/*
* Return the i/o class to issue from, or ZIO_PRIORITY_MAX_QUEUEABLE if
* there is no eligible class.
*/
static zio_priority_t
vdev_queue_class_to_issue(vdev_queue_t *vq)
{
spa_t *spa = vq->vq_vdev->vdev_spa;
zio_priority_t p;
if (avl_numnodes(&vq->vq_active_tree) >= zfs_vdev_max_active)
return (ZIO_PRIORITY_NUM_QUEUEABLE);
/* find a queue that has not reached its minimum # outstanding i/os */
for (p = 0; p < ZIO_PRIORITY_NUM_QUEUEABLE; p++) {
if (avl_numnodes(vdev_queue_class_tree(vq, p)) > 0 &&
vq->vq_class[p].vqc_active <
vdev_queue_class_min_active(p))
return (p);
}
/*
* If we haven't found a queue, look for one that hasn't reached its
* maximum # outstanding i/os.
*/
for (p = 0; p < ZIO_PRIORITY_NUM_QUEUEABLE; p++) {
if (avl_numnodes(vdev_queue_class_tree(vq, p)) > 0 &&
vq->vq_class[p].vqc_active <
vdev_queue_class_max_active(spa, p))
return (p);
}
/* No eligible queued i/os */
return (ZIO_PRIORITY_NUM_QUEUEABLE);
}
void
vdev_queue_init(vdev_t *vd)
{
vdev_queue_t *vq = &vd->vdev_queue;
mutex_init(&vq->vq_lock, NULL, MUTEX_DEFAULT, NULL);
vq->vq_vdev = vd;
avl_create(&vq->vq_active_tree, vdev_queue_offset_compare,
sizeof (zio_t), offsetof(struct zio, io_queue_node));
avl_create(vdev_queue_type_tree(vq, ZIO_TYPE_READ),
vdev_queue_offset_compare, sizeof (zio_t),
offsetof(struct zio, io_offset_node));
avl_create(vdev_queue_type_tree(vq, ZIO_TYPE_WRITE),
vdev_queue_offset_compare, sizeof (zio_t),
offsetof(struct zio, io_offset_node));
for (zio_priority_t p = 0; p < ZIO_PRIORITY_NUM_QUEUEABLE; p++) {
int (*compfn) (const void *, const void *);
/*
* The synchronous i/o queues are dispatched in FIFO rather
* than LBA order. This provides more consistent latency for
* these i/os.
*/
if (p == ZIO_PRIORITY_SYNC_READ || p == ZIO_PRIORITY_SYNC_WRITE)
compfn = vdev_queue_timestamp_compare;
else
compfn = vdev_queue_offset_compare;
avl_create(vdev_queue_class_tree(vq, p), compfn,
sizeof (zio_t), offsetof(struct zio, io_queue_node));
}
}
void
vdev_queue_fini(vdev_t *vd)
{
vdev_queue_t *vq = &vd->vdev_queue;
for (zio_priority_t p = 0; p < ZIO_PRIORITY_NUM_QUEUEABLE; p++)
avl_destroy(vdev_queue_class_tree(vq, p));
avl_destroy(&vq->vq_active_tree);
avl_destroy(vdev_queue_type_tree(vq, ZIO_TYPE_READ));
avl_destroy(vdev_queue_type_tree(vq, ZIO_TYPE_WRITE));
mutex_destroy(&vq->vq_lock);
}
static void
vdev_queue_io_add(vdev_queue_t *vq, zio_t *zio)
{
spa_t *spa = zio->io_spa;
ASSERT3U(zio->io_priority, <, ZIO_PRIORITY_NUM_QUEUEABLE);
avl_add(vdev_queue_class_tree(vq, zio->io_priority), zio);
avl_add(vdev_queue_type_tree(vq, zio->io_type), zio);
mutex_enter(&spa->spa_iokstat_lock);
spa->spa_queue_stats[zio->io_priority].spa_queued++;
if (spa->spa_iokstat != NULL)
kstat_waitq_enter(spa->spa_iokstat->ks_data);
mutex_exit(&spa->spa_iokstat_lock);
}
static void
vdev_queue_io_remove(vdev_queue_t *vq, zio_t *zio)
{
spa_t *spa = zio->io_spa;
spa_stats_history_t *ssh = &spa->spa_stats.io_history;
ASSERT3U(zio->io_priority, <, ZIO_PRIORITY_NUM_QUEUEABLE);
avl_remove(vdev_queue_class_tree(vq, zio->io_priority), zio);
avl_remove(vdev_queue_type_tree(vq, zio->io_type), zio);
if (ssh->kstat != NULL) {
mutex_enter(&ssh->lock);
kstat_waitq_exit(ssh->kstat->ks_data);
mutex_exit(&ssh->lock);
}
}
static void
vdev_queue_io_add(vdev_queue_t *vq, zio_t *zio)
{
spa_t *spa = zio->io_spa;
avl_tree_t *qtt;
ASSERT(MUTEX_HELD(&vq->vq_lock));
ASSERT3U(zio->io_priority, <, ZIO_PRIORITY_NUM_QUEUEABLE);
avl_add(vdev_queue_class_tree(vq, zio->io_priority), zio);
qtt = vdev_queue_type_tree(vq, zio->io_type);
if (qtt)
avl_add(qtt, zio);
#ifdef illumos
mutex_enter(&spa->spa_iokstat_lock);
spa->spa_queue_stats[zio->io_priority].spa_queued++;
if (spa->spa_iokstat != NULL)
kstat_waitq_enter(spa->spa_iokstat->ks_data);
mutex_exit(&spa->spa_iokstat_lock);
#endif
}
