static union dinode *
get_inode(int fd, struct fs *super, ino_t ino)
{
static caddr_t ipbuf;
static struct cg *cgp;
static ino_t last;
static int cg;
struct ufs2_dinode *di2;
if (fd < 0) { /* flush cache */
if (ipbuf) {
free(ipbuf);
ipbuf = NULL;
if (super != NULL && super->fs_magic == FS_UFS2_MAGIC) {
free(cgp);
cgp = NULL;
}
}
return 0;
}
if (!ipbuf || ino < last || ino >= last + INOCNT(super)) {
if (super->fs_magic == FS_UFS2_MAGIC &&
(!cgp || cg != ino_to_cg(super, ino))) {
cg = ino_to_cg(super, ino);
if (!cgp && !(cgp = malloc(super->fs_cgsize)))
errx(1, "allocate cg");
if (pread(fd, cgp, super->fs_cgsize,
(off_t)cgtod(super, cg) << super->fs_fshift)
!= super->fs_cgsize)
if (read(fd, cgp, super->fs_cgsize) != super->fs_cgsize)
err(1, "read cg");
if (!cg_chkmagic(cgp))
errx(1, "cg has bad magic");
}
if (!ipbuf && !(ipbuf = malloc(INOSZ(super))))
err(1, "allocate inodes");
last = (ino / INOCNT(super)) * INOCNT(super);
if (lseek(fd, (off_t)ino_to_fsba(super, last)
<< super->fs_fshift, SEEK_SET) < 0 ||
read(fd, ipbuf, INOSZ(super)) != INOSZ(super)) {
err(1, "read inodes");
}
}
if (super->fs_magic == FS_UFS1_MAGIC)
return ((union dinode *)
&((struct ufs1_dinode *)ipbuf)[ino % INOCNT(super)]);
di2 = &((struct ufs2_dinode *)ipbuf)[ino % INOCNT(super)];
/* If the inode is unused, it might be unallocated too, so zero it. */
if (isclr(cg_inosused(cgp), ino % super->fs_ipg))
memset(di2, 0, sizeof(*di2));
return ((union dinode *)di2);
}
/*
* 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. Otherwise, the policy is to try to allocate the blocks
* contigously. The two fields of the ext2 inode extension (see
* ufs/ufs/inode.h) help this.
*/
daddr_t
ext2fs_blkpref(struct inode *ip, int32_t lbn, int indx, int32_t *bap)
{
struct m_ext2fs *fs;
int cg, i;
fs = ip->i_e2fs;
/*
* if we are doing contigous lbn allocation, try to alloc blocks
* contigously on disk
*/
if ( ip->i_e2fs_last_blk && lbn == ip->i_e2fs_last_lblk + 1) {
return ip->i_e2fs_last_blk + 1;
}
/*
* bap, if provided, gives us a list of blocks to which we want to
* stay close
*/
if (bap) {
for (i = indx; i >= 0 ; i--) {
if (bap[i]) {
return fs2h32(bap[i]) + 1;
}
}
}
/* fall back to the first block of the cylinder containing the inode */
cg = ino_to_cg(fs, ip->i_number);
return fs->e2fs.e2fs_bpg * cg + fs->e2fs.e2fs_first_dblock + 1;
}
/*
* Allocate an inode on the disk
*/
void
iput(union dinode *ip, ino_t ino)
{
ufs2_daddr_t d;
int c;
c = ino_to_cg(&sblock, ino);
bread(&disk, part_ofs + fsbtodb(&sblock, cgtod(&sblock, 0)), (char *)&acg,
sblock.fs_cgsize);
if (acg.cg_magic != CG_MAGIC) {
printf("cg 0: bad magic number\n");
exit(31);
}
acg.cg_cs.cs_nifree--;
setbit(cg_inosused(&acg), ino);
wtfs(fsbtodb(&sblock, cgtod(&sblock, 0)), sblock.fs_cgsize,
(char *)&acg);
sblock.fs_cstotal.cs_nifree--;
fscs[0].cs_nifree--;
if (ino >= (unsigned long)sblock.fs_ipg * sblock.fs_ncg) {
printf("fsinit: inode value out of range (%d).\n", ino);
exit(32);
}
d = fsbtodb(&sblock, ino_to_fsba(&sblock, ino));
bread(&disk, part_ofs + d, (char *)iobuf, sblock.fs_bsize);
if (sblock.fs_magic == FS_UFS1_MAGIC)
((struct ufs1_dinode *)iobuf)[ino_to_fsbo(&sblock, ino)] =
ip->dp1;
else
((struct ufs2_dinode *)iobuf)[ino_to_fsbo(&sblock, ino)] =
ip->dp2;
wtfs(d, sblock.fs_bsize, (char *)iobuf);
}
/*
* Allocate a block in the file system.
*
* 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(struct inode *ip, daddr_t lbn __unused, daddr_t bpref, int size,
daddr_t *bnp)
{
struct fs *fs = ip->i_fs;
daddr_t bno;
int cg;
*bnp = 0;
if (size > fs->fs_bsize || ffs_fragoff(fs, size) != 0) {
errx(1, "ffs_alloc: bad size: bsize %d size %d",
fs->fs_bsize, size);
}
if (size == fs->fs_bsize && fs->fs_cstotal.cs_nbfree == 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, ffs_alloccg);
if (bno > 0) {
DIP_ADD(ip, blocks, size / DEV_BSIZE);
*bnp = bno;
return (0);
}
nospace:
return (ENOSPC);
}
/*
* allocate an unused inode
*/
ino_t
allocino(ino_t request, int type)
{
ino_t ino;
union dinode *dp;
struct bufarea *cgbp;
struct cg *cgp;
int cg;
if (request == 0)
request = ROOTINO;
else if (inoinfo(request)->ino_state != USTATE)
return (0);
for (ino = request; ino < maxino; ino++)
if (inoinfo(ino)->ino_state == USTATE)
break;
if (ino == maxino)
return (0);
cg = ino_to_cg(&sblock, ino);
cgbp = cgget(cg);
cgp = cgbp->b_un.b_cg;
if (!check_cgmagic(cg, cgbp))
return (0);
setbit(cg_inosused(cgp), ino % sblock.fs_ipg);
cgp->cg_cs.cs_nifree--;
switch (type & IFMT) {
case IFDIR:
inoinfo(ino)->ino_state = DSTATE;
cgp->cg_cs.cs_ndir++;
break;
case IFREG:
case IFLNK:
inoinfo(ino)->ino_state = FSTATE;
break;
default:
return (0);
}
dirty(cgbp);
dp = ginode(ino);
DIP_SET(dp, di_db[0], allocblk((long)1));
if (DIP(dp, di_db[0]) == 0) {
inoinfo(ino)->ino_state = USTATE;
return (0);
}
DIP_SET(dp, di_mode, type);
DIP_SET(dp, di_flags, 0);
DIP_SET(dp, di_atime, time(NULL));
DIP_SET(dp, di_ctime, DIP(dp, di_atime));
DIP_SET(dp, di_mtime, DIP(dp, di_ctime));
DIP_SET(dp, di_mtimensec, 0);
DIP_SET(dp, di_ctimensec, 0);
DIP_SET(dp, di_atimensec, 0);
DIP_SET(dp, di_size, sblock.fs_fsize);
DIP_SET(dp, di_blocks, btodb(sblock.fs_fsize));
n_files++;
inodirty();
inoinfo(ino)->ino_type = IFTODT(type);
return (ino);
}
/*
* Change the number of unreferenced inodes.
*/
static int
ufs_gjournal_modref(struct vnode *vp, int count)
{
struct cg *cgp;
struct buf *bp;
ufs2_daddr_t cgbno;
int error, cg;
struct cdev *dev;
struct inode *ip;
struct ufsmount *ump;
struct fs *fs;
struct vnode *devvp;
ino_t ino;
ip = VTOI(vp);
ump = VFSTOUFS(vp->v_mount);
fs = ump->um_fs;
devvp = ump->um_devvp;
ino = ip->i_number;
cg = ino_to_cg(fs, ino);
if (devvp->v_type == VREG) {
/* devvp is a snapshot */
dev = VFSTOUFS(devvp->v_mount)->um_devvp->v_rdev;
cgbno = fragstoblks(fs, cgtod(fs, cg));
} else if (devvp->v_type == VCHR) {
/* devvp is a normal disk device */
dev = devvp->v_rdev;
cgbno = fsbtodb(fs, cgtod(fs, cg));
} else {
bp = NULL;
return (EIO);
}
if ((u_int)ino >= fs->fs_ipg * fs->fs_ncg)
panic("ufs_gjournal_modref: range: dev = %s, ino = %lu, fs = %s",
devtoname(dev), (u_long)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_unrefs += count;
UFS_LOCK(ump);
fs->fs_unrefs += count;
fs->fs_fmod = 1;
ACTIVECLEAR(fs, cg);
UFS_UNLOCK(ump);
bdwrite(bp);
return (0);
}
/*
* allocate an unused inode
*/
ufs1_ino_t
allocino(ufs1_ino_t request, int type)
{
ufs1_ino_t ino;
struct ufs1_dinode *dp;
struct cg *cgp = &cgrp;
int cg;
if (request == 0)
request = ROOTINO;
else if (inoinfo(request)->ino_state != USTATE)
return (0);
for (ino = request; ino < maxino; ino++)
if (inoinfo(ino)->ino_state == USTATE)
break;
if (ino == maxino)
return (0);
cg = ino_to_cg(&sblock, ino);
getblk(&cgblk, cgtod(&sblock, cg), sblock.fs_cgsize);
if (!cg_chkmagic(cgp))
pfatal("CG %d: BAD MAGIC NUMBER\n", cg);
setbit(cg_inosused(cgp), ino % sblock.fs_ipg);
cgp->cg_cs.cs_nifree--;
switch (type & IFMT) {
case IFDIR:
inoinfo(ino)->ino_state = DSTATE;
cgp->cg_cs.cs_ndir++;
break;
case IFREG:
case IFLNK:
inoinfo(ino)->ino_state = FSTATE;
break;
default:
return (0);
}
cgdirty();
dp = ginode(ino);
dp->di_db[0] = allocblk((long)1);
if (dp->di_db[0] == 0) {
inoinfo(ino)->ino_state = USTATE;
return (0);
}
dp->di_mode = type;
dp->di_flags = 0;
dp->di_atime = time(NULL);
dp->di_mtime = dp->di_ctime = dp->di_atime;
dp->di_mtimensec = dp->di_ctimensec = dp->di_atimensec = 0;
dp->di_size = sblock.fs_fsize;
dp->di_blocks = btodb(sblock.fs_fsize);
n_files++;
inodirty();
if (newinofmt)
inoinfo(ino)->ino_type = IFTODT(type);
return (ino);
}
daddr_t
ffs_blkpref_ufs2(struct inode *ip, daddr_t lbn, int indx, int64_t *bap)
{
struct fs *fs;
int cg;
int avgbfree, startcg;
fs = ip->i_fs;
if (indx % fs->fs_maxbpg == 0 || bap[indx - 1] == 0) {
if (lbn < UFS_NDADDR + FFS_NINDIR(fs)) {
cg = ino_to_cg(fs, ip->i_number);
return (fs->fs_fpg * 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,
ufs_rw64(bap[indx - 1], UFS_FSNEEDSWAP(fs)) + 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) {
return (fs->fs_fpg * cg + fs->fs_frag);
}
for (cg = 0; cg < startcg; cg++)
if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) {
return (fs->fs_fpg * cg + fs->fs_frag);
}
return (0);
}
/*
* We just always try to lay things out contiguously.
*/
return ufs_rw64(bap[indx - 1], UFS_FSNEEDSWAP(fs)) + fs->fs_frag;
}
/*
* Allocate an inode in the file system.
*
* If allocating a directory, use ext2fs_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
ext2fs_valloc(struct vnode *pvp, int mode, kauth_cred_t cred,
struct vnode **vpp)
{
struct inode *pip;
struct m_ext2fs *fs;
struct inode *ip;
ino_t ino, ipref;
int cg, error;
*vpp = NULL;
pip = VTOI(pvp);
fs = pip->i_e2fs;
if (fs->e2fs.e2fs_ficount == 0)
goto noinodes;
if ((mode & IFMT) == IFDIR)
cg = ext2fs_dirpref(fs);
else
cg = ino_to_cg(fs, pip->i_number);
ipref = cg * fs->e2fs.e2fs_ipg + 1;
ino = (ino_t)ext2fs_hashalloc(pip, cg, (long)ipref, mode, ext2fs_nodealloccg);
if (ino == 0)
goto noinodes;
error = VFS_VGET(pvp->v_mount, ino, vpp);
if (error) {
ext2fs_vfree(pvp, ino, mode);
return (error);
}
ip = VTOI(*vpp);
if (ip->i_e2fs_mode && ip->i_e2fs_nlink != 0) {
printf("mode = 0%o, nlinks %d, inum = %llu, fs = %s\n",
ip->i_e2fs_mode, ip->i_e2fs_nlink,
(unsigned long long)ip->i_number, fs->e2fs_fsmnt);
panic("ext2fs_valloc: dup alloc");
}
memset(ip->i_din.e2fs_din, 0, sizeof(struct ext2fs_dinode));
/*
* Set up a new generation number for this inode.
*/
if (++ext2gennumber < time_second)
ext2gennumber = time_second;
ip->i_e2fs_gen = ext2gennumber;
return (0);
noinodes:
ext2fs_fserr(fs, kauth_cred_geteuid(cred), "out of inodes");
uprintf("\n%s: create/symlink failed, no inodes free\n", fs->e2fs_fsmnt);
return (ENOSPC);
}
/*
* 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 an inode.
*
* The specified inode is placed back in the free map.
*/
int
ext2fs_vfree(struct vnode *pvp, ino_t ino, int mode)
{
struct m_ext2fs *fs;
char *ibp;
struct inode *pip;
struct buf *bp;
int error, cg;
pip = VTOI(pvp);
fs = pip->i_e2fs;
if ((u_int)ino > fs->e2fs.e2fs_icount || (u_int)ino < EXT2_FIRSTINO)
panic("ifree: range: dev = 0x%llx, ino = %llu, fs = %s",
(unsigned long long)pip->i_dev, (unsigned long long)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, B_MODIFY, &bp);
if (error) {
brelse(bp, 0);
return (0);
}
ibp = (char *)bp->b_data;
ino = (ino - 1) % fs->e2fs.e2fs_ipg;
if (isclr(ibp, ino)) {
printf("dev = 0x%llx, ino = %llu, fs = %s\n",
(unsigned long long)pip->i_dev,
(unsigned long long)ino, fs->e2fs_fsmnt);
if (fs->e2fs_ronly == 0)
panic("ifree: freeing free inode");
}
clrbit(ibp, ino);
fs->e2fs.e2fs_ficount++;
fs->e2fs_gd[cg].ext2bgd_nifree++;
if ((mode & IFMT) == IFDIR) {
fs->e2fs_gd[cg].ext2bgd_ndirs--;
}
fs->e2fs_fmod = 1;
bdwrite(bp);
return (0);
}
/*
* Allocate a block in the file system.
*
* 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
ext2fs_alloc(struct inode *ip, daddr_t lbn, daddr_t bpref,
kauth_cred_t cred, daddr_t *bnp)
{
struct m_ext2fs *fs;
daddr_t bno;
int cg;
*bnp = 0;
fs = ip->i_e2fs;
#ifdef DIAGNOSTIC
if (cred == NOCRED)
panic("ext2fs_alloc: missing credential");
#endif /* DIAGNOSTIC */
if (fs->e2fs.e2fs_fbcount == 0)
goto nospace;
if (kauth_authorize_system(cred, KAUTH_SYSTEM_FS_RESERVEDSPACE, 0, NULL,
NULL, NULL) != 0 &&
freespace(fs) <= 0)
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)ext2fs_hashalloc(ip, cg, bpref, fs->e2fs_bsize,
ext2fs_alloccg);
if (bno > 0) {
ip->i_e2fs_nblock += btodb(fs->e2fs_bsize);
ip->i_flag |= IN_CHANGE | IN_UPDATE;
*bnp = bno;
return (0);
}
nospace:
ext2fs_fserr(fs, kauth_cred_geteuid(cred), "file system full");
uprintf("\n%s: write failed, file system is full\n", fs->e2fs_fsmnt);
return (ENOSPC);
}
/*
* Allocate a block in the file system.
*
* 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) quadratically 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) quadratically rehash into other cylinder groups, until an
* available block is located.
*/
int
ext2fs_alloc(struct inode *ip, int32_t lbn, int32_t bpref,
struct ucred *cred, int32_t *bnp)
{
struct m_ext2fs *fs;
int32_t bno;
int cg;
*bnp = 0;
fs = ip->i_e2fs;
#ifdef DIAGNOSTIC
if (cred == NOCRED)
panic("ext2fs_alloc: missing credential");
#endif /* DIAGNOSTIC */
if (fs->e2fs.e2fs_fbcount == 0)
goto nospace;
if (cred->cr_uid != 0 && freespace(fs) <= 0)
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 = (int32_t)ext2fs_hashalloc(ip, cg, bpref, fs->e2fs_bsize,
ext2fs_alloccg);
if (bno > 0) {
ip->i_e2fs_nblock += btodb(fs->e2fs_bsize);
ip->i_flag |= IN_CHANGE | IN_UPDATE;
*bnp = bno;
return (0);
}
nospace:
ext2fs_fserr(fs, cred->cr_uid, "file system full");
uprintf("\n%s: write failed, file system is full\n", fs->e2fs_fsmnt);
return (ENOSPC);
}
/*
* Look up an 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.
*/
static int
ext2_vget(struct mount *mp, ino_t ino, int flags, struct vnode **vpp)
{
struct m_ext2fs *fs;
struct inode *ip;
struct ext2mount *ump;
struct buf *bp;
struct vnode *vp;
struct cdev *dev;
struct thread *td;
int i, error;
int used_blocks;
td = curthread;
error = vfs_hash_get(mp, ino, flags, td, vpp, NULL, NULL);
if (error || *vpp != NULL)
return (error);
ump = VFSTOEXT2(mp);
dev = ump->um_dev;
/*
* If this malloc() is performed after the getnewvnode()
* it might block, leaving a vnode with a NULL v_data to be
* found by ext2_sync() if a sync happens to fire right then,
* which will cause a panic because ext2_sync() blindly
* dereferences vp->v_data (as well it should).
*/
ip = malloc(sizeof(struct inode), M_EXT2NODE, M_WAITOK | M_ZERO);
/* Allocate a new vnode/inode. */
if ((error = getnewvnode("ext2fs", mp, &ext2_vnodeops, &vp)) != 0) {
*vpp = NULL;
free(ip, M_EXT2NODE);
return (error);
}
vp->v_data = ip;
ip->i_vnode = vp;
ip->i_e2fs = fs = ump->um_e2fs;
ip->i_ump = ump;
ip->i_number = ino;
lockmgr(vp->v_vnlock, LK_EXCLUSIVE, NULL);
error = insmntque(vp, mp);
if (error != 0) {
free(ip, M_EXT2NODE);
*vpp = NULL;
return (error);
}
error = vfs_hash_insert(vp, ino, flags, td, vpp, NULL, NULL);
if (error || *vpp != NULL)
return (error);
/* Read in the disk contents for the inode, copy into the inode. */
if ((error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
(int)fs->e2fs_bsize, NOCRED, &bp)) != 0) {
/*
* 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().
*/
brelse(bp);
vput(vp);
*vpp = NULL;
return (error);
}
/* convert ext2 inode to dinode */
ext2_ei2i((struct ext2fs_dinode *) ((char *)bp->b_data + EXT2_INODE_SIZE(fs) *
ino_to_fsbo(fs, ino)), ip);
ip->i_block_group = ino_to_cg(fs, ino);
ip->i_next_alloc_block = 0;
ip->i_next_alloc_goal = 0;
/*
* Now we want to make sure that block pointers for unused
* blocks are zeroed out - ext2_balloc depends on this
* although for regular files and directories only
*/
if(S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode)) {
used_blocks = (ip->i_size+fs->e2fs_bsize-1) / fs->e2fs_bsize;
for (i = used_blocks; i < EXT2_NDIR_BLOCKS; i++)
ip->i_db[i] = 0;
}
/*
ext2_print_inode(ip);
*/
bqrelse(bp);
/*
* Initialize the vnode from the inode, check for aliases.
* Note that the underlying vnode may have changed.
*/
if ((error = ext2_vinit(mp, &ext2_fifoops, &vp)) != 0) {
vput(vp);
*vpp = NULL;
return (error);
}
/*
* Finish inode initialization.
*/
/*
* 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_gen == 0) {
ip->i_gen = random() / 2 + 1;
if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
ip->i_flag |= IN_MODIFIED;
}
*vpp = vp;
return (0);
}
/*
* Look up an 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.
*/
static int
ext2_vget(struct mount *mp, ino_t ino, int flags, struct vnode **vpp)
{
struct m_ext2fs *fs;
struct inode *ip;
struct ext2mount *ump;
struct buf *bp;
struct vnode *vp;
struct thread *td;
int i, error;
int used_blocks;
td = curthread;
error = vfs_hash_get(mp, ino, flags, td, vpp, NULL, NULL);
if (error || *vpp != NULL)
return (error);
ump = VFSTOEXT2(mp);
ip = malloc(sizeof(struct inode), M_EXT2NODE, M_WAITOK | M_ZERO);
/* Allocate a new vnode/inode. */
if ((error = getnewvnode("ext2fs", mp, &ext2_vnodeops, &vp)) != 0) {
*vpp = NULL;
free(ip, M_EXT2NODE);
return (error);
}
vp->v_data = ip;
ip->i_vnode = vp;
ip->i_e2fs = fs = ump->um_e2fs;
ip->i_ump = ump;
ip->i_number = ino;
lockmgr(vp->v_vnlock, LK_EXCLUSIVE, NULL);
error = insmntque(vp, mp);
if (error != 0) {
free(ip, M_EXT2NODE);
*vpp = NULL;
return (error);
}
error = vfs_hash_insert(vp, ino, flags, td, vpp, NULL, NULL);
if (error || *vpp != NULL)
return (error);
/* Read in the disk contents for the inode, copy into the inode. */
if ((error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
(int)fs->e2fs_bsize, NOCRED, &bp)) != 0) {
/*
* 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().
*/
brelse(bp);
vput(vp);
*vpp = NULL;
return (error);
}
/* convert ext2 inode to dinode */
ext2_ei2i((struct ext2fs_dinode *)((char *)bp->b_data + EXT2_INODE_SIZE(fs) *
ino_to_fsbo(fs, ino)), ip);
ip->i_block_group = ino_to_cg(fs, ino);
ip->i_next_alloc_block = 0;
ip->i_next_alloc_goal = 0;
/*
* Now we want to make sure that block pointers for unused
* blocks are zeroed out - ext2_balloc depends on this
* although for regular files and directories only
*
* If IN_E4EXTENTS is enabled, unused blocks are not zeroed
* out because we could corrupt the extent tree.
*/
if (!(ip->i_flag & IN_E4EXTENTS) &&
(S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode))) {
used_blocks = howmany(ip->i_size, fs->e2fs_bsize);
for (i = used_blocks; i < EXT2_NDIR_BLOCKS; i++)
ip->i_db[i] = 0;
}
#ifdef EXT2FS_DEBUG
ext2_print_inode(ip);
#endif
bqrelse(bp);
/*
* Initialize the vnode from the inode, check for aliases.
* Note that the underlying vnode may have changed.
*/
if ((error = ext2_vinit(mp, &ext2_fifoops, &vp)) != 0) {
vput(vp);
*vpp = NULL;
return (error);
}
/*
* Finish inode initialization.
*/
*vpp = vp;
return (0);
}
/*
* Find a cylinder 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 u_long
ext2_dirpref(struct inode *pip)
{
struct m_ext2fs *fs;
int cg, prefcg, cgsize;
u_int avgifree, avgbfree, avgndir, curdirsize;
u_int minifree, minbfree, maxndir;
u_int mincg, minndir;
u_int dirsize, maxcontigdirs;
mtx_assert(EXT2_MTX(pip->i_ump), MA_OWNED);
fs = pip->i_e2fs;
avgifree = fs->e2fs->e2fs_ficount / fs->e2fs_gcount;
avgbfree = fs->e2fs->e2fs_fbcount / fs->e2fs_gcount;
avgndir = fs->e2fs_total_dir / fs->e2fs_gcount;
/*
* Force allocation in another cg if creating a first level dir.
*/
ASSERT_VOP_LOCKED(ITOV(pip), "ext2fs_dirpref");
if (ITOV(pip)->v_vflag & VV_ROOT) {
prefcg = arc4random() % fs->e2fs_gcount;
mincg = prefcg;
minndir = fs->e2fs_ipg;
for (cg = prefcg; cg < fs->e2fs_gcount; cg++)
if (fs->e2fs_gd[cg].ext2bgd_ndirs < minndir &&
fs->e2fs_gd[cg].ext2bgd_nifree >= avgifree &&
fs->e2fs_gd[cg].ext2bgd_nbfree >= avgbfree) {
mincg = cg;
minndir = fs->e2fs_gd[cg].ext2bgd_ndirs;
}
for (cg = 0; cg < prefcg; cg++)
if (fs->e2fs_gd[cg].ext2bgd_ndirs < minndir &&
fs->e2fs_gd[cg].ext2bgd_nifree >= avgifree &&
fs->e2fs_gd[cg].ext2bgd_nbfree >= avgbfree) {
mincg = cg;
minndir = fs->e2fs_gd[cg].ext2bgd_ndirs;
}
return (mincg);
}
/*
* Count various limits which used for
* optimal allocation of a directory inode.
*/
maxndir = min(avgndir + fs->e2fs_ipg / 16, fs->e2fs_ipg);
minifree = avgifree - avgifree / 4;
if (minifree < 1)
minifree = 1;
minbfree = avgbfree - avgbfree / 4;
if (minbfree < 1)
minbfree = 1;
cgsize = fs->e2fs_fsize * fs->e2fs_fpg;
dirsize = AVGDIRSIZE;
curdirsize = avgndir ? (cgsize - avgbfree * fs->e2fs_bsize) / avgndir : 0;
if (dirsize < curdirsize)
dirsize = curdirsize;
maxcontigdirs = min((avgbfree * fs->e2fs_bsize) / dirsize, 255);
maxcontigdirs = min(maxcontigdirs, fs->e2fs_ipg / AFPDIR);
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->e2fs_gcount; cg++)
if (fs->e2fs_gd[cg].ext2bgd_ndirs < maxndir &&
fs->e2fs_gd[cg].ext2bgd_nifree >= minifree &&
fs->e2fs_gd[cg].ext2bgd_nbfree >= minbfree) {
if (fs->e2fs_contigdirs[cg] < maxcontigdirs)
return (cg);
}
for (cg = 0; cg < prefcg; cg++)
if (fs->e2fs_gd[cg].ext2bgd_ndirs < maxndir &&
fs->e2fs_gd[cg].ext2bgd_nifree >= minifree &&
fs->e2fs_gd[cg].ext2bgd_nbfree >= minbfree) {
if (fs->e2fs_contigdirs[cg] < maxcontigdirs)
return (cg);
}
/*
* This is a backstop when we have deficit in space.
*/
for (cg = prefcg; cg < fs->e2fs_gcount; cg++)
if (fs->e2fs_gd[cg].ext2bgd_nifree >= avgifree)
return (cg);
for (cg = 0; cg < prefcg; cg++)
if (fs->e2fs_gd[cg].ext2bgd_nifree >= avgifree)
break;
return (cg);
}
/*
* 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);
}
