xref: /freebsd/usr.bin/env/envopts.c (revision 0957b409)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
3  *
4  * Copyright (c) 2005  - Garance Alistair Drosehn <gad@FreeBSD.org>.
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  * The views and conclusions contained in the software and documentation
29  * are those of the authors and should not be interpreted as representing
30  * official policies, either expressed or implied, of the FreeBSD Project.
31  */
32 
33 #include <sys/cdefs.h>
34 __FBSDID("$FreeBSD$");
35 
36 #include <sys/stat.h>
37 #include <sys/param.h>
38 #include <err.h>
39 #include <errno.h>
40 #include <ctype.h>
41 #include <stdio.h>
42 #include <stdlib.h>
43 #include <string.h>
44 #include <unistd.h>
45 
46 #include "envopts.h"
47 
48 static const char *
49 		 expand_vars(int in_thisarg, char **thisarg_p, char **dest_p,
50 		     const char **src_p);
51 static int	 is_there(char *candidate);
52 
53 /*
54  * The is*() routines take a parameter of 'int', but expect values in the range
55  * of unsigned char.  Define some wrappers which take a value of type 'char',
56  * whether signed or unsigned, and ensure the value ends up in the right range.
57  */
58 #define	isalnumch(Anychar) isalnum((u_char)(Anychar))
59 #define	isalphach(Anychar) isalpha((u_char)(Anychar))
60 #define	isspacech(Anychar) isspace((u_char)(Anychar))
61 
62 /*
63  * Routine to determine if a given fully-qualified filename is executable.
64  * This is copied almost verbatim from FreeBSD's usr.bin/which/which.c.
65  */
66 static int
67 is_there(char *candidate)
68 {
69         struct stat fin;
70 
71         /* XXX work around access(2) false positives for superuser */
72         if (access(candidate, X_OK) == 0 &&
73             stat(candidate, &fin) == 0 &&
74             S_ISREG(fin.st_mode) &&
75             (getuid() != 0 ||
76             (fin.st_mode & (S_IXUSR | S_IXGRP | S_IXOTH)) != 0)) {
77                 if (env_verbosity > 1)
78 			fprintf(stderr, "#env   matched:\t'%s'\n", candidate);
79                 return (1);
80         }
81         return (0);
82 }
83 
84 /**
85  * Routine to search through an alternate path-list, looking for a given
86  * filename to execute.  If the file is found, replace the original
87  * unqualified name with a fully-qualified path.  This allows `env' to
88  * execute programs from a specific strict list of possible paths, without
89  * changing the value of PATH seen by the program which will be executed.
90  * E.G.:
91  *	#!/usr/bin/env -S-P/usr/local/bin:/usr/bin perl
92  * will execute /usr/local/bin/perl or /usr/bin/perl (whichever is found
93  * first), no matter what the current value of PATH is, and without
94  * changing the value of PATH that the script will see when it runs.
95  *
96  * This is similar to the print_matches() routine in usr.bin/which/which.c.
97  */
98 void
99 search_paths(char *path, char **argv)
100 {
101         char candidate[PATH_MAX];
102         const char *d;
103 	char *filename, *fqname;
104 
105 	/* If the file has a `/' in it, then no search is done */
106 	filename = *argv;
107 	if (strchr(filename, '/') != NULL)
108 		return;
109 
110 	if (env_verbosity > 1) {
111 		fprintf(stderr, "#env Searching:\t'%s'\n", path);
112 		fprintf(stderr, "#env  for file:\t'%s'\n", filename);
113 	}
114 
115 	fqname = NULL;
116         while ((d = strsep(&path, ":")) != NULL) {
117                 if (*d == '\0')
118                         d = ".";
119                 if (snprintf(candidate, sizeof(candidate), "%s/%s", d,
120                     filename) >= (int)sizeof(candidate))
121                         continue;
122                 if (is_there(candidate)) {
123                         fqname = candidate;
124 			break;
125                 }
126         }
127 
128 	if (fqname == NULL) {
129 		errno = ENOENT;
130 		err(127, "%s", filename);
131 	}
132 	*argv = strdup(candidate);
133 }
134 
135 /**
136  * Routine to split a string into multiple parameters, while recognizing a
137  * few special characters.  It recognizes both single and double-quoted
138  * strings.  This processing is designed entirely for the benefit of the
139  * parsing of "#!"-lines (aka "shebang" lines == the first line of an
140  * executable script).  Different operating systems parse that line in very
141  * different ways, and this split-on-spaces processing is meant to provide
142  * ways to specify arbitrary arguments on that line, no matter how the OS
143  * parses it.
144  *
145  * Within a single-quoted string, the two characters "\'" are treated as
146  * a literal "'" character to add to the string, and "\\" are treated as
147  * a literal "\" character to add.  Other than that, all characters are
148  * copied until the processing gets to a terminating "'".
149  *
150  * Within a double-quoted string, many more "\"-style escape sequences
151  * are recognized, mostly copied from what is recognized in the `printf'
152  * command.  Some OS's will not allow a literal blank character to be
153  * included in the one argument that they recognize on a shebang-line,
154  * so a few additional escape-sequences are defined to provide ways to
155  * specify blanks.
156  *
157  * Within a double-quoted string "\_" is turned into a literal blank.
158  * (Inside of a single-quoted string, the two characters are just copied)
159  * Outside of a quoted string, "\_" is treated as both a blank, and the
160  * end of the current argument.  So with a shelbang-line of:
161  *		#!/usr/bin/env -SA=avalue\_perl
162  * the -S value would be broken up into arguments "A=avalue" and "perl".
163  */
164 void
165 split_spaces(const char *str, int *origind, int *origc, char ***origv)
166 {
167 	static const char *nullarg = "";
168 	const char *bq_src, *copystr, *src;
169 	char *dest, **newargv, *newstr, **nextarg, **oldarg;
170 	int addcount, bq_destlen, copychar, found_sep, in_arg, in_dq, in_sq;
171 
172 	/*
173 	 * Ignore leading space on the string, and then malloc enough room
174 	 * to build a copy of it.  The copy might end up shorter than the
175 	 * original, due to quoted strings and '\'-processing.
176 	 */
177 	while (isspacech(*str))
178 		str++;
179 	if (*str == '\0')
180 		return;
181 	newstr = malloc(strlen(str) + 1);
182 
183 	/*
184 	 * Allocate plenty of space for the new array of arg-pointers,
185 	 * and start that array off with the first element of the old
186 	 * array.
187 	 */
188 	newargv = malloc((*origc + (strlen(str) / 2) + 2) * sizeof(char *));
189 	nextarg = newargv;
190 	*nextarg++ = **origv;
191 
192 	/* Come up with the new args by splitting up the given string. */
193 	addcount = 0;
194 	bq_destlen = in_arg = in_dq = in_sq = 0;
195 	bq_src = NULL;
196 	for (src = str, dest = newstr; *src != '\0'; src++) {
197 		/*
198 		 * This switch will look at a character in *src, and decide
199 		 * what should be copied to *dest.  It only decides what
200 		 * character(s) to copy, it should not modify *dest.  In some
201 		 * cases, it will look at multiple characters from *src.
202 		 */
203 		copychar = found_sep = 0;
204 		copystr = NULL;
205 		switch (*src) {
206 		case '"':
207 			if (in_sq)
208 				copychar = *src;
209 			else if (in_dq)
210 				in_dq = 0;
211 			else {
212 				/*
213 				 * Referencing nullarg ensures that a new
214 				 * argument is created, even if this quoted
215 				 * string ends up with zero characters.
216 				 */
217 				copystr = nullarg;
218 				in_dq = 1;
219 				bq_destlen = dest - *(nextarg - 1);
220 				bq_src = src;
221 			}
222 			break;
223 		case '$':
224 			if (in_sq)
225 				copychar = *src;
226 			else {
227 				copystr = expand_vars(in_arg, (nextarg - 1),
228 				    &dest, &src);
229 			}
230 			break;
231 		case '\'':
232 			if (in_dq)
233 				copychar = *src;
234 			else if (in_sq)
235 				in_sq = 0;
236 			else {
237 				/*
238 				 * Referencing nullarg ensures that a new
239 				 * argument is created, even if this quoted
240 				 * string ends up with zero characters.
241 				 */
242 				copystr = nullarg;
243 				in_sq = 1;
244 				bq_destlen = dest - *(nextarg - 1);
245 				bq_src = src;
246 			}
247 			break;
248 		case '\\':
249 			if (in_sq) {
250 				/*
251 				 * Inside single-quoted strings, only the
252 				 * "\'" and "\\" are recognized as special
253 				 * strings.
254 				 */
255 				copychar = *(src + 1);
256 				if (copychar == '\'' || copychar == '\\')
257 					src++;
258 				else
259 					copychar = *src;
260 				break;
261 			}
262 			src++;
263 			switch (*src) {
264 			case '"':
265 			case '#':
266 			case '$':
267 			case '\'':
268 			case '\\':
269 				copychar = *src;
270 				break;
271 			case '_':
272 				/*
273 				 * Alternate way to get a blank, which allows
274 				 * that blank be used to separate arguments
275 				 * when it is not inside a quoted string.
276 				 */
277 				if (in_dq)
278 					copychar = ' ';
279 				else {
280 					found_sep = 1;
281 					src++;
282 				}
283 				break;
284 			case 'c':
285 				/*
286 				 * Ignore remaining characters in the -S string.
287 				 * This would not make sense if found in the
288 				 * middle of a quoted string.
289 				 */
290 				if (in_dq)
291 					errx(1, "Sequence '\\%c' is not allowed"
292 					    " in quoted strings", *src);
293 				goto str_done;
294 			case 'f':
295 				copychar = '\f';
296 				break;
297 			case 'n':
298 				copychar = '\n';
299 				break;
300 			case 'r':
301 				copychar = '\r';
302 				break;
303 			case 't':
304 				copychar = '\t';
305 				break;
306 			case 'v':
307 				copychar = '\v';
308 				break;
309 			default:
310 				if (isspacech(*src))
311 					copychar = *src;
312 				else
313 					errx(1, "Invalid sequence '\\%c' in -S",
314 					    *src);
315 			}
316 			break;
317 		default:
318 			if ((in_dq || in_sq) && in_arg)
319 				copychar = *src;
320 			else if (isspacech(*src))
321 				found_sep = 1;
322 			else {
323 				/*
324 				 * If the first character of a new argument
325 				 * is `#', then ignore the remaining chars.
326 				 */
327 				if (!in_arg && *src == '#')
328 					goto str_done;
329 				copychar = *src;
330 			}
331 		}
332 		/*
333 		 * Now that the switch has determined what (if anything)
334 		 * needs to be copied, copy whatever that is to *dest.
335 		 */
336 		if (copychar || copystr != NULL) {
337 			if (!in_arg) {
338 				/* This is the first byte of a new argument */
339 				*nextarg++ = dest;
340 				addcount++;
341 				in_arg = 1;
342 			}
343 			if (copychar)
344 				*dest++ = (char)copychar;
345 			else if (copystr != NULL)
346 				while (*copystr != '\0')
347 					*dest++ = *copystr++;
348 		} else if (found_sep) {
349 			*dest++ = '\0';
350 			while (isspacech(*src))
351 				src++;
352 			--src;
353 			in_arg = 0;
354 		}
355 	}
356 str_done:
357 	*dest = '\0';
358 	*nextarg = NULL;
359 	if (in_dq || in_sq) {
360 		errx(1, "No terminating quote for string: %.*s%s",
361 		    bq_destlen, *(nextarg - 1), bq_src);
362 	}
363 	if (env_verbosity > 1) {
364 		fprintf(stderr, "#env  split -S:\t'%s'\n", str);
365 		oldarg = newargv + 1;
366 		fprintf(stderr, "#env      into:\t'%s'\n", *oldarg);
367 		for (oldarg++; *oldarg; oldarg++)
368 			fprintf(stderr, "#env          &\t'%s'\n", *oldarg);
369 	}
370 
371 	/* Copy the unprocessed arg-pointers from the original array */
372 	for (oldarg = *origv + *origind; *oldarg; oldarg++)
373 		*nextarg++ = *oldarg;
374 	*nextarg = NULL;
375 
376 	/* Update optind/argc/argv in the calling routine */
377 	*origc += addcount - *origind + 1;
378 	*origv = newargv;
379 	*origind = 1;
380 }
381 
382 /**
383  * Routine to split expand any environment variables referenced in the string
384  * that -S is processing.  For now it only supports the form ${VARNAME}.  It
385  * explicitly does not support $VARNAME, and obviously can not handle special
386  * shell-variables such as $?, $*, $1, etc.  It is called with *src_p pointing
387  * at the initial '$', and if successful it will update *src_p, *dest_p, and
388  * possibly *thisarg_p in the calling routine.
389  */
390 static const char *
391 expand_vars(int in_thisarg, char **thisarg_p, char **dest_p, const char **src_p)
392 {
393 	const char *vbegin, *vend, *vvalue;
394 	char *newstr, *vname;
395 	int bad_reference;
396 	size_t namelen, newlen;
397 
398 	bad_reference = 1;
399 	vbegin = vend = (*src_p) + 1;
400 	if (*vbegin++ == '{')
401 		if (*vbegin == '_' || isalphach(*vbegin)) {
402 			vend = vbegin + 1;
403 			while (*vend == '_' || isalnumch(*vend))
404 				vend++;
405 			if (*vend == '}')
406 				bad_reference = 0;
407 		}
408 	if (bad_reference)
409 		errx(1, "Only ${VARNAME} expansion is supported, error at: %s",
410 		    *src_p);
411 
412 	/*
413 	 * We now know we have a valid environment variable name, so update
414 	 * the caller's source-pointer to the last character in that reference,
415 	 * and then pick up the matching value.  If the variable is not found,
416 	 * or if it has a null value, then our work here is done.
417 	 */
418 	*src_p = vend;
419 	namelen = vend - vbegin + 1;
420 	vname = malloc(namelen);
421 	strlcpy(vname, vbegin, namelen);
422 	vvalue = getenv(vname);
423 	if (vvalue == NULL || *vvalue == '\0') {
424 		if (env_verbosity > 2)
425 			fprintf(stderr,
426 			    "#env  replacing ${%s} with null string\n",
427 			    vname);
428 		free(vname);
429 		return (NULL);
430 	}
431 
432 	if (env_verbosity > 2)
433 		fprintf(stderr, "#env  expanding ${%s} into '%s'\n", vname,
434 		    vvalue);
435 
436 	/*
437 	 * There is some value to copy to the destination.  If the value is
438 	 * shorter than the ${VARNAME} reference that it replaces, then our
439 	 * caller can just copy the value to the existing destination.
440 	 */
441 	if (strlen(vname) + 3 >= strlen(vvalue)) {
442 		free(vname);
443 		return (vvalue);
444 	}
445 
446 	/*
447 	 * The value is longer than the string it replaces, which means the
448 	 * present destination area is too small to hold it.  Create a new
449 	 * destination area, and update the caller's 'dest' variable to match.
450 	 * If the caller has already started copying some info for 'thisarg'
451 	 * into the present destination, then the new destination area must
452 	 * include a copy of that data, and the pointer to 'thisarg' must also
453 	 * be updated.  Note that it is still the caller which copies this
454 	 * vvalue to the new *dest.
455 	 */
456 	newlen = strlen(vvalue) + strlen(*src_p) + 1;
457 	if (in_thisarg) {
458 		**dest_p = '\0';	/* Provide terminator for 'thisarg' */
459 		newlen += strlen(*thisarg_p);
460 		newstr = malloc(newlen);
461 		strcpy(newstr, *thisarg_p);
462 		*thisarg_p = newstr;
463 	} else {
464 		newstr = malloc(newlen);
465 		*newstr = '\0';
466 	}
467 	*dest_p = strchr(newstr, '\0');
468 	free(vname);
469 	return (vvalue);
470 }
471