aio.c 32.7 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>

Al Viro's avatar
Al Viro committed
42
43
#include "internal.h"

44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
#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
66
	atomic_t		dead;
67
68
69
70
71

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

72
73
74
75
	/*
	 * This is what userspace passed to io_setup(), it's not used for
	 * anything but counting against the global max_reqs quota.
	 *
76
	 * The real limit is nr_events - 1, which will be larger (see
77
78
	 * aio_setup_ring())
	 */
79
80
	unsigned		max_reqs;

81
82
	/* Size of ringbuffer, in units of struct io_event */
	unsigned		nr_events;
83

84
85
86
87
88
89
	unsigned long		mmap_base;
	unsigned long		mmap_size;

	struct page		**ring_pages;
	long			nr_pages;

90
91
92
93
94
95
96
97
98
99
100
101
	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;

102
103
	struct {
		struct mutex	ring_lock;
104
105
		wait_queue_head_t wait;
	} ____cacheline_aligned_in_smp;
106
107
108
109

	struct {
		unsigned	tail;
		spinlock_t	completion_lock;
110
	} ____cacheline_aligned_in_smp;
111
112

	struct page		*internal_pages[AIO_RING_PAGES];
113
114
};

Linus Torvalds's avatar
Linus Torvalds committed
115
/*------ sysctl variables----*/
116
117
118
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
119
120
/*----end sysctl variables---*/

121
122
static struct kmem_cache	*kiocb_cachep;
static struct kmem_cache	*kioctx_cachep;
Linus Torvalds's avatar
Linus Torvalds committed
123
124
125
126
127
128
129

/* 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)
{
130
131
	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
132

Kent Overstreet's avatar
Kent Overstreet committed
133
	pr_debug("sizeof(struct page) = %zu\n", sizeof(struct page));
Linus Torvalds's avatar
Linus Torvalds committed
134
135
136

	return 0;
}
137
__initcall(aio_setup);
Linus Torvalds's avatar
Linus Torvalds committed
138
139
140
141
142

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

143
144
	for (i = 0; i < ctx->nr_pages; i++)
		put_page(ctx->ring_pages[i]);
Linus Torvalds's avatar
Linus Torvalds committed
145

146
147
	if (ctx->ring_pages && ctx->ring_pages != ctx->internal_pages)
		kfree(ctx->ring_pages);
Linus Torvalds's avatar
Linus Torvalds committed
148
149
150
151
152
153
}

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
154
	struct mm_struct *mm = current->mm;
155
	unsigned long size, populate;
Linus Torvalds's avatar
Linus Torvalds committed
156
157
158
159
160
161
162
163
164
165
166
167
168
169
	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);

170
171
	ctx->nr_events = 0;
	ctx->ring_pages = ctx->internal_pages;
Linus Torvalds's avatar
Linus Torvalds committed
172
	if (nr_pages > AIO_RING_PAGES) {
173
174
175
		ctx->ring_pages = kcalloc(nr_pages, sizeof(struct page *),
					  GFP_KERNEL);
		if (!ctx->ring_pages)
Linus Torvalds's avatar
Linus Torvalds committed
176
177
178
			return -ENOMEM;
	}

179
180
	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
181
	down_write(&mm->mmap_sem);
182
183
184
185
	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
186
		up_write(&mm->mmap_sem);
187
		ctx->mmap_size = 0;
Linus Torvalds's avatar
Linus Torvalds committed
188
189
190
191
		aio_free_ring(ctx);
		return -EAGAIN;
	}

192
193
194
	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
195
	up_write(&mm->mmap_sem);
Linus Torvalds's avatar
Linus Torvalds committed
196

197
	if (unlikely(ctx->nr_pages != nr_pages)) {
Linus Torvalds's avatar
Linus Torvalds committed
198
199
200
		aio_free_ring(ctx);
		return -EAGAIN;
	}
201
	if (populate)
202
		mm_populate(ctx->mmap_base, populate);
Linus Torvalds's avatar
Linus Torvalds committed
203

204
205
	ctx->user_id = ctx->mmap_base;
	ctx->nr_events = nr_events; /* trusted copy */
Linus Torvalds's avatar
Linus Torvalds committed
206

207
	ring = kmap_atomic(ctx->ring_pages[0]);
Linus Torvalds's avatar
Linus Torvalds committed
208
209
210
211
212
213
214
	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);
215
	kunmap_atomic(ring);
216
	flush_dcache_page(ctx->ring_pages[0]);
Linus Torvalds's avatar
Linus Torvalds committed
217
218
219
220
221
222
223
224

	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)

225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
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
241
242
243
static int kiocb_cancel(struct kioctx *ctx, struct kiocb *kiocb,
			struct io_event *res)
{
244
	kiocb_cancel_fn *old, *cancel;
Kent Overstreet's avatar
Kent Overstreet committed
245
246
	int ret = -EINVAL;

247
248
249
250
251
252
253
254
255
	/*
	 * 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
256

257
258
259
		old = cancel;
		cancel = cmpxchg(&kiocb->ki_cancel, old, KIOCB_CANCELLED);
	} while (cancel != old);
Kent Overstreet's avatar
Kent Overstreet committed
260

261
262
263
264
265
266
267
268
269
	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
270
271
272
273

	return ret;
}

Kent Overstreet's avatar
Kent Overstreet committed
274
275
276
277
278
279
280
281
282
283
284
285
286
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)
{
287
	struct aio_ring *ring;
Kent Overstreet's avatar
Kent Overstreet committed
288
289
	struct io_event res;
	struct kiocb *req;
290
	unsigned head, avail;
Kent Overstreet's avatar
Kent Overstreet committed
291
292
293
294
295
296
297
298
299
300
301
302
303

	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);

304
	ring = kmap_atomic(ctx->ring_pages[0]);
305
306
307
308
	head = ring->head;
	kunmap_atomic(ring);

	while (atomic_read(&ctx->reqs_active) > 0) {
309
310
311
		wait_event(ctx->wait,
				head != ctx->tail ||
				atomic_read(&ctx->reqs_active) <= 0);
312

313
		avail = (head <= ctx->tail ? ctx->tail : ctx->nr_events) - head;
314
315
316

		atomic_sub(avail, &ctx->reqs_active);
		head += avail;
317
		head %= ctx->nr_events;
318
319
320
	}

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

	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);
}

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
342
343
344
345
346
/* 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
347
	struct mm_struct *mm = current->mm;
Linus Torvalds's avatar
Linus Torvalds committed
348
	struct kioctx *ctx;
349
	int err = -ENOMEM;
Linus Torvalds's avatar
Linus Torvalds committed
350
351
352
353
354
355
356
357

	/* 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);
	}

358
	if (!nr_events || (unsigned long)nr_events > aio_max_nr)
Linus Torvalds's avatar
Linus Torvalds committed
359
360
		return ERR_PTR(-EAGAIN);

361
	ctx = kmem_cache_zalloc(kioctx_cachep, GFP_KERNEL);
Linus Torvalds's avatar
Linus Torvalds committed
362
363
364
365
366
	if (!ctx)
		return ERR_PTR(-ENOMEM);

	ctx->max_reqs = nr_events;

Al Viro's avatar
Al Viro committed
367
	atomic_set(&ctx->users, 2);
Kent Overstreet's avatar
Kent Overstreet committed
368
	atomic_set(&ctx->dead, 0);
Linus Torvalds's avatar
Linus Torvalds committed
369
	spin_lock_init(&ctx->ctx_lock);
370
	spin_lock_init(&ctx->completion_lock);
371
	mutex_init(&ctx->ring_lock);
Linus Torvalds's avatar
Linus Torvalds committed
372
373
374
375
376
377
378
379
	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 */
380
	spin_lock(&aio_nr_lock);
381
382
	if (aio_nr + nr_events > aio_max_nr ||
	    aio_nr + nr_events < aio_nr) {
383
		spin_unlock(&aio_nr_lock);
Linus Torvalds's avatar
Linus Torvalds committed
384
		goto out_cleanup;
385
386
	}
	aio_nr += ctx->max_reqs;
387
	spin_unlock(&aio_nr_lock);
Linus Torvalds's avatar
Linus Torvalds committed
388

Jeff Moyer's avatar
Jeff Moyer committed
389
	/* now link into global list. */
390
391
392
	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
393

Kent Overstreet's avatar
Kent Overstreet committed
394
	pr_debug("allocated ioctx %p[%ld]: mm=%p mask=0x%x\n",
395
		 ctx, ctx->user_id, mm, ctx->nr_events);
Linus Torvalds's avatar
Linus Torvalds committed
396
397
398
	return ctx;

out_cleanup:
399
400
	err = -EAGAIN;
	aio_free_ring(ctx);
Linus Torvalds's avatar
Linus Torvalds committed
401
402
out_freectx:
	kmem_cache_free(kioctx_cachep, ctx);
Kent Overstreet's avatar
Kent Overstreet committed
403
	pr_debug("error allocating ioctx %d\n", err);
404
	return ERR_PTR(err);
Linus Torvalds's avatar
Linus Torvalds committed
405
406
}

Kent Overstreet's avatar
Kent Overstreet committed
407
static void kill_ioctx_work(struct work_struct *work)
Linus Torvalds's avatar
Linus Torvalds committed
408
{
Kent Overstreet's avatar
Kent Overstreet committed
409
	struct kioctx *ctx = container_of(work, struct kioctx, rcu_work);
410

Kent Overstreet's avatar
Kent Overstreet committed
411
412
413
	wake_up_all(&ctx->wait);
	put_ioctx(ctx);
}
Kent Overstreet's avatar
Kent Overstreet committed
414

Kent Overstreet's avatar
Kent Overstreet committed
415
416
417
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
418

Kent Overstreet's avatar
Kent Overstreet committed
419
420
421
	INIT_WORK(&ctx->rcu_work, kill_ioctx_work);
	schedule_work(&ctx->rcu_work);
}
Linus Torvalds's avatar
Linus Torvalds committed
422

Kent Overstreet's avatar
Kent Overstreet committed
423
424
425
426
427
428
429
430
431
/* 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);
432

Kent Overstreet's avatar
Kent Overstreet committed
433
		/*
434
435
436
437
438
		 * 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
439
		 */
440
441
442
443
444
445
446
447
448
449
		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);

		/* Between hlist_del_rcu() and dropping the initial ref */
		call_rcu(&ctx->rcu_head, kill_ioctx_rcu);
Kent Overstreet's avatar
Kent Overstreet committed
450
	}
Linus Torvalds's avatar
Linus Torvalds committed
451
452
453
454
455
}

/* wait_on_sync_kiocb:
 *	Waits on the given sync kiocb to complete.
 */
456
ssize_t wait_on_sync_kiocb(struct kiocb *iocb)
Linus Torvalds's avatar
Linus Torvalds committed
457
{
458
	while (atomic_read(&iocb->ki_users)) {
Linus Torvalds's avatar
Linus Torvalds committed
459
		set_current_state(TASK_UNINTERRUPTIBLE);
460
		if (!atomic_read(&iocb->ki_users))
Linus Torvalds's avatar
Linus Torvalds committed
461
			break;
462
		io_schedule();
Linus Torvalds's avatar
Linus Torvalds committed
463
464
465
466
	}
	__set_current_state(TASK_RUNNING);
	return iocb->ki_user_data;
}
467
EXPORT_SYMBOL(wait_on_sync_kiocb);
Linus Torvalds's avatar
Linus Torvalds committed
468

Kent Overstreet's avatar
Kent Overstreet committed
469
470
471
472
473
474
475
/*
 * 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
476
 */
477
void exit_aio(struct mm_struct *mm)
Linus Torvalds's avatar
Linus Torvalds committed
478
{
479
	struct kioctx *ctx;
Kent Overstreet's avatar
Kent Overstreet committed
480
	struct hlist_node *n;
481

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

499
		kill_ioctx(ctx);
Linus Torvalds's avatar
Linus Torvalds committed
500
501
502
503
	}
}

/* aio_get_req
504
 *	Allocate a slot for an aio request.  Increments the ki_users count
Linus Torvalds's avatar
Linus Torvalds committed
505
506
507
 * of the kioctx so that the kioctx stays around until all requests are
 * complete.  Returns NULL if no requests are free.
 *
508
 * Returns with kiocb->ki_users set to 2.  The io submit code path holds
Linus Torvalds's avatar
Linus Torvalds committed
509
510
511
512
 * 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
513
static inline struct kiocb *aio_get_req(struct kioctx *ctx)
Linus Torvalds's avatar
Linus Torvalds committed
514
{
Kent Overstreet's avatar
Kent Overstreet committed
515
516
	struct kiocb *req;

517
	if (atomic_read(&ctx->reqs_active) >= ctx->nr_events)
Kent Overstreet's avatar
Kent Overstreet committed
518
519
		return NULL;

520
	if (atomic_inc_return(&ctx->reqs_active) > ctx->nr_events - 1)
Kent Overstreet's avatar
Kent Overstreet committed
521
		goto out_put;
Linus Torvalds's avatar
Linus Torvalds committed
522

523
	req = kmem_cache_alloc(kiocb_cachep, GFP_KERNEL|__GFP_ZERO);
Linus Torvalds's avatar
Linus Torvalds committed
524
	if (unlikely(!req))
Kent Overstreet's avatar
Kent Overstreet committed
525
		goto out_put;
Linus Torvalds's avatar
Linus Torvalds committed
526

527
	atomic_set(&req->ki_users, 2);
Linus Torvalds's avatar
Linus Torvalds committed
528
529
	req->ki_ctx = ctx;

Jeff Moyer's avatar
Jeff Moyer committed
530
	return req;
Kent Overstreet's avatar
Kent Overstreet committed
531
532
533
out_put:
	atomic_dec(&ctx->reqs_active);
	return NULL;
Linus Torvalds's avatar
Linus Torvalds committed
534
535
}

536
static void kiocb_free(struct kiocb *req)
Linus Torvalds's avatar
Linus Torvalds committed
537
{
538
539
	if (req->ki_filp)
		fput(req->ki_filp);
540
541
	if (req->ki_eventfd != NULL)
		eventfd_ctx_put(req->ki_eventfd);
Linus Torvalds's avatar
Linus Torvalds committed
542
543
	if (req->ki_dtor)
		req->ki_dtor(req);
544
545
	if (req->ki_iovec != &req->ki_inline_vec)
		kfree(req->ki_iovec);
Linus Torvalds's avatar
Linus Torvalds committed
546
547
548
	kmem_cache_free(kiocb_cachep, req);
}

549
void aio_put_req(struct kiocb *req)
Linus Torvalds's avatar
Linus Torvalds committed
550
{
551
552
	if (atomic_dec_and_test(&req->ki_users))
		kiocb_free(req);
Linus Torvalds's avatar
Linus Torvalds committed
553
}
554
EXPORT_SYMBOL(aio_put_req);
Linus Torvalds's avatar
Linus Torvalds committed
555

556
static struct kioctx *lookup_ioctx(unsigned long ctx_id)
Linus Torvalds's avatar
Linus Torvalds committed
557
{
558
	struct mm_struct *mm = current->mm;
559
	struct kioctx *ctx, *ret = NULL;
Linus Torvalds's avatar
Linus Torvalds committed
560

561
562
	rcu_read_lock();

563
	hlist_for_each_entry_rcu(ctx, &mm->ioctx_list, list) {
Kent Overstreet's avatar
Kent Overstreet committed
564
565
		if (ctx->user_id == ctx_id) {
			atomic_inc(&ctx->users);
566
			ret = ctx;
Linus Torvalds's avatar
Linus Torvalds committed
567
568
			break;
		}
569
	}
Linus Torvalds's avatar
Linus Torvalds committed
570

571
	rcu_read_unlock();
572
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
573
574
575
576
577
}

/* aio_complete
 *	Called when the io request on the given iocb is complete.
 */
578
void aio_complete(struct kiocb *iocb, long res, long res2)
Linus Torvalds's avatar
Linus Torvalds committed
579
580
581
{
	struct kioctx	*ctx = iocb->ki_ctx;
	struct aio_ring	*ring;
Kent Overstreet's avatar
Kent Overstreet committed
582
	struct io_event	*ev_page, *event;
Linus Torvalds's avatar
Linus Torvalds committed
583
	unsigned long	flags;
Kent Overstreet's avatar
Kent Overstreet committed
584
	unsigned tail, pos;
Linus Torvalds's avatar
Linus Torvalds committed
585

586
587
588
589
590
591
	/*
	 * 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
592
593
	 */
	if (is_sync_kiocb(iocb)) {
594
		BUG_ON(atomic_read(&iocb->ki_users) != 1);
Linus Torvalds's avatar
Linus Torvalds committed
595
		iocb->ki_user_data = res;
596
		atomic_set(&iocb->ki_users, 0);
Linus Torvalds's avatar
Linus Torvalds committed
597
		wake_up_process(iocb->ki_obj.tsk);
598
		return;
Linus Torvalds's avatar
Linus Torvalds committed
599
600
	}

Kent Overstreet's avatar
Kent Overstreet committed
601
602
603
	/*
	 * 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
604
	 */
Kent Overstreet's avatar
Kent Overstreet committed
605
	rcu_read_lock();
Linus Torvalds's avatar
Linus Torvalds committed
606

607
608
609
610
611
612
613
	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);
	}
614

Linus Torvalds's avatar
Linus Torvalds committed
615
616
617
618
	/*
	 * cancelled requests don't get events, userland was given one
	 * when the event got cancelled.
	 */
619
	if (unlikely(xchg(&iocb->ki_cancel,
620
621
622
			  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
623
		goto put_rq;
624
	}
Linus Torvalds's avatar
Linus Torvalds committed
625

626
627
	/*
	 * Add a completion event to the ring buffer. Must be done holding
628
	 * ctx->completion_lock to prevent other code from messing with the tail
629
630
631
632
	 * pointer since we might be called from irq context.
	 */
	spin_lock_irqsave(&ctx->completion_lock, flags);

633
	tail = ctx->tail;
Kent Overstreet's avatar
Kent Overstreet committed
634
635
	pos = tail + AIO_EVENTS_OFFSET;

636
	if (++tail >= ctx->nr_events)
637
		tail = 0;
Linus Torvalds's avatar
Linus Torvalds committed
638

639
	ev_page = kmap_atomic(ctx->ring_pages[pos / AIO_EVENTS_PER_PAGE]);
Kent Overstreet's avatar
Kent Overstreet committed
640
641
	event = ev_page + pos % AIO_EVENTS_PER_PAGE;

Linus Torvalds's avatar
Linus Torvalds committed
642
643
644
645
646
	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
647
	kunmap_atomic(ev_page);
648
	flush_dcache_page(ctx->ring_pages[pos / AIO_EVENTS_PER_PAGE]);
Kent Overstreet's avatar
Kent Overstreet committed
649
650

	pr_debug("%p[%u]: %p: %p %Lx %lx %lx\n",
Kent Overstreet's avatar
Kent Overstreet committed
651
652
		 ctx, tail, iocb, iocb->ki_obj.user, iocb->ki_user_data,
		 res, res2);
Linus Torvalds's avatar
Linus Torvalds committed
653
654
655
656
657
658

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

659
	ctx->tail = tail;
Linus Torvalds's avatar
Linus Torvalds committed
660

661
	ring = kmap_atomic(ctx->ring_pages[0]);
Kent Overstreet's avatar
Kent Overstreet committed
662
	ring->tail = tail;
663
	kunmap_atomic(ring);
664
	flush_dcache_page(ctx->ring_pages[0]);
Linus Torvalds's avatar
Linus Torvalds committed
665

666
667
	spin_unlock_irqrestore(&ctx->completion_lock, flags);

Kent Overstreet's avatar
Kent Overstreet committed
668
	pr_debug("added to ring %p at [%u]\n", iocb, tail);
Davide Libenzi's avatar
Davide Libenzi committed
669
670
671
672
673
674

	/*
	 * 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.
	 */
675
	if (iocb->ki_eventfd != NULL)
Davide Libenzi's avatar
Davide Libenzi committed
676
677
		eventfd_signal(iocb->ki_eventfd, 1);

Linus Torvalds's avatar
Linus Torvalds committed
678
679
put_rq:
	/* everything turned out well, dispose of the aiocb. */
680
	aio_put_req(iocb);
Linus Torvalds's avatar
Linus Torvalds committed
681

682
683
684
685
686
687
688
689
	/*
	 * 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
690
691
692
	if (waitqueue_active(&ctx->wait))
		wake_up(&ctx->wait);

Kent Overstreet's avatar
Kent Overstreet committed
693
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
694
}
695
EXPORT_SYMBOL(aio_complete);
Linus Torvalds's avatar
Linus Torvalds committed
696

697
698
699
/* 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
700
 */
701
702
static long aio_read_events_ring(struct kioctx *ctx,
				 struct io_event __user *event, long nr)
Linus Torvalds's avatar
Linus Torvalds committed
703
704
{
	struct aio_ring *ring;
705
706
707
708
	unsigned head, pos;
	long ret = 0;
	int copy_ret;

709
	mutex_lock(&ctx->ring_lock);
Linus Torvalds's avatar
Linus Torvalds committed
710

711
	ring = kmap_atomic(ctx->ring_pages[0]);
712
713
714
	head = ring->head;
	kunmap_atomic(ring);

715
	pr_debug("h%u t%u m%u\n", head, ctx->tail, ctx->nr_events);
Linus Torvalds's avatar
Linus Torvalds committed
716

717
	if (head == ctx->tail)
Linus Torvalds's avatar
Linus Torvalds committed
718
719
		goto out;

720
721
722
723
724
	while (ret < nr) {
		long avail;
		struct io_event *ev;
		struct page *page;

725
726
		avail = (head <= ctx->tail ? ctx->tail : ctx->nr_events) - head;
		if (head == ctx->tail)
727
728
729
730
731
732
733
			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;
734
		page = ctx->ring_pages[pos / AIO_EVENTS_PER_PAGE];
735
736
737
738
739
740
741
742
743
744
745
746
747
748
		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;
749
		head %= ctx->nr_events;
Linus Torvalds's avatar
Linus Torvalds committed
750
751
	}

752
	ring = kmap_atomic(ctx->ring_pages[0]);
753
	ring->head = head;
754
	kunmap_atomic(ring);
755
	flush_dcache_page(ctx->ring_pages[0]);
756

757
	pr_debug("%li  h%u t%u\n", ret, head, ctx->tail);
758
759

	atomic_sub(ret, &ctx->reqs_active);
760
out:
761
	mutex_unlock(&ctx->ring_lock);
762

Linus Torvalds's avatar
Linus Torvalds committed
763
764
765
	return ret;
}

766
767
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
768
{
769
	long ret = aio_read_events_ring(ctx, event + *i, nr - *i);
Linus Torvalds's avatar
Linus Torvalds committed
770

771
772
	if (ret > 0)
		*i += ret;
Linus Torvalds's avatar
Linus Torvalds committed
773

774
775
	if (unlikely(atomic_read(&ctx->dead)))
		ret = -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
776

777
778
	if (!*i)
		*i = ret;
Linus Torvalds's avatar
Linus Torvalds committed
779

780
	return ret < 0 || *i >= min_nr;
Linus Torvalds's avatar
Linus Torvalds committed
781
782
}

783
static long read_events(struct kioctx *ctx, long min_nr, long nr,
Linus Torvalds's avatar
Linus Torvalds committed
784
785
786
			struct io_event __user *event,
			struct timespec __user *timeout)
{
787
788
	ktime_t until = { .tv64 = KTIME_MAX };
	long ret = 0;
Linus Torvalds's avatar
Linus Torvalds committed
789
790
791

	if (timeout) {
		struct timespec	ts;
792

Linus Torvalds's avatar
Linus Torvalds committed
793
		if (unlikely(copy_from_user(&ts, timeout, sizeof(ts))))
794
			return -EFAULT;
Linus Torvalds's avatar
Linus Torvalds committed
795

796
		until = timespec_to_ktime(ts);
Linus Torvalds's avatar
Linus Torvalds committed
797
798
	}

799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
	/*
	 * 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
815

816
817
	if (!ret && signal_pending(current))
		ret = -EINTR;
Linus Torvalds's avatar
Linus Torvalds committed
818

819
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
}

/* 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.
 */
835
SYSCALL_DEFINE2(io_setup, unsigned, nr_events, aio_context_t __user *, ctxp)
Linus Torvalds's avatar
Linus Torvalds committed
836
837
838
839
840
841
842
843
844
845
{
	struct kioctx *ioctx = NULL;
	unsigned long ctx;
	long ret;

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

	ret = -EINVAL;
846
847
848
	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
849
850
851
852
853
854
855
		goto out;
	}

	ioctx = ioctx_alloc(nr_events);
	ret = PTR_ERR(ioctx);
	if (!IS_ERR(ioctx)) {
		ret = put_user(ioctx->user_id, ctxp);
856
		if (ret)
Kent Overstreet's avatar
Kent Overstreet committed
857
			kill_ioctx(ioctx);
858
		put_ioctx(ioctx);
Linus Torvalds's avatar
Linus Torvalds committed
859
860
861
862
863
864
865
866
867
	}

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
868
 *	implemented.  May fail with -EINVAL if the context pointed to
Linus Torvalds's avatar
Linus Torvalds committed
869
870
 *	is invalid.
 */
871
SYSCALL_DEFINE1(io_destroy, aio_context_t, ctx)
Linus Torvalds's avatar
Linus Torvalds committed
872
873
874
{
	struct kioctx *ioctx = lookup_ioctx(ctx);
	if (likely(NULL != ioctx)) {
Kent Overstreet's avatar
Kent Overstreet committed
875
		kill_ioctx(ioctx);
876
		put_ioctx(ioctx);
Linus Torvalds's avatar
Linus Torvalds committed
877
878
879
880
881
882
		return 0;
	}
	pr_debug("EINVAL: io_destroy: invalid context id\n");
	return -EINVAL;
}

883
static void aio_advance_iovec(struct kiocb *iocb, ssize_t ret)
Linus Torvalds's avatar
Linus Torvalds committed
884
{
885
886
887
888
889
890
891
892
893
894
895
896
897
	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++;
898
		}
899
	}
Linus Torvalds's avatar
Linus Torvalds committed
900

901
902
903
	/* 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
904
905
}

Kent Overstreet's avatar
Kent Overstreet committed
906
907
908
909
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
910
911
{
	struct file *file = iocb->ki_filp;
912
913
	struct address_space *mapping = file->f_mapping;
	struct inode *inode = mapping->host;
Linus Torvalds's avatar
Linus Torvalds committed
914
915
	ssize_t ret = 0;

916
917
918
919
	/* This matches the pread()/pwrite() logic */
	if (iocb->ki_pos < 0)
		return -EINVAL;

Kent Overstreet's avatar
Kent Overstreet committed
920
	if (rw == WRITE)
921
		file_start_write(file);
922
	do {
923
924
925
926
927
928
929
930
931
		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
932
		 (rw == WRITE ||
933
		  (!S_ISFIFO(inode->i_mode) && !S_ISSOCK(inode->i_mode))));
Kent Overstreet's avatar
Kent Overstreet committed
934
	if (rw == WRITE)
935
		file_end_write(file);
Linus Torvalds's avatar
Linus Torvalds committed
936

937
938
	/* This means we must have transferred all that we could */
	/* No need to retry anymore */
Linus Torvalds's avatar
Linus Torvalds committed
939
940
941
	if ((ret == 0) || (iocb->ki_left == 0))
		ret = iocb->ki_nbytes - iocb->ki_left;

942
943
	/* If we managed to write some out we return that, rather than
	 * the eventual error. */
Kent Overstreet's avatar
Kent Overstreet committed
944
	if (rw == WRITE
Zach Brown's avatar
Zach Brown committed
945
	    && ret < 0 && ret != -EIOCBQUEUED
946
947
948
	    && iocb->ki_nbytes - iocb->ki_left)
		ret = iocb->ki_nbytes - iocb->ki_left;

Linus Torvalds's avatar
Linus Torvalds committed
949
950
951
	return ret;
}

Kent Overstreet's avatar
Kent Overstreet committed
952
static ssize_t aio_setup_vectored_rw(int rw, struct kiocb *kiocb, bool compat)
953
954
955
{
	ssize_t ret;

Kent Overstreet's avatar
Kent Overstreet committed
956
957
	kiocb->ki_nr_segs = kiocb->ki_nbytes;

958
959
#ifdef CONFIG_COMPAT
	if (compat)
Kent Overstreet's avatar
Kent Overstreet committed
960
		ret = compat_rw_copy_check_uvector(rw,
961
				(struct compat_iovec __user *)kiocb->ki_buf,
Kent Overstreet's avatar
Kent Overstreet committed
962
				kiocb->ki_nr_segs, 1, &kiocb->ki_inline_vec,
963
				&kiocb->ki_iovec);
964
965
	else
#endif
Kent Overstreet's avatar
Kent Overstreet committed
966
		ret = rw_copy_check_uvector(rw,
967
				(struct iovec __user *)kiocb->ki_buf,
Kent Overstreet's avatar
Kent Overstreet committed
968
				kiocb->ki_nr_segs, 1, &kiocb->ki_inline_vec,
969
				&kiocb->ki_iovec);
970
	if (ret < 0)
Kent Overstreet's avatar
Kent Overstreet committed
971
		return ret;
972

Kent Overstreet's avatar
Kent Overstreet committed
973
	/* ki_nbytes now reflect bytes instead of segs */
974
	kiocb->ki_nbytes = ret;
Kent Overstreet's avatar
Kent Overstreet committed
975
	return 0;
976
977
}

Kent Overstreet's avatar
Kent Overstreet committed
978
static ssize_t aio_setup_single_vector(int rw, struct kiocb *kiocb)
979
{
Kent Overstreet's avatar
Kent Overstreet committed
980
981
	if (unlikely(!access_ok(!rw, kiocb->ki_buf, kiocb->ki_nbytes)))
		return -EFAULT;
982

983
984
	kiocb->ki_iovec = &kiocb->ki_inline_vec;
	kiocb->ki_iovec->iov_base = kiocb->ki_buf;
Kent Overstreet's avatar
Kent Overstreet committed
985
	kiocb->ki_iovec->iov_len = kiocb->ki_nbytes;
986
987
988
989
	kiocb->ki_nr_segs = 1;
	return 0;
}

Linus Torvalds's avatar
Linus Torvalds committed
990
991
992
993
994
/*
 * 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
995
static ssize_t aio_run_iocb(struct kiocb *req, bool compat)
Linus Torvalds's avatar
Linus Torvalds committed
996
{
Kent Overstreet's avatar
Kent Overstreet committed
997
998
999
1000
1001
	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
1002

Kent Overstreet's avatar
Kent Overstreet committed
1003
	switch (req->ki_opcode) {
Linus Torvalds's avatar
Linus Torvalds committed
1004
	case IOCB_CMD_PREAD:
1005
	case IOCB_CMD_PREADV:
Kent Overstreet's avatar
Kent Overstreet committed
1006
1007
1008
1009
1010
1011
		mode	= FMODE_READ;
		rw	= READ;
		rw_op	= file->f_op->aio_read;
		goto rw_common;

	case IOCB_CMD_PWRITE:
1012
	case IOCB_CMD_PWRITEV:
Kent Overstreet's avatar
Kent Overstreet committed
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
		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);
1028
		if (ret)
Kent Overstreet's avatar
Kent Overstreet committed
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
			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
1039
		break;
Kent Overstreet's avatar
Kent Overstreet committed
1040

Linus Torvalds's avatar
Linus Torvalds committed
1041
	case IOCB_CMD_FDSYNC:
Kent Overstreet's avatar
Kent Overstreet committed
1042
1043
1044
1045
		if (!file->f_op->aio_fsync)
			return -EINVAL;

		ret = file->f_op->aio_fsync(req, 1);
Linus Torvalds's avatar
Linus Torvalds committed
1046
		break;
Kent Overstreet's avatar
Kent Overstreet committed
1047

Linus Torvalds's avatar
Linus Torvalds committed
1048
	case IOCB_CMD_FSYNC:
Kent Overstreet's avatar
Kent Overstreet committed
1049
1050
1051
1052
		if (!file->f_op->aio_fsync)
			return -EINVAL;

		ret = file->f_op->aio_fsync(req, 0);
Linus Torvalds's avatar
Linus Torvalds committed
1053
		break;
Kent Overstreet's avatar
Kent Overstreet committed
1054

Linus Torvalds's avatar
Linus Torvalds committed
1055
	default:
Kent Overstreet's avatar
Kent Overstreet committed
1056
		pr_debug("EINVAL: no operation provided\n");
Kent Overstreet's avatar
Kent Overstreet committed
1057
		return -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
1058
1059
	}

Kent Overstreet's avatar
Kent Overstreet committed
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
	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
1071
1072
1073
1074

	return 0;
}

1075
static int io_submit_one(struct kioctx *ctx, struct iocb __user *user_iocb,
Kent Overstreet's avatar
Kent Overstreet committed
1076
			 struct iocb *iocb, bool compat)
Linus Torvalds's avatar
Linus Torvalds committed
1077
1078
1079
1080
1081
{
	struct kiocb *req;
	ssize_t ret;

	/* enforce forwards compatibility on users */
1082
	if (unlikely(iocb->aio_reserved1 || iocb->aio_reserved2)) {
Kent Overstreet's avatar
Kent Overstreet committed
1083
		pr_debug("EINVAL: reserve field set\n");
Linus Torvalds's avatar
Linus Torvalds committed
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
		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
1097
	req = aio_get_req(ctx);
1098
	if (unlikely(!req))
Linus Torvalds's avatar
Linus Torvalds committed
1099
		return -EAGAIN;
1100
1101
1102
1103
1104

	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
1105
	}
1106

1107
1108
1109
1110
1111
1112
1113
	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.
		 */
1114
		req->ki_eventfd = eventfd_ctx_fdget((int) iocb->aio_resfd);
1115
		if (IS_ERR(req->ki_eventfd)) {
1116
			ret = PTR_ERR(req->ki_eventfd);
1117
			req->ki_eventfd = NULL;
1118
1119
1120
			goto out_put_req;
		}
	}
Linus Torvalds's avatar
Linus Torvalds committed
1121

Kent Overstreet's avatar
Kent Overstreet committed
1122
	ret = put_user(KIOCB_KEY, &user_iocb->aio_key);
Linus Torvalds's avatar
Linus Torvalds committed
1123
	if (unlikely(ret)) {
Kent Overstreet's avatar
Kent Overstreet committed
1124
		pr_debug("EFAULT: aio_key\n");
Linus Torvalds's avatar
Linus Torvalds committed
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
		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
1136
	ret = aio_run_iocb(req, compat);