X-Git-Url: https://git.kernel.dk/?p=fio.git;a=blobdiff_plain;f=stat.c;h=ef9c4af2d81878a1ec6ec427cc804c18f7b07752;hp=fde7af299de449d3f63afec10a548d8984ddc14c;hb=029e6147e9005ddb08774193aa8b3a01174007b1;hpb=8243be59aa35aa016fcbeee99353b08376953911

diff --git a/stat.c b/stat.c
index fde7af29..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)
 {
@@ -37,9 +35,9 @@ void update_rusage_stat(struct thread_data *td)
 	struct thread_stat *ts = &td->ts;
 
 	fio_getrusage(&td->ru_end);
-	ts->usr_time += mtime_since(&td->ru_start.ru_utime,
+	ts->usr_time += mtime_since_tv(&td->ru_start.ru_utime,
 					&td->ru_end.ru_utime);
-	ts->sys_time += mtime_since(&td->ru_start.ru_stime,
+	ts->sys_time += mtime_since_tv(&td->ru_start.ru_stime,
 					&td->ru_end.ru_stime);
 	ts->ctx += td->ru_end.ru_nvcsw + td->ru_end.ru_nivcsw
 			- (td->ru_start.ru_nvcsw + td->ru_start.ru_nivcsw);
@@ -58,7 +56,7 @@ void update_rusage_stat(struct thread_data *td)
  * group by looking at the index bits.
  *
  */
-static unsigned int plat_val_to_idx(unsigned int val)
+static unsigned int plat_val_to_idx(unsigned long long val)
 {
 	unsigned int msb, error_bits, base, offset, idx;
 
@@ -66,7 +64,7 @@ static unsigned int plat_val_to_idx(unsigned int val)
 	if (val == 0)
 		msb = 0;
 	else
-		msb = (sizeof(val)*8) - __builtin_clz(val) - 1;
+		msb = (sizeof(val)*8) - __builtin_clzll(val) - 1;
 
 	/*
 	 * MSB <= (FIO_IO_U_PLAT_BITS-1), cannot be rounded off. Use
@@ -98,9 +96,10 @@ static unsigned int plat_val_to_idx(unsigned int val)
  * Convert the given index of the bucket array to the value
  * represented by the bucket
  */
-static unsigned int plat_idx_to_val(unsigned int idx)
+static unsigned long long plat_idx_to_val(unsigned int idx)
 {
-	unsigned int error_bits, k, base;
+	unsigned int error_bits;
+	unsigned long long k, base;
 
 	assert(idx < FIO_IO_U_PLAT_NR);
 
@@ -111,7 +110,7 @@ static unsigned int plat_idx_to_val(unsigned int idx)
 
 	/* Find the group and compute the minimum value of that group */
 	error_bits = (idx >> FIO_IO_U_PLAT_BITS) - 1;
-	base = 1 << (error_bits + FIO_IO_U_PLAT_BITS);
+	base = ((unsigned long long) 1) << (error_bits + FIO_IO_U_PLAT_BITS);
 
 	/* Find its bucket number of the group */
 	k = idx % FIO_IO_U_PLAT_VAL;
@@ -134,17 +133,17 @@ static int double_cmp(const void *a, const void *b)
 	return cmp;
 }
 
-unsigned int calc_clat_percentiles(unsigned int *io_u_plat, unsigned long nr,
-				   fio_fp64_t *plist, unsigned int **output,
-				   unsigned int *maxv, unsigned int *minv)
+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)
 {
-	unsigned long sum = 0;
+	unsigned long long sum = 0;
 	unsigned int len, i, j = 0;
 	unsigned int oval_len = 0;
-	unsigned int *ovals = NULL;
-	int is_last;
+	unsigned long long *ovals = NULL;
+	bool is_last;
 
-	*minv = -1U;
+	*minv = -1ULL;
 	*maxv = 0;
 
 	len = 0;
@@ -165,7 +164,7 @@ unsigned int calc_clat_percentiles(unsigned int *io_u_plat, unsigned long nr,
 	/*
 	 * Calculate bucket values, note down max and min values
 	 */
-	is_last = 0;
+	is_last = false;
 	for (i = 0; i < FIO_IO_U_PLAT_NR && !is_last; i++) {
 		sum += io_u_plat[i];
 		while (sum >= (plist[j].u.f / 100.0 * nr)) {
@@ -173,7 +172,7 @@ unsigned int calc_clat_percentiles(unsigned int *io_u_plat, unsigned long nr,
 
 			if (j == oval_len) {
 				oval_len += 100;
-				ovals = realloc(ovals, oval_len * sizeof(unsigned int));
+				ovals = realloc(ovals, oval_len * sizeof(*ovals));
 			}
 
 			ovals[j] = plat_idx_to_val(i);
@@ -182,7 +181,7 @@ unsigned int calc_clat_percentiles(unsigned int *io_u_plat, unsigned long nr,
 			if (ovals[j] > *maxv)
 				*maxv = ovals[j];
 
-			is_last = (j == len - 1);
+			is_last = (j == len - 1) != 0;
 			if (is_last)
 				break;
 
@@ -197,53 +196,58 @@ unsigned int calc_clat_percentiles(unsigned int *io_u_plat, unsigned long nr,
 /*
  * Find and display the p-th percentile of clat
  */
-static void show_clat_percentiles(unsigned int *io_u_plat, unsigned long nr,
+static void show_clat_percentiles(uint64_t *io_u_plat, unsigned long long nr,
 				  fio_fp64_t *plist, unsigned int precision,
-				  struct buf_output *out)
+				  const char *pre, struct buf_output *out)
 {
-	unsigned int len, j = 0, minv, maxv;
-	unsigned int *ovals;
-	int is_last, per_line, scale_down;
+	unsigned int divisor, len, i, j = 0;
+	unsigned long long minv, maxv;
+	unsigned long long *ovals;
+	int per_line, scale_down, time_width;
+	bool is_last;
 	char fmt[32];
 
 	len = calc_clat_percentiles(io_u_plat, nr, plist, &ovals, &maxv, &minv);
-	if (!len)
+	if (!len || !ovals)
 		goto out;
 
 	/*
-	 * We default to usecs, but if the value range is such that we
-	 * should scale down to msecs, do that.
+	 * We default to nsecs, but if the value range is such that we
+	 * should scale down to usecs or msecs, do that.
 	 */
-	if (minv > 2000 && maxv > 99999) {
+	if (minv > 2000000 && maxv > 99999999ULL) {
+		scale_down = 2;
+		divisor = 1000000;
+		log_buf(out, "    %s percentiles (msec):\n     |", pre);
+	} else if (minv > 2000 && maxv > 99999) {
 		scale_down = 1;
-		log_buf(out, "    clat percentiles (msec):\n     |");
+		divisor = 1000;
+		log_buf(out, "    %s percentiles (usec):\n     |", pre);
 	} else {
 		scale_down = 0;
-		log_buf(out, "    clat percentiles (usec):\n     |");
+		divisor = 1;
+		log_buf(out, "    %s percentiles (nsec):\n     |", pre);
 	}
 
-	snprintf(fmt, sizeof(fmt), "%%1.%uf", precision);
-	per_line = (80 - 7) / (precision + 14);
 
-	for (j = 0; j < len; j++) {
-		char fbuf[16], *ptr = fbuf;
+	time_width = max(5, (int) (log10(maxv / divisor) + 1));
+	snprintf(fmt, sizeof(fmt), " %%%u.%ufth=[%%%dllu]%%c", precision + 3,
+			precision, time_width);
+	/* fmt will be something like " %5.2fth=[%4llu]%c" */
+	per_line = (80 - 7) / (precision + 10 + time_width);
 
+	for (j = 0; j < len; j++) {
 		/* for formatting */
 		if (j != 0 && (j % per_line) == 0)
 			log_buf(out, "     |");
 
 		/* end of the list */
-		is_last = (j == len - 1);
-
-		if (plist[j].u.f < 10.0)
-			ptr += sprintf(fbuf, " ");
-
-		snprintf(ptr, sizeof(fbuf), fmt, plist[j].u.f);
+		is_last = (j == len - 1) != 0;
 
-		if (scale_down)
+		for (i = 0; i < scale_down; i++)
 			ovals[j] = (ovals[j] + 999) / 1000;
 
-		log_buf(out, " %sth=[%5u]%c", fbuf, ovals[j], is_last ? '\n' : ',');
+		log_buf(out, fmt, plist[j].u.f, ovals[j], is_last ? '\n' : ',');
 
 		if (is_last)
 			break;
@@ -257,8 +261,8 @@ out:
 		free(ovals);
 }
 
-bool calc_lat(struct io_stat *is, unsigned long *min, unsigned long *max,
-	      double *mean, double *dev)
+bool calc_lat(struct io_stat *is, unsigned long long *min,
+	      unsigned long long *max, double *mean, double *dev)
 {
 	double n = (double) is->samples;
 
@@ -292,14 +296,14 @@ void show_group_stats(struct group_run_stats *rs, struct buf_output *out)
 		if (!rs->max_run[i])
 			continue;
 
-		io = num2str(rs->iobytes[i], 4, 1, i2p, N2S_BYTE);
-		ioalt = num2str(rs->iobytes[i], 4, 1, !i2p, N2S_BYTE);
-		agg = num2str(rs->agg[i], 4, 1, i2p, rs->unit_base);
-		aggalt = num2str(rs->agg[i], 4, 1, !i2p, rs->unit_base);
-		min = num2str(rs->min_bw[i], 4, 1, i2p, rs->unit_base);
-		minalt = num2str(rs->min_bw[i], 4, 1, !i2p, rs->unit_base);
-		max = num2str(rs->max_bw[i], 4, 1, i2p, rs->unit_base);
-		maxalt = num2str(rs->max_bw[i], 4, 1, !i2p, rs->unit_base);
+		io = num2str(rs->iobytes[i], rs->sig_figs, 1, i2p, N2S_BYTE);
+		ioalt = num2str(rs->iobytes[i], rs->sig_figs, 1, !i2p, N2S_BYTE);
+		agg = num2str(rs->agg[i], rs->sig_figs, 1, i2p, rs->unit_base);
+		aggalt = num2str(rs->agg[i], rs->sig_figs, 1, !i2p, rs->unit_base);
+		min = num2str(rs->min_bw[i], rs->sig_figs, 1, i2p, rs->unit_base);
+		minalt = num2str(rs->min_bw[i], rs->sig_figs, 1, !i2p, rs->unit_base);
+		max = num2str(rs->max_bw[i], rs->sig_figs, 1, i2p, rs->unit_base);
+		maxalt = num2str(rs->max_bw[i], rs->sig_figs, 1, !i2p, rs->unit_base);
 		log_buf(out, "%s: bw=%s (%s), %s-%s (%s-%s), io=%s (%s), run=%llu-%llumsec\n",
 				rs->unified_rw_rep ? "  MIXED" : str[i],
 				agg, aggalt, min, max, minalt, maxalt, io, ioalt,
@@ -317,7 +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;
 
@@ -336,7 +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;
@@ -355,6 +359,28 @@ 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
+ */
+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;
+
+	stat_calc_lat(ts, io_u_lat_u, ts->io_u_lat_u, FIO_IO_U_LAT_U_NR);
+
+	for (i = 0; i < FIO_IO_U_LAT_N_NR; i++)
+		ntotal += ts->io_u_lat_n[i];
+
+	io_u_lat_u[0] += 100.0 * (double) ntotal / (double) total;
+}
+
+void stat_calc_lat_n(struct thread_stat *ts, double *io_u_lat)
+{
+	stat_calc_lat(ts, io_u_lat, ts->io_u_lat_n, FIO_IO_U_LAT_N_NR);
+}
+
 void stat_calc_lat_u(struct thread_stat *ts, double *io_u_lat)
 {
 	stat_calc_lat(ts, io_u_lat, ts->io_u_lat_u, FIO_IO_U_LAT_U_NR);
@@ -365,14 +391,17 @@ void stat_calc_lat_m(struct thread_stat *ts, double *io_u_lat)
 	stat_calc_lat(ts, io_u_lat, ts->io_u_lat_m, FIO_IO_U_LAT_M_NR);
 }
 
-static void display_lat(const char *name, unsigned long min, unsigned long max,
-			double mean, double dev, struct buf_output *out)
+static void display_lat(const char *name, unsigned long long min,
+			unsigned long long max, double mean, double dev,
+			struct buf_output *out)
 {
-	const char *base = "(usec)";
+	const char *base = "(nsec)";
 	char *minp, *maxp;
 
-	if (usec_to_msec(&min, &max, &mean, &dev))
+	if (nsec_to_msec(&min, &max, &mean, &dev))
 		base = "(msec)";
+	else if (nsec_to_usec(&min, &max, &mean, &dev))
+		base = "(usec)";
 
 	minp = num2str(min, 6, 1, 0, N2S_NONE);
 	maxp = num2str(max, 6, 1, 0, N2S_NONE);
@@ -387,13 +416,26 @@ static void display_lat(const char *name, unsigned long min, unsigned long max,
 static void show_ddir_status(struct group_run_stats *rs, struct thread_stat *ts,
 			     int ddir, struct buf_output *out)
 {
-	const char *str[] = { " read", "write", " trim" };
-	unsigned long min, max, runt;
-	unsigned long long bw, iops;
+	const char *str[] = { " read", "write", " trim", "sync" };
+	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)) {
+		if (calc_lat(&ts->sync_stat, &min, &max, &mean, &dev)) {
+			log_buf(out, "  %s:\n", "fsync/fdatasync/sync_file_range");
+			display_lat(str[ddir], min, max, mean, dev, out);
+			show_clat_percentiles(ts->io_u_sync_plat,
+						ts->sync_stat.samples,
+						ts->percentile_list,
+						ts->percentile_precision,
+						str[ddir], out);
+		}
+		return;
+	}
+
 	assert(ddir_rw(ddir));
 
 	if (!ts->runtime[ddir])
@@ -403,18 +445,22 @@ static void show_ddir_status(struct group_run_stats *rs, struct thread_stat *ts,
 	runt = ts->runtime[ddir];
 
 	bw = (1000 * ts->io_bytes[ddir]) / runt;
-	io_p = num2str(ts->io_bytes[ddir], 4, 1, i2p, N2S_BYTE);
-	bw_p = num2str(bw, 4, 1, i2p, ts->unit_base);
-	bw_p_alt = num2str(bw, 4, 1, !i2p, ts->unit_base);
+	io_p = num2str(ts->io_bytes[ddir], ts->sig_figs, 1, i2p, N2S_BYTE);
+	bw_p = num2str(bw, ts->sig_figs, 1, i2p, ts->unit_base);
+	bw_p_alt = num2str(bw, ts->sig_figs, 1, !i2p, ts->unit_base);
 
 	iops = (1000 * (uint64_t)ts->total_io_u[ddir]) / runt;
-	iops_p = num2str(iops, 4, 1, 0, N2S_NONE);
+	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);
@@ -427,11 +473,19 @@ static void show_ddir_status(struct group_run_stats *rs, struct thread_stat *ts,
 	if (calc_lat(&ts->lat_stat[ddir], &min, &max, &mean, &dev))
 		display_lat(" lat", min, max, mean, dev, out);
 
-	if (ts->clat_percentiles) {
+	if (ts->clat_percentiles || ts->lat_percentiles) {
+		const char *name = ts->clat_percentiles ? "clat" : " lat";
+		uint64_t samples;
+
+		if (ts->clat_percentiles)
+			samples = ts->clat_stat[ddir].samples;
+		else
+			samples = ts->lat_stat[ddir].samples;
+
 		show_clat_percentiles(ts->io_u_plat[ddir],
-					ts->clat_stat[ddir].samples,
+					samples,
 					ts->percentile_list,
-					ts->percentile_precision, out);
+					ts->percentile_precision, name, out);
 	}
 	if (calc_lat(&ts->bw_stat[ddir], &min, &max, &mean, &dev)) {
 		double p_of_agg = 100.0, fkb_base = (double)rs->kb_base;
@@ -446,6 +500,12 @@ static void show_ddir_status(struct group_run_stats *rs, struct thread_stat *ts,
 		else
 			bw_str = "kB";
 
+		if (rs->agg[ddir]) {
+			p_of_agg = mean * 100 / (double) (rs->agg[ddir] / 1024);
+			if (p_of_agg > 100.0)
+				p_of_agg = 100.0;
+		}
+
 		if (rs->unit_base == 1) {
 			min *= 8.0;
 			max *= 8.0;
@@ -453,12 +513,6 @@ static void show_ddir_status(struct group_run_stats *rs, struct thread_stat *ts,
 			dev *= 8.0;
 		}
 
-		if (rs->agg[ddir]) {
-			p_of_agg = mean * 100 / (double) rs->agg[ddir];
-			if (p_of_agg > 100.0)
-				p_of_agg = 100.0;
-		}
-
 		if (mean > fkb_base * fkb_base) {
 			min /= fkb_base;
 			max /= fkb_base;
@@ -467,25 +521,33 @@ static void show_ddir_status(struct group_run_stats *rs, struct thread_stat *ts,
 			bw_str = (rs->unit_base == 1 ? "Mibit" : "MiB");
 		}
 
-		log_buf(out, "   bw (%5s/s): min=%5lu, max=%5lu, per=%3.2f%%, avg=%5.02f, stdev=%5.02f\n",
-			bw_str, min, max, p_of_agg, mean, dev);
+		log_buf(out, "   bw (%5s/s): min=%5llu, max=%5llu, per=%3.2f%%, "
+			"avg=%5.02f, stdev=%5.02f, samples=%" PRIu64 "\n",
+			bw_str, min, max, p_of_agg, mean, dev,
+			(&ts->bw_stat[ddir])->samples);
+	}
+	if (calc_lat(&ts->iops_stat[ddir], &min, &max, &mean, &dev)) {
+		log_buf(out, "   iops        : min=%5llu, max=%5llu, "
+			"avg=%5.02f, stdev=%5.02f, samples=%" PRIu64 "\n",
+			min, max, mean, dev, (&ts->iops_stat[ddir])->samples);
 	}
 }
 
-static int show_lat(double *io_u_lat, int nr, const char **ranges,
-		    const char *msg, struct buf_output *out)
+static bool show_lat(double *io_u_lat, int nr, const char **ranges,
+		     const char *msg, struct buf_output *out)
 {
-	int new_line = 1, i, line = 0, shown = 0;
+	bool new_line = true, shown = false;
+	int i, line = 0;
 
 	for (i = 0; i < nr; i++) {
 		if (io_u_lat[i] <= 0.0)
 			continue;
-		shown = 1;
+		shown = true;
 		if (new_line) {
 			if (line)
 				log_buf(out, "\n");
-			log_buf(out, "    lat (%s) : ", msg);
-			new_line = 0;
+			log_buf(out, "  lat (%s)   : ", msg);
+			new_line = false;
 			line = 0;
 		}
 		if (line)
@@ -493,13 +555,21 @@ static int show_lat(double *io_u_lat, int nr, const char **ranges,
 		log_buf(out, "%s%3.2f%%", ranges[i], io_u_lat[i]);
 		line++;
 		if (line == 5)
-			new_line = 1;
+			new_line = true;
 	}
 
 	if (shown)
 		log_buf(out, "\n");
 
-	return shown;
+	return true;
+}
+
+static void show_lat_n(double *io_u_lat_n, struct buf_output *out)
+{
+	const char *ranges[] = { "2=", "4=", "10=", "20=", "50=", "100=",
+				 "250=", "500=", "750=", "1000=", };
+
+	show_lat(io_u_lat_n, FIO_IO_U_LAT_N_NR, ranges, "nsec", out);
 }
 
 static void show_lat_u(double *io_u_lat_u, struct buf_output *out)
@@ -521,12 +591,15 @@ static void show_lat_m(double *io_u_lat_m, struct buf_output *out)
 
 static void show_latencies(struct thread_stat *ts, struct buf_output *out)
 {
+	double io_u_lat_n[FIO_IO_U_LAT_N_NR];
 	double io_u_lat_u[FIO_IO_U_LAT_U_NR];
 	double io_u_lat_m[FIO_IO_U_LAT_M_NR];
 
+	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);
 
+	show_lat_n(io_u_lat_n, out);
 	show_lat_u(io_u_lat_u, out);
 	show_lat_m(io_u_lat_m, out);
 }
@@ -551,8 +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);
@@ -602,7 +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
@@ -686,17 +758,17 @@ static void show_ss_normal(struct thread_stat *ts, struct buf_output *out)
 	bw_mean = steadystate_bw_mean(ts);
 	iops_mean = steadystate_iops_mean(ts);
 
-	p1 = num2str(bw_mean / ts->kb_base, 4, ts->kb_base, i2p, ts->unit_base);
-	p1alt = num2str(bw_mean / ts->kb_base, 4, ts->kb_base, !i2p, ts->unit_base);
-	p2 = num2str(iops_mean, 4, 1, 0, N2S_NONE);
+	p1 = num2str(bw_mean / ts->kb_base, ts->sig_figs, ts->kb_base, i2p, ts->unit_base);
+	p1alt = num2str(bw_mean / ts->kb_base, ts->sig_figs, ts->kb_base, !i2p, ts->unit_base);
+	p2 = num2str(iops_mean, ts->sig_figs, 1, 0, N2S_NONE);
 
 	log_buf(out, "  steadystate  : attained=%s, bw=%s (%s), iops=%s, %s%s=%.3f%s\n",
-		ts->ss_state & __FIO_SS_ATTAINED ? "yes" : "no",
+		ts->ss_state & FIO_SS_ATTAINED ? "yes" : "no",
 		p1, p1alt, p2,
-		ts->ss_state & __FIO_SS_IOPS ? "iops" : "bw",
-		ts->ss_state & __FIO_SS_SLOPE ? " slope": " mean dev",
+		ts->ss_state & FIO_SS_IOPS ? "iops" : "bw",
+		ts->ss_state & FIO_SS_SLOPE ? " slope": " mean dev",
 		ts->ss_criterion.u.f,
-		ts->ss_state & __FIO_SS_PCT ? "%" : "");
+		ts->ss_state & FIO_SS_PCT ? "%" : "");
 
 	free(p1);
 	free(p1alt);
@@ -744,6 +816,9 @@ static void show_thread_status_normal(struct thread_stat *ts,
 
 	show_latencies(ts, out);
 
+	if (ts->sync_stat.samples)
+		show_ddir_status(rs, ts, DDIR_SYNC, out);
+
 	runtime = ts->total_run_time;
 	if (runtime) {
 		double runt = (double) runtime;
@@ -780,12 +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],
@@ -816,14 +892,13 @@ static void show_thread_status_normal(struct thread_stat *ts,
 
 static void show_ddir_status_terse(struct thread_stat *ts,
 				   struct group_run_stats *rs, int ddir,
-				   struct buf_output *out)
+				   int ver, struct buf_output *out)
 {
-	unsigned long min, max;
-	unsigned long long bw, iops;
-	unsigned int *ovals = NULL;
+	unsigned long long min, max, minv, maxv, bw, iops;
+	unsigned long long *ovals = NULL;
 	double mean, dev;
-	unsigned int len, minv, maxv;
-	int i;
+	unsigned int len;
+	int i, bw_stat;
 
 	assert(ddir_rw(ddir));
 
@@ -840,16 +915,16 @@ static void show_ddir_status_terse(struct thread_stat *ts,
 					(unsigned long long) ts->runtime[ddir]);
 
 	if (calc_lat(&ts->slat_stat[ddir], &min, &max, &mean, &dev))
-		log_buf(out, ";%lu;%lu;%f;%f", min, max, mean, dev);
+		log_buf(out, ";%llu;%llu;%f;%f", min/1000, max/1000, mean/1000, dev/1000);
 	else
-		log_buf(out, ";%lu;%lu;%f;%f", 0UL, 0UL, 0.0, 0.0);
+		log_buf(out, ";%llu;%llu;%f;%f", 0ULL, 0ULL, 0.0, 0.0);
 
 	if (calc_lat(&ts->clat_stat[ddir], &min, &max, &mean, &dev))
-		log_buf(out, ";%lu;%lu;%f;%f", min, max, mean, dev);
+		log_buf(out, ";%llu;%llu;%f;%f", min/1000, max/1000, mean/1000, dev/1000);
 	else
-		log_buf(out, ";%lu;%lu;%f;%f", 0UL, 0UL, 0.0, 0.0);
+		log_buf(out, ";%llu;%llu;%f;%f", 0ULL, 0ULL, 0.0, 0.0);
 
-	if (ts->clat_percentiles) {
+	if (ts->clat_percentiles || ts->lat_percentiles) {
 		len = calc_clat_percentiles(ts->io_u_plat[ddir],
 					ts->clat_stat[ddir].samples,
 					ts->percentile_list, &ovals, &maxv,
@@ -862,46 +937,60 @@ static void show_ddir_status_terse(struct thread_stat *ts,
 			log_buf(out, ";0%%=0");
 			continue;
 		}
-		log_buf(out, ";%f%%=%u", ts->percentile_list[i].u.f, ovals[i]);
+		log_buf(out, ";%f%%=%llu", ts->percentile_list[i].u.f, ovals[i]/1000);
 	}
 
 	if (calc_lat(&ts->lat_stat[ddir], &min, &max, &mean, &dev))
-		log_buf(out, ";%lu;%lu;%f;%f", min, max, mean, dev);
+		log_buf(out, ";%llu;%llu;%f;%f", min/1000, max/1000, mean/1000, dev/1000);
 	else
-		log_buf(out, ";%lu;%lu;%f;%f", 0UL, 0UL, 0.0, 0.0);
+		log_buf(out, ";%llu;%llu;%f;%f", 0ULL, 0ULL, 0.0, 0.0);
 
 	if (ovals)
 		free(ovals);
 
-	if (calc_lat(&ts->bw_stat[ddir], &min, &max, &mean, &dev)) {
+	bw_stat = calc_lat(&ts->bw_stat[ddir], &min, &max, &mean, &dev);
+	if (bw_stat) {
 		double p_of_agg = 100.0;
 
 		if (rs->agg[ddir]) {
-			p_of_agg = mean * 100 / (double) rs->agg[ddir];
+			p_of_agg = mean * 100 / (double) (rs->agg[ddir] / 1024);
 			if (p_of_agg > 100.0)
 				p_of_agg = 100.0;
 		}
 
-		log_buf(out, ";%lu;%lu;%f%%;%f;%f", min, max, p_of_agg, mean, dev);
+		log_buf(out, ";%llu;%llu;%f%%;%f;%f", min, max, p_of_agg, mean, dev);
 	} else
-		log_buf(out, ";%lu;%lu;%f%%;%f;%f", 0UL, 0UL, 0.0, 0.0, 0.0);
+		log_buf(out, ";%llu;%llu;%f%%;%f;%f", 0ULL, 0ULL, 0.0, 0.0, 0.0);
+
+	if (ver == 5) {
+		if (bw_stat)
+			log_buf(out, ";%" PRIu64, (&ts->bw_stat[ddir])->samples);
+		else
+			log_buf(out, ";%lu", 0UL);
+
+		if (calc_lat(&ts->iops_stat[ddir], &min, &max, &mean, &dev))
+			log_buf(out, ";%llu;%llu;%f;%f;%" PRIu64, min, max,
+				mean, dev, (&ts->iops_stat[ddir])->samples);
+		else
+			log_buf(out, ";%llu;%llu;%f;%f;%lu", 0ULL, 0ULL, 0.0, 0.0, 0UL);
+	}
 }
 
 static void add_ddir_status_json(struct thread_stat *ts,
 		struct group_run_stats *rs, int ddir, struct json_object *parent)
 {
-	unsigned long min, max;
-	unsigned long long bw;
-	unsigned int *ovals = NULL;
+	unsigned long long min, max, minv, maxv;
+	unsigned long long bw_bytes, bw;
+	unsigned long long *ovals = NULL;
 	double mean, dev, iops;
-	unsigned int len, minv, maxv;
+	unsigned int len;
 	int i;
-	const char *ddirname[] = {"read", "write", "trim"};
-	struct json_object *dir_object, *tmp_object, *percentile_object, *clat_bins_object;
+	const char *ddirname[] = { "read", "write", "trim", "sync" };
+	struct json_object *dir_object, *tmp_object, *percentile_object, *clat_bins_object = NULL;
 	char buf[120];
 	double p_of_agg = 100.0;
 
-	assert(ddir_rw(ddir));
+	assert(ddir_rw(ddir) || ddir_sync(ddir));
 
 	if (ts->unified_rw_rep && ddir != DDIR_READ)
 		return;
@@ -910,92 +999,137 @@ static void add_ddir_status_json(struct thread_stat *ts,
 	json_object_add_value_object(parent,
 		ts->unified_rw_rep ? "mixed" : ddirname[ddir], dir_object);
 
-	bw = 0;
-	iops = 0.0;
-	if (ts->runtime[ddir]) {
-		uint64_t runt = ts->runtime[ddir];
+	if (ddir_rw(ddir)) {
+		bw_bytes = 0;
+		bw = 0;
+		iops = 0.0;
+		if (ts->runtime[ddir]) {
+			uint64_t runt = ts->runtime[ddir];
 
-		bw = ((1000 * ts->io_bytes[ddir]) / runt) / 1024; /* KiB/s */
-		iops = (1000.0 * (uint64_t) ts->total_io_u[ddir]) / runt;
-	}
+			bw_bytes = ((1000 * ts->io_bytes[ddir]) / runt); /* Bytes/s */
+			bw = bw_bytes / 1024; /* KiB/s */
+			iops = (1000.0 * (uint64_t) ts->total_io_u[ddir]) / runt;
+		}
 
-	json_object_add_value_int(dir_object, "io_bytes", ts->io_bytes[ddir] >> 10);
-	json_object_add_value_int(dir_object, "bw", bw);
-	json_object_add_value_float(dir_object, "iops", iops);
-	json_object_add_value_int(dir_object, "runtime", ts->runtime[ddir]);
-	json_object_add_value_int(dir_object, "total_ios", ts->total_io_u[ddir]);
-	json_object_add_value_int(dir_object, "short_ios", ts->short_io_u[ddir]);
-	json_object_add_value_int(dir_object, "drop_ios", ts->drop_io_u[ddir]);
+		json_object_add_value_int(dir_object, "io_bytes", ts->io_bytes[ddir]);
+		json_object_add_value_int(dir_object, "io_kbytes", ts->io_bytes[ddir] >> 10);
+		json_object_add_value_int(dir_object, "bw_bytes", bw_bytes);
+		json_object_add_value_int(dir_object, "bw", bw);
+		json_object_add_value_float(dir_object, "iops", iops);
+		json_object_add_value_int(dir_object, "runtime", ts->runtime[ddir]);
+		json_object_add_value_int(dir_object, "total_ios", ts->total_io_u[ddir]);
+		json_object_add_value_int(dir_object, "short_ios", ts->short_io_u[ddir]);
+		json_object_add_value_int(dir_object, "drop_ios", ts->drop_io_u[ddir]);
+
+		if (!calc_lat(&ts->slat_stat[ddir], &min, &max, &mean, &dev)) {
+			min = max = 0;
+			mean = dev = 0.0;
+		}
+		tmp_object = json_create_object();
+		json_object_add_value_object(dir_object, "slat_ns", tmp_object);
+		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);
+		json_object_add_value_float(tmp_object, "stddev", dev);
+
+		if (!calc_lat(&ts->clat_stat[ddir], &min, &max, &mean, &dev)) {
+			min = max = 0;
+			mean = dev = 0.0;
+		}
+		tmp_object = json_create_object();
+		json_object_add_value_object(dir_object, "clat_ns", tmp_object);
+		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);
+		json_object_add_value_float(tmp_object, "stddev", dev);
+	} else {
+		if (!calc_lat(&ts->sync_stat, &min, &max, &mean, &dev)) {
+			min = max = 0;
+			mean = dev = 0.0;
+		}
 
-	if (!calc_lat(&ts->slat_stat[ddir], &min, &max, &mean, &dev)) {
-		min = max = 0;
-		mean = dev = 0.0;
+		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);
+		json_object_add_value_float(tmp_object, "stddev", dev);
 	}
-	tmp_object = json_create_object();
-	json_object_add_value_object(dir_object, "slat", tmp_object);
-	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);
-	json_object_add_value_float(tmp_object, "stddev", dev);
 
-	if (!calc_lat(&ts->clat_stat[ddir], &min, &max, &mean, &dev)) {
-		min = max = 0;
-		mean = dev = 0.0;
-	}
-	tmp_object = json_create_object();
-	json_object_add_value_object(dir_object, "clat", tmp_object);
-	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);
-	json_object_add_value_float(tmp_object, "stddev", dev);
+	if (ts->clat_percentiles || ts->lat_percentiles) {
+		if (ddir_rw(ddir)) {
+			uint64_t samples;
 
-	if (ts->clat_percentiles) {
-		len = calc_clat_percentiles(ts->io_u_plat[ddir],
-					ts->clat_stat[ddir].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],
+					samples, ts->percentile_list, &ovals,
+					&maxv, &minv);
+		} else {
+			len = calc_clat_percentiles(ts->io_u_sync_plat,
+					ts->sync_stat.samples,
 					ts->percentile_list, &ovals, &maxv,
 					&minv);
+		}
+
+		if (len > FIO_IO_U_LIST_MAX_LEN)
+			len = FIO_IO_U_LIST_MAX_LEN;
 	} else
 		len = 0;
 
 	percentile_object = json_create_object();
 	json_object_add_value_object(tmp_object, "percentile", percentile_object);
-	for (i = 0; i < FIO_IO_U_LIST_MAX_LEN; i++) {
-		if (i >= len) {
-			json_object_add_value_int(percentile_object, "0.00", 0);
-			continue;
-		}
+	for (i = 0; i < len; i++) {
 		snprintf(buf, sizeof(buf), "%f", ts->percentile_list[i].u.f);
 		json_object_add_value_int(percentile_object, (const char *)buf, ovals[i]);
 	}
 
 	if (output_format & FIO_OUTPUT_JSON_PLUS) {
 		clat_bins_object = json_create_object();
-		json_object_add_value_object(tmp_object, "bins", clat_bins_object);
+		if (ts->clat_percentiles)
+			json_object_add_value_object(tmp_object, "bins", clat_bins_object);
+
 		for(i = 0; i < FIO_IO_U_PLAT_NR; i++) {
-			snprintf(buf, sizeof(buf), "%d", i);
-			json_object_add_value_int(clat_bins_object, (const char *)buf, ts->io_u_plat[ddir][i]);
+			if (ddir_rw(ddir)) {
+				if (ts->io_u_plat[ddir][i]) {
+					snprintf(buf, sizeof(buf), "%llu", plat_idx_to_val(i));
+					json_object_add_value_int(clat_bins_object, (const char *)buf, ts->io_u_plat[ddir][i]);
+				}
+			} else {
+				if (ts->io_u_sync_plat[i]) {
+					snprintf(buf, sizeof(buf), "%llu", plat_idx_to_val(i));
+					json_object_add_value_int(clat_bins_object, (const char *)buf, ts->io_u_sync_plat[i]);
+				}
+			}
 		}
-		json_object_add_value_int(clat_bins_object, "FIO_IO_U_PLAT_BITS", FIO_IO_U_PLAT_BITS);
-		json_object_add_value_int(clat_bins_object, "FIO_IO_U_PLAT_VAL", FIO_IO_U_PLAT_VAL);
-		json_object_add_value_int(clat_bins_object, "FIO_IO_U_PLAT_NR", FIO_IO_U_PLAT_NR);
 	}
 
+	if (!ddir_rw(ddir))
+		return;
+
 	if (!calc_lat(&ts->lat_stat[ddir], &min, &max, &mean, &dev)) {
 		min = max = 0;
 		mean = dev = 0.0;
 	}
 	tmp_object = json_create_object();
-	json_object_add_value_object(dir_object, "lat", tmp_object);
+	json_object_add_value_object(dir_object, "lat_ns", tmp_object);
 	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);
 	json_object_add_value_float(tmp_object, "stddev", dev);
+	if (output_format & FIO_OUTPUT_JSON_PLUS && ts->lat_percentiles)
+		json_object_add_value_object(tmp_object, "bins", clat_bins_object);
+
 	if (ovals)
 		free(ovals);
 
 	if (calc_lat(&ts->bw_stat[ddir], &min, &max, &mean, &dev)) {
 		if (rs->agg[ddir]) {
-			p_of_agg = mean * 100 / (double) rs->agg[ddir];
+			p_of_agg = mean * 100 / (double) (rs->agg[ddir] / 1024);
 			if (p_of_agg > 100.0)
 				p_of_agg = 100.0;
 		}
@@ -1008,74 +1142,24 @@ static void add_ddir_status_json(struct thread_stat *ts,
 	json_object_add_value_float(dir_object, "bw_agg", p_of_agg);
 	json_object_add_value_float(dir_object, "bw_mean", mean);
 	json_object_add_value_float(dir_object, "bw_dev", dev);
-}
+	json_object_add_value_int(dir_object, "bw_samples",
+				(&ts->bw_stat[ddir])->samples);
 
-static void show_thread_status_terse_v2(struct thread_stat *ts,
-					struct group_run_stats *rs,
-					struct buf_output *out)
-{
-	double io_u_dist[FIO_IO_U_MAP_NR];
-	double io_u_lat_u[FIO_IO_U_LAT_U_NR];
-	double io_u_lat_m[FIO_IO_U_LAT_M_NR];
-	double usr_cpu, sys_cpu;
-	int i;
-
-	/* General Info */
-	log_buf(out, "2;%s;%d;%d", ts->name, ts->groupid, ts->error);
-	/* Log Read Status */
-	show_ddir_status_terse(ts, rs, DDIR_READ, out);
-	/* Log Write Status */
-	show_ddir_status_terse(ts, rs, DDIR_WRITE, out);
-	/* Log Trim Status */
-	show_ddir_status_terse(ts, rs, DDIR_TRIM, out);
-
-	/* CPU Usage */
-	if (ts->total_run_time) {
-		double runt = (double) ts->total_run_time;
-
-		usr_cpu = (double) ts->usr_time * 100 / runt;
-		sys_cpu = (double) ts->sys_time * 100 / runt;
-	} else {
-		usr_cpu = 0;
-		sys_cpu = 0;
+	if (!calc_lat(&ts->iops_stat[ddir], &min, &max, &mean, &dev)) {
+		min = max = 0;
+		mean = dev = 0.0;
 	}
-
-	log_buf(out, ";%f%%;%f%%;%llu;%llu;%llu", usr_cpu, sys_cpu,
-						(unsigned long long) ts->ctx,
-						(unsigned long long) ts->majf,
-						(unsigned long long) ts->minf);
-
-	/* Calc % distribution of IO depths, usecond, msecond latency */
-	stat_calc_dist(ts->io_u_map, ddir_rw_sum(ts->total_io_u), io_u_dist);
-	stat_calc_lat_u(ts, io_u_lat_u);
-	stat_calc_lat_m(ts, io_u_lat_m);
-
-	/* Only show fixed 7 I/O depth levels*/
-	log_buf(out, ";%3.1f%%;%3.1f%%;%3.1f%%;%3.1f%%;%3.1f%%;%3.1f%%;%3.1f%%",
-			io_u_dist[0], io_u_dist[1], io_u_dist[2], io_u_dist[3],
-			io_u_dist[4], io_u_dist[5], io_u_dist[6]);
-
-	/* Microsecond latency */
-	for (i = 0; i < FIO_IO_U_LAT_U_NR; i++)
-		log_buf(out, ";%3.2f%%", io_u_lat_u[i]);
-	/* Millisecond latency */
-	for (i = 0; i < FIO_IO_U_LAT_M_NR; i++)
-		log_buf(out, ";%3.2f%%", io_u_lat_m[i]);
-	/* Additional output if continue_on_error set - default off*/
-	if (ts->continue_on_error)
-		log_buf(out, ";%llu;%d", (unsigned long long) ts->total_err_count, ts->first_error);
-	log_buf(out, "\n");
-
-	/* Additional output if description is set */
-	if (strlen(ts->description))
-		log_buf(out, ";%s", ts->description);
-
-	log_buf(out, "\n");
+	json_object_add_value_int(dir_object, "iops_min", min);
+	json_object_add_value_int(dir_object, "iops_max", max);
+	json_object_add_value_float(dir_object, "iops_mean", mean);
+	json_object_add_value_float(dir_object, "iops_stddev", dev);
+	json_object_add_value_int(dir_object, "iops_samples",
+				(&ts->iops_stat[ddir])->samples);
 }
 
-static void show_thread_status_terse_v3_v4(struct thread_stat *ts,
-					   struct group_run_stats *rs, int ver,
-					   struct buf_output *out)
+static void show_thread_status_terse_all(struct thread_stat *ts,
+					 struct group_run_stats *rs, int ver,
+					 struct buf_output *out)
 {
 	double io_u_dist[FIO_IO_U_MAP_NR];
 	double io_u_lat_u[FIO_IO_U_LAT_U_NR];
@@ -1084,15 +1168,19 @@ static void show_thread_status_terse_v3_v4(struct thread_stat *ts,
 	int i;
 
 	/* General Info */
-	log_buf(out, "%d;%s;%s;%d;%d", ver, fio_version_string,
-					ts->name, ts->groupid, ts->error);
+	if (ver == 2)
+		log_buf(out, "2;%s;%d;%d", ts->name, ts->groupid, ts->error);
+	else
+		log_buf(out, "%d;%s;%s;%d;%d", ver, fio_version_string,
+			ts->name, ts->groupid, ts->error);
+
 	/* Log Read Status */
-	show_ddir_status_terse(ts, rs, DDIR_READ, out);
+	show_ddir_status_terse(ts, rs, DDIR_READ, ver, out);
 	/* Log Write Status */
-	show_ddir_status_terse(ts, rs, DDIR_WRITE, out);
+	show_ddir_status_terse(ts, rs, DDIR_WRITE, ver, out);
 	/* Log Trim Status */
-	if (ver == 4)
-		show_ddir_status_terse(ts, rs, DDIR_TRIM, out);
+	if (ver == 2 || ver == 4 || ver == 5)
+		show_ddir_status_terse(ts, rs, DDIR_TRIM, ver, out);
 
 	/* CPU Usage */
 	if (ts->total_run_time) {
@@ -1112,7 +1200,7 @@ static void show_thread_status_terse_v3_v4(struct thread_stat *ts,
 
 	/* Calc % distribution of IO depths, usecond, msecond latency */
 	stat_calc_dist(ts->io_u_map, ddir_rw_sum(ts->total_io_u), io_u_dist);
-	stat_calc_lat_u(ts, io_u_lat_u);
+	stat_calc_lat_nu(ts, io_u_lat_u);
 	stat_calc_lat_m(ts, io_u_lat_m);
 
 	/* Only show fixed 7 I/O depth levels*/
@@ -1128,11 +1216,14 @@ static void show_thread_status_terse_v3_v4(struct thread_stat *ts,
 		log_buf(out, ";%3.2f%%", io_u_lat_m[i]);
 
 	/* disk util stats, if any */
-	show_disk_util(1, NULL, out);
+	if (ver >= 3 && is_running_backend())
+		show_disk_util(1, NULL, out);
 
 	/* Additional output if continue_on_error set - default off*/
 	if (ts->continue_on_error)
 		log_buf(out, ";%llu;%d", (unsigned long long) ts->total_err_count, ts->first_error);
+	if (ver == 2)
+		log_buf(out, "\n");
 
 	/* Additional output if description is set */
 	if (strlen(ts->description))
@@ -1142,7 +1233,7 @@ static void show_thread_status_terse_v3_v4(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;
@@ -1158,8 +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);
@@ -1173,6 +1262,7 @@ static struct json_object *show_thread_status_json(struct thread_stat *ts,
 	struct json_object *root, *tmp;
 	struct jobs_eta *je;
 	double io_u_dist[FIO_IO_U_MAP_NR];
+	double io_u_lat_n[FIO_IO_U_LAT_N_NR];
 	double io_u_lat_u[FIO_IO_U_LAT_U_NR];
 	double io_u_lat_m[FIO_IO_U_LAT_M_NR];
 	double usr_cpu, sys_cpu;
@@ -1192,11 +1282,12 @@ 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);
 	add_ddir_status_json(ts, rs, DDIR_TRIM, root);
+	add_ddir_status_json(ts, rs, DDIR_SYNC, root);
 
 	/* CPU Usage */
 	if (ts->total_run_time) {
@@ -1208,18 +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_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*/
@@ -1232,9 +1320,55 @@ 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, "latency_us", tmp);
+	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);
+	for (i = 0; i < FIO_IO_U_LAT_N_NR; i++) {
+		const char *ranges[] = { "2", "4", "10", "20", "50", "100",
+				 "250", "500", "750", "1000", };
+		json_object_add_value_float(tmp, ranges[i], io_u_lat_n[i]);
+	}
 	/* Microsecond latency */
+	tmp = json_create_object();
+	json_object_add_value_object(root, "latency_us", tmp);
 	for (i = 0; i < FIO_IO_U_LAT_U_NR; i++) {
 		const char *ranges[] = { "2", "4", "10", "20", "50", "100",
 				 "250", "500", "750", "1000", };
@@ -1309,23 +1443,23 @@ 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",
-			ts->ss_state & __FIO_SS_IOPS ? "iops" : "bw",
-			ts->ss_state & __FIO_SS_SLOPE ? "_slope" : "",
+			ts->ss_state & FIO_SS_IOPS ? "iops" : "bw",
+			ts->ss_state & FIO_SS_SLOPE ? "_slope" : "",
 			(float) ts->ss_limit.u.f,
-			ts->ss_state & __FIO_SS_PCT ? "%" : "");
+			ts->ss_state & FIO_SS_PCT ? "%" : "");
 
 		tmp = json_create_object();
 		json_object_add_value_object(root, "steadystate", tmp);
 		json_object_add_value_string(tmp, "ss", ss_buf);
 		json_object_add_value_int(tmp, "duration", (int)ts->ss_dur);
-		json_object_add_value_int(tmp, "attained", (ts->ss_state & __FIO_SS_ATTAINED) > 0);
+		json_object_add_value_int(tmp, "attained", (ts->ss_state & FIO_SS_ATTAINED) > 0);
 
 		snprintf(ss_buf, sizeof(ss_buf), "%f%s", (float) ts->ss_criterion.u.f,
-			ts->ss_state & __FIO_SS_PCT ? "%" : "");
+			ts->ss_state & FIO_SS_PCT ? "%" : "");
 		json_object_add_value_string(tmp, "criterion", ss_buf);
 		json_object_add_value_float(tmp, "max_deviation", ts->ss_deviation.u.f);
 		json_object_add_value_float(tmp, "slope", ts->ss_slope.u.f);
@@ -1341,12 +1475,12 @@ static struct json_object *show_thread_status_json(struct thread_stat *ts,
 		** otherwise it actually points to the second element
 		** in the list
 		*/
-		if ((ts->ss_state & __FIO_SS_ATTAINED) || !(ts->ss_state & __FIO_SS_BUFFER_FULL))
+		if ((ts->ss_state & FIO_SS_ATTAINED) || !(ts->ss_state & FIO_SS_BUFFER_FULL))
 			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]);
 		}
@@ -1363,10 +1497,8 @@ static void show_thread_status_terse(struct thread_stat *ts,
 				     struct group_run_stats *rs,
 				     struct buf_output *out)
 {
-	if (terse_version == 2)
-		show_thread_status_terse_v2(ts, rs, out);
-	else if (terse_version == 3 || terse_version == 4)
-		show_thread_status_terse_v3_v4(ts, rs, terse_version, out);
+	if (terse_version >= 2 && terse_version <= 5)
+		show_thread_status_terse_all(ts, rs, terse_version, out);
 	else
 		log_err("fio: bad terse version!? %d\n", terse_version);
 }
@@ -1445,6 +1577,8 @@ void sum_group_stats(struct group_run_stats *dst, struct group_run_stats *src)
 		dst->kb_base = src->kb_base;
 	if (!dst->unit_base)
 		dst->unit_base = src->unit_base;
+	if (!dst->sig_figs)
+		dst->sig_figs = src->sig_figs;
 }
 
 void sum_thread_stats(struct thread_stat *dst, struct thread_stat *src,
@@ -1458,6 +1592,7 @@ void sum_thread_stats(struct thread_stat *dst, struct thread_stat *src,
 			sum_stat(&dst->slat_stat[l], &src->slat_stat[l], first);
 			sum_stat(&dst->lat_stat[l], &src->lat_stat[l], first);
 			sum_stat(&dst->bw_stat[l], &src->bw_stat[l], first);
+			sum_stat(&dst->iops_stat[l], &src->iops_stat[l], first);
 
 			dst->io_bytes[l] += src->io_bytes[l];
 
@@ -1468,6 +1603,7 @@ void sum_thread_stats(struct thread_stat *dst, struct thread_stat *src,
 			sum_stat(&dst->slat_stat[0], &src->slat_stat[l], first);
 			sum_stat(&dst->lat_stat[0], &src->lat_stat[l], first);
 			sum_stat(&dst->bw_stat[0], &src->bw_stat[l], first);
+			sum_stat(&dst->iops_stat[0], &src->iops_stat[l], first);
 
 			dst->io_bytes[0] += src->io_bytes[l];
 
@@ -1482,22 +1618,25 @@ void sum_thread_stats(struct thread_stat *dst, struct thread_stat *src,
 		}
 	}
 
+	sum_stat(&dst->sync_stat, &src->sync_stat, first);
 	dst->usr_time += src->usr_time;
 	dst->sys_time += src->sys_time;
 	dst->ctx += src->ctx;
 	dst->majf += src->majf;
 	dst->minf += src->minf;
 
-	for (k = 0; k < FIO_IO_U_MAP_NR; k++)
+	for (k = 0; k < FIO_IO_U_MAP_NR; k++) {
 		dst->io_u_map[k] += src->io_u_map[k];
-	for (k = 0; k < FIO_IO_U_MAP_NR; k++)
 		dst->io_u_submit[k] += src->io_u_submit[k];
-	for (k = 0; k < FIO_IO_U_MAP_NR; k++)
 		dst->io_u_complete[k] += src->io_u_complete[k];
-	for (k = 0; k < FIO_IO_U_LAT_U_NR; k++)
+	}
+	for (k = 0; k < FIO_IO_U_LAT_N_NR; k++) {
+		dst->io_u_lat_n[k] += src->io_u_lat_n[k];
 		dst->io_u_lat_u[k] += src->io_u_lat_u[k];
-	for (k = 0; k < FIO_IO_U_LAT_M_NR; k++)
 		dst->io_u_lat_m[k] += src->io_u_lat_m[k];
+	}
+	for (k = 0; k < FIO_IO_U_PLAT_NR; k++)
+		dst->io_u_sync_plat[k] += src->io_u_sync_plat[k];
 
 	for (k = 0; k < DDIR_RWDIR_CNT; k++) {
 		if (!dst->unified_rw_rep) {
@@ -1511,6 +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;
 
@@ -1525,6 +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)
@@ -1547,7 +1689,9 @@ void init_thread_stat(struct thread_stat *ts)
 		ts->clat_stat[j].min_val = -1UL;
 		ts->slat_stat[j].min_val = -1UL;
 		ts->bw_stat[j].min_val = -1UL;
+		ts->iops_stat[j].min_val = -1UL;
 	}
+	ts->sync_stat.min_val = -1UL;
 	ts->groupid = -1;
 }
 
@@ -1557,8 +1701,8 @@ void __show_run_stats(void)
 	struct thread_data *td;
 	struct thread_stat *threadstats, *ts;
 	int i, j, k, nr_ts, last_ts, idx;
-	int kb_base_warned = 0;
-	int unit_base_warned = 0;
+	bool kb_base_warned = false;
+	bool unit_base_warned = false;
 	struct json_object *root = NULL;
 	struct json_array *array = NULL;
 	struct buf_output output[FIO_OUTPUT_NR];
@@ -1614,6 +1758,7 @@ void __show_run_stats(void)
 		ts = &threadstats[j];
 
 		ts->clat_percentiles = td->o.clat_percentiles;
+		ts->lat_percentiles = td->o.lat_percentiles;
 		ts->percentile_precision = td->o.percentile_precision;
 		memcpy(ts->percentile_list, td->o.percentile_list, sizeof(td->o.percentile_list));
 		opt_lists[j] = &td->opt_list;
@@ -1646,15 +1791,16 @@ void __show_run_stats(void)
 
 			ts->kb_base = td->o.kb_base;
 			ts->unit_base = td->o.unit_base;
+			ts->sig_figs = td->o.sig_figs;
 			ts->unified_rw_rep = td->o.unified_rw_rep;
 		} else if (ts->kb_base != td->o.kb_base && !kb_base_warned) {
 			log_info("fio: kb_base differs for jobs in group, using"
 				 " %u as the base\n", ts->kb_base);
-			kb_base_warned = 1;
+			kb_base_warned = true;
 		} else if (ts->unit_base != td->o.unit_base && !unit_base_warned) {
 			log_info("fio: unit_base differs for jobs in group, using"
 				 " %u as the base\n", ts->unit_base);
-			unit_base_warned = 1;
+			unit_base_warned = true;
 		}
 
 		ts->continue_on_error = td->o.continue_on_error;
@@ -1708,6 +1854,7 @@ void __show_run_stats(void)
 		rs = &runstats[ts->groupid];
 		rs->kb_base = ts->kb_base;
 		rs->unit_base = ts->unit_base;
+		rs->sig_figs = ts->sig_figs;
 		rs->unified_rw_rep += ts->unified_rw_rep;
 
 		for (j = 0; j < DDIR_RWDIR_CNT; j++) {
@@ -1755,13 +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';
 
@@ -1771,7 +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);
 	}
@@ -1826,43 +1974,40 @@ void __show_run_stats(void)
 	}
 
 	for (i = 0; i < FIO_OUTPUT_NR; i++) {
-		buf_output_flush(&output[i]);
-		buf_output_free(&output[i]);
+		struct buf_output *out = &output[i];
+
+		log_info_buf(out->buf, out->buflen);
+		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;
 	unsigned long long *rt;
-	struct timeval tv;
+	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(&tv, NULL);
+	fio_gettime(&ts, NULL);
 
 	for_each_td(td, i) {
 		td->update_rusage = 1;
 		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];
-		td->ts.total_run_time = mtime_since(&td->epoch, &tv);
+		td->ts.total_run_time = mtime_since(&td->epoch, &ts);
 
-		rt[i] = mtime_since(&td->start, &tv);
+		rt[i] = mtime_since(&td->start, &ts);
 		if (td_read(td) && td->ts.io_bytes[DDIR_READ])
 			td->ts.runtime[DDIR_READ] += rt[i];
 		if (td_write(td) && td->ts.io_bytes[DDIR_WRITE])
@@ -1876,7 +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;
 	}
@@ -1893,12 +2038,12 @@ void __show_running_run_stats(void)
 	}
 
 	free(rt);
-	fio_mutex_up(stat_mutex);
+	fio_sem_up(stat_sem);
 }
 
-static int status_interval_init;
-static struct timeval status_time;
-static int status_file_disabled;
+static bool status_interval_init;
+static struct timespec status_time;
+static bool status_file_disabled;
 
 #define FIO_STATUS_FILE		"fio-dump-status"
 
@@ -1929,7 +2074,7 @@ static int check_status_file(void)
 		log_err("fio: failed to unlink %s: %s\n", fio_status_file_path,
 							strerror(errno));
 		log_err("fio: disabling status file updates\n");
-		status_file_disabled = 1;
+		status_file_disabled = true;
 	}
 
 	return 1;
@@ -1940,7 +2085,7 @@ void check_for_running_stats(void)
 	if (status_interval) {
 		if (!status_interval_init) {
 			fio_gettime(&status_time, NULL);
-			status_interval_init = 1;
+			status_interval_init = true;
 		} else if (mtime_since_now(&status_time) >= status_interval) {
 			show_running_run_stats();
 			fio_gettime(&status_time, NULL);
@@ -1953,7 +2098,7 @@ void check_for_running_stats(void)
 	}
 }
 
-static inline void add_stat_sample(struct io_stat *is, unsigned long data)
+static inline void add_stat_sample(struct io_stat *is, unsigned long long data)
 {
 	double val = data;
 	double delta;
@@ -2108,17 +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;
@@ -2126,7 +2273,7 @@ static void __add_log_sample(struct io_log *iolog, union io_sample_data data,
 	if (iolog->disabled)
 		return;
 	if (flist_empty(&iolog->io_logs))
-		iolog->avg_last = t;
+		iolog->avg_last[ddir] = t;
 
 	cur_log = get_cur_log(iolog);
 	if (cur_log) {
@@ -2172,26 +2319,35 @@ void reset_io_stats(struct thread_data *td)
 
 		ts->io_bytes[i] = 0;
 		ts->runtime[i] = 0;
+		ts->total_io_u[i] = 0;
+		ts->short_io_u[i] = 0;
+		ts->drop_io_u[i] = 0;
 
-		for (j = 0; j < FIO_IO_U_PLAT_NR; j++)
+		for (j = 0; j < FIO_IO_U_PLAT_NR; j++) {
 			ts->io_u_plat[i][j] = 0;
+			if (!i)
+				ts->io_u_sync_plat[j] = 0;
+		}
 	}
 
+	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;
 		ts->io_u_complete[i] = 0;
+	}
+
+	for (i = 0; i < FIO_IO_U_LAT_N_NR; i++)
+		ts->io_u_lat_n[i] = 0;
+	for (i = 0; i < FIO_IO_U_LAT_U_NR; i++)
 		ts->io_u_lat_u[i] = 0;
+	for (i = 0; i < FIO_IO_U_LAT_M_NR; i++)
 		ts->io_u_lat_m[i] = 0;
-		ts->total_submit = 0;
-		ts->total_complete = 0;
-	}
 
-	for (i = 0; i < 3; i++) {
-		ts->total_io_u[i] = 0;
-		ts->short_io_u[i] = 0;
-		ts->drop_io_u[i] = 0;
-	}
+	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,
@@ -2225,9 +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;
 
@@ -2254,19 +2412,19 @@ static long add_log_sample(struct thread_data *td, struct io_log *iolog,
 	 * If period hasn't passed, adding the above sample is all we
 	 * need to do.
 	 */
-	this_window = elapsed - iolog->avg_last;
-	if (elapsed < iolog->avg_last)
-		return iolog->avg_last - elapsed;
+	this_window = elapsed - iolog->avg_last[ddir];
+	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;
 	}
 
-	_add_stat_to_log(iolog, elapsed, td->o.log_max != 0);
+	__add_stat_to_log(iolog, ddir, elapsed, td->o.log_max != 0);
 
-	iolog->avg_last = elapsed - (this_window - iolog->avg_msec);
+	iolog->avg_last[ddir] = elapsed - (this_window - iolog->avg_msec);
 	return iolog->avg_msec;
 }
 
@@ -2288,7 +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;
 
@@ -2299,32 +2457,44 @@ void add_agg_sample(union io_sample_data data, enum fio_ddir ddir, unsigned int
 	__add_log_sample(iolog, data, ddir, bs, mtime_since_genesis(), 0);
 }
 
+void add_sync_clat_sample(struct thread_stat *ts, unsigned long long nsec)
+{
+	unsigned int idx = plat_val_to_idx(nsec);
+	assert(idx < FIO_IO_U_PLAT_NR);
+
+	ts->io_u_sync_plat[idx]++;
+	add_stat_sample(&ts->sync_stat, nsec);
+}
+
 static void add_clat_percentile_sample(struct thread_stat *ts,
-				unsigned long usec, enum fio_ddir ddir)
+				unsigned long long nsec, enum fio_ddir ddir)
 {
-	unsigned int idx = plat_val_to_idx(usec);
+	unsigned int idx = plat_val_to_idx(nsec);
 	assert(idx < FIO_IO_U_PLAT_NR);
 
 	ts->io_u_plat[ddir][idx]++;
 }
 
 void add_clat_sample(struct thread_data *td, enum fio_ddir ddir,
-		     unsigned long usec, 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], usec);
+	add_stat_sample(&ts->clat_stat[ddir], nsec);
 
 	if (td->clat_log)
-		add_log_sample(td, td->clat_log, sample_val(usec), ddir, bs,
+		add_log_sample(td, td->clat_log, sample_val(nsec), ddir, bs,
 			       offset);
 
 	if (ts->clat_percentiles)
-		add_clat_percentile_sample(ts, usec, ddir);
+		add_clat_percentile_sample(ts, nsec, ddir);
 
 	if (iolog && iolog->hist_msec) {
 		struct io_hist *hw = &iolog->hist_window[ddir];
@@ -2336,7 +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;
 
 			/*
@@ -2346,7 +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));
@@ -2364,58 +2534,71 @@ 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 usec, 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], usec);
+	add_stat_sample(&ts->lat_stat[ddir], nsec);
 
 	if (td->lat_log)
-		add_log_sample(td, td->lat_log, sample_val(usec), ddir, bs,
+		add_log_sample(td, td->lat_log, sample_val(nsec), ddir, bs,
 			       offset);
 
-	td_io_u_unlock(td);
+	if (ts->lat_percentiles)
+		add_clat_percentile_sample(ts, nsec, ddir);
+
+	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 spent)
+		   unsigned int bytes, unsigned long long spent)
 {
+	const bool needs_lock = td_async_processing(td);
 	struct thread_stat *ts = &td->ts;
 	unsigned long rate;
 
 	if (spent)
-		rate = bytes * 1000 / spent;
+		rate = (unsigned long) (bytes * 1000000ULL / spent);
 	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);
 
@@ -2424,24 +2607,30 @@ 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_bw_samples(struct thread_data *td, struct timeval *t)
+static int __add_samples(struct thread_data *td, struct timespec *parent_tv,
+			 struct timespec *t, unsigned int avg_time,
+			 uint64_t *this_io_bytes, uint64_t *stat_io_bytes,
+			 struct io_stat *stat, struct io_log *log,
+			 bool is_kb)
 {
-	struct thread_stat *ts = &td->ts;
+	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 = td->o.bw_avg_time;
+	next_log = avg_time;
 
-	spent = mtime_since(&td->bw_sample_time, t);
-	if (spent < td->o.bw_avg_time &&
-	    td->o.bw_avg_time - spent >= LOG_MSEC_SLACK)
-		return td->o.bw_avg_time - spent;
+	spent = mtime_since(parent_tv, t);
+	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.
@@ -2449,48 +2638,61 @@ static int add_bw_samples(struct thread_data *td, struct timeval *t)
 	for (ddir = 0; ddir < DDIR_RWDIR_CNT; ddir++) {
 		uint64_t delta;
 
-		delta = td->this_io_bytes[ddir] - td->stat_io_bytes[ddir];
+		delta = this_io_bytes[ddir] - stat_io_bytes[ddir];
 		if (!delta)
 			continue; /* No entries for interval */
 
-		if (spent)
-			rate = delta * 1000 / spent / 1024; /* KiB/s */
-		else
+		if (spent) {
+			if (is_kb)
+				rate = delta * 1000 / spent / 1024; /* KiB/s */
+			else
+				rate = (delta * 1000) / spent;
+		} else
 			rate = 0;
 
-		add_stat_sample(&ts->bw_stat[ddir], rate);
+		add_stat_sample(&stat[ddir], rate);
 
-		if (td->bw_log) {
-			unsigned int bs = 0;
+		if (log) {
+			unsigned long long bs = 0;
 
 			if (td->o.min_bs[ddir] == td->o.max_bs[ddir])
 				bs = td->o.min_bs[ddir];
 
-			next = add_log_sample(td, td->bw_log, sample_val(rate),
-					      ddir, bs, 0);
+			next = add_log_sample(td, log, sample_val(rate), ddir, bs, 0);
 			next_log = min(next_log, next);
 		}
 
-		td->stat_io_bytes[ddir] = td->this_io_bytes[ddir];
+		stat_io_bytes[ddir] = this_io_bytes[ddir];
 	}
 
-	timeval_add_msec(&td->bw_sample_time, td->o.bw_avg_time);
+	timespec_add_msec(parent_tv, avg_time);
 
-	td_io_u_unlock(td);
+	if (needs_lock)
+		__td_io_u_unlock(td);
 
-	if (spent <= td->o.bw_avg_time)
-		return min(next_log, td->o.bw_avg_time);
+	if (spent <= avg_time)
+		next = avg_time;
+	else
+		next = avg_time - (1 + spent - avg_time);
 
-	next = td->o.bw_avg_time - (1 + spent - td->o.bw_avg_time);
 	return min(next, next_log);
 }
 
+static int add_bw_samples(struct thread_data *td, struct timespec *t)
+{
+	return __add_samples(td, &td->bw_sample_time, t, td->o.bw_avg_time,
+				td->this_io_bytes, td->stat_io_bytes,
+				td->ts.bw_stat, td->bw_log, true);
+}
+
 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);
 
@@ -2499,65 +2701,16 @@ 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 timeval *t)
+static int add_iops_samples(struct thread_data *td, struct timespec *t)
 {
-	struct thread_stat *ts = &td->ts;
-	unsigned long spent, iops;
-	enum fio_ddir ddir;
-	unsigned int next, next_log;
-
-	next_log = td->o.iops_avg_time;
-
-	spent = mtime_since(&td->iops_sample_time, t);
-	if (spent < td->o.iops_avg_time &&
-	    td->o.iops_avg_time - spent >= LOG_MSEC_SLACK)
-		return td->o.iops_avg_time - spent;
-
-	td_io_u_lock(td);
-
-	/*
-	 * Compute both read and write rates for the interval.
-	 */
-	for (ddir = 0; ddir < DDIR_RWDIR_CNT; ddir++) {
-		uint64_t delta;
-
-		delta = td->this_io_blocks[ddir] - td->stat_io_blocks[ddir];
-		if (!delta)
-			continue; /* No entries for interval */
-
-		if (spent)
-			iops = (delta * 1000) / spent;
-		else
-			iops = 0;
-
-		add_stat_sample(&ts->iops_stat[ddir], iops);
-
-		if (td->iops_log) {
-			unsigned int bs = 0;
-
-			if (td->o.min_bs[ddir] == td->o.max_bs[ddir])
-				bs = td->o.min_bs[ddir];
-
-			next = add_log_sample(td, td->iops_log,
-					      sample_val(iops), ddir, bs, 0);
-			next_log = min(next_log, next);
-		}
-
-		td->stat_io_blocks[ddir] = td->this_io_blocks[ddir];
-	}
-
-	timeval_add_msec(&td->iops_sample_time, td->o.iops_avg_time);
-
-	td_io_u_unlock(td);
-
-	if (spent <= td->o.iops_avg_time)
-		return min(next_log, td->o.iops_avg_time);
-
-	next = td->o.iops_avg_time - (1 + spent - td->o.iops_avg_time);
-	return min(next, next_log);
+	return __add_samples(td, &td->iops_sample_time, t, td->o.iops_avg_time,
+				td->this_io_blocks, td->stat_io_blocks,
+				td->ts.iops_stat, td->iops_log, false);
 }
 
 /*
@@ -2567,7 +2720,7 @@ int calc_log_samples(void)
 {
 	struct thread_data *td;
 	unsigned int next = ~0U, tmp;
-	struct timeval now;
+	struct timespec now;
 	int i;
 
 	fio_gettime(&now, NULL);
@@ -2580,12 +2733,14 @@ int calc_log_samples(void)
 			next = min(td->o.iops_avg_time, td->o.bw_avg_time);
 			continue;
 		}
-		if (td->bw_log && !per_unit_log(td->bw_log)) {
+		if (!td->bw_log ||
+			(td->bw_log && !per_unit_log(td->bw_log))) {
 			tmp = add_bw_samples(td, &now);
 			if (tmp < next)
 				next = tmp;
 		}
-		if (td->iops_log && !per_unit_log(td->iops_log)) {
+		if (!td->iops_log ||
+			(td->iops_log && !per_unit_log(td->iops_log))) {
 			tmp = add_iops_samples(td, &now);
 			if (tmp < next)
 				next = tmp;
@@ -2597,7 +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)
@@ -2606,8 +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);
 }
 
 /*