X-Git-Url: https://git.kernel.dk/?p=fio.git;a=blobdiff_plain;f=stat.c;h=369f96967ae0c6e6084a3dace34e97de5304a877;hp=1c1dceb5beccbe3cd6e1950addf1ab93d5b22937;hb=54c05828da413f6b4f9eec800a30624c453a4d90;hpb=876e10015bc3191a71512c3ae57482bf1add4e0e

diff --git a/stat.c b/stat.c
index 1c1dceb5..369f9696 100644
--- a/stat.c
+++ b/stat.c
@@ -15,6 +15,10 @@
 #include "idletime.h"
 #include "lib/pow2.h"
 #include "lib/output_buffer.h"
+#include "helper_thread.h"
+#include "smalloc.h"
+
+#define LOG_MSEC_SLACK	10
 
 struct fio_mutex *stat_mutex;
 
@@ -33,9 +37,9 @@ 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);
@@ -54,7 +58,7 @@ 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;
 
@@ -62,7 +66,7 @@ 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
@@ -94,7 +98,7 @@ 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;
 
@@ -131,16 +135,16 @@ static int double_cmp(const void *a, const void *b)
 }
 
 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)
+				   fio_fp64_t *plist, unsigned long long **output,
+				   unsigned long long *maxv, unsigned long long *minv)
 {
 	unsigned long sum = 0;
 	unsigned int len, i, j = 0;
 	unsigned int oval_len = 0;
-	unsigned int *ovals = NULL;
+	unsigned long long *ovals = NULL;
 	int is_last;
 
-	*minv = -1U;
+	*minv = -1ULL;
 	*maxv = 0;
 
 	len = 0;
@@ -169,7 +173,7 @@ 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);
@@ -197,9 +201,10 @@ static void show_clat_percentiles(unsigned int *io_u_plat, unsigned long nr,
 				  fio_fp64_t *plist, unsigned int precision,
 				  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 is_last, per_line, scale_down, time_width;
 	char fmt[32];
 
 	len = calc_clat_percentiles(io_u_plat, nr, plist, &ovals, &maxv, &minv);
@@ -207,23 +212,31 @@ static void show_clat_percentiles(unsigned int *io_u_plat, unsigned long nr,
 		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) {
-		scale_down = 1;
+	if (minv > 2000000 && maxv > 99999999ULL) {
+		scale_down = 2;
+		divisor = 1000000;
 		log_buf(out, "    clat percentiles (msec):\n     |");
+	} else if (minv > 2000 && maxv > 99999) {
+		scale_down = 1;
+		divisor = 1000;
+		log_buf(out, "    clat percentiles (usec):\n     |");
 	} else {
 		scale_down = 0;
-		log_buf(out, "    clat percentiles (usec):\n     |");
+		divisor = 1;
+		log_buf(out, "    clat percentiles (nsec):\n     |");
 	}
 
-	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, "     |");
@@ -231,15 +244,10 @@ static void show_clat_percentiles(unsigned int *io_u_plat, unsigned long nr,
 		/* 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);
-
-		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;
@@ -253,13 +261,13 @@ out:
 		free(ovals);
 }
 
-int 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;
 
 	if (n == 0)
-		return 0;
+		return false;
 
 	*min = is->min_val;
 	*max = is->max_val;
@@ -270,12 +278,13 @@ int calc_lat(struct io_stat *is, unsigned long *min, unsigned long *max,
 	else
 		*dev = 0;
 
-	return 1;
+	return true;
 }
 
 void show_group_stats(struct group_run_stats *rs, struct buf_output *out)
 {
-	char *p1, *p2, *p3, *p4;
+	char *io, *agg, *min, *max;
+	char *ioalt, *aggalt, *minalt, *maxalt;
 	const char *str[] = { "   READ", "  WRITE" , "   TRIM"};
 	int i;
 
@@ -287,22 +296,28 @@ void show_group_stats(struct group_run_stats *rs, struct buf_output *out)
 		if (!rs->max_run[i])
 			continue;
 
-		p1 = num2str(rs->io_kb[i], 6, rs->kb_base, i2p, 8);
-		p2 = num2str(rs->agg[i], 6, rs->kb_base, i2p, rs->unit_base);
-		p3 = num2str(rs->min_bw[i], 6, rs->kb_base, i2p, rs->unit_base);
-		p4 = num2str(rs->max_bw[i], 6, rs->kb_base, i2p, rs->unit_base);
-
-		log_buf(out, "%s: io=%s, aggrb=%s/s, minb=%s/s, maxb=%s/s,"
-			 " mint=%llumsec, maxt=%llumsec\n",
+		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);
+		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],
-				p1, p2, p3, p4,
+				agg, aggalt, min, max, minalt, maxalt, io, ioalt,
 				(unsigned long long) rs->min_run[i],
 				(unsigned long long) rs->max_run[i]);
 
-		free(p1);
-		free(p2);
-		free(p3);
-		free(p4);
+		free(io);
+		free(agg);
+		free(min);
+		free(max);
+		free(ioalt);
+		free(aggalt);
+		free(minalt);
+		free(maxalt);
 	}
 }
 
@@ -344,6 +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
+ */
+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);
@@ -354,17 +391,20 @@ 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, 0);
-	maxp = num2str(max, 6, 1, 0, 0);
+	minp = num2str(min, 6, 1, 0, N2S_NONE);
+	maxp = num2str(max, 6, 1, 0, N2S_NONE);
 
 	log_buf(out, "    %s %s: min=%s, max=%s, avg=%5.02f,"
 		 " stdev=%5.02f\n", name, base, minp, maxp, mean, dev);
@@ -376,11 +416,11 @@ 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" };
+	unsigned long runt;
+	unsigned long long min, max, bw, iops;
 	double mean, dev;
-	char *io_p, *bw_p, *iops_p;
+	char *io_p, *bw_p, *bw_p_alt, *iops_p;
 	int i2p;
 
 	assert(ddir_rw(ddir));
@@ -392,19 +432,21 @@ 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], 6, 1, i2p, 8);
-	bw_p = num2str(bw, 6, 1, i2p, ts->unit_base);
+	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);
 
 	iops = (1000 * (uint64_t)ts->total_io_u[ddir]) / runt;
-	iops_p = num2str(iops, 6, 1, 0, 0);
+	iops_p = num2str(iops, 4, 1, 0, N2S_NONE);
 
-	log_buf(out, "  %s: io=%s, bw=%s/s, iops=%s, runt=%6llumsec\n",
-				rs->unified_rw_rep ? "mixed" : str[ddir],
-				io_p, bw_p, iops_p,
-				(unsigned long long) ts->runtime[ddir]);
+	log_buf(out, "  %s: IOPS=%s, BW=%s (%s)(%s/%llumsec)\n",
+			rs->unified_rw_rep ? "mixed" : str[ddir],
+			iops_p, bw_p, bw_p_alt, io_p,
+			(unsigned long long) ts->runtime[ddir]);
 
 	free(io_p);
 	free(bw_p);
+	free(bw_p_alt);
 	free(iops_p);
 
 	if (calc_lat(&ts->slat_stat[ddir], &min, &max, &mean, &dev))
@@ -422,7 +464,16 @@ static void show_ddir_status(struct group_run_stats *rs, struct thread_stat *ts,
 	}
 	if (calc_lat(&ts->bw_stat[ddir], &min, &max, &mean, &dev)) {
 		double p_of_agg = 100.0, fkb_base = (double)rs->kb_base;
-		const char *bw_str = (rs->unit_base == 1 ? "Kbit" : "KB");
+		const char *bw_str;
+
+		if ((rs->unit_base == 1) && i2p)
+			bw_str = "Kibit";
+		else if (rs->unit_base == 1)
+			bw_str = "kbit";
+		else if (i2p)
+			bw_str = "KiB";
+		else
+			bw_str = "kB";
 
 		if (rs->unit_base == 1) {
 			min *= 8.0;
@@ -442,12 +493,18 @@ static void show_ddir_status(struct group_run_stats *rs, struct thread_stat *ts,
 			max /= fkb_base;
 			mean /= fkb_base;
 			dev /= fkb_base;
-			bw_str = (rs->unit_base == 1 ? "Mbit" : "MB");
+			bw_str = (rs->unit_base == 1 ? "Mibit" : "MiB");
 		}
 
-		log_buf(out, "    bw (%-4s/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=%5lu\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=%5lu\n",
+			min, max, mean, dev, (&ts->iops_stat[ddir])->samples);
 	}
 }
 
@@ -481,6 +538,14 @@ static int show_lat(double *io_u_lat, int nr, const char **ranges,
 	return shown;
 }
 
+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)
 {
 	const char *ranges[] = { "2=", "4=", "10=", "20=", "50=", "100=",
@@ -500,12 +565,15 @@ 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);
 }
@@ -653,6 +721,35 @@ static void show_block_infos(int nr_block_infos, uint32_t *block_infos,
 			 i == BLOCK_STATE_COUNT - 1 ? '\n' : ',');
 }
 
+static void show_ss_normal(struct thread_stat *ts, struct buf_output *out)
+{
+	char *p1, *p1alt, *p2;
+	unsigned long long bw_mean, iops_mean;
+	const int i2p = is_power_of_2(ts->kb_base);
+
+	if (!ts->ss_dur)
+		return;
+
+	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);
+
+	log_buf(out, "  steadystate  : attained=%s, bw=%s (%s), iops=%s, %s%s=%.3f%s\n",
+		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_criterion.u.f,
+		ts->ss_state & __FIO_SS_PCT ? "%" : "");
+
+	free(p1);
+	free(p1alt);
+	free(p2);
+}
+
 static void show_thread_status_normal(struct thread_stat *ts,
 				      struct group_run_stats *rs,
 				      struct buf_output *out)
@@ -665,6 +762,8 @@ static void show_thread_status_normal(struct thread_stat *ts,
 
 	if (!ddir_rw_sum(ts->io_bytes) && !ddir_rw_sum(ts->total_io_u))
 		return;
+		
+	memset(time_buf, 0, sizeof(time_buf));
 
 	time(&time_p);
 	os_ctime_r((const time_t *) &time_p, time_buf, sizeof(time_buf));
@@ -728,9 +827,9 @@ 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    : total=r=%llu/w=%llu/d=%llu,"
-				 " short=r=%llu/w=%llu/d=%llu,"
-				 " drop=r=%llu/w=%llu/d=%llu\n",
+	log_buf(out, "     issued rwt: total=%llu,%llu,%llu,"
+				 " short=%llu,%llu,%llu,"
+				 " dropped=%llu,%llu,%llu\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],
@@ -757,17 +856,19 @@ static void show_thread_status_normal(struct thread_stat *ts,
 	if (ts->nr_block_infos)
 		show_block_infos(ts->nr_block_infos, ts->block_infos,
 				  ts->percentile_list, out);
+
+	if (ts->ss_dur)
+		show_ss_normal(ts, out);
 }
 
 static void show_ddir_status_terse(struct thread_stat *ts,
 				   struct group_run_stats *rs, int ddir,
 				   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;
+	unsigned int len;
 	int i;
 
 	assert(ddir_rw(ddir));
@@ -776,7 +877,7 @@ static void show_ddir_status_terse(struct thread_stat *ts,
 	if (ts->runtime[ddir]) {
 		uint64_t runt = ts->runtime[ddir];
 
-		bw = ((1000 * ts->io_bytes[ddir]) / runt) / 1024;
+		bw = ((1000 * ts->io_bytes[ddir]) / runt) / 1024; /* KiB/s */
 		iops = (1000 * (uint64_t) ts->total_io_u[ddir]) / runt;
 	}
 
@@ -785,14 +886,14 @@ 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) {
 		len = calc_clat_percentiles(ts->io_u_plat[ddir],
@@ -807,13 +908,13 @@ 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);
@@ -827,19 +928,19 @@ static void show_ddir_status_terse(struct thread_stat *ts,
 				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);
 }
 
 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 min, max, minv, maxv;
 	unsigned long long bw;
-	unsigned int *ovals = NULL;
+	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;
@@ -860,11 +961,12 @@ static void add_ddir_status_json(struct thread_stat *ts,
 	if (ts->runtime[ddir]) {
 		uint64_t runt = ts->runtime[ddir];
 
-		bw = ((1000 * ts->io_bytes[ddir]) / runt) / 1024;
+		bw = ((1000 * ts->io_bytes[ddir]) / runt) / 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, "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", bw);
 	json_object_add_value_float(dir_object, "iops", iops);
 	json_object_add_value_int(dir_object, "runtime", ts->runtime[ddir]);
@@ -877,7 +979,7 @@ static void add_ddir_status_json(struct thread_stat *ts,
 		mean = dev = 0.0;
 	}
 	tmp_object = json_create_object();
-	json_object_add_value_object(dir_object, "slat", tmp_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);
@@ -888,7 +990,7 @@ static void add_ddir_status_json(struct thread_stat *ts,
 		mean = dev = 0.0;
 	}
 	tmp_object = json_create_object();
-	json_object_add_value_object(dir_object, "clat", tmp_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);
@@ -917,12 +1019,11 @@ static void add_ddir_status_json(struct thread_stat *ts,
 		clat_bins_object = json_create_object();
 		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 (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]);
+			}
 		}
-		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 (!calc_lat(&ts->lat_stat[ddir], &min, &max, &mean, &dev)) {
@@ -930,7 +1031,7 @@ static void add_ddir_status_json(struct thread_stat *ts,
 		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);
@@ -953,6 +1054,19 @@ 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);
+
+	if (!calc_lat(&ts->iops_stat[ddir], &min, &max, &mean, &dev)) {
+		min = max = 0;
+		mean = dev = 0.0;
+	}
+	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_v2(struct thread_stat *ts,
@@ -992,7 +1106,7 @@ static void show_thread_status_terse_v2(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*/
@@ -1057,7 +1171,7 @@ 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*/
@@ -1118,6 +1232,7 @@ 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;
@@ -1162,6 +1277,7 @@ static struct json_object *show_thread_status_json(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_n(ts, io_u_lat_n);
 	stat_calc_lat_u(ts, io_u_lat_u);
 	stat_calc_lat_m(ts, io_u_lat_m);
 
@@ -1177,9 +1293,17 @@ 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]);
 	}
 
+	/* Nanosecond latency */
 	tmp = json_create_object();
-	json_object_add_value_object(root, "latency_us", tmp);
+	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", };
@@ -1251,6 +1375,56 @@ 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;
+		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" : "",
+			(float) ts->ss_limit.u.f,
+			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);
+
+		snprintf(ss_buf, sizeof(ss_buf), "%f%s", (float) ts->ss_criterion.u.f,
+			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);
+
+		data = json_create_object();
+		json_object_add_value_object(tmp, "data", data);
+		bw = json_create_array();
+		iops = json_create_array();
+
+		/*
+		** if ss was attained or the buffer is not full,
+		** ss->head points to the first element in the list.
+		** 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))
+			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;
+			json_array_add_value_int(bw, ts->ss_bw_data[k]);
+			json_array_add_value_int(iops, ts->ss_iops_data[k]);
+		}
+		json_object_add_value_int(data, "bw_mean", steadystate_bw_mean(ts));
+		json_object_add_value_int(data, "iops_mean", steadystate_iops_mean(ts));
+		json_object_add_value_array(data, "iops", iops);
+		json_object_add_value_array(data, "bw", bw);
+	}
+
 	return root;
 }
 
@@ -1268,6 +1442,7 @@ static void show_thread_status_terse(struct thread_stat *ts,
 
 struct json_object *show_thread_status(struct thread_stat *ts,
 				       struct group_run_stats *rs,
+				       struct flist_head *opt_list,
 				       struct buf_output *out)
 {
 	struct json_object *ret = NULL;
@@ -1275,7 +1450,7 @@ struct json_object *show_thread_status(struct thread_stat *ts,
 	if (output_format & FIO_OUTPUT_TERSE)
 		show_thread_status_terse(ts, rs,  out);
 	if (output_format & FIO_OUTPUT_JSON)
-		ret = show_thread_status_json(ts, rs, NULL);
+		ret = show_thread_status_json(ts, rs, opt_list);
 	if (output_format & FIO_OUTPUT_NORMAL)
 		show_thread_status_normal(ts, rs,  out);
 
@@ -1331,7 +1506,7 @@ void sum_group_stats(struct group_run_stats *dst, struct group_run_stats *src)
 		if (dst->min_bw[i] && dst->min_bw[i] > src->min_bw[i])
 			dst->min_bw[i] = src->min_bw[i];
 
-		dst->io_kb[i] += src->io_kb[i];
+		dst->iobytes[i] += src->iobytes[i];
 		dst->agg[i] += src->agg[i];
 	}
 
@@ -1352,6 +1527,7 @@ 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];
 
@@ -1362,6 +1538,7 @@ 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];
 
@@ -1388,6 +1565,8 @@ void sum_thread_stats(struct thread_stat *dst, struct thread_stat *src,
 		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_N_NR; k++)
+		dst->io_u_lat_n[k] += src->io_u_lat_n[k];
 	for (k = 0; k < FIO_IO_U_LAT_U_NR; k++)
 		dst->io_u_lat_u[k] += src->io_u_lat_u[k];
 	for (k = 0; k < FIO_IO_U_LAT_M_NR; k++)
@@ -1441,6 +1620,7 @@ 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->groupid = -1;
 }
@@ -1476,6 +1656,8 @@ void __show_run_stats(void)
 		}
 		if (last_ts == td->groupid)
 			continue;
+		if (!td->o.stats)
+			continue;
 
 		last_ts = td->groupid;
 		nr_ts++;
@@ -1493,6 +1675,8 @@ void __show_run_stats(void)
 	last_ts = -1;
 	idx = 0;
 	for_each_td(td, i) {
+		if (!td->o.stats)
+			continue;
 		if (idx && (!td->o.group_reporting ||
 		    (td->o.group_reporting && last_ts != td->groupid))) {
 			idx = 0;
@@ -1573,12 +1757,28 @@ void __show_run_stats(void)
 			ts->block_infos[k] = td->ts.block_infos[k];
 
 		sum_thread_stats(ts, &td->ts, idx == 1);
+
+		if (td->o.ss_dur) {
+			ts->ss_state = td->ss.state;
+			ts->ss_dur = td->ss.dur;
+			ts->ss_head = td->ss.head;
+			ts->ss_bw_data = td->ss.bw_data;
+			ts->ss_iops_data = td->ss.iops_data;
+			ts->ss_limit.u.f = td->ss.limit;
+			ts->ss_slope.u.f = td->ss.slope;
+			ts->ss_deviation.u.f = td->ss.deviation;
+			ts->ss_criterion.u.f = td->ss.criterion;
+		}
+		else
+			ts->ss_dur = ts->ss_state = 0;
 	}
 
 	for (i = 0; i < nr_ts; i++) {
 		unsigned long long bw;
 
 		ts = &threadstats[i];
+		if (ts->groupid == -1)
+			continue;
 		rs = &runstats[ts->groupid];
 		rs->kb_base = ts->kb_base;
 		rs->unit_base = ts->unit_base;
@@ -1593,19 +1793,14 @@ void __show_run_stats(void)
 				rs->max_run[j] = ts->runtime[j];
 
 			bw = 0;
-			if (ts->runtime[j]) {
-				unsigned long runt = ts->runtime[j];
-				unsigned long long kb;
-
-				kb = ts->io_bytes[j] / rs->kb_base;
-				bw = kb * 1000 / runt;
-			}
+			if (ts->runtime[j])
+				bw = ts->io_bytes[j] * 1000 / ts->runtime[j];
 			if (bw < rs->min_bw[j])
 				rs->min_bw[j] = bw;
 			if (bw > rs->max_bw[j])
 				rs->max_bw[j] = bw;
 
-			rs->io_kb[j] += ts->io_bytes[j] / rs->kb_base;
+			rs->iobytes[j] += ts->io_bytes[j];
 		}
 	}
 
@@ -1616,7 +1811,7 @@ void __show_run_stats(void)
 
 		for (ddir = 0; ddir < DDIR_RWDIR_CNT; ddir++) {
 			if (rs->max_run[ddir])
-				rs->agg[ddir] = (rs->io_kb[ddir] * 1000) /
+				rs->agg[ddir] = (rs->iobytes[ddir] * 1000) /
 						rs->max_run[ddir];
 		}
 	}
@@ -1632,16 +1827,22 @@ void __show_run_stats(void)
 	if (output_format & FIO_OUTPUT_JSON) {
 		struct thread_data *global;
 		char time_buf[32];
-		time_t time_p;
+		struct timeval now;
+		unsigned long long ms_since_epoch;
 
-		time(&time_p);
-		os_ctime_r((const time_t *) &time_p, time_buf,
+		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));
-		time_buf[strlen(time_buf) - 1] = '\0';
+		if (time_buf[strlen(time_buf) - 1] == '\n')
+			time_buf[strlen(time_buf) - 1] = '\0';
 
 		root = json_create_object();
 		json_object_add_value_string(root, "fio version", fio_version_string);
-		json_object_add_value_int(root, "timestamp", time_p);
+		json_object_add_value_int(root, "timestamp", now.tv_sec);
+		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);
@@ -1649,13 +1850,17 @@ void __show_run_stats(void)
 		json_object_add_value_array(root, "jobs", array);
 	}
 
+	if (is_backend)
+		fio_server_send_job_options(&get_global_options()->opt_list, -1U);
+
 	for (i = 0; i < nr_ts; i++) {
 		ts = &threadstats[i];
 		rs = &runstats[ts->groupid];
 
-		if (is_backend)
+		if (is_backend) {
+			fio_server_send_job_options(opt_lists[i], i);
 			fio_server_send_ts(ts, rs);
-		else {
+		} else {
 			if (output_format & FIO_OUTPUT_TERSE)
 				show_thread_status_terse(ts, rs, &output[__FIO_OUTPUT_TERSE]);
 			if (output_format & FIO_OUTPUT_JSON) {
@@ -1695,8 +1900,10 @@ 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);
 	}
 
 	log_info_flush();
@@ -1716,28 +1923,28 @@ 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);
 
 	rt = malloc(thread_number * sizeof(unsigned long long));
-	fio_gettime(&tv, NULL);
+	fio_gettime(&ts, NULL);
 
 	for_each_td(td, i) {
-		rt[i] = mtime_since(&td->start, &tv);
-		if (td_read(td) && td->io_bytes[DDIR_READ])
-			td->ts.runtime[DDIR_READ] += rt[i];
-		if (td_write(td) && td->io_bytes[DDIR_WRITE])
-			td->ts.runtime[DDIR_WRITE] += rt[i];
-		if (td_trim(td) && td->io_bytes[DDIR_TRIM])
-			td->ts.runtime[DDIR_TRIM] += rt[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, &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])
+			td->ts.runtime[DDIR_WRITE] += rt[i];
+		if (td_trim(td) && td->ts.io_bytes[DDIR_TRIM])
+			td->ts.runtime[DDIR_TRIM] += rt[i];
 	}
 
 	for_each_td(td, i) {
@@ -1753,11 +1960,11 @@ void __show_running_run_stats(void)
 	__show_run_stats();
 
 	for_each_td(td, i) {
-		if (td_read(td) && td->io_bytes[DDIR_READ])
+		if (td_read(td) && td->ts.io_bytes[DDIR_READ])
 			td->ts.runtime[DDIR_READ] -= rt[i];
-		if (td_write(td) && td->io_bytes[DDIR_WRITE])
+		if (td_write(td) && td->ts.io_bytes[DDIR_WRITE])
 			td->ts.runtime[DDIR_WRITE] -= rt[i];
-		if (td_trim(td) && td->io_bytes[DDIR_TRIM])
+		if (td_trim(td) && td->ts.io_bytes[DDIR_TRIM])
 			td->ts.runtime[DDIR_TRIM] -= rt[i];
 	}
 
@@ -1766,7 +1973,7 @@ void __show_running_run_stats(void)
 }
 
 static int status_interval_init;
-static struct timeval status_time;
+static struct timespec status_time;
 static int status_file_disabled;
 
 #define FIO_STATUS_FILE		"fio-dump-status"
@@ -1822,7 +2029,7 @@ 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;
@@ -1841,58 +2048,184 @@ static inline void add_stat_sample(struct io_stat *is, unsigned long data)
 	is->samples++;
 }
 
-static void __add_log_sample(struct io_log *iolog, unsigned long val,
+/*
+ * Return a struct io_logs, which is added to the tail of the log
+ * list for 'iolog'.
+ */
+static struct io_logs *get_new_log(struct io_log *iolog)
+{
+	size_t new_size, new_samples;
+	struct io_logs *cur_log;
+
+	/*
+	 * Cap the size at MAX_LOG_ENTRIES, so we don't keep doubling
+	 * forever
+	 */
+	if (!iolog->cur_log_max)
+		new_samples = DEF_LOG_ENTRIES;
+	else {
+		new_samples = iolog->cur_log_max * 2;
+		if (new_samples > MAX_LOG_ENTRIES)
+			new_samples = MAX_LOG_ENTRIES;
+	}
+
+	new_size = new_samples * log_entry_sz(iolog);
+
+	cur_log = smalloc(sizeof(*cur_log));
+	if (cur_log) {
+		INIT_FLIST_HEAD(&cur_log->list);
+		cur_log->log = malloc(new_size);
+		if (cur_log->log) {
+			cur_log->nr_samples = 0;
+			cur_log->max_samples = new_samples;
+			flist_add_tail(&cur_log->list, &iolog->io_logs);
+			iolog->cur_log_max = new_samples;
+			return cur_log;
+		}
+		sfree(cur_log);
+	}
+
+	return NULL;
+}
+
+/*
+ * Add and return a new log chunk, or return current log if big enough
+ */
+static struct io_logs *regrow_log(struct io_log *iolog)
+{
+	struct io_logs *cur_log;
+	int i;
+
+	if (!iolog || iolog->disabled)
+		goto disable;
+
+	cur_log = iolog_cur_log(iolog);
+	if (!cur_log) {
+		cur_log = get_new_log(iolog);
+		if (!cur_log)
+			return NULL;
+	}
+
+	if (cur_log->nr_samples < cur_log->max_samples)
+		return cur_log;
+
+	/*
+	 * No room for a new sample. If we're compressing on the fly, flush
+	 * out the current chunk
+	 */
+	if (iolog->log_gz) {
+		if (iolog_cur_flush(iolog, cur_log)) {
+			log_err("fio: failed flushing iolog! Will stop logging.\n");
+			return NULL;
+		}
+	}
+
+	/*
+	 * Get a new log array, and add to our list
+	 */
+	cur_log = get_new_log(iolog);
+	if (!cur_log) {
+		log_err("fio: failed extending iolog! Will stop logging.\n");
+		return NULL;
+	}
+
+	if (!iolog->pending || !iolog->pending->nr_samples)
+		return cur_log;
+
+	/*
+	 * Flush pending items to new log
+	 */
+	for (i = 0; i < iolog->pending->nr_samples; i++) {
+		struct io_sample *src, *dst;
+
+		src = get_sample(iolog, iolog->pending, i);
+		dst = get_sample(iolog, cur_log, i);
+		memcpy(dst, src, log_entry_sz(iolog));
+	}
+	cur_log->nr_samples = iolog->pending->nr_samples;
+
+	iolog->pending->nr_samples = 0;
+	return cur_log;
+disable:
+	if (iolog)
+		iolog->disabled = true;
+	return NULL;
+}
+
+void regrow_logs(struct thread_data *td)
+{
+	regrow_log(td->slat_log);
+	regrow_log(td->clat_log);
+	regrow_log(td->clat_hist_log);
+	regrow_log(td->lat_log);
+	regrow_log(td->bw_log);
+	regrow_log(td->iops_log);
+	td->flags &= ~TD_F_REGROW_LOGS;
+}
+
+static struct io_logs *get_cur_log(struct io_log *iolog)
+{
+	struct io_logs *cur_log;
+
+	cur_log = iolog_cur_log(iolog);
+	if (!cur_log) {
+		cur_log = get_new_log(iolog);
+		if (!cur_log)
+			return NULL;
+	}
+
+	if (cur_log->nr_samples < cur_log->max_samples)
+		return cur_log;
+
+	/*
+	 * Out of space. If we're in IO offload mode, or we're not doing
+	 * per unit logging (hence logging happens outside of the IO thread
+	 * as well), add a new log chunk inline. If we're doing inline
+	 * 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 ||
+	    !per_unit_log(iolog))
+		return regrow_log(iolog);
+
+	iolog->td->flags |= TD_F_REGROW_LOGS;
+	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,
 			     unsigned long t, uint64_t offset)
 {
-	uint64_t nr_samples = iolog->nr_samples;
-	struct io_sample *s;
+	struct io_logs *cur_log;
 
 	if (iolog->disabled)
 		return;
-
-	if (!iolog->nr_samples)
+	if (flist_empty(&iolog->io_logs))
 		iolog->avg_last = t;
 
-	if (iolog->nr_samples == iolog->max_samples) {
-		size_t new_size;
-		void *new_log;
+	cur_log = get_cur_log(iolog);
+	if (cur_log) {
+		struct io_sample *s;
 
-		new_size = 2 * iolog->max_samples * log_entry_sz(iolog);
-
-		if (iolog->log_gz && (new_size > iolog->log_gz)) {
-			if (iolog_flush(iolog, 0)) {
-				log_err("fio: failed flushing iolog! Will stop logging.\n");
-				iolog->disabled = 1;
-				return;
-			}
-			nr_samples = iolog->nr_samples;
-		} else {
-			new_log = realloc(iolog->log, new_size);
-			if (!new_log) {
-				log_err("fio: failed extending iolog! Will stop logging.\n");
-				iolog->disabled = 1;
-				return;
-			}
-			iolog->log = new_log;
-			iolog->max_samples <<= 1;
-		}
-	}
+		s = get_sample(iolog, cur_log, cur_log->nr_samples);
 
-	s = get_sample(iolog, nr_samples);
+		s->data = data;
+		s->time = t + (iolog->td ? iolog->td->unix_epoch : 0);
+		io_sample_set_ddir(iolog, s, ddir);
+		s->bs = bs;
 
-	s->val = val;
-	s->time = t;
-	io_sample_set_ddir(iolog, s, ddir);
-	s->bs = bs;
+		if (iolog->log_offset) {
+			struct io_sample_offset *so = (void *) s;
 
-	if (iolog->log_offset) {
-		struct io_sample_offset *so = (void *) s;
+			so->offset = offset;
+		}
 
-		so->offset = offset;
+		cur_log->nr_samples++;
+		return;
 	}
 
-	iolog->nr_samples++;
+	iolog->disabled = true;
 }
 
 static inline void reset_io_stat(struct io_stat *ios)
@@ -1915,6 +2248,9 @@ 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++)
 			ts->io_u_plat[i][j] = 0;
@@ -1924,58 +2260,58 @@ void reset_io_stats(struct thread_data *td)
 		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;
 }
 
-static void _add_stat_to_log(struct io_log *iolog, unsigned long elapsed)
+static void __add_stat_to_log(struct io_log *iolog, enum fio_ddir ddir,
+			      unsigned long elapsed, bool log_max)
 {
 	/*
 	 * Note an entry in the log. Use the mean from the logged samples,
 	 * making sure to properly round up. Only write a log entry if we
 	 * had actual samples done.
 	 */
-	if (iolog->avg_window[DDIR_READ].samples) {
-		unsigned long mr;
+	if (iolog->avg_window[ddir].samples) {
+		union io_sample_data data;
 
-		mr = iolog->avg_window[DDIR_READ].mean.u.f + 0.50;
-		__add_log_sample(iolog, mr, DDIR_READ, 0, elapsed, 0);
-	}
-	if (iolog->avg_window[DDIR_WRITE].samples) {
-		unsigned long mw;
+		if (log_max)
+			data.val = iolog->avg_window[ddir].max_val;
+		else
+			data.val = iolog->avg_window[ddir].mean.u.f + 0.50;
 
-		mw = iolog->avg_window[DDIR_WRITE].mean.u.f + 0.50;
-		__add_log_sample(iolog, mw, DDIR_WRITE, 0, elapsed, 0);
+		__add_log_sample(iolog, data, ddir, 0, elapsed, 0);
 	}
-	if (iolog->avg_window[DDIR_TRIM].samples) {
-		unsigned long mw;
 
-		mw = iolog->avg_window[DDIR_TRIM].mean.u.f + 0.50;
-		__add_log_sample(iolog, mw, DDIR_TRIM, 0, elapsed, 0);
-	}
+	reset_io_stat(&iolog->avg_window[ddir]);
+}
+
+static void _add_stat_to_log(struct io_log *iolog, unsigned long elapsed,
+			     bool log_max)
+{
+	int ddir;
 
-	reset_io_stat(&iolog->avg_window[DDIR_READ]);
-	reset_io_stat(&iolog->avg_window[DDIR_WRITE]);
-	reset_io_stat(&iolog->avg_window[DDIR_TRIM]);
+	for (ddir = 0; ddir < DDIR_RWDIR_CNT; ddir++)
+		__add_stat_to_log(iolog, ddir, elapsed, log_max);
 }
 
-static void add_log_sample(struct thread_data *td, struct io_log *iolog,
-			   unsigned long val, enum fio_ddir ddir,
+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)
 {
 	unsigned long elapsed, this_window;
 
 	if (!ddir_rw(ddir))
-		return;
+		return 0;
 
 	elapsed = mtime_since_now(&td->epoch);
 
@@ -1983,48 +2319,55 @@ static void add_log_sample(struct thread_data *td, struct io_log *iolog,
 	 * If no time averaging, just add the log sample.
 	 */
 	if (!iolog->avg_msec) {
-		__add_log_sample(iolog, val, ddir, bs, elapsed, offset);
-		return;
+		__add_log_sample(iolog, data, ddir, bs, elapsed, offset);
+		return 0;
 	}
 
 	/*
 	 * Add the sample. If the time period has passed, then
 	 * add that entry to the log and clear.
 	 */
-	add_stat_sample(&iolog->avg_window[ddir], val);
+	add_stat_sample(&iolog->avg_window[ddir], data.val);
 
 	/*
 	 * If period hasn't passed, adding the above sample is all we
 	 * need to do.
 	 */
 	this_window = elapsed - iolog->avg_last;
-	if (this_window < iolog->avg_msec)
-		return;
+	if (elapsed < iolog->avg_last)
+		return iolog->avg_last - elapsed;
+	else if (this_window < iolog->avg_msec) {
+		int diff = iolog->avg_msec - this_window;
+
+		if (inline_log(iolog) || diff > LOG_MSEC_SLACK)
+			return diff;
+	}
 
-	_add_stat_to_log(iolog, elapsed);
+	_add_stat_to_log(iolog, elapsed, td->o.log_max != 0);
 
-	iolog->avg_last = elapsed;
+	iolog->avg_last = elapsed - (this_window - iolog->avg_msec);
+	return iolog->avg_msec;
 }
 
-void finalize_logs(struct thread_data *td)
+void finalize_logs(struct thread_data *td, bool unit_logs)
 {
 	unsigned long elapsed;
 
 	elapsed = mtime_since_now(&td->epoch);
 
-	if (td->clat_log)
-		_add_stat_to_log(td->clat_log, elapsed);
-	if (td->slat_log)
-		_add_stat_to_log(td->slat_log, elapsed);
-	if (td->lat_log)
-		_add_stat_to_log(td->lat_log, elapsed);
-	if (td->bw_log)
-		_add_stat_to_log(td->bw_log, elapsed);
-	if (td->iops_log)
-		_add_stat_to_log(td->iops_log, elapsed);
+	if (td->clat_log && unit_logs)
+		_add_stat_to_log(td->clat_log, elapsed, td->o.log_max != 0);
+	if (td->slat_log && unit_logs)
+		_add_stat_to_log(td->slat_log, elapsed, td->o.log_max != 0);
+	if (td->lat_log && unit_logs)
+		_add_stat_to_log(td->lat_log, elapsed, td->o.log_max != 0);
+	if (td->bw_log && (unit_logs == per_unit_log(td->bw_log)))
+		_add_stat_to_log(td->bw_log, elapsed, td->o.log_max != 0);
+	if (td->iops_log && (unit_logs == per_unit_log(td->iops_log)))
+		_add_stat_to_log(td->iops_log, elapsed, td->o.log_max != 0);
 }
 
-void add_agg_sample(unsigned long val, enum fio_ddir ddir, unsigned int bs)
+void add_agg_sample(union io_sample_data data, enum fio_ddir ddir, unsigned int bs)
 {
 	struct io_log *iolog;
 
@@ -2032,35 +2375,73 @@ void add_agg_sample(unsigned long val, enum fio_ddir ddir, unsigned int bs)
 		return;
 
 	iolog = agg_io_log[ddir];
-	__add_log_sample(iolog, val, ddir, bs, mtime_since_genesis(), 0);
+	__add_log_sample(iolog, data, ddir, bs, mtime_since_genesis(), 0);
 }
 
 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 int bs, uint64_t offset)
 {
+	unsigned long elapsed, this_window;
 	struct thread_stat *ts = &td->ts;
-
-	if (!ddir_rw(ddir))
-		return;
+	struct io_log *iolog = td->clat_hist_log;
 
 	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, usec, ddir, bs, offset);
+		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];
+
+		hw->samples++;
+		elapsed = mtime_since_now(&td->epoch);
+		if (!hw->hist_last)
+			hw->hist_last = elapsed;
+		this_window = elapsed - hw->hist_last;
+		
+		if (this_window >= iolog->hist_msec) {
+			unsigned int *io_u_plat;
+			struct io_u_plat_entry *dst;
+
+			/*
+			 * Make a byte-for-byte copy of the latency histogram
+			 * stored in td->ts.io_u_plat[ddir], recording it in a
+			 * log sample. Note that the matching call to free() is
+			 * located in iolog.c after printing this sample to the
+			 * log file.
+			 */
+			io_u_plat = (unsigned int *) 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));
+			flist_add(&dst->list, &hw->list);
+			__add_log_sample(iolog, sample_plat(dst), ddir, bs,
+						elapsed, offset);
+
+			/*
+			 * Update the last time we recorded as being now, minus
+			 * any drift in time we encountered before actually
+			 * making the record.
+			 */
+			hw->hist_last = elapsed - (this_window - iolog->hist_msec);
+			hw->samples = 0;
+		}
+	}
 
 	td_io_u_unlock(td);
 }
@@ -2078,13 +2459,13 @@ void add_slat_sample(struct thread_data *td, enum fio_ddir ddir,
 	add_stat_sample(&ts->slat_stat[ddir], usec);
 
 	if (td->slat_log)
-		add_log_sample(td, td->slat_log, usec, ddir, bs, offset);
+		add_log_sample(td, td->slat_log, sample_val(usec), ddir, bs, offset);
 
 	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 int bs, uint64_t offset)
 {
 	struct thread_stat *ts = &td->ts;
 
@@ -2093,96 +2474,167 @@ void add_lat_sample(struct thread_data *td, enum fio_ddir ddir,
 
 	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, usec, ddir, bs, offset);
+		add_log_sample(td, td->lat_log, sample_val(nsec), ddir, bs,
+			       offset);
 
 	td_io_u_unlock(td);
 }
 
-void add_bw_sample(struct thread_data *td, enum fio_ddir ddir, unsigned int bs,
-		   struct timeval *t)
+void add_bw_sample(struct thread_data *td, struct io_u *io_u,
+		   unsigned int bytes, unsigned long long spent)
 {
 	struct thread_stat *ts = &td->ts;
+	unsigned long rate;
+
+	if (spent)
+		rate = (unsigned long) (bytes * 1000000ULL / spent);
+	else
+		rate = 0;
+
+	td_io_u_lock(td);
+
+	add_stat_sample(&ts->bw_stat[io_u->ddir], rate);
+
+	if (td->bw_log)
+		add_log_sample(td, td->bw_log, sample_val(rate), io_u->ddir,
+			       bytes, io_u->offset);
+
+	td->stat_io_bytes[io_u->ddir] = td->this_io_bytes[io_u->ddir];
+	td_io_u_unlock(td);
+}
+
+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)
+{
 	unsigned long spent, rate;
+	enum fio_ddir ddir;
+	unsigned int next, next_log;
 
-	if (!ddir_rw(ddir))
-		return;
+	next_log = avg_time;
 
-	spent = mtime_since(&td->bw_sample_time, t);
-	if (spent < td->o.bw_avg_time)
-		return;
+	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);
 
 	/*
 	 * Compute both read and write rates for the interval.
 	 */
-	for (ddir = DDIR_READ; ddir < DDIR_RWDIR_CNT; ddir++) {
+	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;
-		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 (log) {
+			unsigned int bs = 0;
 
-		if (td->bw_log)
-			add_log_sample(td, td->bw_log, rate, ddir, bs, 0);
+			if (td->o.min_bs[ddir] == td->o.max_bs[ddir])
+				bs = td->o.min_bs[ddir];
+
+			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];
 	}
 
-	fio_gettime(&td->bw_sample_time, NULL);
+	timespec_add_msec(parent_tv, avg_time);
+
 	td_io_u_unlock(td);
+
+	if (spent <= avg_time)
+		next = avg_time;
+	else
+		next = avg_time - (1 + spent - avg_time);
+
+	return min(next, next_log);
 }
 
-void add_iops_sample(struct thread_data *td, enum fio_ddir ddir, unsigned int bs,
-		     struct timeval *t)
+static int add_bw_samples(struct thread_data *td, struct timespec *t)
 {
-	struct thread_stat *ts = &td->ts;
-	unsigned long spent, iops;
-
-	if (!ddir_rw(ddir))
-		return;
+	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);
+}
 
-	spent = mtime_since(&td->iops_sample_time, t);
-	if (spent < td->o.iops_avg_time)
-		return;
+void add_iops_sample(struct thread_data *td, struct io_u *io_u,
+		     unsigned int bytes)
+{
+	struct thread_stat *ts = &td->ts;
 
 	td_io_u_lock(td);
 
-	/*
-	 * Compute both read and write rates for the interval.
-	 */
-	for (ddir = DDIR_READ; ddir < DDIR_RWDIR_CNT; ddir++) {
-		uint64_t delta;
+	add_stat_sample(&ts->iops_stat[io_u->ddir], 1);
 
-		delta = td->this_io_blocks[ddir] - td->stat_io_blocks[ddir];
-		if (!delta)
-			continue; /* No entries for interval */
+	if (td->iops_log)
+		add_log_sample(td, td->iops_log, sample_val(1), io_u->ddir,
+			       bytes, io_u->offset);
 
-		if (spent)
-			iops = (delta * 1000) / spent;
-		else
-			iops = 0;
+	td->stat_io_blocks[io_u->ddir] = td->this_io_blocks[io_u->ddir];
+	td_io_u_unlock(td);
+}
 
-		add_stat_sample(&ts->iops_stat[ddir], iops);
+static int add_iops_samples(struct thread_data *td, struct timespec *t)
+{
+	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);
+}
+
+/*
+ * Returns msecs to next event
+ */
+int calc_log_samples(void)
+{
+	struct thread_data *td;
+	unsigned int next = ~0U, tmp;
+	struct timespec now;
+	int i;
 
-		if (td->iops_log)
-			add_log_sample(td, td->iops_log, iops, ddir, bs, 0);
+	fio_gettime(&now, NULL);
 
-		td->stat_io_blocks[ddir] = td->this_io_blocks[ddir];
+	for_each_td(td, i) {
+		if (!td->o.stats)
+			continue;
+		if (in_ramp_time(td) ||
+		    !(td->runstate == TD_RUNNING || td->runstate == TD_VERIFYING)) {
+			next = min(td->o.iops_avg_time, td->o.bw_avg_time);
+			continue;
+		}
+		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 ||
+			(td->iops_log && !per_unit_log(td->iops_log))) {
+			tmp = add_iops_samples(td, &now);
+			if (tmp < next)
+				next = tmp;
+		}
 	}
 
-	fio_gettime(&td->iops_sample_time, NULL);
-	td_io_u_unlock(td);
+	return next == ~0U ? 0 : next;
 }
 
 void stat_init(void)
@@ -2205,8 +2657,7 @@ void stat_exit(void)
  */
 void show_running_run_stats(void)
 {
-	helper_do_stat = 1;
-	pthread_cond_signal(&helper_cond);
+	helper_do_stat();
 }
 
 uint32_t *io_u_block_info(struct thread_data *td, struct io_u *io_u)