splice.c 46.1 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>
K
KAMEZAWA Hiroyuki 已提交
24
#include <linux/memcontrol.h>
25
#include <linux/mm_inline.h>
26
#include <linux/swap.h>
27
#include <linux/writeback.h>
28
#include <linux/export.h>
29
#include <linux/syscalls.h>
30
#include <linux/uio.h>
31
#include <linux/security.h>
32
#include <linux/gfp.h>
33
#include <linux/socket.h>
34
#include <linux/compat.h>
A
Al Viro 已提交
35
#include <linux/aio.h>
36
#include "internal.h"
37

38 39 40 41 42 43
/*
 * 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.
 */
44
static int page_cache_pipe_buf_steal(struct pipe_inode_info *pipe,
45 46 47
				     struct pipe_buffer *buf)
{
	struct page *page = buf->page;
48
	struct address_space *mapping;
49

50 51
	lock_page(page);

52 53 54
	mapping = page_mapping(page);
	if (mapping) {
		WARN_ON(!PageUptodate(page));
55

56 57 58 59 60 61 62 63 64
		/*
		 * 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);
65

66 67
		if (page_has_private(page) &&
		    !try_to_release_page(page, GFP_KERNEL))
68
			goto out_unlock;
69

70 71 72 73 74 75 76 77
		/*
		 * 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;
		}
78
	}
79

80 81 82 83
	/*
	 * Raced with truncate or failed to remove page from current
	 * address space, unlock and return failure.
	 */
84
out_unlock:
85 86
	unlock_page(page);
	return 1;
87 88
}

89
static void page_cache_pipe_buf_release(struct pipe_inode_info *pipe,
90 91 92
					struct pipe_buffer *buf)
{
	page_cache_release(buf->page);
J
Jens Axboe 已提交
93
	buf->flags &= ~PIPE_BUF_FLAG_LRU;
94 95
}

96 97 98 99
/*
 * Check whether the contents of buf is OK to access. Since the content
 * is a page cache page, IO may be in flight.
 */
100 101
static int page_cache_pipe_buf_confirm(struct pipe_inode_info *pipe,
				       struct pipe_buffer *buf)
102 103
{
	struct page *page = buf->page;
104
	int err;
105 106

	if (!PageUptodate(page)) {
107 108 109 110
		lock_page(page);

		/*
		 * Page got truncated/unhashed. This will cause a 0-byte
I
Ingo Molnar 已提交
111
		 * splice, if this is the first page.
112 113 114 115 116
		 */
		if (!page->mapping) {
			err = -ENODATA;
			goto error;
		}
117

118
		/*
I
Ingo Molnar 已提交
119
		 * Uh oh, read-error from disk.
120 121 122 123 124 125 126
		 */
		if (!PageUptodate(page)) {
			err = -EIO;
			goto error;
		}

		/*
127
		 * Page is ok afterall, we are done.
128
		 */
129 130 131
		unlock_page(page);
	}

132
	return 0;
133 134
error:
	unlock_page(page);
135
	return err;
136 137
}

138
const struct pipe_buf_operations page_cache_pipe_buf_ops = {
139
	.can_merge = 0,
140
	.confirm = page_cache_pipe_buf_confirm,
141
	.release = page_cache_pipe_buf_release,
142
	.steal = page_cache_pipe_buf_steal,
143
	.get = generic_pipe_buf_get,
144 145
};

146 147 148
static int user_page_pipe_buf_steal(struct pipe_inode_info *pipe,
				    struct pipe_buffer *buf)
{
149 150 151
	if (!(buf->flags & PIPE_BUF_FLAG_GIFT))
		return 1;

J
Jens Axboe 已提交
152
	buf->flags |= PIPE_BUF_FLAG_LRU;
153
	return generic_pipe_buf_steal(pipe, buf);
154 155
}

156
static const struct pipe_buf_operations user_page_pipe_buf_ops = {
157
	.can_merge = 0,
158
	.confirm = generic_pipe_buf_confirm,
159 160
	.release = page_cache_pipe_buf_release,
	.steal = user_page_pipe_buf_steal,
161
	.get = generic_pipe_buf_get,
162 163
};

N
Namhyung Kim 已提交
164 165 166 167 168 169 170 171
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);
}

172 173 174 175 176 177
/**
 * splice_to_pipe - fill passed data into a pipe
 * @pipe:	pipe to fill
 * @spd:	data to fill
 *
 * Description:
R
Randy Dunlap 已提交
178
 *    @spd contains a map of pages and len/offset tuples, along with
179 180 181
 *    the struct pipe_buf_operations associated with these pages. This
 *    function will link that data to the pipe.
 *
182
 */
183 184
ssize_t splice_to_pipe(struct pipe_inode_info *pipe,
		       struct splice_pipe_desc *spd)
185
{
186
	unsigned int spd_pages = spd->nr_pages;
187
	int ret, do_wakeup, page_nr;
188 189 190

	ret = 0;
	do_wakeup = 0;
191
	page_nr = 0;
192

193
	pipe_lock(pipe);
194 195

	for (;;) {
196
		if (!pipe->readers) {
197 198 199 200 201 202
			send_sig(SIGPIPE, current, 0);
			if (!ret)
				ret = -EPIPE;
			break;
		}

203 204
		if (pipe->nrbufs < pipe->buffers) {
			int newbuf = (pipe->curbuf + pipe->nrbufs) & (pipe->buffers - 1);
205
			struct pipe_buffer *buf = pipe->bufs + newbuf;
206

207 208 209
			buf->page = spd->pages[page_nr];
			buf->offset = spd->partial[page_nr].offset;
			buf->len = spd->partial[page_nr].len;
210
			buf->private = spd->partial[page_nr].private;
211
			buf->ops = spd->ops;
212 213 214
			if (spd->flags & SPLICE_F_GIFT)
				buf->flags |= PIPE_BUF_FLAG_GIFT;

215
			pipe->nrbufs++;
216 217 218
			page_nr++;
			ret += buf->len;

A
Al Viro 已提交
219
			if (pipe->files)
220
				do_wakeup = 1;
221

222
			if (!--spd->nr_pages)
223
				break;
224
			if (pipe->nrbufs < pipe->buffers)
225 226 227 228 229
				continue;

			break;
		}

230
		if (spd->flags & SPLICE_F_NONBLOCK) {
231 232 233 234 235
			if (!ret)
				ret = -EAGAIN;
			break;
		}

236 237 238 239 240 241 242
		if (signal_pending(current)) {
			if (!ret)
				ret = -ERESTARTSYS;
			break;
		}

		if (do_wakeup) {
243
			smp_mb();
244 245 246
			if (waitqueue_active(&pipe->wait))
				wake_up_interruptible_sync(&pipe->wait);
			kill_fasync(&pipe->fasync_readers, SIGIO, POLL_IN);
247 248 249
			do_wakeup = 0;
		}

250 251 252
		pipe->waiting_writers++;
		pipe_wait(pipe);
		pipe->waiting_writers--;
253 254
	}

255
	pipe_unlock(pipe);
256

N
Namhyung Kim 已提交
257 258
	if (do_wakeup)
		wakeup_pipe_readers(pipe);
259

260
	while (page_nr < spd_pages)
261
		spd->spd_release(spd, page_nr++);
262 263 264 265

	return ret;
}

266
void spd_release_page(struct splice_pipe_desc *spd, unsigned int i)
267 268 269 270
{
	page_cache_release(spd->pages[i]);
}

271 272 273 274
/*
 * Check if we need to grow the arrays holding pages and partial page
 * descriptions.
 */
275
int splice_grow_spd(const struct pipe_inode_info *pipe, struct splice_pipe_desc *spd)
276
{
277 278 279 280
	unsigned int buffers = ACCESS_ONCE(pipe->buffers);

	spd->nr_pages_max = buffers;
	if (buffers <= PIPE_DEF_BUFFERS)
281 282
		return 0;

283 284
	spd->pages = kmalloc(buffers * sizeof(struct page *), GFP_KERNEL);
	spd->partial = kmalloc(buffers * sizeof(struct partial_page), GFP_KERNEL);
285 286 287 288 289 290 291 292 293

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

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

294
void splice_shrink_spd(struct splice_pipe_desc *spd)
295
{
296
	if (spd->nr_pages_max <= PIPE_DEF_BUFFERS)
297 298 299 300 301 302
		return;

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

303
static int
304 305 306
__generic_file_splice_read(struct file *in, loff_t *ppos,
			   struct pipe_inode_info *pipe, size_t len,
			   unsigned int flags)
307 308
{
	struct address_space *mapping = in->f_mapping;
F
Fengguang Wu 已提交
309
	unsigned int loff, nr_pages, req_pages;
310 311
	struct page *pages[PIPE_DEF_BUFFERS];
	struct partial_page partial[PIPE_DEF_BUFFERS];
312
	struct page *page;
313 314
	pgoff_t index, end_index;
	loff_t isize;
315
	int error, page_nr;
316 317 318
	struct splice_pipe_desc spd = {
		.pages = pages,
		.partial = partial,
319
		.nr_pages_max = PIPE_DEF_BUFFERS,
320 321
		.flags = flags,
		.ops = &page_cache_pipe_buf_ops,
322
		.spd_release = spd_release_page,
323
	};
324

325 326 327
	if (splice_grow_spd(pipe, &spd))
		return -ENOMEM;

328
	index = *ppos >> PAGE_CACHE_SHIFT;
329
	loff = *ppos & ~PAGE_CACHE_MASK;
F
Fengguang Wu 已提交
330
	req_pages = (len + loff + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
331
	nr_pages = min(req_pages, spd.nr_pages_max);
332

333 334 335
	/*
	 * Lookup the (hopefully) full range of pages we need.
	 */
336
	spd.nr_pages = find_get_pages_contig(mapping, index, nr_pages, spd.pages);
337
	index += spd.nr_pages;
338

339 340
	/*
	 * If find_get_pages_contig() returned fewer pages than we needed,
341
	 * readahead/allocate the rest and fill in the holes.
342
	 */
343
	if (spd.nr_pages < nr_pages)
344 345
		page_cache_sync_readahead(mapping, &in->f_ra, in,
				index, req_pages - spd.nr_pages);
346

347
	error = 0;
348
	while (spd.nr_pages < nr_pages) {
349
		/*
350 351
		 * Page could be there, find_get_pages_contig() breaks on
		 * the first hole.
352
		 */
353 354 355
		page = find_get_page(mapping, index);
		if (!page) {
			/*
356
			 * page didn't exist, allocate one.
357 358 359 360 361 362
			 */
			page = page_cache_alloc_cold(mapping);
			if (!page)
				break;

			error = add_to_page_cache_lru(page, mapping, index,
363
						GFP_KERNEL);
364 365
			if (unlikely(error)) {
				page_cache_release(page);
366 367
				if (error == -EEXIST)
					continue;
368 369
				break;
			}
370 371 372 373 374
			/*
			 * add_to_page_cache() locks the page, unlock it
			 * to avoid convoluting the logic below even more.
			 */
			unlock_page(page);
375 376
		}

377
		spd.pages[spd.nr_pages++] = page;
378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397
		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);
398
		page = spd.pages[page_nr];
399

400
		if (PageReadahead(page))
401
			page_cache_async_readahead(mapping, &in->f_ra, in,
F
Fengguang Wu 已提交
402
					page, index, req_pages - page_nr);
403

404 405 406 407
		/*
		 * If the page isn't uptodate, we may need to start io on it
		 */
		if (!PageUptodate(page)) {
408
			lock_page(page);
409 410

			/*
411 412 413 414
			 * Page was truncated, or invalidated by the
			 * filesystem.  Redo the find/create, but this time the
			 * page is kept locked, so there's no chance of another
			 * race with truncate/invalidate.
415 416 417
			 */
			if (!page->mapping) {
				unlock_page(page);
418 419 420 421 422 423 424
				page = find_or_create_page(mapping, index,
						mapping_gfp_mask(mapping));

				if (!page) {
					error = -ENOMEM;
					break;
				}
425 426
				page_cache_release(spd.pages[page_nr]);
				spd.pages[page_nr] = page;
427 428 429 430 431 432 433 434
			}
			/*
			 * page was already under io and is now done, great
			 */
			if (PageUptodate(page)) {
				unlock_page(page);
				goto fill_it;
			}
435

436 437 438 439
			/*
			 * need to read in the page
			 */
			error = mapping->a_ops->readpage(in, page);
440
			if (unlikely(error)) {
441 442 443 444 445 446
				/*
				 * 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.
				 */
447
				if (error == AOP_TRUNCATED_PAGE)
448 449
					error = 0;

450 451
				break;
			}
452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467
		}
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;
468 469

			/*
470
			 * max good bytes in this page
471
			 */
472 473
			plen = ((isize - 1) & ~PAGE_CACHE_MASK) + 1;
			if (plen <= loff)
474 475 476
				break;

			/*
477
			 * force quit after adding this page
478
			 */
479 480
			this_len = min(this_len, plen - loff);
			len = this_len;
481
		}
482

483 484
		spd.partial[page_nr].offset = loff;
		spd.partial[page_nr].len = this_len;
485
		len -= this_len;
486
		loff = 0;
487 488
		spd.nr_pages++;
		index++;
489 490
	}

491
	/*
492
	 * Release any pages at the end, if we quit early. 'page_nr' is how far
493 494 495
	 * we got, 'nr_pages' is how many pages are in the map.
	 */
	while (page_nr < nr_pages)
496
		page_cache_release(spd.pages[page_nr++]);
497
	in->f_ra.prev_pos = (loff_t)index << PAGE_CACHE_SHIFT;
498

499
	if (spd.nr_pages)
500
		error = splice_to_pipe(pipe, &spd);
501

502
	splice_shrink_spd(&spd);
503
	return error;
504 505
}

506 507 508
/**
 * generic_file_splice_read - splice data from file to a pipe
 * @in:		file to splice from
509
 * @ppos:	position in @in
510 511 512 513
 * @pipe:	pipe to splice to
 * @len:	number of bytes to splice
 * @flags:	splice modifier flags
 *
514 515 516 517 518
 * 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.
 *
519
 */
520 521 522
ssize_t generic_file_splice_read(struct file *in, loff_t *ppos,
				 struct pipe_inode_info *pipe, size_t len,
				 unsigned int flags)
523
{
524
	loff_t isize, left;
525
	int ret;
526 527 528 529 530 531 532 533

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

	left = isize - *ppos;
	if (unlikely(left < len))
		len = left;
534

535
	ret = __generic_file_splice_read(in, ppos, pipe, len, flags);
536
	if (ret > 0) {
537
		*ppos += ret;
538 539
		file_accessed(in);
	}
540 541 542

	return ret;
}
543 544
EXPORT_SYMBOL(generic_file_splice_read);

545 546 547 548 549 550 551 552
static const struct pipe_buf_operations default_pipe_buf_ops = {
	.can_merge = 0,
	.confirm = generic_pipe_buf_confirm,
	.release = generic_pipe_buf_release,
	.steal = generic_pipe_buf_steal,
	.get = generic_pipe_buf_get,
};

M
Miklos Szeredi 已提交
553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568
static int generic_pipe_buf_nosteal(struct pipe_inode_info *pipe,
				    struct pipe_buffer *buf)
{
	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);

569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584
static ssize_t kernel_readv(struct file *file, const struct iovec *vec,
			    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() */
	res = vfs_readv(file, (const struct iovec __user *)vec, vlen, &pos);
	set_fs(old_fs);

	return res;
}

585
ssize_t kernel_write(struct file *file, const char *buf, size_t count,
586
			    loff_t pos)
587 588 589 590 591 592 593
{
	mm_segment_t old_fs;
	ssize_t res;

	old_fs = get_fs();
	set_fs(get_ds());
	/* The cast to a user pointer is valid due to the set_fs() */
594
	res = vfs_write(file, (__force const char __user *)buf, count, &pos);
595 596 597 598
	set_fs(old_fs);

	return res;
}
599
EXPORT_SYMBOL(kernel_write);
600

601 602 603 604 605 606 607
ssize_t default_file_splice_read(struct file *in, loff_t *ppos,
				 struct pipe_inode_info *pipe, size_t len,
				 unsigned int flags)
{
	unsigned int nr_pages;
	unsigned int nr_freed;
	size_t offset;
608 609 610
	struct page *pages[PIPE_DEF_BUFFERS];
	struct partial_page partial[PIPE_DEF_BUFFERS];
	struct iovec *vec, __vec[PIPE_DEF_BUFFERS];
611 612 613 614 615 616 617
	ssize_t res;
	size_t this_len;
	int error;
	int i;
	struct splice_pipe_desc spd = {
		.pages = pages,
		.partial = partial,
618
		.nr_pages_max = PIPE_DEF_BUFFERS,
619 620 621 622 623
		.flags = flags,
		.ops = &default_pipe_buf_ops,
		.spd_release = spd_release_page,
	};

624 625 626 627 628
	if (splice_grow_spd(pipe, &spd))
		return -ENOMEM;

	res = -ENOMEM;
	vec = __vec;
629 630
	if (spd.nr_pages_max > PIPE_DEF_BUFFERS) {
		vec = kmalloc(spd.nr_pages_max * sizeof(struct iovec), GFP_KERNEL);
631 632 633 634
		if (!vec)
			goto shrink_ret;
	}

635 636 637
	offset = *ppos & ~PAGE_CACHE_MASK;
	nr_pages = (len + offset + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;

638
	for (i = 0; i < nr_pages && i < spd.nr_pages_max && len; i++) {
639 640
		struct page *page;

641
		page = alloc_page(GFP_USER);
642 643 644 645 646
		error = -ENOMEM;
		if (!page)
			goto err;

		this_len = min_t(size_t, len, PAGE_CACHE_SIZE - offset);
647
		vec[i].iov_base = (void __user *) page_address(page);
648
		vec[i].iov_len = this_len;
649
		spd.pages[i] = page;
650 651 652 653 654 655
		spd.nr_pages++;
		len -= this_len;
		offset = 0;
	}

	res = kernel_readv(in, vec, spd.nr_pages, *ppos);
A
Andrew Morton 已提交
656 657
	if (res < 0) {
		error = res;
658
		goto err;
A
Andrew Morton 已提交
659
	}
660 661 662 663 664 665 666 667

	error = 0;
	if (!res)
		goto err;

	nr_freed = 0;
	for (i = 0; i < spd.nr_pages; i++) {
		this_len = min_t(size_t, vec[i].iov_len, res);
668 669
		spd.partial[i].offset = 0;
		spd.partial[i].len = this_len;
670
		if (!this_len) {
671 672
			__free_page(spd.pages[i]);
			spd.pages[i] = NULL;
673 674 675 676 677 678 679 680 681 682
			nr_freed++;
		}
		res -= this_len;
	}
	spd.nr_pages -= nr_freed;

	res = splice_to_pipe(pipe, &spd);
	if (res > 0)
		*ppos += res;

683 684 685
shrink_ret:
	if (vec != __vec)
		kfree(vec);
686
	splice_shrink_spd(&spd);
687 688 689
	return res;

err:
690
	for (i = 0; i < spd.nr_pages; i++)
691
		__free_page(spd.pages[i]);
692

693 694
	res = error;
	goto shrink_ret;
695 696 697
}
EXPORT_SYMBOL(default_file_splice_read);

698
/*
699
 * Send 'sd->len' bytes to socket from 'sd->file' at position 'sd->pos'
J
Jens Axboe 已提交
700
 * using sendpage(). Return the number of bytes sent.
701
 */
702
static int pipe_to_sendpage(struct pipe_inode_info *pipe,
703 704
			    struct pipe_buffer *buf, struct splice_desc *sd)
{
705
	struct file *file = sd->u.file;
706
	loff_t pos = sd->pos;
707
	int more;
708

A
Al Viro 已提交
709
	if (!likely(file->f_op->sendpage))
710 711
		return -EINVAL;

712
	more = (sd->flags & SPLICE_F_MORE) ? MSG_MORE : 0;
713 714

	if (sd->len < sd->total_len && pipe->nrbufs > 1)
715
		more |= MSG_SENDPAGE_NOTLAST;
716

717 718
	return file->f_op->sendpage(file, buf->page, buf->offset,
				    sd->len, &pos, more);
719 720
}

721 722 723 724 725 726 727 728
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);
}

729
/**
730
 * splice_from_pipe_feed - feed available data from a pipe to a file
731 732 733 734 735
 * @pipe:	pipe to splice from
 * @sd:		information to @actor
 * @actor:	handler that splices the data
 *
 * Description:
736 737 738 739 740 741 742
 *    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.
743
 *
744 745 746 747
 *    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.
748
 */
A
Al Viro 已提交
749
static int splice_from_pipe_feed(struct pipe_inode_info *pipe, struct splice_desc *sd,
750
			  splice_actor *actor)
751
{
752
	int ret;
753

754 755 756
	while (pipe->nrbufs) {
		struct pipe_buffer *buf = pipe->bufs + pipe->curbuf;
		const struct pipe_buf_operations *ops = buf->ops;
757

758 759 760
		sd->len = buf->len;
		if (sd->len > sd->total_len)
			sd->len = sd->total_len;
761

762 763
		ret = buf->ops->confirm(pipe, buf);
		if (unlikely(ret)) {
764 765 766 767
			if (ret == -ENODATA)
				ret = 0;
			return ret;
		}
768 769 770 771 772

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

773 774 775 776 777 778 779 780 781 782 783
		buf->offset += ret;
		buf->len -= ret;

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

		if (!buf->len) {
			buf->ops = NULL;
			ops->release(pipe, buf);
784
			pipe->curbuf = (pipe->curbuf + 1) & (pipe->buffers - 1);
785
			pipe->nrbufs--;
A
Al Viro 已提交
786
			if (pipe->files)
787 788
				sd->need_wakeup = true;
		}
789

790 791 792
		if (!sd->total_len)
			return 0;
	}
793

794 795
	return 1;
}
796

797 798 799 800 801 802 803 804 805 806
/**
 * 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
807
static int splice_from_pipe_next(struct pipe_inode_info *pipe, struct splice_desc *sd)
808 809 810 811
{
	while (!pipe->nrbufs) {
		if (!pipe->writers)
			return 0;
J
Jens Axboe 已提交
812

813 814
		if (!pipe->waiting_writers && sd->num_spliced)
			return 0;
I
Ingo Molnar 已提交
815

816 817
		if (sd->flags & SPLICE_F_NONBLOCK)
			return -EAGAIN;
818

819 820
		if (signal_pending(current))
			return -ERESTARTSYS;
821

822 823 824
		if (sd->need_wakeup) {
			wakeup_pipe_writers(pipe);
			sd->need_wakeup = false;
825 826
		}

827 828
		pipe_wait(pipe);
	}
829

830 831
	return 1;
}
832

833 834
/**
 * splice_from_pipe_begin - start splicing from pipe
835
 * @sd:		information about the splice operation
836 837 838 839 840 841
 *
 * 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
842
static void splice_from_pipe_begin(struct splice_desc *sd)
843 844 845 846
{
	sd->num_spliced = 0;
	sd->need_wakeup = false;
}
847

848 849 850 851 852 853 854 855 856 857
/**
 * 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
858
static void splice_from_pipe_end(struct pipe_inode_info *pipe, struct splice_desc *sd)
859 860 861 862
{
	if (sd->need_wakeup)
		wakeup_pipe_writers(pipe);
}
863

864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880
/**
 * __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;
881

882 883 884 885 886 887 888 889 890
	splice_from_pipe_begin(sd);
	do {
		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;
891
}
M
Mark Fasheh 已提交
892
EXPORT_SYMBOL(__splice_from_pipe);
893

894 895 896 897 898 899 900 901 902 903
/**
 * 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:
904
 *    See __splice_from_pipe. This function locks the pipe inode,
905 906 907
 *    otherwise it's identical to __splice_from_pipe().
 *
 */
908 909 910 911 912
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
913 914 915 916
	struct splice_desc sd = {
		.total_len = len,
		.flags = flags,
		.pos = *ppos,
917
		.u.file = out,
J
Jens Axboe 已提交
918
	};
919

920
	pipe_lock(pipe);
J
Jens Axboe 已提交
921
	ret = __splice_from_pipe(pipe, &sd, actor);
922
	pipe_unlock(pipe);
923 924 925 926

	return ret;
}

A
Al Viro 已提交
927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065
/**
 * 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;
		struct kiocb kiocb;
		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;

			ret = buf->ops->confirm(pipe, buf);
			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;
		}

		/* ... iov_iter */
		from.type = ITER_BVEC | WRITE;
		from.bvec = array;
		from.nr_segs = n;
		from.count = sd.total_len - left;
		from.iov_offset = 0;

		/* ... and iocb */
		init_sync_kiocb(&kiocb, out);
		kiocb.ki_pos = sd.pos;
		kiocb.ki_nbytes = sd.total_len - left;

		/* now, send it */
		ret = out->f_op->write_iter(&kiocb, &from);
		if (-EIOCBQUEUED == ret)
			ret = wait_on_sync_kiocb(&kiocb);

		if (ret <= 0)
			break;

		sd.num_spliced += ret;
		sd.total_len -= ret;
		*ppos = sd.pos = kiocb.ki_pos;

		/* dismiss the fully eaten buffers, adjust the partial one */
		while (ret) {
			struct pipe_buffer *buf = pipe->bufs + pipe->curbuf;
			if (ret >= buf->len) {
				const struct pipe_buf_operations *ops = buf->ops;
				ret -= buf->len;
				buf->len = 0;
				buf->ops = NULL;
				ops->release(pipe, buf);
				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);

1066 1067
static int write_pipe_buf(struct pipe_inode_info *pipe, struct pipe_buffer *buf,
			  struct splice_desc *sd)
1068
{
1069 1070
	int ret;
	void *data;
1071
	loff_t tmp = sd->pos;
1072

A
Al Viro 已提交
1073
	data = kmap(buf->page);
1074
	ret = __kernel_write(sd->u.file, data + buf->offset, sd->len, &tmp);
A
Al Viro 已提交
1075
	kunmap(buf->page);
1076 1077

	return ret;
1078 1079 1080 1081 1082 1083
}

static ssize_t default_file_splice_write(struct pipe_inode_info *pipe,
					 struct file *out, loff_t *ppos,
					 size_t len, unsigned int flags)
{
1084
	ssize_t ret;
1085

1086 1087 1088
	ret = splice_from_pipe(pipe, out, ppos, len, flags, write_pipe_buf);
	if (ret > 0)
		*ppos += ret;
1089

1090
	return ret;
1091 1092
}

1093 1094
/**
 * generic_splice_sendpage - splice data from a pipe to a socket
1095
 * @pipe:	pipe to splice from
1096
 * @out:	socket to write to
1097
 * @ppos:	position in @out
1098 1099 1100
 * @len:	number of bytes to splice
 * @flags:	splice modifier flags
 *
1101 1102 1103
 * Description:
 *    Will send @len bytes from the pipe to a network socket. No data copying
 *    is involved.
1104 1105
 *
 */
1106
ssize_t generic_splice_sendpage(struct pipe_inode_info *pipe, struct file *out,
1107
				loff_t *ppos, size_t len, unsigned int flags)
1108
{
1109
	return splice_from_pipe(pipe, out, ppos, len, flags, pipe_to_sendpage);
1110 1111
}

1112
EXPORT_SYMBOL(generic_splice_sendpage);
J
Jeff Garzik 已提交
1113

1114 1115 1116
/*
 * Attempt to initiate a splice from pipe to file.
 */
1117
static long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
1118
			   loff_t *ppos, size_t len, unsigned int flags)
1119
{
1120 1121
	ssize_t (*splice_write)(struct pipe_inode_info *, struct file *,
				loff_t *, size_t, unsigned int);
1122

A
Al Viro 已提交
1123
	if (out->f_op->splice_write)
1124 1125
		splice_write = out->f_op->splice_write;
	else
1126 1127
		splice_write = default_file_splice_write;

1128
	return splice_write(pipe, out, ppos, len, flags);
1129 1130
}

1131 1132 1133
/*
 * Attempt to initiate a splice from a file to a pipe.
 */
1134 1135 1136
static long do_splice_to(struct file *in, loff_t *ppos,
			 struct pipe_inode_info *pipe, size_t len,
			 unsigned int flags)
1137
{
1138 1139
	ssize_t (*splice_read)(struct file *, loff_t *,
			       struct pipe_inode_info *, size_t, unsigned int);
1140 1141
	int ret;

1142
	if (unlikely(!(in->f_mode & FMODE_READ)))
1143 1144
		return -EBADF;

1145
	ret = rw_verify_area(READ, in, ppos, len);
1146 1147 1148
	if (unlikely(ret < 0))
		return ret;

A
Al Viro 已提交
1149
	if (in->f_op->splice_read)
1150 1151
		splice_read = in->f_op->splice_read;
	else
1152 1153 1154
		splice_read = default_file_splice_read;

	return splice_read(in, ppos, pipe, len, flags);
1155 1156
}

1157 1158 1159 1160 1161 1162 1163 1164 1165
/**
 * 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
1166
 *    pipe is cached in the process, and reused during the lifetime of
1167 1168
 *    that process.
 *
J
Jens Axboe 已提交
1169 1170 1171
 */
ssize_t splice_direct_to_actor(struct file *in, struct splice_desc *sd,
			       splice_direct_actor *actor)
1172 1173 1174 1175
{
	struct pipe_inode_info *pipe;
	long ret, bytes;
	umode_t i_mode;
J
Jens Axboe 已提交
1176 1177
	size_t len;
	int i, flags;
1178 1179 1180 1181 1182 1183

	/*
	 * 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
1184
	i_mode = file_inode(in)->i_mode;
1185 1186 1187 1188 1189 1190 1191 1192
	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;
1193
	if (unlikely(!pipe)) {
A
Al Viro 已提交
1194
		pipe = alloc_pipe_info();
1195 1196 1197 1198 1199
		if (!pipe)
			return -ENOMEM;

		/*
		 * We don't have an immediate reader, but we'll read the stuff
1200
		 * out of the pipe right after the splice_to_pipe(). So set
1201 1202 1203 1204 1205 1206 1207 1208
		 * PIPE_READERS appropriately.
		 */
		pipe->readers = 1;

		current->splice_pipe = pipe;
	}

	/*
I
Ingo Molnar 已提交
1209
	 * Do the splice.
1210 1211 1212
	 */
	ret = 0;
	bytes = 0;
J
Jens Axboe 已提交
1213 1214 1215 1216 1217 1218 1219
	len = sd->total_len;
	flags = sd->flags;

	/*
	 * Don't block on output, we have to drain the direct pipe.
	 */
	sd->flags &= ~SPLICE_F_NONBLOCK;
1220 1221

	while (len) {
1222
		size_t read_len;
1223
		loff_t pos = sd->pos, prev_pos = pos;
1224

1225
		ret = do_splice_to(in, &pos, pipe, len, flags);
1226
		if (unlikely(ret <= 0))
1227 1228 1229
			goto out_release;

		read_len = ret;
J
Jens Axboe 已提交
1230
		sd->total_len = read_len;
1231 1232 1233 1234 1235 1236

		/*
		 * 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
1237
		ret = actor(pipe, sd);
1238 1239
		if (unlikely(ret <= 0)) {
			sd->pos = prev_pos;
1240
			goto out_release;
1241
		}
1242 1243 1244

		bytes += ret;
		len -= ret;
1245
		sd->pos = pos;
1246

1247 1248
		if (ret < read_len) {
			sd->pos = prev_pos + ret;
1249
			goto out_release;
1250
		}
1251 1252
	}

1253
done:
1254
	pipe->nrbufs = pipe->curbuf = 0;
1255
	file_accessed(in);
1256 1257 1258 1259 1260 1261 1262
	return bytes;

out_release:
	/*
	 * If we did an incomplete transfer we must release
	 * the pipe buffers in question:
	 */
1263
	for (i = 0; i < pipe->buffers; i++) {
1264 1265 1266 1267 1268 1269 1270 1271
		struct pipe_buffer *buf = pipe->bufs + i;

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

1272 1273
	if (!bytes)
		bytes = ret;
J
Jens Axboe 已提交
1274

1275
	goto done;
J
Jens Axboe 已提交
1276 1277 1278 1279 1280 1281
}
EXPORT_SYMBOL(splice_direct_to_actor);

static int direct_splice_actor(struct pipe_inode_info *pipe,
			       struct splice_desc *sd)
{
1282
	struct file *file = sd->u.file;
J
Jens Axboe 已提交
1283

1284
	return do_splice_from(pipe, file, sd->opos, sd->total_len,
1285
			      sd->flags);
J
Jens Axboe 已提交
1286 1287
}

1288 1289 1290 1291 1292
/**
 * do_splice_direct - splices data directly between two files
 * @in:		file to splice from
 * @ppos:	input file offset
 * @out:	file to splice to
1293
 * @opos:	output file offset
1294 1295 1296 1297 1298 1299 1300 1301 1302 1303
 * @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
1304
long do_splice_direct(struct file *in, loff_t *ppos, struct file *out,
1305
		      loff_t *opos, size_t len, unsigned int flags)
J
Jens Axboe 已提交
1306 1307 1308 1309 1310 1311
{
	struct splice_desc sd = {
		.len		= len,
		.total_len	= len,
		.flags		= flags,
		.pos		= *ppos,
1312
		.u.file		= out,
1313
		.opos		= opos,
J
Jens Axboe 已提交
1314
	};
1315
	long ret;
J
Jens Axboe 已提交
1316

1317 1318 1319 1320 1321 1322 1323 1324 1325 1326
	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
1327
	ret = splice_direct_to_actor(in, &sd, direct_splice_actor);
1328
	if (ret > 0)
1329
		*ppos = sd.pos;
1330

J
Jens Axboe 已提交
1331
	return ret;
1332
}
1333
EXPORT_SYMBOL(do_splice_direct);
1334

1335 1336 1337
static int splice_pipe_to_pipe(struct pipe_inode_info *ipipe,
			       struct pipe_inode_info *opipe,
			       size_t len, unsigned int flags);
1338

1339 1340 1341
/*
 * Determine where to splice to/from.
 */
1342 1343 1344
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)
1345
{
1346 1347
	struct pipe_inode_info *ipipe;
	struct pipe_inode_info *opipe;
1348
	loff_t offset;
J
Jens Axboe 已提交
1349
	long ret;
1350

1351 1352
	ipipe = get_pipe_info(in);
	opipe = get_pipe_info(out);
1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371

	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) {
1372 1373
		if (off_in)
			return -ESPIPE;
1374
		if (off_out) {
1375
			if (!(out->f_mode & FMODE_PWRITE))
1376
				return -EINVAL;
1377
			if (copy_from_user(&offset, off_out, sizeof(loff_t)))
1378
				return -EFAULT;
1379 1380 1381
		} else {
			offset = out->f_pos;
		}
1382

1383 1384 1385 1386 1387 1388 1389 1390 1391 1392
		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;

1393
		file_start_write(out);
1394
		ret = do_splice_from(ipipe, out, &offset, len, flags);
1395
		file_end_write(out);
J
Jens Axboe 已提交
1396

1397 1398 1399
		if (!off_out)
			out->f_pos = offset;
		else if (copy_to_user(off_out, &offset, sizeof(loff_t)))
J
Jens Axboe 已提交
1400 1401 1402
			ret = -EFAULT;

		return ret;
1403
	}
1404

1405
	if (opipe) {
1406 1407
		if (off_out)
			return -ESPIPE;
1408
		if (off_in) {
1409
			if (!(in->f_mode & FMODE_PREAD))
1410
				return -EINVAL;
1411
			if (copy_from_user(&offset, off_in, sizeof(loff_t)))
1412
				return -EFAULT;
1413 1414 1415
		} else {
			offset = in->f_pos;
		}
1416

1417
		ret = do_splice_to(in, &offset, opipe, len, flags);
J
Jens Axboe 已提交
1418

1419 1420 1421
		if (!off_in)
			in->f_pos = offset;
		else if (copy_to_user(off_in, &offset, sizeof(loff_t)))
J
Jens Axboe 已提交
1422 1423 1424
			ret = -EFAULT;

		return ret;
1425
	}
1426 1427 1428 1429

	return -EINVAL;
}

1430 1431 1432 1433 1434 1435 1436 1437 1438
/*
 * 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,
1439
				struct partial_page *partial, bool aligned,
1440
				unsigned int pipe_buffers)
1441 1442 1443 1444 1445
{
	int buffers = 0, error = 0;

	while (nr_vecs) {
		unsigned long off, npages;
1446
		struct iovec entry;
1447 1448 1449 1450
		void __user *base;
		size_t len;
		int i;

1451
		error = -EFAULT;
N
Nick Piggin 已提交
1452
		if (copy_from_user(&entry, iov, sizeof(entry)))
1453 1454
			break;

1455 1456 1457
		base = entry.iov_base;
		len = entry.iov_len;

1458 1459 1460
		/*
		 * Sanity check this iovec. 0 read succeeds.
		 */
1461
		error = 0;
1462 1463 1464
		if (unlikely(!len))
			break;
		error = -EFAULT;
1465
		if (!access_ok(VERIFY_READ, base, len))
1466 1467 1468 1469 1470 1471 1472
			break;

		/*
		 * Get this base offset and number of pages, then map
		 * in the user pages.
		 */
		off = (unsigned long) base & ~PAGE_MASK;
1473 1474 1475 1476 1477 1478 1479 1480 1481

		/*
		 * 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;

1482
		npages = (off + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
1483 1484
		if (npages > pipe_buffers - buffers)
			npages = pipe_buffers - buffers;
1485

N
Nick Piggin 已提交
1486 1487
		error = get_user_pages_fast((unsigned long)base, npages,
					0, &pages[buffers]);
1488 1489 1490 1491 1492 1493 1494 1495

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

		/*
		 * Fill this contiguous range into the partial page map.
		 */
		for (i = 0; i < error; i++) {
1496
			const int plen = min_t(size_t, len, PAGE_SIZE - off);
1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518

			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.
		 */
1519
		if (error < npages || buffers == pipe_buffers)
1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531
			break;

		nr_vecs--;
		iov++;
	}

	if (buffers)
		return buffers;

	return error;
}

1532 1533 1534
static int pipe_to_user(struct pipe_inode_info *pipe, struct pipe_buffer *buf,
			struct splice_desc *sd)
{
1535 1536
	int n = copy_page_to_iter(buf->page, buf->offset, sd->len, sd->u.data);
	return n == sd->len ? n : -EFAULT;
1537 1538 1539 1540 1541 1542
}

/*
 * For lack of a better implementation, implement vmsplice() to userspace
 * as a simple copy of the pipes pages to the user iov.
 */
1543
static long vmsplice_to_user(struct file *file, const struct iovec __user *uiov,
1544 1545 1546 1547 1548
			     unsigned long nr_segs, unsigned int flags)
{
	struct pipe_inode_info *pipe;
	struct splice_desc sd;
	long ret;
1549 1550 1551
	struct iovec iovstack[UIO_FASTIOV];
	struct iovec *iov = iovstack;
	struct iov_iter iter;
M
Miklos Szeredi 已提交
1552
	ssize_t count;
1553

1554
	pipe = get_pipe_info(file);
1555 1556 1557
	if (!pipe)
		return -EBADF;

1558 1559 1560
	ret = rw_copy_check_uvector(READ, uiov, nr_segs,
				    ARRAY_SIZE(iovstack), iovstack, &iov);
	if (ret <= 0)
M
Miklos Szeredi 已提交
1561
		goto out;
1562

M
Miklos Szeredi 已提交
1563
	count = ret;
A
Al Viro 已提交
1564
	iov_iter_init(&iter, READ, iov, nr_segs, count);
1565

1566 1567 1568 1569 1570
	sd.len = 0;
	sd.total_len = count;
	sd.flags = flags;
	sd.u.data = &iter;
	sd.pos = 0;
1571

1572 1573
	pipe_lock(pipe);
	ret = __splice_from_pipe(pipe, &sd, pipe_to_user);
1574
	pipe_unlock(pipe);
1575

M
Miklos Szeredi 已提交
1576
out:
1577 1578
	if (iov != iovstack)
		kfree(iov);
1579 1580 1581 1582

	return ret;
}

1583 1584 1585 1586 1587
/*
 * 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.
 */
1588 1589
static long vmsplice_to_pipe(struct file *file, const struct iovec __user *iov,
			     unsigned long nr_segs, unsigned int flags)
1590
{
1591
	struct pipe_inode_info *pipe;
1592 1593
	struct page *pages[PIPE_DEF_BUFFERS];
	struct partial_page partial[PIPE_DEF_BUFFERS];
1594 1595 1596
	struct splice_pipe_desc spd = {
		.pages = pages,
		.partial = partial,
1597
		.nr_pages_max = PIPE_DEF_BUFFERS,
1598 1599
		.flags = flags,
		.ops = &user_page_pipe_buf_ops,
1600
		.spd_release = spd_release_page,
1601
	};
1602
	long ret;
1603

1604
	pipe = get_pipe_info(file);
1605
	if (!pipe)
1606 1607
		return -EBADF;

1608 1609 1610 1611
	if (splice_grow_spd(pipe, &spd))
		return -ENOMEM;

	spd.nr_pages = get_iovec_page_array(iov, nr_segs, spd.pages,
1612
					    spd.partial, false,
1613
					    spd.nr_pages_max);
1614
	if (spd.nr_pages <= 0)
1615 1616 1617
		ret = spd.nr_pages;
	else
		ret = splice_to_pipe(pipe, &spd);
1618

1619
	splice_shrink_spd(&spd);
1620
	return ret;
1621 1622
}

1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638
/*
 * 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().
 *
 */
1639 1640
SYSCALL_DEFINE4(vmsplice, int, fd, const struct iovec __user *, iov,
		unsigned long, nr_segs, unsigned int, flags)
1641
{
1642
	struct fd f;
1643 1644
	long error;

1645 1646 1647 1648 1649
	if (unlikely(nr_segs > UIO_MAXIOV))
		return -EINVAL;
	else if (unlikely(!nr_segs))
		return 0;

1650
	error = -EBADF;
1651 1652 1653 1654 1655 1656 1657 1658
	f = fdget(fd);
	if (f.file) {
		if (f.file->f_mode & FMODE_WRITE)
			error = vmsplice_to_pipe(f.file, iov, nr_segs, flags);
		else if (f.file->f_mode & FMODE_READ)
			error = vmsplice_to_user(f.file, iov, nr_segs, flags);

		fdput(f);
1659 1660 1661 1662 1663
	}

	return error;
}

1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684
#ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE4(vmsplice, int, fd, const struct compat_iovec __user *, iov32,
		    unsigned int, nr_segs, unsigned int, flags)
{
	unsigned i;
	struct iovec __user *iov;
	if (nr_segs > UIO_MAXIOV)
		return -EINVAL;
	iov = compat_alloc_user_space(nr_segs * sizeof(struct iovec));
	for (i = 0; i < nr_segs; i++) {
		struct compat_iovec v;
		if (get_user(v.iov_base, &iov32[i].iov_base) ||
		    get_user(v.iov_len, &iov32[i].iov_len) ||
		    put_user(compat_ptr(v.iov_base), &iov[i].iov_base) ||
		    put_user(v.iov_len, &iov[i].iov_len))
			return -EFAULT;
	}
	return sys_vmsplice(fd, iov, nr_segs, flags);
}
#endif

1685 1686 1687
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)
1688
{
1689
	struct fd in, out;
1690 1691 1692 1693 1694 1695
	long error;

	if (unlikely(!len))
		return 0;

	error = -EBADF;
1696 1697 1698 1699 1700 1701 1702 1703
	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,
1704
							  len, flags);
1705
				fdput(out);
1706 1707
			}
		}
1708
		fdput(in);
1709 1710 1711
	}
	return error;
}
1712

1713 1714 1715 1716
/*
 * Make sure there's data to read. Wait for input if we can, otherwise
 * return an appropriate error.
 */
1717
static int ipipe_prep(struct pipe_inode_info *pipe, unsigned int flags)
1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728
{
	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;
1729
	pipe_lock(pipe);
1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746

	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);
	}

1747
	pipe_unlock(pipe);
1748 1749 1750 1751 1752 1753 1754
	return ret;
}

/*
 * Make sure there's writeable room. Wait for room if we can, otherwise
 * return an appropriate error.
 */
1755
static int opipe_prep(struct pipe_inode_info *pipe, unsigned int flags)
1756 1757 1758 1759 1760 1761 1762
{
	int ret;

	/*
	 * Check ->nrbufs without the inode lock first. This function
	 * is speculative anyways, so missing one is ok.
	 */
1763
	if (pipe->nrbufs < pipe->buffers)
1764 1765 1766
		return 0;

	ret = 0;
1767
	pipe_lock(pipe);
1768

1769
	while (pipe->nrbufs >= pipe->buffers) {
1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787
		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--;
	}

1788
	pipe_unlock(pipe);
1789 1790 1791
	return ret;
}

1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834
/*
 * 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.
		 */
1835
		if (!ipipe->nrbufs || opipe->nrbufs >= opipe->buffers) {
1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855
			/* 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;
1856
		nbuf = (opipe->curbuf + opipe->nrbufs) & (opipe->buffers - 1);
1857 1858 1859 1860 1861 1862 1863 1864 1865
		obuf = opipe->bufs + nbuf;

		if (len >= ibuf->len) {
			/*
			 * Simply move the whole buffer from ipipe to opipe
			 */
			*obuf = *ibuf;
			ibuf->ops = NULL;
			opipe->nrbufs++;
1866
			ipipe->curbuf = (ipipe->curbuf + 1) & (ipipe->buffers - 1);
1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897
			ipipe->nrbufs--;
			input_wakeup = true;
		} else {
			/*
			 * Get a reference to this pipe buffer,
			 * so we can copy the contents over.
			 */
			ibuf->ops->get(ipipe, ibuf);
			*obuf = *ibuf;

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

			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
1898 1899 1900
	if (ret > 0)
		wakeup_pipe_readers(opipe);

1901 1902 1903 1904 1905 1906
	if (input_wakeup)
		wakeup_pipe_writers(ipipe);

	return ret;
}

1907 1908 1909 1910 1911 1912 1913 1914
/*
 * 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;
1915
	int ret = 0, i = 0, nbuf;
1916 1917 1918

	/*
	 * Potential ABBA deadlock, work around it by ordering lock
1919
	 * grabbing by pipe info address. Otherwise two different processes
1920 1921
	 * could deadlock (one doing tee from A -> B, the other from B -> A).
	 */
1922
	pipe_double_lock(ipipe, opipe);
1923

1924
	do {
1925 1926 1927 1928 1929 1930 1931
		if (!opipe->readers) {
			send_sig(SIGPIPE, current, 0);
			if (!ret)
				ret = -EPIPE;
			break;
		}

1932 1933 1934 1935
		/*
		 * If we have iterated all input buffers or ran out of
		 * output room, break.
		 */
1936
		if (i >= ipipe->nrbufs || opipe->nrbufs >= opipe->buffers)
1937
			break;
1938

1939 1940
		ibuf = ipipe->bufs + ((ipipe->curbuf + i) & (ipipe->buffers-1));
		nbuf = (opipe->curbuf + opipe->nrbufs) & (opipe->buffers - 1);
1941 1942

		/*
1943 1944
		 * Get a reference to this pipe buffer,
		 * so we can copy the contents over.
1945
		 */
1946 1947 1948 1949 1950
		ibuf->ops->get(ipipe, ibuf);

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

1951
		/*
1952 1953
		 * Don't inherit the gift flag, we need to
		 * prevent multiple steals of this page.
1954
		 */
1955
		obuf->flags &= ~PIPE_BUF_FLAG_GIFT;
1956

1957 1958
		if (obuf->len > len)
			obuf->len = len;
1959

1960 1961 1962 1963 1964
		opipe->nrbufs++;
		ret += obuf->len;
		len -= obuf->len;
		i++;
	} while (len);
1965

1966 1967 1968 1969 1970 1971 1972
	/*
	 * 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;

1973 1974
	pipe_unlock(ipipe);
	pipe_unlock(opipe);
1975

1976 1977 1978
	/*
	 * If we put data in the output pipe, wakeup any potential readers.
	 */
N
Namhyung Kim 已提交
1979 1980
	if (ret > 0)
		wakeup_pipe_readers(opipe);
1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993

	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)
{
1994 1995
	struct pipe_inode_info *ipipe = get_pipe_info(in);
	struct pipe_inode_info *opipe = get_pipe_info(out);
1996
	int ret = -EINVAL;
1997 1998

	/*
1999 2000
	 * Duplicate the contents of ipipe to opipe without actually
	 * copying the data.
2001
	 */
2002 2003 2004 2005 2006
	if (ipipe && opipe && ipipe != opipe) {
		/*
		 * Keep going, unless we encounter an error. The ipipe/opipe
		 * ordering doesn't really matter.
		 */
2007
		ret = ipipe_prep(ipipe, flags);
2008
		if (!ret) {
2009
			ret = opipe_prep(opipe, flags);
2010
			if (!ret)
2011 2012 2013
				ret = link_pipe(ipipe, opipe, len, flags);
		}
	}
2014

2015
	return ret;
2016 2017
}

2018
SYSCALL_DEFINE4(tee, int, fdin, int, fdout, size_t, len, unsigned int, flags)
2019
{
2020 2021
	struct fd in;
	int error;
2022 2023 2024 2025 2026

	if (unlikely(!len))
		return 0;

	error = -EBADF;
2027 2028 2029 2030 2031 2032 2033 2034 2035
	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);
2036 2037
			}
		}
2038
 		fdput(in);
2039 2040 2041 2042
 	}

	return error;
}