/*
** 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.
**
** Under Unix, also make sure that the directory entry for the file
** has been created by fsync-ing the directory that contains the file.
** If we do not do this and we encounter a power failure, the directory
** entry for the journal might not exist after we reboot. The next
** SQLite to access the file will not know that the journal exists (because
** the directory entry for the journal was never created) and the transaction
** will not roll back - possibly leading to database corruption.
*/
static int unixSync(sqlite3_file *id, int flags){
int rc;
unixFile *pFile = (unixFile*)id;
int isDataOnly = (flags&SQLITE_SYNC_DATAONLY);
int isFullsync = (flags&0x0F)==SQLITE_SYNC_FULL;
/* Check that one of SQLITE_SYNC_NORMAL or FULL was passed */
assert((flags&0x0F)==SQLITE_SYNC_NORMAL
|| (flags&0x0F)==SQLITE_SYNC_FULL
);
/* Unix cannot, but some systems may return SQLITE_FULL from here. This
** line is to test that doing so does not cause any problems.
*/
SimulateDiskfullError( return SQLITE_FULL );
assert( pFile );
OSTRACE2("SYNC %-3d\n", pFile->h);
rc = full_fsync(pFile->h, isFullsync, isDataOnly);
SimulateIOError( rc=1 );
if( rc ){
return SQLITE_IOERR_FSYNC;
}
return SQLITE_OK;
}
/*
** Write data from a buffer into a file. Return SQLITE_OK on success
** or some other error code on failure.
*/
static int unixWrite(
sqlite3_file *id,
const void *pBuf,
int amt,
sqlite3_int64 offset
){
int wrote = 0;
assert( id );
assert( amt>0 );
while( amt>0 && (wrote = seekAndWrite((unixFile*)id, offset, pBuf, amt))>0 ){
amt -= wrote;
offset += wrote;
pBuf = &((char*)pBuf)[wrote];
}
SimulateIOError(( wrote=(-1), amt=1 ));
SimulateDiskfullError(( wrote=0, amt=1 ));
if( amt>0 ){
if( wrote<0 ){
return SQLITE_IOERR_WRITE;
}else{
return SQLITE_FULL;
}
}
return SQLITE_OK;
}
/*
** Determine the current size of a file in bytes
*/
int sqlite3OsFileSize(OsFile *id, i64 *pSize){
struct stat buf;
assert( id->isOpen );
SimulateIOError(SQLITE_IOERR);
if( fstat(id->h, &buf)!=0 ){
return SQLITE_IOERR;
}
*pSize = buf.st_size;
return SQLITE_OK;
}
/*
** Sync the directory zDirname. This is a no-op on operating systems other
** than UNIX.
**
** This is used to make sure the master journal file has truely been deleted
** before making changes to individual journals on a multi-database commit.
** The F_FULLFSYNC option is not needed here.
*/
int sqlite3OsSyncDirectory(const char *zDirname){
int fd;
int r;
SimulateIOError(SQLITE_IOERR);
fd = open(zDirname, O_RDONLY|O_BINARY, 0);
TRACE3("DIRSYNC %-3d (%s)\n", fd, zDirname);
if( fd<0 ){
return SQLITE_CANTOPEN;
}
r = fsync(fd);
close(fd);
return ((r==0)?SQLITE_OK:SQLITE_IOERR);
}
/*
** Seek to the offset passed as the second argument, then read cnt
** bytes into pBuf. Return the number of bytes actually read.
**
** NB: If you define USE_PREAD or USE_PREAD64, then it might also
** be necessary to define _XOPEN_SOURCE to be 500. This varies from
** one system to another. Since SQLite does not define USE_PREAD
** any any form by default, we will not attempt to define _XOPEN_SOURCE.
** See tickets #2741 and #2681.
*/
static int seekAndRead(unixFile *id, sqlite3_int64 offset, void *pBuf, int cnt){
int got;
i64 newOffset;
TIMER_START;
#if defined(USE_PREAD)
got = pread(id->h, pBuf, cnt, offset);
SimulateIOError( got = -1 );
#elif defined(USE_PREAD64)
got = pread64(id->h, pBuf, cnt, offset);
SimulateIOError( got = -1 );
#else
newOffset = lseek(id->h, offset, SEEK_SET);
SimulateIOError( newOffset-- );
if( newOffset!=offset ){
return -1;
}
got = read(id->h, pBuf, cnt);
#endif
TIMER_END;
OSTRACE5("READ %-3d %5d %7lld %llu\n", id->h, got, offset, TIMER_ELAPSED);
return got;
}
/*
** Read data from a file into a buffer. Return SQLITE_OK if all
** bytes were read successfully and SQLITE_IOERR if anything goes
** wrong.
*/
int sqlite3OsRead(OsFile *id, void *pBuf, int amt){
DWORD got;
assert( id->isOpen );
SimulateIOError(SQLITE_IOERR);
TRACE3("READ %d lock=%d\n", id->h, id->locktype);
if( !ReadFile(id->h, pBuf, amt, &got, 0) ){
got = 0;
}
if( got==(DWORD)amt ){
return SQLITE_OK;
}else{
return SQLITE_IOERR;
}
}
/*
** Make sure all writes to a particular file are committed to disk.
**
** Under Unix, also make sure that the directory entry for the file
** has been created by fsync-ing the directory that contains the file.
** If we do not do this and we encounter a power failure, the directory
** entry for the journal might not exist after we reboot. The next
** SQLite to access the file will not know that the journal exists (because
** the directory entry for the journal was never created) and the transaction
** will not roll back - possibly leading to database corruption.
*/
int sqlite3OsSync(OsFile *id){
assert( id->isOpen );
SimulateIOError(SQLITE_IOERR);
TRACE2("SYNC %-3d\n", id->h);
if( full_fsync(id->h, id->fullSync) ){
return SQLITE_IOERR;
}
if( id->dirfd>=0 ){
TRACE2("DIRSYNC %-3d\n", id->dirfd);
full_fsync(id->dirfd, id->fullSync);
close(id->dirfd); /* Only need to sync once, so close the directory */
id->dirfd = -1; /* when we are done. */
}
return SQLITE_OK;
}
/*
** Read data from a file into a buffer. Return SQLITE_OK if all
** bytes were read successfully and SQLITE_IOERR if anything goes
** wrong.
*/
int sqlite3OsRead(OsFile *id, void *pBuf, int amt){
int got;
assert( id->isOpen );
SimulateIOError(SQLITE_IOERR);
TIMER_START;
got = read(id->h, pBuf, amt);
TIMER_END;
TRACE5("READ %-3d %5d %7d %d\n", id->h, got, last_page, TIMER_ELAPSED);
SEEK(0);
/* if( got<0 ) got = 0; */
if( got==amt ){
return SQLITE_OK;
}else{
return SQLITE_IOERR;
}
}
/*
** Write data from a buffer into a file. Return SQLITE_OK on success
** or some other error code on failure.
*/
int sqlite3OsWrite(OsFile *id, const void *pBuf, int amt){
int rc;
DWORD wrote;
assert( id->isOpen );
SimulateIOError(SQLITE_IOERR);
SimulateDiskfullError;
TRACE3("WRITE %d lock=%d\n", id->h, id->locktype);
while( amt>0 && (rc = WriteFile(id->h, pBuf, amt, &wrote, 0))!=0 && wrote>0 ){
amt -= wrote;
pBuf = &((char*)pBuf)[wrote];
}
if( !rc || amt>(int)wrote ){
return SQLITE_FULL;
}
return SQLITE_OK;
}
/*
** Write data from a buffer into a file. Return SQLITE_OK on success
** or some other error code on failure.
*/
int sqlite3OsWrite(OsFile *id, const void *pBuf, int amt){
int wrote = 0;
assert( id->isOpen );
assert( amt>0 );
SimulateIOError(SQLITE_IOERR);
SimulateDiskfullError;
TIMER_START;
while( amt>0 && (wrote = write(id->h, pBuf, amt))>0 ){
amt -= wrote;
pBuf = &((char*)pBuf)[wrote];
}
TIMER_END;
TRACE5("WRITE %-3d %5d %7d %d\n", id->h, wrote, last_page, TIMER_ELAPSED);
SEEK(0);
if( amt>0 ){
return SQLITE_FULL;
}
return SQLITE_OK;
}
/*
** Determine the current size of a file in bytes
*/
static int unixFileSize(sqlite3_file *id, i64 *pSize){
int rc;
struct stat buf;
assert( id );
rc = fstat(((unixFile*)id)->h, &buf);
SimulateIOError( rc=1 );
if( rc!=0 ){
return SQLITE_IOERR_FSTAT;
}
*pSize = buf.st_size;
/* When opening a zero-size database, the findLockInfo() procedure
** writes a single byte into that file in order to work around a bug
** in the OS-X msdos filesystem. In order to avoid problems with upper
** layers, we need to report this file size as zero even though it is
** really 1. Ticket #3260.
*/
if( *pSize==1 ) *pSize = 0;
return SQLITE_OK;
}
/*
** Truncate an open file to a specified size
*/
int sqlite3OsTruncate(OsFile *id, i64 nByte){
assert( id->isOpen );
SimulateIOError(SQLITE_IOERR);
return ftruncate(id->h, nByte)==0 ? SQLITE_OK : SQLITE_IOERR;
}