xref: /illumos-gate/usr/src/cmd/fs.d/ufs/fsck/utilities.c (revision bfbf29e2)

/*
 * Copyright (c) 1990, 2010, Oracle and/or its affiliates. All rights reserved.
 * Copyright (c) 2016 by Delphix. All rights reserved.
 */

/*	Copyright (c) 1983, 1984, 1985, 1986, 1987, 1988, 1989 AT&T	*/
/*	  All Rights Reserved	*/

/*
 * Copyright (c) 1980, 1986, 1990 The Regents of the University of California.
 * All rights reserved.
 *
 * Redistribution and use in source and binary forms are permitted
 * provided that: (1) source distributions retain this entire copyright
 * notice and comment, and (2) distributions including binaries display
 * the following acknowledgement:  ``This product includes software
 * developed by the University of California, Berkeley and its contributors''
 * in the documentation or other materials provided with the distribution
 * and in all advertising materials mentioning features or use of this
 * software. Neither the name of the University nor the names of its
 * contributors may be used to endorse or promote products derived
 * from this software without specific prior written permission.
 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR
 * IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED
 * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.
 */

#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <stdarg.h>
#include <libadm.h>
#include <note.h>
#include <sys/param.h>
#include <sys/types.h>
#include <sys/mntent.h>
#include <sys/filio.h>
#include <sys/fs/ufs_fs.h>
#include <sys/vnode.h>
#include <sys/fs/ufs_acl.h>
#include <sys/fs/ufs_inode.h>
#include <sys/fs/ufs_log.h>
#define	_KERNEL
#include <sys/fs/ufs_fsdir.h>
#undef _KERNEL
#include <sys/mnttab.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <signal.h>
#include <string.h>
#include <ctype.h>
#include <sys/vfstab.h>
#include <sys/lockfs.h>
#include <errno.h>
#include <sys/cmn_err.h>
#include <sys/dkio.h>
#include <sys/vtoc.h>
#include <sys/efi_partition.h>
#include <fslib.h>
#include <inttypes.h>
#include "fsck.h"

struct bufarea *pbp;
struct bufarea *pdirbp;
caddr_t mount_point = NULL;
static struct bufarea bufhead;	/* head of list of other blks in filesys */
char *elock_combuf;
char *elock_mountp;
static struct lockfs *lfp;		/* current lockfs status */

static int64_t diskreads, totalreads;	/* Disk cache statistics */

static int log_checksum(int32_t *, int32_t *, int);
static void vdirerror(fsck_ino_t, caddr_t, va_list);
static struct mnttab *search_mnttab(caddr_t, caddr_t, caddr_t, size_t);
static struct vfstab *search_vfstab(caddr_t, caddr_t, caddr_t, size_t);
static void vpwarn(caddr_t, va_list);
static int getaline(FILE *, caddr_t, int);
static struct bufarea *alloc_bufarea(void);
static void rwerror(caddr_t, diskaddr_t, int rval);
static void debugclean(void);
static void report_io_prob(caddr_t, diskaddr_t, size_t, ssize_t);
static void freelogblk(daddr32_t);
static void verrexit(caddr_t, va_list);
static void vpfatal(caddr_t, va_list);
static diskaddr_t get_device_size(int, caddr_t);
static diskaddr_t brute_force_get_device_size(int);
static void cg_constants(int, daddr32_t *, daddr32_t *, daddr32_t *,
	    daddr32_t *, daddr32_t *, daddr32_t *);

int
ftypeok(struct dinode *dp)
{
	switch (dp->di_mode & IFMT) {

	case IFDIR:
	case IFREG:
	case IFBLK:
	case IFCHR:
	case IFLNK:
	case IFSOCK:
	case IFIFO:
	case IFSHAD:
	case IFATTRDIR:
		return (1);

	default:
		if (debug)
			(void) printf("bad file type 0%o\n", dp->di_mode);
		return (0);
	}
}

int
acltypeok(struct dinode *dp)
{
	if (CHECK_ACL_ALLOWED(dp->di_mode & IFMT))
		return (1);

	if (debug)
		(void) printf("bad file type for acl I=%d: 0%o\n",
		    dp->di_shadow, dp->di_mode);
	return (0);
}

NOTE(PRINTFLIKE(1))
int
reply(caddr_t fmt, ...)
{
	va_list ap;
	char line[80];

	if (preen)
		pfatal("INTERNAL ERROR: GOT TO reply() in preen mode");

	if (mflag) {
		/*
		 * We don't know what's going on, so don't potentially
		 * make things worse by having errexit() write stuff
		 * out to disk.
		 */
		(void) printf(
		    "\n%s: UNEXPECTED INCONSISTENCY; RUN fsck MANUALLY.\n",
		    devname);
		exit(EXERRFATAL);
	}

	va_start(ap, fmt);
	(void) putchar('\n');
	(void) vprintf(fmt, ap);
	(void) putchar('?');
	(void) putchar(' ');
	va_end(ap);

	if (nflag || fswritefd < 0) {
		(void) printf(" no\n\n");
		return (0);
	}
	if (yflag) {
		(void) printf(" yes\n\n");
		return (1);
	}
	(void) fflush(stdout);
	if (getaline(stdin, line, sizeof (line)) == EOF)
		errexit("\n");
	(void) printf("\n");
	if (line[0] == 'y' || line[0] == 'Y') {
		return (1);
	} else {
		return (0);
	}
}

int
getaline(FILE *fp, caddr_t loc, int maxlen)
{
	int n;
	caddr_t p, lastloc;

	p = loc;
	lastloc = &p[maxlen-1];
	while ((n = getc(fp)) != '\n') {
		if (n == EOF)
			return (EOF);
		if (!isspace(n) && p < lastloc)
			*p++ = (char)n;
	}
	*p = '\0';
	/* LINTED pointer difference won't overflow */
	return (p - loc);
}

/*
 * Malloc buffers and set up cache.
 */
void
bufinit(void)
{
	struct bufarea *bp;
	int bufcnt, i;
	caddr_t bufp;

	bufp = malloc((size_t)sblock.fs_bsize);
	if (bufp == NULL)
		goto nomem;
	initbarea(&cgblk);
	cgblk.b_un.b_buf = bufp;
	bufhead.b_next = bufhead.b_prev = &bufhead;
	bufcnt = MAXBUFSPACE / sblock.fs_bsize;
	if (bufcnt < MINBUFS)
		bufcnt = MINBUFS;
	for (i = 0; i < bufcnt; i++) {
		bp = (struct bufarea *)malloc(sizeof (struct bufarea));
		if (bp == NULL) {
			if (i >= MINBUFS)
				goto noalloc;
			goto nomem;
		}

		bufp = malloc((size_t)sblock.fs_bsize);
		if (bufp == NULL) {
			free((void *)bp);
			if (i >= MINBUFS)
				goto noalloc;
			goto nomem;
		}
		initbarea(bp);
		bp->b_un.b_buf = bufp;
		bp->b_prev = &bufhead;
		bp->b_next = bufhead.b_next;
		bufhead.b_next->b_prev = bp;
		bufhead.b_next = bp;
	}
noalloc:
	bufhead.b_size = i;	/* save number of buffers */
	pbp = pdirbp = NULL;
	return;

nomem:
	errexit("cannot allocate buffer pool\n");
	/* NOTREACHED */
}

/*
 * Undo a bufinit().
 */
void
unbufinit(void)
{
	int cnt;
	struct bufarea *bp, *nbp;

	cnt = 0;
	for (bp = bufhead.b_prev; bp != NULL && bp != &bufhead; bp = nbp) {
		cnt++;
		flush(fswritefd, bp);
		nbp = bp->b_prev;
		/*
		 * We're discarding the entire chain, so this isn't
		 * technically necessary.  However, it doesn't hurt
		 * and lint's data flow analysis is much happier
		 * (this prevents it from thinking there's a chance
		 * of our using memory elsewhere after it's been released).
		 */
		nbp->b_next = bp->b_next;
		bp->b_next->b_prev = nbp;
		free((void *)bp->b_un.b_buf);
		free((void *)bp);
	}

	if (bufhead.b_size != cnt)
		errexit("Panic: cache lost %d buffers\n",
		    bufhead.b_size - cnt);
}

/*
 * Manage a cache of directory blocks.
 */
struct bufarea *
getdatablk(daddr32_t blkno, size_t size)
{
	struct bufarea *bp;

	for (bp = bufhead.b_next; bp != &bufhead; bp = bp->b_next)
		if (bp->b_bno == fsbtodb(&sblock, blkno)) {
			goto foundit;
		}
	for (bp = bufhead.b_prev; bp != &bufhead; bp = bp->b_prev)
		if ((bp->b_flags & B_INUSE) == 0)
			break;
	if (bp == &bufhead) {
		bp = alloc_bufarea();
		if (bp == NULL) {
			errexit("deadlocked buffer pool\n");
			/* NOTREACHED */
		}
	}
	/*
	 * We're at the same logical level as getblk(), so if there
	 * are any errors, we'll let our caller handle them.
	 */
	diskreads++;
	(void) getblk(bp, blkno, size);

foundit:
	totalreads++;
	bp->b_cnt++;
	/*
	 * Move the buffer to head of linked list if it isn't
	 * already there.
	 */
	if (bufhead.b_next != bp) {
		bp->b_prev->b_next = bp->b_next;
		bp->b_next->b_prev = bp->b_prev;
		bp->b_prev = &bufhead;
		bp->b_next = bufhead.b_next;
		bufhead.b_next->b_prev = bp;
		bufhead.b_next = bp;
	}
	bp->b_flags |= B_INUSE;
	return (bp);
}

void
brelse(struct bufarea *bp)
{
	bp->b_cnt--;
	if (bp->b_cnt == 0) {
		bp->b_flags &= ~B_INUSE;
	}
}

struct bufarea *
getblk(struct bufarea *bp, daddr32_t blk, size_t size)
{
	diskaddr_t dblk;

	dblk = fsbtodb(&sblock, blk);
	if (bp->b_bno == dblk)
		return (bp);
	flush(fswritefd, bp);
	bp->b_errs = fsck_bread(fsreadfd, bp->b_un.b_buf, dblk, size);
	bp->b_bno = dblk;
	bp->b_size = size;
	return (bp);
}

void
flush(int fd, struct bufarea *bp)
{
	int i, j;
	caddr_t sip;
	long size;

	if (!bp->b_dirty)
		return;

	/*
	 * It's not our buf, so if there are errors, let whoever
	 * acquired it deal with the actual problem.
	 */
	if (bp->b_errs != 0)
		pfatal("WRITING ZERO'ED BLOCK %lld TO DISK\n", bp->b_bno);
	bp->b_dirty = 0;
	bp->b_errs = 0;
	bwrite(fd, bp->b_un.b_buf, bp->b_bno, (long)bp->b_size);
	if (bp != &sblk) {
		return;
	}

	/*
	 * We're flushing the superblock, so make sure all the
	 * ancillary bits go out as well.
	 */
	sip = (caddr_t)sblock.fs_u.fs_csp;
	for (i = 0, j = 0; i < sblock.fs_cssize; i += sblock.fs_bsize, j++) {
		size = sblock.fs_cssize - i < sblock.fs_bsize ?
		    sblock.fs_cssize - i : sblock.fs_bsize;
		bwrite(fswritefd, sip,
		    fsbtodb(&sblock, sblock.fs_csaddr + j * sblock.fs_frag),
		    size);
		sip += size;
	}
}

static void
rwerror(caddr_t mesg, diskaddr_t blk, int rval)
{
	int olderr = errno;

	if (!preen)
		(void) printf("\n");

	if (rval == -1)
		pfatal("CANNOT %s: DISK BLOCK %lld: %s",
		    mesg, blk, strerror(olderr));
	else
		pfatal("CANNOT %s: DISK BLOCK %lld", mesg, blk);

	if (reply("CONTINUE") == 0) {
		exitstat = EXERRFATAL;
		errexit("Program terminated\n");
	}
}

void
ckfini(void)
{
	int64_t percentage;

	if (fswritefd < 0)
		return;

	flush(fswritefd, &sblk);
	/*
	 * Were we using a backup superblock?
	 */
	if (havesb && sblk.b_bno != SBOFF / dev_bsize) {
		if (preen || reply("UPDATE STANDARD SUPERBLOCK") == 1) {
			sblk.b_bno = SBOFF / dev_bsize;
			sbdirty();
			flush(fswritefd, &sblk);
		}
	}
	flush(fswritefd, &cgblk);
	if (cgblk.b_un.b_buf != NULL) {
		free((void *)cgblk.b_un.b_buf);
		cgblk.b_un.b_buf = NULL;
	}
	unbufinit();
	pbp = NULL;
	pdirbp = NULL;
	if (debug) {
		/*
		 * Note that we only count cache-related reads.
		 * Anything that called fsck_bread() or getblk()
		 * directly are explicitly not cached, so they're not
		 * included here.
		 */
		if (totalreads != 0)
			percentage = diskreads * 100 / totalreads;
		else
			percentage = 0;

		(void) printf("cache missed %lld of %lld reads (%lld%%)\n",
		    (longlong_t)diskreads, (longlong_t)totalreads,
		    (longlong_t)percentage);
	}

	(void) close(fsreadfd);
	(void) close(fswritefd);
	fsreadfd = -1;
	fswritefd = -1;
}

int
fsck_bread(int fd, caddr_t buf, diskaddr_t blk, size_t size)
{
	caddr_t cp;
	int i;
	int errs;
	offset_t offset = ldbtob(blk);
	offset_t addr;

	/*
	 * In our universe, nothing exists before the superblock, so
	 * just pretend it's always zeros.  This is the complement of
	 * bwrite()'s ignoring write requests into that space.
	 */
	if (blk < SBLOCK) {
		if (debug)
			(void) printf(
			    "WARNING: fsck_bread() passed blkno < %d (%lld)\n",
			    SBLOCK, (longlong_t)blk);
		(void) memset(buf, 0, (size_t)size);
		return (1);
	}

	if (llseek(fd, offset, SEEK_SET) < 0) {
		rwerror("SEEK", blk, -1);
	}

	if ((i = read(fd, buf, size)) == size) {
		return (0);
	}
	rwerror("READ", blk, i);
	if (llseek(fd, offset, SEEK_SET) < 0) {
		rwerror("SEEK", blk, -1);
	}
	errs = 0;
	(void) memset(buf, 0, (size_t)size);
	pwarn("THE FOLLOWING SECTORS COULD NOT BE READ:");
	for (cp = buf, i = 0; i < btodb(size); i++, cp += DEV_BSIZE) {
		addr = ldbtob(blk + i);
		if (llseek(fd, addr, SEEK_SET) < 0 ||
		    read(fd, cp, (int)secsize) < 0) {
			iscorrupt = 1;
			(void) printf(" %llu", blk + (u_longlong_t)i);
			errs++;
		}
	}
	(void) printf("\n");
	return (errs);
}

void
bwrite(int fd, caddr_t buf, diskaddr_t blk, int64_t size)
{
	int i;
	int n;
	caddr_t cp;
	offset_t offset = ldbtob(blk);
	offset_t addr;

	if (fd < 0)
		return;
	if (blk < SBLOCK) {
		if (debug)
			(void) printf(
		    "WARNING: Attempt to write illegal blkno %lld on %s\n",
			    (longlong_t)blk, devname);
		return;
	}
	if (llseek(fd, offset, SEEK_SET) < 0) {
		rwerror("SEEK", blk, -1);
	}
	if ((i = write(fd, buf, (int)size)) == size) {
		fsmodified = 1;
		return;
	}
	rwerror("WRITE", blk, i);
	if (llseek(fd, offset, SEEK_SET) < 0) {
		rwerror("SEEK", blk, -1);
	}
	pwarn("THE FOLLOWING SECTORS COULD NOT BE WRITTEN:");
	for (cp = buf, i = 0; i < btodb(size); i++, cp += DEV_BSIZE) {
		n = 0;
		addr = ldbtob(blk + i);
		if (llseek(fd, addr, SEEK_SET) < 0 ||
		    (n = write(fd, cp, DEV_BSIZE)) < 0) {
			iscorrupt = 1;
			(void) printf(" %llu", blk + (u_longlong_t)i);
		} else if (n > 0) {
			fsmodified = 1;
		}

	}
	(void) printf("\n");
}

/*
 * Allocates the specified number of contiguous fragments.
 */
daddr32_t
allocblk(int wantedfrags)
{
	int block, leadfrag, tailfrag;
	daddr32_t selected;
	size_t size;
	struct bufarea *bp;

	/*
	 * It's arguable whether we should just fail, or instead
	 * error out here.  Since we should only ever be asked for
	 * a single fragment or an entire block (i.e., sblock.fs_frag),
	 * we'll fail out because anything else means somebody
	 * changed code without considering all of the ramifications.