X-Git-Url: https://git.kernel.dk/?p=fio.git;a=blobdiff_plain;f=backend.c;h=448fc5981ada3f1724df6982b876554ab22dd10c;hp=a607134b878d4603fcb0d3e32e6273847fa2dc7a;hb=518dac097ec305d76fab3f0f45ce785a3849d8b5;hpb=ea66e04fe1a803f6a9ddf31cb999641d4396d67c diff --git a/backend.c b/backend.c index a607134b..448fc598 100644 --- a/backend.c +++ b/backend.c @@ -52,11 +52,11 @@ #include "server.h" #include "lib/getrusage.h" #include "idletime.h" +#include "err.h" static pthread_t disk_util_thread; static struct fio_mutex *disk_thread_mutex; static struct fio_mutex *startup_mutex; -static struct fio_mutex *writeout_mutex; static struct flist_head *cgroup_list; static char *cgroup_mnt; static int exit_value; @@ -182,7 +182,11 @@ static int __check_min_rate(struct thread_data *td, struct timeval *now, ratemin); return 1; } else { - rate = ((bytes - td->rate_bytes[ddir]) * 1000) / spent; + if (spent) + rate = ((bytes - td->rate_bytes[ddir]) * 1000) / spent; + else + rate = 0; + if (rate < ratemin || bytes < td->rate_bytes[ddir]) { log_err("%s: min rate %u not met, got" @@ -200,7 +204,11 @@ static int __check_min_rate(struct thread_data *td, struct timeval *now, td->o.name, rate_iops); return 1; } else { - rate = ((iops - td->rate_blocks[ddir]) * 1000) / spent; + 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: min iops rate %u not met," @@ -345,7 +353,7 @@ static inline int runtime_exceeded(struct thread_data *td, struct timeval *t) return 0; if (!td->o.timeout) return 0; - if (mtime_since(&td->epoch, t) >= td->o.timeout * 1000) + if (utime_since(&td->epoch, t) >= td->o.timeout) return 1; return 0; @@ -478,6 +486,12 @@ static void do_verify(struct thread_data *td, uint64_t verify_bytes) break; while ((io_u = get_io_u(td)) != NULL) { + if (IS_ERR(io_u)) { + io_u = NULL; + ret = FIO_Q_BUSY; + goto reap; + } + /* * We are only interested in the places where * we wrote or trimmed IOs. Turn those into @@ -574,6 +588,7 @@ sync_done: * completed io_u's first. Note that we can get BUSY even * without IO queued, if the system is resource starved. */ +reap: full = queue_full(td) || (ret == FIO_Q_BUSY && td->cur_depth); if (full || !td->o.iodepth_batch_complete) { min_events = min(td->o.iodepth_batch_complete, @@ -615,9 +630,22 @@ sync_done: dprint(FD_VERIFY, "exiting loop\n"); } +static unsigned int exceeds_number_ios(struct thread_data *td) +{ + unsigned long long number_ios; + + if (!td->o.number_ios) + return 0; + + number_ios = ddir_rw_sum(td->this_io_blocks); + number_ios += td->io_u_queued + td->io_u_in_flight; + + return number_ios >= td->o.number_ios; +} + static int io_bytes_exceeded(struct thread_data *td) { - unsigned long long bytes; + unsigned long long bytes, limit; if (td_rw(td)) bytes = td->this_io_bytes[DDIR_READ] + td->this_io_bytes[DDIR_WRITE]; @@ -628,7 +656,12 @@ static int io_bytes_exceeded(struct thread_data *td) else bytes = td->this_io_bytes[DDIR_TRIM]; - return bytes >= td->o.size; + if (td->o.io_limit) + limit = td->o.io_limit; + else + limit = td->o.size; + + return bytes >= limit || exceeds_number_ios(td); } /* @@ -692,7 +725,14 @@ static uint64_t do_io(struct thread_data *td) break; io_u = get_io_u(td); - if (!io_u) { + if (IS_ERR_OR_NULL(io_u)) { + int err = PTR_ERR(io_u); + + io_u = NULL; + if (err == -EBUSY) { + ret = FIO_Q_BUSY; + goto reap; + } if (td->o.latency_target) goto reap; break; @@ -740,16 +780,21 @@ static uint64_t do_io(struct thread_data *td) case FIO_Q_COMPLETED: if (io_u->error) { ret = -io_u->error; + unlog_io_piece(td, io_u); clear_io_u(td, io_u); } else if (io_u->resid) { int bytes = io_u->xfer_buflen - io_u->resid; struct fio_file *f = io_u->file; bytes_issued += bytes; + + trim_io_piece(td, io_u); + /* * zero read, fail */ if (!bytes) { + unlog_io_piece(td, io_u); td_verror(td, EIO, "full resid"); put_io_u(td, io_u); break; @@ -790,6 +835,7 @@ sync_done: bytes_issued += io_u->xfer_buflen; break; case FIO_Q_BUSY: + unlog_io_piece(td, io_u); requeue_io_u(td, &io_u); ret2 = td_io_commit(td); if (ret2 < 0) @@ -1083,12 +1129,14 @@ static int switch_ioscheduler(struct thread_data *td) /* * Read back and check that the selected scheduler is now the default. */ - ret = fread(tmp, 1, sizeof(tmp), f); + ret = fread(tmp, sizeof(tmp), 1, f); if (ferror(f) || ret < 0) { td_verror(td, errno, "fread"); fclose(f); return 1; } + tmp[sizeof(tmp) - 1] = '\0'; + sprintf(tmp2, "[%s]", td->o.ioscheduler); if (!strstr(tmp, tmp2)) { @@ -1104,6 +1152,8 @@ static int switch_ioscheduler(struct thread_data *td) static int keep_running(struct thread_data *td) { + unsigned long long limit; + if (td->done) return 0; if (td->o.time_based) @@ -1112,18 +1162,28 @@ static int keep_running(struct thread_data *td) td->o.loops--; return 1; } + if (exceeds_number_ios(td)) + return 0; - if (td->o.size != -1ULL && ddir_rw_sum(td->io_bytes) < td->o.size) { + if (td->o.io_limit) + limit = td->o.io_limit; + else + limit = td->o.size; + + if (limit != -1ULL && ddir_rw_sum(td->io_bytes) < limit) { uint64_t diff; /* * If the difference is less than the minimum IO size, we * are done. */ - diff = td->o.size - ddir_rw_sum(td->io_bytes); + diff = limit - ddir_rw_sum(td->io_bytes); if (diff < td_max_bs(td)) return 0; + if (fio_files_done(td)) + return 0; + return 1; } @@ -1178,6 +1238,12 @@ static uint64_t do_dry_run(struct thread_data *td) td->ts.total_io_u[io_u->ddir]++; } + if (td_write(td) && io_u->ddir == DDIR_WRITE && + td->o.do_verify && + td->o.verify != VERIFY_NONE && + !td->o.experimental_verify) + log_io_piece(td, io_u); + ret = io_u_sync_complete(td, io_u, bytes_done); (void) ret; } @@ -1204,11 +1270,6 @@ static void *thread_main(void *data) } else td->pid = gettid(); - /* - * fio_time_init() may not have been called yet if running as a server - */ - fio_time_init(); - fio_local_clock_init(o->use_thread); dprint(FD_PROCESS, "jobs pid=%d started\n", (int) td->pid); @@ -1260,6 +1321,15 @@ static void *thread_main(void *data) * allocations. */ if (o->cpumask_set) { + if (o->cpus_allowed_policy == FIO_CPUS_SPLIT) { + ret = fio_cpus_split(&o->cpumask, td->thread_number - 1); + if (!ret) { + log_err("fio: no CPUs set\n"); + log_err("fio: Try increasing number of available CPUs\n"); + td_verror(td, EINVAL, "cpus_split"); + goto err; + } + } ret = fio_setaffinity(td->pid, o->cpumask); if (ret == -1) { td_verror(td, errno, "cpu_set_affinity"); @@ -1270,6 +1340,7 @@ static void *thread_main(void *data) #ifdef CONFIG_LIBNUMA /* numa node setup */ if (o->numa_cpumask_set || o->numa_memmask_set) { + struct bitmask *mask; int ret; if (numa_available() < 0) { @@ -1278,7 +1349,9 @@ static void *thread_main(void *data) } if (o->numa_cpumask_set) { - ret = numa_run_on_node_mask(o->numa_cpunodesmask); + mask = numa_parse_nodestring(o->numa_cpunodes); + ret = numa_run_on_node_mask(mask); + numa_free_nodemask(mask); if (ret == -1) { td_verror(td, errno, \ "numa_run_on_node_mask failed\n"); @@ -1288,12 +1361,16 @@ static void *thread_main(void *data) if (o->numa_memmask_set) { + mask = NULL; + if (o->numa_memnodes) + mask = numa_parse_nodestring(o->numa_memnodes); + switch (o->numa_mem_mode) { case MPOL_INTERLEAVE: - numa_set_interleave_mask(o->numa_memnodesmask); + numa_set_interleave_mask(mask); break; case MPOL_BIND: - numa_set_membind(o->numa_memnodesmask); + numa_set_membind(mask); break; case MPOL_LOCAL: numa_set_localalloc(); @@ -1306,6 +1383,9 @@ static void *thread_main(void *data) break; } + if (mask) + numa_free_nodemask(mask); + } } #endif @@ -1442,44 +1522,8 @@ static void *thread_main(void *data) fio_unpin_memory(td); - fio_mutex_down(writeout_mutex); - if (td->bw_log) { - if (o->bw_log_file) { - finish_log_named(td, td->bw_log, - o->bw_log_file, "bw"); - } else - finish_log(td, td->bw_log, "bw"); - } - if (td->lat_log) { - if (o->lat_log_file) { - finish_log_named(td, td->lat_log, - o->lat_log_file, "lat"); - } else - finish_log(td, td->lat_log, "lat"); - } - if (td->slat_log) { - if (o->lat_log_file) { - finish_log_named(td, td->slat_log, - o->lat_log_file, "slat"); - } else - finish_log(td, td->slat_log, "slat"); - } - if (td->clat_log) { - if (o->lat_log_file) { - finish_log_named(td, td->clat_log, - o->lat_log_file, "clat"); - } else - finish_log(td, td->clat_log, "clat"); - } - if (td->iops_log) { - if (o->iops_log_file) { - finish_log_named(td, td->iops_log, - o->iops_log_file, "iops"); - } else - finish_log(td, td->iops_log, "iops"); - } + fio_writeout_logs(td); - fio_mutex_up(writeout_mutex); if (o->exec_postrun) exec_string(o, o->exec_postrun, (const char *)"postrun"); @@ -1531,7 +1575,7 @@ static int fork_main(int shmid, int offset) struct thread_data *td; void *data, *ret; -#ifndef __hpux +#if !defined(__hpux) && !defined(CONFIG_NO_SHM) data = shmat(shmid, NULL, 0); if (data == (void *) -1) { int __err = errno; @@ -1664,8 +1708,8 @@ static void do_usleep(unsigned int usecs) static void run_threads(void) { struct thread_data *td; - unsigned long spent; unsigned int i, todo, nr_running, m_rate, t_rate, nr_started; + uint64_t spent; if (fio_gtod_offload && fio_start_gtod_thread()) return; @@ -1763,9 +1807,9 @@ static void run_threads(void) } if (td->o.start_delay) { - spent = mtime_since_genesis(); + spent = utime_since_genesis(); - if (td->o.start_delay * 1000 > spent) + if (td->o.start_delay > spent) continue; } @@ -1976,17 +2020,14 @@ int fio_backend(void) return 0; if (write_bw_log) { - setup_log(&agg_io_log[DDIR_READ], 0, IO_LOG_TYPE_BW); - setup_log(&agg_io_log[DDIR_WRITE], 0, IO_LOG_TYPE_BW); - setup_log(&agg_io_log[DDIR_TRIM], 0, IO_LOG_TYPE_BW); + setup_log(&agg_io_log[DDIR_READ], 0, IO_LOG_TYPE_BW, 0, "agg-read_bw.log"); + setup_log(&agg_io_log[DDIR_WRITE], 0, IO_LOG_TYPE_BW, 0, "agg-write_bw.log"); + setup_log(&agg_io_log[DDIR_TRIM], 0, IO_LOG_TYPE_BW, 0, "agg-trim_bw.log"); } startup_mutex = fio_mutex_init(FIO_MUTEX_LOCKED); if (startup_mutex == NULL) return 1; - writeout_mutex = fio_mutex_init(FIO_MUTEX_UNLOCKED); - if (writeout_mutex == NULL) - return 1; set_genesis_time(); stat_init(); @@ -2000,11 +2041,14 @@ int fio_backend(void) if (!fio_abort) { show_run_stats(); if (write_bw_log) { - __finish_log(agg_io_log[DDIR_READ], "agg-read_bw.log"); - __finish_log(agg_io_log[DDIR_WRITE], - "agg-write_bw.log"); - __finish_log(agg_io_log[DDIR_TRIM], - "agg-write_bw.log"); + int i; + + for (i = 0; i < DDIR_RWDIR_CNT; i++) { + struct io_log *log = agg_io_log[i]; + + __finish_log(log); + free_log(log); + } } } @@ -2017,7 +2061,6 @@ int fio_backend(void) sfree(cgroup_mnt); fio_mutex_remove(startup_mutex); - fio_mutex_remove(writeout_mutex); fio_mutex_remove(disk_thread_mutex); stat_exit(); return exit_value;