splice.c 15.2 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
/*
 * "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
 * Jens to support splicing to files and fixing the initial implementation
 * bugs.
 *
 * Copyright (C) 2005 Jens Axboe <axboe@suse.de>
 * Copyright (C) 2005 Linus Torvalds <torvalds@osdl.org>
 *
 */
#include <linux/fs.h>
#include <linux/file.h>
#include <linux/pagemap.h>
#include <linux/pipe_fs_i.h>
#include <linux/mm_inline.h>
24
#include <linux/swap.h>
25 26
#include <linux/writeback.h>
#include <linux/buffer_head.h>
J
Jeff Garzik 已提交
27
#include <linux/module.h>
28
#include <linux/syscalls.h>
29 30 31 32 33 34 35 36 37 38 39

/*
 * Passed to the actors
 */
struct splice_desc {
	unsigned int len, total_len;	/* current and remaining length */
	unsigned int flags;		/* splice flags */
	struct file *file;		/* file to read/write */
	loff_t pos;			/* file position */
};

40 41 42 43
static int page_cache_pipe_buf_steal(struct pipe_inode_info *info,
				     struct pipe_buffer *buf)
{
	struct page *page = buf->page;
44
	struct address_space *mapping = page_mapping(page);
45 46 47 48

	WARN_ON(!PageLocked(page));
	WARN_ON(!PageUptodate(page));

49 50 51 52
	if (PagePrivate(page))
		try_to_release_page(page, mapping_gfp_mask(mapping));

	if (!remove_mapping(mapping, page))
53 54 55 56 57 58 59 60 61 62 63 64 65 66 67
		return 1;

	if (PageLRU(page)) {
		struct zone *zone = page_zone(page);

		spin_lock_irq(&zone->lru_lock);
		BUG_ON(!PageLRU(page));
		__ClearPageLRU(page);
		del_page_from_lru(zone, page);
		spin_unlock_irq(&zone->lru_lock);
	}

	return 0;
}

68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98
static void page_cache_pipe_buf_release(struct pipe_inode_info *info,
					struct pipe_buffer *buf)
{
	page_cache_release(buf->page);
	buf->page = NULL;
}

static void *page_cache_pipe_buf_map(struct file *file,
				     struct pipe_inode_info *info,
				     struct pipe_buffer *buf)
{
	struct page *page = buf->page;

	lock_page(page);

	if (!PageUptodate(page)) {
		unlock_page(page);
		return ERR_PTR(-EIO);
	}

	if (!page->mapping) {
		unlock_page(page);
		return ERR_PTR(-ENODATA);
	}

	return kmap(buf->page);
}

static void page_cache_pipe_buf_unmap(struct pipe_inode_info *info,
				      struct pipe_buffer *buf)
{
99
	unlock_page(buf->page);
100 101 102 103 104 105 106 107
	kunmap(buf->page);
}

static struct pipe_buf_operations page_cache_pipe_buf_ops = {
	.can_merge = 0,
	.map = page_cache_pipe_buf_map,
	.unmap = page_cache_pipe_buf_unmap,
	.release = page_cache_pipe_buf_release,
108
	.steal = page_cache_pipe_buf_steal,
109 110 111 112
};

static ssize_t move_to_pipe(struct inode *inode, struct page **pages,
			    int nr_pages, unsigned long offset,
113
			    unsigned long len, unsigned int flags)
114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165
{
	struct pipe_inode_info *info;
	int ret, do_wakeup, i;

	ret = 0;
	do_wakeup = 0;
	i = 0;

	mutex_lock(PIPE_MUTEX(*inode));

	info = inode->i_pipe;
	for (;;) {
		int bufs;

		if (!PIPE_READERS(*inode)) {
			send_sig(SIGPIPE, current, 0);
			if (!ret)
				ret = -EPIPE;
			break;
		}

		bufs = info->nrbufs;
		if (bufs < PIPE_BUFFERS) {
			int newbuf = (info->curbuf + bufs) & (PIPE_BUFFERS - 1);
			struct pipe_buffer *buf = info->bufs + newbuf;
			struct page *page = pages[i++];
			unsigned long this_len;

			this_len = PAGE_CACHE_SIZE - offset;
			if (this_len > len)
				this_len = len;

			buf->page = page;
			buf->offset = offset;
			buf->len = this_len;
			buf->ops = &page_cache_pipe_buf_ops;
			info->nrbufs = ++bufs;
			do_wakeup = 1;

			ret += this_len;
			len -= this_len;
			offset = 0;
			if (!--nr_pages)
				break;
			if (!len)
				break;
			if (bufs < PIPE_BUFFERS)
				continue;

			break;
		}

166 167 168 169 170 171
		if (flags & SPLICE_F_NONBLOCK) {
			if (!ret)
				ret = -EAGAIN;
			break;
		}

172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203
		if (signal_pending(current)) {
			if (!ret)
				ret = -ERESTARTSYS;
			break;
		}

		if (do_wakeup) {
			wake_up_interruptible_sync(PIPE_WAIT(*inode));
			kill_fasync(PIPE_FASYNC_READERS(*inode), SIGIO,
				    POLL_IN);
			do_wakeup = 0;
		}

		PIPE_WAITING_WRITERS(*inode)++;
		pipe_wait(inode);
		PIPE_WAITING_WRITERS(*inode)--;
	}

	mutex_unlock(PIPE_MUTEX(*inode));

	if (do_wakeup) {
		wake_up_interruptible(PIPE_WAIT(*inode));
		kill_fasync(PIPE_FASYNC_READERS(*inode), SIGIO, POLL_IN);
	}

	while (i < nr_pages)
		page_cache_release(pages[i++]);

	return ret;
}

static int __generic_file_splice_read(struct file *in, struct inode *pipe,
204
				      size_t len, unsigned int flags)
205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243
{
	struct address_space *mapping = in->f_mapping;
	unsigned int offset, nr_pages;
	struct page *pages[PIPE_BUFFERS], *shadow[PIPE_BUFFERS];
	struct page *page;
	pgoff_t index, pidx;
	int i, j;

	index = in->f_pos >> PAGE_CACHE_SHIFT;
	offset = in->f_pos & ~PAGE_CACHE_MASK;
	nr_pages = (len + offset + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;

	if (nr_pages > PIPE_BUFFERS)
		nr_pages = PIPE_BUFFERS;

	/*
	 * initiate read-ahead on this page range
	 */
	do_page_cache_readahead(mapping, in, index, nr_pages);

	/*
	 * Get as many pages from the page cache as possible..
	 * Start IO on the page cache entries we create (we
	 * can assume that any pre-existing ones we find have
	 * already had IO started on them).
	 */
	i = find_get_pages(mapping, index, nr_pages, pages);

	/*
	 * common case - we found all pages and they are contiguous,
	 * kick them off
	 */
	if (i && (pages[i - 1]->index == index + i - 1))
		goto splice_them;

	/*
	 * fill shadow[] with pages at the right locations, so we only
	 * have to fill holes
	 */
244 245 246
	memset(shadow, 0, nr_pages * sizeof(struct page *));
	for (j = 0; j < i; j++)
		shadow[pages[j]->index - index] = pages[j];
247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291

	/*
	 * now fill in the holes
	 */
	for (i = 0, pidx = index; i < nr_pages; pidx++, i++) {
		int error;

		if (shadow[i])
			continue;

		/*
		 * no page there, look one up / create it
		 */
		page = find_or_create_page(mapping, pidx,
						   mapping_gfp_mask(mapping));
		if (!page)
			break;

		if (PageUptodate(page))
			unlock_page(page);
		else {
			error = mapping->a_ops->readpage(in, page);

			if (unlikely(error)) {
				page_cache_release(page);
				break;
			}
		}
		shadow[i] = page;
	}

	if (!i) {
		for (i = 0; i < nr_pages; i++) {
			 if (shadow[i])
				page_cache_release(shadow[i]);
		}
		return 0;
	}

	memcpy(pages, shadow, i * sizeof(struct page *));

	/*
	 * Now we splice them into the pipe..
	 */
splice_them:
292
	return move_to_pipe(pipe, pages, i, offset, len, flags);
293 294 295 296 297 298 299 300 301 302 303
}

ssize_t generic_file_splice_read(struct file *in, struct inode *pipe,
				 size_t len, unsigned int flags)
{
	ssize_t spliced;
	int ret;

	ret = 0;
	spliced = 0;
	while (len) {
304
		ret = __generic_file_splice_read(in, pipe, len, flags);
305 306 307 308 309 310 311

		if (ret <= 0)
			break;

		in->f_pos += ret;
		len -= ret;
		spliced += ret;
312 313 314 315 316

		if (!(flags & SPLICE_F_NONBLOCK))
			continue;
		ret = -EAGAIN;
		break;
317 318 319 320 321 322 323 324 325
	}

	if (spliced)
		return spliced;

	return ret;
}

/*
326 327
 * Send 'sd->len' bytes to socket from 'sd->file' at position 'sd->pos'
 * using sendpage().
328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385
 */
static int pipe_to_sendpage(struct pipe_inode_info *info,
			    struct pipe_buffer *buf, struct splice_desc *sd)
{
	struct file *file = sd->file;
	loff_t pos = sd->pos;
	unsigned int offset;
	ssize_t ret;
	void *ptr;

	/*
	 * sub-optimal, but we are limited by the pipe ->map. we don't
	 * need a kmap'ed buffer here, we just want to make sure we
	 * have the page pinned if the pipe page originates from the
	 * page cache
	 */
	ptr = buf->ops->map(file, info, buf);
	if (IS_ERR(ptr))
		return PTR_ERR(ptr);

	offset = pos & ~PAGE_CACHE_MASK;

	ret = file->f_op->sendpage(file, buf->page, offset, sd->len, &pos,
					sd->len < sd->total_len);

	buf->ops->unmap(info, buf);
	if (ret == sd->len)
		return 0;

	return -EIO;
}

/*
 * 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.
 *
 * For now we just do the slower thing and always copy pages over, it's
 * easier than migrating pages from the pipe to the target file. For the
 * case of doing file | file splicing, the migrate approach had some LRU
 * nastiness...
 */
static int pipe_to_file(struct pipe_inode_info *info, struct pipe_buffer *buf,
			struct splice_desc *sd)
{
	struct file *file = sd->file;
	struct address_space *mapping = file->f_mapping;
	unsigned int offset;
	struct page *page;
	pgoff_t index;
386
	char *src;
387
	int ret, stolen;
388 389 390 391 392 393 394 395 396 397

	/*
	 * after this, page will be locked and unmapped
	 */
	src = buf->ops->map(file, info, buf);
	if (IS_ERR(src))
		return PTR_ERR(src);

	index = sd->pos >> PAGE_CACHE_SHIFT;
	offset = sd->pos & ~PAGE_CACHE_MASK;
398
	stolen = 0;
399 400

	/*
401
	 * reuse buf page, if SPLICE_F_MOVE is set
402
	 */
403 404 405 406 407
	if (sd->flags & SPLICE_F_MOVE) {
		if (buf->ops->steal(info, buf))
			goto find_page;

		page = buf->page;
408
		stolen = 1;
409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443
		if (add_to_page_cache_lru(page, mapping, index,
						mapping_gfp_mask(mapping)))
			goto find_page;
	} else {
find_page:
		ret = -ENOMEM;
		page = find_or_create_page(mapping, index,
						mapping_gfp_mask(mapping));
		if (!page)
			goto out;

		/*
		 * If the page is uptodate, it is also locked. If it isn't
		 * uptodate, we can mark it uptodate if we are filling the
		 * full page. Otherwise we need to read it in first...
		 */
		if (!PageUptodate(page)) {
			if (sd->len < PAGE_CACHE_SIZE) {
				ret = mapping->a_ops->readpage(file, page);
				if (unlikely(ret))
					goto out;

				lock_page(page);

				if (!PageUptodate(page)) {
					/*
					 * page got invalidated, repeat
					 */
					if (!page->mapping) {
						unlock_page(page);
						page_cache_release(page);
						goto find_page;
					}
					ret = -EIO;
					goto out;
444
				}
445 446 447
			} else {
				WARN_ON(!PageLocked(page));
				SetPageUptodate(page);
448 449 450 451 452
			}
		}
	}

	ret = mapping->a_ops->prepare_write(file, page, 0, sd->len);
453 454 455 456
	if (ret == AOP_TRUNCATED_PAGE) {
		page_cache_release(page);
		goto find_page;
	} else if (ret)
457 458
		goto out;

459
	if (!stolen) {
460 461 462 463 464 465
		char *dst = kmap_atomic(page, KM_USER0);

		memcpy(dst + offset, src + buf->offset, sd->len);
		flush_dcache_page(page);
		kunmap_atomic(dst, KM_USER0);
	}
466 467

	ret = mapping->a_ops->commit_write(file, page, 0, sd->len);
468 469 470 471
	if (ret == AOP_TRUNCATED_PAGE) {
		page_cache_release(page);
		goto find_page;
	} else if (ret)
472 473
		goto out;

474
	balance_dirty_pages_ratelimited(mapping);
475
out:
476
	if (!stolen) {
477
		page_cache_release(page);
478 479
		unlock_page(page);
	}
480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552
	buf->ops->unmap(info, buf);
	return ret;
}

typedef int (splice_actor)(struct pipe_inode_info *, struct pipe_buffer *,
			   struct splice_desc *);

static ssize_t move_from_pipe(struct inode *inode, struct file *out,
			      size_t len, unsigned int flags,
			      splice_actor *actor)
{
	struct pipe_inode_info *info;
	int ret, do_wakeup, err;
	struct splice_desc sd;

	ret = 0;
	do_wakeup = 0;

	sd.total_len = len;
	sd.flags = flags;
	sd.file = out;
	sd.pos = out->f_pos;

	mutex_lock(PIPE_MUTEX(*inode));

	info = inode->i_pipe;
	for (;;) {
		int bufs = info->nrbufs;

		if (bufs) {
			int curbuf = info->curbuf;
			struct pipe_buffer *buf = info->bufs + curbuf;
			struct pipe_buf_operations *ops = buf->ops;

			sd.len = buf->len;
			if (sd.len > sd.total_len)
				sd.len = sd.total_len;

			err = actor(info, buf, &sd);
			if (err) {
				if (!ret && err != -ENODATA)
					ret = err;

				break;
			}

			ret += sd.len;
			buf->offset += sd.len;
			buf->len -= sd.len;
			if (!buf->len) {
				buf->ops = NULL;
				ops->release(info, buf);
				curbuf = (curbuf + 1) & (PIPE_BUFFERS - 1);
				info->curbuf = curbuf;
				info->nrbufs = --bufs;
				do_wakeup = 1;
			}

			sd.pos += sd.len;
			sd.total_len -= sd.len;
			if (!sd.total_len)
				break;
		}

		if (bufs)
			continue;
		if (!PIPE_WRITERS(*inode))
			break;
		if (!PIPE_WAITING_WRITERS(*inode)) {
			if (ret)
				break;
		}

553 554 555 556 557 558
		if (flags & SPLICE_F_NONBLOCK) {
			if (!ret)
				ret = -EAGAIN;
			break;
		}

559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590
		if (signal_pending(current)) {
			if (!ret)
				ret = -ERESTARTSYS;
			break;
		}

		if (do_wakeup) {
			wake_up_interruptible_sync(PIPE_WAIT(*inode));
			kill_fasync(PIPE_FASYNC_WRITERS(*inode),SIGIO,POLL_OUT);
			do_wakeup = 0;
		}

		pipe_wait(inode);
	}

	mutex_unlock(PIPE_MUTEX(*inode));

	if (do_wakeup) {
		wake_up_interruptible(PIPE_WAIT(*inode));
		kill_fasync(PIPE_FASYNC_WRITERS(*inode), SIGIO, POLL_OUT);
	}

	mutex_lock(&out->f_mapping->host->i_mutex);
	out->f_pos = sd.pos;
	mutex_unlock(&out->f_mapping->host->i_mutex);
	return ret;

}

ssize_t generic_file_splice_write(struct inode *inode, struct file *out,
				  size_t len, unsigned int flags)
{
591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611
	struct address_space *mapping = out->f_mapping;
	ssize_t ret = move_from_pipe(inode, out, len, flags, pipe_to_file);

	/*
	 * if file or inode is SYNC and we actually wrote some data, sync it
	 */
	if (unlikely((out->f_flags & O_SYNC) || IS_SYNC(mapping->host))
	    && ret > 0) {
		struct inode *inode = mapping->host;
		int err;

		mutex_lock(&inode->i_mutex);
		err = generic_osync_inode(mapping->host, mapping,
						OSYNC_METADATA|OSYNC_DATA);
		mutex_unlock(&inode->i_mutex);

		if (err)
			ret = err;
	}

	return ret;
612 613 614 615 616 617 618 619
}

ssize_t generic_splice_sendpage(struct inode *inode, struct file *out,
				size_t len, unsigned int flags)
{
	return move_from_pipe(inode, out, len, flags, pipe_to_sendpage);
}

J
Jeff Garzik 已提交
620 621 622
EXPORT_SYMBOL(generic_file_splice_write);
EXPORT_SYMBOL(generic_file_splice_read);

623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712
static long do_splice_from(struct inode *pipe, struct file *out, size_t len,
			   unsigned int flags)
{
	loff_t pos;
	int ret;

	if (!out->f_op || !out->f_op->splice_write)
		return -EINVAL;

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

	pos = out->f_pos;
	ret = rw_verify_area(WRITE, out, &pos, len);
	if (unlikely(ret < 0))
		return ret;

	return out->f_op->splice_write(pipe, out, len, flags);
}

static long do_splice_to(struct file *in, struct inode *pipe, size_t len,
			 unsigned int flags)
{
	loff_t pos, isize, left;
	int ret;

	if (!in->f_op || !in->f_op->splice_read)
		return -EINVAL;

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

	pos = in->f_pos;
	ret = rw_verify_area(READ, in, &pos, len);
	if (unlikely(ret < 0))
		return ret;

	isize = i_size_read(in->f_mapping->host);
	if (unlikely(in->f_pos >= isize))
		return 0;
	
	left = isize - in->f_pos;
	if (left < len)
		len = left;

	return in->f_op->splice_read(in, pipe, len, flags);
}

static long do_splice(struct file *in, struct file *out, size_t len,
		      unsigned int flags)
{
	struct inode *pipe;

	pipe = in->f_dentry->d_inode;
	if (pipe->i_pipe)
		return do_splice_from(pipe, out, len, flags);

	pipe = out->f_dentry->d_inode;
	if (pipe->i_pipe)
		return do_splice_to(in, pipe, len, flags);

	return -EINVAL;
}

asmlinkage long sys_splice(int fdin, int fdout, size_t len, unsigned int flags)
{
	long error;
	struct file *in, *out;
	int fput_in, fput_out;

	if (unlikely(!len))
		return 0;

	error = -EBADF;
	in = fget_light(fdin, &fput_in);
	if (in) {
		if (in->f_mode & FMODE_READ) {
			out = fget_light(fdout, &fput_out);
			if (out) {
				if (out->f_mode & FMODE_WRITE)
					error = do_splice(in, out, len, flags);
				fput_light(out, fput_out);
			}
		}

		fput_light(in, fput_in);
	}

	return error;
}