splice.c 46.4 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

35 36 37 38 39 40
/*
 * 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.
 */
41
static int page_cache_pipe_buf_steal(struct pipe_inode_info *pipe,
42 43 44
				     struct pipe_buffer *buf)
{
	struct page *page = buf->page;
45
	struct address_space *mapping;
46

47 48
	lock_page(page);

49 50 51
	mapping = page_mapping(page);
	if (mapping) {
		WARN_ON(!PageUptodate(page));
52

53 54 55 56 57 58 59 60 61
		/*
		 * 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);
62

63 64
		if (page_has_private(page) &&
		    !try_to_release_page(page, GFP_KERNEL))
65
			goto out_unlock;
66

67 68 69 70 71 72 73 74
		/*
		 * 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;
		}
75
	}
76

77 78 79 80
	/*
	 * Raced with truncate or failed to remove page from current
	 * address space, unlock and return failure.
	 */
81
out_unlock:
82 83
	unlock_page(page);
	return 1;
84 85
}

86
static void page_cache_pipe_buf_release(struct pipe_inode_info *pipe,
87 88 89
					struct pipe_buffer *buf)
{
	page_cache_release(buf->page);
J
Jens Axboe 已提交
90
	buf->flags &= ~PIPE_BUF_FLAG_LRU;
91 92
}

93 94 95 96
/*
 * Check whether the contents of buf is OK to access. Since the content
 * is a page cache page, IO may be in flight.
 */
97 98
static int page_cache_pipe_buf_confirm(struct pipe_inode_info *pipe,
				       struct pipe_buffer *buf)
99 100
{
	struct page *page = buf->page;
101
	int err;
102 103

	if (!PageUptodate(page)) {
104 105 106 107
		lock_page(page);

		/*
		 * Page got truncated/unhashed. This will cause a 0-byte
I
Ingo Molnar 已提交
108
		 * splice, if this is the first page.
109 110 111 112 113
		 */
		if (!page->mapping) {
			err = -ENODATA;
			goto error;
		}
114

115
		/*
I
Ingo Molnar 已提交
116
		 * Uh oh, read-error from disk.
117 118 119 120 121 122 123
		 */
		if (!PageUptodate(page)) {
			err = -EIO;
			goto error;
		}

		/*
124
		 * Page is ok afterall, we are done.
125
		 */
126 127 128
		unlock_page(page);
	}

129
	return 0;
130 131
error:
	unlock_page(page);
132
	return err;
133 134
}

135
const struct pipe_buf_operations page_cache_pipe_buf_ops = {
136
	.can_merge = 0,
137 138
	.map = generic_pipe_buf_map,
	.unmap = generic_pipe_buf_unmap,
139
	.confirm = page_cache_pipe_buf_confirm,
140
	.release = page_cache_pipe_buf_release,
141
	.steal = page_cache_pipe_buf_steal,
142
	.get = generic_pipe_buf_get,
143 144
};

145 146 147
static int user_page_pipe_buf_steal(struct pipe_inode_info *pipe,
				    struct pipe_buffer *buf)
{
148 149 150
	if (!(buf->flags & PIPE_BUF_FLAG_GIFT))
		return 1;

J
Jens Axboe 已提交
151
	buf->flags |= PIPE_BUF_FLAG_LRU;
152
	return generic_pipe_buf_steal(pipe, buf);
153 154
}

155
static const struct pipe_buf_operations user_page_pipe_buf_ops = {
156
	.can_merge = 0,
157 158
	.map = generic_pipe_buf_map,
	.unmap = generic_pipe_buf_unmap,
159
	.confirm = generic_pipe_buf_confirm,
160 161
	.release = page_cache_pipe_buf_release,
	.steal = user_page_pipe_buf_steal,
162
	.get = generic_pipe_buf_get,
163 164
};

N
Namhyung Kim 已提交
165 166 167 168 169 170 171 172
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);
}

173 174 175 176 177 178
/**
 * splice_to_pipe - fill passed data into a pipe
 * @pipe:	pipe to fill
 * @spd:	data to fill
 *
 * Description:
R
Randy Dunlap 已提交
179
 *    @spd contains a map of pages and len/offset tuples, along with
180 181 182
 *    the struct pipe_buf_operations associated with these pages. This
 *    function will link that data to the pipe.
 *
183
 */
184 185
ssize_t splice_to_pipe(struct pipe_inode_info *pipe,
		       struct splice_pipe_desc *spd)
186
{
187
	unsigned int spd_pages = spd->nr_pages;
188
	int ret, do_wakeup, page_nr;
189 190 191

	ret = 0;
	do_wakeup = 0;
192
	page_nr = 0;
193

194
	pipe_lock(pipe);
195 196

	for (;;) {
197
		if (!pipe->readers) {
198 199 200 201 202 203
			send_sig(SIGPIPE, current, 0);
			if (!ret)
				ret = -EPIPE;
			break;
		}

204 205
		if (pipe->nrbufs < pipe->buffers) {
			int newbuf = (pipe->curbuf + pipe->nrbufs) & (pipe->buffers - 1);
206
			struct pipe_buffer *buf = pipe->bufs + newbuf;
207

208 209 210
			buf->page = spd->pages[page_nr];
			buf->offset = spd->partial[page_nr].offset;
			buf->len = spd->partial[page_nr].len;
211
			buf->private = spd->partial[page_nr].private;
212
			buf->ops = spd->ops;
213 214 215
			if (spd->flags & SPLICE_F_GIFT)
				buf->flags |= PIPE_BUF_FLAG_GIFT;

216
			pipe->nrbufs++;
217 218 219
			page_nr++;
			ret += buf->len;

220 221
			if (pipe->inode)
				do_wakeup = 1;
222

223
			if (!--spd->nr_pages)
224
				break;
225
			if (pipe->nrbufs < pipe->buffers)
226 227 228 229 230
				continue;

			break;
		}

231
		if (spd->flags & SPLICE_F_NONBLOCK) {
232 233 234 235 236
			if (!ret)
				ret = -EAGAIN;
			break;
		}

237 238 239 240 241 242 243
		if (signal_pending(current)) {
			if (!ret)
				ret = -ERESTARTSYS;
			break;
		}

		if (do_wakeup) {
244
			smp_mb();
245 246 247
			if (waitqueue_active(&pipe->wait))
				wake_up_interruptible_sync(&pipe->wait);
			kill_fasync(&pipe->fasync_readers, SIGIO, POLL_IN);
248 249 250
			do_wakeup = 0;
		}

251 252 253
		pipe->waiting_writers++;
		pipe_wait(pipe);
		pipe->waiting_writers--;
254 255
	}

256
	pipe_unlock(pipe);
257

N
Namhyung Kim 已提交
258 259
	if (do_wakeup)
		wakeup_pipe_readers(pipe);
260

261
	while (page_nr < spd_pages)
262
		spd->spd_release(spd, page_nr++);
263 264 265 266

	return ret;
}

267
void spd_release_page(struct splice_pipe_desc *spd, unsigned int i)
268 269 270 271
{
	page_cache_release(spd->pages[i]);
}

272 273 274 275
/*
 * Check if we need to grow the arrays holding pages and partial page
 * descriptions.
 */
276
int splice_grow_spd(const struct pipe_inode_info *pipe, struct splice_pipe_desc *spd)
277
{
278 279 280 281
	unsigned int buffers = ACCESS_ONCE(pipe->buffers);

	spd->nr_pages_max = buffers;
	if (buffers <= PIPE_DEF_BUFFERS)
282 283
		return 0;

284 285
	spd->pages = kmalloc(buffers * sizeof(struct page *), GFP_KERNEL);
	spd->partial = kmalloc(buffers * sizeof(struct partial_page), GFP_KERNEL);
286 287 288 289 290 291 292 293 294

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

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

295
void splice_shrink_spd(struct splice_pipe_desc *spd)
296
{
297
	if (spd->nr_pages_max <= PIPE_DEF_BUFFERS)
298 299 300 301 302 303
		return;

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

304
static int
305 306 307
__generic_file_splice_read(struct file *in, loff_t *ppos,
			   struct pipe_inode_info *pipe, size_t len,
			   unsigned int flags)
308 309
{
	struct address_space *mapping = in->f_mapping;
F
Fengguang Wu 已提交
310
	unsigned int loff, nr_pages, req_pages;
311 312
	struct page *pages[PIPE_DEF_BUFFERS];
	struct partial_page partial[PIPE_DEF_BUFFERS];
313
	struct page *page;
314 315
	pgoff_t index, end_index;
	loff_t isize;
316
	int error, page_nr;
317 318 319
	struct splice_pipe_desc spd = {
		.pages = pages,
		.partial = partial,
320
		.nr_pages_max = PIPE_DEF_BUFFERS,
321 322
		.flags = flags,
		.ops = &page_cache_pipe_buf_ops,
323
		.spd_release = spd_release_page,
324
	};
325

326 327 328
	if (splice_grow_spd(pipe, &spd))
		return -ENOMEM;

329
	index = *ppos >> PAGE_CACHE_SHIFT;
330
	loff = *ppos & ~PAGE_CACHE_MASK;
F
Fengguang Wu 已提交
331
	req_pages = (len + loff + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
332
	nr_pages = min(req_pages, spd.nr_pages_max);
333

334 335 336
	/*
	 * Lookup the (hopefully) full range of pages we need.
	 */
337
	spd.nr_pages = find_get_pages_contig(mapping, index, nr_pages, spd.pages);
338
	index += spd.nr_pages;
339

340 341
	/*
	 * If find_get_pages_contig() returned fewer pages than we needed,
342
	 * readahead/allocate the rest and fill in the holes.
343
	 */
344
	if (spd.nr_pages < nr_pages)
345 346
		page_cache_sync_readahead(mapping, &in->f_ra, in,
				index, req_pages - spd.nr_pages);
347

348
	error = 0;
349
	while (spd.nr_pages < nr_pages) {
350
		/*
351 352
		 * Page could be there, find_get_pages_contig() breaks on
		 * the first hole.
353
		 */
354 355 356
		page = find_get_page(mapping, index);
		if (!page) {
			/*
357
			 * page didn't exist, allocate one.
358 359 360 361 362 363
			 */
			page = page_cache_alloc_cold(mapping);
			if (!page)
				break;

			error = add_to_page_cache_lru(page, mapping, index,
364
						GFP_KERNEL);
365 366
			if (unlikely(error)) {
				page_cache_release(page);
367 368
				if (error == -EEXIST)
					continue;
369 370
				break;
			}
371 372 373 374 375
			/*
			 * add_to_page_cache() locks the page, unlock it
			 * to avoid convoluting the logic below even more.
			 */
			unlock_page(page);
376 377
		}

378
		spd.pages[spd.nr_pages++] = page;
379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398
		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);
399
		page = spd.pages[page_nr];
400

401
		if (PageReadahead(page))
402
			page_cache_async_readahead(mapping, &in->f_ra, in,
F
Fengguang Wu 已提交
403
					page, index, req_pages - page_nr);
404

405 406 407 408
		/*
		 * If the page isn't uptodate, we may need to start io on it
		 */
		if (!PageUptodate(page)) {
409
			lock_page(page);
410 411

			/*
412 413 414 415
			 * 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.
416 417 418
			 */
			if (!page->mapping) {
				unlock_page(page);
419 420 421 422 423 424 425
				page = find_or_create_page(mapping, index,
						mapping_gfp_mask(mapping));

				if (!page) {
					error = -ENOMEM;
					break;
				}
426 427
				page_cache_release(spd.pages[page_nr]);
				spd.pages[page_nr] = page;
428 429 430 431 432 433 434 435
			}
			/*
			 * page was already under io and is now done, great
			 */
			if (PageUptodate(page)) {
				unlock_page(page);
				goto fill_it;
			}
436

437 438 439 440
			/*
			 * need to read in the page
			 */
			error = mapping->a_ops->readpage(in, page);
441
			if (unlikely(error)) {
442 443 444 445 446 447
				/*
				 * 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.
				 */
448
				if (error == AOP_TRUNCATED_PAGE)
449 450
					error = 0;

451 452
				break;
			}
453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468
		}
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;
469 470

			/*
471
			 * max good bytes in this page
472
			 */
473 474
			plen = ((isize - 1) & ~PAGE_CACHE_MASK) + 1;
			if (plen <= loff)
475 476 477
				break;

			/*
478
			 * force quit after adding this page
479
			 */
480 481
			this_len = min(this_len, plen - loff);
			len = this_len;
482
		}
483

484 485
		spd.partial[page_nr].offset = loff;
		spd.partial[page_nr].len = this_len;
486
		len -= this_len;
487
		loff = 0;
488 489
		spd.nr_pages++;
		index++;
490 491
	}

492
	/*
493
	 * Release any pages at the end, if we quit early. 'page_nr' is how far
494 495 496
	 * we got, 'nr_pages' is how many pages are in the map.
	 */
	while (page_nr < nr_pages)
497
		page_cache_release(spd.pages[page_nr++]);
498
	in->f_ra.prev_pos = (loff_t)index << PAGE_CACHE_SHIFT;
499

500
	if (spd.nr_pages)
501
		error = splice_to_pipe(pipe, &spd);
502

503
	splice_shrink_spd(&spd);
504
	return error;
505 506
}

507 508 509
/**
 * generic_file_splice_read - splice data from file to a pipe
 * @in:		file to splice from
510
 * @ppos:	position in @in
511 512 513 514
 * @pipe:	pipe to splice to
 * @len:	number of bytes to splice
 * @flags:	splice modifier flags
 *
515 516 517 518 519
 * 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.
 *
520
 */
521 522 523
ssize_t generic_file_splice_read(struct file *in, loff_t *ppos,
				 struct pipe_inode_info *pipe, size_t len,
				 unsigned int flags)
524
{
525
	loff_t isize, left;
526
	int ret;
527 528 529 530 531 532 533 534

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

	left = isize - *ppos;
	if (unlikely(left < len))
		len = left;
535

536
	ret = __generic_file_splice_read(in, ppos, pipe, len, flags);
537
	if (ret > 0) {
538
		*ppos += ret;
539 540
		file_accessed(in);
	}
541 542 543

	return ret;
}
544 545
EXPORT_SYMBOL(generic_file_splice_read);

546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571
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;
}

572 573
static ssize_t kernel_write(struct file *file, const char *buf, size_t count,
			    loff_t pos)
574 575 576 577 578 579 580
{
	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() */
581
	res = vfs_write(file, (const char __user *)buf, count, &pos);
582 583 584 585 586
	set_fs(old_fs);

	return res;
}

587 588 589 590 591 592 593
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;
594 595 596
	struct page *pages[PIPE_DEF_BUFFERS];
	struct partial_page partial[PIPE_DEF_BUFFERS];
	struct iovec *vec, __vec[PIPE_DEF_BUFFERS];
597 598 599 600 601 602 603
	ssize_t res;
	size_t this_len;
	int error;
	int i;
	struct splice_pipe_desc spd = {
		.pages = pages,
		.partial = partial,
604
		.nr_pages_max = PIPE_DEF_BUFFERS,
605 606 607 608 609
		.flags = flags,
		.ops = &default_pipe_buf_ops,
		.spd_release = spd_release_page,
	};

610 611 612 613 614
	if (splice_grow_spd(pipe, &spd))
		return -ENOMEM;

	res = -ENOMEM;
	vec = __vec;
615 616
	if (spd.nr_pages_max > PIPE_DEF_BUFFERS) {
		vec = kmalloc(spd.nr_pages_max * sizeof(struct iovec), GFP_KERNEL);
617 618 619 620
		if (!vec)
			goto shrink_ret;
	}

621 622 623
	offset = *ppos & ~PAGE_CACHE_MASK;
	nr_pages = (len + offset + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;

624
	for (i = 0; i < nr_pages && i < spd.nr_pages_max && len; i++) {
625 626
		struct page *page;

627
		page = alloc_page(GFP_USER);
628 629 630 631 632
		error = -ENOMEM;
		if (!page)
			goto err;

		this_len = min_t(size_t, len, PAGE_CACHE_SIZE - offset);
633
		vec[i].iov_base = (void __user *) page_address(page);
634
		vec[i].iov_len = this_len;
635
		spd.pages[i] = page;
636 637 638 639 640 641
		spd.nr_pages++;
		len -= this_len;
		offset = 0;
	}

	res = kernel_readv(in, vec, spd.nr_pages, *ppos);
A
Andrew Morton 已提交
642 643
	if (res < 0) {
		error = res;
644
		goto err;
A
Andrew Morton 已提交
645
	}
646 647 648 649 650 651 652 653

	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);
654 655
		spd.partial[i].offset = 0;
		spd.partial[i].len = this_len;
656
		if (!this_len) {
657 658
			__free_page(spd.pages[i]);
			spd.pages[i] = NULL;
659 660 661 662 663 664 665 666 667 668
			nr_freed++;
		}
		res -= this_len;
	}
	spd.nr_pages -= nr_freed;

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

669 670 671
shrink_ret:
	if (vec != __vec)
		kfree(vec);
672
	splice_shrink_spd(&spd);
673 674 675
	return res;

err:
676
	for (i = 0; i < spd.nr_pages; i++)
677
		__free_page(spd.pages[i]);
678

679 680
	res = error;
	goto shrink_ret;
681 682 683
}
EXPORT_SYMBOL(default_file_splice_read);

684
/*
685
 * Send 'sd->len' bytes to socket from 'sd->file' at position 'sd->pos'
J
Jens Axboe 已提交
686
 * using sendpage(). Return the number of bytes sent.
687
 */
688
static int pipe_to_sendpage(struct pipe_inode_info *pipe,
689 690
			    struct pipe_buffer *buf, struct splice_desc *sd)
{
691
	struct file *file = sd->u.file;
692
	loff_t pos = sd->pos;
693
	int more;
694

695 696 697
	if (!likely(file->f_op && file->f_op->sendpage))
		return -EINVAL;

698 699 700
	more = (sd->flags & SPLICE_F_MORE) ? MSG_MORE : 0;
	if (sd->len < sd->total_len)
		more |= MSG_SENDPAGE_NOTLAST;
701 702
	return file->f_op->sendpage(file, buf->page, buf->offset,
				    sd->len, &pos, more);
703 704 705 706 707 708 709 710 711 712 713 714 715 716 717
}

/*
 * 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.
 *
718 719 720 721 722 723
 * 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.
724
 */
725 726
int pipe_to_file(struct pipe_inode_info *pipe, struct pipe_buffer *buf,
		 struct splice_desc *sd)
727
{
728
	struct file *file = sd->u.file;
729
	struct address_space *mapping = file->f_mapping;
J
Jens Axboe 已提交
730
	unsigned int offset, this_len;
731
	struct page *page;
732
	void *fsdata;
733
	int ret;
734 735 736

	offset = sd->pos & ~PAGE_CACHE_MASK;

J
Jens Axboe 已提交
737 738 739 740
	this_len = sd->len;
	if (this_len + offset > PAGE_CACHE_SIZE)
		this_len = PAGE_CACHE_SIZE - offset;

741 742 743 744
	ret = pagecache_write_begin(file, mapping, sd->pos, this_len,
				AOP_FLAG_UNINTERRUPTIBLE, &page, &fsdata);
	if (unlikely(ret))
		goto out;
745

746
	if (buf->page != page) {
747
		char *src = buf->ops->map(pipe, buf, 1);
748
		char *dst = kmap_atomic(page);
749

J
Jens Axboe 已提交
750
		memcpy(dst + offset, src + buf->offset, this_len);
751
		flush_dcache_page(page);
752
		kunmap_atomic(dst);
753
		buf->ops->unmap(pipe, buf, src);
754
	}
755 756
	ret = pagecache_write_end(file, mapping, sd->pos, this_len, this_len,
				page, fsdata);
757 758 759
out:
	return ret;
}
760
EXPORT_SYMBOL(pipe_to_file);
761

762 763 764 765 766 767 768 769
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);
}

770
/**
771
 * splice_from_pipe_feed - feed available data from a pipe to a file
772 773 774 775 776
 * @pipe:	pipe to splice from
 * @sd:		information to @actor
 * @actor:	handler that splices the data
 *
 * Description:
777 778 779 780 781 782 783
 *    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.
784
 *
785 786 787 788
 *    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.
789
 */
790 791
int splice_from_pipe_feed(struct pipe_inode_info *pipe, struct splice_desc *sd,
			  splice_actor *actor)
792
{
793
	int ret;
794

795 796 797
	while (pipe->nrbufs) {
		struct pipe_buffer *buf = pipe->bufs + pipe->curbuf;
		const struct pipe_buf_operations *ops = buf->ops;
798

799 800 801
		sd->len = buf->len;
		if (sd->len > sd->total_len)
			sd->len = sd->total_len;
802

803 804
		ret = buf->ops->confirm(pipe, buf);
		if (unlikely(ret)) {
805 806 807 808
			if (ret == -ENODATA)
				ret = 0;
			return ret;
		}
809 810 811 812 813

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

814 815 816 817 818 819 820 821 822 823 824
		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);
825
			pipe->curbuf = (pipe->curbuf + 1) & (pipe->buffers - 1);
826 827 828 829
			pipe->nrbufs--;
			if (pipe->inode)
				sd->need_wakeup = true;
		}
830

831 832 833
		if (!sd->total_len)
			return 0;
	}
834

835 836 837
	return 1;
}
EXPORT_SYMBOL(splice_from_pipe_feed);
838

839 840 841 842 843 844 845 846 847 848 849 850 851 852 853
/**
 * 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
854

855 856
		if (!pipe->waiting_writers && sd->num_spliced)
			return 0;
I
Ingo Molnar 已提交
857

858 859
		if (sd->flags & SPLICE_F_NONBLOCK)
			return -EAGAIN;
860

861 862
		if (signal_pending(current))
			return -ERESTARTSYS;
863

864 865 866
		if (sd->need_wakeup) {
			wakeup_pipe_writers(pipe);
			sd->need_wakeup = false;
867 868
		}

869 870
		pipe_wait(pipe);
	}
871

872 873 874
	return 1;
}
EXPORT_SYMBOL(splice_from_pipe_next);
875

876 877
/**
 * splice_from_pipe_begin - start splicing from pipe
878
 * @sd:		information about the splice operation
879 880 881 882 883 884 885 886 887 888 889 890
 *
 * 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);
891

892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907
/**
 * 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);
908

909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925
/**
 * __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;
926

927 928 929 930 931 932 933 934 935
	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;
936
}
M
Mark Fasheh 已提交
937
EXPORT_SYMBOL(__splice_from_pipe);
938

939 940 941 942 943 944 945 946 947 948
/**
 * 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:
949
 *    See __splice_from_pipe. This function locks the pipe inode,
950 951 952
 *    otherwise it's identical to __splice_from_pipe().
 *
 */
953 954 955 956 957
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
958 959 960 961
	struct splice_desc sd = {
		.total_len = len,
		.flags = flags,
		.pos = *ppos,
962
		.u.file = out,
J
Jens Axboe 已提交
963
	};
964

965
	pipe_lock(pipe);
J
Jens Axboe 已提交
966
	ret = __splice_from_pipe(pipe, &sd, actor);
967
	pipe_unlock(pipe);
968 969 970 971

	return ret;
}

972 973
/**
 * generic_file_splice_write - splice data from a pipe to a file
974
 * @pipe:	pipe info
975
 * @out:	file to write to
976
 * @ppos:	position in @out
977 978 979
 * @len:	number of bytes to splice
 * @flags:	splice modifier flags
 *
980 981 982
 * Description:
 *    Will either move or copy pages (determined by @flags options) from
 *    the given pipe inode to the given file.
983 984
 *
 */
985 986
ssize_t
generic_file_splice_write(struct pipe_inode_info *pipe, struct file *out,
987
			  loff_t *ppos, size_t len, unsigned int flags)
988
{
989
	struct address_space *mapping = out->f_mapping;
990
	struct inode *inode = mapping->host;
991 992 993 994 995 996
	struct splice_desc sd = {
		.total_len = len,
		.flags = flags,
		.pos = *ppos,
		.u.file = out,
	};
997 998
	ssize_t ret;

999 1000
	sb_start_write(inode->i_sb);

1001
	pipe_lock(pipe);
1002 1003 1004 1005 1006 1007 1008 1009 1010

	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);
1011
		if (!ret) {
1012 1013 1014 1015
			ret = file_update_time(out);
			if (!ret)
				ret = splice_from_pipe_feed(pipe, &sd,
							    pipe_to_file);
1016
		}
1017 1018 1019 1020
		mutex_unlock(&inode->i_mutex);
	} while (ret > 0);
	splice_from_pipe_end(pipe, &sd);

1021
	pipe_unlock(pipe);
1022 1023 1024 1025

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

J
Jens Axboe 已提交
1026
	if (ret > 0) {
1027
		unsigned long nr_pages;
1028
		int err;
1029 1030

		nr_pages = (ret + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
J
Jens Axboe 已提交
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_nr(mapping, nr_pages);
1038
	}
1039
	sb_end_write(inode->i_sb);
1040 1041

	return ret;
1042 1043
}

1044 1045
EXPORT_SYMBOL(generic_file_splice_write);

1046 1047
static int write_pipe_buf(struct pipe_inode_info *pipe, struct pipe_buffer *buf,
			  struct splice_desc *sd)
1048
{
1049 1050 1051 1052 1053 1054 1055 1056
	int ret;
	void *data;

	data = buf->ops->map(pipe, buf, 0);
	ret = kernel_write(sd->u.file, data + buf->offset, sd->len, sd->pos);
	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 1118
		splice_write = default_file_splice_write;

	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!
	 */
1174
	i_mode = in->f_path.dentry->d_inode->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)) {
1184 1185 1186 1187 1188 1189
		pipe = alloc_pipe_info(NULL);
		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 1275
	return do_splice_from(pipe, file, &file->f_pos, sd->total_len,
			      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 1294 1295 1296 1297 1298 1299 1300
long do_splice_direct(struct file *in, loff_t *ppos, struct file *out,
		      size_t len, unsigned int flags)
{
	struct splice_desc sd = {
		.len		= len,
		.total_len	= len,
		.flags		= flags,
		.pos		= *ppos,
1301
		.u.file		= out,
J
Jens Axboe 已提交
1302
	};
1303
	long ret;
J
Jens Axboe 已提交
1304 1305

	ret = splice_direct_to_actor(in, &sd, direct_splice_actor);
1306
	if (ret > 0)
1307
		*ppos = sd.pos;
1308

J
Jens Axboe 已提交
1309
	return ret;
1310 1311
}

1312 1313 1314
static int splice_pipe_to_pipe(struct pipe_inode_info *ipipe,
			       struct pipe_inode_info *opipe,
			       size_t len, unsigned int flags);
1315

1316 1317 1318
/*
 * Determine where to splice to/from.
 */
1319 1320 1321
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)
1322
{
1323 1324
	struct pipe_inode_info *ipipe;
	struct pipe_inode_info *opipe;
1325
	loff_t offset, *off;
J
Jens Axboe 已提交
1326
	long ret;
1327

1328 1329
	ipipe = get_pipe_info(in);
	opipe = get_pipe_info(out);
1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348

	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) {
1349 1350
		if (off_in)
			return -ESPIPE;
1351
		if (off_out) {
1352
			if (!(out->f_mode & FMODE_PWRITE))
1353
				return -EINVAL;
1354
			if (copy_from_user(&offset, off_out, sizeof(loff_t)))
1355
				return -EFAULT;
1356 1357 1358
			off = &offset;
		} else
			off = &out->f_pos;
1359

1360
		ret = do_splice_from(ipipe, out, off, len, flags);
J
Jens Axboe 已提交
1361 1362 1363 1364 1365

		if (off_out && copy_to_user(off_out, off, sizeof(loff_t)))
			ret = -EFAULT;

		return ret;
1366
	}
1367

1368
	if (opipe) {
1369 1370
		if (off_out)
			return -ESPIPE;
1371
		if (off_in) {
1372
			if (!(in->f_mode & FMODE_PREAD))
1373
				return -EINVAL;
1374
			if (copy_from_user(&offset, off_in, sizeof(loff_t)))
1375
				return -EFAULT;
1376 1377 1378
			off = &offset;
		} else
			off = &in->f_pos;
1379

1380
		ret = do_splice_to(in, off, opipe, len, flags);
J
Jens Axboe 已提交
1381 1382 1383 1384 1385

		if (off_in && copy_to_user(off_in, off, sizeof(loff_t)))
			ret = -EFAULT;

		return ret;
1386
	}
1387 1388 1389 1390

	return -EINVAL;
}

1391 1392 1393 1394 1395 1396 1397 1398 1399
/*
 * 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,
1400
				struct partial_page *partial, bool aligned,
1401
				unsigned int pipe_buffers)
1402 1403 1404 1405 1406
{
	int buffers = 0, error = 0;

	while (nr_vecs) {
		unsigned long off, npages;
1407
		struct iovec entry;
1408 1409 1410 1411
		void __user *base;
		size_t len;
		int i;

1412
		error = -EFAULT;
N
Nick Piggin 已提交
1413
		if (copy_from_user(&entry, iov, sizeof(entry)))
1414 1415
			break;

1416 1417 1418
		base = entry.iov_base;
		len = entry.iov_len;

1419 1420 1421
		/*
		 * Sanity check this iovec. 0 read succeeds.
		 */
1422
		error = 0;
1423 1424 1425
		if (unlikely(!len))
			break;
		error = -EFAULT;
1426
		if (!access_ok(VERIFY_READ, base, len))
1427 1428 1429 1430 1431 1432 1433
			break;

		/*
		 * Get this base offset and number of pages, then map
		 * in the user pages.
		 */
		off = (unsigned long) base & ~PAGE_MASK;
1434 1435 1436 1437 1438 1439 1440 1441 1442

		/*
		 * 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;

1443
		npages = (off + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
1444 1445
		if (npages > pipe_buffers - buffers)
			npages = pipe_buffers - buffers;
1446

N
Nick Piggin 已提交
1447 1448
		error = get_user_pages_fast((unsigned long)base, npages,
					0, &pages[buffers]);
1449 1450 1451 1452 1453 1454 1455 1456

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

		/*
		 * Fill this contiguous range into the partial page map.
		 */
		for (i = 0; i < error; i++) {
1457
			const int plen = min_t(size_t, len, PAGE_SIZE - off);
1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479

			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.
		 */
1480
		if (error < npages || buffers == pipe_buffers)
1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492
			break;

		nr_vecs--;
		iov++;
	}

	if (buffers)
		return buffers;

	return error;
}

1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522
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;

1523
	buf->ops->unmap(pipe, buf, src);
1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542
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;

1543
	pipe = get_pipe_info(file);
1544 1545 1546
	if (!pipe)
		return -EBADF;

1547
	pipe_lock(pipe);
1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573

	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;
		}

1574 1575 1576 1577 1578
		if (unlikely(!access_ok(VERIFY_WRITE, base, len))) {
			error = -EFAULT;
			break;
		}

1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601
		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++;
	}

1602
	pipe_unlock(pipe);
1603 1604 1605 1606 1607 1608 1609

	if (!ret)
		ret = error;

	return ret;
}

1610 1611 1612 1613 1614
/*
 * 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.
 */
1615 1616
static long vmsplice_to_pipe(struct file *file, const struct iovec __user *iov,
			     unsigned long nr_segs, unsigned int flags)
1617
{
1618
	struct pipe_inode_info *pipe;
1619 1620
	struct page *pages[PIPE_DEF_BUFFERS];
	struct partial_page partial[PIPE_DEF_BUFFERS];
1621 1622 1623
	struct splice_pipe_desc spd = {
		.pages = pages,
		.partial = partial,
1624
		.nr_pages_max = PIPE_DEF_BUFFERS,
1625 1626
		.flags = flags,
		.ops = &user_page_pipe_buf_ops,
1627
		.spd_release = spd_release_page,
1628
	};
1629
	long ret;
1630

1631
	pipe = get_pipe_info(file);
1632
	if (!pipe)
1633 1634
		return -EBADF;

1635 1636 1637 1638
	if (splice_grow_spd(pipe, &spd))
		return -ENOMEM;

	spd.nr_pages = get_iovec_page_array(iov, nr_segs, spd.pages,
1639
					    spd.partial, false,
1640
					    spd.nr_pages_max);
1641
	if (spd.nr_pages <= 0)
1642 1643 1644
		ret = spd.nr_pages;
	else
		ret = splice_to_pipe(pipe, &spd);
1645

1646
	splice_shrink_spd(&spd);
1647
	return ret;
1648 1649
}

1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665
/*
 * 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().
 *
 */
1666 1667
SYSCALL_DEFINE4(vmsplice, int, fd, const struct iovec __user *, iov,
		unsigned long, nr_segs, unsigned int, flags)
1668
{
1669
	struct fd f;
1670 1671
	long error;

1672 1673 1674 1675 1676
	if (unlikely(nr_segs > UIO_MAXIOV))
		return -EINVAL;
	else if (unlikely(!nr_segs))
		return 0;

1677
	error = -EBADF;
1678 1679 1680 1681 1682 1683 1684 1685
	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);
1686 1687 1688 1689 1690
	}

	return error;
}

1691 1692 1693
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)
1694
{
1695
	struct fd in, out;
1696 1697 1698 1699 1700 1701
	long error;

	if (unlikely(!len))
		return 0;

	error = -EBADF;
1702 1703 1704 1705 1706 1707 1708 1709
	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,
1710
							  len, flags);
1711
				fdput(out);
1712 1713
			}
		}
1714
		fdput(in);
1715 1716 1717
	}
	return error;
}
1718

1719 1720 1721 1722
/*
 * Make sure there's data to read. Wait for input if we can, otherwise
 * return an appropriate error.
 */
1723
static int ipipe_prep(struct pipe_inode_info *pipe, unsigned int flags)
1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734
{
	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;
1735
	pipe_lock(pipe);
1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752

	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);
	}

1753
	pipe_unlock(pipe);
1754 1755 1756 1757 1758 1759 1760
	return ret;
}

/*
 * Make sure there's writeable room. Wait for room if we can, otherwise
 * return an appropriate error.
 */
1761
static int opipe_prep(struct pipe_inode_info *pipe, unsigned int flags)
1762 1763 1764 1765 1766 1767 1768
{
	int ret;

	/*
	 * Check ->nrbufs without the inode lock first. This function
	 * is speculative anyways, so missing one is ok.
	 */
1769
	if (pipe->nrbufs < pipe->buffers)
1770 1771 1772
		return 0;

	ret = 0;
1773
	pipe_lock(pipe);
1774

1775
	while (pipe->nrbufs >= pipe->buffers) {
1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793
		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--;
	}

1794
	pipe_unlock(pipe);
1795 1796 1797
	return ret;
}

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 1835 1836 1837 1838 1839 1840
/*
 * 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.
		 */
1841
		if (!ipipe->nrbufs || opipe->nrbufs >= opipe->buffers) {
1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861
			/* 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;
1862
		nbuf = (opipe->curbuf + opipe->nrbufs) & (opipe->buffers - 1);
1863 1864 1865 1866 1867 1868 1869 1870 1871
		obuf = opipe->bufs + nbuf;

		if (len >= ibuf->len) {
			/*
			 * Simply move the whole buffer from ipipe to opipe
			 */
			*obuf = *ibuf;
			ibuf->ops = NULL;
			opipe->nrbufs++;
1872
			ipipe->curbuf = (ipipe->curbuf + 1) & (ipipe->buffers - 1);
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 1898 1899 1900 1901 1902 1903
			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
1904 1905 1906
	if (ret > 0)
		wakeup_pipe_readers(opipe);

1907 1908 1909 1910 1911 1912
	if (input_wakeup)
		wakeup_pipe_writers(ipipe);

	return ret;
}

1913 1914 1915 1916 1917 1918 1919 1920
/*
 * 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;
1921
	int ret = 0, i = 0, nbuf;
1922 1923 1924

	/*
	 * Potential ABBA deadlock, work around it by ordering lock
1925
	 * grabbing by pipe info address. Otherwise two different processes
1926 1927
	 * could deadlock (one doing tee from A -> B, the other from B -> A).
	 */
1928
	pipe_double_lock(ipipe, opipe);
1929

1930
	do {
1931 1932 1933 1934 1935 1936 1937
		if (!opipe->readers) {
			send_sig(SIGPIPE, current, 0);
			if (!ret)
				ret = -EPIPE;
			break;
		}

1938 1939 1940 1941
		/*
		 * If we have iterated all input buffers or ran out of
		 * output room, break.
		 */
1942
		if (i >= ipipe->nrbufs || opipe->nrbufs >= opipe->buffers)
1943
			break;
1944

1945 1946
		ibuf = ipipe->bufs + ((ipipe->curbuf + i) & (ipipe->buffers-1));
		nbuf = (opipe->curbuf + opipe->nrbufs) & (opipe->buffers - 1);
1947 1948

		/*
1949 1950
		 * Get a reference to this pipe buffer,
		 * so we can copy the contents over.
1951
		 */
1952 1953 1954 1955 1956
		ibuf->ops->get(ipipe, ibuf);

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

1957
		/*
1958 1959
		 * Don't inherit the gift flag, we need to
		 * prevent multiple steals of this page.
1960
		 */
1961
		obuf->flags &= ~PIPE_BUF_FLAG_GIFT;
1962

1963 1964
		if (obuf->len > len)
			obuf->len = len;
1965

1966 1967 1968 1969 1970
		opipe->nrbufs++;
		ret += obuf->len;
		len -= obuf->len;
		i++;
	} while (len);
1971

1972 1973 1974 1975 1976 1977 1978
	/*
	 * 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;

1979 1980
	pipe_unlock(ipipe);
	pipe_unlock(opipe);
1981

1982 1983 1984
	/*
	 * If we put data in the output pipe, wakeup any potential readers.
	 */
N
Namhyung Kim 已提交
1985 1986
	if (ret > 0)
		wakeup_pipe_readers(opipe);
1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999

	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)
{
2000 2001
	struct pipe_inode_info *ipipe = get_pipe_info(in);
	struct pipe_inode_info *opipe = get_pipe_info(out);
2002
	int ret = -EINVAL;
2003 2004

	/*
2005 2006
	 * Duplicate the contents of ipipe to opipe without actually
	 * copying the data.
2007
	 */
2008 2009 2010 2011 2012
	if (ipipe && opipe && ipipe != opipe) {
		/*
		 * Keep going, unless we encounter an error. The ipipe/opipe
		 * ordering doesn't really matter.
		 */
2013
		ret = ipipe_prep(ipipe, flags);
2014
		if (!ret) {
2015
			ret = opipe_prep(opipe, flags);
2016
			if (!ret)
2017 2018 2019
				ret = link_pipe(ipipe, opipe, len, flags);
		}
	}
2020

2021
	return ret;
2022 2023
}

2024
SYSCALL_DEFINE4(tee, int, fdin, int, fdout, size_t, len, unsigned int, flags)
2025
{
2026 2027
	struct fd in;
	int error;
2028 2029 2030 2031 2032

	if (unlikely(!len))
		return 0;

	error = -EBADF;
2033 2034 2035 2036 2037 2038 2039 2040 2041
	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);
2042 2043
			}
		}
2044
 		fdput(in);
2045 2046 2047 2048
 	}

	return error;
}