mime-param.c

   1 /*@ S-nail - a mail user agent derived from Berkeley Mail.
   2  *@ MIME parameter handling.
   3  *
   4  * Copyright (c) 2016 - 2017 Steffen (Daode) Nurpmeso <steffen@sdaoden.eu>.
   5  *
   6  * Permission to use, copy, modify, and/or distribute this software for any
   7  * purpose with or without fee is hereby granted, provided that the above
   8  * copyright notice and this permission notice appear in all copies.
   9  *
  10  * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
  11  * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
  12  * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
  13  * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
  14  * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
  15  * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
  16  * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
  17  */
  18 #undef n_FILE
  19 #define n_FILE mime_param
  20
  21 #ifndef HAVE_AMALGAMATION
  22 # include "nail.h"
  23 #endif
  24
  25 struct rfc2231_joiner {
  26    struct rfc2231_joiner *rj_next;
  27    ui32_t      rj_no;            /* Continuation number */
  28    ui32_t      rj_len;           /* of useful data in .rj_dat */
  29    ui32_t      rj_val_off;       /* Start of value data therein */
  30    ui32_t      rj_cs_len;        /* Length of charset part */
  31    bool_t      rj_is_enc;        /* Is percent encoded */
  32    ui8_t       __pad[7];
  33    char const  *rj_dat;
  34 };
  35
  36 struct mime_param_builder {
  37    struct mime_param_builder *mpb_next;
  38    struct str  *mpb_result;
  39    ui32_t      mpb_level;        /* of recursion (<-> continuation number) */
  40    ui32_t      mpb_name_len;     /* of the parameter .mpb_name */
  41    ui32_t      mpb_value_len;    /* of remaining value */
  42    ui32_t      mpb_charset_len;  /* of .mpb_charset (only in outermost level) */
  43    ui32_t      mpb_buf_len;      /* Usable result of this level in .mpb_buf */
  44    bool_t      mpb_is_enc;       /* Level requires encoding */
  45    ui8_t       __dummy[1];
  46    bool_t      mpb_is_utf8;      /* Encoding is UTF-8 */
  47    si8_t       mpb_rv;
  48    char const  *mpb_name;
  49    char const  *mpb_value;       /* Remains of, once the level was entered */
  50    char const  *mpb_charset;     /* *ttycharset* */
  51    char        *mpb_buf;         /* Pointer to on-stack buffer */
  52 };
  53
  54 /* All ASCII characters which cause RFC 2231 to be applied XXX check -1 slots*/
  55 static bool_t const        _rfc2231_etab[] = {
  56     1, 1, 1, 1,  1, 1, 1, 1,  1, 1,-1,-1,  1,-1, 1, 1,   /* NUL..SI */
  57     1, 1, 1, 1,  1, 1, 1, 1,  1, 1, 1, 1,  1, 1, 1, 1,   /* DLE..US */
  58     1, 0, 1, 0,  0, 1, 0, 1,  1, 1, 1, 0,  1, 0, 0, 1,   /* CAN.. / */
  59     0, 0, 0, 0,  0, 0, 0, 0,  0, 0, 1, 1,  1, 1, 1, 1,   /*   0.. ? */
  60
  61     1, 0, 0, 0,  0, 0, 0, 0,  0, 0, 0, 0,  0, 0, 0, 0,   /*   @.. O */
  62     0, 0, 0, 0,  0, 0, 0, 0,  0, 0, 0, 1,  1, 1, 0, 0,   /*   P.. _ */
  63     0, 0, 0, 0,  0, 0, 0, 0,  0, 0, 0, 0,  0, 0, 0, 0,   /*   `.. o */
  64     0, 0, 0, 0,  0, 0, 0, 0,  0, 0, 0, 0,  0, 0, 0, 1,   /*   p..DEL */
  65 };
  66
  67 /* In a headerbody, at a "param=XY" that we're not interested in, skip over the
  68  * entire construct, return pointer to the first byte thereafter or to NUL */
  69 static char const * _mime_param_skip(char const *hbp);
  70
  71 /* Trim value, which points to after the "name[RFC 2231 stuff]=".
  72  * On successful return (1,-1; -1 is returned if the value was quoted via
  73  * double quotation marks) a set end_or_null points to after the value and any
  74  * possible separator and result->s is the salloc()d normalized value */
  75 static si8_t      _mime_param_value_trim(struct str *result, char const *start,
  76                      char const **end_or_null);
  77
  78 /* mime_param_get() found the desired parameter but it seems to use RFC 2231
  79  * extended syntax: perform full RFC 2231 parsing starting at this point.
  80  * Note that _join() returns is-error */
  81 static char *     _rfc2231_param_parse(char const *param, size_t plen,
  82                      char const *hbp);
  83 static bool_t     __rfc2231_join(struct rfc2231_joiner *head, char **result,
  84                      char const **emsg);
  85
  86 /* Recursive parameter builder.  Note we have a magic limit of 999 levels.
  87  * Prepares a portion of output in self->mpb_buf;
  88  * once >mpb_value is worked completely the deepmost level joins the result
  89  * into >mpb_result and unrolls the stack. */
  90 static void       _mime_param_create(struct mime_param_builder *self);
  91 static void       __mime_param_join(struct mime_param_builder *head);
  92
  93 static char const *
  94 _mime_param_skip(char const *hbp)
  95 {
  96    char co, cn;
  97    NYD2_ENTER;
  98
  99    /* Skip over parameter name - note we may have skipped over an entire
 100     * parameter name and thus point to a "="; i haven't yet truly checked
 101     * against MIME RFCs, just test for ";" in the meanwhile XXX */
 102    while ((cn = *hbp) != '\0' && cn != '=' && cn != ';')
 103       ++hbp;
 104    if (cn == '\0')
 105       goto jleave;
 106    ++hbp;
 107    if (cn == ';')
 108       goto jleave;
 109
 110    while (whitechar((cn = *hbp))) /* XXX */
 111       ++hbp;
 112    if (cn == '\0')
 113       goto jleave;
 114
 115    if (cn == '"') {
 116       co = '\0';
 117       while ((cn = *++hbp) != '\0' && (cn != '"' || co == '\\'))
 118          co = (co == '\\') ? '\0' : cn;
 119       if (cn != '\0' && (cn = *++hbp) == ';')
 120          ++hbp;
 121    } else {
 122       for (;; cn = *++hbp)
 123          if (cn == '\0' || cn == ';' || whitechar(cn))
 124             break;
 125       if (cn != '\0')
 126          ++hbp;
 127    }
 128 jleave:
 129    NYD2_LEAVE;
 130    return hbp;
 131 }
 132
 133 static si8_t
 134 _mime_param_value_trim(struct str *result, char const *start,
 135    char const **end_or_null)
 136 {
 137    char const *e;
 138    char co, cn;
 139    size_t i;
 140    si8_t rv;
 141    NYD2_ENTER;
 142
 143    while (whitechar(*start)) /* XXX? */
 144       ++start;
 145
 146    if (*start == '"') {
 147       for (co = '\0', e = ++start;; ++e)
 148          if ((cn = *e) == '\0')
 149             goto jerr;
 150          else if (cn == '"' && co != '\\')
 151             break;
 152          else if (cn == '\\' && co == '\\')
 153             co = '\0';
 154          else
 155             co = cn;
 156       i = PTR2SIZE(e++ - start);
 157       rv = -TRU1;
 158    } else {
 159       for (e = start; (cn = *e) != '\0' && !whitechar(cn) && cn != ';'; ++e)
 160          ;
 161       i = PTR2SIZE(e - start);
 162       rv = TRU1;
 163    }
 164
 165    result->s = salloc(i +1);
 166    if (rv > 0) {
 167       memcpy(result->s, start, result->l = i);
 168       result->s[i] = '\0';
 169    } else {
 170       size_t j;
 171       char *cp;
 172
 173       for (j = 0, cp = result->s, co = '\0'; i-- > 0; co = cn) {
 174          cn = *start++;
 175          if (cn != '\\' || co == '\\') {
 176             cp[j++] = cn;
 177             if (cn == '\\')
 178                cn = '\0';
 179          }
 180       }
 181       cp[j] = '\0';
 182
 183       result->s = cp;
 184       result->l = j;
 185    }
 186
 187    if (end_or_null != NULL) {
 188       while (*e != '\0' && *e == ';')
 189          ++e;
 190       *end_or_null = e;
 191    }
 192 jleave:
 193    NYD2_LEAVE;
 194    return rv;
 195 jerr:
 196    rv = FAL0;
 197    goto jleave;
 198 }
 199
 200 static char *
 201 _rfc2231_param_parse(char const *param, size_t plen, char const *hbp)
 202 {
 203    /* TODO Do it for real and unite with mime_param_get() */
 204    struct str xval;
 205    char nobuf[32], *eptr, *rv = NULL, c;
 206    char const *hbp_base, *cp, *emsg = NULL;
 207    struct rfc2231_joiner *head = NULL, *np;
 208    bool_t errors = FAL0;
 209    size_t i;
 210    NYD2_ENTER;
 211
 212    /* We were called by mime_param_get() after a param name match that
 213     * involved "*", so jump to the matching code */
 214    hbp_base = hbp;
 215    goto jumpin;
 216
 217    for (; *hbp != '\0'; hbp_base = hbp) {
 218       while (whitechar(*hbp))
 219          ++hbp;
 220
 221       if (!ascncasecmp(hbp, param, plen)) {
 222          hbp += plen;
 223          while (whitechar(*hbp))
 224             ++hbp;
 225          if (*hbp++ != '*')
 226                goto jerr;
 227
 228          /* RFC 2231 extensions: "NAME[*DIGITS][*]=", where "*DIGITS" indicates
 229           * parameter continuation and the lone asterisk "*" percent encoded
 230           * values -- if encoding is used the "*0" or lone parameter value
 231           * MUST be encoded and start with a "CHARSET'LANGUAGE'" construct,
 232           * where both of CHARSET and LANGUAGE are optional (we do effectively
 233           * generate error if CHARSET is missing though).
 234           * Continuations may not use that "C'L'" construct, but be tolerant
 235           * and ignore those.  Also encoded and non-encoded continuations may
 236           * occur, i.e., perform percent en-/decoding only as necessary.
 237           * Continuations may occur in any order */
 238          /* xxx RFC 2231 parsing ignores language tags */
 239 jumpin:
 240          for (cp = hbp; digitchar(*cp); ++cp)
 241             ;
 242          i = PTR2SIZE(cp - hbp);
 243          if (i != 0) {
 244             if (i >= sizeof(nobuf)) {
 245                emsg = N_("too many digits to form a valid number");
 246                goto jerr;
 247             } else if ((c = *cp) != '=' && c != '*') {
 248                emsg = N_("expected = or * after leading digits");
 249                goto jerr;
 250             }
 251             memcpy(nobuf, hbp, i);
 252             nobuf[i] = '\0';
 253             if((n_idec_uiz_cp(&i, nobuf, 10, NULL
 254                      ) & (n_IDEC_STATE_EMASK | n_IDEC_STATE_CONSUMED)
 255                   ) != n_IDEC_STATE_CONSUMED || i >= 999){
 256                emsg = N_("invalid continuation sequence number");
 257                goto jerr;
 258             }
 259             hbp = ++cp;
 260
 261             /* Value encoded? */
 262             if (c == '*') {
 263                if (*hbp++ != '=')
 264                   goto jeeqaaster;
 265             } else if (c != '=') {
 266 jeeqaaster:
 267                emsg = N_("expected = after asterisk *");
 268                goto jerr;
 269             }
 270          } else {
 271             /* In continuation mode that is an error, however */
 272             if (head != NULL) {
 273                emsg = N_("missing continuation sequence number");
 274                goto jerr;
 275             }
 276             /* Parameter value is encoded, may define encoding */
 277             c = '*';
 278             if (*cp != '=')
 279                goto jeeqaaster;
 280             hbp = ++cp;
 281             i = 0;
 282          }
 283
 284          /* Create new node and insert it sorted; should be faster than
 285           * creating an unsorted list and sorting it after parsing */
 286          np = smalloc(sizeof *np);
 287          np->rj_next = NULL;
 288          np->rj_no = (ui32_t)i;
 289          np->rj_is_enc = (c == '*');
 290          np->rj_val_off = np->rj_cs_len = 0;
 291
 292          if (head == NULL)
 293             head = np;
 294          else if (i < head->rj_no) {
 295             np->rj_next = head;
 296             head = np;
 297          } else {
 298             struct rfc2231_joiner *l = NULL, *x = head;
 299
 300             while (x != NULL && i > x->rj_no)
 301                l = x, x = x->rj_next;
 302             if (x != NULL)
 303                np->rj_next = x;
 304             assert(l != NULL);
 305             l->rj_next = np;
 306          }
 307
 308          switch (_mime_param_value_trim(&xval, hbp, &cp)) {
 309          default:
 310             emsg = (c == '*') ? N_("invalid value encoding")/* XXX fake */
 311                   : N_("faulty value - missing closing quotation mark \"?");
 312             goto jerr;
 313          case -1:
 314             /* XXX if (np->is_enc && memchr(np->dat, '\'', i) != NULL) {
 315              * XXX    emsg = N_("character set info not allowed here");
 316              * XXX    goto jerr;
 317              * XXX } */np->rj_is_enc = FAL0; /* Silently ignore */
 318             /* FALLTHRU */
 319          case 1:
 320             if (xval.l >= UI32_MAX) {
 321                emsg = N_("parameter value too long");
 322                goto jerr;
 323             }
 324             np->rj_len = (ui32_t)xval.l;
 325             np->rj_dat = xval.s;
 326             break;
 327          }
 328
 329          /* Watch out for character set and language info */
 330          if (np->rj_is_enc && (eptr = memchr(xval.s, '\'', xval.l)) != NULL) {
 331             np->rj_cs_len = PTR2SIZE(eptr - xval.s);
 332             if ((eptr = memchr(eptr + 1, '\'', xval.l - np->rj_cs_len - 1))
 333                   == NULL) {
 334                emsg = N_("faulty RFC 2231 parameter extension");
 335                goto jerr;
 336             }
 337             np->rj_val_off = PTR2SIZE(++eptr - xval.s);
 338          }
 339
 340          hbp = cp;
 341       } else
 342          hbp = _mime_param_skip(hbp);
 343    }
 344    assert(head != NULL); /* (always true due to jumpin:, but..) */
 345
 346    errors |= __rfc2231_join(head, &rv, &emsg);
 347    if (errors && (n_poption & n_PO_D_V_VV)) {
 348       /* TODO should set global flags so that at the end of an operation
 349        * TODO (for a message) a summary can be printed: faulty MIME, xy */
 350       if (emsg == NULL)
 351          emsg = N_("multiple causes");
 352       n_err(_("Message had MIME errors: %s\n"), V_(emsg));
 353    }
 354 jleave:
 355    NYD2_LEAVE;
 356    return rv;
 357
 358 jerr:
 359    while ((np = head) != NULL) {
 360       head = np->rj_next;
 361       free(np);
 362    }
 363    if (n_poption & n_PO_D_V) {
 364       if (emsg == NULL)
 365          emsg = N_("expected asterisk *");
 366       n_err(_("Faulty RFC 2231 MIME parameter value: %s: %s\n"
 367          "Near: %s\n"), param, V_(emsg), hbp_base);
 368    }
 369    rv = NULL;
 370    goto jleave;
 371 }
 372
 373 static bool_t
 374 __rfc2231_join(struct rfc2231_joiner *head, char **result, char const **emsg)
 375 {
 376    struct str sin, sou;
 377    struct rfc2231_joiner *np;
 378    char const *cp;
 379    size_t i;
 380    enum {
 381       _NONE       = 0,
 382       _HAVE_ENC   = 1<<0,
 383       _HAVE_ICONV = 1<<1,
 384       _SEEN_ANY   = 1<<2,
 385       _ERRORS     = 1<<3
 386    } f = _NONE;
 387    ui32_t no;
 388 #ifdef HAVE_ICONV
 389    iconv_t fhicd;
 390 #endif
 391    NYD2_ENTER;
 392
 393 #ifdef HAVE_ICONV
 394    n_UNINIT(fhicd, (iconv_t)-1);
 395
 396    if (head->rj_is_enc) {
 397       char const *tcs;
 398
 399       f |= _HAVE_ENC;
 400       if (head->rj_cs_len == 0) {
 401          /* It is an error if the character set is not set, the language alone
 402           * cannot convert characters, let aside that we don't use it at all */
 403          *emsg = N_("MIME RFC 2231 invalidity: missing character set\n");
 404          f |= _ERRORS;
 405       } else if (ascncasecmp(tcs = ok_vlook(ttycharset),
 406             head->rj_dat, head->rj_cs_len)) {
 407          char *cs = ac_alloc(head->rj_cs_len +1);
 408
 409          memcpy(cs, head->rj_dat, head->rj_cs_len);
 410          cs[head->rj_cs_len] = '\0';
 411          if ((fhicd = n_iconv_open(tcs, cs)) != (iconv_t)-1)
 412             f |= _HAVE_ICONV;
 413          else {
 414             *emsg = N_("necessary character set conversion missing");
 415             f |= _ERRORS;
 416          }
 417          ac_free(cs);
 418       }
 419    }
 420 #endif
 421
 422    if (head->rj_no != 0) {
 423       if (!(f & _ERRORS))
 424          *emsg = N_("First RFC 2231 parameter value chunk number is not 0");
 425       f |= _ERRORS;
 426    }
 427
 428    for (sou.s = NULL, sou.l = 0, no = 0; (np = head) != NULL; free(np)) {
 429       head = np->rj_next;
 430
 431       if (np->rj_no != no++) {
 432          if (!(f & _ERRORS))
 433             *emsg = N_("RFC 2231 parameter value chunks are not contiguous");
 434          f |= _ERRORS;
 435       }
 436
 437       /* RFC 2231 allows such info only in the first continuation, and
 438        * furthermore MUSTs the first to be encoded, then */
 439       if (/*np->rj_is_enc &&*/ np->rj_val_off > 0 &&
 440             (f & (_HAVE_ENC | _SEEN_ANY)) != _HAVE_ENC) {
 441          if (!(f & _ERRORS))
 442             *emsg = N_("invalid redundant RFC 2231 charset/language ignored");
 443          f |= _ERRORS;
 444       }
 445       f |= _SEEN_ANY;
 446
 447       i = np->rj_len - np->rj_val_off;
 448       if (!np->rj_is_enc)
 449          n_str_add_buf(&sou, np->rj_dat + np->rj_val_off, i);
 450       else {
 451          /* Perform percent decoding */
 452          sin.s = smalloc(i +1);
 453          sin.l = 0;
 454
 455          for (cp = np->rj_dat + np->rj_val_off; i > 0;) {
 456             char c;
 457
 458             if ((c = *cp++) == '%') {
 459                si32_t cc;
 460
 461                if (i < 3 || (cc = n_c_from_hex_base16(cp)) < 0) {
 462                   if (!(f & _ERRORS))
 463                      *emsg = N_("invalid RFC 2231 percent encoded sequence");
 464                   f |= _ERRORS;
 465                   goto jhex_putc;
 466                }
 467                sin.s[sin.l++] = (char)cc;
 468                cp += 2;
 469                i -= 3;
 470             } else {
 471 jhex_putc:
 472                sin.s[sin.l++] = c;
 473                --i;
 474             }
 475          }
 476          sin.s[sin.l] = '\0';
 477
 478          n_str_add_buf(&sou, sin.s, sin.l);
 479          free(sin.s);
 480       }
 481    }
 482
 483    /* And add character set conversion on top as necessary.
 484     * RFC 2231 is pragmatic: encode only mentions percent encoding and the
 485     * character set for the entire string ("[no] facility for using more
 486     * than one character set or language"), therefore "continuations may
 487     * contain a mixture of encoded and unencoded segments" applies to
 488     * a contiguous string of a single character set that has been torn in
 489     * pieces due to space restrictions, and it happened that some pieces
 490     * didn't need to be percent encoded.
 491     *
 492     * _In particular_ it therefore doesn't repeat the RFC 2047 paradigm
 493     * that encoded-words-are-atomic, meaning that a single character-set
 494     * conversion run over the final, joined, partially percent-decoded value
 495     * should be sufficient */
 496 #ifdef HAVE_ICONV
 497    if (f & _HAVE_ICONV) {
 498       sin.s = NULL;
 499       sin.l = 0;
 500       if (n_iconv_str(fhicd, n_ICONV_UNIDEFAULT, &sin, &sou, NULL) != 0) {
 501          if (!(f & _ERRORS)) /* XXX won't be reported with _UNIDFEFAULT */
 502             *emsg = N_("character set conversion failed on value");
 503          f |= _ERRORS;
 504       }
 505       free(sou.s);
 506       sou = sin;
 507
 508       n_iconv_close(fhicd);
 509    }
 510 #endif
 511
 512    memcpy(*result = salloc(sou.l +1), sou.s, sou.l +1);
 513    free(sou.s);
 514    NYD2_LEAVE;
 515    return ((f & _ERRORS) != 0);
 516 }
 517
 518 static void
 519 _mime_param_create(struct mime_param_builder *self)
 520 {
 521    struct mime_param_builder next;
 522    /* Don't use MIME_LINELEN_(MAX|LIMIT) stack buffer sizes: normally we won't
 523     * exceed plain MIME_LINELEN, so that this would be a factor 10 wastage.
 524     * On the other hand we may excess _LINELEN to avoid breaking up possible
 525     * multibyte sequences until sizeof(buf) is reached, but since we (a) don't
 526     * support stateful encodings and (b) will try to synchronize on UTF-8 this
 527     * problem is scarce, possibly even artificial */
 528    char buf[n_MIN(MIME_LINELEN_MAX >> 1, MIME_LINELEN * 2)],
 529       *bp, *bp_max, *bp_xmax, *bp_lanoenc;
 530    char const *vb, *vb_lanoenc;
 531    size_t vl;
 532    enum {
 533       _NONE    = 0,
 534       _ISENC   = 1<<0,
 535       _HADRAW  = 1<<1,
 536       _RAW     = 1<<2
 537    } f = _NONE;
 538    NYD2_ENTER;
 539    n_LCTA(sizeof(buf) >= MIME_LINELEN * 2, "Buffer to small for operation");
 540
 541 jneed_enc:
 542    self->mpb_buf = bp = bp_lanoenc = buf;
 543    self->mpb_buf_len = 0;
 544    self->mpb_is_enc = ((f & _ISENC) != 0);
 545    vb_lanoenc = vb = self->mpb_value;
 546    vl = self->mpb_value_len;
 547
 548    /* Configure bp_max to fit in SHOULD, bp_xmax to extent */
 549    bp_max = (buf + MIME_LINELEN) -
 550          (1 + self->mpb_name_len + sizeof("*999*='';") -1 + 2);
 551    bp_xmax = (buf + sizeof(buf)) -
 552          (1 + self->mpb_name_len + sizeof("*999*='';") -1 + 2);
 553    if ((f & _ISENC) && self->mpb_level == 0) {
 554       bp_max -= self->mpb_charset_len;
 555       bp_xmax -= self->mpb_charset_len;
 556    }
 557    if (PTRCMP(bp_max, <=, buf + sizeof("Hunky Dory"))) {
 558       DBG( n_alert("_mime_param_create(): Hunky Dory!"); )
 559       bp_max = buf + (MIME_LINELEN >> 1); /* And then it is SHOULD, anyway */
 560    }
 561    assert(PTRCMP(bp_max + (4 * 3), <=, bp_xmax)); /* UTF-8 extra pad, below */
 562
 563    f &= _ISENC;
 564    while (vl > 0) {
 565       union {char c; ui8_t uc;} u; u.c = *vb;
 566
 567       f |= _RAW;
 568       if (!(f & _ISENC)) {
 569          if (u.uc > 0x7F || cntrlchar(u.c)) { /* XXX reject cntrlchar? */
 570             /* We need to percent encode this character, possibly changing
 571              * overall strategy, but anyway the one of this level, possibly
 572              * rendering invalid any output byte we yet have produced here.
 573              * Instead of throwing away that work just recurse if some fancy
 574              * magic condition is true */
 575              /* *However*, many tested MUAs fail to deal with parameters that
 576               * are splitted across "too many" fields, including ones that
 577               * misread RFC 2231 to allow only one digit, i.e., a maximum of
 578               * ten.  This is plain wrong, but that won't help their users */
 579             if (PTR2SIZE(bp - buf) > /*10 (strawberry) COMPAT*/MIME_LINELEN>>1)
 580                goto jrecurse;
 581             f |= _ISENC;
 582             goto jneed_enc;
 583          }
 584
 585          if (u.uc == '"' || u.uc == '\\') {
 586             f ^= _RAW;
 587             bp[0] = '\\';
 588             bp[1] = u.c;
 589             bp += 2;
 590          }
 591       } else if (u.uc > 0x7F || _rfc2231_etab[u.uc]) {
 592          f ^= _RAW;
 593          bp[0] = '%';
 594          n_c_to_hex_base16(bp + 1, u.c);
 595          bp += 3;
 596       }
 597
 598       ++vb;
 599       --vl;
 600       if (f & _RAW) {
 601          f |= _HADRAW;
 602          vb_lanoenc = vb;
 603          *bp++ = u.c;
 604          bp_lanoenc = bp;
 605       }
 606
 607       /* If all available space has been consumed we must split.
 608        * Due to compatibility reasons we must take care not to break up
 609        * multibyte sequences -- even though RFC 2231 rather implies that the
 610        * splitted value should be joined (after percent encoded fields have
 611        * been percent decoded) and the resulting string be treated in the
 612        * specified character set / language, MUAs have been seen which apply
 613        * the RFC 2047 encoded-words-are-atomic even to RFC 2231 values, even
 614        * if stateful encodings cannot truly be supported like that?!..
 615        *
 616        * So split at 7-bit character if we have seen any and the wastage isn't
 617        * too large; recall that we need to keep the overall number of P=V
 618        * values as low as possible due to compatibility reasons.
 619        * If we haven't seen any plain bytes be laxe and realize that bp_max
 620        * reflects SHOULD lines, and try to extend this as long as possible.
 621        * However, with UTF-8, try to backward synchronize on sequence start */
 622       if (bp <= bp_max)
 623          continue;
 624
 625       if ((f & _HADRAW) && (PTRCMP(bp - bp_lanoenc, <=, bp_lanoenc - buf) ||
 626             (!self->mpb_is_utf8 &&
 627              PTR2SIZE(bp_lanoenc - buf) >= (MIME_LINELEN >> 2)))) {
 628          bp = bp_lanoenc;
 629          vl += PTR2SIZE(vb - vb_lanoenc);
 630          vb = vb_lanoenc;
 631          goto jrecurse;
 632       }
 633
 634       if (self->mpb_is_utf8 && ((ui8_t)(vb[-1]) & 0xC0) != 0x80) {
 635          bp -= 3;
 636          --vb;
 637          ++vl;
 638          goto jrecurse;
 639       }
 640
 641       if (bp <= bp_xmax)
 642          continue;
 643       /* (Shit.) */
 644       goto jrecurse;
 645    }
 646
 647    /* That level made the great and completed encoding.  Build result */
 648    self->mpb_is_enc = ((f & _ISENC) != 0);
 649    self->mpb_buf_len = PTR2SIZE(bp - buf);
 650    __mime_param_join(self);
 651 jleave:
 652    NYD2_LEAVE;
 653    return;
 654
 655    /* Need to recurse, take care not to excess magical limit of 999 levels */
 656 jrecurse:
 657    if (self->mpb_level == 999) {
 658       if (n_poption & n_PO_D_V_VV)
 659          n_err(_("Message RFC 2231 parameters nested too deeply!\n"));
 660       goto jleave;
 661    }
 662
 663    self->mpb_is_enc = ((f & _ISENC) != 0);
 664    self->mpb_buf_len = PTR2SIZE(bp - buf);
 665
 666    memset(&next, 0, sizeof next);
 667    next.mpb_next = self;
 668    next.mpb_level = self->mpb_level + 1;
 669    next.mpb_name_len = self->mpb_name_len;
 670    next.mpb_value_len = vl;
 671    next.mpb_is_utf8 = self->mpb_is_utf8;
 672    next.mpb_name = self->mpb_name;
 673    next.mpb_value = vb;
 674    _mime_param_create(&next);
 675    goto jleave;
 676 }
 677
 678 static void
 679 __mime_param_join(struct mime_param_builder *head)
 680 {
 681    char nobuf[16];
 682    struct mime_param_builder *np;
 683    size_t i, ll;  DBG( size_t len_max; )
 684    struct str *result;
 685    char *cp;
 686    enum {
 687       _NONE    = 0,
 688       _ISENC   = 1<<0,
 689       _ISQUOTE = 1<<1,
 690       _ISCONT  = 1<<2
 691    } f = _NONE;
 692    NYD2_ENTER;
 693
 694    /* Traverse the stack upwards to find out result length (worst case).
 695     * Reverse the list while doing so */
 696    for (i = 0, np = head, head = NULL; np != NULL;) {
 697       struct mime_param_builder *tmp;
 698
 699       i += np->mpb_buf_len + np->mpb_name_len + sizeof(" *999*=\"\";\n") -1;
 700       if (np->mpb_is_enc)
 701          f |= _ISENC;
 702
 703       tmp = np->mpb_next;
 704       np->mpb_next = head;
 705       head = np;
 706       np = tmp;
 707    }
 708    if (f & _ISENC)
 709       i += head->mpb_charset_len; /* sizeof("''") -1 covered by \"\" above */
 710    DBG( len_max = i; )
 711    head->mpb_rv = TRU1;
 712
 713    result = head->mpb_result;
 714    if (head->mpb_next != NULL)
 715       f |= _ISCONT;
 716    cp = result->s = salloc(i +1);
 717
 718    for (ll = 0, np = head;;) {
 719       /* Name part */
 720       memcpy(cp, np->mpb_name, i = np->mpb_name_len);
 721       cp += i;
 722       ll += i;
 723
 724       if (f & _ISCONT) {
 725          char *cpo = cp, *nop = nobuf + sizeof(nobuf);
 726          ui32_t noi = np->mpb_level;
 727
 728          *--nop = '\0';
 729          do
 730             *--nop = "0123456789"[noi % 10];
 731          while ((noi /= 10) != 0);
 732
 733          *cp++ = '*';
 734          while (*nop != '\0')
 735             *cp++ = *nop++;
 736
 737          ll += PTR2SIZE(cp - cpo);
 738       }
 739
 740       if ((f & _ISENC) || np->mpb_is_enc) {
 741          *cp++ = '*';
 742          ++ll;
 743       }
 744       *cp++ = '=';
 745       ++ll;
 746
 747       /* Value part */
 748       if (f & _ISENC) {
 749          f &= ~_ISENC;
 750          memcpy(cp, np->mpb_charset, i = np->mpb_charset_len);
 751          cp += i;
 752          cp[0] = '\'';
 753          cp[1] = '\'';
 754          cp += 2;
 755          ll += i + 2;
 756       } else if (!np->mpb_is_enc) {
 757          f |= _ISQUOTE;
 758          *cp++ = '"';
 759          ++ll;
 760       }
 761
 762       memcpy(cp, np->mpb_buf, i = np->mpb_buf_len);
 763       cp += i;
 764       ll += i;
 765
 766       if (f & _ISQUOTE) {
 767          f ^= _ISQUOTE;
 768          *cp++ = '"';
 769          ++ll;
 770       }
 771
 772       if ((np = np->mpb_next) == NULL)
 773          break;
 774       *cp++ = ';';
 775       ++ll;
 776
 777       i = ll;
 778       i += np->mpb_name_len + np->mpb_buf_len + sizeof(" *999*=\"\";\n") -1;
 779       if (i >= MIME_LINELEN) {
 780          head->mpb_rv = -TRU1;
 781          *cp++ = '\n';
 782          ll = 0;
 783       }
 784
 785       *cp++ = ' ';
 786       ++ll;
 787    }
 788    *cp = '\0';
 789    result->l = PTR2SIZE(cp - result->s);
 790    assert(result->l < len_max);
 791    NYD2_LEAVE;
 792 }
 793
 794 FL char *
 795 mime_param_get(char const *param, char const *headerbody) /* TODO rewr. */
 796 {
 797    struct str xval;
 798    char *rv = NULL;
 799    size_t plen;
 800    char const *p;
 801    NYD_ENTER;
 802
 803    plen = strlen(param);
 804    p = headerbody;
 805
 806    /* At the beginning of headerbody there is no parameter=value pair xxx */
 807    if (!whitechar(*p))
 808       goto jskip1st;
 809
 810    for (;;) {
 811       while (whitechar(*p))
 812          ++p;
 813
 814       if (!ascncasecmp(p, param, plen)) {
 815          p += plen;
 816          while (whitechar(*p)) /* XXX? */
 817             ++p;
 818          switch (*p++) {
 819          case '*':
 820             rv = _rfc2231_param_parse(param, plen, p);
 821             goto jleave;
 822          case '=':
 823             if (!_mime_param_value_trim(&xval, p, NULL)) {
 824                /* XXX LOG? */
 825                goto jleave;
 826             }
 827             rv = xval.s;
 828
 829             /* We do have a result, but some (elder) software (S-nail <v14.8)
 830              * will use RFC 2047 encoded words in  parameter values, too */
 831             /* TODO Automatically check whether the value seems to be RFC 2047
 832              * TODO encwd. -- instead use *rfc2047_parameters* like mutt(1)? */
 833             if ((p = strstr(rv, "=?")) != NULL && strstr(p, "?=") != NULL) {
 834                struct str ti, to;
 835
 836                ti.l = strlen(ti.s = rv);
 837                mime_fromhdr(&ti, &to, TD_ISPR | TD_ICONV | TD_DELCTRL);
 838                rv = savestrbuf(to.s, to.l);
 839                free(to.s);
 840             }
 841             goto jleave;
 842          default:
 843             /* Not our desired parameter, skip and continue */
 844             break;
 845          }
 846       }
 847
 848 jskip1st:
 849       if (*(p = _mime_param_skip(p)) == '\0')
 850          goto jleave;
 851    }
 852
 853 jleave:
 854    NYD_LEAVE;
 855    return rv;
 856 }
 857
 858 FL si8_t
 859 mime_param_create(struct str *result, char const *name, char const *value)
 860 {
 861    /* TODO All this needs rework when we have (1) a real string and even more
 862     * TODO (2) use objects instead of stupid string concat; it's temporary
 863     * TODO I.e., this function should return a HeaderBodyParam */
 864    struct mime_param_builder top;
 865    size_t i;
 866    NYD_ENTER;
 867
 868    memset(result, 0, sizeof *result);
 869
 870    memset(&top, 0, sizeof top);
 871    top.mpb_result = result;
 872    if ((i = strlen(top.mpb_name = name)) > UI32_MAX)
 873       goto jleave;
 874    top.mpb_name_len = (ui32_t)i;
 875    if ((i = strlen(top.mpb_value = value)) > UI32_MAX)
 876       goto jleave;
 877    top.mpb_value_len = (ui32_t)i;
 878    if ((i = strlen(name = ok_vlook(ttycharset))) > UI32_MAX)
 879       goto jleave;
 880    top.mpb_charset = salloc((top.mpb_charset_len = (ui32_t)i) +1);
 881    for (i = 0; *name != '\0'; ++i, ++name)
 882       ((char*)n_UNCONST(top.mpb_charset))[i] = lowerconv(*name);
 883    ((char*)n_UNCONST(top.mpb_charset))[i] = '\0';
 884    if(top.mpb_charset_len >= 4 && !ascncasecmp(top.mpb_charset, "utf", 3) &&
 885          ((top.mpb_charset[3] == '-' && top.mpb_charset[4] == '8' &&
 886           top.mpb_charset_len == 5) || (top.mpb_charset[3] == '8' &&
 887           top.mpb_charset_len == 4)))
 888       top.mpb_is_utf8 = TRU1;
 889    else
 890       top.mpb_is_utf8 = FAL0;
 891
 892    _mime_param_create(&top);
 893 jleave:
 894    NYD_LEAVE;
 895    return top.mpb_rv;
 896 }
 897
 898 FL char *
 899 mime_param_boundary_get(char const *headerbody, size_t *len)
 900 {
 901    char *q = NULL, *p;
 902    NYD_ENTER;
 903
 904    if ((p = mime_param_get("boundary", headerbody)) != NULL) {
 905       size_t sz = strlen(p);
 906
 907       if (len != NULL)
 908          *len = sz + 2;
 909       q = salloc(sz + 2 +1);
 910       q[0] = q[1] = '-';
 911       memcpy(q + 2, p, sz);
 912       *(q + sz + 2) = '\0';
 913    }
 914    NYD_LEAVE;
 915    return q;
 916 }
 917
 918 FL char *
 919 mime_param_boundary_create(void)
 920 {
 921    char *bp;
 922    NYD_ENTER;
 923
 924    bp = salloc(36 + 6 +1);
 925    bp[0] = bp[2] = bp[39] = bp[41] = '=';
 926    bp[1] = bp[40] = '-';
 927    memcpy(bp + 3, n_random_create_cp(36), 36);
 928    bp[42] = '\0';
 929    NYD_LEAVE;
 930    return bp;
 931 }
 932
 933 /* s-it-mode */