fd.c

关系型数据库 Postgresql 6.5.2

	{
		DO_DB(elog(DEBUG, "fileNameOpenFile: not enough descs, retry, er= %d",
				   errno));
		errno = 0;
		ReleaseLruFile();
		goto tryAgain;
	}

	if (vfdP->fd < 0)
	{
		FreeVfd(file);
		return -1;
	}
	++nfile;
	DO_DB(elog(DEBUG, "fileNameOpenFile: success %d",
			   vfdP->fd));

	Insert(file);

	vfdP->fileName = malloc(strlen(fileName) + 1);
	strcpy(vfdP->fileName, fileName);

	vfdP->fileFlags = fileFlags & ~(O_TRUNC | O_EXCL);
	vfdP->fileMode = fileMode;
	vfdP->seekPos = 0;
	vfdP->fdstate = 0x0;

	return file;
}

/*
 * open a file in the database directory ($PGDATA/base/...)
 */
File
FileNameOpenFile(FileName fileName, int fileFlags, int fileMode)
{
	File		fd;
	char	   *fname;

	fname = filepath(fileName);
	fd = fileNameOpenFile(fname, fileFlags, fileMode);
	pfree(fname);
	return fd;
}

/*
 * open a file in an arbitrary directory
 */
File
PathNameOpenFile(FileName fileName, int fileFlags, int fileMode)
{
	return fileNameOpenFile(fileName, fileFlags, fileMode);
}

/*
 * Open a temporary file that will disappear when we close it.
 *
 * This routine takes care of generating an appropriate tempfile name.
 * There's no need to pass in fileFlags or fileMode either, since only
 * one setting makes any sense for a temp file.
 */
File
OpenTemporaryFile(void)
{
	char		tempfilename[64];
	File		file;

	/*
	 * Generate a tempfile name that's unique within the current
	 * transaction
	 */
	snprintf(tempfilename, sizeof(tempfilename),
			 "pg_sorttemp%d.%ld", MyProcPid, tempFileCounter++);

	/* Open the file */
#ifndef __CYGWIN32__
	file = FileNameOpenFile(tempfilename,
							O_RDWR | O_CREAT | O_TRUNC, 0600);
#else
	file = FileNameOpenFile(tempfilename,
							O_RDWR | O_CREAT | O_TRUNC | O_BINARY, 0600);
#endif

	if (file <= 0)
		elog(ERROR, "Failed to create temporary file %s", tempfilename);

	/* Mark it for deletion at close or EOXact */
	VfdCache[file].fdstate |= FD_TEMPORARY;

	return file;
}

/*
 * close a file when done with it
 */
void
FileClose(File file)
{
	int			returnValue;

	Assert(FileIsValid(file));

	DO_DB(elog(DEBUG, "FileClose: %d (%s)",
			   file, VfdCache[file].fileName));

	if (!FileIsNotOpen(file))
	{

		/* remove the file from the lru ring */
		Delete(file);

		/* if we did any writes, sync the file before closing */
		if (VfdCache[file].fdstate & FD_DIRTY)
		{
			returnValue = pg_fsync(VfdCache[file].fd);
			Assert(returnValue != -1);
			VfdCache[file].fdstate &= ~FD_DIRTY;
		}

		/* close the file */
		returnValue = close(VfdCache[file].fd);
		Assert(returnValue != -1);

		--nfile;
		VfdCache[file].fd = VFD_CLOSED;
	}

	/*
	 * Delete the file if it was temporary
	 */
	if (VfdCache[file].fdstate & FD_TEMPORARY)
		unlink(VfdCache[file].fileName);

	/*
	 * Return the Vfd slot to the free list
	 */
	FreeVfd(file);
}

/*
 * close a file and forcibly delete the underlying Unix file
 */
void
FileUnlink(File file)
{
	Assert(FileIsValid(file));

	DO_DB(elog(DEBUG, "FileUnlink: %d (%s)",
			   file, VfdCache[file].fileName));

	/* force FileClose to delete it */
	VfdCache[file].fdstate |= FD_TEMPORARY;

	FileClose(file);
}

int
FileRead(File file, char *buffer, int amount)
{
	int			returnCode;

	Assert(FileIsValid(file));

	DO_DB(elog(DEBUG, "FileRead: %d (%s) %d %p",
			   file, VfdCache[file].fileName, amount, buffer));

	FileAccess(file);
	returnCode = read(VfdCache[file].fd, buffer, amount);
	if (returnCode > 0)
		VfdCache[file].seekPos += returnCode;

	return returnCode;
}

int
FileWrite(File file, char *buffer, int amount)
{
	int			returnCode;

	Assert(FileIsValid(file));

	DO_DB(elog(DEBUG, "FileWrite: %d (%s) %d %p",
			   file, VfdCache[file].fileName, amount, buffer));

	FileAccess(file);
	returnCode = write(VfdCache[file].fd, buffer, amount);
	if (returnCode > 0)
		VfdCache[file].seekPos += returnCode;

	/* record the write */
	VfdCache[file].fdstate |= FD_DIRTY;

	return returnCode;
}

long
FileSeek(File file, long offset, int whence)
{
	Assert(FileIsValid(file));

	DO_DB(elog(DEBUG, "FileSeek: %d (%s) %ld %d",
			   file, VfdCache[file].fileName, offset, whence));

	if (FileIsNotOpen(file))
	{
		switch (whence)
		{
			case SEEK_SET:
				VfdCache[file].seekPos = offset;
				break;
			case SEEK_CUR:
				VfdCache[file].seekPos += offset;
				break;
			case SEEK_END:
				FileAccess(file);
				VfdCache[file].seekPos = lseek(VfdCache[file].fd, offset, whence);
				break;
			default:
				elog(ERROR, "FileSeek: invalid whence: %d", whence);
				break;
		}
	}
	else
		VfdCache[file].seekPos = lseek(VfdCache[file].fd, offset, whence);
	return VfdCache[file].seekPos;
}

/*
 * XXX not actually used but here for completeness
 */
#ifdef NOT_USED
long
FileTell(File file)
{
	Assert(FileIsValid(file));
	DO_DB(elog(DEBUG, "FileTell %d (%s)",
			   file, VfdCache[file].fileName));
	return VfdCache[file].seekPos;
}

#endif

int
FileTruncate(File file, int offset)
{
	int			returnCode;

	Assert(FileIsValid(file));

	DO_DB(elog(DEBUG, "FileTruncate %d (%s)",
			   file, VfdCache[file].fileName));

	FileSync(file);
	FileAccess(file);
	returnCode = ftruncate(VfdCache[file].fd, offset);
	return returnCode;
}

int
FileSync(File file)
{
	int			returnCode;

	Assert(FileIsValid(file));

	/*
	 * If the file isn't open, then we don't need to sync it; we always
	 * sync files when we close them.  Also, if we haven't done any writes
	 * that we haven't already synced, we can ignore the request.
	 */

	if (VfdCache[file].fd < 0 || !(VfdCache[file].fdstate & FD_DIRTY))
		returnCode = 0;
	else
	{
		returnCode = pg_fsync(VfdCache[file].fd);
		VfdCache[file].fdstate &= ~FD_DIRTY;
	}

	return returnCode;
}

int
FileNameUnlink(char *filename)
{
	int			retval;
	char	   *fname;

	fname = filepath(filename);
	retval = unlink(fname);
	pfree(fname);
	return retval;
}

/*
 * Routines that want to use stdio (ie, FILE*) should use AllocateFile
 * rather than plain fopen().  This lets fd.c deal with freeing FDs if
 * necessary to open the file.	When done, call FreeFile rather than fclose.
 *
 * Note that files that will be open for any significant length of time
 * should NOT be handled this way, since they cannot share kernel file
 * descriptors with other files; there is grave risk of running out of FDs
 * if anyone locks down too many FDs.  Most callers of this routine are
 * simply reading a config file that they will read and close immediately.
 *
 * fd.c will automatically close all files opened with AllocateFile at
 * transaction commit or abort; this prevents FD leakage if a routine
 * that calls AllocateFile is terminated prematurely by elog(ERROR).
 */

FILE *
AllocateFile(char *name, char *mode)
{
	FILE	   *file;

	DO_DB(elog(DEBUG, "AllocateFile: Allocated %d.", numAllocatedFiles));

	if (numAllocatedFiles >= MAX_ALLOCATED_FILES)
		elog(ERROR, "AllocateFile: too many private FDs demanded");

TryAgain:
	if ((file = fopen(name, mode)) == NULL)
	{
		if (errno == EMFILE || errno == ENFILE)
		{
			DO_DB(elog(DEBUG, "AllocateFile: not enough descs, retry, er= %d",
					   errno));
			errno = 0;
			ReleaseLruFile();
			goto TryAgain;
		}
	}
	else
		allocatedFiles[numAllocatedFiles++] = file;
	return file;
}

void
FreeFile(FILE *file)
{
	int			i;

	DO_DB(elog(DEBUG, "FreeFile: Allocated %d.", numAllocatedFiles));

	/* Remove file from list of allocated files, if it's present */
	for (i = numAllocatedFiles; --i >= 0;)
	{
		if (allocatedFiles[i] == file)
		{
			allocatedFiles[i] = allocatedFiles[--numAllocatedFiles];
			break;
		}
	}
	if (i < 0)
		elog(NOTICE, "FreeFile: file was not obtained from AllocateFile");

	fclose(file);
}

/*
 * closeAllVfds
 *
 * Force all VFDs into the physically-closed state, so that the fewest
 * possible number of kernel file descriptors are in use.  There is no
 * change in the logical state of the VFDs.
 */
void
closeAllVfds()
{
	Index		i;

	if (SizeVfdCache > 0)
	{
		Assert(FileIsNotOpen(0));		/* Make sure ring not corrupted */
		for (i = 1; i < SizeVfdCache; i++)
		{
			if (!FileIsNotOpen(i))
				LruDelete(i);
		}
	}
}

/*
 * AtEOXact_Files
 *
 * This routine is called during transaction commit or abort (it doesn't
 * particularly care which).  All still-open temporary-file VFDs are closed,
 * which also causes the underlying files to be deleted.  Furthermore,
 * all "allocated" stdio files are closed.
 */
void
AtEOXact_Files(void)
{
	Index		i;

	if (SizeVfdCache > 0)
	{
		Assert(FileIsNotOpen(0));		/* Make sure ring not corrupted */
		for (i = 1; i < SizeVfdCache; i++)
		{
			if ((VfdCache[i].fdstate & FD_TEMPORARY) &&
				VfdCache[i].fileName != NULL)
				FileClose(i);
		}
	}

	while (numAllocatedFiles > 0)
		FreeFile(allocatedFiles[0]);

	/*
	 * Reset the tempfile name counter to 0; not really necessary, but
	 * helps keep the names from growing unreasonably long.
	 */
	tempFileCounter = 0;
}


/*
 * Operations on BufFiles --- a very incomplete emulation of stdio
 * atop virtual Files.	Currently, we only support the buffered-I/O
 * aspect of stdio: a read or write of the low-level File occurs only
 * when the buffer is filled or emptied.  This is an even bigger win
 * for virtual Files than ordinary kernel files, since reducing the
 * frequency with which a virtual File is touched reduces "thrashing"
 * of opening/closing file descriptors.
 *
 * Note that BufFile structs are allocated with palloc(), and therefore
 * will go away automatically at transaction end.  If the underlying
 * virtual File is made with OpenTemporaryFile, then all resources for
 * the file are certain to be cleaned up even if processing is aborted
 * by elog(ERROR).
 */

struct BufFile
{
	File		file;			/* the underlying virtual File */
	bool		dirty;			/* does buffer need to be written? */
	int			pos;			/* next read/write position in buffer */
	int			nbytes;			/* total # of valid bytes in buffer */
	char		buffer[BLCKSZ];
};


/*
 * Create a BufFile and attach it to an (already opened) virtual File.
 *
 * This is comparable to fdopen() in stdio.
 */
BufFile    *
BufFileCreate(File file)
{
	BufFile    *bfile = (BufFile *) palloc(sizeof(BufFile));

	bfile->file = file;
	bfile->dirty = false;
	bfile->pos = 0;
	bfile->nbytes = 0;

	return bfile;
}

/*
 * Close a BufFile
 *
 * Like fclose(), this also implicitly FileCloses the underlying File.
 */
void
BufFileClose(BufFile *file)
{
	/* flush any unwritten data */
	BufFileFlush(file);
	/* close the underlying (with delete if it's a temp file) */
	FileClose(file->file);
	/* release the buffer space */
	pfree(file);
}

/* BufFileRead
 *
 * Like fread() except we assume 1-byte element size.
 */
size_t
BufFileRead(BufFile *file, void *ptr, size_t size)
{
	size_t		nread = 0;
	size_t		nthistime;

	if (file->dirty)
	{
		elog(NOTICE, "BufFileRead: should have flushed after writing");
		BufFileFlush(file);
	}

	while (size > 0)
	{
		if (file->pos >= file->nbytes)
		{
			/* Try to load more data into buffer */
			file->pos = 0;
			file->nbytes = FileRead(file->file, file->buffer,
									sizeof(file->buffer));
			if (file->nbytes < 0)
				file->nbytes = 0;
			if (file->nbytes <= 0)
				break;			/* no more data available */
		}

		nthistime = file->nbytes - file->pos;
		if (nthistime > size)
			nthistime = size;
		Assert(nthistime > 0);

		memcpy(ptr, file->buffer + file->pos, nthistime);

		file->pos += nthistime;
		ptr = (void *) ((char *) ptr + nthistime);
		size -= nthistime;
		nread += nthistime;
	}

	return nread;
}

/* BufFileWrite
 *
 * Like fwrite() except we assume 1-byte element size.
 */
size_t
BufFileWrite(BufFile *file, void *ptr, size_t size)
{
	size_t		nwritten = 0;
	size_t		nthistime;

	while (size > 0)
	{
		if (file->pos >= BLCKSZ)
		{
			/* Buffer full, dump it out */
			if (file->dirty)
			{
				if (FileWrite(file->file, file->buffer, file->nbytes) < 0)
					break;		/* I/O error */
				file->dirty = false;
			}
			file->pos = 0;
			file->nbytes = 0;
		}

		nthistime = BLCKSZ - file->pos;
		if (nthistime > size)
			nthistime = size;
		Assert(nthistime > 0);

		memcpy(file->buffer + file->pos, ptr, nthistime);

		file->dirty = true;
		file->pos += nthistime;
		if (file->nbytes < file->pos)
			file->nbytes = file->pos;
		ptr = (void *) ((char *) ptr + nthistime);
		size -= nthistime;
		nwritten += nthistime;
	}

	return nwritten;
}

/* BufFileFlush
 *
 * Like fflush()
 */
static int
BufFileFlush(BufFile *file)
{
	if (file->dirty)
	{
		if (FileWrite(file->file, file->buffer, file->nbytes) < 0)
			return EOF;
		file->dirty = false;
	}

	return 0;
}

/* BufFileSeek
 *
 * Like fseek(), or really more like lseek() since the return value is
 * the new file offset (or -1 in case of error).
 */
long
BufFileSeek(BufFile *file, long offset, int whence)
{
	if (BufFileFlush(file) < 0)
		return -1L;
	file->pos = 0;
	file->nbytes = 0;
	return FileSeek(file->file, offset, whence);
}