#include "fio.h"
#include "hash.h"
#include "verify.h"
+#include "trim.h"
+#include "lib/rand.h"
struct io_completion_data {
int nr; /* input */
struct fio_file *f = io_u->file;
unsigned long long block;
unsigned int blocks, nr_blocks;
+ int busy_check;
block = (io_u->offset - f->file_offset) / (unsigned long long) min_bs;
nr_blocks = (io_u->buflen + min_bs - 1) / min_bs;
blocks = 0;
+ busy_check = !(io_u->flags & IO_U_F_BUSY_OK);
while (nr_blocks) {
unsigned int this_blocks, mask;
* If we have a mixed random workload, we may
* encounter blocks we already did IO to.
*/
- if ((td->o.ddir_nr == 1) && !random_map_free(f, block)) {
- if (!blocks)
- blocks = 1;
+ if (!busy_check) {
+ blocks = nr_blocks;
break;
}
+ if ((td->o.ddir_seq_nr == 1) && !random_map_free(f, block))
+ break;
idx = RAND_MAP_IDX(f, block);
bit = RAND_MAP_BIT(f, block);
if (this_blocks + bit > BLOCKS_PER_MAP)
this_blocks = BLOCKS_PER_MAP - bit;
- if (this_blocks == BLOCKS_PER_MAP)
- mask = -1U;
- else
- mask = ((1U << this_blocks) - 1) << bit;
+ do {
+ if (this_blocks == BLOCKS_PER_MAP)
+ mask = -1U;
+ else
+ mask = ((1U << this_blocks) - 1) << bit;
+
+ if (!(f->file_map[idx] & mask))
+ break;
+
+ this_blocks--;
+ } while (this_blocks);
+
+ if (!this_blocks)
+ break;
f->file_map[idx] |= mask;
nr_blocks -= this_blocks;
unsigned long long max_blocks;
unsigned long long max_size;
+ assert(ddir_rw(ddir));
+
/*
* Hmm, should we make sure that ->io_size <= ->real_file_size?
*/
return get_next_free_block(td, f, ddir, b);
}
+static int get_next_rand_block(struct thread_data *td, struct fio_file *f,
+ enum fio_ddir ddir, unsigned long long *b)
+{
+ if (get_next_rand_offset(td, f, ddir, b)) {
+ dprint(FD_IO, "%s: rand offset failed, last=%llu, size=%llu\n",
+ f->file_name, f->last_pos, f->real_file_size);
+ return 1;
+ }
+
+ return 0;
+}
+
+static int get_next_seq_block(struct thread_data *td, struct fio_file *f,
+ enum fio_ddir ddir, unsigned long long *b)
+{
+ assert(ddir_rw(ddir));
+
+ if (f->last_pos < f->real_file_size) {
+ *b = (f->last_pos - f->file_offset) / td->o.min_bs[ddir];
+ return 0;
+ }
+
+ return 1;
+}
+
+static int get_next_block(struct thread_data *td, struct io_u *io_u,
+ enum fio_ddir ddir, int rw_seq, unsigned long long *b)
+{
+ struct fio_file *f = io_u->file;
+ int ret;
+
+ assert(ddir_rw(ddir));
+
+ if (rw_seq) {
+ if (td_random(td))
+ ret = get_next_rand_block(td, f, ddir, b);
+ else
+ ret = get_next_seq_block(td, f, ddir, b);
+ } else {
+ io_u->flags |= IO_U_F_BUSY_OK;
+
+ if (td->o.rw_seq == RW_SEQ_SEQ) {
+ ret = get_next_seq_block(td, f, ddir, b);
+ if (ret)
+ ret = get_next_rand_block(td, f, ddir, b);
+ } else if (td->o.rw_seq == RW_SEQ_IDENT) {
+ if (f->last_start != -1ULL)
+ *b = (f->last_start - f->file_offset) / td->o.min_bs[ddir];
+ else
+ *b = 0;
+ ret = 0;
+ } else {
+ log_err("fio: unknown rw_seq=%d\n", td->o.rw_seq);
+ ret = 1;
+ }
+ }
+
+ return ret;
+}
+
/*
* For random io, generate a random new block and see if it's used. Repeat
* 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)
+static int __get_next_offset(struct thread_data *td, struct io_u *io_u)
{
struct fio_file *f = io_u->file;
unsigned long long b;
enum fio_ddir ddir = io_u->ddir;
+ int rw_seq_hit = 0;
- if (td_random(td) && (td->o.ddir_nr && !--td->ddir_nr)) {
- td->ddir_nr = td->o.ddir_nr;
+ assert(ddir_rw(ddir));
- if (get_next_rand_offset(td, f, ddir, &b)) {
- dprint(FD_IO, "%s: getting rand offset failed\n",
- f->file_name);
- return 1;
- }
- } else {
- if (f->last_pos >= f->real_file_size) {
- if (!td_random(td) ||
- get_next_rand_offset(td, f, ddir, &b)) {
- dprint(FD_IO, "%s: pos %llu > size %llu\n",
- f->file_name, f->last_pos,
- f->real_file_size);
- return 1;
- }
- } else
- b = (f->last_pos - f->file_offset) / td->o.min_bs[ddir];
+ if (td->o.ddir_seq_nr && !--td->ddir_seq_nr) {
+ rw_seq_hit = 1;
+ td->ddir_seq_nr = td->o.ddir_seq_nr;
+ }
+
+ if (get_next_block(td, io_u, ddir, rw_seq_hit, &b)) {
+ printf("fail\n");
+ return 1;
}
io_u->offset = b * td->o.ba[ddir];
return 0;
}
-static unsigned int get_next_buflen(struct thread_data *td, struct io_u *io_u)
+static int get_next_offset(struct thread_data *td, struct io_u *io_u)
+{
+ struct prof_io_ops *ops = &td->prof_io_ops;
+
+ if (ops->fill_io_u_off)
+ return ops->fill_io_u_off(td, io_u);
+
+ return __get_next_offset(td, io_u);
+}
+
+static unsigned int __get_next_buflen(struct thread_data *td, struct io_u *io_u)
{
const int ddir = io_u->ddir;
unsigned int uninitialized_var(buflen);
unsigned int minbs, maxbs;
long r;
+ assert(ddir_rw(ddir));
+
minbs = td->o.min_bs[ddir];
maxbs = td->o.max_bs[ddir];
return buflen;
}
+static unsigned int get_next_buflen(struct thread_data *td, struct io_u *io_u)
+{
+ struct prof_io_ops *ops = &td->prof_io_ops;
+
+ if (ops->fill_io_u_size)
+ return ops->fill_io_u_size(td, io_u);
+
+ return __get_next_buflen(td, io_u);
+}
+
static void set_rwmix_bytes(struct thread_data *td)
{
unsigned int diff;
struct timeval t;
long usec;
+ assert(ddir_rw(ddir));
+
if (td->rate_pending_usleep[ddir] <= 0)
return ddir;
td->io_issues[DDIR_WRITE] && should_fsync(td))
return DDIR_DATASYNC;
+ /*
+ * see if it's time to sync_file_range
+ */
+ if (td->sync_file_range_nr &&
+ !(td->io_issues[DDIR_WRITE] % td->sync_file_range_nr) &&
+ td->io_issues[DDIR_WRITE] && should_fsync(td))
+ return DDIR_SYNC_FILE_RANGE;
+
if (td_rw(td)) {
/*
* Check if it's time to seed a new data direction.
{
td_io_u_lock(td);
- assert((io_u->flags & IO_U_F_FREE) == 0);
io_u->flags |= IO_U_F_FREE;
io_u->flags &= ~IO_U_F_FREE_DEF;
put_file_log(td, io_u->file);
io_u->file = NULL;
+ 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);
- td->cur_depth--;
td_io_u_unlock(td);
td_io_u_free_notify(td);
}
td_io_u_lock(td);
__io_u->flags |= IO_U_F_FREE;
- if ((__io_u->flags & IO_U_F_FLIGHT) && !ddir_sync(__io_u->ddir))
+ if ((__io_u->flags & IO_U_F_FLIGHT) && ddir_rw(__io_u->ddir))
td->io_issues[__io_u->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);
- td->cur_depth--;
td_io_u_unlock(td);
*io_u = NULL;
}
io_u->ddir = get_rw_ddir(td);
/*
- * fsync() or fdatasync(), we are done
+ * fsync() or fdatasync() or trim etc, we are done
*/
- if (ddir_sync(io_u->ddir))
+ if (!ddir_rw(io_u->ddir))
goto out;
/*
return f;
}
-static struct fio_file *get_next_file(struct thread_data *td)
+static struct fio_file *__get_next_file(struct thread_data *td)
{
struct fio_file *f;
return f;
}
+static struct fio_file *get_next_file(struct thread_data *td)
+{
+ 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);
+}
+
static int set_io_u_file(struct thread_data *td, struct io_u *io_u)
{
struct fio_file *f;
io_u->end_io = NULL;
}
- /*
- * We ran out, wait for async verify threads to finish and return one
- */
- if (!io_u && td->o.verify_async) {
- pthread_cond_wait(&td->free_cond, &td->io_u_lock);
- goto again;
- }
-
if (io_u) {
assert(io_u->flags & IO_U_F_FREE);
- io_u->flags &= ~IO_U_F_FREE;
- io_u->flags &= ~IO_U_F_FREE_DEF;
+ io_u->flags &= ~(IO_U_F_FREE | IO_U_F_FREE_DEF);
io_u->error = 0;
flist_del(&io_u->list);
flist_add(&io_u->list, &td->io_u_busylist);
td->cur_depth++;
+ io_u->flags |= IO_U_F_IN_CUR_DEPTH;
+ } else if (td->o.verify_async) {
+ /*
+ * We ran out, wait for async verify threads to finish and
+ * return one
+ */
+ pthread_cond_wait(&td->free_cond, &td->io_u_lock);
+ goto again;
}
td_io_u_unlock(td);
return io_u;
}
+static int check_get_trim(struct thread_data *td, struct io_u *io_u)
+{
+ if (td->o.trim_backlog && td->trim_entries) {
+ int get_trim = 0;
+
+ if (td->trim_batch) {
+ td->trim_batch--;
+ get_trim = 1;
+ } else if (!(td->io_hist_len % td->o.trim_backlog) &&
+ td->last_ddir != DDIR_READ) {
+ td->trim_batch = td->o.trim_batch;
+ if (!td->trim_batch)
+ td->trim_batch = td->o.trim_backlog;
+ get_trim = 1;
+ }
+
+ if (get_trim && !get_next_trim(td, io_u))
+ return 1;
+ }
+
+ return 0;
+}
+
+static int check_get_verify(struct thread_data *td, struct io_u *io_u)
+{
+ if (td->o.verify_backlog && td->io_hist_len) {
+ int get_verify = 0;
+
+ if (td->verify_batch) {
+ td->verify_batch--;
+ get_verify = 1;
+ } else if (!(td->io_hist_len % td->o.verify_backlog) &&
+ td->last_ddir != DDIR_READ) {
+ td->verify_batch = td->o.verify_batch;
+ if (!td->verify_batch)
+ td->verify_batch = td->o.verify_backlog;
+ get_verify = 1;
+ }
+
+ if (get_verify && !get_next_verify(td, io_u))
+ return 1;
+ }
+
+ return 0;
+}
+
/*
* 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.
return NULL;
}
+ if (check_get_verify(td, io_u))
+ goto out;
+ if (check_get_trim(td, io_u))
+ goto out;
+
/*
* from a requeue, io_u already setup
*/
f = io_u->file;
assert(fio_file_open(f));
- if (!ddir_sync(io_u->ddir)) {
+ if (ddir_rw(io_u->ddir)) {
if (!io_u->buflen && !(td->io_ops->flags & FIO_NOIO)) {
dprint(FD_IO, "get_io_u: zero buflen on %p\n", io_u);
goto err_put;
}
+ f->last_start = io_u->offset;
f->last_pos = io_u->offset + io_u->buflen;
if (td->o.verify != VERIFY_NONE && io_u->ddir == DDIR_WRITE)
populate_verify_io_u(td, io_u);
else if (td->o.refill_buffers && io_u->ddir == DDIR_WRITE)
io_u_fill_buffer(td, io_u, io_u->xfer_buflen);
+ else if (io_u->ddir == DDIR_READ) {
+ /*
+ * Reset the buf_filled parameters so next time if the
+ * buffer is used for writes it is refilled.
+ */
+ io_u->buf_filled_len = 0;
+ }
}
/*
io_u->xfer_buflen = io_u->buflen;
out:
+ assert(io_u->file);
if (!td_io_prep(td, io_u)) {
if (!td->o.disable_slat)
fio_gettime(&io_u->start_time, NULL);
void io_u_log_error(struct thread_data *td, struct io_u *io_u)
{
- const char *msg[] = { "read", "write", "sync" };
+ const char *msg[] = { "read", "write", "sync", "datasync",
+ "sync_file_range", "wait", "trim" };
+
+
log_err("fio: io_u error");
* initialized, silence that warning.
*/
unsigned long uninitialized_var(usec);
+ struct fio_file *f;
dprint_io_u(io_u, "io complete");
+ td_io_u_lock(td);
assert(io_u->flags & IO_U_F_FLIGHT);
- io_u->flags &= ~IO_U_F_FLIGHT;
+ io_u->flags &= ~(IO_U_F_FLIGHT | IO_U_F_BUSY_OK);
+ td_io_u_unlock(td);
if (ddir_sync(io_u->ddir)) {
td->last_was_sync = 1;
+ f = io_u->file;
+ if (f) {
+ f->first_write = -1ULL;
+ f->last_write = -1ULL;
+ }
return;
}
td->last_was_sync = 0;
+ td->last_ddir = io_u->ddir;
- if (!io_u->error) {
+ if (!io_u->error && ddir_rw(io_u->ddir)) {
unsigned int bytes = io_u->buflen - io_u->resid;
const enum fio_ddir idx = io_u->ddir;
+ const enum fio_ddir odx = io_u->ddir ^ 1;
int ret;
td->io_blocks[idx]++;
td->io_bytes[idx] += bytes;
td->this_io_bytes[idx] += bytes;
+ if (idx == DDIR_WRITE) {
+ f = io_u->file;
+ if (f) {
+ if (f->first_write == -1ULL ||
+ io_u->offset < f->first_write)
+ f->first_write = io_u->offset;
+ if (f->last_write == -1ULL ||
+ ((io_u->offset + bytes) > f->last_write))
+ f->last_write = io_u->offset + bytes;
+ }
+ }
+
if (ramp_time_over(td)) {
unsigned long uninitialized_var(lusec);
- unsigned long uninitialized_var(rusec);
if (!td->o.disable_clat || !td->o.disable_bw)
lusec = utime_since(&io_u->issue_time,
&icd->time);
- if (__should_check_rate(td, idx) ||
- __should_check_rate(td, idx ^ 1))
- rusec = utime_since(&io_u->start_time,
- &icd->time);
+ if (!td->o.disable_lat) {
+ unsigned long tusec;
+ tusec = utime_since(&io_u->start_time,
+ &icd->time);
+ add_lat_sample(td, idx, tusec, bytes);
+ }
if (!td->o.disable_clat) {
add_clat_sample(td, idx, lusec, bytes);
io_u_mark_latency(td, lusec);
if (!td->o.disable_bw)
add_bw_sample(td, idx, bytes, &icd->time);
if (__should_check_rate(td, idx)) {
- td->rate_pending_usleep[idx] +=
- (long) td->rate_usec_cycle[idx] - rusec;
+ td->rate_pending_usleep[idx] =
+ ((td->this_io_bytes[idx] *
+ td->rate_nsec_cycle[idx]) / 1000 -
+ utime_since_now(&td->start));
}
if (__should_check_rate(td, idx ^ 1))
- td->rate_pending_usleep[idx ^ 1] -= rusec;
+ td->rate_pending_usleep[odx] =
+ ((td->this_io_bytes[odx] *
+ td->rate_nsec_cycle[odx]) / 1000 -
+ utime_since_now(&td->start));
}
if (td_write(td) && idx == DDIR_WRITE &&
if (ret && !icd->error)
icd->error = ret;
}
- } else {
+ } else if (io_u->error) {
icd->error = io_u->error;
io_u_log_error(td, io_u);
}
void io_u_fill_buffer(struct thread_data *td, struct io_u *io_u,
unsigned int max_bs)
{
- long *ptr = io_u->buf;
-
- if (!td->o.zero_buffers) {
- while ((void *) ptr - io_u->buf < max_bs) {
- *ptr = rand() * GOLDEN_RATIO_PRIME;
- ptr++;
- }
- } else
- memset(ptr, 0, max_bs);
+ if (!td->o.zero_buffers)
+ fill_random_buf(io_u->buf, max_bs);
+ else
+ memset(io_u->buf, 0, max_bs);
}