xref: /freebsd/lib/libc/gen/wordexp.c (revision 4e8d558c)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause
3  *
4  * Copyright (c) 2002 Tim J. Robbins.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  *
16  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26  * SUCH DAMAGE.
27  */
28 
29 #include "namespace.h"
30 #include <sys/cdefs.h>
31 #include <sys/types.h>
32 #include <sys/wait.h>
33 #include <errno.h>
34 #include <fcntl.h>
35 #include <paths.h>
36 #include <signal.h>
37 #include <stdbool.h>
38 #include <stdio.h>
39 #include <stdlib.h>
40 #include <string.h>
41 #include <unistd.h>
42 #include <wordexp.h>
43 #include "un-namespace.h"
44 #include "libc_private.h"
45 
46 __FBSDID("$FreeBSD$");
47 
48 static int	we_askshell(const char *, wordexp_t *, int);
49 static int	we_check(const char *);
50 
51 /*
52  * wordexp --
53  *	Perform shell word expansion on `words' and place the resulting list
54  *	of words in `we'. See wordexp(3).
55  *
56  *	Specified by IEEE Std. 1003.1-2001.
57  */
58 int
59 wordexp(const char * __restrict words, wordexp_t * __restrict we, int flags)
60 {
61 	int error;
62 
63 	if (flags & WRDE_REUSE)
64 		wordfree(we);
65 	if ((flags & WRDE_APPEND) == 0) {
66 		we->we_wordc = 0;
67 		we->we_wordv = NULL;
68 		we->we_strings = NULL;
69 		we->we_nbytes = 0;
70 	}
71 	if ((error = we_check(words)) != 0) {
72 		wordfree(we);
73 		return (error);
74 	}
75 	if ((error = we_askshell(words, we, flags)) != 0) {
76 		wordfree(we);
77 		return (error);
78 	}
79 	return (0);
80 }
81 
82 static size_t
83 we_read_fully(int fd, char *buffer, size_t len)
84 {
85 	size_t done;
86 	ssize_t nread;
87 
88 	done = 0;
89 	do {
90 		nread = _read(fd, buffer + done, len - done);
91 		if (nread == -1 && errno == EINTR)
92 			continue;
93 		if (nread <= 0)
94 			break;
95 		done += nread;
96 	} while (done != len);
97 	return done;
98 }
99 
100 static bool
101 we_write_fully(int fd, const char *buffer, size_t len)
102 {
103 	size_t done;
104 	ssize_t nwritten;
105 
106 	done = 0;
107 	do {
108 		nwritten = _write(fd, buffer + done, len - done);
109 		if (nwritten == -1 && errno == EINTR)
110 			continue;
111 		if (nwritten <= 0)
112 			return (false);
113 		done += nwritten;
114 	} while (done != len);
115 	return (true);
116 }
117 
118 /*
119  * we_askshell --
120  *	Use the `freebsd_wordexp' /bin/sh builtin function to do most of the
121  *	work in expanding the word string. This function is complicated by
122  *	memory management.
123  */
124 static int
125 we_askshell(const char *words, wordexp_t *we, int flags)
126 {
127 	int pdesw[2];			/* Pipe for writing words */
128 	int pdes[2];			/* Pipe for reading output */
129 	char wfdstr[sizeof(int) * 3 + 1];
130 	char buf[35];			/* Buffer for byte and word count */
131 	long nwords, nbytes;		/* Number of words, bytes from child */
132 	long i;				/* Handy integer */
133 	size_t sofs;			/* Offset into we->we_strings */
134 	size_t vofs;			/* Offset into we->we_wordv */
135 	pid_t pid;			/* Process ID of child */
136 	pid_t wpid;			/* waitpid return value */
137 	int status;			/* Child exit status */
138 	int error;			/* Our return value */
139 	int serrno;			/* errno to return */
140 	char *np, *p;			/* Handy pointers */
141 	char *nstrings;			/* Temporary for realloc() */
142 	char **nwv;			/* Temporary for realloc() */
143 	sigset_t newsigblock, oldsigblock;
144 	const char *ifs;
145 
146 	serrno = errno;
147 	ifs = getenv("IFS");
148 
149 	if (pipe2(pdesw, O_CLOEXEC) < 0)
150 		return (WRDE_NOSPACE);	/* XXX */
151 	snprintf(wfdstr, sizeof(wfdstr), "%d", pdesw[0]);
152 	if (pipe2(pdes, O_CLOEXEC) < 0) {
153 		_close(pdesw[0]);
154 		_close(pdesw[1]);
155 		return (WRDE_NOSPACE);	/* XXX */
156 	}
157 	(void)sigemptyset(&newsigblock);
158 	(void)sigaddset(&newsigblock, SIGCHLD);
159 	(void)__libc_sigprocmask(SIG_BLOCK, &newsigblock, &oldsigblock);
160 	if ((pid = fork()) < 0) {
161 		serrno = errno;
162 		_close(pdesw[0]);
163 		_close(pdesw[1]);
164 		_close(pdes[0]);
165 		_close(pdes[1]);
166 		(void)__libc_sigprocmask(SIG_SETMASK, &oldsigblock, NULL);
167 		errno = serrno;
168 		return (WRDE_NOSPACE);	/* XXX */
169 	}
170 	else if (pid == 0) {
171 		/*
172 		 * We are the child; make /bin/sh expand `words'.
173 		 */
174 		(void)__libc_sigprocmask(SIG_SETMASK, &oldsigblock, NULL);
175 		if ((pdes[1] != STDOUT_FILENO ?
176 		    _dup2(pdes[1], STDOUT_FILENO) :
177 		    _fcntl(pdes[1], F_SETFD, 0)) < 0)
178 			_exit(1);
179 		if (_fcntl(pdesw[0], F_SETFD, 0) < 0)
180 			_exit(1);
181 		execl(_PATH_BSHELL, "sh", flags & WRDE_UNDEF ? "-u" : "+u",
182 		    "-c", "IFS=$1;eval \"$2\";"
183 		    "freebsd_wordexp -f \"$3\" ${4:+\"$4\"}",
184 		    "",
185 		    ifs != NULL ? ifs : " \t\n",
186 		    flags & WRDE_SHOWERR ? "" : "exec 2>/dev/null",
187 		    wfdstr,
188 		    flags & WRDE_NOCMD ? "-p" : "",
189 		    (char *)NULL);
190 		_exit(1);
191 	}
192 
193 	/*
194 	 * We are the parent; write the words.
195 	 */
196 	_close(pdes[1]);
197 	_close(pdesw[0]);
198 	if (!we_write_fully(pdesw[1], words, strlen(words))) {
199 		_close(pdesw[1]);
200 		error = WRDE_SYNTAX;
201 		goto cleanup;
202 	}
203 	_close(pdesw[1]);
204 	/*
205 	 * Read the output of the shell wordexp function,
206 	 * which is a byte indicating that the words were parsed successfully,
207 	 * a 64-bit hexadecimal word count, a dummy byte, a 64-bit hexadecimal
208 	 * byte count (not including terminating null bytes), followed by the
209 	 * expanded words separated by nulls.
210 	 */
211 	switch (we_read_fully(pdes[0], buf, 34)) {
212 	case 1:
213 		error = buf[0] == 'C' ? WRDE_CMDSUB : WRDE_BADVAL;
214 		serrno = errno;
215 		goto cleanup;
216 	case 34:
217 		break;
218 	default:
219 		error = WRDE_SYNTAX;
220 		serrno = errno;
221 		goto cleanup;
222 	}
223 	buf[17] = '\0';
224 	nwords = strtol(buf + 1, NULL, 16);
225 	buf[34] = '\0';
226 	nbytes = strtol(buf + 18, NULL, 16) + nwords;
227 
228 	/*
229 	 * Allocate or reallocate (when flags & WRDE_APPEND) the word vector
230 	 * and string storage buffers for the expanded words we're about to
231 	 * read from the child.
232 	 */
233 	sofs = we->we_nbytes;
234 	vofs = we->we_wordc;
235 	if ((flags & (WRDE_DOOFFS|WRDE_APPEND)) == (WRDE_DOOFFS|WRDE_APPEND))
236 		vofs += we->we_offs;
237 	we->we_wordc += nwords;
238 	we->we_nbytes += nbytes;
239 	if ((nwv = reallocarray(we->we_wordv, (we->we_wordc + 1 +
240 	    (flags & WRDE_DOOFFS ? we->we_offs : 0)),
241 	    sizeof(char *))) == NULL) {
242 		error = WRDE_NOSPACE;
243 		goto cleanup;
244 	}
245 	we->we_wordv = nwv;
246 	if ((nstrings = realloc(we->we_strings, we->we_nbytes)) == NULL) {
247 		error = WRDE_NOSPACE;
248 		goto cleanup;
249 	}
250 	for (i = 0; i < vofs; i++)
251 		if (we->we_wordv[i] != NULL)
252 			we->we_wordv[i] += nstrings - we->we_strings;
253 	we->we_strings = nstrings;
254 
255 	if (we_read_fully(pdes[0], we->we_strings + sofs, nbytes) != nbytes) {
256 		error = WRDE_NOSPACE; /* abort for unknown reason */
257 		serrno = errno;
258 		goto cleanup;
259 	}
260 
261 	error = 0;
262 cleanup:
263 	_close(pdes[0]);
264 	do
265 		wpid = _waitpid(pid, &status, 0);
266 	while (wpid < 0 && errno == EINTR);
267 	(void)__libc_sigprocmask(SIG_SETMASK, &oldsigblock, NULL);
268 	if (error != 0) {
269 		errno = serrno;
270 		return (error);
271 	}
272 	if (wpid < 0 || !WIFEXITED(status) || WEXITSTATUS(status) != 0)
273 		return (WRDE_NOSPACE); /* abort for unknown reason */
274 
275 	/*
276 	 * Break the null-terminated expanded word strings out into
277 	 * the vector.
278 	 */
279 	if (vofs == 0 && flags & WRDE_DOOFFS)
280 		while (vofs < we->we_offs)
281 			we->we_wordv[vofs++] = NULL;
282 	p = we->we_strings + sofs;
283 	while (nwords-- != 0) {
284 		we->we_wordv[vofs++] = p;
285 		if ((np = memchr(p, '\0', nbytes)) == NULL)
286 			return (WRDE_NOSPACE);	/* XXX */
287 		nbytes -= np - p + 1;
288 		p = np + 1;
289 	}
290 	we->we_wordv[vofs] = NULL;
291 
292 	return (0);
293 }
294 
295 /*
296  * we_check --
297  *	Check that the string contains none of the following unquoted
298  *	special characters: <newline> |&;<>(){}
299  *	This mainly serves for {} which are normally legal in sh.
300  *	It deliberately does not attempt to model full sh syntax.
301  */
302 static int
303 we_check(const char *words)
304 {
305 	char c;
306 	/* Saw \ or $, possibly not special: */
307 	bool quote = false, dollar = false;
308 	/* Saw ', ", ${, ` or $(, possibly not special: */
309 	bool have_sq = false, have_dq = false, have_par_begin = false;
310 	bool have_cmd = false;
311 	/* Definitely saw a ', ", ${, ` or $(, need a closing character: */
312 	bool need_sq = false, need_dq = false, need_par_end = false;
313 	bool need_cmd_old = false, need_cmd_new = false;
314 
315 	while ((c = *words++) != '\0') {
316 		switch (c) {
317 		case '\\':
318 			quote = !quote;
319 			continue;
320 		case '$':
321 			if (quote)
322 				quote = false;
323 			else
324 				dollar = !dollar;
325 			continue;
326 		case '\'':
327 			if (!quote && !have_sq && !have_dq)
328 				need_sq = true;
329 			else
330 				need_sq = false;
331 			have_sq = true;
332 			break;
333 		case '"':
334 			if (!quote && !have_sq && !have_dq)
335 				need_dq = true;
336 			else
337 				need_dq = false;
338 			have_dq = true;
339 			break;
340 		case '`':
341 			if (!quote && !have_sq && !have_cmd)
342 				need_cmd_old = true;
343 			else
344 				need_cmd_old = false;
345 			have_cmd = true;
346 			break;
347 		case '{':
348 			if (!quote && !dollar && !have_sq && !have_dq &&
349 			    !have_cmd)
350 				return (WRDE_BADCHAR);
351 			if (dollar) {
352 				if (!quote && !have_sq)
353 					need_par_end = true;
354 				have_par_begin = true;
355 			}
356 			break;
357 		case '}':
358 			if (!quote && !have_sq && !have_dq && !have_par_begin &&
359 			    !have_cmd)
360 				return (WRDE_BADCHAR);
361 			need_par_end = false;
362 			break;
363 		case '(':
364 			if (!quote && !dollar && !have_sq && !have_dq &&
365 			    !have_cmd)
366 				return (WRDE_BADCHAR);
367 			if (dollar) {
368 				if (!quote && !have_sq)
369 					need_cmd_new = true;
370 				have_cmd = true;
371 			}
372 			break;
373 		case ')':
374 			if (!quote && !have_sq && !have_dq && !have_cmd)
375 				return (WRDE_BADCHAR);
376 			need_cmd_new = false;
377 			break;
378 		case '|': case '&': case ';': case '<': case '>': case '\n':
379 			if (!quote && !have_sq && !have_dq && !have_cmd)
380 				return (WRDE_BADCHAR);
381 			break;
382 		default:
383 			break;
384 		}
385 		quote = dollar = false;
386 	}
387 	if (quote || dollar || need_sq || need_dq || need_par_end ||
388 	    need_cmd_old || need_cmd_new)
389 		return (WRDE_SYNTAX);
390 
391 	return (0);
392 }
393 
394 /*
395  * wordfree --
396  *	Free the result of wordexp(). See wordexp(3).
397  *
398  *	Specified by IEEE Std. 1003.1-2001.
399  */
400 void
401 wordfree(wordexp_t *we)
402 {
403 
404 	if (we == NULL)
405 		return;
406 	free(we->we_wordv);
407 	free(we->we_strings);
408 	we->we_wordv = NULL;
409 	we->we_strings = NULL;
410 	we->we_nbytes = 0;
411 	we->we_wordc = 0;
412 }
413