[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 "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))

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 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);

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