xref: /csrg-svn/bin/pax/ftree.c (revision 57111)
1 /*-
2  * Copyright (c) 1992 Keith Muller.
3  * Copyright (c) 1992 The Regents of the University of California.
4  * 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  * %sccs.include.redist.c%
10  */
11 
12 #ifndef lint
13 static char sccsid[] = "@(#)ftree.c	1.1 (Berkeley) 12/13/92";
14 #endif /* not lint */
15 
16 #include <sys/types.h>
17 #include <sys/time.h>
18 #include <sys/stat.h>
19 #include <sys/param.h>
20 #include <unistd.h>
21 #include <string.h>
22 #include <stdio.h>
23 #include <ctype.h>
24 #include <errno.h>
25 #include <stdlib.h>
26 #include <fts.h>
27 #include "pax.h"
28 #include "ftree.h"
29 #include "extern.h"
30 
31 /*
32  * routines to interface with the fts library function.
33  *
34  * file args supplied to pax are stored on a single linked list (of type FTREE)
35  * and given to fts to be processed one at a time. pax "selects" files from
36  * the expansion of each arg into the corresponding file tree (if the arg is a
37  * directory, otherwise the node itself is just passed to pax). The selection
38  * is modified by the -n and -u flags. The user is informed when a specific
39  * file arg does not generate any selected files. -n keeps expanding the file
40  * tree arg until one of its files is selected, then skips to the next file
41  * arg. when the user does not supply the file trees as command line args to
42  * pax, they are read from stdin
43  */
44 
45 static FTS *ftsp = NULL;		/* curent FTS handle */
46 static int ftsopts;			/* options to be used on fts_open */
47 static char *farray[2];			/* array for passing each arg to fts */
48 static FTREE *fthead = NULL;		/* head of linked list of file args */
49 static FTREE *fttail = NULL;		/* tail of linked list of file args */
50 static FTREE *ftcur = NULL;		/* current file arg being processed */
51 static FTSENT *ftent = NULL;		/* current file tree entry */
52 static int refcnt = 1;			/* ref count for stdin supplied arg */
53 static int ftree_skip;			/* when set skip to next file arg */
54 
55 static int ftree_arg __P((void));
56 
57 /*
58  * ftree_start()
59  *	initialize the options passed to fts_open() during this run of pax
60  *	options are based on the selection of pax options by the user
61  *	fts_start() also calls fts_arg() to open the first valid file arg. We
62  *	also attempt to reset directory access times when -t (tflag) is set.
63  * Return:
64  *	0 if there is at least one valid file arg to process, -1 otherwise
65  */
66 
67 #if __STDC__
68 int
69 ftree_start(void)
70 #else
71 int
72 ftree_start()
73 #endif
74 {
75 	/*
76 	 * set up the operation mode of fts, open the first file arg. We must
77 	 * use FTS_NOCHDIR, as the user may have to open multiple archives and
78 	 * if fts did a chdir off into the boondocks, we may create an archive
79 	 * volume in an place where the user did not expect to.
80 	 */
81 	ftsopts = FTS_NOCHDIR;
82 
83 	/*
84 	 * optional user flags that effect file traversal
85 	 * -H command line symlink follow only
86 	 * -L follow sylinks
87 	 * -X do not cross over mount points
88 	 * -t preserve access times on files read.
89 	 * -n select only the first member of a file tree when a match is found
90 	 * -d do not extract subtrees rooted at a directory arg.
91 	 */
92 	if (Lflag)
93 		ftsopts |= FTS_LOGICAL;
94 	else
95 		ftsopts |= FTS_PHYSICAL;
96 	if (Hflag)
97 #	ifdef NET2_FTS
98 		warn(0, "The -H flag is not supported on this version");
99 #	else
100 		ftsopts |= FTS_COMFOLLOW;
101 #	endif
102 	if (Xflag)
103 		ftsopts |= FTS_XDEV;
104 
105 	if ((fthead == NULL) && ((farray[0] = malloc(PAXPATHLEN+2)) == NULL)) {
106 		warn(1, "Unable to allocate memory for file name buffer");
107 		return(-1);
108 	}
109 
110 	if (ftree_arg() < 0)
111 		return(-1);
112 	if (tflag && (atdir_start() < 0))
113 		return(-1);
114 	return(0);
115 }
116 
117 /*
118  * ftree_add()
119  *	add the arg to the linked list of files to process. Each will be
120  *	processed by fts one at a time
121  * Return:
122  *	0 if added to the linked list, -1 if failed
123  */
124 
125 #if __STDC__
126 int
127 ftree_add(register char *str)
128 #else
129 int
130 ftree_add(str)
131 	register char *str;
132 #endif
133 {
134 	register FTREE *ft;
135 	register int len;
136 
137 	/*
138 	 * simple check for bad args
139 	 */
140 	if ((str == NULL) || (*str == '\0')) {
141 		warn(0, "Invalid file name arguement");
142 		return(-1);
143 	}
144 
145 	/*
146 	 * allocate FTREE node and add to the end of the linked list (args are
147 	 * processed in the same order they were passed to pax). Get rid of any
148 	 * trailing / the user may pass us. (watch out for / by itself).
149 	 */
150 	if ((ft = (FTREE *)malloc(sizeof(FTREE))) == NULL) {
151 		warn(0, "Unable to allocate memory for filename");
152 		return(-1);
153 	}
154 
155 	if (((len = strlen(str) - 1) > 0) && (str[len] == '/'))
156 		str[len] = '\0';
157 	ft->fname = str;
158 	ft->refcnt = 0;
159 	ft->fow = NULL;
160 	if (fthead == NULL) {
161 		fttail = fthead = ft;
162 		return(0);
163 	}
164 	fttail->fow = ft;
165 	fttail = ft;
166 	return(0);
167 }
168 
169 /*
170  * ftree_sel()
171  *	this entry has been selected by pax. bump up reference count and handle
172  *	-n and -d processing.
173  */
174 
175 #if __STDC__
176 void
177 ftree_sel(register ARCHD *arcn)
178 #else
179 void
180 ftree_sel(arcn)
181 	register ARCHD *arcn;
182 #endif
183 {
184 	/*
185 	 * set reference bit for this pattern. This linked list is only used
186 	 * when files trees are supplied pax as args. The list is not used when
187 	 * the trees are read from stdin. stdin processing uses a global
188 	 * reference count for the last path input.
189 	 */
190 	if (ftcur != NULL)
191 		ftcur->refcnt = 1;
192 	else
193 		refcnt = 1;
194 
195 	/*
196 	 * if -n we are done with this arg, force a skip to the next arg when
197 	 * pax asks for the next file in next_file().
198 	 * if -d we tell fts only to match the directory (if the arg is a dir)
199 	 * and not the entire file tree rooted at that point.
200 	 */
201 	if (nflag)
202 		ftree_skip = 1;
203 
204 	if (!dflag || (arcn->type != PAX_DIR))
205 		return;
206 	if (ftent != NULL)
207 		(void)fts_set(ftsp, ftent, FTS_SKIP);
208 }
209 
210 /*
211  * ftree_chk()
212  *	called at end on pax execution. Prints all those file args that did not
213  *	have a selected member (reference count still 0)
214  */
215 
216 #if __STDC__
217 void
218 ftree_chk(void)
219 #else
220 void
221 ftree_chk()
222 #endif
223 {
224 	register FTREE *ft;
225 	register int wban = 0;
226 
227 	/*
228 	 * make sure all dir access times were reset.
229 	 */
230 	if (tflag)
231 		atdir_end();
232 
233 	/*
234 	 * walk down list and check reference count. Print out those members
235 	 * that never had a match
236 	 */
237 	for (ft = fthead; ft != NULL; ft = ft->fow) {
238 		if (ft->refcnt > 0)
239 			continue;
240 		if (wban == 0) {
241 			warn(1,"WARNING! These file names were not selected:");
242 			++wban;
243 		}
244 		(void)fprintf(stderr, "%s\n", ft->fname);
245 	}
246 }
247 
248 /*
249  * ftree_arg()
250  *	Get the next file arg for fts to process. Can be from either the linked
251  *	list or read from stdin when the user did not them as args to pax. Each
252  *	arg is processed until the first successful fts_open().
253  * Return:
254  *	0 when the next arg is ready to go, -1 if out of file args (or EOF on
255  *	stdin).
256  */
257 
258 #if __STDC__
259 static int
260 ftree_arg(void)
261 #else
262 static int
263 ftree_arg()
264 #endif
265 {
266 	register 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 	 * keep looping until we get a valid file tree to process. Stop when we
277 	 * reach the end of the list (or get an eof on stdin)
278 	 */
279 	for(;;) {
280 		if (fthead == NULL) {
281 			/*
282 			 * the user didn't supply any args, get the file trees
283 			 * to process from stdin; if the refenece count is 0
284 			 * complain that the last stdin read file tree did not
285 			 * supply any selected members
286 			 */
287 			refcnt = 0;
288 			if (fgets(farray[0], PAXPATHLEN+1, stdin) == NULL)
289 				return(-1);
290 			if ((pt = strchr(farray[0], '\n')) != NULL)
291 				*pt = '\0';
292 		} else {
293 			/*
294 			 * the user supplied the file args as arguements to pax
295 			 * store the current reference count (unless we have not
296 			 * processed any files yet).
297 			 */
298 			if (ftcur != NULL) {
299 				ftcur->refcnt = refcnt;
300 				if ((ftcur = ftcur->fow) == NULL)
301 					return(-1);
302 			} else
303 				ftcur = fthead;
304 			farray[0] = ftcur->fname;
305 		}
306 
307 		/*
308 		 * watch it, fts wants the file arg stored in a array of char
309 		 * ptrs, with the last one a null. we use a two element array
310 		 * and set farray[0] to point at the buffer with the file name
311 		 * in it. We cannnot pass all the file args to fts at one shot
312 		 * as we need to keep a handle on which file arg generates what
313 		 * files (the -n and -d flags need this). If the open is
314 		 * successful, return a 0.
315 		 */
316 		if ((ftsp = fts_open(farray, ftsopts, NULL)) != NULL)
317 			break;
318 	}
319 	return(0);
320 }
321 
322 /*
323  * next_file()
324  *	supplies the next file to process in the supplied archd structure.
325  * Return:
326  *	0 when contents of arcn have been set with the next file, -1 when done.
327  */
328 
329 #if __STDC__
330 int
331 next_file(register ARCHD *arcn)
332 #else
333 int
334 next_file(arcn)
335 	register ARCHD *arcn;
336 #endif
337 {
338 	register int cnt;
339 	time_t atime;
340 	time_t mtime;
341 
342 	/*
343 	 * ftree_sel() might have set the ftree_skip flag if the user has the
344 	 * -n option and a file was selected from this file arg tree. (-n says
345 	 * only one member is matched for each pattern) ftree_skip being 1
346 	 ( forces us to go to the next arg now.
347 	 */
348 	if (ftree_skip) {
349 		/*
350 		 * clear and go to next arg
351 		 */
352 		ftree_skip = 0;
353 		if (ftree_arg() < 0)
354 			return(-1);
355 	}
356 
357 	/*
358 	 * loop until we get a valid file to process
359 	 */
360 	for(;;) {
361 		if ((ftent = fts_read(ftsp)) == NULL) {
362 			/*
363 			 * out of files in this tree, go to next arg, if none
364 			 * we are done
365 			 */
366 			if (ftree_arg() < 0)
367 				return(-1);
368 			continue;
369 		}
370 
371 		/*
372 		 * handle each type of fts_read() flag
373 		 */
374 		switch(ftent->fts_info) {
375 		case FTS_D:
376 		case FTS_DEFAULT:
377 		case FTS_F:
378 		case FTS_SL:
379 		case FTS_SLNONE:
380 			/*
381 			 * these are all ok
382 			 */
383 			break;
384 		case FTS_DP:
385 			/*
386 			 * already saw this directory. If the user wants file
387 			 * access times reset, we use this to restore the
388 			 * access time for this directory since this is the
389 			 * last time we will see it in this file subtree
390 			 * remember to force the time (this is -t on a read
391 			 * directory, not a created directory).
392 			 */
393 #			ifdef NET2_FTS
394 			if (!tflag || (get_atdir(ftent->fts_statb.st_dev,
395 			    ftent->fts_statb.st_ino, &mtime, &atime) < 0))
396 #			else
397 			if (!tflag || (get_atdir(ftent->fts_statp->st_dev,
398 			    ftent->fts_statp->st_ino, &mtime, &atime) < 0))
399 #			endif
400 				continue;
401 			set_ftime(ftent->fts_path, mtime, atime, 1);
402 			continue;
403 		case FTS_DC:
404 			/*
405 			 * fts claims a file system cycle
406 			 */
407 			warn(1,"File system cycle found at %s",ftent->fts_path);
408 			continue;
409 		case FTS_DNR:
410 #			ifdef NET2_FTS
411 			syswarn(1, errno,
412 #			else
413 			syswarn(1, ftent->fts_errno,
414 #			endif
415 			    "Unable to read directory %s", ftent->fts_path);
416 			continue;
417 		case FTS_ERR:
418 #			ifdef NET2_FTS
419 			syswarn(1, errno,
420 #			else
421 			syswarn(1, ftent->fts_errno,
422 #			endif
423 			    "File system traversal error");
424 			continue;
425 		case FTS_NS:
426 		case FTS_NSOK:
427 #			ifdef NET2_FTS
428 			syswarn(1, errno,
429 #			else
430 			syswarn(1, ftent->fts_errno,
431 #			endif
432 			    "Unable to access %s", ftent->fts_path);
433 			continue;
434 		}
435 
436 		/*
437 		 * ok got a file tree node to process. copy info into arcn
438 		 * structure (initialize as required)
439 		 */
440 		arcn->skip = 0;
441 		arcn->pad = 0;
442 		arcn->ln_nlen = 0;
443 		arcn->ln_name[0] = '\0';
444 #		ifdef NET2_FTS
445 		arcn->sb = ftent->fts_statb;
446 #		else
447 		arcn->sb = *(ftent->fts_statp);
448 #		endif
449 
450 		/*
451 		 * file type based set up and copy into the arcn struct
452 		 * SIDE NOTE:
453 		 * we try to reset the access time on all files and directories
454 		 * we may read when the -t flag is specified. files are reset
455 		 * when we close them after copying. we reset the directories
456 		 * when we are done with their file tree (we also clean up at
457 		 * end in case we cut short a file tree traversal). However
458 		 * there is no way to reset access times on symlinks.
459 		 */
460 		switch(S_IFMT & arcn->sb.st_mode) {
461 		case S_IFDIR:
462 			arcn->type = PAX_DIR;
463 			if (!tflag)
464 				break;
465 			add_atdir(ftent->fts_path, arcn->sb.st_dev,
466 			    arcn->sb.st_ino, arcn->sb.st_mtime,
467 			    arcn->sb.st_atime);
468 			break;
469 		case S_IFCHR:
470 			arcn->type = PAX_CHR;
471 			break;
472 		case S_IFBLK:
473 			arcn->type = PAX_BLK;
474 			break;
475 		case S_IFREG:
476 			/*
477 			 * only regular files with have data to store on the
478 			 * archive. all others will store a zero length skip.
479 			 * the skip field is used by pax for actual data it has
480 			 * to read (or skip over).
481 			 */
482 			arcn->type = PAX_REG;
483 			arcn->skip = arcn->sb.st_size;
484 			break;
485 		case S_IFLNK:
486 			arcn->type = PAX_SLK;
487 			/*
488 			 * have to read the symlink path from the file
489 			 */
490 			if ((cnt = readlink(ftent->fts_path, arcn->ln_name,
491 			    PAXPATHLEN)) < 0) {
492 				syswarn(1, errno, "Unable to read symlink %s",
493 				    ftent->fts_path);
494 				continue;
495 			}
496 			/*
497 			 * set link name length, watch out readlink does not
498 			 * allways null terminate the link path
499 			 */
500 			arcn->ln_name[cnt] = '\0';
501 			arcn->ln_nlen = cnt;
502 			break;
503 		case S_IFSOCK:
504 			/*
505 			 * under BSD storing a socket is senseless but we will
506 			 * let the format specific write function make the
507 			 * decision of what to do with it.
508 			 */
509 			arcn->type = PAX_SCK;
510 			break;
511 		case S_IFIFO:
512 			arcn->type = PAX_FIF;
513 			break;
514 		}
515 		break;
516 	}
517 
518 	/*
519 	 * copy file name, set file name length
520 	 */
521 	arcn->nlen = l_strncpy(arcn->name, ftent->fts_path, PAXPATHLEN+1);
522 	arcn->org_name = ftent->fts_path;
523 	return(0);
524 }
525