[fio.git] / fio.h

#ifndef FIO_H
#define FIO_H

#include <sched.h>
#include <limits.h>
#include <pthread.h>
#include <sys/time.h>
#include <sys/resource.h>
#include <errno.h>
#include <stdlib.h>
#include <stdio.h>
#include <unistd.h>
#include <string.h>
#include <getopt.h>
#include <inttypes.h>
#include <assert.h>

#include "compiler/compiler.h"
#include "flist.h"
#include "fifo.h"
#include "rbtree.h"
#include "arch/arch.h"
#include "os/os.h"
#include "mutex.h"
#include "log.h"
#include "debug.h"
#include "file.h"
#include "io_ddir.h"
#include "ioengine.h"
#include "iolog.h"

#ifdef FIO_HAVE_GUASI
#include <guasi.h>
#endif

#ifdef FIO_HAVE_SOLARISAIO
#include <sys/asynch.h>
#endif

struct group_run_stats {
	unsigned long long max_run[2], min_run[2];
	unsigned long long max_bw[2], min_bw[2];
	unsigned long long io_kb[2];
	unsigned long long agg[2];
};

/*
 * What type of allocation to use for io buffers
 */
enum fio_memtype {
	MEM_MALLOC = 0,	/* ordinary malloc */
	MEM_SHM,	/* use shared memory segments */
	MEM_SHMHUGE,	/* use shared memory segments with huge pages */
	MEM_MMAP,	/* use anonynomous mmap */
	MEM_MMAPHUGE,	/* memory mapped huge file */
};

/*
 * How many depth levels to log
 */
#define FIO_IO_U_MAP_NR	8
#define FIO_IO_U_LAT_U_NR 10
#define FIO_IO_U_LAT_M_NR 12

struct thread_stat {
	char *name;
	char *verror;
	int error;
	int groupid;
	pid_t pid;
	char *description;
	int members;

	struct io_log *slat_log;
	struct io_log *clat_log;
	struct io_log *bw_log;

	/*
	 * bandwidth and latency stats
	 */
	struct io_stat clat_stat[2];		/* completion latency */
	struct io_stat slat_stat[2];		/* submission latency */
	struct io_stat bw_stat[2];		/* bandwidth stats */

	unsigned long long stat_io_bytes[2];
	struct timeval stat_sample_time[2];

	/*
	 * fio system usage accounting
	 */
	struct rusage ru_start;
	struct rusage ru_end;
	unsigned long usr_time;
	unsigned long sys_time;
	unsigned long ctx;
	unsigned long minf, majf;

	/*
	 * IO depth and latency stats
	 */
	unsigned int io_u_map[FIO_IO_U_MAP_NR];
	unsigned int io_u_submit[FIO_IO_U_MAP_NR];
	unsigned int io_u_complete[FIO_IO_U_MAP_NR];
	unsigned int io_u_lat_u[FIO_IO_U_LAT_U_NR];
	unsigned int io_u_lat_m[FIO_IO_U_LAT_M_NR];
	unsigned long total_io_u[2];
	unsigned long short_io_u[2];
	unsigned long total_submit;
	unsigned long total_complete;

	unsigned long long io_bytes[2];
	unsigned long runtime[2];
	unsigned long total_run_time;
};

struct bssplit {
	unsigned int bs;
	unsigned char perc;
};

struct thread_options {
	int pad;
	char *description;
	char *name;
	char *directory;
	char *filename;
	char *opendir;
	char *ioengine;
	enum td_ddir td_ddir;
	unsigned int ddir_nr;
	unsigned int iodepth;
	unsigned int iodepth_low;
	unsigned int iodepth_batch;
	unsigned int iodepth_batch_complete;

	unsigned long long size;
	unsigned int fill_device;
	unsigned long long file_size_low;
	unsigned long long file_size_high;
	unsigned long long start_offset;

	unsigned int bs[2];
	unsigned int ba[2];
	unsigned int min_bs[2];
	unsigned int max_bs[2];
	struct bssplit *bssplit[2];
	unsigned int bssplit_nr[2];

	unsigned int nr_files;
	unsigned int open_files;
	enum file_lock_mode file_lock_mode;
	unsigned int lockfile_batch;

	unsigned int odirect;
	unsigned int invalidate_cache;
	unsigned int create_serialize;
	unsigned int create_fsync;
	unsigned int create_on_open;
	unsigned int end_fsync;
	unsigned int pre_read;
	unsigned int sync_io;
	unsigned int verify;
	unsigned int do_verify;
	unsigned int verifysort;
	unsigned int verify_interval;
	unsigned int verify_offset;
	unsigned int verify_pattern;
	unsigned int verify_pattern_bytes;
	unsigned int verify_fatal;
	unsigned int use_thread;
	unsigned int unlink;
	unsigned int do_disk_util;
	unsigned int override_sync;
	unsigned int rand_repeatable;
	unsigned int write_lat_log;
	unsigned int write_bw_log;
	unsigned int norandommap;
	unsigned int softrandommap;
	unsigned int bs_unaligned;
	unsigned int fsync_on_close;

	unsigned int hugepage_size;
	unsigned int rw_min_bs;
	unsigned int thinktime;
	unsigned int thinktime_spin;
	unsigned int thinktime_blocks;
	unsigned int fsync_blocks;
	unsigned int start_delay;
	unsigned long long timeout;
	unsigned long long ramp_time;
	unsigned int overwrite;
	unsigned int bw_avg_time;
	unsigned int loops;
	unsigned long long zone_size;
	unsigned long long zone_skip;
	enum fio_memtype mem_type;

	unsigned int stonewall;
	unsigned int new_group;
	unsigned int numjobs;
	os_cpu_mask_t cpumask;
	unsigned int cpumask_set;
	unsigned int iolog;
	unsigned int rwmixcycle;
	unsigned int rwmix[2];
	unsigned int nice;
	unsigned int file_service_type;
	unsigned int group_reporting;
	unsigned int fadvise_hint;
	unsigned int zero_buffers;
	unsigned int refill_buffers;
	unsigned int time_based;
	unsigned int disable_clat;
	unsigned int disable_slat;
	unsigned int disable_bw;
	unsigned int gtod_reduce;
	unsigned int gtod_cpu;
	unsigned int gtod_offload;

	char *read_iolog_file;
	char *write_iolog_file;
	char *bw_log_file;
	char *lat_log_file;

	/*
	 * Pre-run and post-run shell
	 */
	char *exec_prerun;
	char *exec_postrun;

	unsigned int rate[2];
	unsigned int ratemin[2];
	unsigned int ratecycle;
	unsigned int rate_iops[2];
	unsigned int rate_iops_min[2];

	char *ioscheduler;

	/*
	 * CPU "io" cycle burner
	 */
	unsigned int cpuload;
	unsigned int cpucycle;
};

#define FIO_VERROR_SIZE	128

/*
 * This describes a single thread/process executing a fio job.
 */
struct thread_data {
	struct thread_options o;
	char verror[FIO_VERROR_SIZE];
	pthread_t thread;
	int thread_number;
	int groupid;
	struct thread_stat ts;
	struct fio_file **files;
	unsigned int files_size;
	unsigned int files_index;
	unsigned int nr_open_files;
	unsigned int nr_done_files;
	unsigned int nr_normal_files;
	union {
		unsigned int next_file;
		os_random_state_t next_file_state;
	};
	int error;
	int done;
	pid_t pid;
	char *orig_buffer;
	size_t orig_buffer_size;
	volatile int terminate;
	volatile int runstate;
	unsigned int ioprio;
	unsigned int ioprio_set;
	unsigned int last_was_sync;

	char *mmapfile;
	int mmapfd;

	void *iolog_buf;
	FILE *iolog_f;

	char *sysfs_root;

	unsigned long rand_seeds[6];

	os_random_state_t bsrange_state;
	os_random_state_t verify_state;

	int shm_id;

	/*
	 * IO engine hooks, contains everything needed to submit an io_u
	 * to any of the available IO engines.
	 */
	struct ioengine_ops *io_ops;

	/*
	 * Current IO depth and list of free and busy io_u's.
	 */
	unsigned int cur_depth;
	unsigned int io_u_queued;
	struct flist_head io_u_freelist;
	struct flist_head io_u_busylist;
	struct flist_head io_u_requeues;

	/*
	 * Rate state
	 */
	unsigned long rate_usec_cycle[2];
	long rate_pending_usleep[2];
	unsigned long rate_bytes[2];
	unsigned long rate_blocks[2];
	struct timeval lastrate[2];

	unsigned long long total_io_size;

	unsigned long io_issues[2];
	unsigned long long io_blocks[2];
	unsigned long long io_bytes[2];
	unsigned long long io_skip_bytes;
	unsigned long long this_io_bytes[2];
	unsigned long long zone_bytes;
	struct fio_mutex *mutex;

	/*
	 * State for random io, a bitmap of blocks done vs not done
	 */
	os_random_state_t random_state;

	struct timeval start;	/* start of this loop */
	struct timeval epoch;	/* time job was started */
	struct timeval last_issue;
	struct timeval tv_cache;
	unsigned int tv_cache_nr;
	unsigned int tv_cache_mask;
	unsigned int ramp_time_over;

	/*
	 * read/write mixed workload state
	 */
	os_random_state_t rwmix_state;
	unsigned long rwmix_issues;
	enum fio_ddir rwmix_ddir;
	unsigned int ddir_nr;

	/*
	 * IO history logs for verification. We use a tree for sorting,
	 * if we are overwriting. Otherwise just use a fifo.
	 */
	struct rb_root io_hist_tree;
	struct flist_head io_hist_list;

	/*
	 * For IO replaying
	 */
	struct flist_head io_log_list;

	/*
	 * for fileservice, how often to switch to a new file
	 */
	unsigned int file_service_nr;
	unsigned int file_service_left;
	struct fio_file *file_service_file;

	/*
	 * For generating file sizes
	 */
	os_random_state_t file_size_state;
};

/*
 * when should interactive ETA output be generated
 */
enum {
	FIO_ETA_AUTO,
	FIO_ETA_ALWAYS,
	FIO_ETA_NEVER,
};

#define __td_verror(td, err, msg, func)					\
	do {								\
		if ((td)->error)					\
			break;						\
		int e = (err);						\
		(td)->error = e;					\
		snprintf(td->verror, sizeof(td->verror) - 1, "file:%s:%d, func=%s, error=%s", __FILE__, __LINE__, (func), (msg));	\
	} while (0)


#define td_verror(td, err, func)	\
	__td_verror((td), (err), strerror((err)), (func))
#define td_vmsg(td, err, msg, func)	\
	__td_verror((td), (err), (msg), (func))

extern int exitall_on_terminate;
extern int thread_number;
extern int nr_process, nr_thread;
extern int shm_id;
extern int groupid;
extern int terse_output;
extern int temp_stall_ts;
extern unsigned long long mlock_size;
extern unsigned long page_mask, page_size;
extern int read_only;
extern int eta_print;
extern unsigned long done_secs;
extern char *job_section;
extern int fio_gtod_offload;
extern int fio_gtod_cpu;

extern struct thread_data *threads;

static inline void fio_ro_check(struct thread_data *td, struct io_u *io_u)
{
	assert(!(io_u->ddir == DDIR_WRITE && !td_write(td)));
}

#define BLOCKS_PER_MAP		(8 * sizeof(int))
#define TO_MAP_BLOCK(f, b)	(b)
#define RAND_MAP_IDX(f, b)	(TO_MAP_BLOCK(f, b) / BLOCKS_PER_MAP)
#define RAND_MAP_BIT(f, b)	(TO_MAP_BLOCK(f, b) & (BLOCKS_PER_MAP - 1))

#define MAX_JOBS	(1024)

static inline int should_fsync(struct thread_data *td)
{
	if (td->last_was_sync)
		return 0;
	if (td->o.odirect)
		return 0;
	if (td_write(td) || td_rw(td) || td->o.override_sync)
		return 1;

	return 0;
}

/*
 * Time functions
 */
extern unsigned long long utime_since(struct timeval *, struct timeval *);
extern unsigned long long utime_since_now(struct timeval *);
extern unsigned long mtime_since(struct timeval *, struct timeval *);
extern unsigned long mtime_since_now(struct timeval *);
extern unsigned long time_since_now(struct timeval *);
extern unsigned long mtime_since_genesis(void);
extern void usec_spin(unsigned int);
extern void usec_sleep(struct thread_data *, unsigned long);
extern long rate_throttle(struct thread_data *, unsigned long, unsigned long, enum fio_ddir);
extern void fill_start_time(struct timeval *);
extern void fio_gettime(struct timeval *, void *);
extern void fio_gtod_init(void);
extern void fio_gtod_update(void);
extern void set_genesis_time(void);
extern int ramp_time_over(struct thread_data *);
extern int in_ramp_time(struct thread_data *);

/*
 * Init/option functions
 */
extern int __must_check parse_options(int, char **);
extern int fio_options_parse(struct thread_data *, char **, int);
extern int fio_cmd_option_parse(struct thread_data *, const char *, char *);
extern void fio_fill_default_options(struct thread_data *);
extern int fio_show_option_help(const char *);
extern void fio_options_dup_and_init(struct option *);
extern void options_mem_dupe(struct thread_data *);
extern void options_mem_free(struct thread_data *);
extern void td_fill_rand_seeds(struct thread_data *);
#define FIO_GETOPT_JOB		0x89988998
#define FIO_NR_OPTIONS		128

/*
 * ETA/status stuff
 */
extern void print_thread_status(void);
extern void print_status_init(int);

/*
 * Thread life cycle. Once a thread has a runstate beyond TD_INITIALIZED, it
 * will never back again. It may cycle between running/verififying/fsyncing.
 * Once the thread reaches TD_EXITED, it is just waiting for the core to
 * reap it.
 */
enum {
	TD_NOT_CREATED = 0,
	TD_CREATED,
	TD_INITIALIZED,
	TD_RAMP,
	TD_RUNNING,
	TD_PRE_READING,
	TD_VERIFYING,
	TD_FSYNCING,
	TD_EXITED,
	TD_REAPED,
};

extern void td_set_runstate(struct thread_data *, int);

/*
 * Memory helpers
 */
extern int __must_check fio_pin_memory(void);
extern void fio_unpin_memory(void);
extern int __must_check allocate_io_mem(struct thread_data *);
extern void free_io_mem(struct thread_data *);

/*
 * Reset stats after ramp time completes
 */
extern void reset_all_stats(struct thread_data *);

/*
 * blktrace support
 */
#ifdef FIO_HAVE_BLKTRACE
extern int is_blktrace(const char *);
extern int load_blktrace(struct thread_data *, const char *);
#endif

/*
 * Mark unused variables passed to ops functions as unused, to silence gcc
 */
#define fio_unused	__attribute((__unused__))
#define fio_init	__attribute__((constructor))
#define fio_exit	__attribute__((destructor))

#define for_each_td(td, i)	\
	for ((i) = 0, (td) = &threads[0]; (i) < (int) thread_number; (i)++, (td)++)
#define for_each_file(td, f, i)	\
	if ((td)->files_index)						\
		for ((i) = 0, (f) = (td)->files[0];			\
	    	 (i) < (td)->o.nr_files && ((f) = (td)->files[i]) != NULL; \
		 (i)++)

#define fio_assert(td, cond)	do {	\
	if (!(cond)) {			\
		int *__foo = NULL;	\
		fprintf(stderr, "file:%s:%d, assert %s failed\n", __FILE__, __LINE__, #cond);	\
		(td)->runstate = TD_EXITED;	\
		(td)->error = EFAULT;		\
		*__foo = 0;			\
	}	\
} while (0)

static inline int fio_fill_issue_time(struct thread_data *td)
{
	if (td->o.read_iolog_file ||
	    !td->o.disable_clat || !td->o.disable_slat || !td->o.disable_bw)
		return 1;

	return 0;
}

/*
 * Cheesy number->string conversion, complete with carry rounding error.
 */
static inline char *num2str(unsigned long num, int maxlen, int base, int pow2)
{
	char postfix[] = { ' ', 'K', 'M', 'G', 'P', 'E' };
	unsigned int thousand;
	char *buf;
	int i;

	if (pow2)
		thousand = 1024;
	else
		thousand = 1000;

	buf = malloc(128);

	for (i = 0; base > 1; i++)
		base /= thousand;

	do {
		int len, carry = 0;

		len = sprintf(buf, "%'lu", num);
		if (len <= maxlen) {
			if (i >= 1) {
				buf[len] = postfix[i];
				buf[len + 1] = '\0';
			}
			return buf;
		}

		if ((num % thousand) >= (thousand / 2))
			carry = 1;

		num /= thousand;
		num += carry;
		i++;
	} while (i <= 5);

	return buf;
}

static inline int __should_check_rate(struct thread_data *td,
				      enum fio_ddir ddir)
{
	struct thread_options *o = &td->o;

	/*
	 * If some rate setting was given, we need to check it
	 */
	if (o->rate[ddir] || o->ratemin[ddir] || o->rate_iops[ddir] ||
	    o->rate_iops_min[ddir])
		return 1;

	return 0;
}

static inline int should_check_rate(struct thread_data *td,
				    unsigned long *bytes_done)
{
	int ret = 0;

	if (bytes_done[0])
		ret |= __should_check_rate(td, 0);
	if (bytes_done[1])
		ret |= __should_check_rate(td, 1);

	return ret;
}

#endif
Commit	Line	Data
	1	#ifndef FIO_H
	2	#define FIO_H
	3
	4	#include <sched.h>
	5	#include <limits.h>
	6	#include <pthread.h>
	7	#include <sys/time.h>
	8	#include <sys/resource.h>
	9	#include <errno.h>
	10	#include <stdlib.h>
	11	#include <stdio.h>
	12	#include <unistd.h>
	13	#include <string.h>
	14	#include <getopt.h>
	15	#include <inttypes.h>
	16	#include <assert.h>
	17
	18	#include "compiler/compiler.h"
	19	#include "flist.h"
	20	#include "fifo.h"
	21	#include "rbtree.h"
	22	#include "arch/arch.h"
	23	#include "os/os.h"
	24	#include "mutex.h"
	25	#include "log.h"
	26	#include "debug.h"
	27	#include "file.h"
	28	#include "io_ddir.h"
	29	#include "ioengine.h"
	30	#include "iolog.h"
	31
	32	#ifdef FIO_HAVE_GUASI
	33	#include <guasi.h>
	34	#endif
	35
	36	#ifdef FIO_HAVE_SOLARISAIO
	37	#include <sys/asynch.h>
	38	#endif
	39
	40	struct group_run_stats {
	41	unsigned long long max_run[2], min_run[2];
	42	unsigned long long max_bw[2], min_bw[2];
	43	unsigned long long io_kb[2];
	44	unsigned long long agg[2];
	45	};
	46
	47	/*
	48	* What type of allocation to use for io buffers
	49	*/
	50	enum fio_memtype {
	51	MEM_MALLOC = 0, /* ordinary malloc */
	52	MEM_SHM, /* use shared memory segments */
	53	MEM_SHMHUGE, /* use shared memory segments with huge pages */
	54	MEM_MMAP, /* use anonynomous mmap */
	55	MEM_MMAPHUGE, /* memory mapped huge file */
	56	};
	57
	58	/*
	59	* How many depth levels to log
	60	*/
	61	#define FIO_IO_U_MAP_NR 8
	62	#define FIO_IO_U_LAT_U_NR 10
	63	#define FIO_IO_U_LAT_M_NR 12
	64
	65	struct thread_stat {
	66	char *name;
	67	char *verror;
	68	int error;
	69	int groupid;
	70	pid_t pid;
	71	char *description;
	72	int members;
	73
	74	struct io_log *slat_log;
	75	struct io_log *clat_log;
	76	struct io_log *bw_log;
	77
	78	/*
	79	* bandwidth and latency stats
	80	*/
	81	struct io_stat clat_stat[2]; /* completion latency */
	82	struct io_stat slat_stat[2]; /* submission latency */
	83	struct io_stat bw_stat[2]; /* bandwidth stats */
	84
	85	unsigned long long stat_io_bytes[2];
	86	struct timeval stat_sample_time[2];
	87
	88	/*
	89	* fio system usage accounting
	90	*/
	91	struct rusage ru_start;
	92	struct rusage ru_end;
	93	unsigned long usr_time;
	94	unsigned long sys_time;
	95	unsigned long ctx;
	96	unsigned long minf, majf;
	97
	98	/*
	99	* IO depth and latency stats
	100	*/
	101	unsigned int io_u_map[FIO_IO_U_MAP_NR];
	102	unsigned int io_u_submit[FIO_IO_U_MAP_NR];
	103	unsigned int io_u_complete[FIO_IO_U_MAP_NR];
	104	unsigned int io_u_lat_u[FIO_IO_U_LAT_U_NR];
	105	unsigned int io_u_lat_m[FIO_IO_U_LAT_M_NR];
	106	unsigned long total_io_u[2];
	107	unsigned long short_io_u[2];
	108	unsigned long total_submit;
	109	unsigned long total_complete;
	110
	111	unsigned long long io_bytes[2];
	112	unsigned long runtime[2];
	113	unsigned long total_run_time;
	114	};
	115
	116	struct bssplit {
	117	unsigned int bs;
	118	unsigned char perc;
	119	};
	120
	121	struct thread_options {
	122	int pad;
	123	char *description;
	124	char *name;
	125	char *directory;
	126	char *filename;
	127	char *opendir;
	128	char *ioengine;
	129	enum td_ddir td_ddir;
	130	unsigned int ddir_nr;
	131	unsigned int iodepth;
	132	unsigned int iodepth_low;
	133	unsigned int iodepth_batch;
	134	unsigned int iodepth_batch_complete;
	135
	136	unsigned long long size;
	137	unsigned int fill_device;
	138	unsigned long long file_size_low;
	139	unsigned long long file_size_high;
	140	unsigned long long start_offset;
	141
	142	unsigned int bs[2];
	143	unsigned int ba[2];
	144	unsigned int min_bs[2];
	145	unsigned int max_bs[2];
	146	struct bssplit *bssplit[2];
	147	unsigned int bssplit_nr[2];
	148
	149	unsigned int nr_files;
	150	unsigned int open_files;
	151	enum file_lock_mode file_lock_mode;
	152	unsigned int lockfile_batch;
	153
	154	unsigned int odirect;
	155	unsigned int invalidate_cache;
	156	unsigned int create_serialize;
	157	unsigned int create_fsync;
	158	unsigned int create_on_open;
	159	unsigned int end_fsync;
	160	unsigned int pre_read;
	161	unsigned int sync_io;
	162	unsigned int verify;
	163	unsigned int do_verify;
	164	unsigned int verifysort;
	165	unsigned int verify_interval;
	166	unsigned int verify_offset;
	167	unsigned int verify_pattern;
	168	unsigned int verify_pattern_bytes;
	169	unsigned int verify_fatal;
	170	unsigned int use_thread;
	171	unsigned int unlink;
	172	unsigned int do_disk_util;
	173	unsigned int override_sync;
	174	unsigned int rand_repeatable;
	175	unsigned int write_lat_log;
	176	unsigned int write_bw_log;
	177	unsigned int norandommap;
	178	unsigned int softrandommap;
	179	unsigned int bs_unaligned;
	180	unsigned int fsync_on_close;
	181
	182	unsigned int hugepage_size;
	183	unsigned int rw_min_bs;
	184	unsigned int thinktime;
	185	unsigned int thinktime_spin;
	186	unsigned int thinktime_blocks;
	187	unsigned int fsync_blocks;
	188	unsigned int start_delay;
	189	unsigned long long timeout;
	190	unsigned long long ramp_time;
	191	unsigned int overwrite;
	192	unsigned int bw_avg_time;
	193	unsigned int loops;
	194	unsigned long long zone_size;
	195	unsigned long long zone_skip;
	196	enum fio_memtype mem_type;
	197
	198	unsigned int stonewall;
	199	unsigned int new_group;
	200	unsigned int numjobs;
	201	os_cpu_mask_t cpumask;
	202	unsigned int cpumask_set;
	203	unsigned int iolog;
	204	unsigned int rwmixcycle;
	205	unsigned int rwmix[2];
	206	unsigned int nice;
	207	unsigned int file_service_type;
	208	unsigned int group_reporting;
	209	unsigned int fadvise_hint;
	210	unsigned int zero_buffers;
	211	unsigned int refill_buffers;
	212	unsigned int time_based;
	213	unsigned int disable_clat;
	214	unsigned int disable_slat;
	215	unsigned int disable_bw;
	216	unsigned int gtod_reduce;
	217	unsigned int gtod_cpu;
	218	unsigned int gtod_offload;
	219
	220	char *read_iolog_file;
	221	char *write_iolog_file;
	222	char *bw_log_file;
	223	char *lat_log_file;
	224
	225	/*
	226	* Pre-run and post-run shell
	227	*/
	228	char *exec_prerun;
	229	char *exec_postrun;
	230
	231	unsigned int rate[2];
	232	unsigned int ratemin[2];
	233	unsigned int ratecycle;
	234	unsigned int rate_iops[2];
	235	unsigned int rate_iops_min[2];
	236
	237	char *ioscheduler;
	238
	239	/*
	240	* CPU "io" cycle burner
	241	*/
	242	unsigned int cpuload;
	243	unsigned int cpucycle;
	244	};
	245
	246	#define FIO_VERROR_SIZE 128
	247
	248	/*
	249	* This describes a single thread/process executing a fio job.
	250	*/
	251	struct thread_data {
	252	struct thread_options o;
	253	char verror[FIO_VERROR_SIZE];
	254	pthread_t thread;
	255	int thread_number;
	256	int groupid;
	257	struct thread_stat ts;
	258	struct fio_file **files;
	259	unsigned int files_size;
	260	unsigned int files_index;
	261	unsigned int nr_open_files;
	262	unsigned int nr_done_files;
	263	unsigned int nr_normal_files;
	264	union {
	265	unsigned int next_file;
	266	os_random_state_t next_file_state;
	267	};
	268	int error;
	269	int done;
	270	pid_t pid;
	271	char *orig_buffer;
	272	size_t orig_buffer_size;
	273	volatile int terminate;
	274	volatile int runstate;
	275	unsigned int ioprio;
	276	unsigned int ioprio_set;
	277	unsigned int last_was_sync;
	278
	279	char *mmapfile;
	280	int mmapfd;
	281
	282	void *iolog_buf;
	283	FILE *iolog_f;
	284
	285	char *sysfs_root;
	286
	287	unsigned long rand_seeds[6];
	288
	289	os_random_state_t bsrange_state;
	290	os_random_state_t verify_state;
	291
	292	int shm_id;
	293
	294	/*
	295	* IO engine hooks, contains everything needed to submit an io_u
	296	* to any of the available IO engines.
	297	*/
	298	struct ioengine_ops *io_ops;
	299
	300	/*
	301	* Current IO depth and list of free and busy io_u's.
	302	*/
	303	unsigned int cur_depth;
	304	unsigned int io_u_queued;
	305	struct flist_head io_u_freelist;
	306	struct flist_head io_u_busylist;
	307	struct flist_head io_u_requeues;
	308
	309	/*
	310	* Rate state
	311	*/
	312	unsigned long rate_usec_cycle[2];
	313	long rate_pending_usleep[2];
	314	unsigned long rate_bytes[2];
	315	unsigned long rate_blocks[2];
	316	struct timeval lastrate[2];
	317
	318	unsigned long long total_io_size;
	319
	320	unsigned long io_issues[2];
	321	unsigned long long io_blocks[2];
	322	unsigned long long io_bytes[2];
	323	unsigned long long io_skip_bytes;
	324	unsigned long long this_io_bytes[2];
	325	unsigned long long zone_bytes;
	326	struct fio_mutex *mutex;
	327
	328	/*
	329	* State for random io, a bitmap of blocks done vs not done
	330	*/
	331	os_random_state_t random_state;
	332
	333	struct timeval start; /* start of this loop */
	334	struct timeval epoch; /* time job was started */
	335	struct timeval last_issue;
	336	struct timeval tv_cache;
	337	unsigned int tv_cache_nr;
	338	unsigned int tv_cache_mask;
	339	unsigned int ramp_time_over;
	340
	341	/*
	342	* read/write mixed workload state
	343	*/
	344	os_random_state_t rwmix_state;
	345	unsigned long rwmix_issues;
	346	enum fio_ddir rwmix_ddir;
	347	unsigned int ddir_nr;
	348
	349	/*
	350	* IO history logs for verification. We use a tree for sorting,
	351	* if we are overwriting. Otherwise just use a fifo.
	352	*/
	353	struct rb_root io_hist_tree;
	354	struct flist_head io_hist_list;
	355
	356	/*
	357	* For IO replaying
	358	*/
	359	struct flist_head io_log_list;
	360
	361	/*
	362	* for fileservice, how often to switch to a new file
	363	*/
	364	unsigned int file_service_nr;
	365	unsigned int file_service_left;
	366	struct fio_file *file_service_file;
	367
	368	/*
	369	* For generating file sizes
	370	*/
	371	os_random_state_t file_size_state;
	372	};
	373
	374	/*
	375	* when should interactive ETA output be generated
	376	*/
	377	enum {
	378	FIO_ETA_AUTO,
	379	FIO_ETA_ALWAYS,
	380	FIO_ETA_NEVER,
	381	};
	382
	383	#define __td_verror(td, err, msg, func) \
	384	do { \
	385	if ((td)->error) \
	386	break; \
	387	int e = (err); \
	388	(td)->error = e; \
	389	snprintf(td->verror, sizeof(td->verror) - 1, "file:%s:%d, func=%s, error=%s", __FILE__, __LINE__, (func), (msg)); \
	390	} while (0)
	391
	392
	393	#define td_verror(td, err, func) \
	394	__td_verror((td), (err), strerror((err)), (func))
	395	#define td_vmsg(td, err, msg, func) \
	396	__td_verror((td), (err), (msg), (func))
	397
	398	extern int exitall_on_terminate;
	399	extern int thread_number;
	400	extern int nr_process, nr_thread;
	401	extern int shm_id;
	402	extern int groupid;
	403	extern int terse_output;
	404	extern int temp_stall_ts;
	405	extern unsigned long long mlock_size;
	406	extern unsigned long page_mask, page_size;
	407	extern int read_only;
	408	extern int eta_print;
	409	extern unsigned long done_secs;
	410	extern char *job_section;
	411	extern int fio_gtod_offload;
	412	extern int fio_gtod_cpu;
	413
	414	extern struct thread_data *threads;
	415
	416	static inline void fio_ro_check(struct thread_data td, struct io_u io_u)
	417	{
	418	assert(!(io_u->ddir == DDIR_WRITE && !td_write(td)));
	419	}
	420
	421	#define BLOCKS_PER_MAP (8 * sizeof(int))
	422	#define TO_MAP_BLOCK(f, b) (b)
	423	#define RAND_MAP_IDX(f, b) (TO_MAP_BLOCK(f, b) / BLOCKS_PER_MAP)
	424	#define RAND_MAP_BIT(f, b) (TO_MAP_BLOCK(f, b) & (BLOCKS_PER_MAP - 1))
	425
	426	#define MAX_JOBS (1024)
	427
	428	static inline int should_fsync(struct thread_data *td)
	429	{
	430	if (td->last_was_sync)
	431	return 0;
	432	if (td->o.odirect)
	433	return 0;
	434	if (td_write(td) \|\| td_rw(td) \|\| td->o.override_sync)
	435	return 1;
	436
	437	return 0;
	438	}
	439
	440	/*
	441	* Time functions
	442	*/
	443	extern unsigned long long utime_since(struct timeval , struct timeval );
	444	extern unsigned long long utime_since_now(struct timeval *);
	445	extern unsigned long mtime_since(struct timeval , struct timeval );
	446	extern unsigned long mtime_since_now(struct timeval *);
	447	extern unsigned long time_since_now(struct timeval *);
	448	extern unsigned long mtime_since_genesis(void);
	449	extern void usec_spin(unsigned int);
	450	extern void usec_sleep(struct thread_data *, unsigned long);
	451	extern long rate_throttle(struct thread_data *, unsigned long, unsigned long, enum fio_ddir);
	452	extern void fill_start_time(struct timeval *);
	453	extern void fio_gettime(struct timeval , void );
	454	extern void fio_gtod_init(void);
	455	extern void fio_gtod_update(void);
	456	extern void set_genesis_time(void);
	457	extern int ramp_time_over(struct thread_data *);
	458	extern int in_ramp_time(struct thread_data *);
	459
	460	/*
	461	* Init/option functions
	462	*/
	463	extern int __must_check parse_options(int, char **);
	464	extern int fio_options_parse(struct thread_data , char *, int);
	465	extern int fio_cmd_option_parse(struct thread_data , const char , char *);
	466	extern void fio_fill_default_options(struct thread_data *);
	467	extern int fio_show_option_help(const char *);
	468	extern void fio_options_dup_and_init(struct option *);
	469	extern void options_mem_dupe(struct thread_data *);
	470	extern void options_mem_free(struct thread_data *);
	471	extern void td_fill_rand_seeds(struct thread_data *);
	472	#define FIO_GETOPT_JOB 0x89988998
	473	#define FIO_NR_OPTIONS 128
	474
	475	/*
	476	* ETA/status stuff
	477	*/
	478	extern void print_thread_status(void);
	479	extern void print_status_init(int);
	480
	481	/*
	482	* Thread life cycle. Once a thread has a runstate beyond TD_INITIALIZED, it
	483	* will never back again. It may cycle between running/verififying/fsyncing.
	484	* Once the thread reaches TD_EXITED, it is just waiting for the core to
	485	* reap it.
	486	*/
	487	enum {
	488	TD_NOT_CREATED = 0,
	489	TD_CREATED,
	490	TD_INITIALIZED,
	491	TD_RAMP,
	492	TD_RUNNING,
	493	TD_PRE_READING,
	494	TD_VERIFYING,
	495	TD_FSYNCING,
	496	TD_EXITED,
	497	TD_REAPED,
	498	};
	499
	500	extern void td_set_runstate(struct thread_data *, int);
	501
	502	/*
	503	* Memory helpers
	504	*/
	505	extern int __must_check fio_pin_memory(void);
	506	extern void fio_unpin_memory(void);
	507	extern int __must_check allocate_io_mem(struct thread_data *);
	508	extern void free_io_mem(struct thread_data *);
	509
	510	/*
	511	* Reset stats after ramp time completes
	512	*/
	513	extern void reset_all_stats(struct thread_data *);
	514
	515	/*
	516	* blktrace support
	517	*/
	518	#ifdef FIO_HAVE_BLKTRACE
	519	extern int is_blktrace(const char *);
	520	extern int load_blktrace(struct thread_data , const char );
	521	#endif
	522
	523	/*
	524	* Mark unused variables passed to ops functions as unused, to silence gcc
	525	*/
	526	#define fio_unused __attribute((__unused__))
	527	#define fio_init __attribute__((constructor))
	528	#define fio_exit __attribute__((destructor))
	529
	530	#define for_each_td(td, i) \
	531	for ((i) = 0, (td) = &threads[0]; (i) < (int) thread_number; (i)++, (td)++)
	532	#define for_each_file(td, f, i) \
	533	if ((td)->files_index) \
	534	for ((i) = 0, (f) = (td)->files[0]; \
	535	(i) < (td)->o.nr_files && ((f) = (td)->files[i]) != NULL; \
	536	(i)++)
	537
	538	#define fio_assert(td, cond) do { \
	539	if (!(cond)) { \
	540	int *__foo = NULL; \
	541	fprintf(stderr, "file:%s:%d, assert %s failed\n", __FILE__, __LINE__, #cond); \
	542	(td)->runstate = TD_EXITED; \
	543	(td)->error = EFAULT; \
	544	*__foo = 0; \
	545	} \
	546	} while (0)
	547
	548	static inline int fio_fill_issue_time(struct thread_data *td)
	549	{
	550	if (td->o.read_iolog_file \|\|
	551	!td->o.disable_clat \|\| !td->o.disable_slat \|\| !td->o.disable_bw)
	552	return 1;
	553
	554	return 0;
	555	}
	556
	557	/*
	558	* Cheesy number->string conversion, complete with carry rounding error.
	559	*/
	560	static inline char *num2str(unsigned long num, int maxlen, int base, int pow2)
	561	{
	562	char postfix[] = { ' ', 'K', 'M', 'G', 'P', 'E' };
	563	unsigned int thousand;
	564	char *buf;
	565	int i;
	566
	567	if (pow2)
	568	thousand = 1024;
	569	else
	570	thousand = 1000;
	571
	572	buf = malloc(128);
	573
	574	for (i = 0; base > 1; i++)
	575	base /= thousand;
	576
	577	do {
	578	int len, carry = 0;
	579
	580	len = sprintf(buf, "%'lu", num);
	581	if (len <= maxlen) {
	582	if (i >= 1) {
	583	buf[len] = postfix[i];
	584	buf[len + 1] = '\0';
	585	}
	586	return buf;
	587	}
	588
	589	if ((num % thousand) >= (thousand / 2))
	590	carry = 1;
	591
	592	num /= thousand;
	593	num += carry;
	594	i++;
	595	} while (i <= 5);
	596
	597	return buf;
	598	}
	599
	600	static inline int __should_check_rate(struct thread_data *td,
	601	enum fio_ddir ddir)
	602	{
	603	struct thread_options *o = &td->o;
	604
	605	/*
	606	* If some rate setting was given, we need to check it
	607	*/
	608	if (o->rate[ddir] \|\| o->ratemin[ddir] \|\| o->rate_iops[ddir] \|\|
	609	o->rate_iops_min[ddir])
	610	return 1;
	611
	612	return 0;
	613	}
	614
	615	static inline int should_check_rate(struct thread_data *td,
	616	unsigned long *bytes_done)
	617	{
	618	int ret = 0;
	619
	620	if (bytes_done[0])
	621	ret \|= __should_check_rate(td, 0);
	622	if (bytes_done[1])
	623	ret \|= __should_check_rate(td, 1);
	624
	625	return ret;
	626	}
	627
	628	#endif