1/*-
2 * Copyright (c) 2005  - Garance Alistair Drosehn <gad@FreeBSD.org>.
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 *   1. Redistributions of source code must retain the above copyright
9 *      notice, this list of conditions and the following disclaimer.
10 *   2. Redistributions in binary form must reproduce the above copyright
11 *      notice, this list of conditions and the following disclaimer in the
12 *      documentation and/or other materials provided with the distribution.
13 *
14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17 * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24 * SUCH DAMAGE.
25 *
26 * The views and conclusions contained in the software and documentation
27 * are those of the authors and should not be interpreted as representing
28 * official policies, either expressed or implied, of the FreeBSD Project.
29 */
30
31#include <sys/cdefs.h>
32__FBSDID("$FreeBSD$");
33
34#include <sys/stat.h>
35#include <sys/param.h>
36#include <err.h>
37#include <errno.h>
38#include <ctype.h>
39#include <stdio.h>
40#include <stdlib.h>
41#include <string.h>
42#include <unistd.h>
43
44#include "envopts.h"
45
46static const char *
47		 expand_vars(int in_thisarg, char **thisarg_p, char **dest_p,
48		     const char **src_p);
49static int	 is_there(char *candidate);
50
51/*
52 * The is*() routines take a parameter of 'int', but expect values in the range
53 * of unsigned char.  Define some wrappers which take a value of type 'char',
54 * whether signed or unsigned, and ensure the value ends up in the right range.
55 */
56#define	isalnumch(Anychar) isalnum((u_char)(Anychar))
57#define	isalphach(Anychar) isalpha((u_char)(Anychar))
58#define	isspacech(Anychar) isspace((u_char)(Anychar))
59
60/*
61 * Routine to determine if a given fully-qualified filename is executable.
62 * This is copied almost verbatim from FreeBSD's usr.bin/which/which.c.
63 */
64static int
65is_there(char *candidate)
66{
67        struct stat fin;
68
69        /* XXX work around access(2) false positives for superuser */
70        if (access(candidate, X_OK) == 0 &&
71            stat(candidate, &fin) == 0 &&
72            S_ISREG(fin.st_mode) &&
73            (getuid() != 0 ||
74            (fin.st_mode & (S_IXUSR | S_IXGRP | S_IXOTH)) != 0)) {
75                if (env_verbosity > 1)
76			fprintf(stderr, "#env   matched:\t'%s'\n", candidate);
77                return (1);
78        }
79        return (0);
80}
81
82/**
83 * Routine to search through an alternate path-list, looking for a given
84 * filename to execute.  If the file is found, replace the original
85 * unqualified name with a fully-qualified path.  This allows `env' to
86 * execute programs from a specific strict list of possible paths, without
87 * changing the value of PATH seen by the program which will be executed.
88 * E.G.:
89 *	#!/usr/bin/env -S-P/usr/local/bin:/usr/bin perl
90 * will execute /usr/local/bin/perl or /usr/bin/perl (whichever is found
91 * first), no matter what the current value of PATH is, and without
92 * changing the value of PATH that the script will see when it runs.
93 *
94 * This is similar to the print_matches() routine in usr.bin/which/which.c.
95 */
96void
97search_paths(char *path, char **argv)
98{
99        char candidate[PATH_MAX];
100        const char *d;
101	char *filename, *fqname;
102
103	/* If the file has a `/' in it, then no search is done */
104	filename = *argv;
105	if (strchr(filename, '/') != NULL)
106		return;
107
108	if (env_verbosity > 1) {
109		fprintf(stderr, "#env Searching:\t'%s'\n", path);
110		fprintf(stderr, "#env  for file:\t'%s'\n", filename);
111	}
112
113	fqname = NULL;
114        while ((d = strsep(&path, ":")) != NULL) {
115                if (*d == '\0')
116                        d = ".";
117                if (snprintf(candidate, sizeof(candidate), "%s/%s", d,
118                    filename) >= (int)sizeof(candidate))
119                        continue;
120                if (is_there(candidate)) {
121                        fqname = candidate;
122			break;
123                }
124        }
125
126	if (fqname == NULL) {
127		errno = ENOENT;
128		err(127, "%s", filename);
129	}
130	*argv = strdup(candidate);
131}
132
133/**
134 * Routine to split a string into multiple parameters, while recognizing a
135 * few special characters.  It recognizes both single and double-quoted
136 * strings.  This processing is designed entirely for the benefit of the
137 * parsing of "#!"-lines (aka "shebang" lines == the first line of an
138 * executable script).  Different operating systems parse that line in very
139 * different ways, and this split-on-spaces processing is meant to provide
140 * ways to specify arbitrary arguments on that line, no matter how the OS
141 * parses it.
142 *
143 * Within a single-quoted string, the two characters "\'" are treated as
144 * a literal "'" character to add to the string, and "\\" are treated as
145 * a literal "\" character to add.  Other than that, all characters are
146 * copied until the processing gets to a terminating "'".
147 *
148 * Within a double-quoted string, many more "\"-style escape sequences
149 * are recognized, mostly copied from what is recognized in the `printf'
150 * command.  Some OS's will not allow a literal blank character to be
151 * included in the one argument that they recognize on a shebang-line,
152 * so a few additional escape-sequences are defined to provide ways to
153 * specify blanks.
154 *
155 * Within a double-quoted string "\_" is turned into a literal blank.
156 * (Inside of a single-quoted string, the two characters are just copied)
157 * Outside of a quoted string, "\_" is treated as both a blank, and the
158 * end of the current argument.  So with a shelbang-line of:
159 *		#!/usr/bin/env -SA=avalue\_perl
160 * the -S value would be broken up into arguments "A=avalue" and "perl".
161 */
162void
163split_spaces(const char *str, int *origind, int *origc, char ***origv)
164{
165	static const char *nullarg = "";
166	const char *bq_src, *copystr, *src;
167	char *dest, **newargv, *newstr, **nextarg, **oldarg;
168	int addcount, bq_destlen, copychar, found_sep, in_arg, in_dq, in_sq;
169
170	/*
171	 * Ignore leading space on the string, and then malloc enough room
172	 * to build a copy of it.  The copy might end up shorter than the
173	 * original, due to quoted strings and '\'-processing.
174	 */
175	while (isspacech(*str))
176		str++;
177	if (*str == '\0')
178		return;
179	newstr = malloc(strlen(str) + 1);
180
181	/*
182	 * Allocate plenty of space for the new array of arg-pointers,
183	 * and start that array off with the first element of the old
184	 * array.
185	 */
186	newargv = malloc((*origc + (strlen(str) / 2) + 2) * sizeof(char *));
187	nextarg = newargv;
188	*nextarg++ = **origv;
189
190	/* Come up with the new args by splitting up the given string. */
191	addcount = 0;
192	bq_destlen = in_arg = in_dq = in_sq = 0;
193	bq_src = NULL;
194	for (src = str, dest = newstr; *src != '\0'; src++) {
195		/*
196		 * This switch will look at a character in *src, and decide
197		 * what should be copied to *dest.  It only decides what
198		 * character(s) to copy, it should not modify *dest.  In some
199		 * cases, it will look at multiple characters from *src.
200		 */
201		copychar = found_sep = 0;
202		copystr = NULL;
203		switch (*src) {
204		case '"':
205			if (in_sq)
206				copychar = *src;
207			else if (in_dq)
208				in_dq = 0;
209			else {
210				/*
211				 * Referencing nullarg ensures that a new
212				 * argument is created, even if this quoted
213				 * string ends up with zero characters.
214				 */
215				copystr = nullarg;
216				in_dq = 1;
217				bq_destlen = dest - *(nextarg - 1);
218				bq_src = src;
219			}
220			break;
221		case '$':
222			if (in_sq)
223				copychar = *src;
224			else {
225				copystr = expand_vars(in_arg, (nextarg - 1),
226				    &dest, &src);
227			}
228			break;
229		case '\'':
230			if (in_dq)
231				copychar = *src;
232			else if (in_sq)
233				in_sq = 0;
234			else {
235				/*
236				 * Referencing nullarg ensures that a new
237				 * argument is created, even if this quoted
238				 * string ends up with zero characters.
239				 */
240				copystr = nullarg;
241				in_sq = 1;
242				bq_destlen = dest - *(nextarg - 1);
243				bq_src = src;
244			}
245			break;
246		case '\\':
247			if (in_sq) {
248				/*
249				 * Inside single-quoted strings, only the
250				 * "\'" and "\\" are recognized as special
251				 * strings.
252				 */
253				copychar = *(src + 1);
254				if (copychar == '\'' || copychar == '\\')
255					src++;
256				else
257					copychar = *src;
258				break;
259			}
260			src++;
261			switch (*src) {
262			case '"':
263			case '#':
264			case '$':
265			case '\'':
266			case '\\':
267				copychar = *src;
268				break;
269			case '_':
270				/*
271				 * Alternate way to get a blank, which allows
272				 * that blank be used to separate arguments
273				 * when it is not inside a quoted string.
274				 */
275				if (in_dq)
276					copychar = ' ';
277				else {
278					found_sep = 1;
279					src++;
280				}
281				break;
282			case 'c':
283				/*
284				 * Ignore remaining characters in the -S string.
285				 * This would not make sense if found in the
286				 * middle of a quoted string.
287				 */
288				if (in_dq)
289					errx(1, "Sequence '\\%c' is not allowed"
290					    " in quoted strings", *src);
291				goto str_done;
292			case 'f':
293				copychar = '\f';
294				break;
295			case 'n':
296				copychar = '\n';
297				break;
298			case 'r':
299				copychar = '\r';
300				break;
301			case 't':
302				copychar = '\t';
303				break;
304			case 'v':
305				copychar = '\v';
306				break;
307			default:
308				if (isspacech(*src))
309					copychar = *src;
310				else
311					errx(1, "Invalid sequence '\\%c' in -S",
312					    *src);
313			}
314			break;
315		default:
316			if ((in_dq || in_sq) && in_arg)
317				copychar = *src;
318			else if (isspacech(*src))
319				found_sep = 1;
320			else {
321				/*
322				 * If the first character of a new argument
323				 * is `#', then ignore the remaining chars.
324				 */
325				if (!in_arg && *src == '#')
326					goto str_done;
327				copychar = *src;
328			}
329		}
330		/*
331		 * Now that the switch has determined what (if anything)
332		 * needs to be copied, copy whatever that is to *dest.
333		 */
334		if (copychar || copystr != NULL) {
335			if (!in_arg) {
336				/* This is the first byte of a new argument */
337				*nextarg++ = dest;
338				addcount++;
339				in_arg = 1;
340			}
341			if (copychar)
342				*dest++ = (char)copychar;
343			else if (copystr != NULL)
344				while (*copystr != '\0')
345					*dest++ = *copystr++;
346		} else if (found_sep) {
347			*dest++ = '\0';
348			while (isspacech(*src))
349				src++;
350			--src;
351			in_arg = 0;
352		}
353	}
354str_done:
355	*dest = '\0';
356	*nextarg = NULL;
357	if (in_dq || in_sq) {
358		errx(1, "No terminating quote for string: %.*s%s",
359		    bq_destlen, *(nextarg - 1), bq_src);
360	}
361	if (env_verbosity > 1) {
362		fprintf(stderr, "#env  split -S:\t'%s'\n", str);
363		oldarg = newargv + 1;
364		fprintf(stderr, "#env      into:\t'%s'\n", *oldarg);
365		for (oldarg++; *oldarg; oldarg++)
366			fprintf(stderr, "#env          &\t'%s'\n", *oldarg);
367	}
368
369	/* Copy the unprocessed arg-pointers from the original array */
370	for (oldarg = *origv + *origind; *oldarg; oldarg++)
371		*nextarg++ = *oldarg;
372	*nextarg = NULL;
373
374	/* Update optind/argc/argv in the calling routine */
375	*origind = 1;
376	*origc += addcount;
377	*origv = newargv;
378}
379
380/**
381 * Routine to split expand any environment variables referenced in the string
382 * that -S is processing.  For now it only supports the form ${VARNAME}.  It
383 * explicitly does not support $VARNAME, and obviously can not handle special
384 * shell-variables such as $?, $*, $1, etc.  It is called with *src_p pointing
385 * at the initial '$', and if successful it will update *src_p, *dest_p, and
386 * possibly *thisarg_p in the calling routine.
387 */
388static const char *
389expand_vars(int in_thisarg, char **thisarg_p, char **dest_p, const char **src_p)
390{
391	const char *vbegin, *vend, *vvalue;
392	char *newstr, *vname;
393	int bad_reference;
394	size_t namelen, newlen;
395
396	bad_reference = 1;
397	vbegin = vend = (*src_p) + 1;
398	if (*vbegin++ == '{')
399		if (*vbegin == '_' || isalphach(*vbegin)) {
400			vend = vbegin + 1;
401			while (*vend == '_' || isalnumch(*vend))
402				vend++;
403			if (*vend == '}')
404				bad_reference = 0;
405		}
406	if (bad_reference)
407		errx(1, "Only ${VARNAME} expansion is supported, error at: %s",
408		    *src_p);
409
410	/*
411	 * We now know we have a valid environment variable name, so update
412	 * the caller's source-pointer to the last character in that reference,
413	 * and then pick up the matching value.  If the variable is not found,
414	 * or if it has a null value, then our work here is done.
415	 */
416	*src_p = vend;
417	namelen = vend - vbegin + 1;
418	vname = malloc(namelen);
419	strlcpy(vname, vbegin, namelen);
420	vvalue = getenv(vname);
421	if (vvalue == NULL || *vvalue == '\0') {
422		if (env_verbosity > 2)
423			fprintf(stderr,
424			    "#env  replacing ${%s} with null string\n",
425			    vname);
426		free(vname);
427		return (NULL);
428	}
429
430	if (env_verbosity > 2)
431		fprintf(stderr, "#env  expanding ${%s} into '%s'\n", vname,
432		    vvalue);
433
434	/*
435	 * There is some value to copy to the destination.  If the value is
436	 * shorter than the ${VARNAME} reference that it replaces, then our
437	 * caller can just copy the value to the existing destination.
438	 */
439	if (strlen(vname) + 3 >= strlen(vvalue)) {
440		free(vname);
441		return (vvalue);
442	}
443
444	/*
445	 * The value is longer than the string it replaces, which means the
446	 * present destination area is too small to hold it.  Create a new
447	 * destination area, and update the caller's 'dest' variable to match.
448	 * If the caller has already started copying some info for 'thisarg'
449	 * into the present destination, then the new destination area must
450	 * include a copy of that data, and the pointer to 'thisarg' must also
451	 * be updated.  Note that it is still the caller which copies this
452	 * vvalue to the new *dest.
453	 */
454	newlen = strlen(vvalue) + strlen(*src_p) + 1;
455	if (in_thisarg) {
456		**dest_p = '\0';	/* Provide terminator for 'thisarg' */
457		newlen += strlen(*thisarg_p);
458		newstr = malloc(newlen);
459		strcpy(newstr, *thisarg_p);
460		*thisarg_p = newstr;
461	} else {
462		newstr = malloc(newlen);
463		*newstr = '\0';
464	}
465	*dest_p = strchr(newstr, '\0');
466	free(vname);
467	return (vvalue);
468}
469