bd7e5f992e
Speed up for vfs_bio -- addition of a routine bqrelse to greatly diminish overhead for merged cache. Efficiency improvement for vfs_cluster. It used to do alot of redundant calls to cluster_rbuild. Correct the ordering for vrele of .text and release of credentials. Use the selective tlb update for 486/586/P6. Numerous fixes to the size of objects allocated for files. Additionally, fixes in the various pagers. Fixes for proper positioning of vnode_pager_setsize in msdosfs and ext2fs. Fixes in the swap pager for exhausted resources. The pageout code will not as readily thrash. Change the page queue flags (PG_ACTIVE, PG_INACTIVE, PG_FREE, PG_CACHE) into page queue indices (PQ_ACTIVE, PQ_INACTIVE, PQ_FREE, PQ_CACHE), thereby improving efficiency of several routines. Eliminate even more unnecessary vm_page_protect operations. Significantly speed up process forks. Make vm_object_page_clean more efficient, thereby eliminating the pause that happens every 30seconds. Make sequential clustered writes B_ASYNC instead of B_DELWRI even in the case of filesystems mounted async. Fix a panic with busy pages when write clustering is done for non-VMIO buffers.
523 lines
15 KiB
C
523 lines
15 KiB
C
/*
|
|
* 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 8.5 (Berkeley) 12/30/93
|
|
* $Id: ffs_inode.c,v 1.19 1996/01/05 18:31:48 wollman Exp $
|
|
*/
|
|
|
|
#include "opt_quota.h"
|
|
|
|
#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/resourcevar.h>
|
|
|
|
#include <vm/vm.h>
|
|
#include <vm/vm_param.h>
|
|
#include <vm/vm_extern.h>
|
|
|
|
#include <ufs/ufs/quota.h>
|
|
#include <ufs/ufs/inode.h>
|
|
#include <ufs/ufs/ufsmount.h>
|
|
#include <ufs/ufs/ufs_extern.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,
|
|
long *));
|
|
|
|
int
|
|
ffs_init()
|
|
{
|
|
return (ufs_init());
|
|
}
|
|
|
|
/*
|
|
* 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 even if none
|
|
* of the times needs to be updated. 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 waitfor is set, then wait for the disk
|
|
* write of the inode to complete.
|
|
*/
|
|
int
|
|
ffs_update(ap)
|
|
struct vop_update_args /* {
|
|
struct vnode *a_vp;
|
|
struct timeval *a_access;
|
|
struct timeval *a_modify;
|
|
int a_waitfor;
|
|
} */ *ap;
|
|
{
|
|
register struct fs *fs;
|
|
struct buf *bp;
|
|
struct inode *ip;
|
|
int error;
|
|
time_t tv_sec;
|
|
|
|
ip = VTOI(ap->a_vp);
|
|
if (ap->a_vp->v_mount->mnt_flag & MNT_RDONLY) {
|
|
ip->i_flag &=
|
|
~(IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE);
|
|
return (0);
|
|
}
|
|
if ((ip->i_flag &
|
|
(IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0)
|
|
return (0);
|
|
/*
|
|
* Use a copy of the current time to get consistent timestamps
|
|
* (a_access and a_modify are sometimes aliases for &time).
|
|
*
|
|
* XXX in 2.0, a_access and a_modify are often pointers to the
|
|
* same copy of `time'. This is not as good. Some callers forget
|
|
* to make a copy; others make a copy too early (before the i/o
|
|
* has completed)...
|
|
*
|
|
* XXX there should be a function or macro for reading the time
|
|
* (e.g., some machines may require splclock()).
|
|
*/
|
|
tv_sec = time.tv_sec;
|
|
if (ip->i_flag & IN_ACCESS)
|
|
ip->i_atime.ts_sec =
|
|
(ap->a_access == &time ? tv_sec : ap->a_access->tv_sec);
|
|
if (ip->i_flag & IN_UPDATE) {
|
|
ip->i_mtime.ts_sec =
|
|
(ap->a_modify == &time ? tv_sec : ap->a_modify->tv_sec);
|
|
ip->i_modrev++;
|
|
}
|
|
if (ip->i_flag & IN_CHANGE)
|
|
ip->i_ctime.ts_sec = tv_sec;
|
|
ip->i_flag &= ~(IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE);
|
|
fs = ip->i_fs;
|
|
/*
|
|
* Ensure that uid and gid are correct. This is a temporary
|
|
* fix until fsck has been changed to do the update.
|
|
*/
|
|
if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
|
|
ip->i_din.di_ouid = ip->i_uid; /* XXX */
|
|
ip->i_din.di_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);
|
|
}
|
|
*((struct dinode *)bp->b_data +
|
|
ino_to_fsbo(fs, ip->i_number)) = ip->i_din;
|
|
if (ap->a_waitfor && (ap->a_vp->v_mount->mnt_flag & MNT_ASYNC) == 0)
|
|
return (bwrite(bp));
|
|
else {
|
|
bp->b_flags |= B_CLUSTEROK;
|
|
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(ap)
|
|
struct vop_truncate_args /* {
|
|
struct vnode *a_vp;
|
|
off_t a_length;
|
|
int a_flags;
|
|
struct ucred *a_cred;
|
|
struct proc *a_p;
|
|
} */ *ap;
|
|
{
|
|
register struct vnode *ovp = ap->a_vp;
|
|
register daddr_t lastblock;
|
|
register struct inode *oip;
|
|
daddr_t bn, lbn, lastiblock[NIADDR], indir_lbn[NIADDR];
|
|
daddr_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR];
|
|
off_t length = ap->a_length;
|
|
register struct fs *fs;
|
|
struct buf *bp;
|
|
int offset, size, level;
|
|
long count, nblocks, vflags, blocksreleased = 0;
|
|
struct timeval tv;
|
|
register int i;
|
|
int aflags, error, allerror;
|
|
off_t osize;
|
|
|
|
oip = VTOI(ovp);
|
|
fs = oip->i_fs;
|
|
if (length < 0 || length > fs->fs_maxfilesize)
|
|
return (EINVAL);
|
|
tv = time;
|
|
if (ovp->v_type == VLNK &&
|
|
(oip->i_size < ovp->v_mount->mnt_maxsymlinklen || oip->i_din.di_blocks == 0)) {
|
|
#ifdef DIAGNOSTIC
|
|
if (length != 0)
|
|
panic("ffs_truncate: partial truncate of symlink");
|
|
#endif
|
|
bzero((char *)&oip->i_shortlink, (u_int)oip->i_size);
|
|
oip->i_size = 0;
|
|
oip->i_flag |= IN_CHANGE | IN_UPDATE;
|
|
return (VOP_UPDATE(ovp, &tv, &tv, 1));
|
|
}
|
|
if (oip->i_size == length) {
|
|
oip->i_flag |= IN_CHANGE | IN_UPDATE;
|
|
return (VOP_UPDATE(ovp, &tv, &tv, 0));
|
|
}
|
|
#ifdef QUOTA
|
|
error = getinoquota(oip);
|
|
if (error)
|
|
return (error);
|
|
#endif
|
|
osize = oip->i_size;
|
|
/*
|
|
* 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) {
|
|
offset = blkoff(fs, length - 1);
|
|
lbn = lblkno(fs, length - 1);
|
|
aflags = B_CLRBUF;
|
|
if (ap->a_flags & IO_SYNC)
|
|
aflags |= B_SYNC;
|
|
vnode_pager_setsize(ovp, length);
|
|
error = ffs_balloc(oip, lbn, offset + 1, ap->a_cred,
|
|
&bp, aflags);
|
|
if (error)
|
|
return (error);
|
|
oip->i_size = length;
|
|
if (aflags & B_SYNC)
|
|
bwrite(bp);
|
|
else if (ovp->v_mount->mnt_flag & MNT_ASYNC)
|
|
bdwrite(bp);
|
|
else
|
|
bawrite(bp);
|
|
oip->i_flag |= IN_CHANGE | IN_UPDATE;
|
|
return (VOP_UPDATE(ovp, &tv, &tv, 1));
|
|
}
|
|
/*
|
|
* Shorten the size of the file. If the file is not being
|
|
* truncated to a block boundry, the contents of the
|
|
* partial block following the end of the file must be
|
|
* zero'ed in case it ever become accessable again because
|
|
* of subsequent file growth.
|
|
*/
|
|
offset = blkoff(fs, length);
|
|
if (offset == 0) {
|
|
oip->i_size = length;
|
|
} else {
|
|
lbn = lblkno(fs, length);
|
|
aflags = B_CLRBUF;
|
|
if (ap->a_flags & IO_SYNC)
|
|
aflags |= B_SYNC;
|
|
error = ffs_balloc(oip, lbn, offset, ap->a_cred, &bp, aflags);
|
|
if (error)
|
|
return (error);
|
|
oip->i_size = length;
|
|
size = blksize(fs, oip, lbn);
|
|
bzero((char *)bp->b_data + offset, (u_int)(size - offset));
|
|
allocbuf(bp, size);
|
|
if (aflags & B_SYNC)
|
|
bwrite(bp);
|
|
else if (ovp->v_mount->mnt_flag & MNT_ASYNC)
|
|
bdwrite(bp);
|
|
else
|
|
bawrite(bp);
|
|
}
|
|
/*
|
|
* 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.
|
|
*/
|
|
bcopy((caddr_t)&oip->i_db[0], (caddr_t)oldblks, sizeof oldblks);
|
|
for (level = TRIPLE; level >= SINGLE; level--)
|
|
if (lastiblock[level] < 0) {
|
|
oip->i_ib[level] = 0;
|
|
lastiblock[level] = -1;
|
|
}
|
|
for (i = NDADDR - 1; i > lastblock; i--)
|
|
oip->i_db[i] = 0;
|
|
oip->i_flag |= IN_CHANGE | IN_UPDATE;
|
|
error = VOP_UPDATE(ovp, &tv, &tv, ((length > 0) ? 0 : 1));
|
|
if (error)
|
|
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.
|
|
*/
|
|
bcopy((caddr_t)&oip->i_db[0], (caddr_t)newblks, sizeof newblks);
|
|
bcopy((caddr_t)oldblks, (caddr_t)&oip->i_db[0], sizeof oldblks);
|
|
oip->i_size = osize;
|
|
vflags = ((length > 0) ? V_SAVE : 0) | V_SAVEMETA;
|
|
allerror = vinvalbuf(ovp, vflags, ap->a_cred, ap->a_p, 0, 0);
|
|
|
|
/*
|
|
* 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--) {
|
|
bn = oip->i_ib[level];
|
|
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) {
|
|
oip->i_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--) {
|
|
register long bsize;
|
|
|
|
bn = oip->i_db[i];
|
|
if (bn == 0)
|
|
continue;
|
|
oip->i_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.
|
|
*/
|
|
bn = oip->i_db[lastblock];
|
|
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);
|
|
oip->i_size = length;
|
|
newspace = blksize(fs, oip, lastblock);
|
|
if (newspace == 0)
|
|
panic("ffs_truncate: 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] != oip->i_ib[level])
|
|
panic("ffs_truncate1");
|
|
for (i = 0; i < NDADDR; i++)
|
|
if (newblks[i] != oip->i_db[i])
|
|
panic("ffs_truncate2");
|
|
if (length == 0 &&
|
|
(ovp->v_dirtyblkhd.lh_first || ovp->v_cleanblkhd.lh_first))
|
|
panic("ffs_truncate3");
|
|
#endif /* DIAGNOSTIC */
|
|
/*
|
|
* Put back the real size.
|
|
*/
|
|
oip->i_size = length;
|
|
oip->i_blocks -= blocksreleased;
|
|
if (oip->i_blocks < 0) /* sanity */
|
|
oip->i_blocks = 0;
|
|
oip->i_flag |= IN_CHANGE;
|
|
vnode_pager_setsize(ovp, length);
|
|
#ifdef QUOTA
|
|
(void) chkdq(oip, -blocksreleased, NOCRED, 0);
|
|
#endif
|
|
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)
|
|
register struct inode *ip;
|
|
daddr_t lbn, lastbn;
|
|
daddr_t dbn;
|
|
int level;
|
|
long *countp;
|
|
{
|
|
register int i;
|
|
struct buf *bp;
|
|
register struct fs *fs = ip->i_fs;
|
|
register daddr_t *bap;
|
|
struct vnode *vp;
|
|
daddr_t *copy, nb, nlbn, last;
|
|
long blkcount, factor;
|
|
int nblocks, blocksreleased = 0;
|
|
int error = 0, allerror = 0;
|
|
|
|
/*
|
|
* 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_CACHE) == 0) {
|
|
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;
|
|
vfs_busy_pages(bp, 0);
|
|
VOP_STRATEGY(bp);
|
|
error = biowait(bp);
|
|
}
|
|
if (error) {
|
|
brelse(bp);
|
|
*countp = 0;
|
|
return (error);
|
|
}
|
|
|
|
bap = (daddr_t *)bp->b_data;
|
|
MALLOC(copy, daddr_t *, fs->fs_bsize, M_TEMP, M_WAITOK);
|
|
bcopy((caddr_t)bap, (caddr_t)copy, (u_int)fs->fs_bsize);
|
|
bzero((caddr_t)&bap[last + 1],
|
|
(u_int)(NINDIR(fs) - (last + 1)) * sizeof (daddr_t));
|
|
if (last == -1)
|
|
bp->b_flags |= B_INVAL;
|
|
if ((vp->v_mount->mnt_flag & MNT_ASYNC) == 0) {
|
|
error = bwrite(bp);
|
|
} else {
|
|
bawrite(bp);
|
|
error = 0;
|
|
}
|
|
if (error)
|
|
allerror = error;
|
|
bap = copy;
|
|
|
|
/*
|
|
* Recursively free totally unused blocks.
|
|
*/
|
|
for (i = NINDIR(fs) - 1, nlbn = lbn + 1 - i * factor; i > last;
|
|
i--, nlbn += factor) {
|
|
nb = bap[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 = bap[i];
|
|
if (nb != 0) {
|
|
error = ffs_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
|
|
last, level - 1, &blkcount);
|
|
if (error)
|
|
allerror = error;
|
|
blocksreleased += blkcount;
|
|
}
|
|
}
|
|
FREE(copy, M_TEMP);
|
|
*countp = blocksreleased;
|
|
return (allerror);
|
|
}
|