xref: /linux/fs/unicode/utf8-norm.c (revision 03ab8e6297acd1bc0eedaa050e2a1635c576fd11)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  * Copyright (c) 2014 SGI.
4  * All rights reserved.
5  */
6 
7 #include "utf8n.h"
8 
utf8version_is_supported(const struct unicode_map * um,unsigned int version)9 int utf8version_is_supported(const struct unicode_map *um, unsigned int version)
10 {
11 	int i = um->tables->utf8agetab_size - 1;
12 
13 	while (i >= 0 && um->tables->utf8agetab[i] != 0) {
14 		if (version == um->tables->utf8agetab[i])
15 			return 1;
16 		i--;
17 	}
18 	return 0;
19 }
20 
21 /*
22  * UTF-8 valid ranges.
23  *
24  * The UTF-8 encoding spreads the bits of a 32bit word over several
25  * bytes. This table gives the ranges that can be held and how they'd
26  * be represented.
27  *
28  * 0x00000000 0x0000007F: 0xxxxxxx
29  * 0x00000000 0x000007FF: 110xxxxx 10xxxxxx
30  * 0x00000000 0x0000FFFF: 1110xxxx 10xxxxxx 10xxxxxx
31  * 0x00000000 0x001FFFFF: 11110xxx 10xxxxxx 10xxxxxx 10xxxxxx
32  * 0x00000000 0x03FFFFFF: 111110xx 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx
33  * 0x00000000 0x7FFFFFFF: 1111110x 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx
34  *
35  * There is an additional requirement on UTF-8, in that only the
36  * shortest representation of a 32bit value is to be used.  A decoder
37  * must not decode sequences that do not satisfy this requirement.
38  * Thus the allowed ranges have a lower bound.
39  *
40  * 0x00000000 0x0000007F: 0xxxxxxx
41  * 0x00000080 0x000007FF: 110xxxxx 10xxxxxx
42  * 0x00000800 0x0000FFFF: 1110xxxx 10xxxxxx 10xxxxxx
43  * 0x00010000 0x001FFFFF: 11110xxx 10xxxxxx 10xxxxxx 10xxxxxx
44  * 0x00200000 0x03FFFFFF: 111110xx 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx
45  * 0x04000000 0x7FFFFFFF: 1111110x 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx
46  *
47  * Actual unicode characters are limited to the range 0x0 - 0x10FFFF,
48  * 17 planes of 65536 values.  This limits the sequences actually seen
49  * even more, to just the following.
50  *
51  *          0 -     0x7F: 0                   - 0x7F
52  *       0x80 -    0x7FF: 0xC2 0x80           - 0xDF 0xBF
53  *      0x800 -   0xFFFF: 0xE0 0xA0 0x80      - 0xEF 0xBF 0xBF
54  *    0x10000 - 0x10FFFF: 0xF0 0x90 0x80 0x80 - 0xF4 0x8F 0xBF 0xBF
55  *
56  * Within those ranges the surrogates 0xD800 - 0xDFFF are not allowed.
57  *
58  * Note that the longest sequence seen with valid usage is 4 bytes,
59  * the same a single UTF-32 character.  This makes the UTF-8
60  * representation of Unicode strictly smaller than UTF-32.
61  *
62  * The shortest sequence requirement was introduced by:
63  *    Corrigendum #1: UTF-8 Shortest Form
64  * It can be found here:
65  *    http://www.unicode.org/versions/corrigendum1.html
66  *
67  */
68 
69 /*
70  * Return the number of bytes used by the current UTF-8 sequence.
71  * Assumes the input points to the first byte of a valid UTF-8
72  * sequence.
73  */
utf8clen(const char * s)74 static inline int utf8clen(const char *s)
75 {
76 	unsigned char c = *s;
77 
78 	return 1 + (c >= 0xC0) + (c >= 0xE0) + (c >= 0xF0);
79 }
80 
81 /*
82  * Decode a 3-byte UTF-8 sequence.
83  */
84 static unsigned int
utf8decode3(const char * str)85 utf8decode3(const char *str)
86 {
87 	unsigned int		uc;
88 
89 	uc = *str++ & 0x0F;
90 	uc <<= 6;
91 	uc |= *str++ & 0x3F;
92 	uc <<= 6;
93 	uc |= *str++ & 0x3F;
94 
95 	return uc;
96 }
97 
98 /*
99  * Encode a 3-byte UTF-8 sequence.
100  */
101 static int
utf8encode3(char * str,unsigned int val)102 utf8encode3(char *str, unsigned int val)
103 {
104 	str[2] = (val & 0x3F) | 0x80;
105 	val >>= 6;
106 	str[1] = (val & 0x3F) | 0x80;
107 	val >>= 6;
108 	str[0] = val | 0xE0;
109 
110 	return 3;
111 }
112 
113 /*
114  * utf8trie_t
115  *
116  * A compact binary tree, used to decode UTF-8 characters.
117  *
118  * Internal nodes are one byte for the node itself, and up to three
119  * bytes for an offset into the tree.  The first byte contains the
120  * following information:
121  *  NEXTBYTE  - flag        - advance to next byte if set
122  *  BITNUM    - 3 bit field - the bit number to tested
123  *  OFFLEN    - 2 bit field - number of bytes in the offset
124  * if offlen == 0 (non-branching node)
125  *  RIGHTPATH - 1 bit field - set if the following node is for the
126  *                            right-hand path (tested bit is set)
127  *  TRIENODE  - 1 bit field - set if the following node is an internal
128  *                            node, otherwise it is a leaf node
129  * if offlen != 0 (branching node)
130  *  LEFTNODE  - 1 bit field - set if the left-hand node is internal
131  *  RIGHTNODE - 1 bit field - set if the right-hand node is internal
132  *
133  * Due to the way utf8 works, there cannot be branching nodes with
134  * NEXTBYTE set, and moreover those nodes always have a righthand
135  * descendant.
136  */
137 typedef const unsigned char utf8trie_t;
138 #define BITNUM		0x07
139 #define NEXTBYTE	0x08
140 #define OFFLEN		0x30
141 #define OFFLEN_SHIFT	4
142 #define RIGHTPATH	0x40
143 #define TRIENODE	0x80
144 #define RIGHTNODE	0x40
145 #define LEFTNODE	0x80
146 
147 /*
148  * utf8leaf_t
149  *
150  * The leaves of the trie are embedded in the trie, and so the same
151  * underlying datatype: unsigned char.
152  *
153  * leaf[0]: The unicode version, stored as a generation number that is
154  *          an index into ->utf8agetab[].  With this we can filter code
155  *          points based on the unicode version in which they were
156  *          defined.  The CCC of a non-defined code point is 0.
157  * leaf[1]: Canonical Combining Class. During normalization, we need
158  *          to do a stable sort into ascending order of all characters
159  *          with a non-zero CCC that occur between two characters with
160  *          a CCC of 0, or at the begin or end of a string.
161  *          The unicode standard guarantees that all CCC values are
162  *          between 0 and 254 inclusive, which leaves 255 available as
163  *          a special value.
164  *          Code points with CCC 0 are known as stoppers.
165  * leaf[2]: Decomposition. If leaf[1] == 255, then leaf[2] is the
166  *          start of a NUL-terminated string that is the decomposition
167  *          of the character.
168  *          The CCC of a decomposable character is the same as the CCC
169  *          of the first character of its decomposition.
170  *          Some characters decompose as the empty string: these are
171  *          characters with the Default_Ignorable_Code_Point property.
172  *          These do affect normalization, as they all have CCC 0.
173  *
174  * The decompositions in the trie have been fully expanded, with the
175  * exception of Hangul syllables, which are decomposed algorithmically.
176  *
177  * Casefolding, if applicable, is also done using decompositions.
178  *
179  * The trie is constructed in such a way that leaves exist for all
180  * UTF-8 sequences that match the criteria from the "UTF-8 valid
181  * ranges" comment above, and only for those sequences.  Therefore a
182  * lookup in the trie can be used to validate the UTF-8 input.
183  */
184 typedef const unsigned char utf8leaf_t;
185 
186 #define LEAF_GEN(LEAF)	((LEAF)[0])
187 #define LEAF_CCC(LEAF)	((LEAF)[1])
188 #define LEAF_STR(LEAF)	((const char *)((LEAF) + 2))
189 
190 #define MINCCC		(0)
191 #define MAXCCC		(254)
192 #define STOPPER		(0)
193 #define	DECOMPOSE	(255)
194 
195 /* Marker for hangul syllable decomposition. */
196 #define HANGUL		((char)(255))
197 /* Size of the synthesized leaf used for Hangul syllable decomposition. */
198 #define UTF8HANGULLEAF	(12)
199 
200 /*
201  * Hangul decomposition (algorithm from Section 3.12 of Unicode 6.3.0)
202  *
203  * AC00;<Hangul Syllable, First>;Lo;0;L;;;;;N;;;;;
204  * D7A3;<Hangul Syllable, Last>;Lo;0;L;;;;;N;;;;;
205  *
206  * SBase = 0xAC00
207  * LBase = 0x1100
208  * VBase = 0x1161
209  * TBase = 0x11A7
210  * LCount = 19
211  * VCount = 21
212  * TCount = 28
213  * NCount = 588 (VCount * TCount)
214  * SCount = 11172 (LCount * NCount)
215  *
216  * Decomposition:
217  *   SIndex = s - SBase
218  *
219  * LV (Canonical/Full)
220  *   LIndex = SIndex / NCount
221  *   VIndex = (Sindex % NCount) / TCount
222  *   LPart = LBase + LIndex
223  *   VPart = VBase + VIndex
224  *
225  * LVT (Canonical)
226  *   LVIndex = (SIndex / TCount) * TCount
227  *   TIndex = (Sindex % TCount)
228  *   LVPart = SBase + LVIndex
229  *   TPart = TBase + TIndex
230  *
231  * LVT (Full)
232  *   LIndex = SIndex / NCount
233  *   VIndex = (Sindex % NCount) / TCount
234  *   TIndex = (Sindex % TCount)
235  *   LPart = LBase + LIndex
236  *   VPart = VBase + VIndex
237  *   if (TIndex == 0) {
238  *          d = <LPart, VPart>
239  *   } else {
240  *          TPart = TBase + TIndex
241  *          d = <LPart, TPart, VPart>
242  *   }
243  */
244 
245 /* Constants */
246 #define SB	(0xAC00)
247 #define LB	(0x1100)
248 #define VB	(0x1161)
249 #define TB	(0x11A7)
250 #define LC	(19)
251 #define VC	(21)
252 #define TC	(28)
253 #define NC	(VC * TC)
254 #define SC	(LC * NC)
255 
256 /* Algorithmic decomposition of hangul syllable. */
257 static utf8leaf_t *
utf8hangul(const char * str,unsigned char * hangul)258 utf8hangul(const char *str, unsigned char *hangul)
259 {
260 	unsigned int	si;
261 	unsigned int	li;
262 	unsigned int	vi;
263 	unsigned int	ti;
264 	unsigned char	*h;
265 
266 	/* Calculate the SI, LI, VI, and TI values. */
267 	si = utf8decode3(str) - SB;
268 	li = si / NC;
269 	vi = (si % NC) / TC;
270 	ti = si % TC;
271 
272 	/* Fill in base of leaf. */
273 	h = hangul;
274 	LEAF_GEN(h) = 2;
275 	LEAF_CCC(h) = DECOMPOSE;
276 	h += 2;
277 
278 	/* Add LPart, a 3-byte UTF-8 sequence. */
279 	h += utf8encode3((char *)h, li + LB);
280 
281 	/* Add VPart, a 3-byte UTF-8 sequence. */
282 	h += utf8encode3((char *)h, vi + VB);
283 
284 	/* Add TPart if required, also a 3-byte UTF-8 sequence. */
285 	if (ti)
286 		h += utf8encode3((char *)h, ti + TB);
287 
288 	/* Terminate string. */
289 	h[0] = '\0';
290 
291 	return hangul;
292 }
293 
294 /*
295  * Use trie to scan s, touching at most len bytes.
296  * Returns the leaf if one exists, NULL otherwise.
297  *
298  * A non-NULL return guarantees that the UTF-8 sequence starting at s
299  * is well-formed and corresponds to a known unicode code point.  The
300  * shorthand for this will be "is valid UTF-8 unicode".
301  */
utf8nlookup(const struct unicode_map * um,enum utf8_normalization n,unsigned char * hangul,const char * s,size_t len)302 static utf8leaf_t *utf8nlookup(const struct unicode_map *um,
303 		enum utf8_normalization n, unsigned char *hangul, const char *s,
304 		size_t len)
305 {
306 	utf8trie_t	*trie = um->tables->utf8data + um->ntab[n]->offset;
307 	int		offlen;
308 	int		offset;
309 	int		mask;
310 	int		node;
311 
312 	if (len == 0)
313 		return NULL;
314 
315 	node = 1;
316 	while (node) {
317 		offlen = (*trie & OFFLEN) >> OFFLEN_SHIFT;
318 		if (*trie & NEXTBYTE) {
319 			if (--len == 0)
320 				return NULL;
321 			s++;
322 		}
323 		mask = 1 << (*trie & BITNUM);
324 		if (*s & mask) {
325 			/* Right leg */
326 			if (offlen) {
327 				/* Right node at offset of trie */
328 				node = (*trie & RIGHTNODE);
329 				offset = trie[offlen];
330 				while (--offlen) {
331 					offset <<= 8;
332 					offset |= trie[offlen];
333 				}
334 				trie += offset;
335 			} else if (*trie & RIGHTPATH) {
336 				/* Right node after this node */
337 				node = (*trie & TRIENODE);
338 				trie++;
339 			} else {
340 				/* No right node. */
341 				return NULL;
342 			}
343 		} else {
344 			/* Left leg */
345 			if (offlen) {
346 				/* Left node after this node. */
347 				node = (*trie & LEFTNODE);
348 				trie += offlen + 1;
349 			} else if (*trie & RIGHTPATH) {
350 				/* No left node. */
351 				return NULL;
352 			} else {
353 				/* Left node after this node */
354 				node = (*trie & TRIENODE);
355 				trie++;
356 			}
357 		}
358 	}
359 	/*
360 	 * Hangul decomposition is done algorithmically. These are the
361 	 * codepoints >= 0xAC00 and <= 0xD7A3. Their UTF-8 encoding is
362 	 * always 3 bytes long, so s has been advanced twice, and the
363 	 * start of the sequence is at s-2.
364 	 */
365 	if (LEAF_CCC(trie) == DECOMPOSE && LEAF_STR(trie)[0] == HANGUL)
366 		trie = utf8hangul(s - 2, hangul);
367 	return trie;
368 }
369 
370 /*
371  * Use trie to scan s.
372  * Returns the leaf if one exists, NULL otherwise.
373  *
374  * Forwards to utf8nlookup().
375  */
utf8lookup(const struct unicode_map * um,enum utf8_normalization n,unsigned char * hangul,const char * s)376 static utf8leaf_t *utf8lookup(const struct unicode_map *um,
377 		enum utf8_normalization n, unsigned char *hangul, const char *s)
378 {
379 	return utf8nlookup(um, n, hangul, s, (size_t)-1);
380 }
381 
382 /*
383  * Length of the normalization of s, touch at most len bytes.
384  * Return -1 if s is not valid UTF-8 unicode.
385  */
utf8nlen(const struct unicode_map * um,enum utf8_normalization n,const char * s,size_t len)386 ssize_t utf8nlen(const struct unicode_map *um, enum utf8_normalization n,
387 		const char *s, size_t len)
388 {
389 	utf8leaf_t	*leaf;
390 	size_t		ret = 0;
391 	unsigned char	hangul[UTF8HANGULLEAF];
392 
393 	while (len && *s) {
394 		leaf = utf8nlookup(um, n, hangul, s, len);
395 		if (!leaf)
396 			return -1;
397 		if (um->tables->utf8agetab[LEAF_GEN(leaf)] >
398 		    um->ntab[n]->maxage)
399 			ret += utf8clen(s);
400 		else if (LEAF_CCC(leaf) == DECOMPOSE)
401 			ret += strlen(LEAF_STR(leaf));
402 		else
403 			ret += utf8clen(s);
404 		len -= utf8clen(s);
405 		s += utf8clen(s);
406 	}
407 	return ret;
408 }
409 
410 /*
411  * Set up an utf8cursor for use by utf8byte().
412  *
413  *   u8c    : pointer to cursor.
414  *   data   : const struct utf8data to use for normalization.
415  *   s      : string.
416  *   len    : length of s.
417  *
418  * Returns -1 on error, 0 on success.
419  */
utf8ncursor(struct utf8cursor * u8c,const struct unicode_map * um,enum utf8_normalization n,const char * s,size_t len)420 int utf8ncursor(struct utf8cursor *u8c, const struct unicode_map *um,
421 		enum utf8_normalization n, const char *s, size_t len)
422 {
423 	if (!s)
424 		return -1;
425 	u8c->um = um;
426 	u8c->n = n;
427 	u8c->s = s;
428 	u8c->p = NULL;
429 	u8c->ss = NULL;
430 	u8c->sp = NULL;
431 	u8c->len = len;
432 	u8c->slen = 0;
433 	u8c->ccc = STOPPER;
434 	u8c->nccc = STOPPER;
435 	/* Check we didn't clobber the maximum length. */
436 	if (u8c->len != len)
437 		return -1;
438 	/* The first byte of s may not be an utf8 continuation. */
439 	if (len > 0 && (*s & 0xC0) == 0x80)
440 		return -1;
441 	return 0;
442 }
443 
444 /*
445  * Get one byte from the normalized form of the string described by u8c.
446  *
447  * Returns the byte cast to an unsigned char on succes, and -1 on failure.
448  *
449  * The cursor keeps track of the location in the string in u8c->s.
450  * When a character is decomposed, the current location is stored in
451  * u8c->p, and u8c->s is set to the start of the decomposition. Note
452  * that bytes from a decomposition do not count against u8c->len.
453  *
454  * Characters are emitted if they match the current CCC in u8c->ccc.
455  * Hitting end-of-string while u8c->ccc == STOPPER means we're done,
456  * and the function returns 0 in that case.
457  *
458  * Sorting by CCC is done by repeatedly scanning the string.  The
459  * values of u8c->s and u8c->p are stored in u8c->ss and u8c->sp at
460  * the start of the scan.  The first pass finds the lowest CCC to be
461  * emitted and stores it in u8c->nccc, the second pass emits the
462  * characters with this CCC and finds the next lowest CCC. This limits
463  * the number of passes to 1 + the number of different CCCs in the
464  * sequence being scanned.
465  *
466  * Therefore:
467  *  u8c->p  != NULL -> a decomposition is being scanned.
468  *  u8c->ss != NULL -> this is a repeating scan.
469  *  u8c->ccc == -1   -> this is the first scan of a repeating scan.
470  */
utf8byte(struct utf8cursor * u8c)471 int utf8byte(struct utf8cursor *u8c)
472 {
473 	utf8leaf_t *leaf;
474 	int ccc;
475 
476 	for (;;) {
477 		/* Check for the end of a decomposed character. */
478 		if (u8c->p && *u8c->s == '\0') {
479 			u8c->s = u8c->p;
480 			u8c->p = NULL;
481 		}
482 
483 		/* Check for end-of-string. */
484 		if (!u8c->p && (u8c->len == 0 || *u8c->s == '\0')) {
485 			/* There is no next byte. */
486 			if (u8c->ccc == STOPPER)
487 				return 0;
488 			/* End-of-string during a scan counts as a stopper. */
489 			ccc = STOPPER;
490 			goto ccc_mismatch;
491 		} else if ((*u8c->s & 0xC0) == 0x80) {
492 			/* This is a continuation of the current character. */
493 			if (!u8c->p)
494 				u8c->len--;
495 			return (unsigned char)*u8c->s++;
496 		}
497 
498 		/* Look up the data for the current character. */
499 		if (u8c->p) {
500 			leaf = utf8lookup(u8c->um, u8c->n, u8c->hangul, u8c->s);
501 		} else {
502 			leaf = utf8nlookup(u8c->um, u8c->n, u8c->hangul,
503 					   u8c->s, u8c->len);
504 		}
505 
506 		/* No leaf found implies that the input is a binary blob. */
507 		if (!leaf)
508 			return -1;
509 
510 		ccc = LEAF_CCC(leaf);
511 		/* Characters that are too new have CCC 0. */
512 		if (u8c->um->tables->utf8agetab[LEAF_GEN(leaf)] >
513 		    u8c->um->ntab[u8c->n]->maxage) {
514 			ccc = STOPPER;
515 		} else if (ccc == DECOMPOSE) {
516 			u8c->len -= utf8clen(u8c->s);
517 			u8c->p = u8c->s + utf8clen(u8c->s);
518 			u8c->s = LEAF_STR(leaf);
519 			/* Empty decomposition implies CCC 0. */
520 			if (*u8c->s == '\0') {
521 				if (u8c->ccc == STOPPER)
522 					continue;
523 				ccc = STOPPER;
524 				goto ccc_mismatch;
525 			}
526 
527 			leaf = utf8lookup(u8c->um, u8c->n, u8c->hangul, u8c->s);
528 			if (!leaf)
529 				return -1;
530 			ccc = LEAF_CCC(leaf);
531 		}
532 
533 		/*
534 		 * If this is not a stopper, then see if it updates
535 		 * the next canonical class to be emitted.
536 		 */
537 		if (ccc != STOPPER && u8c->ccc < ccc && ccc < u8c->nccc)
538 			u8c->nccc = ccc;
539 
540 		/*
541 		 * Return the current byte if this is the current
542 		 * combining class.
543 		 */
544 		if (ccc == u8c->ccc) {
545 			if (!u8c->p)
546 				u8c->len--;
547 			return (unsigned char)*u8c->s++;
548 		}
549 
550 		/* Current combining class mismatch. */
551 ccc_mismatch:
552 		if (u8c->nccc == STOPPER) {
553 			/*
554 			 * Scan forward for the first canonical class
555 			 * to be emitted.  Save the position from
556 			 * which to restart.
557 			 */
558 			u8c->ccc = MINCCC - 1;
559 			u8c->nccc = ccc;
560 			u8c->sp = u8c->p;
561 			u8c->ss = u8c->s;
562 			u8c->slen = u8c->len;
563 			if (!u8c->p)
564 				u8c->len -= utf8clen(u8c->s);
565 			u8c->s += utf8clen(u8c->s);
566 		} else if (ccc != STOPPER) {
567 			/* Not a stopper, and not the ccc we're emitting. */
568 			if (!u8c->p)
569 				u8c->len -= utf8clen(u8c->s);
570 			u8c->s += utf8clen(u8c->s);
571 		} else if (u8c->nccc != MAXCCC + 1) {
572 			/* At a stopper, restart for next ccc. */
573 			u8c->ccc = u8c->nccc;
574 			u8c->nccc = MAXCCC + 1;
575 			u8c->s = u8c->ss;
576 			u8c->p = u8c->sp;
577 			u8c->len = u8c->slen;
578 		} else {
579 			/* All done, proceed from here. */
580 			u8c->ccc = STOPPER;
581 			u8c->nccc = STOPPER;
582 			u8c->sp = NULL;
583 			u8c->ss = NULL;
584 			u8c->slen = 0;
585 		}
586 	}
587 }
588 
589 #ifdef CONFIG_UNICODE_NORMALIZATION_SELFTEST_MODULE
590 EXPORT_SYMBOL_GPL(utf8version_is_supported);
591 EXPORT_SYMBOL_GPL(utf8nlen);
592 EXPORT_SYMBOL_GPL(utf8ncursor);
593 EXPORT_SYMBOL_GPL(utf8byte);
594 #endif
595