[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;

#define FIO_MAJOR	1
#define FIO_MINOR	58
#define FIO_PATCH	0

#include "compiler/compiler.h"
#include "flist.h"
#include "fifo.h"
#include "rbtree.h"
#include "arch/arch.h"
#include "os/os.h"
#include "mutex.h"
#include "log.h"
#include "debug.h"
#include "file.h"
#include "io_ddir.h"
#include "ioengine.h"
#include "iolog.h"
#include "helpers.h"
#include "options.h"
#include "profile.h"
#include "time.h"
#include "lib/getopt.h"
#include "lib/rand.h"
#include "server.h"
#include "stat.h"

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

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

/*
 * What type of allocation to use for io buffers
 */
enum fio_memtype {
	MEM_MALLOC = 0,	/* ordinary malloc */
	MEM_SHM,	/* use shared memory segments */
	MEM_SHMHUGE,	/* use shared memory segments with huge pages */
	MEM_MMAP,	/* use anonynomous mmap */
	MEM_MMAPHUGE,	/* memory mapped huge file */
};

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

struct bssplit {
	unsigned int bs;
	unsigned char perc;
};

struct thread_options {
	int pad;
	char *description;
	char *name;
	char *directory;
	char *filename;
	char *opendir;
	char *ioengine;
	enum td_ddir td_ddir;
	unsigned int rw_seq;
	unsigned int kb_base;
	unsigned int ddir_seq_nr;
	long ddir_seq_add;
	unsigned int iodepth;
	unsigned int iodepth_low;
	unsigned int iodepth_batch;
	unsigned int iodepth_batch_complete;

	unsigned long long size;
	unsigned int size_percent;
	unsigned int fill_device;
	unsigned long long file_size_low;
	unsigned long long file_size_high;
	unsigned long long start_offset;

	unsigned int bs[2];
	unsigned int ba[2];
	unsigned int min_bs[2];
	unsigned int max_bs[2];
	struct bssplit *bssplit[2];
	unsigned int bssplit_nr[2];

	unsigned int nr_files;
	unsigned int open_files;
	enum file_lock_mode file_lock_mode;
	unsigned int lockfile_batch;

	unsigned int odirect;
	unsigned int invalidate_cache;
	unsigned int create_serialize;
	unsigned int create_fsync;
	unsigned int create_on_open;
	unsigned int end_fsync;
	unsigned int pre_read;
	unsigned int sync_io;
	unsigned int verify;
	unsigned int do_verify;
	unsigned int verifysort;
	unsigned int verify_interval;
	unsigned int verify_offset;
	char verify_pattern[MAX_PATTERN_SIZE];
	unsigned int verify_pattern_bytes;
	unsigned int verify_fatal;
	unsigned int verify_dump;
	unsigned int verify_async;
	unsigned long long verify_backlog;
	unsigned int verify_batch;
	unsigned int use_thread;
	unsigned int unlink;
	unsigned int do_disk_util;
	unsigned int override_sync;
	unsigned int rand_repeatable;
	unsigned int use_os_rand;
	unsigned int write_lat_log;
	unsigned int write_bw_log;
	unsigned int norandommap;
	unsigned int softrandommap;
	unsigned int bs_unaligned;
	unsigned int fsync_on_close;

	unsigned int hugepage_size;
	unsigned int rw_min_bs;
	unsigned int thinktime;
	unsigned int thinktime_spin;
	unsigned int thinktime_blocks;
	unsigned int fsync_blocks;
	unsigned int fdatasync_blocks;
	unsigned int barrier_blocks;
	unsigned long long start_delay;
	unsigned long long timeout;
	unsigned long long ramp_time;
	unsigned int overwrite;
	unsigned int bw_avg_time;
	unsigned int loops;
	unsigned long long zone_size;
	unsigned long long zone_skip;
	enum fio_memtype mem_type;
	unsigned int mem_align;

	unsigned int stonewall;
	unsigned int new_group;
	unsigned int numjobs;
	os_cpu_mask_t cpumask;
	unsigned int cpumask_set;
	os_cpu_mask_t verify_cpumask;
	unsigned int verify_cpumask_set;
	unsigned int iolog;
	unsigned int rwmixcycle;
	unsigned int rwmix[2];
	unsigned int nice;
	unsigned int file_service_type;
	unsigned int group_reporting;
	unsigned int fadvise_hint;
	enum fio_fallocate_mode fallocate_mode;
	unsigned int zero_buffers;
	unsigned int refill_buffers;
	unsigned int scramble_buffers;
	unsigned int time_based;
	unsigned int disable_lat;
	unsigned int disable_clat;
	unsigned int disable_slat;
	unsigned int disable_bw;
	unsigned int gtod_reduce;
	unsigned int gtod_cpu;
	unsigned int gtod_offload;
	enum fio_cs clocksource;
	unsigned int no_stall;
	unsigned int trim_percentage;
	unsigned int trim_batch;
	unsigned int trim_zero;
	unsigned long long trim_backlog;
	unsigned int clat_percentiles;
	unsigned int overwrite_plist;
	double percentile_list[FIO_IO_U_LIST_MAX_LEN];

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

	/*
	 * Pre-run and post-run shell
	 */
	char *exec_prerun;
	char *exec_postrun;

	unsigned int rate[2];
	unsigned int ratemin[2];
	unsigned int ratecycle;
	unsigned int rate_iops[2];
	unsigned int rate_iops_min[2];

	char *ioscheduler;

	/*
	 * CPU "io" cycle burner
	 */
	unsigned int cpuload;
	unsigned int cpucycle;

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

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

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

	unsigned int uid;
	unsigned int gid;

	unsigned int sync_file_range;

	unsigned int userspace_libaio_reap;
};

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

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

	uint64_t stat_io_bytes[2];
	struct timeval stat_sample_time[2];

	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 done;
	pid_t pid;
	char *orig_buffer;
	size_t orig_buffer_size;
	volatile int terminate;
	volatile int runstate;
	unsigned int ioprio;
	unsigned int ioprio_set;
	unsigned int last_was_sync;
	enum fio_ddir last_ddir;

	char *mmapfile;
	int mmapfd;

	void *iolog_buf;
	FILE *iolog_f;

	char *sysfs_root;

	unsigned long rand_seeds[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;

	/*
	 * Current IO depth and list of free and busy io_u's.
	 */
	unsigned int cur_depth;
	unsigned int io_u_queued;
	struct flist_head io_u_freelist;
	struct flist_head io_u_busylist;
	struct flist_head io_u_requeues;
	pthread_mutex_t io_u_lock;
	pthread_cond_t free_cond;

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

	/*
	 * Rate state
	 */
	unsigned long rate_nsec_cycle[2];
	long rate_pending_usleep[2];
	unsigned long rate_bytes[2];
	unsigned long rate_blocks[2];
	struct timeval lastrate[2];

	unsigned long long total_io_size;
	unsigned long long fill_device_size;

	unsigned long io_issues[2];
	unsigned long long io_blocks[2];
	unsigned long long io_bytes[2];
	unsigned long long io_skip_bytes;
	unsigned long long this_io_bytes[2];
	unsigned long long zone_bytes;
	struct fio_mutex *mutex;

	/*
	 * State for random io, a bitmap of blocks done vs not done
	 */
	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;

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

/*
 * when should interactive ETA output be generated
 */
enum {
	FIO_ETA_AUTO,
	FIO_ETA_ALWAYS,
	FIO_ETA_NEVER,
};

#define __td_verror(td, err, msg, func)					\
	do {								\
		if ((td)->error)					\
			break;						\
		int e = (err);						\
		(td)->error = e;					\
		if (!(td)->first_error)					\
			snprintf(td->verror, sizeof(td->verror) - 1, "file:%s:%d, func=%s, error=%s", __FILE__, __LINE__, (func), (msg));		\
	} while (0)


#define td_clear_error(td)		\
	(td)->error = 0;
#define td_verror(td, err, func)	\
	__td_verror((td), (err), strerror((err)), (func))
#define td_vmsg(td, err, msg, func)	\
	__td_verror((td), (err), (msg), (func))

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

extern int exitall_on_terminate;
extern int thread_number;
extern int nr_process, nr_thread;
extern int shm_id;
extern int groupid;
extern int terse_output;
extern int temp_stall_ts;
extern unsigned long long mlock_size;
extern unsigned long page_mask, page_size;
extern int read_only;
extern int eta_print;
extern unsigned long done_secs;
extern char *job_section;
extern int fio_gtod_offload;
extern int fio_gtod_cpu;
extern enum fio_cs fio_clock_source;
extern int warnings_fatal;
extern int terse_version;
extern int is_backend;
extern int nr_clients;
extern int log_syslog;

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

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

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

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

	return 0;
}

/*
 * Init/option functions
 */
extern int __must_check parse_options(int, char **);
extern int parse_jobs_ini(char *, int, int);
extern int parse_cmd_line(int, char **);
extern int exec_run(void);
extern void reset_fio_state(void);
extern int fio_options_parse(struct thread_data *, char **, int);
extern void fio_keywords_init(void);
extern int fio_cmd_option_parse(struct thread_data *, const char *, char *);
extern void fio_fill_default_options(struct thread_data *);
extern int fio_show_option_help(const char *);
extern void fio_options_dup_and_init(struct option *);
extern void options_mem_dupe(struct thread_data *);
extern void options_mem_free(struct thread_data *);
extern void td_fill_rand_seeds(struct thread_data *);
extern void add_job_opts(const char **);
extern char *num2str(unsigned long, int, int, int);

#define FIO_GETOPT_JOB		0x89988998
#define FIO_NR_OPTIONS		(FIO_MAX_OPTS + 128)

/*
 * ETA/status stuff
 */
extern void print_thread_status(void);
extern void print_status_init(int);

/*
 * Thread life cycle. Once a thread has a runstate beyond TD_INITIALIZED, it
 * will never back again. It may cycle between running/verififying/fsyncing.
 * Once the thread reaches TD_EXITED, it is just waiting for the core to
 * reap it.
 */
enum {
	TD_NOT_CREATED = 0,
	TD_CREATED,
	TD_INITIALIZED,
	TD_RAMP,
	TD_RUNNING,
	TD_PRE_READING,
	TD_VERIFYING,
	TD_FSYNCING,
	TD_EXITED,
	TD_REAPED,
};

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

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

/*
 * Reset stats after ramp time completes
 */
extern void reset_all_stats(struct thread_data *);

/*
 * blktrace support
 */
#ifdef FIO_HAVE_BLKTRACE
extern int is_blktrace(const char *);
extern int load_blktrace(struct thread_data *, const char *);
#endif

/*
 * Mark unused variables passed to ops functions as unused, to silence gcc
 */
#define fio_unused	__attribute((__unused__))
#define fio_init	__attribute__((constructor))
#define fio_exit	__attribute__((destructor))

#define for_each_td(td, i)	\
	for ((i) = 0, (td) = &threads[0]; (i) < (int) thread_number; (i)++, (td)++)
#define for_each_file(td, f, i)	\
	if ((td)->files_index)						\
		for ((i) = 0, (f) = (td)->files[0];			\
	    	 (i) < (td)->o.nr_files && ((f) = (td)->files[i]) != NULL; \
		 (i)++)

#define fio_assert(td, cond)	do {	\
	if (!(cond)) {			\
		int *__foo = NULL;	\
		fprintf(stderr, "file:%s:%d, assert %s failed\n", __FILE__, __LINE__, #cond);	\
		td_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[0])
		ret |= __should_check_rate(td, 0);
	if (bytes_done[1])
		ret |= __should_check_rate(td, 1);

	return ret;
}

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

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

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

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

#endif
Commit	Line	Data
	1	#ifndef FIO_H
	2	#define FIO_H
	3
	4	#include <sched.h>
	5	#include <limits.h>
	6	#include <pthread.h>
	7	#include <sys/time.h>
	8	#include <sys/resource.h>
	9	#include <errno.h>
	10	#include <stdlib.h>
	11	#include <stdio.h>
	12	#include <unistd.h>
	13	#include <string.h>
	14	#include <inttypes.h>
	15	#include <assert.h>
	16
	17	struct thread_data;
	18
	19	#define FIO_MAJOR 1
	20	#define FIO_MINOR 58
	21	#define FIO_PATCH 0
	22
	23	#include "compiler/compiler.h"
	24	#include "flist.h"
	25	#include "fifo.h"
	26	#include "rbtree.h"
	27	#include "arch/arch.h"
	28	#include "os/os.h"
	29	#include "mutex.h"
	30	#include "log.h"
	31	#include "debug.h"
	32	#include "file.h"
	33	#include "io_ddir.h"
	34	#include "ioengine.h"
	35	#include "iolog.h"
	36	#include "helpers.h"
	37	#include "options.h"
	38	#include "profile.h"
	39	#include "time.h"
	40	#include "lib/getopt.h"
	41	#include "lib/rand.h"
	42	#include "server.h"
	43	#include "stat.h"
	44
	45	#ifdef FIO_HAVE_GUASI
	46	#include <guasi.h>
	47	#endif
	48
	49	#ifdef FIO_HAVE_SOLARISAIO
	50	#include <sys/asynch.h>
	51	#endif
	52
	53	/*
	54	* What type of allocation to use for io buffers
	55	*/
	56	enum fio_memtype {
	57	MEM_MALLOC = 0, /* ordinary malloc */
	58	MEM_SHM, /* use shared memory segments */
	59	MEM_SHMHUGE, /* use shared memory segments with huge pages */
	60	MEM_MMAP, /* use anonynomous mmap */
	61	MEM_MMAPHUGE, /* memory mapped huge file */
	62	};
	63
	64	/*
	65	* offset generator types
	66	*/
	67	enum {
	68	RW_SEQ_SEQ = 0,
	69	RW_SEQ_IDENT,
	70	};
	71
	72	struct bssplit {
	73	unsigned int bs;
	74	unsigned char perc;
	75	};
	76
	77	struct thread_options {
	78	int pad;
	79	char *description;
	80	char *name;
	81	char *directory;
	82	char *filename;
	83	char *opendir;
	84	char *ioengine;
	85	enum td_ddir td_ddir;
	86	unsigned int rw_seq;
	87	unsigned int kb_base;
	88	unsigned int ddir_seq_nr;
	89	long ddir_seq_add;
	90	unsigned int iodepth;
	91	unsigned int iodepth_low;
	92	unsigned int iodepth_batch;
	93	unsigned int iodepth_batch_complete;
	94
	95	unsigned long long size;
	96	unsigned int size_percent;
	97	unsigned int fill_device;
	98	unsigned long long file_size_low;
	99	unsigned long long file_size_high;
	100	unsigned long long start_offset;
	101
	102	unsigned int bs[2];
	103	unsigned int ba[2];
	104	unsigned int min_bs[2];
	105	unsigned int max_bs[2];
	106	struct bssplit *bssplit[2];
	107	unsigned int bssplit_nr[2];
	108
	109	unsigned int nr_files;
	110	unsigned int open_files;
	111	enum file_lock_mode file_lock_mode;
	112	unsigned int lockfile_batch;
	113
	114	unsigned int odirect;
	115	unsigned int invalidate_cache;
	116	unsigned int create_serialize;
	117	unsigned int create_fsync;
	118	unsigned int create_on_open;
	119	unsigned int end_fsync;
	120	unsigned int pre_read;
	121	unsigned int sync_io;
	122	unsigned int verify;
	123	unsigned int do_verify;
	124	unsigned int verifysort;
	125	unsigned int verify_interval;
	126	unsigned int verify_offset;
	127	char verify_pattern[MAX_PATTERN_SIZE];
	128	unsigned int verify_pattern_bytes;
	129	unsigned int verify_fatal;
	130	unsigned int verify_dump;
	131	unsigned int verify_async;
	132	unsigned long long verify_backlog;
	133	unsigned int verify_batch;
	134	unsigned int use_thread;
	135	unsigned int unlink;
	136	unsigned int do_disk_util;
	137	unsigned int override_sync;
	138	unsigned int rand_repeatable;
	139	unsigned int use_os_rand;
	140	unsigned int write_lat_log;
	141	unsigned int write_bw_log;
	142	unsigned int norandommap;
	143	unsigned int softrandommap;
	144	unsigned int bs_unaligned;
	145	unsigned int fsync_on_close;
	146
	147	unsigned int hugepage_size;
	148	unsigned int rw_min_bs;
	149	unsigned int thinktime;
	150	unsigned int thinktime_spin;
	151	unsigned int thinktime_blocks;
	152	unsigned int fsync_blocks;
	153	unsigned int fdatasync_blocks;
	154	unsigned int barrier_blocks;
	155	unsigned long long start_delay;
	156	unsigned long long timeout;
	157	unsigned long long ramp_time;
	158	unsigned int overwrite;
	159	unsigned int bw_avg_time;
	160	unsigned int loops;
	161	unsigned long long zone_size;
	162	unsigned long long zone_skip;
	163	enum fio_memtype mem_type;
	164	unsigned int mem_align;
	165
	166	unsigned int stonewall;
	167	unsigned int new_group;
	168	unsigned int numjobs;
	169	os_cpu_mask_t cpumask;
	170	unsigned int cpumask_set;
	171	os_cpu_mask_t verify_cpumask;
	172	unsigned int verify_cpumask_set;
	173	unsigned int iolog;
	174	unsigned int rwmixcycle;
	175	unsigned int rwmix[2];
	176	unsigned int nice;
	177	unsigned int file_service_type;
	178	unsigned int group_reporting;
	179	unsigned int fadvise_hint;
	180	enum fio_fallocate_mode fallocate_mode;
	181	unsigned int zero_buffers;
	182	unsigned int refill_buffers;
	183	unsigned int scramble_buffers;
	184	unsigned int time_based;
	185	unsigned int disable_lat;
	186	unsigned int disable_clat;
	187	unsigned int disable_slat;
	188	unsigned int disable_bw;
	189	unsigned int gtod_reduce;
	190	unsigned int gtod_cpu;
	191	unsigned int gtod_offload;
	192	enum fio_cs clocksource;
	193	unsigned int no_stall;
	194	unsigned int trim_percentage;
	195	unsigned int trim_batch;
	196	unsigned int trim_zero;
	197	unsigned long long trim_backlog;
	198	unsigned int clat_percentiles;
	199	unsigned int overwrite_plist;
	200	double percentile_list[FIO_IO_U_LIST_MAX_LEN];
	201
	202	char *read_iolog_file;
	203	char *write_iolog_file;
	204	char *bw_log_file;
	205	char *lat_log_file;
	206	char *replay_redirect;
	207
	208	/*
	209	* Pre-run and post-run shell
	210	*/
	211	char *exec_prerun;
	212	char *exec_postrun;
	213
	214	unsigned int rate[2];
	215	unsigned int ratemin[2];
	216	unsigned int ratecycle;
	217	unsigned int rate_iops[2];
	218	unsigned int rate_iops_min[2];
	219
	220	char *ioscheduler;
	221
	222	/*
	223	* CPU "io" cycle burner
	224	*/
	225	unsigned int cpuload;
	226	unsigned int cpucycle;
	227
	228	/*
	229	* I/O Error handling
	230	*/
	231	unsigned int continue_on_error;
	232
	233	/*
	234	* Benchmark profile type
	235	*/
	236	char *profile;
	237
	238	/*
	239	* blkio cgroup support
	240	*/
	241	char *cgroup;
	242	unsigned int cgroup_weight;
	243	unsigned int cgroup_nodelete;
	244
	245	unsigned int uid;
	246	unsigned int gid;
	247
	248	unsigned int sync_file_range;
	249
	250	unsigned int userspace_libaio_reap;
	251	};
	252
	253	/*
	254	* This describes a single thread/process executing a fio job.
	255	*/
	256	struct thread_data {
	257	struct thread_options o;
	258	char verror[FIO_VERROR_SIZE];
	259	pthread_t thread;
	260	int thread_number;
	261	int groupid;
	262	struct thread_stat ts;
	263
	264	struct io_log *slat_log;
	265	struct io_log *clat_log;
	266	struct io_log *lat_log;
	267	struct io_log *bw_log;
	268
	269	uint64_t stat_io_bytes[2];
	270	struct timeval stat_sample_time[2];
	271
	272	struct rusage ru_start;
	273	struct rusage ru_end;
	274
	275	struct fio_file **files;
	276	unsigned int files_size;
	277	unsigned int files_index;
	278	unsigned int nr_open_files;
	279	unsigned int nr_done_files;
	280	unsigned int nr_normal_files;
	281	union {
	282	unsigned int next_file;
	283	os_random_state_t next_file_state;
	284	struct frand_state __next_file_state;
	285	};
	286	int error;
	287	int done;
	288	pid_t pid;
	289	char *orig_buffer;
	290	size_t orig_buffer_size;
	291	volatile int terminate;
	292	volatile int runstate;
	293	unsigned int ioprio;
	294	unsigned int ioprio_set;
	295	unsigned int last_was_sync;
	296	enum fio_ddir last_ddir;
	297
	298	char *mmapfile;
	299	int mmapfd;
	300
	301	void *iolog_buf;
	302	FILE *iolog_f;
	303
	304	char *sysfs_root;
	305
	306	unsigned long rand_seeds[8];
	307
	308	union {
	309	os_random_state_t bsrange_state;
	310	struct frand_state __bsrange_state;
	311	};
	312	union {
	313	os_random_state_t verify_state;
	314	struct frand_state __verify_state;
	315	};
	316	union {
	317	os_random_state_t trim_state;
	318	struct frand_state __trim_state;
	319	};
	320
	321	struct frand_state buf_state;
	322
	323	unsigned int verify_batch;
	324	unsigned int trim_batch;
	325
	326	int shm_id;
	327
	328	/*
	329	* IO engine hooks, contains everything needed to submit an io_u
	330	* to any of the available IO engines.
	331	*/
	332	struct ioengine_ops *io_ops;
	333
	334	/*
	335	* Current IO depth and list of free and busy io_u's.
	336	*/
	337	unsigned int cur_depth;
	338	unsigned int io_u_queued;
	339	struct flist_head io_u_freelist;
	340	struct flist_head io_u_busylist;
	341	struct flist_head io_u_requeues;
	342	pthread_mutex_t io_u_lock;
	343	pthread_cond_t free_cond;
	344
	345	/*
	346	* async verify offload
	347	*/
	348	struct flist_head verify_list;
	349	pthread_t *verify_threads;
	350	unsigned int nr_verify_threads;
	351	pthread_cond_t verify_cond;
	352	int verify_thread_exit;
	353
	354	/*
	355	* Rate state
	356	*/
	357	unsigned long rate_nsec_cycle[2];
	358	long rate_pending_usleep[2];
	359	unsigned long rate_bytes[2];
	360	unsigned long rate_blocks[2];
	361	struct timeval lastrate[2];
	362
	363	unsigned long long total_io_size;
	364	unsigned long long fill_device_size;
	365
	366	unsigned long io_issues[2];
	367	unsigned long long io_blocks[2];
	368	unsigned long long io_bytes[2];
	369	unsigned long long io_skip_bytes;
	370	unsigned long long this_io_bytes[2];
	371	unsigned long long zone_bytes;
	372	struct fio_mutex *mutex;
	373
	374	/*
	375	* State for random io, a bitmap of blocks done vs not done
	376	*/
	377	union {
	378	os_random_state_t random_state;
	379	struct frand_state __random_state;
	380	};
	381
	382	struct timeval start; /* start of this loop */
	383	struct timeval epoch; /* time job was started */
	384	struct timeval last_issue;
	385	struct timeval tv_cache;
	386	unsigned int tv_cache_nr;
	387	unsigned int tv_cache_mask;
	388	unsigned int ramp_time_over;
	389
	390	/*
	391	* read/write mixed workload state
	392	*/
	393	union {
	394	os_random_state_t rwmix_state;
	395	struct frand_state __rwmix_state;
	396	};
	397	unsigned long rwmix_issues;
	398	enum fio_ddir rwmix_ddir;
	399	unsigned int ddir_seq_nr;
	400
	401	/*
	402	* IO history logs for verification. We use a tree for sorting,
	403	* if we are overwriting. Otherwise just use a fifo.
	404	*/
	405	struct rb_root io_hist_tree;
	406	struct flist_head io_hist_list;
	407	unsigned long io_hist_len;
	408
	409	/*
	410	* For IO replaying
	411	*/
	412	struct flist_head io_log_list;
	413
	414	/*
	415	* For tracking/handling discards
	416	*/
	417	struct flist_head trim_list;
	418	unsigned long trim_entries;
	419
	420	/*
	421	* for fileservice, how often to switch to a new file
	422	*/
	423	unsigned int file_service_nr;
	424	unsigned int file_service_left;
	425	struct fio_file *file_service_file;
	426
	427	unsigned int sync_file_range_nr;
	428
	429	/*
	430	* For generating file sizes
	431	*/
	432	union {
	433	os_random_state_t file_size_state;
	434	struct frand_state __file_size_state;
	435	};
	436
	437	/*
	438	* Error counts
	439	*/
	440	unsigned int total_err_count;
	441	int first_error;
	442
	443	/*
	444	* Can be overloaded by profiles
	445	*/
	446	struct prof_io_ops prof_io_ops;
	447	void *prof_data;
	448	};
	449
	450	/*
	451	* when should interactive ETA output be generated
	452	*/
	453	enum {
	454	FIO_ETA_AUTO,
	455	FIO_ETA_ALWAYS,
	456	FIO_ETA_NEVER,
	457	};
	458
	459	#define __td_verror(td, err, msg, func) \
	460	do { \
	461	if ((td)->error) \
	462	break; \
	463	int e = (err); \
	464	(td)->error = e; \
	465	if (!(td)->first_error) \
	466	snprintf(td->verror, sizeof(td->verror) - 1, "file:%s:%d, func=%s, error=%s", __FILE__, __LINE__, (func), (msg)); \
	467	} while (0)
	468
	469
	470	#define td_clear_error(td) \
	471	(td)->error = 0;
	472	#define td_verror(td, err, func) \
	473	__td_verror((td), (err), strerror((err)), (func))
	474	#define td_vmsg(td, err, msg, func) \
	475	__td_verror((td), (err), (msg), (func))
	476
	477	#define __fio_stringify_1(x) #x
	478	#define __fio_stringify(x) __fio_stringify_1(x)
	479
	480	extern int exitall_on_terminate;
	481	extern int thread_number;
	482	extern int nr_process, nr_thread;
	483	extern int shm_id;
	484	extern int groupid;
	485	extern int terse_output;
	486	extern int temp_stall_ts;
	487	extern unsigned long long mlock_size;
	488	extern unsigned long page_mask, page_size;
	489	extern int read_only;
	490	extern int eta_print;
	491	extern unsigned long done_secs;
	492	extern char *job_section;
	493	extern int fio_gtod_offload;
	494	extern int fio_gtod_cpu;
	495	extern enum fio_cs fio_clock_source;
	496	extern int warnings_fatal;
	497	extern int terse_version;
	498	extern int is_backend;
	499	extern int nr_clients;
	500	extern int log_syslog;
	501
	502	extern struct thread_data *threads;
	503
	504	static inline void fio_ro_check(struct thread_data td, struct io_u io_u)
	505	{
	506	assert(!(io_u->ddir == DDIR_WRITE && !td_write(td)));
	507	}
	508
	509	#define BLOCKS_PER_MAP (8 * sizeof(unsigned long))
	510	#define TO_MAP_BLOCK(f, b) (b)
	511	#define RAND_MAP_IDX(f, b) (TO_MAP_BLOCK(f, b) / BLOCKS_PER_MAP)
	512	#define RAND_MAP_BIT(f, b) (TO_MAP_BLOCK(f, b) & (BLOCKS_PER_MAP - 1))
	513
	514	#define REAL_MAX_JOBS 2048
	515
	516	#define td_non_fatal_error(e) ((e) == EIO \|\| (e) == EILSEQ)
	517
	518	static inline void update_error_count(struct thread_data *td, int err)
	519	{
	520	td->total_err_count++;
	521	if (td->total_err_count == 1)
	522	td->first_error = err;
	523	}
	524
	525	static inline int should_fsync(struct thread_data *td)
	526	{
	527	if (td->last_was_sync)
	528	return 0;
	529	if (td->o.odirect)
	530	return 0;
	531	if (td_write(td) \|\| td_rw(td) \|\| td->o.override_sync)
	532	return 1;
	533
	534	return 0;
	535	}
	536
	537	/*
	538	* Init/option functions
	539	*/
	540	extern int __must_check parse_options(int, char **);
	541	extern int parse_jobs_ini(char *, int, int);
	542	extern int parse_cmd_line(int, char **);
	543	extern int exec_run(void);
	544	extern void reset_fio_state(void);
	545	extern int fio_options_parse(struct thread_data , char *, int);
	546	extern void fio_keywords_init(void);
	547	extern int fio_cmd_option_parse(struct thread_data , const char , char *);
	548	extern void fio_fill_default_options(struct thread_data *);
	549	extern int fio_show_option_help(const char *);
	550	extern void fio_options_dup_and_init(struct option *);
	551	extern void options_mem_dupe(struct thread_data *);
	552	extern void options_mem_free(struct thread_data *);
	553	extern void td_fill_rand_seeds(struct thread_data *);
	554	extern void add_job_opts(const char **);
	555	extern char *num2str(unsigned long, int, int, int);
	556
	557	#define FIO_GETOPT_JOB 0x89988998
	558	#define FIO_NR_OPTIONS (FIO_MAX_OPTS + 128)
	559
	560	/*
	561	* ETA/status stuff
	562	*/
	563	extern void print_thread_status(void);
	564	extern void print_status_init(int);
	565
	566	/*
	567	* Thread life cycle. Once a thread has a runstate beyond TD_INITIALIZED, it
	568	* will never back again. It may cycle between running/verififying/fsyncing.
	569	* Once the thread reaches TD_EXITED, it is just waiting for the core to
	570	* reap it.
	571	*/
	572	enum {
	573	TD_NOT_CREATED = 0,
	574	TD_CREATED,
	575	TD_INITIALIZED,
	576	TD_RAMP,
	577	TD_RUNNING,
	578	TD_PRE_READING,
	579	TD_VERIFYING,
	580	TD_FSYNCING,
	581	TD_EXITED,
	582	TD_REAPED,
	583	};
	584
	585	extern void td_set_runstate(struct thread_data *, int);
	586	#define TERMINATE_ALL (-1)
	587	extern void fio_terminate_threads(int);
	588
	589	/*
	590	* Memory helpers
	591	*/
	592	extern int __must_check fio_pin_memory(void);
	593	extern void fio_unpin_memory(void);
	594	extern int __must_check allocate_io_mem(struct thread_data *);
	595	extern void free_io_mem(struct thread_data *);
	596
	597	/*
	598	* Reset stats after ramp time completes
	599	*/
	600	extern void reset_all_stats(struct thread_data *);
	601
	602	/*
	603	* blktrace support
	604	*/
	605	#ifdef FIO_HAVE_BLKTRACE
	606	extern int is_blktrace(const char *);
	607	extern int load_blktrace(struct thread_data , const char );
	608	#endif
	609
	610	/*
	611	* Mark unused variables passed to ops functions as unused, to silence gcc
	612	*/
	613	#define fio_unused __attribute((__unused__))
	614	#define fio_init __attribute__((constructor))
	615	#define fio_exit __attribute__((destructor))
	616
	617	#define for_each_td(td, i) \
	618	for ((i) = 0, (td) = &threads[0]; (i) < (int) thread_number; (i)++, (td)++)
	619	#define for_each_file(td, f, i) \
	620	if ((td)->files_index) \
	621	for ((i) = 0, (f) = (td)->files[0]; \
	622	(i) < (td)->o.nr_files && ((f) = (td)->files[i]) != NULL; \
	623	(i)++)
	624
	625	#define fio_assert(td, cond) do { \
	626	if (!(cond)) { \
	627	int *__foo = NULL; \
	628	fprintf(stderr, "file:%s:%d, assert %s failed\n", __FILE__, __LINE__, #cond); \
	629	td_set_runstate((td), TD_EXITED); \
	630	(td)->error = EFAULT; \
	631	*__foo = 0; \
	632	} \
	633	} while (0)
	634
	635	static inline int fio_fill_issue_time(struct thread_data *td)
	636	{
	637	if (td->o.read_iolog_file \|\|
	638	!td->o.disable_clat \|\| !td->o.disable_slat \|\| !td->o.disable_bw)
	639	return 1;
	640
	641	return 0;
	642	}
	643
	644	static inline int __should_check_rate(struct thread_data *td,
	645	enum fio_ddir ddir)
	646	{
	647	struct thread_options *o = &td->o;
	648
	649	/*
	650	* If some rate setting was given, we need to check it
	651	*/
	652	if (o->rate[ddir] \|\| o->ratemin[ddir] \|\| o->rate_iops[ddir] \|\|
	653	o->rate_iops_min[ddir])
	654	return 1;
	655
	656	return 0;
	657	}
	658
	659	static inline int should_check_rate(struct thread_data *td,
	660	unsigned long *bytes_done)
	661	{
	662	int ret = 0;
	663
	664	if (bytes_done[0])
	665	ret \|= __should_check_rate(td, 0);
	666	if (bytes_done[1])
	667	ret \|= __should_check_rate(td, 1);
	668
	669	return ret;
	670	}
	671
	672	static inline int is_power_of_2(unsigned int val)
	673	{
	674	return (val != 0 && ((val & (val - 1)) == 0));
	675	}
	676
	677	/*
	678	* We currently only need to do locking if we have verifier threads
	679	* accessing our internal structures too
	680	*/
	681	static inline void td_io_u_lock(struct thread_data *td)
	682	{
	683	if (td->o.verify_async)
	684	pthread_mutex_lock(&td->io_u_lock);
	685	}
	686
	687	static inline void td_io_u_unlock(struct thread_data *td)
	688	{
	689	if (td->o.verify_async)
	690	pthread_mutex_unlock(&td->io_u_lock);
	691	}
	692
	693	static inline void td_io_u_free_notify(struct thread_data *td)
	694	{
	695	if (td->o.verify_async)
	696	pthread_cond_signal(&td->free_cond);
	697	}
	698
	699	#endif