[fio.git] / stat.h

#ifndef FIO_STAT_H
#define FIO_STAT_H

#include "iolog.h"
#include "lib/output_buffer.h"
#include "diskutil.h"
#include "json.h"

struct group_run_stats {
	uint64_t max_run[DDIR_RWDIR_CNT], min_run[DDIR_RWDIR_CNT];
	uint64_t max_bw[DDIR_RWDIR_CNT], min_bw[DDIR_RWDIR_CNT];
	uint64_t iobytes[DDIR_RWDIR_CNT];
	uint64_t agg[DDIR_RWDIR_CNT];
	uint32_t kb_base;
	uint32_t unit_base;
	uint32_t sig_figs;
	uint32_t groupid;
	uint32_t unified_rw_rep;
} __attribute__((packed));

/*
 * How many depth levels to log
 */
#define FIO_IO_U_MAP_NR	7
#define FIO_IO_U_LAT_N_NR 10
#define FIO_IO_U_LAT_U_NR 10
#define FIO_IO_U_LAT_M_NR 12

/*
 * Constants for clat percentiles
 */
#define FIO_IO_U_PLAT_BITS 6
#define FIO_IO_U_PLAT_VAL (1 << FIO_IO_U_PLAT_BITS)
#define FIO_IO_U_PLAT_GROUP_NR 29
#define FIO_IO_U_PLAT_NR (FIO_IO_U_PLAT_GROUP_NR * FIO_IO_U_PLAT_VAL)
#define FIO_IO_U_LIST_MAX_LEN 20 /* The size of the default and user-specified
					list of percentiles */

/*
 * Aggregate clat samples to report percentile(s) of them.
 *
 * EXECUTIVE SUMMARY
 *
 * FIO_IO_U_PLAT_BITS determines the maximum statistical error on the
 * value of resulting percentiles. The error will be approximately
 * 1/2^(FIO_IO_U_PLAT_BITS+1) of the value.
 *
 * FIO_IO_U_PLAT_GROUP_NR and FIO_IO_U_PLAT_BITS determine the maximum
 * range being tracked for latency samples. The maximum value tracked
 * accurately will be 2^(GROUP_NR + PLAT_BITS - 1) nanoseconds.
 *
 * FIO_IO_U_PLAT_GROUP_NR and FIO_IO_U_PLAT_BITS determine the memory
 * requirement of storing those aggregate counts. The memory used will
 * be (FIO_IO_U_PLAT_GROUP_NR * 2^FIO_IO_U_PLAT_BITS) * sizeof(int)
 * bytes.
 *
 * FIO_IO_U_PLAT_NR is the total number of buckets.
 *
 * DETAILS
 *
 * Suppose the clat varies from 0 to 999 (usec), the straightforward
 * method is to keep an array of (999 + 1) buckets, in which a counter
 * keeps the count of samples which fall in the bucket, e.g.,
 * {[0],[1],...,[999]}. However this consumes a huge amount of space,
 * and can be avoided if an approximation is acceptable.
 *
 * One such method is to let the range of the bucket to be greater
 * than one. This method has low accuracy when the value is small. For
 * example, let the buckets be {[0,99],[100,199],...,[900,999]}, and
 * the represented value of each bucket be the mean of the range. Then
 * a value 0 has an round-off error of 49.5. To improve on this, we
 * use buckets with non-uniform ranges, while bounding the error of
 * each bucket within a ratio of the sample value. A simple example
 * would be when error_bound = 0.005, buckets are {
 * {[0],[1],...,[99]}, {[100,101],[102,103],...,[198,199]},..,
 * {[900,909],[910,919]...}  }. The total range is partitioned into
 * groups with different ranges, then buckets with uniform ranges. An
 * upper bound of the error is (range_of_bucket/2)/value_of_bucket
 *
 * For better efficiency, we implement this using base two. We group
 * samples by their Most Significant Bit (MSB), extract the next M bit
 * of them as an index within the group, and discard the rest of the
 * bits.
 *
 * E.g., assume a sample 'x' whose MSB is bit n (starting from bit 0),
 * and use M bit for indexing
 *
 *        | n |    M bits   | bit (n-M-1) ... bit 0 |
 *
 * Because x is at least 2^n, and bit 0 to bit (n-M-1) is at most
 * (2^(n-M) - 1), discarding bit 0 to (n-M-1) makes the round-off
 * error
 *
 *           2^(n-M)-1    2^(n-M)    1
 *      e <= --------- <= ------- = ---
 *             2^n          2^n     2^M
 *
 * Furthermore, we use "mean" of the range to represent the bucket,
 * the error e can be lowered by half to 1 / 2^(M+1). By using M bits
 * as the index, each group must contains 2^M buckets.
 *
 * E.g. Let M (FIO_IO_U_PLAT_BITS) be 6
 *      Error bound is 1/2^(6+1) = 0.0078125 (< 1%)
 *
 *	Group	MSB	#discarded	range of		#buckets
 *			error_bits	value
 *	----------------------------------------------------------------
 *	0*	0~5	0		[0,63]			64
 *	1*	6	0		[64,127]		64
 *	2	7	1		[128,255]		64
 *	3	8	2		[256,511]		64
 *	4	9	3		[512,1023]		64
 *	...	...	...		[...,...]		...
 *	28	33	27		[8589934592,+inf]**	64
 *
 *  * Special cases: when n < (M-1) or when n == (M-1), in both cases,
 *    the value cannot be rounded off. Use all bits of the sample as
 *    index.
 *
 *  ** If a sample's MSB is greater than 33, it will be counted as 33.
 */

/*
 * Trim cycle count measurements
 */
#define MAX_NR_BLOCK_INFOS	8192
#define BLOCK_INFO_STATE_SHIFT	29
#define BLOCK_INFO_TRIMS(block_info)	\
	((block_info) & ((1 << BLOCK_INFO_STATE_SHIFT) - 1))
#define BLOCK_INFO_STATE(block_info)		\
	((block_info) >> BLOCK_INFO_STATE_SHIFT)
#define BLOCK_INFO(state, trim_cycles)	\
	((trim_cycles) | ((unsigned int) (state) << BLOCK_INFO_STATE_SHIFT))
#define BLOCK_INFO_SET_STATE(block_info, state)	\
	BLOCK_INFO(state, BLOCK_INFO_TRIMS(block_info))
enum block_info_state {
	BLOCK_STATE_UNINIT,
	BLOCK_STATE_TRIMMED,
	BLOCK_STATE_WRITTEN,
	BLOCK_STATE_TRIM_FAILURE,
	BLOCK_STATE_WRITE_FAILURE,
	BLOCK_STATE_COUNT,
};

#define MAX_PATTERN_SIZE	512
#define FIO_JOBNAME_SIZE	128
#define FIO_JOBDESC_SIZE	256
#define FIO_VERROR_SIZE		128

struct thread_stat {
	char name[FIO_JOBNAME_SIZE];
	char verror[FIO_VERROR_SIZE];
	uint32_t error;
	uint32_t thread_number;
	uint32_t groupid;
	uint32_t pid;
	char description[FIO_JOBDESC_SIZE];
	uint32_t members;
	uint32_t unified_rw_rep;

	/*
	 * bandwidth and latency stats
	 */
	struct io_stat sync_stat __attribute__((aligned(8)));/* fsync etc stats */
	struct io_stat clat_stat[DDIR_RWDIR_CNT]; /* completion latency */
	struct io_stat slat_stat[DDIR_RWDIR_CNT]; /* submission latency */
	struct io_stat lat_stat[DDIR_RWDIR_CNT]; /* total latency */
	struct io_stat bw_stat[DDIR_RWDIR_CNT]; /* bandwidth stats */
	struct io_stat iops_stat[DDIR_RWDIR_CNT]; /* IOPS stats */

	/*
	 * fio system usage accounting
	 */
	uint64_t usr_time;
	uint64_t sys_time;
	uint64_t ctx;
	uint64_t minf, majf;

	/*
	 * IO depth and latency stats
	 */
	uint32_t clat_percentiles;
	uint32_t lat_percentiles;
	uint64_t percentile_precision;
	fio_fp64_t percentile_list[FIO_IO_U_LIST_MAX_LEN];

	uint64_t io_u_map[FIO_IO_U_MAP_NR];
	uint64_t io_u_submit[FIO_IO_U_MAP_NR];
	uint64_t io_u_complete[FIO_IO_U_MAP_NR];
	uint64_t io_u_lat_n[FIO_IO_U_LAT_N_NR];
	uint64_t io_u_lat_u[FIO_IO_U_LAT_U_NR];
	uint64_t io_u_lat_m[FIO_IO_U_LAT_M_NR];
	uint64_t io_u_plat[DDIR_RWDIR_CNT][FIO_IO_U_PLAT_NR];
	uint64_t io_u_sync_plat[FIO_IO_U_PLAT_NR];

	uint64_t total_io_u[DDIR_RWDIR_SYNC_CNT];
	uint64_t short_io_u[DDIR_RWDIR_CNT];
	uint64_t drop_io_u[DDIR_RWDIR_CNT];
	uint64_t total_submit;
	uint64_t total_complete;

	uint64_t io_bytes[DDIR_RWDIR_CNT];
	uint64_t runtime[DDIR_RWDIR_CNT];
	uint64_t total_run_time;

	/*
	 * IO Error related stats
	 */
	union {
		uint16_t continue_on_error;
		uint32_t pad2;
	};
	uint32_t first_error;
	uint64_t total_err_count;

	/* ZBD stats */
	uint64_t nr_zone_resets;

	uint64_t nr_block_infos;
	uint32_t block_infos[MAX_NR_BLOCK_INFOS];

	uint32_t kb_base;
	uint32_t unit_base;

	uint32_t latency_depth;
	uint32_t pad3;
	uint64_t latency_target;
	fio_fp64_t latency_percentile;
	uint64_t latency_window;

	uint32_t sig_figs;

	uint64_t ss_dur;
	uint32_t ss_state;
	uint32_t ss_head;

	fio_fp64_t ss_limit;
	fio_fp64_t ss_slope;
	fio_fp64_t ss_deviation;
	fio_fp64_t ss_criterion;

	union {
		uint64_t *ss_iops_data;
		uint64_t pad4;
	};

	union {
		uint64_t *ss_bw_data;
		uint64_t pad5;
	};

	uint64_t cachehit;
	uint64_t cachemiss;
} __attribute__((packed));

#define JOBS_ETA {							\
	uint32_t nr_running;						\
	uint32_t nr_ramp;						\
									\
	uint32_t nr_pending;						\
	uint32_t nr_setting_up;						\
									\
	uint64_t m_rate[DDIR_RWDIR_CNT];				\
	uint64_t t_rate[DDIR_RWDIR_CNT];				\
	uint64_t rate[DDIR_RWDIR_CNT];					\
	uint32_t m_iops[DDIR_RWDIR_CNT];				\
	uint32_t t_iops[DDIR_RWDIR_CNT];				\
	uint32_t iops[DDIR_RWDIR_CNT];					\
	uint32_t pad;							\
	uint64_t elapsed_sec;						\
	uint64_t eta_sec;						\
	uint32_t is_pow2;						\
	uint32_t unit_base;						\
									\
	uint32_t sig_figs;						\
									\
	uint32_t files_open;						\
									\
	/*								\
	 * Network 'copy' of run_str[]					\
	 */								\
	uint32_t nr_threads;						\
	uint32_t pad2;							\
	uint8_t run_str[];						\
}

struct jobs_eta JOBS_ETA;
struct jobs_eta_packed JOBS_ETA __attribute__((packed));

struct io_u_plat_entry {
	struct flist_head list;
	uint64_t io_u_plat[FIO_IO_U_PLAT_NR];
};

extern struct fio_sem *stat_sem;

extern struct jobs_eta *get_jobs_eta(bool force, size_t *size);

extern void stat_init(void);
extern void stat_exit(void);

extern struct json_object * show_thread_status(struct thread_stat *ts, struct group_run_stats *rs, struct flist_head *, struct buf_output *);
extern void show_group_stats(struct group_run_stats *rs, struct buf_output *);
extern bool calc_thread_status(struct jobs_eta *je, int force);
extern void display_thread_status(struct jobs_eta *je);
extern void __show_run_stats(void);
extern void __show_running_run_stats(void);
extern void show_running_run_stats(void);
extern void check_for_running_stats(void);
extern void sum_thread_stats(struct thread_stat *dst, struct thread_stat *src, bool first);
extern void sum_group_stats(struct group_run_stats *dst, struct group_run_stats *src);
extern void init_thread_stat(struct thread_stat *ts);
extern void init_group_run_stat(struct group_run_stats *gs);
extern void eta_to_str(char *str, unsigned long eta_sec);
extern bool calc_lat(struct io_stat *is, unsigned long long *min, unsigned long long *max, double *mean, double *dev);
extern unsigned int calc_clat_percentiles(uint64_t *io_u_plat, unsigned long long nr, fio_fp64_t *plist, unsigned long long **output, unsigned long long *maxv, unsigned long long *minv);
extern void stat_calc_lat_n(struct thread_stat *ts, double *io_u_lat);
extern void stat_calc_lat_m(struct thread_stat *ts, double *io_u_lat);
extern void stat_calc_lat_u(struct thread_stat *ts, double *io_u_lat);
extern void stat_calc_dist(uint64_t *map, unsigned long total, double *io_u_dist);
extern void reset_io_stats(struct thread_data *);
extern void update_rusage_stat(struct thread_data *);
extern void clear_rusage_stat(struct thread_data *);

extern void add_lat_sample(struct thread_data *, enum fio_ddir, unsigned long long,
				unsigned long long, uint64_t);
extern void add_clat_sample(struct thread_data *, enum fio_ddir, unsigned long long,
				unsigned long long, uint64_t);
extern void add_slat_sample(struct thread_data *, enum fio_ddir, unsigned long,
				unsigned long long, uint64_t);
extern void add_agg_sample(union io_sample_data, enum fio_ddir, unsigned long long);
extern void add_iops_sample(struct thread_data *, struct io_u *,
				unsigned int);
extern void add_bw_sample(struct thread_data *, struct io_u *,
				unsigned int, unsigned long long);
extern void add_sync_clat_sample(struct thread_stat *ts,
				unsigned long long nsec);
extern int calc_log_samples(void);

extern void print_disk_util(struct disk_util_stat *, struct disk_util_agg *, int terse, struct buf_output *);
extern void json_array_add_disk_util(struct disk_util_stat *dus,
				struct disk_util_agg *agg, struct json_array *parent);

extern struct io_log *agg_io_log[DDIR_RWDIR_CNT];
extern bool write_bw_log;

static inline bool nsec_to_usec(unsigned long long *min,
				unsigned long long *max, double *mean,
				double *dev)
{
	if (*min > 2000 && *max > 99999 && *dev > 1000.0) {
		*min /= 1000;
		*max /= 1000;
		*mean /= 1000.0;
		*dev /= 1000.0;
		return true;
	}

	return false;
}

static inline bool nsec_to_msec(unsigned long long *min,
				unsigned long long *max, double *mean,
				double *dev)
{
	if (*min > 2000000 && *max > 99999999ULL && *dev > 1000000.0) {
		*min /= 1000000;
		*max /= 1000000;
		*mean /= 1000000.0;
		*dev /= 1000000.0;
		return true;
	}

	return false;
}

/*
 * Worst level condensing would be 1:5, so allow enough room for that
 */
#define __THREAD_RUNSTR_SZ(nr)	((nr) * 5)
#define THREAD_RUNSTR_SZ	__THREAD_RUNSTR_SZ(thread_number)

uint32_t *io_u_block_info(struct thread_data *td, struct io_u *io_u);

#endif
Commit	Line	Data
	1	#ifndef FIO_STAT_H
	2	#define FIO_STAT_H
	3
	4	#include "iolog.h"
	5	#include "lib/output_buffer.h"
	6	#include "diskutil.h"
	7	#include "json.h"
	8
	9	struct group_run_stats {
	10	uint64_t max_run[DDIR_RWDIR_CNT], min_run[DDIR_RWDIR_CNT];
	11	uint64_t max_bw[DDIR_RWDIR_CNT], min_bw[DDIR_RWDIR_CNT];
	12	uint64_t iobytes[DDIR_RWDIR_CNT];
	13	uint64_t agg[DDIR_RWDIR_CNT];
	14	uint32_t kb_base;
	15	uint32_t unit_base;
	16	uint32_t sig_figs;
	17	uint32_t groupid;
	18	uint32_t unified_rw_rep;
	19	} __attribute__((packed));
	20
	21	/*
	22	* How many depth levels to log
	23	*/
	24	#define FIO_IO_U_MAP_NR 7
	25	#define FIO_IO_U_LAT_N_NR 10
	26	#define FIO_IO_U_LAT_U_NR 10
	27	#define FIO_IO_U_LAT_M_NR 12
	28
	29	/*
	30	* Constants for clat percentiles
	31	*/
	32	#define FIO_IO_U_PLAT_BITS 6
	33	#define FIO_IO_U_PLAT_VAL (1 << FIO_IO_U_PLAT_BITS)
	34	#define FIO_IO_U_PLAT_GROUP_NR 29
	35	#define FIO_IO_U_PLAT_NR (FIO_IO_U_PLAT_GROUP_NR * FIO_IO_U_PLAT_VAL)
	36	#define FIO_IO_U_LIST_MAX_LEN 20 /* The size of the default and user-specified
	37	list of percentiles */
	38
	39	/*
	40	* Aggregate clat samples to report percentile(s) of them.
	41	*
	42	* EXECUTIVE SUMMARY
	43	*
	44	* FIO_IO_U_PLAT_BITS determines the maximum statistical error on the
	45	* value of resulting percentiles. The error will be approximately
	46	* 1/2^(FIO_IO_U_PLAT_BITS+1) of the value.
	47	*
	48	* FIO_IO_U_PLAT_GROUP_NR and FIO_IO_U_PLAT_BITS determine the maximum
	49	* range being tracked for latency samples. The maximum value tracked
	50	* accurately will be 2^(GROUP_NR + PLAT_BITS - 1) nanoseconds.
	51	*
	52	* FIO_IO_U_PLAT_GROUP_NR and FIO_IO_U_PLAT_BITS determine the memory
	53	* requirement of storing those aggregate counts. The memory used will
	54	* be (FIO_IO_U_PLAT_GROUP_NR * 2^FIO_IO_U_PLAT_BITS) * sizeof(int)
	55	* bytes.
	56	*
	57	* FIO_IO_U_PLAT_NR is the total number of buckets.
	58	*
	59	* DETAILS
	60	*
	61	* Suppose the clat varies from 0 to 999 (usec), the straightforward
	62	* method is to keep an array of (999 + 1) buckets, in which a counter
	63	* keeps the count of samples which fall in the bucket, e.g.,
	64	* {[0],[1],...,[999]}. However this consumes a huge amount of space,
	65	* and can be avoided if an approximation is acceptable.
	66	*
	67	* One such method is to let the range of the bucket to be greater
	68	* than one. This method has low accuracy when the value is small. For
	69	* example, let the buckets be {[0,99],[100,199],...,[900,999]}, and
	70	* the represented value of each bucket be the mean of the range. Then
	71	* a value 0 has an round-off error of 49.5. To improve on this, we
	72	* use buckets with non-uniform ranges, while bounding the error of
	73	* each bucket within a ratio of the sample value. A simple example
	74	* would be when error_bound = 0.005, buckets are {
	75	* {[0],[1],...,[99]}, {[100,101],[102,103],...,[198,199]},..,
	76	* {[900,909],[910,919]...} }. The total range is partitioned into
	77	* groups with different ranges, then buckets with uniform ranges. An
	78	* upper bound of the error is (range_of_bucket/2)/value_of_bucket
	79	*
	80	* For better efficiency, we implement this using base two. We group
	81	* samples by their Most Significant Bit (MSB), extract the next M bit
	82	* of them as an index within the group, and discard the rest of the
	83	* bits.
	84	*
	85	* E.g., assume a sample 'x' whose MSB is bit n (starting from bit 0),
	86	* and use M bit for indexing
	87	*
	88	* \| n \| M bits \| bit (n-M-1) ... bit 0 \|
	89	*
	90	* Because x is at least 2^n, and bit 0 to bit (n-M-1) is at most
	91	* (2^(n-M) - 1), discarding bit 0 to (n-M-1) makes the round-off
	92	* error
	93	*
	94	* 2^(n-M)-1 2^(n-M) 1
	95	* e <= --------- <= ------- = ---
	96	* 2^n 2^n 2^M
	97	*
	98	* Furthermore, we use "mean" of the range to represent the bucket,
	99	* the error e can be lowered by half to 1 / 2^(M+1). By using M bits
	100	* as the index, each group must contains 2^M buckets.
	101	*
	102	* E.g. Let M (FIO_IO_U_PLAT_BITS) be 6
	103	* Error bound is 1/2^(6+1) = 0.0078125 (< 1%)
	104	*
	105	* Group MSB #discarded range of #buckets
	106	* error_bits value
	107	* ----------------------------------------------------------------
	108	* 0* 0~5 0 [0,63] 64
	109	* 1* 6 0 [64,127] 64
	110	* 2 7 1 [128,255] 64
	111	* 3 8 2 [256,511] 64
	112	* 4 9 3 [512,1023] 64
	113	* ... ... ... [...,...] ...
	114	* 28 33 27 [8589934592,+inf]** 64
	115	*
	116	* * Special cases: when n < (M-1) or when n == (M-1), in both cases,
	117	* the value cannot be rounded off. Use all bits of the sample as
	118	* index.
	119	*
	120	* ** If a sample's MSB is greater than 33, it will be counted as 33.
	121	*/
	122
	123	/*
	124	* Trim cycle count measurements
	125	*/
	126	#define MAX_NR_BLOCK_INFOS 8192
	127	#define BLOCK_INFO_STATE_SHIFT 29
	128	#define BLOCK_INFO_TRIMS(block_info) \
	129	((block_info) & ((1 << BLOCK_INFO_STATE_SHIFT) - 1))
	130	#define BLOCK_INFO_STATE(block_info) \
	131	((block_info) >> BLOCK_INFO_STATE_SHIFT)
	132	#define BLOCK_INFO(state, trim_cycles) \
	133	((trim_cycles) \| ((unsigned int) (state) << BLOCK_INFO_STATE_SHIFT))
	134	#define BLOCK_INFO_SET_STATE(block_info, state) \
	135	BLOCK_INFO(state, BLOCK_INFO_TRIMS(block_info))
	136	enum block_info_state {
	137	BLOCK_STATE_UNINIT,
	138	BLOCK_STATE_TRIMMED,
	139	BLOCK_STATE_WRITTEN,
	140	BLOCK_STATE_TRIM_FAILURE,
	141	BLOCK_STATE_WRITE_FAILURE,
	142	BLOCK_STATE_COUNT,
	143	};
	144
	145	#define MAX_PATTERN_SIZE 512
	146	#define FIO_JOBNAME_SIZE 128
	147	#define FIO_JOBDESC_SIZE 256
	148	#define FIO_VERROR_SIZE 128
	149
	150	struct thread_stat {
	151	char name[FIO_JOBNAME_SIZE];
	152	char verror[FIO_VERROR_SIZE];
	153	uint32_t error;
	154	uint32_t thread_number;
	155	uint32_t groupid;
	156	uint32_t pid;
	157	char description[FIO_JOBDESC_SIZE];
	158	uint32_t members;
	159	uint32_t unified_rw_rep;
	160
	161	/*
	162	* bandwidth and latency stats
	163	*/
	164	struct io_stat sync_stat __attribute__((aligned(8)));/* fsync etc stats */
	165	struct io_stat clat_stat[DDIR_RWDIR_CNT]; /* completion latency */
	166	struct io_stat slat_stat[DDIR_RWDIR_CNT]; /* submission latency */
	167	struct io_stat lat_stat[DDIR_RWDIR_CNT]; /* total latency */
	168	struct io_stat bw_stat[DDIR_RWDIR_CNT]; /* bandwidth stats */
	169	struct io_stat iops_stat[DDIR_RWDIR_CNT]; /* IOPS stats */
	170
	171	/*
	172	* fio system usage accounting
	173	*/
	174	uint64_t usr_time;
	175	uint64_t sys_time;
	176	uint64_t ctx;
	177	uint64_t minf, majf;
	178
	179	/*
	180	* IO depth and latency stats
	181	*/
	182	uint32_t clat_percentiles;
	183	uint32_t lat_percentiles;
	184	uint64_t percentile_precision;
	185	fio_fp64_t percentile_list[FIO_IO_U_LIST_MAX_LEN];
	186
	187	uint64_t io_u_map[FIO_IO_U_MAP_NR];
	188	uint64_t io_u_submit[FIO_IO_U_MAP_NR];
	189	uint64_t io_u_complete[FIO_IO_U_MAP_NR];
	190	uint64_t io_u_lat_n[FIO_IO_U_LAT_N_NR];
	191	uint64_t io_u_lat_u[FIO_IO_U_LAT_U_NR];
	192	uint64_t io_u_lat_m[FIO_IO_U_LAT_M_NR];
	193	uint64_t io_u_plat[DDIR_RWDIR_CNT][FIO_IO_U_PLAT_NR];
	194	uint64_t io_u_sync_plat[FIO_IO_U_PLAT_NR];
	195
	196	uint64_t total_io_u[DDIR_RWDIR_SYNC_CNT];
	197	uint64_t short_io_u[DDIR_RWDIR_CNT];
	198	uint64_t drop_io_u[DDIR_RWDIR_CNT];
	199	uint64_t total_submit;
	200	uint64_t total_complete;
	201
	202	uint64_t io_bytes[DDIR_RWDIR_CNT];
	203	uint64_t runtime[DDIR_RWDIR_CNT];
	204	uint64_t total_run_time;
	205
	206	/*
	207	* IO Error related stats
	208	*/
	209	union {
	210	uint16_t continue_on_error;
	211	uint32_t pad2;
	212	};
	213	uint32_t first_error;
	214	uint64_t total_err_count;
	215
	216	/* ZBD stats */
	217	uint64_t nr_zone_resets;
	218
	219	uint64_t nr_block_infos;
	220	uint32_t block_infos[MAX_NR_BLOCK_INFOS];
	221
	222	uint32_t kb_base;
	223	uint32_t unit_base;
	224
	225	uint32_t latency_depth;
	226	uint32_t pad3;
	227	uint64_t latency_target;
	228	fio_fp64_t latency_percentile;
	229	uint64_t latency_window;
	230
	231	uint32_t sig_figs;
	232
	233	uint64_t ss_dur;
	234	uint32_t ss_state;
	235	uint32_t ss_head;
	236
	237	fio_fp64_t ss_limit;
	238	fio_fp64_t ss_slope;
	239	fio_fp64_t ss_deviation;
	240	fio_fp64_t ss_criterion;
	241
	242	union {
	243	uint64_t *ss_iops_data;
	244	uint64_t pad4;
	245	};
	246
	247	union {
	248	uint64_t *ss_bw_data;
	249	uint64_t pad5;
	250	};
	251
	252	uint64_t cachehit;
	253	uint64_t cachemiss;
	254	} __attribute__((packed));
	255
	256	#define JOBS_ETA { \
	257	uint32_t nr_running; \
	258	uint32_t nr_ramp; \
	259	\
	260	uint32_t nr_pending; \
	261	uint32_t nr_setting_up; \
	262	\
	263	uint64_t m_rate[DDIR_RWDIR_CNT]; \
	264	uint64_t t_rate[DDIR_RWDIR_CNT]; \
	265	uint64_t rate[DDIR_RWDIR_CNT]; \
	266	uint32_t m_iops[DDIR_RWDIR_CNT]; \
	267	uint32_t t_iops[DDIR_RWDIR_CNT]; \
	268	uint32_t iops[DDIR_RWDIR_CNT]; \
	269	uint32_t pad; \
	270	uint64_t elapsed_sec; \
	271	uint64_t eta_sec; \
	272	uint32_t is_pow2; \
	273	uint32_t unit_base; \
	274	\
	275	uint32_t sig_figs; \
	276	\
	277	uint32_t files_open; \
	278	\
	279	/* \
	280	* Network 'copy' of run_str[] \
	281	*/ \
	282	uint32_t nr_threads; \
	283	uint32_t pad2; \
	284	uint8_t run_str[]; \
	285	}
	286
	287	struct jobs_eta JOBS_ETA;
	288	struct jobs_eta_packed JOBS_ETA __attribute__((packed));
	289
	290	struct io_u_plat_entry {
	291	struct flist_head list;
	292	uint64_t io_u_plat[FIO_IO_U_PLAT_NR];
	293	};
	294
	295	extern struct fio_sem *stat_sem;
	296
	297	extern struct jobs_eta get_jobs_eta(bool force, size_t size);
	298
	299	extern void stat_init(void);
	300	extern void stat_exit(void);
	301
	302	extern struct json_object * show_thread_status(struct thread_stat ts, struct group_run_stats rs, struct flist_head , struct buf_output );
	303	extern void show_group_stats(struct group_run_stats rs, struct buf_output );
	304	extern bool calc_thread_status(struct jobs_eta *je, int force);
	305	extern void display_thread_status(struct jobs_eta *je);
	306	extern void __show_run_stats(void);
	307	extern void __show_running_run_stats(void);
	308	extern void show_running_run_stats(void);
	309	extern void check_for_running_stats(void);
	310	extern void sum_thread_stats(struct thread_stat dst, struct thread_stat src, bool first);
	311	extern void sum_group_stats(struct group_run_stats dst, struct group_run_stats src);
	312	extern void init_thread_stat(struct thread_stat *ts);
	313	extern void init_group_run_stat(struct group_run_stats *gs);
	314	extern void eta_to_str(char *str, unsigned long eta_sec);
	315	extern bool calc_lat(struct io_stat is, unsigned long long min, unsigned long long max, double mean, double *dev);
	316	extern unsigned int calc_clat_percentiles(uint64_t io_u_plat, unsigned long long nr, fio_fp64_t plist, unsigned long long *output, unsigned long long maxv, unsigned long long *minv);
	317	extern void stat_calc_lat_n(struct thread_stat ts, double io_u_lat);
	318	extern void stat_calc_lat_m(struct thread_stat ts, double io_u_lat);
	319	extern void stat_calc_lat_u(struct thread_stat ts, double io_u_lat);
	320	extern void stat_calc_dist(uint64_t map, unsigned long total, double io_u_dist);
	321	extern void reset_io_stats(struct thread_data *);
	322	extern void update_rusage_stat(struct thread_data *);
	323	extern void clear_rusage_stat(struct thread_data *);
	324
	325	extern void add_lat_sample(struct thread_data *, enum fio_ddir, unsigned long long,
	326	unsigned long long, uint64_t);
	327	extern void add_clat_sample(struct thread_data *, enum fio_ddir, unsigned long long,
	328	unsigned long long, uint64_t);
	329	extern void add_slat_sample(struct thread_data *, enum fio_ddir, unsigned long,
	330	unsigned long long, uint64_t);
	331	extern void add_agg_sample(union io_sample_data, enum fio_ddir, unsigned long long);
	332	extern void add_iops_sample(struct thread_data , struct io_u ,
	333	unsigned int);
	334	extern void add_bw_sample(struct thread_data , struct io_u ,
	335	unsigned int, unsigned long long);
	336	extern void add_sync_clat_sample(struct thread_stat *ts,
	337	unsigned long long nsec);
	338	extern int calc_log_samples(void);
	339
	340	extern void print_disk_util(struct disk_util_stat , struct disk_util_agg , int terse, struct buf_output *);
	341	extern void json_array_add_disk_util(struct disk_util_stat *dus,
	342	struct disk_util_agg agg, struct json_array parent);
	343
	344	extern struct io_log *agg_io_log[DDIR_RWDIR_CNT];
	345	extern bool write_bw_log;
	346
	347	static inline bool nsec_to_usec(unsigned long long *min,
	348	unsigned long long max, double mean,
	349	double *dev)
	350	{
	351	if (min > 2000 && max > 99999 && *dev > 1000.0) {
	352	*min /= 1000;
	353	*max /= 1000;
	354	*mean /= 1000.0;
	355	*dev /= 1000.0;
	356	return true;
	357	}
	358
	359	return false;
	360	}
	361
	362	static inline bool nsec_to_msec(unsigned long long *min,
	363	unsigned long long max, double mean,
	364	double *dev)
	365	{
	366	if (min > 2000000 && max > 99999999ULL && *dev > 1000000.0) {
	367	*min /= 1000000;
	368	*max /= 1000000;
	369	*mean /= 1000000.0;
	370	*dev /= 1000000.0;
	371	return true;
	372	}
	373
	374	return false;
	375	}
	376
	377	/*
	378	* Worst level condensing would be 1:5, so allow enough room for that
	379	*/
	380	#define __THREAD_RUNSTR_SZ(nr) ((nr) * 5)
	381	#define THREAD_RUNSTR_SZ __THREAD_RUNSTR_SZ(thread_number)
	382
	383	uint32_t io_u_block_info(struct thread_data td, struct io_u *io_u);
	384
	385	#endif