[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 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_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;

	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[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;
	};

	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;

	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 {								\
		if ((td)->error)					\
			break;						\
		int e = (err);						\
		(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 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 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 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 REAL_MAX_JOBS		2048

static inline enum error_type_bit 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_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_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 *);
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,
				    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 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	struct thread_data;
	18
	19	#include "compiler/compiler.h"
	20	#include "thread_options.h"
	21	#include "flist.h"
	22	#include "fifo.h"
	23	#include "lib/rbtree.h"
	24	#include "arch/arch.h"
	25	#include "os/os.h"
	26	#include "mutex.h"
	27	#include "log.h"
	28	#include "debug.h"
	29	#include "file.h"
	30	#include "io_ddir.h"
	31	#include "ioengine.h"
	32	#include "iolog.h"
	33	#include "helpers.h"
	34	#include "options.h"
	35	#include "profile.h"
	36	#include "time.h"
	37	#include "gettime.h"
	38	#include "lib/getopt.h"
	39	#include "lib/rand.h"
	40	#include "client.h"
	41	#include "server.h"
	42	#include "stat.h"
	43	#include "flow.h"
	44
	45	#ifdef CONFIG_SOLARISAIO
	46	#include <sys/asynch.h>
	47	#endif
	48
	49	#ifdef CONFIG_LIBNUMA
	50	#include <linux/mempolicy.h>
	51	#include <numa.h>
	52
	53	/*
	54	* "local" is pseudo-policy
	55	*/
	56	#define MPOL_LOCAL MPOL_MAX
	57	#endif
	58
	59	/*
	60	* offset generator types
	61	*/
	62	enum {
	63	RW_SEQ_SEQ = 0,
	64	RW_SEQ_IDENT,
	65	};
	66
	67	enum {
	68	TD_F_VER_BACKLOG = 1,
	69	TD_F_TRIM_BACKLOG = 2,
	70	TD_F_READ_IOLOG = 4,
	71	TD_F_REFILL_BUFFERS = 8,
	72	TD_F_SCRAMBLE_BUFFERS = 16,
	73	TD_F_VER_NONE = 32,
	74	TD_F_PROFILE_OPS = 64,
	75	};
	76
	77	enum {
	78	FIO_RAND_BS_OFF = 0,
	79	FIO_RAND_VER_OFF,
	80	FIO_RAND_MIX_OFF,
	81	FIO_RAND_FILE_OFF,
	82	FIO_RAND_BLOCK_OFF,
	83	FIO_RAND_FILE_SIZE_OFF,
	84	FIO_RAND_TRIM_OFF,
	85	FIO_RAND_BUF_OFF,
	86	FIO_RAND_NR_OFFS,
	87	};
	88
	89	/*
	90	* This describes a single thread/process executing a fio job.
	91	*/
	92	struct thread_data {
	93	struct thread_options o;
	94	unsigned long flags;
	95	void *eo;
	96	char verror[FIO_VERROR_SIZE];
	97	pthread_t thread;
	98	unsigned int thread_number;
	99	unsigned int groupid;
	100	struct thread_stat ts;
	101
	102	int client_type;
	103
	104	struct io_log *slat_log;
	105	struct io_log *clat_log;
	106	struct io_log *lat_log;
	107	struct io_log *bw_log;
	108	struct io_log *iops_log;
	109
	110	uint64_t stat_io_bytes[DDIR_RWDIR_CNT];
	111	struct timeval bw_sample_time;
	112
	113	uint64_t stat_io_blocks[DDIR_RWDIR_CNT];
	114	struct timeval iops_sample_time;
	115
	116	struct rusage ru_start;
	117	struct rusage ru_end;
	118
	119	struct fio_file **files;
	120	unsigned int files_size;
	121	unsigned int files_index;
	122	unsigned int nr_open_files;
	123	unsigned int nr_done_files;
	124	unsigned int nr_normal_files;
	125	union {
	126	unsigned int next_file;
	127	os_random_state_t next_file_state;
	128	struct frand_state __next_file_state;
	129	};
	130	int error;
	131	int sig;
	132	int done;
	133	pid_t pid;
	134	char *orig_buffer;
	135	size_t orig_buffer_size;
	136	volatile int terminate;
	137	volatile int runstate;
	138	unsigned int last_was_sync;
	139	enum fio_ddir last_ddir;
	140
	141	int mmapfd;
	142
	143	void *iolog_buf;
	144	FILE *iolog_f;
	145
	146	char *sysfs_root;
	147
	148	unsigned long rand_seeds[FIO_RAND_NR_OFFS];
	149
	150	union {
	151	os_random_state_t bsrange_state;
	152	struct frand_state __bsrange_state;
	153	};
	154	union {
	155	os_random_state_t verify_state;
	156	struct frand_state __verify_state;
	157	};
	158	union {
	159	os_random_state_t trim_state;
	160	struct frand_state __trim_state;
	161	};
	162
	163	struct frand_state buf_state;
	164
	165	unsigned int verify_batch;
	166	unsigned int trim_batch;
	167
	168	int shm_id;
	169
	170	/*
	171	* IO engine hooks, contains everything needed to submit an io_u
	172	* to any of the available IO engines.
	173	*/
	174	struct ioengine_ops *io_ops;
	175
	176	/*
	177	* Queue depth of io_u's that fio MIGHT do
	178	*/
	179	unsigned int cur_depth;
	180
	181	/*
	182	* io_u's about to be committed
	183	*/
	184	unsigned int io_u_queued;
	185
	186	/*
	187	* io_u's submitted but not completed yet
	188	*/
	189	unsigned int io_u_in_flight;
	190
	191	/*
	192	* List of free and busy io_u's
	193	*/
	194	struct flist_head io_u_freelist;
	195	struct flist_head io_u_busylist;
	196	struct flist_head io_u_requeues;
	197	pthread_mutex_t io_u_lock;
	198	pthread_cond_t free_cond;
	199
	200	/*
	201	* async verify offload
	202	*/
	203	struct flist_head verify_list;
	204	pthread_t *verify_threads;
	205	unsigned int nr_verify_threads;
	206	pthread_cond_t verify_cond;
	207	int verify_thread_exit;
	208
	209	/*
	210	* Rate state
	211	*/
	212	unsigned long long rate_bps[DDIR_RWDIR_CNT];
	213	long rate_pending_usleep[DDIR_RWDIR_CNT];
	214	unsigned long rate_bytes[DDIR_RWDIR_CNT];
	215	unsigned long rate_blocks[DDIR_RWDIR_CNT];
	216	struct timeval lastrate[DDIR_RWDIR_CNT];
	217
	218	unsigned long long total_io_size;
	219	unsigned long long fill_device_size;
	220
	221	unsigned long io_issues[DDIR_RWDIR_CNT];
	222	unsigned long long io_blocks[DDIR_RWDIR_CNT];
	223	unsigned long long this_io_blocks[DDIR_RWDIR_CNT];
	224	unsigned long long io_bytes[DDIR_RWDIR_CNT];
	225	unsigned long long io_skip_bytes;
	226	unsigned long long this_io_bytes[DDIR_RWDIR_CNT];
	227	unsigned long long zone_bytes;
	228	struct fio_mutex *mutex;
	229
	230	/*
	231	* State for random io, a bitmap of blocks done vs not done
	232	*/
	233	union {
	234	os_random_state_t random_state;
	235	struct frand_state __random_state;
	236	};
	237
	238	struct timeval start; /* start of this loop */
	239	struct timeval epoch; /* time job was started */
	240	struct timeval last_issue;
	241	struct timeval tv_cache;
	242	unsigned int tv_cache_nr;
	243	unsigned int tv_cache_mask;
	244	unsigned int ramp_time_over;
	245
	246	/*
	247	* read/write mixed workload state
	248	*/
	249	union {
	250	os_random_state_t rwmix_state;
	251	struct frand_state __rwmix_state;
	252	};
	253	unsigned long rwmix_issues;
	254	enum fio_ddir rwmix_ddir;
	255	unsigned int ddir_seq_nr;
	256
	257	/*
	258	* IO history logs for verification. We use a tree for sorting,
	259	* if we are overwriting. Otherwise just use a fifo.
	260	*/
	261	struct rb_root io_hist_tree;
	262	struct flist_head io_hist_list;
	263	unsigned long io_hist_len;
	264
	265	/*
	266	* For IO replaying
	267	*/
	268	struct flist_head io_log_list;
	269
	270	/*
	271	* For tracking/handling discards
	272	*/
	273	struct flist_head trim_list;
	274	unsigned long trim_entries;
	275
	276	struct flist_head next_rand_list;
	277
	278	/*
	279	* for fileservice, how often to switch to a new file
	280	*/
	281	unsigned int file_service_nr;
	282	unsigned int file_service_left;
	283	struct fio_file *file_service_file;
	284
	285	unsigned int sync_file_range_nr;
	286
	287	/*
	288	* For generating file sizes
	289	*/
	290	union {
	291	os_random_state_t file_size_state;
	292	struct frand_state __file_size_state;
	293	};
	294
	295	/*
	296	* Error counts
	297	*/
	298	unsigned int total_err_count;
	299	int first_error;
	300
	301	struct fio_flow *flow;
	302
	303	/*
	304	* Can be overloaded by profiles
	305	*/
	306	struct prof_io_ops prof_io_ops;
	307	void *prof_data;
	308
	309	void *pinned_mem;
	310	};
	311
	312	/*
	313	* when should interactive ETA output be generated
	314	*/
	315	enum {
	316	FIO_ETA_AUTO,
	317	FIO_ETA_ALWAYS,
	318	FIO_ETA_NEVER,
	319	};
	320
	321	#define __td_verror(td, err, msg, func) \
	322	do { \
	323	if ((td)->error) \
	324	break; \
	325	int e = (err); \
	326	(td)->error = e; \
	327	if (!(td)->first_error) \
	328	snprintf(td->verror, sizeof(td->verror), "file:%s:%d, func=%s, error=%s", __FILE__, __LINE__, (func), (msg)); \
	329	} while (0)
	330
	331
	332	#define td_clear_error(td) \
	333	(td)->error = 0;
	334	#define td_verror(td, err, func) \
	335	__td_verror((td), (err), strerror((err)), (func))
	336	#define td_vmsg(td, err, msg, func) \
	337	__td_verror((td), (err), (msg), (func))
	338
	339	#define __fio_stringify_1(x) #x
	340	#define __fio_stringify(x) __fio_stringify_1(x)
	341
	342	extern int exitall_on_terminate;
	343	extern unsigned int thread_number;
	344	extern unsigned int stat_number;
	345	extern int shm_id;
	346	extern int groupid;
	347	extern int output_format;
	348	extern int temp_stall_ts;
	349	extern uintptr_t page_mask, page_size;
	350	extern int read_only;
	351	extern int eta_print;
	352	extern unsigned long done_secs;
	353	extern char *job_section;
	354	extern int fio_gtod_offload;
	355	extern int fio_gtod_cpu;
	356	extern enum fio_cs fio_clock_source;
	357	extern int fio_clock_source_set;
	358	extern int warnings_fatal;
	359	extern int terse_version;
	360	extern int is_backend;
	361	extern int nr_clients;
	362	extern int log_syslog;
	363	extern const char fio_version_string[];
	364	extern const fio_fp64_t def_percentile_list[FIO_IO_U_LIST_MAX_LEN];
	365
	366	extern struct thread_data *threads;
	367
	368	static inline void fio_ro_check(struct thread_data td, struct io_u io_u)
	369	{
	370	assert(!(io_u->ddir == DDIR_WRITE && !td_write(td)));
	371	}
	372
	373	#define REAL_MAX_JOBS 2048
	374
	375	static inline enum error_type_bit td_error_type(enum fio_ddir ddir, int err)
	376	{
	377	if (err == EILSEQ)
	378	return ERROR_TYPE_VERIFY_BIT;
	379	if (ddir == DDIR_READ)
	380	return ERROR_TYPE_READ_BIT;
	381	return ERROR_TYPE_WRITE_BIT;
	382	}
	383
	384	static int __NON_FATAL_ERR[] = {EIO, EILSEQ};
	385	static inline int td_non_fatal_error(struct thread_data *td,
	386	enum error_type_bit etype, int err)
	387	{
	388	int i;
	389	if (!td->o.ignore_error[etype]) {
	390	td->o.ignore_error[etype] = __NON_FATAL_ERR;
	391	td->o.ignore_error_nr[etype] = sizeof(__NON_FATAL_ERR)
	392	/ sizeof(int);
	393	}
	394
	395	if (!(td->o.continue_on_error & (1 << etype)))
	396	return 0;
	397	for (i = 0; i < td->o.ignore_error_nr[etype]; i++)
	398	if (td->o.ignore_error[etype][i] == err)
	399	return 1;
	400	return 0;
	401	}
	402
	403	static inline void update_error_count(struct thread_data *td, int err)
	404	{
	405	td->total_err_count++;
	406	if (td->total_err_count == 1)
	407	td->first_error = err;
	408	}
	409
	410	static inline int should_fsync(struct thread_data *td)
	411	{
	412	if (td->last_was_sync)
	413	return 0;
	414	if (td_write(td) \|\| td_rw(td) \|\| td->o.override_sync)
	415	return 1;
	416
	417	return 0;
	418	}
	419
	420	/*
	421	* Init/option functions
	422	*/
	423	extern int __must_check fio_init_options(void);
	424	extern int __must_check parse_options(int, char **);
	425	extern int parse_jobs_ini(char *, int, int, int);
	426	extern int parse_cmd_line(int, char **, int);
	427	extern int fio_backend(void);
	428	extern void reset_fio_state(void);
	429	extern void clear_io_state(struct thread_data *);
	430	extern int fio_options_parse(struct thread_data , char *, int);
	431	extern void fio_keywords_init(void);
	432	extern int fio_cmd_option_parse(struct thread_data , const char , char *);
	433	extern int fio_cmd_ioengine_option_parse(struct thread_data , const char , char *);
	434	extern void fio_fill_default_options(struct thread_data *);
	435	extern int fio_show_option_help(const char *);
	436	extern void fio_options_set_ioengine_opts(struct option long_options, struct thread_data td);
	437	extern void fio_options_dup_and_init(struct option *);
	438	extern void fio_options_mem_dupe(struct thread_data *);
	439	extern void options_mem_dupe(void data, struct fio_option options);
	440	extern void td_fill_rand_seeds(struct thread_data *);
	441	extern void add_job_opts(const char **, int);
	442	extern char *num2str(unsigned long, int, int, int);
	443	extern int ioengine_load(struct thread_data *);
	444
	445	extern unsigned long page_mask;
	446	extern unsigned long page_size;
	447	extern int initialize_fio(char *envp[]);
	448
	449	#define FIO_GETOPT_JOB 0x89000000
	450	#define FIO_GETOPT_IOENGINE 0x98000000
	451	#define FIO_NR_OPTIONS (FIO_MAX_OPTS + 128)
	452
	453	/*
	454	* ETA/status stuff
	455	*/
	456	extern void print_thread_status(void);
	457	extern void print_status_init(int);
	458	extern char *fio_uint_to_kmg(unsigned int val);
	459
	460	/*
	461	* Thread life cycle. Once a thread has a runstate beyond TD_INITIALIZED, it
	462	* will never back again. It may cycle between running/verififying/fsyncing.
	463	* Once the thread reaches TD_EXITED, it is just waiting for the core to
	464	* reap it.
	465	*/
	466	enum {
	467	TD_NOT_CREATED = 0,
	468	TD_CREATED,
	469	TD_INITIALIZED,
	470	TD_RAMP,
	471	TD_SETTING_UP,
	472	TD_RUNNING,
	473	TD_PRE_READING,
	474	TD_VERIFYING,
	475	TD_FSYNCING,
	476	TD_EXITED,
	477	TD_REAPED,
	478	};
	479
	480	extern void td_set_runstate(struct thread_data *, int);
	481	#define TERMINATE_ALL (-1)
	482	extern void fio_terminate_threads(int);
	483
	484	/*
	485	* Memory helpers
	486	*/
	487	extern int __must_check fio_pin_memory(struct thread_data *);
	488	extern void fio_unpin_memory(struct thread_data *);
	489	extern int __must_check allocate_io_mem(struct thread_data *);
	490	extern void free_io_mem(struct thread_data *);
	491	extern void free_threads_shm(void);
	492
	493	/*
	494	* Reset stats after ramp time completes
	495	*/
	496	extern void reset_all_stats(struct thread_data *);
	497
	498	/*
	499	* blktrace support
	500	*/
	501	#ifdef FIO_HAVE_BLKTRACE
	502	extern int is_blktrace(const char *);
	503	extern int load_blktrace(struct thread_data , const char );
	504	#endif
	505
	506	/*
	507	* Mark unused variables passed to ops functions as unused, to silence gcc
	508	*/
	509	#define fio_unused __attribute((__unused__))
	510	#define fio_init __attribute__((constructor))
	511	#define fio_exit __attribute__((destructor))
	512
	513	#define for_each_td(td, i) \
	514	for ((i) = 0, (td) = &threads[0]; (i) < (int) thread_number; (i)++, (td)++)
	515	#define for_each_file(td, f, i) \
	516	if ((td)->files_index) \
	517	for ((i) = 0, (f) = (td)->files[0]; \
	518	(i) < (td)->o.nr_files && ((f) = (td)->files[i]) != NULL; \
	519	(i)++)
	520
	521	#define fio_assert(td, cond) do { \
	522	if (!(cond)) { \
	523	int *__foo = NULL; \
	524	fprintf(stderr, "file:%s:%d, assert %s failed\n", __FILE__, __LINE__, #cond); \
	525	td_set_runstate((td), TD_EXITED); \
	526	(td)->error = EFAULT; \
	527	*__foo = 0; \
	528	} \
	529	} while (0)
	530
	531	static inline int fio_fill_issue_time(struct thread_data *td)
	532	{
	533	if (td->o.read_iolog_file \|\|
	534	!td->o.disable_clat \|\| !td->o.disable_slat \|\| !td->o.disable_bw)
	535	return 1;
	536
	537	return 0;
	538	}
	539
	540	static inline int __should_check_rate(struct thread_data *td,
	541	enum fio_ddir ddir)
	542	{
	543	struct thread_options *o = &td->o;
	544
	545	/*
	546	* If some rate setting was given, we need to check it
	547	*/
	548	if (o->rate[ddir] \|\| o->ratemin[ddir] \|\| o->rate_iops[ddir] \|\|
	549	o->rate_iops_min[ddir])
	550	return 1;
	551
	552	return 0;
	553	}
	554
	555	static inline int should_check_rate(struct thread_data *td,
	556	uint64_t *bytes_done)
	557	{
	558	int ret = 0;
	559
	560	if (bytes_done[DDIR_READ])
	561	ret \|= __should_check_rate(td, DDIR_READ);
	562	if (bytes_done[DDIR_WRITE])
	563	ret \|= __should_check_rate(td, DDIR_WRITE);
	564	if (bytes_done[DDIR_TRIM])
	565	ret \|= __should_check_rate(td, DDIR_TRIM);
	566
	567	return ret;
	568	}
	569
	570	static inline int is_power_of_2(unsigned int val)
	571	{
	572	return (val != 0 && ((val & (val - 1)) == 0));
	573	}
	574
	575	/*
	576	* We currently only need to do locking if we have verifier threads
	577	* accessing our internal structures too
	578	*/
	579	static inline void td_io_u_lock(struct thread_data *td)
	580	{
	581	if (td->o.verify_async)
	582	pthread_mutex_lock(&td->io_u_lock);
	583	}
	584
	585	static inline void td_io_u_unlock(struct thread_data *td)
	586	{
	587	if (td->o.verify_async)
	588	pthread_mutex_unlock(&td->io_u_lock);
	589	}
	590
	591	static inline void td_io_u_free_notify(struct thread_data *td)
	592	{
	593	if (td->o.verify_async)
	594	pthread_cond_signal(&td->free_cond);
	595	}
	596
	597	extern const char *fio_get_arch_string(int);
	598	extern const char *fio_get_os_string(int);
	599
	600	#define ARRAY_SIZE(x) (sizeof((x)) / (sizeof((x)[0])))
	601
	602	enum {
	603	FIO_OUTPUT_TERSE = 0,
	604	FIO_OUTPUT_JSON,
	605	FIO_OUTPUT_NORMAL,
	606	};
	607
	608	enum {
	609	FIO_RAND_DIST_RANDOM = 0,
	610	FIO_RAND_DIST_ZIPF,
	611	FIO_RAND_DIST_PARETO,
	612	};
	613
	614	enum {
	615	FIO_RAND_GEN_TAUSWORTHE = 0,
	616	FIO_RAND_GEN_LFSR,
	617	};
	618
	619	#endif