test/eventdev: add basic SW tests
authorHarry van Haaren <harry.van.haaren@intel.com>
Thu, 30 Mar 2017 19:30:44 +0000 (20:30 +0100)
committerJerin Jacob <jerin.jacob@caviumnetworks.com>
Tue, 4 Apr 2017 17:19:52 +0000 (19:19 +0200)
This commit adds basic enqueue and dequeue unit tests,
some negative invalid tests, and configuration.

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Signed-off-by: David Hunt <david.hunt@intel.com>
Signed-off-by: Harry van Haaren <harry.van.haaren@intel.com>
Acked-by: Anatoly Burakov <anatoly.burakov@intel.com>
test/test/test_eventdev_sw.c

index dfbff00..f294cb9 100644 (file)
@@ -64,6 +64,178 @@ struct test {
        int nb_qids;
 };
 
+static struct rte_event release_ev;
+
+static inline struct rte_mbuf *
+rte_gen_arp(int portid, struct rte_mempool *mp)
+{
+       /*
+        * len = 14 + 46
+        * ARP, Request who-has 10.0.0.1 tell 10.0.0.2, length 46
+        */
+       static const uint8_t arp_request[] = {
+               /*0x0000:*/ 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xec, 0xa8,
+               0x6b, 0xfd, 0x02, 0x29, 0x08, 0x06, 0x00, 0x01,
+               /*0x0010:*/ 0x08, 0x00, 0x06, 0x04, 0x00, 0x01, 0xec, 0xa8,
+               0x6b, 0xfd, 0x02, 0x29, 0x0a, 0x00, 0x00, 0x01,
+               /*0x0020:*/ 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x0a, 0x00,
+               0x00, 0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+               /*0x0030:*/ 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+               0x00, 0x00, 0x00, 0x00
+       };
+       struct rte_mbuf *m;
+       int pkt_len = sizeof(arp_request) - 1;
+
+       m = rte_pktmbuf_alloc(mp);
+       if (!m)
+               return 0;
+
+       memcpy((void *)((uintptr_t)m->buf_addr + m->data_off),
+               arp_request, pkt_len);
+       rte_pktmbuf_pkt_len(m) = pkt_len;
+       rte_pktmbuf_data_len(m) = pkt_len;
+
+       RTE_SET_USED(portid);
+
+       return m;
+}
+
+/* initialization and config */
+static inline int
+init(struct test *t, int nb_queues, int nb_ports)
+{
+       struct rte_event_dev_config config = {
+                       .nb_event_queues = nb_queues,
+                       .nb_event_ports = nb_ports,
+                       .nb_event_queue_flows = 1024,
+                       .nb_events_limit = 4096,
+                       .nb_event_port_dequeue_depth = 128,
+                       .nb_event_port_enqueue_depth = 128,
+       };
+       int ret;
+
+       void *temp = t->mbuf_pool; /* save and restore mbuf pool */
+
+       memset(t, 0, sizeof(*t));
+       t->mbuf_pool = temp;
+
+       ret = rte_event_dev_configure(evdev, &config);
+       if (ret < 0)
+               printf("%d: Error configuring device\n", __LINE__);
+       return ret;
+};
+
+static inline int
+create_ports(struct test *t, int num_ports)
+{
+       int i;
+       static const struct rte_event_port_conf conf = {
+                       .new_event_threshold = 1024,
+                       .dequeue_depth = 32,
+                       .enqueue_depth = 64,
+       };
+       if (num_ports > MAX_PORTS)
+               return -1;
+
+       for (i = 0; i < num_ports; i++) {
+               if (rte_event_port_setup(evdev, i, &conf) < 0) {
+                       printf("Error setting up port %d\n", i);
+                       return -1;
+               }
+               t->port[i] = i;
+       }
+
+       return 0;
+}
+
+static inline int
+create_lb_qids(struct test *t, int num_qids, uint32_t flags)
+{
+       int i;
+
+       /* Q creation */
+       const struct rte_event_queue_conf conf = {
+                       .event_queue_cfg = flags,
+                       .priority = RTE_EVENT_DEV_PRIORITY_NORMAL,
+                       .nb_atomic_flows = 1024,
+                       .nb_atomic_order_sequences = 1024,
+       };
+
+       for (i = t->nb_qids; i < t->nb_qids + num_qids; i++) {
+               if (rte_event_queue_setup(evdev, i, &conf) < 0) {
+                       printf("%d: error creating qid %d\n", __LINE__, i);
+                       return -1;
+               }
+               t->qid[i] = i;
+       }
+       t->nb_qids += num_qids;
+       if (t->nb_qids > MAX_QIDS)
+               return -1;
+
+       return 0;
+}
+
+static inline int
+create_atomic_qids(struct test *t, int num_qids)
+{
+       return create_lb_qids(t, num_qids, RTE_EVENT_QUEUE_CFG_ATOMIC_ONLY);
+}
+
+static inline int
+create_ordered_qids(struct test *t, int num_qids)
+{
+       return create_lb_qids(t, num_qids, RTE_EVENT_QUEUE_CFG_ORDERED_ONLY);
+}
+
+
+static inline int
+create_unordered_qids(struct test *t, int num_qids)
+{
+       return create_lb_qids(t, num_qids, RTE_EVENT_QUEUE_CFG_PARALLEL_ONLY);
+}
+
+static inline int
+create_directed_qids(struct test *t, int num_qids, const uint8_t ports[])
+{
+       int i;
+
+       /* Q creation */
+       static const struct rte_event_queue_conf conf = {
+                       .priority = RTE_EVENT_DEV_PRIORITY_NORMAL,
+                       .event_queue_cfg = RTE_EVENT_QUEUE_CFG_SINGLE_LINK,
+                       .nb_atomic_flows = 1024,
+                       .nb_atomic_order_sequences = 1024,
+       };
+
+       for (i = t->nb_qids; i < t->nb_qids + num_qids; i++) {
+               if (rte_event_queue_setup(evdev, i, &conf) < 0) {
+                       printf("%d: error creating qid %d\n", __LINE__, i);
+                       return -1;
+               }
+               t->qid[i] = i;
+
+               if (rte_event_port_link(evdev, ports[i - t->nb_qids],
+                               &t->qid[i], NULL, 1) != 1) {
+                       printf("%d: error creating link for qid %d\n",
+                                       __LINE__, i);
+                       return -1;
+               }
+       }
+       t->nb_qids += num_qids;
+       if (t->nb_qids > MAX_QIDS)
+               return -1;
+
+       return 0;
+}
+
+/* destruction */
+static inline int
+cleanup(struct test *t __rte_unused)
+{
+       rte_event_dev_stop(evdev);
+       rte_event_dev_close(evdev);
+       return 0;
+};
 
 struct test_event_dev_stats {
        uint64_t rx_pkts;       /**< Total packets received */
@@ -86,12 +258,1055 @@ struct test_event_dev_stats {
        uint64_t qid_tx_pkts[MAX_QIDS];
 };
 
+static inline int
+test_event_dev_stats_get(int dev_id, struct test_event_dev_stats *stats)
+{
+       static uint32_t i;
+       static uint32_t total_ids[3]; /* rx, tx and drop */
+       static uint32_t port_rx_pkts_ids[MAX_PORTS];
+       static uint32_t port_rx_dropped_ids[MAX_PORTS];
+       static uint32_t port_inflight_ids[MAX_PORTS];
+       static uint32_t port_tx_pkts_ids[MAX_PORTS];
+       static uint32_t qid_rx_pkts_ids[MAX_QIDS];
+       static uint32_t qid_rx_dropped_ids[MAX_QIDS];
+       static uint32_t qid_tx_pkts_ids[MAX_QIDS];
+
+
+       stats->rx_pkts = rte_event_dev_xstats_by_name_get(dev_id,
+                       "dev_rx", &total_ids[0]);
+       stats->rx_dropped = rte_event_dev_xstats_by_name_get(dev_id,
+                       "dev_drop", &total_ids[1]);
+       stats->tx_pkts = rte_event_dev_xstats_by_name_get(dev_id,
+                       "dev_tx", &total_ids[2]);
+       for (i = 0; i < MAX_PORTS; i++) {
+               char name[32];
+               snprintf(name, sizeof(name), "port_%u_rx", i);
+               stats->port_rx_pkts[i] = rte_event_dev_xstats_by_name_get(
+                               dev_id, name, &port_rx_pkts_ids[i]);
+               snprintf(name, sizeof(name), "port_%u_drop", i);
+               stats->port_rx_dropped[i] = rte_event_dev_xstats_by_name_get(
+                               dev_id, name, &port_rx_dropped_ids[i]);
+               snprintf(name, sizeof(name), "port_%u_inflight", i);
+               stats->port_inflight[i] = rte_event_dev_xstats_by_name_get(
+                               dev_id, name, &port_inflight_ids[i]);
+               snprintf(name, sizeof(name), "port_%u_tx", i);
+               stats->port_tx_pkts[i] = rte_event_dev_xstats_by_name_get(
+                               dev_id, name, &port_tx_pkts_ids[i]);
+       }
+       for (i = 0; i < MAX_QIDS; i++) {
+               char name[32];
+               snprintf(name, sizeof(name), "qid_%u_rx", i);
+               stats->qid_rx_pkts[i] = rte_event_dev_xstats_by_name_get(
+                               dev_id, name, &qid_rx_pkts_ids[i]);
+               snprintf(name, sizeof(name), "qid_%u_drop", i);
+               stats->qid_rx_dropped[i] = rte_event_dev_xstats_by_name_get(
+                               dev_id, name, &qid_rx_dropped_ids[i]);
+               snprintf(name, sizeof(name), "qid_%u_tx", i);
+               stats->qid_tx_pkts[i] = rte_event_dev_xstats_by_name_get(
+                               dev_id, name, &qid_tx_pkts_ids[i]);
+       }
+
+       return 0;
+}
+
+static int
+test_single_directed_packet(struct test *t)
+{
+       const int rx_enq = 0;
+       const int wrk_enq = 2;
+       int err;
+
+       /* Create instance with 3 directed QIDs going to 3 ports */
+       if (init(t, 3, 3) < 0 ||
+                       create_ports(t, 3) < 0 ||
+                       create_directed_qids(t, 3, t->port) < 0)
+               return -1;
+
+       if (rte_event_dev_start(evdev) < 0) {
+               printf("%d: Error with start call\n", __LINE__);
+               return -1;
+       }
+
+       /************** FORWARD ****************/
+       struct rte_mbuf *arp = rte_gen_arp(0, t->mbuf_pool);
+       struct rte_event ev = {
+                       .op = RTE_EVENT_OP_NEW,
+                       .queue_id = wrk_enq,
+                       .mbuf = arp,
+       };
+
+       if (!arp) {
+               printf("%d: gen of pkt failed\n", __LINE__);
+               return -1;
+       }
+
+       const uint32_t MAGIC_SEQN = 4711;
+       arp->seqn = MAGIC_SEQN;
+
+       /* generate pkt and enqueue */
+       err = rte_event_enqueue_burst(evdev, rx_enq, &ev, 1);
+       if (err < 0) {
+               printf("%d: error failed to enqueue\n", __LINE__);
+               return -1;
+       }
+
+       /* Run schedule() as dir packets may need to be re-ordered */
+       rte_event_schedule(evdev);
+
+       struct test_event_dev_stats stats;
+       err = test_event_dev_stats_get(evdev, &stats);
+       if (err) {
+               printf("%d: error failed to get stats\n", __LINE__);
+               return -1;
+       }
+
+       if (stats.port_rx_pkts[rx_enq] != 1) {
+               printf("%d: error stats incorrect for directed port\n",
+                               __LINE__);
+               return -1;
+       }
+
+       uint32_t deq_pkts;
+       deq_pkts = rte_event_dequeue_burst(evdev, wrk_enq, &ev, 1, 0);
+       if (deq_pkts != 1) {
+               printf("%d: error failed to deq\n", __LINE__);
+               return -1;
+       }
+
+       err = test_event_dev_stats_get(evdev, &stats);
+       if (stats.port_rx_pkts[wrk_enq] != 0 &&
+                       stats.port_rx_pkts[wrk_enq] != 1) {
+               printf("%d: error directed stats post-dequeue\n", __LINE__);
+               return -1;
+       }
+
+       if (ev.mbuf->seqn != MAGIC_SEQN) {
+               printf("%d: error magic sequence number not dequeued\n",
+                               __LINE__);
+               return -1;
+       }
+
+       rte_pktmbuf_free(ev.mbuf);
+       cleanup(t);
+       return 0;
+}
+
+static int
+burst_packets(struct test *t)
+{
+       /************** CONFIG ****************/
+       uint32_t i;
+       int err;
+       int ret;
+
+       /* Create instance with 2 ports and 2 queues */
+       if (init(t, 2, 2) < 0 ||
+                       create_ports(t, 2) < 0 ||
+                       create_atomic_qids(t, 2) < 0) {
+               printf("%d: Error initializing device\n", __LINE__);
+               return -1;
+       }
+
+       /* CQ mapping to QID */
+       ret = rte_event_port_link(evdev, t->port[0], &t->qid[0], NULL, 1);
+       if (ret != 1) {
+               printf("%d: error mapping lb qid0\n", __LINE__);
+               return -1;
+       }
+       ret = rte_event_port_link(evdev, t->port[1], &t->qid[1], NULL, 1);
+       if (ret != 1) {
+               printf("%d: error mapping lb qid1\n", __LINE__);
+               return -1;
+       }
+
+       if (rte_event_dev_start(evdev) < 0) {
+               printf("%d: Error with start call\n", __LINE__);
+               return -1;
+       }
+
+       /************** FORWARD ****************/
+       const uint32_t rx_port = 0;
+       const uint32_t NUM_PKTS = 2;
+
+       for (i = 0; i < NUM_PKTS; i++) {
+               struct rte_mbuf *arp = rte_gen_arp(0, t->mbuf_pool);
+               if (!arp) {
+                       printf("%d: error generating pkt\n", __LINE__);
+                       return -1;
+               }
+
+               struct rte_event ev = {
+                               .op = RTE_EVENT_OP_NEW,
+                               .queue_id = i % 2,
+                               .flow_id = i % 3,
+                               .mbuf = arp,
+               };
+               /* generate pkt and enqueue */
+               err = rte_event_enqueue_burst(evdev, t->port[rx_port], &ev, 1);
+               if (err < 0) {
+                       printf("%d: Failed to enqueue\n", __LINE__);
+                       return -1;
+               }
+       }
+       rte_event_schedule(evdev);
+
+       /* Check stats for all NUM_PKTS arrived to sched core */
+       struct test_event_dev_stats stats;
+
+       err = test_event_dev_stats_get(evdev, &stats);
+       if (err) {
+               printf("%d: failed to get stats\n", __LINE__);
+               return -1;
+       }
+       if (stats.rx_pkts != NUM_PKTS || stats.tx_pkts != NUM_PKTS) {
+               printf("%d: Sched core didn't receive all %d pkts\n",
+                               __LINE__, NUM_PKTS);
+               rte_event_dev_dump(evdev, stdout);
+               return -1;
+       }
+
+       uint32_t deq_pkts;
+       int p;
+
+       deq_pkts = 0;
+       /******** DEQ QID 1 *******/
+       do {
+               struct rte_event ev;
+               p = rte_event_dequeue_burst(evdev, t->port[0], &ev, 1, 0);
+               deq_pkts += p;
+               rte_pktmbuf_free(ev.mbuf);
+       } while (p);
+
+       if (deq_pkts != NUM_PKTS/2) {
+               printf("%d: Half of NUM_PKTS didn't arrive at port 1\n",
+                               __LINE__);
+               return -1;
+       }
+
+       /******** DEQ QID 2 *******/
+       deq_pkts = 0;
+       do {
+               struct rte_event ev;
+               p = rte_event_dequeue_burst(evdev, t->port[1], &ev, 1, 0);
+               deq_pkts += p;
+               rte_pktmbuf_free(ev.mbuf);
+       } while (p);
+       if (deq_pkts != NUM_PKTS/2) {
+               printf("%d: Half of NUM_PKTS didn't arrive at port 2\n",
+                               __LINE__);
+               return -1;
+       }
+
+       cleanup(t);
+       return 0;
+}
+
+static int
+abuse_inflights(struct test *t)
+{
+       const int rx_enq = 0;
+       const int wrk_enq = 2;
+       int err;
+
+       /* Create instance with 4 ports */
+       if (init(t, 1, 4) < 0 ||
+                       create_ports(t, 4) < 0 ||
+                       create_atomic_qids(t, 1) < 0) {
+               printf("%d: Error initializing device\n", __LINE__);
+               return -1;
+       }
+
+       /* CQ mapping to QID */
+       err = rte_event_port_link(evdev, t->port[wrk_enq], NULL, NULL, 0);
+       if (err != 1) {
+               printf("%d: error mapping lb qid\n", __LINE__);
+               cleanup(t);
+               return -1;
+       }
+
+       if (rte_event_dev_start(evdev) < 0) {
+               printf("%d: Error with start call\n", __LINE__);
+               return -1;
+       }
+
+       /* Enqueue op only */
+       err = rte_event_enqueue_burst(evdev, t->port[rx_enq], &release_ev, 1);
+       if (err < 0) {
+               printf("%d: Failed to enqueue\n", __LINE__);
+               return -1;
+       }
+
+       /* schedule */
+       rte_event_schedule(evdev);
+
+       struct test_event_dev_stats stats;
+
+       err = test_event_dev_stats_get(evdev, &stats);
+       if (err) {
+               printf("%d: failed to get stats\n", __LINE__);
+               return -1;
+       }
+
+       if (stats.rx_pkts != 0 ||
+                       stats.tx_pkts != 0 ||
+                       stats.port_inflight[wrk_enq] != 0) {
+               printf("%d: Sched core didn't handle pkt as expected\n",
+                               __LINE__);
+               return -1;
+       }
+
+       cleanup(t);
+       return 0;
+}
+
+static int
+port_reconfig_credits(struct test *t)
+{
+       if (init(t, 1, 1) < 0) {
+               printf("%d: Error initializing device\n", __LINE__);
+               return -1;
+       }
+
+       uint32_t i;
+       const uint32_t NUM_ITERS = 32;
+       for (i = 0; i < NUM_ITERS; i++) {
+               const struct rte_event_queue_conf conf = {
+                       .event_queue_cfg = RTE_EVENT_QUEUE_CFG_ATOMIC_ONLY,
+                       .priority = RTE_EVENT_DEV_PRIORITY_NORMAL,
+                       .nb_atomic_flows = 1024,
+                       .nb_atomic_order_sequences = 1024,
+               };
+               if (rte_event_queue_setup(evdev, 0, &conf) < 0) {
+                       printf("%d: error creating qid\n", __LINE__);
+                       return -1;
+               }
+               t->qid[0] = 0;
+
+               static const struct rte_event_port_conf port_conf = {
+                               .new_event_threshold = 128,
+                               .dequeue_depth = 32,
+                               .enqueue_depth = 64,
+               };
+               if (rte_event_port_setup(evdev, 0, &port_conf) < 0) {
+                       printf("%d Error setting up port\n", __LINE__);
+                       return -1;
+               }
+
+               int links = rte_event_port_link(evdev, 0, NULL, NULL, 0);
+               if (links != 1) {
+                       printf("%d: error mapping lb qid\n", __LINE__);
+                       goto fail;
+               }
+
+               if (rte_event_dev_start(evdev) < 0) {
+                       printf("%d: Error with start call\n", __LINE__);
+                       goto fail;
+               }
+
+               const uint32_t NPKTS = 1;
+               uint32_t j;
+               for (j = 0; j < NPKTS; j++) {
+                       struct rte_event ev;
+                       struct rte_mbuf *arp = rte_gen_arp(0, t->mbuf_pool);
+                       if (!arp) {
+                               printf("%d: gen of pkt failed\n", __LINE__);
+                               goto fail;
+                       }
+                       ev.queue_id = t->qid[0];
+                       ev.op = RTE_EVENT_OP_NEW;
+                       ev.mbuf = arp;
+                       int err = rte_event_enqueue_burst(evdev, 0, &ev, 1);
+                       if (err != 1) {
+                               printf("%d: Failed to enqueue\n", __LINE__);
+                               rte_event_dev_dump(0, stdout);
+                               goto fail;
+                       }
+               }
+
+               rte_event_schedule(evdev);
+
+               struct rte_event ev[NPKTS];
+               int deq = rte_event_dequeue_burst(evdev, t->port[0], ev,
+                                                       NPKTS, 0);
+               if (deq != 1)
+                       printf("%d error; no packet dequeued\n", __LINE__);
+
+               /* let cleanup below stop the device on last iter */
+               if (i != NUM_ITERS-1)
+                       rte_event_dev_stop(evdev);
+       }
+
+       cleanup(t);
+       return 0;
+fail:
+       cleanup(t);
+       return -1;
+}
+
+static int
+port_single_lb_reconfig(struct test *t)
+{
+       if (init(t, 2, 2) < 0) {
+               printf("%d: Error initializing device\n", __LINE__);
+               goto fail;
+       }
+
+       static const struct rte_event_queue_conf conf_lb_atomic = {
+               .priority = RTE_EVENT_DEV_PRIORITY_NORMAL,
+               .event_queue_cfg = RTE_EVENT_QUEUE_CFG_ATOMIC_ONLY,
+               .nb_atomic_flows = 1024,
+               .nb_atomic_order_sequences = 1024,
+       };
+       if (rte_event_queue_setup(evdev, 0, &conf_lb_atomic) < 0) {
+               printf("%d: error creating qid\n", __LINE__);
+               goto fail;
+       }
+
+       static const struct rte_event_queue_conf conf_single_link = {
+               .priority = RTE_EVENT_DEV_PRIORITY_NORMAL,
+               .event_queue_cfg = RTE_EVENT_QUEUE_CFG_SINGLE_LINK,
+               .nb_atomic_flows = 1024,
+               .nb_atomic_order_sequences = 1024,
+       };
+       if (rte_event_queue_setup(evdev, 1, &conf_single_link) < 0) {
+               printf("%d: error creating qid\n", __LINE__);
+               goto fail;
+       }
+
+       struct rte_event_port_conf port_conf = {
+               .new_event_threshold = 128,
+               .dequeue_depth = 32,
+               .enqueue_depth = 64,
+       };
+       if (rte_event_port_setup(evdev, 0, &port_conf) < 0) {
+               printf("%d Error setting up port\n", __LINE__);
+               goto fail;
+       }
+       if (rte_event_port_setup(evdev, 1, &port_conf) < 0) {
+               printf("%d Error setting up port\n", __LINE__);
+               goto fail;
+       }
+
+       /* link port to lb queue */
+       uint8_t queue_id = 0;
+       if (rte_event_port_link(evdev, 0, &queue_id, NULL, 1) != 1) {
+               printf("%d: error creating link for qid\n", __LINE__);
+               goto fail;
+       }
+
+       int ret = rte_event_port_unlink(evdev, 0, &queue_id, 1);
+       if (ret != 1) {
+               printf("%d: Error unlinking lb port\n", __LINE__);
+               goto fail;
+       }
+
+       queue_id = 1;
+       if (rte_event_port_link(evdev, 0, &queue_id, NULL, 1) != 1) {
+               printf("%d: error creating link for qid\n", __LINE__);
+               goto fail;
+       }
+
+       queue_id = 0;
+       int err = rte_event_port_link(evdev, 1, &queue_id, NULL, 1);
+       if (err != 1) {
+               printf("%d: error mapping lb qid\n", __LINE__);
+               goto fail;
+       }
+
+       if (rte_event_dev_start(evdev) < 0) {
+               printf("%d: Error with start call\n", __LINE__);
+               goto fail;
+       }
+
+       cleanup(t);
+       return 0;
+fail:
+       cleanup(t);
+       return -1;
+}
+
+static int
+ordered_reconfigure(struct test *t)
+{
+       if (init(t, 1, 1) < 0 ||
+                       create_ports(t, 1) < 0) {
+               printf("%d: Error initializing device\n", __LINE__);
+               return -1;
+       }
+
+       const struct rte_event_queue_conf conf = {
+                       .event_queue_cfg = RTE_EVENT_QUEUE_CFG_ORDERED_ONLY,
+                       .priority = RTE_EVENT_DEV_PRIORITY_NORMAL,
+                       .nb_atomic_flows = 1024,
+                       .nb_atomic_order_sequences = 1024,
+       };
+
+       if (rte_event_queue_setup(evdev, 0, &conf) < 0) {
+               printf("%d: error creating qid\n", __LINE__);
+               goto failed;
+       }
+
+       if (rte_event_queue_setup(evdev, 0, &conf) < 0) {
+               printf("%d: error creating qid, for 2nd time\n", __LINE__);
+               goto failed;
+       }
+
+       rte_event_port_link(evdev, t->port[0], NULL, NULL, 0);
+       if (rte_event_dev_start(evdev) < 0) {
+               printf("%d: Error with start call\n", __LINE__);
+               return -1;
+       }
+
+       cleanup(t);
+       return 0;
+failed:
+       cleanup(t);
+       return -1;
+}
+
+static int
+invalid_qid(struct test *t)
+{
+       struct test_event_dev_stats stats;
+       const int rx_enq = 0;
+       int err;
+       uint32_t i;
+
+       if (init(t, 1, 4) < 0 ||
+                       create_ports(t, 4) < 0 ||
+                       create_atomic_qids(t, 1) < 0) {
+               printf("%d: Error initializing device\n", __LINE__);
+               return -1;
+       }
+
+       /* CQ mapping to QID */
+       for (i = 0; i < 4; i++) {
+               err = rte_event_port_link(evdev, t->port[i], &t->qid[0],
+                               NULL, 1);
+               if (err != 1) {
+                       printf("%d: error mapping port 1 qid\n", __LINE__);
+                       return -1;
+               }
+       }
+
+       if (rte_event_dev_start(evdev) < 0) {
+               printf("%d: Error with start call\n", __LINE__);
+               return -1;
+       }
+
+       /*
+        * Send in a packet with an invalid qid to the scheduler.
+        * We should see the packed enqueued OK, but the inflights for
+        * that packet should not be incremented, and the rx_dropped
+        * should be incremented.
+        */
+       static uint32_t flows1[] = {20};
+
+       for (i = 0; i < RTE_DIM(flows1); i++) {
+               struct rte_mbuf *arp = rte_gen_arp(0, t->mbuf_pool);
+               if (!arp) {
+                       printf("%d: gen of pkt failed\n", __LINE__);
+                       return -1;
+               }
+
+               struct rte_event ev = {
+                               .op = RTE_EVENT_OP_NEW,
+                               .queue_id = t->qid[0] + flows1[i],
+                               .flow_id = i,
+                               .mbuf = arp,
+               };
+               /* generate pkt and enqueue */
+               err = rte_event_enqueue_burst(evdev, t->port[rx_enq], &ev, 1);
+               if (err < 0) {
+                       printf("%d: Failed to enqueue\n", __LINE__);
+                       return -1;
+               }
+       }
+
+       /* call the scheduler */
+       rte_event_schedule(evdev);
+
+       err = test_event_dev_stats_get(evdev, &stats);
+       if (err) {
+               printf("%d: failed to get stats\n", __LINE__);
+               return -1;
+       }
+
+       /*
+        * Now check the resulting inflights on the port, and the rx_dropped.
+        */
+       if (stats.port_inflight[0] != 0) {
+               printf("%d:%s: port 1 inflight count not correct\n", __LINE__,
+                               __func__);
+               rte_event_dev_dump(evdev, stdout);
+               return -1;
+       }
+       if (stats.port_rx_dropped[0] != 1) {
+               printf("%d:%s: port 1 drops\n", __LINE__, __func__);
+               rte_event_dev_dump(evdev, stdout);
+               return -1;
+       }
+       /* each packet drop should only be counted in one place - port or dev */
+       if (stats.rx_dropped != 0) {
+               printf("%d:%s: port 1 dropped count not correct\n", __LINE__,
+                               __func__);
+               rte_event_dev_dump(evdev, stdout);
+               return -1;
+       }
+
+       cleanup(t);
+       return 0;
+}
+
+static int
+single_packet(struct test *t)
+{
+       const uint32_t MAGIC_SEQN = 7321;
+       struct rte_event ev;
+       struct test_event_dev_stats stats;
+       const int rx_enq = 0;
+       const int wrk_enq = 2;
+       int err;
+
+       /* Create instance with 4 ports */
+       if (init(t, 1, 4) < 0 ||
+                       create_ports(t, 4) < 0 ||
+                       create_atomic_qids(t, 1) < 0) {
+               printf("%d: Error initializing device\n", __LINE__);
+               return -1;
+       }
+
+       /* CQ mapping to QID */
+       err = rte_event_port_link(evdev, t->port[wrk_enq], NULL, NULL, 0);
+       if (err != 1) {
+               printf("%d: error mapping lb qid\n", __LINE__);
+               cleanup(t);
+               return -1;
+       }
+
+       if (rte_event_dev_start(evdev) < 0) {
+               printf("%d: Error with start call\n", __LINE__);
+               return -1;
+       }
+
+       /************** Gen pkt and enqueue ****************/
+       struct rte_mbuf *arp = rte_gen_arp(0, t->mbuf_pool);
+       if (!arp) {
+               printf("%d: gen of pkt failed\n", __LINE__);
+               return -1;
+       }
+
+       ev.op = RTE_EVENT_OP_NEW;
+       ev.priority = RTE_EVENT_DEV_PRIORITY_NORMAL;
+       ev.mbuf = arp;
+       ev.queue_id = 0;
+       ev.flow_id = 3;
+       arp->seqn = MAGIC_SEQN;
+
+       err = rte_event_enqueue_burst(evdev, t->port[rx_enq], &ev, 1);
+       if (err < 0) {
+               printf("%d: Failed to enqueue\n", __LINE__);
+               return -1;
+       }
+
+       rte_event_schedule(evdev);
+
+       err = test_event_dev_stats_get(evdev, &stats);
+       if (err) {
+               printf("%d: failed to get stats\n", __LINE__);
+               return -1;
+       }
+
+       if (stats.rx_pkts != 1 ||
+                       stats.tx_pkts != 1 ||
+                       stats.port_inflight[wrk_enq] != 1) {
+               printf("%d: Sched core didn't handle pkt as expected\n",
+                               __LINE__);
+               rte_event_dev_dump(evdev, stdout);
+               return -1;
+       }
+
+       uint32_t deq_pkts;
+
+       deq_pkts = rte_event_dequeue_burst(evdev, t->port[wrk_enq], &ev, 1, 0);
+       if (deq_pkts < 1) {
+               printf("%d: Failed to deq\n", __LINE__);
+               return -1;
+       }
+
+       err = test_event_dev_stats_get(evdev, &stats);
+       if (err) {
+               printf("%d: failed to get stats\n", __LINE__);
+               return -1;
+       }
+
+       err = test_event_dev_stats_get(evdev, &stats);
+       if (ev.mbuf->seqn != MAGIC_SEQN) {
+               printf("%d: magic sequence number not dequeued\n", __LINE__);
+               return -1;
+       }
+
+       rte_pktmbuf_free(ev.mbuf);
+       err = rte_event_enqueue_burst(evdev, t->port[wrk_enq], &release_ev, 1);
+       if (err < 0) {
+               printf("%d: Failed to enqueue\n", __LINE__);
+               return -1;
+       }
+       rte_event_schedule(evdev);
+
+       err = test_event_dev_stats_get(evdev, &stats);
+       if (stats.port_inflight[wrk_enq] != 0) {
+               printf("%d: port inflight not correct\n", __LINE__);
+               return -1;
+       }
+
+       cleanup(t);
+       return 0;
+}
+
+static int
+inflight_counts(struct test *t)
+{
+       struct rte_event ev;
+       struct test_event_dev_stats stats;
+       const int rx_enq = 0;
+       const int p1 = 1;
+       const int p2 = 2;
+       int err;
+       int i;
+
+       /* Create instance with 4 ports */
+       if (init(t, 2, 3) < 0 ||
+                       create_ports(t, 3) < 0 ||
+                       create_atomic_qids(t, 2) < 0) {
+               printf("%d: Error initializing device\n", __LINE__);
+               return -1;
+       }
+
+       /* CQ mapping to QID */
+       err = rte_event_port_link(evdev, t->port[p1], &t->qid[0], NULL, 1);
+       if (err != 1) {
+               printf("%d: error mapping lb qid\n", __LINE__);
+               cleanup(t);
+               return -1;
+       }
+       err = rte_event_port_link(evdev, t->port[p2], &t->qid[1], NULL, 1);
+       if (err != 1) {
+               printf("%d: error mapping lb qid\n", __LINE__);
+               cleanup(t);
+               return -1;
+       }
+
+       if (rte_event_dev_start(evdev) < 0) {
+               printf("%d: Error with start call\n", __LINE__);
+               return -1;
+       }
+
+       /************** FORWARD ****************/
+#define QID1_NUM 5
+       for (i = 0; i < QID1_NUM; i++) {
+               struct rte_mbuf *arp = rte_gen_arp(0, t->mbuf_pool);
+
+               if (!arp) {
+                       printf("%d: gen of pkt failed\n", __LINE__);
+                       goto err;
+               }
+
+               ev.queue_id =  t->qid[0];
+               ev.op = RTE_EVENT_OP_NEW;
+               ev.mbuf = arp;
+               err = rte_event_enqueue_burst(evdev, t->port[rx_enq], &ev, 1);
+               if (err != 1) {
+                       printf("%d: Failed to enqueue\n", __LINE__);
+                       goto err;
+               }
+       }
+#define QID2_NUM 3
+       for (i = 0; i < QID2_NUM; i++) {
+               struct rte_mbuf *arp = rte_gen_arp(0, t->mbuf_pool);
+
+               if (!arp) {
+                       printf("%d: gen of pkt failed\n", __LINE__);
+                       goto err;
+               }
+               ev.queue_id =  t->qid[1];
+               ev.op = RTE_EVENT_OP_NEW;
+               ev.mbuf = arp;
+               err = rte_event_enqueue_burst(evdev, t->port[rx_enq], &ev, 1);
+               if (err != 1) {
+                       printf("%d: Failed to enqueue\n", __LINE__);
+                       goto err;
+               }
+       }
+
+       /* schedule */
+       rte_event_schedule(evdev);
+
+       err = test_event_dev_stats_get(evdev, &stats);
+       if (err) {
+               printf("%d: failed to get stats\n", __LINE__);
+               goto err;
+       }
+
+       if (stats.rx_pkts != QID1_NUM + QID2_NUM ||
+                       stats.tx_pkts != QID1_NUM + QID2_NUM) {
+               printf("%d: Sched core didn't handle pkt as expected\n",
+                               __LINE__);
+               goto err;
+       }
+
+       if (stats.port_inflight[p1] != QID1_NUM) {
+               printf("%d: %s port 1 inflight not correct\n", __LINE__,
+                               __func__);
+               goto err;
+       }
+       if (stats.port_inflight[p2] != QID2_NUM) {
+               printf("%d: %s port 2 inflight not correct\n", __LINE__,
+                               __func__);
+               goto err;
+       }
+
+       /************** DEQUEUE INFLIGHT COUNT CHECKS  ****************/
+       /* port 1 */
+       struct rte_event events[QID1_NUM + QID2_NUM];
+       uint32_t deq_pkts = rte_event_dequeue_burst(evdev, t->port[p1], events,
+                       RTE_DIM(events), 0);
+
+       if (deq_pkts != QID1_NUM) {
+               printf("%d: Port 1: DEQUEUE inflight failed\n", __LINE__);
+               goto err;
+       }
+       err = test_event_dev_stats_get(evdev, &stats);
+       if (stats.port_inflight[p1] != QID1_NUM) {
+               printf("%d: port 1 inflight decrement after DEQ != 0\n",
+                               __LINE__);
+               goto err;
+       }
+       for (i = 0; i < QID1_NUM; i++) {
+               err = rte_event_enqueue_burst(evdev, t->port[p1], &release_ev,
+                               1);
+               if (err != 1) {
+                       printf("%d: %s rte enqueue of inf release failed\n",
+                               __LINE__, __func__);
+                       goto err;
+               }
+       }
+
+       /*
+        * As the scheduler core decrements inflights, it needs to run to
+        * process packets to act on the drop messages
+        */
+       rte_event_schedule(evdev);
+
+       err = test_event_dev_stats_get(evdev, &stats);
+       if (stats.port_inflight[p1] != 0) {
+               printf("%d: port 1 inflight NON NULL after DROP\n", __LINE__);
+               goto err;
+       }
+
+       /* port2 */
+       deq_pkts = rte_event_dequeue_burst(evdev, t->port[p2], events,
+                       RTE_DIM(events), 0);
+       if (deq_pkts != QID2_NUM) {
+               printf("%d: Port 2: DEQUEUE inflight failed\n", __LINE__);
+               goto err;
+       }
+       err = test_event_dev_stats_get(evdev, &stats);
+       if (stats.port_inflight[p2] != QID2_NUM) {
+               printf("%d: port 1 inflight decrement after DEQ != 0\n",
+                               __LINE__);
+               goto err;
+       }
+       for (i = 0; i < QID2_NUM; i++) {
+               err = rte_event_enqueue_burst(evdev, t->port[p2], &release_ev,
+                               1);
+               if (err != 1) {
+                       printf("%d: %s rte enqueue of inf release failed\n",
+                               __LINE__, __func__);
+                       goto err;
+               }
+       }
+
+       /*
+        * As the scheduler core decrements inflights, it needs to run to
+        * process packets to act on the drop messages
+        */
+       rte_event_schedule(evdev);
+
+       err = test_event_dev_stats_get(evdev, &stats);
+       if (stats.port_inflight[p2] != 0) {
+               printf("%d: port 2 inflight NON NULL after DROP\n", __LINE__);
+               goto err;
+       }
+       cleanup(t);
+       return 0;
+
+err:
+       rte_event_dev_dump(evdev, stdout);
+       cleanup(t);
+       return -1;
+}
+
+static int
+parallel_basic(struct test *t, int check_order)
+{
+       const uint8_t rx_port = 0;
+       const uint8_t w1_port = 1;
+       const uint8_t w3_port = 3;
+       const uint8_t tx_port = 4;
+       int err;
+       int i;
+       uint32_t deq_pkts, j;
+       struct rte_mbuf *mbufs[3];
+       struct rte_mbuf *mbufs_out[3];
+       const uint32_t MAGIC_SEQN = 1234;
+
+       /* Create instance with 4 ports */
+       if (init(t, 2, tx_port + 1) < 0 ||
+                       create_ports(t, tx_port + 1) < 0 ||
+                       (check_order ?  create_ordered_qids(t, 1) :
+                               create_unordered_qids(t, 1)) < 0 ||
+                       create_directed_qids(t, 1, &tx_port)) {
+               printf("%d: Error initializing device\n", __LINE__);
+               return -1;
+       }
+
+       /*
+        * CQ mapping to QID
+        * We need three ports, all mapped to the same ordered qid0. Then we'll
+        * take a packet out to each port, re-enqueue in reverse order,
+        * then make sure the reordering has taken place properly when we
+        * dequeue from the tx_port.
+        *
+        * Simplified test setup diagram:
+        *
+        * rx_port        w1_port
+        *        \     /         \
+        *         qid0 - w2_port - qid1
+        *              \         /     \
+        *                w3_port        tx_port
+        */
+       /* CQ mapping to QID for LB ports (directed mapped on create) */
+       for (i = w1_port; i <= w3_port; i++) {
+               err = rte_event_port_link(evdev, t->port[i], &t->qid[0], NULL,
+                               1);
+               if (err != 1) {
+                       printf("%d: error mapping lb qid\n", __LINE__);
+                       cleanup(t);
+                       return -1;
+               }
+       }
+
+       if (rte_event_dev_start(evdev) < 0) {
+               printf("%d: Error with start call\n", __LINE__);
+               return -1;
+       }
+
+       /* Enqueue 3 packets to the rx port */
+       for (i = 0; i < 3; i++) {
+               struct rte_event ev;
+               mbufs[i] = rte_gen_arp(0, t->mbuf_pool);
+               if (!mbufs[i]) {
+                       printf("%d: gen of pkt failed\n", __LINE__);
+                       return -1;
+               }
+
+               ev.queue_id = t->qid[0];
+               ev.op = RTE_EVENT_OP_NEW;
+               ev.mbuf = mbufs[i];
+               mbufs[i]->seqn = MAGIC_SEQN + i;
+
+               /* generate pkt and enqueue */
+               err = rte_event_enqueue_burst(evdev, t->port[rx_port], &ev, 1);
+               if (err != 1) {
+                       printf("%d: Failed to enqueue pkt %u, retval = %u\n",
+                                       __LINE__, i, err);
+                       return -1;
+               }
+       }
+
+       rte_event_schedule(evdev);
+
+       /* use extra slot to make logic in loops easier */
+       struct rte_event deq_ev[w3_port + 1];
+
+       /* Dequeue the 3 packets, one from each worker port */
+       for (i = w1_port; i <= w3_port; i++) {
+               deq_pkts = rte_event_dequeue_burst(evdev, t->port[i],
+                               &deq_ev[i], 1, 0);
+               if (deq_pkts != 1) {
+                       printf("%d: Failed to deq\n", __LINE__);
+                       rte_event_dev_dump(evdev, stdout);
+                       return -1;
+               }
+       }
+
+       /* Enqueue each packet in reverse order, flushing after each one */
+       for (i = w3_port; i >= w1_port; i--) {
+
+               deq_ev[i].op = RTE_EVENT_OP_FORWARD;
+               deq_ev[i].queue_id = t->qid[1];
+               err = rte_event_enqueue_burst(evdev, t->port[i], &deq_ev[i], 1);
+               if (err != 1) {
+                       printf("%d: Failed to enqueue\n", __LINE__);
+                       return -1;
+               }
+       }
+       rte_event_schedule(evdev);
+
+       /* dequeue from the tx ports, we should get 3 packets */
+       deq_pkts = rte_event_dequeue_burst(evdev, t->port[tx_port], deq_ev,
+                       3, 0);
+
+       /* Check to see if we've got all 3 packets */
+       if (deq_pkts != 3) {
+               printf("%d: expected 3 pkts at tx port got %d from port %d\n",
+                       __LINE__, deq_pkts, tx_port);
+               rte_event_dev_dump(evdev, stdout);
+               return 1;
+       }
+
+       /* Check to see if the sequence numbers are in expected order */
+       if (check_order) {
+               for (j = 0 ; j < deq_pkts ; j++) {
+                       if (deq_ev[j].mbuf->seqn != MAGIC_SEQN + j) {
+                               printf(
+                                       "%d: Incorrect sequence number(%d) from port %d\n",
+                                       __LINE__, mbufs_out[j]->seqn, tx_port);
+                               return -1;
+                       }
+               }
+       }
+
+       /* Destroy the instance */
+       cleanup(t);
+       return 0;
+}
+
+static int
+ordered_basic(struct test *t)
+{
+       return parallel_basic(t, 1);
+}
+
+static int
+unordered_basic(struct test *t)
+{
+       return parallel_basic(t, 0);
+}
+
 static struct rte_mempool *eventdev_func_mempool;
 
 static int
 test_sw_eventdev(void)
 {
        struct test *t = malloc(sizeof(struct test));
+       int ret;
+
+       /* manually initialize the op, older gcc's complain on static
+        * initialization of struct elements that are a bitfield.
+        */
+       release_ev.op = RTE_EVENT_OP_RELEASE;
 
        const char *eventdev_name = "event_sw0";
        evdev = rte_event_dev_get_dev_id(eventdev_name);
@@ -125,6 +1340,72 @@ test_sw_eventdev(void)
        }
        t->mbuf_pool = eventdev_func_mempool;
 
+       printf("*** Running Single Directed Packet test...\n");
+       ret = test_single_directed_packet(t);
+       if (ret != 0) {
+               printf("ERROR - Single Directed Packet test FAILED.\n");
+               return ret;
+       }
+       printf("*** Running Single Load Balanced Packet test...\n");
+       ret = single_packet(t);
+       if (ret != 0) {
+               printf("ERROR - Single Packet test FAILED.\n");
+               return ret;
+       }
+       printf("*** Running Unordered Basic test...\n");
+       ret = unordered_basic(t);
+       if (ret != 0) {
+               printf("ERROR -  Unordered Basic test FAILED.\n");
+               return ret;
+       }
+       printf("*** Running Ordered Basic test...\n");
+       ret = ordered_basic(t);
+       if (ret != 0) {
+               printf("ERROR -  Ordered Basic test FAILED.\n");
+               return ret;
+       }
+       printf("*** Running Burst Packets test...\n");
+       ret = burst_packets(t);
+       if (ret != 0) {
+               printf("ERROR - Burst Packets test FAILED.\n");
+               return ret;
+       }
+       printf("*** Running Invalid QID test...\n");
+       ret = invalid_qid(t);
+       if (ret != 0) {
+               printf("ERROR - Invalid QID test FAILED.\n");
+               return ret;
+       }
+       printf("*** Running Inflight Count test...\n");
+       ret = inflight_counts(t);
+       if (ret != 0) {
+               printf("ERROR - Inflight Count test FAILED.\n");
+               return ret;
+       }
+       printf("*** Running Abuse Inflights test...\n");
+       ret = abuse_inflights(t);
+       if (ret != 0) {
+               printf("ERROR - Abuse Inflights test FAILED.\n");
+               return ret;
+       }
+       printf("*** Running Ordered Reconfigure test...\n");
+       ret = ordered_reconfigure(t);
+       if (ret != 0) {
+               printf("ERROR - Ordered Reconfigure test FAILED.\n");
+               return ret;
+       }
+       printf("*** Running Port LB Single Reconfig test...\n");
+       ret = port_single_lb_reconfig(t);
+       if (ret != 0) {
+               printf("ERROR - Port LB Single Reconfig test FAILED.\n");
+               return ret;
+       }
+       printf("*** Running Port Reconfig Credits test...\n");
+       ret = port_reconfig_credits(t);
+       if (ret != 0) {
+               printf("ERROR - Port Reconfig Credits Reset test FAILED.\n");
+               return ret;
+       }
        /*
         * Free test instance, leaving mempool initialized, and a pointer to it
         * in static eventdev_func_mempool, as it is re-used on re-runs