pass1.c revision 201708 
1/*
2 * Copyright (c) 1980, 1986, 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#if 0
31#ifndef lint
32static const char sccsid[] = "@(#)pass1.c	8.6 (Berkeley) 4/28/95";
33#endif /* not lint */
34#endif
35#include <sys/cdefs.h>
36__FBSDID("$FreeBSD: head/sbin/fsck_ffs/pass1.c 201708 2010-01-07 01:10:49Z mckusick $");
37
38#include <sys/param.h>
39#include <sys/stat.h>
40#include <sys/sysctl.h>
41
42#include <ufs/ufs/dinode.h>
43#include <ufs/ufs/dir.h>
44#include <ufs/ffs/fs.h>
45
46#include <err.h>
47#include <limits.h>
48#include <stdint.h>
49#include <string.h>
50
51#include "fsck.h"
52
53static ufs2_daddr_t badblk;
54static ufs2_daddr_t dupblk;
55static ino_t lastino;		/* last inode in use */
56
57static int checkinode(ino_t inumber, struct inodesc *, int rebuildcg);
58
59void
60pass1(void)
61{
62	struct inostat *info;
63	struct inodesc idesc;
64	ino_t inumber, inosused, mininos;
65	ufs2_daddr_t i, cgd;
66	u_int8_t *cp;
67	int c, rebuildcg;
68
69	/*
70	 * Set file system reserved blocks in used block map.
71	 */
72	for (c = 0; c < sblock.fs_ncg; c++) {
73		cgd = cgdmin(&sblock, c);
74		if (c == 0) {
75			i = cgbase(&sblock, c);
76		} else
77			i = cgsblock(&sblock, c);
78		for (; i < cgd; i++)
79			setbmap(i);
80	}
81	i = sblock.fs_csaddr;
82	cgd = i + howmany(sblock.fs_cssize, sblock.fs_fsize);
83	for (; i < cgd; i++)
84		setbmap(i);
85
86	/*
87	 * Find all allocated blocks.
88	 */
89	memset(&idesc, 0, sizeof(struct inodesc));
90	idesc.id_func = pass1check;
91	n_files = n_blks = 0;
92	for (c = 0; c < sblock.fs_ncg; c++) {
93		inumber = c * sblock.fs_ipg;
94		setinodebuf(inumber);
95		getblk(&cgblk, cgtod(&sblock, c), sblock.fs_cgsize);
96		rebuildcg = 0;
97		if (!check_cgmagic(c, &cgrp))
98			rebuildcg = 1;
99		if (!rebuildcg && sblock.fs_magic == FS_UFS2_MAGIC) {
100			inosused = cgrp.cg_initediblk;
101			if (inosused > sblock.fs_ipg) {
102				pfatal("%s (%d > %d) %s %d\nReset to %d\n",
103				    "Too many initialized inodes", inosused,
104				    sblock.fs_ipg, "in cylinder group", c,
105				    sblock.fs_ipg);
106				inosused = sblock.fs_ipg;
107			}
108		} else {
109			inosused = sblock.fs_ipg;
110		}
111		if (got_siginfo) {
112			printf("%s: phase 1: cyl group %d of %d (%d%%)\n",
113			    cdevname, c, sblock.fs_ncg,
114			    c * 100 / sblock.fs_ncg);
115			got_siginfo = 0;
116		}
117		if (got_sigalarm) {
118			setproctitle("%s p1 %d%%", cdevname,
119			     c * 100 / sblock.fs_ncg);
120			got_sigalarm = 0;
121		}
122		/*
123		 * If we are using soft updates, then we can trust the
124		 * cylinder group inode allocation maps to tell us which
125		 * inodes are allocated. We will scan the used inode map
126		 * to find the inodes that are really in use, and then
127		 * read only those inodes in from disk.
128		 */
129		if ((preen || inoopt) && usedsoftdep && !rebuildcg) {
130			cp = &cg_inosused(&cgrp)[(inosused - 1) / CHAR_BIT];
131			for ( ; inosused > 0; inosused -= CHAR_BIT, cp--) {
132				if (*cp == 0)
133					continue;
134				for (i = 1 << (CHAR_BIT - 1); i > 0; i >>= 1) {
135					if (*cp & i)
136						break;
137					inosused--;
138				}
139				break;
140			}
141			if (inosused < 0)
142				inosused = 0;
143		}
144		/*
145		 * Allocate inoinfo structures for the allocated inodes.
146		 */
147		inostathead[c].il_numalloced = inosused;
148		if (inosused == 0) {
149			inostathead[c].il_stat = 0;
150			continue;
151		}
152		info = calloc((unsigned)inosused, sizeof(struct inostat));
153		if (info == NULL)
154			errx(EEXIT, "cannot alloc %u bytes for inoinfo",
155			    (unsigned)(sizeof(struct inostat) * inosused));
156		inostathead[c].il_stat = info;
157		/*
158		 * Scan the allocated inodes.
159		 */
160		for (i = 0; i < inosused; i++, inumber++) {
161			if (inumber < ROOTINO) {
162				(void)getnextinode(inumber, rebuildcg);
163				continue;
164			}
165			/*
166			 * NULL return indicates probable end of allocated
167			 * inodes during cylinder group rebuild attempt.
168			 * We always keep trying until we get to the minimum
169			 * valid number for this cylinder group.
170			 */
171			if (checkinode(inumber, &idesc, rebuildcg) == 0 &&
172			    i > cgrp.cg_initediblk)
173				break;
174		}
175		/*
176		 * This optimization speeds up future runs of fsck
177		 * by trimming down the number of inodes in cylinder
178		 * groups that formerly had many inodes but now have
179		 * fewer in use.
180		 */
181		mininos = roundup(inosused + INOPB(&sblock), INOPB(&sblock));
182		if (inoopt && !preen && !rebuildcg &&
183		    sblock.fs_magic == FS_UFS2_MAGIC &&
184		    cgrp.cg_initediblk > 2 * INOPB(&sblock) &&
185		    mininos < cgrp.cg_initediblk) {
186			i = cgrp.cg_initediblk;
187			if (mininos < 2 * INOPB(&sblock))
188				cgrp.cg_initediblk = 2 * INOPB(&sblock);
189			else
190				cgrp.cg_initediblk = mininos;
191			pwarn("CYLINDER GROUP %d: RESET FROM %ju TO %d %s\n",
192			    c, i, cgrp.cg_initediblk, "VALID INODES");
193			cgdirty();
194		}
195		if (inosused < sblock.fs_ipg)
196			continue;
197		lastino += 1;
198		if (lastino < (c * sblock.fs_ipg))
199			inosused = 0;
200		else
201			inosused = lastino - (c * sblock.fs_ipg);
202		if (rebuildcg && inosused > cgrp.cg_initediblk &&
203		    sblock.fs_magic == FS_UFS2_MAGIC) {
204			cgrp.cg_initediblk = roundup(inosused, INOPB(&sblock));
205			pwarn("CYLINDER GROUP %d: FOUND %d VALID INODES\n", c,
206			    cgrp.cg_initediblk);
207		}
208		/*
209		 * If we were not able to determine in advance which inodes
210		 * were in use, then reduce the size of the inoinfo structure
211		 * to the size necessary to describe the inodes that we
212		 * really found.
213		 */
214		if (inumber == lastino)
215			continue;
216		inostathead[c].il_numalloced = inosused;
217		if (inosused == 0) {
218			free(inostathead[c].il_stat);
219			inostathead[c].il_stat = 0;
220			continue;
221		}
222		info = calloc((unsigned)inosused, sizeof(struct inostat));
223		if (info == NULL)
224			errx(EEXIT, "cannot alloc %u bytes for inoinfo",
225			    (unsigned)(sizeof(struct inostat) * inosused));
226		memmove(info, inostathead[c].il_stat, inosused * sizeof(*info));
227		free(inostathead[c].il_stat);
228		inostathead[c].il_stat = info;
229	}
230	freeinodebuf();
231}
232
233static int
234checkinode(ino_t inumber, struct inodesc *idesc, int rebuildcg)
235{
236	union dinode *dp;
237	off_t kernmaxfilesize;
238	ufs2_daddr_t ndb;
239	mode_t mode;
240	int j, ret, offset;
241
242	if ((dp = getnextinode(inumber, rebuildcg)) == NULL)
243		return (0);
244	mode = DIP(dp, di_mode) & IFMT;
245	if (mode == 0) {
246		if ((sblock.fs_magic == FS_UFS1_MAGIC &&
247		     (memcmp(dp->dp1.di_db, ufs1_zino.di_db,
248			NDADDR * sizeof(ufs1_daddr_t)) ||
249		      memcmp(dp->dp1.di_ib, ufs1_zino.di_ib,
250			NIADDR * sizeof(ufs1_daddr_t)) ||
251		      dp->dp1.di_mode || dp->dp1.di_size)) ||
252		    (sblock.fs_magic == FS_UFS2_MAGIC &&
253		     (memcmp(dp->dp2.di_db, ufs2_zino.di_db,
254			NDADDR * sizeof(ufs2_daddr_t)) ||
255		      memcmp(dp->dp2.di_ib, ufs2_zino.di_ib,
256			NIADDR * sizeof(ufs2_daddr_t)) ||
257		      dp->dp2.di_mode || dp->dp2.di_size))) {
258			pfatal("PARTIALLY ALLOCATED INODE I=%lu",
259			    (u_long)inumber);
260			if (reply("CLEAR") == 1) {
261				dp = ginode(inumber);
262				clearinode(dp);
263				inodirty();
264			}
265		}
266		inoinfo(inumber)->ino_state = USTATE;
267		return (1);
268	}
269	lastino = inumber;
270	/* This should match the file size limit in ffs_mountfs(). */
271	if (sblock.fs_magic == FS_UFS1_MAGIC)
272		kernmaxfilesize = (off_t)0x40000000 * sblock.fs_bsize - 1;
273	else
274		kernmaxfilesize = sblock.fs_maxfilesize;
275	if (DIP(dp, di_size) > kernmaxfilesize ||
276	    DIP(dp, di_size) > sblock.fs_maxfilesize ||
277	    (mode == IFDIR && DIP(dp, di_size) > MAXDIRSIZE)) {
278		if (debug)
279			printf("bad size %ju:", (uintmax_t)DIP(dp, di_size));
280		goto unknown;
281	}
282	if (!preen && mode == IFMT && reply("HOLD BAD BLOCK") == 1) {
283		dp = ginode(inumber);
284		DIP_SET(dp, di_size, sblock.fs_fsize);
285		DIP_SET(dp, di_mode, IFREG|0600);
286		inodirty();
287	}
288	if ((mode == IFBLK || mode == IFCHR || mode == IFIFO ||
289	     mode == IFSOCK) && DIP(dp, di_size) != 0) {
290		if (debug)
291			printf("bad special-file size %ju:",
292			    (uintmax_t)DIP(dp, di_size));
293		goto unknown;
294	}
295	if ((mode == IFBLK || mode == IFCHR) &&
296	    (dev_t)DIP(dp, di_rdev) == NODEV) {
297		if (debug)
298			printf("bad special-file rdev NODEV:");
299		goto unknown;
300	}
301	ndb = howmany(DIP(dp, di_size), sblock.fs_bsize);
302	if (ndb < 0) {
303		if (debug)
304			printf("bad size %ju ndb %ju:",
305				(uintmax_t)DIP(dp, di_size), (uintmax_t)ndb);
306		goto unknown;
307	}
308	if (mode == IFBLK || mode == IFCHR)
309		ndb++;
310	if (mode == IFLNK) {
311		/*
312		 * Fake ndb value so direct/indirect block checks below
313		 * will detect any garbage after symlink string.
314		 */
315		if (DIP(dp, di_size) < (off_t)sblock.fs_maxsymlinklen) {
316			if (sblock.fs_magic == FS_UFS1_MAGIC)
317				ndb = howmany(DIP(dp, di_size),
318				    sizeof(ufs1_daddr_t));
319			else
320				ndb = howmany(DIP(dp, di_size),
321				    sizeof(ufs2_daddr_t));
322			if (ndb > NDADDR) {
323				j = ndb - NDADDR;
324				for (ndb = 1; j > 1; j--)
325					ndb *= NINDIR(&sblock);
326				ndb += NDADDR;
327			}
328		}
329	}
330	for (j = ndb; ndb < NDADDR && j < NDADDR; j++)
331		if (DIP(dp, di_db[j]) != 0) {
332			if (debug)
333				printf("bad direct addr[%d]: %ju\n", j,
334				    (uintmax_t)DIP(dp, di_db[j]));
335			goto unknown;
336		}
337	for (j = 0, ndb -= NDADDR; ndb > 0; j++)
338		ndb /= NINDIR(&sblock);
339	for (; j < NIADDR; j++)
340		if (DIP(dp, di_ib[j]) != 0) {
341			if (debug)
342				printf("bad indirect addr: %ju\n",
343				    (uintmax_t)DIP(dp, di_ib[j]));
344			goto unknown;
345		}
346	if (ftypeok(dp) == 0)
347		goto unknown;
348	n_files++;
349	inoinfo(inumber)->ino_linkcnt = DIP(dp, di_nlink);
350	if (mode == IFDIR) {
351		if (DIP(dp, di_size) == 0)
352			inoinfo(inumber)->ino_state = DCLEAR;
353		else if (DIP(dp, di_nlink) <= 0)
354			inoinfo(inumber)->ino_state = DZLINK;
355		else
356			inoinfo(inumber)->ino_state = DSTATE;
357		cacheino(dp, inumber);
358		countdirs++;
359	} else if (DIP(dp, di_nlink) <= 0)
360		inoinfo(inumber)->ino_state = FZLINK;
361	else
362		inoinfo(inumber)->ino_state = FSTATE;
363	inoinfo(inumber)->ino_type = IFTODT(mode);
364	badblk = dupblk = 0;
365	idesc->id_number = inumber;
366	if (DIP(dp, di_flags) & SF_SNAPSHOT)
367		idesc->id_type = SNAP;
368	else
369		idesc->id_type = ADDR;
370	(void)ckinode(dp, idesc);
371	if (sblock.fs_magic == FS_UFS2_MAGIC && dp->dp2.di_extsize > 0) {
372		idesc->id_type = ADDR;
373		ndb = howmany(dp->dp2.di_extsize, sblock.fs_bsize);
374		for (j = 0; j < NXADDR; j++) {
375			if (--ndb == 0 &&
376			    (offset = blkoff(&sblock, dp->dp2.di_extsize)) != 0)
377				idesc->id_numfrags = numfrags(&sblock,
378				    fragroundup(&sblock, offset));
379			else
380				idesc->id_numfrags = sblock.fs_frag;
381			if (dp->dp2.di_extb[j] == 0)
382				continue;
383			idesc->id_blkno = dp->dp2.di_extb[j];
384			ret = (*idesc->id_func)(idesc);
385			if (ret & STOP)
386				break;
387		}
388	}
389	if (sblock.fs_magic == FS_UFS2_MAGIC)
390		eascan(idesc, &dp->dp2);
391	idesc->id_entryno *= btodb(sblock.fs_fsize);
392	if (DIP(dp, di_blocks) != idesc->id_entryno) {
393		pwarn("INCORRECT BLOCK COUNT I=%lu (%ju should be %ju)",
394		    (u_long)inumber, (uintmax_t)DIP(dp, di_blocks),
395		    (uintmax_t)idesc->id_entryno);
396		if (preen)
397			printf(" (CORRECTED)\n");
398		else if (reply("CORRECT") == 0)
399			return (1);
400		if (bkgrdflag == 0) {
401			dp = ginode(inumber);
402			DIP_SET(dp, di_blocks, idesc->id_entryno);
403			inodirty();
404		} else {
405			cmd.value = idesc->id_number;
406			cmd.size = idesc->id_entryno - DIP(dp, di_blocks);
407			if (debug)
408				printf("adjblkcnt ino %ju amount %lld\n",
409				    (uintmax_t)cmd.value, (long long)cmd.size);
410			if (sysctl(adjblkcnt, MIBSIZE, 0, 0,
411			    &cmd, sizeof cmd) == -1)
412				rwerror("ADJUST INODE BLOCK COUNT", cmd.value);
413		}
414	}
415	return (1);
416unknown:
417	pfatal("UNKNOWN FILE TYPE I=%lu", (u_long)inumber);
418	inoinfo(inumber)->ino_state = FCLEAR;
419	if (reply("CLEAR") == 1) {
420		inoinfo(inumber)->ino_state = USTATE;
421		dp = ginode(inumber);
422		clearinode(dp);
423		inodirty();
424	}
425	return (1);
426}
427
428int
429pass1check(struct inodesc *idesc)
430{
431	int res = KEEPON;
432	int anyout, nfrags;
433	ufs2_daddr_t blkno = idesc->id_blkno;
434	struct dups *dlp;
435	struct dups *new;
436
437	if (idesc->id_type == SNAP) {
438		if (blkno == BLK_NOCOPY)
439			return (KEEPON);
440		if (idesc->id_number == cursnapshot) {
441			if (blkno == blkstofrags(&sblock, idesc->id_lbn))
442				return (KEEPON);
443			if (blkno == BLK_SNAP) {
444				blkno = blkstofrags(&sblock, idesc->id_lbn);
445				idesc->id_entryno -= idesc->id_numfrags;
446			}
447		} else {
448			if (blkno == BLK_SNAP)
449				return (KEEPON);
450		}
451	}
452	if ((anyout = chkrange(blkno, idesc->id_numfrags)) != 0) {
453		blkerror(idesc->id_number, "BAD", blkno);
454		if (badblk++ >= MAXBAD) {
455			pwarn("EXCESSIVE BAD BLKS I=%lu",
456			    (u_long)idesc->id_number);
457			if (preen)
458				printf(" (SKIPPING)\n");
459			else if (reply("CONTINUE") == 0) {
460				ckfini(0);
461				exit(EEXIT);
462			}
463			return (STOP);
464		}
465	}
466	for (nfrags = idesc->id_numfrags; nfrags > 0; blkno++, nfrags--) {
467		if (anyout && chkrange(blkno, 1)) {
468			res = SKIP;
469		} else if (!testbmap(blkno)) {
470			n_blks++;
471			setbmap(blkno);
472		} else {
473			blkerror(idesc->id_number, "DUP", blkno);
474			if (dupblk++ >= MAXDUP) {
475				pwarn("EXCESSIVE DUP BLKS I=%lu",
476					(u_long)idesc->id_number);
477				if (preen)
478					printf(" (SKIPPING)\n");
479				else if (reply("CONTINUE") == 0) {
480					ckfini(0);
481					exit(EEXIT);
482				}
483				return (STOP);
484			}
485			new = (struct dups *)malloc(sizeof(struct dups));
486			if (new == NULL) {
487				pfatal("DUP TABLE OVERFLOW.");
488				if (reply("CONTINUE") == 0) {
489					ckfini(0);
490					exit(EEXIT);
491				}
492				return (STOP);
493			}
494			new->dup = blkno;
495			if (muldup == 0) {
496				duplist = muldup = new;
497				new->next = 0;
498			} else {
499				new->next = muldup->next;
500				muldup->next = new;
501			}
502			for (dlp = duplist; dlp != muldup; dlp = dlp->next)
503				if (dlp->dup == blkno)
504					break;
505			if (dlp == muldup && dlp->dup != blkno)
506				muldup = new;
507		}
508		/*
509		 * count the number of blocks found in id_entryno
510		 */
511		idesc->id_entryno++;
512	}
513	return (res);
514}
515