aio.c 35.9 KB
Newer Older
L
Linus Torvalds 已提交
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.
 */
K
Kent Overstreet 已提交
11 12
#define pr_fmt(fmt) "%s: " fmt, __func__

L
Linus Torvalds 已提交
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>
L
Linus Torvalds 已提交
19
#include <linux/syscalls.h>
20
#include <linux/backing-dev.h>
21
#include <linux/uio.h>
L
Linus Torvalds 已提交
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>
L
Linus Torvalds 已提交
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>
J
Jeff Moyer 已提交
36
#include <linux/blkdev.h>
37
#include <linux/compat.h>
L
Linus Torvalds 已提交
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 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 aio_ring_info {
	unsigned long		mmap_base;
	unsigned long		mmap_size;

	struct page		**ring_pages;
66
	struct mutex		ring_lock;
67 68 69 70 71 72 73 74 75 76 77 78 79 80 81
	long			nr_pages;

	unsigned		nr, tail;

	struct page		*internal_pages[AIO_RING_PAGES];
};

static inline unsigned aio_ring_avail(struct aio_ring_info *info,
					struct aio_ring *ring)
{
	return (ring->head + info->nr - 1 - ring->tail) % info->nr;
}

struct kioctx {
	atomic_t		users;
K
Kent Overstreet 已提交
82
	atomic_t		dead;
83 84 85 86 87 88 89 90 91

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

	wait_queue_head_t	wait;

	spinlock_t		ctx_lock;

92
	atomic_t		reqs_active;
93 94 95 96 97 98 99
	struct list_head	active_reqs;	/* used for cancellation */

	/* sys_io_setup currently limits this to an unsigned int */
	unsigned		max_reqs;

	struct aio_ring_info	ring_info;

100 101
	spinlock_t		completion_lock;

102
	struct rcu_head		rcu_head;
K
Kent Overstreet 已提交
103
	struct work_struct	rcu_work;
104 105
};

L
Linus Torvalds 已提交
106
/*------ sysctl variables----*/
107 108 109
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 */
L
Linus Torvalds 已提交
110 111
/*----end sysctl variables---*/

112 113
static struct kmem_cache	*kiocb_cachep;
static struct kmem_cache	*kioctx_cachep;
L
Linus Torvalds 已提交
114 115 116 117 118 119 120

/* 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)
{
121 122
	kiocb_cachep = KMEM_CACHE(kiocb, SLAB_HWCACHE_ALIGN|SLAB_PANIC);
	kioctx_cachep = KMEM_CACHE(kioctx,SLAB_HWCACHE_ALIGN|SLAB_PANIC);
L
Linus Torvalds 已提交
123

K
Kent Overstreet 已提交
124
	pr_debug("sizeof(struct page) = %zu\n", sizeof(struct page));
L
Linus Torvalds 已提交
125 126 127

	return 0;
}
128
__initcall(aio_setup);
L
Linus Torvalds 已提交
129 130 131 132 133 134 135 136 137

static void aio_free_ring(struct kioctx *ctx)
{
	struct aio_ring_info *info = &ctx->ring_info;
	long i;

	for (i=0; i<info->nr_pages; i++)
		put_page(info->ring_pages[i]);

138
	if (info->mmap_size) {
A
Al Viro 已提交
139
		vm_munmap(info->mmap_base, info->mmap_size);
140
	}
L
Linus Torvalds 已提交
141 142 143 144 145 146 147 148 149 150 151 152

	if (info->ring_pages && info->ring_pages != info->internal_pages)
		kfree(info->ring_pages);
	info->ring_pages = NULL;
	info->nr = 0;
}

static int aio_setup_ring(struct kioctx *ctx)
{
	struct aio_ring *ring;
	struct aio_ring_info *info = &ctx->ring_info;
	unsigned nr_events = ctx->max_reqs;
Z
Zach Brown 已提交
153
	struct mm_struct *mm = current->mm;
154
	unsigned long size, populate;
L
Linus Torvalds 已提交
155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171
	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);

	info->nr = 0;
	info->ring_pages = info->internal_pages;
	if (nr_pages > AIO_RING_PAGES) {
172
		info->ring_pages = kcalloc(nr_pages, sizeof(struct page *), GFP_KERNEL);
L
Linus Torvalds 已提交
173 174 175 176 177
		if (!info->ring_pages)
			return -ENOMEM;
	}

	info->mmap_size = nr_pages * PAGE_SIZE;
K
Kent Overstreet 已提交
178
	pr_debug("attempting mmap of %lu bytes\n", info->mmap_size);
Z
Zach Brown 已提交
179
	down_write(&mm->mmap_sem);
180 181
	info->mmap_base = do_mmap_pgoff(NULL, 0, info->mmap_size, 
					PROT_READ|PROT_WRITE,
182 183
					MAP_ANONYMOUS|MAP_PRIVATE, 0,
					&populate);
L
Linus Torvalds 已提交
184
	if (IS_ERR((void *)info->mmap_base)) {
Z
Zach Brown 已提交
185
		up_write(&mm->mmap_sem);
L
Linus Torvalds 已提交
186 187 188 189 190
		info->mmap_size = 0;
		aio_free_ring(ctx);
		return -EAGAIN;
	}

K
Kent Overstreet 已提交
191
	pr_debug("mmap address: 0x%08lx\n", info->mmap_base);
Z
Zach Brown 已提交
192
	info->nr_pages = get_user_pages(current, mm, info->mmap_base, nr_pages,
L
Linus Torvalds 已提交
193
					1, 0, info->ring_pages, NULL);
Z
Zach Brown 已提交
194
	up_write(&mm->mmap_sem);
L
Linus Torvalds 已提交
195 196 197 198 199

	if (unlikely(info->nr_pages != nr_pages)) {
		aio_free_ring(ctx);
		return -EAGAIN;
	}
200
	if (populate)
201
		mm_populate(info->mmap_base, populate);
L
Linus Torvalds 已提交
202 203 204 205 206

	ctx->user_id = info->mmap_base;

	info->nr = nr_events;		/* trusted copy */

207
	ring = kmap_atomic(info->ring_pages[0]);
L
Linus Torvalds 已提交
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);
K
Kent Overstreet 已提交
216
	flush_dcache_page(info->ring_pages[0]);
L
Linus Torvalds 已提交
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);

K
Kent Overstreet 已提交
241 242 243
static int kiocb_cancel(struct kioctx *ctx, struct kiocb *kiocb,
			struct io_event *res)
{
244
	kiocb_cancel_fn *old, *cancel;
K
Kent Overstreet 已提交
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;
K
Kent Overstreet 已提交
256

257 258 259
		old = cancel;
		cancel = cmpxchg(&kiocb->ki_cancel, old, KIOCB_CANCELLED);
	} while (cancel != old);
K
Kent Overstreet 已提交
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);
K
Kent Overstreet 已提交
270 271 272 273

	return ret;
}

K
Kent Overstreet 已提交
274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328
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)
{
	struct io_event res;
	struct kiocb *req;

	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);

	wait_event(ctx->wait, !atomic_read(&ctx->reqs_active));

	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);
}

L
Linus Torvalds 已提交
329 330 331 332 333
/* ioctx_alloc
 *	Allocates and initializes an ioctx.  Returns an ERR_PTR if it failed.
 */
static struct kioctx *ioctx_alloc(unsigned nr_events)
{
Z
Zach Brown 已提交
334
	struct mm_struct *mm = current->mm;
L
Linus Torvalds 已提交
335
	struct kioctx *ctx;
336
	int err = -ENOMEM;
L
Linus Torvalds 已提交
337 338 339 340 341 342 343 344

	/* 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);
	}

345
	if (!nr_events || (unsigned long)nr_events > aio_max_nr)
L
Linus Torvalds 已提交
346 347
		return ERR_PTR(-EAGAIN);

348
	ctx = kmem_cache_zalloc(kioctx_cachep, GFP_KERNEL);
L
Linus Torvalds 已提交
349 350 351 352 353
	if (!ctx)
		return ERR_PTR(-ENOMEM);

	ctx->max_reqs = nr_events;

A
Al Viro 已提交
354
	atomic_set(&ctx->users, 2);
K
Kent Overstreet 已提交
355
	atomic_set(&ctx->dead, 0);
L
Linus Torvalds 已提交
356
	spin_lock_init(&ctx->ctx_lock);
357
	spin_lock_init(&ctx->completion_lock);
358
	mutex_init(&ctx->ring_info.ring_lock);
L
Linus Torvalds 已提交
359 360 361 362 363 364 365 366
	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 */
367
	spin_lock(&aio_nr_lock);
368 369
	if (aio_nr + nr_events > aio_max_nr ||
	    aio_nr + nr_events < aio_nr) {
370
		spin_unlock(&aio_nr_lock);
L
Linus Torvalds 已提交
371
		goto out_cleanup;
372 373
	}
	aio_nr += ctx->max_reqs;
374
	spin_unlock(&aio_nr_lock);
L
Linus Torvalds 已提交
375

J
Jeff Moyer 已提交
376
	/* now link into global list. */
J
Jens Axboe 已提交
377 378 379
	spin_lock(&mm->ioctx_lock);
	hlist_add_head_rcu(&ctx->list, &mm->ioctx_list);
	spin_unlock(&mm->ioctx_lock);
L
Linus Torvalds 已提交
380

K
Kent Overstreet 已提交
381
	pr_debug("allocated ioctx %p[%ld]: mm=%p mask=0x%x\n",
Z
Zach Brown 已提交
382
		ctx, ctx->user_id, mm, ctx->ring_info.nr);
L
Linus Torvalds 已提交
383 384 385
	return ctx;

out_cleanup:
386 387
	err = -EAGAIN;
	aio_free_ring(ctx);
L
Linus Torvalds 已提交
388 389
out_freectx:
	kmem_cache_free(kioctx_cachep, ctx);
K
Kent Overstreet 已提交
390
	pr_debug("error allocating ioctx %d\n", err);
391
	return ERR_PTR(err);
L
Linus Torvalds 已提交
392 393
}

K
Kent Overstreet 已提交
394
static void kill_ioctx_work(struct work_struct *work)
L
Linus Torvalds 已提交
395
{
K
Kent Overstreet 已提交
396
	struct kioctx *ctx = container_of(work, struct kioctx, rcu_work);
397

K
Kent Overstreet 已提交
398 399 400
	wake_up_all(&ctx->wait);
	put_ioctx(ctx);
}
K
Kent Overstreet 已提交
401

K
Kent Overstreet 已提交
402 403 404
static void kill_ioctx_rcu(struct rcu_head *head)
{
	struct kioctx *ctx = container_of(head, struct kioctx, rcu_head);
L
Linus Torvalds 已提交
405

K
Kent Overstreet 已提交
406 407 408
	INIT_WORK(&ctx->rcu_work, kill_ioctx_work);
	schedule_work(&ctx->rcu_work);
}
L
Linus Torvalds 已提交
409

K
Kent Overstreet 已提交
410 411 412 413 414 415 416 417 418 419 420
/* 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();
421

K
Kent Overstreet 已提交
422 423 424 425 426 427 428 429 430
		/*
		 * 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);
	}
L
Linus Torvalds 已提交
431 432 433 434 435
}

/* wait_on_sync_kiocb:
 *	Waits on the given sync kiocb to complete.
 */
436
ssize_t wait_on_sync_kiocb(struct kiocb *iocb)
L
Linus Torvalds 已提交
437
{
438
	while (atomic_read(&iocb->ki_users)) {
L
Linus Torvalds 已提交
439
		set_current_state(TASK_UNINTERRUPTIBLE);
440
		if (!atomic_read(&iocb->ki_users))
L
Linus Torvalds 已提交
441
			break;
J
Jeff Moyer 已提交
442
		io_schedule();
L
Linus Torvalds 已提交
443 444 445 446
	}
	__set_current_state(TASK_RUNNING);
	return iocb->ki_user_data;
}
447
EXPORT_SYMBOL(wait_on_sync_kiocb);
L
Linus Torvalds 已提交
448

K
Kent Overstreet 已提交
449 450 451 452 453 454 455
/*
 * 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.
L
Linus Torvalds 已提交
456
 */
457
void exit_aio(struct mm_struct *mm)
L
Linus Torvalds 已提交
458
{
J
Jens Axboe 已提交
459
	struct kioctx *ctx;
K
Kent Overstreet 已提交
460
	struct hlist_node *n;
J
Jens Axboe 已提交
461

K
Kent Overstreet 已提交
462
	hlist_for_each_entry_safe(ctx, n, &mm->ioctx_list, list) {
L
Linus Torvalds 已提交
463 464 465
		if (1 != atomic_read(&ctx->users))
			printk(KERN_DEBUG
				"exit_aio:ioctx still alive: %d %d %d\n",
K
Kent Overstreet 已提交
466 467
				atomic_read(&ctx->users),
				atomic_read(&ctx->dead),
468
				atomic_read(&ctx->reqs_active));
469 470 471 472 473 474 475 476 477
		/*
		 * 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.
		 */
		ctx->ring_info.mmap_size = 0;
K
Kent Overstreet 已提交
478 479 480 481 482

		if (!atomic_xchg(&ctx->dead, 1)) {
			hlist_del_rcu(&ctx->list);
			call_rcu(&ctx->rcu_head, kill_ioctx_rcu);
		}
L
Linus Torvalds 已提交
483 484 485 486
	}
}

/* aio_get_req
487
 *	Allocate a slot for an aio request.  Increments the ki_users count
L
Linus Torvalds 已提交
488 489 490
 * of the kioctx so that the kioctx stays around until all requests are
 * complete.  Returns NULL if no requests are free.
 *
491
 * Returns with kiocb->ki_users set to 2.  The io submit code path holds
L
Linus Torvalds 已提交
492 493 494 495
 * 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.
 */
496
static struct kiocb *__aio_get_req(struct kioctx *ctx)
L
Linus Torvalds 已提交
497 498 499
{
	struct kiocb *req = NULL;

500
	req = kmem_cache_alloc(kiocb_cachep, GFP_KERNEL|__GFP_ZERO);
L
Linus Torvalds 已提交
501 502 503
	if (unlikely(!req))
		return NULL;

504
	atomic_set(&req->ki_users, 2);
L
Linus Torvalds 已提交
505 506
	req->ki_ctx = ctx;

J
Jeff Moyer 已提交
507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525
	return req;
}

/*
 * struct kiocb's are allocated in batches to reduce the number of
 * times the ctx lock is acquired and released.
 */
#define KIOCB_BATCH_SIZE	32L
struct kiocb_batch {
	struct list_head head;
	long count; /* number of requests left to allocate */
};

static void kiocb_batch_init(struct kiocb_batch *batch, long total)
{
	INIT_LIST_HEAD(&batch->head);
	batch->count = total;
}

526
static void kiocb_batch_free(struct kioctx *ctx, struct kiocb_batch *batch)
J
Jeff Moyer 已提交
527 528 529
{
	struct kiocb *req, *n;

530 531 532 533
	if (list_empty(&batch->head))
		return;

	spin_lock_irq(&ctx->ctx_lock);
J
Jeff Moyer 已提交
534 535 536
	list_for_each_entry_safe(req, n, &batch->head, ki_batch) {
		list_del(&req->ki_batch);
		kmem_cache_free(kiocb_cachep, req);
537
		atomic_dec(&ctx->reqs_active);
J
Jeff Moyer 已提交
538
	}
539
	spin_unlock_irq(&ctx->ctx_lock);
J
Jeff Moyer 已提交
540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564
}

/*
 * Allocate a batch of kiocbs.  This avoids taking and dropping the
 * context lock a lot during setup.
 */
static int kiocb_batch_refill(struct kioctx *ctx, struct kiocb_batch *batch)
{
	unsigned short allocated, to_alloc;
	long avail;
	struct kiocb *req, *n;
	struct aio_ring *ring;

	to_alloc = min(batch->count, KIOCB_BATCH_SIZE);
	for (allocated = 0; allocated < to_alloc; allocated++) {
		req = __aio_get_req(ctx);
		if (!req)
			/* allocation failed, go with what we've got */
			break;
		list_add(&req->ki_batch, &batch->head);
	}

	if (allocated == 0)
		goto out;

L
Linus Torvalds 已提交
565
	spin_lock_irq(&ctx->ctx_lock);
J
Jeff Moyer 已提交
566 567
	ring = kmap_atomic(ctx->ring_info.ring_pages[0]);

568 569
	avail = aio_ring_avail(&ctx->ring_info, ring) -
				atomic_read(&ctx->reqs_active);
J
Jeff Moyer 已提交
570 571 572 573 574 575 576 577 578 579 580 581
	BUG_ON(avail < 0);
	if (avail < allocated) {
		/* Trim back the number of requests. */
		list_for_each_entry_safe(req, n, &batch->head, ki_batch) {
			list_del(&req->ki_batch);
			kmem_cache_free(kiocb_cachep, req);
			if (--allocated <= avail)
				break;
		}
	}

	batch->count -= allocated;
582
	atomic_add(allocated, &ctx->reqs_active);
L
Linus Torvalds 已提交
583

J
Jeff Moyer 已提交
584 585
	kunmap_atomic(ring);
	spin_unlock_irq(&ctx->ctx_lock);
L
Linus Torvalds 已提交
586

J
Jeff Moyer 已提交
587 588
out:
	return allocated;
L
Linus Torvalds 已提交
589 590
}

J
Jeff Moyer 已提交
591 592
static inline struct kiocb *aio_get_req(struct kioctx *ctx,
					struct kiocb_batch *batch)
L
Linus Torvalds 已提交
593 594
{
	struct kiocb *req;
J
Jeff Moyer 已提交
595 596 597 598 599 600

	if (list_empty(&batch->head))
		if (kiocb_batch_refill(ctx, batch) == 0)
			return NULL;
	req = list_first_entry(&batch->head, struct kiocb, ki_batch);
	list_del(&req->ki_batch);
L
Linus Torvalds 已提交
601 602 603
	return req;
}

604
static void kiocb_free(struct kiocb *req)
L
Linus Torvalds 已提交
605
{
606 607
	if (req->ki_filp)
		fput(req->ki_filp);
608 609
	if (req->ki_eventfd != NULL)
		eventfd_ctx_put(req->ki_eventfd);
L
Linus Torvalds 已提交
610 611
	if (req->ki_dtor)
		req->ki_dtor(req);
B
Badari Pulavarty 已提交
612 613
	if (req->ki_iovec != &req->ki_inline_vec)
		kfree(req->ki_iovec);
L
Linus Torvalds 已提交
614 615 616
	kmem_cache_free(kiocb_cachep, req);
}

617
void aio_put_req(struct kiocb *req)
L
Linus Torvalds 已提交
618
{
619 620
	if (atomic_dec_and_test(&req->ki_users))
		kiocb_free(req);
L
Linus Torvalds 已提交
621
}
622
EXPORT_SYMBOL(aio_put_req);
L
Linus Torvalds 已提交
623

A
Adrian Bunk 已提交
624
static struct kioctx *lookup_ioctx(unsigned long ctx_id)
L
Linus Torvalds 已提交
625
{
J
Jens Axboe 已提交
626
	struct mm_struct *mm = current->mm;
627
	struct kioctx *ctx, *ret = NULL;
L
Linus Torvalds 已提交
628

J
Jens Axboe 已提交
629 630
	rcu_read_lock();

631
	hlist_for_each_entry_rcu(ctx, &mm->ioctx_list, list) {
K
Kent Overstreet 已提交
632 633
		if (ctx->user_id == ctx_id) {
			atomic_inc(&ctx->users);
634
			ret = ctx;
L
Linus Torvalds 已提交
635 636
			break;
		}
J
Jens Axboe 已提交
637
	}
L
Linus Torvalds 已提交
638

J
Jens Axboe 已提交
639
	rcu_read_unlock();
640
	return ret;
L
Linus Torvalds 已提交
641 642 643 644 645
}

/* aio_complete
 *	Called when the io request on the given iocb is complete.
 */
646
void aio_complete(struct kiocb *iocb, long res, long res2)
L
Linus Torvalds 已提交
647 648 649 650
{
	struct kioctx	*ctx = iocb->ki_ctx;
	struct aio_ring_info	*info;
	struct aio_ring	*ring;
K
Kent Overstreet 已提交
651
	struct io_event	*ev_page, *event;
L
Linus Torvalds 已提交
652
	unsigned long	flags;
K
Kent Overstreet 已提交
653
	unsigned tail, pos;
L
Linus Torvalds 已提交
654

655 656 657 658 659 660
	/*
	 * 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
L
Linus Torvalds 已提交
661 662
	 */
	if (is_sync_kiocb(iocb)) {
663
		BUG_ON(atomic_read(&iocb->ki_users) != 1);
L
Linus Torvalds 已提交
664
		iocb->ki_user_data = res;
665
		atomic_set(&iocb->ki_users, 0);
L
Linus Torvalds 已提交
666
		wake_up_process(iocb->ki_obj.tsk);
667
		return;
L
Linus Torvalds 已提交
668 669 670 671
	}

	info = &ctx->ring_info;

K
Kent Overstreet 已提交
672 673 674
	/*
	 * Take rcu_read_lock() in case the kioctx is being destroyed, as we
	 * need to issue a wakeup after decrementing reqs_active.
L
Linus Torvalds 已提交
675
	 */
K
Kent Overstreet 已提交
676
	rcu_read_lock();
L
Linus Torvalds 已提交
677

678 679 680 681 682 683 684
	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);
	}
685

L
Linus Torvalds 已提交
686 687 688 689
	/*
	 * cancelled requests don't get events, userland was given one
	 * when the event got cancelled.
	 */
690 691
	if (unlikely(xchg(&iocb->ki_cancel,
			  KIOCB_CANCELLED) == KIOCB_CANCELLED))
L
Linus Torvalds 已提交
692 693
		goto put_rq;

694 695 696 697 698 699 700
	/*
	 * 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);

L
Linus Torvalds 已提交
701
	tail = info->tail;
K
Kent Overstreet 已提交
702 703
	pos = tail + AIO_EVENTS_OFFSET;

704 705
	if (++tail >= info->nr)
		tail = 0;
L
Linus Torvalds 已提交
706

K
Kent Overstreet 已提交
707 708 709
	ev_page = kmap_atomic(info->ring_pages[pos / AIO_EVENTS_PER_PAGE]);
	event = ev_page + pos % AIO_EVENTS_PER_PAGE;

L
Linus Torvalds 已提交
710 711 712 713 714
	event->obj = (u64)(unsigned long)iocb->ki_obj.user;
	event->data = iocb->ki_user_data;
	event->res = res;
	event->res2 = res2;

K
Kent Overstreet 已提交
715 716 717 718
	kunmap_atomic(ev_page);
	flush_dcache_page(info->ring_pages[pos / AIO_EVENTS_PER_PAGE]);

	pr_debug("%p[%u]: %p: %p %Lx %lx %lx\n",
K
Kent Overstreet 已提交
719 720
		 ctx, tail, iocb, iocb->ki_obj.user, iocb->ki_user_data,
		 res, res2);
L
Linus Torvalds 已提交
721 722 723 724 725 726 727 728

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

	info->tail = tail;

K
Kent Overstreet 已提交
729 730
	ring = kmap_atomic(info->ring_pages[0]);
	ring->tail = tail;
731
	kunmap_atomic(ring);
K
Kent Overstreet 已提交
732
	flush_dcache_page(info->ring_pages[0]);
L
Linus Torvalds 已提交
733

734 735
	spin_unlock_irqrestore(&ctx->completion_lock, flags);

K
Kent Overstreet 已提交
736
	pr_debug("added to ring %p at [%u]\n", iocb, tail);
D
Davide Libenzi 已提交
737 738 739 740 741 742

	/*
	 * 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.
	 */
743
	if (iocb->ki_eventfd != NULL)
D
Davide Libenzi 已提交
744 745
		eventfd_signal(iocb->ki_eventfd, 1);

L
Linus Torvalds 已提交
746 747
put_rq:
	/* everything turned out well, dispose of the aiocb. */
748 749
	aio_put_req(iocb);
	atomic_dec(&ctx->reqs_active);
L
Linus Torvalds 已提交
750

751 752 753 754 755 756 757 758
	/*
	 * 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();

L
Linus Torvalds 已提交
759 760 761
	if (waitqueue_active(&ctx->wait))
		wake_up(&ctx->wait);

K
Kent Overstreet 已提交
762
	rcu_read_unlock();
L
Linus Torvalds 已提交
763
}
764
EXPORT_SYMBOL(aio_complete);
L
Linus Torvalds 已提交
765

766 767 768
/* aio_read_events
 *	Pull an event off of the ioctx's event ring.  Returns the number of
 *	events fetched
L
Linus Torvalds 已提交
769
 */
770 771
static long aio_read_events_ring(struct kioctx *ctx,
				 struct io_event __user *event, long nr)
L
Linus Torvalds 已提交
772
{
773
	struct aio_ring_info *info = &ctx->ring_info;
L
Linus Torvalds 已提交
774
	struct aio_ring *ring;
775 776 777 778 779
	unsigned head, pos;
	long ret = 0;
	int copy_ret;

	mutex_lock(&info->ring_lock);
L
Linus Torvalds 已提交
780

781
	ring = kmap_atomic(info->ring_pages[0]);
782 783 784 785
	head = ring->head;
	kunmap_atomic(ring);

	pr_debug("h%u t%u m%u\n", head, info->tail, info->nr);
L
Linus Torvalds 已提交
786

787
	if (head == info->tail)
L
Linus Torvalds 已提交
788 789
		goto out;

790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819
	while (ret < nr) {
		long avail;
		struct io_event *ev;
		struct page *page;

		avail = (head <= info->tail ? info->tail : info->nr) - head;
		if (head == info->tail)
			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;
		page = info->ring_pages[pos / AIO_EVENTS_PER_PAGE];
		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;
		head %= info->nr;
L
Linus Torvalds 已提交
820 821
	}

822 823
	ring = kmap_atomic(info->ring_pages[0]);
	ring->head = head;
824
	kunmap_atomic(ring);
K
Kent Overstreet 已提交
825
	flush_dcache_page(info->ring_pages[0]);
826 827 828 829 830

	pr_debug("%li  h%u t%u\n", ret, head, info->tail);
out:
	mutex_unlock(&info->ring_lock);

L
Linus Torvalds 已提交
831 832 833
	return ret;
}

834 835
static bool aio_read_events(struct kioctx *ctx, long min_nr, long nr,
			    struct io_event __user *event, long *i)
L
Linus Torvalds 已提交
836
{
837
	long ret = aio_read_events_ring(ctx, event + *i, nr - *i);
L
Linus Torvalds 已提交
838

839 840
	if (ret > 0)
		*i += ret;
L
Linus Torvalds 已提交
841

842 843
	if (unlikely(atomic_read(&ctx->dead)))
		ret = -EINVAL;
L
Linus Torvalds 已提交
844

845 846
	if (!*i)
		*i = ret;
L
Linus Torvalds 已提交
847

848
	return ret < 0 || *i >= min_nr;
L
Linus Torvalds 已提交
849 850
}

851
static long read_events(struct kioctx *ctx, long min_nr, long nr,
L
Linus Torvalds 已提交
852 853 854
			struct io_event __user *event,
			struct timespec __user *timeout)
{
855 856
	ktime_t until = { .tv64 = KTIME_MAX };
	long ret = 0;
L
Linus Torvalds 已提交
857 858 859

	if (timeout) {
		struct timespec	ts;
860

L
Linus Torvalds 已提交
861
		if (unlikely(copy_from_user(&ts, timeout, sizeof(ts))))
862
			return -EFAULT;
L
Linus Torvalds 已提交
863

864
		until = timespec_to_ktime(ts);
L
Linus Torvalds 已提交
865 866
	}

867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882
	/*
	 * 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);
L
Linus Torvalds 已提交
883

884 885
	if (!ret && signal_pending(current))
		ret = -EINTR;
L
Linus Torvalds 已提交
886

887
	return ret;
L
Linus Torvalds 已提交
888 889 890 891 892 893 894 895 896 897 898 899 900 901 902
}

/* 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.
 */
903
SYSCALL_DEFINE2(io_setup, unsigned, nr_events, aio_context_t __user *, ctxp)
L
Linus Torvalds 已提交
904 905 906 907 908 909 910 911 912 913
{
	struct kioctx *ioctx = NULL;
	unsigned long ctx;
	long ret;

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

	ret = -EINVAL;
914 915 916
	if (unlikely(ctx || nr_events == 0)) {
		pr_debug("EINVAL: io_setup: ctx %lu nr_events %u\n",
		         ctx, nr_events);
L
Linus Torvalds 已提交
917 918 919 920 921 922 923
		goto out;
	}

	ioctx = ioctx_alloc(nr_events);
	ret = PTR_ERR(ioctx);
	if (!IS_ERR(ioctx)) {
		ret = put_user(ioctx->user_id, ctxp);
924
		if (ret)
K
Kent Overstreet 已提交
925
			kill_ioctx(ioctx);
926
		put_ioctx(ioctx);
L
Linus Torvalds 已提交
927 928 929 930 931 932 933 934 935
	}

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
936
 *	implemented.  May fail with -EINVAL if the context pointed to
L
Linus Torvalds 已提交
937 938
 *	is invalid.
 */
939
SYSCALL_DEFINE1(io_destroy, aio_context_t, ctx)
L
Linus Torvalds 已提交
940 941 942
{
	struct kioctx *ioctx = lookup_ioctx(ctx);
	if (likely(NULL != ioctx)) {
K
Kent Overstreet 已提交
943
		kill_ioctx(ioctx);
944
		put_ioctx(ioctx);
L
Linus Torvalds 已提交
945 946 947 948 949 950
		return 0;
	}
	pr_debug("EINVAL: io_destroy: invalid context id\n");
	return -EINVAL;
}

B
Badari Pulavarty 已提交
951
static void aio_advance_iovec(struct kiocb *iocb, ssize_t ret)
L
Linus Torvalds 已提交
952
{
B
Badari Pulavarty 已提交
953 954 955 956 957 958 959 960 961 962 963 964 965
	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++;
966
		}
B
Badari Pulavarty 已提交
967
	}
L
Linus Torvalds 已提交
968

B
Badari Pulavarty 已提交
969 970 971
	/* the caller should not have done more io than what fit in
	 * the remaining iovecs */
	BUG_ON(ret > 0 && iocb->ki_left == 0);
L
Linus Torvalds 已提交
972 973
}

B
Badari Pulavarty 已提交
974
static ssize_t aio_rw_vect_retry(struct kiocb *iocb)
L
Linus Torvalds 已提交
975 976
{
	struct file *file = iocb->ki_filp;
B
Badari Pulavarty 已提交
977 978 979 980
	struct address_space *mapping = file->f_mapping;
	struct inode *inode = mapping->host;
	ssize_t (*rw_op)(struct kiocb *, const struct iovec *,
			 unsigned long, loff_t);
L
Linus Torvalds 已提交
981
	ssize_t ret = 0;
B
Badari Pulavarty 已提交
982 983 984 985 986 987 988 989 990 991
	unsigned short opcode;

	if ((iocb->ki_opcode == IOCB_CMD_PREADV) ||
		(iocb->ki_opcode == IOCB_CMD_PREAD)) {
		rw_op = file->f_op->aio_read;
		opcode = IOCB_CMD_PREADV;
	} else {
		rw_op = file->f_op->aio_write;
		opcode = IOCB_CMD_PWRITEV;
	}
L
Linus Torvalds 已提交
992

993 994 995 996
	/* This matches the pread()/pwrite() logic */
	if (iocb->ki_pos < 0)
		return -EINVAL;

997 998
	if (opcode == IOCB_CMD_PWRITEV)
		file_start_write(file);
999
	do {
B
Badari Pulavarty 已提交
1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010
		ret = rw_op(iocb, &iocb->ki_iovec[iocb->ki_cur_seg],
			    iocb->ki_nr_segs - iocb->ki_cur_seg,
			    iocb->ki_pos);
		if (ret > 0)
			aio_advance_iovec(iocb, ret);

	/* retry all partial writes.  retry partial reads as long as its a
	 * regular file. */
	} while (ret > 0 && iocb->ki_left > 0 &&
		 (opcode == IOCB_CMD_PWRITEV ||
		  (!S_ISFIFO(inode->i_mode) && !S_ISSOCK(inode->i_mode))));
1011 1012
	if (opcode == IOCB_CMD_PWRITEV)
		file_end_write(file);
L
Linus Torvalds 已提交
1013

B
Badari Pulavarty 已提交
1014 1015
	/* This means we must have transferred all that we could */
	/* No need to retry anymore */
L
Linus Torvalds 已提交
1016 1017 1018
	if ((ret == 0) || (iocb->ki_left == 0))
		ret = iocb->ki_nbytes - iocb->ki_left;

1019 1020 1021
	/* If we managed to write some out we return that, rather than
	 * the eventual error. */
	if (opcode == IOCB_CMD_PWRITEV
Z
Zach Brown 已提交
1022
	    && ret < 0 && ret != -EIOCBQUEUED
1023 1024 1025
	    && iocb->ki_nbytes - iocb->ki_left)
		ret = iocb->ki_nbytes - iocb->ki_left;

L
Linus Torvalds 已提交
1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048
	return ret;
}

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

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

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

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

1049
static ssize_t aio_setup_vectored_rw(int type, struct kiocb *kiocb, bool compat)
B
Badari Pulavarty 已提交
1050 1051 1052
{
	ssize_t ret;

1053 1054 1055 1056 1057
#ifdef CONFIG_COMPAT
	if (compat)
		ret = compat_rw_copy_check_uvector(type,
				(struct compat_iovec __user *)kiocb->ki_buf,
				kiocb->ki_nbytes, 1, &kiocb->ki_inline_vec,
1058
				&kiocb->ki_iovec);
1059 1060 1061 1062 1063
	else
#endif
		ret = rw_copy_check_uvector(type,
				(struct iovec __user *)kiocb->ki_buf,
				kiocb->ki_nbytes, 1, &kiocb->ki_inline_vec,
1064
				&kiocb->ki_iovec);
B
Badari Pulavarty 已提交
1065 1066 1067
	if (ret < 0)
		goto out;

1068 1069 1070 1071
	ret = rw_verify_area(type, kiocb->ki_filp, &kiocb->ki_pos, ret);
	if (ret < 0)
		goto out;

B
Badari Pulavarty 已提交
1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082
	kiocb->ki_nr_segs = kiocb->ki_nbytes;
	kiocb->ki_cur_seg = 0;
	/* ki_nbytes/left now reflect bytes instead of segs */
	kiocb->ki_nbytes = ret;
	kiocb->ki_left = ret;

	ret = 0;
out:
	return ret;
}

1083
static ssize_t aio_setup_single_vector(int type, struct file * file, struct kiocb *kiocb)
B
Badari Pulavarty 已提交
1084
{
1085 1086 1087 1088 1089 1090
	int bytes;

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

B
Badari Pulavarty 已提交
1091 1092
	kiocb->ki_iovec = &kiocb->ki_inline_vec;
	kiocb->ki_iovec->iov_base = kiocb->ki_buf;
1093
	kiocb->ki_iovec->iov_len = bytes;
B
Badari Pulavarty 已提交
1094 1095 1096 1097 1098
	kiocb->ki_nr_segs = 1;
	kiocb->ki_cur_seg = 0;
	return 0;
}

L
Linus Torvalds 已提交
1099 1100 1101 1102 1103
/*
 * aio_setup_iocb:
 *	Performs the initial checks and aio retry method
 *	setup for the kiocb at the time of io submission.
 */
1104
static ssize_t aio_setup_iocb(struct kiocb *kiocb, bool compat)
L
Linus Torvalds 已提交
1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117
{
	struct file *file = kiocb->ki_filp;
	ssize_t ret = 0;

	switch (kiocb->ki_opcode) {
	case IOCB_CMD_PREAD:
		ret = -EBADF;
		if (unlikely(!(file->f_mode & FMODE_READ)))
			break;
		ret = -EFAULT;
		if (unlikely(!access_ok(VERIFY_WRITE, kiocb->ki_buf,
			kiocb->ki_left)))
			break;
1118
		ret = aio_setup_single_vector(READ, file, kiocb);
B
Badari Pulavarty 已提交
1119 1120
		if (ret)
			break;
L
Linus Torvalds 已提交
1121 1122
		ret = -EINVAL;
		if (file->f_op->aio_read)
B
Badari Pulavarty 已提交
1123
			kiocb->ki_retry = aio_rw_vect_retry;
L
Linus Torvalds 已提交
1124 1125 1126 1127 1128 1129 1130 1131 1132
		break;
	case IOCB_CMD_PWRITE:
		ret = -EBADF;
		if (unlikely(!(file->f_mode & FMODE_WRITE)))
			break;
		ret = -EFAULT;
		if (unlikely(!access_ok(VERIFY_READ, kiocb->ki_buf,
			kiocb->ki_left)))
			break;
1133
		ret = aio_setup_single_vector(WRITE, file, kiocb);
B
Badari Pulavarty 已提交
1134 1135 1136 1137 1138 1139 1140 1141 1142 1143
		if (ret)
			break;
		ret = -EINVAL;
		if (file->f_op->aio_write)
			kiocb->ki_retry = aio_rw_vect_retry;
		break;
	case IOCB_CMD_PREADV:
		ret = -EBADF;
		if (unlikely(!(file->f_mode & FMODE_READ)))
			break;
1144
		ret = aio_setup_vectored_rw(READ, kiocb, compat);
B
Badari Pulavarty 已提交
1145 1146 1147 1148 1149 1150 1151 1152 1153 1154
		if (ret)
			break;
		ret = -EINVAL;
		if (file->f_op->aio_read)
			kiocb->ki_retry = aio_rw_vect_retry;
		break;
	case IOCB_CMD_PWRITEV:
		ret = -EBADF;
		if (unlikely(!(file->f_mode & FMODE_WRITE)))
			break;
1155
		ret = aio_setup_vectored_rw(WRITE, kiocb, compat);
B
Badari Pulavarty 已提交
1156 1157
		if (ret)
			break;
L
Linus Torvalds 已提交
1158 1159
		ret = -EINVAL;
		if (file->f_op->aio_write)
B
Badari Pulavarty 已提交
1160
			kiocb->ki_retry = aio_rw_vect_retry;
L
Linus Torvalds 已提交
1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172
		break;
	case IOCB_CMD_FDSYNC:
		ret = -EINVAL;
		if (file->f_op->aio_fsync)
			kiocb->ki_retry = aio_fdsync;
		break;
	case IOCB_CMD_FSYNC:
		ret = -EINVAL;
		if (file->f_op->aio_fsync)
			kiocb->ki_retry = aio_fsync;
		break;
	default:
K
Kent Overstreet 已提交
1173
		pr_debug("EINVAL: no operation provided\n");
L
Linus Torvalds 已提交
1174 1175 1176 1177 1178 1179 1180 1181 1182
		ret = -EINVAL;
	}

	if (!kiocb->ki_retry)
		return ret;

	return 0;
}

A
Adrian Bunk 已提交
1183
static int io_submit_one(struct kioctx *ctx, struct iocb __user *user_iocb,
J
Jeff Moyer 已提交
1184 1185
			 struct iocb *iocb, struct kiocb_batch *batch,
			 bool compat)
L
Linus Torvalds 已提交
1186 1187 1188 1189 1190
{
	struct kiocb *req;
	ssize_t ret;

	/* enforce forwards compatibility on users */
1191
	if (unlikely(iocb->aio_reserved1 || iocb->aio_reserved2)) {
K
Kent Overstreet 已提交
1192
		pr_debug("EINVAL: reserve field set\n");
L
Linus Torvalds 已提交
1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205
		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;
	}

J
Jeff Moyer 已提交
1206
	req = aio_get_req(ctx, batch);  /* returns with 2 references to req */
1207
	if (unlikely(!req))
L
Linus Torvalds 已提交
1208
		return -EAGAIN;
1209 1210 1211 1212 1213

	req->ki_filp = fget(iocb->aio_fildes);
	if (unlikely(!req->ki_filp)) {
		ret = -EBADF;
		goto out_put_req;
L
Linus Torvalds 已提交
1214
	}
1215

1216 1217 1218 1219 1220 1221 1222
	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.
		 */
1223
		req->ki_eventfd = eventfd_ctx_fdget((int) iocb->aio_resfd);
1224
		if (IS_ERR(req->ki_eventfd)) {
1225
			ret = PTR_ERR(req->ki_eventfd);
1226
			req->ki_eventfd = NULL;
1227 1228 1229
			goto out_put_req;
		}
	}
L
Linus Torvalds 已提交
1230

1231
	ret = put_user(req->ki_key, &user_iocb->aio_key);
L
Linus Torvalds 已提交
1232
	if (unlikely(ret)) {
K
Kent Overstreet 已提交
1233
		pr_debug("EFAULT: aio_key\n");
L
Linus Torvalds 已提交
1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244
		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;

1245
	ret = aio_setup_iocb(req, compat);
Z
Zach Brown 已提交
1246
	if (ret)
1247
		goto out_put_req;
Z
Zach Brown 已提交
1248

1249
	ret = req->ki_retry(req);
Z
Zach Brown 已提交
1250 1251 1252 1253 1254 1255 1256 1257 1258 1259
	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);
1260
	}
J
Jeff Moyer 已提交
1261

L
Linus Torvalds 已提交
1262 1263 1264 1265
	aio_put_req(req);	/* drop extra ref to req */
	return 0;

out_put_req:
1266
	atomic_dec(&ctx->reqs_active);
L
Linus Torvalds 已提交
1267 1268 1269 1270 1271
	aio_put_req(req);	/* drop extra ref to req */
	aio_put_req(req);	/* drop i/o ref to req */
	return ret;
}

1272 1273
long do_io_submit(aio_context_t ctx_id, long nr,
		  struct iocb __user *__user *iocbpp, bool compat)
L
Linus Torvalds 已提交
1274 1275 1276
{
	struct kioctx *ctx;
	long ret = 0;
J
Jeff Moyer 已提交
1277
	int i = 0;
S
Shaohua Li 已提交
1278
	struct blk_plug plug;
J
Jeff Moyer 已提交
1279
	struct kiocb_batch batch;
L
Linus Torvalds 已提交
1280 1281 1282 1283

	if (unlikely(nr < 0))
		return -EINVAL;

1284 1285 1286
	if (unlikely(nr > LONG_MAX/sizeof(*iocbpp)))
		nr = LONG_MAX/sizeof(*iocbpp);

L
Linus Torvalds 已提交
1287 1288 1289 1290 1291
	if (unlikely(!access_ok(VERIFY_READ, iocbpp, (nr*sizeof(*iocbpp)))))
		return -EFAULT;

	ctx = lookup_ioctx(ctx_id);
	if (unlikely(!ctx)) {
K
Kent Overstreet 已提交
1292
		pr_debug("EINVAL: invalid context id\n");
L
Linus Torvalds 已提交
1293 1294 1295
		return -EINVAL;
	}

J
Jeff Moyer 已提交
1296 1297
	kiocb_batch_init(&batch, nr);

S
Shaohua Li 已提交
1298 1299
	blk_start_plug(&plug);

L
Linus Torvalds 已提交
1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317
	/*
	 * AKPM: should this return a partial result if some of the IOs were
	 * successfully submitted?
	 */
	for (i=0; i<nr; i++) {
		struct iocb __user *user_iocb;
		struct iocb tmp;

		if (unlikely(__get_user(user_iocb, iocbpp + i))) {
			ret = -EFAULT;
			break;
		}

		if (unlikely(copy_from_user(&tmp, user_iocb, sizeof(tmp)))) {
			ret = -EFAULT;
			break;
		}

J
Jeff Moyer 已提交
1318
		ret = io_submit_one(ctx, user_iocb, &tmp, &batch, compat);
L
Linus Torvalds 已提交
1319 1320 1321
		if (ret)
			break;
	}
S
Shaohua Li 已提交
1322
	blk_finish_plug(&plug);
L
Linus Torvalds 已提交
1323

1324
	kiocb_batch_free(ctx, &batch);
L
Linus Torvalds 已提交
1325 1326 1327 1328
	put_ioctx(ctx);
	return i ? i : ret;
}

1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346
/* sys_io_submit:
 *	Queue the nr iocbs pointed to by iocbpp for processing.  Returns
 *	the number of iocbs queued.  May return -EINVAL if the aio_context
 *	specified by ctx_id is invalid, if nr is < 0, if the iocb at
 *	*iocbpp[0] is not properly initialized, if the operation specified
 *	is invalid for the file descriptor in the iocb.  May fail with
 *	-EFAULT if any of the data structures point to invalid data.  May
 *	fail with -EBADF if the file descriptor specified in the first
 *	iocb is invalid.  May fail with -EAGAIN if insufficient resources
 *	are available to queue any iocbs.  Will return 0 if nr is 0.  Will
 *	fail with -ENOSYS if not implemented.
 */
SYSCALL_DEFINE3(io_submit, aio_context_t, ctx_id, long, nr,
		struct iocb __user * __user *, iocbpp)
{
	return do_io_submit(ctx_id, nr, iocbpp, 0);
}

L
Linus Torvalds 已提交
1347 1348 1349
/* lookup_kiocb
 *	Finds a given iocb for cancellation.
 */
1350 1351
static struct kiocb *lookup_kiocb(struct kioctx *ctx, struct iocb __user *iocb,
				  u32 key)
L
Linus Torvalds 已提交
1352 1353
{
	struct list_head *pos;
1354 1355 1356

	assert_spin_locked(&ctx->ctx_lock);

L
Linus Torvalds 已提交
1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375
	/* TODO: use a hash or array, this sucks. */
	list_for_each(pos, &ctx->active_reqs) {
		struct kiocb *kiocb = list_kiocb(pos);
		if (kiocb->ki_obj.user == iocb && kiocb->ki_key == key)
			return kiocb;
	}
	return NULL;
}

/* sys_io_cancel:
 *	Attempts to cancel an iocb previously passed to io_submit.  If
 *	the operation is successfully cancelled, the resulting event is
 *	copied into the memory pointed to by result without being placed
 *	into the completion queue and 0 is returned.  May fail with
 *	-EFAULT if any of the data structures pointed to are invalid.
 *	May fail with -EINVAL if aio_context specified by ctx_id is
 *	invalid.  May fail with -EAGAIN if the iocb specified was not
 *	cancelled.  Will fail with -ENOSYS if not implemented.
 */
1376 1377
SYSCALL_DEFINE3(io_cancel, aio_context_t, ctx_id, struct iocb __user *, iocb,
		struct io_event __user *, result)
L
Linus Torvalds 已提交
1378
{
K
Kent Overstreet 已提交
1379
	struct io_event res;
L
Linus Torvalds 已提交
1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393
	struct kioctx *ctx;
	struct kiocb *kiocb;
	u32 key;
	int ret;

	ret = get_user(key, &iocb->aio_key);
	if (unlikely(ret))
		return -EFAULT;

	ctx = lookup_ioctx(ctx_id);
	if (unlikely(!ctx))
		return -EINVAL;

	spin_lock_irq(&ctx->ctx_lock);
K
Kent Overstreet 已提交
1394

L
Linus Torvalds 已提交
1395
	kiocb = lookup_kiocb(ctx, iocb, key);
K
Kent Overstreet 已提交
1396 1397 1398 1399 1400
	if (kiocb)
		ret = kiocb_cancel(ctx, kiocb, &res);
	else
		ret = -EINVAL;

L
Linus Torvalds 已提交
1401 1402
	spin_unlock_irq(&ctx->ctx_lock);

K
Kent Overstreet 已提交
1403 1404 1405 1406 1407 1408 1409
	if (!ret) {
		/* Cancellation succeeded -- copy the result
		 * into the user's buffer.
		 */
		if (copy_to_user(result, &res, sizeof(res)))
			ret = -EFAULT;
	}
L
Linus Torvalds 已提交
1410 1411 1412 1413 1414 1415 1416 1417

	put_ioctx(ctx);

	return ret;
}

/* io_getevents:
 *	Attempts to read at least min_nr events and up to nr events from
1418 1419 1420 1421 1422 1423 1424 1425 1426 1427
 *	the completion queue for the aio_context specified by ctx_id. If
 *	it succeeds, the number of read events is returned. May fail with
 *	-EINVAL if ctx_id is invalid, if min_nr is out of range, if nr is
 *	out of range, if timeout is out of range.  May fail with -EFAULT
 *	if any of the memory specified is invalid.  May return 0 or
 *	< min_nr if the timeout specified by timeout has elapsed
 *	before sufficient events are available, where timeout == NULL
 *	specifies an infinite timeout. Note that the timeout pointed to by
 *	timeout is relative and will be updated if not NULL and the
 *	operation blocks. Will fail with -ENOSYS if not implemented.
L
Linus Torvalds 已提交
1428
 */
1429 1430 1431 1432 1433
SYSCALL_DEFINE5(io_getevents, aio_context_t, ctx_id,
		long, min_nr,
		long, nr,
		struct io_event __user *, events,
		struct timespec __user *, timeout)
L
Linus Torvalds 已提交
1434 1435 1436 1437 1438
{
	struct kioctx *ioctx = lookup_ioctx(ctx_id);
	long ret = -EINVAL;

	if (likely(ioctx)) {
N
Namhyung Kim 已提交
1439
		if (likely(min_nr <= nr && min_nr >= 0))
L
Linus Torvalds 已提交
1440 1441 1442 1443 1444
			ret = read_events(ioctx, min_nr, nr, events, timeout);
		put_ioctx(ioctx);
	}
	return ret;
}