examples/fips_validation: support self-test only
[dpdk.git] / lib / librte_eventdev / rte_event_timer_adapter.c
index 75a14ac..4c5e49e 100644 (file)
@@ -5,24 +5,38 @@
 
 #include <string.h>
 #include <inttypes.h>
+#include <stdbool.h>
+#include <sys/queue.h>
 
 #include <rte_memzone.h>
 #include <rte_memory.h>
 #include <rte_dev.h>
 #include <rte_errno.h>
+#include <rte_malloc.h>
+#include <rte_ring.h>
+#include <rte_mempool.h>
+#include <rte_common.h>
+#include <rte_timer.h>
+#include <rte_service_component.h>
+#include <rte_cycles.h>
 
 #include "rte_eventdev.h"
 #include "rte_eventdev_pmd.h"
+#include "rte_eventdev_trace.h"
 #include "rte_event_timer_adapter.h"
 #include "rte_event_timer_adapter_pmd.h"
 
 #define DATA_MZ_NAME_MAX_LEN 64
 #define DATA_MZ_NAME_FORMAT "rte_event_timer_adapter_data_%d"
 
-static int evtim_logtype;
+RTE_LOG_REGISTER(evtim_logtype, lib.eventdev.adapter.timer, NOTICE);
+RTE_LOG_REGISTER(evtim_buffer_logtype, lib.eventdev.adapter.timer, NOTICE);
+RTE_LOG_REGISTER(evtim_svc_logtype, lib.eventdev.adapter.timer.svc, NOTICE);
 
 static struct rte_event_timer_adapter adapters[RTE_EVENT_TIMER_ADAPTER_NUM_MAX];
 
+static const struct rte_event_timer_adapter_ops swtim_ops;
+
 #define EVTIM_LOG(level, logtype, ...) \
        rte_log(RTE_LOG_ ## level, logtype, \
                RTE_FMT("EVTIMER: %s() line %u: " RTE_FMT_HEAD(__VA_ARGS__,) \
@@ -33,8 +47,14 @@ static struct rte_event_timer_adapter adapters[RTE_EVENT_TIMER_ADAPTER_NUM_MAX];
 #ifdef RTE_LIBRTE_EVENTDEV_DEBUG
 #define EVTIM_LOG_DBG(...) \
        EVTIM_LOG(DEBUG, evtim_logtype, __VA_ARGS__)
+#define EVTIM_BUF_LOG_DBG(...) \
+       EVTIM_LOG(DEBUG, evtim_buffer_logtype, __VA_ARGS__)
+#define EVTIM_SVC_LOG_DBG(...) \
+       EVTIM_LOG(DEBUG, evtim_svc_logtype, __VA_ARGS__)
 #else
 #define EVTIM_LOG_DBG(...) (void)0
+#define EVTIM_BUF_LOG_DBG(...) (void)0
+#define EVTIM_SVC_LOG_DBG(...) (void)0
 #endif
 
 static int
@@ -98,14 +118,14 @@ default_port_conf_cb(uint16_t id, uint8_t event_dev_id, uint8_t *event_port_id,
        return ret;
 }
 
-struct rte_event_timer_adapter * __rte_experimental
+struct rte_event_timer_adapter *
 rte_event_timer_adapter_create(const struct rte_event_timer_adapter_conf *conf)
 {
        return rte_event_timer_adapter_create_ext(conf, default_port_conf_cb,
                                                  NULL);
 }
 
-struct rte_event_timer_adapter * __rte_experimental
+struct rte_event_timer_adapter *
 rte_event_timer_adapter_create_ext(
                const struct rte_event_timer_adapter_conf *conf,
                rte_event_timer_adapter_port_conf_cb_t conf_cb,
@@ -173,26 +193,32 @@ rte_event_timer_adapter_create_ext(
                                                   &adapter->data->caps,
                                                   &adapter->ops);
        if (ret < 0) {
-               rte_errno = ret;
+               rte_errno = -ret;
                goto free_memzone;
        }
 
        if (!(adapter->data->caps &
              RTE_EVENT_TIMER_ADAPTER_CAP_INTERNAL_PORT)) {
-               FUNC_PTR_OR_NULL_RET_WITH_ERRNO(conf_cb, -EINVAL);
+               FUNC_PTR_OR_NULL_RET_WITH_ERRNO(conf_cb, EINVAL);
                ret = conf_cb(adapter->data->id, adapter->data->event_dev_id,
                              &adapter->data->event_port_id, conf_arg);
                if (ret < 0) {
-                       rte_errno = ret;
+                       rte_errno = -ret;
                        goto free_memzone;
                }
        }
 
+       /* If eventdev PMD did not provide ops, use default software
+        * implementation.
+        */
+       if (adapter->ops == NULL)
+               adapter->ops = &swtim_ops;
+
        /* Allow driver to do some setup */
-       FUNC_PTR_OR_NULL_RET_WITH_ERRNO(adapter->ops->init, -ENOTSUP);
+       FUNC_PTR_OR_NULL_RET_WITH_ERRNO(adapter->ops->init, ENOTSUP);
        ret = adapter->ops->init(adapter);
        if (ret < 0) {
-               rte_errno = ret;
+               rte_errno = -ret;
                goto free_memzone;
        }
 
@@ -203,6 +229,8 @@ rte_event_timer_adapter_create_ext(
 
        adapter->allocated = 1;
 
+       rte_eventdev_trace_timer_adapter_create(adapter_id, adapter, conf,
+               conf_cb);
        return adapter;
 
 free_memzone:
@@ -210,7 +238,7 @@ free_memzone:
        return NULL;
 }
 
-int __rte_experimental
+int
 rte_event_timer_adapter_get_info(const struct rte_event_timer_adapter *adapter,
                struct rte_event_timer_adapter_info *adapter_info)
 {
@@ -228,7 +256,7 @@ rte_event_timer_adapter_get_info(const struct rte_event_timer_adapter *adapter,
        return 0;
 }
 
-int __rte_experimental
+int
 rte_event_timer_adapter_start(const struct rte_event_timer_adapter *adapter)
 {
        int ret;
@@ -236,16 +264,22 @@ rte_event_timer_adapter_start(const struct rte_event_timer_adapter *adapter)
        ADAPTER_VALID_OR_ERR_RET(adapter, -EINVAL);
        FUNC_PTR_OR_ERR_RET(adapter->ops->start, -EINVAL);
 
+       if (adapter->data->started) {
+               EVTIM_LOG_ERR("event timer adapter %"PRIu8" already started",
+                             adapter->data->id);
+               return -EALREADY;
+       }
+
        ret = adapter->ops->start(adapter);
        if (ret < 0)
                return ret;
 
        adapter->data->started = 1;
-
+       rte_eventdev_trace_timer_adapter_start(adapter);
        return 0;
 }
 
-int __rte_experimental
+int
 rte_event_timer_adapter_stop(const struct rte_event_timer_adapter *adapter)
 {
        int ret;
@@ -264,11 +298,11 @@ rte_event_timer_adapter_stop(const struct rte_event_timer_adapter *adapter)
                return ret;
 
        adapter->data->started = 0;
-
+       rte_eventdev_trace_timer_adapter_stop(adapter);
        return 0;
 }
 
-struct rte_event_timer_adapter * __rte_experimental
+struct rte_event_timer_adapter *
 rte_event_timer_adapter_lookup(uint16_t adapter_id)
 {
        char name[DATA_MZ_NAME_MAX_LEN];
@@ -305,6 +339,12 @@ rte_event_timer_adapter_lookup(uint16_t adapter_id)
                return NULL;
        }
 
+       /* If eventdev PMD did not provide ops, use default software
+        * implementation.
+        */
+       if (adapter->ops == NULL)
+               adapter->ops = &swtim_ops;
+
        /* Set fast-path function pointers */
        adapter->arm_burst = adapter->ops->arm_burst;
        adapter->arm_tmo_tick_burst = adapter->ops->arm_tmo_tick_burst;
@@ -315,7 +355,7 @@ rte_event_timer_adapter_lookup(uint16_t adapter_id)
        return adapter;
 }
 
-int __rte_experimental
+int
 rte_event_timer_adapter_free(struct rte_event_timer_adapter *adapter)
 {
        int ret;
@@ -342,10 +382,11 @@ rte_event_timer_adapter_free(struct rte_event_timer_adapter *adapter)
        adapter->data = NULL;
        adapter->allocated = 0;
 
+       rte_eventdev_trace_timer_adapter_free(adapter);
        return 0;
 }
 
-int __rte_experimental
+int
 rte_event_timer_adapter_service_id_get(struct rte_event_timer_adapter *adapter,
                                       uint32_t *service_id)
 {
@@ -357,7 +398,7 @@ rte_event_timer_adapter_service_id_get(struct rte_event_timer_adapter *adapter,
        return adapter->data->service_inited ? 0 : -ESRCH;
 }
 
-int __rte_experimental
+int
 rte_event_timer_adapter_stats_get(struct rte_event_timer_adapter *adapter,
                                  struct rte_event_timer_adapter_stats *stats)
 {
@@ -369,7 +410,7 @@ rte_event_timer_adapter_stats_get(struct rte_event_timer_adapter *adapter,
        return adapter->ops->stats_get(adapter, stats);
 }
 
-int __rte_experimental
+int
 rte_event_timer_adapter_stats_reset(struct rte_event_timer_adapter *adapter)
 {
        ADAPTER_VALID_OR_ERR_RET(adapter, -EINVAL);
@@ -377,11 +418,804 @@ rte_event_timer_adapter_stats_reset(struct rte_event_timer_adapter *adapter)
        return adapter->ops->stats_reset(adapter);
 }
 
-RTE_INIT(event_timer_adapter_init_log);
+/*
+ * Software event timer adapter buffer helper functions
+ */
+
+#define NSECPERSEC 1E9
+
+/* Optimizations used to index into the buffer require that the buffer size
+ * be a power of 2.
+ */
+#define EVENT_BUFFER_SZ 4096
+#define EVENT_BUFFER_BATCHSZ 32
+#define EVENT_BUFFER_MASK (EVENT_BUFFER_SZ - 1)
+
+#define EXP_TIM_BUF_SZ 128
+
+struct event_buffer {
+       size_t head;
+       size_t tail;
+       struct rte_event events[EVENT_BUFFER_SZ];
+} __rte_cache_aligned;
+
+static inline bool
+event_buffer_full(struct event_buffer *bufp)
+{
+       return (bufp->head - bufp->tail) == EVENT_BUFFER_SZ;
+}
+
+static inline bool
+event_buffer_batch_ready(struct event_buffer *bufp)
+{
+       return (bufp->head - bufp->tail) >= EVENT_BUFFER_BATCHSZ;
+}
+
 static void
-event_timer_adapter_init_log(void)
+event_buffer_init(struct event_buffer *bufp)
 {
-       evtim_logtype = rte_log_register("lib.eventdev.adapter.timer");
-       if (evtim_logtype >= 0)
-               rte_log_set_level(evtim_logtype, RTE_LOG_NOTICE);
+       bufp->head = bufp->tail = 0;
+       memset(&bufp->events, 0, sizeof(struct rte_event) * EVENT_BUFFER_SZ);
 }
+
+static int
+event_buffer_add(struct event_buffer *bufp, struct rte_event *eventp)
+{
+       size_t head_idx;
+       struct rte_event *buf_eventp;
+
+       if (event_buffer_full(bufp))
+               return -1;
+
+       /* Instead of modulus, bitwise AND with mask to get head_idx. */
+       head_idx = bufp->head & EVENT_BUFFER_MASK;
+       buf_eventp = &bufp->events[head_idx];
+       rte_memcpy(buf_eventp, eventp, sizeof(struct rte_event));
+
+       /* Wrap automatically when overflow occurs. */
+       bufp->head++;
+
+       return 0;
+}
+
+static void
+event_buffer_flush(struct event_buffer *bufp, uint8_t dev_id, uint8_t port_id,
+                  uint16_t *nb_events_flushed,
+                  uint16_t *nb_events_inv)
+{
+       struct rte_event *events = bufp->events;
+       size_t head_idx, tail_idx;
+       uint16_t n = 0;
+
+       /* Instead of modulus, bitwise AND with mask to get index. */
+       head_idx = bufp->head & EVENT_BUFFER_MASK;
+       tail_idx = bufp->tail & EVENT_BUFFER_MASK;
+
+       RTE_ASSERT(head_idx < EVENT_BUFFER_SZ && tail_idx < EVENT_BUFFER_SZ);
+
+       /* Determine the largest contigous run we can attempt to enqueue to the
+        * event device.
+        */
+       if (head_idx > tail_idx)
+               n = head_idx - tail_idx;
+       else if (head_idx < tail_idx)
+               n = EVENT_BUFFER_SZ - tail_idx;
+       else if (event_buffer_full(bufp))
+               n = EVENT_BUFFER_SZ - tail_idx;
+       else {
+               *nb_events_flushed = 0;
+               return;
+       }
+
+       n = RTE_MIN(EVENT_BUFFER_BATCHSZ, n);
+       *nb_events_inv = 0;
+
+       *nb_events_flushed = rte_event_enqueue_burst(dev_id, port_id,
+                                                    &events[tail_idx], n);
+       if (*nb_events_flushed != n) {
+               if (rte_errno == EINVAL) {
+                       EVTIM_LOG_ERR("failed to enqueue invalid event - "
+                                     "dropping it");
+                       (*nb_events_inv)++;
+               } else if (rte_errno == ENOSPC)
+                       rte_pause();
+       }
+
+       if (*nb_events_flushed > 0)
+               EVTIM_BUF_LOG_DBG("enqueued %"PRIu16" timer events to event "
+                                 "device", *nb_events_flushed);
+
+       bufp->tail = bufp->tail + *nb_events_flushed + *nb_events_inv;
+}
+
+/*
+ * Software event timer adapter implementation
+ */
+struct swtim {
+       /* Identifier of service executing timer management logic. */
+       uint32_t service_id;
+       /* The cycle count at which the adapter should next tick */
+       uint64_t next_tick_cycles;
+       /* The tick resolution used by adapter instance. May have been
+        * adjusted from what user requested
+        */
+       uint64_t timer_tick_ns;
+       /* Maximum timeout in nanoseconds allowed by adapter instance. */
+       uint64_t max_tmo_ns;
+       /* Buffered timer expiry events to be enqueued to an event device. */
+       struct event_buffer buffer;
+       /* Statistics */
+       struct rte_event_timer_adapter_stats stats;
+       /* Mempool of timer objects */
+       struct rte_mempool *tim_pool;
+       /* Back pointer for convenience */
+       struct rte_event_timer_adapter *adapter;
+       /* Identifier of timer data instance */
+       uint32_t timer_data_id;
+       /* Track which cores have actually armed a timer */
+       struct {
+               uint16_t v;
+       } __rte_cache_aligned in_use[RTE_MAX_LCORE];
+       /* Track which cores' timer lists should be polled */
+       unsigned int poll_lcores[RTE_MAX_LCORE];
+       /* The number of lists that should be polled */
+       int n_poll_lcores;
+       /* Timers which have expired and can be returned to a mempool */
+       struct rte_timer *expired_timers[EXP_TIM_BUF_SZ];
+       /* The number of timers that can be returned to a mempool */
+       size_t n_expired_timers;
+};
+
+static inline struct swtim *
+swtim_pmd_priv(const struct rte_event_timer_adapter *adapter)
+{
+       return adapter->data->adapter_priv;
+}
+
+static void
+swtim_callback(struct rte_timer *tim)
+{
+       struct rte_event_timer *evtim = tim->arg;
+       struct rte_event_timer_adapter *adapter;
+       unsigned int lcore = rte_lcore_id();
+       struct swtim *sw;
+       uint16_t nb_evs_flushed = 0;
+       uint16_t nb_evs_invalid = 0;
+       uint64_t opaque;
+       int ret;
+       int n_lcores;
+
+       opaque = evtim->impl_opaque[1];
+       adapter = (struct rte_event_timer_adapter *)(uintptr_t)opaque;
+       sw = swtim_pmd_priv(adapter);
+
+       ret = event_buffer_add(&sw->buffer, &evtim->ev);
+       if (ret < 0) {
+               /* If event buffer is full, put timer back in list with
+                * immediate expiry value, so that we process it again on the
+                * next iteration.
+                */
+               ret = rte_timer_alt_reset(sw->timer_data_id, tim, 0, SINGLE,
+                                         lcore, NULL, evtim);
+               if (ret < 0) {
+                       EVTIM_LOG_DBG("event buffer full, failed to reset "
+                                     "timer with immediate expiry value");
+               } else {
+                       sw->stats.evtim_retry_count++;
+                       EVTIM_LOG_DBG("event buffer full, resetting rte_timer "
+                                     "with immediate expiry value");
+               }
+
+               if (unlikely(sw->in_use[lcore].v == 0)) {
+                       sw->in_use[lcore].v = 1;
+                       n_lcores = __atomic_fetch_add(&sw->n_poll_lcores, 1,
+                                                    __ATOMIC_RELAXED);
+                       __atomic_store_n(&sw->poll_lcores[n_lcores], lcore,
+                                       __ATOMIC_RELAXED);
+               }
+       } else {
+               EVTIM_BUF_LOG_DBG("buffered an event timer expiry event");
+
+               /* Empty the buffer here, if necessary, to free older expired
+                * timers only
+                */
+               if (unlikely(sw->n_expired_timers == EXP_TIM_BUF_SZ)) {
+                       rte_mempool_put_bulk(sw->tim_pool,
+                                            (void **)sw->expired_timers,
+                                            sw->n_expired_timers);
+                       sw->n_expired_timers = 0;
+               }
+
+               sw->expired_timers[sw->n_expired_timers++] = tim;
+               sw->stats.evtim_exp_count++;
+
+               __atomic_store_n(&evtim->state, RTE_EVENT_TIMER_NOT_ARMED,
+                               __ATOMIC_RELEASE);
+       }
+
+       if (event_buffer_batch_ready(&sw->buffer)) {
+               event_buffer_flush(&sw->buffer,
+                                  adapter->data->event_dev_id,
+                                  adapter->data->event_port_id,
+                                  &nb_evs_flushed,
+                                  &nb_evs_invalid);
+
+               sw->stats.ev_enq_count += nb_evs_flushed;
+               sw->stats.ev_inv_count += nb_evs_invalid;
+       }
+}
+
+static __rte_always_inline uint64_t
+get_timeout_cycles(struct rte_event_timer *evtim,
+                  const struct rte_event_timer_adapter *adapter)
+{
+       struct swtim *sw = swtim_pmd_priv(adapter);
+       uint64_t timeout_ns = evtim->timeout_ticks * sw->timer_tick_ns;
+       return timeout_ns * rte_get_timer_hz() / NSECPERSEC;
+}
+
+/* This function returns true if one or more (adapter) ticks have occurred since
+ * the last time it was called.
+ */
+static inline bool
+swtim_did_tick(struct swtim *sw)
+{
+       uint64_t cycles_per_adapter_tick, start_cycles;
+       uint64_t *next_tick_cyclesp;
+
+       next_tick_cyclesp = &sw->next_tick_cycles;
+       cycles_per_adapter_tick = sw->timer_tick_ns *
+                       (rte_get_timer_hz() / NSECPERSEC);
+       start_cycles = rte_get_timer_cycles();
+
+       /* Note: initially, *next_tick_cyclesp == 0, so the clause below will
+        * execute, and set things going.
+        */
+
+       if (start_cycles >= *next_tick_cyclesp) {
+               /* Snap the current cycle count to the preceding adapter tick
+                * boundary.
+                */
+               start_cycles -= start_cycles % cycles_per_adapter_tick;
+               *next_tick_cyclesp = start_cycles + cycles_per_adapter_tick;
+
+               return true;
+       }
+
+       return false;
+}
+
+/* Check that event timer timeout value is in range */
+static __rte_always_inline int
+check_timeout(struct rte_event_timer *evtim,
+             const struct rte_event_timer_adapter *adapter)
+{
+       uint64_t tmo_nsec;
+       struct swtim *sw = swtim_pmd_priv(adapter);
+
+       tmo_nsec = evtim->timeout_ticks * sw->timer_tick_ns;
+       if (tmo_nsec > sw->max_tmo_ns)
+               return -1;
+       if (tmo_nsec < sw->timer_tick_ns)
+               return -2;
+
+       return 0;
+}
+
+/* Check that event timer event queue sched type matches destination event queue
+ * sched type
+ */
+static __rte_always_inline int
+check_destination_event_queue(struct rte_event_timer *evtim,
+                             const struct rte_event_timer_adapter *adapter)
+{
+       int ret;
+       uint32_t sched_type;
+
+       ret = rte_event_queue_attr_get(adapter->data->event_dev_id,
+                                      evtim->ev.queue_id,
+                                      RTE_EVENT_QUEUE_ATTR_SCHEDULE_TYPE,
+                                      &sched_type);
+
+       if ((ret == 0 && evtim->ev.sched_type == sched_type) ||
+           ret == -EOVERFLOW)
+               return 0;
+
+       return -1;
+}
+
+static int
+swtim_service_func(void *arg)
+{
+       struct rte_event_timer_adapter *adapter = arg;
+       struct swtim *sw = swtim_pmd_priv(adapter);
+       uint16_t nb_evs_flushed = 0;
+       uint16_t nb_evs_invalid = 0;
+
+       if (swtim_did_tick(sw)) {
+               rte_timer_alt_manage(sw->timer_data_id,
+                                    sw->poll_lcores,
+                                    sw->n_poll_lcores,
+                                    swtim_callback);
+
+               /* Return expired timer objects back to mempool */
+               rte_mempool_put_bulk(sw->tim_pool, (void **)sw->expired_timers,
+                                    sw->n_expired_timers);
+               sw->n_expired_timers = 0;
+
+               event_buffer_flush(&sw->buffer,
+                                  adapter->data->event_dev_id,
+                                  adapter->data->event_port_id,
+                                  &nb_evs_flushed,
+                                  &nb_evs_invalid);
+
+               sw->stats.ev_enq_count += nb_evs_flushed;
+               sw->stats.ev_inv_count += nb_evs_invalid;
+               sw->stats.adapter_tick_count++;
+       }
+
+       return 0;
+}
+
+/* The adapter initialization function rounds the mempool size up to the next
+ * power of 2, so we can take the difference between that value and what the
+ * user requested, and use the space for caches.  This avoids a scenario where a
+ * user can't arm the number of timers the adapter was configured with because
+ * mempool objects have been lost to caches.
+ *
+ * nb_actual should always be a power of 2, so we can iterate over the powers
+ * of 2 to see what the largest cache size we can use is.
+ */
+static int
+compute_msg_mempool_cache_size(uint64_t nb_requested, uint64_t nb_actual)
+{
+       int i;
+       int size;
+       int cache_size = 0;
+
+       for (i = 0;; i++) {
+               size = 1 << i;
+
+               if (RTE_MAX_LCORE * size < (int)(nb_actual - nb_requested) &&
+                   size < RTE_MEMPOOL_CACHE_MAX_SIZE &&
+                   size <= nb_actual / 1.5)
+                       cache_size = size;
+               else
+                       break;
+       }
+
+       return cache_size;
+}
+
+static int
+swtim_init(struct rte_event_timer_adapter *adapter)
+{
+       int i, ret;
+       struct swtim *sw;
+       unsigned int flags;
+       struct rte_service_spec service;
+
+       /* Allocate storage for private data area */
+#define SWTIM_NAMESIZE 32
+       char swtim_name[SWTIM_NAMESIZE];
+       snprintf(swtim_name, SWTIM_NAMESIZE, "swtim_%"PRIu8,
+                       adapter->data->id);
+       sw = rte_zmalloc_socket(swtim_name, sizeof(*sw), RTE_CACHE_LINE_SIZE,
+                       adapter->data->socket_id);
+       if (sw == NULL) {
+               EVTIM_LOG_ERR("failed to allocate space for private data");
+               rte_errno = ENOMEM;
+               return -1;
+       }
+
+       /* Connect storage to adapter instance */
+       adapter->data->adapter_priv = sw;
+       sw->adapter = adapter;
+
+       sw->timer_tick_ns = adapter->data->conf.timer_tick_ns;
+       sw->max_tmo_ns = adapter->data->conf.max_tmo_ns;
+
+       /* Create a timer pool */
+       char pool_name[SWTIM_NAMESIZE];
+       snprintf(pool_name, SWTIM_NAMESIZE, "swtim_pool_%"PRIu8,
+                adapter->data->id);
+       /* Optimal mempool size is a power of 2 minus one */
+       uint64_t nb_timers = rte_align64pow2(adapter->data->conf.nb_timers);
+       int pool_size = nb_timers - 1;
+       int cache_size = compute_msg_mempool_cache_size(
+                               adapter->data->conf.nb_timers, nb_timers);
+       flags = 0; /* pool is multi-producer, multi-consumer */
+       sw->tim_pool = rte_mempool_create(pool_name, pool_size,
+                       sizeof(struct rte_timer), cache_size, 0, NULL, NULL,
+                       NULL, NULL, adapter->data->socket_id, flags);
+       if (sw->tim_pool == NULL) {
+               EVTIM_LOG_ERR("failed to create timer object mempool");
+               rte_errno = ENOMEM;
+               goto free_alloc;
+       }
+
+       /* Initialize the variables that track in-use timer lists */
+       for (i = 0; i < RTE_MAX_LCORE; i++)
+               sw->in_use[i].v = 0;
+
+       /* Initialize the timer subsystem and allocate timer data instance */
+       ret = rte_timer_subsystem_init();
+       if (ret < 0) {
+               if (ret != -EALREADY) {
+                       EVTIM_LOG_ERR("failed to initialize timer subsystem");
+                       rte_errno = -ret;
+                       goto free_mempool;
+               }
+       }
+
+       ret = rte_timer_data_alloc(&sw->timer_data_id);
+       if (ret < 0) {
+               EVTIM_LOG_ERR("failed to allocate timer data instance");
+               rte_errno = -ret;
+               goto free_mempool;
+       }
+
+       /* Initialize timer event buffer */
+       event_buffer_init(&sw->buffer);
+
+       sw->adapter = adapter;
+
+       /* Register a service component to run adapter logic */
+       memset(&service, 0, sizeof(service));
+       snprintf(service.name, RTE_SERVICE_NAME_MAX,
+                "swtim_svc_%"PRIu8, adapter->data->id);
+       service.socket_id = adapter->data->socket_id;
+       service.callback = swtim_service_func;
+       service.callback_userdata = adapter;
+       service.capabilities &= ~(RTE_SERVICE_CAP_MT_SAFE);
+       ret = rte_service_component_register(&service, &sw->service_id);
+       if (ret < 0) {
+               EVTIM_LOG_ERR("failed to register service %s with id %"PRIu32
+                             ": err = %d", service.name, sw->service_id,
+                             ret);
+
+               rte_errno = ENOSPC;
+               goto free_mempool;
+       }
+
+       EVTIM_LOG_DBG("registered service %s with id %"PRIu32, service.name,
+                     sw->service_id);
+
+       adapter->data->service_id = sw->service_id;
+       adapter->data->service_inited = 1;
+
+       return 0;
+free_mempool:
+       rte_mempool_free(sw->tim_pool);
+free_alloc:
+       rte_free(sw);
+       return -1;
+}
+
+static void
+swtim_free_tim(struct rte_timer *tim, void *arg)
+{
+       struct swtim *sw = arg;
+
+       rte_mempool_put(sw->tim_pool, tim);
+}
+
+/* Traverse the list of outstanding timers and put them back in the mempool
+ * before freeing the adapter to avoid leaking the memory.
+ */
+static int
+swtim_uninit(struct rte_event_timer_adapter *adapter)
+{
+       int ret;
+       struct swtim *sw = swtim_pmd_priv(adapter);
+
+       /* Free outstanding timers */
+       rte_timer_stop_all(sw->timer_data_id,
+                          sw->poll_lcores,
+                          sw->n_poll_lcores,
+                          swtim_free_tim,
+                          sw);
+
+       ret = rte_service_component_unregister(sw->service_id);
+       if (ret < 0) {
+               EVTIM_LOG_ERR("failed to unregister service component");
+               return ret;
+       }
+
+       rte_mempool_free(sw->tim_pool);
+       rte_free(sw);
+       adapter->data->adapter_priv = NULL;
+
+       return 0;
+}
+
+static inline int32_t
+get_mapped_count_for_service(uint32_t service_id)
+{
+       int32_t core_count, i, mapped_count = 0;
+       uint32_t lcore_arr[RTE_MAX_LCORE];
+
+       core_count = rte_service_lcore_list(lcore_arr, RTE_MAX_LCORE);
+
+       for (i = 0; i < core_count; i++)
+               if (rte_service_map_lcore_get(service_id, lcore_arr[i]) == 1)
+                       mapped_count++;
+
+       return mapped_count;
+}
+
+static int
+swtim_start(const struct rte_event_timer_adapter *adapter)
+{
+       int mapped_count;
+       struct swtim *sw = swtim_pmd_priv(adapter);
+
+       /* Mapping the service to more than one service core can introduce
+        * delays while one thread is waiting to acquire a lock, so only allow
+        * one core to be mapped to the service.
+        *
+        * Note: the service could be modified such that it spreads cores to
+        * poll over multiple service instances.
+        */
+       mapped_count = get_mapped_count_for_service(sw->service_id);
+
+       if (mapped_count != 1)
+               return mapped_count < 1 ? -ENOENT : -ENOTSUP;
+
+       return rte_service_component_runstate_set(sw->service_id, 1);
+}
+
+static int
+swtim_stop(const struct rte_event_timer_adapter *adapter)
+{
+       int ret;
+       struct swtim *sw = swtim_pmd_priv(adapter);
+
+       ret = rte_service_component_runstate_set(sw->service_id, 0);
+       if (ret < 0)
+               return ret;
+
+       /* Wait for the service to complete its final iteration */
+       while (rte_service_may_be_active(sw->service_id))
+               rte_pause();
+
+       return 0;
+}
+
+static void
+swtim_get_info(const struct rte_event_timer_adapter *adapter,
+               struct rte_event_timer_adapter_info *adapter_info)
+{
+       struct swtim *sw = swtim_pmd_priv(adapter);
+       adapter_info->min_resolution_ns = sw->timer_tick_ns;
+       adapter_info->max_tmo_ns = sw->max_tmo_ns;
+}
+
+static int
+swtim_stats_get(const struct rte_event_timer_adapter *adapter,
+               struct rte_event_timer_adapter_stats *stats)
+{
+       struct swtim *sw = swtim_pmd_priv(adapter);
+       *stats = sw->stats; /* structure copy */
+       return 0;
+}
+
+static int
+swtim_stats_reset(const struct rte_event_timer_adapter *adapter)
+{
+       struct swtim *sw = swtim_pmd_priv(adapter);
+       memset(&sw->stats, 0, sizeof(sw->stats));
+       return 0;
+}
+
+static uint16_t
+__swtim_arm_burst(const struct rte_event_timer_adapter *adapter,
+               struct rte_event_timer **evtims,
+               uint16_t nb_evtims)
+{
+       int i, ret;
+       struct swtim *sw = swtim_pmd_priv(adapter);
+       uint32_t lcore_id = rte_lcore_id();
+       struct rte_timer *tim, *tims[nb_evtims];
+       uint64_t cycles;
+       int n_lcores;
+       /* Timer list for this lcore is not in use. */
+       uint16_t exp_state = 0;
+       enum rte_event_timer_state n_state;
+
+#ifdef RTE_LIBRTE_EVENTDEV_DEBUG
+       /* Check that the service is running. */
+       if (rte_service_runstate_get(adapter->data->service_id) != 1) {
+               rte_errno = EINVAL;
+               return 0;
+       }
+#endif
+
+       /* Adjust lcore_id if non-EAL thread. Arbitrarily pick the timer list of
+        * the highest lcore to insert such timers into
+        */
+       if (lcore_id == LCORE_ID_ANY)
+               lcore_id = RTE_MAX_LCORE - 1;
+
+       /* If this is the first time we're arming an event timer on this lcore,
+        * mark this lcore as "in use"; this will cause the service
+        * function to process the timer list that corresponds to this lcore.
+        * The atomic compare-and-swap operation can prevent the race condition
+        * on in_use flag between multiple non-EAL threads.
+        */
+       if (unlikely(__atomic_compare_exchange_n(&sw->in_use[lcore_id].v,
+                       &exp_state, 1, 0,
+                       __ATOMIC_RELAXED, __ATOMIC_RELAXED))) {
+               EVTIM_LOG_DBG("Adding lcore id = %u to list of lcores to poll",
+                             lcore_id);
+               n_lcores = __atomic_fetch_add(&sw->n_poll_lcores, 1,
+                                            __ATOMIC_RELAXED);
+               __atomic_store_n(&sw->poll_lcores[n_lcores], lcore_id,
+                               __ATOMIC_RELAXED);
+       }
+
+       ret = rte_mempool_get_bulk(sw->tim_pool, (void **)tims,
+                                  nb_evtims);
+       if (ret < 0) {
+               rte_errno = ENOSPC;
+               return 0;
+       }
+
+       for (i = 0; i < nb_evtims; i++) {
+               n_state = __atomic_load_n(&evtims[i]->state, __ATOMIC_ACQUIRE);
+               if (n_state == RTE_EVENT_TIMER_ARMED) {
+                       rte_errno = EALREADY;
+                       break;
+               } else if (!(n_state == RTE_EVENT_TIMER_NOT_ARMED ||
+                            n_state == RTE_EVENT_TIMER_CANCELED)) {
+                       rte_errno = EINVAL;
+                       break;
+               }
+
+               ret = check_timeout(evtims[i], adapter);
+               if (unlikely(ret == -1)) {
+                       __atomic_store_n(&evtims[i]->state,
+                                       RTE_EVENT_TIMER_ERROR_TOOLATE,
+                                       __ATOMIC_RELAXED);
+                       rte_errno = EINVAL;
+                       break;
+               } else if (unlikely(ret == -2)) {
+                       __atomic_store_n(&evtims[i]->state,
+                                       RTE_EVENT_TIMER_ERROR_TOOEARLY,
+                                       __ATOMIC_RELAXED);
+                       rte_errno = EINVAL;
+                       break;
+               }
+
+               if (unlikely(check_destination_event_queue(evtims[i],
+                                                          adapter) < 0)) {
+                       __atomic_store_n(&evtims[i]->state,
+                                       RTE_EVENT_TIMER_ERROR,
+                                       __ATOMIC_RELAXED);
+                       rte_errno = EINVAL;
+                       break;
+               }
+
+               tim = tims[i];
+               rte_timer_init(tim);
+
+               evtims[i]->impl_opaque[0] = (uintptr_t)tim;
+               evtims[i]->impl_opaque[1] = (uintptr_t)adapter;
+
+               cycles = get_timeout_cycles(evtims[i], adapter);
+               ret = rte_timer_alt_reset(sw->timer_data_id, tim, cycles,
+                                         SINGLE, lcore_id, NULL, evtims[i]);
+               if (ret < 0) {
+                       /* tim was in RUNNING or CONFIG state */
+                       __atomic_store_n(&evtims[i]->state,
+                                       RTE_EVENT_TIMER_ERROR,
+                                       __ATOMIC_RELEASE);
+                       break;
+               }
+
+               EVTIM_LOG_DBG("armed an event timer");
+               /* RELEASE ordering guarantees the adapter specific value
+                * changes observed before the update of state.
+                */
+               __atomic_store_n(&evtims[i]->state, RTE_EVENT_TIMER_ARMED,
+                               __ATOMIC_RELEASE);
+       }
+
+       if (i < nb_evtims)
+               rte_mempool_put_bulk(sw->tim_pool,
+                                    (void **)&tims[i], nb_evtims - i);
+
+       return i;
+}
+
+static uint16_t
+swtim_arm_burst(const struct rte_event_timer_adapter *adapter,
+               struct rte_event_timer **evtims,
+               uint16_t nb_evtims)
+{
+       return __swtim_arm_burst(adapter, evtims, nb_evtims);
+}
+
+static uint16_t
+swtim_cancel_burst(const struct rte_event_timer_adapter *adapter,
+                  struct rte_event_timer **evtims,
+                  uint16_t nb_evtims)
+{
+       int i, ret;
+       struct rte_timer *timp;
+       uint64_t opaque;
+       struct swtim *sw = swtim_pmd_priv(adapter);
+       enum rte_event_timer_state n_state;
+
+#ifdef RTE_LIBRTE_EVENTDEV_DEBUG
+       /* Check that the service is running. */
+       if (rte_service_runstate_get(adapter->data->service_id) != 1) {
+               rte_errno = EINVAL;
+               return 0;
+       }
+#endif
+
+       for (i = 0; i < nb_evtims; i++) {
+               /* Don't modify the event timer state in these cases */
+               /* ACQUIRE ordering guarantees the access of implementation
+                * specific opaque data under the correct state.
+                */
+               n_state = __atomic_load_n(&evtims[i]->state, __ATOMIC_ACQUIRE);
+               if (n_state == RTE_EVENT_TIMER_CANCELED) {
+                       rte_errno = EALREADY;
+                       break;
+               } else if (n_state != RTE_EVENT_TIMER_ARMED) {
+                       rte_errno = EINVAL;
+                       break;
+               }
+
+               opaque = evtims[i]->impl_opaque[0];
+               timp = (struct rte_timer *)(uintptr_t)opaque;
+               RTE_ASSERT(timp != NULL);
+
+               ret = rte_timer_alt_stop(sw->timer_data_id, timp);
+               if (ret < 0) {
+                       /* Timer is running or being configured */
+                       rte_errno = EAGAIN;
+                       break;
+               }
+
+               rte_mempool_put(sw->tim_pool, (void **)timp);
+
+               /* The RELEASE ordering here pairs with atomic ordering
+                * to make sure the state update data observed between
+                * threads.
+                */
+               __atomic_store_n(&evtims[i]->state, RTE_EVENT_TIMER_CANCELED,
+                               __ATOMIC_RELEASE);
+       }
+
+       return i;
+}
+
+static uint16_t
+swtim_arm_tmo_tick_burst(const struct rte_event_timer_adapter *adapter,
+                        struct rte_event_timer **evtims,
+                        uint64_t timeout_ticks,
+                        uint16_t nb_evtims)
+{
+       int i;
+
+       for (i = 0; i < nb_evtims; i++)
+               evtims[i]->timeout_ticks = timeout_ticks;
+
+       return __swtim_arm_burst(adapter, evtims, nb_evtims);
+}
+
+static const struct rte_event_timer_adapter_ops swtim_ops = {
+       .init                   = swtim_init,
+       .uninit                 = swtim_uninit,
+       .start                  = swtim_start,
+       .stop                   = swtim_stop,
+       .get_info               = swtim_get_info,
+       .stats_get              = swtim_stats_get,
+       .stats_reset            = swtim_stats_reset,
+       .arm_burst              = swtim_arm_burst,
+       .arm_tmo_tick_burst     = swtim_arm_tmo_tick_burst,
+       .cancel_burst           = swtim_cancel_burst,
+};