libatalk/unicode/utf8.c

   1 /*
   2    Unix SMB/CIFS implementation.
   3    minimal iconv implementation
   4    Copyright (C) Andrew Tridgell 2001
   5    Copyright (C) Jelmer Vernooij 2002,2003
   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 2 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, write to the Free Software
  19    Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
  20
  21    From samba 3.0 beta and GNU libiconv-1.8
  22    It's bad but most of the time we can't use libc iconv service:
  23    - it doesn't round trip for most encoding
  24    - it doesn't know about Apple extension
  25 */
  26
  27 #ifdef HAVE_CONFIG_H
  28 #include "config.h"
  29 #endif /* HAVE_CONFIG_H */
  30 #include <stdlib.h>
  31 #include <errno.h>
  32
  33 #include <netatalk/endian.h>
  34 #include <atalk/unicode.h>
  35 #include <atalk/logger.h>
  36 #include <atalk/unicode.h>
  37 #include "byteorder.h"
  38
  39 /* Given a trailing UTF-8 byte, get the contribution from it to
  40  * the Unicode scalar value for a particular bit shift amount
  41  */
  42 #define GETUCVAL(utf8_trailbyte,shift)  ((unsigned int) (( utf8_trailbyte & 0x3F) << shift))
  43
  44 /* Given a unicode scalar, get a trail UTF-8 byte for a particular bit shift amount */
  45 #define GETUTF8TRAILBYTE(uc,shift)      ((char)( 0x80 | ((uc >> shift) & 0x3F) ) )
  46
  47
  48
  49 static size_t   utf8_pull(void *,char **, size_t *, char **, size_t *);
  50 static size_t   utf8_push(void *,char **, size_t *, char **, size_t *);
  51
  52 struct charset_functions charset_utf8 =
  53 {
  54         "UTF8",
  55         0x08000103,
  56         utf8_pull,
  57         utf8_push,
  58         CHARSET_VOLUME | CHARSET_MULTIBYTE | CHARSET_PRECOMPOSED,
  59         NULL,
  60         NULL, NULL
  61 };
  62
  63 struct charset_functions charset_utf8_mac =
  64 {
  65         "UTF8-MAC",
  66         0x08000103,
  67         utf8_pull,
  68         utf8_push,
  69         CHARSET_VOLUME | CHARSET_CLIENT | CHARSET_MULTIBYTE | CHARSET_DECOMPOSED,
  70         NULL,
  71         NULL, NULL
  72 };
  73
  74 /* ------------------- Convert from UTF-8 to UCS-2 -------------------*/
  75 static size_t utf8_pull(void *cd _U_, char **inbuf, size_t *inbytesleft,
  76                          char **outbuf, size_t *outbytesleft)
  77 {
  78         ucs2_t uc = 0;
  79         ucs2_t hi, low;     /* surrogate pair */
  80         unsigned int codepoint, surrogate;
  81         int len;
  82
  83         while (*inbytesleft >= 1 && *outbytesleft >= 2) {
  84                 unsigned char *c = (unsigned char *)*inbuf;
  85                 len = 1;
  86
  87                 /* Arrange conditionals in the order of most frequent occurrence
  88                  * for users of Latin-based chars */
  89                 if ((c[0] & 0x80) == 0) {
  90                         uc = c[0];
  91                 } else if ((c[0] & 0xe0) == 0xc0) {
  92                         if (*inbytesleft < 2) {
  93                                 LOG(log_debug, logtype_default, "short utf8 char");
  94                                 goto badseq;
  95                         }
  96                         uc = (ucs2_t) (((c[0] & 0x1f) << 6) | GETUCVAL(c[1],0)) ;
  97                         len = 2;
  98                 } else if ((c[0] & 0xf0) == 0xe0) {
  99                         if (*inbytesleft < 3) {
 100                                 LOG(log_debug, logtype_default, "short utf8 char");
 101                                 goto badseq;
 102                         }
 103                         uc = (ucs2_t) (((c[0] & 0x0f) << 12) | GETUCVAL(c[1],6) | GETUCVAL(c[2],0)) ;
 104                         len = 3;
 105                 } else if ((c[0] & 0xf8) == 0xf0) {
 106                         /* 4 bytes, which happens for surrogate pairs only */
 107                         if (*inbytesleft < 4) {
 108                                 LOG(log_debug, logtype_default, "short utf8 char");
 109                                 goto badseq;
 110                         }
 111                         if (*outbytesleft < 4) {
 112                                 LOG(log_debug, logtype_default, "short ucs-2 write");
 113                                 errno = E2BIG;
 114                                 return -1;
 115                         }
 116                         codepoint = ((c[0] & 0x07) << 18) | GETUCVAL(c[1],12) |
 117                                 GETUCVAL(c[2],6) |  GETUCVAL(c[3],0);
 118                         hi = (ucs2_t)( ((codepoint - 0x10000) >> 10) + 0xD800);
 119                         low = (ucs2_t)(0xDC00 + (codepoint & 0x03FF));
 120                         surrogate = (hi << 16) | low;
 121                         SIVAL(*outbuf,0,surrogate);
 122                         len = 4;
 123                         (*inbuf)  += 4;
 124                         (*inbytesleft)  -= 4;
 125                         (*outbytesleft) -= 4;
 126                         (*outbuf) += 4;
 127                         continue;
 128                 }
 129                 else {
 130                         errno = EINVAL;
 131                         return -1;
 132                 }
 133
 134                 SSVAL(*outbuf,0,uc);
 135                 (*inbuf)  += len;
 136                 (*inbytesleft)  -= len;
 137                 (*outbytesleft) -= 2;
 138                 (*outbuf) += 2;
 139         }
 140
 141         if (*inbytesleft > 0) {
 142                 errno = E2BIG;
 143                 return -1;
 144         }
 145
 146         return 0;
 147
 148 badseq:
 149         errno = EINVAL;
 150         return -1;
 151 }
 152
 153 /* --------------------- Convert from UCS-2 to UTF-8 -----------*/
 154 static size_t utf8_push(void *cd _U_, char **inbuf, size_t *inbytesleft,
 155                          char **outbuf, size_t *outbytesleft)
 156 {
 157         ucs2_t uc=0;
 158         ucs2_t hi, low;
 159         unsigned int surrogatepair, codepoint;
 160         int olen, ilen;
 161
 162         while (*inbytesleft >= 2 && *outbytesleft >= 1) {
 163                 unsigned char *c = (unsigned char *)*outbuf;
 164                 uc = SVAL((*inbuf),0);
 165                 olen=1;
 166                 ilen=2;
 167
 168                 /* Arrange conditionals in the order of most frequent occurrence for
 169                    users of Latin-based chars */
 170                 if (uc < 0x80) {
 171                         c[0] = uc;
 172                 } else if (uc < 0x800) {
 173                         if (*outbytesleft < 2) {
 174                                 LOG(log_debug, logtype_default, "short utf8 write");
 175                                 goto toobig;
 176                         }
 177                         c[1] = GETUTF8TRAILBYTE(uc, 0);
 178                         c[0] = (char)(0xc0 | ((uc >> 6) & 0x1f));
 179                         olen = 2;
 180                 }
 181                 else if ( uc >= 0x202a && uc <= 0x202e ) {
 182                         /* ignore bidi hint characters */
 183                         olen = 0;
 184                 }
 185                 /*
 186                  * A 2-byte uc value represents a stand-alone Unicode character if
 187                  *     0 <= uc < 0xd800 or 0xdfff < uc <= 0xffff.
 188                  * If  0xd800 <= uc <= 0xdfff, uc itself does not represent a Unicode character.
 189                  * Rather, it is just part of a surrogate pair.  A surrogate pair consists of
 190                  * a high surrogate in the range [0xd800 ... 0xdbff] and a low surrogate in the
 191                  * range [0xdc00 ... 0xdfff].  Together the pair maps to a single Unicode character
 192                  * whose scalar value is 64K or larger.  It is this scalar value that is transformed
 193                  * to UTF-8, not the individual surrogates.
 194                  *
 195                  * See www.unicode.org/faq/utf_bom.html for more info.
 196                  */
 197
 198                 else if ( 0xd800 <= uc && uc <= 0xdfff) {
 199                         /* surrogate - needs 4 bytes from input and 4 bytes for output to UTF-8 */
 200                         if (*outbytesleft < 4) {
 201                                 LOG(log_debug, logtype_default, "short utf8 write");
 202                                 goto toobig;
 203                         }
 204                         if (*inbytesleft < 4) {
 205                                 errno = EINVAL;
 206                                 return -1;
 207                         }
 208                         surrogatepair = IVAL((*inbuf),0);
 209                         low = (ucs2_t)surrogatepair;
 210                         hi = (ucs2_t)(surrogatepair >> 16);
 211                         if ( 0xd800 <= hi && hi <= 0xdbff && 0xdc00 <= low && low <= 0xdfff) {
 212                                 codepoint = ((hi - 0xd800) << 10) + (low - 0xdc00) + 0x10000;
 213                                 c[3] = GETUTF8TRAILBYTE(codepoint, 0);
 214                                 c[2] = GETUTF8TRAILBYTE(codepoint, 6);
 215                                 c[1] = GETUTF8TRAILBYTE(codepoint, 12);
 216                                 c[0] = (char)(0xf0 | ((codepoint >> 18) & 0x07));
 217                                 ilen = olen = 4;
 218                         } else { /* invalid values for surrogate */
 219                                 errno = EINVAL;
 220                                 return -1;
 221                         }
 222                 } else {
 223                         if (*outbytesleft < 3) {
 224                                 LOG(log_debug, logtype_default, "short utf8 write");
 225                                 goto toobig;
 226                         }
 227                         c[2] = GETUTF8TRAILBYTE(uc, 0);
 228                         c[1] = GETUTF8TRAILBYTE(uc, 6);
 229                         c[0] = (char)(0xe0 | ((uc >> 12) & 0x0f));
 230                         olen = 3;
 231                 }
 232
 233                 (*inbytesleft)  -= ilen;
 234                 (*outbytesleft) -= olen;
 235                 (*inbuf)  += ilen;
 236                 (*outbuf) += olen;
 237         }
 238
 239         if (*inbytesleft == 1) {
 240                 errno = EINVAL;
 241                 return -1;
 242         }
 243
 244         if (*inbytesleft > 1) {
 245                 errno = E2BIG;
 246                 return -1;
 247         }
 248
 249         return 0;
 250
 251 toobig:
 252         errno = E2BIG;
 253         return -1;
 254 }