xref: /linux/fs/unicode/utf8-norm.c (revision 44f57d78)
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 
9 struct utf8data {
10 	unsigned int maxage;
11 	unsigned int offset;
12 };
13 
14 #define __INCLUDED_FROM_UTF8NORM_C__
15 #include "utf8data.h"
16 #undef __INCLUDED_FROM_UTF8NORM_C__
17 
18 int utf8version_is_supported(u8 maj, u8 min, u8 rev)
19 {
20 	int i = ARRAY_SIZE(utf8agetab) - 1;
21 	unsigned int sb_utf8version = UNICODE_AGE(maj, min, rev);
22 
23 	while (i >= 0 && utf8agetab[i] != 0) {
24 		if (sb_utf8version == utf8agetab[i])
25 			return 1;
26 		i--;
27 	}
28 	return 0;
29 }
30 EXPORT_SYMBOL(utf8version_is_supported);
31 
32 int utf8version_latest(void)
33 {
34 	return utf8vers;
35 }
36 EXPORT_SYMBOL(utf8version_latest);
37 
38 /*
39  * UTF-8 valid ranges.
40  *
41  * The UTF-8 encoding spreads the bits of a 32bit word over several
42  * bytes. This table gives the ranges that can be held and how they'd
43  * be represented.
44  *
45  * 0x00000000 0x0000007F: 0xxxxxxx
46  * 0x00000000 0x000007FF: 110xxxxx 10xxxxxx
47  * 0x00000000 0x0000FFFF: 1110xxxx 10xxxxxx 10xxxxxx
48  * 0x00000000 0x001FFFFF: 11110xxx 10xxxxxx 10xxxxxx 10xxxxxx
49  * 0x00000000 0x03FFFFFF: 111110xx 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx
50  * 0x00000000 0x7FFFFFFF: 1111110x 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx
51  *
52  * There is an additional requirement on UTF-8, in that only the
53  * shortest representation of a 32bit value is to be used.  A decoder
54  * must not decode sequences that do not satisfy this requirement.
55  * Thus the allowed ranges have a lower bound.
56  *
57  * 0x00000000 0x0000007F: 0xxxxxxx
58  * 0x00000080 0x000007FF: 110xxxxx 10xxxxxx
59  * 0x00000800 0x0000FFFF: 1110xxxx 10xxxxxx 10xxxxxx
60  * 0x00010000 0x001FFFFF: 11110xxx 10xxxxxx 10xxxxxx 10xxxxxx
61  * 0x00200000 0x03FFFFFF: 111110xx 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx
62  * 0x04000000 0x7FFFFFFF: 1111110x 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx
63  *
64  * Actual unicode characters are limited to the range 0x0 - 0x10FFFF,
65  * 17 planes of 65536 values.  This limits the sequences actually seen
66  * even more, to just the following.
67  *
68  *          0 -     0x7F: 0                   - 0x7F
69  *       0x80 -    0x7FF: 0xC2 0x80           - 0xDF 0xBF
70  *      0x800 -   0xFFFF: 0xE0 0xA0 0x80      - 0xEF 0xBF 0xBF
71  *    0x10000 - 0x10FFFF: 0xF0 0x90 0x80 0x80 - 0xF4 0x8F 0xBF 0xBF
72  *
73  * Within those ranges the surrogates 0xD800 - 0xDFFF are not allowed.
74  *
75  * Note that the longest sequence seen with valid usage is 4 bytes,
76  * the same a single UTF-32 character.  This makes the UTF-8
77  * representation of Unicode strictly smaller than UTF-32.
78  *
79  * The shortest sequence requirement was introduced by:
80  *    Corrigendum #1: UTF-8 Shortest Form
81  * It can be found here:
82  *    http://www.unicode.org/versions/corrigendum1.html
83  *
84  */
85 
86 /*
87  * Return the number of bytes used by the current UTF-8 sequence.
88  * Assumes the input points to the first byte of a valid UTF-8
89  * sequence.
90  */
91 static inline int utf8clen(const char *s)
92 {
93 	unsigned char c = *s;
94 
95 	return 1 + (c >= 0xC0) + (c >= 0xE0) + (c >= 0xF0);
96 }
97 
98 /*
99  * Decode a 3-byte UTF-8 sequence.
100  */
101 static unsigned int
102 utf8decode3(const char *str)
103 {
104 	unsigned int		uc;
105 
106 	uc = *str++ & 0x0F;
107 	uc <<= 6;
108 	uc |= *str++ & 0x3F;
109 	uc <<= 6;
110 	uc |= *str++ & 0x3F;
111 
112 	return uc;
113 }
114 
115 /*
116  * Encode a 3-byte UTF-8 sequence.
117  */
118 static int
119 utf8encode3(char *str, unsigned int val)
120 {
121 	str[2] = (val & 0x3F) | 0x80;
122 	val >>= 6;
123 	str[1] = (val & 0x3F) | 0x80;
124 	val >>= 6;
125 	str[0] = val | 0xE0;
126 
127 	return 3;
128 }
129 
130 /*
131  * utf8trie_t
132  *
133  * A compact binary tree, used to decode UTF-8 characters.
134  *
135  * Internal nodes are one byte for the node itself, and up to three
136  * bytes for an offset into the tree.  The first byte contains the
137  * following information:
138  *  NEXTBYTE  - flag        - advance to next byte if set
139  *  BITNUM    - 3 bit field - the bit number to tested
140  *  OFFLEN    - 2 bit field - number of bytes in the offset
141  * if offlen == 0 (non-branching node)
142  *  RIGHTPATH - 1 bit field - set if the following node is for the
143  *                            right-hand path (tested bit is set)
144  *  TRIENODE  - 1 bit field - set if the following node is an internal
145  *                            node, otherwise it is a leaf node
146  * if offlen != 0 (branching node)
147  *  LEFTNODE  - 1 bit field - set if the left-hand node is internal
148  *  RIGHTNODE - 1 bit field - set if the right-hand node is internal
149  *
150  * Due to the way utf8 works, there cannot be branching nodes with
151  * NEXTBYTE set, and moreover those nodes always have a righthand
152  * descendant.
153  */
154 typedef const unsigned char utf8trie_t;
155 #define BITNUM		0x07
156 #define NEXTBYTE	0x08
157 #define OFFLEN		0x30
158 #define OFFLEN_SHIFT	4
159 #define RIGHTPATH	0x40
160 #define TRIENODE	0x80
161 #define RIGHTNODE	0x40
162 #define LEFTNODE	0x80
163 
164 /*
165  * utf8leaf_t
166  *
167  * The leaves of the trie are embedded in the trie, and so the same
168  * underlying datatype: unsigned char.
169  *
170  * leaf[0]: The unicode version, stored as a generation number that is
171  *          an index into utf8agetab[].  With this we can filter code
172  *          points based on the unicode version in which they were
173  *          defined.  The CCC of a non-defined code point is 0.
174  * leaf[1]: Canonical Combining Class. During normalization, we need
175  *          to do a stable sort into ascending order of all characters
176  *          with a non-zero CCC that occur between two characters with
177  *          a CCC of 0, or at the begin or end of a string.
178  *          The unicode standard guarantees that all CCC values are
179  *          between 0 and 254 inclusive, which leaves 255 available as
180  *          a special value.
181  *          Code points with CCC 0 are known as stoppers.
182  * leaf[2]: Decomposition. If leaf[1] == 255, then leaf[2] is the
183  *          start of a NUL-terminated string that is the decomposition
184  *          of the character.
185  *          The CCC of a decomposable character is the same as the CCC
186  *          of the first character of its decomposition.
187  *          Some characters decompose as the empty string: these are
188  *          characters with the Default_Ignorable_Code_Point property.
189  *          These do affect normalization, as they all have CCC 0.
190  *
191  * The decompositions in the trie have been fully expanded, with the
192  * exception of Hangul syllables, which are decomposed algorithmically.
193  *
194  * Casefolding, if applicable, is also done using decompositions.
195  *
196  * The trie is constructed in such a way that leaves exist for all
197  * UTF-8 sequences that match the criteria from the "UTF-8 valid
198  * ranges" comment above, and only for those sequences.  Therefore a
199  * lookup in the trie can be used to validate the UTF-8 input.
200  */
201 typedef const unsigned char utf8leaf_t;
202 
203 #define LEAF_GEN(LEAF)	((LEAF)[0])
204 #define LEAF_CCC(LEAF)	((LEAF)[1])
205 #define LEAF_STR(LEAF)	((const char *)((LEAF) + 2))
206 
207 #define MINCCC		(0)
208 #define MAXCCC		(254)
209 #define STOPPER		(0)
210 #define	DECOMPOSE	(255)
211 
212 /* Marker for hangul syllable decomposition. */
213 #define HANGUL		((char)(255))
214 /* Size of the synthesized leaf used for Hangul syllable decomposition. */
215 #define UTF8HANGULLEAF	(12)
216 
217 /*
218  * Hangul decomposition (algorithm from Section 3.12 of Unicode 6.3.0)
219  *
220  * AC00;<Hangul Syllable, First>;Lo;0;L;;;;;N;;;;;
221  * D7A3;<Hangul Syllable, Last>;Lo;0;L;;;;;N;;;;;
222  *
223  * SBase = 0xAC00
224  * LBase = 0x1100
225  * VBase = 0x1161
226  * TBase = 0x11A7
227  * LCount = 19
228  * VCount = 21
229  * TCount = 28
230  * NCount = 588 (VCount * TCount)
231  * SCount = 11172 (LCount * NCount)
232  *
233  * Decomposition:
234  *   SIndex = s - SBase
235  *
236  * LV (Canonical/Full)
237  *   LIndex = SIndex / NCount
238  *   VIndex = (Sindex % NCount) / TCount
239  *   LPart = LBase + LIndex
240  *   VPart = VBase + VIndex
241  *
242  * LVT (Canonical)
243  *   LVIndex = (SIndex / TCount) * TCount
244  *   TIndex = (Sindex % TCount)
245  *   LVPart = SBase + LVIndex
246  *   TPart = TBase + TIndex
247  *
248  * LVT (Full)
249  *   LIndex = SIndex / NCount
250  *   VIndex = (Sindex % NCount) / TCount
251  *   TIndex = (Sindex % TCount)
252  *   LPart = LBase + LIndex
253  *   VPart = VBase + VIndex
254  *   if (TIndex == 0) {
255  *          d = <LPart, VPart>
256  *   } else {
257  *          TPart = TBase + TIndex
258  *          d = <LPart, TPart, VPart>
259  *   }
260  */
261 
262 /* Constants */
263 #define SB	(0xAC00)
264 #define LB	(0x1100)
265 #define VB	(0x1161)
266 #define TB	(0x11A7)
267 #define LC	(19)
268 #define VC	(21)
269 #define TC	(28)
270 #define NC	(VC * TC)
271 #define SC	(LC * NC)
272 
273 /* Algorithmic decomposition of hangul syllable. */
274 static utf8leaf_t *
275 utf8hangul(const char *str, unsigned char *hangul)
276 {
277 	unsigned int	si;
278 	unsigned int	li;
279 	unsigned int	vi;
280 	unsigned int	ti;
281 	unsigned char	*h;
282 
283 	/* Calculate the SI, LI, VI, and TI values. */
284 	si = utf8decode3(str) - SB;
285 	li = si / NC;
286 	vi = (si % NC) / TC;
287 	ti = si % TC;
288 
289 	/* Fill in base of leaf. */
290 	h = hangul;
291 	LEAF_GEN(h) = 2;
292 	LEAF_CCC(h) = DECOMPOSE;
293 	h += 2;
294 
295 	/* Add LPart, a 3-byte UTF-8 sequence. */
296 	h += utf8encode3((char *)h, li + LB);
297 
298 	/* Add VPart, a 3-byte UTF-8 sequence. */
299 	h += utf8encode3((char *)h, vi + VB);
300 
301 	/* Add TPart if required, also a 3-byte UTF-8 sequence. */
302 	if (ti)
303 		h += utf8encode3((char *)h, ti + TB);
304 
305 	/* Terminate string. */
306 	h[0] = '\0';
307 
308 	return hangul;
309 }
310 
311 /*
312  * Use trie to scan s, touching at most len bytes.
313  * Returns the leaf if one exists, NULL otherwise.
314  *
315  * A non-NULL return guarantees that the UTF-8 sequence starting at s
316  * is well-formed and corresponds to a known unicode code point.  The
317  * shorthand for this will be "is valid UTF-8 unicode".
318  */
319 static utf8leaf_t *utf8nlookup(const struct utf8data *data,
320 			       unsigned char *hangul, const char *s, size_t len)
321 {
322 	utf8trie_t	*trie = NULL;
323 	int		offlen;
324 	int		offset;
325 	int		mask;
326 	int		node;
327 
328 	if (!data)
329 		return NULL;
330 	if (len == 0)
331 		return NULL;
332 
333 	trie = utf8data + data->offset;
334 	node = 1;
335 	while (node) {
336 		offlen = (*trie & OFFLEN) >> OFFLEN_SHIFT;
337 		if (*trie & NEXTBYTE) {
338 			if (--len == 0)
339 				return NULL;
340 			s++;
341 		}
342 		mask = 1 << (*trie & BITNUM);
343 		if (*s & mask) {
344 			/* Right leg */
345 			if (offlen) {
346 				/* Right node at offset of trie */
347 				node = (*trie & RIGHTNODE);
348 				offset = trie[offlen];
349 				while (--offlen) {
350 					offset <<= 8;
351 					offset |= trie[offlen];
352 				}
353 				trie += offset;
354 			} else if (*trie & RIGHTPATH) {
355 				/* Right node after this node */
356 				node = (*trie & TRIENODE);
357 				trie++;
358 			} else {
359 				/* No right node. */
360 				return NULL;
361 			}
362 		} else {
363 			/* Left leg */
364 			if (offlen) {
365 				/* Left node after this node. */
366 				node = (*trie & LEFTNODE);
367 				trie += offlen + 1;
368 			} else if (*trie & RIGHTPATH) {
369 				/* No left node. */
370 				return NULL;
371 			} else {
372 				/* Left node after this node */
373 				node = (*trie & TRIENODE);
374 				trie++;
375 			}
376 		}
377 	}
378 	/*
379 	 * Hangul decomposition is done algorithmically. These are the
380 	 * codepoints >= 0xAC00 and <= 0xD7A3. Their UTF-8 encoding is
381 	 * always 3 bytes long, so s has been advanced twice, and the
382 	 * start of the sequence is at s-2.
383 	 */
384 	if (LEAF_CCC(trie) == DECOMPOSE && LEAF_STR(trie)[0] == HANGUL)
385 		trie = utf8hangul(s - 2, hangul);
386 	return trie;
387 }
388 
389 /*
390  * Use trie to scan s.
391  * Returns the leaf if one exists, NULL otherwise.
392  *
393  * Forwards to utf8nlookup().
394  */
395 static utf8leaf_t *utf8lookup(const struct utf8data *data,
396 			      unsigned char *hangul, const char *s)
397 {
398 	return utf8nlookup(data, hangul, s, (size_t)-1);
399 }
400 
401 /*
402  * Maximum age of any character in s.
403  * Return -1 if s is not valid UTF-8 unicode.
404  * Return 0 if only non-assigned code points are used.
405  */
406 int utf8agemax(const struct utf8data *data, const char *s)
407 {
408 	utf8leaf_t	*leaf;
409 	int		age = 0;
410 	int		leaf_age;
411 	unsigned char	hangul[UTF8HANGULLEAF];
412 
413 	if (!data)
414 		return -1;
415 
416 	while (*s) {
417 		leaf = utf8lookup(data, hangul, s);
418 		if (!leaf)
419 			return -1;
420 
421 		leaf_age = utf8agetab[LEAF_GEN(leaf)];
422 		if (leaf_age <= data->maxage && leaf_age > age)
423 			age = leaf_age;
424 		s += utf8clen(s);
425 	}
426 	return age;
427 }
428 EXPORT_SYMBOL(utf8agemax);
429 
430 /*
431  * Minimum age of any character in s.
432  * Return -1 if s is not valid UTF-8 unicode.
433  * Return 0 if non-assigned code points are used.
434  */
435 int utf8agemin(const struct utf8data *data, const char *s)
436 {
437 	utf8leaf_t	*leaf;
438 	int		age;
439 	int		leaf_age;
440 	unsigned char	hangul[UTF8HANGULLEAF];
441 
442 	if (!data)
443 		return -1;
444 	age = data->maxage;
445 	while (*s) {
446 		leaf = utf8lookup(data, hangul, s);
447 		if (!leaf)
448 			return -1;
449 		leaf_age = utf8agetab[LEAF_GEN(leaf)];
450 		if (leaf_age <= data->maxage && leaf_age < age)
451 			age = leaf_age;
452 		s += utf8clen(s);
453 	}
454 	return age;
455 }
456 EXPORT_SYMBOL(utf8agemin);
457 
458 /*
459  * Maximum age of any character in s, touch at most len bytes.
460  * Return -1 if s is not valid UTF-8 unicode.
461  */
462 int utf8nagemax(const struct utf8data *data, const char *s, size_t len)
463 {
464 	utf8leaf_t	*leaf;
465 	int		age = 0;
466 	int		leaf_age;
467 	unsigned char	hangul[UTF8HANGULLEAF];
468 
469 	if (!data)
470 		return -1;
471 
472 	while (len && *s) {
473 		leaf = utf8nlookup(data, hangul, s, len);
474 		if (!leaf)
475 			return -1;
476 		leaf_age = utf8agetab[LEAF_GEN(leaf)];
477 		if (leaf_age <= data->maxage && leaf_age > age)
478 			age = leaf_age;
479 		len -= utf8clen(s);
480 		s += utf8clen(s);
481 	}
482 	return age;
483 }
484 EXPORT_SYMBOL(utf8nagemax);
485 
486 /*
487  * Maximum age of any character in s, touch at most len bytes.
488  * Return -1 if s is not valid UTF-8 unicode.
489  */
490 int utf8nagemin(const struct utf8data *data, const char *s, size_t len)
491 {
492 	utf8leaf_t	*leaf;
493 	int		leaf_age;
494 	int		age;
495 	unsigned char	hangul[UTF8HANGULLEAF];
496 
497 	if (!data)
498 		return -1;
499 	age = data->maxage;
500 	while (len && *s) {
501 		leaf = utf8nlookup(data, hangul, s, len);
502 		if (!leaf)
503 			return -1;
504 		leaf_age = utf8agetab[LEAF_GEN(leaf)];
505 		if (leaf_age <= data->maxage && leaf_age < age)
506 			age = leaf_age;
507 		len -= utf8clen(s);
508 		s += utf8clen(s);
509 	}
510 	return age;
511 }
512 EXPORT_SYMBOL(utf8nagemin);
513 
514 /*
515  * Length of the normalization of s.
516  * Return -1 if s is not valid UTF-8 unicode.
517  *
518  * A string of Default_Ignorable_Code_Point has length 0.
519  */
520 ssize_t utf8len(const struct utf8data *data, const char *s)
521 {
522 	utf8leaf_t	*leaf;
523 	size_t		ret = 0;
524 	unsigned char	hangul[UTF8HANGULLEAF];
525 
526 	if (!data)
527 		return -1;
528 	while (*s) {
529 		leaf = utf8lookup(data, hangul, s);
530 		if (!leaf)
531 			return -1;
532 		if (utf8agetab[LEAF_GEN(leaf)] > data->maxage)
533 			ret += utf8clen(s);
534 		else if (LEAF_CCC(leaf) == DECOMPOSE)
535 			ret += strlen(LEAF_STR(leaf));
536 		else
537 			ret += utf8clen(s);
538 		s += utf8clen(s);
539 	}
540 	return ret;
541 }
542 EXPORT_SYMBOL(utf8len);
543 
544 /*
545  * Length of the normalization of s, touch at most len bytes.
546  * Return -1 if s is not valid UTF-8 unicode.
547  */
548 ssize_t utf8nlen(const struct utf8data *data, const char *s, size_t len)
549 {
550 	utf8leaf_t	*leaf;
551 	size_t		ret = 0;
552 	unsigned char	hangul[UTF8HANGULLEAF];
553 
554 	if (!data)
555 		return -1;
556 	while (len && *s) {
557 		leaf = utf8nlookup(data, hangul, s, len);
558 		if (!leaf)
559 			return -1;
560 		if (utf8agetab[LEAF_GEN(leaf)] > data->maxage)
561 			ret += utf8clen(s);
562 		else if (LEAF_CCC(leaf) == DECOMPOSE)
563 			ret += strlen(LEAF_STR(leaf));
564 		else
565 			ret += utf8clen(s);
566 		len -= utf8clen(s);
567 		s += utf8clen(s);
568 	}
569 	return ret;
570 }
571 EXPORT_SYMBOL(utf8nlen);
572 
573 /*
574  * Set up an utf8cursor for use by utf8byte().
575  *
576  *   u8c    : pointer to cursor.
577  *   data   : const struct utf8data to use for normalization.
578  *   s      : string.
579  *   len    : length of s.
580  *
581  * Returns -1 on error, 0 on success.
582  */
583 int utf8ncursor(struct utf8cursor *u8c, const struct utf8data *data,
584 		const char *s, size_t len)
585 {
586 	if (!data)
587 		return -1;
588 	if (!s)
589 		return -1;
590 	u8c->data = data;
591 	u8c->s = s;
592 	u8c->p = NULL;
593 	u8c->ss = NULL;
594 	u8c->sp = NULL;
595 	u8c->len = len;
596 	u8c->slen = 0;
597 	u8c->ccc = STOPPER;
598 	u8c->nccc = STOPPER;
599 	/* Check we didn't clobber the maximum length. */
600 	if (u8c->len != len)
601 		return -1;
602 	/* The first byte of s may not be an utf8 continuation. */
603 	if (len > 0 && (*s & 0xC0) == 0x80)
604 		return -1;
605 	return 0;
606 }
607 EXPORT_SYMBOL(utf8ncursor);
608 
609 /*
610  * Set up an utf8cursor for use by utf8byte().
611  *
612  *   u8c    : pointer to cursor.
613  *   data   : const struct utf8data to use for normalization.
614  *   s      : NUL-terminated string.
615  *
616  * Returns -1 on error, 0 on success.
617  */
618 int utf8cursor(struct utf8cursor *u8c, const struct utf8data *data,
619 	       const char *s)
620 {
621 	return utf8ncursor(u8c, data, s, (unsigned int)-1);
622 }
623 EXPORT_SYMBOL(utf8cursor);
624 
625 /*
626  * Get one byte from the normalized form of the string described by u8c.
627  *
628  * Returns the byte cast to an unsigned char on succes, and -1 on failure.
629  *
630  * The cursor keeps track of the location in the string in u8c->s.
631  * When a character is decomposed, the current location is stored in
632  * u8c->p, and u8c->s is set to the start of the decomposition. Note
633  * that bytes from a decomposition do not count against u8c->len.
634  *
635  * Characters are emitted if they match the current CCC in u8c->ccc.
636  * Hitting end-of-string while u8c->ccc == STOPPER means we're done,
637  * and the function returns 0 in that case.
638  *
639  * Sorting by CCC is done by repeatedly scanning the string.  The
640  * values of u8c->s and u8c->p are stored in u8c->ss and u8c->sp at
641  * the start of the scan.  The first pass finds the lowest CCC to be
642  * emitted and stores it in u8c->nccc, the second pass emits the
643  * characters with this CCC and finds the next lowest CCC. This limits
644  * the number of passes to 1 + the number of different CCCs in the
645  * sequence being scanned.
646  *
647  * Therefore:
648  *  u8c->p  != NULL -> a decomposition is being scanned.
649  *  u8c->ss != NULL -> this is a repeating scan.
650  *  u8c->ccc == -1   -> this is the first scan of a repeating scan.
651  */
652 int utf8byte(struct utf8cursor *u8c)
653 {
654 	utf8leaf_t *leaf;
655 	int ccc;
656 
657 	for (;;) {
658 		/* Check for the end of a decomposed character. */
659 		if (u8c->p && *u8c->s == '\0') {
660 			u8c->s = u8c->p;
661 			u8c->p = NULL;
662 		}
663 
664 		/* Check for end-of-string. */
665 		if (!u8c->p && (u8c->len == 0 || *u8c->s == '\0')) {
666 			/* There is no next byte. */
667 			if (u8c->ccc == STOPPER)
668 				return 0;
669 			/* End-of-string during a scan counts as a stopper. */
670 			ccc = STOPPER;
671 			goto ccc_mismatch;
672 		} else if ((*u8c->s & 0xC0) == 0x80) {
673 			/* This is a continuation of the current character. */
674 			if (!u8c->p)
675 				u8c->len--;
676 			return (unsigned char)*u8c->s++;
677 		}
678 
679 		/* Look up the data for the current character. */
680 		if (u8c->p) {
681 			leaf = utf8lookup(u8c->data, u8c->hangul, u8c->s);
682 		} else {
683 			leaf = utf8nlookup(u8c->data, u8c->hangul,
684 					   u8c->s, u8c->len);
685 		}
686 
687 		/* No leaf found implies that the input is a binary blob. */
688 		if (!leaf)
689 			return -1;
690 
691 		ccc = LEAF_CCC(leaf);
692 		/* Characters that are too new have CCC 0. */
693 		if (utf8agetab[LEAF_GEN(leaf)] > u8c->data->maxage) {
694 			ccc = STOPPER;
695 		} else if (ccc == DECOMPOSE) {
696 			u8c->len -= utf8clen(u8c->s);
697 			u8c->p = u8c->s + utf8clen(u8c->s);
698 			u8c->s = LEAF_STR(leaf);
699 			/* Empty decomposition implies CCC 0. */
700 			if (*u8c->s == '\0') {
701 				if (u8c->ccc == STOPPER)
702 					continue;
703 				ccc = STOPPER;
704 				goto ccc_mismatch;
705 			}
706 
707 			leaf = utf8lookup(u8c->data, u8c->hangul, u8c->s);
708 			if (!leaf)
709 				return -1;
710 			ccc = LEAF_CCC(leaf);
711 		}
712 
713 		/*
714 		 * If this is not a stopper, then see if it updates
715 		 * the next canonical class to be emitted.
716 		 */
717 		if (ccc != STOPPER && u8c->ccc < ccc && ccc < u8c->nccc)
718 			u8c->nccc = ccc;
719 
720 		/*
721 		 * Return the current byte if this is the current
722 		 * combining class.
723 		 */
724 		if (ccc == u8c->ccc) {
725 			if (!u8c->p)
726 				u8c->len--;
727 			return (unsigned char)*u8c->s++;
728 		}
729 
730 		/* Current combining class mismatch. */
731 ccc_mismatch:
732 		if (u8c->nccc == STOPPER) {
733 			/*
734 			 * Scan forward for the first canonical class
735 			 * to be emitted.  Save the position from
736 			 * which to restart.
737 			 */
738 			u8c->ccc = MINCCC - 1;
739 			u8c->nccc = ccc;
740 			u8c->sp = u8c->p;
741 			u8c->ss = u8c->s;
742 			u8c->slen = u8c->len;
743 			if (!u8c->p)
744 				u8c->len -= utf8clen(u8c->s);
745 			u8c->s += utf8clen(u8c->s);
746 		} else if (ccc != STOPPER) {
747 			/* Not a stopper, and not the ccc we're emitting. */
748 			if (!u8c->p)
749 				u8c->len -= utf8clen(u8c->s);
750 			u8c->s += utf8clen(u8c->s);
751 		} else if (u8c->nccc != MAXCCC + 1) {
752 			/* At a stopper, restart for next ccc. */
753 			u8c->ccc = u8c->nccc;
754 			u8c->nccc = MAXCCC + 1;
755 			u8c->s = u8c->ss;
756 			u8c->p = u8c->sp;
757 			u8c->len = u8c->slen;
758 		} else {
759 			/* All done, proceed from here. */
760 			u8c->ccc = STOPPER;
761 			u8c->nccc = STOPPER;
762 			u8c->sp = NULL;
763 			u8c->ss = NULL;
764 			u8c->slen = 0;
765 		}
766 	}
767 }
768 EXPORT_SYMBOL(utf8byte);
769 
770 const struct utf8data *utf8nfdi(unsigned int maxage)
771 {
772 	int i = ARRAY_SIZE(utf8nfdidata) - 1;
773 
774 	while (maxage < utf8nfdidata[i].maxage)
775 		i--;
776 	if (maxage > utf8nfdidata[i].maxage)
777 		return NULL;
778 	return &utf8nfdidata[i];
779 }
780 EXPORT_SYMBOL(utf8nfdi);
781 
782 const struct utf8data *utf8nfdicf(unsigned int maxage)
783 {
784 	int i = ARRAY_SIZE(utf8nfdicfdata) - 1;
785 
786 	while (maxage < utf8nfdicfdata[i].maxage)
787 		i--;
788 	if (maxage > utf8nfdicfdata[i].maxage)
789 		return NULL;
790 	return &utf8nfdicfdata[i];
791 }
792 EXPORT_SYMBOL(utf8nfdicf);
793