/*
* whiteout vnode call
*/
int
ulfs_whiteout(void *v)
{
struct vop_whiteout_args /* {
struct vnode *a_dvp;
struct componentname *a_cnp;
int a_flags;
} */ *ap = v;
struct vnode *dvp = ap->a_dvp;
struct componentname *cnp = ap->a_cnp;
int error;
struct ulfsmount *ump = VFSTOULFS(dvp->v_mount);
struct lfs *fs = ump->um_lfs;
struct ulfs_lookup_results *ulr;
/* XXX should handle this material another way */
ulr = &VTOI(dvp)->i_crap;
ULFS_CHECK_CRAPCOUNTER(VTOI(dvp));
error = 0;
switch (ap->a_flags) {
case LOOKUP:
/* 4.4 format directories support whiteout operations */
if (fs->um_maxsymlinklen > 0)
return (0);
return (EOPNOTSUPP);
case CREATE:
/* create a new directory whiteout */
fstrans_start(dvp->v_mount, FSTRANS_SHARED);
#ifdef DIAGNOSTIC
if (fs->um_maxsymlinklen <= 0)
panic("ulfs_whiteout: old format filesystem");
#endif
error = ulfs_direnter(dvp, ulr, NULL,
cnp, ULFS_WINO, LFS_DT_WHT, NULL);
break;
case DELETE:
/* remove an existing directory whiteout */
fstrans_start(dvp->v_mount, FSTRANS_SHARED);
#ifdef DIAGNOSTIC
if (fs->um_maxsymlinklen <= 0)
panic("ulfs_whiteout: old format filesystem");
#endif
cnp->cn_flags &= ~DOWHITEOUT;
error = ulfs_dirremove(dvp, ulr, NULL, cnp->cn_flags, 0);
break;
default:
panic("ulfs_whiteout: unknown op");
/* NOTREACHED */
}
fstrans_done(dvp->v_mount);
return (error);
}
/*
* Create a regular file
*/
int
ufs_create(void *v)
{
struct vop_create_v3_args /* {
struct vnode *a_dvp;
struct vnode **a_vpp;
struct componentname *a_cnp;
struct vattr *a_vap;
} */ *ap = v;
int error;
struct vnode *dvp = ap->a_dvp;
struct ufs_lookup_results *ulr;
/* XXX should handle this material another way */
ulr = &VTOI(dvp)->i_crap;
UFS_CHECK_CRAPCOUNTER(VTOI(dvp));
/*
* UFS_WAPBL_BEGIN1(dvp->v_mount, dvp) performed by successful
* ufs_makeinode
*/
fstrans_start(dvp->v_mount, FSTRANS_SHARED);
error =
ufs_makeinode(MAKEIMODE(ap->a_vap->va_type, ap->a_vap->va_mode),
dvp, ulr, ap->a_vpp, ap->a_cnp);
if (error) {
fstrans_done(dvp->v_mount);
return (error);
}
UFS_WAPBL_END1(dvp->v_mount, dvp);
fstrans_done(dvp->v_mount);
VN_KNOTE(dvp, NOTE_WRITE);
VOP_UNLOCK(*ap->a_vpp);
return (0);
}
int
ffs_deleteextattr(void *v)
{
struct vop_deleteextattr_args /* {
struct vnode *a_vp;
int a_attrnamespace;
kauth_cred_t a_cred;
struct proc *a_p;
} */ *ap = v;
struct vnode *vp = ap->a_vp;
struct inode *ip = VTOI(vp);
struct fs *fs = ip->i_fs;
if (fs->fs_magic == FS_UFS1_MAGIC) {
#ifdef UFS_EXTATTR
int error;
fstrans_start(vp->v_mount, FSTRANS_SHARED);
error = ufs_deleteextattr(ap);
fstrans_done(vp->v_mount);
return error;
#else
return (EOPNOTSUPP);
#endif
}
/* XXX Not implemented for UFS2 file systems. */
return (EOPNOTSUPP);
}
/*
* Perform chown operation on inode ip;
* inode must be locked prior to call.
*/
static int
ufs_chown(struct vnode *vp, uid_t uid, gid_t gid, kauth_cred_t cred,
struct lwp *l)
{
struct inode *ip;
int error = 0;
#if defined(QUOTA) || defined(QUOTA2)
uid_t ouid;
gid_t ogid;
int64_t change;
#endif
ip = VTOI(vp);
error = 0;
if (uid == (uid_t)VNOVAL)
uid = ip->i_uid;
if (gid == (gid_t)VNOVAL)
gid = ip->i_gid;
error = kauth_authorize_vnode(cred, KAUTH_VNODE_CHANGE_OWNERSHIP, vp,
NULL, genfs_can_chown(cred, ip->i_uid, ip->i_gid, uid, gid));
if (error)
return (error);
fstrans_start(vp->v_mount, FSTRANS_SHARED);
#if defined(QUOTA) || defined(QUOTA2)
ogid = ip->i_gid;
ouid = ip->i_uid;
change = DIP(ip, blocks);
(void) chkdq(ip, -change, cred, 0);
(void) chkiq(ip, -1, cred, 0);
#endif
ip->i_gid = gid;
DIP_ASSIGN(ip, gid, gid);
ip->i_uid = uid;
DIP_ASSIGN(ip, uid, uid);
#if defined(QUOTA) || defined(QUOTA2)
if ((error = chkdq(ip, change, cred, 0)) == 0) {
if ((error = chkiq(ip, 1, cred, 0)) == 0)
goto good;
else
(void) chkdq(ip, -change, cred, FORCE);
}
ip->i_gid = ogid;
DIP_ASSIGN(ip, gid, ogid);
ip->i_uid = ouid;
DIP_ASSIGN(ip, uid, ouid);
(void) chkdq(ip, change, cred, FORCE);
(void) chkiq(ip, 1, cred, FORCE);
fstrans_done(vp->v_mount);
return (error);
good:
#endif /* QUOTA || QUOTA2 */
ip->i_flag |= IN_CHANGE;
UFS_WAPBL_UPDATE(vp, NULL, NULL, 0);
fstrans_done(vp->v_mount);
return (0);
}
int
msdosfs_sync(struct mount *mp, int waitfor, kauth_cred_t cred)
{
struct vnode *vp;
struct vnode_iterator *marker;
struct denode *dep;
struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
int error, allerror = 0;
/*
* If we ever switch to not updating all of the FATs all the time,
* this would be the place to update them from the first one.
*/
if (pmp->pm_fmod != 0) {
if (pmp->pm_flags & MSDOSFSMNT_RONLY)
panic("msdosfs_sync: rofs mod");
else {
/* update FATs here */
}
}
fstrans_start(mp, FSTRANS_SHARED);
/*
* Write back each (modified) denode.
*/
vfs_vnode_iterator_init(mp, &marker);
while (vfs_vnode_iterator_next(marker, &vp)) {
error = vn_lock(vp, LK_EXCLUSIVE);
if (error) {
vrele(vp);
continue;
}
dep = VTODE(vp);
if (waitfor == MNT_LAZY || vp->v_type == VNON ||
dep == NULL || (((dep->de_flag &
(DE_ACCESS | DE_CREATE | DE_UPDATE | DE_MODIFIED)) == 0) &&
(LIST_EMPTY(&vp->v_dirtyblkhd) &&
UVM_OBJ_IS_CLEAN(&vp->v_uobj)))) {
vput(vp);
continue;
}
if ((error = VOP_FSYNC(vp, cred,
waitfor == MNT_WAIT ? FSYNC_WAIT : 0, 0, 0)) != 0)
allerror = error;
vput(vp);
}
vfs_vnode_iterator_destroy(marker);
/*
* Force stale file system control information to be flushed.
*/
if ((error = VOP_FSYNC(pmp->pm_devvp, cred,
waitfor == MNT_WAIT ? FSYNC_WAIT : 0, 0, 0)) != 0)
allerror = error;
fstrans_done(mp);
return (allerror);
}
/*
* Reclaim an inode so that it can be used for other purposes.
*/
int
ffs_reclaim(void *v)
{
struct vop_reclaim_args /* {
struct vnode *a_vp;
struct lwp *a_l;
} */ *ap = v;
struct vnode *vp = ap->a_vp;
struct inode *ip = VTOI(vp);
struct mount *mp = vp->v_mount;
struct ufsmount *ump = ip->i_ump;
void *data;
int error;
fstrans_start(mp, FSTRANS_LAZY);
/*
* The inode must be freed and updated before being removed
* from its hash chain. Other threads trying to gain a hold
* on the inode will be stalled because it is locked (VI_XLOCK).
*/
error = UFS_WAPBL_BEGIN(mp);
if (error) {
fstrans_done(mp);
return error;
}
if (ip->i_nlink <= 0 && ip->i_omode != 0 &&
(vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
ffs_vfree(vp, ip->i_number, ip->i_omode);
UFS_WAPBL_END(mp);
if ((error = ufs_reclaim(vp)) != 0) {
fstrans_done(mp);
return (error);
}
if (ip->i_din.ffs1_din != NULL) {
if (ump->um_fstype == UFS1)
pool_cache_put(ffs_dinode1_cache, ip->i_din.ffs1_din);
else
pool_cache_put(ffs_dinode2_cache, ip->i_din.ffs2_din);
}
/*
* To interlock with ffs_sync().
*/
genfs_node_destroy(vp);
mutex_enter(vp->v_interlock);
data = vp->v_data;
vp->v_data = NULL;
mutex_exit(vp->v_interlock);
/*
* XXX MFS ends up here, too, to free an inode. Should we create
* XXX a separate pool for MFS inodes?
*/
pool_cache_put(ffs_inode_cache, data);
fstrans_done(mp);
return (0);
}
/*
* Last reference to an inode. If necessary, write or delete it.
*/
int
ulfs_inactive(void *v)
{
struct vop_inactive_args /* {
struct vnode *a_vp;
struct bool *a_recycle;
} */ *ap = v;
struct vnode *vp = ap->a_vp;
struct inode *ip = VTOI(vp);
struct mount *transmp;
mode_t mode;
int error = 0;
transmp = vp->v_mount;
fstrans_start(transmp, FSTRANS_LAZY);
/*
* Ignore inodes related to stale file handles.
*/
if (ip->i_mode == 0)
goto out;
if (ip->i_nlink <= 0 && (vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
#ifdef LFS_EXTATTR
ulfs_extattr_vnode_inactive(vp, curlwp);
#endif
if (ip->i_size != 0) {
error = lfs_truncate(vp, (off_t)0, 0, NOCRED);
}
#if defined(LFS_QUOTA) || defined(LFS_QUOTA2)
(void)lfs_chkiq(ip, -1, NOCRED, 0);
#endif
DIP_ASSIGN(ip, rdev, 0);
mode = ip->i_mode;
ip->i_mode = 0;
ip->i_omode = mode;
DIP_ASSIGN(ip, mode, 0);
ip->i_flag |= IN_CHANGE | IN_UPDATE;
/*
* Defer final inode free and update to ulfs_reclaim().
*/
}
if (ip->i_flag & (IN_CHANGE | IN_UPDATE | IN_MODIFIED)) {
lfs_update(vp, NULL, NULL, 0);
}
out:
/*
* If we are done with the inode, reclaim it
* so that it can be reused immediately.
*/
*ap->a_recycle = (ip->i_mode == 0);
VOP_UNLOCK(vp);
fstrans_done(transmp);
return (error);
}
int
ffs_spec_fsync(void *v)
{
struct vop_fsync_args /* {
struct vnode *a_vp;
kauth_cred_t a_cred;
int a_flags;
off_t a_offlo;
off_t a_offhi;
struct lwp *a_l;
} */ *ap = v;
int error, flags, uflags;
struct vnode *vp;
struct mount *mp;
flags = ap->a_flags;
uflags = UPDATE_CLOSE | ((flags & FSYNC_WAIT) ? UPDATE_WAIT : 0);
vp = ap->a_vp;
mp = vp->v_mount;
fstrans_start(mp, FSTRANS_LAZY);
error = spec_fsync(v);
if (error)
goto out;
#ifdef WAPBL
if (mp && mp->mnt_wapbl) {
/*
* Don't bother writing out metadata if the syncer is
* making the request. We will let the sync vnode
* write it out in a single burst through a call to
* VFS_SYNC().
*/
if ((flags & (FSYNC_DATAONLY | FSYNC_LAZY)) != 0)
goto out;
if ((VTOI(vp)->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE
| IN_MODIFY | IN_MODIFIED | IN_ACCESSED)) != 0) {
error = UFS_WAPBL_BEGIN(mp);
if (error != 0)
goto out;
error = ffs_update(vp, NULL, NULL, uflags);
UFS_WAPBL_END(mp);
}
goto out;
}
#endif /* WAPBL */
error = ffs_update(vp, NULL, NULL, uflags);
out:
fstrans_done(mp);
return error;
}
/*
* symlink -- make a symbolic link
*/
int
ulfs_symlink(void *v)
{
struct vop_symlink_v3_args /* {
struct vnode *a_dvp;
struct vnode **a_vpp;
struct componentname *a_cnp;
struct vattr *a_vap;
char *a_target;
} */ *ap = v;
struct vnode *vp, **vpp;
struct inode *ip;
int len, error;
struct ulfs_lookup_results *ulr;
vpp = ap->a_vpp;
/* XXX should handle this material another way */
ulr = &VTOI(ap->a_dvp)->i_crap;
ULFS_CHECK_CRAPCOUNTER(VTOI(ap->a_dvp));
fstrans_start(ap->a_dvp->v_mount, FSTRANS_SHARED);
error = ulfs_makeinode(LFS_IFLNK | ap->a_vap->va_mode, ap->a_dvp, ulr,
vpp, ap->a_cnp);
if (error)
goto out;
VN_KNOTE(ap->a_dvp, NOTE_WRITE);
vp = *vpp;
len = strlen(ap->a_target);
ip = VTOI(vp);
if (len < ip->i_lfs->um_maxsymlinklen) {
memcpy((char *)SHORTLINK(ip), ap->a_target, len);
ip->i_size = len;
DIP_ASSIGN(ip, size, len);
uvm_vnp_setsize(vp, ip->i_size);
ip->i_flag |= IN_CHANGE | IN_UPDATE;
if (vp->v_mount->mnt_flag & MNT_RELATIME)
ip->i_flag |= IN_ACCESS;
} else
error = vn_rdwr(UIO_WRITE, vp, ap->a_target, len, (off_t)0,
UIO_SYSSPACE, IO_NODELOCKED | IO_JOURNALLOCKED,
ap->a_cnp->cn_cred, NULL, NULL);
VOP_UNLOCK(vp);
if (error)
vrele(vp);
out:
fstrans_done(ap->a_dvp->v_mount);
return (error);
}
static int
ufs_check_possible(struct vnode *vp, struct inode *ip, mode_t mode,
kauth_cred_t cred)
{
#if defined(QUOTA) || defined(QUOTA2)
int error;
#endif
/*
* 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.
*/
if (mode & VWRITE) {
switch (vp->v_type) {
case VDIR:
case VLNK:
case VREG:
if (vp->v_mount->mnt_flag & MNT_RDONLY)
return (EROFS);
#if defined(QUOTA) || defined(QUOTA2)
fstrans_start(vp->v_mount, FSTRANS_SHARED);
error = chkdq(ip, 0, cred, 0);
fstrans_done(vp->v_mount);
if (error != 0)
return error;
#endif
break;
case VBAD:
case VBLK:
case VCHR:
case VSOCK:
case VFIFO:
case VNON:
default:
break;
}
}
/* If it is a snapshot, nobody gets access to it. */
if ((ip->i_flags & SF_SNAPSHOT))
return (EPERM);
/* If immutable bit set, nobody gets to write it. */
if ((mode & VWRITE) && (ip->i_flags & IMMUTABLE))
return (EPERM);
return 0;
}
/* ARGSUSED */
int
ufs_close(void *v)
{
struct vop_close_args /* {
struct vnode *a_vp;
int a_fflag;
kauth_cred_t a_cred;
} */ *ap = v;
struct vnode *vp;
vp = ap->a_vp;
fstrans_start(vp->v_mount, FSTRANS_SHARED);
if (vp->v_usecount > 1)
UFS_ITIMES(vp, NULL, NULL, NULL);
fstrans_done(vp->v_mount);
return (0);
}
int
ufs_remove(void *v)
{
struct vop_remove_args /* {
struct vnode *a_dvp;
struct vnode *a_vp;
struct componentname *a_cnp;
} */ *ap = v;
struct vnode *vp, *dvp;
struct inode *ip;
int error;
struct ufs_lookup_results *ulr;
vp = ap->a_vp;
dvp = ap->a_dvp;
ip = VTOI(vp);
/* XXX should handle this material another way */
ulr = &VTOI(dvp)->i_crap;
UFS_CHECK_CRAPCOUNTER(VTOI(dvp));
fstrans_start(dvp->v_mount, FSTRANS_SHARED);
if (vp->v_type == VDIR || (ip->i_flags & (IMMUTABLE | APPEND)) ||
(VTOI(dvp)->i_flags & APPEND))
error = EPERM;
else {
error = UFS_WAPBL_BEGIN(dvp->v_mount);
if (error == 0) {
error = ufs_dirremove(dvp, ulr,
ip, ap->a_cnp->cn_flags, 0);
UFS_WAPBL_END(dvp->v_mount);
}
}
VN_KNOTE(vp, NOTE_DELETE);
VN_KNOTE(dvp, NOTE_WRITE);
if (dvp == vp)
vrele(vp);
else
vput(vp);
vput(dvp);
fstrans_done(dvp->v_mount);
return (error);
}
/*
* Change the mode on a file.
* Inode must be locked before calling.
*/
static int
ulfs_chmod(struct vnode *vp, int mode, kauth_cred_t cred, struct lwp *l)
{
struct inode *ip;
int error;
ip = VTOI(vp);
error = kauth_authorize_vnode(cred, KAUTH_VNODE_WRITE_SECURITY, vp,
NULL, genfs_can_chmod(vp->v_type, cred, ip->i_uid, ip->i_gid, mode));
if (error)
return (error);
fstrans_start(vp->v_mount, FSTRANS_SHARED);
ip->i_mode &= ~ALLPERMS;
ip->i_mode |= (mode & ALLPERMS);
ip->i_flag |= IN_CHANGE;
DIP_ASSIGN(ip, mode, ip->i_mode);
fstrans_done(vp->v_mount);
return (0);
}
/* ARGSUSED */
int
ufs_mknod(void *v)
{
struct vop_mknod_v3_args /* {
struct vnode *a_dvp;
struct vnode **a_vpp;
struct componentname *a_cnp;
struct vattr *a_vap;
} */ *ap = v;
struct vattr *vap;
struct vnode **vpp;
struct inode *ip;
int error;
struct mount *mp;
ino_t ino;
struct ufs_lookup_results *ulr;
vap = ap->a_vap;
vpp = ap->a_vpp;
/* XXX should handle this material another way */
ulr = &VTOI(ap->a_dvp)->i_crap;
UFS_CHECK_CRAPCOUNTER(VTOI(ap->a_dvp));
/*
* UFS_WAPBL_BEGIN1(dvp->v_mount, dvp) performed by successful
* ufs_makeinode
*/
fstrans_start(ap->a_dvp->v_mount, FSTRANS_SHARED);
if ((error =
ufs_makeinode(MAKEIMODE(vap->va_type, vap->va_mode),
ap->a_dvp, ulr, vpp, ap->a_cnp)) != 0)
goto out;
VN_KNOTE(ap->a_dvp, NOTE_WRITE);
ip = VTOI(*vpp);
mp = (*vpp)->v_mount;
ino = ip->i_number;
ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
if (vap->va_rdev != VNOVAL) {
struct ufsmount *ump = ip->i_ump;
/*
* Want to be able to use this to make badblock
* inodes, so don't truncate the dev number.
*/
if (ump->um_fstype == UFS1)
ip->i_ffs1_rdev = ufs_rw32(vap->va_rdev,
UFS_MPNEEDSWAP(ump));
else
ip->i_ffs2_rdev = ufs_rw64(vap->va_rdev,
UFS_MPNEEDSWAP(ump));
}
UFS_WAPBL_UPDATE(*vpp, NULL, NULL, 0);
UFS_WAPBL_END1(ap->a_dvp->v_mount, ap->a_dvp);
/*
* Remove inode so that it will be reloaded by vcache_get and
* checked to see if it is an alias of an existing entry in
* the inode cache.
*/
(*vpp)->v_type = VNON;
VOP_UNLOCK(*vpp);
vgone(*vpp);
error = vcache_get(mp, &ino, sizeof(ino), vpp);
out:
fstrans_done(ap->a_dvp->v_mount);
if (error != 0) {
*vpp = NULL;
return (error);
}
return (0);
}
/*
* symlink -- make a symbolic link
*/
int
ufs_symlink(void *v)
{
struct vop_symlink_v3_args /* {
struct vnode *a_dvp;
struct vnode **a_vpp;
struct componentname *a_cnp;
struct vattr *a_vap;
char *a_target;
} */ *ap = v;
struct vnode *vp, **vpp;
struct inode *ip;
int len, error;
struct ufs_lookup_results *ulr;
vpp = ap->a_vpp;
/* XXX should handle this material another way */
ulr = &VTOI(ap->a_dvp)->i_crap;
UFS_CHECK_CRAPCOUNTER(VTOI(ap->a_dvp));
/*
* UFS_WAPBL_BEGIN1(dvp->v_mount, dvp) performed by successful
* ufs_makeinode
*/
fstrans_start(ap->a_dvp->v_mount, FSTRANS_SHARED);
error = ufs_makeinode(IFLNK | ap->a_vap->va_mode, ap->a_dvp, ulr,
vpp, ap->a_cnp);
if (error)
goto out;
VN_KNOTE(ap->a_dvp, NOTE_WRITE);
vp = *vpp;
len = strlen(ap->a_target);
ip = VTOI(vp);
/*
* This test is off by one. um_maxsymlinklen contains the
* number of bytes available, and we aren't storing a \0, so
* the test should properly be <=. However, it cannot be
* changed as this would break compatibility with existing fs
* images -- see the way ufs_readlink() works.
*/
if (len < ip->i_ump->um_maxsymlinklen) {
memcpy((char *)SHORTLINK(ip), ap->a_target, len);
ip->i_size = len;
DIP_ASSIGN(ip, size, len);
uvm_vnp_setsize(vp, ip->i_size);
ip->i_flag |= IN_CHANGE | IN_UPDATE;
if (vp->v_mount->mnt_flag & MNT_RELATIME)
ip->i_flag |= IN_ACCESS;
UFS_WAPBL_UPDATE(vp, NULL, NULL, 0);
} else
error = vn_rdwr(UIO_WRITE, vp, ap->a_target, len, (off_t)0,
UIO_SYSSPACE, IO_NODELOCKED | IO_JOURNALLOCKED,
ap->a_cnp->cn_cred, NULL, NULL);
UFS_WAPBL_END1(ap->a_dvp->v_mount, ap->a_dvp);
VOP_UNLOCK(vp);
if (error)
vrele(vp);
out:
fstrans_done(ap->a_dvp->v_mount);
return (error);
}
int
ufs_rmdir(void *v)
{
struct vop_rmdir_args /* {
struct vnode *a_dvp;
struct vnode *a_vp;
struct componentname *a_cnp;
} */ *ap = v;
struct vnode *vp, *dvp;
struct componentname *cnp;
struct inode *ip, *dp;
int error;
struct ufs_lookup_results *ulr;
vp = ap->a_vp;
dvp = ap->a_dvp;
cnp = ap->a_cnp;
ip = VTOI(vp);
dp = VTOI(dvp);
/* XXX should handle this material another way */
ulr = &dp->i_crap;
UFS_CHECK_CRAPCOUNTER(dp);
/*
* No rmdir "." or of mounted directories please.
*/
if (dp == ip || vp->v_mountedhere != NULL) {
if (dp == ip)
vrele(dvp);
else
vput(dvp);
vput(vp);
return (EINVAL);
}
fstrans_start(dvp->v_mount, FSTRANS_SHARED);
/*
* Do not remove a directory that is in the process of being renamed.
* Verify that the directory is empty (and valid). (Rmdir ".." won't
* be valid since ".." will contain a reference to the current
* directory and thus be non-empty.)
*/
error = 0;
if (ip->i_nlink != 2 ||
!ufs_dirempty(ip, dp->i_number, cnp->cn_cred)) {
error = ENOTEMPTY;
goto out;
}
if ((dp->i_flags & APPEND) ||
(ip->i_flags & (IMMUTABLE | APPEND))) {
error = EPERM;
goto out;
}
error = UFS_WAPBL_BEGIN(dvp->v_mount);
if (error)
goto out;
/*
* Delete reference to directory before purging
* inode. If we crash in between, the directory
* will be reattached to lost+found,
*/
error = ufs_dirremove(dvp, ulr, ip, cnp->cn_flags, 1);
if (error) {
UFS_WAPBL_END(dvp->v_mount);
goto out;
}
VN_KNOTE(dvp, NOTE_WRITE | NOTE_LINK);
cache_purge(dvp);
/*
* Truncate inode. The only stuff left in the directory is "." and
* "..". The "." reference is inconsequential since we're quashing
* it.
*/
dp->i_nlink--;
DIP_ASSIGN(dp, nlink, dp->i_nlink);
dp->i_flag |= IN_CHANGE;
UFS_WAPBL_UPDATE(dvp, NULL, NULL, UPDATE_DIROP);
ip->i_nlink--;
DIP_ASSIGN(ip, nlink, ip->i_nlink);
ip->i_flag |= IN_CHANGE;
error = UFS_TRUNCATE(vp, (off_t)0, IO_SYNC, cnp->cn_cred);
cache_purge(vp);
/*
* Unlock the log while we still have reference to unlinked
* directory vp so that it will not get locked for recycling
*/
UFS_WAPBL_END(dvp->v_mount);
#ifdef UFS_DIRHASH
if (ip->i_dirhash != NULL)
ufsdirhash_free(ip);
#endif
out:
VN_KNOTE(vp, NOTE_DELETE);
vput(vp);
fstrans_done(dvp->v_mount);
vput(dvp);
return (error);
}
/*
* Set attribute vnode op. called from several syscalls
*/
int
ufs_setattr(void *v)
{
struct vop_setattr_args /* {
struct vnode *a_vp;
struct vattr *a_vap;
kauth_cred_t a_cred;
} */ *ap = v;
struct vattr *vap;
struct vnode *vp;
struct inode *ip;
kauth_cred_t cred;
struct lwp *l;
int error;
kauth_action_t action;
bool changing_sysflags;
vap = ap->a_vap;
vp = ap->a_vp;
ip = VTOI(vp);
cred = ap->a_cred;
l = curlwp;
action = KAUTH_VNODE_WRITE_FLAGS;
changing_sysflags = false;
/*
* 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);
}
fstrans_start(vp->v_mount, FSTRANS_SHARED);
if (vap->va_flags != VNOVAL) {
if (vp->v_mount->mnt_flag & MNT_RDONLY) {
error = EROFS;
goto out;
}
/* Snapshot flag cannot be set or cleared */
if ((vap->va_flags & (SF_SNAPSHOT | SF_SNAPINVAL)) !=
(ip->i_flags & (SF_SNAPSHOT | SF_SNAPINVAL))) {
error = EPERM;
goto out;
}
if (ip->i_flags & (SF_IMMUTABLE | SF_APPEND)) {
action |= KAUTH_VNODE_HAS_SYSFLAGS;
}
if ((vap->va_flags & SF_SETTABLE) !=
(ip->i_flags & SF_SETTABLE)) {
action |= KAUTH_VNODE_WRITE_SYSFLAGS;
changing_sysflags = true;
}
error = kauth_authorize_vnode(cred, action, vp, NULL,
genfs_can_chflags(cred, vp->v_type, ip->i_uid,
changing_sysflags));
if (error)
goto out;
if (changing_sysflags) {
error = UFS_WAPBL_BEGIN(vp->v_mount);
if (error)
goto out;
ip->i_flags = vap->va_flags;
DIP_ASSIGN(ip, flags, ip->i_flags);
} else {
error = UFS_WAPBL_BEGIN(vp->v_mount);
if (error)
goto out;
ip->i_flags &= SF_SETTABLE;
ip->i_flags |= (vap->va_flags & UF_SETTABLE);
DIP_ASSIGN(ip, flags, ip->i_flags);
}
ip->i_flag |= IN_CHANGE;
UFS_WAPBL_UPDATE(vp, NULL, NULL, 0);
UFS_WAPBL_END(vp->v_mount);
if (vap->va_flags & (IMMUTABLE | APPEND)) {
error = 0;
goto out;
}
}
if (ip->i_flags & (IMMUTABLE | APPEND)) {
error = EPERM;
goto out;
}
/*
* 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) {
error = EROFS;
goto out;
}
error = UFS_WAPBL_BEGIN(vp->v_mount);
if (error)
goto out;
error = ufs_chown(vp, vap->va_uid, vap->va_gid, cred, l);
UFS_WAPBL_END(vp->v_mount);
if (error)
goto out;
}
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:
error = EISDIR;
goto out;
case VCHR:
case VBLK:
case VFIFO:
break;
case VREG:
if (vp->v_mount->mnt_flag & MNT_RDONLY) {
error = EROFS;
goto out;
}
if ((ip->i_flags & SF_SNAPSHOT) != 0) {
error = EPERM;
goto out;
}
error = ufs_truncate(vp, vap->va_size, cred);
if (error)
goto out;
break;
default:
error = EOPNOTSUPP;
goto out;
}
}
ip = VTOI(vp);
if (vap->va_atime.tv_sec != VNOVAL || vap->va_mtime.tv_sec != VNOVAL ||
vap->va_birthtime.tv_sec != VNOVAL) {
if (vp->v_mount->mnt_flag & MNT_RDONLY) {
error = EROFS;
goto out;
}
if ((ip->i_flags & SF_SNAPSHOT) != 0) {
error = EPERM;
goto out;
}
error = kauth_authorize_vnode(cred, KAUTH_VNODE_WRITE_TIMES, vp,
NULL, genfs_can_chtimes(vp, vap->va_vaflags, ip->i_uid, cred));
if (error)
goto out;
error = UFS_WAPBL_BEGIN(vp->v_mount);
if (error)
goto out;
if (vap->va_atime.tv_sec != VNOVAL)
if (!(vp->v_mount->mnt_flag & MNT_NOATIME))
ip->i_flag |= IN_ACCESS;
if (vap->va_mtime.tv_sec != VNOVAL) {
ip->i_flag |= IN_CHANGE | IN_UPDATE;
if (vp->v_mount->mnt_flag & MNT_RELATIME)
ip->i_flag |= IN_ACCESS;
}
if (vap->va_birthtime.tv_sec != VNOVAL &&
ip->i_ump->um_fstype == UFS2) {
ip->i_ffs2_birthtime = vap->va_birthtime.tv_sec;
ip->i_ffs2_birthnsec = vap->va_birthtime.tv_nsec;
}
error = UFS_UPDATE(vp, &vap->va_atime, &vap->va_mtime, 0);
UFS_WAPBL_END(vp->v_mount);
if (error)
goto out;
}
error = 0;
if (vap->va_mode != (mode_t)VNOVAL) {
if (vp->v_mount->mnt_flag & MNT_RDONLY) {
error = EROFS;
goto out;
}
if ((ip->i_flags & SF_SNAPSHOT) != 0 &&
(vap->va_mode & (S_IXUSR | S_IWUSR | S_IXGRP | S_IWGRP |
S_IXOTH | S_IWOTH))) {
error = EPERM;
goto out;
}
error = UFS_WAPBL_BEGIN(vp->v_mount);
if (error)
goto out;
error = ufs_chmod(vp, (int)vap->va_mode, cred, l);
UFS_WAPBL_END(vp->v_mount);
}
VN_KNOTE(vp, NOTE_ATTRIB);
out:
fstrans_done(vp->v_mount);
return (error);
}
int
ffs_fsync(void *v)
{
struct vop_fsync_args /* {
struct vnode *a_vp;
kauth_cred_t a_cred;
int a_flags;
off_t a_offlo;
off_t a_offhi;
struct lwp *a_l;
} */ *ap = v;
struct buf *bp;
int num, error, i;
struct indir ia[NIADDR + 1];
int bsize;
daddr_t blk_high;
struct vnode *vp;
struct mount *mp;
vp = ap->a_vp;
mp = vp->v_mount;
fstrans_start(mp, FSTRANS_LAZY);
if ((ap->a_offlo == 0 && ap->a_offhi == 0) || (vp->v_type != VREG)) {
error = ffs_full_fsync(vp, ap->a_flags);
goto out;
}
bsize = mp->mnt_stat.f_iosize;
blk_high = ap->a_offhi / bsize;
if (ap->a_offhi % bsize != 0)
blk_high++;
/*
* First, flush all pages in range.
*/
mutex_enter(vp->v_interlock);
error = VOP_PUTPAGES(vp, trunc_page(ap->a_offlo),
round_page(ap->a_offhi), PGO_CLEANIT |
((ap->a_flags & FSYNC_WAIT) ? PGO_SYNCIO : 0));
if (error) {
goto out;
}
#ifdef WAPBL
KASSERT(vp->v_type == VREG);
if (mp->mnt_wapbl) {
/*
* Don't bother writing out metadata if the syncer is
* making the request. We will let the sync vnode
* write it out in a single burst through a call to
* VFS_SYNC().
*/
if ((ap->a_flags & (FSYNC_DATAONLY | FSYNC_LAZY)) != 0) {
fstrans_done(mp);
return 0;
}
error = 0;
if (vp->v_tag == VT_UFS && VTOI(vp)->i_flag &
(IN_ACCESS | IN_CHANGE | IN_UPDATE | IN_MODIFY |
IN_MODIFIED | IN_ACCESSED)) {
error = UFS_WAPBL_BEGIN(mp);
if (error) {
fstrans_done(mp);
return error;
}
error = ffs_update(vp, NULL, NULL, UPDATE_CLOSE |
((ap->a_flags & FSYNC_WAIT) ? UPDATE_WAIT : 0));
UFS_WAPBL_END(mp);
}
if (error || (ap->a_flags & FSYNC_NOLOG) != 0) {
fstrans_done(mp);
return error;
}
error = wapbl_flush(mp->mnt_wapbl, 0);
fstrans_done(mp);
return error;
}
#endif /* WAPBL */
/*
* Then, flush indirect blocks.
*/
if (blk_high >= NDADDR) {
error = ufs_getlbns(vp, blk_high, ia, &num);
if (error)
goto out;
mutex_enter(&bufcache_lock);
for (i = 0; i < num; i++) {
if ((bp = incore(vp, ia[i].in_lbn)) == NULL)
continue;
if ((bp->b_cflags & BC_BUSY) != 0 ||
(bp->b_oflags & BO_DELWRI) == 0)
continue;
bp->b_cflags |= BC_BUSY | BC_VFLUSH;
mutex_exit(&bufcache_lock);
bawrite(bp);
mutex_enter(&bufcache_lock);
}
mutex_exit(&bufcache_lock);
}
if (ap->a_flags & FSYNC_WAIT) {
mutex_enter(vp->v_interlock);
while (vp->v_numoutput > 0)
cv_wait(&vp->v_cv, vp->v_interlock);
mutex_exit(vp->v_interlock);
}
error = ffs_update(vp, NULL, NULL, UPDATE_CLOSE |
(((ap->a_flags & (FSYNC_WAIT | FSYNC_DATAONLY)) == FSYNC_WAIT)
? UPDATE_WAIT : 0));
if (error == 0 && ap->a_flags & FSYNC_CACHE) {
int l = 0;
VOP_IOCTL(VTOI(vp)->i_devvp, DIOCCACHESYNC, &l, FWRITE,
curlwp->l_cred);
}
out:
fstrans_done(mp);
return error;
}
/*
* genfs_insane_rename: Generic implementation of the insane API for
* the rename vop.
*
* Arguments:
*
* . fdvp (from directory vnode),
* . fvp (from vnode),
* . fcnp (from component name),
* . tdvp (to directory vnode),
* . tvp (to vnode, or NULL), and
* . tcnp (to component name).
*
* Any pair of vnode parameters may have the same vnode.
*
* On entry,
*
* . fdvp, fvp, tdvp, and tvp are referenced,
* . fdvp and fvp are unlocked, and
* . tdvp and tvp (if nonnull) are locked.
*
* On exit,
*
* . fdvp, fvp, tdvp, and tvp (if nonnull) are unreferenced, and
* . tdvp and tvp (if nonnull) are unlocked.
*/
int
genfs_insane_rename(void *v,
int (*sane_rename)(struct vnode *fdvp, struct componentname *fcnp,
struct vnode *tdvp, struct componentname *tcnp,
kauth_cred_t cred, bool posixly_correct))
{
struct vop_rename_args /* {
struct vnode *a_fdvp;
struct vnode *a_fvp;
struct componentname *a_fcnp;
struct vnode *a_tdvp;
struct vnode *a_tvp;
struct componentname *a_tcnp;
} */ *ap = v;
struct vnode *fdvp = ap->a_fdvp;
struct vnode *fvp = ap->a_fvp;
struct componentname *fcnp = ap->a_fcnp;
struct vnode *tdvp = ap->a_tdvp;
struct vnode *tvp = ap->a_tvp;
struct mount *mp = fdvp->v_mount;
struct componentname *tcnp = ap->a_tcnp;
kauth_cred_t cred;
int error;
KASSERT(fdvp != NULL);
KASSERT(fvp != NULL);
KASSERT(fcnp != NULL);
KASSERT(fcnp->cn_nameptr != NULL);
KASSERT(tdvp != NULL);
KASSERT(tcnp != NULL);
KASSERT(fcnp->cn_nameptr != NULL);
/* KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */
/* KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */
KASSERT(VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE);
KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) == LK_EXCLUSIVE));
KASSERT(fdvp->v_type == VDIR);
KASSERT(tdvp->v_type == VDIR);
fstrans_start(mp, FSTRANS_SHARED);
cred = fcnp->cn_cred;
/*
* XXX Want a better equality test. `tcnp->cn_cred == cred'
* hoses p2k because puffs transmits the creds separately and
* allocates distinct but equivalent structures for them.
*/
KASSERT(kauth_cred_uidmatch(cred, tcnp->cn_cred));
/*
* Sanitize our world from the VFS insanity. Unlock the target
* directory and node, which are locked. Release the children,
* which are referenced, since we'll be looking them up again
* later.
*/
VOP_UNLOCK(tdvp);
if ((tvp != NULL) && (tvp != tdvp))
VOP_UNLOCK(tvp);
vrele(fvp);
if (tvp != NULL)
vrele(tvp);
error = (*sane_rename)(fdvp, fcnp, tdvp, tcnp, cred, false);
/*
* All done, whether with success or failure. Release the
* directory nodes now, as the caller expects from the VFS
* protocol.
*/
vrele(fdvp);
vrele(tdvp);
fstrans_done(mp);
return error;
}
/* ARGSUSED */
int
ufs_getattr(void *v)
{
struct vop_getattr_args /* {
struct vnode *a_vp;
struct vattr *a_vap;
kauth_cred_t a_cred;
} */ *ap = v;
struct vnode *vp;
struct inode *ip;
struct vattr *vap;
vp = ap->a_vp;
ip = VTOI(vp);
vap = ap->a_vap;
fstrans_start(vp->v_mount, FSTRANS_SHARED);
UFS_ITIMES(vp, NULL, NULL, NULL);
/*
* Copy from inode table
*/
vap->va_fsid = ip->i_dev;
vap->va_fileid = ip->i_number;
vap->va_mode = ip->i_mode & ALLPERMS;
vap->va_nlink = ip->i_nlink;
vap->va_uid = ip->i_uid;
vap->va_gid = ip->i_gid;
vap->va_size = vp->v_size;
if (ip->i_ump->um_fstype == UFS1) {
vap->va_rdev = (dev_t)ufs_rw32(ip->i_ffs1_rdev,
UFS_MPNEEDSWAP(ip->i_ump));
vap->va_atime.tv_sec = ip->i_ffs1_atime;
vap->va_atime.tv_nsec = ip->i_ffs1_atimensec;
vap->va_mtime.tv_sec = ip->i_ffs1_mtime;
vap->va_mtime.tv_nsec = ip->i_ffs1_mtimensec;
vap->va_ctime.tv_sec = ip->i_ffs1_ctime;
vap->va_ctime.tv_nsec = ip->i_ffs1_ctimensec;
vap->va_birthtime.tv_sec = 0;
vap->va_birthtime.tv_nsec = 0;
vap->va_bytes = dbtob((u_quad_t)ip->i_ffs1_blocks);
} else {
vap->va_rdev = (dev_t)ufs_rw64(ip->i_ffs2_rdev,
UFS_MPNEEDSWAP(ip->i_ump));
vap->va_atime.tv_sec = ip->i_ffs2_atime;
vap->va_atime.tv_nsec = ip->i_ffs2_atimensec;
vap->va_mtime.tv_sec = ip->i_ffs2_mtime;
vap->va_mtime.tv_nsec = ip->i_ffs2_mtimensec;
vap->va_ctime.tv_sec = ip->i_ffs2_ctime;
vap->va_ctime.tv_nsec = ip->i_ffs2_ctimensec;
vap->va_birthtime.tv_sec = ip->i_ffs2_birthtime;
vap->va_birthtime.tv_nsec = ip->i_ffs2_birthnsec;
vap->va_bytes = dbtob(ip->i_ffs2_blocks);
}
vap->va_gen = ip->i_gen;
vap->va_flags = ip->i_flags;
/* this doesn't belong here */
if (vp->v_type == VBLK)
vap->va_blocksize = BLKDEV_IOSIZE;
else if (vp->v_type == VCHR)
vap->va_blocksize = MAXBSIZE;
else
vap->va_blocksize = vp->v_mount->mnt_stat.f_iosize;
vap->va_type = vp->v_type;
vap->va_filerev = ip->i_modrev;
fstrans_done(vp->v_mount);
return (0);
}
/*
* whiteout vnode call
*/
int
ufs_whiteout(void *v)
{
struct vop_whiteout_args /* {
struct vnode *a_dvp;
struct componentname *a_cnp;
int a_flags;
} */ *ap = v;
struct vnode *dvp = ap->a_dvp;
struct componentname *cnp = ap->a_cnp;
struct direct *newdir;
int error;
struct ufsmount *ump = VFSTOUFS(dvp->v_mount);
struct ufs_lookup_results *ulr;
/* XXX should handle this material another way */
ulr = &VTOI(dvp)->i_crap;
UFS_CHECK_CRAPCOUNTER(VTOI(dvp));
error = 0;
switch (ap->a_flags) {
case LOOKUP:
/* 4.4 format directories support whiteout operations */
if (ump->um_maxsymlinklen > 0)
return (0);
return (EOPNOTSUPP);
case CREATE:
/* create a new directory whiteout */
fstrans_start(dvp->v_mount, FSTRANS_SHARED);
error = UFS_WAPBL_BEGIN(dvp->v_mount);
if (error)
break;
#ifdef DIAGNOSTIC
if (ump->um_maxsymlinklen <= 0)
panic("ufs_whiteout: old format filesystem");
#endif
newdir = pool_cache_get(ufs_direct_cache, PR_WAITOK);
newdir->d_ino = UFS_WINO;
newdir->d_namlen = cnp->cn_namelen;
memcpy(newdir->d_name, cnp->cn_nameptr,
(size_t)cnp->cn_namelen);
newdir->d_name[cnp->cn_namelen] = '\0';
newdir->d_type = DT_WHT;
error = ufs_direnter(dvp, ulr, NULL, newdir, cnp, NULL);
pool_cache_put(ufs_direct_cache, newdir);
break;
case DELETE:
/* remove an existing directory whiteout */
fstrans_start(dvp->v_mount, FSTRANS_SHARED);
error = UFS_WAPBL_BEGIN(dvp->v_mount);
if (error)
break;
#ifdef DIAGNOSTIC
if (ump->um_maxsymlinklen <= 0)
panic("ufs_whiteout: old format filesystem");
#endif
cnp->cn_flags &= ~DOWHITEOUT;
error = ufs_dirremove(dvp, ulr, NULL, cnp->cn_flags, 0);
break;
default:
panic("ufs_whiteout: unknown op");
/* NOTREACHED */
}
UFS_WAPBL_END(dvp->v_mount);
fstrans_done(dvp->v_mount);
return (error);
}
int
msdosfs_sync(struct mount *mp, int waitfor, kauth_cred_t cred)
{
struct vnode *vp, *mvp;
struct denode *dep;
struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
int error, allerror = 0;
/*
* If we ever switch to not updating all of the FATs all the time,
* this would be the place to update them from the first one.
*/
if (pmp->pm_fmod != 0) {
if (pmp->pm_flags & MSDOSFSMNT_RONLY)
panic("msdosfs_sync: rofs mod");
else {
/* update FATs here */
}
}
/* Allocate a marker vnode. */
mvp = vnalloc(mp);
fstrans_start(mp, FSTRANS_SHARED);
/*
* Write back each (modified) denode.
*/
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;
mutex_enter(vp->v_interlock);
dep = VTODE(vp);
if (waitfor == MNT_LAZY || vp->v_type == VNON ||
dep == NULL || (((dep->de_flag &
(DE_ACCESS | DE_CREATE | DE_UPDATE | DE_MODIFIED)) == 0) &&
(LIST_EMPTY(&vp->v_dirtyblkhd) &&
UVM_OBJ_IS_CLEAN(&vp->v_uobj)))) {
mutex_exit(vp->v_interlock);
continue;
}
mutex_exit(&mntvnode_lock);
error = vget(vp, LK_EXCLUSIVE | LK_NOWAIT);
if (error) {
mutex_enter(&mntvnode_lock);
if (error == ENOENT) {
(void)vunmark(mvp);
goto loop;
}
continue;
}
if ((error = VOP_FSYNC(vp, cred,
waitfor == MNT_WAIT ? FSYNC_WAIT : 0, 0, 0)) != 0)
allerror = error;
vput(vp);
mutex_enter(&mntvnode_lock);
}
mutex_exit(&mntvnode_lock);
vnfree(mvp);
/*
* Force stale file system control information to be flushed.
*/
if ((error = VOP_FSYNC(pmp->pm_devvp, cred,
waitfor == MNT_WAIT ? FSYNC_WAIT : 0, 0, 0)) != 0)
allerror = error;
fstrans_done(mp);
return (allerror);
}
/*
* Convert a component of a pathname into a pointer to a locked inode.
* This is a very central and rather complicated routine.
* If the file system is not maintained in a strict tree hierarchy,
* this can result in a deadlock situation (see comments in code below).
*
* The cnp->cn_nameiop argument is LOOKUP, CREATE, RENAME, or DELETE depending
* on whether the name is to be looked up, created, renamed, or deleted.
* When CREATE, RENAME, or DELETE is specified, information usable in
* creating, renaming, or deleting a directory entry may be calculated.
* If flag has LOCKPARENT or'ed into it and the target of the pathname
* exists, lookup returns both the target and its parent directory locked.
* When creating or renaming and LOCKPARENT is specified, the target may
* not be ".". When deleting and LOCKPARENT is specified, the target may
* be "."., but the caller must check to ensure it does an vrele and vput
* instead of two vputs.
*
* Overall outline of ulfs_lookup:
*
* check accessibility of directory
* look for name in cache, if found, then if at end of path
* and deleting or creating, drop it, else return name
* search for name in directory, to found or notfound
* notfound:
* if creating, return locked directory, leaving info on available slots
* else return error
* found:
* if at end of path and deleting, return information to allow delete
* if at end of path and rewriting (RENAME and LOCKPARENT), lock target
* inode and return info to allow rewrite
* if not at end, add name to cache; if at end and neither creating
* nor deleting, add name to cache
*/
int
ulfs_lookup(void *v)
{
struct vop_lookup_v2_args /* {
struct vnode *a_dvp;
struct vnode **a_vpp;
struct componentname *a_cnp;
} */ *ap = v;
struct vnode *vdp = ap->a_dvp; /* vnode for directory being searched */
struct inode *dp = VTOI(vdp); /* inode for directory being searched */
struct buf *bp; /* a buffer of directory entries */
struct lfs_direct *ep; /* the current directory entry */
int entryoffsetinblock; /* offset of ep in bp's buffer */
enum {
NONE, /* need to search a slot for our new entry */
COMPACT, /* a compaction can make a slot in the current
DIRBLKSIZ block */
FOUND, /* found a slot (or no need to search) */
} slotstatus;
doff_t slotoffset; /* offset of area with free space.
a special value -1 for invalid */
int slotsize; /* size of area at slotoffset */
int slotfreespace; /* accumulated amount of space free in
the current DIRBLKSIZ block */
int slotneeded; /* size of the entry we're seeking */
int numdirpasses; /* strategy for directory search */
doff_t endsearch; /* offset to end directory search */
doff_t prevoff; /* previous value of ulr_offset */
struct vnode *pdp; /* saved dp during symlink work */
struct vnode *tdp; /* returned by VFS_VGET */
doff_t enduseful; /* pointer past last used dir slot.
used for directory truncation. */
u_long bmask; /* block offset mask */
int error;
struct vnode **vpp = ap->a_vpp;
struct componentname *cnp = ap->a_cnp;
kauth_cred_t cred = cnp->cn_cred;
int flags;
int nameiop = cnp->cn_nameiop;
struct lfs *fs = dp->i_lfs;
const int needswap = ULFS_MPNEEDSWAP(fs);
int dirblksiz = fs->um_dirblksiz;
ino_t foundino;
struct ulfs_lookup_results *results;
int iswhiteout; /* temp result from cache_lookup() */
flags = cnp->cn_flags;
bp = NULL;
slotoffset = -1;
*vpp = NULL;
endsearch = 0; /* silence compiler warning */
/*
* Produce the auxiliary lookup results into i_crap. Increment
* its serial number so elsewhere we can tell if we're using
* stale results. This should not be done this way. XXX.
*/
results = &dp->i_crap;
dp->i_crapcounter++;
/*
* Check accessiblity of directory.
*/
if ((error = VOP_ACCESS(vdp, VEXEC, cred)) != 0)
return (error);
if ((flags & ISLASTCN) && (vdp->v_mount->mnt_flag & MNT_RDONLY) &&
(nameiop == DELETE || nameiop == RENAME))
return (EROFS);
/*
* We now have a segment name to search for, and a directory to search.
*
* Before tediously performing a linear scan of the directory,
* check the name cache to see if the directory/name pair
* we are looking for is known already.
*/
if (cache_lookup(vdp, cnp->cn_nameptr, cnp->cn_namelen,
cnp->cn_nameiop, cnp->cn_flags, &iswhiteout, vpp)) {
if (iswhiteout) {
cnp->cn_flags |= ISWHITEOUT;
}
return *vpp == NULLVP ? ENOENT : 0;
}
if (iswhiteout) {
/*
* The namecache set iswhiteout without finding a
* cache entry. As of this writing (20121014), this
* can happen if there was a whiteout entry that has
* been invalidated by the lookup. It is not clear if
* it is correct to set ISWHITEOUT in this case or
* not; however, doing so retains the prior behavior,
* so we'll go with that until some clearer answer
* appears. XXX
*/
cnp->cn_flags |= ISWHITEOUT;
}
fstrans_start(vdp->v_mount, FSTRANS_SHARED);
/*
* Suppress search for slots unless creating
* file and at end of pathname, in which case
* we watch for a place to put the new file in
* case it doesn't already exist.
*/
slotstatus = FOUND;
slotfreespace = slotsize = slotneeded = 0;
if ((nameiop == CREATE || nameiop == RENAME) && (flags & ISLASTCN)) {
slotstatus = NONE;
slotneeded = LFS_DIRECTSIZ(cnp->cn_namelen);
}
/*
* If there is cached information on a previous search of
* this directory, pick up where we last left off.
* We cache only lookups as these are the most common
* and have the greatest payoff. Caching CREATE has little
* benefit as it usually must search the entire directory
* to determine that the entry does not exist. Caching the
* location of the last DELETE or RENAME has not reduced
* profiling time and hence has been removed in the interest
* of simplicity.
*/
bmask = vdp->v_mount->mnt_stat.f_iosize - 1;
#ifdef LFS_DIRHASH
/*
* Use dirhash for fast operations on large directories. The logic
* to determine whether to hash the directory is contained within
* ulfsdirhash_build(); a zero return means that it decided to hash
* this directory and it successfully built up the hash table.
*/
if (ulfsdirhash_build(dp) == 0) {
/* Look for a free slot if needed. */
enduseful = dp->i_size;
if (slotstatus != FOUND) {
slotoffset = ulfsdirhash_findfree(dp, slotneeded,
&slotsize);
if (slotoffset >= 0) {
slotstatus = COMPACT;
enduseful = ulfsdirhash_enduseful(dp);
if (enduseful < 0)
enduseful = dp->i_size;
}
}
/* Look up the component. */
numdirpasses = 1;
entryoffsetinblock = 0; /* silence compiler warning */
switch (ulfsdirhash_lookup(dp, cnp->cn_nameptr, cnp->cn_namelen,
&results->ulr_offset, &bp, nameiop == DELETE ? &prevoff : NULL)) {
case 0:
ep = (struct lfs_direct *)((char *)bp->b_data +
(results->ulr_offset & bmask));
goto foundentry;
case ENOENT:
results->ulr_offset = roundup(dp->i_size, dirblksiz);
goto notfound;
default:
/* Something failed; just do a linear search. */
break;
}
}
#endif /* LFS_DIRHASH */
if (nameiop != LOOKUP || results->ulr_diroff == 0 ||
results->ulr_diroff >= dp->i_size) {
entryoffsetinblock = 0;
results->ulr_offset = 0;
numdirpasses = 1;
} else {
results->ulr_offset = results->ulr_diroff;
if ((entryoffsetinblock = results->ulr_offset & bmask) &&
(error = ulfs_blkatoff(vdp, (off_t)results->ulr_offset,
NULL, &bp, false)))
goto out;
numdirpasses = 2;
namecache_count_2passes();
}
prevoff = results->ulr_offset;
endsearch = roundup(dp->i_size, dirblksiz);
enduseful = 0;
searchloop:
while (results->ulr_offset < endsearch) {
if (curcpu()->ci_schedstate.spc_flags & SPCF_SHOULDYIELD)
preempt();
/*
* If necessary, get the next directory block.
*/
if ((results->ulr_offset & bmask) == 0) {
if (bp != NULL)
brelse(bp, 0);
error = ulfs_blkatoff(vdp, (off_t)results->ulr_offset,
NULL, &bp, false);
if (error)
goto out;
entryoffsetinblock = 0;
}
/*
* If still looking for a slot, and at a DIRBLKSIZ
* boundary, have to start looking for free space again.
*/
if (slotstatus == NONE &&
(entryoffsetinblock & (dirblksiz - 1)) == 0) {
slotoffset = -1;
slotfreespace = 0;
}
/*
* Get pointer to next entry.
* Full validation checks are slow, so we only check
* enough to insure forward progress through the
* directory. Complete checks can be run by patching
* "lfs_dirchk" to be true.
*/
KASSERT(bp != NULL);
ep = (struct lfs_direct *)((char *)bp->b_data + entryoffsetinblock);
if (ep->d_reclen == 0 ||
(lfs_dirchk && ulfs_dirbadentry(vdp, ep, entryoffsetinblock))) {
int i;
ulfs_dirbad(dp, results->ulr_offset, "mangled entry");
i = dirblksiz - (entryoffsetinblock & (dirblksiz - 1));
results->ulr_offset += i;
entryoffsetinblock += i;
continue;
}
/*
* If an appropriate sized slot has not yet been found,
* check to see if one is available. Also accumulate space
* in the current block so that we can determine if
* compaction is viable.
*/
if (slotstatus != FOUND) {
int size = ulfs_rw16(ep->d_reclen, needswap);
if (ep->d_ino != 0)
size -= LFS_DIRSIZ(FSFMT(vdp), ep, needswap);
if (size > 0) {
if (size >= slotneeded) {
slotstatus = FOUND;
slotoffset = results->ulr_offset;
slotsize = ulfs_rw16(ep->d_reclen,
needswap);
} else if (slotstatus == NONE) {
slotfreespace += size;
if (slotoffset == -1)
slotoffset = results->ulr_offset;
if (slotfreespace >= slotneeded) {
slotstatus = COMPACT;
slotsize = results->ulr_offset +
ulfs_rw16(ep->d_reclen,
needswap) -
slotoffset;
}
}
}
}
/*
* Check for a name match.
*/
if (ep->d_ino) {
int namlen;
#if (BYTE_ORDER == LITTLE_ENDIAN)
if (FSFMT(vdp) && needswap == 0)
namlen = ep->d_type;
else
namlen = ep->d_namlen;
#else
if (FSFMT(vdp) && needswap != 0)
namlen = ep->d_type;
else
namlen = ep->d_namlen;
#endif
if (namlen == cnp->cn_namelen &&
!memcmp(cnp->cn_nameptr, ep->d_name,
(unsigned)namlen)) {
#ifdef LFS_DIRHASH
foundentry:
#endif
/*
* Save directory entry's inode number and
* reclen, and release directory buffer.
*/
if (!FSFMT(vdp) && ep->d_type == LFS_DT_WHT) {
slotstatus = FOUND;
slotoffset = results->ulr_offset;
slotsize = ulfs_rw16(ep->d_reclen,
needswap);
results->ulr_reclen = slotsize;
/*
* This is used to set
* results->ulr_endoff,
* which may be used by ulfs_direnter()
* as a length to truncate the
* directory to. Therefore, it must
* point past the end of the last
* non-empty directory entry. We don't
* know where that is in this case, so
* we effectively disable shrinking by
* using the existing size of the
* directory.
*
* Note that we wouldn't expect to
* shrink the directory while rewriting
* an existing entry anyway.
*/
enduseful = endsearch;
cnp->cn_flags |= ISWHITEOUT;
numdirpasses--;
goto notfound;
}
foundino = ulfs_rw32(ep->d_ino, needswap);
results->ulr_reclen =
ulfs_rw16(ep->d_reclen, needswap);
goto found;
}
}
prevoff = results->ulr_offset;
results->ulr_offset += ulfs_rw16(ep->d_reclen, needswap);
entryoffsetinblock += ulfs_rw16(ep->d_reclen, needswap);
if (ep->d_ino)
enduseful = results->ulr_offset;
}
notfound:
/*
* If we started in the middle of the directory and failed
* to find our target, we must check the beginning as well.
*/
if (numdirpasses == 2) {
numdirpasses--;
results->ulr_offset = 0;
endsearch = results->ulr_diroff;
goto searchloop;
}
if (bp != NULL)
brelse(bp, 0);
/*
* If creating, and at end of pathname and current
* directory has not been removed, then can consider
* allowing file to be created.
*/
if ((nameiop == CREATE || nameiop == RENAME ||
(nameiop == DELETE &&
(cnp->cn_flags & DOWHITEOUT) &&
(cnp->cn_flags & ISWHITEOUT))) &&
(flags & ISLASTCN) && dp->i_nlink != 0) {
/*
* Access for write is interpreted as allowing
* creation of files in the directory.
*/
error = VOP_ACCESS(vdp, VWRITE, cred);
if (error)
goto out;
/*
* Return an indication of where the new directory
* entry should be put. If we didn't find a slot,
* then set results->ulr_count to 0 indicating
* that the new slot belongs at the end of the
* directory. If we found a slot, then the new entry
* can be put in the range from results->ulr_offset to
* results->ulr_offset + results->ulr_count.
*/
if (slotstatus == NONE) {
results->ulr_offset = roundup(dp->i_size, dirblksiz);
results->ulr_count = 0;
enduseful = results->ulr_offset;
} else if (nameiop == DELETE) {
results->ulr_offset = slotoffset;
if ((results->ulr_offset & (dirblksiz - 1)) == 0)
results->ulr_count = 0;
else
results->ulr_count =
results->ulr_offset - prevoff;
} else {
results->ulr_offset = slotoffset;
results->ulr_count = slotsize;
if (enduseful < slotoffset + slotsize)
enduseful = slotoffset + slotsize;
}
results->ulr_endoff = roundup(enduseful, dirblksiz);
#if 0 /* commented out by dbj. none of the on disk fields changed */
dp->i_flag |= IN_CHANGE | IN_UPDATE;
#endif
/*
* We return with the directory locked, so that
* the parameters we set up above will still be
* valid if we actually decide to do a direnter().
* We return ni_vp == NULL to indicate that the entry
* does not currently exist; we leave a pointer to
* the (locked) directory inode in ndp->ni_dvp.
*
* NB - if the directory is unlocked, then this
* information cannot be used.
*/
error = EJUSTRETURN;
goto out;
}
/*
* Insert name into cache (as non-existent) if appropriate.
*/
if (nameiop != CREATE) {
cache_enter(vdp, *vpp, cnp->cn_nameptr, cnp->cn_namelen,
cnp->cn_flags);
}
error = ENOENT;
goto out;
found:
if (numdirpasses == 2)
namecache_count_pass2();
/*
* Check that directory length properly reflects presence
* of this entry.
*/
if (results->ulr_offset + LFS_DIRSIZ(FSFMT(vdp), ep, needswap) > dp->i_size) {
ulfs_dirbad(dp, results->ulr_offset, "i_size too small");
dp->i_size =
results->ulr_offset + LFS_DIRSIZ(FSFMT(vdp), ep, needswap);
DIP_ASSIGN(dp, size, dp->i_size);
dp->i_flag |= IN_CHANGE | IN_UPDATE;
}
brelse(bp, 0);
/*
* Found component in pathname.
* If the final component of path name, save information
* in the cache as to where the entry was found.
*/
if ((flags & ISLASTCN) && nameiop == LOOKUP)
results->ulr_diroff = results->ulr_offset &~ (dirblksiz - 1);
/*
* If deleting, and at end of pathname, return
* parameters which can be used to remove file.
* Lock the inode, being careful with ".".
*/
if (nameiop == DELETE && (flags & ISLASTCN)) {
/*
* Return pointer to current entry in results->ulr_offset,
* and distance past previous entry (if there
* is a previous entry in this block) in results->ulr_count.
* Save directory inode pointer in ndp->ni_dvp for dirremove().
*/
if ((results->ulr_offset & (dirblksiz - 1)) == 0)
results->ulr_count = 0;
else
results->ulr_count = results->ulr_offset - prevoff;
if (dp->i_number == foundino) {
vref(vdp);
tdp = vdp;
} else {
if (flags & ISDOTDOT)
VOP_UNLOCK(vdp); /* race to get the inode */
error = VFS_VGET(vdp->v_mount, foundino, &tdp);
if (flags & ISDOTDOT)
vn_lock(vdp, LK_EXCLUSIVE | LK_RETRY);
if (error)
goto out;
}
/*
* Write access to directory required to delete files.
*/
error = VOP_ACCESS(vdp, VWRITE, cred);
if (error) {
if (dp->i_number == foundino)
vrele(tdp);
else
vput(tdp);
goto out;
}
/*
* If directory is "sticky", then user must own
* the directory, or the file in it, else she
* may not delete it (unless she's root). This
* implements append-only directories.
*/
if (dp->i_mode & ISVTX) {
error = kauth_authorize_vnode(cred, KAUTH_VNODE_DELETE,
tdp, vdp, genfs_can_sticky(cred, dp->i_uid,
VTOI(tdp)->i_uid));
if (error) {
if (dp->i_number == foundino)
vrele(tdp);
else
vput(tdp);
error = EPERM;
goto out;
}
}
*vpp = tdp;
error = 0;
goto out;
}
/*
* If rewriting (RENAME), return the inode and the
* information required to rewrite the present directory
* Must get inode of directory entry to verify it's a
* regular file, or empty directory.
*/
if (nameiop == RENAME && (flags & ISLASTCN)) {
error = VOP_ACCESS(vdp, VWRITE, cred);
if (error)
goto out;
/*
* Careful about locking second inode.
* This can only occur if the target is ".".
*/
if (dp->i_number == foundino) {
error = EISDIR;
goto out;
}
if (flags & ISDOTDOT)
VOP_UNLOCK(vdp); /* race to get the inode */
error = VFS_VGET(vdp->v_mount, foundino, &tdp);
if (flags & ISDOTDOT)
vn_lock(vdp, LK_EXCLUSIVE | LK_RETRY);
if (error)
goto out;
*vpp = tdp;
error = 0;
goto out;
}
/*
* Step through the translation in the name. We do not `vput' the
* directory because we may need it again if a symbolic link
* is relative to the current directory. Instead we save it
* unlocked as "pdp". We must get the target inode before unlocking
* the directory to insure that the inode will not be removed
* before we get it. We prevent deadlock by always fetching
* inodes from the root, moving down the directory tree. Thus
* when following backward pointers ".." we must unlock the
* parent directory before getting the requested directory.
* There is a potential race condition here if both the current
* and parent directories are removed before the VFS_VGET for the
* inode associated with ".." returns. We hope that this occurs
* infrequently since we cannot avoid this race condition without
* implementing a sophisticated deadlock detection algorithm.
* Note also that this simple deadlock detection scheme will not
* work if the file system has any hard links other than ".."
* that point backwards in the directory structure.
*/
pdp = vdp;
if (flags & ISDOTDOT) {
VOP_UNLOCK(pdp); /* race to get the inode */
error = VFS_VGET(vdp->v_mount, foundino, &tdp);
vn_lock(pdp, LK_EXCLUSIVE | LK_RETRY);
if (error) {
goto out;
}
*vpp = tdp;
} else if (dp->i_number == foundino) {
vref(vdp); /* we want ourself, ie "." */
*vpp = vdp;
} else {
error = VFS_VGET(vdp->v_mount, foundino, &tdp);
if (error)
goto out;
*vpp = tdp;
}
/*
* Insert name into cache if appropriate.
*/
cache_enter(vdp, *vpp, cnp->cn_nameptr, cnp->cn_namelen, cnp->cn_flags);
error = 0;
out:
if (error == 0 && *vpp != vdp)
VOP_UNLOCK(*vpp);
fstrans_done(vdp->v_mount);
return error;
}
/*
* ufs_link: create hard link.
*/
int
ufs_link(void *v)
{
struct vop_link_args /* {
struct vnode *a_dvp;
struct vnode *a_vp;
struct componentname *a_cnp;
} */ *ap = v;
struct vnode *dvp = ap->a_dvp;
struct vnode *vp = ap->a_vp;
struct componentname *cnp = ap->a_cnp;
struct inode *ip;
struct direct *newdir;
int error;
struct ufs_lookup_results *ulr;
KASSERT(dvp != vp);
KASSERT(vp->v_type != VDIR);
KASSERT(dvp->v_mount == vp->v_mount);
/* XXX should handle this material another way */
ulr = &VTOI(dvp)->i_crap;
UFS_CHECK_CRAPCOUNTER(VTOI(dvp));
fstrans_start(dvp->v_mount, FSTRANS_SHARED);
error = vn_lock(vp, LK_EXCLUSIVE);
if (error) {
VOP_ABORTOP(dvp, cnp);
goto out2;
}
ip = VTOI(vp);
if ((nlink_t)ip->i_nlink >= LINK_MAX) {
VOP_ABORTOP(dvp, cnp);
error = EMLINK;
goto out1;
}
if (ip->i_flags & (IMMUTABLE | APPEND)) {
VOP_ABORTOP(dvp, cnp);
error = EPERM;
goto out1;
}
error = UFS_WAPBL_BEGIN(vp->v_mount);
if (error) {
VOP_ABORTOP(dvp, cnp);
goto out1;
}
ip->i_nlink++;
DIP_ASSIGN(ip, nlink, ip->i_nlink);
ip->i_flag |= IN_CHANGE;
error = UFS_UPDATE(vp, NULL, NULL, UPDATE_DIROP);
if (!error) {
newdir = pool_cache_get(ufs_direct_cache, PR_WAITOK);
ufs_makedirentry(ip, cnp, newdir);
error = ufs_direnter(dvp, ulr, vp, newdir, cnp, NULL);
pool_cache_put(ufs_direct_cache, newdir);
}
if (error) {
ip->i_nlink--;
DIP_ASSIGN(ip, nlink, ip->i_nlink);
ip->i_flag |= IN_CHANGE;
UFS_WAPBL_UPDATE(vp, NULL, NULL, UPDATE_DIROP);
}
UFS_WAPBL_END(vp->v_mount);
out1:
VOP_UNLOCK(vp);
out2:
VN_KNOTE(vp, NOTE_LINK);
VN_KNOTE(dvp, NOTE_WRITE);
vput(dvp);
fstrans_done(dvp->v_mount);
return (error);
}
/*
* Last reference to an inode. If necessary, write or delete it.
*/
int
ufs_inactive(void *v)
{
struct vop_inactive_args /* {
struct vnode *a_vp;
struct bool *a_recycle;
} */ *ap = v;
struct vnode *vp = ap->a_vp;
struct inode *ip = VTOI(vp);
struct mount *transmp;
mode_t mode;
int error = 0;
int logged = 0;
UFS_WAPBL_JUNLOCK_ASSERT(vp->v_mount);
transmp = vp->v_mount;
fstrans_start(transmp, FSTRANS_SHARED);
/*
* Ignore inodes related to stale file handles.
*/
if (ip->i_mode == 0)
goto out;
if (ip->i_ffs_effnlink == 0 && DOINGSOFTDEP(vp))
softdep_releasefile(ip);
if (ip->i_nlink <= 0 && (vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
error = UFS_WAPBL_BEGIN(vp->v_mount);
if (error)
goto out;
logged = 1;
#ifdef QUOTA
(void)chkiq(ip, -1, NOCRED, 0);
#endif
#ifdef UFS_EXTATTR
ufs_extattr_vnode_inactive(vp, curlwp);
#endif
if (ip->i_size != 0) {
/*
* When journaling, only truncate one indirect block
* at a time
*/
if (vp->v_mount->mnt_wapbl) {
uint64_t incr = MNINDIR(ip->i_ump) <<
vp->v_mount->mnt_fs_bshift; /* Power of 2 */
uint64_t base = NDADDR <<
vp->v_mount->mnt_fs_bshift;
while (!error && ip->i_size > base + incr) {
/*
* round down to next full indirect
* block boundary.
*/
uint64_t nsize = base +
((ip->i_size - base - 1) &
~(incr - 1));
error = UFS_TRUNCATE(vp, nsize, 0,
NOCRED);
if (error)
break;
UFS_WAPBL_END(vp->v_mount);
error = UFS_WAPBL_BEGIN(vp->v_mount);
if (error)
goto out;
}
}
if (!error)
error = UFS_TRUNCATE(vp, (off_t)0, 0, NOCRED);
}
/*
* Setting the mode to zero needs to wait for the inode
* to be written just as does a change to the link count.
* So, rather than creating a new entry point to do the
* same thing, we just use softdep_change_linkcnt().
*/
DIP_ASSIGN(ip, rdev, 0);
mode = ip->i_mode;
ip->i_mode = 0;
DIP_ASSIGN(ip, mode, 0);
ip->i_flag |= IN_CHANGE | IN_UPDATE;
mutex_enter(&vp->v_interlock);
vp->v_iflag |= VI_FREEING;
mutex_exit(&vp->v_interlock);
if (DOINGSOFTDEP(vp))
softdep_change_linkcnt(ip);
UFS_VFREE(vp, ip->i_number, mode);
}
if (ip->i_flag & (IN_CHANGE | IN_UPDATE | IN_MODIFIED)) {
if (!logged++) {
int err;
err = UFS_WAPBL_BEGIN(vp->v_mount);
if (err)
goto out;
}
UFS_UPDATE(vp, NULL, NULL, 0);
}
if (logged)
UFS_WAPBL_END(vp->v_mount);
out:
/*
* If we are done with the inode, reclaim it
* so that it can be reused immediately.
*/
*ap->a_recycle = (ip->i_mode == 0);
VOP_UNLOCK(vp, 0);
fstrans_done(transmp);
return (error);
}
int
ufs_mkdir(void *v)
{
struct vop_mkdir_v3_args /* {
struct vnode *a_dvp;
struct vnode **a_vpp;
struct componentname *a_cnp;
struct vattr *a_vap;
} */ *ap = v;
struct vnode *dvp = ap->a_dvp, *tvp;
struct vattr *vap = ap->a_vap;
struct componentname *cnp = ap->a_cnp;
struct inode *ip, *dp = VTOI(dvp);
struct buf *bp;
struct dirtemplate dirtemplate;
struct direct *newdir;
int error, dmode;
struct ufsmount *ump = dp->i_ump;
int dirblksiz = ump->um_dirblksiz;
struct ufs_lookup_results *ulr;
fstrans_start(dvp->v_mount, FSTRANS_SHARED);
/* XXX should handle this material another way */
ulr = &dp->i_crap;
UFS_CHECK_CRAPCOUNTER(dp);
if ((nlink_t)dp->i_nlink >= LINK_MAX) {
error = EMLINK;
goto out;
}
dmode = vap->va_mode & ACCESSPERMS;
dmode |= IFDIR;
/*
* Must simulate part of ufs_makeinode here to acquire the inode,
* but not have it entered in the parent directory. The entry is
* made later after writing "." and ".." entries.
*/
if ((error = UFS_VALLOC(dvp, dmode, cnp->cn_cred, ap->a_vpp)) != 0)
goto out;
tvp = *ap->a_vpp;
ip = VTOI(tvp);
error = UFS_WAPBL_BEGIN(ap->a_dvp->v_mount);
if (error) {
UFS_VFREE(tvp, ip->i_number, dmode);
vput(tvp);
goto out;
}
ip->i_uid = kauth_cred_geteuid(cnp->cn_cred);
DIP_ASSIGN(ip, uid, ip->i_uid);
ip->i_gid = dp->i_gid;
DIP_ASSIGN(ip, gid, ip->i_gid);
#if defined(QUOTA) || defined(QUOTA2)
if ((error = chkiq(ip, 1, cnp->cn_cred, 0))) {
UFS_VFREE(tvp, ip->i_number, dmode);
UFS_WAPBL_END(dvp->v_mount);
fstrans_done(dvp->v_mount);
vput(tvp);
return (error);
}
#endif
ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
ip->i_mode = dmode;
DIP_ASSIGN(ip, mode, dmode);
tvp->v_type = VDIR; /* Rest init'd in getnewvnode(). */
ip->i_nlink = 2;
DIP_ASSIGN(ip, nlink, 2);
if (cnp->cn_flags & ISWHITEOUT) {
ip->i_flags |= UF_OPAQUE;
DIP_ASSIGN(ip, flags, ip->i_flags);
}
/*
* Bump link count in parent directory to reflect work done below.
* Should be done before reference is created so cleanup is
* possible if we crash.
*/
dp->i_nlink++;
DIP_ASSIGN(dp, nlink, dp->i_nlink);
dp->i_flag |= IN_CHANGE;
if ((error = UFS_UPDATE(dvp, NULL, NULL, UPDATE_DIROP)) != 0)
goto bad;
/*
* Initialize directory with "." and ".." from static template.
*/
dirtemplate = mastertemplate;
dirtemplate.dotdot_reclen = dirblksiz - dirtemplate.dot_reclen;
dirtemplate.dot_ino = ufs_rw32(ip->i_number, UFS_MPNEEDSWAP(ump));
dirtemplate.dotdot_ino = ufs_rw32(dp->i_number, UFS_MPNEEDSWAP(ump));
dirtemplate.dot_reclen = ufs_rw16(dirtemplate.dot_reclen,
UFS_MPNEEDSWAP(ump));
dirtemplate.dotdot_reclen = ufs_rw16(dirtemplate.dotdot_reclen,
UFS_MPNEEDSWAP(ump));
if (ump->um_maxsymlinklen <= 0) {
#if BYTE_ORDER == LITTLE_ENDIAN
if (UFS_MPNEEDSWAP(ump) == 0)
#else
if (UFS_MPNEEDSWAP(ump) != 0)
#endif
{
dirtemplate.dot_type = dirtemplate.dot_namlen;
dirtemplate.dotdot_type = dirtemplate.dotdot_namlen;
dirtemplate.dot_namlen = dirtemplate.dotdot_namlen = 0;
} else
dirtemplate.dot_type = dirtemplate.dotdot_type = 0;
}
if ((error = UFS_BALLOC(tvp, (off_t)0, dirblksiz, cnp->cn_cred,
B_CLRBUF, &bp)) != 0)
goto bad;
ip->i_size = dirblksiz;
DIP_ASSIGN(ip, size, dirblksiz);
ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
uvm_vnp_setsize(tvp, ip->i_size);
memcpy((void *)bp->b_data, (void *)&dirtemplate, sizeof dirtemplate);
/*
* Directory set up, now install its entry in the parent directory.
* We must write out the buffer containing the new directory body
* before entering the new name in the parent.
*/
if ((error = VOP_BWRITE(bp->b_vp, bp)) != 0)
goto bad;
if ((error = UFS_UPDATE(tvp, NULL, NULL, UPDATE_DIROP)) != 0) {
goto bad;
}
newdir = pool_cache_get(ufs_direct_cache, PR_WAITOK);
ufs_makedirentry(ip, cnp, newdir);
error = ufs_direnter(dvp, ulr, tvp, newdir, cnp, bp);
pool_cache_put(ufs_direct_cache, newdir);
bad:
if (error == 0) {
VN_KNOTE(dvp, NOTE_WRITE | NOTE_LINK);
VOP_UNLOCK(tvp);
UFS_WAPBL_END(dvp->v_mount);
} else {
dp->i_nlink--;
DIP_ASSIGN(dp, nlink, dp->i_nlink);
dp->i_flag |= IN_CHANGE;
UFS_WAPBL_UPDATE(dvp, NULL, NULL, UPDATE_DIROP);
/*
* No need to do an explicit UFS_TRUNCATE here, vrele will
* do this for us because we set the link count to 0.
*/
ip->i_nlink = 0;
DIP_ASSIGN(ip, nlink, 0);
ip->i_flag |= IN_CHANGE;
UFS_WAPBL_UPDATE(tvp, NULL, NULL, UPDATE_DIROP);
UFS_WAPBL_END(dvp->v_mount);
vput(tvp);
}
out:
fstrans_done(dvp->v_mount);
return (error);
}
/*
* ulfs_link: create hard link.
*/
int
ulfs_link(void *v)
{
struct vop_link_v2_args /* {
struct vnode *a_dvp;
struct vnode *a_vp;
struct componentname *a_cnp;
} */ *ap = v;
struct vnode *dvp = ap->a_dvp;
struct vnode *vp = ap->a_vp;
struct componentname *cnp = ap->a_cnp;
struct mount *mp = dvp->v_mount;
struct inode *ip;
int error;
struct ulfs_lookup_results *ulr;
KASSERT(dvp != vp);
KASSERT(vp->v_type != VDIR);
KASSERT(mp == vp->v_mount); /* XXX Not stable without lock. */
/* XXX should handle this material another way */
ulr = &VTOI(dvp)->i_crap;
ULFS_CHECK_CRAPCOUNTER(VTOI(dvp));
fstrans_start(mp, FSTRANS_SHARED);
error = vn_lock(vp, LK_EXCLUSIVE);
if (error) {
VOP_ABORTOP(dvp, cnp);
goto out2;
}
ip = VTOI(vp);
if ((nlink_t)ip->i_nlink >= LINK_MAX) {
VOP_ABORTOP(dvp, cnp);
error = EMLINK;
goto out1;
}
if (ip->i_flags & (IMMUTABLE | APPEND)) {
VOP_ABORTOP(dvp, cnp);
error = EPERM;
goto out1;
}
ip->i_nlink++;
DIP_ASSIGN(ip, nlink, ip->i_nlink);
ip->i_flag |= IN_CHANGE;
error = lfs_update(vp, NULL, NULL, UPDATE_DIROP);
if (!error) {
error = ulfs_direnter(dvp, ulr, vp,
cnp, ip->i_number, LFS_IFTODT(ip->i_mode), NULL);
}
if (error) {
ip->i_nlink--;
DIP_ASSIGN(ip, nlink, ip->i_nlink);
ip->i_flag |= IN_CHANGE;
}
out1:
VOP_UNLOCK(vp);
out2:
VN_KNOTE(vp, NOTE_LINK);
VN_KNOTE(dvp, NOTE_WRITE);
fstrans_done(mp);
return (error);
}