sys/vfs/hammer/hammer_subs.c

   1 /*
   2  * Copyright (c) 2007-2008 The DragonFly Project.  All rights reserved.
   3  *
   4  * This code is derived from software contributed to The DragonFly Project
   5  * by Matthew Dillon <dillon@backplane.com>
   6  *
   7  * Redistribution and use in source and binary forms, with or without
   8  * modification, are permitted provided that the following conditions
   9  * are met:
  10  *
  11  * 1. Redistributions of source code must retain the above copyright
  12  *    notice, this list of conditions and the following disclaimer.
  13  * 2. Redistributions in binary form must reproduce the above copyright
  14  *    notice, this list of conditions and the following disclaimer in
  15  *    the documentation and/or other materials provided with the
  16  *    distribution.
  17  * 3. Neither the name of The DragonFly Project nor the names of its
  18  *    contributors may be used to endorse or promote products derived
  19  *    from this software without specific, prior written permission.
  20  *
  21  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  22  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  23  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
  24  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE
  25  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
  26  * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
  27  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
  28  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
  29  * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
  30  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
  31  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  32  * SUCH DAMAGE.
  33  *
  34  * $DragonFly: src/sys/vfs/hammer/hammer_subs.c,v 1.31 2008/07/07 03:49:51 dillon Exp $
  35  */
  36 /*
  37  * HAMMER structural locking
  38  */
  39
  40 #include "hammer.h"
  41 #include <sys/dirent.h>
  42
  43 void
  44 hammer_lock_ex_ident(struct hammer_lock *lock, const char *ident)
  45 {
  46         thread_t td = curthread;
  47
  48         KKASSERT(lock->refs > 0);
  49         crit_enter();
  50         if (lock->locktd != td) {
  51                 while (lock->locktd != NULL || lock->lockcount) {
  52                         ++lock->exwanted;
  53                         lock->wanted = 1;
  54                         if (hammer_debug_locks) {
  55                                 kprintf("hammer_lock_ex: held by %p\n",
  56                                         lock->locktd);
  57                         }
  58                         ++hammer_contention_count;
  59                         tsleep(lock, 0, ident, 0);
  60                         if (hammer_debug_locks)
  61                                 kprintf("hammer_lock_ex: try again\n");
  62                         --lock->exwanted;
  63                 }
  64                 lock->locktd = td;
  65         }
  66         KKASSERT(lock->lockcount >= 0);
  67         ++lock->lockcount;
  68         crit_exit();
  69 }
  70
  71 /*
  72  * Try to obtain an exclusive lock
  73  */
  74 int
  75 hammer_lock_ex_try(struct hammer_lock *lock)
  76 {
  77         thread_t td = curthread;
  78
  79         KKASSERT(lock->refs > 0);
  80         crit_enter();
  81         if (lock->locktd != td) {
  82                 if (lock->locktd != NULL || lock->lockcount) {
  83                         crit_exit();
  84                         return(EAGAIN);
  85                 }
  86                 lock->locktd = td;
  87         }
  88         KKASSERT(lock->lockcount >= 0);
  89         ++lock->lockcount;
  90         crit_exit();
  91         return(0);
  92 }
  93
  94 /*
  95  * Obtain a shared lock
  96  */
  97 void
  98 hammer_lock_sh(struct hammer_lock *lock)
  99 {
 100         KKASSERT(lock->refs > 0);
 101         crit_enter();
 102         while (lock->locktd != NULL) {
 103                 if (lock->locktd == curthread) {
 104                         Debugger("hammer_lock_sh: lock_sh on exclusive");
 105                         ++lock->lockcount;
 106                         crit_exit();
 107                         return;
 108                 }
 109                 lock->wanted = 1;
 110                 tsleep(lock, 0, "hmrlck", 0);
 111         }
 112         KKASSERT(lock->lockcount <= 0);
 113         --lock->lockcount;
 114         crit_exit();
 115 }
 116
 117 /*
 118  * Obtain a shared lock at a lower priority then thread waiting for an
 119  * exclusive lock.  To avoid a deadlock this may only be done if no other
 120  * shared locks are being held by the caller.
 121  */
 122 void
 123 hammer_lock_sh_lowpri(struct hammer_lock *lock)
 124 {
 125         KKASSERT(lock->refs > 0);
 126         crit_enter();
 127         while (lock->locktd != NULL || lock->exwanted) {
 128                 if (lock->locktd == curthread) {
 129                         Debugger("hammer_lock_sh: lock_sh on exclusive");
 130                         ++lock->lockcount;
 131                         crit_exit();
 132                         return;
 133                 }
 134                 lock->wanted = 1;
 135                 tsleep(lock, 0, "hmrlck", 0);
 136         }
 137         KKASSERT(lock->lockcount <= 0);
 138         --lock->lockcount;
 139         crit_exit();
 140 }
 141
 142 int
 143 hammer_lock_sh_try(struct hammer_lock *lock)
 144 {
 145         KKASSERT(lock->refs > 0);
 146         crit_enter();
 147         if (lock->locktd) {
 148                 crit_exit();
 149                 return(EAGAIN);
 150         }
 151         KKASSERT(lock->lockcount <= 0);
 152         --lock->lockcount;
 153         crit_exit();
 154         return(0);
 155 }
 156
 157 /*
 158  * Upgrade a shared lock to an exclusively held lock.  This function will
 159  * return EDEADLK If there is more then one shared holder.
 160  *
 161  * No error occurs and no action is taken if the lock is already exclusively
 162  * held by the caller.  If the lock is not held at all or held exclusively
 163  * by someone else, this function will panic.
 164  */
 165 int
 166 hammer_lock_upgrade(struct hammer_lock *lock)
 167 {
 168         int error;
 169
 170         crit_enter();
 171         if (lock->lockcount > 0) {
 172                 if (lock->locktd != curthread)
 173                         panic("hammer_lock_upgrade: illegal lock state");
 174                 error = 0;
 175         } else if (lock->lockcount == -1) {
 176                 lock->lockcount = 1;
 177                 lock->locktd = curthread;
 178                 error = 0;
 179         } else if (lock->lockcount != 0) {
 180                 error = EDEADLK;
 181         } else {
 182                 panic("hammer_lock_upgrade: lock is not held");
 183                 /* NOT REACHED */
 184                 error = 0;
 185         }
 186         crit_exit();
 187         return(error);
 188 }
 189
 190 /*
 191  * Downgrade an exclusively held lock to a shared lock.
 192  */
 193 void
 194 hammer_lock_downgrade(struct hammer_lock *lock)
 195 {
 196         KKASSERT(lock->lockcount == 1 && lock->locktd == curthread);
 197         crit_enter();
 198         lock->lockcount = -1;
 199         lock->locktd = NULL;
 200         if (lock->wanted) {
 201                 lock->wanted = 0;
 202                 wakeup(lock);
 203         }
 204         crit_exit();
 205         /* XXX memory barrier */
 206 }
 207
 208 void
 209 hammer_unlock(struct hammer_lock *lock)
 210 {
 211         crit_enter();
 212         KKASSERT(lock->lockcount != 0);
 213         if (lock->lockcount < 0) {
 214                 if (++lock->lockcount == 0 && lock->wanted) {
 215                         lock->wanted = 0;
 216                         wakeup(lock);
 217                 }
 218         } else {
 219                 KKASSERT(lock->locktd == curthread);
 220                 if (--lock->lockcount == 0) {
 221                         lock->locktd = NULL;
 222                         if (lock->wanted) {
 223                                 lock->wanted = 0;
 224                                 wakeup(lock);
 225                         }
 226                 }
 227
 228         }
 229         crit_exit();
 230 }
 231
 232 /*
 233  * The calling thread must be holding a shared or exclusive lock.
 234  * Returns < 0 if lock is held shared, and > 0 if held exlusively.
 235  */
 236 int
 237 hammer_lock_status(struct hammer_lock *lock)
 238 {
 239         if (lock->lockcount < 0)
 240                 return(-1);
 241         if (lock->lockcount > 0)
 242                 return(1);
 243         panic("hammer_lock_status: lock must be held: %p", lock);
 244 }
 245
 246 void
 247 hammer_ref(struct hammer_lock *lock)
 248 {
 249         KKASSERT(lock->refs >= 0);
 250         crit_enter();
 251         ++lock->refs;
 252         crit_exit();
 253 }
 254
 255 void
 256 hammer_unref(struct hammer_lock *lock)
 257 {
 258         KKASSERT(lock->refs > 0);
 259         crit_enter();
 260         --lock->refs;
 261         crit_exit();
 262 }
 263
 264 /*
 265  * The sync_lock must be held when doing any modifying operations on
 266  * meta-data.  The flusher holds the lock exclusively while the reblocker
 267  * and pruner use a shared lock.
 268  *
 269  * Modifying operations can run in parallel until the flusher needs to
 270  * sync the disk media.
 271  */
 272 void
 273 hammer_sync_lock_ex(hammer_transaction_t trans)
 274 {
 275         ++trans->sync_lock_refs;
 276         hammer_lock_ex(&trans->hmp->sync_lock);
 277 }
 278
 279 void
 280 hammer_sync_lock_sh(hammer_transaction_t trans)
 281 {
 282         ++trans->sync_lock_refs;
 283         hammer_lock_sh(&trans->hmp->sync_lock);
 284 }
 285
 286 int
 287 hammer_sync_lock_sh_try(hammer_transaction_t trans)
 288 {
 289         int error;
 290
 291         ++trans->sync_lock_refs;
 292         if ((error = hammer_lock_sh_try(&trans->hmp->sync_lock)) != 0)
 293                 --trans->sync_lock_refs;
 294         return (error);
 295 }
 296
 297 void
 298 hammer_sync_unlock(hammer_transaction_t trans)
 299 {
 300         --trans->sync_lock_refs;
 301         hammer_unlock(&trans->hmp->sync_lock);
 302 }
 303
 304 /*
 305  * Misc
 306  */
 307 u_int32_t
 308 hammer_to_unix_xid(uuid_t *uuid)
 309 {
 310         return(*(u_int32_t *)&uuid->node[2]);
 311 }
 312
 313 void
 314 hammer_guid_to_uuid(uuid_t *uuid, u_int32_t guid)
 315 {
 316         bzero(uuid, sizeof(*uuid));
 317         *(u_int32_t *)&uuid->node[2] = guid;
 318 }
 319
 320 void
 321 hammer_time_to_timespec(u_int64_t xtime, struct timespec *ts)
 322 {
 323         ts->tv_sec = (unsigned long)(xtime / 1000000);
 324         ts->tv_nsec = (unsigned int)(xtime % 1000000) * 1000L;
 325 }
 326
 327 u_int64_t
 328 hammer_timespec_to_time(struct timespec *ts)
 329 {
 330         u_int64_t xtime;
 331
 332         xtime = (unsigned)(ts->tv_nsec / 1000) +
 333                 (unsigned long)ts->tv_sec * 1000000ULL;
 334         return(xtime);
 335 }
 336
 337
 338 /*
 339  * Convert a HAMMER filesystem object type to a vnode type
 340  */
 341 enum vtype
 342 hammer_get_vnode_type(u_int8_t obj_type)
 343 {
 344         switch(obj_type) {
 345         case HAMMER_OBJTYPE_DIRECTORY:
 346                 return(VDIR);
 347         case HAMMER_OBJTYPE_REGFILE:
 348                 return(VREG);
 349         case HAMMER_OBJTYPE_DBFILE:
 350                 return(VDATABASE);
 351         case HAMMER_OBJTYPE_FIFO:
 352                 return(VFIFO);
 353         case HAMMER_OBJTYPE_SOCKET:
 354                 return(VSOCK);
 355         case HAMMER_OBJTYPE_CDEV:
 356                 return(VCHR);
 357         case HAMMER_OBJTYPE_BDEV:
 358                 return(VBLK);
 359         case HAMMER_OBJTYPE_SOFTLINK:
 360                 return(VLNK);
 361         default:
 362                 return(VBAD);
 363         }
 364         /* not reached */
 365 }
 366
 367 int
 368 hammer_get_dtype(u_int8_t obj_type)
 369 {
 370         switch(obj_type) {
 371         case HAMMER_OBJTYPE_DIRECTORY:
 372                 return(DT_DIR);
 373         case HAMMER_OBJTYPE_REGFILE:
 374                 return(DT_REG);
 375         case HAMMER_OBJTYPE_DBFILE:
 376                 return(DT_DBF);
 377         case HAMMER_OBJTYPE_FIFO:
 378                 return(DT_FIFO);
 379         case HAMMER_OBJTYPE_SOCKET:
 380                 return(DT_SOCK);
 381         case HAMMER_OBJTYPE_CDEV:
 382                 return(DT_CHR);
 383         case HAMMER_OBJTYPE_BDEV:
 384                 return(DT_BLK);
 385         case HAMMER_OBJTYPE_SOFTLINK:
 386                 return(DT_LNK);
 387         default:
 388                 return(DT_UNKNOWN);
 389         }
 390         /* not reached */
 391 }
 392
 393 u_int8_t
 394 hammer_get_obj_type(enum vtype vtype)
 395 {
 396         switch(vtype) {
 397         case VDIR:
 398                 return(HAMMER_OBJTYPE_DIRECTORY);
 399         case VREG:
 400                 return(HAMMER_OBJTYPE_REGFILE);
 401         case VDATABASE:
 402                 return(HAMMER_OBJTYPE_DBFILE);
 403         case VFIFO:
 404                 return(HAMMER_OBJTYPE_FIFO);
 405         case VSOCK:
 406                 return(HAMMER_OBJTYPE_SOCKET);
 407         case VCHR:
 408                 return(HAMMER_OBJTYPE_CDEV);
 409         case VBLK:
 410                 return(HAMMER_OBJTYPE_BDEV);
 411         case VLNK:
 412                 return(HAMMER_OBJTYPE_SOFTLINK);
 413         default:
 414                 return(HAMMER_OBJTYPE_UNKNOWN);
 415         }
 416         /* not reached */
 417 }
 418
 419 /*
 420  * Return flags for hammer_delete_at_cursor()
 421  */
 422 int
 423 hammer_nohistory(hammer_inode_t ip)
 424 {
 425         if (ip->hmp->hflags & HMNT_NOHISTORY)
 426                 return(HAMMER_DELETE_DESTROY);
 427         if (ip->ino_data.uflags & (SF_NOHISTORY|UF_NOHISTORY))
 428                 return(HAMMER_DELETE_DESTROY);
 429         return(0);
 430 }
 431
 432 /*
 433  * Return a namekey hash.   The 64 bit namekey hash consists of a 32 bit
 434  * crc in the MSB and 0 in the LSB.  The caller will use the low bits to
 435  * generate a unique key and will scan all entries with the same upper
 436  * 32 bits when issuing a lookup.
 437  *
 438  * We strip bit 63 in order to provide a positive key, this way a seek
 439  * offset of 0 will represent the base of the directory.
 440  *
 441  * This function can never return 0.  We use the MSB-0 space to synthesize
 442  * artificial directory entries such as "." and "..".
 443  */
 444 int64_t
 445 hammer_directory_namekey(const void *name, int len)
 446 {
 447         int64_t key;
 448
 449         key = (int64_t)(crc32(name, len) & 0x7FFFFFFF) << 32;
 450         if (key == 0)
 451                 key |= 0x100000000LL;
 452         return(key);
 453 }
 454
 455 hammer_tid_t
 456 hammer_str_to_tid(const char *str)
 457 {
 458         hammer_tid_t tid;
 459
 460         tid = strtouq(str, NULL, 0);                    /* full TID */
 461         return(tid);
 462 }
 463
 464 void
 465 hammer_crc_set_blockmap(hammer_blockmap_t blockmap)
 466 {
 467         blockmap->entry_crc = crc32(blockmap, HAMMER_BLOCKMAP_CRCSIZE);
 468 }
 469
 470 void
 471 hammer_crc_set_volume(hammer_volume_ondisk_t ondisk)
 472 {
 473         ondisk->vol_crc = crc32(ondisk, HAMMER_VOL_CRCSIZE1) ^
 474                           crc32(&ondisk->vol_crc + 1, HAMMER_VOL_CRCSIZE2);
 475 }
 476
 477 int
 478 hammer_crc_test_blockmap(hammer_blockmap_t blockmap)
 479 {
 480         hammer_crc_t crc;
 481
 482         crc = crc32(blockmap, HAMMER_BLOCKMAP_CRCSIZE);
 483         return (blockmap->entry_crc == crc);
 484 }
 485
 486 int
 487 hammer_crc_test_volume(hammer_volume_ondisk_t ondisk)
 488 {
 489         hammer_crc_t crc;
 490
 491         crc = crc32(ondisk, HAMMER_VOL_CRCSIZE1) ^
 492               crc32(&ondisk->vol_crc + 1, HAMMER_VOL_CRCSIZE2);
 493         return (ondisk->vol_crc == crc);
 494 }
 495
 496 int
 497 hammer_crc_test_btree(hammer_node_ondisk_t ondisk)
 498 {
 499         hammer_crc_t crc;
 500
 501         crc = crc32(&ondisk->crc + 1, HAMMER_BTREE_CRCSIZE);
 502         return (ondisk->crc == crc);
 503 }
 504
 505 /*
 506  * Test or set the leaf->data_crc field.  Deal with any special cases given
 507  * a generic B-Tree leaf element and its data.
 508  *
 509  * NOTE: Inode-data: the atime and mtime fields are not CRCd, allowing them
 510  *       to be updated in-place.
 511  */
 512 int
 513 hammer_crc_test_leaf(void *data, hammer_btree_leaf_elm_t leaf)
 514 {
 515         hammer_crc_t crc;
 516
 517         if (leaf->data_len == 0) {
 518                 crc = 0;
 519         } else {
 520                 switch(leaf->base.rec_type) {
 521                 case HAMMER_RECTYPE_INODE:
 522                         if (leaf->data_len != sizeof(struct hammer_inode_data))
 523                                 return(0);
 524                         crc = crc32(data, HAMMER_INODE_CRCSIZE);
 525                         break;
 526                 default:
 527                         crc = crc32(data, leaf->data_len);
 528                         break;
 529                 }
 530         }
 531         return (leaf->data_crc == crc);
 532 }
 533
 534 void
 535 hammer_crc_set_leaf(void *data, hammer_btree_leaf_elm_t leaf)
 536 {
 537         if (leaf->data_len == 0) {
 538                 leaf->data_crc = 0;
 539         } else {
 540                 switch(leaf->base.rec_type) {
 541                 case HAMMER_RECTYPE_INODE:
 542                         KKASSERT(leaf->data_len ==
 543                                   sizeof(struct hammer_inode_data));
 544                         leaf->data_crc = crc32(data, HAMMER_INODE_CRCSIZE);
 545                         break;
 546                 default:
 547                         leaf->data_crc = crc32(data, leaf->data_len);
 548                         break;
 549                 }
 550         }
 551 }
 552
 553 void
 554 hkprintf(const char *ctl, ...)
 555 {
 556         __va_list va;
 557
 558         if (hammer_debug_debug) {
 559                 __va_start(va, ctl);
 560                 kvprintf(ctl, va);
 561                 __va_end(va);
 562         }
 563 }
 564
 565 /*
 566  * Return the block size at the specified file offset.
 567  */
 568 int
 569 hammer_blocksize(int64_t file_offset)
 570 {
 571         if (file_offset < HAMMER_XDEMARC)
 572                 return(HAMMER_BUFSIZE);
 573         else
 574                 return(HAMMER_XBUFSIZE);
 575 }
 576
 577 /*
 578  * Return the demarkation point between the two offsets where
 579  * the block size changes.
 580  */
 581 int64_t
 582 hammer_blockdemarc(int64_t file_offset1, int64_t file_offset2)
 583 {
 584         if (file_offset1 < HAMMER_XDEMARC) {
 585                 if (file_offset2 <= HAMMER_XDEMARC)
 586                         return(file_offset2);
 587                 return(HAMMER_XDEMARC);
 588         }
 589         panic("hammer_blockdemarc: illegal range %lld %lld\n",
 590               file_offset1, file_offset2);
 591 }
 592
 593 udev_t
 594 hammer_fsid_to_udev(uuid_t *uuid)
 595 {
 596         u_int32_t crc;
 597
 598         crc = crc32(uuid, sizeof(*uuid));
 599         return((udev_t)crc);
 600 }
 601