linux/kernel/trace/ring_buffer_benchmark.c
<<
>>
Prefs
   1/*
   2 * ring buffer tester and benchmark
   3 *
   4 * Copyright (C) 2009 Steven Rostedt <srostedt@redhat.com>
   5 */
   6#include <linux/ring_buffer.h>
   7#include <linux/completion.h>
   8#include <linux/kthread.h>
   9#include <linux/module.h>
  10#include <linux/time.h>
  11#include <asm/local.h>
  12
  13struct rb_page {
  14        u64             ts;
  15        local_t         commit;
  16        char            data[4080];
  17};
  18
  19/* run time and sleep time in seconds */
  20#define RUN_TIME        10
  21#define SLEEP_TIME      10
  22
  23/* number of events for writer to wake up the reader */
  24static int wakeup_interval = 100;
  25
  26static int reader_finish;
  27static struct completion read_start;
  28static struct completion read_done;
  29
  30static struct ring_buffer *buffer;
  31static struct task_struct *producer;
  32static struct task_struct *consumer;
  33static unsigned long read;
  34
  35static int disable_reader;
  36module_param(disable_reader, uint, 0644);
  37MODULE_PARM_DESC(disable_reader, "only run producer");
  38
  39static int write_iteration = 50;
  40module_param(write_iteration, uint, 0644);
  41MODULE_PARM_DESC(write_iteration, "# of writes between timestamp readings");
  42
  43static int producer_nice = 19;
  44static int consumer_nice = 19;
  45
  46static int producer_fifo = -1;
  47static int consumer_fifo = -1;
  48
  49module_param(producer_nice, uint, 0644);
  50MODULE_PARM_DESC(producer_nice, "nice prio for producer");
  51
  52module_param(consumer_nice, uint, 0644);
  53MODULE_PARM_DESC(consumer_nice, "nice prio for consumer");
  54
  55module_param(producer_fifo, uint, 0644);
  56MODULE_PARM_DESC(producer_fifo, "fifo prio for producer");
  57
  58module_param(consumer_fifo, uint, 0644);
  59MODULE_PARM_DESC(consumer_fifo, "fifo prio for consumer");
  60
  61static int read_events;
  62
  63static int kill_test;
  64
  65#define KILL_TEST()                             \
  66        do {                                    \
  67                if (!kill_test) {               \
  68                        kill_test = 1;          \
  69                        WARN_ON(1);             \
  70                }                               \
  71        } while (0)
  72
  73enum event_status {
  74        EVENT_FOUND,
  75        EVENT_DROPPED,
  76};
  77
  78static enum event_status read_event(int cpu)
  79{
  80        struct ring_buffer_event *event;
  81        int *entry;
  82        u64 ts;
  83
  84        event = ring_buffer_consume(buffer, cpu, &ts, NULL);
  85        if (!event)
  86                return EVENT_DROPPED;
  87
  88        entry = ring_buffer_event_data(event);
  89        if (*entry != cpu) {
  90                KILL_TEST();
  91                return EVENT_DROPPED;
  92        }
  93
  94        read++;
  95        return EVENT_FOUND;
  96}
  97
  98static enum event_status read_page(int cpu)
  99{
 100        struct ring_buffer_event *event;
 101        struct rb_page *rpage;
 102        unsigned long commit;
 103        void *bpage;
 104        int *entry;
 105        int ret;
 106        int inc;
 107        int i;
 108
 109        bpage = ring_buffer_alloc_read_page(buffer, cpu);
 110        if (!bpage)
 111                return EVENT_DROPPED;
 112
 113        ret = ring_buffer_read_page(buffer, &bpage, PAGE_SIZE, cpu, 1);
 114        if (ret >= 0) {
 115                rpage = bpage;
 116                /* The commit may have missed event flags set, clear them */
 117                commit = local_read(&rpage->commit) & 0xfffff;
 118                for (i = 0; i < commit && !kill_test; i += inc) {
 119
 120                        if (i >= (PAGE_SIZE - offsetof(struct rb_page, data))) {
 121                                KILL_TEST();
 122                                break;
 123                        }
 124
 125                        inc = -1;
 126                        event = (void *)&rpage->data[i];
 127                        switch (event->type_len) {
 128                        case RINGBUF_TYPE_PADDING:
 129                                /* failed writes may be discarded events */
 130                                if (!event->time_delta)
 131                                        KILL_TEST();
 132                                inc = event->array[0] + 4;
 133                                break;
 134                        case RINGBUF_TYPE_TIME_EXTEND:
 135                                inc = 8;
 136                                break;
 137                        case 0:
 138                                entry = ring_buffer_event_data(event);
 139                                if (*entry != cpu) {
 140                                        KILL_TEST();
 141                                        break;
 142                                }
 143                                read++;
 144                                if (!event->array[0]) {
 145                                        KILL_TEST();
 146                                        break;
 147                                }
 148                                inc = event->array[0] + 4;
 149                                break;
 150                        default:
 151                                entry = ring_buffer_event_data(event);
 152                                if (*entry != cpu) {
 153                                        KILL_TEST();
 154                                        break;
 155                                }
 156                                read++;
 157                                inc = ((event->type_len + 1) * 4);
 158                        }
 159                        if (kill_test)
 160                                break;
 161
 162                        if (inc <= 0) {
 163                                KILL_TEST();
 164                                break;
 165                        }
 166                }
 167        }
 168        ring_buffer_free_read_page(buffer, bpage);
 169
 170        if (ret < 0)
 171                return EVENT_DROPPED;
 172        return EVENT_FOUND;
 173}
 174
 175static void ring_buffer_consumer(void)
 176{
 177        /* toggle between reading pages and events */
 178        read_events ^= 1;
 179
 180        read = 0;
 181        while (!reader_finish && !kill_test) {
 182                int found;
 183
 184                do {
 185                        int cpu;
 186
 187                        found = 0;
 188                        for_each_online_cpu(cpu) {
 189                                enum event_status stat;
 190
 191                                if (read_events)
 192                                        stat = read_event(cpu);
 193                                else
 194                                        stat = read_page(cpu);
 195
 196                                if (kill_test)
 197                                        break;
 198                                if (stat == EVENT_FOUND)
 199                                        found = 1;
 200                        }
 201                } while (found && !kill_test);
 202
 203                set_current_state(TASK_INTERRUPTIBLE);
 204                if (reader_finish)
 205                        break;
 206
 207                schedule();
 208                __set_current_state(TASK_RUNNING);
 209        }
 210        reader_finish = 0;
 211        complete(&read_done);
 212}
 213
 214static void ring_buffer_producer(void)
 215{
 216        struct timeval start_tv;
 217        struct timeval end_tv;
 218        unsigned long long time;
 219        unsigned long long entries;
 220        unsigned long long overruns;
 221        unsigned long missed = 0;
 222        unsigned long hit = 0;
 223        unsigned long avg;
 224        int cnt = 0;
 225
 226        /*
 227         * Hammer the buffer for 10 secs (this may
 228         * make the system stall)
 229         */
 230        trace_printk("Starting ring buffer hammer\n");
 231        do_gettimeofday(&start_tv);
 232        do {
 233                struct ring_buffer_event *event;
 234                int *entry;
 235                int i;
 236
 237                for (i = 0; i < write_iteration; i++) {
 238                        event = ring_buffer_lock_reserve(buffer, 10);
 239                        if (!event) {
 240                                missed++;
 241                        } else {
 242                                hit++;
 243                                entry = ring_buffer_event_data(event);
 244                                *entry = smp_processor_id();
 245                                ring_buffer_unlock_commit(buffer, event);
 246                        }
 247                }
 248                do_gettimeofday(&end_tv);
 249
 250                cnt++;
 251                if (consumer && !(cnt % wakeup_interval))
 252                        wake_up_process(consumer);
 253
 254#ifndef CONFIG_PREEMPT
 255                /*
 256                 * If we are a non preempt kernel, the 10 second run will
 257                 * stop everything while it runs. Instead, we will call
 258                 * cond_resched and also add any time that was lost by a
 259                 * rescedule.
 260                 *
 261                 * Do a cond resched at the same frequency we would wake up
 262                 * the reader.
 263                 */
 264                if (cnt % wakeup_interval)
 265                        cond_resched();
 266#endif
 267
 268        } while (end_tv.tv_sec < (start_tv.tv_sec + RUN_TIME) && !kill_test);
 269        trace_printk("End ring buffer hammer\n");
 270
 271        if (consumer) {
 272                /* Init both completions here to avoid races */
 273                init_completion(&read_start);
 274                init_completion(&read_done);
 275                /* the completions must be visible before the finish var */
 276                smp_wmb();
 277                reader_finish = 1;
 278                /* finish var visible before waking up the consumer */
 279                smp_wmb();
 280                wake_up_process(consumer);
 281                wait_for_completion(&read_done);
 282        }
 283
 284        time = end_tv.tv_sec - start_tv.tv_sec;
 285        time *= USEC_PER_SEC;
 286        time += (long long)((long)end_tv.tv_usec - (long)start_tv.tv_usec);
 287
 288        entries = ring_buffer_entries(buffer);
 289        overruns = ring_buffer_overruns(buffer);
 290
 291        if (kill_test)
 292                trace_printk("ERROR!\n");
 293
 294        if (!disable_reader) {
 295                if (consumer_fifo < 0)
 296                        trace_printk("Running Consumer at nice: %d\n",
 297                                     consumer_nice);
 298                else
 299                        trace_printk("Running Consumer at SCHED_FIFO %d\n",
 300                                     consumer_fifo);
 301        }
 302        if (producer_fifo < 0)
 303                trace_printk("Running Producer at nice: %d\n",
 304                             producer_nice);
 305        else
 306                trace_printk("Running Producer at SCHED_FIFO %d\n",
 307                             producer_fifo);
 308
 309        /* Let the user know that the test is running at low priority */
 310        if (producer_fifo < 0 && consumer_fifo < 0 &&
 311            producer_nice == 19 && consumer_nice == 19)
 312                trace_printk("WARNING!!! This test is running at lowest priority.\n");
 313
 314        trace_printk("Time:     %lld (usecs)\n", time);
 315        trace_printk("Overruns: %lld\n", overruns);
 316        if (disable_reader)
 317                trace_printk("Read:     (reader disabled)\n");
 318        else
 319                trace_printk("Read:     %ld  (by %s)\n", read,
 320                        read_events ? "events" : "pages");
 321        trace_printk("Entries:  %lld\n", entries);
 322        trace_printk("Total:    %lld\n", entries + overruns + read);
 323        trace_printk("Missed:   %ld\n", missed);
 324        trace_printk("Hit:      %ld\n", hit);
 325
 326        /* Convert time from usecs to millisecs */
 327        do_div(time, USEC_PER_MSEC);
 328        if (time)
 329                hit /= (long)time;
 330        else
 331                trace_printk("TIME IS ZERO??\n");
 332
 333        trace_printk("Entries per millisec: %ld\n", hit);
 334
 335        if (hit) {
 336                /* Calculate the average time in nanosecs */
 337                avg = NSEC_PER_MSEC / hit;
 338                trace_printk("%ld ns per entry\n", avg);
 339        }
 340
 341        if (missed) {
 342                if (time)
 343                        missed /= (long)time;
 344
 345                trace_printk("Total iterations per millisec: %ld\n",
 346                             hit + missed);
 347
 348                /* it is possible that hit + missed will overflow and be zero */
 349                if (!(hit + missed)) {
 350                        trace_printk("hit + missed overflowed and totalled zero!\n");
 351                        hit--; /* make it non zero */
 352                }
 353
 354                /* Caculate the average time in nanosecs */
 355                avg = NSEC_PER_MSEC / (hit + missed);
 356                trace_printk("%ld ns per entry\n", avg);
 357        }
 358}
 359
 360static void wait_to_die(void)
 361{
 362        set_current_state(TASK_INTERRUPTIBLE);
 363        while (!kthread_should_stop()) {
 364                schedule();
 365                set_current_state(TASK_INTERRUPTIBLE);
 366        }
 367        __set_current_state(TASK_RUNNING);
 368}
 369
 370static int ring_buffer_consumer_thread(void *arg)
 371{
 372        while (!kthread_should_stop() && !kill_test) {
 373                complete(&read_start);
 374
 375                ring_buffer_consumer();
 376
 377                set_current_state(TASK_INTERRUPTIBLE);
 378                if (kthread_should_stop() || kill_test)
 379                        break;
 380
 381                schedule();
 382                __set_current_state(TASK_RUNNING);
 383        }
 384        __set_current_state(TASK_RUNNING);
 385
 386        if (kill_test)
 387                wait_to_die();
 388
 389        return 0;
 390}
 391
 392static int ring_buffer_producer_thread(void *arg)
 393{
 394        init_completion(&read_start);
 395
 396        while (!kthread_should_stop() && !kill_test) {
 397                ring_buffer_reset(buffer);
 398
 399                if (consumer) {
 400                        smp_wmb();
 401                        wake_up_process(consumer);
 402                        wait_for_completion(&read_start);
 403                }
 404
 405                ring_buffer_producer();
 406
 407                trace_printk("Sleeping for 10 secs\n");
 408                set_current_state(TASK_INTERRUPTIBLE);
 409                schedule_timeout(HZ * SLEEP_TIME);
 410                __set_current_state(TASK_RUNNING);
 411        }
 412
 413        if (kill_test)
 414                wait_to_die();
 415
 416        return 0;
 417}
 418
 419static int __init ring_buffer_benchmark_init(void)
 420{
 421        int ret;
 422
 423        /* make a one meg buffer in overwite mode */
 424        buffer = ring_buffer_alloc(1000000, RB_FL_OVERWRITE);
 425        if (!buffer)
 426                return -ENOMEM;
 427
 428        if (!disable_reader) {
 429                consumer = kthread_create(ring_buffer_consumer_thread,
 430                                          NULL, "rb_consumer");
 431                ret = PTR_ERR(consumer);
 432                if (IS_ERR(consumer))
 433                        goto out_fail;
 434        }
 435
 436        producer = kthread_run(ring_buffer_producer_thread,
 437                               NULL, "rb_producer");
 438        ret = PTR_ERR(producer);
 439
 440        if (IS_ERR(producer))
 441                goto out_kill;
 442
 443        /*
 444         * Run them as low-prio background tasks by default:
 445         */
 446        if (!disable_reader) {
 447                if (consumer_fifo >= 0) {
 448                        struct sched_param param = {
 449                                .sched_priority = consumer_fifo
 450                        };
 451                        sched_setscheduler(consumer, SCHED_FIFO, &param);
 452                } else
 453                        set_user_nice(consumer, consumer_nice);
 454        }
 455
 456        if (producer_fifo >= 0) {
 457                struct sched_param param = {
 458                        .sched_priority = consumer_fifo
 459                };
 460                sched_setscheduler(producer, SCHED_FIFO, &param);
 461        } else
 462                set_user_nice(producer, producer_nice);
 463
 464        return 0;
 465
 466 out_kill:
 467        if (consumer)
 468                kthread_stop(consumer);
 469
 470 out_fail:
 471        ring_buffer_free(buffer);
 472        return ret;
 473}
 474
 475static void __exit ring_buffer_benchmark_exit(void)
 476{
 477        kthread_stop(producer);
 478        if (consumer)
 479                kthread_stop(consumer);
 480        ring_buffer_free(buffer);
 481}
 482
 483module_init(ring_buffer_benchmark_init);
 484module_exit(ring_buffer_benchmark_exit);
 485
 486MODULE_AUTHOR("Steven Rostedt");
 487MODULE_DESCRIPTION("ring_buffer_benchmark");
 488MODULE_LICENSE("GPL");
 489