xref: /openbsd/sys/kern/vfs_lockf.c (revision 0d280c5f)
1 /*	$OpenBSD: vfs_lockf.c,v 1.50 2022/08/14 01:58:28 jsg Exp $	*/
2 /*	$NetBSD: vfs_lockf.c,v 1.7 1996/02/04 02:18:21 christos Exp $	*/
3 
4 /*
5  * Copyright (c) 1982, 1986, 1989, 1993
6  *	The Regents of the University of California.  All rights reserved.
7  *
8  * This code is derived from software contributed to Berkeley by
9  * Scooter Morris at Genentech Inc.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  * 3. Neither the name of the University nor the names of its contributors
20  *    may be used to endorse or promote products derived from this software
21  *    without specific prior written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33  * SUCH DAMAGE.
34  *
35  *	@(#)ufs_lockf.c	8.3 (Berkeley) 1/6/94
36  */
37 
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/proc.h>
41 #include <sys/pool.h>
42 #include <sys/fcntl.h>
43 #include <sys/lockf.h>
44 #include <sys/rwlock.h>
45 #include <sys/unistd.h>
46 
47 /*
48  * The lockf structure is a kernel structure which contains the information
49  * associated with a byte range lock.  The lockf structures are linked into
50  * the inode structure. Locks are sorted by the starting byte of the lock for
51  * efficiency.
52  */
53 TAILQ_HEAD(locklist, lockf);
54 
55 struct lockf {
56 	short	lf_flags;	 /* Lock semantics: F_POSIX, F_FLOCK, F_WAIT */
57 	short	lf_type;	 /* Lock type: F_RDLCK, F_WRLCK */
58 	off_t	lf_start;	 /* The byte # of the start of the lock */
59 	off_t	lf_end;		 /* The byte # of the end of the lock (-1=EOF)*/
60 	caddr_t	lf_id;		 /* The id of the resource holding the lock */
61 	struct	lockf_state *lf_state;	/* State associated with the lock */
62 	TAILQ_ENTRY(lockf) lf_entry;
63 	struct	lockf *lf_blk;	 /* The lock that blocks us */
64 	struct	locklist lf_blkhd;	/* The list of blocked locks */
65 	TAILQ_ENTRY(lockf) lf_block; /* A request waiting for a lock */
66 	uid_t	lf_uid;		/* User ID responsible */
67 	pid_t	lf_pid;		/* POSIX - owner pid */
68 };
69 
70 struct lockf_state {
71 	TAILQ_HEAD(, lockf)	  ls_locks;	/* list of active locks */
72 	TAILQ_HEAD(, lockf)	  ls_pending;	/* list of pending locks */
73 	struct lockf_state	**ls_owner;	/* owner */
74 	int		 	  ls_refs;	/* reference counter */
75 };
76 
77 struct pool lockf_state_pool;
78 struct pool lockf_pool;
79 
80 #define SELF	0x1
81 #define OTHERS	0x2
82 
83 #ifdef LOCKF_DEBUG
84 
85 #define	DEBUG_SETLOCK		0x01
86 #define	DEBUG_CLEARLOCK		0x02
87 #define	DEBUG_GETLOCK		0x04
88 #define	DEBUG_FINDOVR		0x08
89 #define	DEBUG_SPLIT		0x10
90 #define	DEBUG_WAKELOCK		0x20
91 #define	DEBUG_LINK		0x40
92 
93 int	lockf_debug = DEBUG_SETLOCK|DEBUG_CLEARLOCK|DEBUG_WAKELOCK;
94 
95 void	lf_print(const char *, struct lockf *);
96 void	lf_printlist(const char *, struct lockf *);
97 
98 #define	DPRINTF(args, level)	if (lockf_debug & (level)) printf args
99 #define	LFPRINT(args, level)	if (lockf_debug & (level)) lf_print args
100 #else
101 #define	DPRINTF(args, level)
102 #define	LFPRINT(args, level)
103 #endif
104 
105 struct lockf *lf_alloc(uid_t, int);
106 void lf_free(struct lockf *);
107 int lf_clearlock(struct lockf *);
108 int lf_findoverlap(struct lockf *, struct lockf *, int, struct lockf **);
109 struct lockf *lf_getblock(struct lockf *, struct lockf *);
110 int lf_getlock(struct lockf *, struct flock *);
111 int lf_setlock(struct lockf *);
112 void lf_split(struct lockf *, struct lockf *);
113 void lf_wakelock(struct lockf *, int);
114 int lf_deadlock(struct lockf *);
115 void ls_ref(struct lockf_state *);
116 void ls_rele(struct lockf_state *);
117 
118 /*
119  * Serializes access to each instance of struct lockf and struct lockf_state
120  * and each pointer from a vnode to struct lockf_state.
121  */
122 struct rwlock lockf_lock = RWLOCK_INITIALIZER("lockflk");
123 
124 void
lf_init(void)125 lf_init(void)
126 {
127 	pool_init(&lockf_state_pool, sizeof(struct lockf_state), 0, IPL_NONE,
128 	    PR_WAITOK | PR_RWLOCK, "lockfspl", NULL);
129 	pool_init(&lockf_pool, sizeof(struct lockf), 0, IPL_NONE,
130 	    PR_WAITOK | PR_RWLOCK, "lockfpl", NULL);
131 }
132 
133 void
ls_ref(struct lockf_state * ls)134 ls_ref(struct lockf_state *ls)
135 {
136 	rw_assert_wrlock(&lockf_lock);
137 
138 	ls->ls_refs++;
139 }
140 
141 void
ls_rele(struct lockf_state * ls)142 ls_rele(struct lockf_state *ls)
143 {
144 	rw_assert_wrlock(&lockf_lock);
145 
146 	if (--ls->ls_refs > 0)
147 		return;
148 
149 	KASSERT(TAILQ_EMPTY(&ls->ls_locks));
150 	KASSERT(TAILQ_EMPTY(&ls->ls_pending));
151 
152 	*ls->ls_owner = NULL;
153 	pool_put(&lockf_state_pool, ls);
154 }
155 
156 /*
157  * We enforce a limit on locks by uid, so that a single user cannot
158  * run the kernel out of memory.  For now, the limit is pretty coarse.
159  * There is no limit on root.
160  *
161  * Splitting a lock will always succeed, regardless of current allocations.
162  * If you're slightly above the limit, we still have to permit an allocation
163  * so that the unlock can succeed.  If the unlocking causes too many splits,
164  * however, you're totally cutoff.
165  */
166 int maxlocksperuid = 1024;
167 
168 /*
169  * 3 options for allowfail.
170  * 0 - always allocate.  1 - cutoff at limit.  2 - cutoff at double limit.
171  */
172 struct lockf *
lf_alloc(uid_t uid,int allowfail)173 lf_alloc(uid_t uid, int allowfail)
174 {
175 	struct uidinfo *uip;
176 	struct lockf *lock;
177 
178 	uip = uid_find(uid);
179 	if (uid && allowfail && uip->ui_lockcnt >
180 	    (allowfail == 1 ? maxlocksperuid : (maxlocksperuid * 2))) {
181 		uid_release(uip);
182 		return (NULL);
183 	}
184 	uip->ui_lockcnt++;
185 	uid_release(uip);
186 	lock = pool_get(&lockf_pool, PR_WAITOK);
187 	lock->lf_uid = uid;
188 	return (lock);
189 }
190 
191 void
lf_free(struct lockf * lock)192 lf_free(struct lockf *lock)
193 {
194 	struct uidinfo *uip;
195 
196 	rw_assert_wrlock(&lockf_lock);
197 
198 	LFPRINT(("lf_free", lock), DEBUG_LINK);
199 
200 	KASSERT(TAILQ_EMPTY(&lock->lf_blkhd));
201 
202 	ls_rele(lock->lf_state);
203 
204 	uip = uid_find(lock->lf_uid);
205 	uip->ui_lockcnt--;
206 	uid_release(uip);
207 	pool_put(&lockf_pool, lock);
208 }
209 
210 
211 /*
212  * Do an advisory lock operation.
213  */
214 int
lf_advlock(struct lockf_state ** state,off_t size,caddr_t id,int op,struct flock * fl,int flags)215 lf_advlock(struct lockf_state **state, off_t size, caddr_t id, int op,
216     struct flock *fl, int flags)
217 {
218 	struct proc *p = curproc;
219 	struct lockf_state *ls;
220 	struct lockf *lock;
221 	off_t start, end;
222 	int error = 0;
223 
224 	/*
225 	 * Convert the flock structure into a start and end.
226 	 */
227 	switch (fl->l_whence) {
228 	case SEEK_SET:
229 	case SEEK_CUR:
230 		/*
231 		 * Caller is responsible for adding any necessary offset
232 		 * when SEEK_CUR is used.
233 		 */
234 		start = fl->l_start;
235 		break;
236 	case SEEK_END:
237 		start = size + fl->l_start;
238 		break;
239 	default:
240 		return (EINVAL);
241 	}
242 	if (start < 0)
243 		return (EINVAL);
244 	if (fl->l_len > 0) {
245 		if (fl->l_len - 1 > LLONG_MAX - start)
246 			return (EOVERFLOW);
247 		end = start + (fl->l_len - 1);
248 		/* Avoid ambiguity at the end of the range. */
249 		if (end == LLONG_MAX)
250 			end = -1;
251 	} else if (fl->l_len < 0) {
252 		if (start + fl->l_len < 0)
253 			return (EINVAL);
254 		end = start - 1;
255 		start += fl->l_len;
256 	} else {
257 		end = -1;
258 	}
259 
260 	rw_enter_write(&lockf_lock);
261 	ls = *state;
262 
263 	/*
264 	 * Avoid the common case of unlocking when inode has no locks.
265 	 */
266 	if (ls == NULL && op != F_SETLK) {
267 		fl->l_type = F_UNLCK;
268 		goto out;
269 	}
270 
271 	if (ls == NULL) {
272 		ls = pool_get(&lockf_state_pool, PR_WAITOK | PR_ZERO);
273 		ls->ls_owner = state;
274 		TAILQ_INIT(&ls->ls_locks);
275 		TAILQ_INIT(&ls->ls_pending);
276 		*state = ls;
277 	}
278 	ls_ref(ls);
279 
280 	lock = lf_alloc(p->p_ucred->cr_uid, op == F_SETLK ? 1 : 2);
281 	if (!lock) {
282 		ls_rele(ls);
283 		error = ENOLCK;
284 		goto out;
285 	}
286 	lock->lf_flags = flags;
287 	lock->lf_type = fl->l_type;
288 	lock->lf_start = start;
289 	lock->lf_end = end;
290 	lock->lf_id = id;
291 	lock->lf_state = ls;
292 	lock->lf_blk = NULL;
293 	lock->lf_pid = (flags & F_POSIX) ? p->p_p->ps_pid : -1;
294 	TAILQ_INIT(&lock->lf_blkhd);
295 
296 	switch (op) {
297 	case F_SETLK:
298 		error = lf_setlock(lock);
299 		break;
300 	case F_UNLCK:
301 		error = lf_clearlock(lock);
302 		lf_free(lock);
303 		break;
304 	case F_GETLK:
305 		error = lf_getlock(lock, fl);
306 		lf_free(lock);
307 		break;
308 	default:
309 		lf_free(lock);
310 		error = EINVAL;
311 		break;
312 	}
313 
314 out:
315 	rw_exit_write(&lockf_lock);
316 	return (error);
317 }
318 
319 /*
320  * Set a byte-range lock.
321  */
322 int
lf_setlock(struct lockf * lock)323 lf_setlock(struct lockf *lock)
324 {
325 	struct lockf *block;
326 	struct lockf *overlap, *ltmp;
327 	int ovcase, priority, needtolink, error;
328 
329 	rw_assert_wrlock(&lockf_lock);
330 
331 	LFPRINT(("lf_setlock", lock), DEBUG_SETLOCK);
332 
333 	priority = PLOCK;
334 	if (lock->lf_type == F_WRLCK)
335 		priority += 4;
336 	priority |= PCATCH;
337 	/*
338 	 * Scan lock list for this file looking for locks that would block us.
339 	 */
340 	for (;;) {
341 		block = lf_getblock(TAILQ_FIRST(&lock->lf_state->ls_locks),
342 		    lock);
343 		if (block == NULL)
344 			break;
345 
346 		if ((lock->lf_flags & F_WAIT) == 0) {
347 			lf_free(lock);
348 			return (EAGAIN);
349 		}
350 
351 		/*
352 		 * Lock is blocked, check for deadlock before proceeding.
353 		 * Note: flock style locks cover the whole file, there is no
354 		 * chance for deadlock.
355 		 */
356 		if ((lock->lf_flags & F_POSIX) && lf_deadlock(lock)) {
357 			lf_free(lock);
358 			return (EDEADLK);
359 		}
360 
361 		/*
362 		 * For flock type locks, we must first remove
363 		 * any shared locks that we hold before we sleep
364 		 * waiting for an exclusive lock.
365 		 */
366 		if ((lock->lf_flags & F_FLOCK) && lock->lf_type == F_WRLCK) {
367 			lock->lf_type = F_UNLCK;
368 			(void)lf_clearlock(lock);
369 			lock->lf_type = F_WRLCK;
370 		}
371 		/*
372 		 * Add our lock to the blocked list and sleep until we're free.
373 		 * Remember who blocked us (for deadlock detection).
374 		 */
375 		lock->lf_blk = block;
376 		LFPRINT(("lf_setlock", lock), DEBUG_SETLOCK);
377 		LFPRINT(("lf_setlock: blocking on", block), DEBUG_SETLOCK);
378 		TAILQ_INSERT_TAIL(&block->lf_blkhd, lock, lf_block);
379 		TAILQ_INSERT_TAIL(&lock->lf_state->ls_pending, lock, lf_entry);
380 		error = rwsleep_nsec(lock, &lockf_lock, priority, "lockf",
381 		    INFSLP);
382 		TAILQ_REMOVE(&lock->lf_state->ls_pending, lock, lf_entry);
383 		wakeup_one(lock->lf_state);
384 		if (lock->lf_blk != NULL) {
385 			TAILQ_REMOVE(&lock->lf_blk->lf_blkhd, lock, lf_block);
386 			lock->lf_blk = NULL;
387 		}
388 		if (error) {
389 			lf_free(lock);
390 			return (error);
391 		}
392 		if (lock->lf_flags & F_INTR) {
393 			lf_free(lock);
394 			return (EINTR);
395 		}
396 	}
397 	/*
398 	 * No blocks!!  Add the lock.  Note that we will
399 	 * downgrade or upgrade any overlapping locks this
400 	 * process already owns.
401 	 *
402 	 * Skip over locks owned by other processes.
403 	 * Handle any locks that overlap and are owned by ourselves.
404 	 */
405 	block = TAILQ_FIRST(&lock->lf_state->ls_locks);
406 	overlap = NULL;
407 	needtolink = 1;
408 	for (;;) {
409 		ovcase = lf_findoverlap(block, lock, SELF, &overlap);
410 		if (ovcase)
411 			block = TAILQ_NEXT(overlap, lf_entry);
412 		/*
413 		 * Six cases:
414 		 *	0) no overlap
415 		 *	1) overlap == lock
416 		 *	2) overlap contains lock
417 		 *	3) lock contains overlap
418 		 *	4) overlap starts before lock
419 		 *	5) overlap ends after lock
420 		 */
421 		switch (ovcase) {
422 		case 0: /* no overlap */
423 			if (needtolink) {
424 				if (overlap)	/* insert before overlap */
425 					TAILQ_INSERT_BEFORE(overlap, lock,
426 					    lf_entry);
427 				else		/* first or last lock in list */
428 					TAILQ_INSERT_TAIL(&lock->lf_state->ls_locks,
429 					    lock, lf_entry);
430 			}
431 			break;
432 		case 1: /* overlap == lock */
433 			/*
434 			 * If downgrading lock, others may be
435 			 * able to acquire it.
436 			 */
437 			if (lock->lf_type == F_RDLCK &&
438 			    overlap->lf_type == F_WRLCK)
439 				lf_wakelock(overlap, 0);
440 			overlap->lf_type = lock->lf_type;
441 			lf_free(lock);
442 			lock = overlap; /* for debug output below */
443 			break;
444 		case 2: /* overlap contains lock */
445 			/*
446 			 * Check for common starting point and different types.
447 			 */
448 			if (overlap->lf_type == lock->lf_type) {
449 				if (!needtolink)
450 					TAILQ_REMOVE(&lock->lf_state->ls_locks,
451 					    lock, lf_entry);
452 				lf_free(lock);
453 				lock = overlap; /* for debug output below */
454 				break;
455 			}
456 			if (overlap->lf_start == lock->lf_start) {
457 				if (!needtolink)
458 					TAILQ_REMOVE(&lock->lf_state->ls_locks,
459 					    lock, lf_entry);
460 				TAILQ_INSERT_BEFORE(overlap, lock, lf_entry);
461 				overlap->lf_start = lock->lf_end + 1;
462 			} else
463 				lf_split(overlap, lock);
464 			lf_wakelock(overlap, 0);
465 			break;
466 		case 3: /* lock contains overlap */
467 			/*
468 			 * If downgrading lock, others may be able to
469 			 * acquire it, otherwise take the list.
470 			 */
471 			if (lock->lf_type == F_RDLCK &&
472 			    overlap->lf_type == F_WRLCK) {
473 				lf_wakelock(overlap, 0);
474 			} else {
475 				while ((ltmp =
476 				    TAILQ_FIRST(&overlap->lf_blkhd))) {
477 					TAILQ_REMOVE(&overlap->lf_blkhd, ltmp,
478 					    lf_block);
479 					ltmp->lf_blk = lock;
480 					TAILQ_INSERT_TAIL(&lock->lf_blkhd,
481 					    ltmp, lf_block);
482 				}
483 			}
484 			/*
485 			 * Add the new lock if necessary and delete the overlap.
486 			 */
487 			if (needtolink) {
488 				TAILQ_INSERT_BEFORE(overlap, lock, lf_entry);
489 				needtolink = 0;
490 			}
491 			TAILQ_REMOVE(&lock->lf_state->ls_locks, overlap, lf_entry);
492 			lf_free(overlap);
493 			continue;
494 		case 4: /* overlap starts before lock */
495 			/*
496 			 * Add lock after overlap on the list.
497 			 */
498 			if (!needtolink)
499 				TAILQ_REMOVE(&lock->lf_state->ls_locks, lock,
500 				    lf_entry);
501 			TAILQ_INSERT_AFTER(&lock->lf_state->ls_locks, overlap,
502 			    lock, lf_entry);
503 			overlap->lf_end = lock->lf_start - 1;
504 			lf_wakelock(overlap, 0);
505 			needtolink = 0;
506 			continue;
507 		case 5: /* overlap ends after lock */
508 			/*
509 			 * Add the new lock before overlap.
510 			 */
511 			if (needtolink)
512 				TAILQ_INSERT_BEFORE(overlap, lock, lf_entry);
513 			overlap->lf_start = lock->lf_end + 1;
514 			lf_wakelock(overlap, 0);
515 			break;
516 		}
517 		break;
518 	}
519 	LFPRINT(("lf_setlock: got the lock", lock), DEBUG_SETLOCK);
520 	return (0);
521 }
522 
523 /*
524  * Remove a byte-range lock on an inode.
525  *
526  * Generally, find the lock (or an overlap to that lock)
527  * and remove it (or shrink it), then wakeup anyone we can.
528  */
529 int
lf_clearlock(struct lockf * lock)530 lf_clearlock(struct lockf *lock)
531 {
532 	struct lockf *lf, *overlap;
533 	int ovcase;
534 
535 	rw_assert_wrlock(&lockf_lock);
536 
537 	lf = TAILQ_FIRST(&lock->lf_state->ls_locks);
538 	if (lf == NULL)
539 		return (0);
540 
541 	LFPRINT(("lf_clearlock", lock), DEBUG_CLEARLOCK);
542 	while ((ovcase = lf_findoverlap(lf, lock, SELF, &overlap))) {
543 		lf_wakelock(overlap, 0);
544 
545 		switch (ovcase) {
546 		case 1: /* overlap == lock */
547 			TAILQ_REMOVE(&lock->lf_state->ls_locks, overlap,
548 			    lf_entry);
549 			lf_free(overlap);
550 			break;
551 		case 2: /* overlap contains lock: split it */
552 			if (overlap->lf_start == lock->lf_start) {
553 				overlap->lf_start = lock->lf_end + 1;
554 				break;
555 			}
556 			lf_split(overlap, lock);
557 			/*
558 			 * The lock is now part of the list, lf_clearlock() must
559 			 * ensure that the lock remains detached from the list.
560 			 */
561 			TAILQ_REMOVE(&lock->lf_state->ls_locks, lock, lf_entry);
562 			break;
563 		case 3: /* lock contains overlap */
564 			lf = TAILQ_NEXT(overlap, lf_entry);
565 			TAILQ_REMOVE(&lock->lf_state->ls_locks, overlap,
566 			    lf_entry);
567 			lf_free(overlap);
568 			continue;
569 		case 4: /* overlap starts before lock */
570 			overlap->lf_end = lock->lf_start - 1;
571 			lf = TAILQ_NEXT(overlap, lf_entry);
572 			continue;
573 		case 5: /* overlap ends after lock */
574 			overlap->lf_start = lock->lf_end + 1;
575 			break;
576 		}
577 		break;
578 	}
579 	return (0);
580 }
581 
582 /*
583  * Check whether there is a blocking lock,
584  * and if so return its process identifier.
585  */
586 int
lf_getlock(struct lockf * lock,struct flock * fl)587 lf_getlock(struct lockf *lock, struct flock *fl)
588 {
589 	struct lockf *block, *lf;
590 
591 	rw_assert_wrlock(&lockf_lock);
592 
593 	LFPRINT(("lf_getlock", lock), DEBUG_CLEARLOCK);
594 
595 	lf = TAILQ_FIRST(&lock->lf_state->ls_locks);
596 	if ((block = lf_getblock(lf, lock)) != NULL) {
597 		fl->l_type = block->lf_type;
598 		fl->l_whence = SEEK_SET;
599 		fl->l_start = block->lf_start;
600 		if (block->lf_end == -1)
601 			fl->l_len = 0;
602 		else
603 			fl->l_len = block->lf_end - block->lf_start + 1;
604 		fl->l_pid = block->lf_pid;
605 	} else {
606 		fl->l_type = F_UNLCK;
607 	}
608 	return (0);
609 }
610 
611 /*
612  * Walk the list of locks for an inode and
613  * return the first blocking lock.
614  */
615 struct lockf *
lf_getblock(struct lockf * lf,struct lockf * lock)616 lf_getblock(struct lockf *lf, struct lockf *lock)
617 {
618 	struct lockf *overlap;
619 
620 	rw_assert_wrlock(&lockf_lock);
621 
622 	while (lf_findoverlap(lf, lock, OTHERS, &overlap) != 0) {
623 		/*
624 		 * We've found an overlap, see if it blocks us
625 		 */
626 		if ((lock->lf_type == F_WRLCK || overlap->lf_type == F_WRLCK))
627 			return (overlap);
628 		/*
629 		 * Nope, point to the next one on the list and
630 		 * see if it blocks us
631 		 */
632 		lf = TAILQ_NEXT(overlap, lf_entry);
633 	}
634 	return (NULL);
635 }
636 
637 /*
638  * Walk the list of locks for an inode to
639  * find an overlapping lock (if any).
640  *
641  * NOTE: this returns only the FIRST overlapping lock.  There
642  *	 may be more than one.
643  */
644 int
lf_findoverlap(struct lockf * lf,struct lockf * lock,int type,struct lockf ** overlap)645 lf_findoverlap(struct lockf *lf, struct lockf *lock, int type,
646     struct lockf **overlap)
647 {
648 	off_t start, end;
649 
650 	rw_assert_wrlock(&lockf_lock);
651 
652 	LFPRINT(("lf_findoverlap: looking for overlap in", lock), DEBUG_FINDOVR);
653 
654 	*overlap = lf;
655 	start = lock->lf_start;
656 	end = lock->lf_end;
657 	while (lf != NULL) {
658 		if (((type & SELF) && lf->lf_id != lock->lf_id) ||
659 		    ((type & OTHERS) && lf->lf_id == lock->lf_id)) {
660 			*overlap = lf = TAILQ_NEXT(lf, lf_entry);
661 			continue;
662 		}
663 		LFPRINT(("\tchecking", lf), DEBUG_FINDOVR);
664 		/*
665 		 * OK, check for overlap
666 		 *
667 		 * Six cases:
668 		 *	0) no overlap
669 		 *	1) overlap == lock
670 		 *	2) overlap contains lock
671 		 *	3) lock contains overlap
672 		 *	4) overlap starts before lock
673 		 *	5) overlap ends after lock
674 		 */
675 
676 		/* Case 0 */
677 		if ((lf->lf_end != -1 && start > lf->lf_end) ||
678 		    (end != -1 && lf->lf_start > end)) {
679 			DPRINTF(("no overlap\n"), DEBUG_FINDOVR);
680 			if ((type & SELF) && end != -1 && lf->lf_start > end)
681 				return (0);
682 			*overlap = lf = TAILQ_NEXT(lf, lf_entry);
683 			continue;
684 		}
685 		/* Case 1 */
686 		if ((lf->lf_start == start) && (lf->lf_end == end)) {
687 			DPRINTF(("overlap == lock\n"), DEBUG_FINDOVR);
688 			return (1);
689 		}
690 		/* Case 2 */
691 		if ((lf->lf_start <= start) &&
692 		    (lf->lf_end == -1 || (end != -1 && lf->lf_end >= end))) {
693 			DPRINTF(("overlap contains lock\n"), DEBUG_FINDOVR);
694 			return (2);
695 		}
696 		/* Case 3 */
697 		if (start <= lf->lf_start &&
698 		    (end == -1 || (lf->lf_end != -1 && end >= lf->lf_end))) {
699 			DPRINTF(("lock contains overlap\n"), DEBUG_FINDOVR);
700 			return (3);
701 		}
702 		/* Case 4 */
703 		if ((lf->lf_start < start) &&
704 		    ((lf->lf_end >= start) || (lf->lf_end == -1))) {
705 			DPRINTF(("overlap starts before lock\n"),
706 			    DEBUG_FINDOVR);
707 			return (4);
708 		}
709 		/* Case 5 */
710 		if ((lf->lf_start > start) && (end != -1) &&
711 		    ((lf->lf_end > end) || (lf->lf_end == -1))) {
712 			DPRINTF(("overlap ends after lock\n"), DEBUG_FINDOVR);
713 			return (5);
714 		}
715 		panic("lf_findoverlap: default");
716 	}
717 	return (0);
718 }
719 
720 /*
721  * Purge all locks associated with the given lock state.
722  */
723 void
lf_purgelocks(struct lockf_state ** state)724 lf_purgelocks(struct lockf_state **state)
725 {
726 	struct lockf_state *ls;
727 	struct lockf *lock;
728 
729 	rw_enter_write(&lockf_lock);
730 
731 	ls = *state;
732 	if (ls == NULL)
733 		goto out;
734 
735 	ls_ref(ls);
736 
737 	/* Interrupt blocked locks and wait for all of them to finish. */
738 	TAILQ_FOREACH(lock, &ls->ls_locks, lf_entry) {
739 		LFPRINT(("lf_purgelocks: wakeup", lock), DEBUG_SETLOCK);
740 		lf_wakelock(lock, F_INTR);
741 	}
742 	while (!TAILQ_EMPTY(&ls->ls_pending))
743 		rwsleep_nsec(ls, &lockf_lock, PLOCK, "lockfp", INFSLP);
744 
745 	/*
746 	 * Any remaining locks cannot block other locks at this point and can
747 	 * safely be removed.
748 	 */
749 	while ((lock = TAILQ_FIRST(&ls->ls_locks))) {
750 		TAILQ_REMOVE(&ls->ls_locks, lock, lf_entry);
751 		lf_free(lock);
752 	}
753 
754 	/* This is the last expected thread to hold a lock state reference. */
755 	KASSERT(ls->ls_refs == 1);
756 	ls_rele(ls);
757 
758 out:
759 	rw_exit_write(&lockf_lock);
760 }
761 
762 /*
763  * Split a lock and a contained region into
764  * two or three locks as necessary.
765  */
766 void
lf_split(struct lockf * lock1,struct lockf * lock2)767 lf_split(struct lockf *lock1, struct lockf *lock2)
768 {
769 	struct lockf *splitlock;
770 
771 	rw_assert_wrlock(&lockf_lock);
772 
773 	LFPRINT(("lf_split", lock1), DEBUG_SPLIT);
774 	LFPRINT(("splitting from", lock2), DEBUG_SPLIT);
775 
776 	/*
777 	 * Check to see if splitting into only two pieces.
778 	 */
779 	if (lock1->lf_start == lock2->lf_start) {
780 		lock1->lf_start = lock2->lf_end + 1;
781 		TAILQ_INSERT_BEFORE(lock1, lock2, lf_entry);
782 		return;
783 	}
784 	if (lock1->lf_end == lock2->lf_end) {
785 		lock1->lf_end = lock2->lf_start - 1;
786 		TAILQ_INSERT_AFTER(&lock1->lf_state->ls_locks, lock1, lock2,
787 		    lf_entry);
788 		return;
789 	}
790 	/*
791 	 * Make a new lock consisting of the last part of
792 	 * the encompassing lock
793 	 */
794 	splitlock = lf_alloc(lock1->lf_uid, 0);
795 	splitlock->lf_flags = lock1->lf_flags;
796 	splitlock->lf_type = lock1->lf_type;
797 	splitlock->lf_start = lock2->lf_end + 1;
798 	splitlock->lf_end = lock1->lf_end;
799 	splitlock->lf_id = lock1->lf_id;
800 	splitlock->lf_state = lock1->lf_state;
801 	splitlock->lf_blk = NULL;
802 	splitlock->lf_pid = lock1->lf_pid;
803 	TAILQ_INIT(&splitlock->lf_blkhd);
804 	ls_ref(splitlock->lf_state);
805 	lock1->lf_end = lock2->lf_start - 1;
806 
807 	TAILQ_INSERT_AFTER(&lock1->lf_state->ls_locks, lock1, lock2, lf_entry);
808 	TAILQ_INSERT_AFTER(&lock1->lf_state->ls_locks, lock2, splitlock,
809 	    lf_entry);
810 }
811 
812 /*
813  * Wakeup a blocklist
814  */
815 void
lf_wakelock(struct lockf * lock,int flags)816 lf_wakelock(struct lockf *lock, int flags)
817 {
818 	struct lockf *wakelock;
819 
820 	rw_assert_wrlock(&lockf_lock);
821 
822 	while ((wakelock = TAILQ_FIRST(&lock->lf_blkhd))) {
823 		TAILQ_REMOVE(&lock->lf_blkhd, wakelock, lf_block);
824 		wakelock->lf_blk = NULL;
825 		wakelock->lf_flags |= flags;
826 		wakeup_one(wakelock);
827 	}
828 }
829 
830 /*
831  * Returns non-zero if the given lock would cause a deadlock.
832  */
833 int
lf_deadlock(struct lockf * lock)834 lf_deadlock(struct lockf *lock)
835 {
836 	struct lockf *block, *lf, *pending;
837 
838 	lf = TAILQ_FIRST(&lock->lf_state->ls_locks);
839 	for (; (block = lf_getblock(lf, lock)) != NULL;
840 	    lf = TAILQ_NEXT(block, lf_entry)) {
841 		if ((block->lf_flags & F_POSIX) == 0)
842 			continue;
843 
844 		TAILQ_FOREACH(pending, &lock->lf_state->ls_pending, lf_entry) {
845 			if (pending->lf_blk == NULL)
846 				continue; /* lock already unblocked */
847 
848 			if (pending->lf_pid == block->lf_pid &&
849 			    pending->lf_blk->lf_pid == lock->lf_pid)
850 				return (1);
851 		}
852 	}
853 
854 	return (0);
855 }
856 
857 #ifdef LOCKF_DEBUG
858 /*
859  * Print out a lock.
860  */
861 void
lf_print(const char * tag,struct lockf * lock)862 lf_print(const char *tag, struct lockf *lock)
863 {
864 	struct lockf	*block;
865 
866 	if (tag)
867 		printf("%s: ", tag);
868 	printf("lock %p", lock);
869 	if (lock == NULL) {
870 		printf("\n");
871 		return;
872 	}
873 	printf(", %s %p %s, start %lld, end %lld",
874 		lock->lf_flags & F_POSIX ? "posix" : "flock",
875 		lock->lf_id,
876 		lock->lf_type == F_RDLCK ? "shared" :
877 		lock->lf_type == F_WRLCK ? "exclusive" :
878 		lock->lf_type == F_UNLCK ? "unlock" :
879 		"unknown", lock->lf_start, lock->lf_end);
880 	printf(", next %p, state %p",
881 	    TAILQ_NEXT(lock, lf_entry), lock->lf_state);
882 	block = TAILQ_FIRST(&lock->lf_blkhd);
883 	if (block)
884 		printf(", block");
885 	TAILQ_FOREACH(block, &lock->lf_blkhd, lf_block)
886 		printf(" %p,", block);
887 	printf("\n");
888 }
889 
890 void
lf_printlist(const char * tag,struct lockf * lock)891 lf_printlist(const char *tag, struct lockf *lock)
892 {
893 	struct lockf *lf;
894 
895 	printf("%s: Lock list:\n", tag);
896 	TAILQ_FOREACH(lf, &lock->lf_state->ls_locks, lf_entry) {
897 		if (lock == lf)
898 			printf(" * ");
899 		else
900 			printf("   ");
901 		lf_print(NULL, lf);
902 	}
903 }
904 #endif /* LOCKF_DEBUG */
905