xref: /netbsd-src/sys/kern/vfs_lockf.c (revision 404fbe5fb94ca1e054339640cabb2801ce52dd30)
1 /*	$NetBSD: vfs_lockf.c,v 1.69 2008/10/11 13:40:57 pooka 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 1.69 2008/10/11 13:40:57 pooka 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 #include <sys/atomic.h>
50 #include <sys/kauth.h>
51 #include <sys/uidinfo.h>
52 
53 /*
54  * The lockf structure is a kernel structure which contains the information
55  * associated with a byte range lock.  The lockf structures are linked into
56  * the vnode structure.  Locks are sorted by the starting byte of the lock for
57  * efficiency.
58  *
59  * lf_next is used for two purposes, depending on whether the lock is
60  * being held, or is in conflict with an existing lock.  If this lock
61  * is held, it indicates the next lock on the same vnode.
62  * For pending locks, if lock->lf_next is non-NULL, then lock->lf_block
63  * must be queued on the lf_blkhd TAILQ of lock->lf_next.
64  */
65 
66 TAILQ_HEAD(locklist, lockf);
67 
68 struct lockf {
69 	kcondvar_t lf_cv;	 /* Signalling */
70 	short	lf_flags;	 /* Lock semantics: F_POSIX, F_FLOCK, F_WAIT */
71 	short	lf_type;	 /* Lock type: F_RDLCK, F_WRLCK */
72 	off_t	lf_start;	 /* The byte # of the start of the lock */
73 	off_t	lf_end;		 /* The byte # of the end of the lock (-1=EOF)*/
74 	void	*lf_id;		 /* process or file description holding lock */
75 	struct	lockf **lf_head; /* Back pointer to the head of lockf list */
76 	struct	lockf *lf_next;	 /* Next lock on this vnode, or blocking lock */
77 	struct  locklist lf_blkhd; /* List of requests blocked on this lock */
78 	TAILQ_ENTRY(lockf) lf_block;/* A request waiting for a lock */
79 	uid_t	lf_uid;		 /* User ID responsible */
80 };
81 
82 /* Maximum length of sleep chains to traverse to try and detect deadlock. */
83 #define MAXDEPTH 50
84 
85 static pool_cache_t lockf_cache;
86 static kmutex_t *lockf_lock;
87 static char lockstr[] = "lockf";
88 
89 /*
90  * This variable controls the maximum number of processes that will
91  * be checked in doing deadlock detection.
92  */
93 int maxlockdepth = MAXDEPTH;
94 
95 #ifdef LOCKF_DEBUG
96 int	lockf_debug = 0;
97 #endif
98 
99 #define SELF	0x1
100 #define OTHERS	0x2
101 
102 /*
103  * XXX TODO
104  * Misc cleanups: "void *id" should be visible in the API as a
105  * "struct proc *".
106  * (This requires rototilling all VFS's which support advisory locking).
107  */
108 
109 /*
110  * If there's a lot of lock contention on a single vnode, locking
111  * schemes which allow for more paralleism would be needed.  Given how
112  * infrequently byte-range locks are actually used in typical BSD
113  * code, a more complex approach probably isn't worth it.
114  */
115 
116 /*
117  * We enforce a limit on locks by uid, so that a single user cannot
118  * run the kernel out of memory.  For now, the limit is pretty coarse.
119  * There is no limit on root.
120  *
121  * Splitting a lock will always succeed, regardless of current allocations.
122  * If you're slightly above the limit, we still have to permit an allocation
123  * so that the unlock can succeed.  If the unlocking causes too many splits,
124  * however, you're totally cutoff.
125  */
126 int maxlocksperuid = 1024;
127 
128 #ifdef LOCKF_DEBUG
129 /*
130  * Print out a lock.
131  */
132 static void
133 lf_print(const char *tag, struct lockf *lock)
134 {
135 
136 	printf("%s: lock %p for ", tag, lock);
137 	if (lock->lf_flags & F_POSIX)
138 		printf("proc %d", ((struct proc *)lock->lf_id)->p_pid);
139 	else
140 		printf("file %p", (struct file *)lock->lf_id);
141 	printf(" %s, start %qx, end %qx",
142 		lock->lf_type == F_RDLCK ? "shared" :
143 		lock->lf_type == F_WRLCK ? "exclusive" :
144 		lock->lf_type == F_UNLCK ? "unlock" :
145 		"unknown", lock->lf_start, lock->lf_end);
146 	if (TAILQ_FIRST(&lock->lf_blkhd))
147 		printf(" block %p\n", TAILQ_FIRST(&lock->lf_blkhd));
148 	else
149 		printf("\n");
150 }
151 
152 static void
153 lf_printlist(const char *tag, struct lockf *lock)
154 {
155 	struct lockf *lf, *blk;
156 
157 	printf("%s: Lock list:\n", tag);
158 	for (lf = *lock->lf_head; lf; lf = lf->lf_next) {
159 		printf("\tlock %p for ", lf);
160 		if (lf->lf_flags & F_POSIX)
161 			printf("proc %d", ((struct proc *)lf->lf_id)->p_pid);
162 		else
163 			printf("file %p", (struct file *)lf->lf_id);
164 		printf(", %s, start %qx, end %qx",
165 			lf->lf_type == F_RDLCK ? "shared" :
166 			lf->lf_type == F_WRLCK ? "exclusive" :
167 			lf->lf_type == F_UNLCK ? "unlock" :
168 			"unknown", lf->lf_start, lf->lf_end);
169 		TAILQ_FOREACH(blk, &lf->lf_blkhd, lf_block) {
170 			if (blk->lf_flags & F_POSIX)
171 				printf("; proc %d",
172 				    ((struct proc *)blk->lf_id)->p_pid);
173 			else
174 				printf("; file %p", (struct file *)blk->lf_id);
175 			printf(", %s, start %qx, end %qx",
176 				blk->lf_type == F_RDLCK ? "shared" :
177 				blk->lf_type == F_WRLCK ? "exclusive" :
178 				blk->lf_type == F_UNLCK ? "unlock" :
179 				"unknown", blk->lf_start, blk->lf_end);
180 			if (TAILQ_FIRST(&blk->lf_blkhd))
181 				 panic("lf_printlist: bad list");
182 		}
183 		printf("\n");
184 	}
185 }
186 #endif /* LOCKF_DEBUG */
187 
188 /*
189  * 3 options for allowfail.
190  * 0 - always allocate.  1 - cutoff at limit.  2 - cutoff at double limit.
191  */
192 static struct lockf *
193 lf_alloc(uid_t uid, int allowfail)
194 {
195 	struct uidinfo *uip;
196 	struct lockf *lock;
197 	u_long lcnt;
198 
199 	uip = uid_find(uid);
200 	lcnt = atomic_inc_ulong_nv(&uip->ui_lockcnt);
201 	if (uid && allowfail && lcnt >
202 	    (allowfail == 1 ? maxlocksperuid : (maxlocksperuid * 2))) {
203 		atomic_dec_ulong(&uip->ui_lockcnt);
204 		return NULL;
205 	}
206 
207 	lock = pool_cache_get(lockf_cache, PR_WAITOK);
208 	lock->lf_uid = uid;
209 	return lock;
210 }
211 
212 static void
213 lf_free(struct lockf *lock)
214 {
215 	struct uidinfo *uip;
216 
217 	uip = uid_find(lock->lf_uid);
218 	atomic_dec_ulong(&uip->ui_lockcnt);
219 	pool_cache_put(lockf_cache, lock);
220 }
221 
222 static int
223 lf_ctor(void *arg, void *obj, int flag)
224 {
225 	struct lockf *lock;
226 
227 	lock = obj;
228 	cv_init(&lock->lf_cv, lockstr);
229 
230 	return 0;
231 }
232 
233 static void
234 lf_dtor(void *arg, void *obj)
235 {
236 	struct lockf *lock;
237 
238 	lock = obj;
239 	cv_destroy(&lock->lf_cv);
240 }
241 
242 /*
243  * Walk the list of locks for an inode to
244  * find an overlapping lock (if any).
245  *
246  * NOTE: this returns only the FIRST overlapping lock.  There
247  *	 may be more than one.
248  */
249 static int
250 lf_findoverlap(struct lockf *lf, struct lockf *lock, int type,
251     struct lockf ***prev, struct lockf **overlap)
252 {
253 	off_t start, end;
254 
255 	*overlap = lf;
256 	if (lf == NULL)
257 		return 0;
258 #ifdef LOCKF_DEBUG
259 	if (lockf_debug & 2)
260 		lf_print("lf_findoverlap: looking for overlap in", lock);
261 #endif /* LOCKF_DEBUG */
262 	start = lock->lf_start;
263 	end = lock->lf_end;
264 	while (lf != NULL) {
265 		if (((type == SELF) && lf->lf_id != lock->lf_id) ||
266 		    ((type == OTHERS) && lf->lf_id == lock->lf_id)) {
267 			*prev = &lf->lf_next;
268 			*overlap = lf = lf->lf_next;
269 			continue;
270 		}
271 #ifdef LOCKF_DEBUG
272 		if (lockf_debug & 2)
273 			lf_print("\tchecking", lf);
274 #endif /* LOCKF_DEBUG */
275 		/*
276 		 * OK, check for overlap
277 		 *
278 		 * Six cases:
279 		 *	0) no overlap
280 		 *	1) overlap == lock
281 		 *	2) overlap contains lock
282 		 *	3) lock contains overlap
283 		 *	4) overlap starts before lock
284 		 *	5) overlap ends after lock
285 		 */
286 		if ((lf->lf_end != -1 && start > lf->lf_end) ||
287 		    (end != -1 && lf->lf_start > end)) {
288 			/* Case 0 */
289 #ifdef LOCKF_DEBUG
290 			if (lockf_debug & 2)
291 				printf("no overlap\n");
292 #endif /* LOCKF_DEBUG */
293 			if ((type & SELF) && end != -1 && lf->lf_start > end)
294 				return 0;
295 			*prev = &lf->lf_next;
296 			*overlap = lf = lf->lf_next;
297 			continue;
298 		}
299 		if ((lf->lf_start == start) && (lf->lf_end == end)) {
300 			/* Case 1 */
301 #ifdef LOCKF_DEBUG
302 			if (lockf_debug & 2)
303 				printf("overlap == lock\n");
304 #endif /* LOCKF_DEBUG */
305 			return 1;
306 		}
307 		if ((lf->lf_start <= start) &&
308 		    (end != -1) &&
309 		    ((lf->lf_end >= end) || (lf->lf_end == -1))) {
310 			/* Case 2 */
311 #ifdef LOCKF_DEBUG
312 			if (lockf_debug & 2)
313 				printf("overlap contains lock\n");
314 #endif /* LOCKF_DEBUG */
315 			return 2;
316 		}
317 		if (start <= lf->lf_start &&
318 		           (end == -1 ||
319 			   (lf->lf_end != -1 && end >= lf->lf_end))) {
320 			/* Case 3 */
321 #ifdef LOCKF_DEBUG
322 			if (lockf_debug & 2)
323 				printf("lock contains overlap\n");
324 #endif /* LOCKF_DEBUG */
325 			return 3;
326 		}
327 		if ((lf->lf_start < start) &&
328 			((lf->lf_end >= start) || (lf->lf_end == -1))) {
329 			/* Case 4 */
330 #ifdef LOCKF_DEBUG
331 			if (lockf_debug & 2)
332 				printf("overlap starts before lock\n");
333 #endif /* LOCKF_DEBUG */
334 			return 4;
335 		}
336 		if ((lf->lf_start > start) &&
337 			(end != -1) &&
338 			((lf->lf_end > end) || (lf->lf_end == -1))) {
339 			/* Case 5 */
340 #ifdef LOCKF_DEBUG
341 			if (lockf_debug & 2)
342 				printf("overlap ends after lock\n");
343 #endif /* LOCKF_DEBUG */
344 			return 5;
345 		}
346 		panic("lf_findoverlap: default");
347 	}
348 	return 0;
349 }
350 
351 /*
352  * Split a lock and a contained region into
353  * two or three locks as necessary.
354  */
355 static void
356 lf_split(struct lockf *lock1, struct lockf *lock2, struct lockf **sparelock)
357 {
358 	struct lockf *splitlock;
359 
360 #ifdef LOCKF_DEBUG
361 	if (lockf_debug & 2) {
362 		lf_print("lf_split", lock1);
363 		lf_print("splitting from", lock2);
364 	}
365 #endif /* LOCKF_DEBUG */
366 	/*
367 	 * Check to see if spliting into only two pieces.
368 	 */
369 	if (lock1->lf_start == lock2->lf_start) {
370 		lock1->lf_start = lock2->lf_end + 1;
371 		lock2->lf_next = lock1;
372 		return;
373 	}
374 	if (lock1->lf_end == lock2->lf_end) {
375 		lock1->lf_end = lock2->lf_start - 1;
376 		lock2->lf_next = lock1->lf_next;
377 		lock1->lf_next = lock2;
378 		return;
379 	}
380 	/*
381 	 * Make a new lock consisting of the last part of
382 	 * the encompassing lock
383 	 */
384 	splitlock = *sparelock;
385 	*sparelock = NULL;
386 	memcpy(splitlock, lock1, sizeof(*splitlock));
387 	cv_init(&splitlock->lf_cv, lockstr);
388 
389 	splitlock->lf_start = lock2->lf_end + 1;
390 	TAILQ_INIT(&splitlock->lf_blkhd);
391 	lock1->lf_end = lock2->lf_start - 1;
392 	/*
393 	 * OK, now link it in
394 	 */
395 	splitlock->lf_next = lock1->lf_next;
396 	lock2->lf_next = splitlock;
397 	lock1->lf_next = lock2;
398 }
399 
400 /*
401  * Wakeup a blocklist
402  */
403 static void
404 lf_wakelock(struct lockf *listhead)
405 {
406 	struct lockf *wakelock;
407 
408 	while ((wakelock = TAILQ_FIRST(&listhead->lf_blkhd))) {
409 		KASSERT(wakelock->lf_next == listhead);
410 		TAILQ_REMOVE(&listhead->lf_blkhd, wakelock, lf_block);
411 		wakelock->lf_next = NULL;
412 #ifdef LOCKF_DEBUG
413 		if (lockf_debug & 2)
414 			lf_print("lf_wakelock: awakening", wakelock);
415 #endif
416 		cv_broadcast(&wakelock->lf_cv);
417 	}
418 }
419 
420 /*
421  * Remove a byte-range lock on an inode.
422  *
423  * Generally, find the lock (or an overlap to that lock)
424  * and remove it (or shrink it), then wakeup anyone we can.
425  */
426 static int
427 lf_clearlock(struct lockf *unlock, struct lockf **sparelock)
428 {
429 	struct lockf **head = unlock->lf_head;
430 	struct lockf *lf = *head;
431 	struct lockf *overlap, **prev;
432 	int ovcase;
433 
434 	if (lf == NULL)
435 		return 0;
436 #ifdef LOCKF_DEBUG
437 	if (unlock->lf_type != F_UNLCK)
438 		panic("lf_clearlock: bad type");
439 	if (lockf_debug & 1)
440 		lf_print("lf_clearlock", unlock);
441 #endif /* LOCKF_DEBUG */
442 	prev = head;
443 	while ((ovcase = lf_findoverlap(lf, unlock, SELF,
444 	    &prev, &overlap)) != 0) {
445 		/*
446 		 * Wakeup the list of locks to be retried.
447 		 */
448 		lf_wakelock(overlap);
449 
450 		switch (ovcase) {
451 
452 		case 1: /* overlap == lock */
453 			*prev = overlap->lf_next;
454 			lf_free(overlap);
455 			break;
456 
457 		case 2: /* overlap contains lock: split it */
458 			if (overlap->lf_start == unlock->lf_start) {
459 				overlap->lf_start = unlock->lf_end + 1;
460 				break;
461 			}
462 			lf_split(overlap, unlock, sparelock);
463 			overlap->lf_next = unlock->lf_next;
464 			break;
465 
466 		case 3: /* lock contains overlap */
467 			*prev = overlap->lf_next;
468 			lf = overlap->lf_next;
469 			lf_free(overlap);
470 			continue;
471 
472 		case 4: /* overlap starts before lock */
473 			overlap->lf_end = unlock->lf_start - 1;
474 			prev = &overlap->lf_next;
475 			lf = overlap->lf_next;
476 			continue;
477 
478 		case 5: /* overlap ends after lock */
479 			overlap->lf_start = unlock->lf_end + 1;
480 			break;
481 		}
482 		break;
483 	}
484 #ifdef LOCKF_DEBUG
485 	if (lockf_debug & 1)
486 		lf_printlist("lf_clearlock", unlock);
487 #endif /* LOCKF_DEBUG */
488 	return 0;
489 }
490 
491 /*
492  * Walk the list of locks for an inode and
493  * return the first blocking lock.
494  */
495 static struct lockf *
496 lf_getblock(struct lockf *lock)
497 {
498 	struct lockf **prev, *overlap, *lf = *(lock->lf_head);
499 
500 	prev = lock->lf_head;
501 	while (lf_findoverlap(lf, lock, OTHERS, &prev, &overlap) != 0) {
502 		/*
503 		 * We've found an overlap, see if it blocks us
504 		 */
505 		if ((lock->lf_type == F_WRLCK || overlap->lf_type == F_WRLCK))
506 			return overlap;
507 		/*
508 		 * Nope, point to the next one on the list and
509 		 * see if it blocks us
510 		 */
511 		lf = overlap->lf_next;
512 	}
513 	return NULL;
514 }
515 
516 /*
517  * Set a byte-range lock.
518  */
519 static int
520 lf_setlock(struct lockf *lock, struct lockf **sparelock,
521     kmutex_t *interlock)
522 {
523 	struct lockf *block;
524 	struct lockf **head = lock->lf_head;
525 	struct lockf **prev, *overlap, *ltmp;
526 	int ovcase, needtolink, error;
527 
528 #ifdef LOCKF_DEBUG
529 	if (lockf_debug & 1)
530 		lf_print("lf_setlock", lock);
531 #endif /* LOCKF_DEBUG */
532 
533 	/*
534 	 * Scan lock list for this file looking for locks that would block us.
535 	 */
536 	while ((block = lf_getblock(lock)) != NULL) {
537 		/*
538 		 * Free the structure and return if nonblocking.
539 		 */
540 		if ((lock->lf_flags & F_WAIT) == 0) {
541 			lf_free(lock);
542 			return EAGAIN;
543 		}
544 		/*
545 		 * We are blocked. Since flock style locks cover
546 		 * the whole file, there is no chance for deadlock.
547 		 * For byte-range locks we must check for deadlock.
548 		 *
549 		 * Deadlock detection is done by looking through the
550 		 * wait channels to see if there are any cycles that
551 		 * involve us. MAXDEPTH is set just to make sure we
552 		 * do not go off into neverneverland.
553 		 */
554 		if ((lock->lf_flags & F_POSIX) &&
555 		    (block->lf_flags & F_POSIX)) {
556 			struct lwp *wlwp;
557 			volatile const struct lockf *waitblock;
558 			int i = 0;
559 			struct proc *p;
560 
561 			p = (struct proc *)block->lf_id;
562 			KASSERT(p != NULL);
563 			while (i++ < maxlockdepth) {
564 				mutex_enter(p->p_lock);
565 				if (p->p_nlwps > 1) {
566 					mutex_exit(p->p_lock);
567 					break;
568 				}
569 				wlwp = LIST_FIRST(&p->p_lwps);
570 				lwp_lock(wlwp);
571 				if (wlwp->l_wchan == NULL ||
572 				    wlwp->l_wmesg != lockstr) {
573 					lwp_unlock(wlwp);
574 					mutex_exit(p->p_lock);
575 					break;
576 				}
577 				waitblock = wlwp->l_wchan;
578 				lwp_unlock(wlwp);
579 				mutex_exit(p->p_lock);
580 				/* Get the owner of the blocking lock */
581 				waitblock = waitblock->lf_next;
582 				if ((waitblock->lf_flags & F_POSIX) == 0)
583 					break;
584 				p = (struct proc *)waitblock->lf_id;
585 				if (p == curproc) {
586 					lf_free(lock);
587 					return EDEADLK;
588 				}
589 			}
590 			/*
591 			 * If we're still following a dependency chain
592 			 * after maxlockdepth iterations, assume we're in
593 			 * a cycle to be safe.
594 			 */
595 			if (i >= maxlockdepth) {
596 				lf_free(lock);
597 				return EDEADLK;
598 			}
599 		}
600 		/*
601 		 * For flock type locks, we must first remove
602 		 * any shared locks that we hold before we sleep
603 		 * waiting for an exclusive lock.
604 		 */
605 		if ((lock->lf_flags & F_FLOCK) &&
606 		    lock->lf_type == F_WRLCK) {
607 			lock->lf_type = F_UNLCK;
608 			(void) lf_clearlock(lock, NULL);
609 			lock->lf_type = F_WRLCK;
610 		}
611 		/*
612 		 * Add our lock to the blocked list and sleep until we're free.
613 		 * Remember who blocked us (for deadlock detection).
614 		 */
615 		lock->lf_next = block;
616 		TAILQ_INSERT_TAIL(&block->lf_blkhd, lock, lf_block);
617 #ifdef LOCKF_DEBUG
618 		if (lockf_debug & 1) {
619 			lf_print("lf_setlock: blocking on", block);
620 			lf_printlist("lf_setlock", block);
621 		}
622 #endif /* LOCKF_DEBUG */
623 		error = cv_wait_sig(&lock->lf_cv, interlock);
624 
625 		/*
626 		 * We may have been awoken by a signal (in
627 		 * which case we must remove ourselves from the
628 		 * blocked list) and/or by another process
629 		 * releasing a lock (in which case we have already
630 		 * been removed from the blocked list and our
631 		 * lf_next field set to NULL).
632 		 */
633 		if (lock->lf_next != NULL) {
634 			TAILQ_REMOVE(&lock->lf_next->lf_blkhd, lock, lf_block);
635 			lock->lf_next = NULL;
636 		}
637 		if (error) {
638 			lf_free(lock);
639 			return error;
640 		}
641 	}
642 	/*
643 	 * No blocks!!  Add the lock.  Note that we will
644 	 * downgrade or upgrade any overlapping locks this
645 	 * process already owns.
646 	 *
647 	 * Skip over locks owned by other processes.
648 	 * Handle any locks that overlap and are owned by ourselves.
649 	 */
650 	prev = head;
651 	block = *head;
652 	needtolink = 1;
653 	for (;;) {
654 		ovcase = lf_findoverlap(block, lock, SELF, &prev, &overlap);
655 		if (ovcase)
656 			block = overlap->lf_next;
657 		/*
658 		 * Six cases:
659 		 *	0) no overlap
660 		 *	1) overlap == lock
661 		 *	2) overlap contains lock
662 		 *	3) lock contains overlap
663 		 *	4) overlap starts before lock
664 		 *	5) overlap ends after lock
665 		 */
666 		switch (ovcase) {
667 		case 0: /* no overlap */
668 			if (needtolink) {
669 				*prev = lock;
670 				lock->lf_next = overlap;
671 			}
672 			break;
673 
674 		case 1: /* overlap == lock */
675 			/*
676 			 * If downgrading lock, others may be
677 			 * able to acquire it.
678 			 */
679 			if (lock->lf_type == F_RDLCK &&
680 			    overlap->lf_type == F_WRLCK)
681 				lf_wakelock(overlap);
682 			overlap->lf_type = lock->lf_type;
683 			lf_free(lock);
684 			lock = overlap; /* for debug output below */
685 			break;
686 
687 		case 2: /* overlap contains lock */
688 			/*
689 			 * Check for common starting point and different types.
690 			 */
691 			if (overlap->lf_type == lock->lf_type) {
692 				lf_free(lock);
693 				lock = overlap; /* for debug output below */
694 				break;
695 			}
696 			if (overlap->lf_start == lock->lf_start) {
697 				*prev = lock;
698 				lock->lf_next = overlap;
699 				overlap->lf_start = lock->lf_end + 1;
700 			} else
701 				lf_split(overlap, lock, sparelock);
702 			lf_wakelock(overlap);
703 			break;
704 
705 		case 3: /* lock contains overlap */
706 			/*
707 			 * If downgrading lock, others may be able to
708 			 * acquire it, otherwise take the list.
709 			 */
710 			if (lock->lf_type == F_RDLCK &&
711 			    overlap->lf_type == F_WRLCK) {
712 				lf_wakelock(overlap);
713 			} else {
714 				while ((ltmp = TAILQ_FIRST(&overlap->lf_blkhd))) {
715 					KASSERT(ltmp->lf_next == overlap);
716 					TAILQ_REMOVE(&overlap->lf_blkhd, ltmp,
717 					    lf_block);
718 					ltmp->lf_next = lock;
719 					TAILQ_INSERT_TAIL(&lock->lf_blkhd,
720 					    ltmp, lf_block);
721 				}
722 			}
723 			/*
724 			 * Add the new lock if necessary and delete the overlap.
725 			 */
726 			if (needtolink) {
727 				*prev = lock;
728 				lock->lf_next = overlap->lf_next;
729 				prev = &lock->lf_next;
730 				needtolink = 0;
731 			} else
732 				*prev = overlap->lf_next;
733 			lf_free(overlap);
734 			continue;
735 
736 		case 4: /* overlap starts before lock */
737 			/*
738 			 * Add lock after overlap on the list.
739 			 */
740 			lock->lf_next = overlap->lf_next;
741 			overlap->lf_next = lock;
742 			overlap->lf_end = lock->lf_start - 1;
743 			prev = &lock->lf_next;
744 			lf_wakelock(overlap);
745 			needtolink = 0;
746 			continue;
747 
748 		case 5: /* overlap ends after lock */
749 			/*
750 			 * Add the new lock before overlap.
751 			 */
752 			if (needtolink) {
753 				*prev = lock;
754 				lock->lf_next = overlap;
755 			}
756 			overlap->lf_start = lock->lf_end + 1;
757 			lf_wakelock(overlap);
758 			break;
759 		}
760 		break;
761 	}
762 #ifdef LOCKF_DEBUG
763 	if (lockf_debug & 1) {
764 		lf_print("lf_setlock: got the lock", lock);
765 		lf_printlist("lf_setlock", lock);
766 	}
767 #endif /* LOCKF_DEBUG */
768 	return 0;
769 }
770 
771 /*
772  * Check whether there is a blocking lock,
773  * and if so return its process identifier.
774  */
775 static int
776 lf_getlock(struct lockf *lock, struct flock *fl)
777 {
778 	struct lockf *block;
779 
780 #ifdef LOCKF_DEBUG
781 	if (lockf_debug & 1)
782 		lf_print("lf_getlock", lock);
783 #endif /* LOCKF_DEBUG */
784 
785 	if ((block = lf_getblock(lock)) != NULL) {
786 		fl->l_type = block->lf_type;
787 		fl->l_whence = SEEK_SET;
788 		fl->l_start = block->lf_start;
789 		if (block->lf_end == -1)
790 			fl->l_len = 0;
791 		else
792 			fl->l_len = block->lf_end - block->lf_start + 1;
793 		if (block->lf_flags & F_POSIX)
794 			fl->l_pid = ((struct proc *)block->lf_id)->p_pid;
795 		else
796 			fl->l_pid = -1;
797 	} else {
798 		fl->l_type = F_UNLCK;
799 	}
800 	return 0;
801 }
802 
803 /*
804  * Do an advisory lock operation.
805  */
806 int
807 lf_advlock(struct vop_advlock_args *ap, struct lockf **head, off_t size)
808 {
809 	struct lwp *l = curlwp;
810 	struct flock *fl = ap->a_fl;
811 	struct lockf *lock = NULL;
812 	struct lockf *sparelock;
813 	kmutex_t *interlock = lockf_lock;
814 	off_t start, end;
815 	int error = 0;
816 
817 	/*
818 	 * Convert the flock structure into a start and end.
819 	 */
820 	switch (fl->l_whence) {
821 	case SEEK_SET:
822 	case SEEK_CUR:
823 		/*
824 		 * Caller is responsible for adding any necessary offset
825 		 * when SEEK_CUR is used.
826 		 */
827 		start = fl->l_start;
828 		break;
829 
830 	case SEEK_END:
831 		start = size + fl->l_start;
832 		break;
833 
834 	default:
835 		return EINVAL;
836 	}
837 	if (start < 0)
838 		return EINVAL;
839 
840 	/*
841 	 * Allocate locks before acquiring the interlock.  We need two
842 	 * locks in the worst case.
843 	 */
844 	switch (ap->a_op) {
845 	case F_SETLK:
846 	case F_UNLCK:
847 		/*
848 		 * XXX For F_UNLCK case, we can re-use the lock.
849 		 */
850 		if ((ap->a_flags & F_FLOCK) == 0) {
851 			/*
852 			 * Byte-range lock might need one more lock.
853 			 */
854 			sparelock = lf_alloc(kauth_cred_geteuid(l->l_cred), 0);
855 			if (sparelock == NULL) {
856 				error = ENOMEM;
857 				goto quit;
858 			}
859 			break;
860 		}
861 		/* FALLTHROUGH */
862 
863 	case F_GETLK:
864 		sparelock = NULL;
865 		break;
866 
867 	default:
868 		return EINVAL;
869 	}
870 
871 	lock = lf_alloc(kauth_cred_geteuid(l->l_cred),
872 	    ap->a_op != F_UNLCK ? 1 : 2);
873 	if (lock == NULL) {
874 		error = ENOMEM;
875 		goto quit;
876 	}
877 
878 	mutex_enter(interlock);
879 
880 	/*
881 	 * Avoid the common case of unlocking when inode has no locks.
882 	 */
883 	if (*head == (struct lockf *)0) {
884 		if (ap->a_op != F_SETLK) {
885 			fl->l_type = F_UNLCK;
886 			error = 0;
887 			goto quit_unlock;
888 		}
889 	}
890 
891 	if (fl->l_len == 0)
892 		end = -1;
893 	else
894 		end = start + fl->l_len - 1;
895 	/*
896 	 * Create the lockf structure.
897 	 */
898 	lock->lf_start = start;
899 	lock->lf_end = end;
900 	lock->lf_head = head;
901 	lock->lf_type = fl->l_type;
902 	lock->lf_next = (struct lockf *)0;
903 	TAILQ_INIT(&lock->lf_blkhd);
904 	lock->lf_flags = ap->a_flags;
905 	if (lock->lf_flags & F_POSIX) {
906 		KASSERT(curproc == (struct proc *)ap->a_id);
907 	}
908 	lock->lf_id = (struct proc *)ap->a_id;
909 
910 	/*
911 	 * Do the requested operation.
912 	 */
913 	switch (ap->a_op) {
914 
915 	case F_SETLK:
916 		error = lf_setlock(lock, &sparelock, interlock);
917 		lock = NULL; /* lf_setlock freed it */
918 		break;
919 
920 	case F_UNLCK:
921 		error = lf_clearlock(lock, &sparelock);
922 		break;
923 
924 	case F_GETLK:
925 		error = lf_getlock(lock, fl);
926 		break;
927 
928 	default:
929 		break;
930 		/* NOTREACHED */
931 	}
932 
933 quit_unlock:
934 	mutex_exit(interlock);
935 quit:
936 	if (lock)
937 		lf_free(lock);
938 	if (sparelock)
939 		lf_free(sparelock);
940 
941 	return error;
942 }
943 
944 /*
945  * Initialize subsystem.   XXX We use a global lock.  This could be the
946  * vnode interlock, but the deadlock detection code may need to inspect
947  * locks belonging to other files.
948  */
949 void
950 lf_init(void)
951 {
952 
953 	lockf_cache = pool_cache_init(sizeof(struct lockf), 0, 0, 0, "lockf",
954  	    NULL, IPL_NONE, lf_ctor, lf_dtor, NULL);
955         lockf_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE);
956 }
957