1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3 * fs/cifs/cifs_unicode.c
4 *
5 * Copyright (c) International Business Machines Corp., 2000,2009
6 * Modified by Steve French (sfrench@us.ibm.com)
7 */
8 #include <linux/fs.h>
9 #include <linux/slab.h>
10 #include "cifs_fs_sb.h"
11 #include "cifs_unicode.h"
12 #include "cifs_uniupr.h"
13 #include "cifspdu.h"
14 #include "cifsglob.h"
15 #include "cifs_debug.h"
16
17 int cifs_remap(struct cifs_sb_info *cifs_sb)
18 {
19 int map_type;
20
21 if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_MAP_SFM_CHR)
22 map_type = SFM_MAP_UNI_RSVD;
23 else if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_MAP_SPECIAL_CHR)
24 map_type = SFU_MAP_UNI_RSVD;
25 else
26 map_type = NO_MAP_UNI_RSVD;
27
28 return map_type;
29 }
30
31 /* Convert character using the SFU - "Services for Unix" remapping range */
32 static bool
33 convert_sfu_char(const __u16 src_char, char *target)
34 {
35 /*
36 * BB: Cannot handle remapping UNI_SLASH until all the calls to
37 * build_path_from_dentry are modified, as they use slash as
38 * separator.
39 */
40 switch (src_char) {
41 case UNI_COLON:
42 *target = ':';
43 break;
44 case UNI_ASTERISK:
45 *target = '*';
46 break;
47 case UNI_QUESTION:
48 *target = '?';
49 break;
50 case UNI_PIPE:
51 *target = '|';
52 break;
53 case UNI_GRTRTHAN:
54 *target = '>';
55 break;
56 case UNI_LESSTHAN:
57 *target = '<';
58 break;
59 default:
60 return false;
61 }
62 return true;
63 }
64
65 /* Convert character using the SFM - "Services for Mac" remapping range */
66 static bool
67 convert_sfm_char(const __u16 src_char, char *target)
68 {
69 if (src_char >= 0xF001 && src_char <= 0xF01F) {
70 *target = src_char - 0xF000;
71 return true;
72 }
73 switch (src_char) {
74 case SFM_COLON:
75 *target = ':';
76 break;
77 case SFM_DOUBLEQUOTE:
78 *target = '"';
79 break;
80 case SFM_ASTERISK:
81 *target = '*';
82 break;
83 case SFM_QUESTION:
84 *target = '?';
85 break;
86 case SFM_PIPE:
87 *target = '|';
88 break;
89 case SFM_GRTRTHAN:
90 *target = '>';
91 break;
92 case SFM_LESSTHAN:
93 *target = '<';
94 break;
95 case SFM_SPACE:
96 *target = ' ';
97 break;
98 case SFM_PERIOD:
99 *target = '.';
100 break;
101 default:
102 return false;
103 }
104 return true;
105 }
106
107
108 /*
109 * cifs_mapchar - convert a host-endian char to proper char in codepage
110 * @target - where converted character should be copied
111 * @src_char - 2 byte host-endian source character
112 * @cp - codepage to which character should be converted
113 * @map_type - How should the 7 NTFS/SMB reserved characters be mapped to UCS2?
114 *
115 * This function handles the conversion of a single character. It is the
116 * responsibility of the caller to ensure that the target buffer is large
117 * enough to hold the result of the conversion (at least NLS_MAX_CHARSET_SIZE).
118 */
119 static int
120 cifs_mapchar(char *target, const __u16 *from, const struct nls_table *cp,
121 int maptype)
122 {
123 int len = 1;
124 __u16 src_char;
125
126 src_char = *from;
127
128 if ((maptype == SFM_MAP_UNI_RSVD) && convert_sfm_char(src_char, target))
129 return len;
130 else if ((maptype == SFU_MAP_UNI_RSVD) &&
131 convert_sfu_char(src_char, target))
132 return len;
133
134 /* if character not one of seven in special remap set */
135 len = cp->uni2char(src_char, target, NLS_MAX_CHARSET_SIZE);
136 if (len <= 0)
137 goto surrogate_pair;
138
139 return len;
140
141 surrogate_pair:
142 /* convert SURROGATE_PAIR and IVS */
143 if (strcmp(cp->charset, "utf8"))
144 goto unknown;
145 len = utf16s_to_utf8s(from, 3, UTF16_LITTLE_ENDIAN, target, 6);
146 if (len <= 0)
147 goto unknown;
148 return len;
149
150 unknown:
151 *target = '?';
152 len = 1;
153 return len;
154 }
155
156 /*
157 * cifs_from_utf16 - convert utf16le string to local charset
158 * @to - destination buffer
159 * @from - source buffer
160 * @tolen - destination buffer size (in bytes)
161 * @fromlen - source buffer size (in bytes)
162 * @codepage - codepage to which characters should be converted
163 * @mapchar - should characters be remapped according to the mapchars option?
164 *
165 * Convert a little-endian utf16le string (as sent by the server) to a string
166 * in the provided codepage. The tolen and fromlen parameters are to ensure
167 * that the code doesn't walk off of the end of the buffer (which is always
168 * a danger if the alignment of the source buffer is off). The destination
169 * string is always properly null terminated and fits in the destination
170 * buffer. Returns the length of the destination string in bytes (including
171 * null terminator).
172 *
173 * Note that some windows versions actually send multiword UTF-16 characters
174 * instead of straight UTF16-2. The linux nls routines however aren't able to
175 * deal with those characters properly. In the event that we get some of
176 * those characters, they won't be translated properly.
177 */
178 int
179 cifs_from_utf16(char *to, const __le16 *from, int tolen, int fromlen,
180 const struct nls_table *codepage, int map_type)
181 {
182 int i, charlen, safelen;
183 int outlen = 0;
184 int nullsize = nls_nullsize(codepage);
185 int fromwords = fromlen / 2;
186 char tmp[NLS_MAX_CHARSET_SIZE];
187 __u16 ftmp[3]; /* ftmp[3] = 3array x 2bytes = 6bytes UTF-16 */
188
189 /*
190 * because the chars can be of varying widths, we need to take care
191 * not to overflow the destination buffer when we get close to the
192 * end of it. Until we get to this offset, we don't need to check
193 * for overflow however.
194 */
195 safelen = tolen - (NLS_MAX_CHARSET_SIZE + nullsize);
196
197 for (i = 0; i < fromwords; i++) {
198 ftmp[0] = get_unaligned_le16(&from[i]);
199 if (ftmp[0] == 0)
200 break;
201 if (i + 1 < fromwords)
202 ftmp[1] = get_unaligned_le16(&from[i + 1]);
203 else
204 ftmp[1] = 0;
205 if (i + 2 < fromwords)
206 ftmp[2] = get_unaligned_le16(&from[i + 2]);
207 else
208 ftmp[2] = 0;
209
210 /*
211 * check to see if converting this character might make the
212 * conversion bleed into the null terminator
213 */
214 if (outlen >= safelen) {
215 charlen = cifs_mapchar(tmp, ftmp, codepage, map_type);
216 if ((outlen + charlen) > (tolen - nullsize))
217 break;
218 }
219
220 /* put converted char into 'to' buffer */
221 charlen = cifs_mapchar(&to[outlen], ftmp, codepage, map_type);
222 outlen += charlen;
223
224 /* charlen (=bytes of UTF-8 for 1 character)
225 * 4bytes UTF-8(surrogate pair) is charlen=4
226 * (4bytes UTF-16 code)
227 * 7-8bytes UTF-8(IVS) is charlen=3+4 or 4+4
228 * (2 UTF-8 pairs divided to 2 UTF-16 pairs) */
229 if (charlen == 4)
230 i++;
231 else if (charlen >= 5)
232 /* 5-6bytes UTF-8 */
233 i += 2;
234 }
235
236 /* properly null-terminate string */
237 for (i = 0; i < nullsize; i++)
238 to[outlen++] = 0;
239
240 return outlen;
241 }
242
243 /*
244 * NAME: cifs_strtoUTF16()
245 *
246 * FUNCTION: Convert character string to unicode string
247 *
248 */
249 int
250 cifs_strtoUTF16(__le16 *to, const char *from, int len,
251 const struct nls_table *codepage)
252 {
253 int charlen;
254 int i;
255 wchar_t wchar_to; /* needed to quiet sparse */
256
257 /* special case for utf8 to handle no plane0 chars */
258 if (!strcmp(codepage->charset, "utf8")) {
259 /*
260 * convert utf8 -> utf16, we assume we have enough space
261 * as caller should have assumed conversion does not overflow
262 * in destination len is length in wchar_t units (16bits)
263 */
264 i = utf8s_to_utf16s(from, len, UTF16_LITTLE_ENDIAN,
265 (wchar_t *) to, len);
266
267 /* if success terminate and exit */
268 if (i >= 0)
269 goto success;
270 /*
271 * if fails fall back to UCS encoding as this
272 * function should not return negative values
273 * currently can fail only if source contains
274 * invalid encoded characters
275 */
276 }
277
278 for (i = 0; len && *from; i++, from += charlen, len -= charlen) {
279 charlen = codepage->char2uni(from, len, &wchar_to);
280 if (charlen < 1) {
281 cifs_dbg(VFS, "strtoUTF16: char2uni of 0x%x returned %d\n",
282 *from, charlen);
283 /* A question mark */
284 wchar_to = 0x003f;
285 charlen = 1;
286 }
287 put_unaligned_le16(wchar_to, &to[i]);
288 }
289
290 success:
291 put_unaligned_le16(0, &to[i]);
292 return i;
293 }
294
295 /*
296 * cifs_utf16_bytes - how long will a string be after conversion?
297 * @utf16 - pointer to input string
298 * @maxbytes - don't go past this many bytes of input string
299 * @codepage - destination codepage
300 *
301 * Walk a utf16le string and return the number of bytes that the string will
302 * be after being converted to the given charset, not including any null
303 * termination required. Don't walk past maxbytes in the source buffer.
304 */
305 int
306 cifs_utf16_bytes(const __le16 *from, int maxbytes,
307 const struct nls_table *codepage)
308 {
309 int i;
310 int charlen, outlen = 0;
311 int maxwords = maxbytes / 2;
312 char tmp[NLS_MAX_CHARSET_SIZE];
313 __u16 ftmp[3];
314
315 for (i = 0; i < maxwords; i++) {
316 ftmp[0] = get_unaligned_le16(&from[i]);
317 if (ftmp[0] == 0)
318 break;
319 if (i + 1 < maxwords)
320 ftmp[1] = get_unaligned_le16(&from[i + 1]);
321 else
322 ftmp[1] = 0;
323 if (i + 2 < maxwords)
324 ftmp[2] = get_unaligned_le16(&from[i + 2]);
325 else
326 ftmp[2] = 0;
327
328 charlen = cifs_mapchar(tmp, ftmp, codepage, NO_MAP_UNI_RSVD);
329 outlen += charlen;
330 }
331
332 return outlen;
333 }
334
335 /*
336 * cifs_strndup_from_utf16 - copy a string from wire format to the local
337 * codepage
338 * @src - source string
339 * @maxlen - don't walk past this many bytes in the source string
340 * @is_unicode - is this a unicode string?
341 * @codepage - destination codepage
342 *
343 * Take a string given by the server, convert it to the local codepage and
344 * put it in a new buffer. Returns a pointer to the new string or NULL on
345 * error.
346 */
347 char *
348 cifs_strndup_from_utf16(const char *src, const int maxlen,
349 const bool is_unicode, const struct nls_table *codepage)
350 {
351 int len;
352 char *dst;
353
354 if (is_unicode) {
355 len = cifs_utf16_bytes((__le16 *) src, maxlen, codepage);
356 len += nls_nullsize(codepage);
357 dst = kmalloc(len, GFP_KERNEL);
358 if (!dst)
359 return NULL;
360 cifs_from_utf16(dst, (__le16 *) src, len, maxlen, codepage,
361 NO_MAP_UNI_RSVD);
362 } else {
363 len = strnlen(src, maxlen);
364 len++;
365 dst = kmalloc(len, GFP_KERNEL);
366 if (!dst)
367 return NULL;
368 strlcpy(dst, src, len);
369 }
370
371 return dst;
372 }
373
374 static __le16 convert_to_sfu_char(char src_char)
375 {
376 __le16 dest_char;
377
378 switch (src_char) {
379 case ':':
380 dest_char = cpu_to_le16(UNI_COLON);
381 break;
382 case '*':
383 dest_char = cpu_to_le16(UNI_ASTERISK);
384 break;
385 case '?':
386 dest_char = cpu_to_le16(UNI_QUESTION);
387 break;
388 case '<':
389 dest_char = cpu_to_le16(UNI_LESSTHAN);
390 break;
391 case '>':
392 dest_char = cpu_to_le16(UNI_GRTRTHAN);
393 break;
394 case '|':
395 dest_char = cpu_to_le16(UNI_PIPE);
396 break;
397 default:
398 dest_char = 0;
399 }
400
401 return dest_char;
402 }
403
404 static __le16 convert_to_sfm_char(char src_char, bool end_of_string)
405 {
406 __le16 dest_char;
407
408 if (src_char >= 0x01 && src_char <= 0x1F) {
409 dest_char = cpu_to_le16(src_char + 0xF000);
410 return dest_char;
411 }
412 switch (src_char) {
413 case ':':
414 dest_char = cpu_to_le16(SFM_COLON);
415 break;
416 case '"':
417 dest_char = cpu_to_le16(SFM_DOUBLEQUOTE);
418 break;
419 case '*':
420 dest_char = cpu_to_le16(SFM_ASTERISK);
421 break;
422 case '?':
423 dest_char = cpu_to_le16(SFM_QUESTION);
424 break;
425 case '<':
426 dest_char = cpu_to_le16(SFM_LESSTHAN);
427 break;
428 case '>':
429 dest_char = cpu_to_le16(SFM_GRTRTHAN);
430 break;
431 case '|':
432 dest_char = cpu_to_le16(SFM_PIPE);
433 break;
434 case '.':
435 if (end_of_string)
436 dest_char = cpu_to_le16(SFM_PERIOD);
437 else
438 dest_char = 0;
439 break;
440 case ' ':
441 if (end_of_string)
442 dest_char = cpu_to_le16(SFM_SPACE);
443 else
444 dest_char = 0;
445 break;
446 default:
447 dest_char = 0;
448 }
449
450 return dest_char;
451 }
452
453 /*
454 * Convert 16 bit Unicode pathname to wire format from string in current code
455 * page. Conversion may involve remapping up the six characters that are
456 * only legal in POSIX-like OS (if they are present in the string). Path
457 * names are little endian 16 bit Unicode on the wire
458 */
459 int
460 cifsConvertToUTF16(__le16 *target, const char *source, int srclen,
461 const struct nls_table *cp, int map_chars)
462 {
463 int i, charlen;
464 int j = 0;
465 char src_char;
466 __le16 dst_char;
467 wchar_t tmp;
468 wchar_t *wchar_to; /* UTF-16 */
469 int ret;
470 unicode_t u;
471
472 if (map_chars == NO_MAP_UNI_RSVD)
473 return cifs_strtoUTF16(target, source, PATH_MAX, cp);
474
475 wchar_to = kzalloc(6, GFP_KERNEL);
476
477 for (i = 0; i < srclen; j++) {
478 src_char = source[i];
479 charlen = 1;
480
481 /* check if end of string */
482 if (src_char == 0)
483 goto ctoUTF16_out;
484
485 /* see if we must remap this char */
486 if (map_chars == SFU_MAP_UNI_RSVD)
487 dst_char = convert_to_sfu_char(src_char);
488 else if (map_chars == SFM_MAP_UNI_RSVD) {
489 bool end_of_string;
490
491 if (i == srclen - 1)
492 end_of_string = true;
493 else
494 end_of_string = false;
495
496 dst_char = convert_to_sfm_char(src_char, end_of_string);
497 } else
498 dst_char = 0;
499 /*
500 * FIXME: We can not handle remapping backslash (UNI_SLASH)
501 * until all the calls to build_path_from_dentry are modified,
502 * as they use backslash as separator.
503 */
504 if (dst_char == 0) {
505 charlen = cp->char2uni(source + i, srclen - i, &tmp);
506 dst_char = cpu_to_le16(tmp);
507
508 /*
509 * if no match, use question mark, which at least in
510 * some cases serves as wild card
511 */
512 if (charlen > 0)
513 goto ctoUTF16;
514
515 /* convert SURROGATE_PAIR */
516 if (strcmp(cp->charset, "utf8") || !wchar_to)
517 goto unknown;
518 if (*(source + i) & 0x80) {
519 charlen = utf8_to_utf32(source + i, 6, &u);
520 if (charlen < 0)
521 goto unknown;
522 } else
523 goto unknown;
524 ret = utf8s_to_utf16s(source + i, charlen,
525 UTF16_LITTLE_ENDIAN,
526 wchar_to, 6);
527 if (ret < 0)
528 goto unknown;
529
530 i += charlen;
531 dst_char = cpu_to_le16(*wchar_to);
532 if (charlen <= 3)
533 /* 1-3bytes UTF-8 to 2bytes UTF-16 */
534 put_unaligned(dst_char, &target[j]);
535 else if (charlen == 4) {
536 /* 4bytes UTF-8(surrogate pair) to 4bytes UTF-16
537 * 7-8bytes UTF-8(IVS) divided to 2 UTF-16
538 * (charlen=3+4 or 4+4) */
539 put_unaligned(dst_char, &target[j]);
540 dst_char = cpu_to_le16(*(wchar_to + 1));
541 j++;
542 put_unaligned(dst_char, &target[j]);
543 } else if (charlen >= 5) {
544 /* 5-6bytes UTF-8 to 6bytes UTF-16 */
545 put_unaligned(dst_char, &target[j]);
546 dst_char = cpu_to_le16(*(wchar_to + 1));
547 j++;
548 put_unaligned(dst_char, &target[j]);
549 dst_char = cpu_to_le16(*(wchar_to + 2));
550 j++;
551 put_unaligned(dst_char, &target[j]);
552 }
553 continue;
554
555 unknown:
556 dst_char = cpu_to_le16(0x003f);
557 charlen = 1;
558 }
559
560 ctoUTF16:
561 /*
562 * character may take more than one byte in the source string,
563 * but will take exactly two bytes in the target string
564 */
565 i += charlen;
566 put_unaligned(dst_char, &target[j]);
567 }
568
569 ctoUTF16_out:
570 put_unaligned(0, &target[j]); /* Null terminate target unicode string */
571 kfree(wchar_to);
572 return j;
573 }
574
575 /*
576 * cifs_local_to_utf16_bytes - how long will a string be after conversion?
577 * @from - pointer to input string
578 * @maxbytes - don't go past this many bytes of input string
579 * @codepage - source codepage
580 *
581 * Walk a string and return the number of bytes that the string will
582 * be after being converted to the given charset, not including any null
583 * termination required. Don't walk past maxbytes in the source buffer.
584 */
585
586 static int
587 cifs_local_to_utf16_bytes(const char *from, int len,
588 const struct nls_table *codepage)
589 {
590 int charlen;
591 int i;
592 wchar_t wchar_to;
593
594 for (i = 0; len && *from; i++, from += charlen, len -= charlen) {
595 charlen = codepage->char2uni(from, len, &wchar_to);
596 /* Failed conversion defaults to a question mark */
597 if (charlen < 1)
598 charlen = 1;
599 }
600 return 2 * i; /* UTF16 characters are two bytes */
601 }
602
603 /*
604 * cifs_strndup_to_utf16 - copy a string to wire format from the local codepage
605 * @src - source string
606 * @maxlen - don't walk past this many bytes in the source string
607 * @utf16_len - the length of the allocated string in bytes (including null)
608 * @cp - source codepage
609 * @remap - map special chars
610 *
611 * Take a string convert it from the local codepage to UTF16 and
612 * put it in a new buffer. Returns a pointer to the new string or NULL on
613 * error.
614 */
615 __le16 *
616 cifs_strndup_to_utf16(const char *src, const int maxlen, int *utf16_len,
617 const struct nls_table *cp, int remap)
618 {
619 int len;
620 __le16 *dst;
621
622 len = cifs_local_to_utf16_bytes(src, maxlen, cp);
623 len += 2; /* NULL */
624 dst = kmalloc(len, GFP_KERNEL);
625 if (!dst) {
626 *utf16_len = 0;
627 return NULL;
628 }
629 cifsConvertToUTF16(dst, src, strlen(src), cp, remap);
630 *utf16_len = len;
631 return dst;
632 }