1 /*************************************************
2 * Exim - an Internet mail transport agent *
3 *************************************************/
5 /* Copyright (c) University of Cambridge 1995 - 2018 */
6 /* See the file NOTICE for conditions of use and distribution. */
8 /* This file contains a function for decoding message header lines that may
9 contain encoded "words" according to the rules described in
11 RFC-2047 at http://www.ietf.org/rfc/rfc2047.txt
13 The function is a rewritten version of code created by Norihisa Washitake.
14 The original could be used both inside Exim (as part of a patch) or in a
15 freestanding form. The original contained some built-in code conversions; I
16 have chosen only to do code conversions if iconv() is supported by the OS.
17 Because there were quite a lot of hacks to be done, for a variety of reasons,
20 You can find the latest version of the original library at
22 http://washitake.com/mail/exim/mime/
24 The code below is almost completely unlike the original. */
30 /*************************************************
31 * Do a QP conversion *
32 *************************************************/
34 /* This function decodes "quoted printable" into bytes.
37 string the string that includes QP escapes
38 ptrptr where to return pointer to the decoded string
40 Returns: the length of the decoded string, or -1 on failure
44 rfc2047_qpdecode(uschar *string, uschar **ptrptr)
49 ptr = *ptrptr = store_get(Ustrlen(string) + 1, is_tainted(string)); /* No longer than this */
55 if (ch == '_') *ptr++ = ' ';
59 int b = (a == 0)? 0 : string[1];
60 if (!isxdigit(a) || !isxdigit(b)) return -1; /* Bad QP string */
61 *ptr++ = ((Ustrchr(hex_digits, tolower(a)) - hex_digits) << 4) +
62 Ustrchr(hex_digits, tolower(b)) - hex_digits;
65 else if (ch == ' ' || ch == '\t') return -1; /* Whitespace is illegal */
77 /*************************************************
78 * Decode next MIME word *
79 *************************************************/
81 /* Scan a string to see if a MIME word exists; pass back the separator
86 lencheck TRUE to enforce maximum length check
87 q1ptr pass back address of first question mark
88 q2ptr pass back address of second question mark
89 endptr pass back address of final ?=
90 dlenptr pass back length of decoded string
91 dptrptr pass back pointer to decoded string
93 Returns: address of =? or NULL if not present
97 decode_mimeword(uschar *string, BOOL lencheck, uschar **q1ptr, uschar **q2ptr,
98 uschar **endptr, size_t *dlenptr, uschar **dptrptr)
101 for (;; string = mimeword + 2)
106 if ((mimeword = Ustrstr(string, "=?")) == NULL ||
107 (*q1ptr = Ustrchr(mimeword+2, '?')) == NULL ||
108 (*q2ptr = Ustrchr(*q1ptr+1, '?')) == NULL ||
109 (*endptr = Ustrstr(*q2ptr+1, "?=")) == NULL) return NULL;
111 /* We have found =?xxx?xxx?xxx?= in the string. Optionally check the
112 length, and that the second field is just one character long. If not,
113 continue the loop to search again. We must start just after the initial =?
114 because we might have found =?xxx=?xxx?xxx?xxx?=. */
116 if ((lencheck && *endptr - mimeword > 73) || *q2ptr - *q1ptr != 2) continue;
118 /* Get the encoding letter, and decode the data string. */
120 encoding = toupper((*q1ptr)[1]);
123 dlen = b64decode(*q2ptr+1, dptrptr);
124 else if (encoding == 'Q')
125 dlen = rfc2047_qpdecode(*q2ptr+1, dptrptr);
126 **endptr = '?'; /* restore */
128 /* If the decoding succeeded, we are done. Set the length of the decoded
129 string, and pass back the initial pointer. Otherwise, the loop continues. */
133 *dlenptr = (size_t)dlen;
138 /* Control should never actually get here */
143 /*************************************************
144 * Decode and convert an RFC 2047 string *
145 *************************************************/
147 /* There are two functions defined here. The original one was rfc2047_decode()
148 and it was documented in the local_scan() interface. I needed to add an extra
149 argument for use by expand_string(), so I created rfc2047_decode2() for that
150 purpose. The original function became a stub that just supplies NULL for the
151 new argument (sizeptr).
153 An RFC 2047-encoded string may contain one or more "words", each of the
154 form =?...?.?...?= with the first ... specifying the character code, the
155 second being Q (for quoted printable) or B for Base64 encoding. The third ...
158 This function first decodes each "word" into bytes from the Q or B encoding.
159 Then, if provided with the name of a charset encoding, and if iconv() is
160 available, it attempts to translate the result to the named character set.
161 If this fails, the binary string is returned with an error message.
163 If a binary zero is encountered in the decoded string, it is replaced by the
164 contents of the zeroval argument. For use with Exim headers, the value must not
165 be 0 because they are handled as zero-terminated strings. When zeroval==0,
166 lenptr should not be NULL.
169 string the subject string
170 lencheck TRUE to enforce maximum MIME word length
171 target the name of the target encoding for MIME words, or NULL for
172 no charset translation
173 zeroval the value to use for binary zero bytes
174 lenptr if not NULL, the length of the result is returned via
176 sizeptr if not NULL, the length of a new store block in which the
177 result is built is placed here; if no new store is obtained,
178 the value is not changed
179 error for error messages; NULL if no problem; this can be set
180 when the yield is non-NULL if there was a charset
183 Returns: the decoded, converted string, or NULL on error; if there are
184 no MIME words in the string, the original string is returned
188 rfc2047_decode2(uschar *string, BOOL lencheck, uschar *target, int zeroval,
189 int *lenptr, int *sizeptr, uschar **error)
191 int size = Ustrlen(string);
195 uschar *mimeword, *q1, *q2, *endword;
198 mimeword = decode_mimeword(string, lencheck, &q1, &q2, &endword, &dlen, &dptr);
202 if (lenptr) *lenptr = size;
206 /* Scan through the string, decoding MIME words and copying intermediate text,
207 building the result as we go. The result may be longer than the input if it is
208 translated into a multibyte code such as UTF-8. That's why we use the dynamic
209 string building code. */
211 yield = store_get(sizeof(gstring) + ++size, is_tainted(string));
214 yield->s = US(yield + 1);
220 iconv_t icd = (iconv_t)(-1);
223 if (mimeword != string)
224 yield = string_catn(yield, string, mimeword - string);
225 /*XXX that might have to convert an untainted string to a tainted one */
227 /* Do a charset translation if required. This is supported only on hosts
228 that have the iconv() function. Translation errors set error, but carry on,
229 using the untranslated data. If there is more than one error, the message
230 passed back refers to the final one. We use a loop to cater for the case
231 of long strings - the RFC puts limits on the length, but it's best to be
236 if (target != NULL && strcmpic(target, mimeword+2) != 0)
238 icd = iconv_open(CS target, CS(mimeword+2));
240 if (icd == (iconv_t)(-1))
242 *error = string_sprintf("iconv_open(\"%s\", \"%s\") failed: %s%s",
243 target, mimeword+2, strerror(errno),
244 (errno == EINVAL)? " (maybe unsupported conversion)" : "");
252 uschar *tptr = NULL; /* Stops compiler warning */
257 uschar *outptr = tbuffer;
258 size_t outleft = sizeof(tbuffer);
260 /* If translation is required, go for it. */
262 if (icd != (iconv_t)(-1))
264 (void)iconv(icd, (ICONV_ARG2_TYPE)(&dptr), &dlen, CSS &outptr, &outleft);
266 /* If outptr has been adjusted, there is some output. Set up to add it to
267 the output buffer. The function will have adjusted dptr and dlen. If
268 iconv() stopped because of an error, we'll pick it up next time when
271 If there is no output, we expect there to have been a translation
272 error, because we know there was at least one input byte. We leave the
273 value of tlen as -1, which causes the rest of the input to be copied
276 if (outptr > tbuffer)
279 tlen = outptr - tbuffer;
283 DEBUG(D_any) debug_printf("iconv error translating \"%.*s\" to %s: "
284 "%s\n", (int)(endword + 2 - mimeword), mimeword, target, strerror(errno));
290 /* No charset translation is happening or there was a translation error;
291 just set up the original as the string to be added, and mark it all used.
301 /* Deal with zero values; convert them if requested. */
304 for (int i = 0; i < tlen; i++)
305 if (tptr[i] == 0) tptr[i] = zeroval;
307 /* Add the new string onto the result */
309 yield = string_catn(yield, tptr, tlen);
313 if (icd != (iconv_t)(-1)) iconv_close(icd);
316 /* Update string past the MIME word; skip any white space if the next thing
317 is another MIME word. */
319 string = endword + 2;
320 mimeword = decode_mimeword(string, lencheck, &q1, &q2, &endword, &dlen, &dptr);
324 while (isspace(*s)) s++;
325 if (s == mimeword) string = s;
329 /* Copy the remaining characters of the string, zero-terminate it, and return
330 the length as well if requested. */
332 yield = string_cat(yield, string);
334 if (lenptr) *lenptr = yield->ptr;
335 if (sizeptr) *sizeptr = yield->size;
336 return string_from_gstring(yield);
340 /* This is the stub that provides the original interface without the sizeptr
344 rfc2047_decode(uschar *string, BOOL lencheck, uschar *target, int zeroval,
345 int *lenptr, uschar **error)
347 return rfc2047_decode2(string, lencheck, target, zeroval, lenptr, NULL, error);
350 /* End of rfc2047.c */