splice.c 47.5 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 28
#include <linux/writeback.h>
#include <linux/buffer_head.h>
J
Jeff Garzik 已提交
29
#include <linux/module.h>
30
#include <linux/syscalls.h>
31
#include <linux/uio.h>
32
#include <linux/security.h>
33
#include <linux/gfp.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
static 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
};

165 166 167 168 169 170
/**
 * splice_to_pipe - fill passed data into a pipe
 * @pipe:	pipe to fill
 * @spd:	data to fill
 *
 * Description:
R
Randy Dunlap 已提交
171
 *    @spd contains a map of pages and len/offset tuples, along with
172 173 174
 *    the struct pipe_buf_operations associated with these pages. This
 *    function will link that data to the pipe.
 *
175
 */
176 177
ssize_t splice_to_pipe(struct pipe_inode_info *pipe,
		       struct splice_pipe_desc *spd)
178
{
179
	unsigned int spd_pages = spd->nr_pages;
180
	int ret, do_wakeup, page_nr;
181 182 183

	ret = 0;
	do_wakeup = 0;
184
	page_nr = 0;
185

186
	pipe_lock(pipe);
187 188

	for (;;) {
189
		if (!pipe->readers) {
190 191 192 193 194 195
			send_sig(SIGPIPE, current, 0);
			if (!ret)
				ret = -EPIPE;
			break;
		}

196 197
		if (pipe->nrbufs < pipe->buffers) {
			int newbuf = (pipe->curbuf + pipe->nrbufs) & (pipe->buffers - 1);
198
			struct pipe_buffer *buf = pipe->bufs + newbuf;
199

200 201 202
			buf->page = spd->pages[page_nr];
			buf->offset = spd->partial[page_nr].offset;
			buf->len = spd->partial[page_nr].len;
203
			buf->private = spd->partial[page_nr].private;
204
			buf->ops = spd->ops;
205 206 207
			if (spd->flags & SPLICE_F_GIFT)
				buf->flags |= PIPE_BUF_FLAG_GIFT;

208
			pipe->nrbufs++;
209 210 211
			page_nr++;
			ret += buf->len;

212 213
			if (pipe->inode)
				do_wakeup = 1;
214

215
			if (!--spd->nr_pages)
216
				break;
217
			if (pipe->nrbufs < pipe->buffers)
218 219 220 221 222
				continue;

			break;
		}

223
		if (spd->flags & SPLICE_F_NONBLOCK) {
224 225 226 227 228
			if (!ret)
				ret = -EAGAIN;
			break;
		}

229 230 231 232 233 234 235
		if (signal_pending(current)) {
			if (!ret)
				ret = -ERESTARTSYS;
			break;
		}

		if (do_wakeup) {
236
			smp_mb();
237 238 239
			if (waitqueue_active(&pipe->wait))
				wake_up_interruptible_sync(&pipe->wait);
			kill_fasync(&pipe->fasync_readers, SIGIO, POLL_IN);
240 241 242
			do_wakeup = 0;
		}

243 244 245
		pipe->waiting_writers++;
		pipe_wait(pipe);
		pipe->waiting_writers--;
246 247
	}

248
	pipe_unlock(pipe);
249

250 251 252 253 254
	if (do_wakeup) {
		smp_mb();
		if (waitqueue_active(&pipe->wait))
			wake_up_interruptible(&pipe->wait);
		kill_fasync(&pipe->fasync_readers, SIGIO, POLL_IN);
255 256
	}

257
	while (page_nr < spd_pages)
258
		spd->spd_release(spd, page_nr++);
259 260 261 262

	return ret;
}

263 264 265 266 267
static void spd_release_page(struct splice_pipe_desc *spd, unsigned int i)
{
	page_cache_release(spd->pages[i]);
}

268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297
/*
 * Check if we need to grow the arrays holding pages and partial page
 * descriptions.
 */
int splice_grow_spd(struct pipe_inode_info *pipe, struct splice_pipe_desc *spd)
{
	if (pipe->buffers <= PIPE_DEF_BUFFERS)
		return 0;

	spd->pages = kmalloc(pipe->buffers * sizeof(struct page *), GFP_KERNEL);
	spd->partial = kmalloc(pipe->buffers * sizeof(struct partial_page), GFP_KERNEL);

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

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

void splice_shrink_spd(struct pipe_inode_info *pipe,
		       struct splice_pipe_desc *spd)
{
	if (pipe->buffers <= PIPE_DEF_BUFFERS)
		return;

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

298
static int
299 300 301
__generic_file_splice_read(struct file *in, loff_t *ppos,
			   struct pipe_inode_info *pipe, size_t len,
			   unsigned int flags)
302 303
{
	struct address_space *mapping = in->f_mapping;
F
Fengguang Wu 已提交
304
	unsigned int loff, nr_pages, req_pages;
305 306
	struct page *pages[PIPE_DEF_BUFFERS];
	struct partial_page partial[PIPE_DEF_BUFFERS];
307
	struct page *page;
308 309
	pgoff_t index, end_index;
	loff_t isize;
310
	int error, page_nr;
311 312 313 314 315
	struct splice_pipe_desc spd = {
		.pages = pages,
		.partial = partial,
		.flags = flags,
		.ops = &page_cache_pipe_buf_ops,
316
		.spd_release = spd_release_page,
317
	};
318

319 320 321
	if (splice_grow_spd(pipe, &spd))
		return -ENOMEM;

322
	index = *ppos >> PAGE_CACHE_SHIFT;
323
	loff = *ppos & ~PAGE_CACHE_MASK;
F
Fengguang Wu 已提交
324
	req_pages = (len + loff + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
325
	nr_pages = min(req_pages, pipe->buffers);
326

327 328 329
	/*
	 * Lookup the (hopefully) full range of pages we need.
	 */
330
	spd.nr_pages = find_get_pages_contig(mapping, index, nr_pages, spd.pages);
331
	index += spd.nr_pages;
332

333 334
	/*
	 * If find_get_pages_contig() returned fewer pages than we needed,
335
	 * readahead/allocate the rest and fill in the holes.
336
	 */
337
	if (spd.nr_pages < nr_pages)
338 339
		page_cache_sync_readahead(mapping, &in->f_ra, in,
				index, req_pages - spd.nr_pages);
340

341
	error = 0;
342
	while (spd.nr_pages < nr_pages) {
343
		/*
344 345
		 * Page could be there, find_get_pages_contig() breaks on
		 * the first hole.
346
		 */
347 348 349
		page = find_get_page(mapping, index);
		if (!page) {
			/*
350
			 * page didn't exist, allocate one.
351 352 353 354 355 356
			 */
			page = page_cache_alloc_cold(mapping);
			if (!page)
				break;

			error = add_to_page_cache_lru(page, mapping, index,
357
						GFP_KERNEL);
358 359
			if (unlikely(error)) {
				page_cache_release(page);
360 361
				if (error == -EEXIST)
					continue;
362 363
				break;
			}
364 365 366 367 368
			/*
			 * add_to_page_cache() locks the page, unlock it
			 * to avoid convoluting the logic below even more.
			 */
			unlock_page(page);
369 370
		}

371
		spd.pages[spd.nr_pages++] = page;
372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391
		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);
392
		page = spd.pages[page_nr];
393

394
		if (PageReadahead(page))
395
			page_cache_async_readahead(mapping, &in->f_ra, in,
F
Fengguang Wu 已提交
396
					page, index, req_pages - page_nr);
397

398 399 400 401
		/*
		 * If the page isn't uptodate, we may need to start io on it
		 */
		if (!PageUptodate(page)) {
402 403 404 405
			/*
			 * If in nonblock mode then dont block on waiting
			 * for an in-flight io page
			 */
406
			if (flags & SPLICE_F_NONBLOCK) {
N
Nick Piggin 已提交
407
				if (!trylock_page(page)) {
408
					error = -EAGAIN;
409
					break;
410
				}
411 412
			} else
				lock_page(page);
413 414

			/*
415 416 417 418
			 * 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.
419 420 421
			 */
			if (!page->mapping) {
				unlock_page(page);
422 423 424 425 426 427 428
				page = find_or_create_page(mapping, index,
						mapping_gfp_mask(mapping));

				if (!page) {
					error = -ENOMEM;
					break;
				}
429 430
				page_cache_release(spd.pages[page_nr]);
				spd.pages[page_nr] = page;
431 432 433 434 435 436 437 438
			}
			/*
			 * page was already under io and is now done, great
			 */
			if (PageUptodate(page)) {
				unlock_page(page);
				goto fill_it;
			}
439

440 441 442 443
			/*
			 * need to read in the page
			 */
			error = mapping->a_ops->readpage(in, page);
444
			if (unlikely(error)) {
445 446 447 448 449 450
				/*
				 * 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.
				 */
451
				if (error == AOP_TRUNCATED_PAGE)
452 453
					error = 0;

454 455
				break;
			}
456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471
		}
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;
472 473

			/*
474
			 * max good bytes in this page
475
			 */
476 477
			plen = ((isize - 1) & ~PAGE_CACHE_MASK) + 1;
			if (plen <= loff)
478 479 480
				break;

			/*
481
			 * force quit after adding this page
482
			 */
483 484
			this_len = min(this_len, plen - loff);
			len = this_len;
485
		}
486

487 488
		spd.partial[page_nr].offset = loff;
		spd.partial[page_nr].len = this_len;
489
		len -= this_len;
490
		loff = 0;
491 492
		spd.nr_pages++;
		index++;
493 494
	}

495
	/*
496
	 * Release any pages at the end, if we quit early. 'page_nr' is how far
497 498 499
	 * we got, 'nr_pages' is how many pages are in the map.
	 */
	while (page_nr < nr_pages)
500
		page_cache_release(spd.pages[page_nr++]);
501
	in->f_ra.prev_pos = (loff_t)index << PAGE_CACHE_SHIFT;
502

503
	if (spd.nr_pages)
504
		error = splice_to_pipe(pipe, &spd);
505

506
	splice_shrink_spd(pipe, &spd);
507
	return error;
508 509
}

510 511 512
/**
 * generic_file_splice_read - splice data from file to a pipe
 * @in:		file to splice from
513
 * @ppos:	position in @in
514 515 516 517
 * @pipe:	pipe to splice to
 * @len:	number of bytes to splice
 * @flags:	splice modifier flags
 *
518 519 520 521 522
 * 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.
 *
523
 */
524 525 526
ssize_t generic_file_splice_read(struct file *in, loff_t *ppos,
				 struct pipe_inode_info *pipe, size_t len,
				 unsigned int flags)
527
{
528
	loff_t isize, left;
529
	int ret;
530 531 532 533 534 535 536 537

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

	left = isize - *ppos;
	if (unlikely(left < len))
		len = left;
538

539
	ret = __generic_file_splice_read(in, ppos, pipe, len, flags);
540
	if (ret > 0) {
541
		*ppos += ret;
542 543
		file_accessed(in);
	}
544 545 546

	return ret;
}
547 548
EXPORT_SYMBOL(generic_file_splice_read);

549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574
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;
}

575 576
static ssize_t kernel_write(struct file *file, const char *buf, size_t count,
			    loff_t pos)
577 578 579 580 581 582 583
{
	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() */
584
	res = vfs_write(file, (const char __user *)buf, count, &pos);
585 586 587 588 589
	set_fs(old_fs);

	return res;
}

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];
600 601 602 603 604 605 606 607 608 609 610 611 612
	pgoff_t index;
	ssize_t res;
	size_t this_len;
	int error;
	int i;
	struct splice_pipe_desc spd = {
		.pages = pages,
		.partial = partial,
		.flags = flags,
		.ops = &default_pipe_buf_ops,
		.spd_release = spd_release_page,
	};

613 614 615 616 617 618 619 620 621 622 623
	if (splice_grow_spd(pipe, &spd))
		return -ENOMEM;

	res = -ENOMEM;
	vec = __vec;
	if (pipe->buffers > PIPE_DEF_BUFFERS) {
		vec = kmalloc(pipe->buffers * sizeof(struct iovec), GFP_KERNEL);
		if (!vec)
			goto shrink_ret;
	}

624 625 626 627
	index = *ppos >> PAGE_CACHE_SHIFT;
	offset = *ppos & ~PAGE_CACHE_MASK;
	nr_pages = (len + offset + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;

628
	for (i = 0; i < nr_pages && i < pipe->buffers && len; i++) {
629 630
		struct page *page;

631
		page = alloc_page(GFP_USER);
632 633 634 635 636
		error = -ENOMEM;
		if (!page)
			goto err;

		this_len = min_t(size_t, len, PAGE_CACHE_SIZE - offset);
637
		vec[i].iov_base = (void __user *) page_address(page);
638
		vec[i].iov_len = this_len;
639
		spd.pages[i] = page;
640 641 642 643 644 645
		spd.nr_pages++;
		len -= this_len;
		offset = 0;
	}

	res = kernel_readv(in, vec, spd.nr_pages, *ppos);
A
Andrew Morton 已提交
646 647
	if (res < 0) {
		error = res;
648
		goto err;
A
Andrew Morton 已提交
649
	}
650 651 652 653 654 655 656 657

	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);
658 659
		spd.partial[i].offset = 0;
		spd.partial[i].len = this_len;
660
		if (!this_len) {
661 662
			__free_page(spd.pages[i]);
			spd.pages[i] = NULL;
663 664 665 666 667 668 669 670 671 672
			nr_freed++;
		}
		res -= this_len;
	}
	spd.nr_pages -= nr_freed;

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

673 674 675 676
shrink_ret:
	if (vec != __vec)
		kfree(vec);
	splice_shrink_spd(pipe, &spd);
677 678 679
	return res;

err:
680
	for (i = 0; i < spd.nr_pages; i++)
681
		__free_page(spd.pages[i]);
682

683 684
	res = error;
	goto shrink_ret;
685 686 687
}
EXPORT_SYMBOL(default_file_splice_read);

688
/*
689
 * Send 'sd->len' bytes to socket from 'sd->file' at position 'sd->pos'
J
Jens Axboe 已提交
690
 * using sendpage(). Return the number of bytes sent.
691
 */
692
static int pipe_to_sendpage(struct pipe_inode_info *pipe,
693 694
			    struct pipe_buffer *buf, struct splice_desc *sd)
{
695
	struct file *file = sd->u.file;
696
	loff_t pos = sd->pos;
697
	int ret, more;
698

699
	ret = buf->ops->confirm(pipe, buf);
700 701
	if (!ret) {
		more = (sd->flags & SPLICE_F_MORE) || sd->len < sd->total_len;
702 703 704 705 706
		if (file->f_op && file->f_op->sendpage)
			ret = file->f_op->sendpage(file, buf->page, buf->offset,
						   sd->len, &pos, more);
		else
			ret = -EINVAL;
707
	}
708

J
Jens Axboe 已提交
709
	return ret;
710 711 712 713 714 715 716 717 718 719 720 721 722 723 724
}

/*
 * 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.
 *
725 726 727 728 729 730
 * 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.
731
 */
732 733
int pipe_to_file(struct pipe_inode_info *pipe, struct pipe_buffer *buf,
		 struct splice_desc *sd)
734
{
735
	struct file *file = sd->u.file;
736
	struct address_space *mapping = file->f_mapping;
J
Jens Axboe 已提交
737
	unsigned int offset, this_len;
738
	struct page *page;
739
	void *fsdata;
740
	int ret;
741 742

	/*
743
	 * make sure the data in this buffer is uptodate
744
	 */
745
	ret = buf->ops->confirm(pipe, buf);
746 747
	if (unlikely(ret))
		return ret;
748 749 750

	offset = sd->pos & ~PAGE_CACHE_MASK;

J
Jens Axboe 已提交
751 752 753 754
	this_len = sd->len;
	if (this_len + offset > PAGE_CACHE_SIZE)
		this_len = PAGE_CACHE_SIZE - offset;

755 756 757 758
	ret = pagecache_write_begin(file, mapping, sd->pos, this_len,
				AOP_FLAG_UNINTERRUPTIBLE, &page, &fsdata);
	if (unlikely(ret))
		goto out;
759

760
	if (buf->page != page) {
761 762 763
		/*
		 * Careful, ->map() uses KM_USER0!
		 */
764
		char *src = buf->ops->map(pipe, buf, 1);
765
		char *dst = kmap_atomic(page, KM_USER1);
766

J
Jens Axboe 已提交
767
		memcpy(dst + offset, src + buf->offset, this_len);
768
		flush_dcache_page(page);
769
		kunmap_atomic(dst, KM_USER1);
770
		buf->ops->unmap(pipe, buf, src);
771
	}
772 773
	ret = pagecache_write_end(file, mapping, sd->pos, this_len, this_len,
				page, fsdata);
774 775 776
out:
	return ret;
}
777
EXPORT_SYMBOL(pipe_to_file);
778

779 780 781 782 783 784 785 786
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);
}

787
/**
788
 * splice_from_pipe_feed - feed available data from a pipe to a file
789 790 791 792 793
 * @pipe:	pipe to splice from
 * @sd:		information to @actor
 * @actor:	handler that splices the data
 *
 * Description:
794 795 796 797 798 799 800
 *    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.
801
 *
802 803 804 805
 *    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.
806
 */
807 808
int splice_from_pipe_feed(struct pipe_inode_info *pipe, struct splice_desc *sd,
			  splice_actor *actor)
809
{
810
	int ret;
811

812 813 814
	while (pipe->nrbufs) {
		struct pipe_buffer *buf = pipe->bufs + pipe->curbuf;
		const struct pipe_buf_operations *ops = buf->ops;
815

816 817 818
		sd->len = buf->len;
		if (sd->len > sd->total_len)
			sd->len = sd->total_len;
819

820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836
		ret = actor(pipe, buf, sd);
		if (ret <= 0) {
			if (ret == -ENODATA)
				ret = 0;
			return ret;
		}
		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);
837
			pipe->curbuf = (pipe->curbuf + 1) & (pipe->buffers - 1);
838 839 840 841
			pipe->nrbufs--;
			if (pipe->inode)
				sd->need_wakeup = true;
		}
842

843 844 845
		if (!sd->total_len)
			return 0;
	}
846

847 848 849
	return 1;
}
EXPORT_SYMBOL(splice_from_pipe_feed);
850

851 852 853 854 855 856 857 858 859 860 861 862 863 864 865
/**
 * 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
866

867 868
		if (!pipe->waiting_writers && sd->num_spliced)
			return 0;
I
Ingo Molnar 已提交
869

870 871
		if (sd->flags & SPLICE_F_NONBLOCK)
			return -EAGAIN;
872

873 874
		if (signal_pending(current))
			return -ERESTARTSYS;
875

876 877 878
		if (sd->need_wakeup) {
			wakeup_pipe_writers(pipe);
			sd->need_wakeup = false;
879 880
		}

881 882
		pipe_wait(pipe);
	}
883

884 885 886
	return 1;
}
EXPORT_SYMBOL(splice_from_pipe_next);
887

888 889
/**
 * splice_from_pipe_begin - start splicing from pipe
890
 * @sd:		information about the splice operation
891 892 893 894 895 896 897 898 899 900 901 902
 *
 * 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);
903

904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919
/**
 * 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);
920

921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937
/**
 * __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;
938

939 940 941 942 943 944 945 946 947
	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;
948
}
M
Mark Fasheh 已提交
949
EXPORT_SYMBOL(__splice_from_pipe);
950

951 952 953 954 955 956 957 958 959 960
/**
 * 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:
961
 *    See __splice_from_pipe. This function locks the pipe inode,
962 963 964
 *    otherwise it's identical to __splice_from_pipe().
 *
 */
965 966 967 968 969
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
970 971 972 973
	struct splice_desc sd = {
		.total_len = len,
		.flags = flags,
		.pos = *ppos,
974
		.u.file = out,
J
Jens Axboe 已提交
975
	};
976

977
	pipe_lock(pipe);
J
Jens Axboe 已提交
978
	ret = __splice_from_pipe(pipe, &sd, actor);
979
	pipe_unlock(pipe);
980 981 982 983

	return ret;
}

984 985
/**
 * generic_file_splice_write - splice data from a pipe to a file
986
 * @pipe:	pipe info
987
 * @out:	file to write to
988
 * @ppos:	position in @out
989 990 991
 * @len:	number of bytes to splice
 * @flags:	splice modifier flags
 *
992 993 994
 * Description:
 *    Will either move or copy pages (determined by @flags options) from
 *    the given pipe inode to the given file.
995 996
 *
 */
997 998
ssize_t
generic_file_splice_write(struct pipe_inode_info *pipe, struct file *out,
999
			  loff_t *ppos, size_t len, unsigned int flags)
1000
{
1001
	struct address_space *mapping = out->f_mapping;
1002
	struct inode *inode = mapping->host;
1003 1004 1005 1006 1007 1008
	struct splice_desc sd = {
		.total_len = len,
		.flags = flags,
		.pos = *ppos,
		.u.file = out,
	};
1009 1010
	ssize_t ret;

1011
	pipe_lock(pipe);
1012 1013 1014 1015 1016 1017 1018 1019 1020

	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);
1021 1022
		if (!ret) {
			file_update_time(out);
1023
			ret = splice_from_pipe_feed(pipe, &sd, pipe_to_file);
1024
		}
1025 1026 1027 1028
		mutex_unlock(&inode->i_mutex);
	} while (ret > 0);
	splice_from_pipe_end(pipe, &sd);

1029
	pipe_unlock(pipe);
1030 1031 1032 1033

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

J
Jens Axboe 已提交
1034
	if (ret > 0) {
1035
		unsigned long nr_pages;
1036
		int err;
1037 1038

		nr_pages = (ret + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
J
Jens Axboe 已提交
1039

1040 1041 1042 1043 1044
		err = generic_write_sync(out, *ppos, ret);
		if (err)
			ret = err;
		else
			*ppos += ret;
1045
		balance_dirty_pages_ratelimited_nr(mapping, nr_pages);
1046 1047 1048
	}

	return ret;
1049 1050
}

1051 1052
EXPORT_SYMBOL(generic_file_splice_write);

1053 1054
static int write_pipe_buf(struct pipe_inode_info *pipe, struct pipe_buffer *buf,
			  struct splice_desc *sd)
1055
{
1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067
	int ret;
	void *data;

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

	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;
1068 1069 1070 1071 1072 1073
}

static ssize_t default_file_splice_write(struct pipe_inode_info *pipe,
					 struct file *out, loff_t *ppos,
					 size_t len, unsigned int flags)
{
1074
	ssize_t ret;
1075

1076 1077 1078
	ret = splice_from_pipe(pipe, out, ppos, len, flags, write_pipe_buf);
	if (ret > 0)
		*ppos += ret;
1079

1080
	return ret;
1081 1082
}

1083 1084
/**
 * generic_splice_sendpage - splice data from a pipe to a socket
1085
 * @pipe:	pipe to splice from
1086
 * @out:	socket to write to
1087
 * @ppos:	position in @out
1088 1089 1090
 * @len:	number of bytes to splice
 * @flags:	splice modifier flags
 *
1091 1092 1093
 * Description:
 *    Will send @len bytes from the pipe to a network socket. No data copying
 *    is involved.
1094 1095
 *
 */
1096
ssize_t generic_splice_sendpage(struct pipe_inode_info *pipe, struct file *out,
1097
				loff_t *ppos, size_t len, unsigned int flags)
1098
{
1099
	return splice_from_pipe(pipe, out, ppos, len, flags, pipe_to_sendpage);
1100 1101
}

1102
EXPORT_SYMBOL(generic_splice_sendpage);
J
Jeff Garzik 已提交
1103

1104 1105 1106
/*
 * Attempt to initiate a splice from pipe to file.
 */
1107
static long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
1108
			   loff_t *ppos, size_t len, unsigned int flags)
1109
{
1110 1111
	ssize_t (*splice_write)(struct pipe_inode_info *, struct file *,
				loff_t *, size_t, unsigned int);
1112 1113
	int ret;

1114
	if (unlikely(!(out->f_mode & FMODE_WRITE)))
1115 1116
		return -EBADF;

1117 1118 1119
	if (unlikely(out->f_flags & O_APPEND))
		return -EINVAL;

1120
	ret = rw_verify_area(WRITE, out, ppos, len);
1121 1122 1123
	if (unlikely(ret < 0))
		return ret;

1124 1125 1126
	if (out->f_op && out->f_op->splice_write)
		splice_write = out->f_op->splice_write;
	else
1127 1128 1129
		splice_write = default_file_splice_write;

	return splice_write(pipe, out, ppos, len, flags);
1130 1131
}

1132 1133 1134
/*
 * Attempt to initiate a splice from a file to a pipe.
 */
1135 1136 1137
static long do_splice_to(struct file *in, loff_t *ppos,
			 struct pipe_inode_info *pipe, size_t len,
			 unsigned int flags)
1138
{
1139 1140
	ssize_t (*splice_read)(struct file *, loff_t *,
			       struct pipe_inode_info *, size_t, unsigned int);
1141 1142
	int ret;

1143
	if (unlikely(!(in->f_mode & FMODE_READ)))
1144 1145
		return -EBADF;

1146
	ret = rw_verify_area(READ, in, ppos, len);
1147 1148 1149
	if (unlikely(ret < 0))
		return ret;

1150 1151 1152
	if (in->f_op && in->f_op->splice_read)
		splice_read = in->f_op->splice_read;
	else
1153 1154 1155
		splice_read = default_file_splice_read;

	return splice_read(in, ppos, pipe, len, flags);
1156 1157
}

1158 1159 1160 1161 1162 1163 1164 1165 1166
/**
 * 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
1167
 *    pipe is cached in the process, and reused during the lifetime of
1168 1169
 *    that process.
 *
J
Jens Axboe 已提交
1170 1171 1172
 */
ssize_t splice_direct_to_actor(struct file *in, struct splice_desc *sd,
			       splice_direct_actor *actor)
1173 1174 1175 1176
{
	struct pipe_inode_info *pipe;
	long ret, bytes;
	umode_t i_mode;
J
Jens Axboe 已提交
1177 1178
	size_t len;
	int i, flags;
1179 1180 1181 1182 1183 1184

	/*
	 * 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!
	 */
1185
	i_mode = in->f_path.dentry->d_inode->i_mode;
1186 1187 1188 1189 1190 1191 1192 1193
	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;
1194
	if (unlikely(!pipe)) {
1195 1196 1197 1198 1199 1200
		pipe = alloc_pipe_info(NULL);
		if (!pipe)
			return -ENOMEM;

		/*
		 * We don't have an immediate reader, but we'll read the stuff
1201
		 * out of the pipe right after the splice_to_pipe(). So set
1202 1203 1204 1205 1206 1207 1208 1209
		 * PIPE_READERS appropriately.
		 */
		pipe->readers = 1;

		current->splice_pipe = pipe;
	}

	/*
I
Ingo Molnar 已提交
1210
	 * Do the splice.
1211 1212 1213
	 */
	ret = 0;
	bytes = 0;
J
Jens Axboe 已提交
1214 1215 1216 1217 1218 1219 1220
	len = sd->total_len;
	flags = sd->flags;

	/*
	 * Don't block on output, we have to drain the direct pipe.
	 */
	sd->flags &= ~SPLICE_F_NONBLOCK;
1221 1222

	while (len) {
1223
		size_t read_len;
1224
		loff_t pos = sd->pos, prev_pos = pos;
1225

1226
		ret = do_splice_to(in, &pos, pipe, len, flags);
1227
		if (unlikely(ret <= 0))
1228 1229 1230
			goto out_release;

		read_len = ret;
J
Jens Axboe 已提交
1231
		sd->total_len = read_len;
1232 1233 1234 1235 1236 1237

		/*
		 * 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
1238
		ret = actor(pipe, sd);
1239 1240
		if (unlikely(ret <= 0)) {
			sd->pos = prev_pos;
1241
			goto out_release;
1242
		}
1243 1244 1245

		bytes += ret;
		len -= ret;
1246
		sd->pos = pos;
1247

1248 1249
		if (ret < read_len) {
			sd->pos = prev_pos + ret;
1250
			goto out_release;
1251
		}
1252 1253
	}

1254
done:
1255
	pipe->nrbufs = pipe->curbuf = 0;
1256
	file_accessed(in);
1257 1258 1259 1260 1261 1262 1263
	return bytes;

out_release:
	/*
	 * If we did an incomplete transfer we must release
	 * the pipe buffers in question:
	 */
1264
	for (i = 0; i < pipe->buffers; i++) {
1265 1266 1267 1268 1269 1270 1271 1272
		struct pipe_buffer *buf = pipe->bufs + i;

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

1273 1274
	if (!bytes)
		bytes = ret;
J
Jens Axboe 已提交
1275

1276
	goto done;
J
Jens Axboe 已提交
1277 1278 1279 1280 1281 1282
}
EXPORT_SYMBOL(splice_direct_to_actor);

static int direct_splice_actor(struct pipe_inode_info *pipe,
			       struct splice_desc *sd)
{
1283
	struct file *file = sd->u.file;
J
Jens Axboe 已提交
1284

1285 1286
	return do_splice_from(pipe, file, &file->f_pos, sd->total_len,
			      sd->flags);
J
Jens Axboe 已提交
1287 1288
}

1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303
/**
 * 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
1304 1305 1306 1307 1308 1309 1310 1311
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,
1312
		.u.file		= out,
J
Jens Axboe 已提交
1313
	};
1314
	long ret;
J
Jens Axboe 已提交
1315 1316

	ret = splice_direct_to_actor(in, &sd, direct_splice_actor);
1317
	if (ret > 0)
1318
		*ppos = sd.pos;
1319

J
Jens Axboe 已提交
1320
	return ret;
1321 1322
}

1323 1324 1325
static int splice_pipe_to_pipe(struct pipe_inode_info *ipipe,
			       struct pipe_inode_info *opipe,
			       size_t len, unsigned int flags);
1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338
/*
 * After the inode slimming patch, i_pipe/i_bdev/i_cdev share the same
 * location, so checking ->i_pipe is not enough to verify that this is a
 * pipe.
 */
static inline struct pipe_inode_info *pipe_info(struct inode *inode)
{
	if (S_ISFIFO(inode->i_mode))
		return inode->i_pipe;

	return NULL;
}

1339 1340 1341
/*
 * Determine where to splice to/from.
 */
1342 1343 1344
static long do_splice(struct file *in, loff_t __user *off_in,
		      struct file *out, loff_t __user *off_out,
		      size_t len, unsigned int flags)
1345
{
1346 1347
	struct pipe_inode_info *ipipe;
	struct pipe_inode_info *opipe;
1348
	loff_t offset, *off;
J
Jens Axboe 已提交
1349
	long ret;
1350

1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371
	ipipe = pipe_info(in->f_path.dentry->d_inode);
	opipe = pipe_info(out->f_path.dentry->d_inode);

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

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

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

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

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

	if (ipipe) {
1372 1373
		if (off_in)
			return -ESPIPE;
1374
		if (off_out) {
1375 1376
			if (!out->f_op || !out->f_op->llseek ||
			    out->f_op->llseek == no_llseek)
1377
				return -EINVAL;
1378
			if (copy_from_user(&offset, off_out, sizeof(loff_t)))
1379
				return -EFAULT;
1380 1381 1382
			off = &offset;
		} else
			off = &out->f_pos;
1383

1384
		ret = do_splice_from(ipipe, out, off, len, flags);
J
Jens Axboe 已提交
1385 1386 1387 1388 1389

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

		return ret;
1390
	}
1391

1392
	if (opipe) {
1393 1394
		if (off_out)
			return -ESPIPE;
1395
		if (off_in) {
1396 1397
			if (!in->f_op || !in->f_op->llseek ||
			    in->f_op->llseek == no_llseek)
1398
				return -EINVAL;
1399
			if (copy_from_user(&offset, off_in, sizeof(loff_t)))
1400
				return -EFAULT;
1401 1402 1403
			off = &offset;
		} else
			off = &in->f_pos;
1404

1405
		ret = do_splice_to(in, off, opipe, len, flags);
J
Jens Axboe 已提交
1406 1407 1408 1409 1410

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

		return ret;
1411
	}
1412 1413 1414 1415

	return -EINVAL;
}

1416 1417 1418 1419 1420 1421 1422 1423 1424
/*
 * 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,
1425 1426
				struct partial_page *partial, int aligned,
				unsigned int pipe_buffers)
1427 1428 1429 1430 1431
{
	int buffers = 0, error = 0;

	while (nr_vecs) {
		unsigned long off, npages;
1432
		struct iovec entry;
1433 1434 1435 1436
		void __user *base;
		size_t len;
		int i;

1437
		error = -EFAULT;
N
Nick Piggin 已提交
1438
		if (copy_from_user(&entry, iov, sizeof(entry)))
1439 1440
			break;

1441 1442 1443
		base = entry.iov_base;
		len = entry.iov_len;

1444 1445 1446
		/*
		 * Sanity check this iovec. 0 read succeeds.
		 */
1447
		error = 0;
1448 1449 1450
		if (unlikely(!len))
			break;
		error = -EFAULT;
1451
		if (!access_ok(VERIFY_READ, base, len))
1452 1453 1454 1455 1456 1457 1458
			break;

		/*
		 * Get this base offset and number of pages, then map
		 * in the user pages.
		 */
		off = (unsigned long) base & ~PAGE_MASK;
1459 1460 1461 1462 1463 1464 1465 1466 1467

		/*
		 * 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;

1468
		npages = (off + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
1469 1470
		if (npages > pipe_buffers - buffers)
			npages = pipe_buffers - buffers;
1471

N
Nick Piggin 已提交
1472 1473
		error = get_user_pages_fast((unsigned long)base, npages,
					0, &pages[buffers]);
1474 1475 1476 1477 1478 1479 1480 1481

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

		/*
		 * Fill this contiguous range into the partial page map.
		 */
		for (i = 0; i < error; i++) {
1482
			const int plen = min_t(size_t, len, PAGE_SIZE - off);
1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504

			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.
		 */
1505
		if (error < npages || buffers == pipe_buffers)
1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517
			break;

		nr_vecs--;
		iov++;
	}

	if (buffers)
		return buffers;

	return error;
}

1518 1519 1520 1521 1522 1523
static int pipe_to_user(struct pipe_inode_info *pipe, struct pipe_buffer *buf,
			struct splice_desc *sd)
{
	char *src;
	int ret;

1524
	ret = buf->ops->confirm(pipe, buf);
1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551
	if (unlikely(ret))
		return ret;

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

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

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

1552
	buf->ops->unmap(pipe, buf, src);
1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575
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;

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

1576
	pipe_lock(pipe);
1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602

	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;
		}

1603 1604 1605 1606 1607
		if (unlikely(!access_ok(VERIFY_WRITE, base, len))) {
			error = -EFAULT;
			break;
		}

1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630
		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++;
	}

1631
	pipe_unlock(pipe);
1632 1633 1634 1635 1636 1637 1638

	if (!ret)
		ret = error;

	return ret;
}

1639 1640 1641 1642 1643
/*
 * 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.
 */
1644 1645
static long vmsplice_to_pipe(struct file *file, const struct iovec __user *iov,
			     unsigned long nr_segs, unsigned int flags)
1646
{
1647
	struct pipe_inode_info *pipe;
1648 1649
	struct page *pages[PIPE_DEF_BUFFERS];
	struct partial_page partial[PIPE_DEF_BUFFERS];
1650 1651 1652 1653 1654
	struct splice_pipe_desc spd = {
		.pages = pages,
		.partial = partial,
		.flags = flags,
		.ops = &user_page_pipe_buf_ops,
1655
		.spd_release = spd_release_page,
1656
	};
1657
	long ret;
1658

1659
	pipe = pipe_info(file->f_path.dentry->d_inode);
1660
	if (!pipe)
1661 1662
		return -EBADF;

1663 1664 1665 1666 1667 1668
	if (splice_grow_spd(pipe, &spd))
		return -ENOMEM;

	spd.nr_pages = get_iovec_page_array(iov, nr_segs, spd.pages,
					    spd.partial, flags & SPLICE_F_GIFT,
					    pipe->buffers);
1669
	if (spd.nr_pages <= 0)
1670 1671 1672
		ret = spd.nr_pages;
	else
		ret = splice_to_pipe(pipe, &spd);
1673

1674 1675
	splice_shrink_spd(pipe, &spd);
	return ret;
1676 1677
}

1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693
/*
 * 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().
 *
 */
1694 1695
SYSCALL_DEFINE4(vmsplice, int, fd, const struct iovec __user *, iov,
		unsigned long, nr_segs, unsigned int, flags)
1696 1697 1698 1699 1700
{
	struct file *file;
	long error;
	int fput;

1701 1702 1703 1704 1705
	if (unlikely(nr_segs > UIO_MAXIOV))
		return -EINVAL;
	else if (unlikely(!nr_segs))
		return 0;

1706 1707 1708 1709
	error = -EBADF;
	file = fget_light(fd, &fput);
	if (file) {
		if (file->f_mode & FMODE_WRITE)
1710 1711 1712
			error = vmsplice_to_pipe(file, iov, nr_segs, flags);
		else if (file->f_mode & FMODE_READ)
			error = vmsplice_to_user(file, iov, nr_segs, flags);
1713 1714 1715 1716 1717 1718 1719

		fput_light(file, fput);
	}

	return error;
}

1720 1721 1722
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)
1723 1724 1725 1726 1727 1728 1729 1730 1731
{
	long error;
	struct file *in, *out;
	int fput_in, fput_out;

	if (unlikely(!len))
		return 0;

	error = -EBADF;
1732
	in = fget_light(fd_in, &fput_in);
1733 1734
	if (in) {
		if (in->f_mode & FMODE_READ) {
1735
			out = fget_light(fd_out, &fput_out);
1736 1737
			if (out) {
				if (out->f_mode & FMODE_WRITE)
1738 1739 1740
					error = do_splice(in, off_in,
							  out, off_out,
							  len, flags);
1741 1742 1743 1744 1745 1746 1747 1748 1749
				fput_light(out, fput_out);
			}
		}

		fput_light(in, fput_in);
	}

	return error;
}
1750

1751 1752 1753 1754
/*
 * Make sure there's data to read. Wait for input if we can, otherwise
 * return an appropriate error.
 */
1755
static int ipipe_prep(struct pipe_inode_info *pipe, unsigned int flags)
1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766
{
	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;
1767
	pipe_lock(pipe);
1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784

	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);
	}

1785
	pipe_unlock(pipe);
1786 1787 1788 1789 1790 1791 1792
	return ret;
}

/*
 * Make sure there's writeable room. Wait for room if we can, otherwise
 * return an appropriate error.
 */
1793
static int opipe_prep(struct pipe_inode_info *pipe, unsigned int flags)
1794 1795 1796 1797 1798 1799 1800
{
	int ret;

	/*
	 * Check ->nrbufs without the inode lock first. This function
	 * is speculative anyways, so missing one is ok.
	 */
1801
	if (pipe->nrbufs < pipe->buffers)
1802 1803 1804
		return 0;

	ret = 0;
1805
	pipe_lock(pipe);
1806

1807
	while (pipe->nrbufs >= pipe->buffers) {
1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825
		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--;
	}

1826
	pipe_unlock(pipe);
1827 1828 1829
	return ret;
}

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 1869 1870 1871 1872
/*
 * 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.
		 */
1873
		if (!ipipe->nrbufs || opipe->nrbufs >= opipe->buffers) {
1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893
			/* 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;
1894
		nbuf = (opipe->curbuf + opipe->nrbufs) & (opipe->buffers - 1);
1895 1896 1897 1898 1899 1900 1901 1902 1903
		obuf = opipe->bufs + nbuf;

		if (len >= ibuf->len) {
			/*
			 * Simply move the whole buffer from ipipe to opipe
			 */
			*obuf = *ibuf;
			ibuf->ops = NULL;
			opipe->nrbufs++;
1904
			ipipe->curbuf = (ipipe->curbuf + 1) & (ipipe->buffers - 1);
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 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947
			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.
	 */
	if (ret > 0) {
		smp_mb();
		if (waitqueue_active(&opipe->wait))
			wake_up_interruptible(&opipe->wait);
		kill_fasync(&opipe->fasync_readers, SIGIO, POLL_IN);
	}
	if (input_wakeup)
		wakeup_pipe_writers(ipipe);

	return ret;
}

1948 1949 1950 1951 1952 1953 1954 1955
/*
 * 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;
1956
	int ret = 0, i = 0, nbuf;
1957 1958 1959

	/*
	 * Potential ABBA deadlock, work around it by ordering lock
1960
	 * grabbing by pipe info address. Otherwise two different processes
1961 1962
	 * could deadlock (one doing tee from A -> B, the other from B -> A).
	 */
1963
	pipe_double_lock(ipipe, opipe);
1964

1965
	do {
1966 1967 1968 1969 1970 1971 1972
		if (!opipe->readers) {
			send_sig(SIGPIPE, current, 0);
			if (!ret)
				ret = -EPIPE;
			break;
		}

1973 1974 1975 1976
		/*
		 * If we have iterated all input buffers or ran out of
		 * output room, break.
		 */
1977
		if (i >= ipipe->nrbufs || opipe->nrbufs >= opipe->buffers)
1978
			break;
1979

1980 1981
		ibuf = ipipe->bufs + ((ipipe->curbuf + i) & (ipipe->buffers-1));
		nbuf = (opipe->curbuf + opipe->nrbufs) & (opipe->buffers - 1);
1982 1983

		/*
1984 1985
		 * Get a reference to this pipe buffer,
		 * so we can copy the contents over.
1986
		 */
1987 1988 1989 1990 1991
		ibuf->ops->get(ipipe, ibuf);

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

1992
		/*
1993 1994
		 * Don't inherit the gift flag, we need to
		 * prevent multiple steals of this page.
1995
		 */
1996
		obuf->flags &= ~PIPE_BUF_FLAG_GIFT;
1997

1998 1999
		if (obuf->len > len)
			obuf->len = len;
2000

2001 2002 2003 2004 2005
		opipe->nrbufs++;
		ret += obuf->len;
		len -= obuf->len;
		i++;
	} while (len);
2006

2007 2008 2009 2010 2011 2012 2013
	/*
	 * 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;

2014 2015
	pipe_unlock(ipipe);
	pipe_unlock(opipe);
2016

2017 2018 2019 2020
	/*
	 * If we put data in the output pipe, wakeup any potential readers.
	 */
	if (ret > 0) {
2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038
		smp_mb();
		if (waitqueue_active(&opipe->wait))
			wake_up_interruptible(&opipe->wait);
		kill_fasync(&opipe->fasync_readers, SIGIO, POLL_IN);
	}

	return ret;
}

/*
 * This is a tee(1) implementation that works on pipes. It doesn't copy
 * any data, it simply references the 'in' pages on the 'out' pipe.
 * The 'flags' used are the SPLICE_F_* variants, currently the only
 * applicable one is SPLICE_F_NONBLOCK.
 */
static long do_tee(struct file *in, struct file *out, size_t len,
		   unsigned int flags)
{
2039 2040
	struct pipe_inode_info *ipipe = pipe_info(in->f_path.dentry->d_inode);
	struct pipe_inode_info *opipe = pipe_info(out->f_path.dentry->d_inode);
2041
	int ret = -EINVAL;
2042 2043

	/*
2044 2045
	 * Duplicate the contents of ipipe to opipe without actually
	 * copying the data.
2046
	 */
2047 2048 2049 2050 2051
	if (ipipe && opipe && ipipe != opipe) {
		/*
		 * Keep going, unless we encounter an error. The ipipe/opipe
		 * ordering doesn't really matter.
		 */
2052
		ret = ipipe_prep(ipipe, flags);
2053
		if (!ret) {
2054
			ret = opipe_prep(opipe, flags);
2055
			if (!ret)
2056 2057 2058
				ret = link_pipe(ipipe, opipe, len, flags);
		}
	}
2059

2060
	return ret;
2061 2062
}

2063
SYSCALL_DEFINE4(tee, int, fdin, int, fdout, size_t, len, unsigned int, flags)
2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088
{
	struct file *in;
	int error, fput_in;

	if (unlikely(!len))
		return 0;

	error = -EBADF;
	in = fget_light(fdin, &fput_in);
	if (in) {
		if (in->f_mode & FMODE_READ) {
			int fput_out;
			struct file *out = fget_light(fdout, &fput_out);

			if (out) {
				if (out->f_mode & FMODE_WRITE)
					error = do_tee(in, out, len, flags);
				fput_light(out, fput_out);
			}
		}
 		fput_light(in, fput_in);
 	}

	return error;
}