| 1 | |
| 2 | |
| 3 | * Copyright (C) 2021 Red Hat Inc, Daniel Bristot de Oliveira <[email protected]>
|
| 4 |
|
| 5 | #define _GNU_SOURCE
|
| 6 | #include <sys/types.h>
|
| 7 | #include <sys/stat.h>
|
| 8 | #include <pthread.h>
|
| 9 | #include <stdlib.h>
|
| 10 | #include <string.h>
|
| 11 | #include <unistd.h>
|
| 12 | #include <fcntl.h>
|
| 13 | #include <stdio.h>
|
| 14 | #include <sched.h>
|
| 15 |
|
| 16 | #include <linux/compiler.h>
|
| 17 |
|
| 18 | #include "timerlat.h"
|
| 19 | #include "timerlat_aa.h"
|
| 20 | #include "timerlat_bpf.h"
|
| 21 |
|
| 22 | #define DEFAULT_TIMERLAT_PERIOD 1000
|
| 23 |
|
| 24 | static int dma_latency_fd = -1;
|
| 25 |
|
| 26 | |
| 27 | * timerlat_apply_config - apply common configs to the initialized tool
|
| 28 |
|
| 29 | int
|
| 30 | timerlat_apply_config(struct osnoise_tool *tool, struct timerlat_params *params)
|
| 31 | {
|
| 32 | int retval;
|
| 33 | const char *const rtla_no_bpf = getenv("RTLA_NO_BPF");
|
| 34 |
|
| 35 | |
| 36 | * Try to enable BPF, unless disabled explicitly.
|
| 37 | * If BPF enablement fails, fall back to tracefs mode.
|
| 38 |
|
| 39 | if (rtla_no_bpf && strncmp_static(rtla_no_bpf, "1") == 0) {
|
| 40 | debug_msg("RTLA_NO_BPF set, disabling BPF\n");
|
| 41 | params->mode = TRACING_MODE_TRACEFS;
|
| 42 | } else if (!tep_find_event_by_name(tool->trace.tep, "osnoise", "timerlat_sample")) {
|
| 43 | debug_msg("osnoise:timerlat_sample missing, disabling BPF\n");
|
| 44 | params->mode = TRACING_MODE_TRACEFS;
|
| 45 | } else {
|
| 46 | retval = timerlat_bpf_init(params);
|
| 47 | if (retval) {
|
| 48 | debug_msg("Could not enable BPF\n");
|
| 49 | params->mode = TRACING_MODE_TRACEFS;
|
| 50 | }
|
| 51 | }
|
| 52 |
|
| 53 |
|
| 54 | if (params->bpf_action_program) {
|
| 55 | if (params->mode == TRACING_MODE_TRACEFS) {
|
| 56 | err_msg("BPF actions are not supported in tracefs-only mode\n");
|
| 57 | goto out_err;
|
| 58 | }
|
| 59 |
|
| 60 | if (timerlat_load_bpf_action_program(params->bpf_action_program))
|
| 61 | goto out_err;
|
| 62 | }
|
| 63 |
|
| 64 | retval = osnoise_set_timerlat_period_us(tool->context,
|
| 65 | params->timerlat_period_us ?
|
| 66 | params->timerlat_period_us :
|
| 67 | DEFAULT_TIMERLAT_PERIOD);
|
| 68 | if (retval) {
|
| 69 | err_msg("Failed to set timerlat period\n");
|
| 70 | goto out_err;
|
| 71 | }
|
| 72 |
|
| 73 |
|
| 74 | retval = osnoise_set_print_stack(tool->context, params->print_stack);
|
| 75 | if (retval) {
|
| 76 | err_msg("Failed to set print stack\n");
|
| 77 | goto out_err;
|
| 78 | }
|
| 79 |
|
| 80 | retval = osnoise_set_timerlat_align(tool->context, params->timerlat_align);
|
| 81 | if (retval && params->timerlat_align) {
|
| 82 | |
| 83 | * We might be running on a kernel that does not support timerlat align.
|
| 84 | * Unless user requested it explicitly, ignore the error.
|
| 85 |
|
| 86 | err_msg("Failed to enable timerlat align\n");
|
| 87 | goto out_err;
|
| 88 | }
|
| 89 |
|
| 90 | if (params->timerlat_align) {
|
| 91 | retval = osnoise_set_timerlat_align_us(tool->context, params->timerlat_align_us);
|
| 92 | if (retval) {
|
| 93 | err_msg("Failed to set timerlat align us\n");
|
| 94 | goto out_err;
|
| 95 | }
|
| 96 | }
|
| 97 |
|
| 98 | |
| 99 | * If the user did not specify a type of thread, try user-threads first.
|
| 100 | * Fall back to kernel threads otherwise.
|
| 101 |
|
| 102 | if (!params->common.kernel_workload && !params->common.user_data) {
|
| 103 | retval = tracefs_file_exists(NULL, "osnoise/per_cpu/cpu0/timerlat_fd");
|
| 104 | if (retval) {
|
| 105 | debug_msg("User-space interface detected, setting user-threads\n");
|
| 106 | params->common.user_workload = 1;
|
| 107 | params->common.user_data = 1;
|
| 108 | } else {
|
| 109 | debug_msg("User-space interface not detected, setting kernel-threads\n");
|
| 110 | params->common.kernel_workload = 1;
|
| 111 | }
|
| 112 | }
|
| 113 |
|
| 114 | return common_apply_config(tool, ¶ms->common);
|
| 115 |
|
| 116 | out_err:
|
| 117 | return -1;
|
| 118 | }
|
| 119 |
|
| 120 | int timerlat_enable(struct osnoise_tool *tool)
|
| 121 | {
|
| 122 | struct timerlat_params *params = to_timerlat_params(tool->params);
|
| 123 | int retval, i;
|
| 124 |
|
| 125 | if (params->dma_latency >= 0) {
|
| 126 | dma_latency_fd = set_cpu_dma_latency(params->dma_latency);
|
| 127 | if (dma_latency_fd < 0) {
|
| 128 | err_msg("Could not set /dev/cpu_dma_latency.\n");
|
| 129 | return -1;
|
| 130 | }
|
| 131 | }
|
| 132 |
|
| 133 | if (params->deepest_idle_state >= -1) {
|
| 134 | if (!have_libcpupower_support()) {
|
| 135 | err_msg("rtla built without libcpupower, --deepest-idle-state is not supported\n");
|
| 136 | return -1;
|
| 137 | }
|
| 138 |
|
| 139 | for_each_monitored_cpu(i, ¶ms->common) {
|
| 140 | if (save_cpu_idle_disable_state(i) < 0) {
|
| 141 | err_msg("Could not save cpu idle state.\n");
|
| 142 | return -1;
|
| 143 | }
|
| 144 | if (set_deepest_cpu_idle_state(i, params->deepest_idle_state) < 0) {
|
| 145 | err_msg("Could not set deepest cpu idle state.\n");
|
| 146 | return -1;
|
| 147 | }
|
| 148 | }
|
| 149 | }
|
| 150 |
|
| 151 | if (!params->no_aa) {
|
| 152 | tool->aa = osnoise_init_tool("timerlat_aa");
|
| 153 | if (!tool->aa)
|
| 154 | return -1;
|
| 155 |
|
| 156 | retval = timerlat_aa_init(tool->aa, params->dump_tasks, params->stack_format);
|
| 157 | if (retval) {
|
| 158 | err_msg("Failed to enable the auto analysis instance\n");
|
| 159 | return retval;
|
| 160 | }
|
| 161 |
|
| 162 | retval = enable_tracer_by_name(tool->aa->trace.inst, "timerlat");
|
| 163 | if (retval) {
|
| 164 | err_msg("Failed to enable aa tracer\n");
|
| 165 | return retval;
|
| 166 | }
|
| 167 | }
|
| 168 |
|
| 169 | if (params->common.warmup > 0) {
|
| 170 | debug_msg("Warming up for %d seconds\n", params->common.warmup);
|
| 171 | sleep(params->common.warmup);
|
| 172 | if (stop_tracing)
|
| 173 | return -1;
|
| 174 | }
|
| 175 |
|
| 176 | |
| 177 | * Start the tracers here, after having set all instances.
|
| 178 | *
|
| 179 | * Let the trace instance start first for the case of hitting a stop
|
| 180 | * tracing while enabling other instances. The trace instance is the
|
| 181 | * one with most valuable information.
|
| 182 |
|
| 183 | if (tool->record)
|
| 184 | trace_instance_start(&tool->record->trace);
|
| 185 | if (!params->no_aa)
|
| 186 | trace_instance_start(&tool->aa->trace);
|
| 187 | if (params->mode == TRACING_MODE_TRACEFS) {
|
| 188 | trace_instance_start(&tool->trace);
|
| 189 | } else {
|
| 190 | retval = timerlat_bpf_attach();
|
| 191 | if (retval) {
|
| 192 | err_msg("Error attaching BPF program\n");
|
| 193 | return retval;
|
| 194 | }
|
| 195 | }
|
| 196 |
|
| 197 | |
| 198 | * In tracefs and mixed mode, timerlat tracer handles stopping
|
| 199 | * on threshold
|
| 200 |
|
| 201 | if (params->mode != TRACING_MODE_BPF) {
|
| 202 | retval = osn_set_stop(tool);
|
| 203 | if (retval)
|
| 204 | return retval;
|
| 205 | }
|
| 206 |
|
| 207 | return 0;
|
| 208 | }
|
| 209 |
|
| 210 | void timerlat_analyze(struct osnoise_tool *tool, bool stopped)
|
| 211 | {
|
| 212 | struct timerlat_params *params = to_timerlat_params(tool->params);
|
| 213 |
|
| 214 | if (stopped) {
|
| 215 | if (!params->no_aa)
|
| 216 | timerlat_auto_analysis(params->common.stop_us,
|
| 217 | params->common.stop_total_us);
|
| 218 | } else if (params->common.aa_only) {
|
| 219 | char *max_lat;
|
| 220 |
|
| 221 | |
| 222 | * If the trace did not stop with --aa-only, at least print
|
| 223 | * the max known latency.
|
| 224 |
|
| 225 | max_lat = tracefs_instance_file_read(tool->trace.inst, "tracing_max_latency", NULL);
|
| 226 | if (max_lat) {
|
| 227 | printf(" Max latency was %s\n", max_lat);
|
| 228 | free(max_lat);
|
| 229 | }
|
| 230 | }
|
| 231 | }
|
| 232 |
|
| 233 | void timerlat_free(struct osnoise_tool *tool)
|
| 234 | {
|
| 235 | struct timerlat_params *params = to_timerlat_params(tool->params);
|
| 236 | int i;
|
| 237 |
|
| 238 | timerlat_aa_destroy();
|
| 239 | if (dma_latency_fd >= 0)
|
| 240 | close(dma_latency_fd);
|
| 241 | if (params->deepest_idle_state >= -1) {
|
| 242 | for_each_monitored_cpu(i, ¶ms->common) {
|
| 243 | restore_cpu_idle_disable_state(i);
|
| 244 | }
|
| 245 | }
|
| 246 |
|
| 247 | osnoise_destroy_tool(tool->aa);
|
| 248 |
|
| 249 | if (params->mode != TRACING_MODE_TRACEFS)
|
| 250 | timerlat_bpf_destroy();
|
| 251 | free_cpu_idle_disable_states();
|
| 252 | }
|
| 253 |
|
| 254 | __noreturn static void timerlat_usage(int err)
|
| 255 | {
|
| 256 | int i;
|
| 257 |
|
| 258 | static const char * const msg[] = {
|
| 259 | "",
|
| 260 | "timerlat version " VERSION,
|
| 261 | "",
|
| 262 | " usage: [rtla] timerlat [MODE] ...",
|
| 263 | "",
|
| 264 | " modes:",
|
| 265 | " top - prints the summary from timerlat tracer",
|
| 266 | " hist - prints a histogram of timer latencies",
|
| 267 | "",
|
| 268 | "if no MODE is given, the top mode is called, passing the arguments",
|
| 269 | NULL,
|
| 270 | };
|
| 271 |
|
| 272 | for (i = 0; msg[i]; i++)
|
| 273 | fprintf(stderr, "%s\n", msg[i]);
|
| 274 | exit(err);
|
| 275 | }
|
| 276 |
|
| 277 | int timerlat_main(int argc, char *argv[])
|
| 278 | {
|
| 279 | if (argc == 0)
|
| 280 | goto usage;
|
| 281 |
|
| 282 | |
| 283 | * if timerlat was called without any argument, run the
|
| 284 | * default cmdline.
|
| 285 |
|
| 286 | if (argc == 1) {
|
| 287 | run_tool(&timerlat_top_ops, argc, argv);
|
| 288 | exit(0);
|
| 289 | }
|
| 290 |
|
| 291 | if ((strcmp(argv[1], "-h") == 0) || (strcmp(argv[1], "--help") == 0)) {
|
| 292 | timerlat_usage(129);
|
| 293 | } else if (str_has_prefix(argv[1], "-")) {
|
| 294 |
|
| 295 | run_tool(&timerlat_top_ops, argc, argv);
|
| 296 | exit(0);
|
| 297 | } else if (strcmp(argv[1], "top") == 0) {
|
| 298 | run_tool(&timerlat_top_ops, argc-1, &argv[1]);
|
| 299 | exit(0);
|
| 300 | } else if (strcmp(argv[1], "hist") == 0) {
|
| 301 | run_tool(&timerlat_hist_ops, argc-1, &argv[1]);
|
| 302 | exit(0);
|
| 303 | }
|
| 304 |
|
| 305 | usage:
|
| 306 | timerlat_usage(129);
|
| 307 | }
|