1 /* -*- c-file-style: "linux" -*-
3 rsync -- fast file replication program
5 Copyright (C) 1996-2000 by Andrew Tridgell
6 Copyright (C) Paul Mackerras 1996
7 Copyright (C) 2002 by Martin Pool <mbp@samba.org>
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.
14 This program is distributed in the hope that it will be useful,
15 but WITHOUT ANY WARRANTY; without even the implied warranty of
16 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17 GNU General Public License for more details.
19 You should have received a copy of the GNU General Public License
20 along with this program; if not, write to the Free Software
21 Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
28 extern int relative_paths;
29 extern int keep_dirlinks;
30 extern int preserve_links;
32 extern int preserve_devices;
33 extern int preserve_hard_links;
34 extern int preserve_perms;
35 extern int preserve_uid;
36 extern int preserve_gid;
37 extern int preserve_times;
38 extern int omit_dir_times;
39 extern int delete_during;
40 extern int update_only;
41 extern int opt_ignore_existing;
43 extern int make_backups;
44 extern int csum_length;
45 extern int ignore_times;
47 extern OFF_T max_size;
48 extern int io_timeout;
49 extern int protocol_version;
50 extern int fuzzy_basis;
51 extern int always_checksum;
52 extern char *partial_dir;
53 extern char *basis_dir[];
54 extern int compare_dest;
56 extern int whole_file;
57 extern int local_server;
59 extern int read_batch;
60 extern int only_existing;
61 extern int orig_umask;
62 extern int safe_symlinks;
63 extern long block_size; /* "long" because popt can't set an int32. */
65 extern struct filter_list_struct server_filter_list;
67 static int unchanged_attrs(struct file_struct *file, STRUCT_STAT *st)
70 && (st->st_mode & CHMOD_BITS) != (file->mode & CHMOD_BITS))
73 if (am_root && preserve_uid && st->st_uid != file->uid)
76 if (preserve_gid && file->gid != GID_NONE && st->st_gid != file->gid)
82 /* Perform our quick-check heuristic for determining if a file is unchanged. */
83 static int unchanged_file(char *fn, struct file_struct *file, STRUCT_STAT *st)
85 if (st->st_size != file->length)
88 /* if always checksum is set then we use the checksum instead
89 of the file time to determine whether to sync */
90 if (always_checksum && S_ISREG(st->st_mode)) {
91 char sum[MD4_SUM_LENGTH];
92 file_checksum(fn, sum, st->st_size);
93 return memcmp(sum, file->u.sum, protocol_version < 21 ? 2
94 : MD4_SUM_LENGTH) == 0;
103 return cmp_modtime(st->st_mtime, file->modtime) == 0;
108 * set (initialize) the size entries in the per-file sum_struct
109 * calculating dynamic block and checksum sizes.
111 * This is only called from generate_and_send_sums() but is a separate
112 * function to encapsulate the logic.
114 * The block size is a rounded square root of file length.
116 * The checksum size is determined according to:
117 * blocksum_bits = BLOCKSUM_EXP + 2*log2(file_len) - log2(block_len)
118 * provided by Donovan Baarda which gives a probability of rsync
119 * algorithm corrupting data and falling back using the whole md4
122 * This might be made one of several selectable heuristics.
124 static void sum_sizes_sqroot(struct sum_struct *sum, int64 len)
130 blength = block_size;
131 else if (len <= BLOCK_SIZE * BLOCK_SIZE)
132 blength = BLOCK_SIZE;
137 for (c = 1, l = len, cnt = 0; l >>= 2; c <<= 1, cnt++) {}
138 if (cnt >= 31 || c >= MAX_BLOCK_SIZE)
139 blength = MAX_BLOCK_SIZE;
144 if (len < (int64)blength * blength)
147 } while (c >= 8); /* round to multiple of 8 */
148 blength = MAX(blength, BLOCK_SIZE);
152 if (protocol_version < 27) {
153 s2length = csum_length;
154 } else if (csum_length == SUM_LENGTH) {
155 s2length = SUM_LENGTH;
159 int b = BLOCKSUM_BIAS;
160 for (l = len; l >>= 1; b += 2) {}
161 for (c = blength; c >>= 1 && b; b--) {}
162 /* add a bit, subtract rollsum, round up. */
163 s2length = (b + 1 - 32 + 7) / 8; /* --optimize in compiler-- */
164 s2length = MAX(s2length, csum_length);
165 s2length = MIN(s2length, SUM_LENGTH);
169 sum->blength = blength;
170 sum->s2length = s2length;
171 sum->count = (len + (blength - 1)) / blength;
172 sum->remainder = (len % blength);
174 if (sum->count && verbose > 2) {
176 "count=%.0f rem=%ld blength=%ld s2length=%d flength=%.0f\n",
177 (double)sum->count, (long)sum->remainder, (long)sum->blength,
178 sum->s2length, (double)sum->flength);
184 * Generate and send a stream of signatures/checksums that describe a buffer
186 * Generate approximately one checksum every block_len bytes.
188 static void generate_and_send_sums(int fd, OFF_T len, int f_out, int f_copy)
191 struct map_struct *mapbuf;
192 struct sum_struct sum;
195 sum_sizes_sqroot(&sum, len);
198 mapbuf = map_file(fd, len, MAX_MAP_SIZE, sum.blength);
202 write_sum_head(f_out, &sum);
204 for (i = 0; i < sum.count; i++) {
205 int32 n1 = (int32)MIN(len, (OFF_T)sum.blength);
206 char *map = map_ptr(mapbuf, offset, n1);
207 uint32 sum1 = get_checksum1(map, n1);
208 char sum2[SUM_LENGTH];
211 full_write(f_copy, map, n1);
213 get_checksum2(map, n1, sum2);
217 "chunk[%.0f] offset=%.0f len=%ld sum1=%08lx\n",
218 (double)i, (double)offset, (long)n1,
219 (unsigned long)sum1);
221 write_int(f_out, sum1);
222 write_buf(f_out, sum2, sum.s2length);
231 /* Try to find a filename in the same dir as "fname" with a similar name. */
232 static int find_fuzzy(struct file_struct *file, struct file_list *dirlist)
234 int fname_len, fname_suf_len;
235 const char *fname_suf, *fname = file->basename;
236 uint32 lowest_dist = 0x7FFFFFFF;
237 int j, lowest_j = -1;
239 fname_len = strlen(fname);
240 fname_suf = find_filename_suffix(fname, fname_len, &fname_suf_len);
242 for (j = 0; j < dirlist->count; j++) {
243 struct file_struct *fp = dirlist->files[j];
244 const char *suf, *name;
248 if (!S_ISREG(fp->mode) || !fp->length
249 || fp->flags & FLAG_NO_FUZZY)
254 if (fp->length == file->length
255 && fp->modtime == file->modtime) {
258 "fuzzy size/modtime match for %s\n",
265 suf = find_filename_suffix(name, len, &suf_len);
267 dist = fuzzy_distance(name, len, fname, fname_len);
268 /* Add some extra weight to how well the suffixes match. */
269 dist += fuzzy_distance(suf, suf_len, fname_suf, fname_suf_len)
272 rprintf(FINFO, "fuzzy distance for %s = %d.%05d\n",
273 name, (int)(dist>>16), (int)(dist&0xFFFF));
275 if (dist <= lowest_dist) {
285 /* Acts on flist->file's ndx'th item, whose name is fname. If a directory,
286 * make sure it exists, and has the right permissions/timestamp info. For
287 * all other non-regular files (symlinks, etc.) we create them here. For
288 * regular files that have changed, we try to find a basis file and then
289 * start sending checksums.
291 * Note that f_out is set to -1 when doing final directory-permission and
292 * modification-time repair. */
293 static void recv_generator(char *fname, struct file_list *flist,
294 struct file_struct *file, int ndx,
295 int f_out, int f_out_name)
297 static int missing_below = -1;
298 static char *fuzzy_dirname = NULL;
299 static struct file_list *fuzzy_dirlist = NULL;
300 struct file_struct *fuzzy_file = NULL;
301 int fd = -1, f_copy = -1;
302 STRUCT_STAT st, partial_st;
303 struct file_struct *back_file = NULL;
304 int statret, stat_errno;
305 char *fnamecmp, *partialptr, *backupptr = NULL;
306 char fnamecmpbuf[MAXPATHLEN];
314 flist_free(fuzzy_dirlist);
315 fuzzy_dirlist = NULL;
316 fuzzy_dirname = NULL;
318 if (missing_below >= 0) {
326 rprintf(FINFO, "recv_generator(%s,%d)\n",
327 safe_fname(fname), ndx);
330 if (server_filter_list.head
331 && check_filter(&server_filter_list, fname,
332 S_ISDIR(file->mode)) < 0) {
334 rprintf(FINFO, "skipping server-excluded file \"%s\"\n",
340 if (missing_below >= 0 && file->dir.depth <= missing_below) {
348 if (fuzzy_basis && S_ISREG(file->mode)) {
349 char *dn = file->dirname ? file->dirname : ".";
350 /* Yes, identical dirnames are guaranteed to have
351 * identical pointers at this point. */
352 if (fuzzy_dirname != dn) {
354 flist_free(fuzzy_dirlist);
356 fuzzy_dirlist = get_dirlist(fuzzy_dirname, 1);
360 statret = link_stat(fname, &st,
361 keep_dirlinks && S_ISDIR(file->mode));
365 if (only_existing && statret == -1 && stat_errno == ENOENT) {
366 /* we only want to update existing files */
368 rprintf(FINFO, "not creating new file \"%s\"\n",
374 if (statret == 0 && !preserve_perms
375 && S_ISDIR(st.st_mode) == S_ISDIR(file->mode)) {
376 /* if the file exists already and we aren't perserving
377 * permissions then act as though the remote end sent
378 * us the file permissions we already have */
379 file->mode = (file->mode & ~CHMOD_BITS)
380 | (st.st_mode & CHMOD_BITS);
383 if (S_ISDIR(file->mode)) {
384 /* The file to be received is a directory, so we need
385 * to prepare appropriately. If there is already a
386 * file of that name and it is *not* a directory, then
387 * we need to delete it. If it doesn't exist, then
388 * (perhaps recursively) create it. */
389 if (statret == 0 && !S_ISDIR(st.st_mode)) {
390 delete_file(fname, DEL_TERSE);
393 if (dry_run && statret != 0 && missing_below < 0) {
394 missing_below = file->dir.depth;
397 if (statret != 0 && do_mkdir(fname,file->mode) != 0 && errno != EEXIST) {
398 if (!relative_paths || errno != ENOENT
399 || create_directory_path(fname, orig_umask) < 0
400 || do_mkdir(fname, file->mode) < 0) {
401 rsyserr(FERROR, errno,
402 "recv_generator: mkdir %s failed",
406 if (set_perms(fname, file, statret ? NULL : &st, 0)
407 && verbose && f_out != -1)
408 rprintf(FINFO, "%s/\n", safe_fname(fname));
409 if (delete_during && f_out != -1 && csum_length != SUM_LENGTH
410 && (file->flags & FLAG_DEL_HERE))
411 delete_in_dir(flist, fname, file);
413 } else if (max_size && file->length > max_size) {
415 rprintf(FINFO, "%s is over max-size\n",
421 if (preserve_links && S_ISLNK(file->mode)) {
423 if (safe_symlinks && unsafe_symlink(file->u.link, fname)) {
426 "ignoring unsafe symlink %s -> \"%s\"\n",
428 safe_fname(file->u.link));
433 int dflag = S_ISDIR(st.st_mode) ? DEL_DIR : 0;
434 char lnk[MAXPATHLEN];
438 && (len = readlink(fname, lnk, MAXPATHLEN-1)) > 0) {
440 /* A link already pointing to the
441 * right place -- no further action
443 if (strcmp(lnk, file->u.link) == 0) {
444 set_perms(fname, file, &st,
449 /* Not the right symlink (or not a symlink), so
451 delete_file(fname, dflag | DEL_TERSE);
453 if (do_symlink(file->u.link,fname) != 0) {
454 rsyserr(FERROR, errno, "symlink %s -> \"%s\" failed",
455 full_fname(fname), safe_fname(file->u.link));
457 set_perms(fname,file,NULL,0);
459 rprintf(FINFO, "%s -> %s\n", safe_fname(fname),
460 safe_fname(file->u.link));
467 if (am_root && preserve_devices && IS_DEVICE(file->mode)) {
469 st.st_mode != file->mode ||
470 st.st_rdev != file->u.rdev) {
471 int dflag = S_ISDIR(st.st_mode) ? DEL_DIR : 0;
472 delete_file(fname, dflag | DEL_TERSE);
474 rprintf(FINFO,"mknod(%s,0%o,0x%x)\n",
476 (int)file->mode, (int)file->u.rdev);
478 if (do_mknod(fname,file->mode,file->u.rdev) != 0) {
479 rsyserr(FERROR, errno, "mknod %s failed",
482 set_perms(fname,file,NULL,0);
484 rprintf(FINFO, "%s\n",
489 set_perms(fname, file, &st, PERMS_REPORT);
494 if (preserve_hard_links && hard_link_check(file, HL_CHECK_MASTER))
497 if (!S_ISREG(file->mode)) {
498 rprintf(FINFO, "skipping non-regular file \"%s\"\n",
504 fnamecmp_type = FNAMECMP_FNAME;
506 if (statret == -1 && basis_dir[0] != NULL) {
507 int fallback_match = -1;
511 pathjoin(fnamecmpbuf, sizeof fnamecmpbuf,
512 basis_dir[i], fname);
513 if (link_stat(fnamecmpbuf, &st, 0) == 0
514 && S_ISREG(st.st_mode)) {
520 } else if (match_level == 2
521 && !unchanged_attrs(file, &st))
523 if (!unchanged_file(fnamecmpbuf, file, &st))
527 if (!unchanged_attrs(file, &st))
533 } while (basis_dir[++i] != NULL);
535 if (match_level < 3) {
537 pathjoin(fnamecmpbuf, sizeof fnamecmpbuf,
538 basis_dir[i], fname);
541 if (link_dest && match_level == 3 && !dry_run) {
542 if (do_link(fnamecmpbuf, fname) < 0) {
544 rsyserr(FINFO, errno,
546 full_fname(fnamecmpbuf),
549 fnamecmp = fnamecmpbuf;
555 fnamecmp = fnamecmpbuf;
561 if (statret == 0 && !S_ISREG(st.st_mode)) {
562 int dflag = S_ISDIR(st.st_mode) ? DEL_DIR : 0;
563 if (delete_file(fname, dflag | DEL_TERSE) != 0)
569 if (partial_dir && (partialptr = partial_dir_fname(fname)) != NULL
570 && link_stat(partialptr, &partial_st, 0) == 0
571 && S_ISREG(partial_st.st_mode)) {
573 goto prepare_to_open;
577 if (statret == -1 && fuzzy_basis && dry_run <= 1) {
578 int j = find_fuzzy(file, fuzzy_dirlist);
580 fuzzy_file = fuzzy_dirlist->files[j];
581 f_name_to(fuzzy_file, fnamecmpbuf);
583 rprintf(FINFO, "fuzzy basis selected for %s: %s\n",
584 safe_fname(fname), safe_fname(fnamecmpbuf));
586 st.st_mode = fuzzy_file->mode;
587 st.st_size = fuzzy_file->length;
588 st.st_mtime = fuzzy_file->modtime;
590 fnamecmp = fnamecmpbuf;
591 fnamecmp_type = FNAMECMP_FUZZY;
596 if (preserve_hard_links && hard_link_check(file, HL_SKIP))
598 if (stat_errno == ENOENT)
601 rsyserr(FERROR, stat_errno,
602 "recv_generator: failed to stat %s",
608 if (opt_ignore_existing && fnamecmp_type == FNAMECMP_FNAME) {
610 rprintf(FINFO, "%s exists\n", safe_fname(fname));
614 if (update_only && fnamecmp_type == FNAMECMP_FNAME
615 && cmp_modtime(st.st_mtime, file->modtime) > 0) {
617 rprintf(FINFO, "%s is newer\n", safe_fname(fname));
621 if (!compare_dest && fnamecmp_type <= FNAMECMP_BASIS_DIR_HIGH)
623 else if (fnamecmp_type == FNAMECMP_FUZZY)
625 else if (unchanged_file(fnamecmp, file, &st)) {
626 if (fnamecmp_type == FNAMECMP_FNAME)
627 set_perms(fname, file, &st, PERMS_REPORT);
634 fnamecmp = partialptr;
635 fnamecmp_type = FNAMECMP_PARTIAL_DIR;
639 if (dry_run || read_batch)
641 if (whole_file > 0) {
647 int j = flist_find(fuzzy_dirlist, file);
648 if (j >= 0) /* don't use changing file as future fuzzy basis */
649 fuzzy_dirlist->files[j]->flags |= FLAG_NO_FUZZY;
653 fd = do_open(fnamecmp, O_RDONLY, 0);
656 rsyserr(FERROR, errno, "failed to open %s, continuing",
657 full_fname(fnamecmp));
659 /* pretend the file didn't exist */
660 if (preserve_hard_links && hard_link_check(file, HL_SKIP))
666 if (inplace && make_backups) {
667 if (!(backupptr = get_backup_name(fname))) {
671 if (!(back_file = make_file(fname, NULL, NO_FILTERS))) {
673 goto pretend_missing;
675 if (robust_unlink(backupptr) && errno != ENOENT) {
676 rsyserr(FERROR, errno, "unlink %s",
677 full_fname(backupptr));
682 if ((f_copy = do_open(backupptr,
683 O_WRONLY | O_CREAT | O_TRUNC | O_EXCL, 0600)) < 0) {
684 rsyserr(FERROR, errno, "open %s",
685 full_fname(backupptr));
690 fnamecmp_type = FNAMECMP_BACKUP;
694 rprintf(FINFO, "gen mapped %s of size %.0f\n",
695 safe_fname(fnamecmp), (double)st.st_size);
699 rprintf(FINFO, "generating and sending sums for %d\n", ndx);
702 write_int(f_out, ndx);
703 if (protocol_version >= 29 && inplace && !read_batch)
704 write_byte(f_out, fnamecmp_type);
705 if (f_out_name >= 0) {
706 write_byte(f_out_name, fnamecmp_type);
707 if (fnamecmp_type == FNAMECMP_FUZZY) {
708 uchar lenbuf[3], *lb = lenbuf;
709 int len = strlen(fuzzy_file->basename);
711 #if MAXPATHLEN > 0x7FFF
712 *lb++ = len / 0x10000 + 0x80;
715 *lb++ = len / 0x100 + 0x80;
719 write_buf(f_out_name, lenbuf, lb - lenbuf + 1);
720 write_buf(f_out_name, fuzzy_file->basename, len);
724 if (dry_run || read_batch)
728 generate_and_send_sums(fd, st.st_size, f_out, f_copy);
732 set_perms(backupptr, back_file, NULL, 0);
734 rprintf(FINFO, "backed up %s to %s\n",
735 safe_fname(fname), safe_fname(backupptr));
742 write_sum_head(f_out, NULL);
746 void generate_files(int f_out, struct file_list *flist, char *local_name,
751 char fbuf[MAXPATHLEN];
752 int need_retouch_dir_times = preserve_times && !omit_dir_times;
753 int need_retouch_dir_perms = 0;
754 int save_only_existing = only_existing;
755 int save_opt_ignore_existing = opt_ignore_existing;
758 rprintf(FINFO, "generator starting pid=%ld count=%d\n",
759 (long)getpid(), flist->count);
765 ? "delta-transmission disabled for local transfer or --whole-file\n"
766 : "delta transmission enabled\n");
769 /* We expect to just sit around now, so don't exit on a timeout.
770 * If we really get a timeout then the other process should exit. */
773 for (i = 0; i < flist->count; i++) {
774 struct file_struct *file = flist->files[i];
775 struct file_struct copy;
780 /* We need to ensure that any dirs we create have writeable
781 * permissions during the time we are putting files within
782 * them. This is then fixed after the transfer is done. */
783 if (!am_root && S_ISDIR(file->mode) && !(file->mode & S_IWUSR)) {
785 copy.mode |= S_IWUSR; /* user write */
787 need_retouch_dir_perms = 1;
790 recv_generator(local_name ? local_name : f_name_to(file, fbuf),
791 flist, file, i, f_out, f_out_name);
793 recv_generator(NULL, NULL, NULL, 0, -1, -1);
795 delete_in_dir(NULL, NULL, NULL);
798 csum_length = SUM_LENGTH;
799 only_existing = max_size = opt_ignore_existing = 0;
800 update_only = always_checksum = size_only = 0;
804 rprintf(FINFO,"generate_files phase=%d\n",phase);
806 write_int(f_out, -1);
808 /* files can cycle through the system more than once
809 * to catch initial checksum errors */
810 while ((i = get_redo_num()) != -1) {
811 struct file_struct *file = flist->files[i];
812 recv_generator(local_name ? local_name : f_name_to(file, fbuf),
813 flist, file, i, f_out, f_out_name);
817 only_existing = save_only_existing;
818 opt_ignore_existing = save_opt_ignore_existing;
821 rprintf(FINFO,"generate_files phase=%d\n",phase);
823 write_int(f_out, -1);
825 /* Read post-redo-phase MSG_DONE and any prior messages. */
828 if (preserve_hard_links)
831 if ((need_retouch_dir_perms || need_retouch_dir_times)
832 && !list_only && !local_name && !dry_run) {
833 /* Now we need to fix any directory permissions that were
834 * modified during the transfer and/or re-set any tweaked
835 * modified-time values. */
836 for (i = 0; i < flist->count; i++) {
837 struct file_struct *file = flist->files[i];
838 if (!file->basename || !S_ISDIR(file->mode))
840 if (!need_retouch_dir_times && file->mode & S_IWUSR)
842 recv_generator(local_name ? local_name : f_name(file),
843 flist, file, i, -1, -1);
846 recv_generator(NULL, NULL, NULL, 0, -1, -1);
849 rprintf(FINFO,"generate_files finished\n");