aio.c 36.3 KB
Newer Older
Linus Torvalds's avatar
Linus Torvalds committed
1
2
3
4
5
6
7
8
9
10
/*
 *	An async IO implementation for Linux
 *	Written by Benjamin LaHaise <bcrl@kvack.org>
 *
 *	Implements an efficient asynchronous io interface.
 *
 *	Copyright 2000, 2001, 2002 Red Hat, Inc.  All Rights Reserved.
 *
 *	See ../COPYING for licensing terms.
 */
Kent Overstreet's avatar
Kent Overstreet committed
11
12
#define pr_fmt(fmt) "%s: " fmt, __func__

Linus Torvalds's avatar
Linus Torvalds committed
13
14
15
16
17
#include <linux/kernel.h>
#include <linux/init.h>
#include <linux/errno.h>
#include <linux/time.h>
#include <linux/aio_abi.h>
18
#include <linux/export.h>
Linus Torvalds's avatar
Linus Torvalds committed
19
#include <linux/syscalls.h>
20
#include <linux/backing-dev.h>
21
#include <linux/uio.h>
Linus Torvalds's avatar
Linus Torvalds committed
22
23
24
25
26
27

#include <linux/sched.h>
#include <linux/fs.h>
#include <linux/file.h>
#include <linux/mm.h>
#include <linux/mman.h>
28
#include <linux/mmu_context.h>
Kent Overstreet's avatar
Kent Overstreet committed
29
#include <linux/percpu.h>
Linus Torvalds's avatar
Linus Torvalds committed
30
31
32
33
34
35
#include <linux/slab.h>
#include <linux/timer.h>
#include <linux/aio.h>
#include <linux/highmem.h>
#include <linux/workqueue.h>
#include <linux/security.h>
36
#include <linux/eventfd.h>
Jeff Moyer's avatar
Jeff Moyer committed
37
#include <linux/blkdev.h>
38
#include <linux/compat.h>
39
40
41
#include <linux/anon_inodes.h>
#include <linux/migrate.h>
#include <linux/ramfs.h>
Kent Overstreet's avatar
Kent Overstreet committed
42
#include <linux/percpu-refcount.h>
Linus Torvalds's avatar
Linus Torvalds committed
43
44
45
46

#include <asm/kmap_types.h>
#include <asm/uaccess.h>

Al Viro's avatar
Al Viro committed
47
48
#include "internal.h"

49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
#define AIO_RING_MAGIC			0xa10a10a1
#define AIO_RING_COMPAT_FEATURES	1
#define AIO_RING_INCOMPAT_FEATURES	0
struct aio_ring {
	unsigned	id;	/* kernel internal index number */
	unsigned	nr;	/* number of io_events */
	unsigned	head;
	unsigned	tail;

	unsigned	magic;
	unsigned	compat_features;
	unsigned	incompat_features;
	unsigned	header_length;	/* size of aio_ring */


	struct io_event		io_events[0];
}; /* 128 bytes + ring size */

#define AIO_RING_PAGES	8

69
70
71
72
73
74
struct kioctx_table {
	struct rcu_head	rcu;
	unsigned	nr;
	struct kioctx	*table[];
};

Kent Overstreet's avatar
Kent Overstreet committed
75
76
77
78
struct kioctx_cpu {
	unsigned		reqs_available;
};

79
struct kioctx {
Kent Overstreet's avatar
Kent Overstreet committed
80
	struct percpu_ref	users;
Kent Overstreet's avatar
Kent Overstreet committed
81
	atomic_t		dead;
82
83
84

	unsigned long		user_id;

Kent Overstreet's avatar
Kent Overstreet committed
85
86
87
88
89
90
91
	struct __percpu kioctx_cpu *cpu;

	/*
	 * For percpu reqs_available, number of slots we move to/from global
	 * counter at a time:
	 */
	unsigned		req_batch;
92
93
94
95
	/*
	 * This is what userspace passed to io_setup(), it's not used for
	 * anything but counting against the global max_reqs quota.
	 *
96
	 * The real limit is nr_events - 1, which will be larger (see
97
98
	 * aio_setup_ring())
	 */
99
100
	unsigned		max_reqs;

101
102
	/* Size of ringbuffer, in units of struct io_event */
	unsigned		nr_events;
103

104
105
106
107
108
109
	unsigned long		mmap_base;
	unsigned long		mmap_size;

	struct page		**ring_pages;
	long			nr_pages;

110
	struct rcu_head		rcu_head;
Kent Overstreet's avatar
Kent Overstreet committed
111
	struct work_struct	free_work;
112
113

	struct {
114
115
116
117
118
		/*
		 * This counts the number of available slots in the ringbuffer,
		 * so we avoid overflowing it: it's decremented (if positive)
		 * when allocating a kiocb and incremented when the resulting
		 * io_event is pulled off the ringbuffer.
Kent Overstreet's avatar
Kent Overstreet committed
119
120
		 *
		 * We batch accesses to it with a percpu version.
121
122
		 */
		atomic_t	reqs_available;
123
124
125
126
127
128
129
	} ____cacheline_aligned_in_smp;

	struct {
		spinlock_t	ctx_lock;
		struct list_head active_reqs;	/* used for cancellation */
	} ____cacheline_aligned_in_smp;

130
131
	struct {
		struct mutex	ring_lock;
132
133
		wait_queue_head_t wait;
	} ____cacheline_aligned_in_smp;
134
135
136
137

	struct {
		unsigned	tail;
		spinlock_t	completion_lock;
138
	} ____cacheline_aligned_in_smp;
139
140

	struct page		*internal_pages[AIO_RING_PAGES];
141
	struct file		*aio_ring_file;
142
143

	unsigned		id;
144
145
};

Linus Torvalds's avatar
Linus Torvalds committed
146
/*------ sysctl variables----*/
147
148
149
static DEFINE_SPINLOCK(aio_nr_lock);
unsigned long aio_nr;		/* current system wide number of aio requests */
unsigned long aio_max_nr = 0x10000; /* system wide maximum number of aio requests */
Linus Torvalds's avatar
Linus Torvalds committed
150
151
/*----end sysctl variables---*/

152
153
static struct kmem_cache	*kiocb_cachep;
static struct kmem_cache	*kioctx_cachep;
Linus Torvalds's avatar
Linus Torvalds committed
154
155
156
157
158
159
160

/* aio_setup
 *	Creates the slab caches used by the aio routines, panic on
 *	failure as this is done early during the boot sequence.
 */
static int __init aio_setup(void)
{
161
162
	kiocb_cachep = KMEM_CACHE(kiocb, SLAB_HWCACHE_ALIGN|SLAB_PANIC);
	kioctx_cachep = KMEM_CACHE(kioctx,SLAB_HWCACHE_ALIGN|SLAB_PANIC);
Linus Torvalds's avatar
Linus Torvalds committed
163

Kent Overstreet's avatar
Kent Overstreet committed
164
	pr_debug("sizeof(struct page) = %zu\n", sizeof(struct page));
Linus Torvalds's avatar
Linus Torvalds committed
165
166
167

	return 0;
}
168
__initcall(aio_setup);
Linus Torvalds's avatar
Linus Torvalds committed
169
170
171

static void aio_free_ring(struct kioctx *ctx)
{
172
173
	int i;
	struct file *aio_ring_file = ctx->aio_ring_file;
Linus Torvalds's avatar
Linus Torvalds committed
174

175
176
177
	for (i = 0; i < ctx->nr_pages; i++) {
		pr_debug("pid(%d) [%d] page->count=%d\n", current->pid, i,
				page_count(ctx->ring_pages[i]));
178
		put_page(ctx->ring_pages[i]);
179
	}
Linus Torvalds's avatar
Linus Torvalds committed
180

181
182
	if (ctx->ring_pages && ctx->ring_pages != ctx->internal_pages)
		kfree(ctx->ring_pages);
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205

	if (aio_ring_file) {
		truncate_setsize(aio_ring_file->f_inode, 0);
		fput(aio_ring_file);
		ctx->aio_ring_file = NULL;
	}
}

static int aio_ring_mmap(struct file *file, struct vm_area_struct *vma)
{
	vma->vm_ops = &generic_file_vm_ops;
	return 0;
}

static const struct file_operations aio_ring_fops = {
	.mmap = aio_ring_mmap,
};

static int aio_set_page_dirty(struct page *page)
{
	return 0;
}

206
#if IS_ENABLED(CONFIG_MIGRATION)
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
static int aio_migratepage(struct address_space *mapping, struct page *new,
			struct page *old, enum migrate_mode mode)
{
	struct kioctx *ctx = mapping->private_data;
	unsigned long flags;
	unsigned idx = old->index;
	int rc;

	/* Writeback must be complete */
	BUG_ON(PageWriteback(old));
	put_page(old);

	rc = migrate_page_move_mapping(mapping, new, old, NULL, mode);
	if (rc != MIGRATEPAGE_SUCCESS) {
		get_page(old);
		return rc;
	}

	get_page(new);

	spin_lock_irqsave(&ctx->completion_lock, flags);
	migrate_page_copy(new, old);
	ctx->ring_pages[idx] = new;
	spin_unlock_irqrestore(&ctx->completion_lock, flags);

	return rc;
Linus Torvalds's avatar
Linus Torvalds committed
233
}
234
#endif
Linus Torvalds's avatar
Linus Torvalds committed
235

236
237
static const struct address_space_operations aio_ctx_aops = {
	.set_page_dirty = aio_set_page_dirty,
238
#if IS_ENABLED(CONFIG_MIGRATION)
239
	.migratepage	= aio_migratepage,
240
#endif
241
242
};

Linus Torvalds's avatar
Linus Torvalds committed
243
244
245
246
static int aio_setup_ring(struct kioctx *ctx)
{
	struct aio_ring *ring;
	unsigned nr_events = ctx->max_reqs;
Zach Brown's avatar
Zach Brown committed
247
	struct mm_struct *mm = current->mm;
248
	unsigned long size, populate;
Linus Torvalds's avatar
Linus Torvalds committed
249
	int nr_pages;
250
251
	int i;
	struct file *file;
Linus Torvalds's avatar
Linus Torvalds committed
252
253
254
255
256
257
258

	/* Compensate for the ring buffer's head/tail overlap entry */
	nr_events += 2;	/* 1 is required, 2 for good luck */

	size = sizeof(struct aio_ring);
	size += sizeof(struct io_event) * nr_events;

259
	nr_pages = PFN_UP(size);
Linus Torvalds's avatar
Linus Torvalds committed
260
261
262
	if (nr_pages < 0)
		return -EINVAL;

263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
	file = anon_inode_getfile_private("[aio]", &aio_ring_fops, ctx, O_RDWR);
	if (IS_ERR(file)) {
		ctx->aio_ring_file = NULL;
		return -EAGAIN;
	}

	file->f_inode->i_mapping->a_ops = &aio_ctx_aops;
	file->f_inode->i_mapping->private_data = ctx;
	file->f_inode->i_size = PAGE_SIZE * (loff_t)nr_pages;

	for (i = 0; i < nr_pages; i++) {
		struct page *page;
		page = find_or_create_page(file->f_inode->i_mapping,
					   i, GFP_HIGHUSER | __GFP_ZERO);
		if (!page)
			break;
		pr_debug("pid(%d) page[%d]->count=%d\n",
			 current->pid, i, page_count(page));
		SetPageUptodate(page);
		SetPageDirty(page);
		unlock_page(page);
	}
	ctx->aio_ring_file = file;
	nr_events = (PAGE_SIZE * nr_pages - sizeof(struct aio_ring))
			/ sizeof(struct io_event);
Linus Torvalds's avatar
Linus Torvalds committed
288

289
	ctx->ring_pages = ctx->internal_pages;
Linus Torvalds's avatar
Linus Torvalds committed
290
	if (nr_pages > AIO_RING_PAGES) {
291
292
293
		ctx->ring_pages = kcalloc(nr_pages, sizeof(struct page *),
					  GFP_KERNEL);
		if (!ctx->ring_pages)
Linus Torvalds's avatar
Linus Torvalds committed
294
295
296
			return -ENOMEM;
	}

297
298
	ctx->mmap_size = nr_pages * PAGE_SIZE;
	pr_debug("attempting mmap of %lu bytes\n", ctx->mmap_size);
299

Zach Brown's avatar
Zach Brown committed
300
	down_write(&mm->mmap_sem);
301
302
303
	ctx->mmap_base = do_mmap_pgoff(ctx->aio_ring_file, 0, ctx->mmap_size,
				       PROT_READ | PROT_WRITE,
				       MAP_SHARED | MAP_POPULATE, 0, &populate);
304
	if (IS_ERR((void *)ctx->mmap_base)) {
Zach Brown's avatar
Zach Brown committed
305
		up_write(&mm->mmap_sem);
306
		ctx->mmap_size = 0;
Linus Torvalds's avatar
Linus Torvalds committed
307
308
309
		aio_free_ring(ctx);
		return -EAGAIN;
	}
310
311
312
	up_write(&mm->mmap_sem);

	mm_populate(ctx->mmap_base, populate);
Linus Torvalds's avatar
Linus Torvalds committed
313

314
315
316
	pr_debug("mmap address: 0x%08lx\n", ctx->mmap_base);
	ctx->nr_pages = get_user_pages(current, mm, ctx->mmap_base, nr_pages,
				       1, 0, ctx->ring_pages, NULL);
317
318
	for (i = 0; i < ctx->nr_pages; i++)
		put_page(ctx->ring_pages[i]);
Linus Torvalds's avatar
Linus Torvalds committed
319

320
	if (unlikely(ctx->nr_pages != nr_pages)) {
Linus Torvalds's avatar
Linus Torvalds committed
321
322
323
324
		aio_free_ring(ctx);
		return -EAGAIN;
	}

325
326
	ctx->user_id = ctx->mmap_base;
	ctx->nr_events = nr_events; /* trusted copy */
Linus Torvalds's avatar
Linus Torvalds committed
327

328
	ring = kmap_atomic(ctx->ring_pages[0]);
Linus Torvalds's avatar
Linus Torvalds committed
329
	ring->nr = nr_events;	/* user copy */
330
	ring->id = ~0U;
Linus Torvalds's avatar
Linus Torvalds committed
331
332
333
334
335
	ring->head = ring->tail = 0;
	ring->magic = AIO_RING_MAGIC;
	ring->compat_features = AIO_RING_COMPAT_FEATURES;
	ring->incompat_features = AIO_RING_INCOMPAT_FEATURES;
	ring->header_length = sizeof(struct aio_ring);
336
	kunmap_atomic(ring);
337
	flush_dcache_page(ctx->ring_pages[0]);
Linus Torvalds's avatar
Linus Torvalds committed
338
339
340
341
342
343
344
345

	return 0;
}

#define AIO_EVENTS_PER_PAGE	(PAGE_SIZE / sizeof(struct io_event))
#define AIO_EVENTS_FIRST_PAGE	((PAGE_SIZE - sizeof(struct aio_ring)) / sizeof(struct io_event))
#define AIO_EVENTS_OFFSET	(AIO_EVENTS_PER_PAGE - AIO_EVENTS_FIRST_PAGE)

346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
void kiocb_set_cancel_fn(struct kiocb *req, kiocb_cancel_fn *cancel)
{
	struct kioctx *ctx = req->ki_ctx;
	unsigned long flags;

	spin_lock_irqsave(&ctx->ctx_lock, flags);

	if (!req->ki_list.next)
		list_add(&req->ki_list, &ctx->active_reqs);

	req->ki_cancel = cancel;

	spin_unlock_irqrestore(&ctx->ctx_lock, flags);
}
EXPORT_SYMBOL(kiocb_set_cancel_fn);

362
static int kiocb_cancel(struct kioctx *ctx, struct kiocb *kiocb)
Kent Overstreet's avatar
Kent Overstreet committed
363
{
364
	kiocb_cancel_fn *old, *cancel;
Kent Overstreet's avatar
Kent Overstreet committed
365

366
367
368
369
370
371
372
373
	/*
	 * Don't want to set kiocb->ki_cancel = KIOCB_CANCELLED unless it
	 * actually has a cancel function, hence the cmpxchg()
	 */

	cancel = ACCESS_ONCE(kiocb->ki_cancel);
	do {
		if (!cancel || cancel == KIOCB_CANCELLED)
Kent Overstreet's avatar
Kent Overstreet committed
374
			return -EINVAL;
Kent Overstreet's avatar
Kent Overstreet committed
375

376
377
378
		old = cancel;
		cancel = cmpxchg(&kiocb->ki_cancel, old, KIOCB_CANCELLED);
	} while (cancel != old);
Kent Overstreet's avatar
Kent Overstreet committed
379

Kent Overstreet's avatar
Kent Overstreet committed
380
	return cancel(kiocb);
Kent Overstreet's avatar
Kent Overstreet committed
381
382
}

Kent Overstreet's avatar
Kent Overstreet committed
383
384
385
static void free_ioctx_rcu(struct rcu_head *head)
{
	struct kioctx *ctx = container_of(head, struct kioctx, rcu_head);
Kent Overstreet's avatar
Kent Overstreet committed
386
387

	free_percpu(ctx->cpu);
Kent Overstreet's avatar
Kent Overstreet committed
388
389
390
391
392
393
394
395
	kmem_cache_free(kioctx_cachep, ctx);
}

/*
 * When this function runs, the kioctx has been removed from the "hash table"
 * and ctx->users has dropped to 0, so we know no more kiocbs can be submitted -
 * now it's safe to cancel any that need to be.
 */
Kent Overstreet's avatar
Kent Overstreet committed
396
static void free_ioctx(struct work_struct *work)
Kent Overstreet's avatar
Kent Overstreet committed
397
{
Kent Overstreet's avatar
Kent Overstreet committed
398
	struct kioctx *ctx = container_of(work, struct kioctx, free_work);
399
	struct aio_ring *ring;
Kent Overstreet's avatar
Kent Overstreet committed
400
	struct kiocb *req;
401
402
	unsigned cpu, avail;
	DEFINE_WAIT(wait);
Kent Overstreet's avatar
Kent Overstreet committed
403
404
405
406
407
408
409
410

	spin_lock_irq(&ctx->ctx_lock);

	while (!list_empty(&ctx->active_reqs)) {
		req = list_first_entry(&ctx->active_reqs,
				       struct kiocb, ki_list);

		list_del_init(&req->ki_list);
411
		kiocb_cancel(ctx, req);
Kent Overstreet's avatar
Kent Overstreet committed
412
413
414
415
	}

	spin_unlock_irq(&ctx->ctx_lock);

Kent Overstreet's avatar
Kent Overstreet committed
416
417
418
419
420
421
422
	for_each_possible_cpu(cpu) {
		struct kioctx_cpu *kcpu = per_cpu_ptr(ctx->cpu, cpu);

		atomic_add(kcpu->reqs_available, &ctx->reqs_available);
		kcpu->reqs_available = 0;
	}

423
424
	while (1) {
		prepare_to_wait(&ctx->wait, &wait, TASK_UNINTERRUPTIBLE);
425

426
427
428
429
		ring = kmap_atomic(ctx->ring_pages[0]);
		avail = (ring->head <= ring->tail)
			 ? ring->tail - ring->head
			 : ctx->nr_events - ring->head + ring->tail;
430

431
		atomic_add(avail, &ctx->reqs_available);
432
433
434
435
436
437
438
		ring->head = ring->tail;
		kunmap_atomic(ring);

		if (atomic_read(&ctx->reqs_available) >= ctx->nr_events - 1)
			break;

		schedule();
439
	}
440
	finish_wait(&ctx->wait, &wait);
441

442
	WARN_ON(atomic_read(&ctx->reqs_available) > ctx->nr_events - 1);
Kent Overstreet's avatar
Kent Overstreet committed
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457

	aio_free_ring(ctx);

	pr_debug("freeing %p\n", ctx);

	/*
	 * Here the call_rcu() is between the wait_event() for reqs_active to
	 * hit 0, and freeing the ioctx.
	 *
	 * aio_complete() decrements reqs_active, but it has to touch the ioctx
	 * after to issue a wakeup so we use rcu.
	 */
	call_rcu(&ctx->rcu_head, free_ioctx_rcu);
}

Kent Overstreet's avatar
Kent Overstreet committed
458
static void free_ioctx_ref(struct percpu_ref *ref)
Kent Overstreet's avatar
Kent Overstreet committed
459
{
Kent Overstreet's avatar
Kent Overstreet committed
460
461
462
463
	struct kioctx *ctx = container_of(ref, struct kioctx, users);

	INIT_WORK(&ctx->free_work, free_ioctx);
	schedule_work(&ctx->free_work);
Kent Overstreet's avatar
Kent Overstreet committed
464
465
}

466
467
468
469
470
471
472
static int ioctx_add_table(struct kioctx *ctx, struct mm_struct *mm)
{
	unsigned i, new_nr;
	struct kioctx_table *table, *old;
	struct aio_ring *ring;

	spin_lock(&mm->ioctx_lock);
473
	table = rcu_dereference(mm->ioctx_table);
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500

	while (1) {
		if (table)
			for (i = 0; i < table->nr; i++)
				if (!table->table[i]) {
					ctx->id = i;
					table->table[i] = ctx;
					spin_unlock(&mm->ioctx_lock);

					ring = kmap_atomic(ctx->ring_pages[0]);
					ring->id = ctx->id;
					kunmap_atomic(ring);
					return 0;
				}

		new_nr = (table ? table->nr : 1) * 4;

		spin_unlock(&mm->ioctx_lock);

		table = kzalloc(sizeof(*table) + sizeof(struct kioctx *) *
				new_nr, GFP_KERNEL);
		if (!table)
			return -ENOMEM;

		table->nr = new_nr;

		spin_lock(&mm->ioctx_lock);
501
		old = rcu_dereference(mm->ioctx_table);
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517

		if (!old) {
			rcu_assign_pointer(mm->ioctx_table, table);
		} else if (table->nr > old->nr) {
			memcpy(table->table, old->table,
			       old->nr * sizeof(struct kioctx *));

			rcu_assign_pointer(mm->ioctx_table, table);
			kfree_rcu(old, rcu);
		} else {
			kfree(table);
			table = old;
		}
	}
}

Linus Torvalds's avatar
Linus Torvalds committed
518
519
520
521
522
/* ioctx_alloc
 *	Allocates and initializes an ioctx.  Returns an ERR_PTR if it failed.
 */
static struct kioctx *ioctx_alloc(unsigned nr_events)
{
Zach Brown's avatar
Zach Brown committed
523
	struct mm_struct *mm = current->mm;
Linus Torvalds's avatar
Linus Torvalds committed
524
	struct kioctx *ctx;
525
	int err = -ENOMEM;
Linus Torvalds's avatar
Linus Torvalds committed
526

Kent Overstreet's avatar
Kent Overstreet committed
527
528
529
530
531
532
533
534
535
536
537
538
	/*
	 * We keep track of the number of available ringbuffer slots, to prevent
	 * overflow (reqs_available), and we also use percpu counters for this.
	 *
	 * So since up to half the slots might be on other cpu's percpu counters
	 * and unavailable, double nr_events so userspace sees what they
	 * expected: additionally, we move req_batch slots to/from percpu
	 * counters at a time, so make sure that isn't 0:
	 */
	nr_events = max(nr_events, num_possible_cpus() * 4);
	nr_events *= 2;

Linus Torvalds's avatar
Linus Torvalds committed
539
540
541
542
543
544
545
	/* Prevent overflows */
	if ((nr_events > (0x10000000U / sizeof(struct io_event))) ||
	    (nr_events > (0x10000000U / sizeof(struct kiocb)))) {
		pr_debug("ENOMEM: nr_events too high\n");
		return ERR_PTR(-EINVAL);
	}

546
	if (!nr_events || (unsigned long)nr_events > (aio_max_nr * 2UL))
Linus Torvalds's avatar
Linus Torvalds committed
547
548
		return ERR_PTR(-EAGAIN);

549
	ctx = kmem_cache_zalloc(kioctx_cachep, GFP_KERNEL);
Linus Torvalds's avatar
Linus Torvalds committed
550
551
552
553
554
	if (!ctx)
		return ERR_PTR(-ENOMEM);

	ctx->max_reqs = nr_events;

Kent Overstreet's avatar
Kent Overstreet committed
555
556
557
	if (percpu_ref_init(&ctx->users, free_ioctx_ref))
		goto out_freectx;

Linus Torvalds's avatar
Linus Torvalds committed
558
	spin_lock_init(&ctx->ctx_lock);
559
	spin_lock_init(&ctx->completion_lock);
560
	mutex_init(&ctx->ring_lock);
Linus Torvalds's avatar
Linus Torvalds committed
561
562
563
564
	init_waitqueue_head(&ctx->wait);

	INIT_LIST_HEAD(&ctx->active_reqs);

Kent Overstreet's avatar
Kent Overstreet committed
565
566
	ctx->cpu = alloc_percpu(struct kioctx_cpu);
	if (!ctx->cpu)
Kent Overstreet's avatar
Kent Overstreet committed
567
		goto out_freeref;
Linus Torvalds's avatar
Linus Torvalds committed
568

Kent Overstreet's avatar
Kent Overstreet committed
569
570
571
	if (aio_setup_ring(ctx) < 0)
		goto out_freepcpu;

572
	atomic_set(&ctx->reqs_available, ctx->nr_events - 1);
Kent Overstreet's avatar
Kent Overstreet committed
573
	ctx->req_batch = (ctx->nr_events - 1) / (num_possible_cpus() * 4);
574
575
	if (ctx->req_batch < 1)
		ctx->req_batch = 1;
576

Linus Torvalds's avatar
Linus Torvalds committed
577
	/* limit the number of system wide aios */
578
	spin_lock(&aio_nr_lock);
579
	if (aio_nr + nr_events > (aio_max_nr * 2UL) ||
580
	    aio_nr + nr_events < aio_nr) {
581
		spin_unlock(&aio_nr_lock);
Linus Torvalds's avatar
Linus Torvalds committed
582
		goto out_cleanup;
583
584
	}
	aio_nr += ctx->max_reqs;
585
	spin_unlock(&aio_nr_lock);
Linus Torvalds's avatar
Linus Torvalds committed
586

Kent Overstreet's avatar
Kent Overstreet committed
587
588
	percpu_ref_get(&ctx->users); /* io_setup() will drop this ref */

589
590
591
592
	err = ioctx_add_table(ctx, mm);
	if (err)
		goto out_cleanup_put;

Kent Overstreet's avatar
Kent Overstreet committed
593
	pr_debug("allocated ioctx %p[%ld]: mm=%p mask=0x%x\n",
594
		 ctx, ctx->user_id, mm, ctx->nr_events);
Linus Torvalds's avatar
Linus Torvalds committed
595
596
	return ctx;

597
598
out_cleanup_put:
	percpu_ref_put(&ctx->users);
Linus Torvalds's avatar
Linus Torvalds committed
599
out_cleanup:
600
601
	err = -EAGAIN;
	aio_free_ring(ctx);
Kent Overstreet's avatar
Kent Overstreet committed
602
603
out_freepcpu:
	free_percpu(ctx->cpu);
Kent Overstreet's avatar
Kent Overstreet committed
604
605
out_freeref:
	free_percpu(ctx->users.pcpu_count);
Linus Torvalds's avatar
Linus Torvalds committed
606
out_freectx:
607
608
	if (ctx->aio_ring_file)
		fput(ctx->aio_ring_file);
Linus Torvalds's avatar
Linus Torvalds committed
609
	kmem_cache_free(kioctx_cachep, ctx);
Kent Overstreet's avatar
Kent Overstreet committed
610
	pr_debug("error allocating ioctx %d\n", err);
611
	return ERR_PTR(err);
Linus Torvalds's avatar
Linus Torvalds committed
612
613
}

Kent Overstreet's avatar
Kent Overstreet committed
614
615
616
617
618
/* kill_ioctx
 *	Cancels all outstanding aio requests on an aio context.  Used
 *	when the processes owning a context have all exited to encourage
 *	the rapid destruction of the kioctx.
 */
619
static void kill_ioctx(struct mm_struct *mm, struct kioctx *ctx)
Kent Overstreet's avatar
Kent Overstreet committed
620
621
{
	if (!atomic_xchg(&ctx->dead, 1)) {
622
623
624
		struct kioctx_table *table;

		spin_lock(&mm->ioctx_lock);
625
		table = rcu_dereference(mm->ioctx_table);
626
627
628
629
630

		WARN_ON(ctx != table->table[ctx->id]);
		table->table[ctx->id] = NULL;
		spin_unlock(&mm->ioctx_lock);

Kent Overstreet's avatar
Kent Overstreet committed
631
632
		/* percpu_ref_kill() will do the necessary call_rcu() */
		wake_up_all(&ctx->wait);
633

Kent Overstreet's avatar
Kent Overstreet committed
634
		/*
635
636
637
638
639
		 * It'd be more correct to do this in free_ioctx(), after all
		 * the outstanding kiocbs have finished - but by then io_destroy
		 * has already returned, so io_setup() could potentially return
		 * -EAGAIN with no ioctxs actually in use (as far as userspace
		 *  could tell).
Kent Overstreet's avatar
Kent Overstreet committed
640
		 */
641
642
643
644
645
646
647
648
		spin_lock(&aio_nr_lock);
		BUG_ON(aio_nr - ctx->max_reqs > aio_nr);
		aio_nr -= ctx->max_reqs;
		spin_unlock(&aio_nr_lock);

		if (ctx->mmap_size)
			vm_munmap(ctx->mmap_base, ctx->mmap_size);

Kent Overstreet's avatar
Kent Overstreet committed
649
		percpu_ref_kill(&ctx->users);
Kent Overstreet's avatar
Kent Overstreet committed
650
	}
Linus Torvalds's avatar
Linus Torvalds committed
651
652
653
654
655
}

/* wait_on_sync_kiocb:
 *	Waits on the given sync kiocb to complete.
 */
Kent Overstreet's avatar
Kent Overstreet committed
656
ssize_t wait_on_sync_kiocb(struct kiocb *req)
Linus Torvalds's avatar
Linus Torvalds committed
657
{
Kent Overstreet's avatar
Kent Overstreet committed
658
	while (!req->ki_ctx) {
Linus Torvalds's avatar
Linus Torvalds committed
659
		set_current_state(TASK_UNINTERRUPTIBLE);
Kent Overstreet's avatar
Kent Overstreet committed
660
		if (req->ki_ctx)
Linus Torvalds's avatar
Linus Torvalds committed
661
			break;
662
		io_schedule();
Linus Torvalds's avatar
Linus Torvalds committed
663
664
	}
	__set_current_state(TASK_RUNNING);
Kent Overstreet's avatar
Kent Overstreet committed
665
	return req->ki_user_data;
Linus Torvalds's avatar
Linus Torvalds committed
666
}
667
EXPORT_SYMBOL(wait_on_sync_kiocb);
Linus Torvalds's avatar
Linus Torvalds committed
668

Kent Overstreet's avatar
Kent Overstreet committed
669
670
671
672
673
674
675
/*
 * exit_aio: called when the last user of mm goes away.  At this point, there is
 * no way for any new requests to be submited or any of the io_* syscalls to be
 * called on the context.
 *
 * There may be outstanding kiocbs, but free_ioctx() will explicitly wait on
 * them.
Linus Torvalds's avatar
Linus Torvalds committed
676
 */
677
void exit_aio(struct mm_struct *mm)
Linus Torvalds's avatar
Linus Torvalds committed
678
{
679
	struct kioctx_table *table;
680
	struct kioctx *ctx;
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
	unsigned i = 0;

	while (1) {
		rcu_read_lock();
		table = rcu_dereference(mm->ioctx_table);

		do {
			if (!table || i >= table->nr) {
				rcu_read_unlock();
				rcu_assign_pointer(mm->ioctx_table, NULL);
				if (table)
					kfree(table);
				return;
			}

			ctx = table->table[i++];
		} while (!ctx);

		rcu_read_unlock();
700

701
702
703
704
705
706
707
708
		/*
		 * We don't need to bother with munmap() here -
		 * exit_mmap(mm) is coming and it'll unmap everything.
		 * Since aio_free_ring() uses non-zero ->mmap_size
		 * as indicator that it needs to unmap the area,
		 * just set it to 0; aio_free_ring() is the only
		 * place that uses ->mmap_size, so it's safe.
		 */
709
		ctx->mmap_size = 0;
Kent Overstreet's avatar
Kent Overstreet committed
710

711
		kill_ioctx(mm, ctx);
Linus Torvalds's avatar
Linus Torvalds committed
712
713
714
	}
}

Kent Overstreet's avatar
Kent Overstreet committed
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
static void put_reqs_available(struct kioctx *ctx, unsigned nr)
{
	struct kioctx_cpu *kcpu;

	preempt_disable();
	kcpu = this_cpu_ptr(ctx->cpu);

	kcpu->reqs_available += nr;
	while (kcpu->reqs_available >= ctx->req_batch * 2) {
		kcpu->reqs_available -= ctx->req_batch;
		atomic_add(ctx->req_batch, &ctx->reqs_available);
	}

	preempt_enable();
}

static bool get_reqs_available(struct kioctx *ctx)
{
	struct kioctx_cpu *kcpu;
	bool ret = false;

	preempt_disable();
	kcpu = this_cpu_ptr(ctx->cpu);

	if (!kcpu->reqs_available) {
		int old, avail = atomic_read(&ctx->reqs_available);

		do {
			if (avail < ctx->req_batch)
				goto out;

			old = avail;
			avail = atomic_cmpxchg(&ctx->reqs_available,
					       avail, avail - ctx->req_batch);
		} while (avail != old);

		kcpu->reqs_available += ctx->req_batch;
	}

	ret = true;
	kcpu->reqs_available--;
out:
	preempt_enable();
	return ret;
}

Linus Torvalds's avatar
Linus Torvalds committed
761
/* aio_get_req
Kent Overstreet's avatar
Kent Overstreet committed
762
763
 *	Allocate a slot for an aio request.
 * Returns NULL if no requests are free.
Linus Torvalds's avatar
Linus Torvalds committed
764
 */
Kent Overstreet's avatar
Kent Overstreet committed
765
static inline struct kiocb *aio_get_req(struct kioctx *ctx)
Linus Torvalds's avatar
Linus Torvalds committed
766
{
Kent Overstreet's avatar
Kent Overstreet committed
767
768
	struct kiocb *req;

Kent Overstreet's avatar
Kent Overstreet committed
769
	if (!get_reqs_available(ctx))
Kent Overstreet's avatar
Kent Overstreet committed
770
771
		return NULL;

772
	req = kmem_cache_alloc(kiocb_cachep, GFP_KERNEL|__GFP_ZERO);
Linus Torvalds's avatar
Linus Torvalds committed
773
	if (unlikely(!req))
Kent Overstreet's avatar
Kent Overstreet committed
774
		goto out_put;
Linus Torvalds's avatar
Linus Torvalds committed
775
776

	req->ki_ctx = ctx;
Jeff Moyer's avatar
Jeff Moyer committed
777
	return req;
Kent Overstreet's avatar
Kent Overstreet committed
778
out_put:
Kent Overstreet's avatar
Kent Overstreet committed
779
	put_reqs_available(ctx, 1);
Kent Overstreet's avatar
Kent Overstreet committed
780
	return NULL;
Linus Torvalds's avatar
Linus Torvalds committed
781
782
}

783
static void kiocb_free(struct kiocb *req)
Linus Torvalds's avatar
Linus Torvalds committed
784
{
785
786
	if (req->ki_filp)
		fput(req->ki_filp);
787
788
	if (req->ki_eventfd != NULL)
		eventfd_ctx_put(req->ki_eventfd);
Linus Torvalds's avatar
Linus Torvalds committed
789
790
791
	kmem_cache_free(kiocb_cachep, req);
}

792
static struct kioctx *lookup_ioctx(unsigned long ctx_id)
Linus Torvalds's avatar
Linus Torvalds committed
793
{
794
	struct aio_ring __user *ring  = (void __user *)ctx_id;
795
	struct mm_struct *mm = current->mm;
796
	struct kioctx *ctx, *ret = NULL;
797
798
799
800
801
	struct kioctx_table *table;
	unsigned id;

	if (get_user(id, &ring->id))
		return NULL;
Linus Torvalds's avatar
Linus Torvalds committed
802

803
	rcu_read_lock();
804
	table = rcu_dereference(mm->ioctx_table);
805

806
807
	if (!table || id >= table->nr)
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
808

809
	ctx = table->table[id];
810
	if (ctx && ctx->user_id == ctx_id) {
811
812
813
814
		percpu_ref_get(&ctx->users);
		ret = ctx;
	}
out:
815
	rcu_read_unlock();
816
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
817
818
819
820
821
}

/* aio_complete
 *	Called when the io request on the given iocb is complete.
 */
822
void aio_complete(struct kiocb *iocb, long res, long res2)
Linus Torvalds's avatar
Linus Torvalds committed
823
824
825
{
	struct kioctx	*ctx = iocb->ki_ctx;
	struct aio_ring	*ring;
Kent Overstreet's avatar
Kent Overstreet committed
826
	struct io_event	*ev_page, *event;
Linus Torvalds's avatar
Linus Torvalds committed
827
	unsigned long	flags;
Kent Overstreet's avatar
Kent Overstreet committed
828
	unsigned tail, pos;
Linus Torvalds's avatar
Linus Torvalds committed
829

830
831
832
833
834
835
	/*
	 * Special case handling for sync iocbs:
	 *  - events go directly into the iocb for fast handling
	 *  - the sync task with the iocb in its stack holds the single iocb
	 *    ref, no other paths have a way to get another ref
	 *  - the sync task helpfully left a reference to itself in the iocb
Linus Torvalds's avatar
Linus Torvalds committed
836
837
838
	 */
	if (is_sync_kiocb(iocb)) {
		iocb->ki_user_data = res;
Kent Overstreet's avatar
Kent Overstreet committed
839
840
		smp_wmb();
		iocb->ki_ctx = ERR_PTR(-EXDEV);
Linus Torvalds's avatar
Linus Torvalds committed
841
		wake_up_process(iocb->ki_obj.tsk);
842
		return;
Linus Torvalds's avatar
Linus Torvalds committed
843
844
	}

Kent Overstreet's avatar
Kent Overstreet committed
845
846
	/*
	 * Take rcu_read_lock() in case the kioctx is being destroyed, as we
847
	 * need to issue a wakeup after incrementing reqs_available.
Linus Torvalds's avatar
Linus Torvalds committed
848
	 */
Kent Overstreet's avatar
Kent Overstreet committed
849
	rcu_read_lock();
Linus Torvalds's avatar
Linus Torvalds committed
850

851
852
853
854
855
856
857
	if (iocb->ki_list.next) {
		unsigned long flags;

		spin_lock_irqsave(&ctx->ctx_lock, flags);
		list_del(&iocb->ki_list);
		spin_unlock_irqrestore(&ctx->ctx_lock, flags);
	}
858

859
860
	/*
	 * Add a completion event to the ring buffer. Must be done holding
861
	 * ctx->completion_lock to prevent other code from messing with the tail
862
863
864
865
	 * pointer since we might be called from irq context.
	 */
	spin_lock_irqsave(&ctx->completion_lock, flags);

866
	tail = ctx->tail;
Kent Overstreet's avatar
Kent Overstreet committed
867
868
	pos = tail + AIO_EVENTS_OFFSET;

869
	if (++tail >= ctx->nr_events)
870
		tail = 0;
Linus Torvalds's avatar
Linus Torvalds committed
871

872
	ev_page = kmap_atomic(ctx->ring_pages[pos / AIO_EVENTS_PER_PAGE]);
Kent Overstreet's avatar
Kent Overstreet committed
873
874
	event = ev_page + pos % AIO_EVENTS_PER_PAGE;

Linus Torvalds's avatar
Linus Torvalds committed
875
876
877
878
879
	event->obj = (u64)(unsigned long)iocb->ki_obj.user;
	event->data = iocb->ki_user_data;
	event->res = res;
	event->res2 = res2;

Kent Overstreet's avatar
Kent Overstreet committed
880
	kunmap_atomic(ev_page);
881
	flush_dcache_page(ctx->ring_pages[pos / AIO_EVENTS_PER_PAGE]);
Kent Overstreet's avatar
Kent Overstreet committed
882
883

	pr_debug("%p[%u]: %p: %p %Lx %lx %lx\n",
Kent Overstreet's avatar
Kent Overstreet committed
884
885
		 ctx, tail, iocb, iocb->ki_obj.user, iocb->ki_user_data,
		 res, res2);
Linus Torvalds's avatar
Linus Torvalds committed
886
887
888
889
890
891

	/* after flagging the request as done, we
	 * must never even look at it again
	 */
	smp_wmb();	/* make event visible before updating tail */

892
	ctx->tail = tail;
Linus Torvalds's avatar
Linus Torvalds committed
893

894
	ring = kmap_atomic(ctx->ring_pages[0]);
Kent Overstreet's avatar
Kent Overstreet committed
895
	ring->tail = tail;
896
	kunmap_atomic(ring);
897
	flush_dcache_page(ctx->ring_pages[0]);
Linus Torvalds's avatar
Linus Torvalds committed
898

899
900
	spin_unlock_irqrestore(&ctx->completion_lock, flags);

Kent Overstreet's avatar
Kent Overstreet committed
901
	pr_debug("added to ring %p at [%u]\n", iocb, tail);
Davide Libenzi's avatar
Davide Libenzi committed
902
903
904
905
906
907

	/*
	 * Check if the user asked us to deliver the result through an
	 * eventfd. The eventfd_signal() function is safe to be called
	 * from IRQ context.
	 */
908
	if (iocb->ki_eventfd != NULL)
Davide Libenzi's avatar
Davide Libenzi committed
909
910
		eventfd_signal(iocb->ki_eventfd, 1);

Linus Torvalds's avatar
Linus Torvalds committed
911
	/* everything turned out well, dispose of the aiocb. */
Kent Overstreet's avatar
Kent Overstreet committed
912
	kiocb_free(iocb);
Linus Torvalds's avatar
Linus Torvalds committed
913

914
915
916
917
918
919
920
921
	/*
	 * We have to order our ring_info tail store above and test
	 * of the wait list below outside the wait lock.  This is
	 * like in wake_up_bit() where clearing a bit has to be
	 * ordered with the unlocked test.
	 */
	smp_mb();

Linus Torvalds's avatar
Linus Torvalds committed
922
923
924
	if (waitqueue_active(&ctx->wait))
		wake_up(&ctx->wait);

Kent Overstreet's avatar
Kent Overstreet committed
925
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
926
}
927
EXPORT_SYMBOL(aio_complete);
Linus Torvalds's avatar
Linus Torvalds committed
928

929
930
931
/* aio_read_events
 *	Pull an event off of the ioctx's event ring.  Returns the number of
 *	events fetched
Linus Torvalds's avatar
Linus Torvalds committed
932
 */
933
934
static long aio_read_events_ring(struct kioctx *ctx,
				 struct io_event __user *event, long nr)
Linus Torvalds's avatar
Linus Torvalds committed
935
936
{
	struct aio_ring *ring;
937
	unsigned head, tail, pos;
938
939
940
	long ret = 0;
	int copy_ret;

941
	mutex_lock(&ctx->ring_lock);
Linus Torvalds's avatar
Linus Torvalds committed
942

943
	ring = kmap_atomic(ctx->ring_pages[0]);
944
	head = ring->head;
945
	tail = ring->tail;
946
947
	kunmap_atomic(ring);

948
	pr_debug("h%u t%u m%u\n", head, tail, ctx->nr_events);
Linus Torvalds's avatar
Linus Torvalds committed
949

950
	if (head == tail)
Linus Torvalds's avatar
Linus Torvalds committed
951
952
		goto out;

953
954
955
956
957
	while (ret < nr) {
		long avail;
		struct io_event *ev;
		struct page *page;

958
959
		avail = (head <= tail ?  tail : ctx->nr_events) - head;
		if (head == tail)
960
961
962
963
964
965
966
			break;

		avail = min(avail, nr - ret);
		avail = min_t(long, avail, AIO_EVENTS_PER_PAGE -
			    ((head + AIO_EVENTS_OFFSET) % AIO_EVENTS_PER_PAGE));

		pos = head + AIO_EVENTS_OFFSET;
967
		page = ctx->ring_pages[pos / AIO_EVENTS_PER_PAGE];
968
969
970
971
972
973
974
975
976
977
978
979
980
981
		pos %= AIO_EVENTS_PER_PAGE;

		ev = kmap(page);
		copy_ret = copy_to_user(event + ret, ev + pos,
					sizeof(*ev) * avail);
		kunmap(page);

		if (unlikely(copy_ret)) {
			ret = -EFAULT;
			goto out;
		}

		ret += avail;
		head += avail;
982
		head %= ctx->nr_events;
Linus Torvalds's avatar
Linus Torvalds committed
983
984
	}

985
	ring = kmap_atomic(ctx->ring_pages[0]);
986
	ring->head = head;
987
	kunmap_atomic(ring);
988
	flush_dcache_page(ctx->ring_pages[0]);
989

990
	pr_debug("%li  h%u t%u\n", ret, head, tail);
991

Kent Overstreet's avatar
Kent Overstreet committed
992
	put_reqs_available(ctx, ret);
993
out:
994
	mutex_unlock(&ctx->ring_lock);
995

Linus Torvalds's avatar
Linus Torvalds committed
996
997
998
	return ret;
}

999
1000
static bool aio_read_events(struct kioctx *ctx, long min_nr, long nr,
			    struct io_event __user *event, long *i)
Linus Torvalds's avatar
Linus Torvalds committed
1001
{
1002
	long ret = aio_read_events_ring(ctx, event + *i, nr - *i);
Linus Torvalds's avatar
Linus Torvalds committed
1003

1004
1005
	if (ret > 0)
		*i += ret;
Linus Torvalds's avatar
Linus Torvalds committed
1006

1007
1008
	if (unlikely(atomic_read(&ctx->dead)))
		ret = -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
1009

1010
1011
	if (!*i)
		*i = ret;
Linus Torvalds's avatar
Linus Torvalds committed
1012

1013
	return ret < 0 || *i >= min_nr;
Linus Torvalds's avatar
Linus Torvalds committed
1014
1015
}

1016
static long read_events(struct kioctx *ctx, long min_nr, long nr,
Linus Torvalds's avatar
Linus Torvalds committed
1017
1018
1019
			struct io_event __user *event,
			struct timespec __user *timeout)
{
1020
1021
	ktime_t until = { .tv64 = KTIME_MAX };
	long ret = 0;
Linus Torvalds's avatar
Linus Torvalds committed
1022
1023
1024

	if (timeout) {
		struct timespec	ts;
1025

Linus Torvalds's avatar
Linus Torvalds committed
1026
		if (unlikely(copy_from_user(&ts, timeout, sizeof(ts))))
1027
			return -EFAULT;
Linus Torvalds's avatar
Linus Torvalds committed
1028

1029
		until = timespec_to_ktime(ts);
Linus Torvalds's avatar
Linus Torvalds committed
1030
1031
	}

1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
	/*
	 * Note that aio_read_events() is being called as the conditional - i.e.
	 * we're calling it after prepare_to_wait() has set task state to
	 * TASK_INTERRUPTIBLE.
	 *
	 * But aio_read_events() can block, and if it blocks it's going to flip
	 * the task state back to TASK_RUNNING.
	 *
	 * This should be ok, provided it doesn't flip the state back to
	 * TASK_RUNNING and return 0 too much - that causes us to spin. That
	 * will only happen if the mutex_lock() call blocks, and we then find
	 * the ringbuffer empty. So in practice we should be ok, but it's
	 * something to be aware of when touching this code.
	 */
	wait_event_interruptible_hrtimeout(ctx->wait,
			aio_read_events(ctx, min_nr, nr, event, &ret), until);
Linus Torvalds's avatar
Linus Torvalds committed
1048

1049
1050
	if (!ret && signal_pending(current))
		ret = -EINTR;
Linus Torvalds's avatar
Linus Torvalds committed
1051

1052
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
}

/* sys_io_setup:
 *	Create an aio_context capable of receiving at least nr_events.
 *	ctxp must not point to an aio_context that already exists, and
 *	must be initialized to 0 prior to the call.  On successful
 *	creation of the aio_context, *ctxp is filled in with the resulting 
 *	handle.  May fail with -EINVAL if *ctxp is not initialized,
 *	if the specified nr_events exceeds internal limits.  May fail 
 *	with -EAGAIN if the specified nr_events exceeds the user's limit 
 *	of available events.  May fail with -ENOMEM if insufficient kernel
 *	resources are available.  May fail with -EFAULT if an invalid
 *	pointer is passed for ctxp.  Will fail with -ENOSYS if not
 *	implemented.
 */
1068
SYSCALL_DEFINE2(io_setup, unsigned, nr_events, aio_context_t __user *, ctxp)
Linus Torvalds's avatar
Linus Torvalds committed
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
{
	struct kioctx *ioctx = NULL;
	unsigned long ctx;
	long ret;

	ret = get_user(ctx, ctxp);
	if (unlikely(ret))
		goto out;

	ret = -EINVAL;
1079
1080
1081
	if (unlikely(ctx || nr_events == 0)) {
		pr_debug("EINVAL: io_setup: ctx %lu nr_events %u\n",
		         ctx, nr_events);
Linus Torvalds's avatar
Linus Torvalds committed
1082
1083
1084
1085
1086
1087
1088
		goto out;
	}

	ioctx = ioctx_alloc(nr_events);
	ret = PTR_ERR(ioctx);
	if (!IS_ERR(ioctx)) {
		ret = put_user(ioctx->user_id, ctxp);
1089
		if (ret)
1090
			kill_ioctx(current->mm, ioctx);
Kent Overstreet's avatar
Kent Overstreet committed
1091
		percpu_ref_put(&ioctx->users);
Linus Torvalds's avatar
Linus Torvalds committed
1092
1093
1094
1095
1096
1097
1098
1099
1100
	}

out:
	return ret;
}

/* sys_io_destroy:
 *	Destroy the aio_context specified.  May cancel any outstanding 
 *	AIOs and block on completion.  Will fail with -ENOSYS if not
1101
 *	implemented.  May fail with -EINVAL if the context pointed to
Linus Torvalds's avatar
Linus Torvalds committed
1102
1103
 *	is invalid.
 */
1104
SYSCALL_DEFINE1(io_destroy, aio_context_t, ctx)
Linus Torvalds's avatar
Linus Torvalds committed
1105
1106
1107
{
	struct kioctx *ioctx = lookup_ioctx(ctx);
	if (likely(NULL != ioctx)) {
1108
		kill_ioctx(current->mm, ioctx);
Kent Overstreet's avatar
Kent Overstreet committed
1109
		percpu_ref_put(&ioctx->users);
Linus Torvalds's avatar
Linus Torvalds committed
1110
1111
1112
1113
1114
1115
		return 0;
	}
	pr_debug("EINVAL: io_destroy: invalid context id\n");
	return -EINVAL;
}

Kent Overstreet's avatar
Kent Overstreet committed
1116
1117
1118
typedef ssize_t (aio_rw_op)(struct kiocb *, const struct iovec *,
			    unsigned long, loff_t);

1119
1120
1121
1122
1123
static ssize_t aio_setup_vectored_rw(struct kiocb *kiocb,
				     int rw, char __user *buf,
				     unsigned long *nr_segs,
				     struct iovec **iovec,
				     bool compat)
1124
1125
1126
{
	ssize_t ret;

1127
	*nr_segs = kiocb->ki_nbytes;
Kent Overstreet's avatar
Kent Overstreet committed
1128

1129
1130
#ifdef CONFIG_COMPAT
	if (compat)
Kent Overstreet's avatar
Kent Overstreet committed
1131
		ret = compat_rw_copy_check_uvector(rw,
1132
1133
				(struct compat_iovec __user *)buf,
				*nr_segs, 1, *iovec, iovec);
1134
1135
	else
#endif
Kent Overstreet's avatar
Kent Overstreet committed
1136
		ret = rw_copy_check_uvector(rw,
1137
1138
				(struct iovec __user *)buf,
				*nr_segs, 1, *iovec, iovec);
1139
	if (ret < 0)
Kent Overstreet's avatar
Kent Overstreet committed
1140
		return ret;
1141

Kent Overstreet's avatar
Kent Overstreet committed
1142
	/* ki_nbytes now reflect bytes instead of segs */
1143
	kiocb->ki_nbytes = ret;
Kent Overstreet's avatar
Kent Overstreet committed
1144
	return 0;
1145
1146
}

1147
1148
1149
1150
static ssize_t aio_setup_single_vector(struct kiocb *kiocb,
				       int rw, char __user *buf,
				       unsigned long *nr_segs,
				       struct iovec *iovec)
1151
{
1152
	if (unlikely(!access_ok(!rw, buf, kiocb->ki_nbytes)))
Kent Overstreet's avatar
Kent Overstreet committed
1153
		return -EFAULT;
1154

1155
1156
1157
	iovec->iov_base = buf;
	iovec->iov_len = kiocb->ki_nbytes;
	*nr_segs = 1;
1158
1159
1160
	return 0;
}

Linus Torvalds's avatar
Linus Torvalds committed
1161
1162
1163
1164
1165
/*
 * aio_setup_iocb:
 *	Performs the initial checks and aio retry method
 *	setup for the kiocb at the time of io submission.
 */
1166
1167
static ssize_t aio_run_iocb(struct kiocb *req, unsigned opcode,
			    char __user *buf, bool compat)
Linus Torvalds's avatar
Linus Torvalds committed
1168
{
Kent Overstreet's avatar
Kent Overstreet committed
1169
1170
	struct file *file = req->ki_filp;
	ssize_t ret;
1171
	unsigned long nr_segs;
Kent Overstreet's avatar
Kent Overstreet committed
1172
1173
1174
	int rw;
	fmode_t mode;
	aio_rw_op *rw_op;
1175
	struct iovec inline_vec, *iovec = &inline_vec;
Linus Torvalds's avatar
Linus Torvalds committed
1176