os.c

SQLite 2.8.6 源代码,用来在Linux/Unix/Windows上编译安装.它是一个小型的数据库,但是非常好用,速度也快,一般的数据库查询之类的操作据统计比MySQL,PostgreSQL

/*
** 2001 September 16
**
** The author disclaims copyright to this source code.  In place of
** a legal notice, here is a blessing:
**
**    May you do good and not evil.
**    May you find forgiveness for yourself and forgive others.
**    May you share freely, never taking more than you give.
**
******************************************************************************
**
** This file contains code that is specific to particular operating
** systems.  The purpose of this file is to provide a uniform abstraction
** on which the rest of SQLite can operate.
*/
#include "os.h"          /* Must be first to enable large file support */
#include "sqliteInt.h"

#if OS_UNIX
# include <time.h>
# include <errno.h>
# include <unistd.h>
# ifndef O_LARGEFILE
#  define O_LARGEFILE 0
# endif
# ifdef SQLITE_DISABLE_LFS
#  undef O_LARGEFILE
#  define O_LARGEFILE 0
# endif
# ifndef O_NOFOLLOW
#  define O_NOFOLLOW 0
# endif
# ifndef O_BINARY
#  define O_BINARY 0
# endif
#endif


#if OS_WIN
# include <winbase.h>
#endif

#if OS_MAC
# include <extras.h>
# include <path2fss.h>
# include <TextUtils.h>
# include <FinderRegistry.h>
# include <Folders.h>
# include <Timer.h>
# include <OSUtils.h>
#endif

/*
** The DJGPP compiler environment looks mostly like Unix, but it
** lacks the fcntl() system call.  So redefine fcntl() to be something
** that always succeeds.  This means that locking does not occur under
** DJGPP.  But its DOS - what did you expect?
*/
#ifdef __DJGPP__
# define fcntl(A,B,C) 0
#endif

/*
** Macros for performance tracing.  Normally turned off
*/
#if 0
static int last_page = 0;
__inline__ unsigned long long int hwtime(void){
  unsigned long long int x;
  __asm__("rdtsc\n\t"
          "mov %%edx, %%ecx\n\t"
          :"=A" (x));
  return x;
}
static unsigned long long int g_start;
static unsigned int elapse;
#define TIMER_START       g_start=hwtime()
#define TIMER_END         elapse=hwtime()-g_start
#define SEEK(X)           last_page=(X)
#define TRACE1(X)         fprintf(stderr,X)
#define TRACE2(X,Y)       fprintf(stderr,X,Y)
#define TRACE3(X,Y,Z)     fprintf(stderr,X,Y,Z)
#define TRACE4(X,Y,Z,A)   fprintf(stderr,X,Y,Z,A)
#define TRACE5(X,Y,Z,A,B) fprintf(stderr,X,Y,Z,A,B)
#else
#define TIMER_START
#define TIMER_END
#define SEEK(X)
#define TRACE1(X)
#define TRACE2(X,Y)
#define TRACE3(X,Y,Z)
#define TRACE4(X,Y,Z,A)
#define TRACE5(X,Y,Z,A,B)
#endif


#if OS_UNIX
/*
** Here is the dirt on POSIX advisory locks:  ANSI STD 1003.1 (1996)
** section 6.5.2.2 lines 483 through 490 specify that when a process
** sets or clears a lock, that operation overrides any prior locks set
** by the same process.  It does not explicitly say so, but this implies
** that it overrides locks set by the same process using a different
** file descriptor.  Consider this test case:
**
**       int fd1 = open("./file1", O_RDWR|O_CREAT, 0644);
**       int fd2 = open("./file2", O_RDWR|O_CREAT, 0644);
**
** Suppose ./file1 and ./file2 are really the same file (because
** one is a hard or symbolic link to the other) then if you set
** an exclusive lock on fd1, then try to get an exclusive lock
** on fd2, it works.  I would have expected the second lock to
** fail since there was already a lock on the file due to fd1.
** But not so.  Since both locks came from the same process, the
** second overrides the first, even though they were on different
** file descriptors opened on different file names.
**
** Bummer.  If you ask me, this is broken.  Badly broken.  It means
** that we cannot use POSIX locks to synchronize file access among
** competing threads of the same process.  POSIX locks will work fine
** to synchronize access for threads in separate processes, but not
** threads within the same process.
**
** To work around the problem, SQLite has to manage file locks internally
** on its own.  Whenever a new database is opened, we have to find the
** specific inode of the database file (the inode is determined by the
** st_dev and st_ino fields of the stat structure that fstat() fills in)
** and check for locks already existing on that inode.  When locks are
** created or removed, we have to look at our own internal record of the
** locks to see if another thread has previously set a lock on that same
** inode.
**
** The OsFile structure for POSIX is no longer just an integer file
** descriptor.  It is now a structure that holds the integer file
** descriptor and a pointer to a structure that describes the internal
** locks on the corresponding inode.  There is one locking structure
** per inode, so if the same inode is opened twice, both OsFile structures
** point to the same locking structure.  The locking structure keeps
** a reference count (so we will know when to delete it) and a "cnt"
** field that tells us its internal lock status.  cnt==0 means the
** 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. 
*/
struct inodeKey {
  dev_t dev;   /* Device number */
  ino_t ino;   /* Inode number */
};

/*
** An instance of the following structure is allocated for each inode.
** 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 inodeKey key;  /* The lookup key */
  int cnt;              /* 0: unlocked.  -1: write lock.  1...: read lock. */
  int nRef;             /* Number of pointers to this structure */
};

/* 
** This hash table maps inodes (in the form of inodeKey structures) into
** pointers to lockInfo structures.
*/
static Hash lockHash = { SQLITE_HASH_BINARY, 0, 0, 0, 0, 0 };

/*
** Given a file descriptor, locate a lockInfo structure that describes
** that file descriptor.  Create a new one if necessary.  NULL might
** be returned if malloc() fails.
*/
static struct lockInfo *findLockInfo(int fd){
  int rc;
  struct inodeKey key;
  struct stat statbuf;
  struct lockInfo *pInfo;
  rc = fstat(fd, &statbuf);
  if( rc!=0 ) return 0;
  memset(&key, 0, sizeof(key));
  key.dev = statbuf.st_dev;
  key.ino = statbuf.st_ino;
  pInfo = (struct lockInfo*)sqliteHashFind(&lockHash, &key, sizeof(key));
  if( pInfo==0 ){
    struct lockInfo *pOld;
    pInfo = sqliteMalloc( sizeof(*pInfo) );
    if( pInfo==0 ) return 0;
    pInfo->key = key;
    pInfo->nRef = 1;
    pInfo->cnt = 0;
    pOld = sqliteHashInsert(&lockHash, &pInfo->key, sizeof(key), pInfo);
    if( pOld!=0 ){
      assert( pOld==pInfo );
      sqliteFree(pInfo);
      pInfo = 0;
    }
  }else{
    pInfo->nRef++;
  }
  return pInfo;
}

/*
** Release a lockInfo structure previously allocated by findLockInfo().
*/
static void releaseLockInfo(struct lockInfo *pInfo){
  pInfo->nRef--;
  if( pInfo->nRef==0 ){
    sqliteHashInsert(&lockHash, &pInfo->key, sizeof(pInfo->key), 0);
    sqliteFree(pInfo);
  }
}
#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.
*/
#ifdef SQLITE_TEST
int sqlite_io_error_pending = 0;
#define SimulateIOError(A)  \
   if( sqlite_io_error_pending ) \
     if( sqlite_io_error_pending-- == 1 ){ local_ioerr(); return A; }
static void local_ioerr(){
  sqlite_io_error_pending = 0;  /* Really just a place to set a breakpoint */
}
#else
#define SimulateIOError(A)
#endif

/*
** When testing, keep a count of the number of open files.
*/
#ifdef SQLITE_TEST
int sqlite_open_file_count = 0;
#define OpenCounter(X)  sqlite_open_file_count+=(X)
#else
#define OpenCounter(X)
#endif


/*
** Delete the named file
*/
int sqliteOsDelete(const char *zFilename){
#if OS_UNIX
  unlink(zFilename);
#endif
#if OS_WIN
  DeleteFile(zFilename);
#endif
#if OS_MAC
  unlink(zFilename);
#endif
  return SQLITE_OK;
}

/*
** Return TRUE if the named file exists.
*/
int sqliteOsFileExists(const char *zFilename){
#if OS_UNIX
  return access(zFilename, 0)==0;
#endif
#if OS_WIN
  return GetFileAttributes(zFilename) != 0xffffffff;
#endif
#if OS_MAC
  return access(zFilename, 0)==0;
#endif
}


#if 0 /* NOT USED */
/*
** Change the name of an existing file.
*/
int sqliteOsFileRename(const char *zOldName, const char *zNewName){
#if OS_UNIX
  if( link(zOldName, zNewName) ){
    return SQLITE_ERROR;
  }
  unlink(zOldName);
  return SQLITE_OK;
#endif
#if OS_WIN
  if( !MoveFile(zOldName, zNewName) ){
    return SQLITE_ERROR;
  }
  return SQLITE_OK;
#endif
#if OS_MAC
  /**** FIX ME ***/
  return SQLITE_ERROR;
#endif
}
#endif /* NOT USED */

/*
** Attempt to open a file for both reading and writing.  If that
** fails, try opening it read-only.  If the file does not exist,
** try to create it.
**
** On success, a handle for the open file is written to *id
** and *pReadonly is set to 0 if the file was opened for reading and
** writing or 1 if the file was opened read-only.  The function returns
** SQLITE_OK.
**
** On failure, the function returns SQLITE_CANTOPEN and leaves
** *id and *pReadonly unchanged.
*/
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();
  id->pLock = findLockInfo(id->fd);
  sqliteOsLeaveMutex();
  if( id->pLock==0 ){
    close(id->fd);
    return SQLITE_NOMEM;
  }
  id->locked = 0;
  TRACE3("OPEN    %-3d %s\n", id->fd, zFilename);
  OpenCounter(+1);
  return SQLITE_OK;
#endif
#if OS_WIN
  HANDLE h = CreateFile(zFilename,
     GENERIC_READ | GENERIC_WRITE,
     FILE_SHARE_READ | FILE_SHARE_WRITE,
     NULL,
     OPEN_ALWAYS,
     FILE_ATTRIBUTE_NORMAL | FILE_FLAG_RANDOM_ACCESS,
     NULL
  );
  if( h==INVALID_HANDLE_VALUE ){
    h = CreateFile(zFilename,
       GENERIC_READ,
       FILE_SHARE_READ,
       NULL,
       OPEN_ALWAYS,
       FILE_ATTRIBUTE_NORMAL | FILE_FLAG_RANDOM_ACCESS,
       NULL
    );
    if( h==INVALID_HANDLE_VALUE ){
      return SQLITE_CANTOPEN;
    }
    *pReadonly = 1;
  }else{
    *pReadonly = 0;
  }
  id->h = h;
  id->locked = 0;
  OpenCounter(+1);
  return SQLITE_OK;
#endif
#if OS_MAC
  FSSpec fsSpec;
# ifdef _LARGE_FILE
  HFSUniStr255 dfName;
  FSRef fsRef;
  if( __path2fss(zFilename, &fsSpec) != noErr ){
    if( HCreate(fsSpec.vRefNum, fsSpec.parID, fsSpec.name, 'SQLI', cDocumentFile) != noErr )
      return SQLITE_CANTOPEN;
  }
  if( FSpMakeFSRef(&fsSpec, &fsRef) != noErr )
    return SQLITE_CANTOPEN;
  FSGetDataForkName(&dfName);
  if( FSOpenFork(&fsRef, dfName.length, dfName.unicode,
                 fsRdWrShPerm, &(id->refNum)) != noErr ){
    if( FSOpenFork(&fsRef, dfName.length, dfName.unicode,
                   fsRdWrPerm, &(id->refNum)) != noErr ){
      if (FSOpenFork(&fsRef, dfName.length, dfName.unicode,
                   fsRdPerm, &(id->refNum)) != noErr )
        return SQLITE_CANTOPEN;
      else
        *pReadonly = 1;
    } else
      *pReadonly = 0;
  } else
    *pReadonly = 0;
# else
  __path2fss(zFilename, &fsSpec);
  if( !sqliteOsFileExists(zFilename) ){
    if( HCreate(fsSpec.vRefNum, fsSpec.parID, fsSpec.name, 'SQLI', cDocumentFile) != noErr )
      return SQLITE_CANTOPEN;
  }
  if( HOpenDF(fsSpec.vRefNum, fsSpec.parID, fsSpec.name, fsRdWrShPerm, &(id->refNum)) != noErr ){
    if( HOpenDF(fsSpec.vRefNum, fsSpec.parID, fsSpec.name, fsRdWrPerm, &(id->refNum)) != noErr ){
      if( HOpenDF(fsSpec.vRefNum, fsSpec.parID, fsSpec.name, fsRdPerm, &(id->refNum)) != noErr )
        return SQLITE_CANTOPEN;
      else
        *pReadonly = 1;
    } else
      *pReadonly = 0;
  } else
    *pReadonly = 0;
# endif
  if( HOpenRF(fsSpec.vRefNum, fsSpec.parID, fsSpec.name, fsRdWrShPerm, &(id->refNumRF)) != noErr){
    id->refNumRF = -1;
  }
  id->locked = 0;
  id->delOnClose = 0;
  OpenCounter(+1);
  return SQLITE_OK;
#endif
}


/*
** Attempt to open a new file for exclusive access by this process.
** The file will be opened for both reading and writing.  To avoid
** a potential security problem, we do not allow the file to have
** previously existed.  Nor do we allow the file to be a symbolic
** link.
**
** If delFlag is true, then make arrangements to automatically delete
** the file when it is closed.
**
** 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();
  id->pLock = findLockInfo(id->fd);
  sqliteOsLeaveMutex();
  if( id->pLock==0 ){
    close(id->fd);
    unlink(zFilename);
    return SQLITE_NOMEM;
  }
  id->locked = 0;
  if( delFlag ){
    unlink(zFilename);
  }
  TRACE3("OPEN-EX %-3d %s\n", id->fd, zFilename);
  OpenCounter(+1);
  return SQLITE_OK;
#endif
#if OS_WIN
  HANDLE h;
  int fileflags;
  if( delFlag ){
    fileflags = FILE_ATTRIBUTE_TEMPORARY | FILE_FLAG_RANDOM_ACCESS 
                     | FILE_FLAG_DELETE_ON_CLOSE;
  }else{
    fileflags = FILE_FLAG_RANDOM_ACCESS;
  }
  h = CreateFile(zFilename,
     GENERIC_READ | GENERIC_WRITE,
     0,
     NULL,
     CREATE_ALWAYS,
     fileflags,
     NULL
  );
  if( h==INVALID_HANDLE_VALUE ){
    return SQLITE_CANTOPEN;
  }
  id->h = h;
  id->locked = 0;
  OpenCounter(+1);
  return SQLITE_OK;
#endif
#if OS_MAC
  FSSpec fsSpec;
# ifdef _LARGE_FILE
  HFSUniStr255 dfName;
  FSRef fsRef;
  __path2fss(zFilename, &fsSpec);
  if( HCreate(fsSpec.vRefNum, fsSpec.parID, fsSpec.name, 'SQLI', cDocumentFile) != noErr )
    return SQLITE_CANTOPEN;
  if( FSpMakeFSRef(&fsSpec, &fsRef) != noErr )
    return SQLITE_CANTOPEN;
  FSGetDataForkName(&dfName);
  if( FSOpenFork(&fsRef, dfName.length, dfName.unicode,
                 fsRdWrPerm, &(id->refNum)) != noErr )
    return SQLITE_CANTOPEN;
# else
  __path2fss(zFilename, &fsSpec);
  if( HCreate(fsSpec.vRefNum, fsSpec.parID, fsSpec.name, 'SQLI', cDocumentFile) != noErr )
    return SQLITE_CANTOPEN;
  if( HOpenDF(fsSpec.vRefNum, fsSpec.parID, fsSpec.name, fsRdWrPerm, &(id->refNum)) != noErr )
    return SQLITE_CANTOPEN;
# endif
  id->refNumRF = -1;
  id->locked = 0;
  id->delOnClose = delFlag;
  if (delFlag)
    id->pathToDel = sqliteOsFullPathname(zFilename);
  OpenCounter(+1);
  return SQLITE_OK;
#endif
}

/*
** Attempt to open a new file for read-only access.
**
** 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 ){