X-Git-Url: http://git.droids-corp.org/?a=blobdiff_plain;f=app%2Ftest-flow-perf%2Fmain.c;h=99d04634566a46544e0244cd3be3055a89b7cc71;hb=92533e9dfe7501a64b54260420cf00ece7352056;hp=1f693d4ed112daa8f94ca7f728bc6fde7afbd3fd;hpb=ef9ae0cf57821f28c5d271c3d41298e8a8a840f8;p=dpdk.git diff --git a/app/test-flow-perf/main.c b/app/test-flow-perf/main.c index 1f693d4ed1..99d0463456 100644 --- a/app/test-flow-perf/main.c +++ b/app/test-flow-perf/main.c @@ -34,22 +34,28 @@ #include #include #include +#include #include "config.h" #include "flow_gen.h" -#define MAX_ITERATIONS 100 +#define MAX_BATCHES_COUNT 100 #define DEFAULT_RULES_COUNT 4000000 -#define DEFAULT_ITERATION 100000 +#define DEFAULT_RULES_BATCH 100000 +#define DEFAULT_GROUP 0 struct rte_flow *flow; static uint8_t flow_group; +static uint64_t encap_data; +static uint64_t decap_data; + static uint64_t flow_items[MAX_ITEMS_NUM]; static uint64_t flow_actions[MAX_ACTIONS_NUM]; static uint64_t flow_attrs[MAX_ATTRS_NUM]; static uint8_t items_idx, actions_idx, attrs_idx; +static uint64_t ports_mask; static volatile bool force_quit; static bool dump_iterations; static bool delete_flag; @@ -58,8 +64,8 @@ static bool enable_fwd; static struct rte_mempool *mbuf_mp; static uint32_t nb_lcores; -static uint32_t flows_count; -static uint32_t iterations_number; +static uint32_t rules_count; +static uint32_t rules_batch; static uint32_t hairpin_queues_num; /* total hairpin q number - default: 0 */ static uint32_t nb_lcores; @@ -67,7 +73,8 @@ static uint32_t nb_lcores; #define LCORE_MODE_PKT 1 #define LCORE_MODE_STATS 2 #define MAX_STREAMS 64 -#define MAX_LCORES 64 +#define METER_CREATE 1 +#define METER_DELETE 2 struct stream { int tx_port; @@ -87,15 +94,35 @@ struct lcore_info { struct rte_mbuf *pkts[MAX_PKT_BURST]; } __rte_cache_aligned; -static struct lcore_info lcore_infos[MAX_LCORES]; +static struct lcore_info lcore_infos[RTE_MAX_LCORE]; + +struct used_cpu_time { + double insertion[MAX_PORTS][RTE_MAX_LCORE]; + double deletion[MAX_PORTS][RTE_MAX_LCORE]; +}; + +struct multi_cores_pool { + uint32_t cores_count; + uint32_t rules_count; + struct used_cpu_time create_meter; + struct used_cpu_time create_flow; + int64_t last_alloc[RTE_MAX_LCORE]; + int64_t current_alloc[RTE_MAX_LCORE]; +} __rte_cache_aligned; + +static struct multi_cores_pool mc_pool = { + .cores_count = 1, +}; static void usage(char *progname) { printf("\nusage: %s\n", progname); printf("\nControl configurations:\n"); - printf(" --flows-count=N: to set the number of needed" - " flows to insert, default is 4,000,000\n"); + printf(" --rules-count=N: to set the number of needed" + " rules to insert, default is %d\n", DEFAULT_RULES_COUNT); + printf(" --rules-batch=N: set number of batched rules," + " default is %d\n", DEFAULT_RULES_BATCH); printf(" --dump-iterations: To print rates for each" " iteration\n"); printf(" --deletion-rate: Enable deletion rate" @@ -103,13 +130,16 @@ usage(char *progname) printf(" --dump-socket-mem: To dump all socket memory\n"); printf(" --enable-fwd: To enable packets forwarding" " after insertion\n"); + printf(" --portmask=N: hexadecimal bitmask of ports used\n"); printf("To set flow attributes:\n"); printf(" --ingress: set ingress attribute in flows\n"); printf(" --egress: set egress attribute in flows\n"); printf(" --transfer: set transfer attribute in flows\n"); printf(" --group=N: set group for all flows," - " default is 0\n"); + " default is %d\n", DEFAULT_GROUP); + printf(" --cores=N: to set the number of needed " + "cores to insert rte_flow rules, default is 1\n"); printf("To set flow items:\n"); printf(" --ether: add ether layer in flow items\n"); @@ -125,6 +155,8 @@ usage(char *progname) printf(" --gtp: add gtp layer in flow items\n"); printf(" --meta: add meta layer in flow items\n"); printf(" --tag: add tag layer in flow items\n"); + printf(" --icmpv4: add icmpv4 layer in flow items\n"); + printf(" --icmpv6: add icmpv6 layer in flow items\n"); printf("To set flow actions:\n"); printf(" --port-id: add port-id action in flow actions\n"); @@ -170,12 +202,27 @@ usage(char *progname) "ipv4 dscp value to be set is random each flow\n"); printf(" --set-ipv6-dscp: add set ipv6 dscp action to flow actions\n" "ipv6 dscp value to be set is random each flow\n"); + printf(" --flag: add flag action to flow actions\n"); + printf(" --meter: add meter action to flow actions\n"); + printf(" --raw-encap=: add raw encap action to flow actions\n" + "Data is the data needed to be encaped\n" + "Example: raw-encap=ether,ipv4,udp,vxlan\n"); + printf(" --raw-decap=: add raw decap action to flow actions\n" + "Data is the data needed to be decaped\n" + "Example: raw-decap=ether,ipv4,udp,vxlan\n"); + printf(" --vxlan-encap: add vxlan-encap action to flow actions\n" + "Encapped data is fixed with pattern: ether,ipv4,udp,vxlan\n" + "With fixed values\n"); + printf(" --vxlan-decap: add vxlan_decap action to flow actions\n"); } static void args_parse(int argc, char **argv) { + uint64_t pm; char **argvopt; + char *token; + char *end; int n, opt; int opt_idx; size_t i; @@ -265,6 +312,18 @@ args_parse(int argc, char **argv) .map = &flow_items[0], .map_idx = &items_idx }, + { + .str = "icmpv4", + .mask = FLOW_ITEM_MASK(RTE_FLOW_ITEM_TYPE_ICMP), + .map = &flow_items[0], + .map_idx = &items_idx + }, + { + .str = "icmpv6", + .mask = FLOW_ITEM_MASK(RTE_FLOW_ITEM_TYPE_ICMP6), + .map = &flow_items[0], + .map_idx = &items_idx + }, { .str = "ingress", .mask = INGRESS, @@ -465,16 +524,51 @@ args_parse(int argc, char **argv) .map = &flow_actions[0], .map_idx = &actions_idx }, + { + .str = "flag", + .mask = FLOW_ACTION_MASK( + RTE_FLOW_ACTION_TYPE_FLAG + ), + .map = &flow_actions[0], + .map_idx = &actions_idx + }, + { + .str = "meter", + .mask = FLOW_ACTION_MASK( + RTE_FLOW_ACTION_TYPE_METER + ), + .map = &flow_actions[0], + .map_idx = &actions_idx + }, + { + .str = "vxlan-encap", + .mask = FLOW_ACTION_MASK( + RTE_FLOW_ACTION_TYPE_VXLAN_ENCAP + ), + .map = &flow_actions[0], + .map_idx = &actions_idx + }, + { + .str = "vxlan-decap", + .mask = FLOW_ACTION_MASK( + RTE_FLOW_ACTION_TYPE_VXLAN_DECAP + ), + .map = &flow_actions[0], + .map_idx = &actions_idx + }, }; static const struct option lgopts[] = { /* Control */ { "help", 0, 0, 0 }, - { "flows-count", 1, 0, 0 }, + { "rules-count", 1, 0, 0 }, + { "rules-batch", 1, 0, 0 }, { "dump-iterations", 0, 0, 0 }, { "deletion-rate", 0, 0, 0 }, { "dump-socket-mem", 0, 0, 0 }, { "enable-fwd", 0, 0, 0 }, + { "portmask", 1, 0, 0 }, + { "cores", 1, 0, 0 }, /* Attributes */ { "ingress", 0, 0, 0 }, { "egress", 0, 0, 0 }, @@ -494,6 +588,8 @@ args_parse(int argc, char **argv) { "gtp", 0, 0, 0 }, { "meta", 0, 0, 0 }, { "tag", 0, 0, 0 }, + { "icmpv4", 0, 0, 0 }, + { "icmpv6", 0, 0, 0 }, /* Actions */ { "port-id", 0, 0, 0 }, { "rss", 0, 0, 0 }, @@ -522,8 +618,17 @@ args_parse(int argc, char **argv) { "dec-ttl", 0, 0, 0 }, { "set-ipv4-dscp", 0, 0, 0 }, { "set-ipv6-dscp", 0, 0, 0 }, + { "flag", 0, 0, 0 }, + { "meter", 0, 0, 0 }, + { "raw-encap", 1, 0, 0 }, + { "raw-decap", 1, 0, 0 }, + { "vxlan-encap", 0, 0, 0 }, + { "vxlan-decap", 0, 0, 0 }, }; + RTE_ETH_FOREACH_DEV(i) + ports_mask |= 1 << i; + hairpin_queues_num = 0; argvopt = argv; @@ -583,18 +688,80 @@ args_parse(int argc, char **argv) printf("hairpin-queue / "); } + if (strcmp(lgopts[opt_idx].name, "raw-encap") == 0) { + printf("raw-encap "); + flow_actions[actions_idx++] = + FLOW_ITEM_MASK( + RTE_FLOW_ACTION_TYPE_RAW_ENCAP + ); + + token = strtok(optarg, ","); + while (token != NULL) { + for (i = 0; i < RTE_DIM(flow_options); i++) { + if (strcmp(flow_options[i].str, token) == 0) { + printf("%s,", token); + encap_data |= flow_options[i].mask; + break; + } + /* Reached last item with no match */ + if (i == (RTE_DIM(flow_options) - 1)) { + fprintf(stderr, "Invalid encap item: %s\n", token); + usage(argv[0]); + rte_exit(EXIT_SUCCESS, "Invalid encap item\n"); + } + } + token = strtok(NULL, ","); + } + printf(" / "); + } + if (strcmp(lgopts[opt_idx].name, "raw-decap") == 0) { + printf("raw-decap "); + flow_actions[actions_idx++] = + FLOW_ITEM_MASK( + RTE_FLOW_ACTION_TYPE_RAW_DECAP + ); + + token = strtok(optarg, ","); + while (token != NULL) { + for (i = 0; i < RTE_DIM(flow_options); i++) { + if (strcmp(flow_options[i].str, token) == 0) { + printf("%s,", token); + encap_data |= flow_options[i].mask; + break; + } + /* Reached last item with no match */ + if (i == (RTE_DIM(flow_options) - 1)) { + fprintf(stderr, "Invalid decap item: %s\n", token); + usage(argv[0]); + rte_exit(EXIT_SUCCESS, "Invalid decap item\n"); + } + } + token = strtok(NULL, ","); + } + printf(" / "); + } /* Control */ if (strcmp(lgopts[opt_idx].name, - "flows-count") == 0) { + "rules-batch") == 0) { n = atoi(optarg); - if (n > (int) iterations_number) - flows_count = n; + if (n >= DEFAULT_RULES_BATCH) + rules_batch = n; else { - printf("\n\nflows_count should be > %d\n", - iterations_number); + printf("\n\nrules_batch should be >= %d\n", + DEFAULT_RULES_BATCH); rte_exit(EXIT_SUCCESS, " "); } } + if (strcmp(lgopts[opt_idx].name, + "rules-count") == 0) { + n = atoi(optarg); + if (n >= (int) rules_batch) + rules_count = n; + else { + printf("\n\nrules_count should be >= %d\n", + rules_batch); + } + } if (strcmp(lgopts[opt_idx].name, "dump-iterations") == 0) dump_iterations = true; @@ -607,6 +774,30 @@ args_parse(int argc, char **argv) if (strcmp(lgopts[opt_idx].name, "enable-fwd") == 0) enable_fwd = true; + if (strcmp(lgopts[opt_idx].name, + "portmask") == 0) { + /* parse hexadecimal string */ + end = NULL; + pm = strtoull(optarg, &end, 16); + if ((optarg[0] == '\0') || (end == NULL) || (*end != '\0')) + rte_exit(EXIT_FAILURE, "Invalid fwd port mask\n"); + ports_mask = pm; + } + if (strcmp(lgopts[opt_idx].name, "cores") == 0) { + n = atoi(optarg); + if ((int) rte_lcore_count() <= n) { + printf("\nError: you need %d cores to run on multi-cores\n" + "Existing cores are: %d\n", n, rte_lcore_count()); + rte_exit(EXIT_FAILURE, " "); + } + if (n <= RTE_MAX_LCORE && n > 0) + mc_pool.cores_count = n; + else { + printf("Error: cores count must be > 0 " + " and < %d\n", RTE_MAX_LCORE); + rte_exit(EXIT_FAILURE, " "); + } + } break; default: fprintf(stderr, "Invalid option: %s\n", argv[optind]); @@ -683,184 +874,571 @@ print_flow_error(struct rte_flow_error error) } static inline void -destroy_flows(int port_id, struct rte_flow **flow_list) +print_rules_batches(double *cpu_time_per_batch) +{ + uint8_t idx; + double delta; + double rate; + + for (idx = 0; idx < MAX_BATCHES_COUNT; idx++) { + if (!cpu_time_per_batch[idx]) + break; + delta = (double)(rules_batch / cpu_time_per_batch[idx]); + rate = delta / 1000; /* Save rate in K unit. */ + printf(":: Rules batch #%d: %d rules " + "in %f sec[ Rate = %f K Rule/Sec ]\n", + idx, rules_batch, + cpu_time_per_batch[idx], rate); + } +} + + +static inline int +has_meter(void) +{ + int i; + + for (i = 0; i < MAX_ACTIONS_NUM; i++) { + if (flow_actions[i] == 0) + break; + if (flow_actions[i] + & FLOW_ACTION_MASK(RTE_FLOW_ACTION_TYPE_METER)) + return 1; + } + return 0; +} + +static void +create_meter_rule(int port_id, uint32_t counter) +{ + int ret; + struct rte_mtr_params params; + uint32_t default_prof_id = 100; + struct rte_mtr_error error; + + memset(¶ms, 0, sizeof(struct rte_mtr_params)); + params.meter_enable = 1; + params.stats_mask = 0xffff; + params.use_prev_mtr_color = 0; + params.dscp_table = NULL; + + /*create meter*/ + params.meter_profile_id = default_prof_id; + params.action[RTE_COLOR_GREEN] = + MTR_POLICER_ACTION_COLOR_GREEN; + params.action[RTE_COLOR_YELLOW] = + MTR_POLICER_ACTION_COLOR_YELLOW; + params.action[RTE_COLOR_RED] = + MTR_POLICER_ACTION_DROP; + + ret = rte_mtr_create(port_id, counter, ¶ms, 1, &error); + if (ret != 0) { + printf("Port %u create meter idx(%d) error(%d) message: %s\n", + port_id, counter, error.type, + error.message ? error.message : "(no stated reason)"); + rte_exit(EXIT_FAILURE, "error in creating meter"); + } +} + +static void +destroy_meter_rule(int port_id, uint32_t counter) +{ + struct rte_mtr_error error; + + if (rte_mtr_destroy(port_id, counter, &error)) { + printf("Port %u destroy meter(%d) error(%d) message: %s\n", + port_id, counter, error.type, + error.message ? error.message : "(no stated reason)"); + rte_exit(EXIT_FAILURE, "Error in deleting meter rule"); + } +} + +static void +meters_handler(int port_id, uint8_t core_id, uint8_t ops) +{ + uint64_t start_batch; + double cpu_time_used, insertion_rate; + int rules_count_per_core, rules_batch_idx; + uint32_t counter, start_counter = 0, end_counter; + double cpu_time_per_batch[MAX_BATCHES_COUNT] = { 0 }; + + rules_count_per_core = rules_count / mc_pool.cores_count; + + if (core_id) + start_counter = core_id * rules_count_per_core; + end_counter = (core_id + 1) * rules_count_per_core; + + cpu_time_used = 0; + start_batch = rte_rdtsc(); + for (counter = start_counter; counter < end_counter; counter++) { + if (ops == METER_CREATE) + create_meter_rule(port_id, counter); + else + destroy_meter_rule(port_id, counter); + /* + * Save the insertion rate for rules batch. + * Check if the insertion reached the rules + * patch counter, then save the insertion rate + * for this batch. + */ + if (!((counter + 1) % rules_batch)) { + rules_batch_idx = ((counter + 1) / rules_batch) - 1; + cpu_time_per_batch[rules_batch_idx] = + ((double)(rte_rdtsc() - start_batch)) + / rte_get_tsc_hz(); + cpu_time_used += cpu_time_per_batch[rules_batch_idx]; + start_batch = rte_rdtsc(); + } + } + + /* Print insertion rates for all batches */ + if (dump_iterations) + print_rules_batches(cpu_time_per_batch); + + insertion_rate = + ((double) (rules_count_per_core / cpu_time_used) / 1000); + + /* Insertion rate for all rules in one core */ + printf(":: Port %d :: Core %d Meter %s :: start @[%d] - end @[%d]," + " use:%.02fs, rate:%.02fk Rule/Sec\n", + port_id, core_id, ops == METER_CREATE ? "create" : "delete", + start_counter, end_counter - 1, + cpu_time_used, insertion_rate); + + if (ops == METER_CREATE) + mc_pool.create_meter.insertion[port_id][core_id] + = cpu_time_used; + else + mc_pool.create_meter.deletion[port_id][core_id] + = cpu_time_used; +} + +static void +destroy_meter_profile(void) +{ + struct rte_mtr_error error; + uint16_t nr_ports; + int port_id; + + nr_ports = rte_eth_dev_count_avail(); + for (port_id = 0; port_id < nr_ports; port_id++) { + /* If port outside portmask */ + if (!((ports_mask >> port_id) & 0x1)) + continue; + + if (rte_mtr_meter_profile_delete + (port_id, DEFAULT_METER_PROF_ID, &error)) { + printf("Port %u del profile error(%d) message: %s\n", + port_id, error.type, + error.message ? error.message : "(no stated reason)"); + rte_exit(EXIT_FAILURE, "Error: Destroy meter profile Failed!\n"); + } + } +} + +static void +create_meter_profile(void) +{ + uint16_t nr_ports; + int ret, port_id; + struct rte_mtr_meter_profile mp; + struct rte_mtr_error error; + + /* + *currently , only create one meter file for one port + *1 meter profile -> N meter rules -> N rte flows + */ + memset(&mp, 0, sizeof(struct rte_mtr_meter_profile)); + nr_ports = rte_eth_dev_count_avail(); + for (port_id = 0; port_id < nr_ports; port_id++) { + /* If port outside portmask */ + if (!((ports_mask >> port_id) & 0x1)) + continue; + + mp.alg = RTE_MTR_SRTCM_RFC2697; + mp.srtcm_rfc2697.cir = METER_CIR; + mp.srtcm_rfc2697.cbs = METER_CIR / 8; + mp.srtcm_rfc2697.ebs = 0; + + ret = rte_mtr_meter_profile_add + (port_id, DEFAULT_METER_PROF_ID, &mp, &error); + if (ret != 0) { + printf("Port %u create Profile error(%d) message: %s\n", + port_id, error.type, + error.message ? error.message : "(no stated reason)"); + rte_exit(EXIT_FAILURE, "Error: Creation meter profile Failed!\n"); + } + } +} + +static inline void +destroy_flows(int port_id, uint8_t core_id, struct rte_flow **flows_list) { struct rte_flow_error error; - clock_t start_iter, end_iter; + clock_t start_batch, end_batch; double cpu_time_used = 0; - double flows_rate; - double cpu_time_per_iter[MAX_ITERATIONS]; + double deletion_rate; + double cpu_time_per_batch[MAX_BATCHES_COUNT] = { 0 }; double delta; uint32_t i; - int iter_id; + int rules_batch_idx; + int rules_count_per_core; - for (i = 0; i < MAX_ITERATIONS; i++) - cpu_time_per_iter[i] = -1; + rules_count_per_core = rules_count / mc_pool.cores_count; + /* If group > 0 , should add 1 flow which created in group 0 */ + if (flow_group > 0 && core_id == 0) + rules_count_per_core++; - if (iterations_number > flows_count) - iterations_number = flows_count; - - /* Deletion Rate */ - printf("Flows Deletion on port = %d\n", port_id); - start_iter = clock(); - for (i = 0; i < flows_count; i++) { - if (flow_list[i] == 0) + start_batch = rte_rdtsc(); + for (i = 0; i < (uint32_t) rules_count_per_core; i++) { + if (flows_list[i] == 0) break; memset(&error, 0x33, sizeof(error)); - if (rte_flow_destroy(port_id, flow_list[i], &error)) { + if (rte_flow_destroy(port_id, flows_list[i], &error)) { print_flow_error(error); rte_exit(EXIT_FAILURE, "Error in deleting flow"); } - if (i && !((i + 1) % iterations_number)) { - /* Save the deletion rate of each iter */ - end_iter = clock(); - delta = (double) (end_iter - start_iter); - iter_id = ((i + 1) / iterations_number) - 1; - cpu_time_per_iter[iter_id] = - delta / CLOCKS_PER_SEC; - cpu_time_used += cpu_time_per_iter[iter_id]; - start_iter = clock(); + /* + * Save the deletion rate for rules batch. + * Check if the deletion reached the rules + * patch counter, then save the deletion rate + * for this batch. + */ + if (!((i + 1) % rules_batch)) { + end_batch = rte_rdtsc(); + delta = (double) (end_batch - start_batch); + rules_batch_idx = ((i + 1) / rules_batch) - 1; + cpu_time_per_batch[rules_batch_idx] = delta / rte_get_tsc_hz(); + cpu_time_used += cpu_time_per_batch[rules_batch_idx]; + start_batch = rte_rdtsc(); } } - /* Deletion rate per iteration */ + /* Print deletion rates for all batches */ if (dump_iterations) - for (i = 0; i < MAX_ITERATIONS; i++) { - if (cpu_time_per_iter[i] == -1) - continue; - delta = (double)(iterations_number / - cpu_time_per_iter[i]); - flows_rate = delta / 1000; - printf(":: Iteration #%d: %d flows " - "in %f sec[ Rate = %f K/Sec ]\n", - i, iterations_number, - cpu_time_per_iter[i], flows_rate); - } + print_rules_batches(cpu_time_per_batch); - /* Deletion rate for all flows */ - flows_rate = ((double) (flows_count / cpu_time_used) / 1000); - printf("\n:: Total flow deletion rate -> %f K/Sec\n", - flows_rate); - printf(":: The time for deleting %d in flows %f seconds\n", - flows_count, cpu_time_used); + /* Deletion rate for all rules */ + deletion_rate = ((double) (rules_count_per_core / cpu_time_used) / 1000); + printf(":: Port %d :: Core %d :: Rules deletion rate -> %f K Rule/Sec\n", + port_id, core_id, deletion_rate); + printf(":: Port %d :: Core %d :: The time for deleting %d rules is %f seconds\n", + port_id, core_id, rules_count_per_core, cpu_time_used); + + mc_pool.create_flow.deletion[port_id][core_id] = cpu_time_used; } -static inline void -flows_handler(void) +static struct rte_flow ** +insert_flows(int port_id, uint8_t core_id) { - struct rte_flow **flow_list; + struct rte_flow **flows_list; struct rte_flow_error error; - clock_t start_iter, end_iter; + clock_t start_batch, end_batch; double cpu_time_used; - double flows_rate; - double cpu_time_per_iter[MAX_ITERATIONS]; + double insertion_rate; + double cpu_time_per_batch[MAX_BATCHES_COUNT] = { 0 }; double delta; - uint16_t nr_ports; - uint32_t i; - int port_id; - int iter_id; uint32_t flow_index; + uint32_t counter, start_counter = 0, end_counter; uint64_t global_items[MAX_ITEMS_NUM] = { 0 }; uint64_t global_actions[MAX_ACTIONS_NUM] = { 0 }; + int rules_batch_idx; + int rules_count_per_core; + + rules_count_per_core = rules_count / mc_pool.cores_count; + + /* Set boundaries of rules for each core. */ + if (core_id) + start_counter = core_id * rules_count_per_core; + end_counter = (core_id + 1) * rules_count_per_core; global_items[0] = FLOW_ITEM_MASK(RTE_FLOW_ITEM_TYPE_ETH); global_actions[0] = FLOW_ITEM_MASK(RTE_FLOW_ACTION_TYPE_JUMP); - nr_ports = rte_eth_dev_count_avail(); + flows_list = rte_zmalloc("flows_list", + (sizeof(struct rte_flow *) * rules_count_per_core) + 1, 0); + if (flows_list == NULL) + rte_exit(EXIT_FAILURE, "No Memory available!"); - for (i = 0; i < MAX_ITERATIONS; i++) - cpu_time_per_iter[i] = -1; + cpu_time_used = 0; + flow_index = 0; + if (flow_group > 0 && core_id == 0) { + /* + * Create global rule to jump into flow_group, + * this way the app will avoid the default rules. + * + * This rule will be created only once. + * + * Global rule: + * group 0 eth / end actions jump group + */ + flow = generate_flow(port_id, 0, flow_attrs, + global_items, global_actions, + flow_group, 0, 0, 0, 0, core_id, &error); + + if (flow == NULL) { + print_flow_error(error); + rte_exit(EXIT_FAILURE, "error in creating flow"); + } + flows_list[flow_index++] = flow; + } - if (iterations_number > flows_count) - iterations_number = flows_count; + start_batch = rte_rdtsc(); + for (counter = start_counter; counter < end_counter; counter++) { + flow = generate_flow(port_id, flow_group, + flow_attrs, flow_items, flow_actions, + JUMP_ACTION_TABLE, counter, + hairpin_queues_num, + encap_data, decap_data, + core_id, &error); - printf(":: Flows Count per port: %d\n", flows_count); + if (force_quit) + counter = end_counter; - flow_list = rte_zmalloc("flow_list", - (sizeof(struct rte_flow *) * flows_count) + 1, 0); - if (flow_list == NULL) - rte_exit(EXIT_FAILURE, "No Memory available!"); + if (!flow) { + print_flow_error(error); + rte_exit(EXIT_FAILURE, "error in creating flow"); + } + + flows_list[flow_index++] = flow; + + /* + * Save the insertion rate for rules batch. + * Check if the insertion reached the rules + * patch counter, then save the insertion rate + * for this batch. + */ + if (!((counter + 1) % rules_batch)) { + end_batch = rte_rdtsc(); + delta = (double) (end_batch - start_batch); + rules_batch_idx = ((counter + 1) / rules_batch) - 1; + cpu_time_per_batch[rules_batch_idx] = delta / rte_get_tsc_hz(); + cpu_time_used += cpu_time_per_batch[rules_batch_idx]; + start_batch = rte_rdtsc(); + } + } + + /* Print insertion rates for all batches */ + if (dump_iterations) + print_rules_batches(cpu_time_per_batch); + + printf(":: Port %d :: Core %d boundaries :: start @[%d] - end @[%d]\n", + port_id, core_id, start_counter, end_counter - 1); + + /* Insertion rate for all rules in one core */ + insertion_rate = ((double) (rules_count_per_core / cpu_time_used) / 1000); + printf(":: Port %d :: Core %d :: Rules insertion rate -> %f K Rule/Sec\n", + port_id, core_id, insertion_rate); + printf(":: Port %d :: Core %d :: The time for creating %d in rules %f seconds\n", + port_id, core_id, rules_count_per_core, cpu_time_used); + + mc_pool.create_flow.insertion[port_id][core_id] = cpu_time_used; + return flows_list; +} + +static void +flows_handler(uint8_t core_id) +{ + struct rte_flow **flows_list; + uint16_t nr_ports; + int port_id; + + nr_ports = rte_eth_dev_count_avail(); + + if (rules_batch > rules_count) + rules_batch = rules_count; + + printf(":: Rules Count per port: %d\n\n", rules_count); for (port_id = 0; port_id < nr_ports; port_id++) { - cpu_time_used = 0; - flow_index = 0; - if (flow_group > 0) { - /* - * Create global rule to jump into flow_group, - * this way the app will avoid the default rules. - * - * Global rule: - * group 0 eth / end actions jump group - * - */ - flow = generate_flow(port_id, 0, flow_attrs, - global_items, global_actions, - flow_group, 0, 0, &error); + /* If port outside portmask */ + if (!((ports_mask >> port_id) & 0x1)) + continue; - if (flow == NULL) { - print_flow_error(error); - rte_exit(EXIT_FAILURE, "error in creating flow"); - } - flow_list[flow_index++] = flow; + /* Insertion part. */ + mc_pool.last_alloc[core_id] = (int64_t)dump_socket_mem(stdout); + if (has_meter()) + meters_handler(port_id, core_id, METER_CREATE); + flows_list = insert_flows(port_id, core_id); + if (flows_list == NULL) + rte_exit(EXIT_FAILURE, "Error: Insertion Failed!\n"); + mc_pool.current_alloc[core_id] = (int64_t)dump_socket_mem(stdout); + + /* Deletion part. */ + if (delete_flag) { + destroy_flows(port_id, core_id, flows_list); + if (has_meter()) + meters_handler(port_id, core_id, METER_DELETE); } + } +} - /* Insertion Rate */ - printf("Flows insertion on port = %d\n", port_id); - start_iter = clock(); - for (i = 0; i < flows_count; i++) { - flow = generate_flow(port_id, flow_group, - flow_attrs, flow_items, flow_actions, - JUMP_ACTION_TABLE, i, - hairpin_queues_num, &error); - - if (force_quit) - i = flows_count; - - if (!flow) { - print_flow_error(error); - rte_exit(EXIT_FAILURE, "error in creating flow"); - } +static void +dump_used_cpu_time(const char *item, + uint16_t port, struct used_cpu_time *used_time) +{ + uint32_t i; + /* Latency: total count of rte rules divided + * over max time used by thread between all + * threads time. + * + * Throughput: total count of rte rules divided + * over the average of the time cosumed by all + * threads time. + */ + double insertion_latency_time; + double insertion_throughput_time; + double deletion_latency_time; + double deletion_throughput_time; + double insertion_latency, insertion_throughput; + double deletion_latency, deletion_throughput; + + /* Save first insertion/deletion rates from first thread. + * Start comparing with all threads, if any thread used + * time more than current saved, replace it. + * + * Thus in the end we will have the max time used for + * insertion/deletion by one thread. + * + * As for memory consumption, save the min of all threads + * of last alloc, and save the max for all threads for + * current alloc. + */ - flow_list[flow_index++] = flow; - - if (i && !((i + 1) % iterations_number)) { - /* Save the insertion rate of each iter */ - end_iter = clock(); - delta = (double) (end_iter - start_iter); - iter_id = ((i + 1) / iterations_number) - 1; - cpu_time_per_iter[iter_id] = - delta / CLOCKS_PER_SEC; - cpu_time_used += cpu_time_per_iter[iter_id]; - start_iter = clock(); - } + insertion_latency_time = used_time->insertion[port][0]; + deletion_latency_time = used_time->deletion[port][0]; + insertion_throughput_time = used_time->insertion[port][0]; + deletion_throughput_time = used_time->deletion[port][0]; + + i = mc_pool.cores_count; + while (i-- > 1) { + insertion_throughput_time += used_time->insertion[port][i]; + deletion_throughput_time += used_time->deletion[port][i]; + if (insertion_latency_time < used_time->insertion[port][i]) + insertion_latency_time = used_time->insertion[port][i]; + if (deletion_latency_time < used_time->deletion[port][i]) + deletion_latency_time = used_time->deletion[port][i]; + } + + insertion_latency = ((double) (mc_pool.rules_count + / insertion_latency_time) / 1000); + deletion_latency = ((double) (mc_pool.rules_count + / deletion_latency_time) / 1000); + + insertion_throughput_time /= mc_pool.cores_count; + deletion_throughput_time /= mc_pool.cores_count; + insertion_throughput = ((double) (mc_pool.rules_count + / insertion_throughput_time) / 1000); + deletion_throughput = ((double) (mc_pool.rules_count + / deletion_throughput_time) / 1000); + + /* Latency stats */ + printf("\n%s\n:: [Latency | Insertion] All Cores :: Port %d :: ", + item, port); + printf("Total flows insertion rate -> %f K Rules/Sec\n", + insertion_latency); + printf(":: [Latency | Insertion] All Cores :: Port %d :: ", port); + printf("The time for creating %d rules is %f seconds\n", + mc_pool.rules_count, insertion_latency_time); + + /* Throughput stats */ + printf(":: [Throughput | Insertion] All Cores :: Port %d :: ", port); + printf("Total flows insertion rate -> %f K Rules/Sec\n", + insertion_throughput); + printf(":: [Throughput | Insertion] All Cores :: Port %d :: ", port); + printf("The average time for creating %d rules is %f seconds\n", + mc_pool.rules_count, insertion_throughput_time); + + if (delete_flag) { + /* Latency stats */ + printf(":: [Latency | Deletion] All Cores :: Port %d :: Total " + "deletion rate -> %f K Rules/Sec\n", + port, deletion_latency); + printf(":: [Latency | Deletion] All Cores :: Port %d :: ", + port); + printf("The time for deleting %d rules is %f seconds\n", + mc_pool.rules_count, deletion_latency_time); + + /* Throughput stats */ + printf(":: [Throughput | Deletion] All Cores :: Port %d :: Total " + "deletion rate -> %f K Rules/Sec\n", + port, deletion_throughput); + printf(":: [Throughput | Deletion] All Cores :: Port %d :: ", + port); + printf("The average time for deleting %d rules is %f seconds\n", + mc_pool.rules_count, deletion_throughput_time); + } +} + +static void +dump_used_mem(uint16_t port) +{ + uint32_t i; + int64_t last_alloc, current_alloc; + int flow_size_in_bytes; + + last_alloc = mc_pool.last_alloc[0]; + current_alloc = mc_pool.current_alloc[0]; + + i = mc_pool.cores_count; + while (i-- > 1) { + if (last_alloc > mc_pool.last_alloc[i]) + last_alloc = mc_pool.last_alloc[i]; + if (current_alloc < mc_pool.current_alloc[i]) + current_alloc = mc_pool.current_alloc[i]; + } + + flow_size_in_bytes = (current_alloc - last_alloc) / mc_pool.rules_count; + printf("\n:: Port %d :: rte_flow size in DPDK layer: %d Bytes\n", + port, flow_size_in_bytes); +} + +static int +run_rte_flow_handler_cores(void *data __rte_unused) +{ + uint16_t port; + int lcore_counter = 0; + int lcore_id = rte_lcore_id(); + int i; + + RTE_LCORE_FOREACH(i) { + /* If core not needed return. */ + if (lcore_id == i) { + printf(":: lcore %d mapped with index %d\n", lcore_id, lcore_counter); + if (lcore_counter >= (int) mc_pool.cores_count) + return 0; + break; } + lcore_counter++; + } + lcore_id = lcore_counter; - /* Iteration rate per iteration */ - if (dump_iterations) - for (i = 0; i < MAX_ITERATIONS; i++) { - if (cpu_time_per_iter[i] == -1) - continue; - delta = (double)(iterations_number / - cpu_time_per_iter[i]); - flows_rate = delta / 1000; - printf(":: Iteration #%d: %d flows " - "in %f sec[ Rate = %f K/Sec ]\n", - i, iterations_number, - cpu_time_per_iter[i], flows_rate); - } + if (lcore_id >= (int) mc_pool.cores_count) + return 0; + + mc_pool.rules_count = rules_count; + + flows_handler(lcore_id); - /* Insertion rate for all flows */ - flows_rate = ((double) (flows_count / cpu_time_used) / 1000); - printf("\n:: Total flow insertion rate -> %f K/Sec\n", - flows_rate); - printf(":: The time for creating %d in flows %f seconds\n", - flows_count, cpu_time_used); + /* Only main core to print total results. */ + if (lcore_id != 0) + return 0; + + /* Make sure all cores finished insertion/deletion process. */ + rte_eal_mp_wait_lcore(); - if (delete_flag) - destroy_flows(port_id, flow_list); + RTE_ETH_FOREACH_DEV(port) { + if (has_meter()) + dump_used_cpu_time("Meters:", + port, &mc_pool.create_meter); + dump_used_cpu_time("Flows:", + port, &mc_pool.create_flow); + dump_used_mem(port); } + + return 0; } static void @@ -937,12 +1515,12 @@ packet_per_second_stats(void) int i; old = rte_zmalloc("old", - sizeof(struct lcore_info) * MAX_LCORES, 0); + sizeof(struct lcore_info) * RTE_MAX_LCORE, 0); if (old == NULL) rte_exit(EXIT_FAILURE, "No Memory available!"); memcpy(old, lcore_infos, - sizeof(struct lcore_info) * MAX_LCORES); + sizeof(struct lcore_info) * RTE_MAX_LCORE); while (!force_quit) { uint64_t total_tx_pkts = 0; @@ -965,7 +1543,7 @@ packet_per_second_stats(void) printf("%6s %16s %16s %16s\n", "------", "----------------", "----------------", "----------------"); nr_lines = 3; - for (i = 0; i < MAX_LCORES; i++) { + for (i = 0; i < RTE_MAX_LCORE; i++) { li = &lcore_infos[i]; oli = &old[i]; if (li->mode != LCORE_MODE_PKT) @@ -996,7 +1574,7 @@ packet_per_second_stats(void) } memcpy(old, lcore_infos, - sizeof(struct lcore_info) * MAX_LCORES); + sizeof(struct lcore_info) * RTE_MAX_LCORE); } } @@ -1057,7 +1635,7 @@ init_lcore_info(void) * This means that this stream is not used, or not set * yet. */ - for (i = 0; i < MAX_LCORES; i++) + for (i = 0; i < RTE_MAX_LCORE; i++) for (j = 0; j < MAX_STREAMS; j++) { lcore_infos[i].streams[j].tx_port = -1; lcore_infos[i].streams[j].rx_port = -1; @@ -1119,7 +1697,7 @@ init_lcore_info(void) /* Print all streams */ printf(":: Stream -> core id[N]: (rx_port, rx_queue)->(tx_port, tx_queue)\n"); - for (i = 0; i < MAX_LCORES; i++) + for (i = 0; i < RTE_MAX_LCORE; i++) for (j = 0; j < MAX_STREAMS; j++) { /* No streams for this core */ if (lcore_infos[i].streams[j].tx_port == -1) @@ -1273,7 +1851,6 @@ main(int argc, char **argv) int ret; uint16_t port; struct rte_flow_error error; - int64_t alloc, last_alloc; ret = rte_eal_init(argc, argv); if (ret < 0) @@ -1281,11 +1858,11 @@ main(int argc, char **argv) force_quit = false; dump_iterations = false; - flows_count = DEFAULT_RULES_COUNT; - iterations_number = DEFAULT_ITERATION; + rules_count = DEFAULT_RULES_COUNT; + rules_batch = DEFAULT_RULES_BATCH; delete_flag = false; dump_socket_mem_flag = false; - flow_group = 0; + flow_group = DEFAULT_GROUP; signal(SIGINT, signal_handler); signal(SIGTERM, signal_handler); @@ -1301,23 +1878,26 @@ main(int argc, char **argv) if (nb_lcores <= 1) rte_exit(EXIT_FAILURE, "This app needs at least two cores\n"); - last_alloc = (int64_t)dump_socket_mem(stdout); - flows_handler(); - alloc = (int64_t)dump_socket_mem(stdout); - if (last_alloc) - fprintf(stdout, ":: Memory allocation change(M): %.6lf\n", - (alloc - last_alloc) / 1.0e6); + printf(":: Flows Count per port: %d\n\n", rules_count); + + if (has_meter()) + create_meter_profile(); + rte_eal_mp_remote_launch(run_rte_flow_handler_cores, NULL, CALL_MAIN); if (enable_fwd) { init_lcore_info(); - rte_eal_mp_remote_launch(start_forwarding, NULL, CALL_MASTER); + rte_eal_mp_remote_launch(start_forwarding, NULL, CALL_MAIN); } + if (has_meter() && delete_flag) + destroy_meter_profile(); RTE_ETH_FOREACH_DEV(port) { rte_flow_flush(port, &error); - rte_eth_dev_stop(port); + if (rte_eth_dev_stop(port) != 0) + printf("Failed to stop device on port %u\n", port); rte_eth_dev_close(port); } + printf("\nBye ...\n"); return 0; }