splice.c 40.2 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

32 33 34 35 36 37
/*
 * 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.
 */
38
static int page_cache_pipe_buf_steal(struct pipe_inode_info *pipe,
39 40 41
				     struct pipe_buffer *buf)
{
	struct page *page = buf->page;
42
	struct address_space *mapping;
43

44 45
	lock_page(page);

46 47 48
	mapping = page_mapping(page);
	if (mapping) {
		WARN_ON(!PageUptodate(page));
49

50 51 52 53 54 55 56 57 58
		/*
		 * 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);
59

60
		if (PagePrivate(page))
61
			try_to_release_page(page, GFP_KERNEL);
62

63 64 65 66 67 68 69 70
		/*
		 * 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;
		}
71
	}
72

73 74 75 76 77 78
	/*
	 * Raced with truncate or failed to remove page from current
	 * address space, unlock and return failure.
	 */
	unlock_page(page);
	return 1;
79 80
}

81
static void page_cache_pipe_buf_release(struct pipe_inode_info *pipe,
82 83 84
					struct pipe_buffer *buf)
{
	page_cache_release(buf->page);
J
Jens Axboe 已提交
85
	buf->flags &= ~PIPE_BUF_FLAG_LRU;
86 87
}

88 89 90 91
/*
 * Check whether the contents of buf is OK to access. Since the content
 * is a page cache page, IO may be in flight.
 */
92 93
static int page_cache_pipe_buf_confirm(struct pipe_inode_info *pipe,
				       struct pipe_buffer *buf)
94 95
{
	struct page *page = buf->page;
96
	int err;
97 98

	if (!PageUptodate(page)) {
99 100 101 102
		lock_page(page);

		/*
		 * Page got truncated/unhashed. This will cause a 0-byte
I
Ingo Molnar 已提交
103
		 * splice, if this is the first page.
104 105 106 107 108
		 */
		if (!page->mapping) {
			err = -ENODATA;
			goto error;
		}
109

110
		/*
I
Ingo Molnar 已提交
111
		 * Uh oh, read-error from disk.
112 113 114 115 116 117 118
		 */
		if (!PageUptodate(page)) {
			err = -EIO;
			goto error;
		}

		/*
119
		 * Page is ok afterall, we are done.
120
		 */
121 122 123
		unlock_page(page);
	}

124
	return 0;
125 126
error:
	unlock_page(page);
127
	return err;
128 129
}

130
static const struct pipe_buf_operations page_cache_pipe_buf_ops = {
131
	.can_merge = 0,
132 133
	.map = generic_pipe_buf_map,
	.unmap = generic_pipe_buf_unmap,
134
	.confirm = page_cache_pipe_buf_confirm,
135
	.release = page_cache_pipe_buf_release,
136
	.steal = page_cache_pipe_buf_steal,
137
	.get = generic_pipe_buf_get,
138 139
};

140 141 142
static int user_page_pipe_buf_steal(struct pipe_inode_info *pipe,
				    struct pipe_buffer *buf)
{
143 144 145
	if (!(buf->flags & PIPE_BUF_FLAG_GIFT))
		return 1;

J
Jens Axboe 已提交
146
	buf->flags |= PIPE_BUF_FLAG_LRU;
147
	return generic_pipe_buf_steal(pipe, buf);
148 149
}

150
static const struct pipe_buf_operations user_page_pipe_buf_ops = {
151
	.can_merge = 0,
152 153
	.map = generic_pipe_buf_map,
	.unmap = generic_pipe_buf_unmap,
154
	.confirm = generic_pipe_buf_confirm,
155 156
	.release = page_cache_pipe_buf_release,
	.steal = user_page_pipe_buf_steal,
157
	.get = generic_pipe_buf_get,
158 159
};

160 161 162 163 164 165 166 167 168 169
/**
 * 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.
 *
170
 */
171 172
ssize_t splice_to_pipe(struct pipe_inode_info *pipe,
		       struct splice_pipe_desc *spd)
173
{
174
	unsigned int spd_pages = spd->nr_pages;
175
	int ret, do_wakeup, page_nr;
176 177 178

	ret = 0;
	do_wakeup = 0;
179
	page_nr = 0;
180

181 182
	if (pipe->inode)
		mutex_lock(&pipe->inode->i_mutex);
183 184

	for (;;) {
185
		if (!pipe->readers) {
186 187 188 189 190 191
			send_sig(SIGPIPE, current, 0);
			if (!ret)
				ret = -EPIPE;
			break;
		}

192 193
		if (pipe->nrbufs < PIPE_BUFFERS) {
			int newbuf = (pipe->curbuf + pipe->nrbufs) & (PIPE_BUFFERS - 1);
194
			struct pipe_buffer *buf = pipe->bufs + newbuf;
195

196 197 198
			buf->page = spd->pages[page_nr];
			buf->offset = spd->partial[page_nr].offset;
			buf->len = spd->partial[page_nr].len;
199
			buf->private = spd->partial[page_nr].private;
200
			buf->ops = spd->ops;
201 202 203
			if (spd->flags & SPLICE_F_GIFT)
				buf->flags |= PIPE_BUF_FLAG_GIFT;

204
			pipe->nrbufs++;
205 206 207
			page_nr++;
			ret += buf->len;

208 209
			if (pipe->inode)
				do_wakeup = 1;
210

211
			if (!--spd->nr_pages)
212
				break;
213
			if (pipe->nrbufs < PIPE_BUFFERS)
214 215 216 217 218
				continue;

			break;
		}

219
		if (spd->flags & SPLICE_F_NONBLOCK) {
220 221 222 223 224
			if (!ret)
				ret = -EAGAIN;
			break;
		}

225 226 227 228 229 230 231
		if (signal_pending(current)) {
			if (!ret)
				ret = -ERESTARTSYS;
			break;
		}

		if (do_wakeup) {
232
			smp_mb();
233 234 235
			if (waitqueue_active(&pipe->wait))
				wake_up_interruptible_sync(&pipe->wait);
			kill_fasync(&pipe->fasync_readers, SIGIO, POLL_IN);
236 237 238
			do_wakeup = 0;
		}

239 240 241
		pipe->waiting_writers++;
		pipe_wait(pipe);
		pipe->waiting_writers--;
242 243
	}

244
	if (pipe->inode) {
245
		mutex_unlock(&pipe->inode->i_mutex);
246

247 248 249 250 251 252
		if (do_wakeup) {
			smp_mb();
			if (waitqueue_active(&pipe->wait))
				wake_up_interruptible(&pipe->wait);
			kill_fasync(&pipe->fasync_readers, SIGIO, POLL_IN);
		}
253 254
	}

255
	while (page_nr < spd_pages)
256
		page_cache_release(spd->pages[page_nr++]);
257 258 259 260

	return ret;
}

261
static int
262 263 264
__generic_file_splice_read(struct file *in, loff_t *ppos,
			   struct pipe_inode_info *pipe, size_t len,
			   unsigned int flags)
265 266
{
	struct address_space *mapping = in->f_mapping;
267
	unsigned int loff, nr_pages;
268
	struct page *pages[PIPE_BUFFERS];
269
	struct partial_page partial[PIPE_BUFFERS];
270
	struct page *page;
271 272
	pgoff_t index, end_index;
	loff_t isize;
273
	int error, page_nr;
274 275 276 277 278 279
	struct splice_pipe_desc spd = {
		.pages = pages,
		.partial = partial,
		.flags = flags,
		.ops = &page_cache_pipe_buf_ops,
	};
280

281
	index = *ppos >> PAGE_CACHE_SHIFT;
282 283
	loff = *ppos & ~PAGE_CACHE_MASK;
	nr_pages = (len + loff + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
284 285 286 287 288

	if (nr_pages > PIPE_BUFFERS)
		nr_pages = PIPE_BUFFERS;

	/*
289 290
	 * Don't try to 2nd guess the read-ahead logic, call into
	 * page_cache_readahead() like the page cache reads would do.
291
	 */
292
	page_cache_readahead(mapping, &in->f_ra, in, index, nr_pages);
293

294 295 296 297
	/*
	 * Lookup the (hopefully) full range of pages we need.
	 */
	spd.nr_pages = find_get_pages_contig(mapping, index, nr_pages, pages);
298

299 300
	/*
	 * If find_get_pages_contig() returned fewer pages than we needed,
301
	 * allocate the rest and fill in the holes.
302
	 */
303
	error = 0;
304 305
	index += spd.nr_pages;
	while (spd.nr_pages < nr_pages) {
306
		/*
307 308
		 * Page could be there, find_get_pages_contig() breaks on
		 * the first hole.
309
		 */
310 311
		page = find_get_page(mapping, index);
		if (!page) {
312 313 314 315 316 317
			/*
			 * Make sure the read-ahead engine is notified
			 * about this failure.
			 */
			handle_ra_miss(mapping, &in->f_ra, index);

318
			/*
319
			 * page didn't exist, allocate one.
320 321 322 323 324 325
			 */
			page = page_cache_alloc_cold(mapping);
			if (!page)
				break;

			error = add_to_page_cache_lru(page, mapping, index,
326
					      GFP_KERNEL);
327 328
			if (unlikely(error)) {
				page_cache_release(page);
329 330
				if (error == -EEXIST)
					continue;
331 332
				break;
			}
333 334 335 336 337
			/*
			 * add_to_page_cache() locks the page, unlock it
			 * to avoid convoluting the logic below even more.
			 */
			unlock_page(page);
338 339
		}

340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362
		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];

363 364 365 366
		/*
		 * If the page isn't uptodate, we may need to start io on it
		 */
		if (!PageUptodate(page)) {
367 368 369 370
			/*
			 * If in nonblock mode then dont block on waiting
			 * for an in-flight io page
			 */
371 372 373 374 375
			if (flags & SPLICE_F_NONBLOCK) {
				if (TestSetPageLocked(page))
					break;
			} else
				lock_page(page);
376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392

			/*
			 * 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;
			}
393

394 395 396 397
			/*
			 * need to read in the page
			 */
			error = mapping->a_ops->readpage(in, page);
398
			if (unlikely(error)) {
399 400 401 402 403 404
				/*
				 * 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.
				 */
405
				if (error == AOP_TRUNCATED_PAGE)
406 407
					error = 0;

408 409
				break;
			}
410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425
		}
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;
426 427

			/*
428
			 * max good bytes in this page
429
			 */
430 431
			plen = ((isize - 1) & ~PAGE_CACHE_MASK) + 1;
			if (plen <= loff)
432 433 434
				break;

			/*
435
			 * force quit after adding this page
436
			 */
437 438
			this_len = min(this_len, plen - loff);
			len = this_len;
439
		}
440

441 442
		partial[page_nr].offset = loff;
		partial[page_nr].len = this_len;
443
		len -= this_len;
444
		loff = 0;
445 446
		spd.nr_pages++;
		index++;
447 448
	}

449
	/*
450
	 * Release any pages at the end, if we quit early. 'page_nr' is how far
451 452 453 454 455
	 * we got, 'nr_pages' is how many pages are in the map.
	 */
	while (page_nr < nr_pages)
		page_cache_release(pages[page_nr++]);

456
	if (spd.nr_pages)
457
		return splice_to_pipe(pipe, &spd);
458

459
	return error;
460 461
}

462 463 464
/**
 * generic_file_splice_read - splice data from file to a pipe
 * @in:		file to splice from
465
 * @ppos:	position in @in
466 467 468 469
 * @pipe:	pipe to splice to
 * @len:	number of bytes to splice
 * @flags:	splice modifier flags
 *
470 471 472 473 474
 * 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.
 *
475
 */
476 477 478
ssize_t generic_file_splice_read(struct file *in, loff_t *ppos,
				 struct pipe_inode_info *pipe, size_t len,
				 unsigned int flags)
479 480 481
{
	ssize_t spliced;
	int ret;
482 483 484 485 486 487 488 489 490
	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;
491 492 493 494

	ret = 0;
	spliced = 0;
	while (len) {
495
		ret = __generic_file_splice_read(in, ppos, pipe, len, flags);
496

497
		if (ret < 0)
498
			break;
499 500 501 502 503 504 505 506
		else if (!ret) {
			if (spliced)
				break;
			if (flags & SPLICE_F_NONBLOCK) {
				ret = -EAGAIN;
				break;
			}
		}
507

508
		*ppos += ret;
509 510 511 512 513 514 515 516 517 518
		len -= ret;
		spliced += ret;
	}

	if (spliced)
		return spliced;

	return ret;
}

519 520
EXPORT_SYMBOL(generic_file_splice_read);

521
/*
522
 * Send 'sd->len' bytes to socket from 'sd->file' at position 'sd->pos'
J
Jens Axboe 已提交
523
 * using sendpage(). Return the number of bytes sent.
524
 */
525
static int pipe_to_sendpage(struct pipe_inode_info *pipe,
526 527
			    struct pipe_buffer *buf, struct splice_desc *sd)
{
528
	struct file *file = sd->u.file;
529
	loff_t pos = sd->pos;
530
	int ret, more;
531

532
	ret = buf->ops->confirm(pipe, buf);
533 534
	if (!ret) {
		more = (sd->flags & SPLICE_F_MORE) || sd->len < sd->total_len;
535

536 537 538
		ret = file->f_op->sendpage(file, buf->page, buf->offset,
					   sd->len, &pos, more);
	}
539

J
Jens Axboe 已提交
540
	return ret;
541 542 543 544 545 546 547 548 549 550 551 552 553 554 555
}

/*
 * 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.
 *
556 557 558 559 560 561
 * 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.
562
 */
563
static int pipe_to_file(struct pipe_inode_info *pipe, struct pipe_buffer *buf,
564 565
			struct splice_desc *sd)
{
566
	struct file *file = sd->u.file;
567
	struct address_space *mapping = file->f_mapping;
J
Jens Axboe 已提交
568
	unsigned int offset, this_len;
569 570
	struct page *page;
	pgoff_t index;
571
	int ret;
572 573

	/*
574
	 * make sure the data in this buffer is uptodate
575
	 */
576
	ret = buf->ops->confirm(pipe, buf);
577 578
	if (unlikely(ret))
		return ret;
579 580 581 582

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

J
Jens Axboe 已提交
583 584 585 586
	this_len = sd->len;
	if (this_len + offset > PAGE_CACHE_SIZE)
		this_len = PAGE_CACHE_SIZE - offset;

N
Nick Piggin 已提交
587 588 589 590 591 592 593 594
find_page:
	page = find_lock_page(mapping, index);
	if (!page) {
		ret = -ENOMEM;
		page = page_cache_alloc_cold(mapping);
		if (unlikely(!page))
			goto out_ret;

595
		/*
N
Nick Piggin 已提交
596
		 * This will also lock the page
597
		 */
N
Nick Piggin 已提交
598 599 600 601 602
		ret = add_to_page_cache_lru(page, mapping, index,
					    GFP_KERNEL);
		if (unlikely(ret))
			goto out;
	}
603

J
Jens Axboe 已提交
604
	ret = mapping->a_ops->prepare_write(file, page, offset, offset+this_len);
605 606 607 608 609
	if (unlikely(ret)) {
		loff_t isize = i_size_read(mapping->host);

		if (ret != AOP_TRUNCATED_PAGE)
			unlock_page(page);
610
		page_cache_release(page);
611 612 613 614 615 616 617 618 619 620
		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);

621
		goto out_ret;
622
	}
623

624
	if (buf->page != page) {
625 626 627
		/*
		 * Careful, ->map() uses KM_USER0!
		 */
628
		char *src = buf->ops->map(pipe, buf, 1);
629
		char *dst = kmap_atomic(page, KM_USER1);
630

J
Jens Axboe 已提交
631
		memcpy(dst + offset, src + buf->offset, this_len);
632
		flush_dcache_page(page);
633
		kunmap_atomic(dst, KM_USER1);
634
		buf->ops->unmap(pipe, buf, src);
635
	}
636

J
Jens Axboe 已提交
637
	ret = mapping->a_ops->commit_write(file, page, offset, offset+this_len);
638 639 640 641 642 643 644
	if (ret) {
		if (ret == AOP_TRUNCATED_PAGE) {
			page_cache_release(page);
			goto find_page;
		}
		if (ret < 0)
			goto out;
645
		/*
646 647
		 * Partial write has happened, so 'ret' already initialized by
		 * number of bytes written, Where is nothing we have to do here.
648
		 */
649
	} else
650
		ret = this_len;
651 652 653 654 655
	/*
	 * Return the number of bytes written and mark page as
	 * accessed, we are now done!
	 */
	mark_page_accessed(page);
656
out:
657
	page_cache_release(page);
658
	unlock_page(page);
659
out_ret:
660 661 662
	return ret;
}

663 664 665 666 667 668 669 670 671 672 673 674
/**
 * __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.
 *
675
 */
J
Jens Axboe 已提交
676 677
ssize_t __splice_from_pipe(struct pipe_inode_info *pipe, struct splice_desc *sd,
			   splice_actor *actor)
678 679 680 681 682 683 684
{
	int ret, do_wakeup, err;

	ret = 0;
	do_wakeup = 0;

	for (;;) {
685 686
		if (pipe->nrbufs) {
			struct pipe_buffer *buf = pipe->bufs + pipe->curbuf;
687
			const struct pipe_buf_operations *ops = buf->ops;
688

J
Jens Axboe 已提交
689 690 691
			sd->len = buf->len;
			if (sd->len > sd->total_len)
				sd->len = sd->total_len;
692

J
Jens Axboe 已提交
693
			err = actor(pipe, buf, sd);
J
Jens Axboe 已提交
694
			if (err <= 0) {
695 696 697 698 699 700
				if (!ret && err != -ENODATA)
					ret = err;

				break;
			}

J
Jens Axboe 已提交
701 702 703 704
			ret += err;
			buf->offset += err;
			buf->len -= err;

J
Jens Axboe 已提交
705 706 707 708
			sd->len -= err;
			sd->pos += err;
			sd->total_len -= err;
			if (sd->len)
J
Jens Axboe 已提交
709
				continue;
I
Ingo Molnar 已提交
710

711 712
			if (!buf->len) {
				buf->ops = NULL;
713
				ops->release(pipe, buf);
714 715 716 717
				pipe->curbuf = (pipe->curbuf + 1) & (PIPE_BUFFERS - 1);
				pipe->nrbufs--;
				if (pipe->inode)
					do_wakeup = 1;
718 719
			}

J
Jens Axboe 已提交
720
			if (!sd->total_len)
721 722 723
				break;
		}

724
		if (pipe->nrbufs)
725
			continue;
726
		if (!pipe->writers)
727
			break;
728
		if (!pipe->waiting_writers) {
729 730 731 732
			if (ret)
				break;
		}

J
Jens Axboe 已提交
733
		if (sd->flags & SPLICE_F_NONBLOCK) {
734 735 736 737 738
			if (!ret)
				ret = -EAGAIN;
			break;
		}

739 740 741 742 743 744 745
		if (signal_pending(current)) {
			if (!ret)
				ret = -ERESTARTSYS;
			break;
		}

		if (do_wakeup) {
746
			smp_mb();
747 748 749
			if (waitqueue_active(&pipe->wait))
				wake_up_interruptible_sync(&pipe->wait);
			kill_fasync(&pipe->fasync_writers, SIGIO, POLL_OUT);
750 751 752
			do_wakeup = 0;
		}

753
		pipe_wait(pipe);
754 755 756
	}

	if (do_wakeup) {
757
		smp_mb();
758 759 760
		if (waitqueue_active(&pipe->wait))
			wake_up_interruptible(&pipe->wait);
		kill_fasync(&pipe->fasync_writers, SIGIO, POLL_OUT);
761 762 763 764
	}

	return ret;
}
M
Mark Fasheh 已提交
765
EXPORT_SYMBOL(__splice_from_pipe);
766

767 768 769 770 771 772 773 774 775 776 777 778 779 780
/**
 * 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().
 *
 */
781 782 783 784 785 786
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
787 788 789 790
	struct splice_desc sd = {
		.total_len = len,
		.flags = flags,
		.pos = *ppos,
791
		.u.file = out,
J
Jens Axboe 已提交
792
	};
793 794 795 796 797 798 799 800

	/*
	 * 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
801
	ret = __splice_from_pipe(pipe, &sd, actor);
802 803 804 805 806 807 808 809 810
	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
811
 * @ppos:	position in @out
812 813 814
 * @len:	number of bytes to splice
 * @flags:	splice modifier flags
 *
815 816 817 818
 * 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.
819 820 821 822 823 824 825 826
 *
 */
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
827 828 829 830
	struct splice_desc sd = {
		.total_len = len,
		.flags = flags,
		.pos = *ppos,
831
		.u.file = out,
J
Jens Axboe 已提交
832
	};
833 834 835
	ssize_t ret;
	int err;

836
	err = remove_suid(out->f_path.dentry);
837 838 839
	if (unlikely(err))
		return err;

J
Jens Axboe 已提交
840
	ret = __splice_from_pipe(pipe, &sd, pipe_to_file);
841
	if (ret > 0) {
842 843
		unsigned long nr_pages;

844
		*ppos += ret;
845
		nr_pages = (ret + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
846 847 848 849 850 851 852 853 854 855 856 857

		/*
		 * 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;
		}
858
		balance_dirty_pages_ratelimited_nr(mapping, nr_pages);
859 860 861 862 863 864 865
	}

	return ret;
}

EXPORT_SYMBOL(generic_file_splice_write_nolock);

866 867
/**
 * generic_file_splice_write - splice data from a pipe to a file
868
 * @pipe:	pipe info
869
 * @out:	file to write to
870
 * @ppos:	position in @out
871 872 873
 * @len:	number of bytes to splice
 * @flags:	splice modifier flags
 *
874 875 876
 * Description:
 *    Will either move or copy pages (determined by @flags options) from
 *    the given pipe inode to the given file.
877 878
 *
 */
879 880
ssize_t
generic_file_splice_write(struct pipe_inode_info *pipe, struct file *out,
881
			  loff_t *ppos, size_t len, unsigned int flags)
882
{
883
	struct address_space *mapping = out->f_mapping;
884
	struct inode *inode = mapping->host;
885
	ssize_t ret;
886 887
	int err;

888
	err = should_remove_suid(out->f_path.dentry);
889 890
	if (unlikely(err)) {
		mutex_lock(&inode->i_mutex);
891
		err = __remove_suid(out->f_path.dentry, err);
892 893 894 895
		mutex_unlock(&inode->i_mutex);
		if (err)
			return err;
	}
896

897
	ret = splice_from_pipe(pipe, out, ppos, len, flags, pipe_to_file);
J
Jens Axboe 已提交
898
	if (ret > 0) {
899 900
		unsigned long nr_pages;

J
Jens Axboe 已提交
901
		*ppos += ret;
902
		nr_pages = (ret + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
J
Jens Axboe 已提交
903 904 905 906 907 908 909 910 911 912

		/*
		 * 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);
913

J
Jens Axboe 已提交
914 915 916
			if (err)
				ret = err;
		}
917
		balance_dirty_pages_ratelimited_nr(mapping, nr_pages);
918 919 920
	}

	return ret;
921 922
}

923 924
EXPORT_SYMBOL(generic_file_splice_write);

925 926
/**
 * generic_splice_sendpage - splice data from a pipe to a socket
927
 * @pipe:	pipe to splice from
928
 * @out:	socket to write to
929
 * @ppos:	position in @out
930 931 932
 * @len:	number of bytes to splice
 * @flags:	splice modifier flags
 *
933 934 935
 * Description:
 *    Will send @len bytes from the pipe to a network socket. No data copying
 *    is involved.
936 937
 *
 */
938
ssize_t generic_splice_sendpage(struct pipe_inode_info *pipe, struct file *out,
939
				loff_t *ppos, size_t len, unsigned int flags)
940
{
941
	return splice_from_pipe(pipe, out, ppos, len, flags, pipe_to_sendpage);
942 943
}

944
EXPORT_SYMBOL(generic_splice_sendpage);
J
Jeff Garzik 已提交
945

946 947 948
/*
 * Attempt to initiate a splice from pipe to file.
 */
949
static long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
950
			   loff_t *ppos, size_t len, unsigned int flags)
951 952 953
{
	int ret;

954
	if (unlikely(!out->f_op || !out->f_op->splice_write))
955 956
		return -EINVAL;

957
	if (unlikely(!(out->f_mode & FMODE_WRITE)))
958 959
		return -EBADF;

960
	ret = rw_verify_area(WRITE, out, ppos, len);
961 962 963
	if (unlikely(ret < 0))
		return ret;

964
	return out->f_op->splice_write(pipe, out, ppos, len, flags);
965 966
}

967 968 969
/*
 * Attempt to initiate a splice from a file to a pipe.
 */
970 971 972
static long do_splice_to(struct file *in, loff_t *ppos,
			 struct pipe_inode_info *pipe, size_t len,
			 unsigned int flags)
973 974 975
{
	int ret;

976
	if (unlikely(!in->f_op || !in->f_op->splice_read))
977 978
		return -EINVAL;

979
	if (unlikely(!(in->f_mode & FMODE_READ)))
980 981
		return -EBADF;

982
	ret = rw_verify_area(READ, in, ppos, len);
983 984 985
	if (unlikely(ret < 0))
		return ret;

986
	return in->f_op->splice_read(in, ppos, pipe, len, flags);
987 988
}

989 990 991 992 993 994 995 996 997 998 999 1000
/**
 * 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
1001 1002 1003
 */
ssize_t splice_direct_to_actor(struct file *in, struct splice_desc *sd,
			       splice_direct_actor *actor)
1004 1005 1006 1007
{
	struct pipe_inode_info *pipe;
	long ret, bytes;
	umode_t i_mode;
J
Jens Axboe 已提交
1008 1009
	size_t len;
	int i, flags;
1010 1011 1012 1013 1014 1015

	/*
	 * 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!
	 */
1016
	i_mode = in->f_path.dentry->d_inode->i_mode;
1017 1018 1019 1020 1021 1022 1023 1024
	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;
1025
	if (unlikely(!pipe)) {
1026 1027 1028 1029 1030 1031
		pipe = alloc_pipe_info(NULL);
		if (!pipe)
			return -ENOMEM;

		/*
		 * We don't have an immediate reader, but we'll read the stuff
1032
		 * out of the pipe right after the splice_to_pipe(). So set
1033 1034 1035 1036 1037 1038 1039 1040
		 * PIPE_READERS appropriately.
		 */
		pipe->readers = 1;

		current->splice_pipe = pipe;
	}

	/*
I
Ingo Molnar 已提交
1041
	 * Do the splice.
1042 1043 1044
	 */
	ret = 0;
	bytes = 0;
J
Jens Axboe 已提交
1045 1046 1047 1048 1049 1050 1051
	len = sd->total_len;
	flags = sd->flags;

	/*
	 * Don't block on output, we have to drain the direct pipe.
	 */
	sd->flags &= ~SPLICE_F_NONBLOCK;
1052 1053 1054 1055 1056 1057 1058 1059 1060

	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));

J
Jens Axboe 已提交
1061
		ret = do_splice_to(in, &sd->pos, pipe, max_read_len, flags);
1062 1063 1064 1065
		if (unlikely(ret < 0))
			goto out_release;

		read_len = ret;
J
Jens Axboe 已提交
1066
		sd->total_len = read_len;
1067 1068 1069 1070 1071 1072

		/*
		 * 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
1073
		ret = actor(pipe, sd);
1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 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 1111 1112 1113 1114 1115 1116 1117
		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;
J
Jens Axboe 已提交
1118 1119 1120 1121 1122 1123 1124

}
EXPORT_SYMBOL(splice_direct_to_actor);

static int direct_splice_actor(struct pipe_inode_info *pipe,
			       struct splice_desc *sd)
{
1125
	struct file *file = sd->u.file;
J
Jens Axboe 已提交
1126 1127 1128 1129

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

1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144
/**
 * 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
1145 1146 1147 1148 1149 1150 1151 1152
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,
1153
		.u.file		= out,
J
Jens Axboe 已提交
1154 1155 1156 1157 1158 1159
	};
	size_t ret;

	ret = splice_direct_to_actor(in, &sd, direct_splice_actor);
	*ppos = sd.pos;
	return ret;
1160 1161
}

1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174
/*
 * 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;
}

1175 1176 1177
/*
 * Determine where to splice to/from.
 */
1178 1179 1180
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)
1181
{
1182
	struct pipe_inode_info *pipe;
1183
	loff_t offset, *off;
J
Jens Axboe 已提交
1184
	long ret;
1185

1186
	pipe = pipe_info(in->f_path.dentry->d_inode);
1187 1188 1189
	if (pipe) {
		if (off_in)
			return -ESPIPE;
1190 1191 1192
		if (off_out) {
			if (out->f_op->llseek == no_llseek)
				return -EINVAL;
1193
			if (copy_from_user(&offset, off_out, sizeof(loff_t)))
1194
				return -EFAULT;
1195 1196 1197
			off = &offset;
		} else
			off = &out->f_pos;
1198

J
Jens Axboe 已提交
1199 1200 1201 1202 1203 1204
		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;
1205
	}
1206

1207
	pipe = pipe_info(out->f_path.dentry->d_inode);
1208 1209 1210
	if (pipe) {
		if (off_out)
			return -ESPIPE;
1211 1212 1213
		if (off_in) {
			if (in->f_op->llseek == no_llseek)
				return -EINVAL;
1214
			if (copy_from_user(&offset, off_in, sizeof(loff_t)))
1215
				return -EFAULT;
1216 1217 1218
			off = &offset;
		} else
			off = &in->f_pos;
1219

J
Jens Axboe 已提交
1220 1221 1222 1223 1224 1225
		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;
1226
	}
1227 1228 1229 1230

	return -EINVAL;
}

1231 1232 1233 1234 1235 1236 1237 1238 1239
/*
 * 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,
1240
				struct partial_page *partial, int aligned)
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 1275 1276 1277 1278 1279
{
	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;
1280 1281 1282 1283 1284 1285 1286 1287 1288

		/*
		 * 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;

1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303
		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++) {
1304
			const int plen = min_t(size_t, len, PAGE_SIZE - off);
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 1337 1338 1339 1340 1341

			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;
}

1342 1343 1344 1345 1346 1347
static int pipe_to_user(struct pipe_inode_info *pipe, struct pipe_buffer *buf,
			struct splice_desc *sd)
{
	char *src;
	int ret;

1348
	ret = buf->ops->confirm(pipe, buf);
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 1455 1456 1457 1458 1459
	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;
}

1460 1461 1462 1463 1464
/*
 * 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.
 */
1465 1466
static long vmsplice_to_pipe(struct file *file, const struct iovec __user *iov,
			     unsigned long nr_segs, unsigned int flags)
1467
{
1468
	struct pipe_inode_info *pipe;
1469 1470 1471 1472 1473 1474 1475 1476 1477
	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,
	};

1478
	pipe = pipe_info(file->f_path.dentry->d_inode);
1479
	if (!pipe)
1480 1481
		return -EBADF;

1482 1483
	spd.nr_pages = get_iovec_page_array(iov, nr_segs, pages, partial,
					    flags & SPLICE_F_GIFT);
1484 1485 1486
	if (spd.nr_pages <= 0)
		return spd.nr_pages;

1487
	return splice_to_pipe(pipe, &spd);
1488 1489
}

1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505
/*
 * 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().
 *
 */
1506 1507 1508 1509 1510 1511 1512
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;

1513 1514 1515 1516 1517
	if (unlikely(nr_segs > UIO_MAXIOV))
		return -EINVAL;
	else if (unlikely(!nr_segs))
		return 0;

1518 1519 1520 1521
	error = -EBADF;
	file = fget_light(fd, &fput);
	if (file) {
		if (file->f_mode & FMODE_WRITE)
1522 1523 1524
			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);
1525 1526 1527 1528 1529 1530 1531

		fput_light(file, fput);
	}

	return error;
}

1532 1533 1534
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)
1535 1536 1537 1538 1539 1540 1541 1542 1543
{
	long error;
	struct file *in, *out;
	int fput_in, fput_out;

	if (unlikely(!len))
		return 0;

	error = -EBADF;
1544
	in = fget_light(fd_in, &fput_in);
1545 1546
	if (in) {
		if (in->f_mode & FMODE_READ) {
1547
			out = fget_light(fd_out, &fput_out);
1548 1549
			if (out) {
				if (out->f_mode & FMODE_WRITE)
1550 1551 1552
					error = do_splice(in, off_in,
							  out, off_out,
							  len, flags);
1553 1554 1555 1556 1557 1558 1559 1560 1561
				fput_light(out, fput_out);
			}
		}

		fput_light(in, fput_in);
	}

	return error;
}
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 1637 1638 1639 1640 1641
/*
 * 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;
}

1642 1643 1644 1645 1646 1647 1648 1649
/*
 * 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;
1650
	int ret = 0, i = 0, nbuf;
1651 1652 1653 1654 1655 1656

	/*
	 * 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).
	 */
1657
	inode_double_lock(ipipe->inode, opipe->inode);
1658

1659
	do {
1660 1661 1662 1663 1664 1665 1666
		if (!opipe->readers) {
			send_sig(SIGPIPE, current, 0);
			if (!ret)
				ret = -EPIPE;
			break;
		}

1667 1668 1669 1670 1671 1672
		/*
		 * If we have iterated all input buffers or ran out of
		 * output room, break.
		 */
		if (i >= ipipe->nrbufs || opipe->nrbufs >= PIPE_BUFFERS)
			break;
1673

1674 1675
		ibuf = ipipe->bufs + ((ipipe->curbuf + i) & (PIPE_BUFFERS - 1));
		nbuf = (opipe->curbuf + opipe->nrbufs) & (PIPE_BUFFERS - 1);
1676 1677

		/*
1678 1679
		 * Get a reference to this pipe buffer,
		 * so we can copy the contents over.
1680
		 */
1681 1682 1683 1684 1685
		ibuf->ops->get(ipipe, ibuf);

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

1686
		/*
1687 1688
		 * Don't inherit the gift flag, we need to
		 * prevent multiple steals of this page.
1689
		 */
1690
		obuf->flags &= ~PIPE_BUF_FLAG_GIFT;
1691

1692 1693
		if (obuf->len > len)
			obuf->len = len;
1694

1695 1696 1697 1698 1699
		opipe->nrbufs++;
		ret += obuf->len;
		len -= obuf->len;
		i++;
	} while (len);
1700

1701
	inode_double_unlock(ipipe->inode, opipe->inode);
1702

1703 1704 1705 1706
	/*
	 * If we put data in the output pipe, wakeup any potential readers.
	 */
	if (ret > 0) {
1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724
		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)
{
1725 1726
	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);
1727
	int ret = -EINVAL;
1728 1729

	/*
1730 1731
	 * Duplicate the contents of ipipe to opipe without actually
	 * copying the data.
1732
	 */
1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747
	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;
			}
		}
	}
1748

1749
	return ret;
1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777
}

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;
}