[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 "fio_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,
	TD_F_COMPRESS		= 128,
	TD_F_NOIO		= 256,
};

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;
	int latency_end_run;

	/*
	 * 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 {								\
		unsigned 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 int fio_running_or_pending_io_threads(void);
extern int fio_set_fd_nonblocking(int, const char *);

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_FINISHING,
	TD_EXITED,
	TD_REAPED,
};

extern void td_set_runstate(struct thread_data *, int);
extern int td_bump_runstate(struct thread_data *, int);
extern void td_restore_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 *);
extern void lat_target_reset(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 long 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);

#ifdef FIO_INTERNAL
#define ARRAY_SIZE(x) (sizeof((x)) / (sizeof((x)[0])))
#endif

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,
};

enum {
	FIO_CPUS_SHARED		= 0,
	FIO_CPUS_SPLIT,
};

#endif
Commit	Line	Data
ebac4655 JA	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>
3c39a379 JA	9	#include <errno.h>
	10	#include <stdlib.h>
	11	#include <stdio.h>
6d6f031f	12	#include <unistd.h>
34cfcdaf	13	#include <string.h>
cd14cc10	14	#include <inttypes.h>
7101d9c2	15	#include <assert.h>
ebac4655	16
317b95d0	17	#include "compiler/compiler.h"
8062f527	18	#include "thread_options.h"
01743ee1	19	#include "flist.h"
e2887563	20	#include "fifo.h"
317b95d0 JA	21	#include "arch/arch.h"
317b95d0 JA	22	#include "os/os.h"
07739b57	23	#include "mutex.h"
a3d741fa JA	24	#include "log.h"
a3d741fa JA	25	#include "debug.h"
d6aed795 JA	26	#include "file.h"
d6aed795 JA	27	#include "io_ddir.h"
dcefb588	28	#include "ioengine.h"
5995a6a4	29	#include "iolog.h"
c5c8bd5c	30	#include "helpers.h"
07b3232d	31	#include "options.h"
7eb36574	32	#include "profile.h"
ef3c94b6	33	#include "fio_time.h"
a7346816	34	#include "gettime.h"
bf2e821a	35	#include "lib/getopt.h"
2615cc4b	36	#include "lib/rand.h"
7fb6d453	37	#include "lib/rbtree.h"
dd366728	38	#include "client.h"
37db14fe	39	#include "server.h"
a64e88da	40	#include "stat.h"
9e684a49	41	#include "flow.h"
2ae0b204	42	#include "io_u_queue.h"
ebac4655	43
4700b234	44	#ifdef CONFIG_SOLARISAIO
417f0068 JA	45	#include <sys/asynch.h>
	46	#endif
	47
67bf9823	48	#ifdef CONFIG_LIBNUMA
d0b937ed YR	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
38dad62d JA	58	/*
	59	* offset generator types
	60	*/
	61	enum {
	62	RW_SEQ_SEQ = 0,
	63	RW_SEQ_IDENT,
	64	};
	65
d72be545 JA	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,
bedc9dc2	74	TD_F_COMPRESS = 128,
046395d7	75	TD_F_NOIO = 256,
d72be545 JA	76	};
d72be545 JA	77
d24945f0 JA	78	enum {
	79	FIO_RAND_BS_OFF = 0,
	80	FIO_RAND_VER_OFF,
	81	FIO_RAND_MIX_OFF,
	82	FIO_RAND_FILE_OFF,
	83	FIO_RAND_BLOCK_OFF,
	84	FIO_RAND_FILE_SIZE_OFF,
	85	FIO_RAND_TRIM_OFF,
	86	FIO_RAND_BUF_OFF,
d9472271 JA	87	FIO_RAND_SEQ_RAND_READ_OFF,
	88	FIO_RAND_SEQ_RAND_WRITE_OFF,
	89	FIO_RAND_SEQ_RAND_TRIM_OFF,
23ed19b0	90	FIO_RAND_START_DELAY,
d24945f0 JA	91	FIO_RAND_NR_OFFS,
	92	};
	93
2dc1bbeb JA	94	/*
	95	* This describes a single thread/process executing a fio job.
	96	*/
	97	struct thread_data {
	98	struct thread_options o;
d72be545	99	unsigned long flags;
de890a1e	100	void *eo;
e4e33258	101	char verror[FIO_VERROR_SIZE];
2dc1bbeb	102	pthread_t thread;
2f122b13 JA	103	unsigned int thread_number;
2f122b13 JA	104	unsigned int groupid;
2dc1bbeb	105	struct thread_stat ts;
7b9f733a	106
46bcd498 JA	107	int client_type;
46bcd498 JA	108
7b9f733a JA	109	struct io_log *slat_log;
	110	struct io_log *clat_log;
	111	struct io_log *lat_log;
	112	struct io_log *bw_log;
c8eeb9df	113	struct io_log *iops_log;
7b9f733a	114
6eaf09d6	115	uint64_t stat_io_bytes[DDIR_RWDIR_CNT];
c8eeb9df JA	116	struct timeval bw_sample_time;
c8eeb9df JA	117
6eaf09d6	118	uint64_t stat_io_blocks[DDIR_RWDIR_CNT];
c8eeb9df	119	struct timeval iops_sample_time;
f0505a14	120
c97f1ad6 JA	121	volatile int update_rusage;
c97f1ad6 JA	122	struct fio_mutex *rusage_sem;
c8aaba19 JA	123	struct rusage ru_start;
	124	struct rusage ru_end;
	125
126d65c6	126	struct fio_file **files;
d7df1d13	127	unsigned char *file_locks;
dd87b2c9	128	unsigned int files_size;
2dc1bbeb JA	129	unsigned int files_index;
2dc1bbeb JA	130	unsigned int nr_open_files;
1020a139	131	unsigned int nr_done_files;
2dc1bbeb JA	132	unsigned int nr_normal_files;
	133	union {
	134	unsigned int next_file;
	135	os_random_state_t next_file_state;
4c07ad86	136	struct frand_state __next_file_state;
2dc1bbeb JA	137	};
2dc1bbeb JA	138	int error;
a5e371a6	139	int sig;
20e354ef	140	int done;
2dc1bbeb JA	141	pid_t pid;
	142	char *orig_buffer;
	143	size_t orig_buffer_size;
	144	volatile int terminate;
	145	volatile int runstate;
2dc1bbeb	146	unsigned int last_was_sync;
9e144189	147	enum fio_ddir last_ddir;
2dc1bbeb	148
2dc1bbeb JA	149	int mmapfd;
2dc1bbeb JA	150
843a7413 JA	151	void *iolog_buf;
843a7413 JA	152	FILE *iolog_f;
ebac4655	153
da86774e	154	char *sysfs_root;
da86774e	155
d24945f0	156	unsigned long rand_seeds[FIO_RAND_NR_OFFS];
5bfc35d7	157
4c07ad86 JA	158	union {
	159	os_random_state_t bsrange_state;
	160	struct frand_state __bsrange_state;
	161	};
	162	union {
	163	os_random_state_t verify_state;
	164	struct frand_state __verify_state;
	165	};
	166	union {
	167	os_random_state_t trim_state;
	168	struct frand_state __trim_state;
	169	};
23ed19b0 CE	170	union {
	171	os_random_state_t delay_state;
	172	struct frand_state __delay_state;
	173	};
ebac4655	174
3545a109 JA	175	struct frand_state buf_state;
3545a109 JA	176
9e144189	177	unsigned int verify_batch;
0d29de83	178	unsigned int trim_batch;
9e144189	179
ebac4655 JA	180	int shm_id;
ebac4655 JA	181
e9c047a0 JA	182	/*
	183	* IO engine hooks, contains everything needed to submit an io_u
	184	* to any of the available IO engines.
	185	*/
2866c82d	186	struct ioengine_ops *io_ops;
ebac4655	187
e9c047a0	188	/*
422f9e4b	189	* Queue depth of io_u's that fio MIGHT do
e9c047a0	190	*/
ebac4655	191	unsigned int cur_depth;
422f9e4b RM	192
	193	/*
	194	* io_u's about to be committed
	195	*/
d8005759	196	unsigned int io_u_queued;
422f9e4b RM	197
	198	/*
	199	* io_u's submitted but not completed yet
	200	*/
	201	unsigned int io_u_in_flight;
	202
	203	/*
	204	* List of free and busy io_u's
	205	*/
2ae0b204 JA	206	struct io_u_ring io_u_requeues;
	207	struct io_u_queue io_u_freelist;
	208	struct io_u_queue io_u_all;
e8462bd8 JA	209	pthread_mutex_t io_u_lock;
	210	pthread_cond_t free_cond;
	211
	212	/*
	213	* async verify offload
	214	*/
	215	struct flist_head verify_list;
	216	pthread_t *verify_threads;
	217	unsigned int nr_verify_threads;
	218	pthread_cond_t verify_cond;
	219	int verify_thread_exit;
ebac4655	220
e9c047a0 JA	221	/*
	222	* Rate state
	223	*/
bdb6d20a	224	uint64_t rate_bps[DDIR_RWDIR_CNT];
6eaf09d6 SL	225	long rate_pending_usleep[DDIR_RWDIR_CNT];
	226	unsigned long rate_bytes[DDIR_RWDIR_CNT];
	227	unsigned long rate_blocks[DDIR_RWDIR_CNT];
	228	struct timeval lastrate[DDIR_RWDIR_CNT];
ebac4655	229
bdb6d20a JA	230	uint64_t total_io_size;
bdb6d20a JA	231	uint64_t fill_device_size;
ebac4655	232
6eaf09d6	233	unsigned long io_issues[DDIR_RWDIR_CNT];
bdb6d20a JA	234	uint64_t io_blocks[DDIR_RWDIR_CNT];
	235	uint64_t this_io_blocks[DDIR_RWDIR_CNT];
	236	uint64_t io_bytes[DDIR_RWDIR_CNT];
	237	uint64_t io_skip_bytes;
	238	uint64_t this_io_bytes[DDIR_RWDIR_CNT];
	239	uint64_t zone_bytes;
cdd18ad8	240	struct fio_mutex *mutex;
ebac4655	241
e9c047a0 JA	242	/*
	243	* State for random io, a bitmap of blocks done vs not done
	244	*/
4c07ad86 JA	245	union {
	246	os_random_state_t random_state;
	247	struct frand_state __random_state;
	248	};
ebac4655	249
ebac4655 JA	250	struct timeval start; /* start of this loop */
ebac4655 JA	251	struct timeval epoch; /* time job was started */
a61eddec	252	struct timeval last_issue;
993bf48b JA	253	struct timeval tv_cache;
	254	unsigned int tv_cache_nr;
	255	unsigned int tv_cache_mask;
721938ae	256	unsigned int ramp_time_over;
ebac4655	257
3e260a46 JA	258	/*
	259	* Time since last latency_window was started
	260	*/
	261	struct timeval latency_ts;
	262	unsigned int latency_qd;
	263	unsigned int latency_qd_high;
	264	unsigned int latency_qd_low;
	265	unsigned int latency_failed;
	266	uint64_t latency_ios;
6bb58215	267	int latency_end_run;
3e260a46	268
e9c047a0 JA	269	/*
	270	* read/write mixed workload state
	271	*/
4c07ad86 JA	272	union {
	273	os_random_state_t rwmix_state;
	274	struct frand_state __rwmix_state;
	275	};
e4928662	276	unsigned long rwmix_issues;
e9c047a0	277	enum fio_ddir rwmix_ddir;
5736c10d	278	unsigned int ddir_seq_nr;
a6ccc7be	279
211c9b89 JA	280	/*
	281	* rand/seq mixed workload state
	282	*/
	283	union {
d9472271 JA	284	os_random_state_t seq_rand_state[DDIR_RWDIR_CNT];
d9472271 JA	285	struct frand_state __seq_rand_state[DDIR_RWDIR_CNT];
211c9b89 JA	286	};
211c9b89 JA	287
e9c047a0	288	/*
8de8f047 JA	289	* IO history logs for verification. We use a tree for sorting,
8de8f047 JA	290	* if we are overwriting. Otherwise just use a fifo.
e9c047a0	291	*/
4b87898e	292	struct rb_root io_hist_tree;
01743ee1	293	struct flist_head io_hist_list;
9e144189	294	unsigned long io_hist_len;
8de8f047 JA	295
	296	/*
	297	* For IO replaying
	298	*/
01743ee1	299	struct flist_head io_log_list;
433afcb4	300
0d29de83 JA	301	/*
	302	* For tracking/handling discards
	303	*/
	304	struct flist_head trim_list;
	305	unsigned long trim_entries;
	306
1ae83d45 JA	307	struct flist_head next_rand_list;
1ae83d45 JA	308
1907dbc6 JA	309	/*
	310	* for fileservice, how often to switch to a new file
	311	*/
	312	unsigned int file_service_nr;
	313	unsigned int file_service_left;
	314	struct fio_file *file_service_file;
9c60ce64	315
44f29692 JA	316	unsigned int sync_file_range_nr;
44f29692 JA	317
9c60ce64 JA	318	/*
	319	* For generating file sizes
	320	*/
4c07ad86 JA	321	union {
	322	os_random_state_t file_size_state;
	323	struct frand_state __file_size_state;
	324	};
f2bba182 RR	325
	326	/*
	327	* Error counts
	328	*/
	329	unsigned int total_err_count;
	330	int first_error;
15dc1934	331
9e684a49 DE	332	struct fio_flow *flow;
9e684a49 DE	333
15dc1934 JA	334	/*
	335	* Can be overloaded by profiles
	336	*/
7eb36574	337	struct prof_io_ops prof_io_ops;
58c55ba0	338	void *prof_data;
9a3f1100 JA	339
9a3f1100 JA	340	void *pinned_mem;
ebac4655 JA	341	};
ebac4655 JA	342
e592a06b AC	343	/*
	344	* when should interactive ETA output be generated
	345	*/
	346	enum {
	347	FIO_ETA_AUTO,
	348	FIO_ETA_ALWAYS,
	349	FIO_ETA_NEVER,
	350	};
	351
e1161c32	352	#define __td_verror(td, err, msg, func) \
ebac4655	353	do { \
21beb609	354	unsigned int ____e = (err); \
19abcd3d JA	355	if ((td)->error) \
19abcd3d JA	356	break; \
8271963e	357	(td)->error = ____e; \
f2bba182	358	if (!(td)->first_error) \
98ffb8f3	359	snprintf(td->verror, sizeof(td->verror), "file:%s:%d, func=%s, error=%s", __FILE__, __LINE__, (func), (msg)); \
ebac4655 JA	360	} while (0)
ebac4655 JA	361
b990b5c0	362
f2bba182 RR	363	#define td_clear_error(td) \
f2bba182 RR	364	(td)->error = 0;
e1161c32 JA	365	#define td_verror(td, err, func) \
	366	__td_verror((td), (err), strerror((err)), (func))
	367	#define td_vmsg(td, err, msg, func) \
	368	__td_verror((td), (err), (msg), (func))
b990b5c0	369
81179eec JA	370	#define __fio_stringify_1(x) #x
	371	#define __fio_stringify(x) __fio_stringify_1(x)
	372
ebac4655	373	extern int exitall_on_terminate;
11e950bd	374	extern unsigned int thread_number;
108fea77	375	extern unsigned int stat_number;
ebac4655 JA	376	extern int shm_id;
ebac4655 JA	377	extern int groupid;
f3afa57e	378	extern int output_format;
2b8c71b0	379	extern int append_terse_output;
53cdc686	380	extern int temp_stall_ts;
7078106d	381	extern uintptr_t page_mask, page_size;
4241ea8f	382	extern int read_only;
e592a06b	383	extern int eta_print;
e382e661	384	extern int eta_new_line;
d3eeeabc	385	extern unsigned long done_secs;
01f06b63	386	extern char *job_section;
be4ecfdf JA	387	extern int fio_gtod_offload;
be4ecfdf JA	388	extern int fio_gtod_cpu;
c223da83	389	extern enum fio_cs fio_clock_source;
fa80feae	390	extern int fio_clock_source_set;
a9523c6f	391	extern int warnings_fatal;
f57a9c59	392	extern int terse_version;
5c341e9a	393	extern int is_backend;
a37f69b7	394	extern int nr_clients;
e46d8091	395	extern int log_syslog;
06464907	396	extern int status_interval;
5e726d0a	397	extern const char fio_version_string[];
ebac4655 JA	398
	399	extern struct thread_data *threads;
	400
7101d9c2 JA	401	static inline void fio_ro_check(struct thread_data td, struct io_u io_u)
	402	{
	403	assert(!(io_u->ddir == DDIR_WRITE && !td_write(td)));
	404	}
	405
fca70358	406	#define REAL_MAX_JOBS 2048
ebac4655	407
87dc1ab1 JA	408	static inline int should_fsync(struct thread_data *td)
	409	{
	410	if (td->last_was_sync)
	411	return 0;
2dc1bbeb	412	if (td_write(td) \|\| td_rw(td) \|\| td->o.override_sync)
87dc1ab1 JA	413	return 1;
	414
	415	return 0;
	416	}
	417
8914a9d8	418	/*
214e1eca	419	* Init/option functions
8914a9d8	420	*/
0420ba6a	421	extern int __must_check fio_init_options(void);
72cb971b	422	extern int __must_check parse_options(int, char **);
46bcd498 JA	423	extern int parse_jobs_ini(char *, int, int, int);
46bcd498 JA	424	extern int parse_cmd_line(int, char **, int);
2e1df07d	425	extern int fio_backend(void);
50d16976	426	extern void reset_fio_state(void);
2e1df07d	427	extern void clear_io_state(struct thread_data *);
292cc475	428	extern int fio_options_parse(struct thread_data , char *, int, int);
74929ac2	429	extern void fio_keywords_init(void);
214e1eca	430	extern int fio_cmd_option_parse(struct thread_data , const char , char *);
de890a1e	431	extern int fio_cmd_ioengine_option_parse(struct thread_data , const char , char *);
214e1eca JA	432	extern void fio_fill_default_options(struct thread_data *);
214e1eca JA	433	extern int fio_show_option_help(const char *);
de890a1e	434	extern void fio_options_set_ioengine_opts(struct option long_options, struct thread_data td);
214e1eca	435	extern void fio_options_dup_and_init(struct option *);
de890a1e SL	436	extern void fio_options_mem_dupe(struct thread_data *);
de890a1e SL	437	extern void options_mem_dupe(void data, struct fio_option options);
5bfc35d7	438	extern void td_fill_rand_seeds(struct thread_data *);
46bcd498	439	extern void add_job_opts(const char **, int);
73798eb2	440	extern char *num2str(unsigned long, int, int, int, int);
de890a1e	441	extern int ioengine_load(struct thread_data *);
f9633d72	442	extern int parse_dryrun(void);
046395d7	443	extern int fio_running_or_pending_io_threads(void);
3a35845f	444	extern int fio_set_fd_nonblocking(int, const char *);
1ec3d69b	445
a569b45f JA	446	extern uintptr_t page_mask;
	447	extern uintptr_t page_size;
	448	extern int initialize_fio(char *envp[]);
	449
de890a1e SL	450	#define FIO_GETOPT_JOB 0x89000000
de890a1e SL	451	#define FIO_GETOPT_IOENGINE 0x98000000
07b3232d	452	#define FIO_NR_OPTIONS (FIO_MAX_OPTS + 128)
8914a9d8	453
263e529f JA	454	/*
	455	* ETA/status stuff
	456	*/
	457	extern void print_thread_status(void);
	458	extern void print_status_init(int);
0a428b23	459	extern char *fio_uint_to_kmg(unsigned int val);
263e529f JA	460
	461	/*
	462	* Thread life cycle. Once a thread has a runstate beyond TD_INITIALIZED, it
	463	* will never back again. It may cycle between running/verififying/fsyncing.
	464	* Once the thread reaches TD_EXITED, it is just waiting for the core to
	465	* reap it.
	466	*/
	467	enum {
	468	TD_NOT_CREATED = 0,
	469	TD_CREATED,
	470	TD_INITIALIZED,
b29ee5b3	471	TD_RAMP,
36d80bc7	472	TD_SETTING_UP,
263e529f	473	TD_RUNNING,
b0f65863	474	TD_PRE_READING,
263e529f JA	475	TD_VERIFYING,
263e529f JA	476	TD_FSYNCING,
3d434057	477	TD_FINISHING,
263e529f JA	478	TD_EXITED,
	479	TD_REAPED,
	480	};
	481
b29ee5b3	482	extern void td_set_runstate(struct thread_data *, int);
8edd973d JA	483	extern int td_bump_runstate(struct thread_data *, int);
	484	extern void td_restore_runstate(struct thread_data *, int);
	485
cc0df00a JA	486	#define TERMINATE_ALL (-1)
cc0df00a JA	487	extern void fio_terminate_threads(int);
b29ee5b3	488
2f9ade3c JA	489	/*
	490	* Memory helpers
	491	*/
9a3f1100 JA	492	extern int __must_check fio_pin_memory(struct thread_data *);
9a3f1100 JA	493	extern void fio_unpin_memory(struct thread_data *);
b2fdda43	494	extern int __must_check allocate_io_mem(struct thread_data *);
2f9ade3c	495	extern void free_io_mem(struct thread_data *);
2bb3f0a7	496	extern void free_threads_shm(void);
2f9ade3c	497
b29ee5b3 JA	498	/*
	499	* Reset stats after ramp time completes
	500	*/
	501	extern void reset_all_stats(struct thread_data *);
	502
fb7b71a3 JA	503	/*
	504	* blktrace support
	505	*/
5e62c22a	506	#ifdef FIO_HAVE_BLKTRACE
d95b34a6 JA	507	extern int is_blktrace(const char , int );
d95b34a6 JA	508	extern int load_blktrace(struct thread_data , const char , int);
5e62c22a	509	#endif
fb7b71a3	510
3e260a46 JA	511	/*
	512	* Latency target helpers
	513	*/
	514	extern void lat_target_check(struct thread_data *);
	515	extern void lat_target_init(struct thread_data *);
6bb58215	516	extern void lat_target_reset(struct thread_data *);
3e260a46	517
34572e28 JA	518	#define for_each_td(td, i) \
34572e28 JA	519	for ((i) = 0, (td) = &threads[0]; (i) < (int) thread_number; (i)++, (td)++)
53cdc686	520	#define for_each_file(td, f, i) \
691c8fb0 JA	521	if ((td)->files_index) \
	522	for ((i) = 0, (f) = (td)->files[0]; \
	523	(i) < (td)->o.nr_files && ((f) = (td)->files[i]) != NULL; \
	524	(i)++)
53cdc686	525
0032bf9f JA	526	#define fio_assert(td, cond) do { \
0032bf9f JA	527	if (!(cond)) { \
340fd243	528	int *__foo = NULL; \
0032bf9f	529	fprintf(stderr, "file:%s:%d, assert %s failed\n", __FILE__, __LINE__, #cond); \
ac18ea38	530	td_set_runstate((td), TD_EXITED); \
437c9b71	531	(td)->error = EFAULT; \
340fd243	532	*__foo = 0; \
0032bf9f JA	533	} \
	534	} while (0)
	535
12d9d841 JA	536	static inline int fio_fill_issue_time(struct thread_data *td)
	537	{
	538	if (td->o.read_iolog_file \|\|
	539	!td->o.disable_clat \|\| !td->o.disable_slat \|\| !td->o.disable_bw)
	540	return 1;
	541
	542	return 0;
	543	}
	544
581e7141 JA	545	static inline int __should_check_rate(struct thread_data *td,
	546	enum fio_ddir ddir)
	547	{
	548	struct thread_options *o = &td->o;
	549
	550	/*
	551	* If some rate setting was given, we need to check it
	552	*/
	553	if (o->rate[ddir] \|\| o->ratemin[ddir] \|\| o->rate_iops[ddir] \|\|
	554	o->rate_iops_min[ddir])
	555	return 1;
	556
	557	return 0;
	558	}
	559
	560	static inline int should_check_rate(struct thread_data *td,
aa83321f	561	uint64_t *bytes_done)
581e7141 JA	562	{
	563	int ret = 0;
	564
6eaf09d6 SL	565	if (bytes_done[DDIR_READ])
	566	ret \|= __should_check_rate(td, DDIR_READ);
	567	if (bytes_done[DDIR_WRITE])
	568	ret \|= __should_check_rate(td, DDIR_WRITE);
	569	if (bytes_done[DDIR_TRIM])
	570	ret \|= __should_check_rate(td, DDIR_TRIM);
581e7141 JA	571
	572	return ret;
	573	}
	574
74f4b020 JA	575	static inline unsigned int td_max_bs(struct thread_data *td)
	576	{
	577	unsigned int max_bs;
	578
	579	max_bs = max(td->o.max_bs[DDIR_READ], td->o.max_bs[DDIR_WRITE]);
	580	return max(td->o.max_bs[DDIR_TRIM], max_bs);
	581	}
	582
8d916c94 JA	583	static inline unsigned int td_min_bs(struct thread_data *td)
	584	{
	585	unsigned int min_bs;
	586
	587	min_bs = min(td->o.min_bs[DDIR_READ], td->o.min_bs[DDIR_WRITE]);
	588	return min(td->o.min_bs[DDIR_TRIM], min_bs);
	589	}
	590
088b17c5	591	static inline int is_power_of_2(unsigned long val)
d529ee19 JA	592	{
	593	return (val != 0 && ((val & (val - 1)) == 0));
	594	}
	595
e8462bd8 JA	596	/*
	597	* We currently only need to do locking if we have verifier threads
	598	* accessing our internal structures too
	599	*/
	600	static inline void td_io_u_lock(struct thread_data *td)
	601	{
	602	if (td->o.verify_async)
	603	pthread_mutex_lock(&td->io_u_lock);
	604	}
	605
	606	static inline void td_io_u_unlock(struct thread_data *td)
	607	{
	608	if (td->o.verify_async)
	609	pthread_mutex_unlock(&td->io_u_lock);
	610	}
	611
	612	static inline void td_io_u_free_notify(struct thread_data *td)
	613	{
	614	if (td->o.verify_async)
	615	pthread_cond_signal(&td->free_cond);
	616	}
	617
cca84643 JA	618	extern const char *fio_get_arch_string(int);
	619	extern const char *fio_get_os_string(int);
	620
e2c81040	621	#ifdef FIO_INTERNAL
05775438	622	#define ARRAY_SIZE(x) (sizeof((x)) / (sizeof((x)[0])))
e2c81040	623	#endif
05775438	624
f3afa57e JA	625	enum {
	626	FIO_OUTPUT_TERSE = 0,
	627	FIO_OUTPUT_JSON,
	628	FIO_OUTPUT_NORMAL,
	629	};
	630
e25839d4 JA	631	enum {
	632	FIO_RAND_DIST_RANDOM = 0,
	633	FIO_RAND_DIST_ZIPF,
925fee33	634	FIO_RAND_DIST_PARETO,
e25839d4 JA	635	};
e25839d4 JA	636
8055e41d JA	637	enum {
	638	FIO_RAND_GEN_TAUSWORTHE = 0,
	639	FIO_RAND_GEN_LFSR,
	640	};
	641
c2acfbac JA	642	enum {
	643	FIO_CPUS_SHARED = 0,
	644	FIO_CPUS_SPLIT,
	645	};
	646
ebac4655	647	#endif