d/freebsd-skq
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
9f37ee804a
freebsd-skq/sys/ufs/ffs/ffs_alloc.c
Kirk McKusick aa7ddc85c7 When a file is first being written, the dynamic block reallocation 
(implemented by ffs_reallocblks_ufs[12]) relocates the file's blocks
so as to cluster them together into a contiguous set of blocks on
the disk.

When the cluster crosses the boundary into the first indirect block,
the first indirect block is initially allocated in a position
immediately following the last direct block.  Block reallocation
would usually destroy locality by moving the indirect block out of
the way to keep the data blocks contiguous.  This change compensates
for this problem by noting that the first indirect block should be
left immediately following the last direct block.  It then tries
to start a new cluster of contiguous blocks (referenced by the
indirect block) immediately following the indirect block.

We should also do this for other indirect block boundaries, but it
is only important for the first one.

Suggested by: Bruce Evans
MFC:          2 weeks
2012-11-03 18:55:55 +00:00

2958 lines
81 KiB
C
Raw Blame History

/*-
* Copyright (c) 2002 Networks Associates Technology, Inc.
* All rights reserved.
*
* This software was developed for the FreeBSD Project by Marshall
* Kirk McKusick and Network Associates Laboratories, the Security
* Research Division of Network Associates, Inc. under DARPA/SPAWAR
* contract N66001-01-C-8035 ("CBOSS"), as part of the DARPA CHATS
* research program
*
* 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.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR 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 AUTHOR 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.
*
* 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.
* 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_alloc.c 8.18 (Berkeley) 5/26/95
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include "opt_quota.h"
#include <sys/param.h>
#include <sys/capability.h>
#include <sys/systm.h>
#include <sys/bio.h>
#include <sys/buf.h>
#include <sys/conf.h>
#include <sys/fcntl.h>
#include <sys/file.h>
#include <sys/filedesc.h>
#include <sys/priv.h>
#include <sys/proc.h>
#include <sys/vnode.h>
#include <sys/mount.h>
#include <sys/kernel.h>
#include <sys/syscallsubr.h>
#include <sys/sysctl.h>
#include <sys/syslog.h>
#include <sys/taskqueue.h>
#include <security/audit/audit.h>
#include <geom/geom.h>
#include <ufs/ufs/dir.h>
#include <ufs/ufs/extattr.h>
#include <ufs/ufs/quota.h>
#include <ufs/ufs/inode.h>
#include <ufs/ufs/ufs_extern.h>
#include <ufs/ufs/ufsmount.h>
#include <ufs/ffs/fs.h>
#include <ufs/ffs/ffs_extern.h>
#include <ufs/ffs/softdep.h>
typedef ufs2_daddr_t allocfcn_t(struct inode *ip, u_int cg, ufs2_daddr_t bpref,
int size, int rsize);
static ufs2_daddr_t ffs_alloccg(struct inode *, u_int, ufs2_daddr_t, int, int);
static ufs2_daddr_t
ffs_alloccgblk(struct inode *, struct buf *, ufs2_daddr_t, int);
static void ffs_blkfree_cg(struct ufsmount *, struct fs *,
struct vnode *, ufs2_daddr_t, long, ino_t,
struct workhead *);
static void ffs_blkfree_trim_completed(struct bio *);
static void ffs_blkfree_trim_task(void *ctx, int pending __unused);
#ifdef INVARIANTS
static int ffs_checkblk(struct inode *, ufs2_daddr_t, long);
#endif
static ufs2_daddr_t ffs_clusteralloc(struct inode *, u_int, ufs2_daddr_t, int,
int);
static ino_t ffs_dirpref(struct inode *);
static ufs2_daddr_t ffs_fragextend(struct inode *, u_int, ufs2_daddr_t,
int, int);
static ufs2_daddr_t ffs_hashalloc
(struct inode *, u_int, ufs2_daddr_t, int, int, allocfcn_t *);
static ufs2_daddr_t ffs_nodealloccg(struct inode *, u_int, ufs2_daddr_t, int,
int);
static ufs1_daddr_t ffs_mapsearch(struct fs *, struct cg *, ufs2_daddr_t, int);
static int ffs_reallocblks_ufs1(struct vop_reallocblks_args *);
static int ffs_reallocblks_ufs2(struct vop_reallocblks_args *);
/*
* Allocate a block in the filesystem.
*
* The size of the requested block is given, which must be some
* multiple of fs_fsize and <= fs_bsize.
* A preference may be optionally specified. If a preference is given
* the following hierarchy is used to allocate a block:
* 1) allocate the requested block.
* 2) allocate a rotationally optimal block in the same cylinder.
* 3) allocate a block in the same cylinder group.
* 4) quadradically rehash into other cylinder groups, until an
* available block is located.
* If no block preference is given the following hierarchy is used
* to allocate a block:
* 1) allocate a block in the cylinder group that contains the
* inode for the file.
* 2) quadradically rehash into other cylinder groups, until an
* available block is located.
*/
int
ffs_alloc(ip, lbn, bpref, size, flags, cred, bnp)
struct inode *ip;
ufs2_daddr_t lbn, bpref;
int size, flags;
struct ucred *cred;
ufs2_daddr_t *bnp;
{
struct fs *fs;
struct ufsmount *ump;
ufs2_daddr_t bno;
u_int cg, reclaimed;
static struct timeval lastfail;
static int curfail;
int64_t delta;
#ifdef QUOTA
int error;
#endif
*bnp = 0;
fs = ip->i_fs;
ump = ip->i_ump;
mtx_assert(UFS_MTX(ump), MA_OWNED);
#ifdef INVARIANTS
if ((u_int)size > fs->fs_bsize || fragoff(fs, size) != 0) {
printf("dev = %s, bsize = %ld, size = %d, fs = %s\n",
devtoname(ip->i_dev), (long)fs->fs_bsize, size,
fs->fs_fsmnt);
panic("ffs_alloc: bad size");
}
if (cred == NOCRED)
panic("ffs_alloc: missing credential");
#endif /* INVARIANTS */
reclaimed = 0;
retry:
#ifdef QUOTA
UFS_UNLOCK(ump);
error = chkdq(ip, btodb(size), cred, 0);
if (error)
return (error);
UFS_LOCK(ump);
#endif
if (size == fs->fs_bsize && fs->fs_cstotal.cs_nbfree == 0)
goto nospace;
if (priv_check_cred(cred, PRIV_VFS_BLOCKRESERVE, 0) &&
freespace(fs, fs->fs_minfree) - numfrags(fs, size) < 0)
goto nospace;
if (bpref >= fs->fs_size)
bpref = 0;
if (bpref == 0)
cg = ino_to_cg(fs, ip->i_number);
else
cg = dtog(fs, bpref);
bno = ffs_hashalloc(ip, cg, bpref, size, size, ffs_alloccg);
if (bno > 0) {
delta = btodb(size);
DIP_SET(ip, i_blocks, DIP(ip, i_blocks) + delta);
if (flags & IO_EXT)
ip->i_flag |= IN_CHANGE;
else
ip->i_flag |= IN_CHANGE | IN_UPDATE;
*bnp = bno;
return (0);
}
nospace:
#ifdef QUOTA
UFS_UNLOCK(ump);
/*
* Restore user's disk quota because allocation failed.
*/
(void) chkdq(ip, -btodb(size), cred, FORCE);
UFS_LOCK(ump);
#endif
if (reclaimed == 0 && (flags & IO_BUFLOCKED) == 0) {
reclaimed = 1;
softdep_request_cleanup(fs, ITOV(ip), cred, FLUSH_BLOCKS_WAIT);
goto retry;
}
UFS_UNLOCK(ump);
if (reclaimed > 0 && ppsratecheck(&lastfail, &curfail, 1)) {
ffs_fserr(fs, ip->i_number, "filesystem full");
uprintf("\n%s: write failed, filesystem is full\n",
fs->fs_fsmnt);
}
return (ENOSPC);
}
/*
* Reallocate a fragment to a bigger size
*
* The number and size of the old block is given, and a preference
* and new size is also specified. The allocator attempts to extend
* the original block. Failing that, the regular block allocator is
* invoked to get an appropriate block.
*/
int
ffs_realloccg(ip, lbprev, bprev, bpref, osize, nsize, flags, cred, bpp)
struct inode *ip;
ufs2_daddr_t lbprev;
ufs2_daddr_t bprev;
ufs2_daddr_t bpref;
int osize, nsize, flags;
struct ucred *cred;
struct buf **bpp;
{
struct vnode *vp;
struct fs *fs;
struct buf *bp;
struct ufsmount *ump;
u_int cg, request, reclaimed;
int error;
ufs2_daddr_t bno;
static struct timeval lastfail;
static int curfail;
int64_t delta;
*bpp = 0;
vp = ITOV(ip);
fs = ip->i_fs;
bp = NULL;
ump = ip->i_ump;
mtx_assert(UFS_MTX(ump), MA_OWNED);
#ifdef INVARIANTS
if (vp->v_mount->mnt_kern_flag & MNTK_SUSPENDED)
panic("ffs_realloccg: allocation on suspended filesystem");
if ((u_int)osize > fs->fs_bsize || fragoff(fs, osize) != 0 ||
(u_int)nsize > fs->fs_bsize || fragoff(fs, nsize) != 0) {
printf(
"dev = %s, bsize = %ld, osize = %d, nsize = %d, fs = %s\n",
devtoname(ip->i_dev), (long)fs->fs_bsize, osize,
nsize, fs->fs_fsmnt);
panic("ffs_realloccg: bad size");
}
if (cred == NOCRED)
panic("ffs_realloccg: missing credential");
#endif /* INVARIANTS */
reclaimed = 0;
retry:
if (priv_check_cred(cred, PRIV_VFS_BLOCKRESERVE, 0) &&
freespace(fs, fs->fs_minfree) - numfrags(fs, nsize - osize) < 0) {
goto nospace;
}
if (bprev == 0) {
printf("dev = %s, bsize = %ld, bprev = %jd, fs = %s\n",
devtoname(ip->i_dev), (long)fs->fs_bsize, (intmax_t)bprev,
fs->fs_fsmnt);
panic("ffs_realloccg: bad bprev");
}
UFS_UNLOCK(ump);
/*
* Allocate the extra space in the buffer.
*/
error = bread(vp, lbprev, osize, NOCRED, &bp);
if (error) {
brelse(bp);
return (error);
}
if (bp->b_blkno == bp->b_lblkno) {
if (lbprev >= NDADDR)
panic("ffs_realloccg: lbprev out of range");
bp->b_blkno = fsbtodb(fs, bprev);
}
#ifdef QUOTA
error = chkdq(ip, btodb(nsize - osize), cred, 0);
if (error) {
brelse(bp);
return (error);
}
#endif
/*
* Check for extension in the existing location.
*/
cg = dtog(fs, bprev);
UFS_LOCK(ump);
bno = ffs_fragextend(ip, cg, bprev, osize, nsize);
if (bno) {
if (bp->b_blkno != fsbtodb(fs, bno))
panic("ffs_realloccg: bad blockno");
delta = btodb(nsize - osize);
DIP_SET(ip, i_blocks, DIP(ip, i_blocks) + delta);
if (flags & IO_EXT)
ip->i_flag |= IN_CHANGE;
else
ip->i_flag |= IN_CHANGE | IN_UPDATE;
allocbuf(bp, nsize);
bp->b_flags |= B_DONE;
bzero(bp->b_data + osize, nsize - osize);
if ((bp->b_flags & (B_MALLOC | B_VMIO)) == B_VMIO)
vfs_bio_set_valid(bp, osize, nsize - osize);
*bpp = bp;
return (0);
}
/*
* Allocate a new disk location.
*/
if (bpref >= fs->fs_size)
bpref = 0;
switch ((int)fs->fs_optim) {
case FS_OPTSPACE:
/*
* Allocate an exact sized fragment. Although this makes
* best use of space, we will waste time relocating it if
* the file continues to grow. If the fragmentation is
* less than half of the minimum free reserve, we choose
* to begin optimizing for time.
*/
request = nsize;
if (fs->fs_minfree <= 5 ||
fs->fs_cstotal.cs_nffree >
(off_t)fs->fs_dsize * fs->fs_minfree / (2 * 100))
break;
log(LOG_NOTICE, "%s: optimization changed from SPACE to TIME\n",
fs->fs_fsmnt);
fs->fs_optim = FS_OPTTIME;
break;
case FS_OPTTIME:
/*
* At this point we have discovered a file that is trying to
* grow a small fragment to a larger fragment. To save time,
* we allocate a full sized block, then free the unused portion.
* If the file continues to grow, the `ffs_fragextend' call
* above will be able to grow it in place without further
* copying. If aberrant programs cause disk fragmentation to
* grow within 2% of the free reserve, we choose to begin
* optimizing for space.
*/
request = fs->fs_bsize;
if (fs->fs_cstotal.cs_nffree <
(off_t)fs->fs_dsize * (fs->fs_minfree - 2) / 100)
break;
log(LOG_NOTICE, "%s: optimization changed from TIME to SPACE\n",
fs->fs_fsmnt);
fs->fs_optim = FS_OPTSPACE;
break;
default:
printf("dev = %s, optim = %ld, fs = %s\n",
devtoname(ip->i_dev), (long)fs->fs_optim, fs->fs_fsmnt);
panic("ffs_realloccg: bad optim");
/* NOTREACHED */
}
bno = ffs_hashalloc(ip, cg, bpref, request, nsize, ffs_alloccg);
if (bno > 0) {
bp->b_blkno = fsbtodb(fs, bno);
if (!DOINGSOFTDEP(vp))
ffs_blkfree(ump, fs, ip->i_devvp, bprev, (long)osize,
ip->i_number, vp->v_type, NULL);
delta = btodb(nsize - osize);
DIP_SET(ip, i_blocks, DIP(ip, i_blocks) + delta);
if (flags & IO_EXT)
ip->i_flag |= IN_CHANGE;
else
ip->i_flag |= IN_CHANGE | IN_UPDATE;
allocbuf(bp, nsize);
bp->b_flags |= B_DONE;
bzero(bp->b_data + osize, nsize - osize);
if ((bp->b_flags & (B_MALLOC | B_VMIO)) == B_VMIO)
vfs_bio_set_valid(bp, osize, nsize - osize);
*bpp = bp;
return (0);
}
#ifdef QUOTA
UFS_UNLOCK(ump);
/*
* Restore user's disk quota because allocation failed.
*/
(void) chkdq(ip, -btodb(nsize - osize), cred, FORCE);
UFS_LOCK(ump);
#endif
nospace:
/*
* no space available
*/
if (reclaimed == 0 && (flags & IO_BUFLOCKED) == 0) {
reclaimed = 1;
UFS_UNLOCK(ump);
if (bp) {
brelse(bp);
bp = NULL;
}
UFS_LOCK(ump);
softdep_request_cleanup(fs, vp, cred, FLUSH_BLOCKS_WAIT);
goto retry;
}
UFS_UNLOCK(ump);
if (bp)
brelse(bp);
if (reclaimed > 0 && ppsratecheck(&lastfail, &curfail, 1)) {
ffs_fserr(fs, ip->i_number, "filesystem full");
uprintf("\n%s: write failed, filesystem is full\n",
fs->fs_fsmnt);
}
return (ENOSPC);
}
/*
* Reallocate a sequence of blocks into a contiguous sequence of blocks.
*
* The vnode and an array of buffer pointers for a range of sequential
* logical blocks to be made contiguous is given. The allocator attempts
* to find a range of sequential blocks starting as close as possible
* from the end of the allocation for the logical block immediately
* preceding the current range. If successful, the physical block numbers
* in the buffer pointers and in the inode are changed to reflect the new
* allocation. If unsuccessful, the allocation is left unchanged. The
* success in doing the reallocation is returned. Note that the error
* return is not reflected back to the user. Rather the previous block
* allocation will be used.
*/
SYSCTL_NODE(_vfs, OID_AUTO, ffs, CTLFLAG_RW, 0, "FFS filesystem");
static int doasyncfree = 1;
SYSCTL_INT(_vfs_ffs, OID_AUTO, doasyncfree, CTLFLAG_RW, &doasyncfree, 0, "");
static int doreallocblks = 1;
SYSCTL_INT(_vfs_ffs, OID_AUTO, doreallocblks, CTLFLAG_RW, &doreallocblks, 0, "");
#ifdef DEBUG
static volatile int prtrealloc = 0;
#endif
int
ffs_reallocblks(ap)
struct vop_reallocblks_args /* {
struct vnode *a_vp;
struct cluster_save *a_buflist;
} */ *ap;
{
if (doreallocblks == 0)
return (ENOSPC);
/*
* We can't wait in softdep prealloc as it may fsync and recurse
* here. Instead we simply fail to reallocate blocks if this
* rare condition arises.
*/
if (DOINGSOFTDEP(ap->a_vp))
if (softdep_prealloc(ap->a_vp, MNT_NOWAIT) != 0)
return (ENOSPC);
if (VTOI(ap->a_vp)->i_ump->um_fstype == UFS1)
return (ffs_reallocblks_ufs1(ap));
return (ffs_reallocblks_ufs2(ap));
}
static int
ffs_reallocblks_ufs1(ap)
struct vop_reallocblks_args /* {
struct vnode *a_vp;
struct cluster_save *a_buflist;
} */ *ap;
{
struct fs *fs;
struct inode *ip;
struct vnode *vp;
struct buf *sbp, *ebp;
ufs1_daddr_t *bap, *sbap, *ebap = 0;
struct cluster_save *buflist;
struct ufsmount *ump;
ufs_lbn_t start_lbn, end_lbn;
ufs1_daddr_t soff, newblk, blkno;
ufs2_daddr_t pref;
struct indir start_ap[NIADDR + 1], end_ap[NIADDR + 1], *idp;
int i, len, start_lvl, end_lvl, ssize;
vp = ap->a_vp;
ip = VTOI(vp);
fs = ip->i_fs;
ump = ip->i_ump;
if (fs->fs_contigsumsize <= 0)
return (ENOSPC);
buflist = ap->a_buflist;
len = buflist->bs_nchildren;
start_lbn = buflist->bs_children[0]->b_lblkno;
end_lbn = start_lbn + len - 1;
#ifdef INVARIANTS
for (i = 0; i < len; i++)
if (!ffs_checkblk(ip,
dbtofsb(fs, buflist->bs_children[i]->b_blkno), fs->fs_bsize))
panic("ffs_reallocblks: unallocated block 1");
for (i = 1; i < len; i++)
if (buflist->bs_children[i]->b_lblkno != start_lbn + i)
panic("ffs_reallocblks: non-logical cluster");
blkno = buflist->bs_children[0]->b_blkno;
ssize = fsbtodb(fs, fs->fs_frag);
for (i = 1; i < len - 1; i++)
if (buflist->bs_children[i]->b_blkno != blkno + (i * ssize))
panic("ffs_reallocblks: non-physical cluster %d", i);
#endif
/*
* If the cluster crosses the boundary for the first indirect
* block, leave space for the indirect block. Indirect blocks
* are initially laid out in a position after the last direct
* block. Block reallocation would usually destroy locality by
* moving the indirect block out of the way to make room for
* data blocks if we didn't compensate here. We should also do
* this for other indirect block boundaries, but it is only
* important for the first one.
*/
if (start_lbn < NDADDR && end_lbn >= NDADDR)
return (ENOSPC);
/*
* If the latest allocation is in a new cylinder group, assume that
* the filesystem has decided to move and do not force it back to
* the previous cylinder group.
*/
if (dtog(fs, dbtofsb(fs, buflist->bs_children[0]->b_blkno)) !=
dtog(fs, dbtofsb(fs, buflist->bs_children[len - 1]->b_blkno)))
return (ENOSPC);
if (ufs_getlbns(vp, start_lbn, start_ap, &start_lvl) ||
ufs_getlbns(vp, end_lbn, end_ap, &end_lvl))
return (ENOSPC);
/*
* Get the starting offset and block map for the first block.
*/
if (start_lvl == 0) {
sbap = &ip->i_din1->di_db[0];
soff = start_lbn;
} else {
idp = &start_ap[start_lvl - 1];
if (bread(vp, idp->in_lbn, (int)fs->fs_bsize, NOCRED, &sbp)) {
brelse(sbp);
return (ENOSPC);
}
sbap = (ufs1_daddr_t *)sbp->b_data;
soff = idp->in_off;
}
/*
* If the block range spans two block maps, get the second map.
*/
if (end_lvl == 0 || (idp = &end_ap[end_lvl - 1])->in_off + 1 >= len) {
ssize = len;
} else {
#ifdef INVARIANTS
if (start_lvl > 0 &&
start_ap[start_lvl - 1].in_lbn == idp->in_lbn)
panic("ffs_reallocblk: start == end");
#endif
ssize = len - (idp->in_off + 1);
if (bread(vp, idp->in_lbn, (int)fs->fs_bsize, NOCRED, &ebp))
goto fail;
ebap = (ufs1_daddr_t *)ebp->b_data;
}
/*
* Find the preferred location for the cluster.
*/
UFS_LOCK(ump);
pref = ffs_blkpref_ufs1(ip, start_lbn, soff, sbap);
/*
* Search the block map looking for an allocation of the desired size.
*/
if ((newblk = ffs_hashalloc(ip, dtog(fs, pref), pref,
len, len, ffs_clusteralloc)) == 0) {
UFS_UNLOCK(ump);
goto fail;
}
/*
* We have found a new contiguous block.
*
* First we have to replace the old block pointers with the new
* block pointers in the inode and indirect blocks associated
* with the file.
*/
#ifdef DEBUG
if (prtrealloc)
printf("realloc: ino %ju, lbns %jd-%jd\n\told:",
(uintmax_t)ip->i_number,
(intmax_t)start_lbn, (intmax_t)end_lbn);
#endif
blkno = newblk;
for (bap = &sbap[soff], i = 0; i < len; i++, blkno += fs->fs_frag) {
if (i == ssize) {
bap = ebap;
soff = -i;
}
#ifdef INVARIANTS
if (!ffs_checkblk(ip,
dbtofsb(fs, buflist->bs_children[i]->b_blkno), fs->fs_bsize))
panic("ffs_reallocblks: unallocated block 2");
if (dbtofsb(fs, buflist->bs_children[i]->b_blkno) != *bap)
panic("ffs_reallocblks: alloc mismatch");
#endif
#ifdef DEBUG
if (prtrealloc)
printf(" %d,", *bap);
#endif
if (DOINGSOFTDEP(vp)) {
if (sbap == &ip->i_din1->di_db[0] && i < ssize)
softdep_setup_allocdirect(ip, start_lbn + i,
blkno, *bap, fs->fs_bsize, fs->fs_bsize,
buflist->bs_children[i]);
else
softdep_setup_allocindir_page(ip, start_lbn + i,
i < ssize ? sbp : ebp, soff + i, blkno,
*bap, buflist->bs_children[i]);
}
*bap++ = blkno;
}
/*
* Next we must write out the modified inode and indirect blocks.
* For strict correctness, the writes should be synchronous since
* the old block values may have been written to disk. In practise
* they are almost never written, but if we are concerned about
* strict correctness, the `doasyncfree' flag should be set to zero.
*
* The test on `doasyncfree' should be changed to test a flag
* that shows whether the associated buffers and inodes have
* been written. The flag should be set when the cluster is
* started and cleared whenever the buffer or inode is flushed.
* We can then check below to see if it is set, and do the
* synchronous write only when it has been cleared.
*/
if (sbap != &ip->i_din1->di_db[0]) {
if (doasyncfree)
bdwrite(sbp);
else
bwrite(sbp);
} else {
ip->i_flag |= IN_CHANGE | IN_UPDATE;
if (!doasyncfree)
ffs_update(vp, 1);
}
if (ssize < len) {
if (doasyncfree)
bdwrite(ebp);
else
bwrite(ebp);
}
/*
* Last, free the old blocks and assign the new blocks to the buffers.
*/
#ifdef DEBUG
if (prtrealloc)
printf("\n\tnew:");
#endif
for (blkno = newblk, i = 0; i < len; i++, blkno += fs->fs_frag) {
if (!DOINGSOFTDEP(vp))
ffs_blkfree(ump, fs, ip->i_devvp,
dbtofsb(fs, buflist->bs_children[i]->b_blkno),
fs->fs_bsize, ip->i_number, vp->v_type, NULL);
buflist->bs_children[i]->b_blkno = fsbtodb(fs, blkno);
#ifdef INVARIANTS
if (!ffs_checkblk(ip,
dbtofsb(fs, buflist->bs_children[i]->b_blkno), fs->fs_bsize))
panic("ffs_reallocblks: unallocated block 3");
#endif
#ifdef DEBUG
if (prtrealloc)
printf(" %d,", blkno);
#endif
}
#ifdef DEBUG
if (prtrealloc) {
prtrealloc--;
printf("\n");
}
#endif
return (0);
fail:
if (ssize < len)
brelse(ebp);
if (sbap != &ip->i_din1->di_db[0])
brelse(sbp);
return (ENOSPC);
}
static int
ffs_reallocblks_ufs2(ap)
struct vop_reallocblks_args /* {
struct vnode *a_vp;
struct cluster_save *a_buflist;
} */ *ap;
{
struct fs *fs;
struct inode *ip;
struct vnode *vp;
struct buf *sbp, *ebp;
ufs2_daddr_t *bap, *sbap, *ebap = 0;
struct cluster_save *buflist;
struct ufsmount *ump;
ufs_lbn_t start_lbn, end_lbn;
ufs2_daddr_t soff, newblk, blkno, pref;
struct indir start_ap[NIADDR + 1], end_ap[NIADDR + 1], *idp;
int i, len, start_lvl, end_lvl, ssize;
vp = ap->a_vp;
ip = VTOI(vp);
fs = ip->i_fs;
ump = ip->i_ump;
if (fs->fs_contigsumsize <= 0)
return (ENOSPC);
buflist = ap->a_buflist;
len = buflist->bs_nchildren;
start_lbn = buflist->bs_children[0]->b_lblkno;
end_lbn = start_lbn + len - 1;
#ifdef INVARIANTS
for (i = 0; i < len; i++)
if (!ffs_checkblk(ip,
dbtofsb(fs, buflist->bs_children[i]->b_blkno), fs->fs_bsize))
panic("ffs_reallocblks: unallocated block 1");
for (i = 1; i < len; i++)
if (buflist->bs_children[i]->b_lblkno != start_lbn + i)
panic("ffs_reallocblks: non-logical cluster");
blkno = buflist->bs_children[0]->b_blkno;
ssize = fsbtodb(fs, fs->fs_frag);
for (i = 1; i < len - 1; i++)
if (buflist->bs_children[i]->b_blkno != blkno + (i * ssize))
panic("ffs_reallocblks: non-physical cluster %d", i);
#endif
/*
* If the cluster crosses the boundary for the first indirect
* block, do not move anything in it. Indirect blocks are
* usually initially laid out in a position between the data
* blocks. Block reallocation would usually destroy locality by
* moving the indirect block out of the way to make room for
* data blocks if we didn't compensate here. We should also do
* this for other indirect block boundaries, but it is only
* important for the first one.
*/
if (start_lbn < NDADDR && end_lbn >= NDADDR)
return (ENOSPC);
/*
* If the latest allocation is in a new cylinder group, assume that
* the filesystem has decided to move and do not force it back to
* the previous cylinder group.
*/
if (dtog(fs, dbtofsb(fs, buflist->bs_children[0]->b_blkno)) !=
dtog(fs, dbtofsb(fs, buflist->bs_children[len - 1]->b_blkno)))
return (ENOSPC);
if (ufs_getlbns(vp, start_lbn, start_ap, &start_lvl) ||
ufs_getlbns(vp, end_lbn, end_ap, &end_lvl))
return (ENOSPC);
/*
* Get the starting offset and block map for the first block.
*/
if (start_lvl == 0) {
sbap = &ip->i_din2->di_db[0];
soff = start_lbn;
} else {
idp = &start_ap[start_lvl - 1];
if (bread(vp, idp->in_lbn, (int)fs->fs_bsize, NOCRED, &sbp)) {
brelse(sbp);
return (ENOSPC);
}
sbap = (ufs2_daddr_t *)sbp->b_data;
soff = idp->in_off;
}
/*
* If the block range spans two block maps, get the second map.
*/
if (end_lvl == 0 || (idp = &end_ap[end_lvl - 1])->in_off + 1 >= len) {
ssize = len;
} else {
#ifdef INVARIANTS
if (start_lvl > 0 &&
start_ap[start_lvl - 1].in_lbn == idp->in_lbn)
panic("ffs_reallocblk: start == end");
#endif
ssize = len - (idp->in_off + 1);
if (bread(vp, idp->in_lbn, (int)fs->fs_bsize, NOCRED, &ebp))
goto fail;
ebap = (ufs2_daddr_t *)ebp->b_data;
}
/*
* Find the preferred location for the cluster.
*/
UFS_LOCK(ump);
pref = ffs_blkpref_ufs2(ip, start_lbn, soff, sbap);
/*
* Skip a block for the first indirect block. Indirect blocks are
* usually initially laid out in a good position between the data
* blocks, but block reallocation would usually destroy locality by
* moving them out of the way to make room for data blocks if we
* didn't compensate here.
*/
if (start_lbn == NDADDR)
pref += fs->fs_frag;
/*
* Search the block map looking for an allocation of the desired size.
*/
if ((newblk = ffs_hashalloc(ip, dtog(fs, pref), pref,
len, len, ffs_clusteralloc)) == 0) {
UFS_UNLOCK(ump);
goto fail;
}
/*
* We have found a new contiguous block.
*
* First we have to replace the old block pointers with the new
* block pointers in the inode and indirect blocks associated
* with the file.
*/
#ifdef DEBUG
if (prtrealloc)
printf("realloc: ino %d, lbns %jd-%jd\n\told:", ip->i_number,
(intmax_t)start_lbn, (intmax_t)end_lbn);
#endif
blkno = newblk;
for (bap = &sbap[soff], i = 0; i < len; i++, blkno += fs->fs_frag) {
if (i == ssize) {
bap = ebap;
soff = -i;
}
#ifdef INVARIANTS
if (!ffs_checkblk(ip,
dbtofsb(fs, buflist->bs_children[i]->b_blkno), fs->fs_bsize))
panic("ffs_reallocblks: unallocated block 2");
if (dbtofsb(fs, buflist->bs_children[i]->b_blkno) != *bap)
panic("ffs_reallocblks: alloc mismatch");
#endif
#ifdef DEBUG
if (prtrealloc)
printf(" %jd,", (intmax_t)*bap);
#endif
if (DOINGSOFTDEP(vp)) {
if (sbap == &ip->i_din2->di_db[0] && i < ssize)
softdep_setup_allocdirect(ip, start_lbn + i,
blkno, *bap, fs->fs_bsize, fs->fs_bsize,
buflist->bs_children[i]);
else
softdep_setup_allocindir_page(ip, start_lbn + i,
i < ssize ? sbp : ebp, soff + i, blkno,
*bap, buflist->bs_children[i]);
}
*bap++ = blkno;
}
/*
* Next we must write out the modified inode and indirect blocks.
* For strict correctness, the writes should be synchronous since
* the old block values may have been written to disk. In practise
* they are almost never written, but if we are concerned about
* strict correctness, the `doasyncfree' flag should be set to zero.
*
* The test on `doasyncfree' should be changed to test a flag
* that shows whether the associated buffers and inodes have
* been written. The flag should be set when the cluster is
* started and cleared whenever the buffer or inode is flushed.
* We can then check below to see if it is set, and do the
* synchronous write only when it has been cleared.
*/
if (sbap != &ip->i_din2->di_db[0]) {
if (doasyncfree)
bdwrite(sbp);
else
bwrite(sbp);
} else {
ip->i_flag |= IN_CHANGE | IN_UPDATE;
if (!doasyncfree)
ffs_update(vp, 1);
}
if (ssize < len) {
if (doasyncfree)
bdwrite(ebp);
else
bwrite(ebp);
}
/*
* Last, free the old blocks and assign the new blocks to the buffers.
*/
#ifdef DEBUG
if (prtrealloc)
printf("\n\tnew:");
#endif
for (blkno = newblk, i = 0; i < len; i++, blkno += fs->fs_frag) {
if (!DOINGSOFTDEP(vp))
ffs_blkfree(ump, fs, ip->i_devvp,
dbtofsb(fs, buflist->bs_children[i]->b_blkno),
fs->fs_bsize, ip->i_number, vp->v_type, NULL);
buflist->bs_children[i]->b_blkno = fsbtodb(fs, blkno);
#ifdef INVARIANTS
if (!ffs_checkblk(ip,
dbtofsb(fs, buflist->bs_children[i]->b_blkno), fs->fs_bsize))
panic("ffs_reallocblks: unallocated block 3");
#endif
#ifdef DEBUG
if (prtrealloc)
printf(" %jd,", (intmax_t)blkno);
#endif
}
#ifdef DEBUG
if (prtrealloc) {
prtrealloc--;
printf("\n");
}
#endif
return (0);
fail:
if (ssize < len)
brelse(ebp);
if (sbap != &ip->i_din2->di_db[0])
brelse(sbp);
return (ENOSPC);
}
/*
* Allocate an inode in the filesystem.
*
* If allocating a directory, use ffs_dirpref to select the inode.
* If allocating in a directory, the following hierarchy is followed:
* 1) allocate the preferred inode.
* 2) allocate an inode in the same cylinder group.
* 3) quadradically rehash into other cylinder groups, until an
* available inode is located.
* If no inode preference is given the following hierarchy is used
* to allocate an inode:
* 1) allocate an inode in cylinder group 0.
* 2) quadradically rehash into other cylinder groups, until an
* available inode is located.
*/
int
ffs_valloc(pvp, mode, cred, vpp)
struct vnode *pvp;
int mode;
struct ucred *cred;
struct vnode **vpp;
{
struct inode *pip;
struct fs *fs;
struct inode *ip;
struct timespec ts;
struct ufsmount *ump;
ino_t ino, ipref;
u_int cg;
int error, error1, reclaimed;
static struct timeval lastfail;
static int curfail;
*vpp = NULL;
pip = VTOI(pvp);
fs = pip->i_fs;
ump = pip->i_ump;
UFS_LOCK(ump);
reclaimed = 0;
retry:
if (fs->fs_cstotal.cs_nifree == 0)
goto noinodes;
if ((mode & IFMT) == IFDIR)
ipref = ffs_dirpref(pip);
else
ipref = pip->i_number;
if (ipref >= fs->fs_ncg * fs->fs_ipg)
ipref = 0;
cg = ino_to_cg(fs, ipref);
/*
* Track number of dirs created one after another
* in a same cg without intervening by files.
*/
if ((mode & IFMT) == IFDIR) {
if (fs->fs_contigdirs[cg] < 255)
fs->fs_contigdirs[cg]++;
} else {
if (fs->fs_contigdirs[cg] > 0)
fs->fs_contigdirs[cg]--;
}
ino = (ino_t)ffs_hashalloc(pip, cg, ipref, mode, 0,
(allocfcn_t *)ffs_nodealloccg);
if (ino == 0)
goto noinodes;
error = ffs_vget(pvp->v_mount, ino, LK_EXCLUSIVE, vpp);
if (error) {
error1 = ffs_vgetf(pvp->v_mount, ino, LK_EXCLUSIVE, vpp,
FFSV_FORCEINSMQ);
ffs_vfree(pvp, ino, mode);
if (error1 == 0) {
ip = VTOI(*vpp);
if (ip->i_mode)
goto dup_alloc;
ip->i_flag |= IN_MODIFIED;
vput(*vpp);
}
return (error);
}
ip = VTOI(*vpp);
if (ip->i_mode) {
dup_alloc:
printf("mode = 0%o, inum = %lu, fs = %s\n",
ip->i_mode, (u_long)ip->i_number, fs->fs_fsmnt);
panic("ffs_valloc: dup alloc");
}
if (DIP(ip, i_blocks) && (fs->fs_flags & FS_UNCLEAN) == 0) { /* XXX */
printf("free inode %s/%lu had %ld blocks\n",
fs->fs_fsmnt, (u_long)ino, (long)DIP(ip, i_blocks));
DIP_SET(ip, i_blocks, 0);
}
ip->i_flags = 0;
DIP_SET(ip, i_flags, 0);
/*
* Set up a new generation number for this inode.
*/
if (ip->i_gen == 0 || ++ip->i_gen == 0)
ip->i_gen = arc4random() / 2 + 1;
DIP_SET(ip, i_gen, ip->i_gen);
if (fs->fs_magic == FS_UFS2_MAGIC) {
vfs_timestamp(&ts);
ip->i_din2->di_birthtime = ts.tv_sec;
ip->i_din2->di_birthnsec = ts.tv_nsec;
}
ufs_prepare_reclaim(*vpp);
ip->i_flag = 0;
(*vpp)->v_vflag = 0;
(*vpp)->v_type = VNON;
if (fs->fs_magic == FS_UFS2_MAGIC)
(*vpp)->v_op = &ffs_vnodeops2;
else
(*vpp)->v_op = &ffs_vnodeops1;
return (0);
noinodes:
if (reclaimed == 0) {
reclaimed = 1;
softdep_request_cleanup(fs, pvp, cred, FLUSH_INODES_WAIT);
goto retry;
}
UFS_UNLOCK(ump);
if (ppsratecheck(&lastfail, &curfail, 1)) {
ffs_fserr(fs, pip->i_number, "out of inodes");
uprintf("\n%s: create/symlink failed, no inodes free\n",
fs->fs_fsmnt);
}
return (ENOSPC);
}
/*
* Find a cylinder group to place a directory.
*
* The policy implemented by this algorithm is to allocate a
* directory inode in the same cylinder group as its parent
* directory, but also to reserve space for its files inodes
* and data. Restrict the number of directories which may be
* allocated one after another in the same cylinder group
* without intervening allocation of files.
*
* If we allocate a first level directory then force allocation
* in another cylinder group.
*/
static ino_t
ffs_dirpref(pip)
struct inode *pip;
{
struct fs *fs;
u_int cg, prefcg, dirsize, cgsize;
u_int avgifree, avgbfree, avgndir, curdirsize;
u_int minifree, minbfree, maxndir;
u_int mincg, minndir;
u_int maxcontigdirs;
mtx_assert(UFS_MTX(pip->i_ump), MA_OWNED);
fs = pip->i_fs;
avgifree = fs->fs_cstotal.cs_nifree / fs->fs_ncg;
avgbfree = fs->fs_cstotal.cs_nbfree / fs->fs_ncg;
avgndir = fs->fs_cstotal.cs_ndir / fs->fs_ncg;
/*
* Force allocation in another cg if creating a first level dir.
*/
ASSERT_VOP_LOCKED(ITOV(pip), "ffs_dirpref");
if (ITOV(pip)->v_vflag & VV_ROOT) {
prefcg = arc4random() % fs->fs_ncg;
mincg = prefcg;
minndir = fs->fs_ipg;
for (cg = prefcg; cg < fs->fs_ncg; cg++)
if (fs->fs_cs(fs, cg).cs_ndir < minndir &&
fs->fs_cs(fs, cg).cs_nifree >= avgifree &&
fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) {
mincg = cg;
minndir = fs->fs_cs(fs, cg).cs_ndir;
}
for (cg = 0; cg < prefcg; cg++)
if (fs->fs_cs(fs, cg).cs_ndir < minndir &&
fs->fs_cs(fs, cg).cs_nifree >= avgifree &&
fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) {
mincg = cg;
minndir = fs->fs_cs(fs, cg).cs_ndir;
}
return ((ino_t)(fs->fs_ipg * mincg));
}
/*
* Count various limits which used for
* optimal allocation of a directory inode.
*/
maxndir = min(avgndir + fs->fs_ipg / 16, fs->fs_ipg);
minifree = avgifree - avgifree / 4;
if (minifree < 1)
minifree = 1;
minbfree = avgbfree - avgbfree / 4;
if (minbfree < 1)
minbfree = 1;
cgsize = fs->fs_fsize * fs->fs_fpg;
dirsize = fs->fs_avgfilesize * fs->fs_avgfpdir;
curdirsize = avgndir ? (cgsize - avgbfree * fs->fs_bsize) / avgndir : 0;
if (dirsize < curdirsize)
dirsize = curdirsize;
if (dirsize <= 0)
maxcontigdirs = 0; /* dirsize overflowed */
else
maxcontigdirs = min((avgbfree * fs->fs_bsize) / dirsize, 255);
if (fs->fs_avgfpdir > 0)
maxcontigdirs = min(maxcontigdirs,
fs->fs_ipg / fs->fs_avgfpdir);
if (maxcontigdirs == 0)
maxcontigdirs = 1;
/*
* Limit number of dirs in one cg and reserve space for
* regular files, but only if we have no deficit in
* inodes or space.
*/
prefcg = ino_to_cg(fs, pip->i_number);
for (cg = prefcg; cg < fs->fs_ncg; cg++)
if (fs->fs_cs(fs, cg).cs_ndir < maxndir &&
fs->fs_cs(fs, cg).cs_nifree >= minifree &&
fs->fs_cs(fs, cg).cs_nbfree >= minbfree) {
if (fs->fs_contigdirs[cg] < maxcontigdirs)
return ((ino_t)(fs->fs_ipg * cg));
}
for (cg = 0; cg < prefcg; cg++)
if (fs->fs_cs(fs, cg).cs_ndir < maxndir &&
fs->fs_cs(fs, cg).cs_nifree >= minifree &&
fs->fs_cs(fs, cg).cs_nbfree >= minbfree) {
if (fs->fs_contigdirs[cg] < maxcontigdirs)
return ((ino_t)(fs->fs_ipg * cg));
}
/*
* This is a backstop when we have deficit in space.
*/
for (cg = prefcg; cg < fs->fs_ncg; cg++)
if (fs->fs_cs(fs, cg).cs_nifree >= avgifree)
return ((ino_t)(fs->fs_ipg * cg));
for (cg = 0; cg < prefcg; cg++)
if (fs->fs_cs(fs, cg).cs_nifree >= avgifree)
break;
return ((ino_t)(fs->fs_ipg * cg));
}
/*
* Select the desired position for the next block in a file. The file is
* logically divided into sections. The first section is composed of the
* direct blocks. Each additional section contains fs_maxbpg blocks.
*
* If no blocks have been allocated in the first section, the policy is to
* request a block in the same cylinder group as the inode that describes
* the file. If no blocks have been allocated in any other section, the
* policy is to place the section in a cylinder group with a greater than
* average number of free blocks. An appropriate cylinder group is found
* by using a rotor that sweeps the cylinder groups. When a new group of
* blocks is needed, the sweep begins in the cylinder group following the
* cylinder group from which the previous allocation was made. The sweep
* continues until a cylinder group with greater than the average number
* of free blocks is found. If the allocation is for the first block in an
* indirect block, the information on the previous allocation is unavailable;
* here a best guess is made based upon the logical block number being
* allocated.
*
* If a section is already partially allocated, the policy is to
* contiguously allocate fs_maxcontig blocks. The end of one of these
* contiguous blocks and the beginning of the next is laid out
* contiguously if possible.
*/
ufs2_daddr_t
ffs_blkpref_ufs1(ip, lbn, indx, bap)
struct inode *ip;
ufs_lbn_t lbn;
int indx;
ufs1_daddr_t *bap;
{
struct fs *fs;
u_int cg;
u_int avgbfree, startcg;
ufs2_daddr_t pref;
mtx_assert(UFS_MTX(ip->i_ump), MA_OWNED);
fs = ip->i_fs;
/*
* If we are allocating the first indirect block, try to place it
* immediately following the last direct block.
*
* If we are allocating the first data block in the first indirect
* block, try to place it immediately following the indirect block.
*/
if (lbn == NDADDR) {
pref = ip->i_din1->di_db[NDADDR - 1];
if (bap == NULL && pref != 0)
return (pref + fs->fs_frag);
pref = ip->i_din1->di_ib[0];
if (pref != 0)
return (pref + fs->fs_frag);
}
if (indx % fs->fs_maxbpg == 0 || bap[indx - 1] == 0) {
if (lbn < NDADDR + NINDIR(fs)) {
cg = ino_to_cg(fs, ip->i_number);
return (cgbase(fs, cg) + fs->fs_frag);
}
/*
* Find a cylinder with greater than average number of
* unused data blocks.
*/
if (indx == 0 || bap[indx - 1] == 0)
startcg =
ino_to_cg(fs, ip->i_number) + lbn / fs->fs_maxbpg;
else
startcg = dtog(fs, bap[indx - 1]) + 1;
startcg %= fs->fs_ncg;
avgbfree = fs->fs_cstotal.cs_nbfree / fs->fs_ncg;
for (cg = startcg; cg < fs->fs_ncg; cg++)
if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) {
fs->fs_cgrotor = cg;
return (cgbase(fs, cg) + fs->fs_frag);
}
for (cg = 0; cg <= startcg; cg++)
if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) {
fs->fs_cgrotor = cg;
return (cgbase(fs, cg) + fs->fs_frag);
}
return (0);
}
/*
* We just always try to lay things out contiguously.
*/
return (bap[indx - 1] + fs->fs_frag);
}
/*
* Same as above, but for UFS2
*/
ufs2_daddr_t
ffs_blkpref_ufs2(ip, lbn, indx, bap)
struct inode *ip;
ufs_lbn_t lbn;
int indx;
ufs2_daddr_t *bap;
{
struct fs *fs;
u_int cg;
u_int avgbfree, startcg;
ufs2_daddr_t pref;
mtx_assert(UFS_MTX(ip->i_ump), MA_OWNED);
fs = ip->i_fs;
/*
* If we are allocating the first indirect block, try to place it
* immediately following the last direct block.
*
* If we are allocating the first data block in the first indirect
* block, try to place it immediately following the indirect block.
*/
if (lbn == NDADDR) {
pref = ip->i_din1->di_db[NDADDR - 1];
if (bap == NULL && pref != 0)
return (pref + fs->fs_frag);
pref = ip->i_din1->di_ib[0];
if (pref != 0)
return (pref + fs->fs_frag);
}
if (indx % fs->fs_maxbpg == 0 || bap[indx - 1] == 0) {
if (lbn < NDADDR + NINDIR(fs)) {
cg = ino_to_cg(fs, ip->i_number);
return (cgbase(fs, cg) + fs->fs_frag);
}
/*
* Find a cylinder with greater than average number of
* unused data blocks.
*/
if (indx == 0 || bap[indx - 1] == 0)
startcg =
ino_to_cg(fs, ip->i_number) + lbn / fs->fs_maxbpg;
else
startcg = dtog(fs, bap[indx - 1]) + 1;
startcg %= fs->fs_ncg;
avgbfree = fs->fs_cstotal.cs_nbfree / fs->fs_ncg;
for (cg = startcg; cg < fs->fs_ncg; cg++)
if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) {
fs->fs_cgrotor = cg;
return (cgbase(fs, cg) + fs->fs_frag);
}
for (cg = 0; cg <= startcg; cg++)
if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) {
fs->fs_cgrotor = cg;
return (cgbase(fs, cg) + fs->fs_frag);
}
return (0);
}
/*
* We just always try to lay things out contiguously.
*/
return (bap[indx - 1] + fs->fs_frag);
}
/*
* Implement the cylinder overflow algorithm.
*
* The policy implemented by this algorithm is:
* 1) allocate the block in its requested cylinder group.
* 2) quadradically rehash on the cylinder group number.
* 3) brute force search for a free block.
*
* Must be called with the UFS lock held. Will release the lock on success
* and return with it held on failure.
*/
/*VARARGS5*/
static ufs2_daddr_t
ffs_hashalloc(ip, cg, pref, size, rsize, allocator)
struct inode *ip;
u_int cg;
ufs2_daddr_t pref;
int size; /* Search size for data blocks, mode for inodes */
int rsize; /* Real allocated size. */
allocfcn_t *allocator;
{
struct fs *fs;
ufs2_daddr_t result;
u_int i, icg = cg;
mtx_assert(UFS_MTX(ip->i_ump), MA_OWNED);
#ifdef INVARIANTS
if (ITOV(ip)->v_mount->mnt_kern_flag & MNTK_SUSPENDED)
panic("ffs_hashalloc: allocation on suspended filesystem");
#endif
fs = ip->i_fs;
/*
* 1: preferred cylinder group
*/
result = (*allocator)(ip, cg, pref, size, rsize);
if (result)
return (result);
/*
* 2: quadratic rehash
*/
for (i = 1; i < fs->fs_ncg; i *= 2) {
cg += i;
if (cg >= fs->fs_ncg)
cg -= fs->fs_ncg;
result = (*allocator)(ip, cg, 0, size, rsize);
if (result)
return (result);
}
/*
* 3: brute force search
* Note that we start at i == 2, since 0 was checked initially,
* and 1 is always checked in the quadratic rehash.
*/
cg = (icg + 2) % fs->fs_ncg;
for (i = 2; i < fs->fs_ncg; i++) {
result = (*allocator)(ip, cg, 0, size, rsize);
if (result)
return (result);
cg++;
if (cg == fs->fs_ncg)
cg = 0;
}
return (0);
}
/*
* Determine whether a fragment can be extended.
*
* Check to see if the necessary fragments are available, and
* if they are, allocate them.
*/
static ufs2_daddr_t
ffs_fragextend(ip, cg, bprev, osize, nsize)
struct inode *ip;
u_int cg;
ufs2_daddr_t bprev;
int osize, nsize;
{
struct fs *fs;
struct cg *cgp;
struct buf *bp;
struct ufsmount *ump;
int nffree;
long bno;
int frags, bbase;
int i, error;
u_int8_t *blksfree;
ump = ip->i_ump;
fs = ip->i_fs;
if (fs->fs_cs(fs, cg).cs_nffree < numfrags(fs, nsize - osize))
return (0);
frags = numfrags(fs, nsize);
bbase = fragnum(fs, bprev);
if (bbase > fragnum(fs, (bprev + frags - 1))) {
/* cannot extend across a block boundary */
return (0);
}
UFS_UNLOCK(ump);
error = bread(ip->i_devvp, fsbtodb(fs, cgtod(fs, cg)),
(int)fs->fs_cgsize, NOCRED, &bp);
if (error)
goto fail;
cgp = (struct cg *)bp->b_data;
if (!cg_chkmagic(cgp))
goto fail;
bp->b_xflags |= BX_BKGRDWRITE;
cgp->cg_old_time = cgp->cg_time = time_second;
bno = dtogd(fs, bprev);
blksfree = cg_blksfree(cgp);
for (i = numfrags(fs, osize); i < frags; i++)
if (isclr(blksfree, bno + i))
goto fail;
/*
* the current fragment can be extended
* deduct the count on fragment being extended into
* increase the count on the remaining fragment (if any)
* allocate the extended piece
*/
for (i = frags; i < fs->fs_frag - bbase; i++)
if (isclr(blksfree, bno + i))
break;
cgp->cg_frsum[i - numfrags(fs, osize)]--;
if (i != frags)
cgp->cg_frsum[i - frags]++;
for (i = numfrags(fs, osize), nffree = 0; i < frags; i++) {
clrbit(blksfree, bno + i);
cgp->cg_cs.cs_nffree--;
nffree++;
}
UFS_LOCK(ump);
fs->fs_cstotal.cs_nffree -= nffree;
fs->fs_cs(fs, cg).cs_nffree -= nffree;
fs->fs_fmod = 1;
ACTIVECLEAR(fs, cg);
UFS_UNLOCK(ump);
if (DOINGSOFTDEP(ITOV(ip)))
softdep_setup_blkmapdep(bp, UFSTOVFS(ump), bprev,
frags, numfrags(fs, osize));
bdwrite(bp);
return (bprev);
fail:
brelse(bp);
UFS_LOCK(ump);
return (0);
}
/*
* Determine whether a block can be allocated.
*
* Check to see if a block of the appropriate size is available,
* and if it is, allocate it.
*/
static ufs2_daddr_t
ffs_alloccg(ip, cg, bpref, size, rsize)
struct inode *ip;
u_int cg;
ufs2_daddr_t bpref;
int size;
int rsize;
{
struct fs *fs;
struct cg *cgp;
struct buf *bp;
struct ufsmount *ump;
ufs1_daddr_t bno;
ufs2_daddr_t blkno;
int i, allocsiz, error, frags;
u_int8_t *blksfree;
ump = ip->i_ump;
fs = ip->i_fs;
if (fs->fs_cs(fs, cg).cs_nbfree == 0 && size == fs->fs_bsize)
return (0);
UFS_UNLOCK(ump);
error = bread(ip->i_devvp, fsbtodb(fs, cgtod(fs, cg)),
(int)fs->fs_cgsize, NOCRED, &bp);
if (error)
goto fail;
cgp = (struct cg *)bp->b_data;
if (!cg_chkmagic(cgp) ||
(cgp->cg_cs.cs_nbfree == 0 && size == fs->fs_bsize))
goto fail;
bp->b_xflags |= BX_BKGRDWRITE;
cgp->cg_old_time = cgp->cg_time = time_second;
if (size == fs->fs_bsize) {
UFS_LOCK(ump);
blkno = ffs_alloccgblk(ip, bp, bpref, rsize);
ACTIVECLEAR(fs, cg);
UFS_UNLOCK(ump);
bdwrite(bp);
return (blkno);
}
/*
* check to see if any fragments are already available
* allocsiz is the size which will be allocated, hacking
* it down to a smaller size if necessary
*/
blksfree = cg_blksfree(cgp);
frags = numfrags(fs, size);
for (allocsiz = frags; allocsiz < fs->fs_frag; allocsiz++)
if (cgp->cg_frsum[allocsiz] != 0)
break;
if (allocsiz == fs->fs_frag) {
/*
* no fragments were available, so a block will be
* allocated, and hacked up
*/
if (cgp->cg_cs.cs_nbfree == 0)
goto fail;
UFS_LOCK(ump);
blkno = ffs_alloccgblk(ip, bp, bpref, rsize);
ACTIVECLEAR(fs, cg);
UFS_UNLOCK(ump);
bdwrite(bp);
return (blkno);
}
KASSERT(size == rsize,
("ffs_alloccg: size(%d) != rsize(%d)", size, rsize));
bno = ffs_mapsearch(fs, cgp, bpref, allocsiz);
if (bno < 0)
goto fail;
for (i = 0; i < frags; i++)
clrbit(blksfree, bno + i);
cgp->cg_cs.cs_nffree -= frags;
cgp->cg_frsum[allocsiz]--;
if (frags != allocsiz)
cgp->cg_frsum[allocsiz - frags]++;
UFS_LOCK(ump);
fs->fs_cstotal.cs_nffree -= frags;
fs->fs_cs(fs, cg).cs_nffree -= frags;
fs->fs_fmod = 1;
blkno = cgbase(fs, cg) + bno;
ACTIVECLEAR(fs, cg);
UFS_UNLOCK(ump);
if (DOINGSOFTDEP(ITOV(ip)))
softdep_setup_blkmapdep(bp, UFSTOVFS(ump), blkno, frags, 0);
bdwrite(bp);
return (blkno);
fail:
brelse(bp);
UFS_LOCK(ump);
return (0);
}
/*
* Allocate a block in a cylinder group.
*
* This algorithm implements the following policy:
* 1) allocate the requested block.
* 2) allocate a rotationally optimal block in the same cylinder.
* 3) allocate the next available block on the block rotor for the
* specified cylinder group.
* Note that this routine only allocates fs_bsize blocks; these
* blocks may be fragmented by the routine that allocates them.
*/
static ufs2_daddr_t
ffs_alloccgblk(ip, bp, bpref, size)
struct inode *ip;
struct buf *bp;
ufs2_daddr_t bpref;
int size;
{
struct fs *fs;
struct cg *cgp;
struct ufsmount *ump;
ufs1_daddr_t bno;
ufs2_daddr_t blkno;
u_int8_t *blksfree;
int i;
fs = ip->i_fs;
ump = ip->i_ump;
mtx_assert(UFS_MTX(ump), MA_OWNED);
cgp = (struct cg *)bp->b_data;
blksfree = cg_blksfree(cgp);
if (bpref == 0 || dtog(fs, bpref) != cgp->cg_cgx) {
bpref = cgp->cg_rotor;
} else {
bpref = blknum(fs, bpref);
bno = dtogd(fs, bpref);
/*
* if the requested block is available, use it
*/
if (ffs_isblock(fs, blksfree, fragstoblks(fs, bno)))
goto gotit;
}
/*
* Take the next available block in this cylinder group.
*/
bno = ffs_mapsearch(fs, cgp, bpref, (int)fs->fs_frag);
if (bno < 0)
return (0);
cgp->cg_rotor = bno;
gotit:
blkno = fragstoblks(fs, bno);
ffs_clrblock(fs, blksfree, (long)blkno);
ffs_clusteracct(fs, cgp, blkno, -1);
cgp->cg_cs.cs_nbfree--;
fs->fs_cstotal.cs_nbfree--;
fs->fs_cs(fs, cgp->cg_cgx).cs_nbfree--;
fs->fs_fmod = 1;
blkno = cgbase(fs, cgp->cg_cgx) + bno;
/*
* If the caller didn't want the whole block free the frags here.
*/
size = numfrags(fs, size);
if (size != fs->fs_frag) {
bno = dtogd(fs, blkno);
for (i = size; i < fs->fs_frag; i++)
setbit(blksfree, bno + i);
i = fs->fs_frag - size;
cgp->cg_cs.cs_nffree += i;
fs->fs_cstotal.cs_nffree += i;
fs->fs_cs(fs, cgp->cg_cgx).cs_nffree += i;
fs->fs_fmod = 1;
cgp->cg_frsum[i]++;
}
/* XXX Fixme. */
UFS_UNLOCK(ump);
if (DOINGSOFTDEP(ITOV(ip)))
softdep_setup_blkmapdep(bp, UFSTOVFS(ump), blkno,
size, 0);
UFS_LOCK(ump);
return (blkno);
}
/*
* Determine whether a cluster can be allocated.
*
* We do not currently check for optimal rotational layout if there
* are multiple choices in the same cylinder group. Instead we just
* take the first one that we find following bpref.
*/
static ufs2_daddr_t
ffs_clusteralloc(ip, cg, bpref, len, unused)
struct inode *ip;
u_int cg;
ufs2_daddr_t bpref;
int len;
int unused;
{
struct fs *fs;
struct cg *cgp;
struct buf *bp;
struct ufsmount *ump;
int i, run, bit, map, got;
ufs2_daddr_t bno;
u_char *mapp;
int32_t *lp;
u_int8_t *blksfree;
fs = ip->i_fs;
ump = ip->i_ump;
if (fs->fs_maxcluster[cg] < len)
return (0);
UFS_UNLOCK(ump);
if (bread(ip->i_devvp, fsbtodb(fs, cgtod(fs, cg)), (int)fs->fs_cgsize,
NOCRED, &bp))
goto fail_lock;
cgp = (struct cg *)bp->b_data;
if (!cg_chkmagic(cgp))
goto fail_lock;
bp->b_xflags |= BX_BKGRDWRITE;
/*
* Check to see if a cluster of the needed size (or bigger) is
* available in this cylinder group.
*/
lp = &cg_clustersum(cgp)[len];
for (i = len; i <= fs->fs_contigsumsize; i++)
if (*lp++ > 0)
break;
if (i > fs->fs_contigsumsize) {
/*
* This is the first time looking for a cluster in this
* cylinder group. Update the cluster summary information
* to reflect the true maximum sized cluster so that
* future cluster allocation requests can avoid reading
* the cylinder group map only to find no clusters.
*/
lp = &cg_clustersum(cgp)[len - 1];
for (i = len - 1; i > 0; i--)
if (*lp-- > 0)
break;
UFS_LOCK(ump);
fs->fs_maxcluster[cg] = i;
goto fail;
}
/*
* Search the cluster map to find a big enough cluster.
* We take the first one that we find, even if it is larger
* than we need as we prefer to get one close to the previous
* block allocation. We do not search before the current
* preference point as we do not want to allocate a block
* that is allocated before the previous one (as we will
* then have to wait for another pass of the elevator
* algorithm before it will be read). We prefer to fail and
* be recalled to try an allocation in the next cylinder group.
*/
if (dtog(fs, bpref) != cg)
bpref = 0;
else
bpref = fragstoblks(fs, dtogd(fs, blknum(fs, bpref)));
mapp = &cg_clustersfree(cgp)[bpref / NBBY];
map = *mapp++;
bit = 1 << (bpref % NBBY);
for (run = 0, got = bpref; got < cgp->cg_nclusterblks; got++) {
if ((map & bit) == 0) {
run = 0;
} else {
run++;
if (run == len)
break;
}
if ((got & (NBBY - 1)) != (NBBY - 1)) {
bit <<= 1;
} else {
map = *mapp++;
bit = 1;
}
}
if (got >= cgp->cg_nclusterblks)
goto fail_lock;
/*
* Allocate the cluster that we have found.
*/
blksfree = cg_blksfree(cgp);
for (i = 1; i <= len; i++)
if (!ffs_isblock(fs, blksfree, got - run + i))
panic("ffs_clusteralloc: map mismatch");
bno = cgbase(fs, cg) + blkstofrags(fs, got - run + 1);
if (dtog(fs, bno) != cg)
panic("ffs_clusteralloc: allocated out of group");
len = blkstofrags(fs, len);
UFS_LOCK(ump);
for (i = 0; i < len; i += fs->fs_frag)
if (ffs_alloccgblk(ip, bp, bno + i, fs->fs_bsize) != bno + i)
panic("ffs_clusteralloc: lost block");
ACTIVECLEAR(fs, cg);
UFS_UNLOCK(ump);
bdwrite(bp);
return (bno);
fail_lock:
UFS_LOCK(ump);
fail:
brelse(bp);
return (0);
}
/*
* Determine whether an inode can be allocated.
*
* Check to see if an inode is available, and if it is,
* allocate it using the following policy:
* 1) allocate the requested inode.
* 2) allocate the next available inode after the requested
* inode in the specified cylinder group.
*/
static ufs2_daddr_t
ffs_nodealloccg(ip, cg, ipref, mode, unused)
struct inode *ip;
u_int cg;
ufs2_daddr_t ipref;
int mode;
int unused;
{
struct fs *fs;
struct cg *cgp;
struct buf *bp, *ibp;
struct ufsmount *ump;
u_int8_t *inosused, *loc;
struct ufs2_dinode *dp2;
int error, start, len, i;
fs = ip->i_fs;
ump = ip->i_ump;
if (fs->fs_cs(fs, cg).cs_nifree == 0)
return (0);
UFS_UNLOCK(ump);
error = bread(ip->i_devvp, fsbtodb(fs, cgtod(fs, cg)),
(int)fs->fs_cgsize, NOCRED, &bp);
if (error) {
brelse(bp);
UFS_LOCK(ump);
return (0);
}
cgp = (struct cg *)bp->b_data;
if (!cg_chkmagic(cgp) || cgp->cg_cs.cs_nifree == 0) {
brelse(bp);
UFS_LOCK(ump);
return (0);
}
bp->b_xflags |= BX_BKGRDWRITE;
cgp->cg_old_time = cgp->cg_time = time_second;
inosused = cg_inosused(cgp);
if (ipref) {
ipref %= fs->fs_ipg;
if (isclr(inosused, ipref))
goto gotit;
}
start = cgp->cg_irotor / NBBY;
len = howmany(fs->fs_ipg - cgp->cg_irotor, NBBY);
loc = memcchr(&inosused[start], 0xff, len);
if (loc == NULL) {
len = start + 1;
start = 0;
loc = memcchr(&inosused[start], 0xff, len);
if (loc == NULL) {
printf("cg = %d, irotor = %ld, fs = %s\n",
cg, (long)cgp->cg_irotor, fs->fs_fsmnt);
panic("ffs_nodealloccg: map corrupted");
/* NOTREACHED */
}
}
ipref = (loc - inosused) * NBBY + ffs(~*loc) - 1;
cgp->cg_irotor = ipref;
gotit:
/*
* Check to see if we need to initialize more inodes.
*/
ibp = NULL;
if (fs->fs_magic == FS_UFS2_MAGIC &&
ipref + INOPB(fs) > cgp->cg_initediblk &&
cgp->cg_initediblk < cgp->cg_niblk) {
ibp = getblk(ip->i_devvp, fsbtodb(fs,
ino_to_fsba(fs, cg * fs->fs_ipg + cgp->cg_initediblk)),
(int)fs->fs_bsize, 0, 0, 0);
bzero(ibp->b_data, (int)fs->fs_bsize);
dp2 = (struct ufs2_dinode *)(ibp->b_data);
for (i = 0; i < INOPB(fs); i++) {
dp2->di_gen = arc4random() / 2 + 1;
dp2++;
}
cgp->cg_initediblk += INOPB(fs);
}
UFS_LOCK(ump);
ACTIVECLEAR(fs, cg);
setbit(inosused, ipref);
cgp->cg_cs.cs_nifree--;
fs->fs_cstotal.cs_nifree--;
fs->fs_cs(fs, cg).cs_nifree--;
fs->fs_fmod = 1;
if ((mode & IFMT) == IFDIR) {
cgp->cg_cs.cs_ndir++;
fs->fs_cstotal.cs_ndir++;
fs->fs_cs(fs, cg).cs_ndir++;
}
UFS_UNLOCK(ump);
if (DOINGSOFTDEP(ITOV(ip)))
softdep_setup_inomapdep(bp, ip, cg * fs->fs_ipg + ipref, mode);
bdwrite(bp);
if (ibp != NULL)
bawrite(ibp);
return ((ino_t)(cg * fs->fs_ipg + ipref));
}
/*
* Free a block or fragment.
*
* The specified block or fragment is placed back in the
* free map. If a fragment is deallocated, a possible
* block reassembly is checked.
*/
static void
ffs_blkfree_cg(ump, fs, devvp, bno, size, inum, dephd)
struct ufsmount *ump;
struct fs *fs;
struct vnode *devvp;
ufs2_daddr_t bno;
long size;
ino_t inum;
struct workhead *dephd;
{
struct mount *mp;
struct cg *cgp;
struct buf *bp;
ufs1_daddr_t fragno, cgbno;
ufs2_daddr_t cgblkno;
int i, blk, frags, bbase;
u_int cg;
u_int8_t *blksfree;
struct cdev *dev;
cg = dtog(fs, bno);
if (devvp->v_type == VREG) {
/* devvp is a snapshot */
dev = VTOI(devvp)->i_devvp->v_rdev;
cgblkno = fragstoblks(fs, cgtod(fs, cg));
} else {
/* devvp is a normal disk device */
dev = devvp->v_rdev;
cgblkno = fsbtodb(fs, cgtod(fs, cg));
ASSERT_VOP_LOCKED(devvp, "ffs_blkfree_cg");
}
#ifdef INVARIANTS
if ((u_int)size > fs->fs_bsize || fragoff(fs, size) != 0 ||
fragnum(fs, bno) + numfrags(fs, size) > fs->fs_frag) {
printf("dev=%s, bno = %jd, bsize = %ld, size = %ld, fs = %s\n",
devtoname(dev), (intmax_t)bno, (long)fs->fs_bsize,
size, fs->fs_fsmnt);
panic("ffs_blkfree_cg: bad size");
}
#endif
if ((u_int)bno >= fs->fs_size) {
printf("bad block %jd, ino %lu\n", (intmax_t)bno,
(u_long)inum);
ffs_fserr(fs, inum, "bad block");
return;
}
if (bread(devvp, cgblkno, (int)fs->fs_cgsize, NOCRED, &bp)) {
brelse(bp);
return;
}
cgp = (struct cg *)bp->b_data;
if (!cg_chkmagic(cgp)) {
brelse(bp);
return;
}
bp->b_xflags |= BX_BKGRDWRITE;
cgp->cg_old_time = cgp->cg_time = time_second;
cgbno = dtogd(fs, bno);
blksfree = cg_blksfree(cgp);
UFS_LOCK(ump);
if (size == fs->fs_bsize) {
fragno = fragstoblks(fs, cgbno);
if (!ffs_isfreeblock(fs, blksfree, fragno)) {
if (devvp->v_type == VREG) {
UFS_UNLOCK(ump);
/* devvp is a snapshot */
brelse(bp);
return;
}
printf("dev = %s, block = %jd, fs = %s\n",
devtoname(dev), (intmax_t)bno, fs->fs_fsmnt);
panic("ffs_blkfree_cg: freeing free block");
}
ffs_setblock(fs, blksfree, fragno);
ffs_clusteracct(fs, cgp, fragno, 1);
cgp->cg_cs.cs_nbfree++;
fs->fs_cstotal.cs_nbfree++;
fs->fs_cs(fs, cg).cs_nbfree++;
} else {
bbase = cgbno - fragnum(fs, cgbno);
/*
* decrement the counts associated with the old frags
*/
blk = blkmap(fs, blksfree, bbase);
ffs_fragacct(fs, blk, cgp->cg_frsum, -1);
/*
* deallocate the fragment
*/
frags = numfrags(fs, size);
for (i = 0; i < frags; i++) {
if (isset(blksfree, cgbno + i)) {
printf("dev = %s, block = %jd, fs = %s\n",
devtoname(dev), (intmax_t)(bno + i),
fs->fs_fsmnt);
panic("ffs_blkfree_cg: freeing free frag");
}
setbit(blksfree, cgbno + i);
}
cgp->cg_cs.cs_nffree += i;
fs->fs_cstotal.cs_nffree += i;
fs->fs_cs(fs, cg).cs_nffree += i;
/*
* add back in counts associated with the new frags
*/
blk = blkmap(fs, blksfree, bbase);
ffs_fragacct(fs, blk, cgp->cg_frsum, 1);
/*
* if a complete block has been reassembled, account for it
*/
fragno = fragstoblks(fs, bbase);
if (ffs_isblock(fs, blksfree, fragno)) {
cgp->cg_cs.cs_nffree -= fs->fs_frag;
fs->fs_cstotal.cs_nffree -= fs->fs_frag;
fs->fs_cs(fs, cg).cs_nffree -= fs->fs_frag;
ffs_clusteracct(fs, cgp, fragno, 1);
cgp->cg_cs.cs_nbfree++;
fs->fs_cstotal.cs_nbfree++;
fs->fs_cs(fs, cg).cs_nbfree++;
}
}
fs->fs_fmod = 1;
ACTIVECLEAR(fs, cg);
UFS_UNLOCK(ump);
mp = UFSTOVFS(ump);
if (MOUNTEDSOFTDEP(mp) && devvp->v_type != VREG)
softdep_setup_blkfree(UFSTOVFS(ump), bp, bno,
numfrags(fs, size), dephd);
bdwrite(bp);
}
TASKQUEUE_DEFINE_THREAD(ffs_trim);
struct ffs_blkfree_trim_params {
struct task task;
struct ufsmount *ump;
struct vnode *devvp;
ufs2_daddr_t bno;
long size;
ino_t inum;
struct workhead *pdephd;
struct workhead dephd;
};
static void
ffs_blkfree_trim_task(ctx, pending)
void *ctx;
int pending;
{
struct ffs_blkfree_trim_params *tp;
tp = ctx;
ffs_blkfree_cg(tp->ump, tp->ump->um_fs, tp->devvp, tp->bno, tp->size,
tp->inum, tp->pdephd);
vn_finished_secondary_write(UFSTOVFS(tp->ump));
free(tp, M_TEMP);
}
static void
ffs_blkfree_trim_completed(bip)
struct bio *bip;
{
struct ffs_blkfree_trim_params *tp;
tp = bip->bio_caller2;
g_destroy_bio(bip);
TASK_INIT(&tp->task, 0, ffs_blkfree_trim_task, tp);
taskqueue_enqueue(taskqueue_ffs_trim, &tp->task);
}
void
ffs_blkfree(ump, fs, devvp, bno, size, inum, vtype, dephd)
struct ufsmount *ump;
struct fs *fs;
struct vnode *devvp;
ufs2_daddr_t bno;
long size;
ino_t inum;
enum vtype vtype;
struct workhead *dephd;
{
struct mount *mp;
struct bio *bip;
struct ffs_blkfree_trim_params *tp;
/*
* Check to see if a snapshot wants to claim the block.
* Check that devvp is a normal disk device, not a snapshot,
* it has a snapshot(s) associated with it, and one of the
* snapshots wants to claim the block.
*/
if (devvp->v_type != VREG &&
(devvp->v_vflag & VV_COPYONWRITE) &&
ffs_snapblkfree(fs, devvp, bno, size, inum, vtype, dephd)) {
return;
}
/*
* Nothing to delay if TRIM is disabled, or the operation is
* performed on the snapshot.
*/
if (!ump->um_candelete || devvp->v_type == VREG) {
ffs_blkfree_cg(ump, fs, devvp, bno, size, inum, dephd);
return;
}
/*
* Postpone the set of the free bit in the cg bitmap until the
* BIO_DELETE is completed. Otherwise, due to disk queue
* reordering, TRIM might be issued after we reuse the block
* and write some new data into it.
*/
tp = malloc(sizeof(struct ffs_blkfree_trim_params), M_TEMP, M_WAITOK);
tp->ump = ump;
tp->devvp = devvp;
tp->bno = bno;
tp->size = size;
tp->inum = inum;
if (dephd != NULL) {
LIST_INIT(&tp->dephd);
LIST_SWAP(dephd, &tp->dephd, worklist, wk_list);
tp->pdephd = &tp->dephd;
} else
tp->pdephd = NULL;
bip = g_alloc_bio();
bip->bio_cmd = BIO_DELETE;
bip->bio_offset = dbtob(fsbtodb(fs, bno));
bip->bio_done = ffs_blkfree_trim_completed;
bip->bio_length = size;
bip->bio_caller2 = tp;
mp = UFSTOVFS(ump);
vn_start_secondary_write(NULL, &mp, 0);
g_io_request(bip, (struct g_consumer *)devvp->v_bufobj.bo_private);
}
#ifdef INVARIANTS
/*
* Verify allocation of a block or fragment. Returns true if block or
* fragment is allocated, false if it is free.
*/
static int
ffs_checkblk(ip, bno, size)
struct inode *ip;
ufs2_daddr_t bno;
long size;
{
struct fs *fs;
struct cg *cgp;
struct buf *bp;
ufs1_daddr_t cgbno;
int i, error, frags, free;
u_int8_t *blksfree;
fs = ip->i_fs;
if ((u_int)size > fs->fs_bsize || fragoff(fs, size) != 0) {
printf("bsize = %ld, size = %ld, fs = %s\n",
(long)fs->fs_bsize, size, fs->fs_fsmnt);
panic("ffs_checkblk: bad size");
}
if ((u_int)bno >= fs->fs_size)
panic("ffs_checkblk: bad block %jd", (intmax_t)bno);
error = bread(ip->i_devvp, fsbtodb(fs, cgtod(fs, dtog(fs, bno))),
(int)fs->fs_cgsize, NOCRED, &bp);
if (error)
panic("ffs_checkblk: cg bread failed");
cgp = (struct cg *)bp->b_data;
if (!cg_chkmagic(cgp))
panic("ffs_checkblk: cg magic mismatch");
bp->b_xflags |= BX_BKGRDWRITE;
blksfree = cg_blksfree(cgp);
cgbno = dtogd(fs, bno);
if (size == fs->fs_bsize) {
free = ffs_isblock(fs, blksfree, fragstoblks(fs, cgbno));
} else {
frags = numfrags(fs, size);
for (free = 0, i = 0; i < frags; i++)
if (isset(blksfree, cgbno + i))
free++;
if (free != 0 && free != frags)
panic("ffs_checkblk: partially free fragment");
}
brelse(bp);
return (!free);
}
#endif /* INVARIANTS */
/*
* Free an inode.
*/
int
ffs_vfree(pvp, ino, mode)
struct vnode *pvp;
ino_t ino;
int mode;
{
struct inode *ip;
if (DOINGSOFTDEP(pvp)) {
softdep_freefile(pvp, ino, mode);
return (0);
}
ip = VTOI(pvp);
return (ffs_freefile(ip->i_ump, ip->i_fs, ip->i_devvp, ino, mode,
NULL));
}
/*
* Do the actual free operation.
* The specified inode is placed back in the free map.
*/
int
ffs_freefile(ump, fs, devvp, ino, mode, wkhd)
struct ufsmount *ump;
struct fs *fs;
struct vnode *devvp;
ino_t ino;
int mode;
struct workhead *wkhd;
{
struct cg *cgp;
struct buf *bp;
ufs2_daddr_t cgbno;
int error;
u_int cg;
u_int8_t *inosused;
struct cdev *dev;
cg = ino_to_cg(fs, ino);
if (devvp->v_type == VREG) {
/* devvp is a snapshot */
dev = VTOI(devvp)->i_devvp->v_rdev;
cgbno = fragstoblks(fs, cgtod(fs, cg));
} else {
/* devvp is a normal disk device */
dev = devvp->v_rdev;
cgbno = fsbtodb(fs, cgtod(fs, cg));
}
if (ino >= fs->fs_ipg * fs->fs_ncg)
panic("ffs_freefile: range: dev = %s, ino = %ju, fs = %s",
devtoname(dev), (uintmax_t)ino, fs->fs_fsmnt);
if ((error = bread(devvp, cgbno, (int)fs->fs_cgsize, NOCRED, &bp))) {
brelse(bp);
return (error);
}
cgp = (struct cg *)bp->b_data;
if (!cg_chkmagic(cgp)) {
brelse(bp);
return (0);
}
bp->b_xflags |= BX_BKGRDWRITE;
cgp->cg_old_time = cgp->cg_time = time_second;
inosused = cg_inosused(cgp);
ino %= fs->fs_ipg;
if (isclr(inosused, ino)) {
printf("dev = %s, ino = %ju, fs = %s\n", devtoname(dev),
(uintmax_t)(ino + cg * fs->fs_ipg), fs->fs_fsmnt);
if (fs->fs_ronly == 0)
panic("ffs_freefile: freeing free inode");
}
clrbit(inosused, ino);
if (ino < cgp->cg_irotor)
cgp->cg_irotor = ino;
cgp->cg_cs.cs_nifree++;
UFS_LOCK(ump);
fs->fs_cstotal.cs_nifree++;
fs->fs_cs(fs, cg).cs_nifree++;
if ((mode & IFMT) == IFDIR) {
cgp->cg_cs.cs_ndir--;
fs->fs_cstotal.cs_ndir--;
fs->fs_cs(fs, cg).cs_ndir--;
}
fs->fs_fmod = 1;
ACTIVECLEAR(fs, cg);
UFS_UNLOCK(ump);
if (MOUNTEDSOFTDEP(UFSTOVFS(ump)) && devvp->v_type != VREG)
softdep_setup_inofree(UFSTOVFS(ump), bp,
ino + cg * fs->fs_ipg, wkhd);
bdwrite(bp);
return (0);
}
/*
* Check to see if a file is free.
*/
int
ffs_checkfreefile(fs, devvp, ino)
struct fs *fs;
struct vnode *devvp;
ino_t ino;
{
struct cg *cgp;
struct buf *bp;
ufs2_daddr_t cgbno;
int ret;
u_int cg;
u_int8_t *inosused;
cg = ino_to_cg(fs, ino);
if (devvp->v_type == VREG) {
/* devvp is a snapshot */
cgbno = fragstoblks(fs, cgtod(fs, cg));
} else {
/* devvp is a normal disk device */
cgbno = fsbtodb(fs, cgtod(fs, cg));
}
if (ino >= fs->fs_ipg * fs->fs_ncg)
return (1);
if (bread(devvp, cgbno, (int)fs->fs_cgsize, NOCRED, &bp)) {
brelse(bp);
return (1);
}
cgp = (struct cg *)bp->b_data;
if (!cg_chkmagic(cgp)) {
brelse(bp);
return (1);
}
inosused = cg_inosused(cgp);
ino %= fs->fs_ipg;
ret = isclr(inosused, ino);
brelse(bp);
return (ret);
}
/*
* Find a block of the specified size in the specified cylinder group.
*
* It is a panic if a request is made to find a block if none are
* available.
*/
static ufs1_daddr_t
ffs_mapsearch(fs, cgp, bpref, allocsiz)
struct fs *fs;
struct cg *cgp;
ufs2_daddr_t bpref;
int allocsiz;
{
ufs1_daddr_t bno;
int start, len, loc, i;
int blk, field, subfield, pos;
u_int8_t *blksfree;
/*
* find the fragment by searching through the free block
* map for an appropriate bit pattern
*/
if (bpref)
start = dtogd(fs, bpref) / NBBY;
else
start = cgp->cg_frotor / NBBY;
blksfree = cg_blksfree(cgp);
len = howmany(fs->fs_fpg, NBBY) - start;
loc = scanc((u_int)len, (u_char *)&blksfree[start],
fragtbl[fs->fs_frag],
(u_char)(1 << (allocsiz - 1 + (fs->fs_frag % NBBY))));
if (loc == 0) {
len = start + 1;
start = 0;
loc = scanc((u_int)len, (u_char *)&blksfree[0],
fragtbl[fs->fs_frag],
(u_char)(1 << (allocsiz - 1 + (fs->fs_frag % NBBY))));
if (loc == 0) {
printf("start = %d, len = %d, fs = %s\n",
start, len, fs->fs_fsmnt);
panic("ffs_alloccg: map corrupted");
/* NOTREACHED */
}
}
bno = (start + len - loc) * NBBY;
cgp->cg_frotor = bno;
/*
* found the byte in the map
* sift through the bits to find the selected frag
*/
for (i = bno + NBBY; bno < i; bno += fs->fs_frag) {
blk = blkmap(fs, blksfree, bno);
blk <<= 1;
field = around[allocsiz];
subfield = inside[allocsiz];
for (pos = 0; pos <= fs->fs_frag - allocsiz; pos++) {
if ((blk & field) == subfield)
return (bno + pos);
field <<= 1;
subfield <<= 1;
}
}
printf("bno = %lu, fs = %s\n", (u_long)bno, fs->fs_fsmnt);
panic("ffs_alloccg: block not in map");
return (-1);
}
/*
* Fserr prints the name of a filesystem with an error diagnostic.
*
* The form of the error message is:
* fs: error message
*/
void
ffs_fserr(fs, inum, cp)
struct fs *fs;
ino_t inum;
char *cp;
{
struct thread *td = curthread; /* XXX */
struct proc *p = td->td_proc;
log(LOG_ERR, "pid %d (%s), uid %d inumber %ju on %s: %s\n",
p->p_pid, p->p_comm, td->td_ucred->cr_uid, (uintmax_t)inum,
fs->fs_fsmnt, cp);
}
/*
* This function provides the capability for the fsck program to
* update an active filesystem. Fourteen operations are provided:
*
* adjrefcnt(inode, amt) - adjusts the reference count on the
* specified inode by the specified amount. Under normal
* operation the count should always go down. Decrementing
* the count to zero will cause the inode to be freed.
* adjblkcnt(inode, amt) - adjust the number of blocks used by the
* inode by the specified amount.
* adjndir, adjbfree, adjifree, adjffree, adjnumclusters(amt) -
* adjust the superblock summary.
* freedirs(inode, count) - directory inodes [inode..inode + count - 1]
* are marked as free. Inodes should never have to be marked
* as in use.
* freefiles(inode, count) - file inodes [inode..inode + count - 1]
* are marked as free. Inodes should never have to be marked
* as in use.
* freeblks(blockno, size) - blocks [blockno..blockno + size - 1]
* are marked as free. Blocks should never have to be marked
* as in use.
* setflags(flags, set/clear) - the fs_flags field has the specified
* flags set (second parameter +1) or cleared (second parameter -1).
* setcwd(dirinode) - set the current directory to dirinode in the
* filesystem associated with the snapshot.
* setdotdot(oldvalue, newvalue) - Verify that the inode number for ".."
* in the current directory is oldvalue then change it to newvalue.
* unlink(nameptr, oldvalue) - Verify that the inode number associated
* with nameptr in the current directory is oldvalue then unlink it.
*
* The following functions may only be used on a quiescent filesystem
* by the soft updates journal. They are not safe to be run on an active
* filesystem.
*
* setinode(inode, dip) - the specified disk inode is replaced with the
* contents pointed to by dip.
* setbufoutput(fd, flags) - output associated with the specified file
* descriptor (which must reference the character device supporting
* the filesystem) switches from using physio to running through the
* buffer cache when flags is set to 1. The descriptor reverts to
* physio for output when flags is set to zero.
*/
static int sysctl_ffs_fsck(SYSCTL_HANDLER_ARGS);
SYSCTL_PROC(_vfs_ffs, FFS_ADJ_REFCNT, adjrefcnt, CTLFLAG_WR|CTLTYPE_STRUCT,
0, 0, sysctl_ffs_fsck, "S,fsck", "Adjust Inode Reference Count");
static SYSCTL_NODE(_vfs_ffs, FFS_ADJ_BLKCNT, adjblkcnt, CTLFLAG_WR,
sysctl_ffs_fsck, "Adjust Inode Used Blocks Count");
static SYSCTL_NODE(_vfs_ffs, FFS_ADJ_NDIR, adjndir, CTLFLAG_WR,
sysctl_ffs_fsck, "Adjust number of directories");
static SYSCTL_NODE(_vfs_ffs, FFS_ADJ_NBFREE, adjnbfree, CTLFLAG_WR,
sysctl_ffs_fsck, "Adjust number of free blocks");
static SYSCTL_NODE(_vfs_ffs, FFS_ADJ_NIFREE, adjnifree, CTLFLAG_WR,
sysctl_ffs_fsck, "Adjust number of free inodes");
static SYSCTL_NODE(_vfs_ffs, FFS_ADJ_NFFREE, adjnffree, CTLFLAG_WR,
sysctl_ffs_fsck, "Adjust number of free frags");
static SYSCTL_NODE(_vfs_ffs, FFS_ADJ_NUMCLUSTERS, adjnumclusters, CTLFLAG_WR,
sysctl_ffs_fsck, "Adjust number of free clusters");
static SYSCTL_NODE(_vfs_ffs, FFS_DIR_FREE, freedirs, CTLFLAG_WR,
sysctl_ffs_fsck, "Free Range of Directory Inodes");
static SYSCTL_NODE(_vfs_ffs, FFS_FILE_FREE, freefiles, CTLFLAG_WR,
sysctl_ffs_fsck, "Free Range of File Inodes");
static SYSCTL_NODE(_vfs_ffs, FFS_BLK_FREE, freeblks, CTLFLAG_WR,
sysctl_ffs_fsck, "Free Range of Blocks");
static SYSCTL_NODE(_vfs_ffs, FFS_SET_FLAGS, setflags, CTLFLAG_WR,
sysctl_ffs_fsck, "Change Filesystem Flags");
static SYSCTL_NODE(_vfs_ffs, FFS_SET_CWD, setcwd, CTLFLAG_WR,
sysctl_ffs_fsck, "Set Current Working Directory");
static SYSCTL_NODE(_vfs_ffs, FFS_SET_DOTDOT, setdotdot, CTLFLAG_WR,
sysctl_ffs_fsck, "Change Value of .. Entry");
static SYSCTL_NODE(_vfs_ffs, FFS_UNLINK, unlink, CTLFLAG_WR,
sysctl_ffs_fsck, "Unlink a Duplicate Name");
static SYSCTL_NODE(_vfs_ffs, FFS_SET_INODE, setinode, CTLFLAG_WR,
sysctl_ffs_fsck, "Update an On-Disk Inode");
static SYSCTL_NODE(_vfs_ffs, FFS_SET_BUFOUTPUT, setbufoutput, CTLFLAG_WR,
sysctl_ffs_fsck, "Set Buffered Writing for Descriptor");
#define DEBUG 1
#ifdef DEBUG
static int fsckcmds = 0;
SYSCTL_INT(_debug, OID_AUTO, fsckcmds, CTLFLAG_RW, &fsckcmds, 0, "");
#endif /* DEBUG */
static int buffered_write(struct file *, struct uio *, struct ucred *,
int, struct thread *);
static int
sysctl_ffs_fsck(SYSCTL_HANDLER_ARGS)
{
struct thread *td = curthread;
struct fsck_cmd cmd;
struct ufsmount *ump;
struct vnode *vp, *vpold, *dvp, *fdvp;
struct inode *ip, *dp;
struct mount *mp;
struct fs *fs;
ufs2_daddr_t blkno;
long blkcnt, blksize;
struct filedesc *fdp;
struct file *fp, *vfp;
int filetype, error;
static struct fileops *origops, bufferedops;
if (req->newlen > sizeof cmd)
return (EBADRPC);
if ((error = SYSCTL_IN(req, &cmd, sizeof cmd)) != 0)
return (error);
if (cmd.version != FFS_CMD_VERSION)
return (ERPCMISMATCH);
if ((error = getvnode(td->td_proc->p_fd, cmd.handle, CAP_FSCK,
&fp)) != 0)
return (error);
vp = fp->f_data;
if (vp->v_type != VREG && vp->v_type != VDIR) {
fdrop(fp, td);
return (EINVAL);
}
vn_start_write(vp, &mp, V_WAIT);
if (mp == 0 || strncmp(mp->mnt_stat.f_fstypename, "ufs", MFSNAMELEN)) {
vn_finished_write(mp);
fdrop(fp, td);
return (EINVAL);
}
ump = VFSTOUFS(mp);
if ((mp->mnt_flag & MNT_RDONLY) &&
ump->um_fsckpid != td->td_proc->p_pid) {
vn_finished_write(mp);
fdrop(fp, td);
return (EROFS);
}
fs = ump->um_fs;
filetype = IFREG;
switch (oidp->oid_number) {
case FFS_SET_FLAGS:
#ifdef DEBUG
if (fsckcmds)
printf("%s: %s flags\n", mp->mnt_stat.f_mntonname,
cmd.size > 0 ? "set" : "clear");
#endif /* DEBUG */
if (cmd.size > 0)
fs->fs_flags |= (long)cmd.value;
else
fs->fs_flags &= ~(long)cmd.value;
break;
case FFS_ADJ_REFCNT:
#ifdef DEBUG
if (fsckcmds) {
printf("%s: adjust inode %jd link count by %jd\n",
mp->mnt_stat.f_mntonname, (intmax_t)cmd.value,
(intmax_t)cmd.size);
}
#endif /* DEBUG */
if ((error = ffs_vget(mp, (ino_t)cmd.value, LK_EXCLUSIVE, &vp)))
break;
ip = VTOI(vp);
ip->i_nlink += cmd.size;
DIP_SET(ip, i_nlink, ip->i_nlink);
ip->i_effnlink += cmd.size;
ip->i_flag |= IN_CHANGE | IN_MODIFIED;
error = ffs_update(vp, 1);
if (DOINGSOFTDEP(vp))
softdep_change_linkcnt(ip);
vput(vp);
break;
case FFS_ADJ_BLKCNT:
#ifdef DEBUG
if (fsckcmds) {
printf("%s: adjust inode %jd block count by %jd\n",
mp->mnt_stat.f_mntonname, (intmax_t)cmd.value,
(intmax_t)cmd.size);
}
#endif /* DEBUG */
if ((error = ffs_vget(mp, (ino_t)cmd.value, LK_EXCLUSIVE, &vp)))
break;
ip = VTOI(vp);
DIP_SET(ip, i_blocks, DIP(ip, i_blocks) + cmd.size);
ip->i_flag |= IN_CHANGE | IN_MODIFIED;
error = ffs_update(vp, 1);
vput(vp);
break;
case FFS_DIR_FREE:
filetype = IFDIR;
/* fall through */
case FFS_FILE_FREE:
#ifdef DEBUG
if (fsckcmds) {
if (cmd.size == 1)
printf("%s: free %s inode %ju\n",
mp->mnt_stat.f_mntonname,
filetype == IFDIR ? "directory" : "file",
(uintmax_t)cmd.value);
else
printf("%s: free %s inodes %ju-%ju\n",
mp->mnt_stat.f_mntonname,
filetype == IFDIR ? "directory" : "file",
(uintmax_t)cmd.value,
(uintmax_t)(cmd.value + cmd.size - 1));
}
#endif /* DEBUG */
while (cmd.size > 0) {
if ((error = ffs_freefile(ump, fs, ump->um_devvp,
cmd.value, filetype, NULL)))
break;
cmd.size -= 1;
cmd.value += 1;
}
break;
case FFS_BLK_FREE:
#ifdef DEBUG
if (fsckcmds) {
if (cmd.size == 1)
printf("%s: free block %jd\n",
mp->mnt_stat.f_mntonname,
(intmax_t)cmd.value);
else
printf("%s: free blocks %jd-%jd\n",
mp->mnt_stat.f_mntonname,
(intmax_t)cmd.value,
(intmax_t)cmd.value + cmd.size - 1);
}
#endif /* DEBUG */
blkno = cmd.value;
blkcnt = cmd.size;
blksize = fs->fs_frag - (blkno % fs->fs_frag);
while (blkcnt > 0) {
if (blksize > blkcnt)
blksize = blkcnt;
ffs_blkfree(ump, fs, ump->um_devvp, blkno,
blksize * fs->fs_fsize, ROOTINO, VDIR, NULL);
blkno += blksize;
blkcnt -= blksize;
blksize = fs->fs_frag;
}
break;
/*
* Adjust superblock summaries. fsck(8) is expected to
* submit deltas when necessary.
*/
case FFS_ADJ_NDIR:
#ifdef DEBUG
if (fsckcmds) {
printf("%s: adjust number of directories by %jd\n",
mp->mnt_stat.f_mntonname, (intmax_t)cmd.value);
}
#endif /* DEBUG */
fs->fs_cstotal.cs_ndir += cmd.value;
break;
case FFS_ADJ_NBFREE:
#ifdef DEBUG
if (fsckcmds) {
printf("%s: adjust number of free blocks by %+jd\n",
mp->mnt_stat.f_mntonname, (intmax_t)cmd.value);
}
#endif /* DEBUG */
fs->fs_cstotal.cs_nbfree += cmd.value;
break;
case FFS_ADJ_NIFREE:
#ifdef DEBUG
if (fsckcmds) {
printf("%s: adjust number of free inodes by %+jd\n",
mp->mnt_stat.f_mntonname, (intmax_t)cmd.value);
}
#endif /* DEBUG */
fs->fs_cstotal.cs_nifree += cmd.value;
break;
case FFS_ADJ_NFFREE:
#ifdef DEBUG
if (fsckcmds) {
printf("%s: adjust number of free frags by %+jd\n",
mp->mnt_stat.f_mntonname, (intmax_t)cmd.value);
}
#endif /* DEBUG */
fs->fs_cstotal.cs_nffree += cmd.value;
break;
case FFS_ADJ_NUMCLUSTERS:
#ifdef DEBUG
if (fsckcmds) {
printf("%s: adjust number of free clusters by %+jd\n",
mp->mnt_stat.f_mntonname, (intmax_t)cmd.value);
}
#endif /* DEBUG */
fs->fs_cstotal.cs_numclusters += cmd.value;
break;
case FFS_SET_CWD:
#ifdef DEBUG
if (fsckcmds) {
printf("%s: set current directory to inode %jd\n",
mp->mnt_stat.f_mntonname, (intmax_t)cmd.value);
}
#endif /* DEBUG */
if ((error = ffs_vget(mp, (ino_t)cmd.value, LK_SHARED, &vp)))
break;
AUDIT_ARG_VNODE1(vp);
if ((error = change_dir(vp, td)) != 0) {
vput(vp);
break;
}
VOP_UNLOCK(vp, 0);
fdp = td->td_proc->p_fd;
FILEDESC_XLOCK(fdp);
vpold = fdp->fd_cdir;
fdp->fd_cdir = vp;
FILEDESC_XUNLOCK(fdp);
vrele(vpold);
break;
case FFS_SET_DOTDOT:
#ifdef DEBUG
if (fsckcmds) {
printf("%s: change .. in cwd from %jd to %jd\n",
mp->mnt_stat.f_mntonname, (intmax_t)cmd.value,
(intmax_t)cmd.size);
}
#endif /* DEBUG */
/*
* First we have to get and lock the parent directory
* to which ".." points.
*/
error = ffs_vget(mp, (ino_t)cmd.value, LK_EXCLUSIVE, &fdvp);
if (error)
break;
/*
* Now we get and lock the child directory containing "..".
*/
FILEDESC_SLOCK(td->td_proc->p_fd);
dvp = td->td_proc->p_fd->fd_cdir;
FILEDESC_SUNLOCK(td->td_proc->p_fd);
if ((error = vget(dvp, LK_EXCLUSIVE, td)) != 0) {
vput(fdvp);
break;
}
dp = VTOI(dvp);
dp->i_offset = 12; /* XXX mastertemplate.dot_reclen */
error = ufs_dirrewrite(dp, VTOI(fdvp), (ino_t)cmd.size,
DT_DIR, 0);
cache_purge(fdvp);
cache_purge(dvp);
vput(dvp);
vput(fdvp);
break;
case FFS_UNLINK:
#ifdef DEBUG
if (fsckcmds) {
char buf[32];
if (copyinstr((char *)(intptr_t)cmd.value, buf,32,NULL))
strncpy(buf, "Name_too_long", 32);
printf("%s: unlink %s (inode %jd)\n",
mp->mnt_stat.f_mntonname, buf, (intmax_t)cmd.size);
}
#endif /* DEBUG */
/*
* kern_unlinkat will do its own start/finish writes and
* they do not nest, so drop ours here. Setting mp == NULL
* indicates that vn_finished_write is not needed down below.
*/
vn_finished_write(mp);
mp = NULL;
error = kern_unlinkat(td, AT_FDCWD, (char *)(intptr_t)cmd.value,
UIO_USERSPACE, (ino_t)cmd.size);
break;
case FFS_SET_INODE:
if (ump->um_fsckpid != td->td_proc->p_pid) {
error = EPERM;
break;
}
#ifdef DEBUG
if (fsckcmds) {
printf("%s: update inode %jd\n",
mp->mnt_stat.f_mntonname, (intmax_t)cmd.value);
}
#endif /* DEBUG */
if ((error = ffs_vget(mp, (ino_t)cmd.value, LK_EXCLUSIVE, &vp)))
break;
AUDIT_ARG_VNODE1(vp);
ip = VTOI(vp);
if (ip->i_ump->um_fstype == UFS1)
error = copyin((void *)(intptr_t)cmd.size, ip->i_din1,
sizeof(struct ufs1_dinode));
else
error = copyin((void *)(intptr_t)cmd.size, ip->i_din2,
sizeof(struct ufs2_dinode));
if (error) {
vput(vp);
break;
}
ip->i_flag |= IN_CHANGE | IN_MODIFIED;
error = ffs_update(vp, 1);
vput(vp);
break;
case FFS_SET_BUFOUTPUT:
if (ump->um_fsckpid != td->td_proc->p_pid) {
error = EPERM;
break;
}
if (VTOI(vp)->i_ump != ump) {
error = EINVAL;
break;
}
#ifdef DEBUG
if (fsckcmds) {
printf("%s: %s buffered output for descriptor %jd\n",
mp->mnt_stat.f_mntonname,
cmd.size == 1 ? "enable" : "disable",
(intmax_t)cmd.value);
}
#endif /* DEBUG */
if ((error = getvnode(td->td_proc->p_fd, cmd.value,
CAP_FSCK, &vfp)) != 0)
break;
if (vfp->f_vnode->v_type != VCHR) {
fdrop(vfp, td);
error = EINVAL;
break;
}
if (origops == NULL) {
origops = vfp->f_ops;
bcopy((void *)origops, (void *)&bufferedops,
sizeof(bufferedops));
bufferedops.fo_write = buffered_write;
}
if (cmd.size == 1)
atomic_store_rel_ptr((volatile uintptr_t *)&vfp->f_ops,
(uintptr_t)&bufferedops);
else
atomic_store_rel_ptr((volatile uintptr_t *)&vfp->f_ops,
(uintptr_t)origops);
fdrop(vfp, td);
break;
default:
#ifdef DEBUG
if (fsckcmds) {
printf("Invalid request %d from fsck\n",
oidp->oid_number);
}
#endif /* DEBUG */
error = EINVAL;
break;
}
fdrop(fp, td);
vn_finished_write(mp);
return (error);
}
/*
* Function to switch a descriptor to use the buffer cache to stage
* its I/O. This is needed so that writes to the filesystem device
* will give snapshots a chance to copy modified blocks for which it
* needs to retain copies.
*/
static int
buffered_write(fp, uio, active_cred, flags, td)
struct file *fp;
struct uio *uio;
struct ucred *active_cred;
int flags;
struct thread *td;
{
struct vnode *devvp;
struct inode *ip;
struct buf *bp;
struct fs *fs;
int error;
daddr_t lbn;
/*
* The devvp is associated with the /dev filesystem. To discover
* the filesystem with which the device is associated, we depend
* on the application setting the current directory to a location
* within the filesystem being written. Yes, this is an ugly hack.
*/
devvp = fp->f_vnode;
ip = VTOI(td->td_proc->p_fd->fd_cdir);
if (ip->i_devvp != devvp)
return (EINVAL);
fs = ip->i_fs;
foffset_lock_uio(fp, uio, flags);
vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
#ifdef DEBUG
if (fsckcmds) {
printf("%s: buffered write for block %jd\n",
fs->fs_fsmnt, (intmax_t)btodb(uio->uio_offset));
}
#endif /* DEBUG */
/*
* All I/O must be contained within a filesystem block, start on
* a fragment boundary, and be a multiple of fragments in length.
*/
if (uio->uio_resid > fs->fs_bsize - (uio->uio_offset % fs->fs_bsize) ||
fragoff(fs, uio->uio_offset) != 0 ||
fragoff(fs, uio->uio_resid) != 0) {
error = EINVAL;
goto out;
}
lbn = numfrags(fs, uio->uio_offset);
bp = getblk(devvp, lbn, uio->uio_resid, 0, 0, 0);
bp->b_flags |= B_RELBUF;
if ((error = uiomove((char *)bp->b_data, uio->uio_resid, uio)) != 0) {
brelse(bp);
goto out;
}
error = bwrite(bp);
out:
VOP_UNLOCK(devvp, 0);
foffset_unlock_uio(fp, uio, flags | FOF_NEXTOFF);
return (error);
}
Reference in New Issue View Git Blame Copy Permalink