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

38
#include "internal.h"
39

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

52 53
	lock_page(page);

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

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

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

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

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

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

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

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

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

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

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

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

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

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

J
Jens Axboe 已提交
154
	buf->flags |= PIPE_BUF_FLAG_LRU;
155
	return generic_pipe_buf_steal(pipe, buf);
156 157
}

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

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

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

191 192 193
	if (!spd_pages)
		return 0;

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

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

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

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

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

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

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

	return ret;
}
228
EXPORT_SYMBOL_GPL(splice_to_pipe);
229

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

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

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

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

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

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

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

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

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

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

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

	return ret;
}
326 327
EXPORT_SYMBOL(generic_file_splice_read);

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

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

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

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

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

	return res;
}

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

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

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

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

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

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

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

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

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

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

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

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

A
Al Viro 已提交
444
	if (!likely(file->f_op->sendpage))
445 446
		return -EINVAL;

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

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

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

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

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

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

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

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

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

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

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

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

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

527 528
	return 1;
}
529

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

549 550 551
	while (!pipe->nrbufs) {
		if (!pipe->writers)
			return 0;
J
Jens Axboe 已提交
552

553 554
		if (!pipe->waiting_writers && sd->num_spliced)
			return 0;
I
Ingo Molnar 已提交
555

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

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

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

567 568
		pipe_wait(pipe);
	}
569

570 571
	return 1;
}
572

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

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

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

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

	return sd->num_spliced ? sd->num_spliced : ret;
632
}
M
Mark Fasheh 已提交
633
EXPORT_SYMBOL(__splice_from_pipe);
634

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

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

	return ret;
}

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

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

	pipe_lock(pipe);

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

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

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

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

			if (this_len > left)
				this_len = left;

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

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

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

A
Al Viro 已提交
748 749
		iov_iter_bvec(&from, ITER_BVEC | WRITE, array, n,
			      sd.total_len - left);
750
		ret = vfs_iter_write(out, &from, &sd.pos, 0);
A
Al Viro 已提交
751 752 753 754 755
		if (ret <= 0)
			break;

		sd.num_spliced += ret;
		sd.total_len -= ret;
756
		*ppos = sd.pos;
A
Al Viro 已提交
757 758 759 760 761 762 763

		/* dismiss the fully eaten buffers, adjust the partial one */
		while (ret) {
			struct pipe_buffer *buf = pipe->bufs + pipe->curbuf;
			if (ret >= buf->len) {
				ret -= buf->len;
				buf->len = 0;
764
				pipe_buf_release(pipe, buf);
A
Al Viro 已提交
765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789
				pipe->curbuf = (pipe->curbuf + 1) & (pipe->buffers - 1);
				pipe->nrbufs--;
				if (pipe->files)
					sd.need_wakeup = true;
			} else {
				buf->offset += ret;
				buf->len -= ret;
				ret = 0;
			}
		}
	}
done:
	kfree(array);
	splice_from_pipe_end(pipe, &sd);

	pipe_unlock(pipe);

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

	return ret;
}

EXPORT_SYMBOL(iter_file_splice_write);

790 791
static int write_pipe_buf(struct pipe_inode_info *pipe, struct pipe_buffer *buf,
			  struct splice_desc *sd)
792
{
793 794
	int ret;
	void *data;
795
	loff_t tmp = sd->pos;
796

A
Al Viro 已提交
797
	data = kmap(buf->page);
798
	ret = __kernel_write(sd->u.file, data + buf->offset, sd->len, &tmp);
A
Al Viro 已提交
799
	kunmap(buf->page);
800 801

	return ret;
802 803 804 805 806 807
}

static ssize_t default_file_splice_write(struct pipe_inode_info *pipe,
					 struct file *out, loff_t *ppos,
					 size_t len, unsigned int flags)
{
808
	ssize_t ret;
809

810 811 812
	ret = splice_from_pipe(pipe, out, ppos, len, flags, write_pipe_buf);
	if (ret > 0)
		*ppos += ret;
813

814
	return ret;
815 816
}

817 818
/**
 * generic_splice_sendpage - splice data from a pipe to a socket
819
 * @pipe:	pipe to splice from
820
 * @out:	socket to write to
821
 * @ppos:	position in @out
822 823 824
 * @len:	number of bytes to splice
 * @flags:	splice modifier flags
 *
825 826 827
 * Description:
 *    Will send @len bytes from the pipe to a network socket. No data copying
 *    is involved.
828 829
 *
 */
830
ssize_t generic_splice_sendpage(struct pipe_inode_info *pipe, struct file *out,
831
				loff_t *ppos, size_t len, unsigned int flags)
832
{
833
	return splice_from_pipe(pipe, out, ppos, len, flags, pipe_to_sendpage);
834 835
}

836
EXPORT_SYMBOL(generic_splice_sendpage);
J
Jeff Garzik 已提交
837

838 839 840
/*
 * Attempt to initiate a splice from pipe to file.
 */
841
static long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
842
			   loff_t *ppos, size_t len, unsigned int flags)
843
{
844 845
	ssize_t (*splice_write)(struct pipe_inode_info *, struct file *,
				loff_t *, size_t, unsigned int);
846

A
Al Viro 已提交
847
	if (out->f_op->splice_write)
848 849
		splice_write = out->f_op->splice_write;
	else
850 851
		splice_write = default_file_splice_write;

852
	return splice_write(pipe, out, ppos, len, flags);
853 854
}

855 856 857
/*
 * Attempt to initiate a splice from a file to a pipe.
 */
858 859 860
static long do_splice_to(struct file *in, loff_t *ppos,
			 struct pipe_inode_info *pipe, size_t len,
			 unsigned int flags)
861
{
862 863
	ssize_t (*splice_read)(struct file *, loff_t *,
			       struct pipe_inode_info *, size_t, unsigned int);
864 865
	int ret;

866
	if (unlikely(!(in->f_mode & FMODE_READ)))
867 868
		return -EBADF;

869
	ret = rw_verify_area(READ, in, ppos, len);
870 871 872
	if (unlikely(ret < 0))
		return ret;

873 874 875
	if (unlikely(len > MAX_RW_COUNT))
		len = MAX_RW_COUNT;

A
Al Viro 已提交
876
	if (in->f_op->splice_read)
877 878
		splice_read = in->f_op->splice_read;
	else
879 880 881
		splice_read = default_file_splice_read;

	return splice_read(in, ppos, pipe, len, flags);
882 883
}

884 885 886 887 888 889 890 891 892
/**
 * 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
R
Randy Dunlap 已提交
893
 *    pipe is cached in the process, and reused during the lifetime of
894 895
 *    that process.
 *
J
Jens Axboe 已提交
896 897 898
 */
ssize_t splice_direct_to_actor(struct file *in, struct splice_desc *sd,
			       splice_direct_actor *actor)
899 900 901 902
{
	struct pipe_inode_info *pipe;
	long ret, bytes;
	umode_t i_mode;
J
Jens Axboe 已提交
903
	size_t len;
904
	int i, flags, more;
905 906 907 908 909 910

	/*
	 * 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!
	 */
A
Al Viro 已提交
911
	i_mode = file_inode(in)->i_mode;
912 913 914 915 916 917 918 919
	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;
920
	if (unlikely(!pipe)) {
A
Al Viro 已提交
921
		pipe = alloc_pipe_info();
922 923 924 925 926
		if (!pipe)
			return -ENOMEM;

		/*
		 * We don't have an immediate reader, but we'll read the stuff
927
		 * out of the pipe right after the splice_to_pipe(). So set
928 929 930 931 932 933 934 935
		 * PIPE_READERS appropriately.
		 */
		pipe->readers = 1;

		current->splice_pipe = pipe;
	}

	/*
I
Ingo Molnar 已提交
936
	 * Do the splice.
937 938 939
	 */
	ret = 0;
	bytes = 0;
J
Jens Axboe 已提交
940 941 942 943 944 945 946
	len = sd->total_len;
	flags = sd->flags;

	/*
	 * Don't block on output, we have to drain the direct pipe.
	 */
	sd->flags &= ~SPLICE_F_NONBLOCK;
947
	more = sd->flags & SPLICE_F_MORE;
948

949 950
	WARN_ON_ONCE(pipe->nrbufs != 0);

951
	while (len) {
952
		size_t read_len;
953
		loff_t pos = sd->pos, prev_pos = pos;
954

955 956 957 958
		/* Don't try to read more the pipe has space for. */
		read_len = min_t(size_t, len,
				 (pipe->buffers - pipe->nrbufs) << PAGE_SHIFT);
		ret = do_splice_to(in, &pos, pipe, read_len, flags);
959
		if (unlikely(ret <= 0))
960 961 962
			goto out_release;

		read_len = ret;
J
Jens Axboe 已提交
963
		sd->total_len = read_len;
964

965 966 967 968 969 970 971 972 973
		/*
		 * If more data is pending, set SPLICE_F_MORE
		 * If this is the last data and SPLICE_F_MORE was not set
		 * initially, clears it.
		 */
		if (read_len < len)
			sd->flags |= SPLICE_F_MORE;
		else if (!more)
			sd->flags &= ~SPLICE_F_MORE;
974 975 976 977 978
		/*
		 * 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
979
		ret = actor(pipe, sd);
980 981
		if (unlikely(ret <= 0)) {
			sd->pos = prev_pos;
982
			goto out_release;
983
		}
984 985 986

		bytes += ret;
		len -= ret;
987
		sd->pos = pos;
988

989 990
		if (ret < read_len) {
			sd->pos = prev_pos + ret;
991
			goto out_release;
992
		}
993 994
	}

995
done:
996
	pipe->nrbufs = pipe->curbuf = 0;
997
	file_accessed(in);
998 999 1000 1001 1002 1003 1004
	return bytes;

out_release:
	/*
	 * If we did an incomplete transfer we must release
	 * the pipe buffers in question:
	 */
1005
	for (i = 0; i < pipe->buffers; i++) {
1006 1007
		struct pipe_buffer *buf = pipe->bufs + i;

1008 1009
		if (buf->ops)
			pipe_buf_release(pipe, buf);
1010 1011
	}

1012 1013
	if (!bytes)
		bytes = ret;
J
Jens Axboe 已提交
1014

1015
	goto done;
J
Jens Axboe 已提交
1016 1017 1018 1019 1020 1021
}
EXPORT_SYMBOL(splice_direct_to_actor);

static int direct_splice_actor(struct pipe_inode_info *pipe,
			       struct splice_desc *sd)
{
1022
	struct file *file = sd->u.file;
J
Jens Axboe 已提交
1023

1024
	return do_splice_from(pipe, file, sd->opos, sd->total_len,
1025
			      sd->flags);
J
Jens Axboe 已提交
1026 1027
}

1028 1029 1030 1031 1032
/**
 * do_splice_direct - splices data directly between two files
 * @in:		file to splice from
 * @ppos:	input file offset
 * @out:	file to splice to
1033
 * @opos:	output file offset
1034 1035 1036 1037 1038 1039 1040 1041 1042 1043
 * @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
1044
long do_splice_direct(struct file *in, loff_t *ppos, struct file *out,
1045
		      loff_t *opos, size_t len, unsigned int flags)
J
Jens Axboe 已提交
1046 1047 1048 1049 1050 1051
{
	struct splice_desc sd = {
		.len		= len,
		.total_len	= len,
		.flags		= flags,
		.pos		= *ppos,
1052
		.u.file		= out,
1053
		.opos		= opos,
J
Jens Axboe 已提交
1054
	};
1055
	long ret;
J
Jens Axboe 已提交
1056

1057 1058 1059 1060 1061 1062 1063 1064 1065 1066
	if (unlikely(!(out->f_mode & FMODE_WRITE)))
		return -EBADF;

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

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

J
Jens Axboe 已提交
1067
	ret = splice_direct_to_actor(in, &sd, direct_splice_actor);
1068
	if (ret > 0)
1069
		*ppos = sd.pos;
1070

J
Jens Axboe 已提交
1071
	return ret;
1072
}
1073
EXPORT_SYMBOL(do_splice_direct);
1074

1075 1076
static int wait_for_space(struct pipe_inode_info *pipe, unsigned flags)
{
1077 1078 1079 1080 1081 1082 1083
	for (;;) {
		if (unlikely(!pipe->readers)) {
			send_sig(SIGPIPE, current, 0);
			return -EPIPE;
		}
		if (pipe->nrbufs != pipe->buffers)
			return 0;
1084 1085 1086 1087 1088 1089 1090 1091 1092 1093
		if (flags & SPLICE_F_NONBLOCK)
			return -EAGAIN;
		if (signal_pending(current))
			return -ERESTARTSYS;
		pipe->waiting_writers++;
		pipe_wait(pipe);
		pipe->waiting_writers--;
	}
}

1094 1095 1096
static int splice_pipe_to_pipe(struct pipe_inode_info *ipipe,
			       struct pipe_inode_info *opipe,
			       size_t len, unsigned int flags);
1097

1098 1099 1100
/*
 * Determine where to splice to/from.
 */
1101 1102 1103
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)
1104
{
1105 1106
	struct pipe_inode_info *ipipe;
	struct pipe_inode_info *opipe;
1107
	loff_t offset;
J
Jens Axboe 已提交
1108
	long ret;
1109

1110 1111
	ipipe = get_pipe_info(in);
	opipe = get_pipe_info(out);
1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130

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

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

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

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

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

	if (ipipe) {
1131 1132
		if (off_in)
			return -ESPIPE;
1133
		if (off_out) {
1134
			if (!(out->f_mode & FMODE_PWRITE))
1135
				return -EINVAL;
1136
			if (copy_from_user(&offset, off_out, sizeof(loff_t)))
1137
				return -EFAULT;
1138 1139 1140
		} else {
			offset = out->f_pos;
		}
1141

1142 1143 1144 1145 1146 1147 1148 1149 1150 1151
		if (unlikely(!(out->f_mode & FMODE_WRITE)))
			return -EBADF;

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

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

1152
		file_start_write(out);
1153
		ret = do_splice_from(ipipe, out, &offset, len, flags);
1154
		file_end_write(out);
J
Jens Axboe 已提交
1155

1156 1157 1158
		if (!off_out)
			out->f_pos = offset;
		else if (copy_to_user(off_out, &offset, sizeof(loff_t)))
J
Jens Axboe 已提交
1159 1160 1161
			ret = -EFAULT;

		return ret;
1162
	}
1163

1164
	if (opipe) {
1165 1166
		if (off_out)
			return -ESPIPE;
1167
		if (off_in) {
1168
			if (!(in->f_mode & FMODE_PREAD))
1169
				return -EINVAL;
1170
			if (copy_from_user(&offset, off_in, sizeof(loff_t)))
1171
				return -EFAULT;
1172 1173 1174
		} else {
			offset = in->f_pos;
		}
1175

1176 1177 1178 1179 1180 1181 1182
		pipe_lock(opipe);
		ret = wait_for_space(opipe, flags);
		if (!ret)
			ret = do_splice_to(in, &offset, opipe, len, flags);
		pipe_unlock(opipe);
		if (ret > 0)
			wakeup_pipe_readers(opipe);
1183 1184 1185
		if (!off_in)
			in->f_pos = offset;
		else if (copy_to_user(off_in, &offset, sizeof(loff_t)))
J
Jens Axboe 已提交
1186 1187 1188
			ret = -EFAULT;

		return ret;
1189
	}
1190 1191 1192 1193

	return -EINVAL;
}

A
Al Viro 已提交
1194 1195 1196
static int iter_to_pipe(struct iov_iter *from,
			struct pipe_inode_info *pipe,
			unsigned flags)
1197
{
A
Al Viro 已提交
1198 1199 1200 1201 1202 1203 1204 1205 1206 1207
	struct pipe_buffer buf = {
		.ops = &user_page_pipe_buf_ops,
		.flags = flags
	};
	size_t total = 0;
	int ret = 0;
	bool failed = false;

	while (iov_iter_count(from) && !failed) {
		struct page *pages[16];
1208 1209
		ssize_t copied;
		size_t start;
A
Al Viro 已提交
1210
		int n;
1211

A
Al Viro 已提交
1212 1213 1214 1215 1216
		copied = iov_iter_get_pages(from, pages, ~0UL, 16, &start);
		if (copied <= 0) {
			ret = copied;
			break;
		}
1217

A
Al Viro 已提交
1218
		for (n = 0; copied; n++, start = 0) {
1219
			int size = min_t(int, copied, PAGE_SIZE - start);
A
Al Viro 已提交
1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233
			if (!failed) {
				buf.page = pages[n];
				buf.offset = start;
				buf.len = size;
				ret = add_to_pipe(pipe, &buf);
				if (unlikely(ret < 0)) {
					failed = true;
				} else {
					iov_iter_advance(from, ret);
					total += ret;
				}
			} else {
				put_page(pages[n]);
			}
1234
			copied -= size;
1235 1236
		}
	}
A
Al Viro 已提交
1237
	return total ? total : ret;
1238 1239
}

1240 1241 1242
static int pipe_to_user(struct pipe_inode_info *pipe, struct pipe_buffer *buf,
			struct splice_desc *sd)
{
1243 1244
	int n = copy_page_to_iter(buf->page, buf->offset, sd->len, sd->u.data);
	return n == sd->len ? n : -EFAULT;
1245 1246 1247 1248 1249 1250
}

/*
 * For lack of a better implementation, implement vmsplice() to userspace
 * as a simple copy of the pipes pages to the user iov.
 */
1251 1252
static long vmsplice_to_user(struct file *file, struct iov_iter *iter,
			     unsigned int flags)
1253
{
1254 1255 1256 1257 1258 1259 1260
	struct pipe_inode_info *pipe = get_pipe_info(file);
	struct splice_desc sd = {
		.total_len = iov_iter_count(iter),
		.flags = flags,
		.u.data = iter
	};
	long ret = 0;
1261 1262 1263 1264

	if (!pipe)
		return -EBADF;

1265 1266 1267 1268 1269
	if (sd.total_len) {
		pipe_lock(pipe);
		ret = __splice_from_pipe(pipe, &sd, pipe_to_user);
		pipe_unlock(pipe);
	}
1270 1271 1272 1273

	return ret;
}

1274 1275 1276 1277 1278
/*
 * 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.
 */
1279 1280
static long vmsplice_to_pipe(struct file *file, struct iov_iter *iter,
			     unsigned int flags)
1281
{
1282
	struct pipe_inode_info *pipe;
1283
	long ret = 0;
A
Al Viro 已提交
1284 1285 1286 1287
	unsigned buf_flag = 0;

	if (flags & SPLICE_F_GIFT)
		buf_flag = PIPE_BUF_FLAG_GIFT;
1288

1289
	pipe = get_pipe_info(file);
1290
	if (!pipe)
1291 1292
		return -EBADF;

1293 1294
	pipe_lock(pipe);
	ret = wait_for_space(pipe, flags);
A
Al Viro 已提交
1295
	if (!ret)
1296
		ret = iter_to_pipe(iter, pipe, buf_flag);
1297 1298 1299
	pipe_unlock(pipe);
	if (ret > 0)
		wakeup_pipe_readers(pipe);
1300
	return ret;
1301 1302
}

1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317
static int vmsplice_type(struct fd f, int *type)
{
	if (!f.file)
		return -EBADF;
	if (f.file->f_mode & FMODE_WRITE) {
		*type = WRITE;
	} else if (f.file->f_mode & FMODE_READ) {
		*type = READ;
	} else {
		fdput(f);
		return -EBADF;
	}
	return 0;
}

1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333
/*
 * 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().
 *
 */
1334
static long do_vmsplice(struct file *f, struct iov_iter *iter, unsigned int flags)
1335
{
A
Al Viro 已提交
1336 1337
	if (unlikely(flags & ~SPLICE_F_ALL))
		return -EINVAL;
1338

1339 1340
	if (!iov_iter_count(iter))
		return 0;
1341

1342 1343 1344 1345
	if (iov_iter_rw(iter) == WRITE)
		return vmsplice_to_pipe(f, iter, flags);
	else
		return vmsplice_to_user(f, iter, flags);
1346 1347
}

1348
SYSCALL_DEFINE4(vmsplice, int, fd, const struct iovec __user *, uiov,
1349 1350
		unsigned long, nr_segs, unsigned int, flags)
{
1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370
	struct iovec iovstack[UIO_FASTIOV];
	struct iovec *iov = iovstack;
	struct iov_iter iter;
	long error;
	struct fd f;
	int type;

	f = fdget(fd);
	error = vmsplice_type(f, &type);
	if (error)
		return error;

	error = import_iovec(type, uiov, nr_segs,
			     ARRAY_SIZE(iovstack), &iov, &iter);
	if (!error) {
		error = do_vmsplice(f.file, &iter, flags);
		kfree(iov);
	}
	fdput(f);
	return error;
1371 1372
}

1373 1374 1375 1376
#ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE4(vmsplice, int, fd, const struct compat_iovec __user *, iov32,
		    unsigned int, nr_segs, unsigned int, flags)
{
1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393
	struct iovec iovstack[UIO_FASTIOV];
	struct iovec *iov = iovstack;
	struct iov_iter iter;
	long error;
	struct fd f;
	int type;

	f = fdget(fd);
	error = vmsplice_type(f, &type);
	if (error)
		return error;

	error = compat_import_iovec(type, iov32, nr_segs,
			     ARRAY_SIZE(iovstack), &iov, &iter);
	if (!error) {
		error = do_vmsplice(f.file, &iter, flags);
		kfree(iov);
1394
	}
1395 1396
	fdput(f);
	return error;
1397 1398 1399
}
#endif

1400 1401 1402
SYSCALL_DEFINE6(splice, int, fd_in, loff_t __user *, off_in,
		int, fd_out, loff_t __user *, off_out,
		size_t, len, unsigned int, flags)
1403
{
1404
	struct fd in, out;
1405 1406 1407 1408 1409
	long error;

	if (unlikely(!len))
		return 0;

A
Al Viro 已提交
1410 1411 1412
	if (unlikely(flags & ~SPLICE_F_ALL))
		return -EINVAL;

1413
	error = -EBADF;
1414 1415 1416 1417 1418 1419 1420 1421
	in = fdget(fd_in);
	if (in.file) {
		if (in.file->f_mode & FMODE_READ) {
			out = fdget(fd_out);
			if (out.file) {
				if (out.file->f_mode & FMODE_WRITE)
					error = do_splice(in.file, off_in,
							  out.file, off_out,
1422
							  len, flags);
1423
				fdput(out);
1424 1425
			}
		}
1426
		fdput(in);
1427 1428 1429
	}
	return error;
}
1430

1431 1432 1433 1434
/*
 * Make sure there's data to read. Wait for input if we can, otherwise
 * return an appropriate error.
 */
1435
static int ipipe_prep(struct pipe_inode_info *pipe, unsigned int flags)
1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446
{
	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;
1447
	pipe_lock(pipe);
1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464

	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);
	}

1465
	pipe_unlock(pipe);
1466 1467 1468 1469 1470 1471 1472
	return ret;
}

/*
 * Make sure there's writeable room. Wait for room if we can, otherwise
 * return an appropriate error.
 */
1473
static int opipe_prep(struct pipe_inode_info *pipe, unsigned int flags)
1474 1475 1476 1477 1478 1479 1480
{
	int ret;

	/*
	 * Check ->nrbufs without the inode lock first. This function
	 * is speculative anyways, so missing one is ok.
	 */
1481
	if (pipe->nrbufs < pipe->buffers)
1482 1483 1484
		return 0;

	ret = 0;
1485
	pipe_lock(pipe);
1486

1487
	while (pipe->nrbufs >= pipe->buffers) {
1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505
		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--;
	}

1506
	pipe_unlock(pipe);
1507 1508 1509
	return ret;
}

1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552
/*
 * Splice contents of ipipe to opipe.
 */
static int splice_pipe_to_pipe(struct pipe_inode_info *ipipe,
			       struct pipe_inode_info *opipe,
			       size_t len, unsigned int flags)
{
	struct pipe_buffer *ibuf, *obuf;
	int ret = 0, nbuf;
	bool input_wakeup = false;


retry:
	ret = ipipe_prep(ipipe, flags);
	if (ret)
		return ret;

	ret = opipe_prep(opipe, flags);
	if (ret)
		return ret;

	/*
	 * Potential ABBA deadlock, work around it by ordering lock
	 * grabbing by pipe info address. Otherwise two different processes
	 * could deadlock (one doing tee from A -> B, the other from B -> A).
	 */
	pipe_double_lock(ipipe, opipe);

	do {
		if (!opipe->readers) {
			send_sig(SIGPIPE, current, 0);
			if (!ret)
				ret = -EPIPE;
			break;
		}

		if (!ipipe->nrbufs && !ipipe->writers)
			break;

		/*
		 * Cannot make any progress, because either the input
		 * pipe is empty or the output pipe is full.
		 */
1553
		if (!ipipe->nrbufs || opipe->nrbufs >= opipe->buffers) {
1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573
			/* Already processed some buffers, break */
			if (ret)
				break;

			if (flags & SPLICE_F_NONBLOCK) {
				ret = -EAGAIN;
				break;
			}

			/*
			 * We raced with another reader/writer and haven't
			 * managed to process any buffers.  A zero return
			 * value means EOF, so retry instead.
			 */
			pipe_unlock(ipipe);
			pipe_unlock(opipe);
			goto retry;
		}

		ibuf = ipipe->bufs + ipipe->curbuf;
1574
		nbuf = (opipe->curbuf + opipe->nrbufs) & (opipe->buffers - 1);
1575 1576 1577 1578 1579 1580 1581 1582 1583
		obuf = opipe->bufs + nbuf;

		if (len >= ibuf->len) {
			/*
			 * Simply move the whole buffer from ipipe to opipe
			 */
			*obuf = *ibuf;
			ibuf->ops = NULL;
			opipe->nrbufs++;
1584
			ipipe->curbuf = (ipipe->curbuf + 1) & (ipipe->buffers - 1);
1585 1586 1587 1588 1589 1590 1591
			ipipe->nrbufs--;
			input_wakeup = true;
		} else {
			/*
			 * Get a reference to this pipe buffer,
			 * so we can copy the contents over.
			 */
1592 1593 1594 1595 1596
			if (!pipe_buf_get(ipipe, ibuf)) {
				if (ret == 0)
					ret = -EFAULT;
				break;
			}
1597 1598 1599 1600 1601 1602 1603 1604
			*obuf = *ibuf;

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

1605 1606
			pipe_buf_mark_unmergeable(obuf);

1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621
			obuf->len = len;
			opipe->nrbufs++;
			ibuf->offset += obuf->len;
			ibuf->len -= obuf->len;
		}
		ret += obuf->len;
		len -= obuf->len;
	} while (len);

	pipe_unlock(ipipe);
	pipe_unlock(opipe);

	/*
	 * If we put data in the output pipe, wakeup any potential readers.
	 */
N
Namhyung Kim 已提交
1622 1623 1624
	if (ret > 0)
		wakeup_pipe_readers(opipe);

1625 1626 1627 1628 1629 1630
	if (input_wakeup)
		wakeup_pipe_writers(ipipe);

	return ret;
}

1631 1632 1633 1634 1635 1636 1637 1638
/*
 * 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;
1639
	int ret = 0, i = 0, nbuf;
1640 1641 1642

	/*
	 * Potential ABBA deadlock, work around it by ordering lock
1643
	 * grabbing by pipe info address. Otherwise two different processes
1644 1645
	 * could deadlock (one doing tee from A -> B, the other from B -> A).
	 */
1646
	pipe_double_lock(ipipe, opipe);
1647

1648
	do {
1649 1650 1651 1652 1653 1654 1655
		if (!opipe->readers) {
			send_sig(SIGPIPE, current, 0);
			if (!ret)
				ret = -EPIPE;
			break;
		}

1656 1657 1658 1659
		/*
		 * If we have iterated all input buffers or ran out of
		 * output room, break.
		 */
1660
		if (i >= ipipe->nrbufs || opipe->nrbufs >= opipe->buffers)
1661
			break;
1662

1663 1664
		ibuf = ipipe->bufs + ((ipipe->curbuf + i) & (ipipe->buffers-1));
		nbuf = (opipe->curbuf + opipe->nrbufs) & (opipe->buffers - 1);
1665 1666

		/*
1667 1668
		 * Get a reference to this pipe buffer,
		 * so we can copy the contents over.
1669
		 */
1670 1671 1672 1673 1674
		if (!pipe_buf_get(ipipe, ibuf)) {
			if (ret == 0)
				ret = -EFAULT;
			break;
		}
1675 1676 1677 1678

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

1679
		/*
1680 1681
		 * Don't inherit the gift flag, we need to
		 * prevent multiple steals of this page.
1682
		 */
1683
		obuf->flags &= ~PIPE_BUF_FLAG_GIFT;
1684

1685 1686
		pipe_buf_mark_unmergeable(obuf);

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 1697 1698 1699 1700 1701 1702
	/*
	 * return EAGAIN if we have the potential of some data in the
	 * future, otherwise just return 0
	 */
	if (!ret && ipipe->waiting_writers && (flags & SPLICE_F_NONBLOCK))
		ret = -EAGAIN;

1703 1704
	pipe_unlock(ipipe);
	pipe_unlock(opipe);
1705

1706 1707 1708
	/*
	 * If we put data in the output pipe, wakeup any potential readers.
	 */
N
Namhyung Kim 已提交
1709 1710
	if (ret > 0)
		wakeup_pipe_readers(opipe);
1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723

	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)
{
1724 1725
	struct pipe_inode_info *ipipe = get_pipe_info(in);
	struct pipe_inode_info *opipe = get_pipe_info(out);
1726
	int ret = -EINVAL;
1727 1728

	/*
1729 1730
	 * Duplicate the contents of ipipe to opipe without actually
	 * copying the data.
1731
	 */
1732 1733 1734 1735 1736
	if (ipipe && opipe && ipipe != opipe) {
		/*
		 * Keep going, unless we encounter an error. The ipipe/opipe
		 * ordering doesn't really matter.
		 */
1737
		ret = ipipe_prep(ipipe, flags);
1738
		if (!ret) {
1739
			ret = opipe_prep(opipe, flags);
1740
			if (!ret)
1741 1742 1743
				ret = link_pipe(ipipe, opipe, len, flags);
		}
	}
1744

1745
	return ret;
1746 1747
}

1748
SYSCALL_DEFINE4(tee, int, fdin, int, fdout, size_t, len, unsigned int, flags)
1749
{
1750 1751
	struct fd in;
	int error;
1752

A
Al Viro 已提交
1753 1754 1755
	if (unlikely(flags & ~SPLICE_F_ALL))
		return -EINVAL;

1756 1757 1758 1759
	if (unlikely(!len))
		return 0;

	error = -EBADF;
1760 1761 1762 1763 1764 1765 1766 1767 1768
	in = fdget(fdin);
	if (in.file) {
		if (in.file->f_mode & FMODE_READ) {
			struct fd out = fdget(fdout);
			if (out.file) {
				if (out.file->f_mode & FMODE_WRITE)
					error = do_tee(in.file, out.file,
							len, flags);
				fdput(out);
1769 1770
			}
		}
1771
 		fdput(in);
1772 1773 1774 1775
 	}

	return error;
}