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