[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"
#include "helpers.h"
#include "options.h"
#include "profile.h"
#include "time.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];
	unsigned int kb_base;
};

/*
 * 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

#define MAX_PATTERN_SIZE 512

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;

	unsigned int kb_base;
};

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 kb_base;
	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;
	char verify_pattern[MAX_PATTERN_SIZE];
	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 fallocate;
	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;
	enum fio_cs clocksource;

	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;

	/*
	 * 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;
};

#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_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 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;

	unsigned int sync_file_range_nr;

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

/*
 * Init/option functions
 */
extern int __must_check parse_options(int, char **);
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 **);
#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)->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"
c5c8bd5c	31	#include "helpers.h"
07b3232d	32	#include "options.h"
7eb36574	33	#include "profile.h"
c223da83	34	#include "time.h"
ebac4655	35
609342ff DL	36	#ifdef FIO_HAVE_GUASI
	37	#include <guasi.h>
	38	#endif
	39
417f0068 JA	40	#ifdef FIO_HAVE_SOLARISAIO
	41	#include <sys/asynch.h>
	42	#endif
	43
ebac4655	44	struct group_run_stats {
9104f874 JA	45	unsigned long long max_run[2], min_run[2];
9104f874 JA	46	unsigned long long max_bw[2], min_bw[2];
e9b2a3fa	47	unsigned long long io_kb[2];
9104f874	48	unsigned long long agg[2];
90fef2d1	49	unsigned int kb_base;
ebac4655 JA	50	};
ebac4655 JA	51
e9c047a0 JA	52	/*
	53	* What type of allocation to use for io buffers
	54	*/
	55	enum fio_memtype {
	56	MEM_MALLOC = 0, /* ordinary malloc */
	57	MEM_SHM, /* use shared memory segments */
74b025b0	58	MEM_SHMHUGE, /* use shared memory segments with huge pages */
e9c047a0	59	MEM_MMAP, /* use anonynomous mmap */
d0bdaf49	60	MEM_MMAPHUGE, /* memory mapped huge file */
e9c047a0 JA	61	};
e9c047a0 JA	62
b2560f3c JA	63	/*
	64	* How many depth levels to log
	65	*/
	66	#define FIO_IO_U_MAP_NR 8
a9a18a3f JA	67	#define FIO_IO_U_LAT_U_NR 10
a9a18a3f JA	68	#define FIO_IO_U_LAT_M_NR 12
b2560f3c	69
0e92f873 RR	70	#define MAX_PATTERN_SIZE 512
0e92f873 RR	71
079ad09b	72	struct thread_stat {
756867bd JA	73	char *name;
	74	char *verror;
	75	int error;
	76	int groupid;
	77	pid_t pid;
	78	char *description;
6586ee89	79	int members;
756867bd	80
079ad09b JA	81	struct io_log *slat_log;
	82	struct io_log *clat_log;
	83	struct io_log *bw_log;
	84
	85	/*
	86	* bandwidth and latency stats
	87	*/
	88	struct io_stat clat_stat[2]; /* completion latency */
	89	struct io_stat slat_stat[2]; /* submission latency */
	90	struct io_stat bw_stat[2]; /* bandwidth stats */
	91
	92	unsigned long long stat_io_bytes[2];
	93	struct timeval stat_sample_time[2];
	94
	95	/*
	96	* fio system usage accounting
	97	*/
	98	struct rusage ru_start;
	99	struct rusage ru_end;
	100	unsigned long usr_time;
	101	unsigned long sys_time;
	102	unsigned long ctx;
81887d5d	103	unsigned long minf, majf;
079ad09b	104
b2560f3c JA	105	/*
	106	* IO depth and latency stats
	107	*/
	108	unsigned int io_u_map[FIO_IO_U_MAP_NR];
838bc709 JA	109	unsigned int io_u_submit[FIO_IO_U_MAP_NR];
838bc709 JA	110	unsigned int io_u_complete[FIO_IO_U_MAP_NR];
04a0feae JA	111	unsigned int io_u_lat_u[FIO_IO_U_LAT_U_NR];
04a0feae JA	112	unsigned int io_u_lat_m[FIO_IO_U_LAT_M_NR];
b3605062	113	unsigned long total_io_u[2];
30061b97	114	unsigned long short_io_u[2];
838bc709 JA	115	unsigned long total_submit;
838bc709 JA	116	unsigned long total_complete;
756867bd JA	117
	118	unsigned long long io_bytes[2];
	119	unsigned long runtime[2];
	120	unsigned long total_run_time;
f2bba182 RR	121
	122	/*
	123	* IO Error related stats
	124	*/
	125	unsigned continue_on_error;
	126	unsigned long total_err_count;
	127	int first_error;
90fef2d1 JA	128
90fef2d1 JA	129	unsigned int kb_base;
b2560f3c	130	};
71619dc2	131
564ca972 JA	132	struct bssplit {
	133	unsigned int bs;
	134	unsigned char perc;
	135	};
	136
2dc1bbeb	137	struct thread_options {
b1ec1da6	138	int pad;
61697c37	139	char *description;
b4692828	140	char *name;
ef899b63	141	char *directory;
13f8e2d2	142	char *filename;
2dc1bbeb	143	char *opendir;
09629a90	144	char *ioengine;
413dd459	145	enum td_ddir td_ddir;
90fef2d1	146	unsigned int kb_base;
211097b2	147	unsigned int ddir_nr;
2dc1bbeb JA	148	unsigned int iodepth;
	149	unsigned int iodepth_low;
	150	unsigned int iodepth_batch;
4950421a	151	unsigned int iodepth_batch_complete;
2dc1bbeb JA	152
2dc1bbeb JA	153	unsigned long long size;
aa31f1f1	154	unsigned int fill_device;
2dc1bbeb JA	155	unsigned long long file_size_low;
	156	unsigned long long file_size_high;
	157	unsigned long long start_offset;
	158
	159	unsigned int bs[2];
2b7a01d0	160	unsigned int ba[2];
2dc1bbeb JA	161	unsigned int min_bs[2];
2dc1bbeb JA	162	unsigned int max_bs[2];
720e84ad JA	163	struct bssplit *bssplit[2];
720e84ad JA	164	unsigned int bssplit_nr[2];
2dc1bbeb JA	165
	166	unsigned int nr_files;
	167	unsigned int open_files;
4d4e80f2	168	enum file_lock_mode file_lock_mode;
29c1349f	169	unsigned int lockfile_batch;
e9c047a0	170
9158d2f7 JA	171	unsigned int odirect;
	172	unsigned int invalidate_cache;
	173	unsigned int create_serialize;
	174	unsigned int create_fsync;
814452bd	175	unsigned int create_on_open;
9158d2f7	176	unsigned int end_fsync;
afad68f7	177	unsigned int pre_read;
9158d2f7 JA	178	unsigned int sync_io;
9158d2f7 JA	179	unsigned int verify;
e84c73a8	180	unsigned int do_verify;
160b966d	181	unsigned int verifysort;
a59e170d JA	182	unsigned int verify_interval;
a59e170d JA	183	unsigned int verify_offset;
0e92f873	184	char verify_pattern[MAX_PATTERN_SIZE];
90059d65	185	unsigned int verify_pattern_bytes;
a12a3b4d	186	unsigned int verify_fatal;
e8462bd8	187	unsigned int verify_async;
9158d2f7 JA	188	unsigned int use_thread;
	189	unsigned int unlink;
	190	unsigned int do_disk_util;
	191	unsigned int override_sync;
	192	unsigned int rand_repeatable;
	193	unsigned int write_lat_log;
	194	unsigned int write_bw_log;
bb8895e0	195	unsigned int norandommap;
2b386d25	196	unsigned int softrandommap;
690adba3	197	unsigned int bs_unaligned;
ebb1415f	198	unsigned int fsync_on_close;
e9c047a0	199
56bb17f2	200	unsigned int hugepage_size;
a00735e6	201	unsigned int rw_min_bs;
ebac4655	202	unsigned int thinktime;
48097d5c	203	unsigned int thinktime_spin;
9c1f7434	204	unsigned int thinktime_blocks;
ebac4655	205	unsigned int fsync_blocks;
5f9099ea	206	unsigned int fdatasync_blocks;
ebac4655	207	unsigned int start_delay;
0db26797	208	unsigned long long timeout;
721938ae	209	unsigned long long ramp_time;
ebac4655	210	unsigned int overwrite;
ebac4655	211	unsigned int bw_avg_time;
ebac4655	212	unsigned int loops;
20dc95c4 JA	213	unsigned long long zone_size;
20dc95c4 JA	214	unsigned long long zone_skip;
e9c047a0	215	enum fio_memtype mem_type;
d529ee19	216	unsigned int mem_align;
2dc1bbeb	217
ebac4655	218	unsigned int stonewall;
b3d62a75	219	unsigned int new_group;
ebac4655	220	unsigned int numjobs;
ebac4655	221	os_cpu_mask_t cpumask;
375b2695	222	unsigned int cpumask_set;
e8462bd8 JA	223	os_cpu_mask_t verify_cpumask;
e8462bd8 JA	224	unsigned int verify_cpumask_set;
aea47d44	225	unsigned int iolog;
a6ccc7be	226	unsigned int rwmixcycle;
e47f799f	227	unsigned int rwmix[2];
b6f4d880	228	unsigned int nice;
0aabe160	229	unsigned int file_service_type;
b2560f3c	230	unsigned int group_reporting;
d2f3ac35	231	unsigned int fadvise_hint;
7bc8c2cf	232	unsigned int fallocate;
e9459e5a	233	unsigned int zero_buffers;
5973cafb	234	unsigned int refill_buffers;
cf4464ca	235	unsigned int time_based;
9520ebb9 JA	236	unsigned int disable_clat;
	237	unsigned int disable_slat;
	238	unsigned int disable_bw;
993bf48b	239	unsigned int gtod_reduce;
be4ecfdf JA	240	unsigned int gtod_cpu;
be4ecfdf JA	241	unsigned int gtod_offload;
c223da83	242	enum fio_cs clocksource;
aea47d44	243
076efc7c JA	244	char *read_iolog_file;
076efc7c JA	245	char *write_iolog_file;
e3cedca7 JA	246	char *bw_log_file;
e3cedca7 JA	247	char *lat_log_file;
2dc1bbeb JA	248
	249	/*
	250	* Pre-run and post-run shell
	251	*/
	252	char *exec_prerun;
	253	char *exec_postrun;
	254
581e7141 JA	255	unsigned int rate[2];
581e7141 JA	256	unsigned int ratemin[2];
2dc1bbeb	257	unsigned int ratecycle;
581e7141 JA	258	unsigned int rate_iops[2];
581e7141 JA	259	unsigned int rate_iops_min[2];
2dc1bbeb JA	260
	261	char *ioscheduler;
	262
	263	/*
	264	* CPU "io" cycle burner
	265	*/
	266	unsigned int cpuload;
	267	unsigned int cpucycle;
f2bba182 RR	268
	269	/*
	270	* I/O Error handling
	271	*/
	272	unsigned int continue_on_error;
9ac8a797 JA	273
	274	/*
	275	* Benchmark profile type
	276	*/
79d16311	277	char *profile;
a696fa2a JA	278
	279	/*
	280	* blkio cgroup support
	281	*/
a696fa2a JA	282	char *cgroup;
a696fa2a JA	283	unsigned int cgroup_weight;
7de87099	284	unsigned int cgroup_nodelete;
e0b0d892 JA	285
	286	unsigned int uid;
	287	unsigned int gid;
44f29692 JA	288
44f29692 JA	289	unsigned int sync_file_range;
2dc1bbeb JA	290	};
2dc1bbeb JA	291
e4e33258 JA	292	#define FIO_VERROR_SIZE 128
e4e33258 JA	293
2dc1bbeb JA	294	/*
	295	* This describes a single thread/process executing a fio job.
	296	*/
	297	struct thread_data {
	298	struct thread_options o;
e4e33258	299	char verror[FIO_VERROR_SIZE];
2dc1bbeb JA	300	pthread_t thread;
	301	int thread_number;
	302	int groupid;
	303	struct thread_stat ts;
126d65c6	304	struct fio_file **files;
dd87b2c9	305	unsigned int files_size;
2dc1bbeb JA	306	unsigned int files_index;
2dc1bbeb JA	307	unsigned int nr_open_files;
1020a139	308	unsigned int nr_done_files;
2dc1bbeb JA	309	unsigned int nr_normal_files;
	310	union {
	311	unsigned int next_file;
	312	os_random_state_t next_file_state;
	313	};
	314	int error;
20e354ef	315	int done;
2dc1bbeb JA	316	pid_t pid;
	317	char *orig_buffer;
	318	size_t orig_buffer_size;
	319	volatile int terminate;
	320	volatile int runstate;
	321	unsigned int ioprio;
ac684785	322	unsigned int ioprio_set;
2dc1bbeb JA	323	unsigned int last_was_sync;
	324
	325	char *mmapfile;
	326	int mmapfd;
	327
843a7413 JA	328	void *iolog_buf;
843a7413 JA	329	FILE *iolog_f;
ebac4655	330
da86774e	331	char *sysfs_root;
da86774e	332
5bfc35d7 JA	333	unsigned long rand_seeds[6];
5bfc35d7 JA	334
6dfd46b9 JA	335	os_random_state_t bsrange_state;
6dfd46b9 JA	336	os_random_state_t verify_state;
ebac4655 JA	337
	338	int shm_id;
	339
e9c047a0 JA	340	/*
	341	* IO engine hooks, contains everything needed to submit an io_u
	342	* to any of the available IO engines.
	343	*/
2866c82d	344	struct ioengine_ops *io_ops;
ebac4655	345
e9c047a0 JA	346	/*
	347	* Current IO depth and list of free and busy io_u's.
	348	*/
ebac4655	349	unsigned int cur_depth;
d8005759	350	unsigned int io_u_queued;
01743ee1 JA	351	struct flist_head io_u_freelist;
	352	struct flist_head io_u_busylist;
	353	struct flist_head io_u_requeues;
e8462bd8 JA	354	pthread_mutex_t io_u_lock;
	355	pthread_cond_t free_cond;
	356
	357	/*
	358	* async verify offload
	359	*/
	360	struct flist_head verify_list;
	361	pthread_t *verify_threads;
	362	unsigned int nr_verify_threads;
	363	pthread_cond_t verify_cond;
	364	int verify_thread_exit;
ebac4655	365
e9c047a0 JA	366	/*
	367	* Rate state
	368	*/
ba3e4e0c	369	unsigned long rate_nsec_cycle[2];
581e7141 JA	370	long rate_pending_usleep[2];
	371	unsigned long rate_bytes[2];
	372	unsigned long rate_blocks[2];
	373	struct timeval lastrate[2];
ebac4655	374
ebac4655 JA	375	unsigned long long total_io_size;
ebac4655 JA	376
755200a3	377	unsigned long io_issues[2];
9104f874 JA	378	unsigned long long io_blocks[2];
9104f874 JA	379	unsigned long long io_bytes[2];
48f5abd3	380	unsigned long long io_skip_bytes;
9104f874	381	unsigned long long this_io_bytes[2];
079ad09b	382	unsigned long long zone_bytes;
cdd18ad8	383	struct fio_mutex *mutex;
ebac4655	384
e9c047a0 JA	385	/*
	386	* State for random io, a bitmap of blocks done vs not done
	387	*/
6dfd46b9	388	os_random_state_t random_state;
ebac4655	389
ebac4655 JA	390	struct timeval start; /* start of this loop */
ebac4655 JA	391	struct timeval epoch; /* time job was started */
a61eddec	392	struct timeval last_issue;
993bf48b JA	393	struct timeval tv_cache;
	394	unsigned int tv_cache_nr;
	395	unsigned int tv_cache_mask;
721938ae	396	unsigned int ramp_time_over;
ebac4655	397
e9c047a0 JA	398	/*
	399	* read/write mixed workload state
	400	*/
6dfd46b9	401	os_random_state_t rwmix_state;
e4928662	402	unsigned long rwmix_issues;
e9c047a0	403	enum fio_ddir rwmix_ddir;
211097b2	404	unsigned int ddir_nr;
a6ccc7be	405
e9c047a0	406	/*
8de8f047 JA	407	* IO history logs for verification. We use a tree for sorting,
8de8f047 JA	408	* if we are overwriting. Otherwise just use a fifo.
e9c047a0	409	*/
4b87898e	410	struct rb_root io_hist_tree;
01743ee1	411	struct flist_head io_hist_list;
8de8f047 JA	412
	413	/*
	414	* For IO replaying
	415	*/
01743ee1	416	struct flist_head io_log_list;
433afcb4	417
1907dbc6 JA	418	/*
	419	* for fileservice, how often to switch to a new file
	420	*/
	421	unsigned int file_service_nr;
	422	unsigned int file_service_left;
	423	struct fio_file *file_service_file;
9c60ce64	424
44f29692 JA	425	unsigned int sync_file_range_nr;
44f29692 JA	426
9c60ce64 JA	427	/*
	428	* For generating file sizes
	429	*/
	430	os_random_state_t file_size_state;
f2bba182 RR	431
	432	/*
	433	* Error counts
	434	*/
	435	unsigned int total_err_count;
	436	int first_error;
15dc1934 JA	437
	438	/*
	439	* Can be overloaded by profiles
	440	*/
7eb36574	441	struct prof_io_ops prof_io_ops;
58c55ba0	442	void *prof_data;
ebac4655 JA	443	};
ebac4655 JA	444
e592a06b AC	445	/*
	446	* when should interactive ETA output be generated
	447	*/
	448	enum {
	449	FIO_ETA_AUTO,
	450	FIO_ETA_ALWAYS,
	451	FIO_ETA_NEVER,
	452	};
	453
e1161c32	454	#define __td_verror(td, err, msg, func) \
ebac4655	455	do { \
19abcd3d JA	456	if ((td)->error) \
19abcd3d JA	457	break; \
ebac4655 JA	458	int e = (err); \
ebac4655 JA	459	(td)->error = e; \
f2bba182 RR	460	if (!(td)->first_error) \
f2bba182 RR	461	snprintf(td->verror, sizeof(td->verror) - 1, "file:%s:%d, func=%s, error=%s", __FILE__, __LINE__, (func), (msg)); \
ebac4655 JA	462	} while (0)
ebac4655 JA	463
b990b5c0	464
f2bba182 RR	465	#define td_clear_error(td) \
f2bba182 RR	466	(td)->error = 0;
e1161c32 JA	467	#define td_verror(td, err, func) \
	468	__td_verror((td), (err), strerror((err)), (func))
	469	#define td_vmsg(td, err, msg, func) \
	470	__td_verror((td), (err), (msg), (func))
b990b5c0	471
ebac4655 JA	472	extern int exitall_on_terminate;
ebac4655 JA	473	extern int thread_number;
9cedf167	474	extern int nr_process, nr_thread;
ebac4655 JA	475	extern int shm_id;
ebac4655 JA	476	extern int groupid;
c6ae0a5b	477	extern int terse_output;
53cdc686	478	extern int temp_stall_ts;
1e97cce9	479	extern unsigned long long mlock_size;
cfc99db7	480	extern unsigned long page_mask, page_size;
4241ea8f	481	extern int read_only;
e592a06b	482	extern int eta_print;
d3eeeabc	483	extern unsigned long done_secs;
01f06b63	484	extern char *job_section;
be4ecfdf JA	485	extern int fio_gtod_offload;
be4ecfdf JA	486	extern int fio_gtod_cpu;
c223da83	487	extern enum fio_cs fio_clock_source;
ebac4655 JA	488
	489	extern struct thread_data *threads;
	490
7101d9c2 JA	491	static inline void fio_ro_check(struct thread_data td, struct io_u io_u)
	492	{
	493	assert(!(io_u->ddir == DDIR_WRITE && !td_write(td)));
	494	}
	495
de605666	496	#define BLOCKS_PER_MAP (8 * sizeof(int))
aec2de20 JA	497	#define TO_MAP_BLOCK(f, b) (b)
	498	#define RAND_MAP_IDX(f, b) (TO_MAP_BLOCK(f, b) / BLOCKS_PER_MAP)
	499	#define RAND_MAP_BIT(f, b) (TO_MAP_BLOCK(f, b) & (BLOCKS_PER_MAP - 1))
ebac4655 JA	500
	501	#define MAX_JOBS (1024)
	502
1ec99eea	503	#define td_non_fatal_error(e) ((e) == EIO \|\| (e) == EILSEQ)
f2bba182 RR	504
	505	static inline void update_error_count(struct thread_data *td, int err)
	506	{
	507	td->total_err_count++;
	508	if (td->total_err_count == 1)
	509	td->first_error = err;
	510	}
	511
87dc1ab1 JA	512	static inline int should_fsync(struct thread_data *td)
	513	{
	514	if (td->last_was_sync)
	515	return 0;
2dc1bbeb	516	if (td->o.odirect)
87dc1ab1	517	return 0;
2dc1bbeb	518	if (td_write(td) \|\| td_rw(td) \|\| td->o.override_sync)
87dc1ab1 JA	519	return 1;
	520
	521	return 0;
	522	}
	523
8914a9d8	524	/*
214e1eca	525	* Init/option functions
8914a9d8	526	*/
72cb971b	527	extern int __must_check parse_options(int, char **);
3b8b7135	528	extern int fio_options_parse(struct thread_data , char *, int);
74929ac2	529	extern void fio_keywords_init(void);
214e1eca JA	530	extern int fio_cmd_option_parse(struct thread_data , const char , char *);
	531	extern void fio_fill_default_options(struct thread_data *);
	532	extern int fio_show_option_help(const char *);
	533	extern void fio_options_dup_and_init(struct option *);
d23bb327 JA	534	extern void options_mem_dupe(struct thread_data *);
d23bb327 JA	535	extern void options_mem_free(struct thread_data *);
5bfc35d7	536	extern void td_fill_rand_seeds(struct thread_data *);
79d16311	537	extern void add_job_opts(const char **);
214e1eca	538	#define FIO_GETOPT_JOB 0x89988998
07b3232d	539	#define FIO_NR_OPTIONS (FIO_MAX_OPTS + 128)
8914a9d8	540
263e529f JA	541	/*
	542	* ETA/status stuff
	543	*/
	544	extern void print_thread_status(void);
	545	extern void print_status_init(int);
	546
	547	/*
	548	* Thread life cycle. Once a thread has a runstate beyond TD_INITIALIZED, it
	549	* will never back again. It may cycle between running/verififying/fsyncing.
	550	* Once the thread reaches TD_EXITED, it is just waiting for the core to
	551	* reap it.
	552	*/
	553	enum {
	554	TD_NOT_CREATED = 0,
	555	TD_CREATED,
	556	TD_INITIALIZED,
b29ee5b3	557	TD_RAMP,
263e529f	558	TD_RUNNING,
b0f65863	559	TD_PRE_READING,
263e529f JA	560	TD_VERIFYING,
	561	TD_FSYNCING,
	562	TD_EXITED,
	563	TD_REAPED,
	564	};
	565
b29ee5b3 JA	566	extern void td_set_runstate(struct thread_data *, int);
b29ee5b3 JA	567
2f9ade3c JA	568	/*
	569	* Memory helpers
	570	*/
b2fdda43	571	extern int __must_check fio_pin_memory(void);
2f9ade3c	572	extern void fio_unpin_memory(void);
b2fdda43	573	extern int __must_check allocate_io_mem(struct thread_data *);
2f9ade3c JA	574	extern void free_io_mem(struct thread_data *);
2f9ade3c JA	575
b29ee5b3 JA	576	/*
	577	* Reset stats after ramp time completes
	578	*/
	579	extern void reset_all_stats(struct thread_data *);
	580
fb7b71a3 JA	581	/*
	582	* blktrace support
	583	*/
5e62c22a	584	#ifdef FIO_HAVE_BLKTRACE
fb7b71a3 JA	585	extern int is_blktrace(const char *);
fb7b71a3 JA	586	extern int load_blktrace(struct thread_data , const char );
5e62c22a	587	#endif
fb7b71a3	588
2866c82d JA	589	/*
	590	* Mark unused variables passed to ops functions as unused, to silence gcc
	591	*/
	592	#define fio_unused __attribute((__unused__))
5f350952 JA	593	#define fio_init __attribute__((constructor))
5f350952 JA	594	#define fio_exit __attribute__((destructor))
2866c82d	595
34572e28 JA	596	#define for_each_td(td, i) \
34572e28 JA	597	for ((i) = 0, (td) = &threads[0]; (i) < (int) thread_number; (i)++, (td)++)
53cdc686	598	#define for_each_file(td, f, i) \
691c8fb0 JA	599	if ((td)->files_index) \
	600	for ((i) = 0, (f) = (td)->files[0]; \
	601	(i) < (td)->o.nr_files && ((f) = (td)->files[i]) != NULL; \
	602	(i)++)
53cdc686	603
0032bf9f JA	604	#define fio_assert(td, cond) do { \
0032bf9f JA	605	if (!(cond)) { \
340fd243	606	int *__foo = NULL; \
0032bf9f JA	607	fprintf(stderr, "file:%s:%d, assert %s failed\n", __FILE__, __LINE__, #cond); \
0032bf9f JA	608	(td)->runstate = TD_EXITED; \
437c9b71	609	(td)->error = EFAULT; \
340fd243	610	*__foo = 0; \
0032bf9f JA	611	} \
	612	} while (0)
	613
12d9d841 JA	614	static inline int fio_fill_issue_time(struct thread_data *td)
	615	{
	616	if (td->o.read_iolog_file \|\|
	617	!td->o.disable_clat \|\| !td->o.disable_slat \|\| !td->o.disable_bw)
	618	return 1;
	619
	620	return 0;
	621	}
	622
adb02ba8 JA	623	/*
	624	* Cheesy number->string conversion, complete with carry rounding error.
	625	*/
	626	static inline char *num2str(unsigned long num, int maxlen, int base, int pow2)
	627	{
	628	char postfix[] = { ' ', 'K', 'M', 'G', 'P', 'E' };
	629	unsigned int thousand;
	630	char *buf;
	631	int i;
	632
	633	if (pow2)
	634	thousand = 1024;
	635	else
	636	thousand = 1000;
	637
	638	buf = malloc(128);
	639
	640	for (i = 0; base > 1; i++)
	641	base /= thousand;
	642
	643	do {
	644	int len, carry = 0;
	645
	646	len = sprintf(buf, "%'lu", num);
	647	if (len <= maxlen) {
	648	if (i >= 1) {
	649	buf[len] = postfix[i];
	650	buf[len + 1] = '\0';
	651	}
	652	return buf;
	653	}
	654
	655	if ((num % thousand) >= (thousand / 2))
	656	carry = 1;
	657
	658	num /= thousand;
	659	num += carry;
	660	i++;
	661	} while (i <= 5);
	662
	663	return buf;
	664	}
	665
581e7141 JA	666	static inline int __should_check_rate(struct thread_data *td,
	667	enum fio_ddir ddir)
	668	{
	669	struct thread_options *o = &td->o;
	670
	671	/*
	672	* If some rate setting was given, we need to check it
	673	*/
	674	if (o->rate[ddir] \|\| o->ratemin[ddir] \|\| o->rate_iops[ddir] \|\|
	675	o->rate_iops_min[ddir])
	676	return 1;
	677
	678	return 0;
	679	}
	680
	681	static inline int should_check_rate(struct thread_data *td,
	682	unsigned long *bytes_done)
	683	{
	684	int ret = 0;
	685
	686	if (bytes_done[0])
	687	ret \|= __should_check_rate(td, 0);
	688	if (bytes_done[1])
	689	ret \|= __should_check_rate(td, 1);
	690
	691	return ret;
	692	}
	693
d529ee19 JA	694	static inline int is_power_of_2(unsigned int val)
	695	{
	696	return (val != 0 && ((val & (val - 1)) == 0));
	697	}
	698
e8462bd8 JA	699	/*
	700	* We currently only need to do locking if we have verifier threads
	701	* accessing our internal structures too
	702	*/
	703	static inline void td_io_u_lock(struct thread_data *td)
	704	{
	705	if (td->o.verify_async)
	706	pthread_mutex_lock(&td->io_u_lock);
	707	}
	708
	709	static inline void td_io_u_unlock(struct thread_data *td)
	710	{
	711	if (td->o.verify_async)
	712	pthread_mutex_unlock(&td->io_u_lock);
	713	}
	714
	715	static inline void td_io_u_free_notify(struct thread_data *td)
	716	{
	717	if (td->o.verify_async)
	718	pthread_cond_signal(&td->free_cond);
	719	}
	720
ebac4655	721	#endif