xref: /freebsd/lib/libc/locale/utf8.c (revision cd0d51baaa4509a1db83251a601d34404d20c990)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
3  *
4  * Copyright 2013 Garrett D'Amore <garrett@damore.org>
5  * Copyright 2011 Nexenta Systems, Inc.  All rights reserved.
6  * Copyright (c) 2002-2004 Tim J. Robbins
7  * All rights reserved.
8  *
9  * Copyright (c) 2011 The FreeBSD Foundation
10  * All rights reserved.
11  * Portions of this software were developed by David Chisnall
12  * under sponsorship from the FreeBSD Foundation.
13  *
14  * Redistribution and use in source and binary forms, with or without
15  * modification, are permitted provided that the following conditions
16  * are met:
17  * 1. Redistributions of source code must retain the above copyright
18  *    notice, this list of conditions and the following disclaimer.
19  * 2. Redistributions in binary form must reproduce the above copyright
20  *    notice, this list of conditions and the following disclaimer in the
21  *    documentation and/or other materials provided with the distribution.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33  * SUCH DAMAGE.
34  */
35 
36 #include <sys/param.h>
37 __FBSDID("$FreeBSD$");
38 
39 #include <errno.h>
40 #include <limits.h>
41 #include <runetype.h>
42 #include <stdlib.h>
43 #include <string.h>
44 #include <wchar.h>
45 #include "mblocal.h"
46 
47 extern int __mb_sb_limit;
48 
49 static size_t	_UTF8_mbrtowc(wchar_t * __restrict, const char * __restrict,
50 		    size_t, mbstate_t * __restrict);
51 static int	_UTF8_mbsinit(const mbstate_t *);
52 static size_t	_UTF8_mbsnrtowcs(wchar_t * __restrict,
53 		    const char ** __restrict, size_t, size_t,
54 		    mbstate_t * __restrict);
55 static size_t	_UTF8_wcrtomb(char * __restrict, wchar_t,
56 		    mbstate_t * __restrict);
57 static size_t	_UTF8_wcsnrtombs(char * __restrict, const wchar_t ** __restrict,
58 		    size_t, size_t, mbstate_t * __restrict);
59 
60 typedef struct {
61 	wchar_t	ch;
62 	int	want;
63 	wchar_t	lbound;
64 } _UTF8State;
65 
66 int
67 _UTF8_init(struct xlocale_ctype *l, _RuneLocale *rl)
68 {
69 
70 	l->__mbrtowc = _UTF8_mbrtowc;
71 	l->__wcrtomb = _UTF8_wcrtomb;
72 	l->__mbsinit = _UTF8_mbsinit;
73 	l->__mbsnrtowcs = _UTF8_mbsnrtowcs;
74 	l->__wcsnrtombs = _UTF8_wcsnrtombs;
75 	l->runes = rl;
76 	l->__mb_cur_max = 4;
77 	/*
78 	 * UCS-4 encoding used as the internal representation, so
79 	 * slots 0x0080-0x00FF are occuped and must be excluded
80 	 * from the single byte ctype by setting the limit.
81 	 */
82 	l->__mb_sb_limit = 128;
83 
84 	return (0);
85 }
86 
87 static int
88 _UTF8_mbsinit(const mbstate_t *ps)
89 {
90 
91 	return (ps == NULL || ((const _UTF8State *)ps)->want == 0);
92 }
93 
94 static size_t
95 _UTF8_mbrtowc(wchar_t * __restrict pwc, const char * __restrict s, size_t n,
96     mbstate_t * __restrict ps)
97 {
98 	_UTF8State *us;
99 	int ch, i, mask, want;
100 	wchar_t lbound, wch;
101 
102 	us = (_UTF8State *)ps;
103 
104 	if (us->want < 0 || us->want > 6) {
105 		errno = EINVAL;
106 		return ((size_t)-1);
107 	}
108 
109 	if (s == NULL) {
110 		s = "";
111 		n = 1;
112 		pwc = NULL;
113 	}
114 
115 	if (n == 0)
116 		/* Incomplete multibyte sequence */
117 		return ((size_t)-2);
118 
119 	if (us->want == 0) {
120 		/*
121 		 * Determine the number of octets that make up this character
122 		 * from the first octet, and a mask that extracts the
123 		 * interesting bits of the first octet. We already know
124 		 * the character is at least two bytes long.
125 		 *
126 		 * We also specify a lower bound for the character code to
127 		 * detect redundant, non-"shortest form" encodings. For
128 		 * example, the sequence C0 80 is _not_ a legal representation
129 		 * of the null character. This enforces a 1-to-1 mapping
130 		 * between character codes and their multibyte representations.
131 		 */
132 		ch = (unsigned char)*s;
133 		if ((ch & 0x80) == 0) {
134 			/* Fast path for plain ASCII characters. */
135 			if (pwc != NULL)
136 				*pwc = ch;
137 			return (ch != '\0' ? 1 : 0);
138 		}
139 		if ((ch & 0xe0) == 0xc0) {
140 			mask = 0x1f;
141 			want = 2;
142 			lbound = 0x80;
143 		} else if ((ch & 0xf0) == 0xe0) {
144 			mask = 0x0f;
145 			want = 3;
146 			lbound = 0x800;
147 		} else if ((ch & 0xf8) == 0xf0) {
148 			mask = 0x07;
149 			want = 4;
150 			lbound = 0x10000;
151 		} else {
152 			/*
153 			 * Malformed input; input is not UTF-8.
154 			 */
155 			errno = EILSEQ;
156 			return ((size_t)-1);
157 		}
158 	} else {
159 		want = us->want;
160 		lbound = us->lbound;
161 	}
162 
163 	/*
164 	 * Decode the octet sequence representing the character in chunks
165 	 * of 6 bits, most significant first.
166 	 */
167 	if (us->want == 0)
168 		wch = (unsigned char)*s++ & mask;
169 	else
170 		wch = us->ch;
171 
172 	for (i = (us->want == 0) ? 1 : 0; i < MIN(want, n); i++) {
173 		if ((*s & 0xc0) != 0x80) {
174 			/*
175 			 * Malformed input; bad characters in the middle
176 			 * of a character.
177 			 */
178 			errno = EILSEQ;
179 			return ((size_t)-1);
180 		}
181 		wch <<= 6;
182 		wch |= *s++ & 0x3f;
183 	}
184 	if (i < want) {
185 		/* Incomplete multibyte sequence. */
186 		us->want = want - i;
187 		us->lbound = lbound;
188 		us->ch = wch;
189 		return ((size_t)-2);
190 	}
191 	if (wch < lbound) {
192 		/*
193 		 * Malformed input; redundant encoding.
194 		 */
195 		errno = EILSEQ;
196 		return ((size_t)-1);
197 	}
198 	if ((wch >= 0xd800 && wch <= 0xdfff) || wch > 0x10ffff) {
199 		/*
200 		 * Malformed input; invalid code points.
201 		 */
202 		errno = EILSEQ;
203 		return ((size_t)-1);
204 	}
205 	if (pwc != NULL)
206 		*pwc = wch;
207 	us->want = 0;
208 	return (wch == L'\0' ? 0 : want);
209 }
210 
211 static size_t
212 _UTF8_mbsnrtowcs(wchar_t * __restrict dst, const char ** __restrict src,
213     size_t nms, size_t len, mbstate_t * __restrict ps)
214 {
215 	_UTF8State *us;
216 	const char *s;
217 	size_t nchr;
218 	wchar_t wc;
219 	size_t nb;
220 
221 	us = (_UTF8State *)ps;
222 
223 	s = *src;
224 	nchr = 0;
225 
226 	if (dst == NULL) {
227 		/*
228 		 * The fast path in the loop below is not safe if an ASCII
229 		 * character appears as anything but the first byte of a
230 		 * multibyte sequence. Check now to avoid doing it in the loop.
231 		 */
232 		if (nms > 0 && us->want > 0 && (signed char)*s > 0) {
233 			errno = EILSEQ;
234 			return ((size_t)-1);
235 		}
236 		for (;;) {
237 			if (nms > 0 && (signed char)*s > 0)
238 				/*
239 				 * Fast path for plain ASCII characters
240 				 * excluding NUL.
241 				 */
242 				nb = 1;
243 			else if ((nb = _UTF8_mbrtowc(&wc, s, nms, ps)) ==
244 			    (size_t)-1)
245 				/* Invalid sequence - mbrtowc() sets errno. */
246 				return ((size_t)-1);
247 			else if (nb == 0 || nb == (size_t)-2)
248 				return (nchr);
249 			s += nb;
250 			nms -= nb;
251 			nchr++;
252 		}
253 		/*NOTREACHED*/
254 	}
255 
256 	/*
257 	 * The fast path in the loop below is not safe if an ASCII
258 	 * character appears as anything but the first byte of a
259 	 * multibyte sequence. Check now to avoid doing it in the loop.
260 	 */
261 	if (nms > 0 && len > 0 && us->want > 0 && (signed char)*s > 0) {
262 		errno = EILSEQ;
263 		return ((size_t)-1);
264 	}
265 	while (len-- > 0) {
266 		if (nms > 0 && (signed char)*s > 0) {
267 			/*
268 			 * Fast path for plain ASCII characters
269 			 * excluding NUL.
270 			 */
271 			*dst = (wchar_t)*s;
272 			nb = 1;
273 		} else if ((nb = _UTF8_mbrtowc(dst, s, nms, ps)) ==
274 		    (size_t)-1) {
275 			*src = s;
276 			return ((size_t)-1);
277 		} else if (nb == (size_t)-2) {
278 			*src = s + nms;
279 			return (nchr);
280 		} else if (nb == 0) {
281 			*src = NULL;
282 			return (nchr);
283 		}
284 		s += nb;
285 		nms -= nb;
286 		nchr++;
287 		dst++;
288 	}
289 	*src = s;
290 	return (nchr);
291 }
292 
293 static size_t
294 _UTF8_wcrtomb(char * __restrict s, wchar_t wc, mbstate_t * __restrict ps)
295 {
296 	_UTF8State *us;
297 	unsigned char lead;
298 	int i, len;
299 
300 	us = (_UTF8State *)ps;
301 
302 	if (us->want != 0) {
303 		errno = EINVAL;
304 		return ((size_t)-1);
305 	}
306 
307 	if (s == NULL)
308 		/* Reset to initial shift state (no-op) */
309 		return (1);
310 
311 	/*
312 	 * Determine the number of octets needed to represent this character.
313 	 * We always output the shortest sequence possible. Also specify the
314 	 * first few bits of the first octet, which contains the information
315 	 * about the sequence length.
316 	 */
317 	if ((wc & ~0x7f) == 0) {
318 		/* Fast path for plain ASCII characters. */
319 		*s = (char)wc;
320 		return (1);
321 	} else if ((wc & ~0x7ff) == 0) {
322 		lead = 0xc0;
323 		len = 2;
324 	} else if ((wc & ~0xffff) == 0) {
325 		if (wc >= 0xd800 && wc <= 0xdfff) {
326 			errno = EILSEQ;
327 			return ((size_t)-1);
328 		}
329 		lead = 0xe0;
330 		len = 3;
331 	} else if (wc >= 0 && wc <= 0x10ffff) {
332 		lead = 0xf0;
333 		len = 4;
334 	} else {
335 		errno = EILSEQ;
336 		return ((size_t)-1);
337 	}
338 
339 	/*
340 	 * Output the octets representing the character in chunks
341 	 * of 6 bits, least significant last. The first octet is
342 	 * a special case because it contains the sequence length
343 	 * information.
344 	 */
345 	for (i = len - 1; i > 0; i--) {
346 		s[i] = (wc & 0x3f) | 0x80;
347 		wc >>= 6;
348 	}
349 	*s = (wc & 0xff) | lead;
350 
351 	return (len);
352 }
353 
354 static size_t
355 _UTF8_wcsnrtombs(char * __restrict dst, const wchar_t ** __restrict src,
356     size_t nwc, size_t len, mbstate_t * __restrict ps)
357 {
358 	_UTF8State *us;
359 	char buf[MB_LEN_MAX];
360 	const wchar_t *s;
361 	size_t nbytes;
362 	size_t nb;
363 
364 	us = (_UTF8State *)ps;
365 
366 	if (us->want != 0) {
367 		errno = EINVAL;
368 		return ((size_t)-1);
369 	}
370 
371 	s = *src;
372 	nbytes = 0;
373 
374 	if (dst == NULL) {
375 		while (nwc-- > 0) {
376 			if (0 <= *s && *s < 0x80)
377 				/* Fast path for plain ASCII characters. */
378 				nb = 1;
379 			else if ((nb = _UTF8_wcrtomb(buf, *s, ps)) ==
380 			    (size_t)-1)
381 				/* Invalid character - wcrtomb() sets errno. */
382 				return ((size_t)-1);
383 			if (*s == L'\0')
384 				return (nbytes + nb - 1);
385 			s++;
386 			nbytes += nb;
387 		}
388 		return (nbytes);
389 	}
390 
391 	while (len > 0 && nwc-- > 0) {
392 		if (0 <= *s && *s < 0x80) {
393 			/* Fast path for plain ASCII characters. */
394 			nb = 1;
395 			*dst = *s;
396 		} else if (len > (size_t)MB_CUR_MAX) {
397 			/* Enough space to translate in-place. */
398 			if ((nb = _UTF8_wcrtomb(dst, *s, ps)) == (size_t)-1) {
399 				*src = s;
400 				return ((size_t)-1);
401 			}
402 		} else {
403 			/*
404 			 * May not be enough space; use temp. buffer.
405 			 */
406 			if ((nb = _UTF8_wcrtomb(buf, *s, ps)) == (size_t)-1) {
407 				*src = s;
408 				return ((size_t)-1);
409 			}
410 			if (nb > (int)len)
411 				/* MB sequence for character won't fit. */
412 				break;
413 			memcpy(dst, buf, nb);
414 		}
415 		if (*s == L'\0') {
416 			*src = NULL;
417 			return (nbytes + nb - 1);
418 		}
419 		s++;
420 		dst += nb;
421 		len -= nb;
422 		nbytes += nb;
423 	}
424 	*src = s;
425 	return (nbytes);
426 }
427