xref: /netbsd-src/sbin/restore/symtab.c (revision 734527fb7d4ba6a89f7fe2798936ed9501e6e255)
1 /*	$NetBSD: symtab.c,v 1.31 2024/02/10 08:24:50 andvar Exp $	*/
2 
3 /*
4  * Copyright (c) 1983, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  */
31 
32 #include <sys/cdefs.h>
33 #ifndef lint
34 #if 0
35 static char sccsid[] = "@(#)symtab.c	8.3 (Berkeley) 4/28/95";
36 #else
37 __RCSID("$NetBSD: symtab.c,v 1.31 2024/02/10 08:24:50 andvar Exp $");
38 #endif
39 #endif /* not lint */
40 
41 /*
42  * These routines maintain the symbol table which tracks the state
43  * of the file system being restored. They provide lookup by either
44  * name or inode number. They also provide for creation, deletion,
45  * and renaming of entries. Because of the dynamic nature of pathnames,
46  * names should not be saved, but always constructed just before they
47  * are needed, by calling "myname".
48  */
49 
50 #include <sys/param.h>
51 #include <sys/stat.h>
52 
53 #include <ufs/ufs/dinode.h>
54 
55 #include <errno.h>
56 #include <fcntl.h>
57 #include <stdio.h>
58 #include <stdlib.h>
59 #include <string.h>
60 #include <unistd.h>
61 
62 #include "restore.h"
63 #include "extern.h"
64 
65 /*
66  * The following variables define the inode symbol table.
67  * The primary hash table is dynamically allocated based on
68  * the number of inodes in the file system (maxino), scaled by
69  * HASHFACTOR. The variable "entry" points to the hash table;
70  * the variable "entrytblsize" indicates its size (in entries).
71  */
72 #define HASHFACTOR 5
73 static struct entry **entry;
74 static long entrytblsize;
75 
76 static void		 addino(ino_t, struct entry *);
77 static struct entry	*lookupparent(const char *);
78 static void		 removeentry(struct entry *);
79 
80 /*
81  * Look up an entry by inode number
82  */
83 struct entry *
lookupino(ino_t inum)84 lookupino(ino_t inum)
85 {
86 	struct entry *ep;
87 
88 	if (inum < UFS_WINO || inum >= maxino)
89 		return (NULL);
90 	for (ep = entry[inum % entrytblsize]; ep != NULL; ep = ep->e_next)
91 		if (ep->e_ino == inum)
92 			return (ep);
93 	return (NULL);
94 }
95 
96 /*
97  * Add an entry into the entry table
98  */
99 static void
addino(ino_t inum,struct entry * np)100 addino(ino_t inum, struct entry *np)
101 {
102 	struct entry **epp;
103 
104 	if (inum < UFS_WINO || inum >= maxino)
105 		panic("addino: out of range %ju\n", (uintmax_t)inum);
106 	epp = &entry[inum % entrytblsize];
107 	np->e_ino = inum;
108 	np->e_next = *epp;
109 	*epp = np;
110 	if (dflag)
111 		for (np = np->e_next; np != NULL; np = np->e_next)
112 			if (np->e_ino == inum)
113 				badentry(np, "duplicate inum");
114 }
115 
116 /*
117  * Delete an entry from the entry table
118  */
119 void
deleteino(ino_t inum)120 deleteino(ino_t inum)
121 {
122 	struct entry *next;
123 	struct entry **prev;
124 
125 	if (inum < UFS_WINO || inum >= maxino)
126 		panic("deleteino: out of range %ju\n",
127 		    (uintmax_t)inum);
128 	prev = &entry[inum % entrytblsize];
129 	for (next = *prev; next != NULL; next = next->e_next) {
130 		if (next->e_ino == inum) {
131 			next->e_ino = 0;
132 			*prev = next->e_next;
133 			return;
134 		}
135 		prev = &next->e_next;
136 	}
137 	panic("deleteino: %ju not found\n", (uintmax_t)inum);
138 }
139 
140 /*
141  * Look up an entry by name
142  */
143 struct entry *
lookupname(const char * name)144 lookupname(const char *name)
145 {
146 	struct entry *ep;
147 	char *np;
148 	const char *cp;
149 	char buf[MAXPATHLEN];
150 
151 	cp = name;
152 	for (ep = lookupino(UFS_ROOTINO); ep != NULL; ep = ep->e_entries) {
153 		for (np = buf; *cp != '/' && *cp != '\0'; )
154 			*np++ = *cp++;
155 		*np = '\0';
156 		for ( ; ep != NULL; ep = ep->e_sibling)
157 			if (strcmp(ep->e_name, buf) == 0)
158 				break;
159 		if (ep == NULL)
160 			break;
161 		if (*cp++ == '\0')
162 			return (ep);
163 	}
164 	return (NULL);
165 }
166 
167 /*
168  * Look up the parent of a pathname
169  */
170 static struct entry *
lookupparent(const char * name)171 lookupparent(const char *name)
172 {
173 	struct entry *ep;
174 	char *tailindex;
175 
176 	tailindex = strrchr(name, '/');
177 	if (tailindex == NULL)
178 		return (NULL);
179 	*tailindex = '\0';
180 	ep = lookupname(name);
181 	*tailindex = '/';
182 	if (ep == NULL)
183 		return (NULL);
184 	if (ep->e_type != NODE)
185 		panic("%s is not a directory\n", name);
186 	return (ep);
187 }
188 
189 /*
190  * Determine the current pathname of a node or leaf
191  */
192 char *
myname(struct entry * ep)193 myname(struct entry *ep)
194 {
195 	char *cp;
196 	static char namebuf[MAXPATHLEN];
197 
198 	for (cp = &namebuf[MAXPATHLEN - 2]; cp > &namebuf[ep->e_namlen]; ) {
199 		cp -= ep->e_namlen;
200 		memmove(cp, ep->e_name, (long)ep->e_namlen);
201 		if (ep == lookupino(UFS_ROOTINO))
202 			return (cp);
203 		*(--cp) = '/';
204 		ep = ep->e_parent;
205 	}
206 	panic("%s: pathname too long\n", cp);
207 	return(cp);
208 }
209 
210 /*
211  * Unused symbol table entries are linked together on a freelist
212  * headed by the following pointer.
213  */
214 static struct entry *freelist = NULL;
215 
216 /*
217  * add an entry to the symbol table
218  */
219 struct entry *
addentry(const char * name,ino_t inum,int type)220 addentry(const char *name, ino_t inum, int type)
221 {
222 	struct entry *np, *ep;
223 
224 	if (freelist == NULL) {
225 		np = malloc(pagesize);
226 		if (np == NULL)
227 			panic("no memory to extend symbol table\n");
228 		for (ep = (struct entry *)((char *)np + pagesize) - 1;
229 		    np <= ep; np++) {
230 			np->e_next = freelist;
231 			freelist = np;
232 		}
233 	}
234 	np = freelist;
235 	freelist = np->e_next;
236 	memset(np, 0, sizeof(struct entry));
237 
238 	np->e_type = type & ~LINK;
239 	ep = lookupparent(name);
240 	if (ep == NULL) {
241 		if (inum != UFS_ROOTINO || lookupino(UFS_ROOTINO) != NULL)
242 			panic("bad name to addentry %s\n", name);
243 		np->e_name = savename(name);
244 		np->e_namlen = strlen(name);
245 		np->e_parent = np;
246 		addino(UFS_ROOTINO, np);
247 		return (np);
248 	}
249 	np->e_name = savename(strrchr(name, '/') + 1);
250 	np->e_namlen = strlen(np->e_name);
251 	np->e_parent = ep;
252 	np->e_sibling = ep->e_entries;
253 	ep->e_entries = np;
254 	if (type & LINK) {
255 		ep = lookupino(inum);
256 		if (ep == NULL)
257 			panic("link to non-existent name\n");
258 		np->e_ino = inum;
259 		np->e_links = ep->e_links;
260 		ep->e_links = np;
261 	} else if (inum != 0) {
262 		if (lookupino(inum) != NULL)
263 			panic("duplicate entry\n");
264 		addino(inum, np);
265 	}
266 	return (np);
267 }
268 
269 /*
270  * delete an entry from the symbol table
271  */
272 void
freeentry(struct entry * ep)273 freeentry(struct entry *ep)
274 {
275 	struct entry *np;
276 	ino_t inum;
277 
278 	if (ep->e_flags != REMOVED)
279 		badentry(ep, "not marked REMOVED");
280 	if (ep->e_type == NODE) {
281 		if (ep->e_links != NULL)
282 			badentry(ep, "freeing referenced directory");
283 		if (ep->e_entries != NULL)
284 			badentry(ep, "freeing non-empty directory");
285 	}
286 	if (ep->e_ino != 0) {
287 		np = lookupino(ep->e_ino);
288 		if (np == NULL)
289 			badentry(ep, "lookupino failed");
290 		if (np == ep) {
291 			inum = ep->e_ino;
292 			deleteino(inum);
293 			if (ep->e_links != NULL)
294 				addino(inum, ep->e_links);
295 		} else {
296 			for (; np != NULL; np = np->e_links) {
297 				if (np->e_links == ep) {
298 					np->e_links = ep->e_links;
299 					break;
300 				}
301 			}
302 			if (np == NULL)
303 				badentry(ep, "link not found");
304 		}
305 	}
306 	removeentry(ep);
307 	freename(ep->e_name);
308 	ep->e_next = freelist;
309 	freelist = ep;
310 }
311 
312 /*
313  * Relocate an entry in the tree structure
314  */
315 void
moveentry(struct entry * ep,const char * newname)316 moveentry(struct entry *ep, const char *newname)
317 {
318 	struct entry *np;
319 	char *cp;
320 
321 	np = lookupparent(newname);
322 	if (np == NULL)
323 		badentry(ep, "cannot move ROOT");
324 	if (np != ep->e_parent) {
325 		removeentry(ep);
326 		ep->e_parent = np;
327 		ep->e_sibling = np->e_entries;
328 		np->e_entries = ep;
329 	}
330 	cp = strrchr(newname, '/') + 1;
331 	freename(ep->e_name);
332 	ep->e_name = savename(cp);
333 	ep->e_namlen = strlen(cp);
334 	if (strcmp(gentempname(ep), ep->e_name) == 0)
335 		ep->e_flags |= TMPNAME;
336 	else
337 		ep->e_flags &= ~TMPNAME;
338 }
339 
340 /*
341  * Remove an entry in the tree structure
342  */
343 static void
removeentry(struct entry * ep)344 removeentry(struct entry *ep)
345 {
346 	struct entry *np;
347 
348 	np = ep->e_parent;
349 	if (np->e_entries == ep) {
350 		np->e_entries = ep->e_sibling;
351 	} else {
352 		for (np = np->e_entries; np != NULL; np = np->e_sibling) {
353 			if (np->e_sibling == ep) {
354 				np->e_sibling = ep->e_sibling;
355 				break;
356 			}
357 		}
358 		if (np == NULL)
359 			badentry(ep, "cannot find entry in parent list");
360 	}
361 }
362 
363 /*
364  * Table of unused string entries, sorted by length.
365  *
366  * Entries are allocated in STRTBLINCR sized pieces so that names
367  * of similar lengths can use the same entry. The value of STRTBLINCR
368  * is chosen so that every entry has at least enough space to hold
369  * a "struct strtbl" header. Thus every entry can be linked onto an
370  * appropriate free list.
371  *
372  * NB. The macro "allocsize" below assumes that "struct strhdr"
373  *     has a size that is a power of two.
374  */
375 struct strhdr {
376 	struct strhdr *next;
377 };
378 
379 #define STRTBLINCR	(sizeof(struct strhdr))
380 #define allocsize(size)	(((size) + 1 + STRTBLINCR - 1) & ~(STRTBLINCR - 1))
381 
382 static struct strhdr strtblhdr[allocsize(NAME_MAX) / STRTBLINCR];
383 
384 /*
385  * Allocate space for a name. It first looks to see if it already
386  * has an appropriate sized entry, and if not allocates a new one.
387  */
388 char *
savename(const char * name)389 savename(const char *name)
390 {
391 	struct strhdr *np, *tp;
392 	long len, siz;
393 	char *cp, *ep;
394 
395 	if (name == NULL)
396 		panic("bad name\n");
397 	len = strlen(name);
398 	tp = &strtblhdr[len / STRTBLINCR];
399 	if (tp->next == NULL) {
400 		cp = malloc(pagesize);
401 		if (cp == NULL)
402 			panic("no space for string table\n");
403 		for (siz = allocsize(len), ep = (cp + pagesize) - siz;
404 		    cp <= ep; cp += siz) {
405 			np = (struct strhdr *)cp;
406 			np->next = tp->next;
407 			tp->next = np;
408 		}
409 	}
410 	np = tp->next;
411 	tp->next = np->next;
412 	cp = (char *)np;
413 	(void) strcpy(cp, name);
414 	return (cp);
415 }
416 
417 /*
418  * Free space for a name. The resulting entry is linked onto the
419  * appropriate free list.
420  */
421 void
freename(char * name)422 freename(char *name)
423 {
424 	struct strhdr *tp, *np;
425 
426 	tp = &strtblhdr[strlen(name) / STRTBLINCR];
427 	np = (struct strhdr *)name;
428 	np->next = tp->next;
429 	tp->next = np;
430 }
431 
432 /*
433  * Useful quantities placed at the end of a dumped symbol table.
434  */
435 struct symtableheader {
436 	int32_t	volno;
437 	int32_t	stringsize;
438 	int32_t	entrytblsize;
439 	time_t	dumptime;
440 	time_t	dumpdate;
441 	ino_t	maxino;
442 	int32_t	ntrec;
443 };
444 
445 /*
446  * dump a snapshot of the symbol table
447  */
448 void
dumpsymtable(const char * filename,int32_t checkpt)449 dumpsymtable(const char *filename, int32_t checkpt)
450 {
451 	struct entry *ep, *tep;
452 	ino_t i;
453 	long l;
454 	struct entry temp, *tentry;
455 	long mynum = 1, stroff = 0;
456 	FILE *fd;
457 	struct symtableheader hdr;
458 
459 	vprintf(stdout, "Checkpointing the restore\n");
460 	if (Nflag)
461 		return;
462 	if ((fd = fopen(filename, "w")) == NULL) {
463 		fprintf(stderr, "fopen: %s\n", strerror(errno));
464 		panic("cannot create save file %s for symbol table\n",
465 			filename);
466 	}
467 	clearerr(fd);
468 	/*
469 	 * Assign indices to each entry
470 	 * Write out the string entries
471 	 */
472 	for (i = UFS_WINO; i <= maxino; i++) {
473 		for (ep = lookupino(i); ep != NULL; ep = ep->e_links) {
474 			ep->e_index = mynum++;
475 			(void) fwrite(ep->e_name, sizeof(char),
476 			       (int)allocsize(ep->e_namlen), fd);
477 		}
478 	}
479 	/*
480 	 * Convert pointers to indexes, and output
481 	 */
482 	tep = &temp;
483 	stroff = 0;
484 	for (i = UFS_WINO; i <= maxino; i++) {
485 		for (ep = lookupino(i); ep != NULL; ep = ep->e_links) {
486 			memmove(tep, ep, (long)sizeof(struct entry));
487 			tep->e_name = (char *)stroff;
488 			stroff += allocsize(ep->e_namlen);
489 			tep->e_parent = (struct entry *)(long)
490 			    ep->e_parent->e_index;
491 			if (ep->e_links != NULL)
492 				tep->e_links = (struct entry *)(long)
493 				    ep->e_links->e_index;
494 			if (ep->e_sibling != NULL)
495 				tep->e_sibling = (struct entry *)(long)
496 				    ep->e_sibling->e_index;
497 			if (ep->e_entries != NULL)
498 				tep->e_entries = (struct entry *)(long)
499 				    ep->e_entries->e_index;
500 			if (ep->e_next != NULL)
501 				tep->e_next = (struct entry *)(long)
502 				    ep->e_next->e_index;
503 			(void) fwrite((char *)tep, sizeof(struct entry), 1, fd);
504 		}
505 	}
506 	/*
507 	 * Convert entry pointers to indexes, and output
508 	 */
509 	for (l = 0; l < entrytblsize; l++) {
510 		if (entry[l] == NULL)
511 			tentry = NULL;
512 		else
513 			tentry = (struct entry *)(long)entry[l]->e_index;
514 		(void) fwrite((char *)&tentry, sizeof(struct entry *), 1, fd);
515 	}
516 	hdr.volno = checkpt;
517 	hdr.maxino = maxino;
518 	hdr.entrytblsize = entrytblsize;
519 	hdr.stringsize = stroff;
520 	hdr.dumptime = dumptime;
521 	hdr.dumpdate = dumpdate;
522 	hdr.ntrec = ntrec;
523 	(void) fwrite((char *)&hdr, sizeof(struct symtableheader), 1, fd);
524 	if (ferror(fd)) {
525 		fprintf(stderr, "fwrite: %s\n", strerror(errno));
526 		panic("output error to file %s writing symbol table\n",
527 			filename);
528 	}
529 	(void) fclose(fd);
530 }
531 
532 /*
533  * Initialize a symbol table from a file
534  */
535 void
initsymtable(const char * filename)536 initsymtable(const char *filename)
537 {
538 	char *base;
539 	long tblsize;
540 	struct entry *ep;
541 	struct entry *baseep, *lep;
542 	struct symtableheader hdr;
543 	struct stat stbuf;
544 	long i;
545 	int fd;
546 
547 	vprintf(stdout, "Initialize symbol table.\n");
548 	if (filename == NULL) {
549 		entrytblsize = maxino / HASHFACTOR;
550 		entry = (struct entry **)
551 			calloc((unsigned)entrytblsize, sizeof(struct entry *));
552 		if (entry == NULL)
553 			panic("no memory for entry table\n");
554 		ep = addentry(".", UFS_ROOTINO, NODE);
555 		ep->e_flags |= NEW;
556 		return;
557 	}
558 	if ((fd = open(filename, O_RDONLY, 0)) < 0) {
559 		fprintf(stderr, "open: %s\n", strerror(errno));
560 		panic("cannot open symbol table file %s\n", filename);
561 	}
562 	if (fstat(fd, &stbuf) < 0) {
563 		fprintf(stderr, "stat: %s\n", strerror(errno));
564 		panic("cannot stat symbol table file %s\n", filename);
565 	}
566 	tblsize = stbuf.st_size - sizeof(struct symtableheader);
567 	base = calloc((unsigned)tblsize, sizeof(char));
568 	if (base == NULL)
569 		panic("cannot allocate space for symbol table\n");
570 	if (read(fd, base, (int)tblsize) < 0 ||
571 	    read(fd, (char *)&hdr, sizeof(struct symtableheader)) < 0) {
572 		fprintf(stderr, "read: %s\n", strerror(errno));
573 		panic("cannot read symbol table file %s\n", filename);
574 	}
575 	(void)close(fd);
576 	switch (command) {
577 	case 'r':
578 		/*
579 		 * For normal continuation, insure that we are using
580 		 * the next incremental tape
581 		 */
582 		if (hdr.dumpdate != dumptime) {
583 			if (hdr.dumpdate < dumptime)
584 				fprintf(stderr, "Incremental tape too low\n");
585 			else
586 				fprintf(stderr, "Incremental tape too high\n");
587 			exit(1);
588 		}
589 		break;
590 	case 'R':
591 		/*
592 		 * For restart, insure that we are using the same tape
593 		 */
594 		curfile.action = SKIP;
595 		dumptime = hdr.dumptime;
596 		dumpdate = hdr.dumpdate;
597 		if (!bflag)
598 			newtapebuf(hdr.ntrec);
599 		getvol(hdr.volno);
600 		break;
601 	default:
602 		panic("initsymtable called from command %c\n", command);
603 		break;
604 	}
605 	maxino = hdr.maxino;
606 	entrytblsize = hdr.entrytblsize;
607 	entry = (struct entry **)
608 		(base + tblsize - (entrytblsize * sizeof(struct entry *)));
609 	baseep = (struct entry *)(base + hdr.stringsize - sizeof(struct entry));
610 	lep = (struct entry *)entry;
611 	for (i = 0; i < entrytblsize; i++) {
612 		if (entry[i] == NULL)
613 			continue;
614 		entry[i] = &baseep[(long)entry[i]];
615 	}
616 	for (ep = &baseep[1]; ep < lep; ep++) {
617 		ep->e_name = base + (long)ep->e_name;
618 		ep->e_parent = &baseep[(long)ep->e_parent];
619 		if (ep->e_sibling != NULL)
620 			ep->e_sibling = &baseep[(long)ep->e_sibling];
621 		if (ep->e_links != NULL)
622 			ep->e_links = &baseep[(long)ep->e_links];
623 		if (ep->e_entries != NULL)
624 			ep->e_entries = &baseep[(long)ep->e_entries];
625 		if (ep->e_next != NULL)
626 			ep->e_next = &baseep[(long)ep->e_next];
627 	}
628 }
629