static __inline nandfs_lbn_t
lbn_offset(struct nandfs_device *fsdev, int level)
{
nandfs_lbn_t res;
for (res = 1; level > 0; level--)
res *= MNINDIR(fsdev);
return (res);
}
static nandfs_lbn_t
blocks_inside(struct nandfs_device *fsdev, int level, struct nandfs_indir *nip)
{
nandfs_lbn_t blocks;
for (blocks = 1; level >= SINGLE; level--, nip++) {
MPASS(nip->in_off >= 0 && nip->in_off < MNINDIR(fsdev));
blocks += nip->in_off * lbn_offset(fsdev, level);
}
return (blocks);
}
nandfs_lbn_t
get_maxfilesize(struct nandfs_device *fsdev)
{
struct nandfs_indir f[NIADDR];
nandfs_lbn_t max;
int i;
max = NDADDR;
for (i = 0; i < NIADDR; i++) {
f[i].in_off = MNINDIR(fsdev) - 1;
max += blocks_inside(fsdev, i, f);
}
max *= fsdev->nd_blocksize;
return (max);
}
/*
* Create an array of logical block number/offset pairs which represent the
* path of indirect blocks required to access a data block. The first "pair"
* contains the logical block number of the appropriate single, double or
* triple indirect block and the offset into the inode indirect block array.
* Note, the logical block number of the inode single/double/triple indirect
* block appears twice in the array, once with the offset into the i_ib and
* once with the offset into the page itself.
*/
int
ext2_getlbns(struct vnode *vp, daddr_t bn, struct indir *ap, int *nump)
{
long blockcnt;
e2fs_lbn_t metalbn, realbn;
struct ext2mount *ump;
int i, numlevels, off;
int64_t qblockcnt;
ump = VFSTOEXT2(vp->v_mount);
if (nump)
*nump = 0;
numlevels = 0;
realbn = bn;
if ((long)bn < 0)
bn = -(long)bn;
/* The first NDADDR blocks are direct blocks. */
if (bn < NDADDR)
return (0);
/*
* Determine the number of levels of indirection. After this loop
* is done, blockcnt indicates the number of data blocks possible
* at the previous level of indirection, and NIADDR - i is the number
* of levels of indirection needed to locate the requested block.
*/
for (blockcnt = 1, i = NIADDR, bn -= NDADDR;; i--, bn -= blockcnt) {
if (i == 0)
return (EFBIG);
/*
* Use int64_t's here to avoid overflow for triple indirect
* blocks when longs have 32 bits and the block size is more
* than 4K.
*/
qblockcnt = (int64_t)blockcnt * MNINDIR(ump);
if (bn < qblockcnt)
break;
blockcnt = qblockcnt;
}
/* Calculate the address of the first meta-block. */
if (realbn >= 0)
metalbn = -(realbn - bn + NIADDR - i);
else
metalbn = -(-realbn - bn + NIADDR - i);
/*
* At each iteration, off is the offset into the bap array which is
* an array of disk addresses at the current level of indirection.
* The logical block number and the offset in that block are stored
* into the argument array.
*/
ap->in_lbn = metalbn;
ap->in_off = off = NIADDR - i;
ap++;
for (++numlevels; i <= NIADDR; i++) {
/* If searching for a meta-data block, quit when found. */
if (metalbn == realbn)
break;
off = (bn / blockcnt) % MNINDIR(ump);
++numlevels;
ap->in_lbn = metalbn;
ap->in_off = off;
++ap;
metalbn -= -1 + off * blockcnt;
blockcnt /= MNINDIR(ump);
}
if (nump)
*nump = numlevels;
return (0);
}
int
ext2_bmaparray(struct vnode *vp, daddr_t bn, daddr_t *bnp, int *runp, int *runb)
{
struct inode *ip;
struct buf *bp;
struct ext2mount *ump;
struct mount *mp;
struct vnode *devvp;
struct indir a[NIADDR+1], *ap;
daddr_t daddr;
e2fs_lbn_t metalbn;
int error, num, maxrun = 0, bsize;
int *nump;
ap = NULL;
ip = VTOI(vp);
mp = vp->v_mount;
ump = VFSTOEXT2(mp);
devvp = ump->um_devvp;
bsize = EXT2_BLOCK_SIZE(ump->um_e2fs);
if (runp) {
maxrun = mp->mnt_iosize_max / bsize - 1;
*runp = 0;
}
if (runb) {
*runb = 0;
}
ap = a;
nump = #
error = ext2_getlbns(vp, bn, ap, nump);
if (error)
return (error);
num = *nump;
if (num == 0) {
*bnp = blkptrtodb(ump, ip->i_db[bn]);
if (*bnp == 0) {
*bnp = -1;
} else if (runp) {
daddr_t bnb = bn;
for (++bn; bn < NDADDR && *runp < maxrun &&
is_sequential(ump, ip->i_db[bn - 1], ip->i_db[bn]);
++bn, ++*runp);
bn = bnb;
if (runb && (bn > 0)) {
for (--bn; (bn >= 0) && (*runb < maxrun) &&
is_sequential(ump, ip->i_db[bn],
ip->i_db[bn + 1]);
--bn, ++*runb);
}
}
return (0);
}
/* Get disk address out of indirect block array */
daddr = ip->i_ib[ap->in_off];
for (bp = NULL, ++ap; --num; ++ap) {
/*
* Exit the loop if there is no disk address assigned yet and
* the indirect block isn't in the cache, or if we were
* looking for an indirect block and we've found it.
*/
metalbn = ap->in_lbn;
if ((daddr == 0 && !incore(&vp->v_bufobj, metalbn)) || metalbn == bn)
break;
/*
* If we get here, we've either got the block in the cache
* or we have a disk address for it, go fetch it.
*/
if (bp)
bqrelse(bp);
bp = getblk(vp, metalbn, bsize, 0, 0, 0);
if ((bp->b_flags & B_CACHE) == 0) {
#ifdef INVARIANTS
if (!daddr)
panic("ext2_bmaparray: indirect block not in cache");
#endif
bp->b_blkno = blkptrtodb(ump, daddr);
bp->b_iocmd = BIO_READ;
bp->b_flags &= ~B_INVAL;
bp->b_ioflags &= ~BIO_ERROR;
vfs_busy_pages(bp, 0);
bp->b_iooffset = dbtob(bp->b_blkno);
bstrategy(bp);
curthread->td_ru.ru_inblock++;
error = bufwait(bp);
if (error) {
brelse(bp);
return (error);
}
}
daddr = ((e2fs_daddr_t *)bp->b_data)[ap->in_off];
if (num == 1 && daddr && runp) {
for (bn = ap->in_off + 1;
bn < MNINDIR(ump) && *runp < maxrun &&
is_sequential(ump,
((e2fs_daddr_t *)bp->b_data)[bn - 1],
((e2fs_daddr_t *)bp->b_data)[bn]);
++bn, ++*runp);
bn = ap->in_off;
if (runb && bn) {
for (--bn; bn >= 0 && *runb < maxrun &&
is_sequential(ump,
((e2fs_daddr_t *)bp->b_data)[bn],
((e2fs_daddr_t *)bp->b_data)[bn + 1]);
--bn, ++*runb);
}
}
}
if (bp)
bqrelse(bp);
/*
* Since this is FFS independent code, we are out of scope for the
* definitions of BLK_NOCOPY and BLK_SNAP, but we do know that they
* will fall in the range 1..um_seqinc, so we use that test and
* return a request for a zeroed out buffer if attempts are made
* to read a BLK_NOCOPY or BLK_SNAP block.
*/
if ((ip->i_flags & SF_SNAPSHOT) && daddr > 0 && daddr < ump->um_seqinc){
*bnp = -1;
return (0);
}
*bnp = blkptrtodb(ump, daddr);
if (*bnp == 0) {
*bnp = -1;
}
return (0);
}
/*
* Last reference to an inode. If necessary, write or delete it.
*/
int
ufs_inactive(void *v)
{
struct vop_inactive_args /* {
struct vnode *a_vp;
struct bool *a_recycle;
} */ *ap = v;
struct vnode *vp = ap->a_vp;
struct inode *ip = VTOI(vp);
struct mount *transmp;
mode_t mode;
int error = 0;
int logged = 0;
UFS_WAPBL_JUNLOCK_ASSERT(vp->v_mount);
transmp = vp->v_mount;
fstrans_start(transmp, FSTRANS_SHARED);
/*
* Ignore inodes related to stale file handles.
*/
if (ip->i_mode == 0)
goto out;
if (ip->i_ffs_effnlink == 0 && DOINGSOFTDEP(vp))
softdep_releasefile(ip);
if (ip->i_nlink <= 0 && (vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
error = UFS_WAPBL_BEGIN(vp->v_mount);
if (error)
goto out;
logged = 1;
#ifdef QUOTA
(void)chkiq(ip, -1, NOCRED, 0);
#endif
#ifdef UFS_EXTATTR
ufs_extattr_vnode_inactive(vp, curlwp);
#endif
if (ip->i_size != 0) {
/*
* When journaling, only truncate one indirect block
* at a time
*/
if (vp->v_mount->mnt_wapbl) {
uint64_t incr = MNINDIR(ip->i_ump) <<
vp->v_mount->mnt_fs_bshift; /* Power of 2 */
uint64_t base = NDADDR <<
vp->v_mount->mnt_fs_bshift;
while (!error && ip->i_size > base + incr) {
/*
* round down to next full indirect
* block boundary.
*/
uint64_t nsize = base +
((ip->i_size - base - 1) &
~(incr - 1));
error = UFS_TRUNCATE(vp, nsize, 0,
NOCRED);
if (error)
break;
UFS_WAPBL_END(vp->v_mount);
error = UFS_WAPBL_BEGIN(vp->v_mount);
if (error)
goto out;
}
}
if (!error)
error = UFS_TRUNCATE(vp, (off_t)0, 0, NOCRED);
}
/*
* Setting the mode to zero needs to wait for the inode
* to be written just as does a change to the link count.
* So, rather than creating a new entry point to do the
* same thing, we just use softdep_change_linkcnt().
*/
DIP_ASSIGN(ip, rdev, 0);
mode = ip->i_mode;
ip->i_mode = 0;
DIP_ASSIGN(ip, mode, 0);
ip->i_flag |= IN_CHANGE | IN_UPDATE;
mutex_enter(&vp->v_interlock);
vp->v_iflag |= VI_FREEING;
mutex_exit(&vp->v_interlock);
if (DOINGSOFTDEP(vp))
softdep_change_linkcnt(ip);
UFS_VFREE(vp, ip->i_number, mode);
}
if (ip->i_flag & (IN_CHANGE | IN_UPDATE | IN_MODIFIED)) {
if (!logged++) {
int err;
err = UFS_WAPBL_BEGIN(vp->v_mount);
if (err)
goto out;
}
UFS_UPDATE(vp, NULL, NULL, 0);
}
if (logged)
UFS_WAPBL_END(vp->v_mount);
out:
/*
* If we are done with the inode, reclaim it
* so that it can be reused immediately.
*/
*ap->a_recycle = (ip->i_mode == 0);
VOP_UNLOCK(vp, 0);
fstrans_done(transmp);
return (error);
}
/*
* Indirect blocks are now on the vnode for the file. They are given negative
* logical block numbers. Indirect blocks are addressed by the negative
* address of the first data block to which they point. Double indirect blocks
* are addressed by one less than the address of the first indirect block to
* which they point. Triple indirect blocks are addressed by one less than
* the address of the first double indirect block to which they point.
*
* ufs_bmaparray does the bmap conversion, and if requested returns the
* array of logical blocks which must be traversed to get to a block.
* Each entry contains the offset into that block that gets you to the
* next block and the disk address of the block (if it is assigned).
*/
int
ufs_bmaparray(struct vnode *vp, daddr64_t bn, daddr64_t *bnp, struct indir *ap,
int *nump, int *runp)
{
struct inode *ip;
struct buf *bp;
struct ufsmount *ump;
struct mount *mp;
struct vnode *devvp;
struct indir a[NIADDR+1], *xap;
daddr64_t daddr, metalbn;
int error, maxrun = 0, num;
ip = VTOI(vp);
mp = vp->v_mount;
ump = VFSTOUFS(mp);
#ifdef DIAGNOSTIC
if ((ap != NULL && nump == NULL) || (ap == NULL && nump != NULL))
panic("ufs_bmaparray: invalid arguments");
#endif
if (runp) {
/*
* XXX
* If MAXBSIZE is the largest transfer the disks can handle,
* we probably want maxrun to be 1 block less so that we
* don't create a block larger than the device can handle.
*/
*runp = 0;
maxrun = MAXBSIZE / mp->mnt_stat.f_iosize - 1;
}
xap = ap == NULL ? a : ap;
if (!nump)
nump = #
if ((error = ufs_getlbns(vp, bn, xap, nump)) != 0)
return (error);
num = *nump;
if (num == 0) {
*bnp = blkptrtodb(ump, DIP(ip, db[bn]));
if (*bnp == 0)
*bnp = -1;
else if (runp)
for (++bn; bn < NDADDR && *runp < maxrun &&
is_sequential(ump, DIP(ip, db[bn - 1]),
DIP(ip, db[bn]));
++bn, ++*runp);
return (0);
}
/* Get disk address out of indirect block array */
daddr = DIP(ip, ib[xap->in_off]);
devvp = VFSTOUFS(vp->v_mount)->um_devvp;
for (bp = NULL, ++xap; --num; ++xap) {
/*
* Exit the loop if there is no disk address assigned yet and
* the indirect block isn't in the cache, or if we were
* looking for an indirect block and we've found it.
*/
metalbn = xap->in_lbn;
if ((daddr == 0 && !incore(vp, metalbn)) || metalbn == bn)
break;
/*
* If we get here, we've either got the block in the cache
* or we have a disk address for it, go fetch it.
*/
if (bp)
brelse(bp);
xap->in_exists = 1;
bp = getblk(vp, metalbn, mp->mnt_stat.f_iosize, 0, 0);
if (bp->b_flags & (B_DONE | B_DELWRI)) {
;
}
#ifdef DIAGNOSTIC
else if (!daddr)
panic("ufs_bmaparray: indirect block not in cache");
#endif
else {
bp->b_blkno = blkptrtodb(ump, daddr);
bp->b_flags |= B_READ;
bcstats.pendingreads++;
bcstats.numreads++;
VOP_STRATEGY(bp);
curproc->p_ru.ru_inblock++; /* XXX */
if ((error = biowait(bp)) != 0) {
brelse(bp);
return (error);
}
}
#ifdef FFS2
if (ip->i_ump->um_fstype == UM_UFS2) {
daddr = ((int64_t *)bp->b_data)[xap->in_off];
if (num == 1 && daddr && runp)
for (bn = xap->in_off + 1;
bn < MNINDIR(ump) && *runp < maxrun &&
is_sequential(ump,
((int64_t *)bp->b_data)[bn - 1],
((int64_t *)bp->b_data)[bn]);
++bn, ++*runp);
continue;
}
#endif /* FFS2 */
daddr = ((int32_t *)bp->b_data)[xap->in_off];
if (num == 1 && daddr && runp)
for (bn = xap->in_off + 1;
bn < MNINDIR(ump) && *runp < maxrun &&
is_sequential(ump,
((int32_t *)bp->b_data)[bn - 1],
((int32_t *)bp->b_data)[bn]);
++bn, ++*runp);
}
if (bp)
brelse(bp);
daddr = blkptrtodb(ump, daddr);
*bnp = daddr == 0 ? -1 : daddr;
return (0);
}
/*
* Create an array of logical block number/offset pairs which represent the
* path of indirect blocks required to access a data block. The first "pair"
* contains the logical block number of the appropriate single, double or
* triple indirect block and the offset into the inode indirect block array.
* Note, the logical block number of the inode single/double/triple indirect
* block appears twice in the array, once with the offset into the i_ffs_ib and
* once with the offset into the page itself.
*/
int
ufs_getlbns(struct vnode *vp, daddr64_t bn, struct indir *ap, int *nump)
{
daddr64_t metalbn, realbn;
struct ufsmount *ump;
int64_t blockcnt;
int i, numlevels, off;
ump = VFSTOUFS(vp->v_mount);
if (nump)
*nump = 0;
numlevels = 0;
realbn = bn;
if (bn < 0)
bn = -bn;
#ifdef DIAGNOSTIC
if (realbn < 0 && realbn > -NDADDR) {
panic ("ufs_getlbns: Invalid indirect block %lld specified",
realbn);
}
#endif
/* The first NDADDR blocks are direct blocks. */
if (bn < NDADDR)
return (0);
/*
* Determine the number of levels of indirection. After this loop
* is done, blockcnt indicates the number of data blocks possible
* at the given level of indirection, and NIADDR - i is the number
* of levels of indirection needed to locate the requested block.
*/
for (blockcnt = 1, i = NIADDR, bn -= NDADDR;; i--, bn -= blockcnt) {
if (i == 0)
return (EFBIG);
blockcnt *= MNINDIR(ump);
if (bn < blockcnt)
break;
}
/* Calculate the address of the first meta-block. */
if (realbn >= 0)
metalbn = -(realbn - bn + NIADDR - i);
else
metalbn = -(-realbn - bn + NIADDR - i);
/*
* At each iteration, off is the offset into the bap array which is
* an array of disk addresses at the current level of indirection.
* The logical block number and the offset in that block are stored
* into the argument array.
*/
ap->in_lbn = metalbn;
ap->in_off = off = NIADDR - i;
ap->in_exists = 0;
ap++;
for (++numlevels; i <= NIADDR; i++) {
/* If searching for a meta-data block, quit when found. */
if (metalbn == realbn)
break;
blockcnt /= MNINDIR(ump);
off = (bn / blockcnt) % MNINDIR(ump);
++numlevels;
ap->in_lbn = metalbn;
ap->in_off = off;
ap->in_exists = 0;
++ap;
metalbn -= -1 + off * blockcnt;
}
#ifdef DIAGNOSTIC
if (realbn < 0 && metalbn != realbn) {
panic("ufs_getlbns: indirect block %lld not found", realbn);
}
#endif
if (nump)
*nump = numlevels;
return (0);
}
/*
* Last reference to an inode. If necessary, write or delete it.
*/
int
ufs_inactive(void *v)
{
struct vop_inactive_args *ap = v;
struct vnode *vp = ap->a_vp;
struct inode *ip = VTOI(vp);
struct fs *fs = ip->i_fs;
struct proc *p = curproc;
mode_t mode;
int error = 0, logged = 0, truncate_error = 0;
#ifdef DIAGNOSTIC
extern int prtactive;
if (prtactive && vp->v_usecount != 0)
vprint("ufs_inactive: pushing active", vp);
#endif
UFS_WAPBL_JUNLOCK_ASSERT(vp->v_mount);
/*
* Ignore inodes related to stale file handles.
*/
if (ip->i_din1 == NULL || DIP(ip, mode) == 0)
goto out;
if (DIP(ip, nlink) <= 0 && (vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
error = UFS_WAPBL_BEGIN(vp->v_mount);
if (error)
goto out;
logged = 1;
if (getinoquota(ip) == 0)
(void)ufs_quota_free_inode(ip, NOCRED);
if (DIP(ip, size) != 0 && vp->v_mount->mnt_wapbl) {
/*
* When journaling, only truncate one indirect block at
* a time.
*/
uint64_t incr = MNINDIR(ip->i_ump) << fs->fs_bshift;
uint64_t base = NDADDR << fs->fs_bshift;
while (!error && DIP(ip, size) > base + incr) {
/*
* round down to next full indirect block
* boundary.
*/
uint64_t nsize = base +
((DIP(ip, size) - base - 1) &
~(incr - 1));
error = UFS_TRUNCATE(ip, nsize, 0, NOCRED);
if (error)
break;
UFS_WAPBL_END(vp->v_mount);
error = UFS_WAPBL_BEGIN(vp->v_mount);
if (error)
goto out;
}
}
if (error == 0) {
truncate_error = UFS_TRUNCATE(ip, (off_t)0, 0, NOCRED);
/* XXX pedro: remove me */
if (truncate_error)
printf("UFS_TRUNCATE()=%d\n", truncate_error);
}
DIP_ASSIGN(ip, rdev, 0);
mode = DIP(ip, mode);
DIP_ASSIGN(ip, mode, 0);
ip->i_flag |= IN_CHANGE | IN_UPDATE;
/*
* Setting the mode to zero needs to wait for the inode to be
* written just as does a change to the link count. So, rather
* than creating a new entry point to do the same thing, we
* just use softdep_change_linkcnt(). Also, we can't let
* softdep co-opt us to help on its worklist, as we may end up
* trying to recycle vnodes and getting to this same point a
* couple of times, blowing the kernel stack. However, this
* could be optimized by checking if we are coming from
* vrele(), vput() or vclean() (by checking for VXLOCK) and
* just avoiding the co-opt to happen in the last case.
*/
if (DOINGSOFTDEP(vp))
softdep_change_linkcnt(ip, 1);
UFS_INODE_FREE(ip, ip->i_number, mode);
}
if (ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) {
if (!logged++) {
int err;
err = UFS_WAPBL_BEGIN(vp->v_mount);
if (err) {
error = err;
goto out;
}
}
UFS_UPDATE(ip, 0);
}
if (logged)
UFS_WAPBL_END(vp->v_mount);
out:
VOP_UNLOCK(vp, 0);
/*
* If we are done with the inode, reclaim it
* so that it can be reused immediately.
*/
if (error == 0 && truncate_error == 0 &&
(ip->i_din1 == NULL || DIP(ip, mode) == 0))
vrecycle(vp, p);
return (truncate_error ? truncate_error : error);
}
/*
* Indirect blocks are now on the vnode for the file. They are given negative
* logical block numbers. Indirect blocks are addressed by the negative
* address of the first data block to which they point. Double indirect blocks
* are addressed by one less than the address of the first indirect block to
* which they point. Triple indirect blocks are addressed by one less than
* the address of the first double indirect block to which they point.
*
* ext2_bmaparray does the bmap conversion, and if requested returns the
* array of logical blocks which must be traversed to get to a block.
* Each entry contains the offset into that block that gets you to the
* next block and the disk address of the block (if it is assigned).
*/
static
int
ext2_bmaparray(struct vnode *vp, ext2_daddr_t bn, ext2_daddr_t *bnp,
struct indir *ap, int *nump, int *runp, int *runb)
{
struct inode *ip;
struct buf *bp;
struct ext2_mount *ump;
struct mount *mp;
struct ext2_sb_info *fs;
struct indir a[NIADDR+1], *xap;
ext2_daddr_t daddr;
long metalbn;
int error, maxrun, num;
ip = VTOI(vp);
mp = vp->v_mount;
ump = VFSTOEXT2(mp);
fs = ip->i_e2fs;
#ifdef DIAGNOSTIC
if ((ap != NULL && nump == NULL) || (ap == NULL && nump != NULL))
panic("ext2_bmaparray: invalid arguments");
#endif
if (runp) {
*runp = 0;
}
if (runb) {
*runb = 0;
}
maxrun = mp->mnt_iosize_max / mp->mnt_stat.f_iosize - 1;
xap = ap == NULL ? a : ap;
if (!nump)
nump = #
error = ext2_getlbns(vp, bn, xap, nump);
if (error)
return (error);
num = *nump;
if (num == 0) {
*bnp = blkptrtodb(ump, ip->i_db[bn]);
if (*bnp == 0)
*bnp = -1;
else if (runp) {
daddr_t bnb = bn;
for (++bn; bn < NDADDR && *runp < maxrun &&
is_sequential(ump, ip->i_db[bn - 1], ip->i_db[bn]);
++bn, ++*runp);
bn = bnb;
if (runb && (bn > 0)) {
for (--bn; (bn >= 0) && (*runb < maxrun) &&
is_sequential(ump, ip->i_db[bn],
ip->i_db[bn+1]);
--bn, ++*runb);
}
}
return (0);
}
/* Get disk address out of indirect block array */
daddr = ip->i_ib[xap->in_off];
for (bp = NULL, ++xap; --num; ++xap) {
/*
* Exit the loop if there is no disk address assigned yet and
* the indirect block isn't in the cache, or if we were
* looking for an indirect block and we've found it.
*/
metalbn = xap->in_lbn;
if ((daddr == 0 &&
!findblk(vp, dbtodoff(fs, metalbn), FINDBLK_TEST)) ||
metalbn == bn) {
break;
}
/*
* If we get here, we've either got the block in the cache
* or we have a disk address for it, go fetch it.
*/
if (bp)
bqrelse(bp);
xap->in_exists = 1;
bp = getblk(vp, lblktodoff(fs, metalbn),
mp->mnt_stat.f_iosize, 0, 0);
if ((bp->b_flags & B_CACHE) == 0) {
#ifdef DIAGNOSTIC
if (!daddr)
panic("ext2_bmaparray: indirect block not in cache");
#endif
/*
* This runs through ext2_strategy using bio2 to
* cache the disk offset, then comes back through
* bio1. So we want to wait on bio1
*/
bp->b_bio1.bio_done = biodone_sync;
bp->b_bio1.bio_flags |= BIO_SYNC;
bp->b_bio2.bio_offset = fsbtodoff(fs, daddr);
bp->b_flags &= ~(B_INVAL|B_ERROR);
bp->b_cmd = BUF_CMD_READ;
vfs_busy_pages(bp->b_vp, bp);
vn_strategy(bp->b_vp, &bp->b_bio1);
error = biowait(&bp->b_bio1, "biord");
if (error) {
brelse(bp);
return (error);
}
}
daddr = ((ext2_daddr_t *)bp->b_data)[xap->in_off];
if (num == 1 && daddr && runp) {
for (bn = xap->in_off + 1;
bn < MNINDIR(ump) && *runp < maxrun &&
is_sequential(ump,
((ext2_daddr_t *)bp->b_data)[bn - 1],
((ext2_daddr_t *)bp->b_data)[bn]);
++bn, ++*runp);
bn = xap->in_off;
if (runb && bn) {
for(--bn; bn >= 0 && *runb < maxrun &&
is_sequential(ump, ((daddr_t *)bp->b_data)[bn],
((daddr_t *)bp->b_data)[bn+1]);
--bn, ++*runb);
}
}
}
if (bp)
bqrelse(bp);
daddr = blkptrtodb(ump, daddr);
*bnp = daddr == 0 ? -1 : daddr;
return (0);
}
/*
* Create an array of logical block number/offset pairs which represent the
* path of indirect blocks required to access a data block. The first "pair"
* contains the logical block number of the appropriate single, double or
* triple indirect block and the offset into the inode indirect block array.
* Note, the logical block number of the inode single/double/triple indirect
* block appears twice in the array, once with the offset into the i_ib and
* once with the offset into the page itself.
*/
static int
bmap_getlbns(struct nandfs_node *node, nandfs_lbn_t bn, struct nandfs_indir *ap, int *nump)
{
nandfs_daddr_t blockcnt;
nandfs_lbn_t metalbn, realbn;
struct nandfs_device *fsdev;
int i, numlevels, off;
fsdev = node->nn_nandfsdev;
DPRINTF(BMAP, ("%s: node %p bn=%jx mnindir=%zd enter\n", __func__,
node, bn, MNINDIR(fsdev)));
if (nump)
*nump = 0;
numlevels = 0;
realbn = bn;
if (bn < 0)
bn = -bn;
/* The first NDADDR blocks are direct blocks. */
if (bn < NDADDR)
return (0);
/*
* Determine the number of levels of indirection. After this loop
* is done, blockcnt indicates the number of data blocks possible
* at the previous level of indirection, and NIADDR - i is the number
* of levels of indirection needed to locate the requested block.
*/
for (blockcnt = 1, i = NIADDR, bn -= NDADDR;; i--, bn -= blockcnt) {
DPRINTF(BMAP, ("%s: blockcnt=%jd i=%d bn=%jd\n", __func__,
blockcnt, i, bn));
if (i == 0)
return (EFBIG);
blockcnt *= MNINDIR(fsdev);
if (bn < blockcnt)
break;
}
/* Calculate the address of the first meta-block. */
if (realbn >= 0)
metalbn = -(realbn - bn + NIADDR - i);
else
metalbn = -(-realbn - bn + NIADDR - i);
/*
* At each iteration, off is the offset into the bap array which is
* an array of disk addresses at the current level of indirection.
* The logical block number and the offset in that block are stored
* into the argument array.
*/
ap->in_lbn = metalbn;
ap->in_off = off = NIADDR - i;
DPRINTF(BMAP, ("%s: initial: ap->in_lbn=%jx ap->in_off=%d\n", __func__,
metalbn, off));
ap++;
for (++numlevels; i <= NIADDR; i++) {
/* If searching for a meta-data block, quit when found. */
if (metalbn == realbn)
break;
blockcnt /= MNINDIR(fsdev);
off = (bn / blockcnt) % MNINDIR(fsdev);
++numlevels;
ap->in_lbn = metalbn;
ap->in_off = off;
DPRINTF(BMAP, ("%s: in_lbn=%jx in_off=%d\n", __func__,
ap->in_lbn, ap->in_off));
++ap;
metalbn -= -1 + off * blockcnt;
}
if (nump)
*nump = numlevels;
DPRINTF(BMAP, ("%s: numlevels=%d\n", __func__, numlevels));
return (0);
}
int
bmap_truncate_mapping(struct nandfs_node *node, nandfs_lbn_t lastblk,
nandfs_lbn_t todo)
{
struct nandfs_inode *ip;
struct nandfs_indir a[NIADDR + 1], f[NIADDR], *ap;
nandfs_daddr_t indir_lbn[NIADDR];
nandfs_daddr_t *copy;
int error, level;
nandfs_lbn_t left, tosub;
struct nandfs_device *fsdev;
int cleaned, i;
int num, *nump;
DPRINTF(BMAP, ("%s: node %p lastblk %jx truncating by %jx\n", __func__,
node, lastblk, todo));
ip = &node->nn_inode;
fsdev = node->nn_nandfsdev;
ap = a;
nump = #
error = bmap_getlbns(node, lastblk, ap, nump);
if (error)
return (error);
indir_lbn[SINGLE] = -NDADDR;
indir_lbn[DOUBLE] = indir_lbn[SINGLE] - MNINDIR(fsdev) - 1;
indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - MNINDIR(fsdev)
* MNINDIR(fsdev) - 1;
for (i = 0; i < NIADDR; i++) {
f[i].in_off = MNINDIR(fsdev) - 1;
f[i].in_lbn = 0xdeadbeef;
}
left = todo;
#ifdef DEBUG
a[num].in_off = -1;
#endif
ap++;
num -= 2;
if (num < 0)
goto direct;
copy = malloc(MNINDIR(fsdev) * sizeof(nandfs_daddr_t) * (num + 1),
M_NANDFSTEMP, M_WAITOK);
for (level = num; level >= SINGLE && left > 0; level--) {
cleaned = 0;
if (ip->i_ib[level] == 0) {
tosub = blocks_inside(fsdev, level, ap);
if (tosub > left)
left = 0;
else
left -= tosub;
} else {
if (ap == f)
ap->in_lbn = indir_lbn[level];
error = bmap_truncate_indirect(node, level, &left,
&cleaned, ap, f, copy);
if (error) {
nandfs_error("%s: error %d when truncate "
"at level %d\n", __func__, error, level);
return (error);
}
}
if (cleaned) {
nandfs_vblock_end(fsdev, ip->i_ib[level]);
ip->i_ib[level] = 0;
}
ap = f;
}
free(copy, M_NANDFSTEMP);
direct:
if (num < 0)
i = lastblk;
else
i = NDADDR - 1;
for (; i >= 0 && left > 0; i--) {
if (ip->i_db[i] != 0) {
error = nandfs_bdestroy(node, ip->i_db[i]);
if (error) {
nandfs_error("%s: cannot destroy "
"block %jx, error %d\n", __func__,
(uintmax_t)ip->i_db[i], error);
return (error);
}
ip->i_db[i] = 0;
}
left--;
}
KASSERT(left == 0,
("truncated wrong number of blocks (%jd should be 0)", left));
return (error);
}
static int
bmap_truncate_indirect(struct nandfs_node *node, int level, nandfs_lbn_t *left,
int *cleaned, struct nandfs_indir *ap, struct nandfs_indir *fp,
nandfs_daddr_t *copy)
{
struct buf *bp;
nandfs_lbn_t i, lbn, nlbn, factor, tosub;
struct nandfs_device *fsdev;
int error, lcleaned, modified;
DPRINTF(BMAP, ("%s: node %p level %d left %jx\n", __func__,
node, level, *left));
fsdev = node->nn_nandfsdev;
MPASS(ap->in_off >= 0 && ap->in_off < MNINDIR(fsdev));
factor = lbn_offset(fsdev, level);
lbn = ap->in_lbn;
error = nandfs_bread_meta(node, lbn, NOCRED, 0, &bp);
if (error) {
brelse(bp);
return (error);
}
bcopy(bp->b_data, copy, fsdev->nd_blocksize);
bqrelse(bp);
modified = 0;
i = ap->in_off;
if (ap != fp)
ap++;
for (nlbn = lbn + 1 - i * factor; i >= 0 && *left > 0; i--,
nlbn += factor) {
lcleaned = 0;
DPRINTF(BMAP,
("%s: node %p i=%jx nlbn=%jx left=%jx ap=%p vblk %jx\n",
__func__, node, i, nlbn, *left, ap, copy[i]));
if (copy[i] == 0) {
tosub = blocks_inside(fsdev, level - 1, ap);
if (tosub > *left)
tosub = 0;
*left -= tosub;
} else {
if (level > SINGLE) {
if (ap == fp)
ap->in_lbn = nlbn;
error = bmap_truncate_indirect(node, level - 1,
left, &lcleaned, ap, fp,
copy + MNINDIR(fsdev));
if (error)
return (error);
} else {
error = nandfs_bdestroy(node, copy[i]);
if (error)
return (error);
lcleaned = 1;
*left -= 1;
}
}
if (lcleaned) {
if (level > SINGLE) {
error = nandfs_vblock_end(fsdev, copy[i]);
if (error)
return (error);
}
copy[i] = 0;
modified++;
}
ap = fp;
}
if (i == -1)
*cleaned = 1;
error = nandfs_bread_meta(node, lbn, NOCRED, 0, &bp);
if (error) {
brelse(bp);
return (error);
}
if (modified)
bcopy(copy, bp->b_data, fsdev->nd_blocksize);
error = nandfs_dirty_buf_meta(bp, 0);
if (error)
return (error);
return (error);
}
int
ext2fs_bmaparray(struct vnode *vp,
#undef struct
daddr_t bn, daddr_t *bnp, struct indir *ap,
int *nump, int *runp)
{
struct inode *ip;
struct buf *bp, *cbp;
#define struct
// struct ufsmount *ump;
struct mount *mp;
#undef struct
struct indir a[NIADDR+1], *xap;
daddr_t daddr;
daddr_t metalbn;
int error, maxrun = 0, num;
ip = VTOI(vp);
mp = EXT2_SIMPLE_FILE_SYSTEM_PRIVATE_DATA_FROM_THIS(vp->Filesystem);
// mp = vp->v_mount; !!!! need to fix this badly!
// ump = ip->i_ump; NEED TO DO SOMETHING ABOUT ufsmount
#ifdef DIAGNOSTIC
if ((ap != NULL && nump == NULL) || (ap == NULL && nump != NULL))
panic("ext2fs_bmaparray: invalid arguments");
#endif
if (runp) {
/*
* XXX
* If MAXBSIZE is the largest transfer the disks can handle,
* we probably want maxrun to be 1 block less so that we
* don't create a block larger than the device can handle.
*/
*runp = 0;
maxrun = MAXBSIZE / //mp->mnt_stat.f_iosize - 1; NEEDS FIX!!!
mp->fs->e2fs_bsize - 1;
}
if (bn >= 0 && bn < NDADDR) {
/* XXX ondisk32 */
*bnp = blkptrtodb(ump, fs2h32(ip->i_e2fs_blocks[bn]));
if (*bnp == 0)
*bnp = -1;
else if (runp)
/* XXX ondisk32 */
for (++bn; bn < NDADDR && *runp < maxrun &&
is_sequential(ump, (daddr_t)fs2h32(ip->i_e2fs_blocks[bn - 1]),
(daddr_t)fs2h32(ip->i_e2fs_blocks[bn]));
++bn, ++*runp);
return (0);
}
xap = ap == NULL ? a : ap;
if (!nump)
nump = #
if ((error = ufs_getlbns(vp, bn, xap, nump)) != 0)
return (error);
num = *nump;
/* Get disk address out of indirect block array */
/* XXX ondisk32 */
daddr = fs2h32(ip->i_e2fs_blocks[NDADDR + xap->in_off]);
#ifdef DIAGNOSTIC
if (num > NIADDR + 1 || num < 1) {
printf("ext2fs_bmaparray: num=%d\n", num);
panic("ext2fs_bmaparray: num");
}
#endif
for (bp = NULL, ++xap; --num; ++xap) {
/*
* Exit the loop if there is no disk address assigned yet and
* the indirect block isn't in the cache, or if we were
* looking for an indirect block and we've found it.
*/
metalbn = xap->in_lbn;
if (metalbn == bn)
break;
if (daddr == 0) {
mutex_enter(&bufcache_lock);
cbp = incore(vp, metalbn);
mutex_exit(&bufcache_lock);
if (cbp == NULL)
break;
}
/*
* If we get here, we've either got the block in the cache
* or we have a disk address for it, go fetch it.
*/
if (bp)
brelse(bp, 0);
xap->in_exists = 1;
//!!!!!!!!!!!!!!replaced 3rd param with 1 ftw
bp = getblk(vp, metalbn, 1, 0, 0);
if (bp == NULL) {
/*
* getblk() above returns NULL only iff we are
* pagedaemon. See the implementation of getblk
* for detail.
*/
return (ENOMEM);
}
if (bp->b_oflags & (BO_DONE | BO_DELWRI)) {
trace(TR_BREADHIT, pack(vp, size), metalbn);
}
#ifdef DIAGNOSTIC
else if (!daddr)
panic("ext2fs_bmaparry: indirect block not in cache");
#endif
else {
trace(TR_BREADMISS, pack(vp, size), metalbn);
bp->b_blkno = blkptrtodb(ump, daddr);
bp->b_flags |= B_READ;
VOP_STRATEGY(vp, bp);
// curlwp->l_ru.ru_inblock++; *//* XXX */
if ((error = biowait(bp)) != 0) {
brelse(bp, 0);
return (error);
}
}
/* XXX ondisk32 */
daddr = fs2h32(((int32_t *)bp->b_data)[xap->in_off]);
if (num == 1 && daddr && runp)
/* XXX ondisk32 */
for (bn = xap->in_off + 1;
bn < MNINDIR(ump) && *runp < maxrun &&
is_sequential(ump, ((int32_t *)bp->b_data)[bn - 1],
((int32_t *)bp->b_data)[bn]);
++bn, ++*runp);
}
if (bp)
brelse(bp, 0);
daddr = blkptrtodb(ump, daddr);
*bnp = daddr == 0 ? -1 : daddr;
return (0);
}
