aio.c 32.8 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

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

70
71
72
73
	/*
	 * This is what userspace passed to io_setup(), it's not used for
	 * anything but counting against the global max_reqs quota.
	 *
74
	 * The real limit is nr_events - 1, which will be larger (see
75
76
	 * aio_setup_ring())
	 */
77
78
	unsigned		max_reqs;

79
80
	/* Size of ringbuffer, in units of struct io_event */
	unsigned		nr_events;
81

82
83
84
85
86
87
	unsigned long		mmap_base;
	unsigned long		mmap_size;

	struct page		**ring_pages;
	long			nr_pages;

88
89
90
91
92
93
94
95
96
97
98
99
	struct rcu_head		rcu_head;
	struct work_struct	rcu_work;

	struct {
		atomic_t	reqs_active;
	} ____cacheline_aligned_in_smp;

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

100
101
	struct {
		struct mutex	ring_lock;
102
103
		wait_queue_head_t wait;
	} ____cacheline_aligned_in_smp;
104
105
106
107

	struct {
		unsigned	tail;
		spinlock_t	completion_lock;
108
	} ____cacheline_aligned_in_smp;
109
110

	struct page		*internal_pages[AIO_RING_PAGES];
111
112
};

Linus Torvalds's avatar
Linus Torvalds committed
113
/*------ sysctl variables----*/
114
115
116
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
117
118
/*----end sysctl variables---*/

119
120
static struct kmem_cache	*kiocb_cachep;
static struct kmem_cache	*kioctx_cachep;
Linus Torvalds's avatar
Linus Torvalds committed
121
122
123
124
125
126
127

/* 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)
{
128
129
	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
130

Kent Overstreet's avatar
Kent Overstreet committed
131
	pr_debug("sizeof(struct page) = %zu\n", sizeof(struct page));
Linus Torvalds's avatar
Linus Torvalds committed
132
133
134

	return 0;
}
135
__initcall(aio_setup);
Linus Torvalds's avatar
Linus Torvalds committed
136
137
138
139
140

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

141
142
	for (i = 0; i < ctx->nr_pages; i++)
		put_page(ctx->ring_pages[i]);
Linus Torvalds's avatar
Linus Torvalds committed
143

144
145
	if (ctx->mmap_size)
		vm_munmap(ctx->mmap_base, ctx->mmap_size);
Linus Torvalds's avatar
Linus Torvalds committed
146

147
148
	if (ctx->ring_pages && ctx->ring_pages != ctx->internal_pages)
		kfree(ctx->ring_pages);
Linus Torvalds's avatar
Linus Torvalds committed
149
150
151
152
153
154
}

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
155
	struct mm_struct *mm = current->mm;
156
	unsigned long size, populate;
Linus Torvalds's avatar
Linus Torvalds committed
157
158
159
160
161
162
163
164
165
166
167
168
169
170
	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);

171
172
	ctx->nr_events = 0;
	ctx->ring_pages = ctx->internal_pages;
Linus Torvalds's avatar
Linus Torvalds committed
173
	if (nr_pages > AIO_RING_PAGES) {
174
175
176
		ctx->ring_pages = kcalloc(nr_pages, sizeof(struct page *),
					  GFP_KERNEL);
		if (!ctx->ring_pages)
Linus Torvalds's avatar
Linus Torvalds committed
177
178
179
			return -ENOMEM;
	}

180
181
	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
182
	down_write(&mm->mmap_sem);
183
184
185
186
	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
187
		up_write(&mm->mmap_sem);
188
		ctx->mmap_size = 0;
Linus Torvalds's avatar
Linus Torvalds committed
189
190
191
192
		aio_free_ring(ctx);
		return -EAGAIN;
	}

193
194
195
	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
196
	up_write(&mm->mmap_sem);
Linus Torvalds's avatar
Linus Torvalds committed
197

198
	if (unlikely(ctx->nr_pages != nr_pages)) {
Linus Torvalds's avatar
Linus Torvalds committed
199
200
201
		aio_free_ring(ctx);
		return -EAGAIN;
	}
202
	if (populate)
203
		mm_populate(ctx->mmap_base, populate);
Linus Torvalds's avatar
Linus Torvalds committed
204

205
206
	ctx->user_id = ctx->mmap_base;
	ctx->nr_events = nr_events; /* trusted copy */
Linus Torvalds's avatar
Linus Torvalds committed
207

208
	ring = kmap_atomic(ctx->ring_pages[0]);
Linus Torvalds's avatar
Linus Torvalds committed
209
210
211
212
213
214
215
	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);
216
	kunmap_atomic(ring);
217
	flush_dcache_page(ctx->ring_pages[0]);
Linus Torvalds's avatar
Linus Torvalds committed
218
219
220
221
222
223
224
225

	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)

226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
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
242
243
244
static int kiocb_cancel(struct kioctx *ctx, struct kiocb *kiocb,
			struct io_event *res)
{
245
	kiocb_cancel_fn *old, *cancel;
Kent Overstreet's avatar
Kent Overstreet committed
246
247
	int ret = -EINVAL;

248
249
250
251
252
253
254
255
256
	/*
	 * 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
257

258
259
260
		old = cancel;
		cancel = cmpxchg(&kiocb->ki_cancel, old, KIOCB_CANCELLED);
	} while (cancel != old);
Kent Overstreet's avatar
Kent Overstreet committed
261

262
263
264
265
266
267
268
269
270
	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
271
272
273
274

	return ret;
}

Kent Overstreet's avatar
Kent Overstreet committed
275
276
277
278
279
280
281
282
283
284
285
286
287
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)
{
288
	struct aio_ring *ring;
Kent Overstreet's avatar
Kent Overstreet committed
289
290
	struct io_event res;
	struct kiocb *req;
291
	unsigned head, avail;
Kent Overstreet's avatar
Kent Overstreet committed
292
293
294
295
296
297
298
299
300
301
302
303
304

	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);

305
	ring = kmap_atomic(ctx->ring_pages[0]);
306
307
308
309
	head = ring->head;
	kunmap_atomic(ring);

	while (atomic_read(&ctx->reqs_active) > 0) {
310
311
312
		wait_event(ctx->wait,
				head != ctx->tail ||
				atomic_read(&ctx->reqs_active) <= 0);
313

314
		avail = (head <= ctx->tail ? ctx->tail : ctx->nr_events) - head;
315
316
317

		atomic_sub(avail, &ctx->reqs_active);
		head += avail;
318
		head %= ctx->nr_events;
319
320
321
	}

	WARN_ON(atomic_read(&ctx->reqs_active) < 0);
Kent Overstreet's avatar
Kent Overstreet committed
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347

	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
348
349
350
351
352
/* 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
353
	struct mm_struct *mm = current->mm;
Linus Torvalds's avatar
Linus Torvalds committed
354
	struct kioctx *ctx;
355
	int err = -ENOMEM;
Linus Torvalds's avatar
Linus Torvalds committed
356
357
358
359
360
361
362
363

	/* 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);
	}

364
	if (!nr_events || (unsigned long)nr_events > aio_max_nr)
Linus Torvalds's avatar
Linus Torvalds committed
365
366
		return ERR_PTR(-EAGAIN);

367
	ctx = kmem_cache_zalloc(kioctx_cachep, GFP_KERNEL);
Linus Torvalds's avatar
Linus Torvalds committed
368
369
370
371
372
	if (!ctx)
		return ERR_PTR(-ENOMEM);

	ctx->max_reqs = nr_events;

Al Viro's avatar
Al Viro committed
373
	atomic_set(&ctx->users, 2);
Kent Overstreet's avatar
Kent Overstreet committed
374
	atomic_set(&ctx->dead, 0);
Linus Torvalds's avatar
Linus Torvalds committed
375
	spin_lock_init(&ctx->ctx_lock);
376
	spin_lock_init(&ctx->completion_lock);
377
	mutex_init(&ctx->ring_lock);
Linus Torvalds's avatar
Linus Torvalds committed
378
379
380
381
382
383
384
385
	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 */
386
	spin_lock(&aio_nr_lock);
387
388
	if (aio_nr + nr_events > aio_max_nr ||
	    aio_nr + nr_events < aio_nr) {
389
		spin_unlock(&aio_nr_lock);
Linus Torvalds's avatar
Linus Torvalds committed
390
		goto out_cleanup;
391
392
	}
	aio_nr += ctx->max_reqs;
393
	spin_unlock(&aio_nr_lock);
Linus Torvalds's avatar
Linus Torvalds committed
394

Jeff Moyer's avatar
Jeff Moyer committed
395
	/* now link into global list. */
396
397
398
	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
399

Kent Overstreet's avatar
Kent Overstreet committed
400
	pr_debug("allocated ioctx %p[%ld]: mm=%p mask=0x%x\n",
401
		 ctx, ctx->user_id, mm, ctx->nr_events);
Linus Torvalds's avatar
Linus Torvalds committed
402
403
404
	return ctx;

out_cleanup:
405
406
	err = -EAGAIN;
	aio_free_ring(ctx);
Linus Torvalds's avatar
Linus Torvalds committed
407
408
out_freectx:
	kmem_cache_free(kioctx_cachep, ctx);
Kent Overstreet's avatar
Kent Overstreet committed
409
	pr_debug("error allocating ioctx %d\n", err);
410
	return ERR_PTR(err);
Linus Torvalds's avatar
Linus Torvalds committed
411
412
}

Kent Overstreet's avatar
Kent Overstreet committed
413
static void kill_ioctx_work(struct work_struct *work)
Linus Torvalds's avatar
Linus Torvalds committed
414
{
Kent Overstreet's avatar
Kent Overstreet committed
415
	struct kioctx *ctx = container_of(work, struct kioctx, rcu_work);
416

Kent Overstreet's avatar
Kent Overstreet committed
417
418
419
	wake_up_all(&ctx->wait);
	put_ioctx(ctx);
}
Kent Overstreet's avatar
Kent Overstreet committed
420

Kent Overstreet's avatar
Kent Overstreet committed
421
422
423
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
424

Kent Overstreet's avatar
Kent Overstreet committed
425
426
427
	INIT_WORK(&ctx->rcu_work, kill_ioctx_work);
	schedule_work(&ctx->rcu_work);
}
Linus Torvalds's avatar
Linus Torvalds committed
428

Kent Overstreet's avatar
Kent Overstreet committed
429
430
431
432
433
434
435
436
437
438
439
/* 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();
440

Kent Overstreet's avatar
Kent Overstreet committed
441
442
443
444
445
446
447
448
449
		/*
		 * 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
450
451
452
453
454
}

/* wait_on_sync_kiocb:
 *	Waits on the given sync kiocb to complete.
 */
455
ssize_t wait_on_sync_kiocb(struct kiocb *iocb)
Linus Torvalds's avatar
Linus Torvalds committed
456
{
457
	while (atomic_read(&iocb->ki_users)) {
Linus Torvalds's avatar
Linus Torvalds committed
458
		set_current_state(TASK_UNINTERRUPTIBLE);
459
		if (!atomic_read(&iocb->ki_users))
Linus Torvalds's avatar
Linus Torvalds committed
460
			break;
461
		io_schedule();
Linus Torvalds's avatar
Linus Torvalds committed
462
463
464
465
	}
	__set_current_state(TASK_RUNNING);
	return iocb->ki_user_data;
}
466
EXPORT_SYMBOL(wait_on_sync_kiocb);
Linus Torvalds's avatar
Linus Torvalds committed
467

Kent Overstreet's avatar
Kent Overstreet committed
468
469
470
471
472
473
474
/*
 * 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
475
 */
476
void exit_aio(struct mm_struct *mm)
Linus Torvalds's avatar
Linus Torvalds committed
477
{
478
	struct kioctx *ctx;
Kent Overstreet's avatar
Kent Overstreet committed
479
	struct hlist_node *n;
480

Kent Overstreet's avatar
Kent Overstreet committed
481
	hlist_for_each_entry_safe(ctx, n, &mm->ioctx_list, list) {
Linus Torvalds's avatar
Linus Torvalds committed
482
483
484
		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
485
486
				atomic_read(&ctx->users),
				atomic_read(&ctx->dead),
487
				atomic_read(&ctx->reqs_active));
488
489
490
491
492
493
494
495
		/*
		 * 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.
		 */
496
		ctx->mmap_size = 0;
Kent Overstreet's avatar
Kent Overstreet committed
497
498
499
500
501

		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
502
503
504
505
	}
}

/* aio_get_req
506
 *	Allocate a slot for an aio request.  Increments the ki_users count
Linus Torvalds's avatar
Linus Torvalds committed
507
508
509
 * of the kioctx so that the kioctx stays around until all requests are
 * complete.  Returns NULL if no requests are free.
 *
510
 * Returns with kiocb->ki_users set to 2.  The io submit code path holds
Linus Torvalds's avatar
Linus Torvalds committed
511
512
513
514
 * 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
515
static inline struct kiocb *aio_get_req(struct kioctx *ctx)
Linus Torvalds's avatar
Linus Torvalds committed
516
{
Kent Overstreet's avatar
Kent Overstreet committed
517
518
	struct kiocb *req;

519
	if (atomic_read(&ctx->reqs_active) >= ctx->nr_events)
Kent Overstreet's avatar
Kent Overstreet committed
520
521
		return NULL;

522
	if (atomic_inc_return(&ctx->reqs_active) > ctx->nr_events - 1)
Kent Overstreet's avatar
Kent Overstreet committed
523
		goto out_put;
Linus Torvalds's avatar
Linus Torvalds committed
524

525
	req = kmem_cache_alloc(kiocb_cachep, GFP_KERNEL|__GFP_ZERO);
Linus Torvalds's avatar
Linus Torvalds committed
526
	if (unlikely(!req))
Kent Overstreet's avatar
Kent Overstreet committed
527
		goto out_put;
Linus Torvalds's avatar
Linus Torvalds committed
528

529
	atomic_set(&req->ki_users, 2);
Linus Torvalds's avatar
Linus Torvalds committed
530
531
	req->ki_ctx = ctx;

Jeff Moyer's avatar
Jeff Moyer committed
532
	return req;
Kent Overstreet's avatar
Kent Overstreet committed
533
534
535
out_put:
	atomic_dec(&ctx->reqs_active);
	return NULL;
Linus Torvalds's avatar
Linus Torvalds committed
536
537
}

538
static void kiocb_free(struct kiocb *req)
Linus Torvalds's avatar
Linus Torvalds committed
539
{
540
541
	if (req->ki_filp)
		fput(req->ki_filp);
542
543
	if (req->ki_eventfd != NULL)
		eventfd_ctx_put(req->ki_eventfd);
Linus Torvalds's avatar
Linus Torvalds committed
544
545
	if (req->ki_dtor)
		req->ki_dtor(req);
546
547
	if (req->ki_iovec != &req->ki_inline_vec)
		kfree(req->ki_iovec);
Linus Torvalds's avatar
Linus Torvalds committed
548
549
550
	kmem_cache_free(kiocb_cachep, req);
}

551
void aio_put_req(struct kiocb *req)
Linus Torvalds's avatar
Linus Torvalds committed
552
{
553
554
	if (atomic_dec_and_test(&req->ki_users))
		kiocb_free(req);
Linus Torvalds's avatar
Linus Torvalds committed
555
}
556
EXPORT_SYMBOL(aio_put_req);
Linus Torvalds's avatar
Linus Torvalds committed
557

558
static struct kioctx *lookup_ioctx(unsigned long ctx_id)
Linus Torvalds's avatar
Linus Torvalds committed
559
{
560
	struct mm_struct *mm = current->mm;
561
	struct kioctx *ctx, *ret = NULL;
Linus Torvalds's avatar
Linus Torvalds committed
562

563
564
	rcu_read_lock();

565
	hlist_for_each_entry_rcu(ctx, &mm->ioctx_list, list) {
Kent Overstreet's avatar
Kent Overstreet committed
566
567
		if (ctx->user_id == ctx_id) {
			atomic_inc(&ctx->users);
568
			ret = ctx;
Linus Torvalds's avatar
Linus Torvalds committed
569
570
			break;
		}
571
	}
Linus Torvalds's avatar
Linus Torvalds committed
572

573
	rcu_read_unlock();
574
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
575
576
577
578
579
}

/* aio_complete
 *	Called when the io request on the given iocb is complete.
 */
580
void aio_complete(struct kiocb *iocb, long res, long res2)
Linus Torvalds's avatar
Linus Torvalds committed
581
582
583
{
	struct kioctx	*ctx = iocb->ki_ctx;
	struct aio_ring	*ring;
Kent Overstreet's avatar
Kent Overstreet committed
584
	struct io_event	*ev_page, *event;
Linus Torvalds's avatar
Linus Torvalds committed
585
	unsigned long	flags;
Kent Overstreet's avatar
Kent Overstreet committed
586
	unsigned tail, pos;
Linus Torvalds's avatar
Linus Torvalds committed
587

588
589
590
591
592
593
	/*
	 * 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
594
595
	 */
	if (is_sync_kiocb(iocb)) {
596
		BUG_ON(atomic_read(&iocb->ki_users) != 1);
Linus Torvalds's avatar
Linus Torvalds committed
597
		iocb->ki_user_data = res;
598
		atomic_set(&iocb->ki_users, 0);
Linus Torvalds's avatar
Linus Torvalds committed
599
		wake_up_process(iocb->ki_obj.tsk);
600
		return;
Linus Torvalds's avatar
Linus Torvalds committed
601
602
	}

Kent Overstreet's avatar
Kent Overstreet committed
603
604
605
	/*
	 * 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
606
	 */
Kent Overstreet's avatar
Kent Overstreet committed
607
	rcu_read_lock();
Linus Torvalds's avatar
Linus Torvalds committed
608

609
610
611
612
613
614
615
	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);
	}
616

Linus Torvalds's avatar
Linus Torvalds committed
617
618
619
620
	/*
	 * cancelled requests don't get events, userland was given one
	 * when the event got cancelled.
	 */
621
	if (unlikely(xchg(&iocb->ki_cancel,
622
623
624
			  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
625
		goto put_rq;
626
	}
Linus Torvalds's avatar
Linus Torvalds committed
627

628
629
630
631
632
633
634
	/*
	 * 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);

635
	tail = ctx->tail;
Kent Overstreet's avatar
Kent Overstreet committed
636
637
	pos = tail + AIO_EVENTS_OFFSET;

638
	if (++tail >= ctx->nr_events)
639
		tail = 0;
Linus Torvalds's avatar
Linus Torvalds committed
640

641
	ev_page = kmap_atomic(ctx->ring_pages[pos / AIO_EVENTS_PER_PAGE]);
Kent Overstreet's avatar
Kent Overstreet committed
642
643
	event = ev_page + pos % AIO_EVENTS_PER_PAGE;

Linus Torvalds's avatar
Linus Torvalds committed
644
645
646
647
648
	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
649
	kunmap_atomic(ev_page);
650
	flush_dcache_page(ctx->ring_pages[pos / AIO_EVENTS_PER_PAGE]);
Kent Overstreet's avatar
Kent Overstreet committed
651
652

	pr_debug("%p[%u]: %p: %p %Lx %lx %lx\n",
Kent Overstreet's avatar
Kent Overstreet committed
653
654
		 ctx, tail, iocb, iocb->ki_obj.user, iocb->ki_user_data,
		 res, res2);
Linus Torvalds's avatar
Linus Torvalds committed
655
656
657
658
659
660

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

661
	ctx->tail = tail;
Linus Torvalds's avatar
Linus Torvalds committed
662

663
	ring = kmap_atomic(ctx->ring_pages[0]);
Kent Overstreet's avatar
Kent Overstreet committed
664
	ring->tail = tail;
665
	kunmap_atomic(ring);
666
	flush_dcache_page(ctx->ring_pages[0]);
Linus Torvalds's avatar
Linus Torvalds committed
667

668
669
	spin_unlock_irqrestore(&ctx->completion_lock, flags);

Kent Overstreet's avatar
Kent Overstreet committed
670
	pr_debug("added to ring %p at [%u]\n", iocb, tail);
Davide Libenzi's avatar
Davide Libenzi committed
671
672
673
674
675
676

	/*
	 * 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.
	 */
677
	if (iocb->ki_eventfd != NULL)
Davide Libenzi's avatar
Davide Libenzi committed
678
679
		eventfd_signal(iocb->ki_eventfd, 1);

Linus Torvalds's avatar
Linus Torvalds committed
680
681
put_rq:
	/* everything turned out well, dispose of the aiocb. */
682
	aio_put_req(iocb);
Linus Torvalds's avatar
Linus Torvalds committed
683

684
685
686
687
688
689
690
691
	/*
	 * 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
692
693
694
	if (waitqueue_active(&ctx->wait))
		wake_up(&ctx->wait);

Kent Overstreet's avatar
Kent Overstreet committed
695
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
696
}
697
EXPORT_SYMBOL(aio_complete);
Linus Torvalds's avatar
Linus Torvalds committed
698

699
700
701
/* 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
702
 */
703
704
static long aio_read_events_ring(struct kioctx *ctx,
				 struct io_event __user *event, long nr)
Linus Torvalds's avatar
Linus Torvalds committed
705
706
{
	struct aio_ring *ring;
707
708
709
710
	unsigned head, pos;
	long ret = 0;
	int copy_ret;

711
	mutex_lock(&ctx->ring_lock);
Linus Torvalds's avatar
Linus Torvalds committed
712

713
	ring = kmap_atomic(ctx->ring_pages[0]);
714
715
716
	head = ring->head;
	kunmap_atomic(ring);

717
	pr_debug("h%u t%u m%u\n", head, ctx->tail, ctx->nr_events);
Linus Torvalds's avatar
Linus Torvalds committed
718

719
	if (head == ctx->tail)
Linus Torvalds's avatar
Linus Torvalds committed
720
721
		goto out;

722
723
724
725
726
	while (ret < nr) {
		long avail;
		struct io_event *ev;
		struct page *page;

727
728
		avail = (head <= ctx->tail ? ctx->tail : ctx->nr_events) - head;
		if (head == ctx->tail)
729
730
731
732
733
734
735
			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;
736
		page = ctx->ring_pages[pos / AIO_EVENTS_PER_PAGE];
737
738
739
740
741
742
743
744
745
746
747
748
749
750
		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;
751
		head %= ctx->nr_events;
Linus Torvalds's avatar
Linus Torvalds committed
752
753
	}

754
	ring = kmap_atomic(ctx->ring_pages[0]);
755
	ring->head = head;
756
	kunmap_atomic(ring);
757
	flush_dcache_page(ctx->ring_pages[0]);
758

759
	pr_debug("%li  h%u t%u\n", ret, head, ctx->tail);
760
761

	atomic_sub(ret, &ctx->reqs_active);
762
out:
763
	mutex_unlock(&ctx->ring_lock);
764

Linus Torvalds's avatar
Linus Torvalds committed
765
766
767
	return ret;
}

768
769
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
770
{
771
	long ret = aio_read_events_ring(ctx, event + *i, nr - *i);
Linus Torvalds's avatar
Linus Torvalds committed
772

773
774
	if (ret > 0)
		*i += ret;
Linus Torvalds's avatar
Linus Torvalds committed
775

776
777
	if (unlikely(atomic_read(&ctx->dead)))
		ret = -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
778

779
780
	if (!*i)
		*i = ret;
Linus Torvalds's avatar
Linus Torvalds committed
781

782
	return ret < 0 || *i >= min_nr;
Linus Torvalds's avatar
Linus Torvalds committed
783
784
}

785
static long read_events(struct kioctx *ctx, long min_nr, long nr,
Linus Torvalds's avatar
Linus Torvalds committed
786
787
788
			struct io_event __user *event,
			struct timespec __user *timeout)
{
789
790
	ktime_t until = { .tv64 = KTIME_MAX };
	long ret = 0;
Linus Torvalds's avatar
Linus Torvalds committed
791
792
793

	if (timeout) {
		struct timespec	ts;
794

Linus Torvalds's avatar
Linus Torvalds committed
795
		if (unlikely(copy_from_user(&ts, timeout, sizeof(ts))))
796
			return -EFAULT;
Linus Torvalds's avatar
Linus Torvalds committed
797

798
		until = timespec_to_ktime(ts);
Linus Torvalds's avatar
Linus Torvalds committed
799
800
	}

801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
	/*
	 * 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
817

818
819
	if (!ret && signal_pending(current))
		ret = -EINTR;
Linus Torvalds's avatar
Linus Torvalds committed
820

821
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
}

/* 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.
 */
837
SYSCALL_DEFINE2(io_setup, unsigned, nr_events, aio_context_t __user *, ctxp)
Linus Torvalds's avatar
Linus Torvalds committed
838
839
840
841
842
843
844
845
846
847
{
	struct kioctx *ioctx = NULL;
	unsigned long ctx;
	long ret;

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

	ret = -EINVAL;
848
849
850
	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
851
852
853
854
855
856
857
		goto out;
	}

	ioctx = ioctx_alloc(nr_events);
	ret = PTR_ERR(ioctx);
	if (!IS_ERR(ioctx)) {
		ret = put_user(ioctx->user_id, ctxp);
858
		if (ret)
Kent Overstreet's avatar
Kent Overstreet committed
859
			kill_ioctx(ioctx);
860
		put_ioctx(ioctx);
Linus Torvalds's avatar
Linus Torvalds committed
861
862
863
864
865
866
867
868
869
	}

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
870
 *	implemented.  May fail with -EINVAL if the context pointed to
Linus Torvalds's avatar
Linus Torvalds committed
871
872
 *	is invalid.
 */
873
SYSCALL_DEFINE1(io_destroy, aio_context_t, ctx)
Linus Torvalds's avatar
Linus Torvalds committed
874
875
876
{
	struct kioctx *ioctx = lookup_ioctx(ctx);
	if (likely(NULL != ioctx)) {
Kent Overstreet's avatar
Kent Overstreet committed
877
		kill_ioctx(ioctx);
878
		put_ioctx(ioctx);
Linus Torvalds's avatar
Linus Torvalds committed
879
880
881
882
883
884
		return 0;
	}
	pr_debug("EINVAL: io_destroy: invalid context id\n");
	return -EINVAL;
}

885
static void aio_advance_iovec(struct kiocb *iocb, ssize_t ret)
Linus Torvalds's avatar
Linus Torvalds committed
886
{
887
888
889
890
891
892
893
894
895
896
897
898
899
	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++;
900
		}
901
	}
Linus Torvalds's avatar
Linus Torvalds committed
902

903
904
905
	/* 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
906
907
}

Kent Overstreet's avatar
Kent Overstreet committed
908
909
910
911
typedef ssize_t (aio_rw_op)(struct kiocb *, const struct iovec *,
			    unsigned long, loff_t);

static ssize_t aio_rw_vect_retry(struct kiocb *iocb, int rw, aio_rw_op *rw_op)
Linus Torvalds's avatar
Linus Torvalds committed
912
913
{
	struct file *file = iocb->ki_filp;
914
915
	struct address_space *mapping = file->f_mapping;
	struct inode *inode = mapping->host;
Linus Torvalds's avatar
Linus Torvalds committed
916
917
	ssize_t ret = 0;

918
919
920
921
	/* This matches the pread()/pwrite() logic */
	if (iocb->ki_pos < 0)
		return -EINVAL;

Kent Overstreet's avatar
Kent Overstreet committed
922
	if (rw == WRITE)
923
		file_start_write(file);
924
	do {
925
926
927
928
929
930
931
932
933
		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 &&
Kent Overstreet's avatar
Kent Overstreet committed
934
		 (rw == WRITE ||
935
		  (!S_ISFIFO(inode->i_mode) && !S_ISSOCK(inode->i_mode))));
Kent Overstreet's avatar
Kent Overstreet committed
936
	if (rw == WRITE)
937
		file_end_write(file);
Linus Torvalds's avatar
Linus Torvalds committed
938

939
940
	/* This means we must have transferred all that we could */
	/* No need to retry anymore */
Linus Torvalds's avatar
Linus Torvalds committed
941
942
943
	if ((ret == 0) || (iocb->ki_left == 0))
		ret = iocb->ki_nbytes - iocb->ki_left;

944
945
	/* If we managed to write some out we return that, rather than
	 * the eventual error. */
Kent Overstreet's avatar
Kent Overstreet committed
946
	if (rw == WRITE
Zach Brown's avatar
Zach Brown committed
947
	    && ret < 0 && ret != -EIOCBQUEUED
948
949
950
	    && iocb->ki_nbytes - iocb->ki_left)
		ret = iocb->ki_nbytes - iocb->ki_left;

Linus Torvalds's avatar
Linus Torvalds committed
951
952
953
	return ret;
}

Kent Overstreet's avatar
Kent Overstreet committed
954
static ssize_t aio_setup_vectored_rw(int rw, struct kiocb *kiocb, bool compat)
955
956
957
{
	ssize_t ret;

Kent Overstreet's avatar
Kent Overstreet committed
958
959
	kiocb->ki_nr_segs = kiocb->ki_nbytes;

960
961
#ifdef CONFIG_COMPAT
	if (compat)
Kent Overstreet's avatar
Kent Overstreet committed
962
		ret = compat_rw_copy_check_uvector(rw,
963
				(struct compat_iovec __user *)kiocb->ki_buf,
Kent Overstreet's avatar
Kent Overstreet committed
964
				kiocb->ki_nr_segs, 1, &kiocb->ki_inline_vec,
965
				&kiocb->ki_iovec);
966
967
	else
#endif
Kent Overstreet's avatar
Kent Overstreet committed
968
		ret = rw_copy_check_uvector(rw,
969
				(struct iovec __user *)kiocb->ki_buf,
Kent Overstreet's avatar
Kent Overstreet committed
970
				kiocb->ki_nr_segs, 1, &kiocb->ki_inline_vec,
971
				&kiocb->ki_iovec);
972
	if (ret < 0)
Kent Overstreet's avatar
Kent Overstreet committed
973
		return ret;
974

Kent Overstreet's avatar
Kent Overstreet committed
975
	/* ki_nbytes now reflect bytes instead of segs */
976
	kiocb->ki_nbytes = ret;
Kent Overstreet's avatar
Kent Overstreet committed
977
	return 0;
978
979
}

Kent Overstreet's avatar
Kent Overstreet committed
980
static ssize_t aio_setup_single_vector(int rw, struct kiocb *kiocb)
981
{
Kent Overstreet's avatar
Kent Overstreet committed
982
983
	if (unlikely(!access_ok(!rw, kiocb->ki_buf, kiocb->ki_nbytes)))
		return -EFAULT;
984

985
986
	kiocb->ki_iovec = &kiocb->ki_inline_vec;
	kiocb->ki_iovec->iov_base = kiocb->ki_buf;
Kent Overstreet's avatar
Kent Overstreet committed
987
	kiocb->ki_iovec->iov_len = kiocb->ki_nbytes;
988
989
990
991
	kiocb->ki_nr_segs = 1;
	return 0;
}

Linus Torvalds's avatar
Linus Torvalds committed
992
993
994
995
996
/*
 * aio_setup_iocb:
 *	Performs the initial checks and aio retry method
 *	setup for the kiocb at the time of io submission.
 */
Kent Overstreet's avatar
Kent Overstreet committed
997
static ssize_t aio_run_iocb(struct kiocb *req, bool compat)
Linus Torvalds's avatar
Linus Torvalds committed
998
{
Kent Overstreet's avatar
Kent Overstreet committed
999
1000
1001
1002
1003
	struct file *file = req->ki_filp;
	ssize_t ret;
	int rw;
	fmode_t mode;
	aio_rw_op *rw_op;
Linus Torvalds's avatar
Linus Torvalds committed
1004

Kent Overstreet's avatar
Kent Overstreet committed
1005
	switch (req->ki_opcode) {
Linus Torvalds's avatar
Linus Torvalds committed
1006
	case IOCB_CMD_PREAD:
1007
	case IOCB_CMD_PREADV:
Kent Overstreet's avatar
Kent Overstreet committed
1008
1009
1010
1011
1012
1013
		mode	= FMODE_READ;
		rw	= READ;
		rw_op	= file->f_op->aio_read;
		goto rw_common;

	case IOCB_CMD_PWRITE:
1014
	case IOCB_CMD_PWRITEV:
Kent Overstreet's avatar
Kent Overstreet committed
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
		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;

		ret = (req->ki_opcode == IOCB_CMD_PREADV ||
		       req->ki_opcode == IOCB_CMD_PWRITEV)
			? aio_setup_vectored_rw(rw, req, compat)
			: aio_setup_single_vector(rw, req);
1030
		if (ret)
Kent Overstreet's avatar
Kent Overstreet committed
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
			return ret;

		ret = rw_verify_area(rw, file, &req->ki_pos, req->ki_nbytes);
		if (ret < 0)
			return ret;

		req->ki_nbytes = ret;
		req->ki_left = ret;

		ret = aio_rw_vect_retry(req, rw, rw_op);
Linus Torvalds's avatar
Linus Torvalds committed
1041
		break;
Kent Overstreet's avatar
Kent Overstreet committed
1042

Linus Torvalds's avatar
Linus Torvalds committed
1043
	case IOCB_CMD_FDSYNC:
Kent Overstreet's avatar
Kent Overstreet committed
1044
1045
1046
1047
		if (!file->f_op->aio_fsync)
			return -EINVAL;

		ret = file->f_op->aio_fsync(req, 1);
Linus Torvalds's avatar
Linus Torvalds committed
1048
		break;
Kent Overstreet's avatar
Kent Overstreet committed
1049

Linus Torvalds's avatar
Linus Torvalds committed
1050
	case IOCB_CMD_FSYNC:
Kent Overstreet's avatar
Kent Overstreet committed
1051
1052
1053
1054
		if (!file->f_op->aio_fsync)
			return -EINVAL;

		ret = file->f_op->aio_fsync(req, 0);
Linus Torvalds's avatar
Linus Torvalds committed
1055
		break;
Kent Overstreet's avatar
Kent Overstreet committed
1056

Linus Torvalds's avatar
Linus Torvalds committed
1057
	default:
Kent Overstreet's avatar
Kent Overstreet committed
1058
		pr_debug("EINVAL: no operation provided\n");
Kent Overstreet's avatar
Kent Overstreet committed
1059
		return -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
1060
1061
	}

Kent Overstreet's avatar
Kent Overstreet committed
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
	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);
	}
Linus Torvalds's avatar
Linus Torvalds committed
1073
1074
1075
1076

	return 0;
}

1077
static int io_submit_one(struct kioctx *ctx, struct iocb __user *user_iocb,
Kent Overstreet's avatar
Kent Overstreet committed
1078
			 struct iocb *iocb, bool compat)
Linus Torvalds's avatar
Linus Torvalds committed
1079
1080
1081
1082
1083
{
	struct kiocb *req;
	ssize_t ret;

	/* enforce forwards compatibility on users */
1084
	if (unlikely(iocb->aio_reserved1 || iocb->aio_reserved2)) {
Kent Overstreet's avatar
Kent Overstreet committed
1085
		pr_debug("EINVAL: reserve field set\n");
Linus Torvalds's avatar
Linus Torvalds committed
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
		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
1099
	req = aio_get_req(ctx);
1100
	if (unlikely(!req))
Linus Torvalds's avatar
Linus Torvalds committed
1101
		return -EAGAIN;
1102
1103
1104
1105
1106

	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
1107
	}
1108

1109
1110
1111
1112
1113
1114
1115
	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.
		 */
1116
		req->ki_eventfd = eventfd_ctx_fdget((int) iocb->aio_resfd);
1117
		if (IS_ERR(req->ki_eventfd)) {
1118
			ret = PTR_ERR(req->ki_eventfd);
1119
			req->ki_eventfd = NULL;
1120
1121
1122
			goto out_put_req;
		}
	}
Linus Torvalds's avatar
Linus Torvalds committed
1123

Kent Overstreet's avatar
Kent Overstreet committed
1124
	ret = put_user(KIOCB_KEY, &user_iocb->aio_key);
Linus Torvalds's avatar
Linus Torvalds committed
1125
	if (unlikely(ret)) {
Kent Overstreet's avatar
Kent Overstreet committed
1126
		pr_debug("EFAULT: aio_key\n");
Linus Torvalds's avatar
Linus Torvalds committed
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
		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;

Kent Overstreet's avatar
Kent Overstreet committed
1138
	ret = aio_run_iocb(req, compat);
Zach Brown's avatar
Zach Brown committed
1139
	if (ret)
1140
		goto out_put_req;
Zach Brown's avatar
Zach Brown committed
1141

Linus Torvalds's avatar
Linus Torvalds committed
1142
1143
1144
	aio_put_req(req);	/* drop extra ref to req */
	return 0;
out_put_req: