NetBSD/sbin/fsck_lfs/inode.c

860 lines
20 KiB
C
Raw Normal View History

/* $NetBSD: inode.c,v 1.13 2002/05/23 04:05:11 perseant Exp $ */
/*
* Copyright (c) 1997, 1998
* Konrad Schroder. All rights reserved.
* Copyright (c) 1980, 1986, 1993
* The Regents of the University of California. All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by the University of
* California, Berkeley and its contributors.
* 4. 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 BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*/
#include <sys/param.h>
#include <sys/time.h>
#include <ufs/ufs/dinode.h>
#include <ufs/ufs/dir.h>
2000-05-23 05:48:52 +04:00
#include <sys/mount.h> /* XXX */
#include <ufs/lfs/lfs.h>
#include <ufs/lfs/lfs_extern.h>
#ifndef SMALL
#include <pwd.h>
#endif
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include "fsck.h"
#include "fsutil.h"
#include "extern.h"
2000-05-23 05:48:52 +04:00
extern SEGUSE *seg_table;
extern daddr_t *din_table;
2000-05-23 05:48:52 +04:00
static int iblock(struct inodesc *, long, u_int64_t);
int blksreqd(struct lfs *, int);
int lfs_maxino(void);
/* static void dump_inoblk (struct lfs *, struct dinode *); */
/* stolen from lfs_inode.c */
/* Search a block for a specific dinode. */
2000-05-23 05:48:52 +04:00
struct dinode *
lfs_difind(struct lfs * fs, ino_t ino, struct dinode * dip)
{
Merge the short-lived perseant-lfsv2 branch into the trunk. Kernels and tools understand both v1 and v2 filesystems; newfs_lfs generates v2 by default. Changes for the v2 layout include: - Segments of non-PO2 size and arbitrary block offset, so these can be matched to convenient physical characteristics of the partition (e.g., stripe or track size and offset). - Address by fragment instead of by disk sector, paving the way for non-512-byte-sector devices. In theory fragments can be as large as you like, though in reality they must be smaller than MAXBSIZE in size. - Use serial number and filesystem identifier to ensure that roll-forward doesn't get old data and think it's new. Roll-forward is enabled for v2 filesystems, though not for v1 filesystems by default. - The inode free list is now a tailq, paving the way for undelete (undelete is not yet implemented, but can be without further non-backwards-compatible changes to disk structures). - Inode atime information is kept in the Ifile, instead of on the inode; that is, the inode is never written *just* because atime was changed. Because of this the inodes remain near the file data on the disk, rather than wandering all over as the disk is read repeatedly. This speeds up repeated reads by a small but noticeable amount. Other changes of note include: - The ifile written by newfs_lfs can now be of arbitrary length, it is no longer restricted to a single indirect block. - Fixed an old bug where ctime was changed every time a vnode was created. I need to look more closely to make sure that the times are only updated during write(2) and friends, not after-the-fact during a segment write, and certainly not by the cleaner.
2001-07-14 00:30:18 +04:00
struct dinode *ldip, *fin;
Merge the short-lived perseant-lfsv2 branch into the trunk. Kernels and tools understand both v1 and v2 filesystems; newfs_lfs generates v2 by default. Changes for the v2 layout include: - Segments of non-PO2 size and arbitrary block offset, so these can be matched to convenient physical characteristics of the partition (e.g., stripe or track size and offset). - Address by fragment instead of by disk sector, paving the way for non-512-byte-sector devices. In theory fragments can be as large as you like, though in reality they must be smaller than MAXBSIZE in size. - Use serial number and filesystem identifier to ensure that roll-forward doesn't get old data and think it's new. Roll-forward is enabled for v2 filesystems, though not for v1 filesystems by default. - The inode free list is now a tailq, paving the way for undelete (undelete is not yet implemented, but can be without further non-backwards-compatible changes to disk structures). - Inode atime information is kept in the Ifile, instead of on the inode; that is, the inode is never written *just* because atime was changed. Because of this the inodes remain near the file data on the disk, rather than wandering all over as the disk is read repeatedly. This speeds up repeated reads by a small but noticeable amount. Other changes of note include: - The ifile written by newfs_lfs can now be of arbitrary length, it is no longer restricted to a single indirect block. - Fixed an old bug where ctime was changed every time a vnode was created. I need to look more closely to make sure that the times are only updated during write(2) and friends, not after-the-fact during a segment write, and certainly not by the cleaner.
2001-07-14 00:30:18 +04:00
#ifdef LFS_IFILE_FRAG_ADDRESSING
if (fs->lfs_version == 1)
fin = dip + INOPB(fs);
else
fin = dip + INOPF(fs);
#else
fin = dip + INOPB(fs);
#endif
for (ldip = dip; ldip < fin; ++ldip) {
if (ldip->di_inumber == ino)
return ldip;
}
2000-05-23 05:48:52 +04:00
/* printf("lfs_difind: dinode %u not found\n", ino); */
return NULL;
}
/*
* Calculate the number of blocks required to be able to address data block
* blkno (counting, of course, indirect blocks). blkno must >=0.
*/
2000-05-23 05:48:52 +04:00
int
blksreqd(struct lfs * fs, int blkno)
{
2000-05-23 05:48:52 +04:00
long n = blkno;
if (blkno < NDADDR)
return blkno;
n -= NDADDR;
if (n < NINDIR(fs))
return blkno + 1;
n -= NINDIR(fs);
if (n < NINDIR(fs) * NINDIR(fs))
return blkno + 2 + n / NINDIR(fs) + 1;
n -= NINDIR(fs) * NINDIR(fs);
return blkno + 2 + NINDIR(fs) + n / (NINDIR(fs) * NINDIR(fs)) + 1;
}
#define BASE_SINDIR (NDADDR)
#define BASE_DINDIR (NDADDR+NINDIR(fs))
#define BASE_TINDIR (NDADDR+NINDIR(fs)+NINDIR(fs)*NINDIR(fs))
#define D_UNITS (NINDIR(fs))
#define T_UNITS (NINDIR(fs)*NINDIR(fs))
2000-05-23 05:48:52 +04:00
ufs_daddr_t lfs_bmap(struct lfs *, struct dinode *, ufs_daddr_t);
2000-05-23 05:48:52 +04:00
ufs_daddr_t
lfs_bmap(struct lfs * fs, struct dinode * idinode, ufs_daddr_t lbn)
{
2000-05-23 05:48:52 +04:00
ufs_daddr_t residue, up, off = 0;
struct bufarea *bp;
2000-05-23 05:48:52 +04:00
if (lbn > 0 && lbn > (idinode->di_size - 1) / dev_bsize) {
return UNASSIGNED;
}
/*
* Indirect blocks: if it is a first-level indirect, pull its
* address from the inode; otherwise, call ourselves to find the
* address of the parent indirect block, and load that to find
* the desired address.
*/
2000-05-23 05:48:52 +04:00
if (lbn < 0) {
lbn *= -1;
2000-05-23 05:48:52 +04:00
if (lbn == NDADDR) {
/* printf("lbn %d: single indir base\n", -lbn); */
2000-05-23 05:48:52 +04:00
return idinode->di_ib[0]; /* single indirect */
} else if (lbn == BASE_DINDIR + 1) {
/* printf("lbn %d: double indir base\n", -lbn); */
2000-05-23 05:48:52 +04:00
return idinode->di_ib[1]; /* double indirect */
} else if (lbn == BASE_TINDIR + 2) {
/* printf("lbn %d: triple indir base\n", -lbn); */
2000-05-23 05:48:52 +04:00
return idinode->di_ib[2]; /* triple indirect */
}
/*
* Find the immediate parent. This is essentially finding the
* residue of modulus, and then rounding accordingly.
*/
2000-05-23 05:48:52 +04:00
residue = (lbn - NDADDR) % NINDIR(fs);
if (residue == 1) {
/* Double indirect. Parent is the triple. */
up = idinode->di_ib[2];
2000-05-23 05:48:52 +04:00
off = (lbn - 2 - BASE_TINDIR) / (NINDIR(fs) * NINDIR(fs));
if (up == UNASSIGNED || up == LFS_UNUSED_DADDR)
return UNASSIGNED;
/* printf("lbn %d: parent is the triple\n", -lbn); */
2000-05-23 05:48:52 +04:00
bp = getddblk(up, sblock.lfs_bsize);
bp->b_flags &= ~B_INUSE;
return ((daddr_t *)(bp->b_un.b_buf))[off];
2000-05-23 05:48:52 +04:00
} else { /* residue == 0 */
/* Single indirect. Two cases. */
2000-05-23 05:48:52 +04:00
if (lbn < BASE_TINDIR) {
/* Parent is the double, simple */
up = -(BASE_DINDIR) - 1;
2000-05-23 05:48:52 +04:00
off = (lbn - BASE_DINDIR) / D_UNITS;
/*
* printf("lbn %d: parent is %d/%d\n", -lbn,
* up,off);
*/
} else {
/* Ancestor is the triple, more complex */
2000-05-23 05:48:52 +04:00
up = ((lbn - BASE_TINDIR) / T_UNITS)
* T_UNITS + BASE_TINDIR + 1;
2000-05-23 05:48:52 +04:00
off = (lbn / D_UNITS) - (up / D_UNITS);
up = -up;
2000-05-23 05:48:52 +04:00
/*
* printf("lbn %d: parent is %d/%d\n", -lbn,
* up,off);
*/
}
}
} else {
/* Direct block. Its parent must be a single indirect. */
2000-05-23 05:48:52 +04:00
if (lbn < NDADDR)
return idinode->di_db[lbn];
else {
/* Parent is an indirect block. */
2000-05-23 05:48:52 +04:00
up = -(((lbn - NDADDR) / D_UNITS) * D_UNITS + NDADDR);
off = (lbn - NDADDR) % D_UNITS;
/* printf("lbn %d: parent is %d/%d\n", lbn,up,off); */
}
}
2000-05-23 05:48:52 +04:00
up = lfs_bmap(fs, idinode, up);
if (up == UNASSIGNED || up == LFS_UNUSED_DADDR)
return UNASSIGNED;
2000-05-23 05:48:52 +04:00
bp = getddblk(up, sblock.lfs_bsize);
bp->b_flags &= ~B_INUSE;
return ((daddr_t *)(bp->b_un.b_buf))[off];
}
/*
* This is kind of gross. We use this to find the nth block
* from a file whose inode has disk address idaddr. In practice
* we will only use this to find blocks of the ifile.
*/
static struct bufarea empty;
struct bufarea *
2000-05-23 05:48:52 +04:00
getfileblk(struct lfs * fs, struct dinode * idinode, ino_t lbn)
{
2000-05-23 05:48:52 +04:00
struct bufarea *bp;
ufs_daddr_t blkno;
static char empty_buf[65536];
empty.b_un.b_buf = &(empty_buf[0]);
blkno = lfs_bmap(fs, idinode, lbn);
if (blkno == UNASSIGNED || blkno == LFS_UNUSED_DADDR) {
printf("Warning: ifile lbn %d unassigned!\n", lbn);
return &empty;
}
bp = getddblk(blkno, sblock.lfs_bsize);
return bp;
}
#if 0
2000-05-23 05:48:52 +04:00
static struct dinode *
gidinode(void)
{
2000-05-23 05:48:52 +04:00
static struct dinode *idinode;
2000-05-23 05:48:52 +04:00
if (!idinode) { /* only need to do this once */
idinode = lfs_difind(&sblock, sblock.lfs_ifile, &ifblock);
}
return idinode;
}
#endif
2000-05-23 05:48:52 +04:00
struct ifile *
lfs_ientry(ino_t ino, struct bufarea ** bpp)
{
Merge the short-lived perseant-lfsv2 branch into the trunk. Kernels and tools understand both v1 and v2 filesystems; newfs_lfs generates v2 by default. Changes for the v2 layout include: - Segments of non-PO2 size and arbitrary block offset, so these can be matched to convenient physical characteristics of the partition (e.g., stripe or track size and offset). - Address by fragment instead of by disk sector, paving the way for non-512-byte-sector devices. In theory fragments can be as large as you like, though in reality they must be smaller than MAXBSIZE in size. - Use serial number and filesystem identifier to ensure that roll-forward doesn't get old data and think it's new. Roll-forward is enabled for v2 filesystems, though not for v1 filesystems by default. - The inode free list is now a tailq, paving the way for undelete (undelete is not yet implemented, but can be without further non-backwards-compatible changes to disk structures). - Inode atime information is kept in the Ifile, instead of on the inode; that is, the inode is never written *just* because atime was changed. Because of this the inodes remain near the file data on the disk, rather than wandering all over as the disk is read repeatedly. This speeds up repeated reads by a small but noticeable amount. Other changes of note include: - The ifile written by newfs_lfs can now be of arbitrary length, it is no longer restricted to a single indirect block. - Fixed an old bug where ctime was changed every time a vnode was created. I need to look more closely to make sure that the times are only updated during write(2) and friends, not after-the-fact during a segment write, and certainly not by the cleaner.
2001-07-14 00:30:18 +04:00
IFILE *ifp;
2000-05-23 05:48:52 +04:00
*bpp = getfileblk(&sblock, lfs_ginode(LFS_IFILE_INUM),
ino / sblock.lfs_ifpb + sblock.lfs_cleansz +
sblock.lfs_segtabsz);
if (*bpp == &empty) {
printf("Warning: ino %d ientry in unassigned block\n", ino);
}
2000-05-23 05:48:52 +04:00
if (*bpp) {
Merge the short-lived perseant-lfsv2 branch into the trunk. Kernels and tools understand both v1 and v2 filesystems; newfs_lfs generates v2 by default. Changes for the v2 layout include: - Segments of non-PO2 size and arbitrary block offset, so these can be matched to convenient physical characteristics of the partition (e.g., stripe or track size and offset). - Address by fragment instead of by disk sector, paving the way for non-512-byte-sector devices. In theory fragments can be as large as you like, though in reality they must be smaller than MAXBSIZE in size. - Use serial number and filesystem identifier to ensure that roll-forward doesn't get old data and think it's new. Roll-forward is enabled for v2 filesystems, though not for v1 filesystems by default. - The inode free list is now a tailq, paving the way for undelete (undelete is not yet implemented, but can be without further non-backwards-compatible changes to disk structures). - Inode atime information is kept in the Ifile, instead of on the inode; that is, the inode is never written *just* because atime was changed. Because of this the inodes remain near the file data on the disk, rather than wandering all over as the disk is read repeatedly. This speeds up repeated reads by a small but noticeable amount. Other changes of note include: - The ifile written by newfs_lfs can now be of arbitrary length, it is no longer restricted to a single indirect block. - Fixed an old bug where ctime was changed every time a vnode was created. I need to look more closely to make sure that the times are only updated during write(2) and friends, not after-the-fact during a segment write, and certainly not by the cleaner.
2001-07-14 00:30:18 +04:00
if (sblock.lfs_version > 1) {
ifp = (((IFILE *)((*bpp)->b_un.b_buf)) +
(ino % sblock.lfs_ifpb));
} else {
ifp = (IFILE *)(((IFILE_V1 *)
((*bpp)->b_un.b_buf)) +
(ino % sblock.lfs_ifpb));
}
2000-05-23 05:48:52 +04:00
return ifp;
} else
return NULL;
}
2000-05-23 05:48:52 +04:00
SEGUSE *
lfs_gseguse(int segnum, struct bufarea ** bpp)
{
2000-05-23 05:48:52 +04:00
int blkno;
Merge the short-lived perseant-lfsv2 branch into the trunk. Kernels and tools understand both v1 and v2 filesystems; newfs_lfs generates v2 by default. Changes for the v2 layout include: - Segments of non-PO2 size and arbitrary block offset, so these can be matched to convenient physical characteristics of the partition (e.g., stripe or track size and offset). - Address by fragment instead of by disk sector, paving the way for non-512-byte-sector devices. In theory fragments can be as large as you like, though in reality they must be smaller than MAXBSIZE in size. - Use serial number and filesystem identifier to ensure that roll-forward doesn't get old data and think it's new. Roll-forward is enabled for v2 filesystems, though not for v1 filesystems by default. - The inode free list is now a tailq, paving the way for undelete (undelete is not yet implemented, but can be without further non-backwards-compatible changes to disk structures). - Inode atime information is kept in the Ifile, instead of on the inode; that is, the inode is never written *just* because atime was changed. Because of this the inodes remain near the file data on the disk, rather than wandering all over as the disk is read repeatedly. This speeds up repeated reads by a small but noticeable amount. Other changes of note include: - The ifile written by newfs_lfs can now be of arbitrary length, it is no longer restricted to a single indirect block. - Fixed an old bug where ctime was changed every time a vnode was created. I need to look more closely to make sure that the times are only updated during write(2) and friends, not after-the-fact during a segment write, and certainly not by the cleaner.
2001-07-14 00:30:18 +04:00
struct bufarea *bp;
Merge the short-lived perseant-lfsv2 branch into the trunk. Kernels and tools understand both v1 and v2 filesystems; newfs_lfs generates v2 by default. Changes for the v2 layout include: - Segments of non-PO2 size and arbitrary block offset, so these can be matched to convenient physical characteristics of the partition (e.g., stripe or track size and offset). - Address by fragment instead of by disk sector, paving the way for non-512-byte-sector devices. In theory fragments can be as large as you like, though in reality they must be smaller than MAXBSIZE in size. - Use serial number and filesystem identifier to ensure that roll-forward doesn't get old data and think it's new. Roll-forward is enabled for v2 filesystems, though not for v1 filesystems by default. - The inode free list is now a tailq, paving the way for undelete (undelete is not yet implemented, but can be without further non-backwards-compatible changes to disk structures). - Inode atime information is kept in the Ifile, instead of on the inode; that is, the inode is never written *just* because atime was changed. Because of this the inodes remain near the file data on the disk, rather than wandering all over as the disk is read repeatedly. This speeds up repeated reads by a small but noticeable amount. Other changes of note include: - The ifile written by newfs_lfs can now be of arbitrary length, it is no longer restricted to a single indirect block. - Fixed an old bug where ctime was changed every time a vnode was created. I need to look more closely to make sure that the times are only updated during write(2) and friends, not after-the-fact during a segment write, and certainly not by the cleaner.
2001-07-14 00:30:18 +04:00
blkno = segnum / sblock.lfs_sepb + sblock.lfs_cleansz;
(*bpp) = bp = getfileblk(&sblock, lfs_ginode(LFS_IFILE_INUM), blkno);
if (sblock.lfs_version == 1)
return (SEGUSE *)((SEGUSE_V1 *)(bp->b_un.b_buf) +
segnum % sblock.lfs_sepb);
else
return (SEGUSE *)(bp->b_un.b_buf) + segnum % sblock.lfs_sepb;
}
daddr_t
lfs_ino_daddr(ino_t inumber)
{
2000-05-23 05:48:52 +04:00
daddr_t daddr;
IFILE *ifp;
struct bufarea *bp;
2000-05-23 05:48:52 +04:00
if (din_table[inumber]) {
daddr = din_table[inumber];
} else {
2000-05-23 05:48:52 +04:00
if (inumber == LFS_IFILE_INUM)
daddr = idaddr;
else {
ifp = lfs_ientry(inumber, &bp);
2000-05-23 05:48:52 +04:00
if (ifp == NULL) {
return NULL;
}
2000-05-23 05:48:52 +04:00
if (ifp->if_daddr == LFS_UNUSED_DADDR) {
bp->b_flags &= ~B_INUSE;
return NULL;
}
bp->b_flags &= ~B_INUSE;
daddr = ifp->if_daddr;
}
2000-05-23 05:48:52 +04:00
din_table[inumber] = daddr;
Merge the short-lived perseant-lfsv2 branch into the trunk. Kernels and tools understand both v1 and v2 filesystems; newfs_lfs generates v2 by default. Changes for the v2 layout include: - Segments of non-PO2 size and arbitrary block offset, so these can be matched to convenient physical characteristics of the partition (e.g., stripe or track size and offset). - Address by fragment instead of by disk sector, paving the way for non-512-byte-sector devices. In theory fragments can be as large as you like, though in reality they must be smaller than MAXBSIZE in size. - Use serial number and filesystem identifier to ensure that roll-forward doesn't get old data and think it's new. Roll-forward is enabled for v2 filesystems, though not for v1 filesystems by default. - The inode free list is now a tailq, paving the way for undelete (undelete is not yet implemented, but can be without further non-backwards-compatible changes to disk structures). - Inode atime information is kept in the Ifile, instead of on the inode; that is, the inode is never written *just* because atime was changed. Because of this the inodes remain near the file data on the disk, rather than wandering all over as the disk is read repeatedly. This speeds up repeated reads by a small but noticeable amount. Other changes of note include: - The ifile written by newfs_lfs can now be of arbitrary length, it is no longer restricted to a single indirect block. - Fixed an old bug where ctime was changed every time a vnode was created. I need to look more closely to make sure that the times are only updated during write(2) and friends, not after-the-fact during a segment write, and certainly not by the cleaner.
2001-07-14 00:30:18 +04:00
seg_table[dtosn(&sblock, daddr)].su_nbytes += DINODE_SIZE;
}
return daddr;
}
2000-05-23 05:48:52 +04:00
struct dinode *
lfs_ginode(ino_t inumber)
{
2000-05-23 05:48:52 +04:00
struct ifile *ifp;
struct dinode *din;
struct bufarea *bp;
daddr_t daddr;
if (inumber >= maxino)
2000-05-23 05:48:52 +04:00
errexit("bad inode number %d to lfs_ginode\n", inumber);
#if 0
2000-05-23 05:48:52 +04:00
if (inumber == LFS_IFILE_INUM) {
daddr = idaddr;
2000-05-23 05:48:52 +04:00
if (din_table[LFS_IFILE_INUM] == 0) {
din_table[LFS_IFILE_INUM] = daddr;
Merge the short-lived perseant-lfsv2 branch into the trunk. Kernels and tools understand both v1 and v2 filesystems; newfs_lfs generates v2 by default. Changes for the v2 layout include: - Segments of non-PO2 size and arbitrary block offset, so these can be matched to convenient physical characteristics of the partition (e.g., stripe or track size and offset). - Address by fragment instead of by disk sector, paving the way for non-512-byte-sector devices. In theory fragments can be as large as you like, though in reality they must be smaller than MAXBSIZE in size. - Use serial number and filesystem identifier to ensure that roll-forward doesn't get old data and think it's new. Roll-forward is enabled for v2 filesystems, though not for v1 filesystems by default. - The inode free list is now a tailq, paving the way for undelete (undelete is not yet implemented, but can be without further non-backwards-compatible changes to disk structures). - Inode atime information is kept in the Ifile, instead of on the inode; that is, the inode is never written *just* because atime was changed. Because of this the inodes remain near the file data on the disk, rather than wandering all over as the disk is read repeatedly. This speeds up repeated reads by a small but noticeable amount. Other changes of note include: - The ifile written by newfs_lfs can now be of arbitrary length, it is no longer restricted to a single indirect block. - Fixed an old bug where ctime was changed every time a vnode was created. I need to look more closely to make sure that the times are only updated during write(2) and friends, not after-the-fact during a segment write, and certainly not by the cleaner.
2001-07-14 00:30:18 +04:00
seg_table[dtosn(&sblock, daddr)].su_nbytes += DINODE_SIZE;
2000-05-23 05:48:52 +04:00
}
return gidinode();
}
#endif
2000-05-23 05:48:52 +04:00
daddr = lfs_ino_daddr(inumber);
if (daddr == 0)
return NULL;
if (pbp)
pbp->b_flags &= ~B_INUSE;
Merge the short-lived perseant-lfsv2 branch into the trunk. Kernels and tools understand both v1 and v2 filesystems; newfs_lfs generates v2 by default. Changes for the v2 layout include: - Segments of non-PO2 size and arbitrary block offset, so these can be matched to convenient physical characteristics of the partition (e.g., stripe or track size and offset). - Address by fragment instead of by disk sector, paving the way for non-512-byte-sector devices. In theory fragments can be as large as you like, though in reality they must be smaller than MAXBSIZE in size. - Use serial number and filesystem identifier to ensure that roll-forward doesn't get old data and think it's new. Roll-forward is enabled for v2 filesystems, though not for v1 filesystems by default. - The inode free list is now a tailq, paving the way for undelete (undelete is not yet implemented, but can be without further non-backwards-compatible changes to disk structures). - Inode atime information is kept in the Ifile, instead of on the inode; that is, the inode is never written *just* because atime was changed. Because of this the inodes remain near the file data on the disk, rather than wandering all over as the disk is read repeatedly. This speeds up repeated reads by a small but noticeable amount. Other changes of note include: - The ifile written by newfs_lfs can now be of arbitrary length, it is no longer restricted to a single indirect block. - Fixed an old bug where ctime was changed every time a vnode was created. I need to look more closely to make sure that the times are only updated during write(2) and friends, not after-the-fact during a segment write, and certainly not by the cleaner.
2001-07-14 00:30:18 +04:00
if (sblock.lfs_version == 1)
pbp = getddblk(daddr, sblock.lfs_bsize);
else
pbp = getddblk(daddr, sblock.lfs_fsize);
2000-05-23 05:48:52 +04:00
din = lfs_difind(&sblock, inumber, pbp->b_un.b_dinode);
if (din == NULL) {
pfatal("INODE %d NOT FOUND\n", inumber);
if (reply("free")) {
ifp = lfs_ientry(inumber, &bp);
ifp->if_daddr = LFS_UNUSED_DADDR;
ifp->if_nextfree = sblock.lfs_free;
sblock.lfs_free = inumber;
sbdirty();
dirty(bp);
bp->b_flags &= ~B_INUSE;
}
}
return din;
}
/* imported from lfs_vfsops.c */
int
2000-05-23 05:48:52 +04:00
ino_to_fsba(struct lfs * fs, ino_t ino)
{
2000-05-23 05:48:52 +04:00
daddr_t daddr = LFS_UNUSED_DADDR;
struct ifile *ifp;
struct bufarea *bp;
/* Translate the inode number to a disk address. */
if (ino == LFS_IFILE_INUM)
daddr = fs->lfs_idaddr;
else {
ifp = lfs_ientry(ino, &bp);
if (ifp) {
daddr = ifp->if_daddr;
} else {
pwarn("Can't locate inode #%ud\n", ino);
}
bp->b_flags &= ~B_INUSE;
}
return daddr;
}
/*
* Check validity of held (direct) blocks in an inode.
*/
int
2000-05-23 05:48:52 +04:00
ckinode(struct dinode *dp, struct inodesc *idesc)
{
2000-05-23 05:48:52 +04:00
register ufs_daddr_t *ap;
long ret, n, ndb, offset;
struct dinode dino;
u_int64_t remsize, sizepb;
mode_t mode;
char pathbuf[MAXPATHLEN + 1];
if (idesc->id_fix != IGNORE)
idesc->id_fix = DONTKNOW;
idesc->id_entryno = 0;
idesc->id_filesize = dp->di_size;
mode = dp->di_mode & IFMT;
if (mode == IFBLK || mode == IFCHR ||
(mode == IFLNK && (dp->di_size < sblock.lfs_maxsymlinklen ||
(sblock.lfs_maxsymlinklen == 0 &&
dp->di_blocks == 0))))
return (KEEPON);
dino = *dp;
ndb = howmany(dino.di_size, sblock.lfs_bsize);
for (ap = &dino.di_db[0]; ap < &dino.di_db[NDADDR]; ap++) {
if (--ndb == 0 && (offset = blkoff(&sblock, dino.di_size)) != 0) {
idesc->id_numfrags =
numfrags(&sblock, fragroundup(&sblock, offset));
} else
idesc->id_numfrags = sblock.lfs_frag;
if (*ap == 0) {
if (idesc->id_type == DATA && ndb >= 0) {
/* An empty block in a directory XXX */
getpathname(pathbuf, idesc->id_number,
idesc->id_number);
pfatal("DIRECTORY %s: CONTAINS EMPTY BLOCKS",
pathbuf);
if (reply("ADJUST LENGTH") == 1) {
dp = ginode(idesc->id_number);
dp->di_size = (ap - &dino.di_db[0]) *
sblock.lfs_bsize;
printf(
"YOU MUST RERUN FSCK AFTERWARDS\n");
rerun = 1;
inodirty();
}
}
continue;
}
idesc->id_blkno = *ap;
idesc->id_lblkno = ap - &dino.di_db[0];
if (idesc->id_type == ADDR) {
ret = (*idesc->id_func)(idesc);
} else
ret = dirscan(idesc);
idesc->id_lblkno = 0;
if (ret & STOP)
return (ret);
}
idesc->id_numfrags = sblock.lfs_frag;
remsize = dino.di_size - sblock.lfs_bsize * NDADDR;
sizepb = sblock.lfs_bsize;
for (ap = &dino.di_ib[0], n = 1; n <= NIADDR; ap++, n++) {
if (*ap) {
idesc->id_blkno = *ap;
ret = iblock(idesc, n, remsize);
if (ret & STOP)
return (ret);
} else {
if (idesc->id_type == DATA && remsize > 0) {
/* An empty block in a directory XXX */
getpathname(pathbuf, idesc->id_number,
idesc->id_number);
pfatal("DIRECTORY %s: CONTAINS EMPTY BLOCKS",
pathbuf);
if (reply("ADJUST LENGTH") == 1) {
dp = ginode(idesc->id_number);
dp->di_size -= remsize;
remsize = 0;
printf(
"YOU MUST RERUN FSCK AFTERWARDS\n");
rerun = 1;
inodirty();
break;
}
}
}
sizepb *= NINDIR(&sblock);
remsize -= sizepb;
}
return (KEEPON);
}
static int
2000-05-23 05:48:52 +04:00
iblock(struct inodesc * idesc, long ilevel, u_int64_t isize)
{
daddr_t *ap, *aplim;
struct bufarea *bp;
2000-05-23 05:48:52 +04:00
int i, n, (*func)(struct inodesc *), nif;
u_int64_t sizepb;
char pathbuf[MAXPATHLEN + 1], buf[BUFSIZ];
struct dinode *dp;
if (idesc->id_type == ADDR) {
func = idesc->id_func;
n = (*func)(idesc);
if ((n & KEEPON) == 0)
return (n);
} else
func = dirscan;
Merge the short-lived perseant-lfsv2 branch into the trunk. Kernels and tools understand both v1 and v2 filesystems; newfs_lfs generates v2 by default. Changes for the v2 layout include: - Segments of non-PO2 size and arbitrary block offset, so these can be matched to convenient physical characteristics of the partition (e.g., stripe or track size and offset). - Address by fragment instead of by disk sector, paving the way for non-512-byte-sector devices. In theory fragments can be as large as you like, though in reality they must be smaller than MAXBSIZE in size. - Use serial number and filesystem identifier to ensure that roll-forward doesn't get old data and think it's new. Roll-forward is enabled for v2 filesystems, though not for v1 filesystems by default. - The inode free list is now a tailq, paving the way for undelete (undelete is not yet implemented, but can be without further non-backwards-compatible changes to disk structures). - Inode atime information is kept in the Ifile, instead of on the inode; that is, the inode is never written *just* because atime was changed. Because of this the inodes remain near the file data on the disk, rather than wandering all over as the disk is read repeatedly. This speeds up repeated reads by a small but noticeable amount. Other changes of note include: - The ifile written by newfs_lfs can now be of arbitrary length, it is no longer restricted to a single indirect block. - Fixed an old bug where ctime was changed every time a vnode was created. I need to look more closely to make sure that the times are only updated during write(2) and friends, not after-the-fact during a segment write, and certainly not by the cleaner.
2001-07-14 00:30:18 +04:00
if (chkrange(idesc->id_blkno, fragstofsb(&sblock, idesc->id_numfrags)))
2000-05-23 05:48:52 +04:00
return (SKIP);
bp = getddblk(idesc->id_blkno, sblock.lfs_bsize);
ilevel--;
for (sizepb = sblock.lfs_bsize, i = 0; i < ilevel; i++)
sizepb *= NINDIR(&sblock);
if (isize > sizepb * NINDIR(&sblock))
nif = NINDIR(&sblock);
else
nif = howmany(isize, sizepb);
if (idesc->id_func == pass1check && nif < NINDIR(&sblock)) {
aplim = &bp->b_un.b_indir[NINDIR(&sblock)];
for (ap = &bp->b_un.b_indir[nif]; ap < aplim; ap++) {
if (*ap == 0)
continue;
(void)sprintf(buf, "PARTIALLY TRUNCATED INODE I=%u",
idesc->id_number);
2000-05-23 05:48:52 +04:00
if (dofix(idesc, buf)) {
*ap = 0;
dirty(bp);
}
}
flush(fswritefd, bp);
}
aplim = &bp->b_un.b_indir[nif];
for (ap = bp->b_un.b_indir; ap < aplim; ap++) {
if (*ap) {
idesc->id_blkno = *ap;
if (ilevel == 0)
n = (*func)(idesc);
else
n = iblock(idesc, ilevel, isize);
if (n & STOP) {
bp->b_flags &= ~B_INUSE;
return (n);
}
} else {
if (idesc->id_type == DATA && isize > 0) {
/* An empty block in a directory XXX */
getpathname(pathbuf, idesc->id_number,
idesc->id_number);
pfatal("DIRECTORY %s: CONTAINS EMPTY BLOCKS",
pathbuf);
if (reply("ADJUST LENGTH") == 1) {
dp = ginode(idesc->id_number);
dp->di_size -= isize;
isize = 0;
printf(
"YOU MUST RERUN FSCK AFTERWARDS\n");
rerun = 1;
inodirty();
bp->b_flags &= ~B_INUSE;
return (STOP);
}
}
}
isize -= sizepb;
}
bp->b_flags &= ~B_INUSE;
return (KEEPON);
}
/*
* Check that a block in a legal block number.
* Return 0 if in range, 1 if out of range.
*/
int
2000-05-23 05:48:52 +04:00
chkrange(daddr_t blk, int cnt)
{
Merge the short-lived perseant-lfsv2 branch into the trunk. Kernels and tools understand both v1 and v2 filesystems; newfs_lfs generates v2 by default. Changes for the v2 layout include: - Segments of non-PO2 size and arbitrary block offset, so these can be matched to convenient physical characteristics of the partition (e.g., stripe or track size and offset). - Address by fragment instead of by disk sector, paving the way for non-512-byte-sector devices. In theory fragments can be as large as you like, though in reality they must be smaller than MAXBSIZE in size. - Use serial number and filesystem identifier to ensure that roll-forward doesn't get old data and think it's new. Roll-forward is enabled for v2 filesystems, though not for v1 filesystems by default. - The inode free list is now a tailq, paving the way for undelete (undelete is not yet implemented, but can be without further non-backwards-compatible changes to disk structures). - Inode atime information is kept in the Ifile, instead of on the inode; that is, the inode is never written *just* because atime was changed. Because of this the inodes remain near the file data on the disk, rather than wandering all over as the disk is read repeatedly. This speeds up repeated reads by a small but noticeable amount. Other changes of note include: - The ifile written by newfs_lfs can now be of arbitrary length, it is no longer restricted to a single indirect block. - Fixed an old bug where ctime was changed every time a vnode was created. I need to look more closely to make sure that the times are only updated during write(2) and friends, not after-the-fact during a segment write, and certainly not by the cleaner.
2001-07-14 00:30:18 +04:00
if (blk < sntod(&sblock, 0)) {
return (1);
}
if (blk > maxfsblock) {
return (1);
}
if (blk + cnt < sntod(&sblock, 0)) {
return (1);
}
Merge the short-lived perseant-lfsv2 branch into the trunk. Kernels and tools understand both v1 and v2 filesystems; newfs_lfs generates v2 by default. Changes for the v2 layout include: - Segments of non-PO2 size and arbitrary block offset, so these can be matched to convenient physical characteristics of the partition (e.g., stripe or track size and offset). - Address by fragment instead of by disk sector, paving the way for non-512-byte-sector devices. In theory fragments can be as large as you like, though in reality they must be smaller than MAXBSIZE in size. - Use serial number and filesystem identifier to ensure that roll-forward doesn't get old data and think it's new. Roll-forward is enabled for v2 filesystems, though not for v1 filesystems by default. - The inode free list is now a tailq, paving the way for undelete (undelete is not yet implemented, but can be without further non-backwards-compatible changes to disk structures). - Inode atime information is kept in the Ifile, instead of on the inode; that is, the inode is never written *just* because atime was changed. Because of this the inodes remain near the file data on the disk, rather than wandering all over as the disk is read repeatedly. This speeds up repeated reads by a small but noticeable amount. Other changes of note include: - The ifile written by newfs_lfs can now be of arbitrary length, it is no longer restricted to a single indirect block. - Fixed an old bug where ctime was changed every time a vnode was created. I need to look more closely to make sure that the times are only updated during write(2) and friends, not after-the-fact during a segment write, and certainly not by the cleaner.
2001-07-14 00:30:18 +04:00
if (blk + cnt > maxfsblock) {
2000-05-23 05:48:52 +04:00
return (1);
}
return (0);
}
/*
* General purpose interface for reading inodes.
*/
2000-05-23 05:48:52 +04:00
struct dinode *
ginode(ino_t inumber)
{
2000-05-23 05:48:52 +04:00
return lfs_ginode(inumber);
}
/*
* Routines to maintain information about directory inodes.
* This is built during the first pass and used during the
* second and third passes.
*
* Enter inodes into the cache.
*/
void
2000-05-23 05:48:52 +04:00
cacheino(struct dinode *dp, ino_t inumber)
{
2000-05-23 05:48:52 +04:00
register struct inoinfo *inp;
struct inoinfo **inpp;
unsigned int blks;
blks = howmany(dp->di_size, sblock.lfs_bsize);
if (blks > NDADDR)
blks = NDADDR + NIADDR;
inp = (struct inoinfo *)
malloc(sizeof(*inp) + (blks - 1) * sizeof(daddr_t));
if (inp == NULL)
return;
inpp = &inphead[inumber % numdirs];
inp->i_nexthash = *inpp;
*inpp = inp;
inp->i_child = inp->i_sibling = inp->i_parentp = 0;
if (inumber == ROOTINO)
inp->i_parent = ROOTINO;
else
inp->i_parent = (ino_t)0;
inp->i_dotdot = (ino_t)0;
inp->i_number = inumber;
inp->i_isize = dp->di_size;
inp->i_numblks = blks * sizeof(daddr_t);
memcpy(&inp->i_blks[0], &dp->di_db[0], (size_t)inp->i_numblks);
if (inplast == listmax) {
listmax += 100;
inpsort = (struct inoinfo **)realloc((char *) inpsort,
(unsigned)listmax * sizeof(struct inoinfo *));
if (inpsort == NULL)
2001-09-25 04:03:25 +04:00
errexit("cannot increase directory list\n");
2000-05-23 05:48:52 +04:00
}
inpsort[inplast++] = inp;
}
/*
* Look up an inode cache structure.
*/
struct inoinfo *
2000-05-23 05:48:52 +04:00
getinoinfo(ino_t inumber)
{
2000-05-23 05:48:52 +04:00
register struct inoinfo *inp;
for (inp = inphead[inumber % numdirs]; inp; inp = inp->i_nexthash) {
if (inp->i_number != inumber)
continue;
return (inp);
}
errexit("cannot find inode %d\n", inumber);
return ((struct inoinfo *)0);
}
/*
* Clean up all the inode cache structure.
*/
void
inocleanup()
{
2000-05-23 05:48:52 +04:00
register struct inoinfo **inpp;
if (inphead == NULL)
return;
for (inpp = &inpsort[inplast - 1]; inpp >= inpsort; inpp--)
free((char *)(*inpp));
free((char *)inphead);
free((char *)inpsort);
inphead = inpsort = NULL;
}
void
inodirty()
{
2000-05-23 05:48:52 +04:00
dirty(pbp);
}
void
2000-05-23 05:48:52 +04:00
clri(struct inodesc *idesc, char *type, int flag)
{
2000-05-23 05:48:52 +04:00
register struct dinode *dp;
struct bufarea *bp;
IFILE *ifp;
dp = ginode(idesc->id_number);
if (flag == 1) {
pwarn("%s %s", type,
(dp->di_mode & IFMT) == IFDIR ? "DIR" : "FILE");
pinode(idesc->id_number);
}
if (preen || reply("CLEAR") == 1) {
if (preen)
printf(" (CLEARED)\n");
n_files--;
(void)ckinode(dp, idesc);
clearinode(dp);
statemap[idesc->id_number] = USTATE;
inodirty();
/* Send cleared inode to the free list */
ifp = lfs_ientry(idesc->id_number, &bp);
ifp->if_daddr = LFS_UNUSED_DADDR;
ifp->if_nextfree = sblock.lfs_free;
sblock.lfs_free = idesc->id_number;
sbdirty();
dirty(bp);
bp->b_flags &= ~B_INUSE;
}
}
int
2000-05-23 05:48:52 +04:00
findname(struct inodesc *idesc)
{
2000-05-23 05:48:52 +04:00
register struct direct *dirp = idesc->id_dirp;
2000-05-23 05:48:52 +04:00
if (dirp->d_ino != idesc->id_parent)
return (KEEPON);
memcpy(idesc->id_name, dirp->d_name, (size_t)dirp->d_namlen + 1);
return (STOP | FOUND);
}
int
2000-05-23 05:48:52 +04:00
findino(struct inodesc *idesc)
{
2000-05-23 05:48:52 +04:00
register struct direct *dirp = idesc->id_dirp;
if (dirp->d_ino == 0)
return (KEEPON);
if (strcmp(dirp->d_name, idesc->id_name) == 0 &&
dirp->d_ino >= ROOTINO && dirp->d_ino < maxino) {
2000-05-23 05:48:52 +04:00
idesc->id_parent = dirp->d_ino;
return (STOP | FOUND);
}
return (KEEPON);
}
void
2000-05-23 05:48:52 +04:00
pinode(ino_t ino)
{
2000-05-23 05:48:52 +04:00
register struct dinode *dp;
register char *p;
struct passwd *pw;
time_t t;
printf(" I=%u ", ino);
if (ino < ROOTINO || ino >= maxino)
2000-05-23 05:48:52 +04:00
return;
dp = ginode(ino);
if (dp) {
printf(" OWNER=");
#ifndef SMALL
2000-05-23 05:48:52 +04:00
if ((pw = getpwuid((int)dp->di_uid)) != 0)
printf("%s ", pw->pw_name);
else
#endif
2000-05-23 05:48:52 +04:00
printf("%u ", (unsigned)dp->di_uid);
printf("MODE=%o\n", dp->di_mode);
if (preen)
printf("%s: ", cdevname());
printf("SIZE=%llu ", (unsigned long long)dp->di_size);
2000-05-23 05:48:52 +04:00
t = dp->di_mtime;
p = ctime(&t);
printf("MTIME=%12.12s %4.4s ", &p[4], &p[20]);
}
}
void
2000-05-23 05:48:52 +04:00
blkerror(ino_t ino, char *type, daddr_t blk)
{
2000-05-23 05:48:52 +04:00
pfatal("%d %s I=%u", blk, type, ino);
printf("\n");
if (exitonfail)
exit(1);
switch (statemap[ino]) {
2000-05-23 05:48:52 +04:00
case FSTATE:
statemap[ino] = FCLEAR;
return;
2000-05-23 05:48:52 +04:00
case DSTATE:
statemap[ino] = DCLEAR;
return;
2000-05-23 05:48:52 +04:00
case FCLEAR:
case DCLEAR:
return;
2000-05-23 05:48:52 +04:00
default:
2001-09-25 04:03:25 +04:00
errexit("BAD STATE %d TO BLKERR\n", statemap[ino]);
2000-05-23 05:48:52 +04:00
/* NOTREACHED */
}
}
/*
* allocate an unused inode
*/
ino_t
2000-05-23 05:48:52 +04:00
allocino(ino_t request, int type)
{
2000-05-23 05:48:52 +04:00
register ino_t ino;
register struct dinode *dp;
time_t t;
if (request == 0)
request = ROOTINO;
else if (statemap[request] != USTATE)
return (0);
for (ino = request; ino < maxino; ino++)
if (statemap[ino] == USTATE)
break;
if (ino == maxino)
return (0);
switch (type & IFMT) {
case IFDIR:
statemap[ino] = DSTATE;
break;
case IFREG:
case IFLNK:
statemap[ino] = FSTATE;
break;
default:
return (0);
}
dp = ginode(ino);
dp->di_db[0] = allocblk((long)1);
if (dp->di_db[0] == 0) {
statemap[ino] = USTATE;
return (0);
}
dp->di_mode = type;
(void)time(&t);
dp->di_atime = t;
dp->di_mtime = dp->di_ctime = dp->di_atime;
dp->di_size = sblock.lfs_fsize;
Merge the short-lived perseant-lfsv2 branch into the trunk. Kernels and tools understand both v1 and v2 filesystems; newfs_lfs generates v2 by default. Changes for the v2 layout include: - Segments of non-PO2 size and arbitrary block offset, so these can be matched to convenient physical characteristics of the partition (e.g., stripe or track size and offset). - Address by fragment instead of by disk sector, paving the way for non-512-byte-sector devices. In theory fragments can be as large as you like, though in reality they must be smaller than MAXBSIZE in size. - Use serial number and filesystem identifier to ensure that roll-forward doesn't get old data and think it's new. Roll-forward is enabled for v2 filesystems, though not for v1 filesystems by default. - The inode free list is now a tailq, paving the way for undelete (undelete is not yet implemented, but can be without further non-backwards-compatible changes to disk structures). - Inode atime information is kept in the Ifile, instead of on the inode; that is, the inode is never written *just* because atime was changed. Because of this the inodes remain near the file data on the disk, rather than wandering all over as the disk is read repeatedly. This speeds up repeated reads by a small but noticeable amount. Other changes of note include: - The ifile written by newfs_lfs can now be of arbitrary length, it is no longer restricted to a single indirect block. - Fixed an old bug where ctime was changed every time a vnode was created. I need to look more closely to make sure that the times are only updated during write(2) and friends, not after-the-fact during a segment write, and certainly not by the cleaner.
2001-07-14 00:30:18 +04:00
dp->di_blocks = btofsb(&sblock, sblock.lfs_fsize);
2000-05-23 05:48:52 +04:00
n_files++;
inodirty();
if (newinofmt)
typemap[ino] = IFTODT(type);
return (ino);
}
/*
* deallocate an inode
*/
void
2000-05-23 05:48:52 +04:00
freeino(ino_t ino)
{
2000-05-23 05:48:52 +04:00
struct inodesc idesc;
struct dinode *dp;
memset(&idesc, 0, sizeof(struct inodesc));
idesc.id_type = ADDR;
idesc.id_func = pass4check;
idesc.id_number = ino;
dp = ginode(ino);
(void)ckinode(dp, &idesc);
clearinode(dp);
inodirty();
statemap[ino] = USTATE;
n_files--;
}