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 21 22
 *
 */
#include <linux/fs.h>
#include <linux/file.h>
#include <linux/pagemap.h>
23
#include <linux/splice.h>
24
#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
#include <linux/security.h>
32

33 34 35 36 37 38
/*
 * 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.
 */
39
static int page_cache_pipe_buf_steal(struct pipe_inode_info *pipe,
40 41 42
				     struct pipe_buffer *buf)
{
	struct page *page = buf->page;
43
	struct address_space *mapping;
44

45 46
	lock_page(page);

47 48 49
	mapping = page_mapping(page);
	if (mapping) {
		WARN_ON(!PageUptodate(page));
50

51 52 53 54 55 56 57 58 59
		/*
		 * 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);
60

61
		if (PagePrivate(page))
62
			try_to_release_page(page, GFP_KERNEL);
63

64 65 66 67 68 69 70 71
		/*
		 * 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;
		}
72
	}
73

74 75 76 77 78 79
	/*
	 * Raced with truncate or failed to remove page from current
	 * address space, unlock and return failure.
	 */
	unlock_page(page);
	return 1;
80 81
}

82
static void page_cache_pipe_buf_release(struct pipe_inode_info *pipe,
83 84 85
					struct pipe_buffer *buf)
{
	page_cache_release(buf->page);
J
Jens Axboe 已提交
86
	buf->flags &= ~PIPE_BUF_FLAG_LRU;
87 88
}

89 90 91 92
/*
 * Check whether the contents of buf is OK to access. Since the content
 * is a page cache page, IO may be in flight.
 */
93 94
static int page_cache_pipe_buf_confirm(struct pipe_inode_info *pipe,
				       struct pipe_buffer *buf)
95 96
{
	struct page *page = buf->page;
97
	int err;
98 99

	if (!PageUptodate(page)) {
100 101 102 103
		lock_page(page);

		/*
		 * Page got truncated/unhashed. This will cause a 0-byte
I
Ingo Molnar 已提交
104
		 * splice, if this is the first page.
105 106 107 108 109
		 */
		if (!page->mapping) {
			err = -ENODATA;
			goto error;
		}
110

111
		/*
I
Ingo Molnar 已提交
112
		 * Uh oh, read-error from disk.
113 114 115 116 117 118 119
		 */
		if (!PageUptodate(page)) {
			err = -EIO;
			goto error;
		}

		/*
120
		 * Page is ok afterall, we are done.
121
		 */
122 123 124
		unlock_page(page);
	}

125
	return 0;
126 127
error:
	unlock_page(page);
128
	return err;
129 130
}

131
static const struct pipe_buf_operations page_cache_pipe_buf_ops = {
132
	.can_merge = 0,
133 134
	.map = generic_pipe_buf_map,
	.unmap = generic_pipe_buf_unmap,
135
	.confirm = page_cache_pipe_buf_confirm,
136
	.release = page_cache_pipe_buf_release,
137
	.steal = page_cache_pipe_buf_steal,
138
	.get = generic_pipe_buf_get,
139 140
};

141 142 143
static int user_page_pipe_buf_steal(struct pipe_inode_info *pipe,
				    struct pipe_buffer *buf)
{
144 145 146
	if (!(buf->flags & PIPE_BUF_FLAG_GIFT))
		return 1;

J
Jens Axboe 已提交
147
	buf->flags |= PIPE_BUF_FLAG_LRU;
148
	return generic_pipe_buf_steal(pipe, buf);
149 150
}

151
static const struct pipe_buf_operations user_page_pipe_buf_ops = {
152
	.can_merge = 0,
153 154
	.map = generic_pipe_buf_map,
	.unmap = generic_pipe_buf_unmap,
155
	.confirm = generic_pipe_buf_confirm,
156 157
	.release = page_cache_pipe_buf_release,
	.steal = user_page_pipe_buf_steal,
158
	.get = generic_pipe_buf_get,
159 160
};

161 162 163 164 165 166 167 168 169 170
/**
 * splice_to_pipe - fill passed data into a pipe
 * @pipe:	pipe to fill
 * @spd:	data to fill
 *
 * Description:
 *    @spd contains a map of pages and len/offset tupples, a long with
 *    the struct pipe_buf_operations associated with these pages. This
 *    function will link that data to the pipe.
 *
171
 */
172 173
ssize_t splice_to_pipe(struct pipe_inode_info *pipe,
		       struct splice_pipe_desc *spd)
174
{
175
	unsigned int spd_pages = spd->nr_pages;
176
	int ret, do_wakeup, page_nr;
177 178 179

	ret = 0;
	do_wakeup = 0;
180
	page_nr = 0;
181

182 183
	if (pipe->inode)
		mutex_lock(&pipe->inode->i_mutex);
184 185

	for (;;) {
186
		if (!pipe->readers) {
187 188 189 190 191 192
			send_sig(SIGPIPE, current, 0);
			if (!ret)
				ret = -EPIPE;
			break;
		}

193 194
		if (pipe->nrbufs < PIPE_BUFFERS) {
			int newbuf = (pipe->curbuf + pipe->nrbufs) & (PIPE_BUFFERS - 1);
195
			struct pipe_buffer *buf = pipe->bufs + newbuf;
196

197 198 199
			buf->page = spd->pages[page_nr];
			buf->offset = spd->partial[page_nr].offset;
			buf->len = spd->partial[page_nr].len;
200
			buf->private = spd->partial[page_nr].private;
201
			buf->ops = spd->ops;
202 203 204
			if (spd->flags & SPLICE_F_GIFT)
				buf->flags |= PIPE_BUF_FLAG_GIFT;

205
			pipe->nrbufs++;
206 207 208
			page_nr++;
			ret += buf->len;

209 210
			if (pipe->inode)
				do_wakeup = 1;
211

212
			if (!--spd->nr_pages)
213
				break;
214
			if (pipe->nrbufs < PIPE_BUFFERS)
215 216 217 218 219
				continue;

			break;
		}

220
		if (spd->flags & SPLICE_F_NONBLOCK) {
221 222 223 224 225
			if (!ret)
				ret = -EAGAIN;
			break;
		}

226 227 228 229 230 231 232
		if (signal_pending(current)) {
			if (!ret)
				ret = -ERESTARTSYS;
			break;
		}

		if (do_wakeup) {
233
			smp_mb();
234 235 236
			if (waitqueue_active(&pipe->wait))
				wake_up_interruptible_sync(&pipe->wait);
			kill_fasync(&pipe->fasync_readers, SIGIO, POLL_IN);
237 238 239
			do_wakeup = 0;
		}

240 241 242
		pipe->waiting_writers++;
		pipe_wait(pipe);
		pipe->waiting_writers--;
243 244
	}

245
	if (pipe->inode) {
246
		mutex_unlock(&pipe->inode->i_mutex);
247

248 249 250 251 252 253
		if (do_wakeup) {
			smp_mb();
			if (waitqueue_active(&pipe->wait))
				wake_up_interruptible(&pipe->wait);
			kill_fasync(&pipe->fasync_readers, SIGIO, POLL_IN);
		}
254 255
	}

256
	while (page_nr < spd_pages)
257
		page_cache_release(spd->pages[page_nr++]);
258 259 260 261

	return ret;
}

262
static int
263 264 265
__generic_file_splice_read(struct file *in, loff_t *ppos,
			   struct pipe_inode_info *pipe, size_t len,
			   unsigned int flags)
266 267
{
	struct address_space *mapping = in->f_mapping;
F
Fengguang Wu 已提交
268
	unsigned int loff, nr_pages, req_pages;
269
	struct page *pages[PIPE_BUFFERS];
270
	struct partial_page partial[PIPE_BUFFERS];
271
	struct page *page;
272 273
	pgoff_t index, end_index;
	loff_t isize;
274
	int error, page_nr;
275 276 277 278 279 280
	struct splice_pipe_desc spd = {
		.pages = pages,
		.partial = partial,
		.flags = flags,
		.ops = &page_cache_pipe_buf_ops,
	};
281

282
	index = *ppos >> PAGE_CACHE_SHIFT;
283
	loff = *ppos & ~PAGE_CACHE_MASK;
F
Fengguang Wu 已提交
284 285
	req_pages = (len + loff + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
	nr_pages = min(req_pages, (unsigned)PIPE_BUFFERS);
286

287 288 289 290
	/*
	 * Lookup the (hopefully) full range of pages we need.
	 */
	spd.nr_pages = find_get_pages_contig(mapping, index, nr_pages, pages);
291
	index += spd.nr_pages;
292

293 294
	/*
	 * If find_get_pages_contig() returned fewer pages than we needed,
295
	 * readahead/allocate the rest and fill in the holes.
296
	 */
297
	if (spd.nr_pages < nr_pages)
298 299
		page_cache_sync_readahead(mapping, &in->f_ra, in,
				index, req_pages - spd.nr_pages);
300

301
	error = 0;
302
	while (spd.nr_pages < nr_pages) {
303
		/*
304 305
		 * Page could be there, find_get_pages_contig() breaks on
		 * the first hole.
306
		 */
307 308 309
		page = find_get_page(mapping, index);
		if (!page) {
			/*
310
			 * page didn't exist, allocate one.
311 312 313 314 315 316
			 */
			page = page_cache_alloc_cold(mapping);
			if (!page)
				break;

			error = add_to_page_cache_lru(page, mapping, index,
317
					      GFP_KERNEL);
318 319
			if (unlikely(error)) {
				page_cache_release(page);
320 321
				if (error == -EEXIST)
					continue;
322 323
				break;
			}
324 325 326 327 328
			/*
			 * add_to_page_cache() locks the page, unlock it
			 * to avoid convoluting the logic below even more.
			 */
			unlock_page(page);
329 330
		}

331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353
		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];

354
		if (PageReadahead(page))
355
			page_cache_async_readahead(mapping, &in->f_ra, in,
F
Fengguang Wu 已提交
356
					page, index, req_pages - page_nr);
357

358 359 360 361
		/*
		 * If the page isn't uptodate, we may need to start io on it
		 */
		if (!PageUptodate(page)) {
362 363 364 365
			/*
			 * If in nonblock mode then dont block on waiting
			 * for an in-flight io page
			 */
366 367 368 369 370
			if (flags & SPLICE_F_NONBLOCK) {
				if (TestSetPageLocked(page))
					break;
			} else
				lock_page(page);
371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387

			/*
			 * 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;
			}
388

389 390 391 392
			/*
			 * need to read in the page
			 */
			error = mapping->a_ops->readpage(in, page);
393
			if (unlikely(error)) {
394 395 396 397 398 399
				/*
				 * 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.
				 */
400
				if (error == AOP_TRUNCATED_PAGE)
401 402
					error = 0;

403 404
				break;
			}
405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420
		}
fill_it:
		/*
		 * i_size must be checked after PageUptodate.
		 */
		isize = i_size_read(mapping->host);
		end_index = (isize - 1) >> PAGE_CACHE_SHIFT;
		if (unlikely(!isize || index > end_index))
			break;

		/*
		 * if this is the last page, see if we need to shrink
		 * the length and stop
		 */
		if (end_index == index) {
			unsigned int plen;
421 422

			/*
423
			 * max good bytes in this page
424
			 */
425 426
			plen = ((isize - 1) & ~PAGE_CACHE_MASK) + 1;
			if (plen <= loff)
427 428 429
				break;

			/*
430
			 * force quit after adding this page
431
			 */
432 433
			this_len = min(this_len, plen - loff);
			len = this_len;
434
		}
435

436 437
		partial[page_nr].offset = loff;
		partial[page_nr].len = this_len;
438
		len -= this_len;
439
		loff = 0;
440 441
		spd.nr_pages++;
		index++;
442 443
	}

444
	/*
445
	 * Release any pages at the end, if we quit early. 'page_nr' is how far
446 447 448 449
	 * we got, 'nr_pages' is how many pages are in the map.
	 */
	while (page_nr < nr_pages)
		page_cache_release(pages[page_nr++]);
450
	in->f_ra.prev_index = index;
451

452
	if (spd.nr_pages)
453
		return splice_to_pipe(pipe, &spd);
454

455
	return error;
456 457
}

458 459 460
/**
 * generic_file_splice_read - splice data from file to a pipe
 * @in:		file to splice from
461
 * @ppos:	position in @in
462 463 464 465
 * @pipe:	pipe to splice to
 * @len:	number of bytes to splice
 * @flags:	splice modifier flags
 *
466 467 468 469 470
 * Description:
 *    Will read pages from given file and fill them into a pipe. Can be
 *    used as long as the address_space operations for the source implements
 *    a readpage() hook.
 *
471
 */
472 473 474
ssize_t generic_file_splice_read(struct file *in, loff_t *ppos,
				 struct pipe_inode_info *pipe, size_t len,
				 unsigned int flags)
475 476 477
{
	ssize_t spliced;
	int ret;
478 479 480 481 482 483 484 485 486
	loff_t isize, left;

	isize = i_size_read(in->f_mapping->host);
	if (unlikely(*ppos >= isize))
		return 0;

	left = isize - *ppos;
	if (unlikely(left < len))
		len = left;
487 488 489

	ret = 0;
	spliced = 0;
490
	while (len && !spliced) {
491
		ret = __generic_file_splice_read(in, ppos, pipe, len, flags);
492

493
		if (ret < 0)
494
			break;
495 496 497 498 499 500 501 502
		else if (!ret) {
			if (spliced)
				break;
			if (flags & SPLICE_F_NONBLOCK) {
				ret = -EAGAIN;
				break;
			}
		}
503

504
		*ppos += ret;
505 506 507 508 509 510 511 512 513 514
		len -= ret;
		spliced += ret;
	}

	if (spliced)
		return spliced;

	return ret;
}

515 516
EXPORT_SYMBOL(generic_file_splice_read);

517
/*
518
 * Send 'sd->len' bytes to socket from 'sd->file' at position 'sd->pos'
J
Jens Axboe 已提交
519
 * using sendpage(). Return the number of bytes sent.
520
 */
521
static int pipe_to_sendpage(struct pipe_inode_info *pipe,
522 523
			    struct pipe_buffer *buf, struct splice_desc *sd)
{
524
	struct file *file = sd->u.file;
525
	loff_t pos = sd->pos;
526
	int ret, more;
527

528
	ret = buf->ops->confirm(pipe, buf);
529 530
	if (!ret) {
		more = (sd->flags & SPLICE_F_MORE) || sd->len < sd->total_len;
531

532 533 534
		ret = file->f_op->sendpage(file, buf->page, buf->offset,
					   sd->len, &pos, more);
	}
535

J
Jens Axboe 已提交
536
	return ret;
537 538 539 540 541 542 543 544 545 546 547 548 549 550 551
}

/*
 * 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.
 *
552 553 554 555 556 557
 * 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.
558
 */
559
static int pipe_to_file(struct pipe_inode_info *pipe, struct pipe_buffer *buf,
560 561
			struct splice_desc *sd)
{
562
	struct file *file = sd->u.file;
563
	struct address_space *mapping = file->f_mapping;
J
Jens Axboe 已提交
564
	unsigned int offset, this_len;
565 566
	struct page *page;
	pgoff_t index;
567
	int ret;
568 569

	/*
570
	 * make sure the data in this buffer is uptodate
571
	 */
572
	ret = buf->ops->confirm(pipe, buf);
573 574
	if (unlikely(ret))
		return ret;
575 576 577 578

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

J
Jens Axboe 已提交
579 580 581 582
	this_len = sd->len;
	if (this_len + offset > PAGE_CACHE_SIZE)
		this_len = PAGE_CACHE_SIZE - offset;

N
Nick Piggin 已提交
583 584 585 586 587 588 589 590
find_page:
	page = find_lock_page(mapping, index);
	if (!page) {
		ret = -ENOMEM;
		page = page_cache_alloc_cold(mapping);
		if (unlikely(!page))
			goto out_ret;

591
		/*
N
Nick Piggin 已提交
592
		 * This will also lock the page
593
		 */
N
Nick Piggin 已提交
594 595 596
		ret = add_to_page_cache_lru(page, mapping, index,
					    GFP_KERNEL);
		if (unlikely(ret))
597
			goto out_release;
N
Nick Piggin 已提交
598
	}
599

J
Jens Axboe 已提交
600
	ret = mapping->a_ops->prepare_write(file, page, offset, offset+this_len);
601 602 603 604 605
	if (unlikely(ret)) {
		loff_t isize = i_size_read(mapping->host);

		if (ret != AOP_TRUNCATED_PAGE)
			unlock_page(page);
606
		page_cache_release(page);
607 608 609 610 611 612 613 614 615 616
		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);

617
		goto out_ret;
618
	}
619

620
	if (buf->page != page) {
621 622 623
		/*
		 * Careful, ->map() uses KM_USER0!
		 */
624
		char *src = buf->ops->map(pipe, buf, 1);
625
		char *dst = kmap_atomic(page, KM_USER1);
626

J
Jens Axboe 已提交
627
		memcpy(dst + offset, src + buf->offset, this_len);
628
		flush_dcache_page(page);
629
		kunmap_atomic(dst, KM_USER1);
630
		buf->ops->unmap(pipe, buf, src);
631
	}
632

J
Jens Axboe 已提交
633
	ret = mapping->a_ops->commit_write(file, page, offset, offset+this_len);
634 635 636 637 638 639 640
	if (ret) {
		if (ret == AOP_TRUNCATED_PAGE) {
			page_cache_release(page);
			goto find_page;
		}
		if (ret < 0)
			goto out;
641
		/*
642 643
		 * Partial write has happened, so 'ret' already initialized by
		 * number of bytes written, Where is nothing we have to do here.
644
		 */
645
	} else
646
		ret = this_len;
647 648 649 650 651
	/*
	 * Return the number of bytes written and mark page as
	 * accessed, we are now done!
	 */
	mark_page_accessed(page);
652
out:
653
	unlock_page(page);
654 655
out_release:
	page_cache_release(page);
656
out_ret:
657 658 659
	return ret;
}

660 661 662 663 664 665 666 667 668 669 670 671
/**
 * __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.
 *
672
 */
J
Jens Axboe 已提交
673 674
ssize_t __splice_from_pipe(struct pipe_inode_info *pipe, struct splice_desc *sd,
			   splice_actor *actor)
675 676 677 678 679 680 681
{
	int ret, do_wakeup, err;

	ret = 0;
	do_wakeup = 0;

	for (;;) {
682 683
		if (pipe->nrbufs) {
			struct pipe_buffer *buf = pipe->bufs + pipe->curbuf;
684
			const struct pipe_buf_operations *ops = buf->ops;
685

J
Jens Axboe 已提交
686 687 688
			sd->len = buf->len;
			if (sd->len > sd->total_len)
				sd->len = sd->total_len;
689

J
Jens Axboe 已提交
690
			err = actor(pipe, buf, sd);
J
Jens Axboe 已提交
691
			if (err <= 0) {
692 693 694 695 696 697
				if (!ret && err != -ENODATA)
					ret = err;

				break;
			}

J
Jens Axboe 已提交
698 699 700 701
			ret += err;
			buf->offset += err;
			buf->len -= err;

J
Jens Axboe 已提交
702 703 704 705
			sd->len -= err;
			sd->pos += err;
			sd->total_len -= err;
			if (sd->len)
J
Jens Axboe 已提交
706
				continue;
I
Ingo Molnar 已提交
707

708 709
			if (!buf->len) {
				buf->ops = NULL;
710
				ops->release(pipe, buf);
711 712 713 714
				pipe->curbuf = (pipe->curbuf + 1) & (PIPE_BUFFERS - 1);
				pipe->nrbufs--;
				if (pipe->inode)
					do_wakeup = 1;
715 716
			}

J
Jens Axboe 已提交
717
			if (!sd->total_len)
718 719 720
				break;
		}

721
		if (pipe->nrbufs)
722
			continue;
723
		if (!pipe->writers)
724
			break;
725
		if (!pipe->waiting_writers) {
726 727 728 729
			if (ret)
				break;
		}

J
Jens Axboe 已提交
730
		if (sd->flags & SPLICE_F_NONBLOCK) {
731 732 733 734 735
			if (!ret)
				ret = -EAGAIN;
			break;
		}

736 737 738 739 740 741 742
		if (signal_pending(current)) {
			if (!ret)
				ret = -ERESTARTSYS;
			break;
		}

		if (do_wakeup) {
743
			smp_mb();
744 745 746
			if (waitqueue_active(&pipe->wait))
				wake_up_interruptible_sync(&pipe->wait);
			kill_fasync(&pipe->fasync_writers, SIGIO, POLL_OUT);
747 748 749
			do_wakeup = 0;
		}

750
		pipe_wait(pipe);
751 752 753
	}

	if (do_wakeup) {
754
		smp_mb();
755 756 757
		if (waitqueue_active(&pipe->wait))
			wake_up_interruptible(&pipe->wait);
		kill_fasync(&pipe->fasync_writers, SIGIO, POLL_OUT);
758 759 760 761
	}

	return ret;
}
M
Mark Fasheh 已提交
762
EXPORT_SYMBOL(__splice_from_pipe);
763

764 765 766 767 768 769 770 771 772 773 774 775 776 777
/**
 * 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:
 *    See __splice_from_pipe. This function locks the input and output inodes,
 *    otherwise it's identical to __splice_from_pipe().
 *
 */
778 779 780 781 782 783
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;
	struct inode *inode = out->f_mapping->host;
J
Jens Axboe 已提交
784 785 786 787
	struct splice_desc sd = {
		.total_len = len,
		.flags = flags,
		.pos = *ppos,
788
		.u.file = out,
J
Jens Axboe 已提交
789
	};
790 791 792 793 794 795 796 797

	/*
	 * The actor worker might be calling ->prepare_write and
	 * ->commit_write. Most of the time, these expect i_mutex to
	 * be held. Since this may result in an ABBA deadlock with
	 * pipe->inode, we have to order lock acquiry here.
	 */
	inode_double_lock(inode, pipe->inode);
J
Jens Axboe 已提交
798
	ret = __splice_from_pipe(pipe, &sd, actor);
799 800 801 802 803 804 805 806 807
	inode_double_unlock(inode, pipe->inode);

	return ret;
}

/**
 * generic_file_splice_write_nolock - generic_file_splice_write without mutexes
 * @pipe:	pipe info
 * @out:	file to write to
808
 * @ppos:	position in @out
809 810 811
 * @len:	number of bytes to splice
 * @flags:	splice modifier flags
 *
812 813 814 815
 * Description:
 *    Will either move or copy pages (determined by @flags options) from
 *    the given pipe inode to the given file. The caller is responsible
 *    for acquiring i_mutex on both inodes.
816 817 818 819 820 821 822 823
 *
 */
ssize_t
generic_file_splice_write_nolock(struct pipe_inode_info *pipe, struct file *out,
				 loff_t *ppos, size_t len, unsigned int flags)
{
	struct address_space *mapping = out->f_mapping;
	struct inode *inode = mapping->host;
J
Jens Axboe 已提交
824 825 826 827
	struct splice_desc sd = {
		.total_len = len,
		.flags = flags,
		.pos = *ppos,
828
		.u.file = out,
J
Jens Axboe 已提交
829
	};
830 831 832
	ssize_t ret;
	int err;

833
	err = remove_suid(out->f_path.dentry);
834 835 836
	if (unlikely(err))
		return err;

J
Jens Axboe 已提交
837
	ret = __splice_from_pipe(pipe, &sd, pipe_to_file);
838
	if (ret > 0) {
839 840
		unsigned long nr_pages;

841
		*ppos += ret;
842
		nr_pages = (ret + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
843 844 845 846 847 848 849 850 851 852 853 854

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

			if (err)
				ret = err;
		}
855
		balance_dirty_pages_ratelimited_nr(mapping, nr_pages);
856 857 858 859 860 861 862
	}

	return ret;
}

EXPORT_SYMBOL(generic_file_splice_write_nolock);

863 864
/**
 * generic_file_splice_write - splice data from a pipe to a file
865
 * @pipe:	pipe info
866
 * @out:	file to write to
867
 * @ppos:	position in @out
868 869 870
 * @len:	number of bytes to splice
 * @flags:	splice modifier flags
 *
871 872 873
 * Description:
 *    Will either move or copy pages (determined by @flags options) from
 *    the given pipe inode to the given file.
874 875
 *
 */
876 877
ssize_t
generic_file_splice_write(struct pipe_inode_info *pipe, struct file *out,
878
			  loff_t *ppos, size_t len, unsigned int flags)
879
{
880
	struct address_space *mapping = out->f_mapping;
881
	struct inode *inode = mapping->host;
882
	ssize_t ret;
883 884
	int err;

885
	err = should_remove_suid(out->f_path.dentry);
886 887
	if (unlikely(err)) {
		mutex_lock(&inode->i_mutex);
888
		err = __remove_suid(out->f_path.dentry, err);
889 890 891 892
		mutex_unlock(&inode->i_mutex);
		if (err)
			return err;
	}
893

894
	ret = splice_from_pipe(pipe, out, ppos, len, flags, pipe_to_file);
J
Jens Axboe 已提交
895
	if (ret > 0) {
896 897
		unsigned long nr_pages;

J
Jens Axboe 已提交
898
		*ppos += ret;
899
		nr_pages = (ret + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
J
Jens Axboe 已提交
900 901 902 903 904 905 906 907 908 909

		/*
		 * If file or inode is SYNC and we actually wrote some data,
		 * sync it.
		 */
		if (unlikely((out->f_flags & O_SYNC) || IS_SYNC(inode))) {
			mutex_lock(&inode->i_mutex);
			err = generic_osync_inode(inode, mapping,
						  OSYNC_METADATA|OSYNC_DATA);
			mutex_unlock(&inode->i_mutex);
910

J
Jens Axboe 已提交
911 912 913
			if (err)
				ret = err;
		}
914
		balance_dirty_pages_ratelimited_nr(mapping, nr_pages);
915 916 917
	}

	return ret;
918 919
}

920 921
EXPORT_SYMBOL(generic_file_splice_write);

922 923
/**
 * generic_splice_sendpage - splice data from a pipe to a socket
924
 * @pipe:	pipe to splice from
925
 * @out:	socket to write to
926
 * @ppos:	position in @out
927 928 929
 * @len:	number of bytes to splice
 * @flags:	splice modifier flags
 *
930 931 932
 * Description:
 *    Will send @len bytes from the pipe to a network socket. No data copying
 *    is involved.
933 934
 *
 */
935
ssize_t generic_splice_sendpage(struct pipe_inode_info *pipe, struct file *out,
936
				loff_t *ppos, size_t len, unsigned int flags)
937
{
938
	return splice_from_pipe(pipe, out, ppos, len, flags, pipe_to_sendpage);
939 940
}

941
EXPORT_SYMBOL(generic_splice_sendpage);
J
Jeff Garzik 已提交
942

943 944 945
/*
 * Attempt to initiate a splice from pipe to file.
 */
946
static long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
947
			   loff_t *ppos, size_t len, unsigned int flags)
948 949 950
{
	int ret;

951
	if (unlikely(!out->f_op || !out->f_op->splice_write))
952 953
		return -EINVAL;

954
	if (unlikely(!(out->f_mode & FMODE_WRITE)))
955 956
		return -EBADF;

957
	ret = rw_verify_area(WRITE, out, ppos, len);
958 959 960
	if (unlikely(ret < 0))
		return ret;

961 962 963 964
	ret = security_file_permission(out, MAY_WRITE);
	if (unlikely(ret < 0))
		return ret;

965
	return out->f_op->splice_write(pipe, out, ppos, len, flags);
966 967
}

968 969 970
/*
 * Attempt to initiate a splice from a file to a pipe.
 */
971 972 973
static long do_splice_to(struct file *in, loff_t *ppos,
			 struct pipe_inode_info *pipe, size_t len,
			 unsigned int flags)
974 975 976
{
	int ret;

977
	if (unlikely(!in->f_op || !in->f_op->splice_read))
978 979
		return -EINVAL;

980
	if (unlikely(!(in->f_mode & FMODE_READ)))
981 982
		return -EBADF;

983
	ret = rw_verify_area(READ, in, ppos, len);
984 985 986
	if (unlikely(ret < 0))
		return ret;

987 988 989 990
	ret = security_file_permission(in, MAY_READ);
	if (unlikely(ret < 0))
		return ret;

991
	return in->f_op->splice_read(in, ppos, pipe, len, flags);
992 993
}

994 995 996 997 998 999 1000 1001 1002 1003 1004 1005
/**
 * 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
 *    pipe is cached in the process, and reused during the life time of
 *    that process.
 *
J
Jens Axboe 已提交
1006 1007 1008
 */
ssize_t splice_direct_to_actor(struct file *in, struct splice_desc *sd,
			       splice_direct_actor *actor)
1009 1010 1011 1012
{
	struct pipe_inode_info *pipe;
	long ret, bytes;
	umode_t i_mode;
J
Jens Axboe 已提交
1013 1014
	size_t len;
	int i, flags;
1015 1016 1017 1018 1019 1020

	/*
	 * 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!
	 */
1021
	i_mode = in->f_path.dentry->d_inode->i_mode;
1022 1023 1024 1025 1026 1027 1028 1029
	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;
1030
	if (unlikely(!pipe)) {
1031 1032 1033 1034 1035 1036
		pipe = alloc_pipe_info(NULL);
		if (!pipe)
			return -ENOMEM;

		/*
		 * We don't have an immediate reader, but we'll read the stuff
1037
		 * out of the pipe right after the splice_to_pipe(). So set
1038 1039 1040 1041 1042 1043 1044 1045
		 * PIPE_READERS appropriately.
		 */
		pipe->readers = 1;

		current->splice_pipe = pipe;
	}

	/*
I
Ingo Molnar 已提交
1046
	 * Do the splice.
1047 1048 1049
	 */
	ret = 0;
	bytes = 0;
J
Jens Axboe 已提交
1050 1051 1052 1053 1054 1055 1056
	len = sd->total_len;
	flags = sd->flags;

	/*
	 * Don't block on output, we have to drain the direct pipe.
	 */
	sd->flags &= ~SPLICE_F_NONBLOCK;
1057 1058

	while (len) {
1059
		size_t read_len;
1060
		loff_t pos = sd->pos;
1061

1062
		ret = do_splice_to(in, &pos, pipe, len, flags);
1063
		if (unlikely(ret <= 0))
1064 1065 1066
			goto out_release;

		read_len = ret;
J
Jens Axboe 已提交
1067
		sd->total_len = read_len;
1068 1069 1070 1071 1072 1073

		/*
		 * 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:
		 */
J
Jens Axboe 已提交
1074
		ret = actor(pipe, sd);
1075
		if (unlikely(ret <= 0))
1076 1077 1078 1079
			goto out_release;

		bytes += ret;
		len -= ret;
1080
		sd->pos = pos;
1081

1082 1083
		if (ret < read_len)
			goto out_release;
1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110
	}

	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;
J
Jens Axboe 已提交
1111 1112 1113 1114 1115 1116 1117

}
EXPORT_SYMBOL(splice_direct_to_actor);

static int direct_splice_actor(struct pipe_inode_info *pipe,
			       struct splice_desc *sd)
{
1118
	struct file *file = sd->u.file;
J
Jens Axboe 已提交
1119 1120 1121 1122

	return do_splice_from(pipe, file, &sd->pos, sd->total_len, sd->flags);
}

1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137
/**
 * do_splice_direct - splices data directly between two files
 * @in:		file to splice from
 * @ppos:	input file offset
 * @out:	file to splice to
 * @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.
 *
 */
J
Jens Axboe 已提交
1138 1139 1140 1141 1142 1143 1144 1145
long do_splice_direct(struct file *in, loff_t *ppos, struct file *out,
		      size_t len, unsigned int flags)
{
	struct splice_desc sd = {
		.len		= len,
		.total_len	= len,
		.flags		= flags,
		.pos		= *ppos,
1146
		.u.file		= out,
J
Jens Axboe 已提交
1147
	};
1148
	long ret;
J
Jens Axboe 已提交
1149 1150

	ret = splice_direct_to_actor(in, &sd, direct_splice_actor);
1151 1152 1153
	if (ret > 0)
		*ppos += ret;

J
Jens Axboe 已提交
1154
	return ret;
1155 1156
}

1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169
/*
 * After the inode slimming patch, i_pipe/i_bdev/i_cdev share the same
 * location, so checking ->i_pipe is not enough to verify that this is a
 * pipe.
 */
static inline struct pipe_inode_info *pipe_info(struct inode *inode)
{
	if (S_ISFIFO(inode->i_mode))
		return inode->i_pipe;

	return NULL;
}

1170 1171 1172
/*
 * Determine where to splice to/from.
 */
1173 1174 1175
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)
1176
{
1177
	struct pipe_inode_info *pipe;
1178
	loff_t offset, *off;
J
Jens Axboe 已提交
1179
	long ret;
1180

1181
	pipe = pipe_info(in->f_path.dentry->d_inode);
1182 1183 1184
	if (pipe) {
		if (off_in)
			return -ESPIPE;
1185 1186 1187
		if (off_out) {
			if (out->f_op->llseek == no_llseek)
				return -EINVAL;
1188
			if (copy_from_user(&offset, off_out, sizeof(loff_t)))
1189
				return -EFAULT;
1190 1191 1192
			off = &offset;
		} else
			off = &out->f_pos;
1193

J
Jens Axboe 已提交
1194 1195 1196 1197 1198 1199
		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;
1200
	}
1201

1202
	pipe = pipe_info(out->f_path.dentry->d_inode);
1203 1204 1205
	if (pipe) {
		if (off_out)
			return -ESPIPE;
1206 1207 1208
		if (off_in) {
			if (in->f_op->llseek == no_llseek)
				return -EINVAL;
1209
			if (copy_from_user(&offset, off_in, sizeof(loff_t)))
1210
				return -EFAULT;
1211 1212 1213
			off = &offset;
		} else
			off = &in->f_pos;
1214

J
Jens Axboe 已提交
1215 1216 1217 1218 1219 1220
		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;
1221
	}
1222 1223 1224 1225

	return -EINVAL;
}

1226 1227 1228 1229 1230 1231 1232 1233 1234
/*
 * 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,
1235
				struct partial_page *partial, int aligned)
1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274
{
	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;
1275 1276 1277 1278 1279 1280 1281 1282 1283

		/*
		 * 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;

1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298
		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++) {
1299
			const int plen = min_t(size_t, len, PAGE_SIZE - off);
1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336

			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;
}

1337 1338 1339 1340 1341 1342
static int pipe_to_user(struct pipe_inode_info *pipe, struct pipe_buffer *buf,
			struct splice_desc *sd)
{
	char *src;
	int ret;

1343
	ret = buf->ops->confirm(pipe, buf);
1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 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 1417 1418 1419 1420 1421 1422 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
	if (unlikely(ret))
		return ret;

	/*
	 * See if we can use the atomic maps, by prefaulting in the
	 * pages and doing an atomic copy
	 */
	if (!fault_in_pages_writeable(sd->u.userptr, sd->len)) {
		src = buf->ops->map(pipe, buf, 1);
		ret = __copy_to_user_inatomic(sd->u.userptr, src + buf->offset,
							sd->len);
		buf->ops->unmap(pipe, buf, src);
		if (!ret) {
			ret = sd->len;
			goto out;
		}
	}

	/*
	 * No dice, use slow non-atomic map and copy
 	 */
	src = buf->ops->map(pipe, buf, 0);

	ret = sd->len;
	if (copy_to_user(sd->u.userptr, src + buf->offset, sd->len))
		ret = -EFAULT;

out:
	if (ret > 0)
		sd->u.userptr += ret;
	buf->ops->unmap(pipe, buf, src);
	return ret;
}

/*
 * For lack of a better implementation, implement vmsplice() to userspace
 * as a simple copy of the pipes pages to the user iov.
 */
static long vmsplice_to_user(struct file *file, const struct iovec __user *iov,
			     unsigned long nr_segs, unsigned int flags)
{
	struct pipe_inode_info *pipe;
	struct splice_desc sd;
	ssize_t size;
	int error;
	long ret;

	pipe = pipe_info(file->f_path.dentry->d_inode);
	if (!pipe)
		return -EBADF;

	if (pipe->inode)
		mutex_lock(&pipe->inode->i_mutex);

	error = ret = 0;
	while (nr_segs) {
		void __user *base;
		size_t len;

		/*
		 * 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;
		if (unlikely(!base)) {
			error = -EFAULT;
			break;
		}

		sd.len = 0;
		sd.total_len = len;
		sd.flags = flags;
		sd.u.userptr = base;
		sd.pos = 0;

		size = __splice_from_pipe(pipe, &sd, pipe_to_user);
		if (size < 0) {
			if (!ret)
				ret = size;

			break;
		}

		ret += size;

		if (size < len)
			break;

		nr_segs--;
		iov++;
	}

	if (pipe->inode)
		mutex_unlock(&pipe->inode->i_mutex);

	if (!ret)
		ret = error;

	return ret;
}

1455 1456 1457 1458 1459
/*
 * 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.
 */
1460 1461
static long vmsplice_to_pipe(struct file *file, const struct iovec __user *iov,
			     unsigned long nr_segs, unsigned int flags)
1462
{
1463
	struct pipe_inode_info *pipe;
1464 1465 1466 1467 1468 1469 1470 1471 1472
	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,
	};

1473
	pipe = pipe_info(file->f_path.dentry->d_inode);
1474
	if (!pipe)
1475 1476
		return -EBADF;

1477 1478
	spd.nr_pages = get_iovec_page_array(iov, nr_segs, pages, partial,
					    flags & SPLICE_F_GIFT);
1479 1480 1481
	if (spd.nr_pages <= 0)
		return spd.nr_pages;

1482
	return splice_to_pipe(pipe, &spd);
1483 1484
}

1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500
/*
 * 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().
 *
 */
1501 1502 1503 1504 1505 1506 1507
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;

1508 1509 1510 1511 1512
	if (unlikely(nr_segs > UIO_MAXIOV))
		return -EINVAL;
	else if (unlikely(!nr_segs))
		return 0;

1513 1514 1515 1516
	error = -EBADF;
	file = fget_light(fd, &fput);
	if (file) {
		if (file->f_mode & FMODE_WRITE)
1517 1518 1519
			error = vmsplice_to_pipe(file, iov, nr_segs, flags);
		else if (file->f_mode & FMODE_READ)
			error = vmsplice_to_user(file, iov, nr_segs, flags);
1520 1521 1522 1523 1524 1525 1526

		fput_light(file, fput);
	}

	return error;
}

1527 1528 1529
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)
1530 1531 1532 1533 1534 1535 1536 1537 1538
{
	long error;
	struct file *in, *out;
	int fput_in, fput_out;

	if (unlikely(!len))
		return 0;

	error = -EBADF;
1539
	in = fget_light(fd_in, &fput_in);
1540 1541
	if (in) {
		if (in->f_mode & FMODE_READ) {
1542
			out = fget_light(fd_out, &fput_out);
1543 1544
			if (out) {
				if (out->f_mode & FMODE_WRITE)
1545 1546 1547
					error = do_splice(in, off_in,
							  out, off_out,
							  len, flags);
1548 1549 1550 1551 1552 1553 1554 1555 1556
				fput_light(out, fput_out);
			}
		}

		fput_light(in, fput_in);
	}

	return error;
}
1557

1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636
/*
 * Make sure there's data to read. Wait for input if we can, otherwise
 * return an appropriate error.
 */
static int link_ipipe_prep(struct pipe_inode_info *pipe, unsigned int flags)
{
	int ret;

	/*
	 * Check ->nrbufs without the inode lock first. This function
	 * is speculative anyways, so missing one is ok.
	 */
	if (pipe->nrbufs)
		return 0;

	ret = 0;
	mutex_lock(&pipe->inode->i_mutex);

	while (!pipe->nrbufs) {
		if (signal_pending(current)) {
			ret = -ERESTARTSYS;
			break;
		}
		if (!pipe->writers)
			break;
		if (!pipe->waiting_writers) {
			if (flags & SPLICE_F_NONBLOCK) {
				ret = -EAGAIN;
				break;
			}
		}
		pipe_wait(pipe);
	}

	mutex_unlock(&pipe->inode->i_mutex);
	return ret;
}

/*
 * Make sure there's writeable room. Wait for room if we can, otherwise
 * return an appropriate error.
 */
static int link_opipe_prep(struct pipe_inode_info *pipe, unsigned int flags)
{
	int ret;

	/*
	 * Check ->nrbufs without the inode lock first. This function
	 * is speculative anyways, so missing one is ok.
	 */
	if (pipe->nrbufs < PIPE_BUFFERS)
		return 0;

	ret = 0;
	mutex_lock(&pipe->inode->i_mutex);

	while (pipe->nrbufs >= PIPE_BUFFERS) {
		if (!pipe->readers) {
			send_sig(SIGPIPE, current, 0);
			ret = -EPIPE;
			break;
		}
		if (flags & SPLICE_F_NONBLOCK) {
			ret = -EAGAIN;
			break;
		}
		if (signal_pending(current)) {
			ret = -ERESTARTSYS;
			break;
		}
		pipe->waiting_writers++;
		pipe_wait(pipe);
		pipe->waiting_writers--;
	}

	mutex_unlock(&pipe->inode->i_mutex);
	return ret;
}

1637 1638 1639 1640 1641 1642 1643 1644
/*
 * 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;
1645
	int ret = 0, i = 0, nbuf;
1646 1647 1648 1649 1650 1651

	/*
	 * 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).
	 */
1652
	inode_double_lock(ipipe->inode, opipe->inode);
1653

1654
	do {
1655 1656 1657 1658 1659 1660 1661
		if (!opipe->readers) {
			send_sig(SIGPIPE, current, 0);
			if (!ret)
				ret = -EPIPE;
			break;
		}

1662 1663 1664 1665 1666 1667
		/*
		 * If we have iterated all input buffers or ran out of
		 * output room, break.
		 */
		if (i >= ipipe->nrbufs || opipe->nrbufs >= PIPE_BUFFERS)
			break;
1668

1669 1670
		ibuf = ipipe->bufs + ((ipipe->curbuf + i) & (PIPE_BUFFERS - 1));
		nbuf = (opipe->curbuf + opipe->nrbufs) & (PIPE_BUFFERS - 1);
1671 1672

		/*
1673 1674
		 * Get a reference to this pipe buffer,
		 * so we can copy the contents over.
1675
		 */
1676 1677 1678 1679 1680
		ibuf->ops->get(ipipe, ibuf);

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

1681
		/*
1682 1683
		 * Don't inherit the gift flag, we need to
		 * prevent multiple steals of this page.
1684
		 */
1685
		obuf->flags &= ~PIPE_BUF_FLAG_GIFT;
1686

1687 1688
		if (obuf->len > len)
			obuf->len = len;
1689

1690 1691 1692 1693 1694
		opipe->nrbufs++;
		ret += obuf->len;
		len -= obuf->len;
		i++;
	} while (len);
1695

1696
	inode_double_unlock(ipipe->inode, opipe->inode);
1697

1698 1699 1700 1701
	/*
	 * If we put data in the output pipe, wakeup any potential readers.
	 */
	if (ret > 0) {
1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719
		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)
{
1720 1721
	struct pipe_inode_info *ipipe = pipe_info(in->f_path.dentry->d_inode);
	struct pipe_inode_info *opipe = pipe_info(out->f_path.dentry->d_inode);
1722
	int ret = -EINVAL;
1723 1724

	/*
1725 1726
	 * Duplicate the contents of ipipe to opipe without actually
	 * copying the data.
1727
	 */
1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742
	if (ipipe && opipe && ipipe != opipe) {
		/*
		 * Keep going, unless we encounter an error. The ipipe/opipe
		 * ordering doesn't really matter.
		 */
		ret = link_ipipe_prep(ipipe, flags);
		if (!ret) {
			ret = link_opipe_prep(opipe, flags);
			if (!ret) {
				ret = link_pipe(ipipe, opipe, len, flags);
				if (!ret && (flags & SPLICE_F_NONBLOCK))
					ret = -EAGAIN;
			}
		}
	}
1743

1744
	return ret;
1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772
}

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;
}