- The XMIT_SAME_DEV define is now pre-protocol-30 specific.
[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;
61extern alloc_pool_t hlink_pool;
62
63extern char curr_dir[MAXPATHLEN];
64
65extern struct chmod_mode_struct *chmod_modes;
66
67extern struct filter_list_struct filter_list;
68extern struct filter_list_struct server_filter_list;
69
70int io_error;
71int checksum_len;
72dev_t filesystem_dev; /* used to implement -x */
73
74/* The tmp_* vars are used as a cache area by make_file() to store data
75 * that the sender doesn't need to remember in its file list. The data
76 * will survive just long enough to be used by send_file_entry(). */
77static dev_t tmp_rdev;
78#ifdef SUPPORT_HARD_LINKS
79static struct idev tmp_idev;
80#endif
81static char tmp_sum[MD4_SUM_LENGTH];
82
83static char empty_sum[MD4_SUM_LENGTH];
84static int flist_count_offset; /* for --delete --progress */
85
86static void clean_flist(struct file_list *flist, int strip_root, int no_dups);
87static void output_flist(struct file_list *flist);
88
89void init_flist(void)
90{
91 if (verbose > 4) {
92 rprintf(FINFO, "FILE_STRUCT_LEN=%d, EXTRA_LEN=%d\n",
93 (int)FILE_STRUCT_LEN, (int)EXTRA_LEN);
94 }
95 checksum_len = protocol_version < 21 ? 2 : MD4_SUM_LENGTH;
96}
97
98static int show_filelist_p(void)
99{
100 return verbose && xfer_dirs && !am_server;
101}
102
103static void start_filelist_progress(char *kind)
104{
105 rprintf(FCLIENT, "%s ... ", kind);
106 if (verbose > 1 || do_progress)
107 rprintf(FCLIENT, "\n");
108 rflush(FINFO);
109}
110
111static void emit_filelist_progress(int count)
112{
113 rprintf(FCLIENT, " %d files...\r", count);
114}
115
116static void maybe_emit_filelist_progress(int count)
117{
118 if (do_progress && show_filelist_p() && (count % 100) == 0)
119 emit_filelist_progress(count);
120}
121
122static void finish_filelist_progress(const struct file_list *flist)
123{
124 if (do_progress) {
125 /* This overwrites the progress line */
126 rprintf(FINFO, "%d file%sto consider\n",
127 flist->count, flist->count == 1 ? " " : "s ");
128 } else
129 rprintf(FINFO, "done\n");
130}
131
132void show_flist_stats(void)
133{
134 /* Nothing yet */
135}
136
137static void list_file_entry(struct file_struct *f)
138{
139 char permbuf[PERMSTRING_SIZE];
140 double len;
141
142 if (!F_IS_ACTIVE(f)) {
143 /* this can happen if duplicate names were removed */
144 return;
145 }
146
147 permstring(permbuf, f->mode);
148 len = F_LENGTH(f);
149
150#ifdef SUPPORT_LINKS
151 if (preserve_links && S_ISLNK(f->mode)) {
152 rprintf(FINFO, "%s %11.0f %s %s -> %s\n",
153 permbuf, len, timestring(f->modtime),
154 f_name(f, NULL), F_SYMLINK(f));
155 } else
156#endif
157 {
158 rprintf(FINFO, "%s %11.0f %s %s\n",
159 permbuf, len, timestring(f->modtime),
160 f_name(f, NULL));
161 }
162}
163
164/* Stat either a symlink or its referent, depending on the settings of
165 * copy_links, copy_unsafe_links, etc. Returns -1 on error, 0 on success.
166 *
167 * If path is the name of a symlink, then the linkbuf buffer (which must hold
168 * MAXPATHLEN chars) will be set to the symlink's target string.
169 *
170 * The stat structure pointed to by stp will contain information about the
171 * link or the referent as appropriate, if they exist. */
172static int readlink_stat(const char *path, STRUCT_STAT *stp, char *linkbuf)
173{
174#ifdef SUPPORT_LINKS
175 if (link_stat(path, stp, copy_dirlinks) < 0)
176 return -1;
177 if (S_ISLNK(stp->st_mode)) {
178 int llen = readlink(path, linkbuf, MAXPATHLEN - 1);
179 if (llen < 0)
180 return -1;
181 linkbuf[llen] = '\0';
182 if (copy_unsafe_links && unsafe_symlink(linkbuf, path)) {
183 if (verbose > 1) {
184 rprintf(FINFO,"copying unsafe symlink \"%s\" -> \"%s\"\n",
185 path, linkbuf);
186 }
187 return do_stat(path, stp);
188 }
189 }
190 return 0;
191#else
192 return do_stat(path, stp);
193#endif
194}
195
196int link_stat(const char *path, STRUCT_STAT *stp, int follow_dirlinks)
197{
198#ifdef SUPPORT_LINKS
199 if (copy_links)
200 return do_stat(path, stp);
201 if (do_lstat(path, stp) < 0)
202 return -1;
203 if (follow_dirlinks && S_ISLNK(stp->st_mode)) {
204 STRUCT_STAT st;
205 if (do_stat(path, &st) == 0 && S_ISDIR(st.st_mode))
206 *stp = st;
207 }
208 return 0;
209#else
210 return do_stat(path, stp);
211#endif
212}
213
214/* This function is used to check if a file should be included/excluded
215 * from the list of files based on its name and type etc. The value of
216 * filter_level is set to either SERVER_FILTERS or ALL_FILTERS. */
217static int is_excluded(char *fname, int is_dir, int filter_level)
218{
219#if 0 /* This currently never happens, so avoid a useless compare. */
220 if (filter_level == NO_FILTERS)
221 return 0;
222#endif
223 if (fname) {
224 /* never exclude '.', even if somebody does --exclude '*' */
225 if (fname[0] == '.' && !fname[1])
226 return 0;
227 /* Handle the -R version of the '.' dir. */
228 if (fname[0] == '/') {
229 int len = strlen(fname);
230 if (fname[len-1] == '.' && fname[len-2] == '/')
231 return 0;
232 }
233 }
234 if (server_filter_list.head
235 && check_filter(&server_filter_list, fname, is_dir) < 0)
236 return 1;
237 if (filter_level != ALL_FILTERS)
238 return 0;
239 if (filter_list.head
240 && check_filter(&filter_list, fname, is_dir) < 0)
241 return 1;
242 return 0;
243}
244
245static int to_wire_mode(mode_t mode)
246{
247#ifdef SUPPORT_LINKS
248#if _S_IFLNK != 0120000
249 if (S_ISLNK(mode))
250 return (mode & ~(_S_IFMT)) | 0120000;
251#endif
252#endif
253 return mode;
254}
255
256static mode_t from_wire_mode(int mode)
257{
258#if _S_IFLNK != 0120000
259 if ((mode & (_S_IFMT)) == 0120000)
260 return (mode & ~(_S_IFMT)) | _S_IFLNK;
261#endif
262 return mode;
263}
264
265static void send_directory(int f, struct file_list *flist,
266 char *fbuf, int len);
267
268static const char *flist_dir;
269static int flist_dir_len;
270
271
272/**
273 * Make sure @p flist is big enough to hold at least @p flist->count
274 * entries.
275 **/
276void flist_expand(struct file_list *flist)
277{
278 struct file_struct **new_ptr;
279
280 if (flist->count < flist->malloced)
281 return;
282
283 if (flist->malloced < FLIST_START)
284 flist->malloced = FLIST_START;
285 else if (flist->malloced >= FLIST_LINEAR)
286 flist->malloced += FLIST_LINEAR;
287 else
288 flist->malloced *= 2;
289
290 /*
291 * In case count jumped or we are starting the list
292 * with a known size just set it.
293 */
294 if (flist->malloced < flist->count)
295 flist->malloced = flist->count;
296
297 new_ptr = realloc_array(flist->files, struct file_struct *,
298 flist->malloced);
299
300 if (verbose >= 2 && flist->malloced != FLIST_START) {
301 rprintf(FCLIENT, "[%s] expand file_list to %.0f bytes, did%s move\n",
302 who_am_i(),
303 (double)sizeof flist->files[0] * flist->malloced,
304 (new_ptr == flist->files) ? " not" : "");
305 }
306
307 flist->files = new_ptr;
308
309 if (!flist->files)
310 out_of_memory("flist_expand");
311}
312
313static void send_file_entry(struct file_struct *file, int f)
314{
315 unsigned short flags;
316 static time_t modtime;
317 static mode_t mode;
318 static int64 dev;
319 static dev_t rdev;
320 static uint32 rdev_major;
321 static uid_t uid;
322 static gid_t gid;
323 static char lastname[MAXPATHLEN];
324 char fname[MAXPATHLEN];
325 int l1, l2;
326
327 if (f < 0)
328 return;
329
330 if (!file) {
331 write_byte(f, 0);
332 modtime = 0, mode = 0;
333 dev = 0, rdev = MAKEDEV(0, 0);
334 rdev_major = 0;
335 uid = 0, gid = 0;
336 *lastname = '\0';
337 return;
338 }
339
340 f_name(file, fname);
341
342 flags = file->flags & FLAG_TOP_DIR; /* FLAG_TOP_DIR == XMIT_TOP_DIR */
343
344 if (file->mode == mode)
345 flags |= XMIT_SAME_MODE;
346 else
347 mode = file->mode;
348 if ((preserve_devices && IS_DEVICE(mode))
349 || (preserve_specials && IS_SPECIAL(mode))) {
350 if (protocol_version < 28) {
351 if (tmp_rdev == rdev)
352 flags |= XMIT_SAME_RDEV_pre28;
353 else
354 rdev = tmp_rdev;
355 } else {
356 rdev = tmp_rdev;
357 if ((uint32)major(rdev) == rdev_major)
358 flags |= XMIT_SAME_RDEV_MAJOR;
359 else
360 rdev_major = major(rdev);
361 if ((uint32)minor(rdev) <= 0xFFu)
362 flags |= XMIT_RDEV_MINOR_IS_SMALL;
363 }
364 } else if (protocol_version < 28)
365 rdev = MAKEDEV(0, 0);
366 if (preserve_uid) {
367 if (F_UID(file) == uid)
368 flags |= XMIT_SAME_UID;
369 else
370 uid = F_UID(file);
371 }
372 if (preserve_gid) {
373 if (F_GID(file) == gid)
374 flags |= XMIT_SAME_GID;
375 else
376 gid = F_GID(file);
377 }
378 if (file->modtime == modtime)
379 flags |= XMIT_SAME_TIME;
380 else
381 modtime = file->modtime;
382
383#ifdef SUPPORT_HARD_LINKS
384 if (tmp_idev.dev != 0) {
385 if (tmp_idev.dev == dev) {
386 if (protocol_version >= 28)
387 flags |= XMIT_SAME_DEV;
388 } else
389 dev = tmp_idev.dev;
390 flags |= XMIT_HLINKED;
391 }
392#endif
393
394 for (l1 = 0;
395 lastname[l1] && (fname[l1] == lastname[l1]) && (l1 < 255);
396 l1++) {}
397 l2 = strlen(fname+l1);
398
399 if (l1 > 0)
400 flags |= XMIT_SAME_NAME;
401 if (l2 > 255)
402 flags |= XMIT_LONG_NAME;
403
404 /* We must make sure we don't send a zero flag byte or the
405 * other end will terminate the flist transfer. Note that
406 * the use of XMIT_TOP_DIR on a non-dir has no meaning, so
407 * it's harmless way to add a bit to the first flag byte. */
408 if (protocol_version >= 28) {
409 if (!flags && !S_ISDIR(mode))
410 flags |= XMIT_TOP_DIR;
411 if ((flags & 0xFF00) || !flags) {
412 flags |= XMIT_EXTENDED_FLAGS;
413 write_shortint(f, flags);
414 } else
415 write_byte(f, flags);
416 } else {
417 if (!(flags & 0xFF))
418 flags |= S_ISDIR(mode) ? XMIT_LONG_NAME : XMIT_TOP_DIR;
419 write_byte(f, flags);
420 }
421 if (flags & XMIT_SAME_NAME)
422 write_byte(f, l1);
423 if (flags & XMIT_LONG_NAME)
424 write_int(f, l2);
425 else
426 write_byte(f, l2);
427 write_buf(f, fname + l1, l2);
428
429 write_longint(f, F_LENGTH(file));
430 if (!(flags & XMIT_SAME_TIME))
431 write_int(f, modtime);
432 if (!(flags & XMIT_SAME_MODE))
433 write_int(f, to_wire_mode(mode));
434 if (preserve_uid && !(flags & XMIT_SAME_UID)) {
435 if (!numeric_ids)
436 add_uid(uid);
437 write_int(f, uid);
438 }
439 if (preserve_gid && !(flags & XMIT_SAME_GID)) {
440 if (!numeric_ids)
441 add_gid(gid);
442 write_int(f, gid);
443 }
444 if ((preserve_devices && IS_DEVICE(mode))
445 || (preserve_specials && IS_SPECIAL(mode))) {
446 if (protocol_version < 28) {
447 if (!(flags & XMIT_SAME_RDEV_pre28))
448 write_int(f, (int)rdev);
449 } else {
450 if (!(flags & XMIT_SAME_RDEV_MAJOR))
451 write_int(f, major(rdev));
452 if (flags & XMIT_RDEV_MINOR_IS_SMALL)
453 write_byte(f, minor(rdev));
454 else
455 write_int(f, minor(rdev));
456 }
457 }
458
459#ifdef SUPPORT_LINKS
460 if (preserve_links && S_ISLNK(mode)) {
461 const char *sl = F_SYMLINK(file);
462 int len = strlen(sl);
463 write_int(f, len);
464 write_buf(f, sl, len);
465 }
466#endif
467
468#ifdef SUPPORT_HARD_LINKS
469 if (tmp_idev.dev != 0) {
470 if (protocol_version < 26) {
471 /* 32-bit dev_t and ino_t */
472 write_int(f, (int32)dev);
473 write_int(f, (int32)tmp_idev.ino);
474 } else {
475 /* 64-bit dev_t and ino_t */
476 if (!(flags & XMIT_SAME_DEV))
477 write_longint(f, dev);
478 write_longint(f, tmp_idev.ino);
479 }
480 }
481#endif
482
483 if (always_checksum && (S_ISREG(mode) || protocol_version < 28)) {
484 const char *sum;
485 if (S_ISREG(mode))
486 sum = tmp_sum;
487 else {
488 /* Prior to 28, we sent a useless set of nulls. */
489 sum = empty_sum;
490 }
491 write_buf(f, sum, checksum_len);
492 }
493
494 strlcpy(lastname, fname, MAXPATHLEN);
495}
496
497static struct file_struct *recv_file_entry(struct file_list *flist,
498 unsigned short flags, int f)
499{
500 static time_t modtime;
501 static mode_t mode;
502 static int64 dev;
503 static dev_t rdev;
504 static uint32 rdev_major;
505 static uid_t uid;
506 static gid_t gid;
507 static char lastname[MAXPATHLEN], *lastdir;
508 static int lastdir_depth, lastdir_len = -1;
509 static unsigned int del_hier_name_len = 0;
510 static int in_del_hier = 0;
511 char thisname[MAXPATHLEN];
512 unsigned int l1 = 0, l2 = 0;
513 int alloc_len, basename_len, dirname_len, linkname_len;
514 int extra_len = (flist_extra_cnt - 1) * EXTRA_LEN;
515 OFF_T file_length;
516 char *basename, *dirname, *bp;
517 struct file_struct *file;
518
519 if (!flist) {
520 modtime = 0, mode = 0;
521 dev = 0, rdev = MAKEDEV(0, 0);
522 rdev_major = 0;
523 uid = 0, gid = 0;
524 *lastname = '\0';
525 lastdir_len = -1;
526 in_del_hier = 0;
527 return NULL;
528 }
529
530 if (flags & XMIT_SAME_NAME)
531 l1 = read_byte(f);
532
533 if (flags & XMIT_LONG_NAME)
534 l2 = read_int(f);
535 else
536 l2 = read_byte(f);
537
538 if (l2 >= MAXPATHLEN - l1) {
539 rprintf(FERROR,
540 "overflow: flags=0x%x l1=%d l2=%d lastname=%s [%s]\n",
541 flags, l1, l2, lastname, who_am_i());
542 overflow_exit("recv_file_entry");
543 }
544
545 strlcpy(thisname, lastname, l1 + 1);
546 read_sbuf(f, &thisname[l1], l2);
547 thisname[l1 + l2] = 0;
548
549 strlcpy(lastname, thisname, MAXPATHLEN);
550
551 clean_fname(thisname, 0);
552
553 if (sanitize_paths)
554 sanitize_path(thisname, thisname, "", 0, NULL);
555
556 if ((basename = strrchr(thisname, '/')) != NULL) {
557 dirname_len = ++basename - thisname; /* counts future '\0' */
558 if (lastdir_len == dirname_len - 1
559 && strncmp(thisname, lastdir, lastdir_len) == 0) {
560 dirname = lastdir;
561 dirname_len = 0; /* indicates no copy is needed */
562 } else
563 dirname = thisname;
564 } else {
565 basename = thisname;
566 dirname = NULL;
567 dirname_len = 0;
568 }
569 basename_len = strlen(basename) + 1; /* count the '\0' */
570
571 file_length = read_longint(f);
572 if (!(flags & XMIT_SAME_TIME))
573 modtime = (time_t)read_int(f);
574 if (!(flags & XMIT_SAME_MODE))
575 mode = from_wire_mode(read_int(f));
576
577 if (chmod_modes && !S_ISLNK(mode))
578 mode = tweak_mode(mode, chmod_modes);
579
580 if (preserve_uid && !(flags & XMIT_SAME_UID))
581 uid = (uid_t)read_int(f);
582 if (preserve_gid && !(flags & XMIT_SAME_GID))
583 gid = (gid_t)read_int(f);
584
585 if ((preserve_devices && IS_DEVICE(mode))
586 || (preserve_specials && IS_SPECIAL(mode))) {
587 if (protocol_version < 28) {
588 if (!(flags & XMIT_SAME_RDEV_pre28))
589 rdev = (dev_t)read_int(f);
590 } else {
591 uint32 rdev_minor;
592 if (!(flags & XMIT_SAME_RDEV_MAJOR))
593 rdev_major = read_int(f);
594 if (flags & XMIT_RDEV_MINOR_IS_SMALL)
595 rdev_minor = read_byte(f);
596 else
597 rdev_minor = read_int(f);
598 rdev = MAKEDEV(rdev_major, rdev_minor);
599 }
600 extra_len += 2 * EXTRA_LEN;
601 file_length = 0;
602 } else if (protocol_version < 28)
603 rdev = MAKEDEV(0, 0);
604
605#ifdef SUPPORT_LINKS
606 if (preserve_links && S_ISLNK(mode)) {
607 linkname_len = read_int(f) + 1; /* count the '\0' */
608 if (linkname_len <= 0 || linkname_len > MAXPATHLEN) {
609 rprintf(FERROR, "overflow: linkname_len=%d\n",
610 linkname_len - 1);
611 overflow_exit("recv_file_entry");
612 }
613 }
614 else
615#endif
616 linkname_len = 0;
617
618#ifdef SUPPORT_HARD_LINKS
619 if (preserve_hard_links) {
620 if (protocol_version < 28 && S_ISREG(mode))
621 flags |= XMIT_HLINKED;
622 if (flags & XMIT_HLINKED)
623 extra_len += EXTRA_LEN;
624 }
625#endif
626
627 if (always_checksum && S_ISREG(mode))
628 extra_len += SUM_EXTRA_CNT * EXTRA_LEN;
629
630 if (file_length > 0xFFFFFFFFu && S_ISREG(mode))
631 extra_len += EXTRA_LEN;
632
633 alloc_len = FILE_STRUCT_LEN + extra_len + basename_len + dirname_len
634 + linkname_len;
635 bp = pool_alloc(flist->file_pool, alloc_len, "recv_file_entry");
636
637 memset(bp, 0, FILE_STRUCT_LEN + extra_len);
638 bp += extra_len;
639 file = (struct file_struct *)bp;
640 bp += FILE_STRUCT_LEN;
641
642 memcpy(bp, basename, basename_len);
643 bp += basename_len + linkname_len; /* skip space for symlink too */
644
645#ifdef SUPPORT_HARD_LINKS
646 if (flags & XMIT_HLINKED)
647 file->flags |= FLAG_HLINKED;
648#endif
649 file->modtime = modtime;
650 file->len32 = (uint32)file_length;
651 if (file_length > 0xFFFFFFFFu && S_ISREG(mode)) {
652 file->flags |= FLAG_LENGTH64;
653 OPT_EXTRA(file, 0)->unum = (uint32)(file_length >> 32);
654 }
655 file->mode = mode;
656 if (preserve_uid)
657 F_UID(file) = uid;
658 if (preserve_gid)
659 F_GID(file) = gid;
660
661 if (dirname_len) {
662 file->dirname = lastdir = bp;
663 lastdir_len = dirname_len - 1;
664 memcpy(bp, dirname, dirname_len - 1);
665 bp += dirname_len;
666 bp[-1] = '\0';
667 lastdir_depth = count_dir_elements(lastdir);
668 file->dir.depth = lastdir_depth + 1;
669 } else if (dirname) {
670 file->dirname = dirname; /* we're reusing lastname */
671 file->dir.depth = lastdir_depth + 1;
672 } else
673 file->dir.depth = 1;
674
675 if (S_ISDIR(mode)) {
676 if (basename_len == 1+1 && *basename == '.') /* +1 for '\0' */
677 file->dir.depth--;
678 if (flags & XMIT_TOP_DIR) {
679 in_del_hier = recurse;
680 del_hier_name_len = file->dir.depth == 0 ? 0 : l1 + l2;
681 if (relative_paths && del_hier_name_len > 2
682 && lastname[del_hier_name_len-1] == '.'
683 && lastname[del_hier_name_len-2] == '/')
684 del_hier_name_len -= 2;
685 file->flags |= FLAG_TOP_DIR | FLAG_XFER_DIR;
686 } else if (in_del_hier) {
687 if (!relative_paths || !del_hier_name_len
688 || (l1 >= del_hier_name_len
689 && lastname[del_hier_name_len] == '/'))
690 file->flags |= FLAG_XFER_DIR;
691 else
692 in_del_hier = 0;
693 }
694 }
695
696 if ((preserve_devices && IS_DEVICE(mode))
697 || (preserve_specials && IS_SPECIAL(mode))) {
698 uint32 *devp = F_RDEV_P(file);
699 DEV_MAJOR(devp) = major(rdev);
700 DEV_MINOR(devp) = minor(rdev);
701 }
702
703#ifdef SUPPORT_LINKS
704 if (linkname_len) {
705 bp = (char*)F_BASENAME(file) + basename_len;
706 read_sbuf(f, bp, linkname_len - 1);
707 if (sanitize_paths)
708 sanitize_path(bp, bp, "", lastdir_depth, NULL);
709 }
710#endif
711
712#ifdef SUPPORT_HARD_LINKS
713 if (preserve_hard_links && flags & XMIT_HLINKED) {
714 struct idev *idevp = pool_talloc(hlink_pool, struct idev,
715 1, "inode_table");
716 F_HL_IDEV(file) = idevp;
717 if (protocol_version < 26) {
718 idevp->dev = read_int(f);
719 idevp->ino = read_int(f);
720 } else {
721 if (!(flags & XMIT_SAME_DEV))
722 dev = read_longint(f);
723 idevp->dev = dev;
724 idevp->ino = read_longint(f);
725 }
726 }
727#endif
728
729 if (always_checksum && (S_ISREG(mode) || protocol_version < 28)) {
730 if (S_ISREG(mode))
731 bp = (char*)F_SUM(file);
732 else {
733 /* Prior to 28, we get a useless set of nulls. */
734 bp = tmp_sum;
735 }
736 read_buf(f, bp, checksum_len);
737 }
738
739 return file;
740}
741
742/**
743 * Create a file_struct for a named file by reading its stat()
744 * information and performing extensive checks against global
745 * options.
746 *
747 * @return the new file, or NULL if there was an error or this file
748 * should be excluded.
749 *
750 * @todo There is a small optimization opportunity here to avoid
751 * stat()ing the file in some circumstances, which has a certain cost.
752 * We are called immediately after doing readdir(), and so we may
753 * already know the d_type of the file. We could for example avoid
754 * statting directories if we're not recursing, but this is not a very
755 * important case. Some systems may not have d_type.
756 **/
757struct file_struct *make_file(const char *fname, struct file_list *flist,
758 STRUCT_STAT *stp, int flags, int filter_level)
759{
760 static char *lastdir;
761 static int lastdir_len = -1;
762 struct file_struct *file;
763 STRUCT_STAT st;
764 char thisname[MAXPATHLEN];
765 char linkname[MAXPATHLEN];
766 int alloc_len, basename_len, dirname_len, linkname_len;
767 int extra_len = (flist_extra_cnt - 1) * EXTRA_LEN;
768 char *basename, *dirname, *bp;
769
770 if (!flist || !flist->count) /* Ignore lastdir when invalid. */
771 lastdir_len = -1;
772
773 if (strlcpy(thisname, fname, sizeof thisname)
774 >= sizeof thisname - flist_dir_len) {
775 rprintf(FINFO, "skipping overly long name: %s\n", fname);
776 return NULL;
777 }
778 clean_fname(thisname, 0);
779 if (sanitize_paths)
780 sanitize_path(thisname, thisname, "", 0, NULL);
781
782 if (stp && S_ISDIR(stp->st_mode)) {
783 st = *stp; /* Needed for "symlink/." with --relative. */
784 *linkname = '\0'; /* make IBM code checker happy */
785 } else if (readlink_stat(thisname, &st, linkname) != 0) {
786 int save_errno = errno;
787 /* See if file is excluded before reporting an error. */
788 if (filter_level != NO_FILTERS
789 && (is_excluded(thisname, 0, filter_level)
790 || is_excluded(thisname, 1, filter_level))) {
791 if (ignore_perishable && save_errno != ENOENT)
792 non_perishable_cnt++;
793 return NULL;
794 }
795 if (save_errno == ENOENT) {
796#ifdef SUPPORT_LINKS
797 /* Avoid "vanished" error if symlink points nowhere. */
798 if (copy_links && do_lstat(thisname, &st) == 0
799 && S_ISLNK(st.st_mode)) {
800 io_error |= IOERR_GENERAL;
801 rprintf(FERROR, "symlink has no referent: %s\n",
802 full_fname(thisname));
803 } else
804#endif
805 {
806 enum logcode c = am_daemon && protocol_version < 28
807 ? FERROR : FINFO;
808 io_error |= IOERR_VANISHED;
809 rprintf(c, "file has vanished: %s\n",
810 full_fname(thisname));
811 }
812 } else {
813 io_error |= IOERR_GENERAL;
814 rsyserr(FERROR, save_errno, "readlink %s failed",
815 full_fname(thisname));
816 }
817 return NULL;
818 }
819
820 /* backup.c calls us with filter_level set to NO_FILTERS. */
821 if (filter_level == NO_FILTERS)
822 goto skip_filters;
823
824 if (S_ISDIR(st.st_mode) && !xfer_dirs) {
825 rprintf(FINFO, "skipping directory %s\n", thisname);
826 return NULL;
827 }
828
829 /* -x only affects directories because we need to avoid recursing
830 * into a mount-point directory, not to avoid copying a symlinked
831 * file if -L (or similar) was specified. */
832 if (one_file_system && st.st_dev != filesystem_dev
833 && S_ISDIR(st.st_mode)) {
834 if (one_file_system > 1) {
835 if (verbose > 2) {
836 rprintf(FINFO, "skipping mount-point dir %s\n",
837 thisname);
838 }
839 return NULL;
840 }
841 flags |= FLAG_MOUNT_DIR;
842 }
843
844 if (is_excluded(thisname, S_ISDIR(st.st_mode) != 0, filter_level)) {
845 if (ignore_perishable)
846 non_perishable_cnt++;
847 return NULL;
848 }
849
850 if (lp_ignore_nonreadable(module_id)) {
851#ifdef SUPPORT_LINKS
852 if (!S_ISLNK(st.st_mode))
853#endif
854 if (access(thisname, R_OK) != 0)
855 return NULL;
856 }
857
858 skip_filters:
859
860 if (verbose > 2) {
861 rprintf(FINFO, "[%s] make_file(%s,*,%d)\n",
862 who_am_i(), thisname, filter_level);
863 }
864
865 if ((basename = strrchr(thisname, '/')) != NULL) {
866 dirname_len = ++basename - thisname; /* counts future '\0' */
867 if (lastdir_len == dirname_len - 1
868 && strncmp(thisname, lastdir, lastdir_len) == 0) {
869 dirname = lastdir;
870 dirname_len = 0; /* indicates no copy is needed */
871 } else
872 dirname = thisname;
873 } else {
874 basename = thisname;
875 dirname = NULL;
876 dirname_len = 0;
877 }
878 basename_len = strlen(basename) + 1; /* count the '\0' */
879
880#ifdef SUPPORT_LINKS
881 linkname_len = S_ISLNK(st.st_mode) ? strlen(linkname) + 1 : 0;
882#else
883 linkname_len = 0;
884#endif
885
886 if (st.st_size > 0xFFFFFFFFu && S_ISREG(st.st_mode))
887 extra_len += EXTRA_LEN;
888
889 alloc_len = FILE_STRUCT_LEN + extra_len + basename_len + dirname_len
890 + linkname_len;
891 if (flist)
892 bp = pool_alloc(flist->file_pool, alloc_len, "make_file");
893 else {
894 if (!(bp = new_array(char, alloc_len)))
895 out_of_memory("make_file");
896 }
897
898 memset(bp, 0, FILE_STRUCT_LEN + extra_len);
899 bp += extra_len;
900 file = (struct file_struct *)bp;
901 bp += FILE_STRUCT_LEN;
902
903 memcpy(bp, basename, basename_len);
904 bp += basename_len + linkname_len; /* skip space for symlink too */
905
906#ifdef SUPPORT_HARD_LINKS
907 if (preserve_hard_links && flist) {
908 if (protocol_version >= 28
909 ? (!S_ISDIR(st.st_mode) && st.st_nlink > 1)
910 : S_ISREG(st.st_mode)) {
911 tmp_idev.dev = st.st_dev;
912 tmp_idev.ino = st.st_ino;
913 } else
914 tmp_idev.dev = tmp_idev.ino = 0;
915 }
916#endif
917
918#ifdef HAVE_STRUCT_STAT_ST_RDEV
919 if (IS_DEVICE(st.st_mode) || IS_SPECIAL(st.st_mode)) {
920 tmp_rdev = st.st_rdev;
921 st.st_size = 0;
922 }
923#endif
924
925 file->flags = flags;
926 file->modtime = st.st_mtime;
927 file->len32 = (uint32)st.st_size;
928 if (st.st_size > 0xFFFFFFFFu && S_ISREG(st.st_mode)) {
929 file->flags |= FLAG_LENGTH64;
930 OPT_EXTRA(file, 0)->unum = (uint32)(st.st_size >> 32);
931 }
932 file->mode = st.st_mode;
933 if (preserve_uid)
934 F_UID(file) = st.st_uid;
935 if (preserve_gid)
936 F_GID(file) = st.st_gid;
937
938 if (dirname_len) {
939 file->dirname = lastdir = bp;
940 lastdir_len = dirname_len - 1;
941 memcpy(bp, dirname, dirname_len - 1);
942 bp += dirname_len;
943 bp[-1] = '\0';
944 } else if (dirname)
945 file->dirname = dirname;
946
947#ifdef SUPPORT_LINKS
948 if (linkname_len) {
949 bp = (char*)F_BASENAME(file) + basename_len;
950 memcpy(bp, linkname, linkname_len);
951 }
952#endif
953
954 if (always_checksum && am_sender && S_ISREG(st.st_mode))
955 file_checksum(thisname, tmp_sum, st.st_size);
956
957 file->dir.root = flist_dir;
958
959 /* This code is only used by the receiver when it is building
960 * a list of files for a delete pass. */
961 if (keep_dirlinks && linkname_len && flist) {
962 STRUCT_STAT st2;
963 int save_mode = file->mode;
964 file->mode = S_IFDIR; /* Find a directory with our name. */
965 if (flist_find(the_file_list, file) >= 0
966 && do_stat(thisname, &st2) == 0 && S_ISDIR(st2.st_mode)) {
967 file->modtime = st2.st_mtime;
968 file->len32 = 0;
969 file->mode = st2.st_mode;
970 if (preserve_uid)
971 F_UID(file) = st2.st_uid;
972 if (preserve_gid)
973 F_GID(file) = st2.st_gid;
974 } else
975 file->mode = save_mode;
976 }
977
978 if (S_ISREG(st.st_mode) || S_ISLNK(st.st_mode))
979 stats.total_size += st.st_size;
980
981 if (basename_len == 0+1)
982 return NULL;
983
984 return file;
985}
986
987/* Only called for temporary file_struct entries created by make_file(). */
988void unmake_file(struct file_struct *file)
989{
990 int extra_cnt = flist_extra_cnt - 1 + LEN64_BUMP(file);
991 free(file->extras - extra_cnt);
992}
993
994static struct file_struct *send_file_name(int f, struct file_list *flist,
995 char *fname, STRUCT_STAT *stp,
996 unsigned short flags)
997{
998 struct file_struct *file;
999
1000 file = make_file(fname, flist, stp, flags,
1001 f == -2 ? SERVER_FILTERS : ALL_FILTERS);
1002 if (!file)
1003 return NULL;
1004
1005 if (chmod_modes && !S_ISLNK(file->mode))
1006 file->mode = tweak_mode(file->mode, chmod_modes);
1007
1008 maybe_emit_filelist_progress(flist->count + flist_count_offset);
1009
1010 flist_expand(flist);
1011 flist->files[flist->count++] = file;
1012 send_file_entry(file, f);
1013 return file;
1014}
1015
1016static void send_if_directory(int f, struct file_list *flist,
1017 struct file_struct *file,
1018 char *fbuf, unsigned int ol)
1019{
1020 char is_dot_dir = fbuf[ol-1] == '.' && (ol == 1 || fbuf[ol-2] == '/');
1021
1022 if (S_ISDIR(file->mode)
1023 && !(file->flags & FLAG_MOUNT_DIR) && f_name(file, fbuf)) {
1024 void *save_filters;
1025 unsigned int len = strlen(fbuf);
1026 if (len > 1 && fbuf[len-1] == '/')
1027 fbuf[--len] = '\0';
1028 if (len >= MAXPATHLEN - 1) {
1029 io_error |= IOERR_GENERAL;
1030 rprintf(FERROR, "skipping long-named directory: %s\n",
1031 full_fname(fbuf));
1032 return;
1033 }
1034 save_filters = push_local_filters(fbuf, len);
1035 send_directory(f, flist, fbuf, len);
1036 pop_local_filters(save_filters);
1037 fbuf[ol] = '\0';
1038 if (is_dot_dir)
1039 fbuf[ol-1] = '.';
1040 }
1041}
1042
1043/* This function is normally called by the sender, but the receiving side also
1044 * calls it from get_dirlist() with f set to -1 so that we just construct the
1045 * file list in memory without sending it over the wire. Also, get_dirlist()
1046 * might call this with f set to -2, which also indicates that local filter
1047 * rules should be ignored. */
1048static void send_directory(int f, struct file_list *flist, char *fbuf, int len)
1049{
1050 struct dirent *di;
1051 unsigned remainder;
1052 char *p;
1053 DIR *d;
1054 int start = flist->count;
1055
1056 if (!(d = opendir(fbuf))) {
1057 io_error |= IOERR_GENERAL;
1058 rsyserr(FERROR, errno, "opendir %s failed", full_fname(fbuf));
1059 return;
1060 }
1061
1062 p = fbuf + len;
1063 if (len != 1 || *fbuf != '/')
1064 *p++ = '/';
1065 *p = '\0';
1066 remainder = MAXPATHLEN - (p - fbuf);
1067
1068 for (errno = 0, di = readdir(d); di; errno = 0, di = readdir(d)) {
1069 char *dname = d_name(di);
1070 if (dname[0] == '.' && (dname[1] == '\0'
1071 || (dname[1] == '.' && dname[2] == '\0')))
1072 continue;
1073 if (strlcpy(p, dname, remainder) >= remainder) {
1074 io_error |= IOERR_GENERAL;
1075 rprintf(FINFO,
1076 "cannot send long-named file %s\n",
1077 full_fname(fbuf));
1078 continue;
1079 }
1080
1081 send_file_name(f, flist, fbuf, NULL, 0);
1082 }
1083
1084 fbuf[len] = '\0';
1085
1086 if (errno) {
1087 io_error |= IOERR_GENERAL;
1088 rsyserr(FERROR, errno, "readdir(%s)", full_fname(fbuf));
1089 }
1090
1091 closedir(d);
1092
1093 if (recurse) {
1094 int i, end = flist->count - 1;
1095 for (i = start; i <= end; i++)
1096 send_if_directory(f, flist, flist->files[i], fbuf, len);
1097 }
1098}
1099
1100struct file_list *send_file_list(int f, int argc, char *argv[])
1101{
1102 int len;
1103 STRUCT_STAT st;
1104 char *p, *dir, olddir[sizeof curr_dir];
1105 char lastpath[MAXPATHLEN] = "";
1106 struct file_list *flist;
1107 struct timeval start_tv, end_tv;
1108 int64 start_write;
1109 int use_ff_fd = 0;
1110
1111 rprintf(FLOG, "building file list\n");
1112 if (show_filelist_p())
1113 start_filelist_progress("building file list");
1114
1115 start_write = stats.total_written;
1116 gettimeofday(&start_tv, NULL);
1117
1118 flist = flist_new("send_file_list");
1119
1120 io_start_buffering_out();
1121 if (filesfrom_fd >= 0) {
1122 if (argv[0] && !push_dir(argv[0], 0)) {
1123 rsyserr(FERROR, errno, "push_dir %s failed",
1124 full_fname(argv[0]));
1125 exit_cleanup(RERR_FILESELECT);
1126 }
1127 use_ff_fd = 1;
1128 }
1129
1130 while (1) {
1131 char fbuf[MAXPATHLEN];
1132 char *fn;
1133 int is_dot_dir;
1134
1135 if (use_ff_fd) {
1136 if (read_filesfrom_line(filesfrom_fd, fbuf) == 0)
1137 break;
1138 sanitize_path(fbuf, fbuf, "", 0, NULL);
1139 } else {
1140 if (argc-- == 0)
1141 break;
1142 strlcpy(fbuf, *argv++, MAXPATHLEN);
1143 if (sanitize_paths)
1144 sanitize_path(fbuf, fbuf, "", 0, NULL);
1145 }
1146
1147 len = strlen(fbuf);
1148 if (relative_paths) {
1149 /* We clean up fbuf below. */
1150 is_dot_dir = 0;
1151 } else if (!len || fbuf[len - 1] == '/') {
1152 if (len == 2 && fbuf[0] == '.') {
1153 /* Turn "./" into just "." rather than "./." */
1154 fbuf[1] = '\0';
1155 } else {
1156 if (len + 1 >= MAXPATHLEN)
1157 overflow_exit("send_file_list");
1158 fbuf[len++] = '.';
1159 fbuf[len] = '\0';
1160 }
1161 is_dot_dir = 1;
1162 } else if (len > 1 && fbuf[len-1] == '.' && fbuf[len-2] == '.'
1163 && (len == 2 || fbuf[len-3] == '/')) {
1164 if (len + 2 >= MAXPATHLEN)
1165 overflow_exit("send_file_list");
1166 fbuf[len++] = '/';
1167 fbuf[len++] = '.';
1168 fbuf[len] = '\0';
1169 is_dot_dir = 1;
1170 } else {
1171 is_dot_dir = fbuf[len-1] == '.'
1172 && (len == 1 || fbuf[len-2] == '/');
1173 }
1174
1175 if (link_stat(fbuf, &st, copy_dirlinks) != 0) {
1176 io_error |= IOERR_GENERAL;
1177 rsyserr(FERROR, errno, "link_stat %s failed",
1178 full_fname(fbuf));
1179 continue;
1180 }
1181
1182 if (S_ISDIR(st.st_mode) && !xfer_dirs) {
1183 rprintf(FINFO, "skipping directory %s\n", fbuf);
1184 continue;
1185 }
1186
1187 dir = NULL;
1188 olddir[0] = '\0';
1189
1190 if (!relative_paths) {
1191 p = strrchr(fbuf, '/');
1192 if (p) {
1193 *p = '\0';
1194 if (p == fbuf)
1195 dir = "/";
1196 else
1197 dir = fbuf;
1198 len -= p - fbuf + 1;
1199 fn = p + 1;
1200 } else
1201 fn = fbuf;
1202 } else {
1203 if ((p = strstr(fbuf, "/./")) != NULL) {
1204 *p = '\0';
1205 if (p == fbuf)
1206 dir = "/";
1207 else
1208 dir = fbuf;
1209 len -= p - fbuf + 3;
1210 fn = p + 3;
1211 } else
1212 fn = fbuf;
1213 /* Get rid of trailing "/" and "/.". */
1214 while (len) {
1215 if (fn[len - 1] == '/') {
1216 is_dot_dir = 1;
1217 if (!--len && !dir) {
1218 len++;
1219 break;
1220 }
1221 }
1222 else if (len >= 2 && fn[len - 1] == '.'
1223 && fn[len - 2] == '/') {
1224 is_dot_dir = 1;
1225 if (!(len -= 2) && !dir) {
1226 len++;
1227 break;
1228 }
1229 } else
1230 break;
1231 }
1232 if (len == 1 && fn[0] == '/')
1233 fn[len++] = '.';
1234 fn[len] = '\0';
1235 /* Reject a ".." dir in the active part of the path. */
1236 for (p = fn; (p = strstr(p, "..")) != NULL; p += 2) {
1237 if ((p[2] == '/' || p[2] == '\0')
1238 && (p == fn || p[-1] == '/')) {
1239 rprintf(FERROR,
1240 "found \"..\" dir in relative path: %s\n",
1241 fbuf);
1242 exit_cleanup(RERR_SYNTAX);
1243 }
1244 }
1245 }
1246
1247 if (!*fn) {
1248 len = 1;
1249 fn = ".";
1250 }
1251
1252 if (dir && *dir) {
1253 static const char *lastdir;
1254 static int lastdir_len;
1255
1256 strlcpy(olddir, curr_dir, sizeof olddir);
1257
1258 if (!push_dir(dir, 0)) {
1259 io_error |= IOERR_GENERAL;
1260 rsyserr(FERROR, errno, "push_dir %s failed",
1261 full_fname(dir));
1262 continue;
1263 }
1264
1265 if (lastdir && strcmp(lastdir, dir) == 0) {
1266 flist_dir = lastdir;
1267 flist_dir_len = lastdir_len;
1268 } else {
1269 flist_dir = lastdir = strdup(dir);
1270 flist_dir_len = lastdir_len = strlen(dir);
1271 }
1272 }
1273
1274 if (fn != fbuf)
1275 memmove(fbuf, fn, len + 1);
1276
1277 if (implied_dirs && (p=strrchr(fbuf,'/')) && p != fbuf) {
1278 /* Send the implied directories at the start of the
1279 * source spec, so we get their permissions right. */
1280 char *lp = lastpath, *slash = fbuf;
1281 *p = '\0';
1282 /* Skip any initial directories in our path that we
1283 * have in common with lastpath. */
1284 for (fn = fbuf; *fn && *lp == *fn; lp++, fn++) {
1285 if (*fn == '/')
1286 slash = fn;
1287 }
1288 *p = '/';
1289 if (fn != p || (*lp && *lp != '/')) {
1290 int save_copy_links = copy_links;
1291 int save_xfer_dirs = xfer_dirs;
1292 copy_links |= copy_unsafe_links;
1293 xfer_dirs = 1;
1294 while ((slash = strchr(slash+1, '/')) != 0) {
1295 *slash = '\0';
1296 send_file_name(f, flist, fbuf, NULL, 0);
1297 *slash = '/';
1298 }
1299 copy_links = save_copy_links;
1300 xfer_dirs = save_xfer_dirs;
1301 *p = '\0';
1302 strlcpy(lastpath, fbuf, sizeof lastpath);
1303 *p = '/';
1304 }
1305 }
1306
1307 if (one_file_system)
1308 filesystem_dev = st.st_dev;
1309
1310 if (recurse || (xfer_dirs && is_dot_dir)) {
1311 struct file_struct *file;
1312 file = send_file_name(f, flist, fbuf, &st, FLAG_TOP_DIR);
1313 if (file)
1314 send_if_directory(f, flist, file, fbuf, len);
1315 } else
1316 send_file_name(f, flist, fbuf, &st, 0);
1317
1318 if (olddir[0]) {
1319 flist_dir = NULL;
1320 flist_dir_len = 0;
1321 if (!pop_dir(olddir)) {
1322 rsyserr(FERROR, errno, "pop_dir %s failed",
1323 full_fname(olddir));
1324 exit_cleanup(RERR_FILESELECT);
1325 }
1326 }
1327 }
1328
1329 gettimeofday(&end_tv, NULL);
1330 stats.flist_buildtime = (int64)(end_tv.tv_sec - start_tv.tv_sec) * 1000
1331 + (end_tv.tv_usec - start_tv.tv_usec) / 1000;
1332 if (stats.flist_buildtime == 0)
1333 stats.flist_buildtime = 1;
1334 start_tv = end_tv;
1335
1336 send_file_entry(NULL, f);
1337
1338 if (show_filelist_p())
1339 finish_filelist_progress(flist);
1340
1341 gettimeofday(&end_tv, NULL);
1342 stats.flist_xfertime = (int64)(end_tv.tv_sec - start_tv.tv_sec) * 1000
1343 + (end_tv.tv_usec - start_tv.tv_usec) / 1000;
1344
1345 /* Sort the list without removing any duplicates. This allows the
1346 * receiving side to ask for any name they like, which gives us the
1347 * flexibility to change the way we unduplicate names in the future
1348 * without causing a compatibility problem with older versions. */
1349 clean_flist(flist, 0, 0);
1350
1351 if (!numeric_ids)
1352 send_uid_list(f);
1353
1354 /* send the io_error flag */
1355 write_int(f, lp_ignore_errors(module_id) ? 0 : io_error);
1356
1357 io_end_buffering();
1358 stats.flist_size = stats.total_written - start_write;
1359 stats.num_files = flist->count;
1360
1361 if (verbose > 3)
1362 output_flist(flist);
1363
1364 if (verbose > 2)
1365 rprintf(FINFO, "send_file_list done\n");
1366
1367 return flist;
1368}
1369
1370struct file_list *recv_file_list(int f)
1371{
1372 struct file_list *flist;
1373 unsigned short flags;
1374 int64 start_read;
1375
1376 rprintf(FLOG, "receiving file list\n");
1377 if (show_filelist_p())
1378 start_filelist_progress("receiving file list");
1379
1380 start_read = stats.total_read;
1381
1382 flist = flist_new("recv_file_list");
1383
1384
1385 while ((flags = read_byte(f)) != 0) {
1386 struct file_struct *file;
1387
1388 flist_expand(flist);
1389
1390 if (protocol_version >= 28 && (flags & XMIT_EXTENDED_FLAGS))
1391 flags |= read_byte(f) << 8;
1392 file = recv_file_entry(flist, flags, f);
1393
1394 if (S_ISREG(file->mode) || S_ISLNK(file->mode))
1395 stats.total_size += F_LENGTH(file);
1396
1397 flist->files[flist->count++] = file;
1398
1399 maybe_emit_filelist_progress(flist->count);
1400
1401 if (verbose > 2) {
1402 rprintf(FINFO, "recv_file_name(%s)\n",
1403 f_name(file, NULL));
1404 }
1405 }
1406 recv_file_entry(NULL, 0, 0); /* Signal that we're done. */
1407
1408 if (verbose > 2)
1409 rprintf(FINFO, "received %d names\n", flist->count);
1410
1411 if (show_filelist_p())
1412 finish_filelist_progress(flist);
1413
1414 clean_flist(flist, relative_paths, 1);
1415
1416 if (f >= 0) {
1417 recv_uid_list(f, flist);
1418
1419 /* Recv the io_error flag */
1420 if (lp_ignore_errors(module_id) || ignore_errors)
1421 read_int(f);
1422 else
1423 io_error |= read_int(f);
1424 }
1425
1426 if (verbose > 3)
1427 output_flist(flist);
1428
1429 if (list_only) {
1430 int i;
1431 for (i = 0; i < flist->count; i++)
1432 list_file_entry(flist->files[i]);
1433 }
1434
1435 if (verbose > 2)
1436 rprintf(FINFO, "recv_file_list done\n");
1437
1438 stats.flist_size = stats.total_read - start_read;
1439 stats.num_files = flist->count;
1440
1441 return flist;
1442}
1443
1444static int file_compare(struct file_struct **file1, struct file_struct **file2)
1445{
1446 return f_name_cmp(*file1, *file2);
1447}
1448
1449/* Search for an identically-named item in the file list. Note that the
1450 * items must agree in their directory-ness, or no match is returned. */
1451int flist_find(struct file_list *flist, struct file_struct *f)
1452{
1453 int low = flist->low, high = flist->high;
1454 int diff, mid, mid_up;
1455
1456 while (low <= high) {
1457 mid = (low + high) / 2;
1458 if (F_IS_ACTIVE(flist->files[mid]))
1459 mid_up = mid;
1460 else {
1461 /* Scan for the next non-empty entry using the cached
1462 * distance values. If the value isn't fully up-to-
1463 * date, update it. */
1464 mid_up = mid + flist->files[mid]->dir.depth;
1465 if (!F_IS_ACTIVE(flist->files[mid_up])) {
1466 do {
1467 mid_up += flist->files[mid_up]->dir.depth;
1468 } while (!F_IS_ACTIVE(flist->files[mid_up]));
1469 flist->files[mid]->dir.depth = mid_up - mid;
1470 }
1471 if (mid_up > high) {
1472 /* If there's nothing left above us, set high to
1473 * a non-empty entry below us and continue. */
1474 high = mid - (int)flist->files[mid]->len32;
1475 if (!F_IS_ACTIVE(flist->files[high])) {
1476 do {
1477 high -= (int)flist->files[high]->len32;
1478 } while (!F_IS_ACTIVE(flist->files[high]));
1479 flist->files[mid]->len32 = mid - high;
1480 }
1481 continue;
1482 }
1483 }
1484 diff = f_name_cmp(flist->files[mid_up], f);
1485 if (diff == 0) {
1486 if (protocol_version < 29
1487 && S_ISDIR(flist->files[mid_up]->mode)
1488 != S_ISDIR(f->mode))
1489 return -1;
1490 return mid_up;
1491 }
1492 if (diff < 0)
1493 low = mid_up + 1;
1494 else
1495 high = mid - 1;
1496 }
1497 return -1;
1498}
1499
1500/*
1501 * Free up any resources a file_struct has allocated
1502 * and clear the file.
1503 */
1504void clear_file(struct file_struct *file)
1505{
1506 memset((char*)file + EXTRA_LEN, 0, FILE_STRUCT_LEN - EXTRA_LEN + 1);
1507 /* In an empty entry, dir.depth is an offset to the next non-empty
1508 * entry. Likewise for len32 in the opposite direction. We assume
1509 * that we're alone for now since flist_find() will adjust the counts
1510 * it runs into that aren't up-to-date. */
1511 file->len32 = file->dir.depth = 1;
1512}
1513
1514/* Allocate a new file list. */
1515struct file_list *flist_new(char *msg)
1516{
1517 struct file_list *flist;
1518
1519 flist = new(struct file_list);
1520 if (!flist)
1521 out_of_memory(msg);
1522
1523 memset(flist, 0, sizeof flist[0]);
1524
1525 if (!(flist->file_pool = pool_create(FILE_EXTENT, 0, out_of_memory, POOL_INTERN)))
1526 out_of_memory(msg);
1527
1528 return flist;
1529}
1530
1531/* Free up all elements in a flist. */
1532void flist_free(struct file_list *flist)
1533{
1534 pool_destroy(flist->file_pool);
1535 free(flist->files);
1536 free(flist);
1537}
1538
1539/*
1540 * This routine ensures we don't have any duplicate names in our file list.
1541 * duplicate names can cause corruption because of the pipelining
1542 */
1543static void clean_flist(struct file_list *flist, int strip_root, int no_dups)
1544{
1545 char fbuf[MAXPATHLEN];
1546 int i, prev_i = 0;
1547
1548 if (!flist)
1549 return;
1550 if (flist->count == 0) {
1551 flist->high = -1;
1552 return;
1553 }
1554
1555 qsort(flist->files, flist->count,
1556 sizeof flist->files[0], (int (*)())file_compare);
1557
1558 for (i = no_dups? 0 : flist->count; i < flist->count; i++) {
1559 if (F_IS_ACTIVE(flist->files[i])) {
1560 prev_i = i;
1561 break;
1562 }
1563 }
1564 flist->low = prev_i;
1565 while (++i < flist->count) {
1566 int j;
1567 struct file_struct *file = flist->files[i];
1568
1569 if (!F_IS_ACTIVE(file))
1570 continue;
1571 if (f_name_cmp(file, flist->files[prev_i]) == 0)
1572 j = prev_i;
1573 else if (protocol_version >= 29 && S_ISDIR(file->mode)) {
1574 int save_mode = file->mode;
1575 /* Make sure that this directory doesn't duplicate a
1576 * non-directory earlier in the list. */
1577 flist->high = prev_i;
1578 file->mode = S_IFREG;
1579 j = flist_find(flist, file);
1580 file->mode = save_mode;
1581 } else
1582 j = -1;
1583 if (j >= 0) {
1584 struct file_struct *fp = flist->files[j];
1585 int keep, drop;
1586 /* If one is a dir and the other is not, we want to
1587 * keep the dir because it might have contents in the
1588 * list. */
1589 if (S_ISDIR(file->mode) != S_ISDIR(fp->mode)) {
1590 if (S_ISDIR(file->mode))
1591 keep = i, drop = j;
1592 else
1593 keep = j, drop = i;
1594 } else
1595 keep = j, drop = i;
1596 if (verbose > 1 && !am_server) {
1597 rprintf(FINFO,
1598 "removing duplicate name %s from file list (%d)\n",
1599 f_name(file, fbuf), drop);
1600 }
1601 /* Make sure we don't lose track of a user-specified
1602 * top directory. */
1603 flist->files[keep]->flags |= flist->files[drop]->flags
1604 & (FLAG_TOP_DIR|FLAG_XFER_DIR);
1605
1606 clear_file(flist->files[drop]);
1607
1608 if (keep == i) {
1609 if (flist->low == drop) {
1610 for (j = drop + 1;
1611 j < i && !F_IS_ACTIVE(flist->files[j]);
1612 j++) {}
1613 flist->low = j;
1614 }
1615 prev_i = i;
1616 }
1617 } else
1618 prev_i = i;
1619 }
1620 flist->high = no_dups ? prev_i : flist->count - 1;
1621
1622 if (strip_root) {
1623 /* We need to strip off the leading slashes for relative
1624 * paths, but this must be done _after_ the sorting phase. */
1625 for (i = flist->low; i <= flist->high; i++) {
1626 struct file_struct *file = flist->files[i];
1627
1628 if (!file->dirname)
1629 continue;
1630 while (*file->dirname == '/')
1631 file->dirname++;
1632 if (!*file->dirname)
1633 file->dirname = NULL;
1634 }
1635 }
1636
1637 if (prune_empty_dirs && no_dups) {
1638 int j, prev_depth = 0;
1639
1640 prev_i = 0; /* It's OK that this isn't really true. */
1641
1642 for (i = flist->low; i <= flist->high; i++) {
1643 struct file_struct *fp, *file = flist->files[i];
1644
1645 /* This temporarily abuses the dir.depth value for a
1646 * directory that is in a chain that might get pruned.
1647 * We restore the old value if it gets a reprieve. */
1648 if (S_ISDIR(file->mode) && file->dir.depth) {
1649 /* Dump empty dirs when coming back down. */
1650 for (j = prev_depth; j >= file->dir.depth; j--) {
1651 fp = flist->files[prev_i];
1652 if (fp->dir.depth >= 0)
1653 break;
1654 prev_i = -fp->dir.depth-1;
1655 clear_file(fp);
1656 }
1657 prev_depth = file->dir.depth;
1658 if (is_excluded(f_name(file, fbuf), 1,
1659 ALL_FILTERS)) {
1660 /* Keep dirs through this dir. */
1661 for (j = prev_depth-1; ; j--) {
1662 fp = flist->files[prev_i];
1663 if (fp->dir.depth >= 0)
1664 break;
1665 prev_i = -fp->dir.depth-1;
1666 fp->dir.depth = j;
1667 }
1668 } else
1669 file->dir.depth = -prev_i-1;
1670 prev_i = i;
1671 } else {
1672 /* Keep dirs through this non-dir. */
1673 for (j = prev_depth; ; j--) {
1674 fp = flist->files[prev_i];
1675 if (fp->dir.depth >= 0)
1676 break;
1677 prev_i = -fp->dir.depth-1;
1678 fp->dir.depth = j;
1679 }
1680 }
1681 }
1682 /* Dump empty all remaining empty dirs. */
1683 while (1) {
1684 struct file_struct *fp = flist->files[prev_i];
1685 if (fp->dir.depth >= 0)
1686 break;
1687 prev_i = -fp->dir.depth-1;
1688 clear_file(fp);
1689 }
1690
1691 for (i = flist->low; i <= flist->high; i++) {
1692 if (F_IS_ACTIVE(flist->files[i]))
1693 break;
1694 }
1695 flist->low = i;
1696 for (i = flist->high; i >= flist->low; i--) {
1697 if (F_IS_ACTIVE(flist->files[i]))
1698 break;
1699 }
1700 flist->high = i;
1701 }
1702}
1703
1704static void output_flist(struct file_list *flist)
1705{
1706 char uidbuf[16], gidbuf[16], depthbuf[16];
1707 struct file_struct *file;
1708 const char *dir, *slash, *name, *trail;
1709 const char *who = who_am_i();
1710 int i;
1711
1712 for (i = 0; i < flist->count; i++) {
1713 file = flist->files[i];
1714 if ((am_root || am_sender) && preserve_uid) {
1715 snprintf(uidbuf, sizeof uidbuf, " uid=%ld",
1716 (long)F_UID(file));
1717 } else
1718 *uidbuf = '\0';
1719 if (preserve_gid && F_GID(file) != GID_NONE) {
1720 snprintf(gidbuf, sizeof gidbuf, " gid=%ld",
1721 (long)F_GID(file));
1722 } else
1723 *gidbuf = '\0';
1724 if (!am_sender)
1725 snprintf(depthbuf, sizeof depthbuf, "%d", file->dir.depth);
1726 if (F_IS_ACTIVE(file)) {
1727 if ((dir = file->dirname) == NULL)
1728 dir = slash = "";
1729 else
1730 slash = "/";
1731 name = F_BASENAME(file);
1732 trail = S_ISDIR(file->mode) ? "/" : "";
1733 } else
1734 dir = slash = name = trail = "";
1735 rprintf(FINFO, "[%s] i=%d %s %s%s%s%s mode=0%o len=%.0f%s%s flags=%x\n",
1736 who, i, am_sender ? NS(file->dir.root) : depthbuf,
1737 dir, slash, name, trail, (int)file->mode,
1738 (double)F_LENGTH(file), uidbuf, gidbuf, file->flags);
1739 }
1740}
1741
1742enum fnc_state { s_DIR, s_SLASH, s_BASE, s_TRAILING };
1743enum fnc_type { t_PATH, t_ITEM };
1744
1745/* Compare the names of two file_struct entities, similar to how strcmp()
1746 * would do if it were operating on the joined strings.
1747 *
1748 * Some differences beginning with protocol_version 29: (1) directory names
1749 * are compared with an assumed trailing slash so that they compare in a
1750 * way that would cause them to sort immediately prior to any content they
1751 * may have; (2) a directory of any name compares after a non-directory of
1752 * any name at the same depth; (3) a directory with name "." compares prior
1753 * to anything else. These changes mean that a directory and a non-dir
1754 * with the same name will not compare as equal (protocol_version >= 29).
1755 *
1756 * The dirname component can be an empty string, but the basename component
1757 * cannot (and never is in the current codebase). The basename component
1758 * may be NULL (for a removed item), in which case it is considered to be
1759 * after any existing item. */
1760int f_name_cmp(struct file_struct *f1, struct file_struct *f2)
1761{
1762 int dif;
1763 const uchar *c1, *c2;
1764 enum fnc_state state1, state2;
1765 enum fnc_type type1, type2;
1766 enum fnc_type t_path = protocol_version >= 29 ? t_PATH : t_ITEM;
1767
1768 if (!f1 || !F_IS_ACTIVE(f1)) {
1769 if (!f2 || !F_IS_ACTIVE(f2))
1770 return 0;
1771 return -1;
1772 }
1773 if (!f2 || !F_IS_ACTIVE(f2))
1774 return 1;
1775
1776 c1 = (uchar*)f1->dirname;
1777 c2 = (uchar*)f2->dirname;
1778 if (c1 == c2)
1779 c1 = c2 = NULL;
1780 if (!c1) {
1781 type1 = S_ISDIR(f1->mode) ? t_path : t_ITEM;
1782 c1 = (uchar*)F_BASENAME(f1);
1783 if (type1 == t_PATH && *c1 == '.' && !c1[1]) {
1784 type1 = t_ITEM;
1785 state1 = s_TRAILING;
1786 c1 = (uchar*)"";
1787 } else
1788 state1 = s_BASE;
1789 } else {
1790 type1 = t_path;
1791 state1 = s_DIR;
1792 }
1793 if (!c2) {
1794 type2 = S_ISDIR(f2->mode) ? t_path : t_ITEM;
1795 c2 = (uchar*)F_BASENAME(f2);
1796 if (type2 == t_PATH && *c2 == '.' && !c2[1]) {
1797 type2 = t_ITEM;
1798 state2 = s_TRAILING;
1799 c2 = (uchar*)"";
1800 } else
1801 state2 = s_BASE;
1802 } else {
1803 type2 = t_path;
1804 state2 = s_DIR;
1805 }
1806
1807 if (type1 != type2)
1808 return type1 == t_PATH ? 1 : -1;
1809
1810 do {
1811 if (!*c1) {
1812 switch (state1) {
1813 case s_DIR:
1814 state1 = s_SLASH;
1815 c1 = (uchar*)"/";
1816 break;
1817 case s_SLASH:
1818 type1 = S_ISDIR(f1->mode) ? t_path : t_ITEM;
1819 c1 = (uchar*)F_BASENAME(f1);
1820 if (type1 == t_PATH && *c1 == '.' && !c1[1]) {
1821 type1 = t_ITEM;
1822 state1 = s_TRAILING;
1823 c1 = (uchar*)"";
1824 } else
1825 state1 = s_BASE;
1826 break;
1827 case s_BASE:
1828 state1 = s_TRAILING;
1829 if (type1 == t_PATH) {
1830 c1 = (uchar*)"/";
1831 break;
1832 }
1833 /* FALL THROUGH */
1834 case s_TRAILING:
1835 type1 = t_ITEM;
1836 break;
1837 }
1838 if (*c2 && type1 != type2)
1839 return type1 == t_PATH ? 1 : -1;
1840 }
1841 if (!*c2) {
1842 switch (state2) {
1843 case s_DIR:
1844 state2 = s_SLASH;
1845 c2 = (uchar*)"/";
1846 break;
1847 case s_SLASH:
1848 type2 = S_ISDIR(f2->mode) ? t_path : t_ITEM;
1849 c2 = (uchar*)F_BASENAME(f2);
1850 if (type2 == t_PATH && *c2 == '.' && !c2[1]) {
1851 type2 = t_ITEM;
1852 state2 = s_TRAILING;
1853 c2 = (uchar*)"";
1854 } else
1855 state2 = s_BASE;
1856 break;
1857 case s_BASE:
1858 state2 = s_TRAILING;
1859 if (type2 == t_PATH) {
1860 c2 = (uchar*)"/";
1861 break;
1862 }
1863 /* FALL THROUGH */
1864 case s_TRAILING:
1865 if (!*c1)
1866 return 0;
1867 type2 = t_ITEM;
1868 break;
1869 }
1870 if (type1 != type2)
1871 return type1 == t_PATH ? 1 : -1;
1872 }
1873 } while ((dif = (int)*c1++ - (int)*c2++) == 0);
1874
1875 return dif;
1876}
1877
1878char *f_name_buf(void)
1879{
1880 static char names[5][MAXPATHLEN];
1881 static unsigned int n;
1882
1883 n = (n + 1) % (sizeof names / sizeof names[0]);
1884
1885 return names[n];
1886}
1887
1888/* Return a copy of the full filename of a flist entry, using the indicated
1889 * buffer or one of 5 static buffers if fbuf is NULL. No size-checking is
1890 * done because we checked the size when creating the file_struct entry.
1891 */
1892char *f_name(struct file_struct *f, char *fbuf)
1893{
1894 if (!f || !F_IS_ACTIVE(f))
1895 return NULL;
1896
1897 if (!fbuf)
1898 fbuf = f_name_buf();
1899
1900 if (f->dirname) {
1901 int len = strlen(f->dirname);
1902 memcpy(fbuf, f->dirname, len);
1903 fbuf[len] = '/';
1904 strlcpy(fbuf + len + 1, F_BASENAME(f), MAXPATHLEN - (len + 1));
1905 } else
1906 strlcpy(fbuf, F_BASENAME(f), MAXPATHLEN);
1907
1908 return fbuf;
1909}
1910
1911/* Do a non-recursive scan of the named directory, possibly ignoring all
1912 * exclude rules except for the daemon's. If "dlen" is >=0, it is the length
1913 * of the dirname string, and also indicates that "dirname" is a MAXPATHLEN
1914 * buffer (the functions we call will append names onto the end, but the old
1915 * dir value will be restored on exit). */
1916struct file_list *get_dirlist(char *dirname, int dlen, int ignore_filter_rules)
1917{
1918 struct file_list *dirlist;
1919 char dirbuf[MAXPATHLEN];
1920 int save_recurse = recurse;
1921 int save_xfer_dirs = xfer_dirs;
1922
1923 if (dlen < 0) {
1924 dlen = strlcpy(dirbuf, dirname, MAXPATHLEN);
1925 if (dlen >= MAXPATHLEN)
1926 return NULL;
1927 dirname = dirbuf;
1928 }
1929
1930 dirlist = flist_new("get_dirlist");
1931
1932 recurse = 0;
1933 xfer_dirs = 1;
1934 send_directory(ignore_filter_rules ? -2 : -1, dirlist, dirname, dlen);
1935 xfer_dirs = save_xfer_dirs;
1936 recurse = save_recurse;
1937 if (do_progress)
1938 flist_count_offset += dirlist->count;
1939
1940 clean_flist(dirlist, 0, 0);
1941
1942 if (verbose > 3)
1943 output_flist(dirlist);
1944
1945 return dirlist;
1946}