#include <string.h>
#include <errno.h>
#include <sys/ipc.h>
+#ifndef FIO_NO_HAVE_SHM_H
#include <sys/shm.h>
+#endif
#include <sys/types.h>
#include <sys/stat.h>
#include "verify.h"
#include "profile.h"
#include "server.h"
+#include "idletime.h"
#include "lib/getopt.h"
struct thread_data *threads = NULL;
int exitall_on_terminate = 0;
-int terse_output = 0;
+int output_format = FIO_OUTPUT_NORMAL;
int eta_print;
unsigned long long mlock_size = 0;
FILE *f_out = NULL;
.has_arg = optional_argument,
.val = 'm' | FIO_CLIENT_FLAG,
},
+ {
+ .name = (char *) "output-format",
+ .has_arg = optional_argument,
+ .val = 'F' | FIO_CLIENT_FLAG,
+ },
{
.name = (char *) "version",
.has_arg = no_argument,
.has_arg = required_argument,
.val = 'C',
},
+ {
+ .name = (char *) "cpuclock-test",
+ .has_arg = no_argument,
+ .val = 'T',
+ },
+ {
+ .name = (char *) "idle-prof",
+ .has_arg = required_argument,
+ .val = 'I',
+ },
{
.name = NULL,
},
shm_id = shmget(0, size, IPC_CREAT | 0600);
if (shm_id != -1)
break;
- if (errno != EINVAL && errno != ENOMEM) {
+ if (errno != EINVAL && errno != ENOMEM && errno != ENOSPC) {
perror("shmget");
break;
}
profile_add_hooks(td);
td->thread_number = thread_number;
+
+ if (!parent || !parent->o.group_reporting)
+ stat_number++;
+
return td;
}
ret = __setup_rate(td, DDIR_READ);
if (td->o.rate[DDIR_WRITE] || td->o.rate_iops[DDIR_WRITE])
ret |= __setup_rate(td, DDIR_WRITE);
+ if (td->o.rate[DDIR_TRIM] || td->o.rate_iops[DDIR_TRIM])
+ ret |= __setup_rate(td, DDIR_TRIM);
return ret;
}
{
return o->min_bs[DDIR_READ] == o->max_bs[DDIR_READ] &&
o->min_bs[DDIR_WRITE] == o->max_bs[DDIR_WRITE] &&
- o->min_bs[DDIR_READ] == o->min_bs[DDIR_WRITE];
+ o->min_bs[DDIR_TRIM] == o->max_bs[DDIR_TRIM] &&
+ o->min_bs[DDIR_READ] == o->min_bs[DDIR_WRITE] &&
+ o->min_bs[DDIR_READ] == o->min_bs[DDIR_TRIM];
}
/*
o->min_bs[DDIR_WRITE] = o->bs[DDIR_WRITE];
if (!o->max_bs[DDIR_WRITE])
o->max_bs[DDIR_WRITE] = o->bs[DDIR_WRITE];
+ if (!o->min_bs[DDIR_TRIM])
+ o->min_bs[DDIR_TRIM] = o->bs[DDIR_TRIM];
+ if (!o->max_bs[DDIR_TRIM])
+ o->max_bs[DDIR_TRIM] = o->bs[DDIR_TRIM];
+
o->rw_min_bs = min(o->min_bs[DDIR_READ], o->min_bs[DDIR_WRITE]);
+ o->rw_min_bs = min(o->min_bs[DDIR_TRIM], o->rw_min_bs);
/*
* For random IO, allow blockalign offset other than min_bs.
o->ba[DDIR_READ] = o->min_bs[DDIR_READ];
if (!o->ba[DDIR_WRITE] || !td_random(td))
o->ba[DDIR_WRITE] = o->min_bs[DDIR_WRITE];
+ if (!o->ba[DDIR_TRIM] || !td_random(td))
+ o->ba[DDIR_TRIM] = o->min_bs[DDIR_TRIM];
if ((o->ba[DDIR_READ] != o->min_bs[DDIR_READ] ||
- o->ba[DDIR_WRITE] != o->min_bs[DDIR_WRITE]) &&
+ o->ba[DDIR_WRITE] != o->min_bs[DDIR_WRITE] ||
+ o->ba[DDIR_TRIM] != o->min_bs[DDIR_TRIM]) &&
!o->norandommap) {
log_err("fio: Any use of blockalign= turns off randommap\n");
o->norandommap = 1;
/*
* The low water mark cannot be bigger than the iodepth
*/
- if (o->iodepth_low > o->iodepth || !o->iodepth_low) {
- /*
- * syslet work around - if the workload is sequential,
- * we want to let the queue drain all the way down to
- * avoid seeking between async threads
- */
- if (!strcmp(td->io_ops->name, "syslet-rw") && !td_random(td))
- o->iodepth_low = 1;
- else
- o->iodepth_low = o->iodepth;
- }
+ if (o->iodepth_low > o->iodepth || !o->iodepth_low)
+ o->iodepth_low = o->iodepth;
/*
* If batch number isn't set, default to the same as iodepth
if (o->open_files > o->nr_files || !o->open_files)
o->open_files = o->nr_files;
- if (((o->rate[0] + o->rate[1]) && (o->rate_iops[0] + o->rate_iops[1]))||
- ((o->ratemin[0] + o->ratemin[1]) && (o->rate_iops_min[0] +
- o->rate_iops_min[1]))) {
+ if (((o->rate[DDIR_READ] + o->rate[DDIR_WRITE] + o->rate[DDIR_TRIM]) &&
+ (o->rate_iops[DDIR_READ] + o->rate_iops[DDIR_WRITE] + o->rate_iops[DDIR_TRIM])) ||
+ ((o->ratemin[DDIR_READ] + o->ratemin[DDIR_WRITE] + o->ratemin[DDIR_TRIM]) &&
+ (o->rate_iops_min[DDIR_READ] + o->rate_iops_min[DDIR_WRITE] + o->rate_iops_min[DDIR_TRIM]))) {
log_err("fio: rate and rate_iops are mutually exclusive\n");
ret = 1;
}
- if ((o->rate[0] < o->ratemin[0]) || (o->rate[1] < o->ratemin[1]) ||
- (o->rate_iops[0] < o->rate_iops_min[0]) ||
- (o->rate_iops[1] < o->rate_iops_min[1])) {
+ if ((o->rate[DDIR_READ] < o->ratemin[DDIR_READ]) ||
+ (o->rate[DDIR_WRITE] < o->ratemin[DDIR_WRITE]) ||
+ (o->rate[DDIR_TRIM] < o->ratemin[DDIR_TRIM]) ||
+ (o->rate_iops[DDIR_READ] < o->rate_iops_min[DDIR_READ]) ||
+ (o->rate_iops[DDIR_WRITE] < o->rate_iops_min[DDIR_WRITE]) ||
+ (o->rate_iops[DDIR_TRIM] < o->rate_iops_min[DDIR_TRIM])) {
log_err("fio: minimum rate exceeds rate\n");
ret = 1;
}
}
}
-#ifndef FIO_HAVE_FDATASYNC
+#ifndef CONFIG_FDATASYNC
if (o->fdatasync_blocks) {
log_info("fio: this platform does not support fdatasync()"
" falling back to using fsync(). Use the 'fsync'"
td->o.compress_percentage = 0;
}
+ /*
+ * Using a non-uniform random distribution excludes usage of
+ * a random map
+ */
+ if (td->o.random_distribution != FIO_RAND_DIST_RANDOM)
+ td->o.norandommap = 1;
+
return ret;
}
static void td_fill_rand_seeds_os(struct thread_data *td)
{
- os_random_seed(td->rand_seeds[0], &td->bsrange_state);
- os_random_seed(td->rand_seeds[1], &td->verify_state);
- os_random_seed(td->rand_seeds[2], &td->rwmix_state);
+ os_random_seed(td->rand_seeds[FIO_RAND_BS_OFF], &td->bsrange_state);
+ os_random_seed(td->rand_seeds[FIO_RAND_VER_OFF], &td->verify_state);
+ os_random_seed(td->rand_seeds[FIO_RAND_MIX_OFF], &td->rwmix_state);
if (td->o.file_service_type == FIO_FSERVICE_RANDOM)
- os_random_seed(td->rand_seeds[3], &td->next_file_state);
+ os_random_seed(td->rand_seeds[FIO_RAND_FILE_OFF], &td->next_file_state);
- os_random_seed(td->rand_seeds[5], &td->file_size_state);
- os_random_seed(td->rand_seeds[6], &td->trim_state);
+ os_random_seed(td->rand_seeds[FIO_RAND_FILE_SIZE_OFF], &td->file_size_state);
+ os_random_seed(td->rand_seeds[FIO_RAND_TRIM_OFF], &td->trim_state);
if (!td_random(td))
return;
if (td->o.rand_repeatable)
- td->rand_seeds[4] = FIO_RANDSEED * td->thread_number;
+ td->rand_seeds[FIO_RAND_BLOCK_OFF] = FIO_RANDSEED * td->thread_number;
- os_random_seed(td->rand_seeds[4], &td->random_state);
+ os_random_seed(td->rand_seeds[FIO_RAND_BLOCK_OFF], &td->random_state);
}
static void td_fill_rand_seeds_internal(struct thread_data *td)
{
- init_rand_seed(&td->__bsrange_state, td->rand_seeds[0]);
- init_rand_seed(&td->__verify_state, td->rand_seeds[1]);
- init_rand_seed(&td->__rwmix_state, td->rand_seeds[2]);
+ init_rand_seed(&td->__bsrange_state, td->rand_seeds[FIO_RAND_BS_OFF]);
+ init_rand_seed(&td->__verify_state, td->rand_seeds[FIO_RAND_VER_OFF]);
+ init_rand_seed(&td->__rwmix_state, td->rand_seeds[FIO_RAND_MIX_OFF]);
if (td->o.file_service_type == FIO_FSERVICE_RANDOM)
- init_rand_seed(&td->__next_file_state, td->rand_seeds[3]);
+ init_rand_seed(&td->__next_file_state, td->rand_seeds[FIO_RAND_FILE_OFF]);
- init_rand_seed(&td->__file_size_state, td->rand_seeds[5]);
- init_rand_seed(&td->__trim_state, td->rand_seeds[6]);
+ init_rand_seed(&td->__file_size_state, td->rand_seeds[FIO_RAND_FILE_SIZE_OFF]);
+ init_rand_seed(&td->__trim_state, td->rand_seeds[FIO_RAND_TRIM_OFF]);
if (!td_random(td))
return;
if (td->o.rand_repeatable)
- td->rand_seeds[4] = FIO_RANDSEED * td->thread_number;
+ td->rand_seeds[FIO_RAND_BLOCK_OFF] = FIO_RANDSEED * td->thread_number;
- init_rand_seed(&td->__random_state, td->rand_seeds[4]);
+ init_rand_seed(&td->__random_state, td->rand_seeds[FIO_RAND_BLOCK_OFF]);
}
void td_fill_rand_seeds(struct thread_data *td)
else
td_fill_rand_seeds_internal(td);
- init_rand_seed(&td->buf_state, td->rand_seeds[7]);
+ init_rand_seed(&td->buf_state, td->rand_seeds[FIO_RAND_BUF_OFF]);
}
-
/*
* Initializes the ioengine configured for a job, if it has not been done so
* already.
return 0;
}
+static void init_flags(struct thread_data *td)
+{
+ struct thread_options *o = &td->o;
+
+ if (o->verify_backlog)
+ td->flags |= TD_F_VER_BACKLOG;
+ if (o->trim_backlog)
+ td->flags |= TD_F_TRIM_BACKLOG;
+ if (o->read_iolog_file)
+ td->flags |= TD_F_READ_IOLOG;
+ if (o->refill_buffers)
+ td->flags |= TD_F_REFILL_BUFFERS;
+ if (o->scramble_buffers)
+ td->flags |= TD_F_SCRAMBLE_BUFFERS;
+ if (o->verify != VERIFY_NONE)
+ td->flags |= TD_F_VER_NONE;
+}
+
+static int setup_random_seeds(struct thread_data *td)
+{
+ unsigned long seed;
+ unsigned int i;
+
+ if (!td->o.rand_repeatable)
+ return init_random_state(td, td->rand_seeds, sizeof(td->rand_seeds));
+
+ for (seed = 0x89, i = 0; i < 4; i++)
+ seed *= 0x9e370001UL;
+
+ for (i = 0; i < FIO_RAND_NR_OFFS; i++) {
+ td->rand_seeds[i] = seed;
+ seed *= 0x9e370001UL;
+ }
+
+ td_fill_rand_seeds(td);
+ return 0;
+}
+
/*
* Adds a job to the list of things todo. Sanitizes the various options
* to make sure we don't have conflicts, and initializes various
static int add_job(struct thread_data *td, const char *jobname, int job_add_num)
{
const char *ddir_str[] = { NULL, "read", "write", "rw", NULL,
- "randread", "randwrite", "randrw" };
+ "randread", "randwrite", "randrw",
+ "trim", NULL, NULL, NULL, "randtrim" };
unsigned int i;
char fname[PATH_MAX];
int numjobs, file_alloced;
if (td == &def_thread)
return 0;
+ init_flags(td);
+
/*
* if we are just dumping the output command line, don't add the job
*/
else
memcpy(td->ts.percentile_list, def_percentile_list, sizeof(def_percentile_list));
- td->ts.clat_stat[0].min_val = td->ts.clat_stat[1].min_val = ULONG_MAX;
- td->ts.slat_stat[0].min_val = td->ts.slat_stat[1].min_val = ULONG_MAX;
- td->ts.lat_stat[0].min_val = td->ts.lat_stat[1].min_val = ULONG_MAX;
- td->ts.bw_stat[0].min_val = td->ts.bw_stat[1].min_val = ULONG_MAX;
+ for (i = 0; i < DDIR_RWDIR_CNT; i++) {
+ td->ts.clat_stat[i].min_val = ULONG_MAX;
+ td->ts.slat_stat[i].min_val = ULONG_MAX;
+ td->ts.lat_stat[i].min_val = ULONG_MAX;
+ td->ts.bw_stat[i].min_val = ULONG_MAX;
+ }
td->ddir_seq_nr = td->o.ddir_seq_nr;
if ((td->o.stonewall || td->o.new_group) && prev_group_jobs) {
td->groupid = groupid;
prev_group_jobs++;
- if (init_random_state(td, td->rand_seeds, sizeof(td->rand_seeds))) {
+ if (setup_random_seeds(td)) {
td_verror(td, errno, "init_random_state");
goto err;
}
if (!td->o.name)
td->o.name = strdup(jobname);
- if (!terse_output) {
+ if (output_format == FIO_OUTPUT_NORMAL) {
if (!job_add_num) {
if (!strcmp(td->io_ops->name, "cpuio")) {
log_info("%s: ioengine=cpu, cpuload=%u,"
td->o.cpuload,
td->o.cpucycle);
} else {
- char *c1, *c2, *c3, *c4;
+ char *c1, *c2, *c3, *c4, *c5, *c6;
c1 = to_kmg(td->o.min_bs[DDIR_READ]);
c2 = to_kmg(td->o.max_bs[DDIR_READ]);
c3 = to_kmg(td->o.min_bs[DDIR_WRITE]);
c4 = to_kmg(td->o.max_bs[DDIR_WRITE]);
+ c5 = to_kmg(td->o.min_bs[DDIR_TRIM]);
+ c6 = to_kmg(td->o.max_bs[DDIR_TRIM]);
- log_info("%s: (g=%d): rw=%s, bs=%s-%s/%s-%s,"
+ log_info("%s: (g=%d): rw=%s, bs=%s-%s/%s-%s/%s-%s,"
" ioengine=%s, iodepth=%u\n",
td->o.name, td->groupid,
ddir_str[td->o.td_ddir],
- c1, c2, c3, c4,
+ c1, c2, c3, c4, c5, c6,
td->io_ops->name,
td->o.iodepth);
free(c2);
free(c3);
free(c4);
+ free(c5);
+ free(c6);
}
} else if (job_add_num == 1)
log_info("...\n");
fio_getaffinity(getpid(), &def_thread.o.cpumask);
def_thread.o.timeout = def_timeout;
-
+ def_thread.o.error_dump = 1;
/*
* fill default options
*/
"\t\t\tprocess,file,io,mem,blktrace,verify,random,parse,\n"
"\t\t\tdiskutil,job,mutex,profile,time,net\n");
printf(" --output\t\tWrite output to file\n");
- printf(" --timeout\t\tRuntime in seconds\n");
+ printf(" --runtime\t\tRuntime in seconds\n");
printf(" --latency-log\t\tGenerate per-job latency logs\n");
printf(" --bandwidth-log\tGenerate per-job bandwidth logs\n");
printf(" --minimal\t\tMinimal (terse) output\n");
- printf(" --version\t\tPrint version info and exit\n");
+ printf(" --output-format=x\tOutput format (terse,json,normal)\n");
printf(" --terse-version=x\tSet terse version output format to 'x'\n");
+ printf(" --version\t\tPrint version info and exit\n");
printf(" --help\t\tPrint this page\n");
+ printf(" --cpuclock-test\tPerform test/validation of CPU clock\n");
printf(" --cmdhelp=cmd\t\tPrint command help, \"all\" for all of"
" them\n");
printf(" --enghelp=engine\tPrint ioengine help, or list"
printf(" --server=args\t\tStart a backend fio server\n");
printf(" --daemonize=pidfile\tBackground fio server, write pid to file\n");
printf(" --client=hostname\tTalk to remote backend fio server at hostname\n");
+ printf(" --idle-prof=option\tReport cpu idleness on a system or percpu basis\n"
+ "\t\t\t(option=system,percpu) or run unit work\n"
+ "\t\t\tcalibration only (option=calibrate)\n");
printf("\nFio was written by Jens Axboe <jens.axboe@oracle.com>");
- printf("\n Jens Axboe <jaxboe@fusionio.com>\n");
+ printf("\n Jens Axboe <jaxboe@fusionio.com>\n");
}
#ifdef FIO_INC_DEBUG
f_err = f_out;
break;
case 'm':
- terse_output = 1;
+ output_format = FIO_OUTPUT_TERSE;
+ break;
+ case 'F':
+ if (!strcmp(optarg, "minimal") ||
+ !strcmp(optarg, "terse") ||
+ !strcmp(optarg, "csv"))
+ output_format = FIO_OUTPUT_TERSE;
+ else if (!strcmp(optarg, "json"))
+ output_format = FIO_OUTPUT_JSON;
+ else
+ output_format = FIO_OUTPUT_NORMAL;
break;
case 'h':
if (!cur_client) {
break;
case 'V':
terse_version = atoi(optarg);
- if (!(terse_version == 2 || terse_version == 3)) {
+ if (!(terse_version == 2 || terse_version == 3 ||
+ terse_version == 4)) {
log_err("fio: bad terse version format\n");
exit_val = 1;
do_exit++;
case 'D':
pid_file = strdup(optarg);
break;
+ case 'I':
+ if ((ret = fio_idle_prof_parse_opt(optarg))) {
+ /* exit on error and calibration only */
+ do_exit++;
+ if (ret == -1)
+ exit_val = 1;
+ }
+ break;
case 'C':
if (is_backend) {
log_err("fio: can't be both client and server\n");
optind++;
}
break;
+ case 'T':
+ do_exit++;
+ exit_val = fio_monotonic_clocktest();
+ break;
+ case '?':
+ log_err("%s: unrecognized option '%s'\n", argv[0],
+ argv[optind - 1]);
default:
do_exit++;
exit_val = 1;
fio_gtod_cpu = def_thread.o.gtod_cpu;
}
- if (!terse_output)
+ if (output_format == FIO_OUTPUT_NORMAL)
log_info("%s\n", fio_version_string);
return 0;