xref: /netbsd-src/sys/kern/vfs_lockf.c (revision da5f4674a3fc214be3572d358b66af40ab9401e7)
1 /*	$NetBSD: vfs_lockf.c,v 1.33 2003/08/07 16:32:02 agc Exp $	*/
2 
3 /*
4  * Copyright (c) 1982, 1986, 1989, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * This code is derived from software contributed to Berkeley by
8  * Scooter Morris at Genentech Inc.
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  * 3. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  *	@(#)ufs_lockf.c	8.4 (Berkeley) 10/26/94
35  */
36 
37 #include <sys/cdefs.h>
38 __KERNEL_RCSID(0, "$NetBSD: vfs_lockf.c,v 1.33 2003/08/07 16:32:02 agc Exp $");
39 
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/kernel.h>
43 #include <sys/file.h>
44 #include <sys/proc.h>
45 #include <sys/vnode.h>
46 #include <sys/malloc.h>
47 #include <sys/fcntl.h>
48 #include <sys/lockf.h>
49 
50 MALLOC_DEFINE(M_LOCKF, "lockf", "Byte-range locking structures");
51 
52 /*
53  * This variable controls the maximum number of processes that will
54  * be checked in doing deadlock detection.
55  */
56 int maxlockdepth = MAXDEPTH;
57 
58 #ifdef LOCKF_DEBUG
59 int	lockf_debug = 0;
60 #endif
61 
62 #define NOLOCKF (struct lockf *)0
63 #define SELF	0x1
64 #define OTHERS	0x2
65 
66 static int lf_clearlock(struct lockf *, struct lockf **);
67 static int lf_findoverlap(struct lockf *,
68 	    struct lockf *, int, struct lockf ***, struct lockf **);
69 static struct lockf *lf_getblock(struct lockf *);
70 static int lf_getlock(struct lockf *, struct flock *);
71 static int lf_setlock(struct lockf *, struct lockf **, struct simplelock *);
72 static void lf_split(struct lockf *, struct lockf *, struct lockf **);
73 static void lf_wakelock(struct lockf *);
74 
75 #ifdef LOCKF_DEBUG
76 static void lf_print(char *, struct lockf *);
77 static void lf_printlist(char *, struct lockf *);
78 #endif
79 
80 /*
81  * XXX TODO
82  * Misc cleanups: "caddr_t id" should be visible in the API as a
83  * "struct proc *".
84  * (This requires rototilling all VFS's which support advisory locking).
85  *
86  * Use pools for lock allocation.
87  */
88 
89 /*
90  * If there's a lot of lock contention on a single vnode, locking
91  * schemes which allow for more paralleism would be needed.  Given how
92  * infrequently byte-range locks are actually used in typical BSD
93  * code, a more complex approach probably isn't worth it.
94  */
95 
96 /*
97  * Do an advisory lock operation.
98  */
99 int
100 lf_advlock(struct vop_advlock_args *ap, struct lockf **head, off_t size)
101 {
102 	struct flock *fl = ap->a_fl;
103 	struct lockf *lock = NULL;
104 	struct lockf *sparelock;
105 	struct simplelock *interlock = &ap->a_vp->v_interlock;
106 	off_t start, end;
107 	int error;
108 
109 	/*
110 	 * Convert the flock structure into a start and end.
111 	 */
112 	switch (fl->l_whence) {
113 	case SEEK_SET:
114 	case SEEK_CUR:
115 		/*
116 		 * Caller is responsible for adding any necessary offset
117 		 * when SEEK_CUR is used.
118 		 */
119 		start = fl->l_start;
120 		break;
121 
122 	case SEEK_END:
123 		start = size + fl->l_start;
124 		break;
125 
126 	default:
127 		return EINVAL;
128 	}
129 	if (start < 0)
130 		return EINVAL;
131 
132 	/*
133 	 * allocate locks before acquire simple lock.
134 	 * we need two locks in the worst case.
135 	 */
136 	switch (ap->a_op) {
137 	case F_SETLK:
138 	case F_UNLCK:
139 		/*
140 		 * XXX for F_UNLCK case, we can re-use lock.
141 		 */
142 		if ((fl->l_type & F_FLOCK) == 0) {
143 			/*
144 			 * byte-range lock might need one more lock.
145 			 */
146 			MALLOC(sparelock, struct lockf *, sizeof(*lock),
147 			    M_LOCKF, M_WAITOK);
148 			if (sparelock == NULL) {
149 				error = ENOMEM;
150 				goto quit;
151 			}
152 			break;
153 		}
154 		/* FALLTHROUGH */
155 
156 	case F_GETLK:
157 		sparelock = NULL;
158 		break;
159 
160 	default:
161 		return EINVAL;
162 	}
163 
164 	MALLOC(lock, struct lockf *, sizeof(*lock), M_LOCKF, M_WAITOK);
165 	if (lock == NULL) {
166 		error = ENOMEM;
167 		goto quit;
168 	}
169 
170 	simple_lock(interlock);
171 
172 	/*
173 	 * Avoid the common case of unlocking when inode has no locks.
174 	 */
175 	if (*head == (struct lockf *)0) {
176 		if (ap->a_op != F_SETLK) {
177 			fl->l_type = F_UNLCK;
178 			error = 0;
179 			goto quit_unlock;
180 		}
181 	}
182 
183 	if (fl->l_len == 0)
184 		end = -1;
185 	else
186 		end = start + fl->l_len - 1;
187 	/*
188 	 * Create the lockf structure.
189 	 */
190 	lock->lf_start = start;
191 	lock->lf_end = end;
192 	/* XXX NJWLWP
193 	 * I don't want to make the entire VFS universe use LWPs, because
194 	 * they don't need them, for the most part. This is an exception,
195 	 * and a kluge.
196 	 */
197 
198 	lock->lf_head = head;
199 	lock->lf_type = fl->l_type;
200 	lock->lf_next = (struct lockf *)0;
201 	TAILQ_INIT(&lock->lf_blkhd);
202 	lock->lf_flags = ap->a_flags;
203 	if (lock->lf_flags & F_POSIX) {
204 		KASSERT(curproc == (struct proc *)ap->a_id);
205 	}
206 	lock->lf_id = (struct proc *)ap->a_id;
207 	lock->lf_lwp = curlwp;
208 
209 	/*
210 	 * Do the requested operation.
211 	 */
212 	switch (ap->a_op) {
213 
214 	case F_SETLK:
215 		error = lf_setlock(lock, &sparelock, interlock);
216 		lock = NULL; /* lf_setlock freed it */
217 		break;
218 
219 	case F_UNLCK:
220 		error = lf_clearlock(lock, &sparelock);
221 		break;
222 
223 	case F_GETLK:
224 		error = lf_getlock(lock, fl);
225 		break;
226 
227 	default:
228 		break;
229 		/* NOTREACHED */
230 	}
231 
232 quit_unlock:
233 	simple_unlock(interlock);
234 quit:
235 	if (lock)
236 		FREE(lock, M_LOCKF);
237 	if (sparelock)
238 		FREE(sparelock, M_LOCKF);
239 
240 	return error;
241 }
242 
243 /*
244  * Set a byte-range lock.
245  */
246 static int
247 lf_setlock(struct lockf *lock, struct lockf **sparelock,
248     struct simplelock *interlock)
249 {
250 	struct lockf *block;
251 	struct lockf **head = lock->lf_head;
252 	struct lockf **prev, *overlap, *ltmp;
253 	static char lockstr[] = "lockf";
254 	int ovcase, priority, needtolink, error;
255 
256 #ifdef LOCKF_DEBUG
257 	if (lockf_debug & 1)
258 		lf_print("lf_setlock", lock);
259 #endif /* LOCKF_DEBUG */
260 
261 	/*
262 	 * Set the priority
263 	 */
264 	priority = PLOCK;
265 	if (lock->lf_type == F_WRLCK)
266 		priority += 4;
267 	priority |= PCATCH;
268 	/*
269 	 * Scan lock list for this file looking for locks that would block us.
270 	 */
271 	while ((block = lf_getblock(lock)) != NULL) {
272 		/*
273 		 * Free the structure and return if nonblocking.
274 		 */
275 		if ((lock->lf_flags & F_WAIT) == 0) {
276 			FREE(lock, M_LOCKF);
277 			return EAGAIN;
278 		}
279 		/*
280 		 * We are blocked. Since flock style locks cover
281 		 * the whole file, there is no chance for deadlock.
282 		 * For byte-range locks we must check for deadlock.
283 		 *
284 		 * Deadlock detection is done by looking through the
285 		 * wait channels to see if there are any cycles that
286 		 * involve us. MAXDEPTH is set just to make sure we
287 		 * do not go off into neverneverland.
288 		 */
289 		if ((lock->lf_flags & F_POSIX) &&
290 		    (block->lf_flags & F_POSIX)) {
291 			struct lwp *wlwp;
292 			struct lockf *waitblock;
293 			int i = 0;
294 
295 			/*
296 			 * The block is waiting on something.  if_lwp will be
297 			 * 0 once the lock is granted, so we terminate the
298 			 * loop if we find this.
299 			 */
300 			wlwp = block->lf_lwp;
301 			while (wlwp && (i++ < maxlockdepth)) {
302 				waitblock = (struct lockf *)wlwp->l_wchan;
303 				/* Get the owner of the blocking lock */
304 				waitblock = waitblock->lf_next;
305 				if ((waitblock->lf_flags & F_POSIX) == 0)
306 					break;
307 				wlwp = waitblock->lf_lwp;
308 				if (wlwp == lock->lf_lwp) {
309 					free(lock, M_LOCKF);
310 					return EDEADLK;
311 				}
312 			}
313 			/*
314 			 * If we're still following a dependancy chain
315 			 * after maxlockdepth iterations, assume we're in
316 			 * a cycle to be safe.
317 			 */
318 			if (i >= maxlockdepth) {
319 				free(lock, M_LOCKF);
320 				return EDEADLK;
321 			}
322 		}
323 		/*
324 		 * For flock type locks, we must first remove
325 		 * any shared locks that we hold before we sleep
326 		 * waiting for an exclusive lock.
327 		 */
328 		if ((lock->lf_flags & F_FLOCK) &&
329 		    lock->lf_type == F_WRLCK) {
330 			lock->lf_type = F_UNLCK;
331 			(void) lf_clearlock(lock, NULL);
332 			lock->lf_type = F_WRLCK;
333 		}
334 		/*
335 		 * Add our lock to the blocked list and sleep until we're free.
336 		 * Remember who blocked us (for deadlock detection).
337 		 */
338 		lock->lf_next = block;
339 		TAILQ_INSERT_TAIL(&block->lf_blkhd, lock, lf_block);
340 #ifdef LOCKF_DEBUG
341 		if (lockf_debug & 1) {
342 			lf_print("lf_setlock: blocking on", block);
343 			lf_printlist("lf_setlock", block);
344 		}
345 #endif /* LOCKF_DEBUG */
346 		error = ltsleep(lock, priority, lockstr, 0, interlock);
347 
348 		/*
349 		 * We may have been awakened by a signal (in
350 		 * which case we must remove ourselves from the
351 		 * blocked list) and/or by another process
352 		 * releasing a lock (in which case we have already
353 		 * been removed from the blocked list and our
354 		 * lf_next field set to NOLOCKF).
355 		 */
356 		if (lock->lf_next != NOLOCKF) {
357 			TAILQ_REMOVE(&lock->lf_next->lf_blkhd, lock, lf_block);
358 			lock->lf_next = NOLOCKF;
359 		}
360 		if (error) {
361 			free(lock, M_LOCKF);
362 			return error;
363 		}
364 	}
365 	/*
366 	 * No blocks!!  Add the lock.  Note that we will
367 	 * downgrade or upgrade any overlapping locks this
368 	 * process already owns.
369 	 *
370 	 * Skip over locks owned by other processes.
371 	 * Handle any locks that overlap and are owned by ourselves.
372 	 */
373 	lock->lf_lwp = 0;
374 	prev = head;
375 	block = *head;
376 	needtolink = 1;
377 	for (;;) {
378 		ovcase = lf_findoverlap(block, lock, SELF, &prev, &overlap);
379 		if (ovcase)
380 			block = overlap->lf_next;
381 		/*
382 		 * Six cases:
383 		 *	0) no overlap
384 		 *	1) overlap == lock
385 		 *	2) overlap contains lock
386 		 *	3) lock contains overlap
387 		 *	4) overlap starts before lock
388 		 *	5) overlap ends after lock
389 		 */
390 		switch (ovcase) {
391 		case 0: /* no overlap */
392 			if (needtolink) {
393 				*prev = lock;
394 				lock->lf_next = overlap;
395 			}
396 			break;
397 
398 		case 1: /* overlap == lock */
399 			/*
400 			 * If downgrading lock, others may be
401 			 * able to acquire it.
402 			 */
403 			if (lock->lf_type == F_RDLCK &&
404 			    overlap->lf_type == F_WRLCK)
405 				lf_wakelock(overlap);
406 			overlap->lf_type = lock->lf_type;
407 			FREE(lock, M_LOCKF);
408 			lock = overlap; /* for debug output below */
409 			break;
410 
411 		case 2: /* overlap contains lock */
412 			/*
413 			 * Check for common starting point and different types.
414 			 */
415 			if (overlap->lf_type == lock->lf_type) {
416 				free(lock, M_LOCKF);
417 				lock = overlap; /* for debug output below */
418 				break;
419 			}
420 			if (overlap->lf_start == lock->lf_start) {
421 				*prev = lock;
422 				lock->lf_next = overlap;
423 				overlap->lf_start = lock->lf_end + 1;
424 			} else
425 				lf_split(overlap, lock, sparelock);
426 			lf_wakelock(overlap);
427 			break;
428 
429 		case 3: /* lock contains overlap */
430 			/*
431 			 * If downgrading lock, others may be able to
432 			 * acquire it, otherwise take the list.
433 			 */
434 			if (lock->lf_type == F_RDLCK &&
435 			    overlap->lf_type == F_WRLCK) {
436 				lf_wakelock(overlap);
437 			} else {
438 				while ((ltmp = TAILQ_FIRST(&overlap->lf_blkhd))) {
439 					KASSERT(ltmp->lf_next == overlap);
440 					TAILQ_REMOVE(&overlap->lf_blkhd, ltmp,
441 					    lf_block);
442 					ltmp->lf_next = lock;
443 					TAILQ_INSERT_TAIL(&lock->lf_blkhd,
444 					    ltmp, lf_block);
445 				}
446 			}
447 			/*
448 			 * Add the new lock if necessary and delete the overlap.
449 			 */
450 			if (needtolink) {
451 				*prev = lock;
452 				lock->lf_next = overlap->lf_next;
453 				prev = &lock->lf_next;
454 				needtolink = 0;
455 			} else
456 				*prev = overlap->lf_next;
457 			free(overlap, M_LOCKF);
458 			continue;
459 
460 		case 4: /* overlap starts before lock */
461 			/*
462 			 * Add lock after overlap on the list.
463 			 */
464 			lock->lf_next = overlap->lf_next;
465 			overlap->lf_next = lock;
466 			overlap->lf_end = lock->lf_start - 1;
467 			prev = &lock->lf_next;
468 			lf_wakelock(overlap);
469 			needtolink = 0;
470 			continue;
471 
472 		case 5: /* overlap ends after lock */
473 			/*
474 			 * Add the new lock before overlap.
475 			 */
476 			if (needtolink) {
477 				*prev = lock;
478 				lock->lf_next = overlap;
479 			}
480 			overlap->lf_start = lock->lf_end + 1;
481 			lf_wakelock(overlap);
482 			break;
483 		}
484 		break;
485 	}
486 #ifdef LOCKF_DEBUG
487 	if (lockf_debug & 1) {
488 		lf_print("lf_setlock: got the lock", lock);
489 		lf_printlist("lf_setlock", lock);
490 	}
491 #endif /* LOCKF_DEBUG */
492 	return 0;
493 }
494 
495 /*
496  * Remove a byte-range lock on an inode.
497  *
498  * Generally, find the lock (or an overlap to that lock)
499  * and remove it (or shrink it), then wakeup anyone we can.
500  */
501 static int
502 lf_clearlock(struct lockf *unlock, struct lockf **sparelock)
503 {
504 	struct lockf **head = unlock->lf_head;
505 	struct lockf *lf = *head;
506 	struct lockf *overlap, **prev;
507 	int ovcase;
508 
509 	if (lf == NOLOCKF)
510 		return 0;
511 #ifdef LOCKF_DEBUG
512 	if (unlock->lf_type != F_UNLCK)
513 		panic("lf_clearlock: bad type");
514 	if (lockf_debug & 1)
515 		lf_print("lf_clearlock", unlock);
516 #endif /* LOCKF_DEBUG */
517 	prev = head;
518 	while ((ovcase = lf_findoverlap(lf, unlock, SELF,
519 					&prev, &overlap)) != 0) {
520 		/*
521 		 * Wakeup the list of locks to be retried.
522 		 */
523 		lf_wakelock(overlap);
524 
525 		switch (ovcase) {
526 
527 		case 1: /* overlap == lock */
528 			*prev = overlap->lf_next;
529 			FREE(overlap, M_LOCKF);
530 			break;
531 
532 		case 2: /* overlap contains lock: split it */
533 			if (overlap->lf_start == unlock->lf_start) {
534 				overlap->lf_start = unlock->lf_end + 1;
535 				break;
536 			}
537 			lf_split(overlap, unlock, sparelock);
538 			overlap->lf_next = unlock->lf_next;
539 			break;
540 
541 		case 3: /* lock contains overlap */
542 			*prev = overlap->lf_next;
543 			lf = overlap->lf_next;
544 			free(overlap, M_LOCKF);
545 			continue;
546 
547 		case 4: /* overlap starts before lock */
548 			overlap->lf_end = unlock->lf_start - 1;
549 			prev = &overlap->lf_next;
550 			lf = overlap->lf_next;
551 			continue;
552 
553 		case 5: /* overlap ends after lock */
554 			overlap->lf_start = unlock->lf_end + 1;
555 			break;
556 		}
557 		break;
558 	}
559 #ifdef LOCKF_DEBUG
560 	if (lockf_debug & 1)
561 		lf_printlist("lf_clearlock", unlock);
562 #endif /* LOCKF_DEBUG */
563 	return 0;
564 }
565 
566 /*
567  * Check whether there is a blocking lock,
568  * and if so return its process identifier.
569  */
570 static int
571 lf_getlock(struct lockf *lock, struct flock *fl)
572 {
573 	struct lockf *block;
574 
575 #ifdef LOCKF_DEBUG
576 	if (lockf_debug & 1)
577 		lf_print("lf_getlock", lock);
578 #endif /* LOCKF_DEBUG */
579 
580 	if ((block = lf_getblock(lock)) != NULL) {
581 		fl->l_type = block->lf_type;
582 		fl->l_whence = SEEK_SET;
583 		fl->l_start = block->lf_start;
584 		if (block->lf_end == -1)
585 			fl->l_len = 0;
586 		else
587 			fl->l_len = block->lf_end - block->lf_start + 1;
588 		if (block->lf_flags & F_POSIX)
589 			fl->l_pid = ((struct proc *)block->lf_id)->p_pid;
590 		else
591 			fl->l_pid = -1;
592 	} else {
593 		fl->l_type = F_UNLCK;
594 	}
595 	return 0;
596 }
597 
598 /*
599  * Walk the list of locks for an inode and
600  * return the first blocking lock.
601  */
602 static struct lockf *
603 lf_getblock(struct lockf *lock)
604 {
605 	struct lockf **prev, *overlap, *lf = *(lock->lf_head);
606 
607 	prev = lock->lf_head;
608 	while (lf_findoverlap(lf, lock, OTHERS, &prev, &overlap) != 0) {
609 		/*
610 		 * We've found an overlap, see if it blocks us
611 		 */
612 		if ((lock->lf_type == F_WRLCK || overlap->lf_type == F_WRLCK))
613 			return overlap;
614 		/*
615 		 * Nope, point to the next one on the list and
616 		 * see if it blocks us
617 		 */
618 		lf = overlap->lf_next;
619 	}
620 	return NOLOCKF;
621 }
622 
623 /*
624  * Walk the list of locks for an inode to
625  * find an overlapping lock (if any).
626  *
627  * NOTE: this returns only the FIRST overlapping lock.  There
628  *	 may be more than one.
629  */
630 static int
631 lf_findoverlap(struct lockf *lf, struct lockf *lock, int type,
632     struct lockf ***prev, struct lockf **overlap)
633 {
634 	off_t start, end;
635 
636 	*overlap = lf;
637 	if (lf == NOLOCKF)
638 		return 0;
639 #ifdef LOCKF_DEBUG
640 	if (lockf_debug & 2)
641 		lf_print("lf_findoverlap: looking for overlap in", lock);
642 #endif /* LOCKF_DEBUG */
643 	start = lock->lf_start;
644 	end = lock->lf_end;
645 	while (lf != NOLOCKF) {
646 		if (((type == SELF) && lf->lf_id != lock->lf_id) ||
647 		    ((type == OTHERS) && lf->lf_id == lock->lf_id)) {
648 			*prev = &lf->lf_next;
649 			*overlap = lf = lf->lf_next;
650 			continue;
651 		}
652 #ifdef LOCKF_DEBUG
653 		if (lockf_debug & 2)
654 			lf_print("\tchecking", lf);
655 #endif /* LOCKF_DEBUG */
656 		/*
657 		 * OK, check for overlap
658 		 *
659 		 * Six cases:
660 		 *	0) no overlap
661 		 *	1) overlap == lock
662 		 *	2) overlap contains lock
663 		 *	3) lock contains overlap
664 		 *	4) overlap starts before lock
665 		 *	5) overlap ends after lock
666 		 */
667 		if ((lf->lf_end != -1 && start > lf->lf_end) ||
668 		    (end != -1 && lf->lf_start > end)) {
669 			/* Case 0 */
670 #ifdef LOCKF_DEBUG
671 			if (lockf_debug & 2)
672 				printf("no overlap\n");
673 #endif /* LOCKF_DEBUG */
674 			if ((type & SELF) && end != -1 && lf->lf_start > end)
675 				return 0;
676 			*prev = &lf->lf_next;
677 			*overlap = lf = lf->lf_next;
678 			continue;
679 		}
680 		if ((lf->lf_start == start) && (lf->lf_end == end)) {
681 			/* Case 1 */
682 #ifdef LOCKF_DEBUG
683 			if (lockf_debug & 2)
684 				printf("overlap == lock\n");
685 #endif /* LOCKF_DEBUG */
686 			return 1;
687 		}
688 		if ((lf->lf_start <= start) &&
689 		    (end != -1) &&
690 		    ((lf->lf_end >= end) || (lf->lf_end == -1))) {
691 			/* Case 2 */
692 #ifdef LOCKF_DEBUG
693 			if (lockf_debug & 2)
694 				printf("overlap contains lock\n");
695 #endif /* LOCKF_DEBUG */
696 			return 2;
697 		}
698 		if (start <= lf->lf_start &&
699 		           (end == -1 ||
700 			   (lf->lf_end != -1 && end >= lf->lf_end))) {
701 			/* Case 3 */
702 #ifdef LOCKF_DEBUG
703 			if (lockf_debug & 2)
704 				printf("lock contains overlap\n");
705 #endif /* LOCKF_DEBUG */
706 			return 3;
707 		}
708 		if ((lf->lf_start < start) &&
709 			((lf->lf_end >= start) || (lf->lf_end == -1))) {
710 			/* Case 4 */
711 #ifdef LOCKF_DEBUG
712 			if (lockf_debug & 2)
713 				printf("overlap starts before lock\n");
714 #endif /* LOCKF_DEBUG */
715 			return 4;
716 		}
717 		if ((lf->lf_start > start) &&
718 			(end != -1) &&
719 			((lf->lf_end > end) || (lf->lf_end == -1))) {
720 			/* Case 5 */
721 #ifdef LOCKF_DEBUG
722 			if (lockf_debug & 2)
723 				printf("overlap ends after lock\n");
724 #endif /* LOCKF_DEBUG */
725 			return 5;
726 		}
727 		panic("lf_findoverlap: default");
728 	}
729 	return 0;
730 }
731 
732 /*
733  * Split a lock and a contained region into
734  * two or three locks as necessary.
735  */
736 static void
737 lf_split(struct lockf *lock1, struct lockf *lock2, struct lockf **sparelock)
738 {
739 	struct lockf *splitlock;
740 
741 #ifdef LOCKF_DEBUG
742 	if (lockf_debug & 2) {
743 		lf_print("lf_split", lock1);
744 		lf_print("splitting from", lock2);
745 	}
746 #endif /* LOCKF_DEBUG */
747 	/*
748 	 * Check to see if spliting into only two pieces.
749 	 */
750 	if (lock1->lf_start == lock2->lf_start) {
751 		lock1->lf_start = lock2->lf_end + 1;
752 		lock2->lf_next = lock1;
753 		return;
754 	}
755 	if (lock1->lf_end == lock2->lf_end) {
756 		lock1->lf_end = lock2->lf_start - 1;
757 		lock2->lf_next = lock1->lf_next;
758 		lock1->lf_next = lock2;
759 		return;
760 	}
761 	/*
762 	 * Make a new lock consisting of the last part of
763 	 * the encompassing lock
764 	 */
765 	splitlock = *sparelock;
766 	*sparelock = NULL;
767 	memcpy(splitlock, lock1, sizeof(*splitlock));
768 	splitlock->lf_start = lock2->lf_end + 1;
769 	TAILQ_INIT(&splitlock->lf_blkhd);
770 	lock1->lf_end = lock2->lf_start - 1;
771 	/*
772 	 * OK, now link it in
773 	 */
774 	splitlock->lf_next = lock1->lf_next;
775 	lock2->lf_next = splitlock;
776 	lock1->lf_next = lock2;
777 }
778 
779 /*
780  * Wakeup a blocklist
781  */
782 static void
783 lf_wakelock(struct lockf *listhead)
784 {
785 	struct lockf *wakelock;
786 
787 	while ((wakelock = TAILQ_FIRST(&listhead->lf_blkhd))) {
788 		KASSERT(wakelock->lf_next == listhead);
789 		TAILQ_REMOVE(&listhead->lf_blkhd, wakelock, lf_block);
790 		wakelock->lf_next = NOLOCKF;
791 #ifdef LOCKF_DEBUG
792 		if (lockf_debug & 2)
793 			lf_print("lf_wakelock: awakening", wakelock);
794 #endif
795 		wakeup(wakelock);
796 	}
797 }
798 
799 #ifdef LOCKF_DEBUG
800 /*
801  * Print out a lock.
802  */
803 static void
804 lf_print(char *tag, struct lockf *lock)
805 {
806 
807 	printf("%s: lock %p for ", tag, lock);
808 	if (lock->lf_flags & F_POSIX)
809 		printf("proc %d", ((struct proc *)lock->lf_id)->p_pid);
810 	else
811 		printf("file 0x%p", (struct file *)lock->lf_id);
812 	printf(" %s, start %qx, end %qx",
813 		lock->lf_type == F_RDLCK ? "shared" :
814 		lock->lf_type == F_WRLCK ? "exclusive" :
815 		lock->lf_type == F_UNLCK ? "unlock" :
816 		"unknown", lock->lf_start, lock->lf_end);
817 	if (TAILQ_FIRST(&lock->lf_blkhd))
818 		printf(" block %p\n", TAILQ_FIRST(&lock->lf_blkhd));
819 	else
820 		printf("\n");
821 }
822 
823 static void
824 lf_printlist(char *tag, struct lockf *lock)
825 {
826 	struct lockf *lf, *blk;
827 
828 	printf("%s: Lock list:\n", tag);
829 	for (lf = *lock->lf_head; lf; lf = lf->lf_next) {
830 		printf("\tlock %p for ", lf);
831 		if (lf->lf_flags & F_POSIX)
832 			printf("proc %d", ((struct proc *)lf->lf_id)->p_pid);
833 		else
834 			printf("file 0x%p", (struct file *)lf->lf_id);
835 		printf(", %s, start %qx, end %qx",
836 			lf->lf_type == F_RDLCK ? "shared" :
837 			lf->lf_type == F_WRLCK ? "exclusive" :
838 			lf->lf_type == F_UNLCK ? "unlock" :
839 			"unknown", lf->lf_start, lf->lf_end);
840 		TAILQ_FOREACH(blk, &lf->lf_blkhd, lf_block) {
841 			if (blk->lf_flags & F_POSIX)
842 				printf("proc %d",
843 				    ((struct proc *)blk->lf_id)->p_pid);
844 			else
845 				printf("file 0x%p", (struct file *)blk->lf_id);
846 			printf(", %s, start %qx, end %qx",
847 				blk->lf_type == F_RDLCK ? "shared" :
848 				blk->lf_type == F_WRLCK ? "exclusive" :
849 				blk->lf_type == F_UNLCK ? "unlock" :
850 				"unknown", blk->lf_start, blk->lf_end);
851 			if (TAILQ_FIRST(&blk->lf_blkhd))
852 				 panic("lf_printlist: bad list");
853 		}
854 		printf("\n");
855 	}
856 }
857 #endif /* LOCKF_DEBUG */
858