Many hyperlinks are disabled.
Use anonymous login
to enable hyperlinks.
Overview
Comment: | On unix, embargo close() operations until all locks have cleared from the file. Ticket #561. (CVS 1171) |
---|---|
Downloads: | Tarball | ZIP archive |
Timelines: | family | ancestors | descendants | both | trunk |
Files: | files | file ages | folders |
SHA1: |
1ebe5fc7b03a6b070a5d52ffedb95f0d |
User & Date: | drh 2004-01-12 00:39:06.000 |
Context
2004-01-14
| ||
03:12 | Remove an incomplete comment that somehow snuck into the sources. (CVS 1172) (check-in: ea1ad4653e user: drh tags: trunk) | |
2004-01-12
| ||
00:39 | On unix, embargo close() operations until all locks have cleared from the file. Ticket #561. (CVS 1171) (check-in: 1ebe5fc7b0 user: drh tags: trunk) | |
00:38 | Previous commit of changes to the in-memory backend was not quite right. This check-in should square things away. (CVS 1170) (check-in: 75d91e3bca user: drh tags: trunk) | |
Changes
Changes to src/os.c.
︙ | ︙ | |||
160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 | ** file is unlocked. cnt==-1 means the file has an exclusive lock. ** cnt>0 means there are cnt shared locks on the file. ** ** Any attempt to lock or unlock a file first checks the locking ** structure. The fcntl() system call is only invoked to set a ** POSIX lock if the internal lock structure transitions between ** a locked and an unlocked state. */ /* ** An instance of the following structure serves as the key used ** to locate a particular lockInfo structure given its inode. Note ** that we have to include the process ID as part of the key. On some ** threading implementations (ex: linux), each thread has a separate ** process ID. */ struct lockKey { dev_t dev; /* Device number */ ino_t ino; /* Inode number */ pid_t pid; /* Process ID */ }; /* | > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > | > > > | > > > > > > > > > > > > > > > > > > > > > > > > > | | | < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < | | | | | > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > | 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 | ** file is unlocked. cnt==-1 means the file has an exclusive lock. ** cnt>0 means there are cnt shared locks on the file. ** ** Any attempt to lock or unlock a file first checks the locking ** structure. The fcntl() system call is only invoked to set a ** POSIX lock if the internal lock structure transitions between ** a locked and an unlocked state. ** ** 2004-Jan-11: ** More recent discoveries about POSIX advisory locks. (The more ** I discover, the more I realize the a POSIX advisory locks are ** an abomination.) ** ** If you close a file descriptor that points to a file that has locks, ** all locks on that file that are owned by the current process are ** released. To work around this problem, each OsFile structure contains ** a pointer to an openCnt structure. There is one openCnt structure ** per open inode, which means that multiple OsFiles can point to a single ** openCnt. When an attempt is made to close an OsFile, if there are ** other OsFiles open on the same inode that are holding locks, the call ** to close() the file descriptor is deferred until all of the locks clear. ** The openCnt structure keeps a list of file descriptors that need to ** be closed and that list is walked (and cleared) when the last lock ** clears. ** ** First, under Linux threads, because each thread has a separate ** process ID, lock operations in one thread do not override locks ** to the same file in other threads. Linux threads behave like ** separate processes in this respect. But, if you close a file ** descriptor in linux threads, all locks are cleared, even locks ** on other threads and even though the other threads have different ** process IDs. Linux threads is inconsistent in this respect. ** (I'm beginning to think that linux threads is an abomination too.) ** The consequence of this all is that the hash table for the lockInfo ** structure has to include the process id as part of its key because ** locks in different threads are treated as distinct. But the ** openCnt structure should not include the process id in its ** key because close() clears lock on all threads, not just the current ** thread. Were it not for this goofiness in linux threads, we could ** combine the lockInfo and openCnt structures into a single structure. */ /* ** An instance of the following structure serves as the key used ** to locate a particular lockInfo structure given its inode. Note ** that we have to include the process ID as part of the key. On some ** threading implementations (ex: linux), each thread has a separate ** process ID. */ struct lockKey { dev_t dev; /* Device number */ ino_t ino; /* Inode number */ pid_t pid; /* Process ID */ }; /* ** An instance of the following structure is allocated for each open ** inode on each thread with a different process ID. (Threads have ** different process IDs on linux, but not on most other unixes.) ** ** A single inode can have multiple file descriptors, so each OsFile ** structure contains a pointer to an instance of this object and this ** object keeps a count of the number of OsFiles pointing to it. */ struct lockInfo { struct lockKey key; /* The lookup key */ int cnt; /* 0: unlocked. -1: write lock. 1...: read lock. */ int nRef; /* Number of pointers to this structure */ }; /* ** An instance of the following structure serves as the key used ** to locate a particular openCnt structure given its inode. This ** is the same as the lockKey except that the process ID is omitted. */ struct openKey { dev_t dev; /* Device number */ ino_t ino; /* Inode number */ }; /* ** An instance of the following structure is allocated for each open ** inode. This structure keeps track of the number of locks on that ** inode. If a close is attempted against an inode that is holding ** locks, the close is deferred until all locks clear by adding the ** file descriptor to be closed to the pending list. */ struct openCnt { struct openKey key; /* The lookup key */ int nRef; /* Number of pointers to this structure */ int nLock; /* Number of outstanding locks */ int nPending; /* Number of pending close() operations */ int *aPending; /* Malloced space holding fd's awaiting a close() */ }; /* ** These hash table maps inodes and process IDs into lockInfo and openCnt ** structures. Access to these hash tables must be protected by a mutex. */ static Hash lockHash = { SQLITE_HASH_BINARY, 0, 0, 0, 0, 0 }; static Hash openHash = { SQLITE_HASH_BINARY, 0, 0, 0, 0, 0 }; /* ** Release a lockInfo structure previously allocated by findLockInfo(). */ static void releaseLockInfo(struct lockInfo *pLock){ pLock->nRef--; if( pLock->nRef==0 ){ sqliteHashInsert(&lockHash, &pLock->key, sizeof(pLock->key), 0); sqliteFree(pLock); } } /* ** Release a openCnt structure previously allocated by findLockInfo(). */ static void releaseOpenCnt(struct openCnt *pOpen){ pOpen->nRef--; if( pOpen->nRef==0 ){ sqliteHashInsert(&openHash, &pOpen->key, sizeof(pOpen->key), 0); sqliteFree(pOpen->aPending); sqliteFree(pOpen); } } /* ** Given a file descriptor, locate lockInfo and openCnt structures that ** describes that file descriptor. Create a new ones if necessary. The ** return values might be unset if an error occurs. ** ** Return the number of errors. */ int findLockInfo( int fd, /* The file descriptor used in the key */ struct lockInfo **ppLock, /* Return the lockInfo structure here */ struct openCnt **ppOpen /* Return the openCnt structure here */ ){ int rc; struct lockKey key1; struct openKey key2; struct stat statbuf; struct lockInfo *pLock; struct openCnt *pOpen; rc = fstat(fd, &statbuf); if( rc!=0 ) return 1; memset(&key1, 0, sizeof(key1)); key1.dev = statbuf.st_dev; key1.ino = statbuf.st_ino; key1.pid = getpid(); memset(&key2, 0, sizeof(key2)); key2.dev = statbuf.st_dev; key2.ino = statbuf.st_ino; pLock = (struct lockInfo*)sqliteHashFind(&lockHash, &key1, sizeof(key1)); if( pLock==0 ){ struct lockInfo *pOld; pLock = sqliteMallocRaw( sizeof(*pLock) ); if( pLock==0 ) return 1; pLock->key = key1; pLock->nRef = 1; pLock->cnt = 0; pOld = sqliteHashInsert(&lockHash, &pLock->key, sizeof(key1), pLock); if( pOld!=0 ){ assert( pOld==pLock ); sqliteFree(pLock); return 1; } }else{ pLock->nRef++; } *ppLock = pLock; pOpen = (struct openCnt*)sqliteHashFind(&openHash, &key2, sizeof(key2)); if( pOpen==0 ){ struct openCnt *pOld; pOpen = sqliteMallocRaw( sizeof(*pOpen) ); if( pOpen==0 ){ releaseLockInfo(pLock); return 1; } pOpen->key = key2; pOpen->nRef = 1; pOpen->nLock = 0; pOpen->nPending = 0; pOpen->aPending = 0; pOld = sqliteHashInsert(&openHash, &pOpen->key, sizeof(key2), pOpen); if( pOld!=0 ){ assert( pOld==pOpen ); sqliteFree(pOpen); releaseLockInfo(pLock); return 1; } }else{ pOpen->nRef++; } *ppOpen = pOpen; return 0; } #endif /** POSIX advisory lock work-around **/ /* ** If we compile with the SQLITE_TEST macro set, then the following block ** of code will give us the ability to simulate a disk I/O error. This ** is used for testing the I/O recovery logic. */ |
︙ | ︙ | |||
345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 | */ int sqliteOsOpenReadWrite( const char *zFilename, OsFile *id, int *pReadonly ){ #if OS_UNIX id->dirfd = -1; id->fd = open(zFilename, O_RDWR|O_CREAT|O_LARGEFILE|O_BINARY, 0644); if( id->fd<0 ){ id->fd = open(zFilename, O_RDONLY|O_LARGEFILE|O_BINARY); if( id->fd<0 ){ return SQLITE_CANTOPEN; } *pReadonly = 1; }else{ *pReadonly = 0; } sqliteOsEnterMutex(); | > | | | 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 | */ int sqliteOsOpenReadWrite( const char *zFilename, OsFile *id, int *pReadonly ){ #if OS_UNIX int rc; id->dirfd = -1; id->fd = open(zFilename, O_RDWR|O_CREAT|O_LARGEFILE|O_BINARY, 0644); if( id->fd<0 ){ id->fd = open(zFilename, O_RDONLY|O_LARGEFILE|O_BINARY); if( id->fd<0 ){ return SQLITE_CANTOPEN; } *pReadonly = 1; }else{ *pReadonly = 0; } sqliteOsEnterMutex(); rc = findLockInfo(id->fd, &id->pLock, &id->pOpen); sqliteOsLeaveMutex(); if( rc ){ close(id->fd); return SQLITE_NOMEM; } id->locked = 0; TRACE3("OPEN %-3d %s\n", id->fd, zFilename); OpenCounter(+1); return SQLITE_OK; |
︙ | ︙ | |||
467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 | ** ** On success, write the file handle into *id and return SQLITE_OK. ** ** On failure, return SQLITE_CANTOPEN. */ int sqliteOsOpenExclusive(const char *zFilename, OsFile *id, int delFlag){ #if OS_UNIX if( access(zFilename, 0)==0 ){ return SQLITE_CANTOPEN; } id->dirfd = -1; id->fd = open(zFilename, O_RDWR|O_CREAT|O_EXCL|O_NOFOLLOW|O_LARGEFILE|O_BINARY, 0600); if( id->fd<0 ){ return SQLITE_CANTOPEN; } sqliteOsEnterMutex(); | > | | | 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 | ** ** On success, write the file handle into *id and return SQLITE_OK. ** ** On failure, return SQLITE_CANTOPEN. */ int sqliteOsOpenExclusive(const char *zFilename, OsFile *id, int delFlag){ #if OS_UNIX int rc; if( access(zFilename, 0)==0 ){ return SQLITE_CANTOPEN; } id->dirfd = -1; id->fd = open(zFilename, O_RDWR|O_CREAT|O_EXCL|O_NOFOLLOW|O_LARGEFILE|O_BINARY, 0600); if( id->fd<0 ){ return SQLITE_CANTOPEN; } sqliteOsEnterMutex(); rc = findLockInfo(id->fd, &id->pLock, &id->pOpen); sqliteOsLeaveMutex(); if( rc ){ close(id->fd); unlink(zFilename); return SQLITE_NOMEM; } id->locked = 0; if( delFlag ){ unlink(zFilename); |
︙ | ︙ | |||
557 558 559 560 561 562 563 564 565 566 567 568 569 | ** ** On success, write the file handle into *id and return SQLITE_OK. ** ** On failure, return SQLITE_CANTOPEN. */ int sqliteOsOpenReadOnly(const char *zFilename, OsFile *id){ #if OS_UNIX id->dirfd = -1; id->fd = open(zFilename, O_RDONLY|O_LARGEFILE|O_BINARY); if( id->fd<0 ){ return SQLITE_CANTOPEN; } sqliteOsEnterMutex(); | > | | | 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 | ** ** On success, write the file handle into *id and return SQLITE_OK. ** ** On failure, return SQLITE_CANTOPEN. */ int sqliteOsOpenReadOnly(const char *zFilename, OsFile *id){ #if OS_UNIX int rc; id->dirfd = -1; id->fd = open(zFilename, O_RDONLY|O_LARGEFILE|O_BINARY); if( id->fd<0 ){ return SQLITE_CANTOPEN; } sqliteOsEnterMutex(); rc = findLockInfo(id->fd, &id->pLock, &id->pOpen); sqliteOsLeaveMutex(); if( rc ){ close(id->fd); return SQLITE_NOMEM; } id->locked = 0; TRACE3("OPEN-RO %-3d %s\n", id->fd, zFilename); OpenCounter(+1); return SQLITE_OK; |
︙ | ︙ | |||
759 760 761 762 763 764 765 | if( !sqliteOsFileExists(zBuf) ) break; } #endif return SQLITE_OK; } /* | | | > > > > > > > > > > > > > > > > > > > > > | 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 | if( !sqliteOsFileExists(zBuf) ) break; } #endif return SQLITE_OK; } /* ** Close a file. */ int sqliteOsClose(OsFile *id){ #if OS_UNIX sqliteOsUnlock(id); if( id->dirfd>=0 ) close(id->dirfd); id->dirfd = -1; sqliteOsEnterMutex(); if( id->pOpen->nLock ){ /* If there are outstanding locks, do not actually close the file just ** yet because that would clear those locks. Instead, add the file ** descriptor to pOpen->aPending. It will be automatically closed when ** the last lock is cleared. */ int *aNew; struct openCnt *pOpen = id->pOpen; pOpen->nPending++; aNew = sqliteRealloc( pOpen->aPending, pOpen->nPending*sizeof(int) ); if( aNew==0 ){ /* If a malloc fails, just leak the file descriptor */ }else{ pOpen->aPending = aNew; pOpen->aPending[pOpen->nPending-1] = id->fd; } }else{ /* There are no outstanding locks so we can close the file immediately */ close(id->fd); } releaseLockInfo(id->pLock); releaseOpenCnt(id->pOpen); sqliteOsLeaveMutex(); TRACE2("CLOSE %-3d\n", id->fd); OpenCounter(-1); return SQLITE_OK; #endif #if OS_WIN CloseHandle(id->h); |
︙ | ︙ | |||
1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 | #if OS_UNIX int rc; sqliteOsEnterMutex(); if( id->pLock->cnt>0 ){ if( !id->locked ){ id->pLock->cnt++; id->locked = 1; } rc = SQLITE_OK; }else if( id->locked || id->pLock->cnt==0 ){ struct flock lock; int s; lock.l_type = F_RDLCK; lock.l_whence = SEEK_SET; lock.l_start = lock.l_len = 0L; s = fcntl(id->fd, F_SETLK, &lock); if( s!=0 ){ rc = (errno==EINVAL) ? SQLITE_NOLFS : SQLITE_BUSY; }else{ rc = SQLITE_OK; | > > | | > > | 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 | #if OS_UNIX int rc; sqliteOsEnterMutex(); if( id->pLock->cnt>0 ){ if( !id->locked ){ id->pLock->cnt++; id->locked = 1; id->pOpen->nLock++; } rc = SQLITE_OK; }else if( id->locked || id->pLock->cnt==0 ){ struct flock lock; int s; lock.l_type = F_RDLCK; lock.l_whence = SEEK_SET; lock.l_start = lock.l_len = 0L; s = fcntl(id->fd, F_SETLK, &lock); if( s!=0 ){ rc = (errno==EINVAL) ? SQLITE_NOLFS : SQLITE_BUSY; }else{ rc = SQLITE_OK; if( !id->locked ){ id->pOpen->nLock++; id->locked = 1; } id->pLock->cnt = 1; } }else{ rc = SQLITE_BUSY; } sqliteOsLeaveMutex(); return rc; #endif |
︙ | ︙ | |||
1272 1273 1274 1275 1276 1277 1278 | lock.l_whence = SEEK_SET; lock.l_start = lock.l_len = 0L; s = fcntl(id->fd, F_SETLK, &lock); if( s!=0 ){ rc = (errno==EINVAL) ? SQLITE_NOLFS : SQLITE_BUSY; }else{ rc = SQLITE_OK; | > | | > > | 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 | lock.l_whence = SEEK_SET; lock.l_start = lock.l_len = 0L; s = fcntl(id->fd, F_SETLK, &lock); if( s!=0 ){ rc = (errno==EINVAL) ? SQLITE_NOLFS : SQLITE_BUSY; }else{ rc = SQLITE_OK; if( !id->locked ){ id->pOpen->nLock++; id->locked = 1; } id->pLock->cnt = -1; } }else{ rc = SQLITE_BUSY; } sqliteOsLeaveMutex(); return rc; #endif |
︙ | ︙ | |||
1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 | s = fcntl(id->fd, F_SETLK, &lock); if( s!=0 ){ rc = (errno==EINVAL) ? SQLITE_NOLFS : SQLITE_BUSY; }else{ rc = SQLITE_OK; id->pLock->cnt = 0; } } sqliteOsLeaveMutex(); id->locked = 0; return rc; #endif #if OS_WIN int rc; | > > > > > > > > > > > > > > > > > > | 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 | s = fcntl(id->fd, F_SETLK, &lock); if( s!=0 ){ rc = (errno==EINVAL) ? SQLITE_NOLFS : SQLITE_BUSY; }else{ rc = SQLITE_OK; id->pLock->cnt = 0; } } if( rc==SQLITE_OK ){ /* Decrement the count of locks against this same file. When the ** count reaches zero, close any other file descriptors whose close ** was deferred because of outstanding locks. */ struct openCnt *pOpen = id->pOpen; pOpen->nLock--; assert( pOpen->nLock>=0 ); if( pOpen->nLock==0 && pOpen->nPending>0 ){ int i; for(i=0; i<pOpen->nPending; i++){ close(pOpen->aPending[i]); } sqliteFree(pOpen->aPending); pOpen->nPending = 0; pOpen->aPending = 0; } } sqliteOsLeaveMutex(); id->locked = 0; return rc; #endif #if OS_WIN int rc; |
︙ | ︙ |
Changes to src/os.h.
︙ | ︙ | |||
99 100 101 102 103 104 105 | #if OS_UNIX # include <sys/types.h> # include <sys/stat.h> # include <fcntl.h> # include <unistd.h> typedef struct OsFile OsFile; struct OsFile { | > | | | | | 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 | #if OS_UNIX # include <sys/types.h> # include <sys/stat.h> # include <fcntl.h> # include <unistd.h> typedef struct OsFile OsFile; struct OsFile { struct openCnt *pOpen; /* Info about all open fd's on this inode */ struct lockInfo *pLock; /* Info about locks on this inode */ int fd; /* The file descriptor */ int locked; /* True if this instance holds the lock */ int dirfd; /* File descriptor for the directory */ }; # define SQLITE_TEMPNAME_SIZE 200 # if defined(HAVE_USLEEP) && HAVE_USLEEP # define SQLITE_MIN_SLEEP_MS 1 # else # define SQLITE_MIN_SLEEP_MS 1000 # endif |
︙ | ︙ |