splice.c 34.0 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 16 17
 * Copyright (C) 2005-2006 Jens Axboe <axboe@suse.de>
 * Copyright (C) 2005-2006 Linus Torvalds <torvalds@osdl.org>
 * Copyright (C) 2006 Ingo Molnar <mingo@elte.hu>
18 19 20 21 22 23 24
 *
 */
#include <linux/fs.h>
#include <linux/file.h>
#include <linux/pagemap.h>
#include <linux/pipe_fs_i.h>
#include <linux/mm_inline.h>
25
#include <linux/swap.h>
26 27
#include <linux/writeback.h>
#include <linux/buffer_head.h>
J
Jeff Garzik 已提交
28
#include <linux/module.h>
29
#include <linux/syscalls.h>
30
#include <linux/uio.h>
31

32 33 34 35 36 37
struct partial_page {
	unsigned int offset;
	unsigned int len;
};

/*
38
 * Passed to splice_to_pipe
39 40 41 42 43 44 45 46 47
 */
struct splice_pipe_desc {
	struct page **pages;		/* page map */
	struct partial_page *partial;	/* pages[] may not be contig */
	int nr_pages;			/* number of pages in map */
	unsigned int flags;		/* splice flags */
	struct pipe_buf_operations *ops;/* ops associated with output pipe */
};

48 49 50 51 52 53
/*
 * 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.
 */
54 55 56 57
static int page_cache_pipe_buf_steal(struct pipe_inode_info *info,
				     struct pipe_buffer *buf)
{
	struct page *page = buf->page;
58
	struct address_space *mapping = page_mapping(page);
59

60 61
	lock_page(page);

62 63
	WARN_ON(!PageUptodate(page));

64 65 66 67 68 69 70 71 72
	/*
	 * 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);

73 74 75
	if (PagePrivate(page))
		try_to_release_page(page, mapping_gfp_mask(mapping));

76 77
	if (!remove_mapping(mapping, page)) {
		unlock_page(page);
78
		return 1;
79
	}
80 81 82 83

	return 0;
}

84 85 86 87 88 89
static void page_cache_pipe_buf_release(struct pipe_inode_info *info,
					struct pipe_buffer *buf)
{
	page_cache_release(buf->page);
}

90 91
static int page_cache_pipe_buf_pin(struct pipe_inode_info *info,
				   struct pipe_buffer *buf)
92 93
{
	struct page *page = buf->page;
94
	int err;
95 96

	if (!PageUptodate(page)) {
97 98 99 100
		lock_page(page);

		/*
		 * Page got truncated/unhashed. This will cause a 0-byte
I
Ingo Molnar 已提交
101
		 * splice, if this is the first page.
102 103 104 105 106
		 */
		if (!page->mapping) {
			err = -ENODATA;
			goto error;
		}
107

108
		/*
I
Ingo Molnar 已提交
109
		 * Uh oh, read-error from disk.
110 111 112 113 114 115 116
		 */
		if (!PageUptodate(page)) {
			err = -EIO;
			goto error;
		}

		/*
117
		 * Page is ok afterall, we are done.
118
		 */
119 120 121
		unlock_page(page);
	}

122
	return 0;
123 124
error:
	unlock_page(page);
125
	return err;
126 127
}

128 129
static struct pipe_buf_operations page_cache_pipe_buf_ops = {
	.can_merge = 0,
130 131 132
	.map = generic_pipe_buf_map,
	.unmap = generic_pipe_buf_unmap,
	.pin = page_cache_pipe_buf_pin,
133
	.release = page_cache_pipe_buf_release,
134
	.steal = page_cache_pipe_buf_steal,
135
	.get = generic_pipe_buf_get,
136 137
};

138 139 140
static int user_page_pipe_buf_steal(struct pipe_inode_info *pipe,
				    struct pipe_buffer *buf)
{
141 142 143
	if (!(buf->flags & PIPE_BUF_FLAG_GIFT))
		return 1;

144
	return generic_pipe_buf_steal(pipe, buf);
145 146 147 148
}

static struct pipe_buf_operations user_page_pipe_buf_ops = {
	.can_merge = 0,
149 150 151
	.map = generic_pipe_buf_map,
	.unmap = generic_pipe_buf_unmap,
	.pin = generic_pipe_buf_pin,
152 153
	.release = page_cache_pipe_buf_release,
	.steal = user_page_pipe_buf_steal,
154
	.get = generic_pipe_buf_get,
155 156
};

157 158 159 160
/*
 * Pipe output worker. This sets up our pipe format with the page cache
 * pipe buffer operations. Otherwise very similar to the regular pipe_writev().
 */
161 162
static ssize_t splice_to_pipe(struct pipe_inode_info *pipe,
			      struct splice_pipe_desc *spd)
163
{
164
	int ret, do_wakeup, page_nr;
165 166 167

	ret = 0;
	do_wakeup = 0;
168
	page_nr = 0;
169

170 171
	if (pipe->inode)
		mutex_lock(&pipe->inode->i_mutex);
172 173

	for (;;) {
174
		if (!pipe->readers) {
175 176 177 178 179 180
			send_sig(SIGPIPE, current, 0);
			if (!ret)
				ret = -EPIPE;
			break;
		}

181 182
		if (pipe->nrbufs < PIPE_BUFFERS) {
			int newbuf = (pipe->curbuf + pipe->nrbufs) & (PIPE_BUFFERS - 1);
183
			struct pipe_buffer *buf = pipe->bufs + newbuf;
184

185 186 187 188
			buf->page = spd->pages[page_nr];
			buf->offset = spd->partial[page_nr].offset;
			buf->len = spd->partial[page_nr].len;
			buf->ops = spd->ops;
189 190 191
			if (spd->flags & SPLICE_F_GIFT)
				buf->flags |= PIPE_BUF_FLAG_GIFT;

192
			pipe->nrbufs++;
193 194 195
			page_nr++;
			ret += buf->len;

196 197
			if (pipe->inode)
				do_wakeup = 1;
198

199
			if (!--spd->nr_pages)
200
				break;
201
			if (pipe->nrbufs < PIPE_BUFFERS)
202 203 204 205 206
				continue;

			break;
		}

207
		if (spd->flags & SPLICE_F_NONBLOCK) {
208 209 210 211 212
			if (!ret)
				ret = -EAGAIN;
			break;
		}

213 214 215 216 217 218 219
		if (signal_pending(current)) {
			if (!ret)
				ret = -ERESTARTSYS;
			break;
		}

		if (do_wakeup) {
220
			smp_mb();
221 222 223
			if (waitqueue_active(&pipe->wait))
				wake_up_interruptible_sync(&pipe->wait);
			kill_fasync(&pipe->fasync_readers, SIGIO, POLL_IN);
224 225 226
			do_wakeup = 0;
		}

227 228 229
		pipe->waiting_writers++;
		pipe_wait(pipe);
		pipe->waiting_writers--;
230 231
	}

232 233
	if (pipe->inode)
		mutex_unlock(&pipe->inode->i_mutex);
234 235

	if (do_wakeup) {
236
		smp_mb();
237 238 239
		if (waitqueue_active(&pipe->wait))
			wake_up_interruptible(&pipe->wait);
		kill_fasync(&pipe->fasync_readers, SIGIO, POLL_IN);
240 241
	}

242 243
	while (page_nr < spd->nr_pages)
		page_cache_release(spd->pages[page_nr++]);
244 245 246 247

	return ret;
}

248
static int
249 250 251
__generic_file_splice_read(struct file *in, loff_t *ppos,
			   struct pipe_inode_info *pipe, size_t len,
			   unsigned int flags)
252 253
{
	struct address_space *mapping = in->f_mapping;
254
	unsigned int loff, nr_pages;
255
	struct page *pages[PIPE_BUFFERS];
256
	struct partial_page partial[PIPE_BUFFERS];
257
	struct page *page;
258 259
	pgoff_t index, end_index;
	loff_t isize;
260
	size_t total_len;
261
	int error, page_nr;
262 263 264 265 266 267
	struct splice_pipe_desc spd = {
		.pages = pages,
		.partial = partial,
		.flags = flags,
		.ops = &page_cache_pipe_buf_ops,
	};
268

269
	index = *ppos >> PAGE_CACHE_SHIFT;
270 271
	loff = *ppos & ~PAGE_CACHE_MASK;
	nr_pages = (len + loff + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
272 273 274 275 276

	if (nr_pages > PIPE_BUFFERS)
		nr_pages = PIPE_BUFFERS;

	/*
I
Ingo Molnar 已提交
277
	 * Initiate read-ahead on this page range. however, don't call into
278 279
	 * read-ahead if this is a non-zero offset (we are likely doing small
	 * chunk splice and the page is already there) for a single page.
280
	 */
281 282
	if (!loff || nr_pages > 1)
		page_cache_readahead(mapping, &in->f_ra, in, index, nr_pages);
283 284

	/*
I
Ingo Molnar 已提交
285
	 * Now fill in the holes:
286
	 */
287
	error = 0;
288
	total_len = 0;
289

290 291 292 293
	/*
	 * Lookup the (hopefully) full range of pages we need.
	 */
	spd.nr_pages = find_get_pages_contig(mapping, index, nr_pages, pages);
294

295 296 297 298 299 300
	/*
	 * If find_get_pages_contig() returned fewer pages than we needed,
	 * allocate the rest.
	 */
	index += spd.nr_pages;
	while (spd.nr_pages < nr_pages) {
301
		/*
302 303
		 * Page could be there, find_get_pages_contig() breaks on
		 * the first hole.
304
		 */
305 306
		page = find_get_page(mapping, index);
		if (!page) {
307 308 309 310 311 312
			/*
			 * Make sure the read-ahead engine is notified
			 * about this failure.
			 */
			handle_ra_miss(mapping, &in->f_ra, index);

313
			/*
314
			 * page didn't exist, allocate one.
315 316 317 318 319 320
			 */
			page = page_cache_alloc_cold(mapping);
			if (!page)
				break;

			error = add_to_page_cache_lru(page, mapping, index,
321
					      mapping_gfp_mask(mapping));
322 323 324 325
			if (unlikely(error)) {
				page_cache_release(page);
				break;
			}
326 327 328 329 330
			/*
			 * add_to_page_cache() locks the page, unlock it
			 * to avoid convoluting the logic below even more.
			 */
			unlock_page(page);
331 332
		}

333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355
		pages[spd.nr_pages++] = page;
		index++;
	}

	/*
	 * Now loop over the map and see if we need to start IO on any
	 * pages, fill in the partial map, etc.
	 */
	index = *ppos >> PAGE_CACHE_SHIFT;
	nr_pages = spd.nr_pages;
	spd.nr_pages = 0;
	for (page_nr = 0; page_nr < nr_pages; page_nr++) {
		unsigned int this_len;

		if (!len)
			break;

		/*
		 * this_len is the max we'll use from this page
		 */
		this_len = min_t(unsigned long, len, PAGE_CACHE_SIZE - loff);
		page = pages[page_nr];

356 357 358 359
		/*
		 * If the page isn't uptodate, we may need to start io on it
		 */
		if (!PageUptodate(page)) {
360 361 362 363 364 365 366
			/*
			 * If in nonblock mode then dont block on waiting
			 * for an in-flight io page
			 */
			if (flags & SPLICE_F_NONBLOCK)
				break;

367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384
			lock_page(page);

			/*
			 * page was truncated, stop here. if this isn't the
			 * first page, we'll just complete what we already
			 * added
			 */
			if (!page->mapping) {
				unlock_page(page);
				break;
			}
			/*
			 * page was already under io and is now done, great
			 */
			if (PageUptodate(page)) {
				unlock_page(page);
				goto fill_it;
			}
385

386 387 388 389
			/*
			 * need to read in the page
			 */
			error = mapping->a_ops->readpage(in, page);
390
			if (unlikely(error)) {
391 392 393 394 395 396
				/*
				 * We really should re-lookup the page here,
				 * but it complicates things a lot. Instead
				 * lets just do what we already stored, and
				 * we'll get it the next time we are called.
				 */
397
				if (error == AOP_TRUNCATED_PAGE)
398 399
					error = 0;

400 401
				break;
			}
402 403 404 405 406 407

			/*
			 * i_size must be checked after ->readpage().
			 */
			isize = i_size_read(mapping->host);
			end_index = (isize - 1) >> PAGE_CACHE_SHIFT;
408
			if (unlikely(!isize || index > end_index))
409 410 411 412 413 414 415 416
				break;

			/*
			 * if this is the last page, see if we need to shrink
			 * the length and stop
			 */
			if (end_index == index) {
				loff = PAGE_CACHE_SIZE - (isize & ~PAGE_CACHE_MASK);
417
				if (total_len + loff > isize)
418 419 420 421
					break;
				/*
				 * force quit after adding this page
				 */
422
				len = this_len;
423
				this_len = min(this_len, loff);
424
				loff = 0;
425
			}
426
		}
427
fill_it:
428 429
		partial[page_nr].offset = loff;
		partial[page_nr].len = this_len;
430
		len -= this_len;
431
		total_len += this_len;
432
		loff = 0;
433 434
		spd.nr_pages++;
		index++;
435 436
	}

437 438 439 440 441 442 443
	/*
	 * Release any pages at the end, if we quit early. 'i' is how far
	 * we got, 'nr_pages' is how many pages are in the map.
	 */
	while (page_nr < nr_pages)
		page_cache_release(pages[page_nr++]);

444
	if (spd.nr_pages)
445
		return splice_to_pipe(pipe, &spd);
446

447
	return error;
448 449
}

450 451 452 453 454 455 456 457 458
/**
 * generic_file_splice_read - splice data from file to a pipe
 * @in:		file to splice from
 * @pipe:	pipe to splice to
 * @len:	number of bytes to splice
 * @flags:	splice modifier flags
 *
 * Will read pages from given file and fill them into a pipe.
 */
459 460 461
ssize_t generic_file_splice_read(struct file *in, loff_t *ppos,
				 struct pipe_inode_info *pipe, size_t len,
				 unsigned int flags)
462 463 464 465 466 467
{
	ssize_t spliced;
	int ret;

	ret = 0;
	spliced = 0;
468

469
	while (len) {
470
		ret = __generic_file_splice_read(in, ppos, pipe, len, flags);
471

472
		if (ret < 0)
473
			break;
474 475 476 477 478 479 480 481
		else if (!ret) {
			if (spliced)
				break;
			if (flags & SPLICE_F_NONBLOCK) {
				ret = -EAGAIN;
				break;
			}
		}
482

483
		*ppos += ret;
484 485 486 487 488 489 490 491 492 493
		len -= ret;
		spliced += ret;
	}

	if (spliced)
		return spliced;

	return ret;
}

494 495
EXPORT_SYMBOL(generic_file_splice_read);

496
/*
497
 * Send 'sd->len' bytes to socket from 'sd->file' at position 'sd->pos'
J
Jens Axboe 已提交
498
 * using sendpage(). Return the number of bytes sent.
499 500 501 502 503 504
 */
static int pipe_to_sendpage(struct pipe_inode_info *info,
			    struct pipe_buffer *buf, struct splice_desc *sd)
{
	struct file *file = sd->file;
	loff_t pos = sd->pos;
505
	int ret, more;
506

507 508 509
	ret = buf->ops->pin(info, buf);
	if (!ret) {
		more = (sd->flags & SPLICE_F_MORE) || sd->len < sd->total_len;
510

511 512 513
		ret = file->f_op->sendpage(file, buf->page, buf->offset,
					   sd->len, &pos, more);
	}
514

J
Jens Axboe 已提交
515
	return ret;
516 517 518 519 520 521 522 523 524 525 526 527 528 529 530
}

/*
 * This is a little more tricky than the file -> pipe splicing. There are
 * basically three cases:
 *
 *	- Destination page already exists in the address space and there
 *	  are users of it. For that case we have no other option that
 *	  copying the data. Tough luck.
 *	- Destination page already exists in the address space, but there
 *	  are no users of it. Make sure it's uptodate, then drop it. Fall
 *	  through to last case.
 *	- Destination page does not exist, we can add the pipe page to
 *	  the page cache and avoid the copy.
 *
531 532 533 534 535 536
 * If asked to move pages to the output file (SPLICE_F_MOVE is set in
 * sd->flags), we attempt to migrate pages from the pipe to the output
 * file address space page cache. This is possible if no one else has
 * the pipe page referenced outside of the pipe and page cache. If
 * SPLICE_F_MOVE isn't set, or we cannot move the page, we simply create
 * a new page in the output file page cache and fill/dirty that.
537 538 539 540 541 542
 */
static int pipe_to_file(struct pipe_inode_info *info, struct pipe_buffer *buf,
			struct splice_desc *sd)
{
	struct file *file = sd->file;
	struct address_space *mapping = file->f_mapping;
543
	gfp_t gfp_mask = mapping_gfp_mask(mapping);
J
Jens Axboe 已提交
544
	unsigned int offset, this_len;
545 546
	struct page *page;
	pgoff_t index;
547
	int ret;
548 549

	/*
550
	 * make sure the data in this buffer is uptodate
551
	 */
552 553 554
	ret = buf->ops->pin(info, buf);
	if (unlikely(ret))
		return ret;
555 556 557 558

	index = sd->pos >> PAGE_CACHE_SHIFT;
	offset = sd->pos & ~PAGE_CACHE_MASK;

J
Jens Axboe 已提交
559 560 561 562
	this_len = sd->len;
	if (this_len + offset > PAGE_CACHE_SIZE)
		this_len = PAGE_CACHE_SIZE - offset;

563
	/*
564 565
	 * Reuse buf page, if SPLICE_F_MOVE is set and we are doing a full
	 * page.
566
	 */
567
	if ((sd->flags & SPLICE_F_MOVE) && this_len == PAGE_CACHE_SIZE) {
568 569
		/*
		 * If steal succeeds, buf->page is now pruned from the vm
570 571
		 * side (page cache) and we can reuse it. The page will also
		 * be locked on successful return.
572
		 */
573 574 575 576
		if (buf->ops->steal(info, buf))
			goto find_page;

		page = buf->page;
577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594
		page_cache_get(page);

		/*
		 * page must be on the LRU for adding to the pagecache.
		 * Check this without grabbing the zone lock, if it isn't
		 * the do grab the zone lock, recheck, and add if necessary.
		 */
		if (!PageLRU(page)) {
			struct zone *zone = page_zone(page);

			spin_lock_irq(&zone->lru_lock);
			if (!PageLRU(page)) {
				SetPageLRU(page);
				add_page_to_inactive_list(zone, page);
			}
			spin_unlock_irq(&zone->lru_lock);
		}

595
		if (add_to_page_cache(page, mapping, index, gfp_mask)) {
596
			page_cache_release(page);
597
			unlock_page(page);
598
			goto find_page;
599
		}
600 601
	} else {
find_page:
602 603 604 605 606 607 608 609 610 611 612 613 614 615 616
		page = find_lock_page(mapping, index);
		if (!page) {
			ret = -ENOMEM;
			page = page_cache_alloc_cold(mapping);
			if (unlikely(!page))
				goto out_nomem;

			/*
			 * This will also lock the page
			 */
			ret = add_to_page_cache_lru(page, mapping, index,
						    gfp_mask);
			if (unlikely(ret))
				goto out;
		}
617 618

		/*
619 620 621 622
		 * We get here with the page locked. If the page is also
		 * uptodate, we don't need to do more. If it isn't, we
		 * may need to bring it in if we are not going to overwrite
		 * the full page.
623 624
		 */
		if (!PageUptodate(page)) {
J
Jens Axboe 已提交
625
			if (this_len < PAGE_CACHE_SIZE) {
626 627 628 629 630 631 632 633
				ret = mapping->a_ops->readpage(file, page);
				if (unlikely(ret))
					goto out;

				lock_page(page);

				if (!PageUptodate(page)) {
					/*
I
Ingo Molnar 已提交
634
					 * Page got invalidated, repeat.
635 636 637 638 639 640 641 642
					 */
					if (!page->mapping) {
						unlock_page(page);
						page_cache_release(page);
						goto find_page;
					}
					ret = -EIO;
					goto out;
643
				}
644
			} else
645
				SetPageUptodate(page);
646 647 648
		}
	}

J
Jens Axboe 已提交
649
	ret = mapping->a_ops->prepare_write(file, page, offset, offset+this_len);
650 651 652 653 654
	if (unlikely(ret)) {
		loff_t isize = i_size_read(mapping->host);

		if (ret != AOP_TRUNCATED_PAGE)
			unlock_page(page);
655
		page_cache_release(page);
656 657 658 659 660 661 662 663 664 665
		if (ret == AOP_TRUNCATED_PAGE)
			goto find_page;

		/*
		 * prepare_write() may have instantiated a few blocks
		 * outside i_size.  Trim these off again.
		 */
		if (sd->pos + this_len > isize)
			vmtruncate(mapping->host, isize);

666
		goto out;
667
	}
668

669
	if (buf->page != page) {
670 671 672
		/*
		 * Careful, ->map() uses KM_USER0!
		 */
673
		char *src = buf->ops->map(info, buf, 1);
674
		char *dst = kmap_atomic(page, KM_USER1);
675

J
Jens Axboe 已提交
676
		memcpy(dst + offset, src + buf->offset, this_len);
677
		flush_dcache_page(page);
678
		kunmap_atomic(dst, KM_USER1);
679
		buf->ops->unmap(info, buf, src);
680
	}
681

J
Jens Axboe 已提交
682
	ret = mapping->a_ops->commit_write(file, page, offset, offset+this_len);
683 684 685 686 687 688 689 690 691
	if (!ret) {
		/*
		 * Return the number of bytes written and mark page as
		 * accessed, we are now done!
		 */
		ret = this_len;
		mark_page_accessed(page);
		balance_dirty_pages_ratelimited(mapping);
	} else if (ret == AOP_TRUNCATED_PAGE) {
692 693
		page_cache_release(page);
		goto find_page;
694
	}
695
out:
696
	page_cache_release(page);
697
	unlock_page(page);
698
out_nomem:
699 700 701
	return ret;
}

702 703 704 705 706
/*
 * Pipe input worker. Most of this logic works like a regular pipe, the
 * key here is the 'actor' worker passed in that actually moves the data
 * to the wanted destination. See pipe_to_file/pipe_to_sendpage above.
 */
707 708 709
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)
710 711 712 713 714 715 716 717 718 719
{
	int ret, do_wakeup, err;
	struct splice_desc sd;

	ret = 0;
	do_wakeup = 0;

	sd.total_len = len;
	sd.flags = flags;
	sd.file = out;
720
	sd.pos = *ppos;
721

722 723
	if (pipe->inode)
		mutex_lock(&pipe->inode->i_mutex);
724 725

	for (;;) {
726 727
		if (pipe->nrbufs) {
			struct pipe_buffer *buf = pipe->bufs + pipe->curbuf;
728 729 730 731 732 733
			struct pipe_buf_operations *ops = buf->ops;

			sd.len = buf->len;
			if (sd.len > sd.total_len)
				sd.len = sd.total_len;

734
			err = actor(pipe, buf, &sd);
J
Jens Axboe 已提交
735
			if (err <= 0) {
736 737 738 739 740 741
				if (!ret && err != -ENODATA)
					ret = err;

				break;
			}

J
Jens Axboe 已提交
742 743 744 745 746 747 748 749 750
			ret += err;
			buf->offset += err;
			buf->len -= err;

			sd.len -= err;
			sd.pos += err;
			sd.total_len -= err;
			if (sd.len)
				continue;
I
Ingo Molnar 已提交
751

752 753
			if (!buf->len) {
				buf->ops = NULL;
754
				ops->release(pipe, buf);
755 756 757 758
				pipe->curbuf = (pipe->curbuf + 1) & (PIPE_BUFFERS - 1);
				pipe->nrbufs--;
				if (pipe->inode)
					do_wakeup = 1;
759 760 761 762 763 764
			}

			if (!sd.total_len)
				break;
		}

765
		if (pipe->nrbufs)
766
			continue;
767
		if (!pipe->writers)
768
			break;
769
		if (!pipe->waiting_writers) {
770 771 772 773
			if (ret)
				break;
		}

774 775 776 777 778 779
		if (flags & SPLICE_F_NONBLOCK) {
			if (!ret)
				ret = -EAGAIN;
			break;
		}

780 781 782 783 784 785 786
		if (signal_pending(current)) {
			if (!ret)
				ret = -ERESTARTSYS;
			break;
		}

		if (do_wakeup) {
787
			smp_mb();
788 789 790
			if (waitqueue_active(&pipe->wait))
				wake_up_interruptible_sync(&pipe->wait);
			kill_fasync(&pipe->fasync_writers, SIGIO, POLL_OUT);
791 792 793
			do_wakeup = 0;
		}

794
		pipe_wait(pipe);
795 796
	}

797 798
	if (pipe->inode)
		mutex_unlock(&pipe->inode->i_mutex);
799 800

	if (do_wakeup) {
801
		smp_mb();
802 803 804
		if (waitqueue_active(&pipe->wait))
			wake_up_interruptible(&pipe->wait);
		kill_fasync(&pipe->fasync_writers, SIGIO, POLL_OUT);
805 806 807 808 809
	}

	return ret;
}

810 811
/**
 * generic_file_splice_write - splice data from a pipe to a file
812
 * @pipe:	pipe info
813 814 815 816 817 818 819 820
 * @out:	file to write to
 * @len:	number of bytes to splice
 * @flags:	splice modifier flags
 *
 * Will either move or copy pages (determined by @flags options) from
 * the given pipe inode to the given file.
 *
 */
821 822
ssize_t
generic_file_splice_write(struct pipe_inode_info *pipe, struct file *out,
823
			  loff_t *ppos, size_t len, unsigned int flags)
824
{
825
	struct address_space *mapping = out->f_mapping;
826 827
	ssize_t ret;

828
	ret = splice_from_pipe(pipe, out, ppos, len, flags, pipe_to_file);
J
Jens Axboe 已提交
829
	if (ret > 0) {
830 831
		struct inode *inode = mapping->host;

J
Jens Axboe 已提交
832 833 834 835 836 837 838 839 840 841 842 843 844
		*ppos += ret;

		/*
		 * If file or inode is SYNC and we actually wrote some data,
		 * sync it.
		 */
		if (unlikely((out->f_flags & O_SYNC) || IS_SYNC(inode))) {
			int err;

			mutex_lock(&inode->i_mutex);
			err = generic_osync_inode(inode, mapping,
						  OSYNC_METADATA|OSYNC_DATA);
			mutex_unlock(&inode->i_mutex);
845

J
Jens Axboe 已提交
846 847 848
			if (err)
				ret = err;
		}
849 850 851
	}

	return ret;
852 853
}

854 855
EXPORT_SYMBOL(generic_file_splice_write);

856 857 858 859 860 861 862 863 864 865 866
/**
 * generic_splice_sendpage - splice data from a pipe to a socket
 * @inode:	pipe inode
 * @out:	socket to write to
 * @len:	number of bytes to splice
 * @flags:	splice modifier flags
 *
 * Will send @len bytes from the pipe to a network socket. No data copying
 * is involved.
 *
 */
867
ssize_t generic_splice_sendpage(struct pipe_inode_info *pipe, struct file *out,
868
				loff_t *ppos, size_t len, unsigned int flags)
869
{
870
	return splice_from_pipe(pipe, out, ppos, len, flags, pipe_to_sendpage);
871 872
}

873
EXPORT_SYMBOL(generic_splice_sendpage);
J
Jeff Garzik 已提交
874

875 876 877
/*
 * Attempt to initiate a splice from pipe to file.
 */
878
static long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
879
			   loff_t *ppos, size_t len, unsigned int flags)
880 881 882
{
	int ret;

883
	if (unlikely(!out->f_op || !out->f_op->splice_write))
884 885
		return -EINVAL;

886
	if (unlikely(!(out->f_mode & FMODE_WRITE)))
887 888
		return -EBADF;

889
	ret = rw_verify_area(WRITE, out, ppos, len);
890 891 892
	if (unlikely(ret < 0))
		return ret;

893
	return out->f_op->splice_write(pipe, out, ppos, len, flags);
894 895
}

896 897 898
/*
 * Attempt to initiate a splice from a file to a pipe.
 */
899 900 901
static long do_splice_to(struct file *in, loff_t *ppos,
			 struct pipe_inode_info *pipe, size_t len,
			 unsigned int flags)
902
{
903
	loff_t isize, left;
904 905
	int ret;

906
	if (unlikely(!in->f_op || !in->f_op->splice_read))
907 908
		return -EINVAL;

909
	if (unlikely(!(in->f_mode & FMODE_READ)))
910 911
		return -EBADF;

912
	ret = rw_verify_area(READ, in, ppos, len);
913 914 915 916
	if (unlikely(ret < 0))
		return ret;

	isize = i_size_read(in->f_mapping->host);
917
	if (unlikely(*ppos >= isize))
918 919
		return 0;
	
920
	left = isize - *ppos;
921
	if (unlikely(left < len))
922 923
		len = left;

924
	return in->f_op->splice_read(in, ppos, pipe, len, flags);
925 926
}

927 928
long do_splice_direct(struct file *in, loff_t *ppos, struct file *out,
		      size_t len, unsigned int flags)
929 930 931
{
	struct pipe_inode_info *pipe;
	long ret, bytes;
932
	loff_t out_off;
933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949
	umode_t i_mode;
	int i;

	/*
	 * 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!
	 */
	i_mode = in->f_dentry->d_inode->i_mode;
	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;
950
	if (unlikely(!pipe)) {
951 952 953 954 955 956
		pipe = alloc_pipe_info(NULL);
		if (!pipe)
			return -ENOMEM;

		/*
		 * We don't have an immediate reader, but we'll read the stuff
957
		 * out of the pipe right after the splice_to_pipe(). So set
958 959 960 961 962 963 964 965
		 * PIPE_READERS appropriately.
		 */
		pipe->readers = 1;

		current->splice_pipe = pipe;
	}

	/*
I
Ingo Molnar 已提交
966
	 * Do the splice.
967 968 969
	 */
	ret = 0;
	bytes = 0;
970
	out_off = 0;
971 972 973 974 975 976 977 978 979

	while (len) {
		size_t read_len, max_read_len;

		/*
		 * Do at most PIPE_BUFFERS pages worth of transfer:
		 */
		max_read_len = min(len, (size_t)(PIPE_BUFFERS*PAGE_SIZE));

980
		ret = do_splice_to(in, ppos, pipe, max_read_len, flags);
981 982 983 984 985 986 987 988 989 990
		if (unlikely(ret < 0))
			goto out_release;

		read_len = ret;

		/*
		 * 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:
		 */
991
		ret = do_splice_from(pipe, out, &out_off, read_len,
992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040
				     flags & ~SPLICE_F_NONBLOCK);
		if (unlikely(ret < 0))
			goto out_release;

		bytes += ret;
		len -= ret;

		/*
		 * In nonblocking mode, if we got back a short read then
		 * that was due to either an IO error or due to the
		 * pagecache entry not being there. In the IO error case
		 * the _next_ splice attempt will produce a clean IO error
		 * return value (not a short read), so in both cases it's
		 * correct to break out of the loop here:
		 */
		if ((flags & SPLICE_F_NONBLOCK) && (read_len < max_read_len))
			break;
	}

	pipe->nrbufs = pipe->curbuf = 0;

	return bytes;

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

		if (buf->ops) {
			buf->ops->release(pipe, buf);
			buf->ops = NULL;
		}
	}
	pipe->nrbufs = pipe->curbuf = 0;

	/*
	 * If we transferred some data, return the number of bytes:
	 */
	if (bytes > 0)
		return bytes;

	return ret;
}

EXPORT_SYMBOL(do_splice_direct);

1041 1042 1043
/*
 * Determine where to splice to/from.
 */
1044 1045 1046
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)
1047
{
1048
	struct pipe_inode_info *pipe;
1049
	loff_t offset, *off;
J
Jens Axboe 已提交
1050
	long ret;
1051

1052
	pipe = in->f_dentry->d_inode->i_pipe;
1053 1054 1055
	if (pipe) {
		if (off_in)
			return -ESPIPE;
1056 1057 1058
		if (off_out) {
			if (out->f_op->llseek == no_llseek)
				return -EINVAL;
1059
			if (copy_from_user(&offset, off_out, sizeof(loff_t)))
1060
				return -EFAULT;
1061 1062 1063
			off = &offset;
		} else
			off = &out->f_pos;
1064

J
Jens Axboe 已提交
1065 1066 1067 1068 1069 1070
		ret = do_splice_from(pipe, out, off, len, flags);

		if (off_out && copy_to_user(off_out, off, sizeof(loff_t)))
			ret = -EFAULT;

		return ret;
1071
	}
1072

1073
	pipe = out->f_dentry->d_inode->i_pipe;
1074 1075 1076
	if (pipe) {
		if (off_out)
			return -ESPIPE;
1077 1078 1079
		if (off_in) {
			if (in->f_op->llseek == no_llseek)
				return -EINVAL;
1080
			if (copy_from_user(&offset, off_in, sizeof(loff_t)))
1081
				return -EFAULT;
1082 1083 1084
			off = &offset;
		} else
			off = &in->f_pos;
1085

J
Jens Axboe 已提交
1086 1087 1088 1089 1090 1091
		ret = do_splice_to(in, off, pipe, len, flags);

		if (off_in && copy_to_user(off_in, off, sizeof(loff_t)))
			ret = -EFAULT;

		return ret;
1092
	}
1093 1094 1095 1096

	return -EINVAL;
}

1097 1098 1099 1100 1101 1102 1103 1104 1105
/*
 * Map an iov into an array of pages and offset/length tupples. With the
 * partial_page structure, we can map several non-contiguous ranges into
 * our ones pages[] map instead of splitting that operation into pieces.
 * Could easily be exported as a generic helper for other users, in which
 * case one would probably want to add a 'max_nr_pages' parameter as well.
 */
static int get_iovec_page_array(const struct iovec __user *iov,
				unsigned int nr_vecs, struct page **pages,
1106
				struct partial_page *partial, int aligned)
1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145
{
	int buffers = 0, error = 0;

	/*
	 * It's ok to take the mmap_sem for reading, even
	 * across a "get_user()".
	 */
	down_read(&current->mm->mmap_sem);

	while (nr_vecs) {
		unsigned long off, npages;
		void __user *base;
		size_t len;
		int i;

		/*
		 * Get user address base and length for this iovec.
		 */
		error = get_user(base, &iov->iov_base);
		if (unlikely(error))
			break;
		error = get_user(len, &iov->iov_len);
		if (unlikely(error))
			break;

		/*
		 * Sanity check this iovec. 0 read succeeds.
		 */
		if (unlikely(!len))
			break;
		error = -EFAULT;
		if (unlikely(!base))
			break;

		/*
		 * Get this base offset and number of pages, then map
		 * in the user pages.
		 */
		off = (unsigned long) base & ~PAGE_MASK;
1146 1147 1148 1149 1150 1151 1152 1153 1154

		/*
		 * If asked for alignment, the offset must be zero and the
		 * length a multiple of the PAGE_SIZE.
		 */
		error = -EINVAL;
		if (aligned && (off || len & ~PAGE_MASK))
			break;

1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169
		npages = (off + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
		if (npages > PIPE_BUFFERS - buffers)
			npages = PIPE_BUFFERS - buffers;

		error = get_user_pages(current, current->mm,
				       (unsigned long) base, npages, 0, 0,
				       &pages[buffers], NULL);

		if (unlikely(error <= 0))
			break;

		/*
		 * Fill this contiguous range into the partial page map.
		 */
		for (i = 0; i < error; i++) {
1170
			const int plen = min_t(size_t, len, PAGE_SIZE - off);
1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247

			partial[buffers].offset = off;
			partial[buffers].len = plen;

			off = 0;
			len -= plen;
			buffers++;
		}

		/*
		 * We didn't complete this iov, stop here since it probably
		 * means we have to move some of this into a pipe to
		 * be able to continue.
		 */
		if (len)
			break;

		/*
		 * Don't continue if we mapped fewer pages than we asked for,
		 * or if we mapped the max number of pages that we have
		 * room for.
		 */
		if (error < npages || buffers == PIPE_BUFFERS)
			break;

		nr_vecs--;
		iov++;
	}

	up_read(&current->mm->mmap_sem);

	if (buffers)
		return buffers;

	return error;
}

/*
 * 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.
 *
 * Note that vmsplice only supports 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).
 *
 * Alas, it isn't here.
 *
 */
static long do_vmsplice(struct file *file, const struct iovec __user *iov,
			unsigned long nr_segs, unsigned int flags)
{
	struct pipe_inode_info *pipe = file->f_dentry->d_inode->i_pipe;
	struct page *pages[PIPE_BUFFERS];
	struct partial_page partial[PIPE_BUFFERS];
	struct splice_pipe_desc spd = {
		.pages = pages,
		.partial = partial,
		.flags = flags,
		.ops = &user_page_pipe_buf_ops,
	};

	if (unlikely(!pipe))
		return -EBADF;
	if (unlikely(nr_segs > UIO_MAXIOV))
		return -EINVAL;
	else if (unlikely(!nr_segs))
		return 0;

1248 1249
	spd.nr_pages = get_iovec_page_array(iov, nr_segs, pages, partial,
					    flags & SPLICE_F_GIFT);
1250 1251 1252
	if (spd.nr_pages <= 0)
		return spd.nr_pages;

1253
	return splice_to_pipe(pipe, &spd);
1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274
}

asmlinkage long sys_vmsplice(int fd, const struct iovec __user *iov,
			     unsigned long nr_segs, unsigned int flags)
{
	struct file *file;
	long error;
	int fput;

	error = -EBADF;
	file = fget_light(fd, &fput);
	if (file) {
		if (file->f_mode & FMODE_WRITE)
			error = do_vmsplice(file, iov, nr_segs, flags);

		fput_light(file, fput);
	}

	return error;
}

1275 1276 1277
asmlinkage long sys_splice(int fd_in, loff_t __user *off_in,
			   int fd_out, loff_t __user *off_out,
			   size_t len, unsigned int flags)
1278 1279 1280 1281 1282 1283 1284 1285 1286
{
	long error;
	struct file *in, *out;
	int fput_in, fput_out;

	if (unlikely(!len))
		return 0;

	error = -EBADF;
1287
	in = fget_light(fd_in, &fput_in);
1288 1289
	if (in) {
		if (in->f_mode & FMODE_READ) {
1290
			out = fget_light(fd_out, &fput_out);
1291 1292
			if (out) {
				if (out->f_mode & FMODE_WRITE)
1293 1294 1295
					error = do_splice(in, off_in,
							  out, off_out,
							  len, flags);
1296 1297 1298 1299 1300 1301 1302 1303 1304
				fput_light(out, fput_out);
			}
		}

		fput_light(in, fput_in);
	}

	return error;
}
1305 1306 1307 1308 1309 1310 1311 1312 1313

/*
 * Link contents of ipipe to opipe.
 */
static int link_pipe(struct pipe_inode_info *ipipe,
		     struct pipe_inode_info *opipe,
		     size_t len, unsigned int flags)
{
	struct pipe_buffer *ibuf, *obuf;
1314 1315 1316
	int ret, do_wakeup, i, ipipe_first;

	ret = do_wakeup = ipipe_first = 0;
1317 1318 1319 1320 1321 1322 1323

	/*
	 * Potential ABBA deadlock, work around it by ordering lock
	 * grabbing by inode address. Otherwise two different processes
	 * could deadlock (one doing tee from A -> B, the other from B -> A).
	 */
	if (ipipe->inode < opipe->inode) {
1324
		ipipe_first = 1;
1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356
		mutex_lock(&ipipe->inode->i_mutex);
		mutex_lock(&opipe->inode->i_mutex);
	} else {
		mutex_lock(&opipe->inode->i_mutex);
		mutex_lock(&ipipe->inode->i_mutex);
	}

	for (i = 0;; i++) {
		if (!opipe->readers) {
			send_sig(SIGPIPE, current, 0);
			if (!ret)
				ret = -EPIPE;
			break;
		}
		if (ipipe->nrbufs - i) {
			ibuf = ipipe->bufs + ((ipipe->curbuf + i) & (PIPE_BUFFERS - 1));

			/*
			 * If we have room, fill this buffer
			 */
			if (opipe->nrbufs < PIPE_BUFFERS) {
				int nbuf = (opipe->curbuf + opipe->nrbufs) & (PIPE_BUFFERS - 1);

				/*
				 * Get a reference to this pipe buffer,
				 * so we can copy the contents over.
				 */
				ibuf->ops->get(ipipe, ibuf);

				obuf = opipe->bufs + nbuf;
				*obuf = *ibuf;

1357 1358 1359 1360 1361 1362
				/*
				 * Don't inherit the gift flag, we need to
				 * prevent multiple steals of this page.
				 */
				obuf->flags &= ~PIPE_BUF_FLAG_GIFT;

1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378
				if (obuf->len > len)
					obuf->len = len;

				opipe->nrbufs++;
				do_wakeup = 1;
				ret += obuf->len;
				len -= obuf->len;

				if (!len)
					break;
				if (opipe->nrbufs < PIPE_BUFFERS)
					continue;
			}

			/*
			 * We have input available, but no output room.
1379 1380 1381
			 * If we already copied data, return that. If we
			 * need to drop the opipe lock, it must be ordered
			 * last to avoid deadlocks.
1382
			 */
1383
			if ((flags & SPLICE_F_NONBLOCK) || !ipipe_first) {
1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416
				if (!ret)
					ret = -EAGAIN;
				break;
			}
			if (signal_pending(current)) {
				if (!ret)
					ret = -ERESTARTSYS;
				break;
			}
			if (do_wakeup) {
				smp_mb();
				if (waitqueue_active(&opipe->wait))
					wake_up_interruptible(&opipe->wait);
				kill_fasync(&opipe->fasync_readers, SIGIO, POLL_IN);
				do_wakeup = 0;
			}

			opipe->waiting_writers++;
			pipe_wait(opipe);
			opipe->waiting_writers--;
			continue;
		}

		/*
		 * No input buffers, do the usual checks for available
		 * writers and blocking and wait if necessary
		 */
		if (!ipipe->writers)
			break;
		if (!ipipe->waiting_writers) {
			if (ret)
				break;
		}
1417 1418 1419 1420 1421 1422
		/*
		 * pipe_wait() drops the ipipe mutex. To avoid deadlocks
		 * with another process, we can only safely do that if
		 * the ipipe lock is ordered last.
		 */
		if ((flags & SPLICE_F_NONBLOCK) || ipipe_first) {
1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499
			if (!ret)
				ret = -EAGAIN;
			break;
		}
		if (signal_pending(current)) {
			if (!ret)
				ret = -ERESTARTSYS;
			break;
		}

		if (waitqueue_active(&ipipe->wait))
			wake_up_interruptible_sync(&ipipe->wait);
		kill_fasync(&ipipe->fasync_writers, SIGIO, POLL_OUT);

		pipe_wait(ipipe);
	}

	mutex_unlock(&ipipe->inode->i_mutex);
	mutex_unlock(&opipe->inode->i_mutex);

	if (do_wakeup) {
		smp_mb();
		if (waitqueue_active(&opipe->wait))
			wake_up_interruptible(&opipe->wait);
		kill_fasync(&opipe->fasync_readers, SIGIO, POLL_IN);
	}

	return ret;
}

/*
 * This is a tee(1) implementation that works on pipes. It doesn't copy
 * any data, it simply references the 'in' pages on the 'out' pipe.
 * The 'flags' used are the SPLICE_F_* variants, currently the only
 * applicable one is SPLICE_F_NONBLOCK.
 */
static long do_tee(struct file *in, struct file *out, size_t len,
		   unsigned int flags)
{
	struct pipe_inode_info *ipipe = in->f_dentry->d_inode->i_pipe;
	struct pipe_inode_info *opipe = out->f_dentry->d_inode->i_pipe;

	/*
	 * Link ipipe to the two output pipes, consuming as we go along.
	 */
	if (ipipe && opipe)
		return link_pipe(ipipe, opipe, len, flags);

	return -EINVAL;
}

asmlinkage long sys_tee(int fdin, int fdout, size_t len, unsigned int flags)
{
	struct file *in;
	int error, fput_in;

	if (unlikely(!len))
		return 0;

	error = -EBADF;
	in = fget_light(fdin, &fput_in);
	if (in) {
		if (in->f_mode & FMODE_READ) {
			int fput_out;
			struct file *out = fget_light(fdout, &fput_out);

			if (out) {
				if (out->f_mode & FMODE_WRITE)
					error = do_tee(in, out, len, flags);
				fput_light(out, fput_out);
			}
		}
 		fput_light(in, fput_in);
 	}

	return error;
}