xref: /freebsd/bin/pax/ftree.c (revision 6780ab54325a71e7e70112b11657973edde8655e)
1 /*-
2  * Copyright (c) 1992 Keith Muller.
3  * Copyright (c) 1992, 1993
4  *	The Regents of the University of California.  All rights reserved.
5  *
6  * This code is derived from software contributed to Berkeley by
7  * Keith Muller of the University of California, San Diego.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 3. All advertising materials mentioning features or use of this software
18  *    must display the following acknowledgement:
19  *	This product includes software developed by the University of
20  *	California, Berkeley and its contributors.
21  * 4. Neither the name of the University nor the names of its contributors
22  *    may be used to endorse or promote products derived from this software
23  *    without specific prior written permission.
24  *
25  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
26  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
30  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
31  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
34  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35  * SUCH DAMAGE.
36  */
37 
38 #ifndef lint
39 #if 0
40 static char sccsid[] = "@(#)ftree.c	8.2 (Berkeley) 4/18/94";
41 #endif
42 #endif /* not lint */
43 #include <sys/cdefs.h>
44 __FBSDID("$FreeBSD$");
45 
46 #include <sys/types.h>
47 #include <sys/time.h>
48 #include <sys/stat.h>
49 #include <unistd.h>
50 #include <string.h>
51 #include <stdio.h>
52 #include <errno.h>
53 #include <stdlib.h>
54 #include <fts.h>
55 #include "pax.h"
56 #include "ftree.h"
57 #include "extern.h"
58 
59 /*
60  * routines to interface with the fts library function.
61  *
62  * file args supplied to pax are stored on a single linked list (of type FTREE)
63  * and given to fts to be processed one at a time. pax "selects" files from
64  * the expansion of each arg into the corresponding file tree (if the arg is a
65  * directory, otherwise the node itself is just passed to pax). The selection
66  * is modified by the -n and -u flags. The user is informed when a specific
67  * file arg does not generate any selected files. -n keeps expanding the file
68  * tree arg until one of its files is selected, then skips to the next file
69  * arg. when the user does not supply the file trees as command line args to
70  * pax, they are read from stdin
71  */
72 
73 static FTS *ftsp = NULL;		/* current FTS handle */
74 static int ftsopts;			/* options to be used on fts_open */
75 static char *farray[2];			/* array for passing each arg to fts */
76 static FTREE *fthead = NULL;		/* head of linked list of file args */
77 static FTREE *fttail = NULL;		/* tail of linked list of file args */
78 static FTREE *ftcur = NULL;		/* current file arg being processed */
79 static FTSENT *ftent = NULL;		/* current file tree entry */
80 static int ftree_skip;			/* when set skip to next file arg */
81 
82 static int ftree_arg(void);
83 
84 /*
85  * ftree_start()
86  *	initialize the options passed to fts_open() during this run of pax
87  *	options are based on the selection of pax options by the user
88  *	fts_start() also calls fts_arg() to open the first valid file arg. We
89  *	also attempt to reset directory access times when -t (tflag) is set.
90  * Return:
91  *	0 if there is at least one valid file arg to process, -1 otherwise
92  */
93 
94 int
95 ftree_start(void)
96 {
97 	/*
98 	 * Set up the operation mode of fts, open the first file arg. We must
99 	 * use FTS_NOCHDIR, as the user may have to open multiple archives and
100 	 * if fts did a chdir off into the boondocks, we may create an archive
101 	 * volume in a place where the user did not expect to.
102 	 */
103 	ftsopts = FTS_NOCHDIR;
104 
105 	/*
106 	 * optional user flags that effect file traversal
107 	 * -H command line symlink follow only (half follow)
108 	 * -L follow sylinks (logical)
109 	 * -P do not follow sylinks (physical). This is the default.
110 	 * -X do not cross over mount points
111 	 * -t preserve access times on files read.
112 	 * -n select only the first member of a file tree when a match is found
113 	 * -d do not extract subtrees rooted at a directory arg.
114 	 */
115 	if (Lflag)
116 		ftsopts |= FTS_LOGICAL;
117 	else
118 		ftsopts |= FTS_PHYSICAL;
119 	if (Hflag)
120 #	ifdef NET2_FTS
121 		paxwarn(0, "The -H flag is not supported on this version");
122 #	else
123 		ftsopts |= FTS_COMFOLLOW;
124 #	endif
125 	if (Xflag)
126 		ftsopts |= FTS_XDEV;
127 
128 	if ((fthead == NULL) && ((farray[0] = malloc(PAXPATHLEN+2)) == NULL)) {
129 		paxwarn(1, "Unable to allocate memory for file name buffer");
130 		return(-1);
131 	}
132 
133 	if (ftree_arg() < 0)
134 		return(-1);
135 	if (tflag && (atdir_start() < 0))
136 		return(-1);
137 	return(0);
138 }
139 
140 /*
141  * ftree_add()
142  *	add the arg to the linked list of files to process. Each will be
143  *	processed by fts one at a time
144  * Return:
145  *	0 if added to the linked list, -1 if failed
146  */
147 
148 int
149 ftree_add(char *str, int chflg)
150 {
151 	FTREE *ft;
152 	int len;
153 
154 	/*
155 	 * simple check for bad args
156 	 */
157 	if ((str == NULL) || (*str == '\0')) {
158 		paxwarn(0, "Invalid file name argument");
159 		return(-1);
160 	}
161 
162 	/*
163 	 * allocate FTREE node and add to the end of the linked list (args are
164 	 * processed in the same order they were passed to pax). Get rid of any
165 	 * trailing / the user may pass us. (watch out for / by itself).
166 	 */
167 	if ((ft = (FTREE *)malloc(sizeof(FTREE))) == NULL) {
168 		paxwarn(0, "Unable to allocate memory for filename");
169 		return(-1);
170 	}
171 
172 	if (((len = strlen(str) - 1) > 0) && (str[len] == '/'))
173 		str[len] = '\0';
174 	ft->fname = str;
175 	ft->refcnt = 0;
176 	ft->chflg = chflg;
177 	ft->fow = NULL;
178 	if (fthead == NULL) {
179 		fttail = fthead = ft;
180 		return(0);
181 	}
182 	fttail->fow = ft;
183 	fttail = ft;
184 	return(0);
185 }
186 
187 /*
188  * ftree_sel()
189  *	this entry has been selected by pax. bump up reference count and handle
190  *	-n and -d processing.
191  */
192 
193 void
194 ftree_sel(ARCHD *arcn)
195 {
196 	/*
197 	 * set reference bit for this pattern. This linked list is only used
198 	 * when file trees are supplied pax as args. The list is not used when
199 	 * the trees are read from stdin.
200 	 */
201 	if (ftcur != NULL)
202 		ftcur->refcnt = 1;
203 
204 	/*
205 	 * if -n we are done with this arg, force a skip to the next arg when
206 	 * pax asks for the next file in next_file().
207 	 * if -d we tell fts only to match the directory (if the arg is a dir)
208 	 * and not the entire file tree rooted at that point.
209 	 */
210 	if (nflag)
211 		ftree_skip = 1;
212 
213 	if (!dflag || (arcn->type != PAX_DIR))
214 		return;
215 
216 	if (ftent != NULL)
217 		(void)fts_set(ftsp, ftent, FTS_SKIP);
218 }
219 
220 /*
221  * ftree_chk()
222  *	called at end on pax execution. Prints all those file args that did not
223  *	have a selected member (reference count still 0)
224  */
225 
226 void
227 ftree_chk(void)
228 {
229 	FTREE *ft;
230 	int wban = 0;
231 
232 	/*
233 	 * make sure all dir access times were reset.
234 	 */
235 	if (tflag)
236 		atdir_end();
237 
238 	/*
239 	 * walk down list and check reference count. Print out those members
240 	 * that never had a match
241 	 */
242 	for (ft = fthead; ft != NULL; ft = ft->fow) {
243 		if ((ft->refcnt > 0) || ft->chflg)
244 			continue;
245 		if (wban == 0) {
246 			paxwarn(1,"WARNING! These file names were not selected:");
247 			++wban;
248 		}
249 		(void)fprintf(stderr, "%s\n", ft->fname);
250 	}
251 }
252 
253 /*
254  * ftree_arg()
255  *	Get the next file arg for fts to process. Can be from either the linked
256  *	list or read from stdin when the user did not them as args to pax. Each
257  *	arg is processed until the first successful fts_open().
258  * Return:
259  *	0 when the next arg is ready to go, -1 if out of file args (or EOF on
260  *	stdin).
261  */
262 
263 static int
264 ftree_arg(void)
265 {
266 	char *pt;
267 
268 	/*
269 	 * close off the current file tree
270 	 */
271 	if (ftsp != NULL) {
272 		(void)fts_close(ftsp);
273 		ftsp = NULL;
274 	}
275 
276 	/*
277 	 * keep looping until we get a valid file tree to process. Stop when we
278 	 * reach the end of the list (or get an eof on stdin)
279 	 */
280 	for(;;) {
281 		if (fthead == NULL) {
282 			/*
283 			 * the user didn't supply any args, get the file trees
284 			 * to process from stdin;
285 			 */
286 			if (fgets(farray[0], PAXPATHLEN+1, stdin) == NULL)
287 				return(-1);
288 			if ((pt = strchr(farray[0], '\n')) != NULL)
289 				*pt = '\0';
290 		} else {
291 			/*
292 			 * the user supplied the file args as arguments to pax
293 			 */
294 			if (ftcur == NULL)
295 				ftcur = fthead;
296 			else if ((ftcur = ftcur->fow) == NULL)
297 				return(-1);
298 			if (ftcur->chflg) {
299 				/* First fchdir() back... */
300 				if (fchdir(cwdfd) < 0) {
301 					syswarn(1, errno,
302 					  "Can't fchdir to starting directory");
303 					return(-1);
304 				}
305 				if (chdir(ftcur->fname) < 0) {
306 					syswarn(1, errno, "Can't chdir to %s",
307 					    ftcur->fname);
308 					return(-1);
309 				}
310 				continue;
311 			} else
312 				farray[0] = ftcur->fname;
313 		}
314 
315 		/*
316 		 * Watch it, fts wants the file arg stored in an array of char
317 		 * ptrs, with the last one a null. We use a two element array
318 		 * and set farray[0] to point at the buffer with the file name
319 		 * in it. We cannot pass all the file args to fts at one shot
320 		 * as we need to keep a handle on which file arg generates what
321 		 * files (the -n and -d flags need this). If the open is
322 		 * successful, return a 0.
323 		 */
324 		if ((ftsp = fts_open(farray, ftsopts, NULL)) != NULL)
325 			break;
326 	}
327 	return(0);
328 }
329 
330 /*
331  * next_file()
332  *	supplies the next file to process in the supplied archd structure.
333  * Return:
334  *	0 when contents of arcn have been set with the next file, -1 when done.
335  */
336 
337 int
338 next_file(ARCHD *arcn)
339 {
340 	int cnt;
341 	time_t atime;
342 	time_t mtime;
343 
344 	/*
345 	 * ftree_sel() might have set the ftree_skip flag if the user has the
346 	 * -n option and a file was selected from this file arg tree. (-n says
347 	 * only one member is matched for each pattern) ftree_skip being 1
348 	 * forces us to go to the next arg now.
349 	 */
350 	if (ftree_skip) {
351 		/*
352 		 * clear and go to next arg
353 		 */
354 		ftree_skip = 0;
355 		if (ftree_arg() < 0)
356 			return(-1);
357 	}
358 
359 	/*
360 	 * loop until we get a valid file to process
361 	 */
362 	for(;;) {
363 		if ((ftent = fts_read(ftsp)) == NULL) {
364 			/*
365 			 * out of files in this tree, go to next arg, if none
366 			 * we are done
367 			 */
368 			if (ftree_arg() < 0)
369 				return(-1);
370 			continue;
371 		}
372 
373 		/*
374 		 * handle each type of fts_read() flag
375 		 */
376 		switch(ftent->fts_info) {
377 		case FTS_D:
378 		case FTS_DEFAULT:
379 		case FTS_F:
380 		case FTS_SL:
381 		case FTS_SLNONE:
382 			/*
383 			 * these are all ok
384 			 */
385 			break;
386 		case FTS_DP:
387 			/*
388 			 * already saw this directory. If the user wants file
389 			 * access times reset, we use this to restore the
390 			 * access time for this directory since this is the
391 			 * last time we will see it in this file subtree
392 			 * remember to force the time (this is -t on a read
393 			 * directory, not a created directory).
394 			 */
395 #			ifdef NET2_FTS
396 			if (!tflag || (get_atdir(ftent->fts_statb.st_dev,
397 			    ftent->fts_statb.st_ino, &mtime, &atime) < 0))
398 #			else
399 			if (!tflag || (get_atdir(ftent->fts_statp->st_dev,
400 			    ftent->fts_statp->st_ino, &mtime, &atime) < 0))
401 #			endif
402 				continue;
403 			set_ftime(ftent->fts_path, mtime, atime, 1);
404 			continue;
405 		case FTS_DC:
406 			/*
407 			 * fts claims a file system cycle
408 			 */
409 			paxwarn(1,"File system cycle found at %s",ftent->fts_path);
410 			continue;
411 		case FTS_DNR:
412 #			ifdef NET2_FTS
413 			syswarn(1, errno,
414 #			else
415 			syswarn(1, ftent->fts_errno,
416 #			endif
417 			    "Unable to read directory %s", ftent->fts_path);
418 			continue;
419 		case FTS_ERR:
420 #			ifdef NET2_FTS
421 			syswarn(1, errno,
422 #			else
423 			syswarn(1, ftent->fts_errno,
424 #			endif
425 			    "File system traversal error");
426 			continue;
427 		case FTS_NS:
428 		case FTS_NSOK:
429 #			ifdef NET2_FTS
430 			syswarn(1, errno,
431 #			else
432 			syswarn(1, ftent->fts_errno,
433 #			endif
434 			    "Unable to access %s", ftent->fts_path);
435 			continue;
436 		}
437 
438 		/*
439 		 * ok got a file tree node to process. copy info into arcn
440 		 * structure (initialize as required)
441 		 */
442 		arcn->skip = 0;
443 		arcn->pad = 0;
444 		arcn->ln_nlen = 0;
445 		arcn->ln_name[0] = '\0';
446 #		ifdef NET2_FTS
447 		arcn->sb = ftent->fts_statb;
448 #		else
449 		arcn->sb = *(ftent->fts_statp);
450 #		endif
451 
452 		/*
453 		 * file type based set up and copy into the arcn struct
454 		 * SIDE NOTE:
455 		 * we try to reset the access time on all files and directories
456 		 * we may read when the -t flag is specified. files are reset
457 		 * when we close them after copying. we reset the directories
458 		 * when we are done with their file tree (we also clean up at
459 		 * end in case we cut short a file tree traversal). However
460 		 * there is no way to reset access times on symlinks.
461 		 */
462 		switch(S_IFMT & arcn->sb.st_mode) {
463 		case S_IFDIR:
464 			arcn->type = PAX_DIR;
465 			if (!tflag)
466 				break;
467 			add_atdir(ftent->fts_path, arcn->sb.st_dev,
468 			    arcn->sb.st_ino, arcn->sb.st_mtime,
469 			    arcn->sb.st_atime);
470 			break;
471 		case S_IFCHR:
472 			arcn->type = PAX_CHR;
473 			break;
474 		case S_IFBLK:
475 			arcn->type = PAX_BLK;
476 			break;
477 		case S_IFREG:
478 			/*
479 			 * only regular files with have data to store on the
480 			 * archive. all others will store a zero length skip.
481 			 * the skip field is used by pax for actual data it has
482 			 * to read (or skip over).
483 			 */
484 			arcn->type = PAX_REG;
485 			arcn->skip = arcn->sb.st_size;
486 			break;
487 		case S_IFLNK:
488 			arcn->type = PAX_SLK;
489 			/*
490 			 * have to read the symlink path from the file
491 			 */
492 			if ((cnt = readlink(ftent->fts_path, arcn->ln_name,
493 			    PAXPATHLEN - 1)) < 0) {
494 				syswarn(1, errno, "Unable to read symlink %s",
495 				    ftent->fts_path);
496 				continue;
497 			}
498 			/*
499 			 * set link name length, watch out readlink does not
500 			 * always NUL terminate the link path
501 			 */
502 			arcn->ln_name[cnt] = '\0';
503 			arcn->ln_nlen = cnt;
504 			break;
505 		case S_IFSOCK:
506 			/*
507 			 * under BSD storing a socket is senseless but we will
508 			 * let the format specific write function make the
509 			 * decision of what to do with it.
510 			 */
511 			arcn->type = PAX_SCK;
512 			break;
513 		case S_IFIFO:
514 			arcn->type = PAX_FIF;
515 			break;
516 		}
517 		break;
518 	}
519 
520 	/*
521 	 * copy file name, set file name length
522 	 */
523 	arcn->nlen = l_strncpy(arcn->name, ftent->fts_path, sizeof(arcn->name) - 1);
524 	arcn->name[arcn->nlen] = '\0';
525 	arcn->org_name = ftent->fts_path;
526 	return(0);
527 }
528