[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 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 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 stonewall;
	unsigned int new_group;
	unsigned int numjobs;
	os_cpu_mask_t cpumask;
	unsigned int 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;

	/*
	 * 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;
}

#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;
9158d2f7 JA	177	unsigned int use_thread;
	178	unsigned int unlink;
	179	unsigned int do_disk_util;
	180	unsigned int override_sync;
	181	unsigned int rand_repeatable;
	182	unsigned int write_lat_log;
	183	unsigned int write_bw_log;
bb8895e0	184	unsigned int norandommap;
2b386d25	185	unsigned int softrandommap;
690adba3	186	unsigned int bs_unaligned;
ebb1415f	187	unsigned int fsync_on_close;
e9c047a0	188
56bb17f2	189	unsigned int hugepage_size;
a00735e6	190	unsigned int rw_min_bs;
ebac4655	191	unsigned int thinktime;
48097d5c	192	unsigned int thinktime_spin;
9c1f7434	193	unsigned int thinktime_blocks;
ebac4655 JA	194	unsigned int fsync_blocks;
ebac4655 JA	195	unsigned int start_delay;
0db26797	196	unsigned long long timeout;
721938ae	197	unsigned long long ramp_time;
ebac4655	198	unsigned int overwrite;
ebac4655	199	unsigned int bw_avg_time;
ebac4655	200	unsigned int loops;
20dc95c4 JA	201	unsigned long long zone_size;
20dc95c4 JA	202	unsigned long long zone_skip;
e9c047a0	203	enum fio_memtype mem_type;
2dc1bbeb	204
ebac4655	205	unsigned int stonewall;
b3d62a75	206	unsigned int new_group;
ebac4655	207	unsigned int numjobs;
ebac4655	208	os_cpu_mask_t cpumask;
375b2695	209	unsigned int cpumask_set;
aea47d44	210	unsigned int iolog;
a6ccc7be	211	unsigned int rwmixcycle;
e47f799f	212	unsigned int rwmix[2];
b6f4d880	213	unsigned int nice;
0aabe160	214	unsigned int file_service_type;
b2560f3c	215	unsigned int group_reporting;
d2f3ac35	216	unsigned int fadvise_hint;
e9459e5a	217	unsigned int zero_buffers;
5973cafb	218	unsigned int refill_buffers;
cf4464ca	219	unsigned int time_based;
9520ebb9 JA	220	unsigned int disable_clat;
	221	unsigned int disable_slat;
	222	unsigned int disable_bw;
993bf48b	223	unsigned int gtod_reduce;
be4ecfdf JA	224	unsigned int gtod_cpu;
be4ecfdf JA	225	unsigned int gtod_offload;
aea47d44	226
076efc7c JA	227	char *read_iolog_file;
076efc7c JA	228	char *write_iolog_file;
e3cedca7 JA	229	char *bw_log_file;
e3cedca7 JA	230	char *lat_log_file;
2dc1bbeb JA	231
	232	/*
	233	* Pre-run and post-run shell
	234	*/
	235	char *exec_prerun;
	236	char *exec_postrun;
	237
581e7141 JA	238	unsigned int rate[2];
581e7141 JA	239	unsigned int ratemin[2];
2dc1bbeb	240	unsigned int ratecycle;
581e7141 JA	241	unsigned int rate_iops[2];
581e7141 JA	242	unsigned int rate_iops_min[2];
2dc1bbeb JA	243
	244	char *ioscheduler;
	245
	246	/*
	247	* CPU "io" cycle burner
	248	*/
	249	unsigned int cpuload;
	250	unsigned int cpucycle;
f2bba182 RR	251
	252	/*
	253	* I/O Error handling
	254	*/
	255	unsigned int continue_on_error;
2dc1bbeb JA	256	};
2dc1bbeb JA	257
e4e33258 JA	258	#define FIO_VERROR_SIZE 128
e4e33258 JA	259
2dc1bbeb JA	260	/*
	261	* This describes a single thread/process executing a fio job.
	262	*/
	263	struct thread_data {
	264	struct thread_options o;
e4e33258	265	char verror[FIO_VERROR_SIZE];
2dc1bbeb JA	266	pthread_t thread;
	267	int thread_number;
	268	int groupid;
	269	struct thread_stat ts;
126d65c6	270	struct fio_file **files;
dd87b2c9	271	unsigned int files_size;
2dc1bbeb JA	272	unsigned int files_index;
2dc1bbeb JA	273	unsigned int nr_open_files;
1020a139	274	unsigned int nr_done_files;
2dc1bbeb JA	275	unsigned int nr_normal_files;
	276	union {
	277	unsigned int next_file;
	278	os_random_state_t next_file_state;
	279	};
	280	int error;
20e354ef	281	int done;
2dc1bbeb JA	282	pid_t pid;
	283	char *orig_buffer;
	284	size_t orig_buffer_size;
	285	volatile int terminate;
	286	volatile int runstate;
	287	unsigned int ioprio;
ac684785	288	unsigned int ioprio_set;
2dc1bbeb JA	289	unsigned int last_was_sync;
	290
	291	char *mmapfile;
	292	int mmapfd;
	293
843a7413 JA	294	void *iolog_buf;
843a7413 JA	295	FILE *iolog_f;
ebac4655	296
da86774e	297	char *sysfs_root;
da86774e	298
5bfc35d7 JA	299	unsigned long rand_seeds[6];
5bfc35d7 JA	300
6dfd46b9 JA	301	os_random_state_t bsrange_state;
6dfd46b9 JA	302	os_random_state_t verify_state;
ebac4655 JA	303
	304	int shm_id;
	305
e9c047a0 JA	306	/*
	307	* IO engine hooks, contains everything needed to submit an io_u
	308	* to any of the available IO engines.
	309	*/
2866c82d	310	struct ioengine_ops *io_ops;
ebac4655	311
e9c047a0 JA	312	/*
	313	* Current IO depth and list of free and busy io_u's.
	314	*/
ebac4655	315	unsigned int cur_depth;
d8005759	316	unsigned int io_u_queued;
01743ee1 JA	317	struct flist_head io_u_freelist;
	318	struct flist_head io_u_busylist;
	319	struct flist_head io_u_requeues;
ebac4655	320
e9c047a0 JA	321	/*
	322	* Rate state
	323	*/
581e7141 JA	324	unsigned long rate_usec_cycle[2];
	325	long rate_pending_usleep[2];
	326	unsigned long rate_bytes[2];
	327	unsigned long rate_blocks[2];
	328	struct timeval lastrate[2];
ebac4655	329
ebac4655 JA	330	unsigned long long total_io_size;
ebac4655 JA	331
755200a3	332	unsigned long io_issues[2];
9104f874 JA	333	unsigned long long io_blocks[2];
9104f874 JA	334	unsigned long long io_bytes[2];
48f5abd3	335	unsigned long long io_skip_bytes;
9104f874	336	unsigned long long this_io_bytes[2];
079ad09b	337	unsigned long long zone_bytes;
cdd18ad8	338	struct fio_mutex *mutex;
ebac4655	339
e9c047a0 JA	340	/*
	341	* State for random io, a bitmap of blocks done vs not done
	342	*/
6dfd46b9	343	os_random_state_t random_state;
ebac4655	344
ebac4655 JA	345	struct timeval start; /* start of this loop */
ebac4655 JA	346	struct timeval epoch; /* time job was started */
a61eddec	347	struct timeval last_issue;
993bf48b JA	348	struct timeval tv_cache;
	349	unsigned int tv_cache_nr;
	350	unsigned int tv_cache_mask;
721938ae	351	unsigned int ramp_time_over;
ebac4655	352
e9c047a0 JA	353	/*
	354	* read/write mixed workload state
	355	*/
6dfd46b9	356	os_random_state_t rwmix_state;
e4928662	357	unsigned long rwmix_issues;
e9c047a0	358	enum fio_ddir rwmix_ddir;
211097b2	359	unsigned int ddir_nr;
a6ccc7be	360
e9c047a0	361	/*
8de8f047 JA	362	* IO history logs for verification. We use a tree for sorting,
8de8f047 JA	363	* if we are overwriting. Otherwise just use a fifo.
e9c047a0	364	*/
4b87898e	365	struct rb_root io_hist_tree;
01743ee1	366	struct flist_head io_hist_list;
8de8f047 JA	367
	368	/*
	369	* For IO replaying
	370	*/
01743ee1	371	struct flist_head io_log_list;
433afcb4	372
1907dbc6 JA	373	/*
	374	* for fileservice, how often to switch to a new file
	375	*/
	376	unsigned int file_service_nr;
	377	unsigned int file_service_left;
	378	struct fio_file *file_service_file;
9c60ce64 JA	379
	380	/*
	381	* For generating file sizes
	382	*/
	383	os_random_state_t file_size_state;
f2bba182 RR	384
	385	/*
	386	* Error counts
	387	*/
	388	unsigned int total_err_count;
	389	int first_error;
ebac4655 JA	390	};
ebac4655 JA	391
e592a06b AC	392	/*
	393	* when should interactive ETA output be generated
	394	*/
	395	enum {
	396	FIO_ETA_AUTO,
	397	FIO_ETA_ALWAYS,
	398	FIO_ETA_NEVER,
	399	};
	400
e1161c32	401	#define __td_verror(td, err, msg, func) \
ebac4655	402	do { \
19abcd3d JA	403	if ((td)->error) \
19abcd3d JA	404	break; \
ebac4655 JA	405	int e = (err); \
ebac4655 JA	406	(td)->error = e; \
f2bba182 RR	407	if (!(td)->first_error) \
f2bba182 RR	408	snprintf(td->verror, sizeof(td->verror) - 1, "file:%s:%d, func=%s, error=%s", __FILE__, __LINE__, (func), (msg)); \
ebac4655 JA	409	} while (0)
ebac4655 JA	410
b990b5c0	411
f2bba182 RR	412	#define td_clear_error(td) \
f2bba182 RR	413	(td)->error = 0;
e1161c32 JA	414	#define td_verror(td, err, func) \
	415	__td_verror((td), (err), strerror((err)), (func))
	416	#define td_vmsg(td, err, msg, func) \
	417	__td_verror((td), (err), (msg), (func))
b990b5c0	418
ebac4655 JA	419	extern int exitall_on_terminate;
ebac4655 JA	420	extern int thread_number;
9cedf167	421	extern int nr_process, nr_thread;
ebac4655 JA	422	extern int shm_id;
ebac4655 JA	423	extern int groupid;
c6ae0a5b	424	extern int terse_output;
53cdc686	425	extern int temp_stall_ts;
1e97cce9	426	extern unsigned long long mlock_size;
cfc99db7	427	extern unsigned long page_mask, page_size;
4241ea8f	428	extern int read_only;
e592a06b	429	extern int eta_print;
d3eeeabc	430	extern unsigned long done_secs;
01f06b63	431	extern char *job_section;
be4ecfdf JA	432	extern int fio_gtod_offload;
be4ecfdf JA	433	extern int fio_gtod_cpu;
ebac4655 JA	434
	435	extern struct thread_data *threads;
	436
7101d9c2 JA	437	static inline void fio_ro_check(struct thread_data td, struct io_u io_u)
	438	{
	439	assert(!(io_u->ddir == DDIR_WRITE && !td_write(td)));
	440	}
	441
de605666	442	#define BLOCKS_PER_MAP (8 * sizeof(int))
aec2de20 JA	443	#define TO_MAP_BLOCK(f, b) (b)
	444	#define RAND_MAP_IDX(f, b) (TO_MAP_BLOCK(f, b) / BLOCKS_PER_MAP)
	445	#define RAND_MAP_BIT(f, b) (TO_MAP_BLOCK(f, b) & (BLOCKS_PER_MAP - 1))
ebac4655 JA	446
	447	#define MAX_JOBS (1024)
	448
1ec99eea	449	#define td_non_fatal_error(e) ((e) == EIO \|\| (e) == EILSEQ)
f2bba182 RR	450
	451	static inline void update_error_count(struct thread_data *td, int err)
	452	{
	453	td->total_err_count++;
	454	if (td->total_err_count == 1)
	455	td->first_error = err;
	456	}
	457
87dc1ab1 JA	458	static inline int should_fsync(struct thread_data *td)
	459	{
	460	if (td->last_was_sync)
	461	return 0;
2dc1bbeb	462	if (td->o.odirect)
87dc1ab1	463	return 0;
2dc1bbeb	464	if (td_write(td) \|\| td_rw(td) \|\| td->o.override_sync)
87dc1ab1 JA	465	return 1;
	466
	467	return 0;
	468	}
	469
6796209a JA	470	/*
	471	* Time functions
	472	*/
10927316 SL	473	extern unsigned long long utime_since(struct timeval , struct timeval );
10927316 SL	474	extern unsigned long long utime_since_now(struct timeval *);
6796209a JA	475	extern unsigned long mtime_since(struct timeval , struct timeval );
	476	extern unsigned long mtime_since_now(struct timeval *);
	477	extern unsigned long time_since_now(struct timeval *);
263e529f	478	extern unsigned long mtime_since_genesis(void);
6dd6f2cd	479	extern void usec_spin(unsigned int);
6796209a	480	extern void usec_sleep(struct thread_data *, unsigned long);
6043c579	481	extern void fill_start_time(struct timeval *);
02bcaa8c	482	extern void fio_gettime(struct timeval , void );
be4ecfdf JA	483	extern void fio_gtod_init(void);
be4ecfdf JA	484	extern void fio_gtod_update(void);
a2f77c9f	485	extern void set_genesis_time(void);
721938ae	486	extern int ramp_time_over(struct thread_data *);
b29ee5b3	487	extern int in_ramp_time(struct thread_data *);
6796209a	488
8914a9d8	489	/*
214e1eca	490	* Init/option functions
8914a9d8	491	*/
72cb971b	492	extern int __must_check parse_options(int, char **);
3b8b7135	493	extern int fio_options_parse(struct thread_data , char *, int);
214e1eca JA	494	extern int fio_cmd_option_parse(struct thread_data , const char , char *);
	495	extern void fio_fill_default_options(struct thread_data *);
	496	extern int fio_show_option_help(const char *);
	497	extern void fio_options_dup_and_init(struct option *);
d23bb327 JA	498	extern void options_mem_dupe(struct thread_data *);
d23bb327 JA	499	extern void options_mem_free(struct thread_data *);
5bfc35d7	500	extern void td_fill_rand_seeds(struct thread_data *);
214e1eca JA	501	#define FIO_GETOPT_JOB 0x89988998
214e1eca JA	502	#define FIO_NR_OPTIONS 128
8914a9d8	503
263e529f JA	504	/*
	505	* ETA/status stuff
	506	*/
	507	extern void print_thread_status(void);
	508	extern void print_status_init(int);
	509
	510	/*
	511	* Thread life cycle. Once a thread has a runstate beyond TD_INITIALIZED, it
	512	* will never back again. It may cycle between running/verififying/fsyncing.
	513	* Once the thread reaches TD_EXITED, it is just waiting for the core to
	514	* reap it.
	515	*/
	516	enum {
	517	TD_NOT_CREATED = 0,
	518	TD_CREATED,
	519	TD_INITIALIZED,
b29ee5b3	520	TD_RAMP,
263e529f	521	TD_RUNNING,
b0f65863	522	TD_PRE_READING,
263e529f JA	523	TD_VERIFYING,
	524	TD_FSYNCING,
	525	TD_EXITED,
	526	TD_REAPED,
	527	};
	528
b29ee5b3 JA	529	extern void td_set_runstate(struct thread_data *, int);
b29ee5b3 JA	530
2f9ade3c JA	531	/*
	532	* Memory helpers
	533	*/
b2fdda43	534	extern int __must_check fio_pin_memory(void);
2f9ade3c	535	extern void fio_unpin_memory(void);
b2fdda43	536	extern int __must_check allocate_io_mem(struct thread_data *);
2f9ade3c JA	537	extern void free_io_mem(struct thread_data *);
2f9ade3c JA	538
b29ee5b3 JA	539	/*
	540	* Reset stats after ramp time completes
	541	*/
	542	extern void reset_all_stats(struct thread_data *);
	543
fb7b71a3 JA	544	/*
	545	* blktrace support
	546	*/
5e62c22a	547	#ifdef FIO_HAVE_BLKTRACE
fb7b71a3 JA	548	extern int is_blktrace(const char *);
fb7b71a3 JA	549	extern int load_blktrace(struct thread_data , const char );
5e62c22a	550	#endif
fb7b71a3	551
2866c82d JA	552	/*
	553	* Mark unused variables passed to ops functions as unused, to silence gcc
	554	*/
	555	#define fio_unused __attribute((__unused__))
5f350952 JA	556	#define fio_init __attribute__((constructor))
5f350952 JA	557	#define fio_exit __attribute__((destructor))
2866c82d	558
34572e28 JA	559	#define for_each_td(td, i) \
34572e28 JA	560	for ((i) = 0, (td) = &threads[0]; (i) < (int) thread_number; (i)++, (td)++)
53cdc686	561	#define for_each_file(td, f, i) \
691c8fb0 JA	562	if ((td)->files_index) \
	563	for ((i) = 0, (f) = (td)->files[0]; \
	564	(i) < (td)->o.nr_files && ((f) = (td)->files[i]) != NULL; \
	565	(i)++)
53cdc686	566
0032bf9f JA	567	#define fio_assert(td, cond) do { \
0032bf9f JA	568	if (!(cond)) { \
340fd243	569	int *__foo = NULL; \
0032bf9f JA	570	fprintf(stderr, "file:%s:%d, assert %s failed\n", __FILE__, __LINE__, #cond); \
0032bf9f JA	571	(td)->runstate = TD_EXITED; \
437c9b71	572	(td)->error = EFAULT; \
340fd243	573	*__foo = 0; \
0032bf9f JA	574	} \
	575	} while (0)
	576
12d9d841 JA	577	static inline int fio_fill_issue_time(struct thread_data *td)
	578	{
	579	if (td->o.read_iolog_file \|\|
	580	!td->o.disable_clat \|\| !td->o.disable_slat \|\| !td->o.disable_bw)
	581	return 1;
	582
	583	return 0;
	584	}
	585
adb02ba8 JA	586	/*
	587	* Cheesy number->string conversion, complete with carry rounding error.
	588	*/
	589	static inline char *num2str(unsigned long num, int maxlen, int base, int pow2)
	590	{
	591	char postfix[] = { ' ', 'K', 'M', 'G', 'P', 'E' };
	592	unsigned int thousand;
	593	char *buf;
	594	int i;
	595
	596	if (pow2)
	597	thousand = 1024;
	598	else
	599	thousand = 1000;
	600
	601	buf = malloc(128);
	602
	603	for (i = 0; base > 1; i++)
	604	base /= thousand;
	605
	606	do {
	607	int len, carry = 0;
	608
	609	len = sprintf(buf, "%'lu", num);
	610	if (len <= maxlen) {
	611	if (i >= 1) {
	612	buf[len] = postfix[i];
	613	buf[len + 1] = '\0';
	614	}
	615	return buf;
	616	}
	617
	618	if ((num % thousand) >= (thousand / 2))
	619	carry = 1;
	620
	621	num /= thousand;
	622	num += carry;
	623	i++;
	624	} while (i <= 5);
	625
	626	return buf;
	627	}
	628
581e7141 JA	629	static inline int __should_check_rate(struct thread_data *td,
	630	enum fio_ddir ddir)
	631	{
	632	struct thread_options *o = &td->o;
	633
	634	/*
	635	* If some rate setting was given, we need to check it
	636	*/
	637	if (o->rate[ddir] \|\| o->ratemin[ddir] \|\| o->rate_iops[ddir] \|\|
	638	o->rate_iops_min[ddir])
	639	return 1;
	640
	641	return 0;
	642	}
	643
	644	static inline int should_check_rate(struct thread_data *td,
	645	unsigned long *bytes_done)
	646	{
	647	int ret = 0;
	648
	649	if (bytes_done[0])
	650	ret \|= __should_check_rate(td, 0);
	651	if (bytes_done[1])
	652	ret \|= __should_check_rate(td, 1);
	653
	654	return ret;
	655	}
	656
ebac4655	657	#endif