int nr; /* input */
int error; /* output */
- unsigned long bytes_done[DDIR_RWDIR_CNT]; /* output */
+ uint64_t bytes_done[DDIR_RWDIR_CNT]; /* output */
struct timeval time; /* output */
};
* The ->io_axmap contains a map of blocks we have or have not done io
* to yet. Used to make sure we cover the entire range in a fair fashion.
*/
-static int random_map_free(struct fio_file *f, const unsigned long long block)
+static int random_map_free(struct fio_file *f, const uint64_t block)
{
return !axmap_isset(f->io_axmap, block);
}
{
unsigned int min_bs = td->o.rw_min_bs;
struct fio_file *f = io_u->file;
- unsigned long long block;
unsigned int nr_blocks;
+ uint64_t block;
- block = (io_u->offset - f->file_offset) / (unsigned long long) min_bs;
+ block = (io_u->offset - f->file_offset) / (uint64_t) min_bs;
nr_blocks = (io_u->buflen + min_bs - 1) / min_bs;
if (!(io_u->flags & IO_U_F_BUSY_OK))
io_u->buflen = nr_blocks * min_bs;
}
-static unsigned long long last_block(struct thread_data *td, struct fio_file *f,
- enum fio_ddir ddir)
+static uint64_t last_block(struct thread_data *td, struct fio_file *f,
+ enum fio_ddir ddir)
{
- unsigned long long max_blocks;
- unsigned long long max_size;
+ uint64_t max_blocks;
+ uint64_t max_size;
assert(ddir_rw(ddir));
if (td->o.zone_range)
max_size = td->o.zone_range;
- max_blocks = max_size / (unsigned long long) td->o.ba[ddir];
+ max_blocks = max_size / (uint64_t) td->o.ba[ddir];
if (!max_blocks)
return 0;
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, unsigned long long *b)
+ enum fio_ddir ddir, uint64_t *b)
{
- unsigned long long r;
+ uint64_t r, lastb;
- if (td->o.random_generator == FIO_RAND_GEN_TAUSWORTHE) {
- unsigned long long rmax, lastb;
+ lastb = last_block(td, f, ddir);
+ if (!lastb)
+ return 1;
- lastb = last_block(td, f, ddir);
- if (!lastb)
- return 1;
+ if (td->o.random_generator == FIO_RAND_GEN_TAUSWORTHE) {
+ uint64_t rmax;
rmax = td->o.use_os_rand ? OS_RAND_MAX : FRAND_MAX;
r = __rand(&td->__random_state);
}
- dprint(FD_RANDOM, "off rand %llu\n", r);
+ dprint(FD_RANDOM, "off rand %llu\n", (unsigned long long) r);
- *b = (lastb - 1) * (r / ((unsigned long long) rmax + 1.0));
+ *b = (lastb - 1) * (r / ((uint64_t) rmax + 1.0));
} else {
uint64_t off = 0;
- if (lfsr_next(&f->lfsr, &off))
+ if (lfsr_next(&f->lfsr, &off, lastb))
return 1;
*b = off;
if (random_map_free(f, *b))
goto ret;
- dprint(FD_RANDOM, "get_next_rand_offset: offset %llu busy\n", *b);
+ dprint(FD_RANDOM, "get_next_rand_offset: offset %llu busy\n",
+ (unsigned long long) *b);
*b = axmap_next_free(f->io_axmap, *b);
if (*b == (uint64_t) -1ULL)
static int __get_next_rand_offset_zipf(struct thread_data *td,
struct fio_file *f, enum fio_ddir ddir,
- unsigned long long *b)
+ uint64_t *b)
{
*b = zipf_next(&f->zipf);
return 0;
static int __get_next_rand_offset_pareto(struct thread_data *td,
struct fio_file *f, enum fio_ddir ddir,
- unsigned long long *b)
+ uint64_t *b)
{
*b = pareto_next(&f->zipf);
return 0;
}
-static int get_next_rand_offset(struct thread_data *td, struct fio_file *f,
- enum fio_ddir ddir, unsigned long long *b)
+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)
{
if (td->o.random_distribution == FIO_RAND_DIST_RANDOM)
return __get_next_rand_offset(td, f, ddir, b);
return 1;
}
+/*
+ * Sort the reads for a verify phase in batches of verifysort_nr, if
+ * specified.
+ */
+static inline int should_sort_io(struct thread_data *td)
+{
+ if (!td->o.verifysort_nr || !td->o.do_verify)
+ return 0;
+ if (!td_random(td))
+ return 0;
+ if (td->runstate != TD_VERIFYING)
+ return 0;
+ if (td->o.random_generator == FIO_RAND_GEN_TAUSWORTHE)
+ return 0;
+
+ return 1;
+}
+
+static int should_do_random(struct thread_data *td)
+{
+ unsigned int v;
+ unsigned long r;
+
+ if (td->o.perc_rand == 100)
+ return 1;
+
+ if (td->o.use_os_rand) {
+ r = os_random_long(&td->seq_rand_state);
+ v = 1 + (int) (100.0 * (r / (OS_RAND_MAX + 1.0)));
+ } else {
+ r = __rand(&td->__seq_rand_state);
+ v = 1 + (int) (100.0 * (r / (FRAND_MAX + 1.0)));
+ }
+
+ return v <= td->o.perc_rand;
+}
+
+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)) {
+ struct rand_off *r;
+fetch:
+ r = flist_entry(td->next_rand_list.next, 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 int get_next_rand_block(struct thread_data *td, struct fio_file *f,
- enum fio_ddir ddir, unsigned long long *b)
+ enum fio_ddir ddir, uint64_t *b)
{
if (!get_next_rand_offset(td, f, ddir, b))
return 0;
if (td->o.time_based) {
- fio_file_reset(f);
+ fio_file_reset(td, f);
if (!get_next_rand_offset(td, f, ddir, b))
return 0;
}
dprint(FD_IO, "%s: rand offset failed, last=%llu, size=%llu\n",
- f->file_name, f->last_pos, f->real_file_size);
+ f->file_name, (unsigned long long) f->last_pos,
+ (unsigned long long) f->real_file_size);
return 1;
}
static int get_next_seq_offset(struct thread_data *td, struct fio_file *f,
- enum fio_ddir ddir, unsigned long long *offset)
+ enum fio_ddir ddir, uint64_t *offset)
{
assert(ddir_rw(ddir));
f->last_pos = f->last_pos - f->io_size;
if (f->last_pos < f->real_file_size) {
- unsigned long long pos;
+ uint64_t pos;
if (f->last_pos == f->file_offset && td->o.ddir_seq_add < 0)
f->last_pos = f->real_file_size;
enum fio_ddir ddir, int rw_seq)
{
struct fio_file *f = io_u->file;
- unsigned long long b, offset;
+ uint64_t b, offset;
int ret;
assert(ddir_rw(ddir));
b = offset = -1ULL;
if (rw_seq) {
- if (td_random(td))
- ret = get_next_rand_block(td, f, ddir, &b);
- else
+ if (td_random(td)) {
+ if (should_do_random(td))
+ ret = get_next_rand_block(td, f, ddir, &b);
+ else {
+ io_u->flags |= 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
ret = get_next_seq_offset(td, f, ddir, &offset);
} else {
io_u->flags |= IO_U_F_BUSY_OK;
else if (b != -1ULL)
io_u->offset = b * td->o.ba[ddir];
else {
- log_err("fio: bug in offset generation: offset=%llu, b=%llu\n",
- offset, b);
+ log_err("fio: bug in offset generation: offset=%llu, b=%llu\n", (unsigned long long) offset, (unsigned long long) b);
ret = 1;
}
}
if (io_u->offset >= f->io_size) {
dprint(FD_IO, "get_next_offset: offset %llu >= io_size %llu\n",
- io_u->offset, f->io_size);
+ (unsigned long long) io_u->offset,
+ (unsigned long long) f->io_size);
return 1;
}
io_u->offset += f->file_offset;
if (io_u->offset >= f->real_file_size) {
dprint(FD_IO, "get_next_offset: offset %llu >= size %llu\n",
- io_u->offset, f->real_file_size);
+ (unsigned long long) io_u->offset,
+ (unsigned long long) f->real_file_size);
return 1;
}
return DDIR_WRITE;
}
+void io_u_quiesce(struct thread_data *td)
+{
+ /*
+ * We are going to sleep, ensure that we flush anything pending as
+ * not to skew our latency numbers.
+ *
+ * Changed to only monitor 'in flight' requests here instead of the
+ * td->cur_depth, b/c td->cur_depth does not accurately represent
+ * io's that have been actually submitted to an async engine,
+ * and cur_depth is meaningless for sync engines.
+ */
+ while (td->io_u_in_flight) {
+ int fio_unused ret;
+
+ ret = io_u_queued_complete(td, 1, NULL);
+ }
+}
+
static enum fio_ddir rate_ddir(struct thread_data *td, enum fio_ddir ddir)
{
enum fio_ddir odir = ddir ^ 1;
* We have too much pending sleep in this direction. See if we
* should switch.
*/
- if (td_rw(td)) {
+ if (td_rw(td) && td->o.rwmix[odir]) {
/*
* Other direction does not have too much pending, switch
*/
} else
usec = td->rate_pending_usleep[ddir];
- /*
- * We are going to sleep, ensure that we flush anything pending as
- * not to skew our latency numbers.
- *
- * Changed to only monitor 'in flight' requests here instead of the
- * td->cur_depth, b/c td->cur_depth does not accurately represent
- * io's that have been actually submitted to an async engine,
- * and cur_depth is meaningless for sync engines.
- */
- if (td->io_u_in_flight) {
- int fio_unused ret;
-
- ret = io_u_queued_complete(td, td->io_u_in_flight, NULL);
- }
+ io_u_quiesce(td);
fio_gettime(&t, NULL);
usec_sleep(td, usec);
if (ddir_trim(ddir))
return ddir;
+
return ddir;
}
static void set_rw_ddir(struct thread_data *td, struct io_u *io_u)
{
- io_u->ddir = get_rw_ddir(td);
+ io_u->ddir = io_u->acct_ddir = get_rw_ddir(td);
if (io_u->ddir == DDIR_WRITE && (td->io_ops->flags & FIO_BARRIER) &&
td->o.barrier_blocks &&
if (io_u->flags & IO_U_F_IN_CUR_DEPTH)
td->cur_depth--;
- flist_del_init(&io_u->list);
- flist_add(&io_u->list, &td->io_u_freelist);
+ io_u_qpush(&td->io_u_freelist, io_u);
td_io_u_unlock(td);
td_io_u_free_notify(td);
}
void requeue_io_u(struct thread_data *td, struct io_u **io_u)
{
struct io_u *__io_u = *io_u;
+ enum fio_ddir ddir = acct_ddir(__io_u);
dprint(FD_IO, "requeue %p\n", __io_u);
td_io_u_lock(td);
__io_u->flags |= IO_U_F_FREE;
- if ((__io_u->flags & IO_U_F_FLIGHT) && ddir_rw(__io_u->ddir))
- td->io_issues[__io_u->ddir]--;
+ if ((__io_u->flags & IO_U_F_FLIGHT) && ddir_rw(ddir))
+ td->io_issues[ddir]--;
__io_u->flags &= ~IO_U_F_FLIGHT;
if (__io_u->flags & IO_U_F_IN_CUR_DEPTH)
td->cur_depth--;
- flist_del(&__io_u->list);
- flist_add_tail(&__io_u->list, &td->io_u_requeues);
+
+ io_u_rpush(&td->io_u_requeues, __io_u);
td_io_u_unlock(td);
*io_u = NULL;
}
if (io_u->offset + io_u->buflen > io_u->file->real_file_size) {
dprint(FD_IO, "io_u %p, offset too large\n", io_u);
- dprint(FD_IO, " off=%llu/%lu > %llu\n", io_u->offset,
- io_u->buflen, io_u->file->real_file_size);
+ dprint(FD_IO, " off=%llu/%lu > %llu\n",
+ (unsigned long long) io_u->offset, io_u->buflen,
+ (unsigned long long) io_u->file->real_file_size);
return 1;
}
if (td_random(td) && file_randommap(td, io_u->file))
mark_random_map(td, io_u);
- /*
- * If using a write iolog, store this entry.
- */
out:
dprint_io_u(io_u, "fill_io_u");
td->zone_bytes += io_u->buflen;
- log_io_u(td, io_u);
return 0;
}
struct io_u *__get_io_u(struct thread_data *td)
{
- struct io_u *io_u = NULL;
+ struct io_u *io_u;
td_io_u_lock(td);
again:
- if (!flist_empty(&td->io_u_requeues))
- io_u = flist_entry(td->io_u_requeues.next, struct io_u, list);
- else if (!queue_full(td)) {
- io_u = flist_entry(td->io_u_freelist.next, struct io_u, list);
+ if (!io_u_rempty(&td->io_u_requeues))
+ io_u = io_u_rpop(&td->io_u_requeues);
+ else if (!io_u_qempty(&td->io_u_freelist))
+ io_u = io_u_qpop(&td->io_u_freelist);
+ if (io_u) {
io_u->buflen = 0;
io_u->resid = 0;
io_u->file = NULL;
io_u->flags &= ~IO_U_F_VER_LIST;
io_u->error = 0;
- flist_del(&io_u->list);
- flist_add_tail(&io_u->list, &td->io_u_busylist);
+ io_u->acct_ddir = -1;
td->cur_depth++;
io_u->flags |= IO_U_F_IN_CUR_DEPTH;
} else if (td->o.verify_async) {
static void small_content_scramble(struct io_u *io_u)
{
unsigned int i, nr_blocks = io_u->buflen / 512;
- unsigned long long boffset;
+ uint64_t boffset;
unsigned int offset;
void *p, *end;
* and the actual offset.
*/
offset = (io_u->start_time.tv_usec ^ boffset) & 511;
- offset &= ~(sizeof(unsigned long long) - 1);
- if (offset >= 512 - sizeof(unsigned long long))
- offset -= sizeof(unsigned long long);
+ offset &= ~(sizeof(uint64_t) - 1);
+ if (offset >= 512 - sizeof(uint64_t))
+ offset -= sizeof(uint64_t);
memcpy(p + offset, &boffset, sizeof(boffset));
end = p + 512 - sizeof(io_u->start_time);
tusec = utime_since(&io_u->start_time, &icd->time);
add_lat_sample(td, idx, tusec, bytes);
+ if (td->flags & TD_F_PROFILE_OPS) {
+ struct prof_io_ops *ops = &td->prof_io_ops;
+
+ if (ops->io_u_lat)
+ icd->error = ops->io_u_lat(td, tusec);
+ }
+
if (td->o.max_latency && tusec > td->o.max_latency) {
if (!td->error)
log_err("fio: latency of %lu usec exceeds specified max (%u usec)\n", tusec, td->o.max_latency);
static long long usec_for_io(struct thread_data *td, enum fio_ddir ddir)
{
- unsigned long long secs, remainder, bps, bytes;
+ uint64_t secs, remainder, bps, bytes;
+
bytes = td->this_io_bytes[ddir];
bps = td->rate_bps[ddir];
secs = bytes / bps;
if (td_write(td) && idx == DDIR_WRITE &&
td->o.do_verify &&
- td->o.verify != VERIFY_NONE)
+ td->o.verify != VERIFY_NONE &&
+ !td->o.experimental_verify)
log_io_piece(td, io_u);
icd->bytes_done[idx] += bytes;
* Complete a single io_u for the sync engines.
*/
int io_u_sync_complete(struct thread_data *td, struct io_u *io_u,
- unsigned long *bytes)
+ uint64_t *bytes)
{
struct io_completion_data icd;
* Called to complete min_events number of io for the async engines.
*/
int io_u_queued_complete(struct thread_data *td, int min_evts,
- unsigned long *bytes)
+ uint64_t *bytes)
{
struct io_completion_data icd;
struct timespec *tvp = NULL;
}
}
-/*
- * "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)
+void fill_io_buffer(struct thread_data *td, void *buf, unsigned int min_write,
+ unsigned int max_bs)
{
- io_u->buf_filled_len = 0;
-
if (!td->o.zero_buffers) {
unsigned int perc = td->o.compress_percentage;
unsigned int seg = min_write;
seg = min(min_write, td->o.compress_chunk);
- fill_random_buf_percentage(&td->buf_state, io_u->buf,
+ if (!seg)
+ seg = min_write;
+
+ fill_random_buf_percentage(&td->buf_state, buf,
perc, seg, max_bs);
} else
- fill_random_buf(&td->buf_state, io_u->buf, max_bs);
+ fill_random_buf(&td->buf_state, buf, max_bs);
} else
- memset(io_u->buf, 0, max_bs);
+ memset(buf, 0, max_bs);
+}
+
+/*
+ * "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)
+{
+ io_u->buf_filled_len = 0;
+ fill_io_buffer(td, io_u->buf, min_write, max_bs);
}