busybox/mailutils/mime.c

/* vi: set sw=4 ts=4: */
/*
 * makemime: create MIME-encoded message
 * reformime: parse MIME-encoded message
 *
 * Copyright (C) 2008 by Vladimir Dronnikov <dronnikov@gmail.com>
 *
 * Licensed under GPLv2, see file LICENSE in this tarball for details.
 */
#include "libbb.h"
#include "mail.h"

/*
  makemime -c type [-o file] [-e encoding] [-C charset] [-N name] \
                   [-a "Header: Contents"] file
           -m [ type ] [-o file] [-e encoding] [-a "Header: Contents"] file
           -j [-o file] file1 file2
           @file

   file:  filename    - read or write from filename
          -           - read or write from stdin or stdout
          &n          - read or write from file descriptor n
          \( opts \)  - read from child process, that generates [ opts ]

Options:

  -c type         - create a new MIME section from "file" with this
                    Content-Type: (default is application/octet-stream).
  -C charset      - MIME charset of a new text/plain section.
  -N name         - MIME content name of the new mime section.
  -m [ type ]     - create a multipart mime section from "file" of this
                    Content-Type: (default is multipart/mixed).
  -e encoding     - use the given encoding (7bit, 8bit, quoted-printable,
                    or base64), instead of guessing.  Omit "-e" and use
                    -c auto to set Content-Type: to text/plain or
                    application/octet-stream based on picked encoding.
  -j file1 file2  - join mime section file2 to multipart section file1.
  -o file         - write ther result to file, instead of stdout (not
                    allowed in child processes).
  -a header       - prepend an additional header to the output.

  @file - read all of the above options from file, one option or
          value on each line.
*/

int makemime_main(int argc, char **argv) MAIN_EXTERNALLY_VISIBLE;
int makemime_main(int argc UNUSED_PARAM, char **argv)
{
	llist_t *opt_headers = NULL, *l;
	const char *opt_output;
#define boundary opt_output

	enum {
		OPT_c = 1 << 0,         // Content-Type:
		OPT_e = 1 << 1,         // Content-Transfer-Encoding. Ignored. Assumed base64
		OPT_o = 1 << 2,         // output to
		OPT_C = 1 << 3,         // charset
		OPT_N = 1 << 4,         // COMPAT
		OPT_a = 1 << 5,         // additional headers
		OPT_m = 1 << 6,         // COMPAT
		OPT_j = 1 << 7,         // COMPAT
	};

	INIT_G();

	// parse options
	opt_complementary = "a::";
	opts = getopt32(argv,
		"c:e:o:C:N:a:m:j:",
		&G.content_type, NULL, &opt_output, &G.opt_charset, NULL, &opt_headers, NULL, NULL
	);
	//argc -= optind;
	argv += optind;

	// respect -o output
	if (opts & OPT_o)
		freopen(opt_output, "w", stdout);

	// no files given on command line? -> use stdin
	if (!*argv)
		*--argv = (char *)"-";

	// put additional headers
	for (l = opt_headers; l; l = l->link)
		puts(l->data);

	// make a random string -- it will delimit message parts
	srand(monotonic_us());
	boundary = xasprintf("%d-%d-%d", rand(), rand(), rand());

	// put multipart header
	printf(
		"Mime-Version: 1.0\n"
		"Content-Type: multipart/mixed; boundary=\"%s\"\n"
		, boundary
	);

	// put attachments
	while (*argv) {
		printf(
			"\n--%s\n"
			"Content-Type: %s; charset=%s\n"
			"Content-Disposition: inline; filename=\"%s\"\n"
			"Content-Transfer-Encoding: base64\n"
			, boundary
			, G.content_type
			, G.opt_charset
			, bb_get_last_path_component_strip(*argv)
		);
		encode_base64(*argv++, (const char *)stdin, "");
	}

	// put multipart footer
	printf("\n--%s--\n" "\n", boundary);

	return EXIT_SUCCESS;
#undef boundary
}

static const char *find_token(const char *const string_array[], const char *key, const char *defvalue)
{
	const char *r = NULL;
	for (int i = 0; string_array[i] != 0; i++) {
		if (strcasecmp(string_array[i], key) == 0) {
			r = (char *)string_array[i+1];
			break;
		}
	}
	return (r) ? r : defvalue;
}

static const char *xfind_token(const char *const string_array[], const char *key)
{
	const char *r = find_token(string_array, key, NULL);
	if (r)
		return r;
	bb_error_msg_and_die("header: %s", key);
}

enum {
	OPT_x = 1 << 0,
	OPT_X = 1 << 1,
#if ENABLE_FEATURE_REFORMIME_COMPAT
	OPT_d = 1 << 2,
	OPT_e = 1 << 3,
	OPT_i = 1 << 4,
	OPT_s = 1 << 5,
	OPT_r = 1 << 6,
	OPT_c = 1 << 7,
	OPT_m = 1 << 8,
	OPT_h = 1 << 9,
	OPT_o = 1 << 10,
	OPT_O = 1 << 11,
#endif
};

static int parse(const char *boundary, char **argv)
{
	char *line, *s, *p;
	const char *type;
	int boundary_len = strlen(boundary);
	const char *delims = " ;\"\t\r\n";
	const char *uniq;
	int ntokens;
	const char *tokens[32]; // 32 is enough

	// prepare unique string pattern
	uniq = xasprintf("%%llu.%u.%s", (unsigned)getpid(), safe_gethostname());

//bb_info_msg("PARSE[%s]", terminator);

	while ((line = xmalloc_fgets_str(stdin, "\r\n\r\n")) != NULL) {

		// seek to start of MIME section
		// N.B. to avoid false positives let us seek to the _last_ occurance
		p = NULL;
		s = line;
		while ((s=strcasestr(s, "Content-Type:")) != NULL)
			p = s++;
		if (!p)
			goto next;
//bb_info_msg("L[%s]", p);

		// split to tokens
		// TODO: strip of comments which are of form: (comment-text)
		ntokens = 0;
		tokens[ntokens] = NULL;
		for (s = strtok(p, delims); s; s = strtok(NULL, delims)) {
			tokens[ntokens] = s;
			if (ntokens < ARRAY_SIZE(tokens) - 1)
				ntokens++;
//bb_info_msg("L[%d][%s]", ntokens, s);
		}
		tokens[ntokens] = NULL;
//bb_info_msg("N[%d]", ntokens);

		// analyse tokens
		type = find_token(tokens, "Content-Type:", "text/plain");
//bb_info_msg("T[%s]", type);
		if (0 == strncasecmp(type, "multipart/", 10)) {
			if (0 == strcasecmp(type+10, "mixed")) {
				parse(xfind_token(tokens, "boundary="), argv);
			} else
				bb_error_msg_and_die("no support of content type '%s'", type);
		} else {
			pid_t pid = pid;
			int rc;
			FILE *fp;
			// fetch charset
			const char *charset = find_token(tokens, "charset=", CONFIG_FEATURE_MIME_CHARSET);
			// fetch encoding
			const char *encoding = find_token(tokens, "Content-Transfer-Encoding:", "7bit");
			// compose target filename
			char *filename = (char *)find_token(tokens, "filename=", NULL);
			if (!filename)
				filename = xasprintf(uniq, monotonic_us());
			else
				filename = bb_get_last_path_component_strip(xstrdup(filename));

			// start external helper, if any
			if (opts & OPT_X) {
				int fd[2];
				xpipe(fd);
				pid = vfork();
				if (0 == pid) {
					// child reads from fd[0]
					xdup2(fd[0], STDIN_FILENO);
					close(fd[0]); close(fd[1]);
					xsetenv("CONTENT_TYPE", type);
					xsetenv("CHARSET", charset);
					xsetenv("ENCODING", encoding);
					xsetenv("FILENAME", filename);
					BB_EXECVP(*argv, argv);
					_exit(EXIT_FAILURE);
				}
				// parent dumps to fd[1]
				close(fd[0]);
				fp = fdopen(fd[1], "w");
				signal(SIGPIPE, SIG_IGN); // ignore EPIPE
			// or create a file for dump
			} else {
				char *fname = xasprintf("%s%s", *argv, filename);
				fp = xfopen_for_write(fname);
				free(fname);
			}

			// housekeeping
			free(filename);

			// dump to fp
			if (0 == strcasecmp(encoding, "base64")) {
				decode_base64(stdin, fp);
			} else if (0 != strcasecmp(encoding, "7bit")
				&& 0 != strcasecmp(encoding, "8bit")) {
				// quoted-printable, binary, user-defined are unsupported so far
				bb_error_msg_and_die("no support of encoding '%s'", encoding);
			} else {
				// N.B. we have written redundant \n. so truncate the file
				// The following weird 2-tacts reading technique is due to
				// we have to not write extra \n at the end of the file
				// In case of -x option we could truncate the resulting file as
				// fseek(fp, -1, SEEK_END);
				// if (ftruncate(fileno(fp), ftell(fp)))
				//	bb_perror_msg("ftruncate");
				// But in case of -X we have to be much more careful. There is
				// no means to truncate what we already have sent to the helper.
				p = xmalloc_fgets_str(stdin, "\r\n");
				while (p) {
					if ((s = xmalloc_fgets_str(stdin, "\r\n")) == NULL)
						break;
					if ('-' == s[0] && '-' == s[1]
						&& 0 == strncmp(s+2, boundary, boundary_len))
						break;
					fputs(p, fp);
					p = s;
				}

/*
				while ((s = xmalloc_fgetline_str(stdin, "\r\n")) != NULL) {
					if ('-' == s[0] && '-' == s[1]
						&& 0 == strncmp(s+2, boundary, boundary_len))
						break;
					fprintf(fp, "%s\n", s);
				}
				// N.B. we have written redundant \n. so truncate the file
				fseek(fp, -1, SEEK_END);
				if (ftruncate(fileno(fp), ftell(fp)))
					bb_perror_msg("ftruncate");
*/
			}
			fclose(fp);

			// finalize helper
			if (opts & OPT_X) {
				signal(SIGPIPE, SIG_DFL);
				// exit if helper exited >0
				rc = wait4pid(pid);
				if (rc)
					return rc+20;
			}

			// check multipart finalized
			if (s && '-' == s[2+boundary_len] && '-' == s[2+boundary_len+1]) {
				free(line);
				break;
			}
		}
 next:
 		free(line);
	}

//bb_info_msg("ENDPARSE[%s]", boundary);

	return EXIT_SUCCESS;
}

/*
Usage: reformime [options]
    -d - parse a delivery status notification.
    -e - extract contents of MIME section.
    -x - extract MIME section to a file.
    -X - pipe MIME section to a program.
    -i - show MIME info.
    -s n.n.n.n - specify MIME section.
    -r - rewrite message, filling in missing MIME headers.
    -r7 - also convert 8bit/raw encoding to quoted-printable, if possible.
    -r8 - also convert quoted-printable encoding to 8bit, if possible.
    -c charset - default charset for rewriting, -o, and -O.
    -m [file] [file]... - create a MIME message digest.
    -h "header" - decode RFC 2047-encoded header.
    -o "header" - encode unstructured header using RFC 2047.
    -O "header" - encode address list header using RFC 2047.
*/

int reformime_main(int argc, char **argv) MAIN_EXTERNALLY_VISIBLE;
int reformime_main(int argc UNUSED_PARAM, char **argv)
{
	const char *opt_prefix = "";

	INIT_G();

	// parse options
	// N.B. only -x and -X are supported so far
	opt_complementary = "x--X:X--x" USE_FEATURE_REFORMIME_COMPAT(":m::");
	opts = getopt32(argv,
		"x:X" USE_FEATURE_REFORMIME_COMPAT("deis:r:c:m:h:o:O:"),
		&opt_prefix
		USE_FEATURE_REFORMIME_COMPAT(, NULL, NULL, &G.opt_charset, NULL, NULL, NULL, NULL)
	);
	//argc -= optind;
	argv += optind;

	return parse("", (opts & OPT_X) ? argv : (char **)&opt_prefix);
}
1	/* vi: set sw=4 ts=4: */
2	/*
3	* makemime: create MIME-encoded message
4	* reformime: parse MIME-encoded message
5	*
6	* Copyright (C) 2008 by Vladimir Dronnikov <dronnikov@gmail.com>
7	*
8	* Licensed under GPLv2, see file LICENSE in this tarball for details.
9	*/
10	#include "libbb.h"
11	#include "mail.h"
12
13	/*
14	makemime -c type [-o file] [-e encoding] [-C charset] [-N name] \
15	[-a "Header: Contents"] file
16	-m [ type ] [-o file] [-e encoding] [-a "Header: Contents"] file
17	-j [-o file] file1 file2
18	@file
19
20	file: filename - read or write from filename
21	- - read or write from stdin or stdout
22	&n - read or write from file descriptor n
23	\( opts \) - read from child process, that generates [ opts ]
24
25	Options:
26
27	-c type - create a new MIME section from "file" with this
28	Content-Type: (default is application/octet-stream).
29	-C charset - MIME charset of a new text/plain section.
30	-N name - MIME content name of the new mime section.
31	-m [ type ] - create a multipart mime section from "file" of this
32	Content-Type: (default is multipart/mixed).
33	-e encoding - use the given encoding (7bit, 8bit, quoted-printable,
34	or base64), instead of guessing. Omit "-e" and use
35	-c auto to set Content-Type: to text/plain or
36	application/octet-stream based on picked encoding.
37	-j file1 file2 - join mime section file2 to multipart section file1.
38	-o file - write ther result to file, instead of stdout (not
39	allowed in child processes).
40	-a header - prepend an additional header to the output.
41
42	@file - read all of the above options from file, one option or
43	value on each line.
44	*/
45
46	int makemime_main(int argc, char **argv) MAIN_EXTERNALLY_VISIBLE;
47	int makemime_main(int argc UNUSED_PARAM, char **argv)
48	{
49	llist_t opt_headers = NULL, l;
50	const char *opt_output;
51	#define boundary opt_output
52
53	enum {
54	OPT_c = 1 << 0, // Content-Type:
55	OPT_e = 1 << 1, // Content-Transfer-Encoding. Ignored. Assumed base64
56	OPT_o = 1 << 2, // output to
57	OPT_C = 1 << 3, // charset
58	OPT_N = 1 << 4, // COMPAT
59	OPT_a = 1 << 5, // additional headers
60	OPT_m = 1 << 6, // COMPAT
61	OPT_j = 1 << 7, // COMPAT
62	};
63
64	INIT_G();
65
66	// parse options
67	opt_complementary = "a::";
68	opts = getopt32(argv,
69	"c:e:o:C:N:a:m:j:",
70	&G.content_type, NULL, &opt_output, &G.opt_charset, NULL, &opt_headers, NULL, NULL
71	);
72	//argc -= optind;
73	argv += optind;
74
75	// respect -o output
76	if (opts & OPT_o)
77	freopen(opt_output, "w", stdout);
78
79	// no files given on command line? -> use stdin
80	if (!*argv)
81	--argv = (char )"-";
82
83	// put additional headers
84	for (l = opt_headers; l; l = l->link)
85	puts(l->data);
86
87	// make a random string -- it will delimit message parts
88	srand(monotonic_us());
89	boundary = xasprintf("%d-%d-%d", rand(), rand(), rand());
90
91	// put multipart header
92	printf(
93	"Mime-Version: 1.0\n"
94	"Content-Type: multipart/mixed; boundary=\"%s\"\n"
95	, boundary
96	);
97
98	// put attachments
99	while (*argv) {
100	printf(
101	"\n--%s\n"
102	"Content-Type: %s; charset=%s\n"
103	"Content-Disposition: inline; filename=\"%s\"\n"
104	"Content-Transfer-Encoding: base64\n"
105	, boundary
106	, G.content_type
107	, G.opt_charset
108	, bb_get_last_path_component_strip(*argv)
109	);
110	encode_base64(argv++, (const char )stdin, "");
111	}
112
113	// put multipart footer
114	printf("\n--%s--\n" "\n", boundary);
115
116	return EXIT_SUCCESS;
117	#undef boundary
118	}
119
120	static const char find_token(const char const string_array[], const char key, const char defvalue)
121	{
122	const char *r = NULL;
123	for (int i = 0; string_array[i] != 0; i++) {
124	if (strcasecmp(string_array[i], key) == 0) {
125	r = (char *)string_array[i+1];
126	break;
127	}
128	}
129	return (r) ? r : defvalue;
130	}
131
132	static const char xfind_token(const char const string_array[], const char *key)
133	{
134	const char *r = find_token(string_array, key, NULL);
135	if (r)
136	return r;
137	bb_error_msg_and_die("header: %s", key);
138	}
139
140	enum {
141	OPT_x = 1 << 0,
142	OPT_X = 1 << 1,
143	#if ENABLE_FEATURE_REFORMIME_COMPAT
144	OPT_d = 1 << 2,
145	OPT_e = 1 << 3,
146	OPT_i = 1 << 4,
147	OPT_s = 1 << 5,
148	OPT_r = 1 << 6,
149	OPT_c = 1 << 7,
150	OPT_m = 1 << 8,
151	OPT_h = 1 << 9,
152	OPT_o = 1 << 10,
153	OPT_O = 1 << 11,
154	#endif
155	};
156
157	static int parse(const char boundary, char *argv)
158	{
159	char line, s, *p;
160	const char *type;
161	int boundary_len = strlen(boundary);
162	const char *delims = " ;\"\t\r\n";
163	const char *uniq;
164	int ntokens;
165	const char *tokens[32]; // 32 is enough
166
167	// prepare unique string pattern
168	uniq = xasprintf("%%llu.%u.%s", (unsigned)getpid(), safe_gethostname());
169
170	//bb_info_msg("PARSE[%s]", terminator);
171
172	while ((line = xmalloc_fgets_str(stdin, "\r\n\r\n")) != NULL) {
173
174	// seek to start of MIME section
175	// N.B. to avoid false positives let us seek to the _last_ occurance
176	p = NULL;
177	s = line;
178	while ((s=strcasestr(s, "Content-Type:")) != NULL)
179	p = s++;
180	if (!p)
181	goto next;
182	//bb_info_msg("L[%s]", p);
183
184	// split to tokens
185	// TODO: strip of comments which are of form: (comment-text)
186	ntokens = 0;
187	tokens[ntokens] = NULL;
188	for (s = strtok(p, delims); s; s = strtok(NULL, delims)) {
189	tokens[ntokens] = s;
190	if (ntokens < ARRAY_SIZE(tokens) - 1)
191	ntokens++;
192	//bb_info_msg("L[%d][%s]", ntokens, s);
193	}
194	tokens[ntokens] = NULL;
195	//bb_info_msg("N[%d]", ntokens);
196
197	// analyse tokens
198	type = find_token(tokens, "Content-Type:", "text/plain");
199	//bb_info_msg("T[%s]", type);
200	if (0 == strncasecmp(type, "multipart/", 10)) {
201	if (0 == strcasecmp(type+10, "mixed")) {
202	parse(xfind_token(tokens, "boundary="), argv);
203	} else
204	bb_error_msg_and_die("no support of content type '%s'", type);
205	} else {
206	pid_t pid = pid;
207	int rc;
208	FILE *fp;
209	// fetch charset
210	const char *charset = find_token(tokens, "charset=", CONFIG_FEATURE_MIME_CHARSET);
211	// fetch encoding
212	const char *encoding = find_token(tokens, "Content-Transfer-Encoding:", "7bit");
213	// compose target filename
214	char filename = (char )find_token(tokens, "filename=", NULL);
215	if (!filename)
216	filename = xasprintf(uniq, monotonic_us());
217	else
218	filename = bb_get_last_path_component_strip(xstrdup(filename));
219
220	// start external helper, if any
221	if (opts & OPT_X) {
222	int fd[2];
223	xpipe(fd);
224	pid = vfork();
225	if (0 == pid) {
226	// child reads from fd[0]
227	xdup2(fd[0], STDIN_FILENO);
228	close(fd[0]); close(fd[1]);
229	xsetenv("CONTENT_TYPE", type);
230	xsetenv("CHARSET", charset);
231	xsetenv("ENCODING", encoding);
232	xsetenv("FILENAME", filename);
233	BB_EXECVP(*argv, argv);
234	_exit(EXIT_FAILURE);
235	}
236	// parent dumps to fd[1]
237	close(fd[0]);
238	fp = fdopen(fd[1], "w");
239	signal(SIGPIPE, SIG_IGN); // ignore EPIPE
240	// or create a file for dump
241	} else {
242	char fname = xasprintf("%s%s", argv, filename);
243	fp = xfopen_for_write(fname);
244	free(fname);
245	}
246
247	// housekeeping
248	free(filename);
249
250	// dump to fp
251	if (0 == strcasecmp(encoding, "base64")) {
252	decode_base64(stdin, fp);
253	} else if (0 != strcasecmp(encoding, "7bit")
254	&& 0 != strcasecmp(encoding, "8bit")) {
255	// quoted-printable, binary, user-defined are unsupported so far
256	bb_error_msg_and_die("no support of encoding '%s'", encoding);
257	} else {
258	// N.B. we have written redundant \n. so truncate the file
259	// The following weird 2-tacts reading technique is due to
260	// we have to not write extra \n at the end of the file
261	// In case of -x option we could truncate the resulting file as
262	// fseek(fp, -1, SEEK_END);
263	// if (ftruncate(fileno(fp), ftell(fp)))
264	// bb_perror_msg("ftruncate");
265	// But in case of -X we have to be much more careful. There is
266	// no means to truncate what we already have sent to the helper.
267	p = xmalloc_fgets_str(stdin, "\r\n");
268	while (p) {
269	if ((s = xmalloc_fgets_str(stdin, "\r\n")) == NULL)
270	break;
271	if ('-' == s[0] && '-' == s[1]
272	&& 0 == strncmp(s+2, boundary, boundary_len))
273	break;
274	fputs(p, fp);
275	p = s;
276	}
277
278	/*
279	while ((s = xmalloc_fgetline_str(stdin, "\r\n")) != NULL) {
280	if ('-' == s[0] && '-' == s[1]
281	&& 0 == strncmp(s+2, boundary, boundary_len))
282	break;
283	fprintf(fp, "%s\n", s);
284	}
285	// N.B. we have written redundant \n. so truncate the file
286	fseek(fp, -1, SEEK_END);
287	if (ftruncate(fileno(fp), ftell(fp)))
288	bb_perror_msg("ftruncate");
289	*/
290	}
291	fclose(fp);
292
293	// finalize helper
294	if (opts & OPT_X) {
295	signal(SIGPIPE, SIG_DFL);
296	// exit if helper exited >0
297	rc = wait4pid(pid);
298	if (rc)
299	return rc+20;
300	}
301
302	// check multipart finalized
303	if (s && '-' == s[2+boundary_len] && '-' == s[2+boundary_len+1]) {
304	free(line);
305	break;
306	}
307	}
308	next:
309	free(line);
310	}
311
312	//bb_info_msg("ENDPARSE[%s]", boundary);
313
314	return EXIT_SUCCESS;
315	}
316
317	/*
318	Usage: reformime [options]
319	-d - parse a delivery status notification.
320	-e - extract contents of MIME section.
321	-x - extract MIME section to a file.
322	-X - pipe MIME section to a program.
323	-i - show MIME info.
324	-s n.n.n.n - specify MIME section.
325	-r - rewrite message, filling in missing MIME headers.
326	-r7 - also convert 8bit/raw encoding to quoted-printable, if possible.
327	-r8 - also convert quoted-printable encoding to 8bit, if possible.
328	-c charset - default charset for rewriting, -o, and -O.
329	-m [file] [file]... - create a MIME message digest.
330	-h "header" - decode RFC 2047-encoded header.
331	-o "header" - encode unstructured header using RFC 2047.
332	-O "header" - encode address list header using RFC 2047.
333	*/
334
335	int reformime_main(int argc, char **argv) MAIN_EXTERNALLY_VISIBLE;
336	int reformime_main(int argc UNUSED_PARAM, char **argv)
337	{
338	const char *opt_prefix = "";
339
340	INIT_G();
341
342	// parse options
343	// N.B. only -x and -X are supported so far
344	opt_complementary = "x--X:X--x" USE_FEATURE_REFORMIME_COMPAT(":m::");
345	opts = getopt32(argv,
346	"x:X" USE_FEATURE_REFORMIME_COMPAT("deis:r:c:m:h:o:O:"),
347	&opt_prefix
348	USE_FEATURE_REFORMIME_COMPAT(, NULL, NULL, &G.opt_charset, NULL, NULL, NULL, NULL)
349	);
350	//argc -= optind;
351	argv += optind;
352
353	return parse("", (opts & OPT_X) ? argv : (char **)&opt_prefix);
354	}