[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 latency samples for reporting percentile(s).
 *
 * 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 lat 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
#define UNIFIED_SPLIT		0
#define UNIFIED_MIXED		1
#define UNIFIED_BOTH		2

enum fio_lat {
	FIO_SLAT = 0,
	FIO_CLAT,
	FIO_LAT,

	FIO_LAT_CNT = 3,
};

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;
	uint32_t slat_percentiles;
	uint32_t pad;
	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[FIO_LAT_CNT][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;

	uint64_t io_u_plat_high_prio[DDIR_RWDIR_CNT][FIO_IO_U_PLAT_NR] __attribute__((aligned(8)));;
	uint64_t io_u_plat_low_prio[DDIR_RWDIR_CNT][FIO_IO_U_PLAT_NR];
	struct io_stat clat_high_prio_stat[DDIR_RWDIR_CNT] __attribute__((aligned(8)));
	struct io_stat clat_low_prio_stat[DDIR_RWDIR_CNT];

	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 int __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, uint8_t);
extern void add_clat_sample(struct thread_data *, enum fio_ddir, unsigned long long,
				unsigned long long, uint64_t, uint8_t);
extern void add_slat_sample(struct thread_data *, enum fio_ddir, unsigned long long,
				unsigned long long, uint64_t, uint8_t);
extern void add_agg_sample(union io_sample_data, enum fio_ddir, unsigned long long bs,
				uint8_t priority_bit);
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
a64e88da JA	1	#ifndef FIO_STAT_H
	2	#define FIO_STAT_H
	3
ec41265e	4	#include "iolog.h"
a666cab8	5	#include "lib/output_buffer.h"
2a2fdab1 L	6	#include "diskutil.h"
2a2fdab1 L	7	#include "json.h"
ec41265e	8
a64e88da	9	struct group_run_stats {
6eaf09d6 SL	10	uint64_t max_run[DDIR_RWDIR_CNT], min_run[DDIR_RWDIR_CNT];
6eaf09d6 SL	11	uint64_t max_bw[DDIR_RWDIR_CNT], min_bw[DDIR_RWDIR_CNT];
af7f87cb	12	uint64_t iobytes[DDIR_RWDIR_CNT];
6eaf09d6	13	uint64_t agg[DDIR_RWDIR_CNT];
a64e88da	14	uint32_t kb_base;
ad705bcb	15	uint32_t unit_base;
e883cb35	16	uint32_t sig_figs;
a64e88da	17	uint32_t groupid;
771e58be	18	uint32_t unified_rw_rep;
e5021b34	19	} __attribute__((packed));
a64e88da JA	20
	21	/*
	22	* How many depth levels to log
	23	*/
	24	#define FIO_IO_U_MAP_NR 7
d6bb626e	25	#define FIO_IO_U_LAT_N_NR 10
a64e88da JA	26	#define FIO_IO_U_LAT_U_NR 10
	27	#define FIO_IO_U_LAT_M_NR 12
	28
37b08652 VF	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
a64e88da	39	/*
56440e63	40	* Aggregate latency samples for reporting percentile(s).
a64e88da JA	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
37b08652	50	* accurately will be 2^(GROUP_NR + PLAT_BITS - 1) nanoseconds.
a64e88da JA	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	*
56440e63	61	* Suppose the lat varies from 0 to 999 (usec), the straightforward
a64e88da JA	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	* ... ... ... [...,...] ...
37b08652	114	* 28 33 27 [8589934592,+inf]** 64
a64e88da JA	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	*
37b08652	120	* ** If a sample's MSB is greater than 33, it will be counted as 33.
a64e88da JA	121	*/
a64e88da JA	122
66347cfa DE	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) \
fc8d6d05	133	((trim_cycles) \| ((unsigned int) (state) << BLOCK_INFO_STATE_SHIFT))
66347cfa DE	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,
1d6d3455	143	};
66347cfa	144
a64e88da JA	145	#define MAX_PATTERN_SIZE 512
a64e88da JA	146	#define FIO_JOBNAME_SIZE 128
4e59d0f3	147	#define FIO_JOBDESC_SIZE 256
a64e88da	148	#define FIO_VERROR_SIZE 128
5cb8a8cd BP	149	#define UNIFIED_SPLIT 0
	150	#define UNIFIED_MIXED 1
	151	#define UNIFIED_BOTH 2
a64e88da	152
56440e63	153	enum fio_lat {
df8781b6 VF	154	FIO_SLAT = 0,
	155	FIO_CLAT,
	156	FIO_LAT,
	157
	158	FIO_LAT_CNT = 3,
	159	};
	160
a64e88da JA	161	struct thread_stat {
	162	char name[FIO_JOBNAME_SIZE];
	163	char verror[FIO_VERROR_SIZE];
ddcc0b69	164	uint32_t error;
2f122b13	165	uint32_t thread_number;
ddcc0b69	166	uint32_t groupid;
a64e88da	167	uint32_t pid;
4e59d0f3	168	char description[FIO_JOBDESC_SIZE];
a64e88da	169	uint32_t members;
771e58be	170	uint32_t unified_rw_rep;
a64e88da JA	171
	172	/*
	173	* bandwidth and latency stats
	174	*/
1e900e2d	175	struct io_stat sync_stat __attribute__((aligned(8)));/* fsync etc stats */
6eaf09d6 SL	176	struct io_stat clat_stat[DDIR_RWDIR_CNT]; /* completion latency */
	177	struct io_stat slat_stat[DDIR_RWDIR_CNT]; /* submission latency */
	178	struct io_stat lat_stat[DDIR_RWDIR_CNT]; /* total latency */
	179	struct io_stat bw_stat[DDIR_RWDIR_CNT]; /* bandwidth stats */
	180	struct io_stat iops_stat[DDIR_RWDIR_CNT]; /* IOPS stats */
a64e88da JA	181
	182	/*
	183	* fio system usage accounting
	184	*/
	185	uint64_t usr_time;
	186	uint64_t sys_time;
	187	uint64_t ctx;
	188	uint64_t minf, majf;
	189
	190	/*
	191	* IO depth and latency stats
	192	*/
b599759b JA	193	uint32_t clat_percentiles;
b599759b JA	194	uint32_t lat_percentiles;
56440e63 VF	195	uint32_t slat_percentiles;
56440e63 VF	196	uint32_t pad;
435d195a	197	uint64_t percentile_precision;
802ad4a8	198	fio_fp64_t percentile_list[FIO_IO_U_LIST_MAX_LEN];
a64e88da	199
6cc0e5aa AL	200	uint64_t io_u_map[FIO_IO_U_MAP_NR];
	201	uint64_t io_u_submit[FIO_IO_U_MAP_NR];
	202	uint64_t io_u_complete[FIO_IO_U_MAP_NR];
	203	uint64_t io_u_lat_n[FIO_IO_U_LAT_N_NR];
	204	uint64_t io_u_lat_u[FIO_IO_U_LAT_U_NR];
	205	uint64_t io_u_lat_m[FIO_IO_U_LAT_M_NR];
df8781b6	206	uint64_t io_u_plat[FIO_LAT_CNT][DDIR_RWDIR_CNT][FIO_IO_U_PLAT_NR];
6cc0e5aa	207	uint64_t io_u_sync_plat[FIO_IO_U_PLAT_NR];
3d0ebb30	208
7f3ecee2	209	uint64_t total_io_u[DDIR_RWDIR_SYNC_CNT];
de5cdfea JA	210	uint64_t short_io_u[DDIR_RWDIR_CNT];
de5cdfea JA	211	uint64_t drop_io_u[DDIR_RWDIR_CNT];
a64e88da JA	212	uint64_t total_submit;
	213	uint64_t total_complete;
	214
6eaf09d6 SL	215	uint64_t io_bytes[DDIR_RWDIR_CNT];
6eaf09d6 SL	216	uint64_t runtime[DDIR_RWDIR_CNT];
a64e88da JA	217	uint64_t total_run_time;
	218
	219	/*
	220	* IO Error related stats
	221	*/
3d0ebb30 GG	222	union {
3d0ebb30 GG	223	uint16_t continue_on_error;
11e955e3	224	uint32_t pad2;
3d0ebb30	225	};
ddcc0b69	226	uint32_t first_error;
11e955e3	227	uint64_t total_err_count;
a64e88da	228
fd5d733f BVA	229	/* ZBD stats */
	230	uint64_t nr_zone_resets;
	231
66347cfa DE	232	uint64_t nr_block_infos;
	233	uint32_t block_infos[MAX_NR_BLOCK_INFOS];
	234
a64e88da	235	uint32_t kb_base;
ad705bcb	236	uint32_t unit_base;
3e260a46 JA	237
3e260a46 JA	238	uint32_t latency_depth;
11e955e3	239	uint32_t pad3;
3e260a46 JA	240	uint64_t latency_target;
	241	fio_fp64_t latency_percentile;
	242	uint64_t latency_window;
16e56d25	243
e883cb35 JF	244	uint32_t sig_figs;
e883cb35 JF	245
bb49c8bd VF	246	uint64_t ss_dur;
	247	uint32_t ss_state;
	248	uint32_t ss_head;
bb49c8bd	249
bb49c8bd VF	250	fio_fp64_t ss_limit;
	251	fio_fp64_t ss_slope;
	252	fio_fp64_t ss_deviation;
	253	fio_fp64_t ss_criterion;
cb84f1fa	254
4ee9c375	255	uint64_t io_u_plat_high_prio[DDIR_RWDIR_CNT][FIO_IO_U_PLAT_NR] __attribute__((aligned(8)));;
efd78265	256	uint64_t io_u_plat_low_prio[DDIR_RWDIR_CNT][FIO_IO_U_PLAT_NR];
4ee9c375	257	struct io_stat clat_high_prio_stat[DDIR_RWDIR_CNT] __attribute__((aligned(8)));
efd78265	258	struct io_stat clat_low_prio_stat[DDIR_RWDIR_CNT];
b2a432bf	259
cb84f1fa VF	260	union {
cb84f1fa VF	261	uint64_t *ss_iops_data;
11e955e3	262	uint64_t pad4;
cb84f1fa VF	263	};
	264
	265	union {
	266	uint64_t *ss_bw_data;
11e955e3	267	uint64_t pad5;
cb84f1fa	268	};
96563db9 JA	269
	270	uint64_t cachehit;
	271	uint64_t cachemiss;
e5021b34	272	} __attribute__((packed));
a64e88da	273
4c515ab4 BVA	274	#define JOBS_ETA { \
	275	uint32_t nr_running; \
	276	uint32_t nr_ramp; \
	277	\
	278	uint32_t nr_pending; \
	279	uint32_t nr_setting_up; \
	280	\
cd8920a4 JA	281	uint64_t m_rate[DDIR_RWDIR_CNT]; \
cd8920a4 JA	282	uint64_t t_rate[DDIR_RWDIR_CNT]; \
4c515ab4	283	uint64_t rate[DDIR_RWDIR_CNT]; \
cd8920a4 JA	284	uint32_t m_iops[DDIR_RWDIR_CNT]; \
	285	uint32_t t_iops[DDIR_RWDIR_CNT]; \
	286	uint32_t iops[DDIR_RWDIR_CNT]; \
	287	uint32_t pad; \
	288	uint64_t elapsed_sec; \
	289	uint64_t eta_sec; \
4c515ab4 BVA	290	uint32_t is_pow2; \
	291	uint32_t unit_base; \
	292	\
	293	uint32_t sig_figs; \
	294	\
	295	uint32_t files_open; \
	296	\
	297	/* \
	298	* Network 'copy' of run_str[] \
	299	*/ \
	300	uint32_t nr_threads; \
cd8920a4	301	uint32_t pad2; \
4c515ab4 BVA	302	uint8_t run_str[]; \
	303	}
	304
	305	struct jobs_eta JOBS_ETA;
	306	struct jobs_eta_packed JOBS_ETA __attribute__((packed));
b75a394f	307
65a4d15c KC	308	struct io_u_plat_entry {
65a4d15c KC	309	struct flist_head list;
6cc0e5aa	310	uint64_t io_u_plat[FIO_IO_U_PLAT_NR];
65a4d15c KC	311	};
65a4d15c KC	312
971caeb1	313	extern struct fio_sem *stat_sem;
e5437a07	314
c5103619	315	extern struct jobs_eta get_jobs_eta(bool force, size_t size);
723297c9	316
cef9175e JA	317	extern void stat_init(void);
	318	extern void stat_exit(void);
	319
0279b880	320	extern struct json_object * show_thread_status(struct thread_stat ts, struct group_run_stats rs, struct flist_head , struct buf_output );
a666cab8	321	extern void show_group_stats(struct group_run_stats rs, struct buf_output );
8aa89d70	322	extern bool calc_thread_status(struct jobs_eta *je, int force);
cf451d1e	323	extern void display_thread_status(struct jobs_eta *je);
83f7b64e	324	extern void __show_run_stats(void);
966f8ef9	325	extern int __show_running_run_stats(void);
b852e7cf	326	extern void show_running_run_stats(void);
06464907	327	extern void check_for_running_stats(void);
fd595830	328	extern void sum_thread_stats(struct thread_stat dst, struct thread_stat src, bool first);
37f0c1ae JA	329	extern void sum_group_stats(struct group_run_stats dst, struct group_run_stats src);
	330	extern void init_thread_stat(struct thread_stat *ts);
	331	extern void init_group_run_stat(struct group_run_stats *gs);
3e47bd25	332	extern void eta_to_str(char *str, unsigned long eta_sec);
d6bb626e	333	extern bool calc_lat(struct io_stat is, unsigned long long min, unsigned long long max, double mean, double *dev);
6cc0e5aa	334	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);
fea76a0c	335	extern void stat_calc_lat_n(struct thread_stat ts, double io_u_lat);
e5bd1347 JA	336	extern void stat_calc_lat_m(struct thread_stat ts, double io_u_lat);
e5bd1347 JA	337	extern void stat_calc_lat_u(struct thread_stat ts, double io_u_lat);
6cc0e5aa	338	extern void stat_calc_dist(uint64_t map, unsigned long total, double io_u_dist);
6bb58215	339	extern void reset_io_stats(struct thread_data *);
210dd0fc JA	340	extern void update_rusage_stat(struct thread_data *);
210dd0fc JA	341	extern void clear_rusage_stat(struct thread_data *);
2e33101f	342
d6bb626e	343	extern void add_lat_sample(struct thread_data *, enum fio_ddir, unsigned long long,
b2a432bf	344	unsigned long long, uint64_t, uint8_t);
d6bb626e	345	extern void add_clat_sample(struct thread_data *, enum fio_ddir, unsigned long long,
b2a432bf	346	unsigned long long, uint64_t, uint8_t);
56440e63	347	extern void add_slat_sample(struct thread_data *, enum fio_ddir, unsigned long long,
b2a432bf PC	348	unsigned long long, uint64_t, uint8_t);
	349	extern void add_agg_sample(union io_sample_data, enum fio_ddir, unsigned long long bs,
	350	uint8_t priority_bit);
a47591e4 JA	351	extern void add_iops_sample(struct thread_data , struct io_u ,
	352	unsigned int);
	353	extern void add_bw_sample(struct thread_data , struct io_u ,
d6bb626e	354	unsigned int, unsigned long long);
b2b3eefe	355	extern void add_sync_clat_sample(struct thread_stat *ts,
2a2fdab1	356	unsigned long long nsec);
a47591e4	357	extern int calc_log_samples(void);
cf8f852a	358
2a2fdab1 L	359	extern void print_disk_util(struct disk_util_stat , struct disk_util_agg , int terse, struct buf_output *);
	360	extern void json_array_add_disk_util(struct disk_util_stat *dus,
	361	struct disk_util_agg agg, struct json_array parent);
	362
cf8f852a	363	extern struct io_log *agg_io_log[DDIR_RWDIR_CNT];
f1867a7f	364	extern bool write_bw_log;
cf8f852a	365
74558486 JA	366	static inline bool nsec_to_usec(unsigned long long *min,
	367	unsigned long long max, double mean,
	368	double *dev)
b29ad562	369	{
d6bb626e	370	if (min > 2000 && max > 99999 && *dev > 1000.0) {
b29ad562 JA	371	*min /= 1000;
	372	*max /= 1000;
	373	*mean /= 1000.0;
	374	*dev /= 1000.0;
8aa89d70	375	return true;
b29ad562 JA	376	}
b29ad562 JA	377
8aa89d70	378	return false;
b29ad562	379	}
d6bb626e	380
74558486 JA	381	static inline bool nsec_to_msec(unsigned long long *min,
	382	unsigned long long max, double mean,
	383	double *dev)
d6bb626e VF	384	{
	385	if (min > 2000000 && max > 99999999ULL && *dev > 1000000.0) {
	386	*min /= 1000000;
	387	*max /= 1000000;
	388	*mean /= 1000000.0;
	389	*dev /= 1000000.0;
	390	return true;
	391	}
	392
	393	return false;
	394	}
74558486	395
723297c9 JA	396	/*
	397	* Worst level condensing would be 1:5, so allow enough room for that
	398	*/
	399	#define __THREAD_RUNSTR_SZ(nr) ((nr) * 5)
11155857 JA	400	#define THREAD_RUNSTR_SZ __THREAD_RUNSTR_SZ(thread_number)
11155857 JA	401
66347cfa DE	402	uint32_t io_u_block_info(struct thread_data td, struct io_u *io_u);
66347cfa DE	403
a64e88da	404	#endif