lib/util/charset/iconv.c

   1 /*
   2    Unix SMB/CIFS implementation.
   3    minimal iconv implementation
   4    Copyright (C) Andrew Tridgell 2001
   5    Copyright (C) Jelmer Vernooij 2002
   6
   7    This program is free software; you can redistribute it and/or modify
   8    it under the terms of the GNU General Public License as published by
   9    the Free Software Foundation; either version 3 of the License, or
  10    (at your option) any later version.
  11
  12    This program is distributed in the hope that it will be useful,
  13    but WITHOUT ANY WARRANTY; without even the implied warranty of
  14    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  15    GNU General Public License for more details.
  16
  17    You should have received a copy of the GNU General Public License
  18    along with this program.  If not, see <http://www.gnu.org/licenses/>.
  19 */
  20
  21 #include "includes.h"
  22 #include "../lib/util/dlinklist.h"
  23 #include "system/iconv.h"
  24 #include "system/filesys.h"
  25
  26 #ifdef strcasecmp
  27 #undef strcasecmp
  28 #endif
  29
  30 #ifdef static_decl_charset
  31 static_decl_charset;
  32 #endif
  33
  34 /**
  35  * @file
  36  *
  37  * @brief Samba wrapper/stub for iconv character set conversion.
  38  *
  39  * iconv is the XPG2 interface for converting between character
  40  * encodings.  This file provides a Samba wrapper around it, and also
  41  * a simple reimplementation that is used if the system does not
  42  * implement iconv.
  43  *
  44  * Samba only works with encodings that are supersets of ASCII: ascii
  45  * characters like whitespace can be tested for directly, multibyte
  46  * sequences start with a byte with the high bit set, and strings are
  47  * terminated by a nul byte.
  48  *
  49  * Note that the only function provided by iconv is conversion between
  50  * characters.  It doesn't directly support operations like
  51  * uppercasing or comparison.  We have to convert to UTF-16LE and
  52  * compare there.
  53  *
  54  * @sa Samba Developers Guide
  55  **/
  56
  57 static size_t ascii_pull  (void *,const char **, size_t *, char **, size_t *);
  58 static size_t ascii_push  (void *,const char **, size_t *, char **, size_t *);
  59 static size_t latin1_push(void *,const char **, size_t *, char **, size_t *);
  60 static size_t utf8_pull   (void *,const char **, size_t *, char **, size_t *);
  61 static size_t utf8_push   (void *,const char **, size_t *, char **, size_t *);
  62 static size_t utf16_munged_pull(void *,const char **, size_t *, char **, size_t *);
  63 static size_t ucs2hex_pull(void *,const char **, size_t *, char **, size_t *);
  64 static size_t ucs2hex_push(void *,const char **, size_t *, char **, size_t *);
  65 static size_t iconv_copy  (void *,const char **, size_t *, char **, size_t *);
  66 static size_t iconv_swab  (void *,const char **, size_t *, char **, size_t *);
  67
  68 static const struct charset_functions builtin_functions[] = {
  69         /* windows is closest to UTF-16 */
  70         {"UCS-2LE",  iconv_copy, iconv_copy},
  71         {"UTF-16LE",  iconv_copy, iconv_copy},
  72         {"UCS-2BE",  iconv_swab, iconv_swab},
  73         {"UTF-16BE",  iconv_swab, iconv_swab},
  74
  75         /* we include the UTF-8 alias to cope with differing locale settings */
  76         {"UTF8",   utf8_pull,  utf8_push},
  77         {"UTF-8",   utf8_pull,  utf8_push},
  78
  79         /* this handles the munging needed for String2Key */
  80         {"UTF16_MUNGED",   utf16_munged_pull,  iconv_copy},
  81
  82         {"ASCII", ascii_pull, ascii_push},
  83         {"646", ascii_pull, ascii_push},
  84         {"ISO-8859-1", ascii_pull, latin1_push},
  85         {"UCS2-HEX", ucs2hex_pull, ucs2hex_push}
  86 };
  87
  88 static struct charset_functions *charsets = NULL;
  89
  90 static struct charset_functions *find_charset_functions(const char *name)
  91 {
  92         struct charset_functions *c;
  93
  94         /* Check whether we already have this charset... */
  95         for (c = charsets; c != NULL; c = c->next) {
  96                 if(strcasecmp(c->name, name) == 0) {
  97                         return c;
  98                 }
  99                 c = c->next;
 100         }
 101
 102         return NULL;
 103 }
 104
 105 bool smb_register_charset(const struct charset_functions *funcs_in)
 106 {
 107         struct charset_functions *funcs;
 108
 109         DEBUG(5, ("Attempting to register new charset %s\n", funcs_in->name));
 110         /* Check whether we already have this charset... */
 111         if (find_charset_functions(funcs_in->name)) {
 112                 DEBUG(0, ("Duplicate charset %s, not registering\n", funcs_in->name));
 113                 return false;
 114         }
 115
 116         funcs = talloc(NULL, struct charset_functions);
 117         if (!funcs) {
 118                 DEBUG(0, ("Out of memory duplicating charset %s\n", funcs_in->name));
 119                 return false;
 120         }
 121         *funcs = *funcs_in;
 122
 123         funcs->next = funcs->prev = NULL;
 124         DEBUG(5, ("Registered charset %s\n", funcs->name));
 125         DLIST_ADD(charsets, funcs);
 126         return true;
 127 }
 128
 129 static void lazy_initialize_iconv(void)
 130 {
 131 #ifdef static_init_charset
 132         static bool initialized = false;
 133
 134         if (!initialized) {
 135                 static_init_charset;
 136                 initialized = true;
 137         }
 138 #endif
 139 }
 140
 141 #ifdef HAVE_NATIVE_ICONV
 142 /* if there was an error then reset the internal state,
 143    this ensures that we don't have a shift state remaining for
 144    character sets like SJIS */
 145 static size_t sys_iconv(void *cd,
 146                         const char **inbuf, size_t *inbytesleft,
 147                         char **outbuf, size_t *outbytesleft)
 148 {
 149         size_t ret = iconv((iconv_t)cd,
 150                            discard_const_p(char *, inbuf), inbytesleft,
 151                            outbuf, outbytesleft);
 152         if (ret == (size_t)-1) iconv(cd, NULL, NULL, NULL, NULL);
 153         return ret;
 154 }
 155 #endif
 156
 157 /**
 158  * This is a simple portable iconv() implementaion.
 159  *
 160  * It only knows about a very small number of character sets - just
 161  * enough that Samba works on systems that don't have iconv.
 162  **/
 163 _PUBLIC_ size_t smb_iconv(smb_iconv_t cd,
 164                  const char **inbuf, size_t *inbytesleft,
 165                  char **outbuf, size_t *outbytesleft)
 166 {
 167         /* in many cases we can go direct */
 168         if (cd->direct) {
 169                 return cd->direct(cd->cd_direct,
 170                                   inbuf, inbytesleft, outbuf, outbytesleft);
 171         }
 172
 173         /* otherwise we have to do it chunks at a time */
 174         {
 175 #ifndef SMB_ICONV_BUFSIZE
 176 #define SMB_ICONV_BUFSIZE 2048
 177 #endif
 178                 size_t bufsize;
 179                 char *cvtbuf = talloc_array(cd, char, SMB_ICONV_BUFSIZE);
 180
 181                 if (!cvtbuf) {
 182                         return (size_t)-1;
 183                 }
 184
 185                 while (*inbytesleft > 0) {
 186                         char *bufp1 = cvtbuf;
 187                         const char *bufp2 = cvtbuf;
 188
 189                         bufsize = SMB_ICONV_BUFSIZE;
 190
 191                         if (cd->pull(cd->cd_pull,
 192                                      inbuf, inbytesleft, &bufp1, &bufsize) == -1
 193                             && errno != E2BIG) {
 194                                 talloc_free(cvtbuf);
 195                                 return -1;
 196                         }
 197
 198                         bufsize = SMB_ICONV_BUFSIZE - bufsize;
 199
 200                         if (cd->push(cd->cd_push,
 201                                      &bufp2, &bufsize,
 202                                      outbuf, outbytesleft) == -1) {
 203                                 talloc_free(cvtbuf);
 204                                 return -1;
 205                         }
 206                 }
 207                 talloc_free(cvtbuf);
 208         }
 209
 210         return 0;
 211 }
 212
 213 static bool is_utf16(const char *name)
 214 {
 215         return strcasecmp(name, "UCS-2LE") == 0 ||
 216                 strcasecmp(name, "UTF-16LE") == 0;
 217 }
 218
 219 static int smb_iconv_t_destructor(smb_iconv_t hwd)
 220 {
 221 #ifdef HAVE_NATIVE_ICONV
 222         if (hwd->cd_pull != NULL && hwd->cd_pull != (iconv_t)-1)
 223                 iconv_close(hwd->cd_pull);
 224         if (hwd->cd_push != NULL && hwd->cd_push != (iconv_t)-1)
 225                 iconv_close(hwd->cd_push);
 226         if (hwd->cd_direct != NULL && hwd->cd_direct != (iconv_t)-1)
 227                 iconv_close(hwd->cd_direct);
 228 #endif
 229
 230         return 0;
 231 }
 232
 233 _PUBLIC_ smb_iconv_t smb_iconv_open_ex(TALLOC_CTX *mem_ctx, const char *tocode,
 234                               const char *fromcode, bool native_iconv)
 235 {
 236         smb_iconv_t ret;
 237         const struct charset_functions *from=NULL, *to=NULL;
 238         int i;
 239
 240         lazy_initialize_iconv();
 241
 242         ret = (smb_iconv_t)talloc_named(mem_ctx,
 243                                         sizeof(*ret),
 244                                         "iconv(%s,%s)", tocode, fromcode);
 245         if (!ret) {
 246                 errno = ENOMEM;
 247                 return (smb_iconv_t)-1;
 248         }
 249         memset(ret, 0, sizeof(*ret));
 250         talloc_set_destructor(ret, smb_iconv_t_destructor);
 251
 252         /* check for the simplest null conversion */
 253         if (strcmp(fromcode, tocode) == 0) {
 254                 ret->direct = iconv_copy;
 255                 return ret;
 256         }
 257
 258         for (i=0;i<ARRAY_SIZE(builtin_functions);i++) {
 259                 if (strcasecmp(fromcode, builtin_functions[i].name) == 0) {
 260                         from = &builtin_functions[i];
 261                 }
 262                 if (strcasecmp(tocode, builtin_functions[i].name) == 0) {
 263                         to = &builtin_functions[i];
 264                 }
 265         }
 266
 267         if (from == NULL) {
 268                 for (from=charsets; from; from=from->next) {
 269                         if (strcasecmp(from->name, fromcode) == 0) break;
 270                 }
 271         }
 272
 273         if (to == NULL) {
 274                 for (to=charsets; to; to=to->next) {
 275                         if (strcasecmp(to->name, tocode) == 0) break;
 276                 }
 277         }
 278
 279 #ifdef HAVE_NATIVE_ICONV
 280         if ((!from || !to) && !native_iconv) {
 281                 goto failed;
 282         }
 283         if (!from) {
 284                 ret->pull = sys_iconv;
 285                 ret->cd_pull = iconv_open("UTF-16LE", fromcode);
 286                 if (ret->cd_pull == (iconv_t)-1)
 287                         ret->cd_pull = iconv_open("UCS-2LE", fromcode);
 288                 if (ret->cd_pull == (iconv_t)-1) goto failed;
 289         }
 290
 291         if (!to) {
 292                 ret->push = sys_iconv;
 293                 ret->cd_push = iconv_open(tocode, "UTF-16LE");
 294                 if (ret->cd_push == (iconv_t)-1)
 295                         ret->cd_push = iconv_open(tocode, "UCS-2LE");
 296                 if (ret->cd_push == (iconv_t)-1) goto failed;
 297         }
 298 #else
 299         if (!from || !to) {
 300                 goto failed;
 301         }
 302 #endif
 303
 304         /* check for conversion to/from ucs2 */
 305         if (is_utf16(fromcode) && to) {
 306                 ret->direct = to->push;
 307                 return ret;
 308         }
 309         if (is_utf16(tocode) && from) {
 310                 ret->direct = from->pull;
 311                 return ret;
 312         }
 313
 314 #ifdef HAVE_NATIVE_ICONV
 315         if (is_utf16(fromcode)) {
 316                 ret->direct = sys_iconv;
 317                 ret->cd_direct = ret->cd_push;
 318                 ret->cd_push = NULL;
 319                 return ret;
 320         }
 321         if (is_utf16(tocode)) {
 322                 ret->direct = sys_iconv;
 323                 ret->cd_direct = ret->cd_pull;
 324                 ret->cd_pull = NULL;
 325                 return ret;
 326         }
 327 #endif
 328
 329         /* the general case has to go via a buffer */
 330         if (!ret->pull) ret->pull = from->pull;
 331         if (!ret->push) ret->push = to->push;
 332         return ret;
 333
 334 failed:
 335         talloc_free(ret);
 336         errno = EINVAL;
 337         return (smb_iconv_t)-1;
 338 }
 339
 340 /*
 341   simple iconv_open() wrapper
 342  */
 343 _PUBLIC_ smb_iconv_t smb_iconv_open(const char *tocode, const char *fromcode)
 344 {
 345         return smb_iconv_open_ex(NULL, tocode, fromcode, true);
 346 }
 347
 348 /*
 349   simple iconv_close() wrapper
 350 */
 351 _PUBLIC_ int smb_iconv_close(smb_iconv_t cd)
 352 {
 353         talloc_free(cd);
 354         return 0;
 355 }
 356
 357
 358 /**********************************************************************
 359  the following functions implement the builtin character sets in Samba
 360  and also the "test" character sets that are designed to test
 361  multi-byte character set support for english users
 362 ***********************************************************************/
 363 static size_t ascii_pull(void *cd, const char **inbuf, size_t *inbytesleft,
 364                          char **outbuf, size_t *outbytesleft)
 365 {
 366         while (*inbytesleft >= 1 && *outbytesleft >= 2) {
 367                 (*outbuf)[0] = (*inbuf)[0];
 368                 (*outbuf)[1] = 0;
 369                 (*inbytesleft)  -= 1;
 370                 (*outbytesleft) -= 2;
 371                 (*inbuf)  += 1;
 372                 (*outbuf) += 2;
 373         }
 374
 375         if (*inbytesleft > 0) {
 376                 errno = E2BIG;
 377                 return -1;
 378         }
 379
 380         return 0;
 381 }
 382
 383 static size_t ascii_push(void *cd, const char **inbuf, size_t *inbytesleft,
 384                          char **outbuf, size_t *outbytesleft)
 385 {
 386         int ir_count=0;
 387
 388         while (*inbytesleft >= 2 && *outbytesleft >= 1) {
 389                 (*outbuf)[0] = (*inbuf)[0] & 0x7F;
 390                 if ((*inbuf)[1]) ir_count++;
 391                 (*inbytesleft)  -= 2;
 392                 (*outbytesleft) -= 1;
 393                 (*inbuf)  += 2;
 394                 (*outbuf) += 1;
 395         }
 396
 397         if (*inbytesleft == 1) {
 398                 errno = EINVAL;
 399                 return -1;
 400         }
 401
 402         if (*inbytesleft > 1) {
 403                 errno = E2BIG;
 404                 return -1;
 405         }
 406
 407         return ir_count;
 408 }
 409
 410 static size_t latin1_push(void *cd, const char **inbuf, size_t *inbytesleft,
 411                          char **outbuf, size_t *outbytesleft)
 412 {
 413         int ir_count=0;
 414
 415         while (*inbytesleft >= 2 && *outbytesleft >= 1) {
 416                 (*outbuf)[0] = (*inbuf)[0];
 417                 if ((*inbuf)[1]) ir_count++;
 418                 (*inbytesleft)  -= 2;
 419                 (*outbytesleft) -= 1;
 420                 (*inbuf)  += 2;
 421                 (*outbuf) += 1;
 422         }
 423
 424         if (*inbytesleft == 1) {
 425                 errno = EINVAL;
 426                 return -1;
 427         }
 428
 429         if (*inbytesleft > 1) {
 430                 errno = E2BIG;
 431                 return -1;
 432         }
 433
 434         return ir_count;
 435 }
 436
 437 static size_t ucs2hex_pull(void *cd, const char **inbuf, size_t *inbytesleft,
 438                          char **outbuf, size_t *outbytesleft)
 439 {
 440         while (*inbytesleft >= 1 && *outbytesleft >= 2) {
 441                 unsigned int v;
 442
 443                 if ((*inbuf)[0] != '@') {
 444                         /* seven bit ascii case */
 445                         (*outbuf)[0] = (*inbuf)[0];
 446                         (*outbuf)[1] = 0;
 447                         (*inbytesleft)  -= 1;
 448                         (*outbytesleft) -= 2;
 449                         (*inbuf)  += 1;
 450                         (*outbuf) += 2;
 451                         continue;
 452                 }
 453                 /* it's a hex character */
 454                 if (*inbytesleft < 5) {
 455                         errno = EINVAL;
 456                         return -1;
 457                 }
 458
 459                 if (sscanf(&(*inbuf)[1], "%04x", &v) != 1) {
 460                         errno = EILSEQ;
 461                         return -1;
 462                 }
 463
 464                 (*outbuf)[0] = v&0xff;
 465                 (*outbuf)[1] = v>>8;
 466                 (*inbytesleft)  -= 5;
 467                 (*outbytesleft) -= 2;
 468                 (*inbuf)  += 5;
 469                 (*outbuf) += 2;
 470         }
 471
 472         if (*inbytesleft > 0) {
 473                 errno = E2BIG;
 474                 return -1;
 475         }
 476
 477         return 0;
 478 }
 479
 480 static size_t ucs2hex_push(void *cd, const char **inbuf, size_t *inbytesleft,
 481                            char **outbuf, size_t *outbytesleft)
 482 {
 483         while (*inbytesleft >= 2 && *outbytesleft >= 1) {
 484                 char buf[6];
 485
 486                 if ((*inbuf)[1] == 0 &&
 487                     ((*inbuf)[0] & 0x80) == 0 &&
 488                     (*inbuf)[0] != '@') {
 489                         (*outbuf)[0] = (*inbuf)[0];
 490                         (*inbytesleft)  -= 2;
 491                         (*outbytesleft) -= 1;
 492                         (*inbuf)  += 2;
 493                         (*outbuf) += 1;
 494                         continue;
 495                 }
 496                 if (*outbytesleft < 5) {
 497                         errno = E2BIG;
 498                         return -1;
 499                 }
 500                 snprintf(buf, 6, "@%04x", SVAL(*inbuf, 0));
 501                 memcpy(*outbuf, buf, 5);
 502                 (*inbytesleft)  -= 2;
 503                 (*outbytesleft) -= 5;
 504                 (*inbuf)  += 2;
 505                 (*outbuf) += 5;
 506         }
 507
 508         if (*inbytesleft == 1) {
 509                 errno = EINVAL;
 510                 return -1;
 511         }
 512
 513         if (*inbytesleft > 1) {
 514                 errno = E2BIG;
 515                 return -1;
 516         }
 517
 518         return 0;
 519 }
 520
 521 static size_t iconv_swab(void *cd, const char **inbuf, size_t *inbytesleft,
 522                          char **outbuf, size_t *outbytesleft)
 523 {
 524         int n;
 525
 526         n = MIN(*inbytesleft, *outbytesleft);
 527
 528         swab(*inbuf, *outbuf, (n&~1));
 529         if (n&1) {
 530                 (*outbuf)[n-1] = 0;
 531         }
 532
 533         (*inbytesleft) -= n;
 534         (*outbytesleft) -= n;
 535         (*inbuf) += n;
 536         (*outbuf) += n;
 537
 538         if (*inbytesleft > 0) {
 539                 errno = E2BIG;
 540                 return -1;
 541         }
 542
 543         return 0;
 544 }
 545
 546
 547 static size_t iconv_copy(void *cd, const char **inbuf, size_t *inbytesleft,
 548                          char **outbuf, size_t *outbytesleft)
 549 {
 550         int n;
 551
 552         n = MIN(*inbytesleft, *outbytesleft);
 553
 554         memmove(*outbuf, *inbuf, n);
 555
 556         (*inbytesleft) -= n;
 557         (*outbytesleft) -= n;
 558         (*inbuf) += n;
 559         (*outbuf) += n;
 560
 561         if (*inbytesleft > 0) {
 562                 errno = E2BIG;
 563                 return -1;
 564         }
 565
 566         return 0;
 567 }
 568
 569 /*
 570   this takes a UTF8 sequence and produces a UTF16 sequence
 571  */
 572 static size_t utf8_pull(void *cd, const char **inbuf, size_t *inbytesleft,
 573                          char **outbuf, size_t *outbytesleft)
 574 {
 575         size_t in_left=*inbytesleft, out_left=*outbytesleft;
 576         const uint8_t *c = (const uint8_t *)*inbuf;
 577         uint8_t *uc = (uint8_t *)*outbuf;
 578
 579         while (in_left >= 1 && out_left >= 2) {
 580                 if ((c[0] & 0x80) == 0) {
 581                         uc[0] = c[0];
 582                         uc[1] = 0;
 583                         c  += 1;
 584                         in_left  -= 1;
 585                         out_left -= 2;
 586                         uc += 2;
 587                         continue;
 588                 }
 589
 590                 if ((c[0] & 0xe0) == 0xc0) {
 591                         if (in_left < 2 ||
 592                             (c[1] & 0xc0) != 0x80) {
 593                                 errno = EILSEQ;
 594                                 goto error;
 595                         }
 596                         uc[1] = (c[0]>>2) & 0x7;
 597                         uc[0] = (c[0]<<6) | (c[1]&0x3f);
 598                         c  += 2;
 599                         in_left  -= 2;
 600                         out_left -= 2;
 601                         uc += 2;
 602                         continue;
 603                 }
 604
 605                 if ((c[0] & 0xf0) == 0xe0) {
 606                         if (in_left < 3 ||
 607                             (c[1] & 0xc0) != 0x80 ||
 608                             (c[2] & 0xc0) != 0x80) {
 609                                 errno = EILSEQ;
 610                                 goto error;
 611                         }
 612                         uc[1] = ((c[0]&0xF)<<4) | ((c[1]>>2)&0xF);
 613                         uc[0] = (c[1]<<6) | (c[2]&0x3f);
 614                         c  += 3;
 615                         in_left  -= 3;
 616                         out_left -= 2;
 617                         uc += 2;
 618                         continue;
 619                 }
 620
 621                 if ((c[0] & 0xf8) == 0xf0) {
 622                         unsigned int codepoint;
 623                         if (in_left < 4 ||
 624                             (c[1] & 0xc0) != 0x80 ||
 625                             (c[2] & 0xc0) != 0x80 ||
 626                             (c[3] & 0xc0) != 0x80) {
 627                                 errno = EILSEQ;
 628                                 goto error;
 629                         }
 630                         codepoint =
 631                                 (c[3]&0x3f) |
 632                                 ((c[2]&0x3f)<<6) |
 633                                 ((c[1]&0x3f)<<12) |
 634                                 ((c[0]&0x7)<<18);
 635                         if (codepoint < 0x10000) {
 636                                 /* accept UTF-8 characters that are not
 637                                    minimally packed, but pack the result */
 638                                 uc[0] = (codepoint & 0xFF);
 639                                 uc[1] = (codepoint >> 8);
 640                                 c += 4;
 641                                 in_left -= 4;
 642                                 out_left -= 2;
 643                                 uc += 2;
 644                                 continue;
 645                         }
 646
 647                         codepoint -= 0x10000;
 648
 649                         if (out_left < 4) {
 650                                 errno = E2BIG;
 651                                 goto error;
 652                         }
 653
 654                         uc[0] = (codepoint>>10) & 0xFF;
 655                         uc[1] = (codepoint>>18) | 0xd8;
 656                         uc[2] = codepoint & 0xFF;
 657                         uc[3] = ((codepoint>>8) & 0x3) | 0xdc;
 658                         c  += 4;
 659                         in_left  -= 4;
 660                         out_left -= 4;
 661                         uc += 4;
 662                         continue;
 663                 }
 664
 665                 /* we don't handle 5 byte sequences */
 666                 errno = EINVAL;
 667                 goto error;
 668         }
 669
 670         if (in_left > 0) {
 671                 errno = E2BIG;
 672                 goto error;
 673         }
 674
 675         *inbytesleft = in_left;
 676         *outbytesleft = out_left;
 677         *inbuf = (const char *)c;
 678         *outbuf = (char *)uc;
 679         return 0;
 680
 681 error:
 682         *inbytesleft = in_left;
 683         *outbytesleft = out_left;
 684         *inbuf = (const char *)c;
 685         *outbuf = (char *)uc;
 686         return -1;
 687 }
 688
 689
 690 /*
 691   this takes a UTF16 sequence and produces a UTF8 sequence
 692  */
 693 static size_t utf8_push(void *cd, const char **inbuf, size_t *inbytesleft,
 694                         char **outbuf, size_t *outbytesleft)
 695 {
 696         size_t in_left=*inbytesleft, out_left=*outbytesleft;
 697         uint8_t *c = (uint8_t *)*outbuf;
 698         const uint8_t *uc = (const uint8_t *)*inbuf;
 699
 700         while (in_left >= 2 && out_left >= 1) {
 701                 unsigned int codepoint;
 702
 703                 if (uc[1] == 0 && !(uc[0] & 0x80)) {
 704                         /* simplest case */
 705                         c[0] = uc[0];
 706                         in_left  -= 2;
 707                         out_left -= 1;
 708                         uc += 2;
 709                         c  += 1;
 710                         continue;
 711                 }
 712
 713                 if ((uc[1]&0xf8) == 0) {
 714                         /* next simplest case */
 715                         if (out_left < 2) {
 716                                 errno = E2BIG;
 717                                 goto error;
 718                         }
 719                         c[0] = 0xc0 | (uc[0]>>6) | (uc[1]<<2);
 720                         c[1] = 0x80 | (uc[0] & 0x3f);
 721                         in_left  -= 2;
 722                         out_left -= 2;
 723                         uc += 2;
 724                         c  += 2;
 725                         continue;
 726                 }
 727
 728                 if ((uc[1] & 0xfc) == 0xdc) {
 729                         /* its the second part of a 4 byte sequence. Illegal */
 730                         if (in_left < 4) {
 731                                 errno = EINVAL;
 732                         } else {
 733                                 errno = EILSEQ;
 734                         }
 735                         goto error;
 736                 }
 737
 738                 if ((uc[1] & 0xfc) != 0xd8) {
 739                         codepoint = uc[0] | (uc[1]<<8);
 740                         if (out_left < 3) {
 741                                 errno = E2BIG;
 742                                 goto error;
 743                         }
 744                         c[0] = 0xe0 | (codepoint >> 12);
 745                         c[1] = 0x80 | ((codepoint >> 6) & 0x3f);
 746                         c[2] = 0x80 | (codepoint & 0x3f);
 747
 748                         in_left  -= 2;
 749                         out_left -= 3;
 750                         uc  += 2;
 751                         c   += 3;
 752                         continue;
 753                 }
 754
 755                 /* its the first part of a 4 byte sequence */
 756                 if (in_left < 4) {
 757                         errno = EINVAL;
 758                         goto error;
 759                 }
 760                 if ((uc[3] & 0xfc) != 0xdc) {
 761                         errno = EILSEQ;
 762                         goto error;
 763                 }
 764                 codepoint = 0x10000 + (uc[2] | ((uc[3] & 0x3)<<8) |
 765                                        (uc[0]<<10) | ((uc[1] & 0x3)<<18));
 766
 767                 if (out_left < 4) {
 768                         errno = E2BIG;
 769                         goto error;
 770                 }
 771                 c[0] = 0xf0 | (codepoint >> 18);
 772                 c[1] = 0x80 | ((codepoint >> 12) & 0x3f);
 773                 c[2] = 0x80 | ((codepoint >> 6) & 0x3f);
 774                 c[3] = 0x80 | (codepoint & 0x3f);
 775
 776                 in_left  -= 4;
 777                 out_left -= 4;
 778                 uc       += 4;
 779                 c        += 4;
 780         }
 781
 782         if (in_left == 1) {
 783                 errno = EINVAL;
 784                 goto error;
 785         }
 786
 787         if (in_left > 1) {
 788                 errno = E2BIG;
 789                 goto error;
 790         }
 791
 792         *inbytesleft = in_left;
 793         *outbytesleft = out_left;
 794         *inbuf  = (const char *)uc;
 795         *outbuf = (char *)c;
 796
 797         return 0;
 798
 799 error:
 800         *inbytesleft = in_left;
 801         *outbytesleft = out_left;
 802         *inbuf  = (const char *)uc;
 803         *outbuf = (char *)c;
 804         return -1;
 805 }
 806
 807
 808 /*
 809   this takes a UTF16 munged sequence, modifies it according to the
 810   string2key rules, and produces a UTF16 sequence
 811
 812 The rules are:
 813
 814     1) any 0x0000 characters are mapped to 0x0001
 815
 816     2) convert any instance of 0xD800 - 0xDBFF (high surrogate)
 817        without an immediately following 0xDC00 - 0x0xDFFF (low surrogate) to
 818        U+FFFD (OBJECT REPLACEMENT CHARACTER).
 819
 820     3) the same for any low surrogate that was not preceded by a high surrogate.
 821
 822  */
 823 static size_t utf16_munged_pull(void *cd, const char **inbuf, size_t *inbytesleft,
 824                                char **outbuf, size_t *outbytesleft)
 825 {
 826         size_t in_left=*inbytesleft, out_left=*outbytesleft;
 827         uint8_t *c = (uint8_t *)*outbuf;
 828         const uint8_t *uc = (const uint8_t *)*inbuf;
 829
 830         while (in_left >= 2 && out_left >= 2) {
 831                 unsigned int codepoint = uc[0] | (uc[1]<<8);
 832
 833                 if (codepoint == 0) {
 834                         codepoint = 1;
 835                 }
 836
 837                 if ((codepoint & 0xfc00) == 0xd800) {
 838                         /* a high surrogate */
 839                         unsigned int codepoint2;
 840                         if (in_left < 4) {
 841                                 codepoint = 0xfffd;
 842                                 goto codepoint16;
 843                         }
 844                         codepoint2 = uc[2] | (uc[3]<<8);
 845                         if ((codepoint2 & 0xfc00) != 0xdc00) {
 846                                 /* high surrogate not followed by low
 847                                    surrogate: convert to 0xfffd */
 848                                 codepoint = 0xfffd;
 849                                 goto codepoint16;
 850                         }
 851                         if (out_left < 4) {
 852                                 errno = E2BIG;
 853                                 goto error;
 854                         }
 855                         memcpy(c, uc, 4);
 856                         in_left  -= 4;
 857                         out_left -= 4;
 858                         uc       += 4;
 859                         c        += 4;
 860                         continue;
 861                 }
 862
 863                 if ((codepoint & 0xfc00) == 0xdc00) {
 864                         /* low surrogate not preceded by high
 865                            surrogate: convert to 0xfffd */
 866                         codepoint = 0xfffd;
 867                 }
 868
 869         codepoint16:
 870                 c[0] = codepoint & 0xFF;
 871                 c[1] = (codepoint>>8) & 0xFF;
 872
 873                 in_left  -= 2;
 874                 out_left -= 2;
 875                 uc  += 2;
 876                 c   += 2;
 877                 continue;
 878         }
 879
 880         if (in_left == 1) {
 881                 errno = EINVAL;
 882                 goto error;
 883         }
 884
 885         if (in_left > 1) {
 886                 errno = E2BIG;
 887                 goto error;
 888         }
 889
 890         *inbytesleft = in_left;
 891         *outbytesleft = out_left;
 892         *inbuf  = (const char *)uc;
 893         *outbuf = (char *)c;
 894
 895         return 0;
 896
 897 error:
 898         *inbytesleft = in_left;
 899         *outbytesleft = out_left;
 900         *inbuf  = (const char *)uc;
 901         *outbuf = (char *)c;
 902         return -1;
 903 }
 904
 905
 906