xref: /freebsd/lib/libc/locale/utf8.c (revision 9034852c84a13f0e3b5527e1c886ca94b2863b2b)
1 /*-
2  * Copyright 2011 Nexenta Systems, Inc.  All rights reserved.
3  * Copyright (c) 2002-2004 Tim J. Robbins
4  * All rights reserved.
5  *
6  * Copyright (c) 2011 The FreeBSD Foundation
7  * All rights reserved.
8  * Portions of this software were developed by David Chisnall
9  * under sponsorship from the FreeBSD Foundation.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 #include <sys/param.h>
34 __FBSDID("$FreeBSD$");
35 
36 #include <errno.h>
37 #include <limits.h>
38 #include <runetype.h>
39 #include <stdlib.h>
40 #include <string.h>
41 #include <wchar.h>
42 #include "mblocal.h"
43 
44 extern int __mb_sb_limit;
45 
46 static size_t	_UTF8_mbrtowc(wchar_t * __restrict, const char * __restrict,
47 		    size_t, mbstate_t * __restrict);
48 static int	_UTF8_mbsinit(const mbstate_t *);
49 static size_t	_UTF8_mbsnrtowcs(wchar_t * __restrict,
50 		    const char ** __restrict, size_t, size_t,
51 		    mbstate_t * __restrict);
52 static size_t	_UTF8_wcrtomb(char * __restrict, wchar_t,
53 		    mbstate_t * __restrict);
54 static size_t	_UTF8_wcsnrtombs(char * __restrict, const wchar_t ** __restrict,
55 		    size_t, size_t, mbstate_t * __restrict);
56 
57 typedef struct {
58 	wchar_t	ch;
59 	int	want;
60 	wchar_t	lbound;
61 } _UTF8State;
62 
63 int
64 _UTF8_init(struct xlocale_ctype *l, _RuneLocale *rl)
65 {
66 
67 	l->__mbrtowc = _UTF8_mbrtowc;
68 	l->__wcrtomb = _UTF8_wcrtomb;
69 	l->__mbsinit = _UTF8_mbsinit;
70 	l->__mbsnrtowcs = _UTF8_mbsnrtowcs;
71 	l->__wcsnrtombs = _UTF8_wcsnrtombs;
72 	l->runes = rl;
73 	l->__mb_cur_max = 6;
74 	/*
75 	 * UCS-4 encoding used as the internal representation, so
76 	 * slots 0x0080-0x00FF are occuped and must be excluded
77 	 * from the single byte ctype by setting the limit.
78 	 */
79 	l->__mb_sb_limit = 128;
80 
81 	return (0);
82 }
83 
84 static int
85 _UTF8_mbsinit(const mbstate_t *ps)
86 {
87 
88 	return (ps == NULL || ((const _UTF8State *)ps)->want == 0);
89 }
90 
91 static size_t
92 _UTF8_mbrtowc(wchar_t * __restrict pwc, const char * __restrict s, size_t n,
93     mbstate_t * __restrict ps)
94 {
95 	_UTF8State *us;
96 	int ch, i, mask, want;
97 	wchar_t lbound, wch;
98 
99 	us = (_UTF8State *)ps;
100 
101 	if (us->want < 0 || us->want > 6) {
102 		errno = EINVAL;
103 		return ((size_t)-1);
104 	}
105 
106 	if (s == NULL) {
107 		s = "";
108 		n = 1;
109 		pwc = NULL;
110 	}
111 
112 	if (n == 0)
113 		/* Incomplete multibyte sequence */
114 		return ((size_t)-2);
115 
116 	if (us->want == 0) {
117 		/*
118 		 * Determine the number of octets that make up this character
119 		 * from the first octet, and a mask that extracts the
120 		 * interesting bits of the first octet. We already know
121 		 * the character is at least two bytes long.
122 		 *
123 		 * We also specify a lower bound for the character code to
124 		 * detect redundant, non-"shortest form" encodings. For
125 		 * example, the sequence C0 80 is _not_ a legal representation
126 		 * of the null character. This enforces a 1-to-1 mapping
127 		 * between character codes and their multibyte representations.
128 		 */
129 		ch = (unsigned char)*s;
130 		if ((ch & 0x80) == 0) {
131 			/* Fast path for plain ASCII characters. */
132 			if (pwc != NULL)
133 				*pwc = ch;
134 			return (ch != '\0' ? 1 : 0);
135 		}
136 		if ((ch & 0xe0) == 0xc0) {
137 			mask = 0x1f;
138 			want = 2;
139 			lbound = 0x80;
140 		} else if ((ch & 0xf0) == 0xe0) {
141 			mask = 0x0f;
142 			want = 3;
143 			lbound = 0x800;
144 		} else if ((ch & 0xf8) == 0xf0) {
145 			mask = 0x07;
146 			want = 4;
147 			lbound = 0x10000;
148 		} else {
149 			/*
150 			 * Malformed input; input is not UTF-8.
151 			 */
152 			errno = EILSEQ;
153 			return ((size_t)-1);
154 		}
155 	} else {
156 		want = us->want;
157 		lbound = us->lbound;
158 	}
159 
160 	/*
161 	 * Decode the octet sequence representing the character in chunks
162 	 * of 6 bits, most significant first.
163 	 */
164 	if (us->want == 0)
165 		wch = (unsigned char)*s++ & mask;
166 	else
167 		wch = us->ch;
168 	for (i = (us->want == 0) ? 1 : 0; i < MIN(want, n); i++) {
169 		if ((*s & 0xc0) != 0x80) {
170 			/*
171 			 * Malformed input; bad characters in the middle
172 			 * of a character.
173 			 */
174 			errno = EILSEQ;
175 			return ((size_t)-1);
176 		}
177 		wch <<= 6;
178 		wch |= *s++ & 0x3f;
179 	}
180 	if (i < want) {
181 		/* Incomplete multibyte sequence. */
182 		us->want = want - i;
183 		us->lbound = lbound;
184 		us->ch = wch;
185 		return ((size_t)-2);
186 	}
187 	if (wch < lbound) {
188 		/*
189 		 * Malformed input; redundant encoding.
190 		 */
191 		errno = EILSEQ;
192 		return ((size_t)-1);
193 	}
194 	if ((wch >= 0xd800 && wch <= 0xdfff) || wch > 0x10ffff) {
195 		/*
196 		 * Malformed input; invalid code points.
197 		 */
198 		errno = EILSEQ;
199 		return ((size_t)-1);
200 	}
201 	if (pwc != NULL)
202 		*pwc = wch;
203 	us->want = 0;
204 	return (wch == L'\0' ? 0 : want);
205 }
206 
207 static size_t
208 _UTF8_mbsnrtowcs(wchar_t * __restrict dst, const char ** __restrict src,
209     size_t nms, size_t len, mbstate_t * __restrict ps)
210 {
211 	_UTF8State *us;
212 	const char *s;
213 	size_t nchr;
214 	wchar_t wc;
215 	size_t nb;
216 
217 	us = (_UTF8State *)ps;
218 
219 	s = *src;
220 	nchr = 0;
221 
222 	if (dst == NULL) {
223 		/*
224 		 * The fast path in the loop below is not safe if an ASCII
225 		 * character appears as anything but the first byte of a
226 		 * multibyte sequence. Check now to avoid doing it in the loop.
227 		 */
228 		if (nms > 0 && us->want > 0 && (signed char)*s > 0) {
229 			errno = EILSEQ;
230 			return ((size_t)-1);
231 		}
232 		for (;;) {
233 			if (nms > 0 && (signed char)*s > 0)
234 				/*
235 				 * Fast path for plain ASCII characters
236 				 * excluding NUL.
237 				 */
238 				nb = 1;
239 			else if ((nb = _UTF8_mbrtowc(&wc, s, nms, ps)) ==
240 			    (size_t)-1)
241 				/* Invalid sequence - mbrtowc() sets errno. */
242 				return ((size_t)-1);
243 			else if (nb == 0 || nb == (size_t)-2)
244 				return (nchr);
245 			s += nb;
246 			nms -= nb;
247 			nchr++;
248 		}
249 		/*NOTREACHED*/
250 	}
251 
252 	/*
253 	 * The fast path in the loop below is not safe if an ASCII
254 	 * character appears as anything but the first byte of a
255 	 * multibyte sequence. Check now to avoid doing it in the loop.
256 	 */
257 	if (nms > 0 && len > 0 && us->want > 0 && (signed char)*s > 0) {
258 		errno = EILSEQ;
259 		return ((size_t)-1);
260 	}
261 	while (len-- > 0) {
262 		if (nms > 0 && (signed char)*s > 0) {
263 			/*
264 			 * Fast path for plain ASCII characters
265 			 * excluding NUL.
266 			 */
267 			*dst = (wchar_t)*s;
268 			nb = 1;
269 		} else if ((nb = _UTF8_mbrtowc(dst, s, nms, ps)) ==
270 		    (size_t)-1) {
271 			*src = s;
272 			return ((size_t)-1);
273 		} else if (nb == (size_t)-2) {
274 			*src = s + nms;
275 			return (nchr);
276 		} else if (nb == 0) {
277 			*src = NULL;
278 			return (nchr);
279 		}
280 		s += nb;
281 		nms -= nb;
282 		nchr++;
283 		dst++;
284 	}
285 	*src = s;
286 	return (nchr);
287 }
288 
289 static size_t
290 _UTF8_wcrtomb(char * __restrict s, wchar_t wc, mbstate_t * __restrict ps)
291 {
292 	_UTF8State *us;
293 	unsigned char lead;
294 	int i, len;
295 
296 	us = (_UTF8State *)ps;
297 
298 	if (us->want != 0) {
299 		errno = EINVAL;
300 		return ((size_t)-1);
301 	}
302 
303 	if (s == NULL)
304 		/* Reset to initial shift state (no-op) */
305 		return (1);
306 
307 	/*
308 	 * Determine the number of octets needed to represent this character.
309 	 * We always output the shortest sequence possible. Also specify the
310 	 * first few bits of the first octet, which contains the information
311 	 * about the sequence length.
312 	 */
313 	if ((wc & ~0x7f) == 0) {
314 		/* Fast path for plain ASCII characters. */
315 		*s = (char)wc;
316 		return (1);
317 	} else if ((wc & ~0x7ff) == 0) {
318 		lead = 0xc0;
319 		len = 2;
320 	} else if ((wc & ~0xffff) == 0) {
321 		if (wc >= 0xd800 && wc <= 0xdfff) {
322 			errno = EILSEQ;
323 			return ((size_t)-1);
324 		}
325 		lead = 0xe0;
326 		len = 3;
327 	} else if (wc >= 0 && wc <= 0x10ffff) {
328 		lead = 0xf0;
329 		len = 4;
330 	} else {
331 		errno = EILSEQ;
332 		return ((size_t)-1);
333 	}
334 
335 	/*
336 	 * Output the octets representing the character in chunks
337 	 * of 6 bits, least significant last. The first octet is
338 	 * a special case because it contains the sequence length
339 	 * information.
340 	 */
341 	for (i = len - 1; i > 0; i--) {
342 		s[i] = (wc & 0x3f) | 0x80;
343 		wc >>= 6;
344 	}
345 	*s = (wc & 0xff) | lead;
346 
347 	return (len);
348 }
349 
350 static size_t
351 _UTF8_wcsnrtombs(char * __restrict dst, const wchar_t ** __restrict src,
352     size_t nwc, size_t len, mbstate_t * __restrict ps)
353 {
354 	_UTF8State *us;
355 	char buf[MB_LEN_MAX];
356 	const wchar_t *s;
357 	size_t nbytes;
358 	size_t nb;
359 
360 	us = (_UTF8State *)ps;
361 
362 	if (us->want != 0) {
363 		errno = EINVAL;
364 		return ((size_t)-1);
365 	}
366 
367 	s = *src;
368 	nbytes = 0;
369 
370 	if (dst == NULL) {
371 		while (nwc-- > 0) {
372 			if (0 <= *s && *s < 0x80)
373 				/* Fast path for plain ASCII characters. */
374 				nb = 1;
375 			else if ((nb = _UTF8_wcrtomb(buf, *s, ps)) ==
376 			    (size_t)-1)
377 				/* Invalid character - wcrtomb() sets errno. */
378 				return ((size_t)-1);
379 			if (*s == L'\0')
380 				return (nbytes + nb - 1);
381 			s++;
382 			nbytes += nb;
383 		}
384 		return (nbytes);
385 	}
386 
387 	while (len > 0 && nwc-- > 0) {
388 		if (0 <= *s && *s < 0x80) {
389 			/* Fast path for plain ASCII characters. */
390 			nb = 1;
391 			*dst = *s;
392 		} else if (len > (size_t)MB_CUR_MAX) {
393 			/* Enough space to translate in-place. */
394 			if ((nb = _UTF8_wcrtomb(dst, *s, ps)) == (size_t)-1) {
395 				*src = s;
396 				return ((size_t)-1);
397 			}
398 		} else {
399 			/*
400 			 * May not be enough space; use temp. buffer.
401 			 */
402 			if ((nb = _UTF8_wcrtomb(buf, *s, ps)) == (size_t)-1) {
403 				*src = s;
404 				return ((size_t)-1);
405 			}
406 			if (nb > (int)len)
407 				/* MB sequence for character won't fit. */
408 				break;
409 			memcpy(dst, buf, nb);
410 		}
411 		if (*s == L'\0') {
412 			*src = NULL;
413 			return (nbytes + nb - 1);
414 		}
415 		s++;
416 		dst += nb;
417 		len -= nb;
418 		nbytes += nb;
419 	}
420 	*src = s;
421 	return (nbytes);
422 }
423