}
EXPORT_SYMBOL(get_io_context);
-static void cfq_dtor(struct io_context *ioc)
+/*
+ * Releasing ioc may nest into another put_io_context() leading to nested
+ * fast path release. As the ioc's can't be the same, this is okay but
+ * makes lockdep whine. Keep track of nesting and use it as subclass.
+ */
+#ifdef CONFIG_LOCKDEP
+#define ioc_release_depth(q) ((q) ? (q)->ioc_release_depth : 0)
+#define ioc_release_depth_inc(q) (q)->ioc_release_depth++
+#define ioc_release_depth_dec(q) (q)->ioc_release_depth--
+#else
+#define ioc_release_depth(q) 0
+#define ioc_release_depth_inc(q) do { } while (0)
+#define ioc_release_depth_dec(q) do { } while (0)
+#endif
+
+/*
+ * Slow path for ioc release in put_io_context(). Performs double-lock
+ * dancing to unlink all cic's and then frees ioc.
+ */
+static void ioc_release_fn(struct work_struct *work)
{
- if (!hlist_empty(&ioc->cic_list)) {
- struct cfq_io_context *cic;
+ struct io_context *ioc = container_of(work, struct io_context,
+ release_work);
+ struct request_queue *last_q = NULL;
+
+ spin_lock_irq(&ioc->lock);
+
+ while (!hlist_empty(&ioc->cic_list)) {
+ struct cfq_io_context *cic = hlist_entry(ioc->cic_list.first,
+ struct cfq_io_context,
+ cic_list);
+ struct request_queue *this_q = cic->q;
+
+ if (this_q != last_q) {
+ /*
+ * Need to switch to @this_q. Once we release
+ * @ioc->lock, it can go away along with @cic.
+ * Hold on to it.
+ */
+ __blk_get_queue(this_q);
+
+ /*
+ * blk_put_queue() might sleep thanks to kobject
+ * idiocy. Always release both locks, put and
+ * restart.
+ */
+ if (last_q) {
+ spin_unlock(last_q->queue_lock);
+ spin_unlock_irq(&ioc->lock);
+ blk_put_queue(last_q);
+ } else {
+ spin_unlock_irq(&ioc->lock);
+ }
+
+ last_q = this_q;
+ spin_lock_irq(this_q->queue_lock);
+ spin_lock(&ioc->lock);
+ continue;
+ }
+ ioc_release_depth_inc(this_q);
+ cic->exit(cic);
+ cic->release(cic);
+ ioc_release_depth_dec(this_q);
+ }
- cic = hlist_entry(ioc->cic_list.first, struct cfq_io_context,
- cic_list);
- cic->dtor(ioc);
+ if (last_q) {
+ spin_unlock(last_q->queue_lock);
+ spin_unlock_irq(&ioc->lock);
+ blk_put_queue(last_q);
+ } else {
+ spin_unlock_irq(&ioc->lock);
}
+
+ kmem_cache_free(iocontext_cachep, ioc);
}
/**
* put_io_context - put a reference of io_context
* @ioc: io_context to put
+ * @locked_q: request_queue the caller is holding queue_lock of (hint)
*
* Decrement reference count of @ioc and release it if the count reaches
- * zero.
+ * zero. If the caller is holding queue_lock of a queue, it can indicate
+ * that with @locked_q. This is an optimization hint and the caller is
+ * allowed to pass in %NULL even when it's holding a queue_lock.
*/
-void put_io_context(struct io_context *ioc)
+void put_io_context(struct io_context *ioc, struct request_queue *locked_q)
{
+ struct request_queue *last_q = locked_q;
+ unsigned long flags;
+
if (ioc == NULL)
return;
BUG_ON(atomic_long_read(&ioc->refcount) <= 0);
+ if (locked_q)
+ lockdep_assert_held(locked_q->queue_lock);
if (!atomic_long_dec_and_test(&ioc->refcount))
return;
- rcu_read_lock();
- cfq_dtor(ioc);
- rcu_read_unlock();
-
- kmem_cache_free(iocontext_cachep, ioc);
-}
-EXPORT_SYMBOL(put_io_context);
+ /*
+ * Destroy @ioc. This is a bit messy because cic's are chained
+ * from both ioc and queue, and ioc->lock nests inside queue_lock.
+ * The inner ioc->lock should be held to walk our cic_list and then
+ * for each cic the outer matching queue_lock should be grabbed.
+ * ie. We need to do reverse-order double lock dancing.
+ *
+ * Another twist is that we are often called with one of the
+ * matching queue_locks held as indicated by @locked_q, which
+ * prevents performing double-lock dance for other queues.
+ *
+ * So, we do it in two stages. The fast path uses the queue_lock
+ * the caller is holding and, if other queues need to be accessed,
+ * uses trylock to avoid introducing locking dependency. This can
+ * handle most cases, especially if @ioc was performing IO on only
+ * single device.
+ *
+ * If trylock doesn't cut it, we defer to @ioc->release_work which
+ * can do all the double-locking dancing.
+ */
+ spin_lock_irqsave_nested(&ioc->lock, flags,
+ ioc_release_depth(locked_q));
+
+ while (!hlist_empty(&ioc->cic_list)) {
+ struct cfq_io_context *cic = hlist_entry(ioc->cic_list.first,
+ struct cfq_io_context,
+ cic_list);
+ struct request_queue *this_q = cic->q;
+
+ if (this_q != last_q) {
+ if (last_q && last_q != locked_q)
+ spin_unlock(last_q->queue_lock);
+ last_q = NULL;
+
+ if (!spin_trylock(this_q->queue_lock))
+ break;
+ last_q = this_q;
+ continue;
+ }
+ ioc_release_depth_inc(this_q);
+ cic->exit(cic);
+ cic->release(cic);
+ ioc_release_depth_dec(this_q);
+ }
-static void cfq_exit(struct io_context *ioc)
-{
- rcu_read_lock();
+ if (last_q && last_q != locked_q)
+ spin_unlock(last_q->queue_lock);
- if (!hlist_empty(&ioc->cic_list)) {
- struct cfq_io_context *cic;
+ spin_unlock_irqrestore(&ioc->lock, flags);
- cic = hlist_entry(ioc->cic_list.first, struct cfq_io_context,
- cic_list);
- cic->exit(ioc);
- }
- rcu_read_unlock();
+ /* if no cic's left, we're done; otherwise, kick release_work */
+ if (hlist_empty(&ioc->cic_list))
+ kmem_cache_free(iocontext_cachep, ioc);
+ else
+ schedule_work(&ioc->release_work);
}
+EXPORT_SYMBOL(put_io_context);
/* Called by the exiting task */
void exit_io_context(struct task_struct *task)
task->io_context = NULL;
task_unlock(task);
- if (atomic_dec_and_test(&ioc->nr_tasks))
- cfq_exit(ioc);
-
- put_io_context(ioc);
+ atomic_dec(&ioc->nr_tasks);
+ put_io_context(ioc, NULL);
}
static struct io_context *create_task_io_context(struct task_struct *task,
spin_lock_init(&ioc->lock);
INIT_RADIX_TREE(&ioc->radix_root, GFP_ATOMIC | __GFP_HIGH);
INIT_HLIST_HEAD(&ioc->cic_list);
+ INIT_WORK(&ioc->release_work, ioc_release_fn);
/* try to install, somebody might already have beaten us to it */
task_lock(task);
cfqd->active_queue = NULL;
if (cfqd->active_cic) {
- put_io_context(cfqd->active_cic->ioc);
+ put_io_context(cfqd->active_cic->ioc, cfqd->queue);
cfqd->active_cic = NULL;
}
}
}
}
-static void cfq_exit_single_io_context(struct io_context *ioc,
- struct cfq_io_context *cic)
-{
- struct cfq_data *cfqd = cic_to_cfqd(cic);
-
- if (cfqd) {
- struct request_queue *q = cfqd->queue;
- unsigned long flags;
-
- spin_lock_irqsave(q->queue_lock, flags);
-
- /*
- * Ensure we get a fresh copy of the ->key to prevent
- * race between exiting task and queue
- */
- smp_read_barrier_depends();
- if (cic->key == cfqd)
- cfq_exit_cic(cic);
-
- spin_unlock_irqrestore(q->queue_lock, flags);
- }
-}
-
-/*
- * The process that ioc belongs to has exited, we need to clean up
- * and put the internal structures we have that belongs to that process.
- */
-static void cfq_exit_io_context(struct io_context *ioc)
-{
- call_for_each_cic(ioc, cfq_exit_single_io_context);
-}
-
static struct cfq_io_context *
cfq_alloc_io_context(struct cfq_data *cfqd, gfp_t gfp_mask)
{
cic->ttime.last_end_request = jiffies;
INIT_LIST_HEAD(&cic->queue_list);
INIT_HLIST_NODE(&cic->cic_list);
- cic->dtor = cfq_free_io_context;
- cic->exit = cfq_exit_io_context;
+ cic->exit = cfq_exit_cic;
+ cic->release = cfq_release_cic;
elv_ioc_count_inc(cfq_ioc_count);
}
BUG_ON(!cfqq->allocated[rw]);
cfqq->allocated[rw]--;
- put_io_context(RQ_CIC(rq)->ioc);
+ put_io_context(RQ_CIC(rq)->ioc, cfqq->cfqd->queue);
rq->elevator_private[0] = NULL;
rq->elevator_private[1] = NULL;
struct cfq_io_context *cic = list_entry(cfqd->cic_list.next,
struct cfq_io_context,
queue_list);
+ struct io_context *ioc = cic->ioc;
+ spin_lock(&ioc->lock);
cfq_exit_cic(cic);
+ cfq_release_cic(cic);
+ spin_unlock(&ioc->lock);
}
cfq_put_async_queues(cfqd);
#include <linux/radix-tree.h>
#include <linux/rcupdate.h>
+#include <linux/workqueue.h>
struct cfq_queue;
struct cfq_ttime {
unsigned long changed;
- void (*dtor)(struct io_context *); /* destructor */
- void (*exit)(struct io_context *); /* called on task exit */
+ void (*exit)(struct cfq_io_context *);
+ void (*release)(struct cfq_io_context *);
struct rcu_head rcu_head;
};
struct radix_tree_root radix_root;
struct hlist_head cic_list;
void __rcu *ioc_data;
+
+ struct work_struct release_work;
};
static inline struct io_context *ioc_task_link(struct io_context *ioc)
struct task_struct;
#ifdef CONFIG_BLOCK
-void put_io_context(struct io_context *ioc);
+void put_io_context(struct io_context *ioc, struct request_queue *locked_q);
void exit_io_context(struct task_struct *task);
struct io_context *get_task_io_context(struct task_struct *task,
gfp_t gfp_flags, int node);
void ioc_cgroup_changed(struct io_context *ioc);
#else
struct io_context;
-static inline void put_io_context(struct io_context *ioc) { }
+static inline void put_io_context(struct io_context *ioc,
+ struct request_queue *locked_q) { }
static inline void exit_io_context(struct task_struct *task) { }
#endif