1/* 2 * Generate and receive file lists. 3 * 4 * Copyright (C) 1996 Andrew Tridgell 5 * Copyright (C) 1996 Paul Mackerras 6 * Copyright (C) 2001, 2002 Martin Pool <mbp@samba.org> 7 * Copyright (C) 2002, 2003, 2004, 2005, 2006 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 2 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, write to the Free Software Foundation, Inc., 21 * 51 Franklin Street - Fifth Floor, Boston, MA 02110-1301, USA. 22 */ 23 24#include "rsync.h" 25#ifdef HAVE_COPYFILE_H 26#include <copyfile.h> 27#endif /* HAVE_COPYFILE_H */ 28 29extern int verbose; 30extern int list_only; 31extern int am_root; 32extern int am_server; 33extern int am_daemon; 34extern int am_sender; 35extern int do_progress; 36extern int always_checksum; 37extern int module_id; 38extern int ignore_errors; 39extern int numeric_ids; 40extern int recurse; 41extern int xfer_dirs; 42extern int filesfrom_fd; 43extern int one_file_system; 44extern int copy_dirlinks; 45extern int keep_dirlinks; 46extern int preserve_links; 47extern int preserve_hard_links; 48extern int preserve_devices; 49extern int preserve_specials; 50extern int preserve_uid; 51extern int preserve_gid; 52extern int relative_paths; 53extern int implied_dirs; 54extern int prune_empty_dirs; 55extern int copy_links; 56extern int copy_unsafe_links; 57extern int protocol_version; 58extern int sanitize_paths; 59extern int munge_symlinks; 60extern struct stats stats; 61extern struct file_list *the_file_list; 62 63extern char curr_dir[MAXPATHLEN]; 64 65extern struct chmod_mode_struct *chmod_modes; 66 67extern struct filter_list_struct filter_list; 68extern struct filter_list_struct server_filter_list; 69#ifdef EA_SUPPORT 70extern int extended_attributes; 71#endif /* EA_SUPPORT */ 72int io_error; 73int checksum_len; 74dev_t filesystem_dev; /* used to implement -x */ 75unsigned int file_struct_len; 76 77static char empty_sum[MD4_SUM_LENGTH]; 78static int flist_count_offset; 79 80static void clean_flist(struct file_list *flist, int strip_root, int no_dups); 81static void output_flist(struct file_list *flist); 82 83void init_flist(void) 84{ 85 struct file_struct f; 86 87 /* Figure out how big the file_struct is without trailing padding */ 88 file_struct_len = offsetof(struct file_struct, flags) + sizeof f.flags; 89 checksum_len = protocol_version < 21 ? 2 : MD4_SUM_LENGTH; 90} 91 92static int show_filelist_p(void) 93{ 94 return verbose && xfer_dirs && !am_server; 95} 96 97static void start_filelist_progress(char *kind) 98{ 99 rprintf(FCLIENT, "%s ... ", kind); 100 if (verbose > 1 || do_progress) 101 rprintf(FCLIENT, "\n"); 102 rflush(FINFO); 103} 104 105static void emit_filelist_progress(int count) 106{ 107 rprintf(FCLIENT, " %d files...\r", count); 108} 109 110static void maybe_emit_filelist_progress(int count) 111{ 112 if (do_progress && show_filelist_p() && (count % 100) == 0) 113 emit_filelist_progress(count); 114} 115 116static void finish_filelist_progress(const struct file_list *flist) 117{ 118 if (do_progress) { 119 /* This overwrites the progress line */ 120 rprintf(FINFO, "%d file%sto consider\n", 121 flist->count, flist->count == 1 ? " " : "s "); 122 } else 123 rprintf(FINFO, "done\n"); 124} 125 126void show_flist_stats(void) 127{ 128 /* Nothing yet */ 129} 130 131static void list_file_entry(struct file_struct *f) 132{ 133 char permbuf[PERMSTRING_SIZE]; 134 135 if (!f->basename) { 136 /* this can happen if duplicate names were removed */ 137 return; 138 } 139 140 permstring(permbuf, f->mode); 141 142#ifdef SUPPORT_LINKS 143 if (preserve_links && S_ISLNK(f->mode)) { 144 rprintf(FINFO, "%s %11.0f %s %s -> %s\n", 145 permbuf, 146 (double)f->length, timestring(f->modtime), 147 f_name(f, NULL), f->u.link); 148 } else 149#endif 150 { 151 rprintf(FINFO, "%s %11.0f %s %s\n", 152 permbuf, 153 (double)f->length, timestring(f->modtime), 154 f_name(f, NULL)); 155 } 156} 157 158/* Stat either a symlink or its referent, depending on the settings of 159 * copy_links, copy_unsafe_links, etc. Returns -1 on error, 0 on success. 160 * 161 * If path is the name of a symlink, then the linkbuf buffer (which must hold 162 * MAXPATHLEN chars) will be set to the symlink's target string. 163 * 164 * The stat structure pointed to by stp will contain information about the 165 * link or the referent as appropriate, if they exist. */ 166static int readlink_stat(const char *path, STRUCT_STAT *stp, char *linkbuf) 167{ 168#ifdef SUPPORT_LINKS 169 if (link_stat(path, stp, copy_dirlinks) < 0) 170 return -1; 171 if (S_ISLNK(stp->st_mode)) { 172 int llen = readlink(path, linkbuf, MAXPATHLEN - 1); 173 if (llen < 0) 174 return -1; 175 linkbuf[llen] = '\0'; 176 if (copy_unsafe_links && unsafe_symlink(linkbuf, path)) { 177 if (verbose > 1) { 178 rprintf(FINFO,"copying unsafe symlink \"%s\" -> \"%s\"\n", 179 path, linkbuf); 180 } 181 return do_stat(path, stp); 182 } 183 if (munge_symlinks && am_sender && llen > SYMLINK_PREFIX_LEN 184 && strncmp(linkbuf, SYMLINK_PREFIX, SYMLINK_PREFIX_LEN) == 0) { 185 memmove(linkbuf, linkbuf + SYMLINK_PREFIX_LEN, 186 llen - SYMLINK_PREFIX_LEN + 1); 187 } 188 } 189 return 0; 190#else 191 return do_stat(path, stp); 192#endif 193} 194 195int link_stat(const char *path, STRUCT_STAT *stp, int follow_dirlinks) 196{ 197#ifdef SUPPORT_LINKS 198 if (copy_links) 199 return do_stat(path, stp); 200 if (do_lstat(path, stp) < 0) 201 return -1; 202 if (follow_dirlinks && S_ISLNK(stp->st_mode)) { 203 STRUCT_STAT st; 204 if (do_stat(path, &st) == 0 && S_ISDIR(st.st_mode)) 205 *stp = st; 206 } 207 return 0; 208#else 209 return do_stat(path, stp); 210#endif 211} 212 213/* This function is used to check if a file should be included/excluded 214 * from the list of files based on its name and type etc. The value of 215 * filter_level is set to either SERVER_FILTERS or ALL_FILTERS. */ 216static int is_excluded(char *fname, int is_dir, int filter_level) 217{ 218#if 0 /* This currently never happens, so avoid a useless compare. */ 219 if (filter_level == NO_FILTERS) 220 return 0; 221#endif 222 if (fname) { 223 /* never exclude '.', even if somebody does --exclude '*' */ 224 if (fname[0] == '.' && !fname[1]) 225 return 0; 226 /* Handle the -R version of the '.' dir. */ 227 if (fname[0] == '/') { 228 int len = strlen(fname); 229 if (fname[len-1] == '.' && fname[len-2] == '/') 230 return 0; 231 } 232 } 233 if (server_filter_list.head 234 && check_filter(&server_filter_list, fname, is_dir) < 0) 235 return 1; 236 if (filter_level != ALL_FILTERS) 237 return 0; 238 if (filter_list.head 239 && check_filter(&filter_list, fname, is_dir) < 0) 240 return 1; 241 return 0; 242} 243 244static int to_wire_mode(mode_t mode) 245{ 246#ifdef SUPPORT_LINKS 247#if _S_IFLNK != 0120000 248 if (S_ISLNK(mode)) 249 return (mode & ~(_S_IFMT)) | 0120000; 250#endif 251#endif 252 return mode; 253} 254 255static mode_t from_wire_mode(int mode) 256{ 257#if _S_IFLNK != 0120000 258 if ((mode & (_S_IFMT)) == 0120000) 259 return (mode & ~(_S_IFMT)) | _S_IFLNK; 260#endif 261 return mode; 262} 263 264static void send_directory(int f, struct file_list *flist, 265 char *fbuf, int len); 266 267static char *flist_dir; 268static int flist_dir_len; 269 270 271/** 272 * Make sure @p flist is big enough to hold at least @p flist->count 273 * entries. 274 **/ 275void flist_expand(struct file_list *flist) 276{ 277 struct file_struct **new_ptr; 278 279 if (flist->count < flist->malloced) 280 return; 281 282 if (flist->malloced < FLIST_START) 283 flist->malloced = FLIST_START; 284 else if (flist->malloced >= FLIST_LINEAR) 285 flist->malloced += FLIST_LINEAR; 286 else 287 flist->malloced *= 2; 288 289 /* 290 * In case count jumped or we are starting the list 291 * with a known size just set it. 292 */ 293 if (flist->malloced < flist->count) 294 flist->malloced = flist->count; 295 296 new_ptr = realloc_array(flist->files, struct file_struct *, 297 flist->malloced); 298 299 if (verbose >= 2 && flist->malloced != FLIST_START) { 300 rprintf(FCLIENT, "[%s] expand file_list to %.0f bytes, did%s move\n", 301 who_am_i(), 302 (double)sizeof flist->files[0] * flist->malloced, 303 (new_ptr == flist->files) ? " not" : ""); 304 } 305 306 flist->files = new_ptr; 307 308 if (!flist->files) 309 out_of_memory("flist_expand"); 310} 311 312static void send_file_entry(struct file_struct *file, int f) 313{ 314 unsigned short flags; 315 static time_t modtime; 316 static mode_t mode; 317 static int64 dev; 318 static dev_t rdev; 319 static uint32 rdev_major; 320 static uid_t uid; 321 static gid_t gid; 322 static char lastname[MAXPATHLEN]; 323 char fname[MAXPATHLEN]; 324 int l1, l2; 325 326 if (f < 0) 327 return; 328 329 if (!file) { 330 write_byte(f, 0); 331 modtime = 0, mode = 0; 332 dev = 0, rdev = MAKEDEV(0, 0); 333 rdev_major = 0; 334 uid = 0, gid = 0; 335 *lastname = '\0'; 336 return; 337 } 338 339 f_name(file, fname); 340 341 flags = file->flags & XMIT_TOP_DIR; 342 343 if (file->mode == mode) 344 flags |= XMIT_SAME_MODE; 345 else 346 mode = file->mode; 347 if ((preserve_devices && IS_DEVICE(mode)) 348 || (preserve_specials && IS_SPECIAL(mode))) { 349 if (protocol_version < 28) { 350 if (file->u.rdev == rdev) 351 flags |= XMIT_SAME_RDEV_pre28; 352 else 353 rdev = file->u.rdev; 354 } else { 355 rdev = file->u.rdev; 356 if ((uint32)major(rdev) == rdev_major) 357 flags |= XMIT_SAME_RDEV_MAJOR; 358 else 359 rdev_major = major(rdev); 360 if ((uint32)minor(rdev) <= 0xFFu) 361 flags |= XMIT_RDEV_MINOR_IS_SMALL; 362 } 363 } else if (protocol_version < 28) 364 rdev = MAKEDEV(0, 0); 365 if (file->uid == uid) 366 flags |= XMIT_SAME_UID; 367 else 368 uid = file->uid; 369 if (file->gid == gid) 370 flags |= XMIT_SAME_GID; 371 else 372 gid = file->gid; 373 if (file->modtime == modtime) 374 flags |= XMIT_SAME_TIME; 375 else 376 modtime = file->modtime; 377 378#ifdef SUPPORT_HARD_LINKS 379 if (file->link_u.idev) { 380 if (file->F_DEV == dev) { 381 if (protocol_version >= 28) 382 flags |= XMIT_SAME_DEV; 383 } else 384 dev = file->F_DEV; 385 flags |= XMIT_HAS_IDEV_DATA; 386 } 387#endif 388 389 for (l1 = 0; 390 lastname[l1] && (fname[l1] == lastname[l1]) && (l1 < 255); 391 l1++) {} 392 l2 = strlen(fname+l1); 393 394 if (l1 > 0) 395 flags |= XMIT_SAME_NAME; 396 if (l2 > 255) 397 flags |= XMIT_LONG_NAME; 398 399 /* We must make sure we don't send a zero flag byte or the 400 * other end will terminate the flist transfer. Note that 401 * the use of XMIT_TOP_DIR on a non-dir has no meaning, so 402 * it's harmless way to add a bit to the first flag byte. */ 403 if (protocol_version >= 28) { 404 if (!flags && !S_ISDIR(mode)) 405 flags |= XMIT_TOP_DIR; 406 if ((flags & 0xFF00) || !flags) { 407 flags |= XMIT_EXTENDED_FLAGS; 408 write_byte(f, flags); 409 write_byte(f, flags >> 8); 410 } else 411 write_byte(f, flags); 412 } else { 413 if (!(flags & 0xFF)) 414 flags |= S_ISDIR(mode) ? XMIT_LONG_NAME : XMIT_TOP_DIR; 415 write_byte(f, flags); 416 } 417 if (flags & XMIT_SAME_NAME) 418 write_byte(f, l1); 419 if (flags & XMIT_LONG_NAME) 420 write_int(f, l2); 421 else 422 write_byte(f, l2); 423 write_buf(f, fname + l1, l2); 424 425 write_longint(f, file->length); 426 if (!(flags & XMIT_SAME_TIME)) 427 write_int(f, modtime); 428 if (!(flags & XMIT_SAME_MODE)) 429 write_int(f, to_wire_mode(mode)); 430 if (preserve_uid && !(flags & XMIT_SAME_UID)) { 431 if (!numeric_ids) 432 add_uid(uid); 433 write_int(f, uid); 434 } 435 if (preserve_gid && !(flags & XMIT_SAME_GID)) { 436 if (!numeric_ids) 437 add_gid(gid); 438 write_int(f, gid); 439 } 440 if ((preserve_devices && IS_DEVICE(mode)) 441 || (preserve_specials && IS_SPECIAL(mode))) { 442 if (protocol_version < 28) { 443 if (!(flags & XMIT_SAME_RDEV_pre28)) 444 write_int(f, (int)rdev); 445 } else { 446 if (!(flags & XMIT_SAME_RDEV_MAJOR)) 447 write_int(f, major(rdev)); 448 if (flags & XMIT_RDEV_MINOR_IS_SMALL) 449 write_byte(f, minor(rdev)); 450 else 451 write_int(f, minor(rdev)); 452 } 453 } 454 455#ifdef SUPPORT_LINKS 456 if (preserve_links && S_ISLNK(mode)) { 457 int len = strlen(file->u.link); 458 write_int(f, len); 459 write_buf(f, file->u.link, len); 460 } 461#endif 462 463#ifdef SUPPORT_HARD_LINKS 464 if (file->link_u.idev) { 465 if (protocol_version < 26) { 466 /* 32-bit dev_t and ino_t */ 467 write_int(f, dev); 468 write_int(f, file->F_INODE); 469 } else { 470 /* 64-bit dev_t and ino_t */ 471 if (!(flags & XMIT_SAME_DEV)) 472 write_longint(f, dev); 473 write_longint(f, file->F_INODE); 474 } 475 } 476#endif 477 478 if (always_checksum && (S_ISREG(mode) || protocol_version < 28)) { 479 char *sum; 480 if (S_ISREG(mode)) 481 sum = file->u.sum; 482 else { 483 /* Prior to 28, we sent a useless set of nulls. */ 484 sum = empty_sum; 485 } 486 write_buf(f, sum, checksum_len); 487 } 488 489 strlcpy(lastname, fname, MAXPATHLEN); 490} 491 492static struct file_struct *receive_file_entry(struct file_list *flist, 493 unsigned short flags, int f) 494{ 495 static time_t modtime; 496 static mode_t mode; 497 static int64 dev; 498 static dev_t rdev; 499 static uint32 rdev_major; 500 static uid_t uid; 501 static gid_t gid; 502 static char lastname[MAXPATHLEN], *lastdir; 503 static int lastdir_depth, lastdir_len = -1; 504 static unsigned int del_hier_name_len = 0; 505 static int in_del_hier = 0; 506 char thisname[MAXPATHLEN]; 507 unsigned int l1 = 0, l2 = 0; 508 int alloc_len, basename_len, dirname_len, linkname_len, sum_len; 509 OFF_T file_length; 510 char *basename, *dirname, *bp; 511 struct file_struct *file; 512 513 if (!flist) { 514 modtime = 0, mode = 0; 515 dev = 0, rdev = MAKEDEV(0, 0); 516 rdev_major = 0; 517 uid = 0, gid = 0; 518 *lastname = '\0'; 519 lastdir_len = -1; 520 in_del_hier = 0; 521 return NULL; 522 } 523 524 if (flags & XMIT_SAME_NAME) 525 l1 = read_byte(f); 526 527 if (flags & XMIT_LONG_NAME) 528 l2 = read_int(f); 529 else 530 l2 = read_byte(f); 531 532 if (l2 >= MAXPATHLEN - l1) { 533 rprintf(FERROR, 534 "overflow: flags=0x%x l1=%d l2=%d lastname=%s\n", 535 flags, l1, l2, lastname); 536 overflow_exit("receive_file_entry"); 537 } 538 539 strlcpy(thisname, lastname, l1 + 1); 540 read_sbuf(f, &thisname[l1], l2); 541 thisname[l1 + l2] = 0; 542 543 strlcpy(lastname, thisname, MAXPATHLEN); 544 545 clean_fname(thisname, 0); 546 547 if (sanitize_paths) 548 sanitize_path(thisname, thisname, "", 0, NULL); 549 550 if ((basename = strrchr(thisname, '/')) != NULL) { 551 dirname_len = ++basename - thisname; /* counts future '\0' */ 552 if (lastdir_len == dirname_len - 1 553 && strncmp(thisname, lastdir, lastdir_len) == 0) { 554 dirname = lastdir; 555 dirname_len = 0; /* indicates no copy is needed */ 556 } else 557 dirname = thisname; 558 } else { 559 basename = thisname; 560 dirname = NULL; 561 dirname_len = 0; 562 } 563 basename_len = strlen(basename) + 1; /* count the '\0' */ 564 565 file_length = read_longint(f); 566 if (!(flags & XMIT_SAME_TIME)) 567 modtime = (time_t)read_int(f); 568 if (!(flags & XMIT_SAME_MODE)) 569 mode = from_wire_mode(read_int(f)); 570 571 if (chmod_modes && !S_ISLNK(mode)) 572 mode = tweak_mode(mode, chmod_modes); 573 574 if (preserve_uid && !(flags & XMIT_SAME_UID)) 575 uid = (uid_t)read_int(f); 576 if (preserve_gid && !(flags & XMIT_SAME_GID)) 577 gid = (gid_t)read_int(f); 578 579 if ((preserve_devices && IS_DEVICE(mode)) 580 || (preserve_specials && IS_SPECIAL(mode))) { 581 if (protocol_version < 28) { 582 if (!(flags & XMIT_SAME_RDEV_pre28)) 583 rdev = (dev_t)read_int(f); 584 } else { 585 uint32 rdev_minor; 586 if (!(flags & XMIT_SAME_RDEV_MAJOR)) 587 rdev_major = read_int(f); 588 if (flags & XMIT_RDEV_MINOR_IS_SMALL) 589 rdev_minor = read_byte(f); 590 else 591 rdev_minor = read_int(f); 592 rdev = MAKEDEV(rdev_major, rdev_minor); 593 } 594 } else if (protocol_version < 28) 595 rdev = MAKEDEV(0, 0); 596 597#ifdef SUPPORT_LINKS 598 if (preserve_links && S_ISLNK(mode)) { 599 linkname_len = read_int(f) + 1; /* count the '\0' */ 600 if (linkname_len <= 0 || linkname_len > MAXPATHLEN) { 601 rprintf(FERROR, "overflow: linkname_len=%d\n", 602 linkname_len - 1); 603 overflow_exit("receive_file_entry"); 604 } 605 if (munge_symlinks) 606 linkname_len += SYMLINK_PREFIX_LEN; 607 } 608 else 609#endif 610 linkname_len = 0; 611 612 sum_len = always_checksum && S_ISREG(mode) ? MD4_SUM_LENGTH : 0; 613 614 alloc_len = file_struct_len + dirname_len + basename_len 615 + linkname_len + sum_len; 616 bp = pool_alloc(flist->file_pool, alloc_len, "receive_file_entry"); 617 618 file = (struct file_struct *)bp; 619 memset(bp, 0, file_struct_len); 620 bp += file_struct_len; 621 622 file->modtime = modtime; 623 file->length = file_length; 624 file->mode = mode; 625 file->uid = uid; 626 file->gid = gid; 627 628 if (dirname_len) { 629 file->dirname = lastdir = bp; 630 lastdir_len = dirname_len - 1; 631 memcpy(bp, dirname, dirname_len - 1); 632 bp += dirname_len; 633 bp[-1] = '\0'; 634 lastdir_depth = count_dir_elements(lastdir); 635 file->dir.depth = lastdir_depth + 1; 636 } else if (dirname) { 637 file->dirname = dirname; /* we're reusing lastname */ 638 file->dir.depth = lastdir_depth + 1; 639 } else 640 file->dir.depth = 1; 641 642 if (S_ISDIR(mode)) { 643 if (basename_len == 1+1 && *basename == '.') /* +1 for '\0' */ 644 file->dir.depth--; 645 if (flags & XMIT_TOP_DIR) { 646 in_del_hier = recurse; 647 del_hier_name_len = file->dir.depth == 0 ? 0 : l1 + l2; 648 if (relative_paths && del_hier_name_len > 2 649 && lastname[del_hier_name_len-1] == '.' 650 && lastname[del_hier_name_len-2] == '/') 651 del_hier_name_len -= 2; 652 file->flags |= FLAG_TOP_DIR | FLAG_DEL_HERE; 653 } else if (in_del_hier) { 654 if (!relative_paths || !del_hier_name_len 655 || (l1 >= del_hier_name_len 656 && lastname[del_hier_name_len] == '/')) 657 file->flags |= FLAG_DEL_HERE; 658 else 659 in_del_hier = 0; 660 } 661 } 662 663 file->basename = bp; 664 memcpy(bp, basename, basename_len); 665 bp += basename_len; 666 667 if ((preserve_devices && IS_DEVICE(mode)) 668 || (preserve_specials && IS_SPECIAL(mode))) 669 file->u.rdev = rdev; 670 671#ifdef SUPPORT_LINKS 672 if (linkname_len) { 673 file->u.link = bp; 674 if (munge_symlinks) { 675 strlcpy(bp, SYMLINK_PREFIX, linkname_len); 676 bp += SYMLINK_PREFIX_LEN; 677 linkname_len -= SYMLINK_PREFIX_LEN; 678 } 679 read_sbuf(f, bp, linkname_len - 1); 680 if (sanitize_paths && !munge_symlinks) { 681 sanitize_path(bp, bp, "", lastdir_depth, NULL); 682 bp += strlen(bp) + 1; 683 } else 684 bp += linkname_len; 685 } 686#endif 687 688#ifdef SUPPORT_HARD_LINKS 689 if (preserve_hard_links && protocol_version < 28 && S_ISREG(mode)) 690 flags |= XMIT_HAS_IDEV_DATA; 691 if (flags & XMIT_HAS_IDEV_DATA) { 692 int64 inode; 693 if (protocol_version < 26) { 694 dev = read_int(f); 695 inode = read_int(f); 696 } else { 697 if (!(flags & XMIT_SAME_DEV)) 698 dev = read_longint(f); 699 inode = read_longint(f); 700 } 701 if (flist->hlink_pool) { 702 file->link_u.idev = pool_talloc(flist->hlink_pool, 703 struct idev, 1, "inode_table"); 704 file->F_INODE = inode; 705 file->F_DEV = dev; 706 } 707 } 708#endif 709 710 if (always_checksum && (sum_len || protocol_version < 28)) { 711 char *sum; 712 if (sum_len) { 713 file->u.sum = sum = bp; 714 /*bp += sum_len;*/ 715 } else { 716 /* Prior to 28, we get a useless set of nulls. */ 717 sum = empty_sum; 718 } 719 read_buf(f, sum, checksum_len); 720 } 721 722 return file; 723} 724 725/** 726 * Create a file_struct for a named file by reading its stat() 727 * information and performing extensive checks against global 728 * options. 729 * 730 * @return the new file, or NULL if there was an error or this file 731 * should be excluded. 732 * 733 * @todo There is a small optimization opportunity here to avoid 734 * stat()ing the file in some circumstances, which has a certain cost. 735 * We are called immediately after doing readdir(), and so we may 736 * already know the d_type of the file. We could for example avoid 737 * statting directories if we're not recursing, but this is not a very 738 * important case. Some systems may not have d_type. 739 **/ 740struct file_struct *make_file(char *fname, struct file_list *flist, 741 STRUCT_STAT *stp, unsigned short flags, 742 int filter_level) 743{ 744 static char *lastdir; 745 static int lastdir_len = -1; 746 struct file_struct *file; 747 STRUCT_STAT st; 748 char sum[SUM_LENGTH]; 749 char thisname[MAXPATHLEN]; 750 char linkname[MAXPATHLEN]; 751 int alloc_len, basename_len, dirname_len, linkname_len, sum_len; 752 char *basename, *dirname, *bp; 753 754 if (!flist || !flist->count) /* Ignore lastdir when invalid. */ 755 lastdir_len = -1; 756 757 if (strlcpy(thisname, fname, sizeof thisname) 758 >= sizeof thisname - flist_dir_len) { 759 rprintf(FINFO, "skipping overly long name: %s\n", fname); 760 return NULL; 761 } 762 clean_fname(thisname, 0); 763 if (sanitize_paths) 764 sanitize_path(thisname, thisname, "", 0, NULL); 765 766 memset(sum, 0, SUM_LENGTH); 767 768 if (stp && S_ISDIR(stp->st_mode)) { 769 st = *stp; /* Needed for "symlink/." with --relative. */ 770 *linkname = '\0'; /* make IBM code checker happy */ 771 } else if (readlink_stat(thisname, &st, linkname) != 0) { 772 int save_errno = errno; 773 /* See if file is excluded before reporting an error. */ 774 if (filter_level != NO_FILTERS 775 && is_excluded(thisname, 0, filter_level)) 776 return NULL; 777 if (save_errno == ENOENT) { 778#ifdef SUPPORT_LINKS 779 /* Avoid "vanished" error if symlink points nowhere. */ 780 if (copy_links && do_lstat(thisname, &st) == 0 781 && S_ISLNK(st.st_mode)) { 782 io_error |= IOERR_GENERAL; 783 rprintf(FERROR, "symlink has no referent: %s\n", 784 full_fname(thisname)); 785 } else 786#endif 787 { 788 enum logcode c = am_daemon && protocol_version < 28 789 ? FERROR : FINFO; 790 io_error |= IOERR_VANISHED; 791 rprintf(c, "file has vanished: %s\n", 792 full_fname(thisname)); 793 } 794 } else { 795 io_error |= IOERR_GENERAL; 796 rsyserr(FERROR, save_errno, "readlink %s failed", 797 full_fname(thisname)); 798 } 799 return NULL; 800 } 801 802 /* backup.c calls us with filter_level set to NO_FILTERS. */ 803 if (filter_level == NO_FILTERS) 804 goto skip_filters; 805 806 if (S_ISDIR(st.st_mode) && !xfer_dirs) { 807 rprintf(FINFO, "skipping directory %s\n", thisname); 808 return NULL; 809 } 810 811 /* We only care about directories because we need to avoid recursing 812 * into a mount-point directory, not to avoid copying a symlinked 813 * file if -L (or similar) was specified. */ 814 if (one_file_system && st.st_dev != filesystem_dev 815 && S_ISDIR(st.st_mode)) { 816 if (one_file_system > 1) { 817 if (verbose > 2) { 818 rprintf(FINFO, "skipping mount-point dir %s\n", 819 thisname); 820 } 821 return NULL; 822 } 823 flags |= FLAG_MOUNT_POINT; 824 } 825 826 if (is_excluded(thisname, S_ISDIR(st.st_mode) != 0, filter_level)) 827 return NULL; 828 829 if (lp_ignore_nonreadable(module_id)) { 830#ifdef SUPPORT_LINKS 831 if (!S_ISLNK(st.st_mode)) 832#endif 833 if (access(thisname, R_OK) != 0) 834 return NULL; 835 } 836 837 skip_filters: 838 839 if (verbose > 2) { 840 rprintf(FINFO, "[%s] make_file(%s,*,%d)\n", 841 who_am_i(), thisname, filter_level); 842 } 843 844 if ((basename = strrchr(thisname, '/')) != NULL) { 845 dirname_len = ++basename - thisname; /* counts future '\0' */ 846 if (lastdir_len == dirname_len - 1 847 && strncmp(thisname, lastdir, lastdir_len) == 0) { 848 dirname = lastdir; 849 dirname_len = 0; /* indicates no copy is needed */ 850 } else 851 dirname = thisname; 852 } else { 853 basename = thisname; 854 dirname = NULL; 855 dirname_len = 0; 856 } 857 basename_len = strlen(basename) + 1; /* count the '\0' */ 858 859#ifdef SUPPORT_LINKS 860 linkname_len = S_ISLNK(st.st_mode) ? strlen(linkname) + 1 : 0; 861#else 862 linkname_len = 0; 863#endif 864 865 sum_len = always_checksum && am_sender && S_ISREG(st.st_mode) 866 ? MD4_SUM_LENGTH : 0; 867 868 alloc_len = file_struct_len + dirname_len + basename_len 869 + linkname_len + sum_len; 870 if (flist) 871 bp = pool_alloc(flist->file_pool, alloc_len, "make_file"); 872 else { 873 if (!(bp = new_array(char, alloc_len))) 874 out_of_memory("make_file"); 875 } 876 877 file = (struct file_struct *)bp; 878 memset(bp, 0, file_struct_len); 879 bp += file_struct_len; 880 881 file->flags = flags; 882 file->modtime = st.st_mtime; 883 file->length = st.st_size; 884 file->mode = st.st_mode; 885 file->uid = st.st_uid; 886 file->gid = st.st_gid; 887 888#ifdef SUPPORT_HARD_LINKS 889 if (flist && flist->hlink_pool) { 890 if (protocol_version < 28) { 891 if (S_ISREG(st.st_mode)) 892 file->link_u.idev = pool_talloc( 893 flist->hlink_pool, struct idev, 1, 894 "inode_table"); 895 } else { 896 if (!S_ISDIR(st.st_mode) && st.st_nlink > 1) 897 file->link_u.idev = pool_talloc( 898 flist->hlink_pool, struct idev, 1, 899 "inode_table"); 900 } 901 } 902 if (file->link_u.idev) { 903 file->F_DEV = st.st_dev; 904 file->F_INODE = st.st_ino; 905 } 906#endif 907 908 if (dirname_len) { 909 file->dirname = lastdir = bp; 910 lastdir_len = dirname_len - 1; 911 memcpy(bp, dirname, dirname_len - 1); 912 bp += dirname_len; 913 bp[-1] = '\0'; 914 } else if (dirname) 915 file->dirname = dirname; 916 917 file->basename = bp; 918 memcpy(bp, basename, basename_len); 919 bp += basename_len; 920 921#ifdef HAVE_STRUCT_STAT_ST_RDEV 922 if ((preserve_devices && IS_DEVICE(st.st_mode)) 923 || (preserve_specials && IS_SPECIAL(st.st_mode))) 924 file->u.rdev = st.st_rdev; 925#endif 926 927#ifdef SUPPORT_LINKS 928 if (linkname_len) { 929 file->u.link = bp; 930 memcpy(bp, linkname, linkname_len); 931 bp += linkname_len; 932 } 933#endif 934 935 if (sum_len) { 936 file->u.sum = bp; 937 file_checksum(thisname, bp, st.st_size); 938 /*bp += sum_len;*/ 939 } 940 941 file->dir.root = flist_dir; 942 943 /* This code is only used by the receiver when it is building 944 * a list of files for a delete pass. */ 945 if (keep_dirlinks && linkname_len && flist) { 946 STRUCT_STAT st2; 947 int save_mode = file->mode; 948 file->mode = S_IFDIR; /* Find a directory with our name. */ 949 if (flist_find(the_file_list, file) >= 0 950 && do_stat(thisname, &st2) == 0 && S_ISDIR(st2.st_mode)) { 951 file->modtime = st2.st_mtime; 952 file->length = st2.st_size; 953 file->mode = st2.st_mode; 954 file->uid = st2.st_uid; 955 file->gid = st2.st_gid; 956 file->u.link = NULL; 957 } else 958 file->mode = save_mode; 959 } 960 961 if (S_ISREG(st.st_mode) || S_ISLNK(st.st_mode)) 962 stats.total_size += st.st_size; 963 964 return file; 965} 966 967static struct file_struct *send_file_name(int f, struct file_list *flist, 968 char *fname, STRUCT_STAT *stp, 969 unsigned short flags) 970{ 971 struct file_struct *file; 972 973 file = make_file(fname, flist, stp, flags, 974 f == -2 ? SERVER_FILTERS : ALL_FILTERS); 975 if (!file) 976 return NULL; 977 978 if (chmod_modes && !S_ISLNK(file->mode)) 979 file->mode = tweak_mode(file->mode, chmod_modes); 980 981 maybe_emit_filelist_progress(flist->count + flist_count_offset); 982 983 flist_expand(flist); 984 985 if (file->basename[0]) { 986 flist->files[flist->count++] = file; 987 send_file_entry(file, f); 988#ifdef EA_SUPPORT 989 /* If the file doesn't begin with "._", and has 990 * either acls or extended attributes, serialize 991 * the data out and add a fake file to the file 992 * list with the "._" prepended to the basename. 993 * of the original source. 994 */ 995 if (extended_attributes) { 996#ifdef HAVE_COPYFILE 997 if(strncmp(file->basename, "._", 2) 998 && copyfile(fname, NULL, 0, 999 COPYFILE_CHECK | COPYFILE_ACL | COPYFILE_XATTR | (preserve_links ? COPYFILE_NOFOLLOW : 0))) { 1000 char *bp; 1001 struct file_struct *file2 = NULL; 1002 int alloc_len; 1003 1004 if (verbose > 4) 1005 rprintf(FINFO, "added synthetic file for: %s\n", fname); 1006 1007 alloc_len = file_struct_len 1008 + strlen(file->basename) + 1 + 2 // + strlen("._") 1009 + MD4_SUM_LENGTH; 1010 1011 bp = pool_alloc(flist->file_pool, alloc_len, "receive_metadata_entry"); 1012 1013 file2 = (struct file_struct *)bp; 1014 memcpy(bp, file, file_struct_len); 1015 bp += file_struct_len; 1016 1017 file2->basename = bp; 1018 memcpy(bp + 2, file->basename, strlen(file->basename) + 1); 1019 file2->basename[0] = '.'; 1020 file2->basename[1] = '_'; 1021 file2->length = 1; 1022 file2->mode = S_IFREG | S_IRUSR; 1023 file2->modtime = file->modtime; 1024 file2->u.sum = empty_sum; 1025 1026 flist_expand(flist); 1027 flist->files[flist->count++] = file2; 1028 1029 if (f != -1) 1030 write_byte(f, 1); 1031 send_file_entry(file2, f); 1032 } 1033#endif /* HAVE_COPYFILE */ 1034 if (f != -1) 1035 write_byte(f, 0); 1036 } 1037#endif /* EA_SUPPORT */ 1038 } 1039 return file; 1040} 1041 1042static void send_if_directory(int f, struct file_list *flist, 1043 struct file_struct *file, 1044 char *fbuf, unsigned int ol) 1045{ 1046 char is_dot_dir = fbuf[ol-1] == '.' && (ol == 1 || fbuf[ol-2] == '/'); 1047 1048 if (S_ISDIR(file->mode) 1049 && !(file->flags & FLAG_MOUNT_POINT) && f_name(file, fbuf)) { 1050 void *save_filters; 1051 unsigned int len = strlen(fbuf); 1052 if (len > 1 && fbuf[len-1] == '/') 1053 fbuf[--len] = '\0'; 1054 if (len >= MAXPATHLEN - 1) { 1055 io_error |= IOERR_GENERAL; 1056 rprintf(FERROR, "skipping long-named directory: %s\n", 1057 full_fname(fbuf)); 1058 return; 1059 } 1060 save_filters = push_local_filters(fbuf, len); 1061 send_directory(f, flist, fbuf, len); 1062 pop_local_filters(save_filters); 1063 fbuf[ol] = '\0'; 1064 if (is_dot_dir) 1065 fbuf[ol-1] = '.'; 1066 } 1067} 1068 1069/* This function is normally called by the sender, but the receiving side also 1070 * calls it from get_dirlist() with f set to -1 so that we just construct the 1071 * file list in memory without sending it over the wire. Also, get_dirlist() 1072 * might call this with f set to -2, which also indicates that local filter 1073 * rules should be ignored. */ 1074static void send_directory(int f, struct file_list *flist, 1075 char *fbuf, int len) 1076{ 1077 struct dirent *di; 1078 unsigned remainder; 1079 char *p; 1080 DIR *d; 1081 int start = flist->count; 1082 1083 if (!(d = opendir(fbuf))) { 1084 io_error |= IOERR_GENERAL; 1085 rsyserr(FERROR, errno, "opendir %s failed", full_fname(fbuf)); 1086 return; 1087 } 1088 1089 p = fbuf + len; 1090 if (len != 1 || *fbuf != '/') 1091 *p++ = '/'; 1092 *p = '\0'; 1093 remainder = MAXPATHLEN - (p - fbuf); 1094 1095 for (errno = 0, di = readdir(d); di; errno = 0, di = readdir(d)) { 1096 char *dname = d_name(di); 1097 if (dname[0] == '.' && (dname[1] == '\0' 1098 || (dname[1] == '.' && dname[2] == '\0'))) 1099 continue; 1100 if (strlcpy(p, dname, remainder) >= remainder) { 1101 io_error |= IOERR_GENERAL; 1102 rprintf(FINFO, 1103 "cannot send long-named file %s\n", 1104 full_fname(fbuf)); 1105 continue; 1106 } 1107 1108 send_file_name(f, flist, fbuf, NULL, 0); 1109 } 1110 1111 fbuf[len] = '\0'; 1112 1113 if (errno) { 1114 io_error |= IOERR_GENERAL; 1115 rsyserr(FERROR, errno, "readdir(%s)", full_fname(fbuf)); 1116 } 1117 1118 closedir(d); 1119 1120 if (recurse) { 1121 int i, end = flist->count - 1; 1122 for (i = start; i <= end; i++) 1123 send_if_directory(f, flist, flist->files[i], fbuf, len); 1124 } 1125} 1126 1127struct file_list *send_file_list(int f, int argc, char *argv[]) 1128{ 1129 int len; 1130 STRUCT_STAT st; 1131 char *p, *dir, olddir[sizeof curr_dir]; 1132 char lastpath[MAXPATHLEN] = ""; 1133 struct file_list *flist; 1134 struct timeval start_tv, end_tv; 1135 int64 start_write; 1136 int use_ff_fd = 0; 1137 1138 rprintf(FLOG, "building file list\n"); 1139 if (show_filelist_p()) 1140 start_filelist_progress("building file list"); 1141 1142 start_write = stats.total_written; 1143 gettimeofday(&start_tv, NULL); 1144 1145 flist = flist_new(WITH_HLINK, "send_file_list"); 1146 1147 io_start_buffering_out(); 1148 if (filesfrom_fd >= 0) { 1149 if (argv[0] && !push_dir(argv[0], 0)) { 1150 rsyserr(FERROR, errno, "push_dir %s failed", 1151 full_fname(argv[0])); 1152 exit_cleanup(RERR_FILESELECT); 1153 } 1154 use_ff_fd = 1; 1155 } 1156 1157 while (1) { 1158 char fbuf[MAXPATHLEN]; 1159 char *fn; 1160 int is_dot_dir; 1161 1162 if (use_ff_fd) { 1163 if (read_filesfrom_line(filesfrom_fd, fbuf) == 0) 1164 break; 1165 sanitize_path(fbuf, fbuf, "", 0, NULL); 1166 } else { 1167 if (argc-- == 0) 1168 break; 1169 strlcpy(fbuf, *argv++, MAXPATHLEN); 1170 if (sanitize_paths) 1171 sanitize_path(fbuf, fbuf, "", 0, NULL); 1172 } 1173 1174 len = strlen(fbuf); 1175 if (relative_paths) { 1176 /* We clean up fbuf below. */ 1177 is_dot_dir = 0; 1178 } else if (!len || fbuf[len - 1] == '/') { 1179 if (len == 2 && fbuf[0] == '.') { 1180 /* Turn "./" into just "." rather than "./." */ 1181 fbuf[1] = '\0'; 1182 } else { 1183 if (len + 1 >= MAXPATHLEN) 1184 overflow_exit("send_file_list"); 1185 fbuf[len++] = '.'; 1186 fbuf[len] = '\0'; 1187 } 1188 is_dot_dir = 1; 1189 } else if (len > 1 && fbuf[len-1] == '.' && fbuf[len-2] == '.' 1190 && (len == 2 || fbuf[len-3] == '/')) { 1191 if (len + 2 >= MAXPATHLEN) 1192 overflow_exit("send_file_list"); 1193 fbuf[len++] = '/'; 1194 fbuf[len++] = '.'; 1195 fbuf[len] = '\0'; 1196 is_dot_dir = 1; 1197 } else { 1198 is_dot_dir = fbuf[len-1] == '.' 1199 && (len == 1 || fbuf[len-2] == '/'); 1200 } 1201 1202 if (link_stat(fbuf, &st, copy_dirlinks) != 0) { 1203 io_error |= IOERR_GENERAL; 1204 rsyserr(FERROR, errno, "link_stat %s failed", 1205 full_fname(fbuf)); 1206 continue; 1207 } 1208 1209 if (S_ISDIR(st.st_mode) && !xfer_dirs) { 1210 rprintf(FINFO, "skipping directory %s\n", fbuf); 1211 continue; 1212 } 1213 1214 dir = NULL; 1215 olddir[0] = '\0'; 1216 1217 if (!relative_paths) { 1218 p = strrchr(fbuf, '/'); 1219 if (p) { 1220 *p = '\0'; 1221 if (p == fbuf) 1222 dir = "/"; 1223 else 1224 dir = fbuf; 1225 len -= p - fbuf + 1; 1226 fn = p + 1; 1227 } else 1228 fn = fbuf; 1229 } else { 1230 if ((p = strstr(fbuf, "/./")) != NULL) { 1231 *p = '\0'; 1232 if (p == fbuf) 1233 dir = "/"; 1234 else 1235 dir = fbuf; 1236 len -= p - fbuf + 3; 1237 fn = p + 3; 1238 } else 1239 fn = fbuf; 1240 /* Get rid of trailing "/" and "/.". */ 1241 while (len) { 1242 if (fn[len - 1] == '/') { 1243 is_dot_dir = 1; 1244 if (!--len && !dir) { 1245 len++; 1246 break; 1247 } 1248 } 1249 else if (len >= 2 && fn[len - 1] == '.' 1250 && fn[len - 2] == '/') { 1251 is_dot_dir = 1; 1252 if (!(len -= 2) && !dir) { 1253 len++; 1254 break; 1255 } 1256 } else 1257 break; 1258 } 1259 if (len == 1 && fn[0] == '/') 1260 fn[len++] = '.'; 1261 fn[len] = '\0'; 1262 /* Reject a ".." dir in the active part of the path. */ 1263 for (p = fn; (p = strstr(p, "..")) != NULL; p += 2) { 1264 if ((p[2] == '/' || p[2] == '\0') 1265 && (p == fn || p[-1] == '/')) { 1266 rprintf(FERROR, 1267 "found \"..\" dir in relative path: %s\n", 1268 fbuf); 1269 exit_cleanup(RERR_SYNTAX); 1270 } 1271 } 1272 } 1273 1274 if (!*fn) { 1275 len = 1; 1276 fn = "."; 1277 } 1278 1279 if (dir && *dir) { 1280 static char *lastdir; 1281 static int lastdir_len; 1282 1283 strlcpy(olddir, curr_dir, sizeof olddir); 1284 1285 if (!push_dir(dir, 0)) { 1286 io_error |= IOERR_GENERAL; 1287 rsyserr(FERROR, errno, "push_dir %s failed", 1288 full_fname(dir)); 1289 continue; 1290 } 1291 1292 if (lastdir && strcmp(lastdir, dir) == 0) { 1293 flist_dir = lastdir; 1294 flist_dir_len = lastdir_len; 1295 } else { 1296 flist_dir = lastdir = strdup(dir); 1297 flist_dir_len = lastdir_len = strlen(dir); 1298 } 1299 } 1300 1301 if (fn != fbuf) 1302 memmove(fbuf, fn, len + 1); 1303 1304 if (implied_dirs && (p=strrchr(fbuf,'/')) && p != fbuf) { 1305 /* Send the implied directories at the start of the 1306 * source spec, so we get their permissions right. */ 1307 char *lp = lastpath, *slash = fbuf; 1308 *p = '\0'; 1309 /* Skip any initial directories in our path that we 1310 * have in common with lastpath. */ 1311 for (fn = fbuf; *fn && *lp == *fn; lp++, fn++) { 1312 if (*fn == '/') 1313 slash = fn; 1314 } 1315 *p = '/'; 1316 if (fn != p || (*lp && *lp != '/')) { 1317 int save_copy_links = copy_links; 1318 int save_xfer_dirs = xfer_dirs; 1319 copy_links |= copy_unsafe_links; 1320 xfer_dirs = 1; 1321 while ((slash = strchr(slash+1, '/')) != 0) { 1322 *slash = '\0'; 1323 send_file_name(f, flist, fbuf, NULL, 0); 1324 *slash = '/'; 1325 } 1326 copy_links = save_copy_links; 1327 xfer_dirs = save_xfer_dirs; 1328 *p = '\0'; 1329 strlcpy(lastpath, fbuf, sizeof lastpath); 1330 *p = '/'; 1331 } 1332 } 1333 1334 if (one_file_system) 1335 filesystem_dev = st.st_dev; 1336 1337 if (recurse || (xfer_dirs && is_dot_dir)) { 1338 struct file_struct *file; 1339 file = send_file_name(f, flist, fbuf, &st, FLAG_TOP_DIR); 1340 if (file) 1341 send_if_directory(f, flist, file, fbuf, len); 1342 } else 1343 send_file_name(f, flist, fbuf, &st, 0); 1344 1345 if (olddir[0]) { 1346 flist_dir = NULL; 1347 flist_dir_len = 0; 1348 if (!pop_dir(olddir)) { 1349 rsyserr(FERROR, errno, "pop_dir %s failed", 1350 full_fname(olddir)); 1351 exit_cleanup(RERR_FILESELECT); 1352 } 1353 } 1354 } 1355 1356 gettimeofday(&end_tv, NULL); 1357 stats.flist_buildtime = (int64)(end_tv.tv_sec - start_tv.tv_sec) * 1000 1358 + (end_tv.tv_usec - start_tv.tv_usec) / 1000; 1359 if (stats.flist_buildtime == 0) 1360 stats.flist_buildtime = 1; 1361 start_tv = end_tv; 1362 1363 send_file_entry(NULL, f); 1364 1365 if (show_filelist_p()) 1366 finish_filelist_progress(flist); 1367 1368 gettimeofday(&end_tv, NULL); 1369 stats.flist_xfertime = (int64)(end_tv.tv_sec - start_tv.tv_sec) * 1000 1370 + (end_tv.tv_usec - start_tv.tv_usec) / 1000; 1371 1372 if (flist->hlink_pool) { 1373 pool_destroy(flist->hlink_pool); 1374 flist->hlink_pool = NULL; 1375 } 1376 1377 /* Sort the list without removing any duplicates. This allows the 1378 * receiving side to ask for any name they like, which gives us the 1379 * flexibility to change the way we unduplicate names in the future 1380 * without causing a compatibility problem with older versions. */ 1381 clean_flist(flist, 0, 0); 1382 1383 send_uid_list(f); 1384 1385 /* send the io_error flag */ 1386 write_int(f, lp_ignore_errors(module_id) ? 0 : io_error); 1387 1388 io_end_buffering(); 1389 stats.flist_size = stats.total_written - start_write; 1390 stats.num_files = flist->count; 1391 1392 if (verbose > 3) 1393 output_flist(flist); 1394 1395 if (verbose > 2) 1396 rprintf(FINFO, "send_file_list done\n"); 1397 1398 return flist; 1399} 1400 1401struct file_list *recv_file_list(int f) 1402{ 1403 struct file_list *flist; 1404 unsigned short flags; 1405 int64 start_read; 1406 1407 rprintf(FLOG, "receiving file list\n"); 1408 if (show_filelist_p()) 1409 start_filelist_progress("receiving file list"); 1410 1411 start_read = stats.total_read; 1412 1413 flist = flist_new(WITH_HLINK, "recv_file_list"); 1414 1415 flist->count = 0; 1416 flist->malloced = 1000; 1417 flist->files = new_array(struct file_struct *, flist->malloced); 1418 if (!flist->files) 1419 goto oom; 1420 1421 while ((flags = read_byte(f)) != 0) { 1422 struct file_struct *file; 1423 1424 flist_expand(flist); 1425 1426 if (protocol_version >= 28 && (flags & XMIT_EXTENDED_FLAGS)) 1427 flags |= read_byte(f) << 8; 1428 file = receive_file_entry(flist, flags, f); 1429#ifdef EA_SUPPORT 1430 if (extended_attributes) { 1431 if (read_byte(f)) { 1432 file->flags &= ~FLAG_CLEAR_METADATA; 1433 if (verbose > 4) 1434 rprintf(FINFO, "receiving synthetic file entry for: %s/%s\n", 1435 file->dirname ? file->dirname : ".", 1436 file->basename); 1437 } else if (strncmp(file->basename, "._", 2)) { 1438 file->flags |= FLAG_CLEAR_METADATA; 1439 } 1440 } 1441#endif 1442 1443 if (S_ISREG(file->mode) || S_ISLNK(file->mode)) 1444 stats.total_size += file->length; 1445 1446 flist->files[flist->count++] = file; 1447 1448 maybe_emit_filelist_progress(flist->count); 1449 1450 if (verbose > 2) { 1451 rprintf(FINFO, "recv_file_name(%s)\n", 1452 f_name(file, NULL)); 1453 } 1454 } 1455 receive_file_entry(NULL, 0, 0); /* Signal that we're done. */ 1456 1457 if (verbose > 2) 1458 rprintf(FINFO, "received %d names\n", flist->count); 1459 1460 if (show_filelist_p()) 1461 finish_filelist_progress(flist); 1462 1463 clean_flist(flist, relative_paths, 1); 1464 1465 if (f >= 0) { 1466 recv_uid_list(f, flist); 1467 1468 /* Recv the io_error flag */ 1469 if (lp_ignore_errors(module_id) || ignore_errors) 1470 read_int(f); 1471 else 1472 io_error |= read_int(f); 1473 } 1474 1475 if (verbose > 3) 1476 output_flist(flist); 1477 1478 if (list_only) { 1479 int i; 1480 for (i = 0; i < flist->count; i++) 1481 list_file_entry(flist->files[i]); 1482 } 1483 1484 if (verbose > 2) 1485 rprintf(FINFO, "recv_file_list done\n"); 1486 1487 stats.flist_size = stats.total_read - start_read; 1488 stats.num_files = flist->count; 1489 1490 return flist; 1491 1492 oom: 1493 out_of_memory("recv_file_list"); 1494 return NULL; /* not reached */ 1495} 1496 1497static int file_compare(struct file_struct **file1, struct file_struct **file2) 1498{ 1499 return f_name_cmp(*file1, *file2); 1500} 1501 1502/* Search for an identically-named item in the file list. Note that the 1503 * items must agree in their directory-ness, or no match is returned. */ 1504int flist_find(struct file_list *flist, struct file_struct *f) 1505{ 1506 int low = flist->low, high = flist->high; 1507 int diff, mid, mid_up; 1508 1509 while (low <= high) { 1510 mid = (low + high) / 2; 1511 if (flist->files[mid]->basename) 1512 mid_up = mid; 1513 else { 1514 /* Scan for the next non-empty entry using the cached 1515 * distance values. If the value isn't fully up-to- 1516 * date, update it. */ 1517 mid_up = mid + flist->files[mid]->dir.depth; 1518 if (!flist->files[mid_up]->basename) { 1519 do { 1520 mid_up += flist->files[mid_up]->dir.depth; 1521 } while (!flist->files[mid_up]->basename); 1522 flist->files[mid]->dir.depth = mid_up - mid; 1523 } 1524 if (mid_up > high) { 1525 /* If there's nothing left above us, set high to 1526 * a non-empty entry below us and continue. */ 1527 high = mid - flist->files[mid]->length; 1528 if (!flist->files[high]->basename) { 1529 do { 1530 high -= flist->files[high]->length; 1531 } while (!flist->files[high]->basename); 1532 flist->files[mid]->length = mid - high; 1533 } 1534 continue; 1535 } 1536 } 1537 diff = f_name_cmp(flist->files[mid_up], f); 1538 if (diff == 0) { 1539 if (protocol_version < 29 1540 && S_ISDIR(flist->files[mid_up]->mode) 1541 != S_ISDIR(f->mode)) 1542 return -1; 1543 return mid_up; 1544 } 1545 if (diff < 0) 1546 low = mid_up + 1; 1547 else 1548 high = mid - 1; 1549 } 1550 return -1; 1551} 1552 1553/* 1554 * Free up any resources a file_struct has allocated 1555 * and clear the file. 1556 */ 1557void clear_file(struct file_struct *file, struct file_list *flist) 1558{ 1559 if (flist->hlink_pool && file->link_u.idev) 1560 pool_free(flist->hlink_pool, 0, file->link_u.idev); 1561 memset(file, 0, file_struct_len); 1562 /* In an empty entry, dir.depth is an offset to the next non-empty 1563 * entry. Likewise for length in the opposite direction. We assume 1564 * that we're alone for now since flist_find() will adjust the counts 1565 * it runs into that aren't up-to-date. */ 1566 file->length = file->dir.depth = 1; 1567} 1568 1569/* 1570 * allocate a new file list 1571 */ 1572struct file_list *flist_new(int with_hlink, char *msg) 1573{ 1574 struct file_list *flist; 1575 1576 flist = new(struct file_list); 1577 if (!flist) 1578 out_of_memory(msg); 1579 1580 memset(flist, 0, sizeof (struct file_list)); 1581 1582 if (!(flist->file_pool = pool_create(FILE_EXTENT, 0, 1583 out_of_memory, POOL_INTERN))) 1584 out_of_memory(msg); 1585 1586#ifdef SUPPORT_HARD_LINKS 1587 if (with_hlink && preserve_hard_links) { 1588 if (!(flist->hlink_pool = pool_create(HLINK_EXTENT, 1589 sizeof (struct idev), out_of_memory, POOL_INTERN))) 1590 out_of_memory(msg); 1591 } 1592#endif 1593 1594 return flist; 1595} 1596 1597/* 1598 * free up all elements in a flist 1599 */ 1600void flist_free(struct file_list *flist) 1601{ 1602 pool_destroy(flist->file_pool); 1603 pool_destroy(flist->hlink_pool); 1604 free(flist->files); 1605 free(flist); 1606} 1607 1608/* 1609 * This routine ensures we don't have any duplicate names in our file list. 1610 * duplicate names can cause corruption because of the pipelining 1611 */ 1612static void clean_flist(struct file_list *flist, int strip_root, int no_dups) 1613{ 1614 char fbuf[MAXPATHLEN]; 1615 int i, prev_i = 0; 1616 1617 if (!flist) 1618 return; 1619 if (flist->count == 0) { 1620 flist->high = -1; 1621 return; 1622 } 1623 1624 qsort(flist->files, flist->count, 1625 sizeof flist->files[0], (int (*)())file_compare); 1626 1627 for (i = no_dups? 0 : flist->count; i < flist->count; i++) { 1628 if (flist->files[i]->basename) { 1629 prev_i = i; 1630 break; 1631 } 1632 } 1633 flist->low = prev_i; 1634 while (++i < flist->count) { 1635 int j; 1636 struct file_struct *file = flist->files[i]; 1637 1638 if (!file->basename) 1639 continue; 1640 if (f_name_cmp(file, flist->files[prev_i]) == 0) 1641 j = prev_i; 1642 else if (protocol_version >= 29 && S_ISDIR(file->mode)) { 1643 int save_mode = file->mode; 1644 /* Make sure that this directory doesn't duplicate a 1645 * non-directory earlier in the list. */ 1646 flist->high = prev_i; 1647 file->mode = S_IFREG; 1648 j = flist_find(flist, file); 1649 file->mode = save_mode; 1650 } else 1651 j = -1; 1652 if (j >= 0) { 1653 struct file_struct *fp = flist->files[j]; 1654 int keep, drop; 1655 /* If one is a dir and the other is not, we want to 1656 * keep the dir because it might have contents in the 1657 * list. */ 1658 if (S_ISDIR(file->mode) != S_ISDIR(fp->mode)) { 1659 if (S_ISDIR(file->mode)) 1660 keep = i, drop = j; 1661 else 1662 keep = j, drop = i; 1663 } else 1664 keep = j, drop = i; 1665 if (verbose > 1 && !am_server) { 1666 rprintf(FINFO, 1667 "removing duplicate name %s from file list (%d)\n", 1668 f_name(file, fbuf), drop); 1669 } 1670 /* Make sure we don't lose track of a user-specified 1671 * top directory. */ 1672 flist->files[keep]->flags |= flist->files[drop]->flags 1673 & (FLAG_TOP_DIR|FLAG_DEL_HERE); 1674 1675 clear_file(flist->files[drop], flist); 1676 1677 if (keep == i) { 1678 if (flist->low == drop) { 1679 for (j = drop + 1; 1680 j < i && !flist->files[j]->basename; 1681 j++) {} 1682 flist->low = j; 1683 } 1684 prev_i = i; 1685 } 1686 } else 1687 prev_i = i; 1688 } 1689 flist->high = no_dups ? prev_i : flist->count - 1; 1690 1691 if (strip_root) { 1692 /* We need to strip off the leading slashes for relative 1693 * paths, but this must be done _after_ the sorting phase. */ 1694 for (i = flist->low; i <= flist->high; i++) { 1695 struct file_struct *file = flist->files[i]; 1696 1697 if (!file->dirname) 1698 continue; 1699 while (*file->dirname == '/') 1700 file->dirname++; 1701 if (!*file->dirname) 1702 file->dirname = NULL; 1703 } 1704 } 1705 1706 if (prune_empty_dirs && no_dups) { 1707 int j, prev_depth = 0; 1708 1709 prev_i = 0; /* It's OK that this isn't really true. */ 1710 1711 for (i = flist->low; i <= flist->high; i++) { 1712 struct file_struct *fp, *file = flist->files[i]; 1713 1714 /* This temporarily abuses the dir.depth value for a 1715 * directory that is in a chain that might get pruned. 1716 * We restore the old value if it gets a reprieve. */ 1717 if (S_ISDIR(file->mode) && file->dir.depth) { 1718 /* Dump empty dirs when coming back down. */ 1719 for (j = prev_depth; j >= file->dir.depth; j--) { 1720 fp = flist->files[prev_i]; 1721 if (fp->dir.depth >= 0) 1722 break; 1723 prev_i = -fp->dir.depth-1; 1724 clear_file(fp, flist); 1725 } 1726 prev_depth = file->dir.depth; 1727 if (is_excluded(f_name(file, fbuf), 1, 1728 ALL_FILTERS)) { 1729 /* Keep dirs through this dir. */ 1730 for (j = prev_depth-1; ; j--) { 1731 fp = flist->files[prev_i]; 1732 if (fp->dir.depth >= 0) 1733 break; 1734 prev_i = -fp->dir.depth-1; 1735 fp->dir.depth = j; 1736 } 1737 } else 1738 file->dir.depth = -prev_i-1; 1739 prev_i = i; 1740 } else { 1741 /* Keep dirs through this non-dir. */ 1742 for (j = prev_depth; ; j--) { 1743 fp = flist->files[prev_i]; 1744 if (fp->dir.depth >= 0) 1745 break; 1746 prev_i = -fp->dir.depth-1; 1747 fp->dir.depth = j; 1748 } 1749 } 1750 } 1751 /* Dump empty all remaining empty dirs. */ 1752 while (1) { 1753 struct file_struct *fp = flist->files[prev_i]; 1754 if (fp->dir.depth >= 0) 1755 break; 1756 prev_i = -fp->dir.depth-1; 1757 clear_file(fp, flist); 1758 } 1759 1760 for (i = flist->low; i <= flist->high; i++) { 1761 if (flist->files[i]->basename) 1762 break; 1763 } 1764 flist->low = i; 1765 for (i = flist->high; i >= flist->low; i--) { 1766 if (flist->files[i]->basename) 1767 break; 1768 } 1769 flist->high = i; 1770 } 1771} 1772 1773static void output_flist(struct file_list *flist) 1774{ 1775 char uidbuf[16], gidbuf[16], depthbuf[16]; 1776 struct file_struct *file; 1777 const char *who = who_am_i(); 1778 int i; 1779 1780 for (i = 0; i < flist->count; i++) { 1781 file = flist->files[i]; 1782 if ((am_root || am_sender) && preserve_uid) 1783 snprintf(uidbuf, sizeof uidbuf, " uid=%ld", (long)file->uid); 1784 else 1785 *uidbuf = '\0'; 1786 if (preserve_gid && file->gid != GID_NONE) 1787 snprintf(gidbuf, sizeof gidbuf, " gid=%ld", (long)file->gid); 1788 else 1789 *gidbuf = '\0'; 1790 if (!am_sender) 1791 snprintf(depthbuf, sizeof depthbuf, "%d", file->dir.depth); 1792 rprintf(FINFO, "[%s] i=%d %s %s%s%s%s mode=0%o len=%.0f%s%s flags=%x\n", 1793 who, i, am_sender ? NS(file->dir.root) : depthbuf, 1794 file->dirname ? file->dirname : "", 1795 file->dirname ? "/" : "", NS(file->basename), 1796 S_ISDIR(file->mode) ? "/" : "", (int)file->mode, 1797 (double)file->length, uidbuf, gidbuf, file->flags); 1798 } 1799} 1800 1801enum fnc_state { s_DIR, s_SLASH, s_BASE, s_TRAILING }; 1802enum fnc_type { t_PATH, t_ITEM }; 1803 1804/* Compare the names of two file_struct entities, similar to how strcmp() 1805 * would do if it were operating on the joined strings. 1806 * 1807 * Some differences beginning with protocol_version 29: (1) directory names 1808 * are compared with an assumed trailing slash so that they compare in a 1809 * way that would cause them to sort immediately prior to any content they 1810 * may have; (2) a directory of any name compares after a non-directory of 1811 * any name at the same depth; (3) a directory with name "." compares prior 1812 * to anything else. These changes mean that a directory and a non-dir 1813 * with the same name will not compare as equal (protocol_version >= 29). 1814 * 1815 * The dirname component can be an empty string, but the basename component 1816 * cannot (and never is in the current codebase). The basename component 1817 * may be NULL (for a removed item), in which case it is considered to be 1818 * after any existing item. */ 1819int f_name_cmp(struct file_struct *f1, struct file_struct *f2) 1820{ 1821 int dif; 1822 const uchar *c1, *c2; 1823 enum fnc_state state1, state2; 1824 enum fnc_type type1, type2; 1825 enum fnc_type t_path = protocol_version >= 29 ? t_PATH : t_ITEM; 1826 1827 if (!f1 || !f1->basename) { 1828 if (!f2 || !f2->basename) 1829 return 0; 1830 return -1; 1831 } 1832 if (!f2 || !f2->basename) 1833 return 1; 1834 1835 c1 = (uchar*)f1->dirname; 1836 c2 = (uchar*)f2->dirname; 1837 if (c1 == c2) 1838 c1 = c2 = NULL; 1839 if (!c1) { 1840 type1 = S_ISDIR(f1->mode) ? t_path : t_ITEM; 1841 c1 = (uchar*)f1->basename; 1842 if (type1 == t_PATH && *c1 == '.' && !c1[1]) { 1843 type1 = t_ITEM; 1844 state1 = s_TRAILING; 1845 c1 = (uchar*)""; 1846 } else 1847 state1 = s_BASE; 1848 } else { 1849 type1 = t_path; 1850 state1 = s_DIR; 1851 } 1852 if (!c2) { 1853 type2 = S_ISDIR(f2->mode) ? t_path : t_ITEM; 1854 c2 = (uchar*)f2->basename; 1855 if (type2 == t_PATH && *c2 == '.' && !c2[1]) { 1856 type2 = t_ITEM; 1857 state2 = s_TRAILING; 1858 c2 = (uchar*)""; 1859 } else 1860 state2 = s_BASE; 1861 } else { 1862 type2 = t_path; 1863 state2 = s_DIR; 1864 } 1865 1866 if (type1 != type2) 1867 return type1 == t_PATH ? 1 : -1; 1868 1869 do { 1870 if (!*c1) { 1871 switch (state1) { 1872 case s_DIR: 1873 state1 = s_SLASH; 1874 c1 = (uchar*)"/"; 1875 break; 1876 case s_SLASH: 1877 type1 = S_ISDIR(f1->mode) ? t_path : t_ITEM; 1878 c1 = (uchar*)f1->basename; 1879 if (type1 == t_PATH && *c1 == '.' && !c1[1]) { 1880 type1 = t_ITEM; 1881 state1 = s_TRAILING; 1882 c1 = (uchar*)""; 1883 } else 1884 state1 = s_BASE; 1885 break; 1886 case s_BASE: 1887 state1 = s_TRAILING; 1888 if (type1 == t_PATH) { 1889 c1 = (uchar*)"/"; 1890 break; 1891 } 1892 /* FALL THROUGH */ 1893 case s_TRAILING: 1894 type1 = t_ITEM; 1895 break; 1896 } 1897 if (*c2 && type1 != type2) 1898 return type1 == t_PATH ? 1 : -1; 1899 } 1900 if (!*c2) { 1901 switch (state2) { 1902 case s_DIR: 1903 state2 = s_SLASH; 1904 c2 = (uchar*)"/"; 1905 break; 1906 case s_SLASH: 1907 type2 = S_ISDIR(f2->mode) ? t_path : t_ITEM; 1908 c2 = (uchar*)f2->basename; 1909 if (type2 == t_PATH && *c2 == '.' && !c2[1]) { 1910 type2 = t_ITEM; 1911 state2 = s_TRAILING; 1912 c2 = (uchar*)""; 1913 } else 1914 state2 = s_BASE; 1915 break; 1916 case s_BASE: 1917 state2 = s_TRAILING; 1918 if (type2 == t_PATH) { 1919 c2 = (uchar*)"/"; 1920 break; 1921 } 1922 /* FALL THROUGH */ 1923 case s_TRAILING: 1924 if (!*c1) 1925 return 0; 1926 type2 = t_ITEM; 1927 break; 1928 } 1929 if (type1 != type2) 1930 return type1 == t_PATH ? 1 : -1; 1931 } 1932 } while ((dif = (int)*c1++ - (int)*c2++) == 0); 1933 1934 return dif; 1935} 1936 1937/* Return a copy of the full filename of a flist entry, using the indicated 1938 * buffer or one of 5 static buffers if fbuf is NULL. No size-checking is 1939 * done because we checked the size when creating the file_struct entry. 1940 */ 1941char *f_name(struct file_struct *f, char *fbuf) 1942{ 1943 if (!f || !f->basename) 1944 return NULL; 1945 1946 if (!fbuf) { 1947 static char names[5][MAXPATHLEN]; 1948 static unsigned int n; 1949 1950 n = (n + 1) % (sizeof names / sizeof names[0]); 1951 1952 fbuf = names[n]; 1953 } 1954 1955 if (f->dirname) { 1956 int len = strlen(f->dirname); 1957 memcpy(fbuf, f->dirname, len); 1958 fbuf[len] = '/'; 1959 strlcpy(fbuf + len + 1, f->basename, MAXPATHLEN - (len + 1)); 1960 } else 1961 strlcpy(fbuf, f->basename, MAXPATHLEN); 1962 1963 return fbuf; 1964} 1965 1966/* Do a non-recursive scan of the named directory, possibly ignoring all 1967 * exclude rules except for the daemon's. If "dlen" is >=0, it is the length 1968 * of the dirname string, and also indicates that "dirname" is a MAXPATHLEN 1969 * buffer (the functions we call will append names onto the end, but the old 1970 * dir value will be restored on exit). */ 1971struct file_list *get_dirlist(char *dirname, int dlen, 1972 int ignore_filter_rules) 1973{ 1974 struct file_list *dirlist; 1975 char dirbuf[MAXPATHLEN]; 1976 int save_recurse = recurse; 1977 int save_xfer_dirs = xfer_dirs; 1978 1979 if (dlen < 0) { 1980 dlen = strlcpy(dirbuf, dirname, MAXPATHLEN); 1981 if (dlen >= MAXPATHLEN) 1982 return NULL; 1983 dirname = dirbuf; 1984 } 1985 1986 dirlist = flist_new(WITHOUT_HLINK, "get_dirlist"); 1987 1988 recurse = 0; 1989 xfer_dirs = 1; 1990 send_directory(ignore_filter_rules ? -2 : -1, dirlist, dirname, dlen); 1991 xfer_dirs = save_xfer_dirs; 1992 recurse = save_recurse; 1993 if (do_progress) 1994 flist_count_offset += dirlist->count; 1995 1996 clean_flist(dirlist, 0, 0); 1997 1998 if (verbose > 3) 1999 output_flist(dirlist); 2000 2001 return dirlist; 2002} 2003