aio.c 34 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>
Linus Torvalds's avatar
Linus Torvalds committed
29
30
31
32
33
34
#include <linux/slab.h>
#include <linux/timer.h>
#include <linux/aio.h>
#include <linux/highmem.h>
#include <linux/workqueue.h>
#include <linux/security.h>
35
#include <linux/eventfd.h>
Jeff Moyer's avatar
Jeff Moyer committed
36
#include <linux/blkdev.h>
37
#include <linux/compat.h>
Linus Torvalds's avatar
Linus Torvalds committed
38
39
40
41

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

42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
#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

struct kioctx {
	atomic_t		users;
Kent Overstreet's avatar
Kent Overstreet committed
64
	atomic_t		dead;
65
66
67
68
69
70
71
72
73

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

	wait_queue_head_t	wait;

	spinlock_t		ctx_lock;

74
	atomic_t		reqs_active;
75
76
	struct list_head	active_reqs;	/* used for cancellation */

77
78
79
80
	/*
	 * This is what userspace passed to io_setup(), it's not used for
	 * anything but counting against the global max_reqs quota.
	 *
81
	 * The real limit is nr_events - 1, which will be larger (see
82
83
	 * aio_setup_ring())
	 */
84
85
	unsigned		max_reqs;

86
87
	/* Size of ringbuffer, in units of struct io_event */
	unsigned		nr_events;
88

89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
	unsigned long		mmap_base;
	unsigned long		mmap_size;

	struct page		**ring_pages;
	long			nr_pages;

	struct {
		struct mutex	ring_lock;
	} ____cacheline_aligned;

	struct {
		unsigned	tail;
		spinlock_t	completion_lock;
	} ____cacheline_aligned;

	struct page		*internal_pages[AIO_RING_PAGES];
105

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

Linus Torvalds's avatar
Linus Torvalds committed
110
/*------ sysctl variables----*/
111
112
113
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
114
115
/*----end sysctl variables---*/

116
117
static struct kmem_cache	*kiocb_cachep;
static struct kmem_cache	*kioctx_cachep;
Linus Torvalds's avatar
Linus Torvalds committed
118
119
120
121
122
123
124

/* 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)
{
125
126
	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
127

Kent Overstreet's avatar
Kent Overstreet committed
128
	pr_debug("sizeof(struct page) = %zu\n", sizeof(struct page));
Linus Torvalds's avatar
Linus Torvalds committed
129
130
131

	return 0;
}
132
__initcall(aio_setup);
Linus Torvalds's avatar
Linus Torvalds committed
133
134
135
136
137

static void aio_free_ring(struct kioctx *ctx)
{
	long i;

138
139
	for (i = 0; i < ctx->nr_pages; i++)
		put_page(ctx->ring_pages[i]);
Linus Torvalds's avatar
Linus Torvalds committed
140

141
142
	if (ctx->mmap_size)
		vm_munmap(ctx->mmap_base, ctx->mmap_size);
Linus Torvalds's avatar
Linus Torvalds committed
143

144
145
	if (ctx->ring_pages && ctx->ring_pages != ctx->internal_pages)
		kfree(ctx->ring_pages);
Linus Torvalds's avatar
Linus Torvalds committed
146
147
148
149
150
151
}

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
152
	struct mm_struct *mm = current->mm;
153
	unsigned long size, populate;
Linus Torvalds's avatar
Linus Torvalds committed
154
155
156
157
158
159
160
161
162
163
164
165
166
167
	int nr_pages;

	/* 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;
	nr_pages = (size + PAGE_SIZE-1) >> PAGE_SHIFT;

	if (nr_pages < 0)
		return -EINVAL;

	nr_events = (PAGE_SIZE * nr_pages - sizeof(struct aio_ring)) / sizeof(struct io_event);

168
169
	ctx->nr_events = 0;
	ctx->ring_pages = ctx->internal_pages;
Linus Torvalds's avatar
Linus Torvalds committed
170
	if (nr_pages > AIO_RING_PAGES) {
171
172
173
		ctx->ring_pages = kcalloc(nr_pages, sizeof(struct page *),
					  GFP_KERNEL);
		if (!ctx->ring_pages)
Linus Torvalds's avatar
Linus Torvalds committed
174
175
176
			return -ENOMEM;
	}

177
178
	ctx->mmap_size = nr_pages * PAGE_SIZE;
	pr_debug("attempting mmap of %lu bytes\n", ctx->mmap_size);
Zach Brown's avatar
Zach Brown committed
179
	down_write(&mm->mmap_sem);
180
181
182
183
	ctx->mmap_base = do_mmap_pgoff(NULL, 0, ctx->mmap_size,
				       PROT_READ|PROT_WRITE,
				       MAP_ANONYMOUS|MAP_PRIVATE, 0, &populate);
	if (IS_ERR((void *)ctx->mmap_base)) {
Zach Brown's avatar
Zach Brown committed
184
		up_write(&mm->mmap_sem);
185
		ctx->mmap_size = 0;
Linus Torvalds's avatar
Linus Torvalds committed
186
187
188
189
		aio_free_ring(ctx);
		return -EAGAIN;
	}

190
191
192
	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);
Zach Brown's avatar
Zach Brown committed
193
	up_write(&mm->mmap_sem);
Linus Torvalds's avatar
Linus Torvalds committed
194

195
	if (unlikely(ctx->nr_pages != nr_pages)) {
Linus Torvalds's avatar
Linus Torvalds committed
196
197
198
		aio_free_ring(ctx);
		return -EAGAIN;
	}
199
	if (populate)
200
		mm_populate(ctx->mmap_base, populate);
Linus Torvalds's avatar
Linus Torvalds committed
201

202
203
	ctx->user_id = ctx->mmap_base;
	ctx->nr_events = nr_events; /* trusted copy */
Linus Torvalds's avatar
Linus Torvalds committed
204

205
	ring = kmap_atomic(ctx->ring_pages[0]);
Linus Torvalds's avatar
Linus Torvalds committed
206
207
208
209
210
211
212
	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);
213
	kunmap_atomic(ring);
214
	flush_dcache_page(ctx->ring_pages[0]);
Linus Torvalds's avatar
Linus Torvalds committed
215
216
217
218
219
220
221
222

	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)

223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
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);

Kent Overstreet's avatar
Kent Overstreet committed
239
240
241
static int kiocb_cancel(struct kioctx *ctx, struct kiocb *kiocb,
			struct io_event *res)
{
242
	kiocb_cancel_fn *old, *cancel;
Kent Overstreet's avatar
Kent Overstreet committed
243
244
	int ret = -EINVAL;

245
246
247
248
249
250
251
252
253
	/*
	 * 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
254

255
256
257
		old = cancel;
		cancel = cmpxchg(&kiocb->ki_cancel, old, KIOCB_CANCELLED);
	} while (cancel != old);
Kent Overstreet's avatar
Kent Overstreet committed
258

259
260
261
262
263
264
265
266
267
	atomic_inc(&kiocb->ki_users);
	spin_unlock_irq(&ctx->ctx_lock);

	memset(res, 0, sizeof(*res));
	res->obj = (u64)(unsigned long)kiocb->ki_obj.user;
	res->data = kiocb->ki_user_data;
	ret = cancel(kiocb, res);

	spin_lock_irq(&ctx->ctx_lock);
Kent Overstreet's avatar
Kent Overstreet committed
268
269
270
271

	return ret;
}

Kent Overstreet's avatar
Kent Overstreet committed
272
273
274
275
276
277
278
279
280
281
282
283
284
static void free_ioctx_rcu(struct rcu_head *head)
{
	struct kioctx *ctx = container_of(head, struct kioctx, rcu_head);
	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.
 */
static void free_ioctx(struct kioctx *ctx)
{
285
	struct aio_ring *ring;
Kent Overstreet's avatar
Kent Overstreet committed
286
287
	struct io_event res;
	struct kiocb *req;
288
	unsigned head, avail;
Kent Overstreet's avatar
Kent Overstreet committed
289
290
291
292
293
294
295
296
297
298
299
300
301

	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);
		kiocb_cancel(ctx, req, &res);
	}

	spin_unlock_irq(&ctx->ctx_lock);

302
	ring = kmap_atomic(ctx->ring_pages[0]);
303
304
305
306
	head = ring->head;
	kunmap_atomic(ring);

	while (atomic_read(&ctx->reqs_active) > 0) {
307
		wait_event(ctx->wait, head != ctx->tail);
308

309
		avail = (head <= ctx->tail ? ctx->tail : ctx->nr_events) - head;
310
311
312

		atomic_sub(avail, &ctx->reqs_active);
		head += avail;
313
		head %= ctx->nr_events;
314
315
316
	}

	WARN_ON(atomic_read(&ctx->reqs_active) < 0);
Kent Overstreet's avatar
Kent Overstreet committed
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342

	aio_free_ring(ctx);

	spin_lock(&aio_nr_lock);
	BUG_ON(aio_nr - ctx->max_reqs > aio_nr);
	aio_nr -= ctx->max_reqs;
	spin_unlock(&aio_nr_lock);

	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);
}

static void put_ioctx(struct kioctx *ctx)
{
	if (unlikely(atomic_dec_and_test(&ctx->users)))
		free_ioctx(ctx);
}

Linus Torvalds's avatar
Linus Torvalds committed
343
344
345
346
347
/* 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
348
	struct mm_struct *mm = current->mm;
Linus Torvalds's avatar
Linus Torvalds committed
349
	struct kioctx *ctx;
350
	int err = -ENOMEM;
Linus Torvalds's avatar
Linus Torvalds committed
351
352
353
354
355
356
357
358

	/* 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);
	}

359
	if (!nr_events || (unsigned long)nr_events > aio_max_nr)
Linus Torvalds's avatar
Linus Torvalds committed
360
361
		return ERR_PTR(-EAGAIN);

362
	ctx = kmem_cache_zalloc(kioctx_cachep, GFP_KERNEL);
Linus Torvalds's avatar
Linus Torvalds committed
363
364
365
366
367
	if (!ctx)
		return ERR_PTR(-ENOMEM);

	ctx->max_reqs = nr_events;

Al Viro's avatar
Al Viro committed
368
	atomic_set(&ctx->users, 2);
Kent Overstreet's avatar
Kent Overstreet committed
369
	atomic_set(&ctx->dead, 0);
Linus Torvalds's avatar
Linus Torvalds committed
370
	spin_lock_init(&ctx->ctx_lock);
371
	spin_lock_init(&ctx->completion_lock);
372
	mutex_init(&ctx->ring_lock);
Linus Torvalds's avatar
Linus Torvalds committed
373
374
375
376
377
378
379
380
	init_waitqueue_head(&ctx->wait);

	INIT_LIST_HEAD(&ctx->active_reqs);

	if (aio_setup_ring(ctx) < 0)
		goto out_freectx;

	/* limit the number of system wide aios */
381
	spin_lock(&aio_nr_lock);
382
383
	if (aio_nr + nr_events > aio_max_nr ||
	    aio_nr + nr_events < aio_nr) {
384
		spin_unlock(&aio_nr_lock);
Linus Torvalds's avatar
Linus Torvalds committed
385
		goto out_cleanup;
386
387
	}
	aio_nr += ctx->max_reqs;
388
	spin_unlock(&aio_nr_lock);
Linus Torvalds's avatar
Linus Torvalds committed
389

Jeff Moyer's avatar
Jeff Moyer committed
390
	/* now link into global list. */
391
392
393
	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
394

Kent Overstreet's avatar
Kent Overstreet committed
395
	pr_debug("allocated ioctx %p[%ld]: mm=%p mask=0x%x\n",
396
		 ctx, ctx->user_id, mm, ctx->nr_events);
Linus Torvalds's avatar
Linus Torvalds committed
397
398
399
	return ctx;

out_cleanup:
400
401
	err = -EAGAIN;
	aio_free_ring(ctx);
Linus Torvalds's avatar
Linus Torvalds committed
402
403
out_freectx:
	kmem_cache_free(kioctx_cachep, ctx);
Kent Overstreet's avatar
Kent Overstreet committed
404
	pr_debug("error allocating ioctx %d\n", err);
405
	return ERR_PTR(err);
Linus Torvalds's avatar
Linus Torvalds committed
406
407
}

Kent Overstreet's avatar
Kent Overstreet committed
408
static void kill_ioctx_work(struct work_struct *work)
Linus Torvalds's avatar
Linus Torvalds committed
409
{
Kent Overstreet's avatar
Kent Overstreet committed
410
	struct kioctx *ctx = container_of(work, struct kioctx, rcu_work);
411

Kent Overstreet's avatar
Kent Overstreet committed
412
413
414
	wake_up_all(&ctx->wait);
	put_ioctx(ctx);
}
Kent Overstreet's avatar
Kent Overstreet committed
415

Kent Overstreet's avatar
Kent Overstreet committed
416
417
418
static void kill_ioctx_rcu(struct rcu_head *head)
{
	struct kioctx *ctx = container_of(head, struct kioctx, rcu_head);
Linus Torvalds's avatar
Linus Torvalds committed
419

Kent Overstreet's avatar
Kent Overstreet committed
420
421
422
	INIT_WORK(&ctx->rcu_work, kill_ioctx_work);
	schedule_work(&ctx->rcu_work);
}
Linus Torvalds's avatar
Linus Torvalds committed
423

Kent Overstreet's avatar
Kent Overstreet committed
424
425
426
427
428
429
430
431
432
433
434
/* 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);
		/* Between hlist_del_rcu() and dropping the initial ref */
		synchronize_rcu();
435

Kent Overstreet's avatar
Kent Overstreet committed
436
437
438
439
440
441
442
443
444
		/*
		 * We can't punt to workqueue here because put_ioctx() ->
		 * free_ioctx() will unmap the ringbuffer, and that has to be
		 * done in the original process's context. kill_ioctx_rcu/work()
		 * exist for exit_aio(), as in that path free_ioctx() won't do
		 * the unmap.
		 */
		kill_ioctx_work(&ctx->rcu_work);
	}
Linus Torvalds's avatar
Linus Torvalds committed
445
446
447
448
449
}

/* wait_on_sync_kiocb:
 *	Waits on the given sync kiocb to complete.
 */
450
ssize_t wait_on_sync_kiocb(struct kiocb *iocb)
Linus Torvalds's avatar
Linus Torvalds committed
451
{
452
	while (atomic_read(&iocb->ki_users)) {
Linus Torvalds's avatar
Linus Torvalds committed
453
		set_current_state(TASK_UNINTERRUPTIBLE);
454
		if (!atomic_read(&iocb->ki_users))
Linus Torvalds's avatar
Linus Torvalds committed
455
			break;
456
		io_schedule();
Linus Torvalds's avatar
Linus Torvalds committed
457
458
459
460
	}
	__set_current_state(TASK_RUNNING);
	return iocb->ki_user_data;
}
461
EXPORT_SYMBOL(wait_on_sync_kiocb);
Linus Torvalds's avatar
Linus Torvalds committed
462

Kent Overstreet's avatar
Kent Overstreet committed
463
464
465
466
467
468
469
/*
 * 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
470
 */
471
void exit_aio(struct mm_struct *mm)
Linus Torvalds's avatar
Linus Torvalds committed
472
{
473
	struct kioctx *ctx;
Kent Overstreet's avatar
Kent Overstreet committed
474
	struct hlist_node *n;
475

Kent Overstreet's avatar
Kent Overstreet committed
476
	hlist_for_each_entry_safe(ctx, n, &mm->ioctx_list, list) {
Linus Torvalds's avatar
Linus Torvalds committed
477
478
479
		if (1 != atomic_read(&ctx->users))
			printk(KERN_DEBUG
				"exit_aio:ioctx still alive: %d %d %d\n",
Kent Overstreet's avatar
Kent Overstreet committed
480
481
				atomic_read(&ctx->users),
				atomic_read(&ctx->dead),
482
				atomic_read(&ctx->reqs_active));
483
484
485
486
487
488
489
490
		/*
		 * 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.
		 */
491
		ctx->mmap_size = 0;
Kent Overstreet's avatar
Kent Overstreet committed
492
493
494
495
496

		if (!atomic_xchg(&ctx->dead, 1)) {
			hlist_del_rcu(&ctx->list);
			call_rcu(&ctx->rcu_head, kill_ioctx_rcu);
		}
Linus Torvalds's avatar
Linus Torvalds committed
497
498
499
500
	}
}

/* aio_get_req
501
 *	Allocate a slot for an aio request.  Increments the ki_users count
Linus Torvalds's avatar
Linus Torvalds committed
502
503
504
 * of the kioctx so that the kioctx stays around until all requests are
 * complete.  Returns NULL if no requests are free.
 *
505
 * Returns with kiocb->ki_users set to 2.  The io submit code path holds
Linus Torvalds's avatar
Linus Torvalds committed
506
507
508
509
 * 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
510
static inline struct kiocb *aio_get_req(struct kioctx *ctx)
Linus Torvalds's avatar
Linus Torvalds committed
511
{
Kent Overstreet's avatar
Kent Overstreet committed
512
513
	struct kiocb *req;

514
	if (atomic_read(&ctx->reqs_active) >= ctx->nr_events)
Kent Overstreet's avatar
Kent Overstreet committed
515
516
		return NULL;

517
	if (atomic_inc_return(&ctx->reqs_active) > ctx->nr_events - 1)
Kent Overstreet's avatar
Kent Overstreet committed
518
		goto out_put;
Linus Torvalds's avatar
Linus Torvalds committed
519

520
	req = kmem_cache_alloc(kiocb_cachep, GFP_KERNEL|__GFP_ZERO);
Linus Torvalds's avatar
Linus Torvalds committed
521
	if (unlikely(!req))
Kent Overstreet's avatar
Kent Overstreet committed
522
		goto out_put;
Linus Torvalds's avatar
Linus Torvalds committed
523

524
	atomic_set(&req->ki_users, 2);
Linus Torvalds's avatar
Linus Torvalds committed
525
526
	req->ki_ctx = ctx;

Jeff Moyer's avatar
Jeff Moyer committed
527
	return req;
Kent Overstreet's avatar
Kent Overstreet committed
528
529
530
out_put:
	atomic_dec(&ctx->reqs_active);
	return NULL;
Linus Torvalds's avatar
Linus Torvalds committed
531
532
}

533
static void kiocb_free(struct kiocb *req)
Linus Torvalds's avatar
Linus Torvalds committed
534
{
535
536
	if (req->ki_filp)
		fput(req->ki_filp);
537
538
	if (req->ki_eventfd != NULL)
		eventfd_ctx_put(req->ki_eventfd);
Linus Torvalds's avatar
Linus Torvalds committed
539
540
	if (req->ki_dtor)
		req->ki_dtor(req);
541
542
	if (req->ki_iovec != &req->ki_inline_vec)
		kfree(req->ki_iovec);
Linus Torvalds's avatar
Linus Torvalds committed
543
544
545
	kmem_cache_free(kiocb_cachep, req);
}

546
void aio_put_req(struct kiocb *req)
Linus Torvalds's avatar
Linus Torvalds committed
547
{
548
549
	if (atomic_dec_and_test(&req->ki_users))
		kiocb_free(req);
Linus Torvalds's avatar
Linus Torvalds committed
550
}
551
EXPORT_SYMBOL(aio_put_req);
Linus Torvalds's avatar
Linus Torvalds committed
552

553
static struct kioctx *lookup_ioctx(unsigned long ctx_id)
Linus Torvalds's avatar
Linus Torvalds committed
554
{
555
	struct mm_struct *mm = current->mm;
556
	struct kioctx *ctx, *ret = NULL;
Linus Torvalds's avatar
Linus Torvalds committed
557

558
559
	rcu_read_lock();

560
	hlist_for_each_entry_rcu(ctx, &mm->ioctx_list, list) {
Kent Overstreet's avatar
Kent Overstreet committed
561
562
		if (ctx->user_id == ctx_id) {
			atomic_inc(&ctx->users);
563
			ret = ctx;
Linus Torvalds's avatar
Linus Torvalds committed
564
565
			break;
		}
566
	}
Linus Torvalds's avatar
Linus Torvalds committed
567

568
	rcu_read_unlock();
569
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
570
571
572
573
574
}

/* aio_complete
 *	Called when the io request on the given iocb is complete.
 */
575
void aio_complete(struct kiocb *iocb, long res, long res2)
Linus Torvalds's avatar
Linus Torvalds committed
576
577
578
{
	struct kioctx	*ctx = iocb->ki_ctx;
	struct aio_ring	*ring;
Kent Overstreet's avatar
Kent Overstreet committed
579
	struct io_event	*ev_page, *event;
Linus Torvalds's avatar
Linus Torvalds committed
580
	unsigned long	flags;
Kent Overstreet's avatar
Kent Overstreet committed
581
	unsigned tail, pos;
Linus Torvalds's avatar
Linus Torvalds committed
582

583
584
585
586
587
588
	/*
	 * 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
589
590
	 */
	if (is_sync_kiocb(iocb)) {
591
		BUG_ON(atomic_read(&iocb->ki_users) != 1);
Linus Torvalds's avatar
Linus Torvalds committed
592
		iocb->ki_user_data = res;
593
		atomic_set(&iocb->ki_users, 0);
Linus Torvalds's avatar
Linus Torvalds committed
594
		wake_up_process(iocb->ki_obj.tsk);
595
		return;
Linus Torvalds's avatar
Linus Torvalds committed
596
597
	}

Kent Overstreet's avatar
Kent Overstreet committed
598
599
600
	/*
	 * Take rcu_read_lock() in case the kioctx is being destroyed, as we
	 * need to issue a wakeup after decrementing reqs_active.
Linus Torvalds's avatar
Linus Torvalds committed
601
	 */
Kent Overstreet's avatar
Kent Overstreet committed
602
	rcu_read_lock();
Linus Torvalds's avatar
Linus Torvalds committed
603

604
605
606
607
608
609
610
	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);
	}
611

Linus Torvalds's avatar
Linus Torvalds committed
612
613
614
615
	/*
	 * cancelled requests don't get events, userland was given one
	 * when the event got cancelled.
	 */
616
	if (unlikely(xchg(&iocb->ki_cancel,
617
618
619
			  KIOCB_CANCELLED) == KIOCB_CANCELLED)) {
		atomic_dec(&ctx->reqs_active);
		/* Still need the wake_up in case free_ioctx is waiting */
Linus Torvalds's avatar
Linus Torvalds committed
620
		goto put_rq;
621
	}
Linus Torvalds's avatar
Linus Torvalds committed
622

623
624
625
626
627
628
629
	/*
	 * Add a completion event to the ring buffer. Must be done holding
	 * ctx->ctx_lock to prevent other code from messing with the tail
	 * pointer since we might be called from irq context.
	 */
	spin_lock_irqsave(&ctx->completion_lock, flags);

630
	tail = ctx->tail;
Kent Overstreet's avatar
Kent Overstreet committed
631
632
	pos = tail + AIO_EVENTS_OFFSET;

633
	if (++tail >= ctx->nr_events)
634
		tail = 0;
Linus Torvalds's avatar
Linus Torvalds committed
635

636
	ev_page = kmap_atomic(ctx->ring_pages[pos / AIO_EVENTS_PER_PAGE]);
Kent Overstreet's avatar
Kent Overstreet committed
637
638
	event = ev_page + pos % AIO_EVENTS_PER_PAGE;

Linus Torvalds's avatar
Linus Torvalds committed
639
640
641
642
643
	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
644
	kunmap_atomic(ev_page);
645
	flush_dcache_page(ctx->ring_pages[pos / AIO_EVENTS_PER_PAGE]);
Kent Overstreet's avatar
Kent Overstreet committed
646
647

	pr_debug("%p[%u]: %p: %p %Lx %lx %lx\n",
Kent Overstreet's avatar
Kent Overstreet committed
648
649
		 ctx, tail, iocb, iocb->ki_obj.user, iocb->ki_user_data,
		 res, res2);
Linus Torvalds's avatar
Linus Torvalds committed
650
651
652
653
654
655

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

656
	ctx->tail = tail;
Linus Torvalds's avatar
Linus Torvalds committed
657

658
	ring = kmap_atomic(ctx->ring_pages[0]);
Kent Overstreet's avatar
Kent Overstreet committed
659
	ring->tail = tail;
660
	kunmap_atomic(ring);
661
	flush_dcache_page(ctx->ring_pages[0]);
Linus Torvalds's avatar
Linus Torvalds committed
662

663
664
	spin_unlock_irqrestore(&ctx->completion_lock, flags);

Kent Overstreet's avatar
Kent Overstreet committed
665
	pr_debug("added to ring %p at [%u]\n", iocb, tail);
Davide Libenzi's avatar
Davide Libenzi committed
666
667
668
669
670
671

	/*
	 * 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.
	 */
672
	if (iocb->ki_eventfd != NULL)
Davide Libenzi's avatar
Davide Libenzi committed
673
674
		eventfd_signal(iocb->ki_eventfd, 1);

Linus Torvalds's avatar
Linus Torvalds committed
675
676
put_rq:
	/* everything turned out well, dispose of the aiocb. */
677
	aio_put_req(iocb);
Linus Torvalds's avatar
Linus Torvalds committed
678

679
680
681
682
683
684
685
686
	/*
	 * 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
687
688
689
	if (waitqueue_active(&ctx->wait))
		wake_up(&ctx->wait);

Kent Overstreet's avatar
Kent Overstreet committed
690
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
691
}
692
EXPORT_SYMBOL(aio_complete);
Linus Torvalds's avatar
Linus Torvalds committed
693

694
695
696
/* 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
697
 */
698
699
static long aio_read_events_ring(struct kioctx *ctx,
				 struct io_event __user *event, long nr)
Linus Torvalds's avatar
Linus Torvalds committed
700
701
{
	struct aio_ring *ring;
702
703
704
705
	unsigned head, pos;
	long ret = 0;
	int copy_ret;

706
	mutex_lock(&ctx->ring_lock);
Linus Torvalds's avatar
Linus Torvalds committed
707

708
	ring = kmap_atomic(ctx->ring_pages[0]);
709
710
711
	head = ring->head;
	kunmap_atomic(ring);

712
	pr_debug("h%u t%u m%u\n", head, ctx->tail, ctx->nr_events);
Linus Torvalds's avatar
Linus Torvalds committed
713

714
	if (head == ctx->tail)
Linus Torvalds's avatar
Linus Torvalds committed
715
716
		goto out;

717
718
719
720
721
	while (ret < nr) {
		long avail;
		struct io_event *ev;
		struct page *page;

722
723
		avail = (head <= ctx->tail ? ctx->tail : ctx->nr_events) - head;
		if (head == ctx->tail)
724
725
726
727
728
729
730
			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;
731
		page = ctx->ring_pages[pos / AIO_EVENTS_PER_PAGE];
732
733
734
735
736
737
738
739
740
741
742
743
744
745
		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;
746
		head %= ctx->nr_events;
Linus Torvalds's avatar
Linus Torvalds committed
747
748
	}

749
	ring = kmap_atomic(ctx->ring_pages[0]);
750
	ring->head = head;
751
	kunmap_atomic(ring);
752
	flush_dcache_page(ctx->ring_pages[0]);
753

754
	pr_debug("%li  h%u t%u\n", ret, head, ctx->tail);
755
756

	atomic_sub(ret, &ctx->reqs_active);
757
out:
758
	mutex_unlock(&ctx->ring_lock);
759

Linus Torvalds's avatar
Linus Torvalds committed
760
761
762
	return ret;
}

763
764
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
765
{
766
	long ret = aio_read_events_ring(ctx, event + *i, nr - *i);
Linus Torvalds's avatar
Linus Torvalds committed
767

768
769
	if (ret > 0)
		*i += ret;
Linus Torvalds's avatar
Linus Torvalds committed
770

771
772
	if (unlikely(atomic_read(&ctx->dead)))
		ret = -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
773

774
775
	if (!*i)
		*i = ret;
Linus Torvalds's avatar
Linus Torvalds committed
776

777
	return ret < 0 || *i >= min_nr;
Linus Torvalds's avatar
Linus Torvalds committed
778
779
}

780
static long read_events(struct kioctx *ctx, long min_nr, long nr,
Linus Torvalds's avatar
Linus Torvalds committed
781
782
783
			struct io_event __user *event,
			struct timespec __user *timeout)
{
784
785
	ktime_t until = { .tv64 = KTIME_MAX };
	long ret = 0;
Linus Torvalds's avatar
Linus Torvalds committed
786
787
788

	if (timeout) {
		struct timespec	ts;
789

Linus Torvalds's avatar
Linus Torvalds committed
790
		if (unlikely(copy_from_user(&ts, timeout, sizeof(ts))))
791
			return -EFAULT;
Linus Torvalds's avatar
Linus Torvalds committed
792

793
		until = timespec_to_ktime(ts);
Linus Torvalds's avatar
Linus Torvalds committed
794
795
	}

796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
	/*
	 * 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
812

813
814
	if (!ret && signal_pending(current))
		ret = -EINTR;
Linus Torvalds's avatar
Linus Torvalds committed
815

816
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
}

/* 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.
 */
832
SYSCALL_DEFINE2(io_setup, unsigned, nr_events, aio_context_t __user *, ctxp)
Linus Torvalds's avatar
Linus Torvalds committed
833
834
835
836
837
838
839
840
841
842
{
	struct kioctx *ioctx = NULL;
	unsigned long ctx;
	long ret;

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

	ret = -EINVAL;
843
844
845
	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
846
847
848
849
850
851
852
		goto out;
	}

	ioctx = ioctx_alloc(nr_events);
	ret = PTR_ERR(ioctx);
	if (!IS_ERR(ioctx)) {
		ret = put_user(ioctx->user_id, ctxp);
853
		if (ret)
Kent Overstreet's avatar
Kent Overstreet committed
854
			kill_ioctx(ioctx);
855
		put_ioctx(ioctx);
Linus Torvalds's avatar
Linus Torvalds committed
856
857
858
859
860
861
862
863
864
	}

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
865
 *	implemented.  May fail with -EINVAL if the context pointed to
Linus Torvalds's avatar
Linus Torvalds committed
866
867
 *	is invalid.
 */
868
SYSCALL_DEFINE1(io_destroy, aio_context_t, ctx)
Linus Torvalds's avatar
Linus Torvalds committed
869
870
871
{
	struct kioctx *ioctx = lookup_ioctx(ctx);
	if (likely(NULL != ioctx)) {
Kent Overstreet's avatar
Kent Overstreet committed
872
		kill_ioctx(ioctx);
873
		put_ioctx(ioctx);
Linus Torvalds's avatar
Linus Torvalds committed
874
875
876
877
878
879
		return 0;
	}
	pr_debug("EINVAL: io_destroy: invalid context id\n");
	return -EINVAL;
}

880
static void aio_advance_iovec(struct kiocb *iocb, ssize_t ret)
Linus Torvalds's avatar
Linus Torvalds committed
881
{
882
883
884
885
886
887
888
889
890
891
892
893
894
	struct iovec *iov = &iocb->ki_iovec[iocb->ki_cur_seg];

	BUG_ON(ret <= 0);

	while (iocb->ki_cur_seg < iocb->ki_nr_segs && ret > 0) {
		ssize_t this = min((ssize_t)iov->iov_len, ret);
		iov->iov_base += this;
		iov->iov_len -= this;
		iocb->ki_left -= this;
		ret -= this;
		if (iov->iov_len == 0) {
			iocb->ki_cur_seg++;
			iov++;
895
		}
896
	}
Linus Torvalds's avatar
Linus Torvalds committed
897

898
899
900
	/* the caller should not have done more io than what fit in
	 * the remaining iovecs */
	BUG_ON(ret > 0 && iocb->ki_left == 0);
Linus Torvalds's avatar
Linus Torvalds committed
901
902
}

903
static ssize_t aio_rw_vect_retry(struct kiocb *iocb)
Linus Torvalds's avatar
Linus Torvalds committed
904
905
{
	struct file *file = iocb->ki_filp;
906
907
908
909
	struct address_space *mapping = file->f_mapping;
	struct inode *inode = mapping->host;
	ssize_t (*rw_op)(struct kiocb *, const struct iovec *,
			 unsigned long, loff_t);
Linus Torvalds's avatar
Linus Torvalds committed
910
	ssize_t ret = 0;
911
912
913
914
915
916
917
918
919
920
	unsigned short opcode;

	if ((iocb->ki_opcode == IOCB_CMD_PREADV) ||
		(iocb->ki_opcode == IOCB_CMD_PREAD)) {
		rw_op = file->f_op->aio_read;
		opcode = IOCB_CMD_PREADV;
	} else {
		rw_op = file->f_op->aio_write;
		opcode = IOCB_CMD_PWRITEV;
	}
Linus Torvalds's avatar
Linus Torvalds committed
921

922
923
924
925
	/* This matches the pread()/pwrite() logic */
	if (iocb->ki_pos < 0)
		return -EINVAL;

926
927
	if (opcode == IOCB_CMD_PWRITEV)
		file_start_write(file);
928
	do {
929
930
931
932
933
934
935
936
937
938
939
		ret = rw_op(iocb, &iocb->ki_iovec[iocb->ki_cur_seg],
			    iocb->ki_nr_segs - iocb->ki_cur_seg,
			    iocb->ki_pos);
		if (ret > 0)
			aio_advance_iovec(iocb, ret);

	/* retry all partial writes.  retry partial reads as long as its a
	 * regular file. */
	} while (ret > 0 && iocb->ki_left > 0 &&
		 (opcode == IOCB_CMD_PWRITEV ||
		  (!S_ISFIFO(inode->i_mode) && !S_ISSOCK(inode->i_mode))));
940
941
	if (opcode == IOCB_CMD_PWRITEV)
		file_end_write(file);
Linus Torvalds's avatar
Linus Torvalds committed
942

943
944
	/* This means we must have transferred all that we could */
	/* No need to retry anymore */
Linus Torvalds's avatar
Linus Torvalds committed
945
946
947
	if ((ret == 0) || (iocb->ki_left == 0))
		ret = iocb->ki_nbytes - iocb->ki_left;

948
949
950
	/* If we managed to write some out we return that, rather than
	 * the eventual error. */
	if (opcode == IOCB_CMD_PWRITEV
Zach Brown's avatar
Zach Brown committed
951
	    && ret < 0 && ret != -EIOCBQUEUED
952
953
954
	    && iocb->ki_nbytes - iocb->ki_left)
		ret = iocb->ki_nbytes - iocb->ki_left;

Linus Torvalds's avatar
Linus Torvalds committed
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
	return ret;
}

static ssize_t aio_fdsync(struct kiocb *iocb)
{
	struct file *file = iocb->ki_filp;
	ssize_t ret = -EINVAL;

	if (file->f_op->aio_fsync)
		ret = file->f_op->aio_fsync(iocb, 1);
	return ret;
}

static ssize_t aio_fsync(struct kiocb *iocb)
{
	struct file *file = iocb->ki_filp;
	ssize_t ret = -EINVAL;

	if (file->f_op->aio_fsync)
		ret = file->f_op->aio_fsync(iocb, 0);
	return ret;
}

978
static ssize_t aio_setup_vectored_rw(int type, struct kiocb *kiocb, bool compat)
979
980
981
{
	ssize_t ret;

982
983
984
985
986
#ifdef CONFIG_COMPAT
	if (compat)
		ret = compat_rw_copy_check_uvector(type,
				(struct compat_iovec __user *)kiocb->ki_buf,
				kiocb->ki_nbytes, 1, &kiocb->ki_inline_vec,
987
				&kiocb->ki_iovec);
988
989
990
991
992
	else
#endif
		ret = rw_copy_check_uvector(type,
				(struct iovec __user *)kiocb->ki_buf,
				kiocb->ki_nbytes, 1, &kiocb->ki_inline_vec,
993
				&kiocb->ki_iovec);
994
995
996
	if (ret < 0)
		goto out;

997
998
999
1000
	ret = rw_verify_area(type, kiocb->ki_filp, &kiocb->ki_pos, ret);
	if (ret < 0)
		goto out;

1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
	kiocb->ki_nr_segs = kiocb->ki_nbytes;
	kiocb->ki_cur_seg = 0;
	/* ki_nbytes/left now reflect bytes instead of segs */
	kiocb->ki_nbytes = ret;
	kiocb->ki_left = ret;

	ret = 0;
out:
	return ret;
}

1012
static ssize_t aio_setup_single_vector(int type, struct file * file, struct kiocb *kiocb)
1013
{
1014
1015
1016
1017
1018
1019
	int bytes;

	bytes = rw_verify_area(type, file, &kiocb->ki_pos, kiocb->ki_left);
	if (bytes < 0)
		return bytes;

1020
1021
	kiocb->ki_iovec = &kiocb->ki_inline_vec;
	kiocb->ki_iovec->iov_base = kiocb->ki_buf;
1022
	kiocb->ki_iovec->iov_len = bytes;
1023
1024
1025
1026
1027
	kiocb->ki_nr_segs = 1;
	kiocb->ki_cur_seg = 0;
	return 0;
}

Linus Torvalds's avatar
Linus Torvalds committed
1028
1029
1030
1031
1032
/*
 * aio_setup_iocb:
 *	Performs the initial checks and aio retry method
 *	setup for the kiocb at the time of io submission.
 */
1033
static ssize_t aio_setup_iocb(struct kiocb *kiocb, bool compat)
Linus Torvalds's avatar
Linus Torvalds committed
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
{
	struct file *file = kiocb->ki_filp;
	ssize_t ret = 0;

	switch (kiocb->ki_opcode) {
	case IOCB_CMD_PREAD:
		ret = -EBADF;
		if (unlikely(!(file->f_mode & FMODE_READ)))
			break;
		ret = -EFAULT;
		if (unlikely(!access_ok(VERIFY_WRITE, kiocb->ki_buf,
			kiocb->ki_left)))
			break;
1047
		ret = aio_setup_single_vector(READ, file, kiocb);
1048
1049
		if (ret)
			break;
Linus Torvalds's avatar
Linus Torvalds committed
1050
1051
		ret = -EINVAL;
		if (file->f_op->aio_read)
1052
			kiocb->ki_retry = aio_rw_vect_retry;
Linus Torvalds's avatar
Linus Torvalds committed
1053
1054
1055
1056
1057
1058
1059
1060
1061
		break;
	case IOCB_CMD_PWRITE:
		ret = -EBADF;
		if (unlikely(!(file->f_mode & FMODE_WRITE)))
			break;
		ret = -EFAULT;
		if (unlikely(!access_ok(VERIFY_READ, kiocb->ki_buf,
			kiocb->ki_left)))
			break;
1062
		ret = aio_setup_single_vector(WRITE, file, kiocb);
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
		if (ret)
			break;
		ret = -EINVAL;
		if (file->f_op->aio_write)
			kiocb->ki_retry = aio_rw_vect_retry;
		break;
	case IOCB_CMD_PREADV:
		ret = -EBADF;
		if (unlikely(!(file->f_mode & FMODE_READ)))
			break;
1073
		ret = aio_setup_vectored_rw(READ, kiocb, compat);
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
		if (ret)
			break;
		ret = -EINVAL;
		if (file->f_op->aio_read)
			kiocb->ki_retry = aio_rw_vect_retry;
		break;
	case IOCB_CMD_PWRITEV:
		ret = -EBADF;
		if (unlikely(!(file->f_mode & FMODE_WRITE)))
			break;
1084
		ret = aio_setup_vectored_rw(WRITE, kiocb, compat);
1085
1086
		if (ret)
			break;
Linus Torvalds's avatar
Linus Torvalds committed
1087
1088
		ret = -EINVAL;
		if (file->f_op->aio_write)
1089
			kiocb->ki_retry = aio_rw_vect_retry;
Linus Torvalds's avatar
Linus Torvalds committed
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
		break;
	case IOCB_CMD_FDSYNC:
		ret = -EINVAL;
		if (file->f_op->aio_fsync)
			kiocb->ki_retry = aio_fdsync;
		break;
	case IOCB_CMD_FSYNC:
		ret = -EINVAL;
		if (file->f_op->aio_fsync)
			kiocb->ki_retry = aio_fsync;
		break;
	default:
Kent Overstreet's avatar
Kent Overstreet committed
1102
		pr_debug("EINVAL: no operation provided\n");
Linus Torvalds's avatar
Linus Torvalds committed
1103
1104
1105
1106
1107
1108
1109
1110
1111
		ret = -EINVAL;
	}

	if (!kiocb->ki_retry)
		return ret;

	return 0;
}

1112
static int io_submit_one(struct kioctx *ctx, struct iocb __user *user_iocb,
Kent Overstreet's avatar
Kent Overstreet committed
1113
			 struct iocb *iocb, bool compat)
Linus Torvalds's avatar
Linus Torvalds committed
1114
1115
1116
1117
1118
{
	struct kiocb *req;
	ssize_t ret;

	/* enforce forwards compatibility on users */
1119
	if (unlikely(iocb->aio_reserved1 || iocb->aio_reserved2)) {
Kent Overstreet's avatar
Kent Overstreet committed
1120
		pr_debug("EINVAL: reserve field set\n");
Linus Torvalds's avatar
Linus Torvalds committed
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
		return -EINVAL;
	}

	/* prevent overflows */
	if (unlikely(
	    (iocb->aio_buf != (unsigned long)iocb->aio_buf) ||
	    (iocb->aio_nbytes != (size_t)iocb->aio_nbytes) ||
	    ((ssize_t)iocb->aio_nbytes < 0)
	   )) {
		pr_debug("EINVAL: io_submit: overflow check\n");
		return -EINVAL;
	}

Kent Overstreet's avatar
Kent Overstreet committed
1134
	req = aio_get_req(ctx);  /* returns with 2 references to req */
1135
	if (unlikely(!req))
Linus Torvalds's avatar
Linus Torvalds committed
1136
		return -EAGAIN;
1137
1138
1139
1140
1141

	req->ki_filp = fget(iocb->aio_fildes);
	if (unlikely(!req->ki_filp)) {
		ret = -EBADF;
		goto out_put_req;
Linus Torvalds's avatar
Linus Torvalds committed
1142
	}
1143

1144
1145
1146
1147
1148
1149
1150
	if (iocb->aio_flags & IOCB_FLAG_RESFD) {
		/*
		 * If the IOCB_FLAG_RESFD flag of aio_flags is set, get an
		 * instance of the file* now. The file descriptor must be
		 * an eventfd() fd, and will be signaled for each completed
		 * event using the eventfd_signal() function.
		 */
1151
		req->ki_eventfd = eventfd_ctx_fdget((int) iocb->aio_resfd);
1152
		if (IS_ERR(req->ki_eventfd)) {
1153
			ret = PTR_ERR(req->ki_eventfd);
1154
			req->ki_eventfd = NULL;
1155
1156
1157
			goto out_put_req;
		}
	}
Linus Torvalds's avatar
Linus Torvalds committed
1158

1159
	ret = put_user(req->ki_key, &user_iocb->aio_key);
Linus Torvalds's avatar
Linus Torvalds committed
1160
	if (unlikely(ret)) {
Kent Overstreet's avatar
Kent Overstreet committed
1161
		pr_debug("EFAULT: aio_key\n");
Linus Torvalds's avatar
Linus Torvalds committed
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
		goto out_put_req;
	}

	req->ki_obj.user = user_iocb;
	req->ki_user_data = iocb->aio_data;
	req->ki_pos = iocb->aio_offset;

	req->ki_buf = (char __user *)(unsigned long)iocb->aio_buf;
	req->ki_left = req->ki_nbytes = iocb->aio_nbytes;
	req->ki_opcode = iocb->aio_lio_opcode;

1173
	ret = aio_setup_iocb(req, compat);
Zach Brown's avatar
Zach Brown committed
1174
	if (ret)
1175
		goto out_put_req;
Zach Brown's avatar
Zach Brown committed
1176

1177
	ret = req->ki_retry(req);
Zach Brown's avatar
Zach Brown committed
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
	if (ret != -EIOCBQUEUED) {
		/*
		 * There's no easy way to restart the syscall since other AIO's
		 * may be already running. Just fail this IO with EINTR.
		 */
		if (unlikely(ret == -ERESTARTSYS || ret == -ERESTARTNOINTR ||
			     ret == -ERESTARTNOHAND ||
			     ret == -ERESTART_RESTARTBLOCK))
			ret = -EINTR;
		aio_complete(req, ret, 0);
1188
	}
Jeff Moyer's avatar
Jeff Moyer committed
1189

Linus Torvalds's avatar
Linus Torvalds committed
1190
1191
1192
1193
	aio_put_req(req);	/* drop extra ref to req */
	return 0;

out_put_req:
1194
	atomic_dec(&ctx->reqs_active);
Linus Torvalds's avatar
Linus Torvalds committed
1195
1196
1197
1198
1199
	aio_put_req(req);	/* drop extra ref to req */
	aio_put_req(req);	/* drop i/o ref to req */
	return ret;
}

1200
1201
long do_io_submit(aio_context_t ctx_id, long nr,
		  struct iocb __user *__user *iocbpp, bool compat)
Linus Torvalds's avatar
Linus Torvalds committed
1202
1203
1204
{
	struct kioctx *ctx;
	long ret = 0;
Jeff Moyer's avatar
Jeff Moyer committed
1205
	int i = 0;
Shaohua Li's avatar