/*
* Balloc defines the structure of file system storage
* by allocating the physical blocks on a device given
* the inode and the logical block number in a file.
*/
int
ext2fs_balloc(struct inode *ip, daddr_t bn, int size,
kauth_cred_t cred, struct buf **bpp, int flags)
{
struct m_ext2fs *fs;
daddr_t nb;
struct buf *bp, *nbp;
struct vnode *vp = ITOV(ip);
struct indir indirs[EXT2FS_NIADDR + 2];
daddr_t newb, lbn, pref;
int32_t *bap; /* XXX ondisk32 */
int num, i, error;
u_int deallocated;
daddr_t *blkp, *allocblk, allociblk[EXT2FS_NIADDR + 1];
int32_t *allocib; /* XXX ondisk32 */
int unwindidx = -1;
UVMHIST_FUNC("ext2fs_balloc"); UVMHIST_CALLED(ubchist);
UVMHIST_LOG(ubchist, "bn 0x%x", bn,0,0,0);
if (bpp != NULL) {
*bpp = NULL;
}
if (bn < 0)
return (EFBIG);
fs = ip->i_e2fs;
lbn = bn;
/*
* The first EXT2FS_NDADDR blocks are direct blocks
*/
if (bn < EXT2FS_NDADDR) {
/* XXX ondisk32 */
nb = fs2h32(ip->i_e2fs_blocks[bn]);
if (nb != 0) {
/*
* the block is already allocated, just read it.
*/
if (bpp != NULL) {
error = bread(vp, bn, fs->e2fs_bsize, NOCRED,
B_MODIFY, &bp);
if (error) {
return (error);
}
*bpp = bp;
}
return (0);
}
/*
* allocate a new direct block.
*/
error = ext2fs_alloc(ip, bn,
ext2fs_blkpref(ip, bn, bn, &ip->i_e2fs_blocks[0]),
cred, &newb);
if (error)
return (error);
ip->i_e2fs_last_lblk = lbn;
ip->i_e2fs_last_blk = newb;
/* XXX ondisk32 */
ip->i_e2fs_blocks[bn] = h2fs32((int32_t)newb);
ip->i_flag |= IN_CHANGE | IN_UPDATE;
if (bpp != NULL) {
bp = getblk(vp, bn, fs->e2fs_bsize, 0, 0);
bp->b_blkno = EXT2_FSBTODB(fs, newb);
if (flags & B_CLRBUF)
clrbuf(bp);
*bpp = bp;
}
return (0);
}
/*
* Determine the number of levels of indirection.
*/
pref = 0;
if ((error = ufs_getlbns(vp, bn, indirs, &num)) != 0)
return(error);
#ifdef DIAGNOSTIC
if (num < 1)
panic ("ext2fs_balloc: ufs_getlbns returned indirect block\n");
#endif
/*
* Fetch the first indirect block allocating if necessary.
*/
--num;
/* XXX ondisk32 */
nb = fs2h32(ip->i_e2fs_blocks[EXT2FS_NDADDR + indirs[0].in_off]);
allocib = NULL;
allocblk = allociblk;
if (nb == 0) {
pref = ext2fs_blkpref(ip, lbn, 0, (int32_t *)0);
error = ext2fs_alloc(ip, lbn, pref, cred, &newb);
if (error)
return (error);
nb = newb;
*allocblk++ = nb;
ip->i_e2fs_last_blk = newb;
bp = getblk(vp, indirs[1].in_lbn, fs->e2fs_bsize, 0, 0);
bp->b_blkno = EXT2_FSBTODB(fs, newb);
clrbuf(bp);
/*
* Write synchronously so that indirect blocks
* never point at garbage.
*/
if ((error = bwrite(bp)) != 0)
goto fail;
unwindidx = 0;
allocib = &ip->i_e2fs_blocks[EXT2FS_NDADDR + indirs[0].in_off];
/* XXX ondisk32 */
*allocib = h2fs32((int32_t)newb);
ip->i_flag |= IN_CHANGE | IN_UPDATE;
}
/*
* Fetch through the indirect blocks, allocating as necessary.
*/
for (i = 1;;) {
error = bread(vp,
indirs[i].in_lbn, (int)fs->e2fs_bsize, NOCRED, 0, &bp);
if (error) {
goto fail;
}
bap = (int32_t *)bp->b_data; /* XXX ondisk32 */
nb = fs2h32(bap[indirs[i].in_off]);
if (i == num)
break;
i++;
if (nb != 0) {
brelse(bp, 0);
continue;
}
pref = ext2fs_blkpref(ip, lbn, 0, (int32_t *)0);
error = ext2fs_alloc(ip, lbn, pref, cred, &newb);
if (error) {
brelse(bp, 0);
goto fail;
}
nb = newb;
*allocblk++ = nb;
ip->i_e2fs_last_blk = newb;
nbp = getblk(vp, indirs[i].in_lbn, fs->e2fs_bsize, 0, 0);
nbp->b_blkno = EXT2_FSBTODB(fs, nb);
clrbuf(nbp);
/*
* Write synchronously so that indirect blocks
* never point at garbage.
*/
if ((error = bwrite(nbp)) != 0) {
brelse(bp, 0);
goto fail;
}
if (unwindidx < 0)
unwindidx = i - 1;
/* XXX ondisk32 */
bap[indirs[i - 1].in_off] = h2fs32((int32_t)nb);
/*
* If required, write synchronously, otherwise use
* delayed write.
*/
if (flags & B_SYNC) {
bwrite(bp);
} else {
bdwrite(bp);
}
}
/*
* Get the data block, allocating if necessary.
*/
if (nb == 0) {
pref = ext2fs_blkpref(ip, lbn, indirs[num].in_off, &bap[0]);
error = ext2fs_alloc(ip, lbn, pref, cred, &newb);
if (error) {
brelse(bp, 0);
goto fail;
}
nb = newb;
*allocblk++ = nb;
ip->i_e2fs_last_lblk = lbn;
ip->i_e2fs_last_blk = newb;
/* XXX ondisk32 */
bap[indirs[num].in_off] = h2fs32((int32_t)nb);
/*
* If required, write synchronously, otherwise use
* delayed write.
*/
if (flags & B_SYNC) {
bwrite(bp);
} else {
bdwrite(bp);
}
if (bpp != NULL) {
nbp = getblk(vp, lbn, fs->e2fs_bsize, 0, 0);
nbp->b_blkno = EXT2_FSBTODB(fs, nb);
if (flags & B_CLRBUF)
clrbuf(nbp);
*bpp = nbp;
}
return (0);
}
brelse(bp, 0);
if (bpp != NULL) {
if (flags & B_CLRBUF) {
error = bread(vp, lbn, (int)fs->e2fs_bsize, NOCRED,
B_MODIFY, &nbp);
if (error) {
goto fail;
}
} else {
nbp = getblk(vp, lbn, fs->e2fs_bsize, 0, 0);
nbp->b_blkno = EXT2_FSBTODB(fs, nb);
}
*bpp = nbp;
}
return (0);
fail:
/*
* If we have failed part way through block allocation, we
* have to deallocate any indirect blocks that we have allocated.
*/
for (deallocated = 0, blkp = allociblk; blkp < allocblk; blkp++) {
ext2fs_blkfree(ip, *blkp);
deallocated += fs->e2fs_bsize;
}
if (unwindidx >= 0) {
if (unwindidx == 0) {
*allocib = 0;
} else {
int r;
r = bread(vp, indirs[unwindidx].in_lbn,
(int)fs->e2fs_bsize, NOCRED, B_MODIFY, &bp);
if (r) {
panic("Could not unwind indirect block, error %d", r);
} else {
bap = (int32_t *)bp->b_data; /* XXX ondisk32 */
bap[indirs[unwindidx].in_off] = 0;
if (flags & B_SYNC)
bwrite(bp);
else
bdwrite(bp);
}
}
for (i = unwindidx + 1; i <= num; i++) {
bp = getblk(vp, indirs[i].in_lbn, (int)fs->e2fs_bsize,
0, 0);
brelse(bp, BC_INVAL);
}
}
if (deallocated) {
ext2fs_setnblock(ip, ext2fs_nblock(ip) - btodb(deallocated));
ip->i_e2fs_flags |= IN_CHANGE | IN_UPDATE;
}
return error;
}
/*
* Read an inode from disk and initialize this vnode / inode pair.
* Caller assures no other thread will try to load this inode.
*/
int
ext2fs_loadvnode(struct mount *mp, struct vnode *vp,
const void *key, size_t key_len, const void **new_key)
{
ino_t ino;
struct m_ext2fs *fs;
struct inode *ip;
struct ufsmount *ump;
struct buf *bp;
dev_t dev;
int error;
KASSERT(key_len == sizeof(ino));
memcpy(&ino, key, key_len);
ump = VFSTOUFS(mp);
dev = ump->um_dev;
fs = ump->um_e2fs;
/* Read in the disk contents for the inode, copy into the inode. */
error = bread(ump->um_devvp, EXT2_FSBTODB(fs, ino_to_fsba(fs, ino)),
(int)fs->e2fs_bsize, 0, &bp);
if (error)
return error;
/* Allocate and initialize inode. */
ip = pool_get(&ext2fs_inode_pool, PR_WAITOK);
memset(ip, 0, sizeof(struct inode));
vp->v_tag = VT_EXT2FS;
vp->v_op = ext2fs_vnodeop_p;
vp->v_vflag |= VV_LOCKSWORK;
vp->v_data = ip;
ip->i_vnode = vp;
ip->i_ump = ump;
ip->i_e2fs = fs;
ip->i_dev = dev;
ip->i_number = ino;
ip->i_e2fs_last_lblk = 0;
ip->i_e2fs_last_blk = 0;
/* Initialize genfs node. */
genfs_node_init(vp, &ext2fs_genfsops);
error = ext2fs_loadvnode_content(fs, ino, bp, ip);
brelse(bp, 0);
if (error)
return error;
/* If the inode was deleted, reset all fields */
if (ip->i_e2fs_dtime != 0) {
ip->i_e2fs_mode = 0;
(void)ext2fs_setsize(ip, 0);
(void)ext2fs_setnblock(ip, 0);
memset(ip->i_e2fs_blocks, 0, sizeof(ip->i_e2fs_blocks));
}
/* Initialize the vnode from the inode. */
ext2fs_vinit(mp, ext2fs_specop_p, ext2fs_fifoop_p, &vp);
/* Finish inode initialization. */
ip->i_devvp = ump->um_devvp;
vref(ip->i_devvp);
/*
* Set up a generation number for this inode if it does not
* already have one. This should only happen on old filesystems.
*/
if (ip->i_e2fs_gen == 0) {
if (++ext2gennumber < (u_long)time_second)
ext2gennumber = time_second;
ip->i_e2fs_gen = ext2gennumber;
if ((mp->mnt_flag & MNT_RDONLY) == 0)
ip->i_flag |= IN_MODIFIED;
}
uvm_vnp_setsize(vp, ext2fs_size(ip));
*new_key = &ip->i_number;
return 0;
}
/*
* Look up a EXT2FS dinode number to find its incore vnode, otherwise read it
* in from disk. If it is in core, wait for the lock bit to clear, then
* return the inode locked. Detection and handling of mount points must be
* done by the calling routine.
*/
int
ext2fs_vget(struct mount *mp, ino_t ino, struct vnode **vpp)
{
struct m_ext2fs *fs;
struct inode *ip;
struct ufsmount *ump;
struct buf *bp;
struct vnode *vp;
dev_t dev;
int error;
void *cp;
ump = VFSTOUFS(mp);
dev = ump->um_dev;
retry:
if ((*vpp = ufs_ihashget(dev, ino, LK_EXCLUSIVE)) != NULL)
return (0);
/* Allocate a new vnode/inode. */
error = getnewvnode(VT_EXT2FS, mp, ext2fs_vnodeop_p, NULL, &vp);
if (error) {
*vpp = NULL;
return (error);
}
ip = pool_get(&ext2fs_inode_pool, PR_WAITOK);
mutex_enter(&ufs_hashlock);
if ((*vpp = ufs_ihashget(dev, ino, 0)) != NULL) {
mutex_exit(&ufs_hashlock);
ungetnewvnode(vp);
pool_put(&ext2fs_inode_pool, ip);
goto retry;
}
vp->v_vflag |= VV_LOCKSWORK;
memset(ip, 0, sizeof(struct inode));
vp->v_data = ip;
ip->i_vnode = vp;
ip->i_ump = ump;
ip->i_e2fs = fs = ump->um_e2fs;
ip->i_dev = dev;
ip->i_number = ino;
ip->i_e2fs_last_lblk = 0;
ip->i_e2fs_last_blk = 0;
genfs_node_init(vp, &ext2fs_genfsops);
/*
* Put it onto its hash chain and lock it so that other requests for
* this inode will block if they arrive while we are sleeping waiting
* for old data structures to be purged or for the contents of the
* disk portion of this inode to be read.
*/
ufs_ihashins(ip);
mutex_exit(&ufs_hashlock);
/* Read in the disk contents for the inode, copy into the inode. */
error = bread(ump->um_devvp, EXT2_FSBTODB(fs, ino_to_fsba(fs, ino)),
(int)fs->e2fs_bsize, NOCRED, 0, &bp);
if (error) {
/*
* The inode does not contain anything useful, so it would
* be misleading to leave it on its hash chain. With mode
* still zero, it will be unlinked and returned to the free
* list by vput().
*/
vput(vp);
*vpp = NULL;
return (error);
}
cp = (char *)bp->b_data + (ino_to_fsbo(fs, ino) * EXT2_DINODE_SIZE(fs));
ip->i_din.e2fs_din = pool_get(&ext2fs_dinode_pool, PR_WAITOK);
e2fs_iload((struct ext2fs_dinode *)cp, ip->i_din.e2fs_din);
ext2fs_set_inode_guid(ip);
brelse(bp, 0);
/* If the inode was deleted, reset all fields */
if (ip->i_e2fs_dtime != 0) {
ip->i_e2fs_mode = 0;
(void)ext2fs_setsize(ip, 0);
(void)ext2fs_setnblock(ip, 0);
memset(ip->i_e2fs_blocks, 0, sizeof(ip->i_e2fs_blocks));
}
/*
* Initialize the vnode from the inode, check for aliases.
*/
error = ext2fs_vinit(mp, ext2fs_specop_p, ext2fs_fifoop_p, &vp);
if (error) {
vput(vp);
*vpp = NULL;
return (error);
}
/*
* Finish inode initialization now that aliasing has been resolved.
*/
ip->i_devvp = ump->um_devvp;
vref(ip->i_devvp);
/*
* Set up a generation number for this inode if it does not
* already have one. This should only happen on old filesystems.
*/
if (ip->i_e2fs_gen == 0) {
if (++ext2gennumber < (u_long)time_second)
ext2gennumber = time_second;
ip->i_e2fs_gen = ext2gennumber;
if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
ip->i_flag |= IN_MODIFIED;
}
uvm_vnp_setsize(vp, ext2fs_size(ip));
*vpp = vp;
return (0);
}