#include <linux/bio.h>
#include <linux/blktrace_api.h>
#include "blk-cgroup.h"
+#include "blk.h"
/* Max dispatch from a group in 1 round */
static int throtl_grp_quantum = 8;
/* Some throttle limits got updated for the group */
int limits_changed;
+
+ struct rcu_head rcu_head;
};
struct throtl_data
/* service tree for active throtl groups */
struct throtl_rb_root tg_service_tree;
- struct throtl_grp root_tg;
+ struct throtl_grp *root_tg;
struct request_queue *queue;
/* Total Number of queued bios on READ and WRITE lists */
return NULL;
}
-static inline int total_nr_queued(struct throtl_data *td)
+static inline unsigned int total_nr_queued(struct throtl_data *td)
{
- return (td->nr_queued[0] + td->nr_queued[1]);
+ return td->nr_queued[0] + td->nr_queued[1];
}
static inline struct throtl_grp *throtl_ref_get_tg(struct throtl_grp *tg)
return tg;
}
+static void throtl_free_tg(struct rcu_head *head)
+{
+ struct throtl_grp *tg;
+
+ tg = container_of(head, struct throtl_grp, rcu_head);
+ free_percpu(tg->blkg.stats_cpu);
+ kfree(tg);
+}
+
static void throtl_put_tg(struct throtl_grp *tg)
{
BUG_ON(atomic_read(&tg->ref) <= 0);
if (!atomic_dec_and_test(&tg->ref))
return;
- kfree(tg);
+
+ /*
+ * A group is freed in rcu manner. But having an rcu lock does not
+ * mean that one can access all the fields of blkg and assume these
+ * are valid. For example, don't try to follow throtl_data and
+ * request queue links.
+ *
+ * Having a reference to blkg under an rcu allows acess to only
+ * values local to groups like group stats and group rate limits
+ */
+ call_rcu(&tg->rcu_head, throtl_free_tg);
}
static void throtl_init_group(struct throtl_grp *tg)
td->nr_undestroyed_grps++;
}
-static void throtl_init_add_tg_lists(struct throtl_data *td,
- struct throtl_grp *tg, struct blkio_cgroup *blkcg)
+static void
+__throtl_tg_fill_dev_details(struct throtl_data *td, struct throtl_grp *tg)
{
struct backing_dev_info *bdi = &td->queue->backing_dev_info;
unsigned int major, minor;
+ if (!tg || tg->blkg.dev)
+ return;
+
+ /*
+ * Fill in device details for a group which might not have been
+ * filled at group creation time as queue was being instantiated
+ * and driver had not attached a device yet
+ */
+ if (bdi->dev && dev_name(bdi->dev)) {
+ sscanf(dev_name(bdi->dev), "%u:%u", &major, &minor);
+ tg->blkg.dev = MKDEV(major, minor);
+ }
+}
+
+/*
+ * Should be called with without queue lock held. Here queue lock will be
+ * taken rarely. It will be taken only once during life time of a group
+ * if need be
+ */
+static void
+throtl_tg_fill_dev_details(struct throtl_data *td, struct throtl_grp *tg)
+{
+ if (!tg || tg->blkg.dev)
+ return;
+
+ spin_lock_irq(td->queue->queue_lock);
+ __throtl_tg_fill_dev_details(td, tg);
+ spin_unlock_irq(td->queue->queue_lock);
+}
+
+static void throtl_init_add_tg_lists(struct throtl_data *td,
+ struct throtl_grp *tg, struct blkio_cgroup *blkcg)
+{
+ __throtl_tg_fill_dev_details(td, tg);
+
/* Add group onto cgroup list */
- sscanf(dev_name(bdi->dev), "%u:%u", &major, &minor);
blkiocg_add_blkio_group(blkcg, &tg->blkg, (void *)td,
- MKDEV(major, minor), BLKIO_POLICY_THROTL);
+ tg->blkg.dev, BLKIO_POLICY_THROTL);
tg->bps[READ] = blkcg_get_read_bps(blkcg, tg->blkg.dev);
tg->bps[WRITE] = blkcg_get_write_bps(blkcg, tg->blkg.dev);
static struct throtl_grp *throtl_alloc_tg(struct throtl_data *td)
{
struct throtl_grp *tg = NULL;
+ int ret;
tg = kzalloc_node(sizeof(*tg), GFP_ATOMIC, td->queue->node);
if (!tg)
return NULL;
+ ret = blkio_alloc_blkg_stats(&tg->blkg);
+
+ if (ret) {
+ kfree(tg);
+ return NULL;
+ }
+
throtl_init_group(tg);
return tg;
}
{
struct throtl_grp *tg = NULL;
void *key = td;
- struct backing_dev_info *bdi = &td->queue->backing_dev_info;
- unsigned int major, minor;
/*
* This is the common case when there are no blkio cgroups.
* Avoid lookup in this case
*/
if (blkcg == &blkio_root_cgroup)
- tg = &td->root_tg;
+ tg = td->root_tg;
else
tg = tg_of_blkg(blkiocg_lookup_group(blkcg, key));
- /* Fill in device details for root group */
- if (tg && !tg->blkg.dev && bdi->dev && dev_name(bdi->dev)) {
- sscanf(dev_name(bdi->dev), "%u:%u", &major, &minor);
- tg->blkg.dev = MKDEV(major, minor);
- }
-
+ __throtl_tg_fill_dev_details(td, tg);
return tg;
}
-/*
- * This function returns with queue lock unlocked in case of error, like
- * request queue is no more
- */
static struct throtl_grp * throtl_get_tg(struct throtl_data *td)
{
struct throtl_grp *tg = NULL, *__tg = NULL;
struct blkio_cgroup *blkcg;
struct request_queue *q = td->queue;
+ /* no throttling for dead queue */
+ if (unlikely(test_bit(QUEUE_FLAG_DEAD, &q->queue_flags)))
+ return NULL;
+
rcu_read_lock();
blkcg = task_blkio_cgroup(current);
tg = throtl_find_tg(td, blkcg);
/*
* Need to allocate a group. Allocation of group also needs allocation
* of per cpu stats which in-turn takes a mutex() and can block. Hence
- * we need to drop rcu lock and queue_lock before we call alloc
- *
- * Take the request queue reference to make sure queue does not
- * go away once we return from allocation.
+ * we need to drop rcu lock and queue_lock before we call alloc.
*/
- blk_get_queue(q);
rcu_read_unlock();
spin_unlock_irq(q->queue_lock);
tg = throtl_alloc_tg(td);
- /*
- * We might have slept in group allocation. Make sure queue is not
- * dead
- */
- if (unlikely(test_bit(QUEUE_FLAG_DEAD, &q->queue_flags))) {
- blk_put_queue(q);
- if (tg)
- kfree(tg);
-
- return ERR_PTR(-ENODEV);
- }
- blk_put_queue(q);
/* Group allocated and queue is still alive. take the lock */
spin_lock_irq(q->queue_lock);
+ /* Make sure @q is still alive */
+ if (unlikely(test_bit(QUEUE_FLAG_DEAD, &q->queue_flags))) {
+ kfree(tg);
+ return NULL;
+ }
+
/*
* Initialize the new group. After sleeping, read the blkcg again.
*/
/* Group allocation failed. Account the IO to root group */
if (!tg) {
- tg = &td->root_tg;
+ tg = td->root_tg;
return tg;
}
return 0;
}
+static bool tg_no_rule_group(struct throtl_grp *tg, bool rw) {
+ if (tg->bps[rw] == -1 && tg->iops[rw] == -1)
+ return 1;
+ return 0;
+}
+
/*
* Returns whether one can dispatch a bio or not. Also returns approx number
* of jiffies to wait before this bio is with-in IO rate and can be dispatched
static void throtl_charge_bio(struct throtl_grp *tg, struct bio *bio)
{
bool rw = bio_data_dir(bio);
- bool sync = bio->bi_rw & REQ_SYNC;
+ bool sync = rw_is_sync(bio->bi_rw);
/* Charge the bio to the group */
tg->bytes_disp[rw] += bio->bi_size;
tg->io_disp[rw]++;
- /*
- * TODO: This will take blkg->stats_lock. Figure out a way
- * to avoid this cost.
- */
blkiocg_update_dispatch_stats(&tg->blkg, bio->bi_size, rw, sync);
}
bio_list_init(&bio_list_on_stack);
- throtl_log(td, "dispatch nr_queued=%lu read=%u write=%u",
+ throtl_log(td, "dispatch nr_queued=%u read=%u write=%u",
total_nr_queued(td), td->nr_queued[READ],
td->nr_queued[WRITE]);
struct delayed_work *dwork = &td->throtl_work;
/* schedule work if limits changed even if no bio is queued */
- if (total_nr_queued(td) > 0 || td->limits_changed) {
+ if (total_nr_queued(td) || td->limits_changed) {
/*
* We might have a work scheduled to be executed in future.
* Cancel that and schedule a new one.
}
}
-static void throtl_td_free(struct throtl_data *td)
-{
- kfree(td);
-}
-
/*
* Blk cgroup controller notification saying that blkio_group object is being
* delinked as associated cgroup object is going away. That also means that
.plid = BLKIO_POLICY_THROTL,
};
-int blk_throtl_bio(struct request_queue *q, struct bio **biop)
+bool blk_throtl_bio(struct request_queue *q, struct bio *bio)
{
struct throtl_data *td = q->td;
struct throtl_grp *tg;
- struct bio *bio = *biop;
bool rw = bio_data_dir(bio), update_disptime = true;
+ struct blkio_cgroup *blkcg;
+ bool throttled = false;
if (bio->bi_rw & REQ_THROTTLED) {
bio->bi_rw &= ~REQ_THROTTLED;
- return 0;
+ goto out;
}
- spin_lock_irq(q->queue_lock);
- tg = throtl_get_tg(td);
+ /*
+ * A throtl_grp pointer retrieved under rcu can be used to access
+ * basic fields like stats and io rates. If a group has no rules,
+ * just update the dispatch stats in lockless manner and return.
+ */
- if (IS_ERR(tg)) {
- if (PTR_ERR(tg) == -ENODEV) {
- /*
- * Queue is gone. No queue lock held here.
- */
- return -ENODEV;
+ rcu_read_lock();
+ blkcg = task_blkio_cgroup(current);
+ tg = throtl_find_tg(td, blkcg);
+ if (tg) {
+ throtl_tg_fill_dev_details(td, tg);
+
+ if (tg_no_rule_group(tg, rw)) {
+ blkiocg_update_dispatch_stats(&tg->blkg, bio->bi_size,
+ rw, rw_is_sync(bio->bi_rw));
+ rcu_read_unlock();
+ goto out;
}
}
+ rcu_read_unlock();
+
+ /*
+ * Either group has not been allocated yet or it is not an unlimited
+ * IO group
+ */
+ spin_lock_irq(q->queue_lock);
+ tg = throtl_get_tg(td);
+ if (unlikely(!tg))
+ goto out_unlock;
if (tg->nr_queued[rw]) {
/*
* So keep on trimming slice even if bio is not queued.
*/
throtl_trim_slice(td, tg, rw);
- goto out;
+ goto out_unlock;
}
queue_bio:
- throtl_log_tg(td, tg, "[%c] bio. bdisp=%u sz=%u bps=%llu"
+ throtl_log_tg(td, tg, "[%c] bio. bdisp=%llu sz=%u bps=%llu"
" iodisp=%u iops=%u queued=%d/%d",
rw == READ ? 'R' : 'W',
tg->bytes_disp[rw], bio->bi_size, tg->bps[rw],
tg->nr_queued[READ], tg->nr_queued[WRITE]);
throtl_add_bio_tg(q->td, tg, bio);
- *biop = NULL;
+ throttled = true;
if (update_disptime) {
tg_update_disptime(td, tg);
throtl_schedule_next_dispatch(td);
}
+out_unlock:
+ spin_unlock_irq(q->queue_lock);
out:
+ return throttled;
+}
+
+/**
+ * blk_throtl_drain - drain throttled bios
+ * @q: request_queue to drain throttled bios for
+ *
+ * Dispatch all currently throttled bios on @q through ->make_request_fn().
+ */
+void blk_throtl_drain(struct request_queue *q)
+ __releases(q->queue_lock) __acquires(q->queue_lock)
+{
+ struct throtl_data *td = q->td;
+ struct throtl_rb_root *st = &td->tg_service_tree;
+ struct throtl_grp *tg;
+ struct bio_list bl;
+ struct bio *bio;
+
+ WARN_ON_ONCE(!queue_is_locked(q));
+
+ bio_list_init(&bl);
+
+ while ((tg = throtl_rb_first(st))) {
+ throtl_dequeue_tg(td, tg);
+
+ while ((bio = bio_list_peek(&tg->bio_lists[READ])))
+ tg_dispatch_one_bio(td, tg, bio_data_dir(bio), &bl);
+ while ((bio = bio_list_peek(&tg->bio_lists[WRITE])))
+ tg_dispatch_one_bio(td, tg, bio_data_dir(bio), &bl);
+ }
spin_unlock_irq(q->queue_lock);
- return 0;
+
+ while ((bio = bio_list_pop(&bl)))
+ generic_make_request(bio);
+
+ spin_lock_irq(q->queue_lock);
}
int blk_throtl_init(struct request_queue *q)
td->limits_changed = false;
INIT_DELAYED_WORK(&td->throtl_work, blk_throtl_work);
- /* Init root group */
- tg = &td->root_tg;
- throtl_init_group(tg);
+ /* alloc and Init root group. */
+ td->queue = q;
+ tg = throtl_alloc_tg(td);
- /*
- * Set root group reference to 2. One reference will be dropped when
- * all groups on tg_list are being deleted during queue exit. Other
- * reference will remain there as we don't want to delete this group
- * as it is statically allocated and gets destroyed when throtl_data
- * goes away.
- */
- atomic_inc(&tg->ref);
+ if (!tg) {
+ kfree(td);
+ return -ENOMEM;
+ }
+
+ td->root_tg = tg;
rcu_read_lock();
- blkiocg_add_blkio_group(&blkio_root_cgroup, &tg->blkg, (void *)td,
- 0, BLKIO_POLICY_THROTL);
+ throtl_init_add_tg_lists(td, tg, &blkio_root_cgroup);
rcu_read_unlock();
- throtl_add_group_to_td_list(td, tg);
/* Attach throtl data to request queue */
- td->queue = q;
q->td = td;
return 0;
}
* it.
*/
throtl_shutdown_wq(q);
- throtl_td_free(td);
+}
+
+void blk_throtl_release(struct request_queue *q)
+{
+ kfree(q->td);
}
static int __init throtl_init(void)