[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 <getopt.h>
#include <inttypes.h>
#include <assert.h>

#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"

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

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

struct group_run_stats {
	unsigned long long max_run[2], min_run[2];
	unsigned long long max_bw[2], min_bw[2];
	unsigned long long io_kb[2];
	unsigned long long agg[2];
};

/*
 * 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 */
};

/*
 * How many depth levels to log
 */
#define FIO_IO_U_MAP_NR	8
#define FIO_IO_U_LAT_U_NR 10
#define FIO_IO_U_LAT_M_NR 12

struct thread_stat {
	char *name;
	char *verror;
	int error;
	int groupid;
	pid_t pid;
	char *description;
	int members;

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

	/*
	 * bandwidth and latency stats
	 */
	struct io_stat clat_stat[2];		/* completion latency */
	struct io_stat slat_stat[2];		/* submission latency */
	struct io_stat bw_stat[2];		/* bandwidth stats */

	unsigned long long stat_io_bytes[2];
	struct timeval stat_sample_time[2];

	/*
	 * fio system usage accounting
	 */
	struct rusage ru_start;
	struct rusage ru_end;
	unsigned long usr_time;
	unsigned long sys_time;
	unsigned long ctx;
	unsigned long minf, majf;

	/*
	 * IO depth and latency stats
	 */
	unsigned int io_u_map[FIO_IO_U_MAP_NR];
	unsigned int io_u_submit[FIO_IO_U_MAP_NR];
	unsigned int io_u_complete[FIO_IO_U_MAP_NR];
	unsigned int io_u_lat_u[FIO_IO_U_LAT_U_NR];
	unsigned int io_u_lat_m[FIO_IO_U_LAT_M_NR];
	unsigned long total_io_u[2];
	unsigned long short_io_u[2];
	unsigned long total_submit;
	unsigned long total_complete;

	unsigned long long io_bytes[2];
	unsigned long runtime[2];
	unsigned long total_run_time;

	/*
	 * IO Error related stats
	 */
	unsigned continue_on_error;
	unsigned long total_err_count;
	int first_error;
};

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 ddir_nr;
	unsigned int iodepth;
	unsigned int iodepth_low;
	unsigned int iodepth_batch;
	unsigned int iodepth_batch_complete;

	unsigned long long size;
	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;
	unsigned int verify_pattern;
	unsigned int verify_pattern_bytes;
	unsigned int verify_fatal;
	unsigned int verify_async;
	unsigned int use_thread;
	unsigned int unlink;
	unsigned int do_disk_util;
	unsigned int override_sync;
	unsigned int rand_repeatable;
	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 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;
	unsigned int zero_buffers;
	unsigned int refill_buffers;
	unsigned int time_based;
	unsigned int disable_clat;
	unsigned int disable_slat;
	unsigned int disable_bw;
	unsigned int gtod_reduce;
	unsigned int gtod_cpu;
	unsigned int gtod_offload;

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

	/*
	 * 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;
};

#define FIO_VERROR_SIZE	128

/*
 * 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 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;
	};
	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;

	char *mmapfile;
	int mmapfd;

	void *iolog_buf;
	FILE *iolog_f;

	char *sysfs_root;

	unsigned long rand_seeds[6];

	os_random_state_t bsrange_state;
	os_random_state_t verify_state;

	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_usec_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 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
	 */
	os_random_state_t 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
	 */
	os_random_state_t rwmix_state;
	unsigned long rwmix_issues;
	enum fio_ddir rwmix_ddir;
	unsigned int ddir_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;

	/*
	 * For IO replaying
	 */
	struct flist_head io_log_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;

	/*
	 * For generating file sizes
	 */
	os_random_state_t file_size_state;

	/*
	 * Error counts
	 */
	unsigned int total_err_count;
	int first_error;
};

/*
 * 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 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(int))
#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 MAX_JOBS	(1024)

#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;
}

/*
 * Time functions
 */
extern unsigned long long utime_since(struct timeval *, struct timeval *);
extern unsigned long long utime_since_now(struct timeval *);
extern unsigned long mtime_since(struct timeval *, struct timeval *);
extern unsigned long mtime_since_now(struct timeval *);
extern unsigned long time_since_now(struct timeval *);
extern unsigned long mtime_since_genesis(void);
extern void usec_spin(unsigned int);
extern void usec_sleep(struct thread_data *, unsigned long);
extern void fill_start_time(struct timeval *);
extern void fio_gettime(struct timeval *, void *);
extern void fio_gtod_init(void);
extern void fio_gtod_update(void);
extern void set_genesis_time(void);
extern int ramp_time_over(struct thread_data *);
extern int in_ramp_time(struct thread_data *);

/*
 * Init/option functions
 */
extern int __must_check parse_options(int, char **);
extern int fio_options_parse(struct thread_data *, char **, int);
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 *);
#define FIO_GETOPT_JOB		0x89988998
#define FIO_NR_OPTIONS		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)->runstate = 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;
}

/*
 * Cheesy number->string conversion, complete with carry rounding error.
 */
static inline char *num2str(unsigned long num, int maxlen, int base, int pow2)
{
	char postfix[] = { ' ', 'K', 'M', 'G', 'P', 'E' };
	unsigned int thousand;
	char *buf;
	int i;

	if (pow2)
		thousand = 1024;
	else
		thousand = 1000;

	buf = malloc(128);

	for (i = 0; base > 1; i++)
		base /= thousand;

	do {
		int len, carry = 0;

		len = sprintf(buf, "%'lu", num);
		if (len <= maxlen) {
			if (i >= 1) {
				buf[len] = postfix[i];
				buf[len + 1] = '\0';
			}
			return buf;
		}

		if ((num % thousand) >= (thousand / 2))
			carry = 1;

		num /= thousand;
		num += carry;
		i++;
	} while (i <= 5);

	return buf;
}

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
ebac4655 JA	1	#ifndef FIO_H
	2	#define FIO_H
	3
	4	#include <sched.h>
	5	#include <limits.h>
	6	#include <pthread.h>
	7	#include <sys/time.h>
	8	#include <sys/resource.h>
3c39a379 JA	9	#include <errno.h>
	10	#include <stdlib.h>
	11	#include <stdio.h>
6d6f031f	12	#include <unistd.h>
34cfcdaf	13	#include <string.h>
214e1eca	14	#include <getopt.h>
cd14cc10	15	#include <inttypes.h>
7101d9c2	16	#include <assert.h>
ebac4655	17
317b95d0	18	#include "compiler/compiler.h"
01743ee1	19	#include "flist.h"
e2887563	20	#include "fifo.h"
4b87898e	21	#include "rbtree.h"
317b95d0 JA	22	#include "arch/arch.h"
317b95d0 JA	23	#include "os/os.h"
07739b57	24	#include "mutex.h"
a3d741fa JA	25	#include "log.h"
a3d741fa JA	26	#include "debug.h"
d6aed795 JA	27	#include "file.h"
d6aed795 JA	28	#include "io_ddir.h"
dcefb588	29	#include "ioengine.h"
5995a6a4	30	#include "iolog.h"
ebac4655	31
609342ff DL	32	#ifdef FIO_HAVE_GUASI
	33	#include <guasi.h>
	34	#endif
	35
417f0068 JA	36	#ifdef FIO_HAVE_SOLARISAIO
	37	#include <sys/asynch.h>
	38	#endif
	39
ebac4655	40	struct group_run_stats {
9104f874 JA	41	unsigned long long max_run[2], min_run[2];
9104f874 JA	42	unsigned long long max_bw[2], min_bw[2];
e9b2a3fa	43	unsigned long long io_kb[2];
9104f874	44	unsigned long long agg[2];
ebac4655 JA	45	};
ebac4655 JA	46
e9c047a0 JA	47	/*
	48	* What type of allocation to use for io buffers
	49	*/
	50	enum fio_memtype {
	51	MEM_MALLOC = 0, /* ordinary malloc */
	52	MEM_SHM, /* use shared memory segments */
74b025b0	53	MEM_SHMHUGE, /* use shared memory segments with huge pages */
e9c047a0	54	MEM_MMAP, /* use anonynomous mmap */
d0bdaf49	55	MEM_MMAPHUGE, /* memory mapped huge file */
e9c047a0 JA	56	};
e9c047a0 JA	57
b2560f3c JA	58	/*
	59	* How many depth levels to log
	60	*/
	61	#define FIO_IO_U_MAP_NR 8
a9a18a3f JA	62	#define FIO_IO_U_LAT_U_NR 10
a9a18a3f JA	63	#define FIO_IO_U_LAT_M_NR 12
b2560f3c	64
079ad09b	65	struct thread_stat {
756867bd JA	66	char *name;
	67	char *verror;
	68	int error;
	69	int groupid;
	70	pid_t pid;
	71	char *description;
6586ee89	72	int members;
756867bd	73
079ad09b JA	74	struct io_log *slat_log;
	75	struct io_log *clat_log;
	76	struct io_log *bw_log;
	77
	78	/*
	79	* bandwidth and latency stats
	80	*/
	81	struct io_stat clat_stat[2]; /* completion latency */
	82	struct io_stat slat_stat[2]; /* submission latency */
	83	struct io_stat bw_stat[2]; /* bandwidth stats */
	84
	85	unsigned long long stat_io_bytes[2];
	86	struct timeval stat_sample_time[2];
	87
	88	/*
	89	* fio system usage accounting
	90	*/
	91	struct rusage ru_start;
	92	struct rusage ru_end;
	93	unsigned long usr_time;
	94	unsigned long sys_time;
	95	unsigned long ctx;
81887d5d	96	unsigned long minf, majf;
079ad09b	97
b2560f3c JA	98	/*
	99	* IO depth and latency stats
	100	*/
	101	unsigned int io_u_map[FIO_IO_U_MAP_NR];
838bc709 JA	102	unsigned int io_u_submit[FIO_IO_U_MAP_NR];
838bc709 JA	103	unsigned int io_u_complete[FIO_IO_U_MAP_NR];
04a0feae JA	104	unsigned int io_u_lat_u[FIO_IO_U_LAT_U_NR];
04a0feae JA	105	unsigned int io_u_lat_m[FIO_IO_U_LAT_M_NR];
b3605062	106	unsigned long total_io_u[2];
30061b97	107	unsigned long short_io_u[2];
838bc709 JA	108	unsigned long total_submit;
838bc709 JA	109	unsigned long total_complete;
756867bd JA	110
	111	unsigned long long io_bytes[2];
	112	unsigned long runtime[2];
	113	unsigned long total_run_time;
f2bba182 RR	114
	115	/*
	116	* IO Error related stats
	117	*/
	118	unsigned continue_on_error;
	119	unsigned long total_err_count;
	120	int first_error;
b2560f3c	121	};
71619dc2	122
564ca972 JA	123	struct bssplit {
	124	unsigned int bs;
	125	unsigned char perc;
	126	};
	127
2dc1bbeb	128	struct thread_options {
b1ec1da6	129	int pad;
61697c37	130	char *description;
b4692828	131	char *name;
ef899b63	132	char *directory;
13f8e2d2	133	char *filename;
2dc1bbeb	134	char *opendir;
09629a90	135	char *ioengine;
413dd459	136	enum td_ddir td_ddir;
211097b2	137	unsigned int ddir_nr;
2dc1bbeb JA	138	unsigned int iodepth;
	139	unsigned int iodepth_low;
	140	unsigned int iodepth_batch;
4950421a	141	unsigned int iodepth_batch_complete;
2dc1bbeb JA	142
2dc1bbeb JA	143	unsigned long long size;
aa31f1f1	144	unsigned int fill_device;
2dc1bbeb JA	145	unsigned long long file_size_low;
	146	unsigned long long file_size_high;
	147	unsigned long long start_offset;
	148
	149	unsigned int bs[2];
2b7a01d0	150	unsigned int ba[2];
2dc1bbeb JA	151	unsigned int min_bs[2];
2dc1bbeb JA	152	unsigned int max_bs[2];
720e84ad JA	153	struct bssplit *bssplit[2];
720e84ad JA	154	unsigned int bssplit_nr[2];
2dc1bbeb JA	155
	156	unsigned int nr_files;
	157	unsigned int open_files;
4d4e80f2	158	enum file_lock_mode file_lock_mode;
29c1349f	159	unsigned int lockfile_batch;
e9c047a0	160
9158d2f7 JA	161	unsigned int odirect;
	162	unsigned int invalidate_cache;
	163	unsigned int create_serialize;
	164	unsigned int create_fsync;
814452bd	165	unsigned int create_on_open;
9158d2f7	166	unsigned int end_fsync;
afad68f7	167	unsigned int pre_read;
9158d2f7 JA	168	unsigned int sync_io;
9158d2f7 JA	169	unsigned int verify;
e84c73a8	170	unsigned int do_verify;
160b966d	171	unsigned int verifysort;
a59e170d JA	172	unsigned int verify_interval;
a59e170d JA	173	unsigned int verify_offset;
90059d65 JA	174	unsigned int verify_pattern;
90059d65 JA	175	unsigned int verify_pattern_bytes;
a12a3b4d	176	unsigned int verify_fatal;
e8462bd8	177	unsigned int verify_async;
9158d2f7 JA	178	unsigned int use_thread;
	179	unsigned int unlink;
	180	unsigned int do_disk_util;
	181	unsigned int override_sync;
	182	unsigned int rand_repeatable;
	183	unsigned int write_lat_log;
	184	unsigned int write_bw_log;
bb8895e0	185	unsigned int norandommap;
2b386d25	186	unsigned int softrandommap;
690adba3	187	unsigned int bs_unaligned;
ebb1415f	188	unsigned int fsync_on_close;
e9c047a0	189
56bb17f2	190	unsigned int hugepage_size;
a00735e6	191	unsigned int rw_min_bs;
ebac4655	192	unsigned int thinktime;
48097d5c	193	unsigned int thinktime_spin;
9c1f7434	194	unsigned int thinktime_blocks;
ebac4655	195	unsigned int fsync_blocks;
5f9099ea	196	unsigned int fdatasync_blocks;
ebac4655	197	unsigned int start_delay;
0db26797	198	unsigned long long timeout;
721938ae	199	unsigned long long ramp_time;
ebac4655	200	unsigned int overwrite;
ebac4655	201	unsigned int bw_avg_time;
ebac4655	202	unsigned int loops;
20dc95c4 JA	203	unsigned long long zone_size;
20dc95c4 JA	204	unsigned long long zone_skip;
e9c047a0	205	enum fio_memtype mem_type;
d529ee19	206	unsigned int mem_align;
2dc1bbeb	207
ebac4655	208	unsigned int stonewall;
b3d62a75	209	unsigned int new_group;
ebac4655	210	unsigned int numjobs;
ebac4655	211	os_cpu_mask_t cpumask;
375b2695	212	unsigned int cpumask_set;
e8462bd8 JA	213	os_cpu_mask_t verify_cpumask;
e8462bd8 JA	214	unsigned int verify_cpumask_set;
aea47d44	215	unsigned int iolog;
a6ccc7be	216	unsigned int rwmixcycle;
e47f799f	217	unsigned int rwmix[2];
b6f4d880	218	unsigned int nice;
0aabe160	219	unsigned int file_service_type;
b2560f3c	220	unsigned int group_reporting;
d2f3ac35	221	unsigned int fadvise_hint;
e9459e5a	222	unsigned int zero_buffers;
5973cafb	223	unsigned int refill_buffers;
cf4464ca	224	unsigned int time_based;
9520ebb9 JA	225	unsigned int disable_clat;
	226	unsigned int disable_slat;
	227	unsigned int disable_bw;
993bf48b	228	unsigned int gtod_reduce;
be4ecfdf JA	229	unsigned int gtod_cpu;
be4ecfdf JA	230	unsigned int gtod_offload;
aea47d44	231
076efc7c JA	232	char *read_iolog_file;
076efc7c JA	233	char *write_iolog_file;
e3cedca7 JA	234	char *bw_log_file;
e3cedca7 JA	235	char *lat_log_file;
2dc1bbeb JA	236
	237	/*
	238	* Pre-run and post-run shell
	239	*/
	240	char *exec_prerun;
	241	char *exec_postrun;
	242
581e7141 JA	243	unsigned int rate[2];
581e7141 JA	244	unsigned int ratemin[2];
2dc1bbeb	245	unsigned int ratecycle;
581e7141 JA	246	unsigned int rate_iops[2];
581e7141 JA	247	unsigned int rate_iops_min[2];
2dc1bbeb JA	248
	249	char *ioscheduler;
	250
	251	/*
	252	* CPU "io" cycle burner
	253	*/
	254	unsigned int cpuload;
	255	unsigned int cpucycle;
f2bba182 RR	256
	257	/*
	258	* I/O Error handling
	259	*/
	260	unsigned int continue_on_error;
2dc1bbeb JA	261	};
2dc1bbeb JA	262
e4e33258 JA	263	#define FIO_VERROR_SIZE 128
e4e33258 JA	264
2dc1bbeb JA	265	/*
	266	* This describes a single thread/process executing a fio job.
	267	*/
	268	struct thread_data {
	269	struct thread_options o;
e4e33258	270	char verror[FIO_VERROR_SIZE];
2dc1bbeb JA	271	pthread_t thread;
	272	int thread_number;
	273	int groupid;
	274	struct thread_stat ts;
126d65c6	275	struct fio_file **files;
dd87b2c9	276	unsigned int files_size;
2dc1bbeb JA	277	unsigned int files_index;
2dc1bbeb JA	278	unsigned int nr_open_files;
1020a139	279	unsigned int nr_done_files;
2dc1bbeb JA	280	unsigned int nr_normal_files;
	281	union {
	282	unsigned int next_file;
	283	os_random_state_t next_file_state;
	284	};
	285	int error;
20e354ef	286	int done;
2dc1bbeb JA	287	pid_t pid;
	288	char *orig_buffer;
	289	size_t orig_buffer_size;
	290	volatile int terminate;
	291	volatile int runstate;
	292	unsigned int ioprio;
ac684785	293	unsigned int ioprio_set;
2dc1bbeb JA	294	unsigned int last_was_sync;
	295
	296	char *mmapfile;
	297	int mmapfd;
	298
843a7413 JA	299	void *iolog_buf;
843a7413 JA	300	FILE *iolog_f;
ebac4655	301
da86774e	302	char *sysfs_root;
da86774e	303
5bfc35d7 JA	304	unsigned long rand_seeds[6];
5bfc35d7 JA	305
6dfd46b9 JA	306	os_random_state_t bsrange_state;
6dfd46b9 JA	307	os_random_state_t verify_state;
ebac4655 JA	308
	309	int shm_id;
	310
e9c047a0 JA	311	/*
	312	* IO engine hooks, contains everything needed to submit an io_u
	313	* to any of the available IO engines.
	314	*/
2866c82d	315	struct ioengine_ops *io_ops;
ebac4655	316
e9c047a0 JA	317	/*
	318	* Current IO depth and list of free and busy io_u's.
	319	*/
ebac4655	320	unsigned int cur_depth;
d8005759	321	unsigned int io_u_queued;
01743ee1 JA	322	struct flist_head io_u_freelist;
	323	struct flist_head io_u_busylist;
	324	struct flist_head io_u_requeues;
e8462bd8 JA	325	pthread_mutex_t io_u_lock;
	326	pthread_cond_t free_cond;
	327
	328	/*
	329	* async verify offload
	330	*/
	331	struct flist_head verify_list;
	332	pthread_t *verify_threads;
	333	unsigned int nr_verify_threads;
	334	pthread_cond_t verify_cond;
	335	int verify_thread_exit;
ebac4655	336
e9c047a0 JA	337	/*
	338	* Rate state
	339	*/
581e7141 JA	340	unsigned long rate_usec_cycle[2];
	341	long rate_pending_usleep[2];
	342	unsigned long rate_bytes[2];
	343	unsigned long rate_blocks[2];
	344	struct timeval lastrate[2];
ebac4655	345
ebac4655 JA	346	unsigned long long total_io_size;
ebac4655 JA	347
755200a3	348	unsigned long io_issues[2];
9104f874 JA	349	unsigned long long io_blocks[2];
9104f874 JA	350	unsigned long long io_bytes[2];
48f5abd3	351	unsigned long long io_skip_bytes;
9104f874	352	unsigned long long this_io_bytes[2];
079ad09b	353	unsigned long long zone_bytes;
cdd18ad8	354	struct fio_mutex *mutex;
ebac4655	355
e9c047a0 JA	356	/*
	357	* State for random io, a bitmap of blocks done vs not done
	358	*/
6dfd46b9	359	os_random_state_t random_state;
ebac4655	360
ebac4655 JA	361	struct timeval start; /* start of this loop */
ebac4655 JA	362	struct timeval epoch; /* time job was started */
a61eddec	363	struct timeval last_issue;
993bf48b JA	364	struct timeval tv_cache;
	365	unsigned int tv_cache_nr;
	366	unsigned int tv_cache_mask;
721938ae	367	unsigned int ramp_time_over;
ebac4655	368
e9c047a0 JA	369	/*
	370	* read/write mixed workload state
	371	*/
6dfd46b9	372	os_random_state_t rwmix_state;
e4928662	373	unsigned long rwmix_issues;
e9c047a0	374	enum fio_ddir rwmix_ddir;
211097b2	375	unsigned int ddir_nr;
a6ccc7be	376
e9c047a0	377	/*
8de8f047 JA	378	* IO history logs for verification. We use a tree for sorting,
8de8f047 JA	379	* if we are overwriting. Otherwise just use a fifo.
e9c047a0	380	*/
4b87898e	381	struct rb_root io_hist_tree;
01743ee1	382	struct flist_head io_hist_list;
8de8f047 JA	383
	384	/*
	385	* For IO replaying
	386	*/
01743ee1	387	struct flist_head io_log_list;
433afcb4	388
1907dbc6 JA	389	/*
	390	* for fileservice, how often to switch to a new file
	391	*/
	392	unsigned int file_service_nr;
	393	unsigned int file_service_left;
	394	struct fio_file *file_service_file;
9c60ce64 JA	395
	396	/*
	397	* For generating file sizes
	398	*/
	399	os_random_state_t file_size_state;
f2bba182 RR	400
	401	/*
	402	* Error counts
	403	*/
	404	unsigned int total_err_count;
	405	int first_error;
ebac4655 JA	406	};
ebac4655 JA	407
e592a06b AC	408	/*
	409	* when should interactive ETA output be generated
	410	*/
	411	enum {
	412	FIO_ETA_AUTO,
	413	FIO_ETA_ALWAYS,
	414	FIO_ETA_NEVER,
	415	};
	416
e1161c32	417	#define __td_verror(td, err, msg, func) \
ebac4655	418	do { \
19abcd3d JA	419	if ((td)->error) \
19abcd3d JA	420	break; \
ebac4655 JA	421	int e = (err); \
ebac4655 JA	422	(td)->error = e; \
f2bba182 RR	423	if (!(td)->first_error) \
f2bba182 RR	424	snprintf(td->verror, sizeof(td->verror) - 1, "file:%s:%d, func=%s, error=%s", __FILE__, __LINE__, (func), (msg)); \
ebac4655 JA	425	} while (0)
ebac4655 JA	426
b990b5c0	427
f2bba182 RR	428	#define td_clear_error(td) \
f2bba182 RR	429	(td)->error = 0;
e1161c32 JA	430	#define td_verror(td, err, func) \
	431	__td_verror((td), (err), strerror((err)), (func))
	432	#define td_vmsg(td, err, msg, func) \
	433	__td_verror((td), (err), (msg), (func))
b990b5c0	434
ebac4655 JA	435	extern int exitall_on_terminate;
ebac4655 JA	436	extern int thread_number;
9cedf167	437	extern int nr_process, nr_thread;
ebac4655 JA	438	extern int shm_id;
ebac4655 JA	439	extern int groupid;
c6ae0a5b	440	extern int terse_output;
53cdc686	441	extern int temp_stall_ts;
1e97cce9	442	extern unsigned long long mlock_size;
cfc99db7	443	extern unsigned long page_mask, page_size;
4241ea8f	444	extern int read_only;
e592a06b	445	extern int eta_print;
d3eeeabc	446	extern unsigned long done_secs;
01f06b63	447	extern char *job_section;
be4ecfdf JA	448	extern int fio_gtod_offload;
be4ecfdf JA	449	extern int fio_gtod_cpu;
ebac4655 JA	450
	451	extern struct thread_data *threads;
	452
7101d9c2 JA	453	static inline void fio_ro_check(struct thread_data td, struct io_u io_u)
	454	{
	455	assert(!(io_u->ddir == DDIR_WRITE && !td_write(td)));
	456	}
	457
de605666	458	#define BLOCKS_PER_MAP (8 * sizeof(int))
aec2de20 JA	459	#define TO_MAP_BLOCK(f, b) (b)
	460	#define RAND_MAP_IDX(f, b) (TO_MAP_BLOCK(f, b) / BLOCKS_PER_MAP)
	461	#define RAND_MAP_BIT(f, b) (TO_MAP_BLOCK(f, b) & (BLOCKS_PER_MAP - 1))
ebac4655 JA	462
	463	#define MAX_JOBS (1024)
	464
1ec99eea	465	#define td_non_fatal_error(e) ((e) == EIO \|\| (e) == EILSEQ)
f2bba182 RR	466
	467	static inline void update_error_count(struct thread_data *td, int err)
	468	{
	469	td->total_err_count++;
	470	if (td->total_err_count == 1)
	471	td->first_error = err;
	472	}
	473
87dc1ab1 JA	474	static inline int should_fsync(struct thread_data *td)
	475	{
	476	if (td->last_was_sync)
	477	return 0;
2dc1bbeb	478	if (td->o.odirect)
87dc1ab1	479	return 0;
2dc1bbeb	480	if (td_write(td) \|\| td_rw(td) \|\| td->o.override_sync)
87dc1ab1 JA	481	return 1;
	482
	483	return 0;
	484	}
	485
6796209a JA	486	/*
	487	* Time functions
	488	*/
10927316 SL	489	extern unsigned long long utime_since(struct timeval , struct timeval );
10927316 SL	490	extern unsigned long long utime_since_now(struct timeval *);
6796209a JA	491	extern unsigned long mtime_since(struct timeval , struct timeval );
	492	extern unsigned long mtime_since_now(struct timeval *);
	493	extern unsigned long time_since_now(struct timeval *);
263e529f	494	extern unsigned long mtime_since_genesis(void);
6dd6f2cd	495	extern void usec_spin(unsigned int);
6796209a	496	extern void usec_sleep(struct thread_data *, unsigned long);
6043c579	497	extern void fill_start_time(struct timeval *);
02bcaa8c	498	extern void fio_gettime(struct timeval , void );
be4ecfdf JA	499	extern void fio_gtod_init(void);
be4ecfdf JA	500	extern void fio_gtod_update(void);
a2f77c9f	501	extern void set_genesis_time(void);
721938ae	502	extern int ramp_time_over(struct thread_data *);
b29ee5b3	503	extern int in_ramp_time(struct thread_data *);
6796209a	504
8914a9d8	505	/*
214e1eca	506	* Init/option functions
8914a9d8	507	*/
72cb971b	508	extern int __must_check parse_options(int, char **);
3b8b7135	509	extern int fio_options_parse(struct thread_data , char *, int);
214e1eca JA	510	extern int fio_cmd_option_parse(struct thread_data , const char , char *);
	511	extern void fio_fill_default_options(struct thread_data *);
	512	extern int fio_show_option_help(const char *);
	513	extern void fio_options_dup_and_init(struct option *);
d23bb327 JA	514	extern void options_mem_dupe(struct thread_data *);
d23bb327 JA	515	extern void options_mem_free(struct thread_data *);
5bfc35d7	516	extern void td_fill_rand_seeds(struct thread_data *);
214e1eca JA	517	#define FIO_GETOPT_JOB 0x89988998
214e1eca JA	518	#define FIO_NR_OPTIONS 128
8914a9d8	519
263e529f JA	520	/*
	521	* ETA/status stuff
	522	*/
	523	extern void print_thread_status(void);
	524	extern void print_status_init(int);
	525
	526	/*
	527	* Thread life cycle. Once a thread has a runstate beyond TD_INITIALIZED, it
	528	* will never back again. It may cycle between running/verififying/fsyncing.
	529	* Once the thread reaches TD_EXITED, it is just waiting for the core to
	530	* reap it.
	531	*/
	532	enum {
	533	TD_NOT_CREATED = 0,
	534	TD_CREATED,
	535	TD_INITIALIZED,
b29ee5b3	536	TD_RAMP,
263e529f	537	TD_RUNNING,
b0f65863	538	TD_PRE_READING,
263e529f JA	539	TD_VERIFYING,
	540	TD_FSYNCING,
	541	TD_EXITED,
	542	TD_REAPED,
	543	};
	544
b29ee5b3 JA	545	extern void td_set_runstate(struct thread_data *, int);
b29ee5b3 JA	546
2f9ade3c JA	547	/*
	548	* Memory helpers
	549	*/
b2fdda43	550	extern int __must_check fio_pin_memory(void);
2f9ade3c	551	extern void fio_unpin_memory(void);
b2fdda43	552	extern int __must_check allocate_io_mem(struct thread_data *);
2f9ade3c JA	553	extern void free_io_mem(struct thread_data *);
2f9ade3c JA	554
b29ee5b3 JA	555	/*
	556	* Reset stats after ramp time completes
	557	*/
	558	extern void reset_all_stats(struct thread_data *);
	559
fb7b71a3 JA	560	/*
	561	* blktrace support
	562	*/
5e62c22a	563	#ifdef FIO_HAVE_BLKTRACE
fb7b71a3 JA	564	extern int is_blktrace(const char *);
fb7b71a3 JA	565	extern int load_blktrace(struct thread_data , const char );
5e62c22a	566	#endif
fb7b71a3	567
2866c82d JA	568	/*
	569	* Mark unused variables passed to ops functions as unused, to silence gcc
	570	*/
	571	#define fio_unused __attribute((__unused__))
5f350952 JA	572	#define fio_init __attribute__((constructor))
5f350952 JA	573	#define fio_exit __attribute__((destructor))
2866c82d	574
34572e28 JA	575	#define for_each_td(td, i) \
34572e28 JA	576	for ((i) = 0, (td) = &threads[0]; (i) < (int) thread_number; (i)++, (td)++)
53cdc686	577	#define for_each_file(td, f, i) \
691c8fb0 JA	578	if ((td)->files_index) \
	579	for ((i) = 0, (f) = (td)->files[0]; \
	580	(i) < (td)->o.nr_files && ((f) = (td)->files[i]) != NULL; \
	581	(i)++)
53cdc686	582
0032bf9f JA	583	#define fio_assert(td, cond) do { \
0032bf9f JA	584	if (!(cond)) { \
340fd243	585	int *__foo = NULL; \
0032bf9f JA	586	fprintf(stderr, "file:%s:%d, assert %s failed\n", __FILE__, __LINE__, #cond); \
0032bf9f JA	587	(td)->runstate = TD_EXITED; \
437c9b71	588	(td)->error = EFAULT; \
340fd243	589	*__foo = 0; \
0032bf9f JA	590	} \
	591	} while (0)
	592
12d9d841 JA	593	static inline int fio_fill_issue_time(struct thread_data *td)
	594	{
	595	if (td->o.read_iolog_file \|\|
	596	!td->o.disable_clat \|\| !td->o.disable_slat \|\| !td->o.disable_bw)
	597	return 1;
	598
	599	return 0;
	600	}
	601
adb02ba8 JA	602	/*
	603	* Cheesy number->string conversion, complete with carry rounding error.
	604	*/
	605	static inline char *num2str(unsigned long num, int maxlen, int base, int pow2)
	606	{
	607	char postfix[] = { ' ', 'K', 'M', 'G', 'P', 'E' };
	608	unsigned int thousand;
	609	char *buf;
	610	int i;
	611
	612	if (pow2)
	613	thousand = 1024;
	614	else
	615	thousand = 1000;
	616
	617	buf = malloc(128);
	618
	619	for (i = 0; base > 1; i++)
	620	base /= thousand;
	621
	622	do {
	623	int len, carry = 0;
	624
	625	len = sprintf(buf, "%'lu", num);
	626	if (len <= maxlen) {
	627	if (i >= 1) {
	628	buf[len] = postfix[i];
	629	buf[len + 1] = '\0';
	630	}
	631	return buf;
	632	}
	633
	634	if ((num % thousand) >= (thousand / 2))
	635	carry = 1;
	636
	637	num /= thousand;
	638	num += carry;
	639	i++;
	640	} while (i <= 5);
	641
	642	return buf;
	643	}
	644
581e7141 JA	645	static inline int __should_check_rate(struct thread_data *td,
	646	enum fio_ddir ddir)
	647	{
	648	struct thread_options *o = &td->o;
	649
	650	/*
	651	* If some rate setting was given, we need to check it
	652	*/
	653	if (o->rate[ddir] \|\| o->ratemin[ddir] \|\| o->rate_iops[ddir] \|\|
	654	o->rate_iops_min[ddir])
	655	return 1;
	656
	657	return 0;
	658	}
	659
	660	static inline int should_check_rate(struct thread_data *td,
	661	unsigned long *bytes_done)
	662	{
	663	int ret = 0;
	664
	665	if (bytes_done[0])
	666	ret \|= __should_check_rate(td, 0);
	667	if (bytes_done[1])
	668	ret \|= __should_check_rate(td, 1);
	669
	670	return ret;
	671	}
	672
d529ee19 JA	673	static inline int is_power_of_2(unsigned int val)
	674	{
	675	return (val != 0 && ((val & (val - 1)) == 0));
	676	}
	677
e8462bd8 JA	678	/*
	679	* We currently only need to do locking if we have verifier threads
	680	* accessing our internal structures too
	681	*/
	682	static inline void td_io_u_lock(struct thread_data *td)
	683	{
	684	if (td->o.verify_async)
	685	pthread_mutex_lock(&td->io_u_lock);
	686	}
	687
	688	static inline void td_io_u_unlock(struct thread_data *td)
	689	{
	690	if (td->o.verify_async)
	691	pthread_mutex_unlock(&td->io_u_lock);
	692	}
	693
	694	static inline void td_io_u_free_notify(struct thread_data *td)
	695	{
	696	if (td->o.verify_async)
	697	pthread_cond_signal(&td->free_cond);
	698	}
	699
ebac4655	700	#endif