New support function: change_local_filter_dir().
[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#include "rounding.h"
26
27extern int verbose;
28extern int list_only;
29extern int am_root;
30extern int am_server;
31extern int am_daemon;
32extern int am_sender;
33extern int do_progress;
34extern int always_checksum;
35extern int module_id;
36extern int ignore_errors;
37extern int numeric_ids;
38extern int recurse;
39extern int xfer_dirs;
40extern int filesfrom_fd;
41extern int one_file_system;
42extern int copy_dirlinks;
43extern int keep_dirlinks;
44extern int preserve_links;
45extern int preserve_hard_links;
46extern int preserve_devices;
47extern int preserve_specials;
48extern int preserve_uid;
49extern int preserve_gid;
50extern int relative_paths;
51extern int implied_dirs;
52extern int file_extra_cnt;
53extern int ignore_perishable;
54extern int non_perishable_cnt;
55extern int prune_empty_dirs;
56extern int copy_links;
57extern int copy_unsafe_links;
58extern int protocol_version;
59extern int sanitize_paths;
60extern struct stats stats;
61extern struct file_list *the_file_list;
62
63extern char curr_dir[MAXPATHLEN];
64
65extern struct chmod_mode_struct *chmod_modes;
66
67extern struct filter_list_struct filter_list;
68extern struct filter_list_struct server_filter_list;
69
70int io_error;
71int checksum_len;
72dev_t filesystem_dev; /* used to implement -x */
73
74/* The tmp_* vars are used as a cache area by make_file() to store data
75 * that the sender doesn't need to remember in its file list. The data
76 * will survive just long enough to be used by send_file_entry(). */
77static dev_t tmp_rdev;
78#ifdef SUPPORT_HARD_LINKS
79static int64 tmp_dev, tmp_ino;
80#endif
81static char tmp_sum[MD4_SUM_LENGTH];
82
83static char empty_sum[MD4_SUM_LENGTH];
84static int flist_count_offset; /* for --delete --progress */
85
86static void clean_flist(struct file_list *flist, int strip_root, int no_dups);
87static void output_flist(struct file_list *flist);
88
89void init_flist(void)
90{
91 if (verbose > 4) {
92 rprintf(FINFO, "FILE_STRUCT_LEN=%d, EXTRA_LEN=%d\n",
93 (int)FILE_STRUCT_LEN, (int)EXTRA_LEN);
94 }
95 checksum_len = protocol_version < 21 ? 2 : MD4_SUM_LENGTH;
96}
97
98static int show_filelist_p(void)
99{
100 return verbose && xfer_dirs && !am_server;
101}
102
103static void start_filelist_progress(char *kind)
104{
105 rprintf(FCLIENT, "%s ... ", kind);
106 if (verbose > 1 || do_progress)
107 rprintf(FCLIENT, "\n");
108 rflush(FINFO);
109}
110
111static void emit_filelist_progress(int count)
112{
113 rprintf(FCLIENT, " %d files...\r", count);
114}
115
116static void maybe_emit_filelist_progress(int count)
117{
118 if (do_progress && show_filelist_p() && (count % 100) == 0)
119 emit_filelist_progress(count);
120}
121
122static void finish_filelist_progress(const struct file_list *flist)
123{
124 if (do_progress) {
125 /* This overwrites the progress line */
126 rprintf(FINFO, "%d file%sto consider\n",
127 flist->count, flist->count == 1 ? " " : "s ");
128 } else
129 rprintf(FINFO, "done\n");
130}
131
132void show_flist_stats(void)
133{
134 /* Nothing yet */
135}
136
137static void list_file_entry(struct file_struct *f)
138{
139 char permbuf[PERMSTRING_SIZE];
140 double len;
141
142 if (!F_IS_ACTIVE(f)) {
143 /* this can happen if duplicate names were removed */
144 return;
145 }
146
147 permstring(permbuf, f->mode);
148 len = F_LENGTH(f);
149
150#ifdef SUPPORT_LINKS
151 if (preserve_links && S_ISLNK(f->mode)) {
152 rprintf(FINFO, "%s %11.0f %s %s -> %s\n",
153 permbuf, len, timestring(f->modtime),
154 f_name(f, NULL), F_SYMLINK(f));
155 } else
156#endif
157 {
158 rprintf(FINFO, "%s %11.0f %s %s\n",
159 permbuf, len, timestring(f->modtime),
160 f_name(f, NULL));
161 }
162}
163
164/* Stat either a symlink or its referent, depending on the settings of
165 * copy_links, copy_unsafe_links, etc. Returns -1 on error, 0 on success.
166 *
167 * If path is the name of a symlink, then the linkbuf buffer (which must hold
168 * MAXPATHLEN chars) will be set to the symlink's target string.
169 *
170 * The stat structure pointed to by stp will contain information about the
171 * link or the referent as appropriate, if they exist. */
172static int readlink_stat(const char *path, STRUCT_STAT *stp, char *linkbuf)
173{
174#ifdef SUPPORT_LINKS
175 if (link_stat(path, stp, copy_dirlinks) < 0)
176 return -1;
177 if (S_ISLNK(stp->st_mode)) {
178 int llen = readlink(path, linkbuf, MAXPATHLEN - 1);
179 if (llen < 0)
180 return -1;
181 linkbuf[llen] = '\0';
182 if (copy_unsafe_links && unsafe_symlink(linkbuf, path)) {
183 if (verbose > 1) {
184 rprintf(FINFO,"copying unsafe symlink \"%s\" -> \"%s\"\n",
185 path, linkbuf);
186 }
187 return do_stat(path, stp);
188 }
189 }
190 return 0;
191#else
192 return do_stat(path, stp);
193#endif
194}
195
196int link_stat(const char *path, STRUCT_STAT *stp, int follow_dirlinks)
197{
198#ifdef SUPPORT_LINKS
199 if (copy_links)
200 return do_stat(path, stp);
201 if (do_lstat(path, stp) < 0)
202 return -1;
203 if (follow_dirlinks && S_ISLNK(stp->st_mode)) {
204 STRUCT_STAT st;
205 if (do_stat(path, &st) == 0 && S_ISDIR(st.st_mode))
206 *stp = st;
207 }
208 return 0;
209#else
210 return do_stat(path, stp);
211#endif
212}
213
214/* This function is used to check if a file should be included/excluded
215 * from the list of files based on its name and type etc. The value of
216 * filter_level is set to either SERVER_FILTERS or ALL_FILTERS. */
217static int is_excluded(char *fname, int is_dir, int filter_level)
218{
219#if 0 /* This currently never happens, so avoid a useless compare. */
220 if (filter_level == NO_FILTERS)
221 return 0;
222#endif
223 if (fname) {
224 /* never exclude '.', even if somebody does --exclude '*' */
225 if (fname[0] == '.' && !fname[1])
226 return 0;
227 /* Handle the -R version of the '.' dir. */
228 if (fname[0] == '/') {
229 int len = strlen(fname);
230 if (fname[len-1] == '.' && fname[len-2] == '/')
231 return 0;
232 }
233 }
234 if (server_filter_list.head
235 && check_filter(&server_filter_list, fname, is_dir) < 0)
236 return 1;
237 if (filter_level != ALL_FILTERS)
238 return 0;
239 if (filter_list.head
240 && check_filter(&filter_list, fname, is_dir) < 0)
241 return 1;
242 return 0;
243}
244
245static int to_wire_mode(mode_t mode)
246{
247#ifdef SUPPORT_LINKS
248#if _S_IFLNK != 0120000
249 if (S_ISLNK(mode))
250 return (mode & ~(_S_IFMT)) | 0120000;
251#endif
252#endif
253 return mode;
254}
255
256static mode_t from_wire_mode(int mode)
257{
258#if _S_IFLNK != 0120000
259 if ((mode & (_S_IFMT)) == 0120000)
260 return (mode & ~(_S_IFMT)) | _S_IFLNK;
261#endif
262 return mode;
263}
264
265static void send_directory(int f, struct file_list *flist,
266 char *fbuf, int len);
267
268static const char *flist_dir;
269static int flist_dir_len;
270
271
272/**
273 * Make sure @p flist is big enough to hold at least @p flist->count
274 * entries.
275 **/
276void flist_expand(struct file_list *flist)
277{
278 struct file_struct **new_ptr;
279
280 if (flist->count < flist->malloced)
281 return;
282
283 if (flist->malloced < FLIST_START)
284 flist->malloced = FLIST_START;
285 else if (flist->malloced >= FLIST_LINEAR)
286 flist->malloced += FLIST_LINEAR;
287 else
288 flist->malloced *= 2;
289
290 /*
291 * In case count jumped or we are starting the list
292 * with a known size just set it.
293 */
294 if (flist->malloced < flist->count)
295 flist->malloced = flist->count;
296
297 new_ptr = realloc_array(flist->files, struct file_struct *,
298 flist->malloced);
299
300 if (verbose >= 2 && flist->malloced != FLIST_START) {
301 rprintf(FCLIENT, "[%s] expand file_list to %.0f bytes, did%s move\n",
302 who_am_i(),
303 (double)sizeof flist->files[0] * flist->malloced,
304 (new_ptr == flist->files) ? " not" : "");
305 }
306
307 flist->files = new_ptr;
308
309 if (!flist->files)
310 out_of_memory("flist_expand");
311}
312
313static void send_file_entry(struct file_struct *file, int f, int ndx)
314{
315 static time_t modtime;
316 static mode_t mode;
317 static int64 dev;
318 static dev_t rdev;
319 static uint32 rdev_major;
320 static uid_t uid;
321 static gid_t gid;
322 static char lastname[MAXPATHLEN];
323 char fname[MAXPATHLEN];
324 int first_hlink_ndx = -1;
325 int l1, l2;
326 int flags;
327
328 f_name(file, fname);
329
330 flags = file->flags & FLAG_TOP_DIR; /* FLAG_TOP_DIR == 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 (tmp_rdev == rdev)
340 flags |= XMIT_SAME_RDEV_pre28;
341 else
342 rdev = tmp_rdev;
343 } else {
344 rdev = tmp_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 (preserve_uid) {
355 if (F_UID(file) == uid)
356 flags |= XMIT_SAME_UID;
357 else
358 uid = F_UID(file);
359 }
360 if (preserve_gid) {
361 if (F_GID(file) == gid)
362 flags |= XMIT_SAME_GID;
363 else
364 gid = F_GID(file);
365 }
366 if (file->modtime == modtime)
367 flags |= XMIT_SAME_TIME;
368 else
369 modtime = file->modtime;
370
371#ifdef SUPPORT_HARD_LINKS
372 if (tmp_dev != 0) {
373 if (protocol_version >= 30) {
374 struct idev_node *np = idev_node(tmp_dev, tmp_ino);
375 first_hlink_ndx = (int32)np->data - 1;
376 if (first_hlink_ndx < 0) {
377 np->data = (void*)(ndx + 1);
378 flags |= XMIT_HLINK_FIRST;
379 }
380 flags |= XMIT_HLINKED;
381 } else {
382 if (tmp_dev == dev) {
383 if (protocol_version >= 28)
384 flags |= XMIT_SAME_DEV_pre30;
385 } else
386 dev = tmp_dev;
387 flags |= XMIT_HLINKED;
388 }
389 }
390#endif
391
392 for (l1 = 0;
393 lastname[l1] && (fname[l1] == lastname[l1]) && (l1 < 255);
394 l1++) {}
395 l2 = strlen(fname+l1);
396
397 if (l1 > 0)
398 flags |= XMIT_SAME_NAME;
399 if (l2 > 255)
400 flags |= XMIT_LONG_NAME;
401
402 /* We must make sure we don't send a zero flag byte or the
403 * other end will terminate the flist transfer. Note that
404 * the use of XMIT_TOP_DIR on a non-dir has no meaning, so
405 * it's harmless way to add a bit to the first flag byte. */
406 if (protocol_version >= 28) {
407 if (!flags && !S_ISDIR(mode))
408 flags |= XMIT_TOP_DIR;
409 if ((flags & 0xFF00) || !flags) {
410 flags |= XMIT_EXTENDED_FLAGS;
411 write_shortint(f, flags);
412 } else
413 write_byte(f, flags);
414 } else {
415 if (!(flags & 0xFF))
416 flags |= S_ISDIR(mode) ? XMIT_LONG_NAME : XMIT_TOP_DIR;
417 write_byte(f, flags);
418 }
419 if (flags & XMIT_SAME_NAME)
420 write_byte(f, l1);
421 if (flags & XMIT_LONG_NAME)
422 write_int(f, l2);
423 else
424 write_byte(f, l2);
425 write_buf(f, fname + l1, l2);
426
427 if (first_hlink_ndx >= 0) {
428 write_int(f, first_hlink_ndx);
429 goto the_end;
430 }
431
432 write_longint(f, F_LENGTH(file));
433 if (!(flags & XMIT_SAME_TIME))
434 write_int(f, modtime);
435 if (!(flags & XMIT_SAME_MODE))
436 write_int(f, to_wire_mode(mode));
437 if (preserve_uid && !(flags & XMIT_SAME_UID)) {
438 if (!numeric_ids)
439 add_uid(uid);
440 write_int(f, uid);
441 }
442 if (preserve_gid && !(flags & XMIT_SAME_GID)) {
443 if (!numeric_ids)
444 add_gid(gid);
445 write_int(f, gid);
446 }
447 if ((preserve_devices && IS_DEVICE(mode))
448 || (preserve_specials && IS_SPECIAL(mode))) {
449 if (protocol_version < 28) {
450 if (!(flags & XMIT_SAME_RDEV_pre28))
451 write_int(f, (int)rdev);
452 } else {
453 if (!(flags & XMIT_SAME_RDEV_MAJOR))
454 write_int(f, major(rdev));
455 if (flags & XMIT_RDEV_MINOR_IS_SMALL)
456 write_byte(f, minor(rdev));
457 else
458 write_int(f, minor(rdev));
459 }
460 }
461
462#ifdef SUPPORT_LINKS
463 if (preserve_links && S_ISLNK(mode)) {
464 const char *sl = F_SYMLINK(file);
465 int len = strlen(sl);
466 write_int(f, len);
467 write_buf(f, sl, len);
468 }
469#endif
470
471#ifdef SUPPORT_HARD_LINKS
472 if (tmp_dev != 0 && protocol_version < 30) {
473 if (protocol_version < 26) {
474 /* 32-bit dev_t and ino_t */
475 write_int(f, (int32)dev);
476 write_int(f, (int32)tmp_ino);
477 } else {
478 /* 64-bit dev_t and ino_t */
479 if (!(flags & XMIT_SAME_DEV_pre30))
480 write_longint(f, dev);
481 write_longint(f, tmp_ino);
482 }
483 }
484#endif
485
486 if (always_checksum && (S_ISREG(mode) || protocol_version < 28)) {
487 const char *sum;
488 if (S_ISREG(mode))
489 sum = tmp_sum;
490 else {
491 /* Prior to 28, we sent a useless set of nulls. */
492 sum = empty_sum;
493 }
494 write_buf(f, sum, checksum_len);
495 }
496
497 the_end:
498 strlcpy(lastname, fname, MAXPATHLEN);
499}
500
501static struct file_struct *recv_file_entry(struct file_list *flist,
502 int flags, int f)
503{
504 static time_t modtime;
505 static mode_t mode;
506 static int64 dev;
507 static dev_t rdev;
508 static uint32 rdev_major;
509 static uid_t uid;
510 static gid_t gid;
511 static char lastname[MAXPATHLEN], *lastdir;
512 static int lastdir_depth, lastdir_len = -1;
513 static unsigned int del_hier_name_len = 0;
514 static int in_del_hier = 0;
515 char thisname[MAXPATHLEN];
516 unsigned int l1 = 0, l2 = 0;
517 int alloc_len, basename_len, dirname_len, linkname_len;
518 int extra_len = file_extra_cnt * EXTRA_LEN;
519 int first_hlink_ndx = -1;
520 OFF_T file_length;
521 char *basename, *dirname, *bp;
522 struct file_struct *file;
523
524 if (!flist) {
525 modtime = 0, mode = 0;
526 dev = 0, rdev = MAKEDEV(0, 0);
527 rdev_major = 0;
528 uid = 0, gid = 0;
529 *lastname = '\0';
530 lastdir_len = -1;
531 in_del_hier = 0;
532 return NULL;
533 }
534
535 if (flags & XMIT_SAME_NAME)
536 l1 = read_byte(f);
537
538 if (flags & XMIT_LONG_NAME)
539 l2 = read_int(f);
540 else
541 l2 = read_byte(f);
542
543 if (l2 >= MAXPATHLEN - l1) {
544 rprintf(FERROR,
545 "overflow: flags=0x%x l1=%d l2=%d lastname=%s [%s]\n",
546 flags, l1, l2, lastname, who_am_i());
547 overflow_exit("recv_file_entry");
548 }
549
550 strlcpy(thisname, lastname, l1 + 1);
551 read_sbuf(f, &thisname[l1], l2);
552 thisname[l1 + l2] = 0;
553
554 strlcpy(lastname, thisname, MAXPATHLEN);
555
556 clean_fname(thisname, 0);
557
558 if (sanitize_paths)
559 sanitize_path(thisname, thisname, "", 0, NULL);
560
561 if ((basename = strrchr(thisname, '/')) != NULL) {
562 dirname_len = ++basename - thisname; /* counts future '\0' */
563 if (lastdir_len == dirname_len - 1
564 && strncmp(thisname, lastdir, lastdir_len) == 0) {
565 dirname = lastdir;
566 dirname_len = 0; /* indicates no copy is needed */
567 } else
568 dirname = thisname;
569 } else {
570 basename = thisname;
571 dirname = NULL;
572 dirname_len = 0;
573 }
574 basename_len = strlen(basename) + 1; /* count the '\0' */
575
576#ifdef SUPPORT_HARD_LINKS
577 if (protocol_version >= 30
578 && BITS_SETnUNSET(flags, XMIT_HLINKED, XMIT_HLINK_FIRST)) {
579 struct file_struct *first;
580 first_hlink_ndx = read_int(f);
581 if (first_hlink_ndx < 0 || first_hlink_ndx >= flist->count) {
582 rprintf(FERROR,
583 "hard-link reference out of range: %d (%d)\n",
584 first_hlink_ndx, flist->count);
585 exit_cleanup(RERR_PROTOCOL);
586 }
587 first = flist->files[first_hlink_ndx];
588 file_length = F_LENGTH(first);
589 modtime = first->modtime;
590 mode = first->mode;
591 if (preserve_uid)
592 uid = F_UID(first);
593 if (preserve_gid)
594 gid = F_GID(first);
595 if ((preserve_devices && IS_DEVICE(mode))
596 || (preserve_specials && IS_SPECIAL(mode))) {
597 uint32 *devp = F_RDEV_P(first);
598 rdev = MAKEDEV(DEV_MAJOR(devp), DEV_MINOR(devp));
599 }
600 if (preserve_links && S_ISLNK(mode))
601 linkname_len = strlen(F_SYMLINK(first)) + 1;
602 else
603 linkname_len = 0;
604 goto create_object;
605 }
606#endif
607
608 file_length = read_longint(f);
609 if (!(flags & XMIT_SAME_TIME))
610 modtime = (time_t)read_int(f);
611 if (!(flags & XMIT_SAME_MODE))
612 mode = from_wire_mode(read_int(f));
613
614 if (chmod_modes && !S_ISLNK(mode))
615 mode = tweak_mode(mode, chmod_modes);
616
617 if (preserve_uid && !(flags & XMIT_SAME_UID))
618 uid = (uid_t)read_int(f);
619 if (preserve_gid && !(flags & XMIT_SAME_GID))
620 gid = (gid_t)read_int(f);
621
622 if ((preserve_devices && IS_DEVICE(mode))
623 || (preserve_specials && IS_SPECIAL(mode))) {
624 if (protocol_version < 28) {
625 if (!(flags & XMIT_SAME_RDEV_pre28))
626 rdev = (dev_t)read_int(f);
627 } else {
628 uint32 rdev_minor;
629 if (!(flags & XMIT_SAME_RDEV_MAJOR))
630 rdev_major = read_int(f);
631 if (flags & XMIT_RDEV_MINOR_IS_SMALL)
632 rdev_minor = read_byte(f);
633 else
634 rdev_minor = read_int(f);
635 rdev = MAKEDEV(rdev_major, rdev_minor);
636 }
637 extra_len += 2 * EXTRA_LEN;
638 file_length = 0;
639 } else if (protocol_version < 28)
640 rdev = MAKEDEV(0, 0);
641
642#ifdef SUPPORT_LINKS
643 if (preserve_links && S_ISLNK(mode)) {
644 linkname_len = read_int(f) + 1; /* count the '\0' */
645 if (linkname_len <= 0 || linkname_len > MAXPATHLEN) {
646 rprintf(FERROR, "overflow: linkname_len=%d\n",
647 linkname_len - 1);
648 overflow_exit("recv_file_entry");
649 }
650 }
651 else
652#endif
653 linkname_len = 0;
654
655#ifdef SUPPORT_HARD_LINKS
656 create_object:
657 if (preserve_hard_links) {
658 if (protocol_version < 28 && S_ISREG(mode))
659 flags |= XMIT_HLINKED;
660 if (flags & XMIT_HLINKED)
661 extra_len += EXTRA_LEN;
662 }
663#endif
664
665 if (always_checksum && S_ISREG(mode))
666 extra_len += SUM_EXTRA_CNT * EXTRA_LEN;
667
668 if (file_length > 0xFFFFFFFFu && S_ISREG(mode))
669 extra_len += EXTRA_LEN;
670
671#if EXTRA_ROUNDING > 0
672 if (extra_len & (EXTRA_ROUNDING * EXTRA_LEN))
673 extra_len = (extra_len | (EXTRA_ROUNDING * EXTRA_LEN)) + EXTRA_LEN;
674#endif
675
676 alloc_len = FILE_STRUCT_LEN + extra_len + basename_len + dirname_len
677 + linkname_len;
678 bp = pool_alloc(flist->file_pool, alloc_len, "recv_file_entry");
679
680 memset(bp, 0, extra_len + FILE_STRUCT_LEN);
681 bp += extra_len;
682 file = (struct file_struct *)bp;
683 bp += FILE_STRUCT_LEN;
684
685 memcpy(bp, basename, basename_len);
686 bp += basename_len + linkname_len; /* skip space for symlink too */
687
688#ifdef SUPPORT_HARD_LINKS
689 if (flags & XMIT_HLINKED)
690 file->flags |= FLAG_HLINKED;
691#endif
692 file->modtime = modtime;
693 file->len32 = (uint32)file_length;
694 if (file_length > 0xFFFFFFFFu && S_ISREG(mode)) {
695 file->flags |= FLAG_LENGTH64;
696 OPT_EXTRA(file, 0)->unum = (uint32)(file_length >> 32);
697 }
698 file->mode = mode;
699 if (preserve_uid)
700 F_UID(file) = uid;
701 if (preserve_gid)
702 F_GID(file) = gid;
703
704 if (dirname_len) {
705 file->dirname = lastdir = bp;
706 lastdir_len = dirname_len - 1;
707 memcpy(bp, dirname, dirname_len - 1);
708 bp += dirname_len;
709 bp[-1] = '\0';
710 lastdir_depth = count_dir_elements(lastdir);
711 F_DEPTH(file) = lastdir_depth + 1;
712 } else if (dirname) {
713 file->dirname = dirname; /* we're reusing lastname */
714 F_DEPTH(file) = lastdir_depth + 1;
715 } else
716 F_DEPTH(file) = 1;
717
718 if (S_ISDIR(mode)) {
719 if (basename_len == 1+1 && *basename == '.') /* +1 for '\0' */
720 F_DEPTH(file)--;
721 if (flags & XMIT_TOP_DIR) {
722 in_del_hier = recurse;
723 del_hier_name_len = F_DEPTH(file) == 0 ? 0 : l1 + l2;
724 if (relative_paths && del_hier_name_len > 2
725 && lastname[del_hier_name_len-1] == '.'
726 && lastname[del_hier_name_len-2] == '/')
727 del_hier_name_len -= 2;
728 file->flags |= FLAG_TOP_DIR | FLAG_XFER_DIR;
729 } else if (in_del_hier) {
730 if (!relative_paths || !del_hier_name_len
731 || (l1 >= del_hier_name_len
732 && lastname[del_hier_name_len] == '/'))
733 file->flags |= FLAG_XFER_DIR;
734 else
735 in_del_hier = 0;
736 }
737 }
738
739 if ((preserve_devices && IS_DEVICE(mode))
740 || (preserve_specials && IS_SPECIAL(mode))) {
741 uint32 *devp = F_RDEV_P(file);
742 DEV_MAJOR(devp) = major(rdev);
743 DEV_MINOR(devp) = minor(rdev);
744 }
745
746#ifdef SUPPORT_LINKS
747 if (linkname_len) {
748 bp = (char*)file->basename + basename_len;
749 if (first_hlink_ndx >= 0) {
750 struct file_struct *first = flist->files[first_hlink_ndx];
751 memcpy(bp, F_SYMLINK(first), linkname_len);
752 } else
753 read_sbuf(f, bp, linkname_len - 1);
754 if (sanitize_paths)
755 sanitize_path(bp, bp, "", lastdir_depth, NULL);
756 }
757#endif
758
759#ifdef SUPPORT_HARD_LINKS
760 if (preserve_hard_links && flags & XMIT_HLINKED) {
761 if (protocol_version >= 30) {
762 F_HL_GNUM(file) = flags & XMIT_HLINK_FIRST
763 ? flist->count : first_hlink_ndx;
764 } else {
765 static int32 cnt = 0;
766 struct idev_node *np;
767 int64 ino;
768 int32 ndx;
769 if (protocol_version < 26) {
770 dev = read_int(f);
771 ino = read_int(f);
772 } else {
773 if (!(flags & XMIT_SAME_DEV_pre30))
774 dev = read_longint(f);
775 ino = read_longint(f);
776 }
777 np = idev_node(dev, ino);
778 ndx = (int32)np->data - 1;
779 if (ndx < 0) {
780 ndx = cnt++;
781 np->data = (void*)cnt;
782 }
783 F_HL_GNUM(file) = ndx;
784 }
785 }
786#endif
787
788 if (always_checksum && (S_ISREG(mode) || protocol_version < 28)) {
789 if (S_ISREG(mode))
790 bp = (char*)F_SUM(file);
791 else {
792 /* Prior to 28, we get a useless set of nulls. */
793 bp = tmp_sum;
794 }
795 if (first_hlink_ndx >= 0) {
796 struct file_struct *first = flist->files[first_hlink_ndx];
797 memcpy(bp, F_SUM(first), checksum_len);
798 } else
799 read_buf(f, bp, checksum_len);
800 }
801
802 return file;
803}
804
805/**
806 * Create a file_struct for a named file by reading its stat()
807 * information and performing extensive checks against global
808 * options.
809 *
810 * @return the new file, or NULL if there was an error or this file
811 * should be excluded.
812 *
813 * @todo There is a small optimization opportunity here to avoid
814 * stat()ing the file in some circumstances, which has a certain cost.
815 * We are called immediately after doing readdir(), and so we may
816 * already know the d_type of the file. We could for example avoid
817 * statting directories if we're not recursing, but this is not a very
818 * important case. Some systems may not have d_type.
819 **/
820struct file_struct *make_file(const char *fname, struct file_list *flist,
821 STRUCT_STAT *stp, int flags, int filter_level)
822{
823 static char *lastdir;
824 static int lastdir_len = -1;
825 struct file_struct *file;
826 STRUCT_STAT st;
827 char thisname[MAXPATHLEN];
828 char linkname[MAXPATHLEN];
829 int alloc_len, basename_len, dirname_len, linkname_len;
830 int extra_len = file_extra_cnt * EXTRA_LEN;
831 char *basename, *dirname, *bp;
832
833 if (!flist || !flist->count) /* Ignore lastdir when invalid. */
834 lastdir_len = -1;
835
836 if (strlcpy(thisname, fname, sizeof thisname)
837 >= sizeof thisname - flist_dir_len) {
838 rprintf(FINFO, "skipping overly long name: %s\n", fname);
839 return NULL;
840 }
841 clean_fname(thisname, 0);
842 if (sanitize_paths)
843 sanitize_path(thisname, thisname, "", 0, NULL);
844
845 if (stp && S_ISDIR(stp->st_mode)) {
846 st = *stp; /* Needed for "symlink/." with --relative. */
847 *linkname = '\0'; /* make IBM code checker happy */
848 } else if (readlink_stat(thisname, &st, linkname) != 0) {
849 int save_errno = errno;
850 /* See if file is excluded before reporting an error. */
851 if (filter_level != NO_FILTERS
852 && (is_excluded(thisname, 0, filter_level)
853 || is_excluded(thisname, 1, filter_level))) {
854 if (ignore_perishable && save_errno != ENOENT)
855 non_perishable_cnt++;
856 return NULL;
857 }
858 if (save_errno == ENOENT) {
859#ifdef SUPPORT_LINKS
860 /* Avoid "vanished" error if symlink points nowhere. */
861 if (copy_links && do_lstat(thisname, &st) == 0
862 && S_ISLNK(st.st_mode)) {
863 io_error |= IOERR_GENERAL;
864 rprintf(FERROR, "symlink has no referent: %s\n",
865 full_fname(thisname));
866 } else
867#endif
868 {
869 enum logcode c = am_daemon && protocol_version < 28
870 ? FERROR : FINFO;
871 io_error |= IOERR_VANISHED;
872 rprintf(c, "file has vanished: %s\n",
873 full_fname(thisname));
874 }
875 } else {
876 io_error |= IOERR_GENERAL;
877 rsyserr(FERROR, save_errno, "readlink %s failed",
878 full_fname(thisname));
879 }
880 return NULL;
881 }
882
883 /* backup.c calls us with filter_level set to NO_FILTERS. */
884 if (filter_level == NO_FILTERS)
885 goto skip_filters;
886
887 if (S_ISDIR(st.st_mode) && !xfer_dirs) {
888 rprintf(FINFO, "skipping directory %s\n", thisname);
889 return NULL;
890 }
891
892 /* -x only affects directories because we need to avoid recursing
893 * into a mount-point directory, not to avoid copying a symlinked
894 * file if -L (or similar) was specified. */
895 if (one_file_system && st.st_dev != filesystem_dev
896 && S_ISDIR(st.st_mode)) {
897 if (one_file_system > 1) {
898 if (verbose > 2) {
899 rprintf(FINFO, "skipping mount-point dir %s\n",
900 thisname);
901 }
902 return NULL;
903 }
904 flags |= FLAG_MOUNT_DIR;
905 }
906
907 if (is_excluded(thisname, S_ISDIR(st.st_mode) != 0, filter_level)) {
908 if (ignore_perishable)
909 non_perishable_cnt++;
910 return NULL;
911 }
912
913 if (lp_ignore_nonreadable(module_id)) {
914#ifdef SUPPORT_LINKS
915 if (!S_ISLNK(st.st_mode))
916#endif
917 if (access(thisname, R_OK) != 0)
918 return NULL;
919 }
920
921 skip_filters:
922
923 if (verbose > 2) {
924 rprintf(FINFO, "[%s] make_file(%s,*,%d)\n",
925 who_am_i(), thisname, filter_level);
926 }
927
928 if ((basename = strrchr(thisname, '/')) != NULL) {
929 dirname_len = ++basename - thisname; /* counts future '\0' */
930 if (lastdir_len == dirname_len - 1
931 && strncmp(thisname, lastdir, lastdir_len) == 0) {
932 dirname = lastdir;
933 dirname_len = 0; /* indicates no copy is needed */
934 } else
935 dirname = thisname;
936 } else {
937 basename = thisname;
938 dirname = NULL;
939 dirname_len = 0;
940 }
941 basename_len = strlen(basename) + 1; /* count the '\0' */
942
943#ifdef SUPPORT_LINKS
944 linkname_len = S_ISLNK(st.st_mode) ? strlen(linkname) + 1 : 0;
945#else
946 linkname_len = 0;
947#endif
948
949 if (st.st_size > 0xFFFFFFFFu && S_ISREG(st.st_mode))
950 extra_len += EXTRA_LEN;
951
952#if EXTRA_ROUNDING > 0
953 if (extra_len & (EXTRA_ROUNDING * EXTRA_LEN))
954 extra_len = (extra_len | (EXTRA_ROUNDING * EXTRA_LEN)) + EXTRA_LEN;
955#endif
956
957 alloc_len = FILE_STRUCT_LEN + extra_len + basename_len + dirname_len
958 + linkname_len;
959 if (flist)
960 bp = pool_alloc(flist->file_pool, alloc_len, "make_file");
961 else {
962 if (!(bp = new_array(char, alloc_len)))
963 out_of_memory("make_file");
964 }
965
966 memset(bp, 0, extra_len + FILE_STRUCT_LEN);
967 bp += extra_len;
968 file = (struct file_struct *)bp;
969 bp += FILE_STRUCT_LEN;
970
971 memcpy(bp, basename, basename_len);
972 bp += basename_len + linkname_len; /* skip space for symlink too */
973
974#ifdef SUPPORT_HARD_LINKS
975 if (preserve_hard_links && flist) {
976 if (protocol_version >= 28
977 ? (!S_ISDIR(st.st_mode) && st.st_nlink > 1)
978 : S_ISREG(st.st_mode)) {
979 tmp_dev = st.st_dev;
980 tmp_ino = st.st_ino;
981 } else
982 tmp_dev = 0;
983 }
984#endif
985
986#ifdef HAVE_STRUCT_STAT_ST_RDEV
987 if (IS_DEVICE(st.st_mode) || IS_SPECIAL(st.st_mode)) {
988 tmp_rdev = st.st_rdev;
989 st.st_size = 0;
990 }
991#endif
992
993 file->flags = flags;
994 file->modtime = st.st_mtime;
995 file->len32 = (uint32)st.st_size;
996 if (st.st_size > 0xFFFFFFFFu && S_ISREG(st.st_mode)) {
997 file->flags |= FLAG_LENGTH64;
998 OPT_EXTRA(file, 0)->unum = (uint32)(st.st_size >> 32);
999 }
1000 file->mode = st.st_mode;
1001 if (preserve_uid)
1002 F_UID(file) = st.st_uid;
1003 if (preserve_gid)
1004 F_GID(file) = st.st_gid;
1005
1006 if (dirname_len) {
1007 file->dirname = lastdir = bp;
1008 lastdir_len = dirname_len - 1;
1009 memcpy(bp, dirname, dirname_len - 1);
1010 bp += dirname_len;
1011 bp[-1] = '\0';
1012 } else if (dirname)
1013 file->dirname = dirname;
1014
1015#ifdef SUPPORT_LINKS
1016 if (linkname_len) {
1017 bp = (char*)file->basename + basename_len;
1018 memcpy(bp, linkname, linkname_len);
1019 }
1020#endif
1021
1022 if (always_checksum && am_sender && S_ISREG(st.st_mode))
1023 file_checksum(thisname, tmp_sum, st.st_size);
1024
1025 F_ROOTDIR(file) = flist_dir;
1026
1027 /* This code is only used by the receiver when it is building
1028 * a list of files for a delete pass. */
1029 if (keep_dirlinks && linkname_len && flist) {
1030 STRUCT_STAT st2;
1031 int save_mode = file->mode;
1032 file->mode = S_IFDIR; /* Find a directory with our name. */
1033 if (flist_find(the_file_list, file) >= 0
1034 && do_stat(thisname, &st2) == 0 && S_ISDIR(st2.st_mode)) {
1035 file->modtime = st2.st_mtime;
1036 file->len32 = 0;
1037 file->mode = st2.st_mode;
1038 if (preserve_uid)
1039 F_UID(file) = st2.st_uid;
1040 if (preserve_gid)
1041 F_GID(file) = st2.st_gid;
1042 } else
1043 file->mode = save_mode;
1044 }
1045
1046 if (S_ISREG(st.st_mode) || S_ISLNK(st.st_mode))
1047 stats.total_size += st.st_size;
1048
1049 if (basename_len == 0+1)
1050 return NULL;
1051
1052 return file;
1053}
1054
1055/* Only called for temporary file_struct entries created by make_file(). */
1056void unmake_file(struct file_struct *file)
1057{
1058 int extra_cnt = file_extra_cnt + LEN64_BUMP(file);
1059#if EXTRA_ROUNDING > 0
1060 if (extra_cnt & EXTRA_ROUNDING)
1061 extra_cnt = (extra_cnt | EXTRA_ROUNDING) + 1;
1062#endif
1063 free(REQ_EXTRA(file, extra_cnt));
1064}
1065
1066static struct file_struct *send_file_name(int f, struct file_list *flist,
1067 char *fname, STRUCT_STAT *stp,
1068 int flags)
1069{
1070 struct file_struct *file;
1071
1072 file = make_file(fname, flist, stp, flags,
1073 f == -2 ? SERVER_FILTERS : ALL_FILTERS);
1074 if (!file)
1075 return NULL;
1076
1077 if (chmod_modes && !S_ISLNK(file->mode))
1078 file->mode = tweak_mode(file->mode, chmod_modes);
1079
1080 maybe_emit_filelist_progress(flist->count + flist_count_offset);
1081
1082 flist_expand(flist);
1083 flist->files[flist->count++] = file;
1084 if (f >= 0)
1085 send_file_entry(file, f, flist->count - 1);
1086 return file;
1087}
1088
1089static void send_if_directory(int f, struct file_list *flist,
1090 struct file_struct *file,
1091 char *fbuf, unsigned int ol)
1092{
1093 char is_dot_dir = fbuf[ol-1] == '.' && (ol == 1 || fbuf[ol-2] == '/');
1094
1095 if (S_ISDIR(file->mode)
1096 && !(file->flags & FLAG_MOUNT_DIR) && f_name(file, fbuf)) {
1097 void *save_filters;
1098 unsigned int len = strlen(fbuf);
1099 if (len > 1 && fbuf[len-1] == '/')
1100 fbuf[--len] = '\0';
1101 if (len >= MAXPATHLEN - 1) {
1102 io_error |= IOERR_GENERAL;
1103 rprintf(FERROR, "skipping long-named directory: %s\n",
1104 full_fname(fbuf));
1105 return;
1106 }
1107 save_filters = push_local_filters(fbuf, len);
1108 send_directory(f, flist, fbuf, len);
1109 pop_local_filters(save_filters);
1110 fbuf[ol] = '\0';
1111 if (is_dot_dir)
1112 fbuf[ol-1] = '.';
1113 }
1114}
1115
1116/* This function is normally called by the sender, but the receiving side also
1117 * calls it from get_dirlist() with f set to -1 so that we just construct the
1118 * file list in memory without sending it over the wire. Also, get_dirlist()
1119 * might call this with f set to -2, which also indicates that local filter
1120 * rules should be ignored. */
1121static void send_directory(int f, struct file_list *flist, char *fbuf, int len)
1122{
1123 struct dirent *di;
1124 unsigned remainder;
1125 char *p;
1126 DIR *d;
1127 int start = flist->count;
1128
1129 if (!(d = opendir(fbuf))) {
1130 io_error |= IOERR_GENERAL;
1131 rsyserr(FERROR, errno, "opendir %s failed", full_fname(fbuf));
1132 return;
1133 }
1134
1135 p = fbuf + len;
1136 if (len != 1 || *fbuf != '/')
1137 *p++ = '/';
1138 *p = '\0';
1139 remainder = MAXPATHLEN - (p - fbuf);
1140
1141 for (errno = 0, di = readdir(d); di; errno = 0, di = readdir(d)) {
1142 char *dname = d_name(di);
1143 if (dname[0] == '.' && (dname[1] == '\0'
1144 || (dname[1] == '.' && dname[2] == '\0')))
1145 continue;
1146 if (strlcpy(p, dname, remainder) >= remainder) {
1147 io_error |= IOERR_GENERAL;
1148 rprintf(FINFO,
1149 "cannot send long-named file %s\n",
1150 full_fname(fbuf));
1151 continue;
1152 }
1153
1154 send_file_name(f, flist, fbuf, NULL, 0);
1155 }
1156
1157 fbuf[len] = '\0';
1158
1159 if (errno) {
1160 io_error |= IOERR_GENERAL;
1161 rsyserr(FERROR, errno, "readdir(%s)", full_fname(fbuf));
1162 }
1163
1164 closedir(d);
1165
1166 if (recurse) {
1167 int i, end = flist->count - 1;
1168 /* send_if_directory() bumps flist->count, so use "end". */
1169 for (i = start; i <= end; i++)
1170 send_if_directory(f, flist, flist->files[i], fbuf, len);
1171 }
1172}
1173
1174struct file_list *send_file_list(int f, int argc, char *argv[])
1175{
1176 int len;
1177 STRUCT_STAT st;
1178 char *p, *dir, olddir[sizeof curr_dir];
1179 char lastpath[MAXPATHLEN] = "";
1180 struct file_list *flist;
1181 struct timeval start_tv, end_tv;
1182 int64 start_write;
1183 int use_ff_fd = 0;
1184
1185 rprintf(FLOG, "building file list\n");
1186 if (show_filelist_p())
1187 start_filelist_progress("building file list");
1188
1189 start_write = stats.total_written;
1190 gettimeofday(&start_tv, NULL);
1191
1192#ifdef SUPPORT_HARD_LINKS
1193 if (preserve_hard_links && protocol_version >= 30)
1194 init_hard_links();
1195#endif
1196
1197 flist = flist_new("send_file_list");
1198
1199 io_start_buffering_out();
1200 if (filesfrom_fd >= 0) {
1201 if (argv[0] && !push_dir(argv[0], 0)) {
1202 rsyserr(FERROR, errno, "push_dir %s failed",
1203 full_fname(argv[0]));
1204 exit_cleanup(RERR_FILESELECT);
1205 }
1206 use_ff_fd = 1;
1207 }
1208
1209 while (1) {
1210 char fbuf[MAXPATHLEN];
1211 char *fn;
1212 int is_dot_dir;
1213
1214 if (use_ff_fd) {
1215 if (read_filesfrom_line(filesfrom_fd, fbuf) == 0)
1216 break;
1217 sanitize_path(fbuf, fbuf, "", 0, NULL);
1218 } else {
1219 if (argc-- == 0)
1220 break;
1221 strlcpy(fbuf, *argv++, MAXPATHLEN);
1222 if (sanitize_paths)
1223 sanitize_path(fbuf, fbuf, "", 0, NULL);
1224 }
1225
1226 len = strlen(fbuf);
1227 if (relative_paths) {
1228 /* We clean up fbuf below. */
1229 is_dot_dir = 0;
1230 } else if (!len || fbuf[len - 1] == '/') {
1231 if (len == 2 && fbuf[0] == '.') {
1232 /* Turn "./" into just "." rather than "./." */
1233 fbuf[1] = '\0';
1234 } else {
1235 if (len + 1 >= MAXPATHLEN)
1236 overflow_exit("send_file_list");
1237 fbuf[len++] = '.';
1238 fbuf[len] = '\0';
1239 }
1240 is_dot_dir = 1;
1241 } else if (len > 1 && fbuf[len-1] == '.' && fbuf[len-2] == '.'
1242 && (len == 2 || fbuf[len-3] == '/')) {
1243 if (len + 2 >= MAXPATHLEN)
1244 overflow_exit("send_file_list");
1245 fbuf[len++] = '/';
1246 fbuf[len++] = '.';
1247 fbuf[len] = '\0';
1248 is_dot_dir = 1;
1249 } else {
1250 is_dot_dir = fbuf[len-1] == '.'
1251 && (len == 1 || fbuf[len-2] == '/');
1252 }
1253
1254 if (link_stat(fbuf, &st, copy_dirlinks) != 0) {
1255 io_error |= IOERR_GENERAL;
1256 rsyserr(FERROR, errno, "link_stat %s failed",
1257 full_fname(fbuf));
1258 continue;
1259 }
1260
1261 if (S_ISDIR(st.st_mode) && !xfer_dirs) {
1262 rprintf(FINFO, "skipping directory %s\n", fbuf);
1263 continue;
1264 }
1265
1266 dir = NULL;
1267 olddir[0] = '\0';
1268
1269 if (!relative_paths) {
1270 p = strrchr(fbuf, '/');
1271 if (p) {
1272 *p = '\0';
1273 if (p == fbuf)
1274 dir = "/";
1275 else
1276 dir = fbuf;
1277 len -= p - fbuf + 1;
1278 fn = p + 1;
1279 } else
1280 fn = fbuf;
1281 } else {
1282 if ((p = strstr(fbuf, "/./")) != NULL) {
1283 *p = '\0';
1284 if (p == fbuf)
1285 dir = "/";
1286 else
1287 dir = fbuf;
1288 len -= p - fbuf + 3;
1289 fn = p + 3;
1290 } else
1291 fn = fbuf;
1292 /* Get rid of trailing "/" and "/.". */
1293 while (len) {
1294 if (fn[len - 1] == '/') {
1295 is_dot_dir = 1;
1296 if (!--len && !dir) {
1297 len++;
1298 break;
1299 }
1300 }
1301 else if (len >= 2 && fn[len - 1] == '.'
1302 && fn[len - 2] == '/') {
1303 is_dot_dir = 1;
1304 if (!(len -= 2) && !dir) {
1305 len++;
1306 break;
1307 }
1308 } else
1309 break;
1310 }
1311 if (len == 1 && fn[0] == '/')
1312 fn[len++] = '.';
1313 fn[len] = '\0';
1314 /* Reject a ".." dir in the active part of the path. */
1315 for (p = fn; (p = strstr(p, "..")) != NULL; p += 2) {
1316 if ((p[2] == '/' || p[2] == '\0')
1317 && (p == fn || p[-1] == '/')) {
1318 rprintf(FERROR,
1319 "found \"..\" dir in relative path: %s\n",
1320 fbuf);
1321 exit_cleanup(RERR_SYNTAX);
1322 }
1323 }
1324 }
1325
1326 if (!*fn) {
1327 len = 1;
1328 fn = ".";
1329 }
1330
1331 if (dir && *dir) {
1332 static const char *lastdir;
1333 static int lastdir_len;
1334
1335 strlcpy(olddir, curr_dir, sizeof olddir);
1336
1337 if (!push_dir(dir, 0)) {
1338 io_error |= IOERR_GENERAL;
1339 rsyserr(FERROR, errno, "push_dir %s failed",
1340 full_fname(dir));
1341 continue;
1342 }
1343
1344 if (lastdir && strcmp(lastdir, dir) == 0) {
1345 flist_dir = lastdir;
1346 flist_dir_len = lastdir_len;
1347 } else {
1348 flist_dir = lastdir = strdup(dir);
1349 flist_dir_len = lastdir_len = strlen(dir);
1350 }
1351 }
1352
1353 if (fn != fbuf)
1354 memmove(fbuf, fn, len + 1);
1355
1356 if (implied_dirs && (p=strrchr(fbuf,'/')) && p != fbuf) {
1357 /* Send the implied directories at the start of the
1358 * source spec, so we get their permissions right. */
1359 char *lp = lastpath, *slash = fbuf;
1360 *p = '\0';
1361 /* Skip any initial directories in our path that we
1362 * have in common with lastpath. */
1363 for (fn = fbuf; *fn && *lp == *fn; lp++, fn++) {
1364 if (*fn == '/')
1365 slash = fn;
1366 }
1367 *p = '/';
1368 if (fn != p || (*lp && *lp != '/')) {
1369 int save_copy_links = copy_links;
1370 int save_xfer_dirs = xfer_dirs;
1371 copy_links |= copy_unsafe_links;
1372 xfer_dirs = 1;
1373 while ((slash = strchr(slash+1, '/')) != 0) {
1374 *slash = '\0';
1375 send_file_name(f, flist, fbuf, NULL, 0);
1376 *slash = '/';
1377 }
1378 copy_links = save_copy_links;
1379 xfer_dirs = save_xfer_dirs;
1380 *p = '\0';
1381 strlcpy(lastpath, fbuf, sizeof lastpath);
1382 *p = '/';
1383 }
1384 }
1385
1386 if (one_file_system)
1387 filesystem_dev = st.st_dev;
1388
1389 if (recurse || (xfer_dirs && is_dot_dir)) {
1390 struct file_struct *file;
1391 file = send_file_name(f, flist, fbuf, &st, FLAG_TOP_DIR);
1392 if (file)
1393 send_if_directory(f, flist, file, fbuf, len);
1394 } else
1395 send_file_name(f, flist, fbuf, &st, 0);
1396
1397 if (olddir[0]) {
1398 flist_dir = NULL;
1399 flist_dir_len = 0;
1400 if (!pop_dir(olddir)) {
1401 rsyserr(FERROR, errno, "pop_dir %s failed",
1402 full_fname(olddir));
1403 exit_cleanup(RERR_FILESELECT);
1404 }
1405 }
1406 }
1407
1408 gettimeofday(&end_tv, NULL);
1409 stats.flist_buildtime = (int64)(end_tv.tv_sec - start_tv.tv_sec) * 1000
1410 + (end_tv.tv_usec - start_tv.tv_usec) / 1000;
1411 if (stats.flist_buildtime == 0)
1412 stats.flist_buildtime = 1;
1413 start_tv = end_tv;
1414
1415 write_byte(f, 0); /* Indicate end of file list */
1416
1417#ifdef SUPPORT_HARD_LINKS
1418 if (preserve_hard_links && protocol_version >= 30)
1419 idev_destroy();
1420#endif
1421
1422 if (show_filelist_p())
1423 finish_filelist_progress(flist);
1424
1425 gettimeofday(&end_tv, NULL);
1426 stats.flist_xfertime = (int64)(end_tv.tv_sec - start_tv.tv_sec) * 1000
1427 + (end_tv.tv_usec - start_tv.tv_usec) / 1000;
1428
1429 /* Sort the list without removing any duplicates. This allows the
1430 * receiving side to ask for any name they like, which gives us the
1431 * flexibility to change the way we unduplicate names in the future
1432 * without causing a compatibility problem with older versions. */
1433 clean_flist(flist, 0, 0);
1434
1435 if (!numeric_ids)
1436 send_uid_list(f);
1437
1438 /* send the io_error flag */
1439 write_int(f, ignore_errors ? 0 : io_error);
1440
1441 io_end_buffering();
1442 stats.flist_size = stats.total_written - start_write;
1443 stats.num_files = flist->count;
1444
1445 if (verbose > 3)
1446 output_flist(flist);
1447
1448 if (verbose > 2)
1449 rprintf(FINFO, "send_file_list done\n");
1450
1451 return flist;
1452}
1453
1454struct file_list *recv_file_list(int f)
1455{
1456 struct file_list *flist;
1457 int flags;
1458 int64 start_read;
1459
1460 rprintf(FLOG, "receiving file list\n");
1461 if (show_filelist_p())
1462 start_filelist_progress("receiving file list");
1463
1464 start_read = stats.total_read;
1465
1466 flist = flist_new("recv_file_list");
1467
1468#ifdef SUPPORT_HARD_LINKS
1469 if (preserve_hard_links && protocol_version < 30)
1470 init_hard_links();
1471#endif
1472
1473 while ((flags = read_byte(f)) != 0) {
1474 struct file_struct *file;
1475
1476 flist_expand(flist);
1477
1478 if (protocol_version >= 28 && (flags & XMIT_EXTENDED_FLAGS))
1479 flags |= read_byte(f) << 8;
1480 file = recv_file_entry(flist, flags, f);
1481
1482 if (S_ISREG(file->mode) || S_ISLNK(file->mode))
1483 stats.total_size += F_LENGTH(file);
1484
1485 flist->files[flist->count++] = file;
1486
1487 maybe_emit_filelist_progress(flist->count);
1488
1489 if (verbose > 2) {
1490 rprintf(FINFO, "recv_file_name(%s)\n",
1491 f_name(file, NULL));
1492 }
1493 }
1494 recv_file_entry(NULL, 0, 0); /* Signal that we're done. */
1495
1496 if (verbose > 2)
1497 rprintf(FINFO, "received %d names\n", flist->count);
1498
1499 if (show_filelist_p())
1500 finish_filelist_progress(flist);
1501
1502 clean_flist(flist, relative_paths, 1);
1503
1504 if (f >= 0) {
1505 recv_uid_list(f, flist);
1506
1507 /* Recv the io_error flag */
1508 if (ignore_errors)
1509 read_int(f);
1510 else
1511 io_error |= read_int(f);
1512 }
1513
1514 if (verbose > 3)
1515 output_flist(flist);
1516
1517 if (list_only) {
1518 int i;
1519 for (i = 0; i < flist->count; i++)
1520 list_file_entry(flist->files[i]);
1521 }
1522
1523 if (verbose > 2)
1524 rprintf(FINFO, "recv_file_list done\n");
1525
1526 stats.flist_size = stats.total_read - start_read;
1527 stats.num_files = flist->count;
1528
1529 return flist;
1530}
1531
1532static int file_compare(struct file_struct **file1, struct file_struct **file2)
1533{
1534 return f_name_cmp(*file1, *file2);
1535}
1536
1537/* Search for an identically-named item in the file list. Note that the
1538 * items must agree in their directory-ness, or no match is returned. */
1539int flist_find(struct file_list *flist, struct file_struct *f)
1540{
1541 int low = flist->low, high = flist->high;
1542 int diff, mid, mid_up;
1543
1544 while (low <= high) {
1545 mid = (low + high) / 2;
1546 if (F_IS_ACTIVE(flist->files[mid]))
1547 mid_up = mid;
1548 else {
1549 /* Scan for the next non-empty entry using the cached
1550 * distance values. If the value isn't fully up-to-
1551 * date, update it. */
1552 mid_up = mid + F_DEPTH(flist->files[mid]);
1553 if (!F_IS_ACTIVE(flist->files[mid_up])) {
1554 do {
1555 mid_up += F_DEPTH(flist->files[mid_up]);
1556 } while (!F_IS_ACTIVE(flist->files[mid_up]));
1557 F_DEPTH(flist->files[mid]) = mid_up - mid;
1558 }
1559 if (mid_up > high) {
1560 /* If there's nothing left above us, set high to
1561 * a non-empty entry below us and continue. */
1562 high = mid - (int)flist->files[mid]->len32;
1563 if (!F_IS_ACTIVE(flist->files[high])) {
1564 do {
1565 high -= (int)flist->files[high]->len32;
1566 } while (!F_IS_ACTIVE(flist->files[high]));
1567 flist->files[mid]->len32 = mid - high;
1568 }
1569 continue;
1570 }
1571 }
1572 diff = f_name_cmp(flist->files[mid_up], f);
1573 if (diff == 0) {
1574 if (protocol_version < 29
1575 && S_ISDIR(flist->files[mid_up]->mode)
1576 != S_ISDIR(f->mode))
1577 return -1;
1578 return mid_up;
1579 }
1580 if (diff < 0)
1581 low = mid_up + 1;
1582 else
1583 high = mid - 1;
1584 }
1585 return -1;
1586}
1587
1588/*
1589 * Free up any resources a file_struct has allocated
1590 * and clear the file.
1591 */
1592void clear_file(struct file_struct *file)
1593{
1594 /* The +1 zeros out the first char of the basename. */
1595 memset(file, 0, FILE_STRUCT_LEN + 1);
1596 /* In an empty entry, F_DEPTH() is an offset to the next non-empty
1597 * entry. Likewise for len32 in the opposite direction. We assume
1598 * that we're alone for now since flist_find() will adjust the counts
1599 * it runs into that aren't up-to-date. */
1600 file->len32 = F_DEPTH(file) = 1;
1601}
1602
1603/* Allocate a new file list. */
1604struct file_list *flist_new(char *msg)
1605{
1606 struct file_list *flist;
1607
1608 flist = new(struct file_list);
1609 if (!flist)
1610 out_of_memory(msg);
1611
1612 memset(flist, 0, sizeof flist[0]);
1613
1614 if (!(flist->file_pool = pool_create(FILE_EXTENT, 0, out_of_memory, POOL_INTERN)))
1615 out_of_memory(msg);
1616
1617 return flist;
1618}
1619
1620/* Free up all elements in a flist. */
1621void flist_free(struct file_list *flist)
1622{
1623 pool_destroy(flist->file_pool);
1624 free(flist->files);
1625 free(flist);
1626}
1627
1628/*
1629 * This routine ensures we don't have any duplicate names in our file list.
1630 * duplicate names can cause corruption because of the pipelining
1631 */
1632static void clean_flist(struct file_list *flist, int strip_root, int no_dups)
1633{
1634 char fbuf[MAXPATHLEN];
1635 int i, prev_i = 0;
1636
1637 if (!flist)
1638 return;
1639 if (flist->count == 0) {
1640 flist->high = -1;
1641 return;
1642 }
1643
1644 qsort(flist->files, flist->count,
1645 sizeof flist->files[0], (int (*)())file_compare);
1646
1647 for (i = no_dups? 0 : flist->count; i < flist->count; i++) {
1648 if (F_IS_ACTIVE(flist->files[i])) {
1649 prev_i = i;
1650 break;
1651 }
1652 }
1653 flist->low = prev_i;
1654 while (++i < flist->count) {
1655 int j;
1656 struct file_struct *file = flist->files[i];
1657
1658 if (!F_IS_ACTIVE(file))
1659 continue;
1660 if (f_name_cmp(file, flist->files[prev_i]) == 0)
1661 j = prev_i;
1662 else if (protocol_version >= 29 && S_ISDIR(file->mode)) {
1663 int save_mode = file->mode;
1664 /* Make sure that this directory doesn't duplicate a
1665 * non-directory earlier in the list. */
1666 flist->high = prev_i;
1667 file->mode = S_IFREG;
1668 j = flist_find(flist, file);
1669 file->mode = save_mode;
1670 } else
1671 j = -1;
1672 if (j >= 0) {
1673 struct file_struct *fp = flist->files[j];
1674 int keep, drop;
1675 /* If one is a dir and the other is not, we want to
1676 * keep the dir because it might have contents in the
1677 * list. */
1678 if (S_ISDIR(file->mode) != S_ISDIR(fp->mode)) {
1679 if (S_ISDIR(file->mode))
1680 keep = i, drop = j;
1681 else
1682 keep = j, drop = i;
1683 } else
1684 keep = j, drop = i;
1685 if (verbose > 1 && !am_server) {
1686 rprintf(FINFO,
1687 "removing duplicate name %s from file list (%d)\n",
1688 f_name(file, fbuf), drop);
1689 }
1690 /* Make sure we don't lose track of a user-specified
1691 * top directory. */
1692 flist->files[keep]->flags |= flist->files[drop]->flags
1693 & (FLAG_TOP_DIR|FLAG_XFER_DIR);
1694
1695 clear_file(flist->files[drop]);
1696
1697 if (keep == i) {
1698 if (flist->low == drop) {
1699 for (j = drop + 1;
1700 j < i && !F_IS_ACTIVE(flist->files[j]);
1701 j++) {}
1702 flist->low = j;
1703 }
1704 prev_i = i;
1705 }
1706 } else
1707 prev_i = i;
1708 }
1709 flist->high = no_dups ? prev_i : flist->count - 1;
1710
1711 if (strip_root) {
1712 /* We need to strip off the leading slashes for relative
1713 * paths, but this must be done _after_ the sorting phase. */
1714 for (i = flist->low; i <= flist->high; i++) {
1715 struct file_struct *file = flist->files[i];
1716
1717 if (!file->dirname)
1718 continue;
1719 while (*file->dirname == '/')
1720 file->dirname++;
1721 if (!*file->dirname)
1722 file->dirname = NULL;
1723 }
1724 }
1725
1726 if (prune_empty_dirs && no_dups) {
1727 int j, prev_depth = 0;
1728
1729 prev_i = 0; /* It's OK that this isn't really true. */
1730
1731 for (i = flist->low; i <= flist->high; i++) {
1732 struct file_struct *fp, *file = flist->files[i];
1733
1734 /* This temporarily abuses the F_DEPTH() value for a
1735 * directory that is in a chain that might get pruned.
1736 * We restore the old value if it gets a reprieve. */
1737 if (S_ISDIR(file->mode) && F_DEPTH(file)) {
1738 /* Dump empty dirs when coming back down. */
1739 for (j = prev_depth; j >= F_DEPTH(file); j--) {
1740 fp = flist->files[prev_i];
1741 if (F_DEPTH(fp) >= 0)
1742 break;
1743 prev_i = -F_DEPTH(fp)-1;
1744 clear_file(fp);
1745 }
1746 prev_depth = F_DEPTH(file);
1747 if (is_excluded(f_name(file, fbuf), 1,
1748 ALL_FILTERS)) {
1749 /* Keep dirs through this dir. */
1750 for (j = prev_depth-1; ; j--) {
1751 fp = flist->files[prev_i];
1752 if (F_DEPTH(fp) >= 0)
1753 break;
1754 prev_i = -F_DEPTH(fp)-1;
1755 F_DEPTH(fp) = j;
1756 }
1757 } else
1758 F_DEPTH(file) = -prev_i-1;
1759 prev_i = i;
1760 } else {
1761 /* Keep dirs through this non-dir. */
1762 for (j = prev_depth; ; j--) {
1763 fp = flist->files[prev_i];
1764 if (F_DEPTH(fp) >= 0)
1765 break;
1766 prev_i = -F_DEPTH(fp)-1;
1767 F_DEPTH(fp) = j;
1768 }
1769 }
1770 }
1771 /* Dump empty all remaining empty dirs. */
1772 while (1) {
1773 struct file_struct *fp = flist->files[prev_i];
1774 if (F_DEPTH(fp) >= 0)
1775 break;
1776 prev_i = -F_DEPTH(fp)-1;
1777 clear_file(fp);
1778 }
1779
1780 for (i = flist->low; i <= flist->high; i++) {
1781 if (F_IS_ACTIVE(flist->files[i]))
1782 break;
1783 }
1784 flist->low = i;
1785 for (i = flist->high; i >= flist->low; i--) {
1786 if (F_IS_ACTIVE(flist->files[i]))
1787 break;
1788 }
1789 flist->high = i;
1790 }
1791}
1792
1793static void output_flist(struct file_list *flist)
1794{
1795 char uidbuf[16], gidbuf[16], depthbuf[16];
1796 struct file_struct *file;
1797 const char *root, *dir, *slash, *name, *trail;
1798 const char *who = who_am_i();
1799 int i;
1800
1801 for (i = 0; i < flist->count; i++) {
1802 file = flist->files[i];
1803 if ((am_root || am_sender) && preserve_uid) {
1804 snprintf(uidbuf, sizeof uidbuf, " uid=%ld",
1805 (long)F_UID(file));
1806 } else
1807 *uidbuf = '\0';
1808 if (preserve_gid && F_GID(file) != GID_NONE) {
1809 snprintf(gidbuf, sizeof gidbuf, " gid=%ld",
1810 (long)F_GID(file));
1811 } else
1812 *gidbuf = '\0';
1813 if (!am_sender)
1814 snprintf(depthbuf, sizeof depthbuf, "%d", F_DEPTH(file));
1815 if (F_IS_ACTIVE(file)) {
1816 root = am_sender ? NS(F_ROOTDIR(file)) : depthbuf;
1817 if ((dir = file->dirname) == NULL)
1818 dir = slash = "";
1819 else
1820 slash = "/";
1821 name = file->basename;
1822 trail = S_ISDIR(file->mode) ? "/" : "";
1823 } else
1824 root = dir = slash = name = trail = "";
1825 rprintf(FINFO, "[%s] i=%d %s %s%s%s%s mode=0%o len=%.0f%s%s flags=%x\n",
1826 who, i, root, dir, slash, name, trail, (int)file->mode,
1827 (double)F_LENGTH(file), uidbuf, gidbuf, file->flags);
1828 }
1829}
1830
1831enum fnc_state { s_DIR, s_SLASH, s_BASE, s_TRAILING };
1832enum fnc_type { t_PATH, t_ITEM };
1833
1834/* Compare the names of two file_struct entities, similar to how strcmp()
1835 * would do if it were operating on the joined strings.
1836 *
1837 * Some differences beginning with protocol_version 29: (1) directory names
1838 * are compared with an assumed trailing slash so that they compare in a
1839 * way that would cause them to sort immediately prior to any content they
1840 * may have; (2) a directory of any name compares after a non-directory of
1841 * any name at the same depth; (3) a directory with name "." compares prior
1842 * to anything else. These changes mean that a directory and a non-dir
1843 * with the same name will not compare as equal (protocol_version >= 29).
1844 *
1845 * The dirname component can be an empty string, but the basename component
1846 * cannot (and never is in the current codebase). The basename component
1847 * may be NULL (for a removed item), in which case it is considered to be
1848 * after any existing item. */
1849int f_name_cmp(struct file_struct *f1, struct file_struct *f2)
1850{
1851 int dif;
1852 const uchar *c1, *c2;
1853 enum fnc_state state1, state2;
1854 enum fnc_type type1, type2;
1855 enum fnc_type t_path = protocol_version >= 29 ? t_PATH : t_ITEM;
1856
1857 if (!f1 || !F_IS_ACTIVE(f1)) {
1858 if (!f2 || !F_IS_ACTIVE(f2))
1859 return 0;
1860 return -1;
1861 }
1862 if (!f2 || !F_IS_ACTIVE(f2))
1863 return 1;
1864
1865 c1 = (uchar*)f1->dirname;
1866 c2 = (uchar*)f2->dirname;
1867 if (c1 == c2)
1868 c1 = c2 = NULL;
1869 if (!c1) {
1870 type1 = S_ISDIR(f1->mode) ? t_path : t_ITEM;
1871 c1 = (const uchar*)f1->basename;
1872 if (type1 == t_PATH && *c1 == '.' && !c1[1]) {
1873 type1 = t_ITEM;
1874 state1 = s_TRAILING;
1875 c1 = (uchar*)"";
1876 } else
1877 state1 = s_BASE;
1878 } else {
1879 type1 = t_path;
1880 state1 = s_DIR;
1881 }
1882 if (!c2) {
1883 type2 = S_ISDIR(f2->mode) ? t_path : t_ITEM;
1884 c2 = (const uchar*)f2->basename;
1885 if (type2 == t_PATH && *c2 == '.' && !c2[1]) {
1886 type2 = t_ITEM;
1887 state2 = s_TRAILING;
1888 c2 = (uchar*)"";
1889 } else
1890 state2 = s_BASE;
1891 } else {
1892 type2 = t_path;
1893 state2 = s_DIR;
1894 }
1895
1896 if (type1 != type2)
1897 return type1 == t_PATH ? 1 : -1;
1898
1899 do {
1900 if (!*c1) {
1901 switch (state1) {
1902 case s_DIR:
1903 state1 = s_SLASH;
1904 c1 = (uchar*)"/";
1905 break;
1906 case s_SLASH:
1907 type1 = S_ISDIR(f1->mode) ? t_path : t_ITEM;
1908 c1 = (const uchar*)f1->basename;
1909 if (type1 == t_PATH && *c1 == '.' && !c1[1]) {
1910 type1 = t_ITEM;
1911 state1 = s_TRAILING;
1912 c1 = (uchar*)"";
1913 } else
1914 state1 = s_BASE;
1915 break;
1916 case s_BASE:
1917 state1 = s_TRAILING;
1918 if (type1 == t_PATH) {
1919 c1 = (uchar*)"/";
1920 break;
1921 }
1922 /* FALL THROUGH */
1923 case s_TRAILING:
1924 type1 = t_ITEM;
1925 break;
1926 }
1927 if (*c2 && type1 != type2)
1928 return type1 == t_PATH ? 1 : -1;
1929 }
1930 if (!*c2) {
1931 switch (state2) {
1932 case s_DIR:
1933 state2 = s_SLASH;
1934 c2 = (uchar*)"/";
1935 break;
1936 case s_SLASH:
1937 type2 = S_ISDIR(f2->mode) ? t_path : t_ITEM;
1938 c2 = (const uchar*)f2->basename;
1939 if (type2 == t_PATH && *c2 == '.' && !c2[1]) {
1940 type2 = t_ITEM;
1941 state2 = s_TRAILING;
1942 c2 = (uchar*)"";
1943 } else
1944 state2 = s_BASE;
1945 break;
1946 case s_BASE:
1947 state2 = s_TRAILING;
1948 if (type2 == t_PATH) {
1949 c2 = (uchar*)"/";
1950 break;
1951 }
1952 /* FALL THROUGH */
1953 case s_TRAILING:
1954 if (!*c1)
1955 return 0;
1956 type2 = t_ITEM;
1957 break;
1958 }
1959 if (type1 != type2)
1960 return type1 == t_PATH ? 1 : -1;
1961 }
1962 } while ((dif = (int)*c1++ - (int)*c2++) == 0);
1963
1964 return dif;
1965}
1966
1967char *f_name_buf(void)
1968{
1969 static char names[5][MAXPATHLEN];
1970 static unsigned int n;
1971
1972 n = (n + 1) % (sizeof names / sizeof names[0]);
1973
1974 return names[n];
1975}
1976
1977/* Return a copy of the full filename of a flist entry, using the indicated
1978 * buffer or one of 5 static buffers if fbuf is NULL. No size-checking is
1979 * done because we checked the size when creating the file_struct entry.
1980 */
1981char *f_name(struct file_struct *f, char *fbuf)
1982{
1983 if (!f || !F_IS_ACTIVE(f))
1984 return NULL;
1985
1986 if (!fbuf)
1987 fbuf = f_name_buf();
1988
1989 if (f->dirname) {
1990 int len = strlen(f->dirname);
1991 memcpy(fbuf, f->dirname, len);
1992 fbuf[len] = '/';
1993 strlcpy(fbuf + len + 1, f->basename, MAXPATHLEN - (len + 1));
1994 } else
1995 strlcpy(fbuf, f->basename, MAXPATHLEN);
1996
1997 return fbuf;
1998}
1999
2000/* Do a non-recursive scan of the named directory, possibly ignoring all
2001 * exclude rules except for the daemon's. If "dlen" is >=0, it is the length
2002 * of the dirname string, and also indicates that "dirname" is a MAXPATHLEN
2003 * buffer (the functions we call will append names onto the end, but the old
2004 * dir value will be restored on exit). */
2005struct file_list *get_dirlist(char *dirname, int dlen, int ignore_filter_rules)
2006{
2007 struct file_list *dirlist;
2008 char dirbuf[MAXPATHLEN];
2009 int save_recurse = recurse;
2010 int save_xfer_dirs = xfer_dirs;
2011
2012 if (dlen < 0) {
2013 dlen = strlcpy(dirbuf, dirname, MAXPATHLEN);
2014 if (dlen >= MAXPATHLEN)
2015 return NULL;
2016 dirname = dirbuf;
2017 }
2018
2019 dirlist = flist_new("get_dirlist");
2020
2021 recurse = 0;
2022 xfer_dirs = 1;
2023 send_directory(ignore_filter_rules ? -2 : -1, dirlist, dirname, dlen);
2024 xfer_dirs = save_xfer_dirs;
2025 recurse = save_recurse;
2026 if (do_progress)
2027 flist_count_offset += dirlist->count;
2028
2029 clean_flist(dirlist, 0, 0);
2030
2031 if (verbose > 3)
2032 output_flist(dirlist);
2033
2034 return dirlist;
2035}