/*
* Allocate an inode on the disk
*/
void
iput(union dinode *ip, ino_t ino)
{
ufs2_daddr_t d;
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 (%ju).\n",
(uintmax_t)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);
}
/*
* Load up the contents of an inode and copy the appropriate pieces
* to the incore copy.
*/
void
ffs_load_inode(struct buf *bp, struct inode *ip, struct fs *fs, ino_t ino)
{
if (I_IS_UFS1(ip)) {
*ip->i_din1 =
*((struct ufs1_dinode *)bp->b_data + ino_to_fsbo(fs, ino));
ip->i_mode = ip->i_din1->di_mode;
ip->i_nlink = ip->i_din1->di_nlink;
ip->i_size = ip->i_din1->di_size;
ip->i_flags = ip->i_din1->di_flags;
ip->i_gen = ip->i_din1->di_gen;
ip->i_uid = ip->i_din1->di_uid;
ip->i_gid = ip->i_din1->di_gid;
} else {
*ip->i_din2 =
*((struct ufs2_dinode *)bp->b_data + ino_to_fsbo(fs, ino));
ip->i_mode = ip->i_din2->di_mode;
ip->i_nlink = ip->i_din2->di_nlink;
ip->i_size = ip->i_din2->di_size;
ip->i_flags = ip->i_din2->di_flags;
ip->i_gen = ip->i_din2->di_gen;
ip->i_uid = ip->i_din2->di_uid;
ip->i_gid = ip->i_din2->di_gid;
}
}
/*
* Update the access, modified, and inode change times as specified by the
* IN_ACCESS, IN_UPDATE, and IN_CHANGE flags respectively. Write the inode
* to disk if the IN_MODIFIED flag is set (it may be set initially, or by
* the timestamp update). The IN_LAZYMOD flag is set to force a write
* later if not now. If we write now, then clear both IN_MODIFIED and
* IN_LAZYMOD to reflect the presumably successful write, and if waitfor is
* set, then wait for the write to complete.
*/
int
ext2_update(struct vnode *vp, int waitfor)
{
struct m_ext2fs *fs;
struct buf *bp;
struct inode *ip;
int error;
ASSERT_VOP_ELOCKED(vp, "ext2_update");
ext2_itimes(vp);
ip = VTOI(vp);
if ((ip->i_flag & IN_MODIFIED) == 0 && waitfor == 0)
return (0);
ip->i_flag &= ~(IN_LAZYACCESS | IN_LAZYMOD | IN_MODIFIED);
fs = ip->i_e2fs;
if(fs->e2fs_ronly)
return (0);
if ((error = bread(ip->i_devvp,
fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
(int)fs->e2fs_bsize, NOCRED, &bp)) != 0) {
brelse(bp);
return (error);
}
ext2_i2ei(ip, (struct ext2fs_dinode *)((char *)bp->b_data +
EXT2_INODE_SIZE(fs) * ino_to_fsbo(fs, ip->i_number)));
if (waitfor && !DOINGASYNC(vp))
return (bwrite(bp));
else {
bdwrite(bp);
return (0);
}
}
/*
* Sanity check the disk vnode content, and copy it over to inode structure.
*/
static int
ext2fs_loadvnode_content(struct m_ext2fs *fs, ino_t ino, struct buf *bp, struct inode *ip)
{
struct ext2fs_dinode *din;
int error = 0;
din = (struct ext2fs_dinode *)((char *)bp->b_data + (ino_to_fsbo(fs, ino) * EXT2_DINODE_SIZE(fs)));
/* sanity checks - inode data NOT byteswapped at this point */
if (EXT2_DINODE_FITS(din, e2di_extra_isize, EXT2_DINODE_SIZE(fs))
&& (EXT2_DINODE_SIZE(fs) - EXT2_REV0_DINODE_SIZE) < fs2h16(din->e2di_extra_isize))
{
printf("ext2fs: inode %"PRIu64" bad extra_isize %u",
ino, din->e2di_extra_isize);
error = EINVAL;
goto bad;
}
/* everything allright, proceed with copy */
if (ip->i_din.e2fs_din == NULL)
ip->i_din.e2fs_din = kmem_alloc(EXT2_DINODE_SIZE(fs), KM_SLEEP);
e2fs_iload(din, ip->i_din.e2fs_din, EXT2_DINODE_SIZE(fs));
ext2fs_set_inode_guid(ip);
bad:
return error;
}
static long ReadInode( long inodeNum, InodePtr inode, long * flags, long * time )
{
long fragNum = ino_to_fsba(gFS, inodeNum);
long blockOffset = ino_to_fsbo(gFS, inodeNum) * sizeof(Inode);
ReadBlock(fragNum, blockOffset, sizeof(Inode), (char *)inode, 1);
byte_swap_dinode_in(inode);
if (time != 0) *time = inode->di_mtime;
if (flags != 0) {
switch (inode->di_mode & IFMT) {
case IFREG: *flags = kFileTypeFlat; break;
case IFDIR: *flags = kFileTypeDirectory; break;
case IFLNK: *flags = kFileTypeLink; break;
default : *flags = kFileTypeUnknown; break;
}
*flags |= inode->di_mode & kPermMask;
if (inode->di_uid != 0) *flags |= kOwnerNotRoot;
}
return 0;
}
/*
* Update the access, modified, and inode change times as specified by the
* IN_ACCESS, IN_UPDATE, and IN_CHANGE flags respectively. Write the inode
* to disk if the IN_MODIFIED flag is set (it may be set initially, or by
* the timestamp update). The IN_LAZYMOD flag is set to force a write
* later if not now. If we write now, then clear both IN_MODIFIED and
* IN_LAZYMOD to reflect the presumably successful write, and if waitfor is
* set, then wait for the write to complete.
*/
int
ffs_update(struct vnode *vp, int waitfor)
{
struct fs *fs;
struct buf *bp;
struct inode *ip;
int error;
ufs_itimes(vp);
ip = VTOI(vp);
if ((ip->i_flag & IN_MODIFIED) == 0 && waitfor == 0)
return (0);
ip->i_flag &= ~(IN_LAZYMOD | IN_MODIFIED);
fs = ip->i_fs;
if (fs->fs_ronly)
return (0);
/*
* The vnode type is usually set to VBAD if an unrecoverable I/O
* error has occured (such as when reading the inode). Clear the
* modified bits but do not write anything out in this case.
*/
if (vp->v_type == VBAD)
return (0);
/*
* Ensure that uid and gid are correct. This is a temporary
* fix until fsck has been changed to do the update.
*/
if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
ip->i_din.di_ouid = ip->i_uid; /* XXX */
ip->i_din.di_ogid = ip->i_gid; /* XXX */
} /* XXX */
error = bread(ip->i_devvp,
fsbtodoff(fs, ino_to_fsba(fs, ip->i_number)),
(int)fs->fs_bsize, &bp);
if (error) {
brelse(bp);
return (error);
}
if (DOINGSOFTDEP(vp))
softdep_update_inodeblock(ip, bp, waitfor);
else if (ip->i_effnlink != ip->i_nlink)
panic("ffs_update: bad link cnt");
*((struct ufs1_dinode *)bp->b_data +
ino_to_fsbo(fs, ip->i_number)) = ip->i_din;
if (waitfor && !DOINGASYNC(vp)) {
return (bwrite(bp));
} else if (vm_page_count_severe() || buf_dirty_count_severe()) {
return (bwrite(bp));
} else {
if (bp->b_bufsize == fs->fs_bsize)
bp->b_flags |= B_CLUSTEROK;
bdwrite(bp);
return (0);
}
}
/*
* Read a new inode into a file structure.
*/
static int
read_inode(ino32_t inumber, struct open_file *f)
{
struct file *fp = (struct file *)f->f_fsdata;
struct fs *fs = fp->f_fs;
char *buf;
size_t rsize;
int rc;
daddr_t inode_sector = 0; /* XXX: gcc */
#ifdef LIBSA_LFS
struct ufs_dinode *dip;
int cnt;
#endif
#ifdef LIBSA_LFS
if (inumber == fs->lfs_ifile)
inode_sector = FSBTODB(fs, fs->lfs_idaddr);
else if ((rc = find_inode_sector(inumber, f, &inode_sector)) != 0)
return rc;
#else
inode_sector = FSBTODB(fs, ino_to_fsba(fs, inumber));
#endif
/*
* Read inode and save it.
*/
buf = fp->f_buf;
twiddle();
rc = DEV_STRATEGY(f->f_dev)(f->f_devdata, F_READ,
inode_sector, fs->fs_bsize, buf, &rsize);
if (rc)
return rc;
if (rsize != fs->fs_bsize)
return EIO;
#ifdef LIBSA_LFS
cnt = INOPBx(fs);
dip = (struct ufs_dinode *)buf + (cnt - 1);
for (; dip->di_inumber != inumber; --dip) {
/* kernel code panics, but boot blocks which panic are Bad. */
if (--cnt == 0)
return EINVAL;
}
fp->f_di = *dip;
#else
fp->f_di = ((struct ufs_dinode *)buf)[ino_to_fsbo(fs, inumber)];
#endif
/*
* Clear out the old buffers
*/
fp->f_ind_cache_block = ~0;
fp->f_buf_blkno = -1;
return rc;
}
/*
* Allocate an inode on the disk
*/
void
iput(union dinode *ip, ino_t ino)
{
daddr64_t d;
if (Oflag <= 1)
ip->dp1.di_gen = (u_int32_t)arc4random();
else
ip->dp2.di_gen = (u_int32_t)arc4random();
rdfs(fsbtodb(&sblock, cgtod(&sblock, 0)), sblock.fs_cgsize,
(char *)&acg);
if (acg.cg_magic != CG_MAGIC)
errx(41, "cg 0: bad magic number");
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 >= sblock.fs_ipg * sblock.fs_ncg)
errx(32, "fsinit: inode value %d out of range", ino);
d = fsbtodb(&sblock, ino_to_fsba(&sblock, ino));
rdfs(d, sblock.fs_bsize, iobuf);
if (Oflag <= 1)
((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, iobuf);
}
/*
* Read a new inode into a file structure.
*/
static int
read_inode(ino_t inumber, struct open_file *f)
{
struct file *fp = (struct file *)f->f_fsdata;
struct fs *fs = fp->f_fs;
char *buf;
size_t rsize;
int rc;
if (fs == NULL)
panic("fs == NULL");
/*
* Read inode and save it.
*/
buf = malloc(fs->fs_bsize);
twiddle();
rc = (f->f_dev->dv_strategy)(f->f_devdata, F_READ,
fsbtodb(fs, ino_to_fsba(fs, inumber)), fs->fs_bsize,
buf, &rsize);
if (rc)
goto out;
if (rsize != fs->fs_bsize) {
rc = EIO;
goto out;
}
{
struct ufs1_dinode *dp;
dp = (struct ufs1_dinode *)buf;
fp->f_di = dp[ino_to_fsbo(fs, inumber)];
}
/*
* Clear out the old buffers
*/
{
int level;
for (level = 0; level < NIADDR; level++)
fp->f_blkno[level] = -1;
fp->f_buf_blkno = -1;
}
out:
free(buf);
return (rc);
}
int
ffs_reload_vnode(struct vnode *vp, void *args)
{
struct ffs_reload_args *fra = args;
struct inode *ip;
struct buf *bp;
int error;
/*
* Step 4: invalidate all inactive vnodes.
*/
if (vp->v_usecount == 0) {
vgonel(vp, fra->p);
return (0);
}
/*
* Step 5: invalidate all cached file data.
*/
if (vget(vp, LK_EXCLUSIVE, fra->p))
return (0);
if (vinvalbuf(vp, 0, fra->cred, fra->p, 0, 0))
panic("ffs_reload: dirty2");
/*
* Step 6: re-read inode data for all active vnodes.
*/
ip = VTOI(vp);
error = bread(fra->devvp,
fsbtodb(fra->fs, ino_to_fsba(fra->fs, ip->i_number)),
(int)fra->fs->fs_bsize, NOCRED, &bp);
if (error) {
brelse(bp);
vput(vp);
return (error);
}
*ip->i_din1 = *((struct ufs1_dinode *)bp->b_data +
ino_to_fsbo(fra->fs, ip->i_number));
ip->i_effnlink = DIP(ip, nlink);
brelse(bp);
vput(vp);
return (0);
}
/*
* Update the access, modified, and inode change times as specified by the
* IACCESS, IUPDATE, and ICHANGE flags respectively. The IMODIFIED flag is
* used to specify that the inode needs to be updated but that the times have
* already been set. The access and modified times are taken from the second
* and third parameters; the inode change time is always taken from the current
* time. If waitfor is set, then wait for the disk write of the inode to
* complete.
*/
int
ext2fs_update(struct inode *ip, int waitfor)
{
struct m_ext2fs *fs;
struct buf *bp;
int error;
caddr_t cp;
if (ITOV(ip)->v_mount->mnt_flag & MNT_RDONLY)
return (0);
EXT2FS_ITIMES(ip);
if ((ip->i_flag & IN_MODIFIED) == 0)
return (0);
ip->i_flag &= ~IN_MODIFIED;
fs = ip->i_e2fs;
error = bread(ip->i_devvp,
fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
(int)fs->e2fs_bsize, &bp);
if (error) {
brelse(bp);
return (error);
}
ip->i_flag &= ~(IN_MODIFIED);
cp = (caddr_t)bp->b_data +
(ino_to_fsbo(fs, ip->i_number) * EXT2_DINODE_SIZE(fs));
/*
* See note about 16-bit UID/GID limitation in ext2fs_vget(). Now
* that we are about to write the inode, construct the split UID and
* GID fields out of the two 32-bit fields we kept in memory.
*/
ip->i_e2fs_uid_low = (u_int16_t)ip->i_e2fs_uid;
ip->i_e2fs_gid_low = (u_int16_t)ip->i_e2fs_gid;
ip->i_e2fs_uid_high = ip->i_e2fs_uid >> 16;
ip->i_e2fs_gid_high = ip->i_e2fs_gid >> 16;
e2fs_isave(fs, ip->i_e2din, (struct ext2fs_dinode *)cp);
if (waitfor)
return (bwrite(bp));
else {
bdwrite(bp);
return (0);
}
}
int
ext2fs_reload_vnode(struct vnode *vp, void *args)
{
struct ext2fs_reload_args *era = args;
struct buf *bp;
struct inode *ip;
int error;
caddr_t cp;
/*
* Step 4: invalidate all inactive vnodes.
*/
if (vp->v_usecount == 0) {
vgonel(vp, era->p);
return (0);
}
/*
* Step 5: invalidate all cached file data.
*/
if (vget(vp, LK_EXCLUSIVE, era->p))
return (0);
if (vinvalbuf(vp, 0, era->cred, era->p, 0, 0))
panic("ext2fs_reload: dirty2");
/*
* Step 6: re-read inode data for all active vnodes.
*/
ip = VTOI(vp);
error = bread(era->devvp,
fsbtodb(era->fs, ino_to_fsba(era->fs, ip->i_number)),
(int)era->fs->e2fs_bsize, &bp);
if (error) {
vput(vp);
return (error);
}
cp = (caddr_t)bp->b_data +
(ino_to_fsbo(era->fs, ip->i_number) * EXT2_DINODE_SIZE(era->fs));
e2fs_iload((struct ext2fs_dinode *)cp, ip->i_e2din);
brelse(bp);
vput(vp);
return (0);
}
/*
* Read a new inode into a file structure.
*/
static int
read_inode(ino32_t inumber, struct open_file *f)
{
struct file *fp = (struct file *)f->f_fsdata;
struct m_ext2fs *fs = fp->f_fs;
char *buf;
size_t rsize;
int rc;
daddr_t inode_sector;
struct ext2fs_dinode *dip;
inode_sector = FSBTODB(fs, ino_to_fsba(fs, inumber));
/*
* Read inode and save it.
*/
buf = fp->f_buf;
twiddle();
rc = DEV_STRATEGY(f->f_dev)(f->f_devdata, F_READ,
inode_sector, fs->e2fs_bsize, buf, &rsize);
if (rc)
return rc;
if (rsize != fs->e2fs_bsize)
return EIO;
dip = (struct ext2fs_dinode *)(buf +
EXT2_DINODE_SIZE(fs) * ino_to_fsbo(fs, inumber));
e2fs_iload(dip, &fp->f_di, EXT2_DINODE_SIZE(fs));
/*
* Clear out the old buffers
*/
fp->f_ind_cache_block = ~0;
fp->f_buf_blkno = -1;
return rc;
}
/*
* Update the access, modified, and inode change times as specified by the
* IN_ACCESS, IN_UPDATE, and IN_CHANGE flags respectively. Write the inode
* to disk if the IN_MODIFIED flag is set (it may be set initially, or by
* the timestamp update). The IN_LAZYMOD flag is set to force a write
* later if not now. If we write now, then clear both IN_MODIFIED and
* IN_LAZYMOD to reflect the presumably successful write, and if waitfor is
* set, then wait for the write to complete.
*/
int
ext2_update(struct vnode *vp, int waitfor)
{
struct ext2_sb_info *fs;
struct buf *bp;
struct inode *ip;
int error;
ext2_itimes(vp);
ip = VTOI(vp);
if ((ip->i_flag & IN_MODIFIED) == 0)
return (0);
ip->i_flag &= ~(IN_LAZYMOD | IN_MODIFIED);
if (vp->v_mount->mnt_flag & MNT_RDONLY)
return (0);
fs = ip->i_e2fs;
error = bread(ip->i_devvp,
fsbtodoff(fs, ino_to_fsba(fs, ip->i_number)),
(int)fs->s_blocksize, &bp);
if (error) {
brelse(bp);
return (error);
}
ext2_di2ei( &ip->i_din, (struct ext2_inode *) ((char *)bp->b_data + EXT2_INODE_SIZE(fs) *
ino_to_fsbo(fs, ip->i_number)));
/*
if (waitfor && (vp->v_mount->mnt_flag & MNT_ASYNC) == 0)
return (bwrite(bp));
else {
*/
bdwrite(bp);
return (0);
/*
}
*/
}
/*
* Reload all incore data for a filesystem (used after running fsck on
* the root filesystem and finding things to fix). The filesystem must
* be mounted read-only.
*
* Things to do to update the mount:
* 1) invalidate all cached meta-data.
* 2) re-read superblock from disk.
* 3) re-read summary information from disk.
* 4) invalidate all inactive vnodes.
* 5) invalidate all cached file data.
* 6) re-read inode data for all active vnodes.
*/
int
ext2fs_reload(struct mount *mp, kauth_cred_t cred, struct lwp *l)
{
struct vnode *vp, *mvp, *devvp;
struct inode *ip;
struct buf *bp;
struct m_ext2fs *fs;
struct ext2fs *newfs;
int i, error;
void *cp;
struct ufsmount *ump;
if ((mp->mnt_flag & MNT_RDONLY) == 0)
return (EINVAL);
ump = VFSTOUFS(mp);
/*
* Step 1: invalidate all cached meta-data.
*/
devvp = ump->um_devvp;
vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
error = vinvalbuf(devvp, 0, cred, l, 0, 0);
VOP_UNLOCK(devvp);
if (error)
panic("ext2fs_reload: dirty1");
/*
* Step 2: re-read superblock from disk.
*/
error = bread(devvp, SBLOCK, SBSIZE, NOCRED, 0, &bp);
if (error) {
return (error);
}
newfs = (struct ext2fs *)bp->b_data;
error = ext2fs_checksb(newfs, (mp->mnt_flag & MNT_RDONLY) != 0);
if (error) {
brelse(bp, 0);
return (error);
}
fs = ump->um_e2fs;
/*
* copy in new superblock, and compute in-memory values
*/
e2fs_sbload(newfs, &fs->e2fs);
fs->e2fs_ncg =
howmany(fs->e2fs.e2fs_bcount - fs->e2fs.e2fs_first_dblock,
fs->e2fs.e2fs_bpg);
fs->e2fs_fsbtodb = fs->e2fs.e2fs_log_bsize + LOG_MINBSIZE - DEV_BSHIFT;
fs->e2fs_bsize = MINBSIZE << fs->e2fs.e2fs_log_bsize;
fs->e2fs_bshift = LOG_MINBSIZE + fs->e2fs.e2fs_log_bsize;
fs->e2fs_qbmask = fs->e2fs_bsize - 1;
fs->e2fs_bmask = ~fs->e2fs_qbmask;
fs->e2fs_ngdb =
howmany(fs->e2fs_ncg, fs->e2fs_bsize / sizeof(struct ext2_gd));
fs->e2fs_ipb = fs->e2fs_bsize / EXT2_DINODE_SIZE(fs);
fs->e2fs_itpg = fs->e2fs.e2fs_ipg / fs->e2fs_ipb;
brelse(bp, 0);
/*
* Step 3: re-read summary information from disk.
*/
for (i = 0; i < fs->e2fs_ngdb; i++) {
error = bread(devvp ,
EXT2_FSBTODB(fs, fs->e2fs.e2fs_first_dblock +
1 /* superblock */ + i),
fs->e2fs_bsize, NOCRED, 0, &bp);
if (error) {
return (error);
}
e2fs_cgload((struct ext2_gd *)bp->b_data,
&fs->e2fs_gd[i * fs->e2fs_bsize / sizeof(struct ext2_gd)],
fs->e2fs_bsize);
brelse(bp, 0);
}
/* Allocate a marker vnode. */
mvp = vnalloc(mp);
/*
* NOTE: not using the TAILQ_FOREACH here since in this loop vgone()
* and vclean() can be called indirectly
*/
mutex_enter(&mntvnode_lock);
loop:
for (vp = TAILQ_FIRST(&mp->mnt_vnodelist); vp; vp = vunmark(mvp)) {
vmark(mvp, vp);
if (vp->v_mount != mp || vismarker(vp))
continue;
/*
* Step 4: invalidate all inactive vnodes.
*/
if (vrecycle(vp, &mntvnode_lock)) {
mutex_enter(&mntvnode_lock);
(void)vunmark(mvp);
goto loop;
}
/*
* Step 5: invalidate all cached file data.
*/
mutex_enter(vp->v_interlock);
mutex_exit(&mntvnode_lock);
if (vget(vp, LK_EXCLUSIVE)) {
mutex_enter(&mntvnode_lock);
(void)vunmark(mvp);
goto loop;
}
if (vinvalbuf(vp, 0, cred, l, 0, 0))
panic("ext2fs_reload: dirty2");
/*
* Step 6: re-read inode data for all active vnodes.
*/
ip = VTOI(vp);
error = bread(devvp, EXT2_FSBTODB(fs, ino_to_fsba(fs, ip->i_number)),
(int)fs->e2fs_bsize, NOCRED, 0, &bp);
if (error) {
vput(vp);
mutex_enter(&mntvnode_lock);
(void)vunmark(mvp);
break;
}
cp = (char *)bp->b_data +
(ino_to_fsbo(fs, ip->i_number) * EXT2_DINODE_SIZE(fs));
e2fs_iload((struct ext2fs_dinode *)cp, ip->i_din.e2fs_din);
ext2fs_set_inode_guid(ip);
brelse(bp, 0);
vput(vp);
mutex_enter(&mntvnode_lock);
}
mutex_exit(&mntvnode_lock);
vnfree(mvp);
return (error);
}
/*
* 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);
}
int
loadblocknums(char *boot, int devfd)
{
int i, fd;
struct stat statbuf;
struct statvfs statvfsbuf;
struct fs *fs;
char *buf;
daddr_t blk, *ap;
struct ufs1_dinode *ip;
int ndb;
/*
* Open 2nd-level boot program and record the block numbers
* it occupies on the filesystem represented by `devfd'.
*/
/* Make sure the (probably new) boot file is on disk. */
sync(); sleep(1);
if ((fd = open(boot, O_RDONLY)) < 0)
err(1, "open: %s", boot);
if (fstatvfs(fd, &statvfsbuf) != 0)
err(1, "statfs: %s", boot);
if (strncmp(statvfsbuf.f_fstypename, "ffs",
sizeof(statvfsbuf.f_fstypename)) &&
strncmp(statvfsbuf.f_fstypename, "ufs",
sizeof(statvfsbuf.f_fstypename))) {
errx(1, "%s: must be on an FFS filesystem", boot);
}
if (fsync(fd) != 0)
err(1, "fsync: %s", boot);
if (fstat(fd, &statbuf) != 0)
err(1, "fstat: %s", boot);
close(fd);
/* Read superblock */
devread(devfd, sblock, (daddr_t)(BBSIZE / DEV_BSIZE),
SBLOCKSIZE, "superblock");
fs = (struct fs *)sblock;
/* Sanity-check super-block. */
if (fs->fs_magic != FS_UFS1_MAGIC)
errx(1, "Bad magic number in superblock, must be UFS1");
if (fs->fs_inopb <= 0)
err(1, "Bad inopb=%d in superblock", fs->fs_inopb);
/* Read inode */
if ((buf = malloc(fs->fs_bsize)) == NULL)
errx(1, "No memory for filesystem block");
blk = fsbtodb(fs, ino_to_fsba(fs, statbuf.st_ino));
devread(devfd, buf, blk, fs->fs_bsize, "inode");
ip = (struct ufs1_dinode *)(buf) + ino_to_fsbo(fs, statbuf.st_ino);
/*
* Have the inode. Figure out how many blocks we need.
*/
ndb = howmany(ip->di_size, fs->fs_bsize);
if (ndb > maxblocknum)
errx(1, "Too many blocks");
*block_count_p = ndb;
*block_size_p = fs->fs_bsize;
if (verbose)
printf("Will load %d blocks of size %d each.\n",
ndb, fs->fs_bsize);
/*
* Get the block numbers; we don't handle fragments
*/
ap = ip->di_db;
for (i = 0; i < NDADDR && *ap && ndb; i++, ap++, ndb--) {
blk = fsbtodb(fs, *ap);
if (verbose)
printf("%d: %d\n", i, blk);
block_table[i] = blk;
}
if (ndb == 0)
return 0;
/*
* Just one level of indirections; there isn't much room
* for more in the 1st-level bootblocks anyway.
*/
blk = fsbtodb(fs, ip->di_ib[0]);
devread(devfd, buf, blk, fs->fs_bsize, "indirect block");
/* XXX ondisk32 */
ap = (int32_t *)buf;
for (; i < NINDIR(fs) && *ap && ndb; i++, ap++, ndb--) {
blk = fsbtodb(fs, *ap);
if (verbose)
printf("%d: %d\n", i, blk);
block_table[i] = blk;
}
return 0;
}
/*
* Look up a FFS 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
ffs_vget(struct mount *mp, ino_t ino, struct vnode **vpp)
{
struct fs *fs;
struct inode *ip;
struct ufs1_dinode *dp1;
#ifdef FFS2
struct ufs2_dinode *dp2;
#endif
struct ufsmount *ump;
struct buf *bp;
struct vnode *vp;
dev_t dev;
int error;
ump = VFSTOUFS(mp);
dev = ump->um_dev;
retry:
if ((*vpp = ufs_ihashget(dev, ino)) != NULL)
return (0);
/* Allocate a new vnode/inode. */
if ((error = getnewvnode(VT_UFS, mp, ffs_vnodeop_p, &vp)) != 0) {
*vpp = NULL;
return (error);
}
#ifdef VFSDEBUG
vp->v_flag |= VLOCKSWORK;
#endif
/* XXX - we use the same pool for ffs and mfs */
ip = pool_get(&ffs_ino_pool, PR_WAITOK);
bzero((caddr_t)ip, sizeof(struct inode));
lockinit(&ip->i_lock, PINOD, "inode", 0, 0);
ip->i_ump = ump;
VREF(ip->i_devvp);
vp->v_data = ip;
ip->i_vnode = vp;
ip->i_fs = fs = ump->um_fs;
ip->i_dev = dev;
ip->i_number = ino;
ip->i_vtbl = &ffs_vtbl;
/*
* 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.
*/
error = ufs_ihashins(ip);
if (error) {
/*
* VOP_INACTIVE will treat this as a stale file
* and recycle it quickly
*/
vrele(vp);
if (error == EEXIST)
goto retry;
return (error);
}
/* Read in the disk contents for the inode, copy into the inode. */
error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
(int)fs->fs_bsize, NOCRED, &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);
brelse(bp);
*vpp = NULL;
return (error);
}
#ifdef FFS2
if (ip->i_ump->um_fstype == UM_UFS2) {
ip->i_din2 = pool_get(&ffs_dinode2_pool, PR_WAITOK);
dp2 = (struct ufs2_dinode *) bp->b_data + ino_to_fsbo(fs, ino);
*ip->i_din2 = *dp2;
} else
#endif
{
ip->i_din1 = pool_get(&ffs_dinode1_pool, PR_WAITOK);
dp1 = (struct ufs1_dinode *) bp->b_data + ino_to_fsbo(fs, ino);
*ip->i_din1 = *dp1;
}
brelse(bp);
if (DOINGSOFTDEP(vp))
softdep_load_inodeblock(ip);
else
ip->i_effnlink = DIP(ip, nlink);
/*
* Initialize the vnode from the inode, check for aliases.
* Note that the underlying vnode may have changed.
*/
error = ufs_vinit(mp, ffs_specop_p, FFS_FIFOOPS, &vp);
if (error) {
vput(vp);
*vpp = NULL;
return (error);
}
/*
* Set up a generation number for this inode if it does not
* already have one. This should only happen on old filesystems.
*/
if (DIP(ip, gen) == 0) {
DIP_ASSIGN(ip, gen, arc4random() & INT_MAX);
if (DIP(ip, gen) == 0 || DIP(ip, gen) == -1)
DIP_ASSIGN(ip, gen, 1); /* Shouldn't happen */
if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
ip->i_flag |= IN_MODIFIED;
}
/*
* Ensure that uid and gid are correct. This is a temporary
* fix until fsck has been changed to do the update.
*/
if (fs->fs_magic == FS_UFS1_MAGIC && fs->fs_inodefmt < FS_44INODEFMT) {
ip->i_ffs1_uid = ip->i_din1->di_ouid;
ip->i_ffs1_gid = ip->i_din1->di_ogid;
}
*vpp = vp;
return (0);
}
/*
* Update the access, modified, and inode change times as specified by the
* IN_ACCESS, IN_UPDATE, and IN_CHANGE flags respectively. The IN_MODIFIED
* flag is used to specify that the inode needs to be updated but that the
* times have already been set. The access and modified times are taken from
* the second and third parameters; the inode change time is always taken
* from the current time. If waitfor is set, then wait for the disk write
* of the inode to complete.
*/
int
ffs_update(struct inode *ip, struct timespec *atime,
struct timespec *mtime, int waitfor)
{
struct vnode *vp;
struct fs *fs;
struct buf *bp;
int error;
struct timespec ts;
vp = ITOV(ip);
if (vp->v_mount->mnt_flag & MNT_RDONLY) {
ip->i_flag &=
~(IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE);
return (0);
}
if ((ip->i_flag &
(IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
waitfor != MNT_WAIT)
return (0);
getnanotime(&ts);
if (ip->i_flag & IN_ACCESS) {
DIP_ASSIGN(ip, atime, atime ? atime->tv_sec : ts.tv_sec);
DIP_ASSIGN(ip, atimensec, atime ? atime->tv_nsec : ts.tv_nsec);
}
if (ip->i_flag & IN_UPDATE) {
DIP_ASSIGN(ip, mtime, mtime ? mtime->tv_sec : ts.tv_sec);
DIP_ASSIGN(ip, mtimensec, mtime ? mtime->tv_nsec : ts.tv_nsec);
ip->i_modrev++;
}
if (ip->i_flag & IN_CHANGE) {
DIP_ASSIGN(ip, ctime, ts.tv_sec);
DIP_ASSIGN(ip, ctimensec, ts.tv_nsec);
}
ip->i_flag &= ~(IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE);
fs = ip->i_fs;
/*
* Ensure that uid and gid are correct. This is a temporary
* fix until fsck has been changed to do the update.
*/
if (fs->fs_magic == FS_UFS1_MAGIC && fs->fs_inodefmt < FS_44INODEFMT) {
ip->i_din1->di_ouid = ip->i_ffs1_uid;
ip->i_din1->di_ogid = ip->i_ffs1_gid;
}
error = bread(ip->i_devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
(int)fs->fs_bsize, &bp);
if (error) {
brelse(bp);
return (error);
}
if (DOINGSOFTDEP(vp))
softdep_update_inodeblock(ip, bp, waitfor);
else if (ip->i_effnlink != DIP(ip, nlink))
panic("ffs_update: bad link cnt");
#ifdef FFS2
if (ip->i_ump->um_fstype == UM_UFS2)
*((struct ufs2_dinode *)bp->b_data +
ino_to_fsbo(fs, ip->i_number)) = *ip->i_din2;
else
#endif
*((struct ufs1_dinode *)bp->b_data +
ino_to_fsbo(fs, ip->i_number)) = *ip->i_din1;
if (waitfor && !DOINGASYNC(vp)) {
return (bwrite(bp));
} else {
bdwrite(bp);
return (0);
}
}
/*
* Read information about /boot's inode, then put this and filesystem
* parameters from the superblock into pbr_symbols.
*/
static int
getbootparams(char *boot, int devfd, struct disklabel *dl)
{
int fd;
struct stat statbuf, sb;
struct statfs statfsbuf;
struct partition *pl;
struct fs *fs;
char *buf;
daddr_t blk, *ap;
struct ufs1_dinode *ip;
int ndb;
int mib[3];
size_t size;
dev_t dev;
/*
* Open 2nd-level boot program and record enough details about
* where it is on the filesystem represented by `devfd'
* (inode block, offset within that block, and various filesystem
* parameters essentially taken from the superblock) for biosboot
* to be able to load it later.
*/
/* Make sure the (probably new) boot file is on disk. */
sync(); sleep(1);
if ((fd = open(boot, O_RDONLY)) < 0)
err(1, "open: %s", boot);
if (fstatfs(fd, &statfsbuf) != 0)
err(1, "statfs: %s", boot);
if (strncmp(statfsbuf.f_fstypename, "ffs", MFSNAMELEN) &&
strncmp(statfsbuf.f_fstypename, "ufs", MFSNAMELEN) )
errx(1, "%s: not on an FFS filesystem", boot);
#if 0
if (read(fd, &eh, sizeof(eh)) != sizeof(eh))
errx(1, "read: %s", boot);
if (!IS_ELF(eh)) {
errx(1, "%s: bad magic: 0x%02x%02x%02x%02x",
boot,
eh.e_ident[EI_MAG0], eh.e_ident[EI_MAG1],
eh.e_ident[EI_MAG2], eh.e_ident[EI_MAG3]);
}
#endif
if (fsync(fd) != 0)
err(1, "fsync: %s", boot);
if (fstat(fd, &statbuf) != 0)
err(1, "fstat: %s", boot);
if (fstat(devfd, &sb) != 0)
err(1, "fstat: %s", realdev);
/* Check devices. */
mib[0] = CTL_MACHDEP;
mib[1] = CPU_CHR2BLK;
mib[2] = sb.st_rdev;
size = sizeof(dev);
if (sysctl(mib, 3, &dev, &size, NULL, 0) >= 0) {
if (statbuf.st_dev / MAXPARTITIONS != dev / MAXPARTITIONS)
errx(1, "cross-device install");
}
pl = &dl->d_partitions[DISKPART(statbuf.st_dev)];
close(fd);
/* Read superblock. */
devread(devfd, sblock, pl->p_offset + SBLOCK, SBSIZE, "superblock");
fs = (struct fs *)sblock;
/* Sanity-check super-block. */
if (fs->fs_magic != FS_MAGIC)
errx(1, "Bad magic number in superblock");
if (fs->fs_inopb <= 0)
err(1, "Bad inopb=%d in superblock", fs->fs_inopb);
/* Read inode. */
if ((buf = malloc(fs->fs_bsize)) == NULL)
err(1, NULL);
blk = fsbtodb(fs, ino_to_fsba(fs, statbuf.st_ino));
devread(devfd, buf, pl->p_offset + blk, fs->fs_bsize, "inode");
ip = (struct ufs1_dinode *)(buf) + ino_to_fsbo(fs, statbuf.st_ino);
/*
* Have the inode. Figure out how many filesystem blocks (not disk
* sectors) there are for biosboot to load.
*/
ndb = howmany(ip->di_size, fs->fs_bsize);
if (ndb <= 0)
errx(1, "No blocks to load");
/*
* Now set the values that will need to go into biosboot
* (the partition boot record, a.k.a. the PBR).
*/
sym_set_value(pbr_symbols, "_fs_bsize_p", (fs->fs_bsize / 16));
sym_set_value(pbr_symbols, "_fs_bsize_s", (fs->fs_bsize / 512));
sym_set_value(pbr_symbols, "_fsbtodb", fs->fs_fsbtodb);
sym_set_value(pbr_symbols, "_p_offset", pl->p_offset);
sym_set_value(pbr_symbols, "_inodeblk",
ino_to_fsba(fs, statbuf.st_ino));
ap = ip->di_db;
sym_set_value(pbr_symbols, "_inodedbl",
((((char *)ap) - buf) + INODEOFF));
sym_set_value(pbr_symbols, "_nblocks", ndb);
if (verbose) {
fprintf(stderr, "%s is %d blocks x %d bytes\n",
boot, ndb, fs->fs_bsize);
fprintf(stderr, "fs block shift %u; part offset %u; "
"inode block %u, offset %ld\n",
fs->fs_fsbtodb, pl->p_offset,
ino_to_fsba(fs, statbuf.st_ino),
((((char *)ap) - buf) + INODEOFF));
}
return 0;
}
int
ffs_update(struct vnode *vp, const struct timespec *acc,
const struct timespec *mod, int updflags)
{
struct fs *fs;
struct buf *bp;
struct inode *ip;
int error;
void *cp;
int waitfor, flags;
if (vp->v_mount->mnt_flag & MNT_RDONLY)
return (0);
ip = VTOI(vp);
FFS_ITIMES(ip, acc, mod, NULL);
if (updflags & UPDATE_CLOSE)
flags = ip->i_flag & (IN_MODIFIED | IN_ACCESSED);
else
flags = ip->i_flag & IN_MODIFIED;
if (flags == 0)
return (0);
fs = ip->i_fs;
if ((flags & IN_MODIFIED) != 0 &&
(vp->v_mount->mnt_flag & MNT_ASYNC) == 0) {
waitfor = updflags & UPDATE_WAIT;
if ((updflags & UPDATE_DIROP) != 0)
waitfor |= UPDATE_WAIT;
} else
waitfor = 0;
/*
* Ensure that uid and gid are correct. This is a temporary
* fix until fsck has been changed to do the update.
*/
if (fs->fs_magic == FS_UFS1_MAGIC && /* XXX */
fs->fs_old_inodefmt < FS_44INODEFMT) { /* XXX */
ip->i_ffs1_ouid = ip->i_uid; /* XXX */
ip->i_ffs1_ogid = ip->i_gid; /* XXX */
} /* XXX */
error = bread(ip->i_devvp,
FFS_FSBTODB(fs, ino_to_fsba(fs, ip->i_number)),
(int)fs->fs_bsize, B_MODIFY, &bp);
if (error) {
return (error);
}
ip->i_flag &= ~(IN_MODIFIED | IN_ACCESSED);
/* Keep unlinked inode list up to date */
KDASSERTMSG(DIP(ip, nlink) == ip->i_nlink,
"DIP(ip, nlink) [%d] == ip->i_nlink [%d]",
DIP(ip, nlink), ip->i_nlink);
if (ip->i_mode) {
if (ip->i_nlink > 0) {
UFS_WAPBL_UNREGISTER_INODE(ip->i_ump->um_mountp,
ip->i_number, ip->i_mode);
} else {
UFS_WAPBL_REGISTER_INODE(ip->i_ump->um_mountp,
ip->i_number, ip->i_mode);
}
}
if (fs->fs_magic == FS_UFS1_MAGIC) {
cp = (char *)bp->b_data +
(ino_to_fsbo(fs, ip->i_number) * DINODE1_SIZE);
#ifdef FFS_EI
if (UFS_FSNEEDSWAP(fs))
ffs_dinode1_swap(ip->i_din.ffs1_din,
(struct ufs1_dinode *)cp);
else
#endif
memcpy(cp, ip->i_din.ffs1_din, DINODE1_SIZE);
} else {
cp = (char *)bp->b_data +
(ino_to_fsbo(fs, ip->i_number) * DINODE2_SIZE);
#ifdef FFS_EI
if (UFS_FSNEEDSWAP(fs))
ffs_dinode2_swap(ip->i_din.ffs2_din,
(struct ufs2_dinode *)cp);
else
#endif
memcpy(cp, ip->i_din.ffs2_din, DINODE2_SIZE);
}
if (waitfor) {
return (bwrite(bp));
} else {
bdwrite(bp);
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 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);
}
static int ext3fs_mount(void *dev, void *dir) {
struct fs_driver *drv;
struct ext2fs_dinode *dip = sysmalloc(sizeof(struct ext2fs_dinode));
char buf[SECTOR_SIZE * 2];
struct ext2_fs_info *fsi;
int inode_sector, ret, rsize;
struct node *dev_node = dev;
struct nas *dir_nas = ((struct node *)dir)->nas;
journal_t *jp = NULL;
ext3_journal_specific_t *ext3_spec;
journal_fs_specific_t spec = {
.bmap = ext3_journal_bmap,
.commit = ext3_journal_commit,
.update = ext3_journal_update,
.trans_freespace = ext3_journal_trans_freespace
};
if (NULL == (drv = fs_driver_find_drv(EXT2_NAME))) {
return -1;
}
if ((ret = drv->fsop->mount(dev, dir)) < 0) {
return ret;
}
if (NULL == (ext3_spec = objalloc(&ext3_journal_cache))) {
return -1;
}
spec.data = ext3_spec;
if (NULL == (jp = journal_create(&spec))) {
objfree(&ext3_journal_cache, ext3_spec);
return -1;
}
/* Getting first block for inode number EXT3_JOURNAL_SUPERBLOCK_INODE */
dir_nas = ((struct node *)dir)->nas;
fsi = dir_nas->fs->fsi;
inode_sector = ino_to_fsba(fsi, EXT3_JOURNAL_SUPERBLOCK_INODE);
rsize = ext2_read_sector(dir_nas, buf, 1, inode_sector);
if (rsize * fsi->s_block_size != fsi->s_block_size) {
return -EIO;
}
/* set pointer to inode struct in read buffer */
memcpy(dip, (buf
+ EXT2_DINODE_SIZE(fsi) * ino_to_fsbo(fsi, EXT3_JOURNAL_SUPERBLOCK_INODE)),
sizeof(struct ext2fs_dinode));
/* XXX Hack to use ext2 functions */
dir_nas->fs->drv = &ext3fs_driver;
ext3_spec->ext3_journal_inode = dip;
if (0 > ext3_journal_load(jp, (struct block_dev *) dev_node->nas->fi->privdata,
fsbtodb(fsi, dip->i_block[0]))) {
return -EIO;
}
/*
* FIXME Now journal supports block size only equal to filesystem block size
* It is not critical but not flexible enough
*/
assert(jp->j_blocksize == fsi->s_block_size);
fsi->journal = jp;
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 ext2fs_dinode *dp;
struct ufsmount *ump;
struct buf *bp;
struct vnode *vp;
dev_t dev;
int error;
if (ino > (ufsino_t)-1)
panic("ext2fs_vget: alien ino_t %llu",
(unsigned long long)ino);
ump = VFSTOUFS(mp);
dev = ump->um_dev;
retry:
if ((*vpp = ufs_ihashget(dev, ino)) != NULL)
return (0);
/* Allocate a new vnode/inode. */
if ((error = getnewvnode(VT_EXT2FS, mp, &ext2fs_vops, &vp)) != 0) {
*vpp = NULL;
return (error);
}
ip = pool_get(&ext2fs_inode_pool, PR_WAITOK|PR_ZERO);
lockinit(&ip->i_lock, PINOD, "inode", 0, 0);
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;
/*
* 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.
*/
error = ufs_ihashins(ip);
if (error) {
/*
* Inode has not been inserted into the chain, so make sure
* we don't try to remove it.
*/
ip->i_flag |= IN_UNHASHED;
/*
* VOP_INACTIVE will treat this as a stale file
* and recycle it quickly
*/
vrele(vp);
if (error == EEXIST)
goto retry;
return (error);
}
/* Read in the disk contents for the inode, copy into the inode. */
error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
(int)fs->e2fs_bsize, &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);
brelse(bp);
*vpp = NULL;
return (error);
}
dp = (struct ext2fs_dinode *) ((char *)bp->b_data
+ EXT2_DINODE_SIZE(fs) * ino_to_fsbo(fs, ino));
ip->i_e2din = pool_get(&ext2fs_dinode_pool, PR_WAITOK);
e2fs_iload(dp, ip->i_e2din);
brelse(bp);
ip->i_effnlink = ip->i_e2fs_nlink;
/*
* The fields for storing the UID and GID of an ext2fs inode are
* limited to 16 bits. To overcome this limitation, Linux decided to
* scatter the highest bits of these values into a previously reserved
* area on the disk inode. We deal with this situation by having two
* 32-bit fields *out* of the disk inode to hold the complete values.
* Now that we are reading in the inode, compute these fields.
*/
ip->i_e2fs_uid = ip->i_e2fs_uid_low | (ip->i_e2fs_uid_high << 16);
ip->i_e2fs_gid = ip->i_e2fs_gid_low | (ip->i_e2fs_gid_high << 16);
/* If the inode was deleted, reset all fields */
if (ip->i_e2fs_dtime != 0) {
ip->i_e2fs_mode = ip->i_e2fs_nblock = 0;
(void)ext2fs_setsize(ip, 0);
}
/*
* Initialize the vnode from the inode, check for aliases.
* Note that the underlying vnode may have changed.
*/
error = ext2fs_vinit(mp, &ext2fs_specvops, EXT2FS_FIFOOPS, &vp);
if (error) {
vput(vp);
*vpp = NULL;
return (error);
}
/*
* Finish inode initialization now that aliasing has been resolved.
*/
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;
}
*vpp = vp;
return (0);
}
/*
* Update the access, modified, and inode change times as specified by the
* IN_ACCESS, IN_UPDATE, and IN_CHANGE flags respectively. Write the inode
* to disk if the IN_MODIFIED flag is set (it may be set initially, or by
* the timestamp update). The IN_LAZYMOD flag is set to force a write
* later if not now. The IN_LAZYACCESS is set instead of IN_MODIFIED if the fs
* is currently being suspended (or is suspended) and vnode has been accessed.
* If we write now, then clear IN_MODIFIED, IN_LAZYACCESS and IN_LAZYMOD to
* reflect the presumably successful write, and if waitfor is set, then wait
* for the write to complete.
*/
int
ffs_update (vnode *vp, int waitfor)
{
int error = 0;
print("HARVEY TODO: %s\n", __func__);
#if 0
struct fs *fs;
struct buf *bp;
struct inode *ip;
int flags, error;
ASSERT_VOP_ELOCKED(vp, "ffs_update");
ufs_itimes(vp);
ip = VTOI(vp);
if ((ip->i_flag & IN_MODIFIED) == 0 && waitfor == 0)
return (0);
ip->i_flag &= ~(IN_LAZYACCESS | IN_LAZYMOD | IN_MODIFIED);
fs = ITOFS(ip);
if (fs->fs_ronly && ITOUMP(ip)->um_fsckpid == 0)
return (0);
/*
* If we are updating a snapshot and another process is currently
* writing the buffer containing the inode for this snapshot then
* a deadlock can occur when it tries to check the snapshot to see
* if that block needs to be copied. Thus when updating a snapshot
* we check to see if the buffer is already locked, and if it is
* we drop the snapshot lock until the buffer has been written
* and is available to us. We have to grab a reference to the
* snapshot vnode to prevent it from being removed while we are
* waiting for the buffer.
*/
flags = 0;
if (IS_SNAPSHOT(ip))
flags = GB_LOCK_NOWAIT;
loop:
error = breadn_flags(ITODEVVP(ip),
fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
(int) fs->fs_bsize, 0, 0, 0, NOCRED, flags, &bp);
if (error != 0) {
if (error != EBUSY)
return (error);
KASSERT((IS_SNAPSHOT(ip)), ("EBUSY from non-snapshot"));
/*
* Wait for our inode block to become available.
*
* Hold a reference to the vnode to protect against
* ffs_snapgone(). Since we hold a reference, it can only
* get reclaimed (VI_DOOMED flag) in a forcible downgrade
* or unmount. For an unmount, the entire filesystem will be
* gone, so we cannot attempt to touch anything associated
* with it while the vnode is unlocked; all we can do is
* pause briefly and try again. If when we relock the vnode
* we discover that it has been reclaimed, updating it is no
* longer necessary and we can just return an error.
*/
vref(vp);
VOP_UNLOCK(vp, 0);
pause("ffsupd", 1);
vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
vrele(vp);
if ((vp->v_iflag & VI_DOOMED) != 0)
return (ENOENT);
goto loop;
}
if (DOINGSOFTDEP(vp))
softdep_update_inodeblock(ip, bp, waitfor);
else if (ip->i_effnlink != ip->i_nlink)
panic("ffs_update: bad link cnt");
if (I_IS_UFS1(ip)) {
*((struct ufs1_dinode *)bp->b_data +
ino_to_fsbo(fs, ip->i_number)) = *ip->i_din1;
/* XXX: FIX? The entropy here is desirable, but the harvesting may be expensive */
random_harvest_queue(&(ip->i_din1), sizeof(ip->i_din1), 1, RANDOM_FS_ATIME);
} else {
*((struct ufs2_dinode *)bp->b_data +
ino_to_fsbo(fs, ip->i_number)) = *ip->i_din2;
/* XXX: FIX? The entropy here is desirable, but the harvesting may be expensive */
random_harvest_queue(&(ip->i_din2), sizeof(ip->i_din2), 1, RANDOM_FS_ATIME);
}
if (waitfor)
error = bwrite(bp);
else if (vm_page_count_severe() || buf_dirty_count_severe()) {
bawrite(bp);
error = 0;
} else {
if (bp->b_bufsize == fs->fs_bsize)
bp->b_flags |= B_CLUSTEROK;
bdwrite(bp);
error = 0;
}
#endif // 0
return (error);
}
/*
* Read information about /boot's inode, then put this and filesystem
* parameters from the superblock into pbr_symbols.
*/
static int
getbootparams(char *boot, int devfd, struct disklabel *dl)
{
int fd;
struct stat statbuf, sb;
struct statfs statfsbuf;
struct partition *pp;
struct fs *fs;
char *buf;
u_int blk, *ap;
struct ufs1_dinode *ip1;
struct ufs2_dinode *ip2;
int ndb;
int mib[3];
size_t size;
dev_t dev;
int skew;
/*
* Open 2nd-level boot program and record enough details about
* where it is on the filesystem represented by `devfd'
* (inode block, offset within that block, and various filesystem
* parameters essentially taken from the superblock) for biosboot
* to be able to load it later.
*/
/* Make sure the (probably new) boot file is on disk. */
sync(); sleep(1);
if ((fd = open(boot, O_RDONLY)) < 0)
err(1, "open: %s", boot);
if (fstatfs(fd, &statfsbuf) != 0)
err(1, "statfs: %s", boot);
if (strncmp(statfsbuf.f_fstypename, "ffs", MFSNAMELEN) &&
strncmp(statfsbuf.f_fstypename, "ufs", MFSNAMELEN) )
errx(1, "%s: not on an FFS filesystem", boot);
#if 0
if (read(fd, &eh, sizeof(eh)) != sizeof(eh))
errx(1, "read: %s", boot);
if (!IS_ELF(eh)) {
errx(1, "%s: bad magic: 0x%02x%02x%02x%02x",
boot,
eh.e_ident[EI_MAG0], eh.e_ident[EI_MAG1],
eh.e_ident[EI_MAG2], eh.e_ident[EI_MAG3]);
}
#endif
if (fsync(fd) != 0)
err(1, "fsync: %s", boot);
if (fstat(fd, &statbuf) != 0)
err(1, "fstat: %s", boot);
if (fstat(devfd, &sb) != 0)
err(1, "fstat: %s", realdev);
/* Check devices. */
mib[0] = CTL_MACHDEP;
mib[1] = CPU_CHR2BLK;
mib[2] = sb.st_rdev;
size = sizeof(dev);
if (sysctl(mib, 3, &dev, &size, NULL, 0) >= 0) {
if (statbuf.st_dev / MAXPARTITIONS != dev / MAXPARTITIONS)
errx(1, "cross-device install");
}
pp = &dl->d_partitions[DISKPART(statbuf.st_dev)];
close(fd);
sbread(devfd, DL_SECTOBLK(dl, DL_GETPOFFSET(pp)), &fs);
/* Read inode. */
if ((buf = malloc(fs->fs_bsize)) == NULL)
err(1, NULL);
blk = fsbtodb(fs, ino_to_fsba(fs, statbuf.st_ino));
/*
* Have the inode. Figure out how many filesystem blocks (not disk
* sectors) there are for biosboot to load.
*/
devread(devfd, buf, DL_SECTOBLK(dl, pp->p_offset) + blk,
fs->fs_bsize, "inode");
if (fs->fs_magic == FS_UFS2_MAGIC) {
ip2 = (struct ufs2_dinode *)(buf) +
ino_to_fsbo(fs, statbuf.st_ino);
ndb = howmany(ip2->di_size, fs->fs_bsize);
ap = (u_int *)ip2->di_db;
skew = sizeof(u_int32_t);
} else {
ip1 = (struct ufs1_dinode *)(buf) +
ino_to_fsbo(fs, statbuf.st_ino);
ndb = howmany(ip1->di_size, fs->fs_bsize);
ap = (u_int *)ip1->di_db;
skew = 0;
}
if (ndb <= 0)
errx(1, "No blocks to load");
/*
* Now set the values that will need to go into biosboot
* (the partition boot record, a.k.a. the PBR).
*/
sym_set_value(pbr_symbols, "_fs_bsize_p", (fs->fs_bsize / 16));
sym_set_value(pbr_symbols, "_fs_bsize_s", (fs->fs_bsize /
dl->d_secsize));
/*
* fs_fsbtodb is the shift to convert fs_fsize to DEV_BSIZE. The
* ino_to_fsba() return value is the number of fs_fsize units.
* Calculate the shift to convert fs_fsize into physical sectors,
* which are added to p_offset to get the sector address BIOS
* will use.
*
* N.B.: ASSUMES fs_fsize is a power of 2 of d_secsize.
*/
sym_set_value(pbr_symbols, "_fsbtodb",
ffs(fs->fs_fsize / dl->d_secsize) - 1);
if (pp->p_offseth != 0)
errx(1, "partition offset too high");
sym_set_value(pbr_symbols, "_p_offset", pp->p_offset);
sym_set_value(pbr_symbols, "_inodeblk",
ino_to_fsba(fs, statbuf.st_ino));
sym_set_value(pbr_symbols, "_inodedbl",
((((char *)ap) - buf) + INODEOFF));
sym_set_value(pbr_symbols, "_nblocks", ndb);
sym_set_value(pbr_symbols, "_blkskew", skew);
if (verbose) {
fprintf(stderr, "%s is %d blocks x %d bytes\n",
boot, ndb, fs->fs_bsize);
fprintf(stderr, "fs block shift %u; part offset %llu; "
"inode block %lld, offset %u\n",
ffs(fs->fs_fsize / dl->d_secsize) - 1,
DL_GETPOFFSET(pp), ino_to_fsba(fs, statbuf.st_ino),
(unsigned int)((((char *)ap) - buf) + INODEOFF));
fprintf(stderr, "expecting %d-bit fs blocks (skew %d)\n",
skew ? 64 : 32, skew);
}
return 0;
}
/*
* 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;
void *cp;
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);
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. */
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 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);
}
/*
* Reload all incore data for a filesystem (used after running fsck on
* the root filesystem and finding things to fix). The filesystem must
* be mounted read-only.
*
* Things to do to update the mount:
* 1) invalidate all cached meta-data.
* 2) re-read superblock from disk.
* 3) invalidate all cluster summary information.
* 4) invalidate all inactive vnodes.
* 5) invalidate all cached file data.
* 6) re-read inode data for all active vnodes.
* XXX we are missing some steps, in particular # 3, this has to be reviewed.
*/
static int
ext2_reload(struct mount *mp, struct thread *td)
{
struct vnode *vp, *mvp, *devvp;
struct inode *ip;
struct buf *bp;
struct ext2fs *es;
struct m_ext2fs *fs;
struct csum *sump;
int error, i;
int32_t *lp;
if ((mp->mnt_flag & MNT_RDONLY) == 0)
return (EINVAL);
/*
* Step 1: invalidate all cached meta-data.
*/
devvp = VFSTOEXT2(mp)->um_devvp;
vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
if (vinvalbuf(devvp, 0, 0, 0) != 0)
panic("ext2_reload: dirty1");
VOP_UNLOCK(devvp, 0);
/*
* Step 2: re-read superblock from disk.
* constants have been adjusted for ext2
*/
if ((error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) != 0)
return (error);
es = (struct ext2fs *)bp->b_data;
if (ext2_check_sb_compat(es, devvp->v_rdev, 0) != 0) {
brelse(bp);
return (EIO); /* XXX needs translation */
}
fs = VFSTOEXT2(mp)->um_e2fs;
bcopy(bp->b_data, fs->e2fs, sizeof(struct ext2fs));
if((error = compute_sb_data(devvp, es, fs)) != 0) {
brelse(bp);
return (error);
}
#ifdef UNKLAR
if (fs->fs_sbsize < SBSIZE)
bp->b_flags |= B_INVAL;
#endif
brelse(bp);
/*
* Step 3: invalidate all cluster summary information.
*/
if (fs->e2fs_contigsumsize > 0) {
lp = fs->e2fs_maxcluster;
sump = fs->e2fs_clustersum;
for (i = 0; i < fs->e2fs_gcount; i++, sump++) {
*lp++ = fs->e2fs_contigsumsize;
sump->cs_init = 0;
bzero(sump->cs_sum, fs->e2fs_contigsumsize + 1);
}
}
loop:
MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
/*
* Step 4: invalidate all cached file data.
*/
if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, td)) {
MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
goto loop;
}
if (vinvalbuf(vp, 0, 0, 0))
panic("ext2_reload: dirty2");
/*
* Step 5: re-read inode data for all active vnodes.
*/
ip = VTOI(vp);
error = bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
(int)fs->e2fs_bsize, NOCRED, &bp);
if (error) {
VOP_UNLOCK(vp, 0);
vrele(vp);
MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
return (error);
}
ext2_ei2i((struct ext2fs_dinode *) ((char *)bp->b_data +
EXT2_INODE_SIZE(fs) * ino_to_fsbo(fs, ip->i_number)), ip);
brelse(bp);
VOP_UNLOCK(vp, 0);
vrele(vp);
}
return (0);
}
int
main(int argc, char *argv[])
{
struct fs *sbp;
struct ufs1_dinode *dp1;
struct ufs2_dinode *dp2;
char *ibuf[MAXBSIZE];
long generation, bsize;
off_t offset;
int i, fd, inonum;
char *fs, sblock[SBLOCKSIZE];
void *v = ibuf;
if (argc < 3)
usage();
fs = *++argv;
sbp = NULL;
/* get the superblock. */
if ((fd = open(fs, O_RDWR, 0)) < 0)
err(1, "%s", fs);
for (i = 0; sblock_try[i] != -1; i++) {
if (lseek(fd, (off_t)(sblock_try[i]), SEEK_SET) < 0)
err(1, "%s", fs);
if (read(fd, sblock, sizeof(sblock)) != sizeof(sblock))
errx(1, "%s: can't read superblock", fs);
sbp = (struct fs *)sblock;
if ((sbp->fs_magic == FS_UFS1_MAGIC ||
(sbp->fs_magic == FS_UFS2_MAGIC &&
sbp->fs_sblockloc == sblock_try[i])) &&
sbp->fs_bsize <= MAXBSIZE &&
sbp->fs_bsize >= (int)sizeof(struct fs))
break;
}
if (sblock_try[i] == -1)
errx(2, "cannot find file system superblock");
bsize = sbp->fs_bsize;
/* remaining arguments are inode numbers. */
while (*++argv) {
/* get the inode number. */
if ((inonum = atoi(*argv)) <= 0)
errx(1, "%s is not a valid inode number", *argv);
(void)printf("clearing %d\n", inonum);
/* read in the appropriate block. */
offset = ino_to_fsba(sbp, inonum); /* inode to fs blk */
offset = fsbtodb(sbp, offset); /* fs blk disk blk */
offset *= DEV_BSIZE; /* disk blk to bytes */
/* seek and read the block */
if (lseek(fd, offset, SEEK_SET) < 0)
err(1, "%s", fs);
if (read(fd, ibuf, bsize) != bsize)
err(1, "%s", fs);
if (sbp->fs_magic == FS_UFS2_MAGIC) {
/* get the inode within the block. */
dp2 = &(((struct ufs2_dinode *)v)
[ino_to_fsbo(sbp, inonum)]);
/* clear the inode, and bump the generation count. */
generation = dp2->di_gen + 1;
memset(dp2, 0, sizeof(*dp2));
dp2->di_gen = generation;
} else {
/* get the inode within the block. */
dp1 = &(((struct ufs1_dinode *)v)
[ino_to_fsbo(sbp, inonum)]);
/* clear the inode, and bump the generation count. */
generation = dp1->di_gen + 1;
memset(dp1, 0, sizeof(*dp1));
dp1->di_gen = generation;
}
/* backup and write the block */
if (lseek(fd, (off_t)-bsize, SEEK_CUR) < 0)
err(1, "%s", fs);
if (write(fd, ibuf, bsize) != bsize)
err(1, "%s", fs);
(void)fsync(fd);
}
(void)close(fd);
exit(0);
}
