/*
* Allocate a block in the filesystem.
*
* 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
ext2_alloc(struct inode *ip, daddr_t lbn, e4fs_daddr_t bpref, int size,
struct ucred *cred, e4fs_daddr_t *bnp)
{
struct m_ext2fs *fs;
struct ext2mount *ump;
int32_t bno;
int cg;
*bnp = 0;
fs = ip->i_e2fs;
ump = ip->i_ump;
mtx_assert(EXT2_MTX(ump), MA_OWNED);
#ifdef INVARIANTS
if ((u_int)size > fs->e2fs_bsize || blkoff(fs, size) != 0) {
vn_printf(ip->i_devvp, "bsize = %lu, size = %d, fs = %s\n",
(long unsigned int)fs->e2fs_bsize, size, fs->e2fs_fsmnt);
panic("ext2_alloc: bad size");
}
if (cred == NOCRED)
panic("ext2_alloc: missing credential");
#endif /* INVARIANTS */
if (size == fs->e2fs_bsize && fs->e2fs->e2fs_fbcount == 0)
goto nospace;
if (cred->cr_uid != 0 &&
fs->e2fs->e2fs_fbcount < fs->e2fs->e2fs_rbcount)
goto nospace;
if (bpref >= fs->e2fs->e2fs_bcount)
bpref = 0;
if (bpref == 0)
cg = ino_to_cg(fs, ip->i_number);
else
cg = dtog(fs, bpref);
bno = (daddr_t)ext2_hashalloc(ip, cg, bpref, fs->e2fs_bsize,
ext2_alloccg);
if (bno > 0) {
/* set next_alloc fields as done in block_getblk */
ip->i_next_alloc_block = lbn;
ip->i_next_alloc_goal = bno;
ip->i_blocks += btodb(fs->e2fs_bsize);
ip->i_flag |= IN_CHANGE | IN_UPDATE;
*bnp = bno;
return (0);
}
nospace:
EXT2_UNLOCK(ump);
ext2_fserr(fs, cred->cr_uid, "filesystem full");
uprintf("\n%s: write failed, filesystem is full\n", fs->e2fs_fsmnt);
return (ENOSPC);
}
/*
* Free an inode.
*
*/
int
ext2_vfree(struct vnode *pvp, ino_t ino, int mode)
{
struct m_ext2fs *fs;
struct inode *pip;
struct buf *bp;
struct ext2mount *ump;
int error, cg;
char * ibp;
pip = VTOI(pvp);
fs = pip->i_e2fs;
ump = pip->i_ump;
if ((u_int)ino > fs->e2fs_ipg * fs->e2fs_gcount)
panic("ext2_vfree: range: devvp = %p, ino = %ju, fs = %s",
pip->i_devvp, (uintmax_t)ino, fs->e2fs_fsmnt);
cg = ino_to_cg(fs, ino);
error = bread(pip->i_devvp,
fsbtodb(fs, fs->e2fs_gd[cg].ext2bgd_i_bitmap),
(int)fs->e2fs_bsize, NOCRED, &bp);
if (error) {
brelse(bp);
return (0);
}
ibp = (char *)bp->b_data;
ino = (ino - 1) % fs->e2fs->e2fs_ipg;
if (isclr(ibp, ino)) {
printf("ino = %llu, fs = %s\n",
(unsigned long long)ino, fs->e2fs_fsmnt);
if (fs->e2fs_ronly == 0)
panic("ext2_vfree: freeing free inode");
}
clrbit(ibp, ino);
EXT2_LOCK(ump);
fs->e2fs->e2fs_ficount++;
fs->e2fs_gd[cg].ext2bgd_nifree++;
if ((mode & IFMT) == IFDIR) {
fs->e2fs_gd[cg].ext2bgd_ndirs--;
fs->e2fs_total_dir--;
}
fs->e2fs_fmod = 1;
EXT2_UNLOCK(ump);
bdwrite(bp);
return (0);
}
/*
* Free a block or fragment.
*
*/
void
ext2_blkfree(struct inode *ip, e4fs_daddr_t bno, long size)
{
struct m_ext2fs *fs;
struct buf *bp;
struct ext2mount *ump;
int cg, error;
char *bbp;
fs = ip->i_e2fs;
ump = ip->i_ump;
cg = dtog(fs, bno);
if ((u_int)bno >= fs->e2fs->e2fs_bcount) {
printf("bad block %lld, ino %llu\n", (long long)bno,
(unsigned long long)ip->i_number);
ext2_fserr(fs, ip->i_uid, "bad block");
return;
}
error = bread(ip->i_devvp,
fsbtodb(fs, fs->e2fs_gd[cg].ext2bgd_b_bitmap),
(int)fs->e2fs_bsize, NOCRED, &bp);
if (error) {
brelse(bp);
return;
}
bbp = (char *)bp->b_data;
bno = dtogd(fs, bno);
if (isclr(bbp, bno)) {
printf("block = %lld, fs = %s\n",
(long long)bno, fs->e2fs_fsmnt);
panic("ext2_blkfree: freeing free block");
}
clrbit(bbp, bno);
EXT2_LOCK(ump);
ext2_clusteracct(fs, bbp, cg, bno, 1);
fs->e2fs->e2fs_fbcount++;
fs->e2fs_gd[cg].ext2bgd_nbfree++;
fs->e2fs_fmod = 1;
EXT2_UNLOCK(ump);
bdwrite(bp);
}
/*
* Balloc defines the structure of filesystem storage
* by allocating the physical blocks on a device given
* the inode and the logical block number in a file.
*/
int
ext2_balloc(struct inode *ip, e2fs_lbn_t lbn, int size, struct ucred *cred,
struct buf **bpp, int flags)
{
struct m_ext2fs *fs;
struct ext2mount *ump;
struct buf *bp, *nbp;
struct vnode *vp = ITOV(ip);
struct indir indirs[NIADDR + 2];
e4fs_daddr_t nb, newb;
e2fs_daddr_t *bap, pref;
int osize, nsize, num, i, error;
*bpp = NULL;
if (lbn < 0)
return (EFBIG);
fs = ip->i_e2fs;
ump = ip->i_ump;
/*
* check if this is a sequential block allocation.
* If so, increment next_alloc fields to allow ext2_blkpref
* to make a good guess
*/
if (lbn == ip->i_next_alloc_block + 1) {
ip->i_next_alloc_block++;
ip->i_next_alloc_goal++;
}
/*
* The first NDADDR blocks are direct blocks
*/
if (lbn < NDADDR) {
nb = ip->i_db[lbn];
/* no new block is to be allocated, and no need to expand
the file */
if (nb != 0 && ip->i_size >= (lbn + 1) * fs->e2fs_bsize) {
error = bread(vp, lbn, fs->e2fs_bsize, NOCRED, &bp);
if (error) {
brelse(bp);
return (error);
}
bp->b_blkno = fsbtodb(fs, nb);
*bpp = bp;
return (0);
}
if (nb != 0) {
/*
* Consider need to reallocate a fragment.
*/
osize = fragroundup(fs, blkoff(fs, ip->i_size));
nsize = fragroundup(fs, size);
if (nsize <= osize) {
error = bread(vp, lbn, osize, NOCRED, &bp);
if (error) {
brelse(bp);
return (error);
}
bp->b_blkno = fsbtodb(fs, nb);
} else {
/* Godmar thinks: this shouldn't happen w/o fragments */
printf("nsize %d(%d) > osize %d(%d) nb %d\n",
(int)nsize, (int)size, (int)osize,
(int)ip->i_size, (int)nb);
panic(
"ext2_balloc: Something is terribly wrong");
/*
* please note there haven't been any changes from here on -
* FFS seems to work.
*/
}
} else {
if (ip->i_size < (lbn + 1) * fs->e2fs_bsize)
nsize = fragroundup(fs, size);
else
nsize = fs->e2fs_bsize;
EXT2_LOCK(ump);
error = ext2_alloc(ip, lbn,
ext2_blkpref(ip, lbn, (int)lbn, &ip->i_db[0], 0),
nsize, cred, &newb);
if (error)
return (error);
bp = getblk(vp, lbn, nsize, 0, 0, 0);
bp->b_blkno = fsbtodb(fs, newb);
if (flags & BA_CLRBUF)
vfs_bio_clrbuf(bp);
}
ip->i_db[lbn] = dbtofsb(fs, bp->b_blkno);
ip->i_flag |= IN_CHANGE | IN_UPDATE;
*bpp = bp;
return (0);
}
/*
* Determine the number of levels of indirection.
*/
pref = 0;
if ((error = ext2_getlbns(vp, lbn, indirs, &num)) != 0)
return (error);
#ifdef INVARIANTS
if (num < 1)
panic ("ext2_balloc: ext2_getlbns returned indirect block");
#endif
/*
* Fetch the first indirect block allocating if necessary.
*/
--num;
nb = ip->i_ib[indirs[0].in_off];
if (nb == 0) {
EXT2_LOCK(ump);
pref = ext2_blkpref(ip, lbn, indirs[0].in_off +
EXT2_NDIR_BLOCKS, &ip->i_db[0], 0);
if ((error = ext2_alloc(ip, lbn, pref, fs->e2fs_bsize, cred,
&newb)))
return (error);
nb = newb;
bp = getblk(vp, indirs[1].in_lbn, fs->e2fs_bsize, 0, 0, 0);
bp->b_blkno = fsbtodb(fs, newb);
vfs_bio_clrbuf(bp);
/*
* Write synchronously so that indirect blocks
* never point at garbage.
*/
if ((error = bwrite(bp)) != 0) {
ext2_blkfree(ip, nb, fs->e2fs_bsize);
return (error);
}
ip->i_ib[indirs[0].in_off] = 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, &bp);
if (error) {
brelse(bp);
return (error);
}
bap = (e2fs_daddr_t *)bp->b_data;
nb = bap[indirs[i].in_off];
if (i == num)
break;
i += 1;
if (nb != 0) {
bqrelse(bp);
continue;
}
EXT2_LOCK(ump);
if (pref == 0)
pref = ext2_blkpref(ip, lbn, indirs[i].in_off, bap,
bp->b_lblkno);
error = ext2_alloc(ip, lbn, pref, (int)fs->e2fs_bsize, cred, &newb);
if (error) {
brelse(bp);
return (error);
}
nb = newb;
nbp = getblk(vp, indirs[i].in_lbn, fs->e2fs_bsize, 0, 0, 0);
nbp->b_blkno = fsbtodb(fs, nb);
vfs_bio_clrbuf(nbp);
/*
* Write synchronously so that indirect blocks
* never point at garbage.
*/
if ((error = bwrite(nbp)) != 0) {
ext2_blkfree(ip, nb, fs->e2fs_bsize);
EXT2_UNLOCK(ump);
brelse(bp);
return (error);
}
bap[indirs[i - 1].in_off] = nb;
/*
* If required, write synchronously, otherwise use
* delayed write.
*/
if (flags & IO_SYNC) {
bwrite(bp);
} else {
if (bp->b_bufsize == fs->e2fs_bsize)
bp->b_flags |= B_CLUSTEROK;
bdwrite(bp);
}
}
/*
* Get the data block, allocating if necessary.
*/
if (nb == 0) {
EXT2_LOCK(ump);
pref = ext2_blkpref(ip, lbn, indirs[i].in_off, &bap[0],
bp->b_lblkno);
if ((error = ext2_alloc(ip,
lbn, pref, (int)fs->e2fs_bsize, cred, &newb)) != 0) {
brelse(bp);
return (error);
}
nb = newb;
nbp = getblk(vp, lbn, fs->e2fs_bsize, 0, 0, 0);
nbp->b_blkno = fsbtodb(fs, nb);
if (flags & BA_CLRBUF)
vfs_bio_clrbuf(nbp);
bap[indirs[i].in_off] = nb;
/*
* If required, write synchronously, otherwise use
* delayed write.
*/
if (flags & IO_SYNC) {
bwrite(bp);
} else {
if (bp->b_bufsize == fs->e2fs_bsize)
bp->b_flags |= B_CLUSTEROK;
bdwrite(bp);
}
*bpp = nbp;
return (0);
}
brelse(bp);
if (flags & BA_CLRBUF) {
int seqcount = (flags & BA_SEQMASK) >> BA_SEQSHIFT;
if (seqcount && (vp->v_mount->mnt_flag & MNT_NOCLUSTERR) == 0) {
error = cluster_read(vp, ip->i_size, lbn,
(int)fs->e2fs_bsize, NOCRED,
MAXBSIZE, seqcount, 0, &nbp);
} else {
error = bread(vp, lbn, (int)fs->e2fs_bsize, NOCRED, &nbp);
}
if (error) {
brelse(nbp);
return (error);
}
} else {
/*
* Determine whether an inode can be allocated.
*
* Check to see if an inode is available, and if it is,
* allocate it using tode in the specified cylinder group.
*/
static daddr_t
ext2_nodealloccg(struct inode *ip, int cg, daddr_t ipref, int mode)
{
struct m_ext2fs *fs;
struct buf *bp;
struct ext2mount *ump;
int error, start, len;
char *ibp, *loc;
ipref--; /* to avoid a lot of (ipref -1) */
if (ipref == -1)
ipref = 0;
fs = ip->i_e2fs;
ump = ip->i_ump;
if (fs->e2fs_gd[cg].ext2bgd_nifree == 0)
return (0);
EXT2_UNLOCK(ump);
error = bread(ip->i_devvp, fsbtodb(fs,
fs->e2fs_gd[cg].ext2bgd_i_bitmap),
(int)fs->e2fs_bsize, NOCRED, &bp);
if (error) {
brelse(bp);
EXT2_LOCK(ump);
return (0);
}
if (fs->e2fs_gd[cg].ext2bgd_nifree == 0) {
/*
* Another thread allocated the last i-node in this
* group while we were waiting for the buffer.
*/
brelse(bp);
EXT2_LOCK(ump);
return (0);
}
ibp = (char *)bp->b_data;
if (ipref) {
ipref %= fs->e2fs->e2fs_ipg;
if (isclr(ibp, ipref))
goto gotit;
}
start = ipref / NBBY;
len = howmany(fs->e2fs->e2fs_ipg - ipref, NBBY);
loc = memcchr(&ibp[start], 0xff, len);
if (loc == NULL) {
len = start + 1;
start = 0;
loc = memcchr(&ibp[start], 0xff, len);
if (loc == NULL) {
printf("cg = %d, ipref = %lld, fs = %s\n",
cg, (long long)ipref, fs->e2fs_fsmnt);
panic("ext2fs_nodealloccg: map corrupted");
/* NOTREACHED */
}
}
ipref = (loc - ibp) * NBBY + ffs(~*loc) - 1;
gotit:
setbit(ibp, ipref);
EXT2_LOCK(ump);
fs->e2fs_gd[cg].ext2bgd_nifree--;
fs->e2fs->e2fs_ficount--;
fs->e2fs_fmod = 1;
if ((mode & IFMT) == IFDIR) {
fs->e2fs_gd[cg].ext2bgd_ndirs++;
fs->e2fs_total_dir++;
}
EXT2_UNLOCK(ump);
bdwrite(bp);
return (cg * fs->e2fs->e2fs_ipg + ipref +1);
}
/*
* Determine whether a cluster can be allocated.
*/
static daddr_t
ext2_clusteralloc(struct inode *ip, int cg, daddr_t bpref, int len)
{
struct m_ext2fs *fs;
struct ext2mount *ump;
struct buf *bp;
char *bbp;
int bit, error, got, i, loc, run;
int32_t *lp;
daddr_t bno;
fs = ip->i_e2fs;
ump = ip->i_ump;
if (fs->e2fs_maxcluster[cg] < len)
return (0);
EXT2_UNLOCK(ump);
error = bread(ip->i_devvp,
fsbtodb(fs, fs->e2fs_gd[cg].ext2bgd_b_bitmap),
(int)fs->e2fs_bsize, NOCRED, &bp);
if (error)
goto fail_lock;
bbp = (char *)bp->b_data;
EXT2_LOCK(ump);
/*
* Check to see if a cluster of the needed size (or bigger) is
* available in this cylinder group.
*/
lp = &fs->e2fs_clustersum[cg].cs_sum[len];
for (i = len; i <= fs->e2fs_contigsumsize; i++)
if (*lp++ > 0)
break;
if (i > fs->e2fs_contigsumsize) {
/*
* Update the cluster summary information to reflect
* the true maximum-sized cluster so that future cluster
* allocation requests can avoid reading the bitmap only
* to find no cluster.
*/
lp = &fs->e2fs_clustersum[cg].cs_sum[len - 1];
for (i = len - 1; i > 0; i--)
if (*lp-- > 0)
break;
fs->e2fs_maxcluster[cg] = i;
goto fail;
}
EXT2_UNLOCK(ump);
/* Search the bitmap to find a big enough cluster like in FFS. */
if (dtog(fs, bpref) != cg)
bpref = 0;
if (bpref != 0)
bpref = dtogd(fs, bpref);
loc = bpref / NBBY;
bit = 1 << (bpref % NBBY);
for (run = 0, got = bpref; got < fs->e2fs->e2fs_fpg; got++) {
if ((bbp[loc] & bit) != 0)
run = 0;
else {
run++;
if (run == len)
break;
}
if ((got & (NBBY - 1)) != (NBBY - 1))
bit <<= 1;
else {
loc++;
bit = 1;
}
}
if (got >= fs->e2fs->e2fs_fpg)
goto fail_lock;
/* Allocate the cluster that we found. */
for (i = 1; i < len; i++)
if (!isclr(bbp, got - run + i))
panic("ext2_clusteralloc: map mismatch");
bno = got - run + 1;
if (bno >= fs->e2fs->e2fs_fpg)
panic("ext2_clusteralloc: allocated out of group");
EXT2_LOCK(ump);
for (i = 0; i < len; i += fs->e2fs_fpb) {
setbit(bbp, bno + i);
ext2_clusteracct(fs, bbp, cg, bno + i, -1);
fs->e2fs->e2fs_fbcount--;
fs->e2fs_gd[cg].ext2bgd_nbfree--;
}
fs->e2fs_fmod = 1;
EXT2_UNLOCK(ump);
bdwrite(bp);
return (cg * fs->e2fs->e2fs_fpg + fs->e2fs->e2fs_first_dblock + bno);
fail_lock:
EXT2_LOCK(ump);
fail:
brelse(bp);
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 daddr_t
ext2_alloccg(struct inode *ip, int cg, daddr_t bpref, int size)
{
struct m_ext2fs *fs;
struct buf *bp;
struct ext2mount *ump;
daddr_t bno, runstart, runlen;
int bit, loc, end, error, start;
char *bbp;
/* XXX ondisk32 */
fs = ip->i_e2fs;
ump = ip->i_ump;
if (fs->e2fs_gd[cg].ext2bgd_nbfree == 0)
return (0);
EXT2_UNLOCK(ump);
error = bread(ip->i_devvp, fsbtodb(fs,
fs->e2fs_gd[cg].ext2bgd_b_bitmap),
(int)fs->e2fs_bsize, NOCRED, &bp);
if (error) {
brelse(bp);
EXT2_LOCK(ump);
return (0);
}
if (fs->e2fs_gd[cg].ext2bgd_nbfree == 0) {
/*
* Another thread allocated the last block in this
* group while we were waiting for the buffer.
*/
brelse(bp);
EXT2_LOCK(ump);
return (0);
}
bbp = (char *)bp->b_data;
if (dtog(fs, bpref) != cg)
bpref = 0;
if (bpref != 0) {
bpref = dtogd(fs, bpref);
/*
* if the requested block is available, use it
*/
if (isclr(bbp, bpref)) {
bno = bpref;
goto gotit;
}
}
/*
* no blocks in the requested cylinder, so take next
* available one in this cylinder group.
* first try to get 8 contigous blocks, then fall back to a single
* block.
*/
if (bpref)
start = dtogd(fs, bpref) / NBBY;
else
start = 0;
end = howmany(fs->e2fs->e2fs_fpg, NBBY) - start;
retry:
runlen = 0;
runstart = 0;
for (loc = start; loc < end; loc++) {
if (bbp[loc] == (char)0xff) {
runlen = 0;
continue;
}
/* Start of a run, find the number of high clear bits. */
if (runlen == 0) {
bit = fls(bbp[loc]);
runlen = NBBY - bit;
runstart = loc * NBBY + bit;
} else if (bbp[loc] == 0) {
/* Continue a run. */
runlen += NBBY;
} else {
/*
* Finish the current run. If it isn't long
* enough, start a new one.
*/
bit = ffs(bbp[loc]) - 1;
runlen += bit;
if (runlen >= 8) {
bno = runstart;
goto gotit;
}
/* Run was too short, start a new one. */
bit = fls(bbp[loc]);
runlen = NBBY - bit;
runstart = loc * NBBY + bit;
}
/* If the current run is long enough, use it. */
if (runlen >= 8) {
bno = runstart;
goto gotit;
}
}
if (start != 0) {
end = start;
start = 0;
goto retry;
}
bno = ext2_mapsearch(fs, bbp, bpref);
if (bno < 0) {
brelse(bp);
EXT2_LOCK(ump);
return (0);
}
gotit:
#ifdef INVARIANTS
if (isset(bbp, bno)) {
printf("ext2fs_alloccgblk: cg=%d bno=%jd fs=%s\n",
cg, (intmax_t)bno, fs->e2fs_fsmnt);
panic("ext2fs_alloccg: dup alloc");
}
#endif
setbit(bbp, bno);
EXT2_LOCK(ump);
ext2_clusteracct(fs, bbp, cg, bno, -1);
fs->e2fs->e2fs_fbcount--;
fs->e2fs_gd[cg].ext2bgd_nbfree--;
fs->e2fs_fmod = 1;
EXT2_UNLOCK(ump);
bdwrite(bp);
return (cg * fs->e2fs->e2fs_fpg + fs->e2fs->e2fs_first_dblock + bno);
}
/*
* Allocate an inode in the filesystem.
*
*/
int
ext2_valloc(struct vnode *pvp, int mode, struct ucred *cred, struct vnode **vpp)
{
struct timespec ts;
struct inode *pip;
struct m_ext2fs *fs;
struct inode *ip;
struct ext2mount *ump;
ino_t ino, ipref;
int i, error, cg;
*vpp = NULL;
pip = VTOI(pvp);
fs = pip->i_e2fs;
ump = pip->i_ump;
EXT2_LOCK(ump);
if (fs->e2fs->e2fs_ficount == 0)
goto noinodes;
/*
* If it is a directory then obtain a cylinder group based on
* ext2_dirpref else obtain it using ino_to_cg. The preferred inode is
* always the next inode.
*/
if ((mode & IFMT) == IFDIR) {
cg = ext2_dirpref(pip);
if (fs->e2fs_contigdirs[cg] < 255)
fs->e2fs_contigdirs[cg]++;
} else {
cg = ino_to_cg(fs, pip->i_number);
if (fs->e2fs_contigdirs[cg] > 0)
fs->e2fs_contigdirs[cg]--;
}
ipref = cg * fs->e2fs->e2fs_ipg + 1;
ino = (ino_t)ext2_hashalloc(pip, cg, (long)ipref, mode, ext2_nodealloccg);
if (ino == 0)
goto noinodes;
error = VFS_VGET(pvp->v_mount, ino, LK_EXCLUSIVE, vpp);
if (error) {
ext2_vfree(pvp, ino, mode);
return (error);
}
ip = VTOI(*vpp);
/*
* The question is whether using VGET was such good idea at all:
* Linux doesn't read the old inode in when it is allocating a
* new one. I will set at least i_size and i_blocks to zero.
*/
ip->i_size = 0;
ip->i_blocks = 0;
ip->i_mode = 0;
ip->i_flags = 0;
/* now we want to make sure that the block pointers are zeroed out */
for (i = 0; i < NDADDR; i++)
ip->i_db[i] = 0;
for (i = 0; i < NIADDR; i++)
ip->i_ib[i] = 0;
/*
* Set up a new generation number for this inode.
* XXX check if this makes sense in ext2
*/
if (ip->i_gen == 0 || ++ip->i_gen == 0)
ip->i_gen = random() / 2 + 1;
vfs_timestamp(&ts);
ip->i_birthtime = ts.tv_sec;
ip->i_birthnsec = ts.tv_nsec;
/*
printf("ext2_valloc: allocated inode %d\n", ino);
*/
return (0);
noinodes:
EXT2_UNLOCK(ump);
ext2_fserr(fs, cred->cr_uid, "out of inodes");
uprintf("\n%s: create/symlink failed, no inodes free\n", fs->e2fs_fsmnt);
return (ENOSPC);
}
int
ext2_reallocblks(struct vop_reallocblks_args *ap)
{
struct m_ext2fs *fs;
struct inode *ip;
struct vnode *vp;
struct buf *sbp, *ebp;
uint32_t *bap, *sbap, *ebap = 0;
struct ext2mount *ump;
struct cluster_save *buflist;
struct indir start_ap[NIADDR + 1], end_ap[NIADDR + 1], *idp;
e2fs_lbn_t start_lbn, end_lbn;
int soff;
e2fs_daddr_t newblk, blkno;
int i, len, start_lvl, end_lvl, pref, ssize;
if (doreallocblks == 0)
return (ENOSPC);
vp = ap->a_vp;
ip = VTOI(vp);
fs = ip->i_e2fs;
ump = ip->i_ump;
if (fs->e2fs_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 = 1; i < len; i++)
if (buflist->bs_children[i]->b_lblkno != start_lbn + i)
panic("ext2_reallocblks: non-cluster");
#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 (ext2_getlbns(vp, start_lbn, start_ap, &start_lvl) ||
ext2_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_db[0];
soff = start_lbn;
} else {
idp = &start_ap[start_lvl - 1];
if (bread(vp, idp->in_lbn, (int)fs->e2fs_bsize, NOCRED, &sbp)) {
brelse(sbp);
return (ENOSPC);
}
sbap = (u_int *)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_ap[start_lvl-1].in_lbn == idp->in_lbn)
panic("ext2_reallocblks: start == end");
#endif
ssize = len - (idp->in_off + 1);
if (bread(vp, idp->in_lbn, (int)fs->e2fs_bsize, NOCRED, &ebp))
goto fail;
ebap = (u_int *)ebp->b_data;
}
/*
* Find the preferred location for the cluster.
*/
EXT2_LOCK(ump);
pref = ext2_blkpref(ip, start_lbn, soff, sbap, 0);
/*
* Search the block map looking for an allocation of the desired size.
*/
if ((newblk = (e2fs_daddr_t)ext2_hashalloc(ip, dtog(fs, pref), pref,
len, ext2_clusteralloc)) == 0) {
EXT2_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
printf("realloc: ino %d, lbns %jd-%jd\n\told:", ip->i_number,
(intmax_t)start_lbn, (intmax_t)end_lbn);
#endif /* DEBUG */
blkno = newblk;
for (bap = &sbap[soff], i = 0; i < len; i++, blkno += fs->e2fs_fpb) {
if (i == ssize) {
bap = ebap;
soff = -i;
}
#ifdef INVARIANTS
if (buflist->bs_children[i]->b_blkno != fsbtodb(fs, *bap))
panic("ext2_reallocblks: alloc mismatch");
#endif
#ifdef DEBUG
printf(" %d,", *bap);
#endif /* DEBUG */
*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_db[0]) {
if (doasyncfree)
bdwrite(sbp);
else
bwrite(sbp);
} else {
ip->i_flag |= IN_CHANGE | IN_UPDATE;
if (!doasyncfree)
ext2_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
printf("\n\tnew:");
#endif /* DEBUG */
for (blkno = newblk, i = 0; i < len; i++, blkno += fs->e2fs_fpb) {
ext2_blkfree(ip, dbtofsb(fs, buflist->bs_children[i]->b_blkno),
fs->e2fs_bsize);
buflist->bs_children[i]->b_blkno = fsbtodb(fs, blkno);
#ifdef DEBUG
printf(" %d,", blkno);
#endif /* DEBUG */
}
#ifdef DEBUG
printf("\n");
#endif /* DEBUG */
return (0);
fail:
if (ssize < len)
brelse(ebp);
if (sbap != &ip->i_db[0])
brelse(sbp);
return (ENOSPC);
}
