return z->type == ZBD_ZONE_TYPE_SWR;
}
+/**
+ * zbd_zone_end - Return zone end location
+ * @z: zone info pointer.
+ */
+static inline uint64_t zbd_zone_end(const struct fio_zone_info *z)
+{
+ return (z+1)->start;
+}
+
+/**
+ * zbd_zone_capacity_end - Return zone capacity limit end location
+ * @z: zone info pointer.
+ */
+static inline uint64_t zbd_zone_capacity_end(const struct fio_zone_info *z)
+{
+ return z->start + z->capacity;
+}
+
/**
* zbd_zone_full - verify whether a minimum number of bytes remain in a zone
* @f: file pointer.
assert((required & 511) == 0);
return zbd_zone_swr(z) &&
- z->wp + required > z->start + f->zbd_info->zone_size;
+ z->wp + required > zbd_zone_capacity_end(z);
}
static void zone_lock(struct thread_data *td, struct fio_file *f, struct fio_zone_info *z)
z = &f->zbd_info->zone_info[zone_idx];
if ((f->file_offset != z->start) &&
(td->o.td_ddir != TD_DDIR_READ)) {
- new_offset = (z+1)->start;
+ new_offset = zbd_zone_end(z);
if (new_offset >= f->file_offset + f->io_size) {
log_info("%s: io_size must be at least one zone\n",
f->file_name);
f->min_zone = zbd_zone_idx(f, f->file_offset);
f->max_zone = zbd_zone_idx(f, f->file_offset + f->io_size);
+ assert(f->min_zone < f->max_zone);
}
}
uint32_t nr_zones;
struct fio_zone_info *p;
uint64_t zone_size = td->o.zone_size;
+ uint64_t zone_capacity = td->o.zone_capacity;
struct zoned_block_device_info *zbd_info = NULL;
int i;
return 1;
}
+ if (zone_capacity == 0)
+ zone_capacity = zone_size;
+
+ if (zone_capacity > zone_size) {
+ log_err("%s: job parameter zonecapacity %llu is larger than zone size %llu\n",
+ f->file_name, (unsigned long long) td->o.zone_capacity,
+ (unsigned long long) td->o.zone_size);
+ return 1;
+ }
+
nr_zones = (f->real_file_size + zone_size - 1) / zone_size;
zbd_info = scalloc(1, sizeof(*zbd_info) +
(nr_zones + 1) * sizeof(zbd_info->zone_info[0]));
mutex_init_pshared_with_type(&p->mutex,
PTHREAD_MUTEX_RECURSIVE);
p->start = i * zone_size;
- p->wp = p->start + zone_size;
+ p->wp = p->start;
p->type = ZBD_ZONE_TYPE_SWR;
p->cond = ZBD_ZONE_COND_EMPTY;
+ p->capacity = zone_capacity;
}
/* a sentinel */
p->start = nr_zones * zone_size;
mutex_init_pshared_with_type(&p->mutex,
PTHREAD_MUTEX_RECURSIVE);
p->start = z->start;
+ p->capacity = z->capacity;
switch (z->cond) {
case ZBD_ZONE_COND_NOT_WP:
case ZBD_ZONE_COND_FULL:
- p->wp = p->start + zone_size;
+ p->wp = p->start + p->capacity;
break;
default:
assert(z->start <= z->wp);
return ret;
}
+static bool zbd_open_zone(struct thread_data *td, const struct fio_file *f,
+ uint32_t zone_idx);
+static int zbd_reset_zone(struct thread_data *td, struct fio_file *f,
+ struct fio_zone_info *z);
+
int zbd_setup_files(struct thread_data *td)
{
struct fio_file *f;
for_each_file(td, f, i) {
struct zoned_block_device_info *zbd = f->zbd_info;
+ struct fio_zone_info *z;
+ int zi;
if (!zbd)
continue;
log_err("'max_open_zones' value is limited by %u\n", ZBD_MAX_OPEN_ZONES);
return 1;
}
+
+ for (zi = f->min_zone; zi < f->max_zone; zi++) {
+ z = &zbd->zone_info[zi];
+ if (z->cond != ZBD_ZONE_COND_IMP_OPEN &&
+ z->cond != ZBD_ZONE_COND_EXP_OPEN)
+ continue;
+ if (zbd_open_zone(td, f, zi))
+ continue;
+ /*
+ * If the number of open zones exceeds specified limits,
+ * reset all extra open zones.
+ */
+ if (zbd_reset_zone(td, f, z) < 0) {
+ log_err("Failed to reest zone %d\n", zi);
+ return 1;
+ }
+ }
}
return 0;
}
+static unsigned int zbd_zone_nr(struct zoned_block_device_info *zbd_info,
+ struct fio_zone_info *zone)
+{
+ return zone - zbd_info->zone_info;
+}
+
/**
- * zbd_reset_range - reset zones for a range of sectors
+ * zbd_reset_zone - reset the write pointer of a single zone
* @td: FIO thread data.
- * @f: Fio file for which to reset zones
- * @sector: Starting sector in units of 512 bytes
- * @nr_sectors: Number of sectors in units of 512 bytes
+ * @f: FIO file associated with the disk for which to reset a write pointer.
+ * @z: Zone to reset.
*
* Returns 0 upon success and a negative error code upon failure.
+ *
+ * The caller must hold z->mutex.
*/
-static int zbd_reset_range(struct thread_data *td, struct fio_file *f,
- uint64_t offset, uint64_t length)
+static int zbd_reset_zone(struct thread_data *td, struct fio_file *f,
+ struct fio_zone_info *z)
{
- uint32_t zone_idx_b, zone_idx_e;
- struct fio_zone_info *zb, *ze, *z;
+ uint64_t offset = z->start;
+ uint64_t length = (z+1)->start - offset;
int ret = 0;
+ if (z->wp == z->start)
+ return 0;
+
assert(is_valid_offset(f, offset + length - 1));
+ dprint(FD_ZBD, "%s: resetting wp of zone %u.\n", f->file_name,
+ zbd_zone_nr(f->zbd_info, z));
switch (f->zbd_info->model) {
case ZBD_HOST_AWARE:
case ZBD_HOST_MANAGED:
break;
}
- zone_idx_b = zbd_zone_idx(f, offset);
- zb = &f->zbd_info->zone_info[zone_idx_b];
- zone_idx_e = zbd_zone_idx(f, offset + length);
- ze = &f->zbd_info->zone_info[zone_idx_e];
- for (z = zb; z < ze; z++) {
- pthread_mutex_lock(&z->mutex);
- pthread_mutex_lock(&f->zbd_info->mutex);
- f->zbd_info->sectors_with_data -= z->wp - z->start;
- pthread_mutex_unlock(&f->zbd_info->mutex);
- z->wp = z->start;
- z->verify_block = 0;
- pthread_mutex_unlock(&z->mutex);
- }
+ pthread_mutex_lock(&f->zbd_info->mutex);
+ f->zbd_info->sectors_with_data -= z->wp - z->start;
+ pthread_mutex_unlock(&f->zbd_info->mutex);
+ z->wp = z->start;
+ z->verify_block = 0;
- td->ts.nr_zone_resets += ze - zb;
+ td->ts.nr_zone_resets++;
return ret;
}
-static unsigned int zbd_zone_nr(struct zoned_block_device_info *zbd_info,
- struct fio_zone_info *zone)
-{
- return zone - zbd_info->zone_info;
-}
-
-/**
- * zbd_reset_zone - reset the write pointer of a single zone
- * @td: FIO thread data.
- * @f: FIO file associated with the disk for which to reset a write pointer.
- * @z: Zone to reset.
- *
- * Returns 0 upon success and a negative error code upon failure.
- */
-static int zbd_reset_zone(struct thread_data *td, struct fio_file *f,
- struct fio_zone_info *z)
-{
- dprint(FD_ZBD, "%s: resetting wp of zone %u.\n", f->file_name,
- zbd_zone_nr(f->zbd_info, z));
-
- return zbd_reset_range(td, f, z->start, (z+1)->start - z->start);
-}
-
/* The caller must hold f->zbd_info->mutex */
static void zbd_close_zone(struct thread_data *td, const struct fio_file *f,
- unsigned int open_zone_idx)
+ unsigned int zone_idx)
{
- uint32_t zone_idx;
+ uint32_t open_zone_idx = 0;
+
+ for (; open_zone_idx < f->zbd_info->num_open_zones; open_zone_idx++) {
+ if (f->zbd_info->open_zones[open_zone_idx] == zone_idx)
+ break;
+ }
+ if (open_zone_idx == f->zbd_info->num_open_zones) {
+ dprint(FD_ZBD, "%s: zone %d is not open\n",
+ f->file_name, zone_idx);
+ return;
+ }
- assert(open_zone_idx < f->zbd_info->num_open_zones);
- zone_idx = f->zbd_info->open_zones[open_zone_idx];
+ dprint(FD_ZBD, "%s: closing zone %d\n", f->file_name, zone_idx);
memmove(f->zbd_info->open_zones + open_zone_idx,
f->zbd_info->open_zones + open_zone_idx + 1,
(ZBD_MAX_OPEN_ZONES - (open_zone_idx + 1)) *
continue;
zone_lock(td, f, z);
if (all_zones) {
- unsigned int i;
-
pthread_mutex_lock(&f->zbd_info->mutex);
- for (i = 0; i < f->zbd_info->num_open_zones; i++) {
- if (f->zbd_info->open_zones[i] == nz)
- zbd_close_zone(td, f, i);
- }
+ zbd_close_zone(td, f, nz);
pthread_mutex_unlock(&f->zbd_info->mutex);
reset_wp = z->wp != z->start;
struct fio_zone_info *zb, *ze, *z;
uint64_t swd = 0;
- zb = &f->zbd_info->zone_info[zbd_zone_idx(f, f->file_offset)];
- ze = &f->zbd_info->zone_info[zbd_zone_idx(f, f->file_offset +
- f->io_size)];
+ zb = &f->zbd_info->zone_info[f->min_zone];
+ ze = &f->zbd_info->zone_info[f->max_zone];
for (z = zb; z < ze; z++) {
pthread_mutex_lock(&z->mutex);
swd += z->wp - z->start;
* was not yet open and opening a new zone would cause the zone limit to be
* exceeded.
*/
-static bool zbd_open_zone(struct thread_data *td, const struct io_u *io_u,
+static bool zbd_open_zone(struct thread_data *td, const struct fio_file *f,
uint32_t zone_idx)
{
const uint32_t min_bs = td->o.min_bs[DDIR_WRITE];
- const struct fio_file *f = io_u->file;
struct fio_zone_info *z = &f->zbd_info->zone_info[zone_idx];
bool res = true;
return false;
pthread_mutex_lock(&f->zbd_info->mutex);
- if (is_zone_open(td, f, zone_idx))
+ if (is_zone_open(td, f, zone_idx)) {
+ /*
+ * If the zone is already open and going to be full by writes
+ * in-flight, handle it as a full zone instead of an open zone.
+ */
+ if (z->wp >= zbd_zone_capacity_end(z))
+ res = false;
goto out;
+ }
res = false;
/* Zero means no limit */
if (td->o.job_max_open_zones > 0 &&
unsigned int open_zone_idx = -1;
uint32_t zone_idx, new_zone_idx;
int i;
+ bool wait_zone_close;
assert(is_valid_offset(f, io_u->offset));
if (td->o.max_open_zones == 0 && td->o.job_max_open_zones == 0)
goto examine_zone;
if (f->zbd_info->num_open_zones == 0) {
- pthread_mutex_unlock(&f->zbd_info->mutex);
- pthread_mutex_unlock(&z->mutex);
dprint(FD_ZBD, "%s(%s): no zones are open\n",
__func__, f->file_name);
- return NULL;
+ goto open_other_zone;
}
/*
/* Both z->mutex and f->zbd_info->mutex are held. */
examine_zone:
- if (z->wp + min_bs <= (z+1)->start) {
+ if (z->wp + min_bs <= zbd_zone_capacity_end(z)) {
pthread_mutex_unlock(&f->zbd_info->mutex);
goto out;
}
- dprint(FD_ZBD, "%s(%s): closing zone %d\n", __func__, f->file_name,
- zone_idx);
- if (td->o.max_open_zones || td->o.job_max_open_zones)
- zbd_close_zone(td, f, open_zone_idx);
+
+open_other_zone:
+ /* Check if number of open zones reaches one of limits. */
+ wait_zone_close =
+ f->zbd_info->num_open_zones == f->max_zone - f->min_zone ||
+ (td->o.max_open_zones &&
+ f->zbd_info->num_open_zones == td->o.max_open_zones) ||
+ (td->o.job_max_open_zones &&
+ td->num_open_zones == td->o.job_max_open_zones);
+
pthread_mutex_unlock(&f->zbd_info->mutex);
/* Only z->mutex is held. */
+ /*
+ * When number of open zones reaches to one of limits, wait for
+ * zone close before opening a new zone.
+ */
+ if (wait_zone_close) {
+ dprint(FD_ZBD, "%s(%s): quiesce to allow open zones to close\n",
+ __func__, f->file_name);
+ io_u_quiesce(td);
+ }
+
/* Zone 'z' is full, so try to open a new zone. */
for (i = f->io_size / f->zbd_info->zone_size; i > 0; i--) {
zone_idx++;
zone_lock(td, f, z);
if (z->open)
continue;
- if (zbd_open_zone(td, io_u, zone_idx))
+ if (zbd_open_zone(td, f, zone_idx))
goto out;
}
z = &f->zbd_info->zone_info[zone_idx];
zone_lock(td, f, z);
- if (z->wp + min_bs <= (z+1)->start)
+ if (z->wp + min_bs <= zbd_zone_capacity_end(z))
goto out;
pthread_mutex_lock(&f->zbd_info->mutex);
}
const struct fio_file *f = io_u->file;
const uint32_t min_bs = td->o.min_bs[DDIR_WRITE];
- if (!zbd_open_zone(td, io_u, z - f->zbd_info->zone_info)) {
+ if (!zbd_open_zone(td, f, z - f->zbd_info->zone_info)) {
pthread_mutex_unlock(&z->mutex);
z = zbd_convert_to_open_zone(td, io_u);
assert(z);
}
- if (z->verify_block * min_bs >= f->zbd_info->zone_size)
+ if (z->verify_block * min_bs >= z->capacity)
log_err("%s: %d * %d >= %llu\n", f->file_name, z->verify_block,
- min_bs, (unsigned long long) f->zbd_info->zone_size);
+ min_bs, (unsigned long long)z->capacity);
io_u->offset = z->start + z->verify_block++ * min_bs;
return z;
}
struct fio_file *f = io_u->file;
struct fio_zone_info *z1, *z2;
const struct fio_zone_info *const zf =
- &f->zbd_info->zone_info[zbd_zone_idx(f, f->file_offset)];
+ &f->zbd_info->zone_info[f->min_zone];
/*
* Skip to the next non-empty zone in case of sequential I/O and to
return NULL;
}
+/**
+ * zbd_end_zone_io - update zone status at command completion
+ * @io_u: I/O unit
+ * @z: zone info pointer
+ *
+ * If the write command made the zone full, close it.
+ *
+ * The caller must hold z->mutex.
+ */
+static void zbd_end_zone_io(struct thread_data *td, const struct io_u *io_u,
+ struct fio_zone_info *z)
+{
+ const struct fio_file *f = io_u->file;
+
+ if (io_u->ddir == DDIR_WRITE &&
+ io_u->offset + io_u->buflen >= zbd_zone_capacity_end(z)) {
+ pthread_mutex_lock(&f->zbd_info->mutex);
+ zbd_close_zone(td, f, z - f->zbd_info->zone_info);
+ pthread_mutex_unlock(&f->zbd_info->mutex);
+ }
+}
+
/**
* zbd_queue_io - update the write pointer of a sequential zone
* @io_u: I/O unit
* For write and trim operations, update the write pointer of the I/O unit
* target zone.
*/
-static void zbd_queue_io(struct io_u *io_u, int q, bool success)
+static void zbd_queue_io(struct thread_data *td, struct io_u *io_u, int q,
+ bool success)
{
const struct fio_file *f = io_u->file;
struct zoned_block_device_info *zbd_info = f->zbd_info;
switch (io_u->ddir) {
case DDIR_WRITE:
zone_end = min((uint64_t)(io_u->offset + io_u->buflen),
- (z + 1)->start);
+ zbd_zone_capacity_end(z));
pthread_mutex_lock(&zbd_info->mutex);
/*
* z->wp > zone_end means that one or more I/O errors
break;
}
+ if (q == FIO_Q_COMPLETED && !io_u->error)
+ zbd_end_zone_io(td, io_u, z);
+
unlock:
if (!success || q != FIO_Q_QUEUED) {
/* BUSY or COMPLETED: unlock the zone */
* zbd_put_io - Unlock an I/O unit target zone lock
* @io_u: I/O unit
*/
-static void zbd_put_io(const struct io_u *io_u)
+static void zbd_put_io(struct thread_data *td, const struct io_u *io_u)
{
const struct fio_file *f = io_u->file;
struct zoned_block_device_info *zbd_info = f->zbd_info;
"%s: terminate I/O (%lld, %llu) for zone %u\n",
f->file_name, io_u->offset, io_u->buflen, zone_idx);
+ zbd_end_zone_io(td, io_u, z);
+
ret = pthread_mutex_unlock(&z->mutex);
assert(ret == 0);
zbd_check_swd(f);
assert(td->o.zone_mode == ZONE_MODE_ZBD);
assert(td->o.zone_size);
+ zone_idx = zbd_zone_idx(f, f->last_pos[ddir]);
+ z = &f->zbd_info->zone_info[zone_idx];
+
+ /*
+ * When the zone capacity is smaller than the zone size and the I/O is
+ * sequential write, skip to zone end if the latest position is at the
+ * zone capacity limit.
+ */
+ if (z->capacity < f->zbd_info->zone_size && !td_random(td) &&
+ ddir == DDIR_WRITE &&
+ f->last_pos[ddir] >= zbd_zone_capacity_end(z)) {
+ dprint(FD_ZBD,
+ "%s: Jump from zone capacity limit to zone end:"
+ " (%llu -> %llu) for zone %u (%llu)\n",
+ f->file_name, (unsigned long long) f->last_pos[ddir],
+ (unsigned long long) zbd_zone_end(z),
+ zbd_zone_nr(f->zbd_info, z),
+ (unsigned long long) z->capacity);
+ td->io_skip_bytes += zbd_zone_end(z) - f->last_pos[ddir];
+ f->last_pos[ddir] = zbd_zone_end(z);
+ }
+
/*
* zone_skip is valid only for sequential workloads.
*/
* - For reads with td->o.read_beyond_wp == false, the last position
* reached the zone write pointer.
*/
- zone_idx = zbd_zone_idx(f, f->last_pos[ddir]);
- z = &f->zbd_info->zone_info[zone_idx];
-
if (td->zone_bytes >= td->o.zone_size ||
- f->last_pos[ddir] >= (z+1)->start ||
+ f->last_pos[ddir] >= zbd_zone_end(z) ||
(ddir == DDIR_READ &&
(!td->o.read_beyond_wp) && f->last_pos[ddir] >= z->wp)) {
/*
}
/**
- * zbd_adjust_ddir - Adjust an I/O direction for zonedmode=zbd.
+ * zbd_adjust_ddir - Adjust an I/O direction for zonemode=zbd.
*
* @td: FIO thread data.
* @io_u: FIO I/O unit.
switch (io_u->ddir) {
case DDIR_READ:
- if (td->runstate == TD_VERIFYING) {
- if (td_write(td))
- zb = zbd_replay_write_order(td, io_u, zb);
+ if (td->runstate == TD_VERIFYING && td_write(td)) {
+ zb = zbd_replay_write_order(td, io_u, zb);
+ pthread_mutex_unlock(&zb->mutex);
goto accept;
}
/*
if (range < min_bs ||
((!td_random(td)) && (io_u->offset + min_bs > zb->wp))) {
pthread_mutex_unlock(&zb->mutex);
- zl = &f->zbd_info->zone_info[zbd_zone_idx(f,
- f->file_offset + f->io_size)];
+ zl = &f->zbd_info->zone_info[f->max_zone];
zb = zbd_find_zone(td, io_u, zb, zl);
if (!zb) {
dprint(FD_ZBD,
case DDIR_WRITE:
if (io_u->buflen > f->zbd_info->zone_size)
goto eof;
- if (!zbd_open_zone(td, io_u, zone_idx_b)) {
+ if (!zbd_open_zone(td, f, zone_idx_b)) {
pthread_mutex_unlock(&zb->mutex);
zb = zbd_convert_to_open_zone(td, io_u);
if (!zb)
zb->reset_zone = 0;
if (zbd_reset_zone(td, f, zb) < 0)
goto eof;
+
+ if (zb->capacity < min_bs) {
+ log_err("zone capacity %llu smaller than minimum block size %d\n",
+ (unsigned long long)zb->capacity,
+ min_bs);
+ goto eof;
+ }
}
/* Make writes occur at the write pointer */
assert(!zbd_zone_full(f, zb, min_bs));
* small.
*/
new_len = min((unsigned long long)io_u->buflen,
- (zb + 1)->start - io_u->offset);
+ zbd_zone_capacity_end(zb) - io_u->offset);
new_len = new_len / min_bs * min_bs;
if (new_len == io_u->buflen)
goto accept;
goto accept;
}
log_err("Zone remainder %lld smaller than minimum block size %d\n",
- ((zb + 1)->start - io_u->offset),
+ (zbd_zone_capacity_end(zb) - io_u->offset),
min_bs);
goto eof;
case DDIR_TRIM: