test/eventdev: add octeontx flow based max stage
authorJerin Jacob <jerin.jacob@caviumnetworks.com>
Fri, 3 Mar 2017 17:28:16 +0000 (22:58 +0530)
committerJerin Jacob <jerin.jacob@caviumnetworks.com>
Tue, 4 Apr 2017 17:19:53 +0000 (19:19 +0200)
Add flow based pipeline test with maximum number of stages available
in the device.

Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Acked-by: Harry van Haaren <harry.van.haaren@intel.com>
test/test/test_eventdev_octeontx.c

index cf79242..a96fe8e 100644 (file)
@@ -1098,6 +1098,74 @@ test_multi_port_queue_parallel_to_parallel(void)
        return test_multiport_queue_sched_type_test(RTE_SCHED_TYPE_PARALLEL,
                                RTE_SCHED_TYPE_PARALLEL);
 }
+
+static int
+worker_flow_based_pipeline_max_stages_rand_sched_type(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;
+
+       while (rte_atomic32_read(total_events) > 0) {
+               valid_event = rte_event_dequeue_burst(evdev, port, &ev, 1, 0);
+               if (!valid_event)
+                       continue;
+
+               if (ev.sub_event_type == 255) { /* last stage */
+                       rte_pktmbuf_free(ev.mbuf);
+                       rte_atomic32_sub(total_events, 1);
+               } else {
+                       ev.event_type = RTE_EVENT_TYPE_CPU;
+                       ev.sub_event_type++;
+                       ev.sched_type =
+                               rte_rand() % (RTE_SCHED_TYPE_PARALLEL + 1);
+                       ev.op = RTE_EVENT_OP_FORWARD;
+                       rte_event_enqueue_burst(evdev, port, &ev, 1);
+               }
+       }
+       return 0;
+}
+
+static int
+launch_multi_port_max_stages_random_sched_type(int (*fn)(void *))
+{
+       uint8_t nr_ports;
+       int ret;
+
+       nr_ports = RTE_MIN(rte_event_port_count(evdev), rte_lcore_count() - 1);
+
+       if (!nr_ports) {
+               printf("%s: Not enough ports=%d or workers=%d\n", __func__,
+                       rte_event_port_count(evdev), rte_lcore_count() - 1);
+               return TEST_SUCCESS;
+       }
+
+       /* Injects events with m->seqn=0 to total_events */
+       ret = inject_events(
+               0x1 /*flow_id */,
+               RTE_EVENT_TYPE_CPU /* event_type */,
+               0 /* sub_event_type (stage 0) */,
+               rte_rand() % (RTE_SCHED_TYPE_PARALLEL + 1) /* sched_type */,
+               0 /* queue */,
+               0 /* port */,
+               MAX_EVENTS /* events */);
+       if (ret)
+               return TEST_FAILED;
+
+       return launch_workers_and_wait(fn, fn, MAX_EVENTS, nr_ports,
+                                        0xff /* invalid */);
+}
+
+/* Flow based pipeline with maximum stages with random sched type */
+static int
+test_multi_port_flow_max_stages_random_sched_type(void)
+{
+       return launch_multi_port_max_stages_random_sched_type(
+               worker_flow_based_pipeline_max_stages_rand_sched_type);
+}
+
 static struct unit_test_suite eventdev_octeontx_testsuite  = {
        .suite_name = "eventdev octeontx unit test suite",
        .setup = testsuite_setup,
@@ -1155,6 +1223,8 @@ static struct unit_test_suite eventdev_octeontx_testsuite  = {
                        test_multi_port_queue_parallel_to_ordered),
                TEST_CASE_ST(eventdev_setup, eventdev_teardown,
                        test_multi_port_queue_parallel_to_parallel),
+               TEST_CASE_ST(eventdev_setup, eventdev_teardown,
+                       test_multi_port_flow_max_stages_random_sched_type),
                TEST_CASES_END() /**< NULL terminate unit test array */
        }
 };