traverse.c revision 157660
1/*-
2 * Copyright (c) 1980, 1988, 1991, 1993
3 *	The Regents of the University of California.  All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 *    notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 *    notice, this list of conditions and the following disclaimer in the
12 *    documentation and/or other materials provided with the distribution.
13 * 4. Neither the name of the University nor the names of its contributors
14 *    may be used to endorse or promote products derived from this software
15 *    without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 */
29
30#ifndef lint
31#if 0
32static char sccsid[] = "@(#)traverse.c	8.7 (Berkeley) 6/15/95";
33#endif
34static const char rcsid[] =
35  "$FreeBSD: head/sbin/dump/traverse.c 157660 2006-04-11 14:45:42Z dwmalone $";
36#endif /* not lint */
37
38#include <sys/param.h>
39#include <sys/stat.h>
40
41#include <ufs/ufs/dir.h>
42#include <ufs/ufs/dinode.h>
43#include <ufs/ffs/fs.h>
44
45#include <protocols/dumprestore.h>
46
47#include <ctype.h>
48#include <errno.h>
49#include <inttypes.h>
50#include <limits.h>
51#include <stdio.h>
52#include <stdlib.h>
53#include <string.h>
54#include <timeconv.h>
55#include <unistd.h>
56
57#include "dump.h"
58
59union dinode {
60	struct ufs1_dinode dp1;
61	struct ufs2_dinode dp2;
62};
63#define	DIP(dp, field) \
64	((sblock->fs_magic == FS_UFS1_MAGIC) ? \
65	(dp)->dp1.field : (dp)->dp2.field)
66#define DIP_SET(dp, field, val) do {\
67	if (sblock->fs_magic == FS_UFS1_MAGIC) \
68		(dp)->dp1.field = (val); \
69	else \
70		(dp)->dp2.field = (val); \
71	} while (0)
72
73#define	HASDUMPEDFILE	0x1
74#define	HASSUBDIRS	0x2
75
76static	int dirindir(ino_t ino, ufs2_daddr_t blkno, int level, long *size,
77    long *tapesize, int nodump, ino_t maxino);
78static	void dmpindir(ino_t ino, ufs2_daddr_t blk, int level, off_t *size);
79static	int searchdir(ino_t ino, ufs2_daddr_t blkno, long size, long filesize,
80    long *tapesize, int nodump, ino_t maxino);
81static	long blockest(union dinode *dp);
82
83/*
84 * This is an estimation of the number of TP_BSIZE blocks in the file.
85 * It estimates the number of blocks in files with holes by assuming
86 * that all of the blocks accounted for by di_blocks are data blocks
87 * (when some of the blocks are usually used for indirect pointers);
88 * hence the estimate may be high.
89 */
90static long
91blockest(union dinode *dp)
92{
93	long blkest, sizeest;
94
95	/*
96	 * dp->di_size is the size of the file in bytes.
97	 * dp->di_blocks stores the number of sectors actually in the file.
98	 * If there are more sectors than the size would indicate, this just
99	 *	means that there are indirect blocks in the file or unused
100	 *	sectors in the last file block; we can safely ignore these
101	 *	(blkest = sizeest below).
102	 * If the file is bigger than the number of sectors would indicate,
103	 *	then the file has holes in it.	In this case we must use the
104	 *	block count to estimate the number of data blocks used, but
105	 *	we use the actual size for estimating the number of indirect
106	 *	dump blocks (sizeest vs. blkest in the indirect block
107	 *	calculation).
108	 */
109	if ((DIP(dp, di_flags) & SF_SNAPSHOT) != 0)
110		return (1);
111	blkest = howmany(dbtob(DIP(dp, di_blocks)), TP_BSIZE);
112	sizeest = howmany(DIP(dp, di_size), TP_BSIZE);
113	if (blkest > sizeest)
114		blkest = sizeest;
115	if (DIP(dp, di_size) > sblock->fs_bsize * NDADDR) {
116		/* calculate the number of indirect blocks on the dump tape */
117		blkest +=
118			howmany(sizeest - NDADDR * sblock->fs_bsize / TP_BSIZE,
119			TP_NINDIR);
120	}
121	return (blkest + 1);
122}
123
124/* Auxiliary macro to pick up files changed since previous dump. */
125#define	CHANGEDSINCE(dp, t) \
126	(DIP(dp, di_mtime) >= (t) || DIP(dp, di_ctime) >= (t))
127
128/* The WANTTODUMP macro decides whether a file should be dumped. */
129#ifdef UF_NODUMP
130#define	WANTTODUMP(dp) \
131	(CHANGEDSINCE(dp, spcl.c_ddate) && \
132	 (nonodump || (DIP(dp, di_flags) & UF_NODUMP) != UF_NODUMP))
133#else
134#define	WANTTODUMP(dp) CHANGEDSINCE(dp, spcl.c_ddate)
135#endif
136
137/*
138 * Dump pass 1.
139 *
140 * Walk the inode list for a file system to find all allocated inodes
141 * that have been modified since the previous dump time. Also, find all
142 * the directories in the file system.
143 */
144int
145mapfiles(ino_t maxino, long *tapesize)
146{
147	int i, cg, mode, inosused;
148	int anydirskipped = 0;
149	union dinode *dp;
150	struct cg *cgp;
151	ino_t ino;
152	u_char *cp;
153
154	if ((cgp = malloc(sblock->fs_cgsize)) == NULL)
155		quit("mapfiles: cannot allocate memory.\n");
156	for (cg = 0; cg < sblock->fs_ncg; cg++) {
157		ino = cg * sblock->fs_ipg;
158		bread(fsbtodb(sblock, cgtod(sblock, cg)), (char *)cgp,
159		    sblock->fs_cgsize);
160		if (sblock->fs_magic == FS_UFS2_MAGIC)
161			inosused = cgp->cg_initediblk;
162		else
163			inosused = sblock->fs_ipg;
164		/*
165		 * If we are using soft updates, then we can trust the
166		 * cylinder group inode allocation maps to tell us which
167		 * inodes are allocated. We will scan the used inode map
168		 * to find the inodes that are really in use, and then
169		 * read only those inodes in from disk.
170		 */
171		if (sblock->fs_flags & FS_DOSOFTDEP) {
172			if (!cg_chkmagic(cgp))
173				quit("mapfiles: cg %d: bad magic number\n", cg);
174			cp = &cg_inosused(cgp)[(inosused - 1) / CHAR_BIT];
175			for ( ; inosused > 0; inosused -= CHAR_BIT, cp--) {
176				if (*cp == 0)
177					continue;
178				for (i = 1 << (CHAR_BIT - 1); i > 0; i >>= 1) {
179					if (*cp & i)
180						break;
181					inosused--;
182				}
183				break;
184			}
185			if (inosused <= 0)
186				continue;
187		}
188		for (i = 0; i < inosused; i++, ino++) {
189			if (ino < ROOTINO ||
190			    (dp = getino(ino, &mode)) == NULL ||
191			    (mode & IFMT) == 0)
192				continue;
193			if (ino >= maxino) {
194				msg("Skipping inode %d >= maxino %d\n",
195				    ino, maxino);
196				continue;
197			}
198			/*
199			 * Everything must go in usedinomap so that a check
200			 * for "in dumpdirmap but not in usedinomap" to detect
201			 * dirs with nodump set has a chance of succeeding
202			 * (this is used in mapdirs()).
203			 */
204			SETINO(ino, usedinomap);
205			if (mode == IFDIR)
206				SETINO(ino, dumpdirmap);
207			if (WANTTODUMP(dp)) {
208				SETINO(ino, dumpinomap);
209				if (mode != IFREG &&
210				    mode != IFDIR &&
211				    mode != IFLNK)
212					*tapesize += 1;
213				else
214					*tapesize += blockest(dp);
215				continue;
216			}
217			if (mode == IFDIR) {
218				if (!nonodump &&
219				    (DIP(dp, di_flags) & UF_NODUMP))
220					CLRINO(ino, usedinomap);
221				anydirskipped = 1;
222			}
223		}
224	}
225	/*
226	 * Restore gets very upset if the root is not dumped,
227	 * so ensure that it always is dumped.
228	 */
229	SETINO(ROOTINO, dumpinomap);
230	return (anydirskipped);
231}
232
233/*
234 * Dump pass 2.
235 *
236 * Scan each directory on the file system to see if it has any modified
237 * files in it. If it does, and has not already been added to the dump
238 * list (because it was itself modified), then add it. If a directory
239 * has not been modified itself, contains no modified files and has no
240 * subdirectories, then it can be deleted from the dump list and from
241 * the list of directories. By deleting it from the list of directories,
242 * its parent may now qualify for the same treatment on this or a later
243 * pass using this algorithm.
244 */
245int
246mapdirs(ino_t maxino, long *tapesize)
247{
248	union dinode *dp;
249	int i, isdir, nodump;
250	char *map;
251	ino_t ino;
252	union dinode di;
253	long filesize;
254	int ret, change = 0;
255
256	isdir = 0;		/* XXX just to get gcc to shut up */
257	for (map = dumpdirmap, ino = 1; ino < maxino; ino++) {
258		if (((ino - 1) % CHAR_BIT) == 0)	/* map is offset by 1 */
259			isdir = *map++;
260		else
261			isdir >>= 1;
262		/*
263		 * If a directory has been removed from usedinomap, it
264		 * either has the nodump flag set, or has inherited
265		 * it.  Although a directory can't be in dumpinomap if
266		 * it isn't in usedinomap, we have to go through it to
267		 * propagate the nodump flag.
268		 */
269		nodump = !nonodump && (TSTINO(ino, usedinomap) == 0);
270		if ((isdir & 1) == 0 || (TSTINO(ino, dumpinomap) && !nodump))
271			continue;
272		dp = getino(ino, &i);
273		/*
274		 * inode buf may change in searchdir().
275		 */
276		if (sblock->fs_magic == FS_UFS1_MAGIC)
277			di.dp1 = dp->dp1;
278		else
279			di.dp2 = dp->dp2;
280		filesize = DIP(&di, di_size);
281		for (ret = 0, i = 0; filesize > 0 && i < NDADDR; i++) {
282			if (DIP(&di, di_db[i]) != 0)
283				ret |= searchdir(ino, DIP(&di, di_db[i]),
284				    (long)sblksize(sblock, DIP(&di, di_size),
285				    i), filesize, tapesize, nodump, maxino);
286			if (ret & HASDUMPEDFILE)
287				filesize = 0;
288			else
289				filesize -= sblock->fs_bsize;
290		}
291		for (i = 0; filesize > 0 && i < NIADDR; i++) {
292			if (DIP(&di, di_ib[i]) == 0)
293				continue;
294			ret |= dirindir(ino, DIP(&di, di_ib[i]), i, &filesize,
295			    tapesize, nodump, maxino);
296		}
297		if (ret & HASDUMPEDFILE) {
298			SETINO(ino, dumpinomap);
299			*tapesize += blockest(&di);
300			change = 1;
301			continue;
302		}
303		if (nodump) {
304			if (ret & HASSUBDIRS)
305				change = 1;	/* subdirs inherit nodump */
306			CLRINO(ino, dumpdirmap);
307		} else if ((ret & HASSUBDIRS) == 0)
308			if (!TSTINO(ino, dumpinomap)) {
309				CLRINO(ino, dumpdirmap);
310				change = 1;
311			}
312	}
313	return (change);
314}
315
316/*
317 * Read indirect blocks, and pass the data blocks to be searched
318 * as directories. Quit as soon as any entry is found that will
319 * require the directory to be dumped.
320 */
321static int
322dirindir(
323	ino_t ino,
324	ufs2_daddr_t blkno,
325	int ind_level,
326	long *filesize,
327	long *tapesize,
328	int nodump,
329	ino_t maxino)
330{
331	union {
332		ufs1_daddr_t ufs1[MAXBSIZE / sizeof(ufs1_daddr_t)];
333		ufs2_daddr_t ufs2[MAXBSIZE / sizeof(ufs2_daddr_t)];
334	} idblk;
335	int ret = 0;
336	int i;
337
338	bread(fsbtodb(sblock, blkno), (char *)&idblk, (int)sblock->fs_bsize);
339	if (ind_level <= 0) {
340		for (i = 0; *filesize > 0 && i < NINDIR(sblock); i++) {
341			if (sblock->fs_magic == FS_UFS1_MAGIC)
342				blkno = idblk.ufs1[i];
343			else
344				blkno = idblk.ufs2[i];
345			if (blkno != 0)
346				ret |= searchdir(ino, blkno, sblock->fs_bsize,
347					*filesize, tapesize, nodump, maxino);
348			if (ret & HASDUMPEDFILE)
349				*filesize = 0;
350			else
351				*filesize -= sblock->fs_bsize;
352		}
353		return (ret);
354	}
355	ind_level--;
356	for (i = 0; *filesize > 0 && i < NINDIR(sblock); i++) {
357		if (sblock->fs_magic == FS_UFS1_MAGIC)
358			blkno = idblk.ufs1[i];
359		else
360			blkno = idblk.ufs2[i];
361		if (blkno != 0)
362			ret |= dirindir(ino, blkno, ind_level, filesize,
363			    tapesize, nodump, maxino);
364	}
365	return (ret);
366}
367
368/*
369 * Scan a disk block containing directory information looking to see if
370 * any of the entries are on the dump list and to see if the directory
371 * contains any subdirectories.
372 */
373static int
374searchdir(
375	ino_t ino,
376	ufs2_daddr_t blkno,
377	long size,
378	long filesize,
379	long *tapesize,
380	int nodump,
381	ino_t maxino)
382{
383	int mode;
384	struct direct *dp;
385	union dinode *ip;
386	long loc, ret = 0;
387	static caddr_t dblk;
388
389	if (dblk == NULL && (dblk = malloc(sblock->fs_bsize)) == NULL)
390		quit("searchdir: cannot allocate indirect memory.\n");
391	bread(fsbtodb(sblock, blkno), dblk, (int)size);
392	if (filesize < size)
393		size = filesize;
394	for (loc = 0; loc < size; ) {
395		dp = (struct direct *)(dblk + loc);
396		if (dp->d_reclen == 0) {
397			msg("corrupted directory, inumber %d\n", ino);
398			break;
399		}
400		loc += dp->d_reclen;
401		if (dp->d_ino == 0)
402			continue;
403		if (dp->d_ino >= maxino) {
404			msg("corrupted directory entry, d_ino %d >= %d\n",
405			    dp->d_ino, maxino);
406			break;
407		}
408		if (dp->d_name[0] == '.') {
409			if (dp->d_name[1] == '\0')
410				continue;
411			if (dp->d_name[1] == '.' && dp->d_name[2] == '\0')
412				continue;
413		}
414		if (nodump) {
415			ip = getino(dp->d_ino, &mode);
416			if (TSTINO(dp->d_ino, dumpinomap)) {
417				CLRINO(dp->d_ino, dumpinomap);
418				*tapesize -= blockest(ip);
419			}
420			/*
421			 * Add back to dumpdirmap and remove from usedinomap
422			 * to propagate nodump.
423			 */
424			if (mode == IFDIR) {
425				SETINO(dp->d_ino, dumpdirmap);
426				CLRINO(dp->d_ino, usedinomap);
427				ret |= HASSUBDIRS;
428			}
429		} else {
430			if (TSTINO(dp->d_ino, dumpinomap)) {
431				ret |= HASDUMPEDFILE;
432				if (ret & HASSUBDIRS)
433					break;
434			}
435			if (TSTINO(dp->d_ino, dumpdirmap)) {
436				ret |= HASSUBDIRS;
437				if (ret & HASDUMPEDFILE)
438					break;
439			}
440		}
441	}
442	return (ret);
443}
444
445/*
446 * Dump passes 3 and 4.
447 *
448 * Dump the contents of an inode to tape.
449 */
450void
451dumpino(union dinode *dp, ino_t ino)
452{
453	int ind_level, cnt;
454	off_t size;
455	char buf[TP_BSIZE];
456
457	if (newtape) {
458		newtape = 0;
459		dumpmap(dumpinomap, TS_BITS, ino);
460	}
461	CLRINO(ino, dumpinomap);
462	/*
463	 * Zero out the size of a snapshot so that it will be dumped
464	 * as a zero length file.
465	 */
466	if ((DIP(dp, di_flags) & SF_SNAPSHOT) != 0) {
467		DIP_SET(dp, di_size, 0);
468		DIP_SET(dp, di_flags, DIP(dp, di_flags) & ~SF_SNAPSHOT);
469	}
470	if (sblock->fs_magic == FS_UFS1_MAGIC) {
471		spcl.c_mode = dp->dp1.di_mode;
472		spcl.c_size = dp->dp1.di_size;
473		spcl.c_atime = _time32_to_time(dp->dp1.di_atime);
474		spcl.c_atimensec = dp->dp1.di_atimensec;
475		spcl.c_mtime = _time32_to_time(dp->dp1.di_mtime);
476		spcl.c_mtimensec = dp->dp1.di_mtimensec;
477		spcl.c_birthtime = 0;
478		spcl.c_birthtimensec = 0;
479		spcl.c_rdev = dp->dp1.di_rdev;
480		spcl.c_file_flags = dp->dp1.di_flags;
481		spcl.c_uid = dp->dp1.di_uid;
482		spcl.c_gid = dp->dp1.di_gid;
483	} else {
484		spcl.c_mode = dp->dp2.di_mode;
485		spcl.c_size = dp->dp2.di_size;
486		spcl.c_atime = _time64_to_time(dp->dp2.di_atime);
487		spcl.c_atimensec = dp->dp2.di_atimensec;
488		spcl.c_mtime = _time64_to_time(dp->dp2.di_mtime);
489		spcl.c_mtimensec = dp->dp2.di_mtimensec;
490		spcl.c_birthtime = _time64_to_time(dp->dp2.di_birthtime);
491		spcl.c_birthtimensec = dp->dp2.di_birthnsec;
492		spcl.c_rdev = dp->dp2.di_rdev;
493		spcl.c_file_flags = dp->dp2.di_flags;
494		spcl.c_uid = dp->dp2.di_uid;
495		spcl.c_gid = dp->dp2.di_gid;
496	}
497	spcl.c_type = TS_INODE;
498	spcl.c_count = 0;
499	switch (DIP(dp, di_mode) & S_IFMT) {
500
501	case 0:
502		/*
503		 * Freed inode.
504		 */
505		return;
506
507	case S_IFLNK:
508		/*
509		 * Check for short symbolic link.
510		 */
511		if (DIP(dp, di_size) > 0 &&
512		    DIP(dp, di_size) < sblock->fs_maxsymlinklen) {
513			spcl.c_addr[0] = 1;
514			spcl.c_count = 1;
515			writeheader(ino);
516			if (sblock->fs_magic == FS_UFS1_MAGIC)
517				memmove(buf, (caddr_t)dp->dp1.di_db,
518				    (u_long)DIP(dp, di_size));
519			else
520				memmove(buf, (caddr_t)dp->dp2.di_db,
521				    (u_long)DIP(dp, di_size));
522			buf[DIP(dp, di_size)] = '\0';
523			writerec(buf, 0);
524			return;
525		}
526		/* FALLTHROUGH */
527
528	case S_IFDIR:
529	case S_IFREG:
530		if (DIP(dp, di_size) > 0)
531			break;
532		/* FALLTHROUGH */
533
534	case S_IFIFO:
535	case S_IFSOCK:
536	case S_IFCHR:
537	case S_IFBLK:
538		writeheader(ino);
539		return;
540
541	default:
542		msg("Warning: undefined file type 0%o\n",
543		    DIP(dp, di_mode) & IFMT);
544		return;
545	}
546	if (DIP(dp, di_size) > NDADDR * sblock->fs_bsize)
547		cnt = NDADDR * sblock->fs_frag;
548	else
549		cnt = howmany(DIP(dp, di_size), sblock->fs_fsize);
550	if (sblock->fs_magic == FS_UFS1_MAGIC)
551		ufs1_blksout(&dp->dp1.di_db[0], cnt, ino);
552	else
553		ufs2_blksout(&dp->dp2.di_db[0], cnt, ino);
554	if ((size = DIP(dp, di_size) - NDADDR * sblock->fs_bsize) <= 0)
555		return;
556	for (ind_level = 0; ind_level < NIADDR; ind_level++) {
557		dmpindir(ino, DIP(dp, di_ib[ind_level]), ind_level, &size);
558		if (size <= 0)
559			return;
560	}
561}
562
563/*
564 * Read indirect blocks, and pass the data blocks to be dumped.
565 */
566static void
567dmpindir(ino_t ino, ufs2_daddr_t blk, int ind_level, off_t *size)
568{
569	union {
570		ufs1_daddr_t ufs1[MAXBSIZE / sizeof(ufs1_daddr_t)];
571		ufs2_daddr_t ufs2[MAXBSIZE / sizeof(ufs2_daddr_t)];
572	} idblk;
573	int i, cnt;
574
575	if (blk != 0)
576		bread(fsbtodb(sblock, blk), (char *)&idblk,
577		    (int)sblock->fs_bsize);
578	else
579		memset(&idblk, 0, sblock->fs_bsize);
580	if (ind_level <= 0) {
581		if (*size < NINDIR(sblock) * sblock->fs_bsize)
582			cnt = howmany(*size, sblock->fs_fsize);
583		else
584			cnt = NINDIR(sblock) * sblock->fs_frag;
585		*size -= NINDIR(sblock) * sblock->fs_bsize;
586		if (sblock->fs_magic == FS_UFS1_MAGIC)
587			ufs1_blksout(idblk.ufs1, cnt, ino);
588		else
589			ufs2_blksout(idblk.ufs2, cnt, ino);
590		return;
591	}
592	ind_level--;
593	for (i = 0; i < NINDIR(sblock); i++) {
594		if (sblock->fs_magic == FS_UFS1_MAGIC)
595			dmpindir(ino, idblk.ufs1[i], ind_level, size);
596		else
597			dmpindir(ino, idblk.ufs2[i], ind_level, size);
598		if (*size <= 0)
599			return;
600	}
601}
602
603/*
604 * Collect up the data into tape record sized buffers and output them.
605 */
606void
607ufs1_blksout(ufs1_daddr_t *blkp, int frags, ino_t ino)
608{
609	ufs1_daddr_t *bp;
610	int i, j, count, blks, tbperdb;
611
612	blks = howmany(frags * sblock->fs_fsize, TP_BSIZE);
613	tbperdb = sblock->fs_bsize >> tp_bshift;
614	for (i = 0; i < blks; i += TP_NINDIR) {
615		if (i + TP_NINDIR > blks)
616			count = blks;
617		else
618			count = i + TP_NINDIR;
619		for (j = i; j < count; j++)
620			if (blkp[j / tbperdb] != 0)
621				spcl.c_addr[j - i] = 1;
622			else
623				spcl.c_addr[j - i] = 0;
624		spcl.c_count = count - i;
625		writeheader(ino);
626		bp = &blkp[i / tbperdb];
627		for (j = i; j < count; j += tbperdb, bp++)
628			if (*bp != 0) {
629				if (j + tbperdb <= count)
630					dumpblock(*bp, (int)sblock->fs_bsize);
631				else
632					dumpblock(*bp, (count - j) * TP_BSIZE);
633			}
634		spcl.c_type = TS_ADDR;
635	}
636}
637
638/*
639 * Collect up the data into tape record sized buffers and output them.
640 */
641void
642ufs2_blksout(ufs2_daddr_t *blkp, int frags, ino_t ino)
643{
644	ufs2_daddr_t *bp;
645	int i, j, count, blks, tbperdb;
646
647	blks = howmany(frags * sblock->fs_fsize, TP_BSIZE);
648	tbperdb = sblock->fs_bsize >> tp_bshift;
649	for (i = 0; i < blks; i += TP_NINDIR) {
650		if (i + TP_NINDIR > blks)
651			count = blks;
652		else
653			count = i + TP_NINDIR;
654		for (j = i; j < count; j++)
655			if (blkp[j / tbperdb] != 0)
656				spcl.c_addr[j - i] = 1;
657			else
658				spcl.c_addr[j - i] = 0;
659		spcl.c_count = count - i;
660		writeheader(ino);
661		bp = &blkp[i / tbperdb];
662		for (j = i; j < count; j += tbperdb, bp++)
663			if (*bp != 0) {
664				if (j + tbperdb <= count)
665					dumpblock(*bp, (int)sblock->fs_bsize);
666				else
667					dumpblock(*bp, (count - j) * TP_BSIZE);
668			}
669		spcl.c_type = TS_ADDR;
670	}
671}
672
673/*
674 * Dump a map to the tape.
675 */
676void
677dumpmap(char *map, int type, ino_t ino)
678{
679	int i;
680	char *cp;
681
682	spcl.c_type = type;
683	spcl.c_count = howmany(mapsize * sizeof(char), TP_BSIZE);
684	writeheader(ino);
685	for (i = 0, cp = map; i < spcl.c_count; i++, cp += TP_BSIZE)
686		writerec(cp, 0);
687}
688
689/*
690 * Write a header record to the dump tape.
691 */
692void
693writeheader(ino_t ino)
694{
695	int32_t sum, cnt, *lp;
696
697	spcl.c_inumber = ino;
698	spcl.c_magic = FS_UFS2_MAGIC;
699	spcl.c_checksum = 0;
700	lp = (int32_t *)&spcl;
701	sum = 0;
702	cnt = sizeof(union u_spcl) / (4 * sizeof(int32_t));
703	while (--cnt >= 0) {
704		sum += *lp++;
705		sum += *lp++;
706		sum += *lp++;
707		sum += *lp++;
708	}
709	spcl.c_checksum = CHECKSUM - sum;
710	writerec((char *)&spcl, 1);
711}
712
713union dinode *
714getino(ino_t inum, int *modep)
715{
716	static ino_t minino, maxino;
717	static caddr_t inoblock;
718	struct ufs1_dinode *dp1;
719	struct ufs2_dinode *dp2;
720
721	if (inoblock == NULL && (inoblock = malloc(sblock->fs_bsize)) == NULL)
722		quit("cannot allocate inode memory.\n");
723	curino = inum;
724	if (inum >= minino && inum < maxino)
725		goto gotit;
726	bread(fsbtodb(sblock, ino_to_fsba(sblock, inum)), inoblock,
727	    (int)sblock->fs_bsize);
728	minino = inum - (inum % INOPB(sblock));
729	maxino = minino + INOPB(sblock);
730gotit:
731	if (sblock->fs_magic == FS_UFS1_MAGIC) {
732		dp1 = &((struct ufs1_dinode *)inoblock)[inum - minino];
733		*modep = (dp1->di_mode & IFMT);
734		return ((union dinode *)dp1);
735	}
736	dp2 = &((struct ufs2_dinode *)inoblock)[inum - minino];
737	*modep = (dp2->di_mode & IFMT);
738	return ((union dinode *)dp2);
739}
740
741/*
742 * Read a chunk of data from the disk.
743 * Try to recover from hard errors by reading in sector sized pieces.
744 * Error recovery is attempted at most BREADEMAX times before seeking
745 * consent from the operator to continue.
746 */
747int	breaderrors = 0;
748#define	BREADEMAX 32
749
750void
751bread(ufs2_daddr_t blkno, char *buf, int size)
752{
753	int secsize, bytes, resid, xfer, base, cnt, i;
754	static char *tmpbuf;
755	off_t offset;
756
757loop:
758	offset = blkno << dev_bshift;
759	secsize = sblock->fs_fsize;
760	base = offset % secsize;
761	resid = size % secsize;
762	/*
763	 * If the transfer request starts or ends on a non-sector
764	 * boundary, we must read the entire sector and copy out
765	 * just the part that we need.
766	 */
767	if (base == 0 && resid == 0) {
768		cnt = cread(diskfd, buf, size, offset);
769		if (cnt == size)
770			return;
771	} else {
772		if (tmpbuf == NULL && (tmpbuf = malloc(secsize)) == 0)
773			quit("buffer malloc failed\n");
774		xfer = 0;
775		bytes = size;
776		if (base != 0) {
777			cnt = cread(diskfd, tmpbuf, secsize, offset - base);
778			if (cnt != secsize)
779				goto bad;
780			xfer = secsize - base;
781			offset += xfer;
782			bytes -= xfer;
783			resid = bytes % secsize;
784			memcpy(buf, &tmpbuf[base], xfer);
785		}
786		if (bytes >= secsize) {
787			cnt = cread(diskfd, &buf[xfer], bytes - resid, offset);
788			if (cnt != bytes - resid)
789				goto bad;
790			xfer += cnt;
791			offset += cnt;
792		}
793		if (resid == 0)
794			return;
795		cnt = cread(diskfd, tmpbuf, secsize, offset);
796		if (cnt == secsize) {
797			memcpy(&buf[xfer], tmpbuf, resid);
798			return;
799		}
800	}
801bad:
802	if (blkno + (size / dev_bsize) > fsbtodb(sblock, sblock->fs_size)) {
803		/*
804		 * Trying to read the final fragment.
805		 *
806		 * NB - dump only works in TP_BSIZE blocks, hence
807		 * rounds `dev_bsize' fragments up to TP_BSIZE pieces.
808		 * It should be smarter about not actually trying to
809		 * read more than it can get, but for the time being
810		 * we punt and scale back the read only when it gets
811		 * us into trouble. (mkm 9/25/83)
812		 */
813		size -= dev_bsize;
814		goto loop;
815	}
816	if (cnt == -1)
817		msg("read error from %s: %s: [block %jd]: count=%d\n",
818			disk, strerror(errno), (intmax_t)blkno, size);
819	else
820		msg("short read error from %s: [block %jd]: count=%d, got=%d\n",
821			disk, (intmax_t)blkno, size, cnt);
822	if (++breaderrors > BREADEMAX) {
823		msg("More than %d block read errors from %s\n",
824			BREADEMAX, disk);
825		broadcast("DUMP IS AILING!\n");
826		msg("This is an unrecoverable error.\n");
827		if (!query("Do you want to attempt to continue?")){
828			dumpabort(0);
829			/*NOTREACHED*/
830		} else
831			breaderrors = 0;
832	}
833	/*
834	 * Zero buffer, then try to read each sector of buffer separately,
835	 * and bypass the cache.
836	 */
837	memset(buf, 0, size);
838	for (i = 0; i < size; i += dev_bsize, buf += dev_bsize, blkno++) {
839		if ((cnt = pread(diskfd, buf, (int)dev_bsize,
840		    ((off_t)blkno << dev_bshift))) == dev_bsize)
841			continue;
842		if (cnt == -1) {
843			msg("read error from %s: %s: [sector %jd]: count=%ld\n",
844			    disk, strerror(errno), (intmax_t)blkno, dev_bsize);
845			continue;
846		}
847		msg("short read from %s: [sector %jd]: count=%ld, got=%d\n",
848		    disk, (intmax_t)blkno, dev_bsize, cnt);
849	}
850}
851