Get rid of some signed/unsigned-comparison compiler warnings.
[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 flist_extra_cnt;
52extern int ignore_perishable;
53extern int non_perishable_cnt;
54extern int prune_empty_dirs;
55extern int copy_links;
56extern int copy_unsafe_links;
57extern int protocol_version;
58extern int sanitize_paths;
59extern struct stats stats;
60extern struct file_list *the_file_list;
61
62extern char curr_dir[MAXPATHLEN];
63
64extern struct chmod_mode_struct *chmod_modes;
65
66extern struct filter_list_struct filter_list;
67extern struct filter_list_struct server_filter_list;
68
69int io_error;
70int checksum_len;
71dev_t filesystem_dev; /* used to implement -x */
72
73/* The tmp_* vars are used as a cache area by make_file() to store data
74 * that the sender doesn't need to remember in its file list. The data
75 * will survive just long enough to be used by send_file_entry(). */
76static dev_t tmp_rdev;
77static struct idev tmp_idev;
78static char tmp_sum[MD4_SUM_LENGTH];
79
80static char empty_sum[MD4_SUM_LENGTH];
81static int flist_count_offset;
82
83static void clean_flist(struct file_list *flist, int strip_root, int no_dups);
84static void output_flist(struct file_list *flist);
85
86void init_flist(void)
87{
88 if (verbose > 4) {
89 rprintf(FINFO, "FILE_STRUCT_LEN=%d, EXTRA_LEN=%d\n",
90 FILE_STRUCT_LEN, EXTRA_LEN);
91 }
92 checksum_len = protocol_version < 21 ? 2 : MD4_SUM_LENGTH;
93}
94
95static int show_filelist_p(void)
96{
97 return verbose && xfer_dirs && !am_server;
98}
99
100static void start_filelist_progress(char *kind)
101{
102 rprintf(FCLIENT, "%s ... ", kind);
103 if (verbose > 1 || do_progress)
104 rprintf(FCLIENT, "\n");
105 rflush(FINFO);
106}
107
108static void emit_filelist_progress(int count)
109{
110 rprintf(FCLIENT, " %d files...\r", count);
111}
112
113static void maybe_emit_filelist_progress(int count)
114{
115 if (do_progress && show_filelist_p() && (count % 100) == 0)
116 emit_filelist_progress(count);
117}
118
119static void finish_filelist_progress(const struct file_list *flist)
120{
121 if (do_progress) {
122 /* This overwrites the progress line */
123 rprintf(FINFO, "%d file%sto consider\n",
124 flist->count, flist->count == 1 ? " " : "s ");
125 } else
126 rprintf(FINFO, "done\n");
127}
128
129void show_flist_stats(void)
130{
131 /* Nothing yet */
132}
133
134static void list_file_entry(struct file_struct *f)
135{
136 char permbuf[PERMSTRING_SIZE];
137 double len;
138
139 if (!F_IS_ACTIVE(f)) {
140 /* this can happen if duplicate names were removed */
141 return;
142 }
143
144 permstring(permbuf, f->mode);
145 len = F_LENGTH(f);
146
147#ifdef SUPPORT_LINKS
148 if (preserve_links && S_ISLNK(f->mode)) {
149 rprintf(FINFO, "%s %11.0f %s %s -> %s\n",
150 permbuf, len, timestring(f->modtime),
151 f_name(f, NULL), F_SYMLINK(f));
152 } else
153#endif
154 {
155 rprintf(FINFO, "%s %11.0f %s %s\n",
156 permbuf, len, timestring(f->modtime),
157 f_name(f, NULL));
158 }
159}
160
161/* Stat either a symlink or its referent, depending on the settings of
162 * copy_links, copy_unsafe_links, etc. Returns -1 on error, 0 on success.
163 *
164 * If path is the name of a symlink, then the linkbuf buffer (which must hold
165 * MAXPATHLEN chars) will be set to the symlink's target string.
166 *
167 * The stat structure pointed to by stp will contain information about the
168 * link or the referent as appropriate, if they exist. */
169static int readlink_stat(const char *path, STRUCT_STAT *stp, char *linkbuf)
170{
171#ifdef SUPPORT_LINKS
172 if (link_stat(path, stp, copy_dirlinks) < 0)
173 return -1;
174 if (S_ISLNK(stp->st_mode)) {
175 int llen = readlink(path, linkbuf, MAXPATHLEN - 1);
176 if (llen < 0)
177 return -1;
178 linkbuf[llen] = '\0';
179 if (copy_unsafe_links && unsafe_symlink(linkbuf, path)) {
180 if (verbose > 1) {
181 rprintf(FINFO,"copying unsafe symlink \"%s\" -> \"%s\"\n",
182 path, linkbuf);
183 }
184 return do_stat(path, stp);
185 }
186 }
187 return 0;
188#else
189 return do_stat(path, stp);
190#endif
191}
192
193int link_stat(const char *path, STRUCT_STAT *stp, int follow_dirlinks)
194{
195#ifdef SUPPORT_LINKS
196 if (copy_links)
197 return do_stat(path, stp);
198 if (do_lstat(path, stp) < 0)
199 return -1;
200 if (follow_dirlinks && S_ISLNK(stp->st_mode)) {
201 STRUCT_STAT st;
202 if (do_stat(path, &st) == 0 && S_ISDIR(st.st_mode))
203 *stp = st;
204 }
205 return 0;
206#else
207 return do_stat(path, stp);
208#endif
209}
210
211/* This function is used to check if a file should be included/excluded
212 * from the list of files based on its name and type etc. The value of
213 * filter_level is set to either SERVER_FILTERS or ALL_FILTERS. */
214static int is_excluded(char *fname, int is_dir, int filter_level)
215{
216#if 0 /* This currently never happens, so avoid a useless compare. */
217 if (filter_level == NO_FILTERS)
218 return 0;
219#endif
220 if (fname) {
221 /* never exclude '.', even if somebody does --exclude '*' */
222 if (fname[0] == '.' && !fname[1])
223 return 0;
224 /* Handle the -R version of the '.' dir. */
225 if (fname[0] == '/') {
226 int len = strlen(fname);
227 if (fname[len-1] == '.' && fname[len-2] == '/')
228 return 0;
229 }
230 }
231 if (server_filter_list.head
232 && check_filter(&server_filter_list, fname, is_dir) < 0)
233 return 1;
234 if (filter_level != ALL_FILTERS)
235 return 0;
236 if (filter_list.head
237 && check_filter(&filter_list, fname, is_dir) < 0)
238 return 1;
239 return 0;
240}
241
242static int to_wire_mode(mode_t mode)
243{
244#ifdef SUPPORT_LINKS
245#if _S_IFLNK != 0120000
246 if (S_ISLNK(mode))
247 return (mode & ~(_S_IFMT)) | 0120000;
248#endif
249#endif
250 return mode;
251}
252
253static mode_t from_wire_mode(int mode)
254{
255#if _S_IFLNK != 0120000
256 if ((mode & (_S_IFMT)) == 0120000)
257 return (mode & ~(_S_IFMT)) | _S_IFLNK;
258#endif
259 return mode;
260}
261
262static void send_directory(int f, struct file_list *flist,
263 char *fbuf, int len);
264
265static const char *flist_dir;
266static int flist_dir_len;
267
268
269/**
270 * Make sure @p flist is big enough to hold at least @p flist->count
271 * entries.
272 **/
273void flist_expand(struct file_list *flist)
274{
275 struct file_struct **new_ptr;
276
277 if (flist->count < flist->malloced)
278 return;
279
280 if (flist->malloced < FLIST_START)
281 flist->malloced = FLIST_START;
282 else if (flist->malloced >= FLIST_LINEAR)
283 flist->malloced += FLIST_LINEAR;
284 else
285 flist->malloced *= 2;
286
287 /*
288 * In case count jumped or we are starting the list
289 * with a known size just set it.
290 */
291 if (flist->malloced < flist->count)
292 flist->malloced = flist->count;
293
294 new_ptr = realloc_array(flist->files, struct file_struct *,
295 flist->malloced);
296
297 if (verbose >= 2 && flist->malloced != FLIST_START) {
298 rprintf(FCLIENT, "[%s] expand file_list to %.0f bytes, did%s move\n",
299 who_am_i(),
300 (double)sizeof flist->files[0] * flist->malloced,
301 (new_ptr == flist->files) ? " not" : "");
302 }
303
304 flist->files = new_ptr;
305
306 if (!flist->files)
307 out_of_memory("flist_expand");
308}
309
310static void send_file_entry(struct file_struct *file, int f)
311{
312 unsigned short flags;
313 static time_t modtime;
314 static mode_t mode;
315 static int64 dev;
316 static dev_t rdev;
317 static uint32 rdev_major;
318 static uid_t uid;
319 static gid_t gid;
320 static char lastname[MAXPATHLEN];
321 char fname[MAXPATHLEN];
322 int l1, l2;
323
324 if (f < 0)
325 return;
326
327 if (!file) {
328 write_byte(f, 0);
329 modtime = 0, mode = 0;
330 dev = 0, rdev = MAKEDEV(0, 0);
331 rdev_major = 0;
332 uid = 0, gid = 0;
333 *lastname = '\0';
334 return;
335 }
336
337 f_name(file, fname);
338
339 flags = file->flags & FLAG_TOP_DIR; /* FLAG_TOP_DIR == XMIT_TOP_DIR */
340
341 if (file->mode == mode)
342 flags |= XMIT_SAME_MODE;
343 else
344 mode = file->mode;
345 if ((preserve_devices && IS_DEVICE(mode))
346 || (preserve_specials && IS_SPECIAL(mode))) {
347 if (protocol_version < 28) {
348 if (tmp_rdev == rdev)
349 flags |= XMIT_SAME_RDEV_pre28;
350 else
351 rdev = tmp_rdev;
352 } else {
353 rdev = tmp_rdev;
354 if ((uint32)major(rdev) == rdev_major)
355 flags |= XMIT_SAME_RDEV_MAJOR;
356 else
357 rdev_major = major(rdev);
358 if ((uint32)minor(rdev) <= 0xFFu)
359 flags |= XMIT_RDEV_MINOR_IS_SMALL;
360 }
361 } else if (protocol_version < 28)
362 rdev = MAKEDEV(0, 0);
363 if (preserve_uid) {
364 if (F_UID(file) == uid)
365 flags |= XMIT_SAME_UID;
366 else
367 uid = F_UID(file);
368 }
369 if (preserve_gid) {
370 if (F_GID(file) == gid)
371 flags |= XMIT_SAME_GID;
372 else
373 gid = F_GID(file);
374 }
375 if (file->modtime == modtime)
376 flags |= XMIT_SAME_TIME;
377 else
378 modtime = file->modtime;
379
380#ifdef SUPPORT_HARD_LINKS
381 if (tmp_idev.dev != 0) {
382 if (tmp_idev.dev == dev) {
383 if (protocol_version >= 28)
384 flags |= XMIT_SAME_DEV;
385 } else
386 dev = tmp_idev.dev;
387 flags |= XMIT_HAS_IDEV_DATA;
388 }
389#endif
390
391 for (l1 = 0;
392 lastname[l1] && (fname[l1] == lastname[l1]) && (l1 < 255);
393 l1++) {}
394 l2 = strlen(fname+l1);
395
396 if (l1 > 0)
397 flags |= XMIT_SAME_NAME;
398 if (l2 > 255)
399 flags |= XMIT_LONG_NAME;
400
401 /* We must make sure we don't send a zero flag byte or the
402 * other end will terminate the flist transfer. Note that
403 * the use of XMIT_TOP_DIR on a non-dir has no meaning, so
404 * it's harmless way to add a bit to the first flag byte. */
405 if (protocol_version >= 28) {
406 if (!flags && !S_ISDIR(mode))
407 flags |= XMIT_TOP_DIR;
408 if ((flags & 0xFF00) || !flags) {
409 flags |= XMIT_EXTENDED_FLAGS;
410 write_shortint(f, flags);
411 } else
412 write_byte(f, flags);
413 } else {
414 if (!(flags & 0xFF))
415 flags |= S_ISDIR(mode) ? XMIT_LONG_NAME : XMIT_TOP_DIR;
416 write_byte(f, flags);
417 }
418 if (flags & XMIT_SAME_NAME)
419 write_byte(f, l1);
420 if (flags & XMIT_LONG_NAME)
421 write_int(f, l2);
422 else
423 write_byte(f, l2);
424 write_buf(f, fname + l1, l2);
425
426 write_longint(f, F_LENGTH(file));
427 if (!(flags & XMIT_SAME_TIME))
428 write_int(f, modtime);
429 if (!(flags & XMIT_SAME_MODE))
430 write_int(f, to_wire_mode(mode));
431 if (preserve_uid && !(flags & XMIT_SAME_UID)) {
432 if (!numeric_ids)
433 add_uid(uid);
434 write_int(f, uid);
435 }
436 if (preserve_gid && !(flags & XMIT_SAME_GID)) {
437 if (!numeric_ids)
438 add_gid(gid);
439 write_int(f, gid);
440 }
441 if ((preserve_devices && IS_DEVICE(mode))
442 || (preserve_specials && IS_SPECIAL(mode))) {
443 if (protocol_version < 28) {
444 if (!(flags & XMIT_SAME_RDEV_pre28))
445 write_int(f, (int)rdev);
446 } else {
447 if (!(flags & XMIT_SAME_RDEV_MAJOR))
448 write_int(f, major(rdev));
449 if (flags & XMIT_RDEV_MINOR_IS_SMALL)
450 write_byte(f, minor(rdev));
451 else
452 write_int(f, minor(rdev));
453 }
454 }
455
456#ifdef SUPPORT_LINKS
457 if (preserve_links && S_ISLNK(mode)) {
458 const char *sl = F_SYMLINK(file);
459 int len = strlen(sl);
460 write_int(f, len);
461 write_buf(f, sl, len);
462 }
463#endif
464
465#ifdef SUPPORT_HARD_LINKS
466 if (tmp_idev.dev != 0) {
467 if (protocol_version < 26) {
468 /* 32-bit dev_t and ino_t */
469 write_int(f, (int32)dev);
470 write_int(f, (int32)tmp_idev.ino);
471 } else {
472 /* 64-bit dev_t and ino_t */
473 if (!(flags & XMIT_SAME_DEV))
474 write_longint(f, dev);
475 write_longint(f, tmp_idev.ino);
476 }
477 }
478#endif
479
480 if (always_checksum && (S_ISREG(mode) || protocol_version < 28)) {
481 const char *sum;
482 if (S_ISREG(mode))
483 sum = tmp_sum;
484 else {
485 /* Prior to 28, we sent a useless set of nulls. */
486 sum = empty_sum;
487 }
488 write_buf(f, sum, checksum_len);
489 }
490
491 strlcpy(lastname, fname, MAXPATHLEN);
492}
493
494static struct file_struct *recv_file_entry(struct file_list *flist,
495 unsigned short flags, int f)
496{
497 static time_t modtime;
498 static mode_t mode;
499 static int64 dev;
500 static dev_t rdev;
501 static uint32 rdev_major;
502 static uid_t uid;
503 static gid_t gid;
504 static char lastname[MAXPATHLEN], *lastdir;
505 static int lastdir_depth, lastdir_len = -1;
506 static unsigned int del_hier_name_len = 0;
507 static int in_del_hier = 0;
508 char thisname[MAXPATHLEN];
509 unsigned int l1 = 0, l2 = 0;
510 int alloc_len, basename_len, dirname_len, linkname_len;
511 int extra_len = (flist_extra_cnt - 1) * EXTRA_LEN;
512 OFF_T file_length;
513 char *basename, *dirname, *bp;
514 struct file_struct *file;
515
516 if (!flist) {
517 modtime = 0, mode = 0;
518 dev = 0, rdev = MAKEDEV(0, 0);
519 rdev_major = 0;
520 uid = 0, gid = 0;
521 *lastname = '\0';
522 lastdir_len = -1;
523 in_del_hier = 0;
524 return NULL;
525 }
526
527 if (flags & XMIT_SAME_NAME)
528 l1 = read_byte(f);
529
530 if (flags & XMIT_LONG_NAME)
531 l2 = read_int(f);
532 else
533 l2 = read_byte(f);
534
535 if (l2 >= MAXPATHLEN - l1) {
536 rprintf(FERROR,
537 "overflow: flags=0x%x l1=%d l2=%d lastname=%s [%s]\n",
538 flags, l1, l2, lastname, who_am_i());
539 overflow_exit("recv_file_entry");
540 }
541
542 strlcpy(thisname, lastname, l1 + 1);
543 read_sbuf(f, &thisname[l1], l2);
544 thisname[l1 + l2] = 0;
545
546 strlcpy(lastname, thisname, MAXPATHLEN);
547
548 clean_fname(thisname, 0);
549
550 if (sanitize_paths)
551 sanitize_path(thisname, thisname, "", 0, NULL);
552
553 if ((basename = strrchr(thisname, '/')) != NULL) {
554 dirname_len = ++basename - thisname; /* counts future '\0' */
555 if (lastdir_len == dirname_len - 1
556 && strncmp(thisname, lastdir, lastdir_len) == 0) {
557 dirname = lastdir;
558 dirname_len = 0; /* indicates no copy is needed */
559 } else
560 dirname = thisname;
561 } else {
562 basename = thisname;
563 dirname = NULL;
564 dirname_len = 0;
565 }
566 basename_len = strlen(basename) + 1; /* count the '\0' */
567
568 file_length = read_longint(f);
569 if (!(flags & XMIT_SAME_TIME))
570 modtime = (time_t)read_int(f);
571 if (!(flags & XMIT_SAME_MODE))
572 mode = from_wire_mode(read_int(f));
573
574 if (chmod_modes && !S_ISLNK(mode))
575 mode = tweak_mode(mode, chmod_modes);
576
577 if (preserve_uid && !(flags & XMIT_SAME_UID))
578 uid = (uid_t)read_int(f);
579 if (preserve_gid && !(flags & XMIT_SAME_GID))
580 gid = (gid_t)read_int(f);
581
582 if ((preserve_devices && IS_DEVICE(mode))
583 || (preserve_specials && IS_SPECIAL(mode))) {
584 if (protocol_version < 28) {
585 if (!(flags & XMIT_SAME_RDEV_pre28))
586 rdev = (dev_t)read_int(f);
587 } else {
588 uint32 rdev_minor;
589 if (!(flags & XMIT_SAME_RDEV_MAJOR))
590 rdev_major = read_int(f);
591 if (flags & XMIT_RDEV_MINOR_IS_SMALL)
592 rdev_minor = read_byte(f);
593 else
594 rdev_minor = read_int(f);
595 rdev = MAKEDEV(rdev_major, rdev_minor);
596 }
597 extra_len += 2 * EXTRA_LEN;
598 file_length = 0;
599 } else if (protocol_version < 28)
600 rdev = MAKEDEV(0, 0);
601
602#ifdef SUPPORT_LINKS
603 if (preserve_links && S_ISLNK(mode)) {
604 linkname_len = read_int(f) + 1; /* count the '\0' */
605 if (linkname_len <= 0 || linkname_len > MAXPATHLEN) {
606 rprintf(FERROR, "overflow: linkname_len=%d\n",
607 linkname_len - 1);
608 overflow_exit("recv_file_entry");
609 }
610 }
611 else
612#endif
613 linkname_len = 0;
614
615#ifdef SUPPORT_HARD_LINKS
616 if (preserve_hard_links) {
617 if (protocol_version < 28 && S_ISREG(mode))
618 flags |= XMIT_HAS_IDEV_DATA;
619 if (flags & XMIT_HAS_IDEV_DATA)
620 extra_len += EXTRA_LEN;
621 }
622#endif
623
624 if (always_checksum && S_ISREG(mode))
625 extra_len += SUM_EXTRA_CNT * EXTRA_LEN;
626
627 if (file_length > 0xFFFFFFFFu && S_ISREG(mode))
628 extra_len += EXTRA_LEN;
629
630 alloc_len = FILE_STRUCT_LEN + extra_len + basename_len + dirname_len
631 + linkname_len;
632 bp = pool_alloc(flist->file_pool, alloc_len, "recv_file_entry");
633
634 memset(bp, 0, FILE_STRUCT_LEN + extra_len);
635 bp += extra_len;
636 file = (struct file_struct *)bp;
637 bp += FILE_STRUCT_LEN;
638
639 memcpy(bp, basename, basename_len);
640 bp += basename_len + linkname_len; /* skip space for symlink too */
641
642#ifdef SUPPORT_HARD_LINKS
643 if (flags & XMIT_HAS_IDEV_DATA)
644 file->flags |= FLAG_HLINKED;
645#endif
646 file->modtime = modtime;
647 file->len32 = file_length;
648 if (file_length > 0xFFFFFFFFu && S_ISREG(mode)) {
649 file->flags |= FLAG_LENGTH64;
650 OPT_EXTRA(file, 0)->unum = (uint32)(file_length >> 32);
651 }
652 file->mode = mode;
653 if (preserve_uid)
654 F_UID(file) = uid;
655 if (preserve_gid)
656 F_GID(file) = gid;
657
658 if (dirname_len) {
659 file->dirname = lastdir = bp;
660 lastdir_len = dirname_len - 1;
661 memcpy(bp, dirname, dirname_len - 1);
662 bp += dirname_len;
663 bp[-1] = '\0';
664 lastdir_depth = count_dir_elements(lastdir);
665 file->dir.depth = lastdir_depth + 1;
666 } else if (dirname) {
667 file->dirname = dirname; /* we're reusing lastname */
668 file->dir.depth = lastdir_depth + 1;
669 } else
670 file->dir.depth = 1;
671
672 if (S_ISDIR(mode)) {
673 if (basename_len == 1+1 && *basename == '.') /* +1 for '\0' */
674 file->dir.depth--;
675 if (flags & XMIT_TOP_DIR) {
676 in_del_hier = recurse;
677 del_hier_name_len = file->dir.depth == 0 ? 0 : l1 + l2;
678 if (relative_paths && del_hier_name_len > 2
679 && lastname[del_hier_name_len-1] == '.'
680 && lastname[del_hier_name_len-2] == '/')
681 del_hier_name_len -= 2;
682 file->flags |= FLAG_TOP_DIR | FLAG_XFER_DIR;
683 } else if (in_del_hier) {
684 if (!relative_paths || !del_hier_name_len
685 || (l1 >= del_hier_name_len
686 && lastname[del_hier_name_len] == '/'))
687 file->flags |= FLAG_XFER_DIR;
688 else
689 in_del_hier = 0;
690 }
691 }
692
693 if ((preserve_devices && IS_DEVICE(mode))
694 || (preserve_specials && IS_SPECIAL(mode))) {
695 uint32 *devp = F_RDEV_P(file);
696 DEV_MAJOR(devp) = major(rdev);
697 DEV_MINOR(devp) = minor(rdev);
698 }
699
700#ifdef SUPPORT_LINKS
701 if (linkname_len) {
702 bp = (char*)F_BASENAME(file) + basename_len;
703 read_sbuf(f, bp, linkname_len - 1);
704 if (sanitize_paths)
705 sanitize_path(bp, bp, "", lastdir_depth, NULL);
706 }
707#endif
708
709#ifdef SUPPORT_HARD_LINKS
710 if (preserve_hard_links && flags & XMIT_HAS_IDEV_DATA) {
711 struct idev *idevp = pool_talloc(flist->hlink_pool, struct idev,
712 1, "inode_table");
713 F_HL_IDEV(file) = idevp;
714 if (protocol_version < 26) {
715 idevp->dev = read_int(f);
716 idevp->ino = read_int(f);
717 } else {
718 if (!(flags & XMIT_SAME_DEV))
719 dev = read_longint(f);
720 idevp->dev = dev;
721 idevp->ino = read_longint(f);
722 }
723 }
724#endif
725
726 if (always_checksum && (S_ISREG(mode) || protocol_version < 28)) {
727 if (S_ISREG(mode))
728 bp = (char*)F_SUM(file);
729 else {
730 /* Prior to 28, we get a useless set of nulls. */
731 bp = tmp_sum;
732 }
733 read_buf(f, bp, checksum_len);
734 }
735
736 return file;
737}
738
739/**
740 * Create a file_struct for a named file by reading its stat()
741 * information and performing extensive checks against global
742 * options.
743 *
744 * @return the new file, or NULL if there was an error or this file
745 * should be excluded.
746 *
747 * @todo There is a small optimization opportunity here to avoid
748 * stat()ing the file in some circumstances, which has a certain cost.
749 * We are called immediately after doing readdir(), and so we may
750 * already know the d_type of the file. We could for example avoid
751 * statting directories if we're not recursing, but this is not a very
752 * important case. Some systems may not have d_type.
753 **/
754struct file_struct *make_file(const char *fname, struct file_list *flist,
755 STRUCT_STAT *stp, int flags, int filter_level)
756{
757 static char *lastdir;
758 static int lastdir_len = -1;
759 struct file_struct *file;
760 STRUCT_STAT st;
761 char thisname[MAXPATHLEN];
762 char linkname[MAXPATHLEN];
763 int alloc_len, basename_len, dirname_len, linkname_len;
764 int extra_len = (flist_extra_cnt - 1) * EXTRA_LEN;
765 char *basename, *dirname, *bp;
766
767 if (!flist || !flist->count) /* Ignore lastdir when invalid. */
768 lastdir_len = -1;
769
770 if (strlcpy(thisname, fname, sizeof thisname)
771 >= sizeof thisname - flist_dir_len) {
772 rprintf(FINFO, "skipping overly long name: %s\n", fname);
773 return NULL;
774 }
775 clean_fname(thisname, 0);
776 if (sanitize_paths)
777 sanitize_path(thisname, thisname, "", 0, NULL);
778
779 if (stp && S_ISDIR(stp->st_mode)) {
780 st = *stp; /* Needed for "symlink/." with --relative. */
781 *linkname = '\0'; /* make IBM code checker happy */
782 } else if (readlink_stat(thisname, &st, linkname) != 0) {
783 int save_errno = errno;
784 /* See if file is excluded before reporting an error. */
785 if (filter_level != NO_FILTERS
786 && (is_excluded(thisname, 0, filter_level)
787 || is_excluded(thisname, 1, filter_level))) {
788 if (ignore_perishable && save_errno != ENOENT)
789 non_perishable_cnt++;
790 return NULL;
791 }
792 if (save_errno == ENOENT) {
793#ifdef SUPPORT_LINKS
794 /* Avoid "vanished" error if symlink points nowhere. */
795 if (copy_links && do_lstat(thisname, &st) == 0
796 && S_ISLNK(st.st_mode)) {
797 io_error |= IOERR_GENERAL;
798 rprintf(FERROR, "symlink has no referent: %s\n",
799 full_fname(thisname));
800 } else
801#endif
802 {
803 enum logcode c = am_daemon && protocol_version < 28
804 ? FERROR : FINFO;
805 io_error |= IOERR_VANISHED;
806 rprintf(c, "file has vanished: %s\n",
807 full_fname(thisname));
808 }
809 } else {
810 io_error |= IOERR_GENERAL;
811 rsyserr(FERROR, save_errno, "readlink %s failed",
812 full_fname(thisname));
813 }
814 return NULL;
815 }
816
817 /* backup.c calls us with filter_level set to NO_FILTERS. */
818 if (filter_level == NO_FILTERS)
819 goto skip_filters;
820
821 if (S_ISDIR(st.st_mode) && !xfer_dirs) {
822 rprintf(FINFO, "skipping directory %s\n", thisname);
823 return NULL;
824 }
825
826 /* -x only affects directories because we need to avoid recursing
827 * into a mount-point directory, not to avoid copying a symlinked
828 * file if -L (or similar) was specified. */
829 if (one_file_system && st.st_dev != filesystem_dev
830 && S_ISDIR(st.st_mode)) {
831 if (one_file_system > 1) {
832 if (verbose > 2) {
833 rprintf(FINFO, "skipping mount-point dir %s\n",
834 thisname);
835 }
836 return NULL;
837 }
838 flags |= FLAG_MOUNT_DIR;
839 }
840
841 if (is_excluded(thisname, S_ISDIR(st.st_mode) != 0, filter_level)) {
842 if (ignore_perishable)
843 non_perishable_cnt++;
844 return NULL;
845 }
846
847 if (lp_ignore_nonreadable(module_id)) {
848#ifdef SUPPORT_LINKS
849 if (!S_ISLNK(st.st_mode))
850#endif
851 if (access(thisname, R_OK) != 0)
852 return NULL;
853 }
854
855 skip_filters:
856
857 if (verbose > 2) {
858 rprintf(FINFO, "[%s] make_file(%s,*,%d)\n",
859 who_am_i(), thisname, filter_level);
860 }
861
862 if ((basename = strrchr(thisname, '/')) != NULL) {
863 dirname_len = ++basename - thisname; /* counts future '\0' */
864 if (lastdir_len == dirname_len - 1
865 && strncmp(thisname, lastdir, lastdir_len) == 0) {
866 dirname = lastdir;
867 dirname_len = 0; /* indicates no copy is needed */
868 } else
869 dirname = thisname;
870 } else {
871 basename = thisname;
872 dirname = NULL;
873 dirname_len = 0;
874 }
875 basename_len = strlen(basename) + 1; /* count the '\0' */
876
877#ifdef SUPPORT_LINKS
878 linkname_len = S_ISLNK(st.st_mode) ? strlen(linkname) + 1 : 0;
879#else
880 linkname_len = 0;
881#endif
882
883 if (st.st_size > 0xFFFFFFFFu && S_ISREG(st.st_mode))
884 extra_len += EXTRA_LEN;
885
886 alloc_len = FILE_STRUCT_LEN + extra_len + basename_len + dirname_len
887 + linkname_len;
888 if (flist)
889 bp = pool_alloc(flist->file_pool, alloc_len, "make_file");
890 else {
891 if (!(bp = new_array(char, alloc_len)))
892 out_of_memory("make_file");
893 }
894
895 memset(bp, 0, FILE_STRUCT_LEN + extra_len);
896 bp += extra_len;
897 file = (struct file_struct *)bp;
898 bp += FILE_STRUCT_LEN;
899
900 memcpy(bp, basename, basename_len);
901 bp += basename_len + linkname_len; /* skip space for symlink too */
902
903#ifdef SUPPORT_HARD_LINKS
904 if (preserve_hard_links && flist) {
905 if (protocol_version >= 28
906 ? (!S_ISDIR(st.st_mode) && st.st_nlink > 1)
907 : S_ISREG(st.st_mode)) {
908 tmp_idev.dev = st.st_dev;
909 tmp_idev.ino = st.st_ino;
910 } else
911 tmp_idev.dev = tmp_idev.ino = 0;
912 }
913#endif
914
915#ifdef HAVE_STRUCT_STAT_ST_RDEV
916 if (IS_DEVICE(st.st_mode) || IS_SPECIAL(st.st_mode)) {
917 tmp_rdev = st.st_rdev;
918 st.st_size = 0;
919 }
920#endif
921
922 file->flags = flags;
923 file->modtime = st.st_mtime;
924 file->len32 = st.st_size;
925 if (st.st_size > 0xFFFFFFFFu && S_ISREG(st.st_mode)) {
926 file->flags |= FLAG_LENGTH64;
927 OPT_EXTRA(file, 0)->unum = (uint32)(st.st_size >> 32);
928 }
929 file->mode = st.st_mode;
930 if (preserve_uid)
931 F_UID(file) = st.st_uid;
932 if (preserve_gid)
933 F_GID(file) = st.st_gid;
934
935 if (dirname_len) {
936 file->dirname = lastdir = bp;
937 lastdir_len = dirname_len - 1;
938 memcpy(bp, dirname, dirname_len - 1);
939 bp += dirname_len;
940 bp[-1] = '\0';
941 } else if (dirname)
942 file->dirname = dirname;
943
944#ifdef SUPPORT_LINKS
945 if (linkname_len) {
946 bp = (char*)F_BASENAME(file) + basename_len;
947 memcpy(bp, linkname, linkname_len);
948 }
949#endif
950
951 if (always_checksum && am_sender && S_ISREG(st.st_mode))
952 file_checksum(thisname, tmp_sum, st.st_size);
953
954 file->dir.root = flist_dir;
955
956 /* This code is only used by the receiver when it is building
957 * a list of files for a delete pass. */
958 if (keep_dirlinks && linkname_len && flist) {
959 STRUCT_STAT st2;
960 int save_mode = file->mode;
961 file->mode = S_IFDIR; /* Find a directory with our name. */
962 if (flist_find(the_file_list, file) >= 0
963 && do_stat(thisname, &st2) == 0 && S_ISDIR(st2.st_mode)) {
964 file->modtime = st2.st_mtime;
965 file->len32 = 0;
966 file->mode = st2.st_mode;
967 if (preserve_uid)
968 F_UID(file) = st2.st_uid;
969 if (preserve_gid)
970 F_GID(file) = st2.st_gid;
971 } else
972 file->mode = save_mode;
973 }
974
975 if (S_ISREG(st.st_mode) || S_ISLNK(st.st_mode))
976 stats.total_size += st.st_size;
977
978 if (basename_len == 0+1)
979 return NULL;
980
981 return file;
982}
983
984/* Only called for temporary file_struct entries created by make_file(). */
985void unmake_file(struct file_struct *file)
986{
987 int extra_cnt = flist_extra_cnt - 1 + LEN64_BUMP(file);
988 free(file->extras - extra_cnt);
989}
990
991static struct file_struct *send_file_name(int f, struct file_list *flist,
992 char *fname, STRUCT_STAT *stp,
993 unsigned short flags)
994{
995 struct file_struct *file;
996
997 file = make_file(fname, flist, stp, flags,
998 f == -2 ? SERVER_FILTERS : ALL_FILTERS);
999 if (!file)
1000 return NULL;
1001
1002 if (chmod_modes && !S_ISLNK(file->mode))
1003 file->mode = tweak_mode(file->mode, chmod_modes);
1004
1005 maybe_emit_filelist_progress(flist->count + flist_count_offset);
1006
1007 flist_expand(flist);
1008 flist->files[flist->count++] = file;
1009 send_file_entry(file, f);
1010 return file;
1011}
1012
1013static void send_if_directory(int f, struct file_list *flist,
1014 struct file_struct *file,
1015 char *fbuf, unsigned int ol)
1016{
1017 char is_dot_dir = fbuf[ol-1] == '.' && (ol == 1 || fbuf[ol-2] == '/');
1018
1019 if (S_ISDIR(file->mode)
1020 && !(file->flags & FLAG_MOUNT_DIR) && f_name(file, fbuf)) {
1021 void *save_filters;
1022 unsigned int len = strlen(fbuf);
1023 if (len > 1 && fbuf[len-1] == '/')
1024 fbuf[--len] = '\0';
1025 if (len >= MAXPATHLEN - 1) {
1026 io_error |= IOERR_GENERAL;
1027 rprintf(FERROR, "skipping long-named directory: %s\n",
1028 full_fname(fbuf));
1029 return;
1030 }
1031 save_filters = push_local_filters(fbuf, len);
1032 send_directory(f, flist, fbuf, len);
1033 pop_local_filters(save_filters);
1034 fbuf[ol] = '\0';
1035 if (is_dot_dir)
1036 fbuf[ol-1] = '.';
1037 }
1038}
1039
1040/* This function is normally called by the sender, but the receiving side also
1041 * calls it from get_dirlist() with f set to -1 so that we just construct the
1042 * file list in memory without sending it over the wire. Also, get_dirlist()
1043 * might call this with f set to -2, which also indicates that local filter
1044 * rules should be ignored. */
1045static void send_directory(int f, struct file_list *flist,
1046 char *fbuf, int len)
1047{
1048 struct dirent *di;
1049 unsigned remainder;
1050 char *p;
1051 DIR *d;
1052 int start = flist->count;
1053
1054 if (!(d = opendir(fbuf))) {
1055 io_error |= IOERR_GENERAL;
1056 rsyserr(FERROR, errno, "opendir %s failed", full_fname(fbuf));
1057 return;
1058 }
1059
1060 p = fbuf + len;
1061 if (len != 1 || *fbuf != '/')
1062 *p++ = '/';
1063 *p = '\0';
1064 remainder = MAXPATHLEN - (p - fbuf);
1065
1066 for (errno = 0, di = readdir(d); di; errno = 0, di = readdir(d)) {
1067 char *dname = d_name(di);
1068 if (dname[0] == '.' && (dname[1] == '\0'
1069 || (dname[1] == '.' && dname[2] == '\0')))
1070 continue;
1071 if (strlcpy(p, dname, remainder) >= remainder) {
1072 io_error |= IOERR_GENERAL;
1073 rprintf(FINFO,
1074 "cannot send long-named file %s\n",
1075 full_fname(fbuf));
1076 continue;
1077 }
1078
1079 send_file_name(f, flist, fbuf, NULL, 0);
1080 }
1081
1082 fbuf[len] = '\0';
1083
1084 if (errno) {
1085 io_error |= IOERR_GENERAL;
1086 rsyserr(FERROR, errno, "readdir(%s)", full_fname(fbuf));
1087 }
1088
1089 closedir(d);
1090
1091 if (recurse) {
1092 int i, end = flist->count - 1;
1093 for (i = start; i <= end; i++)
1094 send_if_directory(f, flist, flist->files[i], fbuf, len);
1095 }
1096}
1097
1098struct file_list *send_file_list(int f, int argc, char *argv[])
1099{
1100 int len;
1101 STRUCT_STAT st;
1102 char *p, *dir, olddir[sizeof curr_dir];
1103 char lastpath[MAXPATHLEN] = "";
1104 struct file_list *flist;
1105 struct timeval start_tv, end_tv;
1106 int64 start_write;
1107 int use_ff_fd = 0;
1108
1109 rprintf(FLOG, "building file list\n");
1110 if (show_filelist_p())
1111 start_filelist_progress("building file list");
1112
1113 start_write = stats.total_written;
1114 gettimeofday(&start_tv, NULL);
1115
1116 flist = flist_new(0, "send_file_list");
1117
1118 io_start_buffering_out();
1119 if (filesfrom_fd >= 0) {
1120 if (argv[0] && !push_dir(argv[0], 0)) {
1121 rsyserr(FERROR, errno, "push_dir %s failed",
1122 full_fname(argv[0]));
1123 exit_cleanup(RERR_FILESELECT);
1124 }
1125 use_ff_fd = 1;
1126 }
1127
1128 while (1) {
1129 char fbuf[MAXPATHLEN];
1130 char *fn;
1131 int is_dot_dir;
1132
1133 if (use_ff_fd) {
1134 if (read_filesfrom_line(filesfrom_fd, fbuf) == 0)
1135 break;
1136 sanitize_path(fbuf, fbuf, "", 0, NULL);
1137 } else {
1138 if (argc-- == 0)
1139 break;
1140 strlcpy(fbuf, *argv++, MAXPATHLEN);
1141 if (sanitize_paths)
1142 sanitize_path(fbuf, fbuf, "", 0, NULL);
1143 }
1144
1145 len = strlen(fbuf);
1146 if (relative_paths) {
1147 /* We clean up fbuf below. */
1148 is_dot_dir = 0;
1149 } else if (!len || fbuf[len - 1] == '/') {
1150 if (len == 2 && fbuf[0] == '.') {
1151 /* Turn "./" into just "." rather than "./." */
1152 fbuf[1] = '\0';
1153 } else {
1154 if (len + 1 >= MAXPATHLEN)
1155 overflow_exit("send_file_list");
1156 fbuf[len++] = '.';
1157 fbuf[len] = '\0';
1158 }
1159 is_dot_dir = 1;
1160 } else if (len > 1 && fbuf[len-1] == '.' && fbuf[len-2] == '.'
1161 && (len == 2 || fbuf[len-3] == '/')) {
1162 if (len + 2 >= MAXPATHLEN)
1163 overflow_exit("send_file_list");
1164 fbuf[len++] = '/';
1165 fbuf[len++] = '.';
1166 fbuf[len] = '\0';
1167 is_dot_dir = 1;
1168 } else {
1169 is_dot_dir = fbuf[len-1] == '.'
1170 && (len == 1 || fbuf[len-2] == '/');
1171 }
1172
1173 if (link_stat(fbuf, &st, copy_dirlinks) != 0) {
1174 io_error |= IOERR_GENERAL;
1175 rsyserr(FERROR, errno, "link_stat %s failed",
1176 full_fname(fbuf));
1177 continue;
1178 }
1179
1180 if (S_ISDIR(st.st_mode) && !xfer_dirs) {
1181 rprintf(FINFO, "skipping directory %s\n", fbuf);
1182 continue;
1183 }
1184
1185 dir = NULL;
1186 olddir[0] = '\0';
1187
1188 if (!relative_paths) {
1189 p = strrchr(fbuf, '/');
1190 if (p) {
1191 *p = '\0';
1192 if (p == fbuf)
1193 dir = "/";
1194 else
1195 dir = fbuf;
1196 len -= p - fbuf + 1;
1197 fn = p + 1;
1198 } else
1199 fn = fbuf;
1200 } else {
1201 if ((p = strstr(fbuf, "/./")) != NULL) {
1202 *p = '\0';
1203 if (p == fbuf)
1204 dir = "/";
1205 else
1206 dir = fbuf;
1207 len -= p - fbuf + 3;
1208 fn = p + 3;
1209 } else
1210 fn = fbuf;
1211 /* Get rid of trailing "/" and "/.". */
1212 while (len) {
1213 if (fn[len - 1] == '/') {
1214 is_dot_dir = 1;
1215 if (!--len && !dir) {
1216 len++;
1217 break;
1218 }
1219 }
1220 else if (len >= 2 && fn[len - 1] == '.'
1221 && fn[len - 2] == '/') {
1222 is_dot_dir = 1;
1223 if (!(len -= 2) && !dir) {
1224 len++;
1225 break;
1226 }
1227 } else
1228 break;
1229 }
1230 if (len == 1 && fn[0] == '/')
1231 fn[len++] = '.';
1232 fn[len] = '\0';
1233 /* Reject a ".." dir in the active part of the path. */
1234 for (p = fn; (p = strstr(p, "..")) != NULL; p += 2) {
1235 if ((p[2] == '/' || p[2] == '\0')
1236 && (p == fn || p[-1] == '/')) {
1237 rprintf(FERROR,
1238 "found \"..\" dir in relative path: %s\n",
1239 fbuf);
1240 exit_cleanup(RERR_SYNTAX);
1241 }
1242 }
1243 }
1244
1245 if (!*fn) {
1246 len = 1;
1247 fn = ".";
1248 }
1249
1250 if (dir && *dir) {
1251 static const char *lastdir;
1252 static int lastdir_len;
1253
1254 strlcpy(olddir, curr_dir, sizeof olddir);
1255
1256 if (!push_dir(dir, 0)) {
1257 io_error |= IOERR_GENERAL;
1258 rsyserr(FERROR, errno, "push_dir %s failed",
1259 full_fname(dir));
1260 continue;
1261 }
1262
1263 if (lastdir && strcmp(lastdir, dir) == 0) {
1264 flist_dir = lastdir;
1265 flist_dir_len = lastdir_len;
1266 } else {
1267 flist_dir = lastdir = strdup(dir);
1268 flist_dir_len = lastdir_len = strlen(dir);
1269 }
1270 }
1271
1272 if (fn != fbuf)
1273 memmove(fbuf, fn, len + 1);
1274
1275 if (implied_dirs && (p=strrchr(fbuf,'/')) && p != fbuf) {
1276 /* Send the implied directories at the start of the
1277 * source spec, so we get their permissions right. */
1278 char *lp = lastpath, *slash = fbuf;
1279 *p = '\0';
1280 /* Skip any initial directories in our path that we
1281 * have in common with lastpath. */
1282 for (fn = fbuf; *fn && *lp == *fn; lp++, fn++) {
1283 if (*fn == '/')
1284 slash = fn;
1285 }
1286 *p = '/';
1287 if (fn != p || (*lp && *lp != '/')) {
1288 int save_copy_links = copy_links;
1289 int save_xfer_dirs = xfer_dirs;
1290 copy_links |= copy_unsafe_links;
1291 xfer_dirs = 1;
1292 while ((slash = strchr(slash+1, '/')) != 0) {
1293 *slash = '\0';
1294 send_file_name(f, flist, fbuf, NULL, 0);
1295 *slash = '/';
1296 }
1297 copy_links = save_copy_links;
1298 xfer_dirs = save_xfer_dirs;
1299 *p = '\0';
1300 strlcpy(lastpath, fbuf, sizeof lastpath);
1301 *p = '/';
1302 }
1303 }
1304
1305 if (one_file_system)
1306 filesystem_dev = st.st_dev;
1307
1308 if (recurse || (xfer_dirs && is_dot_dir)) {
1309 struct file_struct *file;
1310 file = send_file_name(f, flist, fbuf, &st, FLAG_TOP_DIR);
1311 if (file)
1312 send_if_directory(f, flist, file, fbuf, len);
1313 } else
1314 send_file_name(f, flist, fbuf, &st, 0);
1315
1316 if (olddir[0]) {
1317 flist_dir = NULL;
1318 flist_dir_len = 0;
1319 if (!pop_dir(olddir)) {
1320 rsyserr(FERROR, errno, "pop_dir %s failed",
1321 full_fname(olddir));
1322 exit_cleanup(RERR_FILESELECT);
1323 }
1324 }
1325 }
1326
1327 gettimeofday(&end_tv, NULL);
1328 stats.flist_buildtime = (int64)(end_tv.tv_sec - start_tv.tv_sec) * 1000
1329 + (end_tv.tv_usec - start_tv.tv_usec) / 1000;
1330 if (stats.flist_buildtime == 0)
1331 stats.flist_buildtime = 1;
1332 start_tv = end_tv;
1333
1334 send_file_entry(NULL, f);
1335
1336 if (show_filelist_p())
1337 finish_filelist_progress(flist);
1338
1339 gettimeofday(&end_tv, NULL);
1340 stats.flist_xfertime = (int64)(end_tv.tv_sec - start_tv.tv_sec) * 1000
1341 + (end_tv.tv_usec - start_tv.tv_usec) / 1000;
1342
1343 /* Sort the list without removing any duplicates. This allows the
1344 * receiving side to ask for any name they like, which gives us the
1345 * flexibility to change the way we unduplicate names in the future
1346 * without causing a compatibility problem with older versions. */
1347 clean_flist(flist, 0, 0);
1348
1349 if (!numeric_ids)
1350 send_uid_list(f);
1351
1352 /* send the io_error flag */
1353 write_int(f, lp_ignore_errors(module_id) ? 0 : io_error);
1354
1355 io_end_buffering();
1356 stats.flist_size = stats.total_written - start_write;
1357 stats.num_files = flist->count;
1358
1359 if (verbose > 3)
1360 output_flist(flist);
1361
1362 if (verbose > 2)
1363 rprintf(FINFO, "send_file_list done\n");
1364
1365 return flist;
1366}
1367
1368struct file_list *recv_file_list(int f)
1369{
1370 struct file_list *flist;
1371 unsigned short flags;
1372 int64 start_read;
1373
1374 rprintf(FLOG, "receiving file list\n");
1375 if (show_filelist_p())
1376 start_filelist_progress("receiving file list");
1377
1378 start_read = stats.total_read;
1379
1380 flist = flist_new(WITH_HLINK, "recv_file_list");
1381
1382
1383 while ((flags = read_byte(f)) != 0) {
1384 struct file_struct *file;
1385
1386 flist_expand(flist);
1387
1388 if (protocol_version >= 28 && (flags & XMIT_EXTENDED_FLAGS))
1389 flags |= read_byte(f) << 8;
1390 file = recv_file_entry(flist, flags, f);
1391
1392 if (S_ISREG(file->mode) || S_ISLNK(file->mode))
1393 stats.total_size += F_LENGTH(file);
1394
1395 flist->files[flist->count++] = file;
1396
1397 maybe_emit_filelist_progress(flist->count);
1398
1399 if (verbose > 2) {
1400 rprintf(FINFO, "recv_file_name(%s)\n",
1401 f_name(file, NULL));
1402 }
1403 }
1404 recv_file_entry(NULL, 0, 0); /* Signal that we're done. */
1405
1406 if (verbose > 2)
1407 rprintf(FINFO, "received %d names\n", flist->count);
1408
1409 if (show_filelist_p())
1410 finish_filelist_progress(flist);
1411
1412 clean_flist(flist, relative_paths, 1);
1413
1414 if (f >= 0) {
1415 recv_uid_list(f, flist);
1416
1417 /* Recv the io_error flag */
1418 if (lp_ignore_errors(module_id) || ignore_errors)
1419 read_int(f);
1420 else
1421 io_error |= read_int(f);
1422 }
1423
1424 if (verbose > 3)
1425 output_flist(flist);
1426
1427 if (list_only) {
1428 int i;
1429 for (i = 0; i < flist->count; i++)
1430 list_file_entry(flist->files[i]);
1431 }
1432
1433 if (verbose > 2)
1434 rprintf(FINFO, "recv_file_list done\n");
1435
1436 stats.flist_size = stats.total_read - start_read;
1437 stats.num_files = flist->count;
1438
1439 return flist;
1440}
1441
1442static int file_compare(struct file_struct **file1, struct file_struct **file2)
1443{
1444 return f_name_cmp(*file1, *file2);
1445}
1446
1447/* Search for an identically-named item in the file list. Note that the
1448 * items must agree in their directory-ness, or no match is returned. */
1449int flist_find(struct file_list *flist, struct file_struct *f)
1450{
1451 int low = flist->low, high = flist->high;
1452 int diff, mid, mid_up;
1453
1454 while (low <= high) {
1455 mid = (low + high) / 2;
1456 if (F_IS_ACTIVE(flist->files[mid]))
1457 mid_up = mid;
1458 else {
1459 /* Scan for the next non-empty entry using the cached
1460 * distance values. If the value isn't fully up-to-
1461 * date, update it. */
1462 mid_up = mid + flist->files[mid]->dir.depth;
1463 if (!F_IS_ACTIVE(flist->files[mid_up])) {
1464 do {
1465 mid_up += flist->files[mid_up]->dir.depth;
1466 } while (!F_IS_ACTIVE(flist->files[mid_up]));
1467 flist->files[mid]->dir.depth = mid_up - mid;
1468 }
1469 if (mid_up > high) {
1470 /* If there's nothing left above us, set high to
1471 * a non-empty entry below us and continue. */
1472 high = mid - (int)flist->files[mid]->len32;
1473 if (!F_IS_ACTIVE(flist->files[high])) {
1474 do {
1475 high -= (int)flist->files[high]->len32;
1476 } while (!F_IS_ACTIVE(flist->files[high]));
1477 flist->files[mid]->len32 = mid - high;
1478 }
1479 continue;
1480 }
1481 }
1482 diff = f_name_cmp(flist->files[mid_up], f);
1483 if (diff == 0) {
1484 if (protocol_version < 29
1485 && S_ISDIR(flist->files[mid_up]->mode)
1486 != S_ISDIR(f->mode))
1487 return -1;
1488 return mid_up;
1489 }
1490 if (diff < 0)
1491 low = mid_up + 1;
1492 else
1493 high = mid - 1;
1494 }
1495 return -1;
1496}
1497
1498/*
1499 * Free up any resources a file_struct has allocated
1500 * and clear the file.
1501 */
1502void clear_file(struct file_struct *file)
1503{
1504 memset((char*)file + EXTRA_LEN, 0, FILE_STRUCT_LEN - EXTRA_LEN + 1);
1505 /* In an empty entry, dir.depth is an offset to the next non-empty
1506 * entry. Likewise for len32 in the opposite direction. We assume
1507 * that we're alone for now since flist_find() will adjust the counts
1508 * it runs into that aren't up-to-date. */
1509 file->len32 = file->dir.depth = 1;
1510}
1511
1512/* Allocate a new file list. */
1513struct file_list *flist_new(int with_hlink, char *msg)
1514{
1515 struct file_list *flist;
1516
1517 flist = new(struct file_list);
1518 if (!flist)
1519 out_of_memory(msg);
1520
1521 memset(flist, 0, sizeof flist[0]);
1522
1523 if (!(flist->file_pool = pool_create(FILE_EXTENT, 0, out_of_memory, POOL_INTERN)))
1524 out_of_memory(msg);
1525
1526#ifdef SUPPORT_HARD_LINKS
1527 if (with_hlink && preserve_hard_links) {
1528 if (!(flist->hlink_pool = pool_create(HLINK_EXTENT,
1529 sizeof (struct idev), out_of_memory, POOL_INTERN)))
1530 out_of_memory(msg);
1531 }
1532#endif
1533
1534 return flist;
1535}
1536
1537/* Free up all elements in a flist. */
1538void flist_free(struct file_list *flist)
1539{
1540 pool_destroy(flist->file_pool);
1541 pool_destroy(flist->hlink_pool);
1542 free(flist->files);
1543 free(flist);
1544}
1545
1546/*
1547 * This routine ensures we don't have any duplicate names in our file list.
1548 * duplicate names can cause corruption because of the pipelining
1549 */
1550static void clean_flist(struct file_list *flist, int strip_root, int no_dups)
1551{
1552 char fbuf[MAXPATHLEN];
1553 int i, prev_i = 0;
1554
1555 if (!flist)
1556 return;
1557 if (flist->count == 0) {
1558 flist->high = -1;
1559 return;
1560 }
1561
1562 qsort(flist->files, flist->count,
1563 sizeof flist->files[0], (int (*)())file_compare);
1564
1565 for (i = no_dups? 0 : flist->count; i < flist->count; i++) {
1566 if (F_IS_ACTIVE(flist->files[i])) {
1567 prev_i = i;
1568 break;
1569 }
1570 }
1571 flist->low = prev_i;
1572 while (++i < flist->count) {
1573 int j;
1574 struct file_struct *file = flist->files[i];
1575
1576 if (!F_IS_ACTIVE(file))
1577 continue;
1578 if (f_name_cmp(file, flist->files[prev_i]) == 0)
1579 j = prev_i;
1580 else if (protocol_version >= 29 && S_ISDIR(file->mode)) {
1581 int save_mode = file->mode;
1582 /* Make sure that this directory doesn't duplicate a
1583 * non-directory earlier in the list. */
1584 flist->high = prev_i;
1585 file->mode = S_IFREG;
1586 j = flist_find(flist, file);
1587 file->mode = save_mode;
1588 } else
1589 j = -1;
1590 if (j >= 0) {
1591 struct file_struct *fp = flist->files[j];
1592 int keep, drop;
1593 /* If one is a dir and the other is not, we want to
1594 * keep the dir because it might have contents in the
1595 * list. */
1596 if (S_ISDIR(file->mode) != S_ISDIR(fp->mode)) {
1597 if (S_ISDIR(file->mode))
1598 keep = i, drop = j;
1599 else
1600 keep = j, drop = i;
1601 } else
1602 keep = j, drop = i;
1603 if (verbose > 1 && !am_server) {
1604 rprintf(FINFO,
1605 "removing duplicate name %s from file list (%d)\n",
1606 f_name(file, fbuf), drop);
1607 }
1608 /* Make sure we don't lose track of a user-specified
1609 * top directory. */
1610 flist->files[keep]->flags |= flist->files[drop]->flags
1611 & (FLAG_TOP_DIR|FLAG_XFER_DIR);
1612
1613 clear_file(flist->files[drop]);
1614
1615 if (keep == i) {
1616 if (flist->low == drop) {
1617 for (j = drop + 1;
1618 j < i && !F_IS_ACTIVE(flist->files[j]);
1619 j++) {}
1620 flist->low = j;
1621 }
1622 prev_i = i;
1623 }
1624 } else
1625 prev_i = i;
1626 }
1627 flist->high = no_dups ? prev_i : flist->count - 1;
1628
1629 if (strip_root) {
1630 /* We need to strip off the leading slashes for relative
1631 * paths, but this must be done _after_ the sorting phase. */
1632 for (i = flist->low; i <= flist->high; i++) {
1633 struct file_struct *file = flist->files[i];
1634
1635 if (!file->dirname)
1636 continue;
1637 while (*file->dirname == '/')
1638 file->dirname++;
1639 if (!*file->dirname)
1640 file->dirname = NULL;
1641 }
1642 }
1643
1644 if (prune_empty_dirs && no_dups) {
1645 int j, prev_depth = 0;
1646
1647 prev_i = 0; /* It's OK that this isn't really true. */
1648
1649 for (i = flist->low; i <= flist->high; i++) {
1650 struct file_struct *fp, *file = flist->files[i];
1651
1652 /* This temporarily abuses the dir.depth value for a
1653 * directory that is in a chain that might get pruned.
1654 * We restore the old value if it gets a reprieve. */
1655 if (S_ISDIR(file->mode) && file->dir.depth) {
1656 /* Dump empty dirs when coming back down. */
1657 for (j = prev_depth; j >= file->dir.depth; j--) {
1658 fp = flist->files[prev_i];
1659 if (fp->dir.depth >= 0)
1660 break;
1661 prev_i = -fp->dir.depth-1;
1662 clear_file(fp);
1663 }
1664 prev_depth = file->dir.depth;
1665 if (is_excluded(f_name(file, fbuf), 1,
1666 ALL_FILTERS)) {
1667 /* Keep dirs through this dir. */
1668 for (j = prev_depth-1; ; j--) {
1669 fp = flist->files[prev_i];
1670 if (fp->dir.depth >= 0)
1671 break;
1672 prev_i = -fp->dir.depth-1;
1673 fp->dir.depth = j;
1674 }
1675 } else
1676 file->dir.depth = -prev_i-1;
1677 prev_i = i;
1678 } else {
1679 /* Keep dirs through this non-dir. */
1680 for (j = prev_depth; ; j--) {
1681 fp = flist->files[prev_i];
1682 if (fp->dir.depth >= 0)
1683 break;
1684 prev_i = -fp->dir.depth-1;
1685 fp->dir.depth = j;
1686 }
1687 }
1688 }
1689 /* Dump empty all remaining empty dirs. */
1690 while (1) {
1691 struct file_struct *fp = flist->files[prev_i];
1692 if (fp->dir.depth >= 0)
1693 break;
1694 prev_i = -fp->dir.depth-1;
1695 clear_file(fp);
1696 }
1697
1698 for (i = flist->low; i <= flist->high; i++) {
1699 if (F_IS_ACTIVE(flist->files[i]))
1700 break;
1701 }
1702 flist->low = i;
1703 for (i = flist->high; i >= flist->low; i--) {
1704 if (F_IS_ACTIVE(flist->files[i]))
1705 break;
1706 }
1707 flist->high = i;
1708 }
1709}
1710
1711static void output_flist(struct file_list *flist)
1712{
1713 char uidbuf[16], gidbuf[16], depthbuf[16];
1714 struct file_struct *file;
1715 const char *dir, *slash, *name, *trail;
1716 const char *who = who_am_i();
1717 int i;
1718
1719 for (i = 0; i < flist->count; i++) {
1720 file = flist->files[i];
1721 if ((am_root || am_sender) && preserve_uid) {
1722 snprintf(uidbuf, sizeof uidbuf, " uid=%ld",
1723 (long)F_UID(file));
1724 } else
1725 *uidbuf = '\0';
1726 if (preserve_gid && F_GID(file) != GID_NONE) {
1727 snprintf(gidbuf, sizeof gidbuf, " gid=%ld",
1728 (long)F_GID(file));
1729 } else
1730 *gidbuf = '\0';
1731 if (!am_sender)
1732 snprintf(depthbuf, sizeof depthbuf, "%d", file->dir.depth);
1733 if (F_IS_ACTIVE(file)) {
1734 if ((dir = file->dirname) == NULL)
1735 dir = slash = "";
1736 else
1737 slash = "/";
1738 name = F_BASENAME(file);
1739 trail = S_ISDIR(file->mode) ? "/" : "";
1740 } else
1741 dir = slash = name = trail = "";
1742 rprintf(FINFO, "[%s] i=%d %s %s%s%s%s mode=0%o len=%.0f%s%s flags=%x\n",
1743 who, i, am_sender ? NS(file->dir.root) : depthbuf,
1744 dir, slash, name, trail, (int)file->mode,
1745 (double)F_LENGTH(file), uidbuf, gidbuf, file->flags);
1746 }
1747}
1748
1749enum fnc_state { s_DIR, s_SLASH, s_BASE, s_TRAILING };
1750enum fnc_type { t_PATH, t_ITEM };
1751
1752/* Compare the names of two file_struct entities, similar to how strcmp()
1753 * would do if it were operating on the joined strings.
1754 *
1755 * Some differences beginning with protocol_version 29: (1) directory names
1756 * are compared with an assumed trailing slash so that they compare in a
1757 * way that would cause them to sort immediately prior to any content they
1758 * may have; (2) a directory of any name compares after a non-directory of
1759 * any name at the same depth; (3) a directory with name "." compares prior
1760 * to anything else. These changes mean that a directory and a non-dir
1761 * with the same name will not compare as equal (protocol_version >= 29).
1762 *
1763 * The dirname component can be an empty string, but the basename component
1764 * cannot (and never is in the current codebase). The basename component
1765 * may be NULL (for a removed item), in which case it is considered to be
1766 * after any existing item. */
1767int f_name_cmp(struct file_struct *f1, struct file_struct *f2)
1768{
1769 int dif;
1770 const uchar *c1, *c2;
1771 enum fnc_state state1, state2;
1772 enum fnc_type type1, type2;
1773 enum fnc_type t_path = protocol_version >= 29 ? t_PATH : t_ITEM;
1774
1775 if (!f1 || !F_IS_ACTIVE(f1)) {
1776 if (!f2 || !F_IS_ACTIVE(f2))
1777 return 0;
1778 return -1;
1779 }
1780 if (!f2 || !F_IS_ACTIVE(f2))
1781 return 1;
1782
1783 c1 = (uchar*)f1->dirname;
1784 c2 = (uchar*)f2->dirname;
1785 if (c1 == c2)
1786 c1 = c2 = NULL;
1787 if (!c1) {
1788 type1 = S_ISDIR(f1->mode) ? t_path : t_ITEM;
1789 c1 = (uchar*)F_BASENAME(f1);
1790 if (type1 == t_PATH && *c1 == '.' && !c1[1]) {
1791 type1 = t_ITEM;
1792 state1 = s_TRAILING;
1793 c1 = (uchar*)"";
1794 } else
1795 state1 = s_BASE;
1796 } else {
1797 type1 = t_path;
1798 state1 = s_DIR;
1799 }
1800 if (!c2) {
1801 type2 = S_ISDIR(f2->mode) ? t_path : t_ITEM;
1802 c2 = (uchar*)F_BASENAME(f2);
1803 if (type2 == t_PATH && *c2 == '.' && !c2[1]) {
1804 type2 = t_ITEM;
1805 state2 = s_TRAILING;
1806 c2 = (uchar*)"";
1807 } else
1808 state2 = s_BASE;
1809 } else {
1810 type2 = t_path;
1811 state2 = s_DIR;
1812 }
1813
1814 if (type1 != type2)
1815 return type1 == t_PATH ? 1 : -1;
1816
1817 do {
1818 if (!*c1) {
1819 switch (state1) {
1820 case s_DIR:
1821 state1 = s_SLASH;
1822 c1 = (uchar*)"/";
1823 break;
1824 case s_SLASH:
1825 type1 = S_ISDIR(f1->mode) ? t_path : t_ITEM;
1826 c1 = (uchar*)F_BASENAME(f1);
1827 if (type1 == t_PATH && *c1 == '.' && !c1[1]) {
1828 type1 = t_ITEM;
1829 state1 = s_TRAILING;
1830 c1 = (uchar*)"";
1831 } else
1832 state1 = s_BASE;
1833 break;
1834 case s_BASE:
1835 state1 = s_TRAILING;
1836 if (type1 == t_PATH) {
1837 c1 = (uchar*)"/";
1838 break;
1839 }
1840 /* FALL THROUGH */
1841 case s_TRAILING:
1842 type1 = t_ITEM;
1843 break;
1844 }
1845 if (*c2 && type1 != type2)
1846 return type1 == t_PATH ? 1 : -1;
1847 }
1848 if (!*c2) {
1849 switch (state2) {
1850 case s_DIR:
1851 state2 = s_SLASH;
1852 c2 = (uchar*)"/";
1853 break;
1854 case s_SLASH:
1855 type2 = S_ISDIR(f2->mode) ? t_path : t_ITEM;
1856 c2 = (uchar*)F_BASENAME(f2);
1857 if (type2 == t_PATH && *c2 == '.' && !c2[1]) {
1858 type2 = t_ITEM;
1859 state2 = s_TRAILING;
1860 c2 = (uchar*)"";
1861 } else
1862 state2 = s_BASE;
1863 break;
1864 case s_BASE:
1865 state2 = s_TRAILING;
1866 if (type2 == t_PATH) {
1867 c2 = (uchar*)"/";
1868 break;
1869 }
1870 /* FALL THROUGH */
1871 case s_TRAILING:
1872 if (!*c1)
1873 return 0;
1874 type2 = t_ITEM;
1875 break;
1876 }
1877 if (type1 != type2)
1878 return type1 == t_PATH ? 1 : -1;
1879 }
1880 } while ((dif = (int)*c1++ - (int)*c2++) == 0);
1881
1882 return dif;
1883}
1884
1885char *f_name_buf(void)
1886{
1887 static char names[5][MAXPATHLEN];
1888 static unsigned int n;
1889
1890 n = (n + 1) % (sizeof names / sizeof names[0]);
1891
1892 return names[n];
1893}
1894
1895/* Return a copy of the full filename of a flist entry, using the indicated
1896 * buffer or one of 5 static buffers if fbuf is NULL. No size-checking is
1897 * done because we checked the size when creating the file_struct entry.
1898 */
1899char *f_name(struct file_struct *f, char *fbuf)
1900{
1901 if (!f || !F_IS_ACTIVE(f))
1902 return NULL;
1903
1904 if (!fbuf)
1905 fbuf = f_name_buf();
1906
1907 if (f->dirname) {
1908 int len = strlen(f->dirname);
1909 memcpy(fbuf, f->dirname, len);
1910 fbuf[len] = '/';
1911 strlcpy(fbuf + len + 1, F_BASENAME(f), MAXPATHLEN - (len + 1));
1912 } else
1913 strlcpy(fbuf, F_BASENAME(f), MAXPATHLEN);
1914
1915 return fbuf;
1916}
1917
1918/* Do a non-recursive scan of the named directory, possibly ignoring all
1919 * exclude rules except for the daemon's. If "dlen" is >=0, it is the length
1920 * of the dirname string, and also indicates that "dirname" is a MAXPATHLEN
1921 * buffer (the functions we call will append names onto the end, but the old
1922 * dir value will be restored on exit). */
1923struct file_list *get_dirlist(char *dirname, int dlen,
1924 int ignore_filter_rules)
1925{
1926 struct file_list *dirlist;
1927 char dirbuf[MAXPATHLEN];
1928 int save_recurse = recurse;
1929 int save_xfer_dirs = xfer_dirs;
1930
1931 if (dlen < 0) {
1932 dlen = strlcpy(dirbuf, dirname, MAXPATHLEN);
1933 if (dlen >= MAXPATHLEN)
1934 return NULL;
1935 dirname = dirbuf;
1936 }
1937
1938 dirlist = flist_new(WITHOUT_HLINK, "get_dirlist");
1939
1940 recurse = 0;
1941 xfer_dirs = 1;
1942 send_directory(ignore_filter_rules ? -2 : -1, dirlist, dirname, dlen);
1943 xfer_dirs = save_xfer_dirs;
1944 recurse = save_recurse;
1945 if (do_progress)
1946 flist_count_offset += dirlist->count;
1947
1948 clean_flist(dirlist, 0, 0);
1949
1950 if (verbose > 3)
1951 output_flist(dirlist);
1952
1953 return dirlist;
1954}