charset_macosxfs.c

samba-3.0.22.tar.gz 编译smb服务器的源码

	    (0 != (*outbuf)[outsize*2-1] || 0 != (*outbuf)[outsize*2-2])) {

		if ((outsize*2+2) > *outbytesleft) {
			debug_out("String conversion: "
				  "Output buffer too small\n");
			hexdump("UTF8->UTF16LE (old) input",
				*inbuf, *inbytesleft);
			errno = E2BIG;
			return -1;
		}

		(*outbuf)[outsize*2] = (*outbuf)[outsize*2+1] = 0;
		outsize += 2;
	}

	*inbuf += *inbytesleft;
	*inbytesleft = 0;
	*outbuf += outsize*2;
	*outbytesleft -= outsize*2;

	return 0;
}

static size_t macosxfs_encoding_push(
	void *cd,				/* Encoder handle */
	char **inbuf, size_t *inbytesleft,	/* UTF-16-LE string */
	char **outbuf, size_t *outbytesleft)	/* Script string */
{
	static const int script_code = kCFStringEncodingUTF8;
	static CFMutableStringRef cfstring = NULL;
	static UniChar *buffer = NULL;
	static size_t buflen = 0;
	CFIndex outsize, cfsize, charsconverted;

	(void) cd; /* UNUSED */

	if (0 == *inbytesleft) {
		return 0;
	}

	/*
	 * We need a buffer that can hold 4 times the original data,
	 * because that is the theoretical maximum that decomposition
	 * can create currently (in Unicode 4.0).
	 */
	buffer = set_ucbuffer_with_le_copy(
		buffer, &buflen, *inbuf, *inbytesleft, 3 * *inbytesleft);

	if (NULL == cfstring) {
		cfstring = CFStringCreateMutableWithExternalCharactersNoCopy(
			kCFAllocatorDefault,
			buffer, *inbytesleft/2, buflen/2,
			kCFAllocatorNull);
	} else {
		CFStringSetExternalCharactersNoCopy(
			cfstring,
			buffer, *inbytesleft/2, buflen/2);
	}

	/*
	 * Decompose characters, using the non-canonical decomposition
	 * form.
	 *
	 * NB: This isn't exactly what HFS+ wants (see note on
	 * kCFStringEncodingUseHFSPlusCanonical in
	 * CFStringEncodingConverter.h), but AFAIK it's the best that
	 * the official API can do.
	 */
	CFStringNormalize(cfstring, kCFStringNormalizationFormD);

	cfsize = CFStringGetLength(cfstring);
	charsconverted = CFStringGetBytes(
		cfstring, CFRangeMake(0,cfsize),
		script_code, 0, False,
		*outbuf, *outbytesleft, &outsize);

	if (0 == charsconverted) {
		debug_out("String conversion: "
			  "Buffer too small or not convertable\n");
		hexdump("UTF16LE->UTF8 (old) input",
			*inbuf, *inbytesleft);
		errno = EILSEQ; /* Probably more likely. */
		return -1;
	}

	/*
	 * Add a converted null byte, if the CFString conversions
	 * prevented that until now.
	 */
	if (0 == (*inbuf)[*inbytesleft-1] && 0 == (*inbuf)[*inbytesleft-2] &&
	    (0 != (*outbuf)[outsize-1])) {

		if (((size_t)outsize+1) > *outbytesleft) {
			debug_out("String conversion: "
				  "Output buffer too small\n");
			hexdump("UTF16LE->UTF8 (old) input",
				*inbuf, *inbytesleft);
			errno = E2BIG;
			return -1;
		}

		(*outbuf)[outsize] = 0;
		++outsize;
	}

	*inbuf += *inbytesleft;
	*inbytesleft = 0;
	*outbuf += outsize;
	*outbytesleft -= outsize;

	return 0;
}

#else /* USE_INTERNAL_API */

/*
 * An implementation based on internal code as known from the
 * OpenDarwin CVS.
 *
 * This code doesn't need much memory management because it uses
 * functions that operate on the raw memory directly.
 *
 * The push routine here is faster and more compatible with HFS+ than
 * the other implementation above.  The pull routine is only faster
 * for some strings, slightly slower for others.  The pull routine
 * looses because it has to iterate over the data twice, once to
 * decode UTF-8 and than to do the character composition required by
 * Windows.
 */
static size_t macosxfs_encoding_pull(
	void *cd,				/* Encoder handle */
	char **inbuf, size_t *inbytesleft,	/* Script string */
	char **outbuf, size_t *outbytesleft)	/* UTF-16-LE string */
{
	static const int script_code = kCFStringEncodingUTF8;
	UInt32 srcCharsUsed = 0;
	UInt32 dstCharsUsed = 0;
	UInt32 result;
	uint32_t dstDecomposedUsed = 0;
	uint32_t dstPrecomposedUsed = 0;

	(void) cd; /* UNUSED */

	if (0 == *inbytesleft) {
		return 0;
	}

        result = CFStringEncodingBytesToUnicode(
		script_code, kCFStringEncodingComposeCombinings,
		*inbuf, *inbytesleft, &srcCharsUsed,
		(UniChar*)*outbuf, *outbytesleft, &dstCharsUsed);

	switch(result) {
	case kCFStringEncodingConversionSuccess:
		if (*inbytesleft == srcCharsUsed)
			break;
		else
			; /*fall through*/
	case kCFStringEncodingInsufficientOutputBufferLength:
		debug_out("String conversion: "
			  "Output buffer too small\n");
		hexdump("UTF8->UTF16LE (new) input",
			*inbuf, *inbytesleft);
		errno = E2BIG;
		return -1;
	case kCFStringEncodingInvalidInputStream:
		/*
		 * HACK: smbd/mangle_hash2.c:is_legal_name() expects
		 * errors here.  That function will always pass 2
		 * characters.  smbd/open.c:check_for_pipe() cuts a
		 * patchname to 10 characters blindly.  Suppress the
		 * debug output in those cases.
		 */
		if(2 != *inbytesleft && 10 != *inbytesleft) {
			debug_out("String conversion: "
				  "Invalid input sequence\n");
			hexdump("UTF8->UTF16LE (new) input",
				*inbuf, *inbytesleft);
		}
		errno = EILSEQ;
		return -1;
	case kCFStringEncodingConverterUnavailable:
		debug_out("String conversion: "
			  "Unknown encoding\n");
		hexdump("UTF8->UTF16LE (new) input",
			*inbuf, *inbytesleft);
		errno = EINVAL;
		return -1;
	}

	/*
	 * It doesn't look like CFStringEncodingBytesToUnicode() can
	 * produce precomposed characters (flags=ComposeCombinings
	 * doesn't do it), so we need another pass over the data here.
	 * We can do this in-place, as the string can only get
	 * shorter.
	 *
	 * (Actually in theory there should be an internal
	 * decomposition and reordering before the actual composition
	 * step.  But we should be able to rely on that we always get
	 * fully decomposed strings for input, so this can't create
	 * problems in reality.)
	 */
	CFUniCharPrecompose(
		(const UTF16Char *)*outbuf, dstCharsUsed, &dstDecomposedUsed,
		(UTF16Char *)*outbuf, dstCharsUsed, &dstPrecomposedUsed);

	native_to_le(*outbuf, dstPrecomposedUsed*2);

	*inbuf += srcCharsUsed;
	*inbytesleft -= srcCharsUsed;
	*outbuf += dstPrecomposedUsed*2;
	*outbytesleft -= dstPrecomposedUsed*2;

	return 0;
}

static size_t macosxfs_encoding_push(
	void *cd,				/* Encoder handle */
	char **inbuf, size_t *inbytesleft,	/* UTF-16-LE string */
	char **outbuf, size_t *outbytesleft)	/* Script string */
{
	static const int script_code = kCFStringEncodingUTF8;
	static UniChar *buffer = NULL;
	static size_t buflen = 0;
	UInt32 srcCharsUsed=0, dstCharsUsed=0, result;

	(void) cd; /* UNUSED */

	if (0 == *inbytesleft) {
		return 0;
	}

	buffer = set_ucbuffer_with_le(
		buffer, &buflen, *inbuf, *inbytesleft);

	result = CFStringEncodingUnicodeToBytes(
		script_code, kCFStringEncodingUseHFSPlusCanonical,
		buffer, *inbytesleft/2, &srcCharsUsed,
		*outbuf, *outbytesleft, &dstCharsUsed);

	switch(result) {
	case kCFStringEncodingConversionSuccess:
		if (*inbytesleft/2 == srcCharsUsed)
			break;
		else
			; /*fall through*/
	case kCFStringEncodingInsufficientOutputBufferLength:
		debug_out("String conversion: "
			  "Output buffer too small\n");
		hexdump("UTF16LE->UTF8 (new) input",
			*inbuf, *inbytesleft);
		errno = E2BIG;
		return -1;
	case kCFStringEncodingInvalidInputStream:
		/*
		 * HACK: smbd/open.c:check_for_pipe():is_legal_name()
		 * cuts a pathname to 10 characters blindly.  Suppress
		 * the debug output in those cases.
		 */
		if(10 != *inbytesleft) {
			debug_out("String conversion: "
				  "Invalid input sequence\n");
			hexdump("UTF16LE->UTF8 (new) input",
				*inbuf, *inbytesleft);
		}
		errno = EILSEQ;
		return -1;
	case kCFStringEncodingConverterUnavailable:
		debug_out("String conversion: "
			  "Unknown encoding\n");
		hexdump("UTF16LE->UTF8 (new) input",
			*inbuf, *inbytesleft);
		errno = EINVAL;
		return -1;
	}

	*inbuf += srcCharsUsed*2;
	*inbytesleft -= srcCharsUsed*2;
	*outbuf += dstCharsUsed;
	*outbytesleft -= dstCharsUsed;

	return 0;
}

#endif /* USE_INTERNAL_API */

/*
 * For initialization, actually install the encoding as "macosxfs".
 */
static struct charset_functions macosxfs_encoding_functions = {
	"MACOSXFS", macosxfs_encoding_pull, macosxfs_encoding_push
};

NTSTATUS init_module(void)
{
	return smb_register_charset(&macosxfs_encoding_functions);
}

/* eof */