xref: /titanic_41/usr/src/lib/libc/port/locale/utf8.c (revision b8afd3a780ce850ff107bb3be330465bf47f84bd)
1 /*
2  * Copyright 2011 Nexenta Systems, Inc.  All rights reserved.
3  * Copyright (c) 2002-2004 Tim J. Robbins
4  * All rights reserved.
5  *
6  * Redistribution and use in source and binary forms, with or without
7  * modification, are permitted provided that the following conditions
8  * are met:
9  * 1. Redistributions of source code must retain the above copyright
10  *    notice, this list of conditions and the following disclaimer.
11  * 2. Redistributions in binary form must reproduce the above copyright
12  *    notice, this list of conditions and the following disclaimer in the
13  *    documentation and/or other materials provided with the distribution.
14  *
15  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
16  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
17  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
18  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
19  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
20  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
21  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
22  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
23  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
24  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
25  * SUCH DAMAGE.
26  */
27 
28 #include "lint.h"
29 #include <errno.h>
30 #include <limits.h>
31 #include "runetype.h"
32 #include <stdlib.h>
33 #include <string.h>
34 #include <wchar.h>
35 #include "mblocal.h"
36 
37 static size_t	_UTF8_mbrtowc(wchar_t *_RESTRICT_KYWD,
38 		    const char *_RESTRICT_KYWD,
39 		    size_t, mbstate_t *_RESTRICT_KYWD);
40 static int	_UTF8_mbsinit(const mbstate_t *);
41 static size_t	_UTF8_mbsnrtowcs(wchar_t *_RESTRICT_KYWD,
42 		    const char **_RESTRICT_KYWD, size_t, size_t,
43 		    mbstate_t *_RESTRICT_KYWD);
44 static size_t	_UTF8_wcrtomb(char *_RESTRICT_KYWD, wchar_t,
45 		    mbstate_t *_RESTRICT_KYWD);
46 static size_t	_UTF8_wcsnrtombs(char *_RESTRICT_KYWD,
47 		    const wchar_t **_RESTRICT_KYWD,
48 		    size_t, size_t, mbstate_t *_RESTRICT_KYWD);
49 
50 typedef struct {
51 	wchar_t	ch;
52 	int	want;
53 	wchar_t	lbound;
54 } _UTF8State;
55 
56 int
57 _UTF8_init(_RuneLocale *rl)
58 {
59 	__mbrtowc = _UTF8_mbrtowc;
60 	__wcrtomb = _UTF8_wcrtomb;
61 	__mbsinit = _UTF8_mbsinit;
62 	__mbsnrtowcs = _UTF8_mbsnrtowcs;
63 	__wcsnrtombs = _UTF8_wcsnrtombs;
64 	_CurrentRuneLocale = rl;
65 
66 	charset_is_ascii = 0;
67 
68 	/*
69 	 * In theory up to 6 bytes can be used for the encoding,
70 	 * but only encodings with more than 4 bytes are illegal.
71 	 */
72 	__ctype[520] = 4;
73 	/*
74 	 * Note that the other CSWIDTH members are nonsensical for this
75 	 * this coding.  They only are valid with EUC codings.
76 	 */
77 
78 	return (0);
79 }
80 
81 static int
82 _UTF8_mbsinit(const mbstate_t *ps)
83 {
84 
85 	return (ps == NULL || ((const _UTF8State *)ps)->want == 0);
86 }
87 
88 static size_t
89 _UTF8_mbrtowc(wchar_t *_RESTRICT_KYWD pwc, const char *_RESTRICT_KYWD s,
90     size_t n, mbstate_t *_RESTRICT_KYWD ps)
91 {
92 	_UTF8State *us;
93 	int ch, i, mask, want;
94 	wchar_t lbound, wch;
95 
96 	us = (_UTF8State *)ps;
97 
98 	if (us->want < 0 || us->want > 6) {
99 		errno = EINVAL;
100 		return ((size_t)-1);
101 	}
102 
103 	if (s == NULL) {
104 		s = "";
105 		n = 1;
106 		pwc = NULL;
107 	}
108 
109 	if (n == 0)
110 		/* Incomplete multibyte sequence */
111 		return ((size_t)-2);
112 
113 	if (us->want == 0) {
114 		/*
115 		 * Determine the number of octets that make up this character
116 		 * from the first octet, and a mask that extracts the
117 		 * interesting bits of the first octet. We already know
118 		 * the character is at least two bytes long.
119 		 *
120 		 * We also specify a lower bound for the character code to
121 		 * detect redundant, non-"shortest form" encodings. For
122 		 * example, the sequence C0 80 is _not_ a legal representation
123 		 * of the null character. This enforces a 1-to-1 mapping
124 		 * between character codes and their multibyte representations.
125 		 */
126 		ch = (unsigned char)*s;
127 		if ((ch & 0x80) == 0) {
128 			/* Fast path for plain ASCII characters. */
129 			if (pwc != NULL)
130 				*pwc = ch;
131 			return (ch != '\0' ? 1 : 0);
132 		}
133 		if ((ch & 0xe0) == 0xc0) {
134 			mask = 0x1f;
135 			want = 2;
136 			lbound = 0x80;
137 		} else if ((ch & 0xf0) == 0xe0) {
138 			mask = 0x0f;
139 			want = 3;
140 			lbound = 0x800;
141 		} else if ((ch & 0xf8) == 0xf0) {
142 			mask = 0x07;
143 			want = 4;
144 			lbound = 0x10000;
145 #if 0
146 		/* These would be illegal in the UTF-8 space */
147 
148 		} else if ((ch & 0xfc) == 0xf8) {
149 			mask = 0x03;
150 			want = 5;
151 			lbound = 0x200000;
152 		} else if ((ch & 0xfe) == 0xfc) {
153 			mask = 0x01;
154 			want = 6;
155 			lbound = 0x4000000;
156 #endif
157 		} else {
158 			/*
159 			 * Malformed input; input is not UTF-8.
160 			 */
161 			errno = EILSEQ;
162 			return ((size_t)-1);
163 		}
164 	} else {
165 		want = us->want;
166 		lbound = us->lbound;
167 	}
168 
169 	/*
170 	 * Decode the octet sequence representing the character in chunks
171 	 * of 6 bits, most significant first.
172 	 */
173 	if (us->want == 0)
174 		wch = (unsigned char)*s++ & mask;
175 	else
176 		wch = us->ch;
177 
178 	for (i = (us->want == 0) ? 1 : 0; i < MIN(want, n); i++) {
179 		if ((*s & 0xc0) != 0x80) {
180 			/*
181 			 * Malformed input; bad characters in the middle
182 			 * of a character.
183 			 */
184 			errno = EILSEQ;
185 			return ((size_t)-1);
186 		}
187 		wch <<= 6;
188 		wch |= *s++ & 0x3f;
189 	}
190 	if (i < want) {
191 		/* Incomplete multibyte sequence. */
192 		us->want = want - i;
193 		us->lbound = lbound;
194 		us->ch = wch;
195 		return ((size_t)-2);
196 	}
197 	if (wch < lbound) {
198 		/*
199 		 * Malformed input; redundant encoding.
200 		 */
201 		errno = EILSEQ;
202 		return ((size_t)-1);
203 	}
204 	if (pwc != NULL)
205 		*pwc = wch;
206 	us->want = 0;
207 	return (wch == L'\0' ? 0 : want);
208 }
209 
210 static size_t
211 _UTF8_mbsnrtowcs(wchar_t *_RESTRICT_KYWD dst, const char **_RESTRICT_KYWD src,
212     size_t nms, size_t len, mbstate_t *_RESTRICT_KYWD ps)
213 {
214 	_UTF8State *us;
215 	const char *s;
216 	size_t nchr;
217 	wchar_t wc;
218 	size_t nb;
219 
220 	us = (_UTF8State *)ps;
221 
222 	s = *src;
223 	nchr = 0;
224 
225 	if (dst == NULL) {
226 		/*
227 		 * The fast path in the loop below is not safe if an ASCII
228 		 * character appears as anything but the first byte of a
229 		 * multibyte sequence. Check now to avoid doing it in the loop.
230 		 */
231 		if (nms > 0 && us->want > 0 && (signed char)*s > 0) {
232 			errno = EILSEQ;
233 			return ((size_t)-1);
234 		}
235 		for (;;) {
236 			if (nms > 0 && (signed char)*s > 0)
237 				/*
238 				 * Fast path for plain ASCII characters
239 				 * excluding NUL.
240 				 */
241 				nb = 1;
242 			else if ((nb = _UTF8_mbrtowc(&wc, s, nms, ps)) ==
243 			    (size_t)-1)
244 				/* Invalid sequence - mbrtowc() sets errno. */
245 				return ((size_t)-1);
246 			else if (nb == 0 || nb == (size_t)-2)
247 				return (nchr);
248 			s += nb;
249 			nms -= nb;
250 			nchr++;
251 		}
252 		/*NOTREACHED*/
253 	}
254 
255 	/*
256 	 * The fast path in the loop below is not safe if an ASCII
257 	 * character appears as anything but the first byte of a
258 	 * multibyte sequence. Check now to avoid doing it in the loop.
259 	 */
260 	if (nms > 0 && len > 0 && us->want > 0 && (signed char)*s > 0) {
261 		errno = EILSEQ;
262 		return ((size_t)-1);
263 	}
264 	while (len-- > 0) {
265 		if (nms > 0 && (signed char)*s > 0) {
266 			/*
267 			 * Fast path for plain ASCII characters
268 			 * excluding NUL.
269 			 */
270 			*dst = (wchar_t)*s;
271 			nb = 1;
272 		} else if ((nb = _UTF8_mbrtowc(dst, s, nms, ps)) ==
273 		    (size_t)-1) {
274 			*src = s;
275 			return ((size_t)-1);
276 		} else if (nb == (size_t)-2) {
277 			*src = s + nms;
278 			return (nchr);
279 		} else if (nb == 0) {
280 			*src = NULL;
281 			return (nchr);
282 		}
283 		s += nb;
284 		nms -= nb;
285 		nchr++;
286 		dst++;
287 	}
288 	*src = s;
289 	return (nchr);
290 }
291 
292 static size_t
293 _UTF8_wcrtomb(char *_RESTRICT_KYWD s, wchar_t wc, mbstate_t *_RESTRICT_KYWD ps)
294 {
295 	_UTF8State *us;
296 	unsigned char lead;
297 	int i, len;
298 
299 	us = (_UTF8State *)ps;
300 
301 	if (us->want != 0) {
302 		errno = EINVAL;
303 		return ((size_t)-1);
304 	}
305 
306 	if (s == NULL)
307 		/* Reset to initial shift state (no-op) */
308 		return (1);
309 
310 	/*
311 	 * Determine the number of octets needed to represent this character.
312 	 * We always output the shortest sequence possible. Also specify the
313 	 * first few bits of the first octet, which contains the information
314 	 * about the sequence length.
315 	 */
316 	if ((wc & ~0x7f) == 0) {
317 		/* Fast path for plain ASCII characters. */
318 		*s = (char)wc;
319 		return (1);
320 	} else if ((wc & ~0x7ff) == 0) {
321 		lead = 0xc0;
322 		len = 2;
323 	} else if ((wc & ~0xffff) == 0) {
324 		lead = 0xe0;
325 		len = 3;
326 	} else if ((wc & ~0x1fffff) == 0) {
327 		lead = 0xf0;
328 		len = 4;
329 #if 0
330 	/* Again, 5 and 6 byte encodings are simply not permitted */
331 	} else if ((wc & ~0x3ffffff) == 0) {
332 		lead = 0xf8;
333 		len = 5;
334 	} else if ((wc & ~0x7fffffff) == 0) {
335 		lead = 0xfc;
336 		len = 6;
337 #endif
338 	} else {
339 		errno = EILSEQ;
340 		return ((size_t)-1);
341 	}
342 
343 	/*
344 	 * Output the octets representing the character in chunks
345 	 * of 6 bits, least significant last. The first octet is
346 	 * a special case because it contains the sequence length
347 	 * information.
348 	 */
349 	for (i = len - 1; i > 0; i--) {
350 		s[i] = (wc & 0x3f) | 0x80;
351 		wc >>= 6;
352 	}
353 	*s = (wc & 0xff) | lead;
354 
355 	return (len);
356 }
357 
358 static size_t
359 _UTF8_wcsnrtombs(char *_RESTRICT_KYWD dst, const wchar_t **_RESTRICT_KYWD src,
360     size_t nwc, size_t len, mbstate_t *_RESTRICT_KYWD ps)
361 {
362 	_UTF8State *us;
363 	char buf[MB_LEN_MAX];
364 	const wchar_t *s;
365 	size_t nbytes;
366 	size_t nb;
367 
368 	us = (_UTF8State *)ps;
369 
370 	if (us->want != 0) {
371 		errno = EINVAL;
372 		return ((size_t)-1);
373 	}
374 
375 	s = *src;
376 	nbytes = 0;
377 
378 	if (dst == NULL) {
379 		while (nwc-- > 0) {
380 			if (0 <= *s && *s < 0x80)
381 				/* Fast path for plain ASCII characters. */
382 				nb = 1;
383 			else if ((nb = _UTF8_wcrtomb(buf, *s, ps)) ==
384 			    (size_t)-1)
385 				/* Invalid character - wcrtomb() sets errno. */
386 				return ((size_t)-1);
387 			if (*s == L'\0')
388 				return (nbytes + nb - 1);
389 			s++;
390 			nbytes += nb;
391 		}
392 		return (nbytes);
393 	}
394 
395 	while (len > 0 && nwc-- > 0) {
396 		if (0 <= *s && *s < 0x80) {
397 			/* Fast path for plain ASCII characters. */
398 			nb = 1;
399 			*dst = *s;
400 		} else if (len > (size_t)MB_CUR_MAX) {
401 			/* Enough space to translate in-place. */
402 			if ((nb = _UTF8_wcrtomb(dst, *s, ps)) == (size_t)-1) {
403 				*src = s;
404 				return ((size_t)-1);
405 			}
406 		} else {
407 			/*
408 			 * May not be enough space; use temp. buffer.
409 			 */
410 			if ((nb = _UTF8_wcrtomb(buf, *s, ps)) == (size_t)-1) {
411 				*src = s;
412 				return ((size_t)-1);
413 			}
414 			if (nb > (int)len)
415 				/* MB sequence for character won't fit. */
416 				break;
417 			(void) memcpy(dst, buf, nb);
418 		}
419 		if (*s == L'\0') {
420 			*src = NULL;
421 			return (nbytes + nb - 1);
422 		}
423 		s++;
424 		dst += nb;
425 		len -= nb;
426 		nbytes += nb;
427 	}
428 	*src = s;
429 	return (nbytes);
430 }
431