cifs_unicode.c revision f980d055
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
17int 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 */
32static bool
33convert_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 */
66static bool
67convert_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 */
119static int
120cifs_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
141surrogate_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
150unknown:
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 */
178int
179cifs_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 */
249int
250cifs_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
290success:
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 */
305int
306cifs_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 */
347char *
348cifs_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		dst = kstrndup(src, maxlen, GFP_KERNEL);
364	}
365
366	return dst;
367}
368
369static __le16 convert_to_sfu_char(char src_char)
370{
371	__le16 dest_char;
372
373	switch (src_char) {
374	case ':':
375		dest_char = cpu_to_le16(UNI_COLON);
376		break;
377	case '*':
378		dest_char = cpu_to_le16(UNI_ASTERISK);
379		break;
380	case '?':
381		dest_char = cpu_to_le16(UNI_QUESTION);
382		break;
383	case '<':
384		dest_char = cpu_to_le16(UNI_LESSTHAN);
385		break;
386	case '>':
387		dest_char = cpu_to_le16(UNI_GRTRTHAN);
388		break;
389	case '|':
390		dest_char = cpu_to_le16(UNI_PIPE);
391		break;
392	default:
393		dest_char = 0;
394	}
395
396	return dest_char;
397}
398
399static __le16 convert_to_sfm_char(char src_char, bool end_of_string)
400{
401	__le16 dest_char;
402
403	if (src_char >= 0x01 && src_char <= 0x1F) {
404		dest_char = cpu_to_le16(src_char + 0xF000);
405		return dest_char;
406	}
407	switch (src_char) {
408	case ':':
409		dest_char = cpu_to_le16(SFM_COLON);
410		break;
411	case '"':
412		dest_char = cpu_to_le16(SFM_DOUBLEQUOTE);
413		break;
414	case '*':
415		dest_char = cpu_to_le16(SFM_ASTERISK);
416		break;
417	case '?':
418		dest_char = cpu_to_le16(SFM_QUESTION);
419		break;
420	case '<':
421		dest_char = cpu_to_le16(SFM_LESSTHAN);
422		break;
423	case '>':
424		dest_char = cpu_to_le16(SFM_GRTRTHAN);
425		break;
426	case '|':
427		dest_char = cpu_to_le16(SFM_PIPE);
428		break;
429	case '.':
430		if (end_of_string)
431			dest_char = cpu_to_le16(SFM_PERIOD);
432		else
433			dest_char = 0;
434		break;
435	case ' ':
436		if (end_of_string)
437			dest_char = cpu_to_le16(SFM_SPACE);
438		else
439			dest_char = 0;
440		break;
441	default:
442		dest_char = 0;
443	}
444
445	return dest_char;
446}
447
448/*
449 * Convert 16 bit Unicode pathname to wire format from string in current code
450 * page. Conversion may involve remapping up the six characters that are
451 * only legal in POSIX-like OS (if they are present in the string). Path
452 * names are little endian 16 bit Unicode on the wire
453 */
454int
455cifsConvertToUTF16(__le16 *target, const char *source, int srclen,
456		 const struct nls_table *cp, int map_chars)
457{
458	int i, charlen;
459	int j = 0;
460	char src_char;
461	__le16 dst_char;
462	wchar_t tmp;
463	wchar_t *wchar_to;	/* UTF-16 */
464	int ret;
465	unicode_t u;
466
467	if (map_chars == NO_MAP_UNI_RSVD)
468		return cifs_strtoUTF16(target, source, PATH_MAX, cp);
469
470	wchar_to = kzalloc(6, GFP_KERNEL);
471
472	for (i = 0; i < srclen; j++) {
473		src_char = source[i];
474		charlen = 1;
475
476		/* check if end of string */
477		if (src_char == 0)
478			goto ctoUTF16_out;
479
480		/* see if we must remap this char */
481		if (map_chars == SFU_MAP_UNI_RSVD)
482			dst_char = convert_to_sfu_char(src_char);
483		else if (map_chars == SFM_MAP_UNI_RSVD) {
484			bool end_of_string;
485
486			/**
487			 * Remap spaces and periods found at the end of every
488			 * component of the path. The special cases of '.' and
489			 * '..' do not need to be dealt with explicitly because
490			 * they are addressed in namei.c:link_path_walk().
491			 **/
492			if ((i == srclen - 1) || (source[i+1] == '\\'))
493				end_of_string = true;
494			else
495				end_of_string = false;
496
497			dst_char = convert_to_sfm_char(src_char, end_of_string);
498		} else
499			dst_char = 0;
500		/*
501		 * FIXME: We can not handle remapping backslash (UNI_SLASH)
502		 * until all the calls to build_path_from_dentry are modified,
503		 * as they use backslash as separator.
504		 */
505		if (dst_char == 0) {
506			charlen = cp->char2uni(source + i, srclen - i, &tmp);
507			dst_char = cpu_to_le16(tmp);
508
509			/*
510			 * if no match, use question mark, which at least in
511			 * some cases serves as wild card
512			 */
513			if (charlen > 0)
514				goto ctoUTF16;
515
516			/* convert SURROGATE_PAIR */
517			if (strcmp(cp->charset, "utf8") || !wchar_to)
518				goto unknown;
519			if (*(source + i) & 0x80) {
520				charlen = utf8_to_utf32(source + i, 6, &u);
521				if (charlen < 0)
522					goto unknown;
523			} else
524				goto unknown;
525			ret  = utf8s_to_utf16s(source + i, charlen,
526					       UTF16_LITTLE_ENDIAN,
527					       wchar_to, 6);
528			if (ret < 0)
529				goto unknown;
530
531			i += charlen;
532			dst_char = cpu_to_le16(*wchar_to);
533			if (charlen <= 3)
534				/* 1-3bytes UTF-8 to 2bytes UTF-16 */
535				put_unaligned(dst_char, &target[j]);
536			else if (charlen == 4) {
537				/* 4bytes UTF-8(surrogate pair) to 4bytes UTF-16
538				 * 7-8bytes UTF-8(IVS) divided to 2 UTF-16
539				 *   (charlen=3+4 or 4+4) */
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			} else if (charlen >= 5) {
545				/* 5-6bytes UTF-8 to 6bytes UTF-16 */
546				put_unaligned(dst_char, &target[j]);
547				dst_char = cpu_to_le16(*(wchar_to + 1));
548				j++;
549				put_unaligned(dst_char, &target[j]);
550				dst_char = cpu_to_le16(*(wchar_to + 2));
551				j++;
552				put_unaligned(dst_char, &target[j]);
553			}
554			continue;
555
556unknown:
557			dst_char = cpu_to_le16(0x003f);
558			charlen = 1;
559		}
560
561ctoUTF16:
562		/*
563		 * character may take more than one byte in the source string,
564		 * but will take exactly two bytes in the target string
565		 */
566		i += charlen;
567		put_unaligned(dst_char, &target[j]);
568	}
569
570ctoUTF16_out:
571	put_unaligned(0, &target[j]); /* Null terminate target unicode string */
572	kfree(wchar_to);
573	return j;
574}
575
576/*
577 * cifs_local_to_utf16_bytes - how long will a string be after conversion?
578 * @from - pointer to input string
579 * @maxbytes - don't go past this many bytes of input string
580 * @codepage - source codepage
581 *
582 * Walk a string and return the number of bytes that the string will
583 * be after being converted to the given charset, not including any null
584 * termination required. Don't walk past maxbytes in the source buffer.
585 */
586
587static int
588cifs_local_to_utf16_bytes(const char *from, int len,
589			  const struct nls_table *codepage)
590{
591	int charlen;
592	int i;
593	wchar_t wchar_to;
594
595	for (i = 0; len && *from; i++, from += charlen, len -= charlen) {
596		charlen = codepage->char2uni(from, len, &wchar_to);
597		/* Failed conversion defaults to a question mark */
598		if (charlen < 1)
599			charlen = 1;
600	}
601	return 2 * i; /* UTF16 characters are two bytes */
602}
603
604/*
605 * cifs_strndup_to_utf16 - copy a string to wire format from the local codepage
606 * @src - source string
607 * @maxlen - don't walk past this many bytes in the source string
608 * @utf16_len - the length of the allocated string in bytes (including null)
609 * @cp - source codepage
610 * @remap - map special chars
611 *
612 * Take a string convert it from the local codepage to UTF16 and
613 * put it in a new buffer. Returns a pointer to the new string or NULL on
614 * error.
615 */
616__le16 *
617cifs_strndup_to_utf16(const char *src, const int maxlen, int *utf16_len,
618		      const struct nls_table *cp, int remap)
619{
620	int len;
621	__le16 *dst;
622
623	len = cifs_local_to_utf16_bytes(src, maxlen, cp);
624	len += 2; /* NULL */
625	dst = kmalloc(len, GFP_KERNEL);
626	if (!dst) {
627		*utf16_len = 0;
628		return NULL;
629	}
630	cifsConvertToUTF16(dst, src, strlen(src), cp, remap);
631	*utf16_len = len;
632	return dst;
633}
634