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

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

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

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

/*
 * This describes a single thread/process executing a fio job.
 */
struct thread_data {
	struct thread_options o;
	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;

	struct rusage ru_start;
	struct rusage ru_end;

	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;
		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[8];

	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;
	};

	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 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 long 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];

	unsigned long long total_io_size;
	unsigned long long fill_device_size;

	unsigned long io_issues[DDIR_RWDIR_CNT];
	unsigned long long io_blocks[DDIR_RWDIR_CNT];
	unsigned long long this_io_blocks[DDIR_RWDIR_CNT];
	unsigned long long io_bytes[DDIR_RWDIR_CNT];
	unsigned long long io_skip_bytes;
	unsigned long long this_io_bytes[DDIR_RWDIR_CNT];
	unsigned long long 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;

	/*
	 * 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;

	/*
	 * 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
	 */
	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 {								\
		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))

#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 int shm_id;
extern int groupid;
extern int output_format;
extern int temp_stall_ts;
extern uintptr_t 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 int terse_version;
extern int is_backend;
extern int nr_clients;
extern int log_syslog;
extern const char fio_version_string[];
extern const fio_fp64_t def_percentile_list[FIO_IO_U_LIST_MAX_LEN];

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(unsigned long))
#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 REAL_MAX_JOBS		2048

static inline enum error_type td_error_type(enum fio_ddir ddir, int err)
{
	if (err == EILSEQ)
		return ERROR_TYPE_VERIFY_BIT;
	if (ddir == DDIR_READ)
		return ERROR_TYPE_READ_BIT;
	return ERROR_TYPE_WRITE_BIT;
}

static int __NON_FATAL_ERR[] = {EIO, EILSEQ};
static inline int td_non_fatal_error(struct thread_data *td,
				     enum error_type_bit etype, int err)
{
	int i;
	if (!td->o.ignore_error[etype]) {
		td->o.ignore_error[etype] = __NON_FATAL_ERR;
		td->o.ignore_error_nr[etype] = sizeof(__NON_FATAL_ERR)
			/ sizeof(int);
	}

	if (!(td->o.continue_on_error & (1 << etype)))
		return 0;
	for (i = 0; i < td->o.ignore_error_nr[etype]; i++)
		if (td->o.ignore_error[etype][i] == err)
			return 1;
	return 0;
}

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 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);
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);
extern int ioengine_load(struct thread_data *);

extern unsigned long page_mask;
extern unsigned long 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_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 *);
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_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,
				    unsigned long *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 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,
};

#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
ecc314ba BC	17	struct thread_data;
ecc314ba BC	18
317b95d0	19	#include "compiler/compiler.h"
ec41265e	20	#include "thread_options.h"
01743ee1	21	#include "flist.h"
e2887563	22	#include "fifo.h"
05775438	23	#include "lib/rbtree.h"
317b95d0 JA	24	#include "arch/arch.h"
317b95d0 JA	25	#include "os/os.h"
07739b57	26	#include "mutex.h"
a3d741fa JA	27	#include "log.h"
a3d741fa JA	28	#include "debug.h"
d6aed795 JA	29	#include "file.h"
d6aed795 JA	30	#include "io_ddir.h"
dcefb588	31	#include "ioengine.h"
5995a6a4	32	#include "iolog.h"
c5c8bd5c	33	#include "helpers.h"
07b3232d	34	#include "options.h"
7eb36574	35	#include "profile.h"
c223da83	36	#include "time.h"
a7346816	37	#include "gettime.h"
bf2e821a	38	#include "lib/getopt.h"
2615cc4b	39	#include "lib/rand.h"
dd366728	40	#include "client.h"
37db14fe	41	#include "server.h"
a64e88da	42	#include "stat.h"
9e684a49	43	#include "flow.h"
ebac4655	44
609342ff DL	45	#ifdef FIO_HAVE_GUASI
	46	#include <guasi.h>
	47	#endif
	48
417f0068 JA	49	#ifdef FIO_HAVE_SOLARISAIO
	50	#include <sys/asynch.h>
	51	#endif
	52
38dad62d JA	53	/*
	54	* offset generator types
	55	*/
	56	enum {
	57	RW_SEQ_SEQ = 0,
	58	RW_SEQ_IDENT,
	59	};
	60
2dc1bbeb JA	61	/*
	62	* This describes a single thread/process executing a fio job.
	63	*/
	64	struct thread_data {
	65	struct thread_options o;
de890a1e	66	void *eo;
e4e33258	67	char verror[FIO_VERROR_SIZE];
2dc1bbeb	68	pthread_t thread;
2f122b13 JA	69	unsigned int thread_number;
2f122b13 JA	70	unsigned int groupid;
2dc1bbeb	71	struct thread_stat ts;
7b9f733a	72
46bcd498 JA	73	int client_type;
46bcd498 JA	74
7b9f733a JA	75	struct io_log *slat_log;
	76	struct io_log *clat_log;
	77	struct io_log *lat_log;
	78	struct io_log *bw_log;
c8eeb9df	79	struct io_log *iops_log;
7b9f733a	80
6eaf09d6	81	uint64_t stat_io_bytes[DDIR_RWDIR_CNT];
c8eeb9df JA	82	struct timeval bw_sample_time;
c8eeb9df JA	83
6eaf09d6	84	uint64_t stat_io_blocks[DDIR_RWDIR_CNT];
c8eeb9df	85	struct timeval iops_sample_time;
f0505a14	86
c8aaba19 JA	87	struct rusage ru_start;
	88	struct rusage ru_end;
	89
126d65c6	90	struct fio_file **files;
dd87b2c9	91	unsigned int files_size;
2dc1bbeb JA	92	unsigned int files_index;
2dc1bbeb JA	93	unsigned int nr_open_files;
1020a139	94	unsigned int nr_done_files;
2dc1bbeb JA	95	unsigned int nr_normal_files;
	96	union {
	97	unsigned int next_file;
	98	os_random_state_t next_file_state;
4c07ad86	99	struct frand_state __next_file_state;
2dc1bbeb JA	100	};
2dc1bbeb JA	101	int error;
a5e371a6	102	int sig;
20e354ef	103	int done;
2dc1bbeb JA	104	pid_t pid;
	105	char *orig_buffer;
	106	size_t orig_buffer_size;
	107	volatile int terminate;
	108	volatile int runstate;
2dc1bbeb	109	unsigned int last_was_sync;
9e144189	110	enum fio_ddir last_ddir;
2dc1bbeb	111
2dc1bbeb JA	112	int mmapfd;
2dc1bbeb JA	113
843a7413 JA	114	void *iolog_buf;
843a7413 JA	115	FILE *iolog_f;
ebac4655	116
da86774e	117	char *sysfs_root;
da86774e	118
3545a109	119	unsigned long rand_seeds[8];
5bfc35d7	120
4c07ad86 JA	121	union {
	122	os_random_state_t bsrange_state;
	123	struct frand_state __bsrange_state;
	124	};
	125	union {
	126	os_random_state_t verify_state;
	127	struct frand_state __verify_state;
	128	};
	129	union {
	130	os_random_state_t trim_state;
	131	struct frand_state __trim_state;
	132	};
ebac4655	133
3545a109 JA	134	struct frand_state buf_state;
3545a109 JA	135
9e144189	136	unsigned int verify_batch;
0d29de83	137	unsigned int trim_batch;
9e144189	138
ebac4655 JA	139	int shm_id;
ebac4655 JA	140
e9c047a0 JA	141	/*
	142	* IO engine hooks, contains everything needed to submit an io_u
	143	* to any of the available IO engines.
	144	*/
2866c82d	145	struct ioengine_ops *io_ops;
ebac4655	146
e9c047a0	147	/*
422f9e4b	148	* Queue depth of io_u's that fio MIGHT do
e9c047a0	149	*/
ebac4655	150	unsigned int cur_depth;
422f9e4b RM	151
	152	/*
	153	* io_u's about to be committed
	154	*/
d8005759	155	unsigned int io_u_queued;
422f9e4b RM	156
	157	/*
	158	* io_u's submitted but not completed yet
	159	*/
	160	unsigned int io_u_in_flight;
	161
	162	/*
	163	* List of free and busy io_u's
	164	*/
01743ee1 JA	165	struct flist_head io_u_freelist;
	166	struct flist_head io_u_busylist;
	167	struct flist_head io_u_requeues;
e8462bd8 JA	168	pthread_mutex_t io_u_lock;
	169	pthread_cond_t free_cond;
	170
	171	/*
	172	* async verify offload
	173	*/
	174	struct flist_head verify_list;
	175	pthread_t *verify_threads;
	176	unsigned int nr_verify_threads;
	177	pthread_cond_t verify_cond;
	178	int verify_thread_exit;
ebac4655	179
e9c047a0 JA	180	/*
	181	* Rate state
	182	*/
6eaf09d6 SL	183	unsigned long long rate_bps[DDIR_RWDIR_CNT];
	184	long rate_pending_usleep[DDIR_RWDIR_CNT];
	185	unsigned long rate_bytes[DDIR_RWDIR_CNT];
	186	unsigned long rate_blocks[DDIR_RWDIR_CNT];
	187	struct timeval lastrate[DDIR_RWDIR_CNT];
ebac4655	188
ebac4655	189	unsigned long long total_io_size;
2e3bd4c2	190	unsigned long long fill_device_size;
ebac4655	191
6eaf09d6 SL	192	unsigned long io_issues[DDIR_RWDIR_CNT];
	193	unsigned long long io_blocks[DDIR_RWDIR_CNT];
	194	unsigned long long this_io_blocks[DDIR_RWDIR_CNT];
	195	unsigned long long io_bytes[DDIR_RWDIR_CNT];
48f5abd3	196	unsigned long long io_skip_bytes;
6eaf09d6	197	unsigned long long this_io_bytes[DDIR_RWDIR_CNT];
079ad09b	198	unsigned long long zone_bytes;
cdd18ad8	199	struct fio_mutex *mutex;
ebac4655	200
e9c047a0 JA	201	/*
	202	* State for random io, a bitmap of blocks done vs not done
	203	*/
4c07ad86 JA	204	union {
	205	os_random_state_t random_state;
	206	struct frand_state __random_state;
	207	};
ebac4655	208
ebac4655 JA	209	struct timeval start; /* start of this loop */
ebac4655 JA	210	struct timeval epoch; /* time job was started */
a61eddec	211	struct timeval last_issue;
993bf48b JA	212	struct timeval tv_cache;
	213	unsigned int tv_cache_nr;
	214	unsigned int tv_cache_mask;
721938ae	215	unsigned int ramp_time_over;
ebac4655	216
e9c047a0 JA	217	/*
	218	* read/write mixed workload state
	219	*/
4c07ad86 JA	220	union {
	221	os_random_state_t rwmix_state;
	222	struct frand_state __rwmix_state;
	223	};
e4928662	224	unsigned long rwmix_issues;
e9c047a0	225	enum fio_ddir rwmix_ddir;
5736c10d	226	unsigned int ddir_seq_nr;
a6ccc7be	227
e9c047a0	228	/*
8de8f047 JA	229	* IO history logs for verification. We use a tree for sorting,
8de8f047 JA	230	* if we are overwriting. Otherwise just use a fifo.
e9c047a0	231	*/
4b87898e	232	struct rb_root io_hist_tree;
01743ee1	233	struct flist_head io_hist_list;
9e144189	234	unsigned long io_hist_len;
8de8f047 JA	235
	236	/*
	237	* For IO replaying
	238	*/
01743ee1	239	struct flist_head io_log_list;
433afcb4	240
0d29de83 JA	241	/*
	242	* For tracking/handling discards
	243	*/
	244	struct flist_head trim_list;
	245	unsigned long trim_entries;
	246
1907dbc6 JA	247	/*
	248	* for fileservice, how often to switch to a new file
	249	*/
	250	unsigned int file_service_nr;
	251	unsigned int file_service_left;
	252	struct fio_file *file_service_file;
9c60ce64	253
44f29692 JA	254	unsigned int sync_file_range_nr;
44f29692 JA	255
9c60ce64 JA	256	/*
	257	* For generating file sizes
	258	*/
4c07ad86 JA	259	union {
	260	os_random_state_t file_size_state;
	261	struct frand_state __file_size_state;
	262	};
f2bba182 RR	263
	264	/*
	265	* Error counts
	266	*/
	267	unsigned int total_err_count;
	268	int first_error;
15dc1934	269
9e684a49 DE	270	struct fio_flow *flow;
9e684a49 DE	271
15dc1934 JA	272	/*
	273	* Can be overloaded by profiles
	274	*/
7eb36574	275	struct prof_io_ops prof_io_ops;
58c55ba0	276	void *prof_data;
1b79a070 JA	277
1b79a070 JA	278	void *pinned_mem;
ebac4655 JA	279	};
ebac4655 JA	280
e592a06b AC	281	/*
	282	* when should interactive ETA output be generated
	283	*/
	284	enum {
	285	FIO_ETA_AUTO,
	286	FIO_ETA_ALWAYS,
	287	FIO_ETA_NEVER,
	288	};
	289
e1161c32	290	#define __td_verror(td, err, msg, func) \
ebac4655	291	do { \
19abcd3d JA	292	if ((td)->error) \
19abcd3d JA	293	break; \
ebac4655 JA	294	int e = (err); \
ebac4655 JA	295	(td)->error = e; \
f2bba182 RR	296	if (!(td)->first_error) \
f2bba182 RR	297	snprintf(td->verror, sizeof(td->verror) - 1, "file:%s:%d, func=%s, error=%s", __FILE__, __LINE__, (func), (msg)); \
ebac4655 JA	298	} while (0)
ebac4655 JA	299
b990b5c0	300
f2bba182 RR	301	#define td_clear_error(td) \
f2bba182 RR	302	(td)->error = 0;
e1161c32 JA	303	#define td_verror(td, err, func) \
	304	__td_verror((td), (err), strerror((err)), (func))
	305	#define td_vmsg(td, err, msg, func) \
	306	__td_verror((td), (err), (msg), (func))
b990b5c0	307
81179eec JA	308	#define __fio_stringify_1(x) #x
	309	#define __fio_stringify(x) __fio_stringify_1(x)
	310
ebac4655	311	extern int exitall_on_terminate;
11e950bd	312	extern unsigned int thread_number;
ebac4655 JA	313	extern int shm_id;
ebac4655 JA	314	extern int groupid;
f3afa57e	315	extern int output_format;
53cdc686	316	extern int temp_stall_ts;
7078106d	317	extern uintptr_t page_mask, page_size;
4241ea8f	318	extern int read_only;
e592a06b	319	extern int eta_print;
d3eeeabc	320	extern unsigned long done_secs;
01f06b63	321	extern char *job_section;
be4ecfdf JA	322	extern int fio_gtod_offload;
be4ecfdf JA	323	extern int fio_gtod_cpu;
c223da83	324	extern enum fio_cs fio_clock_source;
a9523c6f	325	extern int warnings_fatal;
f57a9c59	326	extern int terse_version;
5c341e9a	327	extern int is_backend;
a37f69b7	328	extern int nr_clients;
e46d8091	329	extern int log_syslog;
5e726d0a	330	extern const char fio_version_string[];
802ad4a8	331	extern const fio_fp64_t def_percentile_list[FIO_IO_U_LIST_MAX_LEN];
ebac4655 JA	332
	333	extern struct thread_data *threads;
	334
7101d9c2 JA	335	static inline void fio_ro_check(struct thread_data td, struct io_u io_u)
	336	{
	337	assert(!(io_u->ddir == DDIR_WRITE && !td_write(td)));
	338	}
	339
0ce8b119	340	#define BLOCKS_PER_MAP (8 * sizeof(unsigned long))
aec2de20 JA	341	#define TO_MAP_BLOCK(f, b) (b)
	342	#define RAND_MAP_IDX(f, b) (TO_MAP_BLOCK(f, b) / BLOCKS_PER_MAP)
	343	#define RAND_MAP_BIT(f, b) (TO_MAP_BLOCK(f, b) & (BLOCKS_PER_MAP - 1))
ebac4655	344
fca70358	345	#define REAL_MAX_JOBS 2048
ebac4655	346
06842027 SL	347	static inline enum error_type td_error_type(enum fio_ddir ddir, int err)
	348	{
	349	if (err == EILSEQ)
8b28bd41	350	return ERROR_TYPE_VERIFY_BIT;
06842027	351	if (ddir == DDIR_READ)
8b28bd41 DM	352	return ERROR_TYPE_READ_BIT;
	353	return ERROR_TYPE_WRITE_BIT;
	354	}
	355
	356	static int __NON_FATAL_ERR[] = {EIO, EILSEQ};
	357	static inline int td_non_fatal_error(struct thread_data *td,
	358	enum error_type_bit etype, int err)
	359	{
	360	int i;
	361	if (!td->o.ignore_error[etype]) {
	362	td->o.ignore_error[etype] = __NON_FATAL_ERR;
	363	td->o.ignore_error_nr[etype] = sizeof(__NON_FATAL_ERR)
	364	/ sizeof(int);
	365	}
	366
	367	if (!(td->o.continue_on_error & (1 << etype)))
	368	return 0;
	369	for (i = 0; i < td->o.ignore_error_nr[etype]; i++)
	370	if (td->o.ignore_error[etype][i] == err)
	371	return 1;
	372	return 0;
06842027 SL	373	}
06842027 SL	374
f2bba182 RR	375	static inline void update_error_count(struct thread_data *td, int err)
	376	{
	377	td->total_err_count++;
	378	if (td->total_err_count == 1)
	379	td->first_error = err;
	380	}
	381
87dc1ab1 JA	382	static inline int should_fsync(struct thread_data *td)
	383	{
	384	if (td->last_was_sync)
	385	return 0;
2dc1bbeb	386	if (td->o.odirect)
87dc1ab1	387	return 0;
2dc1bbeb	388	if (td_write(td) \|\| td_rw(td) \|\| td->o.override_sync)
87dc1ab1 JA	389	return 1;
	390
	391	return 0;
	392	}
	393
8914a9d8	394	/*
214e1eca	395	* Init/option functions
8914a9d8	396	*/
0420ba6a	397	extern int __must_check fio_init_options(void);
72cb971b	398	extern int __must_check parse_options(int, char **);
46bcd498 JA	399	extern int parse_jobs_ini(char *, int, int, int);
46bcd498 JA	400	extern int parse_cmd_line(int, char **, int);
2e1df07d	401	extern int fio_backend(void);
50d16976	402	extern void reset_fio_state(void);
2e1df07d	403	extern void clear_io_state(struct thread_data *);
3b8b7135	404	extern int fio_options_parse(struct thread_data , char *, int);
74929ac2	405	extern void fio_keywords_init(void);
214e1eca	406	extern int fio_cmd_option_parse(struct thread_data , const char , char *);
de890a1e	407	extern int fio_cmd_ioengine_option_parse(struct thread_data , const char , char *);
214e1eca JA	408	extern void fio_fill_default_options(struct thread_data *);
214e1eca JA	409	extern int fio_show_option_help(const char *);
de890a1e	410	extern void fio_options_set_ioengine_opts(struct option long_options, struct thread_data td);
214e1eca	411	extern void fio_options_dup_and_init(struct option *);
de890a1e SL	412	extern void fio_options_mem_dupe(struct thread_data *);
de890a1e SL	413	extern void options_mem_dupe(void data, struct fio_option options);
5bfc35d7	414	extern void td_fill_rand_seeds(struct thread_data *);
46bcd498	415	extern void add_job_opts(const char **, int);
1ec3d69b	416	extern char *num2str(unsigned long, int, int, int);
de890a1e	417	extern int ioengine_load(struct thread_data *);
1ec3d69b	418
e61ca217 JA	419	extern unsigned long page_mask;
	420	extern unsigned long page_size;
	421	extern int initialize_fio(char *envp[]);
	422
de890a1e SL	423	#define FIO_GETOPT_JOB 0x89000000
de890a1e SL	424	#define FIO_GETOPT_IOENGINE 0x98000000
07b3232d	425	#define FIO_NR_OPTIONS (FIO_MAX_OPTS + 128)
8914a9d8	426
263e529f JA	427	/*
	428	* ETA/status stuff
	429	*/
	430	extern void print_thread_status(void);
	431	extern void print_status_init(int);
99d633af	432	extern char *fio_uint_to_kmg(unsigned int val);
263e529f JA	433
	434	/*
	435	* Thread life cycle. Once a thread has a runstate beyond TD_INITIALIZED, it
	436	* will never back again. It may cycle between running/verififying/fsyncing.
	437	* Once the thread reaches TD_EXITED, it is just waiting for the core to
	438	* reap it.
	439	*/
	440	enum {
	441	TD_NOT_CREATED = 0,
	442	TD_CREATED,
	443	TD_INITIALIZED,
b29ee5b3	444	TD_RAMP,
263e529f	445	TD_RUNNING,
b0f65863	446	TD_PRE_READING,
263e529f JA	447	TD_VERIFYING,
	448	TD_FSYNCING,
	449	TD_EXITED,
	450	TD_REAPED,
	451	};
	452
b29ee5b3	453	extern void td_set_runstate(struct thread_data *, int);
cc0df00a JA	454	#define TERMINATE_ALL (-1)
cc0df00a JA	455	extern void fio_terminate_threads(int);
b29ee5b3	456
2f9ade3c JA	457	/*
	458	* Memory helpers
	459	*/
1b79a070 JA	460	extern int __must_check fio_pin_memory(struct thread_data *);
1b79a070 JA	461	extern void fio_unpin_memory(struct thread_data *);
b2fdda43	462	extern int __must_check allocate_io_mem(struct thread_data *);
2f9ade3c	463	extern void free_io_mem(struct thread_data *);
2bb3f0a7	464	extern void free_threads_shm(void);
2f9ade3c	465
b29ee5b3 JA	466	/*
	467	* Reset stats after ramp time completes
	468	*/
	469	extern void reset_all_stats(struct thread_data *);
	470
fb7b71a3 JA	471	/*
	472	* blktrace support
	473	*/
5e62c22a	474	#ifdef FIO_HAVE_BLKTRACE
fb7b71a3 JA	475	extern int is_blktrace(const char *);
fb7b71a3 JA	476	extern int load_blktrace(struct thread_data , const char );
5e62c22a	477	#endif
fb7b71a3	478
2866c82d JA	479	/*
	480	* Mark unused variables passed to ops functions as unused, to silence gcc
	481	*/
	482	#define fio_unused __attribute((__unused__))
5f350952 JA	483	#define fio_init __attribute__((constructor))
5f350952 JA	484	#define fio_exit __attribute__((destructor))
2866c82d	485
34572e28 JA	486	#define for_each_td(td, i) \
34572e28 JA	487	for ((i) = 0, (td) = &threads[0]; (i) < (int) thread_number; (i)++, (td)++)
53cdc686	488	#define for_each_file(td, f, i) \
691c8fb0 JA	489	if ((td)->files_index) \
	490	for ((i) = 0, (f) = (td)->files[0]; \
	491	(i) < (td)->o.nr_files && ((f) = (td)->files[i]) != NULL; \
	492	(i)++)
53cdc686	493
0032bf9f JA	494	#define fio_assert(td, cond) do { \
0032bf9f JA	495	if (!(cond)) { \
340fd243	496	int *__foo = NULL; \
0032bf9f	497	fprintf(stderr, "file:%s:%d, assert %s failed\n", __FILE__, __LINE__, #cond); \
ac18ea38	498	td_set_runstate((td), TD_EXITED); \
437c9b71	499	(td)->error = EFAULT; \
340fd243	500	*__foo = 0; \
0032bf9f JA	501	} \
	502	} while (0)
	503
12d9d841 JA	504	static inline int fio_fill_issue_time(struct thread_data *td)
	505	{
	506	if (td->o.read_iolog_file \|\|
	507	!td->o.disable_clat \|\| !td->o.disable_slat \|\| !td->o.disable_bw)
	508	return 1;
	509
	510	return 0;
	511	}
	512
581e7141 JA	513	static inline int __should_check_rate(struct thread_data *td,
	514	enum fio_ddir ddir)
	515	{
	516	struct thread_options *o = &td->o;
	517
	518	/*
	519	* If some rate setting was given, we need to check it
	520	*/
	521	if (o->rate[ddir] \|\| o->ratemin[ddir] \|\| o->rate_iops[ddir] \|\|
	522	o->rate_iops_min[ddir])
	523	return 1;
	524
	525	return 0;
	526	}
	527
	528	static inline int should_check_rate(struct thread_data *td,
	529	unsigned long *bytes_done)
	530	{
	531	int ret = 0;
	532
6eaf09d6 SL	533	if (bytes_done[DDIR_READ])
	534	ret \|= __should_check_rate(td, DDIR_READ);
	535	if (bytes_done[DDIR_WRITE])
	536	ret \|= __should_check_rate(td, DDIR_WRITE);
	537	if (bytes_done[DDIR_TRIM])
	538	ret \|= __should_check_rate(td, DDIR_TRIM);
581e7141 JA	539
	540	return ret;
	541	}
	542
d529ee19 JA	543	static inline int is_power_of_2(unsigned int val)
	544	{
	545	return (val != 0 && ((val & (val - 1)) == 0));
	546	}
	547
e8462bd8 JA	548	/*
	549	* We currently only need to do locking if we have verifier threads
	550	* accessing our internal structures too
	551	*/
	552	static inline void td_io_u_lock(struct thread_data *td)
	553	{
	554	if (td->o.verify_async)
	555	pthread_mutex_lock(&td->io_u_lock);
	556	}
	557
	558	static inline void td_io_u_unlock(struct thread_data *td)
	559	{
	560	if (td->o.verify_async)
	561	pthread_mutex_unlock(&td->io_u_lock);
	562	}
	563
	564	static inline void td_io_u_free_notify(struct thread_data *td)
	565	{
	566	if (td->o.verify_async)
	567	pthread_cond_signal(&td->free_cond);
	568	}
	569
cca84643 JA	570	extern const char *fio_get_arch_string(int);
	571	extern const char *fio_get_os_string(int);
	572
05775438 JA	573	#define ARRAY_SIZE(x) (sizeof((x)) / (sizeof((x)[0])))
05775438 JA	574
f3afa57e JA	575	enum {
	576	FIO_OUTPUT_TERSE = 0,
	577	FIO_OUTPUT_JSON,
	578	FIO_OUTPUT_NORMAL,
	579	};
	580
ebac4655	581	#endif