[fio.git] / eta.c

/*
 * Status and ETA code
 */
#include <unistd.h>
#include <fcntl.h>
#include <string.h>

#include "fio.h"

static char __run_str[REAL_MAX_JOBS + 1];
static char run_str[__THREAD_RUNSTR_SZ(REAL_MAX_JOBS)];

static void update_condensed_str(char *run_str, char *run_str_condensed)
{
	int i, ci, last, nr;
	size_t len;

	len = strlen(run_str);
	if (!len)
		return;

	last = 0;
	nr = 0;
	ci = 0;
	for (i = 0; i < len; i++) {
		if (!last) {
new:
			run_str_condensed[ci] = run_str[i];
			last = run_str[i];
			nr = 1;
			ci++;
		} else if (last == run_str[i]) {
			nr++;
		} else {
			ci += sprintf(&run_str_condensed[ci], "(%u),", nr);
			goto new;
		}
	}

	if (nr)
		ci += sprintf(&run_str_condensed[ci], "(%u)", nr);

	run_str_condensed[ci + 1] = '\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];

	switch (td->runstate) {
	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_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->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_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;
	update_condensed_str(__run_str, run_str);
}

/*
 * Convert seconds to a printable string.
 */
void eta_to_str(char *str, unsigned long eta_sec)
{
	unsigned int d, h, m, s;
	int disp_hour = 0;

	s = eta_sec % 60;
	eta_sec /= 60;
	m = eta_sec % 60;
	eta_sec /= 60;
	h = eta_sec % 24;
	eta_sec /= 24;
	d = eta_sec;

	if (d) {
		disp_hour = 1;
		str += sprintf(str, "%02ud:", d);
	}

	if (h || disp_hour)
		str += sprintf(str, "%02uh:", h);

	str += sprintf(str, "%02um:", m);
	str += 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 long bytes_total, bytes_done;
	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->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 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->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->runstate == TD_RUNNING || td->runstate == TD_VERIFYING) {
		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 = (unsigned long) (elapsed * (1.0 / perc)) - 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
			|| td->runstate == TD_SETTING_UP
			|| td->runstate == TD_RAMP
			|| td->runstate == TD_PRE_READING) {
		int 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->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 + ramp_time;
			t_eta /= 1000000ULL;

			if (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 = ddir_rw_sum(td->o.rate);
		if (rate_bytes) {
			r_eta = (bytes_total / 1024) / rate_bytes;
			r_eta += (td->o.start_delay / 1000000ULL);
		}

		if (r_eta && t_eta)
			eta_sec = min(r_eta, t_eta);
		else if (r_eta)
			eta_sec = r_eta;
		else if (t_eta)
			eta_sec = t_eta;
		else
			eta_sec = 0;
	} else {
		/*
		 * thread is already done or waiting for fsync
		 */
		eta_sec = 0;
	}

	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, unsigned int *rate)
{
	int i;

	for (i = 0; i < DDIR_RWDIR_CNT; i++) {
		unsigned long long diff;

		diff = io_bytes[i] - prev_io_bytes[i];
		if (unified_rw_rep) {
			rate[i] = 0;
			rate[0] += ((1000 * diff) / mtime) / 1024;
		} else
			rate[i] = ((1000 * diff) / mtime) / 1024;

		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;

		diff = io_iops[i] - prev_io_iops[i];
		if (unified_rw_rep) {
			iops[i] = 0;
			iops[0] += (diff * 1000) / mtime;
		} else
			iops[i] = (diff * 1000) / mtime;

		prev_io_iops[i] = io_iops[i];
	}
}

/*
 * Print status of the jobs we know about. This includes rate estimates,
 * ETA, thread state, etc.
 */
int calc_thread_status(struct jobs_eta *je, int force)
{
	struct thread_data *td;
	int i, unified_rw_rep;
	unsigned long 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 timeval 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 timeval rate_prev_time, disp_prev_time;

	if (!force) {
		if (output_format != FIO_OUTPUT_NORMAL &&
		    f_out == stdout)
			return 0;
		if (temp_stall_ts || eta_print == FIO_ETA_NEVER)
			return 0;

		if (!isatty(STDOUT_FILENO) && (eta_print != FIO_ETA_ALWAYS))
			return 0;
	}

	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);

	eta_secs = malloc(thread_number * sizeof(unsigned long));
	memset(eta_secs, 0, thread_number * sizeof(unsigned long));

	je->elapsed_sec = (mtime_since_genesis() + 999) / 1000;

	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) {
		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_running++;
			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);

		if (td->runstate > TD_SETTING_UP) {
			int ddir;

			for (ddir = DDIR_READ; 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];
				}
			}
		}
	}

	if (exitall_on_terminate)
		je->eta_sec = INT_MAX;
	else
		je->eta_sec = 0;

	for_each_td(td, i) {
		if (exitall_on_terminate) {
			if (eta_secs[i] < je->eta_sec)
				je->eta_sec = eta_secs[i];
		} else {
			if (eta_secs[i] > je->eta_sec)
				je->eta_sec = eta_secs[i];
		}
	}

	free(eta_secs);

	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(je->rate[DDIR_READ], DDIR_READ, 0);
		add_agg_sample(je->rate[DDIR_WRITE], DDIR_WRITE, 0);
		add_agg_sample(je->rate[DDIR_TRIM], DDIR_TRIM, 0);
	}

	disp_time = mtime_since(&disp_prev_time, &now);

	/*
	 * Allow a little slack, the target is to print it every 1000 msecs
	 */
	if (!force && disp_time < 900)
		return 0;

	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 0;

	je->nr_threads = thread_number;
	update_condensed_str(__run_str, run_str);
	memcpy(je->run_str, run_str, strlen(run_str));
	return 1;
}

void display_thread_status(struct jobs_eta *je)
{
	static struct timeval disp_eta_new_line;
	static int eta_new_line_init, eta_new_line_pending;
	static int linelen_last;
	static int eta_good;
	char output[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);
	}

	if (eta_new_line_pending) {
		eta_new_line_pending = 0;
		p += sprintf(p, "\n");
	}

	p += sprintf(p, "Jobs: %d (f=%d)", je->nr_running, je->files_open);
	if (je->m_rate[0] || je->m_rate[1] || je->t_rate[0] || je->t_rate[1]) {
		char *tr, *mr;

		mr = num2str(je->m_rate[0] + je->m_rate[1], 4, 0, je->is_pow2, 8);
		tr = num2str(je->t_rate[0] + je->t_rate[1], 4, 0, je->is_pow2, 8);
		p += sprintf(p, ", CR=%s/%s KB/s", tr, mr);
		free(tr);
		free(mr);
	} else if (je->m_iops[0] || je->m_iops[1] || je->t_iops[0] || je->t_iops[1]) {
		p += sprintf(p, ", CR=%d/%d IOPS",
					je->t_iops[0] + je->t_iops[1],
					je->m_iops[0] + je->m_iops[1]);
	}
	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;

		if ((!je->eta_sec && !eta_good) || je->nr_ramp == je->nr_running)
			strcpy(perc_str, "-.-% done");
		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%% done", perc);
		}

		for (ddir = DDIR_READ; ddir < DDIR_RWDIR_CNT; ddir++) {
			rate_str[ddir] = num2str(je->rate[ddir], 5,
						1024, je->is_pow2, je->unit_base);
			iops_str[ddir] = num2str(je->iops[ddir], 4, 1, 0, 0);
		}

		left = sizeof(output) - (p - output) - 1;

		l = snprintf(p, left, ": [%s] [%s] [%s/%s/%s /s] [%s/%s/%s iops] [eta %s]",
				je->run_str, perc_str, rate_str[DDIR_READ],
				rate_str[DDIR_WRITE], rate_str[DDIR_TRIM],
				iops_str[DDIR_READ], iops_str[DDIR_WRITE],
				iops_str[DDIR_TRIM], eta_str);
		p += l;
		if (l >= 0 && l < linelen_last)
			p += sprintf(p, "%*s", linelen_last - l, "");
		linelen_last = l;

		for (ddir = DDIR_READ; ddir < DDIR_RWDIR_CNT; ddir++) {
			free(rate_str[ddir]);
			free(iops_str[ddir]);
		}
	}
	p += sprintf(p, "\r");

	printf("%s", output);

	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 * 1000) {
		fio_gettime(&disp_eta_new_line, NULL);
		eta_new_line_pending = 1;
	}

	fflush(stdout);
}

struct jobs_eta *get_jobs_eta(int force, size_t *size)
{
	struct jobs_eta *je;

	if (!thread_number)
		return NULL;

	*size = sizeof(*je) + THREAD_RUNSTR_SZ;
	je = malloc(*size);
	memset(je, 0, *size);

	if (!calc_thread_status(je, 0)) {
		free(je);
		return NULL;
	}

	return je;
}

void print_thread_status(void)
{
	struct jobs_eta *je;
	size_t size;

	je = get_jobs_eta(0, &size);
	if (je)
		display_thread_status(je);

	free(je);
}

void print_status_init(int thr_number)
{
	__run_str[thr_number] = 'P';
	update_condensed_str(__run_str, run_str);
}
Commit	Line	Data
263e529f JA	1	/*
	2	* Status and ETA code
	3	*/
	4	#include <unistd.h>
	5	#include <fcntl.h>
	6	#include <string.h>
	7
	8	#include "fio.h"
263e529f	9
3e2e48a7	10	static char __run_str[REAL_MAX_JOBS + 1];
11155857	11	static char run_str[__THREAD_RUNSTR_SZ(REAL_MAX_JOBS)];
3e2e48a7 JA	12
	13	static void update_condensed_str(char run_str, char run_str_condensed)
	14	{
	15	int i, ci, last, nr;
	16	size_t len;
	17
	18	len = strlen(run_str);
	19	if (!len)
	20	return;
	21
	22	last = 0;
	23	nr = 0;
	24	ci = 0;
	25	for (i = 0; i < len; i++) {
	26	if (!last) {
	27	new:
	28	run_str_condensed[ci] = run_str[i];
	29	last = run_str[i];
	30	nr = 1;
	31	ci++;
	32	} else if (last == run_str[i]) {
	33	nr++;
	34	} else {
6c784104	35	ci += sprintf(&run_str_condensed[ci], "(%u),", nr);
3e2e48a7 JA	36	goto new;
	37	}
	38	}
	39
	40	if (nr)
6c784104	41	ci += sprintf(&run_str_condensed[ci], "(%u)", nr);
3e2e48a7	42
6c784104 JA	43	run_str_condensed[ci + 1] = '\0';
6c784104 JA	44	}
263e529f JA	45
	46	/*
	47	* Sets the status of the 'td' in the printed status map.
	48	*/
	49	static void check_str_update(struct thread_data *td)
	50	{
3e2e48a7	51	char c = __run_str[td->thread_number - 1];
263e529f JA	52
263e529f JA	53	switch (td->runstate) {
5ec10eaa	54	case TD_REAPED:
4f7e57a4 JA	55	if (td->error)
4f7e57a4 JA	56	c = 'X';
a5e371a6 JA	57	else if (td->sig)
a5e371a6 JA	58	c = 'K';
4f7e57a4 JA	59	else
4f7e57a4 JA	60	c = '_';
5ec10eaa JA	61	break;
	62	case TD_EXITED:
	63	c = 'E';
	64	break;
b29ee5b3 JA	65	case TD_RAMP:
	66	c = '/';
	67	break;
5ec10eaa JA	68	case TD_RUNNING:
5ec10eaa JA	69	if (td_rw(td)) {
e3b3f81e JA	70	if (td_random(td)) {
	71	if (td->o.rwmix[DDIR_READ] == 100)
	72	c = 'r';
	73	else if (td->o.rwmix[DDIR_WRITE] == 100)
	74	c = 'w';
	75	else
	76	c = 'm';
	77	} else {
	78	if (td->o.rwmix[DDIR_READ] == 100)
	79	c = 'R';
	80	else if (td->o.rwmix[DDIR_WRITE] == 100)
	81	c = 'W';
	82	else
	83	c = 'M';
	84	}
5ec10eaa JA	85	} else if (td_read(td)) {
	86	if (td_random(td))
	87	c = 'r';
	88	else
	89	c = 'R';
6eaf09d6	90	} else if (td_write(td)) {
5ec10eaa JA	91	if (td_random(td))
	92	c = 'w';
	93	else
	94	c = 'W';
6eaf09d6 SL	95	} else {
	96	if (td_random(td))
	97	c = 'd';
	98	else
	99	c = 'D';
5ec10eaa JA	100	}
5ec10eaa JA	101	break;
b0f65863 JA	102	case TD_PRE_READING:
	103	c = 'p';
	104	break;
5ec10eaa JA	105	case TD_VERIFYING:
	106	c = 'V';
	107	break;
	108	case TD_FSYNCING:
	109	c = 'F';
	110	break;
3d434057 JA	111	case TD_FINISHING:
	112	c = 'f';
	113	break;
5ec10eaa JA	114	case TD_CREATED:
	115	c = 'C';
	116	break;
	117	case TD_INITIALIZED:
9c6f6316	118	case TD_SETTING_UP:
5ec10eaa JA	119	c = 'I';
	120	break;
	121	case TD_NOT_CREATED:
	122	c = 'P';
	123	break;
	124	default:
	125	log_err("state %d\n", td->runstate);
263e529f JA	126	}
263e529f JA	127
3e2e48a7 JA	128	__run_str[td->thread_number - 1] = c;
3e2e48a7 JA	129	update_condensed_str(__run_str, run_str);
263e529f JA	130	}
	131
	132	/*
	133	* Convert seconds to a printable string.
	134	*/
3e47bd25	135	void eta_to_str(char *str, unsigned long eta_sec)
263e529f JA	136	{
263e529f JA	137	unsigned int d, h, m, s;
165faf16	138	int disp_hour = 0;
263e529f	139
263e529f JA	140	s = eta_sec % 60;
	141	eta_sec /= 60;
	142	m = eta_sec % 60;
	143	eta_sec /= 60;
	144	h = eta_sec % 24;
	145	eta_sec /= 24;
	146	d = eta_sec;
	147
165faf16 JA	148	if (d) {
165faf16 JA	149	disp_hour = 1;
1e97cce9	150	str += sprintf(str, "%02ud:", d);
263e529f	151	}
165faf16 JA	152
165faf16 JA	153	if (h \|\| disp_hour)
1e97cce9	154	str += sprintf(str, "%02uh:", h);
263e529f	155
1e97cce9 JA	156	str += sprintf(str, "%02um:", m);
1e97cce9 JA	157	str += sprintf(str, "%02us", s);
263e529f JA	158	}
	159
	160	/*
	161	* Best effort calculation of the estimated pending runtime of a job.
	162	*/
b29ee5b3	163	static int thread_eta(struct thread_data *td)
263e529f JA	164	{
263e529f JA	165	unsigned long long bytes_total, bytes_done;
1e97cce9	166	unsigned long eta_sec = 0;
b29ee5b3	167	unsigned long elapsed;
0de5b26f	168	uint64_t timeout;
b29ee5b3 JA	169
b29ee5b3 JA	170	elapsed = (mtime_since_now(&td->epoch) + 999) / 1000;
0de5b26f	171	timeout = td->o.timeout / 1000000UL;
263e529f JA	172
	173	bytes_total = td->total_io_size;
	174
2e3bd4c2 JA	175	if (td->o.fill_device && td->o.size == -1ULL) {
	176	if (!td->fill_device_size \|\| td->fill_device_size == -1ULL)
	177	return 0;
	178
	179	bytes_total = td->fill_device_size;
	180	}
	181
0d73a2f9 JA	182	if (td->o.zone_size && td->o.zone_skip && bytes_total) {
	183	unsigned int nr_zones;
	184	uint64_t zone_bytes;
	185
	186	zone_bytes = bytes_total + td->o.zone_size + td->o.zone_skip;
	187	nr_zones = (zone_bytes - 1) / (td->o.zone_size + td->o.zone_skip);
	188	bytes_total -= nr_zones * td->o.zone_skip;
	189	}
	190
74939e38	191	/*
165eb050 JA	192	* if writing and verifying afterwards, bytes_total will be twice the
	193	* size. In a mixed workload, verify phase will be the size of the
	194	* first stage writes.
74939e38	195	*/
0dd8377f	196	if (td->o.do_verify && td->o.verify && td_write(td)) {
165eb050 JA	197	if (td_rw(td)) {
	198	unsigned int perc = 50;
	199
	200	if (td->o.rwmix[DDIR_WRITE])
	201	perc = td->o.rwmix[DDIR_WRITE];
	202
	203	bytes_total += (bytes_total * perc) / 100;
	204	} else
74939e38 JA	205	bytes_total <<= 1;
	206	}
	207
263e529f	208	if (td->runstate == TD_RUNNING \|\| td->runstate == TD_VERIFYING) {
cf4464ca	209	double perc, perc_t;
263e529f	210
342f4be4	211	bytes_done = ddir_rw_sum(td->io_bytes);
476882d7 JA	212
	213	if (bytes_total) {
	214	perc = (double) bytes_done / (double) bytes_total;
	215	if (perc > 1.0)
	216	perc = 1.0;
	217	} else
	218	perc = 0.0;
263e529f	219
cf4464ca	220	if (td->o.time_based) {
476882d7 JA	221	if (timeout) {
	222	perc_t = (double) elapsed / (double) timeout;
	223	if (perc_t < perc)
	224	perc = perc_t;
	225	} else {
	226	/*
	227	* Will never hit, we can't have time_based
	228	* without a timeout set.
	229	*/
	230	perc = 0.0;
	231	}
cf4464ca JA	232	}
cf4464ca JA	233
1e97cce9	234	eta_sec = (unsigned long) (elapsed * (1.0 / perc)) - elapsed;
263e529f	235
d3eeeabc	236	if (td->o.timeout &&
0de5b26f JA	237	eta_sec > (timeout + done_secs - elapsed))
0de5b26f JA	238	eta_sec = timeout + done_secs - elapsed;
263e529f	239	} else if (td->runstate == TD_NOT_CREATED \|\| td->runstate == TD_CREATED
b29ee5b3	240	\|\| td->runstate == TD_INITIALIZED
714e85f3	241	\|\| td->runstate == TD_SETTING_UP
b0f65863 JA	242	\|\| td->runstate == TD_RAMP
b0f65863 JA	243	\|\| td->runstate == TD_PRE_READING) {
263e529f	244	int t_eta = 0, r_eta = 0;
342f4be4	245	unsigned long long rate_bytes;
263e529f JA	246
	247	/*
	248	* We can only guess - assume it'll run the full timeout
	249	* if given, otherwise assume it'll run at the specified rate.
	250	*/
b29ee5b3	251	if (td->o.timeout) {
0de5b26f JA	252	uint64_t timeout = td->o.timeout;
	253	uint64_t start_delay = td->o.start_delay;
	254	uint64_t ramp_time = td->o.ramp_time;
	255
	256	t_eta = timeout + start_delay + ramp_time;
	257	t_eta /= 1000000ULL;
b29ee5b3 JA	258
	259	if (in_ramp_time(td)) {
	260	unsigned long ramp_left;
	261
cda99fa0	262	ramp_left = mtime_since_now(&td->epoch);
b29ee5b3 JA	263	ramp_left = (ramp_left + 999) / 1000;
	264	if (ramp_left <= t_eta)
	265	t_eta -= ramp_left;
	266	}
	267	}
342f4be4 JA	268	rate_bytes = ddir_rw_sum(td->o.rate);
	269	if (rate_bytes) {
	270	r_eta = (bytes_total / 1024) / rate_bytes;
0de5b26f	271	r_eta += (td->o.start_delay / 1000000ULL);
263e529f JA	272	}
	273
	274	if (r_eta && t_eta)
	275	eta_sec = min(r_eta, t_eta);
	276	else if (r_eta)
	277	eta_sec = r_eta;
	278	else if (t_eta)
	279	eta_sec = t_eta;
	280	else
	281	eta_sec = 0;
	282	} else {
	283	/*
	284	* thread is already done or waiting for fsync
	285	*/
	286	eta_sec = 0;
	287	}
	288
	289	return eta_sec;
	290	}
	291
771e58be JA	292	static void calc_rate(int unified_rw_rep, unsigned long mtime,
771e58be JA	293	unsigned long long *io_bytes,
46fda8d0 JA	294	unsigned long long prev_io_bytes, unsigned int rate)
46fda8d0 JA	295	{
b7f05eb0 JA	296	int i;
b7f05eb0 JA	297
6eaf09d6	298	for (i = 0; i < DDIR_RWDIR_CNT; i++) {
b7f05eb0 JA	299	unsigned long long diff;
	300
	301	diff = io_bytes[i] - prev_io_bytes[i];
771e58be JA	302	if (unified_rw_rep) {
	303	rate[i] = 0;
	304	rate[0] += ((1000 * diff) / mtime) / 1024;
	305	} else
	306	rate[i] = ((1000 * diff) / mtime) / 1024;
6eaf09d6 SL	307
6eaf09d6 SL	308	prev_io_bytes[i] = io_bytes[i];
b7f05eb0	309	}
46fda8d0	310	}
5ec10eaa	311
771e58be JA	312	static void calc_iops(int unified_rw_rep, unsigned long mtime,
771e58be JA	313	unsigned long long *io_iops,
adb02ba8 JA	314	unsigned long long prev_io_iops, unsigned int iops)
adb02ba8 JA	315	{
6eaf09d6 SL	316	int i;
	317
	318	for (i = 0; i < DDIR_RWDIR_CNT; i++) {
771e58be JA	319	unsigned long long diff;
	320
	321	diff = io_iops[i] - prev_io_iops[i];
	322	if (unified_rw_rep) {
	323	iops[i] = 0;
	324	iops[0] += (diff * 1000) / mtime;
	325	} else
	326	iops[i] = (diff * 1000) / mtime;
	327
6eaf09d6 SL	328	prev_io_iops[i] = io_iops[i];
6eaf09d6 SL	329	}
adb02ba8 JA	330	}
adb02ba8 JA	331
263e529f JA	332	/*
	333	* Print status of the jobs we know about. This includes rate estimates,
	334	* ETA, thread state, etc.
	335	*/
af9c9fb3	336	int calc_thread_status(struct jobs_eta *je, int force)
263e529f	337	{
34572e28	338	struct thread_data *td;
771e58be	339	int i, unified_rw_rep;
b75a394f	340	unsigned long rate_time, disp_time, bw_avg_time, *eta_secs;
6eaf09d6 SL	341	unsigned long long io_bytes[DDIR_RWDIR_CNT];
6eaf09d6 SL	342	unsigned long long io_iops[DDIR_RWDIR_CNT];
46fda8d0 JA	343	struct timeval now;
46fda8d0 JA	344
6eaf09d6 SL	345	static unsigned long long rate_io_bytes[DDIR_RWDIR_CNT];
	346	static unsigned long long disp_io_bytes[DDIR_RWDIR_CNT];
	347	static unsigned long long disp_io_iops[DDIR_RWDIR_CNT];
46fda8d0	348	static struct timeval rate_prev_time, disp_prev_time;
6043c579	349
af9c9fb3	350	if (!force) {
fc47559b JA	351	if (output_format != FIO_OUTPUT_NORMAL &&
fc47559b JA	352	f_out == stdout)
f3afa57e JA	353	return 0;
f3afa57e JA	354	if (temp_stall_ts \|\| eta_print == FIO_ETA_NEVER)
af9c9fb3	355	return 0;
e592a06b	356
af9c9fb3 JA	357	if (!isatty(STDOUT_FILENO) && (eta_print != FIO_ETA_ALWAYS))
	358	return 0;
	359	}
263e529f	360
342f4be4	361	if (!ddir_rw_sum(rate_io_bytes))
46fda8d0	362	fill_start_time(&rate_prev_time);
342f4be4	363	if (!ddir_rw_sum(disp_io_bytes))
46fda8d0	364	fill_start_time(&disp_prev_time);
6043c579	365
d3eeeabc JA	366	eta_secs = malloc(thread_number * sizeof(unsigned long));
d3eeeabc JA	367	memset(eta_secs, 0, thread_number * sizeof(unsigned long));
263e529f	368
b75a394f JA	369	je->elapsed_sec = (mtime_since_genesis() + 999) / 1000;
b75a394f JA	370
6eaf09d6 SL	371	io_bytes[DDIR_READ] = io_bytes[DDIR_WRITE] = io_bytes[DDIR_TRIM] = 0;
6eaf09d6 SL	372	io_iops[DDIR_READ] = io_iops[DDIR_WRITE] = io_iops[DDIR_TRIM] = 0;
bb3884d8	373	bw_avg_time = ULONG_MAX;
771e58be	374	unified_rw_rep = 0;
34572e28	375	for_each_td(td, i) {
771e58be	376	unified_rw_rep += td->o.unified_rw_rep;
b7f05eb0 JA	377	if (is_power_of_2(td->o.kb_base))
b7f05eb0 JA	378	je->is_pow2 = 1;
ad705bcb	379	je->unit_base = td->o.unit_base;
2dc1bbeb JA	380	if (td->o.bw_avg_time < bw_avg_time)
2dc1bbeb JA	381	bw_avg_time = td->o.bw_avg_time;
46fda8d0	382	if (td->runstate == TD_RUNNING \|\| td->runstate == TD_VERIFYING
b0f65863	383	\|\| td->runstate == TD_FSYNCING
3d434057 JA	384	\|\| td->runstate == TD_PRE_READING
3d434057 JA	385	\|\| td->runstate == TD_FINISHING) {
b75a394f	386	je->nr_running++;
242c38db	387	if (td_read(td)) {
c2e9cc4d JA	388	je->t_rate[0] += td->o.rate[DDIR_READ];
	389	je->t_iops[0] += td->o.rate_iops[DDIR_READ];
	390	je->m_rate[0] += td->o.ratemin[DDIR_READ];
	391	je->m_iops[0] += td->o.rate_iops_min[DDIR_READ];
242c38db JA	392	}
242c38db JA	393	if (td_write(td)) {
c2e9cc4d JA	394	je->t_rate[1] += td->o.rate[DDIR_WRITE];
	395	je->t_iops[1] += td->o.rate_iops[DDIR_WRITE];
	396	je->m_rate[1] += td->o.ratemin[DDIR_WRITE];
	397	je->m_iops[1] += td->o.rate_iops_min[DDIR_WRITE];
242c38db	398	}
6eaf09d6	399	if (td_trim(td)) {
d79db122 JA	400	je->t_rate[2] += td->o.rate[DDIR_TRIM];
	401	je->t_iops[2] += td->o.rate_iops[DDIR_TRIM];
	402	je->m_rate[2] += td->o.ratemin[DDIR_TRIM];
	403	je->m_iops[2] += td->o.rate_iops_min[DDIR_TRIM];
6eaf09d6 SL	404	}
6eaf09d6 SL	405
b75a394f	406	je->files_open += td->nr_open_files;
b29ee5b3	407	} else if (td->runstate == TD_RAMP) {
b75a394f JA	408	je->nr_running++;
b75a394f JA	409	je->nr_ramp++;
714e85f3	410	} else if (td->runstate == TD_SETTING_UP) {
9c6f6316	411	je->nr_running++;
714e85f3 JA	412	je->nr_setting_up++;
714e85f3 JA	413	} else if (td->runstate < TD_RUNNING)
b75a394f	414	je->nr_pending++;
263e529f	415
b75a394f	416	if (je->elapsed_sec >= 3)
b29ee5b3	417	eta_secs[i] = thread_eta(td);
263e529f JA	418	else
	419	eta_secs[i] = INT_MAX;
	420
	421	check_str_update(td);
b29ee5b3	422
714e85f3	423	if (td->runstate > TD_SETTING_UP) {
6eaf09d6	424	int ddir;
771e58be	425
6eaf09d6	426	for (ddir = DDIR_READ; ddir < DDIR_RWDIR_CNT; ddir++) {
771e58be JA	427	if (unified_rw_rep) {
	428	io_bytes[0] += td->io_bytes[ddir];
	429	io_iops[0] += td->io_blocks[ddir];
	430	} else {
	431	io_bytes[ddir] += td->io_bytes[ddir];
	432	io_iops[ddir] += td->io_blocks[ddir];
	433	}
6eaf09d6	434	}
b29ee5b3	435	}
263e529f JA	436	}
	437
	438	if (exitall_on_terminate)
b75a394f	439	je->eta_sec = INT_MAX;
263e529f	440	else
b75a394f	441	je->eta_sec = 0;
263e529f	442
34572e28	443	for_each_td(td, i) {
9c5a3854	444	if (exitall_on_terminate) {
b75a394f JA	445	if (eta_secs[i] < je->eta_sec)
b75a394f JA	446	je->eta_sec = eta_secs[i];
9c5a3854	447	} else {
b75a394f JA	448	if (eta_secs[i] > je->eta_sec)
b75a394f JA	449	je->eta_sec = eta_secs[i];
9c5a3854	450	}
263e529f JA	451	}
263e529f JA	452
eecf272f JA	453	free(eta_secs);
eecf272f JA	454
46fda8d0 JA	455	fio_gettime(&now, NULL);
	456	rate_time = mtime_since(&rate_prev_time, &now);
	457
b29ee5b3	458	if (write_bw_log && rate_time > bw_avg_time && !in_ramp_time(td)) {
771e58be JA	459	calc_rate(unified_rw_rep, rate_time, io_bytes, rate_io_bytes,
771e58be JA	460	je->rate);
46fda8d0	461	memcpy(&rate_prev_time, &now, sizeof(now));
b75a394f JA	462	add_agg_sample(je->rate[DDIR_READ], DDIR_READ, 0);
b75a394f JA	463	add_agg_sample(je->rate[DDIR_WRITE], DDIR_WRITE, 0);
6eaf09d6	464	add_agg_sample(je->rate[DDIR_TRIM], DDIR_TRIM, 0);
6043c579 JA	465	}
6043c579 JA	466
46fda8d0	467	disp_time = mtime_since(&disp_prev_time, &now);
a5b01f1b JA	468
	469	/*
	470	* Allow a little slack, the target is to print it every 1000 msecs
	471	*/
af9c9fb3	472	if (!force && disp_time < 900)
b75a394f	473	return 0;
46fda8d0	474
771e58be JA	475	calc_rate(unified_rw_rep, disp_time, io_bytes, disp_io_bytes, je->rate);
771e58be JA	476	calc_iops(unified_rw_rep, disp_time, io_iops, disp_io_iops, je->iops);
adb02ba8	477
46fda8d0 JA	478	memcpy(&disp_prev_time, &now, sizeof(now));
46fda8d0 JA	479
af9c9fb3	480	if (!force && !je->nr_running && !je->nr_pending)
b75a394f JA	481	return 0;
b75a394f JA	482
1d1f45ae	483	je->nr_threads = thread_number;
6c784104 JA	484	update_condensed_str(__run_str, run_str);
6c784104 JA	485	memcpy(je->run_str, run_str, strlen(run_str));
b75a394f JA	486	return 1;
	487	}
	488
cf451d1e	489	void display_thread_status(struct jobs_eta *je)
b75a394f	490	{
e382e661 JA	491	static struct timeval disp_eta_new_line;
e382e661 JA	492	static int eta_new_line_init, eta_new_line_pending;
b75a394f JA	493	static int linelen_last;
b75a394f JA	494	static int eta_good;
79074e7c	495	char output[REAL_MAX_JOBS + 512], *p = output;
b75a394f JA	496	char eta_str[128];
b75a394f JA	497	double perc = 0.0;
b75a394f	498
cf451d1e JA	499	if (je->eta_sec != INT_MAX && je->elapsed_sec) {
	500	perc = (double) je->elapsed_sec / (double) (je->elapsed_sec + je->eta_sec);
	501	eta_to_str(eta_str, je->eta_sec);
b75a394f JA	502	}
b75a394f JA	503
e382e661 JA	504	if (eta_new_line_pending) {
	505	eta_new_line_pending = 0;
	506	p += sprintf(p, "\n");
	507	}
	508
cf451d1e	509	p += sprintf(p, "Jobs: %d (f=%d)", je->nr_running, je->files_open);
3e47bd25	510	if (je->m_rate[0] \|\| je->m_rate[1] \|\| je->t_rate[0] \|\| je->t_rate[1]) {
581e7141 JA	511	char tr, mr;
581e7141 JA	512
22f80458 JA	513	mr = num2str(je->m_rate[0] + je->m_rate[1], 4, 0, je->is_pow2, 8);
22f80458 JA	514	tr = num2str(je->t_rate[0] + je->t_rate[1], 4, 0, je->is_pow2, 8);
37db14fe	515	p += sprintf(p, ", CR=%s/%s KB/s", tr, mr);
581e7141 JA	516	free(tr);
581e7141 JA	517	free(mr);
3e47bd25 JA	518	} else if (je->m_iops[0] \|\| je->m_iops[1] \|\| je->t_iops[0] \|\| je->t_iops[1]) {
	519	p += sprintf(p, ", CR=%d/%d IOPS",
	520	je->t_iops[0] + je->t_iops[1],
88e4beaa	521	je->m_iops[0] + je->m_iops[1]);
3e47bd25	522	}
cf451d1e	523	if (je->eta_sec != INT_MAX && je->nr_running) {
0721d11e	524	char perc_str[32];
6eaf09d6 SL	525	char *iops_str[DDIR_RWDIR_CNT];
6eaf09d6 SL	526	char *rate_str[DDIR_RWDIR_CNT];
79074e7c	527	size_t left;
adb02ba8	528	int l;
6eaf09d6	529	int ddir;
5ec10eaa	530
cf451d1e	531	if ((!je->eta_sec && !eta_good) \|\| je->nr_ramp == je->nr_running)
cac58fd5 JA	532	strcpy(perc_str, "-.-% done");
cac58fd5 JA	533	else {
714e85f3 JA	534	double mult = 100.0;
	535
	536	if (je->nr_setting_up && je->nr_running)
	537	mult *= (1.0 - (double) je->nr_setting_up / (double) je->nr_running);
	538
cac58fd5	539	eta_good = 1;
714e85f3	540	perc *= mult;
0721d11e	541	sprintf(perc_str, "%3.1f%% done", perc);
cac58fd5	542	}
0721d11e	543
6eaf09d6 SL	544	for (ddir = DDIR_READ; ddir < DDIR_RWDIR_CNT; ddir++) {
6eaf09d6 SL	545	rate_str[ddir] = num2str(je->rate[ddir], 5,
ad705bcb	546	1024, je->is_pow2, je->unit_base);
73798eb2	547	iops_str[ddir] = num2str(je->iops[ddir], 4, 1, 0, 0);
6eaf09d6	548	}
adb02ba8	549
79074e7c JA	550	left = sizeof(output) - (p - output) - 1;
79074e7c JA	551
6eaf09d6 SL	552	l = snprintf(p, left, ": [%s] [%s] [%s/%s/%s /s] [%s/%s/%s iops] [eta %s]",
	553	je->run_str, perc_str, rate_str[DDIR_READ],
	554	rate_str[DDIR_WRITE], rate_str[DDIR_TRIM],
	555	iops_str[DDIR_READ], iops_str[DDIR_WRITE],
	556	iops_str[DDIR_TRIM], eta_str);
37db14fe	557	p += l;
adb02ba8	558	if (l >= 0 && l < linelen_last)
37db14fe	559	p += sprintf(p, "%*s", linelen_last - l, "");
adb02ba8	560	linelen_last = l;
d1bd7213	561
6eaf09d6 SL	562	for (ddir = DDIR_READ; ddir < DDIR_RWDIR_CNT; ddir++) {
	563	free(rate_str[ddir]);
	564	free(iops_str[ddir]);
	565	}
263e529f	566	}
37db14fe JA	567	p += sprintf(p, "\r");
37db14fe JA	568
3e47bd25	569	printf("%s", output);
e382e661 JA	570
	571	if (!eta_new_line_init) {
	572	fio_gettime(&disp_eta_new_line, NULL);
	573	eta_new_line_init = 1;
	574	} else if (eta_new_line &&
	575	mtime_since_now(&disp_eta_new_line) > eta_new_line * 1000) {
	576	fio_gettime(&disp_eta_new_line, NULL);
	577	eta_new_line_pending = 1;
	578	}
	579
3e47bd25	580	fflush(stdout);
263e529f JA	581	}
263e529f JA	582
723297c9	583	struct jobs_eta get_jobs_eta(int force, size_t size)
cf451d1e	584	{
1d1f45ae JA	585	struct jobs_eta *je;
1d1f45ae JA	586
b814fb26	587	if (!thread_number)
723297c9	588	return NULL;
cf451d1e	589
723297c9 JA	590	size = sizeof(je) + THREAD_RUNSTR_SZ;
	591	je = malloc(*size);
	592	memset(je, 0, *size);
	593
	594	if (!calc_thread_status(je, 0)) {
	595	free(je);
	596	return NULL;
	597	}
	598
	599	return je;
	600	}
	601
	602	void print_thread_status(void)
	603	{
	604	struct jobs_eta *je;
	605	size_t size;
cf451d1e	606
723297c9 JA	607	je = get_jobs_eta(0, &size);
723297c9 JA	608	if (je)
1d1f45ae	609	display_thread_status(je);
cf451d1e	610
1d1f45ae	611	free(je);
cf451d1e	612	}
b75a394f	613
2b13e716	614	void print_status_init(int thr_number)
263e529f	615	{
3e2e48a7 JA	616	__run_str[thr_number] = 'P';
3e2e48a7 JA	617	update_condensed_str(__run_str, run_str);
263e529f	618	}