xref: /netbsd-src/sys/kern/sys_pipe.c (revision a10c2cec390bbc4a870ed981c04dc105e6820ccf)
1 /*	$NetBSD: sys_pipe.c,v 1.167 2024/02/10 09:21:54 andvar Exp $	*/
2 
3 /*-
4  * Copyright (c) 2003, 2007, 2008, 2009, 2023 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Paul Kranenburg, and by Andrew Doran.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
20  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
21  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
22  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
23  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
24  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
25  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
26  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
27  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
28  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
29  * POSSIBILITY OF SUCH DAMAGE.
30  */
31 
32 /*
33  * Copyright (c) 1996 John S. Dyson
34  * All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice immediately at the beginning of the file, without modification,
41  *    this list of conditions, and the following disclaimer.
42  * 2. Redistributions in binary form must reproduce the above copyright
43  *    notice, this list of conditions and the following disclaimer in the
44  *    documentation and/or other materials provided with the distribution.
45  * 3. Absolutely no warranty of function or purpose is made by the author
46  *    John S. Dyson.
47  * 4. Modifications may be freely made to this file if the above conditions
48  *    are met.
49  */
50 
51 /*
52  * This file contains a high-performance replacement for the socket-based
53  * pipes scheme originally used.  It does not support all features of
54  * sockets, but does do everything that pipes normally do.
55  */
56 
57 #include <sys/cdefs.h>
58 __KERNEL_RCSID(0, "$NetBSD: sys_pipe.c,v 1.167 2024/02/10 09:21:54 andvar Exp $");
59 
60 #include <sys/param.h>
61 #include <sys/systm.h>
62 #include <sys/proc.h>
63 #include <sys/fcntl.h>
64 #include <sys/file.h>
65 #include <sys/filedesc.h>
66 #include <sys/filio.h>
67 #include <sys/kernel.h>
68 #include <sys/ttycom.h>
69 #include <sys/stat.h>
70 #include <sys/poll.h>
71 #include <sys/signalvar.h>
72 #include <sys/vnode.h>
73 #include <sys/uio.h>
74 #include <sys/select.h>
75 #include <sys/mount.h>
76 #include <sys/syscallargs.h>
77 #include <sys/sysctl.h>
78 #include <sys/kauth.h>
79 #include <sys/atomic.h>
80 #include <sys/pipe.h>
81 
82 static int	pipe_read(file_t *, off_t *, struct uio *, kauth_cred_t, int);
83 static int	pipe_write(file_t *, off_t *, struct uio *, kauth_cred_t, int);
84 static int	pipe_close(file_t *);
85 static int	pipe_poll(file_t *, int);
86 static int	pipe_kqfilter(file_t *, struct knote *);
87 static int	pipe_stat(file_t *, struct stat *);
88 static int	pipe_ioctl(file_t *, u_long, void *);
89 static void	pipe_restart(file_t *);
90 static int	pipe_fpathconf(file_t *, int, register_t *);
91 static int	pipe_posix_fadvise(file_t *, off_t, off_t, int);
92 
93 static const struct fileops pipeops = {
94 	.fo_name = "pipe",
95 	.fo_read = pipe_read,
96 	.fo_write = pipe_write,
97 	.fo_ioctl = pipe_ioctl,
98 	.fo_fcntl = fnullop_fcntl,
99 	.fo_poll = pipe_poll,
100 	.fo_stat = pipe_stat,
101 	.fo_close = pipe_close,
102 	.fo_kqfilter = pipe_kqfilter,
103 	.fo_restart = pipe_restart,
104 	.fo_fpathconf = pipe_fpathconf,
105 	.fo_posix_fadvise = pipe_posix_fadvise,
106 };
107 
108 /*
109  * Default pipe buffer size(s), this can be kind-of large now because pipe
110  * space is pageable.  The pipe code will try to maintain locality of
111  * reference for performance reasons, so small amounts of outstanding I/O
112  * will not wipe the cache.
113  */
114 #define	MINPIPESIZE	(PIPE_SIZE / 3)
115 #define	MAXPIPESIZE	(2 * PIPE_SIZE / 3)
116 
117 /*
118  * Limit the number of "big" pipes
119  */
120 #define	LIMITBIGPIPES	32
121 static u_int	maxbigpipes __read_mostly = LIMITBIGPIPES;
122 static u_int	nbigpipe = 0;
123 
124 /*
125  * Amount of KVA consumed by pipe buffers.
126  */
127 static u_int	amountpipekva = 0;
128 
129 static void	pipeclose(struct pipe *);
130 static void	pipe_free_kmem(struct pipe *);
131 static int	pipe_create(struct pipe **, pool_cache_t, struct timespec *);
132 static int	pipelock(struct pipe *, bool);
133 static inline void pipeunlock(struct pipe *);
134 static void	pipeselwakeup(struct pipe *, struct pipe *, int);
135 static int	pipespace(struct pipe *, int);
136 static int	pipe_ctor(void *, void *, int);
137 static void	pipe_dtor(void *, void *);
138 
139 static pool_cache_t	pipe_wr_cache;
140 static pool_cache_t	pipe_rd_cache;
141 
142 void
pipe_init(void)143 pipe_init(void)
144 {
145 
146 	/* Writer side is not automatically allocated KVA. */
147 	pipe_wr_cache = pool_cache_init(sizeof(struct pipe), 0, 0, 0, "pipewr",
148 	    NULL, IPL_NONE, pipe_ctor, pipe_dtor, NULL);
149 	KASSERT(pipe_wr_cache != NULL);
150 
151 	/* Reader side gets preallocated KVA. */
152 	pipe_rd_cache = pool_cache_init(sizeof(struct pipe), 0, 0, 0, "piperd",
153 	    NULL, IPL_NONE, pipe_ctor, pipe_dtor, (void *)1);
154 	KASSERT(pipe_rd_cache != NULL);
155 }
156 
157 static int
pipe_ctor(void * arg,void * obj,int flags)158 pipe_ctor(void *arg, void *obj, int flags)
159 {
160 	struct pipe *pipe;
161 	vaddr_t va;
162 
163 	pipe = obj;
164 
165 	memset(pipe, 0, sizeof(struct pipe));
166 	if (arg != NULL) {
167 		/* Preallocate space. */
168 		va = uvm_km_alloc(kernel_map, PIPE_SIZE, 0,
169 		    UVM_KMF_PAGEABLE | UVM_KMF_WAITVA);
170 		KASSERT(va != 0);
171 		pipe->pipe_kmem = va;
172 		atomic_add_int(&amountpipekva, PIPE_SIZE);
173 	}
174 	cv_init(&pipe->pipe_rcv, "pipe_rd");
175 	cv_init(&pipe->pipe_wcv, "pipe_wr");
176 	cv_init(&pipe->pipe_draincv, "pipe_drn");
177 	cv_init(&pipe->pipe_lkcv, "pipe_lk");
178 	selinit(&pipe->pipe_sel);
179 	pipe->pipe_state = PIPE_SIGNALR;
180 
181 	return 0;
182 }
183 
184 static void
pipe_dtor(void * arg,void * obj)185 pipe_dtor(void *arg, void *obj)
186 {
187 	struct pipe *pipe;
188 
189 	pipe = obj;
190 
191 	cv_destroy(&pipe->pipe_rcv);
192 	cv_destroy(&pipe->pipe_wcv);
193 	cv_destroy(&pipe->pipe_draincv);
194 	cv_destroy(&pipe->pipe_lkcv);
195 	seldestroy(&pipe->pipe_sel);
196 	if (pipe->pipe_kmem != 0) {
197 		uvm_km_free(kernel_map, pipe->pipe_kmem, PIPE_SIZE,
198 		    UVM_KMF_PAGEABLE);
199 		atomic_add_int(&amountpipekva, -PIPE_SIZE);
200 	}
201 }
202 
203 /*
204  * The pipe system call for the DTYPE_PIPE type of pipes
205  */
206 int
pipe1(struct lwp * l,int * fildes,int flags)207 pipe1(struct lwp *l, int *fildes, int flags)
208 {
209 	struct pipe *rpipe, *wpipe;
210 	struct timespec nt;
211 	file_t *rf, *wf;
212 	int fd, error;
213 	proc_t *p;
214 
215 	if (flags & ~(O_CLOEXEC|O_NONBLOCK|O_NOSIGPIPE))
216 		return EINVAL;
217 	p = curproc;
218 	rpipe = wpipe = NULL;
219 	getnanotime(&nt);
220 	if ((error = pipe_create(&rpipe, pipe_rd_cache, &nt)) ||
221 	    (error = pipe_create(&wpipe, pipe_wr_cache, &nt))) {
222 		goto free2;
223 	}
224 	rpipe->pipe_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE);
225 	wpipe->pipe_lock = rpipe->pipe_lock;
226 	mutex_obj_hold(wpipe->pipe_lock);
227 
228 	error = fd_allocfile(&rf, &fd);
229 	if (error)
230 		goto free2;
231 	fildes[0] = fd;
232 
233 	error = fd_allocfile(&wf, &fd);
234 	if (error)
235 		goto free3;
236 	fildes[1] = fd;
237 
238 	rf->f_flag = FREAD | flags;
239 	rf->f_type = DTYPE_PIPE;
240 	rf->f_pipe = rpipe;
241 	rf->f_ops = &pipeops;
242 	fd_set_exclose(l, fildes[0], (flags & O_CLOEXEC) != 0);
243 
244 	wf->f_flag = FWRITE | flags;
245 	wf->f_type = DTYPE_PIPE;
246 	wf->f_pipe = wpipe;
247 	wf->f_ops = &pipeops;
248 	fd_set_exclose(l, fildes[1], (flags & O_CLOEXEC) != 0);
249 
250 	rpipe->pipe_peer = wpipe;
251 	wpipe->pipe_peer = rpipe;
252 
253 	fd_affix(p, rf, fildes[0]);
254 	fd_affix(p, wf, fildes[1]);
255 	return (0);
256 free3:
257 	fd_abort(p, rf, fildes[0]);
258 free2:
259 	pipeclose(wpipe);
260 	pipeclose(rpipe);
261 
262 	return (error);
263 }
264 
265 /*
266  * Allocate kva for pipe circular buffer, the space is pageable
267  * This routine will 'realloc' the size of a pipe safely, if it fails
268  * it will retain the old buffer.
269  * If it fails it will return ENOMEM.
270  */
271 static int
pipespace(struct pipe * pipe,int size)272 pipespace(struct pipe *pipe, int size)
273 {
274 	void *buffer;
275 
276 	/*
277 	 * Allocate pageable virtual address space.  Physical memory is
278 	 * allocated on demand.
279 	 */
280 	if (size == PIPE_SIZE && pipe->pipe_kmem != 0) {
281 		buffer = (void *)pipe->pipe_kmem;
282 	} else {
283 		buffer = (void *)uvm_km_alloc(kernel_map, round_page(size),
284 		    0, UVM_KMF_PAGEABLE);
285 		if (buffer == NULL)
286 			return (ENOMEM);
287 		atomic_add_int(&amountpipekva, size);
288 	}
289 
290 	/* free old resources if we're resizing */
291 	pipe_free_kmem(pipe);
292 	pipe->pipe_buffer.buffer = buffer;
293 	pipe->pipe_buffer.size = size;
294 	pipe->pipe_buffer.in = 0;
295 	pipe->pipe_buffer.out = 0;
296 	pipe->pipe_buffer.cnt = 0;
297 	return (0);
298 }
299 
300 /*
301  * Initialize and allocate VM and memory for pipe.
302  */
303 static int
pipe_create(struct pipe ** pipep,pool_cache_t cache,struct timespec * nt)304 pipe_create(struct pipe **pipep, pool_cache_t cache, struct timespec *nt)
305 {
306 	struct pipe *pipe;
307 	int error;
308 
309 	pipe = pool_cache_get(cache, PR_WAITOK);
310 	KASSERT(pipe != NULL);
311 	*pipep = pipe;
312 	error = 0;
313 	pipe->pipe_atime = pipe->pipe_mtime = pipe->pipe_btime = *nt;
314 	pipe->pipe_lock = NULL;
315 	if (cache == pipe_rd_cache) {
316 		error = pipespace(pipe, PIPE_SIZE);
317 	} else {
318 		pipe->pipe_buffer.buffer = NULL;
319 		pipe->pipe_buffer.size = 0;
320 		pipe->pipe_buffer.in = 0;
321 		pipe->pipe_buffer.out = 0;
322 		pipe->pipe_buffer.cnt = 0;
323 	}
324 	return error;
325 }
326 
327 /*
328  * Lock a pipe for I/O, blocking other access
329  * Called with pipe spin lock held.
330  */
331 static int
pipelock(struct pipe * pipe,bool catch_p)332 pipelock(struct pipe *pipe, bool catch_p)
333 {
334 	int error;
335 
336 	KASSERT(mutex_owned(pipe->pipe_lock));
337 
338 	while (pipe->pipe_state & PIPE_LOCKFL) {
339 		if (catch_p) {
340 			error = cv_wait_sig(&pipe->pipe_lkcv, pipe->pipe_lock);
341 			if (error != 0) {
342 				return error;
343 			}
344 		} else
345 			cv_wait(&pipe->pipe_lkcv, pipe->pipe_lock);
346 	}
347 
348 	pipe->pipe_state |= PIPE_LOCKFL;
349 
350 	return 0;
351 }
352 
353 /*
354  * unlock a pipe I/O lock
355  */
356 static inline void
pipeunlock(struct pipe * pipe)357 pipeunlock(struct pipe *pipe)
358 {
359 
360 	KASSERT(pipe->pipe_state & PIPE_LOCKFL);
361 
362 	pipe->pipe_state &= ~PIPE_LOCKFL;
363 	cv_signal(&pipe->pipe_lkcv);
364 }
365 
366 /*
367  * Select/poll wakeup. This also sends SIGIO to peer connected to
368  * 'sigpipe' side of pipe.
369  */
370 static void
pipeselwakeup(struct pipe * selp,struct pipe * sigp,int code)371 pipeselwakeup(struct pipe *selp, struct pipe *sigp, int code)
372 {
373 	int band;
374 
375 	switch (code) {
376 	case POLL_IN:
377 		band = POLLIN|POLLRDNORM;
378 		break;
379 	case POLL_OUT:
380 		band = POLLOUT|POLLWRNORM;
381 		break;
382 	case POLL_HUP:
383 		band = POLLHUP;
384 		break;
385 	case POLL_ERR:
386 		band = POLLERR;
387 		break;
388 	default:
389 		band = 0;
390 #ifdef DIAGNOSTIC
391 		printf("bad siginfo code %d in pipe notification.\n", code);
392 #endif
393 		break;
394 	}
395 
396 	selnotify(&selp->pipe_sel, band, NOTE_SUBMIT);
397 
398 	if (sigp == NULL || (sigp->pipe_state & PIPE_ASYNC) == 0)
399 		return;
400 
401 	fownsignal(sigp->pipe_pgid, SIGIO, code, band, selp);
402 }
403 
404 static int
pipe_read(file_t * fp,off_t * offset,struct uio * uio,kauth_cred_t cred,int flags)405 pipe_read(file_t *fp, off_t *offset, struct uio *uio, kauth_cred_t cred,
406     int flags)
407 {
408 	struct pipe *rpipe = fp->f_pipe;
409 	struct pipebuf *bp = &rpipe->pipe_buffer;
410 	kmutex_t *lock = rpipe->pipe_lock;
411 	int error;
412 	size_t nread = 0;
413 	size_t size;
414 	size_t ocnt;
415 	unsigned int wakeup_state = 0;
416 
417 	/*
418 	 * Try to avoid locking the pipe if we have nothing to do.
419 	 *
420 	 * There are programs which share one pipe amongst multiple processes
421 	 * and perform non-blocking reads in parallel, even if the pipe is
422 	 * empty.  This in particular is the case with BSD make, which when
423 	 * spawned with a high -j number can find itself with over half of the
424 	 * calls failing to find anything.
425 	 */
426 	if ((fp->f_flag & FNONBLOCK) != 0) {
427 		if (__predict_false(uio->uio_resid == 0))
428 			return (0);
429 		if (atomic_load_relaxed(&bp->cnt) == 0 &&
430 		    (atomic_load_relaxed(&rpipe->pipe_state) & PIPE_EOF) == 0)
431 			return (EAGAIN);
432 	}
433 
434 	mutex_enter(lock);
435 	++rpipe->pipe_busy;
436 	ocnt = bp->cnt;
437 
438 again:
439 	error = pipelock(rpipe, true);
440 	if (error)
441 		goto unlocked_error;
442 
443 	while (uio->uio_resid) {
444 		/*
445 		 * Normal pipe buffer receive.
446 		 */
447 		if (bp->cnt > 0) {
448 			size = bp->size - bp->out;
449 			if (size > bp->cnt)
450 				size = bp->cnt;
451 			if (size > uio->uio_resid)
452 				size = uio->uio_resid;
453 
454 			mutex_exit(lock);
455 			error = uiomove((char *)bp->buffer + bp->out, size, uio);
456 			mutex_enter(lock);
457 			if (error)
458 				break;
459 
460 			bp->out += size;
461 			if (bp->out >= bp->size)
462 				bp->out = 0;
463 
464 			bp->cnt -= size;
465 
466 			/*
467 			 * If there is no more to read in the pipe, reset
468 			 * its pointers to the beginning.  This improves
469 			 * cache hit stats.
470 			 */
471 			if (bp->cnt == 0) {
472 				bp->in = 0;
473 				bp->out = 0;
474 			}
475 			nread += size;
476 			continue;
477 		}
478 
479 		/*
480 		 * Break if some data was read.
481 		 */
482 		if (nread > 0)
483 			break;
484 
485 		/*
486 		 * Detect EOF condition.
487 		 * Read returns 0 on EOF, no need to set error.
488 		 */
489 		if (rpipe->pipe_state & PIPE_EOF)
490 			break;
491 
492 		/*
493 		 * Don't block on non-blocking I/O.
494 		 */
495 		if (fp->f_flag & FNONBLOCK) {
496 			error = EAGAIN;
497 			break;
498 		}
499 
500 		/*
501 		 * Unlock the pipe buffer for our remaining processing.
502 		 * We will either break out with an error or we will
503 		 * sleep and relock to loop.
504 		 */
505 		pipeunlock(rpipe);
506 
507 #if 1   /* XXX (dsl) I'm sure these aren't needed here ... */
508 		/*
509 		 * We want to read more, wake up select/poll.
510 		 */
511 		pipeselwakeup(rpipe, rpipe->pipe_peer, POLL_OUT);
512 
513 		/*
514 		 * If the "write-side" is blocked, wake it up now.
515 		 */
516 		cv_broadcast(&rpipe->pipe_wcv);
517 #endif
518 
519 		if (wakeup_state & PIPE_RESTART) {
520 			error = ERESTART;
521 			goto unlocked_error;
522 		}
523 
524 		/* Now wait until the pipe is filled */
525 		error = cv_wait_sig(&rpipe->pipe_rcv, lock);
526 		if (error != 0)
527 			goto unlocked_error;
528 		wakeup_state = rpipe->pipe_state;
529 		goto again;
530 	}
531 
532 	if (error == 0)
533 		getnanotime(&rpipe->pipe_atime);
534 	pipeunlock(rpipe);
535 
536 unlocked_error:
537 	--rpipe->pipe_busy;
538 	if (rpipe->pipe_busy == 0) {
539 		rpipe->pipe_state &= ~PIPE_RESTART;
540 		cv_broadcast(&rpipe->pipe_draincv);
541 	}
542 	if (bp->cnt < MINPIPESIZE) {
543 		cv_broadcast(&rpipe->pipe_wcv);
544 	}
545 
546 	/*
547 	 * If anything was read off the buffer, signal to the writer it's
548 	 * possible to write more data. Also send signal if we are here for the
549 	 * first time after last write.
550 	 */
551 	if ((bp->size - bp->cnt) >= PIPE_BUF
552 	    && (ocnt != bp->cnt || (rpipe->pipe_state & PIPE_SIGNALR))) {
553 		pipeselwakeup(rpipe, rpipe->pipe_peer, POLL_OUT);
554 		rpipe->pipe_state &= ~PIPE_SIGNALR;
555 	}
556 
557 	mutex_exit(lock);
558 	return (error);
559 }
560 
561 static int
pipe_write(file_t * fp,off_t * offset,struct uio * uio,kauth_cred_t cred,int flags)562 pipe_write(file_t *fp, off_t *offset, struct uio *uio, kauth_cred_t cred,
563     int flags)
564 {
565 	struct pipe *wpipe, *rpipe;
566 	struct pipebuf *bp;
567 	kmutex_t *lock;
568 	int error;
569 	unsigned int wakeup_state = 0;
570 
571 	/* We want to write to our peer */
572 	rpipe = fp->f_pipe;
573 	lock = rpipe->pipe_lock;
574 	error = 0;
575 
576 	mutex_enter(lock);
577 	wpipe = rpipe->pipe_peer;
578 
579 	/*
580 	 * Detect loss of pipe read side, issue SIGPIPE if lost.
581 	 */
582 	if (wpipe == NULL || (wpipe->pipe_state & PIPE_EOF) != 0) {
583 		mutex_exit(lock);
584 		return EPIPE;
585 	}
586 	++wpipe->pipe_busy;
587 
588 	/* Acquire the long-term pipe lock */
589 	if ((error = pipelock(wpipe, true)) != 0) {
590 		--wpipe->pipe_busy;
591 		if (wpipe->pipe_busy == 0) {
592 			wpipe->pipe_state &= ~PIPE_RESTART;
593 			cv_broadcast(&wpipe->pipe_draincv);
594 		}
595 		mutex_exit(lock);
596 		return (error);
597 	}
598 
599 	bp = &wpipe->pipe_buffer;
600 
601 	/*
602 	 * If it is advantageous to resize the pipe buffer, do so.
603 	 */
604 	if ((uio->uio_resid > PIPE_SIZE) &&
605 	    (nbigpipe < maxbigpipes) &&
606 	    (bp->size <= PIPE_SIZE) && (bp->cnt == 0)) {
607 
608 		if (pipespace(wpipe, BIG_PIPE_SIZE) == 0)
609 			atomic_inc_uint(&nbigpipe);
610 	}
611 
612 	while (uio->uio_resid) {
613 		size_t space;
614 
615 		space = bp->size - bp->cnt;
616 
617 		/* Writes of size <= PIPE_BUF must be atomic. */
618 		if ((space < uio->uio_resid) && (uio->uio_resid <= PIPE_BUF))
619 			space = 0;
620 
621 		if (space > 0) {
622 			int size;	/* Transfer size */
623 			int segsize;	/* first segment to transfer */
624 
625 			/*
626 			 * Transfer size is minimum of uio transfer
627 			 * and free space in pipe buffer.
628 			 */
629 			if (space > uio->uio_resid)
630 				size = uio->uio_resid;
631 			else
632 				size = space;
633 			/*
634 			 * First segment to transfer is minimum of
635 			 * transfer size and contiguous space in
636 			 * pipe buffer.  If first segment to transfer
637 			 * is less than the transfer size, we've got
638 			 * a wraparound in the buffer.
639 			 */
640 			segsize = bp->size - bp->in;
641 			if (segsize > size)
642 				segsize = size;
643 
644 			/* Transfer first segment */
645 			mutex_exit(lock);
646 			error = uiomove((char *)bp->buffer + bp->in, segsize,
647 			    uio);
648 
649 			if (error == 0 && segsize < size) {
650 				/*
651 				 * Transfer remaining part now, to
652 				 * support atomic writes.  Wraparound
653 				 * happened.
654 				 */
655 				KASSERT(bp->in + segsize == bp->size);
656 				error = uiomove(bp->buffer,
657 				    size - segsize, uio);
658 			}
659 			mutex_enter(lock);
660 			if (error)
661 				break;
662 
663 			bp->in += size;
664 			if (bp->in >= bp->size) {
665 				KASSERT(bp->in == size - segsize + bp->size);
666 				bp->in = size - segsize;
667 			}
668 
669 			bp->cnt += size;
670 			KASSERT(bp->cnt <= bp->size);
671 			wakeup_state = 0;
672 		} else {
673 			/*
674 			 * If the "read-side" has been blocked, wake it up now.
675 			 */
676 			cv_broadcast(&wpipe->pipe_rcv);
677 
678 			/*
679 			 * Don't block on non-blocking I/O.
680 			 */
681 			if (fp->f_flag & FNONBLOCK) {
682 				error = EAGAIN;
683 				break;
684 			}
685 
686 			/*
687 			 * We have no more space and have something to offer,
688 			 * wake up select/poll.
689 			 */
690 			if (bp->cnt)
691 				pipeselwakeup(wpipe, wpipe, POLL_IN);
692 
693 			if (wakeup_state & PIPE_RESTART) {
694 				error = ERESTART;
695 				break;
696 			}
697 
698 			/*
699 			 * If read side wants to go away, we just issue a signal
700 			 * to ourselves.
701 			 */
702 			if (wpipe->pipe_state & PIPE_EOF) {
703 				error = EPIPE;
704 				break;
705 			}
706 
707 			pipeunlock(wpipe);
708 			error = cv_wait_sig(&wpipe->pipe_wcv, lock);
709 			(void)pipelock(wpipe, false);
710 			if (error != 0)
711 				break;
712 			wakeup_state = wpipe->pipe_state;
713 		}
714 	}
715 
716 	--wpipe->pipe_busy;
717 	if (wpipe->pipe_busy == 0) {
718 		wpipe->pipe_state &= ~PIPE_RESTART;
719 		cv_broadcast(&wpipe->pipe_draincv);
720 	}
721 	if (bp->cnt > 0) {
722 		cv_broadcast(&wpipe->pipe_rcv);
723 	}
724 
725 	/*
726 	 * Don't return EPIPE if I/O was successful
727 	 */
728 	if (error == EPIPE && bp->cnt == 0 && uio->uio_resid == 0)
729 		error = 0;
730 
731 	if (error == 0)
732 		getnanotime(&wpipe->pipe_mtime);
733 
734 	/*
735 	 * We have something to offer, wake up select/poll.
736 	 */
737 	if (bp->cnt)
738 		pipeselwakeup(wpipe, wpipe, POLL_IN);
739 
740 	/*
741 	 * Arrange for next read(2) to do a signal.
742 	 */
743 	wpipe->pipe_state |= PIPE_SIGNALR;
744 
745 	pipeunlock(wpipe);
746 	mutex_exit(lock);
747 	return (error);
748 }
749 
750 /*
751  * We implement a very minimal set of ioctls for compatibility with sockets.
752  */
753 int
pipe_ioctl(file_t * fp,u_long cmd,void * data)754 pipe_ioctl(file_t *fp, u_long cmd, void *data)
755 {
756 	struct pipe *pipe = fp->f_pipe;
757 	kmutex_t *lock = pipe->pipe_lock;
758 
759 	switch (cmd) {
760 
761 	case FIONBIO:
762 		return (0);
763 
764 	case FIOASYNC:
765 		mutex_enter(lock);
766 		if (*(int *)data) {
767 			pipe->pipe_state |= PIPE_ASYNC;
768 		} else {
769 			pipe->pipe_state &= ~PIPE_ASYNC;
770 		}
771 		mutex_exit(lock);
772 		return (0);
773 
774 	case FIONREAD:
775 		mutex_enter(lock);
776 		*(int *)data = pipe->pipe_buffer.cnt;
777 		mutex_exit(lock);
778 		return (0);
779 
780 	case FIONWRITE:
781 		/* Look at other side */
782 		mutex_enter(lock);
783 		pipe = pipe->pipe_peer;
784 		if (pipe == NULL)
785 			*(int *)data = 0;
786 		else
787 			*(int *)data = pipe->pipe_buffer.cnt;
788 		mutex_exit(lock);
789 		return (0);
790 
791 	case FIONSPACE:
792 		/* Look at other side */
793 		mutex_enter(lock);
794 		pipe = pipe->pipe_peer;
795 		if (pipe == NULL)
796 			*(int *)data = 0;
797 		else
798 			*(int *)data = pipe->pipe_buffer.size -
799 			    pipe->pipe_buffer.cnt;
800 		mutex_exit(lock);
801 		return (0);
802 
803 	case TIOCSPGRP:
804 	case FIOSETOWN:
805 		return fsetown(&pipe->pipe_pgid, cmd, data);
806 
807 	case TIOCGPGRP:
808 	case FIOGETOWN:
809 		return fgetown(pipe->pipe_pgid, cmd, data);
810 
811 	}
812 	return (EPASSTHROUGH);
813 }
814 
815 int
pipe_poll(file_t * fp,int events)816 pipe_poll(file_t *fp, int events)
817 {
818 	struct pipe *rpipe = fp->f_pipe;
819 	struct pipe *wpipe;
820 	int eof = 0;
821 	int revents = 0;
822 
823 	mutex_enter(rpipe->pipe_lock);
824 	wpipe = rpipe->pipe_peer;
825 
826 	if (events & (POLLIN | POLLRDNORM))
827 		if ((rpipe->pipe_buffer.cnt > 0) ||
828 		    (rpipe->pipe_state & PIPE_EOF))
829 			revents |= events & (POLLIN | POLLRDNORM);
830 
831 	eof |= (rpipe->pipe_state & PIPE_EOF);
832 
833 	if (wpipe == NULL)
834 		revents |= events & (POLLOUT | POLLWRNORM);
835 	else {
836 		if (events & (POLLOUT | POLLWRNORM))
837 			if ((wpipe->pipe_state & PIPE_EOF) || (
838 			     (wpipe->pipe_buffer.size - wpipe->pipe_buffer.cnt) >= PIPE_BUF))
839 				revents |= events & (POLLOUT | POLLWRNORM);
840 
841 		eof |= (wpipe->pipe_state & PIPE_EOF);
842 	}
843 
844 	if (wpipe == NULL || eof)
845 		revents |= POLLHUP;
846 
847 	if (revents == 0) {
848 		if (events & (POLLIN | POLLRDNORM))
849 			selrecord(curlwp, &rpipe->pipe_sel);
850 
851 		if (events & (POLLOUT | POLLWRNORM))
852 			selrecord(curlwp, &wpipe->pipe_sel);
853 	}
854 	mutex_exit(rpipe->pipe_lock);
855 
856 	return (revents);
857 }
858 
859 static int
pipe_stat(file_t * fp,struct stat * ub)860 pipe_stat(file_t *fp, struct stat *ub)
861 {
862 	struct pipe *pipe = fp->f_pipe;
863 
864 	mutex_enter(pipe->pipe_lock);
865 	memset(ub, 0, sizeof(*ub));
866 	ub->st_mode = S_IFIFO | S_IRUSR | S_IWUSR;
867 	ub->st_blksize = pipe->pipe_buffer.size;
868 	if (ub->st_blksize == 0 && pipe->pipe_peer)
869 		ub->st_blksize = pipe->pipe_peer->pipe_buffer.size;
870 	ub->st_size = pipe->pipe_buffer.cnt;
871 	ub->st_blocks = (ub->st_size) ? 1 : 0;
872 	ub->st_atimespec = pipe->pipe_atime;
873 	ub->st_mtimespec = pipe->pipe_mtime;
874 	ub->st_ctimespec = ub->st_birthtimespec = pipe->pipe_btime;
875 	ub->st_uid = kauth_cred_geteuid(fp->f_cred);
876 	ub->st_gid = kauth_cred_getegid(fp->f_cred);
877 
878 	/*
879 	 * Left as 0: st_dev, st_ino, st_nlink, st_rdev, st_flags, st_gen.
880 	 * XXX (st_dev, st_ino) should be unique.
881 	 */
882 	mutex_exit(pipe->pipe_lock);
883 	return 0;
884 }
885 
886 static int
pipe_close(file_t * fp)887 pipe_close(file_t *fp)
888 {
889 	struct pipe *pipe = fp->f_pipe;
890 
891 	fp->f_pipe = NULL;
892 	pipeclose(pipe);
893 	return (0);
894 }
895 
896 static void
pipe_restart(file_t * fp)897 pipe_restart(file_t *fp)
898 {
899 	struct pipe *pipe = fp->f_pipe;
900 
901 	/*
902 	 * Unblock blocked reads/writes in order to allow close() to complete.
903 	 * System calls return ERESTART so that the fd is revalidated.
904 	 * (Partial writes return the transfer length.)
905 	 */
906 	mutex_enter(pipe->pipe_lock);
907 	pipe->pipe_state |= PIPE_RESTART;
908 	/* Wakeup both cvs, maybe we only need one, but maybe there are some
909 	 * other paths where wakeup is needed, and it saves deciding which! */
910 	cv_broadcast(&pipe->pipe_rcv);
911 	cv_broadcast(&pipe->pipe_wcv);
912 	mutex_exit(pipe->pipe_lock);
913 }
914 
915 static int
pipe_fpathconf(struct file * fp,int name,register_t * retval)916 pipe_fpathconf(struct file *fp, int name, register_t *retval)
917 {
918 
919 	switch (name) {
920 	case _PC_PIPE_BUF:
921 		*retval = PIPE_BUF;
922 		return 0;
923 	default:
924 		return EINVAL;
925 	}
926 }
927 
928 static int
pipe_posix_fadvise(struct file * fp,off_t offset,off_t len,int advice)929 pipe_posix_fadvise(struct file *fp, off_t offset, off_t len, int advice)
930 {
931 
932 	return ESPIPE;
933 }
934 
935 static void
pipe_free_kmem(struct pipe * pipe)936 pipe_free_kmem(struct pipe *pipe)
937 {
938 
939 	if (pipe->pipe_buffer.buffer != NULL) {
940 		if (pipe->pipe_buffer.size > PIPE_SIZE) {
941 			atomic_dec_uint(&nbigpipe);
942 		}
943 		if (pipe->pipe_buffer.buffer != (void *)pipe->pipe_kmem) {
944 			uvm_km_free(kernel_map,
945 			    (vaddr_t)pipe->pipe_buffer.buffer,
946 			    pipe->pipe_buffer.size, UVM_KMF_PAGEABLE);
947 			atomic_add_int(&amountpipekva,
948 			    -pipe->pipe_buffer.size);
949 		}
950 		pipe->pipe_buffer.buffer = NULL;
951 	}
952 }
953 
954 /*
955  * Shutdown the pipe.
956  */
957 static void
pipeclose(struct pipe * pipe)958 pipeclose(struct pipe *pipe)
959 {
960 	kmutex_t *lock;
961 	struct pipe *ppipe;
962 
963 	if (pipe == NULL)
964 		return;
965 
966 	KASSERT(cv_is_valid(&pipe->pipe_rcv));
967 	KASSERT(cv_is_valid(&pipe->pipe_wcv));
968 	KASSERT(cv_is_valid(&pipe->pipe_draincv));
969 	KASSERT(cv_is_valid(&pipe->pipe_lkcv));
970 
971 	lock = pipe->pipe_lock;
972 	if (lock == NULL)
973 		/* Must have failed during create */
974 		goto free_resources;
975 
976 	mutex_enter(lock);
977 	pipeselwakeup(pipe, pipe, POLL_HUP);
978 
979 	/*
980 	 * If the other side is blocked, wake it up saying that
981 	 * we want to close it down.
982 	 */
983 	pipe->pipe_state |= PIPE_EOF;
984 	if (pipe->pipe_busy) {
985 		while (pipe->pipe_busy) {
986 			cv_broadcast(&pipe->pipe_wcv);
987 			cv_wait_sig(&pipe->pipe_draincv, lock);
988 		}
989 	}
990 
991 	/*
992 	 * Disconnect from peer.
993 	 */
994 	if ((ppipe = pipe->pipe_peer) != NULL) {
995 		pipeselwakeup(ppipe, ppipe, POLL_HUP);
996 		ppipe->pipe_state |= PIPE_EOF;
997 		cv_broadcast(&ppipe->pipe_rcv);
998 		ppipe->pipe_peer = NULL;
999 	}
1000 
1001 	/*
1002 	 * Any knote objects still left in the list are
1003 	 * the one attached by peer.  Since no one will
1004 	 * traverse this list, we just clear it.
1005 	 *
1006 	 * XXX Exposes select/kqueue internals.
1007 	 */
1008 	SLIST_INIT(&pipe->pipe_sel.sel_klist);
1009 
1010 	KASSERT((pipe->pipe_state & PIPE_LOCKFL) == 0);
1011 	mutex_exit(lock);
1012 	mutex_obj_free(lock);
1013 
1014 	/*
1015 	 * Free resources.
1016 	 */
1017     free_resources:
1018 	pipe->pipe_pgid = 0;
1019 	pipe->pipe_state = PIPE_SIGNALR;
1020 	pipe->pipe_peer = NULL;
1021 	pipe->pipe_lock = NULL;
1022 	pipe_free_kmem(pipe);
1023 	if (pipe->pipe_kmem != 0) {
1024 		pool_cache_put(pipe_rd_cache, pipe);
1025 	} else {
1026 		pool_cache_put(pipe_wr_cache, pipe);
1027 	}
1028 }
1029 
1030 static void
filt_pipedetach(struct knote * kn)1031 filt_pipedetach(struct knote *kn)
1032 {
1033 	struct pipe *pipe;
1034 	kmutex_t *lock;
1035 
1036 	pipe = ((file_t *)kn->kn_obj)->f_pipe;
1037 	lock = pipe->pipe_lock;
1038 
1039 	mutex_enter(lock);
1040 
1041 	switch(kn->kn_filter) {
1042 	case EVFILT_WRITE:
1043 		/* Need the peer structure, not our own. */
1044 		pipe = pipe->pipe_peer;
1045 
1046 		/* If reader end already closed, just return. */
1047 		if (pipe == NULL) {
1048 			mutex_exit(lock);
1049 			return;
1050 		}
1051 
1052 		break;
1053 	default:
1054 		/* Nothing to do. */
1055 		break;
1056 	}
1057 
1058 	KASSERT(kn->kn_hook == pipe);
1059 	selremove_knote(&pipe->pipe_sel, kn);
1060 	mutex_exit(lock);
1061 }
1062 
1063 static int
filt_piperead(struct knote * kn,long hint)1064 filt_piperead(struct knote *kn, long hint)
1065 {
1066 	struct pipe *rpipe = ((file_t *)kn->kn_obj)->f_pipe;
1067 	struct pipe *wpipe;
1068 	int rv;
1069 
1070 	if ((hint & NOTE_SUBMIT) == 0) {
1071 		mutex_enter(rpipe->pipe_lock);
1072 	}
1073 	wpipe = rpipe->pipe_peer;
1074 	kn->kn_data = rpipe->pipe_buffer.cnt;
1075 
1076 	if ((rpipe->pipe_state & PIPE_EOF) ||
1077 	    (wpipe == NULL) || (wpipe->pipe_state & PIPE_EOF)) {
1078 		knote_set_eof(kn, 0);
1079 		rv = 1;
1080 	} else {
1081 		rv = kn->kn_data > 0;
1082 	}
1083 
1084 	if ((hint & NOTE_SUBMIT) == 0) {
1085 		mutex_exit(rpipe->pipe_lock);
1086 	}
1087 	return rv;
1088 }
1089 
1090 static int
filt_pipewrite(struct knote * kn,long hint)1091 filt_pipewrite(struct knote *kn, long hint)
1092 {
1093 	struct pipe *rpipe = ((file_t *)kn->kn_obj)->f_pipe;
1094 	struct pipe *wpipe;
1095 	int rv;
1096 
1097 	if ((hint & NOTE_SUBMIT) == 0) {
1098 		mutex_enter(rpipe->pipe_lock);
1099 	}
1100 	wpipe = rpipe->pipe_peer;
1101 
1102 	if ((wpipe == NULL) || (wpipe->pipe_state & PIPE_EOF)) {
1103 		kn->kn_data = 0;
1104 		knote_set_eof(kn, 0);
1105 		rv = 1;
1106 	} else {
1107 		kn->kn_data = wpipe->pipe_buffer.size - wpipe->pipe_buffer.cnt;
1108 		rv = kn->kn_data >= PIPE_BUF;
1109 	}
1110 
1111 	if ((hint & NOTE_SUBMIT) == 0) {
1112 		mutex_exit(rpipe->pipe_lock);
1113 	}
1114 	return rv;
1115 }
1116 
1117 static const struct filterops pipe_rfiltops = {
1118 	.f_flags = FILTEROP_ISFD | FILTEROP_MPSAFE,
1119 	.f_attach = NULL,
1120 	.f_detach = filt_pipedetach,
1121 	.f_event = filt_piperead,
1122 };
1123 
1124 static const struct filterops pipe_wfiltops = {
1125 	.f_flags = FILTEROP_ISFD | FILTEROP_MPSAFE,
1126 	.f_attach = NULL,
1127 	.f_detach = filt_pipedetach,
1128 	.f_event = filt_pipewrite,
1129 };
1130 
1131 static int
pipe_kqfilter(file_t * fp,struct knote * kn)1132 pipe_kqfilter(file_t *fp, struct knote *kn)
1133 {
1134 	struct pipe *pipe;
1135 	kmutex_t *lock;
1136 
1137 	pipe = ((file_t *)kn->kn_obj)->f_pipe;
1138 	lock = pipe->pipe_lock;
1139 
1140 	mutex_enter(lock);
1141 
1142 	switch (kn->kn_filter) {
1143 	case EVFILT_READ:
1144 		kn->kn_fop = &pipe_rfiltops;
1145 		break;
1146 	case EVFILT_WRITE:
1147 		kn->kn_fop = &pipe_wfiltops;
1148 		pipe = pipe->pipe_peer;
1149 		if (pipe == NULL) {
1150 			/* Other end of pipe has been closed. */
1151 			mutex_exit(lock);
1152 			return (EBADF);
1153 		}
1154 		break;
1155 	default:
1156 		mutex_exit(lock);
1157 		return (EINVAL);
1158 	}
1159 
1160 	kn->kn_hook = pipe;
1161 	selrecord_knote(&pipe->pipe_sel, kn);
1162 	mutex_exit(lock);
1163 
1164 	return (0);
1165 }
1166 
1167 /*
1168  * Handle pipe sysctls.
1169  */
1170 SYSCTL_SETUP(sysctl_kern_pipe_setup, "sysctl kern.pipe subtree setup")
1171 {
1172 
1173 	sysctl_createv(clog, 0, NULL, NULL,
1174 		       CTLFLAG_PERMANENT,
1175 		       CTLTYPE_NODE, "pipe",
1176 		       SYSCTL_DESCR("Pipe settings"),
1177 		       NULL, 0, NULL, 0,
1178 		       CTL_KERN, KERN_PIPE, CTL_EOL);
1179 
1180 	sysctl_createv(clog, 0, NULL, NULL,
1181 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1182 		       CTLTYPE_INT, "maxbigpipes",
1183 		       SYSCTL_DESCR("Maximum number of \"big\" pipes"),
1184 		       NULL, 0, &maxbigpipes, 0,
1185 		       CTL_KERN, KERN_PIPE, KERN_PIPE_MAXBIGPIPES, CTL_EOL);
1186 	sysctl_createv(clog, 0, NULL, NULL,
1187 		       CTLFLAG_PERMANENT,
1188 		       CTLTYPE_INT, "nbigpipes",
1189 		       SYSCTL_DESCR("Number of \"big\" pipes"),
1190 		       NULL, 0, &nbigpipe, 0,
1191 		       CTL_KERN, KERN_PIPE, KERN_PIPE_NBIGPIPES, CTL_EOL);
1192 	sysctl_createv(clog, 0, NULL, NULL,
1193 		       CTLFLAG_PERMANENT,
1194 		       CTLTYPE_INT, "kvasize",
1195 		       SYSCTL_DESCR("Amount of kernel memory consumed by pipe "
1196 				    "buffers"),
1197 		       NULL, 0, &amountpipekva, 0,
1198 		       CTL_KERN, KERN_PIPE, KERN_PIPE_KVASIZE, CTL_EOL);
1199 }
1200