app/testeventdev: launch order lcores
authorJerin Jacob <jerin.jacob@caviumnetworks.com>
Tue, 4 Jul 2017 04:53:10 +0000 (10:23 +0530)
committerJerin Jacob <jerin.jacob@caviumnetworks.com>
Fri, 7 Jul 2017 07:33:22 +0000 (09:33 +0200)
The event producer and master lcore's test end and
failure detection logic are common for the queue and
all types queue test.Move them as the common function.

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

index 01b73bb..4914808 100644 (file)
@@ -41,6 +41,57 @@ order_test_result(struct evt_test *test, struct evt_options *opt)
        return t->result;
 }
 
+static inline int
+order_producer(void *arg)
+{
+       struct prod_data *p  = arg;
+       struct test_order *t = p->t;
+       struct evt_options *opt = t->opt;
+       const uint8_t dev_id = p->dev_id;
+       const uint8_t port = p->port_id;
+       struct rte_mempool *pool = t->pool;
+       const uint64_t nb_pkts = t->nb_pkts;
+       uint32_t *producer_flow_seq = t->producer_flow_seq;
+       const uint32_t nb_flows = t->nb_flows;
+       uint64_t count = 0;
+       struct rte_mbuf *m;
+       struct rte_event ev;
+
+       if (opt->verbose_level > 1)
+               printf("%s(): lcore %d dev_id %d port=%d queue=%d\n",
+                        __func__, rte_lcore_id(), dev_id, port, p->queue_id);
+
+       ev.event = 0;
+       ev.op = RTE_EVENT_OP_NEW;
+       ev.queue_id = p->queue_id;
+       ev.sched_type = RTE_SCHED_TYPE_ORDERED;
+       ev.priority = RTE_EVENT_DEV_PRIORITY_NORMAL;
+       ev.event_type =  RTE_EVENT_TYPE_CPU;
+       ev.sub_event_type = 0; /* stage 0 */
+
+       while (count < nb_pkts && t->err == false) {
+               m = rte_pktmbuf_alloc(pool);
+               if (m == NULL)
+                       continue;
+
+               const uint32_t flow = (uintptr_t)m % nb_flows;
+               /* Maintain seq number per flow */
+               m->seqn = producer_flow_seq[flow]++;
+
+               ev.flow_id = flow;
+               ev.mbuf = m;
+
+               while (rte_event_enqueue_burst(dev_id, port, &ev, 1) != 1) {
+                       if (t->err)
+                               break;
+                       rte_pause();
+               }
+
+               count++;
+       }
+       return 0;
+}
+
 int
 order_opt_check(struct evt_options *opt)
 {
@@ -206,6 +257,71 @@ order_opt_dump(struct evt_options *opt)
        evt_dump("nb_evdev_ports", "%d", order_nb_event_ports(opt));
 }
 
+int
+order_launch_lcores(struct evt_test *test, struct evt_options *opt,
+                       int (*worker)(void *))
+{
+       int ret, lcore_id;
+       struct test_order *t = evt_test_priv(test);
+
+       int wkr_idx = 0;
+       /* launch workers */
+       RTE_LCORE_FOREACH_SLAVE(lcore_id) {
+               if (!(opt->wlcores[lcore_id]))
+                       continue;
+
+               ret = rte_eal_remote_launch(worker, &t->worker[wkr_idx],
+                                        lcore_id);
+               if (ret) {
+                       evt_err("failed to launch worker %d", lcore_id);
+                       return ret;
+               }
+               wkr_idx++;
+       }
+
+       /* launch producer */
+       int plcore = evt_get_first_active_lcore(opt->plcores);
+
+       ret = rte_eal_remote_launch(order_producer, &t->prod, plcore);
+       if (ret) {
+               evt_err("failed to launch order_producer %d", plcore);
+               return ret;
+       }
+
+       uint64_t cycles = rte_get_timer_cycles();
+       int64_t old_remaining  = -1;
+
+       while (t->err == false) {
+
+               rte_event_schedule(opt->dev_id);
+
+               uint64_t new_cycles = rte_get_timer_cycles();
+               int64_t remaining = rte_atomic64_read(&t->outstand_pkts);
+
+               if (remaining <= 0) {
+                       t->result = EVT_TEST_SUCCESS;
+                       break;
+               }
+
+               if (new_cycles - cycles > rte_get_timer_hz() * 1) {
+                       printf(CLGRN"\r%"PRId64""CLNRM, remaining);
+                       fflush(stdout);
+                       if (old_remaining == remaining) {
+                               rte_event_dev_dump(opt->dev_id, stdout);
+                               evt_err("No schedules for seconds, deadlock");
+                               t->err = true;
+                               rte_smp_wmb();
+                               break;
+                       }
+                       old_remaining = remaining;
+                       cycles = new_cycles;
+               }
+       }
+       printf("\r");
+
+       return 0;
+}
+
 int
 order_event_dev_port_setup(struct evt_test *test, struct evt_options *opt,
                                uint8_t nb_workers, uint8_t nb_queues)
index 1659318..a760b94 100644 (file)
@@ -94,6 +94,8 @@ int order_test_result(struct evt_test *test, struct evt_options *opt);
 int order_opt_check(struct evt_options *opt);
 int order_test_setup(struct evt_test *test, struct evt_options *opt);
 int order_mempool_setup(struct evt_test *test, struct evt_options *opt);
+int order_launch_lcores(struct evt_test *test, struct evt_options *opt,
+                       int (*worker)(void *));
 int order_event_dev_port_setup(struct evt_test *test, struct evt_options *opt,
                                uint8_t nb_workers, uint8_t nb_queues);
 void order_test_destroy(struct evt_test *test, struct evt_options *opt);