xref: /freebsd/lib/libc/locale/utf8.c (revision 193d9e768ba63fcfb187cfd17f461f7d41345048)
1 /*-
2  * Copyright 2013 Garrett D'Amore <garrett@damore.org>
3  * Copyright 2011 Nexenta Systems, Inc.  All rights reserved.
4  * Copyright (c) 2002-2004 Tim J. Robbins
5  * All rights reserved.
6  *
7  * Copyright (c) 2011 The FreeBSD Foundation
8  * All rights reserved.
9  * Portions of this software were developed by David Chisnall
10  * under sponsorship from the FreeBSD Foundation.
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  * 1. Redistributions of source code must retain the above copyright
16  *    notice, this list of conditions and the following disclaimer.
17  * 2. Redistributions in binary form must reproduce the above copyright
18  *    notice, this list of conditions and the following disclaimer in the
19  *    documentation and/or other materials provided with the distribution.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  */
33 
34 #include <sys/param.h>
35 __FBSDID("$FreeBSD$");
36 
37 #include <errno.h>
38 #include <limits.h>
39 #include <runetype.h>
40 #include <stdlib.h>
41 #include <string.h>
42 #include <wchar.h>
43 #include "mblocal.h"
44 
45 extern int __mb_sb_limit;
46 
47 static size_t	_UTF8_mbrtowc(wchar_t * __restrict, const char * __restrict,
48 		    size_t, mbstate_t * __restrict);
49 static int	_UTF8_mbsinit(const mbstate_t *);
50 static size_t	_UTF8_mbsnrtowcs(wchar_t * __restrict,
51 		    const char ** __restrict, size_t, size_t,
52 		    mbstate_t * __restrict);
53 static size_t	_UTF8_wcrtomb(char * __restrict, wchar_t,
54 		    mbstate_t * __restrict);
55 static size_t	_UTF8_wcsnrtombs(char * __restrict, const wchar_t ** __restrict,
56 		    size_t, size_t, mbstate_t * __restrict);
57 
58 typedef struct {
59 	wchar_t	ch;
60 	int	want;
61 	wchar_t	lbound;
62 } _UTF8State;
63 
64 int
65 _UTF8_init(struct xlocale_ctype *l, _RuneLocale *rl)
66 {
67 
68 	l->__mbrtowc = _UTF8_mbrtowc;
69 	l->__wcrtomb = _UTF8_wcrtomb;
70 	l->__mbsinit = _UTF8_mbsinit;
71 	l->__mbsnrtowcs = _UTF8_mbsnrtowcs;
72 	l->__wcsnrtombs = _UTF8_wcsnrtombs;
73 	l->runes = rl;
74 	l->__mb_cur_max = 4;
75 	/*
76 	 * UCS-4 encoding used as the internal representation, so
77 	 * slots 0x0080-0x00FF are occuped and must be excluded
78 	 * from the single byte ctype by setting the limit.
79 	 */
80 	l->__mb_sb_limit = 128;
81 
82 	return (0);
83 }
84 
85 static int
86 _UTF8_mbsinit(const mbstate_t *ps)
87 {
88 
89 	return (ps == NULL || ((const _UTF8State *)ps)->want == 0);
90 }
91 
92 static size_t
93 _UTF8_mbrtowc(wchar_t * __restrict pwc, const char * __restrict s, size_t n,
94     mbstate_t * __restrict ps)
95 {
96 	_UTF8State *us;
97 	int ch, i, mask, want;
98 	wchar_t lbound, wch;
99 
100 	us = (_UTF8State *)ps;
101 
102 	if (us->want < 0 || us->want > 6) {
103 		errno = EINVAL;
104 		return ((size_t)-1);
105 	}
106 
107 	if (s == NULL) {
108 		s = "";
109 		n = 1;
110 		pwc = NULL;
111 	}
112 
113 	if (n == 0)
114 		/* Incomplete multibyte sequence */
115 		return ((size_t)-2);
116 
117 	if (us->want == 0) {
118 		/*
119 		 * Determine the number of octets that make up this character
120 		 * from the first octet, and a mask that extracts the
121 		 * interesting bits of the first octet. We already know
122 		 * the character is at least two bytes long.
123 		 *
124 		 * We also specify a lower bound for the character code to
125 		 * detect redundant, non-"shortest form" encodings. For
126 		 * example, the sequence C0 80 is _not_ a legal representation
127 		 * of the null character. This enforces a 1-to-1 mapping
128 		 * between character codes and their multibyte representations.
129 		 */
130 		ch = (unsigned char)*s;
131 		if ((ch & 0x80) == 0) {
132 			/* Fast path for plain ASCII characters. */
133 			if (pwc != NULL)
134 				*pwc = ch;
135 			return (ch != '\0' ? 1 : 0);
136 		}
137 		if ((ch & 0xe0) == 0xc0) {
138 			mask = 0x1f;
139 			want = 2;
140 			lbound = 0x80;
141 		} else if ((ch & 0xf0) == 0xe0) {
142 			mask = 0x0f;
143 			want = 3;
144 			lbound = 0x800;
145 		} else if ((ch & 0xf8) == 0xf0) {
146 			mask = 0x07;
147 			want = 4;
148 			lbound = 0x10000;
149 		} else {
150 			/*
151 			 * Malformed input; input is not UTF-8.
152 			 */
153 			errno = EILSEQ;
154 			return ((size_t)-1);
155 		}
156 	} else {
157 		want = us->want;
158 		lbound = us->lbound;
159 	}
160 
161 	/*
162 	 * Decode the octet sequence representing the character in chunks
163 	 * of 6 bits, most significant first.
164 	 */
165 	if (us->want == 0)
166 		wch = (unsigned char)*s++ & mask;
167 	else
168 		wch = us->ch;
169 
170 	for (i = (us->want == 0) ? 1 : 0; i < MIN(want, n); i++) {
171 		if ((*s & 0xc0) != 0x80) {
172 			/*
173 			 * Malformed input; bad characters in the middle
174 			 * of a character.
175 			 */
176 			errno = EILSEQ;
177 			return ((size_t)-1);
178 		}
179 		wch <<= 6;
180 		wch |= *s++ & 0x3f;
181 	}
182 	if (i < want) {
183 		/* Incomplete multibyte sequence. */
184 		us->want = want - i;
185 		us->lbound = lbound;
186 		us->ch = wch;
187 		return ((size_t)-2);
188 	}
189 	if (wch < lbound) {
190 		/*
191 		 * Malformed input; redundant encoding.
192 		 */
193 		errno = EILSEQ;
194 		return ((size_t)-1);
195 	}
196 	if ((wch >= 0xd800 && wch <= 0xdfff) || wch > 0x10ffff) {
197 		/*
198 		 * Malformed input; invalid code points.
199 		 */
200 		errno = EILSEQ;
201 		return ((size_t)-1);
202 	}
203 	if (pwc != NULL)
204 		*pwc = wch;
205 	us->want = 0;
206 	return (wch == L'\0' ? 0 : want);
207 }
208 
209 static size_t
210 _UTF8_mbsnrtowcs(wchar_t * __restrict dst, const char ** __restrict src,
211     size_t nms, size_t len, mbstate_t * __restrict ps)
212 {
213 	_UTF8State *us;
214 	const char *s;
215 	size_t nchr;
216 	wchar_t wc;
217 	size_t nb;
218 
219 	us = (_UTF8State *)ps;
220 
221 	s = *src;
222 	nchr = 0;
223 
224 	if (dst == NULL) {
225 		/*
226 		 * The fast path in the loop below is not safe if an ASCII
227 		 * character appears as anything but the first byte of a
228 		 * multibyte sequence. Check now to avoid doing it in the loop.
229 		 */
230 		if (nms > 0 && us->want > 0 && (signed char)*s > 0) {
231 			errno = EILSEQ;
232 			return ((size_t)-1);
233 		}
234 		for (;;) {
235 			if (nms > 0 && (signed char)*s > 0)
236 				/*
237 				 * Fast path for plain ASCII characters
238 				 * excluding NUL.
239 				 */
240 				nb = 1;
241 			else if ((nb = _UTF8_mbrtowc(&wc, s, nms, ps)) ==
242 			    (size_t)-1)
243 				/* Invalid sequence - mbrtowc() sets errno. */
244 				return ((size_t)-1);
245 			else if (nb == 0 || nb == (size_t)-2)
246 				return (nchr);
247 			s += nb;
248 			nms -= nb;
249 			nchr++;
250 		}
251 		/*NOTREACHED*/
252 	}
253 
254 	/*
255 	 * The fast path in the loop below is not safe if an ASCII
256 	 * character appears as anything but the first byte of a
257 	 * multibyte sequence. Check now to avoid doing it in the loop.
258 	 */
259 	if (nms > 0 && len > 0 && us->want > 0 && (signed char)*s > 0) {
260 		errno = EILSEQ;
261 		return ((size_t)-1);
262 	}
263 	while (len-- > 0) {
264 		if (nms > 0 && (signed char)*s > 0) {
265 			/*
266 			 * Fast path for plain ASCII characters
267 			 * excluding NUL.
268 			 */
269 			*dst = (wchar_t)*s;
270 			nb = 1;
271 		} else if ((nb = _UTF8_mbrtowc(dst, s, nms, ps)) ==
272 		    (size_t)-1) {
273 			*src = s;
274 			return ((size_t)-1);
275 		} else if (nb == (size_t)-2) {
276 			*src = s + nms;
277 			return (nchr);
278 		} else if (nb == 0) {
279 			*src = NULL;
280 			return (nchr);
281 		}
282 		s += nb;
283 		nms -= nb;
284 		nchr++;
285 		dst++;
286 	}
287 	*src = s;
288 	return (nchr);
289 }
290 
291 static size_t
292 _UTF8_wcrtomb(char * __restrict s, wchar_t wc, mbstate_t * __restrict ps)
293 {
294 	_UTF8State *us;
295 	unsigned char lead;
296 	int i, len;
297 
298 	us = (_UTF8State *)ps;
299 
300 	if (us->want != 0) {
301 		errno = EINVAL;
302 		return ((size_t)-1);
303 	}
304 
305 	if (s == NULL)
306 		/* Reset to initial shift state (no-op) */
307 		return (1);
308 
309 	/*
310 	 * Determine the number of octets needed to represent this character.
311 	 * We always output the shortest sequence possible. Also specify the
312 	 * first few bits of the first octet, which contains the information
313 	 * about the sequence length.
314 	 */
315 	if ((wc & ~0x7f) == 0) {
316 		/* Fast path for plain ASCII characters. */
317 		*s = (char)wc;
318 		return (1);
319 	} else if ((wc & ~0x7ff) == 0) {
320 		lead = 0xc0;
321 		len = 2;
322 	} else if ((wc & ~0xffff) == 0) {
323 		if (wc >= 0xd800 && wc <= 0xdfff) {
324 			errno = EILSEQ;
325 			return ((size_t)-1);
326 		}
327 		lead = 0xe0;
328 		len = 3;
329 	} else if (wc >= 0 && wc <= 0x10ffff) {
330 		lead = 0xf0;
331 		len = 4;
332 	} else {
333 		errno = EILSEQ;
334 		return ((size_t)-1);
335 	}
336 
337 	/*
338 	 * Output the octets representing the character in chunks
339 	 * of 6 bits, least significant last. The first octet is
340 	 * a special case because it contains the sequence length
341 	 * information.
342 	 */
343 	for (i = len - 1; i > 0; i--) {
344 		s[i] = (wc & 0x3f) | 0x80;
345 		wc >>= 6;
346 	}
347 	*s = (wc & 0xff) | lead;
348 
349 	return (len);
350 }
351 
352 static size_t
353 _UTF8_wcsnrtombs(char * __restrict dst, const wchar_t ** __restrict src,
354     size_t nwc, size_t len, mbstate_t * __restrict ps)
355 {
356 	_UTF8State *us;
357 	char buf[MB_LEN_MAX];
358 	const wchar_t *s;
359 	size_t nbytes;
360 	size_t nb;
361 
362 	us = (_UTF8State *)ps;
363 
364 	if (us->want != 0) {
365 		errno = EINVAL;
366 		return ((size_t)-1);
367 	}
368 
369 	s = *src;
370 	nbytes = 0;
371 
372 	if (dst == NULL) {
373 		while (nwc-- > 0) {
374 			if (0 <= *s && *s < 0x80)
375 				/* Fast path for plain ASCII characters. */
376 				nb = 1;
377 			else if ((nb = _UTF8_wcrtomb(buf, *s, ps)) ==
378 			    (size_t)-1)
379 				/* Invalid character - wcrtomb() sets errno. */
380 				return ((size_t)-1);
381 			if (*s == L'\0')
382 				return (nbytes + nb - 1);
383 			s++;
384 			nbytes += nb;
385 		}
386 		return (nbytes);
387 	}
388 
389 	while (len > 0 && nwc-- > 0) {
390 		if (0 <= *s && *s < 0x80) {
391 			/* Fast path for plain ASCII characters. */
392 			nb = 1;
393 			*dst = *s;
394 		} else if (len > (size_t)MB_CUR_MAX) {
395 			/* Enough space to translate in-place. */
396 			if ((nb = _UTF8_wcrtomb(dst, *s, ps)) == (size_t)-1) {
397 				*src = s;
398 				return ((size_t)-1);
399 			}
400 		} else {
401 			/*
402 			 * May not be enough space; use temp. buffer.
403 			 */
404 			if ((nb = _UTF8_wcrtomb(buf, *s, ps)) == (size_t)-1) {
405 				*src = s;
406 				return ((size_t)-1);
407 			}
408 			if (nb > (int)len)
409 				/* MB sequence for character won't fit. */
410 				break;
411 			memcpy(dst, buf, nb);
412 		}
413 		if (*s == L'\0') {
414 			*src = NULL;
415 			return (nbytes + nb - 1);
416 		}
417 		s++;
418 		dst += nb;
419 		len -= nb;
420 		nbytes += nb;
421 	}
422 	*src = s;
423 	return (nbytes);
424 }
425