[linux.git] / block / blk-ioc.c

// SPDX-License-Identifier: GPL-2.0
/*
 * Functions related to io context handling
 */
#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/init.h>
#include <linux/bio.h>
#include <linux/blkdev.h>
#include <linux/slab.h>
#include <linux/security.h>
#include <linux/sched/task.h>

#include "blk.h"
#include "blk-mq-sched.h"

/*
 * For io context allocations
 */
static struct kmem_cache *iocontext_cachep;

#ifdef CONFIG_BLK_ICQ
/**
 * get_io_context - increment reference count to io_context
 * @ioc: io_context to get
 *
 * Increment reference count to @ioc.
 */
static void get_io_context(struct io_context *ioc)
{
	BUG_ON(atomic_long_read(&ioc->refcount) <= 0);
	atomic_long_inc(&ioc->refcount);
}

/*
 * Exit an icq. Called with ioc locked for blk-mq, and with both ioc
 * and queue locked for legacy.
 */
static void ioc_exit_icq(struct io_cq *icq)
{
	struct elevator_type *et = icq->q->elevator->type;

	if (icq->flags & ICQ_EXITED)
		return;

	if (et->ops.exit_icq)
		et->ops.exit_icq(icq);

	icq->flags |= ICQ_EXITED;
}

static void ioc_exit_icqs(struct io_context *ioc)
{
	struct io_cq *icq;

	spin_lock_irq(&ioc->lock);
	hlist_for_each_entry(icq, &ioc->icq_list, ioc_node)
		ioc_exit_icq(icq);
	spin_unlock_irq(&ioc->lock);
}

/*
 * Release an icq. Called with ioc locked for blk-mq, and with both ioc
 * and queue locked for legacy.
 */
static void ioc_destroy_icq(struct io_cq *icq)
{
	struct io_context *ioc = icq->ioc;
	struct request_queue *q = icq->q;
	struct elevator_type *et = q->elevator->type;

	lockdep_assert_held(&ioc->lock);
	lockdep_assert_held(&q->queue_lock);

	if (icq->flags & ICQ_DESTROYED)
		return;

	radix_tree_delete(&ioc->icq_tree, icq->q->id);
	hlist_del_init(&icq->ioc_node);
	list_del_init(&icq->q_node);

	/*
	 * Both setting lookup hint to and clearing it from @icq are done
	 * under queue_lock.  If it's not pointing to @icq now, it never
	 * will.  Hint assignment itself can race safely.
	 */
	if (rcu_access_pointer(ioc->icq_hint) == icq)
		rcu_assign_pointer(ioc->icq_hint, NULL);

	ioc_exit_icq(icq);

	/*
	 * @icq->q might have gone away by the time RCU callback runs
	 * making it impossible to determine icq_cache.  Record it in @icq.
	 */
	icq->__rcu_icq_cache = et->icq_cache;
	icq->flags |= ICQ_DESTROYED;
	kfree_rcu(icq, __rcu_head);
}

/*
 * Slow path for ioc release in put_io_context().  Performs double-lock
 * dancing to unlink all icq's and then frees ioc.
 */
static void ioc_release_fn(struct work_struct *work)
{
	struct io_context *ioc = container_of(work, struct io_context,
					      release_work);
	spin_lock_irq(&ioc->lock);

	while (!hlist_empty(&ioc->icq_list)) {
		struct io_cq *icq = hlist_entry(ioc->icq_list.first,
						struct io_cq, ioc_node);
		struct request_queue *q = icq->q;

		if (spin_trylock(&q->queue_lock)) {
			ioc_destroy_icq(icq);
			spin_unlock(&q->queue_lock);
		} else {
			/* Make sure q and icq cannot be freed. */
			rcu_read_lock();

			/* Re-acquire the locks in the correct order. */
			spin_unlock(&ioc->lock);
			spin_lock(&q->queue_lock);
			spin_lock(&ioc->lock);

			ioc_destroy_icq(icq);

			spin_unlock(&q->queue_lock);
			rcu_read_unlock();
		}
	}

	spin_unlock_irq(&ioc->lock);

	kmem_cache_free(iocontext_cachep, ioc);
}

/*
 * Releasing icqs requires reverse order double locking and we may already be
 * holding a queue_lock.  Do it asynchronously from a workqueue.
 */
static bool ioc_delay_free(struct io_context *ioc)
{
	unsigned long flags;

	spin_lock_irqsave(&ioc->lock, flags);
	if (!hlist_empty(&ioc->icq_list)) {
		queue_work(system_power_efficient_wq, &ioc->release_work);
		spin_unlock_irqrestore(&ioc->lock, flags);
		return true;
	}
	spin_unlock_irqrestore(&ioc->lock, flags);
	return false;
}

/**
 * ioc_clear_queue - break any ioc association with the specified queue
 * @q: request_queue being cleared
 *
 * Walk @q->icq_list and exit all io_cq's.
 */
void ioc_clear_queue(struct request_queue *q)
{
	spin_lock_irq(&q->queue_lock);
	while (!list_empty(&q->icq_list)) {
		struct io_cq *icq =
			list_first_entry(&q->icq_list, struct io_cq, q_node);

		/*
		 * Other context won't hold ioc lock to wait for queue_lock, see
		 * details in ioc_release_fn().
		 */
		spin_lock(&icq->ioc->lock);
		ioc_destroy_icq(icq);
		spin_unlock(&icq->ioc->lock);
	}
	spin_unlock_irq(&q->queue_lock);
}
#else /* CONFIG_BLK_ICQ */
static inline void ioc_exit_icqs(struct io_context *ioc)
{
}
static inline bool ioc_delay_free(struct io_context *ioc)
{
	return false;
}
#endif /* CONFIG_BLK_ICQ */

/**
 * put_io_context - put a reference of io_context
 * @ioc: io_context to put
 *
 * Decrement reference count of @ioc and release it if the count reaches
 * zero.
 */
void put_io_context(struct io_context *ioc)
{
	BUG_ON(atomic_long_read(&ioc->refcount) <= 0);
	if (atomic_long_dec_and_test(&ioc->refcount) && !ioc_delay_free(ioc))
		kmem_cache_free(iocontext_cachep, ioc);
}
EXPORT_SYMBOL_GPL(put_io_context);

/* Called by the exiting task */
void exit_io_context(struct task_struct *task)
{
	struct io_context *ioc;

	task_lock(task);
	ioc = task->io_context;
	task->io_context = NULL;
	task_unlock(task);

	if (atomic_dec_and_test(&ioc->active_ref)) {
		ioc_exit_icqs(ioc);
		put_io_context(ioc);
	}
}

static struct io_context *alloc_io_context(gfp_t gfp_flags, int node)
{
	struct io_context *ioc;

	ioc = kmem_cache_alloc_node(iocontext_cachep, gfp_flags | __GFP_ZERO,
				    node);
	if (unlikely(!ioc))
		return NULL;

	atomic_long_set(&ioc->refcount, 1);
	atomic_set(&ioc->active_ref, 1);
#ifdef CONFIG_BLK_ICQ
	spin_lock_init(&ioc->lock);
	INIT_RADIX_TREE(&ioc->icq_tree, GFP_ATOMIC);
	INIT_HLIST_HEAD(&ioc->icq_list);
	INIT_WORK(&ioc->release_work, ioc_release_fn);
#endif
	ioc->ioprio = IOPRIO_DEFAULT;

	return ioc;
}

int set_task_ioprio(struct task_struct *task, int ioprio)
{
	int err;
	const struct cred *cred = current_cred(), *tcred;

	rcu_read_lock();
	tcred = __task_cred(task);
	if (!uid_eq(tcred->uid, cred->euid) &&
	    !uid_eq(tcred->uid, cred->uid) && !capable(CAP_SYS_NICE)) {
		rcu_read_unlock();
		return -EPERM;
	}
	rcu_read_unlock();

	err = security_task_setioprio(task, ioprio);
	if (err)
		return err;

	task_lock(task);
	if (unlikely(!task->io_context)) {
		struct io_context *ioc;

		task_unlock(task);

		ioc = alloc_io_context(GFP_ATOMIC, NUMA_NO_NODE);
		if (!ioc)
			return -ENOMEM;

		task_lock(task);
		if (task->flags & PF_EXITING) {
			kmem_cache_free(iocontext_cachep, ioc);
			goto out;
		}
		if (task->io_context)
			kmem_cache_free(iocontext_cachep, ioc);
		else
			task->io_context = ioc;
	}
	task->io_context->ioprio = ioprio;
out:
	task_unlock(task);
	return 0;
}
EXPORT_SYMBOL_GPL(set_task_ioprio);

int __copy_io(unsigned long clone_flags, struct task_struct *tsk)
{
	struct io_context *ioc = current->io_context;

	/*
	 * Share io context with parent, if CLONE_IO is set
	 */
	if (clone_flags & CLONE_IO) {
		atomic_inc(&ioc->active_ref);
		tsk->io_context = ioc;
	} else if (ioprio_valid(ioc->ioprio)) {
		tsk->io_context = alloc_io_context(GFP_KERNEL, NUMA_NO_NODE);
		if (!tsk->io_context)
			return -ENOMEM;
		tsk->io_context->ioprio = ioc->ioprio;
	}

	return 0;
}

#ifdef CONFIG_BLK_ICQ
/**
 * ioc_lookup_icq - lookup io_cq from ioc
 * @q: the associated request_queue
 *
 * Look up io_cq associated with @ioc - @q pair from @ioc.  Must be called
 * with @q->queue_lock held.
 */
struct io_cq *ioc_lookup_icq(struct request_queue *q)
{
	struct io_context *ioc = current->io_context;
	struct io_cq *icq;

	lockdep_assert_held(&q->queue_lock);

	/*
	 * icq's are indexed from @ioc using radix tree and hint pointer,
	 * both of which are protected with RCU.  All removals are done
	 * holding both q and ioc locks, and we're holding q lock - if we
	 * find a icq which points to us, it's guaranteed to be valid.
	 */
	rcu_read_lock();
	icq = rcu_dereference(ioc->icq_hint);
	if (icq && icq->q == q)
		goto out;

	icq = radix_tree_lookup(&ioc->icq_tree, q->id);
	if (icq && icq->q == q)
		rcu_assign_pointer(ioc->icq_hint, icq);	/* allowed to race */
	else
		icq = NULL;
out:
	rcu_read_unlock();
	return icq;
}
EXPORT_SYMBOL(ioc_lookup_icq);

/**
 * ioc_create_icq - create and link io_cq
 * @q: request_queue of interest
 *
 * Make sure io_cq linking @ioc and @q exists.  If icq doesn't exist, they
 * will be created using @gfp_mask.
 *
 * The caller is responsible for ensuring @ioc won't go away and @q is
 * alive and will stay alive until this function returns.
 */
static struct io_cq *ioc_create_icq(struct request_queue *q)
{
	struct io_context *ioc = current->io_context;
	struct elevator_type *et = q->elevator->type;
	struct io_cq *icq;

	/* allocate stuff */
	icq = kmem_cache_alloc_node(et->icq_cache, GFP_ATOMIC | __GFP_ZERO,
				    q->node);
	if (!icq)
		return NULL;

	if (radix_tree_maybe_preload(GFP_ATOMIC) < 0) {
		kmem_cache_free(et->icq_cache, icq);
		return NULL;
	}

	icq->ioc = ioc;
	icq->q = q;
	INIT_LIST_HEAD(&icq->q_node);
	INIT_HLIST_NODE(&icq->ioc_node);

	/* lock both q and ioc and try to link @icq */
	spin_lock_irq(&q->queue_lock);
	spin_lock(&ioc->lock);

	if (likely(!radix_tree_insert(&ioc->icq_tree, q->id, icq))) {
		hlist_add_head(&icq->ioc_node, &ioc->icq_list);
		list_add(&icq->q_node, &q->icq_list);
		if (et->ops.init_icq)
			et->ops.init_icq(icq);
	} else {
		kmem_cache_free(et->icq_cache, icq);
		icq = ioc_lookup_icq(q);
		if (!icq)
			printk(KERN_ERR "cfq: icq link failed!\n");
	}

	spin_unlock(&ioc->lock);
	spin_unlock_irq(&q->queue_lock);
	radix_tree_preload_end();
	return icq;
}

struct io_cq *ioc_find_get_icq(struct request_queue *q)
{
	struct io_context *ioc = current->io_context;
	struct io_cq *icq = NULL;

	if (unlikely(!ioc)) {
		ioc = alloc_io_context(GFP_ATOMIC, q->node);
		if (!ioc)
			return NULL;

		task_lock(current);
		if (current->io_context) {
			kmem_cache_free(iocontext_cachep, ioc);
			ioc = current->io_context;
		} else {
			current->io_context = ioc;
		}

		get_io_context(ioc);
		task_unlock(current);
	} else {
		get_io_context(ioc);

		spin_lock_irq(&q->queue_lock);
		icq = ioc_lookup_icq(q);
		spin_unlock_irq(&q->queue_lock);
	}

	if (!icq) {
		icq = ioc_create_icq(q);
		if (!icq) {
			put_io_context(ioc);
			return NULL;
		}
	}
	return icq;
}
EXPORT_SYMBOL_GPL(ioc_find_get_icq);
#endif /* CONFIG_BLK_ICQ */

static int __init blk_ioc_init(void)
{
	iocontext_cachep = kmem_cache_create("blkdev_ioc",
			sizeof(struct io_context), 0, SLAB_PANIC, NULL);
	return 0;
}
subsys_initcall(blk_ioc_init);
Commit	Line	Data
b2441318	1	// SPDX-License-Identifier: GPL-2.0
86db1e29 JA	2	/*
	3	* Functions related to io context handling
	4	*/
	5	#include <linux/kernel.h>
	6	#include <linux/module.h>
	7	#include <linux/init.h>
	8	#include <linux/bio.h>
	9	#include <linux/blkdev.h>
5a0e3ad6	10	#include <linux/slab.h>
a411cd3c	11	#include <linux/security.h>
f719ff9b	12	#include <linux/sched/task.h>
86db1e29 JA	13
86db1e29 JA	14	#include "blk.h"
2aa7745b	15	#include "blk-mq-sched.h"
86db1e29 JA	16
	17	/*
	18	* For io context allocations
	19	*/
	20	static struct kmem_cache *iocontext_cachep;
	21
5ef16305	22	#ifdef CONFIG_BLK_ICQ
6e736be7 TH	23	/**
	24	* get_io_context - increment reference count to io_context
	25	* @ioc: io_context to get
	26	*
	27	* Increment reference count to @ioc.
	28	*/
87dd1d63	29	static void get_io_context(struct io_context *ioc)
6e736be7 TH	30	{
	31	BUG_ON(atomic_long_read(&ioc->refcount) <= 0);
	32	atomic_long_inc(&ioc->refcount);
	33	}
6e736be7	34
3d492c2e	35	/*
7b36a718 JA	36	* Exit an icq. Called with ioc locked for blk-mq, and with both ioc
7b36a718 JA	37	* and queue locked for legacy.
3d492c2e	38	*/
7e5a8794	39	static void ioc_exit_icq(struct io_cq *icq)
621032ad TH	40	{
	41	struct elevator_type *et = icq->q->elevator->type;
	42
	43	if (icq->flags & ICQ_EXITED)
	44	return;
	45
f9cd4bfe JA	46	if (et->ops.exit_icq)
f9cd4bfe JA	47	et->ops.exit_icq(icq);
621032ad TH	48
	49	icq->flags \|= ICQ_EXITED;
	50	}
	51
4be8a2ea CH	52	static void ioc_exit_icqs(struct io_context *ioc)
	53	{
	54	struct io_cq *icq;
	55
	56	spin_lock_irq(&ioc->lock);
	57	hlist_for_each_entry(icq, &ioc->icq_list, ioc_node)
	58	ioc_exit_icq(icq);
	59	spin_unlock_irq(&ioc->lock);
	60	}
	61
7b36a718 JA	62	/*
	63	* Release an icq. Called with ioc locked for blk-mq, and with both ioc
	64	* and queue locked for legacy.
	65	*/
621032ad	66	static void ioc_destroy_icq(struct io_cq *icq)
7e5a8794 TH	67	{
	68	struct io_context *ioc = icq->ioc;
	69	struct request_queue *q = icq->q;
	70	struct elevator_type *et = q->elevator->type;
	71
	72	lockdep_assert_held(&ioc->lock);
5a0ac57c YK	73	lockdep_assert_held(&q->queue_lock);
	74
	75	if (icq->flags & ICQ_DESTROYED)
	76	return;
7e5a8794 TH	77
	78	radix_tree_delete(&ioc->icq_tree, icq->q->id);
	79	hlist_del_init(&icq->ioc_node);
	80	list_del_init(&icq->q_node);
	81
	82	/*
	83	* Both setting lookup hint to and clearing it from @icq are done
	84	* under queue_lock. If it's not pointing to @icq now, it never
	85	* will. Hint assignment itself can race safely.
	86	*/
ec6c676a	87	if (rcu_access_pointer(ioc->icq_hint) == icq)
7e5a8794 TH	88	rcu_assign_pointer(ioc->icq_hint, NULL);
7e5a8794 TH	89
621032ad	90	ioc_exit_icq(icq);
7e5a8794 TH	91
	92	/*
	93	* @icq->q might have gone away by the time RCU callback runs
	94	* making it impossible to determine icq_cache. Record it in @icq.
	95	*/
	96	icq->__rcu_icq_cache = et->icq_cache;
30a2da7b	97	icq->flags \|= ICQ_DESTROYED;
28878733	98	kfree_rcu(icq, __rcu_head);
7e5a8794 TH	99	}
7e5a8794 TH	100
b2efa052 TH	101	/*
b2efa052 TH	102	* Slow path for ioc release in put_io_context(). Performs double-lock
c5869807	103	* dancing to unlink all icq's and then frees ioc.
b2efa052 TH	104	*/
b2efa052 TH	105	static void ioc_release_fn(struct work_struct *work)
86db1e29	106	{
b2efa052 TH	107	struct io_context *ioc = container_of(work, struct io_context,
b2efa052 TH	108	release_work);
a43f085f	109	spin_lock_irq(&ioc->lock);
b2efa052	110
c5869807 TH	111	while (!hlist_empty(&ioc->icq_list)) {
	112	struct io_cq *icq = hlist_entry(ioc->icq_list.first,
	113	struct io_cq, ioc_node);
2274b029 TH	114	struct request_queue *q = icq->q;
2274b029 TH	115
0d945c1f	116	if (spin_trylock(&q->queue_lock)) {
621032ad	117	ioc_destroy_icq(icq);
0d945c1f	118	spin_unlock(&q->queue_lock);
2274b029	119	} else {
ab96bbab JO	120	/* Make sure q and icq cannot be freed. */
	121	rcu_read_lock();
	122
	123	/* Re-acquire the locks in the correct order. */
	124	spin_unlock(&ioc->lock);
	125	spin_lock(&q->queue_lock);
	126	spin_lock(&ioc->lock);
	127
5a0ac57c	128	ioc_destroy_icq(icq);
ab96bbab JO	129
	130	spin_unlock(&q->queue_lock);
	131	rcu_read_unlock();
b2efa052	132	}
b2efa052	133	}
ffc4e759	134
a43f085f	135	spin_unlock_irq(&ioc->lock);
b2efa052 TH	136
b2efa052 TH	137	kmem_cache_free(iocontext_cachep, ioc);
86db1e29 JA	138	}
86db1e29 JA	139
edf70ff5 CH	140	/*
	141	* Releasing icqs requires reverse order double locking and we may already be
	142	* holding a queue_lock. Do it asynchronously from a workqueue.
	143	*/
	144	static bool ioc_delay_free(struct io_context *ioc)
	145	{
	146	unsigned long flags;
	147
	148	spin_lock_irqsave(&ioc->lock, flags);
	149	if (!hlist_empty(&ioc->icq_list)) {
	150	queue_work(system_power_efficient_wq, &ioc->release_work);
	151	spin_unlock_irqrestore(&ioc->lock, flags);
	152	return true;
	153	}
	154	spin_unlock_irqrestore(&ioc->lock, flags);
	155	return false;
	156	}
	157
5ef16305 CH	158	/**
	159	* ioc_clear_queue - break any ioc association with the specified queue
	160	* @q: request_queue being cleared
	161	*
	162	* Walk @q->icq_list and exit all io_cq's.
	163	*/
	164	void ioc_clear_queue(struct request_queue *q)
	165	{
5ef16305	166	spin_lock_irq(&q->queue_lock);
5a0ac57c	167	while (!list_empty(&q->icq_list)) {
5ef16305	168	struct io_cq *icq =
5a0ac57c	169	list_first_entry(&q->icq_list, struct io_cq, q_node);
5ef16305	170
5a0ac57c YK	171	/*
	172	* Other context won't hold ioc lock to wait for queue_lock, see
	173	* details in ioc_release_fn().
	174	*/
a7cfa0af	175	spin_lock(&icq->ioc->lock);
5a0ac57c	176	ioc_destroy_icq(icq);
a7cfa0af	177	spin_unlock(&icq->ioc->lock);
5ef16305	178	}
5a0ac57c	179	spin_unlock_irq(&q->queue_lock);
5ef16305 CH	180	}
	181	#else /* CONFIG_BLK_ICQ */
	182	static inline void ioc_exit_icqs(struct io_context *ioc)
	183	{
	184	}
	185	static inline bool ioc_delay_free(struct io_context *ioc)
	186	{
	187	return false;
	188	}
	189	#endif /* CONFIG_BLK_ICQ */
	190
42ec57a8 TH	191	/**
	192	* put_io_context - put a reference of io_context
	193	* @ioc: io_context to put
	194	*
	195	* Decrement reference count of @ioc and release it if the count reaches
11a3122f	196	* zero.
86db1e29	197	*/
11a3122f	198	void put_io_context(struct io_context *ioc)
86db1e29	199	{
42ec57a8	200	BUG_ON(atomic_long_read(&ioc->refcount) <= 0);
edf70ff5	201	if (atomic_long_dec_and_test(&ioc->refcount) && !ioc_delay_free(ioc))
ff8c1474	202	kmem_cache_free(iocontext_cachep, ioc);
86db1e29	203	}
222ee581	204	EXPORT_SYMBOL_GPL(put_io_context);
86db1e29	205
f6e8d01b TH	206	/* Called by the exiting task */
	207	void exit_io_context(struct task_struct *task)
	208	{
	209	struct io_context *ioc;
	210
	211	task_lock(task);
	212	ioc = task->io_context;
	213	task->io_context = NULL;
	214	task_unlock(task);
	215
4be8a2ea CH	216	if (atomic_dec_and_test(&ioc->active_ref)) {
	217	ioc_exit_icqs(ioc);
	218	put_io_context(ioc);
	219	}
f6e8d01b TH	220	}
f6e8d01b TH	221
a0f14d8b	222	static struct io_context *alloc_io_context(gfp_t gfp_flags, int node)
86db1e29	223	{
df415656	224	struct io_context *ioc;
86db1e29	225
42ec57a8 TH	226	ioc = kmem_cache_alloc_node(iocontext_cachep, gfp_flags \| __GFP_ZERO,
	227	node);
	228	if (unlikely(!ioc))
a0f14d8b	229	return NULL;
42ec57a8	230
42ec57a8	231	atomic_long_set(&ioc->refcount, 1);
f6e8d01b	232	atomic_set(&ioc->active_ref, 1);
5ef16305	233	#ifdef CONFIG_BLK_ICQ
42ec57a8	234	spin_lock_init(&ioc->lock);
c137969b	235	INIT_RADIX_TREE(&ioc->icq_tree, GFP_ATOMIC);
c5869807	236	INIT_HLIST_HEAD(&ioc->icq_list);
b2efa052	237	INIT_WORK(&ioc->release_work, ioc_release_fn);
5ef16305	238	#endif
e589f464 JK	239	ioc->ioprio = IOPRIO_DEFAULT;
e589f464 JK	240
a0f14d8b CH	241	return ioc;
	242	}
	243
a411cd3c CH	244	int set_task_ioprio(struct task_struct *task, int ioprio)
	245	{
	246	int err;
a411cd3c CH	247	const struct cred cred = current_cred(), tcred;
	248
	249	rcu_read_lock();
	250	tcred = __task_cred(task);
	251	if (!uid_eq(tcred->uid, cred->euid) &&
	252	!uid_eq(tcred->uid, cred->uid) && !capable(CAP_SYS_NICE)) {
	253	rcu_read_unlock();
	254	return -EPERM;
	255	}
	256	rcu_read_unlock();
	257
	258	err = security_task_setioprio(task, ioprio);
	259	if (err)
	260	return err;
	261
8472161b CH	262	task_lock(task);
	263	if (unlikely(!task->io_context)) {
	264	struct io_context *ioc;
a411cd3c	265
8472161b	266	task_unlock(task);
5fc11eeb CH	267
	268	ioc = alloc_io_context(GFP_ATOMIC, NUMA_NO_NODE);
	269	if (!ioc)
	270	return -ENOMEM;
	271
	272	task_lock(task);
a957b612	273	if (task->flags & PF_EXITING) {
a957b612 JA	274	kmem_cache_free(iocontext_cachep, ioc);
	275	goto out;
	276	}
669a0646	277	if (task->io_context)
5fc11eeb	278	kmem_cache_free(iocontext_cachep, ioc);
669a0646	279	else
5fc11eeb	280	task->io_context = ioc;
8472161b CH	281	}
8472161b CH	282	task->io_context->ioprio = ioprio;
a957b612	283	out:
8472161b	284	task_unlock(task);
15583a56	285	return 0;
a411cd3c CH	286	}
	287	EXPORT_SYMBOL_GPL(set_task_ioprio);
	288
88c9a2ce CH	289	int __copy_io(unsigned long clone_flags, struct task_struct *tsk)
	290	{
	291	struct io_context *ioc = current->io_context;
88c9a2ce CH	292
	293	/*
	294	* Share io context with parent, if CLONE_IO is set
	295	*/
	296	if (clone_flags & CLONE_IO) {
50569c24	297	atomic_inc(&ioc->active_ref);
88c9a2ce CH	298	tsk->io_context = ioc;
88c9a2ce CH	299	} else if (ioprio_valid(ioc->ioprio)) {
8ffc1368 CH	300	tsk->io_context = alloc_io_context(GFP_KERNEL, NUMA_NO_NODE);
8ffc1368 CH	301	if (!tsk->io_context)
88c9a2ce	302	return -ENOMEM;
8ffc1368	303	tsk->io_context->ioprio = ioc->ioprio;
88c9a2ce CH	304	}
	305
	306	return 0;
	307	}
	308
5ef16305	309	#ifdef CONFIG_BLK_ICQ
47fdd4ca TH	310	/**
47fdd4ca TH	311	* ioc_lookup_icq - lookup io_cq from ioc
47fdd4ca TH	312	* @q: the associated request_queue
	313	*
	314	* Look up io_cq associated with @ioc - @q pair from @ioc. Must be called
	315	* with @q->queue_lock held.
	316	*/
eca5892a	317	struct io_cq ioc_lookup_icq(struct request_queue q)
47fdd4ca	318	{
eca5892a	319	struct io_context *ioc = current->io_context;
47fdd4ca TH	320	struct io_cq *icq;
47fdd4ca TH	321
0d945c1f	322	lockdep_assert_held(&q->queue_lock);
47fdd4ca TH	323
	324	/*
	325	* icq's are indexed from @ioc using radix tree and hint pointer,
	326	* both of which are protected with RCU. All removals are done
	327	* holding both q and ioc locks, and we're holding q lock - if we
	328	* find a icq which points to us, it's guaranteed to be valid.
	329	*/
	330	rcu_read_lock();
	331	icq = rcu_dereference(ioc->icq_hint);
	332	if (icq && icq->q == q)
	333	goto out;
	334
	335	icq = radix_tree_lookup(&ioc->icq_tree, q->id);
	336	if (icq && icq->q == q)
	337	rcu_assign_pointer(ioc->icq_hint, icq); /* allowed to race */
	338	else
	339	icq = NULL;
	340	out:
	341	rcu_read_unlock();
	342	return icq;
	343	}
	344	EXPORT_SYMBOL(ioc_lookup_icq);
	345
f1f8cc94 TH	346	/**
	347	* ioc_create_icq - create and link io_cq
	348	* @q: request_queue of interest
f1f8cc94	349	*
24acfc34 TH	350	* Make sure io_cq linking @ioc and @q exists. If icq doesn't exist, they
24acfc34 TH	351	* will be created using @gfp_mask.
f1f8cc94 TH	352	*
	353	* The caller is responsible for ensuring @ioc won't go away and @q is
	354	* alive and will stay alive until this function returns.
	355	*/
18b74c4d	356	static struct io_cq ioc_create_icq(struct request_queue q)
f1f8cc94	357	{
18b74c4d	358	struct io_context *ioc = current->io_context;
f1f8cc94	359	struct elevator_type *et = q->elevator->type;
f1f8cc94 TH	360	struct io_cq *icq;
	361
	362	/* allocate stuff */
18b74c4d	363	icq = kmem_cache_alloc_node(et->icq_cache, GFP_ATOMIC \| __GFP_ZERO,
f1f8cc94 TH	364	q->node);
	365	if (!icq)
	366	return NULL;
	367
18b74c4d	368	if (radix_tree_maybe_preload(GFP_ATOMIC) < 0) {
f1f8cc94 TH	369	kmem_cache_free(et->icq_cache, icq);
	370	return NULL;
	371	}
	372
	373	icq->ioc = ioc;
	374	icq->q = q;
	375	INIT_LIST_HEAD(&icq->q_node);
	376	INIT_HLIST_NODE(&icq->ioc_node);
	377
	378	/* lock both q and ioc and try to link @icq */
0d945c1f	379	spin_lock_irq(&q->queue_lock);
f1f8cc94 TH	380	spin_lock(&ioc->lock);
	381
	382	if (likely(!radix_tree_insert(&ioc->icq_tree, q->id, icq))) {
	383	hlist_add_head(&icq->ioc_node, &ioc->icq_list);
	384	list_add(&icq->q_node, &q->icq_list);
f9cd4bfe JA	385	if (et->ops.init_icq)
f9cd4bfe JA	386	et->ops.init_icq(icq);
f1f8cc94 TH	387	} else {
f1f8cc94 TH	388	kmem_cache_free(et->icq_cache, icq);
eca5892a	389	icq = ioc_lookup_icq(q);
f1f8cc94 TH	390	if (!icq)
	391	printk(KERN_ERR "cfq: icq link failed!\n");
	392	}
	393
	394	spin_unlock(&ioc->lock);
0d945c1f	395	spin_unlock_irq(&q->queue_lock);
f1f8cc94 TH	396	radix_tree_preload_end();
	397	return icq;
	398	}
	399
87dd1d63 CH	400	struct io_cq ioc_find_get_icq(struct request_queue q)
87dd1d63 CH	401	{
d538ea4c CH	402	struct io_context *ioc = current->io_context;
d538ea4c CH	403	struct io_cq *icq = NULL;
87dd1d63	404
d538ea4c	405	if (unlikely(!ioc)) {
90b627f5	406	ioc = alloc_io_context(GFP_ATOMIC, q->node);
d538ea4c CH	407	if (!ioc)
d538ea4c CH	408	return NULL;
90b627f5 CH	409
	410	task_lock(current);
	411	if (current->io_context) {
	412	kmem_cache_free(iocontext_cachep, ioc);
	413	ioc = current->io_context;
	414	} else {
	415	current->io_context = ioc;
	416	}
	417
	418	get_io_context(ioc);
	419	task_unlock(current);
d538ea4c CH	420	} else {
d538ea4c CH	421	get_io_context(ioc);
87dd1d63	422
d538ea4c	423	spin_lock_irq(&q->queue_lock);
eca5892a	424	icq = ioc_lookup_icq(q);
d538ea4c CH	425	spin_unlock_irq(&q->queue_lock);
d538ea4c CH	426	}
87dd1d63 CH	427
87dd1d63 CH	428	if (!icq) {
18b74c4d	429	icq = ioc_create_icq(q);
d538ea4c CH	430	if (!icq) {
d538ea4c CH	431	put_io_context(ioc);
87dd1d63	432	return NULL;
d538ea4c	433	}
87dd1d63	434	}
87dd1d63 CH	435	return icq;
	436	}
	437	EXPORT_SYMBOL_GPL(ioc_find_get_icq);
5ef16305	438	#endif /* CONFIG_BLK_ICQ */
87dd1d63	439
13341598	440	static int __init blk_ioc_init(void)
86db1e29 JA	441	{
	442	iocontext_cachep = kmem_cache_create("blkdev_ioc",
	443	sizeof(struct io_context), 0, SLAB_PANIC, NULL);
	444	return 0;
	445	}
	446	subsys_initcall(blk_ioc_init);