xref: /netbsd-src/sys/kern/vfs_lockf.c (revision 23c8222edbfb0f0932d88a8351d3a0cf817dfb9e)
1 /*	$NetBSD: vfs_lockf.c,v 1.35 2004/04/25 16:42:41 simonb 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 1.35 2004/04/25 16:42:41 simonb 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/pool.h>
47 #include <sys/fcntl.h>
48 #include <sys/lockf.h>
49 
50 POOL_INIT(lockfpool, sizeof(struct lockf), 0, 0, 0, "lockfpl",
51     &pool_allocator_nointr);
52 
53 /*
54  * This variable controls the maximum number of processes that will
55  * be checked in doing deadlock detection.
56  */
57 int maxlockdepth = MAXDEPTH;
58 
59 #ifdef LOCKF_DEBUG
60 int	lockf_debug = 0;
61 #endif
62 
63 #define NOLOCKF (struct lockf *)0
64 #define SELF	0x1
65 #define OTHERS	0x2
66 
67 static int lf_clearlock(struct lockf *, struct lockf **);
68 static int lf_findoverlap(struct lockf *,
69 	    struct lockf *, int, struct lockf ***, struct lockf **);
70 static struct lockf *lf_getblock(struct lockf *);
71 static int lf_getlock(struct lockf *, struct flock *);
72 static int lf_setlock(struct lockf *, struct lockf **, struct simplelock *);
73 static void lf_split(struct lockf *, struct lockf *, struct lockf **);
74 static void lf_wakelock(struct lockf *);
75 
76 #ifdef LOCKF_DEBUG
77 static void lf_print(char *, struct lockf *);
78 static void lf_printlist(char *, struct lockf *);
79 #endif
80 
81 /*
82  * XXX TODO
83  * Misc cleanups: "caddr_t id" should be visible in the API as a
84  * "struct proc *".
85  * (This requires rototilling all VFS's which support advisory locking).
86  */
87 
88 /*
89  * If there's a lot of lock contention on a single vnode, locking
90  * schemes which allow for more paralleism would be needed.  Given how
91  * infrequently byte-range locks are actually used in typical BSD
92  * code, a more complex approach probably isn't worth it.
93  */
94 
95 /*
96  * Do an advisory lock operation.
97  */
98 int
99 lf_advlock(struct vop_advlock_args *ap, struct lockf **head, off_t size)
100 {
101 	struct flock *fl = ap->a_fl;
102 	struct lockf *lock = NULL;
103 	struct lockf *sparelock;
104 	struct simplelock *interlock = &ap->a_vp->v_interlock;
105 	off_t start, end;
106 	int error = 0;
107 
108 	/*
109 	 * Convert the flock structure into a start and end.
110 	 */
111 	switch (fl->l_whence) {
112 	case SEEK_SET:
113 	case SEEK_CUR:
114 		/*
115 		 * Caller is responsible for adding any necessary offset
116 		 * when SEEK_CUR is used.
117 		 */
118 		start = fl->l_start;
119 		break;
120 
121 	case SEEK_END:
122 		start = size + fl->l_start;
123 		break;
124 
125 	default:
126 		return EINVAL;
127 	}
128 	if (start < 0)
129 		return EINVAL;
130 
131 	/*
132 	 * allocate locks before acquire simple lock.
133 	 * we need two locks in the worst case.
134 	 */
135 	switch (ap->a_op) {
136 	case F_SETLK:
137 	case F_UNLCK:
138 		/*
139 		 * XXX for F_UNLCK case, we can re-use lock.
140 		 */
141 		if ((fl->l_type & F_FLOCK) == 0) {
142 			/*
143 			 * byte-range lock might need one more lock.
144 			 */
145 			sparelock = pool_get(&lockfpool, PR_WAITOK);
146 			if (sparelock == NULL) {
147 				error = ENOMEM;
148 				goto quit;
149 			}
150 			break;
151 		}
152 		/* FALLTHROUGH */
153 
154 	case F_GETLK:
155 		sparelock = NULL;
156 		break;
157 
158 	default:
159 		return EINVAL;
160 	}
161 
162 	lock = pool_get(&lockfpool, PR_WAITOK);
163 	if (lock == NULL) {
164 		error = ENOMEM;
165 		goto quit;
166 	}
167 
168 	simple_lock(interlock);
169 
170 	/*
171 	 * Avoid the common case of unlocking when inode has no locks.
172 	 */
173 	if (*head == (struct lockf *)0) {
174 		if (ap->a_op != F_SETLK) {
175 			fl->l_type = F_UNLCK;
176 			error = 0;
177 			goto quit_unlock;
178 		}
179 	}
180 
181 	if (fl->l_len == 0)
182 		end = -1;
183 	else
184 		end = start + fl->l_len - 1;
185 	/*
186 	 * Create the lockf structure.
187 	 */
188 	lock->lf_start = start;
189 	lock->lf_end = end;
190 	/* XXX NJWLWP
191 	 * I don't want to make the entire VFS universe use LWPs, because
192 	 * they don't need them, for the most part. This is an exception,
193 	 * and a kluge.
194 	 */
195 
196 	lock->lf_head = head;
197 	lock->lf_type = fl->l_type;
198 	lock->lf_next = (struct lockf *)0;
199 	TAILQ_INIT(&lock->lf_blkhd);
200 	lock->lf_flags = ap->a_flags;
201 	if (lock->lf_flags & F_POSIX) {
202 		KASSERT(curproc == (struct proc *)ap->a_id);
203 	}
204 	lock->lf_id = (struct proc *)ap->a_id;
205 	lock->lf_lwp = curlwp;
206 
207 	/*
208 	 * Do the requested operation.
209 	 */
210 	switch (ap->a_op) {
211 
212 	case F_SETLK:
213 		error = lf_setlock(lock, &sparelock, interlock);
214 		lock = NULL; /* lf_setlock freed it */
215 		break;
216 
217 	case F_UNLCK:
218 		error = lf_clearlock(lock, &sparelock);
219 		break;
220 
221 	case F_GETLK:
222 		error = lf_getlock(lock, fl);
223 		break;
224 
225 	default:
226 		break;
227 		/* NOTREACHED */
228 	}
229 
230 quit_unlock:
231 	simple_unlock(interlock);
232 quit:
233 	if (lock)
234 		pool_put(&lockfpool, lock);
235 	if (sparelock)
236 		pool_put(&lockfpool, sparelock);
237 
238 	return error;
239 }
240 
241 /*
242  * Set a byte-range lock.
243  */
244 static int
245 lf_setlock(struct lockf *lock, struct lockf **sparelock,
246     struct simplelock *interlock)
247 {
248 	struct lockf *block;
249 	struct lockf **head = lock->lf_head;
250 	struct lockf **prev, *overlap, *ltmp;
251 	static char lockstr[] = "lockf";
252 	int ovcase, priority, needtolink, error;
253 
254 #ifdef LOCKF_DEBUG
255 	if (lockf_debug & 1)
256 		lf_print("lf_setlock", lock);
257 #endif /* LOCKF_DEBUG */
258 
259 	/*
260 	 * Set the priority
261 	 */
262 	priority = PLOCK;
263 	if (lock->lf_type == F_WRLCK)
264 		priority += 4;
265 	priority |= PCATCH;
266 	/*
267 	 * Scan lock list for this file looking for locks that would block us.
268 	 */
269 	while ((block = lf_getblock(lock)) != NULL) {
270 		/*
271 		 * Free the structure and return if nonblocking.
272 		 */
273 		if ((lock->lf_flags & F_WAIT) == 0) {
274 			pool_put(&lockfpool, lock);
275 			return EAGAIN;
276 		}
277 		/*
278 		 * We are blocked. Since flock style locks cover
279 		 * the whole file, there is no chance for deadlock.
280 		 * For byte-range locks we must check for deadlock.
281 		 *
282 		 * Deadlock detection is done by looking through the
283 		 * wait channels to see if there are any cycles that
284 		 * involve us. MAXDEPTH is set just to make sure we
285 		 * do not go off into neverneverland.
286 		 */
287 		if ((lock->lf_flags & F_POSIX) &&
288 		    (block->lf_flags & F_POSIX)) {
289 			struct lwp *wlwp;
290 			struct lockf *waitblock;
291 			int i = 0;
292 
293 			/*
294 			 * The block is waiting on something.  if_lwp will be
295 			 * 0 once the lock is granted, so we terminate the
296 			 * loop if we find this.
297 			 */
298 			wlwp = block->lf_lwp;
299 			while (wlwp && (i++ < maxlockdepth)) {
300 				waitblock = (struct lockf *)wlwp->l_wchan;
301 				/* Get the owner of the blocking lock */
302 				waitblock = waitblock->lf_next;
303 				if ((waitblock->lf_flags & F_POSIX) == 0)
304 					break;
305 				wlwp = waitblock->lf_lwp;
306 				if (wlwp == lock->lf_lwp) {
307 					pool_put(&lockfpool, lock);
308 					return EDEADLK;
309 				}
310 			}
311 			/*
312 			 * If we're still following a dependancy chain
313 			 * after maxlockdepth iterations, assume we're in
314 			 * a cycle to be safe.
315 			 */
316 			if (i >= maxlockdepth) {
317 				pool_put(&lockfpool, lock);
318 				return EDEADLK;
319 			}
320 		}
321 		/*
322 		 * For flock type locks, we must first remove
323 		 * any shared locks that we hold before we sleep
324 		 * waiting for an exclusive lock.
325 		 */
326 		if ((lock->lf_flags & F_FLOCK) &&
327 		    lock->lf_type == F_WRLCK) {
328 			lock->lf_type = F_UNLCK;
329 			(void) lf_clearlock(lock, NULL);
330 			lock->lf_type = F_WRLCK;
331 		}
332 		/*
333 		 * Add our lock to the blocked list and sleep until we're free.
334 		 * Remember who blocked us (for deadlock detection).
335 		 */
336 		lock->lf_next = block;
337 		TAILQ_INSERT_TAIL(&block->lf_blkhd, lock, lf_block);
338 #ifdef LOCKF_DEBUG
339 		if (lockf_debug & 1) {
340 			lf_print("lf_setlock: blocking on", block);
341 			lf_printlist("lf_setlock", block);
342 		}
343 #endif /* LOCKF_DEBUG */
344 		error = ltsleep(lock, priority, lockstr, 0, interlock);
345 
346 		/*
347 		 * We may have been awakened by a signal (in
348 		 * which case we must remove ourselves from the
349 		 * blocked list) and/or by another process
350 		 * releasing a lock (in which case we have already
351 		 * been removed from the blocked list and our
352 		 * lf_next field set to NOLOCKF).
353 		 */
354 		if (lock->lf_next != NOLOCKF) {
355 			TAILQ_REMOVE(&lock->lf_next->lf_blkhd, lock, lf_block);
356 			lock->lf_next = NOLOCKF;
357 		}
358 		if (error) {
359 			pool_put(&lockfpool, lock);
360 			return error;
361 		}
362 	}
363 	/*
364 	 * No blocks!!  Add the lock.  Note that we will
365 	 * downgrade or upgrade any overlapping locks this
366 	 * process already owns.
367 	 *
368 	 * Skip over locks owned by other processes.
369 	 * Handle any locks that overlap and are owned by ourselves.
370 	 */
371 	lock->lf_lwp = 0;
372 	prev = head;
373 	block = *head;
374 	needtolink = 1;
375 	for (;;) {
376 		ovcase = lf_findoverlap(block, lock, SELF, &prev, &overlap);
377 		if (ovcase)
378 			block = overlap->lf_next;
379 		/*
380 		 * Six cases:
381 		 *	0) no overlap
382 		 *	1) overlap == lock
383 		 *	2) overlap contains lock
384 		 *	3) lock contains overlap
385 		 *	4) overlap starts before lock
386 		 *	5) overlap ends after lock
387 		 */
388 		switch (ovcase) {
389 		case 0: /* no overlap */
390 			if (needtolink) {
391 				*prev = lock;
392 				lock->lf_next = overlap;
393 			}
394 			break;
395 
396 		case 1: /* overlap == lock */
397 			/*
398 			 * If downgrading lock, others may be
399 			 * able to acquire it.
400 			 */
401 			if (lock->lf_type == F_RDLCK &&
402 			    overlap->lf_type == F_WRLCK)
403 				lf_wakelock(overlap);
404 			overlap->lf_type = lock->lf_type;
405 			pool_put(&lockfpool, lock);
406 			lock = overlap; /* for debug output below */
407 			break;
408 
409 		case 2: /* overlap contains lock */
410 			/*
411 			 * Check for common starting point and different types.
412 			 */
413 			if (overlap->lf_type == lock->lf_type) {
414 				pool_put(&lockfpool, lock);
415 				lock = overlap; /* for debug output below */
416 				break;
417 			}
418 			if (overlap->lf_start == lock->lf_start) {
419 				*prev = lock;
420 				lock->lf_next = overlap;
421 				overlap->lf_start = lock->lf_end + 1;
422 			} else
423 				lf_split(overlap, lock, sparelock);
424 			lf_wakelock(overlap);
425 			break;
426 
427 		case 3: /* lock contains overlap */
428 			/*
429 			 * If downgrading lock, others may be able to
430 			 * acquire it, otherwise take the list.
431 			 */
432 			if (lock->lf_type == F_RDLCK &&
433 			    overlap->lf_type == F_WRLCK) {
434 				lf_wakelock(overlap);
435 			} else {
436 				while ((ltmp = TAILQ_FIRST(&overlap->lf_blkhd))) {
437 					KASSERT(ltmp->lf_next == overlap);
438 					TAILQ_REMOVE(&overlap->lf_blkhd, ltmp,
439 					    lf_block);
440 					ltmp->lf_next = lock;
441 					TAILQ_INSERT_TAIL(&lock->lf_blkhd,
442 					    ltmp, lf_block);
443 				}
444 			}
445 			/*
446 			 * Add the new lock if necessary and delete the overlap.
447 			 */
448 			if (needtolink) {
449 				*prev = lock;
450 				lock->lf_next = overlap->lf_next;
451 				prev = &lock->lf_next;
452 				needtolink = 0;
453 			} else
454 				*prev = overlap->lf_next;
455 			pool_put(&lockfpool, overlap);
456 			continue;
457 
458 		case 4: /* overlap starts before lock */
459 			/*
460 			 * Add lock after overlap on the list.
461 			 */
462 			lock->lf_next = overlap->lf_next;
463 			overlap->lf_next = lock;
464 			overlap->lf_end = lock->lf_start - 1;
465 			prev = &lock->lf_next;
466 			lf_wakelock(overlap);
467 			needtolink = 0;
468 			continue;
469 
470 		case 5: /* overlap ends after lock */
471 			/*
472 			 * Add the new lock before overlap.
473 			 */
474 			if (needtolink) {
475 				*prev = lock;
476 				lock->lf_next = overlap;
477 			}
478 			overlap->lf_start = lock->lf_end + 1;
479 			lf_wakelock(overlap);
480 			break;
481 		}
482 		break;
483 	}
484 #ifdef LOCKF_DEBUG
485 	if (lockf_debug & 1) {
486 		lf_print("lf_setlock: got the lock", lock);
487 		lf_printlist("lf_setlock", lock);
488 	}
489 #endif /* LOCKF_DEBUG */
490 	return 0;
491 }
492 
493 /*
494  * Remove a byte-range lock on an inode.
495  *
496  * Generally, find the lock (or an overlap to that lock)
497  * and remove it (or shrink it), then wakeup anyone we can.
498  */
499 static int
500 lf_clearlock(struct lockf *unlock, struct lockf **sparelock)
501 {
502 	struct lockf **head = unlock->lf_head;
503 	struct lockf *lf = *head;
504 	struct lockf *overlap, **prev;
505 	int ovcase;
506 
507 	if (lf == NOLOCKF)
508 		return 0;
509 #ifdef LOCKF_DEBUG
510 	if (unlock->lf_type != F_UNLCK)
511 		panic("lf_clearlock: bad type");
512 	if (lockf_debug & 1)
513 		lf_print("lf_clearlock", unlock);
514 #endif /* LOCKF_DEBUG */
515 	prev = head;
516 	while ((ovcase = lf_findoverlap(lf, unlock, SELF,
517 					&prev, &overlap)) != 0) {
518 		/*
519 		 * Wakeup the list of locks to be retried.
520 		 */
521 		lf_wakelock(overlap);
522 
523 		switch (ovcase) {
524 
525 		case 1: /* overlap == lock */
526 			*prev = overlap->lf_next;
527 			pool_put(&lockfpool, overlap);
528 			break;
529 
530 		case 2: /* overlap contains lock: split it */
531 			if (overlap->lf_start == unlock->lf_start) {
532 				overlap->lf_start = unlock->lf_end + 1;
533 				break;
534 			}
535 			lf_split(overlap, unlock, sparelock);
536 			overlap->lf_next = unlock->lf_next;
537 			break;
538 
539 		case 3: /* lock contains overlap */
540 			*prev = overlap->lf_next;
541 			lf = overlap->lf_next;
542 			pool_put(&lockfpool, overlap);
543 			continue;
544 
545 		case 4: /* overlap starts before lock */
546 			overlap->lf_end = unlock->lf_start - 1;
547 			prev = &overlap->lf_next;
548 			lf = overlap->lf_next;
549 			continue;
550 
551 		case 5: /* overlap ends after lock */
552 			overlap->lf_start = unlock->lf_end + 1;
553 			break;
554 		}
555 		break;
556 	}
557 #ifdef LOCKF_DEBUG
558 	if (lockf_debug & 1)
559 		lf_printlist("lf_clearlock", unlock);
560 #endif /* LOCKF_DEBUG */
561 	return 0;
562 }
563 
564 /*
565  * Check whether there is a blocking lock,
566  * and if so return its process identifier.
567  */
568 static int
569 lf_getlock(struct lockf *lock, struct flock *fl)
570 {
571 	struct lockf *block;
572 
573 #ifdef LOCKF_DEBUG
574 	if (lockf_debug & 1)
575 		lf_print("lf_getlock", lock);
576 #endif /* LOCKF_DEBUG */
577 
578 	if ((block = lf_getblock(lock)) != NULL) {
579 		fl->l_type = block->lf_type;
580 		fl->l_whence = SEEK_SET;
581 		fl->l_start = block->lf_start;
582 		if (block->lf_end == -1)
583 			fl->l_len = 0;
584 		else
585 			fl->l_len = block->lf_end - block->lf_start + 1;
586 		if (block->lf_flags & F_POSIX)
587 			fl->l_pid = ((struct proc *)block->lf_id)->p_pid;
588 		else
589 			fl->l_pid = -1;
590 	} else {
591 		fl->l_type = F_UNLCK;
592 	}
593 	return 0;
594 }
595 
596 /*
597  * Walk the list of locks for an inode and
598  * return the first blocking lock.
599  */
600 static struct lockf *
601 lf_getblock(struct lockf *lock)
602 {
603 	struct lockf **prev, *overlap, *lf = *(lock->lf_head);
604 
605 	prev = lock->lf_head;
606 	while (lf_findoverlap(lf, lock, OTHERS, &prev, &overlap) != 0) {
607 		/*
608 		 * We've found an overlap, see if it blocks us
609 		 */
610 		if ((lock->lf_type == F_WRLCK || overlap->lf_type == F_WRLCK))
611 			return overlap;
612 		/*
613 		 * Nope, point to the next one on the list and
614 		 * see if it blocks us
615 		 */
616 		lf = overlap->lf_next;
617 	}
618 	return NOLOCKF;
619 }
620 
621 /*
622  * Walk the list of locks for an inode to
623  * find an overlapping lock (if any).
624  *
625  * NOTE: this returns only the FIRST overlapping lock.  There
626  *	 may be more than one.
627  */
628 static int
629 lf_findoverlap(struct lockf *lf, struct lockf *lock, int type,
630     struct lockf ***prev, struct lockf **overlap)
631 {
632 	off_t start, end;
633 
634 	*overlap = lf;
635 	if (lf == NOLOCKF)
636 		return 0;
637 #ifdef LOCKF_DEBUG
638 	if (lockf_debug & 2)
639 		lf_print("lf_findoverlap: looking for overlap in", lock);
640 #endif /* LOCKF_DEBUG */
641 	start = lock->lf_start;
642 	end = lock->lf_end;
643 	while (lf != NOLOCKF) {
644 		if (((type == SELF) && lf->lf_id != lock->lf_id) ||
645 		    ((type == OTHERS) && lf->lf_id == lock->lf_id)) {
646 			*prev = &lf->lf_next;
647 			*overlap = lf = lf->lf_next;
648 			continue;
649 		}
650 #ifdef LOCKF_DEBUG
651 		if (lockf_debug & 2)
652 			lf_print("\tchecking", lf);
653 #endif /* LOCKF_DEBUG */
654 		/*
655 		 * OK, check for overlap
656 		 *
657 		 * Six cases:
658 		 *	0) no overlap
659 		 *	1) overlap == lock
660 		 *	2) overlap contains lock
661 		 *	3) lock contains overlap
662 		 *	4) overlap starts before lock
663 		 *	5) overlap ends after lock
664 		 */
665 		if ((lf->lf_end != -1 && start > lf->lf_end) ||
666 		    (end != -1 && lf->lf_start > end)) {
667 			/* Case 0 */
668 #ifdef LOCKF_DEBUG
669 			if (lockf_debug & 2)
670 				printf("no overlap\n");
671 #endif /* LOCKF_DEBUG */
672 			if ((type & SELF) && end != -1 && lf->lf_start > end)
673 				return 0;
674 			*prev = &lf->lf_next;
675 			*overlap = lf = lf->lf_next;
676 			continue;
677 		}
678 		if ((lf->lf_start == start) && (lf->lf_end == end)) {
679 			/* Case 1 */
680 #ifdef LOCKF_DEBUG
681 			if (lockf_debug & 2)
682 				printf("overlap == lock\n");
683 #endif /* LOCKF_DEBUG */
684 			return 1;
685 		}
686 		if ((lf->lf_start <= start) &&
687 		    (end != -1) &&
688 		    ((lf->lf_end >= end) || (lf->lf_end == -1))) {
689 			/* Case 2 */
690 #ifdef LOCKF_DEBUG
691 			if (lockf_debug & 2)
692 				printf("overlap contains lock\n");
693 #endif /* LOCKF_DEBUG */
694 			return 2;
695 		}
696 		if (start <= lf->lf_start &&
697 		           (end == -1 ||
698 			   (lf->lf_end != -1 && end >= lf->lf_end))) {
699 			/* Case 3 */
700 #ifdef LOCKF_DEBUG
701 			if (lockf_debug & 2)
702 				printf("lock contains overlap\n");
703 #endif /* LOCKF_DEBUG */
704 			return 3;
705 		}
706 		if ((lf->lf_start < start) &&
707 			((lf->lf_end >= start) || (lf->lf_end == -1))) {
708 			/* Case 4 */
709 #ifdef LOCKF_DEBUG
710 			if (lockf_debug & 2)
711 				printf("overlap starts before lock\n");
712 #endif /* LOCKF_DEBUG */
713 			return 4;
714 		}
715 		if ((lf->lf_start > start) &&
716 			(end != -1) &&
717 			((lf->lf_end > end) || (lf->lf_end == -1))) {
718 			/* Case 5 */
719 #ifdef LOCKF_DEBUG
720 			if (lockf_debug & 2)
721 				printf("overlap ends after lock\n");
722 #endif /* LOCKF_DEBUG */
723 			return 5;
724 		}
725 		panic("lf_findoverlap: default");
726 	}
727 	return 0;
728 }
729 
730 /*
731  * Split a lock and a contained region into
732  * two or three locks as necessary.
733  */
734 static void
735 lf_split(struct lockf *lock1, struct lockf *lock2, struct lockf **sparelock)
736 {
737 	struct lockf *splitlock;
738 
739 #ifdef LOCKF_DEBUG
740 	if (lockf_debug & 2) {
741 		lf_print("lf_split", lock1);
742 		lf_print("splitting from", lock2);
743 	}
744 #endif /* LOCKF_DEBUG */
745 	/*
746 	 * Check to see if spliting into only two pieces.
747 	 */
748 	if (lock1->lf_start == lock2->lf_start) {
749 		lock1->lf_start = lock2->lf_end + 1;
750 		lock2->lf_next = lock1;
751 		return;
752 	}
753 	if (lock1->lf_end == lock2->lf_end) {
754 		lock1->lf_end = lock2->lf_start - 1;
755 		lock2->lf_next = lock1->lf_next;
756 		lock1->lf_next = lock2;
757 		return;
758 	}
759 	/*
760 	 * Make a new lock consisting of the last part of
761 	 * the encompassing lock
762 	 */
763 	splitlock = *sparelock;
764 	*sparelock = NULL;
765 	memcpy(splitlock, lock1, sizeof(*splitlock));
766 	splitlock->lf_start = lock2->lf_end + 1;
767 	TAILQ_INIT(&splitlock->lf_blkhd);
768 	lock1->lf_end = lock2->lf_start - 1;
769 	/*
770 	 * OK, now link it in
771 	 */
772 	splitlock->lf_next = lock1->lf_next;
773 	lock2->lf_next = splitlock;
774 	lock1->lf_next = lock2;
775 }
776 
777 /*
778  * Wakeup a blocklist
779  */
780 static void
781 lf_wakelock(struct lockf *listhead)
782 {
783 	struct lockf *wakelock;
784 
785 	while ((wakelock = TAILQ_FIRST(&listhead->lf_blkhd))) {
786 		KASSERT(wakelock->lf_next == listhead);
787 		TAILQ_REMOVE(&listhead->lf_blkhd, wakelock, lf_block);
788 		wakelock->lf_next = NOLOCKF;
789 #ifdef LOCKF_DEBUG
790 		if (lockf_debug & 2)
791 			lf_print("lf_wakelock: awakening", wakelock);
792 #endif
793 		wakeup(wakelock);
794 	}
795 }
796 
797 #ifdef LOCKF_DEBUG
798 /*
799  * Print out a lock.
800  */
801 static void
802 lf_print(char *tag, struct lockf *lock)
803 {
804 
805 	printf("%s: lock %p for ", tag, lock);
806 	if (lock->lf_flags & F_POSIX)
807 		printf("proc %d", ((struct proc *)lock->lf_id)->p_pid);
808 	else
809 		printf("file 0x%p", (struct file *)lock->lf_id);
810 	printf(" %s, start %qx, end %qx",
811 		lock->lf_type == F_RDLCK ? "shared" :
812 		lock->lf_type == F_WRLCK ? "exclusive" :
813 		lock->lf_type == F_UNLCK ? "unlock" :
814 		"unknown", lock->lf_start, lock->lf_end);
815 	if (TAILQ_FIRST(&lock->lf_blkhd))
816 		printf(" block %p\n", TAILQ_FIRST(&lock->lf_blkhd));
817 	else
818 		printf("\n");
819 }
820 
821 static void
822 lf_printlist(char *tag, struct lockf *lock)
823 {
824 	struct lockf *lf, *blk;
825 
826 	printf("%s: Lock list:\n", tag);
827 	for (lf = *lock->lf_head; lf; lf = lf->lf_next) {
828 		printf("\tlock %p for ", lf);
829 		if (lf->lf_flags & F_POSIX)
830 			printf("proc %d", ((struct proc *)lf->lf_id)->p_pid);
831 		else
832 			printf("file 0x%p", (struct file *)lf->lf_id);
833 		printf(", %s, start %qx, end %qx",
834 			lf->lf_type == F_RDLCK ? "shared" :
835 			lf->lf_type == F_WRLCK ? "exclusive" :
836 			lf->lf_type == F_UNLCK ? "unlock" :
837 			"unknown", lf->lf_start, lf->lf_end);
838 		TAILQ_FOREACH(blk, &lf->lf_blkhd, lf_block) {
839 			if (blk->lf_flags & F_POSIX)
840 				printf("proc %d",
841 				    ((struct proc *)blk->lf_id)->p_pid);
842 			else
843 				printf("file 0x%p", (struct file *)blk->lf_id);
844 			printf(", %s, start %qx, end %qx",
845 				blk->lf_type == F_RDLCK ? "shared" :
846 				blk->lf_type == F_WRLCK ? "exclusive" :
847 				blk->lf_type == F_UNLCK ? "unlock" :
848 				"unknown", blk->lf_start, blk->lf_end);
849 			if (TAILQ_FIRST(&blk->lf_blkhd))
850 				 panic("lf_printlist: bad list");
851 		}
852 		printf("\n");
853 	}
854 }
855 #endif /* LOCKF_DEBUG */
856