SQLite

  sqlite3JournalOpen(0, 0, pJfd, 0, -1);
}

#ifdef SQLITE_ENABLE_ATOMIC_WRITE
/*
** If the argument p points to a MemJournal structure that is not an 
** in-memory-only journal file (i.e. is one that was opened with a +ve
** nSpill parameter or as SQLITE_OPEN_MAIN_JOURNAL), and the underlying 
** file has not yet been created, create it now.
*/
int sqlite3JournalCreate(sqlite3_file *pJfd){
  int rc = SQLITE_OK;
  MemJournal *p = (MemJournal*)pJfd;
  if( p->pMethod==&MemJournalMethods 
   && (p->nSpill>0 || (p->flags & SQLITE_OPEN_MAIN_JOURNAL))
  ){
    rc = memjrnlCreateFile(p);
  }
  return rc;
}
#endif

/*
** The file-handle passed as the only argument is open on a journal file.

/*
** standard include files.
*/
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <sys/ioctl.h>
#include <unistd.h>
#include <time.h>
#include <sys/time.h>
#include <errno.h>
#if !defined(SQLITE_OMIT_WAL) || SQLITE_MAX_MMAP_SIZE>0
# include <sys/mman.h>
#endif

#if SQLITE_MAX_MMAP_SIZE>0
  int nFetchOut;                      /* Number of outstanding xFetch refs */
  sqlite3_int64 mmapSize;             /* Usable size of mapping at pMapRegion */
  sqlite3_int64 mmapSizeActual;       /* Actual size of mapping at pMapRegion */
  sqlite3_int64 mmapSizeMax;          /* Configured FCNTL_MMAP_SIZE value */
  void *pMapRegion;                   /* Memory mapped region */
#endif

  int sectorSize;                     /* Device sector size */
  int deviceCharacteristics;          /* Precomputed device characteristics */

#if SQLITE_ENABLE_LOCKING_STYLE
  int openFlags;                      /* The flags specified at open() */
#endif
#if SQLITE_ENABLE_LOCKING_STYLE || defined(__APPLE__)
  unsigned fsFlags;                   /* cached details from statfs() */
#endif
#if OS_VXWORKS

/*
** Explicitly call the 64-bit version of lseek() on Android. Otherwise, lseek()
** is the 32-bit version, even if _FILE_OFFSET_BITS=64 is defined.
*/
#ifdef __ANDROID__
# define lseek lseek64
#endif

#define F2FS_IOCTL_MAGIC        0xf5
#define F2FS_IOC_START_ATOMIC_WRITE     _IO(F2FS_IOCTL_MAGIC, 1)
#define F2FS_IOC_COMMIT_ATOMIC_WRITE    _IO(F2FS_IOCTL_MAGIC, 2)
#define F2FS_IOC_START_VOLATILE_WRITE   _IO(F2FS_IOCTL_MAGIC, 3)
#define F2FS_IOC_ABORT_VOLATILE_WRITE   _IO(F2FS_IOCTL_MAGIC, 5)


/*
** Different Unix systems declare open() in different ways.  Same use
** open(const char*,int,mode_t).  Others use open(const char*,int,...).
** The difference is important when using a pointer to the function.
**
** The safest way to deal with the problem is to always use this wrapper

#if defined(HAVE_LSTAT)
  { "lstat",         (sqlite3_syscall_ptr)lstat,          0 },
#else
  { "lstat",         (sqlite3_syscall_ptr)0,              0 },
#endif
#define osLstat      ((int(*)(const char*,struct stat*))aSyscall[27].pCurrent)

  { "ioctl",         (sqlite3_syscall_ptr)ioctl,          0 },
#define osIoctl ((int(*)(int,int))aSyscall[28].pCurrent)

}; /* End of the overrideable system calls */


/*
** On some systems, calls to fchown() will trigger a message in a security
** log if they come from non-root processes.  So avoid calling fchown() if

/*
** Information and control of an open file handle.
*/
static int unixFileControl(sqlite3_file *id, int op, void *pArg){
  unixFile *pFile = (unixFile*)id;
  switch( op ){
    case SQLITE_FCNTL_BEGIN_ATOMIC_WRITE: {
      int rc = osIoctl(pFile->h, F2FS_IOC_START_ATOMIC_WRITE);
      return rc ? SQLITE_ERROR : SQLITE_OK;
    }
    case SQLITE_FCNTL_COMMIT_ATOMIC_WRITE: {
      int rc = osIoctl(pFile->h, F2FS_IOC_COMMIT_ATOMIC_WRITE);
      return rc ? SQLITE_ERROR : SQLITE_OK;
    }
    case SQLITE_FCNTL_ROLLBACK_ATOMIC_WRITE: {
      int rc = osIoctl(pFile->h, F2FS_IOC_ABORT_VOLATILE_WRITE);
      return rc ? SQLITE_ERROR : SQLITE_OK;
    }

    case SQLITE_FCNTL_LOCKSTATE: {
      *(int*)pArg = pFile->eFileLock;
      return SQLITE_OK;
    }
    case SQLITE_FCNTL_LAST_ERRNO: {
      *(int*)pArg = pFile->lastErrno;
      return SQLITE_OK;

    }
#endif /* SQLITE_ENABLE_LOCKING_STYLE && defined(__APPLE__) */
  }
  return SQLITE_NOTFOUND;
}

/*
** If pFd->sectorSize is non-zero when this function is called, it is a
** no-op. Otherwise, the values of pFd->sectorSize and 
** pFd->deviceCharacteristics are set according to the file-system 
** characteristics. 
**
** There are two versions of this function. One for QNX and one for all


** other systems.
*/
#ifndef __QNXNTO__
static void setDeviceCharacteristics(unixFile *pFd){
  if( pFd->sectorSize==0 ){
    int res;
    assert( pFd->deviceCharacteristics==0 );

    /* Check for support for F2FS atomic batch writes. */
    res = osIoctl(pFd->h, F2FS_IOC_START_VOLATILE_WRITE);
    if( res==SQLITE_OK ){
      osIoctl(pFd->h, F2FS_IOC_ABORT_VOLATILE_WRITE);
      pFd->deviceCharacteristics = 
        SQLITE_IOCAP_BATCH_ATOMIC |
        SQLITE_IOCAP_ATOMIC |
        SQLITE_IOCAP_SEQUENTIAL |
        SQLITE_IOCAP_SAFE_APPEND;
    }

    /* Set the POWERSAFE_OVERWRITE flag if requested. */
    if( pFd->ctrlFlags & UNIXFILE_PSOW ){
      pFd->deviceCharacteristics |= SQLITE_IOCAP_POWERSAFE_OVERWRITE;
    }

    pFd->sectorSize = SQLITE_DEFAULT_SECTOR_SIZE;


  }
}
#else
#include <sys/dcmd_blk.h>
#include <sys/statvfs.h>

static void setDeviceCharacteristics(unixFile *pFile){
  if( pFile->sectorSize == 0 ){
    struct statvfs fsInfo;
       
    /* Set defaults for non-supported filesystems */
    pFile->sectorSize = SQLITE_DEFAULT_SECTOR_SIZE;
    pFile->deviceCharacteristics = 0;
    if( fstatvfs(pFile->h, &fsInfo) == -1 ) {

  }
  /* Last chance verification.  If the sector size isn't a multiple of 512
  ** then it isn't valid.*/
  if( pFile->sectorSize % 512 != 0 ){
    pFile->deviceCharacteristics = 0;
    pFile->sectorSize = SQLITE_DEFAULT_SECTOR_SIZE;
  }

}
#endif

/*
** Return the sector size in bytes of the underlying block device for
** the specified file. This is almost always 512 bytes, but may be
** larger for some devices.
**
** SQLite code assumes this function cannot fail. It also assumes that
** if two files are created in the same file-system directory (i.e.
** a database and its journal file) that the sector size will be the
** same for both.
*/
static int unixSectorSize(sqlite3_file *id){
  unixFile *pFd = (unixFile*)id;
  setDeviceCharacteristics(pFd);
  return pFd->sectorSize;
}

/*
** Return the device characteristics for the file.
**
** This VFS is set up to return SQLITE_IOCAP_POWERSAFE_OVERWRITE by default.
** However, that choice is controversial since technically the underlying
** file system does not always provide powersafe overwrites.  (In other
** words, after a power-loss event, parts of the file that were never
** written might end up being altered.)  However, non-PSOW behavior is very,
** very rare.  And asserting PSOW makes a large reduction in the amount
** of required I/O for journaling, since a lot of padding is eliminated.
**  Hence, while POWERSAFE_OVERWRITE is on by default, there is a file-control
** available to turn it off and URI query parameter available to turn it off.
*/
static int unixDeviceCharacteristics(sqlite3_file *id){
  unixFile *pFd = (unixFile*)id;



  setDeviceCharacteristics(pFd);




  return pFd->deviceCharacteristics;
}

#if !defined(SQLITE_OMIT_WAL) || SQLITE_MAX_MMAP_SIZE>0

/*
** Return the system page size.
**

    UNIXVFS("unix-proxy",    proxyIoFinder ),
#endif
  };
  unsigned int i;          /* Loop counter */

  /* Double-check that the aSyscall[] array has been constructed
  ** correctly.  See ticket [bb3a86e890c8e96ab] */
  assert( ArraySize(aSyscall)==29 );

  /* Register all VFSes defined in the aVfs[] array */
  for(i=0; i<(sizeof(aVfs)/sizeof(sqlite3_vfs)); i++){
    sqlite3_vfs_register(&aVfs[i], i==0);
  }
  return SQLITE_OK; 
}

    case PAGER_WRITER_FINISHED:
      assert( p->eLock==EXCLUSIVE_LOCK );
      assert( pPager->errCode==SQLITE_OK );
      assert( !pagerUseWal(pPager) );
      assert( isOpen(p->jfd) 
           || p->journalMode==PAGER_JOURNALMODE_OFF 
           || p->journalMode==PAGER_JOURNALMODE_WAL 
           || (sqlite3OsDeviceCharacteristics(p->fd)&SQLITE_IOCAP_BATCH_ATOMIC)
      );
      break;

    case PAGER_ERROR:
      /* There must be at least one outstanding reference to the pager if
      ** in ERROR state. Otherwise the pager should have already dropped
      ** back to OPEN state.

  if( !pPager->tempFile ){
    int dc;                           /* Device characteristics */
    int nSector;                      /* Sector size */
    int szPage;                       /* Page size */

    assert( isOpen(pPager->fd) );
    dc = sqlite3OsDeviceCharacteristics(pPager->fd);
    /* use in-memory journal */
    if( dc&SQLITE_IOCAP_BATCH_ATOMIC ){
      return -1;
    }

    nSector = pPager->sectorSize;
    szPage = pPager->pageSize;

    assert(SQLITE_IOCAP_ATOMIC512==(512>>8));
    assert(SQLITE_IOCAP_ATOMIC64K==(65536>>8));
    if( 0==(dc&(SQLITE_IOCAP_ATOMIC|(szPage>>8)) || nSector>szPage) ){
      return 0;

  assert( assert_pager_state(pPager) );
  assert( pPager->eState!=PAGER_ERROR );
  if( pPager->eState<PAGER_WRITER_LOCKED && pPager->eLock<RESERVED_LOCK ){
    return SQLITE_OK;
  }

  releaseAllSavepoints(pPager);
  assert( isOpen(pPager->jfd) || pPager->pInJournal==0 
      || (sqlite3OsDeviceCharacteristics(pPager->fd)&SQLITE_IOCAP_BATCH_ATOMIC)
  );
  if( isOpen(pPager->jfd) ){
    assert( !pagerUseWal(pPager) );

    /* Finalize the journal file. */
    if( sqlite3JournalIsInMemory(pPager->jfd) ){
      /* assert( pPager->journalMode==PAGER_JOURNALMODE_MEMORY ); */
      sqlite3OsClose(pPager->jfd);

  if( pagerUseWal(pPager) ){
    /* Write a single frame for this page to the log. */
    rc = subjournalPageIfRequired(pPg); 
    if( rc==SQLITE_OK ){
      rc = pagerWalFrames(pPager, pPg, 0, 0);
    }
  }else{
    
#ifdef SQLITE_ENABLE_ATOMIC_WRITE
    if( pPager->tempFile==0 ){
      rc = sqlite3JournalCreate(pPager->jfd);
      if( rc!=SQLITE_OK ) return pager_error(pPager, rc);
    }
#endif
  
    /* Sync the journal file if required. */
    if( pPg->flags&PGHDR_NEED_SYNC 
     || pPager->eState==PAGER_WRITER_CACHEMOD
    ){
      rc = syncJournal(pPager, 1);
    }

      **
      ** Otherwise, if the optimization is both enabled and applicable,
      ** then call pager_incr_changecounter() to update the change-counter
      ** in 'direct' mode. In this case the journal file will never be
      ** created for this transaction.
      */
  #ifdef SQLITE_ENABLE_ATOMIC_WRITE
      sqlite3_file *fd = pPager->fd;
      int bBatch = zMaster==0      /* An SQLITE_IOCAP_BATCH_ATOMIC commit */
        && (sqlite3OsDeviceCharacteristics(fd) & SQLITE_IOCAP_BATCH_ATOMIC)
        && pPager->journalMode!=PAGER_JOURNALMODE_MEMORY
        && sqlite3JournalIsInMemory(pPager->jfd);

      if( bBatch==0 ){
        PgHdr *pPg;
        assert( isOpen(pPager->jfd) 
            || pPager->journalMode==PAGER_JOURNALMODE_OFF 
            || pPager->journalMode==PAGER_JOURNALMODE_WAL 
            );
        if( !zMaster && isOpen(pPager->jfd) 
         && pPager->journalOff==jrnlBufferSize(pPager) 
         && pPager->dbSize>=pPager->dbOrigSize
         && (!(pPg = sqlite3PcacheDirtyList(pPager->pPCache)) || 0==pPg->pDirty)
        ){
          /* Update the db file change counter via the direct-write method. The 
          ** following call will modify the in-memory representation of page 1 
          ** to include the updated change counter and then write page 1 
          ** directly to the database file. Because of the atomic-write 
          ** property of the host file-system, this is safe.
          */
          rc = pager_incr_changecounter(pPager, 1);
        }else{
          rc = sqlite3JournalCreate(pPager->jfd);
          if( rc==SQLITE_OK ){
            rc = pager_incr_changecounter(pPager, 0);
          }
        }
      }else

  #endif
      rc = pager_incr_changecounter(pPager, 0);
      if( rc!=SQLITE_OK ) goto commit_phase_one_exit;
  
      /* Write the master journal name into the journal file. If a master 
      ** journal file name has already been written to the journal file, 
      ** or if zMaster is NULL (no master journal), then this call is a no-op.
      */
      rc = writeMasterJournal(pPager, zMaster);

      ** journal requires a sync here. However, in locking_mode=exclusive
      ** on a system under memory pressure it is just possible that this is 
      ** not the case. In this case it is likely enough that the redundant
      ** xSync() call will be changed to a no-op by the OS anyhow. 
      */
      rc = syncJournal(pPager, 0);
      if( rc!=SQLITE_OK ) goto commit_phase_one_exit;

      if( bBatch ){
        /* The pager is now in DBMOD state. But regardless of what happens
        ** next, attempting to play the journal back into the database would
        ** be unsafe. Close it now to make sure that does not happen.  */
        sqlite3OsClose(pPager->jfd);
        rc = sqlite3OsFileControl(fd, SQLITE_FCNTL_BEGIN_ATOMIC_WRITE, 0);
        if( rc!=SQLITE_OK ) goto commit_phase_one_exit;
      }
      rc = pager_write_pagelist(pPager,sqlite3PcacheDirtyList(pPager->pPCache));
      if( bBatch ){
        if( rc==SQLITE_OK ){
          rc = sqlite3OsFileControl(fd, SQLITE_FCNTL_COMMIT_ATOMIC_WRITE, 0);
        }else{
          sqlite3OsFileControl(fd, SQLITE_FCNTL_ROLLBACK_ATOMIC_WRITE, 0);
        }
      }

      if( rc!=SQLITE_OK ){
        assert( rc!=SQLITE_IOERR_BLOCKED );
        goto commit_phase_one_exit;
      }
      sqlite3PcacheCleanAll(pPager->pPCache);

      /* If the file on disk is smaller than the database image, use 

#define SQLITE_IOCAP_ATOMIC32K              0x00000080
#define SQLITE_IOCAP_ATOMIC64K              0x00000100
#define SQLITE_IOCAP_SAFE_APPEND            0x00000200
#define SQLITE_IOCAP_SEQUENTIAL             0x00000400
#define SQLITE_IOCAP_UNDELETABLE_WHEN_OPEN  0x00000800
#define SQLITE_IOCAP_POWERSAFE_OVERWRITE    0x00001000
#define SQLITE_IOCAP_IMMUTABLE              0x00002000
#define SQLITE_IOCAP_BATCH_ATOMIC           0x00004000

/*
** CAPI3REF: File Locking Levels
**
** SQLite uses one of these integer values as the second
** argument to calls it makes to the xLock() and xUnlock() methods
** of an [sqlite3_io_methods] object.

** <li> [SQLITE_IOCAP_ATOMIC32K]
** <li> [SQLITE_IOCAP_ATOMIC64K]
** <li> [SQLITE_IOCAP_SAFE_APPEND]
** <li> [SQLITE_IOCAP_SEQUENTIAL]
** <li> [SQLITE_IOCAP_UNDELETABLE_WHEN_OPEN]
** <li> [SQLITE_IOCAP_POWERSAFE_OVERWRITE]
** <li> [SQLITE_IOCAP_IMMUTABLE]
** <li> [SQLITE_IOCAP_BATCH_ATOMIC]
** </ul>
**
** The SQLITE_IOCAP_ATOMIC property means that all writes of
** any size are atomic.  The SQLITE_IOCAP_ATOMICnnn values
** mean that writes of blocks that are nnn bytes in size and
** are aligned to an address which is an integer multiple of
** nnn are atomic.  The SQLITE_IOCAP_SAFE_APPEND value means

#define SQLITE_FCNTL_WAL_BLOCK              24
#define SQLITE_FCNTL_ZIPVFS                 25
#define SQLITE_FCNTL_RBU                    26
#define SQLITE_FCNTL_VFS_POINTER            27
#define SQLITE_FCNTL_JOURNAL_POINTER        28
#define SQLITE_FCNTL_WIN32_GET_HANDLE       29
#define SQLITE_FCNTL_PDB                    30

#define SQLITE_FCNTL_BEGIN_ATOMIC_WRITE     31
#define SQLITE_FCNTL_COMMIT_ATOMIC_WRITE    32
#define SQLITE_FCNTL_ROLLBACK_ATOMIC_WRITE  33

/* deprecated names */
#define SQLITE_GET_LOCKPROXYFILE      SQLITE_FCNTL_GET_LOCKPROXYFILE
#define SQLITE_SET_LOCKPROXYFILE      SQLITE_FCNTL_SET_LOCKPROXYFILE
#define SQLITE_LAST_ERRNO             SQLITE_FCNTL_LAST_ERRNO

# Tests for the xNextSystemCall method.
#
foreach s {
    open close access getcwd stat fstat ftruncate
    fcntl read pread write pwrite fchmod fallocate
    pread64 pwrite64 unlink openDirectory mkdir rmdir 
    statvfs fchown geteuid umask mmap munmap mremap
    getpagesize readlink lstat ioctl
} {
  if {[test_syscall exists $s]} {lappend syscall_list $s}
}
do_test 3.1 { lsort [test_syscall list] } [lsort $syscall_list]

#-------------------------------------------------------------------------
# This test verifies that if a call to open() fails and errno is set to