*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software
- * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
*
*/
#include <unistd.h>
-#include <fcntl.h>
#include <string.h>
-#include <limits.h>
#include <signal.h>
-#include <time.h>
-#include <locale.h>
#include <assert.h>
-#include <time.h>
#include <inttypes.h>
#include <sys/stat.h>
#include <sys/wait.h>
-#include <sys/ipc.h>
-#include <sys/mman.h>
#include <math.h>
+#include <pthread.h>
#include "fio.h"
-#ifndef FIO_NO_HAVE_SHM_H
-#include <sys/shm.h>
-#endif
-#include "hash.h"
#include "smalloc.h"
#include "verify.h"
-#include "trim.h"
#include "diskutil.h"
#include "cgroup.h"
#include "profile.h"
#include "workqueue.h"
#include "lib/mountcheck.h"
#include "rate-submit.h"
+#include "helper_thread.h"
+#include "pshared.h"
+#include "zone-dist.h"
-static pthread_t helper_thread;
-static pthread_mutex_t helper_lock;
-pthread_cond_t helper_cond;
-int helper_do_stat = 0;
-
-static struct fio_mutex *startup_mutex;
+static struct fio_sem *startup_sem;
static struct flist_head *cgroup_list;
-static char *cgroup_mnt;
+static struct cgroup_mnt *cgroup_mnt;
static int exit_value;
-static volatile int fio_abort;
+static volatile bool fio_abort;
static unsigned int nr_process = 0;
static unsigned int nr_thread = 0;
int shm_id = 0;
int temp_stall_ts;
unsigned long done_secs = 0;
-volatile int helper_exit = 0;
-
-#define PAGE_ALIGN(buf) \
- (char *) (((uintptr_t) (buf) + page_mask) & ~page_mask)
+pthread_mutex_t overlap_check = PTHREAD_MUTEX_INITIALIZER;
#define JOB_START_TIMEOUT (5 * 1000)
/*
* Check if we are above the minimum rate given.
*/
-static bool __check_min_rate(struct thread_data *td, struct timeval *now,
+static bool __check_min_rate(struct thread_data *td, struct timespec *now,
enum fio_ddir ddir)
{
unsigned long long bytes = 0;
* check bandwidth specified rate
*/
if (bytes < td->rate_bytes[ddir]) {
- log_err("%s: min rate %u not met\n", td->o.name,
- ratemin);
+ log_err("%s: rate_min=%uB/s not met, only transferred %lluB\n",
+ td->o.name, ratemin, bytes);
return true;
} else {
if (spent)
if (rate < ratemin ||
bytes < td->rate_bytes[ddir]) {
- log_err("%s: min rate %u not met, got"
- " %luKB/sec\n", td->o.name,
- ratemin, rate);
+ log_err("%s: rate_min=%uB/s not met, got %luB/s\n",
+ td->o.name, ratemin, rate);
return true;
}
}
* checks iops specified rate
*/
if (iops < rate_iops) {
- log_err("%s: min iops rate %u not met\n",
- td->o.name, rate_iops);
+ log_err("%s: rate_iops_min=%u not met, only performed %lu IOs\n",
+ td->o.name, rate_iops, iops);
return true;
} else {
if (spent)
if (rate < rate_iops_min ||
iops < td->rate_blocks[ddir]) {
- log_err("%s: min iops rate %u not met,"
- " got %lu\n", td->o.name,
- rate_iops_min, rate);
+ log_err("%s: rate_iops_min=%u not met, got %lu IOPS\n",
+ td->o.name, rate_iops_min, rate);
return true;
}
}
return false;
}
-static bool check_min_rate(struct thread_data *td, struct timeval *now)
+static bool check_min_rate(struct thread_data *td, struct timespec *now)
{
bool ret = false;
{
int r;
+ if (td->error)
+ return;
+
/*
* get immediately available events, if any
*/
static bool fio_io_sync(struct thread_data *td, struct fio_file *f)
{
struct io_u *io_u = __get_io_u(td);
- int ret;
+ enum fio_q_status ret;
if (!io_u)
return true;
requeue:
ret = td_io_queue(td, io_u);
- if (ret < 0) {
- td_verror(td, io_u->error, "td_io_queue");
- put_io_u(td, io_u);
- return true;
- } else if (ret == FIO_Q_QUEUED) {
+ switch (ret) {
+ case FIO_Q_QUEUED:
+ td_io_commit(td);
if (io_u_queued_complete(td, 1) < 0)
return true;
- } else if (ret == FIO_Q_COMPLETED) {
+ break;
+ case FIO_Q_COMPLETED:
if (io_u->error) {
td_verror(td, io_u->error, "td_io_queue");
return true;
if (io_u_sync_complete(td, io_u) < 0)
return true;
- } else if (ret == FIO_Q_BUSY) {
- if (td_io_commit(td))
- return true;
+ break;
+ case FIO_Q_BUSY:
+ td_io_commit(td);
goto requeue;
}
return ret;
}
-static inline void __update_tv_cache(struct thread_data *td)
+static inline void __update_ts_cache(struct thread_data *td)
{
- fio_gettime(&td->tv_cache, NULL);
+ fio_gettime(&td->ts_cache, NULL);
}
-static inline void update_tv_cache(struct thread_data *td)
+static inline void update_ts_cache(struct thread_data *td)
{
- if ((++td->tv_cache_nr & td->tv_cache_mask) == td->tv_cache_mask)
- __update_tv_cache(td);
+ if ((++td->ts_cache_nr & td->ts_cache_mask) == td->ts_cache_mask)
+ __update_ts_cache(td);
}
-static inline bool runtime_exceeded(struct thread_data *td, struct timeval *t)
+static inline bool runtime_exceeded(struct thread_data *td, struct timespec *t)
{
if (in_ramp_time(td))
return false;
if (td->update_rusage) {
td->update_rusage = 0;
update_rusage_stat(td);
- fio_mutex_up(td->rusage_sem);
+ fio_sem_up(td->rusage_sem);
}
}
-static int wait_for_completions(struct thread_data *td, struct timeval *time)
+static int wait_for_completions(struct thread_data *td, struct timespec *time)
{
const int full = queue_full(td);
int min_evts = 0;
int ret;
+ if (td->flags & TD_F_REGROW_LOGS)
+ return io_u_quiesce(td);
+
/*
* if the queue is full, we MUST reap at least 1 event
*/
if ((full && !min_evts) || !td->o.iodepth_batch_complete_min)
min_evts = 1;
- if (time && (__should_check_rate(td, DDIR_READ) ||
- __should_check_rate(td, DDIR_WRITE) ||
- __should_check_rate(td, DDIR_TRIM)))
+ if (time && __should_check_rate(td))
fio_gettime(time, NULL);
do {
int io_queue_event(struct thread_data *td, struct io_u *io_u, int *ret,
enum fio_ddir ddir, uint64_t *bytes_issued, int from_verify,
- struct timeval *comp_time)
+ struct timespec *comp_time)
{
- int ret2;
-
switch (*ret) {
case FIO_Q_COMPLETED:
if (io_u->error) {
*ret = -io_u->error;
clear_io_u(td, io_u);
} else if (io_u->resid) {
- int bytes = io_u->xfer_buflen - io_u->resid;
+ long long bytes = io_u->xfer_buflen - io_u->resid;
struct fio_file *f = io_u->file;
if (bytes_issued)
*bytes_issued += bytes;
if (!from_verify)
- trim_io_piece(td, io_u);
+ trim_io_piece(io_u);
/*
* zero read, fail
if (ddir_rw(io_u->ddir))
td->ts.short_io_u[io_u->ddir]++;
- f = io_u->file;
if (io_u->offset == f->real_file_size)
goto sync_done;
requeue_io_u(td, &io_u);
} else {
sync_done:
- if (comp_time && (__should_check_rate(td, DDIR_READ) ||
- __should_check_rate(td, DDIR_WRITE) ||
- __should_check_rate(td, DDIR_TRIM)))
+ if (comp_time && __should_check_rate(td))
fio_gettime(comp_time, NULL);
*ret = io_u_sync_complete(td, io_u);
break;
}
+ if (td->flags & TD_F_REGROW_LOGS)
+ regrow_logs(td);
+
/*
* when doing I/O (not when verifying),
* check for any errors that are to be ignored
if (!from_verify)
unlog_io_piece(td, io_u);
requeue_io_u(td, &io_u);
- ret2 = td_io_commit(td);
- if (ret2 < 0)
- *ret = ret2;
+ td_io_commit(td);
break;
default:
assert(*ret < 0);
return !td->o.iodepth_batch_complete_min &&
!td->o.iodepth_batch_complete_max;
}
+/*
+ * Unlinks files from thread data fio_file structure
+ */
+static int unlink_all_files(struct thread_data *td)
+{
+ struct fio_file *f;
+ unsigned int i;
+ int ret = 0;
+
+ for_each_file(td, f, i) {
+ if (f->filetype != FIO_TYPE_FILE)
+ continue;
+ ret = td_io_unlink_file(td, f);
+ if (ret)
+ break;
+ }
+
+ if (ret)
+ td_verror(td, ret, "unlink_all_files");
+
+ return ret;
+}
+
+/*
+ * Check if io_u will overlap an in-flight IO in the queue
+ */
+bool in_flight_overlap(struct io_u_queue *q, struct io_u *io_u)
+{
+ bool overlap;
+ struct io_u *check_io_u;
+ unsigned long long x1, x2, y1, y2;
+ int i;
+
+ x1 = io_u->offset;
+ x2 = io_u->offset + io_u->buflen;
+ overlap = false;
+ io_u_qiter(q, check_io_u, i) {
+ if (check_io_u->flags & IO_U_F_FLIGHT) {
+ y1 = check_io_u->offset;
+ y2 = check_io_u->offset + check_io_u->buflen;
+
+ if (x1 < y2 && y1 < x2) {
+ overlap = true;
+ dprint(FD_IO, "in-flight overlap: %llu/%llu, %llu/%llu\n",
+ x1, io_u->buflen,
+ y1, check_io_u->buflen);
+ break;
+ }
+ }
+ }
+
+ return overlap;
+}
+
+static enum fio_q_status io_u_submit(struct thread_data *td, struct io_u *io_u)
+{
+ /*
+ * Check for overlap if the user asked us to, and we have
+ * at least one IO in flight besides this one.
+ */
+ if (td->o.serialize_overlap && td->cur_depth > 1 &&
+ in_flight_overlap(&td->io_u_all, io_u))
+ return FIO_Q_BUSY;
+
+ return td_io_queue(td, io_u);
+}
/*
* The main verify engine. Runs over the writes we previously submitted,
if (td->error)
return;
+ /*
+ * verify_state needs to be reset before verification
+ * proceeds so that expected random seeds match actual
+ * random seeds in headers. The main loop will reset
+ * all random number generators if randrepeat is set.
+ */
+ if (!td->o.rand_repeatable)
+ td_fill_verify_state_seed(td);
+
td_set_runstate(td, TD_VERIFYING);
io_u = NULL;
enum fio_ddir ddir;
int full;
- update_tv_cache(td);
+ update_ts_cache(td);
check_update_rusage(td);
- if (runtime_exceeded(td, &td->tv_cache)) {
- __update_tv_cache(td);
- if (runtime_exceeded(td, &td->tv_cache)) {
+ if (runtime_exceeded(td, &td->ts_cache)) {
+ __update_ts_cache(td);
+ if (runtime_exceeded(td, &td->ts_cache)) {
fio_mark_td_terminate(td);
break;
}
break;
while ((io_u = get_io_u(td)) != NULL) {
- if (IS_ERR(io_u)) {
+ if (IS_ERR_OR_NULL(io_u)) {
io_u = NULL;
ret = FIO_Q_BUSY;
goto reap;
continue;
} else if (io_u->ddir == DDIR_TRIM) {
io_u->ddir = DDIR_READ;
- io_u_set(io_u, IO_U_F_TRIMMED);
+ io_u_set(td, io_u, IO_U_F_TRIMMED);
break;
} else if (io_u->ddir == DDIR_WRITE) {
io_u->ddir = DDIR_READ;
+ populate_verify_io_u(td, io_u);
break;
} else {
put_io_u(td, io_u);
if (!td->o.disable_slat)
fio_gettime(&io_u->start_time, NULL);
- ret = td_io_queue(td, io_u);
+ ret = io_u_submit(td, io_u);
if (io_queue_event(td, io_u, &ret, ddir, NULL, 1, NULL))
break;
return number_ios >= (td->o.number_ios * td->loops);
}
-static bool io_issue_bytes_exceeded(struct thread_data *td)
+static bool io_bytes_exceeded(struct thread_data *td, uint64_t *this_bytes)
{
unsigned long long bytes, limit;
if (td_rw(td))
- bytes = td->io_issue_bytes[DDIR_READ] + td->io_issue_bytes[DDIR_WRITE];
+ bytes = this_bytes[DDIR_READ] + this_bytes[DDIR_WRITE];
else if (td_write(td))
- bytes = td->io_issue_bytes[DDIR_WRITE];
+ bytes = this_bytes[DDIR_WRITE];
else if (td_read(td))
- bytes = td->io_issue_bytes[DDIR_READ];
+ bytes = this_bytes[DDIR_READ];
else
- bytes = td->io_issue_bytes[DDIR_TRIM];
+ bytes = this_bytes[DDIR_TRIM];
- if (td->o.io_limit)
- limit = td->o.io_limit;
+ if (td->o.io_size)
+ limit = td->o.io_size;
else
limit = td->o.size;
return bytes >= limit || exceeds_number_ios(td);
}
-static bool io_complete_bytes_exceeded(struct thread_data *td)
+static bool io_issue_bytes_exceeded(struct thread_data *td)
{
- unsigned long long bytes, limit;
-
- if (td_rw(td))
- bytes = td->this_io_bytes[DDIR_READ] + td->this_io_bytes[DDIR_WRITE];
- else if (td_write(td))
- bytes = td->this_io_bytes[DDIR_WRITE];
- else if (td_read(td))
- bytes = td->this_io_bytes[DDIR_READ];
- else
- bytes = td->this_io_bytes[DDIR_TRIM];
-
- if (td->o.io_limit)
- limit = td->o.io_limit;
- else
- limit = td->o.size;
+ return io_bytes_exceeded(td, td->io_issue_bytes);
+}
- limit *= td->loops;
- return bytes >= limit || exceeds_number_ios(td);
+static bool io_complete_bytes_exceeded(struct thread_data *td)
+{
+ return io_bytes_exceeded(td, td->this_io_bytes);
}
/*
*/
static long long usec_for_io(struct thread_data *td, enum fio_ddir ddir)
{
- uint64_t secs, remainder, bps, bytes, iops;
+ uint64_t bps = td->rate_bps[ddir];
assert(!(td->flags & TD_F_CHILD));
- bytes = td->rate_io_issue_bytes[ddir];
- bps = td->rate_bps[ddir];
if (td->o.rate_process == RATE_PROCESS_POISSON) {
- uint64_t val;
+ uint64_t val, iops;
+
iops = bps / td->o.bs[ddir];
val = (int64_t) (1000000 / iops) *
- -logf(__rand_0_1(&td->poisson_state));
+ -logf(__rand_0_1(&td->poisson_state[ddir]));
if (val) {
- dprint(FD_RATE, "poisson rate iops=%llu\n",
- (unsigned long long) 1000000 / val);
+ dprint(FD_RATE, "poisson rate iops=%llu, ddir=%d\n",
+ (unsigned long long) 1000000 / val,
+ ddir);
}
- td->last_usec += val;
- return td->last_usec;
+ td->last_usec[ddir] += val;
+ return td->last_usec[ddir];
} else if (bps) {
- secs = bytes / bps;
- remainder = bytes % bps;
+ uint64_t bytes = td->rate_io_issue_bytes[ddir];
+ uint64_t secs = bytes / bps;
+ uint64_t remainder = bytes % bps;
+
return remainder * 1000000 / bps + secs * 1000000;
}
return 0;
}
+static void handle_thinktime(struct thread_data *td, enum fio_ddir ddir)
+{
+ unsigned long long b;
+ uint64_t total;
+ int left;
+
+ b = ddir_rw_sum(td->io_blocks);
+ if (b % td->o.thinktime_blocks)
+ return;
+
+ io_u_quiesce(td);
+
+ total = 0;
+ if (td->o.thinktime_spin)
+ total = usec_spin(td->o.thinktime_spin);
+
+ left = td->o.thinktime - total;
+ if (left)
+ total += usec_sleep(td, left);
+
+ /*
+ * If we're ignoring thinktime for the rate, add the number of bytes
+ * we would have done while sleeping, minus one block to ensure we
+ * start issuing immediately after the sleep.
+ */
+ if (total && td->rate_bps[ddir] && td->o.rate_ign_think) {
+ uint64_t missed = (td->rate_bps[ddir] * total) / 1000000ULL;
+ uint64_t bs = td->o.min_bs[ddir];
+ uint64_t usperop = bs * 1000000ULL / td->rate_bps[ddir];
+ uint64_t over;
+
+ if (usperop <= total)
+ over = bs;
+ else
+ over = (usperop - total) / usperop * -bs;
+
+ td->rate_io_issue_bytes[ddir] += (missed - over);
+ /* adjust for rate_process=poisson */
+ td->last_usec[ddir] += total;
+ }
+}
+
/*
* Main IO worker function. It retrieves io_u's to process and queues
* and reaps them, checking for rate and errors along the way.
total_bytes = td->o.size;
/*
- * Allow random overwrite workloads to write up to io_limit
+ * Allow random overwrite workloads to write up to io_size
* before starting verification phase as 'size' doesn't apply.
*/
if (td_write(td) && td_random(td) && td->o.norandommap)
- total_bytes = max(total_bytes, (uint64_t) td->o.io_limit);
+ total_bytes = max(total_bytes, (uint64_t) td->o.io_size);
/*
* If verify_backlog is enabled, we'll run the verify in this
* handler as well. For that case, we may need up to twice the
while ((td->o.read_iolog_file && !flist_empty(&td->io_log_list)) ||
(!flist_empty(&td->trim_list)) || !io_issue_bytes_exceeded(td) ||
td->o.time_based) {
- struct timeval comp_time;
+ struct timespec comp_time;
struct io_u *io_u;
int full;
enum fio_ddir ddir;
if (td->terminate || td->done)
break;
- update_tv_cache(td);
+ update_ts_cache(td);
- if (runtime_exceeded(td, &td->tv_cache)) {
- __update_tv_cache(td);
- if (runtime_exceeded(td, &td->tv_cache)) {
+ if (runtime_exceeded(td, &td->ts_cache)) {
+ __update_ts_cache(td);
+ if (runtime_exceeded(td, &td->ts_cache)) {
fio_mark_td_terminate(td);
break;
}
if (flow_threshold_exceeded(td))
continue;
- if (!td->o.time_based && bytes_issued >= total_bytes)
+ /*
+ * Break if we exceeded the bytes. The exception is time
+ * based runs, but we still need to break out of the loop
+ * for those to run verification, if enabled.
+ * Jobs read from iolog do not use this stop condition.
+ */
+ if (bytes_issued >= total_bytes &&
+ !td->o.read_iolog_file &&
+ (!td->o.time_based ||
+ (td->o.time_based && td->o.verify != VERIFY_NONE)))
break;
io_u = get_io_u(td);
int err = PTR_ERR(io_u);
io_u = NULL;
+ ddir = DDIR_INVAL;
if (err == -EBUSY) {
ret = FIO_Q_BUSY;
goto reap;
break;
}
+ if (io_u->ddir == DDIR_WRITE && td->flags & TD_F_DO_VERIFY)
+ populate_verify_io_u(td, io_u);
+
ddir = io_u->ddir;
/*
log_io_piece(td, io_u);
if (td->o.io_submit_mode == IO_MODE_OFFLOAD) {
- const unsigned long blen = io_u->xfer_buflen;
- const enum fio_ddir ddir = acct_ddir(io_u);
+ const unsigned long long blen = io_u->xfer_buflen;
+ const enum fio_ddir __ddir = acct_ddir(io_u);
if (td->error)
break;
workqueue_enqueue(&td->io_wq, &io_u->work);
ret = FIO_Q_QUEUED;
- if (ddir_rw(ddir)) {
- td->io_issues[ddir]++;
- td->io_issue_bytes[ddir] += blen;
- td->rate_io_issue_bytes[ddir] += blen;
+ if (ddir_rw(__ddir)) {
+ td->io_issues[__ddir]++;
+ td->io_issue_bytes[__ddir] += blen;
+ td->rate_io_issue_bytes[__ddir] += blen;
}
if (should_check_rate(td))
- td->rate_next_io_time[ddir] = usec_for_io(td, ddir);
+ td->rate_next_io_time[__ddir] = usec_for_io(td, __ddir);
} else {
- ret = td_io_queue(td, io_u);
+ ret = io_u_submit(td, io_u);
if (should_check_rate(td))
td->rate_next_io_time[ddir] = usec_for_io(td, ddir);
if (ret < 0)
break;
if (!ddir_rw_sum(td->bytes_done) &&
- !(td->io_ops->flags & FIO_NOIO))
+ !td_ioengine_flagged(td, FIO_NOIO))
continue;
if (!in_ramp_time(td) && should_check_rate(td)) {
if (!in_ramp_time(td) && td->o.latency_target)
lat_target_check(td);
- if (td->o.thinktime) {
- unsigned long long b;
-
- b = ddir_rw_sum(td->io_blocks);
- if (!(b % td->o.thinktime_blocks)) {
- int left;
-
- io_u_quiesce(td);
-
- if (td->o.thinktime_spin)
- usec_spin(td->o.thinktime_spin);
-
- left = td->o.thinktime - td->o.thinktime_spin;
- if (left)
- usec_sleep(td, left);
- }
- }
+ if (ddir_rw(ddir) && td->o.thinktime)
+ handle_thinktime(td, ddir);
}
check_update_rusage(td);
bytes_done[i] = td->bytes_done[i] - bytes_done[i];
}
+static void free_file_completion_logging(struct thread_data *td)
+{
+ struct fio_file *f;
+ unsigned int i;
+
+ for_each_file(td, f, i) {
+ if (!f->last_write_comp)
+ break;
+ sfree(f->last_write_comp);
+ }
+}
+
+static int init_file_completion_logging(struct thread_data *td,
+ unsigned int depth)
+{
+ struct fio_file *f;
+ unsigned int i;
+
+ if (td->o.verify == VERIFY_NONE || !td->o.verify_state_save)
+ return 0;
+
+ for_each_file(td, f, i) {
+ f->last_write_comp = scalloc(depth, sizeof(uint64_t));
+ if (!f->last_write_comp)
+ goto cleanup;
+ }
+
+ return 0;
+
+cleanup:
+ free_file_completion_logging(td);
+ log_err("fio: failed to alloc write comp data\n");
+ return 1;
+}
+
static void cleanup_io_u(struct thread_data *td)
{
struct io_u *io_u;
if (td->io_ops->io_u_free)
td->io_ops->io_u_free(td, io_u);
- fio_memfree(io_u, sizeof(*io_u));
+ fio_memfree(io_u, sizeof(*io_u), td_offload_overlap(td));
}
free_io_mem(td);
io_u_rexit(&td->io_u_requeues);
- io_u_qexit(&td->io_u_freelist);
- io_u_qexit(&td->io_u_all);
+ io_u_qexit(&td->io_u_freelist, false);
+ io_u_qexit(&td->io_u_all, td_offload_overlap(td));
- if (td->last_write_comp)
- sfree(td->last_write_comp);
+ free_file_completion_logging(td);
}
static int init_io_u(struct thread_data *td)
{
struct io_u *io_u;
- unsigned int max_bs, min_write;
int cl_align, i, max_units;
- int data_xfer = 1, err;
- char *p;
+ int err;
max_units = td->o.iodepth;
- max_bs = td_max_bs(td);
- min_write = td->o.min_bs[DDIR_WRITE];
- td->orig_buffer_size = (unsigned long long) max_bs
- * (unsigned long long) max_units;
-
- if ((td->io_ops->flags & FIO_NOIO) || !(td_read(td) || td_write(td)))
- data_xfer = 0;
err = 0;
- err += io_u_rinit(&td->io_u_requeues, td->o.iodepth);
- err += io_u_qinit(&td->io_u_freelist, td->o.iodepth);
- err += io_u_qinit(&td->io_u_all, td->o.iodepth);
+ err += !io_u_rinit(&td->io_u_requeues, td->o.iodepth);
+ err += !io_u_qinit(&td->io_u_freelist, td->o.iodepth, false);
+ err += !io_u_qinit(&td->io_u_all, td->o.iodepth, td_offload_overlap(td));
if (err) {
log_err("fio: failed setting up IO queues\n");
return 1;
}
+ cl_align = os_cache_line_size();
+
+ for (i = 0; i < max_units; i++) {
+ void *ptr;
+
+ if (td->terminate)
+ return 1;
+
+ ptr = fio_memalign(cl_align, sizeof(*io_u), td_offload_overlap(td));
+ if (!ptr) {
+ log_err("fio: unable to allocate aligned memory\n");
+ break;
+ }
+
+ io_u = ptr;
+ memset(io_u, 0, sizeof(*io_u));
+ INIT_FLIST_HEAD(&io_u->verify_list);
+ dprint(FD_MEM, "io_u alloc %p, index %u\n", io_u, i);
+
+ io_u->index = i;
+ io_u->flags = IO_U_F_FREE;
+ io_u_qpush(&td->io_u_freelist, io_u);
+
+ /*
+ * io_u never leaves this stack, used for iteration of all
+ * io_u buffers.
+ */
+ io_u_qpush(&td->io_u_all, io_u);
+
+ if (td->io_ops->io_u_init) {
+ int ret = td->io_ops->io_u_init(td, io_u);
+
+ if (ret) {
+ log_err("fio: failed to init engine data: %d\n", ret);
+ return 1;
+ }
+ }
+ }
+
+ init_io_u_buffers(td);
+
+ if (init_file_completion_logging(td, max_units))
+ return 1;
+
+ return 0;
+}
+
+int init_io_u_buffers(struct thread_data *td)
+{
+ struct io_u *io_u;
+ unsigned long long max_bs, min_write;
+ int i, max_units;
+ int data_xfer = 1;
+ char *p;
+
+ max_units = td->o.iodepth;
+ max_bs = td_max_bs(td);
+ min_write = td->o.min_bs[DDIR_WRITE];
+ td->orig_buffer_size = (unsigned long long) max_bs
+ * (unsigned long long) max_units;
+
+ if (td_ioengine_flagged(td, FIO_NOIO) || !(td_read(td) || td_write(td)))
+ data_xfer = 0;
+
/*
* if we may later need to do address alignment, then add any
* possible adjustment here so that we don't cause a buffer
* lucky and the allocator gives us an aligned address.
*/
if (td->o.odirect || td->o.mem_align || td->o.oatomic ||
- (td->io_ops->flags & FIO_RAWIO))
+ td_ioengine_flagged(td, FIO_RAWIO))
td->orig_buffer_size += page_mask + td->o.mem_align;
if (td->o.mem_type == MEM_SHMHUGE || td->o.mem_type == MEM_MMAPHUGE) {
- unsigned long bs;
+ unsigned long long bs;
bs = td->orig_buffer_size + td->o.hugepage_size - 1;
td->orig_buffer_size = bs & ~(td->o.hugepage_size - 1);
return 1;
if (td->o.odirect || td->o.mem_align || td->o.oatomic ||
- (td->io_ops->flags & FIO_RAWIO))
- p = PAGE_ALIGN(td->orig_buffer) + td->o.mem_align;
+ td_ioengine_flagged(td, FIO_RAWIO))
+ p = PTR_ALIGN(td->orig_buffer, page_mask) + td->o.mem_align;
else
p = td->orig_buffer;
- cl_align = os_cache_line_size();
-
for (i = 0; i < max_units; i++) {
- void *ptr;
-
- if (td->terminate)
- return 1;
-
- ptr = fio_memalign(cl_align, sizeof(*io_u));
- if (!ptr) {
- log_err("fio: unable to allocate aligned memory\n");
- break;
- }
-
- io_u = ptr;
- memset(io_u, 0, sizeof(*io_u));
- INIT_FLIST_HEAD(&io_u->verify_list);
+ io_u = td->io_u_all.io_us[i];
dprint(FD_MEM, "io_u alloc %p, index %u\n", io_u, i);
if (data_xfer) {
fill_verify_pattern(td, io_u->buf, max_bs, io_u, 0, 0);
}
}
-
- io_u->index = i;
- io_u->flags = IO_U_F_FREE;
- io_u_qpush(&td->io_u_freelist, io_u);
-
- /*
- * io_u never leaves this stack, used for iteration of all
- * io_u buffers.
- */
- io_u_qpush(&td->io_u_all, io_u);
-
- if (td->io_ops->io_u_init) {
- int ret = td->io_ops->io_u_init(td, io_u);
-
- if (ret) {
- log_err("fio: failed to init engine data: %d\n", ret);
- return 1;
- }
- }
-
p += max_bs;
}
- if (td->o.verify != VERIFY_NONE) {
- td->last_write_comp = scalloc(max_units, sizeof(uint64_t));
- if (!td->last_write_comp) {
- log_err("fio: failed to alloc write comp data\n");
- return 1;
- }
- }
-
return 0;
}
+/*
+ * This function is Linux specific.
+ * FIO_HAVE_IOSCHED_SWITCH enabled currently means it's Linux.
+ */
static int switch_ioscheduler(struct thread_data *td)
{
- char tmp[256], tmp2[128];
+#ifdef FIO_HAVE_IOSCHED_SWITCH
+ char tmp[256], tmp2[128], *p;
FILE *f;
int ret;
- if (td->io_ops->flags & FIO_DISKLESSIO)
+ if (td_ioengine_flagged(td, FIO_DISKLESSIO))
return 0;
- sprintf(tmp, "%s/queue/scheduler", td->sysfs_root);
+ assert(td->files && td->files[0]);
+ sprintf(tmp, "%s/queue/scheduler", td->files[0]->du->sysfs_root);
f = fopen(tmp, "r+");
if (!f) {
/*
* Read back and check that the selected scheduler is now the default.
*/
- memset(tmp, 0, sizeof(tmp));
- ret = fread(tmp, sizeof(tmp), 1, f);
+ ret = fread(tmp, 1, sizeof(tmp) - 1, f);
if (ferror(f) || ret < 0) {
td_verror(td, errno, "fread");
fclose(f);
return 1;
}
+ tmp[ret] = '\0';
/*
- * either a list of io schedulers or "none\n" is expected.
+ * either a list of io schedulers or "none\n" is expected. Strip the
+ * trailing newline.
*/
- tmp[strlen(tmp) - 1] = '\0';
+ p = tmp;
+ strsep(&p, "\n");
+ /*
+ * Write to "none" entry doesn't fail, so check the result here.
+ */
+ if (!strcmp(tmp, "none")) {
+ log_err("fio: io scheduler is not tunable\n");
+ fclose(f);
+ return 0;
+ }
sprintf(tmp2, "[%s]", td->o.ioscheduler);
if (!strstr(tmp, tmp2)) {
fclose(f);
return 0;
+#else
+ return 0;
+#endif
}
static bool keep_running(struct thread_data *td)
if (td->done)
return false;
+ if (td->terminate)
+ return false;
if (td->o.time_based)
return true;
if (td->o.loops) {
if (exceeds_number_ios(td))
return false;
- if (td->o.io_limit)
- limit = td->o.io_limit;
+ if (td->o.io_size)
+ limit = td->o.io_size;
else
limit = td->o.size;
uint64_t diff;
/*
- * If the difference is less than the minimum IO size, we
+ * If the difference is less than the maximum IO size, we
* are done.
*/
diff = limit - ddir_rw_sum(td->io_bytes);
if (diff < td_max_bs(td))
return false;
- if (fio_files_done(td) && !td->o.io_limit)
+ if (fio_files_done(td) && !td->o.io_size)
return false;
return true;
break;
io_u = get_io_u(td);
- if (!io_u)
+ if (IS_ERR_OR_NULL(io_u))
break;
- io_u_set(io_u, IO_U_F_FLIGHT);
+ io_u_set(td, io_u, IO_U_F_FLIGHT);
io_u->error = 0;
io_u->resid = 0;
if (ddir_rw(acct_ddir(io_u)))
struct thread_data *td = fd->td;
struct thread_options *o = &td->o;
struct sk_out *sk_out = fd->sk_out;
- pthread_condattr_t attr;
- int clear_state;
+ uint64_t bytes_done[DDIR_RWDIR_CNT];
+ int deadlock_loop_cnt;
+ bool clear_state;
int ret;
sk_out_assign(sk_out);
} else
td->pid = gettid();
- fio_local_clock_init(o->use_thread);
+ fio_local_clock_init();
dprint(FD_PROCESS, "jobs pid=%d started\n", (int) td->pid);
INIT_FLIST_HEAD(&td->io_hist_list);
INIT_FLIST_HEAD(&td->verify_list);
INIT_FLIST_HEAD(&td->trim_list);
- INIT_FLIST_HEAD(&td->next_rand_list);
- pthread_mutex_init(&td->io_u_lock, NULL);
td->io_hist_tree = RB_ROOT;
- pthread_condattr_init(&attr);
- pthread_cond_init(&td->verify_cond, &attr);
- pthread_cond_init(&td->free_cond, &attr);
+ ret = mutex_cond_init_pshared(&td->io_u_lock, &td->free_cond);
+ if (ret) {
+ td_verror(td, ret, "mutex_cond_init_pshared");
+ goto err;
+ }
+ ret = cond_init_pshared(&td->verify_cond);
+ if (ret) {
+ td_verror(td, ret, "mutex_cond_pshared");
+ goto err;
+ }
td_set_runstate(td, TD_INITIALIZED);
- dprint(FD_MUTEX, "up startup_mutex\n");
- fio_mutex_up(startup_mutex);
- dprint(FD_MUTEX, "wait on td->mutex\n");
- fio_mutex_down(td->mutex);
- dprint(FD_MUTEX, "done waiting on td->mutex\n");
+ dprint(FD_MUTEX, "up startup_sem\n");
+ fio_sem_up(startup_sem);
+ dprint(FD_MUTEX, "wait on td->sem\n");
+ fio_sem_down(td->sem);
+ dprint(FD_MUTEX, "done waiting on td->sem\n");
/*
* A new gid requires privilege, so we need to do this before setting
goto err;
}
+ td_zone_gen_index(td);
+
+ /*
+ * Do this early, we don't want the compress threads to be limited
+ * to the same CPUs as the IO workers. So do this before we set
+ * any potential CPU affinity
+ */
+ if (iolog_compress_init(td, sk_out))
+ goto err;
+
/*
* If we have a gettimeofday() thread, make sure we exclude that
* thread from this job
* May alter parameters that init_io_u() will use, so we need to
* do this first.
*/
- if (init_iolog(td))
+ if (!init_iolog(td))
+ goto err;
+
+ if (td_io_init(td))
goto err;
if (init_io_u(td))
goto err;
+ if (td->io_ops->post_init && td->io_ops->post_init(td))
+ goto err;
+
if (o->verify_async && verify_async_init(td))
goto err;
if (!o->create_serialize && setup_files(td))
goto err;
- if (td_io_init(td))
- goto err;
-
- if (init_random_map(td))
+ if (!init_random_map(td))
goto err;
if (o->exec_prerun && exec_string(o, o->exec_prerun, (const char *)"prerun"))
goto err;
- if (o->pre_read) {
- if (pre_read_files(td) < 0)
- goto err;
- }
-
- if (iolog_compress_init(td, sk_out))
+ if (o->pre_read && !pre_read_files(td))
goto err;
fio_verify_init(td);
if (rate_submit_init(td, sk_out))
goto err;
- fio_gettime(&td->epoch, NULL);
+ set_epoch_time(td, o->log_unix_epoch);
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));
if (o->ratemin[DDIR_READ] || o->ratemin[DDIR_WRITE] ||
o->ratemin[DDIR_TRIM]) {
sizeof(td->bw_sample_time));
}
- clear_state = 0;
+ memset(bytes_done, 0, sizeof(bytes_done));
+ clear_state = false;
+
while (keep_running(td)) {
uint64_t verify_bytes;
fio_gettime(&td->start, NULL);
- memcpy(&td->tv_cache, &td->start, sizeof(td->start));
+ memcpy(&td->ts_cache, &td->start, sizeof(td->start));
- if (clear_state)
+ if (clear_state) {
clear_io_state(td, 0);
+ if (o->unlink_each_loop && unlink_all_files(td))
+ break;
+ }
+
prune_io_piece_log(td);
- if (td->o.verify_only && (td_write(td) || td_rw(td)))
+ if (td->o.verify_only && td_write(td))
verify_bytes = do_dry_run(td);
else {
- uint64_t bytes_done[DDIR_RWDIR_CNT];
-
do_io(td, bytes_done);
if (!ddir_rw_sum(bytes_done)) {
}
}
- clear_state = 1;
+ /*
+ * If we took too long to shut down, the main thread could
+ * already consider us reaped/exited. If that happens, break
+ * out and clean up.
+ */
+ if (td->runstate >= TD_EXITED)
+ break;
+
+ clear_state = true;
/*
* Make sure we've successfully updated the rusage stats
* the rusage_sem, which would never get upped because
* this thread is waiting for the stat mutex.
*/
- check_update_rusage(td);
+ deadlock_loop_cnt = 0;
+ do {
+ check_update_rusage(td);
+ if (!fio_sem_down_trylock(stat_sem))
+ break;
+ usleep(1000);
+ if (deadlock_loop_cnt++ > 5000) {
+ log_err("fio seems to be stuck grabbing stat_sem, forcibly exiting\n");
+ td->error = EDEADLK;
+ goto err;
+ }
+ } while (1);
- fio_mutex_down(stat_mutex);
if (td_read(td) && td->io_bytes[DDIR_READ])
update_runtime(td, elapsed_us, DDIR_READ);
if (td_write(td) && td->io_bytes[DDIR_WRITE])
if (td_trim(td) && td->io_bytes[DDIR_TRIM])
update_runtime(td, elapsed_us, DDIR_TRIM);
fio_gettime(&td->start, NULL);
- fio_mutex_up(stat_mutex);
+ fio_sem_up(stat_sem);
if (td->error || td->terminate)
break;
if (!o->do_verify ||
o->verify == VERIFY_NONE ||
- (td->io_ops->flags & FIO_UNIDIR))
+ td_ioengine_flagged(td, FIO_UNIDIR))
continue;
clear_io_state(td, 0);
*/
check_update_rusage(td);
- fio_mutex_down(stat_mutex);
+ fio_sem_down(stat_sem);
update_runtime(td, elapsed_us, DDIR_READ);
fio_gettime(&td->start, NULL);
- fio_mutex_up(stat_mutex);
+ fio_sem_up(stat_sem);
if (td->error || td->terminate)
break;
}
+ /*
+ * Acquire this lock if we were doing overlap checking in
+ * 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);
+ td_set_runstate(td, TD_FINISHING);
+ if (td_offload_overlap(td))
+ pthread_mutex_unlock(&overlap_check);
+
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];
fio_unpin_memory(td);
- fio_writeout_logs(td);
+ td_writeout_logs(td, true);
iolog_compress_exit(td);
rate_submit_exit(td);
close_and_free_files(td);
cleanup_io_u(td);
close_ioengine(td);
- cgroup_shutdown(td, &cgroup_mnt);
+ cgroup_shutdown(td, cgroup_mnt);
verify_free_state(td);
+ td_zone_free_index(td);
if (fio_option_is_set(o, cpumask)) {
ret = fio_cpuset_exit(&o->cpumask);
*/
if (o->write_iolog_file)
write_iolog_close(td);
-
- fio_mutex_remove(td->mutex);
- td->mutex = NULL;
+ if (td->io_log_rfile)
+ fclose(td->io_log_rfile);
td_set_runstate(td, TD_EXITED);
return (void *) (uintptr_t) td->error;
}
-
-/*
- * We cannot pass the td data into a forked process, so attach the td and
- * pass it to the thread worker.
- */
-static int fork_main(struct sk_out *sk_out, int shmid, int offset)
-{
- struct fork_data *fd;
- void *data, *ret;
-
-#if !defined(__hpux) && !defined(CONFIG_NO_SHM)
- data = shmat(shmid, NULL, 0);
- if (data == (void *) -1) {
- int __err = errno;
-
- perror("shmat");
- return __err;
- }
-#else
- /*
- * HP-UX inherits shm mappings?
- */
- data = threads;
-#endif
-
- fd = calloc(1, sizeof(*fd));
- fd->td = data + offset * sizeof(struct thread_data);
- fd->sk_out = sk_out;
- ret = thread_main(fd);
- shmdt(data);
- return (int) (uintptr_t) ret;
-}
-
-static void dump_td_info(struct thread_data *td)
-{
- log_err("fio: job '%s' hasn't exited in %lu seconds, it appears to "
- "be stuck. Doing forceful exit of this job.\n", td->o.name,
- (unsigned long) time_since_now(&td->terminate_time));
-}
-
/*
* Run over the job map and reap the threads that have exited, if any.
*/
-static void reap_threads(unsigned int *nr_running, unsigned int *t_rate,
- unsigned int *m_rate)
+static void reap_threads(unsigned int *nr_running, uint64_t *t_rate,
+ uint64_t *m_rate)
{
struct thread_data *td;
unsigned int cputhreads, realthreads, pending;
for_each_td(td, i) {
int flags = 0;
- /*
- * ->io_ops is NULL for a thread that has closed its
- * io engine
- */
- if (td->io_ops && !strcmp(td->io_ops->name, "cpuio"))
+ if (!strcmp(td->o.ioengine, "cpuio"))
cputhreads++;
else
realthreads++;
* move on.
*/
if (td->terminate &&
+ td->runstate < TD_FSYNCING &&
time_since_now(&td->terminate_time) >= FIO_REAP_TIMEOUT) {
- dump_td_info(td);
+ log_err("fio: job '%s' (state=%d) hasn't exited in "
+ "%lu seconds, it appears to be stuck. Doing "
+ "forceful exit of this job.\n",
+ td->o.name, td->runstate,
+ (unsigned long) time_since_now(&td->terminate_time));
td_set_runstate(td, TD_REAPED);
goto reaped;
}
static bool trigger_timedout(void)
{
if (trigger_timeout)
- return time_since_genesis() >= trigger_timeout;
+ if (time_since_genesis() >= trigger_timeout) {
+ trigger_timeout = 0;
+ return true;
+ }
return false;
}
{
int ret;
- if (!cmd)
+ if (!cmd || cmd[0] == '\0')
return;
ret = system(cmd);
if (is_backend) {
void *data;
- int ver;
ret = fio_server_get_verify_state(td->o.name,
- td->thread_number - 1, &data, &ver);
+ td->thread_number - 1, &data);
if (!ret)
- verify_convert_assign_state(td, data, ver);
+ verify_assign_state(td, data);
} else
ret = verify_load_state(td, "local");
if (!td_write(td) || td->o.allow_mounted_write)
return false;
+ /*
+ * If FIO_HAVE_CHARDEV_SIZE is defined, it's likely that chrdevs
+ * are mkfs'd and mounted.
+ */
for_each_file(td, f, i) {
- if (f->filetype != FIO_TYPE_BD)
+#ifdef FIO_HAVE_CHARDEV_SIZE
+ if (f->filetype != FIO_TYPE_BLOCK && f->filetype != FIO_TYPE_CHAR)
+#else
+ if (f->filetype != FIO_TYPE_BLOCK)
+#endif
continue;
if (device_is_mounted(f->file_name))
goto mounted;
return false;
mounted:
- log_err("fio: %s appears mounted, and 'allow_mounted_write' isn't set. Aborting.", f->file_name);
+ log_err("fio: %s appears mounted, and 'allow_mounted_write' isn't set. Aborting.\n", f->file_name);
return true;
}
static void run_threads(struct sk_out *sk_out)
{
struct thread_data *td;
- unsigned int i, todo, nr_running, m_rate, t_rate, nr_started;
+ unsigned int i, todo, nr_running, nr_started;
+ uint64_t m_rate, t_rate;
uint64_t spent;
if (fio_gtod_offload && fio_start_gtod_thread())
}
if (output_format & FIO_OUTPUT_NORMAL) {
- log_info("Starting ");
+ struct buf_output out;
+
+ buf_output_init(&out);
+ __log_buf(&out, "Starting ");
if (nr_thread)
- log_info("%d thread%s", nr_thread,
+ __log_buf(&out, "%d thread%s", nr_thread,
nr_thread > 1 ? "s" : "");
if (nr_process) {
if (nr_thread)
- log_info(" and ");
- log_info("%d process%s", nr_process,
+ __log_buf(&out, " and ");
+ __log_buf(&out, "%d process%s", nr_process,
nr_process > 1 ? "es" : "");
}
- log_info("\n");
- log_info_flush();
+ __log_buf(&out, "\n");
+ log_info_buf(out.buf, out.buflen);
+ buf_output_free(&out);
}
todo = thread_number;
while (todo) {
struct thread_data *map[REAL_MAX_JOBS];
- struct timeval this_start;
+ struct timespec this_start;
int this_jobs = 0, left;
+ struct fork_data *fd;
/*
* create threads (TD_NOT_CREATED -> TD_CREATED)
init_disk_util(td);
- td->rusage_sem = fio_mutex_init(FIO_MUTEX_LOCKED);
+ td->rusage_sem = fio_sem_init(FIO_SEM_LOCKED);
td->update_rusage = 0;
/*
map[this_jobs++] = td;
nr_started++;
+ fd = calloc(1, sizeof(*fd));
+ fd->td = td;
+ fd->sk_out = sk_out;
+
if (td->o.use_thread) {
- struct fork_data *fd;
int ret;
- fd = calloc(1, sizeof(*fd));
- fd->td = td;
- fd->sk_out = sk_out;
-
dprint(FD_PROCESS, "will pthread_create\n");
ret = pthread_create(&td->thread, NULL,
thread_main, fd);
nr_started--;
break;
}
+ fd = NULL;
ret = pthread_detach(td->thread);
if (ret)
log_err("pthread_detach: %s",
dprint(FD_PROCESS, "will fork\n");
pid = fork();
if (!pid) {
- int ret = fork_main(sk_out, shm_id, i);
+ int ret;
+ ret = (int)(uintptr_t)thread_main(fd);
_exit(ret);
} else if (i == fio_debug_jobno)
*fio_debug_jobp = pid;
}
- dprint(FD_MUTEX, "wait on startup_mutex\n");
- if (fio_mutex_down_timeout(startup_mutex, 10000)) {
+ dprint(FD_MUTEX, "wait on startup_sem\n");
+ if (fio_sem_down_timeout(startup_sem, 10000)) {
log_err("fio: job startup hung? exiting.\n");
fio_terminate_threads(TERMINATE_ALL);
- fio_abort = 1;
+ fio_abort = true;
nr_started--;
+ free(fd);
break;
}
- dprint(FD_MUTEX, "done waiting on startup_mutex\n");
+ dprint(FD_MUTEX, "done waiting on startup_sem\n");
}
/*
m_rate += ddir_rw_sum(td->o.ratemin);
t_rate += ddir_rw_sum(td->o.rate);
todo--;
- fio_mutex_up(td->mutex);
+ fio_sem_up(td->sem);
}
reap_threads(&nr_running, &t_rate, &m_rate);
update_io_ticks();
}
-static void wait_for_helper_thread_exit(void)
-{
- void *ret;
-
- helper_exit = 1;
- pthread_cond_signal(&helper_cond);
- pthread_join(helper_thread, &ret);
-}
-
static void free_disk_util(void)
{
disk_util_prune_entries();
-
- pthread_cond_destroy(&helper_cond);
-}
-
-static void *helper_thread_main(void *data)
-{
- struct sk_out *sk_out = data;
- int ret = 0;
-
- sk_out_assign(sk_out);
-
- fio_mutex_up(startup_mutex);
-
- while (!ret) {
- uint64_t sec = DISK_UTIL_MSEC / 1000;
- uint64_t nsec = (DISK_UTIL_MSEC % 1000) * 1000000;
- struct timespec ts;
- struct timeval tv;
-
- gettimeofday(&tv, NULL);
- ts.tv_sec = tv.tv_sec + sec;
- ts.tv_nsec = (tv.tv_usec * 1000) + nsec;
-
- if (ts.tv_nsec >= 1000000000ULL) {
- ts.tv_nsec -= 1000000000ULL;
- ts.tv_sec++;
- }
-
- pthread_cond_timedwait(&helper_cond, &helper_lock, &ts);
-
- ret = update_io_ticks();
-
- if (helper_do_stat) {
- helper_do_stat = 0;
- __show_running_run_stats();
- }
-
- if (!is_backend)
- print_thread_status();
- }
-
- sk_out_drop();
- return NULL;
-}
-
-static int create_helper_thread(struct sk_out *sk_out)
-{
- int ret;
-
- setup_disk_util();
-
- pthread_cond_init(&helper_cond, NULL);
- pthread_mutex_init(&helper_lock, NULL);
-
- ret = pthread_create(&helper_thread, NULL, helper_thread_main, sk_out);
- if (ret) {
- log_err("Can't create helper thread: %s\n", strerror(ret));
- return 1;
- }
-
- dprint(FD_MUTEX, "wait on startup_mutex\n");
- fio_mutex_down(startup_mutex);
- dprint(FD_MUTEX, "done waiting on startup_mutex\n");
- return 0;
+ helper_thread_destroy();
}
int fio_backend(struct sk_out *sk_out)
setup_log(&agg_io_log[DDIR_TRIM], &p, "agg-trim_bw.log");
}
- startup_mutex = fio_mutex_init(FIO_MUTEX_LOCKED);
- if (startup_mutex == NULL)
+ startup_sem = fio_sem_init(FIO_SEM_LOCKED);
+ if (!sk_out)
+ is_local_backend = true;
+ if (startup_sem == NULL)
return 1;
set_genesis_time();
stat_init();
- create_helper_thread(sk_out);
+ helper_thread_create(startup_sem, sk_out);
cgroup_list = smalloc(sizeof(*cgroup_list));
- INIT_FLIST_HEAD(cgroup_list);
+ if (cgroup_list)
+ INIT_FLIST_HEAD(cgroup_list);
run_threads(sk_out);
- wait_for_helper_thread_exit();
+ helper_thread_exit();
if (!fio_abort) {
__show_run_stats();
for (i = 0; i < DDIR_RWDIR_CNT; i++) {
struct io_log *log = agg_io_log[i];
- flush_log(log, 0);
+ flush_log(log, false);
free_log(log);
}
}
}
for_each_td(td, i) {
+ steadystate_free(td);
fio_options_free(td);
if (td->rusage_sem) {
- fio_mutex_remove(td->rusage_sem);
+ fio_sem_remove(td->rusage_sem);
td->rusage_sem = NULL;
}
+ fio_sem_remove(td->sem);
+ td->sem = NULL;
}
free_disk_util();
- cgroup_kill(cgroup_list);
- sfree(cgroup_list);
- sfree(cgroup_mnt);
+ if (cgroup_list) {
+ cgroup_kill(cgroup_list);
+ sfree(cgroup_list);
+ }
- fio_mutex_remove(startup_mutex);
+ fio_sem_remove(startup_sem);
stat_exit();
return exit_value;
}