event/dlb2: add xstats
authorTimothy McDaniel <timothy.mcdaniel@intel.com>
Sun, 1 Nov 2020 23:37:49 +0000 (17:37 -0600)
committerJerin Jacob <jerinj@marvell.com>
Mon, 2 Nov 2020 08:40:12 +0000 (09:40 +0100)
Add support for DLB2 xstats.  Perform initialization and add
standard xstats entry points.

Signed-off-by: Timothy McDaniel <timothy.mcdaniel@intel.com>
Reviewed-by: Gage Eads <gage.eads@intel.com>
Reviewed-by: Mike Ximing Chen <mike.ximing.chen@intel.com>
drivers/event/dlb2/dlb2.c
drivers/event/dlb2/dlb2_xstats.c [new file with mode: 0644]
drivers/event/dlb2/meson.build

index 16653db..925824e 100644 (file)
@@ -70,6 +70,21 @@ static struct rte_event_dev_info evdev_dlb2_default_info = {
 struct process_local_port_data
 dlb2_port[DLB2_MAX_NUM_PORTS][DLB2_NUM_PORT_TYPES];
 
+/*
+ * DUMMY - added so that xstats path will compile/link.
+ * Will be replaced by real version in a subsequent
+ * patch.
+ */
+uint32_t
+dlb2_get_queue_depth(struct dlb2_eventdev *dlb2,
+                    struct dlb2_eventdev_queue *queue)
+{
+       RTE_SET_USED(dlb2);
+       RTE_SET_USED(queue);
+
+       return 0;
+}
+
 /* override defaults with value(s) provided on command line */
 static void
 dlb2_init_queue_depth_thresholds(struct dlb2_eventdev *dlb2,
@@ -337,9 +352,16 @@ set_qid_depth_thresh(const char *key __rte_unused,
 static void
 dlb2_entry_points_init(struct rte_eventdev *dev)
 {
-       RTE_SET_USED(dev);
-
-       /* Eventdev PMD entry points */
+       /* Expose PMD's eventdev interface */
+       static struct rte_eventdev_ops dlb2_eventdev_entry_ops = {
+               .dump             = dlb2_eventdev_dump,
+               .xstats_get       = dlb2_eventdev_xstats_get,
+               .xstats_get_names = dlb2_eventdev_xstats_get_names,
+               .xstats_get_by_name = dlb2_eventdev_xstats_get_by_name,
+               .xstats_reset       = dlb2_eventdev_xstats_reset,
+       };
+
+       dev->dev_ops = &dlb2_eventdev_entry_ops;
 }
 
 int
@@ -391,6 +413,13 @@ dlb2_primary_eventdev_probe(struct rte_eventdev *dev,
                return err;
        }
 
+       /* Complete xtstats runtime initialization */
+       err = dlb2_xstats_init(dlb2);
+       if (err) {
+               DLB2_LOG_ERR("dlb2: failed to init xstats, err=%d\n", err);
+               return err;
+       }
+
        rte_spinlock_init(&dlb2->qm_instance.resource_lock);
 
        dlb2_iface_low_level_io_init();
diff --git a/drivers/event/dlb2/dlb2_xstats.c b/drivers/event/dlb2/dlb2_xstats.c
new file mode 100644 (file)
index 0000000..8c3c3cd
--- /dev/null
@@ -0,0 +1,1230 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2016-2020 Intel Corporation
+ */
+
+#include <inttypes.h>
+
+#include <rte_malloc.h>
+#include <rte_eventdev.h>
+
+#include "dlb2_priv.h"
+#include "dlb2_inline_fns.h"
+
+enum dlb2_xstats_type {
+       /* common to device and port */
+       rx_ok,                          /**< Receive an event */
+       rx_drop,                        /**< Error bit set in received QE */
+       rx_interrupt_wait,              /**< Wait on an interrupt */
+       rx_umonitor_umwait,             /**< Block using umwait */
+       tx_ok,                          /**< Transmit an event */
+       total_polls,                    /**< Call dequeue_burst */
+       zero_polls,                     /**< Call dequeue burst and return 0 */
+       tx_nospc_ldb_hw_credits,        /**< Insufficient LDB h/w credits */
+       tx_nospc_dir_hw_credits,        /**< Insufficient DIR h/w credits */
+       tx_nospc_inflight_max,          /**< Reach the new_event_threshold */
+       tx_nospc_new_event_limit,       /**< Insufficient s/w credits */
+       tx_nospc_inflight_credits,      /**< Port has too few s/w credits */
+       /* device specific */
+       nb_events_limit,
+       inflight_events,
+       ldb_pool_size,
+       dir_pool_size,
+       /* port specific */
+       tx_new,                         /**< Send an OP_NEW event */
+       tx_fwd,                         /**< Send an OP_FORWARD event */
+       tx_rel,                         /**< Send an OP_RELEASE event */
+       tx_implicit_rel,                /**< Issue an implicit event release */
+       tx_sched_ordered,               /**< Send a SCHED_TYPE_ORDERED event */
+       tx_sched_unordered,             /**< Send a SCHED_TYPE_PARALLEL event */
+       tx_sched_atomic,                /**< Send a SCHED_TYPE_ATOMIC event */
+       tx_sched_directed,              /**< Send a directed event */
+       tx_invalid,                     /**< Send an event with an invalid op */
+       outstanding_releases,           /**< # of releases a port owes */
+       max_outstanding_releases,       /**< max # of releases a port can owe */
+       rx_sched_ordered,               /**< Dequeue an ordered event */
+       rx_sched_unordered,             /**< Dequeue an unordered event */
+       rx_sched_atomic,                /**< Dequeue an atomic event */
+       rx_sched_directed,              /**< Dequeue an directed event */
+       rx_sched_invalid,               /**< Dequeue event sched type invalid */
+       /* common to port and queue */
+       is_configured,                  /**< Port is configured */
+       is_load_balanced,               /**< Port is LDB */
+       hw_id,                          /**< Hardware ID */
+       /* queue specific */
+       num_links,                      /**< Number of ports linked */
+       sched_type,                     /**< Queue sched type */
+       enq_ok,                         /**< # events enqueued to the queue */
+       current_depth,                  /**< Current queue depth */
+       depth_threshold,                /**< Programmed depth threshold */
+       depth_le50_threshold,
+       /**< Depth LE to 50% of the configured hardware threshold */
+       depth_gt50_le75_threshold,
+       /**< Depth GT 50%, but LE to 75% of the configured hardware threshold */
+       depth_gt75_le100_threshold,
+       /**< Depth GT 75%. but LE to the configured hardware threshold */
+       depth_gt100_threshold
+       /**< Depth GT 100% of the configured hw threshold */
+};
+
+typedef uint64_t (*dlb2_xstats_fn)(struct dlb2_eventdev *dlb2,
+               uint16_t obj_idx, /* port or queue id */
+               enum dlb2_xstats_type stat, int extra_arg);
+
+enum dlb2_xstats_fn_type {
+       DLB2_XSTATS_FN_DEV,
+       DLB2_XSTATS_FN_PORT,
+       DLB2_XSTATS_FN_QUEUE
+};
+
+struct dlb2_xstats_entry {
+       struct rte_event_dev_xstats_name name;
+       uint64_t reset_value; /* an offset to be taken away to emulate resets */
+       enum dlb2_xstats_fn_type fn_id;
+       enum dlb2_xstats_type stat;
+       enum rte_event_dev_xstats_mode mode;
+       int extra_arg;
+       uint16_t obj_idx;
+       uint8_t reset_allowed; /* when set, this value can be reset */
+};
+
+/* Some device stats are simply a summation of the corresponding port values */
+static uint64_t
+dlb2_device_traffic_stat_get(struct dlb2_eventdev *dlb2,
+                            int which_stat)
+{
+       int i;
+       uint64_t val = 0;
+
+       for (i = 0; i < DLB2_MAX_NUM_PORTS; i++) {
+               struct dlb2_eventdev_port *port = &dlb2->ev_ports[i];
+
+               if (!port->setup_done)
+                       continue;
+
+               switch (which_stat) {
+               case rx_ok:
+                       val += port->stats.traffic.rx_ok;
+                       break;
+               case rx_drop:
+                       val += port->stats.traffic.rx_drop;
+                       break;
+               case rx_interrupt_wait:
+                       val += port->stats.traffic.rx_interrupt_wait;
+                       break;
+               case rx_umonitor_umwait:
+                       val += port->stats.traffic.rx_umonitor_umwait;
+                       break;
+               case tx_ok:
+                       val += port->stats.traffic.tx_ok;
+                       break;
+               case total_polls:
+                       val += port->stats.traffic.total_polls;
+                       break;
+               case zero_polls:
+                       val += port->stats.traffic.zero_polls;
+                       break;
+               case tx_nospc_ldb_hw_credits:
+                       val += port->stats.traffic.tx_nospc_ldb_hw_credits;
+                       break;
+               case tx_nospc_dir_hw_credits:
+                       val += port->stats.traffic.tx_nospc_dir_hw_credits;
+                       break;
+               case tx_nospc_inflight_max:
+                       val += port->stats.traffic.tx_nospc_inflight_max;
+                       break;
+               case tx_nospc_new_event_limit:
+                       val += port->stats.traffic.tx_nospc_new_event_limit;
+                       break;
+               case tx_nospc_inflight_credits:
+                       val += port->stats.traffic.tx_nospc_inflight_credits;
+                       break;
+               default:
+                       return -1;
+               }
+       }
+       return val;
+}
+
+static uint64_t
+get_dev_stat(struct dlb2_eventdev *dlb2, uint16_t obj_idx __rte_unused,
+            enum dlb2_xstats_type type, int extra_arg __rte_unused)
+{
+       switch (type) {
+       case rx_ok:
+       case rx_drop:
+       case rx_interrupt_wait:
+       case rx_umonitor_umwait:
+       case tx_ok:
+       case total_polls:
+       case zero_polls:
+       case tx_nospc_ldb_hw_credits:
+       case tx_nospc_dir_hw_credits:
+       case tx_nospc_inflight_max:
+       case tx_nospc_new_event_limit:
+       case tx_nospc_inflight_credits:
+               return dlb2_device_traffic_stat_get(dlb2, type);
+       case nb_events_limit:
+               return dlb2->new_event_limit;
+       case inflight_events:
+               return __atomic_load_n(&dlb2->inflights, __ATOMIC_SEQ_CST);
+       case ldb_pool_size:
+               return dlb2->num_ldb_credits;
+       case dir_pool_size:
+               return dlb2->num_dir_credits;
+       default: return -1;
+       }
+}
+
+static uint64_t
+get_port_stat(struct dlb2_eventdev *dlb2, uint16_t obj_idx,
+             enum dlb2_xstats_type type, int extra_arg __rte_unused)
+{
+       struct dlb2_eventdev_port *ev_port = &dlb2->ev_ports[obj_idx];
+
+       switch (type) {
+       case rx_ok: return ev_port->stats.traffic.rx_ok;
+
+       case rx_drop: return ev_port->stats.traffic.rx_drop;
+
+       case rx_interrupt_wait: return ev_port->stats.traffic.rx_interrupt_wait;
+
+       case rx_umonitor_umwait:
+               return ev_port->stats.traffic.rx_umonitor_umwait;
+
+       case tx_ok: return ev_port->stats.traffic.tx_ok;
+
+       case total_polls: return ev_port->stats.traffic.total_polls;
+
+       case zero_polls: return ev_port->stats.traffic.zero_polls;
+
+       case tx_nospc_ldb_hw_credits:
+               return ev_port->stats.traffic.tx_nospc_ldb_hw_credits;
+
+       case tx_nospc_dir_hw_credits:
+               return ev_port->stats.traffic.tx_nospc_dir_hw_credits;
+
+       case tx_nospc_inflight_max:
+               return ev_port->stats.traffic.tx_nospc_inflight_max;
+
+       case tx_nospc_new_event_limit:
+               return ev_port->stats.traffic.tx_nospc_new_event_limit;
+
+       case tx_nospc_inflight_credits:
+               return ev_port->stats.traffic.tx_nospc_inflight_credits;
+
+       case is_configured: return ev_port->setup_done;
+
+       case is_load_balanced: return !ev_port->qm_port.is_directed;
+
+       case hw_id: return ev_port->qm_port.id;
+
+       case tx_new: return ev_port->stats.tx_op_cnt[RTE_EVENT_OP_NEW];
+
+       case tx_fwd: return ev_port->stats.tx_op_cnt[RTE_EVENT_OP_FORWARD];
+
+       case tx_rel: return ev_port->stats.tx_op_cnt[RTE_EVENT_OP_RELEASE];
+
+       case tx_implicit_rel: return ev_port->stats.tx_implicit_rel;
+
+       case tx_sched_ordered:
+               return ev_port->stats.tx_sched_cnt[DLB2_SCHED_ORDERED];
+
+       case tx_sched_unordered:
+               return ev_port->stats.tx_sched_cnt[DLB2_SCHED_UNORDERED];
+
+       case tx_sched_atomic:
+               return ev_port->stats.tx_sched_cnt[DLB2_SCHED_ATOMIC];
+
+       case tx_sched_directed:
+               return ev_port->stats.tx_sched_cnt[DLB2_SCHED_DIRECTED];
+
+       case tx_invalid: return ev_port->stats.tx_invalid;
+
+       case outstanding_releases: return ev_port->outstanding_releases;
+
+       case max_outstanding_releases:
+               return DLB2_NUM_HIST_LIST_ENTRIES_PER_LDB_PORT;
+
+       case rx_sched_ordered:
+               return ev_port->stats.rx_sched_cnt[DLB2_SCHED_ORDERED];
+
+       case rx_sched_unordered:
+               return ev_port->stats.rx_sched_cnt[DLB2_SCHED_UNORDERED];
+
+       case rx_sched_atomic:
+               return ev_port->stats.rx_sched_cnt[DLB2_SCHED_ATOMIC];
+
+       case rx_sched_directed:
+               return ev_port->stats.rx_sched_cnt[DLB2_SCHED_DIRECTED];
+
+       case rx_sched_invalid: return ev_port->stats.rx_sched_invalid;
+
+       default: return -1;
+       }
+}
+
+static uint64_t
+dlb2_get_threshold_stat(struct dlb2_eventdev *dlb2, int qid, int stat)
+{
+       int port = 0;
+       uint64_t tally = 0;
+
+       for (port = 0; port < DLB2_MAX_NUM_PORTS; port++)
+               tally += dlb2->ev_ports[port].stats.queue[qid].qid_depth[stat];
+
+       return tally;
+}
+
+static uint64_t
+dlb2_get_enq_ok_stat(struct dlb2_eventdev *dlb2, int qid)
+{
+       int port = 0;
+       uint64_t enq_ok_tally = 0;
+
+       for (port = 0; port < DLB2_MAX_NUM_PORTS; port++)
+               enq_ok_tally += dlb2->ev_ports[port].stats.queue[qid].enq_ok;
+
+       return enq_ok_tally;
+}
+
+static uint64_t
+get_queue_stat(struct dlb2_eventdev *dlb2, uint16_t obj_idx,
+              enum dlb2_xstats_type type, int extra_arg __rte_unused)
+{
+       struct dlb2_eventdev_queue *ev_queue =
+               &dlb2->ev_queues[obj_idx];
+
+       switch (type) {
+       case is_configured: return ev_queue->setup_done;
+
+       case is_load_balanced: return !ev_queue->qm_queue.is_directed;
+
+       case hw_id: return ev_queue->qm_queue.id;
+
+       case num_links: return ev_queue->num_links;
+
+       case sched_type: return ev_queue->qm_queue.sched_type;
+
+       case enq_ok: return dlb2_get_enq_ok_stat(dlb2, obj_idx);
+
+       case current_depth: return dlb2_get_queue_depth(dlb2, ev_queue);
+
+       case depth_threshold: return ev_queue->depth_threshold;
+
+       case depth_le50_threshold:
+               return dlb2_get_threshold_stat(dlb2, ev_queue->id,
+                                              DLB2_QID_DEPTH_LE50);
+
+       case depth_gt50_le75_threshold:
+               return dlb2_get_threshold_stat(dlb2, ev_queue->id,
+                                              DLB2_QID_DEPTH_GT50_LE75);
+
+       case depth_gt75_le100_threshold:
+               return dlb2_get_threshold_stat(dlb2, ev_queue->id,
+                                              DLB2_QID_DEPTH_GT75_LE100);
+
+       case depth_gt100_threshold:
+               return dlb2_get_threshold_stat(dlb2, ev_queue->id,
+                                              DLB2_QID_DEPTH_GT100);
+
+       default: return -1;
+       }
+}
+
+int
+dlb2_xstats_init(struct dlb2_eventdev *dlb2)
+{
+       /*
+        * define the stats names and types. Used to build up the device
+        * xstats array
+        * There are multiple set of stats:
+        *   - device-level,
+        *   - per-port,
+        *   - per-qid,
+        *
+        * For each of these sets, we have three parallel arrays, one for the
+        * names, the other for the stat type parameter to be passed in the fn
+        * call to get that stat. The third array allows resetting or not.
+        * All these arrays must be kept in sync
+        */
+       static const char * const dev_stats[] = {
+               "rx_ok",
+               "rx_drop",
+               "rx_interrupt_wait",
+               "rx_umonitor_umwait",
+               "tx_ok",
+               "total_polls",
+               "zero_polls",
+               "tx_nospc_ldb_hw_credits",
+               "tx_nospc_dir_hw_credits",
+               "tx_nospc_inflight_max",
+               "tx_nospc_new_event_limit",
+               "tx_nospc_inflight_credits",
+               "nb_events_limit",
+               "inflight_events",
+               "ldb_pool_size",
+               "dir_pool_size",
+       };
+       static const enum dlb2_xstats_type dev_types[] = {
+               rx_ok,
+               rx_drop,
+               rx_interrupt_wait,
+               rx_umonitor_umwait,
+               tx_ok,
+               total_polls,
+               zero_polls,
+               tx_nospc_ldb_hw_credits,
+               tx_nospc_dir_hw_credits,
+               tx_nospc_inflight_max,
+               tx_nospc_new_event_limit,
+               tx_nospc_inflight_credits,
+               nb_events_limit,
+               inflight_events,
+               ldb_pool_size,
+               dir_pool_size,
+       };
+       /* Note: generated device stats are not allowed to be reset. */
+       static const uint8_t dev_reset_allowed[] = {
+               0, /* rx_ok */
+               0, /* rx_drop */
+               0, /* rx_interrupt_wait */
+               0, /* rx_umonitor_umwait */
+               0, /* tx_ok */
+               0, /* total_polls */
+               0, /* zero_polls */
+               0, /* tx_nospc_ldb_hw_credits */
+               0, /* tx_nospc_dir_hw_credits */
+               0, /* tx_nospc_inflight_max */
+               0, /* tx_nospc_new_event_limit */
+               0, /* tx_nospc_inflight_credits */
+               0, /* nb_events_limit */
+               0, /* inflight_events */
+               0, /* ldb_pool_size */
+               0, /* dir_pool_size */
+       };
+       static const char * const port_stats[] = {
+               "is_configured",
+               "is_load_balanced",
+               "hw_id",
+               "rx_ok",
+               "rx_drop",
+               "rx_interrupt_wait",
+               "rx_umonitor_umwait",
+               "tx_ok",
+               "total_polls",
+               "zero_polls",
+               "tx_nospc_ldb_hw_credits",
+               "tx_nospc_dir_hw_credits",
+               "tx_nospc_inflight_max",
+               "tx_nospc_new_event_limit",
+               "tx_nospc_inflight_credits",
+               "tx_new",
+               "tx_fwd",
+               "tx_rel",
+               "tx_implicit_rel",
+               "tx_sched_ordered",
+               "tx_sched_unordered",
+               "tx_sched_atomic",
+               "tx_sched_directed",
+               "tx_invalid",
+               "outstanding_releases",
+               "max_outstanding_releases",
+               "rx_sched_ordered",
+               "rx_sched_unordered",
+               "rx_sched_atomic",
+               "rx_sched_directed",
+               "rx_sched_invalid"
+       };
+       static const enum dlb2_xstats_type port_types[] = {
+               is_configured,
+               is_load_balanced,
+               hw_id,
+               rx_ok,
+               rx_drop,
+               rx_interrupt_wait,
+               rx_umonitor_umwait,
+               tx_ok,
+               total_polls,
+               zero_polls,
+               tx_nospc_ldb_hw_credits,
+               tx_nospc_dir_hw_credits,
+               tx_nospc_inflight_max,
+               tx_nospc_new_event_limit,
+               tx_nospc_inflight_credits,
+               tx_new,
+               tx_fwd,
+               tx_rel,
+               tx_implicit_rel,
+               tx_sched_ordered,
+               tx_sched_unordered,
+               tx_sched_atomic,
+               tx_sched_directed,
+               tx_invalid,
+               outstanding_releases,
+               max_outstanding_releases,
+               rx_sched_ordered,
+               rx_sched_unordered,
+               rx_sched_atomic,
+               rx_sched_directed,
+               rx_sched_invalid
+       };
+       static const uint8_t port_reset_allowed[] = {
+               0, /* is_configured */
+               0, /* is_load_balanced */
+               0, /* hw_id */
+               1, /* rx_ok */
+               1, /* rx_drop */
+               1, /* rx_interrupt_wait */
+               1, /* rx_umonitor_umwait */
+               1, /* tx_ok */
+               1, /* total_polls */
+               1, /* zero_polls */
+               1, /* tx_nospc_ldb_hw_credits */
+               1, /* tx_nospc_dir_hw_credits */
+               1, /* tx_nospc_inflight_max */
+               1, /* tx_nospc_new_event_limit */
+               1, /* tx_nospc_inflight_credits */
+               1, /* tx_new */
+               1, /* tx_fwd */
+               1, /* tx_rel */
+               1, /* tx_implicit_rel */
+               1, /* tx_sched_ordered */
+               1, /* tx_sched_unordered */
+               1, /* tx_sched_atomic */
+               1, /* tx_sched_directed */
+               1, /* tx_invalid */
+               0, /* outstanding_releases */
+               0, /* max_outstanding_releases */
+               1, /* rx_sched_ordered */
+               1, /* rx_sched_unordered */
+               1, /* rx_sched_atomic */
+               1, /* rx_sched_directed */
+               1  /* rx_sched_invalid */
+       };
+
+       /* QID specific stats */
+       static const char * const qid_stats[] = {
+               "is_configured",
+               "is_load_balanced",
+               "hw_id",
+               "num_links",
+               "sched_type",
+               "enq_ok",
+               "current_depth",
+               "depth_threshold",
+               "depth_le50_threshold",
+               "depth_gt50_le75_threshold",
+               "depth_gt75_le100_threshold",
+               "depth_gt100_threshold",
+       };
+       static const enum dlb2_xstats_type qid_types[] = {
+               is_configured,
+               is_load_balanced,
+               hw_id,
+               num_links,
+               sched_type,
+               enq_ok,
+               current_depth,
+               depth_threshold,
+               depth_le50_threshold,
+               depth_gt50_le75_threshold,
+               depth_gt75_le100_threshold,
+               depth_gt100_threshold,
+       };
+       static const uint8_t qid_reset_allowed[] = {
+               0, /* is_configured */
+               0, /* is_load_balanced */
+               0, /* hw_id */
+               0, /* num_links */
+               0, /* sched_type */
+               1, /* enq_ok */
+               0, /* current_depth */
+               0, /* depth_threshold */
+               1, /* depth_le50_threshold */
+               1, /* depth_gt50_le75_threshold */
+               1, /* depth_gt75_le100_threshold */
+               1, /* depth_gt100_threshold */
+       };
+
+       /* ---- end of stat definitions ---- */
+
+       /* check sizes, since a missed comma can lead to strings being
+        * joined by the compiler.
+        */
+       RTE_BUILD_BUG_ON(RTE_DIM(dev_stats) != RTE_DIM(dev_types));
+       RTE_BUILD_BUG_ON(RTE_DIM(port_stats) != RTE_DIM(port_types));
+       RTE_BUILD_BUG_ON(RTE_DIM(qid_stats) != RTE_DIM(qid_types));
+
+       RTE_BUILD_BUG_ON(RTE_DIM(dev_stats) != RTE_DIM(dev_reset_allowed));
+       RTE_BUILD_BUG_ON(RTE_DIM(port_stats) != RTE_DIM(port_reset_allowed));
+       RTE_BUILD_BUG_ON(RTE_DIM(qid_stats) != RTE_DIM(qid_reset_allowed));
+
+       /* other vars */
+       const unsigned int count = RTE_DIM(dev_stats) +
+                       DLB2_MAX_NUM_PORTS * RTE_DIM(port_stats) +
+                       DLB2_MAX_NUM_QUEUES * RTE_DIM(qid_stats);
+       unsigned int i, port, qid, stat_id = 0;
+
+       dlb2->xstats = rte_zmalloc_socket(NULL,
+                       sizeof(dlb2->xstats[0]) * count, 0,
+                       dlb2->qm_instance.info.socket_id);
+       if (dlb2->xstats == NULL)
+               return -ENOMEM;
+
+#define sname dlb2->xstats[stat_id].name.name
+       for (i = 0; i < RTE_DIM(dev_stats); i++, stat_id++) {
+               dlb2->xstats[stat_id] = (struct dlb2_xstats_entry) {
+                       .fn_id = DLB2_XSTATS_FN_DEV,
+                       .stat = dev_types[i],
+                       .mode = RTE_EVENT_DEV_XSTATS_DEVICE,
+                       .reset_allowed = dev_reset_allowed[i],
+               };
+               snprintf(sname, sizeof(sname), "dev_%s", dev_stats[i]);
+       }
+       dlb2->xstats_count_mode_dev = stat_id;
+
+       for (port = 0; port < DLB2_MAX_NUM_PORTS; port++) {
+               dlb2->xstats_offset_for_port[port] = stat_id;
+
+               uint32_t count_offset = stat_id;
+
+               for (i = 0; i < RTE_DIM(port_stats); i++, stat_id++) {
+                       dlb2->xstats[stat_id] = (struct dlb2_xstats_entry){
+                               .fn_id = DLB2_XSTATS_FN_PORT,
+                               .obj_idx = port,
+                               .stat = port_types[i],
+                               .mode = RTE_EVENT_DEV_XSTATS_PORT,
+                               .reset_allowed = port_reset_allowed[i],
+                       };
+                       snprintf(sname, sizeof(sname), "port_%u_%s",
+                                port, port_stats[i]);
+               }
+
+               dlb2->xstats_count_per_port[port] = stat_id - count_offset;
+       }
+
+       dlb2->xstats_count_mode_port = stat_id - dlb2->xstats_count_mode_dev;
+
+       for (qid = 0; qid < DLB2_MAX_NUM_QUEUES; qid++) {
+               uint32_t count_offset = stat_id;
+
+               dlb2->xstats_offset_for_qid[qid] = stat_id;
+
+               for (i = 0; i < RTE_DIM(qid_stats); i++, stat_id++) {
+                       dlb2->xstats[stat_id] = (struct dlb2_xstats_entry){
+                               .fn_id = DLB2_XSTATS_FN_QUEUE,
+                               .obj_idx = qid,
+                               .stat = qid_types[i],
+                               .mode = RTE_EVENT_DEV_XSTATS_QUEUE,
+                               .reset_allowed = qid_reset_allowed[i],
+                       };
+                       snprintf(sname, sizeof(sname), "qid_%u_%s",
+                                qid, qid_stats[i]);
+               }
+
+               dlb2->xstats_count_per_qid[qid] = stat_id - count_offset;
+       }
+
+       dlb2->xstats_count_mode_queue = stat_id -
+               (dlb2->xstats_count_mode_dev + dlb2->xstats_count_mode_port);
+#undef sname
+
+       dlb2->xstats_count = stat_id;
+
+       return 0;
+}
+
+void
+dlb2_xstats_uninit(struct dlb2_eventdev *dlb2)
+{
+       rte_free(dlb2->xstats);
+       dlb2->xstats_count = 0;
+}
+
+int
+dlb2_eventdev_xstats_get_names(const struct rte_eventdev *dev,
+               enum rte_event_dev_xstats_mode mode, uint8_t queue_port_id,
+               struct rte_event_dev_xstats_name *xstats_names,
+               unsigned int *ids, unsigned int size)
+{
+       const struct dlb2_eventdev *dlb2 = dlb2_pmd_priv(dev);
+       unsigned int i;
+       unsigned int xidx = 0;
+       uint32_t xstats_mode_count = 0;
+       uint32_t start_offset = 0;
+
+       switch (mode) {
+       case RTE_EVENT_DEV_XSTATS_DEVICE:
+               xstats_mode_count = dlb2->xstats_count_mode_dev;
+               break;
+       case RTE_EVENT_DEV_XSTATS_PORT:
+               if (queue_port_id >= DLB2_MAX_NUM_PORTS)
+                       break;
+               xstats_mode_count = dlb2->xstats_count_per_port[queue_port_id];
+               start_offset = dlb2->xstats_offset_for_port[queue_port_id];
+               break;
+       case RTE_EVENT_DEV_XSTATS_QUEUE:
+#if (DLB2_MAX_NUM_QUEUES <= 255) /* max 8 bit value */
+               if (queue_port_id >= DLB2_MAX_NUM_QUEUES)
+                       break;
+#endif
+               xstats_mode_count = dlb2->xstats_count_per_qid[queue_port_id];
+               start_offset = dlb2->xstats_offset_for_qid[queue_port_id];
+               break;
+       default:
+               return -EINVAL;
+       };
+
+       if (xstats_mode_count > size || ids == NULL || xstats_names == NULL)
+               return xstats_mode_count;
+
+       for (i = 0; i < dlb2->xstats_count && xidx < size; i++) {
+               if (dlb2->xstats[i].mode != mode)
+                       continue;
+
+               if (mode != RTE_EVENT_DEV_XSTATS_DEVICE &&
+                   queue_port_id != dlb2->xstats[i].obj_idx)
+                       continue;
+
+               xstats_names[xidx] = dlb2->xstats[i].name;
+               if (ids)
+                       ids[xidx] = start_offset + xidx;
+               xidx++;
+       }
+       return xidx;
+}
+
+static int
+dlb2_xstats_update(struct dlb2_eventdev *dlb2,
+               enum rte_event_dev_xstats_mode mode,
+               uint8_t queue_port_id, const unsigned int ids[],
+               uint64_t values[], unsigned int n, const uint32_t reset)
+{
+       unsigned int i;
+       unsigned int xidx = 0;
+       uint32_t xstats_mode_count = 0;
+
+       switch (mode) {
+       case RTE_EVENT_DEV_XSTATS_DEVICE:
+               xstats_mode_count = dlb2->xstats_count_mode_dev;
+               break;
+       case RTE_EVENT_DEV_XSTATS_PORT:
+               if (queue_port_id >= DLB2_MAX_NUM_PORTS)
+                       goto invalid_value;
+               xstats_mode_count = dlb2->xstats_count_per_port[queue_port_id];
+               break;
+       case RTE_EVENT_DEV_XSTATS_QUEUE:
+#if (DLB2_MAX_NUM_QUEUES <= 255) /* max 8 bit value */
+               if (queue_port_id >= DLB2_MAX_NUM_QUEUES)
+                       goto invalid_value;
+#endif
+               xstats_mode_count = dlb2->xstats_count_per_qid[queue_port_id];
+               break;
+       default:
+               goto invalid_value;
+       };
+
+       for (i = 0; i < n && xidx < xstats_mode_count; i++) {
+               struct dlb2_xstats_entry *xs = &dlb2->xstats[ids[i]];
+               dlb2_xstats_fn fn;
+
+               if (ids[i] > dlb2->xstats_count || xs->mode != mode)
+                       continue;
+
+               if (mode != RTE_EVENT_DEV_XSTATS_DEVICE &&
+                   queue_port_id != xs->obj_idx)
+                       continue;
+
+               switch (xs->fn_id) {
+               case DLB2_XSTATS_FN_DEV:
+                       fn = get_dev_stat;
+                       break;
+               case DLB2_XSTATS_FN_PORT:
+                       fn = get_port_stat;
+                       break;
+               case DLB2_XSTATS_FN_QUEUE:
+                       fn = get_queue_stat;
+                       break;
+               default:
+                       DLB2_LOG_ERR("Unexpected xstat fn_id %d\n", xs->fn_id);
+                       goto invalid_value;
+               }
+
+               uint64_t val = fn(dlb2, xs->obj_idx, xs->stat,
+                                 xs->extra_arg) - xs->reset_value;
+
+               if (values)
+                       values[xidx] = val;
+
+               if (xs->reset_allowed && reset)
+                       xs->reset_value += val;
+
+               xidx++;
+       }
+
+       return xidx;
+
+invalid_value:
+       return -EINVAL;
+}
+
+int
+dlb2_eventdev_xstats_get(const struct rte_eventdev *dev,
+               enum rte_event_dev_xstats_mode mode, uint8_t queue_port_id,
+               const unsigned int ids[], uint64_t values[], unsigned int n)
+{
+       struct dlb2_eventdev *dlb2 = dlb2_pmd_priv(dev);
+       const uint32_t reset = 0;
+
+       return dlb2_xstats_update(dlb2, mode, queue_port_id, ids, values, n,
+                                 reset);
+}
+
+uint64_t
+dlb2_eventdev_xstats_get_by_name(const struct rte_eventdev *dev,
+                                const char *name, unsigned int *id)
+{
+       struct dlb2_eventdev *dlb2 = dlb2_pmd_priv(dev);
+       unsigned int i;
+       dlb2_xstats_fn fn;
+
+       for (i = 0; i < dlb2->xstats_count; i++) {
+               struct dlb2_xstats_entry *xs = &dlb2->xstats[i];
+
+               if (strncmp(xs->name.name, name,
+                           RTE_EVENT_DEV_XSTATS_NAME_SIZE) == 0){
+                       if (id != NULL)
+                               *id = i;
+
+                       switch (xs->fn_id) {
+                       case DLB2_XSTATS_FN_DEV:
+                               fn = get_dev_stat;
+                               break;
+                       case DLB2_XSTATS_FN_PORT:
+                               fn = get_port_stat;
+                               break;
+                       case DLB2_XSTATS_FN_QUEUE:
+                               fn = get_queue_stat;
+                               break;
+                       default:
+                               DLB2_LOG_ERR("Unexpected xstat fn_id %d\n",
+                                         xs->fn_id);
+                               return (uint64_t)-1;
+                       }
+
+                       return fn(dlb2, xs->obj_idx, xs->stat,
+                                 xs->extra_arg) - xs->reset_value;
+               }
+       }
+       if (id != NULL)
+               *id = (uint32_t)-1;
+       return (uint64_t)-1;
+}
+
+static void
+dlb2_xstats_reset_range(struct dlb2_eventdev *dlb2, uint32_t start,
+                       uint32_t num)
+{
+       uint32_t i;
+       dlb2_xstats_fn fn;
+
+       for (i = start; i < start + num; i++) {
+               struct dlb2_xstats_entry *xs = &dlb2->xstats[i];
+
+               if (!xs->reset_allowed)
+                       continue;
+
+               switch (xs->fn_id) {
+               case DLB2_XSTATS_FN_DEV:
+                       fn = get_dev_stat;
+                       break;
+               case DLB2_XSTATS_FN_PORT:
+                       fn = get_port_stat;
+                       break;
+               case DLB2_XSTATS_FN_QUEUE:
+                       fn = get_queue_stat;
+                       break;
+               default:
+                       DLB2_LOG_ERR("Unexpected xstat fn_id %d\n", xs->fn_id);
+                       return;
+               }
+
+               uint64_t val = fn(dlb2, xs->obj_idx, xs->stat, xs->extra_arg);
+               xs->reset_value = val;
+       }
+}
+
+static int
+dlb2_xstats_reset_queue(struct dlb2_eventdev *dlb2, uint8_t queue_id,
+                       const uint32_t ids[], uint32_t nb_ids)
+{
+       const uint32_t reset = 1;
+
+       if (ids) {
+               uint32_t nb_reset = dlb2_xstats_update(dlb2,
+                                       RTE_EVENT_DEV_XSTATS_QUEUE,
+                                       queue_id, ids, NULL, nb_ids,
+                                       reset);
+               return nb_reset == nb_ids ? 0 : -EINVAL;
+       }
+
+       if (ids == NULL)
+               dlb2_xstats_reset_range(dlb2,
+                       dlb2->xstats_offset_for_qid[queue_id],
+                       dlb2->xstats_count_per_qid[queue_id]);
+
+       return 0;
+}
+
+static int
+dlb2_xstats_reset_port(struct dlb2_eventdev *dlb2, uint8_t port_id,
+                      const uint32_t ids[], uint32_t nb_ids)
+{
+       const uint32_t reset = 1;
+       int offset = dlb2->xstats_offset_for_port[port_id];
+       int nb_stat = dlb2->xstats_count_per_port[port_id];
+
+       if (ids) {
+               uint32_t nb_reset = dlb2_xstats_update(dlb2,
+                                       RTE_EVENT_DEV_XSTATS_PORT, port_id,
+                                       ids, NULL, nb_ids,
+                                       reset);
+               return nb_reset == nb_ids ? 0 : -EINVAL;
+       }
+
+       dlb2_xstats_reset_range(dlb2, offset, nb_stat);
+       return 0;
+}
+
+static int
+dlb2_xstats_reset_dev(struct dlb2_eventdev *dlb2, const uint32_t ids[],
+                     uint32_t nb_ids)
+{
+       uint32_t i;
+
+       if (ids) {
+               for (i = 0; i < nb_ids; i++) {
+                       uint32_t id = ids[i];
+
+                       if (id >= dlb2->xstats_count_mode_dev)
+                               return -EINVAL;
+                       dlb2_xstats_reset_range(dlb2, id, 1);
+               }
+       } else {
+               for (i = 0; i < dlb2->xstats_count_mode_dev; i++)
+                       dlb2_xstats_reset_range(dlb2, i, 1);
+       }
+
+       return 0;
+}
+
+int
+dlb2_eventdev_xstats_reset(struct rte_eventdev *dev,
+                          enum rte_event_dev_xstats_mode mode,
+                          int16_t queue_port_id,
+                          const uint32_t ids[],
+                          uint32_t nb_ids)
+{
+       struct dlb2_eventdev *dlb2 = dlb2_pmd_priv(dev);
+       uint32_t i;
+
+       /* handle -1 for queue_port_id here, looping over all ports/queues */
+       switch (mode) {
+       case RTE_EVENT_DEV_XSTATS_DEVICE:
+               if (dlb2_xstats_reset_dev(dlb2, ids, nb_ids))
+                       return -EINVAL;
+               break;
+       case RTE_EVENT_DEV_XSTATS_PORT:
+               if (queue_port_id == -1) {
+                       for (i = 0; i < DLB2_MAX_NUM_PORTS; i++) {
+                               if (dlb2_xstats_reset_port(dlb2, i,
+                                                          ids, nb_ids))
+                                       return -EINVAL;
+                       }
+               } else if (queue_port_id < DLB2_MAX_NUM_PORTS) {
+                       if (dlb2_xstats_reset_port(dlb2, queue_port_id,
+                                                  ids, nb_ids))
+                               return -EINVAL;
+               }
+               break;
+       case RTE_EVENT_DEV_XSTATS_QUEUE:
+               if (queue_port_id == -1) {
+                       for (i = 0; i < DLB2_MAX_NUM_QUEUES; i++) {
+                               if (dlb2_xstats_reset_queue(dlb2, i,
+                                                           ids, nb_ids))
+                                       return -EINVAL;
+                       }
+               } else if (queue_port_id < DLB2_MAX_NUM_QUEUES) {
+                       if (dlb2_xstats_reset_queue(dlb2, queue_port_id,
+                                                   ids, nb_ids))
+                               return -EINVAL;
+               }
+               break;
+       };
+
+       return 0;
+}
+
+void
+dlb2_eventdev_dump(struct rte_eventdev *dev, FILE *f)
+{
+       struct dlb2_eventdev *dlb2;
+       struct dlb2_hw_dev *handle;
+       int i;
+
+       dlb2 = dlb2_pmd_priv(dev);
+
+       if (dlb2 == NULL) {
+               fprintf(f, "DLB2 Event device cannot be dumped!\n");
+               return;
+       }
+
+       if (!dlb2->configured)
+               fprintf(f, "DLB2 Event device is not configured\n");
+
+       handle = &dlb2->qm_instance;
+
+       fprintf(f, "================\n");
+       fprintf(f, "DLB2 Device Dump\n");
+       fprintf(f, "================\n");
+
+       fprintf(f, "Processor supports umonitor/umwait instructions = %s\n",
+               dlb2->umwait_allowed ? "yes" : "no");
+
+       /* Generic top level device information */
+
+       fprintf(f, "device is configured and run state =");
+       if (dlb2->run_state == DLB2_RUN_STATE_STOPPED)
+               fprintf(f, "STOPPED\n");
+       else if (dlb2->run_state == DLB2_RUN_STATE_STOPPING)
+               fprintf(f, "STOPPING\n");
+       else if (dlb2->run_state == DLB2_RUN_STATE_STARTING)
+               fprintf(f, "STARTING\n");
+       else if (dlb2->run_state == DLB2_RUN_STATE_STARTED)
+               fprintf(f, "STARTED\n");
+       else
+               fprintf(f, "UNEXPECTED\n");
+
+       fprintf(f, "domain ID=%u, socket_id=%u, evdev=%p\n",
+               handle->domain_id, handle->info.socket_id, dlb2->event_dev);
+
+       fprintf(f, "num dir ports=%u, num dir queues=%u\n",
+               dlb2->num_dir_ports, dlb2->num_dir_queues);
+
+       fprintf(f, "num ldb ports=%u, num ldb queues=%u\n",
+               dlb2->num_ldb_ports, dlb2->num_ldb_queues);
+
+       fprintf(f, "num atomic inflights=%u, hist list entries=%u\n",
+               handle->cfg.resources.num_atomic_inflights,
+               handle->cfg.resources.num_hist_list_entries);
+
+       fprintf(f, "results from most recent hw resource query:\n");
+
+       fprintf(f, "\tnum_sched_domains = %u\n",
+               dlb2->hw_rsrc_query_results.num_sched_domains);
+
+       fprintf(f, "\tnum_ldb_queues = %u\n",
+               dlb2->hw_rsrc_query_results.num_ldb_queues);
+
+       fprintf(f, "\tnum_ldb_ports = %u\n",
+               dlb2->hw_rsrc_query_results.num_ldb_ports);
+
+       fprintf(f, "\tnum_dir_ports = %u\n",
+               dlb2->hw_rsrc_query_results.num_dir_ports);
+
+       fprintf(f, "\tnum_atomic_inflights = %u\n",
+               dlb2->hw_rsrc_query_results.num_atomic_inflights);
+
+       fprintf(f, "\tnum_hist_list_entries = %u\n",
+               dlb2->hw_rsrc_query_results.num_hist_list_entries);
+
+       fprintf(f, "\tmax_contiguous_hist_list_entries = %u\n",
+               dlb2->hw_rsrc_query_results.max_contiguous_hist_list_entries);
+
+       fprintf(f, "\tnum_ldb_credits = %u\n",
+               dlb2->hw_rsrc_query_results.num_ldb_credits);
+
+       fprintf(f, "\tnum_dir_credits = %u\n",
+               dlb2->hw_rsrc_query_results.num_dir_credits);
+
+       /* Port level information */
+
+       for (i = 0; i < dlb2->num_ports; i++) {
+               struct dlb2_eventdev_port *p = &dlb2->ev_ports[i];
+               int j;
+
+               if (!p->enq_configured)
+                       fprintf(f, "Port_%d is not configured\n", i);
+
+               fprintf(f, "Port_%d\n", i);
+               fprintf(f, "=======\n");
+
+               fprintf(f, "\tevport_%u is configured, setup done=%d\n",
+                       p->id, p->setup_done);
+
+               fprintf(f, "\tconfig state=%d, port state=%d\n",
+                       p->qm_port.config_state, p->qm_port.state);
+
+               fprintf(f, "\tport is %s\n",
+                       p->qm_port.is_directed ? "directed" : "load balanced");
+
+               fprintf(f, "\toutstanding releases=%u\n",
+                       p->outstanding_releases);
+
+               fprintf(f, "\tinflight max=%u, inflight credits=%u\n",
+                       p->inflight_max, p->inflight_credits);
+
+               fprintf(f, "\tcredit update quanta=%u, implicit release =%u\n",
+                       p->credit_update_quanta, p->implicit_release);
+
+               fprintf(f, "\tnum_links=%d, queues -> ", p->num_links);
+
+               for (j = 0; j < DLB2_MAX_NUM_QIDS_PER_LDB_CQ; j++) {
+                       if (p->link[j].valid)
+                               fprintf(f, "id=%u prio=%u ",
+                                       p->link[j].queue_id,
+                                       p->link[j].priority);
+               }
+               fprintf(f, "\n");
+
+               fprintf(f, "\thardware port id=%u\n", p->qm_port.id);
+
+               fprintf(f, "\tcached_ldb_credits=%u\n",
+                       p->qm_port.cached_ldb_credits);
+
+               fprintf(f, "\tldb_credits = %u\n",
+                       p->qm_port.ldb_credits);
+
+               fprintf(f, "\tcached_dir_credits = %u\n",
+                       p->qm_port.cached_dir_credits);
+
+               fprintf(f, "\tdir_credits = %u\n",
+                       p->qm_port.dir_credits);
+
+               fprintf(f, "\tgenbit=%d, cq_idx=%d, cq_depth=%d\n",
+                       p->qm_port.gen_bit,
+                       p->qm_port.cq_idx,
+                       p->qm_port.cq_depth);
+
+               fprintf(f, "\tinterrupt armed=%d\n",
+                       p->qm_port.int_armed);
+
+               fprintf(f, "\tPort statistics\n");
+
+               fprintf(f, "\t\trx_ok %" PRIu64 "\n",
+                       p->stats.traffic.rx_ok);
+
+               fprintf(f, "\t\trx_drop %" PRIu64 "\n",
+                       p->stats.traffic.rx_drop);
+
+               fprintf(f, "\t\trx_interrupt_wait %" PRIu64 "\n",
+                       p->stats.traffic.rx_interrupt_wait);
+
+               fprintf(f, "\t\trx_umonitor_umwait %" PRIu64 "\n",
+                       p->stats.traffic.rx_umonitor_umwait);
+
+               fprintf(f, "\t\ttx_ok %" PRIu64 "\n",
+                       p->stats.traffic.tx_ok);
+
+               fprintf(f, "\t\ttotal_polls %" PRIu64 "\n",
+                       p->stats.traffic.total_polls);
+
+               fprintf(f, "\t\tzero_polls %" PRIu64 "\n",
+                       p->stats.traffic.zero_polls);
+
+               fprintf(f, "\t\ttx_nospc_ldb_hw_credits %" PRIu64 "\n",
+                       p->stats.traffic.tx_nospc_ldb_hw_credits);
+
+               fprintf(f, "\t\ttx_nospc_dir_hw_credits %" PRIu64 "\n",
+                       p->stats.traffic.tx_nospc_dir_hw_credits);
+
+               fprintf(f, "\t\ttx_nospc_inflight_max %" PRIu64 "\n",
+                       p->stats.traffic.tx_nospc_inflight_max);
+
+               fprintf(f, "\t\ttx_nospc_new_event_limit %" PRIu64 "\n",
+                       p->stats.traffic.tx_nospc_new_event_limit);
+
+               fprintf(f, "\t\ttx_nospc_inflight_credits %" PRIu64 "\n",
+                       p->stats.traffic.tx_nospc_inflight_credits);
+
+               fprintf(f, "\t\ttx_new %" PRIu64 "\n",
+                       p->stats.tx_op_cnt[RTE_EVENT_OP_NEW]);
+
+               fprintf(f, "\t\ttx_fwd %" PRIu64 "\n",
+                       p->stats.tx_op_cnt[RTE_EVENT_OP_FORWARD]);
+
+               fprintf(f, "\t\ttx_rel %" PRIu64 "\n",
+                       p->stats.tx_op_cnt[RTE_EVENT_OP_RELEASE]);
+
+               fprintf(f, "\t\ttx_implicit_rel %" PRIu64 "\n",
+                       p->stats.tx_implicit_rel);
+
+               fprintf(f, "\t\ttx_sched_ordered %" PRIu64 "\n",
+                       p->stats.tx_sched_cnt[DLB2_SCHED_ORDERED]);
+
+               fprintf(f, "\t\ttx_sched_unordered %" PRIu64 "\n",
+                       p->stats.tx_sched_cnt[DLB2_SCHED_UNORDERED]);
+
+               fprintf(f, "\t\ttx_sched_atomic %" PRIu64 "\n",
+                       p->stats.tx_sched_cnt[DLB2_SCHED_ATOMIC]);
+
+               fprintf(f, "\t\ttx_sched_directed %" PRIu64 "\n",
+                       p->stats.tx_sched_cnt[DLB2_SCHED_DIRECTED]);
+
+               fprintf(f, "\t\ttx_invalid %" PRIu64 "\n",
+                       p->stats.tx_invalid);
+
+               fprintf(f, "\t\trx_sched_ordered %" PRIu64 "\n",
+                       p->stats.rx_sched_cnt[DLB2_SCHED_ORDERED]);
+
+               fprintf(f, "\t\trx_sched_unordered %" PRIu64 "\n",
+                       p->stats.rx_sched_cnt[DLB2_SCHED_UNORDERED]);
+
+               fprintf(f, "\t\trx_sched_atomic %" PRIu64 "\n",
+                       p->stats.rx_sched_cnt[DLB2_SCHED_ATOMIC]);
+
+               fprintf(f, "\t\trx_sched_directed %" PRIu64 "\n",
+                       p->stats.rx_sched_cnt[DLB2_SCHED_DIRECTED]);
+
+               fprintf(f, "\t\trx_sched_invalid %" PRIu64 "\n",
+                       p->stats.rx_sched_invalid);
+       }
+
+       /* Queue level information */
+
+       for (i = 0; i < dlb2->num_queues; i++) {
+               struct dlb2_eventdev_queue *q = &dlb2->ev_queues[i];
+               int j, k;
+
+               if (!q->setup_done)
+                       fprintf(f, "Queue_%d is not configured\n", i);
+
+               fprintf(f, "Queue_%d\n", i);
+               fprintf(f, "========\n");
+
+               fprintf(f, "\tevqueue_%u is set up\n", q->id);
+
+               fprintf(f, "\tqueue is %s\n",
+                       q->qm_queue.is_directed ? "directed" : "load balanced");
+
+               fprintf(f, "\tnum_links=%d, ports -> ", q->num_links);
+
+               for (j = 0; j < dlb2->num_ports; j++) {
+                       struct dlb2_eventdev_port *p = &dlb2->ev_ports[j];
+
+                       for (k = 0; k < DLB2_MAX_NUM_QIDS_PER_LDB_CQ; k++) {
+                               if (p->link[k].valid &&
+                                   p->link[k].queue_id == q->id)
+                                       fprintf(f, "id=%u prio=%u ",
+                                               p->id, p->link[k].priority);
+                       }
+               }
+               fprintf(f, "\n");
+
+               fprintf(f, "\tcurrent depth: %u events\n",
+                       dlb2_get_queue_depth(dlb2, q));
+
+               fprintf(f, "\tnum qid inflights=%u, sched_type=%d\n",
+                       q->qm_queue.num_qid_inflights, q->qm_queue.sched_type);
+       }
+}
index d7390ec..a80d7ab 100644 (file)
@@ -9,6 +9,7 @@ endif
 
 sources = files('dlb2.c',
                'dlb2_iface.c',
+               'dlb2_xstats.c',
                'pf/dlb2_main.c',
                'pf/dlb2_pf.c',
                'pf/base/dlb2_resource.c'