NetBSD/sys/ufs/ffs/ffs_inode.c

648 lines
19 KiB
C

/* $NetBSD: ffs_inode.c,v 1.55 2003/04/02 10:39:37 fvdl Exp $ */
/*
* Copyright (c) 1982, 1986, 1989, 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.
*
* @(#)ffs_inode.c 8.13 (Berkeley) 4/21/95
*/
#include <sys/cdefs.h>
__KERNEL_RCSID(0, "$NetBSD: ffs_inode.c,v 1.55 2003/04/02 10:39:37 fvdl Exp $");
#if defined(_KERNEL_OPT)
#include "opt_ffs.h"
#include "opt_quota.h"
#endif
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/mount.h>
#include <sys/proc.h>
#include <sys/file.h>
#include <sys/buf.h>
#include <sys/vnode.h>
#include <sys/kernel.h>
#include <sys/malloc.h>
#include <sys/trace.h>
#include <sys/resourcevar.h>
#include <ufs/ufs/quota.h>
#include <ufs/ufs/inode.h>
#include <ufs/ufs/ufsmount.h>
#include <ufs/ufs/ufs_extern.h>
#include <ufs/ufs/ufs_bswap.h>
#include <ufs/ffs/fs.h>
#include <ufs/ffs/ffs_extern.h>
static int ffs_indirtrunc __P((struct inode *, daddr_t, daddr_t,
daddr_t, int, int64_t *));
/*
* Update the access, modified, and inode change times as specified
* by the IN_ACCESS, IN_UPDATE, and IN_CHANGE flags respectively.
* The IN_MODIFIED flag is used to specify that the inode needs to be
* updated but that the times have already been set. The access
* and modified times are taken from the second and third parameters;
* the inode change time is always taken from the current time. If
* UPDATE_WAIT flag is set, or UPDATE_DIROP is set and we are not doing
* softupdates, then wait for the disk write of the inode to complete.
*/
int
ffs_update(v)
void *v;
{
struct vop_update_args /* {
struct vnode *a_vp;
struct timespec *a_access;
struct timespec *a_modify;
int a_flags;
} */ *ap = v;
struct fs *fs;
struct buf *bp;
struct inode *ip;
int error;
struct timespec ts;
caddr_t cp;
int waitfor, flags;
if (ap->a_vp->v_mount->mnt_flag & MNT_RDONLY)
return (0);
ip = VTOI(ap->a_vp);
TIMEVAL_TO_TIMESPEC(&time, &ts);
FFS_ITIMES(ip,
ap->a_access ? ap->a_access : &ts,
ap->a_modify ? ap->a_modify : &ts, &ts);
flags = ip->i_flag & (IN_MODIFIED | IN_ACCESSED);
if (flags == 0)
return (0);
fs = ip->i_fs;
if ((flags & IN_MODIFIED) != 0 &&
(ap->a_vp->v_mount->mnt_flag & MNT_ASYNC) == 0) {
waitfor = ap->a_flags & UPDATE_WAIT;
if ((ap->a_flags & UPDATE_DIROP) && !DOINGSOFTDEP(ap->a_vp))
waitfor |= UPDATE_WAIT;
} else
waitfor = 0;
/*
* Ensure that uid and gid are correct. This is a temporary
* fix until fsck has been changed to do the update.
*/
if (fs->fs_magic == FS_UFS1_MAGIC && /* XXX */
fs->fs_old_inodefmt < FS_44INODEFMT) { /* XXX */
ip->i_ffs1_ouid = ip->i_uid; /* XXX */
ip->i_ffs1_ogid = ip->i_gid; /* XXX */
} /* XXX */
error = bread(ip->i_devvp,
fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
(int)fs->fs_bsize, NOCRED, &bp);
if (error) {
brelse(bp);
return (error);
}
ip->i_flag &= ~(IN_MODIFIED | IN_ACCESSED);
if (DOINGSOFTDEP(ap->a_vp))
softdep_update_inodeblock(ip, bp, waitfor);
else if (ip->i_ffs_effnlink != ip->i_nlink)
panic("ffs_update: bad link cnt");
if (fs->fs_magic == FS_UFS1_MAGIC) {
cp = (caddr_t)bp->b_data +
(ino_to_fsbo(fs, ip->i_number) * DINODE1_SIZE);
#ifdef FFS_EI
if (UFS_FSNEEDSWAP(fs))
ffs_dinode1_swap(ip->i_din.ffs1_din,
(struct ufs1_dinode *)cp);
else
#endif
memcpy(cp, ip->i_din.ffs1_din, DINODE1_SIZE);
} else {
cp = (caddr_t)bp->b_data +
(ino_to_fsbo(fs, ip->i_number) * DINODE2_SIZE);
#ifdef FFS_EI
if (UFS_FSNEEDSWAP(fs))
ffs_dinode2_swap(ip->i_din.ffs2_din,
(struct ufs2_dinode *)cp);
else
#endif
memcpy(cp, ip->i_din.ffs2_din, DINODE2_SIZE);
}
if (waitfor) {
return (bwrite(bp));
} else {
bdwrite(bp);
return (0);
}
}
#define SINGLE 0 /* index of single indirect block */
#define DOUBLE 1 /* index of double indirect block */
#define TRIPLE 2 /* index of triple indirect block */
/*
* Truncate the inode oip to at most length size, freeing the
* disk blocks.
*/
int
ffs_truncate(v)
void *v;
{
struct vop_truncate_args /* {
struct vnode *a_vp;
off_t a_length;
int a_flags;
struct ucred *a_cred;
struct proc *a_p;
} */ *ap = v;
struct vnode *ovp = ap->a_vp;
struct genfs_node *gp = VTOG(ovp);
daddr_t lastblock;
struct inode *oip;
daddr_t bn, lastiblock[NIADDR], indir_lbn[NIADDR];
daddr_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR];
off_t length = ap->a_length;
struct fs *fs;
int offset, size, level;
int64_t count, blocksreleased = 0;
int i, ioflag, aflag, nblocks;
int error, allerror = 0;
off_t osize;
if (length < 0)
return (EINVAL);
oip = VTOI(ovp);
if (ovp->v_type == VLNK &&
(oip->i_size < ovp->v_mount->mnt_maxsymlinklen ||
(ovp->v_mount->mnt_maxsymlinklen == 0 &&
DIP(oip, blocks) == 0))) {
KDASSERT(length == 0);
memset(SHORTLINK(oip), 0, (size_t)oip->i_size);
oip->i_size = 0;
DIP(oip, size) = 0;
oip->i_flag |= IN_CHANGE | IN_UPDATE;
return (VOP_UPDATE(ovp, NULL, NULL, UPDATE_WAIT));
}
if (oip->i_size == length) {
oip->i_flag |= IN_CHANGE | IN_UPDATE;
return (VOP_UPDATE(ovp, NULL, NULL, 0));
}
#ifdef QUOTA
if ((error = getinoquota(oip)) != 0)
return (error);
#endif
fs = oip->i_fs;
if (length > fs->fs_maxfilesize)
return (EFBIG);
osize = oip->i_size;
ioflag = ap->a_flags;
aflag = ioflag & IO_SYNC ? B_SYNC : 0;
/*
* Lengthen the size of the file. We must ensure that the
* last byte of the file is allocated. Since the smallest
* value of osize is 0, length will be at least 1.
*/
if (osize < length) {
if (lblkno(fs, osize) < NDADDR &&
lblkno(fs, osize) != lblkno(fs, length) &&
blkroundup(fs, osize) != osize) {
error = ufs_balloc_range(ovp, osize,
blkroundup(fs, osize) - osize, ap->a_cred, aflag);
if (error) {
return error;
}
if (ioflag & IO_SYNC) {
ovp->v_size = blkroundup(fs, osize);
simple_lock(&ovp->v_interlock);
VOP_PUTPAGES(ovp,
trunc_page(osize & ~(fs->fs_bsize - 1)),
round_page(ovp->v_size),
PGO_CLEANIT | PGO_SYNCIO);
}
}
error = ufs_balloc_range(ovp, length - 1, 1, ap->a_cred,
aflag);
if (error) {
(void) VOP_TRUNCATE(ovp, osize, ioflag & IO_SYNC,
ap->a_cred, ap->a_p);
return error;
}
uvm_vnp_setsize(ovp, length);
oip->i_flag |= IN_CHANGE | IN_UPDATE;
KASSERT(ovp->v_size == oip->i_size);
return (VOP_UPDATE(ovp, NULL, NULL, 1));
}
/*
* When truncating a regular file down to a non-block-aligned size,
* we must zero the part of last block which is past the new EOF.
* We must synchronously flush the zeroed pages to disk
* since the new pages will be invalidated as soon as we
* inform the VM system of the new, smaller size.
* We must do this before acquiring the GLOCK, since fetching
* the pages will acquire the GLOCK internally.
* So there is a window where another thread could see a whole
* zeroed page past EOF, but that's life.
*/
offset = blkoff(fs, length);
if (ovp->v_type == VREG && length < osize && offset != 0) {
voff_t eoz;
error = ufs_balloc_range(ovp, length - 1, 1, ap->a_cred,
aflag);
if (error) {
return error;
}
size = blksize(fs, oip, lblkno(fs, length));
eoz = MIN(lblktosize(fs, lblkno(fs, length)) + size, osize);
uvm_vnp_zerorange(ovp, length, eoz - length);
simple_lock(&ovp->v_interlock);
error = VOP_PUTPAGES(ovp, trunc_page(length), round_page(eoz),
PGO_CLEANIT | PGO_DEACTIVATE | PGO_SYNCIO);
if (error) {
return error;
}
}
lockmgr(&gp->g_glock, LK_EXCLUSIVE, NULL);
if (DOINGSOFTDEP(ovp)) {
if (length > 0) {
/*
* If a file is only partially truncated, then
* we have to clean up the data structures
* describing the allocation past the truncation
* point. Finding and deallocating those structures
* is a lot of work. Since partial truncation occurs
* rarely, we solve the problem by syncing the file
* so that it will have no data structures left.
*/
if ((error = VOP_FSYNC(ovp, ap->a_cred, FSYNC_WAIT,
0, 0, ap->a_p)) != 0) {
lockmgr(&gp->g_glock, LK_RELEASE, NULL);
return (error);
}
if (oip->i_flag & IN_SPACECOUNTED)
fs->fs_pendingblocks -= DIP(oip, blocks);
} else {
uvm_vnp_setsize(ovp, length);
#ifdef QUOTA
(void) chkdq(oip, -DIP(oip, blocks), NOCRED, 0);
#endif
softdep_setup_freeblocks(oip, length, 0);
(void) vinvalbuf(ovp, 0, ap->a_cred, ap->a_p, 0, 0);
lockmgr(&gp->g_glock, LK_RELEASE, NULL);
oip->i_flag |= IN_CHANGE | IN_UPDATE;
return (VOP_UPDATE(ovp, NULL, NULL, 0));
}
}
if (oip->i_ump->um_fstype != UFS2 &&
oip->i_fs->fs_magic == FS_UFS2_MAGIC)
panic("fstype inconsistency (%ld)", oip->i_ump->um_fstype);
oip->i_size = length;
DIP(oip, size) = length;
uvm_vnp_setsize(ovp, length);
/*
* Calculate index into inode's block list of
* last direct and indirect blocks (if any)
* which we want to keep. Lastblock is -1 when
* the file is truncated to 0.
*/
lastblock = lblkno(fs, length + fs->fs_bsize - 1) - 1;
lastiblock[SINGLE] = lastblock - NDADDR;
lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs);
lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs);
nblocks = btodb(fs->fs_bsize);
/*
* Update file and block pointers on disk before we start freeing
* blocks. If we crash before free'ing blocks below, the blocks
* will be returned to the free list. lastiblock values are also
* normalized to -1 for calls to ffs_indirtrunc below.
*/
for (level = TRIPLE; level >= SINGLE; level--) {
oldblks[NDADDR + level] = DIP(oip, ib[level]);
if (lastiblock[level] < 0) {
DIP(oip, ib[level]) = 0;
lastiblock[level] = -1;
}
}
for (i = 0; i < NDADDR; i++) {
oldblks[i] = DIP(oip, db[i]);
if (i > lastblock)
DIP(oip, db[i]) = 0;
}
oip->i_flag |= IN_CHANGE | IN_UPDATE;
error = VOP_UPDATE(ovp, NULL, NULL, UPDATE_WAIT);
if (error && !allerror)
allerror = error;
/*
* Having written the new inode to disk, save its new configuration
* and put back the old block pointers long enough to process them.
* Note that we save the new block configuration so we can check it
* when we are done.
*/
for (i = 0; i < NDADDR; i++) {
newblks[i] = DIP(oip, db[i]);
DIP(oip, db[i]) = oldblks[i];
}
for (i = 0; i < NIADDR; i++) {
newblks[NDADDR + i] = DIP(oip, ib[i]);
DIP(oip, ib[i]) = oldblks[NDADDR + i];
}
if (oip->i_ump->um_fstype != UFS2 &&
oip->i_fs->fs_magic == FS_UFS2_MAGIC)
panic("fstype inconsistency (%ld)", oip->i_ump->um_fstype);
oip->i_size = osize;
DIP(oip, size) = osize;
error = vtruncbuf(ovp, lastblock + 1, 0, 0);
if (error && !allerror)
allerror = error;
/*
* Indirect blocks first.
*/
indir_lbn[SINGLE] = -NDADDR;
indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) - 1;
indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1;
for (level = TRIPLE; level >= SINGLE; level--) {
if (oip->i_ump->um_fstype == UFS1)
bn = ufs_rw32(oip->i_ffs1_ib[level],UFS_FSNEEDSWAP(fs));
else
bn = ufs_rw64(oip->i_ffs2_ib[level],UFS_FSNEEDSWAP(fs));
if (bn != 0) {
error = ffs_indirtrunc(oip, indir_lbn[level],
fsbtodb(fs, bn), lastiblock[level], level, &count);
if (error)
allerror = error;
blocksreleased += count;
if (lastiblock[level] < 0) {
DIP(oip, ib[level]) = 0;
ffs_blkfree(oip, bn, fs->fs_bsize);
blocksreleased += nblocks;
}
}
if (lastiblock[level] >= 0)
goto done;
}
/*
* All whole direct blocks or frags.
*/
for (i = NDADDR - 1; i > lastblock; i--) {
long bsize;
if (oip->i_ump->um_fstype == UFS1)
bn = ufs_rw32(oip->i_ffs1_db[i], UFS_FSNEEDSWAP(fs));
else
bn = ufs_rw64(oip->i_ffs2_db[i], UFS_FSNEEDSWAP(fs));
if (bn == 0)
continue;
DIP(oip, db[i]) = 0;
bsize = blksize(fs, oip, i);
ffs_blkfree(oip, bn, bsize);
blocksreleased += btodb(bsize);
}
if (lastblock < 0)
goto done;
/*
* Finally, look for a change in size of the
* last direct block; release any frags.
*/
if (oip->i_ump->um_fstype == UFS1)
bn = ufs_rw32(oip->i_ffs1_db[lastblock], UFS_FSNEEDSWAP(fs));
else
bn = ufs_rw64(oip->i_ffs2_db[lastblock], UFS_FSNEEDSWAP(fs));
if (bn != 0) {
long oldspace, newspace;
/*
* Calculate amount of space we're giving
* back as old block size minus new block size.
*/
oldspace = blksize(fs, oip, lastblock);
if (oip->i_ump->um_fstype != UFS2 &&
oip->i_fs->fs_magic == FS_UFS2_MAGIC)
panic("fstype inconsistency (%ld)", oip->i_ump->um_fstype);
oip->i_size = length;
DIP(oip, size) = length;
newspace = blksize(fs, oip, lastblock);
if (newspace == 0)
panic("itrunc: newspace");
if (oldspace - newspace > 0) {
/*
* Block number of space to be free'd is
* the old block # plus the number of frags
* required for the storage we're keeping.
*/
bn += numfrags(fs, newspace);
ffs_blkfree(oip, bn, oldspace - newspace);
blocksreleased += btodb(oldspace - newspace);
}
}
done:
#ifdef DIAGNOSTIC
for (level = SINGLE; level <= TRIPLE; level++)
if (newblks[NDADDR + level] != DIP(oip, ib[level]))
panic("itrunc1");
for (i = 0; i < NDADDR; i++)
if (newblks[i] != DIP(oip, db[i]))
panic("itrunc2");
if (length == 0 &&
(!LIST_EMPTY(&ovp->v_cleanblkhd) || !LIST_EMPTY(&ovp->v_dirtyblkhd)))
panic("itrunc3");
#endif /* DIAGNOSTIC */
/*
* Put back the real size.
*/
if (oip->i_ump->um_fstype != UFS2 &&
oip->i_fs->fs_magic == FS_UFS2_MAGIC)
panic("fstype inconsistency (%ld)", oip->i_ump->um_fstype);
oip->i_size = length;
DIP(oip, size) = length;
DIP(oip, blocks) -= blocksreleased;
lockmgr(&gp->g_glock, LK_RELEASE, NULL);
oip->i_flag |= IN_CHANGE;
#ifdef QUOTA
(void) chkdq(oip, -blocksreleased, NOCRED, 0);
#endif
KASSERT(ovp->v_type != VREG || ovp->v_size == oip->i_size);
return (allerror);
}
/*
* Release blocks associated with the inode ip and stored in the indirect
* block bn. Blocks are free'd in LIFO order up to (but not including)
* lastbn. If level is greater than SINGLE, the block is an indirect block
* and recursive calls to indirtrunc must be used to cleanse other indirect
* blocks.
*
* NB: triple indirect blocks are untested.
*/
static int
ffs_indirtrunc(ip, lbn, dbn, lastbn, level, countp)
struct inode *ip;
daddr_t lbn, lastbn;
daddr_t dbn;
int level;
int64_t *countp;
{
int i;
struct buf *bp;
struct fs *fs = ip->i_fs;
int32_t *bap1 = NULL;
int64_t *bap2 = NULL;
struct vnode *vp;
daddr_t nb, nlbn, last;
char *copy = NULL;
int64_t blkcount, factor, blocksreleased = 0;
int nblocks;
int error = 0, allerror = 0;
#ifdef FFS_EI
const int needswap = UFS_FSNEEDSWAP(fs);
#endif
#define RBAP(ip, i) (((ip)->i_ump->um_fstype == UFS1) ? \
ufs_rw32(bap1[i], needswap) : ufs_rw64(bap2[i], needswap))
#define BAP(ip, i) (((ip)->i_ump->um_fstype == UFS1) ? bap1[i] : bap2[i])
/*
* Calculate index in current block of last
* block to be kept. -1 indicates the entire
* block so we need not calculate the index.
*/
factor = 1;
for (i = SINGLE; i < level; i++)
factor *= NINDIR(fs);
last = lastbn;
if (lastbn > 0)
last /= factor;
nblocks = btodb(fs->fs_bsize);
/*
* Get buffer of block pointers, zero those entries corresponding
* to blocks to be free'd, and update on disk copy first. Since
* double(triple) indirect before single(double) indirect, calls
* to bmap on these blocks will fail. However, we already have
* the on disk address, so we have to set the b_blkno field
* explicitly instead of letting bread do everything for us.
*/
vp = ITOV(ip);
bp = getblk(vp, lbn, (int)fs->fs_bsize, 0, 0);
if (bp->b_flags & (B_DONE | B_DELWRI)) {
/* Braces must be here in case trace evaluates to nothing. */
trace(TR_BREADHIT, pack(vp, fs->fs_bsize), lbn);
} else {
trace(TR_BREADMISS, pack(vp, fs->fs_bsize), lbn);
curproc->p_stats->p_ru.ru_inblock++; /* pay for read */
bp->b_flags |= B_READ;
if (bp->b_bcount > bp->b_bufsize)
panic("ffs_indirtrunc: bad buffer size");
bp->b_blkno = dbn;
VOP_STRATEGY(bp);
error = biowait(bp);
}
if (error) {
brelse(bp);
*countp = 0;
return (error);
}
if (ip->i_ump->um_fstype == UFS1)
bap1 = (int32_t *)bp->b_data;
else
bap2 = (int64_t *)bp->b_data;
if (lastbn >= 0) {
copy = malloc(fs->fs_bsize, M_TEMP, M_WAITOK);
memcpy((caddr_t)copy, bp->b_data, (u_int)fs->fs_bsize);
for (i = last + 1; i < NINDIR(fs); i++)
BAP(ip, i) = 0;
error = bwrite(bp);
if (error)
allerror = error;
if (ip->i_ump->um_fstype == UFS1)
bap1 = (int32_t *)copy;
else
bap2 = (int64_t *)copy;
}
/*
* Recursively free totally unused blocks.
*/
for (i = NINDIR(fs) - 1, nlbn = lbn + 1 - i * factor; i > last;
i--, nlbn += factor) {
nb = RBAP(ip, i);
if (nb == 0)
continue;
if (level > SINGLE) {
error = ffs_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
(daddr_t)-1, level - 1,
&blkcount);
if (error)
allerror = error;
blocksreleased += blkcount;
}
ffs_blkfree(ip, nb, fs->fs_bsize);
blocksreleased += nblocks;
}
/*
* Recursively free last partial block.
*/
if (level > SINGLE && lastbn >= 0) {
last = lastbn % factor;
nb = RBAP(ip, i);
if (nb != 0) {
error = ffs_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
last, level - 1, &blkcount);
if (error)
allerror = error;
blocksreleased += blkcount;
}
}
if (copy != NULL) {
FREE(copy, M_TEMP);
} else {
bp->b_flags |= B_INVAL;
brelse(bp);
}
*countp = blocksreleased;
return (allerror);
}