1/* POSIX.2 wordexp implementation.
2 Copyright (C) 1997-2019 Free Software Foundation, Inc.
3 This file is part of the GNU C Library.
4 Contributed by Tim Waugh <tim@cyberelk.demon.co.uk>.
5
6 The GNU C Library is free software; you can redistribute it and/or
7 modify it under the terms of the GNU Lesser General Public
8 License as published by the Free Software Foundation; either
9 version 2.1 of the License, or (at your option) any later version.
10
11 The GNU C Library is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 Lesser General Public License for more details.
15
16 You should have received a copy of the GNU Lesser General Public
17 License along with the GNU C Library; if not, see
18 <http://www.gnu.org/licenses/>. */
19
20#include <ctype.h>
21#include <errno.h>
22#include <fcntl.h>
23#include <fnmatch.h>
24#include <glob.h>
25#include <libintl.h>
26#include <paths.h>
27#include <pwd.h>
28#include <signal.h>
29#include <stdbool.h>
30#include <stdio.h>
31#include <stdlib.h>
32#include <string.h>
33#include <sys/param.h>
34#include <sys/stat.h>
35#include <sys/time.h>
36#include <sys/types.h>
37#include <sys/types.h>
38#include <sys/wait.h>
39#include <unistd.h>
40#include <wchar.h>
41#include <wordexp.h>
42#include <kernel-features.h>
43#include <scratch_buffer.h>
44
45#include <libc-lock.h>
46#include <_itoa.h>
47
48/* Undefine the following line for the production version. */
49/* #define NDEBUG 1 */
50#include <assert.h>
51
52/* Get some device information. */
53#include <device-nrs.h>
54
55/*
56 * This is a recursive-descent-style word expansion routine.
57 */
58
59/* These variables are defined and initialized in the startup code. */
60extern int __libc_argc attribute_hidden;
61extern char **__libc_argv attribute_hidden;
62
63/* Some forward declarations */
64static int parse_dollars (char **word, size_t *word_length, size_t *max_length,
65 const char *words, size_t *offset, int flags,
66 wordexp_t *pwordexp, const char *ifs,
67 const char *ifs_white, int quoted);
68static int parse_backtick (char **word, size_t *word_length,
69 size_t *max_length, const char *words,
70 size_t *offset, int flags, wordexp_t *pwordexp,
71 const char *ifs, const char *ifs_white);
72static int parse_dquote (char **word, size_t *word_length, size_t *max_length,
73 const char *words, size_t *offset, int flags,
74 wordexp_t *pwordexp, const char *ifs,
75 const char *ifs_white);
76static int eval_expr (char *expr, long int *result);
77
78/* The w_*() functions manipulate word lists. */
79
80#define W_CHUNK (100)
81
82/* Result of w_newword will be ignored if it's the last word. */
83static inline char *
84w_newword (size_t *actlen, size_t *maxlen)
85{
86 *actlen = *maxlen = 0;
87 return NULL;
88}
89
90static char *
91w_addchar (char *buffer, size_t *actlen, size_t *maxlen, char ch)
92 /* (lengths exclude trailing zero) */
93{
94 /* Add a character to the buffer, allocating room for it if needed. */
95
96 if (*actlen == *maxlen)
97 {
98 char *old_buffer = buffer;
99 assert (buffer == NULL || *maxlen != 0);
100 *maxlen += W_CHUNK;
101 buffer = (char *) realloc (buffer, 1 + *maxlen);
102
103 if (buffer == NULL)
104 free (old_buffer);
105 }
106
107 if (buffer != NULL)
108 {
109 buffer[*actlen] = ch;
110 buffer[++(*actlen)] = '\0';
111 }
112
113 return buffer;
114}
115
116static char *
117w_addmem (char *buffer, size_t *actlen, size_t *maxlen, const char *str,
118 size_t len)
119{
120 /* Add a string to the buffer, allocating room for it if needed.
121 */
122 if (*actlen + len > *maxlen)
123 {
124 char *old_buffer = buffer;
125 assert (buffer == NULL || *maxlen != 0);
126 *maxlen += MAX (2 * len, W_CHUNK);
127 buffer = realloc (old_buffer, 1 + *maxlen);
128
129 if (buffer == NULL)
130 free (old_buffer);
131 }
132
133 if (buffer != NULL)
134 {
135 *((char *) __mempcpy (&buffer[*actlen], str, len)) = '\0';
136 *actlen += len;
137 }
138
139 return buffer;
140}
141
142static char *
143w_addstr (char *buffer, size_t *actlen, size_t *maxlen, const char *str)
144 /* (lengths exclude trailing zero) */
145{
146 /* Add a string to the buffer, allocating room for it if needed.
147 */
148 size_t len;
149
150 assert (str != NULL); /* w_addstr only called from this file */
151 len = strlen (str);
152
153 return w_addmem (buffer, actlen, maxlen, str, len);
154}
155
156static int
157w_addword (wordexp_t *pwordexp, char *word)
158{
159 /* Add a word to the wordlist */
160 size_t num_p;
161 char **new_wordv;
162 bool allocated = false;
163
164 /* Internally, NULL acts like "". Convert NULLs to "" before
165 * the caller sees them.
166 */
167 if (word == NULL)
168 {
169 word = __strdup ("");
170 if (word == NULL)
171 goto no_space;
172 allocated = true;
173 }
174
175 num_p = 2 + pwordexp->we_wordc + pwordexp->we_offs;
176 new_wordv = realloc (pwordexp->we_wordv, sizeof (char *) * num_p);
177 if (new_wordv != NULL)
178 {
179 pwordexp->we_wordv = new_wordv;
180 pwordexp->we_wordv[pwordexp->we_offs + pwordexp->we_wordc++] = word;
181 pwordexp->we_wordv[pwordexp->we_offs + pwordexp->we_wordc] = NULL;
182 return 0;
183 }
184
185 if (allocated)
186 free (word);
187
188no_space:
189 return WRDE_NOSPACE;
190}
191
192/* The parse_*() functions should leave *offset being the offset in 'words'
193 * to the last character processed.
194 */
195
196static int
197parse_backslash (char **word, size_t *word_length, size_t *max_length,
198 const char *words, size_t *offset)
199{
200 /* We are poised _at_ a backslash, not in quotes */
201
202 switch (words[1 + *offset])
203 {
204 case 0:
205 /* Backslash is last character of input words */
206 return WRDE_SYNTAX;
207
208 case '\n':
209 ++(*offset);
210 break;
211
212 default:
213 *word = w_addchar (*word, word_length, max_length, words[1 + *offset]);
214 if (*word == NULL)
215 return WRDE_NOSPACE;
216
217 ++(*offset);
218 break;
219 }
220
221 return 0;
222}
223
224static int
225parse_qtd_backslash (char **word, size_t *word_length, size_t *max_length,
226 const char *words, size_t *offset)
227{
228 /* We are poised _at_ a backslash, inside quotes */
229
230 switch (words[1 + *offset])
231 {
232 case 0:
233 /* Backslash is last character of input words */
234 return WRDE_SYNTAX;
235
236 case '\n':
237 ++(*offset);
238 break;
239
240 case '$':
241 case '`':
242 case '"':
243 case '\\':
244 *word = w_addchar (*word, word_length, max_length, words[1 + *offset]);
245 if (*word == NULL)
246 return WRDE_NOSPACE;
247
248 ++(*offset);
249 break;
250
251 default:
252 *word = w_addchar (*word, word_length, max_length, words[*offset]);
253 if (*word != NULL)
254 *word = w_addchar (*word, word_length, max_length, words[1 + *offset]);
255
256 if (*word == NULL)
257 return WRDE_NOSPACE;
258
259 ++(*offset);
260 break;
261 }
262
263 return 0;
264}
265
266static int
267parse_tilde (char **word, size_t *word_length, size_t *max_length,
268 const char *words, size_t *offset, size_t wordc)
269{
270 /* We are poised _at_ a tilde */
271 size_t i;
272
273 if (*word_length != 0)
274 {
275 if (!((*word)[*word_length - 1] == '=' && wordc == 0))
276 {
277 if (!((*word)[*word_length - 1] == ':'
278 && strchr (*word, '=') && wordc == 0))
279 {
280 *word = w_addchar (*word, word_length, max_length, '~');
281 return *word ? 0 : WRDE_NOSPACE;
282 }
283 }
284 }
285
286 for (i = 1 + *offset; words[i]; i++)
287 {
288 if (words[i] == ':' || words[i] == '/' || words[i] == ' ' ||
289 words[i] == '\t' || words[i] == 0 )
290 break;
291
292 if (words[i] == '\\')
293 {
294 *word = w_addchar (*word, word_length, max_length, '~');
295 return *word ? 0 : WRDE_NOSPACE;
296 }
297 }
298
299 if (i == 1 + *offset)
300 {
301 /* Tilde appears on its own */
302 char* home;
303
304 /* POSIX.2 says ~ expands to $HOME and if HOME is unset the
305 results are unspecified. We do a lookup on the uid if
306 HOME is unset. */
307
308 home = getenv ("HOME");
309 if (home != NULL)
310 {
311 *word = w_addstr (*word, word_length, max_length, home);
312 if (*word == NULL)
313 return WRDE_NOSPACE;
314 }
315 else
316 {
317 struct passwd pwd, *tpwd;
318 uid_t uid = __getuid ();
319 int result;
320 struct scratch_buffer tmpbuf;
321 scratch_buffer_init (&tmpbuf);
322
323 while ((result = __getpwuid_r (uid, &pwd,
324 tmpbuf.data, tmpbuf.length,
325 &tpwd)) != 0
326 && errno == ERANGE)
327 if (!scratch_buffer_grow (&tmpbuf))
328 return WRDE_NOSPACE;
329
330 if (result == 0 && tpwd != NULL && pwd.pw_dir != NULL)
331 {
332 *word = w_addstr (*word, word_length, max_length, pwd.pw_dir);
333 if (*word == NULL)
334 {
335 scratch_buffer_free (&tmpbuf);
336 return WRDE_NOSPACE;
337 }
338 }
339 else
340 {
341 *word = w_addchar (*word, word_length, max_length, '~');
342 if (*word == NULL)
343 {
344 scratch_buffer_free (&tmpbuf);
345 return WRDE_NOSPACE;
346 }
347 }
348 scratch_buffer_free (&tmpbuf);
349 }
350 }
351 else
352 {
353 /* Look up user name in database to get home directory */
354 char *user = strndupa (&words[1 + *offset], i - (1 + *offset));
355 struct passwd pwd, *tpwd;
356 int result;
357 struct scratch_buffer tmpbuf;
358 scratch_buffer_init (&tmpbuf);
359
360 while ((result = __getpwnam_r (user, &pwd, tmpbuf.data, tmpbuf.length,
361 &tpwd)) != 0
362 && errno == ERANGE)
363 if (!scratch_buffer_grow (&tmpbuf))
364 return WRDE_NOSPACE;
365
366 if (result == 0 && tpwd != NULL && pwd.pw_dir)
367 *word = w_addstr (*word, word_length, max_length, pwd.pw_dir);
368 else
369 {
370 /* (invalid login name) */
371 *word = w_addchar (*word, word_length, max_length, '~');
372 if (*word != NULL)
373 *word = w_addstr (*word, word_length, max_length, user);
374 }
375
376 scratch_buffer_free (&tmpbuf);
377
378 *offset = i - 1;
379 }
380 return *word ? 0 : WRDE_NOSPACE;
381}
382
383
384static int
385do_parse_glob (const char *glob_word, char **word, size_t *word_length,
386 size_t *max_length, wordexp_t *pwordexp, const char *ifs,
387 const char *ifs_white)
388{
389 int error;
390 unsigned int match;
391 glob_t globbuf;
392
393 error = glob (glob_word, GLOB_NOCHECK, NULL, &globbuf);
394
395 if (error != 0)
396 {
397 /* We can only run into memory problems. */
398 assert (error == GLOB_NOSPACE);
399 return WRDE_NOSPACE;
400 }
401
402 if (ifs && !*ifs)
403 {
404 /* No field splitting allowed. */
405 assert (globbuf.gl_pathv[0] != NULL);
406 *word = w_addstr (*word, word_length, max_length, globbuf.gl_pathv[0]);
407 for (match = 1; match < globbuf.gl_pathc && *word != NULL; ++match)
408 {
409 *word = w_addchar (*word, word_length, max_length, ' ');
410 if (*word != NULL)
411 *word = w_addstr (*word, word_length, max_length,
412 globbuf.gl_pathv[match]);
413 }
414
415 globfree (&globbuf);
416 return *word ? 0 : WRDE_NOSPACE;
417 }
418
419 assert (ifs == NULL || *ifs != '\0');
420 if (*word != NULL)
421 {
422 free (*word);
423 *word = w_newword (word_length, max_length);
424 }
425
426 for (match = 0; match < globbuf.gl_pathc; ++match)
427 {
428 char *matching_word = __strdup (globbuf.gl_pathv[match]);
429 if (matching_word == NULL || w_addword (pwordexp, matching_word))
430 {
431 globfree (&globbuf);
432 return WRDE_NOSPACE;
433 }
434 }
435
436 globfree (&globbuf);
437 return 0;
438}
439
440static int
441parse_glob (char **word, size_t *word_length, size_t *max_length,
442 const char *words, size_t *offset, int flags,
443 wordexp_t *pwordexp, const char *ifs, const char *ifs_white)
444{
445 /* We are poised just after a '*', a '[' or a '?'. */
446 int error = WRDE_NOSPACE;
447 int quoted = 0; /* 1 if singly-quoted, 2 if doubly */
448 size_t i;
449 wordexp_t glob_list; /* List of words to glob */
450
451 glob_list.we_wordc = 0;
452 glob_list.we_wordv = NULL;
453 glob_list.we_offs = 0;
454 for (; words[*offset] != '\0'; ++*offset)
455 {
456 if (strchr (ifs, words[*offset]) != NULL)
457 /* Reached IFS */
458 break;
459
460 /* Sort out quoting */
461 if (words[*offset] == '\'')
462 {
463 if (quoted == 0)
464 {
465 quoted = 1;
466 continue;
467 }
468 else if (quoted == 1)
469 {
470 quoted = 0;
471 continue;
472 }
473 }
474 else if (words[*offset] == '"')
475 {
476 if (quoted == 0)
477 {
478 quoted = 2;
479 continue;
480 }
481 else if (quoted == 2)
482 {
483 quoted = 0;
484 continue;
485 }
486 }
487
488 /* Sort out other special characters */
489 if (quoted != 1 && words[*offset] == '$')
490 {
491 error = parse_dollars (word, word_length, max_length, words,
492 offset, flags, &glob_list, ifs, ifs_white,
493 quoted == 2);
494 if (error)
495 goto tidy_up;
496
497 continue;
498 }
499 else if (words[*offset] == '\\')
500 {
501 if (quoted)
502 error = parse_qtd_backslash (word, word_length, max_length,
503 words, offset);
504 else
505 error = parse_backslash (word, word_length, max_length,
506 words, offset);
507
508 if (error)
509 goto tidy_up;
510
511 continue;
512 }
513
514 *word = w_addchar (*word, word_length, max_length, words[*offset]);
515 if (*word == NULL)
516 goto tidy_up;
517 }
518
519 /* Don't forget to re-parse the character we stopped at. */
520 --*offset;
521
522 /* Glob the words */
523 error = w_addword (&glob_list, *word);
524 *word = w_newword (word_length, max_length);
525 for (i = 0; error == 0 && i < glob_list.we_wordc; i++)
526 error = do_parse_glob (glob_list.we_wordv[i], word, word_length,
527 max_length, pwordexp, ifs, ifs_white);
528
529 /* Now tidy up */
530tidy_up:
531 wordfree (&glob_list);
532 return error;
533}
534
535static int
536parse_squote (char **word, size_t *word_length, size_t *max_length,
537 const char *words, size_t *offset)
538{
539 /* We are poised just after a single quote */
540 for (; words[*offset]; ++(*offset))
541 {
542 if (words[*offset] != '\'')
543 {
544 *word = w_addchar (*word, word_length, max_length, words[*offset]);
545 if (*word == NULL)
546 return WRDE_NOSPACE;
547 }
548 else return 0;
549 }
550
551 /* Unterminated string */
552 return WRDE_SYNTAX;
553}
554
555/* Functions to evaluate an arithmetic expression */
556static int
557eval_expr_val (char **expr, long int *result)
558{
559 char *digit;
560
561 /* Skip white space */
562 for (digit = *expr; digit && *digit && isspace (*digit); ++digit);
563
564 if (*digit == '(')
565 {
566 /* Scan for closing paren */
567 for (++digit; **expr && **expr != ')'; ++(*expr));
568
569 /* Is there one? */
570 if (!**expr)
571 return WRDE_SYNTAX;
572
573 *(*expr)++ = 0;
574
575 if (eval_expr (digit, result))
576 return WRDE_SYNTAX;
577
578 return 0;
579 }
580
581 /* POSIX requires that decimal, octal, and hexadecimal constants are
582 recognized. Therefore we pass 0 as the third parameter to strtol. */
583 *result = strtol (digit, expr, 0);
584 if (digit == *expr)
585 return WRDE_SYNTAX;
586
587 return 0;
588}
589
590static int
591eval_expr_multdiv (char **expr, long int *result)
592{
593 long int arg;
594
595 /* Read a Value */
596 if (eval_expr_val (expr, result) != 0)
597 return WRDE_SYNTAX;
598
599 while (**expr)
600 {
601 /* Skip white space */
602 for (; *expr && **expr && isspace (**expr); ++(*expr));
603
604 if (**expr == '*')
605 {
606 ++(*expr);
607 if (eval_expr_val (expr, &arg) != 0)
608 return WRDE_SYNTAX;
609
610 *result *= arg;
611 }
612 else if (**expr == '/')
613 {
614 ++(*expr);
615 if (eval_expr_val (expr, &arg) != 0)
616 return WRDE_SYNTAX;
617
618 /* Division by zero or integer overflow. */
619 if (arg == 0 || (arg == -1 && *result == LONG_MIN))
620 return WRDE_SYNTAX;
621
622 *result /= arg;
623 }
624 else break;
625 }
626
627 return 0;
628}
629
630static int
631eval_expr (char *expr, long int *result)
632{
633 long int arg;
634
635 /* Read a Multdiv */
636 if (eval_expr_multdiv (&expr, result) != 0)
637 return WRDE_SYNTAX;
638
639 while (*expr)
640 {
641 /* Skip white space */
642 for (; expr && *expr && isspace (*expr); ++expr);
643
644 if (*expr == '+')
645 {
646 ++expr;
647 if (eval_expr_multdiv (&expr, &arg) != 0)
648 return WRDE_SYNTAX;
649
650 *result += arg;
651 }
652 else if (*expr == '-')
653 {
654 ++expr;
655 if (eval_expr_multdiv (&expr, &arg) != 0)
656 return WRDE_SYNTAX;
657
658 *result -= arg;
659 }
660 else break;
661 }
662
663 return 0;
664}
665
666static int
667parse_arith (char **word, size_t *word_length, size_t *max_length,
668 const char *words, size_t *offset, int flags, int bracket)
669{
670 /* We are poised just after "$((" or "$[" */
671 int error;
672 int paren_depth = 1;
673 size_t expr_length;
674 size_t expr_maxlen;
675 char *expr;
676
677 expr = w_newword (&expr_length, &expr_maxlen);
678 for (; words[*offset]; ++(*offset))
679 {
680 switch (words[*offset])
681 {
682 case '$':
683 error = parse_dollars (&expr, &expr_length, &expr_maxlen,
684 words, offset, flags, NULL, NULL, NULL, 1);
685 /* The ``1'' here is to tell parse_dollars not to
686 * split the fields.
687 */
688 if (error)
689 {
690 free (expr);
691 return error;
692 }
693 break;
694
695 case '`':
696 (*offset)++;
697 error = parse_backtick (&expr, &expr_length, &expr_maxlen,
698 words, offset, flags, NULL, NULL, NULL);
699 /* The first NULL here is to tell parse_backtick not to
700 * split the fields.
701 */
702 if (error)
703 {
704 free (expr);
705 return error;
706 }
707 break;
708
709 case '\\':
710 error = parse_qtd_backslash (&expr, &expr_length, &expr_maxlen,
711 words, offset);
712 if (error)
713 {
714 free (expr);
715 return error;
716 }
717 /* I think that a backslash within an
718 * arithmetic expansion is bound to
719 * cause an error sooner or later anyway though.
720 */
721 break;
722
723 case ')':
724 if (--paren_depth == 0)
725 {
726 char result[21]; /* 21 = ceil(log10(2^64)) + 1 */
727 long int numresult = 0;
728 long long int convertme;
729
730 if (bracket || words[1 + *offset] != ')')
731 {
732 free (expr);
733 return WRDE_SYNTAX;
734 }
735
736 ++(*offset);
737
738 /* Go - evaluate. */
739 if (*expr && eval_expr (expr, &numresult) != 0)
740 {
741 free (expr);
742 return WRDE_SYNTAX;
743 }
744
745 if (numresult < 0)
746 {
747 convertme = -numresult;
748 *word = w_addchar (*word, word_length, max_length, '-');
749 if (!*word)
750 {
751 free (expr);
752 return WRDE_NOSPACE;
753 }
754 }
755 else
756 convertme = numresult;
757
758 result[20] = '\0';
759 *word = w_addstr (*word, word_length, max_length,
760 _itoa (convertme, &result[20], 10, 0));
761 free (expr);
762 return *word ? 0 : WRDE_NOSPACE;
763 }
764 expr = w_addchar (expr, &expr_length, &expr_maxlen, words[*offset]);
765 if (expr == NULL)
766 return WRDE_NOSPACE;
767
768 break;
769
770 case ']':
771 if (bracket && paren_depth == 1)
772 {
773 char result[21]; /* 21 = ceil(log10(2^64)) + 1 */
774 long int numresult = 0;
775
776 /* Go - evaluate. */
777 if (*expr && eval_expr (expr, &numresult) != 0)
778 {
779 free (expr);
780 return WRDE_SYNTAX;
781 }
782
783 result[20] = '\0';
784 *word = w_addstr (*word, word_length, max_length,
785 _itoa_word (numresult, &result[20], 10, 0));
786 free (expr);
787 return *word ? 0 : WRDE_NOSPACE;
788 }
789
790 free (expr);
791 return WRDE_SYNTAX;
792
793 case '\n':
794 case ';':
795 case '{':
796 case '}':
797 free (expr);
798 return WRDE_BADCHAR;
799
800 case '(':
801 ++paren_depth;
802 default:
803 expr = w_addchar (expr, &expr_length, &expr_maxlen, words[*offset]);
804 if (expr == NULL)
805 return WRDE_NOSPACE;
806 }
807 }
808
809 /* Premature end */
810 free (expr);
811 return WRDE_SYNTAX;
812}
813
814/* Function called by child process in exec_comm() */
815static inline void
816__attribute__ ((always_inline))
817exec_comm_child (char *comm, int *fildes, int showerr, int noexec)
818{
819 const char *args[4] = { _PATH_BSHELL, "-c", comm, NULL };
820
821 /* Execute the command, or just check syntax? */
822 if (noexec)
823 args[1] = "-nc";
824
825 /* Redirect output. */
826 if (__glibc_likely (fildes[1] != STDOUT_FILENO))
827 {
828 __dup2 (fildes[1], STDOUT_FILENO);
829 __close (fildes[1]);
830 }
831 else
832 /* Reset the close-on-exec flag (if necessary). */
833 __fcntl (fildes[1], F_SETFD, 0);
834
835 /* Redirect stderr to /dev/null if we have to. */
836 if (showerr == 0)
837 {
838 struct stat64 st;
839 int fd;
840 __close (STDERR_FILENO);
841 fd = __open (_PATH_DEVNULL, O_WRONLY);
842 if (fd >= 0 && fd != STDERR_FILENO)
843 {
844 __dup2 (fd, STDERR_FILENO);
845 __close (fd);
846 }
847 /* Be paranoid. Check that we actually opened the /dev/null
848 device. */
849 if (__builtin_expect (__fxstat64 (_STAT_VER, STDERR_FILENO, &st), 0) != 0
850 || __builtin_expect (S_ISCHR (st.st_mode), 1) == 0
851#if defined DEV_NULL_MAJOR && defined DEV_NULL_MINOR
852 || st.st_rdev != __gnu_dev_makedev (DEV_NULL_MAJOR, DEV_NULL_MINOR)
853#endif
854 )
855 /* It's not the /dev/null device. Stop right here. The
856 problem is: how do we stop? We use _exit() with an
857 hopefully unusual exit code. */
858 _exit (90);
859 }
860
861 /* Make sure the subshell doesn't field-split on our behalf. */
862 __unsetenv ("IFS");
863
864 __close (fildes[0]);
865 __execve (_PATH_BSHELL, (char *const *) args, __environ);
866
867 /* Bad. What now? */
868 abort ();
869}
870
871/* Function to execute a command and retrieve the results */
872/* pwordexp contains NULL if field-splitting is forbidden */
873static int
874exec_comm (char *comm, char **word, size_t *word_length, size_t *max_length,
875 int flags, wordexp_t *pwordexp, const char *ifs,
876 const char *ifs_white)
877{
878 int fildes[2];
879#define bufsize 128
880 int buflen;
881 int i;
882 int status = 0;
883 size_t maxnewlines = 0;
884 char buffer[bufsize];
885 pid_t pid;
886 int noexec = 0;
887
888 /* Do nothing if command substitution should not succeed. */
889 if (flags & WRDE_NOCMD)
890 return WRDE_CMDSUB;
891
892 /* Don't fork() unless necessary */
893 if (!comm || !*comm)
894 return 0;
895
896 if (__pipe2 (fildes, O_CLOEXEC) < 0)
897 return WRDE_NOSPACE;
898
899 again:
900 if ((pid = __fork ()) < 0)
901 {
902 /* Bad */
903 __close (fildes[0]);
904 __close (fildes[1]);
905 return WRDE_NOSPACE;
906 }
907
908 if (pid == 0)
909 exec_comm_child (comm, fildes, noexec ? 0 : flags & WRDE_SHOWERR, noexec);
910
911 /* Parent */
912
913 /* If we are just testing the syntax, only wait. */
914 if (noexec)
915 return (TEMP_FAILURE_RETRY (__waitpid (pid, &status, 0)) == pid
916 && status != 0) ? WRDE_SYNTAX : 0;
917
918 __close (fildes[1]);
919 fildes[1] = -1;
920
921 if (!pwordexp)
922 /* Quoted - no field splitting */
923 {
924 while (1)
925 {
926 if ((buflen = TEMP_FAILURE_RETRY (__read (fildes[0], buffer,
927 bufsize))) < 1)
928 {
929 /* If read returned 0 then the process has closed its
930 stdout. Don't use WNOHANG in that case to avoid busy
931 looping until the process eventually exits. */
932 if (TEMP_FAILURE_RETRY (__waitpid (pid, &status,
933 buflen == 0 ? 0 : WNOHANG))
934 == 0)
935 continue;
936 if ((buflen = TEMP_FAILURE_RETRY (__read (fildes[0], buffer,
937 bufsize))) < 1)
938 break;
939 }
940
941 maxnewlines += buflen;
942
943 *word = w_addmem (*word, word_length, max_length, buffer, buflen);
944 if (*word == NULL)
945 goto no_space;
946 }
947 }
948 else
949 /* Not quoted - split fields */
950 {
951 int copying = 0;
952 /* 'copying' is:
953 * 0 when searching for first character in a field not IFS white space
954 * 1 when copying the text of a field
955 * 2 when searching for possible non-whitespace IFS
956 * 3 when searching for non-newline after copying field
957 */
958
959 while (1)
960 {
961 if ((buflen = TEMP_FAILURE_RETRY (__read (fildes[0], buffer,
962 bufsize))) < 1)
963 {
964 /* If read returned 0 then the process has closed its
965 stdout. Don't use WNOHANG in that case to avoid busy
966 looping until the process eventually exits. */
967 if (TEMP_FAILURE_RETRY (__waitpid (pid, &status,
968 buflen == 0 ? 0 : WNOHANG))
969 == 0)
970 continue;
971 if ((buflen = TEMP_FAILURE_RETRY (__read (fildes[0], buffer,
972 bufsize))) < 1)
973 break;
974 }
975
976 for (i = 0; i < buflen; ++i)
977 {
978 if (strchr (ifs, buffer[i]) != NULL)
979 {
980 /* Current character is IFS */
981 if (strchr (ifs_white, buffer[i]) == NULL)
982 {
983 /* Current character is IFS but not whitespace */
984 if (copying == 2)
985 {
986 /* current character
987 * |
988 * V
989 * eg: text<space><comma><space>moretext
990 *
991 * So, strip whitespace IFS (like at the start)
992 */
993 copying = 0;
994 continue;
995 }
996
997 copying = 0;
998 /* fall through and delimit field.. */
999 }
1000 else
1001 {
1002 if (buffer[i] == '\n')
1003 {
1004 /* Current character is (IFS) newline */
1005
1006 /* If copying a field, this is the end of it,
1007 but maybe all that's left is trailing newlines.
1008 So start searching for a non-newline. */
1009 if (copying == 1)
1010 copying = 3;
1011
1012 continue;
1013 }
1014 else
1015 {
1016 /* Current character is IFS white space, but
1017 not a newline */
1018
1019 /* If not either copying a field or searching
1020 for non-newline after a field, ignore it */
1021 if (copying != 1 && copying != 3)
1022 continue;
1023
1024 /* End of field (search for non-ws IFS afterwards) */
1025 copying = 2;
1026 }
1027 }
1028
1029 /* First IFS white space (non-newline), or IFS non-whitespace.
1030 * Delimit the field. Nulls are converted by w_addword. */
1031 if (w_addword (pwordexp, *word) == WRDE_NOSPACE)
1032 goto no_space;
1033
1034 *word = w_newword (word_length, max_length);
1035
1036 maxnewlines = 0;
1037 /* fall back round the loop.. */
1038 }
1039 else
1040 {
1041 /* Not IFS character */
1042
1043 if (copying == 3)
1044 {
1045 /* Nothing but (IFS) newlines since the last field,
1046 so delimit it here before starting new word */
1047 if (w_addword (pwordexp, *word) == WRDE_NOSPACE)
1048 goto no_space;
1049
1050 *word = w_newword (word_length, max_length);
1051 }
1052
1053 copying = 1;
1054
1055 if (buffer[i] == '\n') /* happens if newline not in IFS */
1056 maxnewlines++;
1057 else
1058 maxnewlines = 0;
1059
1060 *word = w_addchar (*word, word_length, max_length,
1061 buffer[i]);
1062 if (*word == NULL)
1063 goto no_space;
1064 }
1065 }
1066 }
1067 }
1068
1069 /* Chop off trailing newlines (required by POSIX.2) */
1070 /* Ensure we don't go back further than the beginning of the
1071 substitution (i.e. remove maxnewlines bytes at most) */
1072 while (maxnewlines-- != 0 &&
1073 *word_length > 0 && (*word)[*word_length - 1] == '\n')
1074 {
1075 (*word)[--*word_length] = '\0';
1076
1077 /* If the last word was entirely newlines, turn it into a new word
1078 * which can be ignored if there's nothing following it. */
1079 if (*word_length == 0)
1080 {
1081 free (*word);
1082 *word = w_newword (word_length, max_length);
1083 break;
1084 }
1085 }
1086
1087 __close (fildes[0]);
1088 fildes[0] = -1;
1089
1090 /* Check for syntax error (re-execute but with "-n" flag) */
1091 if (buflen < 1 && status != 0)
1092 {
1093 noexec = 1;
1094 goto again;
1095 }
1096
1097 return 0;
1098
1099no_space:
1100 __kill (pid, SIGKILL);
1101 TEMP_FAILURE_RETRY (__waitpid (pid, NULL, 0));
1102 __close (fildes[0]);
1103 return WRDE_NOSPACE;
1104}
1105
1106static int
1107parse_comm (char **word, size_t *word_length, size_t *max_length,
1108 const char *words, size_t *offset, int flags, wordexp_t *pwordexp,
1109 const char *ifs, const char *ifs_white)
1110{
1111 /* We are poised just after "$(" */
1112 int paren_depth = 1;
1113 int error = 0;
1114 int quoted = 0; /* 1 for singly-quoted, 2 for doubly-quoted */
1115 size_t comm_length;
1116 size_t comm_maxlen;
1117 char *comm = w_newword (&comm_length, &comm_maxlen);
1118
1119 for (; words[*offset]; ++(*offset))
1120 {
1121 switch (words[*offset])
1122 {
1123 case '\'':
1124 if (quoted == 0)
1125 quoted = 1;
1126 else if (quoted == 1)
1127 quoted = 0;
1128
1129 break;
1130
1131 case '"':
1132 if (quoted == 0)
1133 quoted = 2;
1134 else if (quoted == 2)
1135 quoted = 0;
1136
1137 break;
1138
1139 case ')':
1140 if (!quoted && --paren_depth == 0)
1141 {
1142 /* Go -- give script to the shell */
1143 if (comm)
1144 {
1145#ifdef __libc_ptf_call
1146 /* We do not want the exec_comm call to be cut short
1147 by a thread cancellation since cleanup is very
1148 ugly. Therefore disable cancellation for
1149 now. */
1150 // XXX Ideally we do want the thread being cancelable.
1151 // XXX If demand is there we'll change it.
1152 int state = PTHREAD_CANCEL_ENABLE;
1153 __libc_ptf_call (__pthread_setcancelstate,
1154 (PTHREAD_CANCEL_DISABLE, &state), 0);
1155#endif
1156
1157 error = exec_comm (comm, word, word_length, max_length,
1158 flags, pwordexp, ifs, ifs_white);
1159
1160#ifdef __libc_ptf_call
1161 __libc_ptf_call (__pthread_setcancelstate,
1162 (state, NULL), 0);
1163#endif
1164
1165 free (comm);
1166 }
1167
1168 return error;
1169 }
1170
1171 /* This is just part of the script */
1172 break;
1173
1174 case '(':
1175 if (!quoted)
1176 ++paren_depth;
1177 }
1178
1179 comm = w_addchar (comm, &comm_length, &comm_maxlen, words[*offset]);
1180 if (comm == NULL)
1181 return WRDE_NOSPACE;
1182 }
1183
1184 /* Premature end. */
1185 free (comm);
1186
1187 return WRDE_SYNTAX;
1188}
1189
1190#define CHAR_IN_SET(ch, char_set) \
1191 (memchr (char_set "", ch, sizeof (char_set) - 1) != NULL)
1192
1193static int
1194parse_param (char **word, size_t *word_length, size_t *max_length,
1195 const char *words, size_t *offset, int flags, wordexp_t *pwordexp,
1196 const char *ifs, const char *ifs_white, int quoted)
1197{
1198 /* We are poised just after "$" */
1199 enum action
1200 {
1201 ACT_NONE,
1202 ACT_RP_SHORT_LEFT = '#',
1203 ACT_RP_LONG_LEFT = 'L',
1204 ACT_RP_SHORT_RIGHT = '%',
1205 ACT_RP_LONG_RIGHT = 'R',
1206 ACT_NULL_ERROR = '?',
1207 ACT_NULL_SUBST = '-',
1208 ACT_NONNULL_SUBST = '+',
1209 ACT_NULL_ASSIGN = '='
1210 };
1211 size_t env_length;
1212 size_t env_maxlen;
1213 size_t pat_length;
1214 size_t pat_maxlen;
1215 size_t start = *offset;
1216 char *env;
1217 char *pattern;
1218 char *value = NULL;
1219 enum action action = ACT_NONE;
1220 int depth = 0;
1221 int colon_seen = 0;
1222 int seen_hash = 0;
1223 int free_value = 0;
1224 int pattern_is_quoted = 0; /* 1 for singly-quoted, 2 for doubly-quoted */
1225 int error;
1226 int special = 0;
1227 char buffer[21];
1228 int brace = words[*offset] == '{';
1229
1230 env = w_newword (&env_length, &env_maxlen);
1231 pattern = w_newword (&pat_length, &pat_maxlen);
1232
1233 if (brace)
1234 ++*offset;
1235
1236 /* First collect the parameter name. */
1237
1238 if (words[*offset] == '#')
1239 {
1240 seen_hash = 1;
1241 if (!brace)
1242 goto envsubst;
1243 ++*offset;
1244 }
1245
1246 if (isalpha (words[*offset]) || words[*offset] == '_')
1247 {
1248 /* Normal parameter name. */
1249 do
1250 {
1251 env = w_addchar (env, &env_length, &env_maxlen,
1252 words[*offset]);
1253 if (env == NULL)
1254 goto no_space;
1255 }
1256 while (isalnum (words[++*offset]) || words[*offset] == '_');
1257 }
1258 else if (isdigit (words[*offset]))
1259 {
1260 /* Numeric parameter name. */
1261 special = 1;
1262 do
1263 {
1264 env = w_addchar (env, &env_length, &env_maxlen,
1265 words[*offset]);
1266 if (env == NULL)
1267 goto no_space;
1268 if (!brace)
1269 goto envsubst;
1270 }
1271 while (isdigit(words[++*offset]));
1272 }
1273 else if (CHAR_IN_SET (words[*offset], "*@$"))
1274 {
1275 /* Special parameter. */
1276 special = 1;
1277 env = w_addchar (env, &env_length, &env_maxlen,
1278 words[*offset]);
1279 if (env == NULL)
1280 goto no_space;
1281 ++*offset;
1282 }
1283 else
1284 {
1285 if (brace)
1286 goto syntax;
1287 }
1288
1289 if (brace)
1290 {
1291 /* Check for special action to be applied to the value. */
1292 switch (words[*offset])
1293 {
1294 case '}':
1295 /* Evaluate. */
1296 goto envsubst;
1297
1298 case '#':
1299 action = ACT_RP_SHORT_LEFT;
1300 if (words[1 + *offset] == '#')
1301 {
1302 ++*offset;
1303 action = ACT_RP_LONG_LEFT;
1304 }
1305 break;
1306
1307 case '%':
1308 action = ACT_RP_SHORT_RIGHT;
1309 if (words[1 + *offset] == '%')
1310 {
1311 ++*offset;
1312 action = ACT_RP_LONG_RIGHT;
1313 }
1314 break;
1315
1316 case ':':
1317 if (!CHAR_IN_SET (words[1 + *offset], "-=?+"))
1318 goto syntax;
1319
1320 colon_seen = 1;
1321 action = words[++*offset];
1322 break;
1323
1324 case '-':
1325 case '=':
1326 case '?':
1327 case '+':
1328 action = words[*offset];
1329 break;
1330
1331 default:
1332 goto syntax;
1333 }
1334
1335 /* Now collect the pattern, but don't expand it yet. */
1336 ++*offset;
1337 for (; words[*offset]; ++(*offset))
1338 {
1339 switch (words[*offset])
1340 {
1341 case '{':
1342 if (!pattern_is_quoted)
1343 ++depth;
1344 break;
1345
1346 case '}':
1347 if (!pattern_is_quoted)
1348 {
1349 if (depth == 0)
1350 goto envsubst;
1351 --depth;
1352 }
1353 break;
1354
1355 case '\\':
1356 if (pattern_is_quoted)
1357 /* Quoted; treat as normal character. */
1358 break;
1359
1360 /* Otherwise, it's an escape: next character is literal. */
1361 if (words[++*offset] == '\0')
1362 goto syntax;
1363
1364 pattern = w_addchar (pattern, &pat_length, &pat_maxlen, '\\');
1365 if (pattern == NULL)
1366 goto no_space;
1367
1368 break;
1369
1370 case '\'':
1371 if (pattern_is_quoted == 0)
1372 pattern_is_quoted = 1;
1373 else if (pattern_is_quoted == 1)
1374 pattern_is_quoted = 0;
1375
1376 break;
1377
1378 case '"':
1379 if (pattern_is_quoted == 0)
1380 pattern_is_quoted = 2;
1381 else if (pattern_is_quoted == 2)
1382 pattern_is_quoted = 0;
1383
1384 break;
1385 }
1386
1387 pattern = w_addchar (pattern, &pat_length, &pat_maxlen,
1388 words[*offset]);
1389 if (pattern == NULL)
1390 goto no_space;
1391 }
1392 }
1393
1394 /* End of input string -- remember to reparse the character that we
1395 * stopped at. */
1396 --(*offset);
1397
1398envsubst:
1399 if (words[start] == '{' && words[*offset] != '}')
1400 goto syntax;
1401
1402 if (env == NULL)
1403 {
1404 if (seen_hash)
1405 {
1406 /* $# expands to the number of positional parameters */
1407 buffer[20] = '\0';
1408 value = _itoa_word (__libc_argc - 1, &buffer[20], 10, 0);
1409 seen_hash = 0;
1410 }
1411 else
1412 {
1413 /* Just $ on its own */
1414 *offset = start - 1;
1415 *word = w_addchar (*word, word_length, max_length, '$');
1416 return *word ? 0 : WRDE_NOSPACE;
1417 }
1418 }
1419 /* Is it a numeric parameter? */
1420 else if (isdigit (env[0]))
1421 {
1422 int n = atoi (env);
1423
1424 if (n >= __libc_argc)
1425 /* Substitute NULL. */
1426 value = NULL;
1427 else
1428 /* Replace with appropriate positional parameter. */
1429 value = __libc_argv[n];
1430 }
1431 /* Is it a special parameter? */
1432 else if (special)
1433 {
1434 /* Is it `$$'? */
1435 if (*env == '$')
1436 {
1437 buffer[20] = '\0';
1438 value = _itoa_word (__getpid (), &buffer[20], 10, 0);
1439 }
1440 /* Is it `${#*}' or `${#@}'? */
1441 else if ((*env == '*' || *env == '@') && seen_hash)
1442 {
1443 buffer[20] = '\0';
1444 value = _itoa_word (__libc_argc > 0 ? __libc_argc - 1 : 0,
1445 &buffer[20], 10, 0);
1446 *word = w_addstr (*word, word_length, max_length, value);
1447 free (env);
1448 free (pattern);
1449 return *word ? 0 : WRDE_NOSPACE;
1450 }
1451 /* Is it `$*' or `$@' (unquoted) ? */
1452 else if (*env == '*' || (*env == '@' && !quoted))
1453 {
1454 size_t plist_len = 0;
1455 int p;
1456 char *end;
1457
1458 /* Build up value parameter by parameter (copy them) */
1459 for (p = 1; __libc_argv[p]; ++p)
1460 plist_len += strlen (__libc_argv[p]) + 1; /* for space */
1461 value = malloc (plist_len);
1462 if (value == NULL)
1463 goto no_space;
1464 end = value;
1465 *end = 0;
1466 for (p = 1; __libc_argv[p]; ++p)
1467 {
1468 if (p > 1)
1469 *end++ = ' ';
1470 end = __stpcpy (end, __libc_argv[p]);
1471 }
1472
1473 free_value = 1;
1474 }
1475 else
1476 {
1477 /* Must be a quoted `$@' */
1478 assert (*env == '@' && quoted);
1479
1480 /* Each parameter is a separate word ("$@") */
1481 if (__libc_argc == 2)
1482 value = __libc_argv[1];
1483 else if (__libc_argc > 2)
1484 {
1485 int p;
1486
1487 /* Append first parameter to current word. */
1488 value = w_addstr (*word, word_length, max_length,
1489 __libc_argv[1]);
1490 if (value == NULL || w_addword (pwordexp, value))
1491 goto no_space;
1492
1493 for (p = 2; __libc_argv[p + 1]; p++)
1494 {
1495 char *newword = __strdup (__libc_argv[p]);
1496 if (newword == NULL || w_addword (pwordexp, newword))
1497 goto no_space;
1498 }
1499
1500 /* Start a new word with the last parameter. */
1501 *word = w_newword (word_length, max_length);
1502 value = __libc_argv[p];
1503 }
1504 else
1505 {
1506 free (env);
1507 free (pattern);
1508 return 0;
1509 }
1510 }
1511 }
1512 else
1513 value = getenv (env);
1514
1515 if (value == NULL && (flags & WRDE_UNDEF))
1516 {
1517 /* Variable not defined. */
1518 error = WRDE_BADVAL;
1519 goto do_error;
1520 }
1521
1522 if (action != ACT_NONE)
1523 {
1524 int expand_pattern = 0;
1525
1526 /* First, find out if we need to expand pattern (i.e. if we will
1527 * use it). */
1528 switch (action)
1529 {
1530 case ACT_RP_SHORT_LEFT:
1531 case ACT_RP_LONG_LEFT:
1532 case ACT_RP_SHORT_RIGHT:
1533 case ACT_RP_LONG_RIGHT:
1534 /* Always expand for these. */
1535 expand_pattern = 1;
1536 break;
1537
1538 case ACT_NULL_ERROR:
1539 case ACT_NULL_SUBST:
1540 case ACT_NULL_ASSIGN:
1541 if (!value || (!*value && colon_seen))
1542 /* If param is unset, or set but null and a colon has been seen,
1543 the expansion of the pattern will be needed. */
1544 expand_pattern = 1;
1545
1546 break;
1547
1548 case ACT_NONNULL_SUBST:
1549 /* Expansion of word will be needed if parameter is set and not null,
1550 or set null but no colon has been seen. */
1551 if (value && (*value || !colon_seen))
1552 expand_pattern = 1;
1553
1554 break;
1555
1556 default:
1557 assert (! "Unrecognised action!");
1558 }
1559
1560 if (expand_pattern)
1561 {
1562 /* We need to perform tilde expansion, parameter expansion,
1563 command substitution, and arithmetic expansion. We also
1564 have to be a bit careful with wildcard characters, as
1565 pattern might be given to fnmatch soon. To do this, we
1566 convert quotes to escapes. */
1567
1568 char *expanded;
1569 size_t exp_len;
1570 size_t exp_maxl;
1571 char *p;
1572 int quoted = 0; /* 1: single quotes; 2: double */
1573
1574 expanded = w_newword (&exp_len, &exp_maxl);
1575 for (p = pattern; p && *p; p++)
1576 {
1577 size_t offset;
1578
1579 switch (*p)
1580 {
1581 case '"':
1582 if (quoted == 2)
1583 quoted = 0;
1584 else if (quoted == 0)
1585 quoted = 2;
1586 else break;
1587
1588 continue;
1589
1590 case '\'':
1591 if (quoted == 1)
1592 quoted = 0;
1593 else if (quoted == 0)
1594 quoted = 1;
1595 else break;
1596
1597 continue;
1598
1599 case '*':
1600 case '?':
1601 if (quoted)
1602 {
1603 /* Convert quoted wildchar to escaped wildchar. */
1604 expanded = w_addchar (expanded, &exp_len,
1605 &exp_maxl, '\\');
1606
1607 if (expanded == NULL)
1608 goto no_space;
1609 }
1610 break;
1611
1612 case '$':
1613 offset = 0;
1614 error = parse_dollars (&expanded, &exp_len, &exp_maxl, p,
1615 &offset, flags, NULL, NULL, NULL, 1);
1616 if (error)
1617 {
1618 if (free_value)
1619 free (value);
1620
1621 free (expanded);
1622
1623 goto do_error;
1624 }
1625
1626 p += offset;
1627 continue;
1628
1629 case '~':
1630 if (quoted || exp_len)
1631 break;
1632
1633 offset = 0;
1634 error = parse_tilde (&expanded, &exp_len, &exp_maxl, p,
1635 &offset, 0);
1636 if (error)
1637 {
1638 if (free_value)
1639 free (value);
1640
1641 free (expanded);
1642
1643 goto do_error;
1644 }
1645
1646 p += offset;
1647 continue;
1648
1649 case '\\':
1650 expanded = w_addchar (expanded, &exp_len, &exp_maxl, '\\');
1651 ++p;
1652 assert (*p); /* checked when extracted initially */
1653 if (expanded == NULL)
1654 goto no_space;
1655 }
1656
1657 expanded = w_addchar (expanded, &exp_len, &exp_maxl, *p);
1658
1659 if (expanded == NULL)
1660 goto no_space;
1661 }
1662
1663 free (pattern);
1664
1665 pattern = expanded;
1666 }
1667
1668 switch (action)
1669 {
1670 case ACT_RP_SHORT_LEFT:
1671 case ACT_RP_LONG_LEFT:
1672 case ACT_RP_SHORT_RIGHT:
1673 case ACT_RP_LONG_RIGHT:
1674 {
1675 char *p;
1676 char c;
1677 char *end;
1678
1679 if (value == NULL || pattern == NULL || *pattern == '\0')
1680 break;
1681
1682 end = value + strlen (value);
1683
1684 switch (action)
1685 {
1686 case ACT_RP_SHORT_LEFT:
1687 for (p = value; p <= end; ++p)
1688 {
1689 c = *p;
1690 *p = '\0';
1691 if (fnmatch (pattern, value, 0) != FNM_NOMATCH)
1692 {
1693 *p = c;
1694 if (free_value)
1695 {
1696 char *newval = __strdup (p);
1697 if (newval == NULL)
1698 {
1699 free (value);
1700 goto no_space;
1701 }
1702 free (value);
1703 value = newval;
1704 }
1705 else
1706 value = p;
1707 break;
1708 }
1709 *p = c;
1710 }
1711
1712 break;
1713
1714 case ACT_RP_LONG_LEFT:
1715 for (p = end; p >= value; --p)
1716 {
1717 c = *p;
1718 *p = '\0';
1719 if (fnmatch (pattern, value, 0) != FNM_NOMATCH)
1720 {
1721 *p = c;
1722 if (free_value)
1723 {
1724 char *newval = __strdup (p);
1725 if (newval == NULL)
1726 {
1727 free (value);
1728 goto no_space;
1729 }
1730 free (value);
1731 value = newval;
1732 }
1733 else
1734 value = p;
1735 break;
1736 }
1737 *p = c;
1738 }
1739
1740 break;
1741
1742 case ACT_RP_SHORT_RIGHT:
1743 for (p = end; p >= value; --p)
1744 {
1745 if (fnmatch (pattern, p, 0) != FNM_NOMATCH)
1746 {
1747 char *newval;
1748 newval = malloc (p - value + 1);
1749
1750 if (newval == NULL)
1751 {
1752 if (free_value)
1753 free (value);
1754 goto no_space;
1755 }
1756
1757 *(char *) __mempcpy (newval, value, p - value) = '\0';
1758 if (free_value)
1759 free (value);
1760 value = newval;
1761 free_value = 1;
1762 break;
1763 }
1764 }
1765
1766 break;
1767
1768 case ACT_RP_LONG_RIGHT:
1769 for (p = value; p <= end; ++p)
1770 {
1771 if (fnmatch (pattern, p, 0) != FNM_NOMATCH)
1772 {
1773 char *newval;
1774 newval = malloc (p - value + 1);
1775
1776 if (newval == NULL)
1777 {
1778 if (free_value)
1779 free (value);
1780 goto no_space;
1781 }
1782
1783 *(char *) __mempcpy (newval, value, p - value) = '\0';
1784 if (free_value)
1785 free (value);
1786 value = newval;
1787 free_value = 1;
1788 break;
1789 }
1790 }
1791
1792 break;
1793
1794 default:
1795 break;
1796 }
1797
1798 break;
1799 }
1800
1801 case ACT_NULL_ERROR:
1802 if (value && *value)
1803 /* Substitute parameter */
1804 break;
1805
1806 error = 0;
1807 if (!colon_seen && value)
1808 /* Substitute NULL */
1809 ;
1810 else
1811 {
1812 const char *str = pattern;
1813
1814 if (str[0] == '\0')
1815 str = _("parameter null or not set");
1816
1817 __fxprintf (NULL, "%s: %s\n", env, str);
1818 }
1819
1820 if (free_value)
1821 free (value);
1822 goto do_error;
1823
1824 case ACT_NULL_SUBST:
1825 if (value && *value)
1826 /* Substitute parameter */
1827 break;
1828
1829 if (free_value)
1830 free (value);
1831
1832 if (!colon_seen && value)
1833 /* Substitute NULL */
1834 goto success;
1835
1836 value = pattern ? __strdup (pattern) : pattern;
1837 free_value = 1;
1838
1839 if (pattern && !value)
1840 goto no_space;
1841
1842 break;
1843
1844 case ACT_NONNULL_SUBST:
1845 if (value && (*value || !colon_seen))
1846 {
1847 if (free_value)
1848 free (value);
1849
1850 value = pattern ? __strdup (pattern) : pattern;
1851 free_value = 1;
1852
1853 if (pattern && !value)
1854 goto no_space;
1855
1856 break;
1857 }
1858
1859 /* Substitute NULL */
1860 if (free_value)
1861 free (value);
1862 goto success;
1863
1864 case ACT_NULL_ASSIGN:
1865 if (value && *value)
1866 /* Substitute parameter */
1867 break;
1868
1869 if (!colon_seen && value)
1870 {
1871 /* Substitute NULL */
1872 if (free_value)
1873 free (value);
1874 goto success;
1875 }
1876
1877 if (free_value)
1878 free (value);
1879
1880 value = pattern ? __strdup (pattern) : pattern;
1881 free_value = 1;
1882
1883 if (pattern && !value)
1884 goto no_space;
1885
1886 __setenv (env, value ?: "", 1);
1887 break;
1888
1889 default:
1890 assert (! "Unrecognised action!");
1891 }
1892 }
1893
1894 free (env);
1895 env = NULL;
1896 free (pattern);
1897 pattern = NULL;
1898
1899 if (seen_hash)
1900 {
1901 char param_length[21];
1902 param_length[20] = '\0';
1903 *word = w_addstr (*word, word_length, max_length,
1904 _itoa_word (value ? strlen (value) : 0,
1905 &param_length[20], 10, 0));
1906 if (free_value)
1907 {
1908 assert (value != NULL);
1909 free (value);
1910 }
1911
1912 return *word ? 0 : WRDE_NOSPACE;
1913 }
1914
1915 if (value == NULL)
1916 return 0;
1917
1918 if (quoted || !pwordexp)
1919 {
1920 /* Quoted - no field split */
1921 *word = w_addstr (*word, word_length, max_length, value);
1922 if (free_value)
1923 free (value);
1924
1925 return *word ? 0 : WRDE_NOSPACE;
1926 }
1927 else
1928 {
1929 /* Need to field-split */
1930 char *value_copy = __strdup (value); /* Don't modify value */
1931 char *field_begin = value_copy;
1932 int seen_nonws_ifs = 0;
1933
1934 if (free_value)
1935 free (value);
1936
1937 if (value_copy == NULL)
1938 goto no_space;
1939
1940 do
1941 {
1942 char *field_end = field_begin;
1943 char *next_field;
1944
1945 /* If this isn't the first field, start a new word */
1946 if (field_begin != value_copy)
1947 {
1948 if (w_addword (pwordexp, *word) == WRDE_NOSPACE)
1949 {
1950 free (value_copy);
1951 goto no_space;
1952 }
1953
1954 *word = w_newword (word_length, max_length);
1955 }
1956
1957 /* Skip IFS whitespace before the field */
1958 field_begin += strspn (field_begin, ifs_white);
1959
1960 if (!seen_nonws_ifs && *field_begin == 0)
1961 /* Nothing but whitespace */
1962 break;
1963
1964 /* Search for the end of the field */
1965 field_end = field_begin + strcspn (field_begin, ifs);
1966
1967 /* Set up pointer to the character after end of field and
1968 skip whitespace IFS after it. */
1969 next_field = field_end + strspn (field_end, ifs_white);
1970
1971 /* Skip at most one non-whitespace IFS character after the field */
1972 seen_nonws_ifs = 0;
1973 if (*next_field && strchr (ifs, *next_field))
1974 {
1975 seen_nonws_ifs = 1;
1976 next_field++;
1977 }
1978
1979 /* Null-terminate it */
1980 *field_end = 0;
1981
1982 /* Tag a copy onto the current word */
1983 *word = w_addstr (*word, word_length, max_length, field_begin);
1984
1985 if (*word == NULL && *field_begin != '\0')
1986 {
1987 free (value_copy);
1988 goto no_space;
1989 }
1990
1991 field_begin = next_field;
1992 }
1993 while (seen_nonws_ifs || *field_begin);
1994
1995 free (value_copy);
1996 }
1997
1998 return 0;
1999
2000success:
2001 error = 0;
2002 goto do_error;
2003
2004no_space:
2005 error = WRDE_NOSPACE;
2006 goto do_error;
2007
2008syntax:
2009 error = WRDE_SYNTAX;
2010
2011do_error:
2012 free (env);
2013
2014 free (pattern);
2015
2016 return error;
2017}
2018
2019#undef CHAR_IN_SET
2020
2021static int
2022parse_dollars (char **word, size_t *word_length, size_t *max_length,
2023 const char *words, size_t *offset, int flags,
2024 wordexp_t *pwordexp, const char *ifs, const char *ifs_white,
2025 int quoted)
2026{
2027 /* We are poised _at_ "$" */
2028 switch (words[1 + *offset])
2029 {
2030 case '"':
2031 case '\'':
2032 case 0:
2033 *word = w_addchar (*word, word_length, max_length, '$');
2034 return *word ? 0 : WRDE_NOSPACE;
2035
2036 case '(':
2037 if (words[2 + *offset] == '(')
2038 {
2039 /* Differentiate between $((1+3)) and $((echo);(ls)) */
2040 int i = 3 + *offset;
2041 int depth = 0;
2042 while (words[i] && !(depth == 0 && words[i] == ')'))
2043 {
2044 if (words[i] == '(')
2045 ++depth;
2046 else if (words[i] == ')')
2047 --depth;
2048
2049 ++i;
2050 }
2051
2052 if (words[i] == ')' && words[i + 1] == ')')
2053 {
2054 (*offset) += 3;
2055 /* Call parse_arith -- 0 is for "no brackets" */
2056 return parse_arith (word, word_length, max_length, words, offset,
2057 flags, 0);
2058 }
2059 }
2060
2061 (*offset) += 2;
2062 return parse_comm (word, word_length, max_length, words, offset, flags,
2063 quoted? NULL : pwordexp, ifs, ifs_white);
2064
2065 case '[':
2066 (*offset) += 2;
2067 /* Call parse_arith -- 1 is for "brackets" */
2068 return parse_arith (word, word_length, max_length, words, offset, flags,
2069 1);
2070
2071 case '{':
2072 default:
2073 ++(*offset); /* parse_param needs to know if "{" is there */
2074 return parse_param (word, word_length, max_length, words, offset, flags,
2075 pwordexp, ifs, ifs_white, quoted);
2076 }
2077}
2078
2079static int
2080parse_backtick (char **word, size_t *word_length, size_t *max_length,
2081 const char *words, size_t *offset, int flags,
2082 wordexp_t *pwordexp, const char *ifs, const char *ifs_white)
2083{
2084 /* We are poised just after "`" */
2085 int error;
2086 int squoting = 0;
2087 size_t comm_length;
2088 size_t comm_maxlen;
2089 char *comm = w_newword (&comm_length, &comm_maxlen);
2090
2091 for (; words[*offset]; ++(*offset))
2092 {
2093 switch (words[*offset])
2094 {
2095 case '`':
2096 /* Go -- give the script to the shell */
2097 error = exec_comm (comm, word, word_length, max_length, flags,
2098 pwordexp, ifs, ifs_white);
2099 free (comm);
2100 return error;
2101
2102 case '\\':
2103 if (squoting)
2104 {
2105 error = parse_qtd_backslash (&comm, &comm_length, &comm_maxlen,
2106 words, offset);
2107
2108 if (error)
2109 {
2110 free (comm);
2111 return error;
2112 }
2113
2114 break;
2115 }
2116
2117 error = parse_backslash (&comm, &comm_length, &comm_maxlen, words,
2118 offset);
2119
2120 if (error)
2121 {
2122 free (comm);
2123 return error;
2124 }
2125
2126 break;
2127
2128 case '\'':
2129 squoting = 1 - squoting;
2130 default:
2131 comm = w_addchar (comm, &comm_length, &comm_maxlen, words[*offset]);
2132 if (comm == NULL)
2133 return WRDE_NOSPACE;
2134 }
2135 }
2136
2137 /* Premature end */
2138 free (comm);
2139 return WRDE_SYNTAX;
2140}
2141
2142static int
2143parse_dquote (char **word, size_t *word_length, size_t *max_length,
2144 const char *words, size_t *offset, int flags,
2145 wordexp_t *pwordexp, const char * ifs, const char * ifs_white)
2146{
2147 /* We are poised just after a double-quote */
2148 int error;
2149
2150 for (; words[*offset]; ++(*offset))
2151 {
2152 switch (words[*offset])
2153 {
2154 case '"':
2155 return 0;
2156
2157 case '$':
2158 error = parse_dollars (word, word_length, max_length, words, offset,
2159 flags, pwordexp, ifs, ifs_white, 1);
2160 /* The ``1'' here is to tell parse_dollars not to
2161 * split the fields. It may need to, however ("$@").
2162 */
2163 if (error)
2164 return error;
2165
2166 break;
2167
2168 case '`':
2169 ++(*offset);
2170 error = parse_backtick (word, word_length, max_length, words,
2171 offset, flags, NULL, NULL, NULL);
2172 /* The first NULL here is to tell parse_backtick not to
2173 * split the fields.
2174 */
2175 if (error)
2176 return error;
2177
2178 break;
2179
2180 case '\\':
2181 error = parse_qtd_backslash (word, word_length, max_length, words,
2182 offset);
2183
2184 if (error)
2185 return error;
2186
2187 break;
2188
2189 default:
2190 *word = w_addchar (*word, word_length, max_length, words[*offset]);
2191 if (*word == NULL)
2192 return WRDE_NOSPACE;
2193 }
2194 }
2195
2196 /* Unterminated string */
2197 return WRDE_SYNTAX;
2198}
2199
2200/*
2201 * wordfree() is to be called after pwordexp is finished with.
2202 */
2203
2204void
2205wordfree (wordexp_t *pwordexp)
2206{
2207
2208 /* wordexp can set pwordexp to NULL */
2209 if (pwordexp && pwordexp->we_wordv)
2210 {
2211 char **wordv = pwordexp->we_wordv;
2212
2213 for (wordv += pwordexp->we_offs; *wordv; ++wordv)
2214 free (*wordv);
2215
2216 free (pwordexp->we_wordv);
2217 pwordexp->we_wordv = NULL;
2218 }
2219}
2220libc_hidden_def (wordfree)
2221
2222/*
2223 * wordexp()
2224 */
2225
2226int
2227wordexp (const char *words, wordexp_t *pwordexp, int flags)
2228{
2229 size_t words_offset;
2230 size_t word_length;
2231 size_t max_length;
2232 char *word = w_newword (&word_length, &max_length);
2233 int error;
2234 char *ifs;
2235 char ifs_white[4];
2236 wordexp_t old_word = *pwordexp;
2237
2238 if (flags & WRDE_REUSE)
2239 {
2240 /* Minimal implementation of WRDE_REUSE for now */
2241 wordfree (pwordexp);
2242 old_word.we_wordv = NULL;
2243 }
2244
2245 if ((flags & WRDE_APPEND) == 0)
2246 {
2247 pwordexp->we_wordc = 0;
2248
2249 if (flags & WRDE_DOOFFS)
2250 {
2251 pwordexp->we_wordv = calloc (1 + pwordexp->we_offs, sizeof (char *));
2252 if (pwordexp->we_wordv == NULL)
2253 {
2254 error = WRDE_NOSPACE;
2255 goto do_error;
2256 }
2257 }
2258 else
2259 {
2260 pwordexp->we_wordv = calloc (1, sizeof (char *));
2261 if (pwordexp->we_wordv == NULL)
2262 {
2263 error = WRDE_NOSPACE;
2264 goto do_error;
2265 }
2266
2267 pwordexp->we_offs = 0;
2268 }
2269 }
2270
2271 /* Find out what the field separators are.
2272 * There are two types: whitespace and non-whitespace.
2273 */
2274 ifs = getenv ("IFS");
2275
2276 if (ifs == NULL)
2277 /* IFS unset - use <space><tab><newline>. */
2278 ifs = strcpy (ifs_white, " \t\n");
2279 else
2280 {
2281 char *ifsch = ifs;
2282 char *whch = ifs_white;
2283
2284 while (*ifsch != '\0')
2285 {
2286 if (*ifsch == ' ' || *ifsch == '\t' || *ifsch == '\n')
2287 {
2288 /* Whitespace IFS. See first whether it is already in our
2289 collection. */
2290 char *runp = ifs_white;
2291
2292 while (runp < whch && *runp != *ifsch)
2293 ++runp;
2294
2295 if (runp == whch)
2296 *whch++ = *ifsch;
2297 }
2298
2299 ++ifsch;
2300 }
2301 *whch = '\0';
2302 }
2303
2304 for (words_offset = 0 ; words[words_offset] ; ++words_offset)
2305 switch (words[words_offset])
2306 {
2307 case '\\':
2308 error = parse_backslash (&word, &word_length, &max_length, words,
2309 &words_offset);
2310
2311 if (error)
2312 goto do_error;
2313
2314 break;
2315
2316 case '$':
2317 error = parse_dollars (&word, &word_length, &max_length, words,
2318 &words_offset, flags, pwordexp, ifs, ifs_white,
2319 0);
2320
2321 if (error)
2322 goto do_error;
2323
2324 break;
2325
2326 case '`':
2327 ++words_offset;
2328 error = parse_backtick (&word, &word_length, &max_length, words,
2329 &words_offset, flags, pwordexp, ifs,
2330 ifs_white);
2331
2332 if (error)
2333 goto do_error;
2334
2335 break;
2336
2337 case '"':
2338 ++words_offset;
2339 error = parse_dquote (&word, &word_length, &max_length, words,
2340 &words_offset, flags, pwordexp, ifs, ifs_white);
2341
2342 if (error)
2343 goto do_error;
2344
2345 if (!word_length)
2346 {
2347 error = w_addword (pwordexp, NULL);
2348
2349 if (error)
2350 return error;
2351 }
2352
2353 break;
2354
2355 case '\'':
2356 ++words_offset;
2357 error = parse_squote (&word, &word_length, &max_length, words,
2358 &words_offset);
2359
2360 if (error)
2361 goto do_error;
2362
2363 if (!word_length)
2364 {
2365 error = w_addword (pwordexp, NULL);
2366
2367 if (error)
2368 return error;
2369 }
2370
2371 break;
2372
2373 case '~':
2374 error = parse_tilde (&word, &word_length, &max_length, words,
2375 &words_offset, pwordexp->we_wordc);
2376
2377 if (error)
2378 goto do_error;
2379
2380 break;
2381
2382 case '*':
2383 case '[':
2384 case '?':
2385 error = parse_glob (&word, &word_length, &max_length, words,
2386 &words_offset, flags, pwordexp, ifs, ifs_white);
2387
2388 if (error)
2389 goto do_error;
2390
2391 break;
2392
2393 default:
2394 /* Is it a word separator? */
2395 if (strchr (" \t", words[words_offset]) == NULL)
2396 {
2397 char ch = words[words_offset];
2398
2399 /* Not a word separator -- but is it a valid word char? */
2400 if (strchr ("\n|&;<>(){}", ch))
2401 {
2402 /* Fail */
2403 error = WRDE_BADCHAR;
2404 goto do_error;
2405 }
2406
2407 /* "Ordinary" character -- add it to word */
2408 word = w_addchar (word, &word_length, &max_length,
2409 ch);
2410 if (word == NULL)
2411 {
2412 error = WRDE_NOSPACE;
2413 goto do_error;
2414 }
2415
2416 break;
2417 }
2418
2419 /* If a word has been delimited, add it to the list. */
2420 if (word != NULL)
2421 {
2422 error = w_addword (pwordexp, word);
2423 if (error)
2424 goto do_error;
2425 }
2426
2427 word = w_newword (&word_length, &max_length);
2428 }
2429
2430 /* End of string */
2431
2432 /* There was a word separator at the end */
2433 if (word == NULL) /* i.e. w_newword */
2434 return 0;
2435
2436 /* There was no field separator at the end */
2437 return w_addword (pwordexp, word);
2438
2439do_error:
2440 /* Error:
2441 * free memory used (unless error is WRDE_NOSPACE), and
2442 * set pwordexp members back to what they were.
2443 */
2444
2445 free (word);
2446
2447 if (error == WRDE_NOSPACE)
2448 return WRDE_NOSPACE;
2449
2450 if ((flags & WRDE_APPEND) == 0)
2451 wordfree (pwordexp);
2452
2453 *pwordexp = old_word;
2454 return error;
2455}
2456