X-Git-Url: https://git.kernel.dk/?p=fio.git;a=blobdiff_plain;f=io_u.c;h=c58dcf0493bb50347367f8a1d6886ded60ec7748;hp=1d6872edb2876794e9c4c07d5ea5e05a726f34dd;hb=a76e1995b08ffe645b2d9f81d258afc38688884b;hpb=53bb5d9c037e842970b32bc828dbe809b42c144d diff --git a/io_u.c b/io_u.c index 1d6872ed..c58dcf04 100644 --- a/io_u.c +++ b/io_u.c @@ -1,12 +1,8 @@ #include -#include #include -#include -#include #include #include "fio.h" -#include "hash.h" #include "verify.h" #include "trim.h" #include "lib/rand.h" @@ -37,9 +33,9 @@ static bool random_map_free(struct fio_file *f, const uint64_t block) */ static void mark_random_map(struct thread_data *td, struct io_u *io_u) { - unsigned int min_bs = td->o.min_bs[io_u->ddir]; + unsigned long long min_bs = td->o.min_bs[io_u->ddir]; struct fio_file *f = io_u->file; - unsigned int nr_blocks; + unsigned long long nr_blocks; uint64_t block; block = (io_u->offset - f->file_offset) / (uint64_t) min_bs; @@ -81,11 +77,6 @@ static uint64_t last_block(struct thread_data *td, struct fio_file *f, return max_blocks; } -struct rand_off { - struct flist_head list; - uint64_t off; -}; - static int __get_next_rand_offset(struct thread_data *td, struct fio_file *f, enum fio_ddir ddir, uint64_t *b, uint64_t lastb) @@ -163,7 +154,6 @@ static int __get_next_rand_offset_zoned_abs(struct thread_data *td, { struct zone_split_index *zsi; uint64_t lastb, send, stotal; - static int warned; unsigned int v; lastb = last_block(td, f, ddir); @@ -178,7 +168,7 @@ bail: /* * Generate a value, v, between 1 and 100, both inclusive */ - v = rand32_between(&td->zone_state, 1, 100); + v = rand_between(&td->zone_state, 1, 100); /* * Find our generated table. 'send' is the end block of this zone, @@ -192,10 +182,8 @@ bail: * Should never happen */ if (send == -1U) { - if (!warned) { + if (!fio_did_warn(FIO_WARN_ZONED_BUG)) log_err("fio: bug in zoned generation\n"); - warned = 1; - } goto bail; } else if (send > lastb) { /* @@ -223,7 +211,6 @@ static int __get_next_rand_offset_zoned(struct thread_data *td, { unsigned int v, send, stotal; uint64_t offset, lastb; - static int warned; struct zone_split_index *zsi; lastb = last_block(td, f, ddir); @@ -238,7 +225,7 @@ bail: /* * Generate a value, v, between 1 and 100, both inclusive */ - v = rand32_between(&td->zone_state, 1, 100); + v = rand_between(&td->zone_state, 1, 100); zsi = &td->zone_state_index[ddir][v - 1]; stotal = zsi->size_perc_prev; @@ -248,10 +235,8 @@ bail: * Should never happen */ if (send == -1U) { - if (!warned) { + if (!fio_did_warn(FIO_WARN_ZONED_BUG)) log_err("fio: bug in zoned generation\n"); - warned = 1; - } goto bail; } @@ -282,16 +267,8 @@ bail: return 0; } -static int flist_cmp(void *data, struct flist_head *a, struct flist_head *b) -{ - struct rand_off *r1 = flist_entry(a, struct rand_off, list); - struct rand_off *r2 = flist_entry(b, struct rand_off, list); - - return r1->off - r2->off; -} - -static int get_off_from_method(struct thread_data *td, struct fio_file *f, - enum fio_ddir ddir, uint64_t *b) +static int get_next_rand_offset(struct thread_data *td, struct fio_file *f, + enum fio_ddir ddir, uint64_t *b) { if (td->o.random_distribution == FIO_RAND_DIST_RANDOM) { uint64_t lastb; @@ -316,25 +293,6 @@ static int get_off_from_method(struct thread_data *td, struct fio_file *f, return 1; } -/* - * Sort the reads for a verify phase in batches of verifysort_nr, if - * specified. - */ -static inline bool should_sort_io(struct thread_data *td) -{ - if (!td->o.verifysort_nr || !td->o.do_verify) - return false; - if (!td_random(td)) - return false; - if (td->runstate != TD_VERIFYING) - return false; - if (td->o.random_generator == FIO_RAND_GEN_TAUSWORTHE || - td->o.random_generator == FIO_RAND_GEN_TAUSWORTHE64) - return false; - - return true; -} - static bool should_do_random(struct thread_data *td, enum fio_ddir ddir) { unsigned int v; @@ -342,49 +300,11 @@ static bool should_do_random(struct thread_data *td, enum fio_ddir ddir) if (td->o.perc_rand[ddir] == 100) return true; - v = rand32_between(&td->seq_rand_state[ddir], 1, 100); + v = rand_between(&td->seq_rand_state[ddir], 1, 100); return v <= td->o.perc_rand[ddir]; } -static int get_next_rand_offset(struct thread_data *td, struct fio_file *f, - enum fio_ddir ddir, uint64_t *b) -{ - struct rand_off *r; - int i, ret = 1; - - if (!should_sort_io(td)) - return get_off_from_method(td, f, ddir, b); - - if (!flist_empty(&td->next_rand_list)) { -fetch: - r = flist_first_entry(&td->next_rand_list, struct rand_off, list); - flist_del(&r->list); - *b = r->off; - free(r); - return 0; - } - - for (i = 0; i < td->o.verifysort_nr; i++) { - r = malloc(sizeof(*r)); - - ret = get_off_from_method(td, f, ddir, &r->off); - if (ret) { - free(r); - break; - } - - flist_add(&r->list, &td->next_rand_list); - } - - if (ret && !i) - return ret; - - assert(!flist_empty(&td->next_rand_list)); - flist_sort(NULL, &td->next_rand_list, flist_cmp); - goto fetch; -} - static void loop_cache_invalidate(struct thread_data *td, struct fio_file *f) { struct thread_options *o = &td->o; @@ -405,9 +325,9 @@ static int get_next_rand_block(struct thread_data *td, struct fio_file *f, if (td->o.time_based || (td->o.file_service_type & __FIO_FSERVICE_NONUNIFORM)) { fio_file_reset(td, f); + loop_cache_invalidate(td, f); if (!get_next_rand_offset(td, f, ddir, b)) return 0; - loop_cache_invalidate(td, f); } dprint(FD_IO, "%s: rand offset failed, last=%llu, size=%llu\n", @@ -436,7 +356,11 @@ static int get_next_seq_offset(struct thread_data *td, struct fio_file *f, if (f->last_pos[ddir] < f->real_file_size) { uint64_t pos; - if (f->last_pos[ddir] == f->file_offset && o->ddir_seq_add < 0) { + /* + * Only rewind if we already hit the end + */ + if (f->last_pos[ddir] == f->file_offset && + f->file_offset && o->ddir_seq_add < 0) { if (f->real_file_size > f->io_size) f->last_pos[ddir] = f->io_size; else @@ -476,7 +400,7 @@ static int get_next_seq_offset(struct thread_data *td, struct fio_file *f, static int get_next_block(struct thread_data *td, struct io_u *io_u, enum fio_ddir ddir, int rw_seq, - unsigned int *is_random) + bool *is_random) { struct fio_file *f = io_u->file; uint64_t b, offset; @@ -490,27 +414,27 @@ static int get_next_block(struct thread_data *td, struct io_u *io_u, if (td_random(td)) { if (should_do_random(td, ddir)) { ret = get_next_rand_block(td, f, ddir, &b); - *is_random = 1; + *is_random = true; } else { - *is_random = 0; + *is_random = false; io_u_set(td, io_u, IO_U_F_BUSY_OK); ret = get_next_seq_offset(td, f, ddir, &offset); if (ret) ret = get_next_rand_block(td, f, ddir, &b); } } else { - *is_random = 0; + *is_random = false; ret = get_next_seq_offset(td, f, ddir, &offset); } } else { io_u_set(td, io_u, IO_U_F_BUSY_OK); - *is_random = 0; + *is_random = false; if (td->o.rw_seq == RW_SEQ_SEQ) { ret = get_next_seq_offset(td, f, ddir, &offset); if (ret) { ret = get_next_rand_block(td, f, ddir, &b); - *is_random = 0; + *is_random = false; } } else if (td->o.rw_seq == RW_SEQ_IDENT) { if (f->last_start[ddir] != -1ULL) @@ -543,8 +467,8 @@ static int get_next_block(struct thread_data *td, struct io_u *io_u, * until we find a free one. For sequential io, just return the end of * the last io issued. */ -static int __get_next_offset(struct thread_data *td, struct io_u *io_u, - unsigned int *is_random) +static int get_next_offset(struct thread_data *td, struct io_u *io_u, + bool *is_random) { struct fio_file *f = io_u->file; enum fio_ddir ddir = io_u->ddir; @@ -578,40 +502,27 @@ static int __get_next_offset(struct thread_data *td, struct io_u *io_u, return 0; } -static int get_next_offset(struct thread_data *td, struct io_u *io_u, - unsigned int *is_random) -{ - if (td->flags & TD_F_PROFILE_OPS) { - struct prof_io_ops *ops = &td->prof_io_ops; - - if (ops->fill_io_u_off) - return ops->fill_io_u_off(td, io_u, is_random); - } - - return __get_next_offset(td, io_u, is_random); -} - static inline bool io_u_fits(struct thread_data *td, struct io_u *io_u, - unsigned int buflen) + unsigned long long buflen) { struct fio_file *f = io_u->file; return io_u->offset + buflen <= f->io_size + get_start_offset(td, f); } -static unsigned int __get_next_buflen(struct thread_data *td, struct io_u *io_u, - unsigned int is_random) +static unsigned long long get_next_buflen(struct thread_data *td, struct io_u *io_u, + bool is_random) { int ddir = io_u->ddir; - unsigned int buflen = 0; - unsigned int minbs, maxbs; + unsigned long long buflen = 0; + unsigned long long minbs, maxbs; uint64_t frand_max, r; bool power_2; assert(ddir_rw(ddir)); if (td->o.bs_is_seq_rand) - ddir = is_random ? DDIR_WRITE: DDIR_READ; + ddir = is_random ? DDIR_WRITE : DDIR_READ; minbs = td->o.min_bs[ddir]; maxbs = td->o.max_bs[ddir]; @@ -630,10 +541,8 @@ static unsigned int __get_next_buflen(struct thread_data *td, struct io_u *io_u, r = __rand(&td->bsrange_state[ddir]); if (!td->o.bssplit_nr[ddir]) { - buflen = 1 + (unsigned int) ((double) maxbs * + buflen = minbs + (unsigned long long) ((double) maxbs * (r / (frand_max + 1.0))); - if (buflen < minbs) - buflen = minbs; } else { long long perc = 0; unsigned int i; @@ -661,19 +570,6 @@ static unsigned int __get_next_buflen(struct thread_data *td, struct io_u *io_u, return buflen; } -static unsigned int get_next_buflen(struct thread_data *td, struct io_u *io_u, - unsigned int is_random) -{ - if (td->flags & TD_F_PROFILE_OPS) { - struct prof_io_ops *ops = &td->prof_io_ops; - - if (ops->fill_io_u_size) - return ops->fill_io_u_size(td, io_u, is_random); - } - - return __get_next_buflen(td, io_u, is_random); -} - static void set_rwmix_bytes(struct thread_data *td) { unsigned int diff; @@ -691,7 +587,7 @@ static inline enum fio_ddir get_rand_ddir(struct thread_data *td) { unsigned int v; - v = rand32_between(&td->rwmix_state, 1, 100); + v = rand_between(&td->rwmix_state, 1, 100); if (v <= td->o.rwmix[DDIR_READ]) return DDIR_READ; @@ -712,11 +608,8 @@ int io_u_quiesce(struct thread_data *td) * io's that have been actually submitted to an async engine, * and cur_depth is meaningless for sync engines. */ - if (td->io_u_queued || td->cur_depth) { - int fio_unused ret; - - ret = td_io_commit(td); - } + if (td->io_u_queued || td->cur_depth) + td_io_commit(td); while (td->io_u_in_flight) { int ret; @@ -884,8 +777,8 @@ void put_io_u(struct thread_data *td, struct io_u *io_u) assert(!(td->flags & TD_F_CHILD)); } io_u_qpush(&td->io_u_freelist, io_u); - td_io_u_unlock(td); td_io_u_free_notify(td); + td_io_u_unlock(td); } void clear_io_u(struct thread_data *td, struct io_u *io_u) @@ -917,8 +810,8 @@ void requeue_io_u(struct thread_data *td, struct io_u **io_u) } io_u_rpush(&td->io_u_requeues, __io_u); - td_io_u_unlock(td); td_io_u_free_notify(td); + td_io_u_unlock(td); *io_u = NULL; } @@ -937,15 +830,16 @@ static void __fill_io_u_zone(struct thread_data *td, struct io_u *io_u) * Wrap from the beginning, if we exceed the file size */ if (f->file_offset >= f->real_file_size) - f->file_offset = f->real_file_size - f->file_offset; + f->file_offset = get_start_offset(td, f); + f->last_pos[io_u->ddir] = f->file_offset; td->io_skip_bytes += td->o.zone_skip; } /* - * If zone_size > zone_range, then maintain the same zone until - * zone_bytes >= zone_size. - */ + * If zone_size > zone_range, then maintain the same zone until + * zone_bytes >= zone_size. + */ if (f->last_pos[io_u->ddir] >= (f->file_offset + td->o.zone_range)) { dprint(FD_IO, "io_u maintain zone offset=%" PRIu64 "/last_pos=%" PRIu64 "\n", f->file_offset, f->last_pos[io_u->ddir]); @@ -956,14 +850,13 @@ static void __fill_io_u_zone(struct thread_data *td, struct io_u *io_u) * For random: if 'norandommap' is not set and zone_size > zone_range, * map needs to be reset as it's done with zone_range everytime. */ - if ((td->zone_bytes % td->o.zone_range) == 0) { + if ((td->zone_bytes % td->o.zone_range) == 0) fio_file_reset(td, f); - } } static int fill_io_u(struct thread_data *td, struct io_u *io_u) { - unsigned int is_random; + bool is_random; if (td_ioengine_flagged(td, FIO_NOIO)) goto out; @@ -979,9 +872,8 @@ static int fill_io_u(struct thread_data *td, struct io_u *io_u) /* * When file is zoned zone_range is always positive */ - if (td->o.zone_range) { + if (td->o.zone_range) __fill_io_u_zone(td, io_u); - } /* * No log, let the seq/rand engine retrieve the next buflen and @@ -999,7 +891,7 @@ static int fill_io_u(struct thread_data *td, struct io_u *io_u) } if (io_u->offset + io_u->buflen > io_u->file->real_file_size) { - dprint(FD_IO, "io_u %p, off=0x%llx + len=0x%lx exceeds file size=0x%llx\n", + dprint(FD_IO, "io_u %p, off=0x%llx + len=0x%llx exceeds file size=0x%llx\n", io_u, (unsigned long long) io_u->offset, io_u->buflen, (unsigned long long) io_u->file->real_file_size); @@ -1018,7 +910,7 @@ out: return 0; } -static void __io_u_mark_map(unsigned int *map, unsigned int nr) +static void __io_u_mark_map(uint64_t *map, unsigned int nr) { int idx = 0; @@ -1393,13 +1285,6 @@ out: static struct fio_file *get_next_file(struct thread_data *td) { - if (td->flags & TD_F_PROFILE_OPS) { - struct prof_io_ops *ops = &td->prof_io_ops; - - if (ops->get_next_file) - return ops->get_next_file(td); - } - return __get_next_file(td); } @@ -1593,6 +1478,7 @@ bool queue_full(const struct thread_data *td) struct io_u *__get_io_u(struct thread_data *td) { struct io_u *io_u = NULL; + int ret; if (td->stop_io) return NULL; @@ -1629,7 +1515,8 @@ again: * return one */ assert(!(td->flags & TD_F_CHILD)); - assert(!pthread_cond_wait(&td->free_cond, &td->io_u_lock)); + ret = pthread_cond_wait(&td->free_cond, &td->io_u_lock); + assert(ret == 0); goto again; } @@ -1695,7 +1582,7 @@ static bool check_get_verify(struct thread_data *td, struct io_u *io_u) */ static void small_content_scramble(struct io_u *io_u) { - unsigned int i, nr_blocks = io_u->buflen >> 9; + unsigned long long i, nr_blocks = io_u->buflen >> 9; unsigned int offset; uint64_t boffset, *iptr; char *p; @@ -1736,7 +1623,7 @@ static void small_content_scramble(struct io_u *io_u) /* * Return an io_u to be processed. Gets a buflen and offset, sets direction, - * etc. The returned io_u is fully ready to be prepped and submitted. + * etc. The returned io_u is fully ready to be prepped, populated and submitted. */ struct io_u *get_io_u(struct thread_data *td) { @@ -1797,12 +1684,9 @@ struct io_u *get_io_u(struct thread_data *td) td->o.min_bs[DDIR_WRITE], io_u->buflen); } else if ((td->flags & TD_F_SCRAMBLE_BUFFERS) && - !(td->flags & TD_F_COMPRESS)) + !(td->flags & TD_F_COMPRESS) && + !(td->flags & TD_F_DO_VERIFY)) do_scramble = 1; - if (td->flags & TD_F_VER_NONE) { - populate_verify_io_u(td, io_u); - do_scramble = 0; - } } else if (io_u->ddir == DDIR_READ) { /* * Reset the buf_filled parameters so next time if the @@ -1842,7 +1726,7 @@ static void __io_u_log_error(struct thread_data *td, struct io_u *io_u) if (td_non_fatal_error(td, eb, io_u->error) && !td->o.error_dump) return; - log_err("fio: io_u error%s%s: %s: %s offset=%llu, buflen=%lu\n", + log_err("fio: io_u error%s%s: %s: %s offset=%llu, buflen=%llu\n", io_u->file ? " on file " : "", io_u->file ? io_u->file->file_name : "", strerror(io_u->error), @@ -1921,7 +1805,8 @@ static void account_io_completion(struct thread_data *td, struct io_u *io_u, if (no_reduce && per_unit_log(td->iops_log)) add_iops_sample(td, io_u, bytes); - } + } else if (ddir_sync(idx) && !td->o.disable_clat) + add_sync_clat_sample(&td->ts, llnsec); if (td->ts.nr_block_infos && io_u->ddir == DDIR_TRIM) { uint32_t *info = io_u_block_info(td, io_u); @@ -1959,6 +1844,12 @@ static void file_log_write_comp(const struct thread_data *td, struct fio_file *f f->last_write_idx = 0; } +static bool should_account(struct thread_data *td) +{ + return ramp_time_over(td) && (td->runstate == TD_RUNNING || + td->runstate == TD_VERIFYING); +} + static void io_completed(struct thread_data *td, struct io_u **io_u_ptr, struct io_completion_data *icd) { @@ -1987,19 +1878,21 @@ static void io_completed(struct thread_data *td, struct io_u **io_u_ptr, } if (ddir_sync(ddir)) { - td->last_was_sync = 1; + td->last_was_sync = true; if (f) { f->first_write = -1ULL; f->last_write = -1ULL; } + if (should_account(td)) + account_io_completion(td, io_u, icd, ddir, io_u->buflen); return; } - td->last_was_sync = 0; + td->last_was_sync = false; td->last_ddir = ddir; if (!io_u->error && ddir_rw(ddir)) { - unsigned int bytes = io_u->buflen - io_u->resid; + unsigned long long bytes = io_u->buflen - io_u->resid; int ret; td->io_blocks[ddir]++; @@ -2013,8 +1906,7 @@ static void io_completed(struct thread_data *td, struct io_u **io_u_ptr, if (ddir == DDIR_WRITE) file_log_write_comp(td, f, io_u->offset, bytes); - if (ramp_time_over(td) && (td->runstate == TD_RUNNING || - td->runstate == TD_VERIFYING)) + if (should_account(td)) account_io_completion(td, io_u, icd, ddir, bytes); icd->bytes_done[ddir] += bytes; @@ -2174,7 +2066,7 @@ static struct frand_state *get_buf_state(struct thread_data *td) return &td->buf_state; } - v = rand32_between(&td->dedupe_state, 1, 100); + v = rand_between(&td->dedupe_state, 1, 100); if (v <= td->o.dedupe_percentage) return &td->buf_state_prev; @@ -2190,8 +2082,8 @@ static void save_buf_state(struct thread_data *td, struct frand_state *rs) frand_copy(&td->buf_state_prev, rs); } -void fill_io_buffer(struct thread_data *td, void *buf, unsigned int min_write, - unsigned int max_bs) +void fill_io_buffer(struct thread_data *td, void *buf, unsigned long long min_write, + unsigned long long max_bs) { struct thread_options *o = &td->o; @@ -2201,8 +2093,8 @@ void fill_io_buffer(struct thread_data *td, void *buf, unsigned int min_write, if (o->compress_percentage || o->dedupe_percentage) { unsigned int perc = td->o.compress_percentage; struct frand_state *rs; - unsigned int left = max_bs; - unsigned int this_write; + unsigned long long left = max_bs; + unsigned long long this_write; do { rs = get_buf_state(td); @@ -2211,7 +2103,7 @@ void fill_io_buffer(struct thread_data *td, void *buf, unsigned int min_write, if (perc) { this_write = min_not_zero(min_write, - td->o.compress_chunk); + (unsigned long long) td->o.compress_chunk); fill_random_buf_percentage(rs, buf, perc, this_write, this_write, @@ -2238,7 +2130,7 @@ void fill_io_buffer(struct thread_data *td, void *buf, unsigned int min_write, * "randomly" fill the buffer contents */ void io_u_fill_buffer(struct thread_data *td, struct io_u *io_u, - unsigned int min_write, unsigned int max_bs) + unsigned long long min_write, unsigned long long max_bs) { io_u->buf_filled_len = 0; fill_io_buffer(td, io_u->buf, min_write, max_bs);