xref: /linux/fs/smb/client/cifs_unicode.c (revision 42eb01783091e49020221a8a7d6c00e154ae7e58)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *
4  *   Copyright (c) International Business Machines  Corp., 2000,2009
5  *   Modified by Steve French (sfrench@us.ibm.com)
6  */
7 #include <linux/fs.h>
8 #include <linux/slab.h>
9 #include "cifs_fs_sb.h"
10 #include "cifs_unicode.h"
11 #include "cifsglob.h"
12 #include "cifs_debug.h"
13 
14 /* Convert character using the SFU - "Services for Unix" remapping range */
15 static bool
convert_sfu_char(const __u16 src_char,char * target)16 convert_sfu_char(const __u16 src_char, char *target)
17 {
18 	/*
19 	 * BB: Cannot handle remapping UNI_SLASH until all the calls to
20 	 *     build_path_from_dentry are modified, as they use slash as
21 	 *     separator.
22 	 */
23 	switch (src_char) {
24 	case UNI_COLON:
25 		*target = ':';
26 		break;
27 	case UNI_ASTERISK:
28 		*target = '*';
29 		break;
30 	case UNI_QUESTION:
31 		*target = '?';
32 		break;
33 	case UNI_PIPE:
34 		*target = '|';
35 		break;
36 	case UNI_GRTRTHAN:
37 		*target = '>';
38 		break;
39 	case UNI_LESSTHAN:
40 		*target = '<';
41 		break;
42 	default:
43 		return false;
44 	}
45 	return true;
46 }
47 
48 /* Convert character using the SFM - "Services for Mac" remapping range */
49 static bool
convert_sfm_char(const __u16 src_char,char * target)50 convert_sfm_char(const __u16 src_char, char *target)
51 {
52 	if (src_char >= 0xF001 && src_char <= 0xF01F) {
53 		*target = src_char - 0xF000;
54 		return true;
55 	}
56 	switch (src_char) {
57 	case SFM_COLON:
58 		*target = ':';
59 		break;
60 	case SFM_DOUBLEQUOTE:
61 		*target = '"';
62 		break;
63 	case SFM_ASTERISK:
64 		*target = '*';
65 		break;
66 	case SFM_QUESTION:
67 		*target = '?';
68 		break;
69 	case SFM_PIPE:
70 		*target = '|';
71 		break;
72 	case SFM_GRTRTHAN:
73 		*target = '>';
74 		break;
75 	case SFM_LESSTHAN:
76 		*target = '<';
77 		break;
78 	case SFM_SPACE:
79 		*target = ' ';
80 		break;
81 	case SFM_PERIOD:
82 		*target = '.';
83 		break;
84 	default:
85 		return false;
86 	}
87 	return true;
88 }
89 
90 
91 /*
92  * cifs_mapchar - convert a host-endian char to proper char in codepage
93  * @target - where converted character should be copied
94  * @src_char - 2 byte host-endian source character
95  * @cp - codepage to which character should be converted
96  * @map_type - How should the 7 NTFS/SMB reserved characters be mapped to UCS2?
97  *
98  * This function handles the conversion of a single character. It is the
99  * responsibility of the caller to ensure that the target buffer is large
100  * enough to hold the result of the conversion (at least NLS_MAX_CHARSET_SIZE).
101  */
102 static int
cifs_mapchar(char * target,const __u16 * from,const struct nls_table * cp,int maptype)103 cifs_mapchar(char *target, const __u16 *from, const struct nls_table *cp,
104 	     int maptype)
105 {
106 	int len = 1;
107 	__u16 src_char;
108 
109 	src_char = *from;
110 
111 	if ((maptype == SFM_MAP_UNI_RSVD) && convert_sfm_char(src_char, target))
112 		return len;
113 	else if ((maptype == SFU_MAP_UNI_RSVD) &&
114 		  convert_sfu_char(src_char, target))
115 		return len;
116 
117 	/* if character not one of seven in special remap set */
118 	len = cp->uni2char(src_char, target, NLS_MAX_CHARSET_SIZE);
119 	if (len <= 0)
120 		goto surrogate_pair;
121 
122 	return len;
123 
124 surrogate_pair:
125 	/* convert SURROGATE_PAIR and IVS */
126 	if (strcmp(cp->charset, "utf8"))
127 		goto unknown;
128 	len = utf16s_to_utf8s(from, 3, UTF16_LITTLE_ENDIAN, target, 6);
129 	if (len <= 0)
130 		goto unknown;
131 	return len;
132 
133 unknown:
134 	*target = '?';
135 	len = 1;
136 	return len;
137 }
138 
139 /*
140  * cifs_from_utf16 - convert utf16le string to local charset
141  * @to - destination buffer
142  * @from - source buffer
143  * @tolen - destination buffer size (in bytes)
144  * @fromlen - source buffer size (in bytes)
145  * @codepage - codepage to which characters should be converted
146  * @mapchar - should characters be remapped according to the mapchars option?
147  *
148  * Convert a little-endian utf16le string (as sent by the server) to a string
149  * in the provided codepage. The tolen and fromlen parameters are to ensure
150  * that the code doesn't walk off of the end of the buffer (which is always
151  * a danger if the alignment of the source buffer is off). The destination
152  * string is always properly null terminated and fits in the destination
153  * buffer. Returns the length of the destination string in bytes (including
154  * null terminator).
155  *
156  * Note that some windows versions actually send multiword UTF-16 characters
157  * instead of straight UTF16-2. The linux nls routines however aren't able to
158  * deal with those characters properly. In the event that we get some of
159  * those characters, they won't be translated properly.
160  */
161 int
cifs_from_utf16(char * to,const __le16 * from,int tolen,int fromlen,const struct nls_table * codepage,int map_type)162 cifs_from_utf16(char *to, const __le16 *from, int tolen, int fromlen,
163 		const struct nls_table *codepage, int map_type)
164 {
165 	int i, charlen, safelen;
166 	int outlen = 0;
167 	int nullsize = nls_nullsize(codepage);
168 	int fromwords = fromlen / 2;
169 	char tmp[NLS_MAX_CHARSET_SIZE];
170 	__u16 ftmp[3];		/* ftmp[3] = 3array x 2bytes = 6bytes UTF-16 */
171 
172 	/*
173 	 * because the chars can be of varying widths, we need to take care
174 	 * not to overflow the destination buffer when we get close to the
175 	 * end of it. Until we get to this offset, we don't need to check
176 	 * for overflow however.
177 	 */
178 	safelen = tolen - (NLS_MAX_CHARSET_SIZE + nullsize);
179 
180 	for (i = 0; i < fromwords; i++) {
181 		ftmp[0] = get_unaligned_le16(&from[i]);
182 		if (ftmp[0] == 0)
183 			break;
184 		if (i + 1 < fromwords)
185 			ftmp[1] = get_unaligned_le16(&from[i + 1]);
186 		else
187 			ftmp[1] = 0;
188 		if (i + 2 < fromwords)
189 			ftmp[2] = get_unaligned_le16(&from[i + 2]);
190 		else
191 			ftmp[2] = 0;
192 
193 		/*
194 		 * check to see if converting this character might make the
195 		 * conversion bleed into the null terminator
196 		 */
197 		if (outlen >= safelen) {
198 			charlen = cifs_mapchar(tmp, ftmp, codepage, map_type);
199 			if ((outlen + charlen) > (tolen - nullsize))
200 				break;
201 		}
202 
203 		/* put converted char into 'to' buffer */
204 		charlen = cifs_mapchar(&to[outlen], ftmp, codepage, map_type);
205 		outlen += charlen;
206 
207 		/* charlen (=bytes of UTF-8 for 1 character)
208 		 * 4bytes UTF-8(surrogate pair) is charlen=4
209 		 *   (4bytes UTF-16 code)
210 		 * 7-8bytes UTF-8(IVS) is charlen=3+4 or 4+4
211 		 *   (2 UTF-8 pairs divided to 2 UTF-16 pairs) */
212 		if (charlen == 4)
213 			i++;
214 		else if (charlen >= 5)
215 			/* 5-6bytes UTF-8 */
216 			i += 2;
217 	}
218 
219 	/* properly null-terminate string */
220 	for (i = 0; i < nullsize; i++)
221 		to[outlen++] = 0;
222 
223 	return outlen;
224 }
225 
226 /*
227  * NAME:	cifs_strtoUTF16()
228  *
229  * FUNCTION:	Convert character string to unicode string
230  *
231  */
232 int
cifs_strtoUTF16(__le16 * to,const char * from,int len,const struct nls_table * codepage)233 cifs_strtoUTF16(__le16 *to, const char *from, int len,
234 	      const struct nls_table *codepage)
235 {
236 	int charlen;
237 	int i;
238 	wchar_t wchar_to; /* needed to quiet sparse */
239 
240 	/* special case for utf8 to handle no plane0 chars */
241 	if (!strcmp(codepage->charset, "utf8")) {
242 		/*
243 		 * convert utf8 -> utf16, we assume we have enough space
244 		 * as caller should have assumed conversion does not overflow
245 		 * in destination len is length in wchar_t units (16bits)
246 		 */
247 		i  = utf8s_to_utf16s(from, len, UTF16_LITTLE_ENDIAN,
248 				       (wchar_t *) to, len);
249 
250 		/* if success terminate and exit */
251 		if (i >= 0)
252 			goto success;
253 		/*
254 		 * if fails fall back to UCS encoding as this
255 		 * function should not return negative values
256 		 * currently can fail only if source contains
257 		 * invalid encoded characters
258 		 */
259 	}
260 
261 	for (i = 0; len && *from; i++, from += charlen, len -= charlen) {
262 		charlen = codepage->char2uni(from, len, &wchar_to);
263 		if (charlen < 1) {
264 			cifs_dbg(VFS, "strtoUTF16: char2uni of 0x%x returned %d\n",
265 				 *from, charlen);
266 			/* A question mark */
267 			wchar_to = 0x003f;
268 			charlen = 1;
269 		}
270 		put_unaligned_le16(wchar_to, &to[i]);
271 	}
272 
273 success:
274 	put_unaligned_le16(0, &to[i]);
275 	return i;
276 }
277 
278 /*
279  * cifs_utf16_bytes - how long will a string be after conversion?
280  * @utf16 - pointer to input string
281  * @maxbytes - don't go past this many bytes of input string
282  * @codepage - destination codepage
283  *
284  * Walk a utf16le string and return the number of bytes that the string will
285  * be after being converted to the given charset, not including any null
286  * termination required. Don't walk past maxbytes in the source buffer.
287  */
288 int
cifs_utf16_bytes(const __le16 * from,int maxbytes,const struct nls_table * codepage)289 cifs_utf16_bytes(const __le16 *from, int maxbytes,
290 		const struct nls_table *codepage)
291 {
292 	int i;
293 	int charlen, outlen = 0;
294 	int maxwords = maxbytes / 2;
295 	char tmp[NLS_MAX_CHARSET_SIZE];
296 	__u16 ftmp[3];
297 
298 	for (i = 0; i < maxwords; i++) {
299 		ftmp[0] = get_unaligned_le16(&from[i]);
300 		if (ftmp[0] == 0)
301 			break;
302 		if (i + 1 < maxwords)
303 			ftmp[1] = get_unaligned_le16(&from[i + 1]);
304 		else
305 			ftmp[1] = 0;
306 		if (i + 2 < maxwords)
307 			ftmp[2] = get_unaligned_le16(&from[i + 2]);
308 		else
309 			ftmp[2] = 0;
310 
311 		charlen = cifs_mapchar(tmp, ftmp, codepage, NO_MAP_UNI_RSVD);
312 		outlen += charlen;
313 	}
314 
315 	return outlen;
316 }
317 
318 /*
319  * cifs_strndup_from_utf16 - copy a string from wire format to the local
320  * codepage
321  * @src - source string
322  * @maxlen - don't walk past this many bytes in the source string
323  * @is_unicode - is this a unicode string?
324  * @codepage - destination codepage
325  *
326  * Take a string given by the server, convert it to the local codepage and
327  * put it in a new buffer. Returns a pointer to the new string or NULL on
328  * error.
329  */
330 char *
cifs_strndup_from_utf16(const char * src,const int maxlen,const bool is_unicode,const struct nls_table * codepage)331 cifs_strndup_from_utf16(const char *src, const int maxlen,
332 			const bool is_unicode, const struct nls_table *codepage)
333 {
334 	int len;
335 	char *dst;
336 
337 	if (is_unicode) {
338 		len = cifs_utf16_bytes((__le16 *) src, maxlen, codepage);
339 		len += nls_nullsize(codepage);
340 		dst = kmalloc(len, GFP_KERNEL);
341 		if (!dst)
342 			return NULL;
343 		cifs_from_utf16(dst, (__le16 *) src, len, maxlen, codepage,
344 				NO_MAP_UNI_RSVD);
345 	} else {
346 		dst = kstrndup(src, maxlen, GFP_KERNEL);
347 	}
348 
349 	return dst;
350 }
351 
convert_to_sfu_char(char src_char)352 static __le16 convert_to_sfu_char(char src_char)
353 {
354 	__le16 dest_char;
355 
356 	switch (src_char) {
357 	case ':':
358 		dest_char = cpu_to_le16(UNI_COLON);
359 		break;
360 	case '*':
361 		dest_char = cpu_to_le16(UNI_ASTERISK);
362 		break;
363 	case '?':
364 		dest_char = cpu_to_le16(UNI_QUESTION);
365 		break;
366 	case '<':
367 		dest_char = cpu_to_le16(UNI_LESSTHAN);
368 		break;
369 	case '>':
370 		dest_char = cpu_to_le16(UNI_GRTRTHAN);
371 		break;
372 	case '|':
373 		dest_char = cpu_to_le16(UNI_PIPE);
374 		break;
375 	default:
376 		dest_char = 0;
377 	}
378 
379 	return dest_char;
380 }
381 
convert_to_sfm_char(char src_char,bool end_of_string)382 static __le16 convert_to_sfm_char(char src_char, bool end_of_string)
383 {
384 	__le16 dest_char;
385 
386 	if (src_char >= 0x01 && src_char <= 0x1F) {
387 		dest_char = cpu_to_le16(src_char + 0xF000);
388 		return dest_char;
389 	}
390 	switch (src_char) {
391 	case ':':
392 		dest_char = cpu_to_le16(SFM_COLON);
393 		break;
394 	case '"':
395 		dest_char = cpu_to_le16(SFM_DOUBLEQUOTE);
396 		break;
397 	case '*':
398 		dest_char = cpu_to_le16(SFM_ASTERISK);
399 		break;
400 	case '?':
401 		dest_char = cpu_to_le16(SFM_QUESTION);
402 		break;
403 	case '<':
404 		dest_char = cpu_to_le16(SFM_LESSTHAN);
405 		break;
406 	case '>':
407 		dest_char = cpu_to_le16(SFM_GRTRTHAN);
408 		break;
409 	case '|':
410 		dest_char = cpu_to_le16(SFM_PIPE);
411 		break;
412 	case '.':
413 		if (end_of_string)
414 			dest_char = cpu_to_le16(SFM_PERIOD);
415 		else
416 			dest_char = 0;
417 		break;
418 	case ' ':
419 		if (end_of_string)
420 			dest_char = cpu_to_le16(SFM_SPACE);
421 		else
422 			dest_char = 0;
423 		break;
424 	default:
425 		dest_char = 0;
426 	}
427 
428 	return dest_char;
429 }
430 
431 /*
432  * Convert 16 bit Unicode pathname to wire format from string in current code
433  * page. Conversion may involve remapping up the six characters that are
434  * only legal in POSIX-like OS (if they are present in the string). Path
435  * names are little endian 16 bit Unicode on the wire
436  */
437 int
cifsConvertToUTF16(__le16 * target,const char * source,int srclen,const struct nls_table * cp,int map_chars)438 cifsConvertToUTF16(__le16 *target, const char *source, int srclen,
439 		 const struct nls_table *cp, int map_chars)
440 {
441 	int i, charlen;
442 	int j = 0;
443 	char src_char;
444 	__le16 dst_char;
445 	wchar_t tmp;
446 	wchar_t *wchar_to;	/* UTF-16 */
447 	int ret;
448 	unicode_t u;
449 
450 	if (map_chars == NO_MAP_UNI_RSVD)
451 		return cifs_strtoUTF16(target, source, PATH_MAX, cp);
452 
453 	wchar_to = kzalloc(6, GFP_KERNEL);
454 
455 	for (i = 0; i < srclen; j++) {
456 		src_char = source[i];
457 		charlen = 1;
458 
459 		/* check if end of string */
460 		if (src_char == 0)
461 			goto ctoUTF16_out;
462 
463 		/* see if we must remap this char */
464 		if (map_chars == SFU_MAP_UNI_RSVD)
465 			dst_char = convert_to_sfu_char(src_char);
466 		else if (map_chars == SFM_MAP_UNI_RSVD) {
467 			bool end_of_string;
468 
469 			/**
470 			 * Remap spaces and periods found at the end of every
471 			 * component of the path. The special cases of '.' and
472 			 * '..' are need to be handled because of symlinks.
473 			 * They are treated as non-end-of-string to avoid
474 			 * remapping and breaking symlinks pointing to . or ..
475 			 **/
476 			if ((i == 0 || source[i-1] == '\\') &&
477 			    source[i] == '.' &&
478 			    (i == srclen-1 || source[i+1] == '\\'))
479 				end_of_string = false; /* "." case */
480 			else if (i >= 1 &&
481 				 (i == 1 || source[i-2] == '\\') &&
482 				 source[i-1] == '.' &&
483 				 source[i] == '.' &&
484 				 (i == srclen-1 || source[i+1] == '\\'))
485 				end_of_string = false; /* ".." case */
486 			else if ((i == srclen - 1) || (source[i+1] == '\\'))
487 				end_of_string = true;
488 			else
489 				end_of_string = false;
490 
491 			dst_char = convert_to_sfm_char(src_char, end_of_string);
492 		} else
493 			dst_char = 0;
494 		/*
495 		 * FIXME: We can not handle remapping backslash (UNI_SLASH)
496 		 * until all the calls to build_path_from_dentry are modified,
497 		 * as they use backslash as separator.
498 		 */
499 		if (dst_char == 0) {
500 			charlen = cp->char2uni(source + i, srclen - i, &tmp);
501 			dst_char = cpu_to_le16(tmp);
502 
503 			/*
504 			 * if no match, use question mark, which at least in
505 			 * some cases serves as wild card
506 			 */
507 			if (charlen > 0)
508 				goto ctoUTF16;
509 
510 			/* convert SURROGATE_PAIR */
511 			if (strcmp(cp->charset, "utf8") || !wchar_to)
512 				goto unknown;
513 			if (*(source + i) & 0x80) {
514 				charlen = utf8_to_utf32(source + i, 6, &u);
515 				if (charlen < 0)
516 					goto unknown;
517 			} else
518 				goto unknown;
519 			ret  = utf8s_to_utf16s(source + i, charlen,
520 					       UTF16_LITTLE_ENDIAN,
521 					       wchar_to, 6);
522 			if (ret < 0)
523 				goto unknown;
524 
525 			i += charlen;
526 			dst_char = cpu_to_le16(*wchar_to);
527 			if (charlen <= 3)
528 				/* 1-3bytes UTF-8 to 2bytes UTF-16 */
529 				put_unaligned(dst_char, &target[j]);
530 			else if (charlen == 4) {
531 				/* 4bytes UTF-8(surrogate pair) to 4bytes UTF-16
532 				 * 7-8bytes UTF-8(IVS) divided to 2 UTF-16
533 				 *   (charlen=3+4 or 4+4) */
534 				put_unaligned(dst_char, &target[j]);
535 				dst_char = cpu_to_le16(*(wchar_to + 1));
536 				j++;
537 				put_unaligned(dst_char, &target[j]);
538 			} else if (charlen >= 5) {
539 				/* 5-6bytes UTF-8 to 6bytes UTF-16 */
540 				put_unaligned(dst_char, &target[j]);
541 				dst_char = cpu_to_le16(*(wchar_to + 1));
542 				j++;
543 				put_unaligned(dst_char, &target[j]);
544 				dst_char = cpu_to_le16(*(wchar_to + 2));
545 				j++;
546 				put_unaligned(dst_char, &target[j]);
547 			}
548 			continue;
549 
550 unknown:
551 			dst_char = cpu_to_le16(0x003f);
552 			charlen = 1;
553 		}
554 
555 ctoUTF16:
556 		/*
557 		 * character may take more than one byte in the source string,
558 		 * but will take exactly two bytes in the target string
559 		 */
560 		i += charlen;
561 		put_unaligned(dst_char, &target[j]);
562 	}
563 
564 ctoUTF16_out:
565 	put_unaligned(0, &target[j]); /* Null terminate target unicode string */
566 	kfree(wchar_to);
567 	return j;
568 }
569 
570 /*
571  * cifs_local_to_utf16_bytes - how long will a string be after conversion?
572  * @from - pointer to input string
573  * @maxbytes - don't go past this many bytes of input string
574  * @codepage - source codepage
575  *
576  * Walk a string and return the number of bytes that the string will
577  * be after being converted to the given charset, not including any null
578  * termination required. Don't walk past maxbytes in the source buffer.
579  */
580 
581 static int
cifs_local_to_utf16_bytes(const char * from,int len,const struct nls_table * codepage)582 cifs_local_to_utf16_bytes(const char *from, int len,
583 			  const struct nls_table *codepage)
584 {
585 	int charlen;
586 	int i;
587 	wchar_t wchar_to;
588 
589 	for (i = 0; len && *from; i++, from += charlen, len -= charlen) {
590 		charlen = codepage->char2uni(from, len, &wchar_to);
591 		/* Failed conversion defaults to a question mark */
592 		if (charlen < 1)
593 			charlen = 1;
594 	}
595 	return 2 * i; /* UTF16 characters are two bytes */
596 }
597 
598 /*
599  * cifs_strndup_to_utf16 - copy a string to wire format from the local codepage
600  * @src - source string
601  * @maxlen - don't walk past this many bytes in the source string
602  * @utf16_len - the length of the allocated string in bytes (including null)
603  * @cp - source codepage
604  * @remap - map special chars
605  *
606  * Take a string convert it from the local codepage to UTF16 and
607  * put it in a new buffer. Returns a pointer to the new string or NULL on
608  * error.
609  */
610 __le16 *
cifs_strndup_to_utf16(const char * src,const int maxlen,int * utf16_len,const struct nls_table * cp,int remap)611 cifs_strndup_to_utf16(const char *src, const int maxlen, int *utf16_len,
612 		      const struct nls_table *cp, int remap)
613 {
614 	int len;
615 	__le16 *dst;
616 
617 	if (!src)
618 		return NULL;
619 
620 	len = cifs_local_to_utf16_bytes(src, maxlen, cp);
621 	len += 2; /* NULL */
622 	dst = kmalloc(len, GFP_KERNEL);
623 	if (!dst) {
624 		*utf16_len = 0;
625 		return NULL;
626 	}
627 	cifsConvertToUTF16(dst, src, strlen(src), cp, remap);
628 	*utf16_len = len;
629 	return dst;
630 }
631