X-Git-Url: https://git.kernel.dk/?p=fio.git;a=blobdiff_plain;f=eta.c;h=970a67dfd0ac8d6672758b99d528f32526f26e4e;hp=99e59ec51d78d4bac7ff45180f16f3ce9407a7d2;hb=4806b82473fea74c517e5e0c7665b0ca0542b3ec;hpb=02bcaa8c31feb93c61b701d143a7eea3efd2124d

diff --git a/eta.c b/eta.c
index 99e59ec5..970a67df 100644
--- a/eta.c
+++ b/eta.c
@@ -2,77 +2,135 @@
  * Status and ETA code
  */
 #include <unistd.h>
-#include <fcntl.h>
 #include <string.h>
+#ifdef CONFIG_VALGRIND_DEV
+#include <valgrind/drd.h>
+#else
+#define DRD_IGNORE_VAR(x) do { } while (0)
+#endif
 
 #include "fio.h"
-#include "os.h"
+#include "lib/pow2.h"
 
-static char run_str[MAX_JOBS + 1];
+static char __run_str[REAL_MAX_JOBS + 1];
+static char run_str[__THREAD_RUNSTR_SZ(REAL_MAX_JOBS) + 1];
+
+static void update_condensed_str(char *rstr, char *run_str_condensed)
+{
+	if (*rstr) {
+		while (*rstr) {
+			int nr = 1;
+
+			*run_str_condensed++ = *rstr++;
+			while (*(rstr - 1) == *rstr) {
+				rstr++;
+				nr++;
+			}
+			run_str_condensed += sprintf(run_str_condensed, "(%u),", nr);
+		}
+		run_str_condensed--;
+	}
+	*run_str_condensed = '\0';
+}
 
 /*
  * Sets the status of the 'td' in the printed status map.
  */
 static void check_str_update(struct thread_data *td)
 {
-	char c = run_str[td->thread_number - 1];
+	char c = __run_str[td->thread_number - 1];
 
 	switch (td->runstate) {
-		case TD_REAPED:
+	case TD_REAPED:
+		if (td->error)
+			c = 'X';
+		else if (td->sig)
+			c = 'K';
+		else
 			c = '_';
-			break;
-		case TD_EXITED:
-			c = 'E';
-			break;
-		case TD_RUNNING:
-			if (td_rw(td)) {
-				if (td->sequential)
-					c = 'M';
+		break;
+	case TD_EXITED:
+		c = 'E';
+		break;
+	case TD_RAMP:
+		c = '/';
+		break;
+	case TD_RUNNING:
+		if (td_rw(td)) {
+			if (td_random(td)) {
+				if (td->o.rwmix[DDIR_READ] == 100)
+					c = 'r';
+				else if (td->o.rwmix[DDIR_WRITE] == 100)
+					c = 'w';
 				else
 					c = 'm';
-			} else if (td_read(td)) {
-				if (td->sequential)
-					c = 'R';
-				else
-					c = 'r';
 			} else {
-				if (td->sequential)
+				if (td->o.rwmix[DDIR_READ] == 100)
+					c = 'R';
+				else if (td->o.rwmix[DDIR_WRITE] == 100)
 					c = 'W';
 				else
-					c = 'w';
+					c = 'M';
 			}
-			break;
-		case TD_VERIFYING:
-			c = 'V';
-			break;
-		case TD_FSYNCING:
-			c = 'F';
-			break;
-		case TD_CREATED:
-			c = 'C';
-			break;
-		case TD_INITIALIZED:
-			c = 'I';
-			break;
-		case TD_NOT_CREATED:
-			c = 'P';
-			break;
-		default:
-			log_err("state %d\n", td->runstate);
+		} else if (td_read(td)) {
+			if (td_random(td))
+				c = 'r';
+			else
+				c = 'R';
+		} else if (td_write(td)) {
+			if (td_random(td))
+				c = 'w';
+			else
+				c = 'W';
+		} else {
+			if (td_random(td))
+				c = 'd';
+			else
+				c = 'D';
+		}
+		break;
+	case TD_PRE_READING:
+		c = 'p';
+		break;
+	case TD_VERIFYING:
+		c = 'V';
+		break;
+	case TD_FSYNCING:
+		c = 'F';
+		break;
+	case TD_FINISHING:
+		c = 'f';
+		break;
+	case TD_CREATED:
+		c = 'C';
+		break;
+	case TD_INITIALIZED:
+	case TD_SETTING_UP:
+		c = 'I';
+		break;
+	case TD_NOT_CREATED:
+		c = 'P';
+		break;
+	default:
+		log_err("state %d\n", td->runstate);
 	}
 
-	run_str[td->thread_number - 1] = c;
+	__run_str[td->thread_number - 1] = c;
+	update_condensed_str(__run_str, run_str);
 }
 
 /*
  * Convert seconds to a printable string.
  */
-static void eta_to_str(char *str, int eta_sec)
+void eta_to_str(char *str, unsigned long eta_sec)
 {
 	unsigned int d, h, m, s;
-	static int always_d, always_h;
+	int disp_hour = 0;
 
-	d = h = m = s = 0;
+	if (eta_sec == -1) {
+		sprintf(str, "--");
+		return;
+	}
 
 	s = eta_sec % 60;
 	eta_sec /= 60;
@@ -82,68 +140,147 @@ static void eta_to_str(char *str, int eta_sec)
 	eta_sec /= 24;
 	d = eta_sec;
 
-	if (d || always_d) {
-		always_d = 1;
-		str += sprintf(str, "%02dd:", d);
-	}
-	if (h || always_h) {
-		always_h = 1;
-		str += sprintf(str, "%02dh:", h);
+	if (d) {
+		disp_hour = 1;
+		str += sprintf(str, "%02ud:", d);
 	}
 
-	str += sprintf(str, "%02dm:", m);
-	str += sprintf(str, "%02ds", s);
+	if (h || disp_hour)
+		str += sprintf(str, "%02uh:", h);
+
+	str += sprintf(str, "%02um:", m);
+	sprintf(str, "%02us", s);
 }
 
 /*
  * Best effort calculation of the estimated pending runtime of a job.
  */
-static int thread_eta(struct thread_data *td, unsigned long elapsed)
+static unsigned long thread_eta(struct thread_data *td)
 {
 	unsigned long long bytes_total, bytes_done;
-	unsigned int eta_sec = 0;
+	unsigned long eta_sec = 0;
+	unsigned long elapsed;
+	uint64_t timeout;
+
+	elapsed = (mtime_since_now(&td->epoch) + 999) / 1000;
+	timeout = td->o.timeout / 1000000UL;
 
 	bytes_total = td->total_io_size;
 
+	if (td->flags & TD_F_NO_PROGRESS)
+		return -1;
+
+	if (td->o.fill_device && td->o.size  == -1ULL) {
+		if (!td->fill_device_size || td->fill_device_size == -1ULL)
+			return 0;
+
+		bytes_total = td->fill_device_size;
+	}
+
+	if (td->o.zone_size && td->o.zone_skip && bytes_total) {
+		unsigned int nr_zones;
+		uint64_t zone_bytes;
+
+		zone_bytes = bytes_total + td->o.zone_size + td->o.zone_skip;
+		nr_zones = (zone_bytes - 1) / (td->o.zone_size + td->o.zone_skip);
+		bytes_total -= nr_zones * td->o.zone_skip;
+	}
+
 	/*
-	 * if writing, bytes_total will be twice the size. If mixing,
-	 * assume a 50/50 split and thus bytes_total will be 50% larger.
+	 * if writing and verifying afterwards, bytes_total will be twice the
+	 * size. In a mixed workload, verify phase will be the size of the
+	 * first stage writes.
 	 */
-	if (td->verify) {
-		if (td_rw(td))
-			bytes_total = bytes_total * 3 / 2;
-		else
+	if (td->o.do_verify && td->o.verify && td_write(td)) {
+		if (td_rw(td)) {
+			unsigned int perc = 50;
+
+			if (td->o.rwmix[DDIR_WRITE])
+				perc = td->o.rwmix[DDIR_WRITE];
+
+			bytes_total += (bytes_total * perc) / 100;
+		} else
 			bytes_total <<= 1;
 	}
 
-	if (td->zone_size && td->zone_skip)
-		bytes_total /= (td->zone_skip / td->zone_size);
-
 	if (td->runstate == TD_RUNNING || td->runstate == TD_VERIFYING) {
-		double perc;
-
-		bytes_done = td->io_bytes[DDIR_READ] + td->io_bytes[DDIR_WRITE];
-		perc = (double) bytes_done / (double) bytes_total;
-		if (perc > 1.0)
-			perc = 1.0;
+		double perc, perc_t;
+
+		bytes_done = ddir_rw_sum(td->io_bytes);
+
+		if (bytes_total) {
+			perc = (double) bytes_done / (double) bytes_total;
+			if (perc > 1.0)
+				perc = 1.0;
+		} else
+			perc = 0.0;
+
+		if (td->o.time_based) {
+			if (timeout) {
+				perc_t = (double) elapsed / (double) timeout;
+				if (perc_t < perc)
+					perc = perc_t;
+			} else {
+				/*
+				 * Will never hit, we can't have time_based
+				 * without a timeout set.
+				 */
+				perc = 0.0;
+			}
+		}
 
-		eta_sec = (elapsed * (1.0 / perc)) - elapsed;
+		if (perc == 0.0) {
+			eta_sec = timeout;
+		} else {
+			eta_sec = (unsigned long) (elapsed * (1.0 / perc)) - elapsed;
+		}
 
-		if (td->timeout && eta_sec > (td->timeout - elapsed))
-			eta_sec = td->timeout - elapsed;
+		if (td->o.timeout &&
+		    eta_sec > (timeout + done_secs - elapsed))
+			eta_sec = timeout + done_secs - elapsed;
 	} else if (td->runstate == TD_NOT_CREATED || td->runstate == TD_CREATED
-			|| td->runstate == TD_INITIALIZED) {
-		int t_eta = 0, r_eta = 0;
+			|| td->runstate == TD_INITIALIZED
+			|| td->runstate == TD_SETTING_UP
+			|| td->runstate == TD_RAMP
+			|| td->runstate == TD_PRE_READING) {
+		int64_t t_eta = 0, r_eta = 0;
+		unsigned long long rate_bytes;
 
 		/*
 		 * We can only guess - assume it'll run the full timeout
 		 * if given, otherwise assume it'll run at the specified rate.
 		 */
-		if (td->timeout)
-			t_eta = td->timeout + td->start_delay - elapsed;
-		if (td->rate) {
-			r_eta = (bytes_total / 1024) / td->rate;
-			r_eta += td->start_delay - elapsed;
+		if (td->o.timeout) {
+			uint64_t __timeout = td->o.timeout;
+			uint64_t start_delay = td->o.start_delay;
+			uint64_t ramp_time = td->o.ramp_time;
+
+			t_eta = __timeout + start_delay;
+			if (!td->ramp_time_over) {
+				t_eta += ramp_time;
+			}
+			t_eta /= 1000000ULL;
+
+			if ((td->runstate == TD_RAMP) && in_ramp_time(td)) {
+				unsigned long ramp_left;
+
+				ramp_left = mtime_since_now(&td->epoch);
+				ramp_left = (ramp_left + 999) / 1000;
+				if (ramp_left <= t_eta)
+					t_eta -= ramp_left;
+			}
+		}
+		rate_bytes = 0;
+		if (td_read(td))
+			rate_bytes  = td->o.rate[DDIR_READ];
+		if (td_write(td))
+			rate_bytes += td->o.rate[DDIR_WRITE];
+		if (td_trim(td))
+			rate_bytes += td->o.rate[DDIR_TRIM];
+
+		if (rate_bytes) {
+			r_eta = bytes_total / rate_bytes;
+			r_eta += (td->o.start_delay / 1000000ULL);
 		}
 
 		if (r_eta && t_eta)
@@ -164,99 +301,424 @@ static int thread_eta(struct thread_data *td, unsigned long elapsed)
 	return eta_sec;
 }
 
+static void calc_rate(int unified_rw_rep, unsigned long mtime,
+		      unsigned long long *io_bytes,
+		      unsigned long long *prev_io_bytes, uint64_t *rate)
+{
+	int i;
+
+	for (i = 0; i < DDIR_RWDIR_CNT; i++) {
+		unsigned long long diff, this_rate;
+
+		diff = io_bytes[i] - prev_io_bytes[i];
+		if (mtime)
+			this_rate = ((1000 * diff) / mtime) / 1024; /* KiB/s */
+		else
+			this_rate = 0;
+
+		if (unified_rw_rep) {
+			rate[i] = 0;
+			rate[0] += this_rate;
+		} else
+			rate[i] = this_rate;
+
+		prev_io_bytes[i] = io_bytes[i];
+	}
+}
+
+static void calc_iops(int unified_rw_rep, unsigned long mtime,
+		      unsigned long long *io_iops,
+		      unsigned long long *prev_io_iops, unsigned int *iops)
+{
+	int i;
+
+	for (i = 0; i < DDIR_RWDIR_CNT; i++) {
+		unsigned long long diff, this_iops;
+
+		diff = io_iops[i] - prev_io_iops[i];
+		if (mtime)
+			this_iops = (diff * 1000) / mtime;
+		else
+			this_iops = 0;
+
+		if (unified_rw_rep) {
+			iops[i] = 0;
+			iops[0] += this_iops;
+		} else
+			iops[i] = this_iops;
+
+		prev_io_iops[i] = io_iops[i];
+	}
+}
+
+/*
+ * Allow a little slack - if we're within 95% of the time, allow ETA.
+ */
+bool eta_time_within_slack(unsigned int time)
+{
+	return time > ((eta_interval_msec * 95) / 100);
+}
+
 /*
  * Print status of the jobs we know about. This includes rate estimates,
  * ETA, thread state, etc.
  */
-void print_thread_status(void)
+bool calc_thread_status(struct jobs_eta *je, int force)
 {
-	unsigned long elapsed = mtime_since_genesis() / 1000;
-	int i, nr_running, nr_pending, t_rate, m_rate, *eta_secs, eta_sec;
 	struct thread_data *td;
-	char eta_str[32];
-	double perc = 0.0;
+	int i, unified_rw_rep;
+	uint64_t rate_time, disp_time, bw_avg_time, *eta_secs;
+	unsigned long long io_bytes[DDIR_RWDIR_CNT];
+	unsigned long long io_iops[DDIR_RWDIR_CNT];
+	struct timespec now;
+
+	static unsigned long long rate_io_bytes[DDIR_RWDIR_CNT];
+	static unsigned long long disp_io_bytes[DDIR_RWDIR_CNT];
+	static unsigned long long disp_io_iops[DDIR_RWDIR_CNT];
+	static struct timespec rate_prev_time, disp_prev_time;
+
+	if (!force) {
+		if (!(output_format & FIO_OUTPUT_NORMAL) &&
+		    f_out == stdout)
+			return false;
+		if (temp_stall_ts || eta_print == FIO_ETA_NEVER)
+			return false;
+
+		if (!isatty(STDOUT_FILENO) && (eta_print != FIO_ETA_ALWAYS))
+			return false;
+	}
 
-	static unsigned long long prev_io_bytes[2];
-	static struct timeval prev_time;
-	static unsigned int r_rate, w_rate;
-	unsigned long long io_bytes[2];
-	unsigned long mtime;
+	if (!ddir_rw_sum(rate_io_bytes))
+		fill_start_time(&rate_prev_time);
+	if (!ddir_rw_sum(disp_io_bytes))
+		fill_start_time(&disp_prev_time);
 
-	if (temp_stall_ts || terse_output)
-		return;
+	eta_secs = malloc(thread_number * sizeof(uint64_t));
+	memset(eta_secs, 0, thread_number * sizeof(uint64_t));
 
-	if (!prev_io_bytes[0] && !prev_io_bytes[1])
-		fill_start_time(&prev_time);
+	je->elapsed_sec = (mtime_since_genesis() + 999) / 1000;
 
-	eta_secs = malloc(thread_number * sizeof(int));
-	memset(eta_secs, 0, thread_number * sizeof(int));
-
-	io_bytes[0] = io_bytes[1] = 0;
-	nr_pending = nr_running = t_rate = m_rate = 0;
+	io_bytes[DDIR_READ] = io_bytes[DDIR_WRITE] = io_bytes[DDIR_TRIM] = 0;
+	io_iops[DDIR_READ] = io_iops[DDIR_WRITE] = io_iops[DDIR_TRIM] = 0;
+	bw_avg_time = ULONG_MAX;
+	unified_rw_rep = 0;
 	for_each_td(td, i) {
-		if (td->runstate == TD_RUNNING || td->runstate == TD_VERIFYING||
-		    td->runstate == TD_FSYNCING) {
-			nr_running++;
-			t_rate += td->rate;
-			m_rate += td->ratemin;
-		} else if (td->runstate < TD_RUNNING)
-			nr_pending++;
-
-		if (elapsed >= 3)
-			eta_secs[i] = thread_eta(td, elapsed);
+		unified_rw_rep += td->o.unified_rw_rep;
+		if (is_power_of_2(td->o.kb_base))
+			je->is_pow2 = 1;
+		je->unit_base = td->o.unit_base;
+		if (td->o.bw_avg_time < bw_avg_time)
+			bw_avg_time = td->o.bw_avg_time;
+		if (td->runstate == TD_RUNNING || td->runstate == TD_VERIFYING
+		    || td->runstate == TD_FSYNCING
+		    || td->runstate == TD_PRE_READING
+		    || td->runstate == TD_FINISHING) {
+			je->nr_running++;
+			if (td_read(td)) {
+				je->t_rate[0] += td->o.rate[DDIR_READ];
+				je->t_iops[0] += td->o.rate_iops[DDIR_READ];
+				je->m_rate[0] += td->o.ratemin[DDIR_READ];
+				je->m_iops[0] += td->o.rate_iops_min[DDIR_READ];
+			}
+			if (td_write(td)) {
+				je->t_rate[1] += td->o.rate[DDIR_WRITE];
+				je->t_iops[1] += td->o.rate_iops[DDIR_WRITE];
+				je->m_rate[1] += td->o.ratemin[DDIR_WRITE];
+				je->m_iops[1] += td->o.rate_iops_min[DDIR_WRITE];
+			}
+			if (td_trim(td)) {
+				je->t_rate[2] += td->o.rate[DDIR_TRIM];
+				je->t_iops[2] += td->o.rate_iops[DDIR_TRIM];
+				je->m_rate[2] += td->o.ratemin[DDIR_TRIM];
+				je->m_iops[2] += td->o.rate_iops_min[DDIR_TRIM];
+			}
+
+			je->files_open += td->nr_open_files;
+		} else if (td->runstate == TD_RAMP) {
+			je->nr_running++;
+			je->nr_ramp++;
+		} else if (td->runstate == TD_SETTING_UP)
+			je->nr_setting_up++;
+		else if (td->runstate < TD_RUNNING)
+			je->nr_pending++;
+
+		if (je->elapsed_sec >= 3)
+			eta_secs[i] = thread_eta(td);
 		else
 			eta_secs[i] = INT_MAX;
 
 		check_str_update(td);
-		io_bytes[0] += td->io_bytes[0];
-		io_bytes[1] += td->io_bytes[1];
-	}
 
-	if (exitall_on_terminate)
-		eta_sec = INT_MAX;
-	else
-		eta_sec = 0;
+		if (td->runstate > TD_SETTING_UP) {
+			int ddir;
+
+			for (ddir = 0; ddir < DDIR_RWDIR_CNT; ddir++) {
+				if (unified_rw_rep) {
+					io_bytes[0] += td->io_bytes[ddir];
+					io_iops[0] += td->io_blocks[ddir];
+				} else {
+					io_bytes[ddir] += td->io_bytes[ddir];
+					io_iops[ddir] += td->io_blocks[ddir];
+				}
+			}
+		}
+	}
 
-	for_each_td(td, i) {
-		if (exitall_on_terminate) {
-			if (eta_secs[i] < eta_sec)
-				eta_sec = eta_secs[i];
-		} else {
-			if (eta_secs[i] > eta_sec)
-				eta_sec = eta_secs[i];
+	if (exitall_on_terminate) {
+		je->eta_sec = INT_MAX;
+		for_each_td(td, i) {
+			if (eta_secs[i] < je->eta_sec)
+				je->eta_sec = eta_secs[i];
+		}
+	} else {
+		unsigned long eta_stone = 0;
+
+		je->eta_sec = 0;
+		for_each_td(td, i) {
+			if ((td->runstate == TD_NOT_CREATED) && td->o.stonewall)
+				eta_stone += eta_secs[i];
+			else {
+				if (eta_secs[i] > je->eta_sec)
+					je->eta_sec = eta_secs[i];
+			}
 		}
+		je->eta_sec += eta_stone;
 	}
 
 	free(eta_secs);
 
-	if (eta_sec != INT_MAX && elapsed) {
-		perc = (double) elapsed / (double) (elapsed + eta_sec);
-		eta_to_str(eta_str, eta_sec);
+	fio_gettime(&now, NULL);
+	rate_time = mtime_since(&rate_prev_time, &now);
+
+	if (write_bw_log && rate_time > bw_avg_time && !in_ramp_time(td)) {
+		calc_rate(unified_rw_rep, rate_time, io_bytes, rate_io_bytes,
+				je->rate);
+		memcpy(&rate_prev_time, &now, sizeof(now));
+		add_agg_sample(sample_val(je->rate[DDIR_READ]), DDIR_READ, 0);
+		add_agg_sample(sample_val(je->rate[DDIR_WRITE]), DDIR_WRITE, 0);
+		add_agg_sample(sample_val(je->rate[DDIR_TRIM]), DDIR_TRIM, 0);
+	}
+
+	disp_time = mtime_since(&disp_prev_time, &now);
+
+	if (!force && !eta_time_within_slack(disp_time))
+		return false;
+
+	calc_rate(unified_rw_rep, disp_time, io_bytes, disp_io_bytes, je->rate);
+	calc_iops(unified_rw_rep, disp_time, io_iops, disp_io_iops, je->iops);
+
+	memcpy(&disp_prev_time, &now, sizeof(now));
+
+	if (!force && !je->nr_running && !je->nr_pending)
+		return false;
+
+	je->nr_threads = thread_number;
+	update_condensed_str(__run_str, run_str);
+	memcpy(je->run_str, run_str, strlen(run_str));
+	return true;
+}
+
+static int gen_eta_str(struct jobs_eta *je, char *p, size_t left,
+		       char **rate_str, char **iops_str)
+{
+	bool has_r = je->rate[DDIR_READ] || je->iops[DDIR_READ];
+	bool has_w = je->rate[DDIR_WRITE] || je->iops[DDIR_WRITE];
+	bool has_t = je->rate[DDIR_TRIM] || je->iops[DDIR_TRIM];
+	int l = 0;
+
+	if (!has_r && !has_w && !has_t)
+		return 0;
+
+	if (has_r) {
+		l += snprintf(p + l, left - l, "[r=%s", rate_str[DDIR_READ]);
+		if (!has_w)
+			l += snprintf(p + l, left - l, "]");
+	}
+	if (has_w) {
+		if (has_r)
+			l += snprintf(p + l, left - l, ",");
+		else
+			l += snprintf(p + l, left - l, "[");
+		l += snprintf(p + l, left - l, "w=%s", rate_str[DDIR_WRITE]);
+		if (!has_t)
+			l += snprintf(p + l, left - l, "]");
+	}
+	if (has_t) {
+		if (has_r || has_w)
+			l += snprintf(p + l, left - l, ",");
+		else if (!has_r && !has_w)
+			l += snprintf(p + l, left - l, "[");
+		l += snprintf(p + l, left - l, "t=%s]", rate_str[DDIR_TRIM]);
+	}
+	if (has_r) {
+		l += snprintf(p + l, left - l, "[r=%s", iops_str[DDIR_READ]);
+		if (!has_w)
+			l += snprintf(p + l, left - l, " IOPS]");
+	}
+	if (has_w) {
+		if (has_r)
+			l += snprintf(p + l, left - l, ",");
+		else
+			l += snprintf(p + l, left - l, "[");
+		l += snprintf(p + l, left - l, "w=%s", iops_str[DDIR_WRITE]);
+		if (!has_t)
+			l += snprintf(p + l, left - l, " IOPS]");
+	}
+	if (has_t) {
+		if (has_r || has_w)
+			l += snprintf(p + l, left - l, ",");
+		else if (!has_r && !has_w)
+			l += snprintf(p + l, left - l, "[");
+		l += snprintf(p + l, left - l, "t=%s IOPS]", iops_str[DDIR_TRIM]);
+	}
+
+	return l;
+}
+
+void display_thread_status(struct jobs_eta *je)
+{
+	static struct timespec disp_eta_new_line;
+	static int eta_new_line_init, eta_new_line_pending;
+	static int linelen_last;
+	static int eta_good;
+	char output[__THREAD_RUNSTR_SZ(REAL_MAX_JOBS) + 512], *p = output;
+	char eta_str[128];
+	double perc = 0.0;
+
+	if (je->eta_sec != INT_MAX && je->elapsed_sec) {
+		perc = (double) je->elapsed_sec / (double) (je->elapsed_sec + je->eta_sec);
+		eta_to_str(eta_str, je->eta_sec);
 	}
 
-	mtime = mtime_since_now(&prev_time);
-	if (mtime > 1000) {
-		r_rate = (io_bytes[0] - prev_io_bytes[0]) / mtime;
-		w_rate = (io_bytes[1] - prev_io_bytes[1]) / mtime;
-		fio_gettime(&prev_time, NULL);
-		memcpy(prev_io_bytes, io_bytes, sizeof(io_bytes));
+	if (eta_new_line_pending) {
+		eta_new_line_pending = 0;
+		linelen_last = 0;
+		p += sprintf(p, "\n");
 	}
 
-	if (!nr_running && !nr_pending)
-		return;
+	p += sprintf(p, "Jobs: %d (f=%d)", je->nr_running, je->files_open);
+
+	/* rate limits, if any */
+	if (je->m_rate[0] || je->m_rate[1] || je->m_rate[2] ||
+	    je->t_rate[0] || je->t_rate[1] || je->t_rate[2]) {
+		char *tr, *mr;
+
+		mr = num2str(je->m_rate[0] + je->m_rate[1] + je->m_rate[2],
+				je->sig_figs, 0, je->is_pow2, N2S_BYTEPERSEC);
+		tr = num2str(je->t_rate[0] + je->t_rate[1] + je->t_rate[2],
+				je->sig_figs, 0, je->is_pow2, N2S_BYTEPERSEC);
+
+		p += sprintf(p, ", %s-%s", mr, tr);
+		free(tr);
+		free(mr);
+	} else if (je->m_iops[0] || je->m_iops[1] || je->m_iops[2] ||
+		   je->t_iops[0] || je->t_iops[1] || je->t_iops[2]) {
+		p += sprintf(p, ", %d-%d IOPS",
+					je->m_iops[0] + je->m_iops[1] + je->m_iops[2],
+					je->t_iops[0] + je->t_iops[1] + je->t_iops[2]);
+	}
+
+	/* current run string, % done, bandwidth, iops, eta */
+	if (je->eta_sec != INT_MAX && je->nr_running) {
+		char perc_str[32];
+		char *iops_str[DDIR_RWDIR_CNT];
+		char *rate_str[DDIR_RWDIR_CNT];
+		size_t left;
+		int l;
+		int ddir;
+		int linelen;
+
+		if ((!je->eta_sec && !eta_good) || je->nr_ramp == je->nr_running ||
+		    je->eta_sec == -1)
+			strcpy(perc_str, "-.-%");
+		else {
+			double mult = 100.0;
+
+			if (je->nr_setting_up && je->nr_running)
+				mult *= (1.0 - (double) je->nr_setting_up / (double) je->nr_running);
+
+			eta_good = 1;
+			perc *= mult;
+			sprintf(perc_str, "%3.1f%%", perc);
+		}
+
+		for (ddir = 0; ddir < DDIR_RWDIR_CNT; ddir++) {
+			rate_str[ddir] = num2str(je->rate[ddir], 4,
+						1024, je->is_pow2, je->unit_base);
+			iops_str[ddir] = num2str(je->iops[ddir], 4, 1, 0, N2S_NONE);
+		}
+
+		left = sizeof(output) - (p - output) - 1;
+		l = snprintf(p, left, ": [%s][%s]", je->run_str, perc_str);
+		l += gen_eta_str(je, p + l, left - l, rate_str, iops_str);
+		l += snprintf(p + l, left - l, "[eta %s]", eta_str);
+
+		/* If truncation occurred adjust l so p is on the null */
+		if (l >= left)
+			l = left - 1;
+		p += l;
+		linelen = p - output;
+		if (l >= 0 && linelen < linelen_last)
+			p += sprintf(p, "%*s", linelen_last - linelen, "");
+		linelen_last = linelen;
+
+		for (ddir = 0; ddir < DDIR_RWDIR_CNT; ddir++) {
+			free(rate_str[ddir]);
+			free(iops_str[ddir]);
+		}
+	}
+	sprintf(p, "\r");
+
+	printf("%s", output);
 
-	printf("Threads running: %d", nr_running);
-	if (m_rate || t_rate)
-		printf(", commitrate %d/%dKiB/sec", t_rate, m_rate);
-	if (eta_sec != INT_MAX && nr_running) {
-		perc *= 100.0;
-		printf(": [%s] [%3.2f%% done] [%6u/%6u kb/s] [eta %s]", run_str, perc, r_rate, w_rate, eta_str);
+	if (!eta_new_line_init) {
+		fio_gettime(&disp_eta_new_line, NULL);
+		eta_new_line_init = 1;
+	} else if (eta_new_line && mtime_since_now(&disp_eta_new_line) > eta_new_line) {
+		fio_gettime(&disp_eta_new_line, NULL);
+		eta_new_line_pending = 1;
 	}
-	printf("\r");
+
 	fflush(stdout);
 }
 
-void print_status_init(int thread_number)
+struct jobs_eta *get_jobs_eta(bool force, size_t *size)
+{
+	struct jobs_eta *je;
+
+	if (!thread_number)
+		return NULL;
+
+	*size = sizeof(*je) + THREAD_RUNSTR_SZ + 8;
+	je = malloc(*size);
+	if (!je)
+		return NULL;
+	memset(je, 0, *size);
+
+	if (!calc_thread_status(je, force)) {
+		free(je);
+		return NULL;
+	}
+
+	*size = sizeof(*je) + strlen((char *) je->run_str) + 1;
+	return je;
+}
+
+void print_thread_status(void)
+{
+	struct jobs_eta *je;
+	size_t size;
+
+	je = get_jobs_eta(false, &size);
+	if (je)
+		display_thread_status(je);
+
+	free(je);
+}
+
+void print_status_init(int thr_number)
 {
-	run_str[thread_number] = 'P';
+	DRD_IGNORE_VAR(__run_str);
+	__run_str[thr_number] = 'P';
+	update_condensed_str(__run_str, run_str);
 }