X-Git-Url: https://git.kernel.dk/?p=fio.git;a=blobdiff_plain;f=fio.h;h=6abbd79e2dfd7e737f57ffdfca83a5c2e9b9bf8b;hp=c1c71465b8789ada64ebf51a8c611358ca1e6fac;hb=4cd02b3f7f4eda0ec263f04422ab6424b7693605;hpb=433afcb4fe81e775c15af9d39a6f4db8a53d693a diff --git a/fio.h b/fio.h index c1c71465..19f7713b 100644 --- a/fio.h +++ b/fio.h @@ -11,22 +11,29 @@ #include #include #include - -#include "list.h" -#include "md5.h" -#include "crc32.h" -#include "arch.h" -#include "os.h" - -#ifdef FIO_HAVE_SYSLET -#include "syslet.h" +#include +#include +#include + +#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" + +#ifdef FIO_HAVE_GUASI +#include #endif -enum fio_ddir { - DDIR_READ = 0, - DDIR_WRITE, - DDIR_SYNC, -}; +#ifdef FIO_HAVE_SOLARISAIO +#include +#endif /* * Use for maintaining statistics @@ -47,6 +54,7 @@ struct io_sample { unsigned long time; unsigned long val; enum fio_ddir ddir; + unsigned int bs; }; /* @@ -62,25 +70,30 @@ struct io_log { * When logging io actions, this matches a single sent io_u */ struct io_piece { - struct list_head list; - struct fio_file *file; + union { + struct rb_node rb_node; + struct flist_head list; + }; + union { + int fileno; + struct fio_file *file; + }; unsigned long long offset; unsigned long len; enum fio_ddir ddir; + union { + unsigned long delay; + unsigned int file_action; + }; }; -#ifdef FIO_HAVE_SYSLET -struct syslet_req { - struct syslet_uatom atom; - long ret; -}; -#endif - enum { IO_U_F_FREE = 1 << 0, IO_U_F_FLIGHT = 1 << 1, }; +struct thread_data; + /* * The io unit */ @@ -95,9 +108,13 @@ struct io_u { #ifdef FIO_HAVE_SGIO struct sg_io_hdr hdr; #endif -#ifdef FIO_HAVE_SYSLET - struct syslet_req req; +#ifdef FIO_HAVE_GUASI + guasi_req_t greq; #endif +#ifdef FIO_HAVE_SOLARISAIO + aio_result_t resultp; +#endif + void *mmap_data; }; struct timeval start_time; struct timeval issue_time; @@ -127,13 +144,19 @@ struct io_u { union { unsigned int index; unsigned int seen; + void *engine_data; }; unsigned int flags; struct fio_file *file; - struct list_head list; + struct flist_head list; + + /* + * Callback for io completion + */ + int (*end_io)(struct thread_data *, struct io_u *); }; /* @@ -145,27 +168,6 @@ enum { FIO_Q_BUSY = 2, /* no more room, call ->commit() */ }; -#define FIO_HDR_MAGIC 0xf00baaef - -enum { - VERIFY_NONE = 0, /* no verification */ - VERIFY_MD5, /* md5 sum data blocks */ - VERIFY_CRC32, /* crc32 sum data blocks */ -}; - -/* - * A header structure associated with each checksummed data block - */ -struct verify_header { - unsigned int fio_magic; - unsigned int len; - unsigned int verify_type; - union { - char md5_digest[MD5_HASH_WORDS * 4]; - unsigned long crc32; - }; -}; - struct group_run_stats { unsigned long long max_run[2], min_run[2]; unsigned long long max_bw[2], min_bw[2]; @@ -184,55 +186,33 @@ enum fio_memtype { MEM_MMAPHUGE, /* memory mapped huge file */ }; -/* - * The type of object we are working on - */ -enum fio_filetype { - FIO_TYPE_FILE = 1, /* plain file */ - FIO_TYPE_BD, /* block device */ - FIO_TYPE_CHAR, /* character device */ -}; - enum fio_ioengine_flags { FIO_SYNCIO = 1 << 0, /* io engine has synchronous ->queue */ - FIO_CPUIO = 1 << 1, /* cpu burner, doesn't do real io */ - FIO_MMAPIO = 1 << 2, /* uses memory mapped io */ - FIO_RAWIO = 1 << 3, /* some sort of direct/raw io */ - FIO_NETIO = 1 << 4, /* networked io */ - FIO_NULLIO = 1 << 5, /* no real data transfer (cpu/null) */ + FIO_RAWIO = 1 << 1, /* some sort of direct/raw io */ + FIO_DISKLESSIO = 1 << 2, /* no disk involved */ + FIO_NOEXTEND = 1 << 3, /* engine can't extend file */ + FIO_NODISKUTIL = 1 << 4, /* diskutil can't handle filename */ + FIO_UNIDIR = 1 << 5, /* engine is uni-directional */ + FIO_NOIO = 1 << 6, /* thread does only pseudo IO */ + FIO_SIGQUIT = 1 << 7, /* needs SIGQUIT to exit */ }; /* - * Each thread_data structure has a number of files associated with it, - * this structure holds state information for a single file. + * How many depth levels to log */ -struct fio_file { - /* - * A file may not be a file descriptor, let the io engine decide - */ - union { - unsigned long file_data; - int fd; - }; - char *file_name; - void *mmap; - unsigned long long file_size; - unsigned long long real_file_size; - unsigned long long file_offset; - unsigned long long last_pos; - unsigned long long last_completed_pos; - - /* - * block map for random io - */ - unsigned long *file_map; - unsigned int num_maps; - unsigned int last_free_lookup; - - unsigned int unlink; -}; +#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; @@ -255,51 +235,80 @@ struct thread_stat { 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; }; -/* - * How many depth levels to log - */ -#define FIO_IO_U_MAP_NR 8 -#define FIO_IO_U_LAT_NR 12 +struct bssplit { + unsigned int bs; + unsigned char perc; +}; -/* - * This describes a single thread/process executing a fio job. - */ -struct thread_data { +struct thread_options { + int pad; char *description; char *name; char *directory; char *filename; - char verror[80]; - pthread_t thread; - int thread_number; - int groupid; - struct thread_stat ts; - enum fio_filetype filetype; - struct fio_file *files; + 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 nr_uniq_files; - unsigned int next_file; - int error; - pid_t pid; - char *orig_buffer; - size_t orig_buffer_size; - volatile int terminate; - volatile int runstate; - enum fio_ddir ddir; - unsigned int iomix; - unsigned int ioprio; - unsigned int last_was_sync; + unsigned int open_files; + enum file_lock_mode file_lock_mode; + unsigned int lockfile_batch; - unsigned int sequential; 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; @@ -308,11 +317,10 @@ struct thread_data { 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 bs[2]; - unsigned int min_bs[2]; - unsigned int max_bs[2]; unsigned int hugepage_size; unsigned int rw_min_bs; unsigned int thinktime; @@ -320,34 +328,105 @@ struct thread_data { unsigned int thinktime_blocks; unsigned int fsync_blocks; unsigned int start_delay; - unsigned long timeout; + 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; - char *mmapfile; - int mmapfd; + unsigned int stonewall; + unsigned int new_group; unsigned int numjobs; - unsigned int iodepth; - unsigned int iodepth_low; os_cpu_mask_t cpumask; + unsigned int cpumask_set; unsigned int iolog; - unsigned int read_iolog; unsigned int rwmixcycle; - unsigned int rwmixread; - unsigned int rwmixwrite; + 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; + unsigned int ratemin; + unsigned int ratecycle; + unsigned int rate_iops; + unsigned int rate_iops_min; + + char *ioscheduler; + + /* + * CPU "io" cycle burner + */ + unsigned int cpuload; + unsigned int cpucycle; +}; + +#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; - char *ioscheduler; + + unsigned long rand_seeds[6]; os_random_state_t bsrange_state; os_random_state_t verify_state; @@ -364,118 +443,145 @@ struct thread_data { * Current IO depth and list of free and busy io_u's. */ unsigned int cur_depth; - unsigned int io_u_map[FIO_IO_U_MAP_NR]; - unsigned int io_u_lat[FIO_IO_U_LAT_NR]; - unsigned long total_io_u; - struct list_head io_u_freelist; - struct list_head io_u_busylist; - struct list_head io_u_requeues; + 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 int rate; - unsigned int ratemin; - unsigned int ratecycle; unsigned long rate_usec_cycle; long rate_pending_usleep; unsigned long rate_bytes; + unsigned long rate_blocks; struct timeval lastrate; - unsigned long runtime[2]; /* msec */ - unsigned long long io_size; - unsigned long long total_file_size; - unsigned long long start_offset; 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; - volatile int mutex; + struct fio_mutex *mutex; /* * State for random io, a bitmap of blocks done vs not done */ os_random_state_t random_state; - /* - * CPU "io" cycle burner - */ - unsigned int cpuload; - unsigned int cpucycle; - struct timeval start; /* start of this loop */ struct timeval epoch; /* time job was started */ - struct timeval end_time;/* time job ended */ + struct timeval rw_end[2]; + struct timeval last_issue; + struct timeval tv_cache; + unsigned int tv_cache_nr; + unsigned int tv_cache_mask; + unsigned int rw_end_set[2]; + unsigned int ramp_time_over; /* * read/write mixed workload state */ os_random_state_t rwmix_state; - struct timeval rwmix_switch; + unsigned long rwmix_issues; enum fio_ddir rwmix_ddir; + unsigned int ddir_nr; /* - * Pre-run and post-run shell + * IO history logs for verification. We use a tree for sorting, + * if we are overwriting. Otherwise just use a fifo. */ - char *exec_prerun; - char *exec_postrun; + struct rb_root io_hist_tree; + struct flist_head io_hist_list; /* - * IO historic logs + * For IO replaying */ - struct list_head io_hist_list; - struct list_head io_log_list; + struct flist_head io_log_list; /* - * timeout handling + * for fileservice, how often to switch to a new file */ - struct timeval timeout_end; - struct itimerval timer; + 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; }; /* - * 30 second per-io_u timeout, with 5 second intervals to avoid resetting - * the timer on each queue operation. + * roundrobin available files, or choose one at random, or do each one + * serially. */ -#define IO_U_TIMEOUT_INC 5 -#define IO_U_TIMEOUT 30 +enum { + FIO_FSERVICE_RANDOM = 1, + FIO_FSERVICE_RR = 2, + FIO_FSERVICE_SEQ = 3, +}; -#define __td_verror(td, err, msg) \ +/* + * 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; \ - snprintf(td->verror, sizeof(td->verror) - 1, "file:%s:%d, error=%s", __FILE__, __LINE__, (msg)); \ + snprintf(td->verror, sizeof(td->verror) - 1, "file:%s:%d, func=%s, error=%s", __FILE__, __LINE__, (func), (msg)); \ } while (0) -#define td_verror(td, err) __td_verror((td), (err), strerror((err))) -#define td_vmsg(td, err, msg) __td_verror((td), (err), (msg)) +#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 FILE *f_out; -extern FILE *f_err; 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; -#define td_read(td) ((td)->ddir == DDIR_READ) -#define td_write(td) ((td)->ddir == DDIR_WRITE) -#define td_rw(td) ((td)->iomix != 0) +#define td_read(td) ((td)->o.td_ddir & TD_DDIR_READ) +#define td_write(td) ((td)->o.td_ddir & TD_DDIR_WRITE) +#define td_rw(td) (((td)->o.td_ddir & TD_DDIR_RW) == TD_DDIR_RW) +#define td_random(td) ((td)->o.td_ddir & TD_DDIR_RAND) +#define file_randommap(td, f) (!(td)->o.norandommap && (f)->file_map) + +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(long)) -#define TO_MAP_BLOCK(td, f, b) ((b) - ((f)->file_offset / (td)->rw_min_bs)) -#define RAND_MAP_IDX(td, f, b) (TO_MAP_BLOCK(td, f, b) / BLOCKS_PER_MAP) -#define RAND_MAP_BIT(td, f, b) (TO_MAP_BLOCK(td, f, b) & (BLOCKS_PER_MAP - 1)) +#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) @@ -483,9 +589,9 @@ static inline int should_fsync(struct thread_data *td) { if (td->last_was_sync) return 0; - if (td->odirect) + if (td->o.odirect) return 0; - if (td_write(td) || td_rw(td) || td->override_sync) + if (td_write(td) || td_rw(td) || td->o.override_sync) return 1; return 0; @@ -507,90 +613,130 @@ struct disk_util_stat { * Per-device disk util management */ struct disk_util { - struct list_head list; + struct flist_head list; + /* If this disk is a slave, hook it into the master's + * list using this head. + */ + struct flist_head slavelist; char *name; + char *sysfs_root; char path[256]; - dev_t dev; + int major, minor; struct disk_util_stat dus; struct disk_util_stat last_dus; + /* For software raids, this entry maintains pointers to the + * entries for the slave devices. The disk_util entries for + * the slaves devices should primarily be maintained through + * the disk_list list, i.e. for memory allocation and + * de-allocation, etc. Whereas this list should be used only + * for aggregating a software RAID's disk util figures. + */ + struct flist_head slaves; + unsigned long msec; struct timeval time; + + struct fio_mutex *lock; + unsigned long users; }; -/* - * Callback for io completion - */ -typedef int (endio_handler)(struct io_u *); +static inline void disk_util_inc(struct disk_util *du) +{ + if (du) { + fio_mutex_down(du->lock); + du->users++; + fio_mutex_up(du->lock); + } +} -#define DISK_UTIL_MSEC (250) +static inline void disk_util_dec(struct disk_util *du) +{ + if (du) { + fio_mutex_down(du->lock); + du->users--; + fio_mutex_up(du->lock); + } +} -#ifndef min -#define min(a, b) ((a) < (b) ? (a) : (b)) -#endif -#ifndef max -#define max(a, b) ((a) > (b) ? (a) : (b)) -#endif +#define DISK_UTIL_MSEC (250) /* * Log exports */ +enum file_log_act { + FIO_LOG_ADD_FILE, + FIO_LOG_OPEN_FILE, + FIO_LOG_CLOSE_FILE, + FIO_LOG_UNLINK_FILE, +}; + extern int __must_check read_iolog_get(struct thread_data *, struct io_u *); -extern void write_iolog_put(struct thread_data *, struct io_u *); +extern void log_io_u(struct thread_data *, struct io_u *); +extern void log_file(struct thread_data *, struct fio_file *, enum file_log_act); extern int __must_check init_iolog(struct thread_data *td); extern void log_io_piece(struct thread_data *, struct io_u *); +extern void queue_io_piece(struct thread_data *, struct io_piece *); extern void prune_io_piece_log(struct thread_data *); extern void write_iolog_close(struct thread_data *); /* * Logging */ -extern void add_clat_sample(struct thread_data *, enum fio_ddir, unsigned long); -extern void add_slat_sample(struct thread_data *, enum fio_ddir, unsigned long); -extern void add_bw_sample(struct thread_data *, enum fio_ddir, struct timeval *); +extern void add_clat_sample(struct thread_data *, enum fio_ddir, unsigned long, + unsigned int); +extern void add_slat_sample(struct thread_data *, enum fio_ddir, unsigned long, + unsigned int); +extern void add_bw_sample(struct thread_data *, enum fio_ddir, unsigned int, + struct timeval *); extern void show_run_stats(void); extern void init_disk_util(struct thread_data *); extern void update_rusage_stat(struct thread_data *); extern void update_io_ticks(void); -extern void disk_util_timer_arm(void); extern void setup_log(struct io_log **); extern void finish_log(struct thread_data *, struct io_log *, const char *); +extern void finish_log_named(struct thread_data *, struct io_log *, const char *, const char *); extern void __finish_log(struct io_log *, const char *); -extern int setup_rate(struct thread_data *); extern struct io_log *agg_io_log[2]; extern int write_bw_log; -extern void add_agg_sample(unsigned long, enum fio_ddir); +extern void add_agg_sample(unsigned long, enum fio_ddir, unsigned int); /* * Time functions */ -extern unsigned long utime_since(struct timeval *, struct timeval *); -extern unsigned long utime_since_now(struct timeval *); +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_sleep(unsigned int); +extern void usec_spin(unsigned int); extern void usec_sleep(struct thread_data *, unsigned long); -extern void rate_throttle(struct thread_data *, unsigned long, unsigned int, int); +extern void rate_throttle(struct thread_data *, unsigned long, unsigned int); 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 functions + * Init/option functions */ extern int __must_check parse_options(int, char **); -extern int __must_check init_random_state(struct thread_data *); - -/* - * File setup/shutdown - */ -extern void close_files(struct thread_data *); -extern int __must_check setup_files(struct thread_data *); -extern int __must_check open_files(struct thread_data *); -extern int __must_check file_invalidate_cache(struct thread_data *, struct fio_file *); +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 @@ -598,6 +744,19 @@ extern int __must_check file_invalidate_cache(struct thread_data *, struct fio_f extern void print_thread_status(void); extern void print_status_init(int); +/* + * disk util stuff + */ +#ifdef FIO_HAVE_DISK_UTIL +extern void show_disk_util(void); +extern void init_disk_util(struct thread_data *); +extern void update_io_ticks(void); +#else +#define show_disk_util() +#define init_disk_util(td) +#define update_io_ticks() +#endif + /* * Thread life cycle. Once a thread has a runstate beyond TD_INITIALIZED, it * will never back again. It may cycle between running/verififying/fsyncing. @@ -608,19 +767,16 @@ enum { TD_NOT_CREATED = 0, TD_CREATED, TD_INITIALIZED, + TD_RAMP, TD_RUNNING, + TD_PRE_READING, TD_VERIFYING, TD_FSYNCING, TD_EXITED, TD_REAPED, }; -/* - * Verify helpers - */ -extern void populate_verify_io_u(struct thread_data *, struct io_u *); -extern int __must_check get_next_verify(struct thread_data *td, struct io_u *); -extern int __must_check verify_io_u(struct io_u *); +extern void td_set_runstate(struct thread_data *, int); /* * Memory helpers @@ -633,16 +789,24 @@ extern void free_io_mem(struct thread_data *); /* * io unit handling */ -#define queue_full(td) list_empty(&(td)->io_u_freelist) +#define queue_full(td) flist_empty(&(td)->io_u_freelist) extern struct io_u *__get_io_u(struct thread_data *); extern struct io_u *get_io_u(struct thread_data *); extern void put_io_u(struct thread_data *, struct io_u *); extern void requeue_io_u(struct thread_data *, struct io_u **); -extern long __must_check io_u_sync_complete(struct thread_data *, struct io_u *, endio_handler *); -extern long __must_check io_u_queued_complete(struct thread_data *, int, endio_handler *); +extern long __must_check io_u_sync_complete(struct thread_data *, struct io_u *); +extern long __must_check io_u_queued_complete(struct thread_data *, int); extern void io_u_queued(struct thread_data *, struct io_u *); -extern void io_u_init_timeout(void); -extern void io_u_set_timeout(struct thread_data *); +extern void io_u_log_error(struct thread_data *, struct io_u *); +extern void io_u_mark_depth(struct thread_data *, unsigned int); +extern void io_u_fill_buffer(struct thread_data *td, struct io_u *, unsigned int); +void io_u_mark_complete(struct thread_data *, unsigned int); +void io_u_mark_submit(struct thread_data *, unsigned int); + +/* + * Reset stats after ramp time completes + */ +extern void reset_all_stats(struct thread_data *); /* * io engine entry points @@ -651,44 +815,22 @@ extern int __must_check td_io_init(struct thread_data *); extern int __must_check td_io_prep(struct thread_data *, struct io_u *); extern int __must_check td_io_queue(struct thread_data *, struct io_u *); extern int __must_check td_io_sync(struct thread_data *, struct fio_file *); -extern int __must_check td_io_getevents(struct thread_data *, int, int, struct timespec *); +extern int __must_check td_io_getevents(struct thread_data *, unsigned int, unsigned int, struct timespec *); extern int __must_check td_io_commit(struct thread_data *); +extern int __must_check td_io_open_file(struct thread_data *, struct fio_file *); +extern int td_io_close_file(struct thread_data *, struct fio_file *); +extern int __must_check td_io_get_file_size(struct thread_data *, struct fio_file *); /* - * This is a pretty crappy semaphore implementation, but with the use that fio - * has (just signalling start/go conditions), it doesn't have to be better. - * Naturally this would not work for any type of contended semaphore or - * for real locking. + * blktrace support */ -static inline void fio_sem_init(volatile int *sem, int val) -{ - *sem = val; -} - -static inline void fio_sem_down(volatile int *sem) -{ - while (*sem == 0) - usleep(10000); - - (*sem)--; -} - -static inline void fio_sem_up(volatile int *sem) -{ - (*sem)++; -} - -/* - * If logging output to a file, stderr should go to both stderr and f_err - */ -#define log_err(args...) do { \ - fprintf(f_err, ##args); \ - if (f_err != stderr) \ - fprintf(stderr, ##args); \ - } while (0) +#ifdef FIO_HAVE_BLKTRACE +extern int is_blktrace(const char *); +extern int load_blktrace(struct thread_data *, const char *); +#endif struct ioengine_ops { - struct list_head list; + struct flist_head list; char name[16]; int version; int flags; @@ -697,16 +839,18 @@ struct ioengine_ops { int (*prep)(struct thread_data *, struct io_u *); int (*queue)(struct thread_data *, struct io_u *); int (*commit)(struct thread_data *); - int (*getevents)(struct thread_data *, int, int, struct timespec *); + int (*getevents)(struct thread_data *, unsigned int, unsigned int, struct timespec *); struct io_u *(*event)(struct thread_data *, int); int (*cancel)(struct thread_data *, struct io_u *); void (*cleanup)(struct thread_data *); + int (*open_file)(struct thread_data *, struct fio_file *); + int (*close_file)(struct thread_data *, struct fio_file *); + int (*get_file_size)(struct thread_data *, struct fio_file *); void *data; void *dlhandle; - unsigned long priv; }; -#define FIO_IOOPS_VERSION 5 +#define FIO_IOOPS_VERSION 10 extern struct ioengine_ops *load_ioengine(struct thread_data *, const char *); extern void register_ioengine(struct ioengine_ops *); @@ -723,7 +867,10 @@ extern void close_ioengine(struct thread_data *); #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) \ - for ((i) = 0, (f) = &(td)->files[0]; (i) < (int) (td)->nr_files; (i)++, (f)++) + 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)) { \ @@ -735,4 +882,46 @@ extern void close_ioengine(struct thread_data *); } \ } while (0) +static inline void fio_file_reset(struct fio_file *f) +{ + f->last_free_lookup = 0; + f->last_pos = f->file_offset; + if (f->file_map) + memset(f->file_map, 0, f->num_maps * sizeof(int)); +} + +static inline void clear_error(struct thread_data *td) +{ + td->error = 0; + td->verror[0] = '\0'; +} + +#ifdef FIO_INC_DEBUG +static inline void dprint_io_u(struct io_u *io_u, const char *p) +{ + struct fio_file *f = io_u->file; + + dprint(FD_IO, "%s: io_u %p: off=%llu/len=%lu/ddir=%d", p, io_u, + (unsigned long long) io_u->offset, + io_u->buflen, io_u->ddir); + if (fio_debug & (1 << FD_IO)) { + if (f) + log_info("/%s", f->file_name); + + log_info("\n"); + } +} +#else +#define dprint_io_u(io_u, p) +#endif + +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; +} + #endif