| 1 | /* |
| 2 | * The filter include/exclude routines. |
| 3 | * |
| 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) 2003-2009 Wayne Davison |
| 8 | * |
| 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. |
| 13 | * |
| 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. |
| 18 | * |
| 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. |
| 21 | */ |
| 22 | |
| 23 | #include "rsync.h" |
| 24 | |
| 25 | extern int am_server; |
| 26 | extern int am_sender; |
| 27 | extern int eol_nulls; |
| 28 | extern int io_error; |
| 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; |
| 37 | extern int module_id; |
| 38 | |
| 39 | extern char curr_dir[MAXPATHLEN]; |
| 40 | extern unsigned int curr_dir_len; |
| 41 | extern unsigned int module_dirlen; |
| 42 | |
| 43 | filter_rule_list filter_list = { .debug_type = "" }; |
| 44 | filter_rule_list cvs_filter_list = { .debug_type = " [global CVS]" }; |
| 45 | filter_rule_list daemon_filter_list = { .debug_type = " [daemon]" }; |
| 46 | |
| 47 | /* Need room enough for ":MODS " prefix plus some room to grow. */ |
| 48 | #define MAX_RULE_PREFIX (16) |
| 49 | |
| 50 | #define SLASH_WILD3_SUFFIX "/***" |
| 51 | |
| 52 | /* The dirbuf is set by push_local_filters() to the current subdirectory |
| 53 | * relative to curr_dir that is being processed. The path always has a |
| 54 | * trailing slash appended, and the variable dirbuf_len contains the length |
| 55 | * of this path prefix. The path is always absolute. */ |
| 56 | static char dirbuf[MAXPATHLEN+1]; |
| 57 | static unsigned int dirbuf_len = 0; |
| 58 | static int dirbuf_depth; |
| 59 | |
| 60 | /* This is True when we're scanning parent dirs for per-dir merge-files. */ |
| 61 | static BOOL parent_dirscan = False; |
| 62 | |
| 63 | /* This array contains a list of all the currently active per-dir merge |
| 64 | * files. This makes it easier to save the appropriate values when we |
| 65 | * "push" down into each subdirectory. */ |
| 66 | static filter_rule **mergelist_parents; |
| 67 | static int mergelist_cnt = 0; |
| 68 | static int mergelist_size = 0; |
| 69 | |
| 70 | /* Each filter_list_struct describes a singly-linked list by keeping track |
| 71 | * of both the head and tail pointers. The list is slightly unusual in that |
| 72 | * a parent-dir's content can be appended to the end of the local list in a |
| 73 | * special way: the last item in the local list has its "next" pointer set |
| 74 | * to point to the inherited list, but the local list's tail pointer points |
| 75 | * at the end of the local list. Thus, if the local list is empty, the head |
| 76 | * will be pointing at the inherited content but the tail will be NULL. To |
| 77 | * help you visualize this, here are the possible list arrangements: |
| 78 | * |
| 79 | * Completely Empty Local Content Only |
| 80 | * ================================== ==================================== |
| 81 | * head -> NULL head -> Local1 -> Local2 -> NULL |
| 82 | * tail -> NULL tail -------------^ |
| 83 | * |
| 84 | * Inherited Content Only Both Local and Inherited Content |
| 85 | * ================================== ==================================== |
| 86 | * head -> Parent1 -> Parent2 -> NULL head -> L1 -> L2 -> P1 -> P2 -> NULL |
| 87 | * tail -> NULL tail ---------^ |
| 88 | * |
| 89 | * This means that anyone wanting to traverse the whole list to use it just |
| 90 | * needs to start at the head and use the "next" pointers until it goes |
| 91 | * NULL. To add new local content, we insert the item after the tail item |
| 92 | * and update the tail (obviously, if "tail" was NULL, we insert it at the |
| 93 | * head). To clear the local list, WE MUST NOT FREE THE INHERITED CONTENT |
| 94 | * because it is shared between the current list and our parent list(s). |
| 95 | * The easiest way to handle this is to simply truncate the list after the |
| 96 | * tail item and then free the local list from the head. When inheriting |
| 97 | * the list for a new local dir, we just save off the filter_list_struct |
| 98 | * values (so we can pop back to them later) and set the tail to NULL. |
| 99 | */ |
| 100 | |
| 101 | static void teardown_mergelist(filter_rule *ex) |
| 102 | { |
| 103 | if (DEBUG_GTE(FILTER, 2)) { |
| 104 | rprintf(FINFO, "[%s] deactivating mergelist #%d%s\n", |
| 105 | who_am_i(), mergelist_cnt - 1, |
| 106 | ex->u.mergelist->debug_type); |
| 107 | } |
| 108 | |
| 109 | /* We should deactivate mergelists in LIFO order. */ |
| 110 | assert(mergelist_cnt > 0); |
| 111 | assert(ex == mergelist_parents[mergelist_cnt - 1]); |
| 112 | |
| 113 | /* The parent_dirscan filters should have been freed. */ |
| 114 | assert(ex->u.mergelist->parent_dirscan_head == NULL); |
| 115 | |
| 116 | free(ex->u.mergelist->debug_type); |
| 117 | free(ex->u.mergelist); |
| 118 | mergelist_cnt--; |
| 119 | } |
| 120 | |
| 121 | static void free_filter(filter_rule *ex) |
| 122 | { |
| 123 | free(ex->pattern); |
| 124 | free(ex); |
| 125 | } |
| 126 | |
| 127 | static void free_filters(filter_rule *head) |
| 128 | { |
| 129 | filter_rule *rev_head = NULL; |
| 130 | |
| 131 | /* Reverse the list so we deactivate mergelists in the proper LIFO |
| 132 | * order. */ |
| 133 | while (head) { |
| 134 | filter_rule *next = head->next; |
| 135 | head->next = rev_head; |
| 136 | rev_head = head; |
| 137 | head = next; |
| 138 | } |
| 139 | |
| 140 | while (rev_head) { |
| 141 | filter_rule *prev = rev_head->next; |
| 142 | /* Tear down mergelists here, not in free_filter, so that we |
| 143 | * affect only real filter lists and not temporarily allocated |
| 144 | * filters. */ |
| 145 | if (rev_head->rflags & FILTRULE_PERDIR_MERGE) |
| 146 | teardown_mergelist(rev_head); |
| 147 | free_filter(rev_head); |
| 148 | rev_head = prev; |
| 149 | } |
| 150 | } |
| 151 | |
| 152 | /* Build a filter structure given a filter pattern. The value in "pat" |
| 153 | * is not null-terminated. "rule" is either held or freed, so the |
| 154 | * caller should not free it. */ |
| 155 | static void add_rule(filter_rule_list *listp, const char *pat, unsigned int pat_len, |
| 156 | filter_rule *rule, int xflags) |
| 157 | { |
| 158 | const char *cp; |
| 159 | unsigned int pre_len, suf_len, slash_cnt = 0; |
| 160 | |
| 161 | if (DEBUG_GTE(FILTER, 2)) { |
| 162 | rprintf(FINFO, "[%s] add_rule(%s%.*s%s)%s\n", |
| 163 | who_am_i(), get_rule_prefix(rule, pat, 0, NULL), |
| 164 | (int)pat_len, pat, |
| 165 | (rule->rflags & FILTRULE_DIRECTORY) ? "/" : "", |
| 166 | listp->debug_type); |
| 167 | } |
| 168 | |
| 169 | /* These flags also indicate that we're reading a list that |
| 170 | * needs to be filtered now, not post-filtered later. */ |
| 171 | if (xflags & (XFLG_ANCHORED2ABS|XFLG_ABS_IF_SLASH) |
| 172 | && (rule->rflags & FILTRULES_SIDES) |
| 173 | == (am_sender ? FILTRULE_RECEIVER_SIDE : FILTRULE_SENDER_SIDE)) { |
| 174 | /* This filter applies only to the other side. Drop it. */ |
| 175 | free_filter(rule); |
| 176 | return; |
| 177 | } |
| 178 | |
| 179 | if (pat_len > 1 && pat[pat_len-1] == '/') { |
| 180 | pat_len--; |
| 181 | rule->rflags |= FILTRULE_DIRECTORY; |
| 182 | } |
| 183 | |
| 184 | for (cp = pat; cp < pat + pat_len; cp++) { |
| 185 | if (*cp == '/') |
| 186 | slash_cnt++; |
| 187 | } |
| 188 | |
| 189 | if (!(rule->rflags & (FILTRULE_ABS_PATH | FILTRULE_MERGE_FILE)) |
| 190 | && ((xflags & (XFLG_ANCHORED2ABS|XFLG_ABS_IF_SLASH) && *pat == '/') |
| 191 | || (xflags & XFLG_ABS_IF_SLASH && slash_cnt))) { |
| 192 | rule->rflags |= FILTRULE_ABS_PATH; |
| 193 | if (*pat == '/') |
| 194 | pre_len = dirbuf_len - module_dirlen - 1; |
| 195 | else |
| 196 | pre_len = 0; |
| 197 | } else |
| 198 | pre_len = 0; |
| 199 | |
| 200 | /* The daemon wants dir-exclude rules to get an appended "/" + "***". */ |
| 201 | if (xflags & XFLG_DIR2WILD3 |
| 202 | && BITS_SETnUNSET(rule->rflags, FILTRULE_DIRECTORY, FILTRULE_INCLUDE)) { |
| 203 | rule->rflags &= ~FILTRULE_DIRECTORY; |
| 204 | suf_len = sizeof SLASH_WILD3_SUFFIX - 1; |
| 205 | } else |
| 206 | suf_len = 0; |
| 207 | |
| 208 | if (!(rule->pattern = new_array(char, pre_len + pat_len + suf_len + 1))) |
| 209 | out_of_memory("add_rule"); |
| 210 | if (pre_len) { |
| 211 | memcpy(rule->pattern, dirbuf + module_dirlen, pre_len); |
| 212 | for (cp = rule->pattern; cp < rule->pattern + pre_len; cp++) { |
| 213 | if (*cp == '/') |
| 214 | slash_cnt++; |
| 215 | } |
| 216 | } |
| 217 | strlcpy(rule->pattern + pre_len, pat, pat_len + 1); |
| 218 | pat_len += pre_len; |
| 219 | if (suf_len) { |
| 220 | memcpy(rule->pattern + pat_len, SLASH_WILD3_SUFFIX, suf_len+1); |
| 221 | pat_len += suf_len; |
| 222 | slash_cnt++; |
| 223 | } |
| 224 | |
| 225 | if (strpbrk(rule->pattern, "*[?")) { |
| 226 | rule->rflags |= FILTRULE_WILD; |
| 227 | if ((cp = strstr(rule->pattern, "**")) != NULL) { |
| 228 | rule->rflags |= FILTRULE_WILD2; |
| 229 | /* If the pattern starts with **, note that. */ |
| 230 | if (cp == rule->pattern) |
| 231 | rule->rflags |= FILTRULE_WILD2_PREFIX; |
| 232 | /* If the pattern ends with ***, note that. */ |
| 233 | if (pat_len >= 3 |
| 234 | && rule->pattern[pat_len-3] == '*' |
| 235 | && rule->pattern[pat_len-2] == '*' |
| 236 | && rule->pattern[pat_len-1] == '*') |
| 237 | rule->rflags |= FILTRULE_WILD3_SUFFIX; |
| 238 | } |
| 239 | } |
| 240 | |
| 241 | if (rule->rflags & FILTRULE_PERDIR_MERGE) { |
| 242 | filter_rule_list *lp; |
| 243 | unsigned int len; |
| 244 | int i; |
| 245 | |
| 246 | if ((cp = strrchr(rule->pattern, '/')) != NULL) |
| 247 | cp++; |
| 248 | else |
| 249 | cp = rule->pattern; |
| 250 | |
| 251 | /* If the local merge file was already mentioned, don't |
| 252 | * add it again. */ |
| 253 | for (i = 0; i < mergelist_cnt; i++) { |
| 254 | filter_rule *ex = mergelist_parents[i]; |
| 255 | const char *s = strrchr(ex->pattern, '/'); |
| 256 | if (s) |
| 257 | s++; |
| 258 | else |
| 259 | s = ex->pattern; |
| 260 | len = strlen(s); |
| 261 | if (len == pat_len - (cp - rule->pattern) && memcmp(s, cp, len) == 0) { |
| 262 | free_filter(rule); |
| 263 | return; |
| 264 | } |
| 265 | } |
| 266 | |
| 267 | if (!(lp = new_array(filter_rule_list, 1))) |
| 268 | out_of_memory("add_rule"); |
| 269 | lp->head = lp->tail = lp->parent_dirscan_head = NULL; |
| 270 | if (asprintf(&lp->debug_type, " [per-dir %s]", cp) < 0) |
| 271 | out_of_memory("add_rule"); |
| 272 | rule->u.mergelist = lp; |
| 273 | |
| 274 | if (mergelist_cnt == mergelist_size) { |
| 275 | mergelist_size += 5; |
| 276 | mergelist_parents = realloc_array(mergelist_parents, |
| 277 | filter_rule *, |
| 278 | mergelist_size); |
| 279 | if (!mergelist_parents) |
| 280 | out_of_memory("add_rule"); |
| 281 | } |
| 282 | if (DEBUG_GTE(FILTER, 2)) { |
| 283 | rprintf(FINFO, "[%s] activating mergelist #%d%s\n", |
| 284 | who_am_i(), mergelist_cnt, lp->debug_type); |
| 285 | } |
| 286 | mergelist_parents[mergelist_cnt++] = rule; |
| 287 | } else |
| 288 | rule->u.slash_cnt = slash_cnt; |
| 289 | |
| 290 | if (!listp->tail) { |
| 291 | rule->next = listp->head; |
| 292 | listp->head = listp->tail = rule; |
| 293 | } else { |
| 294 | rule->next = listp->tail->next; |
| 295 | listp->tail->next = rule; |
| 296 | listp->tail = rule; |
| 297 | } |
| 298 | } |
| 299 | |
| 300 | static void clear_filter_list(filter_rule_list *listp) |
| 301 | { |
| 302 | if (listp->tail) { |
| 303 | /* Truncate any inherited items from the local list. */ |
| 304 | listp->tail->next = NULL; |
| 305 | /* Now free everything that is left. */ |
| 306 | free_filters(listp->head); |
| 307 | } |
| 308 | |
| 309 | listp->head = listp->tail = NULL; |
| 310 | } |
| 311 | |
| 312 | /* This returns an expanded (absolute) filename for the merge-file name if |
| 313 | * the name has any slashes in it OR if the parent_dirscan var is True; |
| 314 | * otherwise it returns the original merge_file name. If the len_ptr value |
| 315 | * is non-NULL the merge_file name is limited by the referenced length |
| 316 | * value and will be updated with the length of the resulting name. We |
| 317 | * always return a name that is null terminated, even if the merge_file |
| 318 | * name was not. */ |
| 319 | static char *parse_merge_name(const char *merge_file, unsigned int *len_ptr, |
| 320 | unsigned int prefix_skip) |
| 321 | { |
| 322 | static char buf[MAXPATHLEN]; |
| 323 | char *fn, tmpbuf[MAXPATHLEN]; |
| 324 | unsigned int fn_len; |
| 325 | |
| 326 | if (!parent_dirscan && *merge_file != '/') { |
| 327 | /* Return the name unchanged it doesn't have any slashes. */ |
| 328 | if (len_ptr) { |
| 329 | const char *p = merge_file + *len_ptr; |
| 330 | while (--p > merge_file && *p != '/') {} |
| 331 | if (p == merge_file) { |
| 332 | strlcpy(buf, merge_file, *len_ptr + 1); |
| 333 | return buf; |
| 334 | } |
| 335 | } else if (strchr(merge_file, '/') == NULL) |
| 336 | return (char *)merge_file; |
| 337 | } |
| 338 | |
| 339 | fn = *merge_file == '/' ? buf : tmpbuf; |
| 340 | if (sanitize_paths) { |
| 341 | const char *r = prefix_skip ? "/" : NULL; |
| 342 | /* null-terminate the name if it isn't already */ |
| 343 | if (len_ptr && merge_file[*len_ptr]) { |
| 344 | char *to = fn == buf ? tmpbuf : buf; |
| 345 | strlcpy(to, merge_file, *len_ptr + 1); |
| 346 | merge_file = to; |
| 347 | } |
| 348 | if (!sanitize_path(fn, merge_file, r, dirbuf_depth, SP_DEFAULT)) { |
| 349 | rprintf(FERROR, "merge-file name overflows: %s\n", |
| 350 | merge_file); |
| 351 | return NULL; |
| 352 | } |
| 353 | fn_len = strlen(fn); |
| 354 | } else { |
| 355 | strlcpy(fn, merge_file, len_ptr ? *len_ptr + 1 : MAXPATHLEN); |
| 356 | fn_len = clean_fname(fn, CFN_COLLAPSE_DOT_DOT_DIRS); |
| 357 | } |
| 358 | |
| 359 | /* If the name isn't in buf yet, it's wasn't absolute. */ |
| 360 | if (fn != buf) { |
| 361 | int d_len = dirbuf_len - prefix_skip; |
| 362 | if (d_len + fn_len >= MAXPATHLEN) { |
| 363 | rprintf(FERROR, "merge-file name overflows: %s\n", fn); |
| 364 | return NULL; |
| 365 | } |
| 366 | memcpy(buf, dirbuf + prefix_skip, d_len); |
| 367 | memcpy(buf + d_len, fn, fn_len + 1); |
| 368 | fn_len = clean_fname(buf, CFN_COLLAPSE_DOT_DOT_DIRS); |
| 369 | } |
| 370 | |
| 371 | if (len_ptr) |
| 372 | *len_ptr = fn_len; |
| 373 | return buf; |
| 374 | } |
| 375 | |
| 376 | /* Sets the dirbuf and dirbuf_len values. */ |
| 377 | void set_filter_dir(const char *dir, unsigned int dirlen) |
| 378 | { |
| 379 | unsigned int len; |
| 380 | if (*dir != '/') { |
| 381 | memcpy(dirbuf, curr_dir, curr_dir_len); |
| 382 | dirbuf[curr_dir_len] = '/'; |
| 383 | len = curr_dir_len + 1; |
| 384 | if (len + dirlen >= MAXPATHLEN) |
| 385 | dirlen = 0; |
| 386 | } else |
| 387 | len = 0; |
| 388 | memcpy(dirbuf + len, dir, dirlen); |
| 389 | dirbuf[dirlen + len] = '\0'; |
| 390 | dirbuf_len = clean_fname(dirbuf, CFN_COLLAPSE_DOT_DOT_DIRS); |
| 391 | if (dirbuf_len > 1 && dirbuf[dirbuf_len-1] == '.' |
| 392 | && dirbuf[dirbuf_len-2] == '/') |
| 393 | dirbuf_len -= 2; |
| 394 | if (dirbuf_len != 1) |
| 395 | dirbuf[dirbuf_len++] = '/'; |
| 396 | dirbuf[dirbuf_len] = '\0'; |
| 397 | if (sanitize_paths) |
| 398 | dirbuf_depth = count_dir_elements(dirbuf + module_dirlen); |
| 399 | } |
| 400 | |
| 401 | /* This routine takes a per-dir merge-file entry and finishes its setup. |
| 402 | * If the name has a path portion then we check to see if it refers to a |
| 403 | * parent directory of the first transfer dir. If it does, we scan all the |
| 404 | * dirs from that point through the parent dir of the transfer dir looking |
| 405 | * for the per-dir merge-file in each one. */ |
| 406 | static BOOL setup_merge_file(int mergelist_num, filter_rule *ex, |
| 407 | filter_rule_list *lp) |
| 408 | { |
| 409 | char buf[MAXPATHLEN]; |
| 410 | char *x, *y, *pat = ex->pattern; |
| 411 | unsigned int len; |
| 412 | |
| 413 | if (!(x = parse_merge_name(pat, NULL, 0)) || *x != '/') |
| 414 | return 0; |
| 415 | |
| 416 | if (DEBUG_GTE(FILTER, 2)) { |
| 417 | rprintf(FINFO, "[%s] performing parent_dirscan for mergelist #%d%s\n", |
| 418 | who_am_i(), mergelist_num, lp->debug_type); |
| 419 | } |
| 420 | y = strrchr(x, '/'); |
| 421 | *y = '\0'; |
| 422 | ex->pattern = strdup(y+1); |
| 423 | if (!*x) |
| 424 | x = "/"; |
| 425 | if (*x == '/') |
| 426 | strlcpy(buf, x, MAXPATHLEN); |
| 427 | else |
| 428 | pathjoin(buf, MAXPATHLEN, dirbuf, x); |
| 429 | |
| 430 | len = clean_fname(buf, CFN_COLLAPSE_DOT_DOT_DIRS); |
| 431 | if (len != 1 && len < MAXPATHLEN-1) { |
| 432 | buf[len++] = '/'; |
| 433 | buf[len] = '\0'; |
| 434 | } |
| 435 | /* This ensures that the specified dir is a parent of the transfer. */ |
| 436 | for (x = buf, y = dirbuf; *x && *x == *y; x++, y++) {} |
| 437 | if (*x) |
| 438 | y += strlen(y); /* nope -- skip the scan */ |
| 439 | |
| 440 | parent_dirscan = True; |
| 441 | while (*y) { |
| 442 | char save[MAXPATHLEN]; |
| 443 | strlcpy(save, y, MAXPATHLEN); |
| 444 | *y = '\0'; |
| 445 | dirbuf_len = y - dirbuf; |
| 446 | strlcpy(x, ex->pattern, MAXPATHLEN - (x - buf)); |
| 447 | parse_filter_file(lp, buf, ex, XFLG_ANCHORED2ABS); |
| 448 | if (ex->rflags & FILTRULE_NO_INHERIT) { |
| 449 | /* Free the undesired rules to clean up any per-dir |
| 450 | * mergelists they defined. Otherwise pop_local_filters |
| 451 | * may crash trying to restore nonexistent state for |
| 452 | * those mergelists. */ |
| 453 | free_filters(lp->head); |
| 454 | lp->head = NULL; |
| 455 | } |
| 456 | lp->tail = NULL; |
| 457 | strlcpy(y, save, MAXPATHLEN); |
| 458 | while ((*x++ = *y++) != '/') {} |
| 459 | } |
| 460 | /* Save current head for freeing when the mergelist becomes inactive. */ |
| 461 | lp->parent_dirscan_head = lp->head; |
| 462 | parent_dirscan = False; |
| 463 | if (DEBUG_GTE(FILTER, 2)) { |
| 464 | rprintf(FINFO, "[%s] completed parent_dirscan for mergelist #%d%s\n", |
| 465 | who_am_i(), mergelist_num, lp->debug_type); |
| 466 | } |
| 467 | free(pat); |
| 468 | return 1; |
| 469 | } |
| 470 | |
| 471 | struct local_filter_state { |
| 472 | int mergelist_cnt; |
| 473 | filter_rule_list mergelists[1]; |
| 474 | }; |
| 475 | |
| 476 | /* Each time rsync changes to a new directory it call this function to |
| 477 | * handle all the per-dir merge-files. The "dir" value is the current path |
| 478 | * relative to curr_dir (which might not be null-terminated). We copy it |
| 479 | * into dirbuf so that we can easily append a file name on the end. */ |
| 480 | void *push_local_filters(const char *dir, unsigned int dirlen) |
| 481 | { |
| 482 | struct local_filter_state *push; |
| 483 | int i; |
| 484 | |
| 485 | set_filter_dir(dir, dirlen); |
| 486 | if (DEBUG_GTE(FILTER, 2)) { |
| 487 | rprintf(FINFO, "[%s] pushing local filters for %s\n", |
| 488 | who_am_i(), dirbuf); |
| 489 | } |
| 490 | |
| 491 | if (!mergelist_cnt) { |
| 492 | /* No old state to save and no new merge files to push. */ |
| 493 | return NULL; |
| 494 | } |
| 495 | |
| 496 | push = (struct local_filter_state *)new_array(char, |
| 497 | sizeof (struct local_filter_state) |
| 498 | + (mergelist_cnt-1) * sizeof (filter_rule_list)); |
| 499 | if (!push) |
| 500 | out_of_memory("push_local_filters"); |
| 501 | |
| 502 | push->mergelist_cnt = mergelist_cnt; |
| 503 | for (i = 0; i < mergelist_cnt; i++) { |
| 504 | memcpy(&push->mergelists[i], mergelist_parents[i]->u.mergelist, |
| 505 | sizeof (filter_rule_list)); |
| 506 | } |
| 507 | |
| 508 | /* Note: parse_filter_file() might increase mergelist_cnt, so keep |
| 509 | * this loop separate from the above loop. */ |
| 510 | for (i = 0; i < mergelist_cnt; i++) { |
| 511 | filter_rule *ex = mergelist_parents[i]; |
| 512 | filter_rule_list *lp = ex->u.mergelist; |
| 513 | |
| 514 | if (DEBUG_GTE(FILTER, 2)) { |
| 515 | rprintf(FINFO, "[%s] pushing mergelist #%d%s\n", |
| 516 | who_am_i(), i, lp->debug_type); |
| 517 | } |
| 518 | |
| 519 | lp->tail = NULL; /* Switch any local rules to inherited. */ |
| 520 | if (ex->rflags & FILTRULE_NO_INHERIT) |
| 521 | lp->head = NULL; |
| 522 | |
| 523 | if (ex->rflags & FILTRULE_FINISH_SETUP) { |
| 524 | ex->rflags &= ~FILTRULE_FINISH_SETUP; |
| 525 | if (setup_merge_file(i, ex, lp)) |
| 526 | set_filter_dir(dir, dirlen); |
| 527 | } |
| 528 | |
| 529 | if (strlcpy(dirbuf + dirbuf_len, ex->pattern, |
| 530 | MAXPATHLEN - dirbuf_len) < MAXPATHLEN - dirbuf_len) { |
| 531 | parse_filter_file(lp, dirbuf, ex, |
| 532 | XFLG_ANCHORED2ABS); |
| 533 | } else { |
| 534 | io_error |= IOERR_GENERAL; |
| 535 | rprintf(FERROR, |
| 536 | "cannot add local filter rules in long-named directory: %s\n", |
| 537 | full_fname(dirbuf)); |
| 538 | } |
| 539 | dirbuf[dirbuf_len] = '\0'; |
| 540 | } |
| 541 | |
| 542 | return (void*)push; |
| 543 | } |
| 544 | |
| 545 | void pop_local_filters(void *mem) |
| 546 | { |
| 547 | struct local_filter_state *pop = (struct local_filter_state *)mem; |
| 548 | int i; |
| 549 | int old_mergelist_cnt = pop ? pop->mergelist_cnt : 0; |
| 550 | |
| 551 | if (DEBUG_GTE(FILTER, 2)) |
| 552 | rprintf(FINFO, "[%s] popping local filters\n", who_am_i()); |
| 553 | |
| 554 | for (i = mergelist_cnt; i-- > 0; ) { |
| 555 | filter_rule *ex = mergelist_parents[i]; |
| 556 | filter_rule_list *lp = ex->u.mergelist; |
| 557 | |
| 558 | if (DEBUG_GTE(FILTER, 2)) { |
| 559 | rprintf(FINFO, "[%s] popping mergelist #%d%s\n", |
| 560 | who_am_i(), i, lp->debug_type); |
| 561 | } |
| 562 | |
| 563 | clear_filter_list(lp); |
| 564 | |
| 565 | if (i >= old_mergelist_cnt) { |
| 566 | /* This mergelist does not exist in the state to be |
| 567 | * restored. Free its parent_dirscan list to clean up |
| 568 | * any per-dir mergelists defined there so we don't |
| 569 | * crash trying to restore nonexistent state for them |
| 570 | * below. (Counterpart to setup_merge_file call in |
| 571 | * push_local_filters. Must be done here, not in |
| 572 | * free_filter, for LIFO order.) */ |
| 573 | if (DEBUG_GTE(FILTER, 2)) { |
| 574 | rprintf(FINFO, "[%s] freeing parent_dirscan filters of mergelist #%d%s\n", |
| 575 | who_am_i(), i, ex->u.mergelist->debug_type); |
| 576 | } |
| 577 | free_filters(lp->parent_dirscan_head); |
| 578 | lp->parent_dirscan_head = NULL; |
| 579 | } |
| 580 | } |
| 581 | |
| 582 | /* If we cleaned things up properly, the only still-active mergelists |
| 583 | * should be those with a state to be restored. */ |
| 584 | assert(mergelist_cnt == old_mergelist_cnt); |
| 585 | |
| 586 | if (!pop) { |
| 587 | /* No state to restore. */ |
| 588 | return; |
| 589 | } |
| 590 | |
| 591 | for (i = 0; i < mergelist_cnt; i++) { |
| 592 | memcpy(mergelist_parents[i]->u.mergelist, &pop->mergelists[i], |
| 593 | sizeof (filter_rule_list)); |
| 594 | } |
| 595 | |
| 596 | free(pop); |
| 597 | } |
| 598 | |
| 599 | void change_local_filter_dir(const char *dname, int dlen, int dir_depth) |
| 600 | { |
| 601 | static int cur_depth = -1; |
| 602 | static void *filt_array[MAXPATHLEN/2+1]; |
| 603 | |
| 604 | if (!dname) { |
| 605 | for ( ; cur_depth >= 0; cur_depth--) { |
| 606 | if (filt_array[cur_depth]) { |
| 607 | pop_local_filters(filt_array[cur_depth]); |
| 608 | filt_array[cur_depth] = NULL; |
| 609 | } |
| 610 | } |
| 611 | return; |
| 612 | } |
| 613 | |
| 614 | assert(dir_depth < MAXPATHLEN/2+1); |
| 615 | |
| 616 | for ( ; cur_depth >= dir_depth; cur_depth--) { |
| 617 | if (filt_array[cur_depth]) { |
| 618 | pop_local_filters(filt_array[cur_depth]); |
| 619 | filt_array[cur_depth] = NULL; |
| 620 | } |
| 621 | } |
| 622 | |
| 623 | cur_depth = dir_depth; |
| 624 | filt_array[cur_depth] = push_local_filters(dname, dlen); |
| 625 | } |
| 626 | |
| 627 | static int rule_matches(const char *fname, filter_rule *ex, int name_is_dir) |
| 628 | { |
| 629 | int slash_handling, str_cnt = 0, anchored_match = 0; |
| 630 | int ret_match = ex->rflags & FILTRULE_NEGATE ? 0 : 1; |
| 631 | char *p, *pattern = ex->pattern; |
| 632 | const char *strings[16]; /* more than enough */ |
| 633 | const char *name = fname + (*fname == '/'); |
| 634 | |
| 635 | if (!*name) |
| 636 | return 0; |
| 637 | |
| 638 | if (!ex->u.slash_cnt && !(ex->rflags & FILTRULE_WILD2)) { |
| 639 | /* If the pattern does not have any slashes AND it does |
| 640 | * not have a "**" (which could match a slash), then we |
| 641 | * just match the name portion of the path. */ |
| 642 | if ((p = strrchr(name,'/')) != NULL) |
| 643 | name = p+1; |
| 644 | } else if (ex->rflags & FILTRULE_ABS_PATH && *fname != '/' |
| 645 | && curr_dir_len > module_dirlen + 1) { |
| 646 | /* If we're matching against an absolute-path pattern, |
| 647 | * we need to prepend our full path info. */ |
| 648 | strings[str_cnt++] = curr_dir + module_dirlen + 1; |
| 649 | strings[str_cnt++] = "/"; |
| 650 | } else if (ex->rflags & FILTRULE_WILD2_PREFIX && *fname != '/') { |
| 651 | /* Allow "**"+"/" to match at the start of the string. */ |
| 652 | strings[str_cnt++] = "/"; |
| 653 | } |
| 654 | strings[str_cnt++] = name; |
| 655 | if (name_is_dir) { |
| 656 | /* Allow a trailing "/"+"***" to match the directory. */ |
| 657 | if (ex->rflags & FILTRULE_WILD3_SUFFIX) |
| 658 | strings[str_cnt++] = "/"; |
| 659 | } else if (ex->rflags & FILTRULE_DIRECTORY) |
| 660 | return !ret_match; |
| 661 | strings[str_cnt] = NULL; |
| 662 | |
| 663 | if (*pattern == '/') { |
| 664 | anchored_match = 1; |
| 665 | pattern++; |
| 666 | } |
| 667 | |
| 668 | if (!anchored_match && ex->u.slash_cnt |
| 669 | && !(ex->rflags & FILTRULE_WILD2)) { |
| 670 | /* A non-anchored match with an infix slash and no "**" |
| 671 | * needs to match the last slash_cnt+1 name elements. */ |
| 672 | slash_handling = ex->u.slash_cnt + 1; |
| 673 | } else if (!anchored_match && !(ex->rflags & FILTRULE_WILD2_PREFIX) |
| 674 | && ex->rflags & FILTRULE_WILD2) { |
| 675 | /* A non-anchored match with an infix or trailing "**" (but not |
| 676 | * a prefixed "**") needs to try matching after every slash. */ |
| 677 | slash_handling = -1; |
| 678 | } else { |
| 679 | /* The pattern matches only at the start of the path or name. */ |
| 680 | slash_handling = 0; |
| 681 | } |
| 682 | |
| 683 | if (ex->rflags & FILTRULE_WILD) { |
| 684 | if (wildmatch_array(pattern, strings, slash_handling)) |
| 685 | return ret_match; |
| 686 | } else if (str_cnt > 1) { |
| 687 | if (litmatch_array(pattern, strings, slash_handling)) |
| 688 | return ret_match; |
| 689 | } else if (anchored_match) { |
| 690 | if (strcmp(name, pattern) == 0) |
| 691 | return ret_match; |
| 692 | } else { |
| 693 | int l1 = strlen(name); |
| 694 | int l2 = strlen(pattern); |
| 695 | if (l2 <= l1 && |
| 696 | strcmp(name+(l1-l2),pattern) == 0 && |
| 697 | (l1==l2 || name[l1-(l2+1)] == '/')) { |
| 698 | return ret_match; |
| 699 | } |
| 700 | } |
| 701 | |
| 702 | return !ret_match; |
| 703 | } |
| 704 | |
| 705 | static void report_filter_result(enum logcode code, char const *name, |
| 706 | filter_rule const *ent, |
| 707 | int name_is_dir, const char *type) |
| 708 | { |
| 709 | /* If a trailing slash is present to match only directories, |
| 710 | * then it is stripped out by add_rule(). So as a special |
| 711 | * case we add it back in here. */ |
| 712 | |
| 713 | if (DEBUG_GTE(FILTER, 1)) { |
| 714 | static char *actions[2][2] |
| 715 | = { {"show", "hid"}, {"risk", "protect"} }; |
| 716 | const char *w = who_am_i(); |
| 717 | rprintf(code, "[%s] %sing %s %s because of pattern %s%s%s\n", |
| 718 | w, actions[*w!='s'][!(ent->rflags & FILTRULE_INCLUDE)], |
| 719 | name_is_dir ? "directory" : "file", name, ent->pattern, |
| 720 | ent->rflags & FILTRULE_DIRECTORY ? "/" : "", type); |
| 721 | } |
| 722 | } |
| 723 | |
| 724 | /* Return -1 if file "name" is defined to be excluded by the specified |
| 725 | * exclude list, 1 if it is included, and 0 if it was not matched. */ |
| 726 | int check_filter(filter_rule_list *listp, enum logcode code, |
| 727 | const char *name, int name_is_dir) |
| 728 | { |
| 729 | filter_rule *ent; |
| 730 | |
| 731 | for (ent = listp->head; ent; ent = ent->next) { |
| 732 | if (ignore_perishable && ent->rflags & FILTRULE_PERISHABLE) |
| 733 | continue; |
| 734 | if (ent->rflags & FILTRULE_PERDIR_MERGE) { |
| 735 | int rc = check_filter(ent->u.mergelist, code, name, |
| 736 | name_is_dir); |
| 737 | if (rc) |
| 738 | return rc; |
| 739 | continue; |
| 740 | } |
| 741 | if (ent->rflags & FILTRULE_CVS_IGNORE) { |
| 742 | int rc = check_filter(&cvs_filter_list, code, name, |
| 743 | name_is_dir); |
| 744 | if (rc) |
| 745 | return rc; |
| 746 | continue; |
| 747 | } |
| 748 | if (rule_matches(name, ent, name_is_dir)) { |
| 749 | report_filter_result(code, name, ent, name_is_dir, |
| 750 | listp->debug_type); |
| 751 | return ent->rflags & FILTRULE_INCLUDE ? 1 : -1; |
| 752 | } |
| 753 | } |
| 754 | |
| 755 | return 0; |
| 756 | } |
| 757 | |
| 758 | #define RULE_STRCMP(s,r) rule_strcmp((s), (r), sizeof (r) - 1) |
| 759 | |
| 760 | static const uchar *rule_strcmp(const uchar *str, const char *rule, int rule_len) |
| 761 | { |
| 762 | if (strncmp((char*)str, rule, rule_len) != 0) |
| 763 | return NULL; |
| 764 | if (isspace(str[rule_len]) || str[rule_len] == '_' || !str[rule_len]) |
| 765 | return str + rule_len - 1; |
| 766 | if (str[rule_len] == ',') |
| 767 | return str + rule_len; |
| 768 | return NULL; |
| 769 | } |
| 770 | |
| 771 | #define FILTRULES_FROM_CONTAINER (FILTRULE_ABS_PATH | FILTRULE_INCLUDE \ |
| 772 | | FILTRULE_DIRECTORY | FILTRULE_NEGATE \ |
| 773 | | FILTRULE_PERISHABLE) |
| 774 | |
| 775 | /* Gets the next include/exclude rule from *rulestr_ptr and advances |
| 776 | * *rulestr_ptr to point beyond it. Stores the pattern's start (within |
| 777 | * *rulestr_ptr) and length in *pat_ptr and *pat_len_ptr, and returns a newly |
| 778 | * allocated filter_rule containing the rest of the information. Returns |
| 779 | * NULL if there are no more rules in the input. |
| 780 | * |
| 781 | * The template provides defaults for the new rule to inherit, and the |
| 782 | * template rflags and the xflags additionally affect parsing. */ |
| 783 | static filter_rule *parse_rule_tok(const char **rulestr_ptr, |
| 784 | const filter_rule *template, int xflags, |
| 785 | const char **pat_ptr, unsigned int *pat_len_ptr) |
| 786 | { |
| 787 | const uchar *s = (const uchar *)*rulestr_ptr; |
| 788 | filter_rule *rule; |
| 789 | unsigned int len; |
| 790 | |
| 791 | if (template->rflags & FILTRULE_WORD_SPLIT) { |
| 792 | /* Skip over any initial whitespace. */ |
| 793 | while (isspace(*s)) |
| 794 | s++; |
| 795 | /* Update to point to real start of rule. */ |
| 796 | *rulestr_ptr = (const char *)s; |
| 797 | } |
| 798 | if (!*s) |
| 799 | return NULL; |
| 800 | |
| 801 | if (!(rule = new0(filter_rule))) |
| 802 | out_of_memory("parse_rule_tok"); |
| 803 | |
| 804 | /* Inherit from the template. Don't inherit FILTRULES_SIDES; we check |
| 805 | * that later. */ |
| 806 | rule->rflags = template->rflags & FILTRULES_FROM_CONTAINER; |
| 807 | |
| 808 | /* Figure out what kind of a filter rule "s" is pointing at. Note |
| 809 | * that if FILTRULE_NO_PREFIXES is set, the rule is either an include |
| 810 | * or an exclude based on the inheritance of the FILTRULE_INCLUDE |
| 811 | * flag (above). XFLG_OLD_PREFIXES indicates a compatibility mode |
| 812 | * for old include/exclude patterns where just "+ " and "- " are |
| 813 | * allowed as optional prefixes. */ |
| 814 | if (template->rflags & FILTRULE_NO_PREFIXES) { |
| 815 | if (*s == '!' && template->rflags & FILTRULE_CVS_IGNORE) |
| 816 | rule->rflags |= FILTRULE_CLEAR_LIST; /* Tentative! */ |
| 817 | } else if (xflags & XFLG_OLD_PREFIXES) { |
| 818 | if (*s == '-' && s[1] == ' ') { |
| 819 | rule->rflags &= ~FILTRULE_INCLUDE; |
| 820 | s += 2; |
| 821 | } else if (*s == '+' && s[1] == ' ') { |
| 822 | rule->rflags |= FILTRULE_INCLUDE; |
| 823 | s += 2; |
| 824 | } else if (*s == '!') |
| 825 | rule->rflags |= FILTRULE_CLEAR_LIST; /* Tentative! */ |
| 826 | } else { |
| 827 | char ch = 0; |
| 828 | BOOL prefix_specifies_side = False; |
| 829 | switch (*s) { |
| 830 | case 'c': |
| 831 | if ((s = RULE_STRCMP(s, "clear")) != NULL) |
| 832 | ch = '!'; |
| 833 | break; |
| 834 | case 'd': |
| 835 | if ((s = RULE_STRCMP(s, "dir-merge")) != NULL) |
| 836 | ch = ':'; |
| 837 | break; |
| 838 | case 'e': |
| 839 | if ((s = RULE_STRCMP(s, "exclude")) != NULL) |
| 840 | ch = '-'; |
| 841 | break; |
| 842 | case 'h': |
| 843 | if ((s = RULE_STRCMP(s, "hide")) != NULL) |
| 844 | ch = 'H'; |
| 845 | break; |
| 846 | case 'i': |
| 847 | if ((s = RULE_STRCMP(s, "include")) != NULL) |
| 848 | ch = '+'; |
| 849 | break; |
| 850 | case 'm': |
| 851 | if ((s = RULE_STRCMP(s, "merge")) != NULL) |
| 852 | ch = '.'; |
| 853 | break; |
| 854 | case 'p': |
| 855 | if ((s = RULE_STRCMP(s, "protect")) != NULL) |
| 856 | ch = 'P'; |
| 857 | break; |
| 858 | case 'r': |
| 859 | if ((s = RULE_STRCMP(s, "risk")) != NULL) |
| 860 | ch = 'R'; |
| 861 | break; |
| 862 | case 's': |
| 863 | if ((s = RULE_STRCMP(s, "show")) != NULL) |
| 864 | ch = 'S'; |
| 865 | break; |
| 866 | default: |
| 867 | ch = *s; |
| 868 | if (s[1] == ',') |
| 869 | s++; |
| 870 | break; |
| 871 | } |
| 872 | switch (ch) { |
| 873 | case ':': |
| 874 | rule->rflags |= FILTRULE_PERDIR_MERGE |
| 875 | | FILTRULE_FINISH_SETUP; |
| 876 | /* FALL THROUGH */ |
| 877 | case '.': |
| 878 | rule->rflags |= FILTRULE_MERGE_FILE; |
| 879 | break; |
| 880 | case '+': |
| 881 | rule->rflags |= FILTRULE_INCLUDE; |
| 882 | break; |
| 883 | case '-': |
| 884 | break; |
| 885 | case 'S': |
| 886 | rule->rflags |= FILTRULE_INCLUDE; |
| 887 | /* FALL THROUGH */ |
| 888 | case 'H': |
| 889 | rule->rflags |= FILTRULE_SENDER_SIDE; |
| 890 | prefix_specifies_side = True; |
| 891 | break; |
| 892 | case 'R': |
| 893 | rule->rflags |= FILTRULE_INCLUDE; |
| 894 | /* FALL THROUGH */ |
| 895 | case 'P': |
| 896 | rule->rflags |= FILTRULE_RECEIVER_SIDE; |
| 897 | prefix_specifies_side = True; |
| 898 | break; |
| 899 | case '!': |
| 900 | rule->rflags |= FILTRULE_CLEAR_LIST; |
| 901 | break; |
| 902 | default: |
| 903 | rprintf(FERROR, "Unknown filter rule: `%s'\n", *rulestr_ptr); |
| 904 | exit_cleanup(RERR_SYNTAX); |
| 905 | } |
| 906 | while (ch != '!' && *++s && *s != ' ' && *s != '_') { |
| 907 | if (template->rflags & FILTRULE_WORD_SPLIT && isspace(*s)) { |
| 908 | s--; |
| 909 | break; |
| 910 | } |
| 911 | switch (*s) { |
| 912 | default: |
| 913 | invalid: |
| 914 | rprintf(FERROR, |
| 915 | "invalid modifier '%c' at position %d in filter rule: %s\n", |
| 916 | *s, (int)(s - (const uchar *)*rulestr_ptr), *rulestr_ptr); |
| 917 | exit_cleanup(RERR_SYNTAX); |
| 918 | case '-': |
| 919 | if (!BITS_SETnUNSET(rule->rflags, FILTRULE_MERGE_FILE, FILTRULE_NO_PREFIXES)) |
| 920 | goto invalid; |
| 921 | rule->rflags |= FILTRULE_NO_PREFIXES; |
| 922 | break; |
| 923 | case '+': |
| 924 | if (!BITS_SETnUNSET(rule->rflags, FILTRULE_MERGE_FILE, FILTRULE_NO_PREFIXES)) |
| 925 | goto invalid; |
| 926 | rule->rflags |= FILTRULE_NO_PREFIXES |
| 927 | | FILTRULE_INCLUDE; |
| 928 | break; |
| 929 | case '/': |
| 930 | rule->rflags |= FILTRULE_ABS_PATH; |
| 931 | break; |
| 932 | case '!': |
| 933 | /* Negation really goes with the pattern, so it |
| 934 | * isn't useful as a merge-file default. */ |
| 935 | if (rule->rflags & FILTRULE_MERGE_FILE) |
| 936 | goto invalid; |
| 937 | rule->rflags |= FILTRULE_NEGATE; |
| 938 | break; |
| 939 | case 'C': |
| 940 | if (rule->rflags & FILTRULE_NO_PREFIXES || prefix_specifies_side) |
| 941 | goto invalid; |
| 942 | rule->rflags |= FILTRULE_NO_PREFIXES |
| 943 | | FILTRULE_WORD_SPLIT |
| 944 | | FILTRULE_NO_INHERIT |
| 945 | | FILTRULE_CVS_IGNORE; |
| 946 | break; |
| 947 | case 'e': |
| 948 | if (!(rule->rflags & FILTRULE_MERGE_FILE)) |
| 949 | goto invalid; |
| 950 | rule->rflags |= FILTRULE_EXCLUDE_SELF; |
| 951 | break; |
| 952 | case 'n': |
| 953 | if (!(rule->rflags & FILTRULE_MERGE_FILE)) |
| 954 | goto invalid; |
| 955 | rule->rflags |= FILTRULE_NO_INHERIT; |
| 956 | break; |
| 957 | case 'p': |
| 958 | rule->rflags |= FILTRULE_PERISHABLE; |
| 959 | break; |
| 960 | case 'r': |
| 961 | if (prefix_specifies_side) |
| 962 | goto invalid; |
| 963 | rule->rflags |= FILTRULE_RECEIVER_SIDE; |
| 964 | break; |
| 965 | case 's': |
| 966 | if (prefix_specifies_side) |
| 967 | goto invalid; |
| 968 | rule->rflags |= FILTRULE_SENDER_SIDE; |
| 969 | break; |
| 970 | case 'w': |
| 971 | if (!(rule->rflags & FILTRULE_MERGE_FILE)) |
| 972 | goto invalid; |
| 973 | rule->rflags |= FILTRULE_WORD_SPLIT; |
| 974 | break; |
| 975 | } |
| 976 | } |
| 977 | if (*s) |
| 978 | s++; |
| 979 | } |
| 980 | if (template->rflags & FILTRULES_SIDES) { |
| 981 | if (rule->rflags & FILTRULES_SIDES) { |
| 982 | /* The filter and template both specify side(s). This |
| 983 | * is dodgy (and won't work correctly if the template is |
| 984 | * a one-sided per-dir merge rule), so reject it. */ |
| 985 | rprintf(FERROR, |
| 986 | "specified-side merge file contains specified-side filter: %s\n", |
| 987 | *rulestr_ptr); |
| 988 | exit_cleanup(RERR_SYNTAX); |
| 989 | } |
| 990 | rule->rflags |= template->rflags & FILTRULES_SIDES; |
| 991 | } |
| 992 | |
| 993 | if (template->rflags & FILTRULE_WORD_SPLIT) { |
| 994 | const uchar *cp = s; |
| 995 | /* Token ends at whitespace or the end of the string. */ |
| 996 | while (!isspace(*cp) && *cp != '\0') |
| 997 | cp++; |
| 998 | len = cp - s; |
| 999 | } else |
| 1000 | len = strlen((char*)s); |
| 1001 | |
| 1002 | if (rule->rflags & FILTRULE_CLEAR_LIST) { |
| 1003 | if (!(rule->rflags & FILTRULE_NO_PREFIXES) |
| 1004 | && !(xflags & XFLG_OLD_PREFIXES) && len) { |
| 1005 | rprintf(FERROR, |
| 1006 | "'!' rule has trailing characters: %s\n", *rulestr_ptr); |
| 1007 | exit_cleanup(RERR_SYNTAX); |
| 1008 | } |
| 1009 | if (len > 1) |
| 1010 | rule->rflags &= ~FILTRULE_CLEAR_LIST; |
| 1011 | } else if (!len && !(rule->rflags & FILTRULE_CVS_IGNORE)) { |
| 1012 | rprintf(FERROR, "unexpected end of filter rule: %s\n", *rulestr_ptr); |
| 1013 | exit_cleanup(RERR_SYNTAX); |
| 1014 | } |
| 1015 | |
| 1016 | /* --delete-excluded turns an un-modified include/exclude into a |
| 1017 | * sender-side rule. We also affect per-dir merge files that take |
| 1018 | * no prefixes as a simple optimization. */ |
| 1019 | if (delete_excluded |
| 1020 | && !(rule->rflags & FILTRULES_SIDES) |
| 1021 | && (!(rule->rflags & FILTRULE_PERDIR_MERGE) |
| 1022 | || rule->rflags & FILTRULE_NO_PREFIXES)) |
| 1023 | rule->rflags |= FILTRULE_SENDER_SIDE; |
| 1024 | |
| 1025 | *pat_ptr = (const char *)s; |
| 1026 | *pat_len_ptr = len; |
| 1027 | *rulestr_ptr = *pat_ptr + len; |
| 1028 | return rule; |
| 1029 | } |
| 1030 | |
| 1031 | static char default_cvsignore[] = |
| 1032 | /* These default ignored items come from the CVS manual. */ |
| 1033 | "RCS SCCS CVS CVS.adm RCSLOG cvslog.* tags TAGS" |
| 1034 | " .make.state .nse_depinfo *~ #* .#* ,* _$* *$" |
| 1035 | " *.old *.bak *.BAK *.orig *.rej .del-*" |
| 1036 | " *.a *.olb *.o *.obj *.so *.exe" |
| 1037 | " *.Z *.elc *.ln core" |
| 1038 | /* The rest we added to suit ourself. */ |
| 1039 | " .svn/ .git/ .bzr/"; |
| 1040 | |
| 1041 | static void get_cvs_excludes(uint32 rflags) |
| 1042 | { |
| 1043 | static int initialized = 0; |
| 1044 | char *p, fname[MAXPATHLEN]; |
| 1045 | |
| 1046 | if (initialized) |
| 1047 | return; |
| 1048 | initialized = 1; |
| 1049 | |
| 1050 | parse_filter_str(&cvs_filter_list, default_cvsignore, |
| 1051 | rule_template(rflags | (protocol_version >= 30 ? FILTRULE_PERISHABLE : 0)), |
| 1052 | 0); |
| 1053 | |
| 1054 | p = module_id >= 0 && lp_use_chroot(module_id) ? "/" : getenv("HOME"); |
| 1055 | if (p && pathjoin(fname, MAXPATHLEN, p, ".cvsignore") < MAXPATHLEN) |
| 1056 | parse_filter_file(&cvs_filter_list, fname, rule_template(rflags), 0); |
| 1057 | |
| 1058 | parse_filter_str(&cvs_filter_list, getenv("CVSIGNORE"), rule_template(rflags), 0); |
| 1059 | } |
| 1060 | |
| 1061 | const filter_rule *rule_template(uint32 rflags) |
| 1062 | { |
| 1063 | static filter_rule template; /* zero-initialized */ |
| 1064 | template.rflags = rflags; |
| 1065 | return &template; |
| 1066 | } |
| 1067 | |
| 1068 | void parse_filter_str(filter_rule_list *listp, const char *rulestr, |
| 1069 | const filter_rule *template, int xflags) |
| 1070 | { |
| 1071 | filter_rule *rule; |
| 1072 | const char *pat; |
| 1073 | unsigned int pat_len; |
| 1074 | |
| 1075 | if (!rulestr) |
| 1076 | return; |
| 1077 | |
| 1078 | while (1) { |
| 1079 | uint32 new_rflags; |
| 1080 | |
| 1081 | /* Remember that the returned string is NOT '\0' terminated! */ |
| 1082 | if (!(rule = parse_rule_tok(&rulestr, template, xflags, &pat, &pat_len))) |
| 1083 | break; |
| 1084 | |
| 1085 | if (pat_len >= MAXPATHLEN) { |
| 1086 | rprintf(FERROR, "discarding over-long filter: %.*s\n", |
| 1087 | (int)pat_len, pat); |
| 1088 | free_continue: |
| 1089 | free_filter(rule); |
| 1090 | continue; |
| 1091 | } |
| 1092 | |
| 1093 | new_rflags = rule->rflags; |
| 1094 | if (new_rflags & FILTRULE_CLEAR_LIST) { |
| 1095 | if (DEBUG_GTE(FILTER, 2)) { |
| 1096 | rprintf(FINFO, |
| 1097 | "[%s] clearing filter list%s\n", |
| 1098 | who_am_i(), listp->debug_type); |
| 1099 | } |
| 1100 | clear_filter_list(listp); |
| 1101 | goto free_continue; |
| 1102 | } |
| 1103 | |
| 1104 | if (new_rflags & FILTRULE_MERGE_FILE) { |
| 1105 | if (!pat_len) { |
| 1106 | pat = ".cvsignore"; |
| 1107 | pat_len = 10; |
| 1108 | } |
| 1109 | if (new_rflags & FILTRULE_EXCLUDE_SELF) { |
| 1110 | const char *name; |
| 1111 | filter_rule *excl_self; |
| 1112 | |
| 1113 | if (!(excl_self = new0(filter_rule))) |
| 1114 | out_of_memory("parse_filter_str"); |
| 1115 | /* Find the beginning of the basename and add an exclude for it. */ |
| 1116 | for (name = pat + pat_len; name > pat && name[-1] != '/'; name--) {} |
| 1117 | add_rule(listp, name, (pat + pat_len) - name, excl_self, 0); |
| 1118 | rule->rflags &= ~FILTRULE_EXCLUDE_SELF; |
| 1119 | } |
| 1120 | if (new_rflags & FILTRULE_PERDIR_MERGE) { |
| 1121 | if (parent_dirscan) { |
| 1122 | const char *p; |
| 1123 | unsigned int len = pat_len; |
| 1124 | if ((p = parse_merge_name(pat, &len, module_dirlen))) |
| 1125 | add_rule(listp, p, len, rule, 0); |
| 1126 | else |
| 1127 | free_filter(rule); |
| 1128 | continue; |
| 1129 | } |
| 1130 | } else { |
| 1131 | const char *p; |
| 1132 | unsigned int len = pat_len; |
| 1133 | if ((p = parse_merge_name(pat, &len, 0))) |
| 1134 | parse_filter_file(listp, p, rule, XFLG_FATAL_ERRORS); |
| 1135 | free_filter(rule); |
| 1136 | continue; |
| 1137 | } |
| 1138 | } |
| 1139 | |
| 1140 | add_rule(listp, pat, pat_len, rule, xflags); |
| 1141 | |
| 1142 | if (new_rflags & FILTRULE_CVS_IGNORE |
| 1143 | && !(new_rflags & FILTRULE_MERGE_FILE)) |
| 1144 | get_cvs_excludes(new_rflags); |
| 1145 | } |
| 1146 | } |
| 1147 | |
| 1148 | void parse_filter_file(filter_rule_list *listp, const char *fname, const filter_rule *template, int xflags) |
| 1149 | { |
| 1150 | FILE *fp; |
| 1151 | char line[BIGPATHBUFLEN]; |
| 1152 | char *eob = line + sizeof line - 1; |
| 1153 | BOOL word_split = (template->rflags & FILTRULE_WORD_SPLIT) != 0; |
| 1154 | |
| 1155 | if (!fname || !*fname) |
| 1156 | return; |
| 1157 | |
| 1158 | if (*fname != '-' || fname[1] || am_server) { |
| 1159 | if (daemon_filter_list.head) { |
| 1160 | strlcpy(line, fname, sizeof line); |
| 1161 | clean_fname(line, CFN_COLLAPSE_DOT_DOT_DIRS); |
| 1162 | if (check_filter(&daemon_filter_list, FLOG, line, 0) < 0) |
| 1163 | fp = NULL; |
| 1164 | else |
| 1165 | fp = fopen(line, "rb"); |
| 1166 | } else |
| 1167 | fp = fopen(fname, "rb"); |
| 1168 | } else |
| 1169 | fp = stdin; |
| 1170 | |
| 1171 | if (DEBUG_GTE(FILTER, 2)) { |
| 1172 | rprintf(FINFO, "[%s] parse_filter_file(%s,%x,%x)%s\n", |
| 1173 | who_am_i(), fname, template->rflags, xflags, |
| 1174 | fp ? "" : " [not found]"); |
| 1175 | } |
| 1176 | |
| 1177 | if (!fp) { |
| 1178 | if (xflags & XFLG_FATAL_ERRORS) { |
| 1179 | rsyserr(FERROR, errno, |
| 1180 | "failed to open %sclude file %s", |
| 1181 | template->rflags & FILTRULE_INCLUDE ? "in" : "ex", |
| 1182 | fname); |
| 1183 | exit_cleanup(RERR_FILEIO); |
| 1184 | } |
| 1185 | return; |
| 1186 | } |
| 1187 | dirbuf[dirbuf_len] = '\0'; |
| 1188 | |
| 1189 | while (1) { |
| 1190 | char *s = line; |
| 1191 | int ch, overflow = 0; |
| 1192 | while (1) { |
| 1193 | if ((ch = getc(fp)) == EOF) { |
| 1194 | if (ferror(fp) && errno == EINTR) { |
| 1195 | clearerr(fp); |
| 1196 | continue; |
| 1197 | } |
| 1198 | break; |
| 1199 | } |
| 1200 | if (word_split && isspace(ch)) |
| 1201 | break; |
| 1202 | if (eol_nulls? !ch : (ch == '\n' || ch == '\r')) |
| 1203 | break; |
| 1204 | if (s < eob) |
| 1205 | *s++ = ch; |
| 1206 | else |
| 1207 | overflow = 1; |
| 1208 | } |
| 1209 | if (overflow) { |
| 1210 | rprintf(FERROR, "discarding over-long filter: %s...\n", line); |
| 1211 | s = line; |
| 1212 | } |
| 1213 | *s = '\0'; |
| 1214 | /* Skip an empty token and (when line parsing) comments. */ |
| 1215 | if (*line && (word_split || (*line != ';' && *line != '#'))) |
| 1216 | parse_filter_str(listp, line, template, xflags); |
| 1217 | if (ch == EOF) |
| 1218 | break; |
| 1219 | } |
| 1220 | fclose(fp); |
| 1221 | } |
| 1222 | |
| 1223 | /* If the "for_xfer" flag is set, the prefix is made compatible with the |
| 1224 | * current protocol_version (if possible) or a NULL is returned (if not |
| 1225 | * possible). */ |
| 1226 | char *get_rule_prefix(filter_rule *rule, const char *pat, int for_xfer, |
| 1227 | unsigned int *plen_ptr) |
| 1228 | { |
| 1229 | static char buf[MAX_RULE_PREFIX+1]; |
| 1230 | char *op = buf; |
| 1231 | int legal_len = for_xfer && protocol_version < 29 ? 1 : MAX_RULE_PREFIX-1; |
| 1232 | |
| 1233 | if (rule->rflags & FILTRULE_PERDIR_MERGE) { |
| 1234 | if (legal_len == 1) |
| 1235 | return NULL; |
| 1236 | *op++ = ':'; |
| 1237 | } else if (rule->rflags & FILTRULE_INCLUDE) |
| 1238 | *op++ = '+'; |
| 1239 | else if (legal_len != 1 |
| 1240 | || ((*pat == '-' || *pat == '+') && pat[1] == ' ')) |
| 1241 | *op++ = '-'; |
| 1242 | else |
| 1243 | legal_len = 0; |
| 1244 | |
| 1245 | if (rule->rflags & FILTRULE_ABS_PATH) |
| 1246 | *op++ = '/'; |
| 1247 | if (rule->rflags & FILTRULE_NEGATE) |
| 1248 | *op++ = '!'; |
| 1249 | if (rule->rflags & FILTRULE_CVS_IGNORE) |
| 1250 | *op++ = 'C'; |
| 1251 | else { |
| 1252 | if (rule->rflags & FILTRULE_NO_INHERIT) |
| 1253 | *op++ = 'n'; |
| 1254 | if (rule->rflags & FILTRULE_WORD_SPLIT) |
| 1255 | *op++ = 'w'; |
| 1256 | if (rule->rflags & FILTRULE_NO_PREFIXES) { |
| 1257 | if (rule->rflags & FILTRULE_INCLUDE) |
| 1258 | *op++ = '+'; |
| 1259 | else |
| 1260 | *op++ = '-'; |
| 1261 | } |
| 1262 | } |
| 1263 | if (rule->rflags & FILTRULE_EXCLUDE_SELF) |
| 1264 | *op++ = 'e'; |
| 1265 | if (rule->rflags & FILTRULE_SENDER_SIDE |
| 1266 | && (!for_xfer || protocol_version >= 29)) |
| 1267 | *op++ = 's'; |
| 1268 | if (rule->rflags & FILTRULE_RECEIVER_SIDE |
| 1269 | && (!for_xfer || protocol_version >= 29 |
| 1270 | || (delete_excluded && am_sender))) |
| 1271 | *op++ = 'r'; |
| 1272 | if (rule->rflags & FILTRULE_PERISHABLE) { |
| 1273 | if (!for_xfer || protocol_version >= 30) |
| 1274 | *op++ = 'p'; |
| 1275 | else if (am_sender) |
| 1276 | return NULL; |
| 1277 | } |
| 1278 | if (op - buf > legal_len) |
| 1279 | return NULL; |
| 1280 | if (legal_len) |
| 1281 | *op++ = ' '; |
| 1282 | *op = '\0'; |
| 1283 | if (plen_ptr) |
| 1284 | *plen_ptr = op - buf; |
| 1285 | return buf; |
| 1286 | } |
| 1287 | |
| 1288 | static void send_rules(int f_out, filter_rule_list *flp) |
| 1289 | { |
| 1290 | filter_rule *ent, *prev = NULL; |
| 1291 | |
| 1292 | for (ent = flp->head; ent; ent = ent->next) { |
| 1293 | unsigned int len, plen, dlen; |
| 1294 | int elide = 0; |
| 1295 | char *p; |
| 1296 | |
| 1297 | /* Note we need to check delete_excluded here in addition to |
| 1298 | * the code in parse_rule_tok() because some rules may have |
| 1299 | * been added before we found the --delete-excluded option. |
| 1300 | * We must also elide any CVS merge-file rules to avoid a |
| 1301 | * backward compatibility problem, and we elide any no-prefix |
| 1302 | * merge files as an optimization (since they can only have |
| 1303 | * include/exclude rules). */ |
| 1304 | if (ent->rflags & FILTRULE_SENDER_SIDE) |
| 1305 | elide = am_sender ? 1 : -1; |
| 1306 | if (ent->rflags & FILTRULE_RECEIVER_SIDE) |
| 1307 | elide = elide ? 0 : am_sender ? -1 : 1; |
| 1308 | else if (delete_excluded && !elide |
| 1309 | && (!(ent->rflags & FILTRULE_PERDIR_MERGE) |
| 1310 | || ent->rflags & FILTRULE_NO_PREFIXES)) |
| 1311 | elide = am_sender ? 1 : -1; |
| 1312 | if (elide < 0) { |
| 1313 | if (prev) |
| 1314 | prev->next = ent->next; |
| 1315 | else |
| 1316 | flp->head = ent->next; |
| 1317 | } else |
| 1318 | prev = ent; |
| 1319 | if (elide > 0) |
| 1320 | continue; |
| 1321 | if (ent->rflags & FILTRULE_CVS_IGNORE |
| 1322 | && !(ent->rflags & FILTRULE_MERGE_FILE)) { |
| 1323 | int f = am_sender || protocol_version < 29 ? f_out : -2; |
| 1324 | send_rules(f, &cvs_filter_list); |
| 1325 | if (f == f_out) |
| 1326 | continue; |
| 1327 | } |
| 1328 | p = get_rule_prefix(ent, ent->pattern, 1, &plen); |
| 1329 | if (!p) { |
| 1330 | rprintf(FERROR, |
| 1331 | "filter rules are too modern for remote rsync.\n"); |
| 1332 | exit_cleanup(RERR_PROTOCOL); |
| 1333 | } |
| 1334 | if (f_out < 0) |
| 1335 | continue; |
| 1336 | len = strlen(ent->pattern); |
| 1337 | dlen = ent->rflags & FILTRULE_DIRECTORY ? 1 : 0; |
| 1338 | if (!(plen + len + dlen)) |
| 1339 | continue; |
| 1340 | write_int(f_out, plen + len + dlen); |
| 1341 | if (plen) |
| 1342 | write_buf(f_out, p, plen); |
| 1343 | write_buf(f_out, ent->pattern, len); |
| 1344 | if (dlen) |
| 1345 | write_byte(f_out, '/'); |
| 1346 | } |
| 1347 | flp->tail = prev; |
| 1348 | } |
| 1349 | |
| 1350 | /* This is only called by the client. */ |
| 1351 | void send_filter_list(int f_out) |
| 1352 | { |
| 1353 | int receiver_wants_list = prune_empty_dirs |
| 1354 | || (delete_mode && (!delete_excluded || protocol_version >= 29)); |
| 1355 | |
| 1356 | if (local_server || (am_sender && !receiver_wants_list)) |
| 1357 | f_out = -1; |
| 1358 | if (cvs_exclude && am_sender) { |
| 1359 | if (protocol_version >= 29) |
| 1360 | parse_filter_str(&filter_list, ":C", rule_template(0), 0); |
| 1361 | parse_filter_str(&filter_list, "-C", rule_template(0), 0); |
| 1362 | } |
| 1363 | |
| 1364 | send_rules(f_out, &filter_list); |
| 1365 | |
| 1366 | if (f_out >= 0) |
| 1367 | write_int(f_out, 0); |
| 1368 | |
| 1369 | if (cvs_exclude) { |
| 1370 | if (!am_sender || protocol_version < 29) |
| 1371 | parse_filter_str(&filter_list, ":C", rule_template(0), 0); |
| 1372 | if (!am_sender) |
| 1373 | parse_filter_str(&filter_list, "-C", rule_template(0), 0); |
| 1374 | } |
| 1375 | } |
| 1376 | |
| 1377 | /* This is only called by the server. */ |
| 1378 | void recv_filter_list(int f_in) |
| 1379 | { |
| 1380 | char line[BIGPATHBUFLEN]; |
| 1381 | int xflags = protocol_version >= 29 ? 0 : XFLG_OLD_PREFIXES; |
| 1382 | int receiver_wants_list = prune_empty_dirs |
| 1383 | || (delete_mode |
| 1384 | && (!delete_excluded || protocol_version >= 29)); |
| 1385 | unsigned int len; |
| 1386 | |
| 1387 | if (!local_server && (am_sender || receiver_wants_list)) { |
| 1388 | while ((len = read_int(f_in)) != 0) { |
| 1389 | if (len >= sizeof line) |
| 1390 | overflow_exit("recv_rules"); |
| 1391 | read_sbuf(f_in, line, len); |
| 1392 | parse_filter_str(&filter_list, line, rule_template(0), xflags); |
| 1393 | } |
| 1394 | } |
| 1395 | |
| 1396 | if (cvs_exclude) { |
| 1397 | if (local_server || am_sender || protocol_version < 29) |
| 1398 | parse_filter_str(&filter_list, ":C", rule_template(0), 0); |
| 1399 | if (local_server || am_sender) |
| 1400 | parse_filter_str(&filter_list, "-C", rule_template(0), 0); |
| 1401 | } |
| 1402 | |
| 1403 | if (local_server) /* filter out any rules that aren't for us. */ |
| 1404 | send_rules(-1, &filter_list); |
| 1405 | } |