A helper file for cleanup.c.
[rsync/rsync.git] / flist.c
... / ...
CommitLineData
1/*
2 * Generate and receive file lists.
3 *
4 * Copyright (C) 1996 Andrew Tridgell
5 * Copyright (C) 1996 Paul Mackerras
6 * Copyright (C) 2001, 2002 Martin Pool <mbp@samba.org>
7 * Copyright (C) 2002, 2003, 2004, 2005, 2006 Wayne Davison
8 *
9 * This program is free software; you can redistribute it and/or modify
10 * it under the terms of the GNU General Public License as published by
11 * the Free Software Foundation; either version 2 of the License, or
12 * (at your option) any later version.
13 *
14 * This program is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17 * GNU General Public License for more details.
18 *
19 * You should have received a copy of the GNU General Public License along
20 * with this program; if not, write to the Free Software Foundation, Inc.,
21 * 51 Franklin Street - Fifth Floor, Boston, MA 02110-1301, USA.
22 */
23
24#include "rsync.h"
25
26extern int verbose;
27extern int list_only;
28extern int am_root;
29extern int am_server;
30extern int am_daemon;
31extern int am_sender;
32extern int do_progress;
33extern int always_checksum;
34extern int module_id;
35extern int ignore_errors;
36extern int numeric_ids;
37extern int recurse;
38extern int xfer_dirs;
39extern int filesfrom_fd;
40extern int one_file_system;
41extern int copy_dirlinks;
42extern int keep_dirlinks;
43extern int preserve_links;
44extern int preserve_hard_links;
45extern int preserve_devices;
46extern int preserve_specials;
47extern int preserve_uid;
48extern int preserve_gid;
49extern int relative_paths;
50extern int implied_dirs;
51extern int prune_empty_dirs;
52extern int copy_links;
53extern int copy_unsafe_links;
54extern int protocol_version;
55extern int sanitize_paths;
56extern struct stats stats;
57extern struct file_list *the_file_list;
58
59extern char curr_dir[MAXPATHLEN];
60
61extern struct chmod_mode_struct *chmod_modes;
62
63extern struct filter_list_struct filter_list;
64extern struct filter_list_struct server_filter_list;
65
66int io_error;
67int checksum_len;
68dev_t filesystem_dev; /* used to implement -x */
69unsigned int file_struct_len;
70
71static char empty_sum[MD4_SUM_LENGTH];
72static int flist_count_offset;
73
74static void clean_flist(struct file_list *flist, int strip_root, int no_dups);
75static void output_flist(struct file_list *flist);
76
77void init_flist(void)
78{
79 struct file_struct f;
80
81 /* Figure out how big the file_struct is without trailing padding */
82 file_struct_len = offsetof(struct file_struct, flags) + sizeof f.flags;
83 checksum_len = protocol_version < 21 ? 2 : MD4_SUM_LENGTH;
84}
85
86static int show_filelist_p(void)
87{
88 return verbose && xfer_dirs && !am_server;
89}
90
91static void start_filelist_progress(char *kind)
92{
93 rprintf(FCLIENT, "%s ... ", kind);
94 if (verbose > 1 || do_progress)
95 rprintf(FCLIENT, "\n");
96 rflush(FINFO);
97}
98
99static void emit_filelist_progress(int count)
100{
101 rprintf(FCLIENT, " %d files...\r", count);
102}
103
104static void maybe_emit_filelist_progress(int count)
105{
106 if (do_progress && show_filelist_p() && (count % 100) == 0)
107 emit_filelist_progress(count);
108}
109
110static void finish_filelist_progress(const struct file_list *flist)
111{
112 if (do_progress) {
113 /* This overwrites the progress line */
114 rprintf(FINFO, "%d file%sto consider\n",
115 flist->count, flist->count == 1 ? " " : "s ");
116 } else
117 rprintf(FINFO, "done\n");
118}
119
120void show_flist_stats(void)
121{
122 /* Nothing yet */
123}
124
125static void list_file_entry(struct file_struct *f)
126{
127 char permbuf[PERMSTRING_SIZE];
128
129 if (!f->basename) {
130 /* this can happen if duplicate names were removed */
131 return;
132 }
133
134 permstring(permbuf, f->mode);
135
136#ifdef SUPPORT_LINKS
137 if (preserve_links && S_ISLNK(f->mode)) {
138 rprintf(FINFO, "%s %11.0f %s %s -> %s\n",
139 permbuf,
140 (double)f->length, timestring(f->modtime),
141 f_name(f, NULL), f->u.link);
142 } else
143#endif
144 {
145 rprintf(FINFO, "%s %11.0f %s %s\n",
146 permbuf,
147 (double)f->length, timestring(f->modtime),
148 f_name(f, NULL));
149 }
150}
151
152/* Stat either a symlink or its referent, depending on the settings of
153 * copy_links, copy_unsafe_links, etc. Returns -1 on error, 0 on success.
154 *
155 * If path is the name of a symlink, then the linkbuf buffer (which must hold
156 * MAXPATHLEN chars) will be set to the symlink's target string.
157 *
158 * The stat structure pointed to by stp will contain information about the
159 * link or the referent as appropriate, if they exist. */
160static int readlink_stat(const char *path, STRUCT_STAT *stp, char *linkbuf)
161{
162#ifdef SUPPORT_LINKS
163 if (link_stat(path, stp, copy_dirlinks) < 0)
164 return -1;
165 if (S_ISLNK(stp->st_mode)) {
166 int llen = readlink(path, linkbuf, MAXPATHLEN - 1);
167 if (llen < 0)
168 return -1;
169 linkbuf[llen] = '\0';
170 if (copy_unsafe_links && unsafe_symlink(linkbuf, path)) {
171 if (verbose > 1) {
172 rprintf(FINFO,"copying unsafe symlink \"%s\" -> \"%s\"\n",
173 path, linkbuf);
174 }
175 return do_stat(path, stp);
176 }
177 }
178 return 0;
179#else
180 return do_stat(path, stp);
181#endif
182}
183
184int link_stat(const char *path, STRUCT_STAT *stp, int follow_dirlinks)
185{
186#ifdef SUPPORT_LINKS
187 if (copy_links)
188 return do_stat(path, stp);
189 if (do_lstat(path, stp) < 0)
190 return -1;
191 if (follow_dirlinks && S_ISLNK(stp->st_mode)) {
192 STRUCT_STAT st;
193 if (do_stat(path, &st) == 0 && S_ISDIR(st.st_mode))
194 *stp = st;
195 }
196 return 0;
197#else
198 return do_stat(path, stp);
199#endif
200}
201
202/* This function is used to check if a file should be included/excluded
203 * from the list of files based on its name and type etc. The value of
204 * filter_level is set to either SERVER_FILTERS or ALL_FILTERS. */
205static int is_excluded(char *fname, int is_dir, int filter_level)
206{
207#if 0 /* This currently never happens, so avoid a useless compare. */
208 if (filter_level == NO_FILTERS)
209 return 0;
210#endif
211 if (fname) {
212 /* never exclude '.', even if somebody does --exclude '*' */
213 if (fname[0] == '.' && !fname[1])
214 return 0;
215 /* Handle the -R version of the '.' dir. */
216 if (fname[0] == '/') {
217 int len = strlen(fname);
218 if (fname[len-1] == '.' && fname[len-2] == '/')
219 return 0;
220 }
221 }
222 if (server_filter_list.head
223 && check_filter(&server_filter_list, fname, is_dir) < 0)
224 return 1;
225 if (filter_level != ALL_FILTERS)
226 return 0;
227 if (filter_list.head
228 && check_filter(&filter_list, fname, is_dir) < 0)
229 return 1;
230 return 0;
231}
232
233static int to_wire_mode(mode_t mode)
234{
235#ifdef SUPPORT_LINKS
236#if _S_IFLNK != 0120000
237 if (S_ISLNK(mode))
238 return (mode & ~(_S_IFMT)) | 0120000;
239#endif
240#endif
241 return mode;
242}
243
244static mode_t from_wire_mode(int mode)
245{
246#if _S_IFLNK != 0120000
247 if ((mode & (_S_IFMT)) == 0120000)
248 return (mode & ~(_S_IFMT)) | _S_IFLNK;
249#endif
250 return mode;
251}
252
253static void send_directory(int f, struct file_list *flist,
254 char *fbuf, int len);
255
256static char *flist_dir;
257static int flist_dir_len;
258
259
260/**
261 * Make sure @p flist is big enough to hold at least @p flist->count
262 * entries.
263 **/
264void flist_expand(struct file_list *flist)
265{
266 struct file_struct **new_ptr;
267
268 if (flist->count < flist->malloced)
269 return;
270
271 if (flist->malloced < FLIST_START)
272 flist->malloced = FLIST_START;
273 else if (flist->malloced >= FLIST_LINEAR)
274 flist->malloced += FLIST_LINEAR;
275 else
276 flist->malloced *= 2;
277
278 /*
279 * In case count jumped or we are starting the list
280 * with a known size just set it.
281 */
282 if (flist->malloced < flist->count)
283 flist->malloced = flist->count;
284
285 new_ptr = realloc_array(flist->files, struct file_struct *,
286 flist->malloced);
287
288 if (verbose >= 2 && flist->malloced != FLIST_START) {
289 rprintf(FCLIENT, "[%s] expand file_list to %.0f bytes, did%s move\n",
290 who_am_i(),
291 (double)sizeof flist->files[0] * flist->malloced,
292 (new_ptr == flist->files) ? " not" : "");
293 }
294
295 flist->files = new_ptr;
296
297 if (!flist->files)
298 out_of_memory("flist_expand");
299}
300
301static void send_file_entry(struct file_struct *file, int f)
302{
303 unsigned short flags;
304 static time_t modtime;
305 static mode_t mode;
306 static int64 dev;
307 static dev_t rdev;
308 static uint32 rdev_major;
309 static uid_t uid;
310 static gid_t gid;
311 static char lastname[MAXPATHLEN];
312 char fname[MAXPATHLEN];
313 int l1, l2;
314
315 if (f < 0)
316 return;
317
318 if (!file) {
319 write_byte(f, 0);
320 modtime = 0, mode = 0;
321 dev = 0, rdev = MAKEDEV(0, 0);
322 rdev_major = 0;
323 uid = 0, gid = 0;
324 *lastname = '\0';
325 return;
326 }
327
328 f_name(file, fname);
329
330 flags = file->flags & XMIT_TOP_DIR;
331
332 if (file->mode == mode)
333 flags |= XMIT_SAME_MODE;
334 else
335 mode = file->mode;
336 if ((preserve_devices && IS_DEVICE(mode))
337 || (preserve_specials && IS_SPECIAL(mode))) {
338 if (protocol_version < 28) {
339 if (file->u.rdev == rdev)
340 flags |= XMIT_SAME_RDEV_pre28;
341 else
342 rdev = file->u.rdev;
343 } else {
344 rdev = file->u.rdev;
345 if ((uint32)major(rdev) == rdev_major)
346 flags |= XMIT_SAME_RDEV_MAJOR;
347 else
348 rdev_major = major(rdev);
349 if ((uint32)minor(rdev) <= 0xFFu)
350 flags |= XMIT_RDEV_MINOR_IS_SMALL;
351 }
352 } else if (protocol_version < 28)
353 rdev = MAKEDEV(0, 0);
354 if (file->uid == uid)
355 flags |= XMIT_SAME_UID;
356 else
357 uid = file->uid;
358 if (file->gid == gid)
359 flags |= XMIT_SAME_GID;
360 else
361 gid = file->gid;
362 if (file->modtime == modtime)
363 flags |= XMIT_SAME_TIME;
364 else
365 modtime = file->modtime;
366
367#ifdef SUPPORT_HARD_LINKS
368 if (file->link_u.idev) {
369 if (file->F_DEV == dev) {
370 if (protocol_version >= 28)
371 flags |= XMIT_SAME_DEV;
372 } else
373 dev = file->F_DEV;
374 flags |= XMIT_HAS_IDEV_DATA;
375 }
376#endif
377
378 for (l1 = 0;
379 lastname[l1] && (fname[l1] == lastname[l1]) && (l1 < 255);
380 l1++) {}
381 l2 = strlen(fname+l1);
382
383 if (l1 > 0)
384 flags |= XMIT_SAME_NAME;
385 if (l2 > 255)
386 flags |= XMIT_LONG_NAME;
387
388 /* We must make sure we don't send a zero flag byte or the
389 * other end will terminate the flist transfer. Note that
390 * the use of XMIT_TOP_DIR on a non-dir has no meaning, so
391 * it's harmless way to add a bit to the first flag byte. */
392 if (protocol_version >= 28) {
393 if (!flags && !S_ISDIR(mode))
394 flags |= XMIT_TOP_DIR;
395 if ((flags & 0xFF00) || !flags) {
396 flags |= XMIT_EXTENDED_FLAGS;
397 write_byte(f, flags);
398 write_byte(f, flags >> 8);
399 } else
400 write_byte(f, flags);
401 } else {
402 if (!(flags & 0xFF))
403 flags |= S_ISDIR(mode) ? XMIT_LONG_NAME : XMIT_TOP_DIR;
404 write_byte(f, flags);
405 }
406 if (flags & XMIT_SAME_NAME)
407 write_byte(f, l1);
408 if (flags & XMIT_LONG_NAME)
409 write_int(f, l2);
410 else
411 write_byte(f, l2);
412 write_buf(f, fname + l1, l2);
413
414 write_longint(f, file->length);
415 if (!(flags & XMIT_SAME_TIME))
416 write_int(f, modtime);
417 if (!(flags & XMIT_SAME_MODE))
418 write_int(f, to_wire_mode(mode));
419 if (preserve_uid && !(flags & XMIT_SAME_UID)) {
420 if (!numeric_ids)
421 add_uid(uid);
422 write_int(f, uid);
423 }
424 if (preserve_gid && !(flags & XMIT_SAME_GID)) {
425 if (!numeric_ids)
426 add_gid(gid);
427 write_int(f, gid);
428 }
429 if ((preserve_devices && IS_DEVICE(mode))
430 || (preserve_specials && IS_SPECIAL(mode))) {
431 if (protocol_version < 28) {
432 if (!(flags & XMIT_SAME_RDEV_pre28))
433 write_int(f, (int)rdev);
434 } else {
435 if (!(flags & XMIT_SAME_RDEV_MAJOR))
436 write_int(f, major(rdev));
437 if (flags & XMIT_RDEV_MINOR_IS_SMALL)
438 write_byte(f, minor(rdev));
439 else
440 write_int(f, minor(rdev));
441 }
442 }
443
444#ifdef SUPPORT_LINKS
445 if (preserve_links && S_ISLNK(mode)) {
446 int len = strlen(file->u.link);
447 write_int(f, len);
448 write_buf(f, file->u.link, len);
449 }
450#endif
451
452#ifdef SUPPORT_HARD_LINKS
453 if (file->link_u.idev) {
454 if (protocol_version < 26) {
455 /* 32-bit dev_t and ino_t */
456 write_int(f, dev);
457 write_int(f, file->F_INODE);
458 } else {
459 /* 64-bit dev_t and ino_t */
460 if (!(flags & XMIT_SAME_DEV))
461 write_longint(f, dev);
462 write_longint(f, file->F_INODE);
463 }
464 }
465#endif
466
467 if (always_checksum && (S_ISREG(mode) || protocol_version < 28)) {
468 char *sum;
469 if (S_ISREG(mode))
470 sum = file->u.sum;
471 else {
472 /* Prior to 28, we sent a useless set of nulls. */
473 sum = empty_sum;
474 }
475 write_buf(f, sum, checksum_len);
476 }
477
478 strlcpy(lastname, fname, MAXPATHLEN);
479}
480
481static struct file_struct *receive_file_entry(struct file_list *flist,
482 unsigned short flags, int f)
483{
484 static time_t modtime;
485 static mode_t mode;
486 static int64 dev;
487 static dev_t rdev;
488 static uint32 rdev_major;
489 static uid_t uid;
490 static gid_t gid;
491 static char lastname[MAXPATHLEN], *lastdir;
492 static int lastdir_depth, lastdir_len = -1;
493 static unsigned int del_hier_name_len = 0;
494 static int in_del_hier = 0;
495 char thisname[MAXPATHLEN];
496 unsigned int l1 = 0, l2 = 0;
497 int alloc_len, basename_len, dirname_len, linkname_len, sum_len;
498 OFF_T file_length;
499 char *basename, *dirname, *bp;
500 struct file_struct *file;
501
502 if (!flist) {
503 modtime = 0, mode = 0;
504 dev = 0, rdev = MAKEDEV(0, 0);
505 rdev_major = 0;
506 uid = 0, gid = 0;
507 *lastname = '\0';
508 lastdir_len = -1;
509 in_del_hier = 0;
510 return NULL;
511 }
512
513 if (flags & XMIT_SAME_NAME)
514 l1 = read_byte(f);
515
516 if (flags & XMIT_LONG_NAME)
517 l2 = read_int(f);
518 else
519 l2 = read_byte(f);
520
521 if (l2 >= MAXPATHLEN - l1) {
522 rprintf(FERROR,
523 "overflow: flags=0x%x l1=%d l2=%d lastname=%s\n",
524 flags, l1, l2, lastname);
525 overflow_exit("receive_file_entry");
526 }
527
528 strlcpy(thisname, lastname, l1 + 1);
529 read_sbuf(f, &thisname[l1], l2);
530 thisname[l1 + l2] = 0;
531
532 strlcpy(lastname, thisname, MAXPATHLEN);
533
534 clean_fname(thisname, 0);
535
536 if (sanitize_paths)
537 sanitize_path(thisname, thisname, "", 0, NULL);
538
539 if ((basename = strrchr(thisname, '/')) != NULL) {
540 dirname_len = ++basename - thisname; /* counts future '\0' */
541 if (lastdir_len == dirname_len - 1
542 && strncmp(thisname, lastdir, lastdir_len) == 0) {
543 dirname = lastdir;
544 dirname_len = 0; /* indicates no copy is needed */
545 } else
546 dirname = thisname;
547 } else {
548 basename = thisname;
549 dirname = NULL;
550 dirname_len = 0;
551 }
552 basename_len = strlen(basename) + 1; /* count the '\0' */
553
554 file_length = read_longint(f);
555 if (!(flags & XMIT_SAME_TIME))
556 modtime = (time_t)read_int(f);
557 if (!(flags & XMIT_SAME_MODE))
558 mode = from_wire_mode(read_int(f));
559
560 if (chmod_modes && !S_ISLNK(mode))
561 mode = tweak_mode(mode, chmod_modes);
562
563 if (preserve_uid && !(flags & XMIT_SAME_UID))
564 uid = (uid_t)read_int(f);
565 if (preserve_gid && !(flags & XMIT_SAME_GID))
566 gid = (gid_t)read_int(f);
567
568 if ((preserve_devices && IS_DEVICE(mode))
569 || (preserve_specials && IS_SPECIAL(mode))) {
570 if (protocol_version < 28) {
571 if (!(flags & XMIT_SAME_RDEV_pre28))
572 rdev = (dev_t)read_int(f);
573 } else {
574 uint32 rdev_minor;
575 if (!(flags & XMIT_SAME_RDEV_MAJOR))
576 rdev_major = read_int(f);
577 if (flags & XMIT_RDEV_MINOR_IS_SMALL)
578 rdev_minor = read_byte(f);
579 else
580 rdev_minor = read_int(f);
581 rdev = MAKEDEV(rdev_major, rdev_minor);
582 }
583 } else if (protocol_version < 28)
584 rdev = MAKEDEV(0, 0);
585
586#ifdef SUPPORT_LINKS
587 if (preserve_links && S_ISLNK(mode)) {
588 linkname_len = read_int(f) + 1; /* count the '\0' */
589 if (linkname_len <= 0 || linkname_len > MAXPATHLEN) {
590 rprintf(FERROR, "overflow: linkname_len=%d\n",
591 linkname_len - 1);
592 overflow_exit("receive_file_entry");
593 }
594 }
595 else
596#endif
597 linkname_len = 0;
598
599 sum_len = always_checksum && S_ISREG(mode) ? MD4_SUM_LENGTH : 0;
600
601 alloc_len = file_struct_len + dirname_len + basename_len
602 + linkname_len + sum_len;
603 bp = pool_alloc(flist->file_pool, alloc_len, "receive_file_entry");
604
605 file = (struct file_struct *)bp;
606 memset(bp, 0, file_struct_len);
607 bp += file_struct_len;
608
609 file->modtime = modtime;
610 file->length = file_length;
611 file->mode = mode;
612 file->uid = uid;
613 file->gid = gid;
614
615 if (dirname_len) {
616 file->dirname = lastdir = bp;
617 lastdir_len = dirname_len - 1;
618 memcpy(bp, dirname, dirname_len - 1);
619 bp += dirname_len;
620 bp[-1] = '\0';
621 lastdir_depth = count_dir_elements(lastdir);
622 file->dir.depth = lastdir_depth + 1;
623 } else if (dirname) {
624 file->dirname = dirname; /* we're reusing lastname */
625 file->dir.depth = lastdir_depth + 1;
626 } else
627 file->dir.depth = 1;
628
629 if (S_ISDIR(mode)) {
630 if (basename_len == 1+1 && *basename == '.') /* +1 for '\0' */
631 file->dir.depth--;
632 if (flags & XMIT_TOP_DIR) {
633 in_del_hier = recurse;
634 del_hier_name_len = file->dir.depth == 0 ? 0 : l1 + l2;
635 if (relative_paths && del_hier_name_len > 2
636 && lastname[del_hier_name_len-1] == '.'
637 && lastname[del_hier_name_len-2] == '/')
638 del_hier_name_len -= 2;
639 file->flags |= FLAG_TOP_DIR | FLAG_DEL_HERE;
640 } else if (in_del_hier) {
641 if (!relative_paths || !del_hier_name_len
642 || (l1 >= del_hier_name_len
643 && lastname[del_hier_name_len] == '/'))
644 file->flags |= FLAG_DEL_HERE;
645 else
646 in_del_hier = 0;
647 }
648 }
649
650 file->basename = bp;
651 memcpy(bp, basename, basename_len);
652 bp += basename_len;
653
654 if ((preserve_devices && IS_DEVICE(mode))
655 || (preserve_specials && IS_SPECIAL(mode)))
656 file->u.rdev = rdev;
657
658#ifdef SUPPORT_LINKS
659 if (linkname_len) {
660 file->u.link = bp;
661 read_sbuf(f, bp, linkname_len - 1);
662 if (sanitize_paths)
663 sanitize_path(bp, bp, "", lastdir_depth, NULL);
664 bp += linkname_len;
665 }
666#endif
667
668#ifdef SUPPORT_HARD_LINKS
669 if (preserve_hard_links && protocol_version < 28 && S_ISREG(mode))
670 flags |= XMIT_HAS_IDEV_DATA;
671 if (flags & XMIT_HAS_IDEV_DATA) {
672 int64 inode;
673 if (protocol_version < 26) {
674 dev = read_int(f);
675 inode = read_int(f);
676 } else {
677 if (!(flags & XMIT_SAME_DEV))
678 dev = read_longint(f);
679 inode = read_longint(f);
680 }
681 if (flist->hlink_pool) {
682 file->link_u.idev = pool_talloc(flist->hlink_pool,
683 struct idev, 1, "inode_table");
684 file->F_INODE = inode;
685 file->F_DEV = dev;
686 }
687 }
688#endif
689
690 if (always_checksum && (sum_len || protocol_version < 28)) {
691 char *sum;
692 if (sum_len) {
693 file->u.sum = sum = bp;
694 /*bp += sum_len;*/
695 } else {
696 /* Prior to 28, we get a useless set of nulls. */
697 sum = empty_sum;
698 }
699 read_buf(f, sum, checksum_len);
700 }
701
702 return file;
703}
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 file_list *flist,
721 STRUCT_STAT *stp, unsigned short flags,
722 int filter_level)
723{
724 static char *lastdir;
725 static int lastdir_len = -1;
726 struct file_struct *file;
727 STRUCT_STAT st;
728 char sum[SUM_LENGTH];
729 char thisname[MAXPATHLEN];
730 char linkname[MAXPATHLEN];
731 int alloc_len, basename_len, dirname_len, linkname_len, sum_len;
732 char *basename, *dirname, *bp;
733
734 if (!flist || !flist->count) /* Ignore lastdir when invalid. */
735 lastdir_len = -1;
736
737 if (strlcpy(thisname, fname, sizeof thisname)
738 >= sizeof thisname - flist_dir_len) {
739 rprintf(FINFO, "skipping overly long name: %s\n", fname);
740 return NULL;
741 }
742 clean_fname(thisname, 0);
743 if (sanitize_paths)
744 sanitize_path(thisname, thisname, "", 0, NULL);
745
746 memset(sum, 0, SUM_LENGTH);
747
748 if (stp && S_ISDIR(stp->st_mode)) {
749 st = *stp; /* Needed for "symlink/." with --relative. */
750 *linkname = '\0'; /* make IBM code checker happy */
751 } else if (readlink_stat(thisname, &st, linkname) != 0) {
752 int save_errno = errno;
753 /* See if file is excluded before reporting an error. */
754 if (filter_level != NO_FILTERS
755 && is_excluded(thisname, 0, filter_level))
756 return NULL;
757 if (save_errno == ENOENT) {
758#ifdef SUPPORT_LINKS
759 /* Avoid "vanished" error if symlink points nowhere. */
760 if (copy_links && do_lstat(thisname, &st) == 0
761 && S_ISLNK(st.st_mode)) {
762 io_error |= IOERR_GENERAL;
763 rprintf(FERROR, "symlink has no referent: %s\n",
764 full_fname(thisname));
765 } else
766#endif
767 {
768 enum logcode c = am_daemon && protocol_version < 28
769 ? FERROR : FINFO;
770 io_error |= IOERR_VANISHED;
771 rprintf(c, "file has vanished: %s\n",
772 full_fname(thisname));
773 }
774 } else {
775 io_error |= IOERR_GENERAL;
776 rsyserr(FERROR, save_errno, "readlink %s failed",
777 full_fname(thisname));
778 }
779 return NULL;
780 }
781
782 /* backup.c calls us with filter_level set to NO_FILTERS. */
783 if (filter_level == NO_FILTERS)
784 goto skip_filters;
785
786 if (S_ISDIR(st.st_mode) && !xfer_dirs) {
787 rprintf(FINFO, "skipping directory %s\n", thisname);
788 return NULL;
789 }
790
791 /* We only care about directories because we need to avoid recursing
792 * into a mount-point directory, not to avoid copying a symlinked
793 * file if -L (or similar) was specified. */
794 if (one_file_system && st.st_dev != filesystem_dev
795 && S_ISDIR(st.st_mode)) {
796 if (one_file_system > 1) {
797 if (verbose > 2) {
798 rprintf(FINFO, "skipping mount-point dir %s\n",
799 thisname);
800 }
801 return NULL;
802 }
803 flags |= FLAG_MOUNT_POINT;
804 }
805
806 if (is_excluded(thisname, S_ISDIR(st.st_mode) != 0, filter_level))
807 return NULL;
808
809 if (lp_ignore_nonreadable(module_id)) {
810#ifdef SUPPORT_LINKS
811 if (!S_ISLNK(st.st_mode))
812#endif
813 if (access(thisname, R_OK) != 0)
814 return NULL;
815 }
816
817 skip_filters:
818
819 if (verbose > 2) {
820 rprintf(FINFO, "[%s] make_file(%s,*,%d)\n",
821 who_am_i(), thisname, filter_level);
822 }
823
824 if ((basename = strrchr(thisname, '/')) != NULL) {
825 dirname_len = ++basename - thisname; /* counts future '\0' */
826 if (lastdir_len == dirname_len - 1
827 && strncmp(thisname, lastdir, lastdir_len) == 0) {
828 dirname = lastdir;
829 dirname_len = 0; /* indicates no copy is needed */
830 } else
831 dirname = thisname;
832 } else {
833 basename = thisname;
834 dirname = NULL;
835 dirname_len = 0;
836 }
837 basename_len = strlen(basename) + 1; /* count the '\0' */
838
839#ifdef SUPPORT_LINKS
840 linkname_len = S_ISLNK(st.st_mode) ? strlen(linkname) + 1 : 0;
841#else
842 linkname_len = 0;
843#endif
844
845 sum_len = always_checksum && am_sender && S_ISREG(st.st_mode)
846 ? MD4_SUM_LENGTH : 0;
847
848 alloc_len = file_struct_len + dirname_len + basename_len
849 + linkname_len + sum_len;
850 if (flist)
851 bp = pool_alloc(flist->file_pool, alloc_len, "make_file");
852 else {
853 if (!(bp = new_array(char, alloc_len)))
854 out_of_memory("make_file");
855 }
856
857 file = (struct file_struct *)bp;
858 memset(bp, 0, file_struct_len);
859 bp += file_struct_len;
860
861 file->flags = flags;
862 file->modtime = st.st_mtime;
863 file->length = st.st_size;
864 file->mode = st.st_mode;
865 file->uid = st.st_uid;
866 file->gid = st.st_gid;
867
868#ifdef SUPPORT_HARD_LINKS
869 if (flist && flist->hlink_pool) {
870 if (protocol_version < 28) {
871 if (S_ISREG(st.st_mode))
872 file->link_u.idev = pool_talloc(
873 flist->hlink_pool, struct idev, 1,
874 "inode_table");
875 } else {
876 if (!S_ISDIR(st.st_mode) && st.st_nlink > 1)
877 file->link_u.idev = pool_talloc(
878 flist->hlink_pool, struct idev, 1,
879 "inode_table");
880 }
881 }
882 if (file->link_u.idev) {
883 file->F_DEV = st.st_dev;
884 file->F_INODE = st.st_ino;
885 }
886#endif
887
888 if (dirname_len) {
889 file->dirname = lastdir = bp;
890 lastdir_len = dirname_len - 1;
891 memcpy(bp, dirname, dirname_len - 1);
892 bp += dirname_len;
893 bp[-1] = '\0';
894 } else if (dirname)
895 file->dirname = dirname;
896
897 file->basename = bp;
898 memcpy(bp, basename, basename_len);
899 bp += basename_len;
900
901#ifdef HAVE_STRUCT_STAT_ST_RDEV
902 if ((preserve_devices && IS_DEVICE(st.st_mode))
903 || (preserve_specials && IS_SPECIAL(st.st_mode)))
904 file->u.rdev = st.st_rdev;
905#endif
906
907#ifdef SUPPORT_LINKS
908 if (linkname_len) {
909 file->u.link = bp;
910 memcpy(bp, linkname, linkname_len);
911 bp += linkname_len;
912 }
913#endif
914
915 if (sum_len) {
916 file->u.sum = bp;
917 file_checksum(thisname, bp, st.st_size);
918 /*bp += sum_len;*/
919 }
920
921 file->dir.root = flist_dir;
922
923 /* This code is only used by the receiver when it is building
924 * a list of files for a delete pass. */
925 if (keep_dirlinks && linkname_len && flist) {
926 STRUCT_STAT st2;
927 int save_mode = file->mode;
928 file->mode = S_IFDIR; /* Find a directory with our name. */
929 if (flist_find(the_file_list, file) >= 0
930 && do_stat(thisname, &st2) == 0 && S_ISDIR(st2.st_mode)) {
931 file->modtime = st2.st_mtime;
932 file->length = st2.st_size;
933 file->mode = st2.st_mode;
934 file->uid = st2.st_uid;
935 file->gid = st2.st_gid;
936 file->u.link = NULL;
937 } else
938 file->mode = save_mode;
939 }
940
941 if (S_ISREG(st.st_mode) || S_ISLNK(st.st_mode))
942 stats.total_size += st.st_size;
943
944 return file;
945}
946
947static struct file_struct *send_file_name(int f, struct file_list *flist,
948 char *fname, STRUCT_STAT *stp,
949 unsigned short flags)
950{
951 struct file_struct *file;
952
953 file = make_file(fname, flist, stp, flags,
954 f == -2 ? SERVER_FILTERS : ALL_FILTERS);
955 if (!file)
956 return NULL;
957
958 if (chmod_modes && !S_ISLNK(file->mode))
959 file->mode = tweak_mode(file->mode, chmod_modes);
960
961 maybe_emit_filelist_progress(flist->count + flist_count_offset);
962
963 flist_expand(flist);
964
965 if (file->basename[0]) {
966 flist->files[flist->count++] = file;
967 send_file_entry(file, f);
968 }
969 return file;
970}
971
972static void send_if_directory(int f, struct file_list *flist,
973 struct file_struct *file,
974 char *fbuf, unsigned int ol)
975{
976 char is_dot_dir = fbuf[ol-1] == '.' && (ol == 1 || fbuf[ol-2] == '/');
977
978 if (S_ISDIR(file->mode)
979 && !(file->flags & FLAG_MOUNT_POINT) && f_name(file, fbuf)) {
980 void *save_filters;
981 unsigned int len = strlen(fbuf);
982 if (len > 1 && fbuf[len-1] == '/')
983 fbuf[--len] = '\0';
984 if (len >= MAXPATHLEN - 1) {
985 io_error |= IOERR_GENERAL;
986 rprintf(FERROR, "skipping long-named directory: %s\n",
987 full_fname(fbuf));
988 return;
989 }
990 save_filters = push_local_filters(fbuf, len);
991 send_directory(f, flist, fbuf, len);
992 pop_local_filters(save_filters);
993 fbuf[ol] = '\0';
994 if (is_dot_dir)
995 fbuf[ol-1] = '.';
996 }
997}
998
999/* This function is normally called by the sender, but the receiving side also
1000 * calls it from get_dirlist() with f set to -1 so that we just construct the
1001 * file list in memory without sending it over the wire. Also, get_dirlist()
1002 * might call this with f set to -2, which also indicates that local filter
1003 * rules should be ignored. */
1004static void send_directory(int f, struct file_list *flist,
1005 char *fbuf, int len)
1006{
1007 struct dirent *di;
1008 unsigned remainder;
1009 char *p;
1010 DIR *d;
1011 int start = flist->count;
1012
1013 if (!(d = opendir(fbuf))) {
1014 io_error |= IOERR_GENERAL;
1015 rsyserr(FERROR, errno, "opendir %s failed", full_fname(fbuf));
1016 return;
1017 }
1018
1019 p = fbuf + len;
1020 if (len != 1 || *fbuf != '/')
1021 *p++ = '/';
1022 *p = '\0';
1023 remainder = MAXPATHLEN - (p - fbuf);
1024
1025 for (errno = 0, di = readdir(d); di; errno = 0, di = readdir(d)) {
1026 char *dname = d_name(di);
1027 if (dname[0] == '.' && (dname[1] == '\0'
1028 || (dname[1] == '.' && dname[2] == '\0')))
1029 continue;
1030 if (strlcpy(p, dname, remainder) >= remainder) {
1031 io_error |= IOERR_GENERAL;
1032 rprintf(FINFO,
1033 "cannot send long-named file %s\n",
1034 full_fname(fbuf));
1035 continue;
1036 }
1037
1038 send_file_name(f, flist, fbuf, NULL, 0);
1039 }
1040
1041 fbuf[len] = '\0';
1042
1043 if (errno) {
1044 io_error |= IOERR_GENERAL;
1045 rsyserr(FERROR, errno, "readdir(%s)", full_fname(fbuf));
1046 }
1047
1048 closedir(d);
1049
1050 if (recurse) {
1051 int i, end = flist->count - 1;
1052 for (i = start; i <= end; i++)
1053 send_if_directory(f, flist, flist->files[i], fbuf, len);
1054 }
1055}
1056
1057struct file_list *send_file_list(int f, int argc, char *argv[])
1058{
1059 int len;
1060 STRUCT_STAT st;
1061 char *p, *dir, olddir[sizeof curr_dir];
1062 char lastpath[MAXPATHLEN] = "";
1063 struct file_list *flist;
1064 struct timeval start_tv, end_tv;
1065 int64 start_write;
1066 int use_ff_fd = 0;
1067
1068 rprintf(FLOG, "building file list\n");
1069 if (show_filelist_p())
1070 start_filelist_progress("building file list");
1071
1072 start_write = stats.total_written;
1073 gettimeofday(&start_tv, NULL);
1074
1075 flist = flist_new(WITH_HLINK, "send_file_list");
1076
1077 io_start_buffering_out();
1078 if (filesfrom_fd >= 0) {
1079 if (argv[0] && !push_dir(argv[0], 0)) {
1080 rsyserr(FERROR, errno, "push_dir %s failed",
1081 full_fname(argv[0]));
1082 exit_cleanup(RERR_FILESELECT);
1083 }
1084 use_ff_fd = 1;
1085 }
1086
1087 while (1) {
1088 char fbuf[MAXPATHLEN];
1089 char *fn;
1090 int is_dot_dir;
1091
1092 if (use_ff_fd) {
1093 if (read_filesfrom_line(filesfrom_fd, fbuf) == 0)
1094 break;
1095 sanitize_path(fbuf, fbuf, "", 0, NULL);
1096 } else {
1097 if (argc-- == 0)
1098 break;
1099 strlcpy(fbuf, *argv++, MAXPATHLEN);
1100 if (sanitize_paths)
1101 sanitize_path(fbuf, fbuf, "", 0, NULL);
1102 }
1103
1104 len = strlen(fbuf);
1105 if (relative_paths) {
1106 /* We clean up fbuf below. */
1107 is_dot_dir = 0;
1108 } else if (!len || fbuf[len - 1] == '/') {
1109 if (len == 2 && fbuf[0] == '.') {
1110 /* Turn "./" into just "." rather than "./." */
1111 fbuf[1] = '\0';
1112 } else {
1113 if (len + 1 >= MAXPATHLEN)
1114 overflow_exit("send_file_list");
1115 fbuf[len++] = '.';
1116 fbuf[len] = '\0';
1117 }
1118 is_dot_dir = 1;
1119 } else if (len > 1 && fbuf[len-1] == '.' && fbuf[len-2] == '.'
1120 && (len == 2 || fbuf[len-3] == '/')) {
1121 if (len + 2 >= MAXPATHLEN)
1122 overflow_exit("send_file_list");
1123 fbuf[len++] = '/';
1124 fbuf[len++] = '.';
1125 fbuf[len] = '\0';
1126 is_dot_dir = 1;
1127 } else {
1128 is_dot_dir = fbuf[len-1] == '.'
1129 && (len == 1 || fbuf[len-2] == '/');
1130 }
1131
1132 if (link_stat(fbuf, &st, copy_dirlinks) != 0) {
1133 io_error |= IOERR_GENERAL;
1134 rsyserr(FERROR, errno, "link_stat %s failed",
1135 full_fname(fbuf));
1136 continue;
1137 }
1138
1139 if (S_ISDIR(st.st_mode) && !xfer_dirs) {
1140 rprintf(FINFO, "skipping directory %s\n", fbuf);
1141 continue;
1142 }
1143
1144 dir = NULL;
1145 olddir[0] = '\0';
1146
1147 if (!relative_paths) {
1148 p = strrchr(fbuf, '/');
1149 if (p) {
1150 *p = '\0';
1151 if (p == fbuf)
1152 dir = "/";
1153 else
1154 dir = fbuf;
1155 len -= p - fbuf + 1;
1156 fn = p + 1;
1157 } else
1158 fn = fbuf;
1159 } else {
1160 if ((p = strstr(fbuf, "/./")) != NULL) {
1161 *p = '\0';
1162 if (p == fbuf)
1163 dir = "/";
1164 else
1165 dir = fbuf;
1166 len -= p - fbuf + 3;
1167 fn = p + 3;
1168 } else
1169 fn = fbuf;
1170 /* Get rid of trailing "/" and "/.". */
1171 while (len) {
1172 if (fn[len - 1] == '/') {
1173 is_dot_dir = 1;
1174 if (!--len && !dir) {
1175 len++;
1176 break;
1177 }
1178 }
1179 else if (len >= 2 && fn[len - 1] == '.'
1180 && fn[len - 2] == '/') {
1181 is_dot_dir = 1;
1182 if (!(len -= 2) && !dir) {
1183 len++;
1184 break;
1185 }
1186 } else
1187 break;
1188 }
1189 if (len == 1 && fn[0] == '/')
1190 fn[len++] = '.';
1191 fn[len] = '\0';
1192 /* Reject a ".." dir in the active part of the path. */
1193 for (p = fn; (p = strstr(p, "..")) != NULL; p += 2) {
1194 if ((p[2] == '/' || p[2] == '\0')
1195 && (p == fn || p[-1] == '/')) {
1196 rprintf(FERROR,
1197 "found \"..\" dir in relative path: %s\n",
1198 fbuf);
1199 exit_cleanup(RERR_SYNTAX);
1200 }
1201 }
1202 }
1203
1204 if (!*fn) {
1205 len = 1;
1206 fn = ".";
1207 }
1208
1209 if (dir && *dir) {
1210 static char *lastdir;
1211 static int lastdir_len;
1212
1213 strlcpy(olddir, curr_dir, sizeof olddir);
1214
1215 if (!push_dir(dir, 0)) {
1216 io_error |= IOERR_GENERAL;
1217 rsyserr(FERROR, errno, "push_dir %s failed",
1218 full_fname(dir));
1219 continue;
1220 }
1221
1222 if (lastdir && strcmp(lastdir, dir) == 0) {
1223 flist_dir = lastdir;
1224 flist_dir_len = lastdir_len;
1225 } else {
1226 flist_dir = lastdir = strdup(dir);
1227 flist_dir_len = lastdir_len = strlen(dir);
1228 }
1229 }
1230
1231 if (fn != fbuf)
1232 memmove(fbuf, fn, len + 1);
1233
1234 if (implied_dirs && (p=strrchr(fbuf,'/')) && p != fbuf) {
1235 /* Send the implied directories at the start of the
1236 * source spec, so we get their permissions right. */
1237 char *lp = lastpath, *slash = fbuf;
1238 *p = '\0';
1239 /* Skip any initial directories in our path that we
1240 * have in common with lastpath. */
1241 for (fn = fbuf; *fn && *lp == *fn; lp++, fn++) {
1242 if (*fn == '/')
1243 slash = fn;
1244 }
1245 *p = '/';
1246 if (fn != p || (*lp && *lp != '/')) {
1247 int save_copy_links = copy_links;
1248 int save_xfer_dirs = xfer_dirs;
1249 copy_links |= copy_unsafe_links;
1250 xfer_dirs = 1;
1251 while ((slash = strchr(slash+1, '/')) != 0) {
1252 *slash = '\0';
1253 send_file_name(f, flist, fbuf, NULL, 0);
1254 *slash = '/';
1255 }
1256 copy_links = save_copy_links;
1257 xfer_dirs = save_xfer_dirs;
1258 *p = '\0';
1259 strlcpy(lastpath, fbuf, sizeof lastpath);
1260 *p = '/';
1261 }
1262 }
1263
1264 if (one_file_system)
1265 filesystem_dev = st.st_dev;
1266
1267 if (recurse || (xfer_dirs && is_dot_dir)) {
1268 struct file_struct *file;
1269 file = send_file_name(f, flist, fbuf, &st, FLAG_TOP_DIR);
1270 if (file)
1271 send_if_directory(f, flist, file, fbuf, len);
1272 } else
1273 send_file_name(f, flist, fbuf, &st, 0);
1274
1275 if (olddir[0]) {
1276 flist_dir = NULL;
1277 flist_dir_len = 0;
1278 if (!pop_dir(olddir)) {
1279 rsyserr(FERROR, errno, "pop_dir %s failed",
1280 full_fname(olddir));
1281 exit_cleanup(RERR_FILESELECT);
1282 }
1283 }
1284 }
1285
1286 gettimeofday(&end_tv, NULL);
1287 stats.flist_buildtime = (int64)(end_tv.tv_sec - start_tv.tv_sec) * 1000
1288 + (end_tv.tv_usec - start_tv.tv_usec) / 1000;
1289 if (stats.flist_buildtime == 0)
1290 stats.flist_buildtime = 1;
1291 start_tv = end_tv;
1292
1293 send_file_entry(NULL, f);
1294
1295 if (show_filelist_p())
1296 finish_filelist_progress(flist);
1297
1298 gettimeofday(&end_tv, NULL);
1299 stats.flist_xfertime = (int64)(end_tv.tv_sec - start_tv.tv_sec) * 1000
1300 + (end_tv.tv_usec - start_tv.tv_usec) / 1000;
1301
1302 if (flist->hlink_pool) {
1303 pool_destroy(flist->hlink_pool);
1304 flist->hlink_pool = NULL;
1305 }
1306
1307 /* Sort the list without removing any duplicates. This allows the
1308 * receiving side to ask for any name they like, which gives us the
1309 * flexibility to change the way we unduplicate names in the future
1310 * without causing a compatibility problem with older versions. */
1311 clean_flist(flist, 0, 0);
1312
1313 send_uid_list(f);
1314
1315 /* send the io_error flag */
1316 write_int(f, lp_ignore_errors(module_id) ? 0 : io_error);
1317
1318 io_end_buffering();
1319 stats.flist_size = stats.total_written - start_write;
1320 stats.num_files = flist->count;
1321
1322 if (verbose > 3)
1323 output_flist(flist);
1324
1325 if (verbose > 2)
1326 rprintf(FINFO, "send_file_list done\n");
1327
1328 return flist;
1329}
1330
1331struct file_list *recv_file_list(int f)
1332{
1333 struct file_list *flist;
1334 unsigned short flags;
1335 int64 start_read;
1336
1337 rprintf(FLOG, "receiving file list\n");
1338 if (show_filelist_p())
1339 start_filelist_progress("receiving file list");
1340
1341 start_read = stats.total_read;
1342
1343 flist = flist_new(WITH_HLINK, "recv_file_list");
1344
1345 flist->count = 0;
1346 flist->malloced = 1000;
1347 flist->files = new_array(struct file_struct *, flist->malloced);
1348 if (!flist->files)
1349 goto oom;
1350
1351 while ((flags = read_byte(f)) != 0) {
1352 struct file_struct *file;
1353
1354 flist_expand(flist);
1355
1356 if (protocol_version >= 28 && (flags & XMIT_EXTENDED_FLAGS))
1357 flags |= read_byte(f) << 8;
1358 file = receive_file_entry(flist, flags, f);
1359
1360 if (S_ISREG(file->mode) || S_ISLNK(file->mode))
1361 stats.total_size += file->length;
1362
1363 flist->files[flist->count++] = file;
1364
1365 maybe_emit_filelist_progress(flist->count);
1366
1367 if (verbose > 2) {
1368 rprintf(FINFO, "recv_file_name(%s)\n",
1369 f_name(file, NULL));
1370 }
1371 }
1372 receive_file_entry(NULL, 0, 0); /* Signal that we're done. */
1373
1374 if (verbose > 2)
1375 rprintf(FINFO, "received %d names\n", flist->count);
1376
1377 if (show_filelist_p())
1378 finish_filelist_progress(flist);
1379
1380 clean_flist(flist, relative_paths, 1);
1381
1382 if (f >= 0) {
1383 recv_uid_list(f, flist);
1384
1385 /* Recv the io_error flag */
1386 if (lp_ignore_errors(module_id) || ignore_errors)
1387 read_int(f);
1388 else
1389 io_error |= read_int(f);
1390 }
1391
1392 if (verbose > 3)
1393 output_flist(flist);
1394
1395 if (list_only) {
1396 int i;
1397 for (i = 0; i < flist->count; i++)
1398 list_file_entry(flist->files[i]);
1399 }
1400
1401 if (verbose > 2)
1402 rprintf(FINFO, "recv_file_list done\n");
1403
1404 stats.flist_size = stats.total_read - start_read;
1405 stats.num_files = flist->count;
1406
1407 return flist;
1408
1409 oom:
1410 out_of_memory("recv_file_list");
1411 return NULL; /* not reached */
1412}
1413
1414static int file_compare(struct file_struct **file1, struct file_struct **file2)
1415{
1416 return f_name_cmp(*file1, *file2);
1417}
1418
1419/* Search for an identically-named item in the file list. Note that the
1420 * items must agree in their directory-ness, or no match is returned. */
1421int flist_find(struct file_list *flist, struct file_struct *f)
1422{
1423 int low = flist->low, high = flist->high;
1424 int diff, mid, mid_up;
1425
1426 while (low <= high) {
1427 mid = (low + high) / 2;
1428 if (flist->files[mid]->basename)
1429 mid_up = mid;
1430 else {
1431 /* Scan for the next non-empty entry using the cached
1432 * distance values. If the value isn't fully up-to-
1433 * date, update it. */
1434 mid_up = mid + flist->files[mid]->dir.depth;
1435 if (!flist->files[mid_up]->basename) {
1436 do {
1437 mid_up += flist->files[mid_up]->dir.depth;
1438 } while (!flist->files[mid_up]->basename);
1439 flist->files[mid]->dir.depth = mid_up - mid;
1440 }
1441 if (mid_up > high) {
1442 /* If there's nothing left above us, set high to
1443 * a non-empty entry below us and continue. */
1444 high = mid - flist->files[mid]->length;
1445 if (!flist->files[high]->basename) {
1446 do {
1447 high -= flist->files[high]->length;
1448 } while (!flist->files[high]->basename);
1449 flist->files[mid]->length = mid - high;
1450 }
1451 continue;
1452 }
1453 }
1454 diff = f_name_cmp(flist->files[mid_up], f);
1455 if (diff == 0) {
1456 if (protocol_version < 29
1457 && S_ISDIR(flist->files[mid_up]->mode)
1458 != S_ISDIR(f->mode))
1459 return -1;
1460 return mid_up;
1461 }
1462 if (diff < 0)
1463 low = mid_up + 1;
1464 else
1465 high = mid - 1;
1466 }
1467 return -1;
1468}
1469
1470/*
1471 * Free up any resources a file_struct has allocated
1472 * and clear the file.
1473 */
1474void clear_file(struct file_struct *file, struct file_list *flist)
1475{
1476 if (flist->hlink_pool && file->link_u.idev)
1477 pool_free(flist->hlink_pool, 0, file->link_u.idev);
1478 memset(file, 0, file_struct_len);
1479 /* In an empty entry, dir.depth is an offset to the next non-empty
1480 * entry. Likewise for length in the opposite direction. We assume
1481 * that we're alone for now since flist_find() will adjust the counts
1482 * it runs into that aren't up-to-date. */
1483 file->length = file->dir.depth = 1;
1484}
1485
1486/*
1487 * allocate a new file list
1488 */
1489struct file_list *flist_new(int with_hlink, char *msg)
1490{
1491 struct file_list *flist;
1492
1493 flist = new(struct file_list);
1494 if (!flist)
1495 out_of_memory(msg);
1496
1497 memset(flist, 0, sizeof (struct file_list));
1498
1499 if (!(flist->file_pool = pool_create(FILE_EXTENT, 0,
1500 out_of_memory, POOL_INTERN)))
1501 out_of_memory(msg);
1502
1503#ifdef SUPPORT_HARD_LINKS
1504 if (with_hlink && preserve_hard_links) {
1505 if (!(flist->hlink_pool = pool_create(HLINK_EXTENT,
1506 sizeof (struct idev), out_of_memory, POOL_INTERN)))
1507 out_of_memory(msg);
1508 }
1509#endif
1510
1511 return flist;
1512}
1513
1514/*
1515 * free up all elements in a flist
1516 */
1517void flist_free(struct file_list *flist)
1518{
1519 pool_destroy(flist->file_pool);
1520 pool_destroy(flist->hlink_pool);
1521 free(flist->files);
1522 free(flist);
1523}
1524
1525/*
1526 * This routine ensures we don't have any duplicate names in our file list.
1527 * duplicate names can cause corruption because of the pipelining
1528 */
1529static void clean_flist(struct file_list *flist, int strip_root, int no_dups)
1530{
1531 char fbuf[MAXPATHLEN];
1532 int i, prev_i = 0;
1533
1534 if (!flist)
1535 return;
1536 if (flist->count == 0) {
1537 flist->high = -1;
1538 return;
1539 }
1540
1541 qsort(flist->files, flist->count,
1542 sizeof flist->files[0], (int (*)())file_compare);
1543
1544 for (i = no_dups? 0 : flist->count; i < flist->count; i++) {
1545 if (flist->files[i]->basename) {
1546 prev_i = i;
1547 break;
1548 }
1549 }
1550 flist->low = prev_i;
1551 while (++i < flist->count) {
1552 int j;
1553 struct file_struct *file = flist->files[i];
1554
1555 if (!file->basename)
1556 continue;
1557 if (f_name_cmp(file, flist->files[prev_i]) == 0)
1558 j = prev_i;
1559 else if (protocol_version >= 29 && S_ISDIR(file->mode)) {
1560 int save_mode = file->mode;
1561 /* Make sure that this directory doesn't duplicate a
1562 * non-directory earlier in the list. */
1563 flist->high = prev_i;
1564 file->mode = S_IFREG;
1565 j = flist_find(flist, file);
1566 file->mode = save_mode;
1567 } else
1568 j = -1;
1569 if (j >= 0) {
1570 struct file_struct *fp = flist->files[j];
1571 int keep, drop;
1572 /* If one is a dir and the other is not, we want to
1573 * keep the dir because it might have contents in the
1574 * list. */
1575 if (S_ISDIR(file->mode) != S_ISDIR(fp->mode)) {
1576 if (S_ISDIR(file->mode))
1577 keep = i, drop = j;
1578 else
1579 keep = j, drop = i;
1580 } else
1581 keep = j, drop = i;
1582 if (verbose > 1 && !am_server) {
1583 rprintf(FINFO,
1584 "removing duplicate name %s from file list (%d)\n",
1585 f_name(file, fbuf), drop);
1586 }
1587 /* Make sure we don't lose track of a user-specified
1588 * top directory. */
1589 flist->files[keep]->flags |= flist->files[drop]->flags
1590 & (FLAG_TOP_DIR|FLAG_DEL_HERE);
1591
1592 clear_file(flist->files[drop], flist);
1593
1594 if (keep == i) {
1595 if (flist->low == drop) {
1596 for (j = drop + 1;
1597 j < i && !flist->files[j]->basename;
1598 j++) {}
1599 flist->low = j;
1600 }
1601 prev_i = i;
1602 }
1603 } else
1604 prev_i = i;
1605 }
1606 flist->high = no_dups ? prev_i : flist->count - 1;
1607
1608 if (strip_root) {
1609 /* We need to strip off the leading slashes for relative
1610 * paths, but this must be done _after_ the sorting phase. */
1611 for (i = flist->low; i <= flist->high; i++) {
1612 struct file_struct *file = flist->files[i];
1613
1614 if (!file->dirname)
1615 continue;
1616 while (*file->dirname == '/')
1617 file->dirname++;
1618 if (!*file->dirname)
1619 file->dirname = NULL;
1620 }
1621 }
1622
1623 if (prune_empty_dirs && no_dups) {
1624 int j, prev_depth = 0;
1625
1626 prev_i = 0; /* It's OK that this isn't really true. */
1627
1628 for (i = flist->low; i <= flist->high; i++) {
1629 struct file_struct *fp, *file = flist->files[i];
1630
1631 /* This temporarily abuses the dir.depth value for a
1632 * directory that is in a chain that might get pruned.
1633 * We restore the old value if it gets a reprieve. */
1634 if (S_ISDIR(file->mode) && file->dir.depth) {
1635 /* Dump empty dirs when coming back down. */
1636 for (j = prev_depth; j >= file->dir.depth; j--) {
1637 fp = flist->files[prev_i];
1638 if (fp->dir.depth >= 0)
1639 break;
1640 prev_i = -fp->dir.depth-1;
1641 clear_file(fp, flist);
1642 }
1643 prev_depth = file->dir.depth;
1644 if (is_excluded(f_name(file, fbuf), 1,
1645 ALL_FILTERS)) {
1646 /* Keep dirs through this dir. */
1647 for (j = prev_depth-1; ; j--) {
1648 fp = flist->files[prev_i];
1649 if (fp->dir.depth >= 0)
1650 break;
1651 prev_i = -fp->dir.depth-1;
1652 fp->dir.depth = j;
1653 }
1654 } else
1655 file->dir.depth = -prev_i-1;
1656 prev_i = i;
1657 } else {
1658 /* Keep dirs through this non-dir. */
1659 for (j = prev_depth; ; j--) {
1660 fp = flist->files[prev_i];
1661 if (fp->dir.depth >= 0)
1662 break;
1663 prev_i = -fp->dir.depth-1;
1664 fp->dir.depth = j;
1665 }
1666 }
1667 }
1668 /* Dump empty all remaining empty dirs. */
1669 while (1) {
1670 struct file_struct *fp = flist->files[prev_i];
1671 if (fp->dir.depth >= 0)
1672 break;
1673 prev_i = -fp->dir.depth-1;
1674 clear_file(fp, flist);
1675 }
1676
1677 for (i = flist->low; i <= flist->high; i++) {
1678 if (flist->files[i]->basename)
1679 break;
1680 }
1681 flist->low = i;
1682 for (i = flist->high; i >= flist->low; i--) {
1683 if (flist->files[i]->basename)
1684 break;
1685 }
1686 flist->high = i;
1687 }
1688}
1689
1690static void output_flist(struct file_list *flist)
1691{
1692 char uidbuf[16], gidbuf[16], depthbuf[16];
1693 struct file_struct *file;
1694 const char *who = who_am_i();
1695 int i;
1696
1697 for (i = 0; i < flist->count; i++) {
1698 file = flist->files[i];
1699 if ((am_root || am_sender) && preserve_uid)
1700 snprintf(uidbuf, sizeof uidbuf, " uid=%ld", (long)file->uid);
1701 else
1702 *uidbuf = '\0';
1703 if (preserve_gid && file->gid != GID_NONE)
1704 snprintf(gidbuf, sizeof gidbuf, " gid=%ld", (long)file->gid);
1705 else
1706 *gidbuf = '\0';
1707 if (!am_sender)
1708 snprintf(depthbuf, sizeof depthbuf, "%d", file->dir.depth);
1709 rprintf(FINFO, "[%s] i=%d %s %s%s%s%s mode=0%o len=%.0f%s%s flags=%x\n",
1710 who, i, am_sender ? NS(file->dir.root) : depthbuf,
1711 file->dirname ? file->dirname : "",
1712 file->dirname ? "/" : "", NS(file->basename),
1713 S_ISDIR(file->mode) ? "/" : "", (int)file->mode,
1714 (double)file->length, uidbuf, gidbuf, file->flags);
1715 }
1716}
1717
1718enum fnc_state { s_DIR, s_SLASH, s_BASE, s_TRAILING };
1719enum fnc_type { t_PATH, t_ITEM };
1720
1721/* Compare the names of two file_struct entities, similar to how strcmp()
1722 * would do if it were operating on the joined strings.
1723 *
1724 * Some differences beginning with protocol_version 29: (1) directory names
1725 * are compared with an assumed trailing slash so that they compare in a
1726 * way that would cause them to sort immediately prior to any content they
1727 * may have; (2) a directory of any name compares after a non-directory of
1728 * any name at the same depth; (3) a directory with name "." compares prior
1729 * to anything else. These changes mean that a directory and a non-dir
1730 * with the same name will not compare as equal (protocol_version >= 29).
1731 *
1732 * The dirname component can be an empty string, but the basename component
1733 * cannot (and never is in the current codebase). The basename component
1734 * may be NULL (for a removed item), in which case it is considered to be
1735 * after any existing item. */
1736int f_name_cmp(struct file_struct *f1, struct file_struct *f2)
1737{
1738 int dif;
1739 const uchar *c1, *c2;
1740 enum fnc_state state1, state2;
1741 enum fnc_type type1, type2;
1742 enum fnc_type t_path = protocol_version >= 29 ? t_PATH : t_ITEM;
1743
1744 if (!f1 || !f1->basename) {
1745 if (!f2 || !f2->basename)
1746 return 0;
1747 return -1;
1748 }
1749 if (!f2 || !f2->basename)
1750 return 1;
1751
1752 c1 = (uchar*)f1->dirname;
1753 c2 = (uchar*)f2->dirname;
1754 if (c1 == c2)
1755 c1 = c2 = NULL;
1756 if (!c1) {
1757 type1 = S_ISDIR(f1->mode) ? t_path : t_ITEM;
1758 c1 = (uchar*)f1->basename;
1759 if (type1 == t_PATH && *c1 == '.' && !c1[1]) {
1760 type1 = t_ITEM;
1761 state1 = s_TRAILING;
1762 c1 = (uchar*)"";
1763 } else
1764 state1 = s_BASE;
1765 } else {
1766 type1 = t_path;
1767 state1 = s_DIR;
1768 }
1769 if (!c2) {
1770 type2 = S_ISDIR(f2->mode) ? t_path : t_ITEM;
1771 c2 = (uchar*)f2->basename;
1772 if (type2 == t_PATH && *c2 == '.' && !c2[1]) {
1773 type2 = t_ITEM;
1774 state2 = s_TRAILING;
1775 c2 = (uchar*)"";
1776 } else
1777 state2 = s_BASE;
1778 } else {
1779 type2 = t_path;
1780 state2 = s_DIR;
1781 }
1782
1783 if (type1 != type2)
1784 return type1 == t_PATH ? 1 : -1;
1785
1786 do {
1787 if (!*c1) {
1788 switch (state1) {
1789 case s_DIR:
1790 state1 = s_SLASH;
1791 c1 = (uchar*)"/";
1792 break;
1793 case s_SLASH:
1794 type1 = S_ISDIR(f1->mode) ? t_path : t_ITEM;
1795 c1 = (uchar*)f1->basename;
1796 if (type1 == t_PATH && *c1 == '.' && !c1[1]) {
1797 type1 = t_ITEM;
1798 state1 = s_TRAILING;
1799 c1 = (uchar*)"";
1800 } else
1801 state1 = s_BASE;
1802 break;
1803 case s_BASE:
1804 state1 = s_TRAILING;
1805 if (type1 == t_PATH) {
1806 c1 = (uchar*)"/";
1807 break;
1808 }
1809 /* FALL THROUGH */
1810 case s_TRAILING:
1811 type1 = t_ITEM;
1812 break;
1813 }
1814 if (*c2 && type1 != type2)
1815 return type1 == t_PATH ? 1 : -1;
1816 }
1817 if (!*c2) {
1818 switch (state2) {
1819 case s_DIR:
1820 state2 = s_SLASH;
1821 c2 = (uchar*)"/";
1822 break;
1823 case s_SLASH:
1824 type2 = S_ISDIR(f2->mode) ? t_path : t_ITEM;
1825 c2 = (uchar*)f2->basename;
1826 if (type2 == t_PATH && *c2 == '.' && !c2[1]) {
1827 type2 = t_ITEM;
1828 state2 = s_TRAILING;
1829 c2 = (uchar*)"";
1830 } else
1831 state2 = s_BASE;
1832 break;
1833 case s_BASE:
1834 state2 = s_TRAILING;
1835 if (type2 == t_PATH) {
1836 c2 = (uchar*)"/";
1837 break;
1838 }
1839 /* FALL THROUGH */
1840 case s_TRAILING:
1841 if (!*c1)
1842 return 0;
1843 type2 = t_ITEM;
1844 break;
1845 }
1846 if (type1 != type2)
1847 return type1 == t_PATH ? 1 : -1;
1848 }
1849 } while ((dif = (int)*c1++ - (int)*c2++) == 0);
1850
1851 return dif;
1852}
1853
1854/* Return a copy of the full filename of a flist entry, using the indicated
1855 * buffer or one of 5 static buffers if fbuf is NULL. No size-checking is
1856 * done because we checked the size when creating the file_struct entry.
1857 */
1858char *f_name(struct file_struct *f, char *fbuf)
1859{
1860 if (!f || !f->basename)
1861 return NULL;
1862
1863 if (!fbuf) {
1864 static char names[5][MAXPATHLEN];
1865 static unsigned int n;
1866
1867 n = (n + 1) % (sizeof names / sizeof names[0]);
1868
1869 fbuf = names[n];
1870 }
1871
1872 if (f->dirname) {
1873 int len = strlen(f->dirname);
1874 memcpy(fbuf, f->dirname, len);
1875 fbuf[len] = '/';
1876 strlcpy(fbuf + len + 1, f->basename, MAXPATHLEN - (len + 1));
1877 } else
1878 strlcpy(fbuf, f->basename, MAXPATHLEN);
1879
1880 return fbuf;
1881}
1882
1883/* Do a non-recursive scan of the named directory, possibly ignoring all
1884 * exclude rules except for the daemon's. If "dlen" is >=0, it is the length
1885 * of the dirname string, and also indicates that "dirname" is a MAXPATHLEN
1886 * buffer (the functions we call will append names onto the end, but the old
1887 * dir value will be restored on exit). */
1888struct file_list *get_dirlist(char *dirname, int dlen,
1889 int ignore_filter_rules)
1890{
1891 struct file_list *dirlist;
1892 char dirbuf[MAXPATHLEN];
1893 int save_recurse = recurse;
1894 int save_xfer_dirs = xfer_dirs;
1895
1896 if (dlen < 0) {
1897 dlen = strlcpy(dirbuf, dirname, MAXPATHLEN);
1898 if (dlen >= MAXPATHLEN)
1899 return NULL;
1900 dirname = dirbuf;
1901 }
1902
1903 dirlist = flist_new(WITHOUT_HLINK, "get_dirlist");
1904
1905 recurse = 0;
1906 xfer_dirs = 1;
1907 send_directory(ignore_filter_rules ? -2 : -1, dirlist, dirname, dlen);
1908 xfer_dirs = save_xfer_dirs;
1909 recurse = save_recurse;
1910 if (do_progress)
1911 flist_count_offset += dirlist->count;
1912
1913 clean_flist(dirlist, 0, 0);
1914
1915 if (verbose > 3)
1916 output_flist(dirlist);
1917
1918 return dirlist;
1919}