       dev_conf->nb_event_port_dequeue_depth =
+                       info->max_event_port_dequeue_depth;
+       dev_conf->nb_event_port_enqueue_depth =
+                       info->max_event_port_enqueue_depth;
+       dev_conf->nb_event_port_enqueue_depth =
+                       info->max_event_port_enqueue_depth;
+       dev_conf->nb_events_limit =
+                       info->max_num_events;
+}
+
+enum {
+       TEST_EVENTDEV_SETUP_DEFAULT,
+       TEST_EVENTDEV_SETUP_PRIORITY,
+       TEST_EVENTDEV_SETUP_DEQUEUE_TIMEOUT,
+};
+
+static int
+_eventdev_setup(int mode)
+{
+       int i, ret;
+       struct rte_event_dev_config dev_conf;
+       struct rte_event_dev_info info;
+       const char *pool_name = "evdev_dpaa2_test_pool";
+
+       /* Create and destrory pool for each test case to make it standalone */
+       eventdev_test_mempool = rte_pktmbuf_pool_create(pool_name,
+                                       MAX_EVENTS,
+                                       0 /*MBUF_CACHE_SIZE*/,
+                                       0,
+                                       512, /* Use very small mbufs */
+                                       rte_socket_id());
+       if (!eventdev_test_mempool) {
+               dpaa2_evdev_err("ERROR creating mempool");
+               return -1;
+       }
+
+       ret = rte_event_dev_info_get(evdev, &info);
+       RTE_TEST_ASSERT_SUCCESS(ret, "Failed to get event dev info");
+       RTE_TEST_ASSERT(info.max_num_events >= (int32_t)MAX_EVENTS,
+                       "ERROR max_num_events=%d < max_events=%d",
+                               info.max_num_events, MAX_EVENTS);
+
+       devconf_set_default_sane_values(&dev_conf, &info);
+       if (mode == TEST_EVENTDEV_SETUP_DEQUEUE_TIMEOUT)
+               dev_conf.event_dev_cfg |= RTE_EVENT_DEV_CFG_PER_DEQUEUE_TIMEOUT;
+
+       ret = rte_event_dev_configure(evdev, &dev_conf);
+       RTE_TEST_ASSERT_SUCCESS(ret, "Failed to configure eventdev");
+
+       uint32_t queue_count;
+       RTE_TEST_ASSERT_SUCCESS(rte_event_dev_attr_get(evdev,
+                           RTE_EVENT_DEV_ATTR_QUEUE_COUNT,
+                           &queue_count), "Queue count get failed");
+
+       if (mode == TEST_EVENTDEV_SETUP_PRIORITY) {
+               if (queue_count > 8) {
+                       dpaa2_evdev_err(
+                               "test expects the unique priority per queue");
+                       return -ENOTSUP;
+               }
+
+               /* Configure event queues(0 to n) with
+                * RTE_EVENT_DEV_PRIORITY_HIGHEST to
+                * RTE_EVENT_DEV_PRIORITY_LOWEST
+                */
+               uint8_t step = (RTE_EVENT_DEV_PRIORITY_LOWEST + 1) /
+                               queue_count;
+               for (i = 0; i < (int)queue_count; i++) {
+                       struct rte_event_queue_conf queue_conf;
+
+                       ret = rte_event_queue_default_conf_get(evdev, i,
+                                               &queue_conf);
+                       RTE_TEST_ASSERT_SUCCESS(ret, "Failed to get def_conf%d",
+                                       i);
+                       queue_conf.priority = i * step;
+                       ret = rte_event_queue_setup(evdev, i, &queue_conf);
+                       RTE_TEST_ASSERT_SUCCESS(ret, "Failed to setup queue=%d",
+                                       i);
+               }
+
+       } else {
+               /* Configure event queues with default priority */
+               for (i = 0; i < (int)queue_count; i++) {
+                       ret = rte_event_queue_setup(evdev, i, NULL);
+                       RTE_TEST_ASSERT_SUCCESS(ret, "Failed to setup queue=%d",
+                                       i);
+               }
+       }
+       /* Configure event ports */
+       uint32_t port_count;
+       RTE_TEST_ASSERT_SUCCESS(rte_event_dev_attr_get(evdev,
+                               RTE_EVENT_DEV_ATTR_PORT_COUNT,
+                               &port_count), "Port count get failed");
+       for (i = 0; i < (int)port_count; i++) {
+               ret = rte_event_port_setup(evdev, i, NULL);
+               RTE_TEST_ASSERT_SUCCESS(ret, "Failed to setup port=%d", i);
+               ret = rte_event_port_link(evdev, i, NULL, NULL, 0);
+               RTE_TEST_ASSERT(ret >= 0, "Failed to link all queues port=%d",
+                               i);
+       }
+
+       ret = rte_event_dev_start(evdev);
+       RTE_TEST_ASSERT_SUCCESS(ret, "Failed to start device");
+
+       return 0;
+}
+
+static int
+eventdev_setup(void)
+{
+       return _eventdev_setup(TEST_EVENTDEV_SETUP_DEFAULT);
+}
+
+static void
+eventdev_teardown(void)
+{
+       rte_event_dev_stop(evdev);
+       rte_mempool_free(eventdev_test_mempool);
+}
+
+static void
+update_event_and_validation_attr(struct rte_mbuf *m, struct rte_event *ev,
+                       uint32_t flow_id, uint8_t event_type,
+                       uint8_t sub_event_type, uint8_t sched_type,
+                       uint8_t queue, uint8_t port, uint8_t seq)
+{
+       struct event_attr *attr;
+
+       /* Store the event attributes in mbuf for future reference */
+       attr = rte_pktmbuf_mtod(m, struct event_attr *);
+       attr->flow_id = flow_id;
+       attr->event_type = event_type;
+       attr->sub_event_type = sub_event_type;
+       attr->sched_type = sched_type;
+       attr->queue = queue;
+       attr->port = port;
+       attr->seq = seq;
+
+       ev->flow_id = flow_id;
+       ev->sub_event_type = sub_event_type;
+       ev->event_type = event_type;
+       /* Inject the new event */
+       ev->op = RTE_EVENT_OP_NEW;
+       ev->sched_type = sched_type;
+       ev->queue_id = queue;
+       ev->mbuf = m;
+}
+
+static int
+inject_events(uint32_t flow_id, uint8_t event_type, uint8_t sub_event_type,
+               uint8_t sched_type, uint8_t queue, uint8_t port,
+               unsigned int events)
+{
+       struct rte_mbuf *m;
+       unsigned int i;
+
+       for (i = 0; i < events; i++) {
+               struct rte_event ev = {.event = 0, .u64 = 0};
+
+               m = rte_pktmbuf_alloc(eventdev_test_mempool);
+               RTE_TEST_ASSERT_NOT_NULL(m, "mempool alloc failed");
+
+               update_event_and_validation_attr(m, &ev, flow_id, event_type,
+                       sub_event_type, sched_type, queue, port, i);
+               rte_event_enqueue_burst(evdev, port, &ev, 1);
+       }
+       return 0;
+}
+
+static int
+check_excess_events(uint8_t port)
+{
+       int i;
+       uint16_t valid_event;
+       struct rte_event ev;
+
+       /* Check for excess events, try for a few times and exit */
+       for (i = 0; i < 32; i++) {
+               valid_event = rte_event_dequeue_burst(evdev, port, &ev, 1, 0);
+
+               RTE_TEST_ASSERT_SUCCESS(valid_event,
+                               "Unexpected valid event=%d", ev.mbuf->seqn);
+       }
+       return 0;
+}
+
+static int
+generate_random_events(const unsigned int total_events)
+{
+       struct rte_event_dev_info info;
+       unsigned int i;
+       int ret;
+
+       uint32_t queue_count;
+       RTE_TEST_ASSERT_SUCCESS(rte_event_dev_attr_get(evdev,
+                           RTE_EVENT_DEV_ATTR_QUEUE_COUNT,
+                           &queue_count), "Queue count get failed");
+
+       ret = rte_event_dev_info_get(evdev, &info);
+       RTE_TEST_ASSERT_SUCCESS(ret, "Failed to get event dev info");
+       for (i = 0; i < total_events; i++) {
+               ret = inject_events(
+                       rte_rand() % info.max_event_queue_flows /*flow_id */,
+                       RTE_EVENT_TYPE_CPU /* event_type */,
+                       rte_rand() % 256 /* sub_event_type */,
+                       rte_rand() % (RTE_SCHED_TYPE_PARALLEL + 1),
+                       rte_rand() % queue_count /* queue */,
+                       0 /* port */,
+                       1 /* events */);
+               if (ret)
+                       return -1;
+       }
+       return ret;
+}
+
+
+static int
+validate_event(struct rte_event *ev)
+{
+       struct event_attr *attr;
+
+       attr = rte_pktmbuf_mtod(ev->mbuf, struct event_attr *);
+       RTE_TEST_ASSERT_EQUAL(attr->flow_id, ev->flow_id,
+                       "flow_id mismatch enq=%d deq =%d",
+                       attr->flow_id, ev->flow_id);
+       RTE_TEST_ASSERT_EQUAL(attr->event_type, ev->event_type,
+                       "event_type mismatch enq=%d deq =%d",
+                       attr->event_type, ev->event_type);
+       RTE_TEST_ASSERT_EQUAL(attr->sub_event_type, ev->sub_event_type,
+                       "sub_event_type mismatch enq=%d deq =%d",
+                       attr->sub_event_type, ev->sub_event_type);
+       RTE_TEST_ASSERT_EQUAL(attr->sched_type, ev->sched_type,
+                       "sched_type mismatch enq=%d deq =%d",
+                       attr->sched_type, ev->sched_type);
+       RTE_TEST_ASSERT_EQUAL(attr->queue, ev->queue_id,
+                       "queue mismatch enq=%d deq =%d",
+                       attr->queue, ev->queue_id);
+       return 0;
+}
+
+typedef int (*validate_event_cb)(uint32_t index, uint8_t port,
+                                struct rte_event *ev);
+
+static int
+consume_events(uint8_t port, const uint32_t total_events, validate_event_cb fn)
+{
+       int ret;
+       uint16_t valid_event;
+       uint32_t events = 0, forward_progress_cnt = 0, index = 0;
+       struct rte_event ev;
+
+       while (1) {
+               if (++forward_progress_cnt > UINT16_MAX) {
+                       dpaa2_evdev_err("Detected deadlock");
+                       return -1;
+               }
+
+               valid_event = rte_event_dequeue_burst(evdev, port, &ev, 1, 0);
+               if (!valid_event)
+                       continue;
+
+               forward_progress_cnt = 0;
+               ret = validate_event(&ev);
+               if (ret)
+                       return -1;
+
+               if (fn != NULL) {
+                       ret = fn(index, port, &ev);
+                       RTE_TEST_ASSERT_SUCCESS(ret,
+                               "Failed to validate test specific event");
+               }
+
+               ++index;
+
+               rte_pktmbuf_free(ev.mbuf);
+               if (++events >= total_events)
+                       break;
+       }
+
+       return check_excess_events(port);
+}
+
+static int
+validate_simple_enqdeq(uint32_t index, uint8_t port, struct rte_event *ev)
+{
+       struct event_attr *attr;
+
+       attr = rte_pktmbuf_mtod(ev->mbuf, struct event_attr *);
+
+       RTE_SET_USED(port);
+       RTE_TEST_ASSERT_EQUAL(index, attr->seq,
+               "index=%d != seqn=%d", index, attr->seq);
+       return 0;
+}
+
+static int
+test_simple_enqdeq(uint8_t sched_type)
+{
+       int ret;
+
+       ret = inject_events(0 /*flow_id */,
+                               RTE_EVENT_TYPE_CPU /* event_type */,
+                               0 /* sub_event_type */,
+                               sched_type,
+                               0 /* queue */,
+                               0 /* port */,
+                               MAX_EVENTS);
+       if (ret)
+               return -1;
+
+       return consume_events(0 /* port */, MAX_EVENTS, validate_simple_enqdeq);
+}
+
+static int
+test_simple_enqdeq_atomic(void)
+{
+       return test_simple_enqdeq(RTE_SCHED_TYPE_ATOMIC);
+}
+
+static int
+test_simple_enqdeq_parallel(void)
+{
+       return test_simple_enqdeq(RTE_SCHED_TYPE_PARALLEL);
+}
+
+/*
+ * Generate a prescribed number of events and spread them across available
+ * queues. On dequeue, using single event port(port 0) verify the enqueued
+ * event attributes
+ */
+static int
+test_multi_queue_enq_single_port_deq(void)
+{
+       int ret;
+
+       ret = generate_random_events(MAX_EVENTS);
+       if (ret)
+               return -1;
+
+       return consume_events(0 /* port */, MAX_EVENTS, NULL);
+}
+
+static int
+worker_multi_port_fn(void *arg)
+{
+       struct test_core_param *param = arg;
+       struct rte_event ev;
+       uint16_t valid_event;
+       uint8_t port = param->port;
+       rte_atomic32_t *total_events = param->total_events;
+       int ret;
+
+       while (rte_atomic32_read(total_events) > 0) {
+               valid_event = rte_event_dequeue_burst(evdev, port, &ev, 1, 0);
+               if (!valid_event)
+                       continue;
+
+               ret = validate_event(&ev);
+               RTE_TEST_ASSERT_SUCCESS(ret, "Failed to validate event");
+               rte_pktmbuf_free(ev.mbuf);
+               rte_atomic32_sub(total_events, 1);
+       }
+       return 0;
+}
+
+static int
+wait_workers_to_join(int lcore, const rte_atomic32_t *count)
+{
+       uint64_t cycles, print_cycles;
+
+       RTE_SET_USED(count);
+
+       print_cycles = cycles = rte_get_timer_cycles();
+       while (rte_eal_get_lcore_state(lcore) != FINISHED) {
+               uint64_t new_cycles = rte_get_timer_cycles();
+
+               if (new_cycles - print_cycles > rte_get_timer_hz()) {
+                       dpaa2_evdev_dbg("\r%s: events %d", __func__,
+                               rte_atomic32_read(count));
+                       print_cycles = new_cycles;
+               }
+               if (new_cycles - cycles > rte_get_timer_hz() * 10) {
+                       dpaa2_evdev_info(
+                               "%s: No schedules for seconds, deadlock (%d)",
+                               __func__,
+                               rte_atomic32_read(count));
+                       rte_event_dev_dump(evdev, stdout);
+                       cycles = new_cycles;
+                       return -1;
+               }
+       }
+       rte_eal_mp_wait_lcore();
+       return 0;
+}
+
+
+static int
+launch_workers_and_wait(int (*master_worker)(void *),
+                       int (*slave_workers)(void *), uint32_t total_events,
+                       uint8_t nb_workers, uint8_t sched_type)
+{
+       uint8_t port = 0;
+       int w_lcore;
+       int ret;
+       struct test_core_param *param;
+       rte_atomic32_t atomic_total_events;
+       uint64_t dequeue_tmo_ticks;
+
+       if (!nb_workers)
+               return 0;
+
+       rte_atomic32_set(&atomic_total_events, total_events);
+       seqn_list_init();
+
+       param = malloc(sizeof(struct test_core_param) * nb_workers);
+       if (!param)
+               return -1;
+
+       ret = rte_event_dequeue_timeout_ticks(evdev,
+               rte_rand() % 10000000/* 10ms */, &dequeue_tmo_ticks);
+       if (ret) {
+               free(param);
+               return -1;
+       }
+
+       param[0].total_events = &atomic_total_events;
+       param[0].sched_type = sched_type;
+       param[0].port = 0;
+       param[0].dequeue_tmo_ticks = dequeue_tmo_ticks;
+       rte_smp_wmb();
+
+       w_lcore = rte_get_next_lcore(
+                       /* start core */ -1,
+                       /* skip master */ 1,
+                       /* wrap */ 0);
+       rte_eal_remote_launch(master_worker, &param[0], w_lcore);
+
+       for (port = 1; port < nb_workers; port++) {
+               param[port].total_events = &atomic_total_events;
+               param[port].sched_type = sched_type;
+               param[port].port = port;
+               param[port].dequeue_tmo_ticks = dequeue_tmo_ticks;
+               rte_smp_wmb();
+               w_lcore = rte_get_next_lcore(w_lcore, 1, 0);
+               rte_eal_remote_launch(slave_workers, &param[port], w_lcore);
+       }
+
+       ret = wait_workers_to_join(w_lcore, &atomic_total_events);
+       free(param);
+       return ret;
+}
+
+/*
+ * Generate a prescribed number of events and spread them across available
+ * queues. Dequeue the events through multiple ports and verify the enqueued
+ * event attributes
+ */
+static int
+test_multi_queue_enq_multi_port_deq(void)
+{
+       const unsigned int total_events = MAX_EVENTS;
+       uint32_t nr_ports;
+       int ret;
+
+       ret = generate_random_events(total_events);
+       if (ret)
+               return -1;
+
+       RTE_TEST_ASSERT_SUCCESS(rte_event_dev_attr_get(evdev,
+                               RTE_EVENT_DEV_ATTR_PORT_COUNT,
+                               &nr_ports), "Port count get failed");
+       nr_ports = RTE_MIN(nr_ports, rte_lcore_count() - 1);
+
+       if (!nr_ports) {
+               dpaa2_evdev_err("%s: Not enough ports=%d or workers=%d",
+                               __func__, nr_ports, rte_lcore_count() - 1);
+               return 0;
+       }
+
+       return launch_workers_and_wait(worker_multi_port_fn,
+                                       worker_multi_port_fn, total_events,
+                                       nr_ports, 0xff /* invalid */);
+}
+
+static
+void flush(uint8_t dev_id, struct rte_event event, void *arg)
+{
+       unsigned int *count = arg;
+
+       RTE_SET_USED(dev_id);
+       if (event.event_type == RTE_EVENT_TYPE_CPU)
+               *count = *count + 1;
+
+}
+
+static int
+test_dev_stop_flush(void)
+{
+       unsigned int total_events = MAX_EVENTS, count = 0;
+       int ret;
+
+       ret = generate_random_events(total_events);
+       if (ret)
+               return -1;
+
+       ret = rte_event_dev_stop_flush_callback_register(evdev, flush, &count);
+       if (ret)
+               return -2;
+       rte_event_dev_stop(evdev);
+       ret = rte_event_dev_stop_flush_callback_register(evdev, NULL, NULL);
+       if (ret)
+               return -3;
+       RTE_TEST_ASSERT_EQUAL(total_events, count,
+                               "count mismatch total_events=%d count=%d",
+                               total_events, count);
+       return 0;
+}
+
+static int
+validate_queue_to_port_single_link(uint32_t index, uint8_t port,
+                       struct rte_event *ev)
+{
+       RTE_SET_USED(index);
+       RTE_TEST_ASSERT_EQUAL(port, ev->queue_id,
+                               "queue mismatch enq=%d deq =%d",
+                               port, ev->queue_id);
+       return 0;
+}
+
+/*
+ * Link queue x to port x and check correctness of link by checking
+ * queue_id == x on dequeue on the specific port x
+ */
+static int
+test_queue_to_port_single_link(void)
+{
+       int i, nr_links, ret;
+
+       uint32_t port_count;
+
+       RTE_TEST_ASSERT_SUCCESS(rte_event_dev_attr_get(evdev,
+                               RTE_EVENT_DEV_ATTR_PORT_COUNT,
+                               &port_count), "Port count get failed");
+
+       /* Unlink all connections that created in eventdev_setup */
+       for (i = 0; i < (int)port_count; i++) {
+               ret = rte_event_port_unlink(evdev, i, NULL, 0);
+               RTE_TEST_ASSERT(ret >= 0,
+                               "Failed to unlink all queues port=%d", i);
+       }
+
+       uint32_t queue_count;
+
+       RTE_TEST_ASSERT_SUCCESS(rte_event_dev_attr_get(evdev,
+                           RTE_EVENT_DEV_ATTR_QUEUE_COUNT,
+                           &queue_count), "Queue count get failed");
+
+       nr_links = RTE_MIN(port_count, queue_count);
+       const unsigned int total_events = MAX_EVENTS / nr_links;
+
+       /* Link queue x to port x and inject events to queue x through port x */
+       for (i = 0; i < nr_links; i++) {
+               uint8_t queue = (uint8_t)i;
+
+               ret = rte_event_port_link(evdev, i, &queue, NULL, 1);
+               RTE_TEST_ASSERT(ret == 1, "Failed to link queue to port %d", i);
+
+               ret = inject_events(
+                       0x100 /*flow_id */,
+                       RTE_EVENT_TYPE_CPU /* event_type */,
+                       rte_rand() % 256 /* sub_event_type */,
+                       rte_rand() % (RTE_SCHED_TYPE_PARALLEL + 1),
+                       queue /* queue */,
+                       i /* port */,
+                       total_events /* events */);
+               if (ret)
+                       return -1;
+       }
+
+       /* Verify the events generated from correct queue */
+       for (i = 0; i < nr_links; i++) {
+               ret = consume_events(i /* port */, total_events,
+                               validate_queue_to_port_single_link);
+               if (ret)
+                       return -1;
+       }
+
+       return 0;
+}
+
+static int
+validate_queue_to_port_multi_link(uint32_t index, uint8_t port,
+                       struct rte_event *ev)
+{
+       RTE_SET_USED(index);
+       RTE_TEST_ASSERT_EQUAL(port, (ev->queue_id & 0x1),
+                               "queue mismatch enq=%d deq =%d",
+                               port, ev->queue_id);
+       return 0;
+}
+
+/*
+ * Link all even number of queues to port 0 and all odd number of queues to
+ * port 1 and verify the link connection on dequeue
+ */
+static int
+test_queue_to_port_multi_link(void)
+{
+       int ret, port0_events = 0, port1_events = 0;
+       uint8_t queue, port;
+       uint32_t nr_queues = 0;
+       uint32_t nr_ports = 0;
+
+       RTE_TEST_ASSERT_SUCCESS(rte_event_dev_attr_get(evdev,
+                           RTE_EVENT_DEV_ATTR_QUEUE_COUNT,
+                           &nr_queues), "Queue count get failed");
+
+       RTE_TEST_ASSERT_SUCCESS(rte_event_dev_attr_get(evdev,
+                               RTE_EVENT_DEV_ATTR_QUEUE_COUNT,
+                               &nr_queues), "Queue count get failed");
+       RTE_TEST_ASSERT_SUCCESS(rte_event_dev_attr_get(evdev,
+                               RTE_EVENT_DEV_ATTR_PORT_COUNT,
+                               &nr_ports), "Port count get failed");
+
+       if (nr_ports < 2) {
+               dpaa2_evdev_err("%s: Not enough ports to test ports=%d",
+                               __func__, nr_ports);
+               return 0;
+       }
+
+       /* Unlink all connections that created in eventdev_setup */
+       for (port = 0; port < nr_ports; port++) {
+               ret = rte_event_port_unlink(evdev, port, NULL, 0);
+               RTE_TEST_ASSERT(ret >= 0, "Failed to unlink all queues port=%d",
+                                       port);
+       }
+
+       const unsigned int total_events = MAX_EVENTS / nr_queues;
+
+       /* Link all even number of queues to port0 and odd numbers to port 1*/
+       for (queue = 0; queue < nr_queues; queue++) {
+               port = queue & 0x1;
+               ret = rte_event_port_link(evdev, port, &queue, NULL, 1);
+               RTE_TEST_ASSERT(ret == 1, "Failed to link queue=%d to port=%d",
+                                       queue, port);
+
+               ret = inject_events(
+                       0x100 /*flow_id */,
+                       RTE_EVENT_TYPE_CPU /* event_type */,
+                       rte_rand() % 256 /* sub_event_type */,
+                       rte_rand() % (RTE_SCHED_TYPE_PARALLEL + 1),
+                       queue /* queue */,
+                       port /* port */,
+                       total_events /* events */);
+               if (ret)
+                       return -1;
+
+               if (port == 0)
+                       port0_events += total_events;
+               else
+                       port1_events += total_events;
+       }
+
+       ret = consume_events(0 /* port */, port0_events,
+                               validate_queue_to_port_multi_link);
+       if (ret)
+               return -1;
+       ret = consume_events(1 /* port */, port1_events,
+                               validate_queue_to_port_multi_link);
+       if (ret)
+               return -1;
+
+       return 0;
+}
+
+static void dpaa2_test_run(int (*setup)(void), void (*tdown)(void),
+               int (*test)(void), const char *name)
+{
+       if (setup() < 0) {
+               RTE_LOG(INFO, PMD, "Error setting up test %s", name);
+               unsupported++;
+       } else {
+               if (test() < 0) {
+                       failed++;
+                       RTE_LOG(INFO, PMD, "%s Failed\n", name);
+               } else {
+                       passed++;
+                       RTE_LOG(INFO, PMD, "%s Passed", name);
+               }
+       }
+
+       total++;
+       tdown();
+}
+
+int
+test_eventdev_dpaa2(void)
+{
+       testsuite_setup();
+
+       DPAA2_TEST_RUN(eventdev_setup, eventdev_teardown,
+                       test_simple_enqdeq_atomic);
+       DPAA2_TEST_RUN(eventdev_setup, eventdev_teardown,
+                       test_simple_enqdeq_parallel);
+       DPAA2_TEST_RUN(eventdev_setup, eventdev_teardown,
+                       test_multi_queue_enq_single_port_deq);
+       DPAA2_TEST_RUN(eventdev_setup, eventdev_teardown,
+                       test_dev_stop_flush);
+       DPAA2_TEST_RUN(eventdev_setup, eventdev_teardown,
+                       test_multi_queue_enq_multi_port_deq);
+       DPAA2_TEST_RUN(eventdev_setup, eventdev_teardown,
+                       test_queue_to_port_single_link);
+       DPAA2_TEST_RUN(eventdev_setup, eventdev_teardown,
+                       test_queue_to_port_multi_link);
+
+       DPAA2_EVENTDEV_INFO("Total tests   : %d", total);
+       DPAA2_EVENTDEV_INFO("Passed        : %d", passed);
+       DPAA2_EVENTDEV_INFO("Failed        : %d", failed);
+       DPAA2_EVENTDEV_INFO("Not supported : %d", unsupported);
+
+       testsuite_teardown();
+
+       if (failed)
+               return -1;
+
+       return 0;
+}
index f7da7fa..72f97d4 100644 (file)
@@ -9,7 +9,8 @@ if not is_linux
 endif
 deps += ['bus_vdev', 'pmd_dpaa2', 'pmd_dpaa2_sec']
 sources = files('dpaa2_hw_dpcon.c',
-               'dpaa2_eventdev.c')
+               'dpaa2_eventdev.c',
+               'dpaa2_eventdev_selftest.c')
 
 allow_experimental_apis = true
 includes += include_directories('../../crypto/dpaa2_sec/')