2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 1992 Keith Muller.
5 * Copyright (c) 1992, 1993
6 * The Regents of the University of California. All rights reserved.
8 * This code is derived from software contributed to Berkeley by
9 * Keith Muller of the University of California, San Diego.
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
14 * 1. Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 * notice, this list of conditions and the following disclaimer in the
18 * documentation and/or other materials provided with the distribution.
19 * 3. Neither the name of the University nor the names of its contributors
20 * may be used to endorse or promote products derived from this software
21 * without specific prior written permission.
23 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36 #include <sys/types.h>
47 * routines to handle pattern matching, name modification (regular expression
48 * substitution and interactive renames), and destination name modification for
49 * copy (-rw). Both file name and link names are adjusted as required in these
53 #define MAXSUBEXP 10 /* max subexpressions, DO NOT CHANGE */
54 static PATTERN *pathead = NULL; /* file pattern match list head */
55 static PATTERN *pattail = NULL; /* file pattern match list tail */
56 static REPLACE *rephead = NULL; /* replacement string list head */
57 static REPLACE *reptail = NULL; /* replacement string list tail */
59 static int rep_name(char *, int *, int);
60 static int tty_rename(ARCHD *);
61 static int fix_path(char *, int *, char *, int);
62 static int fn_match(char *, char *, char **);
63 static char * range_match(char *, int);
64 static int resub(regex_t *, regmatch_t *, char *, char *, char *, char *);
68 * parses the -s replacement string; compiles the regular expression
69 * and stores the compiled value and it's replacement string together in
70 * replacement string list. Input to this function is of the form:
72 * The first char in the string specifies the delimiter used by this
73 * replacement string. "Old" is a regular expression in "ed" format which
74 * is compiled by regcomp() and is applied to filenames. "new" is the
75 * substitution string; p and g are options flags for printing and global
76 * replacement (over the single filename)
78 * 0 if a proper replacement string and regular expression was added to
79 * the list of replacement patterns; -1 otherwise.
92 * throw out the bad parameters
94 if ((str == NULL) || (*str == '\0')) {
95 paxwarn(1, "Empty replacement string");
100 * first character in the string specifies what the delimiter is for
103 if ((pt1 = strchr(str+1, *str)) == NULL) {
104 paxwarn(1, "Invalid replacement string %s", str);
109 * allocate space for the node that handles this replacement pattern
110 * and split out the regular expression and try to compile it
112 if ((rep = (REPLACE *)malloc(sizeof(REPLACE))) == NULL) {
113 paxwarn(1, "Unable to allocate memory for replacement string");
118 if ((res = regcomp(&(rep->rcmp), str+1, 0)) != 0) {
119 regerror(res, &(rep->rcmp), rebuf, sizeof(rebuf));
120 paxwarn(1, "%s while compiling regular expression %s", rebuf, str);
126 * put the delimiter back in case we need an error message and
127 * locate the delimiter at the end of the replacement string
128 * we then point the node at the new substitution string
131 if ((pt2 = strchr(pt1, *str)) == NULL) {
134 paxwarn(1, "Invalid replacement string %s", str);
144 * set the options if any
146 while (*pt2 != '\0') {
160 paxwarn(1, "Invalid replacement string option %s", str);
167 * all done, link it in at the end
170 if (rephead == NULL) {
171 reptail = rephead = rep;
181 * add a pattern match to the pattern match list. Pattern matches are used
182 * to select which archive members are extracted. (They appear as
183 * arguments to pax in the list and read modes). If no patterns are
184 * supplied to pax, all members in the archive will be selected (and the
185 * pattern match list is empty).
187 * 0 if the pattern was added to the list, -1 otherwise
191 pat_add(char *str, char *chdnam)
198 if ((str == NULL) || (*str == '\0')) {
199 paxwarn(1, "Empty pattern string");
204 * allocate space for the pattern and store the pattern. the pattern is
205 * part of argv so do not bother to copy it, just point at it. Add the
206 * node to the end of the pattern list
208 if ((pt = (PATTERN *)malloc(sizeof(PATTERN))) == NULL) {
209 paxwarn(1, "Unable to allocate memory for pattern string");
215 pt->plen = strlen(str);
218 pt->chdname = chdnam;
220 if (pathead == NULL) {
221 pattail = pathead = pt;
231 * complain if any the user supplied pattern did not result in a match to
232 * a selected archive member.
242 * walk down the list checking the flags to make sure MTCH was set,
245 for (pt = pathead; pt != NULL; pt = pt->fow) {
249 paxwarn(1, "WARNING! These patterns were not matched:");
252 (void)fprintf(stderr, "%s\n", pt->pstr);
258 * the archive member which matches a pattern was selected. Mark the
259 * pattern as having selected an archive member. arcn->pat points at the
260 * pattern that was matched. arcn->pat is set in pat_match()
262 * NOTE: When the -c option is used, we are called when there was no match
263 * by pat_match() (that means we did match before the inverted sense of
264 * the logic). Now this seems really strange at first, but with -c we
265 * need to keep track of those patterns that cause an archive member to NOT
266 * be selected (it found an archive member with a specified pattern)
268 * 0 if the pattern pointed at by arcn->pat was tagged as creating a
269 * match, -1 otherwise.
280 * if no patterns just return
282 if ((pathead == NULL) || ((pt = arcn->pat) == NULL))
286 * when we are NOT limited to a single match per pattern mark the
295 * we reach this point only when we allow a single selected match per
296 * pattern, if the pattern matches a directory and we do not have -d
297 * (dflag) we are done with this pattern. We may also be handed a file
298 * in the subtree of a directory. in that case when we are operating
299 * with -d, this pattern was already selected and we are done
301 if (pt->flgs & DIR_MTCH)
304 if (!dflag && ((pt->pend != NULL) || (arcn->type == PAX_DIR))) {
306 * ok we matched a directory and we are allowing
307 * subtree matches but because of the -n only its children will
308 * match. This is tagged as a DIR_MTCH type.
309 * WATCH IT, the code assumes that pt->pend points
310 * into arcn->name and arcn->name has not been modified.
311 * If not we will have a big mess. Yup this is another kludge
315 * if this was a prefix match, remove trailing part of path
316 * so we can copy it. Future matches will be exact prefix match
318 if (pt->pend != NULL)
321 if ((pt->pstr = strdup(arcn->name)) == NULL) {
322 paxwarn(1, "Pattern select out of memory");
323 if (pt->pend != NULL)
330 * put the trailing / back in the source string
332 if (pt->pend != NULL) {
336 pt->plen = strlen(pt->pstr);
339 * strip off any trailing /, this should really never happen
342 if (*(pt->pstr + len) == '/') {
343 *(pt->pstr + len) = '\0';
346 pt->flgs = DIR_MTCH | MTCH;
352 * we are then done with this pattern, so we delete it from the list
353 * because it can never be used for another match.
354 * Seems kind of strange to do for a -c, but the pax spec is really
355 * vague on the interaction of -c -n and -d. We assume that when -c
356 * and the pattern rejects a member (i.e. it matched it) it is done.
357 * In effect we place the order of the flags as having -c last.
361 while ((pt != NULL) && (pt != arcn->pat)) {
368 * should never happen....
370 paxwarn(1, "Pattern list inconsistent");
381 * see if this archive member matches any supplied pattern, if a match
382 * is found, arcn->pat is set to point at the potential pattern. Later if
383 * this archive member is "selected" we process and mark the pattern as
384 * one which matched a selected archive member (see pat_sel())
386 * 0 if this archive member should be processed, 1 if it should be
387 * skipped and -1 if we are done with all patterns (and pax should quit
388 * looking for more members)
392 pat_match(ARCHD *arcn)
399 * if there are no more patterns and we have -n (and not -c) we are
400 * done. otherwise with no patterns to match, matches all
402 if (pathead == NULL) {
409 * have to search down the list one at a time looking for a match.
414 * check for a file name match unless we have DIR_MTCH set in
415 * this pattern then we want a prefix match
417 if (pt->flgs & DIR_MTCH) {
419 * this pattern was matched before to a directory
420 * as we must have -n set for this (but not -d). We can
421 * only match CHILDREN of that directory so we must use
422 * an exact prefix match (no wildcards).
424 if ((arcn->name[pt->plen] == '/') &&
425 (strncmp(pt->pstr, arcn->name, pt->plen) == 0))
427 } else if (fn_match(pt->pstr, arcn->name, &pt->pend) == 0)
433 * return the result, remember that cflag (-c) inverts the sense of a
437 return(cflag ? 0 : 1);
440 * We had a match, now when we invert the sense (-c) we reject this
441 * member. However we have to tag the pattern a being successful, (in a
442 * match, not in selecting an archive member) so we call pat_sel() here.
448 if (pat_sel(arcn) < 0)
457 * 0 if this archive member should be processed, 1 if it should be
458 * skipped and -1 if we are done with all patterns (and pax should quit
459 * looking for more members)
460 * Note: *pend may be changed to show where the prefix ends.
464 fn_match(char *pattern, char *string, char **pend)
471 switch (c = *pattern++) {
474 * Ok we found an exact match
480 * Check if it is a prefix match
482 if ((dflag == 1) || (*string != '/'))
486 * It is a prefix match, remember where the trailing
492 if ((test = *string++) == '\0')
498 * Collapse multiple *'s.
504 * Optimized hack for pattern with a * at the end
510 * General case, use recursion.
512 while ((test = *string) != '\0') {
513 if (!fn_match(pattern, string, pend))
522 if (((test = *string++) == '\0') ||
523 ((pattern = range_match(pattern, test)) == NULL))
537 range_match(char *pattern, int test)
544 if ((negate = (*pattern == '!')) != 0)
547 while ((c = *pattern++) != ']') {
554 if ((*pattern == '-') && ((c2 = pattern[1]) != '\0') &&
556 if ((c <= test) && (test <= c2))
559 } else if (c == test)
562 return (ok == negate ? NULL : pattern);
567 * modify a selected file name. first attempt to apply replacement string
568 * expressions, then apply interactive file rename. We apply replacement
569 * string expressions to both filenames and file links (if we didn't the
570 * links would point to the wrong place, and we could never be able to
571 * move an archive that has a file link in it). When we rename files
572 * interactively, we store that mapping (old name to user input name) so
573 * if we spot any file links to the old file name in the future, we will
574 * know exactly how to fix the file link.
576 * 0 continue to process file, 1 skip this file, -1 pax is finished
580 mod_name(ARCHD *arcn)
585 * Strip off leading '/' if appropriate.
586 * Currently, this option is only set for the tar format.
588 if (rmleadslash && arcn->name[0] == '/') {
589 if (arcn->name[1] == '\0') {
592 (void)memmove(arcn->name, &arcn->name[1],
596 if (rmleadslash < 2) {
598 paxwarn(0, "Removing leading / from absolute path names in the archive");
601 if (rmleadslash && arcn->ln_name[0] == '/' &&
602 (arcn->type == PAX_HLK || arcn->type == PAX_HRG)) {
603 if (arcn->ln_name[1] == '\0') {
604 arcn->ln_name[0] = '.';
606 (void)memmove(arcn->ln_name, &arcn->ln_name[1],
607 strlen(arcn->ln_name));
610 if (rmleadslash < 2) {
612 paxwarn(0, "Removing leading / from absolute path names in the archive");
617 * IMPORTANT: We have a problem. what do we do with symlinks?
618 * Modifying a hard link name makes sense, as we know the file it
619 * points at should have been seen already in the archive (and if it
620 * wasn't seen because of a read error or a bad archive, we lose
621 * anyway). But there are no such requirements for symlinks. On one
622 * hand the symlink that refers to a file in the archive will have to
623 * be modified to so it will still work at its new location in the
624 * file system. On the other hand a symlink that points elsewhere (and
625 * should continue to do so) should not be modified. There is clearly
626 * no perfect solution here. So we handle them like hardlinks. Clearly
627 * a replacement made by the interactive rename mapping is very likely
628 * to be correct since it applies to a single file and is an exact
629 * match. The regular expression replacements are a little harder to
630 * justify though. We claim that the symlink name is only likely
631 * to be replaced when it points within the file tree being moved and
632 * in that case it should be modified. what we really need to do is to
633 * call an oracle here. :)
635 if (rephead != NULL) {
637 * we have replacement strings, modify the name and the link
640 if ((res = rep_name(arcn->name, &(arcn->nlen), 1)) != 0)
643 if (((arcn->type == PAX_SLK) || (arcn->type == PAX_HLK) ||
644 (arcn->type == PAX_HRG)) &&
645 ((res = rep_name(arcn->ln_name, &(arcn->ln_nlen), 0)) != 0))
651 * perform interactive file rename, then map the link if any
653 if ((res = tty_rename(arcn)) != 0)
655 if ((arcn->type == PAX_SLK) || (arcn->type == PAX_HLK) ||
656 (arcn->type == PAX_HRG))
657 sub_name(arcn->ln_name, &(arcn->ln_nlen), sizeof(arcn->ln_name));
664 * Prompt the user for a replacement file name. A "." keeps the old name,
665 * a empty line skips the file, and an EOF on reading the tty, will cause
666 * pax to stop processing and exit. Otherwise the file name input, replaces
669 * 0 process this file, 1 skip this file, -1 we need to exit pax
673 tty_rename(ARCHD *arcn)
675 char tmpname[PAXPATHLEN+2];
679 * prompt user for the replacement name for a file, keep trying until
680 * we get some reasonable input. Archives may have more than one file
681 * on them with the same name (from updates etc). We print verbose info
682 * on the file so the user knows what is up.
684 tty_prnt("\nATTENTION: %s interactive file rename operation.\n", argv0);
688 tty_prnt("Input new name, or a \".\" to keep the old name, ");
689 tty_prnt("or a \"return\" to skip this file.\n");
690 tty_prnt("Input > ");
691 if (tty_read(tmpname, sizeof(tmpname)) < 0)
693 if (strcmp(tmpname, "..") == 0) {
694 tty_prnt("Try again, illegal file name: ..\n");
697 if (strlen(tmpname) > PAXPATHLEN) {
698 tty_prnt("Try again, file name too long\n");
705 * empty file name, skips this file. a "." leaves it alone
707 if (tmpname[0] == '\0') {
708 tty_prnt("Skipping file.\n");
711 if ((tmpname[0] == '.') && (tmpname[1] == '\0')) {
712 tty_prnt("Processing continues, name unchanged.\n");
717 * ok the name changed. We may run into links that point at this
718 * file later. we have to remember where the user sent the file
719 * in order to repair any links.
721 tty_prnt("Processing continues, name changed to: %s\n", tmpname);
722 res = add_name(arcn->name, arcn->nlen, tmpname);
723 arcn->nlen = l_strncpy(arcn->name, tmpname, sizeof(arcn->name) - 1);
724 arcn->name[arcn->nlen] = '\0';
732 * fix up the file name and the link name (if any) so this file will land
733 * in the destination directory (used during copy() -rw).
735 * 0 if ok, -1 if failure (name too long)
739 set_dest(ARCHD *arcn, char *dest_dir, int dir_len)
741 if (fix_path(arcn->name, &(arcn->nlen), dest_dir, dir_len) < 0)
745 * It is really hard to deal with symlinks here, we cannot be sure
746 * if the name they point was moved (or will be moved). It is best to
749 if ((arcn->type != PAX_HLK) && (arcn->type != PAX_HRG))
752 if (fix_path(arcn->ln_name, &(arcn->ln_nlen), dest_dir, dir_len) < 0)
759 * concatenate dir_name and or_name and store the result in or_name (if
760 * it fits). This is one ugly function.
762 * 0 if ok, -1 if the final name is too long
766 fix_path( char *or_name, int *or_len, char *dir_name, int dir_len)
774 * we shift the or_name to the right enough to tack in the dir_name
775 * at the front. We make sure we have enough space for it all before
776 * we start. since dest always ends in a slash, we skip of or_name
777 * if it also starts with one.
780 src = start + *or_len;
781 dest = src + dir_len;
786 if ((len = dest - or_name) > PAXPATHLEN) {
787 paxwarn(1, "File name %s/%s, too long", dir_name, start);
793 * enough space, shift
797 src = dir_name + dir_len - 1;
800 * splice in the destination directory name
802 while (src >= dir_name)
805 *(or_name + len) = '\0';
811 * walk down the list of replacement strings applying each one in order.
812 * when we find one with a successful substitution, we modify the name
813 * as specified. if required, we print the results. if the resulting name
814 * is empty, we will skip this archive member. We use the regexp(3)
815 * routines (regexp() ought to win a prize as having the most cryptic
816 * library function manual page).
818 * name is the file name we are going to apply the regular expressions to
819 * (and may be modified)
820 * nlen is the length of this name (and is modified to hold the length of
822 * prnt is a flag that says whether to print the final result.
824 * 0 if substitution was successful, 1 if we are to skip the file (the name
829 rep_name(char *name, int *nlen, int prnt)
838 regmatch_t pm[MAXSUBEXP];
839 char nname[PAXPATHLEN+1]; /* final result of all replacements */
840 char buf1[PAXPATHLEN+1]; /* where we work on the name */
843 * copy the name into buf1, where we will work on it. We need to keep
844 * the orig string around so we can print out the result of the final
845 * replacement. We build up the final result in nname. inpt points at
846 * the string we apply the regular expression to. prnt is used to
847 * suppress printing when we handle replacements on the link field
848 * (the user already saw that substitution go by)
851 (void)strlcpy(buf1, name, sizeof(buf1));
854 endpt = outpt + PAXPATHLEN;
857 * try each replacement string in order
862 * check for a successful substitution, if not go to
863 * the next pattern, or cleanup if we were global
865 if (regexec(&(pt->rcmp), inpt, MAXSUBEXP, pm, 0) != 0)
869 * ok we found one. We have three parts, the prefix
870 * which did not match, the section that did and the
871 * tail (that also did not match). Copy the prefix to
872 * the final output buffer (watching to make sure we
873 * do not create a string too long).
876 rpt = inpt + pm[0].rm_so;
878 while ((inpt < rpt) && (outpt < endpt))
884 * for the second part (which matched the regular
885 * expression) apply the substitution using the
886 * replacement string and place it the prefix in the
887 * final output. If we have problems, skip it.
889 if ((res = resub(&(pt->rcmp),pm,inpt,pt->nstr,outpt,endpt))
892 paxwarn(1, "Replacement name error %s",
899 * we set up to look again starting at the first
900 * character in the tail (of the input string right
901 * after the last character matched by the regular
902 * expression (inpt always points at the first char in
903 * the string to process). If we are not doing a global
904 * substitution, we will use inpt to copy the tail to
905 * the final result. Make sure we do not overrun the
908 inpt += pm[0].rm_eo - pm[0].rm_so;
910 if ((outpt == endpt) || (*inpt == '\0'))
914 * if the user wants global we keep trying to
915 * substitute until it fails, then we are done.
917 } while (pt->flgs & GLOB);
923 * a successful substitution did NOT occur, try the next one
930 * we had a substitution, copy the last tail piece (if there is
931 * room) to the final result
933 while ((outpt < endpt) && (*inpt != '\0'))
937 if ((outpt == endpt) && (*inpt != '\0')) {
939 paxwarn(1,"Replacement name too long %s >> %s",
945 * inform the user of the result if wanted
947 if (prnt && (pt->flgs & PRNT)) {
949 (void)fprintf(stderr,"%s >> <empty string>\n",
952 (void)fprintf(stderr,"%s >> %s\n", name, nname);
956 * if empty inform the caller this file is to be skipped
957 * otherwise copy the new name over the orig name and return
961 *nlen = l_strncpy(name, nname, PAXPATHLEN + 1);
962 name[PAXPATHLEN] = '\0';
970 * apply the replacement to the matched expression. expand out the old
971 * style ed(1) subexpression expansion.
973 * -1 if error, or the number of characters added to the destination.
977 resub(regex_t *rp, regmatch_t *pm, char *orig, char *src, char *dest,
989 subexcnt = rp->re_nsub;
990 while ((dpt < destend) && ((c = *spt++) != '\0')) {
992 * see if we just have an ordinary replacement character
993 * or we refer to a subexpression.
997 } else if ((c == '\\') && (*spt >= '0') && (*spt <= '9')) {
999 * make sure there is a subexpression as specified
1001 if ((len = *spt++ - '0') > subexcnt)
1006 * Ordinary character, just copy it
1008 if ((c == '\\') && ((*spt == '\\') || (*spt == '&')))
1015 * continue if the subexpression is bogus
1017 if ((pmpt->rm_so < 0) || (pmpt->rm_eo < 0) ||
1018 ((len = pmpt->rm_eo - pmpt->rm_so) <= 0))
1022 * copy the subexpression to the destination.
1023 * fail if we run out of space or the match string is damaged
1025 if (len > (destend - dpt))
1026 len = destend - dpt;
1027 if (l_strncpy(dpt, orig + pmpt->rm_so, len) != len)