[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 <inttypes.h>
#include <assert.h>

struct thread_data;

#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"
#include "helpers.h"
#include "options.h"
#include "profile.h"
#include "time.h"
#include "lib/getopt.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];
	unsigned int kb_base;
};

/*
 * 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 */
};

/*
 * offset generator types
 */
enum {
	RW_SEQ_SEQ	= 0,
	RW_SEQ_IDENT,
};

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

#define MAX_PATTERN_SIZE 512

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 *lat_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 lat_stat[2];		/* total 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[3];
	unsigned long short_io_u[3];
	unsigned long total_submit;
	unsigned long total_complete;

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

	/*
	 * IO Error related stats
	 */
	unsigned continue_on_error;
	unsigned long total_err_count;
	int first_error;

	unsigned int kb_base;
};

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 rw_seq;
	unsigned int kb_base;
	unsigned int ddir_seq_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;
	char verify_pattern[MAX_PATTERN_SIZE];
	unsigned int verify_pattern_bytes;
	unsigned int verify_fatal;
	unsigned int verify_dump;
	unsigned int verify_async;
	unsigned long long verify_backlog;
	unsigned int verify_batch;
	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 fdatasync_blocks;
	unsigned int barrier_blocks;
	unsigned long 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 mem_align;

	unsigned int stonewall;
	unsigned int new_group;
	unsigned int numjobs;
	os_cpu_mask_t cpumask;
	unsigned int cpumask_set;
	os_cpu_mask_t verify_cpumask;
	unsigned int verify_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 fallocate;
	unsigned int zero_buffers;
	unsigned int refill_buffers;
	unsigned int time_based;
	unsigned int disable_lat;
	unsigned int disable_clat;
	unsigned int disable_slat;
	unsigned int disable_bw;
	unsigned int gtod_reduce;
	unsigned int gtod_cpu;
	unsigned int gtod_offload;
	enum fio_cs clocksource;
	unsigned int no_stall;
	unsigned int trim_percentage;
	unsigned int trim_batch;
	unsigned int trim_zero;
	unsigned long long trim_backlog;

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

	/*
	 * 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;

	/*
	 * I/O Error handling
	 */
	unsigned int continue_on_error;

	/*
	 * Benchmark profile type
	 */
	char *profile;

	/*
	 * blkio cgroup support
	 */
	char *cgroup;
	unsigned int cgroup_weight;
	unsigned int cgroup_nodelete;

	unsigned int uid;
	unsigned int gid;

	unsigned int sync_file_range;
};

#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;
	enum fio_ddir last_ddir;

	char *mmapfile;
	int mmapfd;

	void *iolog_buf;
	FILE *iolog_f;

	char *sysfs_root;

	unsigned long rand_seeds[7];

	os_random_state_t bsrange_state;
	os_random_state_t verify_state;
	os_random_state_t trim_state;

	unsigned int verify_batch;
	unsigned int trim_batch;

	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;
	pthread_mutex_t io_u_lock;
	pthread_cond_t free_cond;

	/*
	 * async verify offload
	 */
	struct flist_head verify_list;
	pthread_t *verify_threads;
	unsigned int nr_verify_threads;
	pthread_cond_t verify_cond;
	int verify_thread_exit;

	/*
	 * Rate state
	 */
	unsigned long rate_nsec_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 long fill_device_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_seq_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;
	unsigned long io_hist_len;

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

	/*
	 * For tracking/handling discards
	 */
	struct flist_head trim_list;
	unsigned long trim_entries;

	/*
	 * 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;

	unsigned int sync_file_range_nr;

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

	/*
	 * Error counts
	 */
	unsigned int total_err_count;
	int first_error;

	/*
	 * Can be overloaded by profiles
	 */
	struct prof_io_ops prof_io_ops;
	void *prof_data;
};

/*
 * 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;					\
		if (!(td)->first_error)					\
			snprintf(td->verror, sizeof(td->verror) - 1, "file:%s:%d, func=%s, error=%s", __FILE__, __LINE__, (func), (msg));		\
	} while (0)


#define td_clear_error(td)		\
	(td)->error = 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 enum fio_cs fio_clock_source;
extern int warnings_fatal;

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)

#define td_non_fatal_error(e)	((e) == EIO || (e) == EILSEQ)

static inline void update_error_count(struct thread_data *td, int err)
{
	td->total_err_count++;
	if (td->total_err_count == 1)
		td->first_error = err;
}

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;
}

/*
 * Init/option functions
 */
extern int __must_check parse_options(int, char **);
extern int fio_options_parse(struct thread_data *, char **, int);
extern void fio_keywords_init(void);
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 *);
extern void add_job_opts(const char **);
extern char *num2str(unsigned long, int, int, int);

#define FIO_GETOPT_JOB		0x89988998
#define FIO_NR_OPTIONS		(FIO_MAX_OPTS + 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;
}

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;
}

static inline int is_power_of_2(unsigned int val)
{
	return (val != 0 && ((val & (val - 1)) == 0));
}

/*
 * We currently only need to do locking if we have verifier threads
 * accessing our internal structures too
 */
static inline void td_io_u_lock(struct thread_data *td)
{
	if (td->o.verify_async)
		pthread_mutex_lock(&td->io_u_lock);
}

static inline void td_io_u_unlock(struct thread_data *td)
{
	if (td->o.verify_async)
		pthread_mutex_unlock(&td->io_u_lock);
}

static inline void td_io_u_free_notify(struct thread_data *td)
{
	if (td->o.verify_async)
		pthread_cond_signal(&td->free_cond);
}

#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 <inttypes.h>
	15	#include <assert.h>
	16
	17	struct thread_data;
	18
	19	#include "compiler/compiler.h"
	20	#include "flist.h"
	21	#include "fifo.h"
	22	#include "rbtree.h"
	23	#include "arch/arch.h"
	24	#include "os/os.h"
	25	#include "mutex.h"
	26	#include "log.h"
	27	#include "debug.h"
	28	#include "file.h"
	29	#include "io_ddir.h"
	30	#include "ioengine.h"
	31	#include "iolog.h"
	32	#include "helpers.h"
	33	#include "options.h"
	34	#include "profile.h"
	35	#include "time.h"
	36	#include "lib/getopt.h"
	37
	38	#ifdef FIO_HAVE_GUASI
	39	#include <guasi.h>
	40	#endif
	41
	42	#ifdef FIO_HAVE_SOLARISAIO
	43	#include <sys/asynch.h>
	44	#endif
	45
	46	struct group_run_stats {
	47	unsigned long long max_run[2], min_run[2];
	48	unsigned long long max_bw[2], min_bw[2];
	49	unsigned long long io_kb[2];
	50	unsigned long long agg[2];
	51	unsigned int kb_base;
	52	};
	53
	54	/*
	55	* What type of allocation to use for io buffers
	56	*/
	57	enum fio_memtype {
	58	MEM_MALLOC = 0, /* ordinary malloc */
	59	MEM_SHM, /* use shared memory segments */
	60	MEM_SHMHUGE, /* use shared memory segments with huge pages */
	61	MEM_MMAP, /* use anonynomous mmap */
	62	MEM_MMAPHUGE, /* memory mapped huge file */
	63	};
	64
	65	/*
	66	* offset generator types
	67	*/
	68	enum {
	69	RW_SEQ_SEQ = 0,
	70	RW_SEQ_IDENT,
	71	};
	72
	73	/*
	74	* How many depth levels to log
	75	*/
	76	#define FIO_IO_U_MAP_NR 7
	77	#define FIO_IO_U_LAT_U_NR 10
	78	#define FIO_IO_U_LAT_M_NR 12
	79
	80	#define MAX_PATTERN_SIZE 512
	81
	82	struct thread_stat {
	83	char *name;
	84	char *verror;
	85	int error;
	86	int groupid;
	87	pid_t pid;
	88	char *description;
	89	int members;
	90
	91	struct io_log *slat_log;
	92	struct io_log *clat_log;
	93	struct io_log *lat_log;
	94	struct io_log *bw_log;
	95
	96	/*
	97	* bandwidth and latency stats
	98	*/
	99	struct io_stat clat_stat[2]; /* completion latency */
	100	struct io_stat slat_stat[2]; /* submission latency */
	101	struct io_stat lat_stat[2]; /* total latency */
	102	struct io_stat bw_stat[2]; /* bandwidth stats */
	103
	104	unsigned long long stat_io_bytes[2];
	105	struct timeval stat_sample_time[2];
	106
	107	/*
	108	* fio system usage accounting
	109	*/
	110	struct rusage ru_start;
	111	struct rusage ru_end;
	112	unsigned long usr_time;
	113	unsigned long sys_time;
	114	unsigned long ctx;
	115	unsigned long minf, majf;
	116
	117	/*
	118	* IO depth and latency stats
	119	*/
	120	unsigned int io_u_map[FIO_IO_U_MAP_NR];
	121	unsigned int io_u_submit[FIO_IO_U_MAP_NR];
	122	unsigned int io_u_complete[FIO_IO_U_MAP_NR];
	123	unsigned int io_u_lat_u[FIO_IO_U_LAT_U_NR];
	124	unsigned int io_u_lat_m[FIO_IO_U_LAT_M_NR];
	125	unsigned long total_io_u[3];
	126	unsigned long short_io_u[3];
	127	unsigned long total_submit;
	128	unsigned long total_complete;
	129
	130	unsigned long long io_bytes[2];
	131	unsigned long long runtime[2];
	132	unsigned long total_run_time;
	133
	134	/*
	135	* IO Error related stats
	136	*/
	137	unsigned continue_on_error;
	138	unsigned long total_err_count;
	139	int first_error;
	140
	141	unsigned int kb_base;
	142	};
	143
	144	struct bssplit {
	145	unsigned int bs;
	146	unsigned char perc;
	147	};
	148
	149	struct thread_options {
	150	int pad;
	151	char *description;
	152	char *name;
	153	char *directory;
	154	char *filename;
	155	char *opendir;
	156	char *ioengine;
	157	enum td_ddir td_ddir;
	158	unsigned int rw_seq;
	159	unsigned int kb_base;
	160	unsigned int ddir_seq_nr;
	161	unsigned int iodepth;
	162	unsigned int iodepth_low;
	163	unsigned int iodepth_batch;
	164	unsigned int iodepth_batch_complete;
	165
	166	unsigned long long size;
	167	unsigned int fill_device;
	168	unsigned long long file_size_low;
	169	unsigned long long file_size_high;
	170	unsigned long long start_offset;
	171
	172	unsigned int bs[2];
	173	unsigned int ba[2];
	174	unsigned int min_bs[2];
	175	unsigned int max_bs[2];
	176	struct bssplit *bssplit[2];
	177	unsigned int bssplit_nr[2];
	178
	179	unsigned int nr_files;
	180	unsigned int open_files;
	181	enum file_lock_mode file_lock_mode;
	182	unsigned int lockfile_batch;
	183
	184	unsigned int odirect;
	185	unsigned int invalidate_cache;
	186	unsigned int create_serialize;
	187	unsigned int create_fsync;
	188	unsigned int create_on_open;
	189	unsigned int end_fsync;
	190	unsigned int pre_read;
	191	unsigned int sync_io;
	192	unsigned int verify;
	193	unsigned int do_verify;
	194	unsigned int verifysort;
	195	unsigned int verify_interval;
	196	unsigned int verify_offset;
	197	char verify_pattern[MAX_PATTERN_SIZE];
	198	unsigned int verify_pattern_bytes;
	199	unsigned int verify_fatal;
	200	unsigned int verify_dump;
	201	unsigned int verify_async;
	202	unsigned long long verify_backlog;
	203	unsigned int verify_batch;
	204	unsigned int use_thread;
	205	unsigned int unlink;
	206	unsigned int do_disk_util;
	207	unsigned int override_sync;
	208	unsigned int rand_repeatable;
	209	unsigned int write_lat_log;
	210	unsigned int write_bw_log;
	211	unsigned int norandommap;
	212	unsigned int softrandommap;
	213	unsigned int bs_unaligned;
	214	unsigned int fsync_on_close;
	215
	216	unsigned int hugepage_size;
	217	unsigned int rw_min_bs;
	218	unsigned int thinktime;
	219	unsigned int thinktime_spin;
	220	unsigned int thinktime_blocks;
	221	unsigned int fsync_blocks;
	222	unsigned int fdatasync_blocks;
	223	unsigned int barrier_blocks;
	224	unsigned long start_delay;
	225	unsigned long long timeout;
	226	unsigned long long ramp_time;
	227	unsigned int overwrite;
	228	unsigned int bw_avg_time;
	229	unsigned int loops;
	230	unsigned long long zone_size;
	231	unsigned long long zone_skip;
	232	enum fio_memtype mem_type;
	233	unsigned int mem_align;
	234
	235	unsigned int stonewall;
	236	unsigned int new_group;
	237	unsigned int numjobs;
	238	os_cpu_mask_t cpumask;
	239	unsigned int cpumask_set;
	240	os_cpu_mask_t verify_cpumask;
	241	unsigned int verify_cpumask_set;
	242	unsigned int iolog;
	243	unsigned int rwmixcycle;
	244	unsigned int rwmix[2];
	245	unsigned int nice;
	246	unsigned int file_service_type;
	247	unsigned int group_reporting;
	248	unsigned int fadvise_hint;
	249	unsigned int fallocate;
	250	unsigned int zero_buffers;
	251	unsigned int refill_buffers;
	252	unsigned int time_based;
	253	unsigned int disable_lat;
	254	unsigned int disable_clat;
	255	unsigned int disable_slat;
	256	unsigned int disable_bw;
	257	unsigned int gtod_reduce;
	258	unsigned int gtod_cpu;
	259	unsigned int gtod_offload;
	260	enum fio_cs clocksource;
	261	unsigned int no_stall;
	262	unsigned int trim_percentage;
	263	unsigned int trim_batch;
	264	unsigned int trim_zero;
	265	unsigned long long trim_backlog;
	266
	267	char *read_iolog_file;
	268	char *write_iolog_file;
	269	char *bw_log_file;
	270	char *lat_log_file;
	271	char *replay_redirect;
	272
	273	/*
	274	* Pre-run and post-run shell
	275	*/
	276	char *exec_prerun;
	277	char *exec_postrun;
	278
	279	unsigned int rate[2];
	280	unsigned int ratemin[2];
	281	unsigned int ratecycle;
	282	unsigned int rate_iops[2];
	283	unsigned int rate_iops_min[2];
	284
	285	char *ioscheduler;
	286
	287	/*
	288	* CPU "io" cycle burner
	289	*/
	290	unsigned int cpuload;
	291	unsigned int cpucycle;
	292
	293	/*
	294	* I/O Error handling
	295	*/
	296	unsigned int continue_on_error;
	297
	298	/*
	299	* Benchmark profile type
	300	*/
	301	char *profile;
	302
	303	/*
	304	* blkio cgroup support
	305	*/
	306	char *cgroup;
	307	unsigned int cgroup_weight;
	308	unsigned int cgroup_nodelete;
	309
	310	unsigned int uid;
	311	unsigned int gid;
	312
	313	unsigned int sync_file_range;
	314	};
	315
	316	#define FIO_VERROR_SIZE 128
	317
	318	/*
	319	* This describes a single thread/process executing a fio job.
	320	*/
	321	struct thread_data {
	322	struct thread_options o;
	323	char verror[FIO_VERROR_SIZE];
	324	pthread_t thread;
	325	int thread_number;
	326	int groupid;
	327	struct thread_stat ts;
	328	struct fio_file **files;
	329	unsigned int files_size;
	330	unsigned int files_index;
	331	unsigned int nr_open_files;
	332	unsigned int nr_done_files;
	333	unsigned int nr_normal_files;
	334	union {
	335	unsigned int next_file;
	336	os_random_state_t next_file_state;
	337	};
	338	int error;
	339	int done;
	340	pid_t pid;
	341	char *orig_buffer;
	342	size_t orig_buffer_size;
	343	volatile int terminate;
	344	volatile int runstate;
	345	unsigned int ioprio;
	346	unsigned int ioprio_set;
	347	unsigned int last_was_sync;
	348	enum fio_ddir last_ddir;
	349
	350	char *mmapfile;
	351	int mmapfd;
	352
	353	void *iolog_buf;
	354	FILE *iolog_f;
	355
	356	char *sysfs_root;
	357
	358	unsigned long rand_seeds[7];
	359
	360	os_random_state_t bsrange_state;
	361	os_random_state_t verify_state;
	362	os_random_state_t trim_state;
	363
	364	unsigned int verify_batch;
	365	unsigned int trim_batch;
	366
	367	int shm_id;
	368
	369	/*
	370	* IO engine hooks, contains everything needed to submit an io_u
	371	* to any of the available IO engines.
	372	*/
	373	struct ioengine_ops *io_ops;
	374
	375	/*
	376	* Current IO depth and list of free and busy io_u's.
	377	*/
	378	unsigned int cur_depth;
	379	unsigned int io_u_queued;
	380	struct flist_head io_u_freelist;
	381	struct flist_head io_u_busylist;
	382	struct flist_head io_u_requeues;
	383	pthread_mutex_t io_u_lock;
	384	pthread_cond_t free_cond;
	385
	386	/*
	387	* async verify offload
	388	*/
	389	struct flist_head verify_list;
	390	pthread_t *verify_threads;
	391	unsigned int nr_verify_threads;
	392	pthread_cond_t verify_cond;
	393	int verify_thread_exit;
	394
	395	/*
	396	* Rate state
	397	*/
	398	unsigned long rate_nsec_cycle[2];
	399	long rate_pending_usleep[2];
	400	unsigned long rate_bytes[2];
	401	unsigned long rate_blocks[2];
	402	struct timeval lastrate[2];
	403
	404	unsigned long long total_io_size;
	405	unsigned long long fill_device_size;
	406
	407	unsigned long io_issues[2];
	408	unsigned long long io_blocks[2];
	409	unsigned long long io_bytes[2];
	410	unsigned long long io_skip_bytes;
	411	unsigned long long this_io_bytes[2];
	412	unsigned long long zone_bytes;
	413	struct fio_mutex *mutex;
	414
	415	/*
	416	* State for random io, a bitmap of blocks done vs not done
	417	*/
	418	os_random_state_t random_state;
	419
	420	struct timeval start; /* start of this loop */
	421	struct timeval epoch; /* time job was started */
	422	struct timeval last_issue;
	423	struct timeval tv_cache;
	424	unsigned int tv_cache_nr;
	425	unsigned int tv_cache_mask;
	426	unsigned int ramp_time_over;
	427
	428	/*
	429	* read/write mixed workload state
	430	*/
	431	os_random_state_t rwmix_state;
	432	unsigned long rwmix_issues;
	433	enum fio_ddir rwmix_ddir;
	434	unsigned int ddir_seq_nr;
	435
	436	/*
	437	* IO history logs for verification. We use a tree for sorting,
	438	* if we are overwriting. Otherwise just use a fifo.
	439	*/
	440	struct rb_root io_hist_tree;
	441	struct flist_head io_hist_list;
	442	unsigned long io_hist_len;
	443
	444	/*
	445	* For IO replaying
	446	*/
	447	struct flist_head io_log_list;
	448
	449	/*
	450	* For tracking/handling discards
	451	*/
	452	struct flist_head trim_list;
	453	unsigned long trim_entries;
	454
	455	/*
	456	* for fileservice, how often to switch to a new file
	457	*/
	458	unsigned int file_service_nr;
	459	unsigned int file_service_left;
	460	struct fio_file *file_service_file;
	461
	462	unsigned int sync_file_range_nr;
	463
	464	/*
	465	* For generating file sizes
	466	*/
	467	os_random_state_t file_size_state;
	468
	469	/*
	470	* Error counts
	471	*/
	472	unsigned int total_err_count;
	473	int first_error;
	474
	475	/*
	476	* Can be overloaded by profiles
	477	*/
	478	struct prof_io_ops prof_io_ops;
	479	void *prof_data;
	480	};
	481
	482	/*
	483	* when should interactive ETA output be generated
	484	*/
	485	enum {
	486	FIO_ETA_AUTO,
	487	FIO_ETA_ALWAYS,
	488	FIO_ETA_NEVER,
	489	};
	490
	491	#define __td_verror(td, err, msg, func) \
	492	do { \
	493	if ((td)->error) \
	494	break; \
	495	int e = (err); \
	496	(td)->error = e; \
	497	if (!(td)->first_error) \
	498	snprintf(td->verror, sizeof(td->verror) - 1, "file:%s:%d, func=%s, error=%s", __FILE__, __LINE__, (func), (msg)); \
	499	} while (0)
	500
	501
	502	#define td_clear_error(td) \
	503	(td)->error = 0;
	504	#define td_verror(td, err, func) \
	505	__td_verror((td), (err), strerror((err)), (func))
	506	#define td_vmsg(td, err, msg, func) \
	507	__td_verror((td), (err), (msg), (func))
	508
	509	extern int exitall_on_terminate;
	510	extern int thread_number;
	511	extern int nr_process, nr_thread;
	512	extern int shm_id;
	513	extern int groupid;
	514	extern int terse_output;
	515	extern int temp_stall_ts;
	516	extern unsigned long long mlock_size;
	517	extern unsigned long page_mask, page_size;
	518	extern int read_only;
	519	extern int eta_print;
	520	extern unsigned long done_secs;
	521	extern char *job_section;
	522	extern int fio_gtod_offload;
	523	extern int fio_gtod_cpu;
	524	extern enum fio_cs fio_clock_source;
	525	extern int warnings_fatal;
	526
	527	extern struct thread_data *threads;
	528
	529	static inline void fio_ro_check(struct thread_data td, struct io_u io_u)
	530	{
	531	assert(!(io_u->ddir == DDIR_WRITE && !td_write(td)));
	532	}
	533
	534	#define BLOCKS_PER_MAP (8 * sizeof(int))
	535	#define TO_MAP_BLOCK(f, b) (b)
	536	#define RAND_MAP_IDX(f, b) (TO_MAP_BLOCK(f, b) / BLOCKS_PER_MAP)
	537	#define RAND_MAP_BIT(f, b) (TO_MAP_BLOCK(f, b) & (BLOCKS_PER_MAP - 1))
	538
	539	#define MAX_JOBS (1024)
	540
	541	#define td_non_fatal_error(e) ((e) == EIO \|\| (e) == EILSEQ)
	542
	543	static inline void update_error_count(struct thread_data *td, int err)
	544	{
	545	td->total_err_count++;
	546	if (td->total_err_count == 1)
	547	td->first_error = err;
	548	}
	549
	550	static inline int should_fsync(struct thread_data *td)
	551	{
	552	if (td->last_was_sync)
	553	return 0;
	554	if (td->o.odirect)
	555	return 0;
	556	if (td_write(td) \|\| td_rw(td) \|\| td->o.override_sync)
	557	return 1;
	558
	559	return 0;
	560	}
	561
	562	/*
	563	* Init/option functions
	564	*/
	565	extern int __must_check parse_options(int, char **);
	566	extern int fio_options_parse(struct thread_data , char *, int);
	567	extern void fio_keywords_init(void);
	568	extern int fio_cmd_option_parse(struct thread_data , const char , char *);
	569	extern void fio_fill_default_options(struct thread_data *);
	570	extern int fio_show_option_help(const char *);
	571	extern void fio_options_dup_and_init(struct option *);
	572	extern void options_mem_dupe(struct thread_data *);
	573	extern void options_mem_free(struct thread_data *);
	574	extern void td_fill_rand_seeds(struct thread_data *);
	575	extern void add_job_opts(const char **);
	576	extern char *num2str(unsigned long, int, int, int);
	577
	578	#define FIO_GETOPT_JOB 0x89988998
	579	#define FIO_NR_OPTIONS (FIO_MAX_OPTS + 128)
	580
	581	/*
	582	* ETA/status stuff
	583	*/
	584	extern void print_thread_status(void);
	585	extern void print_status_init(int);
	586
	587	/*
	588	* Thread life cycle. Once a thread has a runstate beyond TD_INITIALIZED, it
	589	* will never back again. It may cycle between running/verififying/fsyncing.
	590	* Once the thread reaches TD_EXITED, it is just waiting for the core to
	591	* reap it.
	592	*/
	593	enum {
	594	TD_NOT_CREATED = 0,
	595	TD_CREATED,
	596	TD_INITIALIZED,
	597	TD_RAMP,
	598	TD_RUNNING,
	599	TD_PRE_READING,
	600	TD_VERIFYING,
	601	TD_FSYNCING,
	602	TD_EXITED,
	603	TD_REAPED,
	604	};
	605
	606	extern void td_set_runstate(struct thread_data *, int);
	607
	608	/*
	609	* Memory helpers
	610	*/
	611	extern int __must_check fio_pin_memory(void);
	612	extern void fio_unpin_memory(void);
	613	extern int __must_check allocate_io_mem(struct thread_data *);
	614	extern void free_io_mem(struct thread_data *);
	615
	616	/*
	617	* Reset stats after ramp time completes
	618	*/
	619	extern void reset_all_stats(struct thread_data *);
	620
	621	/*
	622	* blktrace support
	623	*/
	624	#ifdef FIO_HAVE_BLKTRACE
	625	extern int is_blktrace(const char *);
	626	extern int load_blktrace(struct thread_data , const char );
	627	#endif
	628
	629	/*
	630	* Mark unused variables passed to ops functions as unused, to silence gcc
	631	*/
	632	#define fio_unused __attribute((__unused__))
	633	#define fio_init __attribute__((constructor))
	634	#define fio_exit __attribute__((destructor))
	635
	636	#define for_each_td(td, i) \
	637	for ((i) = 0, (td) = &threads[0]; (i) < (int) thread_number; (i)++, (td)++)
	638	#define for_each_file(td, f, i) \
	639	if ((td)->files_index) \
	640	for ((i) = 0, (f) = (td)->files[0]; \
	641	(i) < (td)->o.nr_files && ((f) = (td)->files[i]) != NULL; \
	642	(i)++)
	643
	644	#define fio_assert(td, cond) do { \
	645	if (!(cond)) { \
	646	int *__foo = NULL; \
	647	fprintf(stderr, "file:%s:%d, assert %s failed\n", __FILE__, __LINE__, #cond); \
	648	(td)->runstate = TD_EXITED; \
	649	(td)->error = EFAULT; \
	650	*__foo = 0; \
	651	} \
	652	} while (0)
	653
	654	static inline int fio_fill_issue_time(struct thread_data *td)
	655	{
	656	if (td->o.read_iolog_file \|\|
	657	!td->o.disable_clat \|\| !td->o.disable_slat \|\| !td->o.disable_bw)
	658	return 1;
	659
	660	return 0;
	661	}
	662
	663	static inline int __should_check_rate(struct thread_data *td,
	664	enum fio_ddir ddir)
	665	{
	666	struct thread_options *o = &td->o;
	667
	668	/*
	669	* If some rate setting was given, we need to check it
	670	*/
	671	if (o->rate[ddir] \|\| o->ratemin[ddir] \|\| o->rate_iops[ddir] \|\|
	672	o->rate_iops_min[ddir])
	673	return 1;
	674
	675	return 0;
	676	}
	677
	678	static inline int should_check_rate(struct thread_data *td,
	679	unsigned long *bytes_done)
	680	{
	681	int ret = 0;
	682
	683	if (bytes_done[0])
	684	ret \|= __should_check_rate(td, 0);
	685	if (bytes_done[1])
	686	ret \|= __should_check_rate(td, 1);
	687
	688	return ret;
	689	}
	690
	691	static inline int is_power_of_2(unsigned int val)
	692	{
	693	return (val != 0 && ((val & (val - 1)) == 0));
	694	}
	695
	696	/*
	697	* We currently only need to do locking if we have verifier threads
	698	* accessing our internal structures too
	699	*/
	700	static inline void td_io_u_lock(struct thread_data *td)
	701	{
	702	if (td->o.verify_async)
	703	pthread_mutex_lock(&td->io_u_lock);
	704	}
	705
	706	static inline void td_io_u_unlock(struct thread_data *td)
	707	{
	708	if (td->o.verify_async)
	709	pthread_mutex_unlock(&td->io_u_lock);
	710	}
	711
	712	static inline void td_io_u_free_notify(struct thread_data *td)
	713	{
	714	if (td->o.verify_async)
	715	pthread_cond_signal(&td->free_cond);
	716	}
	717
	718	#endif