Home | History | Annotate | Line # | Download | only in kern
vfs_lockf.c revision 1.33
      1 /*	$NetBSD: vfs_lockf.c,v 1.33 2003/08/07 16:32:02 agc Exp $	*/
      2 
      3 /*
      4  * Copyright (c) 1982, 1986, 1989, 1993
      5  *	The Regents of the University of California.  All rights reserved.
      6  *
      7  * This code is derived from software contributed to Berkeley by
      8  * Scooter Morris at Genentech Inc.
      9  *
     10  * Redistribution and use in source and binary forms, with or without
     11  * modification, are permitted provided that the following conditions
     12  * are met:
     13  * 1. Redistributions of source code must retain the above copyright
     14  *    notice, this list of conditions and the following disclaimer.
     15  * 2. Redistributions in binary form must reproduce the above copyright
     16  *    notice, this list of conditions and the following disclaimer in the
     17  *    documentation and/or other materials provided with the distribution.
     18  * 3. Neither the name of the University nor the names of its contributors
     19  *    may be used to endorse or promote products derived from this software
     20  *    without specific prior written permission.
     21  *
     22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     32  * SUCH DAMAGE.
     33  *
     34  *	@(#)ufs_lockf.c	8.4 (Berkeley) 10/26/94
     35  */
     36 
     37 #include <sys/cdefs.h>
     38 __KERNEL_RCSID(0, "$NetBSD: vfs_lockf.c,v 1.33 2003/08/07 16:32:02 agc Exp $");
     39 
     40 #include <sys/param.h>
     41 #include <sys/systm.h>
     42 #include <sys/kernel.h>
     43 #include <sys/file.h>
     44 #include <sys/proc.h>
     45 #include <sys/vnode.h>
     46 #include <sys/malloc.h>
     47 #include <sys/fcntl.h>
     48 #include <sys/lockf.h>
     49 
     50 MALLOC_DEFINE(M_LOCKF, "lockf", "Byte-range locking structures");
     51 
     52 /*
     53  * This variable controls the maximum number of processes that will
     54  * be checked in doing deadlock detection.
     55  */
     56 int maxlockdepth = MAXDEPTH;
     57 
     58 #ifdef LOCKF_DEBUG
     59 int	lockf_debug = 0;
     60 #endif
     61 
     62 #define NOLOCKF (struct lockf *)0
     63 #define SELF	0x1
     64 #define OTHERS	0x2
     65 
     66 static int lf_clearlock(struct lockf *, struct lockf **);
     67 static int lf_findoverlap(struct lockf *,
     68 	    struct lockf *, int, struct lockf ***, struct lockf **);
     69 static struct lockf *lf_getblock(struct lockf *);
     70 static int lf_getlock(struct lockf *, struct flock *);
     71 static int lf_setlock(struct lockf *, struct lockf **, struct simplelock *);
     72 static void lf_split(struct lockf *, struct lockf *, struct lockf **);
     73 static void lf_wakelock(struct lockf *);
     74 
     75 #ifdef LOCKF_DEBUG
     76 static void lf_print(char *, struct lockf *);
     77 static void lf_printlist(char *, struct lockf *);
     78 #endif
     79 
     80 /*
     81  * XXX TODO
     82  * Misc cleanups: "caddr_t id" should be visible in the API as a
     83  * "struct proc *".
     84  * (This requires rototilling all VFS's which support advisory locking).
     85  *
     86  * Use pools for lock allocation.
     87  */
     88 
     89 /*
     90  * If there's a lot of lock contention on a single vnode, locking
     91  * schemes which allow for more paralleism would be needed.  Given how
     92  * infrequently byte-range locks are actually used in typical BSD
     93  * code, a more complex approach probably isn't worth it.
     94  */
     95 
     96 /*
     97  * Do an advisory lock operation.
     98  */
     99 int
    100 lf_advlock(struct vop_advlock_args *ap, struct lockf **head, off_t size)
    101 {
    102 	struct flock *fl = ap->a_fl;
    103 	struct lockf *lock = NULL;
    104 	struct lockf *sparelock;
    105 	struct simplelock *interlock = &ap->a_vp->v_interlock;
    106 	off_t start, end;
    107 	int error;
    108 
    109 	/*
    110 	 * Convert the flock structure into a start and end.
    111 	 */
    112 	switch (fl->l_whence) {
    113 	case SEEK_SET:
    114 	case SEEK_CUR:
    115 		/*
    116 		 * Caller is responsible for adding any necessary offset
    117 		 * when SEEK_CUR is used.
    118 		 */
    119 		start = fl->l_start;
    120 		break;
    121 
    122 	case SEEK_END:
    123 		start = size + fl->l_start;
    124 		break;
    125 
    126 	default:
    127 		return EINVAL;
    128 	}
    129 	if (start < 0)
    130 		return EINVAL;
    131 
    132 	/*
    133 	 * allocate locks before acquire simple lock.
    134 	 * we need two locks in the worst case.
    135 	 */
    136 	switch (ap->a_op) {
    137 	case F_SETLK:
    138 	case F_UNLCK:
    139 		/*
    140 		 * XXX for F_UNLCK case, we can re-use lock.
    141 		 */
    142 		if ((fl->l_type & F_FLOCK) == 0) {
    143 			/*
    144 			 * byte-range lock might need one more lock.
    145 			 */
    146 			MALLOC(sparelock, struct lockf *, sizeof(*lock),
    147 			    M_LOCKF, M_WAITOK);
    148 			if (sparelock == NULL) {
    149 				error = ENOMEM;
    150 				goto quit;
    151 			}
    152 			break;
    153 		}
    154 		/* FALLTHROUGH */
    155 
    156 	case F_GETLK:
    157 		sparelock = NULL;
    158 		break;
    159 
    160 	default:
    161 		return EINVAL;
    162 	}
    163 
    164 	MALLOC(lock, struct lockf *, sizeof(*lock), M_LOCKF, M_WAITOK);
    165 	if (lock == NULL) {
    166 		error = ENOMEM;
    167 		goto quit;
    168 	}
    169 
    170 	simple_lock(interlock);
    171 
    172 	/*
    173 	 * Avoid the common case of unlocking when inode has no locks.
    174 	 */
    175 	if (*head == (struct lockf *)0) {
    176 		if (ap->a_op != F_SETLK) {
    177 			fl->l_type = F_UNLCK;
    178 			error = 0;
    179 			goto quit_unlock;
    180 		}
    181 	}
    182 
    183 	if (fl->l_len == 0)
    184 		end = -1;
    185 	else
    186 		end = start + fl->l_len - 1;
    187 	/*
    188 	 * Create the lockf structure.
    189 	 */
    190 	lock->lf_start = start;
    191 	lock->lf_end = end;
    192 	/* XXX NJWLWP
    193 	 * I don't want to make the entire VFS universe use LWPs, because
    194 	 * they don't need them, for the most part. This is an exception,
    195 	 * and a kluge.
    196 	 */
    197 
    198 	lock->lf_head = head;
    199 	lock->lf_type = fl->l_type;
    200 	lock->lf_next = (struct lockf *)0;
    201 	TAILQ_INIT(&lock->lf_blkhd);
    202 	lock->lf_flags = ap->a_flags;
    203 	if (lock->lf_flags & F_POSIX) {
    204 		KASSERT(curproc == (struct proc *)ap->a_id);
    205 	}
    206 	lock->lf_id = (struct proc *)ap->a_id;
    207 	lock->lf_lwp = curlwp;
    208 
    209 	/*
    210 	 * Do the requested operation.
    211 	 */
    212 	switch (ap->a_op) {
    213 
    214 	case F_SETLK:
    215 		error = lf_setlock(lock, &sparelock, interlock);
    216 		lock = NULL; /* lf_setlock freed it */
    217 		break;
    218 
    219 	case F_UNLCK:
    220 		error = lf_clearlock(lock, &sparelock);
    221 		break;
    222 
    223 	case F_GETLK:
    224 		error = lf_getlock(lock, fl);
    225 		break;
    226 
    227 	default:
    228 		break;
    229 		/* NOTREACHED */
    230 	}
    231 
    232 quit_unlock:
    233 	simple_unlock(interlock);
    234 quit:
    235 	if (lock)
    236 		FREE(lock, M_LOCKF);
    237 	if (sparelock)
    238 		FREE(sparelock, M_LOCKF);
    239 
    240 	return error;
    241 }
    242 
    243 /*
    244  * Set a byte-range lock.
    245  */
    246 static int
    247 lf_setlock(struct lockf *lock, struct lockf **sparelock,
    248     struct simplelock *interlock)
    249 {
    250 	struct lockf *block;
    251 	struct lockf **head = lock->lf_head;
    252 	struct lockf **prev, *overlap, *ltmp;
    253 	static char lockstr[] = "lockf";
    254 	int ovcase, priority, needtolink, error;
    255 
    256 #ifdef LOCKF_DEBUG
    257 	if (lockf_debug & 1)
    258 		lf_print("lf_setlock", lock);
    259 #endif /* LOCKF_DEBUG */
    260 
    261 	/*
    262 	 * Set the priority
    263 	 */
    264 	priority = PLOCK;
    265 	if (lock->lf_type == F_WRLCK)
    266 		priority += 4;
    267 	priority |= PCATCH;
    268 	/*
    269 	 * Scan lock list for this file looking for locks that would block us.
    270 	 */
    271 	while ((block = lf_getblock(lock)) != NULL) {
    272 		/*
    273 		 * Free the structure and return if nonblocking.
    274 		 */
    275 		if ((lock->lf_flags & F_WAIT) == 0) {
    276 			FREE(lock, M_LOCKF);
    277 			return EAGAIN;
    278 		}
    279 		/*
    280 		 * We are blocked. Since flock style locks cover
    281 		 * the whole file, there is no chance for deadlock.
    282 		 * For byte-range locks we must check for deadlock.
    283 		 *
    284 		 * Deadlock detection is done by looking through the
    285 		 * wait channels to see if there are any cycles that
    286 		 * involve us. MAXDEPTH is set just to make sure we
    287 		 * do not go off into neverneverland.
    288 		 */
    289 		if ((lock->lf_flags & F_POSIX) &&
    290 		    (block->lf_flags & F_POSIX)) {
    291 			struct lwp *wlwp;
    292 			struct lockf *waitblock;
    293 			int i = 0;
    294 
    295 			/*
    296 			 * The block is waiting on something.  if_lwp will be
    297 			 * 0 once the lock is granted, so we terminate the
    298 			 * loop if we find this.
    299 			 */
    300 			wlwp = block->lf_lwp;
    301 			while (wlwp && (i++ < maxlockdepth)) {
    302 				waitblock = (struct lockf *)wlwp->l_wchan;
    303 				/* Get the owner of the blocking lock */
    304 				waitblock = waitblock->lf_next;
    305 				if ((waitblock->lf_flags & F_POSIX) == 0)
    306 					break;
    307 				wlwp = waitblock->lf_lwp;
    308 				if (wlwp == lock->lf_lwp) {
    309 					free(lock, M_LOCKF);
    310 					return EDEADLK;
    311 				}
    312 			}
    313 			/*
    314 			 * If we're still following a dependancy chain
    315 			 * after maxlockdepth iterations, assume we're in
    316 			 * a cycle to be safe.
    317 			 */
    318 			if (i >= maxlockdepth) {
    319 				free(lock, M_LOCKF);
    320 				return EDEADLK;
    321 			}
    322 		}
    323 		/*
    324 		 * For flock type locks, we must first remove
    325 		 * any shared locks that we hold before we sleep
    326 		 * waiting for an exclusive lock.
    327 		 */
    328 		if ((lock->lf_flags & F_FLOCK) &&
    329 		    lock->lf_type == F_WRLCK) {
    330 			lock->lf_type = F_UNLCK;
    331 			(void) lf_clearlock(lock, NULL);
    332 			lock->lf_type = F_WRLCK;
    333 		}
    334 		/*
    335 		 * Add our lock to the blocked list and sleep until we're free.
    336 		 * Remember who blocked us (for deadlock detection).
    337 		 */
    338 		lock->lf_next = block;
    339 		TAILQ_INSERT_TAIL(&block->lf_blkhd, lock, lf_block);
    340 #ifdef LOCKF_DEBUG
    341 		if (lockf_debug & 1) {
    342 			lf_print("lf_setlock: blocking on", block);
    343 			lf_printlist("lf_setlock", block);
    344 		}
    345 #endif /* LOCKF_DEBUG */
    346 		error = ltsleep(lock, priority, lockstr, 0, interlock);
    347 
    348 		/*
    349 		 * We may have been awakened by a signal (in
    350 		 * which case we must remove ourselves from the
    351 		 * blocked list) and/or by another process
    352 		 * releasing a lock (in which case we have already
    353 		 * been removed from the blocked list and our
    354 		 * lf_next field set to NOLOCKF).
    355 		 */
    356 		if (lock->lf_next != NOLOCKF) {
    357 			TAILQ_REMOVE(&lock->lf_next->lf_blkhd, lock, lf_block);
    358 			lock->lf_next = NOLOCKF;
    359 		}
    360 		if (error) {
    361 			free(lock, M_LOCKF);
    362 			return error;
    363 		}
    364 	}
    365 	/*
    366 	 * No blocks!!  Add the lock.  Note that we will
    367 	 * downgrade or upgrade any overlapping locks this
    368 	 * process already owns.
    369 	 *
    370 	 * Skip over locks owned by other processes.
    371 	 * Handle any locks that overlap and are owned by ourselves.
    372 	 */
    373 	lock->lf_lwp = 0;
    374 	prev = head;
    375 	block = *head;
    376 	needtolink = 1;
    377 	for (;;) {
    378 		ovcase = lf_findoverlap(block, lock, SELF, &prev, &overlap);
    379 		if (ovcase)
    380 			block = overlap->lf_next;
    381 		/*
    382 		 * Six cases:
    383 		 *	0) no overlap
    384 		 *	1) overlap == lock
    385 		 *	2) overlap contains lock
    386 		 *	3) lock contains overlap
    387 		 *	4) overlap starts before lock
    388 		 *	5) overlap ends after lock
    389 		 */
    390 		switch (ovcase) {
    391 		case 0: /* no overlap */
    392 			if (needtolink) {
    393 				*prev = lock;
    394 				lock->lf_next = overlap;
    395 			}
    396 			break;
    397 
    398 		case 1: /* overlap == lock */
    399 			/*
    400 			 * If downgrading lock, others may be
    401 			 * able to acquire it.
    402 			 */
    403 			if (lock->lf_type == F_RDLCK &&
    404 			    overlap->lf_type == F_WRLCK)
    405 				lf_wakelock(overlap);
    406 			overlap->lf_type = lock->lf_type;
    407 			FREE(lock, M_LOCKF);
    408 			lock = overlap; /* for debug output below */
    409 			break;
    410 
    411 		case 2: /* overlap contains lock */
    412 			/*
    413 			 * Check for common starting point and different types.
    414 			 */
    415 			if (overlap->lf_type == lock->lf_type) {
    416 				free(lock, M_LOCKF);
    417 				lock = overlap; /* for debug output below */
    418 				break;
    419 			}
    420 			if (overlap->lf_start == lock->lf_start) {
    421 				*prev = lock;
    422 				lock->lf_next = overlap;
    423 				overlap->lf_start = lock->lf_end + 1;
    424 			} else
    425 				lf_split(overlap, lock, sparelock);
    426 			lf_wakelock(overlap);
    427 			break;
    428 
    429 		case 3: /* lock contains overlap */
    430 			/*
    431 			 * If downgrading lock, others may be able to
    432 			 * acquire it, otherwise take the list.
    433 			 */
    434 			if (lock->lf_type == F_RDLCK &&
    435 			    overlap->lf_type == F_WRLCK) {
    436 				lf_wakelock(overlap);
    437 			} else {
    438 				while ((ltmp = TAILQ_FIRST(&overlap->lf_blkhd))) {
    439 					KASSERT(ltmp->lf_next == overlap);
    440 					TAILQ_REMOVE(&overlap->lf_blkhd, ltmp,
    441 					    lf_block);
    442 					ltmp->lf_next = lock;
    443 					TAILQ_INSERT_TAIL(&lock->lf_blkhd,
    444 					    ltmp, lf_block);
    445 				}
    446 			}
    447 			/*
    448 			 * Add the new lock if necessary and delete the overlap.
    449 			 */
    450 			if (needtolink) {
    451 				*prev = lock;
    452 				lock->lf_next = overlap->lf_next;
    453 				prev = &lock->lf_next;
    454 				needtolink = 0;
    455 			} else
    456 				*prev = overlap->lf_next;
    457 			free(overlap, M_LOCKF);
    458 			continue;
    459 
    460 		case 4: /* overlap starts before lock */
    461 			/*
    462 			 * Add lock after overlap on the list.
    463 			 */
    464 			lock->lf_next = overlap->lf_next;
    465 			overlap->lf_next = lock;
    466 			overlap->lf_end = lock->lf_start - 1;
    467 			prev = &lock->lf_next;
    468 			lf_wakelock(overlap);
    469 			needtolink = 0;
    470 			continue;
    471 
    472 		case 5: /* overlap ends after lock */
    473 			/*
    474 			 * Add the new lock before overlap.
    475 			 */
    476 			if (needtolink) {
    477 				*prev = lock;
    478 				lock->lf_next = overlap;
    479 			}
    480 			overlap->lf_start = lock->lf_end + 1;
    481 			lf_wakelock(overlap);
    482 			break;
    483 		}
    484 		break;
    485 	}
    486 #ifdef LOCKF_DEBUG
    487 	if (lockf_debug & 1) {
    488 		lf_print("lf_setlock: got the lock", lock);
    489 		lf_printlist("lf_setlock", lock);
    490 	}
    491 #endif /* LOCKF_DEBUG */
    492 	return 0;
    493 }
    494 
    495 /*
    496  * Remove a byte-range lock on an inode.
    497  *
    498  * Generally, find the lock (or an overlap to that lock)
    499  * and remove it (or shrink it), then wakeup anyone we can.
    500  */
    501 static int
    502 lf_clearlock(struct lockf *unlock, struct lockf **sparelock)
    503 {
    504 	struct lockf **head = unlock->lf_head;
    505 	struct lockf *lf = *head;
    506 	struct lockf *overlap, **prev;
    507 	int ovcase;
    508 
    509 	if (lf == NOLOCKF)
    510 		return 0;
    511 #ifdef LOCKF_DEBUG
    512 	if (unlock->lf_type != F_UNLCK)
    513 		panic("lf_clearlock: bad type");
    514 	if (lockf_debug & 1)
    515 		lf_print("lf_clearlock", unlock);
    516 #endif /* LOCKF_DEBUG */
    517 	prev = head;
    518 	while ((ovcase = lf_findoverlap(lf, unlock, SELF,
    519 					&prev, &overlap)) != 0) {
    520 		/*
    521 		 * Wakeup the list of locks to be retried.
    522 		 */
    523 		lf_wakelock(overlap);
    524 
    525 		switch (ovcase) {
    526 
    527 		case 1: /* overlap == lock */
    528 			*prev = overlap->lf_next;
    529 			FREE(overlap, M_LOCKF);
    530 			break;
    531 
    532 		case 2: /* overlap contains lock: split it */
    533 			if (overlap->lf_start == unlock->lf_start) {
    534 				overlap->lf_start = unlock->lf_end + 1;
    535 				break;
    536 			}
    537 			lf_split(overlap, unlock, sparelock);
    538 			overlap->lf_next = unlock->lf_next;
    539 			break;
    540 
    541 		case 3: /* lock contains overlap */
    542 			*prev = overlap->lf_next;
    543 			lf = overlap->lf_next;
    544 			free(overlap, M_LOCKF);
    545 			continue;
    546 
    547 		case 4: /* overlap starts before lock */
    548 			overlap->lf_end = unlock->lf_start - 1;
    549 			prev = &overlap->lf_next;
    550 			lf = overlap->lf_next;
    551 			continue;
    552 
    553 		case 5: /* overlap ends after lock */
    554 			overlap->lf_start = unlock->lf_end + 1;
    555 			break;
    556 		}
    557 		break;
    558 	}
    559 #ifdef LOCKF_DEBUG
    560 	if (lockf_debug & 1)
    561 		lf_printlist("lf_clearlock", unlock);
    562 #endif /* LOCKF_DEBUG */
    563 	return 0;
    564 }
    565 
    566 /*
    567  * Check whether there is a blocking lock,
    568  * and if so return its process identifier.
    569  */
    570 static int
    571 lf_getlock(struct lockf *lock, struct flock *fl)
    572 {
    573 	struct lockf *block;
    574 
    575 #ifdef LOCKF_DEBUG
    576 	if (lockf_debug & 1)
    577 		lf_print("lf_getlock", lock);
    578 #endif /* LOCKF_DEBUG */
    579 
    580 	if ((block = lf_getblock(lock)) != NULL) {
    581 		fl->l_type = block->lf_type;
    582 		fl->l_whence = SEEK_SET;
    583 		fl->l_start = block->lf_start;
    584 		if (block->lf_end == -1)
    585 			fl->l_len = 0;
    586 		else
    587 			fl->l_len = block->lf_end - block->lf_start + 1;
    588 		if (block->lf_flags & F_POSIX)
    589 			fl->l_pid = ((struct proc *)block->lf_id)->p_pid;
    590 		else
    591 			fl->l_pid = -1;
    592 	} else {
    593 		fl->l_type = F_UNLCK;
    594 	}
    595 	return 0;
    596 }
    597 
    598 /*
    599  * Walk the list of locks for an inode and
    600  * return the first blocking lock.
    601  */
    602 static struct lockf *
    603 lf_getblock(struct lockf *lock)
    604 {
    605 	struct lockf **prev, *overlap, *lf = *(lock->lf_head);
    606 
    607 	prev = lock->lf_head;
    608 	while (lf_findoverlap(lf, lock, OTHERS, &prev, &overlap) != 0) {
    609 		/*
    610 		 * We've found an overlap, see if it blocks us
    611 		 */
    612 		if ((lock->lf_type == F_WRLCK || overlap->lf_type == F_WRLCK))
    613 			return overlap;
    614 		/*
    615 		 * Nope, point to the next one on the list and
    616 		 * see if it blocks us
    617 		 */
    618 		lf = overlap->lf_next;
    619 	}
    620 	return NOLOCKF;
    621 }
    622 
    623 /*
    624  * Walk the list of locks for an inode to
    625  * find an overlapping lock (if any).
    626  *
    627  * NOTE: this returns only the FIRST overlapping lock.  There
    628  *	 may be more than one.
    629  */
    630 static int
    631 lf_findoverlap(struct lockf *lf, struct lockf *lock, int type,
    632     struct lockf ***prev, struct lockf **overlap)
    633 {
    634 	off_t start, end;
    635 
    636 	*overlap = lf;
    637 	if (lf == NOLOCKF)
    638 		return 0;
    639 #ifdef LOCKF_DEBUG
    640 	if (lockf_debug & 2)
    641 		lf_print("lf_findoverlap: looking for overlap in", lock);
    642 #endif /* LOCKF_DEBUG */
    643 	start = lock->lf_start;
    644 	end = lock->lf_end;
    645 	while (lf != NOLOCKF) {
    646 		if (((type == SELF) && lf->lf_id != lock->lf_id) ||
    647 		    ((type == OTHERS) && lf->lf_id == lock->lf_id)) {
    648 			*prev = &lf->lf_next;
    649 			*overlap = lf = lf->lf_next;
    650 			continue;
    651 		}
    652 #ifdef LOCKF_DEBUG
    653 		if (lockf_debug & 2)
    654 			lf_print("\tchecking", lf);
    655 #endif /* LOCKF_DEBUG */
    656 		/*
    657 		 * OK, check for overlap
    658 		 *
    659 		 * Six cases:
    660 		 *	0) no overlap
    661 		 *	1) overlap == lock
    662 		 *	2) overlap contains lock
    663 		 *	3) lock contains overlap
    664 		 *	4) overlap starts before lock
    665 		 *	5) overlap ends after lock
    666 		 */
    667 		if ((lf->lf_end != -1 && start > lf->lf_end) ||
    668 		    (end != -1 && lf->lf_start > end)) {
    669 			/* Case 0 */
    670 #ifdef LOCKF_DEBUG
    671 			if (lockf_debug & 2)
    672 				printf("no overlap\n");
    673 #endif /* LOCKF_DEBUG */
    674 			if ((type & SELF) && end != -1 && lf->lf_start > end)
    675 				return 0;
    676 			*prev = &lf->lf_next;
    677 			*overlap = lf = lf->lf_next;
    678 			continue;
    679 		}
    680 		if ((lf->lf_start == start) && (lf->lf_end == end)) {
    681 			/* Case 1 */
    682 #ifdef LOCKF_DEBUG
    683 			if (lockf_debug & 2)
    684 				printf("overlap == lock\n");
    685 #endif /* LOCKF_DEBUG */
    686 			return 1;
    687 		}
    688 		if ((lf->lf_start <= start) &&
    689 		    (end != -1) &&
    690 		    ((lf->lf_end >= end) || (lf->lf_end == -1))) {
    691 			/* Case 2 */
    692 #ifdef LOCKF_DEBUG
    693 			if (lockf_debug & 2)
    694 				printf("overlap contains lock\n");
    695 #endif /* LOCKF_DEBUG */
    696 			return 2;
    697 		}
    698 		if (start <= lf->lf_start &&
    699 		           (end == -1 ||
    700 			   (lf->lf_end != -1 && end >= lf->lf_end))) {
    701 			/* Case 3 */
    702 #ifdef LOCKF_DEBUG
    703 			if (lockf_debug & 2)
    704 				printf("lock contains overlap\n");
    705 #endif /* LOCKF_DEBUG */
    706 			return 3;
    707 		}
    708 		if ((lf->lf_start < start) &&
    709 			((lf->lf_end >= start) || (lf->lf_end == -1))) {
    710 			/* Case 4 */
    711 #ifdef LOCKF_DEBUG
    712 			if (lockf_debug & 2)
    713 				printf("overlap starts before lock\n");
    714 #endif /* LOCKF_DEBUG */
    715 			return 4;
    716 		}
    717 		if ((lf->lf_start > start) &&
    718 			(end != -1) &&
    719 			((lf->lf_end > end) || (lf->lf_end == -1))) {
    720 			/* Case 5 */
    721 #ifdef LOCKF_DEBUG
    722 			if (lockf_debug & 2)
    723 				printf("overlap ends after lock\n");
    724 #endif /* LOCKF_DEBUG */
    725 			return 5;
    726 		}
    727 		panic("lf_findoverlap: default");
    728 	}
    729 	return 0;
    730 }
    731 
    732 /*
    733  * Split a lock and a contained region into
    734  * two or three locks as necessary.
    735  */
    736 static void
    737 lf_split(struct lockf *lock1, struct lockf *lock2, struct lockf **sparelock)
    738 {
    739 	struct lockf *splitlock;
    740 
    741 #ifdef LOCKF_DEBUG
    742 	if (lockf_debug & 2) {
    743 		lf_print("lf_split", lock1);
    744 		lf_print("splitting from", lock2);
    745 	}
    746 #endif /* LOCKF_DEBUG */
    747 	/*
    748 	 * Check to see if spliting into only two pieces.
    749 	 */
    750 	if (lock1->lf_start == lock2->lf_start) {
    751 		lock1->lf_start = lock2->lf_end + 1;
    752 		lock2->lf_next = lock1;
    753 		return;
    754 	}
    755 	if (lock1->lf_end == lock2->lf_end) {
    756 		lock1->lf_end = lock2->lf_start - 1;
    757 		lock2->lf_next = lock1->lf_next;
    758 		lock1->lf_next = lock2;
    759 		return;
    760 	}
    761 	/*
    762 	 * Make a new lock consisting of the last part of
    763 	 * the encompassing lock
    764 	 */
    765 	splitlock = *sparelock;
    766 	*sparelock = NULL;
    767 	memcpy(splitlock, lock1, sizeof(*splitlock));
    768 	splitlock->lf_start = lock2->lf_end + 1;
    769 	TAILQ_INIT(&splitlock->lf_blkhd);
    770 	lock1->lf_end = lock2->lf_start - 1;
    771 	/*
    772 	 * OK, now link it in
    773 	 */
    774 	splitlock->lf_next = lock1->lf_next;
    775 	lock2->lf_next = splitlock;
    776 	lock1->lf_next = lock2;
    777 }
    778 
    779 /*
    780  * Wakeup a blocklist
    781  */
    782 static void
    783 lf_wakelock(struct lockf *listhead)
    784 {
    785 	struct lockf *wakelock;
    786 
    787 	while ((wakelock = TAILQ_FIRST(&listhead->lf_blkhd))) {
    788 		KASSERT(wakelock->lf_next == listhead);
    789 		TAILQ_REMOVE(&listhead->lf_blkhd, wakelock, lf_block);
    790 		wakelock->lf_next = NOLOCKF;
    791 #ifdef LOCKF_DEBUG
    792 		if (lockf_debug & 2)
    793 			lf_print("lf_wakelock: awakening", wakelock);
    794 #endif
    795 		wakeup(wakelock);
    796 	}
    797 }
    798 
    799 #ifdef LOCKF_DEBUG
    800 /*
    801  * Print out a lock.
    802  */
    803 static void
    804 lf_print(char *tag, struct lockf *lock)
    805 {
    806 
    807 	printf("%s: lock %p for ", tag, lock);
    808 	if (lock->lf_flags & F_POSIX)
    809 		printf("proc %d", ((struct proc *)lock->lf_id)->p_pid);
    810 	else
    811 		printf("file 0x%p", (struct file *)lock->lf_id);
    812 	printf(" %s, start %qx, end %qx",
    813 		lock->lf_type == F_RDLCK ? "shared" :
    814 		lock->lf_type == F_WRLCK ? "exclusive" :
    815 		lock->lf_type == F_UNLCK ? "unlock" :
    816 		"unknown", lock->lf_start, lock->lf_end);
    817 	if (TAILQ_FIRST(&lock->lf_blkhd))
    818 		printf(" block %p\n", TAILQ_FIRST(&lock->lf_blkhd));
    819 	else
    820 		printf("\n");
    821 }
    822 
    823 static void
    824 lf_printlist(char *tag, struct lockf *lock)
    825 {
    826 	struct lockf *lf, *blk;
    827 
    828 	printf("%s: Lock list:\n", tag);
    829 	for (lf = *lock->lf_head; lf; lf = lf->lf_next) {
    830 		printf("\tlock %p for ", lf);
    831 		if (lf->lf_flags & F_POSIX)
    832 			printf("proc %d", ((struct proc *)lf->lf_id)->p_pid);
    833 		else
    834 			printf("file 0x%p", (struct file *)lf->lf_id);
    835 		printf(", %s, start %qx, end %qx",
    836 			lf->lf_type == F_RDLCK ? "shared" :
    837 			lf->lf_type == F_WRLCK ? "exclusive" :
    838 			lf->lf_type == F_UNLCK ? "unlock" :
    839 			"unknown", lf->lf_start, lf->lf_end);
    840 		TAILQ_FOREACH(blk, &lf->lf_blkhd, lf_block) {
    841 			if (blk->lf_flags & F_POSIX)
    842 				printf("proc %d",
    843 				    ((struct proc *)blk->lf_id)->p_pid);
    844 			else
    845 				printf("file 0x%p", (struct file *)blk->lf_id);
    846 			printf(", %s, start %qx, end %qx",
    847 				blk->lf_type == F_RDLCK ? "shared" :
    848 				blk->lf_type == F_WRLCK ? "exclusive" :
    849 				blk->lf_type == F_UNLCK ? "unlock" :
    850 				"unknown", blk->lf_start, blk->lf_end);
    851 			if (TAILQ_FIRST(&blk->lf_blkhd))
    852 				 panic("lf_printlist: bad list");
    853 		}
    854 		printf("\n");
    855 	}
    856 }
    857 #endif /* LOCKF_DEBUG */
    858