pipe.c 24.4 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
/*
 *  linux/fs/pipe.c
 *
 *  Copyright (C) 1991, 1992, 1999  Linus Torvalds
 */

#include <linux/mm.h>
#include <linux/file.h>
#include <linux/poll.h>
#include <linux/slab.h>
#include <linux/module.h>
#include <linux/init.h>
#include <linux/fs.h>
#include <linux/mount.h>
#include <linux/pipe_fs_i.h>
#include <linux/uio.h>
#include <linux/highmem.h>
18
#include <linux/pagemap.h>
A
Al Viro 已提交
19
#include <linux/audit.h>
L
Linus Torvalds 已提交
20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39

#include <asm/uaccess.h>
#include <asm/ioctls.h>

/*
 * We use a start+len construction, which provides full use of the 
 * allocated memory.
 * -- Florian Coosmann (FGC)
 * 
 * Reads with count = 0 should always return 0.
 * -- Julian Bradfield 1999-06-07.
 *
 * FIFOs and Pipes now generate SIGIO for both readers and writers.
 * -- Jeremy Elson <jelson@circlemud.org> 2001-08-16
 *
 * pipe_read & write cleanup
 * -- Manfred Spraul <manfred@colorfullife.com> 2002-05-09
 */

/* Drop the inode semaphore and wait for a pipe event, atomically */
40
void pipe_wait(struct pipe_inode_info *pipe)
L
Linus Torvalds 已提交
41 42 43
{
	DEFINE_WAIT(wait);

I
Ingo Molnar 已提交
44 45 46 47
	/*
	 * Pipes are system-local resources, so sleeping on them
	 * is considered a noninteractive wait:
	 */
48
	prepare_to_wait(&pipe->wait, &wait, TASK_INTERRUPTIBLE);
49 50
	if (pipe->inode)
		mutex_unlock(&pipe->inode->i_mutex);
L
Linus Torvalds 已提交
51
	schedule();
52 53 54
	finish_wait(&pipe->wait, &wait);
	if (pipe->inode)
		mutex_lock(&pipe->inode->i_mutex);
L
Linus Torvalds 已提交
55 56
}

57
static int
58 59
pipe_iov_copy_from_user(void *to, struct iovec *iov, unsigned long len,
			int atomic)
L
Linus Torvalds 已提交
60 61 62 63 64 65 66 67
{
	unsigned long copy;

	while (len > 0) {
		while (!iov->iov_len)
			iov++;
		copy = min_t(unsigned long, len, iov->iov_len);

68 69 70 71 72 73 74
		if (atomic) {
			if (__copy_from_user_inatomic(to, iov->iov_base, copy))
				return -EFAULT;
		} else {
			if (copy_from_user(to, iov->iov_base, copy))
				return -EFAULT;
		}
L
Linus Torvalds 已提交
75 76 77 78 79 80 81 82
		to += copy;
		len -= copy;
		iov->iov_base += copy;
		iov->iov_len -= copy;
	}
	return 0;
}

83
static int
84 85
pipe_iov_copy_to_user(struct iovec *iov, const void *from, unsigned long len,
		      int atomic)
L
Linus Torvalds 已提交
86 87 88 89 90 91 92 93
{
	unsigned long copy;

	while (len > 0) {
		while (!iov->iov_len)
			iov++;
		copy = min_t(unsigned long, len, iov->iov_len);

94 95 96 97 98 99 100
		if (atomic) {
			if (__copy_to_user_inatomic(iov->iov_base, from, copy))
				return -EFAULT;
		} else {
			if (copy_to_user(iov->iov_base, from, copy))
				return -EFAULT;
		}
L
Linus Torvalds 已提交
101 102 103 104 105 106 107 108
		from += copy;
		len -= copy;
		iov->iov_base += copy;
		iov->iov_len -= copy;
	}
	return 0;
}

109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149
/*
 * Attempt to pre-fault in the user memory, so we can use atomic copies.
 * Returns the number of bytes not faulted in.
 */
static int iov_fault_in_pages_write(struct iovec *iov, unsigned long len)
{
	while (!iov->iov_len)
		iov++;

	while (len > 0) {
		unsigned long this_len;

		this_len = min_t(unsigned long, len, iov->iov_len);
		if (fault_in_pages_writeable(iov->iov_base, this_len))
			break;

		len -= this_len;
		iov++;
	}

	return len;
}

/*
 * Pre-fault in the user memory, so we can use atomic copies.
 */
static void iov_fault_in_pages_read(struct iovec *iov, unsigned long len)
{
	while (!iov->iov_len)
		iov++;

	while (len > 0) {
		unsigned long this_len;

		this_len = min_t(unsigned long, len, iov->iov_len);
		fault_in_pages_readable(iov->iov_base, this_len);
		len -= this_len;
		iov++;
	}
}

150 151
static void anon_pipe_buf_release(struct pipe_inode_info *pipe,
				  struct pipe_buffer *buf)
L
Linus Torvalds 已提交
152 153 154
{
	struct page *page = buf->page;

155 156 157
	/*
	 * If nobody else uses this page, and we don't already have a
	 * temporary page, let's keep track of it as a one-deep
158
	 * allocation cache. (Otherwise just release our reference to it)
159
	 */
160
	if (page_count(page) == 1 && !pipe->tmp_page)
161
		pipe->tmp_page = page;
162 163
	else
		page_cache_release(page);
L
Linus Torvalds 已提交
164 165
}

166 167 168 169 170 171 172 173 174 175 176 177 178 179
/**
 * generic_pipe_buf_map - virtually map a pipe buffer
 * @pipe:	the pipe that the buffer belongs to
 * @buf:	the buffer that should be mapped
 * @atomic:	whether to use an atomic map
 *
 * Description:
 *	This function returns a kernel virtual address mapping for the
 *	passed in @pipe_buffer. If @atomic is set, an atomic map is provided
 *	and the caller has to be careful not to fault before calling
 *	the unmap function.
 *
 *	Note that this function occupies KM_USER0 if @atomic != 0.
 */
180
void *generic_pipe_buf_map(struct pipe_inode_info *pipe,
181
			   struct pipe_buffer *buf, int atomic)
L
Linus Torvalds 已提交
182
{
183 184 185 186 187
	if (atomic) {
		buf->flags |= PIPE_BUF_FLAG_ATOMIC;
		return kmap_atomic(buf->page, KM_USER0);
	}

L
Linus Torvalds 已提交
188 189 190
	return kmap(buf->page);
}

191 192 193 194 195 196 197 198 199
/**
 * generic_pipe_buf_unmap - unmap a previously mapped pipe buffer
 * @pipe:	the pipe that the buffer belongs to
 * @buf:	the buffer that should be unmapped
 * @map_data:	the data that the mapping function returned
 *
 * Description:
 *	This function undoes the mapping that ->map() provided.
 */
200
void generic_pipe_buf_unmap(struct pipe_inode_info *pipe,
201
			    struct pipe_buffer *buf, void *map_data)
L
Linus Torvalds 已提交
202
{
203 204 205 206 207
	if (buf->flags & PIPE_BUF_FLAG_ATOMIC) {
		buf->flags &= ~PIPE_BUF_FLAG_ATOMIC;
		kunmap_atomic(map_data, KM_USER0);
	} else
		kunmap(buf->page);
L
Linus Torvalds 已提交
208 209
}

210 211 212 213 214 215 216 217 218 219 220 221
/**
 * generic_pipe_buf_steal - attempt to take ownership of a @pipe_buffer
 * @pipe:	the pipe that the buffer belongs to
 * @buf:	the buffer to attempt to steal
 *
 * Description:
 *	This function attempts to steal the @struct page attached to
 *	@buf. If successful, this function returns 0 and returns with
 *	the page locked. The caller may then reuse the page for whatever
 *	he wishes, the typical use is insertion into a different file
 *	page cache.
 */
222 223
int generic_pipe_buf_steal(struct pipe_inode_info *pipe,
			   struct pipe_buffer *buf)
224
{
225 226
	struct page *page = buf->page;

227 228 229 230 231
	/*
	 * A reference of one is golden, that means that the owner of this
	 * page is the only one holding a reference to it. lock the page
	 * and return OK.
	 */
232 233 234 235 236 237
	if (page_count(page) == 1) {
		lock_page(page);
		return 0;
	}

	return 1;
238 239
}

240 241 242 243 244 245 246 247 248 249 250
/**
 * generic_pipe_buf_get - get a reference to a @struct pipe_buffer
 * @pipe:	the pipe that the buffer belongs to
 * @buf:	the buffer to get a reference to
 *
 * Description:
 *	This function grabs an extra reference to @buf. It's used in
 *	in the tee() system call, when we duplicate the buffers in one
 *	pipe into another.
 */
void generic_pipe_buf_get(struct pipe_inode_info *pipe, struct pipe_buffer *buf)
251 252 253 254
{
	page_cache_get(buf->page);
}

255 256
/**
 * generic_pipe_buf_confirm - verify contents of the pipe buffer
R
Randy Dunlap 已提交
257
 * @info:	the pipe that the buffer belongs to
258 259 260 261 262 263
 * @buf:	the buffer to confirm
 *
 * Description:
 *	This function does nothing, because the generic pipe code uses
 *	pages that are always good when inserted into the pipe.
 */
264 265
int generic_pipe_buf_confirm(struct pipe_inode_info *info,
			     struct pipe_buffer *buf)
266 267 268 269
{
	return 0;
}

270
static const struct pipe_buf_operations anon_pipe_buf_ops = {
L
Linus Torvalds 已提交
271
	.can_merge = 1,
272 273
	.map = generic_pipe_buf_map,
	.unmap = generic_pipe_buf_unmap,
274
	.confirm = generic_pipe_buf_confirm,
L
Linus Torvalds 已提交
275
	.release = anon_pipe_buf_release,
276
	.steal = generic_pipe_buf_steal,
277
	.get = generic_pipe_buf_get,
L
Linus Torvalds 已提交
278 279 280
};

static ssize_t
281 282
pipe_read(struct kiocb *iocb, const struct iovec *_iov,
	   unsigned long nr_segs, loff_t pos)
L
Linus Torvalds 已提交
283
{
284
	struct file *filp = iocb->ki_filp;
285
	struct inode *inode = filp->f_path.dentry->d_inode;
286
	struct pipe_inode_info *pipe;
L
Linus Torvalds 已提交
287 288 289 290 291 292 293 294 295 296 297 298
	int do_wakeup;
	ssize_t ret;
	struct iovec *iov = (struct iovec *)_iov;
	size_t total_len;

	total_len = iov_length(iov, nr_segs);
	/* Null read succeeds. */
	if (unlikely(total_len == 0))
		return 0;

	do_wakeup = 0;
	ret = 0;
299
	mutex_lock(&inode->i_mutex);
300
	pipe = inode->i_pipe;
L
Linus Torvalds 已提交
301
	for (;;) {
302
		int bufs = pipe->nrbufs;
L
Linus Torvalds 已提交
303
		if (bufs) {
304 305
			int curbuf = pipe->curbuf;
			struct pipe_buffer *buf = pipe->bufs + curbuf;
306
			const struct pipe_buf_operations *ops = buf->ops;
L
Linus Torvalds 已提交
307 308
			void *addr;
			size_t chars = buf->len;
309
			int error, atomic;
L
Linus Torvalds 已提交
310 311 312 313

			if (chars > total_len)
				chars = total_len;

314
			error = ops->confirm(pipe, buf);
315
			if (error) {
316
				if (!ret)
317
					error = ret;
318 319
				break;
			}
320

321 322 323 324 325
			atomic = !iov_fault_in_pages_write(iov, chars);
redo:
			addr = ops->map(pipe, buf, atomic);
			error = pipe_iov_copy_to_user(iov, addr + buf->offset, chars, atomic);
			ops->unmap(pipe, buf, addr);
L
Linus Torvalds 已提交
326
			if (unlikely(error)) {
327 328 329 330 331 332 333
				/*
				 * Just retry with the slow path if we failed.
				 */
				if (atomic) {
					atomic = 0;
					goto redo;
				}
334
				if (!ret)
335
					ret = error;
L
Linus Torvalds 已提交
336 337 338 339 340 341 342
				break;
			}
			ret += chars;
			buf->offset += chars;
			buf->len -= chars;
			if (!buf->len) {
				buf->ops = NULL;
343
				ops->release(pipe, buf);
L
Linus Torvalds 已提交
344
				curbuf = (curbuf + 1) & (PIPE_BUFFERS-1);
345 346
				pipe->curbuf = curbuf;
				pipe->nrbufs = --bufs;
L
Linus Torvalds 已提交
347 348 349 350 351 352 353 354
				do_wakeup = 1;
			}
			total_len -= chars;
			if (!total_len)
				break;	/* common path: read succeeded */
		}
		if (bufs)	/* More to do? */
			continue;
355
		if (!pipe->writers)
L
Linus Torvalds 已提交
356
			break;
357
		if (!pipe->waiting_writers) {
L
Linus Torvalds 已提交
358 359 360 361 362 363 364 365 366 367 368 369 370
			/* syscall merging: Usually we must not sleep
			 * if O_NONBLOCK is set, or if we got some data.
			 * But if a writer sleeps in kernel space, then
			 * we can wait for that data without violating POSIX.
			 */
			if (ret)
				break;
			if (filp->f_flags & O_NONBLOCK) {
				ret = -EAGAIN;
				break;
			}
		}
		if (signal_pending(current)) {
371 372
			if (!ret)
				ret = -ERESTARTSYS;
L
Linus Torvalds 已提交
373 374 375
			break;
		}
		if (do_wakeup) {
376 377
			wake_up_interruptible_sync(&pipe->wait);
 			kill_fasync(&pipe->fasync_writers, SIGIO, POLL_OUT);
L
Linus Torvalds 已提交
378
		}
379
		pipe_wait(pipe);
L
Linus Torvalds 已提交
380
	}
381
	mutex_unlock(&inode->i_mutex);
382 383

	/* Signal writers asynchronously that there is more room. */
L
Linus Torvalds 已提交
384
	if (do_wakeup) {
385 386
		wake_up_interruptible(&pipe->wait);
		kill_fasync(&pipe->fasync_writers, SIGIO, POLL_OUT);
L
Linus Torvalds 已提交
387 388 389 390 391 392 393
	}
	if (ret > 0)
		file_accessed(filp);
	return ret;
}

static ssize_t
394 395
pipe_write(struct kiocb *iocb, const struct iovec *_iov,
	    unsigned long nr_segs, loff_t ppos)
L
Linus Torvalds 已提交
396
{
397
	struct file *filp = iocb->ki_filp;
398
	struct inode *inode = filp->f_path.dentry->d_inode;
399
	struct pipe_inode_info *pipe;
L
Linus Torvalds 已提交
400 401 402 403 404 405 406 407 408 409 410 411 412
	ssize_t ret;
	int do_wakeup;
	struct iovec *iov = (struct iovec *)_iov;
	size_t total_len;
	ssize_t chars;

	total_len = iov_length(iov, nr_segs);
	/* Null write succeeds. */
	if (unlikely(total_len == 0))
		return 0;

	do_wakeup = 0;
	ret = 0;
413
	mutex_lock(&inode->i_mutex);
414
	pipe = inode->i_pipe;
L
Linus Torvalds 已提交
415

416
	if (!pipe->readers) {
L
Linus Torvalds 已提交
417 418 419 420 421 422 423
		send_sig(SIGPIPE, current, 0);
		ret = -EPIPE;
		goto out;
	}

	/* We try to merge small writes */
	chars = total_len & (PAGE_SIZE-1); /* size of the last buffer */
424
	if (pipe->nrbufs && chars != 0) {
425 426
		int lastbuf = (pipe->curbuf + pipe->nrbufs - 1) &
							(PIPE_BUFFERS-1);
427
		struct pipe_buffer *buf = pipe->bufs + lastbuf;
428
		const struct pipe_buf_operations *ops = buf->ops;
L
Linus Torvalds 已提交
429
		int offset = buf->offset + buf->len;
430

L
Linus Torvalds 已提交
431
		if (ops->can_merge && offset + chars <= PAGE_SIZE) {
432
			int error, atomic = 1;
433 434
			void *addr;

435
			error = ops->confirm(pipe, buf);
436
			if (error)
437
				goto out;
438

439 440 441
			iov_fault_in_pages_read(iov, chars);
redo1:
			addr = ops->map(pipe, buf, atomic);
442
			error = pipe_iov_copy_from_user(offset + addr, iov,
443 444
							chars, atomic);
			ops->unmap(pipe, buf, addr);
L
Linus Torvalds 已提交
445 446
			ret = error;
			do_wakeup = 1;
447 448 449 450 451
			if (error) {
				if (atomic) {
					atomic = 0;
					goto redo1;
				}
L
Linus Torvalds 已提交
452
				goto out;
453
			}
L
Linus Torvalds 已提交
454 455 456 457 458 459 460 461 462 463
			buf->len += chars;
			total_len -= chars;
			ret = chars;
			if (!total_len)
				goto out;
		}
	}

	for (;;) {
		int bufs;
464

465
		if (!pipe->readers) {
L
Linus Torvalds 已提交
466
			send_sig(SIGPIPE, current, 0);
467 468
			if (!ret)
				ret = -EPIPE;
L
Linus Torvalds 已提交
469 470
			break;
		}
471
		bufs = pipe->nrbufs;
L
Linus Torvalds 已提交
472
		if (bufs < PIPE_BUFFERS) {
473 474 475
			int newbuf = (pipe->curbuf + bufs) & (PIPE_BUFFERS-1);
			struct pipe_buffer *buf = pipe->bufs + newbuf;
			struct page *page = pipe->tmp_page;
476 477
			char *src;
			int error, atomic = 1;
L
Linus Torvalds 已提交
478 479 480 481 482 483 484

			if (!page) {
				page = alloc_page(GFP_HIGHUSER);
				if (unlikely(!page)) {
					ret = ret ? : -ENOMEM;
					break;
				}
485
				pipe->tmp_page = page;
L
Linus Torvalds 已提交
486
			}
487
			/* Always wake up, even if the copy fails. Otherwise
L
Linus Torvalds 已提交
488 489 490 491 492 493 494 495 496
			 * we lock up (O_NONBLOCK-)readers that sleep due to
			 * syscall merging.
			 * FIXME! Is this really true?
			 */
			do_wakeup = 1;
			chars = PAGE_SIZE;
			if (chars > total_len)
				chars = total_len;

497 498 499 500 501 502 503 504 505 506 507 508 509 510
			iov_fault_in_pages_read(iov, chars);
redo2:
			if (atomic)
				src = kmap_atomic(page, KM_USER0);
			else
				src = kmap(page);

			error = pipe_iov_copy_from_user(src, iov, chars,
							atomic);
			if (atomic)
				kunmap_atomic(src, KM_USER0);
			else
				kunmap(page);

L
Linus Torvalds 已提交
511
			if (unlikely(error)) {
512 513 514 515
				if (atomic) {
					atomic = 0;
					goto redo2;
				}
516
				if (!ret)
517
					ret = error;
L
Linus Torvalds 已提交
518 519 520 521 522 523 524 525 526
				break;
			}
			ret += chars;

			/* Insert it into the buffer array */
			buf->page = page;
			buf->ops = &anon_pipe_buf_ops;
			buf->offset = 0;
			buf->len = chars;
527 528
			pipe->nrbufs = ++bufs;
			pipe->tmp_page = NULL;
L
Linus Torvalds 已提交
529 530 531 532 533 534 535 536

			total_len -= chars;
			if (!total_len)
				break;
		}
		if (bufs < PIPE_BUFFERS)
			continue;
		if (filp->f_flags & O_NONBLOCK) {
537 538
			if (!ret)
				ret = -EAGAIN;
L
Linus Torvalds 已提交
539 540 541
			break;
		}
		if (signal_pending(current)) {
542 543
			if (!ret)
				ret = -ERESTARTSYS;
L
Linus Torvalds 已提交
544 545 546
			break;
		}
		if (do_wakeup) {
547 548
			wake_up_interruptible_sync(&pipe->wait);
			kill_fasync(&pipe->fasync_readers, SIGIO, POLL_IN);
L
Linus Torvalds 已提交
549 550
			do_wakeup = 0;
		}
551 552 553
		pipe->waiting_writers++;
		pipe_wait(pipe);
		pipe->waiting_writers--;
L
Linus Torvalds 已提交
554 555
	}
out:
556
	mutex_unlock(&inode->i_mutex);
L
Linus Torvalds 已提交
557
	if (do_wakeup) {
558 559
		wake_up_interruptible(&pipe->wait);
		kill_fasync(&pipe->fasync_readers, SIGIO, POLL_IN);
L
Linus Torvalds 已提交
560 561
	}
	if (ret > 0)
562
		file_update_time(filp);
L
Linus Torvalds 已提交
563 564 565 566 567 568 569 570 571 572
	return ret;
}

static ssize_t
bad_pipe_r(struct file *filp, char __user *buf, size_t count, loff_t *ppos)
{
	return -EBADF;
}

static ssize_t
573 574
bad_pipe_w(struct file *filp, const char __user *buf, size_t count,
	   loff_t *ppos)
L
Linus Torvalds 已提交
575 576 577 578 579 580 581 582
{
	return -EBADF;
}

static int
pipe_ioctl(struct inode *pino, struct file *filp,
	   unsigned int cmd, unsigned long arg)
{
583
	struct inode *inode = filp->f_path.dentry->d_inode;
584
	struct pipe_inode_info *pipe;
L
Linus Torvalds 已提交
585 586 587 588
	int count, buf, nrbufs;

	switch (cmd) {
		case FIONREAD:
589
			mutex_lock(&inode->i_mutex);
590
			pipe = inode->i_pipe;
L
Linus Torvalds 已提交
591
			count = 0;
592 593
			buf = pipe->curbuf;
			nrbufs = pipe->nrbufs;
L
Linus Torvalds 已提交
594
			while (--nrbufs >= 0) {
595
				count += pipe->bufs[buf].len;
L
Linus Torvalds 已提交
596 597
				buf = (buf+1) & (PIPE_BUFFERS-1);
			}
598
			mutex_unlock(&inode->i_mutex);
599

L
Linus Torvalds 已提交
600 601 602 603 604 605 606 607 608 609 610
			return put_user(count, (int __user *)arg);
		default:
			return -EINVAL;
	}
}

/* No kernel lock held - fine */
static unsigned int
pipe_poll(struct file *filp, poll_table *wait)
{
	unsigned int mask;
611
	struct inode *inode = filp->f_path.dentry->d_inode;
612
	struct pipe_inode_info *pipe = inode->i_pipe;
L
Linus Torvalds 已提交
613 614
	int nrbufs;

615
	poll_wait(filp, &pipe->wait, wait);
L
Linus Torvalds 已提交
616 617

	/* Reading only -- no need for acquiring the semaphore.  */
618
	nrbufs = pipe->nrbufs;
L
Linus Torvalds 已提交
619 620 621
	mask = 0;
	if (filp->f_mode & FMODE_READ) {
		mask = (nrbufs > 0) ? POLLIN | POLLRDNORM : 0;
622
		if (!pipe->writers && filp->f_version != pipe->w_counter)
L
Linus Torvalds 已提交
623 624 625 626 627
			mask |= POLLHUP;
	}

	if (filp->f_mode & FMODE_WRITE) {
		mask |= (nrbufs < PIPE_BUFFERS) ? POLLOUT | POLLWRNORM : 0;
628 629 630 631
		/*
		 * Most Unices do not set POLLERR for FIFOs but on Linux they
		 * behave exactly like pipes for poll().
		 */
632
		if (!pipe->readers)
L
Linus Torvalds 已提交
633 634 635 636 637 638 639 640 641
			mask |= POLLERR;
	}

	return mask;
}

static int
pipe_release(struct inode *inode, int decr, int decw)
{
642 643
	struct pipe_inode_info *pipe;

644
	mutex_lock(&inode->i_mutex);
645 646 647
	pipe = inode->i_pipe;
	pipe->readers -= decr;
	pipe->writers -= decw;
648

649
	if (!pipe->readers && !pipe->writers) {
L
Linus Torvalds 已提交
650 651
		free_pipe_info(inode);
	} else {
652 653 654
		wake_up_interruptible(&pipe->wait);
		kill_fasync(&pipe->fasync_readers, SIGIO, POLL_IN);
		kill_fasync(&pipe->fasync_writers, SIGIO, POLL_OUT);
L
Linus Torvalds 已提交
655
	}
656
	mutex_unlock(&inode->i_mutex);
L
Linus Torvalds 已提交
657 658 659 660 661 662 663

	return 0;
}

static int
pipe_read_fasync(int fd, struct file *filp, int on)
{
664
	struct inode *inode = filp->f_path.dentry->d_inode;
L
Linus Torvalds 已提交
665 666
	int retval;

667 668 669
	mutex_lock(&inode->i_mutex);
	retval = fasync_helper(fd, filp, on, &inode->i_pipe->fasync_readers);
	mutex_unlock(&inode->i_mutex);
L
Linus Torvalds 已提交
670 671 672 673 674 675 676 677 678 679 680

	if (retval < 0)
		return retval;

	return 0;
}


static int
pipe_write_fasync(int fd, struct file *filp, int on)
{
681
	struct inode *inode = filp->f_path.dentry->d_inode;
L
Linus Torvalds 已提交
682 683
	int retval;

684 685 686
	mutex_lock(&inode->i_mutex);
	retval = fasync_helper(fd, filp, on, &inode->i_pipe->fasync_writers);
	mutex_unlock(&inode->i_mutex);
L
Linus Torvalds 已提交
687 688 689 690 691 692 693 694 695 696 697

	if (retval < 0)
		return retval;

	return 0;
}


static int
pipe_rdwr_fasync(int fd, struct file *filp, int on)
{
698
	struct inode *inode = filp->f_path.dentry->d_inode;
699
	struct pipe_inode_info *pipe = inode->i_pipe;
L
Linus Torvalds 已提交
700 701
	int retval;

702
	mutex_lock(&inode->i_mutex);
L
Linus Torvalds 已提交
703

704
	retval = fasync_helper(fd, filp, on, &pipe->fasync_readers);
L
Linus Torvalds 已提交
705 706

	if (retval >= 0)
707
		retval = fasync_helper(fd, filp, on, &pipe->fasync_writers);
L
Linus Torvalds 已提交
708

709
	mutex_unlock(&inode->i_mutex);
L
Linus Torvalds 已提交
710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747

	if (retval < 0)
		return retval;

	return 0;
}


static int
pipe_read_release(struct inode *inode, struct file *filp)
{
	pipe_read_fasync(-1, filp, 0);
	return pipe_release(inode, 1, 0);
}

static int
pipe_write_release(struct inode *inode, struct file *filp)
{
	pipe_write_fasync(-1, filp, 0);
	return pipe_release(inode, 0, 1);
}

static int
pipe_rdwr_release(struct inode *inode, struct file *filp)
{
	int decr, decw;

	pipe_rdwr_fasync(-1, filp, 0);
	decr = (filp->f_mode & FMODE_READ) != 0;
	decw = (filp->f_mode & FMODE_WRITE) != 0;
	return pipe_release(inode, decr, decw);
}

static int
pipe_read_open(struct inode *inode, struct file *filp)
{
	/* We could have perhaps used atomic_t, but this and friends
	   below are the only places.  So it doesn't seem worthwhile.  */
748 749 750
	mutex_lock(&inode->i_mutex);
	inode->i_pipe->readers++;
	mutex_unlock(&inode->i_mutex);
L
Linus Torvalds 已提交
751 752 753 754 755 756 757

	return 0;
}

static int
pipe_write_open(struct inode *inode, struct file *filp)
{
758 759 760
	mutex_lock(&inode->i_mutex);
	inode->i_pipe->writers++;
	mutex_unlock(&inode->i_mutex);
L
Linus Torvalds 已提交
761 762 763 764 765 766 767

	return 0;
}

static int
pipe_rdwr_open(struct inode *inode, struct file *filp)
{
768
	mutex_lock(&inode->i_mutex);
L
Linus Torvalds 已提交
769
	if (filp->f_mode & FMODE_READ)
770
		inode->i_pipe->readers++;
L
Linus Torvalds 已提交
771
	if (filp->f_mode & FMODE_WRITE)
772 773
		inode->i_pipe->writers++;
	mutex_unlock(&inode->i_mutex);
L
Linus Torvalds 已提交
774 775 776 777 778 779 780 781

	return 0;
}

/*
 * The file_operations structs are not static because they
 * are also used in linux/fs/fifo.c to do operations on FIFOs.
 */
782
const struct file_operations read_fifo_fops = {
L
Linus Torvalds 已提交
783
	.llseek		= no_llseek,
784 785
	.read		= do_sync_read,
	.aio_read	= pipe_read,
L
Linus Torvalds 已提交
786
	.write		= bad_pipe_w,
787
	.poll		= pipe_poll,
L
Linus Torvalds 已提交
788 789 790 791 792 793
	.ioctl		= pipe_ioctl,
	.open		= pipe_read_open,
	.release	= pipe_read_release,
	.fasync		= pipe_read_fasync,
};

794
const struct file_operations write_fifo_fops = {
L
Linus Torvalds 已提交
795 796
	.llseek		= no_llseek,
	.read		= bad_pipe_r,
797 798
	.write		= do_sync_write,
	.aio_write	= pipe_write,
799
	.poll		= pipe_poll,
L
Linus Torvalds 已提交
800 801 802 803 804 805
	.ioctl		= pipe_ioctl,
	.open		= pipe_write_open,
	.release	= pipe_write_release,
	.fasync		= pipe_write_fasync,
};

806
const struct file_operations rdwr_fifo_fops = {
L
Linus Torvalds 已提交
807
	.llseek		= no_llseek,
808 809 810 811
	.read		= do_sync_read,
	.aio_read	= pipe_read,
	.write		= do_sync_write,
	.aio_write	= pipe_write,
812
	.poll		= pipe_poll,
L
Linus Torvalds 已提交
813 814 815 816 817 818
	.ioctl		= pipe_ioctl,
	.open		= pipe_rdwr_open,
	.release	= pipe_rdwr_release,
	.fasync		= pipe_rdwr_fasync,
};

819
static const struct file_operations read_pipe_fops = {
L
Linus Torvalds 已提交
820
	.llseek		= no_llseek,
821 822
	.read		= do_sync_read,
	.aio_read	= pipe_read,
L
Linus Torvalds 已提交
823 824 825 826 827 828 829 830
	.write		= bad_pipe_w,
	.poll		= pipe_poll,
	.ioctl		= pipe_ioctl,
	.open		= pipe_read_open,
	.release	= pipe_read_release,
	.fasync		= pipe_read_fasync,
};

831
static const struct file_operations write_pipe_fops = {
L
Linus Torvalds 已提交
832 833
	.llseek		= no_llseek,
	.read		= bad_pipe_r,
834 835
	.write		= do_sync_write,
	.aio_write	= pipe_write,
L
Linus Torvalds 已提交
836 837 838 839 840 841 842
	.poll		= pipe_poll,
	.ioctl		= pipe_ioctl,
	.open		= pipe_write_open,
	.release	= pipe_write_release,
	.fasync		= pipe_write_fasync,
};

843
static const struct file_operations rdwr_pipe_fops = {
L
Linus Torvalds 已提交
844
	.llseek		= no_llseek,
845 846 847 848
	.read		= do_sync_read,
	.aio_read	= pipe_read,
	.write		= do_sync_write,
	.aio_write	= pipe_write,
L
Linus Torvalds 已提交
849 850 851 852 853 854 855
	.poll		= pipe_poll,
	.ioctl		= pipe_ioctl,
	.open		= pipe_rdwr_open,
	.release	= pipe_rdwr_release,
	.fasync		= pipe_rdwr_fasync,
};

856 857
struct pipe_inode_info * alloc_pipe_info(struct inode *inode)
{
858
	struct pipe_inode_info *pipe;
859

860 861 862 863 864
	pipe = kzalloc(sizeof(struct pipe_inode_info), GFP_KERNEL);
	if (pipe) {
		init_waitqueue_head(&pipe->wait);
		pipe->r_counter = pipe->w_counter = 1;
		pipe->inode = inode;
865 866
	}

867
	return pipe;
868 869
}

870
void __free_pipe_info(struct pipe_inode_info *pipe)
L
Linus Torvalds 已提交
871 872 873 874
{
	int i;

	for (i = 0; i < PIPE_BUFFERS; i++) {
875
		struct pipe_buffer *buf = pipe->bufs + i;
L
Linus Torvalds 已提交
876
		if (buf->ops)
877
			buf->ops->release(pipe, buf);
L
Linus Torvalds 已提交
878
	}
879 880 881
	if (pipe->tmp_page)
		__free_page(pipe->tmp_page);
	kfree(pipe);
L
Linus Torvalds 已提交
882 883
}

884 885 886 887 888 889
void free_pipe_info(struct inode *inode)
{
	__free_pipe_info(inode->i_pipe);
	inode->i_pipe = NULL;
}

890
static struct vfsmount *pipe_mnt __read_mostly;
L
Linus Torvalds 已提交
891 892
static int pipefs_delete_dentry(struct dentry *dentry)
{
893 894 895 896 897 898 899 900
	/*
	 * At creation time, we pretended this dentry was hashed
	 * (by clearing DCACHE_UNHASHED bit in d_flags)
	 * At delete time, we restore the truth : not hashed.
	 * (so that dput() can proceed correctly)
	 */
	dentry->d_flags |= DCACHE_UNHASHED;
	return 0;
L
Linus Torvalds 已提交
901
}
902

903 904 905 906 907 908 909 910 911
/*
 * pipefs_dname() is called from d_path().
 */
static char *pipefs_dname(struct dentry *dentry, char *buffer, int buflen)
{
	return dynamic_dname(dentry, buffer, buflen, "pipe:[%lu]",
				dentry->d_inode->i_ino);
}

L
Linus Torvalds 已提交
912 913
static struct dentry_operations pipefs_dentry_operations = {
	.d_delete	= pipefs_delete_dentry,
914
	.d_dname	= pipefs_dname,
L
Linus Torvalds 已提交
915 916 917 918 919
};

static struct inode * get_pipe_inode(void)
{
	struct inode *inode = new_inode(pipe_mnt->mnt_sb);
920
	struct pipe_inode_info *pipe;
L
Linus Torvalds 已提交
921 922 923 924

	if (!inode)
		goto fail_inode;

925 926
	pipe = alloc_pipe_info(inode);
	if (!pipe)
L
Linus Torvalds 已提交
927
		goto fail_iput;
928
	inode->i_pipe = pipe;
929

930
	pipe->readers = pipe->writers = 1;
L
Linus Torvalds 已提交
931 932 933 934 935 936 937 938 939 940 941 942 943
	inode->i_fop = &rdwr_pipe_fops;

	/*
	 * Mark the inode dirty from the very beginning,
	 * that way it will never be moved to the dirty
	 * list because "mark_inode_dirty()" will think
	 * that it already _is_ on the dirty list.
	 */
	inode->i_state = I_DIRTY;
	inode->i_mode = S_IFIFO | S_IRUSR | S_IWUSR;
	inode->i_uid = current->fsuid;
	inode->i_gid = current->fsgid;
	inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
944

L
Linus Torvalds 已提交
945 946 947 948
	return inode;

fail_iput:
	iput(inode);
949

L
Linus Torvalds 已提交
950 951 952 953
fail_inode:
	return NULL;
}

A
Andi Kleen 已提交
954
struct file *create_write_pipe(void)
L
Linus Torvalds 已提交
955
{
A
Andi Kleen 已提交
956 957 958
	int err;
	struct inode *inode;
	struct file *f;
L
Linus Torvalds 已提交
959
	struct dentry *dentry;
960
	struct qstr name = { .name = "" };
L
Linus Torvalds 已提交
961

A
Andi Kleen 已提交
962 963 964 965
	f = get_empty_filp();
	if (!f)
		return ERR_PTR(-ENFILE);
	err = -ENFILE;
L
Linus Torvalds 已提交
966 967
	inode = get_pipe_inode();
	if (!inode)
A
Andi Kleen 已提交
968
		goto err_file;
L
Linus Torvalds 已提交
969

A
Andi Kleen 已提交
970
	err = -ENOMEM;
971
	dentry = d_alloc(pipe_mnt->mnt_sb->s_root, &name);
L
Linus Torvalds 已提交
972
	if (!dentry)
A
Andi Kleen 已提交
973
		goto err_inode;
974

L
Linus Torvalds 已提交
975
	dentry->d_op = &pipefs_dentry_operations;
976 977 978 979 980 981 982
	/*
	 * We dont want to publish this dentry into global dentry hash table.
	 * We pretend dentry is already hashed, by unsetting DCACHE_UNHASHED
	 * This permits a working /proc/$pid/fd/XXX on pipes
	 */
	dentry->d_flags &= ~DCACHE_UNHASHED;
	d_instantiate(dentry, inode);
983 984
	f->f_path.mnt = mntget(pipe_mnt);
	f->f_path.dentry = dentry;
A
Andi Kleen 已提交
985
	f->f_mapping = inode->i_mapping;
986

A
Andi Kleen 已提交
987 988 989 990 991 992
	f->f_flags = O_WRONLY;
	f->f_op = &write_pipe_fops;
	f->f_mode = FMODE_WRITE;
	f->f_version = 0;

	return f;
L
Linus Torvalds 已提交
993

A
Andi Kleen 已提交
994
 err_inode:
L
Linus Torvalds 已提交
995 996
	free_pipe_info(inode);
	iput(inode);
A
Andi Kleen 已提交
997 998 999 1000 1001 1002 1003
 err_file:
	put_filp(f);
	return ERR_PTR(err);
}

void free_write_pipe(struct file *f)
{
1004
	free_pipe_info(f->f_dentry->d_inode);
1005
	dput(f->f_path.dentry);
1006
	mntput(f->f_path.mnt);
A
Andi Kleen 已提交
1007 1008 1009 1010 1011 1012 1013 1014 1015 1016
	put_filp(f);
}

struct file *create_read_pipe(struct file *wrf)
{
	struct file *f = get_empty_filp();
	if (!f)
		return ERR_PTR(-ENFILE);

	/* Grab pipe from the writer */
1017 1018 1019
	f->f_path.mnt = mntget(wrf->f_path.mnt);
	f->f_path.dentry = dget(wrf->f_path.dentry);
	f->f_mapping = wrf->f_path.dentry->d_inode->i_mapping;
A
Andi Kleen 已提交
1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053

	f->f_pos = 0;
	f->f_flags = O_RDONLY;
	f->f_op = &read_pipe_fops;
	f->f_mode = FMODE_READ;
	f->f_version = 0;

	return f;
}

int do_pipe(int *fd)
{
	struct file *fw, *fr;
	int error;
	int fdw, fdr;

	fw = create_write_pipe();
	if (IS_ERR(fw))
		return PTR_ERR(fw);
	fr = create_read_pipe(fw);
	error = PTR_ERR(fr);
	if (IS_ERR(fr))
		goto err_write_pipe;

	error = get_unused_fd();
	if (error < 0)
		goto err_read_pipe;
	fdr = error;

	error = get_unused_fd();
	if (error < 0)
		goto err_fdr;
	fdw = error;

A
Al Viro 已提交
1054 1055 1056 1057
	error = audit_fd_pair(fdr, fdw);
	if (error < 0)
		goto err_fdw;

A
Andi Kleen 已提交
1058 1059 1060 1061 1062 1063 1064
	fd_install(fdr, fr);
	fd_install(fdw, fw);
	fd[0] = fdr;
	fd[1] = fdw;

	return 0;

A
Al Viro 已提交
1065 1066
 err_fdw:
	put_unused_fd(fdw);
A
Andi Kleen 已提交
1067 1068 1069
 err_fdr:
	put_unused_fd(fdr);
 err_read_pipe:
1070 1071
	dput(fr->f_dentry);
	mntput(fr->f_vfsmnt);
A
Andi Kleen 已提交
1072 1073 1074 1075
	put_filp(fr);
 err_write_pipe:
	free_write_pipe(fw);
	return error;
L
Linus Torvalds 已提交
1076 1077 1078 1079 1080 1081 1082 1083
}

/*
 * pipefs should _never_ be mounted by userland - too much of security hassle,
 * no real gain from having the whole whorehouse mounted. So we don't need
 * any operations on the root directory. However, we need a non-trivial
 * d_name - pipe: will go nicely and kill the special-casing in procfs.
 */
1084 1085 1086
static int pipefs_get_sb(struct file_system_type *fs_type,
			 int flags, const char *dev_name, void *data,
			 struct vfsmount *mnt)
L
Linus Torvalds 已提交
1087
{
1088
	return get_sb_pseudo(fs_type, "pipe:", NULL, PIPEFS_MAGIC, mnt);
L
Linus Torvalds 已提交
1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099
}

static struct file_system_type pipe_fs_type = {
	.name		= "pipefs",
	.get_sb		= pipefs_get_sb,
	.kill_sb	= kill_anon_super,
};

static int __init init_pipe_fs(void)
{
	int err = register_filesystem(&pipe_fs_type);
1100

L
Linus Torvalds 已提交
1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118
	if (!err) {
		pipe_mnt = kern_mount(&pipe_fs_type);
		if (IS_ERR(pipe_mnt)) {
			err = PTR_ERR(pipe_mnt);
			unregister_filesystem(&pipe_fs_type);
		}
	}
	return err;
}

static void __exit exit_pipe_fs(void)
{
	unregister_filesystem(&pipe_fs_type);
	mntput(pipe_mnt);
}

fs_initcall(init_pipe_fs);
module_exit(exit_pipe_fs);