static int
zfs_replay_remove(void *arg1, char *arg2, boolean_t byteswap)
{
zfsvfs_t *zfsvfs = (zfsvfs_t *)arg1;
lr_remove_t *lr = (lr_remove_t *)arg2;
char *name = (char *)(lr + 1); /* name follows lr_remove_t */
znode_t *dzp;
int error;
int vflg = 0;
if (byteswap)
byteswap_uint64_array(lr, sizeof (*lr));
if ((error = zfs_zget(zfsvfs, lr->lr_doid, &dzp)) != 0)
return (error);
if (lr->lr_common.lrc_txtype & TX_CI)
vflg |= FIGNORECASE;
switch ((int)lr->lr_common.lrc_txtype) {
case TX_REMOVE:
error = VOP_REMOVE(ZTOV(dzp), name, kcred, NULL, vflg);
break;
case TX_RMDIR:
error = VOP_RMDIR(ZTOV(dzp), name, NULL, kcred, NULL, vflg);
break;
default:
error = ENOTSUP;
}
VN_RELE(ZTOV(dzp));
return (error);
}
static int zfsfuse_unlink(fuse_req_t req, fuse_ino_t parent, const char *name)
{
if(strlen(name) >= MAXNAMELEN)
return ENAMETOOLONG;
vfs_t *vfs = (vfs_t *) fuse_req_userdata(req);
zfsvfs_t *zfsvfs = vfs->vfs_data;
ZFS_ENTER(zfsvfs);
znode_t *znode;
int error = zfs_zget(zfsvfs, parent, &znode, B_FALSE);
if(error) {
ZFS_EXIT(zfsvfs);
/* If the inode we are trying to get was recently deleted
dnode_hold_impl will return EEXIST instead of ENOENT */
return error == EEXIST ? ENOENT : error;
}
ASSERT(znode != NULL);
vnode_t *dvp = ZTOV(znode);
ASSERT(dvp != NULL);
cred_t cred;
zfsfuse_getcred(req, &cred);
error = VOP_REMOVE(dvp, (char *) name, &cred, NULL, 0);
VN_RELE(dvp);
ZFS_EXIT(zfsvfs);
return error;
}
/*
* union_remove:
*
* Remove the specified cnp. The dvp and vp are passed to us locked
* and must remain locked on return.
*
* union_remove(struct vnode *a_dvp, struct vnode *a_vp,
* struct componentname *a_cnp)
*/
static int
union_remove(struct vop_old_remove_args *ap)
{
struct union_node *dun = VTOUNION(ap->a_dvp);
struct union_node *un = VTOUNION(ap->a_vp);
struct componentname *cnp = ap->a_cnp;
struct thread *td = cnp->cn_td;
struct vnode *uppervp;
struct vnode *upperdvp;
int error;
if ((upperdvp = union_lock_upper(dun, td)) == NULLVP)
panic("union remove: null upper vnode");
if ((uppervp = union_lock_upper(un, td)) != NULLVP) {
if (union_dowhiteout(un, cnp->cn_cred, td))
cnp->cn_flags |= CNP_DOWHITEOUT;
error = VOP_REMOVE(upperdvp, uppervp, cnp);
#if 0
/* XXX */
if (!error)
union_removed_upper(un);
#endif
union_unlock_upper(uppervp, td);
} else {
error = union_mkwhiteout(
MOUNTTOUNIONMOUNT(ap->a_dvp->v_mount),
upperdvp, ap->a_cnp, un->un_path);
}
union_unlock_upper(upperdvp, td);
return (error);
}
static int
auto_remove(
vnode_t *dvp,
char *nm,
cred_t *cred,
caller_context_t *ct,
int flags)
{
vnode_t *newvp;
int error;
AUTOFS_DPRINT((4, "auto_remove dvp %p nm %s\n", (void *)dvp, nm));
if (error = auto_trigger_mount(dvp, cred, &newvp))
goto done;
if (newvp != NULL) {
/*
* Node is now mounted on.
*/
if (vn_is_readonly(newvp))
error = EROFS;
else
error = VOP_REMOVE(newvp, nm, cred, ct, flags);
VN_RELE(newvp);
} else
error = ENOSYS;
done:
AUTOFS_DPRINT((5, "auto_remove: error=%d\n", error));
return (error);
}
static int
zfs_replay_remove(zfsvfs_t *zfsvfs, void *data, boolean_t byteswap)
{
#ifndef TODO_OSV
kprintf("TX_REMOVE\n");
return EOPNOTSUPP;
#else
lr_remove_t *lr = data;
char *name = (char *)(lr + 1); /* name follows lr_remove_t */
znode_t *dzp;
struct componentname cn;
vnode_t *vp;
int error;
int vflg = 0;
if (byteswap)
byteswap_uint64_array(lr, sizeof (*lr));
if ((error = zfs_zget(zfsvfs, lr->lr_doid, &dzp)) != 0)
return (error);
if (lr->lr_common.lrc_txtype & TX_CI)
vflg |= FIGNORECASE;
cn.cn_nameptr = name;
cn.cn_namelen = strlen(name);
cn.cn_nameiop = DELETE;
cn.cn_flags = ISLASTCN | SAVENAME;
cn.cn_lkflags = LK_EXCLUSIVE | LK_RETRY;
cn.cn_cred = kcred;
cn.cn_thread = curthread;
vn_lock(ZTOV(dzp), LK_EXCLUSIVE | LK_RETRY);
error = VOP_LOOKUP(ZTOV(dzp), &vp, &cn);
if (error != 0) {
VOP_UNLOCK(ZTOV(dzp), 0);
goto fail;
}
switch ((int)lr->lr_common.lrc_txtype) {
case TX_REMOVE:
error = VOP_REMOVE(ZTOV(dzp), vp, &cn /*,vflg*/);
break;
case TX_RMDIR:
error = VOP_RMDIR(ZTOV(dzp), vp, &cn /*,vflg*/);
break;
default:
error = ENOTSUP;
}
vput(vp);
VOP_UNLOCK(ZTOV(dzp), 0);
fail:
VN_RELE(ZTOV(dzp));
return (error);
#endif
}
int
RUMP_VOP_REMOVE(struct vnode *dvp,
struct vnode *vp,
struct componentname *cnp)
{
int error;
rump_schedule();
error = VOP_REMOVE(dvp, vp, cnp);
rump_unschedule();
return error;
}
extern int
vnode_iop_unlink(
INODE_T *dir,
DENT_T *dent
)
{
int err = 0;
VNODE_T *obj;
struct unlink_ctx ctx;
CALL_DATA_T cd;
struct dentry *peer;
ASSERT_I_SEM_MINE(dent->d_inode);
ASSERT_I_SEM_MINE(dir);
ASSERT(MDKI_INOISMVFS(dir));
if (!S_ISDIR(dir->i_mode)) {
/* bogus */
return -ENOTDIR;
}
peer = vnlayer_dentry_peer(dent);
mdki_linux_init_call_data(&cd);
ctx.dentry = dent;
ctx.done = FALSE;
if (dent->d_inode && MDKI_INOISMVFS(dent->d_inode))
obj = ITOV(dent->d_inode); /* no extra reference (be careful) */
else
obj = NULL;
err = VOP_REMOVE(ITOV(dir), obj, (char *)dent->d_name.name, &cd, &ctx);
/* XXX pullup attributes on removed object, if it's not gone yet? */
err = mdki_errno_unix_to_linux(err);
mdki_linux_destroy_call_data(&cd);
/* XXX Don't d_delete(dentry), our caller will do that */
if (peer != NULL) {
/*
* unhash peer name in other mode. Don't d_delete in case
* we're racing
*/
if (err == 0)
d_drop(peer);
VNODE_DPUT(peer);
}
return err;
}
static int
xattr_dir_remove(vnode_t *dvp, char *name, cred_t *cr, caller_context_t *ct,
int flags)
{
vnode_t *pvp;
int error;
if (is_sattr_name(name)) {
return (EACCES);
}
error = xattr_dir_realdir(dvp, &pvp, LOOKUP_XATTR, cr, ct);
if (error == 0) {
error = VOP_REMOVE(pvp, name, cr, ct, flags);
}
return (error);
}
/* Does most of the work for remove(). */
int
vfs_remove(char *path)
{
struct vnode *dir;
char name[NAME_MAX+1];
int result;
result = vfs_lookparent(path, &dir, name, sizeof(name));
if (result) {
return result;
}
result = VOP_REMOVE(dir, name);
VOP_DECREF(dir);
return result;
}
int
smb_vop_remove(vnode_t *dvp, char *name, int flags, cred_t *cr)
{
int error;
int option_flags = 0;
char *np = name;
char namebuf[MAXNAMELEN];
if (flags & SMB_IGNORE_CASE)
option_flags = FIGNORECASE;
if (flags & SMB_CATIA)
np = smb_vop_catia_v5tov4(name, namebuf, sizeof (namebuf));
error = VOP_REMOVE(dvp, np, cr, &smb_ct, option_flags);
return (error);
}
STATIC int
linvfs_unlink(
struct inode *dir,
struct dentry *dentry)
{
struct inode *inode;
vnode_t *dvp; /* directory containing name to remove */
int error;
inode = dentry->d_inode;
dvp = LINVFS_GET_VP(dir);
VOP_REMOVE(dvp, dentry, NULL, error);
if (!error) {
validate_fields(dir); /* For size only */
validate_fields(inode);
}
return -error;
}
static int
unionfs_remove(void *v)
{
struct vop_remove_args *ap = v;
int error;
struct unionfs_node *dunp;
struct unionfs_node *unp;
struct unionfs_mount *ump;
struct vnode *udvp;
struct vnode *uvp;
struct vnode *lvp;
struct componentname *cnp;
UNIONFS_INTERNAL_DEBUG("unionfs_remove: enter\n");
error = 0;
dunp = VTOUNIONFS(ap->a_dvp);
unp = VTOUNIONFS(ap->a_vp);
udvp = dunp->un_uppervp;
uvp = unp->un_uppervp;
lvp = unp->un_lowervp;
cnp = ap->a_cnp;
if (udvp == NULLVP)
return (EROFS);
if (uvp != NULLVP) {
ump = MOUNTTOUNIONFSMOUNT(ap->a_vp->v_mount);
if (ump->um_whitemode == UNIONFS_WHITE_ALWAYS || lvp != NULLVP)
cnp->cn_flags |= DOWHITEOUT;
error = VOP_REMOVE(udvp, uvp, cnp);
} else if (lvp != NULLVP)
error = unionfs_mkwhiteout(udvp, cnp, unp->un_path);
UNIONFS_INTERNAL_DEBUG("unionfs_remove: leave (%d)\n", error);
return (error);
}
STATIC inline void
cleanup_inode(
vnode_t *dvp,
vnode_t *vp,
struct dentry *dentry,
int mode)
{
struct dentry teardown = {};
int err2;
/* Oh, the horror.
* If we can't add the ACL or we fail in
* linvfs_init_security we must back out.
* ENOSPC can hit here, among other things.
*/
teardown.d_inode = LINVFS_GET_IP(vp);
teardown.d_name = dentry->d_name;
if (S_ISDIR(mode))
VOP_RMDIR(dvp, &teardown, NULL, err2);
else
VOP_REMOVE(dvp, &teardown, NULL, err2);
VN_RELE(vp);
}
STATIC int
linvfs_mknod(
struct inode *dir,
struct dentry *dentry,
int mode,
int rdev)
{
struct inode *ip;
vattr_t va;
vnode_t *vp = NULL, *dvp = LINVFS_GET_VP(dir);
xfs_acl_t *default_acl = NULL;
attrexists_t test_default_acl = _ACL_DEFAULT_EXISTS;
int error;
if (test_default_acl && test_default_acl(dvp)) {
if (!_ACL_ALLOC(default_acl))
return -ENOMEM;
if (!_ACL_GET_DEFAULT(dvp, default_acl)) {
_ACL_FREE(default_acl);
default_acl = NULL;
}
}
#ifdef CONFIG_XFS_POSIX_ACL
/*
* Conditionally compiled so that the ACL base kernel changes can be
* split out into separate patches - remove this once MS_POSIXACL is
* accepted, or some other way to implement this exists.
*/
if (IS_POSIXACL(dir) && !default_acl && has_fs_struct(current))
mode &= ~current->fs->umask;
#endif
memset(&va, 0, sizeof(va));
va.va_mask = XFS_AT_TYPE|XFS_AT_MODE;
va.va_type = IFTOVT(mode);
va.va_mode = mode;
switch (mode & S_IFMT) {
case S_IFCHR: case S_IFBLK: case S_IFIFO: case S_IFSOCK:
va.va_rdev = XFS_MKDEV(MAJOR(rdev), MINOR(rdev));
va.va_mask |= XFS_AT_RDEV;
/*FALLTHROUGH*/
case S_IFREG:
VOP_CREATE(dvp, dentry, &va, &vp, NULL, error);
break;
case S_IFDIR:
VOP_MKDIR(dvp, dentry, &va, &vp, NULL, error);
break;
default:
error = EINVAL;
break;
}
if (default_acl) {
if (!error) {
error = _ACL_INHERIT(vp, &va, default_acl);
if (!error) {
VMODIFY(vp);
} else {
struct dentry teardown = {};
int err2;
/* Oh, the horror.
* If we can't add the ACL we must back out.
* ENOSPC can hit here, among other things.
*/
teardown.d_inode = ip = LINVFS_GET_IP(vp);
teardown.d_name = dentry->d_name;
vn_mark_bad(vp);
if (S_ISDIR(mode))
VOP_RMDIR(dvp, &teardown, NULL, err2);
else
VOP_REMOVE(dvp, &teardown, NULL, err2);
VN_RELE(vp);
}
}
_ACL_FREE(default_acl);
}
if (!error) {
ASSERT(vp);
ip = LINVFS_GET_IP(vp);
if (S_ISCHR(mode) || S_ISBLK(mode))
ip->i_rdev = to_kdev_t(rdev);
else if (S_ISDIR(mode))
validate_fields(ip);
d_instantiate(dentry, ip);
validate_fields(dir);
}
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.
*/
int
ext2fs_rename(void *v)
{
struct vop_rename_args *ap = v;
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 proc *p = fcnp->cn_proc;
struct ext2fs_dirtemplate dirbuf;
/* struct timespec ts; */
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("ext2fs_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:
VOP_ABORTOP(tdvp, tcnp); /* XXX, why not in NFS? */
if (tdvp == tvp)
vrele(tdvp);
else
vput(tdvp);
if (tvp)
vput(tvp);
VOP_ABORTOP(fdvp, fcnp); /* XXX, why not in NFS? */
vrele(fdvp);
vrele(fvp);
return (error);
}
/*
* Check if just deleting a link name.
*/
if (tvp && ((VTOI(tvp)->i_e2fs_flags & (EXT2_IMMUTABLE | EXT2_APPEND)) ||
(VTOI(tdvp)->i_e2fs_flags & EXT2_APPEND))) {
error = EPERM;
goto abortit;
}
if (fvp == tvp) {
if (fvp->v_type == VDIR) {
error = EINVAL;
goto abortit;
}
/* Release destination completely. */
VOP_ABORTOP(tdvp, tcnp);
vput(tdvp);
vput(tvp);
/* Delete source. */
vrele(fdvp);
vrele(fvp);
fcnp->cn_flags &= ~MODMASK;
fcnp->cn_flags |= LOCKPARENT | LOCKLEAF;
if ((fcnp->cn_flags & SAVESTART) == 0)
panic("ext2fs_rename: lost from startdir");
fcnp->cn_nameiop = DELETE;
(void) vfs_relookup(fdvp, &fvp, fcnp);
return (VOP_REMOVE(fdvp, fvp, fcnp));
}
if ((error = vn_lock(fvp, LK_EXCLUSIVE, p)) != 0)
goto abortit;
dp = VTOI(fdvp);
ip = VTOI(fvp);
if ((nlink_t)ip->i_e2fs_nlink >= LINK_MAX) {
VOP_UNLOCK(fvp, 0);
error = EMLINK;
goto abortit;
}
if ((ip->i_e2fs_flags & (EXT2_IMMUTABLE | EXT2_APPEND)) ||
(dp->i_e2fs_flags & EXT2_APPEND)) {
VOP_UNLOCK(fvp, 0);
error = EPERM;
goto abortit;
}
if ((ip->i_e2fs_mode & IFMT) == IFDIR) {
error = VOP_ACCESS(fvp, VWRITE, tcnp->cn_cred);
if (!error && tvp)
error = VOP_ACCESS(tvp, VWRITE, tcnp->cn_cred);
if (error) {
VOP_UNLOCK(fvp, 0);
error = EACCES;
goto abortit;
}
/*
* Avoid ".", "..", and aliases of "." for obvious reasons.
*/
if ((fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.') ||
dp == ip ||
(fcnp->cn_flags&ISDOTDOT) ||
(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_e2fs_nlink++;
ip->i_flag |= IN_CHANGE;
if ((error = ext2fs_update(ip, NULL, NULL, 1)) != 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);
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 = ext2fs_checkpath(ip, dp, tcnp->cn_cred);
if (error != 0)
goto out;
if ((tcnp->cn_flags & SAVESTART) == 0)
panic("ext2fs_rename: lost to startdir");
if ((error = vfs_relookup(tdvp, &tvp, tcnp)) != 0)
goto out;
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_dev != ip->i_dev)
panic("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_e2fs_nlink >= LINK_MAX) {
error = EMLINK;
goto bad;
}
dp->i_e2fs_nlink++;
dp->i_flag |= IN_CHANGE;
if ((error = ext2fs_update(dp, NULL, NULL, 1)) != 0)
goto bad;
}
error = ext2fs_direnter(ip, tdvp, tcnp);
if (error != 0) {
if (doingdirectory && newparent) {
dp->i_e2fs_nlink--;
dp->i_flag |= IN_CHANGE;
(void)ext2fs_update(dp, NULL, NULL, 1);
}
goto bad;
}
vput(tdvp);
} else {
if (xp->i_dev != dp->i_dev || xp->i_dev != ip->i_dev)
panic("rename: EXDEV");
/*
* Short circuit rename(foo, foo).
*/
if (xp->i_number == ip->i_number)
panic("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_e2fs_mode & S_ISTXT) && tcnp->cn_cred->cr_uid != 0 &&
tcnp->cn_cred->cr_uid != dp->i_e2fs_uid &&
xp->i_e2fs_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_e2fs_mode & IFMT) == IFDIR) {
if (!ext2fs_dirempty(xp, dp->i_number, tcnp->cn_cred) ||
xp->i_e2fs_nlink > 2) {
error = ENOTEMPTY;
goto bad;
}
if (!doingdirectory) {
error = ENOTDIR;
goto bad;
}
cache_purge(tdvp);
} else if (doingdirectory) {
error = EISDIR;
goto bad;
}
error = ext2fs_dirrewrite(dp, ip, tcnp);
if (error != 0)
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_e2fs_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_e2fs_nlink--;
if (doingdirectory) {
if (--xp->i_e2fs_nlink != 0)
panic("rename: linked directory");
error = ext2fs_truncate(xp, (off_t)0, IO_SYNC,
tcnp->cn_cred);
}
xp->i_flag |= IN_CHANGE;
vput(tvp);
xp = NULL;
}
/*
* 3) Unlink the source.
*/
fcnp->cn_flags &= ~MODMASK;
fcnp->cn_flags |= LOCKPARENT | LOCKLEAF;
if ((fcnp->cn_flags & SAVESTART) == 0)
panic("ext2fs_rename: lost from startdir");
(void) vfs_relookup(fdvp, &fvp, fcnp);
if (fvp != NULL) {
xp = VTOI(fvp);
dp = VTOI(fdvp);
} else {
/*
* From name has disappeared.
*/
if (doingdirectory)
panic("ext2fs_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 IRENAME flag ensures that it cannot be moved by another
* rename.
*/
if (xp != ip) {
if (doingdirectory)
panic("ext2fs_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_e2fs_nlink--;
dp->i_flag |= IN_CHANGE;
error = vn_rdwr(UIO_READ, fvp, (caddr_t)&dirbuf,
sizeof (struct ext2fs_dirtemplate), (off_t)0,
UIO_SYSSPACE, IO_NODELOCKED,
tcnp->cn_cred, NULL, curproc);
if (error == 0) {
namlen = dirbuf.dotdot_namlen;
if (namlen != 2 ||
dirbuf.dotdot_name[0] != '.' ||
dirbuf.dotdot_name[1] != '.') {
ufs_dirbad(xp, (doff_t)12,
"ext2fs_rename: mangled dir");
} else {
dirbuf.dotdot_ino = h2fs32(newparent);
(void) vn_rdwr(UIO_WRITE, fvp,
(caddr_t)&dirbuf,
sizeof (struct dirtemplate),
(off_t)0, UIO_SYSSPACE,
IO_NODELOCKED|IO_SYNC,
tcnp->cn_cred, NULL, curproc);
cache_purge(fdvp);
}
}
}
error = ext2fs_dirremove(fdvp, fcnp);
if (!error) {
xp->i_e2fs_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, p) == 0) {
ip->i_e2fs_nlink--;
ip->i_flag |= IN_CHANGE;
vput(fvp);
} else
vrele(fvp);
return (error);
}
/*ARGSUSED4*/
static int
devfs_setattr(
struct vnode *vp,
struct vattr *vap,
int flags,
struct cred *cr,
caller_context_t *ct)
{
struct dv_node *dv = VTODV(vp);
struct dv_node *ddv;
struct vnode *dvp;
struct vattr *map;
long int mask;
int error = 0;
struct vattr *free_vattr = NULL;
struct vattr *vattrp = NULL;
mperm_t mp;
int persist;
/*
* Message goes to console only. Otherwise, the message
* causes devfs_getattr to be invoked again... infinite loop
*/
dcmn_err2(("?devfs_setattr %s\n", dv->dv_name));
ASSERT(dv->dv_attr || dv->dv_attrvp);
if (!(vp->v_type == VDIR || vp->v_type == VCHR || vp->v_type == VBLK)) {
cmn_err(CE_WARN, /* panic ? */
"?%s: getattr on vnode type %d", dvnm, vp->v_type);
return (ENOENT);
}
if (vap->va_mask & AT_NOSET)
return (EINVAL);
/*
* If we are changing something we don't care about
* the persistence of, return success.
*/
if ((vap->va_mask &
(AT_MODE|AT_UID|AT_GID|AT_ATIME|AT_MTIME)) == 0)
return (0);
/*
* If driver overrides fs perm, disallow chmod
* and do not create attribute nodes.
*/
if (dv->dv_flags & DV_NO_FSPERM) {
ASSERT(dv->dv_attr);
if (vap->va_mask & (AT_MODE | AT_UID | AT_GID))
return (EPERM);
if ((vap->va_mask & (AT_ATIME|AT_MTIME)) == 0)
return (0);
rw_enter(&dv->dv_contents, RW_WRITER);
if (vap->va_mask & AT_ATIME)
dv->dv_attr->va_atime = vap->va_atime;
if (vap->va_mask & AT_MTIME)
dv->dv_attr->va_mtime = vap->va_mtime;
rw_exit(&dv->dv_contents);
return (0);
}
/*
* Directories are always created but device nodes are
* only used to persist non-default permissions.
*/
if (vp->v_type == VDIR) {
ASSERT(dv->dv_attr || dv->dv_attrvp);
return (devfs_setattr_dir(dv, vp, vap, flags, cr));
}
/*
* Allocate now before we take any locks
*/
vattrp = kmem_zalloc(sizeof (*vattrp), KM_SLEEP);
/* to ensure consistency, single thread setting of attributes */
rw_enter(&dv->dv_contents, RW_WRITER);
/*
* We don't need to create an attribute node
* to persist access or modification times.
*/
persist = (vap->va_mask & (AT_MODE | AT_UID | AT_GID));
/*
* If persisting something, get the default permissions
* for this minor to compare against what the attributes
* are now being set to. Default ordering is:
* - minor_perm match for this minor
* - mode supplied by ddi_create_priv_minor_node
* - devfs defaults
*/
if (persist) {
if (dev_minorperm(dv->dv_devi, dv->dv_name, &mp) != 0) {
mp.mp_uid = dv_vattr_file.va_uid;
mp.mp_gid = dv_vattr_file.va_gid;
mp.mp_mode = dv_vattr_file.va_mode;
if (dv->dv_flags & DV_DFLT_MODE) {
ASSERT((dv->dv_dflt_mode & ~S_IAMB) == 0);
mp.mp_mode &= ~S_IAMB;
mp.mp_mode |= dv->dv_dflt_mode;
dcmn_err5(("%s: setattr priv default 0%o\n",
dv->dv_name, mp.mp_mode));
} else {
dcmn_err5(("%s: setattr devfs default 0%o\n",
dv->dv_name, mp.mp_mode));
}
} else {
dcmn_err5(("%s: setattr minor perm default 0%o\n",
dv->dv_name, mp.mp_mode));
}
}
/*
* If we don't have a vattr for this node, construct one.
*/
if (dv->dv_attr) {
free_vattr = vattrp;
vattrp = NULL;
} else {
ASSERT(dv->dv_attrvp);
ASSERT(vp->v_type != VDIR);
*vattrp = dv_vattr_file;
error = VOP_GETATTR(dv->dv_attrvp, vattrp, 0, cr);
dsysdebug(error, ("vop_getattr %s %d\n",
dv->dv_name, error));
if (error)
goto out;
dv->dv_attr = vattrp;
dv_vattr_merge(dv, dv->dv_attr);
vattrp = NULL;
}
error = secpolicy_vnode_setattr(cr, vp, vap, dv->dv_attr,
flags, devfs_unlocked_access, dv);
if (error) {
dsysdebug(error, ("devfs_setattr %s secpolicy error %d\n",
dv->dv_name, error));
goto out;
}
/*
* Apply changes to the memory based attribute. This code
* is modeled after the tmpfs implementation of memory
* based vnodes
*/
map = dv->dv_attr;
mask = vap->va_mask;
/* Change file access modes. */
if (mask & AT_MODE) {
map->va_mode &= S_IFMT;
map->va_mode |= vap->va_mode & ~S_IFMT;
}
if (mask & AT_UID)
map->va_uid = vap->va_uid;
if (mask & AT_GID)
map->va_gid = vap->va_gid;
if (mask & AT_ATIME)
map->va_atime = vap->va_atime;
if (mask & AT_MTIME)
map->va_mtime = vap->va_mtime;
if (mask & (AT_MODE | AT_UID | AT_GID | AT_MTIME)) {
gethrestime(&map->va_ctime);
}
/*
* A setattr to defaults means we no longer need the
* shadow node as a persistent store, unless there
* are ACLs. Otherwise create a shadow node if one
* doesn't exist yet.
*/
if (persist) {
if ((dv_setattr_cmp(map, &mp) == 0) &&
((dv->dv_flags & DV_ACL) == 0)) {
if (dv->dv_attrvp) {
ddv = dv->dv_dotdot;
ASSERT(ddv->dv_attrvp);
error = VOP_REMOVE(ddv->dv_attrvp,
dv->dv_name, cr);
dsysdebug(error,
("vop_remove %s %s %d\n",
ddv->dv_name, dv->dv_name, error));
if (error == EROFS)
error = 0;
VN_RELE(dv->dv_attrvp);
dv->dv_attrvp = NULL;
}
ASSERT(dv->dv_attr);
} else {
if (mask & AT_MODE)
dcmn_err5(("%s persisting mode 0%o\n",
dv->dv_name, vap->va_mode));
if (mask & AT_UID)
dcmn_err5(("%s persisting uid %d\n",
dv->dv_name, vap->va_uid));
if (mask & AT_GID)
dcmn_err5(("%s persisting gid %d\n",
dv->dv_name, vap->va_gid));
if (dv->dv_attrvp == NULL) {
dvp = DVTOV(dv->dv_dotdot);
dv_shadow_node(dvp, dv->dv_name, vp,
NULL, NULLVP, cr,
DV_SHADOW_CREATE | DV_SHADOW_WRITE_HELD);
}
if (dv->dv_attrvp) {
error = VOP_SETATTR(dv->dv_attrvp,
vap, flags, cr, NULL);
dsysdebug(error, ("vop_setattr %s %d\n",
dv->dv_name, error));
}
/*
* Some file systems may return EROFS for a setattr
* on a readonly file system. In this case save
* as our own memory based attribute.
* NOTE: ufs is NOT one of these (see ufs_iupdat).
*/
if (dv->dv_attr && dv->dv_attrvp && error == 0) {
vattrp = dv->dv_attr;
dv->dv_attr = NULL;
} else if (error == EROFS)
error = 0;
}
}
out:
rw_exit(&dv->dv_contents);
if (vattrp)
kmem_free(vattrp, sizeof (*vattrp));
if (free_vattr)
kmem_free(free_vattr, sizeof (*free_vattr));
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.
*/
int
ufs_rename(void *v)
{
struct vop_rename_args *ap = v;
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 proc *p = fcnp->cn_proc;
struct inode *ip, *xp, *dp;
struct direct newdir;
int doingdirectory = 0, oldparent = 0, newparent = 0;
int error = 0;
#ifdef DIAGNOSTIC
if ((tcnp->cn_flags & HASBUF) == 0 ||
(fcnp->cn_flags & HASBUF) == 0)
panic("ufs_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:
VOP_ABORTOP(tdvp, tcnp);
if (tdvp == tvp)
vrele(tdvp);
else
vput(tdvp);
if (tvp)
vput(tvp);
VOP_ABORTOP(fdvp, fcnp);
vrele(fdvp);
vrele(fvp);
return (error);
}
if (tvp && ((DIP(VTOI(tvp), flags) & (IMMUTABLE | APPEND)) ||
(DIP(VTOI(tdvp), flags) & APPEND))) {
error = EPERM;
goto abortit;
}
/*
* Check if just deleting a link name or if we've lost a race.
* If another process completes the same rename after we've looked
* up the source and have blocked looking up the target, then the
* source and target inodes may be identical now although the
* names were never linked.
*/
if (fvp == tvp) {
if (fvp->v_type == VDIR) {
/*
* Linked directories are impossible, so we must
* have lost the race. Pretend that the rename
* completed before the lookup.
*/
error = ENOENT;
goto abortit;
}
/* Release destination completely. */
VOP_ABORTOP(tdvp, tcnp);
vput(tdvp);
vput(tvp);
/*
* Delete source. There is another race now that everything
* is unlocked, but this doesn't cause any new complications.
* relookup() may find a file that is unrelated to the
* original one, or it may fail. Too bad.
*/
vrele(fvp);
fcnp->cn_flags &= ~MODMASK;
fcnp->cn_flags |= LOCKPARENT | LOCKLEAF;
if ((fcnp->cn_flags & SAVESTART) == 0)
panic("ufs_rename: lost from startdir");
fcnp->cn_nameiop = DELETE;
if ((error = vfs_relookup(fdvp, &fvp, fcnp)) != 0)
return (error); /* relookup did vrele() */
vrele(fdvp);
return (VOP_REMOVE(fdvp, fvp, fcnp));
}
if ((error = vn_lock(fvp, LK_EXCLUSIVE, p)) != 0)
goto abortit;
/* fvp, tdvp, tvp now locked */
dp = VTOI(fdvp);
ip = VTOI(fvp);
if ((nlink_t) DIP(ip, nlink) >= LINK_MAX) {
VOP_UNLOCK(fvp, 0);
error = EMLINK;
goto abortit;
}
if ((DIP(ip, flags) & (IMMUTABLE | APPEND)) ||
(DIP(dp, flags) & APPEND)) {
VOP_UNLOCK(fvp, 0);
error = EPERM;
goto abortit;
}
if ((DIP(ip, mode) & IFMT) == IFDIR) {
error = VOP_ACCESS(fvp, VWRITE, tcnp->cn_cred);
if (!error && tvp)
error = VOP_ACCESS(tvp, VWRITE, tcnp->cn_cred);
if (error) {
VOP_UNLOCK(fvp, 0);
error = EACCES;
goto abortit;
}
/*
* Avoid ".", "..", and aliases of "." for obvious reasons.
*/
if ((fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.') ||
dp == ip ||
(fcnp->cn_flags & ISDOTDOT) ||
(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 = 1;
}
VN_KNOTE(fdvp, NOTE_WRITE); /* XXX right place? */
/*
* 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_effnlink++;
DIP_ADD(ip, nlink, 1);
ip->i_flag |= IN_CHANGE;
if (DOINGSOFTDEP(fvp))
softdep_change_linkcnt(ip, 0);
if ((error = UFS_UPDATE(ip, !DOINGSOFTDEP(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);
VOP_UNLOCK(fvp, 0);
/* tdvp and tvp locked */
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);
/*
* Compensate for the reference ufs_checkpath() loses.
*/
vref(tdvp);
/* Only tdvp is locked */
if ((error = ufs_checkpath(ip, dp, tcnp->cn_cred)) != 0) {
vrele(tdvp);
goto out;
}
if ((tcnp->cn_flags & SAVESTART) == 0)
panic("ufs_rename: lost to startdir");
if ((error = vfs_relookup(tdvp, &tvp, tcnp)) != 0)
goto out;
vrele(tdvp); /* relookup() acquired a reference */
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_dev != ip->i_dev)
panic("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) DIP(dp, nlink) >= LINK_MAX) {
error = EMLINK;
goto bad;
}
dp->i_effnlink++;
DIP_ADD(dp, nlink, 1);
dp->i_flag |= IN_CHANGE;
if (DOINGSOFTDEP(tdvp))
softdep_change_linkcnt(dp, 0);
if ((error = UFS_UPDATE(dp, !DOINGSOFTDEP(tdvp)))
!= 0) {
dp->i_effnlink--;
DIP_ADD(dp, nlink, -1);
dp->i_flag |= IN_CHANGE;
if (DOINGSOFTDEP(tdvp))
softdep_change_linkcnt(dp, 0);
goto bad;
}
}
ufs_makedirentry(ip, tcnp, &newdir);
if ((error = ufs_direnter(tdvp, NULL, &newdir, tcnp, NULL)) != 0) {
if (doingdirectory && newparent) {
dp->i_effnlink--;
DIP_ADD(dp, nlink, -1);
dp->i_flag |= IN_CHANGE;
if (DOINGSOFTDEP(tdvp))
softdep_change_linkcnt(dp, 0);
(void)UFS_UPDATE(dp, 1);
}
goto bad;
}
VN_KNOTE(tdvp, NOTE_WRITE);
vput(tdvp);
} else {
if (xp->i_dev != dp->i_dev || xp->i_dev != ip->i_dev)
panic("rename: EXDEV");
/*
* Short circuit rename(foo, foo).
*/
if (xp->i_number == ip->i_number)
panic("ufs_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 ((DIP(dp, mode) & S_ISTXT) && tcnp->cn_cred->cr_uid != 0 &&
tcnp->cn_cred->cr_uid != DIP(dp, uid) &&
DIP(xp, 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 ((DIP(xp, mode) & IFMT) == IFDIR) {
if (xp->i_effnlink > 2 ||
!ufs_dirempty(xp, dp->i_number, tcnp->cn_cred)) {
error = ENOTEMPTY;
goto bad;
}
if (!doingdirectory) {
error = ENOTDIR;
goto bad;
}
cache_purge(tdvp);
} else if (doingdirectory) {
error = EISDIR;
goto bad;
}
if ((error = ufs_dirrewrite(dp, xp, ip->i_number,
IFTODT(DIP(ip, mode)), (doingdirectory && newparent) ?
newparent : doingdirectory)) != 0)
goto bad;
if (doingdirectory) {
if (!newparent) {
dp->i_effnlink--;
if (DOINGSOFTDEP(tdvp))
softdep_change_linkcnt(dp, 0);
}
xp->i_effnlink--;
if (DOINGSOFTDEP(tvp))
softdep_change_linkcnt(xp, 0);
}
if (doingdirectory && !DOINGSOFTDEP(tvp)) {
/*
* Truncate inode. The only stuff left in the directory
* is "." and "..". The "." reference is inconsequential
* since we are quashing it. We have removed the "."
* reference and the reference in the parent directory,
* but there may be other hard links. The soft
* dependency code will arrange to do these operations
* after the parent directory entry has been deleted on
* disk, so when running with that code we avoid doing
* them now.
*/
if (!newparent) {
DIP_ADD(dp, nlink, -1);
dp->i_flag |= IN_CHANGE;
}
DIP_ADD(xp, nlink, -1);
xp->i_flag |= IN_CHANGE;
if ((error = UFS_TRUNCATE(VTOI(tvp), (off_t)0, IO_SYNC,
tcnp->cn_cred)) != 0)
goto bad;
}
VN_KNOTE(tdvp, NOTE_WRITE);
vput(tdvp);
VN_KNOTE(tvp, NOTE_DELETE);
vput(tvp);
xp = NULL;
}
/*
* 3) Unlink the source.
*/
fcnp->cn_flags &= ~MODMASK;
fcnp->cn_flags |= LOCKPARENT | LOCKLEAF;
if ((fcnp->cn_flags & SAVESTART) == 0)
panic("ufs_rename: lost from startdir");
if ((error = vfs_relookup(fdvp, &fvp, fcnp)) != 0) {
vrele(ap->a_fvp);
return (error);
}
vrele(fdvp);
if (fvp == NULL) {
/*
* From name has disappeared.
*/
if (doingdirectory)
panic("ufs_rename: lost dir entry");
vrele(ap->a_fvp);
return (0);
}
xp = VTOI(fvp);
dp = VTOI(fdvp);
/*
* 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; the IN_RENAME
* flag ensures that it cannot be moved by another rename or removed
* by a rmdir.
*/
if (xp != ip) {
if (doingdirectory)
panic("ufs_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) {
xp->i_offset = mastertemplate.dot_reclen;
ufs_dirrewrite(xp, dp, newparent, DT_DIR, 0);
cache_purge(fdvp);
}
error = ufs_dirremove(fdvp, xp, fcnp->cn_flags, 0);
xp->i_flag &= ~IN_RENAME;
}
VN_KNOTE(fvp, NOTE_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:
vrele(fdvp);
if (doingdirectory)
ip->i_flag &= ~IN_RENAME;
if (vn_lock(fvp, LK_EXCLUSIVE, p) == 0) {
ip->i_effnlink--;
DIP_ADD(ip, nlink, -1);
ip->i_flag |= IN_CHANGE;
ip->i_flag &= ~IN_RENAME;
if (DOINGSOFTDEP(fvp))
softdep_change_linkcnt(ip, 0);
vput(fvp);
} else
vrele(fvp);
return (error);
}
