| 1 | /* |
| 2 | * Copyright (c) 2015 Apple Computer, Inc. All rights reserved. |
| 3 | * |
| 4 | * @APPLE_OSREFERENCE_LICENSE_HEADER_START@ |
| 5 | * |
| 6 | * This file contains Original Code and/or Modifications of Original Code |
| 7 | * as defined in and that are subject to the Apple Public Source License |
| 8 | * Version 2.0 (the 'License'). You may not use this file except in |
| 9 | * compliance with the License. The rights granted to you under the License |
| 10 | * may not be used to create, or enable the creation or redistribution of, |
| 11 | * unlawful or unlicensed copies of an Apple operating system, or to |
| 12 | * circumvent, violate, or enable the circumvention or violation of, any |
| 13 | * terms of an Apple operating system software license agreement. |
| 14 | * |
| 15 | * Please obtain a copy of the License at |
| 16 | * http://www.opensource.apple.com/apsl/ and read it before using this file. |
| 17 | * |
| 18 | * The Original Code and all software distributed under the License are |
| 19 | * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER |
| 20 | * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES, |
| 21 | * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY, |
| 22 | * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT. |
| 23 | * Please see the License for the specific language governing rights and |
| 24 | * limitations under the License. |
| 25 | * |
| 26 | * @APPLE_OSREFERENCE_LICENSE_HEADER_END@ |
| 27 | */ |
| 28 | /* |
| 29 | * Copyright (c) 1982, 1986, 1989, 1993 |
| 30 | * The Regents of the University of California. All rights reserved. |
| 31 | * |
| 32 | * This code is derived from software contributed to Berkeley by |
| 33 | * Scooter Morris at Genentech Inc. |
| 34 | * |
| 35 | * Redistribution and use in source and binary forms, with or without |
| 36 | * modification, are permitted provided that the following conditions |
| 37 | * are met: |
| 38 | * 1. Redistributions of source code must retain the above copyright |
| 39 | * notice, this list of conditions and the following disclaimer. |
| 40 | * 2. Redistributions in binary form must reproduce the above copyright |
| 41 | * notice, this list of conditions and the following disclaimer in the |
| 42 | * documentation and/or other materials provided with the distribution. |
| 43 | * 4. Neither the name of the University nor the names of its contributors |
| 44 | * may be used to endorse or promote products derived from this software |
| 45 | * without specific prior written permission. |
| 46 | * |
| 47 | * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND |
| 48 | * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
| 49 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE |
| 50 | * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE |
| 51 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL |
| 52 | * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS |
| 53 | * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) |
| 54 | * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT |
| 55 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY |
| 56 | * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF |
| 57 | * SUCH DAMAGE. |
| 58 | * |
| 59 | * @(#)ufs_lockf.c 8.3 (Berkeley) 1/6/94 |
| 60 | */ |
| 61 | |
| 62 | #include <sys/cdefs.h> |
| 63 | #include <sys/param.h> |
| 64 | #include <sys/systm.h> |
| 65 | #include <sys/kernel.h> |
| 66 | #include <sys/lock.h> |
| 67 | #include <sys/mount.h> |
| 68 | #include <sys/proc.h> |
| 69 | #include <sys/signalvar.h> |
| 70 | #include <sys/unistd.h> |
| 71 | #include <sys/user.h> |
| 72 | #include <sys/vnode.h> |
| 73 | #include <sys/vnode_internal.h> |
| 74 | #include <sys/vnode_if.h> |
| 75 | #include <sys/malloc.h> |
| 76 | #include <sys/fcntl.h> |
| 77 | #include <sys/lockf.h> |
| 78 | #include <sys/sdt.h> |
| 79 | #include <kern/policy_internal.h> |
| 80 | |
| 81 | #include <sys/file_internal.h> |
| 82 | |
| 83 | /* |
| 84 | * This variable controls the maximum number of processes that will |
| 85 | * be checked in doing deadlock detection. |
| 86 | */ |
| 87 | static int maxlockdepth = MAXDEPTH; |
| 88 | |
| 89 | #if (DEVELOPMENT || DEBUG) |
| 90 | #define LOCKF_DEBUGGING 1 |
| 91 | #endif |
| 92 | |
| 93 | #ifdef LOCKF_DEBUGGING |
| 94 | #include <sys/sysctl.h> |
| 95 | void lf_print(const char *tag, struct lockf *lock); |
| 96 | void lf_printlist(const char *tag, struct lockf *lock); |
| 97 | |
| 98 | #define LF_DBG_LOCKOP (1 << 0) /* setlk, getlk, clearlk */ |
| 99 | #define LF_DBG_LIST (1 << 1) /* split, coalesce */ |
| 100 | #define LF_DBG_IMPINH (1 << 2) /* importance inheritance */ |
| 101 | #define LF_DBG_TRACE (1 << 3) /* errors, exit */ |
| 102 | |
| 103 | static int lockf_debug = 0; /* was 2, could be 3 ;-) */ |
| 104 | SYSCTL_INT(_debug, OID_AUTO, lockf_debug, CTLFLAG_RW | CTLFLAG_LOCKED, &lockf_debug, 0, ""); |
| 105 | |
| 106 | /* |
| 107 | * If there is no mask bit selector, or there is one, and the selector is |
| 108 | * set, then output the debugging diagnostic. |
| 109 | */ |
| 110 | #define LOCKF_DEBUG(mask, ...) \ |
| 111 | do { \ |
| 112 | if( !(mask) || ((mask) & lockf_debug)) { \ |
| 113 | printf(__VA_ARGS__); \ |
| 114 | } \ |
| 115 | } while(0) |
| 116 | #else /* !LOCKF_DEBUGGING */ |
| 117 | #define LOCKF_DEBUG(mask, ...) /* mask */ |
| 118 | #endif /* !LOCKF_DEBUGGING */ |
| 119 | |
| 120 | MALLOC_DEFINE(M_LOCKF, "lockf", "Byte-range locking structures"); |
| 121 | |
| 122 | #define NOLOCKF (struct lockf *)0 |
| 123 | #define SELF 0x1 |
| 124 | #define OTHERS 0x2 |
| 125 | #define OFF_MAX 0x7fffffffffffffffULL /* max off_t */ |
| 126 | |
| 127 | /* |
| 128 | * Overlapping lock states |
| 129 | */ |
| 130 | typedef enum { |
| 131 | OVERLAP_NONE = 0, |
| 132 | OVERLAP_EQUALS_LOCK, |
| 133 | OVERLAP_CONTAINS_LOCK, |
| 134 | OVERLAP_CONTAINED_BY_LOCK, |
| 135 | OVERLAP_STARTS_BEFORE_LOCK, |
| 136 | OVERLAP_ENDS_AFTER_LOCK |
| 137 | } overlap_t; |
| 138 | |
| 139 | static int lf_clearlock(struct lockf *); |
| 140 | static overlap_t lf_findoverlap(struct lockf *, |
| 141 | struct lockf *, int, struct lockf ***, struct lockf **); |
| 142 | static struct lockf *lf_getblock(struct lockf *, pid_t); |
| 143 | static int lf_getlock(struct lockf *, struct flock *, pid_t); |
| 144 | static int lf_setlock(struct lockf *, struct timespec *); |
| 145 | static int lf_split(struct lockf *, struct lockf *); |
| 146 | static void lf_wakelock(struct lockf *, boolean_t); |
| 147 | #if IMPORTANCE_INHERITANCE |
| 148 | static void lf_hold_assertion(task_t, struct lockf *); |
| 149 | static void lf_jump_to_queue_head(struct lockf *, struct lockf *); |
| 150 | static void lf_drop_assertion(struct lockf *); |
| 151 | static void lf_boost_blocking_proc(struct lockf *, struct lockf *); |
| 152 | #endif /* IMPORTANCE_INHERITANCE */ |
| 153 | |
| 154 | /* |
| 155 | * lf_advlock |
| 156 | * |
| 157 | * Description: Advisory record locking support |
| 158 | * |
| 159 | * Parameters: ap Argument pointer to a vnop_advlock_args |
| 160 | * argument descriptor structure for the |
| 161 | * lock operation to be attempted. |
| 162 | * |
| 163 | * Returns: 0 Success |
| 164 | * EOVERFLOW |
| 165 | * EINVAL |
| 166 | * ENOLCK Number of locked regions exceeds limit |
| 167 | * lf_setlock:EAGAIN |
| 168 | * lf_setlock:EDEADLK |
| 169 | * lf_setlock:EINTR |
| 170 | * lf_setlock:ENOLCK |
| 171 | * lf_setlock:ETIMEDOUT |
| 172 | * lf_clearlock:ENOLCK |
| 173 | * vnode_size:??? |
| 174 | * |
| 175 | * Notes: We return ENOLCK when we run out of memory to support locks; as |
| 176 | * such, there is no specific expectation limit other than the |
| 177 | * amount of available resources. |
| 178 | */ |
| 179 | int |
| 180 | lf_advlock(struct vnop_advlock_args *ap) |
| 181 | { |
| 182 | struct vnode *vp = ap->a_vp; |
| 183 | struct flock *fl = ap->a_fl; |
| 184 | vfs_context_t context = ap->a_context; |
| 185 | struct lockf *lock; |
| 186 | off_t start, end, oadd; |
| 187 | u_quad_t size; |
| 188 | int error; |
| 189 | struct lockf **head = &vp->v_lockf; |
| 190 | |
| 191 | /* XXX HFS may need a !vnode_isreg(vp) EISDIR error here */ |
| 192 | |
| 193 | /* |
| 194 | * Avoid the common case of unlocking when inode has no locks. |
| 195 | */ |
| 196 | if (*head == (struct lockf *)0) { |
| 197 | if (ap->a_op != F_SETLK) { |
| 198 | fl->l_type = F_UNLCK; |
| 199 | LOCKF_DEBUG(LF_DBG_TRACE, |
| 200 | "lf_advlock: '%s' unlock without lock\n", |
| 201 | vfs_context_proc(context)->p_comm); |
| 202 | return (0); |
| 203 | } |
| 204 | } |
| 205 | |
| 206 | /* |
| 207 | * Convert the flock structure into a start and end. |
| 208 | */ |
| 209 | switch (fl->l_whence) { |
| 210 | |
| 211 | case SEEK_SET: |
| 212 | case SEEK_CUR: |
| 213 | /* |
| 214 | * Caller is responsible for adding any necessary offset |
| 215 | * when SEEK_CUR is used. |
| 216 | */ |
| 217 | start = fl->l_start; |
| 218 | break; |
| 219 | |
| 220 | case SEEK_END: |
| 221 | |
| 222 | /* |
| 223 | * It's OK to cast the u_quad_t to and off_t here, since they |
| 224 | * are the same storage size, and the value of the returned |
| 225 | * contents will never overflow into the sign bit. We need to |
| 226 | * do this because we will use size to force range checks. |
| 227 | */ |
| 228 | if ((error = vnode_size(vp, (off_t *)&size, context))) { |
| 229 | LOCKF_DEBUG(LF_DBG_TRACE, |
| 230 | "lf_advlock: vnode_getattr failed: %d\n", error); |
| 231 | return (error); |
| 232 | } |
| 233 | |
| 234 | if (size > OFF_MAX || |
| 235 | (fl->l_start > 0 && |
| 236 | size > (u_quad_t)(OFF_MAX - fl->l_start))) |
| 237 | return (EOVERFLOW); |
| 238 | start = size + fl->l_start; |
| 239 | break; |
| 240 | |
| 241 | default: |
| 242 | LOCKF_DEBUG(LF_DBG_TRACE, "lf_advlock: unknown whence %d\n", |
| 243 | fl->l_whence); |
| 244 | return (EINVAL); |
| 245 | } |
| 246 | if (start < 0) { |
| 247 | LOCKF_DEBUG(LF_DBG_TRACE, "lf_advlock: start < 0 (%qd)\n", |
| 248 | start); |
| 249 | return (EINVAL); |
| 250 | } |
| 251 | if (fl->l_len < 0) { |
| 252 | if (start == 0) { |
| 253 | LOCKF_DEBUG(LF_DBG_TRACE, |
| 254 | "lf_advlock: len < 0 & start == 0\n"); |
| 255 | return (EINVAL); |
| 256 | } |
| 257 | end = start - 1; |
| 258 | start += fl->l_len; |
| 259 | if (start < 0) { |
| 260 | LOCKF_DEBUG(LF_DBG_TRACE, |
| 261 | "lf_advlock: start < 0 (%qd)\n", start); |
| 262 | return (EINVAL); |
| 263 | } |
| 264 | } else if (fl->l_len == 0) |
| 265 | end = -1; |
| 266 | else { |
| 267 | oadd = fl->l_len - 1; |
| 268 | if (oadd > (off_t)(OFF_MAX - start)) { |
| 269 | LOCKF_DEBUG(LF_DBG_TRACE, "lf_advlock: overflow\n"); |
| 270 | return (EOVERFLOW); |
| 271 | } |
| 272 | end = start + oadd; |
| 273 | } |
| 274 | /* |
| 275 | * Create the lockf structure |
| 276 | */ |
| 277 | MALLOC(lock, struct lockf *, sizeof *lock, M_LOCKF, M_WAITOK); |
| 278 | if (lock == NULL) |
| 279 | return (ENOLCK); |
| 280 | lock->lf_start = start; |
| 281 | lock->lf_end = end; |
| 282 | lock->lf_id = ap->a_id; |
| 283 | lock->lf_vnode = vp; |
| 284 | lock->lf_type = fl->l_type; |
| 285 | lock->lf_head = head; |
| 286 | lock->lf_next = (struct lockf *)0; |
| 287 | TAILQ_INIT(&lock->lf_blkhd); |
| 288 | lock->lf_flags = ap->a_flags; |
| 289 | #if IMPORTANCE_INHERITANCE |
| 290 | lock->lf_boosted = LF_NOT_BOOSTED; |
| 291 | #endif |
| 292 | if (ap->a_flags & F_POSIX) |
| 293 | lock->lf_owner = (struct proc *)lock->lf_id; |
| 294 | else |
| 295 | lock->lf_owner = NULL; |
| 296 | |
| 297 | if (ap->a_flags & F_FLOCK) |
| 298 | lock->lf_flags |= F_WAKE1_SAFE; |
| 299 | |
| 300 | lck_mtx_lock(&vp->v_lock); /* protect the lockf list */ |
| 301 | /* |
| 302 | * Do the requested operation. |
| 303 | */ |
| 304 | switch(ap->a_op) { |
| 305 | case F_SETLK: |
| 306 | /* |
| 307 | * For F_OFD_* locks, lf_id is the fileglob. |
| 308 | * Record an "lf_owner" iff this is a confined fd |
| 309 | * i.e. it cannot escape this process and will be |
| 310 | * F_UNLCKed before the owner exits. (This is |
| 311 | * the implicit guarantee needed to ensure lf_owner |
| 312 | * remains a valid reference here.) |
| 313 | */ |
| 314 | if (ap->a_flags & F_OFD_LOCK) { |
| 315 | struct fileglob *fg = (void *)lock->lf_id; |
| 316 | if (fg->fg_lflags & FG_CONFINED) |
| 317 | lock->lf_owner = current_proc(); |
| 318 | } |
| 319 | error = lf_setlock(lock, ap->a_timeout); |
| 320 | break; |
| 321 | |
| 322 | case F_UNLCK: |
| 323 | error = lf_clearlock(lock); |
| 324 | FREE(lock, M_LOCKF); |
| 325 | break; |
| 326 | |
| 327 | case F_GETLK: |
| 328 | error = lf_getlock(lock, fl, -1); |
| 329 | FREE(lock, M_LOCKF); |
| 330 | break; |
| 331 | |
| 332 | |
| 333 | default: |
| 334 | FREE(lock, M_LOCKF); |
| 335 | error = EINVAL; |
| 336 | break; |
| 337 | } |
| 338 | lck_mtx_unlock(&vp->v_lock); /* done manipulating the list */ |
| 339 | |
| 340 | LOCKF_DEBUG(LF_DBG_TRACE, "lf_advlock: normal exit: %d\n", error); |
| 341 | return (error); |
| 342 | } |
| 343 | |
| 344 | /* |
| 345 | * Empty the queue of msleeping requests for a lock on the given vnode. |
| 346 | * Called with the vnode already locked. Used for forced unmount, where |
| 347 | * a flock(2) invoker sleeping on a blocked lock holds an iocount reference |
| 348 | * that prevents the vnode from ever being drained. Force unmounting wins. |
| 349 | */ |
| 350 | void |
| 351 | lf_abort_advlocks(vnode_t vp) |
| 352 | { |
| 353 | struct lockf *lock; |
| 354 | |
| 355 | if ((lock = vp->v_lockf) == NULL) |
| 356 | return; |
| 357 | |
| 358 | lck_mtx_assert(&vp->v_lock, LCK_MTX_ASSERT_OWNED); |
| 359 | |
| 360 | if (!TAILQ_EMPTY(&lock->lf_blkhd)) { |
| 361 | struct lockf *tlock; |
| 362 | |
| 363 | TAILQ_FOREACH(tlock, &lock->lf_blkhd, lf_block) { |
| 364 | /* |
| 365 | * Setting this flag should cause all |
| 366 | * currently blocked F_SETLK request to |
| 367 | * return to userland with an errno. |
| 368 | */ |
| 369 | tlock->lf_flags |= F_ABORT; |
| 370 | } |
| 371 | lf_wakelock(lock, TRUE); |
| 372 | } |
| 373 | } |
| 374 | |
| 375 | /* |
| 376 | * Take any lock attempts which are currently blocked by a given lock ("from") |
| 377 | * and mark them as blocked by a different lock ("to"). Used in the case |
| 378 | * where a byte range currently occupied by "from" is to be occupied by "to." |
| 379 | */ |
| 380 | static void |
| 381 | lf_move_blocked(struct lockf *to, struct lockf *from) |
| 382 | { |
| 383 | struct lockf *tlock; |
| 384 | |
| 385 | TAILQ_FOREACH(tlock, &from->lf_blkhd, lf_block) { |
| 386 | tlock->lf_next = to; |
| 387 | } |
| 388 | |
| 389 | TAILQ_CONCAT(&to->lf_blkhd, &from->lf_blkhd, lf_block); |
| 390 | } |
| 391 | |
| 392 | /* |
| 393 | * lf_coalesce_adjacent |
| 394 | * |
| 395 | * Description: Helper function: when setting a lock, coalesce adjacent |
| 396 | * locks. Needed because adjacent locks are not overlapping, |
| 397 | * but POSIX requires that they be coalesced. |
| 398 | * |
| 399 | * Parameters: lock The new lock which may be adjacent |
| 400 | * to already locked regions, and which |
| 401 | * should therefore be coalesced with them |
| 402 | * |
| 403 | * Returns: <void> |
| 404 | */ |
| 405 | static void |
| 406 | lf_coalesce_adjacent(struct lockf *lock) |
| 407 | { |
| 408 | struct lockf **lf = lock->lf_head; |
| 409 | |
| 410 | while (*lf != NOLOCKF) { |
| 411 | /* reject locks that obviously could not be coalesced */ |
| 412 | if ((*lf == lock) || |
| 413 | ((*lf)->lf_id != lock->lf_id) || |
| 414 | ((*lf)->lf_type != lock->lf_type)) { |
| 415 | lf = &(*lf)->lf_next; |
| 416 | continue; |
| 417 | } |
| 418 | |
| 419 | /* |
| 420 | * NOTE: Assumes that if two locks are adjacent on the number line |
| 421 | * and belong to the same owner, then they are adjacent on the list. |
| 422 | */ |
| 423 | if ((*lf)->lf_end != -1 && |
| 424 | ((*lf)->lf_end + 1) == lock->lf_start) { |
| 425 | struct lockf *adjacent = *lf; |
| 426 | |
| 427 | LOCKF_DEBUG(LF_DBG_LIST, "lf_coalesce_adjacent: coalesce adjacent previous\n"); |
| 428 | lock->lf_start = (*lf)->lf_start; |
| 429 | *lf = lock; |
| 430 | lf = &(*lf)->lf_next; |
| 431 | |
| 432 | lf_move_blocked(lock, adjacent); |
| 433 | |
| 434 | FREE(adjacent, M_LOCKF); |
| 435 | continue; |
| 436 | } |
| 437 | /* If the lock starts adjacent to us, we can coalesce it */ |
| 438 | if (lock->lf_end != -1 && |
| 439 | (lock->lf_end + 1) == (*lf)->lf_start) { |
| 440 | struct lockf *adjacent = *lf; |
| 441 | |
| 442 | LOCKF_DEBUG(LF_DBG_LIST, "lf_coalesce_adjacent: coalesce adjacent following\n"); |
| 443 | lock->lf_end = (*lf)->lf_end; |
| 444 | lock->lf_next = (*lf)->lf_next; |
| 445 | lf = &lock->lf_next; |
| 446 | |
| 447 | lf_move_blocked(lock, adjacent); |
| 448 | |
| 449 | FREE(adjacent, M_LOCKF); |
| 450 | continue; |
| 451 | } |
| 452 | |
| 453 | /* no matching conditions; go on to next lock */ |
| 454 | lf = &(*lf)->lf_next; |
| 455 | } |
| 456 | } |
| 457 | |
| 458 | /* |
| 459 | * lf_setlock |
| 460 | * |
| 461 | * Description: Set a byte-range lock. |
| 462 | * |
| 463 | * Parameters: lock The lock structure describing the lock |
| 464 | * to be set; allocated by the caller, it |
| 465 | * will be linked into the lock list if |
| 466 | * the set is successful, and freed if the |
| 467 | * set is unsuccessful. |
| 468 | * |
| 469 | * timeout Timeout specified in the case of |
| 470 | * SETLKWTIMEOUT. |
| 471 | * |
| 472 | * Returns: 0 Success |
| 473 | * EAGAIN |
| 474 | * EDEADLK |
| 475 | * lf_split:ENOLCK |
| 476 | * lf_clearlock:ENOLCK |
| 477 | * msleep:EINTR |
| 478 | * msleep:ETIMEDOUT |
| 479 | * |
| 480 | * Notes: We add the lock to the provisional lock list. We do not |
| 481 | * coalesce at this time; this has implications for other lock |
| 482 | * requestors in the blocker search mechanism. |
| 483 | */ |
| 484 | static int |
| 485 | lf_setlock(struct lockf *lock, struct timespec *timeout) |
| 486 | { |
| 487 | struct lockf *block; |
| 488 | struct lockf **head = lock->lf_head; |
| 489 | struct lockf **prev, *overlap, *ltmp; |
| 490 | static char lockstr[] = "lockf"; |
| 491 | int priority, needtolink, error; |
| 492 | struct vnode *vp = lock->lf_vnode; |
| 493 | overlap_t ovcase; |
| 494 | |
| 495 | #ifdef LOCKF_DEBUGGING |
| 496 | if (lockf_debug & LF_DBG_LOCKOP) { |
| 497 | lf_print("lf_setlock", lock); |
| 498 | lf_printlist("lf_setlock(in)", lock); |
| 499 | } |
| 500 | #endif /* LOCKF_DEBUGGING */ |
| 501 | |
| 502 | /* |
| 503 | * Set the priority |
| 504 | */ |
| 505 | priority = PLOCK; |
| 506 | if (lock->lf_type == F_WRLCK) |
| 507 | priority += 4; |
| 508 | priority |= PCATCH; |
| 509 | /* |
| 510 | * Scan lock list for this file looking for locks that would block us. |
| 511 | */ |
| 512 | while ((block = lf_getblock(lock, -1))) { |
| 513 | /* |
| 514 | * Free the structure and return if nonblocking. |
| 515 | */ |
| 516 | if ((lock->lf_flags & F_WAIT) == 0) { |
| 517 | DTRACE_FSINFO(advlock__nowait, vnode_t, vp); |
| 518 | FREE(lock, M_LOCKF); |
| 519 | return (EAGAIN); |
| 520 | } |
| 521 | |
| 522 | /* |
| 523 | * We are blocked. Since flock style locks cover |
| 524 | * the whole file, there is no chance for deadlock. |
| 525 | * |
| 526 | * OFD byte-range locks currently do NOT support |
| 527 | * deadlock detection. |
| 528 | * |
| 529 | * For POSIX byte-range locks we must check for deadlock. |
| 530 | * |
| 531 | * Deadlock detection is done by looking through the |
| 532 | * wait channels to see if there are any cycles that |
| 533 | * involve us. MAXDEPTH is set just to make sure we |
| 534 | * do not go off into neverland. |
| 535 | */ |
| 536 | if ((lock->lf_flags & F_POSIX) && |
| 537 | (block->lf_flags & F_POSIX)) { |
| 538 | struct proc *wproc, *bproc; |
| 539 | struct uthread *ut; |
| 540 | struct lockf *waitblock; |
| 541 | int i = 0; |
| 542 | |
| 543 | /* The block is waiting on something */ |
| 544 | wproc = block->lf_owner; |
| 545 | proc_lock(wproc); |
| 546 | TAILQ_FOREACH(ut, &wproc->p_uthlist, uu_list) { |
| 547 | /* |
| 548 | * While the thread is asleep (uu_wchan != 0) |
| 549 | * in this code (uu_wmesg == lockstr) |
| 550 | * and we have not exceeded the maximum cycle |
| 551 | * depth (i < maxlockdepth), then check for a |
| 552 | * cycle to see if the lock is blocked behind |
| 553 | * someone blocked behind us. |
| 554 | */ |
| 555 | while (((waitblock = (struct lockf *)ut->uu_wchan) != NULL) && |
| 556 | ut->uu_wmesg == lockstr && |
| 557 | (i++ < maxlockdepth)) { |
| 558 | waitblock = (struct lockf *)ut->uu_wchan; |
| 559 | /* |
| 560 | * Get the lock blocking the lock |
| 561 | * which would block us, and make |
| 562 | * certain it hasn't come unblocked |
| 563 | * (been granted, e.g. between the time |
| 564 | * we called lf_getblock, and the time |
| 565 | * we successfully acquired the |
| 566 | * proc_lock). |
| 567 | */ |
| 568 | waitblock = waitblock->lf_next; |
| 569 | if (waitblock == NULL) |
| 570 | break; |
| 571 | |
| 572 | /* |
| 573 | * Make sure it's an advisory range |
| 574 | * lock and not any other kind of lock; |
| 575 | * if we mix lock types, it's our own |
| 576 | * fault. |
| 577 | */ |
| 578 | if ((waitblock->lf_flags & F_POSIX) == 0) |
| 579 | break; |
| 580 | |
| 581 | /* |
| 582 | * If the owner of the lock that's |
| 583 | * blocking a lock that's blocking us |
| 584 | * getting the requested lock, then we |
| 585 | * would deadlock, so error out. |
| 586 | */ |
| 587 | bproc = waitblock->lf_owner; |
| 588 | if (bproc == lock->lf_owner) { |
| 589 | proc_unlock(wproc); |
| 590 | FREE(lock, M_LOCKF); |
| 591 | return (EDEADLK); |
| 592 | } |
| 593 | } |
| 594 | } |
| 595 | proc_unlock(wproc); |
| 596 | } |
| 597 | |
| 598 | /* |
| 599 | * For flock type locks, we must first remove |
| 600 | * any shared locks that we hold before we sleep |
| 601 | * waiting for an exclusive lock. |
| 602 | */ |
| 603 | if ((lock->lf_flags & F_FLOCK) && |
| 604 | lock->lf_type == F_WRLCK) { |
| 605 | lock->lf_type = F_UNLCK; |
| 606 | if ((error = lf_clearlock(lock)) != 0) { |
| 607 | FREE(lock, M_LOCKF); |
| 608 | return (error); |
| 609 | } |
| 610 | lock->lf_type = F_WRLCK; |
| 611 | } |
| 612 | /* |
| 613 | * Add our lock to the blocked list and sleep until we're free. |
| 614 | * Remember who blocked us (for deadlock detection). |
| 615 | */ |
| 616 | lock->lf_next = block; |
| 617 | TAILQ_INSERT_TAIL(&block->lf_blkhd, lock, lf_block); |
| 618 | |
| 619 | if ( !(lock->lf_flags & F_FLOCK)) |
| 620 | block->lf_flags &= ~F_WAKE1_SAFE; |
| 621 | |
| 622 | #if IMPORTANCE_INHERITANCE |
| 623 | /* |
| 624 | * Importance donation is done only for cases where the |
| 625 | * owning task can be unambiguously determined. |
| 626 | * |
| 627 | * POSIX type locks are not inherited by child processes; |
| 628 | * we maintain a 1:1 mapping between a lock and its owning |
| 629 | * process. |
| 630 | * |
| 631 | * Flock type locks are inherited across fork() and there is |
| 632 | * no 1:1 mapping in the general case. However, the fileglobs |
| 633 | * used by OFD locks *may* be confined to the process that |
| 634 | * created them, and thus have an "owner", in which case |
| 635 | * we also attempt importance donation. |
| 636 | */ |
| 637 | if ((lock->lf_flags & block->lf_flags & F_POSIX) != 0) |
| 638 | lf_boost_blocking_proc(lock, block); |
| 639 | else if ((lock->lf_flags & block->lf_flags & F_OFD_LOCK) && |
| 640 | lock->lf_owner != block->lf_owner && |
| 641 | NULL != lock->lf_owner && NULL != block->lf_owner) |
| 642 | lf_boost_blocking_proc(lock, block); |
| 643 | #endif /* IMPORTANCE_INHERITANCE */ |
| 644 | |
| 645 | #ifdef LOCKF_DEBUGGING |
| 646 | if (lockf_debug & LF_DBG_LOCKOP) { |
| 647 | lf_print("lf_setlock: blocking on", block); |
| 648 | lf_printlist("lf_setlock(block)", block); |
| 649 | } |
| 650 | #endif /* LOCKF_DEBUGGING */ |
| 651 | DTRACE_FSINFO(advlock__wait, vnode_t, vp); |
| 652 | |
| 653 | error = msleep(lock, &vp->v_lock, priority, lockstr, timeout); |
| 654 | |
| 655 | if (error == 0 && (lock->lf_flags & F_ABORT) != 0) |
| 656 | error = EBADF; |
| 657 | |
| 658 | if (lock->lf_next) { |
| 659 | /* |
| 660 | * lf_wakelock() always sets wakelock->lf_next to |
| 661 | * NULL before a wakeup; so we've been woken early |
| 662 | * - perhaps by a debugger, signal or other event. |
| 663 | * |
| 664 | * Remove 'lock' from the block list (avoids double-add |
| 665 | * in the spurious case, which would create a cycle) |
| 666 | */ |
| 667 | TAILQ_REMOVE(&lock->lf_next->lf_blkhd, lock, lf_block); |
| 668 | lock->lf_next = NULL; |
| 669 | |
| 670 | if (error == 0) { |
| 671 | /* |
| 672 | * If this was a spurious wakeup, retry |
| 673 | */ |
| 674 | printf("%s: spurious wakeup, retrying lock\n", |
| 675 | __func__); |
| 676 | continue; |
| 677 | } |
| 678 | } |
| 679 | |
| 680 | if (!TAILQ_EMPTY(&lock->lf_blkhd)) { |
| 681 | if ((block = lf_getblock(lock, -1)) != NULL) |
| 682 | lf_move_blocked(block, lock); |
| 683 | } |
| 684 | |
| 685 | if (error) { |
| 686 | if (!TAILQ_EMPTY(&lock->lf_blkhd)) |
| 687 | lf_wakelock(lock, TRUE); |
| 688 | FREE(lock, M_LOCKF); |
| 689 | /* Return ETIMEDOUT if timeout occoured. */ |
| 690 | if (error == EWOULDBLOCK) { |
| 691 | error = ETIMEDOUT; |
| 692 | } |
| 693 | return (error); |
| 694 | } |
| 695 | } |
| 696 | |
| 697 | /* |
| 698 | * No blocks!! Add the lock. Note that we will |
| 699 | * downgrade or upgrade any overlapping locks this |
| 700 | * process already owns. |
| 701 | * |
| 702 | * Skip over locks owned by other processes. |
| 703 | * Handle any locks that overlap and are owned by ourselves. |
| 704 | */ |
| 705 | prev = head; |
| 706 | block = *head; |
| 707 | needtolink = 1; |
| 708 | for (;;) { |
| 709 | ovcase = lf_findoverlap(block, lock, SELF, &prev, &overlap); |
| 710 | if (ovcase) |
| 711 | block = overlap->lf_next; |
| 712 | /* |
| 713 | * Six cases: |
| 714 | * 0) no overlap |
| 715 | * 1) overlap == lock |
| 716 | * 2) overlap contains lock |
| 717 | * 3) lock contains overlap |
| 718 | * 4) overlap starts before lock |
| 719 | * 5) overlap ends after lock |
| 720 | */ |
| 721 | switch (ovcase) { |
| 722 | case OVERLAP_NONE: |
| 723 | if (needtolink) { |
| 724 | *prev = lock; |
| 725 | lock->lf_next = overlap; |
| 726 | } |
| 727 | break; |
| 728 | |
| 729 | case OVERLAP_EQUALS_LOCK: |
| 730 | /* |
| 731 | * If downgrading lock, others may be |
| 732 | * able to acquire it. |
| 733 | */ |
| 734 | if (lock->lf_type == F_RDLCK && |
| 735 | overlap->lf_type == F_WRLCK) |
| 736 | lf_wakelock(overlap, TRUE); |
| 737 | overlap->lf_type = lock->lf_type; |
| 738 | FREE(lock, M_LOCKF); |
| 739 | lock = overlap; /* for lf_coalesce_adjacent() */ |
| 740 | break; |
| 741 | |
| 742 | case OVERLAP_CONTAINS_LOCK: |
| 743 | /* |
| 744 | * Check for common starting point and different types. |
| 745 | */ |
| 746 | if (overlap->lf_type == lock->lf_type) { |
| 747 | FREE(lock, M_LOCKF); |
| 748 | lock = overlap; /* for lf_coalesce_adjacent() */ |
| 749 | break; |
| 750 | } |
| 751 | if (overlap->lf_start == lock->lf_start) { |
| 752 | *prev = lock; |
| 753 | lock->lf_next = overlap; |
| 754 | overlap->lf_start = lock->lf_end + 1; |
| 755 | } else { |
| 756 | /* |
| 757 | * If we can't split the lock, we can't |
| 758 | * grant it. Claim a system limit for the |
| 759 | * resource shortage. |
| 760 | */ |
| 761 | if (lf_split(overlap, lock)) { |
| 762 | FREE(lock, M_LOCKF); |
| 763 | return (ENOLCK); |
| 764 | } |
| 765 | } |
| 766 | lf_wakelock(overlap, TRUE); |
| 767 | break; |
| 768 | |
| 769 | case OVERLAP_CONTAINED_BY_LOCK: |
| 770 | /* |
| 771 | * If downgrading lock, others may be able to |
| 772 | * acquire it, otherwise take the list. |
| 773 | */ |
| 774 | if (lock->lf_type == F_RDLCK && |
| 775 | overlap->lf_type == F_WRLCK) { |
| 776 | lf_wakelock(overlap, TRUE); |
| 777 | } else { |
| 778 | while (!TAILQ_EMPTY(&overlap->lf_blkhd)) { |
| 779 | ltmp = TAILQ_FIRST(&overlap->lf_blkhd); |
| 780 | TAILQ_REMOVE(&overlap->lf_blkhd, ltmp, |
| 781 | lf_block); |
| 782 | TAILQ_INSERT_TAIL(&lock->lf_blkhd, |
| 783 | ltmp, lf_block); |
| 784 | ltmp->lf_next = lock; |
| 785 | } |
| 786 | } |
| 787 | /* |
| 788 | * Add the new lock if necessary and delete the overlap. |
| 789 | */ |
| 790 | if (needtolink) { |
| 791 | *prev = lock; |
| 792 | lock->lf_next = overlap->lf_next; |
| 793 | prev = &lock->lf_next; |
| 794 | needtolink = 0; |
| 795 | } else |
| 796 | *prev = overlap->lf_next; |
| 797 | FREE(overlap, M_LOCKF); |
| 798 | continue; |
| 799 | |
| 800 | case OVERLAP_STARTS_BEFORE_LOCK: |
| 801 | /* |
| 802 | * Add lock after overlap on the list. |
| 803 | */ |
| 804 | lock->lf_next = overlap->lf_next; |
| 805 | overlap->lf_next = lock; |
| 806 | overlap->lf_end = lock->lf_start - 1; |
| 807 | prev = &lock->lf_next; |
| 808 | lf_wakelock(overlap, TRUE); |
| 809 | needtolink = 0; |
| 810 | continue; |
| 811 | |
| 812 | case OVERLAP_ENDS_AFTER_LOCK: |
| 813 | /* |
| 814 | * Add the new lock before overlap. |
| 815 | */ |
| 816 | if (needtolink) { |
| 817 | *prev = lock; |
| 818 | lock->lf_next = overlap; |
| 819 | } |
| 820 | overlap->lf_start = lock->lf_end + 1; |
| 821 | lf_wakelock(overlap, TRUE); |
| 822 | break; |
| 823 | } |
| 824 | break; |
| 825 | } |
| 826 | /* Coalesce adjacent locks with identical attributes */ |
| 827 | lf_coalesce_adjacent(lock); |
| 828 | #ifdef LOCKF_DEBUGGING |
| 829 | if (lockf_debug & LF_DBG_LOCKOP) { |
| 830 | lf_print("lf_setlock: got the lock", lock); |
| 831 | lf_printlist("lf_setlock(out)", lock); |
| 832 | } |
| 833 | #endif /* LOCKF_DEBUGGING */ |
| 834 | return (0); |
| 835 | } |
| 836 | |
| 837 | |
| 838 | /* |
| 839 | * lf_clearlock |
| 840 | * |
| 841 | * Description: Remove a byte-range lock on an vnode. Generally, find the |
| 842 | * lock (or an overlap to that lock) and remove it (or shrink |
| 843 | * it), then wakeup anyone we can. |
| 844 | * |
| 845 | * Parameters: unlock The lock to clear |
| 846 | * |
| 847 | * Returns: 0 Success |
| 848 | * lf_split:ENOLCK |
| 849 | * |
| 850 | * Notes: A caller may unlock all the locks owned by the caller by |
| 851 | * specifying the entire file range; locks owned by other |
| 852 | * callers are not effected by this operation. |
| 853 | */ |
| 854 | static int |
| 855 | lf_clearlock(struct lockf *unlock) |
| 856 | { |
| 857 | struct lockf **head = unlock->lf_head; |
| 858 | struct lockf *lf = *head; |
| 859 | struct lockf *overlap, **prev; |
| 860 | overlap_t ovcase; |
| 861 | |
| 862 | if (lf == NOLOCKF) |
| 863 | return (0); |
| 864 | #ifdef LOCKF_DEBUGGING |
| 865 | if (unlock->lf_type != F_UNLCK) |
| 866 | panic("lf_clearlock: bad type"); |
| 867 | if (lockf_debug & LF_DBG_LOCKOP) |
| 868 | lf_print("lf_clearlock", unlock); |
| 869 | #endif /* LOCKF_DEBUGGING */ |
| 870 | prev = head; |
| 871 | while ((ovcase = lf_findoverlap(lf, unlock, SELF, &prev, &overlap)) != OVERLAP_NONE) { |
| 872 | /* |
| 873 | * Wakeup the list of locks to be retried. |
| 874 | */ |
| 875 | lf_wakelock(overlap, FALSE); |
| 876 | #if IMPORTANCE_INHERITANCE |
| 877 | if (overlap->lf_boosted == LF_BOOSTED) { |
| 878 | lf_drop_assertion(overlap); |
| 879 | } |
| 880 | #endif /* IMPORTANCE_INHERITANCE */ |
| 881 | |
| 882 | switch (ovcase) { |
| 883 | case OVERLAP_NONE: /* satisfy compiler enum/switch */ |
| 884 | break; |
| 885 | |
| 886 | case OVERLAP_EQUALS_LOCK: |
| 887 | *prev = overlap->lf_next; |
| 888 | FREE(overlap, M_LOCKF); |
| 889 | break; |
| 890 | |
| 891 | case OVERLAP_CONTAINS_LOCK: /* split it */ |
| 892 | if (overlap->lf_start == unlock->lf_start) { |
| 893 | overlap->lf_start = unlock->lf_end + 1; |
| 894 | break; |
| 895 | } |
| 896 | /* |
| 897 | * If we can't split the lock, we can't grant it. |
| 898 | * Claim a system limit for the resource shortage. |
| 899 | */ |
| 900 | if (lf_split(overlap, unlock)) |
| 901 | return (ENOLCK); |
| 902 | overlap->lf_next = unlock->lf_next; |
| 903 | break; |
| 904 | |
| 905 | case OVERLAP_CONTAINED_BY_LOCK: |
| 906 | *prev = overlap->lf_next; |
| 907 | lf = overlap->lf_next; |
| 908 | FREE(overlap, M_LOCKF); |
| 909 | continue; |
| 910 | |
| 911 | case OVERLAP_STARTS_BEFORE_LOCK: |
| 912 | overlap->lf_end = unlock->lf_start - 1; |
| 913 | prev = &overlap->lf_next; |
| 914 | lf = overlap->lf_next; |
| 915 | continue; |
| 916 | |
| 917 | case OVERLAP_ENDS_AFTER_LOCK: |
| 918 | overlap->lf_start = unlock->lf_end + 1; |
| 919 | break; |
| 920 | } |
| 921 | break; |
| 922 | } |
| 923 | #ifdef LOCKF_DEBUGGING |
| 924 | if (lockf_debug & LF_DBG_LOCKOP) |
| 925 | lf_printlist("lf_clearlock", unlock); |
| 926 | #endif /* LOCKF_DEBUGGING */ |
| 927 | return (0); |
| 928 | } |
| 929 | |
| 930 | |
| 931 | /* |
| 932 | * lf_getlock |
| 933 | * |
| 934 | * Description: Check whether there is a blocking lock, and if so return |
| 935 | * its process identifier into the lock being requested. |
| 936 | * |
| 937 | * Parameters: lock Pointer to lock to test for blocks |
| 938 | * fl Pointer to flock structure to receive |
| 939 | * the blocking lock information, if a |
| 940 | * blocking lock is found. |
| 941 | * matchpid -1, or pid value to match in lookup. |
| 942 | * |
| 943 | * Returns: 0 Success |
| 944 | * |
| 945 | * Implicit Returns: |
| 946 | * *fl Contents modified to reflect the |
| 947 | * blocking lock, if one is found; not |
| 948 | * modified otherwise |
| 949 | * |
| 950 | * Notes: fl->l_pid will be (-1) for file locks and will only be set to |
| 951 | * the blocking process ID for advisory record locks. |
| 952 | */ |
| 953 | static int |
| 954 | lf_getlock(struct lockf *lock, struct flock *fl, pid_t matchpid) |
| 955 | { |
| 956 | struct lockf *block; |
| 957 | |
| 958 | #ifdef LOCKF_DEBUGGING |
| 959 | if (lockf_debug & LF_DBG_LOCKOP) |
| 960 | lf_print("lf_getlock", lock); |
| 961 | #endif /* LOCKF_DEBUGGING */ |
| 962 | |
| 963 | if ((block = lf_getblock(lock, matchpid))) { |
| 964 | fl->l_type = block->lf_type; |
| 965 | fl->l_whence = SEEK_SET; |
| 966 | fl->l_start = block->lf_start; |
| 967 | if (block->lf_end == -1) |
| 968 | fl->l_len = 0; |
| 969 | else |
| 970 | fl->l_len = block->lf_end - block->lf_start + 1; |
| 971 | if (NULL != block->lf_owner) { |
| 972 | /* |
| 973 | * lf_owner is only non-NULL when the lock |
| 974 | * "owner" can be unambiguously determined |
| 975 | */ |
| 976 | fl->l_pid = proc_pid(block->lf_owner); |
| 977 | } else |
| 978 | fl->l_pid = -1; |
| 979 | } else { |
| 980 | fl->l_type = F_UNLCK; |
| 981 | } |
| 982 | return (0); |
| 983 | } |
| 984 | |
| 985 | /* |
| 986 | * lf_getblock |
| 987 | * |
| 988 | * Description: Walk the list of locks for an inode and return the first |
| 989 | * blocking lock. A lock is considered blocking if we are not |
| 990 | * the lock owner; otherwise, we are permitted to upgrade or |
| 991 | * downgrade it, and it's not considered blocking. |
| 992 | * |
| 993 | * Parameters: lock The lock for which we are interested |
| 994 | * in obtaining the blocking lock, if any |
| 995 | * matchpid -1, or pid value to match in lookup. |
| 996 | * |
| 997 | * Returns: NOLOCKF No blocking lock exists |
| 998 | * !NOLOCKF The address of the blocking lock's |
| 999 | * struct lockf. |
| 1000 | */ |
| 1001 | static struct lockf * |
| 1002 | lf_getblock(struct lockf *lock, pid_t matchpid) |
| 1003 | { |
| 1004 | struct lockf **prev, *overlap, *lf = *(lock->lf_head); |
| 1005 | |
| 1006 | for (prev = lock->lf_head; |
| 1007 | lf_findoverlap(lf, lock, OTHERS, &prev, &overlap) != OVERLAP_NONE; |
| 1008 | lf = overlap->lf_next) { |
| 1009 | /* |
| 1010 | * Found an overlap. |
| 1011 | * |
| 1012 | * If we're matching pids, and it's a record lock, |
| 1013 | * or it's an OFD lock on a process-confined fd, |
| 1014 | * but the pid doesn't match, then keep on looking .. |
| 1015 | */ |
| 1016 | if (matchpid != -1 && |
| 1017 | (overlap->lf_flags & (F_POSIX|F_OFD_LOCK)) != 0 && |
| 1018 | proc_pid(overlap->lf_owner) != matchpid) |
| 1019 | continue; |
| 1020 | |
| 1021 | /* |
| 1022 | * does it block us? |
| 1023 | */ |
| 1024 | if ((lock->lf_type == F_WRLCK || overlap->lf_type == F_WRLCK)) |
| 1025 | return (overlap); |
| 1026 | } |
| 1027 | return (NOLOCKF); |
| 1028 | } |
| 1029 | |
| 1030 | |
| 1031 | /* |
| 1032 | * lf_findoverlap |
| 1033 | * |
| 1034 | * Description: Walk the list of locks to find an overlapping lock (if any). |
| 1035 | * |
| 1036 | * Parameters: lf First lock on lock list |
| 1037 | * lock The lock we are checking for an overlap |
| 1038 | * check Check type |
| 1039 | * prev pointer to pointer pointer to contain |
| 1040 | * address of pointer to previous lock |
| 1041 | * pointer to overlapping lock, if overlap |
| 1042 | * overlap pointer to pointer to contain address |
| 1043 | * of overlapping lock |
| 1044 | * |
| 1045 | * Returns: OVERLAP_NONE |
| 1046 | * OVERLAP_EQUALS_LOCK |
| 1047 | * OVERLAP_CONTAINS_LOCK |
| 1048 | * OVERLAP_CONTAINED_BY_LOCK |
| 1049 | * OVERLAP_STARTS_BEFORE_LOCK |
| 1050 | * OVERLAP_ENDS_AFTER_LOCK |
| 1051 | * |
| 1052 | * Implicit Returns: |
| 1053 | * *prev The address of the next pointer in the |
| 1054 | * lock previous to the overlapping lock; |
| 1055 | * this is generally used to relink the |
| 1056 | * lock list, avoiding a second iteration. |
| 1057 | * *overlap The pointer to the overlapping lock |
| 1058 | * itself; this is used to return data in |
| 1059 | * the check == OTHERS case, and for the |
| 1060 | * caller to modify the overlapping lock, |
| 1061 | * in the check == SELF case |
| 1062 | * |
| 1063 | * Note: This returns only the FIRST overlapping lock. There may be |
| 1064 | * more than one. lf_getlock will return the first blocking lock, |
| 1065 | * while lf_setlock will iterate over all overlapping locks to |
| 1066 | * |
| 1067 | * The check parameter can be SELF, meaning we are looking for |
| 1068 | * overlapping locks owned by us, or it can be OTHERS, meaning |
| 1069 | * we are looking for overlapping locks owned by someone else so |
| 1070 | * we can report a blocking lock on an F_GETLK request. |
| 1071 | * |
| 1072 | * The value of *overlap and *prev are modified, even if there is |
| 1073 | * no overlapping lock found; always check the return code. |
| 1074 | */ |
| 1075 | static overlap_t |
| 1076 | lf_findoverlap(struct lockf *lf, struct lockf *lock, int type, |
| 1077 | struct lockf ***prev, struct lockf **overlap) |
| 1078 | { |
| 1079 | off_t start, end; |
| 1080 | int found_self = 0; |
| 1081 | |
| 1082 | *overlap = lf; |
| 1083 | if (lf == NOLOCKF) |
| 1084 | return (0); |
| 1085 | #ifdef LOCKF_DEBUGGING |
| 1086 | if (lockf_debug & LF_DBG_LIST) |
| 1087 | lf_print("lf_findoverlap: looking for overlap in", lock); |
| 1088 | #endif /* LOCKF_DEBUGGING */ |
| 1089 | start = lock->lf_start; |
| 1090 | end = lock->lf_end; |
| 1091 | while (lf != NOLOCKF) { |
| 1092 | if (((type & SELF) && lf->lf_id != lock->lf_id) || |
| 1093 | ((type & OTHERS) && lf->lf_id == lock->lf_id)) { |
| 1094 | /* |
| 1095 | * Locks belonging to one process are adjacent on the |
| 1096 | * list, so if we've found any locks belonging to us, |
| 1097 | * and we're now seeing something else, then we've |
| 1098 | * examined all "self" locks. Note that bailing out |
| 1099 | * here is quite important; for coalescing, we assume |
| 1100 | * numerically adjacent locks from the same owner to |
| 1101 | * be adjacent on the list. |
| 1102 | */ |
| 1103 | if ((type & SELF) && found_self) { |
| 1104 | return OVERLAP_NONE; |
| 1105 | } |
| 1106 | |
| 1107 | *prev = &lf->lf_next; |
| 1108 | *overlap = lf = lf->lf_next; |
| 1109 | continue; |
| 1110 | } |
| 1111 | |
| 1112 | if ((type & SELF)) { |
| 1113 | found_self = 1; |
| 1114 | } |
| 1115 | |
| 1116 | #ifdef LOCKF_DEBUGGING |
| 1117 | if (lockf_debug & LF_DBG_LIST) |
| 1118 | lf_print("\tchecking", lf); |
| 1119 | #endif /* LOCKF_DEBUGGING */ |
| 1120 | /* |
| 1121 | * OK, check for overlap |
| 1122 | */ |
| 1123 | if ((lf->lf_end != -1 && start > lf->lf_end) || |
| 1124 | (end != -1 && lf->lf_start > end)) { |
| 1125 | /* Case 0 */ |
| 1126 | LOCKF_DEBUG(LF_DBG_LIST, "no overlap\n"); |
| 1127 | |
| 1128 | /* |
| 1129 | * NOTE: assumes that locks for the same process are |
| 1130 | * nonintersecting and ordered. |
| 1131 | */ |
| 1132 | if ((type & SELF) && end != -1 && lf->lf_start > end) |
| 1133 | return (OVERLAP_NONE); |
| 1134 | *prev = &lf->lf_next; |
| 1135 | *overlap = lf = lf->lf_next; |
| 1136 | continue; |
| 1137 | } |
| 1138 | if ((lf->lf_start == start) && (lf->lf_end == end)) { |
| 1139 | LOCKF_DEBUG(LF_DBG_LIST, "overlap == lock\n"); |
| 1140 | return (OVERLAP_EQUALS_LOCK); |
| 1141 | } |
| 1142 | if ((lf->lf_start <= start) && |
| 1143 | (end != -1) && |
| 1144 | ((lf->lf_end >= end) || (lf->lf_end == -1))) { |
| 1145 | LOCKF_DEBUG(LF_DBG_LIST, "overlap contains lock\n"); |
| 1146 | return (OVERLAP_CONTAINS_LOCK); |
| 1147 | } |
| 1148 | if (start <= lf->lf_start && |
| 1149 | (end == -1 || |
| 1150 | (lf->lf_end != -1 && end >= lf->lf_end))) { |
| 1151 | LOCKF_DEBUG(LF_DBG_LIST, "lock contains overlap\n"); |
| 1152 | return (OVERLAP_CONTAINED_BY_LOCK); |
| 1153 | } |
| 1154 | if ((lf->lf_start < start) && |
| 1155 | ((lf->lf_end >= start) || (lf->lf_end == -1))) { |
| 1156 | LOCKF_DEBUG(LF_DBG_LIST, "overlap starts before lock\n"); |
| 1157 | return (OVERLAP_STARTS_BEFORE_LOCK); |
| 1158 | } |
| 1159 | if ((lf->lf_start > start) && |
| 1160 | (end != -1) && |
| 1161 | ((lf->lf_end > end) || (lf->lf_end == -1))) { |
| 1162 | LOCKF_DEBUG(LF_DBG_LIST, "overlap ends after lock\n"); |
| 1163 | return (OVERLAP_ENDS_AFTER_LOCK); |
| 1164 | } |
| 1165 | panic("lf_findoverlap: default"); |
| 1166 | } |
| 1167 | return (OVERLAP_NONE); |
| 1168 | } |
| 1169 | |
| 1170 | |
| 1171 | /* |
| 1172 | * lf_split |
| 1173 | * |
| 1174 | * Description: Split a lock and a contained region into two or three locks |
| 1175 | * as necessary. |
| 1176 | * |
| 1177 | * Parameters: lock1 Lock to split |
| 1178 | * lock2 Overlapping lock region requiring the |
| 1179 | * split (upgrade/downgrade/unlock) |
| 1180 | * |
| 1181 | * Returns: 0 Success |
| 1182 | * ENOLCK No memory for new lock |
| 1183 | * |
| 1184 | * Implicit Returns: |
| 1185 | * *lock1 Modified original lock |
| 1186 | * *lock2 Overlapping lock (inserted into list) |
| 1187 | * (new lock) Potential new lock inserted into list |
| 1188 | * if split results in 3 locks |
| 1189 | * |
| 1190 | * Notes: This operation can only fail if the split would result in three |
| 1191 | * locks, and there is insufficient memory to allocate the third |
| 1192 | * lock; in that case, neither of the locks will be modified. |
| 1193 | */ |
| 1194 | static int |
| 1195 | lf_split(struct lockf *lock1, struct lockf *lock2) |
| 1196 | { |
| 1197 | struct lockf *splitlock; |
| 1198 | |
| 1199 | #ifdef LOCKF_DEBUGGING |
| 1200 | if (lockf_debug & LF_DBG_LIST) { |
| 1201 | lf_print("lf_split", lock1); |
| 1202 | lf_print("splitting from", lock2); |
| 1203 | } |
| 1204 | #endif /* LOCKF_DEBUGGING */ |
| 1205 | /* |
| 1206 | * Check to see if splitting into only two pieces. |
| 1207 | */ |
| 1208 | if (lock1->lf_start == lock2->lf_start) { |
| 1209 | lock1->lf_start = lock2->lf_end + 1; |
| 1210 | lock2->lf_next = lock1; |
| 1211 | return (0); |
| 1212 | } |
| 1213 | if (lock1->lf_end == lock2->lf_end) { |
| 1214 | lock1->lf_end = lock2->lf_start - 1; |
| 1215 | lock2->lf_next = lock1->lf_next; |
| 1216 | lock1->lf_next = lock2; |
| 1217 | return (0); |
| 1218 | } |
| 1219 | /* |
| 1220 | * Make a new lock consisting of the last part of |
| 1221 | * the encompassing lock |
| 1222 | */ |
| 1223 | MALLOC(splitlock, struct lockf *, sizeof *splitlock, M_LOCKF, M_WAITOK); |
| 1224 | if (splitlock == NULL) |
| 1225 | return (ENOLCK); |
| 1226 | bcopy(lock1, splitlock, sizeof *splitlock); |
| 1227 | splitlock->lf_start = lock2->lf_end + 1; |
| 1228 | TAILQ_INIT(&splitlock->lf_blkhd); |
| 1229 | lock1->lf_end = lock2->lf_start - 1; |
| 1230 | /* |
| 1231 | * OK, now link it in |
| 1232 | */ |
| 1233 | splitlock->lf_next = lock1->lf_next; |
| 1234 | lock2->lf_next = splitlock; |
| 1235 | lock1->lf_next = lock2; |
| 1236 | |
| 1237 | return (0); |
| 1238 | } |
| 1239 | |
| 1240 | |
| 1241 | /* |
| 1242 | * lf_wakelock |
| 1243 | * |
| 1244 | * Wakeup a blocklist in the case of a downgrade or unlock, since others |
| 1245 | * waiting on the lock may now be able to acquire it. |
| 1246 | * |
| 1247 | * Parameters: listhead Lock list head on which waiters may |
| 1248 | * have pending locks |
| 1249 | * |
| 1250 | * Returns: <void> |
| 1251 | * |
| 1252 | * Notes: This function iterates a list of locks and wakes all waiters, |
| 1253 | * rather than only waiters for the contended regions. Because |
| 1254 | * of this, for heavily contended files, this can result in a |
| 1255 | * "thundering herd" situation. Refactoring the code could make |
| 1256 | * this operation more efficient, if heavy contention ever results |
| 1257 | * in a real-world performance problem. |
| 1258 | */ |
| 1259 | static void |
| 1260 | lf_wakelock(struct lockf *listhead, boolean_t force_all) |
| 1261 | { |
| 1262 | struct lockf *wakelock; |
| 1263 | boolean_t wake_all = TRUE; |
| 1264 | |
| 1265 | if (force_all == FALSE && (listhead->lf_flags & F_WAKE1_SAFE)) |
| 1266 | wake_all = FALSE; |
| 1267 | |
| 1268 | while (!TAILQ_EMPTY(&listhead->lf_blkhd)) { |
| 1269 | wakelock = TAILQ_FIRST(&listhead->lf_blkhd); |
| 1270 | TAILQ_REMOVE(&listhead->lf_blkhd, wakelock, lf_block); |
| 1271 | |
| 1272 | wakelock->lf_next = NOLOCKF; |
| 1273 | #ifdef LOCKF_DEBUGGING |
| 1274 | if (lockf_debug & LF_DBG_LOCKOP) |
| 1275 | lf_print("lf_wakelock: awakening", wakelock); |
| 1276 | #endif /* LOCKF_DEBUGGING */ |
| 1277 | if (wake_all == FALSE) { |
| 1278 | /* |
| 1279 | * If there are items on the list head block list, |
| 1280 | * move them to the wakelock list instead, and then |
| 1281 | * correct their lf_next pointers. |
| 1282 | */ |
| 1283 | if (!TAILQ_EMPTY(&listhead->lf_blkhd)) { |
| 1284 | TAILQ_CONCAT(&wakelock->lf_blkhd, &listhead->lf_blkhd, lf_block); |
| 1285 | |
| 1286 | struct lockf *tlock; |
| 1287 | |
| 1288 | TAILQ_FOREACH(tlock, &wakelock->lf_blkhd, lf_block) { |
| 1289 | if (TAILQ_NEXT(tlock, lf_block) == tlock) { |
| 1290 | /* See rdar://10887303 */ |
| 1291 | panic("cycle in wakelock list"); |
| 1292 | } |
| 1293 | tlock->lf_next = wakelock; |
| 1294 | } |
| 1295 | } |
| 1296 | } |
| 1297 | wakeup(wakelock); |
| 1298 | |
| 1299 | if (wake_all == FALSE) |
| 1300 | break; |
| 1301 | } |
| 1302 | } |
| 1303 | |
| 1304 | |
| 1305 | #ifdef LOCKF_DEBUGGING |
| 1306 | #define GET_LF_OWNER_PID(lf) (proc_pid((lf)->lf_owner)) |
| 1307 | |
| 1308 | /* |
| 1309 | * lf_print DEBUG |
| 1310 | * |
| 1311 | * Print out a lock; lock information is prefixed by the string in 'tag' |
| 1312 | * |
| 1313 | * Parameters: tag A string tag for debugging |
| 1314 | * lock The lock whose information should be |
| 1315 | * displayed |
| 1316 | * |
| 1317 | * Returns: <void> |
| 1318 | */ |
| 1319 | void |
| 1320 | lf_print(const char *tag, struct lockf *lock) |
| 1321 | { |
| 1322 | printf("%s: lock %p for ", tag, (void *)lock); |
| 1323 | if (lock->lf_flags & F_POSIX) |
| 1324 | printf("proc %p (owner %d)", |
| 1325 | lock->lf_id, GET_LF_OWNER_PID(lock)); |
| 1326 | else if (lock->lf_flags & F_OFD_LOCK) |
| 1327 | printf("fg %p (owner %d)", |
| 1328 | lock->lf_id, GET_LF_OWNER_PID(lock)); |
| 1329 | else |
| 1330 | printf("id %p", (void *)lock->lf_id); |
| 1331 | if (lock->lf_vnode != 0) |
| 1332 | printf(" in vno %p, %s, start 0x%016llx, end 0x%016llx", |
| 1333 | lock->lf_vnode, |
| 1334 | lock->lf_type == F_RDLCK ? "shared" : |
| 1335 | lock->lf_type == F_WRLCK ? "exclusive" : |
| 1336 | lock->lf_type == F_UNLCK ? "unlock" : "unknown", |
| 1337 | (intmax_t)lock->lf_start, (intmax_t)lock->lf_end); |
| 1338 | else |
| 1339 | printf(" %s, start 0x%016llx, end 0x%016llx", |
| 1340 | lock->lf_type == F_RDLCK ? "shared" : |
| 1341 | lock->lf_type == F_WRLCK ? "exclusive" : |
| 1342 | lock->lf_type == F_UNLCK ? "unlock" : "unknown", |
| 1343 | (intmax_t)lock->lf_start, (intmax_t)lock->lf_end); |
| 1344 | if (!TAILQ_EMPTY(&lock->lf_blkhd)) |
| 1345 | printf(" block %p\n", (void *)TAILQ_FIRST(&lock->lf_blkhd)); |
| 1346 | else |
| 1347 | printf("\n"); |
| 1348 | } |
| 1349 | |
| 1350 | |
| 1351 | /* |
| 1352 | * lf_printlist DEBUG |
| 1353 | * |
| 1354 | * Print out a lock list for the vnode associated with 'lock'; lock information |
| 1355 | * is prefixed by the string in 'tag' |
| 1356 | * |
| 1357 | * Parameters: tag A string tag for debugging |
| 1358 | * lock The lock whose vnode's lock list should |
| 1359 | * be displayed |
| 1360 | * |
| 1361 | * Returns: <void> |
| 1362 | */ |
| 1363 | void |
| 1364 | lf_printlist(const char *tag, struct lockf *lock) |
| 1365 | { |
| 1366 | struct lockf *lf, *blk; |
| 1367 | |
| 1368 | if (lock->lf_vnode == 0) |
| 1369 | return; |
| 1370 | |
| 1371 | printf("%s: Lock list for vno %p:\n", |
| 1372 | tag, lock->lf_vnode); |
| 1373 | for (lf = lock->lf_vnode->v_lockf; lf; lf = lf->lf_next) { |
| 1374 | printf("\tlock %p for ",(void *)lf); |
| 1375 | if (lf->lf_flags & F_POSIX) |
| 1376 | printf("proc %p (owner %d)", |
| 1377 | lf->lf_id, GET_LF_OWNER_PID(lf)); |
| 1378 | else if (lf->lf_flags & F_OFD_LOCK) |
| 1379 | printf("fg %p (owner %d)", |
| 1380 | lf->lf_id, GET_LF_OWNER_PID(lf)); |
| 1381 | else |
| 1382 | printf("id %p", (void *)lf->lf_id); |
| 1383 | printf(", %s, start 0x%016llx, end 0x%016llx", |
| 1384 | lf->lf_type == F_RDLCK ? "shared" : |
| 1385 | lf->lf_type == F_WRLCK ? "exclusive" : |
| 1386 | lf->lf_type == F_UNLCK ? "unlock" : |
| 1387 | "unknown", (intmax_t)lf->lf_start, (intmax_t)lf->lf_end); |
| 1388 | TAILQ_FOREACH(blk, &lf->lf_blkhd, lf_block) { |
| 1389 | printf("\n\t\tlock request %p for ", (void *)blk); |
| 1390 | if (blk->lf_flags & F_POSIX) |
| 1391 | printf("proc %p (owner %d)", |
| 1392 | blk->lf_id, GET_LF_OWNER_PID(blk)); |
| 1393 | else if (blk->lf_flags & F_OFD_LOCK) |
| 1394 | printf("fg %p (owner %d)", |
| 1395 | blk->lf_id, GET_LF_OWNER_PID(blk)); |
| 1396 | else |
| 1397 | printf("id %p", (void *)blk->lf_id); |
| 1398 | printf(", %s, start 0x%016llx, end 0x%016llx", |
| 1399 | blk->lf_type == F_RDLCK ? "shared" : |
| 1400 | blk->lf_type == F_WRLCK ? "exclusive" : |
| 1401 | blk->lf_type == F_UNLCK ? "unlock" : |
| 1402 | "unknown", (intmax_t)blk->lf_start, |
| 1403 | (intmax_t)blk->lf_end); |
| 1404 | if (!TAILQ_EMPTY(&blk->lf_blkhd)) |
| 1405 | panic("lf_printlist: bad list"); |
| 1406 | } |
| 1407 | printf("\n"); |
| 1408 | } |
| 1409 | } |
| 1410 | #endif /* LOCKF_DEBUGGING */ |
| 1411 | |
| 1412 | #if IMPORTANCE_INHERITANCE |
| 1413 | |
| 1414 | /* |
| 1415 | * lf_hold_assertion |
| 1416 | * |
| 1417 | * Call task importance hold assertion on the owner of the lock. |
| 1418 | * |
| 1419 | * Parameters: block_task Owner of the lock blocking |
| 1420 | * current thread. |
| 1421 | * |
| 1422 | * block lock on which the current thread |
| 1423 | * is blocking on. |
| 1424 | * |
| 1425 | * Returns: <void> |
| 1426 | * |
| 1427 | * Notes: The task reference on block_task is not needed to be hold since |
| 1428 | * the current thread has vnode lock and block_task has a file |
| 1429 | * lock, thus removing file lock in exit requires block_task to |
| 1430 | * grab the vnode lock. |
| 1431 | */ |
| 1432 | static void |
| 1433 | lf_hold_assertion(task_t block_task, struct lockf *block) |
| 1434 | { |
| 1435 | if (task_importance_hold_file_lock_assertion(block_task, 1) == 0) { |
| 1436 | block->lf_boosted = LF_BOOSTED; |
| 1437 | LOCKF_DEBUG(LF_DBG_IMPINH, |
| 1438 | "lf: importance hold file lock assert on pid %d lock %p\n", |
| 1439 | proc_pid(block->lf_owner), block); |
| 1440 | } |
| 1441 | } |
| 1442 | |
| 1443 | |
| 1444 | /* |
| 1445 | * lf_jump_to_queue_head |
| 1446 | * |
| 1447 | * Jump the lock from the tail of the block queue to the head of |
| 1448 | * the queue. |
| 1449 | * |
| 1450 | * Parameters: block lockf struct containing the |
| 1451 | * block queue. |
| 1452 | * lock lockf struct to be jumped to the |
| 1453 | * front. |
| 1454 | * |
| 1455 | * Returns: <void> |
| 1456 | */ |
| 1457 | static void |
| 1458 | lf_jump_to_queue_head(struct lockf *block, struct lockf *lock) |
| 1459 | { |
| 1460 | /* Move the lock to the head of the block queue. */ |
| 1461 | TAILQ_REMOVE(&block->lf_blkhd, lock, lf_block); |
| 1462 | TAILQ_INSERT_HEAD(&block->lf_blkhd, lock, lf_block); |
| 1463 | } |
| 1464 | |
| 1465 | |
| 1466 | /* |
| 1467 | * lf_drop_assertion |
| 1468 | * |
| 1469 | * Drops the task hold assertion. |
| 1470 | * |
| 1471 | * Parameters: block lockf struct holding the assertion. |
| 1472 | * |
| 1473 | * Returns: <void> |
| 1474 | */ |
| 1475 | static void |
| 1476 | lf_drop_assertion(struct lockf *block) |
| 1477 | { |
| 1478 | LOCKF_DEBUG(LF_DBG_IMPINH, "lf: %d: dropping assertion for lock %p\n", |
| 1479 | proc_pid(block->lf_owner), block); |
| 1480 | |
| 1481 | task_t current_task = proc_task(block->lf_owner); |
| 1482 | task_importance_drop_file_lock_assertion(current_task, 1); |
| 1483 | block->lf_boosted = LF_NOT_BOOSTED; |
| 1484 | } |
| 1485 | |
| 1486 | static void |
| 1487 | lf_boost_blocking_proc(struct lockf *lock, struct lockf *block) |
| 1488 | { |
| 1489 | task_t ltask = proc_task(lock->lf_owner); |
| 1490 | task_t btask = proc_task(block->lf_owner); |
| 1491 | |
| 1492 | /* |
| 1493 | * Check if ltask can donate importance. The |
| 1494 | * check of imp_donor bit is done without holding |
| 1495 | * any lock. The value may change after you read it, |
| 1496 | * but it is ok to boost a task while someone else is |
| 1497 | * unboosting you. |
| 1498 | * |
| 1499 | * TODO: Support live inheritance on file locks. |
| 1500 | */ |
| 1501 | if (task_is_importance_donor(ltask)) { |
| 1502 | LOCKF_DEBUG(LF_DBG_IMPINH, |
| 1503 | "lf: %d: attempt to boost pid %d that holds lock %p\n", |
| 1504 | proc_pid(lock->lf_owner), proc_pid(block->lf_owner), block); |
| 1505 | |
| 1506 | if (block->lf_boosted != LF_BOOSTED && |
| 1507 | task_is_importance_receiver_type(btask)) { |
| 1508 | lf_hold_assertion(btask, block); |
| 1509 | } |
| 1510 | lf_jump_to_queue_head(block, lock); |
| 1511 | } |
| 1512 | } |
| 1513 | #endif /* IMPORTANCE_INHERITANCE */ |