diff options
Diffstat (limited to 'lib/librte_eventdev')
| -rw-r--r-- | lib/librte_eventdev/Makefile | 15 | ||||
| -rw-r--r-- | lib/librte_eventdev/meson.build | 19 | ||||
| -rw-r--r-- | lib/librte_eventdev/rte_event_crypto_adapter.c | 1128 | ||||
| -rw-r--r-- | lib/librte_eventdev/rte_event_crypto_adapter.h | 575 | ||||
| -rw-r--r-- | lib/librte_eventdev/rte_event_eth_rx_adapter.c | 1662 | ||||
| -rw-r--r-- | lib/librte_eventdev/rte_event_eth_rx_adapter.h | 129 | ||||
| -rw-r--r-- | lib/librte_eventdev/rte_event_ring.c | 15 | ||||
| -rw-r--r-- | lib/librte_eventdev/rte_event_ring.h | 4 | ||||
| -rw-r--r-- | lib/librte_eventdev/rte_event_timer_adapter.c | 1299 | ||||
| -rw-r--r-- | lib/librte_eventdev/rte_event_timer_adapter.h | 766 | ||||
| -rw-r--r-- | lib/librte_eventdev/rte_event_timer_adapter_pmd.h | 114 | ||||
| -rw-r--r-- | lib/librte_eventdev/rte_eventdev.c | 79 | ||||
| -rw-r--r-- | lib/librte_eventdev/rte_eventdev.h | 167 | ||||
| -rw-r--r-- | lib/librte_eventdev/rte_eventdev_pmd.h | 225 | ||||
| -rw-r--r-- | lib/librte_eventdev/rte_eventdev_version.map | 39 |
15 files changed, 5910 insertions, 326 deletions
diff --git a/lib/librte_eventdev/Makefile b/lib/librte_eventdev/Makefile index d27dd070..47f599a6 100644 --- a/lib/librte_eventdev/Makefile +++ b/lib/librte_eventdev/Makefile @@ -8,18 +8,26 @@ include $(RTE_SDK)/mk/rte.vars.mk LIB = librte_eventdev.a # library version -LIBABIVER := 3 +LIBABIVER := 5 # build flags CFLAGS += -DALLOW_EXPERIMENTAL_API CFLAGS += -O3 CFLAGS += $(WERROR_FLAGS) -LDLIBS += -lrte_eal -lrte_ring -lrte_ethdev -lrte_hash +ifeq ($(CONFIG_RTE_EXEC_ENV_LINUXAPP),y) +CFLAGS += -DLINUX +else +CFLAGS += -DBSD +endif +LDLIBS += -lrte_eal -lrte_ring -lrte_ethdev -lrte_hash -lrte_mempool -lrte_timer +LDLIBS += -lrte_mbuf -lrte_cryptodev -lpthread # library source files SRCS-y += rte_eventdev.c SRCS-y += rte_event_ring.c SRCS-y += rte_event_eth_rx_adapter.c +SRCS-y += rte_event_timer_adapter.c +SRCS-y += rte_event_crypto_adapter.c # export include files SYMLINK-y-include += rte_eventdev.h @@ -28,6 +36,9 @@ SYMLINK-y-include += rte_eventdev_pmd_pci.h SYMLINK-y-include += rte_eventdev_pmd_vdev.h SYMLINK-y-include += rte_event_ring.h SYMLINK-y-include += rte_event_eth_rx_adapter.h +SYMLINK-y-include += rte_event_timer_adapter.h +SYMLINK-y-include += rte_event_timer_adapter_pmd.h +SYMLINK-y-include += rte_event_crypto_adapter.h # versioning export map EXPORT_MAP := rte_eventdev_version.map diff --git a/lib/librte_eventdev/meson.build b/lib/librte_eventdev/meson.build index d1a99602..3cbaf298 100644 --- a/lib/librte_eventdev/meson.build +++ b/lib/librte_eventdev/meson.build @@ -1,14 +1,27 @@ # SPDX-License-Identifier: BSD-3-Clause # Copyright(c) 2017 Intel Corporation +version = 5 allow_experimental_apis = true + +if host_machine.system() == 'linux' + cflags += '-DLINUX' +else + cflags += '-DBSD' +endif + sources = files('rte_eventdev.c', 'rte_event_ring.c', - 'rte_event_eth_rx_adapter.c') + 'rte_event_eth_rx_adapter.c', + 'rte_event_timer_adapter.c', + 'rte_event_crypto_adapter.c') headers = files('rte_eventdev.h', 'rte_eventdev_pmd.h', 'rte_eventdev_pmd_pci.h', 'rte_eventdev_pmd_vdev.h', 'rte_event_ring.h', - 'rte_event_eth_rx_adapter.h') -deps += ['ring', 'ethdev', 'hash'] + 'rte_event_eth_rx_adapter.h', + 'rte_event_timer_adapter.h', + 'rte_event_timer_adapter_pmd.h', + 'rte_event_crypto_adapter.h') +deps += ['ring', 'ethdev', 'hash', 'mempool', 'mbuf', 'timer', 'cryptodev'] diff --git a/lib/librte_eventdev/rte_event_crypto_adapter.c b/lib/librte_eventdev/rte_event_crypto_adapter.c new file mode 100644 index 00000000..11b28ca9 --- /dev/null +++ b/lib/librte_eventdev/rte_event_crypto_adapter.c @@ -0,0 +1,1128 @@ +/* SPDX-License-Identifier: BSD-3-Clause + * Copyright(c) 2018 Intel Corporation. + * All rights reserved. + */ + +#include <string.h> +#include <stdbool.h> +#include <rte_common.h> +#include <rte_dev.h> +#include <rte_errno.h> +#include <rte_cryptodev.h> +#include <rte_cryptodev_pmd.h> +#include <rte_log.h> +#include <rte_malloc.h> +#include <rte_service_component.h> + +#include "rte_eventdev.h" +#include "rte_eventdev_pmd.h" +#include "rte_event_crypto_adapter.h" + +#define BATCH_SIZE 32 +#define DEFAULT_MAX_NB 128 +#define CRYPTO_ADAPTER_NAME_LEN 32 +#define CRYPTO_ADAPTER_MEM_NAME_LEN 32 +#define CRYPTO_ADAPTER_MAX_EV_ENQ_RETRIES 100 + +/* Flush an instance's enqueue buffers every CRYPTO_ENQ_FLUSH_THRESHOLD + * iterations of eca_crypto_adapter_enq_run() + */ +#define CRYPTO_ENQ_FLUSH_THRESHOLD 1024 + +struct rte_event_crypto_adapter { + /* Event device identifier */ + uint8_t eventdev_id; + /* Event port identifier */ + uint8_t event_port_id; + /* Store event device's implicit release capability */ + uint8_t implicit_release_disabled; + /* Max crypto ops processed in any service function invocation */ + uint32_t max_nb; + /* Lock to serialize config updates with service function */ + rte_spinlock_t lock; + /* Next crypto device to be processed */ + uint16_t next_cdev_id; + /* Per crypto device structure */ + struct crypto_device_info *cdevs; + /* Loop counter to flush crypto ops */ + uint16_t transmit_loop_count; + /* Per instance stats structure */ + struct rte_event_crypto_adapter_stats crypto_stats; + /* Configuration callback for rte_service configuration */ + rte_event_crypto_adapter_conf_cb conf_cb; + /* Configuration callback argument */ + void *conf_arg; + /* Set if default_cb is being used */ + int default_cb_arg; + /* Service initialization state */ + uint8_t service_inited; + /* Memory allocation name */ + char mem_name[CRYPTO_ADAPTER_MEM_NAME_LEN]; + /* Socket identifier cached from eventdev */ + int socket_id; + /* Per adapter EAL service */ + uint32_t service_id; + /* No. of queue pairs configured */ + uint16_t nb_qps; + /* Adapter mode */ + enum rte_event_crypto_adapter_mode mode; +} __rte_cache_aligned; + +/* Per crypto device information */ +struct crypto_device_info { + /* Pointer to cryptodev */ + struct rte_cryptodev *dev; + /* Pointer to queue pair info */ + struct crypto_queue_pair_info *qpairs; + /* Next queue pair to be processed */ + uint16_t next_queue_pair_id; + /* Set to indicate cryptodev->eventdev packet + * transfer uses a hardware mechanism + */ + uint8_t internal_event_port; + /* Set to indicate processing has been started */ + uint8_t dev_started; + /* If num_qpairs > 0, the start callback will + * be invoked if not already invoked + */ + uint16_t num_qpairs; +} __rte_cache_aligned; + +/* Per queue pair information */ +struct crypto_queue_pair_info { + /* Set to indicate queue pair is enabled */ + bool qp_enabled; + /* Pointer to hold rte_crypto_ops for batching */ + struct rte_crypto_op **op_buffer; + /* No of crypto ops accumulated */ + uint8_t len; +} __rte_cache_aligned; + +static struct rte_event_crypto_adapter **event_crypto_adapter; + +/* Macros to check for valid adapter */ +#define EVENT_CRYPTO_ADAPTER_ID_VALID_OR_ERR_RET(id, retval) do { \ + if (!eca_valid_id(id)) { \ + RTE_EDEV_LOG_ERR("Invalid crypto adapter id = %d\n", id); \ + return retval; \ + } \ +} while (0) + +static inline int +eca_valid_id(uint8_t id) +{ + return id < RTE_EVENT_CRYPTO_ADAPTER_MAX_INSTANCE; +} + +static int +eca_init(void) +{ + const char *name = "crypto_adapter_array"; + const struct rte_memzone *mz; + unsigned int sz; + + sz = sizeof(*event_crypto_adapter) * + RTE_EVENT_CRYPTO_ADAPTER_MAX_INSTANCE; + sz = RTE_ALIGN(sz, RTE_CACHE_LINE_SIZE); + + mz = rte_memzone_lookup(name); + if (mz == NULL) { + mz = rte_memzone_reserve_aligned(name, sz, rte_socket_id(), 0, + RTE_CACHE_LINE_SIZE); + if (mz == NULL) { + RTE_EDEV_LOG_ERR("failed to reserve memzone err = %" + PRId32, rte_errno); + return -rte_errno; + } + } + + event_crypto_adapter = mz->addr; + return 0; +} + +static inline struct rte_event_crypto_adapter * +eca_id_to_adapter(uint8_t id) +{ + return event_crypto_adapter ? + event_crypto_adapter[id] : NULL; +} + +static int +eca_default_config_cb(uint8_t id, uint8_t dev_id, + struct rte_event_crypto_adapter_conf *conf, void *arg) +{ + struct rte_event_dev_config dev_conf; + struct rte_eventdev *dev; + uint8_t port_id; + int started; + int ret; + struct rte_event_port_conf *port_conf = arg; + struct rte_event_crypto_adapter *adapter = eca_id_to_adapter(id); + + dev = &rte_eventdevs[adapter->eventdev_id]; + dev_conf = dev->data->dev_conf; + + started = dev->data->dev_started; + if (started) + rte_event_dev_stop(dev_id); + port_id = dev_conf.nb_event_ports; + dev_conf.nb_event_ports += 1; + ret = rte_event_dev_configure(dev_id, &dev_conf); + if (ret) { + RTE_EDEV_LOG_ERR("failed to configure event dev %u\n", dev_id); + if (started) { + if (rte_event_dev_start(dev_id)) + return -EIO; + } + return ret; + } + + ret = rte_event_port_setup(dev_id, port_id, port_conf); + if (ret) { + RTE_EDEV_LOG_ERR("failed to setup event port %u\n", port_id); + return ret; + } + + conf->event_port_id = port_id; + conf->max_nb = DEFAULT_MAX_NB; + if (started) + ret = rte_event_dev_start(dev_id); + + adapter->default_cb_arg = 1; + return ret; +} + +int __rte_experimental +rte_event_crypto_adapter_create_ext(uint8_t id, uint8_t dev_id, + rte_event_crypto_adapter_conf_cb conf_cb, + enum rte_event_crypto_adapter_mode mode, + void *conf_arg) +{ + struct rte_event_crypto_adapter *adapter; + char mem_name[CRYPTO_ADAPTER_NAME_LEN]; + struct rte_event_dev_info dev_info; + int socket_id; + uint8_t i; + int ret; + + EVENT_CRYPTO_ADAPTER_ID_VALID_OR_ERR_RET(id, -EINVAL); + RTE_EVENTDEV_VALID_DEVID_OR_ERR_RET(dev_id, -EINVAL); + if (conf_cb == NULL) + return -EINVAL; + + if (event_crypto_adapter == NULL) { + ret = eca_init(); + if (ret) + return ret; + } + + adapter = eca_id_to_adapter(id); + if (adapter != NULL) { + RTE_EDEV_LOG_ERR("Crypto adapter id %u already exists!", id); + return -EEXIST; + } + + socket_id = rte_event_dev_socket_id(dev_id); + snprintf(mem_name, CRYPTO_ADAPTER_MEM_NAME_LEN, + "rte_event_crypto_adapter_%d", id); + + adapter = rte_zmalloc_socket(mem_name, sizeof(*adapter), + RTE_CACHE_LINE_SIZE, socket_id); + if (adapter == NULL) { + RTE_EDEV_LOG_ERR("Failed to get mem for event crypto adapter!"); + return -ENOMEM; + } + + ret = rte_event_dev_info_get(dev_id, &dev_info); + if (ret < 0) { + RTE_EDEV_LOG_ERR("Failed to get info for eventdev %d: %s!", + dev_id, dev_info.driver_name); + return ret; + } + + adapter->implicit_release_disabled = (dev_info.event_dev_cap & + RTE_EVENT_DEV_CAP_IMPLICIT_RELEASE_DISABLE); + adapter->eventdev_id = dev_id; + adapter->socket_id = socket_id; + adapter->conf_cb = conf_cb; + adapter->conf_arg = conf_arg; + adapter->mode = mode; + strcpy(adapter->mem_name, mem_name); + adapter->cdevs = rte_zmalloc_socket(adapter->mem_name, + rte_cryptodev_count() * + sizeof(struct crypto_device_info), 0, + socket_id); + if (adapter->cdevs == NULL) { + RTE_EDEV_LOG_ERR("Failed to get mem for crypto devices\n"); + rte_free(adapter); + return -ENOMEM; + } + + rte_spinlock_init(&adapter->lock); + for (i = 0; i < rte_cryptodev_count(); i++) + adapter->cdevs[i].dev = rte_cryptodev_pmd_get_dev(i); + + event_crypto_adapter[id] = adapter; + + return 0; +} + + +int __rte_experimental +rte_event_crypto_adapter_create(uint8_t id, uint8_t dev_id, + struct rte_event_port_conf *port_config, + enum rte_event_crypto_adapter_mode mode) +{ + struct rte_event_port_conf *pc; + int ret; + + if (port_config == NULL) + return -EINVAL; + EVENT_CRYPTO_ADAPTER_ID_VALID_OR_ERR_RET(id, -EINVAL); + + pc = rte_malloc(NULL, sizeof(*pc), 0); + if (pc == NULL) + return -ENOMEM; + *pc = *port_config; + ret = rte_event_crypto_adapter_create_ext(id, dev_id, + eca_default_config_cb, + mode, + pc); + if (ret) + rte_free(pc); + + return ret; +} + +int __rte_experimental +rte_event_crypto_adapter_free(uint8_t id) +{ + struct rte_event_crypto_adapter *adapter; + + EVENT_CRYPTO_ADAPTER_ID_VALID_OR_ERR_RET(id, -EINVAL); + + adapter = eca_id_to_adapter(id); + if (adapter == NULL) + return -EINVAL; + + if (adapter->nb_qps) { + RTE_EDEV_LOG_ERR("%" PRIu16 "Queue pairs not deleted", + adapter->nb_qps); + return -EBUSY; + } + + if (adapter->default_cb_arg) + rte_free(adapter->conf_arg); + rte_free(adapter->cdevs); + rte_free(adapter); + event_crypto_adapter[id] = NULL; + + return 0; +} + +static inline unsigned int +eca_enq_to_cryptodev(struct rte_event_crypto_adapter *adapter, + struct rte_event *ev, unsigned int cnt) +{ + struct rte_event_crypto_adapter_stats *stats = &adapter->crypto_stats; + union rte_event_crypto_metadata *m_data = NULL; + struct crypto_queue_pair_info *qp_info = NULL; + struct rte_crypto_op *crypto_op; + unsigned int i, n; + uint16_t qp_id, len, ret; + uint8_t cdev_id; + + len = 0; + ret = 0; + n = 0; + stats->event_deq_count += cnt; + + for (i = 0; i < cnt; i++) { + crypto_op = ev[i].event_ptr; + if (crypto_op == NULL) + continue; + if (crypto_op->sess_type == RTE_CRYPTO_OP_WITH_SESSION) { + m_data = rte_cryptodev_sym_session_get_user_data( + crypto_op->sym->session); + if (m_data == NULL) { + rte_pktmbuf_free(crypto_op->sym->m_src); + rte_crypto_op_free(crypto_op); + continue; + } + + cdev_id = m_data->request_info.cdev_id; + qp_id = m_data->request_info.queue_pair_id; + qp_info = &adapter->cdevs[cdev_id].qpairs[qp_id]; + if (qp_info == NULL) { + rte_pktmbuf_free(crypto_op->sym->m_src); + rte_crypto_op_free(crypto_op); + continue; + } + len = qp_info->len; + qp_info->op_buffer[len] = crypto_op; + len++; + } else if (crypto_op->sess_type == RTE_CRYPTO_OP_SESSIONLESS && + crypto_op->private_data_offset) { + m_data = (union rte_event_crypto_metadata *) + ((uint8_t *)crypto_op + + crypto_op->private_data_offset); + cdev_id = m_data->request_info.cdev_id; + qp_id = m_data->request_info.queue_pair_id; + qp_info = &adapter->cdevs[cdev_id].qpairs[qp_id]; + if (qp_info == NULL) { + rte_pktmbuf_free(crypto_op->sym->m_src); + rte_crypto_op_free(crypto_op); + continue; + } + len = qp_info->len; + qp_info->op_buffer[len] = crypto_op; + len++; + } else { + rte_pktmbuf_free(crypto_op->sym->m_src); + rte_crypto_op_free(crypto_op); + continue; + } + + if (len == BATCH_SIZE) { + struct rte_crypto_op **op_buffer = qp_info->op_buffer; + ret = rte_cryptodev_enqueue_burst(cdev_id, + qp_id, + op_buffer, + BATCH_SIZE); + + stats->crypto_enq_count += ret; + + while (ret < len) { + struct rte_crypto_op *op; + op = op_buffer[ret++]; + stats->crypto_enq_fail++; + rte_pktmbuf_free(op->sym->m_src); + rte_crypto_op_free(op); + } + + len = 0; + } + + if (qp_info) + qp_info->len = len; + n += ret; + } + + return n; +} + +static unsigned int +eca_crypto_enq_flush(struct rte_event_crypto_adapter *adapter) +{ + struct rte_event_crypto_adapter_stats *stats = &adapter->crypto_stats; + struct crypto_device_info *curr_dev; + struct crypto_queue_pair_info *curr_queue; + struct rte_crypto_op **op_buffer; + struct rte_cryptodev *dev; + uint8_t cdev_id; + uint16_t qp; + uint16_t ret; + uint16_t num_cdev = rte_cryptodev_count(); + + ret = 0; + for (cdev_id = 0; cdev_id < num_cdev; cdev_id++) { + curr_dev = &adapter->cdevs[cdev_id]; + if (curr_dev == NULL) + continue; + dev = curr_dev->dev; + + for (qp = 0; qp < dev->data->nb_queue_pairs; qp++) { + + curr_queue = &curr_dev->qpairs[qp]; + if (!curr_queue->qp_enabled) + continue; + + op_buffer = curr_queue->op_buffer; + ret = rte_cryptodev_enqueue_burst(cdev_id, + qp, + op_buffer, + curr_queue->len); + stats->crypto_enq_count += ret; + + while (ret < curr_queue->len) { + struct rte_crypto_op *op; + op = op_buffer[ret++]; + stats->crypto_enq_fail++; + rte_pktmbuf_free(op->sym->m_src); + rte_crypto_op_free(op); + } + curr_queue->len = 0; + } + } + + return ret; +} + +static int +eca_crypto_adapter_enq_run(struct rte_event_crypto_adapter *adapter, + unsigned int max_enq) +{ + struct rte_event_crypto_adapter_stats *stats = &adapter->crypto_stats; + struct rte_event ev[BATCH_SIZE]; + unsigned int nb_enq, nb_enqueued; + uint16_t n; + uint8_t event_dev_id = adapter->eventdev_id; + uint8_t event_port_id = adapter->event_port_id; + + nb_enqueued = 0; + if (adapter->mode == RTE_EVENT_CRYPTO_ADAPTER_OP_NEW) + return 0; + + for (nb_enq = 0; nb_enq < max_enq; nb_enq += n) { + stats->event_poll_count++; + n = rte_event_dequeue_burst(event_dev_id, + event_port_id, ev, BATCH_SIZE, 0); + + if (!n) + break; + + nb_enqueued += eca_enq_to_cryptodev(adapter, ev, n); + } + + if ((++adapter->transmit_loop_count & + (CRYPTO_ENQ_FLUSH_THRESHOLD - 1)) == 0) { + nb_enqueued += eca_crypto_enq_flush(adapter); + } + + return nb_enqueued; +} + +static inline void +eca_ops_enqueue_burst(struct rte_event_crypto_adapter *adapter, + struct rte_crypto_op **ops, uint16_t num) +{ + struct rte_event_crypto_adapter_stats *stats = &adapter->crypto_stats; + union rte_event_crypto_metadata *m_data = NULL; + uint8_t event_dev_id = adapter->eventdev_id; + uint8_t event_port_id = adapter->event_port_id; + struct rte_event events[BATCH_SIZE]; + uint16_t nb_enqueued, nb_ev; + uint8_t retry; + uint8_t i; + + nb_ev = 0; + retry = 0; + nb_enqueued = 0; + num = RTE_MIN(num, BATCH_SIZE); + for (i = 0; i < num; i++) { + struct rte_event *ev = &events[nb_ev++]; + if (ops[i]->sess_type == RTE_CRYPTO_OP_WITH_SESSION) { + m_data = rte_cryptodev_sym_session_get_user_data( + ops[i]->sym->session); + } else if (ops[i]->sess_type == RTE_CRYPTO_OP_SESSIONLESS && + ops[i]->private_data_offset) { + m_data = (union rte_event_crypto_metadata *) + ((uint8_t *)ops[i] + + ops[i]->private_data_offset); + } + + if (unlikely(m_data == NULL)) { + rte_pktmbuf_free(ops[i]->sym->m_src); + rte_crypto_op_free(ops[i]); + continue; + } + + rte_memcpy(ev, &m_data->response_info, sizeof(*ev)); + ev->event_ptr = ops[i]; + ev->event_type = RTE_EVENT_TYPE_CRYPTODEV; + if (adapter->implicit_release_disabled) + ev->op = RTE_EVENT_OP_FORWARD; + else + ev->op = RTE_EVENT_OP_NEW; + } + + do { + nb_enqueued += rte_event_enqueue_burst(event_dev_id, + event_port_id, + &events[nb_enqueued], + nb_ev - nb_enqueued); + } while (retry++ < CRYPTO_ADAPTER_MAX_EV_ENQ_RETRIES && + nb_enqueued < nb_ev); + + /* Free mbufs and rte_crypto_ops for failed events */ + for (i = nb_enqueued; i < nb_ev; i++) { + struct rte_crypto_op *op = events[i].event_ptr; + rte_pktmbuf_free(op->sym->m_src); + rte_crypto_op_free(op); + } + + stats->event_enq_fail_count += nb_ev - nb_enqueued; + stats->event_enq_count += nb_enqueued; + stats->event_enq_retry_count += retry - 1; +} + +static inline unsigned int +eca_crypto_adapter_deq_run(struct rte_event_crypto_adapter *adapter, + unsigned int max_deq) +{ + struct rte_event_crypto_adapter_stats *stats = &adapter->crypto_stats; + struct crypto_device_info *curr_dev; + struct crypto_queue_pair_info *curr_queue; + struct rte_crypto_op *ops[BATCH_SIZE]; + uint16_t n, nb_deq; + struct rte_cryptodev *dev; + uint8_t cdev_id; + uint16_t qp, dev_qps; + bool done; + uint16_t num_cdev = rte_cryptodev_count(); + + nb_deq = 0; + do { + uint16_t queues = 0; + done = true; + + for (cdev_id = adapter->next_cdev_id; + cdev_id < num_cdev; cdev_id++) { + curr_dev = &adapter->cdevs[cdev_id]; + if (curr_dev == NULL) + continue; + dev = curr_dev->dev; + dev_qps = dev->data->nb_queue_pairs; + + for (qp = curr_dev->next_queue_pair_id; + queues < dev_qps; qp = (qp + 1) % dev_qps, + queues++) { + + curr_queue = &curr_dev->qpairs[qp]; + if (!curr_queue->qp_enabled) + continue; + + n = rte_cryptodev_dequeue_burst(cdev_id, qp, + ops, BATCH_SIZE); + if (!n) + continue; + + done = false; + stats->crypto_deq_count += n; + eca_ops_enqueue_burst(adapter, ops, n); + nb_deq += n; + + if (nb_deq > max_deq) { + if ((qp + 1) == dev_qps) { + adapter->next_cdev_id = + (cdev_id + 1) + % num_cdev; + } + curr_dev->next_queue_pair_id = (qp + 1) + % dev->data->nb_queue_pairs; + + return nb_deq; + } + } + } + } while (done == false); + return nb_deq; +} + +static void +eca_crypto_adapter_run(struct rte_event_crypto_adapter *adapter, + unsigned int max_ops) +{ + while (max_ops) { + unsigned int e_cnt, d_cnt; + + e_cnt = eca_crypto_adapter_deq_run(adapter, max_ops); + max_ops -= RTE_MIN(max_ops, e_cnt); + + d_cnt = eca_crypto_adapter_enq_run(adapter, max_ops); + max_ops -= RTE_MIN(max_ops, d_cnt); + + if (e_cnt == 0 && d_cnt == 0) + break; + + } +} + +static int +eca_service_func(void *args) +{ + struct rte_event_crypto_adapter *adapter = args; + + if (rte_spinlock_trylock(&adapter->lock) == 0) + return 0; + eca_crypto_adapter_run(adapter, adapter->max_nb); + rte_spinlock_unlock(&adapter->lock); + + return 0; +} + +static int +eca_init_service(struct rte_event_crypto_adapter *adapter, uint8_t id) +{ + struct rte_event_crypto_adapter_conf adapter_conf; + struct rte_service_spec service; + int ret; + + if (adapter->service_inited) + return 0; + + memset(&service, 0, sizeof(service)); + snprintf(service.name, CRYPTO_ADAPTER_NAME_LEN, + "rte_event_crypto_adapter_%d", id); + service.socket_id = adapter->socket_id; + service.callback = eca_service_func; + service.callback_userdata = adapter; + /* Service function handles locking for queue add/del updates */ + service.capabilities = RTE_SERVICE_CAP_MT_SAFE; + ret = rte_service_component_register(&service, &adapter->service_id); + if (ret) { + RTE_EDEV_LOG_ERR("failed to register service %s err = %" PRId32, + service.name, ret); + return ret; + } + + ret = adapter->conf_cb(id, adapter->eventdev_id, + &adapter_conf, adapter->conf_arg); + if (ret) { + RTE_EDEV_LOG_ERR("configuration callback failed err = %" PRId32, + ret); + return ret; + } + + adapter->max_nb = adapter_conf.max_nb; + adapter->event_port_id = adapter_conf.event_port_id; + adapter->service_inited = 1; + + return ret; +} + +static void +eca_update_qp_info(struct rte_event_crypto_adapter *adapter, + struct crypto_device_info *dev_info, + int32_t queue_pair_id, + uint8_t add) +{ + struct crypto_queue_pair_info *qp_info; + int enabled; + uint16_t i; + + if (dev_info->qpairs == NULL) + return; + + if (queue_pair_id == -1) { + for (i = 0; i < dev_info->dev->data->nb_queue_pairs; i++) + eca_update_qp_info(adapter, dev_info, i, add); + } else { + qp_info = &dev_info->qpairs[queue_pair_id]; + enabled = qp_info->qp_enabled; + if (add) { + adapter->nb_qps += !enabled; + dev_info->num_qpairs += !enabled; + } else { + adapter->nb_qps -= enabled; + dev_info->num_qpairs -= enabled; + } + qp_info->qp_enabled = !!add; + } +} + +static int +eca_add_queue_pair(struct rte_event_crypto_adapter *adapter, + uint8_t cdev_id, + int queue_pair_id) +{ + struct crypto_device_info *dev_info = &adapter->cdevs[cdev_id]; + struct crypto_queue_pair_info *qpairs; + uint32_t i; + + if (dev_info->qpairs == NULL) { + dev_info->qpairs = + rte_zmalloc_socket(adapter->mem_name, + dev_info->dev->data->nb_queue_pairs * + sizeof(struct crypto_queue_pair_info), + 0, adapter->socket_id); + if (dev_info->qpairs == NULL) + return -ENOMEM; + + qpairs = dev_info->qpairs; + qpairs->op_buffer = rte_zmalloc_socket(adapter->mem_name, + BATCH_SIZE * + sizeof(struct rte_crypto_op *), + 0, adapter->socket_id); + if (!qpairs->op_buffer) { + rte_free(qpairs); + return -ENOMEM; + } + } + + if (queue_pair_id == -1) { + for (i = 0; i < dev_info->dev->data->nb_queue_pairs; i++) + eca_update_qp_info(adapter, dev_info, i, 1); + } else + eca_update_qp_info(adapter, dev_info, + (uint16_t)queue_pair_id, 1); + + return 0; +} + +int __rte_experimental +rte_event_crypto_adapter_queue_pair_add(uint8_t id, + uint8_t cdev_id, + int32_t queue_pair_id, + const struct rte_event *event) +{ + struct rte_event_crypto_adapter *adapter; + struct rte_eventdev *dev; + struct crypto_device_info *dev_info; + uint32_t cap; + int ret; + + EVENT_CRYPTO_ADAPTER_ID_VALID_OR_ERR_RET(id, -EINVAL); + + if (!rte_cryptodev_pmd_is_valid_dev(cdev_id)) { + RTE_EDEV_LOG_ERR("Invalid dev_id=%" PRIu8, cdev_id); + return -EINVAL; + } + + adapter = eca_id_to_adapter(id); + if (adapter == NULL) + return -EINVAL; + + dev = &rte_eventdevs[adapter->eventdev_id]; + ret = rte_event_crypto_adapter_caps_get(adapter->eventdev_id, + cdev_id, + &cap); + if (ret) { + RTE_EDEV_LOG_ERR("Failed to get adapter caps dev %" PRIu8 + " cdev %" PRIu8, id, cdev_id); + return ret; + } + + if ((cap & RTE_EVENT_CRYPTO_ADAPTER_CAP_INTERNAL_PORT_QP_EV_BIND) && + (event == NULL)) { + RTE_EDEV_LOG_ERR("Conf value can not be NULL for dev_id=%u", + cdev_id); + return -EINVAL; + } + + dev_info = &adapter->cdevs[cdev_id]; + + if (queue_pair_id != -1 && + (uint16_t)queue_pair_id >= dev_info->dev->data->nb_queue_pairs) { + RTE_EDEV_LOG_ERR("Invalid queue_pair_id %" PRIu16, + (uint16_t)queue_pair_id); + return -EINVAL; + } + + /* In case HW cap is RTE_EVENT_CRYPTO_ADAPTER_CAP_INTERNAL_PORT_OP_FWD, + * no need of service core as HW supports event forward capability. + */ + if ((cap & RTE_EVENT_CRYPTO_ADAPTER_CAP_INTERNAL_PORT_OP_FWD) || + (cap & RTE_EVENT_CRYPTO_ADAPTER_CAP_INTERNAL_PORT_QP_EV_BIND && + adapter->mode == RTE_EVENT_CRYPTO_ADAPTER_OP_NEW) || + (cap & RTE_EVENT_CRYPTO_ADAPTER_CAP_INTERNAL_PORT_OP_NEW && + adapter->mode == RTE_EVENT_CRYPTO_ADAPTER_OP_NEW)) { + RTE_FUNC_PTR_OR_ERR_RET( + *dev->dev_ops->crypto_adapter_queue_pair_add, + -ENOTSUP); + if (dev_info->qpairs == NULL) { + dev_info->qpairs = + rte_zmalloc_socket(adapter->mem_name, + dev_info->dev->data->nb_queue_pairs * + sizeof(struct crypto_queue_pair_info), + 0, adapter->socket_id); + if (dev_info->qpairs == NULL) + return -ENOMEM; + } + + ret = (*dev->dev_ops->crypto_adapter_queue_pair_add)(dev, + dev_info->dev, + queue_pair_id, + event); + if (ret) + return ret; + + else + eca_update_qp_info(adapter, &adapter->cdevs[cdev_id], + queue_pair_id, 1); + } + + /* In case HW cap is RTE_EVENT_CRYPTO_ADAPTER_CAP_INTERNAL_PORT_OP_NEW, + * or SW adapter, initiate services so the application can choose + * which ever way it wants to use the adapter. + * Case 1: RTE_EVENT_CRYPTO_ADAPTER_CAP_INTERNAL_PORT_OP_NEW + * Application may wants to use one of below two mode + * a. OP_FORWARD mode -> HW Dequeue + SW enqueue + * b. OP_NEW mode -> HW Dequeue + * Case 2: No HW caps, use SW adapter + * a. OP_FORWARD mode -> SW enqueue & dequeue + * b. OP_NEW mode -> SW Dequeue + */ + if ((cap & RTE_EVENT_CRYPTO_ADAPTER_CAP_INTERNAL_PORT_OP_NEW && + adapter->mode == RTE_EVENT_CRYPTO_ADAPTER_OP_FORWARD) || + (!(cap & RTE_EVENT_CRYPTO_ADAPTER_CAP_INTERNAL_PORT_OP_NEW) && + !(cap & RTE_EVENT_CRYPTO_ADAPTER_CAP_INTERNAL_PORT_OP_FWD) && + !(cap & RTE_EVENT_CRYPTO_ADAPTER_CAP_INTERNAL_PORT_QP_EV_BIND) && + (cap & RTE_EVENT_CRYPTO_ADAPTER_CAP_SESSION_PRIVATE_DATA))) { + rte_spinlock_lock(&adapter->lock); + ret = eca_init_service(adapter, id); + if (ret == 0) + ret = eca_add_queue_pair(adapter, cdev_id, + queue_pair_id); + rte_spinlock_unlock(&adapter->lock); + + if (ret) + return ret; + + rte_service_component_runstate_set(adapter->service_id, 1); + } + + return 0; +} + +int __rte_experimental +rte_event_crypto_adapter_queue_pair_del(uint8_t id, uint8_t cdev_id, + int32_t queue_pair_id) +{ + struct rte_event_crypto_adapter *adapter; + struct crypto_device_info *dev_info; + struct rte_eventdev *dev; + int ret; + uint32_t cap; + uint16_t i; + + EVENT_CRYPTO_ADAPTER_ID_VALID_OR_ERR_RET(id, -EINVAL); + + if (!rte_cryptodev_pmd_is_valid_dev(cdev_id)) { + RTE_EDEV_LOG_ERR("Invalid dev_id=%" PRIu8, cdev_id); + return -EINVAL; + } + + adapter = eca_id_to_adapter(id); + if (adapter == NULL) + return -EINVAL; + + dev = &rte_eventdevs[adapter->eventdev_id]; + ret = rte_event_crypto_adapter_caps_get(adapter->eventdev_id, + cdev_id, + &cap); + if (ret) + return ret; + + dev_info = &adapter->cdevs[cdev_id]; + + if (queue_pair_id != -1 && + (uint16_t)queue_pair_id >= dev_info->dev->data->nb_queue_pairs) { + RTE_EDEV_LOG_ERR("Invalid queue_pair_id %" PRIu16, + (uint16_t)queue_pair_id); + return -EINVAL; + } + + if ((cap & RTE_EVENT_CRYPTO_ADAPTER_CAP_INTERNAL_PORT_OP_FWD) || + (cap & RTE_EVENT_CRYPTO_ADAPTER_CAP_INTERNAL_PORT_OP_NEW && + adapter->mode == RTE_EVENT_CRYPTO_ADAPTER_OP_NEW)) { + RTE_FUNC_PTR_OR_ERR_RET( + *dev->dev_ops->crypto_adapter_queue_pair_del, + -ENOTSUP); + ret = (*dev->dev_ops->crypto_adapter_queue_pair_del)(dev, + dev_info->dev, + queue_pair_id); + if (ret == 0) { + eca_update_qp_info(adapter, + &adapter->cdevs[cdev_id], + queue_pair_id, + 0); + if (dev_info->num_qpairs == 0) { + rte_free(dev_info->qpairs); + dev_info->qpairs = NULL; + } + } + } else { + if (adapter->nb_qps == 0) + return 0; + + rte_spinlock_lock(&adapter->lock); + if (queue_pair_id == -1) { + for (i = 0; i < dev_info->dev->data->nb_queue_pairs; + i++) + eca_update_qp_info(adapter, dev_info, + queue_pair_id, 0); + } else { + eca_update_qp_info(adapter, dev_info, + (uint16_t)queue_pair_id, 0); + } + + if (dev_info->num_qpairs == 0) { + rte_free(dev_info->qpairs); + dev_info->qpairs = NULL; + } + + rte_spinlock_unlock(&adapter->lock); + rte_service_component_runstate_set(adapter->service_id, + adapter->nb_qps); + } + + return ret; +} + +static int +eca_adapter_ctrl(uint8_t id, int start) +{ + struct rte_event_crypto_adapter *adapter; + struct crypto_device_info *dev_info; + struct rte_eventdev *dev; + uint32_t i; + int use_service; + int stop = !start; + + use_service = 0; + EVENT_CRYPTO_ADAPTER_ID_VALID_OR_ERR_RET(id, -EINVAL); + adapter = eca_id_to_adapter(id); + if (adapter == NULL) + return -EINVAL; + + dev = &rte_eventdevs[adapter->eventdev_id]; + + for (i = 0; i < rte_cryptodev_count(); i++) { + dev_info = &adapter->cdevs[i]; + /* if start check for num queue pairs */ + if (start && !dev_info->num_qpairs) + continue; + /* if stop check if dev has been started */ + if (stop && !dev_info->dev_started) + continue; + use_service |= !dev_info->internal_event_port; + dev_info->dev_started = start; + if (dev_info->internal_event_port == 0) + continue; + start ? (*dev->dev_ops->crypto_adapter_start)(dev, + &dev_info->dev[i]) : + (*dev->dev_ops->crypto_adapter_stop)(dev, + &dev_info->dev[i]); + } + + if (use_service) + rte_service_runstate_set(adapter->service_id, start); + + return 0; +} + +int __rte_experimental +rte_event_crypto_adapter_start(uint8_t id) +{ + struct rte_event_crypto_adapter *adapter; + + EVENT_CRYPTO_ADAPTER_ID_VALID_OR_ERR_RET(id, -EINVAL); + adapter = eca_id_to_adapter(id); + if (adapter == NULL) + return -EINVAL; + + return eca_adapter_ctrl(id, 1); +} + +int __rte_experimental +rte_event_crypto_adapter_stop(uint8_t id) +{ + return eca_adapter_ctrl(id, 0); +} + +int __rte_experimental +rte_event_crypto_adapter_stats_get(uint8_t id, + struct rte_event_crypto_adapter_stats *stats) +{ + struct rte_event_crypto_adapter *adapter; + struct rte_event_crypto_adapter_stats dev_stats_sum = { 0 }; + struct rte_event_crypto_adapter_stats dev_stats; + struct rte_eventdev *dev; + struct crypto_device_info *dev_info; + uint32_t i; + int ret; + + EVENT_CRYPTO_ADAPTER_ID_VALID_OR_ERR_RET(id, -EINVAL); + + adapter = eca_id_to_adapter(id); + if (adapter == NULL || stats == NULL) + return -EINVAL; + + dev = &rte_eventdevs[adapter->eventdev_id]; + memset(stats, 0, sizeof(*stats)); + for (i = 0; i < rte_cryptodev_count(); i++) { + dev_info = &adapter->cdevs[i]; + if (dev_info->internal_event_port == 0 || + dev->dev_ops->crypto_adapter_stats_get == NULL) + continue; + ret = (*dev->dev_ops->crypto_adapter_stats_get)(dev, + dev_info->dev, + &dev_stats); + if (ret) + continue; + + dev_stats_sum.crypto_deq_count += dev_stats.crypto_deq_count; + dev_stats_sum.event_enq_count += + dev_stats.event_enq_count; + } + + if (adapter->service_inited) + *stats = adapter->crypto_stats; + + stats->crypto_deq_count += dev_stats_sum.crypto_deq_count; + stats->event_enq_count += dev_stats_sum.event_enq_count; + + return 0; +} + +int __rte_experimental +rte_event_crypto_adapter_stats_reset(uint8_t id) +{ + struct rte_event_crypto_adapter *adapter; + struct crypto_device_info *dev_info; + struct rte_eventdev *dev; + uint32_t i; + + EVENT_CRYPTO_ADAPTER_ID_VALID_OR_ERR_RET(id, -EINVAL); + + adapter = eca_id_to_adapter(id); + if (adapter == NULL) + return -EINVAL; + + dev = &rte_eventdevs[adapter->eventdev_id]; + for (i = 0; i < rte_cryptodev_count(); i++) { + dev_info = &adapter->cdevs[i]; + if (dev_info->internal_event_port == 0 || + dev->dev_ops->crypto_adapter_stats_reset == NULL) + continue; + (*dev->dev_ops->crypto_adapter_stats_reset)(dev, + dev_info->dev); + } + + memset(&adapter->crypto_stats, 0, sizeof(adapter->crypto_stats)); + return 0; +} + +int __rte_experimental +rte_event_crypto_adapter_service_id_get(uint8_t id, uint32_t *service_id) +{ + struct rte_event_crypto_adapter *adapter; + + EVENT_CRYPTO_ADAPTER_ID_VALID_OR_ERR_RET(id, -EINVAL); + + adapter = eca_id_to_adapter(id); + if (adapter == NULL || service_id == NULL) + return -EINVAL; + + if (adapter->service_inited) + *service_id = adapter->service_id; + + return adapter->service_inited ? 0 : -ESRCH; +} + +int __rte_experimental +rte_event_crypto_adapter_event_port_get(uint8_t id, uint8_t *event_port_id) +{ + struct rte_event_crypto_adapter *adapter; + + EVENT_CRYPTO_ADAPTER_ID_VALID_OR_ERR_RET(id, -EINVAL); + + adapter = eca_id_to_adapter(id); + if (adapter == NULL || event_port_id == NULL) + return -EINVAL; + + *event_port_id = adapter->event_port_id; + + return 0; +} diff --git a/lib/librte_eventdev/rte_event_crypto_adapter.h b/lib/librte_eventdev/rte_event_crypto_adapter.h new file mode 100644 index 00000000..d367309c --- /dev/null +++ b/lib/librte_eventdev/rte_event_crypto_adapter.h @@ -0,0 +1,575 @@ +/* SPDX-License-Identifier: BSD-3-Clause + * Copyright(c) 2018 Intel Corporation. + * All rights reserved. + */ + +#ifndef _RTE_EVENT_CRYPTO_ADAPTER_ +#define _RTE_EVENT_CRYPTO_ADAPTER_ + +/** + * @file + * + * RTE Event crypto adapter + * + * Eventdev library provides couple of adapters to bridge between various + * components for providing new event source. The event crypto adapter is + * one of those adapters which is intended to bridge between event devices + * and crypto devices. + * + * The crypto adapter adds support to enqueue/dequeue crypto operations to/ + * from event device. The packet flow between crypto device and the event + * device can be accomplished using both SW and HW based transfer mechanisms. + * The adapter uses an EAL service core function for SW based packet transfer + * and uses the eventdev PMD functions to configure HW based packet transfer + * between the crypto device and the event device. + * + * The application can choose to submit a crypto operation directly to + * crypto device or send it to the crypto adapter via eventdev based on + * RTE_EVENT_CRYPTO_ADAPTER_CAP_INTERNAL_PORT_OP_FWD capability. + * The first mode is known as the event new(RTE_EVENT_CRYPTO_ADAPTER_OP_NEW) + * mode and the second as the event forward(RTE_EVENT_CRYPTO_ADAPTER_OP_FORWARD) + * mode. The choice of mode can be specified while creating the adapter. + * In the former mode, it is an application responsibility to enable ingress + * packet ordering. In the latter mode, it is the adapter responsibility to + * enable the ingress packet ordering. + * + * + * Working model of RTE_EVENT_CRYPTO_ADAPTER_OP_NEW mode: + * + * +--------------+ +--------------+ + * | | | Crypto stage | + * | Application |---[2]-->| + enqueue to | + * | | | cryptodev | + * +--------------+ +--------------+ + * ^ ^ | + * | | [3] + * [6] [1] | + * | | | + * +--------------+ | + * | | | + * | Event device | | + * | | | + * +--------------+ | + * ^ | + * | | + * [5] | + * | v + * +--------------+ +--------------+ + * | | | | + * |Crypto adapter|<--[4]---| Cryptodev | + * | | | | + * +--------------+ +--------------+ + * + * + * [1] Application dequeues events from the previous stage. + * [2] Application prepares the crypto operations. + * [3] Crypto operations are submitted to cryptodev by application. + * [4] Crypto adapter dequeues crypto completions from cryptodev. + * [5] Crypto adapter enqueues events to the eventdev. + * [6] Application dequeues from eventdev and prepare for further + * processing. + * + * In the RTE_EVENT_CRYPTO_ADAPTER_OP_NEW mode, application submits crypto + * operations directly to crypto device. The adapter then dequeues crypto + * completions from crypto device and enqueue events to the event device. + * This mode does not ensure ingress ordering, if the application directly + * enqueues to cryptodev without going through crypto/atomic stage i.e. + * removing item [1] and [2]. + * Events dequeued from the adapter will be treated as new events. + * In this mode, application needs to specify event information (response + * information) which is needed to enqueue an event after the crypto operation + * is completed. + * + * + * Working model of RTE_EVENT_CRYPTO_ADAPTER_OP_FORWARD mode: + * + * +--------------+ +--------------+ + * --[1]-->| |---[2]-->| Application | + * | Event device | | in | + * <--[8]--| |<--[3]---| Ordered stage| + * +--------------+ +--------------+ + * ^ | + * | [4] + * [7] | + * | v + * +----------------+ +--------------+ + * | |--[5]->| | + * | Crypto adapter | | Cryptodev | + * | |<-[6]--| | + * +----------------+ +--------------+ + * + * + * [1] Events from the previous stage. + * [2] Application in ordered stage dequeues events from eventdev. + * [3] Application enqueues crypto operations as events to eventdev. + * [4] Crypto adapter dequeues event from eventdev. + * [5] Crypto adapter submits crypto operations to cryptodev + * (Atomic stage). + * [6] Crypto adapter dequeues crypto completions from cryptodev + * [7] Crypto adapter enqueues events to the eventdev + * [8] Events to the next stage + * + * In the RTE_EVENT_CRYPTO_ADAPTER_OP_FORWARD mode, if HW supports + * RTE_EVENT_CRYPTO_ADAPTER_CAP_INTERNAL_PORT_OP_FWD capability the application + * can directly submit the crypto operations to the cryptodev. + * If not, application retrieves crypto adapter's event port using + * rte_event_crypto_adapter_event_port_get() API. Then, links its event + * queue to this port and starts enqueuing crypto operations as events + * to the eventdev. The adapter then dequeues the events and submits the + * crypto operations to the cryptodev. After the crypto completions, the + * adapter enqueues events to the event device. + * Application can use this mode, when ingress packet ordering is needed. + * Events dequeued from the adapter will be treated as forwarded events. + * In this mode, the application needs to specify the cryptodev ID + * and queue pair ID (request information) needed to enqueue a crypto + * operation in addition to the event information (response information) + * needed to enqueue an event after the crypto operation has completed. + * + * + * The event crypto adapter provides common APIs to configure the packet flow + * from the crypto device to event devices for both SW and HW based transfers. + * The crypto event adapter's functions are: + * - rte_event_crypto_adapter_create_ext() + * - rte_event_crypto_adapter_create() + * - rte_event_crypto_adapter_free() + * - rte_event_crypto_adapter_queue_pair_add() + * - rte_event_crypto_adapter_queue_pair_del() + * - rte_event_crypto_adapter_start() + * - rte_event_crypto_adapter_stop() + * - rte_event_crypto_adapter_stats_get() + * - rte_event_crypto_adapter_stats_reset() + + * The applicaton creates an instance using rte_event_crypto_adapter_create() + * or rte_event_crypto_adapter_create_ext(). + * + * Cryptodev queue pair addition/deletion is done using the + * rte_event_crypto_adapter_queue_pair_xxx() APIs. If HW supports + * RTE_EVENT_CRYPTO_ADAPTER_CAP_INTERNAL_PORT_QP_EV_BIND capability, event + * information must be passed to the add API. + * + * The SW adapter or HW PMD uses rte_crypto_op::sess_type to decide whether + * request/response(private) data is located in the crypto/security session + * or at an offset in the rte_crypto_op. + * + * For session-based operations, the set and get API provides a mechanism for + * an application to store and retrieve the data information stored + * along with the crypto session. + * The RTE_EVENT_CRYPTO_ADAPTER_CAP_SESSION_PRIVATE_DATA capability indicates + * whether HW or SW supports this feature. + * + * For session-less mode, the adapter gets the private data information placed + * along with the ``struct rte_crypto_op``. + * The rte_crypto_op::private_data_offset provides an offset to locate the + * request/response information in the rte_crypto_op. This offset is counted + * from the start of the rte_crypto_op including initialization vector (IV). + */ + +#ifdef __cplusplus +extern "C" { +#endif + +#include <stdint.h> + +#include "rte_eventdev.h" + +/** + * @warning + * @b EXPERIMENTAL: this enum may change without prior notice + * + * Crypto event adapter mode + */ +enum rte_event_crypto_adapter_mode { + RTE_EVENT_CRYPTO_ADAPTER_OP_NEW, + /**< Start the crypto adapter in event new mode. + * @see RTE_EVENT_OP_NEW. + * Application submits crypto operations to the cryptodev. + * Adapter only dequeues the crypto completions from cryptodev + * and enqueue events to the eventdev. + */ + RTE_EVENT_CRYPTO_ADAPTER_OP_FORWARD, + /**< Start the crypto adapter in event forward mode. + * @see RTE_EVENT_OP_FORWARD. + * Application submits crypto requests as events to the crypto + * adapter or crypto device based on + * RTE_EVENT_CRYPTO_ADAPTER_CAP_INTERNAL_PORT_OP_FWD capability. + * Crypto completions are enqueued back to the eventdev by + * crypto adapter. + */ +}; + +/** + * @warning + * @b EXPERIMENTAL: this structure may change without prior notice + * + * Crypto event request structure will be filled by application to + * provide event request information to the adapter. + */ +struct rte_event_crypto_request { + uint8_t resv[8]; + /**< Overlaps with first 8 bytes of struct rte_event + * that encode the response event information. Application + * is expected to fill in struct rte_event response_info. + */ + uint16_t cdev_id; + /**< cryptodev ID to be used */ + uint16_t queue_pair_id; + /**< cryptodev queue pair ID to be used */ + uint32_t resv1; + /**< Reserved bits */ +}; + +/** + * @warning + * @b EXPERIMENTAL: this structure may change without prior notice + * + * Crypto event metadata structure will be filled by application + * to provide crypto request and event response information. + * + * If crypto events are enqueued using a HW mechanism, the cryptodev + * PMD will use the event response information to set up the event + * that is enqueued back to eventdev after completion of the crypto + * operation. If the transfer is done by SW, event response information + * will be used by the adapter. + */ +union rte_event_crypto_metadata { + struct rte_event_crypto_request request_info; + /**< Request information to be filled in by application + * for RTE_EVENT_CRYPTO_ADAPTER_OP_FORWARD mode. + */ + struct rte_event response_info; + /**< Response information to be filled in by application + * for RTE_EVENT_CRYPTO_ADAPTER_OP_NEW and + * RTE_EVENT_CRYPTO_ADAPTER_OP_FORWARD mode. + */ +}; + +/** + * @warning + * @b EXPERIMENTAL: this structure may change without prior notice + * + * Adapter configuration structure that the adapter configuration callback + * function is expected to fill out + * @see rte_event_crypto_adapter_conf_cb + */ +struct rte_event_crypto_adapter_conf { + uint8_t event_port_id; + /**< Event port identifier, the adapter enqueues events to this + * port and dequeues crypto request events in + * RTE_EVENT_CRYPTO_ADAPTER_OP_FORWARD mode. + */ + uint32_t max_nb; + /**< The adapter can return early if it has processed at least + * max_nb crypto ops. This isn't treated as a requirement; batching + * may cause the adapter to process more than max_nb crypto ops. + */ +}; + +/** + * @warning + * @b EXPERIMENTAL: this API may change without prior notice + * + * Function type used for adapter configuration callback. The callback is + * used to fill in members of the struct rte_event_crypto_adapter_conf, this + * callback is invoked when creating a SW service for packet transfer from + * cryptodev queue pair to the event device. The SW service is created within + * the rte_event_crypto_adapter_queue_pair_add() function if SW based packet + * transfers from cryptodev queue pair to the event device are required. + * + * @param id + * Adapter identifier. + * + * @param dev_id + * Event device identifier. + * + * @param conf + * Structure that needs to be populated by this callback. + * + * @param arg + * Argument to the callback. This is the same as the conf_arg passed to the + * rte_event_crypto_adapter_create_ext(). + */ +typedef int (*rte_event_crypto_adapter_conf_cb) (uint8_t id, uint8_t dev_id, + struct rte_event_crypto_adapter_conf *conf, + void *arg); + +/** + * @warning + * @b EXPERIMENTAL: this structure may change without prior notice + * + * A structure used to retrieve statistics for an event crypto adapter + * instance. + */ + +struct rte_event_crypto_adapter_stats { + uint64_t event_poll_count; + /**< Event port poll count */ + uint64_t event_deq_count; + /**< Event dequeue count */ + uint64_t crypto_enq_count; + /**< Cryptodev enqueue count */ + uint64_t crypto_enq_fail; + /**< Cryptodev enqueue failed count */ + uint64_t crypto_deq_count; + /**< Cryptodev dequeue count */ + uint64_t event_enq_count; + /**< Event enqueue count */ + uint64_t event_enq_retry_count; + /**< Event enqueue retry count */ + uint64_t event_enq_fail_count; + /**< Event enqueue fail count */ +}; + +/** + * @warning + * @b EXPERIMENTAL: this API may change without prior notice + * + * Create a new event crypto adapter with the specified identifier. + * + * @param id + * Adapter identifier. + * + * @param dev_id + * Event device identifier. + * + * @param conf_cb + * Callback function that fills in members of a + * struct rte_event_crypto_adapter_conf struct passed into + * it. + * + * @param mode + * Flag to indicate the mode of the adapter. + * @see rte_event_crypto_adapter_mode + * + * @param conf_arg + * Argument that is passed to the conf_cb function. + * + * @return + * - 0: Success + * - <0: Error code on failure + */ +int __rte_experimental +rte_event_crypto_adapter_create_ext(uint8_t id, uint8_t dev_id, + rte_event_crypto_adapter_conf_cb conf_cb, + enum rte_event_crypto_adapter_mode mode, + void *conf_arg); + +/** + * @warning + * @b EXPERIMENTAL: this API may change without prior notice + * + * Create a new event crypto adapter with the specified identifier. + * This function uses an internal configuration function that creates an event + * port. This default function reconfigures the event device with an + * additional event port and set up the event port using the port_config + * parameter passed into this function. In case the application needs more + * control in configuration of the service, it should use the + * rte_event_crypto_adapter_create_ext() version. + * + * @param id + * Adapter identifier. + * + * @param dev_id + * Event device identifier. + * + * @param port_config + * Argument of type *rte_event_port_conf* that is passed to the conf_cb + * function. + * + * @param mode + * Flag to indicate the mode of the adapter. + * @see rte_event_crypto_adapter_mode + * + * @return + * - 0: Success + * - <0: Error code on failure + */ +int __rte_experimental +rte_event_crypto_adapter_create(uint8_t id, uint8_t dev_id, + struct rte_event_port_conf *port_config, + enum rte_event_crypto_adapter_mode mode); + +/** + * @warning + * @b EXPERIMENTAL: this API may change without prior notice + * + * Free an event crypto adapter + * + * @param id + * Adapter identifier. + * + * @return + * - 0: Success + * - <0: Error code on failure, If the adapter still has queue pairs + * added to it, the function returns -EBUSY. + */ +int __rte_experimental +rte_event_crypto_adapter_free(uint8_t id); + +/** + * @warning + * @b EXPERIMENTAL: this API may change without prior notice + * + * Add a queue pair to an event crypto adapter. + * + * @param id + * Adapter identifier. + * + * @param cdev_id + * Cryptodev identifier. + * + * @param queue_pair_id + * Cryptodev queue pair identifier. If queue_pair_id is set -1, + * adapter adds all the pre configured queue pairs to the instance. + * + * @param event + * if HW supports cryptodev queue pair to event queue binding, application is + * expected to fill in event information, else it will be NULL. + * @see RTE_EVENT_CRYPTO_ADAPTER_CAP_INTERNAL_PORT_QP_EV_BIND + * + * @return + * - 0: Success, queue pair added correctly. + * - <0: Error code on failure. + */ +int __rte_experimental +rte_event_crypto_adapter_queue_pair_add(uint8_t id, + uint8_t cdev_id, + int32_t queue_pair_id, + const struct rte_event *event); + +/** + * @warning + * @b EXPERIMENTAL: this API may change without prior notice + * + * Delete a queue pair from an event crypto adapter. + * + * @param id + * Adapter identifier. + * + * @param cdev_id + * Cryptodev identifier. + * + * @param queue_pair_id + * Cryptodev queue pair identifier. + * + * @return + * - 0: Success, queue pair deleted successfully. + * - <0: Error code on failure. + */ +int __rte_experimental +rte_event_crypto_adapter_queue_pair_del(uint8_t id, uint8_t cdev_id, + int32_t queue_pair_id); + +/** + * @warning + * @b EXPERIMENTAL: this API may change without prior notice + * + * Start event crypto adapter + * + * @param id + * Adapter identifier. + * + * + * @return + * - 0: Success, adapter started successfully. + * - <0: Error code on failure. + */ +int __rte_experimental +rte_event_crypto_adapter_start(uint8_t id); + +/** + * @warning + * @b EXPERIMENTAL: this API may change without prior notice + * + * Stop event crypto adapter + * + * @param id + * Adapter identifier. + * + * @return + * - 0: Success, adapter stopped successfully. + * - <0: Error code on failure. + */ +int __rte_experimental +rte_event_crypto_adapter_stop(uint8_t id); + +/** + * @warning + * @b EXPERIMENTAL: this API may change without prior notice + * + * Retrieve statistics for an adapter + * + * @param id + * Adapter identifier. + * + * @param [out] stats + * A pointer to structure used to retrieve statistics for an adapter. + * + * @return + * - 0: Success, retrieved successfully. + * - <0: Error code on failure. + */ +int __rte_experimental +rte_event_crypto_adapter_stats_get(uint8_t id, + struct rte_event_crypto_adapter_stats *stats); + +/** + * @warning + * @b EXPERIMENTAL: this API may change without prior notice + * + * Reset statistics for an adapter. + * + * @param id + * Adapter identifier. + * + * @return + * - 0: Success, statistics reset successfully. + * - <0: Error code on failure. + */ +int __rte_experimental +rte_event_crypto_adapter_stats_reset(uint8_t id); + +/** + * @warning + * @b EXPERIMENTAL: this API may change without prior notice + * + * Retrieve the service ID of an adapter. If the adapter doesn't use + * a rte_service function, this function returns -ESRCH. + * + * @param id + * Adapter identifier. + * + * @param [out] service_id + * A pointer to a uint32_t, to be filled in with the service id. + * + * @return + * - 0: Success + * - <0: Error code on failure, if the adapter doesn't use a rte_service + * function, this function returns -ESRCH. + */ +int __rte_experimental +rte_event_crypto_adapter_service_id_get(uint8_t id, uint32_t *service_id); + +/** + * @warning + * @b EXPERIMENTAL: this API may change without prior notice + * + * Retrieve the event port of an adapter. + * + * @param id + * Adapter identifier. + * + * @param [out] event_port_id + * Application links its event queue to this adapter port which is used + * in RTE_EVENT_CRYPTO_ADAPTER_OP_FORWARD mode. + * + * @return + * - 0: Success + * - <0: Error code on failure. + */ +int __rte_experimental +rte_event_crypto_adapter_event_port_get(uint8_t id, uint8_t *event_port_id); + +#ifdef __cplusplus +} +#endif +#endif /* _RTE_EVENT_CRYPTO_ADAPTER_ */ diff --git a/lib/librte_eventdev/rte_event_eth_rx_adapter.c b/lib/librte_eventdev/rte_event_eth_rx_adapter.c index 9aece9f8..f5e5a0b5 100644 --- a/lib/librte_eventdev/rte_event_eth_rx_adapter.c +++ b/lib/librte_eventdev/rte_event_eth_rx_adapter.c @@ -1,3 +1,12 @@ +/* SPDX-License-Identifier: BSD-3-Clause + * Copyright(c) 2017 Intel Corporation. + * All rights reserved. + */ +#if defined(LINUX) +#include <sys/epoll.h> +#endif +#include <unistd.h> + #include <rte_cycles.h> #include <rte_common.h> #include <rte_dev.h> @@ -7,6 +16,7 @@ #include <rte_malloc.h> #include <rte_service_component.h> #include <rte_thash.h> +#include <rte_interrupts.h> #include "rte_eventdev.h" #include "rte_eventdev_pmd.h" @@ -20,6 +30,22 @@ #define ETH_RX_ADAPTER_MEM_NAME_LEN 32 #define RSS_KEY_SIZE 40 +/* value written to intr thread pipe to signal thread exit */ +#define ETH_BRIDGE_INTR_THREAD_EXIT 1 +/* Sentinel value to detect initialized file handle */ +#define INIT_FD -1 + +/* + * Used to store port and queue ID of interrupting Rx queue + */ +union queue_data { + RTE_STD_C11 + void *ptr; + struct { + uint16_t port; + uint16_t queue; + }; +}; /* * There is an instance of this struct per polled Rx queue added to the @@ -27,7 +53,7 @@ */ struct eth_rx_poll_entry { /* Eth port to poll */ - uint8_t eth_dev_id; + uint16_t eth_dev_id; /* Eth rx queue to poll */ uint16_t eth_rx_qid; }; @@ -71,6 +97,30 @@ struct rte_event_eth_rx_adapter { uint16_t enq_block_count; /* Block start ts */ uint64_t rx_enq_block_start_ts; + /* epoll fd used to wait for Rx interrupts */ + int epd; + /* Num of interrupt driven interrupt queues */ + uint32_t num_rx_intr; + /* Used to send <dev id, queue id> of interrupting Rx queues from + * the interrupt thread to the Rx thread + */ + struct rte_ring *intr_ring; + /* Rx Queue data (dev id, queue id) for the last non-empty + * queue polled + */ + union queue_data qd; + /* queue_data is valid */ + int qd_valid; + /* Interrupt ring lock, synchronizes Rx thread + * and interrupt thread + */ + rte_spinlock_t intr_ring_lock; + /* event array passed to rte_poll_wait */ + struct rte_epoll_event *epoll_events; + /* Count of interrupt vectors in use */ + uint32_t num_intr_vec; + /* Thread blocked on Rx interrupts */ + pthread_t rx_intr_thread; /* Configuration callback for rte_service configuration */ rte_event_eth_rx_adapter_conf_cb conf_cb; /* Configuration callback argument */ @@ -87,12 +137,20 @@ struct rte_event_eth_rx_adapter { int socket_id; /* Per adapter EAL service */ uint32_t service_id; + /* Adapter started flag */ + uint8_t rxa_started; + /* Adapter ID */ + uint8_t id; } __rte_cache_aligned; /* Per eth device */ struct eth_device_info { struct rte_eth_dev *dev; struct eth_rx_queue_info *rx_queue; + /* Rx callback */ + rte_event_eth_rx_adapter_cb_fn cb_fn; + /* Rx callback argument */ + void *cb_arg; /* Set if ethdev->eventdev packet transfer uses a * hardware mechanism */ @@ -103,15 +161,42 @@ struct eth_device_info { * rx_adapter_stop callback needs to be invoked */ uint8_t dev_rx_started; - /* If nb_dev_queues > 0, the start callback will + /* Number of queues added for this device */ + uint16_t nb_dev_queues; + /* Number of poll based queues + * If nb_rx_poll > 0, the start callback will * be invoked if not already invoked */ - uint16_t nb_dev_queues; + uint16_t nb_rx_poll; + /* Number of interrupt based queues + * If nb_rx_intr > 0, the start callback will + * be invoked if not already invoked. + */ + uint16_t nb_rx_intr; + /* Number of queues that use the shared interrupt */ + uint16_t nb_shared_intr; + /* sum(wrr(q)) for all queues within the device + * useful when deleting all device queues + */ + uint32_t wrr_len; + /* Intr based queue index to start polling from, this is used + * if the number of shared interrupts is non-zero + */ + uint16_t next_q_idx; + /* Intr based queue indices */ + uint16_t *intr_queue; + /* device generates per Rx queue interrupt for queue index + * for queue indices < RTE_MAX_RXTX_INTR_VEC_ID - 1 + */ + int multi_intr_cap; + /* shared interrupt enabled */ + int shared_intr_enabled; }; /* Per Rx queue */ struct eth_rx_queue_info { int queue_enabled; /* True if added */ + int intr_enabled; uint16_t wt; /* Polling weight */ uint8_t event_queue_id; /* Event queue to enqueue packets to */ uint8_t sched_type; /* Sched type for events */ @@ -123,30 +208,30 @@ struct eth_rx_queue_info { static struct rte_event_eth_rx_adapter **event_eth_rx_adapter; static inline int -valid_id(uint8_t id) +rxa_validate_id(uint8_t id) { return id < RTE_EVENT_ETH_RX_ADAPTER_MAX_INSTANCE; } #define RTE_EVENT_ETH_RX_ADAPTER_ID_VALID_OR_ERR_RET(id, retval) do { \ - if (!valid_id(id)) { \ + if (!rxa_validate_id(id)) { \ RTE_EDEV_LOG_ERR("Invalid eth Rx adapter id = %d\n", id); \ return retval; \ } \ } while (0) static inline int -sw_rx_adapter_queue_count(struct rte_event_eth_rx_adapter *rx_adapter) +rxa_sw_adapter_queue_count(struct rte_event_eth_rx_adapter *rx_adapter) { - return rx_adapter->num_rx_polled; + return rx_adapter->num_rx_polled + rx_adapter->num_rx_intr; } /* Greatest common divisor */ -static uint16_t gcd_u16(uint16_t a, uint16_t b) +static uint16_t rxa_gcd_u16(uint16_t a, uint16_t b) { uint16_t r = a % b; - return r ? gcd_u16(b, r) : b; + return r ? rxa_gcd_u16(b, r) : b; } /* Returns the next queue in the polling sequence @@ -154,7 +239,7 @@ static uint16_t gcd_u16(uint16_t a, uint16_t b) * http://kb.linuxvirtualserver.org/wiki/Weighted_Round-Robin_Scheduling */ static int -wrr_next(struct rte_event_eth_rx_adapter *rx_adapter, +rxa_wrr_next(struct rte_event_eth_rx_adapter *rx_adapter, unsigned int n, int *cw, struct eth_rx_poll_entry *eth_rx_poll, uint16_t max_wt, uint16_t gcd, int prev) @@ -164,7 +249,7 @@ wrr_next(struct rte_event_eth_rx_adapter *rx_adapter, while (1) { uint16_t q; - uint8_t d; + uint16_t d; i = (i + 1) % n; if (i == 0) { @@ -182,13 +267,298 @@ wrr_next(struct rte_event_eth_rx_adapter *rx_adapter, } } -/* Precalculate WRR polling sequence for all queues in rx_adapter */ +static inline int +rxa_shared_intr(struct eth_device_info *dev_info, + int rx_queue_id) +{ + int multi_intr_cap; + + if (dev_info->dev->intr_handle == NULL) + return 0; + + multi_intr_cap = rte_intr_cap_multiple(dev_info->dev->intr_handle); + return !multi_intr_cap || + rx_queue_id >= RTE_MAX_RXTX_INTR_VEC_ID - 1; +} + +static inline int +rxa_intr_queue(struct eth_device_info *dev_info, + int rx_queue_id) +{ + struct eth_rx_queue_info *queue_info; + + queue_info = &dev_info->rx_queue[rx_queue_id]; + return dev_info->rx_queue && + !dev_info->internal_event_port && + queue_info->queue_enabled && queue_info->wt == 0; +} + +static inline int +rxa_polled_queue(struct eth_device_info *dev_info, + int rx_queue_id) +{ + struct eth_rx_queue_info *queue_info; + + queue_info = &dev_info->rx_queue[rx_queue_id]; + return !dev_info->internal_event_port && + dev_info->rx_queue && + queue_info->queue_enabled && queue_info->wt != 0; +} + +/* Calculate change in number of vectors after Rx queue ID is add/deleted */ static int -eth_poll_wrr_calc(struct rte_event_eth_rx_adapter *rx_adapter) +rxa_nb_intr_vect(struct eth_device_info *dev_info, int rx_queue_id, int add) { - uint8_t d; + uint16_t i; + int n, s; + uint16_t nbq; + + nbq = dev_info->dev->data->nb_rx_queues; + n = 0; /* non shared count */ + s = 0; /* shared count */ + + if (rx_queue_id == -1) { + for (i = 0; i < nbq; i++) { + if (!rxa_shared_intr(dev_info, i)) + n += add ? !rxa_intr_queue(dev_info, i) : + rxa_intr_queue(dev_info, i); + else + s += add ? !rxa_intr_queue(dev_info, i) : + rxa_intr_queue(dev_info, i); + } + + if (s > 0) { + if ((add && dev_info->nb_shared_intr == 0) || + (!add && dev_info->nb_shared_intr)) + n += 1; + } + } else { + if (!rxa_shared_intr(dev_info, rx_queue_id)) + n = add ? !rxa_intr_queue(dev_info, rx_queue_id) : + rxa_intr_queue(dev_info, rx_queue_id); + else + n = add ? !dev_info->nb_shared_intr : + dev_info->nb_shared_intr == 1; + } + + return add ? n : -n; +} + +/* Calculate nb_rx_intr after deleting interrupt mode rx queues + */ +static void +rxa_calc_nb_post_intr_del(struct rte_event_eth_rx_adapter *rx_adapter, + struct eth_device_info *dev_info, + int rx_queue_id, + uint32_t *nb_rx_intr) +{ + uint32_t intr_diff; + + if (rx_queue_id == -1) + intr_diff = dev_info->nb_rx_intr; + else + intr_diff = rxa_intr_queue(dev_info, rx_queue_id); + + *nb_rx_intr = rx_adapter->num_rx_intr - intr_diff; +} + +/* Calculate nb_rx_* after adding interrupt mode rx queues, newly added + * interrupt queues could currently be poll mode Rx queues + */ +static void +rxa_calc_nb_post_add_intr(struct rte_event_eth_rx_adapter *rx_adapter, + struct eth_device_info *dev_info, + int rx_queue_id, + uint32_t *nb_rx_poll, + uint32_t *nb_rx_intr, + uint32_t *nb_wrr) +{ + uint32_t intr_diff; + uint32_t poll_diff; + uint32_t wrr_len_diff; + + if (rx_queue_id == -1) { + intr_diff = dev_info->dev->data->nb_rx_queues - + dev_info->nb_rx_intr; + poll_diff = dev_info->nb_rx_poll; + wrr_len_diff = dev_info->wrr_len; + } else { + intr_diff = !rxa_intr_queue(dev_info, rx_queue_id); + poll_diff = rxa_polled_queue(dev_info, rx_queue_id); + wrr_len_diff = poll_diff ? dev_info->rx_queue[rx_queue_id].wt : + 0; + } + + *nb_rx_intr = rx_adapter->num_rx_intr + intr_diff; + *nb_rx_poll = rx_adapter->num_rx_polled - poll_diff; + *nb_wrr = rx_adapter->wrr_len - wrr_len_diff; +} + +/* Calculate size of the eth_rx_poll and wrr_sched arrays + * after deleting poll mode rx queues + */ +static void +rxa_calc_nb_post_poll_del(struct rte_event_eth_rx_adapter *rx_adapter, + struct eth_device_info *dev_info, + int rx_queue_id, + uint32_t *nb_rx_poll, + uint32_t *nb_wrr) +{ + uint32_t poll_diff; + uint32_t wrr_len_diff; + + if (rx_queue_id == -1) { + poll_diff = dev_info->nb_rx_poll; + wrr_len_diff = dev_info->wrr_len; + } else { + poll_diff = rxa_polled_queue(dev_info, rx_queue_id); + wrr_len_diff = poll_diff ? dev_info->rx_queue[rx_queue_id].wt : + 0; + } + + *nb_rx_poll = rx_adapter->num_rx_polled - poll_diff; + *nb_wrr = rx_adapter->wrr_len - wrr_len_diff; +} + +/* Calculate nb_rx_* after adding poll mode rx queues + */ +static void +rxa_calc_nb_post_add_poll(struct rte_event_eth_rx_adapter *rx_adapter, + struct eth_device_info *dev_info, + int rx_queue_id, + uint16_t wt, + uint32_t *nb_rx_poll, + uint32_t *nb_rx_intr, + uint32_t *nb_wrr) +{ + uint32_t intr_diff; + uint32_t poll_diff; + uint32_t wrr_len_diff; + + if (rx_queue_id == -1) { + intr_diff = dev_info->nb_rx_intr; + poll_diff = dev_info->dev->data->nb_rx_queues - + dev_info->nb_rx_poll; + wrr_len_diff = wt*dev_info->dev->data->nb_rx_queues + - dev_info->wrr_len; + } else { + intr_diff = rxa_intr_queue(dev_info, rx_queue_id); + poll_diff = !rxa_polled_queue(dev_info, rx_queue_id); + wrr_len_diff = rxa_polled_queue(dev_info, rx_queue_id) ? + wt - dev_info->rx_queue[rx_queue_id].wt : + wt; + } + + *nb_rx_poll = rx_adapter->num_rx_polled + poll_diff; + *nb_rx_intr = rx_adapter->num_rx_intr - intr_diff; + *nb_wrr = rx_adapter->wrr_len + wrr_len_diff; +} + +/* Calculate nb_rx_* after adding rx_queue_id */ +static void +rxa_calc_nb_post_add(struct rte_event_eth_rx_adapter *rx_adapter, + struct eth_device_info *dev_info, + int rx_queue_id, + uint16_t wt, + uint32_t *nb_rx_poll, + uint32_t *nb_rx_intr, + uint32_t *nb_wrr) +{ + if (wt != 0) + rxa_calc_nb_post_add_poll(rx_adapter, dev_info, rx_queue_id, + wt, nb_rx_poll, nb_rx_intr, nb_wrr); + else + rxa_calc_nb_post_add_intr(rx_adapter, dev_info, rx_queue_id, + nb_rx_poll, nb_rx_intr, nb_wrr); +} + +/* Calculate nb_rx_* after deleting rx_queue_id */ +static void +rxa_calc_nb_post_del(struct rte_event_eth_rx_adapter *rx_adapter, + struct eth_device_info *dev_info, + int rx_queue_id, + uint32_t *nb_rx_poll, + uint32_t *nb_rx_intr, + uint32_t *nb_wrr) +{ + rxa_calc_nb_post_poll_del(rx_adapter, dev_info, rx_queue_id, nb_rx_poll, + nb_wrr); + rxa_calc_nb_post_intr_del(rx_adapter, dev_info, rx_queue_id, + nb_rx_intr); +} + +/* + * Allocate the rx_poll array + */ +static struct eth_rx_poll_entry * +rxa_alloc_poll(struct rte_event_eth_rx_adapter *rx_adapter, + uint32_t num_rx_polled) +{ + size_t len; + + len = RTE_ALIGN(num_rx_polled * sizeof(*rx_adapter->eth_rx_poll), + RTE_CACHE_LINE_SIZE); + return rte_zmalloc_socket(rx_adapter->mem_name, + len, + RTE_CACHE_LINE_SIZE, + rx_adapter->socket_id); +} + +/* + * Allocate the WRR array + */ +static uint32_t * +rxa_alloc_wrr(struct rte_event_eth_rx_adapter *rx_adapter, int nb_wrr) +{ + size_t len; + + len = RTE_ALIGN(nb_wrr * sizeof(*rx_adapter->wrr_sched), + RTE_CACHE_LINE_SIZE); + return rte_zmalloc_socket(rx_adapter->mem_name, + len, + RTE_CACHE_LINE_SIZE, + rx_adapter->socket_id); +} + +static int +rxa_alloc_poll_arrays(struct rte_event_eth_rx_adapter *rx_adapter, + uint32_t nb_poll, + uint32_t nb_wrr, + struct eth_rx_poll_entry **rx_poll, + uint32_t **wrr_sched) +{ + + if (nb_poll == 0) { + *rx_poll = NULL; + *wrr_sched = NULL; + return 0; + } + + *rx_poll = rxa_alloc_poll(rx_adapter, nb_poll); + if (*rx_poll == NULL) { + *wrr_sched = NULL; + return -ENOMEM; + } + + *wrr_sched = rxa_alloc_wrr(rx_adapter, nb_wrr); + if (*wrr_sched == NULL) { + rte_free(*rx_poll); + return -ENOMEM; + } + return 0; +} + +/* Precalculate WRR polling sequence for all queues in rx_adapter */ +static void +rxa_calc_wrr_sequence(struct rte_event_eth_rx_adapter *rx_adapter, + struct eth_rx_poll_entry *rx_poll, + uint32_t *rx_wrr) +{ + uint16_t d; uint16_t q; unsigned int i; + int prev = -1; + int cw = -1; /* Initialize variables for calculation of wrr schedule */ uint16_t max_wrr_pos = 0; @@ -196,79 +566,52 @@ eth_poll_wrr_calc(struct rte_event_eth_rx_adapter *rx_adapter) uint16_t max_wt = 0; uint16_t gcd = 0; - struct eth_rx_poll_entry *rx_poll = NULL; - uint32_t *rx_wrr = NULL; + if (rx_poll == NULL) + return; - if (rx_adapter->num_rx_polled) { - size_t len = RTE_ALIGN(rx_adapter->num_rx_polled * - sizeof(*rx_adapter->eth_rx_poll), - RTE_CACHE_LINE_SIZE); - rx_poll = rte_zmalloc_socket(rx_adapter->mem_name, - len, - RTE_CACHE_LINE_SIZE, - rx_adapter->socket_id); - if (rx_poll == NULL) - return -ENOMEM; + /* Generate array of all queues to poll, the size of this + * array is poll_q + */ + RTE_ETH_FOREACH_DEV(d) { + uint16_t nb_rx_queues; + struct eth_device_info *dev_info = + &rx_adapter->eth_devices[d]; + nb_rx_queues = dev_info->dev->data->nb_rx_queues; + if (dev_info->rx_queue == NULL) + continue; + if (dev_info->internal_event_port) + continue; + dev_info->wrr_len = 0; + for (q = 0; q < nb_rx_queues; q++) { + struct eth_rx_queue_info *queue_info = + &dev_info->rx_queue[q]; + uint16_t wt; - /* Generate array of all queues to poll, the size of this - * array is poll_q - */ - for (d = 0; d < rte_eth_dev_count(); d++) { - uint16_t nb_rx_queues; - struct eth_device_info *dev_info = - &rx_adapter->eth_devices[d]; - nb_rx_queues = dev_info->dev->data->nb_rx_queues; - if (dev_info->rx_queue == NULL) + if (!rxa_polled_queue(dev_info, q)) continue; - for (q = 0; q < nb_rx_queues; q++) { - struct eth_rx_queue_info *queue_info = - &dev_info->rx_queue[q]; - if (queue_info->queue_enabled == 0) - continue; - - uint16_t wt = queue_info->wt; - rx_poll[poll_q].eth_dev_id = d; - rx_poll[poll_q].eth_rx_qid = q; - max_wrr_pos += wt; - max_wt = RTE_MAX(max_wt, wt); - gcd = (gcd) ? gcd_u16(gcd, wt) : wt; - poll_q++; - } - } - - len = RTE_ALIGN(max_wrr_pos * sizeof(*rx_wrr), - RTE_CACHE_LINE_SIZE); - rx_wrr = rte_zmalloc_socket(rx_adapter->mem_name, - len, - RTE_CACHE_LINE_SIZE, - rx_adapter->socket_id); - if (rx_wrr == NULL) { - rte_free(rx_poll); - return -ENOMEM; - } - - /* Generate polling sequence based on weights */ - int prev = -1; - int cw = -1; - for (i = 0; i < max_wrr_pos; i++) { - rx_wrr[i] = wrr_next(rx_adapter, poll_q, &cw, - rx_poll, max_wt, gcd, prev); - prev = rx_wrr[i]; + wt = queue_info->wt; + rx_poll[poll_q].eth_dev_id = d; + rx_poll[poll_q].eth_rx_qid = q; + max_wrr_pos += wt; + dev_info->wrr_len += wt; + max_wt = RTE_MAX(max_wt, wt); + gcd = (gcd) ? rxa_gcd_u16(gcd, wt) : wt; + poll_q++; } } - rte_free(rx_adapter->eth_rx_poll); - rte_free(rx_adapter->wrr_sched); - - rx_adapter->eth_rx_poll = rx_poll; - rx_adapter->wrr_sched = rx_wrr; - rx_adapter->wrr_len = max_wrr_pos; - - return 0; + /* Generate polling sequence based on weights */ + prev = -1; + cw = -1; + for (i = 0; i < max_wrr_pos; i++) { + rx_wrr[i] = rxa_wrr_next(rx_adapter, poll_q, &cw, + rx_poll, max_wt, gcd, prev); + prev = rx_wrr[i]; + } } static inline void -mtoip(struct rte_mbuf *m, struct ipv4_hdr **ipv4_hdr, +rxa_mtoip(struct rte_mbuf *m, struct ipv4_hdr **ipv4_hdr, struct ipv6_hdr **ipv6_hdr) { struct ether_hdr *eth_hdr = rte_pktmbuf_mtod(m, struct ether_hdr *); @@ -307,7 +650,7 @@ mtoip(struct rte_mbuf *m, struct ipv4_hdr **ipv4_hdr, /* Calculate RSS hash for IPv4/6 */ static inline uint32_t -do_softrss(struct rte_mbuf *m, const uint8_t *rss_key_be) +rxa_do_softrss(struct rte_mbuf *m, const uint8_t *rss_key_be) { uint32_t input_len; void *tuple; @@ -316,7 +659,7 @@ do_softrss(struct rte_mbuf *m, const uint8_t *rss_key_be) struct ipv4_hdr *ipv4_hdr; struct ipv6_hdr *ipv6_hdr; - mtoip(m, &ipv4_hdr, &ipv6_hdr); + rxa_mtoip(m, &ipv4_hdr, &ipv6_hdr); if (ipv4_hdr) { ipv4_tuple.src_addr = rte_be_to_cpu_32(ipv4_hdr->src_addr); @@ -335,13 +678,13 @@ do_softrss(struct rte_mbuf *m, const uint8_t *rss_key_be) } static inline int -rx_enq_blocked(struct rte_event_eth_rx_adapter *rx_adapter) +rxa_enq_blocked(struct rte_event_eth_rx_adapter *rx_adapter) { return !!rx_adapter->enq_block_count; } static inline void -rx_enq_block_start_ts(struct rte_event_eth_rx_adapter *rx_adapter) +rxa_enq_block_start_ts(struct rte_event_eth_rx_adapter *rx_adapter) { if (rx_adapter->rx_enq_block_start_ts) return; @@ -354,13 +697,13 @@ rx_enq_block_start_ts(struct rte_event_eth_rx_adapter *rx_adapter) } static inline void -rx_enq_block_end_ts(struct rte_event_eth_rx_adapter *rx_adapter, +rxa_enq_block_end_ts(struct rte_event_eth_rx_adapter *rx_adapter, struct rte_event_eth_rx_adapter_stats *stats) { if (unlikely(!stats->rx_enq_start_ts)) stats->rx_enq_start_ts = rte_get_tsc_cycles(); - if (likely(!rx_enq_blocked(rx_adapter))) + if (likely(!rxa_enq_blocked(rx_adapter))) return; rx_adapter->enq_block_count = 0; @@ -376,8 +719,8 @@ rx_enq_block_end_ts(struct rte_event_eth_rx_adapter *rx_adapter, * this function */ static inline void -buf_event_enqueue(struct rte_event_eth_rx_adapter *rx_adapter, - struct rte_event *ev) +rxa_buffer_event(struct rte_event_eth_rx_adapter *rx_adapter, + struct rte_event *ev) { struct rte_eth_event_enqueue_buffer *buf = &rx_adapter->event_enqueue_buffer; @@ -386,7 +729,7 @@ buf_event_enqueue(struct rte_event_eth_rx_adapter *rx_adapter, /* Enqueue buffered events to event device */ static inline uint16_t -flush_event_buffer(struct rte_event_eth_rx_adapter *rx_adapter) +rxa_flush_event_buffer(struct rte_event_eth_rx_adapter *rx_adapter) { struct rte_eth_event_enqueue_buffer *buf = &rx_adapter->event_enqueue_buffer; @@ -403,8 +746,8 @@ flush_event_buffer(struct rte_event_eth_rx_adapter *rx_adapter) stats->rx_enq_retry++; } - n ? rx_enq_block_end_ts(rx_adapter, stats) : - rx_enq_block_start_ts(rx_adapter); + n ? rxa_enq_block_end_ts(rx_adapter, stats) : + rxa_enq_block_start_ts(rx_adapter); buf->count -= n; stats->rx_enq_count += n; @@ -413,18 +756,19 @@ flush_event_buffer(struct rte_event_eth_rx_adapter *rx_adapter) } static inline void -fill_event_buffer(struct rte_event_eth_rx_adapter *rx_adapter, - uint8_t dev_id, - uint16_t rx_queue_id, - struct rte_mbuf **mbufs, - uint16_t num) +rxa_buffer_mbufs(struct rte_event_eth_rx_adapter *rx_adapter, + uint16_t eth_dev_id, + uint16_t rx_queue_id, + struct rte_mbuf **mbufs, + uint16_t num) { uint32_t i; - struct eth_device_info *eth_device_info = - &rx_adapter->eth_devices[dev_id]; + struct eth_device_info *dev_info = + &rx_adapter->eth_devices[eth_dev_id]; struct eth_rx_queue_info *eth_rx_queue_info = - ð_device_info->rx_queue[rx_queue_id]; - + &dev_info->rx_queue[rx_queue_id]; + struct rte_eth_event_enqueue_buffer *buf = + &rx_adapter->event_enqueue_buffer; int32_t qid = eth_rx_queue_info->event_queue_id; uint8_t sched_type = eth_rx_queue_info->sched_type; uint8_t priority = eth_rx_queue_info->priority; @@ -434,22 +778,48 @@ fill_event_buffer(struct rte_event_eth_rx_adapter *rx_adapter, uint32_t rss_mask; uint32_t rss; int do_rss; + uint64_t ts; + struct rte_mbuf *cb_mbufs[BATCH_SIZE]; + uint16_t nb_cb; /* 0xffff ffff if PKT_RX_RSS_HASH is set, otherwise 0 */ rss_mask = ~(((m->ol_flags & PKT_RX_RSS_HASH) != 0) - 1); do_rss = !rss_mask && !eth_rx_queue_info->flow_id_mask; + if ((m->ol_flags & PKT_RX_TIMESTAMP) == 0) { + ts = rte_get_tsc_cycles(); + for (i = 0; i < num; i++) { + m = mbufs[i]; + + m->timestamp = ts; + m->ol_flags |= PKT_RX_TIMESTAMP; + } + } + + + nb_cb = dev_info->cb_fn ? dev_info->cb_fn(eth_dev_id, rx_queue_id, + ETH_EVENT_BUFFER_SIZE, + buf->count, mbufs, + num, + dev_info->cb_arg, + cb_mbufs) : + num; + if (nb_cb < num) { + mbufs = cb_mbufs; + num = nb_cb; + } + for (i = 0; i < num; i++) { m = mbufs[i]; struct rte_event *ev = &events[i]; rss = do_rss ? - do_softrss(m, rx_adapter->rss_key_be) : m->hash.rss; + rxa_do_softrss(m, rx_adapter->rss_key_be) : + m->hash.rss; flow_id = eth_rx_queue_info->flow_id & eth_rx_queue_info->flow_id_mask; flow_id |= rss & ~eth_rx_queue_info->flow_id_mask; - ev->flow_id = flow_id; ev->op = RTE_EVENT_OP_NEW; ev->sched_type = sched_type; @@ -459,8 +829,275 @@ fill_event_buffer(struct rte_event_eth_rx_adapter *rx_adapter, ev->priority = priority; ev->mbuf = m; - buf_event_enqueue(rx_adapter, ev); + rxa_buffer_event(rx_adapter, ev); + } +} + +/* Enqueue packets from <port, q> to event buffer */ +static inline uint32_t +rxa_eth_rx(struct rte_event_eth_rx_adapter *rx_adapter, + uint16_t port_id, + uint16_t queue_id, + uint32_t rx_count, + uint32_t max_rx, + int *rxq_empty) +{ + struct rte_mbuf *mbufs[BATCH_SIZE]; + struct rte_eth_event_enqueue_buffer *buf = + &rx_adapter->event_enqueue_buffer; + struct rte_event_eth_rx_adapter_stats *stats = + &rx_adapter->stats; + uint16_t n; + uint32_t nb_rx = 0; + + if (rxq_empty) + *rxq_empty = 0; + /* Don't do a batch dequeue from the rx queue if there isn't + * enough space in the enqueue buffer. + */ + while (BATCH_SIZE <= (RTE_DIM(buf->events) - buf->count)) { + if (buf->count >= BATCH_SIZE) + rxa_flush_event_buffer(rx_adapter); + + stats->rx_poll_count++; + n = rte_eth_rx_burst(port_id, queue_id, mbufs, BATCH_SIZE); + if (unlikely(!n)) { + if (rxq_empty) + *rxq_empty = 1; + break; + } + rxa_buffer_mbufs(rx_adapter, port_id, queue_id, mbufs, n); + nb_rx += n; + if (rx_count + nb_rx > max_rx) + break; + } + + if (buf->count >= BATCH_SIZE) + rxa_flush_event_buffer(rx_adapter); + + return nb_rx; +} + +static inline void +rxa_intr_ring_enqueue(struct rte_event_eth_rx_adapter *rx_adapter, + void *data) +{ + uint16_t port_id; + uint16_t queue; + int err; + union queue_data qd; + struct eth_device_info *dev_info; + struct eth_rx_queue_info *queue_info; + int *intr_enabled; + + qd.ptr = data; + port_id = qd.port; + queue = qd.queue; + + dev_info = &rx_adapter->eth_devices[port_id]; + queue_info = &dev_info->rx_queue[queue]; + rte_spinlock_lock(&rx_adapter->intr_ring_lock); + if (rxa_shared_intr(dev_info, queue)) + intr_enabled = &dev_info->shared_intr_enabled; + else + intr_enabled = &queue_info->intr_enabled; + + if (*intr_enabled) { + *intr_enabled = 0; + err = rte_ring_enqueue(rx_adapter->intr_ring, data); + /* Entry should always be available. + * The ring size equals the maximum number of interrupt + * vectors supported (an interrupt vector is shared in + * case of shared interrupts) + */ + if (err) + RTE_EDEV_LOG_ERR("Failed to enqueue interrupt" + " to ring: %s", strerror(err)); + else + rte_eth_dev_rx_intr_disable(port_id, queue); + } + rte_spinlock_unlock(&rx_adapter->intr_ring_lock); +} + +static int +rxa_intr_ring_check_avail(struct rte_event_eth_rx_adapter *rx_adapter, + uint32_t num_intr_vec) +{ + if (rx_adapter->num_intr_vec + num_intr_vec > + RTE_EVENT_ETH_INTR_RING_SIZE) { + RTE_EDEV_LOG_ERR("Exceeded intr ring slots current" + " %d needed %d limit %d", rx_adapter->num_intr_vec, + num_intr_vec, RTE_EVENT_ETH_INTR_RING_SIZE); + return -ENOSPC; + } + + return 0; +} + +/* Delete entries for (dev, queue) from the interrupt ring */ +static void +rxa_intr_ring_del_entries(struct rte_event_eth_rx_adapter *rx_adapter, + struct eth_device_info *dev_info, + uint16_t rx_queue_id) +{ + int i, n; + union queue_data qd; + + rte_spinlock_lock(&rx_adapter->intr_ring_lock); + + n = rte_ring_count(rx_adapter->intr_ring); + for (i = 0; i < n; i++) { + rte_ring_dequeue(rx_adapter->intr_ring, &qd.ptr); + if (!rxa_shared_intr(dev_info, rx_queue_id)) { + if (qd.port == dev_info->dev->data->port_id && + qd.queue == rx_queue_id) + continue; + } else { + if (qd.port == dev_info->dev->data->port_id) + continue; + } + rte_ring_enqueue(rx_adapter->intr_ring, qd.ptr); + } + + rte_spinlock_unlock(&rx_adapter->intr_ring_lock); +} + +/* pthread callback handling interrupt mode receive queues + * After receiving an Rx interrupt, it enqueues the port id and queue id of the + * interrupting queue to the adapter's ring buffer for interrupt events. + * These events are picked up by rxa_intr_ring_dequeue() which is invoked from + * the adapter service function. + */ +static void * +rxa_intr_thread(void *arg) +{ + struct rte_event_eth_rx_adapter *rx_adapter = arg; + struct rte_epoll_event *epoll_events = rx_adapter->epoll_events; + int n, i; + + while (1) { + n = rte_epoll_wait(rx_adapter->epd, epoll_events, + RTE_EVENT_ETH_INTR_RING_SIZE, -1); + if (unlikely(n < 0)) + RTE_EDEV_LOG_ERR("rte_epoll_wait returned error %d", + n); + for (i = 0; i < n; i++) { + rxa_intr_ring_enqueue(rx_adapter, + epoll_events[i].epdata.data); + } + } + + return NULL; +} + +/* Dequeue <port, q> from interrupt ring and enqueue received + * mbufs to eventdev + */ +static inline uint32_t +rxa_intr_ring_dequeue(struct rte_event_eth_rx_adapter *rx_adapter) +{ + uint32_t n; + uint32_t nb_rx = 0; + int rxq_empty; + struct rte_eth_event_enqueue_buffer *buf; + rte_spinlock_t *ring_lock; + uint8_t max_done = 0; + + if (rx_adapter->num_rx_intr == 0) + return 0; + + if (rte_ring_count(rx_adapter->intr_ring) == 0 + && !rx_adapter->qd_valid) + return 0; + + buf = &rx_adapter->event_enqueue_buffer; + ring_lock = &rx_adapter->intr_ring_lock; + + if (buf->count >= BATCH_SIZE) + rxa_flush_event_buffer(rx_adapter); + + while (BATCH_SIZE <= (RTE_DIM(buf->events) - buf->count)) { + struct eth_device_info *dev_info; + uint16_t port; + uint16_t queue; + union queue_data qd = rx_adapter->qd; + int err; + + if (!rx_adapter->qd_valid) { + struct eth_rx_queue_info *queue_info; + + rte_spinlock_lock(ring_lock); + err = rte_ring_dequeue(rx_adapter->intr_ring, &qd.ptr); + if (err) { + rte_spinlock_unlock(ring_lock); + break; + } + + port = qd.port; + queue = qd.queue; + rx_adapter->qd = qd; + rx_adapter->qd_valid = 1; + dev_info = &rx_adapter->eth_devices[port]; + if (rxa_shared_intr(dev_info, queue)) + dev_info->shared_intr_enabled = 1; + else { + queue_info = &dev_info->rx_queue[queue]; + queue_info->intr_enabled = 1; + } + rte_eth_dev_rx_intr_enable(port, queue); + rte_spinlock_unlock(ring_lock); + } else { + port = qd.port; + queue = qd.queue; + + dev_info = &rx_adapter->eth_devices[port]; + } + + if (rxa_shared_intr(dev_info, queue)) { + uint16_t i; + uint16_t nb_queues; + + nb_queues = dev_info->dev->data->nb_rx_queues; + n = 0; + for (i = dev_info->next_q_idx; i < nb_queues; i++) { + uint8_t enq_buffer_full; + + if (!rxa_intr_queue(dev_info, i)) + continue; + n = rxa_eth_rx(rx_adapter, port, i, nb_rx, + rx_adapter->max_nb_rx, + &rxq_empty); + nb_rx += n; + + enq_buffer_full = !rxq_empty && n == 0; + max_done = nb_rx > rx_adapter->max_nb_rx; + + if (enq_buffer_full || max_done) { + dev_info->next_q_idx = i; + goto done; + } + } + + rx_adapter->qd_valid = 0; + + /* Reinitialize for next interrupt */ + dev_info->next_q_idx = dev_info->multi_intr_cap ? + RTE_MAX_RXTX_INTR_VEC_ID - 1 : + 0; + } else { + n = rxa_eth_rx(rx_adapter, port, queue, nb_rx, + rx_adapter->max_nb_rx, + &rxq_empty); + rx_adapter->qd_valid = !rxq_empty; + nb_rx += n; + if (nb_rx > rx_adapter->max_nb_rx) + break; + } } + +done: + rx_adapter->stats.rx_intr_packets += nb_rx; + return nb_rx; } /* @@ -477,12 +1114,10 @@ fill_event_buffer(struct rte_event_eth_rx_adapter *rx_adapter, * it. */ static inline uint32_t -eth_rx_poll(struct rte_event_eth_rx_adapter *rx_adapter) +rxa_poll(struct rte_event_eth_rx_adapter *rx_adapter) { uint32_t num_queue; - uint16_t n; uint32_t nb_rx = 0; - struct rte_mbuf *mbufs[BATCH_SIZE]; struct rte_eth_event_enqueue_buffer *buf; uint32_t wrr_pos; uint32_t max_nb_rx; @@ -490,57 +1125,54 @@ eth_rx_poll(struct rte_event_eth_rx_adapter *rx_adapter) wrr_pos = rx_adapter->wrr_pos; max_nb_rx = rx_adapter->max_nb_rx; buf = &rx_adapter->event_enqueue_buffer; - struct rte_event_eth_rx_adapter_stats *stats = &rx_adapter->stats; + stats = &rx_adapter->stats; /* Iterate through a WRR sequence */ for (num_queue = 0; num_queue < rx_adapter->wrr_len; num_queue++) { unsigned int poll_idx = rx_adapter->wrr_sched[wrr_pos]; uint16_t qid = rx_adapter->eth_rx_poll[poll_idx].eth_rx_qid; - uint8_t d = rx_adapter->eth_rx_poll[poll_idx].eth_dev_id; + uint16_t d = rx_adapter->eth_rx_poll[poll_idx].eth_dev_id; /* Don't do a batch dequeue from the rx queue if there isn't * enough space in the enqueue buffer. */ if (buf->count >= BATCH_SIZE) - flush_event_buffer(rx_adapter); - if (BATCH_SIZE > (ETH_EVENT_BUFFER_SIZE - buf->count)) - break; - - stats->rx_poll_count++; - n = rte_eth_rx_burst(d, qid, mbufs, BATCH_SIZE); + rxa_flush_event_buffer(rx_adapter); + if (BATCH_SIZE > (ETH_EVENT_BUFFER_SIZE - buf->count)) { + rx_adapter->wrr_pos = wrr_pos; + return nb_rx; + } - if (n) { - stats->rx_packets += n; - /* The check before rte_eth_rx_burst() ensures that - * all n mbufs can be buffered - */ - fill_event_buffer(rx_adapter, d, qid, mbufs, n); - nb_rx += n; - if (nb_rx > max_nb_rx) { - rx_adapter->wrr_pos = + nb_rx += rxa_eth_rx(rx_adapter, d, qid, nb_rx, max_nb_rx, + NULL); + if (nb_rx > max_nb_rx) { + rx_adapter->wrr_pos = (wrr_pos + 1) % rx_adapter->wrr_len; - return nb_rx; - } + break; } if (++wrr_pos == rx_adapter->wrr_len) wrr_pos = 0; } - return nb_rx; } static int -event_eth_rx_adapter_service_func(void *args) +rxa_service_func(void *args) { struct rte_event_eth_rx_adapter *rx_adapter = args; - struct rte_eth_event_enqueue_buffer *buf; + struct rte_event_eth_rx_adapter_stats *stats; - buf = &rx_adapter->event_enqueue_buffer; if (rte_spinlock_trylock(&rx_adapter->rx_lock) == 0) return 0; - if (eth_rx_poll(rx_adapter) == 0 && buf->count) - flush_event_buffer(rx_adapter); + if (!rx_adapter->rxa_started) { + return 0; + rte_spinlock_unlock(&rx_adapter->rx_lock); + } + + stats = &rx_adapter->stats; + stats->rx_packets += rxa_intr_ring_dequeue(rx_adapter); + stats->rx_packets += rxa_poll(rx_adapter); rte_spinlock_unlock(&rx_adapter->rx_lock); return 0; } @@ -572,14 +1204,14 @@ rte_event_eth_rx_adapter_init(void) } static inline struct rte_event_eth_rx_adapter * -id_to_rx_adapter(uint8_t id) +rxa_id_to_adapter(uint8_t id) { return event_eth_rx_adapter ? event_eth_rx_adapter[id] : NULL; } static int -default_conf_cb(uint8_t id, uint8_t dev_id, +rxa_default_conf_cb(uint8_t id, uint8_t dev_id, struct rte_event_eth_rx_adapter_conf *conf, void *arg) { int ret; @@ -588,7 +1220,7 @@ default_conf_cb(uint8_t id, uint8_t dev_id, int started; uint8_t port_id; struct rte_event_port_conf *port_conf = arg; - struct rte_event_eth_rx_adapter *rx_adapter = id_to_rx_adapter(id); + struct rte_event_eth_rx_adapter *rx_adapter = rxa_id_to_adapter(id); dev = &rte_eventdevs[rx_adapter->eventdev_id]; dev_conf = dev->data->dev_conf; @@ -625,7 +1257,351 @@ default_conf_cb(uint8_t id, uint8_t dev_id, } static int -init_service(struct rte_event_eth_rx_adapter *rx_adapter, uint8_t id) +rxa_epoll_create1(void) +{ +#if defined(LINUX) + int fd; + fd = epoll_create1(EPOLL_CLOEXEC); + return fd < 0 ? -errno : fd; +#elif defined(BSD) + return -ENOTSUP; +#endif +} + +static int +rxa_init_epd(struct rte_event_eth_rx_adapter *rx_adapter) +{ + if (rx_adapter->epd != INIT_FD) + return 0; + + rx_adapter->epd = rxa_epoll_create1(); + if (rx_adapter->epd < 0) { + int err = rx_adapter->epd; + rx_adapter->epd = INIT_FD; + RTE_EDEV_LOG_ERR("epoll_create1() failed, err %d", err); + return err; + } + + return 0; +} + +static int +rxa_create_intr_thread(struct rte_event_eth_rx_adapter *rx_adapter) +{ + int err; + char thread_name[RTE_MAX_THREAD_NAME_LEN]; + + if (rx_adapter->intr_ring) + return 0; + + rx_adapter->intr_ring = rte_ring_create("intr_ring", + RTE_EVENT_ETH_INTR_RING_SIZE, + rte_socket_id(), 0); + if (!rx_adapter->intr_ring) + return -ENOMEM; + + rx_adapter->epoll_events = rte_zmalloc_socket(rx_adapter->mem_name, + RTE_EVENT_ETH_INTR_RING_SIZE * + sizeof(struct rte_epoll_event), + RTE_CACHE_LINE_SIZE, + rx_adapter->socket_id); + if (!rx_adapter->epoll_events) { + err = -ENOMEM; + goto error; + } + + rte_spinlock_init(&rx_adapter->intr_ring_lock); + + snprintf(thread_name, RTE_MAX_THREAD_NAME_LEN, + "rx-intr-thread-%d", rx_adapter->id); + + err = rte_ctrl_thread_create(&rx_adapter->rx_intr_thread, thread_name, + NULL, rxa_intr_thread, rx_adapter); + if (!err) { + rte_thread_setname(rx_adapter->rx_intr_thread, thread_name); + return 0; + } + + RTE_EDEV_LOG_ERR("Failed to create interrupt thread err = %d\n", err); +error: + rte_ring_free(rx_adapter->intr_ring); + rx_adapter->intr_ring = NULL; + rx_adapter->epoll_events = NULL; + return err; +} + +static int +rxa_destroy_intr_thread(struct rte_event_eth_rx_adapter *rx_adapter) +{ + int err; + + err = pthread_cancel(rx_adapter->rx_intr_thread); + if (err) + RTE_EDEV_LOG_ERR("Can't cancel interrupt thread err = %d\n", + err); + + err = pthread_join(rx_adapter->rx_intr_thread, NULL); + if (err) + RTE_EDEV_LOG_ERR("Can't join interrupt thread err = %d\n", err); + + rte_free(rx_adapter->epoll_events); + rte_ring_free(rx_adapter->intr_ring); + rx_adapter->intr_ring = NULL; + rx_adapter->epoll_events = NULL; + return 0; +} + +static int +rxa_free_intr_resources(struct rte_event_eth_rx_adapter *rx_adapter) +{ + int ret; + + if (rx_adapter->num_rx_intr == 0) + return 0; + + ret = rxa_destroy_intr_thread(rx_adapter); + if (ret) + return ret; + + close(rx_adapter->epd); + rx_adapter->epd = INIT_FD; + + return ret; +} + +static int +rxa_disable_intr(struct rte_event_eth_rx_adapter *rx_adapter, + struct eth_device_info *dev_info, + uint16_t rx_queue_id) +{ + int err; + uint16_t eth_dev_id = dev_info->dev->data->port_id; + int sintr = rxa_shared_intr(dev_info, rx_queue_id); + + err = rte_eth_dev_rx_intr_disable(eth_dev_id, rx_queue_id); + if (err) { + RTE_EDEV_LOG_ERR("Could not disable interrupt for Rx queue %u", + rx_queue_id); + return err; + } + + err = rte_eth_dev_rx_intr_ctl_q(eth_dev_id, rx_queue_id, + rx_adapter->epd, + RTE_INTR_EVENT_DEL, + 0); + if (err) + RTE_EDEV_LOG_ERR("Interrupt event deletion failed %d", err); + + if (sintr) + dev_info->rx_queue[rx_queue_id].intr_enabled = 0; + else + dev_info->shared_intr_enabled = 0; + return err; +} + +static int +rxa_del_intr_queue(struct rte_event_eth_rx_adapter *rx_adapter, + struct eth_device_info *dev_info, + int rx_queue_id) +{ + int err; + int i; + int s; + + if (dev_info->nb_rx_intr == 0) + return 0; + + err = 0; + if (rx_queue_id == -1) { + s = dev_info->nb_shared_intr; + for (i = 0; i < dev_info->nb_rx_intr; i++) { + int sintr; + uint16_t q; + + q = dev_info->intr_queue[i]; + sintr = rxa_shared_intr(dev_info, q); + s -= sintr; + + if (!sintr || s == 0) { + + err = rxa_disable_intr(rx_adapter, dev_info, + q); + if (err) + return err; + rxa_intr_ring_del_entries(rx_adapter, dev_info, + q); + } + } + } else { + if (!rxa_intr_queue(dev_info, rx_queue_id)) + return 0; + if (!rxa_shared_intr(dev_info, rx_queue_id) || + dev_info->nb_shared_intr == 1) { + err = rxa_disable_intr(rx_adapter, dev_info, + rx_queue_id); + if (err) + return err; + rxa_intr_ring_del_entries(rx_adapter, dev_info, + rx_queue_id); + } + + for (i = 0; i < dev_info->nb_rx_intr; i++) { + if (dev_info->intr_queue[i] == rx_queue_id) { + for (; i < dev_info->nb_rx_intr - 1; i++) + dev_info->intr_queue[i] = + dev_info->intr_queue[i + 1]; + break; + } + } + } + + return err; +} + +static int +rxa_config_intr(struct rte_event_eth_rx_adapter *rx_adapter, + struct eth_device_info *dev_info, + uint16_t rx_queue_id) +{ + int err, err1; + uint16_t eth_dev_id = dev_info->dev->data->port_id; + union queue_data qd; + int init_fd; + uint16_t *intr_queue; + int sintr = rxa_shared_intr(dev_info, rx_queue_id); + + if (rxa_intr_queue(dev_info, rx_queue_id)) + return 0; + + intr_queue = dev_info->intr_queue; + if (dev_info->intr_queue == NULL) { + size_t len = + dev_info->dev->data->nb_rx_queues * sizeof(uint16_t); + dev_info->intr_queue = + rte_zmalloc_socket( + rx_adapter->mem_name, + len, + 0, + rx_adapter->socket_id); + if (dev_info->intr_queue == NULL) + return -ENOMEM; + } + + init_fd = rx_adapter->epd; + err = rxa_init_epd(rx_adapter); + if (err) + goto err_free_queue; + + qd.port = eth_dev_id; + qd.queue = rx_queue_id; + + err = rte_eth_dev_rx_intr_ctl_q(eth_dev_id, rx_queue_id, + rx_adapter->epd, + RTE_INTR_EVENT_ADD, + qd.ptr); + if (err) { + RTE_EDEV_LOG_ERR("Failed to add interrupt event for" + " Rx Queue %u err %d", rx_queue_id, err); + goto err_del_fd; + } + + err = rte_eth_dev_rx_intr_enable(eth_dev_id, rx_queue_id); + if (err) { + RTE_EDEV_LOG_ERR("Could not enable interrupt for" + " Rx Queue %u err %d", rx_queue_id, err); + + goto err_del_event; + } + + err = rxa_create_intr_thread(rx_adapter); + if (!err) { + if (sintr) + dev_info->shared_intr_enabled = 1; + else + dev_info->rx_queue[rx_queue_id].intr_enabled = 1; + return 0; + } + + + err = rte_eth_dev_rx_intr_disable(eth_dev_id, rx_queue_id); + if (err) + RTE_EDEV_LOG_ERR("Could not disable interrupt for" + " Rx Queue %u err %d", rx_queue_id, err); +err_del_event: + err1 = rte_eth_dev_rx_intr_ctl_q(eth_dev_id, rx_queue_id, + rx_adapter->epd, + RTE_INTR_EVENT_DEL, + 0); + if (err1) { + RTE_EDEV_LOG_ERR("Could not delete event for" + " Rx Queue %u err %d", rx_queue_id, err1); + } +err_del_fd: + if (init_fd == INIT_FD) { + close(rx_adapter->epd); + rx_adapter->epd = -1; + } +err_free_queue: + if (intr_queue == NULL) + rte_free(dev_info->intr_queue); + + return err; +} + +static int +rxa_add_intr_queue(struct rte_event_eth_rx_adapter *rx_adapter, + struct eth_device_info *dev_info, + int rx_queue_id) + +{ + int i, j, err; + int si = -1; + int shared_done = (dev_info->nb_shared_intr > 0); + + if (rx_queue_id != -1) { + if (rxa_shared_intr(dev_info, rx_queue_id) && shared_done) + return 0; + return rxa_config_intr(rx_adapter, dev_info, rx_queue_id); + } + + err = 0; + for (i = 0; i < dev_info->dev->data->nb_rx_queues; i++) { + + if (rxa_shared_intr(dev_info, i) && shared_done) + continue; + + err = rxa_config_intr(rx_adapter, dev_info, i); + + shared_done = err == 0 && rxa_shared_intr(dev_info, i); + if (shared_done) { + si = i; + dev_info->shared_intr_enabled = 1; + } + if (err) + break; + } + + if (err == 0) + return 0; + + shared_done = (dev_info->nb_shared_intr > 0); + for (j = 0; j < i; j++) { + if (rxa_intr_queue(dev_info, j)) + continue; + if (rxa_shared_intr(dev_info, j) && si != j) + continue; + err = rxa_disable_intr(rx_adapter, dev_info, j); + if (err) + break; + + } + + return err; +} + + +static int +rxa_init_service(struct rte_event_eth_rx_adapter *rx_adapter, uint8_t id) { int ret; struct rte_service_spec service; @@ -638,7 +1614,7 @@ init_service(struct rte_event_eth_rx_adapter *rx_adapter, uint8_t id) snprintf(service.name, ETH_RX_ADAPTER_SERVICE_NAME_LEN, "rte_event_eth_rx_adapter_%d", id); service.socket_id = rx_adapter->socket_id; - service.callback = event_eth_rx_adapter_service_func; + service.callback = rxa_service_func; service.callback_userdata = rx_adapter; /* Service function handles locking for queue add/del updates */ service.capabilities = RTE_SERVICE_CAP_MT_SAFE; @@ -659,6 +1635,7 @@ init_service(struct rte_event_eth_rx_adapter *rx_adapter, uint8_t id) rx_adapter->event_port_id = rx_adapter_conf.event_port_id; rx_adapter->max_nb_rx = rx_adapter_conf.max_nb_rx; rx_adapter->service_inited = 1; + rx_adapter->epd = INIT_FD; return 0; err_done: @@ -666,9 +1643,8 @@ err_done: return ret; } - static void -update_queue_info(struct rte_event_eth_rx_adapter *rx_adapter, +rxa_update_queue(struct rte_event_eth_rx_adapter *rx_adapter, struct eth_device_info *dev_info, int32_t rx_queue_id, uint8_t add) @@ -682,7 +1658,7 @@ update_queue_info(struct rte_event_eth_rx_adapter *rx_adapter, if (rx_queue_id == -1) { for (i = 0; i < dev_info->dev->data->nb_rx_queues; i++) - update_queue_info(rx_adapter, dev_info, i, add); + rxa_update_queue(rx_adapter, dev_info, i, add); } else { queue_info = &dev_info->rx_queue[rx_queue_id]; enabled = queue_info->queue_enabled; @@ -697,31 +1673,65 @@ update_queue_info(struct rte_event_eth_rx_adapter *rx_adapter, } } -static int -event_eth_rx_adapter_queue_del(struct rte_event_eth_rx_adapter *rx_adapter, - struct eth_device_info *dev_info, - uint16_t rx_queue_id) +static void +rxa_sw_del(struct rte_event_eth_rx_adapter *rx_adapter, + struct eth_device_info *dev_info, + int32_t rx_queue_id) { - struct eth_rx_queue_info *queue_info; + int pollq; + int intrq; + int sintrq; + if (rx_adapter->nb_queues == 0) - return 0; + return; - queue_info = &dev_info->rx_queue[rx_queue_id]; - rx_adapter->num_rx_polled -= queue_info->queue_enabled; - update_queue_info(rx_adapter, dev_info, rx_queue_id, 0); - return 0; + if (rx_queue_id == -1) { + uint16_t nb_rx_queues; + uint16_t i; + + nb_rx_queues = dev_info->dev->data->nb_rx_queues; + for (i = 0; i < nb_rx_queues; i++) + rxa_sw_del(rx_adapter, dev_info, i); + return; + } + + pollq = rxa_polled_queue(dev_info, rx_queue_id); + intrq = rxa_intr_queue(dev_info, rx_queue_id); + sintrq = rxa_shared_intr(dev_info, rx_queue_id); + rxa_update_queue(rx_adapter, dev_info, rx_queue_id, 0); + rx_adapter->num_rx_polled -= pollq; + dev_info->nb_rx_poll -= pollq; + rx_adapter->num_rx_intr -= intrq; + dev_info->nb_rx_intr -= intrq; + dev_info->nb_shared_intr -= intrq && sintrq; } static void -event_eth_rx_adapter_queue_add(struct rte_event_eth_rx_adapter *rx_adapter, - struct eth_device_info *dev_info, - uint16_t rx_queue_id, - const struct rte_event_eth_rx_adapter_queue_conf *conf) - +rxa_add_queue(struct rte_event_eth_rx_adapter *rx_adapter, + struct eth_device_info *dev_info, + int32_t rx_queue_id, + const struct rte_event_eth_rx_adapter_queue_conf *conf) { struct eth_rx_queue_info *queue_info; const struct rte_event *ev = &conf->ev; + int pollq; + int intrq; + int sintrq; + + if (rx_queue_id == -1) { + uint16_t nb_rx_queues; + uint16_t i; + + nb_rx_queues = dev_info->dev->data->nb_rx_queues; + for (i = 0; i < nb_rx_queues; i++) + rxa_add_queue(rx_adapter, dev_info, i, conf); + return; + } + + pollq = rxa_polled_queue(dev_info, rx_queue_id); + intrq = rxa_intr_queue(dev_info, rx_queue_id); + sintrq = rxa_shared_intr(dev_info, rx_queue_id); queue_info = &dev_info->rx_queue[rx_queue_id]; queue_info->event_queue_id = ev->queue_id; @@ -735,69 +1745,162 @@ event_eth_rx_adapter_queue_add(struct rte_event_eth_rx_adapter *rx_adapter, queue_info->flow_id_mask = ~0; } - /* The same queue can be added more than once */ - rx_adapter->num_rx_polled += !queue_info->queue_enabled; - update_queue_info(rx_adapter, dev_info, rx_queue_id, 1); + rxa_update_queue(rx_adapter, dev_info, rx_queue_id, 1); + if (rxa_polled_queue(dev_info, rx_queue_id)) { + rx_adapter->num_rx_polled += !pollq; + dev_info->nb_rx_poll += !pollq; + rx_adapter->num_rx_intr -= intrq; + dev_info->nb_rx_intr -= intrq; + dev_info->nb_shared_intr -= intrq && sintrq; + } + + if (rxa_intr_queue(dev_info, rx_queue_id)) { + rx_adapter->num_rx_polled -= pollq; + dev_info->nb_rx_poll -= pollq; + rx_adapter->num_rx_intr += !intrq; + dev_info->nb_rx_intr += !intrq; + dev_info->nb_shared_intr += !intrq && sintrq; + if (dev_info->nb_shared_intr == 1) { + if (dev_info->multi_intr_cap) + dev_info->next_q_idx = + RTE_MAX_RXTX_INTR_VEC_ID - 1; + else + dev_info->next_q_idx = 0; + } + } } -static int add_rx_queue(struct rte_event_eth_rx_adapter *rx_adapter, - uint8_t eth_dev_id, +static int rxa_sw_add(struct rte_event_eth_rx_adapter *rx_adapter, + uint16_t eth_dev_id, int rx_queue_id, const struct rte_event_eth_rx_adapter_queue_conf *queue_conf) { struct eth_device_info *dev_info = &rx_adapter->eth_devices[eth_dev_id]; struct rte_event_eth_rx_adapter_queue_conf temp_conf; - uint32_t i; int ret; + struct eth_rx_poll_entry *rx_poll; + struct eth_rx_queue_info *rx_queue; + uint32_t *rx_wrr; + uint16_t nb_rx_queues; + uint32_t nb_rx_poll, nb_wrr; + uint32_t nb_rx_intr; + int num_intr_vec; + uint16_t wt; if (queue_conf->servicing_weight == 0) { - struct rte_eth_dev_data *data = dev_info->dev->data; - if (data->dev_conf.intr_conf.rxq) { - RTE_EDEV_LOG_ERR("Interrupt driven queues" - " not supported"); - return -ENOTSUP; - } - temp_conf = *queue_conf; - /* If Rx interrupts are disabled set wt = 1 */ - temp_conf.servicing_weight = 1; + temp_conf = *queue_conf; + if (!data->dev_conf.intr_conf.rxq) { + /* If Rx interrupts are disabled set wt = 1 */ + temp_conf.servicing_weight = 1; + } queue_conf = &temp_conf; } + nb_rx_queues = dev_info->dev->data->nb_rx_queues; + rx_queue = dev_info->rx_queue; + wt = queue_conf->servicing_weight; + if (dev_info->rx_queue == NULL) { dev_info->rx_queue = rte_zmalloc_socket(rx_adapter->mem_name, - dev_info->dev->data->nb_rx_queues * + nb_rx_queues * sizeof(struct eth_rx_queue_info), 0, rx_adapter->socket_id); if (dev_info->rx_queue == NULL) return -ENOMEM; } + rx_wrr = NULL; + rx_poll = NULL; - if (rx_queue_id == -1) { - for (i = 0; i < dev_info->dev->data->nb_rx_queues; i++) - event_eth_rx_adapter_queue_add(rx_adapter, - dev_info, i, - queue_conf); + rxa_calc_nb_post_add(rx_adapter, dev_info, rx_queue_id, + queue_conf->servicing_weight, + &nb_rx_poll, &nb_rx_intr, &nb_wrr); + + if (dev_info->dev->intr_handle) + dev_info->multi_intr_cap = + rte_intr_cap_multiple(dev_info->dev->intr_handle); + + ret = rxa_alloc_poll_arrays(rx_adapter, nb_rx_poll, nb_wrr, + &rx_poll, &rx_wrr); + if (ret) + goto err_free_rxqueue; + + if (wt == 0) { + num_intr_vec = rxa_nb_intr_vect(dev_info, rx_queue_id, 1); + + ret = rxa_intr_ring_check_avail(rx_adapter, num_intr_vec); + if (ret) + goto err_free_rxqueue; + + ret = rxa_add_intr_queue(rx_adapter, dev_info, rx_queue_id); + if (ret) + goto err_free_rxqueue; } else { - event_eth_rx_adapter_queue_add(rx_adapter, dev_info, - (uint16_t)rx_queue_id, - queue_conf); + + num_intr_vec = 0; + if (rx_adapter->num_rx_intr > nb_rx_intr) { + num_intr_vec = rxa_nb_intr_vect(dev_info, + rx_queue_id, 0); + /* interrupt based queues are being converted to + * poll mode queues, delete the interrupt configuration + * for those. + */ + ret = rxa_del_intr_queue(rx_adapter, + dev_info, rx_queue_id); + if (ret) + goto err_free_rxqueue; + } } - ret = eth_poll_wrr_calc(rx_adapter); - if (ret) { - event_eth_rx_adapter_queue_del(rx_adapter, - dev_info, rx_queue_id); - return ret; + if (nb_rx_intr == 0) { + ret = rxa_free_intr_resources(rx_adapter); + if (ret) + goto err_free_rxqueue; } - return ret; + if (wt == 0) { + uint16_t i; + + if (rx_queue_id == -1) { + for (i = 0; i < dev_info->dev->data->nb_rx_queues; i++) + dev_info->intr_queue[i] = i; + } else { + if (!rxa_intr_queue(dev_info, rx_queue_id)) + dev_info->intr_queue[nb_rx_intr - 1] = + rx_queue_id; + } + } + + + + rxa_add_queue(rx_adapter, dev_info, rx_queue_id, queue_conf); + rxa_calc_wrr_sequence(rx_adapter, rx_poll, rx_wrr); + + rte_free(rx_adapter->eth_rx_poll); + rte_free(rx_adapter->wrr_sched); + + rx_adapter->eth_rx_poll = rx_poll; + rx_adapter->wrr_sched = rx_wrr; + rx_adapter->wrr_len = nb_wrr; + rx_adapter->num_intr_vec += num_intr_vec; + return 0; + +err_free_rxqueue: + if (rx_queue == NULL) { + rte_free(dev_info->rx_queue); + dev_info->rx_queue = NULL; + } + + rte_free(rx_poll); + rte_free(rx_wrr); + + return 0; } static int -rx_adapter_ctrl(uint8_t id, int start) +rxa_ctrl(uint8_t id, int start) { struct rte_event_eth_rx_adapter *rx_adapter; struct rte_eventdev *dev; @@ -807,13 +1910,13 @@ rx_adapter_ctrl(uint8_t id, int start) int stop = !start; RTE_EVENT_ETH_RX_ADAPTER_ID_VALID_OR_ERR_RET(id, -EINVAL); - rx_adapter = id_to_rx_adapter(id); + rx_adapter = rxa_id_to_adapter(id); if (rx_adapter == NULL) return -EINVAL; dev = &rte_eventdevs[rx_adapter->eventdev_id]; - for (i = 0; i < rte_eth_dev_count(); i++) { + RTE_ETH_FOREACH_DEV(i) { dev_info = &rx_adapter->eth_devices[i]; /* if start check for num dev queues */ if (start && !dev_info->nb_dev_queues) @@ -831,8 +1934,12 @@ rx_adapter_ctrl(uint8_t id, int start) &rte_eth_devices[i]); } - if (use_service) + if (use_service) { + rte_spinlock_lock(&rx_adapter->rx_lock); + rx_adapter->rxa_started = start; rte_service_runstate_set(rx_adapter->service_id, start); + rte_spinlock_unlock(&rx_adapter->rx_lock); + } return 0; } @@ -845,7 +1952,7 @@ rte_event_eth_rx_adapter_create_ext(uint8_t id, uint8_t dev_id, struct rte_event_eth_rx_adapter *rx_adapter; int ret; int socket_id; - uint8_t i; + uint16_t i; char mem_name[ETH_RX_ADAPTER_SERVICE_NAME_LEN]; const uint8_t default_rss_key[] = { 0x6d, 0x5a, 0x56, 0xda, 0x25, 0x5b, 0x0e, 0xc2, @@ -866,7 +1973,7 @@ rte_event_eth_rx_adapter_create_ext(uint8_t id, uint8_t dev_id, return ret; } - rx_adapter = id_to_rx_adapter(id); + rx_adapter = rxa_id_to_adapter(id); if (rx_adapter != NULL) { RTE_EDEV_LOG_ERR("Eth Rx adapter exists id = %" PRIu8, id); return -EEXIST; @@ -888,9 +1995,11 @@ rte_event_eth_rx_adapter_create_ext(uint8_t id, uint8_t dev_id, rx_adapter->socket_id = socket_id; rx_adapter->conf_cb = conf_cb; rx_adapter->conf_arg = conf_arg; + rx_adapter->id = id; strcpy(rx_adapter->mem_name, mem_name); rx_adapter->eth_devices = rte_zmalloc_socket(rx_adapter->mem_name, - rte_eth_dev_count() * + /* FIXME: incompatible with hotplug */ + rte_eth_dev_count_total() * sizeof(struct eth_device_info), 0, socket_id); rte_convert_rss_key((const uint32_t *)default_rss_key, @@ -903,11 +2012,11 @@ rte_event_eth_rx_adapter_create_ext(uint8_t id, uint8_t dev_id, return -ENOMEM; } rte_spinlock_init(&rx_adapter->rx_lock); - for (i = 0; i < rte_eth_dev_count(); i++) + RTE_ETH_FOREACH_DEV(i) rx_adapter->eth_devices[i].dev = &rte_eth_devices[i]; event_eth_rx_adapter[id] = rx_adapter; - if (conf_cb == default_conf_cb) + if (conf_cb == rxa_default_conf_cb) rx_adapter->default_cb_arg = 1; return 0; } @@ -928,7 +2037,7 @@ rte_event_eth_rx_adapter_create(uint8_t id, uint8_t dev_id, return -ENOMEM; *pc = *port_config; ret = rte_event_eth_rx_adapter_create_ext(id, dev_id, - default_conf_cb, + rxa_default_conf_cb, pc); if (ret) rte_free(pc); @@ -942,7 +2051,7 @@ rte_event_eth_rx_adapter_free(uint8_t id) RTE_EVENT_ETH_RX_ADAPTER_ID_VALID_OR_ERR_RET(id, -EINVAL); - rx_adapter = id_to_rx_adapter(id); + rx_adapter = rxa_id_to_adapter(id); if (rx_adapter == NULL) return -EINVAL; @@ -963,7 +2072,7 @@ rte_event_eth_rx_adapter_free(uint8_t id) int rte_event_eth_rx_adapter_queue_add(uint8_t id, - uint8_t eth_dev_id, + uint16_t eth_dev_id, int32_t rx_queue_id, const struct rte_event_eth_rx_adapter_queue_conf *queue_conf) { @@ -972,12 +2081,11 @@ rte_event_eth_rx_adapter_queue_add(uint8_t id, struct rte_event_eth_rx_adapter *rx_adapter; struct rte_eventdev *dev; struct eth_device_info *dev_info; - int start_service; RTE_EVENT_ETH_RX_ADAPTER_ID_VALID_OR_ERR_RET(id, -EINVAL); RTE_ETH_VALID_PORTID_OR_ERR_RET(eth_dev_id, -EINVAL); - rx_adapter = id_to_rx_adapter(id); + rx_adapter = rxa_id_to_adapter(id); if ((rx_adapter == NULL) || (queue_conf == NULL)) return -EINVAL; @@ -987,7 +2095,7 @@ rte_event_eth_rx_adapter_queue_add(uint8_t id, &cap); if (ret) { RTE_EDEV_LOG_ERR("Failed to get adapter caps edev %" PRIu8 - "eth port %" PRIu8, id, eth_dev_id); + "eth port %" PRIu16, id, eth_dev_id); return ret; } @@ -995,7 +2103,7 @@ rte_event_eth_rx_adapter_queue_add(uint8_t id, && (queue_conf->rx_queue_flags & RTE_EVENT_ETH_RX_ADAPTER_QUEUE_FLOW_ID_VALID)) { RTE_EDEV_LOG_ERR("Flow ID override is not supported," - " eth port: %" PRIu8 " adapter id: %" PRIu8, + " eth port: %" PRIu16 " adapter id: %" PRIu8, eth_dev_id, id); return -EINVAL; } @@ -1003,7 +2111,8 @@ rte_event_eth_rx_adapter_queue_add(uint8_t id, if ((cap & RTE_EVENT_ETH_RX_ADAPTER_CAP_MULTI_EVENTQ) == 0 && (rx_queue_id != -1)) { RTE_EDEV_LOG_ERR("Rx queues can only be connected to single " - "event queue id %u eth port %u", id, eth_dev_id); + "event queue, eth port: %" PRIu16 " adapter id: %" + PRIu8, eth_dev_id, id); return -EINVAL; } @@ -1014,7 +2123,6 @@ rte_event_eth_rx_adapter_queue_add(uint8_t id, return -EINVAL; } - start_service = 0; dev_info = &rx_adapter->eth_devices[eth_dev_id]; if (cap & RTE_EVENT_ETH_RX_ADAPTER_CAP_INTERNAL_PORT) { @@ -1034,33 +2142,34 @@ rte_event_eth_rx_adapter_queue_add(uint8_t id, &rte_eth_devices[eth_dev_id], rx_queue_id, queue_conf); if (ret == 0) { - update_queue_info(rx_adapter, + dev_info->internal_event_port = 1; + rxa_update_queue(rx_adapter, &rx_adapter->eth_devices[eth_dev_id], rx_queue_id, 1); } } else { rte_spinlock_lock(&rx_adapter->rx_lock); - ret = init_service(rx_adapter, id); - if (ret == 0) - ret = add_rx_queue(rx_adapter, eth_dev_id, rx_queue_id, + dev_info->internal_event_port = 0; + ret = rxa_init_service(rx_adapter, id); + if (ret == 0) { + uint32_t service_id = rx_adapter->service_id; + ret = rxa_sw_add(rx_adapter, eth_dev_id, rx_queue_id, queue_conf); + rte_service_component_runstate_set(service_id, + rxa_sw_adapter_queue_count(rx_adapter)); + } rte_spinlock_unlock(&rx_adapter->rx_lock); - if (ret == 0) - start_service = !!sw_rx_adapter_queue_count(rx_adapter); } if (ret) return ret; - if (start_service) - rte_service_component_runstate_set(rx_adapter->service_id, 1); - return 0; } int -rte_event_eth_rx_adapter_queue_del(uint8_t id, uint8_t eth_dev_id, +rte_event_eth_rx_adapter_queue_del(uint8_t id, uint16_t eth_dev_id, int32_t rx_queue_id) { int ret = 0; @@ -1068,12 +2177,17 @@ rte_event_eth_rx_adapter_queue_del(uint8_t id, uint8_t eth_dev_id, struct rte_event_eth_rx_adapter *rx_adapter; struct eth_device_info *dev_info; uint32_t cap; - uint16_t i; + uint32_t nb_rx_poll = 0; + uint32_t nb_wrr = 0; + uint32_t nb_rx_intr; + struct eth_rx_poll_entry *rx_poll = NULL; + uint32_t *rx_wrr = NULL; + int num_intr_vec; RTE_EVENT_ETH_RX_ADAPTER_ID_VALID_OR_ERR_RET(id, -EINVAL); RTE_ETH_VALID_PORTID_OR_ERR_RET(eth_dev_id, -EINVAL); - rx_adapter = id_to_rx_adapter(id); + rx_adapter = rxa_id_to_adapter(id); if (rx_adapter == NULL) return -EINVAL; @@ -1100,7 +2214,7 @@ rte_event_eth_rx_adapter_queue_del(uint8_t id, uint8_t eth_dev_id, &rte_eth_devices[eth_dev_id], rx_queue_id); if (ret == 0) { - update_queue_info(rx_adapter, + rxa_update_queue(rx_adapter, &rx_adapter->eth_devices[eth_dev_id], rx_queue_id, 0); @@ -1110,48 +2224,78 @@ rte_event_eth_rx_adapter_queue_del(uint8_t id, uint8_t eth_dev_id, } } } else { - int rc; + rxa_calc_nb_post_del(rx_adapter, dev_info, rx_queue_id, + &nb_rx_poll, &nb_rx_intr, &nb_wrr); + + ret = rxa_alloc_poll_arrays(rx_adapter, nb_rx_poll, nb_wrr, + &rx_poll, &rx_wrr); + if (ret) + return ret; + rte_spinlock_lock(&rx_adapter->rx_lock); - if (rx_queue_id == -1) { - for (i = 0; i < dev_info->dev->data->nb_rx_queues; i++) - event_eth_rx_adapter_queue_del(rx_adapter, - dev_info, - i); - } else { - event_eth_rx_adapter_queue_del(rx_adapter, - dev_info, - (uint16_t)rx_queue_id); + + num_intr_vec = 0; + if (rx_adapter->num_rx_intr > nb_rx_intr) { + + num_intr_vec = rxa_nb_intr_vect(dev_info, + rx_queue_id, 0); + ret = rxa_del_intr_queue(rx_adapter, dev_info, + rx_queue_id); + if (ret) + goto unlock_ret; + } + + if (nb_rx_intr == 0) { + ret = rxa_free_intr_resources(rx_adapter); + if (ret) + goto unlock_ret; + } + + rxa_sw_del(rx_adapter, dev_info, rx_queue_id); + rxa_calc_wrr_sequence(rx_adapter, rx_poll, rx_wrr); + + rte_free(rx_adapter->eth_rx_poll); + rte_free(rx_adapter->wrr_sched); + + if (nb_rx_intr == 0) { + rte_free(dev_info->intr_queue); + dev_info->intr_queue = NULL; } - rc = eth_poll_wrr_calc(rx_adapter); - if (rc) - RTE_EDEV_LOG_ERR("WRR recalculation failed %" PRId32, - rc); + rx_adapter->eth_rx_poll = rx_poll; + rx_adapter->wrr_sched = rx_wrr; + rx_adapter->wrr_len = nb_wrr; + rx_adapter->num_intr_vec += num_intr_vec; if (dev_info->nb_dev_queues == 0) { rte_free(dev_info->rx_queue); dev_info->rx_queue = NULL; } - +unlock_ret: rte_spinlock_unlock(&rx_adapter->rx_lock); + if (ret) { + rte_free(rx_poll); + rte_free(rx_wrr); + return ret; + } + rte_service_component_runstate_set(rx_adapter->service_id, - sw_rx_adapter_queue_count(rx_adapter)); + rxa_sw_adapter_queue_count(rx_adapter)); } return ret; } - int rte_event_eth_rx_adapter_start(uint8_t id) { - return rx_adapter_ctrl(id, 1); + return rxa_ctrl(id, 1); } int rte_event_eth_rx_adapter_stop(uint8_t id) { - return rx_adapter_ctrl(id, 0); + return rxa_ctrl(id, 0); } int @@ -1168,13 +2312,13 @@ rte_event_eth_rx_adapter_stats_get(uint8_t id, RTE_EVENT_ETH_RX_ADAPTER_ID_VALID_OR_ERR_RET(id, -EINVAL); - rx_adapter = id_to_rx_adapter(id); + rx_adapter = rxa_id_to_adapter(id); if (rx_adapter == NULL || stats == NULL) return -EINVAL; dev = &rte_eventdevs[rx_adapter->eventdev_id]; memset(stats, 0, sizeof(*stats)); - for (i = 0; i < rte_eth_dev_count(); i++) { + RTE_ETH_FOREACH_DEV(i) { dev_info = &rx_adapter->eth_devices[i]; if (dev_info->internal_event_port == 0 || dev->dev_ops->eth_rx_adapter_stats_get == NULL) @@ -1206,12 +2350,12 @@ rte_event_eth_rx_adapter_stats_reset(uint8_t id) RTE_EVENT_ETH_RX_ADAPTER_ID_VALID_OR_ERR_RET(id, -EINVAL); - rx_adapter = id_to_rx_adapter(id); + rx_adapter = rxa_id_to_adapter(id); if (rx_adapter == NULL) return -EINVAL; dev = &rte_eventdevs[rx_adapter->eventdev_id]; - for (i = 0; i < rte_eth_dev_count(); i++) { + RTE_ETH_FOREACH_DEV(i) { dev_info = &rx_adapter->eth_devices[i]; if (dev_info->internal_event_port == 0 || dev->dev_ops->eth_rx_adapter_stats_reset == NULL) @@ -1231,7 +2375,7 @@ rte_event_eth_rx_adapter_service_id_get(uint8_t id, uint32_t *service_id) RTE_EVENT_ETH_RX_ADAPTER_ID_VALID_OR_ERR_RET(id, -EINVAL); - rx_adapter = id_to_rx_adapter(id); + rx_adapter = rxa_id_to_adapter(id); if (rx_adapter == NULL || service_id == NULL) return -EINVAL; @@ -1240,3 +2384,47 @@ rte_event_eth_rx_adapter_service_id_get(uint8_t id, uint32_t *service_id) return rx_adapter->service_inited ? 0 : -ESRCH; } + +int rte_event_eth_rx_adapter_cb_register(uint8_t id, + uint16_t eth_dev_id, + rte_event_eth_rx_adapter_cb_fn cb_fn, + void *cb_arg) +{ + struct rte_event_eth_rx_adapter *rx_adapter; + struct eth_device_info *dev_info; + uint32_t cap; + int ret; + + RTE_EVENT_ETH_RX_ADAPTER_ID_VALID_OR_ERR_RET(id, -EINVAL); + RTE_ETH_VALID_PORTID_OR_ERR_RET(eth_dev_id, -EINVAL); + + rx_adapter = rxa_id_to_adapter(id); + if (rx_adapter == NULL) + return -EINVAL; + + dev_info = &rx_adapter->eth_devices[eth_dev_id]; + if (dev_info->rx_queue == NULL) + return -EINVAL; + + ret = rte_event_eth_rx_adapter_caps_get(rx_adapter->eventdev_id, + eth_dev_id, + &cap); + if (ret) { + RTE_EDEV_LOG_ERR("Failed to get adapter caps edev %" PRIu8 + "eth port %" PRIu16, id, eth_dev_id); + return ret; + } + + if (cap & RTE_EVENT_ETH_RX_ADAPTER_CAP_INTERNAL_PORT) { + RTE_EDEV_LOG_ERR("Rx callback not supported for eth port %" + PRIu16, eth_dev_id); + return -EINVAL; + } + + rte_spinlock_lock(&rx_adapter->rx_lock); + dev_info->cb_fn = cb_fn; + dev_info->cb_arg = cb_arg; + rte_spinlock_unlock(&rx_adapter->rx_lock); + + return 0; +} diff --git a/lib/librte_eventdev/rte_event_eth_rx_adapter.h b/lib/librte_eventdev/rte_event_eth_rx_adapter.h index c20507b2..332ee216 100644 --- a/lib/librte_eventdev/rte_event_eth_rx_adapter.h +++ b/lib/librte_eventdev/rte_event_eth_rx_adapter.h @@ -1,32 +1,6 @@ -/* - * Copyright(c) 2017 Intel Corporation. All rights reserved. - * All rights reserved. - * - * Redistribution and use in source and binary forms, with or without - * modification, are permitted provided that the following conditions - * are met: - * - * * Redistributions of source code must retain the above copyright - * notice, this list of conditions and the following disclaimer. - * * Redistributions in binary form must reproduce the above copyright - * notice, this list of conditions and the following disclaimer in - * the documentation and/or other materials provided with the - * distribution. - * * Neither the name of Intel Corporation nor the names of its - * contributors may be used to endorse or promote products derived - * from this software without specific prior written permission. - * - * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS - * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT - * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR - * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT - * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, - * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT - * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, - * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY - * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT - * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE - * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +/* SPDX-License-Identifier: BSD-3-Clause + * Copyright(c) 2017 Intel Corporation. + * All rights reserved. */ #ifndef _RTE_EVENT_ETH_RX_ADAPTER_ @@ -47,7 +21,11 @@ * * The adapter uses a EAL service core function for SW based packet transfer * and uses the eventdev PMD functions to configure HW based packet transfer - * between the ethernet device and the event device. + * between the ethernet device and the event device. For SW based packet + * transfer, if the mbuf does not have a timestamp set, the adapter adds a + * timestamp to the mbuf using rte_get_tsc_cycles(), this provides a more + * accurate timestamp as compared to if the application were to set the time + * stamp since it avoids event device schedule latency. * * The ethernet Rx event adapter's functions are: * - rte_event_eth_rx_adapter_create_ext() @@ -85,7 +63,23 @@ * rte_event_eth_rx_adapter_service_id_get() function can be used to retrieve * the service function ID of the adapter in this case. * - * Note: Interrupt driven receive queues are currently unimplemented. + * For SW based packet transfers, i.e., when the + * RTE_EVENT_ETH_RX_ADAPTER_CAP_INTERNAL_PORT is not set in the adapter's + * capabilities flags for a particular ethernet device, the service function + * temporarily enqueues mbufs to an event buffer before batch enqueueing these + * to the event device. If the buffer fills up, the service function stops + * dequeueing packets from the ethernet device. The application may want to + * monitor the buffer fill level and instruct the service function to + * selectively buffer packets. The application may also use some other + * criteria to decide which packets should enter the event device even when + * the event buffer fill level is low. The + * rte_event_eth_rx_adapter_cb_register() function allows the + * application to register a callback that selects which packets to enqueue + * to the event device. + * + * Note: + * 1) Devices created after an instance of rte_event_eth_rx_adapter_create + * should be added to a new instance of the rx adapter. */ #ifdef __cplusplus @@ -218,12 +212,55 @@ struct rte_event_eth_rx_adapter_stats { * block cycles can be used to compute the percentage of * cycles the service is blocked by the event device. */ + uint64_t rx_intr_packets; + /**< Received packet count for interrupt mode Rx queues */ }; /** * @warning * @b EXPERIMENTAL: this API may change without prior notice * + * Callback function invoked by the SW adapter before it continues + * to process packets. The callback is passed the size of the enqueue + * buffer in the SW adapter and the occupancy of the buffer. The + * callback can use these values to decide which mbufs should be + * enqueued to the event device. If the return value of the callback + * is less than nb_mbuf then the SW adapter uses the return value to + * enqueue enq_mbuf[] to the event device. + * + * @param eth_dev_id + * Port identifier of the Ethernet device. + * @param queue_id + * Receive queue index. + * @param enqueue_buf_size + * Total enqueue buffer size. + * @param enqueue_buf_count + * mbuf count in enqueue buffer. + * @param mbuf + * mbuf array. + * @param nb_mbuf + * mbuf count. + * @param cb_arg + * Callback argument. + * @param[out] enq_mbuf + * The adapter enqueues enq_mbuf[] if the return value of the + * callback is less than nb_mbuf + * @return + * Returns the number of mbufs should be enqueued to eventdev + */ +typedef uint16_t (*rte_event_eth_rx_adapter_cb_fn)(uint16_t eth_dev_id, + uint16_t queue_id, + uint32_t enqueue_buf_size, + uint32_t enqueue_buf_count, + struct rte_mbuf **mbuf, + uint16_t nb_mbuf, + void *cb_arg, + struct rte_mbuf **enq_buf); + +/** + * @warning + * @b EXPERIMENTAL: this API may change without prior notice + * * Create a new ethernet Rx event adapter with the specified identifier. * * @param id @@ -329,7 +366,7 @@ int rte_event_eth_rx_adapter_free(uint8_t id); * combination of the two error codes. */ int rte_event_eth_rx_adapter_queue_add(uint8_t id, - uint8_t eth_dev_id, + uint16_t eth_dev_id, int32_t rx_queue_id, const struct rte_event_eth_rx_adapter_queue_conf *conf); @@ -357,7 +394,7 @@ int rte_event_eth_rx_adapter_queue_add(uint8_t id, * - 0: Success, Receive queue deleted correctly. * - <0: Error code on failure. */ -int rte_event_eth_rx_adapter_queue_del(uint8_t id, uint8_t eth_dev_id, +int rte_event_eth_rx_adapter_queue_del(uint8_t id, uint16_t eth_dev_id, int32_t rx_queue_id); /** @@ -444,6 +481,32 @@ int rte_event_eth_rx_adapter_stats_reset(uint8_t id); */ int rte_event_eth_rx_adapter_service_id_get(uint8_t id, uint32_t *service_id); +/** + * @warning + * @b EXPERIMENTAL: this API may change without prior notice + * + * Register callback to process Rx packets, this is supported for + * SW based packet transfers. + * @see rte_event_eth_rx_cb_fn + * + * @param id + * Adapter identifier. + * @param eth_dev_id + * Port identifier of Ethernet device. + * @param cb_fn + * Callback function. + * @param cb_arg + * Callback arg. + * @return + * - 0: Success + * - <0: Error code on failure. + */ +int __rte_experimental +rte_event_eth_rx_adapter_cb_register(uint8_t id, + uint16_t eth_dev_id, + rte_event_eth_rx_adapter_cb_fn cb_fn, + void *cb_arg); + #ifdef __cplusplus } #endif diff --git a/lib/librte_eventdev/rte_event_ring.c b/lib/librte_eventdev/rte_event_ring.c index eb67751d..16d02a95 100644 --- a/lib/librte_eventdev/rte_event_ring.c +++ b/lib/librte_eventdev/rte_event_ring.c @@ -82,11 +82,16 @@ rte_event_ring_create(const char *name, unsigned int count, int socket_id, mz = rte_memzone_reserve(mz_name, ring_size, socket_id, mz_flags); if (mz != NULL) { r = mz->addr; - /* - * no need to check return value here, we already checked the - * arguments above - */ - rte_event_ring_init(r, name, requested_count, flags); + /* Check return value in case rte_ring_init() fails on size */ + int err = rte_event_ring_init(r, name, requested_count, flags); + if (err) { + RTE_LOG(ERR, RING, "Ring init failed\n"); + if (rte_memzone_free(mz) != 0) + RTE_LOG(ERR, RING, "Cannot free memzone\n"); + rte_free(te); + rte_rwlock_write_unlock(RTE_EAL_TAILQ_RWLOCK); + return NULL; + } te->data = (void *) r; r->r.memzone = mz; diff --git a/lib/librte_eventdev/rte_event_ring.h b/lib/librte_eventdev/rte_event_ring.h index 29d4228a..827a3209 100644 --- a/lib/librte_eventdev/rte_event_ring.h +++ b/lib/librte_eventdev/rte_event_ring.h @@ -99,7 +99,7 @@ rte_event_ring_enqueue_burst(struct rte_event_ring *r, ENQUEUE_PTRS(&r->r, &r[1], prod_head, events, n, struct rte_event); - update_tail(&r->r.prod, prod_head, prod_next, 1, 1); + update_tail(&r->r.prod, prod_head, prod_next, r->r.prod.single, 1); end: if (free_space != NULL) *free_space = free_entries - n; @@ -140,7 +140,7 @@ rte_event_ring_dequeue_burst(struct rte_event_ring *r, DEQUEUE_PTRS(&r->r, &r[1], cons_head, events, n, struct rte_event); - update_tail(&r->r.cons, cons_head, cons_next, 1, 0); + update_tail(&r->r.cons, cons_head, cons_next, r->r.cons.single, 0); end: if (available != NULL) diff --git a/lib/librte_eventdev/rte_event_timer_adapter.c b/lib/librte_eventdev/rte_event_timer_adapter.c new file mode 100644 index 00000000..79070d48 --- /dev/null +++ b/lib/librte_eventdev/rte_event_timer_adapter.c @@ -0,0 +1,1299 @@ +/* SPDX-License-Identifier: BSD-3-Clause + * Copyright(c) 2017-2018 Intel Corporation. + * All rights reserved. + */ + +#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_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; +static int evtim_svc_logtype; +static int evtim_buffer_logtype; + +static struct rte_event_timer_adapter adapters[RTE_EVENT_TIMER_ADAPTER_NUM_MAX]; + +static const struct rte_event_timer_adapter_ops sw_event_adapter_timer_ops; + +#define EVTIM_LOG(level, logtype, ...) \ + rte_log(RTE_LOG_ ## level, logtype, \ + RTE_FMT("EVTIMER: %s() line %u: " RTE_FMT_HEAD(__VA_ARGS__,) \ + "\n", __func__, __LINE__, RTE_FMT_TAIL(__VA_ARGS__,))) + +#define EVTIM_LOG_ERR(...) EVTIM_LOG(ERR, evtim_logtype, __VA_ARGS__) + +#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 +default_port_conf_cb(uint16_t id, uint8_t event_dev_id, uint8_t *event_port_id, + void *conf_arg) +{ + struct rte_event_timer_adapter *adapter; + struct rte_eventdev *dev; + struct rte_event_dev_config dev_conf; + struct rte_event_port_conf *port_conf, def_port_conf = {0}; + int started; + uint8_t port_id; + uint8_t dev_id; + int ret; + + RTE_SET_USED(event_dev_id); + + adapter = &adapters[id]; + dev = &rte_eventdevs[adapter->data->event_dev_id]; + dev_id = dev->data->dev_id; + dev_conf = dev->data->dev_conf; + + started = dev->data->dev_started; + if (started) + rte_event_dev_stop(dev_id); + + port_id = dev_conf.nb_event_ports; + dev_conf.nb_event_ports += 1; + ret = rte_event_dev_configure(dev_id, &dev_conf); + if (ret < 0) { + EVTIM_LOG_ERR("failed to configure event dev %u\n", dev_id); + if (started) + if (rte_event_dev_start(dev_id)) + return -EIO; + + return ret; + } + + if (conf_arg != NULL) + port_conf = conf_arg; + else { + port_conf = &def_port_conf; + ret = rte_event_port_default_conf_get(dev_id, port_id, + port_conf); + if (ret < 0) + return ret; + } + + ret = rte_event_port_setup(dev_id, port_id, port_conf); + if (ret < 0) { + EVTIM_LOG_ERR("failed to setup event port %u on event dev %u\n", + port_id, dev_id); + return ret; + } + + *event_port_id = port_id; + + if (started) + ret = rte_event_dev_start(dev_id); + + return ret; +} + +struct rte_event_timer_adapter * __rte_experimental +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 +rte_event_timer_adapter_create_ext( + const struct rte_event_timer_adapter_conf *conf, + rte_event_timer_adapter_port_conf_cb_t conf_cb, + void *conf_arg) +{ + uint16_t adapter_id; + struct rte_event_timer_adapter *adapter; + const struct rte_memzone *mz; + char mz_name[DATA_MZ_NAME_MAX_LEN]; + int n, ret; + struct rte_eventdev *dev; + + if (conf == NULL) { + rte_errno = EINVAL; + return NULL; + } + + /* Check eventdev ID */ + if (!rte_event_pmd_is_valid_dev(conf->event_dev_id)) { + rte_errno = EINVAL; + return NULL; + } + dev = &rte_eventdevs[conf->event_dev_id]; + + adapter_id = conf->timer_adapter_id; + + /* Check that adapter_id is in range */ + if (adapter_id >= RTE_EVENT_TIMER_ADAPTER_NUM_MAX) { + rte_errno = EINVAL; + return NULL; + } + + /* Check adapter ID not already allocated */ + adapter = &adapters[adapter_id]; + if (adapter->allocated) { + rte_errno = EEXIST; + return NULL; + } + + /* Create shared data area. */ + n = snprintf(mz_name, sizeof(mz_name), DATA_MZ_NAME_FORMAT, adapter_id); + if (n >= (int)sizeof(mz_name)) { + rte_errno = EINVAL; + return NULL; + } + mz = rte_memzone_reserve(mz_name, + sizeof(struct rte_event_timer_adapter_data), + conf->socket_id, 0); + if (mz == NULL) + /* rte_errno set by rte_memzone_reserve */ + return NULL; + + adapter->data = mz->addr; + memset(adapter->data, 0, sizeof(struct rte_event_timer_adapter_data)); + + adapter->data->mz = mz; + adapter->data->event_dev_id = conf->event_dev_id; + adapter->data->id = adapter_id; + adapter->data->socket_id = conf->socket_id; + adapter->data->conf = *conf; /* copy conf structure */ + + /* Query eventdev PMD for timer adapter capabilities and ops */ + ret = dev->dev_ops->timer_adapter_caps_get(dev, + adapter->data->conf.flags, + &adapter->data->caps, + &adapter->ops); + if (ret < 0) { + 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); + ret = conf_cb(adapter->data->id, adapter->data->event_dev_id, + &adapter->data->event_port_id, conf_arg); + if (ret < 0) { + rte_errno = ret; + goto free_memzone; + } + } + + /* If eventdev PMD did not provide ops, use default software + * implementation. + */ + if (adapter->ops == NULL) + adapter->ops = &sw_event_adapter_timer_ops; + + /* Allow driver to do some setup */ + FUNC_PTR_OR_NULL_RET_WITH_ERRNO(adapter->ops->init, -ENOTSUP); + ret = adapter->ops->init(adapter); + if (ret < 0) { + rte_errno = ret; + goto free_memzone; + } + + /* Set fast-path function pointers */ + adapter->arm_burst = adapter->ops->arm_burst; + adapter->arm_tmo_tick_burst = adapter->ops->arm_tmo_tick_burst; + adapter->cancel_burst = adapter->ops->cancel_burst; + + adapter->allocated = 1; + + return adapter; + +free_memzone: + rte_memzone_free(adapter->data->mz); + return NULL; +} + +int __rte_experimental +rte_event_timer_adapter_get_info(const struct rte_event_timer_adapter *adapter, + struct rte_event_timer_adapter_info *adapter_info) +{ + ADAPTER_VALID_OR_ERR_RET(adapter, -EINVAL); + + if (adapter->ops->get_info) + /* let driver set values it knows */ + adapter->ops->get_info(adapter, adapter_info); + + /* Set common values */ + adapter_info->conf = adapter->data->conf; + adapter_info->event_dev_port_id = adapter->data->event_port_id; + adapter_info->caps = adapter->data->caps; + + return 0; +} + +int __rte_experimental +rte_event_timer_adapter_start(const struct rte_event_timer_adapter *adapter) +{ + int ret; + + ADAPTER_VALID_OR_ERR_RET(adapter, -EINVAL); + FUNC_PTR_OR_ERR_RET(adapter->ops->start, -EINVAL); + + ret = adapter->ops->start(adapter); + if (ret < 0) + return ret; + + adapter->data->started = 1; + + return 0; +} + +int __rte_experimental +rte_event_timer_adapter_stop(const struct rte_event_timer_adapter *adapter) +{ + int ret; + + ADAPTER_VALID_OR_ERR_RET(adapter, -EINVAL); + FUNC_PTR_OR_ERR_RET(adapter->ops->stop, -EINVAL); + + if (adapter->data->started == 0) { + EVTIM_LOG_ERR("event timer adapter %"PRIu8" already stopped", + adapter->data->id); + return 0; + } + + ret = adapter->ops->stop(adapter); + if (ret < 0) + return ret; + + adapter->data->started = 0; + + return 0; +} + +struct rte_event_timer_adapter * __rte_experimental +rte_event_timer_adapter_lookup(uint16_t adapter_id) +{ + char name[DATA_MZ_NAME_MAX_LEN]; + const struct rte_memzone *mz; + struct rte_event_timer_adapter_data *data; + struct rte_event_timer_adapter *adapter; + int ret; + struct rte_eventdev *dev; + + if (adapters[adapter_id].allocated) + return &adapters[adapter_id]; /* Adapter is already loaded */ + + snprintf(name, DATA_MZ_NAME_MAX_LEN, DATA_MZ_NAME_FORMAT, adapter_id); + mz = rte_memzone_lookup(name); + if (mz == NULL) { + rte_errno = ENOENT; + return NULL; + } + + data = mz->addr; + + adapter = &adapters[data->id]; + adapter->data = data; + + dev = &rte_eventdevs[adapter->data->event_dev_id]; + + /* Query eventdev PMD for timer adapter capabilities and ops */ + ret = dev->dev_ops->timer_adapter_caps_get(dev, + adapter->data->conf.flags, + &adapter->data->caps, + &adapter->ops); + if (ret < 0) { + rte_errno = EINVAL; + return NULL; + } + + /* If eventdev PMD did not provide ops, use default software + * implementation. + */ + if (adapter->ops == NULL) + adapter->ops = &sw_event_adapter_timer_ops; + + /* Set fast-path function pointers */ + adapter->arm_burst = adapter->ops->arm_burst; + adapter->arm_tmo_tick_burst = adapter->ops->arm_tmo_tick_burst; + adapter->cancel_burst = adapter->ops->cancel_burst; + + adapter->allocated = 1; + + return adapter; +} + +int __rte_experimental +rte_event_timer_adapter_free(struct rte_event_timer_adapter *adapter) +{ + int ret; + + ADAPTER_VALID_OR_ERR_RET(adapter, -EINVAL); + FUNC_PTR_OR_ERR_RET(adapter->ops->uninit, -EINVAL); + + if (adapter->data->started == 1) { + EVTIM_LOG_ERR("event timer adapter %"PRIu8" must be stopped " + "before freeing", adapter->data->id); + return -EBUSY; + } + + /* free impl priv data */ + ret = adapter->ops->uninit(adapter); + if (ret < 0) + return ret; + + /* free shared data area */ + ret = rte_memzone_free(adapter->data->mz); + if (ret < 0) + return ret; + + adapter->data = NULL; + adapter->allocated = 0; + + return 0; +} + +int __rte_experimental +rte_event_timer_adapter_service_id_get(struct rte_event_timer_adapter *adapter, + uint32_t *service_id) +{ + ADAPTER_VALID_OR_ERR_RET(adapter, -EINVAL); + + if (adapter->data->service_inited && service_id != NULL) + *service_id = adapter->data->service_id; + + return adapter->data->service_inited ? 0 : -ESRCH; +} + +int __rte_experimental +rte_event_timer_adapter_stats_get(struct rte_event_timer_adapter *adapter, + struct rte_event_timer_adapter_stats *stats) +{ + ADAPTER_VALID_OR_ERR_RET(adapter, -EINVAL); + FUNC_PTR_OR_ERR_RET(adapter->ops->stats_get, -EINVAL); + if (stats == NULL) + return -EINVAL; + + return adapter->ops->stats_get(adapter, stats); +} + +int __rte_experimental +rte_event_timer_adapter_stats_reset(struct rte_event_timer_adapter *adapter) +{ + ADAPTER_VALID_OR_ERR_RET(adapter, -EINVAL); + FUNC_PTR_OR_ERR_RET(adapter->ops->stats_reset, -EINVAL); + return adapter->ops->stats_reset(adapter); +} + +/* + * 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) + +struct event_buffer { + uint16_t head; + uint16_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_buffer_init(struct event_buffer *bufp) +{ + 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) +{ + uint16_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) +{ + uint16_t head_idx, tail_idx, n = 0; + struct rte_event *events = bufp->events; + + /* Instead of modulus, bitwise AND with mask to get index. */ + head_idx = bufp->head & EVENT_BUFFER_MASK; + tail_idx = bufp->tail & EVENT_BUFFER_MASK; + + /* 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 { + *nb_events_flushed = 0; + return; + } + + *nb_events_inv = 0; + *nb_events_flushed = rte_event_enqueue_burst(dev_id, port_id, + &events[tail_idx], n); + if (*nb_events_flushed != n && rte_errno == -EINVAL) { + EVTIM_LOG_ERR("failed to enqueue invalid event - dropping it"); + (*nb_events_inv)++; + } + + bufp->tail = bufp->tail + *nb_events_flushed + *nb_events_inv; +} + +/* + * Software event timer adapter implementation + */ + +struct rte_event_timer_adapter_sw_data { + /* List of messages for outstanding timers */ + TAILQ_HEAD(, msg) msgs_tailq_head; + /* Lock to guard tailq and armed count */ + rte_spinlock_t msgs_tailq_sl; + /* 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; + /* Incremented as the service moves through phases of an iteration */ + volatile int service_phase; + /* 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; + /* Ring containing messages to arm or cancel event timers */ + struct rte_ring *msg_ring; + /* Mempool containing msg objects */ + struct rte_mempool *msg_pool; + /* Buffered timer expiry events to be enqueued to an event device. */ + struct event_buffer buffer; + /* Statistics */ + struct rte_event_timer_adapter_stats stats; + /* The number of threads currently adding to the message ring */ + rte_atomic16_t message_producer_count; +}; + +enum msg_type {MSG_TYPE_ARM, MSG_TYPE_CANCEL}; + +struct msg { + enum msg_type type; + struct rte_event_timer *evtim; + struct rte_timer tim; + TAILQ_ENTRY(msg) msgs; +}; + +static void +sw_event_timer_cb(struct rte_timer *tim, void *arg) +{ + int ret; + uint16_t nb_evs_flushed = 0; + uint16_t nb_evs_invalid = 0; + uint64_t opaque; + struct rte_event_timer *evtim; + struct rte_event_timer_adapter *adapter; + struct rte_event_timer_adapter_sw_data *sw_data; + + evtim = arg; + opaque = evtim->impl_opaque[1]; + adapter = (struct rte_event_timer_adapter *)(uintptr_t)opaque; + sw_data = adapter->data->adapter_priv; + + ret = event_buffer_add(&sw_data->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. + */ + rte_timer_reset_sync(tim, 0, SINGLE, rte_lcore_id(), + sw_event_timer_cb, evtim); + + sw_data->stats.evtim_retry_count++; + EVTIM_LOG_DBG("event buffer full, resetting rte_timer with " + "immediate expiry value"); + } else { + struct msg *m = container_of(tim, struct msg, tim); + TAILQ_REMOVE(&sw_data->msgs_tailq_head, m, msgs); + EVTIM_BUF_LOG_DBG("buffered an event timer expiry event"); + evtim->state = RTE_EVENT_TIMER_NOT_ARMED; + + /* Free the msg object containing the rte_timer now that + * we've buffered its event successfully. + */ + rte_mempool_put(sw_data->msg_pool, m); + + /* Bump the count when we successfully add an expiry event to + * the buffer. + */ + sw_data->stats.evtim_exp_count++; + } + + if (event_buffer_batch_ready(&sw_data->buffer)) { + event_buffer_flush(&sw_data->buffer, + adapter->data->event_dev_id, + adapter->data->event_port_id, + &nb_evs_flushed, + &nb_evs_invalid); + + sw_data->stats.ev_enq_count += nb_evs_flushed; + sw_data->stats.ev_inv_count += nb_evs_invalid; + } +} + +static __rte_always_inline uint64_t +get_timeout_cycles(struct rte_event_timer *evtim, + struct rte_event_timer_adapter *adapter) +{ + uint64_t timeout_ns; + struct rte_event_timer_adapter_sw_data *sw_data; + + sw_data = adapter->data->adapter_priv; + timeout_ns = evtim->timeout_ticks * sw_data->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 +adapter_did_tick(struct rte_event_timer_adapter *adapter) +{ + uint64_t cycles_per_adapter_tick, start_cycles; + uint64_t *next_tick_cyclesp; + struct rte_event_timer_adapter_sw_data *sw_data; + + sw_data = adapter->data->adapter_priv; + next_tick_cyclesp = &sw_data->next_tick_cycles; + + cycles_per_adapter_tick = sw_data->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 rte_event_timer_adapter_sw_data *sw_data; + + sw_data = adapter->data->adapter_priv; + tmo_nsec = evtim->timeout_ticks * sw_data->timer_tick_ns; + + if (tmo_nsec > sw_data->max_tmo_ns) + return -1; + + if (tmo_nsec < sw_data->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 && ret != -EOVERFLOW) || + evtim->ev.sched_type != sched_type) + return -1; + + return 0; +} + +#define NB_OBJS 32 +static int +sw_event_timer_adapter_service_func(void *arg) +{ + int i, num_msgs; + uint64_t cycles, opaque; + uint16_t nb_evs_flushed = 0; + uint16_t nb_evs_invalid = 0; + struct rte_event_timer_adapter *adapter; + struct rte_event_timer_adapter_sw_data *sw_data; + struct rte_event_timer *evtim = NULL; + struct rte_timer *tim = NULL; + struct msg *msg, *msgs[NB_OBJS]; + + adapter = arg; + sw_data = adapter->data->adapter_priv; + + sw_data->service_phase = 1; + rte_smp_wmb(); + + while (rte_atomic16_read(&sw_data->message_producer_count) > 0 || + !rte_ring_empty(sw_data->msg_ring)) { + + num_msgs = rte_ring_dequeue_burst(sw_data->msg_ring, + (void **)msgs, NB_OBJS, NULL); + + for (i = 0; i < num_msgs; i++) { + int ret = 0; + + RTE_SET_USED(ret); + + msg = msgs[i]; + evtim = msg->evtim; + + switch (msg->type) { + case MSG_TYPE_ARM: + EVTIM_SVC_LOG_DBG("dequeued ARM message from " + "ring"); + tim = &msg->tim; + rte_timer_init(tim); + cycles = get_timeout_cycles(evtim, + adapter); + ret = rte_timer_reset(tim, cycles, SINGLE, + rte_lcore_id(), + sw_event_timer_cb, + evtim); + RTE_ASSERT(ret == 0); + + evtim->impl_opaque[0] = (uintptr_t)tim; + evtim->impl_opaque[1] = (uintptr_t)adapter; + + TAILQ_INSERT_TAIL(&sw_data->msgs_tailq_head, + msg, + msgs); + break; + case MSG_TYPE_CANCEL: + EVTIM_SVC_LOG_DBG("dequeued CANCEL message " + "from ring"); + opaque = evtim->impl_opaque[0]; + tim = (struct rte_timer *)(uintptr_t)opaque; + RTE_ASSERT(tim != NULL); + + ret = rte_timer_stop(tim); + RTE_ASSERT(ret == 0); + + /* Free the msg object for the original arm + * request. + */ + struct msg *m; + m = container_of(tim, struct msg, tim); + TAILQ_REMOVE(&sw_data->msgs_tailq_head, m, + msgs); + rte_mempool_put(sw_data->msg_pool, m); + + /* Free the msg object for the current msg */ + rte_mempool_put(sw_data->msg_pool, msg); + + evtim->impl_opaque[0] = 0; + evtim->impl_opaque[1] = 0; + + break; + } + } + } + + sw_data->service_phase = 2; + rte_smp_wmb(); + + if (adapter_did_tick(adapter)) { + rte_timer_manage(); + + event_buffer_flush(&sw_data->buffer, + adapter->data->event_dev_id, + adapter->data->event_port_id, + &nb_evs_flushed, &nb_evs_invalid); + + sw_data->stats.ev_enq_count += nb_evs_flushed; + sw_data->stats.ev_inv_count += nb_evs_invalid; + sw_data->stats.adapter_tick_count++; + } + + sw_data->service_phase = 0; + rte_smp_wmb(); + + 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; +} + +#define SW_MIN_INTERVAL 1E5 + +static int +sw_event_timer_adapter_init(struct rte_event_timer_adapter *adapter) +{ + int ret; + struct rte_event_timer_adapter_sw_data *sw_data; + uint64_t nb_timers; + unsigned int flags; + struct rte_service_spec service; + static bool timer_subsystem_inited; // static initialized to false + + /* Allocate storage for SW implementation data */ + char priv_data_name[RTE_RING_NAMESIZE]; + snprintf(priv_data_name, RTE_RING_NAMESIZE, "sw_evtim_adap_priv_%"PRIu8, + adapter->data->id); + adapter->data->adapter_priv = rte_zmalloc_socket( + priv_data_name, + sizeof(struct rte_event_timer_adapter_sw_data), + RTE_CACHE_LINE_SIZE, + adapter->data->socket_id); + if (adapter->data->adapter_priv == NULL) { + EVTIM_LOG_ERR("failed to allocate space for private data"); + rte_errno = ENOMEM; + return -1; + } + + if (adapter->data->conf.timer_tick_ns < SW_MIN_INTERVAL) { + EVTIM_LOG_ERR("failed to create adapter with requested tick " + "interval"); + rte_errno = EINVAL; + return -1; + } + + sw_data = adapter->data->adapter_priv; + + sw_data->timer_tick_ns = adapter->data->conf.timer_tick_ns; + sw_data->max_tmo_ns = adapter->data->conf.max_tmo_ns; + + TAILQ_INIT(&sw_data->msgs_tailq_head); + rte_spinlock_init(&sw_data->msgs_tailq_sl); + rte_atomic16_init(&sw_data->message_producer_count); + + /* Rings require power of 2, so round up to next such value */ + nb_timers = rte_align64pow2(adapter->data->conf.nb_timers); + + char msg_ring_name[RTE_RING_NAMESIZE]; + snprintf(msg_ring_name, RTE_RING_NAMESIZE, + "sw_evtim_adap_msg_ring_%"PRIu8, adapter->data->id); + flags = adapter->data->conf.flags & RTE_EVENT_TIMER_ADAPTER_F_SP_PUT ? + RING_F_SP_ENQ | RING_F_SC_DEQ : + RING_F_SC_DEQ; + sw_data->msg_ring = rte_ring_create(msg_ring_name, nb_timers, + adapter->data->socket_id, flags); + if (sw_data->msg_ring == NULL) { + EVTIM_LOG_ERR("failed to create message ring"); + rte_errno = ENOMEM; + goto free_priv_data; + } + + char pool_name[RTE_RING_NAMESIZE]; + snprintf(pool_name, RTE_RING_NAMESIZE, "sw_evtim_adap_msg_pool_%"PRIu8, + adapter->data->id); + + /* Both the arming/canceling thread and the service thread will do puts + * to the mempool, but if the SP_PUT flag is enabled, we can specify + * single-consumer get for the mempool. + */ + flags = adapter->data->conf.flags & RTE_EVENT_TIMER_ADAPTER_F_SP_PUT ? + MEMPOOL_F_SC_GET : 0; + + /* The usable size of a ring is count - 1, so subtract one here to + * make the counts agree. + */ + int pool_size = nb_timers - 1; + int cache_size = compute_msg_mempool_cache_size( + adapter->data->conf.nb_timers, nb_timers); + sw_data->msg_pool = rte_mempool_create(pool_name, pool_size, + sizeof(struct msg), cache_size, + 0, NULL, NULL, NULL, NULL, + adapter->data->socket_id, flags); + if (sw_data->msg_pool == NULL) { + EVTIM_LOG_ERR("failed to create message object mempool"); + rte_errno = ENOMEM; + goto free_msg_ring; + } + + event_buffer_init(&sw_data->buffer); + + /* Register a service component to run adapter logic */ + memset(&service, 0, sizeof(service)); + snprintf(service.name, RTE_SERVICE_NAME_MAX, + "sw_evimer_adap_svc_%"PRIu8, adapter->data->id); + service.socket_id = adapter->data->socket_id; + service.callback = sw_event_timer_adapter_service_func; + service.callback_userdata = adapter; + service.capabilities &= ~(RTE_SERVICE_CAP_MT_SAFE); + ret = rte_service_component_register(&service, &sw_data->service_id); + if (ret < 0) { + EVTIM_LOG_ERR("failed to register service %s with id %"PRIu32 + ": err = %d", service.name, sw_data->service_id, + ret); + + rte_errno = ENOSPC; + goto free_msg_pool; + } + + EVTIM_LOG_DBG("registered service %s with id %"PRIu32, service.name, + sw_data->service_id); + + adapter->data->service_id = sw_data->service_id; + adapter->data->service_inited = 1; + + if (!timer_subsystem_inited) { + rte_timer_subsystem_init(); + timer_subsystem_inited = true; + } + + return 0; + +free_msg_pool: + rte_mempool_free(sw_data->msg_pool); +free_msg_ring: + rte_ring_free(sw_data->msg_ring); +free_priv_data: + rte_free(sw_data); + return -1; +} + +static int +sw_event_timer_adapter_uninit(struct rte_event_timer_adapter *adapter) +{ + int ret; + struct msg *m1, *m2; + struct rte_event_timer_adapter_sw_data *sw_data = + adapter->data->adapter_priv; + + rte_spinlock_lock(&sw_data->msgs_tailq_sl); + + /* Cancel outstanding rte_timers and free msg objects */ + m1 = TAILQ_FIRST(&sw_data->msgs_tailq_head); + while (m1 != NULL) { + EVTIM_LOG_DBG("freeing outstanding timer"); + m2 = TAILQ_NEXT(m1, msgs); + + rte_timer_stop_sync(&m1->tim); + rte_mempool_put(sw_data->msg_pool, m1); + + m1 = m2; + } + + rte_spinlock_unlock(&sw_data->msgs_tailq_sl); + + ret = rte_service_component_unregister(sw_data->service_id); + if (ret < 0) { + EVTIM_LOG_ERR("failed to unregister service component"); + return ret; + } + + rte_ring_free(sw_data->msg_ring); + rte_mempool_free(sw_data->msg_pool); + rte_free(adapter->data->adapter_priv); + + 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 +sw_event_timer_adapter_start(const struct rte_event_timer_adapter *adapter) +{ + int mapped_count; + struct rte_event_timer_adapter_sw_data *sw_data; + + sw_data = adapter->data->adapter_priv; + + /* 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. + */ + mapped_count = get_mapped_count_for_service(sw_data->service_id); + + if (mapped_count == 1) + return rte_service_component_runstate_set(sw_data->service_id, + 1); + + return mapped_count < 1 ? -ENOENT : -ENOTSUP; +} + +static int +sw_event_timer_adapter_stop(const struct rte_event_timer_adapter *adapter) +{ + int ret; + struct rte_event_timer_adapter_sw_data *sw_data = + adapter->data->adapter_priv; + + ret = rte_service_component_runstate_set(sw_data->service_id, 0); + if (ret < 0) + return ret; + + /* Wait for the service to complete its final iteration before + * stopping. + */ + while (sw_data->service_phase != 0) + rte_pause(); + + rte_smp_rmb(); + + return 0; +} + +static void +sw_event_timer_adapter_get_info(const struct rte_event_timer_adapter *adapter, + struct rte_event_timer_adapter_info *adapter_info) +{ + struct rte_event_timer_adapter_sw_data *sw_data; + sw_data = adapter->data->adapter_priv; + + adapter_info->min_resolution_ns = sw_data->timer_tick_ns; + adapter_info->max_tmo_ns = sw_data->max_tmo_ns; +} + +static int +sw_event_timer_adapter_stats_get(const struct rte_event_timer_adapter *adapter, + struct rte_event_timer_adapter_stats *stats) +{ + struct rte_event_timer_adapter_sw_data *sw_data; + sw_data = adapter->data->adapter_priv; + *stats = sw_data->stats; + return 0; +} + +static int +sw_event_timer_adapter_stats_reset( + const struct rte_event_timer_adapter *adapter) +{ + struct rte_event_timer_adapter_sw_data *sw_data; + sw_data = adapter->data->adapter_priv; + memset(&sw_data->stats, 0, sizeof(sw_data->stats)); + return 0; +} + +static __rte_always_inline uint16_t +__sw_event_timer_arm_burst(const struct rte_event_timer_adapter *adapter, + struct rte_event_timer **evtims, + uint16_t nb_evtims) +{ + uint16_t i; + int ret; + struct rte_event_timer_adapter_sw_data *sw_data; + struct msg *msgs[nb_evtims]; + +#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 + + sw_data = adapter->data->adapter_priv; + + ret = rte_mempool_get_bulk(sw_data->msg_pool, (void **)msgs, nb_evtims); + if (ret < 0) { + rte_errno = ENOSPC; + return 0; + } + + /* Let the service know we're producing messages for it to process */ + rte_atomic16_inc(&sw_data->message_producer_count); + + /* If the service is managing timers, wait for it to finish */ + while (sw_data->service_phase == 2) + rte_pause(); + + rte_smp_rmb(); + + for (i = 0; i < nb_evtims; i++) { + /* Don't modify the event timer state in these cases */ + if (evtims[i]->state == RTE_EVENT_TIMER_ARMED) { + rte_errno = EALREADY; + break; + } else if (!(evtims[i]->state == RTE_EVENT_TIMER_NOT_ARMED || + evtims[i]->state == RTE_EVENT_TIMER_CANCELED)) { + rte_errno = EINVAL; + break; + } + + ret = check_timeout(evtims[i], adapter); + if (ret == -1) { + evtims[i]->state = RTE_EVENT_TIMER_ERROR_TOOLATE; + rte_errno = EINVAL; + break; + } + if (ret == -2) { + evtims[i]->state = RTE_EVENT_TIMER_ERROR_TOOEARLY; + rte_errno = EINVAL; + break; + } + + if (check_destination_event_queue(evtims[i], adapter) < 0) { + evtims[i]->state = RTE_EVENT_TIMER_ERROR; + rte_errno = EINVAL; + break; + } + + /* Checks passed, set up a message to enqueue */ + msgs[i]->type = MSG_TYPE_ARM; + msgs[i]->evtim = evtims[i]; + + /* Set the payload pointer if not set. */ + if (evtims[i]->ev.event_ptr == NULL) + evtims[i]->ev.event_ptr = evtims[i]; + + /* msg objects that get enqueued successfully will be freed + * either by a future cancel operation or by the timer + * expiration callback. + */ + if (rte_ring_enqueue(sw_data->msg_ring, msgs[i]) < 0) { + rte_errno = ENOSPC; + break; + } + + EVTIM_LOG_DBG("enqueued ARM message to ring"); + + evtims[i]->state = RTE_EVENT_TIMER_ARMED; + } + + /* Let the service know we're done producing messages */ + rte_atomic16_dec(&sw_data->message_producer_count); + + if (i < nb_evtims) + rte_mempool_put_bulk(sw_data->msg_pool, (void **)&msgs[i], + nb_evtims - i); + + return i; +} + +static uint16_t +sw_event_timer_arm_burst(const struct rte_event_timer_adapter *adapter, + struct rte_event_timer **evtims, + uint16_t nb_evtims) +{ + return __sw_event_timer_arm_burst(adapter, evtims, nb_evtims); +} + +static uint16_t +sw_event_timer_cancel_burst(const struct rte_event_timer_adapter *adapter, + struct rte_event_timer **evtims, + uint16_t nb_evtims) +{ + uint16_t i; + int ret; + struct rte_event_timer_adapter_sw_data *sw_data; + struct msg *msgs[nb_evtims]; + +#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 + + sw_data = adapter->data->adapter_priv; + + ret = rte_mempool_get_bulk(sw_data->msg_pool, (void **)msgs, nb_evtims); + if (ret < 0) { + rte_errno = ENOSPC; + return 0; + } + + /* Let the service know we're producing messages for it to process */ + rte_atomic16_inc(&sw_data->message_producer_count); + + /* If the service could be modifying event timer states, wait */ + while (sw_data->service_phase == 2) + rte_pause(); + + rte_smp_rmb(); + + for (i = 0; i < nb_evtims; i++) { + /* Don't modify the event timer state in these cases */ + if (evtims[i]->state == RTE_EVENT_TIMER_CANCELED) { + rte_errno = EALREADY; + break; + } else if (evtims[i]->state != RTE_EVENT_TIMER_ARMED) { + rte_errno = EINVAL; + break; + } + + msgs[i]->type = MSG_TYPE_CANCEL; + msgs[i]->evtim = evtims[i]; + + if (rte_ring_enqueue(sw_data->msg_ring, msgs[i]) < 0) { + rte_errno = ENOSPC; + break; + } + + EVTIM_LOG_DBG("enqueued CANCEL message to ring"); + + evtims[i]->state = RTE_EVENT_TIMER_CANCELED; + } + + /* Let the service know we're done producing messages */ + rte_atomic16_dec(&sw_data->message_producer_count); + + if (i < nb_evtims) + rte_mempool_put_bulk(sw_data->msg_pool, (void **)&msgs[i], + nb_evtims - i); + + return i; +} + +static uint16_t +sw_event_timer_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 __sw_event_timer_arm_burst(adapter, evtims, nb_evtims); +} + +static const struct rte_event_timer_adapter_ops sw_event_adapter_timer_ops = { + .init = sw_event_timer_adapter_init, + .uninit = sw_event_timer_adapter_uninit, + .start = sw_event_timer_adapter_start, + .stop = sw_event_timer_adapter_stop, + .get_info = sw_event_timer_adapter_get_info, + .stats_get = sw_event_timer_adapter_stats_get, + .stats_reset = sw_event_timer_adapter_stats_reset, + .arm_burst = sw_event_timer_arm_burst, + .arm_tmo_tick_burst = sw_event_timer_arm_tmo_tick_burst, + .cancel_burst = sw_event_timer_cancel_burst, +}; + +RTE_INIT(event_timer_adapter_init_log) +{ + evtim_logtype = rte_log_register("lib.eventdev.adapter.timer"); + if (evtim_logtype >= 0) + rte_log_set_level(evtim_logtype, RTE_LOG_NOTICE); + + evtim_buffer_logtype = rte_log_register("lib.eventdev.adapter.timer." + "buffer"); + if (evtim_buffer_logtype >= 0) + rte_log_set_level(evtim_buffer_logtype, RTE_LOG_NOTICE); + + evtim_svc_logtype = rte_log_register("lib.eventdev.adapter.timer.svc"); + if (evtim_svc_logtype >= 0) + rte_log_set_level(evtim_svc_logtype, RTE_LOG_NOTICE); +} diff --git a/lib/librte_eventdev/rte_event_timer_adapter.h b/lib/librte_eventdev/rte_event_timer_adapter.h new file mode 100644 index 00000000..d4ea6f17 --- /dev/null +++ b/lib/librte_eventdev/rte_event_timer_adapter.h @@ -0,0 +1,766 @@ +/* SPDX-License-Identifier: BSD-3-Clause + * Copyright(c) 2017 Cavium, Inc. + * Copyright(c) 2017-2018 Intel Corporation. + * All rights reserved. + */ + +#ifndef __RTE_EVENT_TIMER_ADAPTER_H__ +#define __RTE_EVENT_TIMER_ADAPTER_H__ + +/** + * @file + * + * RTE Event Timer Adapter + * + * An event timer adapter has the following abstract working model: + * + * timer_tick_ns + * + + * +-------+ | + * | | | + * +-------+ bkt 0 +----v---+ + * | | | | + * | +-------+ | + * +---+---+ +---+---+ +---+---+---+---+ + * | | | | | | | | | + * | bkt n | | bkt 1 |<-> t0| t1| t2| tn| + * | | | | | | | | | + * +---+---+ +---+---+ +---+---+---+---+ + * | Timer adapter | + * +---+---+ +---+---+ + * | | | | + * | bkt 4 | | bkt 2 |<--- Current bucket + * | | | | + * +---+---+ +---+---+ + * | +-------+ | + * | | | | + * +------+ bkt 3 +-------+ + * | | + * +-------+ + * + * - It has a virtual monotonically increasing 64-bit timer adapter clock based + * on *enum rte_event_timer_adapter_clk_src* clock source. The clock source + * could be a CPU clock, or a platform dependent external clock. + * + * - The application creates a timer adapter instance with given the clock + * source, the total number of event timers, and a resolution(expressed in ns) + * to traverse between the buckets. + * + * - Each timer adapter may have 0 to n buckets based on the configured + * max timeout(max_tmo_ns) and resolution(timer_tick_ns). Upon starting the + * timer adapter, the adapter starts ticking at *timer_tick_ns* resolution. + * + * - The application arms an event timer that will expire *timer_tick_ns* + * from now. + * + * - The application can cancel an armed timer and no timer expiry event will be + * generated. + * + * - If a timer expires then the library injects the timer expiry event in + * the designated event queue. + * + * - The timer expiry event will be received through *rte_event_dequeue_burst*. + * + * - The application frees the timer adapter instance. + * + * Multiple timer adapters can be created with a varying level of resolution + * for various expiry use cases that run in parallel. + * + * Before using the timer adapter, the application has to create and configure + * an event device along with the event port. Based on the event device + * capability it might require creating an additional event port to be used + * by the timer adapter. + * + * The application creates the event timer adapter using the + * ``rte_event_timer_adapter_create()``. The event device id is passed to this + * function, inside this function the event device capability is checked, + * and if an in-built port is absent the application uses the default + * function to create a new producer port. + * + * The application may also use the function + * ``rte_event_timer_adapter_create_ext()`` to have granular control over + * producer port creation in a case where the in-built port is absent. + * + * After creating the timer adapter, the application has to start it + * using ``rte_event_timer_adapter_start()``. The buckets are traversed from + * 0 to n; when the adapter ticks, the next bucket is visited. Each time, + * the list per bucket is processed, and timer expiry events are sent to the + * designated event queue. + * + * The application can arm one or more event timers using the + * ``rte_event_timer_arm_burst()``. The *timeout_ticks* represents the number + * of *timer_tick_ns* after which the timer has to expire. The timeout at + * which the timers expire can be grouped or be independent of each + * event timer instance. ``rte_event_timer_arm_tmo_tick_burst()`` addresses the + * former case and ``rte_event_timer_arm_burst()`` addresses the latter case. + * + * The application can cancel the timers from expiring using the + * ``rte_event_timer_cancel_burst()``. + * + * On the secondary process, ``rte_event_timer_adapter_lookup()`` can be used + * to get the timer adapter pointer from its id and use it to invoke fastpath + * operations such as arm and cancel. + * + * Some of the use cases of event timer adapter are Beacon Timers, + * Generic SW Timeout, Wireless MAC Scheduling, 3G Frame Protocols, + * Packet Scheduling, Protocol Retransmission Timers, Supervision Timers. + * All these use cases require high resolution and low time drift. + */ + +#ifdef __cplusplus +extern "C" { +#endif + +#include <rte_spinlock.h> +#include <rte_memory.h> + +#include "rte_eventdev.h" + +/** + * @warning + * @b EXPERIMENTAL: this enum may change without prior notice + * + * Timer adapter clock source + */ +enum rte_event_timer_adapter_clk_src { + RTE_EVENT_TIMER_ADAPTER_CPU_CLK, + /**< Use CPU clock as the clock source. */ + RTE_EVENT_TIMER_ADAPTER_EXT_CLK0, + /**< Platform dependent external clock source 0. */ + RTE_EVENT_TIMER_ADAPTER_EXT_CLK1, + /**< Platform dependent external clock source 1. */ + RTE_EVENT_TIMER_ADAPTER_EXT_CLK2, + /**< Platform dependent external clock source 2. */ + RTE_EVENT_TIMER_ADAPTER_EXT_CLK3, + /**< Platform dependent external clock source 3. */ +}; + +#define RTE_EVENT_TIMER_ADAPTER_F_ADJUST_RES (1ULL << 0) +/**< The event timer adapter implementation may have constraints on the + * resolution (timer_tick_ns) and maximum timer expiry timeout(max_tmo_ns) + * based on the given timer adapter or system. If this flag is set, the + * implementation adjusts the resolution and maximum timeout to the best + * possible configuration. On successful timer adapter creation, the + * application can get the configured resolution and max timeout with + * ``rte_event_timer_adapter_get_info()``. + * + * @see struct rte_event_timer_adapter_info::min_resolution_ns + * @see struct rte_event_timer_adapter_info::max_tmo_ns + */ +#define RTE_EVENT_TIMER_ADAPTER_F_SP_PUT (1ULL << 1) +/**< ``rte_event_timer_arm_burst()`` API to be used in single producer mode. + * + * @see struct rte_event_timer_adapter_conf::flags + */ + +/** + * @warning + * @b EXPERIMENTAL: this structure may change without prior notice + * + * Timer adapter configuration structure + */ +struct rte_event_timer_adapter_conf { + uint8_t event_dev_id; + /**< Event device identifier */ + uint16_t timer_adapter_id; + /**< Event timer adapter identifier */ + uint32_t socket_id; + /**< Identifier of socket from which to allocate memory for adapter */ + enum rte_event_timer_adapter_clk_src clk_src; + /**< Clock source for timer adapter */ + uint64_t timer_tick_ns; + /**< Timer adapter resolution in ns */ + uint64_t max_tmo_ns; + /**< Maximum timer timeout(expiry) in ns */ + uint64_t nb_timers; + /**< Total number of timers per adapter */ + uint64_t flags; + /**< Timer adapter config flags (RTE_EVENT_TIMER_ADAPTER_F_*) */ +}; + +/** + * @warning + * @b EXPERIMENTAL: this structure may change without prior notice + * + * Event timer adapter stats structure + */ +struct rte_event_timer_adapter_stats { + uint64_t evtim_exp_count; + /**< Number of event timers that have expired. */ + uint64_t ev_enq_count; + /**< Eventdev enqueue count */ + uint64_t ev_inv_count; + /**< Invalid expiry event count */ + uint64_t evtim_retry_count; + /**< Event timer retry count */ + uint64_t adapter_tick_count; + /**< Tick count for the adapter, at its resolution */ +}; + +struct rte_event_timer_adapter; + +/** + * @warning + * @b EXPERIMENTAL: this API may change without prior notice + * + * Callback function type for producer port creation. + */ +typedef int (*rte_event_timer_adapter_port_conf_cb_t)(uint16_t id, + uint8_t event_dev_id, + uint8_t *event_port_id, + void *conf_arg); + +/** + * @warning + * @b EXPERIMENTAL: this API may change without prior notice + * + * Create an event timer adapter. + * + * This function must be invoked first before any other function in the API. + * + * @param conf + * The event timer adapter configuration structure. + * + * @return + * A pointer to the new allocated event timer adapter on success. + * NULL on error with rte_errno set appropriately. + * Possible rte_errno values include: + * - ERANGE: timer_tick_ns is not in supported range. + * - ENOMEM: unable to allocate sufficient memory for adapter instances + * - EINVAL: invalid event device identifier specified in config + * - ENOSPC: maximum number of adapters already created + * - EIO: event device reconfiguration and restart error. The adapter + * reconfigures the event device with an additional port by default if it is + * required to use a service to manage timers. If the device had been started + * before this call, this error code indicates an error in restart following + * an error in reconfiguration, i.e., a combination of the two error codes. + */ +struct rte_event_timer_adapter * __rte_experimental +rte_event_timer_adapter_create(const struct rte_event_timer_adapter_conf *conf); + +/** + * @warning + * @b EXPERIMENTAL: this API may change without prior notice + * + * Create a timer adapter with the supplied callback. + * + * This function can be used to have a more granular control over the timer + * adapter creation. If a built-in port is absent, then the function uses the + * callback provided to create and get the port id to be used as a producer + * port. + * + * @param conf + * The timer adapter configuration structure + * @param conf_cb + * The port config callback function. + * @param conf_arg + * Opaque pointer to the argument for the callback function + * + * @return + * A pointer to the new allocated event timer adapter on success. + * NULL on error with rte_errno set appropriately. + * Possible rte_errno values include: + * - ERANGE: timer_tick_ns is not in supported range. + * - ENOMEM: unable to allocate sufficient memory for adapter instances + * - EINVAL: invalid event device identifier specified in config + * - ENOSPC: maximum number of adapters already created + */ +struct rte_event_timer_adapter * __rte_experimental +rte_event_timer_adapter_create_ext( + const struct rte_event_timer_adapter_conf *conf, + rte_event_timer_adapter_port_conf_cb_t conf_cb, + void *conf_arg); + +/** + * @warning + * @b EXPERIMENTAL: this structure may change without prior notice + * + * Timer adapter info structure. + */ +struct rte_event_timer_adapter_info { + uint64_t min_resolution_ns; + /**< Minimum timer adapter resolution in ns */ + uint64_t max_tmo_ns; + /**< Maximum timer timeout(expire) in ns */ + struct rte_event_timer_adapter_conf conf; + /**< Configured timer adapter attributes */ + uint32_t caps; + /**< Event timer adapter capabilities */ + int16_t event_dev_port_id; + /**< Event device port ID, if applicable */ +}; + +/** + * @warning + * @b EXPERIMENTAL: this API may change without prior notice + * + * Retrieve the contextual information of an event timer adapter. + * + * @param adapter + * A pointer to the event timer adapter structure. + * + * @param[out] adapter_info + * A pointer to a structure of type *rte_event_timer_adapter_info* to be + * filled with the contextual information of the adapter. + * + * @return + * - 0: Success, driver updates the contextual information of the + * timer adapter + * - <0: Error code returned by the driver info get function. + * - -EINVAL: adapter identifier invalid + * + * @see RTE_EVENT_TIMER_ADAPTER_F_ADJUST_RES, + * struct rte_event_timer_adapter_info + * + */ +int __rte_experimental +rte_event_timer_adapter_get_info( + const struct rte_event_timer_adapter *adapter, + struct rte_event_timer_adapter_info *adapter_info); + +/** + * @warning + * @b EXPERIMENTAL: this API may change without prior notice + * + * Start a timer adapter. + * + * The adapter start step is the last one and consists of setting the timer + * adapter to start accepting the timers and schedules to event queues. + * + * On success, all basic functions exported by the API (timer arm, + * timer cancel and so on) can be invoked. + * + * @param adapter + * A pointer to the event timer adapter structure. + * + * @return + * - 0: Success, adapter started. + * - <0: Error code returned by the driver start function. + * - -EINVAL if adapter identifier invalid + * - -ENOENT if software adapter but no service core mapped + * - -ENOTSUP if software adapter and more than one service core mapped + */ +int __rte_experimental +rte_event_timer_adapter_start( + const struct rte_event_timer_adapter *adapter); + +/** + * @warning + * @b EXPERIMENTAL: this API may change without prior notice + * + * Stop an event timer adapter. + * + * The adapter can be restarted with a call to + * ``rte_event_timer_adapter_start()``. + * + * @param adapter + * A pointer to the event timer adapter structure. + * + * @return + * - 0: Success, adapter stopped. + * - <0: Error code returned by the driver stop function. + * - -EINVAL if adapter identifier invalid + */ +int __rte_experimental +rte_event_timer_adapter_stop(const struct rte_event_timer_adapter *adapter); + +/** + * @warning + * @b EXPERIMENTAL: this API may change without prior notice + * + * Lookup an event timer adapter using its identifier. + * + * If an event timer adapter was created in another process with the same + * identifier, this function will locate its state and set up access to it + * so that it can be used in this process. + * + * @param adapter_id + * The event timer adapter identifier. + * + * @return + * A pointer to the event timer adapter matching the identifier on success. + * NULL on error with rte_errno set appropriately. + * Possible rte_errno values include: + * - ENOENT - requested entry not available to return. + */ +struct rte_event_timer_adapter * __rte_experimental +rte_event_timer_adapter_lookup(uint16_t adapter_id); + +/** + * @warning + * @b EXPERIMENTAL: this API may change without prior notice + * + * Free an event timer adapter. + * + * Destroy an event timer adapter, freeing all resources. + * + * Before invoking this function, the application must wait for all the + * armed timers to expire or cancel the outstanding armed timers. + * + * @param adapter + * A pointer to an event timer adapter structure. + * + * @return + * - 0: Successfully freed the event timer adapter resources. + * - <0: Failed to free the event timer adapter resources. + * - -EAGAIN: adapter is busy; timers outstanding + * - -EBUSY: stop hasn't been called for this adapter yet + * - -EINVAL: adapter id invalid, or adapter invalid + */ +int __rte_experimental +rte_event_timer_adapter_free(struct rte_event_timer_adapter *adapter); + +/** + * Retrieve the service ID of the event timer adapter. If the adapter doesn't + * use an rte_service function, this function returns -ESRCH. + * + * @param adapter + * A pointer to an event timer adapter. + * + * @param [out] service_id + * A pointer to a uint32_t, to be filled in with the service id. + * + * @return + * - 0: Success + * - <0: Error code on failure + * - -ESRCH: the adapter does not require a service to operate + */ +int __rte_experimental +rte_event_timer_adapter_service_id_get(struct rte_event_timer_adapter *adapter, + uint32_t *service_id); + +/** + * @warning + * @b EXPERIMENTAL: this API may change without prior notice + * + * Retrieve statistics for an event timer adapter instance. + * + * @param adapter + * A pointer to an event timer adapter structure. + * @param[out] stats + * A pointer to a structure to fill with statistics. + * + * @return + * - 0: Successfully retrieved. + * - <0: Failure; error code returned. + */ +int __rte_experimental +rte_event_timer_adapter_stats_get(struct rte_event_timer_adapter *adapter, + struct rte_event_timer_adapter_stats *stats); + +/** + * @warning + * @b EXPERIMENTAL: this API may change without prior notice + * + * Reset statistics for an event timer adapter instance. + * + * @param adapter + * A pointer to an event timer adapter structure. + * + * @return + * - 0: Successfully reset; + * - <0: Failure; error code returned. + */ +int __rte_experimental rte_event_timer_adapter_stats_reset( + struct rte_event_timer_adapter *adapter); + +/** + * Retrieve the service ID of the event timer adapter. If the adapter doesn't + * use an rte_service function, this function returns -ESRCH. + * + * @param adapter + * A pointer to an event timer adapter. + * + * @param [out] service_id + * A pointer to a uint32_t, to be filled in with the service id. + * + * @return + * - 0: Success + * - <0: Error code on failure, if the event dev doesn't use a rte_service + * function, this function returns -ESRCH. + */ +int +rte_event_timer_adapter_service_id_get(struct rte_event_timer_adapter *adapter, + uint32_t *service_id); + +/** + * @warning + * @b EXPERIMENTAL: this API may change without prior notice + * + * Retrieve statistics for an event timer adapter instance. + * + * @param adapter + * A pointer to an event timer adapter structure. + * @param[out] stats + * A pointer to a structure to fill with statistics. + * + * @return + * - 0: Successfully retrieved. + * - <0: Failure; error code returned. + */ +int rte_event_timer_adapter_stats_get(struct rte_event_timer_adapter *adapter, + struct rte_event_timer_adapter_stats *stats); + +/** + * @warning + * @b EXPERIMENTAL: this API may change without prior notice + * + * Reset statistics for an event timer adapter instance. + * + * @param adapter + * A pointer to an event timer adapter structure. + * + * @return + * - 0: Successfully reset; + * - <0: Failure; error code returned. + */ +int rte_event_timer_adapter_stats_reset( + struct rte_event_timer_adapter *adapter); + +/** + * @warning + * @b EXPERIMENTAL: this structure may change without prior notice + * + * Event timer state. + */ +enum rte_event_timer_state { + RTE_EVENT_TIMER_NOT_ARMED = 0, + /**< Event timer not armed. */ + RTE_EVENT_TIMER_ARMED = 1, + /**< Event timer successfully armed. */ + RTE_EVENT_TIMER_CANCELED = 2, + /**< Event timer successfully canceled. */ + RTE_EVENT_TIMER_ERROR = -1, + /**< Generic event timer error. */ + RTE_EVENT_TIMER_ERROR_TOOEARLY = -2, + /**< Event timer timeout tick value is too small for the adapter to + * handle, given its configured resolution. + */ + RTE_EVENT_TIMER_ERROR_TOOLATE = -3, + /**< Event timer timeout tick is greater than the maximum timeout.*/ +}; + +/** + * @warning + * @b EXPERIMENTAL: this structure may change without prior notice + * + * The generic *rte_event_timer* structure to hold the event timer attributes + * for arm and cancel operations. + */ +RTE_STD_C11 +struct rte_event_timer { + struct rte_event ev; + /**< + * Expiry event attributes. On successful event timer timeout, + * the following attributes will be used to inject the expiry event to + * the eventdev: + * - event_queue_id: Targeted event queue id for expiry events. + * - event_priority: Event priority of the event expiry event in the + * event queue relative to other events. + * - sched_type: Scheduling type of the expiry event. + * - flow_id: Flow id of the expiry event. + * - op: RTE_EVENT_OP_NEW + * - event_type: RTE_EVENT_TYPE_TIMER + */ + volatile enum rte_event_timer_state state; + /**< State of the event timer. */ + uint64_t timeout_ticks; + /**< Expiry timer ticks expressed in number of *timer_ticks_ns* from + * now. + * @see struct rte_event_timer_adapter_info::adapter_conf::timer_tick_ns + */ + uint64_t impl_opaque[2]; + /**< Implementation-specific opaque data. + * An event timer adapter implementation use this field to hold + * implementation specific values to share between the arm and cancel + * operations. The application should not modify this field. + */ + uint8_t user_meta[0]; + /**< Memory to store user specific metadata. + * The event timer adapter implementation should not modify this area. + */ +} __rte_cache_aligned; + +typedef uint16_t (*rte_event_timer_arm_burst_t)( + const struct rte_event_timer_adapter *adapter, + struct rte_event_timer **tims, + uint16_t nb_tims); +/**< @internal Enable event timers to enqueue timer events upon expiry */ +typedef uint16_t (*rte_event_timer_arm_tmo_tick_burst_t)( + const struct rte_event_timer_adapter *adapter, + struct rte_event_timer **tims, + uint64_t timeout_tick, + uint16_t nb_tims); +/**< @internal Enable event timers with common expiration time */ +typedef uint16_t (*rte_event_timer_cancel_burst_t)( + const struct rte_event_timer_adapter *adapter, + struct rte_event_timer **tims, + uint16_t nb_tims); +/**< @internal Prevent event timers from enqueuing timer events */ + +/** + * @internal Data structure associated with each event timer adapter. + */ +struct rte_event_timer_adapter { + rte_event_timer_arm_burst_t arm_burst; + /**< Pointer to driver arm_burst function. */ + rte_event_timer_arm_tmo_tick_burst_t arm_tmo_tick_burst; + /**< Pointer to driver arm_tmo_tick_burst function. */ + rte_event_timer_cancel_burst_t cancel_burst; + /**< Pointer to driver cancel function. */ + struct rte_event_timer_adapter_data *data; + /**< Pointer to shared adapter data */ + const struct rte_event_timer_adapter_ops *ops; + /**< Functions exported by adapter driver */ + + RTE_STD_C11 + uint8_t allocated : 1; + /**< Flag to indicate that this adapter has been allocated */ +} __rte_cache_aligned; + +#define ADAPTER_VALID_OR_ERR_RET(adapter, retval) do { \ + if (adapter == NULL || !adapter->allocated) \ + return retval; \ +} while (0) + +#define FUNC_PTR_OR_ERR_RET(func, errval) do { \ + if ((func) == NULL) \ + return errval; \ +} while (0) + +#define FUNC_PTR_OR_NULL_RET_WITH_ERRNO(func, errval) do { \ + if ((func) == NULL) { \ + rte_errno = errval; \ + return NULL; \ + } \ +} while (0) + +/** + * @warning + * @b EXPERIMENTAL: this API may change without prior notice + * + * Arm a burst of event timers with separate expiration timeout tick for each + * event timer. + * + * Before calling this function, the application allocates + * ``struct rte_event_timer`` objects from mempool or huge page backed + * application buffers of desired size. On successful allocation, + * application updates the `struct rte_event_timer`` attributes such as + * expiry event attributes, timeout ticks from now. + * This function submits the event timer arm requests to the event timer adapter + * and on expiry, the events will be injected to designated event queue. + * + * @param adapter + * A pointer to an event timer adapter structure. + * @param evtims + * Pointer to an array of objects of type *rte_event_timer* structure. + * @param nb_evtims + * Number of event timers in the supplied array. + * + * @return + * The number of successfully armed event timers. The return value can be less + * than the value of the *nb_evtims* parameter. If the return value is less + * than *nb_evtims*, the remaining event timers at the end of *evtims* + * are not consumed, and the caller has to take care of them, and rte_errno + * is set accordingly. Possible errno values include: + * - EINVAL Invalid timer adapter, expiry event queue ID is invalid, or an + * expiry event's sched type doesn't match the capabilities of the + * destination event queue. + * - EAGAIN Specified timer adapter is not running + * - EALREADY A timer was encountered that was already armed + */ +static inline uint16_t __rte_experimental +rte_event_timer_arm_burst(const struct rte_event_timer_adapter *adapter, + struct rte_event_timer **evtims, + uint16_t nb_evtims) +{ +#ifdef RTE_LIBRTE_EVENTDEV_DEBUG + ADAPTER_VALID_OR_ERR_RET(adapter, -EINVAL); + FUNC_PTR_OR_ERR_RET(adapter->arm_burst, -EINVAL); +#endif + return adapter->arm_burst(adapter, evtims, nb_evtims); +} + +/** + * @warning + * @b EXPERIMENTAL: this API may change without prior notice + * + * Arm a burst of event timers with same expiration timeout tick. + * + * Provides the same functionality as ``rte_event_timer_arm_burst()``, except + * that application can use this API when all the event timers have the + * same timeout expiration tick. This specialized function can provide the + * additional hint to the adapter implementation and optimize if possible. + * + * @param adapter + * A pointer to an event timer adapter structure. + * @param evtims + * Points to an array of objects of type *rte_event_timer* structure. + * @param timeout_ticks + * The number of ticks in which the timers should expire. + * @param nb_evtims + * Number of event timers in the supplied array. + * + * @return + * The number of successfully armed event timers. The return value can be less + * than the value of the *nb_evtims* parameter. If the return value is less + * than *nb_evtims*, the remaining event timers at the end of *evtims* + * are not consumed, and the caller has to take care of them, and rte_errno + * is set accordingly. Possible errno values include: + * - EINVAL Invalid timer adapter, expiry event queue ID is invalid, or an + * expiry event's sched type doesn't match the capabilities of the + * destination event queue. + * - EAGAIN Specified event timer adapter is not running + * - EALREADY A timer was encountered that was already armed + */ +static inline uint16_t __rte_experimental +rte_event_timer_arm_tmo_tick_burst( + const struct rte_event_timer_adapter *adapter, + struct rte_event_timer **evtims, + const uint64_t timeout_ticks, + const uint16_t nb_evtims) +{ +#ifdef RTE_LIBRTE_EVENTDEV_DEBUG + ADAPTER_VALID_OR_ERR_RET(adapter, -EINVAL); + FUNC_PTR_OR_ERR_RET(adapter->arm_tmo_tick_burst, -EINVAL); +#endif + return adapter->arm_tmo_tick_burst(adapter, evtims, timeout_ticks, + nb_evtims); +} + +/** + * @warning + * @b EXPERIMENTAL: this API may change without prior notice + * + * Cancel a burst of event timers from being scheduled to the event device. + * + * @param adapter + * A pointer to an event timer adapter structure. + * @param evtims + * Points to an array of objects of type *rte_event_timer* structure + * @param nb_evtims + * Number of event timer instances in the supplied array. + * + * @return + * The number of successfully canceled event timers. The return value can be + * less than the value of the *nb_evtims* parameter. If the return value is + * less than *nb_evtims*, the remaining event timers at the end of *evtims* + * are not consumed, and the caller has to take care of them, and rte_errno + * is set accordingly. Possible errno values include: + * - EINVAL Invalid timer adapter identifier + * - EAGAIN Specified timer adapter is not running + * - EALREADY A timer was encountered that was already canceled + */ +static inline uint16_t __rte_experimental +rte_event_timer_cancel_burst(const struct rte_event_timer_adapter *adapter, + struct rte_event_timer **evtims, + uint16_t nb_evtims) +{ +#ifdef RTE_LIBRTE_EVENTDEV_DEBUG + ADAPTER_VALID_OR_ERR_RET(adapter, -EINVAL); + FUNC_PTR_OR_ERR_RET(adapter->cancel_burst, -EINVAL); +#endif + return adapter->cancel_burst(adapter, evtims, nb_evtims); +} + +#endif /* __RTE_EVENT_TIMER_ADAPTER_H__ */ diff --git a/lib/librte_eventdev/rte_event_timer_adapter_pmd.h b/lib/librte_eventdev/rte_event_timer_adapter_pmd.h new file mode 100644 index 00000000..cf3509dc --- /dev/null +++ b/lib/librte_eventdev/rte_event_timer_adapter_pmd.h @@ -0,0 +1,114 @@ +/* SPDX-License-Identifier: BSD-3-Clause + * Copyright(c) 2017-2018 Intel Corporation. + * All rights reserved. + */ + +#ifndef __RTE_EVENT_TIMER_ADAPTER_PMD_H__ +#define __RTE_EVENT_TIMER_ADAPTER_PMD_H__ + +/** + * @file + * RTE Event Timer Adapter API (PMD Side) + * + * @note + * This file provides implementation helpers for internal use by PMDs. They + * are not intended to be exposed to applications and are not subject to ABI + * versioning. + * + */ + +#ifdef __cplusplus +extern "C" { +#endif + +#include "rte_event_timer_adapter.h" + +/* + * Definitions of functions exported by an event timer adapter implementation + * through *rte_event_timer_adapter_ops* structure supplied in the + * *rte_event_timer_adapter* structure associated with an event timer adapter. + */ + +typedef int (*rte_event_timer_adapter_init_t)( + struct rte_event_timer_adapter *adapter); +/**< @internal Event timer adapter implementation setup */ +typedef int (*rte_event_timer_adapter_uninit_t)( + struct rte_event_timer_adapter *adapter); +/**< @internal Event timer adapter implementation teardown */ +typedef int (*rte_event_timer_adapter_start_t)( + const struct rte_event_timer_adapter *adapter); +/**< @internal Start running event timer adapter */ +typedef int (*rte_event_timer_adapter_stop_t)( + const struct rte_event_timer_adapter *adapter); +/**< @internal Stop running event timer adapter */ +typedef void (*rte_event_timer_adapter_get_info_t)( + const struct rte_event_timer_adapter *adapter, + struct rte_event_timer_adapter_info *adapter_info); +/**< @internal Get contextual information for event timer adapter */ +typedef int (*rte_event_timer_adapter_stats_get_t)( + const struct rte_event_timer_adapter *adapter, + struct rte_event_timer_adapter_stats *stats); +/**< @internal Get statistics for event timer adapter */ +typedef int (*rte_event_timer_adapter_stats_reset_t)( + const struct rte_event_timer_adapter *adapter); +/**< @internal Reset statistics for event timer adapter */ + +/** + * @internal Structure containing the functions exported by an event timer + * adapter implementation. + */ +struct rte_event_timer_adapter_ops { + rte_event_timer_adapter_init_t init; /**< Set up adapter */ + rte_event_timer_adapter_uninit_t uninit;/**< Tear down adapter */ + rte_event_timer_adapter_start_t start; /**< Start adapter */ + rte_event_timer_adapter_stop_t stop; /**< Stop adapter */ + rte_event_timer_adapter_get_info_t get_info; + /**< Get info from driver */ + rte_event_timer_adapter_stats_get_t stats_get; + /**< Get adapter statistics */ + rte_event_timer_adapter_stats_reset_t stats_reset; + /**< Reset adapter statistics */ + rte_event_timer_arm_burst_t arm_burst; + /**< Arm one or more event timers */ + rte_event_timer_arm_tmo_tick_burst_t arm_tmo_tick_burst; + /**< Arm event timers with same expiration time */ + rte_event_timer_cancel_burst_t cancel_burst; + /**< Cancel one or more event timers */ +}; + +/** + * @internal Adapter data; structure to be placed in shared memory to be + * accessible by various processes in a multi-process configuration. + */ +struct rte_event_timer_adapter_data { + uint8_t id; + /**< Event timer adapter ID */ + uint8_t event_dev_id; + /**< Event device ID */ + uint32_t socket_id; + /**< Socket ID where memory is allocated */ + uint8_t event_port_id; + /**< Optional: event port ID used when the inbuilt port is absent */ + const struct rte_memzone *mz; + /**< Event timer adapter memzone pointer */ + struct rte_event_timer_adapter_conf conf; + /**< Configuration used to configure the adapter. */ + uint32_t caps; + /**< Adapter capabilities */ + void *adapter_priv; + /**< Timer adapter private data*/ + uint8_t service_inited; + /**< Service initialization state */ + uint32_t service_id; + /**< Service ID*/ + + RTE_STD_C11 + uint8_t started : 1; + /**< Flag to indicate adapter started. */ +} __rte_cache_aligned; + +#ifdef __cplusplus +} +#endif + +#endif /* __RTE_EVENT_TIMER_ADAPTER_PMD_H__ */ diff --git a/lib/librte_eventdev/rte_eventdev.c b/lib/librte_eventdev/rte_eventdev.c index 851a1190..801810ed 100644 --- a/lib/librte_eventdev/rte_eventdev.c +++ b/lib/librte_eventdev/rte_eventdev.c @@ -29,6 +29,8 @@ #include <rte_malloc.h> #include <rte_errno.h> #include <rte_ethdev.h> +#include <rte_cryptodev.h> +#include <rte_cryptodev_pmd.h> #include "rte_eventdev.h" #include "rte_eventdev_pmd.h" @@ -55,16 +57,21 @@ int rte_event_dev_get_dev_id(const char *name) { int i; + uint8_t cmp; if (!name) return -EINVAL; - for (i = 0; i < rte_eventdev_globals->nb_devs; i++) - if ((strcmp(rte_event_devices[i].data->name, name) - == 0) && - (rte_event_devices[i].attached == - RTE_EVENTDEV_ATTACHED)) + for (i = 0; i < rte_eventdev_globals->nb_devs; i++) { + cmp = (strncmp(rte_event_devices[i].data->name, name, + RTE_EVENTDEV_NAME_MAX_LEN) == 0) || + (rte_event_devices[i].dev ? (strncmp( + rte_event_devices[i].dev->driver->name, name, + RTE_EVENTDEV_NAME_MAX_LEN) == 0) : 0); + if (cmp && (rte_event_devices[i].attached == + RTE_EVENTDEV_ATTACHED)) return i; + } return -ENODEV; } @@ -123,6 +130,51 @@ rte_event_eth_rx_adapter_caps_get(uint8_t dev_id, uint8_t eth_port_id, : 0; } +int __rte_experimental +rte_event_timer_adapter_caps_get(uint8_t dev_id, uint32_t *caps) +{ + struct rte_eventdev *dev; + const struct rte_event_timer_adapter_ops *ops; + + RTE_EVENTDEV_VALID_DEVID_OR_ERR_RET(dev_id, -EINVAL); + + dev = &rte_eventdevs[dev_id]; + + if (caps == NULL) + return -EINVAL; + *caps = 0; + + return dev->dev_ops->timer_adapter_caps_get ? + (*dev->dev_ops->timer_adapter_caps_get)(dev, + 0, + caps, + &ops) + : 0; +} + +int __rte_experimental +rte_event_crypto_adapter_caps_get(uint8_t dev_id, uint8_t cdev_id, + uint32_t *caps) +{ + struct rte_eventdev *dev; + struct rte_cryptodev *cdev; + + RTE_EVENTDEV_VALID_DEVID_OR_ERR_RET(dev_id, -EINVAL); + if (!rte_cryptodev_pmd_is_valid_dev(cdev_id)) + return -EINVAL; + + dev = &rte_eventdevs[dev_id]; + cdev = rte_cryptodev_pmd_get_dev(cdev_id); + + if (caps == NULL) + return -EINVAL; + *caps = 0; + + return dev->dev_ops->crypto_adapter_caps_get ? + (*dev->dev_ops->crypto_adapter_caps_get) + (dev, cdev, caps) : -ENOTSUP; +} + static inline int rte_event_dev_queue_config(struct rte_eventdev *dev, uint8_t nb_queues) { @@ -1123,6 +1175,23 @@ rte_event_dev_start(uint8_t dev_id) return 0; } +int +rte_event_dev_stop_flush_callback_register(uint8_t dev_id, + eventdev_stop_flush_t callback, void *userdata) +{ + struct rte_eventdev *dev; + + RTE_EDEV_LOG_DEBUG("Stop flush register dev_id=%" PRIu8, dev_id); + + RTE_EVENTDEV_VALID_DEVID_OR_ERR_RET(dev_id, -EINVAL); + dev = &rte_eventdevs[dev_id]; + + dev->dev_ops->dev_stop_flush = callback; + dev->data->dev_stop_flush_arg = userdata; + + return 0; +} + void rte_event_dev_stop(uint8_t dev_id) { diff --git a/lib/librte_eventdev/rte_eventdev.h b/lib/librte_eventdev/rte_eventdev.h index b21c2717..b6fd6ee7 100644 --- a/lib/librte_eventdev/rte_eventdev.h +++ b/lib/librte_eventdev/rte_eventdev.h @@ -1,35 +1,8 @@ -/* - * BSD LICENSE - * - * Copyright 2016 Cavium, Inc. - * Copyright 2016 Intel Corporation. - * Copyright 2016 NXP. - * - * Redistribution and use in source and binary forms, with or without - * modification, are permitted provided that the following conditions - * are met: - * - * * Redistributions of source code must retain the above copyright - * notice, this list of conditions and the following disclaimer. - * * Redistributions in binary form must reproduce the above copyright - * notice, this list of conditions and the following disclaimer in - * the documentation and/or other materials provided with the - * distribution. - * * Neither the name of Cavium, Inc nor the names of its - * contributors may be used to endorse or promote products derived - * from this software without specific prior written permission. - * - * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS - * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT - * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR - * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT - * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, - * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT - * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, - * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY - * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT - * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE - * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +/* SPDX-License-Identifier: BSD-3-Clause + * Copyright(c) 2016 Cavium, Inc. + * Copyright(c) 2016-2018 Intel Corporation. + * Copyright 2016 NXP + * All rights reserved. */ #ifndef _RTE_EVENTDEV_H_ @@ -244,6 +217,7 @@ extern "C" { #include <rte_errno.h> struct rte_mbuf; /* we just use mbuf pointers; no need to include rte_mbuf.h */ +struct rte_event; /* Event device capability bitmap flags */ #define RTE_EVENT_DEV_CAP_QUEUE_QOS (1ULL << 0) @@ -835,15 +809,60 @@ int rte_event_dev_start(uint8_t dev_id); /** - * Stop an event device. The device can be restarted with a call to - * rte_event_dev_start() + * Stop an event device. + * + * This function causes all queued events to be drained, including those + * residing in event ports. While draining events out of the device, this + * function calls the user-provided flush callback (if one was registered) once + * per event. + * + * The device can be restarted with a call to rte_event_dev_start(). Threads + * that continue to enqueue/dequeue while the device is stopped, or being + * stopped, will result in undefined behavior. This includes event adapters, + * which must be stopped prior to stopping the eventdev. * * @param dev_id * Event device identifier. + * + * @see rte_event_dev_stop_flush_callback_register() */ void rte_event_dev_stop(uint8_t dev_id); +typedef void (*eventdev_stop_flush_t)(uint8_t dev_id, struct rte_event event, + void *arg); +/**< Callback function called during rte_event_dev_stop(), invoked once per + * flushed event. + */ + +/** + * Registers a callback function to be invoked during rte_event_dev_stop() for + * each flushed event. This function can be used to properly dispose of queued + * events, for example events containing memory pointers. + * + * The callback function is only registered for the calling process. The + * callback function must be registered in every process that can call + * rte_event_dev_stop(). + * + * To unregister a callback, call this function with a NULL callback pointer. + * + * @param dev_id + * The identifier of the device. + * @param callback + * Callback function invoked once per flushed event. + * @param userdata + * Argument supplied to callback. + * + * @return + * - 0 on success. + * - -EINVAL if *dev_id* is invalid + * + * @see rte_event_dev_stop() + */ +int +rte_event_dev_stop_flush_callback_register(uint8_t dev_id, + eventdev_stop_flush_t callback, void *userdata); + /** * Close an event device. The device cannot be restarted! * @@ -923,8 +942,8 @@ rte_event_dev_close(uint8_t dev_id); /**< The event generated from ethdev subsystem */ #define RTE_EVENT_TYPE_CRYPTODEV 0x1 /**< The event generated from crypodev subsystem */ -#define RTE_EVENT_TYPE_TIMERDEV 0x2 -/**< The event generated from timerdev subsystem */ +#define RTE_EVENT_TYPE_TIMER 0x2 +/**< The event generated from event timer adapter */ #define RTE_EVENT_TYPE_CPU 0x3 /**< The event generated from cpu for pipelining. * Application may use *sub_event_type* to further classify the event @@ -1096,7 +1115,77 @@ int rte_event_eth_rx_adapter_caps_get(uint8_t dev_id, uint8_t eth_port_id, uint32_t *caps); -struct rte_eventdev_driver; +#define RTE_EVENT_TIMER_ADAPTER_CAP_INTERNAL_PORT (1ULL << 0) +/**< This flag is set when the timer mechanism is in HW. */ + +/** + * Retrieve the event device's timer adapter capabilities. + * + * @param dev_id + * The identifier of the device. + * + * @param[out] caps + * A pointer to memory to be filled with event timer adapter capabilities. + * + * @return + * - 0: Success, driver provided event timer adapter capabilities. + * - <0: Error code returned by the driver function. + */ +int __rte_experimental +rte_event_timer_adapter_caps_get(uint8_t dev_id, uint32_t *caps); + +/* Crypto adapter capability bitmap flag */ +#define RTE_EVENT_CRYPTO_ADAPTER_CAP_INTERNAL_PORT_OP_NEW 0x1 +/**< Flag indicates HW is capable of generating events in + * RTE_EVENT_OP_NEW enqueue operation. Cryptodev will send + * packets to the event device as new events using an internal + * event port. + */ + +#define RTE_EVENT_CRYPTO_ADAPTER_CAP_INTERNAL_PORT_OP_FWD 0x2 +/**< Flag indicates HW is capable of generating events in + * RTE_EVENT_OP_FORWARD enqueue operation. Cryptodev will send + * packets to the event device as forwarded event using an + * internal event port. + */ + +#define RTE_EVENT_CRYPTO_ADAPTER_CAP_INTERNAL_PORT_QP_EV_BIND 0x4 +/**< Flag indicates HW is capable of mapping crypto queue pair to + * event queue. + */ + +#define RTE_EVENT_CRYPTO_ADAPTER_CAP_SESSION_PRIVATE_DATA 0x8 +/**< Flag indicates HW/SW suports a mechanism to store and retrieve + * the private data information along with the crypto session. + */ + +/** + * @warning + * @b EXPERIMENTAL: this API may change without prior notice + * + * Retrieve the event device's crypto adapter capabilities for the + * specified cryptodev device + * + * @param dev_id + * The identifier of the device. + * + * @param cdev_id + * The identifier of the cryptodev device. + * + * @param[out] caps + * A pointer to memory filled with event adapter capabilities. + * It is expected to be pre-allocated & initialized by caller. + * + * @return + * - 0: Success, driver provides event adapter capabilities for the + * cryptodev device. + * - <0: Error code returned by the driver function. + * + */ +int __rte_experimental +rte_event_crypto_adapter_caps_get(uint8_t dev_id, uint8_t cdev_id, + uint32_t *caps); + struct rte_eventdev_ops; struct rte_eventdev; @@ -1152,6 +1241,8 @@ struct rte_eventdev_data { /* Service initialization state */ uint32_t service_id; /* Service ID*/ + void *dev_stop_flush_arg; + /**< User-provided argument for event flush function */ RTE_STD_C11 uint8_t dev_started : 1; @@ -1178,7 +1269,7 @@ struct rte_eventdev { struct rte_eventdev_data *data; /**< Pointer to device data */ - const struct rte_eventdev_ops *dev_ops; + struct rte_eventdev_ops *dev_ops; /**< Functions exported by PMD */ struct rte_device *dev; /**< Device info. supplied by probing */ diff --git a/lib/librte_eventdev/rte_eventdev_pmd.h b/lib/librte_eventdev/rte_eventdev_pmd.h index 31343b51..3fbb4d2b 100644 --- a/lib/librte_eventdev/rte_eventdev_pmd.h +++ b/lib/librte_eventdev/rte_eventdev_pmd.h @@ -26,6 +26,7 @@ extern "C" { #include <rte_malloc.h> #include "rte_eventdev.h" +#include "rte_event_timer_adapter_pmd.h" /* Logging Macros */ #define RTE_EDEV_LOG_ERR(...) \ @@ -69,6 +70,9 @@ extern "C" { ((RTE_EVENT_ETH_RX_ADAPTER_CAP_OVERRIDE_FLOW_ID) | \ (RTE_EVENT_ETH_RX_ADAPTER_CAP_MULTI_EVENTQ)) +#define RTE_EVENT_CRYPTO_ADAPTER_SW_CAP \ + RTE_EVENT_CRYPTO_ADAPTER_CAP_SESSION_PRIVATE_DATA + /**< Ethernet Rx adapter cap to return If the packet transfers from * the ethdev to eventdev use a SW service function */ @@ -449,6 +453,37 @@ typedef int (*eventdev_eth_rx_adapter_caps_get_t) struct rte_event_eth_rx_adapter_queue_conf *queue_conf; /** + * Retrieve the event device's timer adapter capabilities, as well as the ops + * structure that an event timer adapter should call through to enter the + * driver + * + * @param dev + * Event device pointer + * + * @param flags + * Flags that can be used to determine how to select an event timer + * adapter ops structure + * + * @param[out] caps + * A pointer to memory filled with Rx event adapter capabilities. + * + * @param[out] ops + * A pointer to the ops pointer to set with the address of the desired ops + * structure + * + * @return + * - 0: Success, driver provides Rx event adapter capabilities for the + * ethernet device. + * - <0: Error code returned by the driver function. + * + */ +typedef int (*eventdev_timer_adapter_caps_get_t)( + const struct rte_eventdev *dev, + uint64_t flags, + uint32_t *caps, + const struct rte_event_timer_adapter_ops **ops); + +/** * Add ethernet Rx queues to event device. This callback is invoked if * the caps returned from rte_eventdev_eth_rx_adapter_caps_get(, eth_port_id) * has RTE_EVENT_ETH_RX_ADAPTER_CAP_INTERNAL_PORT set. @@ -585,6 +620,175 @@ typedef int (*eventdev_eth_rx_adapter_stats_reset) */ typedef int (*eventdev_selftest)(void); + +struct rte_cryptodev; + +/** + * This API may change without prior notice + * + * Retrieve the event device's crypto adapter capabilities for the + * specified cryptodev + * + * @param dev + * Event device pointer + * + * @param cdev + * cryptodev pointer + * + * @param[out] caps + * A pointer to memory filled with event adapter capabilities. + * It is expected to be pre-allocated & initialized by caller. + * + * @return + * - 0: Success, driver provides event adapter capabilities for the + * cryptodev. + * - <0: Error code returned by the driver function. + * + */ +typedef int (*eventdev_crypto_adapter_caps_get_t) + (const struct rte_eventdev *dev, + const struct rte_cryptodev *cdev, + uint32_t *caps); + +/** + * This API may change without prior notice + * + * Add crypto queue pair to event device. This callback is invoked if + * the caps returned from rte_event_crypto_adapter_caps_get(, cdev_id) + * has RTE_EVENT_CRYPTO_ADAPTER_CAP_INTERNAL_PORT_* set. + * + * @param dev + * Event device pointer + * + * @param cdev + * cryptodev pointer + * + * @param queue_pair_id + * cryptodev queue pair identifier. + * + * @param event + * Event information required for binding cryptodev queue pair to event queue. + * This structure will have a valid value for only those HW PMDs supporting + * @see RTE_EVENT_CRYPTO_ADAPTER_CAP_INTERNAL_PORT_QP_EV_BIND capability. + * + * @return + * - 0: Success, cryptodev queue pair added successfully. + * - <0: Error code returned by the driver function. + * + */ +typedef int (*eventdev_crypto_adapter_queue_pair_add_t) + (const struct rte_eventdev *dev, + const struct rte_cryptodev *cdev, + int32_t queue_pair_id, + const struct rte_event *event); + + +/** + * This API may change without prior notice + * + * Delete crypto queue pair to event device. This callback is invoked if + * the caps returned from rte_event_crypto_adapter_caps_get(, cdev_id) + * has RTE_EVENT_CRYPTO_ADAPTER_CAP_INTERNAL_PORT_* set. + * + * @param dev + * Event device pointer + * + * @param cdev + * cryptodev pointer + * + * @param queue_pair_id + * cryptodev queue pair identifier. + * + * @return + * - 0: Success, cryptodev queue pair deleted successfully. + * - <0: Error code returned by the driver function. + * + */ +typedef int (*eventdev_crypto_adapter_queue_pair_del_t) + (const struct rte_eventdev *dev, + const struct rte_cryptodev *cdev, + int32_t queue_pair_id); + +/** + * Start crypto adapter. This callback is invoked if + * the caps returned from rte_event_crypto_adapter_caps_get(.., cdev_id) + * has RTE_EVENT_CRYPTO_ADAPTER_CAP_INTERNAL_PORT_* set and queue pairs + * from cdev_id have been added to the event device. + * + * @param dev + * Event device pointer + * + * @param cdev + * Crypto device pointer + * + * @return + * - 0: Success, crypto adapter started successfully. + * - <0: Error code returned by the driver function. + */ +typedef int (*eventdev_crypto_adapter_start_t) + (const struct rte_eventdev *dev, + const struct rte_cryptodev *cdev); + +/** + * Stop crypto adapter. This callback is invoked if + * the caps returned from rte_event_crypto_adapter_caps_get(.., cdev_id) + * has RTE_EVENT_CRYPTO_ADAPTER_CAP_INTERNAL_PORT_* set and queue pairs + * from cdev_id have been added to the event device. + * + * @param dev + * Event device pointer + * + * @param cdev + * Crypto device pointer + * + * @return + * - 0: Success, crypto adapter stopped successfully. + * - <0: Error code returned by the driver function. + */ +typedef int (*eventdev_crypto_adapter_stop_t) + (const struct rte_eventdev *dev, + const struct rte_cryptodev *cdev); + +struct rte_event_crypto_adapter_stats; + +/** + * Retrieve crypto adapter statistics. + * + * @param dev + * Event device pointer + * + * @param cdev + * Crypto device pointer + * + * @param[out] stats + * Pointer to stats structure + * + * @return + * Return 0 on success. + */ + +typedef int (*eventdev_crypto_adapter_stats_get) + (const struct rte_eventdev *dev, + const struct rte_cryptodev *cdev, + struct rte_event_crypto_adapter_stats *stats); + +/** + * Reset crypto adapter statistics. + * + * @param dev + * Event device pointer + * + * @param cdev + * Crypto device pointer + * + * @return + * Return 0 on success. + */ + +typedef int (*eventdev_crypto_adapter_stats_reset) + (const struct rte_eventdev *dev, + const struct rte_cryptodev *cdev); + /** Event device operations function pointer table */ struct rte_eventdev_ops { eventdev_info_get_t dev_infos_get; /**< Get device info. */ @@ -640,8 +844,29 @@ struct rte_eventdev_ops { eventdev_eth_rx_adapter_stats_reset eth_rx_adapter_stats_reset; /**< Reset ethernet Rx stats */ + eventdev_timer_adapter_caps_get_t timer_adapter_caps_get; + /**< Get timer adapter capabilities */ + + eventdev_crypto_adapter_caps_get_t crypto_adapter_caps_get; + /**< Get crypto adapter capabilities */ + eventdev_crypto_adapter_queue_pair_add_t crypto_adapter_queue_pair_add; + /**< Add queue pair to crypto adapter */ + eventdev_crypto_adapter_queue_pair_del_t crypto_adapter_queue_pair_del; + /**< Delete queue pair from crypto adapter */ + eventdev_crypto_adapter_start_t crypto_adapter_start; + /**< Start crypto adapter */ + eventdev_crypto_adapter_stop_t crypto_adapter_stop; + /**< Stop crypto adapter */ + eventdev_crypto_adapter_stats_get crypto_adapter_stats_get; + /**< Get crypto stats */ + eventdev_crypto_adapter_stats_reset crypto_adapter_stats_reset; + /**< Reset crypto stats */ + eventdev_selftest dev_selftest; /**< Start eventdev Selftest */ + + eventdev_stop_flush_t dev_stop_flush; + /**< User-provided event flush function */ }; /** diff --git a/lib/librte_eventdev/rte_eventdev_version.map b/lib/librte_eventdev/rte_eventdev_version.map index 2aef470b..12835e9f 100644 --- a/lib/librte_eventdev/rte_eventdev_version.map +++ b/lib/librte_eventdev/rte_eventdev_version.map @@ -66,7 +66,6 @@ DPDK_17.11 { rte_event_eth_rx_adapter_stats_get; rte_event_eth_rx_adapter_stats_reset; rte_event_eth_rx_adapter_stop; - } DPDK_17.08; DPDK_18.02 { @@ -74,3 +73,41 @@ DPDK_18.02 { rte_event_dev_selftest; } DPDK_17.11; + +DPDK_18.05 { + global: + + rte_event_dev_stop_flush_callback_register; +} DPDK_18.02; + +EXPERIMENTAL { + global: + + rte_event_crypto_adapter_caps_get; + rte_event_crypto_adapter_create; + rte_event_crypto_adapter_create_ext; + rte_event_crypto_adapter_event_port_get; + rte_event_crypto_adapter_free; + rte_event_crypto_adapter_queue_pair_add; + rte_event_crypto_adapter_queue_pair_del; + rte_event_crypto_adapter_service_id_get; + rte_event_crypto_adapter_start; + rte_event_crypto_adapter_stats_get; + rte_event_crypto_adapter_stats_reset; + rte_event_crypto_adapter_stop; + rte_event_eth_rx_adapter_cb_register; + rte_event_timer_adapter_caps_get; + rte_event_timer_adapter_create; + rte_event_timer_adapter_create_ext; + rte_event_timer_adapter_free; + rte_event_timer_adapter_get_info; + rte_event_timer_adapter_lookup; + rte_event_timer_adapter_service_id_get; + rte_event_timer_adapter_start; + rte_event_timer_adapter_stats_get; + rte_event_timer_adapter_stats_reset; + rte_event_timer_adapter_stop; + rte_event_timer_arm_burst; + rte_event_timer_arm_tmo_tick_burst; + rte_event_timer_cancel_burst; +}; |