splice.c 39.9 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11
/*
 * "splice": joining two ropes together by interweaving their strands.
 *
 * This is the "extended pipe" functionality, where a pipe is used as
 * an arbitrary in-memory buffer. Think of a pipe as a small kernel
 * buffer that you can use to transfer data from one end to the other.
 *
 * The traditional unix read/write is extended with a "splice()" operation
 * that transfers data buffers to or from a pipe buffer.
 *
 * Named by Larry McVoy, original implementation from Linus, extended by
12 13
 * Jens to support splicing to files, network, direct splicing, etc and
 * fixing lots of bugs.
14
 *
15
 * Copyright (C) 2005-2006 Jens Axboe <axboe@kernel.dk>
16 17
 * Copyright (C) 2005-2006 Linus Torvalds <torvalds@osdl.org>
 * Copyright (C) 2006 Ingo Molnar <mingo@elte.hu>
18 19
 *
 */
20
#include <linux/bvec.h>
21 22 23
#include <linux/fs.h>
#include <linux/file.h>
#include <linux/pagemap.h>
24
#include <linux/splice.h>
25
#include <linux/memcontrol.h>
26
#include <linux/mm_inline.h>
27
#include <linux/swap.h>
28
#include <linux/writeback.h>
29
#include <linux/export.h>
30
#include <linux/syscalls.h>
31
#include <linux/uio.h>
32
#include <linux/security.h>
33
#include <linux/gfp.h>
34
#include <linux/socket.h>
35
#include <linux/compat.h>
36 37
#include <linux/sched/signal.h>

38
#include "internal.h"
39

40 41 42 43 44 45
/*
 * Attempt to steal a page from a pipe buffer. This should perhaps go into
 * a vm helper function, it's already simplified quite a bit by the
 * addition of remove_mapping(). If success is returned, the caller may
 * attempt to reuse this page for another destination.
 */
46
static int page_cache_pipe_buf_steal(struct pipe_inode_info *pipe,
47 48 49
				     struct pipe_buffer *buf)
{
	struct page *page = buf->page;
50
	struct address_space *mapping;
51

52 53
	lock_page(page);

54 55 56
	mapping = page_mapping(page);
	if (mapping) {
		WARN_ON(!PageUptodate(page));
57

58 59 60 61 62 63 64 65 66
		/*
		 * At least for ext2 with nobh option, we need to wait on
		 * writeback completing on this page, since we'll remove it
		 * from the pagecache.  Otherwise truncate wont wait on the
		 * page, allowing the disk blocks to be reused by someone else
		 * before we actually wrote our data to them. fs corruption
		 * ensues.
		 */
		wait_on_page_writeback(page);
67

68 69
		if (page_has_private(page) &&
		    !try_to_release_page(page, GFP_KERNEL))
70
			goto out_unlock;
71

72 73 74 75 76 77 78 79
		/*
		 * If we succeeded in removing the mapping, set LRU flag
		 * and return good.
		 */
		if (remove_mapping(mapping, page)) {
			buf->flags |= PIPE_BUF_FLAG_LRU;
			return 0;
		}
80
	}
81

82 83 84 85
	/*
	 * Raced with truncate or failed to remove page from current
	 * address space, unlock and return failure.
	 */
86
out_unlock:
87 88
	unlock_page(page);
	return 1;
89 90
}

91
static void page_cache_pipe_buf_release(struct pipe_inode_info *pipe,
92 93
					struct pipe_buffer *buf)
{
94
	put_page(buf->page);
Jens Axboe's avatar
Jens Axboe committed
95
	buf->flags &= ~PIPE_BUF_FLAG_LRU;
96 97
}

98 99 100 101
/*
 * Check whether the contents of buf is OK to access. Since the content
 * is a page cache page, IO may be in flight.
 */
102 103
static int page_cache_pipe_buf_confirm(struct pipe_inode_info *pipe,
				       struct pipe_buffer *buf)
104 105
{
	struct page *page = buf->page;
106
	int err;
107 108

	if (!PageUptodate(page)) {
109 110 111 112
		lock_page(page);

		/*
		 * Page got truncated/unhashed. This will cause a 0-byte
113
		 * splice, if this is the first page.
114 115 116 117 118
		 */
		if (!page->mapping) {
			err = -ENODATA;
			goto error;
		}
119

120
		/*
121
		 * Uh oh, read-error from disk.
122 123 124 125 126 127 128
		 */
		if (!PageUptodate(page)) {
			err = -EIO;
			goto error;
		}

		/*
129
		 * Page is ok afterall, we are done.
130
		 */
131 132 133
		unlock_page(page);
	}

134
	return 0;
135 136
error:
	unlock_page(page);
137
	return err;
138 139
}

140
const struct pipe_buf_operations page_cache_pipe_buf_ops = {
141
	.can_merge = 0,
142
	.confirm = page_cache_pipe_buf_confirm,
143
	.release = page_cache_pipe_buf_release,
144
	.steal = page_cache_pipe_buf_steal,
145
	.get = generic_pipe_buf_get,
146 147
};

148 149 150
static int user_page_pipe_buf_steal(struct pipe_inode_info *pipe,
				    struct pipe_buffer *buf)
{
151 152 153
	if (!(buf->flags & PIPE_BUF_FLAG_GIFT))
		return 1;

Jens Axboe's avatar
Jens Axboe committed
154
	buf->flags |= PIPE_BUF_FLAG_LRU;
155
	return generic_pipe_buf_steal(pipe, buf);
156 157
}

158
static const struct pipe_buf_operations user_page_pipe_buf_ops = {
159
	.can_merge = 0,
160
	.confirm = generic_pipe_buf_confirm,
161 162
	.release = page_cache_pipe_buf_release,
	.steal = user_page_pipe_buf_steal,
163
	.get = generic_pipe_buf_get,
164 165
};

166 167 168 169 170 171 172 173
static void wakeup_pipe_readers(struct pipe_inode_info *pipe)
{
	smp_mb();
	if (waitqueue_active(&pipe->wait))
		wake_up_interruptible(&pipe->wait);
	kill_fasync(&pipe->fasync_readers, SIGIO, POLL_IN);
}

174 175 176 177 178 179
/**
 * splice_to_pipe - fill passed data into a pipe
 * @pipe:	pipe to fill
 * @spd:	data to fill
 *
 * Description:
180
 *    @spd contains a map of pages and len/offset tuples, along with
181 182 183
 *    the struct pipe_buf_operations associated with these pages. This
 *    function will link that data to the pipe.
 *
184
 */
185 186
ssize_t splice_to_pipe(struct pipe_inode_info *pipe,
		       struct splice_pipe_desc *spd)
187
{
188
	unsigned int spd_pages = spd->nr_pages;
189
	int ret = 0, page_nr = 0;
190

191 192 193
	if (!spd_pages)
		return 0;

194 195 196 197 198
	if (unlikely(!pipe->readers)) {
		send_sig(SIGPIPE, current, 0);
		ret = -EPIPE;
		goto out;
	}
199

200 201 202
	while (pipe->nrbufs < pipe->buffers) {
		int newbuf = (pipe->curbuf + pipe->nrbufs) & (pipe->buffers - 1);
		struct pipe_buffer *buf = pipe->bufs + newbuf;
203

204 205 206 207 208
		buf->page = spd->pages[page_nr];
		buf->offset = spd->partial[page_nr].offset;
		buf->len = spd->partial[page_nr].len;
		buf->private = spd->partial[page_nr].private;
		buf->ops = spd->ops;
209
		buf->flags = 0;
210

211 212 213
		pipe->nrbufs++;
		page_nr++;
		ret += buf->len;
214

215
		if (!--spd->nr_pages)
216 217 218
			break;
	}

219 220
	if (!ret)
		ret = -EAGAIN;
221

222
out:
223
	while (page_nr < spd_pages)
224
		spd->spd_release(spd, page_nr++);
225 226 227

	return ret;
}
228
EXPORT_SYMBOL_GPL(splice_to_pipe);
229

Al Viro's avatar
Al Viro committed
230 231 232 233 234 235 236 237 238 239 240 241 242 243 244
ssize_t add_to_pipe(struct pipe_inode_info *pipe, struct pipe_buffer *buf)
{
	int ret;

	if (unlikely(!pipe->readers)) {
		send_sig(SIGPIPE, current, 0);
		ret = -EPIPE;
	} else if (pipe->nrbufs == pipe->buffers) {
		ret = -EAGAIN;
	} else {
		int newbuf = (pipe->curbuf + pipe->nrbufs) & (pipe->buffers - 1);
		pipe->bufs[newbuf] = *buf;
		pipe->nrbufs++;
		return buf->len;
	}
245
	pipe_buf_release(pipe, buf);
Al Viro's avatar
Al Viro committed
246 247 248 249
	return ret;
}
EXPORT_SYMBOL(add_to_pipe);

250 251 252 253
/*
 * Check if we need to grow the arrays holding pages and partial page
 * descriptions.
 */
254
int splice_grow_spd(const struct pipe_inode_info *pipe, struct splice_pipe_desc *spd)
255
{
256
	unsigned int buffers = READ_ONCE(pipe->buffers);
257 258 259

	spd->nr_pages_max = buffers;
	if (buffers <= PIPE_DEF_BUFFERS)
260 261
		return 0;

262 263 264
	spd->pages = kmalloc_array(buffers, sizeof(struct page *), GFP_KERNEL);
	spd->partial = kmalloc_array(buffers, sizeof(struct partial_page),
				     GFP_KERNEL);
265 266 267 268 269 270 271 272 273

	if (spd->pages && spd->partial)
		return 0;

	kfree(spd->pages);
	kfree(spd->partial);
	return -ENOMEM;
}

274
void splice_shrink_spd(struct splice_pipe_desc *spd)
275
{
276
	if (spd->nr_pages_max <= PIPE_DEF_BUFFERS)
277 278 279 280 281 282
		return;

	kfree(spd->pages);
	kfree(spd->partial);
}

283 284 285
/**
 * generic_file_splice_read - splice data from file to a pipe
 * @in:		file to splice from
286
 * @ppos:	position in @in
287 288 289 290
 * @pipe:	pipe to splice to
 * @len:	number of bytes to splice
 * @flags:	splice modifier flags
 *
291 292
 * Description:
 *    Will read pages from given file and fill them into a pipe. Can be
293
 *    used as long as it has more or less sane ->read_iter().
294
 *
295
 */
296 297 298
ssize_t generic_file_splice_read(struct file *in, loff_t *ppos,
				 struct pipe_inode_info *pipe, size_t len,
				 unsigned int flags)
299
{
300 301 302
	struct iov_iter to;
	struct kiocb kiocb;
	int idx, ret;
303

304
	iov_iter_pipe(&to, READ, pipe, len);
305 306 307
	idx = to.idx;
	init_sync_kiocb(&kiocb, in);
	kiocb.ki_pos = *ppos;
308
	ret = call_read_iter(in, &kiocb, &to);
309
	if (ret > 0) {
310
		*ppos = kiocb.ki_pos;
311
		file_accessed(in);
312
	} else if (ret < 0) {
313 314 315
		to.idx = idx;
		to.iov_offset = 0;
		iov_iter_advance(&to, 0); /* to free what was emitted */
316 317 318 319 320 321
		/*
		 * callers of ->splice_read() expect -EAGAIN on
		 * "can't put anything in there", rather than -EFAULT.
		 */
		if (ret == -EFAULT)
			ret = -EAGAIN;
322
	}
323 324 325

	return ret;
}
326 327
EXPORT_SYMBOL(generic_file_splice_read);

328
const struct pipe_buf_operations default_pipe_buf_ops = {
329 330 331 332 333 334 335
	.can_merge = 0,
	.confirm = generic_pipe_buf_confirm,
	.release = generic_pipe_buf_release,
	.steal = generic_pipe_buf_steal,
	.get = generic_pipe_buf_get,
};

336 337
int generic_pipe_buf_nosteal(struct pipe_inode_info *pipe,
			     struct pipe_buffer *buf)
338 339 340 341 342 343 344 345 346 347 348 349 350 351
{
	return 1;
}

/* Pipe buffer operations for a socket and similar. */
const struct pipe_buf_operations nosteal_pipe_buf_ops = {
	.can_merge = 0,
	.confirm = generic_pipe_buf_confirm,
	.release = generic_pipe_buf_release,
	.steal = generic_pipe_buf_nosteal,
	.get = generic_pipe_buf_get,
};
EXPORT_SYMBOL(nosteal_pipe_buf_ops);

352
static ssize_t kernel_readv(struct file *file, const struct kvec *vec,
353 354 355 356 357 358 359 360 361
			    unsigned long vlen, loff_t offset)
{
	mm_segment_t old_fs;
	loff_t pos = offset;
	ssize_t res;

	old_fs = get_fs();
	set_fs(get_ds());
	/* The cast to a user pointer is valid due to the set_fs() */
362
	res = vfs_readv(file, (const struct iovec __user *)vec, vlen, &pos, 0);
363 364 365 366 367
	set_fs(old_fs);

	return res;
}

368
static ssize_t default_file_splice_read(struct file *in, loff_t *ppos,
369 370 371
				 struct pipe_inode_info *pipe, size_t len,
				 unsigned int flags)
{
372 373 374
	struct kvec *vec, __vec[PIPE_DEF_BUFFERS];
	struct iov_iter to;
	struct page **pages;
375
	unsigned int nr_pages;
376
	size_t offset, base, copied = 0;
377 378
	ssize_t res;
	int i;
379

380 381
	if (pipe->nrbufs == pipe->buffers)
		return -EAGAIN;
382

383 384 385 386
	/*
	 * Try to keep page boundaries matching to source pagecache ones -
	 * it probably won't be much help, but...
	 */
387
	offset = *ppos & ~PAGE_MASK;
388

389
	iov_iter_pipe(&to, READ, pipe, len + offset);
390

391
	res = iov_iter_get_pages_alloc(&to, &pages, len + offset, &base);
392 393
	if (res <= 0)
		return -ENOMEM;
394

395
	nr_pages = DIV_ROUND_UP(res + base, PAGE_SIZE);
396

397 398
	vec = __vec;
	if (nr_pages > PIPE_DEF_BUFFERS) {
399
		vec = kmalloc_array(nr_pages, sizeof(struct kvec), GFP_KERNEL);
400 401 402 403
		if (unlikely(!vec)) {
			res = -ENOMEM;
			goto out;
		}
404
	}
405

406 407 408 409 410 411 412 413 414
	pipe->bufs[to.idx].offset = offset;
	pipe->bufs[to.idx].len -= offset;

	for (i = 0; i < nr_pages; i++) {
		size_t this_len = min_t(size_t, len, PAGE_SIZE - offset);
		vec[i].iov_base = page_address(pages[i]) + offset;
		vec[i].iov_len = this_len;
		len -= this_len;
		offset = 0;
415 416
	}

417 418 419
	res = kernel_readv(in, vec, nr_pages, *ppos);
	if (res > 0) {
		copied = res;
420
		*ppos += res;
421
	}
422

423 424
	if (vec != __vec)
		kfree(vec);
425 426 427 428 429
out:
	for (i = 0; i < nr_pages; i++)
		put_page(pages[i]);
	kvfree(pages);
	iov_iter_advance(&to, copied);	/* truncates and discards */
430 431 432
	return res;
}

433
/*
434
 * Send 'sd->len' bytes to socket from 'sd->file' at position 'sd->pos'
435
 * using sendpage(). Return the number of bytes sent.
436
 */
437
static int pipe_to_sendpage(struct pipe_inode_info *pipe,
438 439
			    struct pipe_buffer *buf, struct splice_desc *sd)
{
440
	struct file *file = sd->u.file;
441
	loff_t pos = sd->pos;
442
	int more;
443

Al Viro's avatar
Al Viro committed
444
	if (!likely(file->f_op->sendpage))
445 446
		return -EINVAL;

447
	more = (sd->flags & SPLICE_F_MORE) ? MSG_MORE : 0;
448 449

	if (sd->len < sd->total_len && pipe->nrbufs > 1)
450
		more |= MSG_SENDPAGE_NOTLAST;
451

452 453
	return file->f_op->sendpage(file, buf->page, buf->offset,
				    sd->len, &pos, more);
454 455
}

456 457 458 459 460 461 462 463
static void wakeup_pipe_writers(struct pipe_inode_info *pipe)
{
	smp_mb();
	if (waitqueue_active(&pipe->wait))
		wake_up_interruptible(&pipe->wait);
	kill_fasync(&pipe->fasync_writers, SIGIO, POLL_OUT);
}

464
/**
465
 * splice_from_pipe_feed - feed available data from a pipe to a file
466 467 468 469 470
 * @pipe:	pipe to splice from
 * @sd:		information to @actor
 * @actor:	handler that splices the data
 *
 * Description:
471 472 473 474 475 476 477
 *    This function loops over the pipe and calls @actor to do the
 *    actual moving of a single struct pipe_buffer to the desired
 *    destination.  It returns when there's no more buffers left in
 *    the pipe or if the requested number of bytes (@sd->total_len)
 *    have been copied.  It returns a positive number (one) if the
 *    pipe needs to be filled with more data, zero if the required
 *    number of bytes have been copied and -errno on error.
478
 *
479 480 481 482
 *    This, together with splice_from_pipe_{begin,end,next}, may be
 *    used to implement the functionality of __splice_from_pipe() when
 *    locking is required around copying the pipe buffers to the
 *    destination.
483
 */
484
static int splice_from_pipe_feed(struct pipe_inode_info *pipe, struct splice_desc *sd,
485
			  splice_actor *actor)
486
{
487
	int ret;
488

489 490
	while (pipe->nrbufs) {
		struct pipe_buffer *buf = pipe->bufs + pipe->curbuf;
491

492 493 494
		sd->len = buf->len;
		if (sd->len > sd->total_len)
			sd->len = sd->total_len;
495

496
		ret = pipe_buf_confirm(pipe, buf);
497
		if (unlikely(ret)) {
498 499 500 501
			if (ret == -ENODATA)
				ret = 0;
			return ret;
		}
502 503 504 505 506

		ret = actor(pipe, buf, sd);
		if (ret <= 0)
			return ret;

507 508 509 510 511 512 513 514 515
		buf->offset += ret;
		buf->len -= ret;

		sd->num_spliced += ret;
		sd->len -= ret;
		sd->pos += ret;
		sd->total_len -= ret;

		if (!buf->len) {
516
			pipe_buf_release(pipe, buf);
517
			pipe->curbuf = (pipe->curbuf + 1) & (pipe->buffers - 1);
518
			pipe->nrbufs--;
Al Viro's avatar
Al Viro committed
519
			if (pipe->files)
520 521
				sd->need_wakeup = true;
		}
522

523 524 525
		if (!sd->total_len)
			return 0;
	}
526

527 528
	return 1;
}
529

530 531 532 533 534 535 536 537 538 539
/**
 * splice_from_pipe_next - wait for some data to splice from
 * @pipe:	pipe to splice from
 * @sd:		information about the splice operation
 *
 * Description:
 *    This function will wait for some data and return a positive
 *    value (one) if pipe buffers are available.  It will return zero
 *    or -errno if no more data needs to be spliced.
 */
540
static int splice_from_pipe_next(struct pipe_inode_info *pipe, struct splice_desc *sd)
541
{
542 543 544 545 546 547 548
	/*
	 * Check for signal early to make process killable when there are
	 * always buffers available
	 */
	if (signal_pending(current))
		return -ERESTARTSYS;

549 550 551
	while (!pipe->nrbufs) {
		if (!pipe->writers)
			return 0;
552

553 554
		if (!pipe->waiting_writers && sd->num_spliced)
			return 0;
555

556 557
		if (sd->flags & SPLICE_F_NONBLOCK)
			return -EAGAIN;
558

559 560
		if (signal_pending(current))
			return -ERESTARTSYS;
561

562 563 564
		if (sd->need_wakeup) {
			wakeup_pipe_writers(pipe);
			sd->need_wakeup = false;
565 566
		}

567 568
		pipe_wait(pipe);
	}
569

570 571
	return 1;
}
572

573 574
/**
 * splice_from_pipe_begin - start splicing from pipe
575
 * @sd:		information about the splice operation
576 577 578 579 580 581
 *
 * Description:
 *    This function should be called before a loop containing
 *    splice_from_pipe_next() and splice_from_pipe_feed() to
 *    initialize the necessary fields of @sd.
 */
582
static void splice_from_pipe_begin(struct splice_desc *sd)
583 584 585 586
{
	sd->num_spliced = 0;
	sd->need_wakeup = false;
}
587

588 589 590 591 592 593 594 595 596 597
/**
 * splice_from_pipe_end - finish splicing from pipe
 * @pipe:	pipe to splice from
 * @sd:		information about the splice operation
 *
 * Description:
 *    This function will wake up pipe writers if necessary.  It should
 *    be called after a loop containing splice_from_pipe_next() and
 *    splice_from_pipe_feed().
 */
598
static void splice_from_pipe_end(struct pipe_inode_info *pipe, struct splice_desc *sd)
599 600 601 602
{
	if (sd->need_wakeup)
		wakeup_pipe_writers(pipe);
}
603

604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620
/**
 * __splice_from_pipe - splice data from a pipe to given actor
 * @pipe:	pipe to splice from
 * @sd:		information to @actor
 * @actor:	handler that splices the data
 *
 * Description:
 *    This function does little more than loop over the pipe and call
 *    @actor to do the actual moving of a single struct pipe_buffer to
 *    the desired destination. See pipe_to_file, pipe_to_sendpage, or
 *    pipe_to_user.
 *
 */
ssize_t __splice_from_pipe(struct pipe_inode_info *pipe, struct splice_desc *sd,
			   splice_actor *actor)
{
	int ret;
621

622 623
	splice_from_pipe_begin(sd);
	do {
624
		cond_resched();
625 626 627 628 629 630 631
		ret = splice_from_pipe_next(pipe, sd);
		if (ret > 0)
			ret = splice_from_pipe_feed(pipe, sd, actor);
	} while (ret > 0);
	splice_from_pipe_end(pipe, sd);

	return sd->num_spliced ? sd->num_spliced : ret;
632
}
633
EXPORT_SYMBOL(__splice_from_pipe);
634

635 636 637 638 639 640 641 642 643 644
/**
 * splice_from_pipe - splice data from a pipe to a file
 * @pipe:	pipe to splice from
 * @out:	file to splice to
 * @ppos:	position in @out
 * @len:	how many bytes to splice
 * @flags:	splice modifier flags
 * @actor:	handler that splices the data
 *
 * Description:
645
 *    See __splice_from_pipe. This function locks the pipe inode,
646 647 648
 *    otherwise it's identical to __splice_from_pipe().
 *
 */
649 650 651 652 653
ssize_t splice_from_pipe(struct pipe_inode_info *pipe, struct file *out,
			 loff_t *ppos, size_t len, unsigned int flags,
			 splice_actor *actor)
{
	ssize_t ret;
654 655 656 657
	struct splice_desc sd = {
		.total_len = len,
		.flags = flags,
		.pos = *ppos,
658
		.u.file = out,
659
	};
660

661
	pipe_lock(pipe);
662
	ret = __splice_from_pipe(pipe, &sd, actor);
663
	pipe_unlock(pipe);
664 665 666 667

	return ret;
}

668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734
/**
 * iter_file_splice_write - splice data from a pipe to a file
 * @pipe:	pipe info
 * @out:	file to write to
 * @ppos:	position in @out
 * @len:	number of bytes to splice
 * @flags:	splice modifier flags
 *
 * Description:
 *    Will either move or copy pages (determined by @flags options) from
 *    the given pipe inode to the given file.
 *    This one is ->write_iter-based.
 *
 */
ssize_t
iter_file_splice_write(struct pipe_inode_info *pipe, struct file *out,
			  loff_t *ppos, size_t len, unsigned int flags)
{
	struct splice_desc sd = {
		.total_len = len,
		.flags = flags,
		.pos = *ppos,
		.u.file = out,
	};
	int nbufs = pipe->buffers;
	struct bio_vec *array = kcalloc(nbufs, sizeof(struct bio_vec),
					GFP_KERNEL);
	ssize_t ret;

	if (unlikely(!array))
		return -ENOMEM;

	pipe_lock(pipe);

	splice_from_pipe_begin(&sd);
	while (sd.total_len) {
		struct iov_iter from;
		size_t left;
		int n, idx;

		ret = splice_from_pipe_next(pipe, &sd);
		if (ret <= 0)
			break;

		if (unlikely(nbufs < pipe->buffers)) {
			kfree(array);
			nbufs = pipe->buffers;
			array = kcalloc(nbufs, sizeof(struct bio_vec),
					GFP_KERNEL);
			if (!array) {
				ret = -ENOMEM;
				break;
			}
		}

		/* build the vector */
		left = sd.total_len;
		for (n = 0, idx = pipe->curbuf; left && n < pipe->nrbufs; n++, idx++) {
			struct pipe_buffer *buf = pipe->bufs + idx;
			size_t this_len = buf->len;

			if (this_len > left)
				this_len = left;

			if (idx == pipe->buffers - 1)
				idx = -1;

735
			ret = pipe_buf_confirm(pipe, buf);
736 737 738 739 740 741 742 743 744 745 746 747
			if (unlikely(ret)) {
				if (ret == -ENODATA)
					ret = 0;
				goto done;
			}

			array[n].bv_page = buf->page;
			array[n].bv_len = this_len;
			array[n].bv_offset = buf->offset;
			left -= this_len;
		}

748
		iov_iter_bvec(&from, WRITE, array, n, sd.total_len - left);
749
		ret = vfs_iter_write(out, &from, &sd.pos, 0);
750 751 752 753 754
		if (ret <= 0)
			break;

		sd.num_spliced += ret;
		sd.total_len -= ret;
755
		*ppos = sd.pos;
756 757 758 759 760 761 762

		/* dismiss the fully eaten buffers, adjust the partial one */
		while (ret) {
			struct pipe_buffer *buf = pipe->bufs + pipe->curbuf;
			if (ret >= buf->len) {
				ret -= buf->len;
				buf->len = 0;
763
				pipe_buf_release(pipe, buf);
764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788
				pipe->curbuf = (pipe->curbuf + 1) & (pipe->buffers - 1);
				pipe->nrbufs--;
				if (pipe->files)
					sd.need_wakeup = true;
			} else {
				buf->offset += ret;
				buf->len -= ret;
				ret = 0;
			}
		}
	}
done:
	kfree(array);
	splice_from_pipe_end(pipe, &sd);

	pipe_unlock(pipe);

	if (sd.num_spliced)
		ret = sd.num_spliced;

	return ret;
}

EXPORT_SYMBOL(iter_file_splice_write);

789 790
static int write_pipe_buf(struct pipe_inode_info *pipe, struct pipe_buffer *buf,
			  struct splice_desc *sd)
791
{
792 793
	int ret;
	void *data;
794
	loff_t tmp = sd->pos;
795

796
	data = kmap(buf->page);
797
	ret = __kernel_write(sd->u.file, data + buf->offset, sd->len, &tmp);
798
	kunmap(buf->page);
799 800

	return ret;
801 802 803 804 805 806
}

static ssize_t default_file_splice_write(struct pipe_inode_info *pipe,
					 struct file *out, loff_t *ppos,
					 size_t len, unsigned int flags)
{
807
	ssize_t ret;
808

809 810 811
	ret = splice_from_pipe(pipe, out, ppos, len, flags, write_pipe_buf);
	if (ret > 0)
		*ppos += ret;
812

813
	return ret;
814 815
}

816 817
/**
 * generic_splice_sendpage - splice data from a pipe to a socket
818
 * @pipe:	pipe to splice from
819
 * @out:	socket to write to
820
 * @ppos:	position in @out
821 822 823
 * @len:	number of bytes to splice
 * @flags:	splice modifier flags
 *
824 825 826
 * Description:
 *    Will send @len bytes from the pipe to a network socket. No data copying
 *    is involved.
827 828
 *
 */
829
ssize_t generic_splice_sendpage(struct pipe_inode_info *pipe, struct file *out,
830
				loff_t *ppos, size_t len, unsigned int flags)
831
{
832
	return splice_from_pipe(pipe, out, ppos, len, flags, pipe_to_sendpage);
833 834
}

835
EXPORT_SYMBOL(generic_splice_sendpage);
Jeff Garzik's avatar
Jeff Garzik committed
836

837 838 839
/*
 * Attempt to initiate a splice from pipe to file.
 */
840
static long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
841
			   loff_t *ppos, size_t len, unsigned int flags)
842
{
843 844
	ssize_t (*splice_write)(struct pipe_inode_info *, struct file *,
				loff_t *, size_t, unsigned int);
845

Al Viro's avatar
Al Viro committed
846
	if (out->f_op->splice_write)
847 848
		splice_write = out->f_op->splice_write;
	else
849 850
		splice_write = default_file_splice_write;

851
	return splice_write(pipe, out, ppos, len, flags);
852 853
}

854 855 856
/*
 * Attempt to initiate a splice from a file to a pipe.
 */
857 858 859
static long do_splice_to(struct file *in, loff_t *ppos,
			 struct pipe_inode_info *pipe, size_t len,
			 unsigned int flags)
860
{
861 862
	ssize_t (*splice_read)(struct file *, loff_t *,
			       struct pipe_inode_info *, size_t, unsigned int);
863 864
	int ret;

865
	if (unlikely(!(in->f_mode & FMODE_READ)))
866 867
		return -EBADF;

868
	ret = rw_verify_area(READ, in, ppos, len);
869 870 871
	if (unlikely(ret < 0))
		return ret;

872 873 874
	if (unlikely(len > MAX_RW_COUNT))
		len = MAX_RW_COUNT;

Al Viro's avatar
Al Viro committed
875
	if (in->f_op->splice_read)
876 877
		splice_read = in->f_op->splice_read;
	else
878 879 880
		splice_read = default_file_splice_read;

	return splice_read(in, ppos, pipe, len, flags);
881 882
}

883 884 885 886 887 888 889 890 891
/**
 * splice_direct_to_actor - splices data directly between two non-pipes
 * @in:		file to splice from
 * @sd:		actor information on where to splice to
 * @actor:	handles the data splicing
 *
 * Description:
 *    This is a special case helper to splice directly between two
 *    points, without requiring an explicit pipe. Internally an allocated
892
 *    pipe is cached in the process, and reused during the lifetime of
893 894
 *    that process.
 *
895 896 897
 */
ssize_t splice_direct_to_actor(struct file *in, struct splice_desc *sd,
			       splice_direct_actor *actor)
898 899 900 901
{
	struct pipe_inode_info *pipe;
	long ret, bytes;
	umode_t i_mode;
902
	size_t len;
903
	int i, flags, more;
904 905 906 907 908 909

	/*
	 * We require the input being a regular file, as we don't want to
	 * randomly drop data for eg socket -> socket splicing. Use the
	 * piped splicing for that!
	 */
Al Viro's avatar
Al Viro committed
910
	i_mode = file_inode(in)->i_mode;
911 912 913 914 915 916 917 918
	if (unlikely(!S_ISREG(i_mode) && !S_ISBLK(i_mode)))
		return -EINVAL;

	/*
	 * neither in nor out is a pipe, setup an internal pipe attached to
	 * 'out' and transfer the wanted data from 'in' to 'out' through that
	 */
	pipe = current->splice_pipe;
919
	if (unlikely(!pipe)) {
920
		pipe = alloc_pipe_info();
921 922 923 924 925
		if (!pipe)
			return -ENOMEM;

		/*
		 * We don't have an immediate reader, but we'll read the stuff
926
		 * out of the pipe right after the splice_to_pipe(). So set
927 928 929 930 931 932 933 934
		 * PIPE_READERS appropriately.
		 */
		pipe->readers = 1;

		current->splice_pipe = pipe;
	}

	/*
935
	 * Do the splice.
936 937 938
	 */
	ret = 0;
	bytes = 0;
939 940 941 942 943 944 945
	len = sd->total_len;
	flags = sd->flags;

	/*
	 * Don't block on output, we have to drain the direct pipe.
	 */
	sd->flags &= ~SPLICE_F_NONBLOCK;
946
	more = sd->flags & SPLICE_F_MORE;
947

948 949
	WARN_ON_ONCE(pipe->nrbufs != 0);

950
	while (len) {
951
		size_t read_len;
952
		loff_t pos = sd->pos, prev_pos = pos;
953

954 955 956 957
		/* Don't try to read more the pipe has space for. */
		read_len = min_t(size_t, len,
				 (pipe->buffers - pipe->nrbufs) << PAGE_SHIFT);
		ret = do_splice_to(in, &pos, pipe, read_len, flags);
958
		if (unlikely(ret <= 0))
959 960 961
			goto out_release;

		read_len = ret;
962
		sd->total_len = read_len;
963

964 965 966 967 968 969 970 971 972
		/*
		 * If more data is pending, set SPLICE_F_MORE
		 * If this is the last data and SPLICE_F_MORE was not set
		 * initially, clears it.
		 */
		if (read_len < len)
			sd->flags |= SPLICE_F_MORE;
		else if (!more)
			sd->flags &= ~SPLICE_F_MORE;
973 974 975 976 977
		/*
		 * NOTE: nonblocking mode only applies to the input. We
		 * must not do the output in nonblocking mode as then we
		 * could get stuck data in the internal pipe:
		 */
978
		ret = actor(pipe, sd);
979 980
		if (unlikely(ret <= 0)) {
			sd->pos = prev_pos;
981
			goto out_release;
982
		}
983 984 985

		bytes += ret;
		len -= ret;
986
		sd->pos = pos;
987

988 989
		if (ret < read_len) {
			sd->pos = prev_pos + ret;
990
			goto out_release;
991
		}
992 993
	}

994
done:
995
	pipe->nrbufs = pipe->curbuf = 0;
996
	file_accessed(in);
997 998 999 1000 1001 1002 1003
	return bytes;

out_release:
	/*
	 * If we did an incomplete transfer we must release
	 * the pipe buffers in question:
	 */
1004
	for (i = 0; i < pipe->buffers; i++) {
1005 1006
		struct pipe_buffer *buf = pipe->bufs + i;

1007 1008
		if (buf->ops)
			pipe_buf_release(pipe, buf);
1009 1010
	}

1011 1012
	if (!bytes)
		bytes = ret;
1013

1014
	goto done;
1015 1016 1017 1018 1019 1020
}
EXPORT_SYMBOL(splice_direct_to_actor);

static int direct_splice_actor(struct pipe_inode_info *pipe,
			       struct splice_desc *sd)
{
1021
	struct file *file = sd->u.file;
1022

1023
	return do_splice_from(pipe, file, sd->opos, sd->total_len,
1024
			      sd->flags);
1025 1026
}

1027 1028 1029 1030 1031
/**
 * do_splice_direct - splices data directly between two files
 * @in:		file to splice from
 * @ppos:	input file offset
 * @out:	file to splice to
1032
 * @opos:	output file offset
1033 1034 1035 1036 1037 1038 1039 1040 1041 1042
 * @len:	number of bytes to splice
 * @flags:	splice modifier flags
 *
 * Description:
 *    For use by do_sendfile(). splice can easily emulate sendfile, but
 *    doing it in the application would incur an extra system call
 *    (splice in + splice out, as compared to just sendfile()). So this helper
 *    can splice directly through a process-private pipe.
 *
 */
1043
long do_splice_direct(struct file *in, loff_t *ppos, struct file *out,
1044
		      loff_t *opos, size_t len, unsigned int flags)
1045 1046 1047 1048 1049 1050
{
	struct splice_desc sd = {
		.len		= len,
		.total_len	= len,
		.flags		= flags,
		.pos		= *ppos,
1051
		.u.file		= out,
1052
		.opos		= opos,
1053
	};
1054
	long ret;
1055

1056 1057 1058 1059 1060 1061 1062 1063 1064 1065
	if (unlikely(!(out->f_mode & FMODE_WRITE)))
		return -EBADF;

	if (unlikely(out->f_flags & O_APPEND))
		return -EINVAL;

	ret = rw_verify_area(WRITE, out, opos, len);
	if (unlikely(ret < 0))
		return ret;

1066
	ret = splice_direct_to_actor(in, &sd, direct_splice_actor);
1067
	if (ret > 0)
1068
		*ppos = sd.pos;
1069

1070
	return ret;
1071
}
1072
EXPORT_SYMBOL(do_splice_direct);
1073

1074 1075
static int wait_for_space(struct pipe_inode_info *pipe, unsigned flags)
{
1076 1077 1078 1079 1080 1081 1082
	for (;;) {
		if (unlikely(!pipe->readers)) {
			send_sig(SIGPIPE, current, 0);
			return -EPIPE;
		}
		if (pipe->nrbufs != pipe->buffers)
			return 0;
1083 1084 1085 1086 1087 1088 1089 1090 1091 1092
		if (flags & SPLICE_F_NONBLOCK)
			return -EAGAIN;
		if (signal_pending(current))
			return -ERESTARTSYS;
		pipe->waiting_writers++;
		pipe_wait(pipe);
		pipe->waiting_writers--;
	}
}

1093 1094 1095
static int splice_pipe_to_pipe(struct pipe_inode_info *ipipe,
			       struct pipe_inode_info *opipe,
			       size_t len, unsigned int flags);
1096

1097 1098 1099
/*
 * Determine where to splice to/from.
 */
1100 1101 1102
static long do_splice(struct file *in, loff_t __user *off_in,
		      struct file *out, loff_t __user *off_out,
		      size_t len, unsigned int flags)
1103
{
1104 1105
	struct pipe_inode_info *ipipe;
	struct pipe_inode_info *opipe;
1106
	loff_t offset;
1107
	long ret;
1108

1109 1110
	ipipe = get_pipe_info(in);
	opipe = get_pipe_info(out);
1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129

	if (ipipe && opipe) {
		if (off_in || off_out)
			return -ESPIPE;

		if (!(in->f_mode & FMODE_READ))
			return -EBADF;

		if (!(out->f_mode & FMODE_WRITE))
			return -EBADF;

		/* Splicing to self would be fun, but... */
		if (ipipe == opipe)
			return -EINVAL;

		return splice_pipe_to_pipe(ipipe, opipe, len, flags);
	}

	if (ipipe) {
1130 1131
		if (off_in)
			return -ESPIPE;
1132
		if (off_out) {
1133
			if (!(out->f_mode & FMODE_PWRITE))
1134
				return -EINVAL;
1135
			if (copy_from_user(&offset, off_out, sizeof(loff_t)))
1136
				return -EFAULT;
1137 1138 1139
		} else {
			offset = out->f_pos;
		}
1140

1141 1142 1143 1144 1145 1146 1147 1148 1149 1150
		if (unlikely(!(out->f_mode & FMODE_WRITE)))
			return -EBADF;

		if (unlikely(out->f_flags & O_APPEND))
			return -EINVAL;

		ret = rw_verify_area(WRITE, out, &offset, len);
		if (unlikely(ret < 0))
			return ret;

1151
		file_start_write(out);
1152
		ret = do_splice_from(ipipe, out, &offset, len, flags);
1153
		file_end_write(out);
1154

1155 1156 1157
		if (!off_out)
			out->f_pos = offset;
		else if (copy_to_user(off_out, &offset, sizeof(loff_t)))
1158 1159 1160
			ret = -EFAULT;

		return ret;
1161
	}
1162

1163
	if (opipe) {
1164 1165
		if (off_out)
			return -ESPIPE;
1166
		if (off_in) {
1167
			if (!(in->f_mode & FMODE_PREAD))
1168
				return -EINVAL;
1169
			if (copy_from_user(&offset, off_in, sizeof(loff_t)))
1170
				return -EFAULT;
1171 1172 1173
		} else {
			offset = in->f_pos;
		}
1174

1175 1176 1177 1178 1179 1180 1181
		pipe_lock(opipe);
		ret = wait_for_space(opipe, flags);
		if (!ret)
			ret = do_splice_to(in, &offset, opipe, len, flags);
		pipe_unlock(opipe);
		if (ret > 0)
			wakeup_pipe_readers(opipe);
1182 1183 1184
		if (!off_in)
			in->f_pos = offset;
		else if (copy_to_user(off_in, &offset, sizeof(loff_t)))
1185 1186 1187
			ret = -EFAULT;

		return ret;
1188
	}
1189 1190 1191 1192

	return -EINVAL;
}

Al Viro's avatar
Al Viro committed
1193 1194 1195
static int iter_to_pipe(struct iov_iter *from,
			struct pipe_inode_info *pipe,
			unsigned flags)
1196
{
Al Viro's avatar
Al Viro committed
1197 1198 1199 1200 1201 1202 1203 1204 1205 1206
	struct pipe_buffer buf = {
		.ops = &user_page_pipe_buf_ops,
		.flags = flags
	};
	size_t total = 0;
	int ret = 0;
	bool failed = false;

	while (iov_iter_count(from) && !failed) {
		struct page *pages[16];
1207 1208
		ssize_t copied;
		size_t start;
Al Viro's avatar
Al Viro committed
1209
		int n;
1210

Al Viro's avatar
Al Viro committed
1211 1212 1213 1214 1215
		copied = iov_iter_get_pages(from, pages, ~0UL, 16, &start);
		if (copied <= 0) {
			ret = copied;
			break;
		}
1216

Al Viro's avatar
Al Viro committed
1217
		for (n = 0; copied; n++, start = 0) {
1218
			int size = min_t(int, copied, PAGE_SIZE - start);
Al Viro's avatar
Al Viro committed
1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232
			if (!failed) {
				buf.page = pages[n];
				buf.offset = start;
				buf.len = size;
				ret = add_to_pipe(pipe, &buf);
				if (unlikely(ret < 0)) {
					failed = true;
				} else {
					iov_iter_advance(from, ret);
					total += ret;
				}
			} else {
				put_page(pages[n]);
			}
1233
			copied -= size;
1234 1235
		}
	}
Al Viro's avatar
Al Viro committed
1236
	return total ? total : ret;
1237 1238
}

1239 1240 1241
static int pipe_to_user(struct pipe_inode_info *pipe, struct pipe_buffer *buf,
			struct splice_desc *sd)
{
1242 1243
	int n = copy_page_to_iter(buf->page, buf->offset, sd->len, sd->u.data);
	return n == sd->len ? n : -EFAULT;
1244 1245 1246 1247 1248 1249
}

/*
 * For lack of a better implementation, implement vmsplice() to userspace
 * as a simple copy of the pipes pages to the user iov.
 */
1250 1251
static long vmsplice_to_user(struct file *file, struct iov_iter *iter,
			     unsigned int flags)
1252
{
1253 1254 1255 1256 1257 1258 1259
	struct pipe_inode_info *pipe = get_pipe_info(file);
	struct splice_desc sd = {
		.total_len = iov_iter_count(iter),
		.flags = flags,
		.u.data = iter
	};
	long ret = 0;
1260 1261 1262 1263

	if (!pipe)
		return -EBADF;

1264 1265 1266 1267 1268
	if (sd.total_len) {
		pipe_lock(pipe);
		ret = __splice_from_pipe(pipe, &sd, pipe_to_user);
		pipe_unlock(pipe);
	}
1269 1270 1271 1272

	return ret;
}

1273 1274 1275 1276 1277
/*
 * vmsplice splices a user address range into a pipe. It can be thought of
 * as splice-from-memory, where the regular splice is splice-from-file (or
 * to file). In both cases the output is a pipe, naturally.
 */
1278 1279
static long vmsplice_to_pipe(struct file *file, struct iov_iter *iter,
			     unsigned int flags)
1280
{
1281
	struct pipe_inode_info *pipe;
1282
	long ret = 0;
Al Viro's avatar
Al Viro committed
1283 1284 1285 1286
	unsigned buf_flag = 0;

	if (flags & SPLICE_F_GIFT)
		buf_flag = PIPE_BUF_FLAG_GIFT;
1287

1288
	pipe = get_pipe_info(file);
1289
	if (!pipe)
1290 1291
		return -EBADF;

1292 1293
	pipe_lock(pipe);
	ret = wait_for_space(pipe, flags);
Al Viro's avatar
Al Viro committed
1294
	if (!ret)
1295
		ret = iter_to_pipe(iter, pipe, buf_flag);
1296 1297 1298
	pipe_unlock(pipe);
	if (ret > 0)
		wakeup_pipe_readers(pipe);
1299
	return ret;
1300 1301
}

1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316
static int vmsplice_type(struct fd f, int *type)
{
	if (!f.file)
		return -EBADF;
	if (f.file->f_mode & FMODE_WRITE) {
		*type = WRITE;
	} else if (f.file->f_mode & FMODE_READ) {
		*type = READ;
	} else {
		fdput(f);
		return -EBADF;
	}
	return 0;
}

1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332
/*
 * Note that vmsplice only really supports true splicing _from_ user memory
 * to a pipe, not the other way around. Splicing from user memory is a simple
 * operation that can be supported without any funky alignment restrictions
 * or nasty vm tricks. We simply map in the user memory and fill them into
 * a pipe. The reverse isn't quite as easy, though. There are two possible
 * solutions for that:
 *
 *	- memcpy() the data internally, at which point we might as well just
 *	  do a regular read() on the buffer anyway.
 *	- Lots of nasty vm tricks, that are neither fast nor flexible (it
 *	  has restriction limitations on both ends of the pipe).
 *
 * Currently we punt and implement it as a normal copy, see pipe_to_user().
 *
 */
1333
static long do_vmsplice(struct file *f, struct iov_iter *iter, unsigned int flags)
1334
{