diff options
Diffstat (limited to 'examples/ip_pipeline/thread.c')
-rw-r--r-- | examples/ip_pipeline/thread.c | 3053 |
1 files changed, 2865 insertions, 188 deletions
diff --git a/examples/ip_pipeline/thread.c b/examples/ip_pipeline/thread.c index 9013afd1..7fc03332 100644 --- a/examples/ip_pipeline/thread.c +++ b/examples/ip_pipeline/thread.c @@ -1,291 +1,2968 @@ /* SPDX-License-Identifier: BSD-3-Clause - * Copyright(c) 2010-2015 Intel Corporation + * Copyright(c) 2010-2018 Intel Corporation */ +#include <stdlib.h> + #include <rte_common.h> #include <rte_cycles.h> -#include <rte_pipeline.h> +#include <rte_lcore.h> +#include <rte_ring.h> + +#include <rte_table_acl.h> +#include <rte_table_array.h> +#include <rte_table_hash.h> +#include <rte_table_lpm.h> +#include <rte_table_lpm_ipv6.h> -#include "pipeline_common_be.h" -#include "app.h" +#include "common.h" #include "thread.h" +#include "pipeline.h" -#if APP_THREAD_HEADROOM_STATS_COLLECT - -#define PIPELINE_RUN_REGULAR(thread, pipeline) \ -do { \ - uint64_t t0 = rte_rdtsc_precise(); \ - int n_pkts = rte_pipeline_run(pipeline->p); \ - \ - if (n_pkts == 0) { \ - uint64_t t1 = rte_rdtsc_precise(); \ - \ - thread->headroom_cycles += t1 - t0; \ - } \ -} while (0) - - -#define PIPELINE_RUN_CUSTOM(thread, data) \ -do { \ - uint64_t t0 = rte_rdtsc_precise(); \ - int n_pkts = data->f_run(data->be); \ - \ - if (n_pkts == 0) { \ - uint64_t t1 = rte_rdtsc_precise(); \ - \ - thread->headroom_cycles += t1 - t0; \ - } \ -} while (0) - -#else - -#define PIPELINE_RUN_REGULAR(thread, pipeline) \ - rte_pipeline_run(pipeline->p) - -#define PIPELINE_RUN_CUSTOM(thread, data) \ - data->f_run(data->be) +#ifndef THREAD_PIPELINES_MAX +#define THREAD_PIPELINES_MAX 256 +#endif +#ifndef THREAD_MSGQ_SIZE +#define THREAD_MSGQ_SIZE 64 #endif -static inline void * -thread_msg_recv(struct rte_ring *r) +#ifndef THREAD_TIMER_PERIOD_MS +#define THREAD_TIMER_PERIOD_MS 100 +#endif + +/** + * Master thead: data plane thread context + */ +struct thread { + struct rte_ring *msgq_req; + struct rte_ring *msgq_rsp; + + uint32_t enabled; +}; + +static struct thread thread[RTE_MAX_LCORE]; + +/** + * Data plane threads: context + */ +struct table_data { + struct rte_table_action *a; +}; + +struct pipeline_data { + struct rte_pipeline *p; + struct table_data table_data[RTE_PIPELINE_TABLE_MAX]; + uint32_t n_tables; + + struct rte_ring *msgq_req; + struct rte_ring *msgq_rsp; + uint64_t timer_period; /* Measured in CPU cycles. */ + uint64_t time_next; + + uint8_t buffer[TABLE_RULE_ACTION_SIZE_MAX]; +}; + +struct thread_data { + struct rte_pipeline *p[THREAD_PIPELINES_MAX]; + uint32_t n_pipelines; + + struct pipeline_data pipeline_data[THREAD_PIPELINES_MAX]; + struct rte_ring *msgq_req; + struct rte_ring *msgq_rsp; + uint64_t timer_period; /* Measured in CPU cycles. */ + uint64_t time_next; + uint64_t time_next_min; +} __rte_cache_aligned; + +static struct thread_data thread_data[RTE_MAX_LCORE]; + +/** + * Master thread: data plane thread init + */ +static void +thread_free(void) +{ + uint32_t i; + + for (i = 0; i < RTE_MAX_LCORE; i++) { + struct thread *t = &thread[i]; + + if (!rte_lcore_is_enabled(i)) + continue; + + /* MSGQs */ + if (t->msgq_req) + rte_ring_free(t->msgq_req); + + if (t->msgq_rsp) + rte_ring_free(t->msgq_rsp); + } +} + +int +thread_init(void) +{ + uint32_t i; + + RTE_LCORE_FOREACH_SLAVE(i) { + char name[NAME_MAX]; + struct rte_ring *msgq_req, *msgq_rsp; + struct thread *t = &thread[i]; + struct thread_data *t_data = &thread_data[i]; + uint32_t cpu_id = rte_lcore_to_socket_id(i); + + /* MSGQs */ + snprintf(name, sizeof(name), "THREAD-%04x-MSGQ-REQ", i); + + msgq_req = rte_ring_create(name, + THREAD_MSGQ_SIZE, + cpu_id, + RING_F_SP_ENQ | RING_F_SC_DEQ); + + if (msgq_req == NULL) { + thread_free(); + return -1; + } + + snprintf(name, sizeof(name), "THREAD-%04x-MSGQ-RSP", i); + + msgq_rsp = rte_ring_create(name, + THREAD_MSGQ_SIZE, + cpu_id, + RING_F_SP_ENQ | RING_F_SC_DEQ); + + if (msgq_rsp == NULL) { + thread_free(); + return -1; + } + + /* Master thread records */ + t->msgq_req = msgq_req; + t->msgq_rsp = msgq_rsp; + t->enabled = 1; + + /* Data plane thread records */ + t_data->n_pipelines = 0; + t_data->msgq_req = msgq_req; + t_data->msgq_rsp = msgq_rsp; + t_data->timer_period = + (rte_get_tsc_hz() * THREAD_TIMER_PERIOD_MS) / 1000; + t_data->time_next = rte_get_tsc_cycles() + t_data->timer_period; + t_data->time_next_min = t_data->time_next; + } + + return 0; +} + +static inline int +thread_is_running(uint32_t thread_id) +{ + enum rte_lcore_state_t thread_state; + + thread_state = rte_eal_get_lcore_state(thread_id); + return (thread_state == RUNNING) ? 1 : 0; +} + +/** + * Pipeline is running when: + * (A) Pipeline is mapped to a data plane thread AND + * (B) Its data plane thread is in RUNNING state. + */ +static inline int +pipeline_is_running(struct pipeline *p) +{ + if (p->enabled == 0) + return 0; + + return thread_is_running(p->thread_id); +} + +/** + * Master thread & data plane threads: message passing + */ +enum thread_req_type { + THREAD_REQ_PIPELINE_ENABLE = 0, + THREAD_REQ_PIPELINE_DISABLE, + THREAD_REQ_MAX +}; + +struct thread_msg_req { + enum thread_req_type type; + + union { + struct { + struct rte_pipeline *p; + struct { + struct rte_table_action *a; + } table[RTE_PIPELINE_TABLE_MAX]; + struct rte_ring *msgq_req; + struct rte_ring *msgq_rsp; + uint32_t timer_period_ms; + uint32_t n_tables; + } pipeline_enable; + + struct { + struct rte_pipeline *p; + } pipeline_disable; + }; +}; + +struct thread_msg_rsp { + int status; +}; + +/** + * Master thread + */ +static struct thread_msg_req * +thread_msg_alloc(void) +{ + size_t size = RTE_MAX(sizeof(struct thread_msg_req), + sizeof(struct thread_msg_rsp)); + + return calloc(1, size); +} + +static void +thread_msg_free(struct thread_msg_rsp *rsp) +{ + free(rsp); +} + +static struct thread_msg_rsp * +thread_msg_send_recv(uint32_t thread_id, + struct thread_msg_req *req) +{ + struct thread *t = &thread[thread_id]; + struct rte_ring *msgq_req = t->msgq_req; + struct rte_ring *msgq_rsp = t->msgq_rsp; + struct thread_msg_rsp *rsp; + int status; + + /* send */ + do { + status = rte_ring_sp_enqueue(msgq_req, req); + } while (status == -ENOBUFS); + + /* recv */ + do { + status = rte_ring_sc_dequeue(msgq_rsp, (void **) &rsp); + } while (status != 0); + + return rsp; +} + +int +thread_pipeline_enable(uint32_t thread_id, + const char *pipeline_name) +{ + struct pipeline *p = pipeline_find(pipeline_name); + struct thread *t; + struct thread_msg_req *req; + struct thread_msg_rsp *rsp; + uint32_t i; + int status; + + /* Check input params */ + if ((thread_id >= RTE_MAX_LCORE) || + (p == NULL) || + (p->n_ports_in == 0) || + (p->n_ports_out == 0) || + (p->n_tables == 0)) + return -1; + + t = &thread[thread_id]; + if ((t->enabled == 0) || + p->enabled) + return -1; + + if (!thread_is_running(thread_id)) { + struct thread_data *td = &thread_data[thread_id]; + struct pipeline_data *tdp = &td->pipeline_data[td->n_pipelines]; + + if (td->n_pipelines >= THREAD_PIPELINES_MAX) + return -1; + + /* Data plane thread */ + td->p[td->n_pipelines] = p->p; + + tdp->p = p->p; + for (i = 0; i < p->n_tables; i++) + tdp->table_data[i].a = p->table[i].a; + + tdp->n_tables = p->n_tables; + + tdp->msgq_req = p->msgq_req; + tdp->msgq_rsp = p->msgq_rsp; + tdp->timer_period = (rte_get_tsc_hz() * p->timer_period_ms) / 1000; + tdp->time_next = rte_get_tsc_cycles() + tdp->timer_period; + + td->n_pipelines++; + + /* Pipeline */ + p->thread_id = thread_id; + p->enabled = 1; + + return 0; + } + + /* Allocate request */ + req = thread_msg_alloc(); + if (req == NULL) + return -1; + + /* Write request */ + req->type = THREAD_REQ_PIPELINE_ENABLE; + req->pipeline_enable.p = p->p; + for (i = 0; i < p->n_tables; i++) + req->pipeline_enable.table[i].a = + p->table[i].a; + req->pipeline_enable.msgq_req = p->msgq_req; + req->pipeline_enable.msgq_rsp = p->msgq_rsp; + req->pipeline_enable.timer_period_ms = p->timer_period_ms; + req->pipeline_enable.n_tables = p->n_tables; + + /* Send request and wait for response */ + rsp = thread_msg_send_recv(thread_id, req); + if (rsp == NULL) + return -1; + + /* Read response */ + status = rsp->status; + + /* Free response */ + thread_msg_free(rsp); + + /* Request completion */ + if (status) + return status; + + p->thread_id = thread_id; + p->enabled = 1; + + return 0; +} + +int +thread_pipeline_disable(uint32_t thread_id, + const char *pipeline_name) +{ + struct pipeline *p = pipeline_find(pipeline_name); + struct thread *t; + struct thread_msg_req *req; + struct thread_msg_rsp *rsp; + int status; + + /* Check input params */ + if ((thread_id >= RTE_MAX_LCORE) || + (p == NULL)) + return -1; + + t = &thread[thread_id]; + if (t->enabled == 0) + return -1; + + if (p->enabled == 0) + return 0; + + if (p->thread_id != thread_id) + return -1; + + if (!thread_is_running(thread_id)) { + struct thread_data *td = &thread_data[thread_id]; + uint32_t i; + + for (i = 0; i < td->n_pipelines; i++) { + struct pipeline_data *tdp = &td->pipeline_data[i]; + + if (tdp->p != p->p) + continue; + + /* Data plane thread */ + if (i < td->n_pipelines - 1) { + struct rte_pipeline *pipeline_last = + td->p[td->n_pipelines - 1]; + struct pipeline_data *tdp_last = + &td->pipeline_data[td->n_pipelines - 1]; + + td->p[i] = pipeline_last; + memcpy(tdp, tdp_last, sizeof(*tdp)); + } + + td->n_pipelines--; + + /* Pipeline */ + p->enabled = 0; + + break; + } + + return 0; + } + + /* Allocate request */ + req = thread_msg_alloc(); + if (req == NULL) + return -1; + + /* Write request */ + req->type = THREAD_REQ_PIPELINE_DISABLE; + req->pipeline_disable.p = p->p; + + /* Send request and wait for response */ + rsp = thread_msg_send_recv(thread_id, req); + if (rsp == NULL) + return -1; + + /* Read response */ + status = rsp->status; + + /* Free response */ + thread_msg_free(rsp); + + /* Request completion */ + if (status) + return status; + + p->enabled = 0; + + return 0; +} + +/** + * Data plane threads: message handling + */ +static inline struct thread_msg_req * +thread_msg_recv(struct rte_ring *msgq_req) { - void *msg; - int status = rte_ring_sc_dequeue(r, &msg); + struct thread_msg_req *req; + + int status = rte_ring_sc_dequeue(msgq_req, (void **) &req); if (status != 0) return NULL; - return msg; + return req; } static inline void -thread_msg_send(struct rte_ring *r, - void *msg) +thread_msg_send(struct rte_ring *msgq_rsp, + struct thread_msg_rsp *rsp) { int status; do { - status = rte_ring_sp_enqueue(r, msg); + status = rte_ring_sp_enqueue(msgq_rsp, rsp); } while (status == -ENOBUFS); } +static struct thread_msg_rsp * +thread_msg_handle_pipeline_enable(struct thread_data *t, + struct thread_msg_req *req) +{ + struct thread_msg_rsp *rsp = (struct thread_msg_rsp *) req; + struct pipeline_data *p = &t->pipeline_data[t->n_pipelines]; + uint32_t i; + + /* Request */ + if (t->n_pipelines >= THREAD_PIPELINES_MAX) { + rsp->status = -1; + return rsp; + } + + t->p[t->n_pipelines] = req->pipeline_enable.p; + + p->p = req->pipeline_enable.p; + for (i = 0; i < req->pipeline_enable.n_tables; i++) + p->table_data[i].a = + req->pipeline_enable.table[i].a; + + p->n_tables = req->pipeline_enable.n_tables; + + p->msgq_req = req->pipeline_enable.msgq_req; + p->msgq_rsp = req->pipeline_enable.msgq_rsp; + p->timer_period = + (rte_get_tsc_hz() * req->pipeline_enable.timer_period_ms) / 1000; + p->time_next = rte_get_tsc_cycles() + p->timer_period; + + t->n_pipelines++; + + /* Response */ + rsp->status = 0; + return rsp; +} + +static struct thread_msg_rsp * +thread_msg_handle_pipeline_disable(struct thread_data *t, + struct thread_msg_req *req) +{ + struct thread_msg_rsp *rsp = (struct thread_msg_rsp *) req; + uint32_t n_pipelines = t->n_pipelines; + struct rte_pipeline *pipeline = req->pipeline_disable.p; + uint32_t i; + + /* find pipeline */ + for (i = 0; i < n_pipelines; i++) { + struct pipeline_data *p = &t->pipeline_data[i]; + + if (p->p != pipeline) + continue; + + if (i < n_pipelines - 1) { + struct rte_pipeline *pipeline_last = + t->p[n_pipelines - 1]; + struct pipeline_data *p_last = + &t->pipeline_data[n_pipelines - 1]; + + t->p[i] = pipeline_last; + memcpy(p, p_last, sizeof(*p)); + } + + t->n_pipelines--; + + rsp->status = 0; + return rsp; + } + + /* should not get here */ + rsp->status = 0; + return rsp; +} + +static void +thread_msg_handle(struct thread_data *t) +{ + for ( ; ; ) { + struct thread_msg_req *req; + struct thread_msg_rsp *rsp; + + req = thread_msg_recv(t->msgq_req); + if (req == NULL) + break; + + switch (req->type) { + case THREAD_REQ_PIPELINE_ENABLE: + rsp = thread_msg_handle_pipeline_enable(t, req); + break; + + case THREAD_REQ_PIPELINE_DISABLE: + rsp = thread_msg_handle_pipeline_disable(t, req); + break; + + default: + rsp = (struct thread_msg_rsp *) req; + rsp->status = -1; + } + + thread_msg_send(t->msgq_rsp, rsp); + } +} + +/** + * Master thread & data plane threads: message passing + */ +enum pipeline_req_type { + /* Port IN */ + PIPELINE_REQ_PORT_IN_STATS_READ, + PIPELINE_REQ_PORT_IN_ENABLE, + PIPELINE_REQ_PORT_IN_DISABLE, + + /* Port OUT */ + PIPELINE_REQ_PORT_OUT_STATS_READ, + + /* Table */ + PIPELINE_REQ_TABLE_STATS_READ, + PIPELINE_REQ_TABLE_RULE_ADD, + PIPELINE_REQ_TABLE_RULE_ADD_DEFAULT, + PIPELINE_REQ_TABLE_RULE_ADD_BULK, + PIPELINE_REQ_TABLE_RULE_DELETE, + PIPELINE_REQ_TABLE_RULE_DELETE_DEFAULT, + PIPELINE_REQ_TABLE_RULE_STATS_READ, + PIPELINE_REQ_TABLE_MTR_PROFILE_ADD, + PIPELINE_REQ_TABLE_MTR_PROFILE_DELETE, + PIPELINE_REQ_TABLE_RULE_MTR_READ, + PIPELINE_REQ_TABLE_DSCP_TABLE_UPDATE, + PIPELINE_REQ_TABLE_RULE_TTL_READ, + PIPELINE_REQ_MAX +}; + +struct pipeline_msg_req_port_in_stats_read { + int clear; +}; + +struct pipeline_msg_req_port_out_stats_read { + int clear; +}; + +struct pipeline_msg_req_table_stats_read { + int clear; +}; + +struct pipeline_msg_req_table_rule_add { + struct table_rule_match match; + struct table_rule_action action; +}; + +struct pipeline_msg_req_table_rule_add_default { + struct table_rule_action action; +}; + +struct pipeline_msg_req_table_rule_add_bulk { + struct table_rule_match *match; + struct table_rule_action *action; + void **data; + uint32_t n_rules; + int bulk; +}; + +struct pipeline_msg_req_table_rule_delete { + struct table_rule_match match; +}; + +struct pipeline_msg_req_table_rule_stats_read { + void *data; + int clear; +}; + +struct pipeline_msg_req_table_mtr_profile_add { + uint32_t meter_profile_id; + struct rte_table_action_meter_profile profile; +}; + +struct pipeline_msg_req_table_mtr_profile_delete { + uint32_t meter_profile_id; +}; + +struct pipeline_msg_req_table_rule_mtr_read { + void *data; + uint32_t tc_mask; + int clear; +}; + +struct pipeline_msg_req_table_dscp_table_update { + uint64_t dscp_mask; + struct rte_table_action_dscp_table dscp_table; +}; + +struct pipeline_msg_req_table_rule_ttl_read { + void *data; + int clear; +}; + +struct pipeline_msg_req { + enum pipeline_req_type type; + uint32_t id; /* Port IN, port OUT or table ID */ + + RTE_STD_C11 + union { + struct pipeline_msg_req_port_in_stats_read port_in_stats_read; + struct pipeline_msg_req_port_out_stats_read port_out_stats_read; + struct pipeline_msg_req_table_stats_read table_stats_read; + struct pipeline_msg_req_table_rule_add table_rule_add; + struct pipeline_msg_req_table_rule_add_default table_rule_add_default; + struct pipeline_msg_req_table_rule_add_bulk table_rule_add_bulk; + struct pipeline_msg_req_table_rule_delete table_rule_delete; + struct pipeline_msg_req_table_rule_stats_read table_rule_stats_read; + struct pipeline_msg_req_table_mtr_profile_add table_mtr_profile_add; + struct pipeline_msg_req_table_mtr_profile_delete table_mtr_profile_delete; + struct pipeline_msg_req_table_rule_mtr_read table_rule_mtr_read; + struct pipeline_msg_req_table_dscp_table_update table_dscp_table_update; + struct pipeline_msg_req_table_rule_ttl_read table_rule_ttl_read; + }; +}; + +struct pipeline_msg_rsp_port_in_stats_read { + struct rte_pipeline_port_in_stats stats; +}; + +struct pipeline_msg_rsp_port_out_stats_read { + struct rte_pipeline_port_out_stats stats; +}; + +struct pipeline_msg_rsp_table_stats_read { + struct rte_pipeline_table_stats stats; +}; + +struct pipeline_msg_rsp_table_rule_add { + void *data; +}; + +struct pipeline_msg_rsp_table_rule_add_default { + void *data; +}; + +struct pipeline_msg_rsp_table_rule_add_bulk { + uint32_t n_rules; +}; + +struct pipeline_msg_rsp_table_rule_stats_read { + struct rte_table_action_stats_counters stats; +}; + +struct pipeline_msg_rsp_table_rule_mtr_read { + struct rte_table_action_mtr_counters stats; +}; + +struct pipeline_msg_rsp_table_rule_ttl_read { + struct rte_table_action_ttl_counters stats; +}; + +struct pipeline_msg_rsp { + int status; + + RTE_STD_C11 + union { + struct pipeline_msg_rsp_port_in_stats_read port_in_stats_read; + struct pipeline_msg_rsp_port_out_stats_read port_out_stats_read; + struct pipeline_msg_rsp_table_stats_read table_stats_read; + struct pipeline_msg_rsp_table_rule_add table_rule_add; + struct pipeline_msg_rsp_table_rule_add_default table_rule_add_default; + struct pipeline_msg_rsp_table_rule_add_bulk table_rule_add_bulk; + struct pipeline_msg_rsp_table_rule_stats_read table_rule_stats_read; + struct pipeline_msg_rsp_table_rule_mtr_read table_rule_mtr_read; + struct pipeline_msg_rsp_table_rule_ttl_read table_rule_ttl_read; + }; +}; + +/** + * Master thread + */ +static struct pipeline_msg_req * +pipeline_msg_alloc(void) +{ + size_t size = RTE_MAX(sizeof(struct pipeline_msg_req), + sizeof(struct pipeline_msg_rsp)); + + return calloc(1, size); +} + +static void +pipeline_msg_free(struct pipeline_msg_rsp *rsp) +{ + free(rsp); +} + +static struct pipeline_msg_rsp * +pipeline_msg_send_recv(struct pipeline *p, + struct pipeline_msg_req *req) +{ + struct rte_ring *msgq_req = p->msgq_req; + struct rte_ring *msgq_rsp = p->msgq_rsp; + struct pipeline_msg_rsp *rsp; + int status; + + /* send */ + do { + status = rte_ring_sp_enqueue(msgq_req, req); + } while (status == -ENOBUFS); + + /* recv */ + do { + status = rte_ring_sc_dequeue(msgq_rsp, (void **) &rsp); + } while (status != 0); + + return rsp; +} + +int +pipeline_port_in_stats_read(const char *pipeline_name, + uint32_t port_id, + struct rte_pipeline_port_in_stats *stats, + int clear) +{ + struct pipeline *p; + struct pipeline_msg_req *req; + struct pipeline_msg_rsp *rsp; + int status; + + /* Check input params */ + if ((pipeline_name == NULL) || + (stats == NULL)) + return -1; + + p = pipeline_find(pipeline_name); + if ((p == NULL) || + (port_id >= p->n_ports_in)) + return -1; + + if (!pipeline_is_running(p)) { + status = rte_pipeline_port_in_stats_read(p->p, + port_id, + stats, + clear); + + return status; + } + + /* Allocate request */ + req = pipeline_msg_alloc(); + if (req == NULL) + return -1; + + /* Write request */ + req->type = PIPELINE_REQ_PORT_IN_STATS_READ; + req->id = port_id; + req->port_in_stats_read.clear = clear; + + /* Send request and wait for response */ + rsp = pipeline_msg_send_recv(p, req); + if (rsp == NULL) + return -1; + + /* Read response */ + status = rsp->status; + if (status) + memcpy(stats, &rsp->port_in_stats_read.stats, sizeof(*stats)); + + /* Free response */ + pipeline_msg_free(rsp); + + return status; +} + +int +pipeline_port_in_enable(const char *pipeline_name, + uint32_t port_id) +{ + struct pipeline *p; + struct pipeline_msg_req *req; + struct pipeline_msg_rsp *rsp; + int status; + + /* Check input params */ + if (pipeline_name == NULL) + return -1; + + p = pipeline_find(pipeline_name); + if ((p == NULL) || + (port_id >= p->n_ports_in)) + return -1; + + if (!pipeline_is_running(p)) { + status = rte_pipeline_port_in_enable(p->p, port_id); + return status; + } + + /* Allocate request */ + req = pipeline_msg_alloc(); + if (req == NULL) + return -1; + + /* Write request */ + req->type = PIPELINE_REQ_PORT_IN_ENABLE; + req->id = port_id; + + /* Send request and wait for response */ + rsp = pipeline_msg_send_recv(p, req); + if (rsp == NULL) + return -1; + + /* Read response */ + status = rsp->status; + + /* Free response */ + pipeline_msg_free(rsp); + + return status; +} + +int +pipeline_port_in_disable(const char *pipeline_name, + uint32_t port_id) +{ + struct pipeline *p; + struct pipeline_msg_req *req; + struct pipeline_msg_rsp *rsp; + int status; + + /* Check input params */ + if (pipeline_name == NULL) + return -1; + + p = pipeline_find(pipeline_name); + if ((p == NULL) || + (port_id >= p->n_ports_in)) + return -1; + + if (!pipeline_is_running(p)) { + status = rte_pipeline_port_in_disable(p->p, port_id); + return status; + } + + /* Allocate request */ + req = pipeline_msg_alloc(); + if (req == NULL) + return -1; + + /* Write request */ + req->type = PIPELINE_REQ_PORT_IN_DISABLE; + req->id = port_id; + + /* Send request and wait for response */ + rsp = pipeline_msg_send_recv(p, req); + if (rsp == NULL) + return -1; + + /* Read response */ + status = rsp->status; + + /* Free response */ + pipeline_msg_free(rsp); + + return status; +} + +int +pipeline_port_out_stats_read(const char *pipeline_name, + uint32_t port_id, + struct rte_pipeline_port_out_stats *stats, + int clear) +{ + struct pipeline *p; + struct pipeline_msg_req *req; + struct pipeline_msg_rsp *rsp; + int status; + + /* Check input params */ + if ((pipeline_name == NULL) || + (stats == NULL)) + return -1; + + p = pipeline_find(pipeline_name); + if ((p == NULL) || + (port_id >= p->n_ports_out)) + return -1; + + if (!pipeline_is_running(p)) { + status = rte_pipeline_port_out_stats_read(p->p, + port_id, + stats, + clear); + + return status; + } + + /* Allocate request */ + req = pipeline_msg_alloc(); + if (req == NULL) + return -1; + + /* Write request */ + req->type = PIPELINE_REQ_PORT_OUT_STATS_READ; + req->id = port_id; + req->port_out_stats_read.clear = clear; + + /* Send request and wait for response */ + rsp = pipeline_msg_send_recv(p, req); + if (rsp == NULL) + return -1; + + /* Read response */ + status = rsp->status; + if (status) + memcpy(stats, &rsp->port_out_stats_read.stats, sizeof(*stats)); + + /* Free response */ + pipeline_msg_free(rsp); + + return status; +} + +int +pipeline_table_stats_read(const char *pipeline_name, + uint32_t table_id, + struct rte_pipeline_table_stats *stats, + int clear) +{ + struct pipeline *p; + struct pipeline_msg_req *req; + struct pipeline_msg_rsp *rsp; + int status; + + /* Check input params */ + if ((pipeline_name == NULL) || + (stats == NULL)) + return -1; + + p = pipeline_find(pipeline_name); + if ((p == NULL) || + (table_id >= p->n_tables)) + return -1; + + if (!pipeline_is_running(p)) { + status = rte_pipeline_table_stats_read(p->p, + table_id, + stats, + clear); + + return status; + } + + /* Allocate request */ + req = pipeline_msg_alloc(); + if (req == NULL) + return -1; + + /* Write request */ + req->type = PIPELINE_REQ_TABLE_STATS_READ; + req->id = table_id; + req->table_stats_read.clear = clear; + + /* Send request and wait for response */ + rsp = pipeline_msg_send_recv(p, req); + if (rsp == NULL) + return -1; + + /* Read response */ + status = rsp->status; + if (status) + memcpy(stats, &rsp->table_stats_read.stats, sizeof(*stats)); + + /* Free response */ + pipeline_msg_free(rsp); + + return status; +} + +static int +match_check(struct table_rule_match *match, + struct pipeline *p, + uint32_t table_id) +{ + struct table *table; + + if ((match == NULL) || + (p == NULL) || + (table_id >= p->n_tables)) + return -1; + + table = &p->table[table_id]; + if (match->match_type != table->params.match_type) + return -1; + + switch (match->match_type) { + case TABLE_ACL: + { + struct table_acl_params *t = &table->params.match.acl; + struct table_rule_match_acl *r = &match->match.acl; + + if ((r->ip_version && (t->ip_version == 0)) || + ((r->ip_version == 0) && t->ip_version)) + return -1; + + if (r->ip_version) { + if ((r->sa_depth > 32) || + (r->da_depth > 32)) + return -1; + } else { + if ((r->sa_depth > 128) || + (r->da_depth > 128)) + return -1; + } + return 0; + } + + case TABLE_ARRAY: + return 0; + + case TABLE_HASH: + return 0; + + case TABLE_LPM: + { + struct table_lpm_params *t = &table->params.match.lpm; + struct table_rule_match_lpm *r = &match->match.lpm; + + if ((r->ip_version && (t->key_size != 4)) || + ((r->ip_version == 0) && (t->key_size != 16))) + return -1; + + if (r->ip_version) { + if (r->depth > 32) + return -1; + } else { + if (r->depth > 128) + return -1; + } + return 0; + } + + case TABLE_STUB: + return -1; + + default: + return -1; + } +} + static int -thread_pipeline_enable(struct app_thread_data *t, - struct thread_pipeline_enable_msg_req *req) +action_check(struct table_rule_action *action, + struct pipeline *p, + uint32_t table_id) { - struct app_thread_pipeline_data *p; + struct table_action_profile *ap; + + if ((action == NULL) || + (p == NULL) || + (table_id >= p->n_tables)) + return -1; + + ap = p->table[table_id].ap; + if (action->action_mask != ap->params.action_mask) + return -1; + + if (action->action_mask & (1LLU << RTE_TABLE_ACTION_FWD)) { + if ((action->fwd.action == RTE_PIPELINE_ACTION_PORT) && + (action->fwd.id >= p->n_ports_out)) + return -1; - if (req->f_run == NULL) { - if (t->n_regular >= APP_MAX_THREAD_PIPELINES) + if ((action->fwd.action == RTE_PIPELINE_ACTION_TABLE) && + (action->fwd.id >= p->n_tables)) return -1; - } else { - if (t->n_custom >= APP_MAX_THREAD_PIPELINES) + } + + if (action->action_mask & (1LLU << RTE_TABLE_ACTION_MTR)) { + uint32_t tc_mask0 = (1 << ap->params.mtr.n_tc) - 1; + uint32_t tc_mask1 = action->mtr.tc_mask; + + if (tc_mask1 != tc_mask0) return -1; } - p = (req->f_run == NULL) ? - &t->regular[t->n_regular] : - &t->custom[t->n_custom]; + if (action->action_mask & (1LLU << RTE_TABLE_ACTION_TM)) { + uint32_t n_subports_per_port = + ap->params.tm.n_subports_per_port; + uint32_t n_pipes_per_subport = + ap->params.tm.n_pipes_per_subport; + uint32_t subport_id = action->tm.subport_id; + uint32_t pipe_id = action->tm.pipe_id; - p->pipeline_id = req->pipeline_id; - p->be = req->be; - p->f_run = req->f_run; - p->f_timer = req->f_timer; - p->timer_period = req->timer_period; - p->deadline = 0; + if ((subport_id >= n_subports_per_port) || + (pipe_id >= n_pipes_per_subport)) + return -1; + } - if (req->f_run == NULL) - t->n_regular++; - else - t->n_custom++; + if (action->action_mask & (1LLU << RTE_TABLE_ACTION_ENCAP)) { + uint64_t encap_mask = ap->params.encap.encap_mask; + enum rte_table_action_encap_type type = action->encap.type; + + if ((encap_mask & (1LLU << type)) == 0) + return -1; + } + + if (action->action_mask & (1LLU << RTE_TABLE_ACTION_NAT)) { + int ip_version0 = ap->params.common.ip_version; + int ip_version1 = action->nat.ip_version; + + if ((ip_version1 && (ip_version0 == 0)) || + ((ip_version1 == 0) && ip_version0)) + return -1; + } return 0; } static int -thread_pipeline_disable(struct app_thread_data *t, - struct thread_pipeline_disable_msg_req *req) +action_default_check(struct table_rule_action *action, + struct pipeline *p, + uint32_t table_id) { - uint32_t n_regular = RTE_MIN(t->n_regular, RTE_DIM(t->regular)); - uint32_t n_custom = RTE_MIN(t->n_custom, RTE_DIM(t->custom)); - uint32_t i; + if ((action == NULL) || + (action->action_mask != (1LLU << RTE_TABLE_ACTION_FWD)) || + (p == NULL) || + (table_id >= p->n_tables)) + return -1; - /* search regular pipelines of current thread */ - for (i = 0; i < n_regular; i++) { - if (t->regular[i].pipeline_id != req->pipeline_id) - continue; + if (action->action_mask & (1LLU << RTE_TABLE_ACTION_FWD)) { + if ((action->fwd.action == RTE_PIPELINE_ACTION_PORT) && + (action->fwd.id >= p->n_ports_out)) + return -1; + + if ((action->fwd.action == RTE_PIPELINE_ACTION_TABLE) && + (action->fwd.id >= p->n_tables)) + return -1; + } - if (i < n_regular - 1) - memcpy(&t->regular[i], - &t->regular[i+1], - (n_regular - 1 - i) * sizeof(struct app_thread_pipeline_data)); + return 0; +} - n_regular--; - t->n_regular = n_regular; +union table_rule_match_low_level { + struct rte_table_acl_rule_add_params acl_add; + struct rte_table_acl_rule_delete_params acl_delete; + struct rte_table_array_key array; + uint8_t hash[TABLE_RULE_MATCH_SIZE_MAX]; + struct rte_table_lpm_key lpm_ipv4; + struct rte_table_lpm_ipv6_key lpm_ipv6; +}; +static int +match_convert(struct table_rule_match *mh, + union table_rule_match_low_level *ml, + int add); + +static int +action_convert(struct rte_table_action *a, + struct table_rule_action *action, + struct rte_pipeline_table_entry *data); + +int +pipeline_table_rule_add(const char *pipeline_name, + uint32_t table_id, + struct table_rule_match *match, + struct table_rule_action *action, + void **data) +{ + struct pipeline *p; + struct pipeline_msg_req *req; + struct pipeline_msg_rsp *rsp; + int status; + + /* Check input params */ + if ((pipeline_name == NULL) || + (match == NULL) || + (action == NULL) || + (data == NULL)) + return -1; + + p = pipeline_find(pipeline_name); + if ((p == NULL) || + (table_id >= p->n_tables) || + match_check(match, p, table_id) || + action_check(action, p, table_id)) + return -1; + + if (!pipeline_is_running(p)) { + struct rte_table_action *a = p->table[table_id].a; + union table_rule_match_low_level match_ll; + struct rte_pipeline_table_entry *data_in, *data_out; + int key_found; + uint8_t *buffer; + + buffer = calloc(TABLE_RULE_ACTION_SIZE_MAX, sizeof(uint8_t)); + if (buffer == NULL) + return -1; + + /* Table match-action rule conversion */ + data_in = (struct rte_pipeline_table_entry *)buffer; + + status = match_convert(match, &match_ll, 1); + if (status) { + free(buffer); + return -1; + } + + status = action_convert(a, action, data_in); + if (status) { + free(buffer); + return -1; + } + + /* Add rule (match, action) to table */ + status = rte_pipeline_table_entry_add(p->p, + table_id, + &match_ll, + data_in, + &key_found, + &data_out); + if (status) { + free(buffer); + return -1; + } + + /* Write Response */ + *data = data_out; + + free(buffer); return 0; } - /* search custom pipelines of current thread */ - for (i = 0; i < n_custom; i++) { - if (t->custom[i].pipeline_id != req->pipeline_id) - continue; + /* Allocate request */ + req = pipeline_msg_alloc(); + if (req == NULL) + return -1; + + /* Write request */ + req->type = PIPELINE_REQ_TABLE_RULE_ADD; + req->id = table_id; + memcpy(&req->table_rule_add.match, match, sizeof(*match)); + memcpy(&req->table_rule_add.action, action, sizeof(*action)); + + /* Send request and wait for response */ + rsp = pipeline_msg_send_recv(p, req); + if (rsp == NULL) + return -1; - if (i < n_custom - 1) - memcpy(&t->custom[i], - &t->custom[i+1], - (n_custom - 1 - i) * sizeof(struct app_thread_pipeline_data)); + /* Read response */ + status = rsp->status; + if (status == 0) + *data = rsp->table_rule_add.data; - n_custom--; - t->n_custom = n_custom; + /* Free response */ + pipeline_msg_free(rsp); + return status; +} + +int +pipeline_table_rule_add_default(const char *pipeline_name, + uint32_t table_id, + struct table_rule_action *action, + void **data) +{ + struct pipeline *p; + struct pipeline_msg_req *req; + struct pipeline_msg_rsp *rsp; + int status; + + /* Check input params */ + if ((pipeline_name == NULL) || + (action == NULL) || + (data == NULL)) + return -1; + + p = pipeline_find(pipeline_name); + if ((p == NULL) || + (table_id >= p->n_tables) || + action_default_check(action, p, table_id)) + return -1; + + if (!pipeline_is_running(p)) { + struct rte_pipeline_table_entry *data_in, *data_out; + uint8_t *buffer; + + buffer = calloc(TABLE_RULE_ACTION_SIZE_MAX, sizeof(uint8_t)); + if (buffer == NULL) + return -1; + + /* Apply actions */ + data_in = (struct rte_pipeline_table_entry *)buffer; + + data_in->action = action->fwd.action; + if (action->fwd.action == RTE_PIPELINE_ACTION_PORT) + data_in->port_id = action->fwd.id; + if (action->fwd.action == RTE_PIPELINE_ACTION_TABLE) + data_in->table_id = action->fwd.id; + + /* Add default rule to table */ + status = rte_pipeline_table_default_entry_add(p->p, + table_id, + data_in, + &data_out); + if (status) { + free(buffer); + return -1; + } + + /* Write Response */ + *data = data_out; + + free(buffer); return 0; } - /* return if pipeline not found */ - return -1; + /* Allocate request */ + req = pipeline_msg_alloc(); + if (req == NULL) + return -1; + + /* Write request */ + req->type = PIPELINE_REQ_TABLE_RULE_ADD_DEFAULT; + req->id = table_id; + memcpy(&req->table_rule_add_default.action, action, sizeof(*action)); + + /* Send request and wait for response */ + rsp = pipeline_msg_send_recv(p, req); + if (rsp == NULL) + return -1; + + /* Read response */ + status = rsp->status; + if (status == 0) + *data = rsp->table_rule_add_default.data; + + /* Free response */ + pipeline_msg_free(rsp); + + return status; } -static int -thread_msg_req_handle(struct app_thread_data *t) +int +pipeline_table_rule_add_bulk(const char *pipeline_name, + uint32_t table_id, + struct table_rule_match *match, + struct table_rule_action *action, + void **data, + uint32_t *n_rules) { - void *msg_ptr; - struct thread_msg_req *req; - struct thread_msg_rsp *rsp; + struct pipeline *p; + struct pipeline_msg_req *req; + struct pipeline_msg_rsp *rsp; + uint32_t i; + int status; - msg_ptr = thread_msg_recv(t->msgq_in); - req = msg_ptr; - rsp = msg_ptr; + /* Check input params */ + if ((pipeline_name == NULL) || + (match == NULL) || + (action == NULL) || + (data == NULL) || + (n_rules == NULL) || + (*n_rules == 0)) + return -1; - if (req != NULL) - switch (req->type) { - case THREAD_MSG_REQ_PIPELINE_ENABLE: { - rsp->status = thread_pipeline_enable(t, - (struct thread_pipeline_enable_msg_req *) req); - thread_msg_send(t->msgq_out, rsp); - break; + p = pipeline_find(pipeline_name); + if ((p == NULL) || + (table_id >= p->n_tables)) + return -1; + + for (i = 0; i < *n_rules; i++) + if (match_check(match, p, table_id) || + action_check(action, p, table_id)) + return -1; + + if (!pipeline_is_running(p)) { + struct rte_table_action *a = p->table[table_id].a; + union table_rule_match_low_level *match_ll; + uint8_t *action_ll; + void **match_ll_ptr; + struct rte_pipeline_table_entry **action_ll_ptr; + struct rte_pipeline_table_entry **entries_ptr = + (struct rte_pipeline_table_entry **)data; + uint32_t bulk = + (p->table[table_id].params.match_type == TABLE_ACL) ? 1 : 0; + int *found; + + /* Memory allocation */ + match_ll = calloc(*n_rules, sizeof(union table_rule_match_low_level)); + action_ll = calloc(*n_rules, TABLE_RULE_ACTION_SIZE_MAX); + match_ll_ptr = calloc(*n_rules, sizeof(void *)); + action_ll_ptr = + calloc(*n_rules, sizeof(struct rte_pipeline_table_entry *)); + found = calloc(*n_rules, sizeof(int)); + + if (match_ll == NULL || + action_ll == NULL || + match_ll_ptr == NULL || + action_ll_ptr == NULL || + found == NULL) + goto fail; + + for (i = 0; i < *n_rules; i++) { + match_ll_ptr[i] = (void *)&match_ll[i]; + action_ll_ptr[i] = + (struct rte_pipeline_table_entry *)&action_ll[i * TABLE_RULE_ACTION_SIZE_MAX]; } - case THREAD_MSG_REQ_PIPELINE_DISABLE: { - rsp->status = thread_pipeline_disable(t, - (struct thread_pipeline_disable_msg_req *) req); - thread_msg_send(t->msgq_out, rsp); - break; + /* Rule match conversion */ + for (i = 0; i < *n_rules; i++) { + status = match_convert(&match[i], match_ll_ptr[i], 1); + if (status) + goto fail; } - case THREAD_MSG_REQ_HEADROOM_READ: { - struct thread_headroom_read_msg_rsp *rsp = - (struct thread_headroom_read_msg_rsp *) - req; + /* Rule action conversion */ + for (i = 0; i < *n_rules; i++) { + status = action_convert(a, &action[i], action_ll_ptr[i]); + if (status) + goto fail; + } - rsp->headroom_ratio = t->headroom_ratio; - rsp->status = 0; - thread_msg_send(t->msgq_out, rsp); - break; + /* Add rule (match, action) to table */ + if (bulk) { + status = rte_pipeline_table_entry_add_bulk(p->p, + table_id, + match_ll_ptr, + action_ll_ptr, + *n_rules, + found, + entries_ptr); + if (status) + *n_rules = 0; + } else { + for (i = 0; i < *n_rules; i++) { + status = rte_pipeline_table_entry_add(p->p, + table_id, + match_ll_ptr[i], + action_ll_ptr[i], + &found[i], + &entries_ptr[i]); + if (status) { + *n_rules = i; + break; + } + } } - default: - break; + + /* Free */ + free(found); + free(action_ll_ptr); + free(match_ll_ptr); + free(action_ll); + free(match_ll); + + return status; + +fail: + free(found); + free(action_ll_ptr); + free(match_ll_ptr); + free(action_ll); + free(match_ll); + + *n_rules = 0; + return -1; + } + + /* Allocate request */ + req = pipeline_msg_alloc(); + if (req == NULL) + return -1; + + /* Write request */ + req->type = PIPELINE_REQ_TABLE_RULE_ADD_BULK; + req->id = table_id; + req->table_rule_add_bulk.match = match; + req->table_rule_add_bulk.action = action; + req->table_rule_add_bulk.data = data; + req->table_rule_add_bulk.n_rules = *n_rules; + req->table_rule_add_bulk.bulk = + (p->table[table_id].params.match_type == TABLE_ACL) ? 1 : 0; + + /* Send request and wait for response */ + rsp = pipeline_msg_send_recv(p, req); + if (rsp == NULL) + return -1; + + /* Read response */ + status = rsp->status; + if (status == 0) + *n_rules = rsp->table_rule_add_bulk.n_rules; + + /* Free response */ + pipeline_msg_free(rsp); + + return status; +} + +int +pipeline_table_rule_delete(const char *pipeline_name, + uint32_t table_id, + struct table_rule_match *match) +{ + struct pipeline *p; + struct pipeline_msg_req *req; + struct pipeline_msg_rsp *rsp; + int status; + + /* Check input params */ + if ((pipeline_name == NULL) || + (match == NULL)) + return -1; + + p = pipeline_find(pipeline_name); + if ((p == NULL) || + (table_id >= p->n_tables) || + match_check(match, p, table_id)) + return -1; + + if (!pipeline_is_running(p)) { + union table_rule_match_low_level match_ll; + int key_found; + + status = match_convert(match, &match_ll, 0); + if (status) + return -1; + + status = rte_pipeline_table_entry_delete(p->p, + table_id, + &match_ll, + &key_found, + NULL); + + return status; + } + + /* Allocate request */ + req = pipeline_msg_alloc(); + if (req == NULL) + return -1; + + /* Write request */ + req->type = PIPELINE_REQ_TABLE_RULE_DELETE; + req->id = table_id; + memcpy(&req->table_rule_delete.match, match, sizeof(*match)); + + /* Send request and wait for response */ + rsp = pipeline_msg_send_recv(p, req); + if (rsp == NULL) + return -1; + + /* Read response */ + status = rsp->status; + + /* Free response */ + pipeline_msg_free(rsp); + + return status; +} + +int +pipeline_table_rule_delete_default(const char *pipeline_name, + uint32_t table_id) +{ + struct pipeline *p; + struct pipeline_msg_req *req; + struct pipeline_msg_rsp *rsp; + int status; + + /* Check input params */ + if (pipeline_name == NULL) + return -1; + + p = pipeline_find(pipeline_name); + if ((p == NULL) || + (table_id >= p->n_tables)) + return -1; + + if (!pipeline_is_running(p)) { + status = rte_pipeline_table_default_entry_delete(p->p, + table_id, + NULL); + + return status; + } + + /* Allocate request */ + req = pipeline_msg_alloc(); + if (req == NULL) + return -1; + + /* Write request */ + req->type = PIPELINE_REQ_TABLE_RULE_DELETE_DEFAULT; + req->id = table_id; + + /* Send request and wait for response */ + rsp = pipeline_msg_send_recv(p, req); + if (rsp == NULL) + return -1; + + /* Read response */ + status = rsp->status; + + /* Free response */ + pipeline_msg_free(rsp); + + return status; +} + +int +pipeline_table_rule_stats_read(const char *pipeline_name, + uint32_t table_id, + void *data, + struct rte_table_action_stats_counters *stats, + int clear) +{ + struct pipeline *p; + struct pipeline_msg_req *req; + struct pipeline_msg_rsp *rsp; + int status; + + /* Check input params */ + if ((pipeline_name == NULL) || + (data == NULL) || + (stats == NULL)) + return -1; + + p = pipeline_find(pipeline_name); + if ((p == NULL) || + (table_id >= p->n_tables)) + return -1; + + if (!pipeline_is_running(p)) { + struct rte_table_action *a = p->table[table_id].a; + + status = rte_table_action_stats_read(a, + data, + stats, + clear); + + return status; + } + + /* Allocate request */ + req = pipeline_msg_alloc(); + if (req == NULL) + return -1; + + /* Write request */ + req->type = PIPELINE_REQ_TABLE_RULE_STATS_READ; + req->id = table_id; + req->table_rule_stats_read.data = data; + req->table_rule_stats_read.clear = clear; + + /* Send request and wait for response */ + rsp = pipeline_msg_send_recv(p, req); + if (rsp == NULL) + return -1; + + /* Read response */ + status = rsp->status; + if (status) + memcpy(stats, &rsp->table_rule_stats_read.stats, sizeof(*stats)); + + /* Free response */ + pipeline_msg_free(rsp); + + return status; +} + +int +pipeline_table_mtr_profile_add(const char *pipeline_name, + uint32_t table_id, + uint32_t meter_profile_id, + struct rte_table_action_meter_profile *profile) +{ + struct pipeline *p; + struct pipeline_msg_req *req; + struct pipeline_msg_rsp *rsp; + int status; + + /* Check input params */ + if ((pipeline_name == NULL) || + (profile == NULL)) + return -1; + + p = pipeline_find(pipeline_name); + if ((p == NULL) || + (table_id >= p->n_tables)) + return -1; + + if (!pipeline_is_running(p)) { + struct rte_table_action *a = p->table[table_id].a; + + status = rte_table_action_meter_profile_add(a, + meter_profile_id, + profile); + + return status; + } + + /* Allocate request */ + req = pipeline_msg_alloc(); + if (req == NULL) + return -1; + + /* Write request */ + req->type = PIPELINE_REQ_TABLE_MTR_PROFILE_ADD; + req->id = table_id; + req->table_mtr_profile_add.meter_profile_id = meter_profile_id; + memcpy(&req->table_mtr_profile_add.profile, profile, sizeof(*profile)); + + /* Send request and wait for response */ + rsp = pipeline_msg_send_recv(p, req); + if (rsp == NULL) + return -1; + + /* Read response */ + status = rsp->status; + + /* Free response */ + pipeline_msg_free(rsp); + + return status; +} + +int +pipeline_table_mtr_profile_delete(const char *pipeline_name, + uint32_t table_id, + uint32_t meter_profile_id) +{ + struct pipeline *p; + struct pipeline_msg_req *req; + struct pipeline_msg_rsp *rsp; + int status; + + /* Check input params */ + if (pipeline_name == NULL) + return -1; + + p = pipeline_find(pipeline_name); + if ((p == NULL) || + (table_id >= p->n_tables)) + return -1; + + if (!pipeline_is_running(p)) { + struct rte_table_action *a = p->table[table_id].a; + + status = rte_table_action_meter_profile_delete(a, + meter_profile_id); + + return status; + } + + /* Allocate request */ + req = pipeline_msg_alloc(); + if (req == NULL) + return -1; + + /* Write request */ + req->type = PIPELINE_REQ_TABLE_MTR_PROFILE_DELETE; + req->id = table_id; + req->table_mtr_profile_delete.meter_profile_id = meter_profile_id; + + /* Send request and wait for response */ + rsp = pipeline_msg_send_recv(p, req); + if (rsp == NULL) + return -1; + + /* Read response */ + status = rsp->status; + + /* Free response */ + pipeline_msg_free(rsp); + + return status; +} + +int +pipeline_table_rule_mtr_read(const char *pipeline_name, + uint32_t table_id, + void *data, + uint32_t tc_mask, + struct rte_table_action_mtr_counters *stats, + int clear) +{ + struct pipeline *p; + struct pipeline_msg_req *req; + struct pipeline_msg_rsp *rsp; + int status; + + /* Check input params */ + if ((pipeline_name == NULL) || + (data == NULL) || + (stats == NULL)) + return -1; + + p = pipeline_find(pipeline_name); + if ((p == NULL) || + (table_id >= p->n_tables)) + return -1; + + if (!pipeline_is_running(p)) { + struct rte_table_action *a = p->table[table_id].a; + + status = rte_table_action_meter_read(a, + data, + tc_mask, + stats, + clear); + + return status; + } + + /* Allocate request */ + req = pipeline_msg_alloc(); + if (req == NULL) + return -1; + + /* Write request */ + req->type = PIPELINE_REQ_TABLE_RULE_MTR_READ; + req->id = table_id; + req->table_rule_mtr_read.data = data; + req->table_rule_mtr_read.tc_mask = tc_mask; + req->table_rule_mtr_read.clear = clear; + + /* Send request and wait for response */ + rsp = pipeline_msg_send_recv(p, req); + if (rsp == NULL) + return -1; + + /* Read response */ + status = rsp->status; + if (status) + memcpy(stats, &rsp->table_rule_mtr_read.stats, sizeof(*stats)); + + /* Free response */ + pipeline_msg_free(rsp); + + return status; +} + +int +pipeline_table_dscp_table_update(const char *pipeline_name, + uint32_t table_id, + uint64_t dscp_mask, + struct rte_table_action_dscp_table *dscp_table) +{ + struct pipeline *p; + struct pipeline_msg_req *req; + struct pipeline_msg_rsp *rsp; + int status; + + /* Check input params */ + if ((pipeline_name == NULL) || + (dscp_table == NULL)) + return -1; + + p = pipeline_find(pipeline_name); + if ((p == NULL) || + (table_id >= p->n_tables)) + return -1; + + if (!pipeline_is_running(p)) { + struct rte_table_action *a = p->table[table_id].a; + + status = rte_table_action_dscp_table_update(a, + dscp_mask, + dscp_table); + + return status; + } + + /* Allocate request */ + req = pipeline_msg_alloc(); + if (req == NULL) + return -1; + + /* Write request */ + req->type = PIPELINE_REQ_TABLE_DSCP_TABLE_UPDATE; + req->id = table_id; + req->table_dscp_table_update.dscp_mask = dscp_mask; + memcpy(&req->table_dscp_table_update.dscp_table, + dscp_table, sizeof(*dscp_table)); + + /* Send request and wait for response */ + rsp = pipeline_msg_send_recv(p, req); + if (rsp == NULL) + return -1; + + /* Read response */ + status = rsp->status; + + /* Free response */ + pipeline_msg_free(rsp); + + return status; +} + +int +pipeline_table_rule_ttl_read(const char *pipeline_name, + uint32_t table_id, + void *data, + struct rte_table_action_ttl_counters *stats, + int clear) +{ + struct pipeline *p; + struct pipeline_msg_req *req; + struct pipeline_msg_rsp *rsp; + int status; + + /* Check input params */ + if ((pipeline_name == NULL) || + (data == NULL) || + (stats == NULL)) + return -1; + + p = pipeline_find(pipeline_name); + if ((p == NULL) || + (table_id >= p->n_tables)) + return -1; + + if (!pipeline_is_running(p)) { + struct rte_table_action *a = p->table[table_id].a; + + status = rte_table_action_ttl_read(a, + data, + stats, + clear); + + return status; + } + + /* Allocate request */ + req = pipeline_msg_alloc(); + if (req == NULL) + return -1; + + /* Write request */ + req->type = PIPELINE_REQ_TABLE_RULE_TTL_READ; + req->id = table_id; + req->table_rule_ttl_read.data = data; + req->table_rule_ttl_read.clear = clear; + + /* Send request and wait for response */ + rsp = pipeline_msg_send_recv(p, req); + if (rsp == NULL) + return -1; + + /* Read response */ + status = rsp->status; + if (status) + memcpy(stats, &rsp->table_rule_ttl_read.stats, sizeof(*stats)); + + /* Free response */ + pipeline_msg_free(rsp); + + return status; +} + +/** + * Data plane threads: message handling + */ +static inline struct pipeline_msg_req * +pipeline_msg_recv(struct rte_ring *msgq_req) +{ + struct pipeline_msg_req *req; + + int status = rte_ring_sc_dequeue(msgq_req, (void **) &req); + + if (status != 0) + return NULL; + + return req; +} + +static inline void +pipeline_msg_send(struct rte_ring *msgq_rsp, + struct pipeline_msg_rsp *rsp) +{ + int status; + + do { + status = rte_ring_sp_enqueue(msgq_rsp, rsp); + } while (status == -ENOBUFS); +} + +static struct pipeline_msg_rsp * +pipeline_msg_handle_port_in_stats_read(struct pipeline_data *p, + struct pipeline_msg_req *req) +{ + struct pipeline_msg_rsp *rsp = (struct pipeline_msg_rsp *) req; + uint32_t port_id = req->id; + int clear = req->port_in_stats_read.clear; + + rsp->status = rte_pipeline_port_in_stats_read(p->p, + port_id, + &rsp->port_in_stats_read.stats, + clear); + + return rsp; +} + +static struct pipeline_msg_rsp * +pipeline_msg_handle_port_in_enable(struct pipeline_data *p, + struct pipeline_msg_req *req) +{ + struct pipeline_msg_rsp *rsp = (struct pipeline_msg_rsp *) req; + uint32_t port_id = req->id; + + rsp->status = rte_pipeline_port_in_enable(p->p, + port_id); + + return rsp; +} + +static struct pipeline_msg_rsp * +pipeline_msg_handle_port_in_disable(struct pipeline_data *p, + struct pipeline_msg_req *req) +{ + struct pipeline_msg_rsp *rsp = (struct pipeline_msg_rsp *) req; + uint32_t port_id = req->id; + + rsp->status = rte_pipeline_port_in_disable(p->p, + port_id); + + return rsp; +} + +static struct pipeline_msg_rsp * +pipeline_msg_handle_port_out_stats_read(struct pipeline_data *p, + struct pipeline_msg_req *req) +{ + struct pipeline_msg_rsp *rsp = (struct pipeline_msg_rsp *) req; + uint32_t port_id = req->id; + int clear = req->port_out_stats_read.clear; + + rsp->status = rte_pipeline_port_out_stats_read(p->p, + port_id, + &rsp->port_out_stats_read.stats, + clear); + + return rsp; +} + +static struct pipeline_msg_rsp * +pipeline_msg_handle_table_stats_read(struct pipeline_data *p, + struct pipeline_msg_req *req) +{ + struct pipeline_msg_rsp *rsp = (struct pipeline_msg_rsp *) req; + uint32_t port_id = req->id; + int clear = req->table_stats_read.clear; + + rsp->status = rte_pipeline_table_stats_read(p->p, + port_id, + &rsp->table_stats_read.stats, + clear); + + return rsp; +} + +static int +match_convert_ipv6_depth(uint32_t depth, uint32_t *depth32) +{ + if (depth > 128) + return -1; + + switch (depth / 32) { + case 0: + depth32[0] = depth; + depth32[1] = 0; + depth32[2] = 0; + depth32[3] = 0; + return 0; + + case 1: + depth32[0] = 32; + depth32[1] = depth - 32; + depth32[2] = 0; + depth32[3] = 0; + return 0; + + case 2: + depth32[0] = 32; + depth32[1] = 32; + depth32[2] = depth - 64; + depth32[3] = 0; + return 0; + + case 3: + depth32[0] = 32; + depth32[1] = 32; + depth32[2] = 32; + depth32[3] = depth - 96; + return 0; + + case 4: + depth32[0] = 32; + depth32[1] = 32; + depth32[2] = 32; + depth32[3] = 32; + return 0; + + default: + return -1; + } +} + +static int +match_convert(struct table_rule_match *mh, + union table_rule_match_low_level *ml, + int add) +{ + memset(ml, 0, sizeof(*ml)); + + switch (mh->match_type) { + case TABLE_ACL: + if (mh->match.acl.ip_version) + if (add) { + ml->acl_add.field_value[0].value.u8 = + mh->match.acl.proto; + ml->acl_add.field_value[0].mask_range.u8 = + mh->match.acl.proto_mask; + + ml->acl_add.field_value[1].value.u32 = + mh->match.acl.ipv4.sa; + ml->acl_add.field_value[1].mask_range.u32 = + mh->match.acl.sa_depth; + + ml->acl_add.field_value[2].value.u32 = + mh->match.acl.ipv4.da; + ml->acl_add.field_value[2].mask_range.u32 = + mh->match.acl.da_depth; + + ml->acl_add.field_value[3].value.u16 = + mh->match.acl.sp0; + ml->acl_add.field_value[3].mask_range.u16 = + mh->match.acl.sp1; + + ml->acl_add.field_value[4].value.u16 = + mh->match.acl.dp0; + ml->acl_add.field_value[4].mask_range.u16 = + mh->match.acl.dp1; + + ml->acl_add.priority = + (int32_t) mh->match.acl.priority; + } else { + ml->acl_delete.field_value[0].value.u8 = + mh->match.acl.proto; + ml->acl_delete.field_value[0].mask_range.u8 = + mh->match.acl.proto_mask; + + ml->acl_delete.field_value[1].value.u32 = + mh->match.acl.ipv4.sa; + ml->acl_delete.field_value[1].mask_range.u32 = + mh->match.acl.sa_depth; + + ml->acl_delete.field_value[2].value.u32 = + mh->match.acl.ipv4.da; + ml->acl_delete.field_value[2].mask_range.u32 = + mh->match.acl.da_depth; + + ml->acl_delete.field_value[3].value.u16 = + mh->match.acl.sp0; + ml->acl_delete.field_value[3].mask_range.u16 = + mh->match.acl.sp1; + + ml->acl_delete.field_value[4].value.u16 = + mh->match.acl.dp0; + ml->acl_delete.field_value[4].mask_range.u16 = + mh->match.acl.dp1; + } + else + if (add) { + uint32_t *sa32 = + (uint32_t *) mh->match.acl.ipv6.sa; + uint32_t *da32 = + (uint32_t *) mh->match.acl.ipv6.da; + uint32_t sa32_depth[4], da32_depth[4]; + int status; + + status = match_convert_ipv6_depth( + mh->match.acl.sa_depth, + sa32_depth); + if (status) + return status; + + status = match_convert_ipv6_depth( + mh->match.acl.da_depth, + da32_depth); + if (status) + return status; + + ml->acl_add.field_value[0].value.u8 = + mh->match.acl.proto; + ml->acl_add.field_value[0].mask_range.u8 = + mh->match.acl.proto_mask; + + ml->acl_add.field_value[1].value.u32 = sa32[0]; + ml->acl_add.field_value[1].mask_range.u32 = + sa32_depth[0]; + ml->acl_add.field_value[2].value.u32 = sa32[1]; + ml->acl_add.field_value[2].mask_range.u32 = + sa32_depth[1]; + ml->acl_add.field_value[3].value.u32 = sa32[2]; + ml->acl_add.field_value[3].mask_range.u32 = + sa32_depth[2]; + ml->acl_add.field_value[4].value.u32 = sa32[3]; + ml->acl_add.field_value[4].mask_range.u32 = + sa32_depth[3]; + + ml->acl_add.field_value[5].value.u32 = da32[0]; + ml->acl_add.field_value[5].mask_range.u32 = + da32_depth[0]; + ml->acl_add.field_value[6].value.u32 = da32[1]; + ml->acl_add.field_value[6].mask_range.u32 = + da32_depth[1]; + ml->acl_add.field_value[7].value.u32 = da32[2]; + ml->acl_add.field_value[7].mask_range.u32 = + da32_depth[2]; + ml->acl_add.field_value[8].value.u32 = da32[3]; + ml->acl_add.field_value[8].mask_range.u32 = + da32_depth[3]; + + ml->acl_add.field_value[9].value.u16 = + mh->match.acl.sp0; + ml->acl_add.field_value[9].mask_range.u16 = + mh->match.acl.sp1; + + ml->acl_add.field_value[10].value.u16 = + mh->match.acl.dp0; + ml->acl_add.field_value[10].mask_range.u16 = + mh->match.acl.dp1; + + ml->acl_add.priority = + (int32_t) mh->match.acl.priority; + } else { + uint32_t *sa32 = + (uint32_t *) mh->match.acl.ipv6.sa; + uint32_t *da32 = + (uint32_t *) mh->match.acl.ipv6.da; + uint32_t sa32_depth[4], da32_depth[4]; + int status; + + status = match_convert_ipv6_depth( + mh->match.acl.sa_depth, + sa32_depth); + if (status) + return status; + + status = match_convert_ipv6_depth( + mh->match.acl.da_depth, + da32_depth); + if (status) + return status; + + ml->acl_delete.field_value[0].value.u8 = + mh->match.acl.proto; + ml->acl_delete.field_value[0].mask_range.u8 = + mh->match.acl.proto_mask; + + ml->acl_delete.field_value[1].value.u32 = + sa32[0]; + ml->acl_delete.field_value[1].mask_range.u32 = + sa32_depth[0]; + ml->acl_delete.field_value[2].value.u32 = + sa32[1]; + ml->acl_delete.field_value[2].mask_range.u32 = + sa32_depth[1]; + ml->acl_delete.field_value[3].value.u32 = + sa32[2]; + ml->acl_delete.field_value[3].mask_range.u32 = + sa32_depth[2]; + ml->acl_delete.field_value[4].value.u32 = + sa32[3]; + ml->acl_delete.field_value[4].mask_range.u32 = + sa32_depth[3]; + + ml->acl_delete.field_value[5].value.u32 = + da32[0]; + ml->acl_delete.field_value[5].mask_range.u32 = + da32_depth[0]; + ml->acl_delete.field_value[6].value.u32 = + da32[1]; + ml->acl_delete.field_value[6].mask_range.u32 = + da32_depth[1]; + ml->acl_delete.field_value[7].value.u32 = + da32[2]; + ml->acl_delete.field_value[7].mask_range.u32 = + da32_depth[2]; + ml->acl_delete.field_value[8].value.u32 = + da32[3]; + ml->acl_delete.field_value[8].mask_range.u32 = + da32_depth[3]; + + ml->acl_delete.field_value[9].value.u16 = + mh->match.acl.sp0; + ml->acl_delete.field_value[9].mask_range.u16 = + mh->match.acl.sp1; + + ml->acl_delete.field_value[10].value.u16 = + mh->match.acl.dp0; + ml->acl_delete.field_value[10].mask_range.u16 = + mh->match.acl.dp1; + } + return 0; + + case TABLE_ARRAY: + ml->array.pos = mh->match.array.pos; + return 0; + + case TABLE_HASH: + memcpy(ml->hash, mh->match.hash.key, sizeof(ml->hash)); + return 0; + + case TABLE_LPM: + if (mh->match.lpm.ip_version) { + ml->lpm_ipv4.ip = mh->match.lpm.ipv4; + ml->lpm_ipv4.depth = mh->match.lpm.depth; + } else { + memcpy(ml->lpm_ipv6.ip, + mh->match.lpm.ipv6, sizeof(ml->lpm_ipv6.ip)); + ml->lpm_ipv6.depth = mh->match.lpm.depth; } + return 0; + + default: + return -1; + } +} + +static int +action_convert(struct rte_table_action *a, + struct table_rule_action *action, + struct rte_pipeline_table_entry *data) +{ + int status; + + /* Apply actions */ + if (action->action_mask & (1LLU << RTE_TABLE_ACTION_FWD)) { + status = rte_table_action_apply(a, + data, + RTE_TABLE_ACTION_FWD, + &action->fwd); + + if (status) + return status; + } + + if (action->action_mask & (1LLU << RTE_TABLE_ACTION_LB)) { + status = rte_table_action_apply(a, + data, + RTE_TABLE_ACTION_LB, + &action->lb); + + if (status) + return status; + } + + if (action->action_mask & (1LLU << RTE_TABLE_ACTION_MTR)) { + status = rte_table_action_apply(a, + data, + RTE_TABLE_ACTION_MTR, + &action->mtr); + + if (status) + return status; + } + + if (action->action_mask & (1LLU << RTE_TABLE_ACTION_TM)) { + status = rte_table_action_apply(a, + data, + RTE_TABLE_ACTION_TM, + &action->tm); + + if (status) + return status; + } + + if (action->action_mask & (1LLU << RTE_TABLE_ACTION_ENCAP)) { + status = rte_table_action_apply(a, + data, + RTE_TABLE_ACTION_ENCAP, + &action->encap); + + if (status) + return status; + } + + if (action->action_mask & (1LLU << RTE_TABLE_ACTION_NAT)) { + status = rte_table_action_apply(a, + data, + RTE_TABLE_ACTION_NAT, + &action->nat); + + if (status) + return status; + } + + if (action->action_mask & (1LLU << RTE_TABLE_ACTION_TTL)) { + status = rte_table_action_apply(a, + data, + RTE_TABLE_ACTION_TTL, + &action->ttl); + + if (status) + return status; + } + + if (action->action_mask & (1LLU << RTE_TABLE_ACTION_STATS)) { + status = rte_table_action_apply(a, + data, + RTE_TABLE_ACTION_STATS, + &action->stats); + + if (status) + return status; + } + + if (action->action_mask & (1LLU << RTE_TABLE_ACTION_TIME)) { + status = rte_table_action_apply(a, + data, + RTE_TABLE_ACTION_TIME, + &action->time); + + if (status) + return status; + } + return 0; } -static void -thread_headroom_update(struct app_thread_data *t, uint64_t time) +static struct pipeline_msg_rsp * +pipeline_msg_handle_table_rule_add(struct pipeline_data *p, + struct pipeline_msg_req *req) { - uint64_t time_diff = time - t->headroom_time; + union table_rule_match_low_level match_ll; + struct pipeline_msg_rsp *rsp = (struct pipeline_msg_rsp *) req; + struct table_rule_match *match = &req->table_rule_add.match; + struct table_rule_action *action = &req->table_rule_add.action; + struct rte_pipeline_table_entry *data_in, *data_out; + uint32_t table_id = req->id; + int key_found, status; + struct rte_table_action *a = p->table_data[table_id].a; + + /* Apply actions */ + memset(p->buffer, 0, sizeof(p->buffer)); + data_in = (struct rte_pipeline_table_entry *) p->buffer; + + status = match_convert(match, &match_ll, 1); + if (status) { + rsp->status = -1; + return rsp; + } + + status = action_convert(a, action, data_in); + if (status) { + rsp->status = -1; + return rsp; + } + + status = rte_pipeline_table_entry_add(p->p, + table_id, + &match_ll, + data_in, + &key_found, + &data_out); + if (status) { + rsp->status = -1; + return rsp; + } - t->headroom_ratio = - ((double) t->headroom_cycles) / ((double) time_diff); + /* Write response */ + rsp->status = 0; + rsp->table_rule_add.data = data_out; - t->headroom_cycles = 0; - t->headroom_time = rte_rdtsc_precise(); + return rsp; } -int -app_thread(void *arg) +static struct pipeline_msg_rsp * +pipeline_msg_handle_table_rule_add_default(struct pipeline_data *p, + struct pipeline_msg_req *req) { - struct app_params *app = (struct app_params *) arg; - uint32_t core_id = rte_lcore_id(), i, j; - struct app_thread_data *t = &app->thread_data[core_id]; + struct pipeline_msg_rsp *rsp = (struct pipeline_msg_rsp *) req; + struct table_rule_action *action = &req->table_rule_add_default.action; + struct rte_pipeline_table_entry *data_in, *data_out; + uint32_t table_id = req->id; + int status; - for (i = 0; ; i++) { - uint32_t n_regular = RTE_MIN(t->n_regular, RTE_DIM(t->regular)); - uint32_t n_custom = RTE_MIN(t->n_custom, RTE_DIM(t->custom)); + /* Apply actions */ + memset(p->buffer, 0, sizeof(p->buffer)); + data_in = (struct rte_pipeline_table_entry *) p->buffer; + + data_in->action = action->fwd.action; + if (action->fwd.action == RTE_PIPELINE_ACTION_PORT) + data_in->port_id = action->fwd.id; + if (action->fwd.action == RTE_PIPELINE_ACTION_TABLE) + data_in->table_id = action->fwd.id; + + /* Add default rule to table */ + status = rte_pipeline_table_default_entry_add(p->p, + table_id, + data_in, + &data_out); + if (status) { + rsp->status = -1; + return rsp; + } - /* Run regular pipelines */ - for (j = 0; j < n_regular; j++) { - struct app_thread_pipeline_data *data = &t->regular[j]; - struct pipeline *p = data->be; + /* Write response */ + rsp->status = 0; + rsp->table_rule_add_default.data = data_out; - PIPELINE_RUN_REGULAR(t, p); + return rsp; +} + +static struct pipeline_msg_rsp * +pipeline_msg_handle_table_rule_add_bulk(struct pipeline_data *p, + struct pipeline_msg_req *req) +{ + + struct pipeline_msg_rsp *rsp = (struct pipeline_msg_rsp *) req; + + uint32_t table_id = req->id; + struct table_rule_match *match = req->table_rule_add_bulk.match; + struct table_rule_action *action = req->table_rule_add_bulk.action; + struct rte_pipeline_table_entry **data = + (struct rte_pipeline_table_entry **)req->table_rule_add_bulk.data; + uint32_t n_rules = req->table_rule_add_bulk.n_rules; + uint32_t bulk = req->table_rule_add_bulk.bulk; + + struct rte_table_action *a = p->table_data[table_id].a; + union table_rule_match_low_level *match_ll; + uint8_t *action_ll; + void **match_ll_ptr; + struct rte_pipeline_table_entry **action_ll_ptr; + int *found, status; + uint32_t i; + + /* Memory allocation */ + match_ll = calloc(n_rules, sizeof(union table_rule_match_low_level)); + action_ll = calloc(n_rules, TABLE_RULE_ACTION_SIZE_MAX); + match_ll_ptr = calloc(n_rules, sizeof(void *)); + action_ll_ptr = + calloc(n_rules, sizeof(struct rte_pipeline_table_entry *)); + found = calloc(n_rules, sizeof(int)); + + if ((match_ll == NULL) || + (action_ll == NULL) || + (match_ll_ptr == NULL) || + (action_ll_ptr == NULL) || + (found == NULL)) + goto fail; + + for (i = 0; i < n_rules; i++) { + match_ll_ptr[i] = (void *)&match_ll[i]; + action_ll_ptr[i] = + (struct rte_pipeline_table_entry *)&action_ll[i * TABLE_RULE_ACTION_SIZE_MAX]; + } + + /* Rule match conversion */ + for (i = 0; i < n_rules; i++) { + status = match_convert(&match[i], match_ll_ptr[i], 1); + if (status) + goto fail; + } + + /* Rule action conversion */ + for (i = 0; i < n_rules; i++) { + status = action_convert(a, &action[i], action_ll_ptr[i]); + if (status) + goto fail; + } + + /* Add rule (match, action) to table */ + if (bulk) { + status = rte_pipeline_table_entry_add_bulk(p->p, + table_id, + match_ll_ptr, + action_ll_ptr, + n_rules, + found, + data); + if (status) + n_rules = 0; + } else + for (i = 0; i < n_rules; i++) { + status = rte_pipeline_table_entry_add(p->p, + table_id, + match_ll_ptr[i], + action_ll_ptr[i], + &found[i], + &data[i]); + if (status) { + n_rules = i; + break; + } } - /* Run custom pipelines */ - for (j = 0; j < n_custom; j++) { - struct app_thread_pipeline_data *data = &t->custom[j]; + /* Write response */ + rsp->status = 0; + rsp->table_rule_add_bulk.n_rules = n_rules; + + /* Free */ + free(found); + free(action_ll_ptr); + free(match_ll_ptr); + free(action_ll); + free(match_ll); + + return rsp; + +fail: + free(found); + free(action_ll_ptr); + free(match_ll_ptr); + free(action_ll); + free(match_ll); + + rsp->status = -1; + rsp->table_rule_add_bulk.n_rules = 0; + return rsp; +} + +static struct pipeline_msg_rsp * +pipeline_msg_handle_table_rule_delete(struct pipeline_data *p, + struct pipeline_msg_req *req) +{ + union table_rule_match_low_level match_ll; + struct pipeline_msg_rsp *rsp = (struct pipeline_msg_rsp *) req; + struct table_rule_match *match = &req->table_rule_delete.match; + uint32_t table_id = req->id; + int key_found, status; + + status = match_convert(match, &match_ll, 0); + if (status) { + rsp->status = -1; + return rsp; + } + + rsp->status = rte_pipeline_table_entry_delete(p->p, + table_id, + &match_ll, + &key_found, + NULL); + + return rsp; +} + +static struct pipeline_msg_rsp * +pipeline_msg_handle_table_rule_delete_default(struct pipeline_data *p, + struct pipeline_msg_req *req) +{ + struct pipeline_msg_rsp *rsp = (struct pipeline_msg_rsp *) req; + uint32_t table_id = req->id; + + rsp->status = rte_pipeline_table_default_entry_delete(p->p, + table_id, + NULL); + + return rsp; +} + +static struct pipeline_msg_rsp * +pipeline_msg_handle_table_rule_stats_read(struct pipeline_data *p, + struct pipeline_msg_req *req) +{ + struct pipeline_msg_rsp *rsp = (struct pipeline_msg_rsp *) req; + uint32_t table_id = req->id; + void *data = req->table_rule_stats_read.data; + int clear = req->table_rule_stats_read.clear; + struct rte_table_action *a = p->table_data[table_id].a; + + rsp->status = rte_table_action_stats_read(a, + data, + &rsp->table_rule_stats_read.stats, + clear); + + return rsp; +} + +static struct pipeline_msg_rsp * +pipeline_msg_handle_table_mtr_profile_add(struct pipeline_data *p, + struct pipeline_msg_req *req) +{ + struct pipeline_msg_rsp *rsp = (struct pipeline_msg_rsp *) req; + uint32_t table_id = req->id; + uint32_t meter_profile_id = req->table_mtr_profile_add.meter_profile_id; + struct rte_table_action_meter_profile *profile = + &req->table_mtr_profile_add.profile; + struct rte_table_action *a = p->table_data[table_id].a; + + rsp->status = rte_table_action_meter_profile_add(a, + meter_profile_id, + profile); + + return rsp; +} + +static struct pipeline_msg_rsp * +pipeline_msg_handle_table_mtr_profile_delete(struct pipeline_data *p, + struct pipeline_msg_req *req) +{ + struct pipeline_msg_rsp *rsp = (struct pipeline_msg_rsp *) req; + uint32_t table_id = req->id; + uint32_t meter_profile_id = + req->table_mtr_profile_delete.meter_profile_id; + struct rte_table_action *a = p->table_data[table_id].a; + + rsp->status = rte_table_action_meter_profile_delete(a, + meter_profile_id); + + return rsp; +} + +static struct pipeline_msg_rsp * +pipeline_msg_handle_table_rule_mtr_read(struct pipeline_data *p, + struct pipeline_msg_req *req) +{ + struct pipeline_msg_rsp *rsp = (struct pipeline_msg_rsp *) req; + uint32_t table_id = req->id; + void *data = req->table_rule_mtr_read.data; + uint32_t tc_mask = req->table_rule_mtr_read.tc_mask; + int clear = req->table_rule_mtr_read.clear; + struct rte_table_action *a = p->table_data[table_id].a; + + rsp->status = rte_table_action_meter_read(a, + data, + tc_mask, + &rsp->table_rule_mtr_read.stats, + clear); + + return rsp; +} + +static struct pipeline_msg_rsp * +pipeline_msg_handle_table_dscp_table_update(struct pipeline_data *p, + struct pipeline_msg_req *req) +{ + struct pipeline_msg_rsp *rsp = (struct pipeline_msg_rsp *) req; + uint32_t table_id = req->id; + uint64_t dscp_mask = req->table_dscp_table_update.dscp_mask; + struct rte_table_action_dscp_table *dscp_table = + &req->table_dscp_table_update.dscp_table; + struct rte_table_action *a = p->table_data[table_id].a; + + rsp->status = rte_table_action_dscp_table_update(a, + dscp_mask, + dscp_table); + + return rsp; +} + +static struct pipeline_msg_rsp * +pipeline_msg_handle_table_rule_ttl_read(struct pipeline_data *p, + struct pipeline_msg_req *req) +{ + struct pipeline_msg_rsp *rsp = (struct pipeline_msg_rsp *) req; + uint32_t table_id = req->id; + void *data = req->table_rule_ttl_read.data; + int clear = req->table_rule_ttl_read.clear; + struct rte_table_action *a = p->table_data[table_id].a; + + rsp->status = rte_table_action_ttl_read(a, + data, + &rsp->table_rule_ttl_read.stats, + clear); + + return rsp; +} + +static void +pipeline_msg_handle(struct pipeline_data *p) +{ + for ( ; ; ) { + struct pipeline_msg_req *req; + struct pipeline_msg_rsp *rsp; - PIPELINE_RUN_CUSTOM(t, data); + req = pipeline_msg_recv(p->msgq_req); + if (req == NULL) + break; + + switch (req->type) { + case PIPELINE_REQ_PORT_IN_STATS_READ: + rsp = pipeline_msg_handle_port_in_stats_read(p, req); + break; + + case PIPELINE_REQ_PORT_IN_ENABLE: + rsp = pipeline_msg_handle_port_in_enable(p, req); + break; + + case PIPELINE_REQ_PORT_IN_DISABLE: + rsp = pipeline_msg_handle_port_in_disable(p, req); + break; + + case PIPELINE_REQ_PORT_OUT_STATS_READ: + rsp = pipeline_msg_handle_port_out_stats_read(p, req); + break; + + case PIPELINE_REQ_TABLE_STATS_READ: + rsp = pipeline_msg_handle_table_stats_read(p, req); + break; + + case PIPELINE_REQ_TABLE_RULE_ADD: + rsp = pipeline_msg_handle_table_rule_add(p, req); + break; + + case PIPELINE_REQ_TABLE_RULE_ADD_DEFAULT: + rsp = pipeline_msg_handle_table_rule_add_default(p, req); + break; + + case PIPELINE_REQ_TABLE_RULE_ADD_BULK: + rsp = pipeline_msg_handle_table_rule_add_bulk(p, req); + break; + + case PIPELINE_REQ_TABLE_RULE_DELETE: + rsp = pipeline_msg_handle_table_rule_delete(p, req); + break; + + case PIPELINE_REQ_TABLE_RULE_DELETE_DEFAULT: + rsp = pipeline_msg_handle_table_rule_delete_default(p, req); + break; + + case PIPELINE_REQ_TABLE_RULE_STATS_READ: + rsp = pipeline_msg_handle_table_rule_stats_read(p, req); + break; + + case PIPELINE_REQ_TABLE_MTR_PROFILE_ADD: + rsp = pipeline_msg_handle_table_mtr_profile_add(p, req); + break; + + case PIPELINE_REQ_TABLE_MTR_PROFILE_DELETE: + rsp = pipeline_msg_handle_table_mtr_profile_delete(p, req); + break; + + case PIPELINE_REQ_TABLE_RULE_MTR_READ: + rsp = pipeline_msg_handle_table_rule_mtr_read(p, req); + break; + + case PIPELINE_REQ_TABLE_DSCP_TABLE_UPDATE: + rsp = pipeline_msg_handle_table_dscp_table_update(p, req); + break; + + case PIPELINE_REQ_TABLE_RULE_TTL_READ: + rsp = pipeline_msg_handle_table_rule_ttl_read(p, req); + break; + + default: + rsp = (struct pipeline_msg_rsp *) req; + rsp->status = -1; } - /* Timer */ - if ((i & 0xF) == 0) { - uint64_t time = rte_get_tsc_cycles(); - uint64_t t_deadline = UINT64_MAX; + pipeline_msg_send(p->msgq_rsp, rsp); + } +} - if (time < t->deadline) - continue; +/** + * Data plane threads: main + */ +int +thread_main(void *arg __rte_unused) +{ + struct thread_data *t; + uint32_t thread_id, i; - /* Timer for regular pipelines */ - for (j = 0; j < n_regular; j++) { - struct app_thread_pipeline_data *data = - &t->regular[j]; - uint64_t p_deadline = data->deadline; + thread_id = rte_lcore_id(); + t = &thread_data[thread_id]; - if (p_deadline <= time) { - data->f_timer(data->be); - p_deadline = time + data->timer_period; - data->deadline = p_deadline; - } + /* Dispatch loop */ + for (i = 0; ; i++) { + uint32_t j; - if (p_deadline < t_deadline) - t_deadline = p_deadline; - } + /* Data Plane */ + for (j = 0; j < t->n_pipelines; j++) + rte_pipeline_run(t->p[j]); - /* Timer for custom pipelines */ - for (j = 0; j < n_custom; j++) { - struct app_thread_pipeline_data *data = - &t->custom[j]; - uint64_t p_deadline = data->deadline; + /* Control Plane */ + if ((i & 0xF) == 0) { + uint64_t time = rte_get_tsc_cycles(); + uint64_t time_next_min = UINT64_MAX; + + if (time < t->time_next_min) + continue; + + /* Pipeline message queues */ + for (j = 0; j < t->n_pipelines; j++) { + struct pipeline_data *p = + &t->pipeline_data[j]; + uint64_t time_next = p->time_next; - if (p_deadline <= time) { - data->f_timer(data->be); - p_deadline = time + data->timer_period; - data->deadline = p_deadline; + if (time_next <= time) { + pipeline_msg_handle(p); + rte_pipeline_flush(p->p); + time_next = time + p->timer_period; + p->time_next = time_next; } - if (p_deadline < t_deadline) - t_deadline = p_deadline; + if (time_next < time_next_min) + time_next_min = time_next; } - /* Timer for thread message request */ + /* Thread message queues */ { - uint64_t deadline = t->thread_req_deadline; + uint64_t time_next = t->time_next; - if (deadline <= time) { - thread_msg_req_handle(t); - thread_headroom_update(t, time); - deadline = time + t->timer_period; - t->thread_req_deadline = deadline; + if (time_next <= time) { + thread_msg_handle(t); + time_next = time + t->timer_period; + t->time_next = time_next; } - if (deadline < t_deadline) - t_deadline = deadline; + if (time_next < time_next_min) + time_next_min = time_next; } - - t->deadline = t_deadline; + t->time_next_min = time_next_min; } } |