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