aio.c 35.2 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>
38
39
40
#include <linux/anon_inodes.h>
#include <linux/migrate.h>
#include <linux/ramfs.h>
Linus Torvalds's avatar
Linus Torvalds committed
41
42
43
44

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

Al Viro's avatar
Al Viro committed
45
46
#include "internal.h"

47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
#define AIO_RING_MAGIC			0xa10a10a1
#define AIO_RING_COMPAT_FEATURES	1
#define AIO_RING_INCOMPAT_FEATURES	0
struct aio_ring {
	unsigned	id;	/* kernel internal index number */
	unsigned	nr;	/* number of io_events */
	unsigned	head;
	unsigned	tail;

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


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

#define AIO_RING_PAGES	8

struct kioctx {
	atomic_t		users;
Kent Overstreet's avatar
Kent Overstreet committed
69
	atomic_t		dead;
70
71
72
73
74

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

75
76
77
78
	/*
	 * This is what userspace passed to io_setup(), it's not used for
	 * anything but counting against the global max_reqs quota.
	 *
79
	 * The real limit is nr_events - 1, which will be larger (see
80
81
	 * aio_setup_ring())
	 */
82
83
	unsigned		max_reqs;

84
85
	/* Size of ringbuffer, in units of struct io_event */
	unsigned		nr_events;
86

87
88
89
90
91
92
	unsigned long		mmap_base;
	unsigned long		mmap_size;

	struct page		**ring_pages;
	long			nr_pages;

93
94
95
96
97
98
99
100
101
102
103
104
	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;

105
106
	struct {
		struct mutex	ring_lock;
107
108
		wait_queue_head_t wait;
	} ____cacheline_aligned_in_smp;
109
110
111
112

	struct {
		unsigned	tail;
		spinlock_t	completion_lock;
113
	} ____cacheline_aligned_in_smp;
114
115

	struct page		*internal_pages[AIO_RING_PAGES];
116
	struct file		*aio_ring_file;
117
118
};

Linus Torvalds's avatar
Linus Torvalds committed
119
/*------ sysctl variables----*/
120
121
122
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
123
124
/*----end sysctl variables---*/

125
126
static struct kmem_cache	*kiocb_cachep;
static struct kmem_cache	*kioctx_cachep;
Linus Torvalds's avatar
Linus Torvalds committed
127
128
129
130
131
132
133

/* 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)
{
134
135
	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
136

Kent Overstreet's avatar
Kent Overstreet committed
137
	pr_debug("sizeof(struct page) = %zu\n", sizeof(struct page));
Linus Torvalds's avatar
Linus Torvalds committed
138
139
140

	return 0;
}
141
__initcall(aio_setup);
Linus Torvalds's avatar
Linus Torvalds committed
142
143
144

static void aio_free_ring(struct kioctx *ctx)
{
145
146
	int i;
	struct file *aio_ring_file = ctx->aio_ring_file;
Linus Torvalds's avatar
Linus Torvalds committed
147

148
149
150
	for (i = 0; i < ctx->nr_pages; i++) {
		pr_debug("pid(%d) [%d] page->count=%d\n", current->pid, i,
				page_count(ctx->ring_pages[i]));
151
		put_page(ctx->ring_pages[i]);
152
	}
Linus Torvalds's avatar
Linus Torvalds committed
153

154
155
	if (ctx->ring_pages && ctx->ring_pages != ctx->internal_pages)
		kfree(ctx->ring_pages);
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183

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

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

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

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

184
#if IS_ENABLED(CONFIG_MIGRATION)
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
static int aio_migratepage(struct address_space *mapping, struct page *new,
			struct page *old, enum migrate_mode mode)
{
	struct kioctx *ctx = mapping->private_data;
	unsigned long flags;
	unsigned idx = old->index;
	int rc;

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

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

	get_page(new);

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

	return rc;
Linus Torvalds's avatar
Linus Torvalds committed
211
}
212
#endif
Linus Torvalds's avatar
Linus Torvalds committed
213

214
215
static const struct address_space_operations aio_ctx_aops = {
	.set_page_dirty = aio_set_page_dirty,
216
#if IS_ENABLED(CONFIG_MIGRATION)
217
	.migratepage	= aio_migratepage,
218
#endif
219
220
};

Linus Torvalds's avatar
Linus Torvalds committed
221
222
223
224
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
225
	struct mm_struct *mm = current->mm;
226
	unsigned long size, populate;
Linus Torvalds's avatar
Linus Torvalds committed
227
	int nr_pages;
228
229
	int i;
	struct file *file;
Linus Torvalds's avatar
Linus Torvalds committed
230
231
232
233
234
235
236

	/* 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;

237
	nr_pages = PFN_UP(size);
Linus Torvalds's avatar
Linus Torvalds committed
238
239
240
	if (nr_pages < 0)
		return -EINVAL;

241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
	file = anon_inode_getfile_private("[aio]", &aio_ring_fops, ctx, O_RDWR);
	if (IS_ERR(file)) {
		ctx->aio_ring_file = NULL;
		return -EAGAIN;
	}

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

	for (i = 0; i < nr_pages; i++) {
		struct page *page;
		page = find_or_create_page(file->f_inode->i_mapping,
					   i, GFP_HIGHUSER | __GFP_ZERO);
		if (!page)
			break;
		pr_debug("pid(%d) page[%d]->count=%d\n",
			 current->pid, i, page_count(page));
		SetPageUptodate(page);
		SetPageDirty(page);
		unlock_page(page);
	}
	ctx->aio_ring_file = file;
	nr_events = (PAGE_SIZE * nr_pages - sizeof(struct aio_ring))
			/ sizeof(struct io_event);
Linus Torvalds's avatar
Linus Torvalds committed
266

267
	ctx->ring_pages = ctx->internal_pages;
Linus Torvalds's avatar
Linus Torvalds committed
268
	if (nr_pages > AIO_RING_PAGES) {
269
270
271
		ctx->ring_pages = kcalloc(nr_pages, sizeof(struct page *),
					  GFP_KERNEL);
		if (!ctx->ring_pages)
Linus Torvalds's avatar
Linus Torvalds committed
272
273
274
			return -ENOMEM;
	}

275
276
	ctx->mmap_size = nr_pages * PAGE_SIZE;
	pr_debug("attempting mmap of %lu bytes\n", ctx->mmap_size);
277

Zach Brown's avatar
Zach Brown committed
278
	down_write(&mm->mmap_sem);
279
280
281
	ctx->mmap_base = do_mmap_pgoff(ctx->aio_ring_file, 0, ctx->mmap_size,
				       PROT_READ | PROT_WRITE,
				       MAP_SHARED | MAP_POPULATE, 0, &populate);
282
	if (IS_ERR((void *)ctx->mmap_base)) {
Zach Brown's avatar
Zach Brown committed
283
		up_write(&mm->mmap_sem);
284
		ctx->mmap_size = 0;
Linus Torvalds's avatar
Linus Torvalds committed
285
286
287
		aio_free_ring(ctx);
		return -EAGAIN;
	}
288
289
290
	up_write(&mm->mmap_sem);

	mm_populate(ctx->mmap_base, populate);
Linus Torvalds's avatar
Linus Torvalds committed
291

292
293
294
	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);
295
296
	for (i = 0; i < ctx->nr_pages; i++)
		put_page(ctx->ring_pages[i]);
Linus Torvalds's avatar
Linus Torvalds committed
297

298
	if (unlikely(ctx->nr_pages != nr_pages)) {
Linus Torvalds's avatar
Linus Torvalds committed
299
300
301
302
		aio_free_ring(ctx);
		return -EAGAIN;
	}

303
304
	ctx->user_id = ctx->mmap_base;
	ctx->nr_events = nr_events; /* trusted copy */
Linus Torvalds's avatar
Linus Torvalds committed
305

306
	ring = kmap_atomic(ctx->ring_pages[0]);
Linus Torvalds's avatar
Linus Torvalds committed
307
308
309
310
311
312
313
	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);
314
	kunmap_atomic(ring);
315
	flush_dcache_page(ctx->ring_pages[0]);
Linus Torvalds's avatar
Linus Torvalds committed
316
317
318
319
320
321
322
323

	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)

324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
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
340
341
342
static int kiocb_cancel(struct kioctx *ctx, struct kiocb *kiocb,
			struct io_event *res)
{
343
	kiocb_cancel_fn *old, *cancel;
Kent Overstreet's avatar
Kent Overstreet committed
344
345
	int ret = -EINVAL;

346
347
348
349
350
351
352
353
354
	/*
	 * 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
355

356
357
358
		old = cancel;
		cancel = cmpxchg(&kiocb->ki_cancel, old, KIOCB_CANCELLED);
	} while (cancel != old);
Kent Overstreet's avatar
Kent Overstreet committed
359

360
361
362
363
364
365
366
367
368
	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
369
370
371
372

	return ret;
}

Kent Overstreet's avatar
Kent Overstreet committed
373
374
375
376
377
378
379
380
381
382
383
384
385
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)
{
386
	struct aio_ring *ring;
Kent Overstreet's avatar
Kent Overstreet committed
387
388
	struct io_event res;
	struct kiocb *req;
389
	unsigned head, avail;
Kent Overstreet's avatar
Kent Overstreet committed
390
391
392
393
394
395
396
397
398
399
400
401
402

	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);

403
	ring = kmap_atomic(ctx->ring_pages[0]);
404
405
406
407
	head = ring->head;
	kunmap_atomic(ring);

	while (atomic_read(&ctx->reqs_active) > 0) {
408
409
410
		wait_event(ctx->wait,
				head != ctx->tail ||
				atomic_read(&ctx->reqs_active) <= 0);
411

412
		avail = (head <= ctx->tail ? ctx->tail : ctx->nr_events) - head;
413
414
415

		atomic_sub(avail, &ctx->reqs_active);
		head += avail;
416
		head %= ctx->nr_events;
417
418
419
	}

	WARN_ON(atomic_read(&ctx->reqs_active) < 0);
Kent Overstreet's avatar
Kent Overstreet committed
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440

	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
441
442
443
444
445
/* 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
446
	struct mm_struct *mm = current->mm;
Linus Torvalds's avatar
Linus Torvalds committed
447
	struct kioctx *ctx;
448
	int err = -ENOMEM;
Linus Torvalds's avatar
Linus Torvalds committed
449
450
451
452
453
454
455
456

	/* 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);
	}

457
	if (!nr_events || (unsigned long)nr_events > aio_max_nr)
Linus Torvalds's avatar
Linus Torvalds committed
458
459
		return ERR_PTR(-EAGAIN);

460
	ctx = kmem_cache_zalloc(kioctx_cachep, GFP_KERNEL);
Linus Torvalds's avatar
Linus Torvalds committed
461
462
463
464
465
	if (!ctx)
		return ERR_PTR(-ENOMEM);

	ctx->max_reqs = nr_events;

Al Viro's avatar
Al Viro committed
466
	atomic_set(&ctx->users, 2);
Kent Overstreet's avatar
Kent Overstreet committed
467
	atomic_set(&ctx->dead, 0);
Linus Torvalds's avatar
Linus Torvalds committed
468
	spin_lock_init(&ctx->ctx_lock);
469
	spin_lock_init(&ctx->completion_lock);
470
	mutex_init(&ctx->ring_lock);
Linus Torvalds's avatar
Linus Torvalds committed
471
472
473
474
475
476
477
478
	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 */
479
	spin_lock(&aio_nr_lock);
480
481
	if (aio_nr + nr_events > aio_max_nr ||
	    aio_nr + nr_events < aio_nr) {
482
		spin_unlock(&aio_nr_lock);
Linus Torvalds's avatar
Linus Torvalds committed
483
		goto out_cleanup;
484
485
	}
	aio_nr += ctx->max_reqs;
486
	spin_unlock(&aio_nr_lock);
Linus Torvalds's avatar
Linus Torvalds committed
487

Jeff Moyer's avatar
Jeff Moyer committed
488
	/* now link into global list. */
489
490
491
	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
492

Kent Overstreet's avatar
Kent Overstreet committed
493
	pr_debug("allocated ioctx %p[%ld]: mm=%p mask=0x%x\n",
494
		 ctx, ctx->user_id, mm, ctx->nr_events);
Linus Torvalds's avatar
Linus Torvalds committed
495
496
497
	return ctx;

out_cleanup:
498
499
	err = -EAGAIN;
	aio_free_ring(ctx);
Linus Torvalds's avatar
Linus Torvalds committed
500
out_freectx:
501
502
	if (ctx->aio_ring_file)
		fput(ctx->aio_ring_file);
Linus Torvalds's avatar
Linus Torvalds committed
503
	kmem_cache_free(kioctx_cachep, ctx);
Kent Overstreet's avatar
Kent Overstreet committed
504
	pr_debug("error allocating ioctx %d\n", err);
505
	return ERR_PTR(err);
Linus Torvalds's avatar
Linus Torvalds committed
506
507
}

Kent Overstreet's avatar
Kent Overstreet committed
508
static void kill_ioctx_work(struct work_struct *work)
Linus Torvalds's avatar
Linus Torvalds committed
509
{
Kent Overstreet's avatar
Kent Overstreet committed
510
	struct kioctx *ctx = container_of(work, struct kioctx, rcu_work);
511

Kent Overstreet's avatar
Kent Overstreet committed
512
513
514
	wake_up_all(&ctx->wait);
	put_ioctx(ctx);
}
Kent Overstreet's avatar
Kent Overstreet committed
515

Kent Overstreet's avatar
Kent Overstreet committed
516
517
518
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
519

Kent Overstreet's avatar
Kent Overstreet committed
520
521
522
	INIT_WORK(&ctx->rcu_work, kill_ioctx_work);
	schedule_work(&ctx->rcu_work);
}
Linus Torvalds's avatar
Linus Torvalds committed
523

Kent Overstreet's avatar
Kent Overstreet committed
524
525
526
527
528
529
530
531
532
/* 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);
533

Kent Overstreet's avatar
Kent Overstreet committed
534
		/*
535
536
537
538
539
		 * 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
540
		 */
541
542
543
544
545
546
547
548
549
550
		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
551
	}
Linus Torvalds's avatar
Linus Torvalds committed
552
553
554
555
556
}

/* wait_on_sync_kiocb:
 *	Waits on the given sync kiocb to complete.
 */
557
ssize_t wait_on_sync_kiocb(struct kiocb *iocb)
Linus Torvalds's avatar
Linus Torvalds committed
558
{
559
	while (atomic_read(&iocb->ki_users)) {
Linus Torvalds's avatar
Linus Torvalds committed
560
		set_current_state(TASK_UNINTERRUPTIBLE);
561
		if (!atomic_read(&iocb->ki_users))
Linus Torvalds's avatar
Linus Torvalds committed
562
			break;
563
		io_schedule();
Linus Torvalds's avatar
Linus Torvalds committed
564
565
566
567
	}
	__set_current_state(TASK_RUNNING);
	return iocb->ki_user_data;
}
568
EXPORT_SYMBOL(wait_on_sync_kiocb);
Linus Torvalds's avatar
Linus Torvalds committed
569

Kent Overstreet's avatar
Kent Overstreet committed
570
571
572
573
574
575
576
/*
 * 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
577
 */
578
void exit_aio(struct mm_struct *mm)
Linus Torvalds's avatar
Linus Torvalds committed
579
{
580
	struct kioctx *ctx;
Kent Overstreet's avatar
Kent Overstreet committed
581
	struct hlist_node *n;
582

Kent Overstreet's avatar
Kent Overstreet committed
583
	hlist_for_each_entry_safe(ctx, n, &mm->ioctx_list, list) {
Linus Torvalds's avatar
Linus Torvalds committed
584
585
586
		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
587
588
				atomic_read(&ctx->users),
				atomic_read(&ctx->dead),
589
				atomic_read(&ctx->reqs_active));
590
591
592
593
594
595
596
597
		/*
		 * 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.
		 */
598
		ctx->mmap_size = 0;
Kent Overstreet's avatar
Kent Overstreet committed
599

600
		kill_ioctx(ctx);
Linus Torvalds's avatar
Linus Torvalds committed
601
602
603
604
	}
}

/* aio_get_req
605
 *	Allocate a slot for an aio request.  Increments the ki_users count
Linus Torvalds's avatar
Linus Torvalds committed
606
607
608
 * of the kioctx so that the kioctx stays around until all requests are
 * complete.  Returns NULL if no requests are free.
 *
609
 * Returns with kiocb->ki_users set to 2.  The io submit code path holds
Linus Torvalds's avatar
Linus Torvalds committed
610
611
612
613
 * 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
614
static inline struct kiocb *aio_get_req(struct kioctx *ctx)
Linus Torvalds's avatar
Linus Torvalds committed
615
{
Kent Overstreet's avatar
Kent Overstreet committed
616
617
	struct kiocb *req;

618
	if (atomic_read(&ctx->reqs_active) >= ctx->nr_events)
Kent Overstreet's avatar
Kent Overstreet committed
619
620
		return NULL;

621
	if (atomic_inc_return(&ctx->reqs_active) > ctx->nr_events - 1)
Kent Overstreet's avatar
Kent Overstreet committed
622
		goto out_put;
Linus Torvalds's avatar
Linus Torvalds committed
623

624
	req = kmem_cache_alloc(kiocb_cachep, GFP_KERNEL|__GFP_ZERO);
Linus Torvalds's avatar
Linus Torvalds committed
625
	if (unlikely(!req))
Kent Overstreet's avatar
Kent Overstreet committed
626
		goto out_put;
Linus Torvalds's avatar
Linus Torvalds committed
627

628
	atomic_set(&req->ki_users, 2);
Linus Torvalds's avatar
Linus Torvalds committed
629
630
	req->ki_ctx = ctx;

Jeff Moyer's avatar
Jeff Moyer committed
631
	return req;
Kent Overstreet's avatar
Kent Overstreet committed
632
633
634
out_put:
	atomic_dec(&ctx->reqs_active);
	return NULL;
Linus Torvalds's avatar
Linus Torvalds committed
635
636
}

637
static void kiocb_free(struct kiocb *req)
Linus Torvalds's avatar
Linus Torvalds committed
638
{
639
640
	if (req->ki_filp)
		fput(req->ki_filp);
641
642
	if (req->ki_eventfd != NULL)
		eventfd_ctx_put(req->ki_eventfd);
Linus Torvalds's avatar
Linus Torvalds committed
643
644
	if (req->ki_dtor)
		req->ki_dtor(req);
645
646
	if (req->ki_iovec != &req->ki_inline_vec)
		kfree(req->ki_iovec);
Linus Torvalds's avatar
Linus Torvalds committed
647
648
649
	kmem_cache_free(kiocb_cachep, req);
}

650
void aio_put_req(struct kiocb *req)
Linus Torvalds's avatar
Linus Torvalds committed
651
{
652
653
	if (atomic_dec_and_test(&req->ki_users))
		kiocb_free(req);
Linus Torvalds's avatar
Linus Torvalds committed
654
}
655
EXPORT_SYMBOL(aio_put_req);
Linus Torvalds's avatar
Linus Torvalds committed
656

657
static struct kioctx *lookup_ioctx(unsigned long ctx_id)
Linus Torvalds's avatar
Linus Torvalds committed
658
{
659
	struct mm_struct *mm = current->mm;
660
	struct kioctx *ctx, *ret = NULL;
Linus Torvalds's avatar
Linus Torvalds committed
661

662
663
	rcu_read_lock();

664
	hlist_for_each_entry_rcu(ctx, &mm->ioctx_list, list) {
Kent Overstreet's avatar
Kent Overstreet committed
665
666
		if (ctx->user_id == ctx_id) {
			atomic_inc(&ctx->users);
667
			ret = ctx;
Linus Torvalds's avatar
Linus Torvalds committed
668
669
			break;
		}
670
	}
Linus Torvalds's avatar
Linus Torvalds committed
671

672
	rcu_read_unlock();
673
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
674
675
676
677
678
}

/* aio_complete
 *	Called when the io request on the given iocb is complete.
 */
679
void aio_complete(struct kiocb *iocb, long res, long res2)
Linus Torvalds's avatar
Linus Torvalds committed
680
681
682
{
	struct kioctx	*ctx = iocb->ki_ctx;
	struct aio_ring	*ring;
Kent Overstreet's avatar
Kent Overstreet committed
683
	struct io_event	*ev_page, *event;
Linus Torvalds's avatar
Linus Torvalds committed
684
	unsigned long	flags;
Kent Overstreet's avatar
Kent Overstreet committed
685
	unsigned tail, pos;
Linus Torvalds's avatar
Linus Torvalds committed
686

687
688
689
690
691
692
	/*
	 * 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
693
694
	 */
	if (is_sync_kiocb(iocb)) {
695
		BUG_ON(atomic_read(&iocb->ki_users) != 1);
Linus Torvalds's avatar
Linus Torvalds committed
696
		iocb->ki_user_data = res;
697
		atomic_set(&iocb->ki_users, 0);
Linus Torvalds's avatar
Linus Torvalds committed
698
		wake_up_process(iocb->ki_obj.tsk);
699
		return;
Linus Torvalds's avatar
Linus Torvalds committed
700
701
	}

Kent Overstreet's avatar
Kent Overstreet committed
702
703
704
	/*
	 * 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
705
	 */
Kent Overstreet's avatar
Kent Overstreet committed
706
	rcu_read_lock();
Linus Torvalds's avatar
Linus Torvalds committed
707

708
709
710
711
712
713
714
	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);
	}
715

Linus Torvalds's avatar
Linus Torvalds committed
716
717
718
719
	/*
	 * cancelled requests don't get events, userland was given one
	 * when the event got cancelled.
	 */
720
	if (unlikely(xchg(&iocb->ki_cancel,
721
722
723
			  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
724
		goto put_rq;
725
	}
Linus Torvalds's avatar
Linus Torvalds committed
726

727
728
	/*
	 * Add a completion event to the ring buffer. Must be done holding
729
	 * ctx->completion_lock to prevent other code from messing with the tail
730
731
732
733
	 * pointer since we might be called from irq context.
	 */
	spin_lock_irqsave(&ctx->completion_lock, flags);

734
	tail = ctx->tail;
Kent Overstreet's avatar
Kent Overstreet committed
735
736
	pos = tail + AIO_EVENTS_OFFSET;

737
	if (++tail >= ctx->nr_events)
738
		tail = 0;
Linus Torvalds's avatar
Linus Torvalds committed
739

740
	ev_page = kmap_atomic(ctx->ring_pages[pos / AIO_EVENTS_PER_PAGE]);
Kent Overstreet's avatar
Kent Overstreet committed
741
742
	event = ev_page + pos % AIO_EVENTS_PER_PAGE;

Linus Torvalds's avatar
Linus Torvalds committed
743
744
745
746
747
	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
748
	kunmap_atomic(ev_page);
749
	flush_dcache_page(ctx->ring_pages[pos / AIO_EVENTS_PER_PAGE]);
Kent Overstreet's avatar
Kent Overstreet committed
750
751

	pr_debug("%p[%u]: %p: %p %Lx %lx %lx\n",
Kent Overstreet's avatar
Kent Overstreet committed
752
753
		 ctx, tail, iocb, iocb->ki_obj.user, iocb->ki_user_data,
		 res, res2);
Linus Torvalds's avatar
Linus Torvalds committed
754
755
756
757
758
759

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

760
	ctx->tail = tail;
Linus Torvalds's avatar
Linus Torvalds committed
761

762
	ring = kmap_atomic(ctx->ring_pages[0]);
Kent Overstreet's avatar
Kent Overstreet committed
763
	ring->tail = tail;
764
	kunmap_atomic(ring);
765
	flush_dcache_page(ctx->ring_pages[0]);
Linus Torvalds's avatar
Linus Torvalds committed
766

767
768
	spin_unlock_irqrestore(&ctx->completion_lock, flags);

Kent Overstreet's avatar
Kent Overstreet committed
769
	pr_debug("added to ring %p at [%u]\n", iocb, tail);
Davide Libenzi's avatar
Davide Libenzi committed
770
771
772
773
774
775

	/*
	 * 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.
	 */
776
	if (iocb->ki_eventfd != NULL)
Davide Libenzi's avatar
Davide Libenzi committed
777
778
		eventfd_signal(iocb->ki_eventfd, 1);

Linus Torvalds's avatar
Linus Torvalds committed
779
780
put_rq:
	/* everything turned out well, dispose of the aiocb. */
781
	aio_put_req(iocb);
Linus Torvalds's avatar
Linus Torvalds committed
782

783
784
785
786
787
788
789
790
	/*
	 * 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
791
792
793
	if (waitqueue_active(&ctx->wait))
		wake_up(&ctx->wait);

Kent Overstreet's avatar
Kent Overstreet committed
794
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
795
}
796
EXPORT_SYMBOL(aio_complete);
Linus Torvalds's avatar
Linus Torvalds committed
797

798
799
800
/* 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
801
 */
802
803
static long aio_read_events_ring(struct kioctx *ctx,
				 struct io_event __user *event, long nr)
Linus Torvalds's avatar
Linus Torvalds committed
804
805
{
	struct aio_ring *ring;
806
807
808
809
	unsigned head, pos;
	long ret = 0;
	int copy_ret;

810
	mutex_lock(&ctx->ring_lock);
Linus Torvalds's avatar
Linus Torvalds committed
811

812
	ring = kmap_atomic(ctx->ring_pages[0]);
813
814
815
	head = ring->head;
	kunmap_atomic(ring);

816
	pr_debug("h%u t%u m%u\n", head, ctx->tail, ctx->nr_events);
Linus Torvalds's avatar
Linus Torvalds committed
817

818
	if (head == ctx->tail)
Linus Torvalds's avatar
Linus Torvalds committed
819
820
		goto out;

821
822
823
824
825
	while (ret < nr) {
		long avail;
		struct io_event *ev;
		struct page *page;

826
827
		avail = (head <= ctx->tail ? ctx->tail : ctx->nr_events) - head;
		if (head == ctx->tail)
828
829
830
831
832
833
834
			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;
835
		page = ctx->ring_pages[pos / AIO_EVENTS_PER_PAGE];
836
837
838
839
840
841
842
843
844
845
846
847
848
849
		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;
850
		head %= ctx->nr_events;
Linus Torvalds's avatar
Linus Torvalds committed
851
852
	}

853
	ring = kmap_atomic(ctx->ring_pages[0]);
854
	ring->head = head;
855
	kunmap_atomic(ring);
856
	flush_dcache_page(ctx->ring_pages[0]);
857

858
	pr_debug("%li  h%u t%u\n", ret, head, ctx->tail);
859
860

	atomic_sub(ret, &ctx->reqs_active);
861
out:
862
	mutex_unlock(&ctx->ring_lock);
863

Linus Torvalds's avatar
Linus Torvalds committed
864
865
866
	return ret;
}

867
868
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
869
{
870
	long ret = aio_read_events_ring(ctx, event + *i, nr - *i);
Linus Torvalds's avatar
Linus Torvalds committed
871

872
873
	if (ret > 0)
		*i += ret;
Linus Torvalds's avatar
Linus Torvalds committed
874

875
876
	if (unlikely(atomic_read(&ctx->dead)))
		ret = -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
877

878
879
	if (!*i)
		*i = ret;
Linus Torvalds's avatar
Linus Torvalds committed
880

881
	return ret < 0 || *i >= min_nr;
Linus Torvalds's avatar
Linus Torvalds committed
882
883
}

884
static long read_events(struct kioctx *ctx, long min_nr, long nr,
Linus Torvalds's avatar
Linus Torvalds committed
885
886
887
			struct io_event __user *event,
			struct timespec __user *timeout)
{
888
889
	ktime_t until = { .tv64 = KTIME_MAX };
	long ret = 0;
Linus Torvalds's avatar
Linus Torvalds committed
890
891
892

	if (timeout) {
		struct timespec	ts;
893

Linus Torvalds's avatar
Linus Torvalds committed
894
		if (unlikely(copy_from_user(&ts, timeout, sizeof(ts))))
895
			return -EFAULT;
Linus Torvalds's avatar
Linus Torvalds committed
896

897
		until = timespec_to_ktime(ts);
Linus Torvalds's avatar
Linus Torvalds committed
898
899
	}

900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
	/*
	 * 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
916

917
918
	if (!ret && signal_pending(current))
		ret = -EINTR;
Linus Torvalds's avatar
Linus Torvalds committed
919

920
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
}

/* 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.
 */
936
SYSCALL_DEFINE2(io_setup, unsigned, nr_events, aio_context_t __user *, ctxp)
Linus Torvalds's avatar
Linus Torvalds committed
937
938
939
940
941
942
943
944
945
946
{
	struct kioctx *ioctx = NULL;
	unsigned long ctx;
	long ret;

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

	ret = -EINVAL;
947
948
949
	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
950
951
952
953
954
955
956
		goto out;
	}

	ioctx = ioctx_alloc(nr_events);
	ret = PTR_ERR(ioctx);
	if (!IS_ERR(ioctx)) {
		ret = put_user(ioctx->user_id, ctxp);
957
		if (ret)
Kent Overstreet's avatar
Kent Overstreet committed
958
			kill_ioctx(ioctx);
959
		put_ioctx(ioctx);
Linus Torvalds's avatar
Linus Torvalds committed
960
961
962
963
964
965
966
967
968
	}

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
969
 *	implemented.  May fail with -EINVAL if the context pointed to
Linus Torvalds's avatar
Linus Torvalds committed
970
971
 *	is invalid.
 */
972
SYSCALL_DEFINE1(io_destroy, aio_context_t, ctx)
Linus Torvalds's avatar
Linus Torvalds committed
973
974
975
{
	struct kioctx *ioctx = lookup_ioctx(ctx);
	if (likely(NULL != ioctx)) {
Kent Overstreet's avatar
Kent Overstreet committed
976
		kill_ioctx(ioctx);
977
		put_ioctx(ioctx);
Linus Torvalds's avatar
Linus Torvalds committed
978
979
980
981
982
983
		return 0;
	}
	pr_debug("EINVAL: io_destroy: invalid context id\n");
	return -EINVAL;
}

984
static void aio_advance_iovec(struct kiocb *iocb, ssize_t ret)
Linus Torvalds's avatar
Linus Torvalds committed
985
{
986
987
988
989
990
991
992
993
994
995
996
997
998
	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++;
999
		}
1000
	}
Linus Torvalds's avatar
Linus Torvalds committed
1001

1002
1003
1004
	/* 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
1005
1006
}

Kent Overstreet's avatar
Kent Overstreet committed
1007
1008
1009
1010
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
1011
1012
{
	struct file *file = iocb->ki_filp;
1013
1014
	struct address_space *mapping = file->f_mapping;
	struct inode *inode = mapping->host;
Linus Torvalds's avatar
Linus Torvalds committed
1015
1016
	ssize_t ret = 0;

1017
1018
1019
1020
	/* This matches the pread()/pwrite() logic */
	if (iocb->ki_pos < 0)
		return -EINVAL;

Kent Overstreet's avatar
Kent Overstreet committed
1021
	if (rw == WRITE)
1022
		file_start_write(file);
1023
	do {
1024
1025
1026
1027
1028
1029
1030
1031
1032
		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
1033
		 (rw == WRITE ||
1034
		  (!S_ISFIFO(inode->i_mode) && !S_ISSOCK(inode->i_mode))));
Kent Overstreet's avatar
Kent Overstreet committed
1035
	if (rw == WRITE)
1036
		file_end_write(file);
Linus Torvalds's avatar
Linus Torvalds committed
1037

1038
1039
	/* This means we must have transferred all that we could */
	/* No need to retry anymore */
Linus Torvalds's avatar
Linus Torvalds committed
1040
1041
1042
	if ((ret == 0) || (iocb->ki_left == 0))
		ret = iocb->ki_nbytes - iocb->ki_left;

1043
1044
	/* If we managed to write some out we return that, rather than
	 * the eventual error. */
Kent Overstreet's avatar
Kent Overstreet committed
1045
	if (rw == WRITE
Zach Brown's avatar
Zach Brown committed
1046
	    && ret < 0 && ret != -EIOCBQUEUED
1047
1048
1049
	    && iocb->ki_nbytes - iocb->ki_left)
		ret = iocb->ki_nbytes - iocb->ki_left;

Linus Torvalds's avatar
Linus Torvalds committed
1050
1051
1052
	return ret;
}

Kent Overstreet's avatar
Kent Overstreet committed
1053
static ssize_t aio_setup_vectored_rw(int rw, struct kiocb *kiocb, bool compat)
1054
1055
1056
{
	ssize_t ret;

Kent Overstreet's avatar
Kent Overstreet committed
1057
1058
	kiocb->ki_nr_segs = kiocb->ki_nbytes;

1059
1060
#ifdef CONFIG_COMPAT
	if (compat)
Kent Overstreet's avatar
Kent Overstreet committed
1061
		ret = compat_rw_copy_check_uvector(rw,
1062
				(struct compat_iovec __user *)kiocb->ki_buf,
Kent Overstreet's avatar
Kent Overstreet committed
1063
				kiocb->ki_nr_segs, 1, &kiocb->ki_inline_vec,
1064
				&kiocb->ki_iovec);
1065
1066
	else
#endif
Kent Overstreet's avatar
Kent Overstreet committed
1067
		ret = rw_copy_check_uvector(rw,
1068
				(struct iovec __user *)kiocb->ki_buf,
Kent Overstreet's avatar
Kent Overstreet committed
1069
				kiocb->ki_nr_segs, 1, &kiocb->ki_inline_vec,
1070
				&kiocb->ki_iovec);
1071
	if (ret < 0)
Kent Overstreet's avatar
Kent Overstreet committed
1072
		return ret;
1073

Kent Overstreet's avatar
Kent Overstreet committed
1074
	/* ki_nbytes now reflect bytes instead of segs */
1075
	kiocb->ki_nbytes = ret;
Kent Overstreet's avatar
Kent Overstreet committed
1076
	return 0;
1077
1078
}

Kent Overstreet's avatar
Kent Overstreet committed
1079
static ssize_t aio_setup_single_vector(int rw, struct kiocb *kiocb)
1080
{
Kent Overstreet's avatar
Kent Overstreet committed
1081
1082
	if (unlikely(!access_ok(!rw, kiocb->ki_buf, kiocb->ki_nbytes)))
		return -EFAULT;
1083

1084
1085
	kiocb->ki_iovec = &kiocb->ki_inline_vec;
	kiocb->ki_iovec->iov_base = kiocb->ki_buf;
Kent Overstreet's avatar
Kent Overstreet committed
1086
	kiocb->ki_iovec->iov_len = kiocb->ki_nbytes;
1087
1088
1089
1090
	kiocb->ki_nr_segs = 1;
	return 0;
}

Linus Torvalds's avatar
Linus Torvalds committed
1091
1092
1093
1094
1095
/*
 * 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
1096
static ssize_t aio_run_iocb(struct kiocb *req, bool compat)
Linus Torvalds's avatar
Linus Torvalds committed
1097
{
Kent Overstreet's avatar
Kent Overstreet committed
1098
1099
1100
1101
1102
	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
1103

Kent Overstreet's avatar
Kent Overstreet committed
1104
	switch (req->ki_opcode) {
Linus Torvalds's avatar
Linus Torvalds committed
1105
	case IOCB_CMD_PREAD:
1106
	case IOCB_CMD_PREADV:
Kent Overstreet's avatar
Kent Overstreet committed
1107
1108
1109
1110
1111
1112
		mode	= FMODE_READ;
		rw	= READ;
		rw_op	= file->f_op->aio_read;
		goto rw_common;

	case IOCB_CMD_PWRITE:
1113
	case IOCB_CMD_PWRITEV:
Kent Overstreet's avatar
Kent Overstreet committed
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
		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);
1129
		if (ret)
Kent Overstreet's avatar
Kent Overstreet committed
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
			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
1140
		break;
Kent Overstreet's avatar
Kent Overstreet committed
1141

Linus Torvalds's avatar
Linus Torvalds committed
1142
	case IOCB_CMD_FDSYNC:
Kent Overstreet's avatar
Kent Overstreet committed
1143
1144
1145
1146
		if (!file->f_op->aio_fsync)
			return -EINVAL;

		ret = file->f_op->aio_fsync(req, 1);
Linus Torvalds's avatar
Linus Torvalds committed
1147
		break;
Kent Overstreet's avatar
Kent Overstreet committed
1148

Linus Torvalds's avatar
Linus Torvalds committed
1149
	case IOCB_CMD_FSYNC:
Kent Overstreet's avatar
Kent Overstreet committed
1150
1151
1152
1153
		if (!file->f_op->aio_fsync)
			return -EINVAL;

		ret = file->f_op->aio_fsync(req, 0);