aio.c 35.5 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

Kent Overstreet's avatar
Kent Overstreet committed
69
70
71
72
struct kioctx_cpu {
	unsigned		reqs_available;
};

73
struct kioctx {
Kent Overstreet's avatar
Kent Overstreet committed
74
	struct percpu_ref	users;
Kent Overstreet's avatar
Kent Overstreet committed
75
	atomic_t		dead;
76
77
78
79
80

	/* This needs improving */
	unsigned long		user_id;
	struct hlist_node	list;

Kent Overstreet's avatar
Kent Overstreet committed
81
82
83
84
85
86
87
	struct __percpu kioctx_cpu *cpu;

	/*
	 * For percpu reqs_available, number of slots we move to/from global
	 * counter at a time:
	 */
	unsigned		req_batch;
88
89
90
91
	/*
	 * This is what userspace passed to io_setup(), it's not used for
	 * anything but counting against the global max_reqs quota.
	 *
92
	 * The real limit is nr_events - 1, which will be larger (see
93
94
	 * aio_setup_ring())
	 */
95
96
	unsigned		max_reqs;

97
98
	/* Size of ringbuffer, in units of struct io_event */
	unsigned		nr_events;
99

100
101
102
103
104
105
	unsigned long		mmap_base;
	unsigned long		mmap_size;

	struct page		**ring_pages;
	long			nr_pages;

106
	struct rcu_head		rcu_head;
Kent Overstreet's avatar
Kent Overstreet committed
107
	struct work_struct	free_work;
108
109

	struct {
110
111
112
113
114
		/*
		 * 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
115
116
		 *
		 * We batch accesses to it with a percpu version.
117
118
		 */
		atomic_t	reqs_available;
119
120
121
122
123
124
125
	} ____cacheline_aligned_in_smp;

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

126
127
	struct {
		struct mutex	ring_lock;
128
129
		wait_queue_head_t wait;
	} ____cacheline_aligned_in_smp;
130
131
132
133

	struct {
		unsigned	tail;
		spinlock_t	completion_lock;
134
	} ____cacheline_aligned_in_smp;
135
136

	struct page		*internal_pages[AIO_RING_PAGES];
137
	struct file		*aio_ring_file;
138
139
};

Linus Torvalds's avatar
Linus Torvalds committed
140
/*------ sysctl variables----*/
141
142
143
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
144
145
/*----end sysctl variables---*/

146
147
static struct kmem_cache	*kiocb_cachep;
static struct kmem_cache	*kioctx_cachep;
Linus Torvalds's avatar
Linus Torvalds committed
148
149
150
151
152
153
154

/* 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)
{
155
156
	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
157

Kent Overstreet's avatar
Kent Overstreet committed
158
	pr_debug("sizeof(struct page) = %zu\n", sizeof(struct page));
Linus Torvalds's avatar
Linus Torvalds committed
159
160
161

	return 0;
}
162
__initcall(aio_setup);
Linus Torvalds's avatar
Linus Torvalds committed
163
164
165

static void aio_free_ring(struct kioctx *ctx)
{
166
167
	int i;
	struct file *aio_ring_file = ctx->aio_ring_file;
Linus Torvalds's avatar
Linus Torvalds committed
168

169
170
171
	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]));
172
		put_page(ctx->ring_pages[i]);
173
	}
Linus Torvalds's avatar
Linus Torvalds committed
174

175
176
	if (ctx->ring_pages && ctx->ring_pages != ctx->internal_pages)
		kfree(ctx->ring_pages);
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204

	if (aio_ring_file) {
		truncate_setsize(aio_ring_file->f_inode, 0);
		pr_debug("pid(%d) i_nlink=%u d_count=%d d_unhashed=%d i_count=%d\n",
			current->pid, aio_ring_file->f_inode->i_nlink,
			aio_ring_file->f_path.dentry->d_count,
			d_unhashed(aio_ring_file->f_path.dentry),
			atomic_read(&aio_ring_file->f_inode->i_count));
		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;
}

205
#if IS_ENABLED(CONFIG_MIGRATION)
206
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
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
232
}
233
#endif
Linus Torvalds's avatar
Linus Torvalds committed
234

235
236
static const struct address_space_operations aio_ctx_aops = {
	.set_page_dirty = aio_set_page_dirty,
237
#if IS_ENABLED(CONFIG_MIGRATION)
238
	.migratepage	= aio_migratepage,
239
#endif
240
241
};

Linus Torvalds's avatar
Linus Torvalds committed
242
243
244
245
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
246
	struct mm_struct *mm = current->mm;
247
	unsigned long size, populate;
Linus Torvalds's avatar
Linus Torvalds committed
248
	int nr_pages;
249
250
	int i;
	struct file *file;
Linus Torvalds's avatar
Linus Torvalds committed
251
252
253
254
255
256
257

	/* 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;

258
	nr_pages = PFN_UP(size);
Linus Torvalds's avatar
Linus Torvalds committed
259
260
261
	if (nr_pages < 0)
		return -EINVAL;

262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
	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
287

288
	ctx->ring_pages = ctx->internal_pages;
Linus Torvalds's avatar
Linus Torvalds committed
289
	if (nr_pages > AIO_RING_PAGES) {
290
291
292
		ctx->ring_pages = kcalloc(nr_pages, sizeof(struct page *),
					  GFP_KERNEL);
		if (!ctx->ring_pages)
Linus Torvalds's avatar
Linus Torvalds committed
293
294
295
			return -ENOMEM;
	}

296
297
	ctx->mmap_size = nr_pages * PAGE_SIZE;
	pr_debug("attempting mmap of %lu bytes\n", ctx->mmap_size);
298

Zach Brown's avatar
Zach Brown committed
299
	down_write(&mm->mmap_sem);
300
301
302
	ctx->mmap_base = do_mmap_pgoff(ctx->aio_ring_file, 0, ctx->mmap_size,
				       PROT_READ | PROT_WRITE,
				       MAP_SHARED | MAP_POPULATE, 0, &populate);
303
	if (IS_ERR((void *)ctx->mmap_base)) {
Zach Brown's avatar
Zach Brown committed
304
		up_write(&mm->mmap_sem);
305
		ctx->mmap_size = 0;
Linus Torvalds's avatar
Linus Torvalds committed
306
307
308
		aio_free_ring(ctx);
		return -EAGAIN;
	}
309
310
311
	up_write(&mm->mmap_sem);

	mm_populate(ctx->mmap_base, populate);
Linus Torvalds's avatar
Linus Torvalds committed
312

313
314
315
	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);
316
317
	for (i = 0; i < ctx->nr_pages; i++)
		put_page(ctx->ring_pages[i]);
Linus Torvalds's avatar
Linus Torvalds committed
318

319
	if (unlikely(ctx->nr_pages != nr_pages)) {
Linus Torvalds's avatar
Linus Torvalds committed
320
321
322
323
		aio_free_ring(ctx);
		return -EAGAIN;
	}

324
325
	ctx->user_id = ctx->mmap_base;
	ctx->nr_events = nr_events; /* trusted copy */
Linus Torvalds's avatar
Linus Torvalds committed
326

327
	ring = kmap_atomic(ctx->ring_pages[0]);
Linus Torvalds's avatar
Linus Torvalds committed
328
329
330
331
332
333
334
	ring->nr = nr_events;	/* user copy */
	ring->id = ctx->user_id;
	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);
335
	kunmap_atomic(ring);
336
	flush_dcache_page(ctx->ring_pages[0]);
Linus Torvalds's avatar
Linus Torvalds committed
337
338
339
340
341
342
343
344

	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)

345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
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);

361
static int kiocb_cancel(struct kioctx *ctx, struct kiocb *kiocb)
Kent Overstreet's avatar
Kent Overstreet committed
362
{
363
	kiocb_cancel_fn *old, *cancel;
Kent Overstreet's avatar
Kent Overstreet committed
364
365
	int ret = -EINVAL;

366
367
368
369
370
371
372
373
374
	/*
	 * 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)
			return ret;
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

380
381
382
	atomic_inc(&kiocb->ki_users);
	spin_unlock_irq(&ctx->ctx_lock);

383
	ret = cancel(kiocb);
384
385

	spin_lock_irq(&ctx->ctx_lock);
386
	aio_put_req(kiocb);
Kent Overstreet's avatar
Kent Overstreet committed
387
388
389
390

	return ret;
}

Kent Overstreet's avatar
Kent Overstreet committed
391
392
393
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
394
395

	free_percpu(ctx->cpu);
Kent Overstreet's avatar
Kent Overstreet committed
396
397
398
399
400
401
402
403
	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
404
static void free_ioctx(struct work_struct *work)
Kent Overstreet's avatar
Kent Overstreet committed
405
{
Kent Overstreet's avatar
Kent Overstreet committed
406
	struct kioctx *ctx = container_of(work, struct kioctx, free_work);
407
	struct aio_ring *ring;
Kent Overstreet's avatar
Kent Overstreet committed
408
	struct kiocb *req;
409
410
	unsigned cpu, avail;
	DEFINE_WAIT(wait);
Kent Overstreet's avatar
Kent Overstreet committed
411
412
413
414
415
416
417
418

	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);
419
		kiocb_cancel(ctx, req);
Kent Overstreet's avatar
Kent Overstreet committed
420
421
422
423
	}

	spin_unlock_irq(&ctx->ctx_lock);

Kent Overstreet's avatar
Kent Overstreet committed
424
425
426
427
428
429
430
	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;
	}

431
432
	while (1) {
		prepare_to_wait(&ctx->wait, &wait, TASK_UNINTERRUPTIBLE);
433

434
435
436
437
		ring = kmap_atomic(ctx->ring_pages[0]);
		avail = (ring->head <= ring->tail)
			 ? ring->tail - ring->head
			 : ctx->nr_events - ring->head + ring->tail;
438

439
		atomic_add(avail, &ctx->reqs_available);
440
441
442
443
444
445
446
		ring->head = ring->tail;
		kunmap_atomic(ring);

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

		schedule();
447
	}
448
	finish_wait(&ctx->wait, &wait);
449

450
	WARN_ON(atomic_read(&ctx->reqs_available) > ctx->nr_events - 1);
Kent Overstreet's avatar
Kent Overstreet committed
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465

	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
466
static void free_ioctx_ref(struct percpu_ref *ref)
Kent Overstreet's avatar
Kent Overstreet committed
467
{
Kent Overstreet's avatar
Kent Overstreet committed
468
469
470
471
	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
472
473
}

Linus Torvalds's avatar
Linus Torvalds committed
474
475
476
477
478
/* 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
479
	struct mm_struct *mm = current->mm;
Linus Torvalds's avatar
Linus Torvalds committed
480
	struct kioctx *ctx;
481
	int err = -ENOMEM;
Linus Torvalds's avatar
Linus Torvalds committed
482

Kent Overstreet's avatar
Kent Overstreet committed
483
484
485
486
487
488
489
490
491
492
493
494
	/*
	 * 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
495
496
497
498
499
500
501
	/* 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);
	}

502
	if (!nr_events || (unsigned long)nr_events > aio_max_nr)
Linus Torvalds's avatar
Linus Torvalds committed
503
504
		return ERR_PTR(-EAGAIN);

505
	ctx = kmem_cache_zalloc(kioctx_cachep, GFP_KERNEL);
Linus Torvalds's avatar
Linus Torvalds committed
506
507
508
509
510
	if (!ctx)
		return ERR_PTR(-ENOMEM);

	ctx->max_reqs = nr_events;

Kent Overstreet's avatar
Kent Overstreet committed
511
512
513
	if (percpu_ref_init(&ctx->users, free_ioctx_ref))
		goto out_freectx;

Linus Torvalds's avatar
Linus Torvalds committed
514
	spin_lock_init(&ctx->ctx_lock);
515
	spin_lock_init(&ctx->completion_lock);
516
	mutex_init(&ctx->ring_lock);
Linus Torvalds's avatar
Linus Torvalds committed
517
518
519
520
	init_waitqueue_head(&ctx->wait);

	INIT_LIST_HEAD(&ctx->active_reqs);

Kent Overstreet's avatar
Kent Overstreet committed
521
522
	ctx->cpu = alloc_percpu(struct kioctx_cpu);
	if (!ctx->cpu)
Kent Overstreet's avatar
Kent Overstreet committed
523
		goto out_freeref;
Linus Torvalds's avatar
Linus Torvalds committed
524

Kent Overstreet's avatar
Kent Overstreet committed
525
526
527
	if (aio_setup_ring(ctx) < 0)
		goto out_freepcpu;

528
	atomic_set(&ctx->reqs_available, ctx->nr_events - 1);
Kent Overstreet's avatar
Kent Overstreet committed
529
530
	ctx->req_batch = (ctx->nr_events - 1) / (num_possible_cpus() * 4);
	BUG_ON(!ctx->req_batch);
531

Linus Torvalds's avatar
Linus Torvalds committed
532
	/* limit the number of system wide aios */
533
	spin_lock(&aio_nr_lock);
534
535
	if (aio_nr + nr_events > aio_max_nr ||
	    aio_nr + nr_events < aio_nr) {
536
		spin_unlock(&aio_nr_lock);
Linus Torvalds's avatar
Linus Torvalds committed
537
		goto out_cleanup;
538
539
	}
	aio_nr += ctx->max_reqs;
540
	spin_unlock(&aio_nr_lock);
Linus Torvalds's avatar
Linus Torvalds committed
541

Kent Overstreet's avatar
Kent Overstreet committed
542
543
	percpu_ref_get(&ctx->users); /* io_setup() will drop this ref */

Jeff Moyer's avatar
Jeff Moyer committed
544
	/* now link into global list. */
545
546
547
	spin_lock(&mm->ioctx_lock);
	hlist_add_head_rcu(&ctx->list, &mm->ioctx_list);
	spin_unlock(&mm->ioctx_lock);
Linus Torvalds's avatar
Linus Torvalds committed
548

Kent Overstreet's avatar
Kent Overstreet committed
549
	pr_debug("allocated ioctx %p[%ld]: mm=%p mask=0x%x\n",
550
		 ctx, ctx->user_id, mm, ctx->nr_events);
Linus Torvalds's avatar
Linus Torvalds committed
551
552
553
	return ctx;

out_cleanup:
554
555
	err = -EAGAIN;
	aio_free_ring(ctx);
Kent Overstreet's avatar
Kent Overstreet committed
556
557
out_freepcpu:
	free_percpu(ctx->cpu);
Kent Overstreet's avatar
Kent Overstreet committed
558
559
out_freeref:
	free_percpu(ctx->users.pcpu_count);
Linus Torvalds's avatar
Linus Torvalds committed
560
out_freectx:
561
562
	if (ctx->aio_ring_file)
		fput(ctx->aio_ring_file);
Linus Torvalds's avatar
Linus Torvalds committed
563
	kmem_cache_free(kioctx_cachep, ctx);
Kent Overstreet's avatar
Kent Overstreet committed
564
	pr_debug("error allocating ioctx %d\n", err);
565
	return ERR_PTR(err);
Linus Torvalds's avatar
Linus Torvalds committed
566
567
}

Kent Overstreet's avatar
Kent Overstreet committed
568
569
570
571
572
573
574
575
576
/* 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.
 */
static void kill_ioctx(struct kioctx *ctx)
{
	if (!atomic_xchg(&ctx->dead, 1)) {
		hlist_del_rcu(&ctx->list);
Kent Overstreet's avatar
Kent Overstreet committed
577
578
		/* percpu_ref_kill() will do the necessary call_rcu() */
		wake_up_all(&ctx->wait);
579

Kent Overstreet's avatar
Kent Overstreet committed
580
		/*
581
582
583
584
585
		 * 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
586
		 */
587
588
589
590
591
592
593
594
		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
595
		percpu_ref_kill(&ctx->users);
Kent Overstreet's avatar
Kent Overstreet committed
596
	}
Linus Torvalds's avatar
Linus Torvalds committed
597
598
599
600
601
}

/* wait_on_sync_kiocb:
 *	Waits on the given sync kiocb to complete.
 */
602
ssize_t wait_on_sync_kiocb(struct kiocb *iocb)
Linus Torvalds's avatar
Linus Torvalds committed
603
{
604
	while (atomic_read(&iocb->ki_users)) {
Linus Torvalds's avatar
Linus Torvalds committed
605
		set_current_state(TASK_UNINTERRUPTIBLE);
606
		if (!atomic_read(&iocb->ki_users))
Linus Torvalds's avatar
Linus Torvalds committed
607
			break;
608
		io_schedule();
Linus Torvalds's avatar
Linus Torvalds committed
609
610
611
612
	}
	__set_current_state(TASK_RUNNING);
	return iocb->ki_user_data;
}
613
EXPORT_SYMBOL(wait_on_sync_kiocb);
Linus Torvalds's avatar
Linus Torvalds committed
614

Kent Overstreet's avatar
Kent Overstreet committed
615
616
617
618
619
620
621
/*
 * 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
622
 */
623
void exit_aio(struct mm_struct *mm)
Linus Torvalds's avatar
Linus Torvalds committed
624
{
625
	struct kioctx *ctx;
Kent Overstreet's avatar
Kent Overstreet committed
626
	struct hlist_node *n;
627

Kent Overstreet's avatar
Kent Overstreet committed
628
	hlist_for_each_entry_safe(ctx, n, &mm->ioctx_list, list) {
629
630
631
632
633
634
635
636
		/*
		 * 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.
		 */
637
		ctx->mmap_size = 0;
Kent Overstreet's avatar
Kent Overstreet committed
638

639
		kill_ioctx(ctx);
Linus Torvalds's avatar
Linus Torvalds committed
640
641
642
	}
}

Kent Overstreet's avatar
Kent Overstreet committed
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
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
689
/* aio_get_req
690
 *	Allocate a slot for an aio request.  Increments the ki_users count
Linus Torvalds's avatar
Linus Torvalds committed
691
692
693
 * of the kioctx so that the kioctx stays around until all requests are
 * complete.  Returns NULL if no requests are free.
 *
694
 * Returns with kiocb->ki_users set to 2.  The io submit code path holds
Linus Torvalds's avatar
Linus Torvalds committed
695
696
697
698
 * an extra reference while submitting the i/o.
 * This prevents races between the aio code path referencing the
 * req (after submitting it) and aio_complete() freeing the req.
 */
Kent Overstreet's avatar
Kent Overstreet committed
699
static inline struct kiocb *aio_get_req(struct kioctx *ctx)
Linus Torvalds's avatar
Linus Torvalds committed
700
{
Kent Overstreet's avatar
Kent Overstreet committed
701
702
	struct kiocb *req;

Kent Overstreet's avatar
Kent Overstreet committed
703
	if (!get_reqs_available(ctx))
Kent Overstreet's avatar
Kent Overstreet committed
704
705
		return NULL;

706
	req = kmem_cache_alloc(kiocb_cachep, GFP_KERNEL|__GFP_ZERO);
Linus Torvalds's avatar
Linus Torvalds committed
707
	if (unlikely(!req))
Kent Overstreet's avatar
Kent Overstreet committed
708
		goto out_put;
Linus Torvalds's avatar
Linus Torvalds committed
709

710
	atomic_set(&req->ki_users, 1);
Linus Torvalds's avatar
Linus Torvalds committed
711
	req->ki_ctx = ctx;
Jeff Moyer's avatar
Jeff Moyer committed
712
	return req;
Kent Overstreet's avatar
Kent Overstreet committed
713
out_put:
Kent Overstreet's avatar
Kent Overstreet committed
714
	put_reqs_available(ctx, 1);
Kent Overstreet's avatar
Kent Overstreet committed
715
	return NULL;
Linus Torvalds's avatar
Linus Torvalds committed
716
717
}

718
static void kiocb_free(struct kiocb *req)
Linus Torvalds's avatar
Linus Torvalds committed
719
{
720
721
	if (req->ki_filp)
		fput(req->ki_filp);
722
723
	if (req->ki_eventfd != NULL)
		eventfd_ctx_put(req->ki_eventfd);
Linus Torvalds's avatar
Linus Torvalds committed
724
725
726
727
728
	if (req->ki_dtor)
		req->ki_dtor(req);
	kmem_cache_free(kiocb_cachep, req);
}

729
void aio_put_req(struct kiocb *req)
Linus Torvalds's avatar
Linus Torvalds committed
730
{
731
732
	if (atomic_dec_and_test(&req->ki_users))
		kiocb_free(req);
Linus Torvalds's avatar
Linus Torvalds committed
733
}
734
EXPORT_SYMBOL(aio_put_req);
Linus Torvalds's avatar
Linus Torvalds committed
735

736
static struct kioctx *lookup_ioctx(unsigned long ctx_id)
Linus Torvalds's avatar
Linus Torvalds committed
737
{
738
	struct mm_struct *mm = current->mm;
739
	struct kioctx *ctx, *ret = NULL;
Linus Torvalds's avatar
Linus Torvalds committed
740

741
742
	rcu_read_lock();

743
	hlist_for_each_entry_rcu(ctx, &mm->ioctx_list, list) {
Kent Overstreet's avatar
Kent Overstreet committed
744
		if (ctx->user_id == ctx_id) {
Kent Overstreet's avatar
Kent Overstreet committed
745
			percpu_ref_get(&ctx->users);
746
			ret = ctx;
Linus Torvalds's avatar
Linus Torvalds committed
747
748
			break;
		}
749
	}
Linus Torvalds's avatar
Linus Torvalds committed
750

751
	rcu_read_unlock();
752
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
753
754
755
756
757
}

/* aio_complete
 *	Called when the io request on the given iocb is complete.
 */
758
void aio_complete(struct kiocb *iocb, long res, long res2)
Linus Torvalds's avatar
Linus Torvalds committed
759
760
761
{
	struct kioctx	*ctx = iocb->ki_ctx;
	struct aio_ring	*ring;
Kent Overstreet's avatar
Kent Overstreet committed
762
	struct io_event	*ev_page, *event;
Linus Torvalds's avatar
Linus Torvalds committed
763
	unsigned long	flags;
Kent Overstreet's avatar
Kent Overstreet committed
764
	unsigned tail, pos;
Linus Torvalds's avatar
Linus Torvalds committed
765

766
767
768
769
770
771
	/*
	 * 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
772
773
	 */
	if (is_sync_kiocb(iocb)) {
774
		BUG_ON(atomic_read(&iocb->ki_users) != 1);
Linus Torvalds's avatar
Linus Torvalds committed
775
		iocb->ki_user_data = res;
776
		atomic_set(&iocb->ki_users, 0);
Linus Torvalds's avatar
Linus Torvalds committed
777
		wake_up_process(iocb->ki_obj.tsk);
778
		return;
Linus Torvalds's avatar
Linus Torvalds committed
779
780
	}

Kent Overstreet's avatar
Kent Overstreet committed
781
782
	/*
	 * Take rcu_read_lock() in case the kioctx is being destroyed, as we
783
	 * need to issue a wakeup after incrementing reqs_available.
Linus Torvalds's avatar
Linus Torvalds committed
784
	 */
Kent Overstreet's avatar
Kent Overstreet committed
785
	rcu_read_lock();
Linus Torvalds's avatar
Linus Torvalds committed
786

787
788
789
790
791
792
793
	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);
	}
794

795
796
	/*
	 * Add a completion event to the ring buffer. Must be done holding
797
	 * ctx->completion_lock to prevent other code from messing with the tail
798
799
800
801
	 * pointer since we might be called from irq context.
	 */
	spin_lock_irqsave(&ctx->completion_lock, flags);

802
	tail = ctx->tail;
Kent Overstreet's avatar
Kent Overstreet committed
803
804
	pos = tail + AIO_EVENTS_OFFSET;

805
	if (++tail >= ctx->nr_events)
806
		tail = 0;
Linus Torvalds's avatar
Linus Torvalds committed
807

808
	ev_page = kmap_atomic(ctx->ring_pages[pos / AIO_EVENTS_PER_PAGE]);
Kent Overstreet's avatar
Kent Overstreet committed
809
810
	event = ev_page + pos % AIO_EVENTS_PER_PAGE;

Linus Torvalds's avatar
Linus Torvalds committed
811
812
813
814
815
	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
816
	kunmap_atomic(ev_page);
817
	flush_dcache_page(ctx->ring_pages[pos / AIO_EVENTS_PER_PAGE]);
Kent Overstreet's avatar
Kent Overstreet committed
818
819

	pr_debug("%p[%u]: %p: %p %Lx %lx %lx\n",
Kent Overstreet's avatar
Kent Overstreet committed
820
821
		 ctx, tail, iocb, iocb->ki_obj.user, iocb->ki_user_data,
		 res, res2);
Linus Torvalds's avatar
Linus Torvalds committed
822
823
824
825
826
827

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

828
	ctx->tail = tail;
Linus Torvalds's avatar
Linus Torvalds committed
829

830
	ring = kmap_atomic(ctx->ring_pages[0]);
Kent Overstreet's avatar
Kent Overstreet committed
831
	ring->tail = tail;
832
	kunmap_atomic(ring);
833
	flush_dcache_page(ctx->ring_pages[0]);
Linus Torvalds's avatar
Linus Torvalds committed
834

835
836
	spin_unlock_irqrestore(&ctx->completion_lock, flags);

Kent Overstreet's avatar
Kent Overstreet committed
837
	pr_debug("added to ring %p at [%u]\n", iocb, tail);
Davide Libenzi's avatar
Davide Libenzi committed
838
839
840
841
842
843

	/*
	 * 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.
	 */
844
	if (iocb->ki_eventfd != NULL)
Davide Libenzi's avatar
Davide Libenzi committed
845
846
		eventfd_signal(iocb->ki_eventfd, 1);

Linus Torvalds's avatar
Linus Torvalds committed
847
	/* everything turned out well, dispose of the aiocb. */
848
	aio_put_req(iocb);
Linus Torvalds's avatar
Linus Torvalds committed
849

850
851
852
853
854
855
856
857
	/*
	 * 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
858
859
860
	if (waitqueue_active(&ctx->wait))
		wake_up(&ctx->wait);

Kent Overstreet's avatar
Kent Overstreet committed
861
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
862
}
863
EXPORT_SYMBOL(aio_complete);
Linus Torvalds's avatar
Linus Torvalds committed
864

865
866
867
/* 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
868
 */
869
870
static long aio_read_events_ring(struct kioctx *ctx,
				 struct io_event __user *event, long nr)
Linus Torvalds's avatar
Linus Torvalds committed
871
872
{
	struct aio_ring *ring;
873
	unsigned head, tail, pos;
874
875
876
	long ret = 0;
	int copy_ret;

877
	mutex_lock(&ctx->ring_lock);
Linus Torvalds's avatar
Linus Torvalds committed
878

879
	ring = kmap_atomic(ctx->ring_pages[0]);
880
	head = ring->head;
881
	tail = ring->tail;
882
883
	kunmap_atomic(ring);

884
	pr_debug("h%u t%u m%u\n", head, tail, ctx->nr_events);
Linus Torvalds's avatar
Linus Torvalds committed
885

886
	if (head == tail)
Linus Torvalds's avatar
Linus Torvalds committed
887
888
		goto out;

889
890
891
892
893
	while (ret < nr) {
		long avail;
		struct io_event *ev;
		struct page *page;

894
895
		avail = (head <= tail ?  tail : ctx->nr_events) - head;
		if (head == tail)
896
897
898
899
900
901
902
			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;
903
		page = ctx->ring_pages[pos / AIO_EVENTS_PER_PAGE];
904
905
906
907
908
909
910
911
912
913
914
915
916
917
		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;
918
		head %= ctx->nr_events;
Linus Torvalds's avatar
Linus Torvalds committed
919
920
	}

921
	ring = kmap_atomic(ctx->ring_pages[0]);
922
	ring->head = head;
923
	kunmap_atomic(ring);
924
	flush_dcache_page(ctx->ring_pages[0]);
925

926
	pr_debug("%li  h%u t%u\n", ret, head, tail);
927

Kent Overstreet's avatar
Kent Overstreet committed
928
	put_reqs_available(ctx, ret);
929
out:
930
	mutex_unlock(&ctx->ring_lock);
931

Linus Torvalds's avatar
Linus Torvalds committed
932
933
934
	return ret;
}

935
936
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
937
{
938
	long ret = aio_read_events_ring(ctx, event + *i, nr - *i);
Linus Torvalds's avatar
Linus Torvalds committed
939

940
941
	if (ret > 0)
		*i += ret;
Linus Torvalds's avatar
Linus Torvalds committed
942

943
944
	if (unlikely(atomic_read(&ctx->dead)))
		ret = -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
945

946
947
	if (!*i)
		*i = ret;
Linus Torvalds's avatar
Linus Torvalds committed
948

949
	return ret < 0 || *i >= min_nr;
Linus Torvalds's avatar
Linus Torvalds committed
950
951
}

952
static long read_events(struct kioctx *ctx, long min_nr, long nr,
Linus Torvalds's avatar
Linus Torvalds committed
953
954
955
			struct io_event __user *event,
			struct timespec __user *timeout)
{
956
957
	ktime_t until = { .tv64 = KTIME_MAX };
	long ret = 0;
Linus Torvalds's avatar
Linus Torvalds committed
958
959
960

	if (timeout) {
		struct timespec	ts;
961

Linus Torvalds's avatar
Linus Torvalds committed
962
		if (unlikely(copy_from_user(&ts, timeout, sizeof(ts))))
963
			return -EFAULT;
Linus Torvalds's avatar
Linus Torvalds committed
964

965
		until = timespec_to_ktime(ts);
Linus Torvalds's avatar
Linus Torvalds committed
966
967
	}

968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
	/*
	 * 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
984

985
986
	if (!ret && signal_pending(current))
		ret = -EINTR;
Linus Torvalds's avatar
Linus Torvalds committed
987

988
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
}

/* 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.
 */
1004
SYSCALL_DEFINE2(io_setup, unsigned, nr_events, aio_context_t __user *, ctxp)
Linus Torvalds's avatar
Linus Torvalds committed
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
{
	struct kioctx *ioctx = NULL;
	unsigned long ctx;
	long ret;

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

	ret = -EINVAL;
1015
1016
1017
	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
1018
1019
1020
1021
1022
1023
1024
		goto out;
	}

	ioctx = ioctx_alloc(nr_events);
	ret = PTR_ERR(ioctx);
	if (!IS_ERR(ioctx)) {
		ret = put_user(ioctx->user_id, ctxp);
1025
		if (ret)
Kent Overstreet's avatar
Kent Overstreet committed
1026
			kill_ioctx(ioctx);
Kent Overstreet's avatar
Kent Overstreet committed
1027
		percpu_ref_put(&ioctx->users);
Linus Torvalds's avatar
Linus Torvalds committed
1028
1029
1030
1031
1032
1033
1034
1035
1036
	}

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
1037
 *	implemented.  May fail with -EINVAL if the context pointed to
Linus Torvalds's avatar
Linus Torvalds committed
1038
1039
 *	is invalid.
 */
1040
SYSCALL_DEFINE1(io_destroy, aio_context_t, ctx)
Linus Torvalds's avatar
Linus Torvalds committed
1041
1042
1043
{
	struct kioctx *ioctx = lookup_ioctx(ctx);
	if (likely(NULL != ioctx)) {
Kent Overstreet's avatar
Kent Overstreet committed
1044
		kill_ioctx(ioctx);
Kent Overstreet's avatar
Kent Overstreet committed
1045
		percpu_ref_put(&ioctx->users);
Linus Torvalds's avatar
Linus Torvalds committed
1046
1047
1048
1049
1050
1051
		return 0;
	}
	pr_debug("EINVAL: io_destroy: invalid context id\n");
	return -EINVAL;
}

Kent Overstreet's avatar
Kent Overstreet committed
1052
1053
1054
typedef ssize_t (aio_rw_op)(struct kiocb *, const struct iovec *,
			    unsigned long, loff_t);

1055
1056
1057
1058
1059
static ssize_t aio_setup_vectored_rw(struct kiocb *kiocb,
				     int rw, char __user *buf,
				     unsigned long *nr_segs,
				     struct iovec **iovec,
				     bool compat)
1060
1061
1062
{
	ssize_t ret;

1063
	*nr_segs = kiocb->ki_nbytes;
Kent Overstreet's avatar
Kent Overstreet committed
1064

1065
1066
#ifdef CONFIG_COMPAT
	if (compat)
Kent Overstreet's avatar
Kent Overstreet committed
1067
		ret = compat_rw_copy_check_uvector(rw,
1068
1069
				(struct compat_iovec __user *)buf,
				*nr_segs, 1, *iovec, iovec);
1070
1071
	else
#endif
Kent Overstreet's avatar
Kent Overstreet committed
1072
		ret = rw_copy_check_uvector(rw,
1073
1074
				(struct iovec __user *)buf,
				*nr_segs, 1, *iovec, iovec);
1075
	if (ret < 0)
Kent Overstreet's avatar
Kent Overstreet committed
1076
		return ret;
1077

Kent Overstreet's avatar
Kent Overstreet committed
1078
	/* ki_nbytes now reflect bytes instead of segs */
1079
	kiocb->ki_nbytes = ret;
Kent Overstreet's avatar
Kent Overstreet committed
1080
	return 0;
1081
1082
}

1083
1084
1085
1086
static ssize_t aio_setup_single_vector(struct kiocb *kiocb,
				       int rw, char __user *buf,
				       unsigned long *nr_segs,
				       struct iovec *iovec)
1087
{
1088
	if (unlikely(!access_ok(!rw, buf, kiocb->ki_nbytes)))
Kent Overstreet's avatar
Kent Overstreet committed
1089
		return -EFAULT;
1090

1091
1092
1093
	iovec->iov_base = buf;
	iovec->iov_len = kiocb->ki_nbytes;
	*nr_segs = 1;
1094
1095
1096
	return 0;
}

Linus Torvalds's avatar
Linus Torvalds committed
1097
1098
1099
1100
1101
/*
 * aio_setup_iocb:
 *	Performs the initial checks and aio retry method
 *	setup for the kiocb at the time of io submission.
 */
1102
1103
static ssize_t aio_run_iocb(struct kiocb *req, unsigned opcode,
			    char __user *buf, bool compat)
Linus Torvalds's avatar
Linus Torvalds committed
1104
{
Kent Overstreet's avatar
Kent Overstreet committed
1105
1106
	struct file *file = req->ki_filp;
	ssize_t ret;
1107
	unsigned long nr_segs;
Kent Overstreet's avatar
Kent Overstreet committed
1108
1109
1110
	int rw;
	fmode_t mode;
	aio_rw_op *rw_op;
1111
	struct iovec inline_vec, *iovec = &inline_vec;
Linus Torvalds's avatar
Linus Torvalds committed
1112

1113
	switch (opcode) {
Linus Torvalds's avatar
Linus Torvalds committed
1114
	case IOCB_CMD_PREAD:
1115
	case IOCB_CMD_PREADV:
Kent Overstreet's avatar
Kent Overstreet committed
1116
1117
1118
1119
1120
1121
		mode	= FMODE_READ;
		rw	= READ;
		rw_op	= file->f_op->aio_read;
		goto rw_common;

	case IOCB_CMD_PWRITE:
1122
	case IOCB_CMD_PWRITEV:
Kent Overstreet's avatar
Kent Overstreet committed
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
		mode	= FMODE_WRITE;
		rw	= WRITE;
		rw_op	= file->f_op->aio_write;
		goto rw_common;
rw_common:
		if (unlikely(!(file->f_mode & mode)))
			return -EBADF;

		if (!rw_op)
			return -EINVAL;

1134
1135
1136
1137
1138
1139
		ret = (opcode == IOCB_CMD_PREADV ||
		       opcode == IOCB_CMD_PWRITEV)
			? aio_setup_vectored_rw(req, rw, buf, &nr_segs,
						&iovec, compat)
			: aio_setup_single_vector(req, rw, buf, &nr_segs,
						  iovec);
1140
		if (ret)
Kent Overstreet's avatar
Kent Overstreet committed
1141
1142
1143
			return ret;

		ret = rw_verify_area(rw, file, &req->ki_pos, req->ki_nbytes);
1144
1145
1146
		if (ret < 0) {
			if (iovec != &inline_vec)
				kfree(iovec);
Kent Overstreet's avatar
Kent Overstreet committed
1147
			return ret;
1148
		}
Kent Overstreet's avatar
Kent Overstreet committed
1149
1150
1151

		req->ki_nbytes = ret;

1152
1153
1154
1155
1156
1157