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 * $FreeBSD: head/usr.bin/env/envopts.c 326276 2017-11-27 15:37:16Z pfg $
33 */
34
35 #include <sys/stat.h>
36 #include <sys/param.h>
37 #include <err.h>
38 #include <errno.h>
39 #include <ctype.h>
40 #include <stdio.h>
41 #include <stdlib.h>
42 #include <string.h>
43 #include <unistd.h>
44
45 #include "envopts.h"
46
47 static const char *
48 expand_vars(int in_thisarg, char **thisarg_p, char **dest_p,
49 const char **src_p);
50 static int is_there(char *candidate);
51
52 /*
53 * The is*() routines take a parameter of 'int', but expect values in the range
54 * of unsigned char. Define some wrappers which take a value of type 'char',
55 * whether signed or unsigned, and ensure the value ends up in the right range.
56 */
57 #define isalnumch(Anychar) isalnum((u_char)(Anychar))
58 #define isalphach(Anychar) isalpha((u_char)(Anychar))
59 #define isspacech(Anychar) isspace((u_char)(Anychar))
60
61 /*
62 * Routine to determine if a given fully-qualified filename is executable.
63 * This is copied almost verbatim from FreeBSD's usr.bin/which/which.c.
64 */
65 static int
is_there(char * candidate)66 is_there(char *candidate)
67 {
68 struct stat fin;
69
70 /* XXX work around access(2) false positives for superuser */
71 if (access(candidate, X_OK) == 0 &&
72 stat(candidate, &fin) == 0 &&
73 S_ISREG(fin.st_mode) &&
74 (getuid() != 0 ||
75 (fin.st_mode & (S_IXUSR | S_IXGRP | S_IXOTH)) != 0)) {
76 if (env_verbosity > 1)
77 fprintf(stderr, "#env matched:\t'%s'\n", candidate);
78 return (1);
79 }
80 return (0);
81 }
82
83 /**
84 * Routine to search through an alternate path-list, looking for a given
85 * filename to execute. If the file is found, replace the original
86 * unqualified name with a fully-qualified path. This allows `env' to
87 * execute programs from a specific strict list of possible paths, without
88 * changing the value of PATH seen by the program which will be executed.
89 * E.G.:
90 * #!/usr/bin/env -S-P/usr/local/bin:/usr/bin perl
91 * will execute /usr/local/bin/perl or /usr/bin/perl (whichever is found
92 * first), no matter what the current value of PATH is, and without
93 * changing the value of PATH that the script will see when it runs.
94 *
95 * This is similar to the print_matches() routine in usr.bin/which/which.c.
96 */
97 void
search_paths(char * path,char ** argv)98 search_paths(char *path, char **argv)
99 {
100 char candidate[PATH_MAX];
101 const char *d;
102 char *filename, *fqname;
103
104 /* If the file has a `/' in it, then no search is done */
105 filename = *argv;
106 if (strchr(filename, '/') != NULL)
107 return;
108
109 if (env_verbosity > 1) {
110 fprintf(stderr, "#env Searching:\t'%s'\n", path);
111 fprintf(stderr, "#env for file:\t'%s'\n", filename);
112 }
113
114 fqname = NULL;
115 while ((d = strsep(&path, ":")) != NULL) {
116 if (*d == '\0')
117 d = ".";
118 if (snprintf(candidate, sizeof(candidate), "%s/%s", d,
119 filename) >= (int)sizeof(candidate))
120 continue;
121 if (is_there(candidate)) {
122 fqname = candidate;
123 break;
124 }
125 }
126
127 if (fqname == NULL) {
128 errno = ENOENT;
129 err(127, "%s", filename);
130 }
131 *argv = strdup(candidate);
132 }
133
134 /**
135 * Routine to split a string into multiple parameters, while recognizing a
136 * few special characters. It recognizes both single and double-quoted
137 * strings. This processing is designed entirely for the benefit of the
138 * parsing of "#!"-lines (aka "shebang" lines == the first line of an
139 * executable script). Different operating systems parse that line in very
140 * different ways, and this split-on-spaces processing is meant to provide
141 * ways to specify arbitrary arguments on that line, no matter how the OS
142 * parses it.
143 *
144 * Within a single-quoted string, the two characters "\'" are treated as
145 * a literal "'" character to add to the string, and "\\" are treated as
146 * a literal "\" character to add. Other than that, all characters are
147 * copied until the processing gets to a terminating "'".
148 *
149 * Within a double-quoted string, many more "\"-style escape sequences
150 * are recognized, mostly copied from what is recognized in the `printf'
151 * command. Some OS's will not allow a literal blank character to be
152 * included in the one argument that they recognize on a shebang-line,
153 * so a few additional escape-sequences are defined to provide ways to
154 * specify blanks.
155 *
156 * Within a double-quoted string "\_" is turned into a literal blank.
157 * (Inside of a single-quoted string, the two characters are just copied)
158 * Outside of a quoted string, "\_" is treated as both a blank, and the
159 * end of the current argument. So with a shelbang-line of:
160 * #!/usr/bin/env -SA=avalue\_perl
161 * the -S value would be broken up into arguments "A=avalue" and "perl".
162 */
163 void
split_spaces(const char * str,int * origind,int * origc,char *** origv)164 split_spaces(const char *str, int *origind, int *origc, char ***origv)
165 {
166 static const char *nullarg = "";
167 const char *bq_src, *copystr, *src;
168 char *dest, **newargv, *newstr, **nextarg, **oldarg;
169 int addcount, bq_destlen, copychar, found_sep, in_arg, in_dq, in_sq;
170
171 /*
172 * Ignore leading space on the string, and then malloc enough room
173 * to build a copy of it. The copy might end up shorter than the
174 * original, due to quoted strings and '\'-processing.
175 */
176 while (isspacech(*str))
177 str++;
178 if (*str == '\0')
179 return;
180 newstr = malloc(strlen(str) + 1);
181
182 /*
183 * Allocate plenty of space for the new array of arg-pointers,
184 * and start that array off with the first element of the old
185 * array.
186 */
187 newargv = malloc((*origc + (strlen(str) / 2) + 2) * sizeof(char *));
188 nextarg = newargv;
189 *nextarg++ = **origv;
190
191 /* Come up with the new args by splitting up the given string. */
192 addcount = 0;
193 bq_destlen = in_arg = in_dq = in_sq = 0;
194 bq_src = NULL;
195 for (src = str, dest = newstr; *src != '\0'; src++) {
196 /*
197 * This switch will look at a character in *src, and decide
198 * what should be copied to *dest. It only decides what
199 * character(s) to copy, it should not modify *dest. In some
200 * cases, it will look at multiple characters from *src.
201 */
202 copychar = found_sep = 0;
203 copystr = NULL;
204 switch (*src) {
205 case '"':
206 if (in_sq)
207 copychar = *src;
208 else if (in_dq)
209 in_dq = 0;
210 else {
211 /*
212 * Referencing nullarg ensures that a new
213 * argument is created, even if this quoted
214 * string ends up with zero characters.
215 */
216 copystr = nullarg;
217 in_dq = 1;
218 bq_destlen = dest - *(nextarg - 1);
219 bq_src = src;
220 }
221 break;
222 case '$':
223 if (in_sq)
224 copychar = *src;
225 else {
226 copystr = expand_vars(in_arg, (nextarg - 1),
227 &dest, &src);
228 }
229 break;
230 case '\'':
231 if (in_dq)
232 copychar = *src;
233 else if (in_sq)
234 in_sq = 0;
235 else {
236 /*
237 * Referencing nullarg ensures that a new
238 * argument is created, even if this quoted
239 * string ends up with zero characters.
240 */
241 copystr = nullarg;
242 in_sq = 1;
243 bq_destlen = dest - *(nextarg - 1);
244 bq_src = src;
245 }
246 break;
247 case '\\':
248 if (in_sq) {
249 /*
250 * Inside single-quoted strings, only the
251 * "\'" and "\\" are recognized as special
252 * strings.
253 */
254 copychar = *(src + 1);
255 if (copychar == '\'' || copychar == '\\')
256 src++;
257 else
258 copychar = *src;
259 break;
260 }
261 src++;
262 switch (*src) {
263 case '"':
264 case '#':
265 case '$':
266 case '\'':
267 case '\\':
268 copychar = *src;
269 break;
270 case '_':
271 /*
272 * Alternate way to get a blank, which allows
273 * that blank be used to separate arguments
274 * when it is not inside a quoted string.
275 */
276 if (in_dq)
277 copychar = ' ';
278 else {
279 found_sep = 1;
280 src++;
281 }
282 break;
283 case 'c':
284 /*
285 * Ignore remaining characters in the -S string.
286 * This would not make sense if found in the
287 * middle of a quoted string.
288 */
289 if (in_dq)
290 errx(1, "Sequence '\\%c' is not allowed"
291 " in quoted strings", *src);
292 goto str_done;
293 case 'f':
294 copychar = '\f';
295 break;
296 case 'n':
297 copychar = '\n';
298 break;
299 case 'r':
300 copychar = '\r';
301 break;
302 case 't':
303 copychar = '\t';
304 break;
305 case 'v':
306 copychar = '\v';
307 break;
308 default:
309 if (isspacech(*src))
310 copychar = *src;
311 else
312 errx(1, "Invalid sequence '\\%c' in -S",
313 *src);
314 }
315 break;
316 default:
317 if ((in_dq || in_sq) && in_arg)
318 copychar = *src;
319 else if (isspacech(*src))
320 found_sep = 1;
321 else {
322 /*
323 * If the first character of a new argument
324 * is `#', then ignore the remaining chars.
325 */
326 if (!in_arg && *src == '#')
327 goto str_done;
328 copychar = *src;
329 }
330 }
331 /*
332 * Now that the switch has determined what (if anything)
333 * needs to be copied, copy whatever that is to *dest.
334 */
335 if (copychar || copystr != NULL) {
336 if (!in_arg) {
337 /* This is the first byte of a new argument */
338 *nextarg++ = dest;
339 addcount++;
340 in_arg = 1;
341 }
342 if (copychar)
343 *dest++ = (char)copychar;
344 else if (copystr != NULL)
345 while (*copystr != '\0')
346 *dest++ = *copystr++;
347 } else if (found_sep) {
348 *dest++ = '\0';
349 while (isspacech(*src))
350 src++;
351 --src;
352 in_arg = 0;
353 }
354 }
355 str_done:
356 *dest = '\0';
357 *nextarg = NULL;
358 if (in_dq || in_sq) {
359 errx(1, "No terminating quote for string: %.*s%s",
360 bq_destlen, *(nextarg - 1), bq_src);
361 }
362 if (env_verbosity > 1) {
363 fprintf(stderr, "#env split -S:\t'%s'\n", str);
364 oldarg = newargv + 1;
365 fprintf(stderr, "#env into:\t'%s'\n", *oldarg);
366 for (oldarg++; *oldarg; oldarg++)
367 fprintf(stderr, "#env &\t'%s'\n", *oldarg);
368 }
369
370 /* Copy the unprocessed arg-pointers from the original array */
371 for (oldarg = *origv + *origind; *oldarg; oldarg++)
372 *nextarg++ = *oldarg;
373 *nextarg = NULL;
374
375 /* Update optind/argc/argv in the calling routine */
376 *origc += addcount - *origind + 1;
377 *origv = newargv;
378 *origind = 1;
379 }
380
381 /**
382 * Routine to split expand any environment variables referenced in the string
383 * that -S is processing. For now it only supports the form ${VARNAME}. It
384 * explicitly does not support $VARNAME, and obviously can not handle special
385 * shell-variables such as $?, $*, $1, etc. It is called with *src_p pointing
386 * at the initial '$', and if successful it will update *src_p, *dest_p, and
387 * possibly *thisarg_p in the calling routine.
388 */
389 static const char *
expand_vars(int in_thisarg,char ** thisarg_p,char ** dest_p,const char ** src_p)390 expand_vars(int in_thisarg, char **thisarg_p, char **dest_p, const char **src_p)
391 {
392 const char *vbegin, *vend, *vvalue;
393 char *newstr, *vname;
394 int bad_reference;
395 size_t namelen, newlen;
396
397 bad_reference = 1;
398 vbegin = vend = (*src_p) + 1;
399 if (*vbegin++ == '{')
400 if (*vbegin == '_' || isalphach(*vbegin)) {
401 vend = vbegin + 1;
402 while (*vend == '_' || isalnumch(*vend))
403 vend++;
404 if (*vend == '}')
405 bad_reference = 0;
406 }
407 if (bad_reference)
408 errx(1, "Only ${VARNAME} expansion is supported, error at: %s",
409 *src_p);
410
411 /*
412 * We now know we have a valid environment variable name, so update
413 * the caller's source-pointer to the last character in that reference,
414 * and then pick up the matching value. If the variable is not found,
415 * or if it has a null value, then our work here is done.
416 */
417 *src_p = vend;
418 namelen = vend - vbegin + 1;
419 vname = malloc(namelen);
420 strlcpy(vname, vbegin, namelen);
421 vvalue = getenv(vname);
422 if (vvalue == NULL || *vvalue == '\0') {
423 if (env_verbosity > 2)
424 fprintf(stderr,
425 "#env replacing ${%s} with null string\n",
426 vname);
427 free(vname);
428 return (NULL);
429 }
430
431 if (env_verbosity > 2)
432 fprintf(stderr, "#env expanding ${%s} into '%s'\n", vname,
433 vvalue);
434
435 /*
436 * There is some value to copy to the destination. If the value is
437 * shorter than the ${VARNAME} reference that it replaces, then our
438 * caller can just copy the value to the existing destination.
439 */
440 if (strlen(vname) + 3 >= strlen(vvalue)) {
441 free(vname);
442 return (vvalue);
443 }
444
445 /*
446 * The value is longer than the string it replaces, which means the
447 * present destination area is too small to hold it. Create a new
448 * destination area, and update the caller's 'dest' variable to match.
449 * If the caller has already started copying some info for 'thisarg'
450 * into the present destination, then the new destination area must
451 * include a copy of that data, and the pointer to 'thisarg' must also
452 * be updated. Note that it is still the caller which copies this
453 * vvalue to the new *dest.
454 */
455 newlen = strlen(vvalue) + strlen(*src_p) + 1;
456 if (in_thisarg) {
457 **dest_p = '\0'; /* Provide terminator for 'thisarg' */
458 newlen += strlen(*thisarg_p);
459 newstr = malloc(newlen);
460 strcpy(newstr, *thisarg_p);
461 *thisarg_p = newstr;
462 } else {
463 newstr = malloc(newlen);
464 *newstr = '\0';
465 }
466 *dest_p = strchr(newstr, '\0');
467 free(vname);
468 return (vvalue);
469 }
470