[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>

#include "compiler/compiler.h"
#include "thread_options.h"
#include "flist.h"
#include "fifo.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 "gettime.h"
#include "lib/getopt.h"
#include "lib/rand.h"
#include "lib/rbtree.h"
#include "client.h"
#include "server.h"
#include "stat.h"
#include "flow.h"
#include "io_u_queue.h"

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

#ifdef CONFIG_LIBNUMA
#include <linux/mempolicy.h>
#include <numa.h>

/*
 * "local" is pseudo-policy
 */
#define MPOL_LOCAL MPOL_MAX
#endif

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

enum {
	TD_F_VER_BACKLOG	= 1,
	TD_F_TRIM_BACKLOG	= 2,
	TD_F_READ_IOLOG		= 4,
	TD_F_REFILL_BUFFERS	= 8,
	TD_F_SCRAMBLE_BUFFERS	= 16,
	TD_F_VER_NONE		= 32,
	TD_F_PROFILE_OPS	= 64,
};

enum {
	FIO_RAND_BS_OFF		= 0,
	FIO_RAND_VER_OFF,
	FIO_RAND_MIX_OFF,
	FIO_RAND_FILE_OFF,
	FIO_RAND_BLOCK_OFF,
	FIO_RAND_FILE_SIZE_OFF,
	FIO_RAND_TRIM_OFF,
	FIO_RAND_BUF_OFF,
	FIO_RAND_SEQ_RAND_READ_OFF,
	FIO_RAND_SEQ_RAND_WRITE_OFF,
	FIO_RAND_SEQ_RAND_TRIM_OFF,
	FIO_RAND_START_DELAY,
	FIO_RAND_NR_OFFS,
};

/*
 * This describes a single thread/process executing a fio job.
 */
struct thread_data {
	struct thread_options o;
	unsigned long flags;
	void *eo;
	char verror[FIO_VERROR_SIZE];
	pthread_t thread;
	unsigned int thread_number;
	unsigned int groupid;
	struct thread_stat ts;

	int client_type;

	struct io_log *slat_log;
	struct io_log *clat_log;
	struct io_log *lat_log;
	struct io_log *bw_log;
	struct io_log *iops_log;

	uint64_t stat_io_bytes[DDIR_RWDIR_CNT];
	struct timeval bw_sample_time;

	uint64_t stat_io_blocks[DDIR_RWDIR_CNT];
	struct timeval iops_sample_time;

	volatile int update_rusage;
	struct fio_mutex *rusage_sem;
	struct rusage ru_start;
	struct rusage ru_end;

	struct fio_file **files;
	unsigned char *file_locks;
	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;
		struct frand_state __next_file_state;
	};
	int error;
	int sig;
	int done;
	pid_t pid;
	char *orig_buffer;
	size_t orig_buffer_size;
	volatile int terminate;
	volatile int runstate;
	unsigned int last_was_sync;
	enum fio_ddir last_ddir;

	int mmapfd;

	void *iolog_buf;
	FILE *iolog_f;

	char *sysfs_root;

	unsigned long rand_seeds[FIO_RAND_NR_OFFS];

	union {
		os_random_state_t bsrange_state;
		struct frand_state __bsrange_state;
	};
	union {
		os_random_state_t verify_state;
		struct frand_state __verify_state;
	};
	union {
		os_random_state_t trim_state;
		struct frand_state __trim_state;
	};
	union {
		os_random_state_t delay_state;
		struct frand_state __delay_state;
	};

	struct frand_state buf_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;

	/*
	 * Queue depth of io_u's that fio MIGHT do
	 */
	unsigned int cur_depth;

	/*
	 * io_u's about to be committed
	 */
	unsigned int io_u_queued;

	/*
	 * io_u's submitted but not completed yet
	 */
	unsigned int io_u_in_flight;

	/*
	 * List of free and busy io_u's
	 */
	struct io_u_ring io_u_requeues;
	struct io_u_queue io_u_freelist;
	struct io_u_queue io_u_all;
	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
	 */
	uint64_t rate_bps[DDIR_RWDIR_CNT];
	long rate_pending_usleep[DDIR_RWDIR_CNT];
	unsigned long rate_bytes[DDIR_RWDIR_CNT];
	unsigned long rate_blocks[DDIR_RWDIR_CNT];
	struct timeval lastrate[DDIR_RWDIR_CNT];

	uint64_t total_io_size;
	uint64_t fill_device_size;

	unsigned long io_issues[DDIR_RWDIR_CNT];
	uint64_t io_blocks[DDIR_RWDIR_CNT];
	uint64_t this_io_blocks[DDIR_RWDIR_CNT];
	uint64_t io_bytes[DDIR_RWDIR_CNT];
	uint64_t io_skip_bytes;
	uint64_t this_io_bytes[DDIR_RWDIR_CNT];
	uint64_t zone_bytes;
	struct fio_mutex *mutex;

	/*
	 * State for random io, a bitmap of blocks done vs not done
	 */
	union {
		os_random_state_t random_state;
		struct frand_state __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;

	/*
	 * Time since last latency_window was started
	 */
	struct timeval latency_ts;
	unsigned int latency_qd;
	unsigned int latency_qd_high;
	unsigned int latency_qd_low;
	unsigned int latency_failed;
	uint64_t latency_ios;

	/*
	 * read/write mixed workload state
	 */
	union {
		os_random_state_t rwmix_state;
		struct frand_state __rwmix_state;
	};
	unsigned long rwmix_issues;
	enum fio_ddir rwmix_ddir;
	unsigned int ddir_seq_nr;

	/*
	 * rand/seq mixed workload state
	 */
	union {
		os_random_state_t seq_rand_state[DDIR_RWDIR_CNT];
		struct frand_state __seq_rand_state[DDIR_RWDIR_CNT];
	};

	/*
	 * 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;

	struct flist_head next_rand_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;

	unsigned int sync_file_range_nr;

	/*
	 * For generating file sizes
	 */
	union {
		os_random_state_t file_size_state;
		struct frand_state __file_size_state;
	};

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

	struct fio_flow *flow;

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

	void *pinned_mem;
};

/*
 * 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 {								\
		int ____e = (err);						\
		if ((td)->error)					\
			break;						\
		(td)->error = ____e;					\
		if (!(td)->first_error)					\
			snprintf(td->verror, sizeof(td->verror), "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))

#define __fio_stringify_1(x)	#x
#define __fio_stringify(x)	__fio_stringify_1(x)

extern int exitall_on_terminate;
extern unsigned int thread_number;
extern unsigned int stat_number;
extern int shm_id;
extern int groupid;
extern int output_format;
extern int append_terse_output;
extern int temp_stall_ts;
extern uintptr_t page_mask, page_size;
extern int read_only;
extern int eta_print;
extern int eta_new_line;
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 fio_clock_source_set;
extern int warnings_fatal;
extern int terse_version;
extern int is_backend;
extern int nr_clients;
extern int log_syslog;
extern int status_interval;
extern const char fio_version_string[];

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 REAL_MAX_JOBS		2048

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

	return 0;
}

/*
 * Init/option functions
 */
extern int __must_check fio_init_options(void);
extern int __must_check parse_options(int, char **);
extern int parse_jobs_ini(char *, int, int, int);
extern int parse_cmd_line(int, char **, int);
extern int fio_backend(void);
extern void reset_fio_state(void);
extern void clear_io_state(struct thread_data *);
extern int fio_options_parse(struct thread_data *, char **, int, int);
extern void fio_keywords_init(void);
extern int fio_cmd_option_parse(struct thread_data *, const char *, char *);
extern int fio_cmd_ioengine_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_set_ioengine_opts(struct option *long_options, struct thread_data *td);
extern void fio_options_dup_and_init(struct option *);
extern void fio_options_mem_dupe(struct thread_data *);
extern void options_mem_dupe(void *data, struct fio_option *options);
extern void td_fill_rand_seeds(struct thread_data *);
extern void add_job_opts(const char **, int);
extern char *num2str(unsigned long, int, int, int, int);
extern int ioengine_load(struct thread_data *);
extern int parse_dryrun(void);

extern uintptr_t page_mask;
extern uintptr_t page_size;
extern int initialize_fio(char *envp[]);

#define FIO_GETOPT_JOB		0x89000000
#define FIO_GETOPT_IOENGINE	0x98000000
#define FIO_NR_OPTIONS		(FIO_MAX_OPTS + 128)

/*
 * ETA/status stuff
 */
extern void print_thread_status(void);
extern void print_status_init(int);
extern char *fio_uint_to_kmg(unsigned int val);

/*
 * 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_SETTING_UP,
	TD_RUNNING,
	TD_PRE_READING,
	TD_VERIFYING,
	TD_FSYNCING,
	TD_EXITED,
	TD_REAPED,
};

extern void td_set_runstate(struct thread_data *, int);
#define TERMINATE_ALL		(-1)
extern void fio_terminate_threads(int);

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

/*
 * 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 *, int *);
extern int load_blktrace(struct thread_data *, const char *, int);
#endif

/*
 * Latency target helpers
 */
extern void lat_target_check(struct thread_data *);
extern void lat_target_init(struct thread_data *);

#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_set_runstate((td), 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,
				    uint64_t *bytes_done)
{
	int ret = 0;

	if (bytes_done[DDIR_READ])
		ret |= __should_check_rate(td, DDIR_READ);
	if (bytes_done[DDIR_WRITE])
		ret |= __should_check_rate(td, DDIR_WRITE);
	if (bytes_done[DDIR_TRIM])
		ret |= __should_check_rate(td, DDIR_TRIM);

	return ret;
}

static inline unsigned int td_max_bs(struct thread_data *td)
{
	unsigned int max_bs;

	max_bs = max(td->o.max_bs[DDIR_READ], td->o.max_bs[DDIR_WRITE]);
	return max(td->o.max_bs[DDIR_TRIM], max_bs);
}

static inline unsigned int td_min_bs(struct thread_data *td)
{
	unsigned int min_bs;

	min_bs = min(td->o.min_bs[DDIR_READ], td->o.min_bs[DDIR_WRITE]);
	return min(td->o.min_bs[DDIR_TRIM], min_bs);
}

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);
}

extern const char *fio_get_arch_string(int);
extern const char *fio_get_os_string(int);

#define ARRAY_SIZE(x) (sizeof((x)) / (sizeof((x)[0])))

enum {
	FIO_OUTPUT_TERSE	= 0,
	FIO_OUTPUT_JSON,
	FIO_OUTPUT_NORMAL,
};

enum {
	FIO_RAND_DIST_RANDOM	= 0,
	FIO_RAND_DIST_ZIPF,
	FIO_RAND_DIST_PARETO,
};

enum {
	FIO_RAND_GEN_TAUSWORTHE = 0,
	FIO_RAND_GEN_LFSR,
};

#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	#include "compiler/compiler.h"
	18	#include "thread_options.h"
	19	#include "flist.h"
	20	#include "fifo.h"
	21	#include "arch/arch.h"
	22	#include "os/os.h"
	23	#include "mutex.h"
	24	#include "log.h"
	25	#include "debug.h"
	26	#include "file.h"
	27	#include "io_ddir.h"
	28	#include "ioengine.h"
	29	#include "iolog.h"
	30	#include "helpers.h"
	31	#include "options.h"
	32	#include "profile.h"
	33	#include "time.h"
	34	#include "gettime.h"
	35	#include "lib/getopt.h"
	36	#include "lib/rand.h"
	37	#include "lib/rbtree.h"
	38	#include "client.h"
	39	#include "server.h"
	40	#include "stat.h"
	41	#include "flow.h"
	42	#include "io_u_queue.h"
	43
	44	#ifdef CONFIG_SOLARISAIO
	45	#include <sys/asynch.h>
	46	#endif
	47
	48	#ifdef CONFIG_LIBNUMA
	49	#include <linux/mempolicy.h>
	50	#include <numa.h>
	51
	52	/*
	53	* "local" is pseudo-policy
	54	*/
	55	#define MPOL_LOCAL MPOL_MAX
	56	#endif
	57
	58	/*
	59	* offset generator types
	60	*/
	61	enum {
	62	RW_SEQ_SEQ = 0,
	63	RW_SEQ_IDENT,
	64	};
	65
	66	enum {
	67	TD_F_VER_BACKLOG = 1,
	68	TD_F_TRIM_BACKLOG = 2,
	69	TD_F_READ_IOLOG = 4,
	70	TD_F_REFILL_BUFFERS = 8,
	71	TD_F_SCRAMBLE_BUFFERS = 16,
	72	TD_F_VER_NONE = 32,
	73	TD_F_PROFILE_OPS = 64,
	74	};
	75
	76	enum {
	77	FIO_RAND_BS_OFF = 0,
	78	FIO_RAND_VER_OFF,
	79	FIO_RAND_MIX_OFF,
	80	FIO_RAND_FILE_OFF,
	81	FIO_RAND_BLOCK_OFF,
	82	FIO_RAND_FILE_SIZE_OFF,
	83	FIO_RAND_TRIM_OFF,
	84	FIO_RAND_BUF_OFF,
	85	FIO_RAND_SEQ_RAND_READ_OFF,
	86	FIO_RAND_SEQ_RAND_WRITE_OFF,
	87	FIO_RAND_SEQ_RAND_TRIM_OFF,
	88	FIO_RAND_START_DELAY,
	89	FIO_RAND_NR_OFFS,
	90	};
	91
	92	/*
	93	* This describes a single thread/process executing a fio job.
	94	*/
	95	struct thread_data {
	96	struct thread_options o;
	97	unsigned long flags;
	98	void *eo;
	99	char verror[FIO_VERROR_SIZE];
	100	pthread_t thread;
	101	unsigned int thread_number;
	102	unsigned int groupid;
	103	struct thread_stat ts;
	104
	105	int client_type;
	106
	107	struct io_log *slat_log;
	108	struct io_log *clat_log;
	109	struct io_log *lat_log;
	110	struct io_log *bw_log;
	111	struct io_log *iops_log;
	112
	113	uint64_t stat_io_bytes[DDIR_RWDIR_CNT];
	114	struct timeval bw_sample_time;
	115
	116	uint64_t stat_io_blocks[DDIR_RWDIR_CNT];
	117	struct timeval iops_sample_time;
	118
	119	volatile int update_rusage;
	120	struct fio_mutex *rusage_sem;
	121	struct rusage ru_start;
	122	struct rusage ru_end;
	123
	124	struct fio_file **files;
	125	unsigned char *file_locks;
	126	unsigned int files_size;
	127	unsigned int files_index;
	128	unsigned int nr_open_files;
	129	unsigned int nr_done_files;
	130	unsigned int nr_normal_files;
	131	union {
	132	unsigned int next_file;
	133	os_random_state_t next_file_state;
	134	struct frand_state __next_file_state;
	135	};
	136	int error;
	137	int sig;
	138	int done;
	139	pid_t pid;
	140	char *orig_buffer;
	141	size_t orig_buffer_size;
	142	volatile int terminate;
	143	volatile int runstate;
	144	unsigned int last_was_sync;
	145	enum fio_ddir last_ddir;
	146
	147	int mmapfd;
	148
	149	void *iolog_buf;
	150	FILE *iolog_f;
	151
	152	char *sysfs_root;
	153
	154	unsigned long rand_seeds[FIO_RAND_NR_OFFS];
	155
	156	union {
	157	os_random_state_t bsrange_state;
	158	struct frand_state __bsrange_state;
	159	};
	160	union {
	161	os_random_state_t verify_state;
	162	struct frand_state __verify_state;
	163	};
	164	union {
	165	os_random_state_t trim_state;
	166	struct frand_state __trim_state;
	167	};
	168	union {
	169	os_random_state_t delay_state;
	170	struct frand_state __delay_state;
	171	};
	172
	173	struct frand_state buf_state;
	174
	175	unsigned int verify_batch;
	176	unsigned int trim_batch;
	177
	178	int shm_id;
	179
	180	/*
	181	* IO engine hooks, contains everything needed to submit an io_u
	182	* to any of the available IO engines.
	183	*/
	184	struct ioengine_ops *io_ops;
	185
	186	/*
	187	* Queue depth of io_u's that fio MIGHT do
	188	*/
	189	unsigned int cur_depth;
	190
	191	/*
	192	* io_u's about to be committed
	193	*/
	194	unsigned int io_u_queued;
	195
	196	/*
	197	* io_u's submitted but not completed yet
	198	*/
	199	unsigned int io_u_in_flight;
	200
	201	/*
	202	* List of free and busy io_u's
	203	*/
	204	struct io_u_ring io_u_requeues;
	205	struct io_u_queue io_u_freelist;
	206	struct io_u_queue io_u_all;
	207	pthread_mutex_t io_u_lock;
	208	pthread_cond_t free_cond;
	209
	210	/*
	211	* async verify offload
	212	*/
	213	struct flist_head verify_list;
	214	pthread_t *verify_threads;
	215	unsigned int nr_verify_threads;
	216	pthread_cond_t verify_cond;
	217	int verify_thread_exit;
	218
	219	/*
	220	* Rate state
	221	*/
	222	uint64_t rate_bps[DDIR_RWDIR_CNT];
	223	long rate_pending_usleep[DDIR_RWDIR_CNT];
	224	unsigned long rate_bytes[DDIR_RWDIR_CNT];
	225	unsigned long rate_blocks[DDIR_RWDIR_CNT];
	226	struct timeval lastrate[DDIR_RWDIR_CNT];
	227
	228	uint64_t total_io_size;
	229	uint64_t fill_device_size;
	230
	231	unsigned long io_issues[DDIR_RWDIR_CNT];
	232	uint64_t io_blocks[DDIR_RWDIR_CNT];
	233	uint64_t this_io_blocks[DDIR_RWDIR_CNT];
	234	uint64_t io_bytes[DDIR_RWDIR_CNT];
	235	uint64_t io_skip_bytes;
	236	uint64_t this_io_bytes[DDIR_RWDIR_CNT];
	237	uint64_t zone_bytes;
	238	struct fio_mutex *mutex;
	239
	240	/*
	241	* State for random io, a bitmap of blocks done vs not done
	242	*/
	243	union {
	244	os_random_state_t random_state;
	245	struct frand_state __random_state;
	246	};
	247
	248	struct timeval start; /* start of this loop */
	249	struct timeval epoch; /* time job was started */
	250	struct timeval last_issue;
	251	struct timeval tv_cache;
	252	unsigned int tv_cache_nr;
	253	unsigned int tv_cache_mask;
	254	unsigned int ramp_time_over;
	255
	256	/*
	257	* Time since last latency_window was started
	258	*/
	259	struct timeval latency_ts;
	260	unsigned int latency_qd;
	261	unsigned int latency_qd_high;
	262	unsigned int latency_qd_low;
	263	unsigned int latency_failed;
	264	uint64_t latency_ios;
	265
	266	/*
	267	* read/write mixed workload state
	268	*/
	269	union {
	270	os_random_state_t rwmix_state;
	271	struct frand_state __rwmix_state;
	272	};
	273	unsigned long rwmix_issues;
	274	enum fio_ddir rwmix_ddir;
	275	unsigned int ddir_seq_nr;
	276
	277	/*
	278	* rand/seq mixed workload state
	279	*/
	280	union {
	281	os_random_state_t seq_rand_state[DDIR_RWDIR_CNT];
	282	struct frand_state __seq_rand_state[DDIR_RWDIR_CNT];
	283	};
	284
	285	/*
	286	* IO history logs for verification. We use a tree for sorting,
	287	* if we are overwriting. Otherwise just use a fifo.
	288	*/
	289	struct rb_root io_hist_tree;
	290	struct flist_head io_hist_list;
	291	unsigned long io_hist_len;
	292
	293	/*
	294	* For IO replaying
	295	*/
	296	struct flist_head io_log_list;
	297
	298	/*
	299	* For tracking/handling discards
	300	*/
	301	struct flist_head trim_list;
	302	unsigned long trim_entries;
	303
	304	struct flist_head next_rand_list;
	305
	306	/*
	307	* for fileservice, how often to switch to a new file
	308	*/
	309	unsigned int file_service_nr;
	310	unsigned int file_service_left;
	311	struct fio_file *file_service_file;
	312
	313	unsigned int sync_file_range_nr;
	314
	315	/*
	316	* For generating file sizes
	317	*/
	318	union {
	319	os_random_state_t file_size_state;
	320	struct frand_state __file_size_state;
	321	};
	322
	323	/*
	324	* Error counts
	325	*/
	326	unsigned int total_err_count;
	327	int first_error;
	328
	329	struct fio_flow *flow;
	330
	331	/*
	332	* Can be overloaded by profiles
	333	*/
	334	struct prof_io_ops prof_io_ops;
	335	void *prof_data;
	336
	337	void *pinned_mem;
	338	};
	339
	340	/*
	341	* when should interactive ETA output be generated
	342	*/
	343	enum {
	344	FIO_ETA_AUTO,
	345	FIO_ETA_ALWAYS,
	346	FIO_ETA_NEVER,
	347	};
	348
	349	#define __td_verror(td, err, msg, func) \
	350	do { \
	351	int ____e = (err); \
	352	if ((td)->error) \
	353	break; \
	354	(td)->error = ____e; \
	355	if (!(td)->first_error) \
	356	snprintf(td->verror, sizeof(td->verror), "file:%s:%d, func=%s, error=%s", __FILE__, __LINE__, (func), (msg)); \
	357	} while (0)
	358
	359
	360	#define td_clear_error(td) \
	361	(td)->error = 0;
	362	#define td_verror(td, err, func) \
	363	__td_verror((td), (err), strerror((err)), (func))
	364	#define td_vmsg(td, err, msg, func) \
	365	__td_verror((td), (err), (msg), (func))
	366
	367	#define __fio_stringify_1(x) #x
	368	#define __fio_stringify(x) __fio_stringify_1(x)
	369
	370	extern int exitall_on_terminate;
	371	extern unsigned int thread_number;
	372	extern unsigned int stat_number;
	373	extern int shm_id;
	374	extern int groupid;
	375	extern int output_format;
	376	extern int append_terse_output;
	377	extern int temp_stall_ts;
	378	extern uintptr_t page_mask, page_size;
	379	extern int read_only;
	380	extern int eta_print;
	381	extern int eta_new_line;
	382	extern unsigned long done_secs;
	383	extern char *job_section;
	384	extern int fio_gtod_offload;
	385	extern int fio_gtod_cpu;
	386	extern enum fio_cs fio_clock_source;
	387	extern int fio_clock_source_set;
	388	extern int warnings_fatal;
	389	extern int terse_version;
	390	extern int is_backend;
	391	extern int nr_clients;
	392	extern int log_syslog;
	393	extern int status_interval;
	394	extern const char fio_version_string[];
	395
	396	extern struct thread_data *threads;
	397
	398	static inline void fio_ro_check(struct thread_data td, struct io_u io_u)
	399	{
	400	assert(!(io_u->ddir == DDIR_WRITE && !td_write(td)));
	401	}
	402
	403	#define REAL_MAX_JOBS 2048
	404
	405	static inline int should_fsync(struct thread_data *td)
	406	{
	407	if (td->last_was_sync)
	408	return 0;
	409	if (td_write(td) \|\| td_rw(td) \|\| td->o.override_sync)
	410	return 1;
	411
	412	return 0;
	413	}
	414
	415	/*
	416	* Init/option functions
	417	*/
	418	extern int __must_check fio_init_options(void);
	419	extern int __must_check parse_options(int, char **);
	420	extern int parse_jobs_ini(char *, int, int, int);
	421	extern int parse_cmd_line(int, char **, int);
	422	extern int fio_backend(void);
	423	extern void reset_fio_state(void);
	424	extern void clear_io_state(struct thread_data *);
	425	extern int fio_options_parse(struct thread_data , char *, int, int);
	426	extern void fio_keywords_init(void);
	427	extern int fio_cmd_option_parse(struct thread_data , const char , char *);
	428	extern int fio_cmd_ioengine_option_parse(struct thread_data , const char , char *);
	429	extern void fio_fill_default_options(struct thread_data *);
	430	extern int fio_show_option_help(const char *);
	431	extern void fio_options_set_ioengine_opts(struct option long_options, struct thread_data td);
	432	extern void fio_options_dup_and_init(struct option *);
	433	extern void fio_options_mem_dupe(struct thread_data *);
	434	extern void options_mem_dupe(void data, struct fio_option options);
	435	extern void td_fill_rand_seeds(struct thread_data *);
	436	extern void add_job_opts(const char **, int);
	437	extern char *num2str(unsigned long, int, int, int, int);
	438	extern int ioengine_load(struct thread_data *);
	439	extern int parse_dryrun(void);
	440
	441	extern uintptr_t page_mask;
	442	extern uintptr_t page_size;
	443	extern int initialize_fio(char *envp[]);
	444
	445	#define FIO_GETOPT_JOB 0x89000000
	446	#define FIO_GETOPT_IOENGINE 0x98000000
	447	#define FIO_NR_OPTIONS (FIO_MAX_OPTS + 128)
	448
	449	/*
	450	* ETA/status stuff
	451	*/
	452	extern void print_thread_status(void);
	453	extern void print_status_init(int);
	454	extern char *fio_uint_to_kmg(unsigned int val);
	455
	456	/*
	457	* Thread life cycle. Once a thread has a runstate beyond TD_INITIALIZED, it
	458	* will never back again. It may cycle between running/verififying/fsyncing.
	459	* Once the thread reaches TD_EXITED, it is just waiting for the core to
	460	* reap it.
	461	*/
	462	enum {
	463	TD_NOT_CREATED = 0,
	464	TD_CREATED,
	465	TD_INITIALIZED,
	466	TD_RAMP,
	467	TD_SETTING_UP,
	468	TD_RUNNING,
	469	TD_PRE_READING,
	470	TD_VERIFYING,
	471	TD_FSYNCING,
	472	TD_EXITED,
	473	TD_REAPED,
	474	};
	475
	476	extern void td_set_runstate(struct thread_data *, int);
	477	#define TERMINATE_ALL (-1)
	478	extern void fio_terminate_threads(int);
	479
	480	/*
	481	* Memory helpers
	482	*/
	483	extern int __must_check fio_pin_memory(struct thread_data *);
	484	extern void fio_unpin_memory(struct thread_data *);
	485	extern int __must_check allocate_io_mem(struct thread_data *);
	486	extern void free_io_mem(struct thread_data *);
	487	extern void free_threads_shm(void);
	488
	489	/*
	490	* Reset stats after ramp time completes
	491	*/
	492	extern void reset_all_stats(struct thread_data *);
	493
	494	/*
	495	* blktrace support
	496	*/
	497	#ifdef FIO_HAVE_BLKTRACE
	498	extern int is_blktrace(const char , int );
	499	extern int load_blktrace(struct thread_data , const char , int);
	500	#endif
	501
	502	/*
	503	* Latency target helpers
	504	*/
	505	extern void lat_target_check(struct thread_data *);
	506	extern void lat_target_init(struct thread_data *);
	507
	508	#define for_each_td(td, i) \
	509	for ((i) = 0, (td) = &threads[0]; (i) < (int) thread_number; (i)++, (td)++)
	510	#define for_each_file(td, f, i) \
	511	if ((td)->files_index) \
	512	for ((i) = 0, (f) = (td)->files[0]; \
	513	(i) < (td)->o.nr_files && ((f) = (td)->files[i]) != NULL; \
	514	(i)++)
	515
	516	#define fio_assert(td, cond) do { \
	517	if (!(cond)) { \
	518	int *__foo = NULL; \
	519	fprintf(stderr, "file:%s:%d, assert %s failed\n", __FILE__, __LINE__, #cond); \
	520	td_set_runstate((td), TD_EXITED); \
	521	(td)->error = EFAULT; \
	522	*__foo = 0; \
	523	} \
	524	} while (0)
	525
	526	static inline int fio_fill_issue_time(struct thread_data *td)
	527	{
	528	if (td->o.read_iolog_file \|\|
	529	!td->o.disable_clat \|\| !td->o.disable_slat \|\| !td->o.disable_bw)
	530	return 1;
	531
	532	return 0;
	533	}
	534
	535	static inline int __should_check_rate(struct thread_data *td,
	536	enum fio_ddir ddir)
	537	{
	538	struct thread_options *o = &td->o;
	539
	540	/*
	541	* If some rate setting was given, we need to check it
	542	*/
	543	if (o->rate[ddir] \|\| o->ratemin[ddir] \|\| o->rate_iops[ddir] \|\|
	544	o->rate_iops_min[ddir])
	545	return 1;
	546
	547	return 0;
	548	}
	549
	550	static inline int should_check_rate(struct thread_data *td,
	551	uint64_t *bytes_done)
	552	{
	553	int ret = 0;
	554
	555	if (bytes_done[DDIR_READ])
	556	ret \|= __should_check_rate(td, DDIR_READ);
	557	if (bytes_done[DDIR_WRITE])
	558	ret \|= __should_check_rate(td, DDIR_WRITE);
	559	if (bytes_done[DDIR_TRIM])
	560	ret \|= __should_check_rate(td, DDIR_TRIM);
	561
	562	return ret;
	563	}
	564
	565	static inline unsigned int td_max_bs(struct thread_data *td)
	566	{
	567	unsigned int max_bs;
	568
	569	max_bs = max(td->o.max_bs[DDIR_READ], td->o.max_bs[DDIR_WRITE]);
	570	return max(td->o.max_bs[DDIR_TRIM], max_bs);
	571	}
	572
	573	static inline unsigned int td_min_bs(struct thread_data *td)
	574	{
	575	unsigned int min_bs;
	576
	577	min_bs = min(td->o.min_bs[DDIR_READ], td->o.min_bs[DDIR_WRITE]);
	578	return min(td->o.min_bs[DDIR_TRIM], min_bs);
	579	}
	580
	581	static inline int is_power_of_2(unsigned int val)
	582	{
	583	return (val != 0 && ((val & (val - 1)) == 0));
	584	}
	585
	586	/*
	587	* We currently only need to do locking if we have verifier threads
	588	* accessing our internal structures too
	589	*/
	590	static inline void td_io_u_lock(struct thread_data *td)
	591	{
	592	if (td->o.verify_async)
	593	pthread_mutex_lock(&td->io_u_lock);
	594	}
	595
	596	static inline void td_io_u_unlock(struct thread_data *td)
	597	{
	598	if (td->o.verify_async)
	599	pthread_mutex_unlock(&td->io_u_lock);
	600	}
	601
	602	static inline void td_io_u_free_notify(struct thread_data *td)
	603	{
	604	if (td->o.verify_async)
	605	pthread_cond_signal(&td->free_cond);
	606	}
	607
	608	extern const char *fio_get_arch_string(int);
	609	extern const char *fio_get_os_string(int);
	610
	611	#define ARRAY_SIZE(x) (sizeof((x)) / (sizeof((x)[0])))
	612
	613	enum {
	614	FIO_OUTPUT_TERSE = 0,
	615	FIO_OUTPUT_JSON,
	616	FIO_OUTPUT_NORMAL,
	617	};
	618
	619	enum {
	620	FIO_RAND_DIST_RANDOM = 0,
	621	FIO_RAND_DIST_ZIPF,
	622	FIO_RAND_DIST_PARETO,
	623	};
	624
	625	enum {
	626	FIO_RAND_GEN_TAUSWORTHE = 0,
	627	FIO_RAND_GEN_LFSR,
	628	};
	629
	630	#endif