Fixed the optional sanitization of the files_from arg.
[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, "[%s] expand file_list to %.0f bytes, did%s move\n",
346 who_am_i(),
347 (double)sizeof(flist->files[0])
348 * flist->malloced,
349 (new_ptr == flist->files) ? " not" : "");
350 }
351
352 flist->files = (struct file_struct **) new_ptr;
353
354 if (!flist->files)
355 out_of_memory("flist_expand");
356 }
357}
358
359void send_file_entry(struct file_struct *file, int f, unsigned short base_flags)
360{
361 unsigned short flags;
362 static time_t modtime;
363 static mode_t mode;
364 static DEV64_T rdev; /* just high bytes in p28 onward */
365 static DEV64_T dev;
366 static uid_t uid;
367 static gid_t gid;
368 static char lastname[MAXPATHLEN];
369 char *fname, fbuf[MAXPATHLEN];
370 int l1, l2;
371
372 if (f == -1)
373 return;
374
375 if (!file) {
376 write_byte(f, 0);
377 modtime = 0, mode = 0;
378 rdev = 0, dev = 0;
379 uid = 0, gid = 0;
380 *lastname = '\0';
381 return;
382 }
383
384 io_write_phase = "send_file_entry";
385
386 fname = f_name_to(file, fbuf);
387
388 flags = base_flags;
389
390 if (file->mode == mode)
391 flags |= XMIT_SAME_MODE;
392 else
393 mode = file->mode;
394 if (preserve_devices) {
395 if (protocol_version < 28) {
396 if (IS_DEVICE(mode)) {
397 if (file->u.rdev == rdev) {
398 /* Set both flags to simplify the test
399 * when writing the data. */
400 flags |= XMIT_SAME_RDEV_pre28
401 | XMIT_SAME_HIGH_RDEV;
402 } else
403 rdev = file->u.rdev;
404 } else
405 rdev = 0;
406 } else if (IS_DEVICE(mode)) {
407 if ((file->u.rdev & ~0xFF) == rdev)
408 flags |= XMIT_SAME_HIGH_RDEV;
409 else
410 rdev = file->u.rdev & ~0xFF;
411 }
412 }
413 if (file->uid == uid)
414 flags |= XMIT_SAME_UID;
415 else
416 uid = file->uid;
417 if (file->gid == gid)
418 flags |= XMIT_SAME_GID;
419 else
420 gid = file->gid;
421 if (file->modtime == modtime)
422 flags |= XMIT_SAME_TIME;
423 else
424 modtime = file->modtime;
425 if (file->link_u.idev) {
426 if (file->F_DEV == dev) {
427 if (protocol_version >= 28)
428 flags |= XMIT_SAME_DEV;
429 } else
430 dev = file->F_DEV;
431 flags |= XMIT_HAS_IDEV_DATA;
432 }
433
434 for (l1 = 0;
435 lastname[l1] && (fname[l1] == lastname[l1]) && (l1 < 255);
436 l1++) {}
437 l2 = strlen(fname+l1);
438
439 if (l1 > 0)
440 flags |= XMIT_SAME_NAME;
441 if (l2 > 255)
442 flags |= XMIT_LONG_NAME;
443
444 /* We must make sure we don't send a zero flags byte or
445 * the other end will terminate the flist transfer. */
446 if (flags == 0 && !S_ISDIR(mode))
447 flags |= XMIT_TOP_DIR; /* NOTE: no meaning for non-dir */
448 if (protocol_version >= 28) {
449 if ((flags & 0xFF00) || flags == 0) {
450 flags |= XMIT_EXTENDED_FLAGS;
451 write_byte(f, flags);
452 write_byte(f, flags >> 8);
453 } else
454 write_byte(f, flags);
455 } else {
456 if (flags == 0)
457 flags |= XMIT_LONG_NAME;
458 write_byte(f, flags);
459 }
460 if (flags & XMIT_SAME_NAME)
461 write_byte(f, l1);
462 if (flags & XMIT_LONG_NAME)
463 write_int(f, l2);
464 else
465 write_byte(f, l2);
466 write_buf(f, fname + l1, l2);
467
468 write_longint(f, file->length);
469 if (!(flags & XMIT_SAME_TIME))
470 write_int(f, modtime);
471 if (!(flags & XMIT_SAME_MODE))
472 write_int(f, to_wire_mode(mode));
473 if (preserve_uid && !(flags & XMIT_SAME_UID)) {
474 add_uid(uid);
475 write_int(f, uid);
476 }
477 if (preserve_gid && !(flags & XMIT_SAME_GID)) {
478 add_gid(gid);
479 write_int(f, gid);
480 }
481 if (preserve_devices && IS_DEVICE(mode)) {
482 /* If XMIT_SAME_HIGH_RDEV is off, XMIT_SAME_RDEV_pre28 is
483 * also off. Also, avoid using "rdev" because it may be
484 * incomplete. */
485 if (!(flags & XMIT_SAME_HIGH_RDEV))
486 write_int(f, file->u.rdev);
487 else if (protocol_version >= 28)
488 write_byte(f, file->u.rdev);
489 }
490
491#if SUPPORT_LINKS
492 if (preserve_links && S_ISLNK(mode)) {
493 write_int(f, strlen(file->u.link));
494 write_buf(f, file->u.link, strlen(file->u.link));
495 }
496#endif
497
498#if SUPPORT_HARD_LINKS
499 if (flags & XMIT_HAS_IDEV_DATA) {
500 if (protocol_version < 26) {
501 /* 32-bit dev_t and ino_t */
502 write_int(f, dev);
503 write_int(f, file->F_INODE);
504 } else {
505 /* 64-bit dev_t and ino_t */
506 if (!(flags & XMIT_SAME_DEV))
507 write_longint(f, dev);
508 write_longint(f, file->F_INODE);
509 }
510 }
511#endif
512
513 if (always_checksum) {
514 char *sum;
515 if (S_ISREG(mode))
516 sum = file->u.sum;
517 else if (protocol_version < 28) {
518 /* Prior to 28, we sent a useless set of nulls. */
519 sum = empty_sum;
520 } else
521 sum = NULL;
522 if (sum) {
523 write_buf(f, sum, protocol_version < 21? 2
524 : MD4_SUM_LENGTH);
525 }
526 }
527
528 strlcpy(lastname, fname, MAXPATHLEN);
529
530 io_write_phase = "unknown";
531}
532
533
534
535void receive_file_entry(struct file_struct **fptr, unsigned short flags, int f)
536{
537 static time_t modtime;
538 static mode_t mode;
539 static DEV64_T rdev; /* just high bytes in p28 onward */
540 static DEV64_T dev;
541 static uid_t uid;
542 static gid_t gid;
543 static char lastname[MAXPATHLEN];
544 char thisname[MAXPATHLEN];
545 unsigned int l1 = 0, l2 = 0;
546 char *p;
547 struct file_struct *file;
548
549 if (!fptr) {
550 modtime = 0, mode = 0;
551 rdev = 0, dev = 0;
552 uid = 0, gid = 0;
553 *lastname = '\0';
554 return;
555 }
556
557 if (flags & XMIT_SAME_NAME)
558 l1 = read_byte(f);
559
560 if (flags & XMIT_LONG_NAME)
561 l2 = read_int(f);
562 else
563 l2 = read_byte(f);
564
565 file = new(struct file_struct);
566 if (!file)
567 out_of_memory("receive_file_entry");
568 memset((char *) file, 0, sizeof(*file));
569 (*fptr) = file;
570
571 if (l2 >= MAXPATHLEN - l1) {
572 rprintf(FERROR,
573 "overflow: flags=0x%x l1=%d l2=%d lastname=%s\n",
574 flags, l1, l2, lastname);
575 overflow("receive_file_entry");
576 }
577
578 strlcpy(thisname, lastname, l1 + 1);
579 read_sbuf(f, &thisname[l1], l2);
580 thisname[l1 + l2] = 0;
581
582 strlcpy(lastname, thisname, MAXPATHLEN);
583 lastname[MAXPATHLEN - 1] = 0;
584
585 clean_fname(thisname);
586
587 if (sanitize_paths) {
588 sanitize_path(thisname, NULL);
589 }
590
591 if ((p = strrchr(thisname, '/'))) {
592 static char *lastdir;
593 *p = 0;
594 if (lastdir && strcmp(thisname, lastdir) == 0)
595 file->dirname = lastdir;
596 else {
597 file->dirname = strdup(thisname);
598 lastdir = file->dirname;
599 }
600 file->basename = strdup(p + 1);
601 } else {
602 file->dirname = NULL;
603 file->basename = strdup(thisname);
604 }
605
606 if (!file->basename)
607 out_of_memory("receive_file_entry 1");
608
609 file->flags = flags & XMIT_TOP_DIR ? FLAG_TOP_DIR : 0;
610 file->length = read_longint(f);
611 if (!(flags & XMIT_SAME_TIME))
612 modtime = (time_t)read_int(f);
613 file->modtime = modtime;
614 if (!(flags & XMIT_SAME_MODE))
615 mode = from_wire_mode(read_int(f));
616 file->mode = mode;
617
618 if (preserve_uid) {
619 if (!(flags & XMIT_SAME_UID))
620 uid = (uid_t)read_int(f);
621 file->uid = uid;
622 }
623 if (preserve_gid) {
624 if (!(flags & XMIT_SAME_GID))
625 gid = (gid_t)read_int(f);
626 file->gid = gid;
627 }
628 if (preserve_devices) {
629 if (protocol_version < 28) {
630 if (IS_DEVICE(mode)) {
631 if (!(flags & XMIT_SAME_RDEV_pre28))
632 rdev = (DEV64_T)read_int(f);
633 file->u.rdev = rdev;
634 } else
635 rdev = 0;
636 } else if (IS_DEVICE(mode)) {
637 if (!(flags & XMIT_SAME_HIGH_RDEV)) {
638 file->u.rdev = (DEV64_T)read_int(f);
639 rdev = file->u.rdev & ~0xFF;
640 } else
641 file->u.rdev = rdev | (DEV64_T)read_byte(f);
642 }
643 }
644
645 if (preserve_links && S_ISLNK(mode)) {
646 int l = read_int(f);
647 if (l < 0) {
648 rprintf(FERROR, "overflow: l=%d\n", l);
649 overflow("receive_file_entry");
650 }
651 if (!(file->u.link = new_array(char, l + 1)))
652 out_of_memory("receive_file_entry 2");
653 read_sbuf(f, file->u.link, l);
654 if (sanitize_paths)
655 sanitize_path(file->u.link, file->dirname);
656 }
657#if SUPPORT_HARD_LINKS
658 if (preserve_hard_links && protocol_version < 28 && S_ISREG(mode))
659 flags |= XMIT_HAS_IDEV_DATA;
660 if (flags & XMIT_HAS_IDEV_DATA) {
661 if (!(file->link_u.idev = new(struct idev)))
662 out_of_memory("file inode data");
663 if (protocol_version < 26) {
664 dev = read_int(f);
665 file->F_INODE = read_int(f);
666 } else {
667 if (!(flags & XMIT_SAME_DEV))
668 dev = read_longint(f);
669 file->F_INODE = read_longint(f);
670 }
671 file->F_DEV = dev;
672 }
673#endif
674
675 if (always_checksum) {
676 char *sum;
677 if (S_ISREG(mode)) {
678 sum = file->u.sum = new_array(char, MD4_SUM_LENGTH);
679 if (!sum)
680 out_of_memory("md4 sum");
681 } else if (protocol_version < 28) {
682 /* Prior to 28, we get a useless set of nulls. */
683 sum = empty_sum;
684 } else
685 sum = NULL;
686 if (sum) {
687 read_buf(f, sum, protocol_version < 21? 2
688 : MD4_SUM_LENGTH);
689 }
690 }
691
692 if (!preserve_perms) {
693 extern int orig_umask;
694 /* set an appropriate set of permissions based on original
695 * permissions and umask. This emulates what GNU cp does */
696 file->mode &= ~orig_umask;
697 }
698}
699
700
701#define STRDUP(ap, p) (ap ? string_area_strdup(ap, p) : strdup(p))
702/* IRIX cc cares that the operands to the ternary have the same type. */
703#define MALLOC(ap, i) (ap ? (void*) string_area_malloc(ap, i) : malloc(i))
704
705/**
706 * Create a file_struct for a named file by reading its stat()
707 * information and performing extensive checks against global
708 * options.
709 *
710 * @return the new file, or NULL if there was an error or this file
711 * should be excluded.
712 *
713 * @todo There is a small optimization opportunity here to avoid
714 * stat()ing the file in some circumstances, which has a certain cost.
715 * We are called immediately after doing readdir(), and so we may
716 * already know the d_type of the file. We could for example avoid
717 * statting directories if we're not recursing, but this is not a very
718 * important case. Some systems may not have d_type.
719 **/
720struct file_struct *make_file(char *fname, struct string_area **ap,
721 int exclude_level)
722{
723 struct file_struct *file;
724 STRUCT_STAT st;
725 char sum[SUM_LENGTH];
726 char *p;
727 char cleaned_name[MAXPATHLEN];
728 char linkbuf[MAXPATHLEN];
729 unsigned short flags = 0;
730
731 if (strlcpy(cleaned_name, fname, sizeof cleaned_name)
732 >= sizeof cleaned_name - flist_dir_len) {
733 rprintf(FINFO, "skipping overly long name: %s\n", fname);
734 return NULL;
735 }
736 clean_fname(cleaned_name);
737 if (sanitize_paths)
738 sanitize_path(cleaned_name, NULL);
739 fname = cleaned_name;
740
741 memset(sum, 0, SUM_LENGTH);
742
743 if (readlink_stat(fname, &st, linkbuf) != 0) {
744 int save_errno = errno;
745 if (errno == ENOENT && exclude_level != NO_EXCLUDES) {
746 /* either symlink pointing nowhere or file that
747 * was removed during rsync run; see if excluded
748 * before reporting an error */
749 if (check_exclude_file(fname, 0, exclude_level)) {
750 /* file is excluded anyway, ignore silently */
751 return NULL;
752 }
753 }
754 io_error |= IOERR_GENERAL;
755 rprintf(FERROR, "readlink %s failed: %s\n",
756 full_fname(fname), strerror(save_errno));
757 return NULL;
758 }
759
760 /* backup.c calls us with exclude_level set to NO_EXCLUDES. */
761 if (exclude_level == NO_EXCLUDES)
762 goto skip_excludes;
763
764 if (S_ISDIR(st.st_mode) && !recurse && !files_from) {
765 rprintf(FINFO, "skipping directory %s\n", fname);
766 return NULL;
767 }
768
769 if (one_file_system && st.st_dev != filesystem_dev) {
770 /* We allow a directory though to preserve the mount point.
771 * However, flag it so that we don't recurse. */
772 if (!S_ISDIR(st.st_mode))
773 return NULL;
774 flags |= FLAG_MOUNT_POINT;
775 }
776
777 if (check_exclude_file(fname, S_ISDIR(st.st_mode) != 0, exclude_level))
778 return NULL;
779
780 if (lp_ignore_nonreadable(module_id) && access(fname, R_OK) != 0)
781 return NULL;
782
783 skip_excludes:
784
785 if (verbose > 2) {
786 rprintf(FINFO, "[%s] make_file(%s,*,%d)\n",
787 who_am_i(), fname, exclude_level);
788 }
789
790 file = new(struct file_struct);
791 if (!file)
792 out_of_memory("make_file");
793 memset((char *) file, 0, sizeof(*file));
794 file->flags = flags;
795
796 if ((p = strrchr(fname, '/'))) {
797 static char *lastdir;
798 *p = 0;
799 if (lastdir && strcmp(fname, lastdir) == 0)
800 file->dirname = lastdir;
801 else {
802 file->dirname = strdup(fname);
803 lastdir = file->dirname;
804 }
805 file->basename = STRDUP(ap, p + 1);
806 *p = '/';
807 } else {
808 file->dirname = NULL;
809 file->basename = STRDUP(ap, fname);
810 }
811
812 file->modtime = st.st_mtime;
813 file->length = st.st_size;
814 file->mode = st.st_mode;
815 file->uid = st.st_uid;
816 file->gid = st.st_gid;
817 if (preserve_hard_links) {
818 if (protocol_version < 28 ? S_ISREG(st.st_mode)
819 : !S_ISDIR(st.st_mode) && st.st_nlink > 1) {
820 if (!(file->link_u.idev = new(struct idev)))
821 out_of_memory("file inode data");
822 file->F_DEV = st.st_dev;
823 file->F_INODE = st.st_ino;
824 }
825 }
826#ifdef HAVE_STRUCT_STAT_ST_RDEV
827 if (IS_DEVICE(st.st_mode))
828 file->u.rdev = st.st_rdev;
829#endif
830
831#if SUPPORT_LINKS
832 if (S_ISLNK(st.st_mode))
833 file->u.link = STRDUP(ap, linkbuf);
834#endif
835
836 if (always_checksum && S_ISREG(st.st_mode)) {
837 if (!(file->u.sum = (char*)MALLOC(ap, MD4_SUM_LENGTH)))
838 out_of_memory("md4 sum");
839 file_checksum(fname, file->u.sum, st.st_size);
840 }
841
842 file->basedir = flist_dir;
843
844 if (!S_ISDIR(st.st_mode))
845 stats.total_size += st.st_size;
846
847 return file;
848}
849
850
851void send_file_name(int f, struct file_list *flist, char *fname,
852 int recursive, unsigned short base_flags)
853{
854 struct file_struct *file;
855 char fbuf[MAXPATHLEN];
856 extern int delete_excluded;
857
858 /* f is set to -1 when calculating deletion file list */
859 file = make_file(fname, &flist->string_area,
860 f == -1 && delete_excluded? SERVER_EXCLUDES
861 : ALL_EXCLUDES);
862
863 if (!file)
864 return;
865
866 maybe_emit_filelist_progress(flist);
867
868 flist_expand(flist);
869
870 if (write_batch)
871 file->flags |= FLAG_TOP_DIR;
872
873 if (file->basename[0]) {
874 flist->files[flist->count++] = file;
875 send_file_entry(file, f, base_flags);
876 }
877
878 if (recursive && S_ISDIR(file->mode)
879 && !(file->flags & FLAG_MOUNT_POINT)) {
880 struct exclude_struct **last_exclude_list = local_exclude_list;
881 send_directory(f, flist, f_name_to(file, fbuf));
882 local_exclude_list = last_exclude_list;
883 return;
884 }
885}
886
887
888static void send_directory(int f, struct file_list *flist, char *dir)
889{
890 DIR *d;
891 struct dirent *di;
892 char fname[MAXPATHLEN];
893 unsigned int offset;
894 char *p;
895
896 d = opendir(dir);
897 if (!d) {
898 io_error |= IOERR_GENERAL;
899 rprintf(FERROR, "opendir %s failed: %s\n",
900 full_fname(dir), strerror(errno));
901 return;
902 }
903
904 offset = strlcpy(fname, dir, MAXPATHLEN);
905 p = fname + offset;
906 if (offset >= MAXPATHLEN || p[-1] != '/') {
907 if (offset >= MAXPATHLEN - 1) {
908 io_error |= IOERR_GENERAL;
909 rprintf(FERROR, "skipping long-named directory: %s\n",
910 full_fname(fname));
911 closedir(d);
912 return;
913 }
914 *p++ = '/';
915 offset++;
916 }
917
918 local_exclude_list = NULL;
919
920 if (cvs_exclude) {
921 if (strlcpy(p, ".cvsignore", MAXPATHLEN - offset)
922 < MAXPATHLEN - offset)
923 add_exclude_file(&local_exclude_list,fname,MISSING_OK,ADD_EXCLUDE);
924 else {
925 io_error |= IOERR_GENERAL;
926 rprintf(FINFO,
927 "cannot cvs-exclude in long-named directory %s\n",
928 full_fname(fname));
929 }
930 }
931
932 for (errno = 0, di = readdir(d); di; errno = 0, di = readdir(d)) {
933 char *dname = d_name(di);
934 if (dname[0] == '.' && (dname[1] == '\0'
935 || (dname[1] == '.' && dname[2] == '\0')))
936 continue;
937 if (strlcpy(p, dname, MAXPATHLEN - offset) < MAXPATHLEN - offset)
938 send_file_name(f, flist, fname, recurse, 0);
939 else {
940 io_error |= IOERR_GENERAL;
941 rprintf(FINFO,
942 "cannot send long-named file %s\n",
943 full_fname(fname));
944 }
945 }
946 if (errno) {
947 io_error |= IOERR_GENERAL;
948 rprintf(FERROR, "readdir(%s): (%d) %s\n",
949 dir, errno, strerror(errno));
950 }
951
952 if (local_exclude_list)
953 free_exclude_list(&local_exclude_list); /* Zeros pointer too */
954
955 closedir(d);
956}
957
958
959/**
960 * The delete_files() function in receiver.c sets f to -1 so that we just
961 * construct the file list in memory without sending it over the wire. It
962 * also has the side-effect of ignoring user-excludes if delete_excluded
963 * is set (so that the delete list includes user-excluded files).
964 **/
965struct file_list *send_file_list(int f, int argc, char *argv[])
966{
967 int l;
968 STRUCT_STAT st;
969 char *p, *dir, olddir[sizeof curr_dir];
970 char lastpath[MAXPATHLEN] = "";
971 struct file_list *flist;
972 int64 start_write;
973 int use_ff_fd = 0;
974
975 if (show_filelist_p() && f != -1)
976 start_filelist_progress("building file list");
977
978 start_write = stats.total_written;
979
980 flist = flist_new();
981
982 if (f != -1) {
983 io_start_buffering_out(f);
984 if (filesfrom_fd >= 0) {
985 if (argv[0] && !push_dir(argv[0])) {
986 rprintf(FERROR, "push_dir %s failed: %s\n",
987 full_fname(argv[0]), strerror(errno));
988 exit_cleanup(RERR_FILESELECT);
989 }
990 use_ff_fd = 1;
991 }
992 }
993
994 while (1) {
995 char fname2[MAXPATHLEN];
996 char *fname = fname2;
997
998 if (use_ff_fd) {
999 if (read_filesfrom_line(filesfrom_fd, fname) == 0)
1000 break;
1001 sanitize_path(fname, NULL);
1002 } else {
1003 if (argc-- == 0)
1004 break;
1005 strlcpy(fname, *argv++, MAXPATHLEN);
1006 if (sanitize_paths)
1007 sanitize_path(fname, NULL);
1008 }
1009
1010 l = strlen(fname);
1011 if (fname[l - 1] == '/') {
1012 if (l == 2 && fname[0] == '.') {
1013 /* Turn "./" into just "." rather than "./." */
1014 fname[1] = '\0';
1015 } else if (l < MAXPATHLEN) {
1016 fname[l++] = '.';
1017 fname[l] = '\0';
1018 }
1019 }
1020
1021 if (link_stat(fname, &st) != 0) {
1022 if (f != -1) {
1023 io_error |= IOERR_GENERAL;
1024 rprintf(FERROR, "link_stat %s failed: %s\n",
1025 full_fname(fname), strerror(errno));
1026 }
1027 continue;
1028 }
1029
1030 if (S_ISDIR(st.st_mode) && !recurse && !files_from) {
1031 rprintf(FINFO, "skipping directory %s\n", fname);
1032 continue;
1033 }
1034
1035 dir = NULL;
1036 olddir[0] = '\0';
1037
1038 if (!relative_paths) {
1039 p = strrchr(fname, '/');
1040 if (p) {
1041 *p = 0;
1042 if (p == fname)
1043 dir = "/";
1044 else
1045 dir = fname;
1046 fname = p + 1;
1047 }
1048 } else if (f != -1 && implied_dirs && (p=strrchr(fname,'/')) && p != fname) {
1049 /* this ensures we send the intermediate directories,
1050 thus getting their permissions right */
1051 char *lp = lastpath, *fn = fname, *slash = fname;
1052 *p = 0;
1053 /* Skip any initial directories in our path that we
1054 * have in common with lastpath. */
1055 while (*fn && *lp == *fn) {
1056 if (*fn == '/')
1057 slash = fn;
1058 lp++, fn++;
1059 }
1060 *p = '/';
1061 if (fn != p || (*lp && *lp != '/')) {
1062 int copy_links_saved = copy_links;
1063 int recurse_saved = recurse;
1064 copy_links = copy_unsafe_links;
1065 /* set recurse to 1 to prevent make_file
1066 * from ignoring directory, but still
1067 * turn off the recursive parameter to
1068 * send_file_name */
1069 recurse = 1;
1070 while ((slash = strchr(slash+1, '/')) != 0) {
1071 *slash = 0;
1072 send_file_name(f, flist, fname, 0, 0);
1073 *slash = '/';
1074 }
1075 copy_links = copy_links_saved;
1076 recurse = recurse_saved;
1077 *p = 0;
1078 strlcpy(lastpath, fname, sizeof lastpath);
1079 *p = '/';
1080 }
1081 }
1082
1083 if (!*fname)
1084 fname = ".";
1085
1086 if (dir && *dir) {
1087 static char *lastdir;
1088 static int lastdir_len;
1089
1090 strcpy(olddir, curr_dir); /* can't overflow */
1091
1092 if (!push_dir(dir)) {
1093 io_error |= IOERR_GENERAL;
1094 rprintf(FERROR, "push_dir %s failed: %s\n",
1095 full_fname(dir), strerror(errno));
1096 continue;
1097 }
1098
1099 if (lastdir && strcmp(lastdir, dir) == 0) {
1100 flist_dir = lastdir;
1101 flist_dir_len = lastdir_len;
1102 } else {
1103 flist_dir = lastdir = strdup(dir);
1104 flist_dir_len = lastdir_len = strlen(dir);
1105 }
1106 }
1107
1108 if (one_file_system)
1109 set_filesystem(fname);
1110
1111 send_file_name(f, flist, fname, recurse, XMIT_TOP_DIR);
1112
1113 if (olddir[0]) {
1114 flist_dir = NULL;
1115 flist_dir_len = 0;
1116 if (!pop_dir(olddir)) {
1117 rprintf(FERROR, "pop_dir %s failed: %s\n",
1118 full_fname(dir), strerror(errno));
1119 exit_cleanup(RERR_FILESELECT);
1120 }
1121 }
1122 }
1123
1124 if (f != -1) {
1125 send_file_entry(NULL, f, 0);
1126
1127 if (show_filelist_p())
1128 finish_filelist_progress(flist);
1129 }
1130
1131 clean_flist(flist, 0, 0);
1132
1133 if (f != -1) {
1134 /* Now send the uid/gid list. This was introduced in
1135 * protocol version 15 */
1136 send_uid_list(f);
1137
1138 /* send the io_error flag */
1139 write_int(f, lp_ignore_errors(module_id) ? 0 : io_error);
1140
1141 io_end_buffering();
1142 stats.flist_size = stats.total_written - start_write;
1143 stats.num_files = flist->count;
1144 if (write_batch)
1145 write_batch_flist_info(flist->count, flist->files);
1146 }
1147
1148 if (verbose > 2)
1149 rprintf(FINFO, "send_file_list done\n");
1150
1151 return flist;
1152}
1153
1154
1155struct file_list *recv_file_list(int f)
1156{
1157 struct file_list *flist;
1158 unsigned short flags;
1159 int64 start_read;
1160 extern int list_only;
1161
1162 if (show_filelist_p())
1163 start_filelist_progress("receiving file list");
1164
1165 start_read = stats.total_read;
1166
1167 flist = new(struct file_list);
1168 if (!flist)
1169 goto oom;
1170
1171 flist->count = 0;
1172 flist->malloced = 1000;
1173 flist->files = new_array(struct file_struct *, flist->malloced);
1174 if (!flist->files)
1175 goto oom;
1176
1177
1178 while ((flags = read_byte(f)) != 0) {
1179 int i = flist->count;
1180
1181 flist_expand(flist);
1182
1183 if (protocol_version >= 28 && (flags & XMIT_EXTENDED_FLAGS))
1184 flags |= read_byte(f) << 8;
1185 receive_file_entry(&flist->files[i], flags, f);
1186
1187 if (S_ISREG(flist->files[i]->mode))
1188 stats.total_size += flist->files[i]->length;
1189
1190 flist->count++;
1191
1192 maybe_emit_filelist_progress(flist);
1193
1194 if (verbose > 2) {
1195 rprintf(FINFO, "recv_file_name(%s)\n",
1196 f_name(flist->files[i]));
1197 }
1198 }
1199 receive_file_entry(NULL, 0, 0); /* Signal that we're done. */
1200
1201 if (verbose > 2)
1202 rprintf(FINFO, "received %d names\n", flist->count);
1203
1204 if (show_filelist_p())
1205 finish_filelist_progress(flist);
1206
1207 clean_flist(flist, relative_paths, 1);
1208
1209 if (f != -1) {
1210 /* Now send the uid/gid list. This was introduced in
1211 * protocol version 15 */
1212 recv_uid_list(f, flist);
1213
1214 if (!read_batch) {
1215 /* Recv the io_error flag */
1216 if (lp_ignore_errors(module_id) || ignore_errors)
1217 read_int(f);
1218 else
1219 io_error |= read_int(f);
1220 }
1221 }
1222
1223 if (list_only) {
1224 int i;
1225 for (i = 0; i < flist->count; i++)
1226 list_file_entry(flist->files[i]);
1227 }
1228
1229 if (verbose > 2)
1230 rprintf(FINFO, "recv_file_list done\n");
1231
1232 stats.flist_size = stats.total_read - start_read;
1233 stats.num_files = flist->count;
1234
1235 return flist;
1236
1237 oom:
1238 out_of_memory("recv_file_list");
1239 return NULL; /* not reached */
1240}
1241
1242
1243int file_compare(struct file_struct **file1, struct file_struct **file2)
1244{
1245 struct file_struct *f1 = *file1;
1246 struct file_struct *f2 = *file2;
1247
1248 if (!f1->basename && !f2->basename)
1249 return 0;
1250 if (!f1->basename)
1251 return -1;
1252 if (!f2->basename)
1253 return 1;
1254 if (f1->dirname == f2->dirname)
1255 return u_strcmp(f1->basename, f2->basename);
1256 return f_name_cmp(f1, f2);
1257}
1258
1259
1260int flist_find(struct file_list *flist, struct file_struct *f)
1261{
1262 int low = 0, high = flist->count - 1;
1263
1264 while (high >= 0 && !flist->files[high]->basename) high--;
1265
1266 if (high < 0)
1267 return -1;
1268
1269 while (low != high) {
1270 int mid = (low + high) / 2;
1271 int ret = file_compare(&flist->files[flist_up(flist, mid)],&f);
1272 if (ret == 0)
1273 return flist_up(flist, mid);
1274 if (ret > 0)
1275 high = mid;
1276 else
1277 low = mid + 1;
1278 }
1279
1280 if (file_compare(&flist->files[flist_up(flist, low)], &f) == 0)
1281 return flist_up(flist, low);
1282 return -1;
1283}
1284
1285
1286/*
1287 * free up one file
1288 */
1289void free_file(struct file_struct *file)
1290{
1291 if (!file)
1292 return;
1293 if (file->basename)
1294 free(file->basename);
1295 if (!IS_DEVICE(file->mode) && file->u.link)
1296 free(file->u.link); /* Handles u.sum too. */
1297 if (file->link_u.idev)
1298 free((char*)file->link_u.idev); /* Handles link_u.links too. */
1299 *file = null_file;
1300}
1301
1302
1303/*
1304 * allocate a new file list
1305 */
1306struct file_list *flist_new(void)
1307{
1308 struct file_list *flist;
1309
1310 flist = new(struct file_list);
1311 if (!flist)
1312 out_of_memory("send_file_list");
1313
1314 flist->count = 0;
1315 flist->malloced = 0;
1316 flist->files = NULL;
1317
1318#if ARENA_SIZE > 0
1319 flist->string_area = string_area_new(0);
1320#else
1321 flist->string_area = NULL;
1322#endif
1323 return flist;
1324}
1325
1326/*
1327 * free up all elements in a flist
1328 */
1329void flist_free(struct file_list *flist)
1330{
1331 int i;
1332 for (i = 1; i < flist->count; i++) {
1333 if (!flist->string_area)
1334 free_file(flist->files[i]);
1335 free(flist->files[i]);
1336 }
1337 /* FIXME: I don't think we generally need to blank the flist
1338 * since it's about to be freed. This will just cause more
1339 * memory traffic. If you want a freed-memory debugger, you
1340 * know where to get it. */
1341 memset((char *) flist->files, 0,
1342 sizeof(flist->files[0]) * flist->count);
1343 free(flist->files);
1344 if (flist->string_area)
1345 string_area_free(flist->string_area);
1346 memset((char *) flist, 0, sizeof(*flist));
1347 free(flist);
1348}
1349
1350
1351/*
1352 * This routine ensures we don't have any duplicate names in our file list.
1353 * duplicate names can cause corruption because of the pipelining
1354 */
1355static void clean_flist(struct file_list *flist, int strip_root, int no_dups)
1356{
1357 int i, prev_i = 0;
1358
1359 if (!flist || flist->count == 0)
1360 return;
1361
1362 qsort(flist->files, flist->count,
1363 sizeof(flist->files[0]), (int (*)()) file_compare);
1364
1365 for (i = no_dups? 0 : flist->count; i < flist->count; i++) {
1366 if (flist->files[i]->basename) {
1367 prev_i = i;
1368 break;
1369 }
1370 }
1371 while (++i < flist->count) {
1372 if (!flist->files[i]->basename)
1373 continue;
1374 if (f_name_cmp(flist->files[i], flist->files[prev_i]) == 0) {
1375 if (verbose > 1 && !am_server) {
1376 rprintf(FINFO,
1377 "removing duplicate name %s from file list %d\n",
1378 f_name(flist->files[i]), i);
1379 }
1380 /* Make sure that if we unduplicate '.', that we don't
1381 * lose track of a user-specified starting point (or
1382 * else deletions will mysteriously fail with -R). */
1383 if (flist->files[i]->flags & FLAG_TOP_DIR)
1384 flist->files[prev_i]->flags |= FLAG_TOP_DIR;
1385 /* it's not great that the flist knows the semantics of
1386 * the file memory usage, but i'd rather not add a flag
1387 * byte to that struct.
1388 * XXX can i use a bit in the flags field? */
1389 if (flist->string_area)
1390 flist->files[i][0] = null_file;
1391 else
1392 free_file(flist->files[i]);
1393 } else
1394 prev_i = i;
1395 }
1396
1397 if (strip_root) {
1398 /* we need to strip off the root directory in the case
1399 of relative paths, but this must be done _after_
1400 the sorting phase */
1401 for (i = 0; i < flist->count; i++) {
1402 if (flist->files[i]->dirname &&
1403 flist->files[i]->dirname[0] == '/') {
1404 memmove(&flist->files[i]->dirname[0],
1405 &flist->files[i]->dirname[1],
1406 strlen(flist->files[i]->dirname));
1407 }
1408
1409 if (flist->files[i]->dirname &&
1410 !flist->files[i]->dirname[0]) {
1411 flist->files[i]->dirname = NULL;
1412 }
1413 }
1414 }
1415
1416 if (verbose <= 3)
1417 return;
1418
1419 for (i = 0; i < flist->count; i++) {
1420 rprintf(FINFO, "[%s] i=%d %s %s %s mode=0%o len=%.0f\n",
1421 who_am_i(), i,
1422 NS(flist->files[i]->basedir),
1423 NS(flist->files[i]->dirname),
1424 NS(flist->files[i]->basename),
1425 (int) flist->files[i]->mode,
1426 (double) flist->files[i]->length);
1427 }
1428}
1429
1430
1431enum fnc_state { fnc_DIR, fnc_SLASH, fnc_BASE };
1432
1433/* Compare the names of two file_struct entities, just like strcmp()
1434 * would do if it were operating on the joined strings. We assume
1435 * that there are no 0-length strings.
1436 */
1437int f_name_cmp(struct file_struct *f1, struct file_struct *f2)
1438{
1439 int dif;
1440 const uchar *c1, *c2;
1441 enum fnc_state state1, state2;
1442
1443 if (!f1 || !f1->basename) {
1444 if (!f2 || !f2->basename)
1445 return 0;
1446 return -1;
1447 }
1448 if (!f2 || !f2->basename)
1449 return 1;
1450
1451 if (!(c1 = (uchar*)f1->dirname)) {
1452 state1 = fnc_BASE;
1453 c1 = (uchar*)f1->basename;
1454 } else
1455 state1 = fnc_DIR;
1456 if (!(c2 = (uchar*)f2->dirname)) {
1457 state2 = fnc_BASE;
1458 c2 = (uchar*)f2->basename;
1459 } else
1460 state2 = fnc_DIR;
1461
1462 while (1) {
1463 if ((dif = (int)*c1 - (int)*c2) != 0)
1464 break;
1465 if (!*++c1) {
1466 switch (state1) {
1467 case fnc_DIR:
1468 state1 = fnc_SLASH;
1469 c1 = (uchar*)"/";
1470 break;
1471 case fnc_SLASH:
1472 state1 = fnc_BASE;
1473 c1 = (uchar*)f1->basename;
1474 break;
1475 case fnc_BASE:
1476 break;
1477 }
1478 }
1479 if (!*++c2) {
1480 switch (state2) {
1481 case fnc_DIR:
1482 state2 = fnc_SLASH;
1483 c2 = (uchar*)"/";
1484 break;
1485 case fnc_SLASH:
1486 state2 = fnc_BASE;
1487 c2 = (uchar*)f2->basename;
1488 break;
1489 case fnc_BASE:
1490 if (!*c1)
1491 return 0;
1492 break;
1493 }
1494 }
1495 }
1496
1497 return dif;
1498}
1499
1500
1501/* Return a copy of the full filename of a flist entry, using the indicated
1502 * buffer. No size-checking is done because we checked the size when creating
1503 * the file_struct entry.
1504 */
1505char *f_name_to(struct file_struct *f, char *fbuf)
1506{
1507 if (!f || !f->basename)
1508 return NULL;
1509
1510 if (f->dirname) {
1511 int len = strlen(f->dirname);
1512 memcpy(fbuf, f->dirname, len);
1513 fbuf[len] = '/';
1514 strcpy(fbuf + len + 1, f->basename);
1515 } else
1516 strcpy(fbuf, f->basename);
1517 return fbuf;
1518}
1519
1520
1521/* Like f_name_to(), but we rotate through 5 static buffers of our own.
1522 */
1523char *f_name(struct file_struct *f)
1524{
1525 static char names[5][MAXPATHLEN];
1526 static unsigned int n;
1527
1528 n = (n + 1) % (sizeof names / sizeof names[0]);
1529
1530 return f_name_to(f, names[n]);
1531}