A couple comment tweaks.
[rsync/rsync.git] / flist.c
... / ...
CommitLineData
1/*
2 Copyright (C) Andrew Tridgell 1996
3 Copyright (C) Paul Mackerras 1996
4 Copyright (C) 2001, 2002 by Martin Pool <mbp@samba.org>
5
6 This program is free software; you can redistribute it and/or modify
7 it under the terms of the GNU General Public License as published by
8 the Free Software Foundation; either version 2 of the License, or
9 (at your option) any later version.
10
11 This program is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 GNU General Public License for more details.
15
16 You should have received a copy of the GNU General Public License
17 along with this program; if not, write to the Free Software
18 Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
19*/
20
21/** @file flist.c
22 * Generate and receive file lists
23 *
24 * @todo Get rid of the string_area optimization. Efficiently
25 * allocating blocks is the responsibility of the system's malloc
26 * library, not of rsync.
27 *
28 * @sa http://lists.samba.org/pipermail/rsync/2000-June/002351.html
29 *
30 **/
31
32#include "rsync.h"
33
34extern struct stats stats;
35
36extern int verbose;
37extern int do_progress;
38extern int am_server;
39extern int always_checksum;
40extern int module_id;
41extern int ignore_errors;
42
43extern int cvs_exclude;
44
45extern int recurse;
46extern char curr_dir[MAXPATHLEN];
47extern char *files_from;
48extern int filesfrom_fd;
49
50extern int one_file_system;
51extern int make_backups;
52extern int preserve_links;
53extern int preserve_hard_links;
54extern int preserve_perms;
55extern int preserve_devices;
56extern int preserve_uid;
57extern int preserve_gid;
58extern int preserve_times;
59extern int relative_paths;
60extern int implied_dirs;
61extern int copy_links;
62extern int copy_unsafe_links;
63extern int protocol_version;
64extern int sanitize_paths;
65
66extern int read_batch;
67extern int write_batch;
68
69extern struct exclude_struct **exclude_list;
70extern struct exclude_struct **server_exclude_list;
71extern struct exclude_struct **local_exclude_list;
72
73int io_error;
74
75static struct file_struct null_file;
76static char empty_sum[MD4_SUM_LENGTH];
77
78static void clean_flist(struct file_list *flist, int strip_root, int no_dups);
79
80
81static int show_filelist_p(void)
82{
83 return verbose && (recurse || files_from) && !am_server;
84}
85
86static void start_filelist_progress(char *kind)
87{
88 rprintf(FINFO, "%s ... ", kind);
89 if ((verbose > 1) || do_progress)
90 rprintf(FINFO, "\n");
91 rflush(FINFO);
92}
93
94
95static void emit_filelist_progress(const struct file_list *flist)
96{
97 rprintf(FINFO, " %d files...\r", flist->count);
98}
99
100
101static void maybe_emit_filelist_progress(const struct file_list *flist)
102{
103 if (do_progress && show_filelist_p() && ((flist->count % 100) == 0))
104 emit_filelist_progress(flist);
105}
106
107
108static void finish_filelist_progress(const struct file_list *flist)
109{
110 if (do_progress) {
111 /* This overwrites the progress line */
112 rprintf(FINFO, "%d file%sto consider\n",
113 flist->count, flist->count == 1 ? " " : "s ");
114 } else
115 rprintf(FINFO, "done\n");
116}
117
118void show_flist_stats(void)
119{
120 /* Nothing yet */
121}
122
123
124static struct string_area *string_area_new(int size)
125{
126 struct string_area *a;
127
128 if (size <= 0)
129 size = ARENA_SIZE;
130 a = new(struct string_area);
131 if (!a)
132 out_of_memory("string_area_new");
133 a->current = a->base = new_array(char, size);
134 if (!a->current)
135 out_of_memory("string_area_new buffer");
136 a->end = a->base + size;
137 a->next = NULL;
138
139 return a;
140}
141
142static void string_area_free(struct string_area *a)
143{
144 struct string_area *next;
145
146 for (; a; a = next) {
147 next = a->next;
148 free(a->base);
149 }
150}
151
152static char *string_area_malloc(struct string_area **ap, int size)
153{
154 char *p;
155 struct string_area *a;
156
157 /* does the request fit into the current space? */
158 a = *ap;
159 if (a->current + size >= a->end) {
160 /* no; get space, move new string_area to front of the list */
161 a = string_area_new(size > ARENA_SIZE ? size : ARENA_SIZE);
162 a->next = *ap;
163 *ap = a;
164 }
165
166 /* have space; do the "allocation." */
167 p = a->current;
168 a->current += size;
169 return p;
170}
171
172static char *string_area_strdup(struct string_area **ap, const char *src)
173{
174 char *dest = string_area_malloc(ap, strlen(src) + 1);
175 return strcpy(dest, src);
176}
177
178static void list_file_entry(struct file_struct *f)
179{
180 char perms[11];
181
182 if (!f->basename)
183 /* this can happen if duplicate names were removed */
184 return;
185
186 permstring(perms, f->mode);
187
188 if (preserve_links && S_ISLNK(f->mode)) {
189 rprintf(FINFO, "%s %11.0f %s %s -> %s\n",
190 perms,
191 (double) f->length, timestring(f->modtime),
192 f_name(f), f->u.link);
193 } else {
194 rprintf(FINFO, "%s %11.0f %s %s\n",
195 perms,
196 (double) f->length, timestring(f->modtime),
197 f_name(f));
198 }
199}
200
201
202/**
203 * Stat either a symlink or its referent, depending on the settings of
204 * copy_links, copy_unsafe_links, etc.
205 *
206 * @retval -1 on error
207 *
208 * @retval 0 for success
209 *
210 * @post If @p path is a symlink, then @p linkbuf (of size @c
211 * MAXPATHLEN) contains the symlink target.
212 *
213 * @post @p buffer contains information about the link or the
214 * referrent as appropriate, if they exist.
215 **/
216int readlink_stat(const char *path, STRUCT_STAT *buffer, char *linkbuf)
217{
218#if SUPPORT_LINKS
219 if (copy_links)
220 return do_stat(path, buffer);
221 if (do_lstat(path, buffer) == -1)
222 return -1;
223 if (S_ISLNK(buffer->st_mode)) {
224 int l = readlink((char *) path, linkbuf, MAXPATHLEN - 1);
225 if (l == -1)
226 return -1;
227 linkbuf[l] = 0;
228 if (copy_unsafe_links && unsafe_symlink(linkbuf, path)) {
229 if (verbose > 1) {
230 rprintf(FINFO,"copying unsafe symlink \"%s\" -> \"%s\"\n",
231 path, linkbuf);
232 }
233 return do_stat(path, buffer);
234 }
235 }
236 return 0;
237#else
238 return do_stat(path, buffer);
239#endif
240}
241
242int link_stat(const char *path, STRUCT_STAT * buffer)
243{
244#if SUPPORT_LINKS
245 if (copy_links)
246 return do_stat(path, buffer);
247 return do_lstat(path, buffer);
248#else
249 return do_stat(path, buffer);
250#endif
251}
252
253/*
254 * This function is used to check if a file should be included/excluded
255 * from the list of files based on its name and type etc. The value of
256 * exclude_level is set to either SERVER_EXCLUDES or ALL_EXCLUDES.
257 */
258static int check_exclude_file(char *fname, int is_dir, int exclude_level)
259{
260#if 0 /* This currently never happens, so avoid a useless compare. */
261 if (exclude_level == NO_EXCLUDES)
262 return 0;
263#endif
264 if (fname) {
265 /* never exclude '.', even if somebody does --exclude '*' */
266 if (fname[0] == '.' && !fname[1])
267 return 0;
268 /* Handle the -R version of the '.' dir. */
269 if (fname[0] == '/') {
270 int len = strlen(fname);
271 if (fname[len-1] == '.' && fname[len-2] == '/')
272 return 0;
273 }
274 }
275 if (server_exclude_list
276 && check_exclude(server_exclude_list, fname, is_dir))
277 return 1;
278 if (exclude_level != ALL_EXCLUDES)
279 return 0;
280 if (exclude_list && check_exclude(exclude_list, fname, is_dir))
281 return 1;
282 if (local_exclude_list
283 && check_exclude(local_exclude_list, fname, is_dir))
284 return 1;
285 return 0;
286}
287
288/* used by the one_file_system code */
289static dev_t filesystem_dev;
290
291static void set_filesystem(char *fname)
292{
293 STRUCT_STAT st;
294 if (link_stat(fname, &st) != 0)
295 return;
296 filesystem_dev = st.st_dev;
297}
298
299
300static int to_wire_mode(mode_t mode)
301{
302 if (S_ISLNK(mode) && (_S_IFLNK != 0120000))
303 return (mode & ~(_S_IFMT)) | 0120000;
304 return (int) mode;
305}
306
307static mode_t from_wire_mode(int mode)
308{
309 if ((mode & (_S_IFMT)) == 0120000 && (_S_IFLNK != 0120000))
310 return (mode & ~(_S_IFMT)) | _S_IFLNK;
311 return (mode_t) mode;
312}
313
314
315static void send_directory(int f, struct file_list *flist, char *dir);
316
317static char *flist_dir;
318static int flist_dir_len;
319
320
321/**
322 * Make sure @p flist is big enough to hold at least @p flist->count
323 * entries.
324 **/
325static void flist_expand(struct file_list *flist)
326{
327 if (flist->count >= flist->malloced) {
328 void *new_ptr;
329
330 if (flist->malloced < 1000)
331 flist->malloced += 1000;
332 else
333 flist->malloced *= 2;
334
335 if (flist->files) {
336 new_ptr = realloc_array(flist->files,
337 struct file_struct *,
338 flist->malloced);
339 } else {
340 new_ptr = new_array(struct file_struct *,
341 flist->malloced);
342 }
343
344 if (verbose >= 2) {
345 rprintf(FINFO, "expand file_list to %.0f bytes, did%s move\n",
346 (double)sizeof(flist->files[0])
347 * flist->malloced,
348 (new_ptr == flist->files) ? " not" : "");
349 }
350
351 flist->files = (struct file_struct **) new_ptr;
352
353 if (!flist->files)
354 out_of_memory("flist_expand");
355 }
356}
357
358void send_file_entry(struct file_struct *file, int f, unsigned short base_flags)
359{
360 unsigned short flags;
361 static time_t modtime;
362 static mode_t mode;
363 static DEV64_T rdev; /* just high bytes in p28 onward */
364 static DEV64_T dev;
365 static uid_t uid;
366 static gid_t gid;
367 static char lastname[MAXPATHLEN];
368 char *fname, fbuf[MAXPATHLEN];
369 int l1, l2;
370
371 if (f == -1)
372 return;
373
374 if (!file) {
375 write_byte(f, 0);
376 modtime = 0, mode = 0;
377 rdev = 0, dev = 0;
378 uid = 0, gid = 0;
379 *lastname = '\0';
380 return;
381 }
382
383 io_write_phase = "send_file_entry";
384
385 fname = f_name_to(file, fbuf);
386
387 flags = base_flags;
388
389 if (file->mode == mode)
390 flags |= SAME_MODE;
391 else
392 mode = file->mode;
393 if (preserve_devices) {
394 if (protocol_version < 28) {
395 if (IS_DEVICE(mode)) {
396 if (file->u.rdev == rdev) {
397 /* Set both flags so that the test when
398 * writing the data is simpler. */
399 flags |= SAME_RDEV_pre28|SAME_HIGH_RDEV;
400 } else
401 rdev = file->u.rdev;
402 } else
403 rdev = 0;
404 } else if (IS_DEVICE(mode)) {
405 if ((file->u.rdev & ~0xFF) == rdev)
406 flags |= SAME_HIGH_RDEV;
407 else
408 rdev = file->u.rdev & ~0xFF;
409 }
410 }
411 if (file->uid == uid)
412 flags |= SAME_UID;
413 else
414 uid = file->uid;
415 if (file->gid == gid)
416 flags |= SAME_GID;
417 else
418 gid = file->gid;
419 if (file->modtime == modtime)
420 flags |= SAME_TIME;
421 else
422 modtime = file->modtime;
423 if (file->link_u.idev) {
424 if (file->F_DEV == dev) {
425 if (protocol_version >= 28)
426 flags |= SAME_DEV;
427 } else
428 dev = file->F_DEV;
429 flags |= HAS_INODE_DATA;
430 }
431
432 for (l1 = 0;
433 lastname[l1] && (fname[l1] == lastname[l1]) && (l1 < 255);
434 l1++) {}
435 l2 = strlen(fname+l1);
436
437 if (l1 > 0)
438 flags |= SAME_NAME;
439 if (l2 > 255)
440 flags |= LONG_NAME;
441
442 /* We must make sure we don't send a zero flags byte or
443 * the other end will terminate the flist transfer. */
444 if (flags == 0 && !S_ISDIR(mode))
445 flags |= FLAG_DELETE; /* NOTE: no meaning for non-dir */
446 if (protocol_version >= 28) {
447 if ((flags & 0xFF00) || flags == 0) {
448 flags |= EXTENDED_FLAGS;
449 write_byte(f, flags);
450 write_byte(f, flags >> 8);
451 } else
452 write_byte(f, flags);
453 } else {
454 if (flags == 0)
455 flags |= LONG_NAME;
456 write_byte(f, flags);
457 }
458 if (flags & SAME_NAME)
459 write_byte(f, l1);
460 if (flags & LONG_NAME)
461 write_int(f, l2);
462 else
463 write_byte(f, l2);
464 write_buf(f, fname + l1, l2);
465
466 write_longint(f, file->length);
467 if (!(flags & SAME_TIME))
468 write_int(f, modtime);
469 if (!(flags & SAME_MODE))
470 write_int(f, to_wire_mode(mode));
471 if (preserve_uid && !(flags & SAME_UID)) {
472 add_uid(uid);
473 write_int(f, uid);
474 }
475 if (preserve_gid && !(flags & SAME_GID)) {
476 add_gid(gid);
477 write_int(f, gid);
478 }
479 if (preserve_devices && IS_DEVICE(mode)) {
480 /* If SAME_HIGH_RDEV is off, SAME_RDEV_pre28 is also off.
481 * Also, avoid using "rdev" because it may be incomplete. */
482 if (!(flags & SAME_HIGH_RDEV))
483 write_int(f, file->u.rdev);
484 else if (protocol_version >= 28)
485 write_byte(f, file->u.rdev);
486 }
487
488#if SUPPORT_LINKS
489 if (preserve_links && S_ISLNK(mode)) {
490 write_int(f, strlen(file->u.link));
491 write_buf(f, file->u.link, strlen(file->u.link));
492 }
493#endif
494
495#if SUPPORT_HARD_LINKS
496 if (flags & HAS_INODE_DATA) {
497 if (protocol_version < 26) {
498 /* 32-bit dev_t and ino_t */
499 write_int(f, dev);
500 write_int(f, file->F_INODE);
501 } else {
502 /* 64-bit dev_t and ino_t */
503 if (!(flags & SAME_DEV))
504 write_longint(f, dev);
505 write_longint(f, file->F_INODE);
506 }
507 }
508#endif
509
510 if (always_checksum) {
511 char *sum;
512 if (S_ISREG(mode))
513 sum = file->u.sum;
514 else if (protocol_version < 28) {
515 /* Prior to 28, we sent a useless set of nulls. */
516 sum = empty_sum;
517 } else
518 sum = NULL;
519 if (sum) {
520 write_buf(f, sum, protocol_version < 21? 2
521 : MD4_SUM_LENGTH);
522 }
523 }
524
525 strlcpy(lastname, fname, MAXPATHLEN);
526
527 io_write_phase = "unknown";
528}
529
530
531
532void receive_file_entry(struct file_struct **fptr, unsigned short flags, int f)
533{
534 static time_t modtime;
535 static mode_t mode;
536 static DEV64_T rdev; /* just high bytes in p28 onward */
537 static DEV64_T dev;
538 static uid_t uid;
539 static gid_t gid;
540 static char lastname[MAXPATHLEN];
541 char thisname[MAXPATHLEN];
542 unsigned int l1 = 0, l2 = 0;
543 char *p;
544 struct file_struct *file;
545
546 if (!fptr) {
547 modtime = 0, mode = 0;
548 rdev = 0, dev = 0;
549 uid = 0, gid = 0;
550 *lastname = '\0';
551 return;
552 }
553
554 if (flags & SAME_NAME)
555 l1 = read_byte(f);
556
557 if (flags & LONG_NAME)
558 l2 = read_int(f);
559 else
560 l2 = read_byte(f);
561
562 file = new(struct file_struct);
563 if (!file)
564 out_of_memory("receive_file_entry");
565 memset((char *) file, 0, sizeof(*file));
566 (*fptr) = file;
567
568 if (l2 >= MAXPATHLEN - l1) {
569 rprintf(FERROR,
570 "overflow: flags=0x%x l1=%d l2=%d lastname=%s\n",
571 flags, l1, l2, lastname);
572 overflow("receive_file_entry");
573 }
574
575 strlcpy(thisname, lastname, l1 + 1);
576 read_sbuf(f, &thisname[l1], l2);
577 thisname[l1 + l2] = 0;
578
579 strlcpy(lastname, thisname, MAXPATHLEN);
580 lastname[MAXPATHLEN - 1] = 0;
581
582 clean_fname(thisname);
583
584 if (sanitize_paths) {
585 sanitize_path(thisname, NULL);
586 }
587
588 if ((p = strrchr(thisname, '/'))) {
589 static char *lastdir;
590 *p = 0;
591 if (lastdir && strcmp(thisname, lastdir) == 0)
592 file->dirname = lastdir;
593 else {
594 file->dirname = strdup(thisname);
595 lastdir = file->dirname;
596 }
597 file->basename = strdup(p + 1);
598 } else {
599 file->dirname = NULL;
600 file->basename = strdup(thisname);
601 }
602
603 if (!file->basename)
604 out_of_memory("receive_file_entry 1");
605
606 file->flags = flags;
607 file->length = read_longint(f);
608 if (!(flags & SAME_TIME))
609 modtime = (time_t)read_int(f);
610 file->modtime = modtime;
611 if (!(flags & SAME_MODE))
612 mode = from_wire_mode(read_int(f));
613 file->mode = mode;
614
615 if (preserve_uid) {
616 if (!(flags & SAME_UID))
617 uid = (uid_t)read_int(f);
618 file->uid = uid;
619 }
620 if (preserve_gid) {
621 if (!(flags & SAME_GID))
622 gid = (gid_t)read_int(f);
623 file->gid = gid;
624 }
625 if (preserve_devices) {
626 if (protocol_version < 28) {
627 if (IS_DEVICE(mode)) {
628 if (!(flags & SAME_RDEV_pre28))
629 rdev = (DEV64_T)read_int(f);
630 file->u.rdev = rdev;
631 } else
632 rdev = 0;
633 } else if (IS_DEVICE(mode)) {
634 if (!(flags & SAME_HIGH_RDEV)) {
635 file->u.rdev = (DEV64_T)read_int(f);
636 rdev = file->u.rdev & ~0xFF;
637 } else
638 file->u.rdev = rdev | (DEV64_T)read_byte(f);
639 }
640 }
641
642 if (preserve_links && S_ISLNK(mode)) {
643 int l = read_int(f);
644 if (l < 0) {
645 rprintf(FERROR, "overflow: l=%d\n", l);
646 overflow("receive_file_entry");
647 }
648 if (!(file->u.link = new_array(char, l + 1)))
649 out_of_memory("receive_file_entry 2");
650 read_sbuf(f, file->u.link, l);
651 if (sanitize_paths)
652 sanitize_path(file->u.link, file->dirname);
653 }
654#if SUPPORT_HARD_LINKS
655 if (preserve_hard_links && protocol_version < 28 && S_ISREG(mode))
656 flags |= HAS_INODE_DATA;
657 if (flags & HAS_INODE_DATA) {
658 if (!(file->link_u.idev = new(struct idev)))
659 out_of_memory("file inode data");
660 if (protocol_version < 26) {
661 dev = read_int(f);
662 file->F_INODE = read_int(f);
663 } else {
664 if (!(flags & SAME_DEV))
665 dev = read_longint(f);
666 file->F_INODE = read_longint(f);
667 }
668 file->F_DEV = dev;
669 }
670#endif
671
672 if (always_checksum) {
673 char *sum;
674 if (S_ISREG(mode)) {
675 sum = file->u.sum = new_array(char, MD4_SUM_LENGTH);
676 if (!sum)
677 out_of_memory("md4 sum");
678 } else if (protocol_version < 28) {
679 /* Prior to 28, we get a useless set of nulls. */
680 sum = empty_sum;
681 } else
682 sum = NULL;
683 if (sum) {
684 read_buf(f, sum, protocol_version < 21? 2
685 : MD4_SUM_LENGTH);
686 }
687 }
688
689 if (!preserve_perms) {
690 extern int orig_umask;
691 /* set an appropriate set of permissions based on original
692 * permissions and umask. This emulates what GNU cp does */
693 file->mode &= ~orig_umask;
694 }
695}
696
697
698/* determine if a file in a different filesstem should be skipped
699 when one_file_system is set. We bascally only want to include
700 the mount points - but they can be hard to find! */
701static int skip_filesystem(char *fname, STRUCT_STAT * st)
702{
703 STRUCT_STAT st2;
704 char *p = strrchr(fname, '/');
705
706 /* skip all but directories */
707 if (!S_ISDIR(st->st_mode))
708 return 1;
709
710 /* if its not a subdirectory then allow */
711 if (!p)
712 return 0;
713
714 *p = 0;
715 if (link_stat(fname, &st2)) {
716 *p = '/';
717 return 0;
718 }
719 *p = '/';
720
721 return (st2.st_dev != filesystem_dev);
722}
723
724#define STRDUP(ap, p) (ap ? string_area_strdup(ap, p) : strdup(p))
725/* IRIX cc cares that the operands to the ternary have the same type. */
726#define MALLOC(ap, i) (ap ? (void*) string_area_malloc(ap, i) : malloc(i))
727
728/**
729 * Create a file_struct for a named file by reading its stat()
730 * information and performing extensive checks against global
731 * options.
732 *
733 * @return the new file, or NULL if there was an error or this file
734 * should be excluded.
735 *
736 * @todo There is a small optimization opportunity here to avoid
737 * stat()ing the file in some circumstances, which has a certain cost.
738 * We are called immediately after doing readdir(), and so we may
739 * already know the d_type of the file. We could for example avoid
740 * statting directories if we're not recursing, but this is not a very
741 * important case. Some systems may not have d_type.
742 **/
743struct file_struct *make_file(char *fname, struct string_area **ap,
744 int exclude_level)
745{
746 struct file_struct *file;
747 STRUCT_STAT st;
748 char sum[SUM_LENGTH];
749 char *p;
750 char cleaned_name[MAXPATHLEN];
751 char linkbuf[MAXPATHLEN];
752
753 if (strlcpy(cleaned_name, fname, sizeof cleaned_name)
754 >= sizeof cleaned_name - flist_dir_len) {
755 rprintf(FINFO, "skipping overly long name: %s\n", fname);
756 return NULL;
757 }
758 clean_fname(cleaned_name);
759 if (sanitize_paths)
760 sanitize_path(cleaned_name, NULL);
761 fname = cleaned_name;
762
763 memset(sum, 0, SUM_LENGTH);
764
765 if (readlink_stat(fname, &st, linkbuf) != 0) {
766 int save_errno = errno;
767 if (errno == ENOENT && exclude_level != NO_EXCLUDES) {
768 /* either symlink pointing nowhere or file that
769 * was removed during rsync run; see if excluded
770 * before reporting an error */
771 if (check_exclude_file(fname, 0, exclude_level)) {
772 /* file is excluded anyway, ignore silently */
773 return NULL;
774 }
775 }
776 io_error |= IOERR_GENERAL;
777 rprintf(FERROR, "readlink %s failed: %s\n",
778 full_fname(fname), strerror(save_errno));
779 return NULL;
780 }
781
782 /* backup.c calls us with exclude_level set to NO_EXCLUDES. */
783 if (exclude_level == NO_EXCLUDES)
784 goto skip_excludes;
785
786 if (S_ISDIR(st.st_mode) && !recurse && !files_from) {
787 rprintf(FINFO, "skipping directory %s\n", fname);
788 return NULL;
789 }
790
791 if (one_file_system && st.st_dev != filesystem_dev) {
792 if (skip_filesystem(fname, &st))
793 return NULL;
794 }
795
796 if (check_exclude_file(fname, S_ISDIR(st.st_mode) != 0, exclude_level))
797 return NULL;
798
799 if (lp_ignore_nonreadable(module_id) && access(fname, R_OK) != 0)
800 return NULL;
801
802 skip_excludes:
803
804 if (verbose > 2)
805 rprintf(FINFO, "make_file(%s,*,%d)\n", fname, exclude_level);
806
807 file = new(struct file_struct);
808 if (!file)
809 out_of_memory("make_file");
810 memset((char *) file, 0, sizeof(*file));
811
812 if ((p = strrchr(fname, '/'))) {
813 static char *lastdir;
814 *p = 0;
815 if (lastdir && strcmp(fname, lastdir) == 0)
816 file->dirname = lastdir;
817 else {
818 file->dirname = strdup(fname);
819 lastdir = file->dirname;
820 }
821 file->basename = STRDUP(ap, p + 1);
822 *p = '/';
823 } else {
824 file->dirname = NULL;
825 file->basename = STRDUP(ap, fname);
826 }
827
828 file->modtime = st.st_mtime;
829 file->length = st.st_size;
830 file->mode = st.st_mode;
831 file->uid = st.st_uid;
832 file->gid = st.st_gid;
833 if (preserve_hard_links) {
834 if (protocol_version < 28 ? S_ISREG(st.st_mode)
835 : !S_ISDIR(st.st_mode) && st.st_nlink > 1) {
836 if (!(file->link_u.idev = new(struct idev)))
837 out_of_memory("file inode data");
838 file->F_DEV = st.st_dev;
839 file->F_INODE = st.st_ino;
840 }
841 }
842#ifdef HAVE_STRUCT_STAT_ST_RDEV
843 if (IS_DEVICE(st.st_mode))
844 file->u.rdev = st.st_rdev;
845#endif
846
847#if SUPPORT_LINKS
848 if (S_ISLNK(st.st_mode))
849 file->u.link = STRDUP(ap, linkbuf);
850#endif
851
852 if (always_checksum && S_ISREG(st.st_mode)) {
853 if (!(file->u.sum = (char*)MALLOC(ap, MD4_SUM_LENGTH)))
854 out_of_memory("md4 sum");
855 file_checksum(fname, file->u.sum, st.st_size);
856 }
857
858 file->basedir = flist_dir;
859
860 if (!S_ISDIR(st.st_mode))
861 stats.total_size += st.st_size;
862
863 return file;
864}
865
866
867void send_file_name(int f, struct file_list *flist, char *fname,
868 int recursive, unsigned short base_flags)
869{
870 struct file_struct *file;
871 char fbuf[MAXPATHLEN];
872 extern int delete_excluded;
873
874 /* f is set to -1 when calculating deletion file list */
875 file = make_file(fname, &flist->string_area,
876 f == -1 && delete_excluded? SERVER_EXCLUDES
877 : ALL_EXCLUDES);
878
879 if (!file)
880 return;
881
882 maybe_emit_filelist_progress(flist);
883
884 flist_expand(flist);
885
886 if (write_batch)
887 file->flags |= FLAG_DELETE;
888
889 if (file->basename[0]) {
890 flist->files[flist->count++] = file;
891 send_file_entry(file, f, base_flags);
892 }
893
894 if (S_ISDIR(file->mode) && recursive) {
895 struct exclude_struct **last_exclude_list =
896 local_exclude_list;
897 send_directory(f, flist, f_name_to(file, fbuf));
898 local_exclude_list = last_exclude_list;
899 return;
900 }
901}
902
903
904static void send_directory(int f, struct file_list *flist, char *dir)
905{
906 DIR *d;
907 struct dirent *di;
908 char fname[MAXPATHLEN];
909 unsigned int offset;
910 char *p;
911
912 d = opendir(dir);
913 if (!d) {
914 io_error |= IOERR_GENERAL;
915 rprintf(FERROR, "opendir %s failed: %s\n",
916 full_fname(dir), strerror(errno));
917 return;
918 }
919
920 offset = strlcpy(fname, dir, MAXPATHLEN);
921 p = fname + offset;
922 if (offset >= MAXPATHLEN || p[-1] != '/') {
923 if (offset >= MAXPATHLEN - 1) {
924 io_error |= IOERR_GENERAL;
925 rprintf(FERROR, "skipping long-named directory: %s\n",
926 full_fname(fname));
927 closedir(d);
928 return;
929 }
930 *p++ = '/';
931 offset++;
932 }
933
934 local_exclude_list = NULL;
935
936 if (cvs_exclude) {
937 if (strlcpy(p, ".cvsignore", MAXPATHLEN - offset)
938 < MAXPATHLEN - offset)
939 add_exclude_file(&local_exclude_list,fname,MISSING_OK,ADD_EXCLUDE);
940 else {
941 io_error |= IOERR_GENERAL;
942 rprintf(FINFO,
943 "cannot cvs-exclude in long-named directory %s\n",
944 full_fname(fname));
945 }
946 }
947
948 for (errno = 0, di = readdir(d); di; errno = 0, di = readdir(d)) {
949 char *dname = d_name(di);
950 if (dname[0] == '.' && (dname[1] == '\0'
951 || (dname[1] == '.' && dname[2] == '\0')))
952 continue;
953 if (strlcpy(p, dname, MAXPATHLEN - offset) < MAXPATHLEN - offset)
954 send_file_name(f, flist, fname, recurse, 0);
955 else {
956 io_error |= IOERR_GENERAL;
957 rprintf(FINFO,
958 "cannot send long-named file %s\n",
959 full_fname(fname));
960 }
961 }
962 if (errno) {
963 io_error |= IOERR_GENERAL;
964 rprintf(FERROR, "readdir(%s): (%d) %s\n",
965 dir, errno, strerror(errno));
966 }
967
968 if (local_exclude_list)
969 free_exclude_list(&local_exclude_list); /* Zeros pointer too */
970
971 closedir(d);
972}
973
974
975/**
976 * The delete_files() function in receiver.c sets f to -1 so that we just
977 * construct the file list in memory without sending it over the wire. It
978 * also has the side-effect of ignoring user-excludes if delete_excluded
979 * is set (so that the delete list includes user-excluded files).
980 **/
981struct file_list *send_file_list(int f, int argc, char *argv[])
982{
983 int l;
984 STRUCT_STAT st;
985 char *p, *dir, olddir[sizeof curr_dir];
986 char lastpath[MAXPATHLEN] = "";
987 struct file_list *flist;
988 int64 start_write;
989 int use_ff_fd = 0;
990
991 if (show_filelist_p() && f != -1)
992 start_filelist_progress("building file list");
993
994 start_write = stats.total_written;
995
996 flist = flist_new();
997
998 if (f != -1) {
999 io_start_buffering_out(f);
1000 if (filesfrom_fd >= 0) {
1001 if (argv[0] && !push_dir(argv[0])) {
1002 rprintf(FERROR, "push_dir %s failed: %s\n",
1003 full_fname(argv[0]), strerror(errno));
1004 exit_cleanup(RERR_FILESELECT);
1005 }
1006 use_ff_fd = 1;
1007 }
1008 }
1009
1010 while (1) {
1011 char fname2[MAXPATHLEN];
1012 char *fname = fname2;
1013
1014 if (use_ff_fd) {
1015 if (read_filesfrom_line(filesfrom_fd, fname) == 0)
1016 break;
1017 sanitize_path(fname, NULL);
1018 } else {
1019 if (argc-- == 0)
1020 break;
1021 strlcpy(fname, *argv++, MAXPATHLEN);
1022 if (sanitize_paths)
1023 sanitize_path(fname, NULL);
1024 }
1025
1026 l = strlen(fname);
1027 if (fname[l - 1] == '/') {
1028 if (l == 2 && fname[0] == '.') {
1029 /* Turn "./" into just "." rather than "./." */
1030 fname[1] = '\0';
1031 } else if (l < MAXPATHLEN) {
1032 fname[l++] = '.';
1033 fname[l] = '\0';
1034 }
1035 }
1036
1037 if (link_stat(fname, &st) != 0) {
1038 if (f != -1) {
1039 io_error |= IOERR_GENERAL;
1040 rprintf(FERROR, "link_stat %s failed: %s\n",
1041 full_fname(fname), strerror(errno));
1042 }
1043 continue;
1044 }
1045
1046 if (S_ISDIR(st.st_mode) && !recurse && !files_from) {
1047 rprintf(FINFO, "skipping directory %s\n", fname);
1048 continue;
1049 }
1050
1051 dir = NULL;
1052 olddir[0] = '\0';
1053
1054 if (!relative_paths) {
1055 p = strrchr(fname, '/');
1056 if (p) {
1057 *p = 0;
1058 if (p == fname)
1059 dir = "/";
1060 else
1061 dir = fname;
1062 fname = p + 1;
1063 }
1064 } else if (f != -1 && implied_dirs && (p=strrchr(fname,'/')) && p != fname) {
1065 /* this ensures we send the intermediate directories,
1066 thus getting their permissions right */
1067 char *lp = lastpath, *fn = fname, *slash = fname;
1068 *p = 0;
1069 /* Skip any initial directories in our path that we
1070 * have in common with lastpath. */
1071 while (*fn && *lp == *fn) {
1072 if (*fn == '/')
1073 slash = fn;
1074 lp++, fn++;
1075 }
1076 *p = '/';
1077 if (fn != p || (*lp && *lp != '/')) {
1078 int copy_links_saved = copy_links;
1079 int recurse_saved = recurse;
1080 copy_links = copy_unsafe_links;
1081 /* set recurse to 1 to prevent make_file
1082 * from ignoring directory, but still
1083 * turn off the recursive parameter to
1084 * send_file_name */
1085 recurse = 1;
1086 while ((slash = strchr(slash+1, '/')) != 0) {
1087 *slash = 0;
1088 send_file_name(f, flist, fname, 0, 0);
1089 *slash = '/';
1090 }
1091 copy_links = copy_links_saved;
1092 recurse = recurse_saved;
1093 *p = 0;
1094 strlcpy(lastpath, fname, sizeof lastpath);
1095 *p = '/';
1096 }
1097 }
1098
1099 if (!*fname)
1100 fname = ".";
1101
1102 if (dir && *dir) {
1103 static char *lastdir;
1104 static int lastdir_len;
1105
1106 strcpy(olddir, curr_dir); /* can't overflow */
1107
1108 if (!push_dir(dir)) {
1109 io_error |= IOERR_GENERAL;
1110 rprintf(FERROR, "push_dir %s failed: %s\n",
1111 full_fname(dir), strerror(errno));
1112 continue;
1113 }
1114
1115 if (lastdir && strcmp(lastdir, dir) == 0) {
1116 flist_dir = lastdir;
1117 flist_dir_len = lastdir_len;
1118 } else {
1119 if (lastdir)
1120 free(lastdir);
1121 flist_dir = lastdir = strdup(dir);
1122 flist_dir_len = lastdir_len = strlen(dir);
1123 }
1124 }
1125
1126 if (one_file_system)
1127 set_filesystem(fname);
1128
1129 send_file_name(f, flist, fname, recurse, FLAG_DELETE);
1130
1131 if (olddir[0]) {
1132 flist_dir = NULL;
1133 flist_dir_len = 0;
1134 if (!pop_dir(olddir)) {
1135 rprintf(FERROR, "pop_dir %s failed: %s\n",
1136 full_fname(dir), strerror(errno));
1137 exit_cleanup(RERR_FILESELECT);
1138 }
1139 }
1140 }
1141
1142 if (f != -1) {
1143 send_file_entry(NULL, f, 0);
1144
1145 if (show_filelist_p())
1146 finish_filelist_progress(flist);
1147 }
1148
1149 clean_flist(flist, 0, 0);
1150
1151 if (f != -1) {
1152 /* Now send the uid/gid list. This was introduced in
1153 * protocol version 15 */
1154 send_uid_list(f);
1155
1156 /* send the io_error flag */
1157 write_int(f, lp_ignore_errors(module_id) ? 0 : io_error);
1158
1159 io_end_buffering();
1160 stats.flist_size = stats.total_written - start_write;
1161 stats.num_files = flist->count;
1162 if (write_batch)
1163 write_batch_flist_info(flist->count, flist->files);
1164 }
1165
1166 if (verbose > 2)
1167 rprintf(FINFO, "send_file_list done\n");
1168
1169 return flist;
1170}
1171
1172
1173struct file_list *recv_file_list(int f)
1174{
1175 struct file_list *flist;
1176 unsigned short flags;
1177 int64 start_read;
1178 extern int list_only;
1179
1180 if (show_filelist_p())
1181 start_filelist_progress("receiving file list");
1182
1183 start_read = stats.total_read;
1184
1185 flist = new(struct file_list);
1186 if (!flist)
1187 goto oom;
1188
1189 flist->count = 0;
1190 flist->malloced = 1000;
1191 flist->files = new_array(struct file_struct *, flist->malloced);
1192 if (!flist->files)
1193 goto oom;
1194
1195
1196 while ((flags = read_byte(f)) != 0) {
1197 int i = flist->count;
1198
1199 flist_expand(flist);
1200
1201 if (protocol_version >= 28 && (flags & EXTENDED_FLAGS))
1202 flags |= read_byte(f) << 8;
1203 receive_file_entry(&flist->files[i], flags, f);
1204
1205 if (S_ISREG(flist->files[i]->mode))
1206 stats.total_size += flist->files[i]->length;
1207
1208 flist->count++;
1209
1210 maybe_emit_filelist_progress(flist);
1211
1212 if (verbose > 2) {
1213 rprintf(FINFO, "recv_file_name(%s)\n",
1214 f_name(flist->files[i]));
1215 }
1216 }
1217 receive_file_entry(NULL, 0, 0); /* Signal that we're done. */
1218
1219 if (verbose > 2)
1220 rprintf(FINFO, "received %d names\n", flist->count);
1221
1222 if (show_filelist_p())
1223 finish_filelist_progress(flist);
1224
1225 clean_flist(flist, relative_paths, 1);
1226
1227 if (f != -1) {
1228 /* Now send the uid/gid list. This was introduced in
1229 * protocol version 15 */
1230 recv_uid_list(f, flist);
1231
1232 if (!read_batch) {
1233 /* Recv the io_error flag */
1234 if (lp_ignore_errors(module_id) || ignore_errors)
1235 read_int(f);
1236 else
1237 io_error |= read_int(f);
1238 }
1239 }
1240
1241 if (list_only) {
1242 int i;
1243 for (i = 0; i < flist->count; i++)
1244 list_file_entry(flist->files[i]);
1245 }
1246
1247 if (verbose > 2)
1248 rprintf(FINFO, "recv_file_list done\n");
1249
1250 stats.flist_size = stats.total_read - start_read;
1251 stats.num_files = flist->count;
1252
1253 return flist;
1254
1255 oom:
1256 out_of_memory("recv_file_list");
1257 return NULL; /* not reached */
1258}
1259
1260
1261int file_compare(struct file_struct **file1, struct file_struct **file2)
1262{
1263 struct file_struct *f1 = *file1;
1264 struct file_struct *f2 = *file2;
1265
1266 if (!f1->basename && !f2->basename)
1267 return 0;
1268 if (!f1->basename)
1269 return -1;
1270 if (!f2->basename)
1271 return 1;
1272 if (f1->dirname == f2->dirname)
1273 return u_strcmp(f1->basename, f2->basename);
1274 return f_name_cmp(f1, f2);
1275}
1276
1277
1278int flist_find(struct file_list *flist, struct file_struct *f)
1279{
1280 int low = 0, high = flist->count - 1;
1281
1282 while (high >= 0 && !flist->files[high]->basename) high--;
1283
1284 if (high < 0)
1285 return -1;
1286
1287 while (low != high) {
1288 int mid = (low + high) / 2;
1289 int ret = file_compare(&flist->files[flist_up(flist, mid)],&f);
1290 if (ret == 0)
1291 return flist_up(flist, mid);
1292 if (ret > 0)
1293 high = mid;
1294 else
1295 low = mid + 1;
1296 }
1297
1298 if (file_compare(&flist->files[flist_up(flist, low)], &f) == 0)
1299 return flist_up(flist, low);
1300 return -1;
1301}
1302
1303
1304/*
1305 * free up one file
1306 */
1307void free_file(struct file_struct *file)
1308{
1309 if (!file)
1310 return;
1311 if (file->basename)
1312 free(file->basename);
1313 if (!IS_DEVICE(file->mode) && file->u.link)
1314 free(file->u.link); /* Handles u.sum too. */
1315 if (file->link_u.idev)
1316 free((char*)file->link_u.idev); /* Handles link_u.links too. */
1317 *file = null_file;
1318}
1319
1320
1321/*
1322 * allocate a new file list
1323 */
1324struct file_list *flist_new(void)
1325{
1326 struct file_list *flist;
1327
1328 flist = new(struct file_list);
1329 if (!flist)
1330 out_of_memory("send_file_list");
1331
1332 flist->count = 0;
1333 flist->malloced = 0;
1334 flist->files = NULL;
1335
1336#if ARENA_SIZE > 0
1337 flist->string_area = string_area_new(0);
1338#else
1339 flist->string_area = NULL;
1340#endif
1341 return flist;
1342}
1343
1344/*
1345 * free up all elements in a flist
1346 */
1347void flist_free(struct file_list *flist)
1348{
1349 int i;
1350 for (i = 1; i < flist->count; i++) {
1351 if (!flist->string_area)
1352 free_file(flist->files[i]);
1353 free(flist->files[i]);
1354 }
1355 /* FIXME: I don't think we generally need to blank the flist
1356 * since it's about to be freed. This will just cause more
1357 * memory traffic. If you want a freed-memory debugger, you
1358 * know where to get it. */
1359 memset((char *) flist->files, 0,
1360 sizeof(flist->files[0]) * flist->count);
1361 free(flist->files);
1362 if (flist->string_area)
1363 string_area_free(flist->string_area);
1364 memset((char *) flist, 0, sizeof(*flist));
1365 free(flist);
1366}
1367
1368
1369/*
1370 * This routine ensures we don't have any duplicate names in our file list.
1371 * duplicate names can cause corruption because of the pipelining
1372 */
1373static void clean_flist(struct file_list *flist, int strip_root, int no_dups)
1374{
1375 int i, prev_i = 0;
1376
1377 if (!flist || flist->count == 0)
1378 return;
1379
1380 qsort(flist->files, flist->count,
1381 sizeof(flist->files[0]), (int (*)()) file_compare);
1382
1383 for (i = no_dups? 0 : flist->count; i < flist->count; i++) {
1384 if (flist->files[i]->basename) {
1385 prev_i = i;
1386 break;
1387 }
1388 }
1389 while (++i < flist->count) {
1390 if (!flist->files[i]->basename)
1391 continue;
1392 if (f_name_cmp(flist->files[i], flist->files[prev_i]) == 0) {
1393 if (verbose > 1 && !am_server) {
1394 rprintf(FINFO,
1395 "removing duplicate name %s from file list %d\n",
1396 f_name(flist->files[i]), i);
1397 }
1398 /* Make sure that if we unduplicate '.', that we don't
1399 * lose track of a user-specified starting point (or
1400 * else deletions will mysteriously fail with -R). */
1401 if (flist->files[i]->flags & FLAG_DELETE)
1402 flist->files[prev_i]->flags |= FLAG_DELETE;
1403 /* it's not great that the flist knows the semantics of
1404 * the file memory usage, but i'd rather not add a flag
1405 * byte to that struct.
1406 * XXX can i use a bit in the flags field? */
1407 if (flist->string_area)
1408 flist->files[i][0] = null_file;
1409 else
1410 free_file(flist->files[i]);
1411 } else
1412 prev_i = i;
1413 }
1414
1415 if (strip_root) {
1416 /* we need to strip off the root directory in the case
1417 of relative paths, but this must be done _after_
1418 the sorting phase */
1419 for (i = 0; i < flist->count; i++) {
1420 if (flist->files[i]->dirname &&
1421 flist->files[i]->dirname[0] == '/') {
1422 memmove(&flist->files[i]->dirname[0],
1423 &flist->files[i]->dirname[1],
1424 strlen(flist->files[i]->dirname));
1425 }
1426
1427 if (flist->files[i]->dirname &&
1428 !flist->files[i]->dirname[0]) {
1429 flist->files[i]->dirname = NULL;
1430 }
1431 }
1432 }
1433
1434 if (verbose <= 3)
1435 return;
1436
1437 for (i = 0; i < flist->count; i++) {
1438 rprintf(FINFO, "[%ld] i=%d %s %s mode=0%o len=%.0f\n",
1439 (long) getpid(), i,
1440 NS(flist->files[i]->dirname),
1441 NS(flist->files[i]->basename),
1442 (int) flist->files[i]->mode,
1443 (double) flist->files[i]->length);
1444 }
1445}
1446
1447
1448enum fnc_state { fnc_DIR, fnc_SLASH, fnc_BASE };
1449
1450/* Compare the names of two file_struct entities, just like strcmp()
1451 * would do if it were operating on the joined strings. We assume
1452 * that there are no 0-length strings.
1453 */
1454int f_name_cmp(struct file_struct *f1, struct file_struct *f2)
1455{
1456 int dif;
1457 const uchar *c1, *c2;
1458 enum fnc_state state1, state2;
1459
1460 if (!f1 || !f1->basename) {
1461 if (!f2 || !f2->basename)
1462 return 0;
1463 return -1;
1464 }
1465 if (!f2 || !f2->basename)
1466 return 1;
1467
1468 if (!(c1 = (uchar*)f1->dirname)) {
1469 state1 = fnc_BASE;
1470 c1 = (uchar*)f1->basename;
1471 } else
1472 state1 = fnc_DIR;
1473 if (!(c2 = (uchar*)f2->dirname)) {
1474 state2 = fnc_BASE;
1475 c2 = (uchar*)f2->basename;
1476 } else
1477 state2 = fnc_DIR;
1478
1479 while (1) {
1480 if ((dif = (int)*c1 - (int)*c2) != 0)
1481 break;
1482 if (!*++c1) {
1483 switch (state1) {
1484 case fnc_DIR:
1485 state1 = fnc_SLASH;
1486 c1 = (uchar*)"/";
1487 break;
1488 case fnc_SLASH:
1489 state1 = fnc_BASE;
1490 c1 = (uchar*)f1->basename;
1491 break;
1492 case fnc_BASE:
1493 break;
1494 }
1495 }
1496 if (!*++c2) {
1497 switch (state2) {
1498 case fnc_DIR:
1499 state2 = fnc_SLASH;
1500 c2 = (uchar*)"/";
1501 break;
1502 case fnc_SLASH:
1503 state2 = fnc_BASE;
1504 c2 = (uchar*)f2->basename;
1505 break;
1506 case fnc_BASE:
1507 if (!*c1)
1508 return 0;
1509 break;
1510 }
1511 }
1512 }
1513
1514 return dif;
1515}
1516
1517
1518/* Return a copy of the full filename of a flist entry, using the indicated
1519 * buffer. No size-checking is done because we checked the size when creating
1520 * the file_struct entry.
1521 */
1522char *f_name_to(struct file_struct *f, char *fbuf)
1523{
1524 if (!f || !f->basename)
1525 return NULL;
1526
1527 if (f->dirname) {
1528 int len = strlen(f->dirname);
1529 memcpy(fbuf, f->dirname, len);
1530 fbuf[len] = '/';
1531 strcpy(fbuf + len + 1, f->basename);
1532 } else
1533 strcpy(fbuf, f->basename);
1534 return fbuf;
1535}
1536
1537
1538/* Like f_name_to(), but we rotate through 5 static buffers of our own.
1539 */
1540char *f_name(struct file_struct *f)
1541{
1542 static char names[5][MAXPATHLEN];
1543 static unsigned int n;
1544
1545 n = (n + 1) % (sizeof names / sizeof names[0]);
1546
1547 return f_name_to(f, names[n]);
1548}