splice.c 47.1 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>
35
#include "internal.h"
36

37 38 39 40 41 42
/*
 * 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.
 */
43
static int page_cache_pipe_buf_steal(struct pipe_inode_info *pipe,
44 45 46
				     struct pipe_buffer *buf)
{
	struct page *page = buf->page;
47
	struct address_space *mapping;
48

49 50
	lock_page(page);

51 52 53
	mapping = page_mapping(page);
	if (mapping) {
		WARN_ON(!PageUptodate(page));
54

55 56 57 58 59 60 61 62 63
		/*
		 * 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);
64

65 66
		if (page_has_private(page) &&
		    !try_to_release_page(page, GFP_KERNEL))
67
			goto out_unlock;
68

69 70 71 72 73 74 75 76
		/*
		 * 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;
		}
77
	}
78

79 80 81 82
	/*
	 * Raced with truncate or failed to remove page from current
	 * address space, unlock and return failure.
	 */
83
out_unlock:
84 85
	unlock_page(page);
	return 1;
86 87
}

88
static void page_cache_pipe_buf_release(struct pipe_inode_info *pipe,
89 90 91
					struct pipe_buffer *buf)
{
	page_cache_release(buf->page);
J
Jens Axboe 已提交
92
	buf->flags &= ~PIPE_BUF_FLAG_LRU;
93 94
}

95 96 97 98
/*
 * Check whether the contents of buf is OK to access. Since the content
 * is a page cache page, IO may be in flight.
 */
99 100
static int page_cache_pipe_buf_confirm(struct pipe_inode_info *pipe,
				       struct pipe_buffer *buf)
101 102
{
	struct page *page = buf->page;
103
	int err;
104 105

	if (!PageUptodate(page)) {
106 107 108 109
		lock_page(page);

		/*
		 * Page got truncated/unhashed. This will cause a 0-byte
I
Ingo Molnar 已提交
110
		 * splice, if this is the first page.
111 112 113 114 115
		 */
		if (!page->mapping) {
			err = -ENODATA;
			goto error;
		}
116

117
		/*
I
Ingo Molnar 已提交
118
		 * Uh oh, read-error from disk.
119 120 121 122 123 124 125
		 */
		if (!PageUptodate(page)) {
			err = -EIO;
			goto error;
		}

		/*
126
		 * Page is ok afterall, we are done.
127
		 */
128 129 130
		unlock_page(page);
	}

131
	return 0;
132 133
error:
	unlock_page(page);
134
	return err;
135 136
}

137
const struct pipe_buf_operations page_cache_pipe_buf_ops = {
138
	.can_merge = 0,
139 140
	.map = generic_pipe_buf_map,
	.unmap = generic_pipe_buf_unmap,
141
	.confirm = page_cache_pipe_buf_confirm,
142
	.release = page_cache_pipe_buf_release,
143
	.steal = page_cache_pipe_buf_steal,
144
	.get = generic_pipe_buf_get,
145 146
};

147 148 149
static int user_page_pipe_buf_steal(struct pipe_inode_info *pipe,
				    struct pipe_buffer *buf)
{
150 151 152
	if (!(buf->flags & PIPE_BUF_FLAG_GIFT))
		return 1;

J
Jens Axboe 已提交
153
	buf->flags |= PIPE_BUF_FLAG_LRU;
154
	return generic_pipe_buf_steal(pipe, buf);
155 156
}

157
static const struct pipe_buf_operations user_page_pipe_buf_ops = {
158
	.can_merge = 0,
159 160
	.map = generic_pipe_buf_map,
	.unmap = generic_pipe_buf_unmap,
161
	.confirm = generic_pipe_buf_confirm,
162 163
	.release = page_cache_pipe_buf_release,
	.steal = user_page_pipe_buf_steal,
164
	.get = generic_pipe_buf_get,
165 166
};

N
Namhyung Kim 已提交
167 168 169 170 171 172 173 174
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);
}

175 176 177 178 179 180
/**
 * splice_to_pipe - fill passed data into a pipe
 * @pipe:	pipe to fill
 * @spd:	data to fill
 *
 * Description:
R
Randy Dunlap 已提交
181
 *    @spd contains a map of pages and len/offset tuples, along with
182 183 184
 *    the struct pipe_buf_operations associated with these pages. This
 *    function will link that data to the pipe.
 *
185
 */
186 187
ssize_t splice_to_pipe(struct pipe_inode_info *pipe,
		       struct splice_pipe_desc *spd)
188
{
189
	unsigned int spd_pages = spd->nr_pages;
190
	int ret, do_wakeup, page_nr;
191 192 193

	ret = 0;
	do_wakeup = 0;
194
	page_nr = 0;
195

196
	pipe_lock(pipe);
197 198

	for (;;) {
199
		if (!pipe->readers) {
200 201 202 203 204 205
			send_sig(SIGPIPE, current, 0);
			if (!ret)
				ret = -EPIPE;
			break;
		}

206 207
		if (pipe->nrbufs < pipe->buffers) {
			int newbuf = (pipe->curbuf + pipe->nrbufs) & (pipe->buffers - 1);
208
			struct pipe_buffer *buf = pipe->bufs + newbuf;
209

210 211 212
			buf->page = spd->pages[page_nr];
			buf->offset = spd->partial[page_nr].offset;
			buf->len = spd->partial[page_nr].len;
213
			buf->private = spd->partial[page_nr].private;
214
			buf->ops = spd->ops;
215 216 217
			if (spd->flags & SPLICE_F_GIFT)
				buf->flags |= PIPE_BUF_FLAG_GIFT;

218
			pipe->nrbufs++;
219 220 221
			page_nr++;
			ret += buf->len;

A
Al Viro 已提交
222
			if (pipe->files)
223
				do_wakeup = 1;
224

225
			if (!--spd->nr_pages)
226
				break;
227
			if (pipe->nrbufs < pipe->buffers)
228 229 230 231 232
				continue;

			break;
		}

233
		if (spd->flags & SPLICE_F_NONBLOCK) {
234 235 236 237 238
			if (!ret)
				ret = -EAGAIN;
			break;
		}

239 240 241 242 243 244 245
		if (signal_pending(current)) {
			if (!ret)
				ret = -ERESTARTSYS;
			break;
		}

		if (do_wakeup) {
246
			smp_mb();
247 248 249
			if (waitqueue_active(&pipe->wait))
				wake_up_interruptible_sync(&pipe->wait);
			kill_fasync(&pipe->fasync_readers, SIGIO, POLL_IN);
250 251 252
			do_wakeup = 0;
		}

253 254 255
		pipe->waiting_writers++;
		pipe_wait(pipe);
		pipe->waiting_writers--;
256 257
	}

258
	pipe_unlock(pipe);
259

N
Namhyung Kim 已提交
260 261
	if (do_wakeup)
		wakeup_pipe_readers(pipe);
262

263
	while (page_nr < spd_pages)
264
		spd->spd_release(spd, page_nr++);
265 266 267 268

	return ret;
}

269
void spd_release_page(struct splice_pipe_desc *spd, unsigned int i)
270 271 272 273
{
	page_cache_release(spd->pages[i]);
}

274 275 276 277
/*
 * Check if we need to grow the arrays holding pages and partial page
 * descriptions.
 */
278
int splice_grow_spd(const struct pipe_inode_info *pipe, struct splice_pipe_desc *spd)
279
{
280 281 282 283
	unsigned int buffers = ACCESS_ONCE(pipe->buffers);

	spd->nr_pages_max = buffers;
	if (buffers <= PIPE_DEF_BUFFERS)
284 285
		return 0;

286 287
	spd->pages = kmalloc(buffers * sizeof(struct page *), GFP_KERNEL);
	spd->partial = kmalloc(buffers * sizeof(struct partial_page), GFP_KERNEL);
288 289 290 291 292 293 294 295 296

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

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

297
void splice_shrink_spd(struct splice_pipe_desc *spd)
298
{
299
	if (spd->nr_pages_max <= PIPE_DEF_BUFFERS)
300 301 302 303 304 305
		return;

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

306
static int
307 308 309
__generic_file_splice_read(struct file *in, loff_t *ppos,
			   struct pipe_inode_info *pipe, size_t len,
			   unsigned int flags)
310 311
{
	struct address_space *mapping = in->f_mapping;
F
Fengguang Wu 已提交
312
	unsigned int loff, nr_pages, req_pages;
313 314
	struct page *pages[PIPE_DEF_BUFFERS];
	struct partial_page partial[PIPE_DEF_BUFFERS];
315
	struct page *page;
316 317
	pgoff_t index, end_index;
	loff_t isize;
318
	int error, page_nr;
319 320 321
	struct splice_pipe_desc spd = {
		.pages = pages,
		.partial = partial,
322
		.nr_pages_max = PIPE_DEF_BUFFERS,
323 324
		.flags = flags,
		.ops = &page_cache_pipe_buf_ops,
325
		.spd_release = spd_release_page,
326
	};
327

328 329 330
	if (splice_grow_spd(pipe, &spd))
		return -ENOMEM;

331
	index = *ppos >> PAGE_CACHE_SHIFT;
332
	loff = *ppos & ~PAGE_CACHE_MASK;
F
Fengguang Wu 已提交
333
	req_pages = (len + loff + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
334
	nr_pages = min(req_pages, spd.nr_pages_max);
335

336 337 338
	/*
	 * Lookup the (hopefully) full range of pages we need.
	 */
339
	spd.nr_pages = find_get_pages_contig(mapping, index, nr_pages, spd.pages);
340
	index += spd.nr_pages;
341

342 343
	/*
	 * If find_get_pages_contig() returned fewer pages than we needed,
344
	 * readahead/allocate the rest and fill in the holes.
345
	 */
346
	if (spd.nr_pages < nr_pages)
347 348
		page_cache_sync_readahead(mapping, &in->f_ra, in,
				index, req_pages - spd.nr_pages);
349

350
	error = 0;
351
	while (spd.nr_pages < nr_pages) {
352
		/*
353 354
		 * Page could be there, find_get_pages_contig() breaks on
		 * the first hole.
355
		 */
356 357 358
		page = find_get_page(mapping, index);
		if (!page) {
			/*
359
			 * page didn't exist, allocate one.
360 361 362 363 364 365
			 */
			page = page_cache_alloc_cold(mapping);
			if (!page)
				break;

			error = add_to_page_cache_lru(page, mapping, index,
366
						GFP_KERNEL);
367 368
			if (unlikely(error)) {
				page_cache_release(page);
369 370
				if (error == -EEXIST)
					continue;
371 372
				break;
			}
373 374 375 376 377
			/*
			 * add_to_page_cache() locks the page, unlock it
			 * to avoid convoluting the logic below even more.
			 */
			unlock_page(page);
378 379
		}

380
		spd.pages[spd.nr_pages++] = page;
381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400
		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);
401
		page = spd.pages[page_nr];
402

403
		if (PageReadahead(page))
404
			page_cache_async_readahead(mapping, &in->f_ra, in,
F
Fengguang Wu 已提交
405
					page, index, req_pages - page_nr);
406

407 408 409 410
		/*
		 * If the page isn't uptodate, we may need to start io on it
		 */
		if (!PageUptodate(page)) {
411
			lock_page(page);
412 413

			/*
414 415 416 417
			 * 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.
418 419 420
			 */
			if (!page->mapping) {
				unlock_page(page);
421 422 423 424 425 426 427
				page = find_or_create_page(mapping, index,
						mapping_gfp_mask(mapping));

				if (!page) {
					error = -ENOMEM;
					break;
				}
428 429
				page_cache_release(spd.pages[page_nr]);
				spd.pages[page_nr] = page;
430 431 432 433 434 435 436 437
			}
			/*
			 * page was already under io and is now done, great
			 */
			if (PageUptodate(page)) {
				unlock_page(page);
				goto fill_it;
			}
438

439 440 441 442
			/*
			 * need to read in the page
			 */
			error = mapping->a_ops->readpage(in, page);
443
			if (unlikely(error)) {
444 445 446 447 448 449
				/*
				 * 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.
				 */
450
				if (error == AOP_TRUNCATED_PAGE)
451 452
					error = 0;

453 454
				break;
			}
455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470
		}
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;
471 472

			/*
473
			 * max good bytes in this page
474
			 */
475 476
			plen = ((isize - 1) & ~PAGE_CACHE_MASK) + 1;
			if (plen <= loff)
477 478 479
				break;

			/*
480
			 * force quit after adding this page
481
			 */
482 483
			this_len = min(this_len, plen - loff);
			len = this_len;
484
		}
485

486 487
		spd.partial[page_nr].offset = loff;
		spd.partial[page_nr].len = this_len;
488
		len -= this_len;
489
		loff = 0;
490 491
		spd.nr_pages++;
		index++;
492 493
	}

494
	/*
495
	 * Release any pages at the end, if we quit early. 'page_nr' is how far
496 497 498
	 * we got, 'nr_pages' is how many pages are in the map.
	 */
	while (page_nr < nr_pages)
499
		page_cache_release(spd.pages[page_nr++]);
500
	in->f_ra.prev_pos = (loff_t)index << PAGE_CACHE_SHIFT;
501

502
	if (spd.nr_pages)
503
		error = splice_to_pipe(pipe, &spd);
504

505
	splice_shrink_spd(&spd);
506
	return error;
507 508
}

509 510 511
/**
 * generic_file_splice_read - splice data from file to a pipe
 * @in:		file to splice from
512
 * @ppos:	position in @in
513 514 515 516
 * @pipe:	pipe to splice to
 * @len:	number of bytes to splice
 * @flags:	splice modifier flags
 *
517 518 519 520 521
 * 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.
 *
522
 */
523 524 525
ssize_t generic_file_splice_read(struct file *in, loff_t *ppos,
				 struct pipe_inode_info *pipe, size_t len,
				 unsigned int flags)
526
{
527
	loff_t isize, left;
528
	int ret;
529 530 531 532 533 534 535 536

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

	left = isize - *ppos;
	if (unlikely(left < len))
		len = left;
537

538
	ret = __generic_file_splice_read(in, ppos, pipe, len, flags);
539
	if (ret > 0) {
540
		*ppos += ret;
541 542
		file_accessed(in);
	}
543 544 545

	return ret;
}
546 547
EXPORT_SYMBOL(generic_file_splice_read);

548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573
static const struct pipe_buf_operations default_pipe_buf_ops = {
	.can_merge = 0,
	.map = generic_pipe_buf_map,
	.unmap = generic_pipe_buf_unmap,
	.confirm = generic_pipe_buf_confirm,
	.release = generic_pipe_buf_release,
	.steal = generic_pipe_buf_steal,
	.get = generic_pipe_buf_get,
};

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;
}

574
ssize_t kernel_write(struct file *file, const char *buf, size_t count,
575
			    loff_t pos)
576 577 578 579 580 581 582
{
	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() */
583
	res = vfs_write(file, (__force const char __user *)buf, count, &pos);
584 585 586 587
	set_fs(old_fs);

	return res;
}
588
EXPORT_SYMBOL(kernel_write);
589

590 591 592 593 594 595 596
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;
597 598 599
	struct page *pages[PIPE_DEF_BUFFERS];
	struct partial_page partial[PIPE_DEF_BUFFERS];
	struct iovec *vec, __vec[PIPE_DEF_BUFFERS];
600 601 602 603 604 605 606
	ssize_t res;
	size_t this_len;
	int error;
	int i;
	struct splice_pipe_desc spd = {
		.pages = pages,
		.partial = partial,
607
		.nr_pages_max = PIPE_DEF_BUFFERS,
608 609 610 611 612
		.flags = flags,
		.ops = &default_pipe_buf_ops,
		.spd_release = spd_release_page,
	};

613 614 615 616 617
	if (splice_grow_spd(pipe, &spd))
		return -ENOMEM;

	res = -ENOMEM;
	vec = __vec;
618 619
	if (spd.nr_pages_max > PIPE_DEF_BUFFERS) {
		vec = kmalloc(spd.nr_pages_max * sizeof(struct iovec), GFP_KERNEL);
620 621 622 623
		if (!vec)
			goto shrink_ret;
	}

624 625 626
	offset = *ppos & ~PAGE_CACHE_MASK;
	nr_pages = (len + offset + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;

627
	for (i = 0; i < nr_pages && i < spd.nr_pages_max && len; i++) {
628 629
		struct page *page;

630
		page = alloc_page(GFP_USER);
631 632 633 634 635
		error = -ENOMEM;
		if (!page)
			goto err;

		this_len = min_t(size_t, len, PAGE_CACHE_SIZE - offset);
636
		vec[i].iov_base = (void __user *) page_address(page);
637
		vec[i].iov_len = this_len;
638
		spd.pages[i] = page;
639 640 641 642 643 644
		spd.nr_pages++;
		len -= this_len;
		offset = 0;
	}

	res = kernel_readv(in, vec, spd.nr_pages, *ppos);
A
Andrew Morton 已提交
645 646
	if (res < 0) {
		error = res;
647
		goto err;
A
Andrew Morton 已提交
648
	}
649 650 651 652 653 654 655 656

	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);
657 658
		spd.partial[i].offset = 0;
		spd.partial[i].len = this_len;
659
		if (!this_len) {
660 661
			__free_page(spd.pages[i]);
			spd.pages[i] = NULL;
662 663 664 665 666 667 668 669 670 671
			nr_freed++;
		}
		res -= this_len;
	}
	spd.nr_pages -= nr_freed;

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

672 673 674
shrink_ret:
	if (vec != __vec)
		kfree(vec);
675
	splice_shrink_spd(&spd);
676 677 678
	return res;

err:
679
	for (i = 0; i < spd.nr_pages; i++)
680
		__free_page(spd.pages[i]);
681

682 683
	res = error;
	goto shrink_ret;
684 685 686
}
EXPORT_SYMBOL(default_file_splice_read);

687
/*
688
 * Send 'sd->len' bytes to socket from 'sd->file' at position 'sd->pos'
J
Jens Axboe 已提交
689
 * using sendpage(). Return the number of bytes sent.
690
 */
691
static int pipe_to_sendpage(struct pipe_inode_info *pipe,
692 693
			    struct pipe_buffer *buf, struct splice_desc *sd)
{
694
	struct file *file = sd->u.file;
695
	loff_t pos = sd->pos;
696
	int more;
697

698 699 700
	if (!likely(file->f_op && file->f_op->sendpage))
		return -EINVAL;

701
	more = (sd->flags & SPLICE_F_MORE) ? MSG_MORE : 0;
702 703

	if (sd->len < sd->total_len && pipe->nrbufs > 1)
704
		more |= MSG_SENDPAGE_NOTLAST;
705

706 707
	return file->f_op->sendpage(file, buf->page, buf->offset,
				    sd->len, &pos, more);
708 709 710 711 712 713 714 715 716 717 718 719 720 721 722
}

/*
 * This is a little more tricky than the file -> pipe splicing. There are
 * basically three cases:
 *
 *	- Destination page already exists in the address space and there
 *	  are users of it. For that case we have no other option that
 *	  copying the data. Tough luck.
 *	- Destination page already exists in the address space, but there
 *	  are no users of it. Make sure it's uptodate, then drop it. Fall
 *	  through to last case.
 *	- Destination page does not exist, we can add the pipe page to
 *	  the page cache and avoid the copy.
 *
723 724 725 726 727 728
 * If asked to move pages to the output file (SPLICE_F_MOVE is set in
 * sd->flags), we attempt to migrate pages from the pipe to the output
 * file address space page cache. This is possible if no one else has
 * the pipe page referenced outside of the pipe and page cache. If
 * SPLICE_F_MOVE isn't set, or we cannot move the page, we simply create
 * a new page in the output file page cache and fill/dirty that.
729
 */
730 731
int pipe_to_file(struct pipe_inode_info *pipe, struct pipe_buffer *buf,
		 struct splice_desc *sd)
732
{
733
	struct file *file = sd->u.file;
734
	struct address_space *mapping = file->f_mapping;
J
Jens Axboe 已提交
735
	unsigned int offset, this_len;
736
	struct page *page;
737
	void *fsdata;
738
	int ret;
739 740 741

	offset = sd->pos & ~PAGE_CACHE_MASK;

J
Jens Axboe 已提交
742 743 744 745
	this_len = sd->len;
	if (this_len + offset > PAGE_CACHE_SIZE)
		this_len = PAGE_CACHE_SIZE - offset;

746 747 748 749
	ret = pagecache_write_begin(file, mapping, sd->pos, this_len,
				AOP_FLAG_UNINTERRUPTIBLE, &page, &fsdata);
	if (unlikely(ret))
		goto out;
750

751
	if (buf->page != page) {
752
		char *src = buf->ops->map(pipe, buf, 1);
753
		char *dst = kmap_atomic(page);
754

J
Jens Axboe 已提交
755
		memcpy(dst + offset, src + buf->offset, this_len);
756
		flush_dcache_page(page);
757
		kunmap_atomic(dst);
758
		buf->ops->unmap(pipe, buf, src);
759
	}
760 761
	ret = pagecache_write_end(file, mapping, sd->pos, this_len, this_len,
				page, fsdata);
762 763 764
out:
	return ret;
}
765
EXPORT_SYMBOL(pipe_to_file);
766

767 768 769 770 771 772 773 774
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);
}

775
/**
776
 * splice_from_pipe_feed - feed available data from a pipe to a file
777 778 779 780 781
 * @pipe:	pipe to splice from
 * @sd:		information to @actor
 * @actor:	handler that splices the data
 *
 * Description:
782 783 784 785 786 787 788
 *    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.
789
 *
790 791 792 793
 *    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.
794
 */
795 796
int splice_from_pipe_feed(struct pipe_inode_info *pipe, struct splice_desc *sd,
			  splice_actor *actor)
797
{
798
	int ret;
799

800 801 802
	while (pipe->nrbufs) {
		struct pipe_buffer *buf = pipe->bufs + pipe->curbuf;
		const struct pipe_buf_operations *ops = buf->ops;
803

804 805 806
		sd->len = buf->len;
		if (sd->len > sd->total_len)
			sd->len = sd->total_len;
807

808 809
		ret = buf->ops->confirm(pipe, buf);
		if (unlikely(ret)) {
810 811 812 813
			if (ret == -ENODATA)
				ret = 0;
			return ret;
		}
814 815 816 817 818

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

819 820 821 822 823 824 825 826 827 828 829
		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);
830
			pipe->curbuf = (pipe->curbuf + 1) & (pipe->buffers - 1);
831
			pipe->nrbufs--;
A
Al Viro 已提交
832
			if (pipe->files)
833 834
				sd->need_wakeup = true;
		}
835

836 837 838
		if (!sd->total_len)
			return 0;
	}
839

840 841 842
	return 1;
}
EXPORT_SYMBOL(splice_from_pipe_feed);
843

844 845 846 847 848 849 850 851 852 853 854 855 856 857 858
/**
 * 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.
 */
int splice_from_pipe_next(struct pipe_inode_info *pipe, struct splice_desc *sd)
{
	while (!pipe->nrbufs) {
		if (!pipe->writers)
			return 0;
J
Jens Axboe 已提交
859

860 861
		if (!pipe->waiting_writers && sd->num_spliced)
			return 0;
I
Ingo Molnar 已提交
862

863 864
		if (sd->flags & SPLICE_F_NONBLOCK)
			return -EAGAIN;
865

866 867
		if (signal_pending(current))
			return -ERESTARTSYS;
868

869 870 871
		if (sd->need_wakeup) {
			wakeup_pipe_writers(pipe);
			sd->need_wakeup = false;
872 873
		}

874 875
		pipe_wait(pipe);
	}
876

877 878 879
	return 1;
}
EXPORT_SYMBOL(splice_from_pipe_next);
880

881 882
/**
 * splice_from_pipe_begin - start splicing from pipe
883
 * @sd:		information about the splice operation
884 885 886 887 888 889 890 891 892 893 894 895
 *
 * 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.
 */
void splice_from_pipe_begin(struct splice_desc *sd)
{
	sd->num_spliced = 0;
	sd->need_wakeup = false;
}
EXPORT_SYMBOL(splice_from_pipe_begin);
896

897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912
/**
 * 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().
 */
void splice_from_pipe_end(struct pipe_inode_info *pipe, struct splice_desc *sd)
{
	if (sd->need_wakeup)
		wakeup_pipe_writers(pipe);
}
EXPORT_SYMBOL(splice_from_pipe_end);
913

914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930
/**
 * __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;
931

932 933 934 935 936 937 938 939 940
	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;
941
}
M
Mark Fasheh 已提交
942
EXPORT_SYMBOL(__splice_from_pipe);
943

944 945 946 947 948 949 950 951 952 953
/**
 * 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:
954
 *    See __splice_from_pipe. This function locks the pipe inode,
955 956 957
 *    otherwise it's identical to __splice_from_pipe().
 *
 */
958 959 960 961 962
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
963 964 965 966
	struct splice_desc sd = {
		.total_len = len,
		.flags = flags,
		.pos = *ppos,
967
		.u.file = out,
J
Jens Axboe 已提交
968
	};
969

970
	pipe_lock(pipe);
J
Jens Axboe 已提交
971
	ret = __splice_from_pipe(pipe, &sd, actor);
972
	pipe_unlock(pipe);
973 974 975 976

	return ret;
}

977 978
/**
 * generic_file_splice_write - splice data from a pipe to a file
979
 * @pipe:	pipe info
980
 * @out:	file to write to
981
 * @ppos:	position in @out
982 983 984
 * @len:	number of bytes to splice
 * @flags:	splice modifier flags
 *
985 986 987
 * Description:
 *    Will either move or copy pages (determined by @flags options) from
 *    the given pipe inode to the given file.
988 989
 *
 */
990 991
ssize_t
generic_file_splice_write(struct pipe_inode_info *pipe, struct file *out,
992
			  loff_t *ppos, size_t len, unsigned int flags)
993
{
994
	struct address_space *mapping = out->f_mapping;
995
	struct inode *inode = mapping->host;
996 997 998 999 1000 1001
	struct splice_desc sd = {
		.total_len = len,
		.flags = flags,
		.pos = *ppos,
		.u.file = out,
	};
1002 1003
	ssize_t ret;

1004
	pipe_lock(pipe);
1005 1006 1007 1008 1009 1010 1011 1012 1013

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

		mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
		ret = file_remove_suid(out);
1014
		if (!ret) {
1015 1016 1017 1018
			ret = file_update_time(out);
			if (!ret)
				ret = splice_from_pipe_feed(pipe, &sd,
							    pipe_to_file);
1019
		}
1020 1021 1022 1023
		mutex_unlock(&inode->i_mutex);
	} while (ret > 0);
	splice_from_pipe_end(pipe, &sd);

1024
	pipe_unlock(pipe);
1025 1026 1027 1028

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

J
Jens Axboe 已提交
1029
	if (ret > 0) {
1030
		int err;
1031

1032 1033 1034 1035 1036
		err = generic_write_sync(out, *ppos, ret);
		if (err)
			ret = err;
		else
			*ppos += ret;
1037
		balance_dirty_pages_ratelimited(mapping);
1038 1039 1040
	}

	return ret;
1041 1042
}

1043 1044
EXPORT_SYMBOL(generic_file_splice_write);

1045 1046
static int write_pipe_buf(struct pipe_inode_info *pipe, struct pipe_buffer *buf,
			  struct splice_desc *sd)
1047
{
1048 1049
	int ret;
	void *data;
1050
	loff_t tmp = sd->pos;
1051 1052

	data = buf->ops->map(pipe, buf, 0);
1053
	ret = __kernel_write(sd->u.file, data + buf->offset, sd->len, &tmp);
1054 1055 1056
	buf->ops->unmap(pipe, buf, data);

	return ret;
1057 1058 1059 1060 1061 1062
}

static ssize_t default_file_splice_write(struct pipe_inode_info *pipe,
					 struct file *out, loff_t *ppos,
					 size_t len, unsigned int flags)
{
1063
	ssize_t ret;
1064

1065 1066 1067
	ret = splice_from_pipe(pipe, out, ppos, len, flags, write_pipe_buf);
	if (ret > 0)
		*ppos += ret;
1068

1069
	return ret;
1070 1071
}

1072 1073
/**
 * generic_splice_sendpage - splice data from a pipe to a socket
1074
 * @pipe:	pipe to splice from
1075
 * @out:	socket to write to
1076
 * @ppos:	position in @out
1077 1078 1079
 * @len:	number of bytes to splice
 * @flags:	splice modifier flags
 *
1080 1081 1082
 * Description:
 *    Will send @len bytes from the pipe to a network socket. No data copying
 *    is involved.
1083 1084
 *
 */
1085
ssize_t generic_splice_sendpage(struct pipe_inode_info *pipe, struct file *out,
1086
				loff_t *ppos, size_t len, unsigned int flags)
1087
{
1088
	return splice_from_pipe(pipe, out, ppos, len, flags, pipe_to_sendpage);
1089 1090
}

1091
EXPORT_SYMBOL(generic_splice_sendpage);
J
Jeff Garzik 已提交
1092

1093 1094 1095
/*
 * Attempt to initiate a splice from pipe to file.
 */
1096
static long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
1097
			   loff_t *ppos, size_t len, unsigned int flags)
1098
{
1099 1100
	ssize_t (*splice_write)(struct pipe_inode_info *, struct file *,
				loff_t *, size_t, unsigned int);
1101 1102
	int ret;

1103
	if (unlikely(!(out->f_mode & FMODE_WRITE)))
1104 1105
		return -EBADF;

1106 1107 1108
	if (unlikely(out->f_flags & O_APPEND))
		return -EINVAL;

1109
	ret = rw_verify_area(WRITE, out, ppos, len);
1110 1111 1112
	if (unlikely(ret < 0))
		return ret;

1113 1114 1115
	if (out->f_op && out->f_op->splice_write)
		splice_write = out->f_op->splice_write;
	else
1116 1117
		splice_write = default_file_splice_write;

1118
	return splice_write(pipe, out, ppos, len, flags);
1119 1120
}

1121 1122 1123
/*
 * Attempt to initiate a splice from a file to a pipe.
 */
1124 1125 1126
static long do_splice_to(struct file *in, loff_t *ppos,
			 struct pipe_inode_info *pipe, size_t len,
			 unsigned int flags)
1127
{
1128 1129
	ssize_t (*splice_read)(struct file *, loff_t *,
			       struct pipe_inode_info *, size_t, unsigned int);
1130 1131
	int ret;

1132
	if (unlikely(!(in->f_mode & FMODE_READ)))
1133 1134
		return -EBADF;

1135
	ret = rw_verify_area(READ, in, ppos, len);
1136 1137 1138
	if (unlikely(ret < 0))
		return ret;

1139 1140 1141
	if (in->f_op && in->f_op->splice_read)
		splice_read = in->f_op->splice_read;
	else
1142 1143 1144
		splice_read = default_file_splice_read;

	return splice_read(in, ppos, pipe, len, flags);
1145 1146
}

1147 1148 1149 1150 1151 1152 1153 1154 1155
/**
 * 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
1156
 *    pipe is cached in the process, and reused during the lifetime of
1157 1158
 *    that process.
 *
J
Jens Axboe 已提交
1159 1160 1161
 */
ssize_t splice_direct_to_actor(struct file *in, struct splice_desc *sd,
			       splice_direct_actor *actor)
1162 1163 1164 1165
{
	struct pipe_inode_info *pipe;
	long ret, bytes;
	umode_t i_mode;
J
Jens Axboe 已提交
1166 1167
	size_t len;
	int i, flags;
1168 1169 1170 1171 1172 1173

	/*
	 * 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
1174
	i_mode = file_inode(in)->i_mode;
1175 1176 1177 1178 1179 1180 1181 1182
	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;
1183
	if (unlikely(!pipe)) {
A
Al Viro 已提交
1184
		pipe = alloc_pipe_info();
1185 1186 1187 1188 1189
		if (!pipe)
			return -ENOMEM;

		/*
		 * We don't have an immediate reader, but we'll read the stuff
1190
		 * out of the pipe right after the splice_to_pipe(). So set
1191 1192 1193 1194 1195 1196 1197 1198
		 * PIPE_READERS appropriately.
		 */
		pipe->readers = 1;

		current->splice_pipe = pipe;
	}

	/*
I
Ingo Molnar 已提交
1199
	 * Do the splice.
1200 1201 1202
	 */
	ret = 0;
	bytes = 0;
J
Jens Axboe 已提交
1203 1204 1205 1206 1207 1208 1209
	len = sd->total_len;
	flags = sd->flags;

	/*
	 * Don't block on output, we have to drain the direct pipe.
	 */
	sd->flags &= ~SPLICE_F_NONBLOCK;
1210 1211

	while (len) {
1212
		size_t read_len;
1213
		loff_t pos = sd->pos, prev_pos = pos;
1214

1215
		ret = do_splice_to(in, &pos, pipe, len, flags);
1216
		if (unlikely(ret <= 0))
1217 1218 1219
			goto out_release;

		read_len = ret;
J
Jens Axboe 已提交
1220
		sd->total_len = read_len;
1221 1222 1223 1224 1225 1226

		/*
		 * 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
1227
		ret = actor(pipe, sd);
1228 1229
		if (unlikely(ret <= 0)) {
			sd->pos = prev_pos;
1230
			goto out_release;
1231
		}
1232 1233 1234

		bytes += ret;
		len -= ret;
1235
		sd->pos = pos;
1236

1237 1238
		if (ret < read_len) {
			sd->pos = prev_pos + ret;
1239
			goto out_release;
1240
		}
1241 1242
	}

1243
done:
1244
	pipe->nrbufs = pipe->curbuf = 0;
1245
	file_accessed(in);
1246 1247 1248 1249 1250 1251 1252
	return bytes;

out_release:
	/*
	 * If we did an incomplete transfer we must release
	 * the pipe buffers in question:
	 */
1253
	for (i = 0; i < pipe->buffers; i++) {
1254 1255 1256 1257 1258 1259 1260 1261
		struct pipe_buffer *buf = pipe->bufs + i;

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

1262 1263
	if (!bytes)
		bytes = ret;
J
Jens Axboe 已提交
1264

1265
	goto done;
J
Jens Axboe 已提交
1266 1267 1268 1269 1270 1271
}
EXPORT_SYMBOL(splice_direct_to_actor);

static int direct_splice_actor(struct pipe_inode_info *pipe,
			       struct splice_desc *sd)
{
1272
	struct file *file = sd->u.file;
J
Jens Axboe 已提交
1273

1274
	return do_splice_from(pipe, file, sd->opos, sd->total_len,
1275
			      sd->flags);
J
Jens Axboe 已提交
1276 1277
}

1278 1279 1280 1281 1282 1283 1284 1285 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
 * @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
1293
long do_splice_direct(struct file *in, loff_t *ppos, struct file *out,
1294
		      loff_t *opos, size_t len, unsigned int flags)
J
Jens Axboe 已提交
1295 1296 1297 1298 1299 1300
{
	struct splice_desc sd = {
		.len		= len,
		.total_len	= len,
		.flags		= flags,
		.pos		= *ppos,
1301
		.u.file		= out,
1302
		.opos		= opos,
J
Jens Axboe 已提交
1303
	};
1304
	long ret;
J
Jens Axboe 已提交
1305 1306

	ret = splice_direct_to_actor(in, &sd, direct_splice_actor);
1307
	if (ret > 0)
1308
		*ppos = sd.pos;
1309

J
Jens Axboe 已提交
1310
	return ret;
1311 1312
}

1313 1314 1315
static int splice_pipe_to_pipe(struct pipe_inode_info *ipipe,
			       struct pipe_inode_info *opipe,
			       size_t len, unsigned int flags);
1316

1317 1318 1319
/*
 * Determine where to splice to/from.
 */
1320 1321 1322
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)
1323
{
1324 1325
	struct pipe_inode_info *ipipe;
	struct pipe_inode_info *opipe;
1326
	loff_t offset;
J
Jens Axboe 已提交
1327
	long ret;
1328

1329 1330
	ipipe = get_pipe_info(in);
	opipe = get_pipe_info(out);
1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349

	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) {
1350 1351
		if (off_in)
			return -ESPIPE;
1352
		if (off_out) {
1353
			if (!(out->f_mode & FMODE_PWRITE))
1354
				return -EINVAL;
1355
			if (copy_from_user(&offset, off_out, sizeof(loff_t)))
1356
				return -EFAULT;
1357 1358 1359
		} else {
			offset = out->f_pos;
		}
1360

1361
		file_start_write(out);
1362
		ret = do_splice_from(ipipe, out, &offset, len, flags);
1363
		file_end_write(out);
J
Jens Axboe 已提交
1364

1365 1366 1367
		if (!off_out)
			out->f_pos = offset;
		else if (copy_to_user(off_out, &offset, sizeof(loff_t)))
J
Jens Axboe 已提交
1368 1369 1370
			ret = -EFAULT;

		return ret;
1371
	}
1372

1373
	if (opipe) {
1374 1375
		if (off_out)
			return -ESPIPE;
1376
		if (off_in) {
1377
			if (!(in->f_mode & FMODE_PREAD))
1378
				return -EINVAL;
1379
			if (copy_from_user(&offset, off_in, sizeof(loff_t)))
1380
				return -EFAULT;
1381 1382 1383
		} else {
			offset = in->f_pos;
		}
1384

1385
		ret = do_splice_to(in, &offset, opipe, len, flags);
J
Jens Axboe 已提交
1386

1387 1388 1389
		if (!off_in)
			in->f_pos = offset;
		else if (copy_to_user(off_in, &offset, sizeof(loff_t)))
J
Jens Axboe 已提交
1390 1391 1392
			ret = -EFAULT;

		return ret;
1393
	}
1394 1395 1396 1397

	return -EINVAL;
}

1398 1399 1400 1401 1402 1403 1404 1405 1406
/*
 * 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,
1407
				struct partial_page *partial, bool aligned,
1408
				unsigned int pipe_buffers)
1409 1410 1411 1412 1413
{
	int buffers = 0, error = 0;

	while (nr_vecs) {
		unsigned long off, npages;
1414
		struct iovec entry;
1415 1416 1417 1418
		void __user *base;
		size_t len;
		int i;

1419
		error = -EFAULT;
N
Nick Piggin 已提交
1420
		if (copy_from_user(&entry, iov, sizeof(entry)))
1421 1422
			break;

1423 1424 1425
		base = entry.iov_base;
		len = entry.iov_len;

1426 1427 1428
		/*
		 * Sanity check this iovec. 0 read succeeds.
		 */
1429
		error = 0;
1430 1431 1432
		if (unlikely(!len))
			break;
		error = -EFAULT;
1433
		if (!access_ok(VERIFY_READ, base, len))
1434 1435 1436 1437 1438 1439 1440
			break;

		/*
		 * Get this base offset and number of pages, then map
		 * in the user pages.
		 */
		off = (unsigned long) base & ~PAGE_MASK;
1441 1442 1443 1444 1445 1446 1447 1448 1449

		/*
		 * 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;

1450
		npages = (off + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
1451 1452
		if (npages > pipe_buffers - buffers)
			npages = pipe_buffers - buffers;
1453

N
Nick Piggin 已提交
1454 1455
		error = get_user_pages_fast((unsigned long)base, npages,
					0, &pages[buffers]);
1456 1457 1458 1459 1460 1461 1462 1463

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

		/*
		 * Fill this contiguous range into the partial page map.
		 */
		for (i = 0; i < error; i++) {
1464
			const int plen = min_t(size_t, len, PAGE_SIZE - off);
1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486

			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.
		 */
1487
		if (error < npages || buffers == pipe_buffers)
1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499
			break;

		nr_vecs--;
		iov++;
	}

	if (buffers)
		return buffers;

	return error;
}

1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529
static int pipe_to_user(struct pipe_inode_info *pipe, struct pipe_buffer *buf,
			struct splice_desc *sd)
{
	char *src;
	int ret;

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

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

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

1530
	buf->ops->unmap(pipe, buf, src);
1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549
out:
	if (ret > 0)
		sd->u.userptr += ret;
	return ret;
}

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

1550
	pipe = get_pipe_info(file);
1551 1552 1553
	if (!pipe)
		return -EBADF;

1554
	pipe_lock(pipe);
1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580

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

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

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

1581 1582 1583 1584 1585
		if (unlikely(!access_ok(VERIFY_WRITE, base, len))) {
			error = -EFAULT;
			break;
		}

1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608
		sd.len = 0;
		sd.total_len = len;
		sd.flags = flags;
		sd.u.userptr = base;
		sd.pos = 0;

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

			break;
		}

		ret += size;

		if (size < len)
			break;

		nr_segs--;
		iov++;
	}

1609
	pipe_unlock(pipe);
1610 1611 1612 1613 1614 1615 1616

	if (!ret)
		ret = error;

	return ret;
}

1617 1618 1619 1620 1621
/*
 * 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.
 */
1622 1623
static long vmsplice_to_pipe(struct file *file, const struct iovec __user *iov,
			     unsigned long nr_segs, unsigned int flags)
1624
{
1625
	struct pipe_inode_info *pipe;
1626 1627
	struct page *pages[PIPE_DEF_BUFFERS];
	struct partial_page partial[PIPE_DEF_BUFFERS];
1628 1629 1630
	struct splice_pipe_desc spd = {
		.pages = pages,
		.partial = partial,
1631
		.nr_pages_max = PIPE_DEF_BUFFERS,
1632 1633
		.flags = flags,
		.ops = &user_page_pipe_buf_ops,
1634
		.spd_release = spd_release_page,
1635
	};
1636
	long ret;
1637

1638
	pipe = get_pipe_info(file);
1639
	if (!pipe)
1640 1641
		return -EBADF;

1642 1643 1644 1645
	if (splice_grow_spd(pipe, &spd))
		return -ENOMEM;

	spd.nr_pages = get_iovec_page_array(iov, nr_segs, spd.pages,
1646
					    spd.partial, false,
1647
					    spd.nr_pages_max);
1648
	if (spd.nr_pages <= 0)
1649 1650 1651
		ret = spd.nr_pages;
	else
		ret = splice_to_pipe(pipe, &spd);
1652

1653
	splice_shrink_spd(&spd);
1654
	return ret;
1655 1656
}

1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672
/*
 * 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().
 *
 */
1673 1674
SYSCALL_DEFINE4(vmsplice, int, fd, const struct iovec __user *, iov,
		unsigned long, nr_segs, unsigned int, flags)
1675
{
1676
	struct fd f;
1677 1678
	long error;

1679 1680 1681 1682 1683
	if (unlikely(nr_segs > UIO_MAXIOV))
		return -EINVAL;
	else if (unlikely(!nr_segs))
		return 0;

1684
	error = -EBADF;
1685 1686 1687 1688 1689 1690 1691 1692
	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);
1693 1694 1695 1696 1697
	}

	return error;
}

1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718
#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

1719 1720 1721
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)
1722
{
1723
	struct fd in, out;
1724 1725 1726 1727 1728 1729
	long error;

	if (unlikely(!len))
		return 0;

	error = -EBADF;
1730 1731 1732 1733 1734 1735 1736 1737
	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,
1738
							  len, flags);
1739
				fdput(out);
1740 1741
			}
		}
1742
		fdput(in);
1743 1744 1745
	}
	return error;
}
1746

1747 1748 1749 1750
/*
 * Make sure there's data to read. Wait for input if we can, otherwise
 * return an appropriate error.
 */
1751
static int ipipe_prep(struct pipe_inode_info *pipe, unsigned int flags)
1752 1753 1754 1755 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.
	 */
	if (pipe->nrbufs)
		return 0;

	ret = 0;
1763
	pipe_lock(pipe);
1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780

	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);
	}

1781
	pipe_unlock(pipe);
1782 1783 1784 1785 1786 1787 1788
	return ret;
}

/*
 * Make sure there's writeable room. Wait for room if we can, otherwise
 * return an appropriate error.
 */
1789
static int opipe_prep(struct pipe_inode_info *pipe, unsigned int flags)
1790 1791 1792 1793 1794 1795 1796
{
	int ret;

	/*
	 * Check ->nrbufs without the inode lock first. This function
	 * is speculative anyways, so missing one is ok.
	 */
1797
	if (pipe->nrbufs < pipe->buffers)
1798 1799 1800
		return 0;

	ret = 0;
1801
	pipe_lock(pipe);
1802

1803
	while (pipe->nrbufs >= pipe->buffers) {
1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821
		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--;
	}

1822
	pipe_unlock(pipe);
1823 1824 1825
	return ret;
}

1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868
/*
 * 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.
		 */
1869
		if (!ipipe->nrbufs || opipe->nrbufs >= opipe->buffers) {
1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889
			/* 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;
1890
		nbuf = (opipe->curbuf + opipe->nrbufs) & (opipe->buffers - 1);
1891 1892 1893 1894 1895 1896 1897 1898 1899
		obuf = opipe->bufs + nbuf;

		if (len >= ibuf->len) {
			/*
			 * Simply move the whole buffer from ipipe to opipe
			 */
			*obuf = *ibuf;
			ibuf->ops = NULL;
			opipe->nrbufs++;
1900
			ipipe->curbuf = (ipipe->curbuf + 1) & (ipipe->buffers - 1);
1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931
			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
1932 1933 1934
	if (ret > 0)
		wakeup_pipe_readers(opipe);

1935 1936 1937 1938 1939 1940
	if (input_wakeup)
		wakeup_pipe_writers(ipipe);

	return ret;
}

1941 1942 1943 1944 1945 1946 1947 1948
/*
 * 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;
1949
	int ret = 0, i = 0, nbuf;
1950 1951 1952

	/*
	 * Potential ABBA deadlock, work around it by ordering lock
1953
	 * grabbing by pipe info address. Otherwise two different processes
1954 1955
	 * could deadlock (one doing tee from A -> B, the other from B -> A).
	 */
1956
	pipe_double_lock(ipipe, opipe);
1957

1958
	do {
1959 1960 1961 1962 1963 1964 1965
		if (!opipe->readers) {
			send_sig(SIGPIPE, current, 0);
			if (!ret)
				ret = -EPIPE;
			break;
		}

1966 1967 1968 1969
		/*
		 * If we have iterated all input buffers or ran out of
		 * output room, break.
		 */
1970
		if (i >= ipipe->nrbufs || opipe->nrbufs >= opipe->buffers)
1971
			break;
1972

1973 1974
		ibuf = ipipe->bufs + ((ipipe->curbuf + i) & (ipipe->buffers-1));
		nbuf = (opipe->curbuf + opipe->nrbufs) & (opipe->buffers - 1);
1975 1976

		/*
1977 1978
		 * Get a reference to this pipe buffer,
		 * so we can copy the contents over.
1979
		 */
1980 1981 1982 1983 1984
		ibuf->ops->get(ipipe, ibuf);

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

1985
		/*
1986 1987
		 * Don't inherit the gift flag, we need to
		 * prevent multiple steals of this page.
1988
		 */
1989
		obuf->flags &= ~PIPE_BUF_FLAG_GIFT;
1990

1991 1992
		if (obuf->len > len)
			obuf->len = len;
1993

1994 1995 1996 1997 1998
		opipe->nrbufs++;
		ret += obuf->len;
		len -= obuf->len;
		i++;
	} while (len);
1999

2000 2001 2002 2003 2004 2005 2006
	/*
	 * 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;

2007 2008
	pipe_unlock(ipipe);
	pipe_unlock(opipe);
2009

2010 2011 2012
	/*
	 * If we put data in the output pipe, wakeup any potential readers.
	 */
N
Namhyung Kim 已提交
2013 2014
	if (ret > 0)
		wakeup_pipe_readers(opipe);
2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027

	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)
{
2028 2029
	struct pipe_inode_info *ipipe = get_pipe_info(in);
	struct pipe_inode_info *opipe = get_pipe_info(out);
2030
	int ret = -EINVAL;
2031 2032

	/*
2033 2034
	 * Duplicate the contents of ipipe to opipe without actually
	 * copying the data.
2035
	 */
2036 2037 2038 2039 2040
	if (ipipe && opipe && ipipe != opipe) {
		/*
		 * Keep going, unless we encounter an error. The ipipe/opipe
		 * ordering doesn't really matter.
		 */
2041
		ret = ipipe_prep(ipipe, flags);
2042
		if (!ret) {
2043
			ret = opipe_prep(opipe, flags);
2044
			if (!ret)
2045 2046 2047
				ret = link_pipe(ipipe, opipe, len, flags);
		}
	}
2048

2049
	return ret;
2050 2051
}

2052
SYSCALL_DEFINE4(tee, int, fdin, int, fdout, size_t, len, unsigned int, flags)
2053
{
2054 2055
	struct fd in;
	int error;
2056 2057 2058 2059 2060

	if (unlikely(!len))
		return 0;

	error = -EBADF;
2061 2062 2063 2064 2065 2066 2067 2068 2069
	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);
2070 2071
			}
		}
2072
 		fdput(in);
2073 2074 2075 2076
 	}

	return error;
}