X-Git-Url: https://git.kernel.dk/?p=fio.git;a=blobdiff_plain;f=backend.c;h=e5bb4e259034b29af86fbaed018ae64c7bb288db;hp=f519728c2131f073f55869ed765118fbbeafe8b9;hb=b3d5e3fd80e3834097578e92d1b788065b8346e1;hpb=0e21f5c6f64a73bcede20e1a29a885845e453b8e diff --git a/backend.c b/backend.c index f519728c..e5bb4e25 100644 --- a/backend.c +++ b/backend.c @@ -62,17 +62,22 @@ struct io_log *agg_io_log[DDIR_RWDIR_CNT]; int groupid = 0; unsigned int thread_number = 0; +unsigned int nr_segments = 0; +unsigned int cur_segment = 0; unsigned int stat_number = 0; -int shm_id = 0; int temp_stall_ts; unsigned long done_secs = 0; +#ifdef PTHREAD_ERRORCHECK_MUTEX_INITIALIZER_NP +pthread_mutex_t overlap_check = PTHREAD_ERRORCHECK_MUTEX_INITIALIZER_NP; +#else pthread_mutex_t overlap_check = PTHREAD_MUTEX_INITIALIZER; +#endif #define JOB_START_TIMEOUT (5 * 1000) static void sig_int(int sig) { - if (threads) { + if (nr_segments) { if (is_backend) fio_server_got_signal(sig); else { @@ -131,13 +136,10 @@ static void set_sig_handlers(void) static bool __check_min_rate(struct thread_data *td, struct timespec *now, enum fio_ddir ddir) { - unsigned long long bytes = 0; - unsigned long iops = 0; - unsigned long spent; - unsigned long long rate; - unsigned long long ratemin = 0; - unsigned int rate_iops = 0; - unsigned int rate_iops_min = 0; + unsigned long long current_rate_check_bytes = td->this_io_bytes[ddir]; + unsigned long current_rate_check_blocks = td->this_io_blocks[ddir]; + unsigned long long option_rate_bytes_min = td->o.ratemin[ddir]; + unsigned int option_rate_iops_min = td->o.rate_iops_min[ddir]; assert(ddir_rw(ddir)); @@ -150,68 +152,44 @@ static bool __check_min_rate(struct thread_data *td, struct timespec *now, if (mtime_since(&td->start, now) < 2000) return false; - iops += td->this_io_blocks[ddir]; - bytes += td->this_io_bytes[ddir]; - ratemin += td->o.ratemin[ddir]; - rate_iops += td->o.rate_iops[ddir]; - rate_iops_min += td->o.rate_iops_min[ddir]; - /* - * if rate blocks is set, sample is running + * if last_rate_check_blocks or last_rate_check_bytes is set, + * we can compute a rate per ratecycle */ - if (td->rate_bytes[ddir] || td->rate_blocks[ddir]) { - spent = mtime_since(&td->lastrate[ddir], now); - if (spent < td->o.ratecycle) + if (td->last_rate_check_bytes[ddir] || td->last_rate_check_blocks[ddir]) { + unsigned long spent = mtime_since(&td->last_rate_check_time[ddir], now); + if (spent < td->o.ratecycle || spent==0) return false; - if (td->o.rate[ddir] || td->o.ratemin[ddir]) { + if (td->o.ratemin[ddir]) { /* * check bandwidth specified rate */ - if (bytes < td->rate_bytes[ddir]) { - log_err("%s: rate_min=%lluB/s not met, only transferred %lluB\n", - td->o.name, ratemin, bytes); + unsigned long long current_rate_bytes = + ((current_rate_check_bytes - td->last_rate_check_bytes[ddir]) * 1000) / spent; + if (current_rate_bytes < option_rate_bytes_min) { + log_err("%s: rate_min=%lluB/s not met, got %lluB/s\n", + td->o.name, option_rate_bytes_min, current_rate_bytes); return true; - } else { - if (spent) - rate = ((bytes - td->rate_bytes[ddir]) * 1000) / spent; - else - rate = 0; - - if (rate < ratemin || - bytes < td->rate_bytes[ddir]) { - log_err("%s: rate_min=%lluB/s not met, got %lluB/s\n", - td->o.name, ratemin, rate); - return true; - } } } else { /* * checks iops specified rate */ - if (iops < rate_iops) { - log_err("%s: rate_iops_min=%u not met, only performed %lu IOs\n", - td->o.name, rate_iops, iops); + unsigned long long current_rate_iops = + ((current_rate_check_blocks - td->last_rate_check_blocks[ddir]) * 1000) / spent; + + if (current_rate_iops < option_rate_iops_min) { + log_err("%s: rate_iops_min=%u not met, got %llu IOPS\n", + td->o.name, option_rate_iops_min, current_rate_iops); return true; - } else { - if (spent) - rate = ((iops - td->rate_blocks[ddir]) * 1000) / spent; - else - rate = 0; - - if (rate < rate_iops_min || - iops < td->rate_blocks[ddir]) { - log_err("%s: rate_iops_min=%u not met, got %llu IOPS\n", - td->o.name, rate_iops_min, rate); - return true; - } } } } - td->rate_bytes[ddir] = bytes; - td->rate_blocks[ddir] = iops; - memcpy(&td->lastrate[ddir], now, sizeof(*now)); + td->last_rate_check_bytes[ddir] = current_rate_check_bytes; + td->last_rate_check_blocks[ddir] = current_rate_check_blocks; + memcpy(&td->last_rate_check_time[ddir], now, sizeof(*now)); return false; } @@ -219,12 +197,10 @@ static bool check_min_rate(struct thread_data *td, struct timespec *now) { bool ret = false; - if (td->bytes_done[DDIR_READ]) - ret |= __check_min_rate(td, now, DDIR_READ); - if (td->bytes_done[DDIR_WRITE]) - ret |= __check_min_rate(td, now, DDIR_WRITE); - if (td->bytes_done[DDIR_TRIM]) - ret |= __check_min_rate(td, now, DDIR_TRIM); + for_each_rw_ddir(ddir) { + if (td->bytes_done[ddir]) + ret |= __check_min_rate(td, now, ddir); + } return ret; } @@ -390,7 +366,7 @@ static bool break_on_this_error(struct thread_data *td, enum fio_ddir ddir, td_clear_error(td); *retptr = 0; return false; - } else if (td->o.fill_device && err == ENOSPC) { + } else if (td->o.fill_device && (err == ENOSPC || err == EDQUOT)) { /* * We expect to hit this error if * fill_device option is set. @@ -436,7 +412,7 @@ static int wait_for_completions(struct thread_data *td, struct timespec *time) if ((full && !min_evts) || !td->o.iodepth_batch_complete_min) min_evts = 1; - if (time && __should_check_rate(td)) + if (time && should_check_rate(td)) fio_gettime(time, NULL); do { @@ -491,7 +467,7 @@ int io_queue_event(struct thread_data *td, struct io_u *io_u, int *ret, requeue_io_u(td, &io_u); } else { sync_done: - if (comp_time && __should_check_rate(td)) + if (comp_time && should_check_rate(td)) fio_gettime(comp_time, NULL); *ret = io_u_sync_complete(td, io_u); @@ -834,7 +810,7 @@ static long long usec_for_io(struct thread_data *td, enum fio_ddir ddir) if (td->o.rate_process == RATE_PROCESS_POISSON) { uint64_t val, iops; - iops = bps / td->o.bs[ddir]; + iops = bps / td->o.min_bs[ddir]; val = (int64_t) (1000000 / iops) * -logf(__rand_0_1(&td->poisson_state[ddir])); if (val) { @@ -855,14 +831,47 @@ static long long usec_for_io(struct thread_data *td, enum fio_ddir ddir) return 0; } -static void handle_thinktime(struct thread_data *td, enum fio_ddir ddir) +static void init_thinktime(struct thread_data *td) +{ + if (td->o.thinktime_blocks_type == THINKTIME_BLOCKS_TYPE_COMPLETE) + td->thinktime_blocks_counter = td->io_blocks; + else + td->thinktime_blocks_counter = td->io_issues; + td->last_thinktime = td->epoch; + td->last_thinktime_blocks = 0; +} + +static void handle_thinktime(struct thread_data *td, enum fio_ddir ddir, + struct timespec *time) { unsigned long long b; uint64_t total; int left; + struct timespec now; + bool stall = false; + + if (td->o.thinktime_iotime) { + fio_gettime(&now, NULL); + if (utime_since(&td->last_thinktime, &now) + >= td->o.thinktime_iotime + td->o.thinktime) { + stall = true; + } else if (!fio_option_is_set(&td->o, thinktime_blocks)) { + /* + * When thinktime_iotime is set and thinktime_blocks is + * not set, skip the thinktime_blocks check, since + * thinktime_blocks default value 1 does not work + * together with thinktime_iotime. + */ + return; + } - b = ddir_rw_sum(td->io_blocks); - if (b % td->o.thinktime_blocks) + } + + b = ddir_rw_sum(td->thinktime_blocks_counter); + if (b >= td->last_thinktime_blocks + td->o.thinktime_blocks) + stall = true; + + if (!stall) return; io_u_quiesce(td); @@ -895,6 +904,13 @@ static void handle_thinktime(struct thread_data *td, enum fio_ddir ddir) /* adjust for rate_process=poisson */ td->last_usec[ddir] += total; } + + if (time && should_check_rate(td)) + fio_gettime(time, NULL); + + td->last_thinktime_blocks = b; + if (td->o.thinktime_iotime) + td->last_thinktime = now; } /* @@ -1006,12 +1022,6 @@ static void do_io(struct thread_data *td, uint64_t *bytes_done) if (td->o.verify != VERIFY_NONE && io_u->ddir == DDIR_READ && ((io_u->flags & IO_U_F_VER_LIST) || !td_rw(td))) { - if (!td->o.verify_pattern_bytes) { - io_u->rand_seed = __rand(&td->verify_state); - if (sizeof(int) != sizeof(long *)) - io_u->rand_seed *= __rand(&td->verify_state); - } - if (verify_state_should_stop(td, io_u)) { put_io_u(td, io_u); break; @@ -1054,8 +1064,10 @@ static void do_io(struct thread_data *td, uint64_t *bytes_done) td->rate_io_issue_bytes[__ddir] += blen; } - if (should_check_rate(td)) + if (should_check_rate(td)) { td->rate_next_io_time[__ddir] = usec_for_io(td, __ddir); + fio_gettime(&comp_time, NULL); + } } else { ret = io_u_submit(td, io_u); @@ -1079,6 +1091,10 @@ reap: } if (ret < 0) break; + + if (ddir_rw(ddir) && td->o.thinktime) + handle_thinktime(td, ddir, &comp_time); + if (!ddir_rw_sum(td->bytes_done) && !td_ioengine_flagged(td, FIO_NOIO)) continue; @@ -1093,9 +1109,6 @@ reap: } if (!in_ramp_time(td) && td->o.latency_target) lat_target_check(td); - - if (ddir_rw(ddir) && td->o.thinktime) - handle_thinktime(td, ddir); } check_update_rusage(td); @@ -1103,7 +1116,7 @@ reap: if (td->trim_entries) log_err("fio: %lu trim entries leaked?\n", td->trim_entries); - if (td->o.fill_device && td->error == ENOSPC) { + if (td->o.fill_device && (td->error == ENOSPC || td->error == EDQUOT)) { td->error = 0; fio_mark_td_terminate(td); } @@ -1118,7 +1131,8 @@ reap: if (i) { ret = io_u_queued_complete(td, i); - if (td->o.fill_device && td->error == ENOSPC) + if (td->o.fill_device && + (td->error == ENOSPC || td->error == EDQUOT)) td->error = 0; } @@ -1133,8 +1147,11 @@ reap: f->file_name); } } - } else + } else { + if (td->o.io_submit_mode == IO_MODE_OFFLOAD) + workqueue_flush(&td->io_wq); cleanup_pending_aio(td); + } /* * stop job if we failed doing any IO @@ -1339,22 +1356,19 @@ int init_io_u_buffers(struct thread_data *td) return 0; } +#ifdef FIO_HAVE_IOSCHED_SWITCH /* - * This function is Linux specific. + * These functions are Linux specific. * FIO_HAVE_IOSCHED_SWITCH enabled currently means it's Linux. */ -static int switch_ioscheduler(struct thread_data *td) +static int set_ioscheduler(struct thread_data *td, struct fio_file *file) { -#ifdef FIO_HAVE_IOSCHED_SWITCH char tmp[256], tmp2[128], *p; FILE *f; int ret; - if (td_ioengine_flagged(td, FIO_DISKLESSIO)) - return 0; - - assert(td->files && td->files[0]); - sprintf(tmp, "%s/queue/scheduler", td->files[0]->du->sysfs_root); + assert(file->du && file->du->sysfs_root); + sprintf(tmp, "%s/queue/scheduler", file->du->sysfs_root); f = fopen(tmp, "r+"); if (!f) { @@ -1407,7 +1421,7 @@ static int switch_ioscheduler(struct thread_data *td) sprintf(tmp2, "[%s]", td->o.ioscheduler); if (!strstr(tmp, tmp2)) { - log_err("fio: io scheduler %s not found\n", td->o.ioscheduler); + log_err("fio: unable to set io scheduler to %s\n", td->o.ioscheduler); td_verror(td, EINVAL, "iosched_switch"); fclose(f); return 1; @@ -1415,11 +1429,55 @@ static int switch_ioscheduler(struct thread_data *td) fclose(f); return 0; +} + +static int switch_ioscheduler(struct thread_data *td) +{ + struct fio_file *f; + unsigned int i; + int ret = 0; + + if (td_ioengine_flagged(td, FIO_DISKLESSIO)) + return 0; + + assert(td->files && td->files[0]); + + for_each_file(td, f, i) { + + /* Only consider regular files and block device files */ + switch (f->filetype) { + case FIO_TYPE_FILE: + case FIO_TYPE_BLOCK: + /* + * Make sure that the device hosting the file could + * be determined. + */ + if (!f->du) + continue; + break; + case FIO_TYPE_CHAR: + case FIO_TYPE_PIPE: + default: + continue; + } + + ret = set_ioscheduler(td, f); + if (ret) + return ret; + } + + return 0; +} + #else + +static int switch_ioscheduler(struct thread_data *td) +{ return 0; -#endif } +#endif /* FIO_HAVE_IOSCHED_SWITCH */ + static bool keep_running(struct thread_data *td) { unsigned long long limit; @@ -1462,16 +1520,17 @@ static bool keep_running(struct thread_data *td) return false; } -static int exec_string(struct thread_options *o, const char *string, const char *mode) +static int exec_string(struct thread_options *o, const char *string, + const char *mode) { - size_t newlen = strlen(string) + strlen(o->name) + strlen(mode) + 13 + 1; int ret; char *str; - str = malloc(newlen); - sprintf(str, "%s > %s.%s.txt 2>&1", string, o->name, mode); + if (asprintf(&str, "%s > %s.%s.txt 2>&1", string, o->name, mode) < 0) + return -1; - log_info("%s : Saving output of %s in %s.%s.txt\n",o->name, mode, o->name, mode); + log_info("%s : Saving output of %s in %s.%s.txt\n", o->name, mode, + o->name, mode); ret = system(str); if (ret == -1) log_err("fio: exec of cmd <%s> failed\n", str); @@ -1541,7 +1600,7 @@ static void *thread_main(void *data) uint64_t bytes_done[DDIR_RWDIR_CNT]; int deadlock_loop_cnt; bool clear_state; - int ret; + int res, ret; sk_out_assign(sk_out); free(fd); @@ -1696,6 +1755,18 @@ static void *thread_main(void *data) if (!init_iolog(td)) goto err; + /* ioprio_set() has to be done before td_io_init() */ + if (fio_option_is_set(o, ioprio) || + fio_option_is_set(o, ioprio_class)) { + ret = ioprio_set(IOPRIO_WHO_PROCESS, 0, o->ioprio_class, o->ioprio); + if (ret == -1) { + td_verror(td, errno, "ioprio_set"); + goto err; + } + td->ioprio = ioprio_value(o->ioprio_class, o->ioprio); + td->ts.ioprio = td->ioprio; + } + if (td_io_init(td)) goto err; @@ -1708,15 +1779,6 @@ static void *thread_main(void *data) if (o->verify_async && verify_async_init(td)) goto err; - if (fio_option_is_set(o, ioprio) || - fio_option_is_set(o, ioprio_class)) { - ret = ioprio_set(IOPRIO_WHO_PROCESS, 0, o->ioprio_class, o->ioprio); - if (ret == -1) { - td_verror(td, errno, "ioprio_set"); - goto err; - } - } - if (o->cgroup && cgroup_setup(td, cgroup_list, &cgroup_mnt)) goto err; @@ -1735,7 +1797,7 @@ static void *thread_main(void *data) if (!init_random_map(td)) goto err; - if (o->exec_prerun && exec_string(o, o->exec_prerun, (const char *)"prerun")) + if (o->exec_prerun && exec_string(o, o->exec_prerun, "prerun")) goto err; if (o->pre_read && !pre_read_files(td)) @@ -1746,19 +1808,21 @@ static void *thread_main(void *data) if (rate_submit_init(td, sk_out)) goto err; - set_epoch_time(td, o->log_unix_epoch); + set_epoch_time(td, o->log_unix_epoch | o->log_alternate_epoch, o->log_alternate_epoch_clock_id); fio_getrusage(&td->ru_start); memcpy(&td->bw_sample_time, &td->epoch, sizeof(td->epoch)); memcpy(&td->iops_sample_time, &td->epoch, sizeof(td->epoch)); memcpy(&td->ss.prev_time, &td->epoch, sizeof(td->epoch)); + init_thinktime(td); + if (o->ratemin[DDIR_READ] || o->ratemin[DDIR_WRITE] || o->ratemin[DDIR_TRIM]) { - memcpy(&td->lastrate[DDIR_READ], &td->bw_sample_time, + memcpy(&td->last_rate_check_time[DDIR_READ], &td->bw_sample_time, sizeof(td->bw_sample_time)); - memcpy(&td->lastrate[DDIR_WRITE], &td->bw_sample_time, + memcpy(&td->last_rate_check_time[DDIR_WRITE], &td->bw_sample_time, sizeof(td->bw_sample_time)); - memcpy(&td->lastrate[DDIR_TRIM], &td->bw_sample_time, + memcpy(&td->last_rate_check_time[DDIR_TRIM], &td->bw_sample_time, sizeof(td->bw_sample_time)); } @@ -1866,17 +1930,21 @@ static void *thread_main(void *data) * offload mode so that we don't clean up this job while * another thread is checking its io_u's for overlap */ - if (td_offload_overlap(td)) - pthread_mutex_lock(&overlap_check); + if (td_offload_overlap(td)) { + int res = pthread_mutex_lock(&overlap_check); + assert(res == 0); + } td_set_runstate(td, TD_FINISHING); - if (td_offload_overlap(td)) - pthread_mutex_unlock(&overlap_check); + if (td_offload_overlap(td)) { + res = pthread_mutex_unlock(&overlap_check); + assert(res == 0); + } update_rusage_stat(td); td->ts.total_run_time = mtime_since_now(&td->epoch); - td->ts.io_bytes[DDIR_READ] = td->io_bytes[DDIR_READ]; - td->ts.io_bytes[DDIR_WRITE] = td->io_bytes[DDIR_WRITE]; - td->ts.io_bytes[DDIR_TRIM] = td->io_bytes[DDIR_TRIM]; + for_each_rw_ddir(ddir) { + td->ts.io_bytes[ddir] = td->io_bytes[ddir]; + } if (td->o.verify_state_save && !(td->flags & TD_F_VSTATE_SAVED) && (td->o.verify != VERIFY_NONE && td_write(td))) @@ -1890,7 +1958,7 @@ static void *thread_main(void *data) rate_submit_exit(td); if (o->exec_postrun) - exec_string(o, o->exec_postrun, (const char *)"postrun"); + exec_string(o, o->exec_postrun, "postrun"); if (exitall_on_terminate || (o->exitall_error && td->error)) fio_terminate_threads(td->groupid, td->o.exit_what); @@ -1953,7 +2021,7 @@ static void reap_threads(unsigned int *nr_running, uint64_t *t_rate, for_each_td(td, i) { int flags = 0; - if (!strcmp(td->o.ioengine, "cpuio")) + if (!strcmp(td->o.ioengine, "cpuio")) cputhreads++; else realthreads++; @@ -2042,6 +2110,7 @@ reaped: done_secs += mtime_since_now(&td->epoch) / 1000; profile_td_exit(td); + flow_exit_job(td); } if (*nr_running == cputhreads && !pending && realthreads) @@ -2363,7 +2432,12 @@ reap: strerror(ret)); } else { pid_t pid; + struct fio_file **files; + void *eo; dprint(FD_PROCESS, "will fork\n"); + files = td->files; + eo = td->eo; + read_barrier(); pid = fork(); if (!pid) { int ret; @@ -2372,6 +2446,12 @@ reap: _exit(ret); } else if (i == fio_debug_jobno) *fio_debug_jobp = pid; + // freeing previously allocated memory for files + // this memory freed MUST NOT be shared between processes, only the pointer itself may be shared within TD + free(files); + free(eo); + free(fd); + fd = NULL; } dprint(FD_MUTEX, "wait on startup_sem\n"); if (fio_sem_down_timeout(startup_sem, 10000)) { @@ -2490,6 +2570,11 @@ int fio_backend(struct sk_out *sk_out) setup_log(&agg_io_log[DDIR_TRIM], &p, "agg-trim_bw.log"); } + if (init_global_dedupe_working_set_seeds()) { + log_err("fio: failed to initialize global dedupe working set\n"); + return 1; + } + startup_sem = fio_sem_init(FIO_SEM_LOCKED); if (!sk_out) is_local_backend = true; @@ -2522,8 +2607,12 @@ int fio_backend(struct sk_out *sk_out) } for_each_td(td, i) { + struct thread_stat *ts = &td->ts; + + free_clat_prio_stats(ts); steadystate_free(td); fio_options_free(td); + fio_dump_options_free(td); if (td->rusage_sem) { fio_sem_remove(td->rusage_sem); td->rusage_sem = NULL;