X-Git-Url: https://git.kernel.dk/?p=fio.git;a=blobdiff_plain;f=stat.c;h=ef9c4af2d81878a1ec6ec427cc804c18f7b07752;hp=3a014d6c1aa8f8e7f10690bf09caba68396e52d8;hb=029e6147e9005ddb08774193aa8b3a01174007b1;hpb=b2b3eefe6db21daac41f162a102c0a7c5241acfd

diff --git a/stat.c b/stat.c
index 3a014d6c..ef9c4af2 100644
--- a/stat.c
+++ b/stat.c
@@ -1,10 +1,7 @@
 #include <stdio.h>
 #include <string.h>
 #include <sys/time.h>
-#include <sys/types.h>
 #include <sys/stat.h>
-#include <dirent.h>
-#include <libgen.h>
 #include <math.h>
 
 #include "fio.h"
@@ -17,10 +14,11 @@
 #include "lib/output_buffer.h"
 #include "helper_thread.h"
 #include "smalloc.h"
+#include "zbd.h"
 
-#define LOG_MSEC_SLACK	10
+#define LOG_MSEC_SLACK	1
 
-struct fio_mutex *stat_mutex;
+struct fio_sem *stat_sem;
 
 void clear_rusage_stat(struct thread_data *td)
 {
@@ -135,7 +133,7 @@ static int double_cmp(const void *a, const void *b)
 	return cmp;
 }
 
-unsigned int calc_clat_percentiles(unsigned int *io_u_plat, unsigned long long nr,
+unsigned int calc_clat_percentiles(uint64_t *io_u_plat, unsigned long long nr,
 				   fio_fp64_t *plist, unsigned long long **output,
 				   unsigned long long *maxv, unsigned long long *minv)
 {
@@ -198,7 +196,7 @@ unsigned int calc_clat_percentiles(unsigned int *io_u_plat, unsigned long long n
 /*
  * Find and display the p-th percentile of clat
  */
-static void show_clat_percentiles(unsigned int *io_u_plat, unsigned long long nr,
+static void show_clat_percentiles(uint64_t *io_u_plat, unsigned long long nr,
 				  fio_fp64_t *plist, unsigned int precision,
 				  const char *pre, struct buf_output *out)
 {
@@ -323,7 +321,7 @@ void show_group_stats(struct group_run_stats *rs, struct buf_output *out)
 	}
 }
 
-void stat_calc_dist(unsigned int *map, unsigned long total, double *io_u_dist)
+void stat_calc_dist(uint64_t *map, unsigned long total, double *io_u_dist)
 {
 	int i;
 
@@ -342,7 +340,7 @@ void stat_calc_dist(unsigned int *map, unsigned long total, double *io_u_dist)
 }
 
 static void stat_calc_lat(struct thread_stat *ts, double *dst,
-			  unsigned int *src, int nr)
+			  uint64_t *src, int nr)
 {
 	unsigned long total = ddir_rw_sum(ts->total_io_u);
 	int i;
@@ -365,7 +363,7 @@ static void stat_calc_lat(struct thread_stat *ts, double *dst,
  * To keep the terse format unaltered, add all of the ns latency
  * buckets to the first us latency bucket
  */
-void stat_calc_lat_nu(struct thread_stat *ts, double *io_u_lat_u)
+static void stat_calc_lat_nu(struct thread_stat *ts, double *io_u_lat_u)
 {
 	unsigned long ntotal = 0, total = ddir_rw_sum(ts->total_io_u);
 	int i;
@@ -422,7 +420,7 @@ static void show_ddir_status(struct group_run_stats *rs, struct thread_stat *ts,
 	unsigned long runt;
 	unsigned long long min, max, bw, iops;
 	double mean, dev;
-	char *io_p, *bw_p, *bw_p_alt, *iops_p;
+	char *io_p, *bw_p, *bw_p_alt, *iops_p, *zbd_w_st = NULL;
 	int i2p;
 
 	if (ddir_sync(ddir)) {
@@ -453,12 +451,16 @@ static void show_ddir_status(struct group_run_stats *rs, struct thread_stat *ts,
 
 	iops = (1000 * (uint64_t)ts->total_io_u[ddir]) / runt;
 	iops_p = num2str(iops, ts->sig_figs, 1, 0, N2S_NONE);
+	if (ddir == DDIR_WRITE)
+		zbd_w_st = zbd_write_status(ts);
 
-	log_buf(out, "  %s: IOPS=%s, BW=%s (%s)(%s/%llumsec)\n",
+	log_buf(out, "  %s: IOPS=%s, BW=%s (%s)(%s/%llumsec)%s\n",
 			rs->unified_rw_rep ? "mixed" : str[ddir],
 			iops_p, bw_p, bw_p_alt, io_p,
-			(unsigned long long) ts->runtime[ddir]);
+			(unsigned long long) ts->runtime[ddir],
+			zbd_w_st ? : "");
 
+	free(zbd_w_st);
 	free(io_p);
 	free(bw_p);
 	free(bw_p_alt);
@@ -622,8 +624,8 @@ static int block_state_category(int block_state)
 
 static int compare_block_infos(const void *bs1, const void *bs2)
 {
-	uint32_t block1 = *(uint32_t *)bs1;
-	uint32_t block2 = *(uint32_t *)bs2;
+	uint64_t block1 = *(uint64_t *)bs1;
+	uint64_t block2 = *(uint64_t *)bs2;
 	int state1 = BLOCK_INFO_STATE(block1);
 	int state2 = BLOCK_INFO_STATE(block2);
 	int bscat1 = block_state_category(state1);
@@ -673,7 +675,6 @@ static int calc_block_percentiles(int nr_block_infos, uint32_t *block_infos,
 	if (len > 1)
 		qsort((void *)plist, len, sizeof(plist[0]), double_cmp);
 
-	nr_uninit = 0;
 	/* Start only after the uninit entries end */
 	for (nr_uninit = 0;
 	     nr_uninit < nr_block_infos
@@ -854,12 +855,13 @@ static void show_thread_status_normal(struct thread_stat *ts,
 					io_u_dist[1], io_u_dist[2],
 					io_u_dist[3], io_u_dist[4],
 					io_u_dist[5], io_u_dist[6]);
-	log_buf(out, "     issued rwt: total=%llu,%llu,%llu,"
-				 " short=%llu,%llu,%llu,"
-				 " dropped=%llu,%llu,%llu\n",
+	log_buf(out, "     issued rwts: total=%llu,%llu,%llu,%llu"
+				 " short=%llu,%llu,%llu,0"
+				 " dropped=%llu,%llu,%llu,0\n",
 					(unsigned long long) ts->total_io_u[0],
 					(unsigned long long) ts->total_io_u[1],
 					(unsigned long long) ts->total_io_u[2],
+					(unsigned long long) ts->total_io_u[3],
 					(unsigned long long) ts->short_io_u[0],
 					(unsigned long long) ts->short_io_u[1],
 					(unsigned long long) ts->short_io_u[2],
@@ -1048,6 +1050,7 @@ static void add_ddir_status_json(struct thread_stat *ts,
 
 		tmp_object = json_create_object();
 		json_object_add_value_object(dir_object, "lat_ns", tmp_object);
+		json_object_add_value_int(dir_object, "total_ios", ts->total_io_u[DDIR_SYNC]);
 		json_object_add_value_int(tmp_object, "min", min);
 		json_object_add_value_int(tmp_object, "max", max);
 		json_object_add_value_float(tmp_object, "mean", mean);
@@ -1056,10 +1059,16 @@ static void add_ddir_status_json(struct thread_stat *ts,
 
 	if (ts->clat_percentiles || ts->lat_percentiles) {
 		if (ddir_rw(ddir)) {
+			uint64_t samples;
+
+			if (ts->clat_percentiles)
+				samples = ts->clat_stat[ddir].samples;
+			else
+				samples = ts->lat_stat[ddir].samples;
+
 			len = calc_clat_percentiles(ts->io_u_plat[ddir],
-					ts->clat_stat[ddir].samples,
-					ts->percentile_list, &ovals, &maxv,
-					&minv);
+					samples, ts->percentile_list, &ovals,
+					&maxv, &minv);
 		} else {
 			len = calc_clat_percentiles(ts->io_u_sync_plat,
 					ts->sync_stat.samples,
@@ -1207,7 +1216,7 @@ static void show_thread_status_terse_all(struct thread_stat *ts,
 		log_buf(out, ";%3.2f%%", io_u_lat_m[i]);
 
 	/* disk util stats, if any */
-	if (ver >= 3)
+	if (ver >= 3 && is_running_backend())
 		show_disk_util(1, NULL, out);
 
 	/* Additional output if continue_on_error set - default off*/
@@ -1224,7 +1233,7 @@ static void show_thread_status_terse_all(struct thread_stat *ts,
 }
 
 static void json_add_job_opts(struct json_object *root, const char *name,
-			      struct flist_head *opt_list, bool num_jobs)
+			      struct flist_head *opt_list)
 {
 	struct json_object *dir_object;
 	struct flist_head *entry;
@@ -1240,8 +1249,6 @@ static void json_add_job_opts(struct json_object *root, const char *name,
 		const char *pos = "";
 
 		p = flist_entry(entry, struct print_option, list);
-		if (!num_jobs && !strcmp(p->name, "numjobs"))
-			continue;
 		if (p->value)
 			pos = p->value;
 		json_object_add_value_string(dir_object, p->name, pos);
@@ -1275,7 +1282,7 @@ static struct json_object *show_thread_status_json(struct thread_stat *ts,
 	}
 
 	if (opt_list)
-		json_add_job_opts(root, "job options", opt_list, true);
+		json_add_job_opts(root, "job options", opt_list);
 
 	add_ddir_status_json(ts, rs, DDIR_READ, root);
 	add_ddir_status_json(ts, rs, DDIR_WRITE, root);
@@ -1292,19 +1299,15 @@ static struct json_object *show_thread_status_json(struct thread_stat *ts,
 		usr_cpu = 0;
 		sys_cpu = 0;
 	}
+	json_object_add_value_int(root, "job_runtime", ts->total_run_time);
 	json_object_add_value_float(root, "usr_cpu", usr_cpu);
 	json_object_add_value_float(root, "sys_cpu", sys_cpu);
 	json_object_add_value_int(root, "ctx", ts->ctx);
 	json_object_add_value_int(root, "majf", ts->majf);
 	json_object_add_value_int(root, "minf", ts->minf);
 
-
-	/* Calc % distribution of IO depths, usecond, msecond latency */
+	/* Calc % distribution of IO depths */
 	stat_calc_dist(ts->io_u_map, ddir_rw_sum(ts->total_io_u), io_u_dist);
-	stat_calc_lat_n(ts, io_u_lat_n);
-	stat_calc_lat_u(ts, io_u_lat_u);
-	stat_calc_lat_m(ts, io_u_lat_m);
-
 	tmp = json_create_object();
 	json_object_add_value_object(root, "iodepth_level", tmp);
 	/* Only show fixed 7 I/O depth levels*/
@@ -1317,6 +1320,44 @@ static struct json_object *show_thread_status_json(struct thread_stat *ts,
 		json_object_add_value_float(tmp, (const char *)name, io_u_dist[i]);
 	}
 
+	/* Calc % distribution of submit IO depths */
+	stat_calc_dist(ts->io_u_submit, ts->total_submit, io_u_dist);
+	tmp = json_create_object();
+	json_object_add_value_object(root, "iodepth_submit", tmp);
+	/* Only show fixed 7 I/O depth levels*/
+	for (i = 0; i < 7; i++) {
+		char name[20];
+		if (i == 0)
+			snprintf(name, 20, "0");
+		else if (i < 6)
+			snprintf(name, 20, "%d", 1 << (i+1));
+		else
+			snprintf(name, 20, ">=%d", 1 << i);
+		json_object_add_value_float(tmp, (const char *)name, io_u_dist[i]);
+	}
+
+	/* Calc % distribution of completion IO depths */
+	stat_calc_dist(ts->io_u_complete, ts->total_complete, io_u_dist);
+	tmp = json_create_object();
+	json_object_add_value_object(root, "iodepth_complete", tmp);
+	/* Only show fixed 7 I/O depth levels*/
+	for (i = 0; i < 7; i++) {
+		char name[20];
+		if (i == 0)
+			snprintf(name, 20, "0");
+		else if (i < 6)
+			snprintf(name, 20, "%d", 1 << (i+1));
+		else
+			snprintf(name, 20, ">=%d", 1 << i);
+		json_object_add_value_float(tmp, (const char *)name, io_u_dist[i]);
+	}
+
+	/* Calc % distribution of nsecond, usecond, msecond latency */
+	stat_calc_dist(ts->io_u_map, ddir_rw_sum(ts->total_io_u), io_u_dist);
+	stat_calc_lat_n(ts, io_u_lat_n);
+	stat_calc_lat_u(ts, io_u_lat_u);
+	stat_calc_lat_m(ts, io_u_lat_m);
+
 	/* Nanosecond latency */
 	tmp = json_create_object();
 	json_object_add_value_object(root, "latency_ns", tmp);
@@ -1402,7 +1443,7 @@ static struct json_object *show_thread_status_json(struct thread_stat *ts,
 	if (ts->ss_dur) {
 		struct json_object *data;
 		struct json_array *iops, *bw;
-		int i, j, k;
+		int j, k, l;
 		char ss_buf[64];
 
 		snprintf(ss_buf, sizeof(ss_buf), "%s%s:%f%s",
@@ -1438,8 +1479,8 @@ static struct json_object *show_thread_status_json(struct thread_stat *ts,
 			j = ts->ss_head;
 		else
 			j = ts->ss_head == 0 ? ts->ss_dur - 1 : ts->ss_head - 1;
-		for (i = 0; i < ts->ss_dur; i++) {
-			k = (j + i) % ts->ss_dur;
+		for (l = 0; l < ts->ss_dur; l++) {
+			k = (j + l) % ts->ss_dur;
 			json_array_add_value_int(bw, ts->ss_bw_data[k]);
 			json_array_add_value_int(iops, ts->ss_iops_data[k]);
 		}
@@ -1609,6 +1650,8 @@ void sum_thread_stats(struct thread_stat *dst, struct thread_stat *src,
 		}
 	}
 
+	dst->total_io_u[DDIR_SYNC] += src->total_io_u[DDIR_SYNC];
+
 	for (k = 0; k < DDIR_RWDIR_CNT; k++) {
 		int m;
 
@@ -1623,6 +1666,7 @@ void sum_thread_stats(struct thread_stat *dst, struct thread_stat *src,
 	dst->total_run_time += src->total_run_time;
 	dst->total_submit += src->total_submit;
 	dst->total_complete += src->total_complete;
+	dst->nr_zone_resets += src->nr_zone_resets;
 }
 
 void init_group_run_stat(struct group_run_stats *gs)
@@ -1647,6 +1691,7 @@ void init_thread_stat(struct thread_stat *ts)
 		ts->bw_stat[j].min_val = -1UL;
 		ts->iops_stat[j].min_val = -1UL;
 	}
+	ts->sync_stat.min_val = -1UL;
 	ts->groupid = -1;
 }
 
@@ -1857,13 +1902,14 @@ void __show_run_stats(void)
 		char time_buf[32];
 		struct timeval now;
 		unsigned long long ms_since_epoch;
+		time_t tv_sec;
 
 		gettimeofday(&now, NULL);
 		ms_since_epoch = (unsigned long long)(now.tv_sec) * 1000 +
 		                 (unsigned long long)(now.tv_usec) / 1000;
 
-		os_ctime_r((const time_t *) &now.tv_sec, time_buf,
-				sizeof(time_buf));
+		tv_sec = now.tv_sec;
+		os_ctime_r(&tv_sec, time_buf, sizeof(time_buf));
 		if (time_buf[strlen(time_buf) - 1] == '\n')
 			time_buf[strlen(time_buf) - 1] = '\0';
 
@@ -1873,7 +1919,7 @@ void __show_run_stats(void)
 		json_object_add_value_int(root, "timestamp_ms", ms_since_epoch);
 		json_object_add_value_string(root, "time", time_buf);
 		global = get_global_options();
-		json_add_job_opts(root, "global options", &global->opt_list, false);
+		json_add_job_opts(root, "global options", &global->opt_list);
 		array = json_create_array();
 		json_object_add_value_array(root, "jobs", array);
 	}
@@ -1934,19 +1980,14 @@ void __show_run_stats(void)
 		buf_output_free(out);
 	}
 
+	fio_idle_prof_cleanup();
+
 	log_info_flush();
 	free(runstats);
 	free(threadstats);
 	free(opt_lists);
 }
 
-void show_run_stats(void)
-{
-	fio_mutex_down(stat_mutex);
-	__show_run_stats();
-	fio_mutex_up(stat_mutex);
-}
-
 void __show_running_run_stats(void)
 {
 	struct thread_data *td;
@@ -1954,7 +1995,7 @@ void __show_running_run_stats(void)
 	struct timespec ts;
 	int i;
 
-	fio_mutex_down(stat_mutex);
+	fio_sem_down(stat_sem);
 
 	rt = malloc(thread_number * sizeof(unsigned long long));
 	fio_gettime(&ts, NULL);
@@ -1980,7 +2021,7 @@ void __show_running_run_stats(void)
 			continue;
 		if (td->rusage_sem) {
 			td->update_rusage = 1;
-			fio_mutex_down(td->rusage_sem);
+			fio_sem_down(td->rusage_sem);
 		}
 		td->update_rusage = 0;
 	}
@@ -1997,7 +2038,7 @@ void __show_running_run_stats(void)
 	}
 
 	free(rt);
-	fio_mutex_up(stat_mutex);
+	fio_sem_up(stat_sem);
 }
 
 static bool status_interval_init;
@@ -2212,17 +2253,19 @@ static struct io_logs *get_cur_log(struct io_log *iolog)
 	 * submissions, flag 'td' as needing a log regrow and we'll take
 	 * care of it on the submission side.
 	 */
-	if (iolog->td->o.io_submit_mode == IO_MODE_OFFLOAD ||
+	if ((iolog->td && iolog->td->o.io_submit_mode == IO_MODE_OFFLOAD) ||
 	    !per_unit_log(iolog))
 		return regrow_log(iolog);
 
-	iolog->td->flags |= TD_F_REGROW_LOGS;
-	assert(iolog->pending->nr_samples < iolog->pending->max_samples);
+	if (iolog->td)
+		iolog->td->flags |= TD_F_REGROW_LOGS;
+	if (iolog->pending)
+		assert(iolog->pending->nr_samples < iolog->pending->max_samples);
 	return iolog->pending;
 }
 
 static void __add_log_sample(struct io_log *iolog, union io_sample_data data,
-			     enum fio_ddir ddir, unsigned int bs,
+			     enum fio_ddir ddir, unsigned long long bs,
 			     unsigned long t, uint64_t offset)
 {
 	struct io_logs *cur_log;
@@ -2287,6 +2330,8 @@ void reset_io_stats(struct thread_data *td)
 		}
 	}
 
+	ts->total_io_u[DDIR_SYNC] = 0;
+
 	for (i = 0; i < FIO_IO_U_MAP_NR; i++) {
 		ts->io_u_map[i] = 0;
 		ts->io_u_submit[i] = 0;
@@ -2302,6 +2347,7 @@ void reset_io_stats(struct thread_data *td)
 
 	ts->total_submit = 0;
 	ts->total_complete = 0;
+	ts->nr_zone_resets = 0;
 }
 
 static void __add_stat_to_log(struct io_log *iolog, enum fio_ddir ddir,
@@ -2335,9 +2381,11 @@ static void _add_stat_to_log(struct io_log *iolog, unsigned long elapsed,
 		__add_stat_to_log(iolog, ddir, elapsed, log_max);
 }
 
-static long add_log_sample(struct thread_data *td, struct io_log *iolog,
-			   union io_sample_data data, enum fio_ddir ddir,
-			   unsigned int bs, uint64_t offset)
+static unsigned long add_log_sample(struct thread_data *td,
+				    struct io_log *iolog,
+				    union io_sample_data data,
+				    enum fio_ddir ddir, unsigned long long bs,
+				    uint64_t offset)
 {
 	unsigned long elapsed, this_window;
 
@@ -2368,7 +2416,7 @@ static long add_log_sample(struct thread_data *td, struct io_log *iolog,
 	if (elapsed < iolog->avg_last[ddir])
 		return iolog->avg_last[ddir] - elapsed;
 	else if (this_window < iolog->avg_msec) {
-		int diff = iolog->avg_msec - this_window;
+		unsigned long diff = iolog->avg_msec - this_window;
 
 		if (inline_log(iolog) || diff > LOG_MSEC_SLACK)
 			return diff;
@@ -2398,7 +2446,7 @@ void finalize_logs(struct thread_data *td, bool unit_logs)
 		_add_stat_to_log(td->iops_log, elapsed, td->o.log_max != 0);
 }
 
-void add_agg_sample(union io_sample_data data, enum fio_ddir ddir, unsigned int bs)
+void add_agg_sample(union io_sample_data data, enum fio_ddir ddir, unsigned long long bs)
 {
 	struct io_log *iolog;
 
@@ -2428,13 +2476,16 @@ static void add_clat_percentile_sample(struct thread_stat *ts,
 }
 
 void add_clat_sample(struct thread_data *td, enum fio_ddir ddir,
-		     unsigned long long nsec, unsigned int bs, uint64_t offset)
+		     unsigned long long nsec, unsigned long long bs,
+		     uint64_t offset)
 {
+	const bool needs_lock = td_async_processing(td);
 	unsigned long elapsed, this_window;
 	struct thread_stat *ts = &td->ts;
 	struct io_log *iolog = td->clat_hist_log;
 
-	td_io_u_lock(td);
+	if (needs_lock)
+		__td_io_u_lock(td);
 
 	add_stat_sample(&ts->clat_stat[ddir], nsec);
 
@@ -2455,7 +2506,7 @@ void add_clat_sample(struct thread_data *td, enum fio_ddir ddir,
 		this_window = elapsed - hw->hist_last;
 		
 		if (this_window >= iolog->hist_msec) {
-			unsigned int *io_u_plat;
+			uint64_t *io_u_plat;
 			struct io_u_plat_entry *dst;
 
 			/*
@@ -2465,7 +2516,7 @@ void add_clat_sample(struct thread_data *td, enum fio_ddir ddir,
 			 * located in iolog.c after printing this sample to the
 			 * log file.
 			 */
-			io_u_plat = (unsigned int *) td->ts.io_u_plat[ddir];
+			io_u_plat = (uint64_t *) td->ts.io_u_plat[ddir];
 			dst = malloc(sizeof(struct io_u_plat_entry));
 			memcpy(&(dst->io_u_plat), io_u_plat,
 				FIO_IO_U_PLAT_NR * sizeof(unsigned int));
@@ -2483,36 +2534,43 @@ void add_clat_sample(struct thread_data *td, enum fio_ddir ddir,
 		}
 	}
 
-	td_io_u_unlock(td);
+	if (needs_lock)
+		__td_io_u_unlock(td);
 }
 
 void add_slat_sample(struct thread_data *td, enum fio_ddir ddir,
-		     unsigned long usec, unsigned int bs, uint64_t offset)
+		     unsigned long usec, unsigned long long bs, uint64_t offset)
 {
+	const bool needs_lock = td_async_processing(td);
 	struct thread_stat *ts = &td->ts;
 
 	if (!ddir_rw(ddir))
 		return;
 
-	td_io_u_lock(td);
+	if (needs_lock)
+		__td_io_u_lock(td);
 
 	add_stat_sample(&ts->slat_stat[ddir], usec);
 
 	if (td->slat_log)
 		add_log_sample(td, td->slat_log, sample_val(usec), ddir, bs, offset);
 
-	td_io_u_unlock(td);
+	if (needs_lock)
+		__td_io_u_unlock(td);
 }
 
 void add_lat_sample(struct thread_data *td, enum fio_ddir ddir,
-		    unsigned long long nsec, unsigned int bs, uint64_t offset)
+		    unsigned long long nsec, unsigned long long bs,
+		    uint64_t offset)
 {
+	const bool needs_lock = td_async_processing(td);
 	struct thread_stat *ts = &td->ts;
 
 	if (!ddir_rw(ddir))
 		return;
 
-	td_io_u_lock(td);
+	if (needs_lock)
+		__td_io_u_lock(td);
 
 	add_stat_sample(&ts->lat_stat[ddir], nsec);
 
@@ -2523,12 +2581,14 @@ void add_lat_sample(struct thread_data *td, enum fio_ddir ddir,
 	if (ts->lat_percentiles)
 		add_clat_percentile_sample(ts, nsec, ddir);
 
-	td_io_u_unlock(td);
+	if (needs_lock)
+		__td_io_u_unlock(td);
 }
 
 void add_bw_sample(struct thread_data *td, struct io_u *io_u,
 		   unsigned int bytes, unsigned long long spent)
 {
+	const bool needs_lock = td_async_processing(td);
 	struct thread_stat *ts = &td->ts;
 	unsigned long rate;
 
@@ -2537,7 +2597,8 @@ void add_bw_sample(struct thread_data *td, struct io_u *io_u,
 	else
 		rate = 0;
 
-	td_io_u_lock(td);
+	if (needs_lock)
+		__td_io_u_lock(td);
 
 	add_stat_sample(&ts->bw_stat[io_u->ddir], rate);
 
@@ -2546,7 +2607,9 @@ void add_bw_sample(struct thread_data *td, struct io_u *io_u,
 			       bytes, io_u->offset);
 
 	td->stat_io_bytes[io_u->ddir] = td->this_io_bytes[io_u->ddir];
-	td_io_u_unlock(td);
+
+	if (needs_lock)
+		__td_io_u_unlock(td);
 }
 
 static int __add_samples(struct thread_data *td, struct timespec *parent_tv,
@@ -2555,9 +2618,10 @@ static int __add_samples(struct thread_data *td, struct timespec *parent_tv,
 			 struct io_stat *stat, struct io_log *log,
 			 bool is_kb)
 {
+	const bool needs_lock = td_async_processing(td);
 	unsigned long spent, rate;
 	enum fio_ddir ddir;
-	unsigned int next, next_log;
+	unsigned long next, next_log;
 
 	next_log = avg_time;
 
@@ -2565,7 +2629,8 @@ static int __add_samples(struct thread_data *td, struct timespec *parent_tv,
 	if (spent < avg_time && avg_time - spent >= LOG_MSEC_SLACK)
 		return avg_time - spent;
 
-	td_io_u_lock(td);
+	if (needs_lock)
+		__td_io_u_lock(td);
 
 	/*
 	 * Compute both read and write rates for the interval.
@@ -2588,7 +2653,7 @@ static int __add_samples(struct thread_data *td, struct timespec *parent_tv,
 		add_stat_sample(&stat[ddir], rate);
 
 		if (log) {
-			unsigned int bs = 0;
+			unsigned long long bs = 0;
 
 			if (td->o.min_bs[ddir] == td->o.max_bs[ddir])
 				bs = td->o.min_bs[ddir];
@@ -2602,7 +2667,8 @@ static int __add_samples(struct thread_data *td, struct timespec *parent_tv,
 
 	timespec_add_msec(parent_tv, avg_time);
 
-	td_io_u_unlock(td);
+	if (needs_lock)
+		__td_io_u_unlock(td);
 
 	if (spent <= avg_time)
 		next = avg_time;
@@ -2622,9 +2688,11 @@ static int add_bw_samples(struct thread_data *td, struct timespec *t)
 void add_iops_sample(struct thread_data *td, struct io_u *io_u,
 		     unsigned int bytes)
 {
+	const bool needs_lock = td_async_processing(td);
 	struct thread_stat *ts = &td->ts;
 
-	td_io_u_lock(td);
+	if (needs_lock)
+		__td_io_u_lock(td);
 
 	add_stat_sample(&ts->iops_stat[io_u->ddir], 1);
 
@@ -2633,7 +2701,9 @@ void add_iops_sample(struct thread_data *td, struct io_u *io_u,
 			       bytes, io_u->offset);
 
 	td->stat_io_blocks[io_u->ddir] = td->this_io_blocks[io_u->ddir];
-	td_io_u_unlock(td);
+
+	if (needs_lock)
+		__td_io_u_unlock(td);
 }
 
 static int add_iops_samples(struct thread_data *td, struct timespec *t)
@@ -2682,7 +2752,7 @@ int calc_log_samples(void)
 
 void stat_init(void)
 {
-	stat_mutex = fio_mutex_init(FIO_MUTEX_UNLOCKED);
+	stat_sem = fio_sem_init(FIO_SEM_UNLOCKED);
 }
 
 void stat_exit(void)
@@ -2691,8 +2761,8 @@ void stat_exit(void)
 	 * When we have the mutex, we know out-of-band access to it
 	 * have ended.
 	 */
-	fio_mutex_down(stat_mutex);
-	fio_mutex_remove(stat_mutex);
+	fio_sem_down(stat_sem);
+	fio_sem_remove(stat_sem);
 }
 
 /*