1 // SPDX-License-Identifier: GPL-2.0-only
3 * Copyright (c) 2014 SGI.
14 #define __INCLUDED_FROM_UTF8NORM_C__
16 #undef __INCLUDED_FROM_UTF8NORM_C__
18 int utf8version_is_supported(u8 maj, u8 min, u8 rev)
20 int i = ARRAY_SIZE(utf8agetab) - 1;
21 unsigned int sb_utf8version = UNICODE_AGE(maj, min, rev);
23 while (i >= 0 && utf8agetab[i] != 0) {
24 if (sb_utf8version == utf8agetab[i])
30 EXPORT_SYMBOL(utf8version_is_supported);
32 int utf8version_latest(void)
36 EXPORT_SYMBOL(utf8version_latest);
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
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
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.
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
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.
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
73 * Within those ranges the surrogates 0xD800 - 0xDFFF are not allowed.
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.
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
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
91 static inline int utf8clen(const char *s)
95 return 1 + (c >= 0xC0) + (c >= 0xE0) + (c >= 0xF0);
99 * Decode a 3-byte UTF-8 sequence.
102 utf8decode3(const char *str)
116 * Encode a 3-byte UTF-8 sequence.
119 utf8encode3(char *str, unsigned int val)
121 str[2] = (val & 0x3F) | 0x80;
123 str[1] = (val & 0x3F) | 0x80;
133 * A compact binary tree, used to decode UTF-8 characters.
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
150 * Due to the way utf8 works, there cannot be branching nodes with
151 * NEXTBYTE set, and moreover those nodes always have a righthand
154 typedef const unsigned char utf8trie_t;
156 #define NEXTBYTE 0x08
158 #define OFFLEN_SHIFT 4
159 #define RIGHTPATH 0x40
160 #define TRIENODE 0x80
161 #define RIGHTNODE 0x40
162 #define LEFTNODE 0x80
167 * The leaves of the trie are embedded in the trie, and so the same
168 * underlying datatype: unsigned char.
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
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
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.
191 * The decompositions in the trie have been fully expanded, with the
192 * exception of Hangul syllables, which are decomposed algorithmically.
194 * Casefolding, if applicable, is also done using decompositions.
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.
201 typedef const unsigned char utf8leaf_t;
203 #define LEAF_GEN(LEAF) ((LEAF)[0])
204 #define LEAF_CCC(LEAF) ((LEAF)[1])
205 #define LEAF_STR(LEAF) ((const char *)((LEAF) + 2))
210 #define DECOMPOSE (255)
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)
218 * Hangul decomposition (algorithm from Section 3.12 of Unicode 6.3.0)
220 * AC00;<Hangul Syllable, First>;Lo;0;L;;;;;N;;;;;
221 * D7A3;<Hangul Syllable, Last>;Lo;0;L;;;;;N;;;;;
230 * NCount = 588 (VCount * TCount)
231 * SCount = 11172 (LCount * NCount)
236 * LV (Canonical/Full)
237 * LIndex = SIndex / NCount
238 * VIndex = (Sindex % NCount) / TCount
239 * LPart = LBase + LIndex
240 * VPart = VBase + VIndex
243 * LVIndex = (SIndex / TCount) * TCount
244 * TIndex = (Sindex % TCount)
245 * LVPart = SBase + LVIndex
246 * TPart = TBase + TIndex
249 * LIndex = SIndex / NCount
250 * VIndex = (Sindex % NCount) / TCount
251 * TIndex = (Sindex % TCount)
252 * LPart = LBase + LIndex
253 * VPart = VBase + VIndex
257 * TPart = TBase + TIndex
258 * d = <LPart, TPart, VPart>
273 /* Algorithmic decomposition of hangul syllable. */
275 utf8hangul(const char *str, unsigned char *hangul)
283 /* Calculate the SI, LI, VI, and TI values. */
284 si = utf8decode3(str) - SB;
289 /* Fill in base of leaf. */
292 LEAF_CCC(h) = DECOMPOSE;
295 /* Add LPart, a 3-byte UTF-8 sequence. */
296 h += utf8encode3((char *)h, li + LB);
298 /* Add VPart, a 3-byte UTF-8 sequence. */
299 h += utf8encode3((char *)h, vi + VB);
301 /* Add TPart if required, also a 3-byte UTF-8 sequence. */
303 h += utf8encode3((char *)h, ti + TB);
305 /* Terminate string. */
312 * Use trie to scan s, touching at most len bytes.
313 * Returns the leaf if one exists, NULL otherwise.
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".
319 static utf8leaf_t *utf8nlookup(const struct utf8data *data,
320 unsigned char *hangul, const char *s, size_t len)
322 utf8trie_t *trie = NULL;
333 trie = utf8data + data->offset;
336 offlen = (*trie & OFFLEN) >> OFFLEN_SHIFT;
337 if (*trie & NEXTBYTE) {
342 mask = 1 << (*trie & BITNUM);
346 /* Right node at offset of trie */
347 node = (*trie & RIGHTNODE);
348 offset = trie[offlen];
351 offset |= trie[offlen];
354 } else if (*trie & RIGHTPATH) {
355 /* Right node after this node */
356 node = (*trie & TRIENODE);
365 /* Left node after this node. */
366 node = (*trie & LEFTNODE);
368 } else if (*trie & RIGHTPATH) {
372 /* Left node after this node */
373 node = (*trie & TRIENODE);
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.
384 if (LEAF_CCC(trie) == DECOMPOSE && LEAF_STR(trie)[0] == HANGUL)
385 trie = utf8hangul(s - 2, hangul);
390 * Use trie to scan s.
391 * Returns the leaf if one exists, NULL otherwise.
393 * Forwards to utf8nlookup().
395 static utf8leaf_t *utf8lookup(const struct utf8data *data,
396 unsigned char *hangul, const char *s)
398 return utf8nlookup(data, hangul, s, (size_t)-1);
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.
406 int utf8agemax(const struct utf8data *data, const char *s)
411 unsigned char hangul[UTF8HANGULLEAF];
417 leaf = utf8lookup(data, hangul, s);
421 leaf_age = utf8agetab[LEAF_GEN(leaf)];
422 if (leaf_age <= data->maxage && leaf_age > age)
428 EXPORT_SYMBOL(utf8agemax);
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.
435 int utf8agemin(const struct utf8data *data, const char *s)
440 unsigned char hangul[UTF8HANGULLEAF];
446 leaf = utf8lookup(data, hangul, s);
449 leaf_age = utf8agetab[LEAF_GEN(leaf)];
450 if (leaf_age <= data->maxage && leaf_age < age)
456 EXPORT_SYMBOL(utf8agemin);
459 * Maximum age of any character in s, touch at most len bytes.
460 * Return -1 if s is not valid UTF-8 unicode.
462 int utf8nagemax(const struct utf8data *data, const char *s, size_t len)
467 unsigned char hangul[UTF8HANGULLEAF];
473 leaf = utf8nlookup(data, hangul, s, len);
476 leaf_age = utf8agetab[LEAF_GEN(leaf)];
477 if (leaf_age <= data->maxage && leaf_age > age)
484 EXPORT_SYMBOL(utf8nagemax);
487 * Maximum age of any character in s, touch at most len bytes.
488 * Return -1 if s is not valid UTF-8 unicode.
490 int utf8nagemin(const struct utf8data *data, const char *s, size_t len)
495 unsigned char hangul[UTF8HANGULLEAF];
501 leaf = utf8nlookup(data, hangul, s, len);
504 leaf_age = utf8agetab[LEAF_GEN(leaf)];
505 if (leaf_age <= data->maxage && leaf_age < age)
512 EXPORT_SYMBOL(utf8nagemin);
515 * Length of the normalization of s.
516 * Return -1 if s is not valid UTF-8 unicode.
518 * A string of Default_Ignorable_Code_Point has length 0.
520 ssize_t utf8len(const struct utf8data *data, const char *s)
524 unsigned char hangul[UTF8HANGULLEAF];
529 leaf = utf8lookup(data, hangul, s);
532 if (utf8agetab[LEAF_GEN(leaf)] > data->maxage)
534 else if (LEAF_CCC(leaf) == DECOMPOSE)
535 ret += strlen(LEAF_STR(leaf));
542 EXPORT_SYMBOL(utf8len);
545 * Length of the normalization of s, touch at most len bytes.
546 * Return -1 if s is not valid UTF-8 unicode.
548 ssize_t utf8nlen(const struct utf8data *data, const char *s, size_t len)
552 unsigned char hangul[UTF8HANGULLEAF];
557 leaf = utf8nlookup(data, hangul, s, len);
560 if (utf8agetab[LEAF_GEN(leaf)] > data->maxage)
562 else if (LEAF_CCC(leaf) == DECOMPOSE)
563 ret += strlen(LEAF_STR(leaf));
571 EXPORT_SYMBOL(utf8nlen);
574 * Set up an utf8cursor for use by utf8byte().
576 * u8c : pointer to cursor.
577 * data : const struct utf8data to use for normalization.
581 * Returns -1 on error, 0 on success.
583 int utf8ncursor(struct utf8cursor *u8c, const struct utf8data *data,
584 const char *s, size_t len)
599 /* Check we didn't clobber the maximum length. */
602 /* The first byte of s may not be an utf8 continuation. */
603 if (len > 0 && (*s & 0xC0) == 0x80)
607 EXPORT_SYMBOL(utf8ncursor);
610 * Set up an utf8cursor for use by utf8byte().
612 * u8c : pointer to cursor.
613 * data : const struct utf8data to use for normalization.
614 * s : NUL-terminated string.
616 * Returns -1 on error, 0 on success.
618 int utf8cursor(struct utf8cursor *u8c, const struct utf8data *data,
621 return utf8ncursor(u8c, data, s, (unsigned int)-1);
623 EXPORT_SYMBOL(utf8cursor);
626 * Get one byte from the normalized form of the string described by u8c.
628 * Returns the byte cast to an unsigned char on succes, and -1 on failure.
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.
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.
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.
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.
652 int utf8byte(struct utf8cursor *u8c)
658 /* Check for the end of a decomposed character. */
659 if (u8c->p && *u8c->s == '\0') {
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)
669 /* End-of-string during a scan counts as a stopper. */
672 } else if ((*u8c->s & 0xC0) == 0x80) {
673 /* This is a continuation of the current character. */
676 return (unsigned char)*u8c->s++;
679 /* Look up the data for the current character. */
681 leaf = utf8lookup(u8c->data, u8c->hangul, u8c->s);
683 leaf = utf8nlookup(u8c->data, u8c->hangul,
687 /* No leaf found implies that the input is a binary blob. */
691 ccc = LEAF_CCC(leaf);
692 /* Characters that are too new have CCC 0. */
693 if (utf8agetab[LEAF_GEN(leaf)] > u8c->data->maxage) {
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)
707 leaf = utf8lookup(u8c->data, u8c->hangul, u8c->s);
710 ccc = LEAF_CCC(leaf);
714 * If this is not a stopper, then see if it updates
715 * the next canonical class to be emitted.
717 if (ccc != STOPPER && u8c->ccc < ccc && ccc < u8c->nccc)
721 * Return the current byte if this is the current
724 if (ccc == u8c->ccc) {
727 return (unsigned char)*u8c->s++;
730 /* Current combining class mismatch. */
732 if (u8c->nccc == STOPPER) {
734 * Scan forward for the first canonical class
735 * to be emitted. Save the position from
738 u8c->ccc = MINCCC - 1;
742 u8c->slen = u8c->len;
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. */
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;
757 u8c->len = u8c->slen;
759 /* All done, proceed from here. */
768 EXPORT_SYMBOL(utf8byte);
770 const struct utf8data *utf8nfdi(unsigned int maxage)
772 int i = ARRAY_SIZE(utf8nfdidata) - 1;
774 while (maxage < utf8nfdidata[i].maxage)
776 if (maxage > utf8nfdidata[i].maxage)
778 return &utf8nfdidata[i];
780 EXPORT_SYMBOL(utf8nfdi);
782 const struct utf8data *utf8nfdicf(unsigned int maxage)
784 int i = ARRAY_SIZE(utf8nfdicfdata) - 1;
786 while (maxage < utf8nfdicfdata[i].maxage)
788 if (maxage > utf8nfdicfdata[i].maxage)
790 return &utf8nfdicfdata[i];
792 EXPORT_SYMBOL(utf8nfdicf);