2 Copyright (C) 2000-2004 Kern Sibbald and John Walker
4 This program is free software; you can redistribute it and/or
5 modify it under the terms of the GNU General Public License as
6 published by the Free Software Foundation; either version 2 of
7 the License, or (at your option) any later version.
9 This program is distributed in the hope that it will be useful,
10 but WITHOUT ANY WARRANTY; without even the implied warranty of
11 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
12 General Public License for more details.
14 You should have received a copy of the GNU General Public
15 License along with this program; if not, write to the Free
16 Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
19 This file is based on GNU TAR source code. Except for a few key
20 ideas, it has been rewritten for Bacula.
24 Thanks to the TAR programmers.
36 extern int32_t name_max; /* filename max length */
37 extern int32_t path_max; /* path name max length */
40 * Structure for keeping track of hard linked files, we
41 * keep an entry for each hardlinked file that we save,
42 * which is the first one found. For all the other files that
43 * are linked to this one, we save only the directory
44 * entry so we can link it.
48 dev_t dev; /* device */
49 ino_t ino; /* inode with device is unique */
51 uint32_t FileIndex; /* Bacula FileIndex of this file */
52 char name[1]; /* The name */
55 static void free_dir_ff_pkt(FF_PKT *dir_ff_pkt)
57 free(dir_ff_pkt->fname);
58 free(dir_ff_pkt->link);
59 free_pool_memory(dir_ff_pkt->sys_fname);
64 * Check to see if we allow the file system type of a file or directory.
65 * If we do not have a list of file system types, we accept anything.
67 static int accept_fstype(FF_PKT *ff, void *dummy) {
72 if (ff->fstypes.size()) {
74 fs = fstype(ff->fname);
76 Dmsg1(50, "Cannot determine file system type for \"%s\"\n", ff->fname);
78 for (i = 0; i < ff->fstypes.size(); ++i) {
79 if (strcmp(fs, (char *)ff->fstypes.get(i)) == 0) {
80 Dmsg2(100, "Accepting fstype %s for \"%s\"\n", fs, ff->fname);
84 Dmsg3(200, "fstype %s for \"%s\" does not match %s\n", fs,
85 ff->fname, ff->fstypes.get(i));
95 * handle_file is the callback for handling the file.
97 * parent_device is the device we are currently on
98 * top_level is 1 when not recursing or 0 when
99 * descending into a directory.
102 find_one_file(JCR *jcr, FF_PKT *ff_pkt, int handle_file(FF_PKT *ff, void *hpkt),
103 void *pkt, char *fname, dev_t parent_device, int top_level)
105 struct utimbuf restore_times;
108 ff_pkt->fname = ff_pkt->link = fname;
110 if (lstat(fname, &ff_pkt->statp) != 0) {
111 /* Cannot stat file */
112 ff_pkt->type = FT_NOSTAT;
113 ff_pkt->ff_errno = errno;
114 return handle_file(ff_pkt, pkt);
117 Dmsg1(300, "File ----: %s\n", fname);
119 /* Save current times of this directory in case we need to
120 * reset them because the user doesn't want them changed.
122 restore_times.actime = ff_pkt->statp.st_atime;
123 restore_times.modtime = ff_pkt->statp.st_mtime;
126 * We check for allowed fstypes at top_level and fstype change (below).
128 if (top_level && !accept_fstype(ff_pkt, NULL)) {
129 ff_pkt->type = FT_INVALIDFS;
130 if (ff_pkt->flags & FO_KEEPATIME) {
131 utime(fname, &restore_times);
133 Jmsg1(jcr, M_ERROR, 0, _("Top level directory \"%s\" has an unlisted fstype\n"), fname);
134 return 1; /* Just ignore this error - or the whole backup is cancelled */
138 * If this is an Incremental backup, see if file was modified
139 * since our last "save_time", presumably the last Full save
142 if (ff_pkt->incremental && !S_ISDIR(ff_pkt->statp.st_mode)) {
143 Dmsg1(300, "Non-directory incremental: %s\n", ff_pkt->fname);
144 /* Not a directory */
145 if (ff_pkt->statp.st_mtime < ff_pkt->save_time
146 && ((ff_pkt->flags & FO_MTIMEONLY) ||
147 ff_pkt->statp.st_ctime < ff_pkt->save_time)) {
148 /* Incremental option, file not changed */
149 ff_pkt->type = FT_NOCHG;
150 return handle_file(ff_pkt, pkt);
154 #ifdef HAVE_DARWIN_OS
155 if (S_ISREG(ff_pkt->statp.st_mode) && ff_pkt->flags & FO_HFSPLUS) {
156 /* TODO: initialise attrList once elsewhere? */
157 struct attrlist attrList;
158 memset(&attrList, 0, sizeof(attrList));
159 attrList.bitmapcount = ATTR_BIT_MAP_COUNT;
160 attrList.commonattr = ATTR_CMN_FNDRINFO;
161 attrList.fileattr = ATTR_FILE_RSRCLENGTH;
162 if (getattrlist(fname, &attrList, &ff_pkt->hfsinfo,
163 sizeof(ff_pkt->hfsinfo), 0) != 0) {
164 ff_pkt->type = FT_NOSTAT;
165 ff_pkt->ff_errno = errno;
166 return handle_file(ff_pkt, pkt);
171 /* ***FIXME*** implement this */
173 /* See if we are trying to dump the archive. */
174 if (ar_dev && ff_pkt->statp.st_dev == ar_dev && ff_pkt->statp.st_ino == ar_ino) {
175 ff_pkt->type = FT_ISARCH;
176 return handle_file(ff_pkt, pkt);
181 * Handle hard linked files
183 * Maintain a list of hard linked files already backed up. This
184 * allows us to ensure that the data of each file gets backed
187 if (!(ff_pkt->flags & FO_NO_HARDLINK)
188 && ff_pkt->statp.st_nlink > 1
189 && (S_ISREG(ff_pkt->statp.st_mode)
190 || S_ISCHR(ff_pkt->statp.st_mode)
191 || S_ISBLK(ff_pkt->statp.st_mode)
192 || S_ISFIFO(ff_pkt->statp.st_mode)
193 || S_ISSOCK(ff_pkt->statp.st_mode))) {
197 /* Search link list of hard linked files */
198 for (lp = ff_pkt->linklist; lp; lp = lp->next)
199 if (lp->ino == (ino_t)ff_pkt->statp.st_ino &&
200 lp->dev == (dev_t)ff_pkt->statp.st_dev) {
201 /* If we have already backed up the hard linked file don't do it again */
202 if (strcmp(lp->name, fname) == 0) {
203 Jmsg1(jcr, M_WARNING, 0, _("Attempt to backup hard linked file %s twice ignored.\n"),
205 return 1; /* ignore */
207 ff_pkt->link = lp->name;
208 ff_pkt->type = FT_LNKSAVED; /* Handle link, file already saved */
209 ff_pkt->LinkFI = lp->FileIndex;
210 return handle_file(ff_pkt, pkt);
213 /* File not previously dumped. Chain it into our list. */
214 lp = (struct f_link *)bmalloc(sizeof(struct f_link) + strlen(fname) +1);
215 lp->ino = ff_pkt->statp.st_ino;
216 lp->dev = ff_pkt->statp.st_dev;
217 strcpy(lp->name, fname);
218 lp->next = ff_pkt->linklist;
219 ff_pkt->linklist = lp;
220 ff_pkt->linked = lp; /* mark saved link */
222 ff_pkt->linked = NULL;
225 /* This is not a link to a previously dumped file, so dump it. */
226 if (S_ISREG(ff_pkt->statp.st_mode)) {
229 sizeleft = ff_pkt->statp.st_size;
231 /* Don't bother opening empty, world readable files. Also do not open
232 files when archive is meant for /dev/null. */
233 if (ff_pkt->null_output_device || (sizeleft == 0
234 && MODE_RALL == (MODE_RALL & ff_pkt->statp.st_mode))) {
235 ff_pkt->type = FT_REGE;
237 ff_pkt->type = FT_REG;
239 rtn_stat = handle_file(ff_pkt, pkt);
240 if (ff_pkt->linked) {
241 ff_pkt->linked->FileIndex = ff_pkt->FileIndex;
246 } else if (S_ISLNK(ff_pkt->statp.st_mode)) { /* soft link */
248 char *buffer = (char *)alloca(path_max + name_max + 102);
250 size = readlink(fname, buffer, path_max + name_max + 101);
252 /* Could not follow link */
253 ff_pkt->type = FT_NOFOLLOW;
254 ff_pkt->ff_errno = errno;
255 rtn_stat = handle_file(ff_pkt, pkt);
256 if (ff_pkt->linked) {
257 ff_pkt->linked->FileIndex = ff_pkt->FileIndex;
262 ff_pkt->link = buffer; /* point to link */
263 ff_pkt->type = FT_LNK; /* got a real link */
264 rtn_stat = handle_file(ff_pkt, pkt);
265 if (ff_pkt->linked) {
266 ff_pkt->linked->FileIndex = ff_pkt->FileIndex;
270 } else if (S_ISDIR(ff_pkt->statp.st_mode)) {
272 struct dirent *entry, *result;
277 dev_t our_device = ff_pkt->statp.st_dev;
281 * If we are using Win32 (non-portable) backup API, don't check
282 * access as everything is more complicated, and
283 * in principle, we should be able to access everything.
285 if (!have_win32_api() || (ff_pkt->flags & FO_PORTABLE)) {
286 if (access(fname, R_OK) == -1 && geteuid() != 0) {
287 /* Could not access() directory */
288 ff_pkt->type = FT_NOACCESS;
289 ff_pkt->ff_errno = errno;
290 rtn_stat = handle_file(ff_pkt, pkt);
291 if (ff_pkt->linked) {
292 ff_pkt->linked->FileIndex = ff_pkt->FileIndex;
298 /* Build a canonical directory name with a trailing slash in link var */
300 link_len = len + 200;
301 link = (char *)bmalloc(link_len + 2);
302 bstrncpy(link, fname, link_len);
303 /* Strip all trailing slashes */
304 while (len >= 1 && link[len - 1] == '/')
306 link[len++] = '/'; /* add back one */
310 if (ff_pkt->incremental &&
311 (ff_pkt->statp.st_mtime < ff_pkt->save_time &&
312 ff_pkt->statp.st_ctime < ff_pkt->save_time)) {
313 /* Incremental option, directory entry not changed */
314 ff_pkt->type = FT_DIRNOCHG;
316 ff_pkt->type = FT_DIRBEGIN;
319 * Note, we return the directory to the calling program (handle_file)
320 * when we first see the directory (FT_DIRBEGIN.
321 * This allows the program to apply matches and make a
322 * choice whether or not to accept it. If it is accepted, we
323 * do not immediately save it, but do so only after everything
324 * in the directory is seen (i.e. the FT_DIREND).
326 rtn_stat = handle_file(ff_pkt, pkt);
327 if (rtn_stat < 1) { /* ignore or error status */
331 /* Done with DIRBEGIN, next call will be DIREND */
332 if (ff_pkt->type == FT_DIRBEGIN) {
333 ff_pkt->type = FT_DIREND;
337 * Create a temporary ff packet for this directory
338 * entry, and defer handling the directory until
339 * we have recursed into it. This saves the
340 * directory after all files have been processed, and
341 * during the restore, the directory permissions will
342 * be reset after all the files have been restored.
344 Dmsg1(300, "Create temp ff packet for dir: %s\n", ff_pkt->fname);
345 FF_PKT *dir_ff_pkt = (FF_PKT *)bmalloc(sizeof(FF_PKT));
346 memcpy(dir_ff_pkt, ff_pkt, sizeof(FF_PKT));
347 dir_ff_pkt->fname = bstrdup(ff_pkt->fname);
348 dir_ff_pkt->link = bstrdup(ff_pkt->link);
349 dir_ff_pkt->sys_fname = get_pool_memory(PM_FNAME);
350 dir_ff_pkt->included_files_list = NULL;
351 dir_ff_pkt->excluded_files_list = NULL;
352 dir_ff_pkt->excluded_paths_list = NULL;
353 dir_ff_pkt->linklist = NULL;
356 * Do not descend into subdirectories (recurse) if the
357 * user has turned it off for this directory.
359 * If we are crossing file systems, we are either not allowed
360 * to cross, or we may be restricted by a list of permitted
363 if (ff_pkt->flags & FO_NO_RECURSION) {
364 ff_pkt->type = FT_NORECURSE;
366 } else if (!top_level && parent_device != ff_pkt->statp.st_dev) {
367 if(!(ff_pkt->flags & FO_MULTIFS)) {
368 ff_pkt->type = FT_NOFSCHG;
370 } else if (!accept_fstype(ff_pkt, NULL)) {
371 ff_pkt->type = FT_INVALIDFS;
376 rtn_stat = handle_file(ff_pkt, pkt);
377 if (ff_pkt->linked) {
378 ff_pkt->linked->FileIndex = ff_pkt->FileIndex;
381 free_dir_ff_pkt(dir_ff_pkt);
382 ff_pkt->link = ff_pkt->fname; /* reset "link" */
383 if (ff_pkt->flags & FO_KEEPATIME) {
384 utime(fname, &restore_times);
389 ff_pkt->link = ff_pkt->fname; /* reset "link" */
392 * Descend into or "recurse" into the directory to read
393 * all the files in it.
396 if ((directory = opendir(fname)) == NULL) {
397 ff_pkt->type = FT_NOOPEN;
398 ff_pkt->ff_errno = errno;
399 rtn_stat = handle_file(ff_pkt, pkt);
400 if (ff_pkt->linked) {
401 ff_pkt->linked->FileIndex = ff_pkt->FileIndex;
404 free_dir_ff_pkt(dir_ff_pkt);
409 * Process all files in this directory entry (recursing).
410 * This would possibly run faster if we chdir to the directory
411 * before traversing it.
414 entry = (struct dirent *)malloc(sizeof(struct dirent) + name_max + 100);
415 for ( ; !job_canceled(jcr); ) {
419 status = readdir_r(directory, entry, &result);
420 if (status != 0 || result == NULL) {
421 // Dmsg2(99, "readdir returned stat=%d result=0x%x\n",
422 // status, (long)result);
425 ASSERT(name_max+1 > (int)sizeof(struct dirent) + (int)NAMELEN(entry));
427 /* Skip `.', `..', and excluded file names. */
428 if (p[0] == '\0' || (p[0] == '.' && (p[1] == '\0' ||
429 (p[1] == '.' && p[2] == '\0')))) {
433 if ((int)NAMELEN(entry) + len >= link_len) {
434 link_len = len + NAMELEN(entry) + 1;
435 link = (char *)brealloc(link, link_len + 1);
438 for (i=0; i < (int)NAMELEN(entry); i++) {
442 if (!file_is_excluded(ff_pkt, link)) {
443 rtn_stat = find_one_file(jcr, ff_pkt, handle_file, pkt, link, our_device, 0);
444 if (ff_pkt->linked) {
445 ff_pkt->linked->FileIndex = ff_pkt->FileIndex;
454 * Now that we have recursed through all the files in the
455 * directory, we "save" the directory so that after all
456 * the files are restored, this entry will serve to reset
457 * the directory modes and dates. Temp directory values
458 * were used without this record.
460 handle_file(dir_ff_pkt, pkt); /* handle directory entry */
461 if (ff_pkt->linked) {
462 ff_pkt->linked->FileIndex = dir_ff_pkt->FileIndex;
464 free_dir_ff_pkt(dir_ff_pkt);
466 if (ff_pkt->flags & FO_KEEPATIME) {
467 utime(fname, &restore_times);
470 } /* end check for directory */
473 * If it is explicitly mentioned (i.e. top_level) and is
474 * a block device, we do a raw backup of it or if it is
475 * a fifo, we simply read it.
477 #ifdef HAVE_FREEBSD_OS
479 * On FreeBSD, all block devices are character devices, so
480 * to be able to read a raw disk, we need the check for
481 * a character device.
482 * crw-r----- 1 root operator - 116, 0x00040002 Jun 9 19:32 /dev/ad0s3
483 * crw-r----- 1 root operator - 116, 0x00040002 Jun 9 19:32 /dev/rad0s3
485 if (top_level && (S_ISBLK(ff_pkt->statp.st_mode) || S_ISCHR(ff_pkt->statp.st_mode))) {
487 if (top_level && S_ISBLK(ff_pkt->statp.st_mode)) {
489 ff_pkt->type = FT_RAW; /* raw partition */
490 } else if (top_level && S_ISFIFO(ff_pkt->statp.st_mode) &&
491 ff_pkt->flags & FO_READFIFO) {
492 ff_pkt->type = FT_FIFO;
494 /* The only remaining types are special (character, ...) files */
495 ff_pkt->type = FT_SPEC;
497 rtn_stat = handle_file(ff_pkt, pkt);
498 if (ff_pkt->linked) {
499 ff_pkt->linked->FileIndex = ff_pkt->FileIndex;
504 int term_find_one(FF_PKT *ff)
506 struct f_link *lp, *lc;
509 /* Free up list of hard linked files */
510 for (lp = ff->linklist; lp;) {