vfs_lockf.c revision 1.42 
1/*	$NetBSD: vfs_lockf.c,v 1.42 2005/05/09 23:40:20 christos 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 1.42 2005/05/09 23:40:20 christos 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
50POOL_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 */
57int maxlockdepth = MAXDEPTH;
58
59#ifdef LOCKF_DEBUG
60int	lockf_debug = 0;
61#endif
62
63#define NOLOCKF (struct lockf *)0
64#define SELF	0x1
65#define OTHERS	0x2
66
67static int lf_clearlock(struct lockf *, struct lockf **);
68static int lf_findoverlap(struct lockf *,
69	    struct lockf *, int, struct lockf ***, struct lockf **);
70static struct lockf *lf_getblock(struct lockf *);
71static int lf_getlock(struct lockf *, struct flock *);
72static int lf_setlock(struct lockf *, struct lockf **, struct simplelock *);
73static void lf_split(struct lockf *, struct lockf *, struct lockf **);
74static void lf_wakelock(struct lockf *);
75static struct lockf *lf_alloc(uid_t, int);
76static void lf_free(struct lockf *);
77
78
79#ifdef LOCKF_DEBUG
80static void lf_print(char *, struct lockf *);
81static void lf_printlist(char *, struct lockf *);
82#endif
83
84/*
85 * XXX TODO
86 * Misc cleanups: "caddr_t id" should be visible in the API as a
87 * "struct proc *".
88 * (This requires rototilling all VFS's which support advisory locking).
89 */
90
91/*
92 * If there's a lot of lock contention on a single vnode, locking
93 * schemes which allow for more paralleism would be needed.  Given how
94 * infrequently byte-range locks are actually used in typical BSD
95 * code, a more complex approach probably isn't worth it.
96 */
97
98/*
99 * We enforce a limit on locks by uid, so that a single user cannot
100 * run the kernel out of memory.  For now, the limit is pretty coarse.
101 * There is no limit on root.
102 *
103 * Splitting a lock will always succeed, regardless of current allocations.
104 * If you're slightly above the limit, we still have to permit an allocation
105 * so that the unlock can succeed.  If the unlocking causes too many splits,
106 * however, you're totally cutoff.
107 */
108int maxlocksperuid = 1024;
109
110/*
111 * 3 options for allowfail.
112 * 0 - always allocate.  1 - cutoff at limit.  2 - cutoff at double limit.
113 */
114struct lockf *
115lf_alloc(uid_t uid, int allowfail)
116{
117	struct uidinfo *uip;
118	struct lockf *lock;
119	int s;
120
121	uip = uid_find(uid);
122	UILOCK(uip, s);
123	if (uid && allowfail && uip->ui_lockcnt >
124	    (allowfail == 1 ? maxlocksperuid : (maxlocksperuid * 2))) {
125		UIUNLOCK(uip, s);
126		return NULL;
127	}
128	uip->ui_lockcnt++;
129	UIUNLOCK(uip, s);
130	lock = pool_get(&lockfpool, PR_WAITOK);
131	lock->lf_uid = uid;
132	return lock;
133}
134
135void
136lf_free(struct lockf *lock)
137{
138	struct uidinfo *uip;
139	int s;
140
141	uip = uid_find(lock->lf_uid);
142	UILOCK(uip, s);
143	uip->ui_lockcnt--;
144	UIUNLOCK(uip, s);
145	pool_put(&lockfpool, lock);
146}
147
148/*
149 * Do an advisory lock operation.
150 */
151int
152lf_advlock(struct vop_advlock_args *ap, struct lockf **head, off_t size)
153{
154	struct proc *p = curproc;
155	struct flock *fl = ap->a_fl;
156	struct lockf *lock = NULL;
157	struct lockf *sparelock;
158	struct simplelock *interlock = &ap->a_vp->v_interlock;
159	off_t start, end;
160	int error = 0;
161
162	/*
163	 * Convert the flock structure into a start and end.
164	 */
165	switch (fl->l_whence) {
166	case SEEK_SET:
167	case SEEK_CUR:
168		/*
169		 * Caller is responsible for adding any necessary offset
170		 * when SEEK_CUR is used.
171		 */
172		start = fl->l_start;
173		break;
174
175	case SEEK_END:
176		start = size + fl->l_start;
177		break;
178
179	default:
180		return EINVAL;
181	}
182	if (start < 0)
183		return EINVAL;
184
185	/*
186	 * allocate locks before acquire simple lock.
187	 * we need two locks in the worst case.
188	 */
189	switch (ap->a_op) {
190	case F_SETLK:
191	case F_UNLCK:
192		/*
193		 * XXX for F_UNLCK case, we can re-use lock.
194		 */
195		if ((fl->l_type & F_FLOCK) == 0) {
196			/*
197			 * byte-range lock might need one more lock.
198			 */
199			sparelock = lf_alloc(p->p_ucred->cr_uid, 0);
200			if (sparelock == NULL) {
201				error = ENOMEM;
202				goto quit;
203			}
204			break;
205		}
206		/* FALLTHROUGH */
207
208	case F_GETLK:
209		sparelock = NULL;
210		break;
211
212	default:
213		return EINVAL;
214	}
215
216	lock = lf_alloc(p->p_ucred->cr_uid, ap->a_op != F_UNLCK ? 1 : 2);
217	if (lock == NULL) {
218		error = ENOMEM;
219		goto quit;
220	}
221
222	simple_lock(interlock);
223
224	/*
225	 * Avoid the common case of unlocking when inode has no locks.
226	 */
227	if (*head == (struct lockf *)0) {
228		if (ap->a_op != F_SETLK) {
229			fl->l_type = F_UNLCK;
230			error = 0;
231			goto quit_unlock;
232		}
233	}
234
235	if (fl->l_len == 0)
236		end = -1;
237	else
238		end = start + fl->l_len - 1;
239	/*
240	 * Create the lockf structure.
241	 */
242	lock->lf_start = start;
243	lock->lf_end = end;
244	/* XXX NJWLWP
245	 * I don't want to make the entire VFS universe use LWPs, because
246	 * they don't need them, for the most part. This is an exception,
247	 * and a kluge.
248	 */
249
250	lock->lf_head = head;
251	lock->lf_type = fl->l_type;
252	lock->lf_next = (struct lockf *)0;
253	TAILQ_INIT(&lock->lf_blkhd);
254	lock->lf_flags = ap->a_flags;
255	if (lock->lf_flags & F_POSIX) {
256		KASSERT(curproc == (struct proc *)ap->a_id);
257	}
258	lock->lf_id = (struct proc *)ap->a_id;
259	lock->lf_lwp = curlwp;
260
261	/*
262	 * Do the requested operation.
263	 */
264	switch (ap->a_op) {
265
266	case F_SETLK:
267		error = lf_setlock(lock, &sparelock, interlock);
268		lock = NULL; /* lf_setlock freed it */
269		break;
270
271	case F_UNLCK:
272		error = lf_clearlock(lock, &sparelock);
273		break;
274
275	case F_GETLK:
276		error = lf_getlock(lock, fl);
277		break;
278
279	default:
280		break;
281		/* NOTREACHED */
282	}
283
284quit_unlock:
285	simple_unlock(interlock);
286quit:
287	if (lock)
288		lf_free(lock);
289	if (sparelock)
290		lf_free(sparelock);
291
292	return error;
293}
294
295/*
296 * Set a byte-range lock.
297 */
298static int
299lf_setlock(struct lockf *lock, struct lockf **sparelock,
300    struct simplelock *interlock)
301{
302	struct lockf *block;
303	struct lockf **head = lock->lf_head;
304	struct lockf **prev, *overlap, *ltmp;
305	static char lockstr[] = "lockf";
306	int ovcase, priority, needtolink, error;
307
308#ifdef LOCKF_DEBUG
309	if (lockf_debug & 1)
310		lf_print("lf_setlock", lock);
311#endif /* LOCKF_DEBUG */
312
313	/*
314	 * Set the priority
315	 */
316	priority = PLOCK;
317	if (lock->lf_type == F_WRLCK)
318		priority += 4;
319	priority |= PCATCH;
320	/*
321	 * Scan lock list for this file looking for locks that would block us.
322	 */
323	while ((block = lf_getblock(lock)) != NULL) {
324		/*
325		 * Free the structure and return if nonblocking.
326		 */
327		if ((lock->lf_flags & F_WAIT) == 0) {
328			lf_free(lock);
329			return EAGAIN;
330		}
331		/*
332		 * We are blocked. Since flock style locks cover
333		 * the whole file, there is no chance for deadlock.
334		 * For byte-range locks we must check for deadlock.
335		 *
336		 * Deadlock detection is done by looking through the
337		 * wait channels to see if there are any cycles that
338		 * involve us. MAXDEPTH is set just to make sure we
339		 * do not go off into neverneverland.
340		 */
341		if ((lock->lf_flags & F_POSIX) &&
342		    (block->lf_flags & F_POSIX)) {
343			struct lwp *wlwp;
344			struct lockf *waitblock;
345			int i = 0;
346
347			/*
348			 * The block is waiting on something.  if_lwp will be
349			 * 0 once the lock is granted, so we terminate the
350			 * loop if we find this.
351			 */
352			wlwp = block->lf_lwp;
353			while (wlwp && (i++ < maxlockdepth)) {
354				waitblock = (struct lockf *)wlwp->l_wchan;
355				/* Get the owner of the blocking lock */
356				waitblock = waitblock->lf_next;
357				if ((waitblock->lf_flags & F_POSIX) == 0)
358					break;
359				wlwp = waitblock->lf_lwp;
360				if (wlwp == lock->lf_lwp) {
361					lf_free(lock);
362					return EDEADLK;
363				}
364			}
365			/*
366			 * If we're still following a dependency chain
367			 * after maxlockdepth iterations, assume we're in
368			 * a cycle to be safe.
369			 */
370			if (i >= maxlockdepth) {
371				lf_free(lock);
372				return EDEADLK;
373			}
374		}
375		/*
376		 * For flock type locks, we must first remove
377		 * any shared locks that we hold before we sleep
378		 * waiting for an exclusive lock.
379		 */
380		if ((lock->lf_flags & F_FLOCK) &&
381		    lock->lf_type == F_WRLCK) {
382			lock->lf_type = F_UNLCK;
383			(void) lf_clearlock(lock, NULL);
384			lock->lf_type = F_WRLCK;
385		}
386		/*
387		 * Add our lock to the blocked list and sleep until we're free.
388		 * Remember who blocked us (for deadlock detection).
389		 */
390		lock->lf_next = block;
391		TAILQ_INSERT_TAIL(&block->lf_blkhd, lock, lf_block);
392#ifdef LOCKF_DEBUG
393		if (lockf_debug & 1) {
394			lf_print("lf_setlock: blocking on", block);
395			lf_printlist("lf_setlock", block);
396		}
397#endif /* LOCKF_DEBUG */
398		error = ltsleep(lock, priority, lockstr, 0, interlock);
399
400		/*
401		 * We may have been awakened by a signal (in
402		 * which case we must remove ourselves from the
403		 * blocked list) and/or by another process
404		 * releasing a lock (in which case we have already
405		 * been removed from the blocked list and our
406		 * lf_next field set to NOLOCKF).
407		 */
408		if (lock->lf_next != NOLOCKF) {
409			TAILQ_REMOVE(&lock->lf_next->lf_blkhd, lock, lf_block);
410			lock->lf_next = NOLOCKF;
411		}
412		if (error) {
413			lf_free(lock);
414			return error;
415		}
416	}
417	/*
418	 * No blocks!!  Add the lock.  Note that we will
419	 * downgrade or upgrade any overlapping locks this
420	 * process already owns.
421	 *
422	 * Skip over locks owned by other processes.
423	 * Handle any locks that overlap and are owned by ourselves.
424	 */
425	lock->lf_lwp = 0;
426	prev = head;
427	block = *head;
428	needtolink = 1;
429	for (;;) {
430		ovcase = lf_findoverlap(block, lock, SELF, &prev, &overlap);
431		if (ovcase)
432			block = overlap->lf_next;
433		/*
434		 * Six cases:
435		 *	0) no overlap
436		 *	1) overlap == lock
437		 *	2) overlap contains lock
438		 *	3) lock contains overlap
439		 *	4) overlap starts before lock
440		 *	5) overlap ends after lock
441		 */
442		switch (ovcase) {
443		case 0: /* no overlap */
444			if (needtolink) {
445				*prev = lock;
446				lock->lf_next = overlap;
447			}
448			break;
449
450		case 1: /* overlap == lock */
451			/*
452			 * If downgrading lock, others may be
453			 * able to acquire it.
454			 */
455			if (lock->lf_type == F_RDLCK &&
456			    overlap->lf_type == F_WRLCK)
457				lf_wakelock(overlap);
458			overlap->lf_type = lock->lf_type;
459			lf_free(lock);
460			lock = overlap; /* for debug output below */
461			break;
462
463		case 2: /* overlap contains lock */
464			/*
465			 * Check for common starting point and different types.
466			 */
467			if (overlap->lf_type == lock->lf_type) {
468				lf_free(lock);
469				lock = overlap; /* for debug output below */
470				break;
471			}
472			if (overlap->lf_start == lock->lf_start) {
473				*prev = lock;
474				lock->lf_next = overlap;
475				overlap->lf_start = lock->lf_end + 1;
476			} else
477				lf_split(overlap, lock, sparelock);
478			lf_wakelock(overlap);
479			break;
480
481		case 3: /* lock contains overlap */
482			/*
483			 * If downgrading lock, others may be able to
484			 * acquire it, otherwise take the list.
485			 */
486			if (lock->lf_type == F_RDLCK &&
487			    overlap->lf_type == F_WRLCK) {
488				lf_wakelock(overlap);
489			} else {
490				while ((ltmp = TAILQ_FIRST(&overlap->lf_blkhd))) {
491					KASSERT(ltmp->lf_next == overlap);
492					TAILQ_REMOVE(&overlap->lf_blkhd, ltmp,
493					    lf_block);
494					ltmp->lf_next = lock;
495					TAILQ_INSERT_TAIL(&lock->lf_blkhd,
496					    ltmp, lf_block);
497				}
498			}
499			/*
500			 * Add the new lock if necessary and delete the overlap.
501			 */
502			if (needtolink) {
503				*prev = lock;
504				lock->lf_next = overlap->lf_next;
505				prev = &lock->lf_next;
506				needtolink = 0;
507			} else
508				*prev = overlap->lf_next;
509			lf_free(overlap);
510			continue;
511
512		case 4: /* overlap starts before lock */
513			/*
514			 * Add lock after overlap on the list.
515			 */
516			lock->lf_next = overlap->lf_next;
517			overlap->lf_next = lock;
518			overlap->lf_end = lock->lf_start - 1;
519			prev = &lock->lf_next;
520			lf_wakelock(overlap);
521			needtolink = 0;
522			continue;
523
524		case 5: /* overlap ends after lock */
525			/*
526			 * Add the new lock before overlap.
527			 */
528			if (needtolink) {
529				*prev = lock;
530				lock->lf_next = overlap;
531			}
532			overlap->lf_start = lock->lf_end + 1;
533			lf_wakelock(overlap);
534			break;
535		}
536		break;
537	}
538#ifdef LOCKF_DEBUG
539	if (lockf_debug & 1) {
540		lf_print("lf_setlock: got the lock", lock);
541		lf_printlist("lf_setlock", lock);
542	}
543#endif /* LOCKF_DEBUG */
544	return 0;
545}
546
547/*
548 * Remove a byte-range lock on an inode.
549 *
550 * Generally, find the lock (or an overlap to that lock)
551 * and remove it (or shrink it), then wakeup anyone we can.
552 */
553static int
554lf_clearlock(struct lockf *unlock, struct lockf **sparelock)
555{
556	struct lockf **head = unlock->lf_head;
557	struct lockf *lf = *head;
558	struct lockf *overlap, **prev;
559	int ovcase;
560
561	if (lf == NOLOCKF)
562		return 0;
563#ifdef LOCKF_DEBUG
564	if (unlock->lf_type != F_UNLCK)
565		panic("lf_clearlock: bad type");
566	if (lockf_debug & 1)
567		lf_print("lf_clearlock", unlock);
568#endif /* LOCKF_DEBUG */
569	prev = head;
570	while ((ovcase = lf_findoverlap(lf, unlock, SELF,
571					&prev, &overlap)) != 0) {
572		/*
573		 * Wakeup the list of locks to be retried.
574		 */
575		lf_wakelock(overlap);
576
577		switch (ovcase) {
578
579		case 1: /* overlap == lock */
580			*prev = overlap->lf_next;
581			lf_free(overlap);
582			break;
583
584		case 2: /* overlap contains lock: split it */
585			if (overlap->lf_start == unlock->lf_start) {
586				overlap->lf_start = unlock->lf_end + 1;
587				break;
588			}
589			lf_split(overlap, unlock, sparelock);
590			overlap->lf_next = unlock->lf_next;
591			break;
592
593		case 3: /* lock contains overlap */
594			*prev = overlap->lf_next;
595			lf = overlap->lf_next;
596			lf_free(overlap);
597			continue;
598
599		case 4: /* overlap starts before lock */
600			overlap->lf_end = unlock->lf_start - 1;
601			prev = &overlap->lf_next;
602			lf = overlap->lf_next;
603			continue;
604
605		case 5: /* overlap ends after lock */
606			overlap->lf_start = unlock->lf_end + 1;
607			break;
608		}
609		break;
610	}
611#ifdef LOCKF_DEBUG
612	if (lockf_debug & 1)
613		lf_printlist("lf_clearlock", unlock);
614#endif /* LOCKF_DEBUG */
615	return 0;
616}
617
618/*
619 * Check whether there is a blocking lock,
620 * and if so return its process identifier.
621 */
622static int
623lf_getlock(struct lockf *lock, struct flock *fl)
624{
625	struct lockf *block;
626
627#ifdef LOCKF_DEBUG
628	if (lockf_debug & 1)
629		lf_print("lf_getlock", lock);
630#endif /* LOCKF_DEBUG */
631
632	if ((block = lf_getblock(lock)) != NULL) {
633		fl->l_type = block->lf_type;
634		fl->l_whence = SEEK_SET;
635		fl->l_start = block->lf_start;
636		if (block->lf_end == -1)
637			fl->l_len = 0;
638		else
639			fl->l_len = block->lf_end - block->lf_start + 1;
640		if (block->lf_flags & F_POSIX)
641			fl->l_pid = ((struct proc *)block->lf_id)->p_pid;
642		else
643			fl->l_pid = -1;
644	} else {
645		fl->l_type = F_UNLCK;
646	}
647	return 0;
648}
649
650/*
651 * Walk the list of locks for an inode and
652 * return the first blocking lock.
653 */
654static struct lockf *
655lf_getblock(struct lockf *lock)
656{
657	struct lockf **prev, *overlap, *lf = *(lock->lf_head);
658
659	prev = lock->lf_head;
660	while (lf_findoverlap(lf, lock, OTHERS, &prev, &overlap) != 0) {
661		/*
662		 * We've found an overlap, see if it blocks us
663		 */
664		if ((lock->lf_type == F_WRLCK || overlap->lf_type == F_WRLCK))
665			return overlap;
666		/*
667		 * Nope, point to the next one on the list and
668		 * see if it blocks us
669		 */
670		lf = overlap->lf_next;
671	}
672	return NOLOCKF;
673}
674
675/*
676 * Walk the list of locks for an inode to
677 * find an overlapping lock (if any).
678 *
679 * NOTE: this returns only the FIRST overlapping lock.  There
680 *	 may be more than one.
681 */
682static int
683lf_findoverlap(struct lockf *lf, struct lockf *lock, int type,
684    struct lockf ***prev, struct lockf **overlap)
685{
686	off_t start, end;
687
688	*overlap = lf;
689	if (lf == NOLOCKF)
690		return 0;
691#ifdef LOCKF_DEBUG
692	if (lockf_debug & 2)
693		lf_print("lf_findoverlap: looking for overlap in", lock);
694#endif /* LOCKF_DEBUG */
695	start = lock->lf_start;
696	end = lock->lf_end;
697	while (lf != NOLOCKF) {
698		if (((type == SELF) && lf->lf_id != lock->lf_id) ||
699		    ((type == OTHERS) && lf->lf_id == lock->lf_id)) {
700			*prev = &lf->lf_next;
701			*overlap = lf = lf->lf_next;
702			continue;
703		}
704#ifdef LOCKF_DEBUG
705		if (lockf_debug & 2)
706			lf_print("\tchecking", lf);
707#endif /* LOCKF_DEBUG */
708		/*
709		 * OK, check for overlap
710		 *
711		 * Six cases:
712		 *	0) no overlap
713		 *	1) overlap == lock
714		 *	2) overlap contains lock
715		 *	3) lock contains overlap
716		 *	4) overlap starts before lock
717		 *	5) overlap ends after lock
718		 */
719		if ((lf->lf_end != -1 && start > lf->lf_end) ||
720		    (end != -1 && lf->lf_start > end)) {
721			/* Case 0 */
722#ifdef LOCKF_DEBUG
723			if (lockf_debug & 2)
724				printf("no overlap\n");
725#endif /* LOCKF_DEBUG */
726			if ((type & SELF) && end != -1 && lf->lf_start > end)
727				return 0;
728			*prev = &lf->lf_next;
729			*overlap = lf = lf->lf_next;
730			continue;
731		}
732		if ((lf->lf_start == start) && (lf->lf_end == end)) {
733			/* Case 1 */
734#ifdef LOCKF_DEBUG
735			if (lockf_debug & 2)
736				printf("overlap == lock\n");
737#endif /* LOCKF_DEBUG */
738			return 1;
739		}
740		if ((lf->lf_start <= start) &&
741		    (end != -1) &&
742		    ((lf->lf_end >= end) || (lf->lf_end == -1))) {
743			/* Case 2 */
744#ifdef LOCKF_DEBUG
745			if (lockf_debug & 2)
746				printf("overlap contains lock\n");
747#endif /* LOCKF_DEBUG */
748			return 2;
749		}
750		if (start <= lf->lf_start &&
751		           (end == -1 ||
752			   (lf->lf_end != -1 && end >= lf->lf_end))) {
753			/* Case 3 */
754#ifdef LOCKF_DEBUG
755			if (lockf_debug & 2)
756				printf("lock contains overlap\n");
757#endif /* LOCKF_DEBUG */
758			return 3;
759		}
760		if ((lf->lf_start < start) &&
761			((lf->lf_end >= start) || (lf->lf_end == -1))) {
762			/* Case 4 */
763#ifdef LOCKF_DEBUG
764			if (lockf_debug & 2)
765				printf("overlap starts before lock\n");
766#endif /* LOCKF_DEBUG */
767			return 4;
768		}
769		if ((lf->lf_start > start) &&
770			(end != -1) &&
771			((lf->lf_end > end) || (lf->lf_end == -1))) {
772			/* Case 5 */
773#ifdef LOCKF_DEBUG
774			if (lockf_debug & 2)
775				printf("overlap ends after lock\n");
776#endif /* LOCKF_DEBUG */
777			return 5;
778		}
779		panic("lf_findoverlap: default");
780	}
781	return 0;
782}
783
784/*
785 * Split a lock and a contained region into
786 * two or three locks as necessary.
787 */
788static void
789lf_split(struct lockf *lock1, struct lockf *lock2, struct lockf **sparelock)
790{
791	struct lockf *splitlock;
792
793#ifdef LOCKF_DEBUG
794	if (lockf_debug & 2) {
795		lf_print("lf_split", lock1);
796		lf_print("splitting from", lock2);
797	}
798#endif /* LOCKF_DEBUG */
799	/*
800	 * Check to see if spliting into only two pieces.
801	 */
802	if (lock1->lf_start == lock2->lf_start) {
803		lock1->lf_start = lock2->lf_end + 1;
804		lock2->lf_next = lock1;
805		return;
806	}
807	if (lock1->lf_end == lock2->lf_end) {
808		lock1->lf_end = lock2->lf_start - 1;
809		lock2->lf_next = lock1->lf_next;
810		lock1->lf_next = lock2;
811		return;
812	}
813	/*
814	 * Make a new lock consisting of the last part of
815	 * the encompassing lock
816	 */
817	splitlock = *sparelock;
818	*sparelock = NULL;
819	memcpy(splitlock, lock1, sizeof(*splitlock));
820	splitlock->lf_start = lock2->lf_end + 1;
821	TAILQ_INIT(&splitlock->lf_blkhd);
822	lock1->lf_end = lock2->lf_start - 1;
823	/*
824	 * OK, now link it in
825	 */
826	splitlock->lf_next = lock1->lf_next;
827	lock2->lf_next = splitlock;
828	lock1->lf_next = lock2;
829}
830
831/*
832 * Wakeup a blocklist
833 */
834static void
835lf_wakelock(struct lockf *listhead)
836{
837	struct lockf *wakelock;
838
839	while ((wakelock = TAILQ_FIRST(&listhead->lf_blkhd))) {
840		KASSERT(wakelock->lf_next == listhead);
841		TAILQ_REMOVE(&listhead->lf_blkhd, wakelock, lf_block);
842		wakelock->lf_next = NOLOCKF;
843#ifdef LOCKF_DEBUG
844		if (lockf_debug & 2)
845			lf_print("lf_wakelock: awakening", wakelock);
846#endif
847		wakeup(wakelock);
848	}
849}
850
851#ifdef LOCKF_DEBUG
852/*
853 * Print out a lock.
854 */
855static void
856lf_print(char *tag, struct lockf *lock)
857{
858
859	printf("%s: lock %p for ", tag, lock);
860	if (lock->lf_flags & F_POSIX)
861		printf("proc %d", ((struct proc *)lock->lf_id)->p_pid);
862	else
863		printf("file 0x%p", (struct file *)lock->lf_id);
864	printf(" %s, start %qx, end %qx",
865		lock->lf_type == F_RDLCK ? "shared" :
866		lock->lf_type == F_WRLCK ? "exclusive" :
867		lock->lf_type == F_UNLCK ? "unlock" :
868		"unknown", lock->lf_start, lock->lf_end);
869	if (TAILQ_FIRST(&lock->lf_blkhd))
870		printf(" block %p\n", TAILQ_FIRST(&lock->lf_blkhd));
871	else
872		printf("\n");
873}
874
875static void
876lf_printlist(char *tag, struct lockf *lock)
877{
878	struct lockf *lf, *blk;
879
880	printf("%s: Lock list:\n", tag);
881	for (lf = *lock->lf_head; lf; lf = lf->lf_next) {
882		printf("\tlock %p for ", lf);
883		if (lf->lf_flags & F_POSIX)
884			printf("proc %d", ((struct proc *)lf->lf_id)->p_pid);
885		else
886			printf("file 0x%p", (struct file *)lf->lf_id);
887		printf(", %s, start %qx, end %qx",
888			lf->lf_type == F_RDLCK ? "shared" :
889			lf->lf_type == F_WRLCK ? "exclusive" :
890			lf->lf_type == F_UNLCK ? "unlock" :
891			"unknown", lf->lf_start, lf->lf_end);
892		TAILQ_FOREACH(blk, &lf->lf_blkhd, lf_block) {
893			if (blk->lf_flags & F_POSIX)
894				printf("proc %d",
895				    ((struct proc *)blk->lf_id)->p_pid);
896			else
897				printf("file 0x%p", (struct file *)blk->lf_id);
898			printf(", %s, start %qx, end %qx",
899				blk->lf_type == F_RDLCK ? "shared" :
900				blk->lf_type == F_WRLCK ? "exclusive" :
901				blk->lf_type == F_UNLCK ? "unlock" :
902				"unknown", blk->lf_start, blk->lf_end);
903			if (TAILQ_FIRST(&blk->lf_blkhd))
904				 panic("lf_printlist: bad list");
905		}
906		printf("\n");
907	}
908}
909#endif /* LOCKF_DEBUG */
910