2 * The filter include/exclude routines.
4 * Copyright (C) 1996-2001 Andrew Tridgell <tridge@samba.org>
5 * Copyright (C) 1996 Paul Mackerras
6 * Copyright (C) 2002 Martin Pool
7 * Copyright (C) 2003-2008 Wayne Davison
9 * This program is free software; you can redistribute it and/or modify
10 * it under the terms of the GNU General Public License as published by
11 * the Free Software Foundation; either version 3 of the License, or
12 * (at your option) any later version.
14 * This program is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17 * GNU General Public License for more details.
19 * You should have received a copy of the GNU General Public License along
20 * with this program; if not, visit the http://fsf.org website.
29 extern int local_server;
30 extern int prune_empty_dirs;
31 extern int ignore_perishable;
32 extern int delete_mode;
33 extern int delete_excluded;
34 extern int cvs_exclude;
35 extern int sanitize_paths;
36 extern int protocol_version;
39 extern char curr_dir[MAXPATHLEN];
40 extern unsigned int curr_dir_len;
41 extern unsigned int module_dirlen;
43 struct filter_list_struct filter_list = { .debug_type = "" };
44 struct filter_list_struct cvs_filter_list = { .debug_type = " [global CVS]" };
45 struct filter_list_struct daemon_filter_list = { .debug_type = " [daemon]" };
47 /* Need room enough for ":MODS " prefix plus some room to grow. */
48 #define MAX_RULE_PREFIX (16)
50 #define MODIFIERS_MERGE_FILE "-+Cenw"
51 #define MODIFIERS_INCL_EXCL "/!Crsp"
52 #define MODIFIERS_HIDE_PROTECT "/!p"
54 #define SLASH_WILD3_SUFFIX "/***"
56 /* The dirbuf is set by push_local_filters() to the current subdirectory
57 * relative to curr_dir that is being processed. The path always has a
58 * trailing slash appended, and the variable dirbuf_len contains the length
59 * of this path prefix. The path is always absolute. */
60 static char dirbuf[MAXPATHLEN+1];
61 static unsigned int dirbuf_len = 0;
62 static int dirbuf_depth;
64 /* This is True when we're scanning parent dirs for per-dir merge-files. */
65 static BOOL parent_dirscan = False;
67 /* This array contains a list of all the currently active per-dir merge
68 * files. This makes it easier to save the appropriate values when we
69 * "push" down into each subdirectory. */
70 static struct filter_struct **mergelist_parents;
71 static int mergelist_cnt = 0;
72 static int mergelist_size = 0;
74 /* Each filter_list_struct describes a singly-linked list by keeping track
75 * of both the head and tail pointers. The list is slightly unusual in that
76 * a parent-dir's content can be appended to the end of the local list in a
77 * special way: the last item in the local list has its "next" pointer set
78 * to point to the inherited list, but the local list's tail pointer points
79 * at the end of the local list. Thus, if the local list is empty, the head
80 * will be pointing at the inherited content but the tail will be NULL. To
81 * help you visualize this, here are the possible list arrangements:
83 * Completely Empty Local Content Only
84 * ================================== ====================================
85 * head -> NULL head -> Local1 -> Local2 -> NULL
86 * tail -> NULL tail -------------^
88 * Inherited Content Only Both Local and Inherited Content
89 * ================================== ====================================
90 * head -> Parent1 -> Parent2 -> NULL head -> L1 -> L2 -> P1 -> P2 -> NULL
91 * tail -> NULL tail ---------^
93 * This means that anyone wanting to traverse the whole list to use it just
94 * needs to start at the head and use the "next" pointers until it goes
95 * NULL. To add new local content, we insert the item after the tail item
96 * and update the tail (obviously, if "tail" was NULL, we insert it at the
97 * head). To clear the local list, WE MUST NOT FREE THE INHERITED CONTENT
98 * because it is shared between the current list and our parent list(s).
99 * The easiest way to handle this is to simply truncate the list after the
100 * tail item and then free the local list from the head. When inheriting
101 * the list for a new local dir, we just save off the filter_list_struct
102 * values (so we can pop back to them later) and set the tail to NULL.
105 static void teardown_mergelist(struct filter_struct *ex)
107 if (DEBUG_GTE(FILTER, 2)) {
108 rprintf(FINFO, "[%s] deactivating mergelist #%d%s\n",
109 who_am_i(), mergelist_cnt - 1,
110 ex->u.mergelist->debug_type);
113 /* We should deactivate mergelists in LIFO order. */
114 assert(mergelist_cnt > 0);
115 assert(ex == mergelist_parents[mergelist_cnt - 1]);
117 /* The parent_dirscan filters should have been freed. */
118 assert(ex->u.mergelist->parent_dirscan_head == NULL);
120 free(ex->u.mergelist->debug_type);
121 free(ex->u.mergelist);
125 static void free_filter(struct filter_struct *ex)
127 if (ex->match_flags & MATCHFLG_PERDIR_MERGE)
128 teardown_mergelist(ex);
133 static void free_filters(struct filter_struct *head)
135 struct filter_struct *rev_head = NULL;
137 /* Reverse the list so we deactivate mergelists in the proper LIFO
140 struct filter_struct *next = head->next;
141 head->next = rev_head;
147 struct filter_struct *prev = rev_head->next;
148 free_filter(rev_head);
153 /* Helper for identifying duplicate per-dir merge rules. */
154 static const char *perdir_merge_relative_part(const char *merge_file)
156 if (merge_file[0] == '/' || strncmp(merge_file, "../", 3) == 0)
157 return strrchr(merge_file, '/') + 1; /* There is a slash. */
162 /* Build a filter structure given a filter pattern. The value in "pat"
163 * is not null-terminated. */
164 static void add_rule(struct filter_list_struct *listp, const char *pat,
165 unsigned int pat_len, uint32 mflags, int xflags)
167 struct filter_struct *ret;
169 unsigned int pre_len, suf_len, slash_cnt = 0;
171 if (DEBUG_GTE(FILTER, 2)) {
172 rprintf(FINFO, "[%s] add_rule(%s%.*s%s)%s\n",
173 who_am_i(), get_rule_prefix(mflags, pat, 0, NULL),
175 (mflags & MATCHFLG_DIRECTORY) ? "/" : "",
179 /* These flags also indicate that we're reading a list that
180 * needs to be filtered now, not post-filtered later. */
181 if (xflags & (XFLG_ANCHORED2ABS|XFLG_ABS_IF_SLASH)) {
182 uint32 mf = mflags & (MATCHFLG_RECEIVER_SIDE|MATCHFLG_SENDER_SIDE);
184 if (mf == MATCHFLG_RECEIVER_SIDE)
187 if (mf == MATCHFLG_SENDER_SIDE)
192 if (!(ret = new0(struct filter_struct)))
193 out_of_memory("add_rule");
195 if (pat_len > 1 && pat[pat_len-1] == '/') {
197 mflags |= MATCHFLG_DIRECTORY;
200 for (cp = pat; cp < pat + pat_len; cp++) {
205 if (!(mflags & (MATCHFLG_ABS_PATH | MATCHFLG_MERGE_FILE))
206 && ((xflags & (XFLG_ANCHORED2ABS|XFLG_ABS_IF_SLASH) && *pat == '/')
207 || (xflags & XFLG_ABS_IF_SLASH && slash_cnt))) {
208 mflags |= MATCHFLG_ABS_PATH;
210 pre_len = dirbuf_len - module_dirlen - 1;
216 /* The daemon wants dir-exclude rules to get an appended "/" + "***". */
217 if (xflags & XFLG_DIR2WILD3
218 && BITS_SETnUNSET(mflags, MATCHFLG_DIRECTORY, MATCHFLG_INCLUDE)) {
219 mflags &= ~MATCHFLG_DIRECTORY;
220 suf_len = sizeof SLASH_WILD3_SUFFIX - 1;
224 if (!(ret->pattern = new_array(char, pre_len + pat_len + suf_len + 1)))
225 out_of_memory("add_rule");
227 memcpy(ret->pattern, dirbuf + module_dirlen, pre_len);
228 for (cp = ret->pattern; cp < ret->pattern + pre_len; cp++) {
233 strlcpy(ret->pattern + pre_len, pat, pat_len + 1);
236 memcpy(ret->pattern + pat_len, SLASH_WILD3_SUFFIX, suf_len+1);
241 if (strpbrk(ret->pattern, "*[?")) {
242 mflags |= MATCHFLG_WILD;
243 if ((cp = strstr(ret->pattern, "**")) != NULL) {
244 mflags |= MATCHFLG_WILD2;
245 /* If the pattern starts with **, note that. */
246 if (cp == ret->pattern)
247 mflags |= MATCHFLG_WILD2_PREFIX;
248 /* If the pattern ends with ***, note that. */
250 && ret->pattern[pat_len-3] == '*'
251 && ret->pattern[pat_len-2] == '*'
252 && ret->pattern[pat_len-1] == '*')
253 mflags |= MATCHFLG_WILD3_SUFFIX;
257 if (mflags & MATCHFLG_PERDIR_MERGE) {
258 struct filter_list_struct *lp;
261 clean_fname(ret->pattern, CFN_COLLAPSE_DOT_DOT_DIRS);
262 /* Throw out a corner case so that we are left with either a
263 * ..-free path (relative) or an absolute or ../-starting path
264 * (parent_dirscanning).. */
265 if (strcmp(ret->pattern, "..") == 0) {
266 rprintf(FERROR, "Dir-merging .. is not allowed.\n");
267 exit_cleanup(RERR_UNSUPPORTED);
269 /* If this per-dir merge file was already mentioned, don't
271 cp = perdir_merge_relative_part(ret->pattern);
272 for (i = 0; i < mergelist_cnt; i++) {
273 struct filter_struct *ex = mergelist_parents[i];
274 if (strcmp(cp, perdir_merge_relative_part(ex->pattern)) == 0) {
280 if (!(lp = new_array(struct filter_list_struct, 1)))
281 out_of_memory("add_rule");
282 lp->head = lp->tail = lp->parent_dirscan_head = NULL;
283 if (asprintf(&lp->debug_type, " [per-dir %s]", cp) < 0)
284 out_of_memory("add_rule");
285 ret->u.mergelist = lp;
287 if (mergelist_cnt == mergelist_size) {
289 mergelist_parents = realloc_array(mergelist_parents,
290 struct filter_struct *,
292 if (!mergelist_parents)
293 out_of_memory("add_rule");
295 if (DEBUG_GTE(FILTER, 2)) {
296 rprintf(FINFO, "[%s] activating mergelist #%d%s\n",
297 who_am_i(), mergelist_cnt, lp->debug_type);
299 mergelist_parents[mergelist_cnt++] = ret;
301 ret->u.slash_cnt = slash_cnt;
303 ret->match_flags = mflags;
306 ret->next = listp->head;
307 listp->head = listp->tail = ret;
309 ret->next = listp->tail->next;
310 listp->tail->next = ret;
315 static void clear_filter_list(struct filter_list_struct *listp)
318 /* Truncate any inherited items from the local list. */
319 listp->tail->next = NULL;
320 /* Now free everything that is left. */
321 free_filters(listp->head);
324 listp->head = listp->tail = NULL;
327 /* parse_merge_name is called to process merge-file names in three cases:
329 * 1. When an non-per-dir merge rule is added.
330 * - Sanitize path using NULL (module_dir) and dirbuf_depth (which is 0 as
331 * desired if the rule came from initial filter processing rather than a
332 * per-dir merge file). If non-sanitizing, just cancel .. components.
333 * - If the path is relative, prepend dirbuf. During initial filter
334 * processing, dirbuf is empty, so this step has no effect and we will
335 * follow the path relative to the current directory, as desired. From a
336 * per-dir merge file, this step makes the path absolute so that we
337 * interpret it correctly.
338 * 2. When a dir-merge rule is set up (setup_merge_file), with dirbuf containing
339 * the topmost dir to which it will apply.
340 * - Sanitize path using NULL (module_dir) and dirbuf_depth. If non-
341 * sanitizing, just cancel .. components.
342 * - If the path begins with ../, then prepend the dirbuf, yielding an
344 * - A path that is absolute at this point will activate the parent_dirscan
345 * in setup_merge_file, while a relative path (which contains no .. due to
346 * sanitization/cleaning) will be stored for following from each dir.
347 * 3. When a dir-merge rule is added during another parent_dirscan.
348 * - We want to interpret the rule relative to the ancestor dir in which it
349 * was defined, even though we don't actually push into that dir to set up
350 * the rule. Thus, we apply the same processing as in #2 while we have the
351 * desired dirbuf value, leaving an absolute or ..-free relative path that
352 * #2 won't modify further, except...
353 * - #2 will prepend the module_dir to an absolute path, so we are passed
354 * prefix_skip == module_dirlen to avoid prepending a duplicate module_dir
357 * FIXME: A per-dir merge rule added during a parent-dirscan needs to load
358 * files in the same ancestor dirs, so it should be made absolute
359 * unconditionally. This might make a relative rule parent-dirscanning, and
360 * it will mess up multicomponent relative rules. Quite possibly I should
361 * introduce a better syntax that allows separate specification of the relative
362 * and parent-dirscanning parts of the path, like ./ for --relative.
364 * When this is called on a rule addition (#1 or #3), merge_file is not null-
365 * terminated, but len_ptr points to its length and is updated with the length
366 * of the result. In all cases, the result is null-terminated. */
367 static char *parse_merge_name(const char *merge_file, unsigned int *len_ptr,
368 unsigned int prefix_skip, BOOL is_perdir)
370 static char buf[MAXPATHLEN];
371 char *fn, tmpbuf[MAXPATHLEN];
374 fn = *merge_file == '/' ? buf : tmpbuf;
375 if (sanitize_paths) {
376 const char *r = prefix_skip ? "/" : NULL;
377 /* null-terminate the name if it isn't already */
378 if (len_ptr && merge_file[*len_ptr]) {
379 char *to = fn == buf ? tmpbuf : buf;
380 strlcpy(to, merge_file, *len_ptr + 1);
383 if (!sanitize_path(fn, merge_file, r, dirbuf_depth, SP_DEFAULT)) {
384 rprintf(FERROR, "merge-file name overflows: %s\n",
390 strlcpy(fn, merge_file, len_ptr ? *len_ptr + 1 : MAXPATHLEN);
391 fn_len = clean_fname(fn, CFN_COLLAPSE_DOT_DOT_DIRS);
394 /* If the name isn't in buf yet, it's relative. In case #1, we want to
395 * prepend the dirbuf unconditionally. Otherwise, we want to make
396 * parent_dirscan-activating paths absolute but leave others alone. */
397 if (fn != buf && (!is_perdir || strncmp(fn, "../", 3) == 0)) {
398 int d_len = dirbuf_len - prefix_skip;
399 if (d_len + fn_len >= MAXPATHLEN) {
400 rprintf(FERROR, "merge-file name overflows: %s\n", fn);
403 memcpy(buf, dirbuf + prefix_skip, d_len);
404 memcpy(buf + d_len, fn, fn_len + 1);
406 fn_len = clean_fname(buf, CFN_COLLAPSE_DOT_DOT_DIRS);
414 /* Sets the dirbuf and dirbuf_len values. */
415 void set_filter_dir(const char *dir, unsigned int dirlen)
419 memcpy(dirbuf, curr_dir, curr_dir_len);
420 dirbuf[curr_dir_len] = '/';
421 len = curr_dir_len + 1;
422 if (len + dirlen >= MAXPATHLEN)
426 memcpy(dirbuf + len, dir, dirlen);
427 dirbuf[dirlen + len] = '\0';
428 dirbuf_len = clean_fname(dirbuf, CFN_COLLAPSE_DOT_DOT_DIRS);
429 if (dirbuf_len > 1 && dirbuf[dirbuf_len-1] == '.'
430 && dirbuf[dirbuf_len-2] == '/')
433 dirbuf[dirbuf_len++] = '/';
434 dirbuf[dirbuf_len] = '\0';
436 dirbuf_depth = count_dir_elements(dirbuf + module_dirlen);
439 /* This routine takes a per-dir merge-file entry and finishes its setup.
440 * If the name has a path portion then we check to see if it refers to a
441 * parent directory of the first transfer dir. If it does, we scan all the
442 * dirs from that point through the parent dir of the transfer dir looking
443 * for the per-dir merge-file in each one. */
444 static BOOL setup_merge_file(int mergelist_num, struct filter_struct *ex,
445 struct filter_list_struct *lp)
447 char buf[MAXPATHLEN];
448 char *x, *y, *pat = ex->pattern;
451 /* Case #2 in parse_merge_name comment: make a parent_dirscanning path
453 if (!(x = parse_merge_name(pat, NULL, 0, True)) || *x != '/')
456 if (DEBUG_GTE(FILTER, 2)) {
457 rprintf(FINFO, "[%s] performing parent_dirscan for mergelist #%d%s\n",
458 who_am_i(), mergelist_num, lp->debug_type);
462 ex->pattern = strdup(y+1);
466 strlcpy(buf, x, MAXPATHLEN);
468 pathjoin(buf, MAXPATHLEN, dirbuf, x);
470 len = clean_fname(buf, CFN_COLLAPSE_DOT_DOT_DIRS);
471 if (len != 1 && len < MAXPATHLEN-1) {
475 /* This ensures that the specified dir is a parent of the transfer. */
476 for (x = buf, y = dirbuf; *x && *x == *y; x++, y++) {}
478 y += strlen(y); /* nope -- skip the scan */
480 parent_dirscan = True;
482 char save[MAXPATHLEN];
483 strlcpy(save, y, MAXPATHLEN);
485 dirbuf_len = y - dirbuf;
486 strlcpy(x, ex->pattern, MAXPATHLEN - (x - buf));
487 parse_filter_file(lp, buf, ex->match_flags, XFLG_ANCHORED2ABS);
488 if (ex->match_flags & MATCHFLG_NO_INHERIT) {
489 /* Free the undesired rules to clean up any per-dir
490 * mergelists they defined. Otherwise pop_local_filters
491 * may crash trying to restore nonexistent state for
492 * those mergelists. */
493 free_filters(lp->head);
497 strlcpy(y, save, MAXPATHLEN);
498 while ((*x++ = *y++) != '/') {}
500 /* Save current head for freeing when the mergelist becomes inactive. */
501 lp->parent_dirscan_head = lp->head;
502 parent_dirscan = False;
503 if (DEBUG_GTE(FILTER, 2)) {
504 rprintf(FINFO, "[%s] completed parent_dirscan for mergelist #%d%s\n",
505 who_am_i(), mergelist_num, lp->debug_type);
511 struct local_filter_state {
513 struct filter_list_struct mergelists[1];
516 /* Each time rsync changes to a new directory it call this function to
517 * handle all the per-dir merge-files. The "dir" value is the current path
518 * relative to curr_dir (which might not be null-terminated). We copy it
519 * into dirbuf so that we can easily append a file name on the end. */
520 void *push_local_filters(const char *dir, unsigned int dirlen)
522 struct local_filter_state *push;
525 set_filter_dir(dir, dirlen);
526 if (DEBUG_GTE(FILTER, 2)) {
527 rprintf(FINFO, "[%s] pushing local filters for %s\n",
531 if (!mergelist_cnt) {
532 /* No old state to save and no new merge files to push. */
536 push = (struct local_filter_state *)new_array(char,
537 sizeof (struct local_filter_state)
538 + (mergelist_cnt-1) * sizeof (struct filter_list_struct));
540 out_of_memory("push_local_filters");
542 push->mergelist_cnt = mergelist_cnt;
543 for (i = 0; i < mergelist_cnt; i++) {
544 memcpy(&push->mergelists[i], mergelist_parents[i]->u.mergelist,
545 sizeof (struct filter_list_struct));
548 /* Note: parse_filter_file() might increase mergelist_cnt, so keep
549 * this loop separate from the above loop. */
550 for (i = 0; i < mergelist_cnt; i++) {
551 struct filter_struct *ex = mergelist_parents[i];
552 struct filter_list_struct *lp = ex->u.mergelist;
554 if (DEBUG_GTE(FILTER, 2)) {
555 rprintf(FINFO, "[%s] pushing mergelist #%d%s\n",
556 who_am_i(), i, lp->debug_type);
559 lp->tail = NULL; /* Switch any local rules to inherited. */
560 if (ex->match_flags & MATCHFLG_NO_INHERIT)
563 if (ex->match_flags & MATCHFLG_FINISH_SETUP) {
564 ex->match_flags &= ~MATCHFLG_FINISH_SETUP;
565 if (setup_merge_file(i, ex, lp))
566 set_filter_dir(dir, dirlen);
569 if (strlcpy(dirbuf + dirbuf_len, ex->pattern,
570 MAXPATHLEN - dirbuf_len) < MAXPATHLEN - dirbuf_len) {
571 parse_filter_file(lp, dirbuf, ex->match_flags,
574 io_error |= IOERR_GENERAL;
576 "cannot add local filter rules in long-named directory: %s\n",
579 dirbuf[dirbuf_len] = '\0';
585 void pop_local_filters(void *mem)
587 struct local_filter_state *pop = (struct local_filter_state *)mem;
589 int old_mergelist_cnt = pop ? pop->mergelist_cnt : 0;
591 if (DEBUG_GTE(FILTER, 2))
592 rprintf(FINFO, "[%s] popping local filters\n", who_am_i());
594 for (i = mergelist_cnt; i-- > 0; ) {
595 struct filter_struct *ex = mergelist_parents[i];
596 struct filter_list_struct *lp = ex->u.mergelist;
598 if (DEBUG_GTE(FILTER, 2)) {
599 rprintf(FINFO, "[%s] popping mergelist #%d%s\n",
600 who_am_i(), i, lp->debug_type);
603 clear_filter_list(lp);
605 if (i >= old_mergelist_cnt) {
606 /* This mergelist does not exist in the state to be
607 * restored. Free its parent_dirscan list to clean up
608 * any per-dir mergelists defined there so we don't
609 * crash trying to restore nonexistent state for them
610 * below. (Counterpart to setup_merge_file call in
611 * push_local_filters. Must be done here, not in
612 * free_filter, for LIFO order.) */
613 if (DEBUG_GTE(FILTER, 2)) {
614 rprintf(FINFO, "[%s] freeing parent_dirscan filters of mergelist #%d%s\n",
615 who_am_i(), i, ex->u.mergelist->debug_type);
617 free_filters(lp->parent_dirscan_head);
618 lp->parent_dirscan_head = NULL;
622 /* If we cleaned things up properly, the only still-active mergelists
623 * should be those with a state to be restored. */
624 assert(mergelist_cnt == old_mergelist_cnt);
627 /* No state to restore. */
631 for (i = 0; i < mergelist_cnt; i++) {
632 memcpy(mergelist_parents[i]->u.mergelist, &pop->mergelists[i],
633 sizeof (struct filter_list_struct));
639 void change_local_filter_dir(const char *dname, int dlen, int dir_depth)
641 static int cur_depth = -1;
642 static void *filt_array[MAXPATHLEN/2+1];
645 for ( ; cur_depth >= 0; cur_depth--) {
646 if (filt_array[cur_depth]) {
647 pop_local_filters(filt_array[cur_depth]);
648 filt_array[cur_depth] = NULL;
654 assert(dir_depth < MAXPATHLEN/2+1);
656 for ( ; cur_depth >= dir_depth; cur_depth--) {
657 if (filt_array[cur_depth]) {
658 pop_local_filters(filt_array[cur_depth]);
659 filt_array[cur_depth] = NULL;
663 cur_depth = dir_depth;
664 filt_array[cur_depth] = push_local_filters(dname, dlen);
667 static int rule_matches(const char *fname, struct filter_struct *ex, int name_is_dir)
669 int slash_handling, str_cnt = 0, anchored_match = 0;
670 int ret_match = ex->match_flags & MATCHFLG_NEGATE ? 0 : 1;
671 char *p, *pattern = ex->pattern;
672 const char *strings[16]; /* more than enough */
673 const char *name = fname + (*fname == '/');
678 if (!ex->u.slash_cnt && !(ex->match_flags & MATCHFLG_WILD2)) {
679 /* If the pattern does not have any slashes AND it does
680 * not have a "**" (which could match a slash), then we
681 * just match the name portion of the path. */
682 if ((p = strrchr(name,'/')) != NULL)
684 } else if (ex->match_flags & MATCHFLG_ABS_PATH && *fname != '/'
685 && curr_dir_len > module_dirlen + 1) {
686 /* If we're matching against an absolute-path pattern,
687 * we need to prepend our full path info. */
688 strings[str_cnt++] = curr_dir + module_dirlen + 1;
689 strings[str_cnt++] = "/";
690 } else if (ex->match_flags & MATCHFLG_WILD2_PREFIX && *fname != '/') {
691 /* Allow "**"+"/" to match at the start of the string. */
692 strings[str_cnt++] = "/";
694 strings[str_cnt++] = name;
696 /* Allow a trailing "/"+"***" to match the directory. */
697 if (ex->match_flags & MATCHFLG_WILD3_SUFFIX)
698 strings[str_cnt++] = "/";
699 } else if (ex->match_flags & MATCHFLG_DIRECTORY)
701 strings[str_cnt] = NULL;
703 if (*pattern == '/') {
708 if (!anchored_match && ex->u.slash_cnt
709 && !(ex->match_flags & MATCHFLG_WILD2)) {
710 /* A non-anchored match with an infix slash and no "**"
711 * needs to match the last slash_cnt+1 name elements. */
712 slash_handling = ex->u.slash_cnt + 1;
713 } else if (!anchored_match && !(ex->match_flags & MATCHFLG_WILD2_PREFIX)
714 && ex->match_flags & MATCHFLG_WILD2) {
715 /* A non-anchored match with an infix or trailing "**" (but not
716 * a prefixed "**") needs to try matching after every slash. */
719 /* The pattern matches only at the start of the path or name. */
723 if (ex->match_flags & MATCHFLG_WILD) {
724 if (wildmatch_array(pattern, strings, slash_handling))
726 } else if (str_cnt > 1) {
727 if (litmatch_array(pattern, strings, slash_handling))
729 } else if (anchored_match) {
730 if (strcmp(name, pattern) == 0)
733 int l1 = strlen(name);
734 int l2 = strlen(pattern);
736 strcmp(name+(l1-l2),pattern) == 0 &&
737 (l1==l2 || name[l1-(l2+1)] == '/')) {
746 static void report_filter_result(enum logcode code, char const *name,
747 struct filter_struct const *ent,
748 int name_is_dir, const char *type)
750 /* If a trailing slash is present to match only directories,
751 * then it is stripped out by add_rule(). So as a special
752 * case we add it back in here. */
754 if (DEBUG_GTE(FILTER, 1)) {
755 static char *actions[2][2]
756 = { {"show", "hid"}, {"risk", "protect"} };
757 const char *w = who_am_i();
758 rprintf(code, "[%s] %sing %s %s because of pattern %s%s%s\n",
759 w, actions[*w!='s'][!(ent->match_flags&MATCHFLG_INCLUDE)],
760 name_is_dir ? "directory" : "file", name, ent->pattern,
761 ent->match_flags & MATCHFLG_DIRECTORY ? "/" : "", type);
767 * Return -1 if file "name" is defined to be excluded by the specified
768 * exclude list, 1 if it is included, and 0 if it was not matched.
770 int check_filter(struct filter_list_struct *listp, enum logcode code,
771 const char *name, int name_is_dir)
773 struct filter_struct *ent;
775 for (ent = listp->head; ent; ent = ent->next) {
776 if (ignore_perishable && ent->match_flags & MATCHFLG_PERISHABLE)
778 if (ent->match_flags & MATCHFLG_PERDIR_MERGE) {
779 int rc = check_filter(ent->u.mergelist, code, name,
785 if (ent->match_flags & MATCHFLG_CVS_IGNORE) {
786 int rc = check_filter(&cvs_filter_list, code, name,
792 if (rule_matches(name, ent, name_is_dir)) {
793 report_filter_result(code, name, ent, name_is_dir,
795 return ent->match_flags & MATCHFLG_INCLUDE ? 1 : -1;
802 #define RULE_STRCMP(s,r) rule_strcmp((s), (r), sizeof (r) - 1)
804 static const uchar *rule_strcmp(const uchar *str, const char *rule, int rule_len)
806 if (strncmp((char*)str, rule, rule_len) != 0)
808 if (isspace(str[rule_len]) || str[rule_len] == '_' || !str[rule_len])
809 return str + rule_len - 1;
810 if (str[rule_len] == ',')
811 return str + rule_len;
815 /* Get the next include/exclude arg from the string. The token will not
816 * be '\0' terminated, so use the returned length to limit the string.
817 * Also, be sure to add this length to the returned pointer before passing
818 * it back to ask for the next token. This routine parses the "!" (list-
819 * clearing) token and (depending on the mflags) the various prefixes.
820 * The *mflags_ptr value will be set on exit to the new MATCHFLG_* bits
821 * for the current token. */
822 static const char *parse_rule_tok(const char *p, uint32 mflags, int xflags,
823 unsigned int *len_ptr, uint32 *mflags_ptr)
825 const uchar *s = (const uchar *)p;
829 if (mflags & MATCHFLG_WORD_SPLIT) {
830 /* Skip over any initial whitespace. */
833 /* Update to point to real start of rule. */
839 new_mflags = mflags & MATCHFLGS_FROM_CONTAINER;
841 /* Figure out what kind of a filter rule "s" is pointing at. Note
842 * that if MATCHFLG_NO_PREFIXES is set, the rule is either an include
843 * or an exclude based on the inheritance of the MATCHFLG_INCLUDE
844 * flag (above). XFLG_OLD_PREFIXES indicates a compatibility mode
845 * for old include/exclude patterns where just "+ " and "- " are
846 * allowed as optional prefixes. */
847 if (mflags & MATCHFLG_NO_PREFIXES) {
848 if (*s == '!' && mflags & MATCHFLG_CVS_IGNORE)
849 new_mflags |= MATCHFLG_CLEAR_LIST; /* Tentative! */
850 } else if (xflags & XFLG_OLD_PREFIXES) {
851 if (*s == '-' && s[1] == ' ') {
852 new_mflags &= ~MATCHFLG_INCLUDE;
854 } else if (*s == '+' && s[1] == ' ') {
855 new_mflags |= MATCHFLG_INCLUDE;
857 } else if (*s == '!')
858 new_mflags |= MATCHFLG_CLEAR_LIST; /* Tentative! */
860 char ch = 0, *mods = "";
863 if ((s = RULE_STRCMP(s, "clear")) != NULL)
867 if ((s = RULE_STRCMP(s, "dir-merge")) != NULL)
871 if ((s = RULE_STRCMP(s, "exclude")) != NULL)
875 if ((s = RULE_STRCMP(s, "hide")) != NULL)
879 if ((s = RULE_STRCMP(s, "include")) != NULL)
883 if ((s = RULE_STRCMP(s, "merge")) != NULL)
887 if ((s = RULE_STRCMP(s, "protect")) != NULL)
891 if ((s = RULE_STRCMP(s, "risk")) != NULL)
895 if ((s = RULE_STRCMP(s, "show")) != NULL)
906 new_mflags |= MATCHFLG_PERDIR_MERGE
907 | MATCHFLG_FINISH_SETUP;
910 new_mflags |= MATCHFLG_MERGE_FILE;
911 mods = MODIFIERS_INCL_EXCL MODIFIERS_MERGE_FILE;
914 new_mflags |= MATCHFLG_INCLUDE;
917 mods = MODIFIERS_INCL_EXCL;
920 new_mflags |= MATCHFLG_INCLUDE;
923 new_mflags |= MATCHFLG_SENDER_SIDE;
924 mods = MODIFIERS_HIDE_PROTECT;
927 new_mflags |= MATCHFLG_INCLUDE;
930 new_mflags |= MATCHFLG_RECEIVER_SIDE;
931 mods = MODIFIERS_HIDE_PROTECT;
934 new_mflags |= MATCHFLG_CLEAR_LIST;
938 rprintf(FERROR, "Unknown filter rule: `%s'\n", p);
939 exit_cleanup(RERR_SYNTAX);
941 while (mods && *++s && *s != ' ' && *s != '_') {
942 if (strchr(mods, *s) == NULL) {
943 if (mflags & MATCHFLG_WORD_SPLIT && isspace(*s)) {
949 "invalid modifier sequence at '%c' in filter rule: %s\n",
951 exit_cleanup(RERR_SYNTAX);
955 if (new_mflags & MATCHFLG_NO_PREFIXES)
957 new_mflags |= MATCHFLG_NO_PREFIXES;
960 if (new_mflags & MATCHFLG_NO_PREFIXES)
962 new_mflags |= MATCHFLG_NO_PREFIXES
966 new_mflags |= MATCHFLG_ABS_PATH;
969 new_mflags |= MATCHFLG_NEGATE;
972 if (new_mflags & MATCHFLG_NO_PREFIXES)
974 new_mflags |= MATCHFLG_NO_PREFIXES
975 | MATCHFLG_WORD_SPLIT
976 | MATCHFLG_NO_INHERIT
977 | MATCHFLG_CVS_IGNORE;
980 new_mflags |= MATCHFLG_EXCLUDE_SELF;
983 new_mflags |= MATCHFLG_NO_INHERIT;
986 new_mflags |= MATCHFLG_PERISHABLE;
989 new_mflags |= MATCHFLG_RECEIVER_SIDE;
992 new_mflags |= MATCHFLG_SENDER_SIDE;
995 new_mflags |= MATCHFLG_WORD_SPLIT;
1003 if (mflags & MATCHFLG_WORD_SPLIT) {
1004 const uchar *cp = s;
1005 /* Token ends at whitespace or the end of the string. */
1006 while (!isspace(*cp) && *cp != '\0')
1010 len = strlen((char*)s);
1012 if (new_mflags & MATCHFLG_CLEAR_LIST) {
1013 if (!(mflags & MATCHFLG_NO_PREFIXES)
1014 && !(xflags & XFLG_OLD_PREFIXES) && len) {
1016 "'!' rule has trailing characters: %s\n", p);
1017 exit_cleanup(RERR_SYNTAX);
1020 new_mflags &= ~MATCHFLG_CLEAR_LIST;
1021 } else if (!len && !(new_mflags & MATCHFLG_CVS_IGNORE)) {
1022 rprintf(FERROR, "unexpected end of filter rule: %s\n", p);
1023 exit_cleanup(RERR_SYNTAX);
1026 /* --delete-excluded turns an un-modified include/exclude into a
1027 * sender-side rule. We also affect per-dir merge files that take
1028 * no prefixes as a simple optimization. */
1030 && !(new_mflags & (MATCHFLG_RECEIVER_SIDE|MATCHFLG_SENDER_SIDE))
1031 && (!(new_mflags & MATCHFLG_PERDIR_MERGE)
1032 || new_mflags & MATCHFLG_NO_PREFIXES))
1033 new_mflags |= MATCHFLG_SENDER_SIDE;
1036 *mflags_ptr = new_mflags;
1037 return (const char *)s;
1041 static char default_cvsignore[] =
1042 /* These default ignored items come from the CVS manual. */
1043 "RCS SCCS CVS CVS.adm RCSLOG cvslog.* tags TAGS"
1044 " .make.state .nse_depinfo *~ #* .#* ,* _$* *$"
1045 " *.old *.bak *.BAK *.orig *.rej .del-*"
1046 " *.a *.olb *.o *.obj *.so *.exe"
1047 " *.Z *.elc *.ln core"
1048 /* The rest we added to suit ourself. */
1049 " .svn/ .git/ .bzr/";
1051 static void get_cvs_excludes(uint32 mflags)
1053 static int initialized = 0;
1054 char *p, fname[MAXPATHLEN];
1060 parse_rule(&cvs_filter_list, default_cvsignore,
1061 mflags | (protocol_version >= 30 ? MATCHFLG_PERISHABLE : 0),
1064 p = module_id >= 0 && lp_use_chroot(module_id) ? "/" : getenv("HOME");
1065 if (p && pathjoin(fname, MAXPATHLEN, p, ".cvsignore") < MAXPATHLEN)
1066 parse_filter_file(&cvs_filter_list, fname, mflags, 0);
1068 parse_rule(&cvs_filter_list, getenv("CVSIGNORE"), mflags, 0);
1072 void parse_rule(struct filter_list_struct *listp, const char *pattern,
1073 uint32 mflags, int xflags)
1075 unsigned int pat_len;
1083 /* Remember that the returned string is NOT '\0' terminated! */
1084 cp = parse_rule_tok(pattern, mflags, xflags,
1085 &pat_len, &new_mflags);
1089 pattern = cp + pat_len;
1091 if (pat_len >= MAXPATHLEN) {
1092 rprintf(FERROR, "discarding over-long filter: %.*s\n",
1097 if (new_mflags & MATCHFLG_CLEAR_LIST) {
1098 if (DEBUG_GTE(FILTER, 2)) {
1100 "[%s] clearing filter list%s\n",
1101 who_am_i(), listp->debug_type);
1103 clear_filter_list(listp);
1107 if (new_mflags & MATCHFLG_MERGE_FILE) {
1114 if (new_mflags & MATCHFLG_EXCLUDE_SELF) {
1115 const char *name = cp + len;
1116 while (name > cp && name[-1] != '/') name--;
1118 add_rule(listp, name, len, 0, 0);
1119 new_mflags &= ~MATCHFLG_EXCLUDE_SELF;
1122 if (new_mflags & MATCHFLG_PERDIR_MERGE) {
1123 if (parent_dirscan) {
1124 /* Case #3 in parse_merge_name comment:
1125 * record correct starting dir for a
1126 * parent_dirscanning per-dir merge rule
1127 * found during another parent_dirscan. */
1128 if (!(p = parse_merge_name(cp, &len,
1129 module_dirlen, True)))
1131 add_rule(listp, p, len, new_mflags, 0);
1135 /* Case #1 in parse_merge_name comment:
1136 * interpret the path w.r.t. the correct
1138 if (!(p = parse_merge_name(cp, &len, 0, False)))
1140 parse_filter_file(listp, p, new_mflags,
1146 add_rule(listp, cp, pat_len, new_mflags, xflags);
1148 if (new_mflags & MATCHFLG_CVS_IGNORE
1149 && !(new_mflags & MATCHFLG_MERGE_FILE))
1150 get_cvs_excludes(new_mflags);
1155 void parse_filter_file(struct filter_list_struct *listp, const char *fname,
1156 uint32 mflags, int xflags)
1159 char line[BIGPATHBUFLEN];
1160 char *eob = line + sizeof line - 1;
1161 int word_split = mflags & MATCHFLG_WORD_SPLIT;
1163 if (!fname || !*fname)
1166 if (*fname != '-' || fname[1] || am_server) {
1167 if (daemon_filter_list.head) {
1168 strlcpy(line, fname, sizeof line);
1169 clean_fname(line, CFN_COLLAPSE_DOT_DOT_DIRS);
1170 if (check_filter(&daemon_filter_list, FLOG, line, 0) < 0)
1173 fp = fopen(line, "rb");
1175 fp = fopen(fname, "rb");
1179 if (DEBUG_GTE(FILTER, 2)) {
1180 rprintf(FINFO, "[%s] parse_filter_file(%s,%x,%x)%s\n",
1181 who_am_i(), fname, mflags, xflags,
1182 fp ? "" : " [not found]");
1186 if (xflags & XFLG_FATAL_ERRORS) {
1187 rsyserr(FERROR, errno,
1188 "failed to open %sclude file %s",
1189 mflags & MATCHFLG_INCLUDE ? "in" : "ex",
1191 exit_cleanup(RERR_FILEIO);
1195 dirbuf[dirbuf_len] = '\0';
1199 int ch, overflow = 0;
1201 if ((ch = getc(fp)) == EOF) {
1202 if (ferror(fp) && errno == EINTR) {
1208 if (word_split && isspace(ch))
1210 if (eol_nulls? !ch : (ch == '\n' || ch == '\r'))
1218 rprintf(FERROR, "discarding over-long filter: %s...\n", line);
1222 /* Skip an empty token and (when line parsing) comments. */
1223 if (*line && (word_split || (*line != ';' && *line != '#')))
1224 parse_rule(listp, line, mflags, xflags);
1231 /* If the "for_xfer" flag is set, the prefix is made compatible with the
1232 * current protocol_version (if possible) or a NULL is returned (if not
1234 char *get_rule_prefix(int match_flags, const char *pat, int for_xfer,
1235 unsigned int *plen_ptr)
1237 static char buf[MAX_RULE_PREFIX+1];
1239 int legal_len = for_xfer && protocol_version < 29 ? 1 : MAX_RULE_PREFIX-1;
1241 if (match_flags & MATCHFLG_PERDIR_MERGE) {
1245 } else if (match_flags & MATCHFLG_INCLUDE)
1247 else if (legal_len != 1
1248 || ((*pat == '-' || *pat == '+') && pat[1] == ' '))
1253 if (match_flags & MATCHFLG_NEGATE)
1255 if (match_flags & MATCHFLG_CVS_IGNORE)
1258 if (match_flags & MATCHFLG_NO_INHERIT)
1260 if (match_flags & MATCHFLG_WORD_SPLIT)
1262 if (match_flags & MATCHFLG_NO_PREFIXES) {
1263 if (match_flags & MATCHFLG_INCLUDE)
1269 if (match_flags & MATCHFLG_EXCLUDE_SELF)
1271 if (match_flags & MATCHFLG_SENDER_SIDE
1272 && (!for_xfer || protocol_version >= 29))
1274 if (match_flags & MATCHFLG_RECEIVER_SIDE
1275 && (!for_xfer || protocol_version >= 29
1276 || (delete_excluded && am_sender)))
1278 if (match_flags & MATCHFLG_PERISHABLE) {
1279 if (!for_xfer || protocol_version >= 30)
1284 if (op - buf > legal_len)
1290 *plen_ptr = op - buf;
1294 static void send_rules(int f_out, struct filter_list_struct *flp)
1296 struct filter_struct *ent, *prev = NULL;
1298 for (ent = flp->head; ent; ent = ent->next) {
1299 unsigned int len, plen, dlen;
1303 /* Note we need to check delete_excluded here in addition to
1304 * the code in parse_rule_tok() because some rules may have
1305 * been added before we found the --delete-excluded option.
1306 * We must also elide any CVS merge-file rules to avoid a
1307 * backward compatibility problem, and we elide any no-prefix
1308 * merge files as an optimization (since they can only have
1309 * include/exclude rules). */
1310 if (ent->match_flags & MATCHFLG_SENDER_SIDE)
1311 elide = am_sender ? 1 : -1;
1312 if (ent->match_flags & MATCHFLG_RECEIVER_SIDE)
1313 elide = elide ? 0 : am_sender ? -1 : 1;
1314 else if (delete_excluded && !elide
1315 && (!(ent->match_flags & MATCHFLG_PERDIR_MERGE)
1316 || ent->match_flags & MATCHFLG_NO_PREFIXES))
1317 elide = am_sender ? 1 : -1;
1320 prev->next = ent->next;
1322 flp->head = ent->next;
1327 if (ent->match_flags & MATCHFLG_CVS_IGNORE
1328 && !(ent->match_flags & MATCHFLG_MERGE_FILE)) {
1329 int f = am_sender || protocol_version < 29 ? f_out : -2;
1330 send_rules(f, &cvs_filter_list);
1334 p = get_rule_prefix(ent->match_flags, ent->pattern, 1, &plen);
1337 "filter rules are too modern for remote rsync.\n");
1338 exit_cleanup(RERR_PROTOCOL);
1342 len = strlen(ent->pattern);
1343 dlen = ent->match_flags & MATCHFLG_DIRECTORY ? 1 : 0;
1344 if (!(plen + len + dlen))
1346 write_int(f_out, plen + len + dlen);
1348 write_buf(f_out, p, plen);
1349 write_buf(f_out, ent->pattern, len);
1351 write_byte(f_out, '/');
1356 /* This is only called by the client. */
1357 void send_filter_list(int f_out)
1359 int receiver_wants_list = prune_empty_dirs
1360 || (delete_mode && (!delete_excluded || protocol_version >= 29));
1362 if (local_server || (am_sender && !receiver_wants_list))
1364 if (cvs_exclude && am_sender) {
1365 if (protocol_version >= 29)
1366 parse_rule(&filter_list, ":C", 0, 0);
1367 parse_rule(&filter_list, "-C", 0, 0);
1370 send_rules(f_out, &filter_list);
1373 write_int(f_out, 0);
1376 if (!am_sender || protocol_version < 29)
1377 parse_rule(&filter_list, ":C", 0, 0);
1379 parse_rule(&filter_list, "-C", 0, 0);
1383 /* This is only called by the server. */
1384 void recv_filter_list(int f_in)
1386 char line[BIGPATHBUFLEN];
1387 int xflags = protocol_version >= 29 ? 0 : XFLG_OLD_PREFIXES;
1388 int receiver_wants_list = prune_empty_dirs
1390 && (!delete_excluded || protocol_version >= 29));
1393 if (!local_server && (am_sender || receiver_wants_list)) {
1394 while ((len = read_int(f_in)) != 0) {
1395 if (len >= sizeof line)
1396 overflow_exit("recv_rules");
1397 read_sbuf(f_in, line, len);
1398 parse_rule(&filter_list, line, 0, xflags);
1403 if (local_server || am_sender || protocol_version < 29)
1404 parse_rule(&filter_list, ":C", 0, 0);
1405 if (local_server || am_sender)
1406 parse_rule(&filter_list, "-C", 0, 0);
1409 if (local_server) /* filter out any rules that aren't for us. */
1410 send_rules(-1, &filter_list);