Many hyperlinks are disabled.
Use anonymous login
to enable hyperlinks.
Overview
Comment: | In os_unix.c, do not open the directory containing the database file when the database file is opened. Instead, wait until time to fsync() the directory. And do not report an error if the open fails, since some systems (Ex: AIX and a chromium sandbox) are unable to open and fsync directories. |
---|---|
Downloads: | Tarball | ZIP archive | SQL archive |
Timelines: | family | ancestors | descendants | both | trunk |
Files: | files | file ages | folders |
SHA1: |
713b1b7dc1296e9cee42aeaad8c85528 |
User & Date: | drh 2011-08-08 23:48:40 |
Context
2011-08-09
| ||
18:14 | Add command-line utilities "offsets.c" and "extract.c" for use in low-level analyzsis of database files. (check-in: dfa22ed4 user: drh tags: trunk) | |
2011-08-08
| ||
23:48 | In os_unix.c, do not open the directory containing the database file when the database file is opened. Instead, wait until time to fsync() the directory. And do not report an error if the open fails, since some systems (Ex: AIX and a chromium sandbox) are unable to open and fsync directories. (check-in: 713b1b7d user: drh tags: trunk) | |
23:18 | Allow the unlink() system call to be overridden in os_unix.c. (check-in: 8d1b5c3a user: drh tags: trunk) | |
Changes
Changes to src/os_unix.c.
︙ | ︙ | |||
204 205 206 207 208 209 210 | ** VFS implementations. */ typedef struct unixFile unixFile; struct unixFile { sqlite3_io_methods const *pMethod; /* Always the first entry */ unixInodeInfo *pInode; /* Info about locks on this inode */ int h; /* The file descriptor */ | < | 204 205 206 207 208 209 210 211 212 213 214 215 216 217 | ** VFS implementations. */ typedef struct unixFile unixFile; struct unixFile { sqlite3_io_methods const *pMethod; /* Always the first entry */ unixInodeInfo *pInode; /* Info about locks on this inode */ int h; /* The file descriptor */ unsigned char eFileLock; /* The type of lock held on this fd */ unsigned char ctrlFlags; /* Behavioral bits. UNIXFILE_* flags */ int lastErrno; /* The unix errno from last I/O error */ void *lockingContext; /* Locking style specific state */ UnixUnusedFd *pUnused; /* Pre-allocated UnixUnusedFd */ const char *zPath; /* Name of the file */ unixShm *pShm; /* Shared memory segment information */ |
︙ | ︙ | |||
249 250 251 252 253 254 255 256 257 258 259 260 261 262 | /* ** Allowed values for the unixFile.ctrlFlags bitmask: */ #define UNIXFILE_EXCL 0x01 /* Connections from one process only */ #define UNIXFILE_RDONLY 0x02 /* Connection is read only */ #define UNIXFILE_PERSIST_WAL 0x04 /* Persistent WAL mode */ /* ** Include code that is common to all os_*.c files */ #include "os_common.h" /* | > | 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 | /* ** Allowed values for the unixFile.ctrlFlags bitmask: */ #define UNIXFILE_EXCL 0x01 /* Connections from one process only */ #define UNIXFILE_RDONLY 0x02 /* Connection is read only */ #define UNIXFILE_PERSIST_WAL 0x04 /* Persistent WAL mode */ #define UNIXFILE_DIRSYNC 0x08 /* Directory sync needed */ /* ** Include code that is common to all os_*.c files */ #include "os_common.h" /* |
︙ | ︙ | |||
1749 1750 1751 1752 1753 1754 1755 | ** ** It is *not* necessary to hold the mutex when this routine is called, ** even on VxWorks. A mutex will be acquired on VxWorks by the ** vxworksReleaseFileId() routine. */ static int closeUnixFile(sqlite3_file *id){ unixFile *pFile = (unixFile*)id; | < < < < | 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 | ** ** It is *not* necessary to hold the mutex when this routine is called, ** even on VxWorks. A mutex will be acquired on VxWorks by the ** vxworksReleaseFileId() routine. */ static int closeUnixFile(sqlite3_file *id){ unixFile *pFile = (unixFile*)id; if( pFile->h>=0 ){ robust_close(pFile, pFile->h, __LINE__); pFile->h = -1; } #if OS_VXWORKS if( pFile->pId ){ if( pFile->isDelete ){ |
︙ | ︙ | |||
3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 | #endif /* ifdef SQLITE_NO_SYNC elif HAVE_FULLFSYNC */ if( OS_VXWORKS && rc!= -1 ){ rc = 0; } return rc; } /* ** Make sure all writes to a particular file are committed to disk. ** ** If dataOnly==0 then both the file itself and its metadata (file ** size, access time, etc) are synced. If dataOnly!=0 then only the ** file data is synced. | > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > | 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284 | #endif /* ifdef SQLITE_NO_SYNC elif HAVE_FULLFSYNC */ if( OS_VXWORKS && rc!= -1 ){ rc = 0; } return rc; } /* ** Open a file descriptor to the directory containing file zFilename. ** If successful, *pFd is set to the opened file descriptor and ** SQLITE_OK is returned. If an error occurs, either SQLITE_NOMEM ** or SQLITE_CANTOPEN is returned and *pFd is set to an undefined ** value. ** ** If SQLITE_OK is returned, the caller is responsible for closing ** the file descriptor *pFd using close(). */ static int openDirectory(const char *zFilename, int *pFd){ int ii; int fd = -1; char zDirname[MAX_PATHNAME+1]; sqlite3_snprintf(MAX_PATHNAME, zDirname, "%s", zFilename); for(ii=(int)strlen(zDirname); ii>1 && zDirname[ii]!='/'; ii--); if( ii>0 ){ zDirname[ii] = '\0'; fd = robust_open(zDirname, O_RDONLY|O_BINARY, 0); if( fd>=0 ){ #ifdef FD_CLOEXEC osFcntl(fd, F_SETFD, osFcntl(fd, F_GETFD, 0) | FD_CLOEXEC); #endif OSTRACE(("OPENDIR %-3d %s\n", fd, zDirname)); } } *pFd = fd; return (fd>=0?SQLITE_OK:unixLogError(SQLITE_CANTOPEN_BKPT, "open", zDirname)); } /* ** Make sure all writes to a particular file are committed to disk. ** ** If dataOnly==0 then both the file itself and its metadata (file ** size, access time, etc) are synced. If dataOnly!=0 then only the ** file data is synced. |
︙ | ︙ | |||
3285 3286 3287 3288 3289 3290 3291 | OSTRACE(("SYNC %-3d\n", pFile->h)); rc = full_fsync(pFile->h, isFullsync, isDataOnly); SimulateIOError( rc=1 ); if( rc ){ pFile->lastErrno = errno; return unixLogError(SQLITE_IOERR_FSYNC, "full_fsync", pFile->zPath); } | > > > > > > | > | < | < | < | < < < < < < | < < < < | | 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324 3325 3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 | OSTRACE(("SYNC %-3d\n", pFile->h)); rc = full_fsync(pFile->h, isFullsync, isDataOnly); SimulateIOError( rc=1 ); if( rc ){ pFile->lastErrno = errno; return unixLogError(SQLITE_IOERR_FSYNC, "full_fsync", pFile->zPath); } /* Also fsync the directory containing the file if the DIRSYNC flag ** is set. This is a one-time occurrance. Many systems (examples: AIX ** or any process running inside a chromium sandbox) are unable to fsync a ** directory, so ignore errors. */ if( pFile->ctrlFlags & UNIXFILE_DIRSYNC ){ int dirfd; OSTRACE(("DIRSYNC %s (have_fullfsync=%d fullsync=%d)\n", pFile->zPath, HAVE_FULLFSYNC, isFullsync)); openDirectory(pFile->zPath, &dirfd); if( dirfd>=0 ){ full_fsync(dirfd, 0, 0); robust_close(pFile, dirfd, __LINE__); } pFile->ctrlFlags &= ~UNIXFILE_DIRSYNC; } return rc; } /* ** Truncate an open file to a specified size */ |
︙ | ︙ | |||
4474 4475 4476 4477 4478 4479 4480 | /* ** Initialize the contents of the unixFile structure pointed to by pId. */ static int fillInUnixFile( sqlite3_vfs *pVfs, /* Pointer to vfs object */ int h, /* Open file descriptor of file being opened */ | | | 4495 4496 4497 4498 4499 4500 4501 4502 4503 4504 4505 4506 4507 4508 4509 | /* ** Initialize the contents of the unixFile structure pointed to by pId. */ static int fillInUnixFile( sqlite3_vfs *pVfs, /* Pointer to vfs object */ int h, /* Open file descriptor of file being opened */ int syncDir, /* True to sync directory on first sync */ sqlite3_file *pId, /* Write to the unixFile structure here */ const char *zFilename, /* Name of the file being opened */ int noLock, /* Omit locking if true */ int isDelete, /* Delete on close if true */ int isReadOnly /* True if the file is opened read-only */ ){ const sqlite3_io_methods *pLockingStyle; |
︙ | ︙ | |||
4505 4506 4507 4508 4509 4510 4511 | || pVfs->pAppData==(void*)&autolockIoFinder ); #else assert( zFilename==0 || zFilename[0]=='/' ); #endif OSTRACE(("OPEN %-3d %s\n", h, zFilename)); pNew->h = h; | < > > > | 4526 4527 4528 4529 4530 4531 4532 4533 4534 4535 4536 4537 4538 4539 4540 4541 4542 4543 4544 4545 4546 4547 4548 4549 4550 4551 | || pVfs->pAppData==(void*)&autolockIoFinder ); #else assert( zFilename==0 || zFilename[0]=='/' ); #endif OSTRACE(("OPEN %-3d %s\n", h, zFilename)); pNew->h = h; pNew->zPath = zFilename; if( memcmp(pVfs->zName,"unix-excl",10)==0 ){ pNew->ctrlFlags = UNIXFILE_EXCL; }else{ pNew->ctrlFlags = 0; } if( isReadOnly ){ pNew->ctrlFlags |= UNIXFILE_RDONLY; } if( syncDir ){ pNew->ctrlFlags |= UNIXFILE_DIRSYNC; } #if OS_VXWORKS pNew->pId = vxworksFindFileId(zFilename); if( pNew->pId==0 ){ noLock = 1; rc = SQLITE_NOMEM; } |
︙ | ︙ | |||
4647 4648 4649 4650 4651 4652 4653 | h = -1; osUnlink(zFilename); isDelete = 0; } pNew->isDelete = isDelete; #endif if( rc!=SQLITE_OK ){ | < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < < | 4670 4671 4672 4673 4674 4675 4676 4677 4678 4679 4680 4681 4682 4683 4684 4685 4686 4687 4688 4689 4690 4691 | h = -1; osUnlink(zFilename); isDelete = 0; } pNew->isDelete = isDelete; #endif if( rc!=SQLITE_OK ){ if( h>=0 ) robust_close(pNew, h, __LINE__); }else{ pNew->pMethod = pLockingStyle; OpenCounter(+1); } return rc; } /* ** Return the name of a directory in which to put temporary files. ** If no suitable temporary file directory can be found, return NULL. */ static const char *unixTempFileDir(void){ static const char *azDirs[] = { 0, |
︙ | ︙ | |||
4919 4920 4921 4922 4923 4924 4925 | const char *zPath, /* Pathname of file to be opened */ sqlite3_file *pFile, /* The file descriptor to be filled in */ int flags, /* Input flags to control the opening */ int *pOutFlags /* Output flags returned to SQLite core */ ){ unixFile *p = (unixFile *)pFile; int fd = -1; /* File descriptor returned by open() */ | < | | 4910 4911 4912 4913 4914 4915 4916 4917 4918 4919 4920 4921 4922 4923 4924 4925 4926 4927 4928 4929 4930 4931 4932 4933 4934 4935 4936 4937 4938 4939 4940 4941 4942 | const char *zPath, /* Pathname of file to be opened */ sqlite3_file *pFile, /* The file descriptor to be filled in */ int flags, /* Input flags to control the opening */ int *pOutFlags /* Output flags returned to SQLite core */ ){ unixFile *p = (unixFile *)pFile; int fd = -1; /* File descriptor returned by open() */ int openFlags = 0; /* Flags to pass to open() */ int eType = flags&0xFFFFFF00; /* Type of file to open */ int noLock; /* True to omit locking primitives */ int rc = SQLITE_OK; /* Function Return Code */ int isExclusive = (flags & SQLITE_OPEN_EXCLUSIVE); int isDelete = (flags & SQLITE_OPEN_DELETEONCLOSE); int isCreate = (flags & SQLITE_OPEN_CREATE); int isReadonly = (flags & SQLITE_OPEN_READONLY); int isReadWrite = (flags & SQLITE_OPEN_READWRITE); #if SQLITE_ENABLE_LOCKING_STYLE int isAutoProxy = (flags & SQLITE_OPEN_AUTOPROXY); #endif /* If creating a master or main-file journal, this function will open ** a file-descriptor on the directory too. The first time unixSync() ** is called the directory file descriptor will be fsync()ed and close()d. */ int syncDir = (isCreate && ( eType==SQLITE_OPEN_MASTER_JOURNAL || eType==SQLITE_OPEN_MAIN_JOURNAL || eType==SQLITE_OPEN_WAL )); /* If argument zPath is a NULL pointer, this function is required to open ** a temporary file. Use this buffer to store the file name in. |
︙ | ︙ | |||
4992 4993 4994 4995 4996 4997 4998 | if( !pUnused ){ return SQLITE_NOMEM; } } p->pUnused = pUnused; }else if( !zName ){ /* If zName is NULL, the upper layer is requesting a temp file. */ | | | 4982 4983 4984 4985 4986 4987 4988 4989 4990 4991 4992 4993 4994 4995 4996 | if( !pUnused ){ return SQLITE_NOMEM; } } p->pUnused = pUnused; }else if( !zName ){ /* If zName is NULL, the upper layer is requesting a temp file. */ assert(isDelete && !syncDir); rc = unixGetTempname(MAX_PATHNAME+1, zTmpname); if( rc!=SQLITE_OK ){ return rc; } zName = zTmpname; } |
︙ | ︙ | |||
5057 5058 5059 5060 5061 5062 5063 | } #if SQLITE_ENABLE_LOCKING_STYLE else{ p->openFlags = openFlags; } #endif | < < < < < < < < < < < < < < | 5047 5048 5049 5050 5051 5052 5053 5054 5055 5056 5057 5058 5059 5060 5061 5062 5063 5064 5065 5066 5067 5068 5069 5070 5071 | } #if SQLITE_ENABLE_LOCKING_STYLE else{ p->openFlags = openFlags; } #endif #ifdef FD_CLOEXEC osFcntl(fd, F_SETFD, osFcntl(fd, F_GETFD, 0) | FD_CLOEXEC); #endif noLock = eType!=SQLITE_OPEN_MAIN_DB; #if defined(__APPLE__) || SQLITE_ENABLE_LOCKING_STYLE struct statfs fsInfo; if( fstatfs(fd, &fsInfo) == -1 ){ ((unixFile*)pFile)->lastErrno = errno; robust_close(p, fd, __LINE__); return SQLITE_IOERR_ACCESS; } if (0 == strncmp("msdos", fsInfo.f_fstypename, 5)) { ((unixFile*)pFile)->fsFlags |= SQLITE_FSFLAGS_IS_MSDOS; } #endif |
︙ | ︙ | |||
5113 5114 5115 5116 5117 5118 5119 | ** with fd is a database file, and there are other connections open ** on that file that are currently holding advisory locks on it, ** then the call to close() will cancel those locks. In practice, ** we're assuming that statfs() doesn't fail very often. At least ** not while other file descriptors opened by the same process on ** the same file are working. */ p->lastErrno = errno; | < < < | | | 5089 5090 5091 5092 5093 5094 5095 5096 5097 5098 5099 5100 5101 5102 5103 5104 5105 5106 5107 5108 5109 5110 5111 5112 5113 5114 5115 5116 5117 5118 5119 5120 5121 5122 5123 5124 5125 5126 5127 5128 | ** with fd is a database file, and there are other connections open ** on that file that are currently holding advisory locks on it, ** then the call to close() will cancel those locks. In practice, ** we're assuming that statfs() doesn't fail very often. At least ** not while other file descriptors opened by the same process on ** the same file are working. */ p->lastErrno = errno; robust_close(p, fd, __LINE__); rc = SQLITE_IOERR_ACCESS; goto open_finished; } useProxy = !(fsInfo.f_flags&MNT_LOCAL); } if( useProxy ){ rc = fillInUnixFile(pVfs, fd, syncDir, pFile, zPath, noLock, isDelete, isReadonly); if( rc==SQLITE_OK ){ rc = proxyTransformUnixFile((unixFile*)pFile, ":auto:"); if( rc!=SQLITE_OK ){ /* Use unixClose to clean up the resources added in fillInUnixFile ** and clear all the structure's references. Specifically, ** pFile->pMethods will be NULL so sqlite3OsClose will be a no-op */ unixClose(pFile); return rc; } } goto open_finished; } } #endif rc = fillInUnixFile(pVfs, fd, syncDir, pFile, zPath, noLock, isDelete, isReadonly); open_finished: if( rc!=SQLITE_OK ){ sqlite3_free(p->pUnused); } return rc; } |
︙ | ︙ | |||
5741 5742 5743 5744 5745 5746 5747 | */ static int proxyCreateUnixFile( const char *path, /* path for the new unixFile */ unixFile **ppFile, /* unixFile created and returned by ref */ int islockfile /* if non zero missing dirs will be created */ ) { int fd = -1; | < | 5714 5715 5716 5717 5718 5719 5720 5721 5722 5723 5724 5725 5726 5727 | */ static int proxyCreateUnixFile( const char *path, /* path for the new unixFile */ unixFile **ppFile, /* unixFile created and returned by ref */ int islockfile /* if non zero missing dirs will be created */ ) { int fd = -1; unixFile *pNew; int rc = SQLITE_OK; int openFlags = O_RDWR | O_CREAT; sqlite3_vfs dummyVfs; int terrno = 0; UnixUnusedFd *pUnused = NULL; |
︙ | ︙ | |||
5806 5807 5808 5809 5810 5811 5812 | memset(&dummyVfs, 0, sizeof(dummyVfs)); dummyVfs.pAppData = (void*)&autolockIoFinder; dummyVfs.zName = "dummy"; pUnused->fd = fd; pUnused->flags = openFlags; pNew->pUnused = pUnused; | | | 5778 5779 5780 5781 5782 5783 5784 5785 5786 5787 5788 5789 5790 5791 5792 | memset(&dummyVfs, 0, sizeof(dummyVfs)); dummyVfs.pAppData = (void*)&autolockIoFinder; dummyVfs.zName = "dummy"; pUnused->fd = fd; pUnused->flags = openFlags; pNew->pUnused = pUnused; rc = fillInUnixFile(&dummyVfs, fd, 0, (sqlite3_file*)pNew, path, 0, 0, 0); if( rc==SQLITE_OK ){ *ppFile = pNew; return SQLITE_OK; } end_create_proxy: robust_close(pNew, fd, __LINE__); sqlite3_free(pNew); |
︙ | ︙ |