#ifdef calls to hard-link functions.
[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 && protocol_version < 28 && S_ISREG(mode))
617 flags |= XMIT_HAS_IDEV_DATA;
618 if (flags & XMIT_HAS_IDEV_DATA) {
619 extra_len += EXTRA_LEN;
620 assert(flist->hlink_pool != NULL);
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 F_DMAJOR(file) = major(rdev);
696 F_DMINOR(file) = minor(rdev);
697 }
698
699#ifdef SUPPORT_LINKS
700 if (linkname_len) {
701 bp = (char*)F_BASENAME(file) + basename_len;
702 read_sbuf(f, bp, linkname_len - 1);
703 if (sanitize_paths)
704 sanitize_path(bp, bp, "", lastdir_depth, NULL);
705 }
706#endif
707
708#ifdef SUPPORT_HARD_LINKS
709 if (flags & XMIT_HAS_IDEV_DATA) {
710 struct idev *idevp = pool_talloc(flist->hlink_pool, struct idev,
711 1, "inode_table");
712 F_IDEV(file) = idevp;
713 if (protocol_version < 26) {
714 idevp->dev = read_int(f);
715 idevp->ino = read_int(f);
716 } else {
717 if (!(flags & XMIT_SAME_DEV))
718 dev = read_longint(f);
719 idevp->dev = dev;
720 idevp->ino = read_longint(f);
721 }
722 }
723#endif
724
725 if (always_checksum && (S_ISREG(mode) || protocol_version < 28)) {
726 if (S_ISREG(mode))
727 bp = (char*)F_SUM(file);
728 else {
729 /* Prior to 28, we get a useless set of nulls. */
730 bp = tmp_sum;
731 }
732 read_buf(f, bp, checksum_len);
733 }
734
735 return file;
736}
737
738/**
739 * Create a file_struct for a named file by reading its stat()
740 * information and performing extensive checks against global
741 * options.
742 *
743 * @return the new file, or NULL if there was an error or this file
744 * should be excluded.
745 *
746 * @todo There is a small optimization opportunity here to avoid
747 * stat()ing the file in some circumstances, which has a certain cost.
748 * We are called immediately after doing readdir(), and so we may
749 * already know the d_type of the file. We could for example avoid
750 * statting directories if we're not recursing, but this is not a very
751 * important case. Some systems may not have d_type.
752 **/
753struct file_struct *make_file(const char *fname, struct file_list *flist,
754 STRUCT_STAT *stp, int flags, int filter_level)
755{
756 static char *lastdir;
757 static int lastdir_len = -1;
758 struct file_struct *file;
759 STRUCT_STAT st;
760 char thisname[MAXPATHLEN];
761 char linkname[MAXPATHLEN];
762 int alloc_len, basename_len, dirname_len, linkname_len;
763 int extra_len = (flist_extra_cnt - 1) * EXTRA_LEN;
764 char *basename, *dirname, *bp;
765
766 if (!flist || !flist->count) /* Ignore lastdir when invalid. */
767 lastdir_len = -1;
768
769 if (strlcpy(thisname, fname, sizeof thisname)
770 >= sizeof thisname - flist_dir_len) {
771 rprintf(FINFO, "skipping overly long name: %s\n", fname);
772 return NULL;
773 }
774 clean_fname(thisname, 0);
775 if (sanitize_paths)
776 sanitize_path(thisname, thisname, "", 0, NULL);
777
778 if (stp && S_ISDIR(stp->st_mode)) {
779 st = *stp; /* Needed for "symlink/." with --relative. */
780 *linkname = '\0'; /* make IBM code checker happy */
781 } else if (readlink_stat(thisname, &st, linkname) != 0) {
782 int save_errno = errno;
783 /* See if file is excluded before reporting an error. */
784 if (filter_level != NO_FILTERS
785 && (is_excluded(thisname, 0, filter_level)
786 || is_excluded(thisname, 1, filter_level))) {
787 if (ignore_perishable && save_errno != ENOENT)
788 non_perishable_cnt++;
789 return NULL;
790 }
791 if (save_errno == ENOENT) {
792#ifdef SUPPORT_LINKS
793 /* Avoid "vanished" error if symlink points nowhere. */
794 if (copy_links && do_lstat(thisname, &st) == 0
795 && S_ISLNK(st.st_mode)) {
796 io_error |= IOERR_GENERAL;
797 rprintf(FERROR, "symlink has no referent: %s\n",
798 full_fname(thisname));
799 } else
800#endif
801 {
802 enum logcode c = am_daemon && protocol_version < 28
803 ? FERROR : FINFO;
804 io_error |= IOERR_VANISHED;
805 rprintf(c, "file has vanished: %s\n",
806 full_fname(thisname));
807 }
808 } else {
809 io_error |= IOERR_GENERAL;
810 rsyserr(FERROR, save_errno, "readlink %s failed",
811 full_fname(thisname));
812 }
813 return NULL;
814 }
815
816 /* backup.c calls us with filter_level set to NO_FILTERS. */
817 if (filter_level == NO_FILTERS)
818 goto skip_filters;
819
820 if (S_ISDIR(st.st_mode) && !xfer_dirs) {
821 rprintf(FINFO, "skipping directory %s\n", thisname);
822 return NULL;
823 }
824
825 /* -x only affects directories because we need to avoid recursing
826 * into a mount-point directory, not to avoid copying a symlinked
827 * file if -L (or similar) was specified. */
828 if (one_file_system && st.st_dev != filesystem_dev
829 && S_ISDIR(st.st_mode)) {
830 if (one_file_system > 1) {
831 if (verbose > 2) {
832 rprintf(FINFO, "skipping mount-point dir %s\n",
833 thisname);
834 }
835 return NULL;
836 }
837 flags |= FLAG_MOUNT_DIR;
838 }
839
840 if (is_excluded(thisname, S_ISDIR(st.st_mode) != 0, filter_level)) {
841 if (ignore_perishable)
842 non_perishable_cnt++;
843 return NULL;
844 }
845
846 if (lp_ignore_nonreadable(module_id)) {
847#ifdef SUPPORT_LINKS
848 if (!S_ISLNK(st.st_mode))
849#endif
850 if (access(thisname, R_OK) != 0)
851 return NULL;
852 }
853
854 skip_filters:
855
856 if (verbose > 2) {
857 rprintf(FINFO, "[%s] make_file(%s,*,%d)\n",
858 who_am_i(), thisname, filter_level);
859 }
860
861 if ((basename = strrchr(thisname, '/')) != NULL) {
862 dirname_len = ++basename - thisname; /* counts future '\0' */
863 if (lastdir_len == dirname_len - 1
864 && strncmp(thisname, lastdir, lastdir_len) == 0) {
865 dirname = lastdir;
866 dirname_len = 0; /* indicates no copy is needed */
867 } else
868 dirname = thisname;
869 } else {
870 basename = thisname;
871 dirname = NULL;
872 dirname_len = 0;
873 }
874 basename_len = strlen(basename) + 1; /* count the '\0' */
875
876#ifdef SUPPORT_LINKS
877 linkname_len = S_ISLNK(st.st_mode) ? strlen(linkname) + 1 : 0;
878#else
879 linkname_len = 0;
880#endif
881
882 if (st.st_size > 0xFFFFFFFFu && S_ISREG(st.st_mode))
883 extra_len += EXTRA_LEN;
884
885 alloc_len = FILE_STRUCT_LEN + extra_len + basename_len + dirname_len
886 + linkname_len;
887 if (flist)
888 bp = pool_alloc(flist->file_pool, alloc_len, "make_file");
889 else {
890 if (!(bp = new_array(char, alloc_len)))
891 out_of_memory("make_file");
892 }
893
894 memset(bp, 0, FILE_STRUCT_LEN + extra_len);
895 bp += extra_len;
896 file = (struct file_struct *)bp;
897 bp += FILE_STRUCT_LEN;
898
899 memcpy(bp, basename, basename_len);
900 bp += basename_len + linkname_len; /* skip space for symlink too */
901
902#ifdef SUPPORT_HARD_LINKS
903 if (preserve_hard_links && flist) {
904 if (protocol_version >= 28
905 ? (!S_ISDIR(st.st_mode) && st.st_nlink > 1)
906 : S_ISREG(st.st_mode)) {
907 tmp_idev.dev = st.st_dev;
908 tmp_idev.ino = st.st_ino;
909 } else
910 tmp_idev.dev = tmp_idev.ino = 0;
911 }
912#endif
913
914#ifdef HAVE_STRUCT_STAT_ST_RDEV
915 if (IS_DEVICE(st.st_mode) || IS_SPECIAL(st.st_mode)) {
916 tmp_rdev = st.st_rdev;
917 st.st_size = 0;
918 }
919#endif
920
921 file->flags = flags;
922 file->modtime = st.st_mtime;
923 file->len32 = st.st_size;
924 if (st.st_size > 0xFFFFFFFFu && S_ISREG(st.st_mode)) {
925 file->flags |= FLAG_LENGTH64;
926 OPT_EXTRA(file, 0)->unum = (uint32)(st.st_size >> 32);
927 }
928 file->mode = st.st_mode;
929 if (preserve_uid)
930 F_UID(file) = st.st_uid;
931 if (preserve_gid)
932 F_GID(file) = st.st_gid;
933
934 if (dirname_len) {
935 file->dirname = lastdir = bp;
936 lastdir_len = dirname_len - 1;
937 memcpy(bp, dirname, dirname_len - 1);
938 bp += dirname_len;
939 bp[-1] = '\0';
940 } else if (dirname)
941 file->dirname = dirname;
942
943#ifdef SUPPORT_LINKS
944 if (linkname_len) {
945 bp = (char*)F_BASENAME(file) + basename_len;
946 memcpy(bp, linkname, linkname_len);
947 }
948#endif
949
950 if (always_checksum && am_sender && S_ISREG(st.st_mode))
951 file_checksum(thisname, tmp_sum, st.st_size);
952
953 file->dir.root = flist_dir;
954
955 /* This code is only used by the receiver when it is building
956 * a list of files for a delete pass. */
957 if (keep_dirlinks && linkname_len && flist) {
958 STRUCT_STAT st2;
959 int save_mode = file->mode;
960 file->mode = S_IFDIR; /* Find a directory with our name. */
961 if (flist_find(the_file_list, file) >= 0
962 && do_stat(thisname, &st2) == 0 && S_ISDIR(st2.st_mode)) {
963 file->modtime = st2.st_mtime;
964 file->len32 = 0;
965 file->mode = st2.st_mode;
966 if (preserve_uid)
967 F_UID(file) = st2.st_uid;
968 if (preserve_gid)
969 F_GID(file) = st2.st_gid;
970 } else
971 file->mode = save_mode;
972 }
973
974 if (S_ISREG(st.st_mode) || S_ISLNK(st.st_mode))
975 stats.total_size += st.st_size;
976
977 if (basename_len == 0+1)
978 return NULL;
979
980 return file;
981}
982
983/* Only called for temporary file_struct entries created by make_file(). */
984void unmake_file(struct file_struct *file)
985{
986 int extra_cnt = flist_extra_cnt - 1 + LEN64_BUMP(file);
987 free(file->extras - extra_cnt);
988}
989
990static struct file_struct *send_file_name(int f, struct file_list *flist,
991 char *fname, STRUCT_STAT *stp,
992 unsigned short flags)
993{
994 struct file_struct *file;
995
996 file = make_file(fname, flist, stp, flags,
997 f == -2 ? SERVER_FILTERS : ALL_FILTERS);
998 if (!file)
999 return NULL;
1000
1001 if (chmod_modes && !S_ISLNK(file->mode))
1002 file->mode = tweak_mode(file->mode, chmod_modes);
1003
1004 maybe_emit_filelist_progress(flist->count + flist_count_offset);
1005
1006 flist_expand(flist);
1007 flist->files[flist->count++] = file;
1008 send_file_entry(file, f);
1009 return file;
1010}
1011
1012static void send_if_directory(int f, struct file_list *flist,
1013 struct file_struct *file,
1014 char *fbuf, unsigned int ol)
1015{
1016 char is_dot_dir = fbuf[ol-1] == '.' && (ol == 1 || fbuf[ol-2] == '/');
1017
1018 if (S_ISDIR(file->mode)
1019 && !(file->flags & FLAG_MOUNT_DIR) && f_name(file, fbuf)) {
1020 void *save_filters;
1021 unsigned int len = strlen(fbuf);
1022 if (len > 1 && fbuf[len-1] == '/')
1023 fbuf[--len] = '\0';
1024 if (len >= MAXPATHLEN - 1) {
1025 io_error |= IOERR_GENERAL;
1026 rprintf(FERROR, "skipping long-named directory: %s\n",
1027 full_fname(fbuf));
1028 return;
1029 }
1030 save_filters = push_local_filters(fbuf, len);
1031 send_directory(f, flist, fbuf, len);
1032 pop_local_filters(save_filters);
1033 fbuf[ol] = '\0';
1034 if (is_dot_dir)
1035 fbuf[ol-1] = '.';
1036 }
1037}
1038
1039/* This function is normally called by the sender, but the receiving side also
1040 * calls it from get_dirlist() with f set to -1 so that we just construct the
1041 * file list in memory without sending it over the wire. Also, get_dirlist()
1042 * might call this with f set to -2, which also indicates that local filter
1043 * rules should be ignored. */
1044static void send_directory(int f, struct file_list *flist,
1045 char *fbuf, int len)
1046{
1047 struct dirent *di;
1048 unsigned remainder;
1049 char *p;
1050 DIR *d;
1051 int start = flist->count;
1052
1053 if (!(d = opendir(fbuf))) {
1054 io_error |= IOERR_GENERAL;
1055 rsyserr(FERROR, errno, "opendir %s failed", full_fname(fbuf));
1056 return;
1057 }
1058
1059 p = fbuf + len;
1060 if (len != 1 || *fbuf != '/')
1061 *p++ = '/';
1062 *p = '\0';
1063 remainder = MAXPATHLEN - (p - fbuf);
1064
1065 for (errno = 0, di = readdir(d); di; errno = 0, di = readdir(d)) {
1066 char *dname = d_name(di);
1067 if (dname[0] == '.' && (dname[1] == '\0'
1068 || (dname[1] == '.' && dname[2] == '\0')))
1069 continue;
1070 if (strlcpy(p, dname, remainder) >= remainder) {
1071 io_error |= IOERR_GENERAL;
1072 rprintf(FINFO,
1073 "cannot send long-named file %s\n",
1074 full_fname(fbuf));
1075 continue;
1076 }
1077
1078 send_file_name(f, flist, fbuf, NULL, 0);
1079 }
1080
1081 fbuf[len] = '\0';
1082
1083 if (errno) {
1084 io_error |= IOERR_GENERAL;
1085 rsyserr(FERROR, errno, "readdir(%s)", full_fname(fbuf));
1086 }
1087
1088 closedir(d);
1089
1090 if (recurse) {
1091 int i, end = flist->count - 1;
1092 for (i = start; i <= end; i++)
1093 send_if_directory(f, flist, flist->files[i], fbuf, len);
1094 }
1095}
1096
1097struct file_list *send_file_list(int f, int argc, char *argv[])
1098{
1099 int len;
1100 STRUCT_STAT st;
1101 char *p, *dir, olddir[sizeof curr_dir];
1102 char lastpath[MAXPATHLEN] = "";
1103 struct file_list *flist;
1104 struct timeval start_tv, end_tv;
1105 int64 start_write;
1106 int use_ff_fd = 0;
1107
1108 rprintf(FLOG, "building file list\n");
1109 if (show_filelist_p())
1110 start_filelist_progress("building file list");
1111
1112 start_write = stats.total_written;
1113 gettimeofday(&start_tv, NULL);
1114
1115 flist = flist_new(0, "send_file_list");
1116
1117 io_start_buffering_out();
1118 if (filesfrom_fd >= 0) {
1119 if (argv[0] && !push_dir(argv[0], 0)) {
1120 rsyserr(FERROR, errno, "push_dir %s failed",
1121 full_fname(argv[0]));
1122 exit_cleanup(RERR_FILESELECT);
1123 }
1124 use_ff_fd = 1;
1125 }
1126
1127 while (1) {
1128 char fbuf[MAXPATHLEN];
1129 char *fn;
1130 int is_dot_dir;
1131
1132 if (use_ff_fd) {
1133 if (read_filesfrom_line(filesfrom_fd, fbuf) == 0)
1134 break;
1135 sanitize_path(fbuf, fbuf, "", 0, NULL);
1136 } else {
1137 if (argc-- == 0)
1138 break;
1139 strlcpy(fbuf, *argv++, MAXPATHLEN);
1140 if (sanitize_paths)
1141 sanitize_path(fbuf, fbuf, "", 0, NULL);
1142 }
1143
1144 len = strlen(fbuf);
1145 if (relative_paths) {
1146 /* We clean up fbuf below. */
1147 is_dot_dir = 0;
1148 } else if (!len || fbuf[len - 1] == '/') {
1149 if (len == 2 && fbuf[0] == '.') {
1150 /* Turn "./" into just "." rather than "./." */
1151 fbuf[1] = '\0';
1152 } else {
1153 if (len + 1 >= MAXPATHLEN)
1154 overflow_exit("send_file_list");
1155 fbuf[len++] = '.';
1156 fbuf[len] = '\0';
1157 }
1158 is_dot_dir = 1;
1159 } else if (len > 1 && fbuf[len-1] == '.' && fbuf[len-2] == '.'
1160 && (len == 2 || fbuf[len-3] == '/')) {
1161 if (len + 2 >= MAXPATHLEN)
1162 overflow_exit("send_file_list");
1163 fbuf[len++] = '/';
1164 fbuf[len++] = '.';
1165 fbuf[len] = '\0';
1166 is_dot_dir = 1;
1167 } else {
1168 is_dot_dir = fbuf[len-1] == '.'
1169 && (len == 1 || fbuf[len-2] == '/');
1170 }
1171
1172 if (link_stat(fbuf, &st, copy_dirlinks) != 0) {
1173 io_error |= IOERR_GENERAL;
1174 rsyserr(FERROR, errno, "link_stat %s failed",
1175 full_fname(fbuf));
1176 continue;
1177 }
1178
1179 if (S_ISDIR(st.st_mode) && !xfer_dirs) {
1180 rprintf(FINFO, "skipping directory %s\n", fbuf);
1181 continue;
1182 }
1183
1184 dir = NULL;
1185 olddir[0] = '\0';
1186
1187 if (!relative_paths) {
1188 p = strrchr(fbuf, '/');
1189 if (p) {
1190 *p = '\0';
1191 if (p == fbuf)
1192 dir = "/";
1193 else
1194 dir = fbuf;
1195 len -= p - fbuf + 1;
1196 fn = p + 1;
1197 } else
1198 fn = fbuf;
1199 } else {
1200 if ((p = strstr(fbuf, "/./")) != NULL) {
1201 *p = '\0';
1202 if (p == fbuf)
1203 dir = "/";
1204 else
1205 dir = fbuf;
1206 len -= p - fbuf + 3;
1207 fn = p + 3;
1208 } else
1209 fn = fbuf;
1210 /* Get rid of trailing "/" and "/.". */
1211 while (len) {
1212 if (fn[len - 1] == '/') {
1213 is_dot_dir = 1;
1214 if (!--len && !dir) {
1215 len++;
1216 break;
1217 }
1218 }
1219 else if (len >= 2 && fn[len - 1] == '.'
1220 && fn[len - 2] == '/') {
1221 is_dot_dir = 1;
1222 if (!(len -= 2) && !dir) {
1223 len++;
1224 break;
1225 }
1226 } else
1227 break;
1228 }
1229 if (len == 1 && fn[0] == '/')
1230 fn[len++] = '.';
1231 fn[len] = '\0';
1232 /* Reject a ".." dir in the active part of the path. */
1233 for (p = fn; (p = strstr(p, "..")) != NULL; p += 2) {
1234 if ((p[2] == '/' || p[2] == '\0')
1235 && (p == fn || p[-1] == '/')) {
1236 rprintf(FERROR,
1237 "found \"..\" dir in relative path: %s\n",
1238 fbuf);
1239 exit_cleanup(RERR_SYNTAX);
1240 }
1241 }
1242 }
1243
1244 if (!*fn) {
1245 len = 1;
1246 fn = ".";
1247 }
1248
1249 if (dir && *dir) {
1250 static const char *lastdir;
1251 static int lastdir_len;
1252
1253 strlcpy(olddir, curr_dir, sizeof olddir);
1254
1255 if (!push_dir(dir, 0)) {
1256 io_error |= IOERR_GENERAL;
1257 rsyserr(FERROR, errno, "push_dir %s failed",
1258 full_fname(dir));
1259 continue;
1260 }
1261
1262 if (lastdir && strcmp(lastdir, dir) == 0) {
1263 flist_dir = lastdir;
1264 flist_dir_len = lastdir_len;
1265 } else {
1266 flist_dir = lastdir = strdup(dir);
1267 flist_dir_len = lastdir_len = strlen(dir);
1268 }
1269 }
1270
1271 if (fn != fbuf)
1272 memmove(fbuf, fn, len + 1);
1273
1274 if (implied_dirs && (p=strrchr(fbuf,'/')) && p != fbuf) {
1275 /* Send the implied directories at the start of the
1276 * source spec, so we get their permissions right. */
1277 char *lp = lastpath, *slash = fbuf;
1278 *p = '\0';
1279 /* Skip any initial directories in our path that we
1280 * have in common with lastpath. */
1281 for (fn = fbuf; *fn && *lp == *fn; lp++, fn++) {
1282 if (*fn == '/')
1283 slash = fn;
1284 }
1285 *p = '/';
1286 if (fn != p || (*lp && *lp != '/')) {
1287 int save_copy_links = copy_links;
1288 int save_xfer_dirs = xfer_dirs;
1289 copy_links |= copy_unsafe_links;
1290 xfer_dirs = 1;
1291 while ((slash = strchr(slash+1, '/')) != 0) {
1292 *slash = '\0';
1293 send_file_name(f, flist, fbuf, NULL, 0);
1294 *slash = '/';
1295 }
1296 copy_links = save_copy_links;
1297 xfer_dirs = save_xfer_dirs;
1298 *p = '\0';
1299 strlcpy(lastpath, fbuf, sizeof lastpath);
1300 *p = '/';
1301 }
1302 }
1303
1304 if (one_file_system)
1305 filesystem_dev = st.st_dev;
1306
1307 if (recurse || (xfer_dirs && is_dot_dir)) {
1308 struct file_struct *file;
1309 file = send_file_name(f, flist, fbuf, &st, FLAG_TOP_DIR);
1310 if (file)
1311 send_if_directory(f, flist, file, fbuf, len);
1312 } else
1313 send_file_name(f, flist, fbuf, &st, 0);
1314
1315 if (olddir[0]) {
1316 flist_dir = NULL;
1317 flist_dir_len = 0;
1318 if (!pop_dir(olddir)) {
1319 rsyserr(FERROR, errno, "pop_dir %s failed",
1320 full_fname(olddir));
1321 exit_cleanup(RERR_FILESELECT);
1322 }
1323 }
1324 }
1325
1326 gettimeofday(&end_tv, NULL);
1327 stats.flist_buildtime = (int64)(end_tv.tv_sec - start_tv.tv_sec) * 1000
1328 + (end_tv.tv_usec - start_tv.tv_usec) / 1000;
1329 if (stats.flist_buildtime == 0)
1330 stats.flist_buildtime = 1;
1331 start_tv = end_tv;
1332
1333 send_file_entry(NULL, f);
1334
1335 if (show_filelist_p())
1336 finish_filelist_progress(flist);
1337
1338 gettimeofday(&end_tv, NULL);
1339 stats.flist_xfertime = (int64)(end_tv.tv_sec - start_tv.tv_sec) * 1000
1340 + (end_tv.tv_usec - start_tv.tv_usec) / 1000;
1341
1342 /* Sort the list without removing any duplicates. This allows the
1343 * receiving side to ask for any name they like, which gives us the
1344 * flexibility to change the way we unduplicate names in the future
1345 * without causing a compatibility problem with older versions. */
1346 clean_flist(flist, 0, 0);
1347
1348 if (!numeric_ids)
1349 send_uid_list(f);
1350
1351 /* send the io_error flag */
1352 write_int(f, lp_ignore_errors(module_id) ? 0 : io_error);
1353
1354 io_end_buffering();
1355 stats.flist_size = stats.total_written - start_write;
1356 stats.num_files = flist->count;
1357
1358 if (verbose > 3)
1359 output_flist(flist);
1360
1361 if (verbose > 2)
1362 rprintf(FINFO, "send_file_list done\n");
1363
1364 return flist;
1365}
1366
1367struct file_list *recv_file_list(int f)
1368{
1369 struct file_list *flist;
1370 unsigned short flags;
1371 int64 start_read;
1372
1373 rprintf(FLOG, "receiving file list\n");
1374 if (show_filelist_p())
1375 start_filelist_progress("receiving file list");
1376
1377 start_read = stats.total_read;
1378
1379 flist = flist_new(WITH_HLINK, "recv_file_list");
1380
1381
1382 while ((flags = read_byte(f)) != 0) {
1383 struct file_struct *file;
1384
1385 flist_expand(flist);
1386
1387 if (protocol_version >= 28 && (flags & XMIT_EXTENDED_FLAGS))
1388 flags |= read_byte(f) << 8;
1389 file = recv_file_entry(flist, flags, f);
1390
1391 if (S_ISREG(file->mode) || S_ISLNK(file->mode))
1392 stats.total_size += F_LENGTH(file);
1393
1394 flist->files[flist->count++] = file;
1395
1396 maybe_emit_filelist_progress(flist->count);
1397
1398 if (verbose > 2) {
1399 rprintf(FINFO, "recv_file_name(%s)\n",
1400 f_name(file, NULL));
1401 }
1402 }
1403 recv_file_entry(NULL, 0, 0); /* Signal that we're done. */
1404
1405 if (verbose > 2)
1406 rprintf(FINFO, "received %d names\n", flist->count);
1407
1408 if (show_filelist_p())
1409 finish_filelist_progress(flist);
1410
1411 clean_flist(flist, relative_paths, 1);
1412
1413 if (f >= 0) {
1414 recv_uid_list(f, flist);
1415
1416 /* Recv the io_error flag */
1417 if (lp_ignore_errors(module_id) || ignore_errors)
1418 read_int(f);
1419 else
1420 io_error |= read_int(f);
1421 }
1422
1423 if (verbose > 3)
1424 output_flist(flist);
1425
1426 if (list_only) {
1427 int i;
1428 for (i = 0; i < flist->count; i++)
1429 list_file_entry(flist->files[i]);
1430 }
1431
1432 if (verbose > 2)
1433 rprintf(FINFO, "recv_file_list done\n");
1434
1435 stats.flist_size = stats.total_read - start_read;
1436 stats.num_files = flist->count;
1437
1438 return flist;
1439}
1440
1441static int file_compare(struct file_struct **file1, struct file_struct **file2)
1442{
1443 return f_name_cmp(*file1, *file2);
1444}
1445
1446/* Search for an identically-named item in the file list. Note that the
1447 * items must agree in their directory-ness, or no match is returned. */
1448int flist_find(struct file_list *flist, struct file_struct *f)
1449{
1450 int low = flist->low, high = flist->high;
1451 int diff, mid, mid_up;
1452
1453 while (low <= high) {
1454 mid = (low + high) / 2;
1455 if (F_IS_ACTIVE(flist->files[mid]))
1456 mid_up = mid;
1457 else {
1458 /* Scan for the next non-empty entry using the cached
1459 * distance values. If the value isn't fully up-to-
1460 * date, update it. */
1461 mid_up = mid + flist->files[mid]->dir.depth;
1462 if (!F_IS_ACTIVE(flist->files[mid_up])) {
1463 do {
1464 mid_up += flist->files[mid_up]->dir.depth;
1465 } while (!F_IS_ACTIVE(flist->files[mid_up]));
1466 flist->files[mid]->dir.depth = mid_up - mid;
1467 }
1468 if (mid_up > high) {
1469 /* If there's nothing left above us, set high to
1470 * a non-empty entry below us and continue. */
1471 high = mid - (int)flist->files[mid]->len32;
1472 if (!F_IS_ACTIVE(flist->files[high])) {
1473 do {
1474 high -= (int)flist->files[high]->len32;
1475 } while (!F_IS_ACTIVE(flist->files[high]));
1476 flist->files[mid]->len32 = mid - high;
1477 }
1478 continue;
1479 }
1480 }
1481 diff = f_name_cmp(flist->files[mid_up], f);
1482 if (diff == 0) {
1483 if (protocol_version < 29
1484 && S_ISDIR(flist->files[mid_up]->mode)
1485 != S_ISDIR(f->mode))
1486 return -1;
1487 return mid_up;
1488 }
1489 if (diff < 0)
1490 low = mid_up + 1;
1491 else
1492 high = mid - 1;
1493 }
1494 return -1;
1495}
1496
1497/*
1498 * Free up any resources a file_struct has allocated
1499 * and clear the file.
1500 */
1501void clear_file(struct file_struct *file)
1502{
1503 memset((char*)file + EXTRA_LEN, 0, FILE_STRUCT_LEN - EXTRA_LEN + 1);
1504 /* In an empty entry, dir.depth is an offset to the next non-empty
1505 * entry. Likewise for len32 in the opposite direction. We assume
1506 * that we're alone for now since flist_find() will adjust the counts
1507 * it runs into that aren't up-to-date. */
1508 file->len32 = file->dir.depth = 1;
1509}
1510
1511/*
1512 * allocate a new file list
1513 */
1514struct file_list *flist_new(int with_hlink, char *msg)
1515{
1516 struct file_list *flist;
1517
1518 flist = new(struct file_list);
1519 if (!flist)
1520 out_of_memory(msg);
1521
1522 memset(flist, 0, sizeof flist[0]);
1523
1524 if (!(flist->file_pool = pool_create(FILE_EXTENT, 0, out_of_memory, POOL_INTERN)))
1525 out_of_memory(msg);
1526
1527#ifdef SUPPORT_HARD_LINKS
1528 if (with_hlink && preserve_hard_links) {
1529 if (!(flist->hlink_pool = pool_create(HLINK_EXTENT,
1530 sizeof (struct idev), out_of_memory, POOL_INTERN)))
1531 out_of_memory(msg);
1532 }
1533#endif
1534
1535 return flist;
1536}
1537
1538/*
1539 * free up all elements in a flist
1540 */
1541void flist_free(struct file_list *flist)
1542{
1543 pool_destroy(flist->file_pool);
1544 pool_destroy(flist->hlink_pool);
1545 free(flist->files);
1546 free(flist);
1547}
1548
1549/*
1550 * This routine ensures we don't have any duplicate names in our file list.
1551 * duplicate names can cause corruption because of the pipelining
1552 */
1553static void clean_flist(struct file_list *flist, int strip_root, int no_dups)
1554{
1555 char fbuf[MAXPATHLEN];
1556 int i, prev_i = 0;
1557
1558 if (!flist)
1559 return;
1560 if (flist->count == 0) {
1561 flist->high = -1;
1562 return;
1563 }
1564
1565 qsort(flist->files, flist->count,
1566 sizeof flist->files[0], (int (*)())file_compare);
1567
1568 for (i = no_dups? 0 : flist->count; i < flist->count; i++) {
1569 if (F_IS_ACTIVE(flist->files[i])) {
1570 prev_i = i;
1571 break;
1572 }
1573 }
1574 flist->low = prev_i;
1575 while (++i < flist->count) {
1576 int j;
1577 struct file_struct *file = flist->files[i];
1578
1579 if (!F_IS_ACTIVE(file))
1580 continue;
1581 if (f_name_cmp(file, flist->files[prev_i]) == 0)
1582 j = prev_i;
1583 else if (protocol_version >= 29 && S_ISDIR(file->mode)) {
1584 int save_mode = file->mode;
1585 /* Make sure that this directory doesn't duplicate a
1586 * non-directory earlier in the list. */
1587 flist->high = prev_i;
1588 file->mode = S_IFREG;
1589 j = flist_find(flist, file);
1590 file->mode = save_mode;
1591 } else
1592 j = -1;
1593 if (j >= 0) {
1594 struct file_struct *fp = flist->files[j];
1595 int keep, drop;
1596 /* If one is a dir and the other is not, we want to
1597 * keep the dir because it might have contents in the
1598 * list. */
1599 if (S_ISDIR(file->mode) != S_ISDIR(fp->mode)) {
1600 if (S_ISDIR(file->mode))
1601 keep = i, drop = j;
1602 else
1603 keep = j, drop = i;
1604 } else
1605 keep = j, drop = i;
1606 if (verbose > 1 && !am_server) {
1607 rprintf(FINFO,
1608 "removing duplicate name %s from file list (%d)\n",
1609 f_name(file, fbuf), drop);
1610 }
1611 /* Make sure we don't lose track of a user-specified
1612 * top directory. */
1613 flist->files[keep]->flags |= flist->files[drop]->flags
1614 & (FLAG_TOP_DIR|FLAG_XFER_DIR);
1615
1616 clear_file(flist->files[drop]);
1617
1618 if (keep == i) {
1619 if (flist->low == drop) {
1620 for (j = drop + 1;
1621 j < i && !F_IS_ACTIVE(flist->files[j]);
1622 j++) {}
1623 flist->low = j;
1624 }
1625 prev_i = i;
1626 }
1627 } else
1628 prev_i = i;
1629 }
1630 flist->high = no_dups ? prev_i : flist->count - 1;
1631
1632 if (strip_root) {
1633 /* We need to strip off the leading slashes for relative
1634 * paths, but this must be done _after_ the sorting phase. */
1635 for (i = flist->low; i <= flist->high; i++) {
1636 struct file_struct *file = flist->files[i];
1637
1638 if (!file->dirname)
1639 continue;
1640 while (*file->dirname == '/')
1641 file->dirname++;
1642 if (!*file->dirname)
1643 file->dirname = NULL;
1644 }
1645 }
1646
1647 if (prune_empty_dirs && no_dups) {
1648 int j, prev_depth = 0;
1649
1650 prev_i = 0; /* It's OK that this isn't really true. */
1651
1652 for (i = flist->low; i <= flist->high; i++) {
1653 struct file_struct *fp, *file = flist->files[i];
1654
1655 /* This temporarily abuses the dir.depth value for a
1656 * directory that is in a chain that might get pruned.
1657 * We restore the old value if it gets a reprieve. */
1658 if (S_ISDIR(file->mode) && file->dir.depth) {
1659 /* Dump empty dirs when coming back down. */
1660 for (j = prev_depth; j >= file->dir.depth; j--) {
1661 fp = flist->files[prev_i];
1662 if (fp->dir.depth >= 0)
1663 break;
1664 prev_i = -fp->dir.depth-1;
1665 clear_file(fp);
1666 }
1667 prev_depth = file->dir.depth;
1668 if (is_excluded(f_name(file, fbuf), 1,
1669 ALL_FILTERS)) {
1670 /* Keep dirs through this dir. */
1671 for (j = prev_depth-1; ; j--) {
1672 fp = flist->files[prev_i];
1673 if (fp->dir.depth >= 0)
1674 break;
1675 prev_i = -fp->dir.depth-1;
1676 fp->dir.depth = j;
1677 }
1678 } else
1679 file->dir.depth = -prev_i-1;
1680 prev_i = i;
1681 } else {
1682 /* Keep dirs through this non-dir. */
1683 for (j = prev_depth; ; j--) {
1684 fp = flist->files[prev_i];
1685 if (fp->dir.depth >= 0)
1686 break;
1687 prev_i = -fp->dir.depth-1;
1688 fp->dir.depth = j;
1689 }
1690 }
1691 }
1692 /* Dump empty all remaining empty dirs. */
1693 while (1) {
1694 struct file_struct *fp = flist->files[prev_i];
1695 if (fp->dir.depth >= 0)
1696 break;
1697 prev_i = -fp->dir.depth-1;
1698 clear_file(fp);
1699 }
1700
1701 for (i = flist->low; i <= flist->high; i++) {
1702 if (F_IS_ACTIVE(flist->files[i]))
1703 break;
1704 }
1705 flist->low = i;
1706 for (i = flist->high; i >= flist->low; i--) {
1707 if (F_IS_ACTIVE(flist->files[i]))
1708 break;
1709 }
1710 flist->high = i;
1711 }
1712}
1713
1714static void output_flist(struct file_list *flist)
1715{
1716 char uidbuf[16], gidbuf[16], depthbuf[16];
1717 struct file_struct *file;
1718 const char *dir, *slash, *name, *trail;
1719 const char *who = who_am_i();
1720 int i;
1721
1722 for (i = 0; i < flist->count; i++) {
1723 file = flist->files[i];
1724 if ((am_root || am_sender) && preserve_uid) {
1725 snprintf(uidbuf, sizeof uidbuf, " uid=%ld",
1726 (long)F_UID(file));
1727 } else
1728 *uidbuf = '\0';
1729 if (preserve_gid && F_GID(file) != GID_NONE) {
1730 snprintf(gidbuf, sizeof gidbuf, " gid=%ld",
1731 (long)F_GID(file));
1732 } else
1733 *gidbuf = '\0';
1734 if (!am_sender)
1735 snprintf(depthbuf, sizeof depthbuf, "%d", file->dir.depth);
1736 if (F_IS_ACTIVE(file)) {
1737 if ((dir = file->dirname) == NULL)
1738 dir = slash = "";
1739 else
1740 slash = "/";
1741 name = F_BASENAME(file);
1742 trail = S_ISDIR(file->mode) ? "/" : "";
1743 } else
1744 dir = slash = name = trail = "";
1745 rprintf(FINFO, "[%s] i=%d %s %s%s%s%s mode=0%o len=%.0f%s%s flags=%x\n",
1746 who, i, am_sender ? NS(file->dir.root) : depthbuf,
1747 dir, slash, name, trail, (int)file->mode,
1748 (double)F_LENGTH(file), uidbuf, gidbuf, file->flags);
1749 }
1750}
1751
1752enum fnc_state { s_DIR, s_SLASH, s_BASE, s_TRAILING };
1753enum fnc_type { t_PATH, t_ITEM };
1754
1755/* Compare the names of two file_struct entities, similar to how strcmp()
1756 * would do if it were operating on the joined strings.
1757 *
1758 * Some differences beginning with protocol_version 29: (1) directory names
1759 * are compared with an assumed trailing slash so that they compare in a
1760 * way that would cause them to sort immediately prior to any content they
1761 * may have; (2) a directory of any name compares after a non-directory of
1762 * any name at the same depth; (3) a directory with name "." compares prior
1763 * to anything else. These changes mean that a directory and a non-dir
1764 * with the same name will not compare as equal (protocol_version >= 29).
1765 *
1766 * The dirname component can be an empty string, but the basename component
1767 * cannot (and never is in the current codebase). The basename component
1768 * may be NULL (for a removed item), in which case it is considered to be
1769 * after any existing item. */
1770int f_name_cmp(struct file_struct *f1, struct file_struct *f2)
1771{
1772 int dif;
1773 const uchar *c1, *c2;
1774 enum fnc_state state1, state2;
1775 enum fnc_type type1, type2;
1776 enum fnc_type t_path = protocol_version >= 29 ? t_PATH : t_ITEM;
1777
1778 if (!f1 || !F_IS_ACTIVE(f1)) {
1779 if (!f2 || !F_IS_ACTIVE(f2))
1780 return 0;
1781 return -1;
1782 }
1783 if (!f2 || !F_IS_ACTIVE(f2))
1784 return 1;
1785
1786 c1 = (uchar*)f1->dirname;
1787 c2 = (uchar*)f2->dirname;
1788 if (c1 == c2)
1789 c1 = c2 = NULL;
1790 if (!c1) {
1791 type1 = S_ISDIR(f1->mode) ? t_path : t_ITEM;
1792 c1 = (uchar*)F_BASENAME(f1);
1793 if (type1 == t_PATH && *c1 == '.' && !c1[1]) {
1794 type1 = t_ITEM;
1795 state1 = s_TRAILING;
1796 c1 = (uchar*)"";
1797 } else
1798 state1 = s_BASE;
1799 } else {
1800 type1 = t_path;
1801 state1 = s_DIR;
1802 }
1803 if (!c2) {
1804 type2 = S_ISDIR(f2->mode) ? t_path : t_ITEM;
1805 c2 = (uchar*)F_BASENAME(f2);
1806 if (type2 == t_PATH && *c2 == '.' && !c2[1]) {
1807 type2 = t_ITEM;
1808 state2 = s_TRAILING;
1809 c2 = (uchar*)"";
1810 } else
1811 state2 = s_BASE;
1812 } else {
1813 type2 = t_path;
1814 state2 = s_DIR;
1815 }
1816
1817 if (type1 != type2)
1818 return type1 == t_PATH ? 1 : -1;
1819
1820 do {
1821 if (!*c1) {
1822 switch (state1) {
1823 case s_DIR:
1824 state1 = s_SLASH;
1825 c1 = (uchar*)"/";
1826 break;
1827 case s_SLASH:
1828 type1 = S_ISDIR(f1->mode) ? t_path : t_ITEM;
1829 c1 = (uchar*)F_BASENAME(f1);
1830 if (type1 == t_PATH && *c1 == '.' && !c1[1]) {
1831 type1 = t_ITEM;
1832 state1 = s_TRAILING;
1833 c1 = (uchar*)"";
1834 } else
1835 state1 = s_BASE;
1836 break;
1837 case s_BASE:
1838 state1 = s_TRAILING;
1839 if (type1 == t_PATH) {
1840 c1 = (uchar*)"/";
1841 break;
1842 }
1843 /* FALL THROUGH */
1844 case s_TRAILING:
1845 type1 = t_ITEM;
1846 break;
1847 }
1848 if (*c2 && type1 != type2)
1849 return type1 == t_PATH ? 1 : -1;
1850 }
1851 if (!*c2) {
1852 switch (state2) {
1853 case s_DIR:
1854 state2 = s_SLASH;
1855 c2 = (uchar*)"/";
1856 break;
1857 case s_SLASH:
1858 type2 = S_ISDIR(f2->mode) ? t_path : t_ITEM;
1859 c2 = (uchar*)F_BASENAME(f2);
1860 if (type2 == t_PATH && *c2 == '.' && !c2[1]) {
1861 type2 = t_ITEM;
1862 state2 = s_TRAILING;
1863 c2 = (uchar*)"";
1864 } else
1865 state2 = s_BASE;
1866 break;
1867 case s_BASE:
1868 state2 = s_TRAILING;
1869 if (type2 == t_PATH) {
1870 c2 = (uchar*)"/";
1871 break;
1872 }
1873 /* FALL THROUGH */
1874 case s_TRAILING:
1875 if (!*c1)
1876 return 0;
1877 type2 = t_ITEM;
1878 break;
1879 }
1880 if (type1 != type2)
1881 return type1 == t_PATH ? 1 : -1;
1882 }
1883 } while ((dif = (int)*c1++ - (int)*c2++) == 0);
1884
1885 return dif;
1886}
1887
1888char *f_name_buf(void)
1889{
1890 static char names[5][MAXPATHLEN];
1891 static unsigned int n;
1892
1893 n = (n + 1) % (sizeof names / sizeof names[0]);
1894
1895 return names[n];
1896}
1897
1898/* Return a copy of the full filename of a flist entry, using the indicated
1899 * buffer or one of 5 static buffers if fbuf is NULL. No size-checking is
1900 * done because we checked the size when creating the file_struct entry.
1901 */
1902char *f_name(struct file_struct *f, char *fbuf)
1903{
1904 if (!f || !F_IS_ACTIVE(f))
1905 return NULL;
1906
1907 if (!fbuf)
1908 fbuf = f_name_buf();
1909
1910 if (f->dirname) {
1911 int len = strlen(f->dirname);
1912 memcpy(fbuf, f->dirname, len);
1913 fbuf[len] = '/';
1914 strlcpy(fbuf + len + 1, F_BASENAME(f), MAXPATHLEN - (len + 1));
1915 } else
1916 strlcpy(fbuf, F_BASENAME(f), MAXPATHLEN);
1917
1918 return fbuf;
1919}
1920
1921/* Do a non-recursive scan of the named directory, possibly ignoring all
1922 * exclude rules except for the daemon's. If "dlen" is >=0, it is the length
1923 * of the dirname string, and also indicates that "dirname" is a MAXPATHLEN
1924 * buffer (the functions we call will append names onto the end, but the old
1925 * dir value will be restored on exit). */
1926struct file_list *get_dirlist(char *dirname, int dlen,
1927 int ignore_filter_rules)
1928{
1929 struct file_list *dirlist;
1930 char dirbuf[MAXPATHLEN];
1931 int save_recurse = recurse;
1932 int save_xfer_dirs = xfer_dirs;
1933
1934 if (dlen < 0) {
1935 dlen = strlcpy(dirbuf, dirname, MAXPATHLEN);
1936 if (dlen >= MAXPATHLEN)
1937 return NULL;
1938 dirname = dirbuf;
1939 }
1940
1941 dirlist = flist_new(WITHOUT_HLINK, "get_dirlist");
1942
1943 recurse = 0;
1944 xfer_dirs = 1;
1945 send_directory(ignore_filter_rules ? -2 : -1, dirlist, dirname, dlen);
1946 xfer_dirs = save_xfer_dirs;
1947 recurse = save_recurse;
1948 if (do_progress)
1949 flist_count_offset += dirlist->count;
1950
1951 clean_flist(dirlist, 0, 0);
1952
1953 if (verbose > 3)
1954 output_flist(dirlist);
1955
1956 return dirlist;
1957}