test/eventdev: add octeontx flow based two stage sched
authorJerin Jacob <jerin.jacob@caviumnetworks.com>
Fri, 3 Mar 2017 17:28:14 +0000 (22:58 +0530)
committerJerin Jacob <jerin.jacob@caviumnetworks.com>
Tue, 4 Apr 2017 17:19:53 +0000 (19:19 +0200)
Add flow based two stage pipeline test with all combination
of schedule types.

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 cb73883..84d8e39 100644 (file)
@@ -74,6 +74,30 @@ seqn_list_init(void)
        seqn_list_index = 0;
 }
 
+static inline int
+seqn_list_update(int val)
+{
+       if (seqn_list_index >= NUM_PACKETS)
+               return TEST_FAILED;
+
+       seqn_list[seqn_list_index++] = val;
+       rte_smp_wmb();
+       return TEST_SUCCESS;
+}
+
+static inline int
+seqn_list_check(int limit)
+{
+       int i;
+
+       for (i = 0; i < limit; i++) {
+               if (seqn_list[i] != i) {
+                       printf("Seqn mismatch %d %d\n", seqn_list[i], i);
+                       return TEST_FAILED;
+               }
+       }
+       return TEST_SUCCESS;
+}
 
 struct test_core_param {
        rte_atomic32_t *total_events;
@@ -771,6 +795,158 @@ test_queue_to_port_multi_link(void)
        return TEST_SUCCESS;
 }
 
+static int
+worker_flow_based_pipeline(void *arg)
+{
+       struct test_core_param *param = arg;
+       struct rte_event ev;
+       uint16_t valid_event;
+       uint8_t port = param->port;
+       uint8_t new_sched_type = param->sched_type;
+       rte_atomic32_t *total_events = param->total_events;
+       uint64_t dequeue_tmo_ticks = param->dequeue_tmo_ticks;
+
+       while (rte_atomic32_read(total_events) > 0) {
+               valid_event = rte_event_dequeue_burst(evdev, port, &ev, 1,
+                                       dequeue_tmo_ticks);
+               if (!valid_event)
+                       continue;
+
+               /* Events from stage 0 */
+               if (ev.sub_event_type == 0) {
+                       /* Move to atomic flow to maintain the ordering */
+                       ev.flow_id = 0x2;
+                       ev.event_type = RTE_EVENT_TYPE_CPU;
+                       ev.sub_event_type = 1; /* stage 1 */
+                       ev.sched_type = new_sched_type;
+                       ev.op = RTE_EVENT_OP_FORWARD;
+                       rte_event_enqueue_burst(evdev, port, &ev, 1);
+               } else if (ev.sub_event_type == 1) { /* Events from stage 1*/
+                       if (seqn_list_update(ev.mbuf->seqn) == TEST_SUCCESS) {
+                               rte_pktmbuf_free(ev.mbuf);
+                               rte_atomic32_sub(total_events, 1);
+                       } else {
+                               printf("Failed to update seqn_list\n");
+                               return TEST_FAILED;
+                       }
+               } else {
+                       printf("Invalid ev.sub_event_type = %d\n",
+                                       ev.sub_event_type);
+                       return TEST_FAILED;
+               }
+       }
+       return 0;
+}
+
+static int
+test_multiport_flow_sched_type_test(uint8_t in_sched_type,
+                       uint8_t out_sched_type)
+{
+       const unsigned int total_events = MAX_EVENTS;
+       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) */,
+               in_sched_type,
+               0 /* queue */,
+               0 /* port */,
+               total_events /* events */);
+       if (ret)
+               return TEST_FAILED;
+
+       ret = launch_workers_and_wait(worker_flow_based_pipeline,
+                                       worker_flow_based_pipeline,
+                                       total_events, nr_ports, out_sched_type);
+       if (ret)
+               return TEST_FAILED;
+
+       if (in_sched_type != RTE_SCHED_TYPE_PARALLEL &&
+                       out_sched_type == RTE_SCHED_TYPE_ATOMIC) {
+               /* Check the events order maintained or not */
+               return seqn_list_check(total_events);
+       }
+       return TEST_SUCCESS;
+}
+
+
+/* Multi port ordered to atomic transaction */
+static int
+test_multi_port_flow_ordered_to_atomic(void)
+{
+       /* Ingress event order test */
+       return test_multiport_flow_sched_type_test(RTE_SCHED_TYPE_ORDERED,
+                               RTE_SCHED_TYPE_ATOMIC);
+}
+
+static int
+test_multi_port_flow_ordered_to_ordered(void)
+{
+       return test_multiport_flow_sched_type_test(RTE_SCHED_TYPE_ORDERED,
+                               RTE_SCHED_TYPE_ORDERED);
+}
+
+static int
+test_multi_port_flow_ordered_to_parallel(void)
+{
+       return test_multiport_flow_sched_type_test(RTE_SCHED_TYPE_ORDERED,
+                               RTE_SCHED_TYPE_PARALLEL);
+}
+
+static int
+test_multi_port_flow_atomic_to_atomic(void)
+{
+       /* Ingress event order test */
+       return test_multiport_flow_sched_type_test(RTE_SCHED_TYPE_ATOMIC,
+                               RTE_SCHED_TYPE_ATOMIC);
+}
+
+static int
+test_multi_port_flow_atomic_to_ordered(void)
+{
+       return test_multiport_flow_sched_type_test(RTE_SCHED_TYPE_ATOMIC,
+                               RTE_SCHED_TYPE_ORDERED);
+}
+
+static int
+test_multi_port_flow_atomic_to_parallel(void)
+{
+       return test_multiport_flow_sched_type_test(RTE_SCHED_TYPE_ATOMIC,
+                               RTE_SCHED_TYPE_PARALLEL);
+}
+
+static int
+test_multi_port_flow_parallel_to_atomic(void)
+{
+       return test_multiport_flow_sched_type_test(RTE_SCHED_TYPE_PARALLEL,
+                               RTE_SCHED_TYPE_ATOMIC);
+}
+
+static int
+test_multi_port_flow_parallel_to_ordered(void)
+{
+       return test_multiport_flow_sched_type_test(RTE_SCHED_TYPE_PARALLEL,
+                               RTE_SCHED_TYPE_ORDERED);
+}
+
+static int
+test_multi_port_flow_parallel_to_parallel(void)
+{
+       return test_multiport_flow_sched_type_test(RTE_SCHED_TYPE_PARALLEL,
+                               RTE_SCHED_TYPE_PARALLEL);
+}
+
 static struct unit_test_suite eventdev_octeontx_testsuite  = {
        .suite_name = "eventdev octeontx unit test suite",
        .setup = testsuite_setup,
@@ -792,6 +968,24 @@ static struct unit_test_suite eventdev_octeontx_testsuite  = {
                        test_queue_to_port_single_link),
                TEST_CASE_ST(eventdev_setup, eventdev_teardown,
                        test_queue_to_port_multi_link),
+               TEST_CASE_ST(eventdev_setup, eventdev_teardown,
+                       test_multi_port_flow_ordered_to_atomic),
+               TEST_CASE_ST(eventdev_setup, eventdev_teardown,
+                       test_multi_port_flow_ordered_to_ordered),
+               TEST_CASE_ST(eventdev_setup, eventdev_teardown,
+                       test_multi_port_flow_ordered_to_parallel),
+               TEST_CASE_ST(eventdev_setup, eventdev_teardown,
+                       test_multi_port_flow_atomic_to_atomic),
+               TEST_CASE_ST(eventdev_setup, eventdev_teardown,
+                       test_multi_port_flow_atomic_to_ordered),
+               TEST_CASE_ST(eventdev_setup, eventdev_teardown,
+                       test_multi_port_flow_atomic_to_parallel),
+               TEST_CASE_ST(eventdev_setup, eventdev_teardown,
+                       test_multi_port_flow_parallel_to_atomic),
+               TEST_CASE_ST(eventdev_setup, eventdev_teardown,
+                       test_multi_port_flow_parallel_to_ordered),
+               TEST_CASE_ST(eventdev_setup, eventdev_teardown,
+                       test_multi_port_flow_parallel_to_parallel),
                TEST_CASES_END() /**< NULL terminate unit test array */
        }
 };