/*
* discard preallocated blocks
*/
int
ext2_inactive(struct vop_inactive_args *ap)
{
struct vnode *vp = ap->a_vp;
struct inode *ip = VTOI(vp);
struct thread *td = ap->a_td;
int mode, error = 0;
/*
* Ignore inodes related to stale file handles.
*/
if (ip->i_mode == 0)
goto out;
if (ip->i_nlink <= 0) {
error = ext2_truncate(vp, (off_t)0, 0, NOCRED, td);
ip->i_rdev = 0;
mode = ip->i_mode;
ip->i_mode = 0;
ip->i_flag |= IN_CHANGE | IN_UPDATE;
ext2_vfree(vp, ip->i_number, mode);
}
if (ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE))
ext2_update(vp, 0);
out:
/*
* If we are done with the inode, reclaim it
* so that it can be reused immediately.
*/
if (ip->i_mode == 0)
vrecycle(vp);
return (error);
}
/* ARGSUSED */
static int
ext2_fsync(struct vop_fsync_args *ap)
{
/*
* Flush all dirty buffers associated with a vnode.
*/
vop_stdfsync(ap);
return (ext2_update(ap->a_vp, ap->a_waitfor == MNT_WAIT));
}
/*
* Reclaim an inode so that it can be used for other purposes.
*/
int
ext2_reclaim(struct vop_reclaim_args *ap)
{
struct inode *ip;
struct vnode *vp = ap->a_vp;
ip = VTOI(vp);
if (ip->i_flag & IN_LAZYMOD) {
ip->i_flag |= IN_MODIFIED;
ext2_update(vp, 0);
}
vfs_hash_remove(vp);
free(vp->v_data, M_EXT2NODE);
vp->v_data = 0;
vnode_destroy_vobject(vp);
return (0);
}
/*
* link vnode call
*/
static int
ext2_link(struct vop_link_args *ap)
{
struct vnode *vp = ap->a_vp;
struct vnode *tdvp = ap->a_tdvp;
struct componentname *cnp = ap->a_cnp;
struct inode *ip;
int error;
#ifdef DIAGNOSTIC
if ((cnp->cn_flags & HASBUF) == 0)
panic("ext2_link: no name");
#endif
if (tdvp->v_mount != vp->v_mount) {
error = EXDEV;
goto out;
}
ip = VTOI(vp);
if ((nlink_t)ip->i_nlink >= EXT2_LINK_MAX) {
error = EMLINK;
goto out;
}
if (ip->i_flags & (IMMUTABLE | APPEND)) {
error = EPERM;
goto out;
}
ip->i_nlink++;
ip->i_flag |= IN_CHANGE;
error = ext2_update(vp, !DOINGASYNC(vp));
if (!error)
error = ext2_direnter(ip, tdvp, cnp);
if (error) {
ip->i_nlink--;
ip->i_flag |= IN_CHANGE;
}
out:
return (error);
}
/*
* Rename system call.
* rename("foo", "bar");
* is essentially
* unlink("bar");
* link("foo", "bar");
* unlink("foo");
* but ``atomically''. Can't do full commit without saving state in the
* inode on disk which isn't feasible at this time. Best we can do is
* always guarantee the target exists.
*
* Basic algorithm is:
*
* 1) Bump link count on source while we're linking it to the
* target. This also ensure the inode won't be deleted out
* from underneath us while we work (it may be truncated by
* a concurrent `trunc' or `open' for creation).
* 2) Link source to destination. If destination already exists,
* delete it first.
* 3) Unlink source reference to inode if still around. If a
* directory was moved and the parent of the destination
* is different from the source, patch the ".." entry in the
* directory.
*/
static int
ext2_rename(struct vop_rename_args *ap)
{
struct vnode *tvp = ap->a_tvp;
struct vnode *tdvp = ap->a_tdvp;
struct vnode *fvp = ap->a_fvp;
struct vnode *fdvp = ap->a_fdvp;
struct componentname *tcnp = ap->a_tcnp;
struct componentname *fcnp = ap->a_fcnp;
struct inode *ip, *xp, *dp;
struct dirtemplate dirbuf;
int doingdirectory = 0, oldparent = 0, newparent = 0;
int error = 0;
u_char namlen;
#ifdef DIAGNOSTIC
if ((tcnp->cn_flags & HASBUF) == 0 ||
(fcnp->cn_flags & HASBUF) == 0)
panic("ext2_rename: no name");
#endif
/*
* Check for cross-device rename.
*/
if ((fvp->v_mount != tdvp->v_mount) ||
(tvp && (fvp->v_mount != tvp->v_mount))) {
error = EXDEV;
abortit:
if (tdvp == tvp)
vrele(tdvp);
else
vput(tdvp);
if (tvp)
vput(tvp);
vrele(fdvp);
vrele(fvp);
return (error);
}
if (tvp && ((VTOI(tvp)->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
(VTOI(tdvp)->i_flags & APPEND))) {
error = EPERM;
goto abortit;
}
/*
* Renaming a file to itself has no effect. The upper layers should
* not call us in that case. Temporarily just warn if they do.
*/
if (fvp == tvp) {
printf("ext2_rename: fvp == tvp (can't happen)\n");
error = 0;
goto abortit;
}
if ((error = vn_lock(fvp, LK_EXCLUSIVE)) != 0)
goto abortit;
dp = VTOI(fdvp);
ip = VTOI(fvp);
if (ip->i_nlink >= EXT2_LINK_MAX) {
VOP_UNLOCK(fvp, 0);
error = EMLINK;
goto abortit;
}
if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))
|| (dp->i_flags & APPEND)) {
VOP_UNLOCK(fvp, 0);
error = EPERM;
goto abortit;
}
if ((ip->i_mode & IFMT) == IFDIR) {
/*
* Avoid ".", "..", and aliases of "." for obvious reasons.
*/
if ((fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.') ||
dp == ip || (fcnp->cn_flags | tcnp->cn_flags) & ISDOTDOT ||
(ip->i_flag & IN_RENAME)) {
VOP_UNLOCK(fvp, 0);
error = EINVAL;
goto abortit;
}
ip->i_flag |= IN_RENAME;
oldparent = dp->i_number;
doingdirectory++;
}
vrele(fdvp);
/*
* When the target exists, both the directory
* and target vnodes are returned locked.
*/
dp = VTOI(tdvp);
xp = NULL;
if (tvp)
xp = VTOI(tvp);
/*
* 1) Bump link count while we're moving stuff
* around. If we crash somewhere before
* completing our work, the link count
* may be wrong, but correctable.
*/
ip->i_nlink++;
ip->i_flag |= IN_CHANGE;
if ((error = ext2_update(fvp, !DOINGASYNC(fvp))) != 0) {
VOP_UNLOCK(fvp, 0);
goto bad;
}
/*
* If ".." must be changed (ie the directory gets a new
* parent) then the source directory must not be in the
* directory hierarchy above the target, as this would
* orphan everything below the source directory. Also
* the user must have write permission in the source so
* as to be able to change "..". We must repeat the call
* to namei, as the parent directory is unlocked by the
* call to checkpath().
*/
error = VOP_ACCESS(fvp, VWRITE, tcnp->cn_cred, tcnp->cn_thread);
VOP_UNLOCK(fvp, 0);
if (oldparent != dp->i_number)
newparent = dp->i_number;
if (doingdirectory && newparent) {
if (error) /* write access check above */
goto bad;
if (xp != NULL)
vput(tvp);
error = ext2_checkpath(ip, dp, tcnp->cn_cred);
if (error)
goto out;
VREF(tdvp);
error = relookup(tdvp, &tvp, tcnp);
if (error)
goto out;
vrele(tdvp);
dp = VTOI(tdvp);
xp = NULL;
if (tvp)
xp = VTOI(tvp);
}
/*
* 2) If target doesn't exist, link the target
* to the source and unlink the source.
* Otherwise, rewrite the target directory
* entry to reference the source inode and
* expunge the original entry's existence.
*/
if (xp == NULL) {
if (dp->i_devvp != ip->i_devvp)
panic("ext2_rename: EXDEV");
/*
* Account for ".." in new directory.
* When source and destination have the same
* parent we don't fool with the link count.
*/
if (doingdirectory && newparent) {
if ((nlink_t)dp->i_nlink >= EXT2_LINK_MAX) {
error = EMLINK;
goto bad;
}
dp->i_nlink++;
dp->i_flag |= IN_CHANGE;
error = ext2_update(tdvp, !DOINGASYNC(tdvp));
if (error)
goto bad;
}
error = ext2_direnter(ip, tdvp, tcnp);
if (error) {
if (doingdirectory && newparent) {
dp->i_nlink--;
dp->i_flag |= IN_CHANGE;
(void)ext2_update(tdvp, 1);
}
goto bad;
}
vput(tdvp);
} else {
if (xp->i_devvp != dp->i_devvp || xp->i_devvp != ip->i_devvp)
panic("ext2_rename: EXDEV");
/*
* Short circuit rename(foo, foo).
*/
if (xp->i_number == ip->i_number)
panic("ext2_rename: same file");
/*
* If the parent directory is "sticky", then the user must
* own the parent directory, or the destination of the rename,
* otherwise the destination may not be changed (except by
* root). This implements append-only directories.
*/
if ((dp->i_mode & S_ISTXT) && tcnp->cn_cred->cr_uid != 0 &&
tcnp->cn_cred->cr_uid != dp->i_uid &&
xp->i_uid != tcnp->cn_cred->cr_uid) {
error = EPERM;
goto bad;
}
/*
* Target must be empty if a directory and have no links
* to it. Also, ensure source and target are compatible
* (both directories, or both not directories).
*/
if ((xp->i_mode&IFMT) == IFDIR) {
if (! ext2_dirempty(xp, dp->i_number, tcnp->cn_cred) ||
xp->i_nlink > 2) {
error = ENOTEMPTY;
goto bad;
}
if (!doingdirectory) {
error = ENOTDIR;
goto bad;
}
cache_purge(tdvp);
} else if (doingdirectory) {
error = EISDIR;
goto bad;
}
error = ext2_dirrewrite(dp, ip, tcnp);
if (error)
goto bad;
/*
* If the target directory is in the same
* directory as the source directory,
* decrement the link count on the parent
* of the target directory.
*/
if (doingdirectory && !newparent) {
dp->i_nlink--;
dp->i_flag |= IN_CHANGE;
}
vput(tdvp);
/*
* Adjust the link count of the target to
* reflect the dirrewrite above. If this is
* a directory it is empty and there are
* no links to it, so we can squash the inode and
* any space associated with it. We disallowed
* renaming over top of a directory with links to
* it above, as the remaining link would point to
* a directory without "." or ".." entries.
*/
xp->i_nlink--;
if (doingdirectory) {
if (--xp->i_nlink != 0)
panic("ext2_rename: linked directory");
error = ext2_truncate(tvp, (off_t)0, IO_SYNC,
tcnp->cn_cred, tcnp->cn_thread);
}
xp->i_flag |= IN_CHANGE;
vput(tvp);
xp = NULL;
}
/*
* 3) Unlink the source.
*/
fcnp->cn_flags &= ~MODMASK;
fcnp->cn_flags |= LOCKPARENT | LOCKLEAF;
VREF(fdvp);
error = relookup(fdvp, &fvp, fcnp);
if (error == 0)
vrele(fdvp);
if (fvp != NULL) {
xp = VTOI(fvp);
dp = VTOI(fdvp);
} else {
/*
* From name has disappeared.
*/
if (doingdirectory)
panic("ext2_rename: lost dir entry");
vrele(ap->a_fvp);
return (0);
}
/*
* Ensure that the directory entry still exists and has not
* changed while the new name has been entered. If the source is
* a file then the entry may have been unlinked or renamed. In
* either case there is no further work to be done. If the source
* is a directory then it cannot have been rmdir'ed; its link
* count of three would cause a rmdir to fail with ENOTEMPTY.
* The IN_RENAME flag ensures that it cannot be moved by another
* rename.
*/
if (xp != ip) {
if (doingdirectory)
panic("ext2_rename: lost dir entry");
} else {
/*
* If the source is a directory with a
* new parent, the link count of the old
* parent directory must be decremented
* and ".." set to point to the new parent.
*/
if (doingdirectory && newparent) {
dp->i_nlink--;
dp->i_flag |= IN_CHANGE;
error = vn_rdwr(UIO_READ, fvp, (caddr_t)&dirbuf,
sizeof(struct dirtemplate), (off_t)0,
UIO_SYSSPACE, IO_NODELOCKED | IO_NOMACCHECK,
tcnp->cn_cred, NOCRED, NULL, NULL);
if (error == 0) {
/* Like ufs little-endian: */
namlen = dirbuf.dotdot_type;
if (namlen != 2 ||
dirbuf.dotdot_name[0] != '.' ||
dirbuf.dotdot_name[1] != '.') {
ext2_dirbad(xp, (doff_t)12,
"rename: mangled dir");
} else {
dirbuf.dotdot_ino = newparent;
(void) vn_rdwr(UIO_WRITE, fvp,
(caddr_t)&dirbuf,
sizeof(struct dirtemplate),
(off_t)0, UIO_SYSSPACE,
IO_NODELOCKED | IO_SYNC |
IO_NOMACCHECK, tcnp->cn_cred,
NOCRED, NULL, NULL);
cache_purge(fdvp);
}
}
}
error = ext2_dirremove(fdvp, fcnp);
if (!error) {
xp->i_nlink--;
xp->i_flag |= IN_CHANGE;
}
xp->i_flag &= ~IN_RENAME;
}
if (dp)
vput(fdvp);
if (xp)
vput(fvp);
vrele(ap->a_fvp);
return (error);
bad:
if (xp)
vput(ITOV(xp));
vput(ITOV(dp));
out:
if (doingdirectory)
ip->i_flag &= ~IN_RENAME;
if (vn_lock(fvp, LK_EXCLUSIVE) == 0) {
ip->i_nlink--;
ip->i_flag |= IN_CHANGE;
ip->i_flag &= ~IN_RENAME;
vput(fvp);
} else
vrele(fvp);
return (error);
}
/*
* Set attribute vnode op. called from several syscalls
*/
static int
ext2_setattr(struct vop_setattr_args *ap)
{
struct vattr *vap = ap->a_vap;
struct vnode *vp = ap->a_vp;
struct inode *ip = VTOI(vp);
struct ucred *cred = ap->a_cred;
struct thread *td = curthread;
int error;
/*
* Check for unsettable attributes.
*/
if ((vap->va_type != VNON) || (vap->va_nlink != VNOVAL) ||
(vap->va_fsid != VNOVAL) || (vap->va_fileid != VNOVAL) ||
(vap->va_blocksize != VNOVAL) || (vap->va_rdev != VNOVAL) ||
((int)vap->va_bytes != VNOVAL) || (vap->va_gen != VNOVAL)) {
return (EINVAL);
}
if (vap->va_flags != VNOVAL) {
/* Disallow flags not supported by ext2fs. */
if(vap->va_flags & ~(SF_APPEND | SF_IMMUTABLE | UF_NODUMP))
return (EOPNOTSUPP);
if (vp->v_mount->mnt_flag & MNT_RDONLY)
return (EROFS);
/*
* Callers may only modify the file flags on objects they
* have VADMIN rights for.
*/
if ((error = VOP_ACCESS(vp, VADMIN, cred, td)))
return (error);
/*
* Unprivileged processes and privileged processes in
* jail() are not permitted to unset system flags, or
* modify flags if any system flags are set.
* Privileged non-jail processes may not modify system flags
* if securelevel > 0 and any existing system flags are set.
*/
if (!priv_check_cred(cred, PRIV_VFS_SYSFLAGS, 0)) {
if (ip->i_flags & (SF_IMMUTABLE | SF_APPEND)) {
error = securelevel_gt(cred, 0);
if (error)
return (error);
}
} else {
if (ip->i_flags & (SF_IMMUTABLE | SF_APPEND) ||
((vap->va_flags ^ ip->i_flags) & SF_SETTABLE))
return (EPERM);
}
ip->i_flags = vap->va_flags;
ip->i_flag |= IN_CHANGE;
if (ip->i_flags & (IMMUTABLE | APPEND))
return (0);
}
if (ip->i_flags & (IMMUTABLE | APPEND))
return (EPERM);
/*
* Go through the fields and update iff not VNOVAL.
*/
if (vap->va_uid != (uid_t)VNOVAL || vap->va_gid != (gid_t)VNOVAL) {
if (vp->v_mount->mnt_flag & MNT_RDONLY)
return (EROFS);
if ((error = ext2_chown(vp, vap->va_uid, vap->va_gid, cred,
td)) != 0)
return (error);
}
if (vap->va_size != VNOVAL) {
/*
* Disallow write attempts on read-only file systems;
* unless the file is a socket, fifo, or a block or
* character device resident on the file system.
*/
switch (vp->v_type) {
case VDIR:
return (EISDIR);
case VLNK:
case VREG:
if (vp->v_mount->mnt_flag & MNT_RDONLY)
return (EROFS);
break;
default:
break;
}
if ((error = ext2_truncate(vp, vap->va_size, 0, cred, td)) != 0)
return (error);
}
if (vap->va_atime.tv_sec != VNOVAL || vap->va_mtime.tv_sec != VNOVAL) {
if (vp->v_mount->mnt_flag & MNT_RDONLY)
return (EROFS);
/*
* From utimes(2):
* If times is NULL, ... The caller must be the owner of
* the file, have permission to write the file, or be the
* super-user.
* If times is non-NULL, ... The caller must be the owner of
* the file or be the super-user.
*/
if ((error = VOP_ACCESS(vp, VADMIN, cred, td)) &&
((vap->va_vaflags & VA_UTIMES_NULL) == 0 ||
(error = VOP_ACCESS(vp, VWRITE, cred, td))))
return (error);
if (vap->va_atime.tv_sec != VNOVAL)
ip->i_flag |= IN_ACCESS;
if (vap->va_mtime.tv_sec != VNOVAL)
ip->i_flag |= IN_CHANGE | IN_UPDATE;
ext2_itimes(vp);
if (vap->va_atime.tv_sec != VNOVAL) {
ip->i_atime = vap->va_atime.tv_sec;
ip->i_atimensec = vap->va_atime.tv_nsec;
}
if (vap->va_mtime.tv_sec != VNOVAL) {
ip->i_mtime = vap->va_mtime.tv_sec;
ip->i_mtimensec = vap->va_mtime.tv_nsec;
}
ip->i_birthtime = vap->va_birthtime.tv_sec;
ip->i_birthnsec = vap->va_birthtime.tv_nsec;
error = ext2_update(vp, 0);
if (error)
return (error);
}
error = 0;
if (vap->va_mode != (mode_t)VNOVAL) {
if (vp->v_mount->mnt_flag & MNT_RDONLY)
return (EROFS);
error = ext2_chmod(vp, (int)vap->va_mode, cred, td);
}
return (error);
}
/*
* Vnode op for writing.
*/
static int
ext2_write(struct vop_write_args *ap)
{
struct vnode *vp;
struct uio *uio;
struct inode *ip;
struct m_ext2fs *fs;
struct buf *bp;
daddr_t lbn;
off_t osize;
int blkoffset, error, flags, ioflag, resid, size, seqcount, xfersize;
ioflag = ap->a_ioflag;
uio = ap->a_uio;
vp = ap->a_vp;
seqcount = ioflag >> IO_SEQSHIFT;
ip = VTOI(vp);
#ifdef INVARIANTS
if (uio->uio_rw != UIO_WRITE)
panic("%s: mode", "ext2_write");
#endif
switch (vp->v_type) {
case VREG:
if (ioflag & IO_APPEND)
uio->uio_offset = ip->i_size;
if ((ip->i_flags & APPEND) && uio->uio_offset != ip->i_size)
return (EPERM);
/* FALLTHROUGH */
case VLNK:
break;
case VDIR:
/* XXX differs from ffs -- this is called from ext2_mkdir(). */
if ((ioflag & IO_SYNC) == 0)
panic("ext2_write: nonsync dir write");
break;
default:
panic("ext2_write: type %p %d (%jd,%jd)", (void *)vp,
vp->v_type, (intmax_t)uio->uio_offset,
(intmax_t)uio->uio_resid);
}
KASSERT(uio->uio_resid >= 0, ("ext2_write: uio->uio_resid < 0"));
KASSERT(uio->uio_offset >= 0, ("ext2_write: uio->uio_offset < 0"));
fs = ip->i_e2fs;
if ((uoff_t)uio->uio_offset + uio->uio_resid > fs->e2fs_maxfilesize)
return (EFBIG);
/*
* Maybe this should be above the vnode op call, but so long as
* file servers have no limits, I don't think it matters.
*/
if (vn_rlimit_fsize(vp, uio, uio->uio_td))
return (EFBIG);
resid = uio->uio_resid;
osize = ip->i_size;
if (seqcount > BA_SEQMAX)
flags = BA_SEQMAX << BA_SEQSHIFT;
else
flags = seqcount << BA_SEQSHIFT;
if ((ioflag & IO_SYNC) && !DOINGASYNC(vp))
flags |= IO_SYNC;
for (error = 0; uio->uio_resid > 0;) {
lbn = lblkno(fs, uio->uio_offset);
blkoffset = blkoff(fs, uio->uio_offset);
xfersize = fs->e2fs_fsize - blkoffset;
if (uio->uio_resid < xfersize)
xfersize = uio->uio_resid;
if (uio->uio_offset + xfersize > ip->i_size)
vnode_pager_setsize(vp, uio->uio_offset + xfersize);
/*
* We must perform a read-before-write if the transfer size
* does not cover the entire buffer.
*/
if (fs->e2fs_bsize > xfersize)
flags |= BA_CLRBUF;
else
flags &= ~BA_CLRBUF;
error = ext2_balloc(ip, lbn, blkoffset + xfersize,
ap->a_cred, &bp, flags);
if (error != 0)
break;
/*
* If the buffer is not valid and we did not clear garbage
* out above, we have to do so here even though the write
* covers the entire buffer in order to avoid a mmap()/write
* race where another process may see the garbage prior to
* the uiomove() for a write replacing it.
*/
if ((bp->b_flags & B_CACHE) == 0 && fs->e2fs_bsize <= xfersize)
vfs_bio_clrbuf(bp);
if ((ioflag & (IO_SYNC|IO_INVAL)) == (IO_SYNC|IO_INVAL))
bp->b_flags |= B_NOCACHE;
if (uio->uio_offset + xfersize > ip->i_size)
ip->i_size = uio->uio_offset + xfersize;
size = blksize(fs, ip, lbn) - bp->b_resid;
if (size < xfersize)
xfersize = size;
error =
uiomove((char *)bp->b_data + blkoffset, (int)xfersize, uio);
if (ioflag & (IO_VMIO|IO_DIRECT)) {
bp->b_flags |= B_RELBUF;
}
/*
* If IO_SYNC each buffer is written synchronously. Otherwise
* if we have a severe page deficiency write the buffer
* asynchronously. Otherwise try to cluster, and if that
* doesn't do it then either do an async write (if O_DIRECT),
* or a delayed write (if not).
*/
if (ioflag & IO_SYNC) {
(void)bwrite(bp);
} else if (vm_page_count_severe() ||
buf_dirty_count_severe() ||
(ioflag & IO_ASYNC)) {
bp->b_flags |= B_CLUSTEROK;
bawrite(bp);
} else if (xfersize + blkoffset == fs->e2fs_fsize) {
if ((vp->v_mount->mnt_flag & MNT_NOCLUSTERW) == 0) {
bp->b_flags |= B_CLUSTEROK;
cluster_write(vp, bp, ip->i_size, seqcount, 0);
} else {
bawrite(bp);
}
} else if (ioflag & IO_DIRECT) {
bp->b_flags |= B_CLUSTEROK;
bawrite(bp);
} else {
bp->b_flags |= B_CLUSTEROK;
bdwrite(bp);
}
if (error || xfersize == 0)
break;
}
/*
* If we successfully wrote any data, and we are not the superuser
* we clear the setuid and setgid bits as a precaution against
* tampering.
*/
if ((ip->i_mode & (ISUID | ISGID)) && resid > uio->uio_resid &&
ap->a_cred) {
if (priv_check_cred(ap->a_cred, PRIV_VFS_RETAINSUGID, 0))
ip->i_mode &= ~(ISUID | ISGID);
}
if (error) {
if (ioflag & IO_UNIT) {
(void)ext2_truncate(vp, osize,
ioflag & IO_SYNC, ap->a_cred, uio->uio_td);
uio->uio_offset -= resid - uio->uio_resid;
uio->uio_resid = resid;
}
}
if (uio->uio_resid != resid) {
ip->i_flag |= IN_CHANGE | IN_UPDATE;
if (ioflag & IO_SYNC)
error = ext2_update(vp, 1);
}
return (error);
}
/*
* Allocate a new inode.
*/
static int
ext2_makeinode(int mode, struct vnode *dvp, struct vnode **vpp,
struct componentname *cnp)
{
struct inode *ip, *pdir;
struct vnode *tvp;
int error;
pdir = VTOI(dvp);
#ifdef DIAGNOSTIC
if ((cnp->cn_flags & HASBUF) == 0)
panic("ext2_makeinode: no name");
#endif
*vpp = NULL;
if ((mode & IFMT) == 0)
mode |= IFREG;
error = ext2_valloc(dvp, mode, cnp->cn_cred, &tvp);
if (error) {
return (error);
}
ip = VTOI(tvp);
ip->i_gid = pdir->i_gid;
#ifdef SUIDDIR
{
/*
* if we are
* not the owner of the directory,
* and we are hacking owners here, (only do this where told to)
* and we are not giving it TOO root, (would subvert quotas)
* then go ahead and give it to the other user.
* Note that this drops off the execute bits for security.
*/
if ( (dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
(pdir->i_mode & ISUID) &&
(pdir->i_uid != cnp->cn_cred->cr_uid) && pdir->i_uid) {
ip->i_uid = pdir->i_uid;
mode &= ~07111;
} else {
ip->i_uid = cnp->cn_cred->cr_uid;
}
}
#else
ip->i_uid = cnp->cn_cred->cr_uid;
#endif
ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
ip->i_mode = mode;
tvp->v_type = IFTOVT(mode); /* Rest init'd in getnewvnode(). */
ip->i_nlink = 1;
if ((ip->i_mode & ISGID) && !groupmember(ip->i_gid, cnp->cn_cred)) {
if (priv_check_cred(cnp->cn_cred, PRIV_VFS_RETAINSUGID, 0))
ip->i_mode &= ~ISGID;
}
if (cnp->cn_flags & ISWHITEOUT)
ip->i_flags |= UF_OPAQUE;
/*
* Make sure inode goes to disk before directory entry.
*/
error = ext2_update(tvp, !DOINGASYNC(tvp));
if (error)
goto bad;
error = ext2_direnter(ip, dvp, cnp);
if (error)
goto bad;
*vpp = tvp;
return (0);
bad:
/*
* Write error occurred trying to update the inode
* or the directory so must deallocate the inode.
*/
ip->i_nlink = 0;
ip->i_flag |= IN_CHANGE;
vput(tvp);
return (error);
}
/*
* Mkdir system call
*/
static int
ext2_mkdir(struct vop_mkdir_args *ap)
{
struct vnode *dvp = ap->a_dvp;
struct vattr *vap = ap->a_vap;
struct componentname *cnp = ap->a_cnp;
struct inode *ip, *dp;
struct vnode *tvp;
struct dirtemplate dirtemplate, *dtp;
int error, dmode;
#ifdef DIAGNOSTIC
if ((cnp->cn_flags & HASBUF) == 0)
panic("ext2_mkdir: no name");
#endif
dp = VTOI(dvp);
if ((nlink_t)dp->i_nlink >= EXT2_LINK_MAX) {
error = EMLINK;
goto out;
}
dmode = vap->va_mode & 0777;
dmode |= IFDIR;
/*
* Must simulate part of ext2_makeinode here to acquire the inode,
* but not have it entered in the parent directory. The entry is
* made later after writing "." and ".." entries.
*/
error = ext2_valloc(dvp, dmode, cnp->cn_cred, &tvp);
if (error)
goto out;
ip = VTOI(tvp);
ip->i_gid = dp->i_gid;
#ifdef SUIDDIR
{
/*
* if we are hacking owners here, (only do this where told to)
* and we are not giving it TOO root, (would subvert quotas)
* then go ahead and give it to the other user.
* The new directory also inherits the SUID bit.
* If user's UID and dir UID are the same,
* 'give it away' so that the SUID is still forced on.
*/
if ( (dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
(dp->i_mode & ISUID) && dp->i_uid) {
dmode |= ISUID;
ip->i_uid = dp->i_uid;
} else {
ip->i_uid = cnp->cn_cred->cr_uid;
}
}
#else
ip->i_uid = cnp->cn_cred->cr_uid;
#endif
ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
ip->i_mode = dmode;
tvp->v_type = VDIR; /* Rest init'd in getnewvnode(). */
ip->i_nlink = 2;
if (cnp->cn_flags & ISWHITEOUT)
ip->i_flags |= UF_OPAQUE;
error = ext2_update(tvp, 1);
/*
* Bump link count in parent directory
* to reflect work done below. Should
* be done before reference is created
* so reparation is possible if we crash.
*/
dp->i_nlink++;
dp->i_flag |= IN_CHANGE;
error = ext2_update(dvp, !DOINGASYNC(dvp));
if (error)
goto bad;
/* Initialize directory with "." and ".." from static template. */
if (EXT2_HAS_INCOMPAT_FEATURE(ip->i_e2fs,
EXT2F_INCOMPAT_FTYPE))
dtp = &mastertemplate;
else
dtp = &omastertemplate;
dirtemplate = *dtp;
dirtemplate.dot_ino = ip->i_number;
dirtemplate.dotdot_ino = dp->i_number;
/* note that in ext2 DIRBLKSIZ == blocksize, not DEV_BSIZE
* so let's just redefine it - for this function only
*/
#undef DIRBLKSIZ
#define DIRBLKSIZ VTOI(dvp)->i_e2fs->e2fs_bsize
dirtemplate.dotdot_reclen = DIRBLKSIZ - 12;
error = vn_rdwr(UIO_WRITE, tvp, (caddr_t)&dirtemplate,
sizeof(dirtemplate), (off_t)0, UIO_SYSSPACE,
IO_NODELOCKED | IO_SYNC | IO_NOMACCHECK, cnp->cn_cred, NOCRED,
NULL, NULL);
if (error) {
dp->i_nlink--;
dp->i_flag |= IN_CHANGE;
goto bad;
}
if (DIRBLKSIZ > VFSTOEXT2(dvp->v_mount)->um_mountp->mnt_stat.f_bsize)
/* XXX should grow with balloc() */
panic("ext2_mkdir: blksize");
else {
ip->i_size = DIRBLKSIZ;
ip->i_flag |= IN_CHANGE;
}
/* Directory set up, now install its entry in the parent directory. */
error = ext2_direnter(ip, dvp, cnp);
if (error) {
dp->i_nlink--;
dp->i_flag |= IN_CHANGE;
}
bad:
/*
* No need to do an explicit VOP_TRUNCATE here, vrele will do this
* for us because we set the link count to 0.
*/
if (error) {
ip->i_nlink = 0;
ip->i_flag |= IN_CHANGE;
vput(tvp);
} else
*ap->a_vpp = tvp;
out:
return (error);
#undef DIRBLKSIZ
#define DIRBLKSIZ DEV_BSIZE
}
/*
* Truncate the inode oip to at most length size, freeing the
* disk blocks.
*/
int
ext2_truncate(struct vnode *vp, off_t length, int flags, struct ucred *cred,
struct thread *td)
{
struct vnode *ovp = vp;
int32_t lastblock;
struct inode *oip;
int32_t bn, lbn, lastiblock[NIADDR], indir_lbn[NIADDR];
uint32_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR];
struct m_ext2fs *fs;
struct buf *bp;
int offset, size, level;
e4fs_daddr_t count, nblocks, blocksreleased = 0;
int error, i, allerror;
off_t osize;
#ifdef INVARIANTS
struct bufobj *bo;
#endif
oip = VTOI(ovp);
#ifdef INVARIANTS
bo = &ovp->v_bufobj;
#endif
ASSERT_VOP_LOCKED(vp, "ext2_truncate");
if (length < 0)
return (EINVAL);
if (ovp->v_type == VLNK &&
oip->i_size < ovp->v_mount->mnt_maxsymlinklen) {
#ifdef INVARIANTS
if (length != 0)
panic("ext2_truncate: partial truncate of symlink");
#endif
bzero((char *)&oip->i_shortlink, (u_int)oip->i_size);
oip->i_size = 0;
oip->i_flag |= IN_CHANGE | IN_UPDATE;
return (ext2_update(ovp, 1));
}
if (oip->i_size == length) {
oip->i_flag |= IN_CHANGE | IN_UPDATE;
return (ext2_update(ovp, 0));
}
fs = oip->i_e2fs;
osize = oip->i_size;
/*
* Lengthen the size of the file. We must ensure that the
* last byte of the file is allocated. Since the smallest
* value of osize is 0, length will be at least 1.
*/
if (osize < length) {
if (length > oip->i_e2fs->e2fs_maxfilesize)
return (EFBIG);
vnode_pager_setsize(ovp, length);
offset = blkoff(fs, length - 1);
lbn = lblkno(fs, length - 1);
flags |= BA_CLRBUF;
error = ext2_balloc(oip, lbn, offset + 1, cred, &bp, flags);
if (error) {
vnode_pager_setsize(vp, osize);
return (error);
}
oip->i_size = length;
if (bp->b_bufsize == fs->e2fs_bsize)
bp->b_flags |= B_CLUSTEROK;
if (flags & IO_SYNC)
bwrite(bp);
else if (DOINGASYNC(ovp))
bdwrite(bp);
else
bawrite(bp);
oip->i_flag |= IN_CHANGE | IN_UPDATE;
return (ext2_update(ovp, !DOINGASYNC(ovp)));
}
/*
* Shorten the size of the file. If the file is not being
* truncated to a block boundry, the contents of the
* partial block following the end of the file must be
* zero'ed in case it ever become accessible again because
* of subsequent file growth.
*/
/* I don't understand the comment above */
offset = blkoff(fs, length);
if (offset == 0) {
oip->i_size = length;
} else {
lbn = lblkno(fs, length);
flags |= BA_CLRBUF;
error = ext2_balloc(oip, lbn, offset, cred, &bp, flags);
if (error)
return (error);
oip->i_size = length;
size = blksize(fs, oip, lbn);
bzero((char *)bp->b_data + offset, (u_int)(size - offset));
allocbuf(bp, size);
if (bp->b_bufsize == fs->e2fs_bsize)
bp->b_flags |= B_CLUSTEROK;
if (flags & IO_SYNC)
bwrite(bp);
else if (DOINGASYNC(ovp))
bdwrite(bp);
else
bawrite(bp);
}
/*
* Calculate index into inode's block list of
* last direct and indirect blocks (if any)
* which we want to keep. Lastblock is -1 when
* the file is truncated to 0.
*/
lastblock = lblkno(fs, length + fs->e2fs_bsize - 1) - 1;
lastiblock[SINGLE] = lastblock - NDADDR;
lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs);
lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs);
nblocks = btodb(fs->e2fs_bsize);
/*
* Update file and block pointers on disk before we start freeing
* blocks. If we crash before free'ing blocks below, the blocks
* will be returned to the free list. lastiblock values are also
* normalized to -1 for calls to ext2_indirtrunc below.
*/
for (level = TRIPLE; level >= SINGLE; level--) {
oldblks[NDADDR + level] = oip->i_ib[level];
if (lastiblock[level] < 0) {
oip->i_ib[level] = 0;
lastiblock[level] = -1;
}
}
for (i = 0; i < NDADDR; i++) {
oldblks[i] = oip->i_db[i];
if (i > lastblock)
oip->i_db[i] = 0;
}
oip->i_flag |= IN_CHANGE | IN_UPDATE;
allerror = ext2_update(ovp, !DOINGASYNC(ovp));
/*
* Having written the new inode to disk, save its new configuration
* and put back the old block pointers long enough to process them.
* Note that we save the new block configuration so we can check it
* when we are done.
*/
for (i = 0; i < NDADDR; i++) {
newblks[i] = oip->i_db[i];
oip->i_db[i] = oldblks[i];
}
for (i = 0; i < NIADDR; i++) {
newblks[NDADDR + i] = oip->i_ib[i];
oip->i_ib[i] = oldblks[NDADDR + i];
}
oip->i_size = osize;
error = vtruncbuf(ovp, cred, length, (int)fs->e2fs_bsize);
if (error && (allerror == 0))
allerror = error;
vnode_pager_setsize(ovp, length);
/*
* Indirect blocks first.
*/
indir_lbn[SINGLE] = -NDADDR;
indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) - 1;
indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1;
for (level = TRIPLE; level >= SINGLE; level--) {
bn = oip->i_ib[level];
if (bn != 0) {
error = ext2_indirtrunc(oip, indir_lbn[level],
fsbtodb(fs, bn), lastiblock[level], level, &count);
if (error)
allerror = error;
blocksreleased += count;
if (lastiblock[level] < 0) {
oip->i_ib[level] = 0;
ext2_blkfree(oip, bn, fs->e2fs_fsize);
blocksreleased += nblocks;
}
}
if (lastiblock[level] >= 0)
goto done;
}
/*
* All whole direct blocks or frags.
*/
for (i = NDADDR - 1; i > lastblock; i--) {
long bsize;
bn = oip->i_db[i];
if (bn == 0)
continue;
oip->i_db[i] = 0;
bsize = blksize(fs, oip, i);
ext2_blkfree(oip, bn, bsize);
blocksreleased += btodb(bsize);
}
if (lastblock < 0)
goto done;
/*
* Finally, look for a change in size of the
* last direct block; release any frags.
*/
bn = oip->i_db[lastblock];
if (bn != 0) {
long oldspace, newspace;
/*
* Calculate amount of space we're giving
* back as old block size minus new block size.
*/
oldspace = blksize(fs, oip, lastblock);
oip->i_size = length;
newspace = blksize(fs, oip, lastblock);
if (newspace == 0)
panic("ext2_truncate: newspace");
if (oldspace - newspace > 0) {
/*
* Block number of space to be free'd is
* the old block # plus the number of frags
* required for the storage we're keeping.
*/
bn += numfrags(fs, newspace);
ext2_blkfree(oip, bn, oldspace - newspace);
blocksreleased += btodb(oldspace - newspace);
}
}
done:
#ifdef INVARIANTS
for (level = SINGLE; level <= TRIPLE; level++)
if (newblks[NDADDR + level] != oip->i_ib[level])
panic("itrunc1");
for (i = 0; i < NDADDR; i++)
if (newblks[i] != oip->i_db[i])
panic("itrunc2");
BO_LOCK(bo);
if (length == 0 && (bo->bo_dirty.bv_cnt != 0 ||
bo->bo_clean.bv_cnt != 0))
panic("itrunc3");
BO_UNLOCK(bo);
#endif /* INVARIANTS */
/*
* Put back the real size.
*/
oip->i_size = length;
if (oip->i_blocks >= blocksreleased)
oip->i_blocks -= blocksreleased;
else /* sanity */
oip->i_blocks = 0;
oip->i_flag |= IN_CHANGE;
vnode_pager_setsize(ovp, length);
return (allerror);
}
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);
}
