X-Git-Url: https://git.kernel.dk/?p=fio.git;a=blobdiff_plain;f=fio.h;h=20d14082d3281b03e44f04823de5f714a81f65da;hp=faa0cfb7427a5549bc31119d5808fd222dfd968e;hb=4241ea8fb0606f9bd5b7ed1278d3b7825883acf6;hpb=480936dd2270baa028a8ad8ab20719caa0429968 diff --git a/fio.h b/fio.h index faa0cfb7..20d14082 100644 --- a/fio.h +++ b/fio.h @@ -11,21 +11,30 @@ #include #include #include +#include +#include +#include "compiler/compiler.h" #include "list.h" -#include "md5.h" -#include "crc32.h" -#include "arch.h" -#include "os.h" +#include "fifo.h" +#include "rbtree.h" +#include "arch/arch.h" +#include "os/os.h" +#include "mutex.h" #ifdef FIO_HAVE_SYSLET #include "syslet.h" #endif +#ifdef FIO_HAVE_GUASI +#include +#endif + enum fio_ddir { DDIR_READ = 0, DDIR_WRITE, DDIR_SYNC, + DDIR_INVAL = -1, }; enum td_ddir { @@ -34,7 +43,7 @@ enum td_ddir { TD_DDIR_RAND = 1 << 2, TD_DDIR_RW = TD_DDIR_READ | TD_DDIR_WRITE, TD_DDIR_RANDREAD = TD_DDIR_READ | TD_DDIR_RAND, - TD_DDIR_RANDWRITE = TD_DDIR_WRITE | TD_DDIR_READ, + TD_DDIR_RANDWRITE = TD_DDIR_WRITE | TD_DDIR_RAND, TD_DDIR_RANDRW = TD_DDIR_RW | TD_DDIR_RAND, }; @@ -72,11 +81,21 @@ 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 list_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 @@ -92,6 +111,8 @@ enum { IO_U_F_FLIGHT = 1 << 1, }; +struct thread_data; + /* * The io unit */ @@ -108,6 +129,9 @@ struct io_u { #endif #ifdef FIO_HAVE_SYSLET struct syslet_req req; +#endif +#ifdef FIO_HAVE_GUASI + guasi_req_t greq; #endif }; struct timeval start_time; @@ -119,6 +143,7 @@ struct io_u { void *buf; unsigned long buflen; unsigned long long offset; + unsigned long long endpos; /* * IO engine state, may be different from above when we get @@ -149,7 +174,7 @@ struct io_u { /* * Callback for io completion */ - int (*end_io)(struct io_u *); + int (*end_io)(struct thread_data *, struct io_u *); }; /* @@ -166,20 +191,55 @@ enum { enum { VERIFY_NONE = 0, /* no verification */ VERIFY_MD5, /* md5 sum data blocks */ + VERIFY_CRC64, /* crc64 sum data blocks */ VERIFY_CRC32, /* crc32 sum data blocks */ + VERIFY_CRC16, /* crc16 sum data blocks */ + VERIFY_CRC7, /* crc7 sum data blocks */ + VERIFY_SHA256, /* sha256 sum data blocks */ + VERIFY_SHA512, /* sha512 sum data blocks */ + VERIFY_META, /* block_num, timestamp etc. */ + VERIFY_PATTERN, /* verify a specific pattern */ + VERIFY_NULL, /* pretend to verify */ }; /* - * A header structure associated with each checksummed data block + * A header structure associated with each checksummed data block. It is + * followed by a checksum specific header that contains the verification + * data. */ 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 vhdr_md5 { + uint32_t md5_digest[16]; +}; +struct vhdr_sha512 { + uint8_t sha512[128]; +}; +struct vhdr_sha256 { + uint8_t sha256[128]; +}; +struct vhdr_crc64 { + uint64_t crc64; +}; +struct vhdr_crc32 { + uint32_t crc32; +}; +struct vhdr_crc16 { + uint16_t crc16; +}; +struct vhdr_crc7 { + uint8_t crc7; +}; +struct vhdr_meta { + uint64_t offset; + unsigned char thread; + unsigned short numberio; + unsigned long time_sec; + unsigned long time_usec; }; struct group_run_stats { @@ -207,15 +267,26 @@ enum fio_filetype { FIO_TYPE_FILE = 1, /* plain file */ FIO_TYPE_BD, /* block device */ FIO_TYPE_CHAR, /* character device */ + FIO_TYPE_PIPE, /* pipe */ }; 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_DISKLESSIO = 1 << 4, /* no disk involved */ - FIO_SELFOPEN = 1 << 5, /* opens its own devices */ + 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 */ +}; + +enum fio_file_flags { + FIO_FILE_OPEN = 1 << 0, /* file is open */ + FIO_FILE_CLOSING = 1 << 1, /* file being closed */ + FIO_FILE_EXISTS = 1 << 2, /* file there */ + FIO_FILE_EXTEND = 1 << 3, /* needs extend */ + FIO_FILE_NOSORT = 1 << 4, /* don't sort verify blocks */ + FIO_FILE_DONE = 1 << 5, /* io completed to this file */ + FIO_SIZE_KNOWN = 1 << 6, /* size has been set */ }; /* @@ -223,6 +294,8 @@ enum fio_ioengine_flags { * this structure holds state information for a single file. */ struct fio_file { + enum fio_filetype filetype; + /* * A file may not be a file descriptor, let the io engine decide */ @@ -230,11 +303,21 @@ struct fio_file { unsigned long file_data; int fd; }; + + /* + * filename and possible memory mapping + */ char *file_name; void *mmap; - unsigned long long file_size; + unsigned int major, minor; + + /* + * size of the file, offset into file, and io size from that offset + */ unsigned long long real_file_size; unsigned long long file_offset; + unsigned long long io_size; + unsigned long long last_pos; unsigned long long last_completed_pos; @@ -245,10 +328,26 @@ struct fio_file { unsigned int num_maps; unsigned int last_free_lookup; - unsigned int unlink; + int references; + enum fio_file_flags flags; }; +/* + * 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; @@ -271,46 +370,47 @@ struct thread_stat { unsigned long usr_time; unsigned long sys_time; unsigned long ctx; -}; + unsigned long minf, majf; -/* - * How many depth levels to log - */ -#define FIO_IO_U_MAP_NR 8 -#define FIO_IO_U_LAT_NR 12 + /* + * IO depth and latency stats + */ + unsigned int io_u_map[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]; -/* - * This describes a single thread/process executing a fio job. - */ -struct thread_data { + unsigned long long io_bytes[2]; + unsigned long runtime[2]; + unsigned long total_run_time; +}; + +struct thread_options { int pad; char *description; char *name; char *directory; char *filename; - char verror[128]; - pthread_t thread; - int thread_number; - int groupid; - struct thread_stat ts; - enum fio_filetype filetype; - struct fio_file *files; - unsigned int nr_files; - unsigned int nr_open_files; - unsigned int nr_uniq_files; - union { - unsigned int next_file; - os_random_state_t next_file_state; - }; - int error; - pid_t pid; - char *orig_buffer; - size_t orig_buffer_size; - volatile int terminate; - volatile int runstate; + char *opendir; + char *ioengine; enum td_ddir td_ddir; - unsigned int ioprio; - unsigned int last_was_sync; + unsigned int ddir_nr; + unsigned int iodepth; + unsigned int iodepth_low; + unsigned int iodepth_batch; + + unsigned long long size; + unsigned long long file_size_low; + unsigned long long file_size_high; + unsigned long long start_offset; + + unsigned int bs[2]; + unsigned int min_bs[2]; + unsigned int max_bs[2]; + + unsigned int nr_files; + unsigned int open_files; unsigned int odirect; unsigned int invalidate_cache; @@ -319,6 +419,13 @@ struct thread_data { unsigned int end_fsync; 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; @@ -328,10 +435,8 @@ struct thread_data { unsigned int write_bw_log; unsigned int norandommap; 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; @@ -339,36 +444,91 @@ struct thread_data { unsigned int thinktime_blocks; unsigned int fsync_blocks; unsigned int start_delay; - unsigned long timeout; + unsigned long long timeout; 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; - unsigned int iodepth_batch; 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 time_based; char *read_iolog_file; char *write_iolog_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_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 last_was_sync; + + char *mmapfile; + int mmapfd; + void *iolog_buf; FILE *iolog_f; char *sysfs_root; - char *ioscheduler; os_random_state_t bsrange_state; os_random_state_t verify_state; @@ -385,9 +545,6 @@ 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; @@ -396,59 +553,52 @@ struct thread_data { /* * 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_sem *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; + unsigned int rw_end_set[2]; /* * read/write mixed workload state */ os_random_state_t rwmix_state; + unsigned long long rwmix_bytes; struct timeval rwmix_switch; 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 list_head io_hist_list; /* - * IO historic logs + * For IO replaying */ - struct list_head io_hist_list; struct list_head io_log_list; /* @@ -456,6 +606,18 @@ struct thread_data { */ struct timeval timeout_end; struct itimerval timer; + + /* + * 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; }; /* @@ -490,6 +652,7 @@ enum { 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; @@ -497,16 +660,18 @@ 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 struct thread_data *threads; -#define td_read(td) ((td)->td_ddir & TD_DDIR_READ) -#define td_write(td) ((td)->td_ddir & TD_DDIR_WRITE) -#define td_rw(td) (((td)->td_ddir & TD_DDIR_RW) == TD_DDIR_RW) -#define td_random(td) ((td)->td_ddir & TD_DDIR_RAND) +#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 BLOCKS_PER_MAP (8 * sizeof(long)) -#define TO_MAP_BLOCK(td, f, b) ((b) - ((f)->file_offset / (td)->rw_min_bs)) +#define TO_MAP_BLOCK(td, f, b) ((b) - ((f)->file_offset / (td)->o.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)) @@ -516,9 +681,9 @@ 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; @@ -543,8 +708,9 @@ struct disk_util { struct list_head list; 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; @@ -555,18 +721,18 @@ struct disk_util { #define DISK_UTIL_MSEC (250) -#ifndef min -#define min(a, b) ((a) < (b) ? (a) : (b)) -#endif -#ifndef max -#define max(a, b) ((a) > (b) ? (a) : (b)) -#endif - /* * Log exports */ +enum file_log_act { + FIO_LOG_ADD_FILE, + FIO_LOG_OPEN_FILE, + FIO_LOG_CLOSE_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 prune_io_piece_log(struct thread_data *); @@ -586,7 +752,6 @@ 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(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); @@ -608,10 +773,18 @@ extern void fio_gettime(struct timeval *, void *); extern void set_genesis_time(void); /* - * Init functions + * Init/option functions */ extern int __must_check parse_options(int, char **); -extern int __must_check init_random_state(struct thread_data *); +extern int fio_option_parse(struct thread_data *, const char *); +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 *); +#define FIO_GETOPT_JOB 0x89988998 +#define FIO_NR_OPTIONS 128 /* * File setup/shutdown @@ -619,9 +792,17 @@ extern int __must_check init_random_state(struct thread_data *); 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 open_file(struct thread_data *, struct fio_file *, int, int); -extern void close_file(struct thread_data *, struct fio_file *); extern int __must_check file_invalidate_cache(struct thread_data *, struct fio_file *); +extern int __must_check generic_open_file(struct thread_data *, struct fio_file *); +extern void generic_close_file(struct thread_data *, struct fio_file *); +extern int add_file(struct thread_data *, const char *); +extern void get_file(struct fio_file *); +extern void put_file(struct thread_data *, struct fio_file *); +extern int add_dir_files(struct thread_data *, const char *); +extern int init_random_map(struct thread_data *); +extern void dup_files(struct thread_data *, struct thread_data *); +extern int get_fileno(struct thread_data *, const char *); +extern void free_release_files(struct thread_data *); /* * ETA/status stuff @@ -629,6 +810,21 @@ 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 disk_util_timer_arm(void); +extern void init_disk_util(struct thread_data *); +extern void update_io_ticks(void); +#else +#define show_disk_util() +#define disk_util_timer_arm() +#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. @@ -651,7 +847,7 @@ enum { */ 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 int __must_check verify_io_u(struct thread_data *, struct io_u *); /* * Memory helpers @@ -672,8 +868,10 @@ 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 *); 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_log_error(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_mark_depth(struct thread_data *, struct io_u *); /* * io engine entry points @@ -684,30 +882,16 @@ 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_commit(struct thread_data *); +extern int __must_check td_io_open_file(struct thread_data *, struct fio_file *); +extern void td_io_close_file(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)++; -} +#ifdef FIO_HAVE_BLKTRACE +extern int is_blktrace(const char *); +extern int load_blktrace(struct thread_data *, const char *); +#endif /* * If logging output to a file, stderr should go to both stderr and f_err @@ -718,6 +902,8 @@ static inline void fio_sem_up(volatile int *sem) fprintf(stderr, ##args); \ } while (0) +#define log_info(args...) fprintf(f_out, ##args) + FILE *get_f_out(void); FILE *get_f_err(void); @@ -735,12 +921,13 @@ struct ioengine_ops { 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 *); + void (*close_file)(struct thread_data *, struct fio_file *); void *data; void *dlhandle; - unsigned long priv; }; -#define FIO_IOOPS_VERSION 5 +#define FIO_IOOPS_VERSION 7 extern struct ioengine_ops *load_ioengine(struct thread_data *, const char *); extern void register_ioengine(struct ioengine_ops *); @@ -757,7 +944,7 @@ 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)++) + for ((i) = 0, (f) = &(td)->files[0]; (i) < (td)->o.nr_files; (i)++, (f)++) #define fio_assert(td, cond) do { \ if (!(cond)) { \ @@ -769,4 +956,10 @@ extern void close_ioengine(struct thread_data *); } \ } while (0) +static inline void clear_error(struct thread_data *td) +{ + td->error = 0; + td->verror[0] = '\0'; +} + #endif