- Define the new CFN_* flags for clean_fname().
[rsync.git] / exclude.c
blobb7a00c602159e7ad699e483cd4529d87f5025c84
1 /*
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) 2003-2007 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.
23 #include "rsync.h"
25 extern int verbose;
26 extern int am_server;
27 extern int am_sender;
28 extern int eol_nulls;
29 extern int recurse;
30 extern int io_error;
31 extern int local_server;
32 extern int prune_empty_dirs;
33 extern int ignore_perishable;
34 extern int delete_mode;
35 extern int delete_excluded;
36 extern int cvs_exclude;
37 extern int sanitize_paths;
38 extern int protocol_version;
39 extern int module_id;
41 extern char curr_dir[];
42 extern unsigned int curr_dir_len;
43 extern unsigned int module_dirlen;
45 struct filter_list_struct filter_list = { 0, 0, "" };
46 struct filter_list_struct cvs_filter_list = { 0, 0, " [global CVS]" };
47 struct filter_list_struct server_filter_list = { 0, 0, " [daemon]" };
49 /* Need room enough for ":MODS " prefix plus some room to grow. */
50 #define MAX_RULE_PREFIX (16)
52 #define MODIFIERS_MERGE_FILE "-+Cenw"
53 #define MODIFIERS_INCL_EXCL "/!Crsp"
54 #define MODIFIERS_HIDE_PROTECT "/!p"
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 free_filter(struct filter_struct *ex)
107 if (ex->match_flags & MATCHFLG_PERDIR_MERGE) {
108 free(ex->u.mergelist->debug_type);
109 free(ex->u.mergelist);
110 mergelist_cnt--;
112 free(ex->pattern);
113 free(ex);
116 /* Build a filter structure given a filter pattern. The value in "pat"
117 * is not null-terminated. */
118 static void add_rule(struct filter_list_struct *listp, const char *pat,
119 unsigned int pat_len, uint32 mflags, int xflags)
121 struct filter_struct *ret;
122 const char *cp;
123 unsigned int ex_len;
125 if (verbose > 2) {
126 rprintf(FINFO, "[%s] add_rule(%s%.*s%s)%s\n",
127 who_am_i(), get_rule_prefix(mflags, pat, 0, NULL),
128 (int)pat_len, pat,
129 (mflags & MATCHFLG_DIRECTORY) ? "/" : "",
130 listp->debug_type);
133 /* These flags also indicate that we're reading a list that
134 * needs to be filtered now, not post-filtered later. */
135 if (xflags & (XFLG_ANCHORED2ABS|XFLG_ABS_IF_SLASH)) {
136 uint32 mf = mflags & (MATCHFLG_RECEIVER_SIDE|MATCHFLG_SENDER_SIDE);
137 if (am_sender) {
138 if (mf == MATCHFLG_RECEIVER_SIDE)
139 return;
140 } else {
141 if (mf == MATCHFLG_SENDER_SIDE)
142 return;
146 if (!(ret = new0(struct filter_struct)))
147 out_of_memory("add_rule");
149 if (!(mflags & (MATCHFLG_ABS_PATH | MATCHFLG_MERGE_FILE))
150 && ((xflags & (XFLG_ANCHORED2ABS|XFLG_ABS_IF_SLASH) && *pat == '/')
151 || (xflags & XFLG_ABS_IF_SLASH && strchr(pat, '/') != NULL))) {
152 mflags |= MATCHFLG_ABS_PATH;
153 if (*pat == '/')
154 ex_len = dirbuf_len - module_dirlen - 1;
155 else
156 ex_len = 0;
157 } else
158 ex_len = 0;
159 if (!(ret->pattern = new_array(char, ex_len + pat_len + 1)))
160 out_of_memory("add_rule");
161 if (ex_len)
162 memcpy(ret->pattern, dirbuf + module_dirlen, ex_len);
163 strlcpy(ret->pattern + ex_len, pat, pat_len + 1);
164 pat_len += ex_len;
166 if (strpbrk(ret->pattern, "*[?")) {
167 mflags |= MATCHFLG_WILD;
168 if ((cp = strstr(ret->pattern, "**")) != NULL) {
169 mflags |= MATCHFLG_WILD2;
170 /* If the pattern starts with **, note that. */
171 if (cp == ret->pattern)
172 mflags |= MATCHFLG_WILD2_PREFIX;
173 /* If the pattern ends with ***, note that. */
174 if (pat_len >= 3
175 && ret->pattern[pat_len-3] == '*'
176 && ret->pattern[pat_len-2] == '*'
177 && ret->pattern[pat_len-1] == '*')
178 mflags |= MATCHFLG_WILD3_SUFFIX;
182 if (pat_len > 1 && ret->pattern[pat_len-1] == '/') {
183 ret->pattern[pat_len-1] = 0;
184 mflags |= MATCHFLG_DIRECTORY;
187 if (mflags & MATCHFLG_PERDIR_MERGE) {
188 struct filter_list_struct *lp;
189 unsigned int len;
190 int i;
192 if ((cp = strrchr(ret->pattern, '/')) != NULL)
193 cp++;
194 else
195 cp = ret->pattern;
197 /* If the local merge file was already mentioned, don't
198 * add it again. */
199 for (i = 0; i < mergelist_cnt; i++) {
200 struct filter_struct *ex = mergelist_parents[i];
201 const char *s = strrchr(ex->pattern, '/');
202 if (s)
203 s++;
204 else
205 s = ex->pattern;
206 len = strlen(s);
207 if (len == pat_len - (cp - ret->pattern)
208 && memcmp(s, cp, len) == 0) {
209 free_filter(ret);
210 return;
214 if (!(lp = new_array(struct filter_list_struct, 1)))
215 out_of_memory("add_rule");
216 lp->head = lp->tail = NULL;
217 if (asprintf(&lp->debug_type, " [per-dir %s]", cp) < 0)
218 out_of_memory("add_rule");
219 ret->u.mergelist = lp;
221 if (mergelist_cnt == mergelist_size) {
222 mergelist_size += 5;
223 mergelist_parents = realloc_array(mergelist_parents,
224 struct filter_struct *,
225 mergelist_size);
226 if (!mergelist_parents)
227 out_of_memory("add_rule");
229 mergelist_parents[mergelist_cnt++] = ret;
230 } else {
231 for (cp = ret->pattern; (cp = strchr(cp, '/')) != NULL; cp++)
232 ret->u.slash_cnt++;
235 ret->match_flags = mflags;
237 if (!listp->tail) {
238 ret->next = listp->head;
239 listp->head = listp->tail = ret;
240 } else {
241 ret->next = listp->tail->next;
242 listp->tail->next = ret;
243 listp->tail = ret;
247 static void clear_filter_list(struct filter_list_struct *listp)
249 if (listp->tail) {
250 struct filter_struct *ent, *next;
251 /* Truncate any inherited items from the local list. */
252 listp->tail->next = NULL;
253 /* Now free everything that is left. */
254 for (ent = listp->head; ent; ent = next) {
255 next = ent->next;
256 free_filter(ent);
260 listp->head = listp->tail = NULL;
263 /* This returns an expanded (absolute) filename for the merge-file name if
264 * the name has any slashes in it OR if the parent_dirscan var is True;
265 * otherwise it returns the original merge_file name. If the len_ptr value
266 * is non-NULL the merge_file name is limited by the referenced length
267 * value and will be updated with the length of the resulting name. We
268 * always return a name that is null terminated, even if the merge_file
269 * name was not. */
270 static char *parse_merge_name(const char *merge_file, unsigned int *len_ptr,
271 unsigned int prefix_skip)
273 static char buf[MAXPATHLEN];
274 char *fn, tmpbuf[MAXPATHLEN];
275 unsigned int fn_len;
277 if (!parent_dirscan && *merge_file != '/') {
278 /* Return the name unchanged it doesn't have any slashes. */
279 if (len_ptr) {
280 const char *p = merge_file + *len_ptr;
281 while (--p > merge_file && *p != '/') {}
282 if (p == merge_file) {
283 strlcpy(buf, merge_file, *len_ptr + 1);
284 return buf;
286 } else if (strchr(merge_file, '/') == NULL)
287 return (char *)merge_file;
290 fn = *merge_file == '/' ? buf : tmpbuf;
291 if (sanitize_paths) {
292 const char *r = prefix_skip ? "/" : NULL;
293 /* null-terminate the name if it isn't already */
294 if (len_ptr && merge_file[*len_ptr]) {
295 char *to = fn == buf ? tmpbuf : buf;
296 strlcpy(to, merge_file, *len_ptr + 1);
297 merge_file = to;
299 if (!sanitize_path(fn, merge_file, r, dirbuf_depth, NULL)) {
300 rprintf(FERROR, "merge-file name overflows: %s\n",
301 merge_file);
302 return NULL;
304 fn_len = strlen(fn);
305 } else {
306 strlcpy(fn, merge_file, len_ptr ? *len_ptr + 1 : MAXPATHLEN);
307 fn_len = clean_fname(fn, 1);
310 /* If the name isn't in buf yet, it's wasn't absolute. */
311 if (fn != buf) {
312 if (dirbuf_len + fn_len >= MAXPATHLEN) {
313 rprintf(FERROR, "merge-file name overflows: %s\n", fn);
314 return NULL;
316 memcpy(buf, dirbuf + prefix_skip, dirbuf_len - prefix_skip);
317 memcpy(buf + dirbuf_len - prefix_skip, fn, fn_len + 1);
318 fn_len = clean_fname(buf, 1);
321 if (len_ptr)
322 *len_ptr = fn_len;
323 return buf;
326 /* Sets the dirbuf and dirbuf_len values. */
327 void set_filter_dir(const char *dir, unsigned int dirlen)
329 unsigned int len;
330 if (*dir != '/') {
331 memcpy(dirbuf, curr_dir, curr_dir_len);
332 dirbuf[curr_dir_len] = '/';
333 len = curr_dir_len + 1;
334 if (len + dirlen >= MAXPATHLEN)
335 dirlen = 0;
336 } else
337 len = 0;
338 memcpy(dirbuf + len, dir, dirlen);
339 dirbuf[dirlen + len] = '\0';
340 dirbuf_len = clean_fname(dirbuf, 1);
341 if (dirbuf_len > 1 && dirbuf[dirbuf_len-1] == '.'
342 && dirbuf[dirbuf_len-2] == '/')
343 dirbuf_len -= 2;
344 if (dirbuf_len != 1)
345 dirbuf[dirbuf_len++] = '/';
346 dirbuf[dirbuf_len] = '\0';
347 if (sanitize_paths)
348 dirbuf_depth = count_dir_elements(dirbuf + module_dirlen);
351 /* This routine takes a per-dir merge-file entry and finishes its setup.
352 * If the name has a path portion then we check to see if it refers to a
353 * parent directory of the first transfer dir. If it does, we scan all the
354 * dirs from that point through the parent dir of the transfer dir looking
355 * for the per-dir merge-file in each one. */
356 static BOOL setup_merge_file(struct filter_struct *ex,
357 struct filter_list_struct *lp)
359 char buf[MAXPATHLEN];
360 char *x, *y, *pat = ex->pattern;
361 unsigned int len;
363 if (!(x = parse_merge_name(pat, NULL, 0)) || *x != '/')
364 return 0;
366 y = strrchr(x, '/');
367 *y = '\0';
368 ex->pattern = strdup(y+1);
369 if (!*x)
370 x = "/";
371 if (*x == '/')
372 strlcpy(buf, x, MAXPATHLEN);
373 else
374 pathjoin(buf, MAXPATHLEN, dirbuf, x);
376 len = clean_fname(buf, 1);
377 if (len != 1 && len < MAXPATHLEN-1) {
378 buf[len++] = '/';
379 buf[len] = '\0';
381 /* This ensures that the specified dir is a parent of the transfer. */
382 for (x = buf, y = dirbuf; *x && *x == *y; x++, y++) {}
383 if (*x)
384 y += strlen(y); /* nope -- skip the scan */
386 parent_dirscan = True;
387 while (*y) {
388 char save[MAXPATHLEN];
389 strlcpy(save, y, MAXPATHLEN);
390 *y = '\0';
391 dirbuf_len = y - dirbuf;
392 strlcpy(x, ex->pattern, MAXPATHLEN - (x - buf));
393 parse_filter_file(lp, buf, ex->match_flags, XFLG_ANCHORED2ABS);
394 if (ex->match_flags & MATCHFLG_NO_INHERIT)
395 lp->head = NULL;
396 lp->tail = NULL;
397 strlcpy(y, save, MAXPATHLEN);
398 while ((*x++ = *y++) != '/') {}
400 parent_dirscan = False;
401 free(pat);
402 return 1;
405 /* Each time rsync changes to a new directory it call this function to
406 * handle all the per-dir merge-files. The "dir" value is the current path
407 * relative to curr_dir (which might not be null-terminated). We copy it
408 * into dirbuf so that we can easily append a file name on the end. */
409 void *push_local_filters(const char *dir, unsigned int dirlen)
411 struct filter_list_struct *ap, *push;
412 int i;
414 set_filter_dir(dir, dirlen);
416 if (!mergelist_cnt)
417 return NULL;
419 push = new_array(struct filter_list_struct, mergelist_cnt);
420 if (!push)
421 out_of_memory("push_local_filters");
423 for (i = 0, ap = push; i < mergelist_cnt; i++) {
424 memcpy(ap++, mergelist_parents[i]->u.mergelist,
425 sizeof (struct filter_list_struct));
428 /* Note: parse_filter_file() might increase mergelist_cnt, so keep
429 * this loop separate from the above loop. */
430 for (i = 0; i < mergelist_cnt; i++) {
431 struct filter_struct *ex = mergelist_parents[i];
432 struct filter_list_struct *lp = ex->u.mergelist;
434 if (verbose > 2) {
435 rprintf(FINFO, "[%s] pushing filter list%s\n",
436 who_am_i(), lp->debug_type);
439 lp->tail = NULL; /* Switch any local rules to inherited. */
440 if (ex->match_flags & MATCHFLG_NO_INHERIT)
441 lp->head = NULL;
443 if (ex->match_flags & MATCHFLG_FINISH_SETUP) {
444 ex->match_flags &= ~MATCHFLG_FINISH_SETUP;
445 if (setup_merge_file(ex, lp))
446 set_filter_dir(dir, dirlen);
449 if (strlcpy(dirbuf + dirbuf_len, ex->pattern,
450 MAXPATHLEN - dirbuf_len) < MAXPATHLEN - dirbuf_len) {
451 parse_filter_file(lp, dirbuf, ex->match_flags,
452 XFLG_ANCHORED2ABS);
453 } else {
454 io_error |= IOERR_GENERAL;
455 rprintf(FINFO,
456 "cannot add local filter rules in long-named directory: %s\n",
457 full_fname(dirbuf));
459 dirbuf[dirbuf_len] = '\0';
462 return (void*)push;
465 void pop_local_filters(void *mem)
467 struct filter_list_struct *ap, *pop = (struct filter_list_struct*)mem;
468 int i;
470 for (i = mergelist_cnt; i-- > 0; ) {
471 struct filter_struct *ex = mergelist_parents[i];
472 struct filter_list_struct *lp = ex->u.mergelist;
474 if (verbose > 2) {
475 rprintf(FINFO, "[%s] popping filter list%s\n",
476 who_am_i(), lp->debug_type);
479 clear_filter_list(lp);
482 if (!pop)
483 return;
485 for (i = 0, ap = pop; i < mergelist_cnt; i++) {
486 memcpy(mergelist_parents[i]->u.mergelist, ap++,
487 sizeof (struct filter_list_struct));
490 free(pop);
493 void change_local_filter_dir(const char *dname, int dlen, int dir_depth)
495 static int cur_depth = -1;
496 static void *filt_array[MAXPATHLEN/2+1];
498 if (!dname) {
499 for ( ; cur_depth >= 0; cur_depth--) {
500 if (filt_array[cur_depth]) {
501 pop_local_filters(filt_array[cur_depth]);
502 filt_array[cur_depth] = NULL;
505 return;
508 assert(dir_depth < MAXPATHLEN/2+1);
510 for ( ; cur_depth >= dir_depth; cur_depth--) {
511 if (filt_array[cur_depth]) {
512 pop_local_filters(filt_array[cur_depth]);
513 filt_array[cur_depth] = NULL;
517 cur_depth = dir_depth;
518 filt_array[cur_depth] = push_local_filters(dname, dlen);
521 static int rule_matches(char *fname, struct filter_struct *ex, int name_is_dir)
523 int slash_handling, str_cnt = 0, anchored_match = 0;
524 int ret_match = ex->match_flags & MATCHFLG_NEGATE ? 0 : 1;
525 char *p, *pattern = ex->pattern;
526 const char *strings[16]; /* more than enough */
527 char *name = fname + (*fname == '/');
529 if (!*name)
530 return 0;
532 if (!ex->u.slash_cnt && !(ex->match_flags & MATCHFLG_WILD2)) {
533 /* If the pattern does not have any slashes AND it does
534 * not have a "**" (which could match a slash), then we
535 * just match the name portion of the path. */
536 if ((p = strrchr(name,'/')) != NULL)
537 name = p+1;
538 } else if (ex->match_flags & MATCHFLG_ABS_PATH && *fname != '/'
539 && curr_dir_len > module_dirlen + 1) {
540 /* If we're matching against an absolute-path pattern,
541 * we need to prepend our full path info. */
542 strings[str_cnt++] = curr_dir + module_dirlen + 1;
543 strings[str_cnt++] = "/";
544 } else if (ex->match_flags & MATCHFLG_WILD2_PREFIX && *fname != '/') {
545 /* Allow "**"+"/" to match at the start of the string. */
546 strings[str_cnt++] = "/";
548 strings[str_cnt++] = name;
549 if (name_is_dir) {
550 /* Allow a trailing "/"+"***" to match the directory. */
551 if (ex->match_flags & MATCHFLG_WILD3_SUFFIX)
552 strings[str_cnt++] = "/";
553 } else if (ex->match_flags & MATCHFLG_DIRECTORY)
554 return !ret_match;
555 strings[str_cnt] = NULL;
557 if (*pattern == '/') {
558 anchored_match = 1;
559 pattern++;
562 if (!anchored_match && ex->u.slash_cnt
563 && !(ex->match_flags & MATCHFLG_WILD2)) {
564 /* A non-anchored match with an infix slash and no "**"
565 * needs to match the last slash_cnt+1 name elements. */
566 slash_handling = ex->u.slash_cnt + 1;
567 } else if (!anchored_match && !(ex->match_flags & MATCHFLG_WILD2_PREFIX)
568 && ex->match_flags & MATCHFLG_WILD2) {
569 /* A non-anchored match with an infix or trailing "**" (but not
570 * a prefixed "**") needs to try matching after every slash. */
571 slash_handling = -1;
572 } else {
573 /* The pattern matches only at the start of the path or name. */
574 slash_handling = 0;
577 if (ex->match_flags & MATCHFLG_WILD) {
578 if (wildmatch_array(pattern, strings, slash_handling))
579 return ret_match;
580 } else if (str_cnt > 1) {
581 if (litmatch_array(pattern, strings, slash_handling))
582 return ret_match;
583 } else if (anchored_match) {
584 if (strcmp(name, pattern) == 0)
585 return ret_match;
586 } else {
587 int l1 = strlen(name);
588 int l2 = strlen(pattern);
589 if (l2 <= l1 &&
590 strcmp(name+(l1-l2),pattern) == 0 &&
591 (l1==l2 || name[l1-(l2+1)] == '/')) {
592 return ret_match;
596 return !ret_match;
600 static void report_filter_result(char const *name,
601 struct filter_struct const *ent,
602 int name_is_dir, const char *type)
604 /* If a trailing slash is present to match only directories,
605 * then it is stripped out by add_rule(). So as a special
606 * case we add it back in here. */
608 if (verbose >= 2) {
609 static char *actions[2][2]
610 = { {"show", "hid"}, {"risk", "protect"} };
611 const char *w = who_am_i();
612 rprintf(FINFO, "[%s] %sing %s %s because of pattern %s%s%s\n",
613 w, actions[*w!='s'][!(ent->match_flags&MATCHFLG_INCLUDE)],
614 name_is_dir ? "directory" : "file", name, ent->pattern,
615 ent->match_flags & MATCHFLG_DIRECTORY ? "/" : "", type);
621 * Return -1 if file "name" is defined to be excluded by the specified
622 * exclude list, 1 if it is included, and 0 if it was not matched.
624 int check_filter(struct filter_list_struct *listp, char *name, int name_is_dir)
626 struct filter_struct *ent;
628 for (ent = listp->head; ent; ent = ent->next) {
629 if (ignore_perishable && ent->match_flags & MATCHFLG_PERISHABLE)
630 continue;
631 if (ent->match_flags & MATCHFLG_PERDIR_MERGE) {
632 int rc = check_filter(ent->u.mergelist, name,
633 name_is_dir);
634 if (rc)
635 return rc;
636 continue;
638 if (ent->match_flags & MATCHFLG_CVS_IGNORE) {
639 int rc = check_filter(&cvs_filter_list, name,
640 name_is_dir);
641 if (rc)
642 return rc;
643 continue;
645 if (rule_matches(name, ent, name_is_dir)) {
646 report_filter_result(name, ent, name_is_dir,
647 listp->debug_type);
648 return ent->match_flags & MATCHFLG_INCLUDE ? 1 : -1;
652 return 0;
655 #define RULE_STRCMP(s,r) rule_strcmp((s), (r), sizeof (r) - 1)
657 static const uchar *rule_strcmp(const uchar *str, const char *rule, int rule_len)
659 if (strncmp((char*)str, rule, rule_len) != 0)
660 return NULL;
661 if (isspace(str[rule_len]) || str[rule_len] == '_' || !str[rule_len])
662 return str + rule_len - 1;
663 if (str[rule_len] == ',')
664 return str + rule_len;
665 return NULL;
668 /* Get the next include/exclude arg from the string. The token will not
669 * be '\0' terminated, so use the returned length to limit the string.
670 * Also, be sure to add this length to the returned pointer before passing
671 * it back to ask for the next token. This routine parses the "!" (list-
672 * clearing) token and (depending on the mflags) the various prefixes.
673 * The *mflags_ptr value will be set on exit to the new MATCHFLG_* bits
674 * for the current token. */
675 static const char *parse_rule_tok(const char *p, uint32 mflags, int xflags,
676 unsigned int *len_ptr, uint32 *mflags_ptr)
678 const uchar *s = (const uchar *)p;
679 uint32 new_mflags;
680 unsigned int len;
682 if (mflags & MATCHFLG_WORD_SPLIT) {
683 /* Skip over any initial whitespace. */
684 while (isspace(*s))
685 s++;
686 /* Update to point to real start of rule. */
687 p = (const char *)s;
689 if (!*s)
690 return NULL;
692 new_mflags = mflags & MATCHFLGS_FROM_CONTAINER;
694 /* Figure out what kind of a filter rule "s" is pointing at. Note
695 * that if MATCHFLG_NO_PREFIXES is set, the rule is either an include
696 * or an exclude based on the inheritance of the MATCHFLG_INCLUDE
697 * flag (above). XFLG_OLD_PREFIXES indicates a compatibility mode
698 * for old include/exclude patterns where just "+ " and "- " are
699 * allowed as optional prefixes. */
700 if (mflags & MATCHFLG_NO_PREFIXES) {
701 if (*s == '!' && mflags & MATCHFLG_CVS_IGNORE)
702 new_mflags |= MATCHFLG_CLEAR_LIST; /* Tentative! */
703 } else if (xflags & XFLG_OLD_PREFIXES) {
704 if (*s == '-' && s[1] == ' ') {
705 new_mflags &= ~MATCHFLG_INCLUDE;
706 s += 2;
707 } else if (*s == '+' && s[1] == ' ') {
708 new_mflags |= MATCHFLG_INCLUDE;
709 s += 2;
710 } else if (*s == '!')
711 new_mflags |= MATCHFLG_CLEAR_LIST; /* Tentative! */
712 } else {
713 char ch = 0, *mods = "";
714 switch (*s) {
715 case 'c':
716 if ((s = RULE_STRCMP(s, "clear")) != NULL)
717 ch = '!';
718 break;
719 case 'd':
720 if ((s = RULE_STRCMP(s, "dir-merge")) != NULL)
721 ch = ':';
722 break;
723 case 'e':
724 if ((s = RULE_STRCMP(s, "exclude")) != NULL)
725 ch = '-';
726 break;
727 case 'h':
728 if ((s = RULE_STRCMP(s, "hide")) != NULL)
729 ch = 'H';
730 break;
731 case 'i':
732 if ((s = RULE_STRCMP(s, "include")) != NULL)
733 ch = '+';
734 break;
735 case 'm':
736 if ((s = RULE_STRCMP(s, "merge")) != NULL)
737 ch = '.';
738 break;
739 case 'p':
740 if ((s = RULE_STRCMP(s, "protect")) != NULL)
741 ch = 'P';
742 break;
743 case 'r':
744 if ((s = RULE_STRCMP(s, "risk")) != NULL)
745 ch = 'R';
746 break;
747 case 's':
748 if ((s = RULE_STRCMP(s, "show")) != NULL)
749 ch = 'S';
750 break;
751 default:
752 ch = *s;
753 if (s[1] == ',')
754 s++;
755 break;
757 switch (ch) {
758 case ':':
759 new_mflags |= MATCHFLG_PERDIR_MERGE
760 | MATCHFLG_FINISH_SETUP;
761 /* FALL THROUGH */
762 case '.':
763 new_mflags |= MATCHFLG_MERGE_FILE;
764 mods = MODIFIERS_INCL_EXCL MODIFIERS_MERGE_FILE;
765 break;
766 case '+':
767 new_mflags |= MATCHFLG_INCLUDE;
768 /* FALL THROUGH */
769 case '-':
770 mods = MODIFIERS_INCL_EXCL;
771 break;
772 case 'S':
773 new_mflags |= MATCHFLG_INCLUDE;
774 /* FALL THROUGH */
775 case 'H':
776 new_mflags |= MATCHFLG_SENDER_SIDE;
777 mods = MODIFIERS_HIDE_PROTECT;
778 break;
779 case 'R':
780 new_mflags |= MATCHFLG_INCLUDE;
781 /* FALL THROUGH */
782 case 'P':
783 new_mflags |= MATCHFLG_RECEIVER_SIDE;
784 mods = MODIFIERS_HIDE_PROTECT;
785 break;
786 case '!':
787 new_mflags |= MATCHFLG_CLEAR_LIST;
788 mods = NULL;
789 break;
790 default:
791 rprintf(FERROR, "Unknown filter rule: `%s'\n", p);
792 exit_cleanup(RERR_SYNTAX);
794 while (mods && *++s && *s != ' ' && *s != '_') {
795 if (strchr(mods, *s) == NULL) {
796 if (mflags & MATCHFLG_WORD_SPLIT && isspace(*s)) {
797 s--;
798 break;
800 invalid:
801 rprintf(FERROR,
802 "invalid modifier sequence at '%c' in filter rule: %s\n",
803 *s, p);
804 exit_cleanup(RERR_SYNTAX);
806 switch (*s) {
807 case '-':
808 if (new_mflags & MATCHFLG_NO_PREFIXES)
809 goto invalid;
810 new_mflags |= MATCHFLG_NO_PREFIXES;
811 break;
812 case '+':
813 if (new_mflags & MATCHFLG_NO_PREFIXES)
814 goto invalid;
815 new_mflags |= MATCHFLG_NO_PREFIXES
816 | MATCHFLG_INCLUDE;
817 break;
818 case '/':
819 new_mflags |= MATCHFLG_ABS_PATH;
820 break;
821 case '!':
822 new_mflags |= MATCHFLG_NEGATE;
823 break;
824 case 'C':
825 if (new_mflags & MATCHFLG_NO_PREFIXES)
826 goto invalid;
827 new_mflags |= MATCHFLG_NO_PREFIXES
828 | MATCHFLG_WORD_SPLIT
829 | MATCHFLG_NO_INHERIT
830 | MATCHFLG_CVS_IGNORE;
831 break;
832 case 'e':
833 new_mflags |= MATCHFLG_EXCLUDE_SELF;
834 break;
835 case 'n':
836 new_mflags |= MATCHFLG_NO_INHERIT;
837 break;
838 case 'p':
839 new_mflags |= MATCHFLG_PERISHABLE;
840 break;
841 case 'r':
842 new_mflags |= MATCHFLG_RECEIVER_SIDE;
843 break;
844 case 's':
845 new_mflags |= MATCHFLG_SENDER_SIDE;
846 break;
847 case 'w':
848 new_mflags |= MATCHFLG_WORD_SPLIT;
849 break;
852 if (*s)
853 s++;
856 if (mflags & MATCHFLG_WORD_SPLIT) {
857 const uchar *cp = s;
858 /* Token ends at whitespace or the end of the string. */
859 while (!isspace(*cp) && *cp != '\0')
860 cp++;
861 len = cp - s;
862 } else
863 len = strlen((char*)s);
865 if (new_mflags & MATCHFLG_CLEAR_LIST) {
866 if (!(mflags & MATCHFLG_NO_PREFIXES)
867 && !(xflags & XFLG_OLD_PREFIXES) && len) {
868 rprintf(FERROR,
869 "'!' rule has trailing characters: %s\n", p);
870 exit_cleanup(RERR_SYNTAX);
872 if (len > 1)
873 new_mflags &= ~MATCHFLG_CLEAR_LIST;
874 } else if (!len && !(new_mflags & MATCHFLG_CVS_IGNORE)) {
875 rprintf(FERROR, "unexpected end of filter rule: %s\n", p);
876 exit_cleanup(RERR_SYNTAX);
879 /* --delete-excluded turns an un-modified include/exclude into a
880 * sender-side rule. We also affect per-dir merge files that take
881 * no prefixes as a simple optimization. */
882 if (delete_excluded
883 && !(new_mflags & (MATCHFLG_RECEIVER_SIDE|MATCHFLG_SENDER_SIDE))
884 && (!(new_mflags & MATCHFLG_PERDIR_MERGE)
885 || new_mflags & MATCHFLG_NO_PREFIXES))
886 new_mflags |= MATCHFLG_SENDER_SIDE;
888 *len_ptr = len;
889 *mflags_ptr = new_mflags;
890 return (const char *)s;
894 static char default_cvsignore[] =
895 /* These default ignored items come from the CVS manual. */
896 "RCS SCCS CVS CVS.adm RCSLOG cvslog.* tags TAGS"
897 " .make.state .nse_depinfo *~ #* .#* ,* _$* *$"
898 " *.old *.bak *.BAK *.orig *.rej .del-*"
899 " *.a *.olb *.o *.obj *.so *.exe"
900 " *.Z *.elc *.ln core"
901 /* The rest we added to suit ourself. */
902 " .svn/ .git/ .bzr/";
904 static void get_cvs_excludes(uint32 mflags)
906 static int initialized = 0;
907 char *p, fname[MAXPATHLEN];
909 if (initialized)
910 return;
911 initialized = 1;
913 parse_rule(&cvs_filter_list, default_cvsignore,
914 mflags | (protocol_version >= 30 ? MATCHFLG_PERISHABLE : 0),
917 p = module_id >= 0 && lp_use_chroot(module_id) ? "/" : getenv("HOME");
918 if (p && pathjoin(fname, MAXPATHLEN, p, ".cvsignore") < MAXPATHLEN)
919 parse_filter_file(&cvs_filter_list, fname, mflags, 0);
921 parse_rule(&cvs_filter_list, getenv("CVSIGNORE"), mflags, 0);
925 void parse_rule(struct filter_list_struct *listp, const char *pattern,
926 uint32 mflags, int xflags)
928 unsigned int pat_len;
929 uint32 new_mflags;
930 const char *cp, *p;
932 if (!pattern)
933 return;
935 while (1) {
936 /* Remember that the returned string is NOT '\0' terminated! */
937 cp = parse_rule_tok(pattern, mflags, xflags,
938 &pat_len, &new_mflags);
939 if (!cp)
940 break;
942 pattern = cp + pat_len;
944 if (pat_len >= MAXPATHLEN) {
945 rprintf(FERROR, "discarding over-long filter: %.*s\n",
946 (int)pat_len, cp);
947 continue;
950 if (new_mflags & MATCHFLG_CLEAR_LIST) {
951 if (verbose > 2) {
952 rprintf(FINFO,
953 "[%s] clearing filter list%s\n",
954 who_am_i(), listp->debug_type);
956 clear_filter_list(listp);
957 continue;
960 if (new_mflags & MATCHFLG_MERGE_FILE) {
961 unsigned int len;
962 if (!pat_len) {
963 cp = ".cvsignore";
964 pat_len = 10;
966 len = pat_len;
967 if (new_mflags & MATCHFLG_EXCLUDE_SELF) {
968 const char *name = cp + len;
969 while (name > cp && name[-1] != '/') name--;
970 len -= name - cp;
971 add_rule(listp, name, len, 0, 0);
972 new_mflags &= ~MATCHFLG_EXCLUDE_SELF;
973 len = pat_len;
975 if (new_mflags & MATCHFLG_PERDIR_MERGE) {
976 if (parent_dirscan) {
977 if (!(p = parse_merge_name(cp, &len,
978 module_dirlen)))
979 continue;
980 add_rule(listp, p, len, new_mflags, 0);
981 continue;
983 } else {
984 if (!(p = parse_merge_name(cp, &len, 0)))
985 continue;
986 parse_filter_file(listp, p, new_mflags,
987 XFLG_FATAL_ERRORS);
988 continue;
992 add_rule(listp, cp, pat_len, new_mflags, xflags);
994 if (new_mflags & MATCHFLG_CVS_IGNORE
995 && !(new_mflags & MATCHFLG_MERGE_FILE))
996 get_cvs_excludes(new_mflags);
1001 void parse_filter_file(struct filter_list_struct *listp, const char *fname,
1002 uint32 mflags, int xflags)
1004 FILE *fp;
1005 char line[BIGPATHBUFLEN];
1006 char *eob = line + sizeof line - 1;
1007 int word_split = mflags & MATCHFLG_WORD_SPLIT;
1009 if (!fname || !*fname)
1010 return;
1012 if (*fname != '-' || fname[1] || am_server) {
1013 if (server_filter_list.head) {
1014 strlcpy(line, fname, sizeof line);
1015 clean_fname(line, 1);
1016 if (check_filter(&server_filter_list, line, 0) < 0)
1017 fp = NULL;
1018 else
1019 fp = fopen(line, "rb");
1020 } else
1021 fp = fopen(fname, "rb");
1022 } else
1023 fp = stdin;
1025 if (verbose > 2) {
1026 rprintf(FINFO, "[%s] parse_filter_file(%s,%x,%x)%s\n",
1027 who_am_i(), fname, mflags, xflags,
1028 fp ? "" : " [not found]");
1031 if (!fp) {
1032 if (xflags & XFLG_FATAL_ERRORS) {
1033 rsyserr(FERROR, errno,
1034 "failed to open %sclude file %s",
1035 mflags & MATCHFLG_INCLUDE ? "in" : "ex",
1036 fname);
1037 exit_cleanup(RERR_FILEIO);
1039 return;
1041 dirbuf[dirbuf_len] = '\0';
1043 while (1) {
1044 char *s = line;
1045 int ch, overflow = 0;
1046 while (1) {
1047 if ((ch = getc(fp)) == EOF) {
1048 if (ferror(fp) && errno == EINTR) {
1049 clearerr(fp);
1050 continue;
1052 break;
1054 if (word_split && isspace(ch))
1055 break;
1056 if (eol_nulls? !ch : (ch == '\n' || ch == '\r'))
1057 break;
1058 if (s < eob)
1059 *s++ = ch;
1060 else
1061 overflow = 1;
1063 if (overflow) {
1064 rprintf(FERROR, "discarding over-long filter: %s...\n", line);
1065 s = line;
1067 *s = '\0';
1068 /* Skip an empty token and (when line parsing) comments. */
1069 if (*line && (word_split || (*line != ';' && *line != '#')))
1070 parse_rule(listp, line, mflags, xflags);
1071 if (ch == EOF)
1072 break;
1074 fclose(fp);
1077 /* If the "for_xfer" flag is set, the prefix is made compatible with the
1078 * current protocol_version (if possible) or a NULL is returned (if not
1079 * possible). */
1080 char *get_rule_prefix(int match_flags, const char *pat, int for_xfer,
1081 unsigned int *plen_ptr)
1083 static char buf[MAX_RULE_PREFIX+1];
1084 char *op = buf;
1085 int legal_len = for_xfer && protocol_version < 29 ? 1 : MAX_RULE_PREFIX-1;
1087 if (match_flags & MATCHFLG_PERDIR_MERGE) {
1088 if (legal_len == 1)
1089 return NULL;
1090 *op++ = ':';
1091 } else if (match_flags & MATCHFLG_INCLUDE)
1092 *op++ = '+';
1093 else if (legal_len != 1
1094 || ((*pat == '-' || *pat == '+') && pat[1] == ' '))
1095 *op++ = '-';
1096 else
1097 legal_len = 0;
1099 if (match_flags & MATCHFLG_NEGATE)
1100 *op++ = '!';
1101 if (match_flags & MATCHFLG_CVS_IGNORE)
1102 *op++ = 'C';
1103 else {
1104 if (match_flags & MATCHFLG_NO_INHERIT)
1105 *op++ = 'n';
1106 if (match_flags & MATCHFLG_WORD_SPLIT)
1107 *op++ = 'w';
1108 if (match_flags & MATCHFLG_NO_PREFIXES) {
1109 if (match_flags & MATCHFLG_INCLUDE)
1110 *op++ = '+';
1111 else
1112 *op++ = '-';
1115 if (match_flags & MATCHFLG_EXCLUDE_SELF)
1116 *op++ = 'e';
1117 if (match_flags & MATCHFLG_SENDER_SIDE
1118 && (!for_xfer || protocol_version >= 29))
1119 *op++ = 's';
1120 if (match_flags & MATCHFLG_RECEIVER_SIDE
1121 && (!for_xfer || protocol_version >= 29
1122 || (delete_excluded && am_sender)))
1123 *op++ = 'r';
1124 if (match_flags & MATCHFLG_PERISHABLE) {
1125 if (!for_xfer || protocol_version >= 30)
1126 *op++ = 'p';
1127 else if (am_sender)
1128 return NULL;
1130 if (op - buf > legal_len)
1131 return NULL;
1132 if (legal_len)
1133 *op++ = ' ';
1134 *op = '\0';
1135 if (plen_ptr)
1136 *plen_ptr = op - buf;
1137 return buf;
1140 static void send_rules(int f_out, struct filter_list_struct *flp)
1142 struct filter_struct *ent, *prev = NULL;
1144 for (ent = flp->head; ent; ent = ent->next) {
1145 unsigned int len, plen, dlen;
1146 int elide = 0;
1147 char *p;
1149 /* Note we need to check delete_excluded here in addition to
1150 * the code in parse_rule_tok() because some rules may have
1151 * been added before we found the --delete-excluded option.
1152 * We must also elide any CVS merge-file rules to avoid a
1153 * backward compatibility problem, and we elide any no-prefix
1154 * merge files as an optimization (since they can only have
1155 * include/exclude rules). */
1156 if (ent->match_flags & MATCHFLG_SENDER_SIDE)
1157 elide = am_sender ? 1 : -1;
1158 if (ent->match_flags & MATCHFLG_RECEIVER_SIDE)
1159 elide = elide ? 0 : am_sender ? -1 : 1;
1160 else if (delete_excluded && !elide
1161 && (!(ent->match_flags & MATCHFLG_PERDIR_MERGE)
1162 || ent->match_flags & MATCHFLG_NO_PREFIXES))
1163 elide = am_sender ? 1 : -1;
1164 if (elide < 0) {
1165 if (prev)
1166 prev->next = ent->next;
1167 else
1168 flp->head = ent->next;
1169 } else
1170 prev = ent;
1171 if (elide > 0)
1172 continue;
1173 if (ent->match_flags & MATCHFLG_CVS_IGNORE
1174 && !(ent->match_flags & MATCHFLG_MERGE_FILE)) {
1175 int f = am_sender || protocol_version < 29 ? f_out : -2;
1176 send_rules(f, &cvs_filter_list);
1177 if (f == f_out)
1178 continue;
1180 p = get_rule_prefix(ent->match_flags, ent->pattern, 1, &plen);
1181 if (!p) {
1182 rprintf(FERROR,
1183 "filter rules are too modern for remote rsync.\n");
1184 exit_cleanup(RERR_PROTOCOL);
1186 if (f_out < 0)
1187 continue;
1188 len = strlen(ent->pattern);
1189 dlen = ent->match_flags & MATCHFLG_DIRECTORY ? 1 : 0;
1190 if (!(plen + len + dlen))
1191 continue;
1192 write_int(f_out, plen + len + dlen);
1193 if (plen)
1194 write_buf(f_out, p, plen);
1195 write_buf(f_out, ent->pattern, len);
1196 if (dlen)
1197 write_byte(f_out, '/');
1199 flp->tail = prev;
1202 /* This is only called by the client. */
1203 void send_filter_list(int f_out)
1205 int receiver_wants_list = prune_empty_dirs
1206 || (delete_mode && (!delete_excluded || protocol_version >= 29));
1208 if (local_server || (am_sender && !receiver_wants_list))
1209 f_out = -1;
1210 if (cvs_exclude && am_sender) {
1211 if (protocol_version >= 29)
1212 parse_rule(&filter_list, ":C", 0, 0);
1213 parse_rule(&filter_list, "-C", 0, 0);
1216 send_rules(f_out, &filter_list);
1218 if (f_out >= 0)
1219 write_int(f_out, 0);
1221 if (cvs_exclude) {
1222 if (!am_sender || protocol_version < 29)
1223 parse_rule(&filter_list, ":C", 0, 0);
1224 if (!am_sender)
1225 parse_rule(&filter_list, "-C", 0, 0);
1229 /* This is only called by the server. */
1230 void recv_filter_list(int f_in)
1232 char line[BIGPATHBUFLEN];
1233 int xflags = protocol_version >= 29 ? 0 : XFLG_OLD_PREFIXES;
1234 int receiver_wants_list = prune_empty_dirs
1235 || (delete_mode
1236 && (!delete_excluded || protocol_version >= 29));
1237 unsigned int len;
1239 if (!local_server && (am_sender || receiver_wants_list)) {
1240 while ((len = read_int(f_in)) != 0) {
1241 if (len >= sizeof line)
1242 overflow_exit("recv_rules");
1243 read_sbuf(f_in, line, len);
1244 parse_rule(&filter_list, line, 0, xflags);
1248 if (cvs_exclude) {
1249 if (local_server || am_sender || protocol_version < 29)
1250 parse_rule(&filter_list, ":C", 0, 0);
1251 if (local_server || am_sender)
1252 parse_rule(&filter_list, "-C", 0, 0);
1255 if (local_server) /* filter out any rules that aren't for us. */
1256 send_rules(-1, &filter_list);