/*
* Look up an entry in a directory.
*
* NOTE: '.' and '..' are handled as special cases because
* no directory entries are actually stored for them. If this is
* the root of a filesystem, then '.zfs' is also treated as a
* special pseudo-directory.
*/
int
zfs_dirlook(znode_t *dzp, char *name, struct inode **ipp, int flags,
int *deflg, pathname_t *rpnp)
{
zfs_dirlock_t *dl;
znode_t *zp;
int error = 0;
uint64_t parent;
if (name[0] == 0 || (name[0] == '.' && name[1] == 0)) {
*ipp = ZTOI(dzp);
igrab(*ipp);
} else if (name[0] == '.' && name[1] == '.' && name[2] == 0) {
zfs_sb_t *zsb = ZTOZSB(dzp);
/*
* If we are a snapshot mounted under .zfs, return
* the vp for the snapshot directory.
*/
if ((error = sa_lookup(dzp->z_sa_hdl,
SA_ZPL_PARENT(zsb), &parent, sizeof (parent))) != 0)
return (error);
#ifdef HAVE_SNAPSHOT
if (parent == dzp->z_id && zsb->z_parent != zsb) {
error = zfsctl_root_lookup(zsb->z_parent->z_ctldir,
"snapshot", ipp, NULL, 0, NULL, kcred,
NULL, NULL, NULL);
return (error);
}
#endif /* HAVE_SNAPSHOT */
rw_enter(&dzp->z_parent_lock, RW_READER);
error = zfs_zget(zsb, parent, &zp);
if (error == 0)
*ipp = ZTOI(zp);
rw_exit(&dzp->z_parent_lock);
#ifdef HAVE_SNAPSHOT
} else if (zfs_has_ctldir(dzp) && strcmp(name, ZFS_CTLDIR_NAME) == 0) {
*ipp = zfsctl_root(dzp);
#endif /* HAVE_SNAPSHOT */
} else {
int zf;
zf = ZEXISTS | ZSHARED;
if (flags & FIGNORECASE)
zf |= ZCILOOK;
error = zfs_dirent_lock(&dl, dzp, name, &zp, zf, deflg, rpnp);
if (error == 0) {
*ipp = ZTOI(zp);
zfs_dirent_unlock(dl);
dzp->z_zn_prefetch = B_TRUE; /* enable prefetching */
}
rpnp = NULL;
}
if ((flags & FIGNORECASE) && rpnp && !error)
(void) strlcpy(rpnp->pn_buf, name, rpnp->pn_bufsize);
return (error);
}
int __osd_attr_init(const struct lu_env *env, struct osd_device *osd,
sa_handle_t *sa_hdl, dmu_tx_t *tx,
struct lu_attr *la, uint64_t parent)
{
sa_bulk_attr_t *bulk = osd_oti_get(env)->oti_attr_bulk;
struct osa_attr *osa = &osd_oti_get(env)->oti_osa;
uint64_t gen;
uint64_t crtime[2];
timestruc_t now;
int cnt;
int rc;
LASSERT(sa_hdl);
gen = dmu_tx_get_txg(tx);
gethrestime(&now);
ZFS_TIME_ENCODE(&now, crtime);
osa->atime[0] = la->la_atime;
osa->ctime[0] = la->la_ctime;
osa->mtime[0] = la->la_mtime;
osa->mode = la->la_mode;
osa->uid = la->la_uid;
osa->gid = la->la_gid;
osa->rdev = la->la_rdev;
osa->nlink = la->la_nlink;
osa->flags = attrs_fs2zfs(la->la_flags);
osa->size = la->la_size;
/*
* we need to create all SA below upon object create.
*
* XXX The attribute order matters since the accounting callback relies
* on static offsets (i.e. SA_*_OFFSET, see zfs_space_delta_cb()) to
* look up the UID/GID attributes. Moreover, the callback does not seem
* to support the spill block.
* We define attributes in the same order as SA_*_OFFSET in order to
* work around the problem. See ORI-610.
*/
cnt = 0;
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MODE(osd), NULL, &osa->mode, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_SIZE(osd), NULL, &osa->size, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_GEN(osd), NULL, &gen, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_UID(osd), NULL, &osa->uid, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_GID(osd), NULL, &osa->gid, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_PARENT(osd), NULL, &parent, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_FLAGS(osd), NULL, &osa->flags, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_ATIME(osd), NULL, osa->atime, 16);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(osd), NULL, osa->mtime, 16);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(osd), NULL, osa->ctime, 16);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CRTIME(osd), NULL, crtime, 16);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_LINKS(osd), NULL, &osa->nlink, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_RDEV(osd), NULL, &osa->rdev, 8);
LASSERT(cnt <= ARRAY_SIZE(osd_oti_get(env)->oti_attr_bulk));
rc = -sa_replace_all_by_template(sa_hdl, bulk, cnt, tx);
return rc;
}
zfs_dirlook(znode_t *dzp, char *name, vnode_t **vpp)
#endif
{
zfs_dirlock_t *dl;
znode_t *zp;
int error = 0;
uint64_t parent;
#ifdef __APPLE__
char *name = cnp->cn_nameptr;
#endif
if (name[0] == 0 || (name[0] == '.' && name[1] == 0)) {
*vpp = ZTOV(dzp);
VN_HOLD(*vpp);
} else if (name[0] == '.' && name[1] == '.' && name[2] == 0) {
zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
/*
* If we are a snapshot mounted under .zfs, return
* the vp for the snapshot directory.
*/
if ((error = sa_lookup(dzp->z_sa_hdl,
SA_ZPL_PARENT(zfsvfs), &parent, sizeof (parent))) != 0)
return (error);
if (parent == dzp->z_id && zfsvfs->z_parent != zfsvfs) {
error = zfsctl_root_lookup(zfsvfs->z_parent->z_ctldir,
"snapshot", vpp, NULL, 0, NULL, kcred
/*, NULL, NULL, NULL*/);
return (error);
}
rw_enter(&dzp->z_parent_lock, RW_READER);
error = zfs_zget(zfsvfs, parent, &zp);
if (error == 0)
*vpp = ZTOV(zp);
rw_exit(&dzp->z_parent_lock);
} else if (zfs_has_ctldir(dzp) && strcmp(name, ZFS_CTLDIR_NAME) == 0) {
*vpp = zfsctl_root(dzp);
} else {
#ifdef __APPLE__
error = zfs_dirent_lock(&dl, dzp, cnp, &zp, ZEXISTS | ZSHARED);
#else
error = zfs_dirent_lock(&dl, dzp, name, &zp, ZEXISTS | ZSHARED);
#endif
if (error == 0) {
*vpp = ZTOV(zp);
zfs_dirent_unlock(dl);
dzp->z_zn_prefetch = B_TRUE; /* enable prefetching */
}
}
return (error);
}
static int
zfs_dd_lookup(znode_t *dzp, znode_t **zpp)
{
zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
znode_t *zp;
uint64_t parent;
int error;
ASSERT_VOP_LOCKED(ZTOV(dzp), __func__);
ASSERT(RRM_READ_HELD(&zfsvfs->z_teardown_lock));
if (dzp->z_unlinked)
return (ENOENT);
if ((error = sa_lookup(dzp->z_sa_hdl,
SA_ZPL_PARENT(zfsvfs), &parent, sizeof (parent))) != 0)
return (error);
error = zfs_zget(zfsvfs, parent, &zp);
if (error == 0)
*zpp = zp;
return (error);
}
/*
* Link zp into dl. Can only fail if zp has been unlinked.
*/
int
zfs_make_xattrdir(znode_t *zp, vattr_t *vap, vnode_t **xvpp, cred_t *cr)
{
zfsvfs_t *zfsvfs = zp->z_zfsvfs;
znode_t *xzp;
dmu_tx_t *tx;
int error;
zfs_acl_ids_t acl_ids;
boolean_t fuid_dirtied;
uint64_t parent;
*xvpp = NULL;
if ((error = zfs_zaccess(zp, ACE_WRITE_NAMED_ATTRS, 0, B_FALSE, cr)))
return (error);
if ((error = zfs_acl_ids_create(zp, IS_XATTR, vap, cr, NULL,
&acl_ids)) != 0)
return (error);
if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) {
zfs_acl_ids_free(&acl_ids);
return (EDQUOT);
}
top:
tx = dmu_tx_create(zfsvfs->z_os);
dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes +
ZFS_SA_BASE_ATTR_SIZE);
dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
fuid_dirtied = zfsvfs->z_fuid_dirty;
if (fuid_dirtied)
zfs_fuid_txhold(zfsvfs, tx);
error = dmu_tx_assign(tx, TXG_NOWAIT);
if (error) {
if (error == ERESTART) {
dmu_tx_wait(tx);
dmu_tx_abort(tx);
goto top;
}
zfs_acl_ids_free(&acl_ids);
dmu_tx_abort(tx);
return (error);
}
zfs_mknode(zp, vap, tx, cr, IS_XATTR, &xzp, &acl_ids);
#ifdef HAVE_ZPL
if (fuid_dirtied)
zfs_fuid_sync(zfsvfs, tx);
#endif /* HAVE_ZPL */
#ifdef DEBUG
error = sa_lookup(xzp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs),
&parent, sizeof (parent));
ASSERT(error == 0 && parent == zp->z_id);
#endif
VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), &xzp->z_id,
sizeof (xzp->z_id), tx));
(void) zfs_log_create(zfsvfs->z_log, tx, TX_MKXATTR, zp,
xzp, "", NULL, acl_ids.z_fuidp, vap);
zfs_acl_ids_free(&acl_ids);
dmu_tx_commit(tx);
*xvpp = ZTOV(xzp);
return (0);
}
void
zfs_sa_upgrade(sa_handle_t *hdl, dmu_tx_t *tx)
{
dmu_buf_t *db = sa_get_db(hdl);
znode_t *zp = sa_get_userdata(hdl);
zfsvfs_t *zfsvfs = zp->z_zfsvfs;
sa_bulk_attr_t bulk[20];
int count = 0;
sa_bulk_attr_t sa_attrs[20] = { { 0 } };
zfs_acl_locator_cb_t locate = { 0 };
uint64_t uid, gid, mode, rdev, xattr, parent;
uint64_t crtime[2], mtime[2], ctime[2];
zfs_acl_phys_t znode_acl;
char scanstamp[AV_SCANSTAMP_SZ];
boolean_t drop_lock = B_FALSE;
/*
* No upgrade if ACL isn't cached
* since we won't know which locks are held
* and ready the ACL would require special "locked"
* interfaces that would be messy
*/
if (zp->z_acl_cached == NULL || vnode_islnk(ZTOV(zp)))
return;
/*
* If the z_lock is held and we aren't the owner
* the just return since we don't want to deadlock
* trying to update the status of z_is_sa. This
* file can then be upgraded at a later time.
*
* Otherwise, we know we are doing the
* sa_update() that caused us to enter this function.
*/
if (mutex_owner(&zp->z_lock) != curthread) {
if (mutex_tryenter(&zp->z_lock) == 0)
return;
else
drop_lock = B_TRUE;
}
/* First do a bulk query of the attributes that aren't cached */
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CRTIME(zfsvfs), NULL, &crtime, 16);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL, &parent, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_XATTR(zfsvfs), NULL, &xattr, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_RDEV(zfsvfs), NULL, &rdev, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, &uid, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL, &gid, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ZNODE_ACL(zfsvfs), NULL,
&znode_acl, 88);
if (sa_bulk_lookup_locked(hdl, bulk, count) != 0)
goto done;
/*
* While the order here doesn't matter its best to try and organize
* it is such a way to pick up an already existing layout number
*/
count = 0;
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8);
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_SIZE(zfsvfs), NULL,
&zp->z_size, 8);
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_GEN(zfsvfs),
NULL, &zp->z_gen, 8);
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_UID(zfsvfs), NULL, &uid, 8);
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_GID(zfsvfs), NULL, &gid, 8);
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_PARENT(zfsvfs),
NULL, &parent, 8);
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_FLAGS(zfsvfs), NULL,
&zp->z_pflags, 8);
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_ATIME(zfsvfs), NULL,
zp->z_atime, 16);
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_MTIME(zfsvfs), NULL,
&mtime, 16);
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_CTIME(zfsvfs), NULL,
&ctime, 16);
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_CRTIME(zfsvfs), NULL,
&crtime, 16);
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_LINKS(zfsvfs), NULL,
&zp->z_links, 8);
if (vnode_isblk(zp->z_vnode) || vnode_islnk(zp->z_vnode))
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_RDEV(zfsvfs), NULL,
&rdev, 8);
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_DACL_COUNT(zfsvfs), NULL,
&zp->z_acl_cached->z_acl_count, 8);
if (zp->z_acl_cached->z_version < ZFS_ACL_VERSION_FUID)
zfs_acl_xform(zp, zp->z_acl_cached, CRED());
locate.cb_aclp = zp->z_acl_cached;
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_DACL_ACES(zfsvfs),
zfs_acl_data_locator, &locate, zp->z_acl_cached->z_acl_bytes);
if (xattr)
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_XATTR(zfsvfs),
NULL, &xattr, 8);
/* if scanstamp then add scanstamp */
if (zp->z_pflags & ZFS_BONUS_SCANSTAMP) {
bcopy((caddr_t)db->db_data + ZFS_OLD_ZNODE_PHYS_SIZE,
scanstamp, AV_SCANSTAMP_SZ);
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_SCANSTAMP(zfsvfs),
NULL, scanstamp, AV_SCANSTAMP_SZ);
zp->z_pflags &= ~ZFS_BONUS_SCANSTAMP;
}
VERIFY(dmu_set_bonustype(db, DMU_OT_SA, tx) == 0);
VERIFY(sa_replace_all_by_template_locked(hdl, sa_attrs,
count, tx) == 0);
if (znode_acl.z_acl_extern_obj)
VERIFY(0 == dmu_object_free(zfsvfs->z_os,
znode_acl.z_acl_extern_obj, tx));
zp->z_is_sa = B_TRUE;
done:
if (drop_lock)
mutex_exit(&zp->z_lock);
}
/*
* Link zp into dl. Can only fail if zp has been unlinked.
*/
int
zfs_link_create(zfs_dirlock_t *dl, znode_t *zp, dmu_tx_t *tx, int flag)
{
znode_t *dzp = dl->dl_dzp;
zfsvfs_t *zfsvfs = ZTOZSB(zp);
uint64_t value;
int zp_is_dir = S_ISDIR(ZTOI(zp)->i_mode);
sa_bulk_attr_t bulk[5];
uint64_t mtime[2], ctime[2];
uint64_t links;
int count = 0;
int error;
mutex_enter(&zp->z_lock);
if (!(flag & ZRENAMING)) {
if (zp->z_unlinked) { /* no new links to unlinked zp */
ASSERT(!(flag & (ZNEW | ZEXISTS)));
mutex_exit(&zp->z_lock);
return (SET_ERROR(ENOENT));
}
if (!(flag & ZNEW)) {
/*
* ZNEW nodes come from zfs_mknode() where the link
* count has already been initialised
*/
inc_nlink(ZTOI(zp));
links = ZTOI(zp)->i_nlink;
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs),
NULL, &links, sizeof (links));
}
}
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL,
&dzp->z_id, sizeof (dzp->z_id));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
&zp->z_pflags, sizeof (zp->z_pflags));
if (!(flag & ZNEW)) {
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
ctime, sizeof (ctime));
zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime,
ctime);
}
error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
ASSERT(error == 0);
mutex_exit(&zp->z_lock);
mutex_enter(&dzp->z_lock);
dzp->z_size++;
if (zp_is_dir)
inc_nlink(ZTOI(dzp));
links = ZTOI(dzp)->i_nlink;
count = 0;
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
&dzp->z_size, sizeof (dzp->z_size));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
&links, sizeof (links));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL,
mtime, sizeof (mtime));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
ctime, sizeof (ctime));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
&dzp->z_pflags, sizeof (dzp->z_pflags));
zfs_tstamp_update_setup(dzp, CONTENT_MODIFIED, mtime, ctime);
error = sa_bulk_update(dzp->z_sa_hdl, bulk, count, tx);
ASSERT(error == 0);
mutex_exit(&dzp->z_lock);
value = zfs_dirent(zp, zp->z_mode);
error = zap_add(ZTOZSB(zp)->z_os, dzp->z_id, dl->dl_name,
8, 1, &value, tx);
ASSERT(error == 0);
return (0);
}
int __osd_attr_init(const struct lu_env *env, udmu_objset_t *uos,
uint64_t oid, dmu_tx_t *tx, struct lu_attr *la)
{
sa_bulk_attr_t *bulk;
sa_handle_t *sa_hdl;
struct osa_attr *osa = &osd_oti_get(env)->oti_osa;
uint64_t gen;
uint64_t parent;
uint64_t crtime[2];
timestruc_t now;
int cnt;
int rc;
gethrestime(&now);
gen = dmu_tx_get_txg(tx);
ZFS_TIME_ENCODE(&now, crtime);
/* XXX: this should be real id of parent for ZPL access, but we have no
* such info in OSD, probably it can be part of dt_object_format */
parent = 0;
osa->atime[0] = la->la_atime;
osa->ctime[0] = la->la_ctime;
osa->mtime[0] = la->la_mtime;
osa->mode = la->la_mode;
osa->uid = la->la_uid;
osa->gid = la->la_gid;
osa->rdev = la->la_rdev;
osa->nlink = la->la_nlink;
osa->flags = la->la_flags;
osa->size = la->la_size;
/* Now add in all of the "SA" attributes */
rc = -sa_handle_get(uos->os, oid, NULL, SA_HDL_PRIVATE, &sa_hdl);
if (rc)
return rc;
OBD_ALLOC(bulk, sizeof(sa_bulk_attr_t) * 13);
if (bulk == NULL) {
rc = -ENOMEM;
goto out;
}
/*
* we need to create all SA below upon object create.
*
* XXX The attribute order matters since the accounting callback relies
* on static offsets (i.e. SA_*_OFFSET, see zfs_space_delta_cb()) to
* look up the UID/GID attributes. Moreover, the callback does not seem
* to support the spill block.
* We define attributes in the same order as SA_*_OFFSET in order to
* work around the problem. See ORI-610.
*/
cnt = 0;
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MODE(uos), NULL, &osa->mode, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_SIZE(uos), NULL, &osa->size, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_GEN(uos), NULL, &gen, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_UID(uos), NULL, &osa->uid, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_GID(uos), NULL, &osa->gid, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_PARENT(uos), NULL, &parent, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_FLAGS(uos), NULL, &osa->flags, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_ATIME(uos), NULL, osa->atime, 16);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(uos), NULL, osa->mtime, 16);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(uos), NULL, osa->ctime, 16);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CRTIME(uos), NULL, crtime, 16);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_LINKS(uos), NULL, &osa->nlink, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_RDEV(uos), NULL, &osa->rdev, 8);
rc = -sa_replace_all_by_template(sa_hdl, bulk, cnt, tx);
OBD_FREE(bulk, sizeof(sa_bulk_attr_t) * 13);
out:
sa_handle_destroy(sa_hdl);
return rc;
}
int
zfs_make_xattrdir(znode_t *zp, vattr_t *vap, struct inode **xipp, cred_t *cr)
{
zfs_sb_t *zsb = ZTOZSB(zp);
znode_t *xzp;
dmu_tx_t *tx;
int error;
zfs_acl_ids_t acl_ids;
boolean_t fuid_dirtied;
#ifdef DEBUG
uint64_t parent;
#endif
*xipp = NULL;
if ((error = zfs_zaccess(zp, ACE_WRITE_NAMED_ATTRS, 0, B_FALSE, cr)))
return (error);
if ((error = zfs_acl_ids_create(zp, IS_XATTR, vap, cr, NULL,
&acl_ids)) != 0)
return (error);
if (zfs_acl_ids_overquota(zsb, &acl_ids)) {
zfs_acl_ids_free(&acl_ids);
return (SET_ERROR(EDQUOT));
}
tx = dmu_tx_create(zsb->z_os);
dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes +
ZFS_SA_BASE_ATTR_SIZE);
dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
fuid_dirtied = zsb->z_fuid_dirty;
if (fuid_dirtied)
zfs_fuid_txhold(zsb, tx);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
zfs_acl_ids_free(&acl_ids);
dmu_tx_abort(tx);
return (error);
}
zfs_mknode(zp, vap, tx, cr, IS_XATTR, &xzp, &acl_ids);
if (fuid_dirtied)
zfs_fuid_sync(zsb, tx);
#ifdef DEBUG
error = sa_lookup(xzp->z_sa_hdl, SA_ZPL_PARENT(zsb),
&parent, sizeof (parent));
ASSERT(error == 0 && parent == zp->z_id);
#endif
VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_XATTR(zsb), &xzp->z_id,
sizeof (xzp->z_id), tx));
if (!zp->z_unlinked)
(void) zfs_log_create(zsb->z_log, tx, TX_MKXATTR, zp,
xzp, "", NULL, acl_ids.z_fuidp, vap);
zfs_acl_ids_free(&acl_ids);
dmu_tx_commit(tx);
*xipp = ZTOI(xzp);
return (0);
}
/*
* Create a new DMU object to hold a zfs znode.
*
* IN: dzp - parent directory for new znode
* vap - file attributes for new znode
* tx - dmu transaction id for zap operations
* cr - credentials of caller
* flag - flags:
* IS_ROOT_NODE - new object will be root
* IS_XATTR - new object is an attribute
* bonuslen - length of bonus buffer
* setaclp - File/Dir initial ACL
* fuidp - Tracks fuid allocation.
*
* OUT: zpp - allocated znode
*
*/
void
zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr,
uint_t flag, znode_t **zpp, zfs_acl_ids_t *acl_ids)
{
uint64_t crtime[2], atime[2], mtime[2], ctime[2];
uint64_t mode, size, links, parent, pflags;
uint64_t dzp_pflags = 0;
uint64_t rdev = 0;
zfs_sb_t *zsb = ZTOZSB(dzp);
dmu_buf_t *db;
timestruc_t now;
uint64_t gen, obj;
int bonuslen;
sa_handle_t *sa_hdl;
dmu_object_type_t obj_type;
sa_bulk_attr_t *sa_attrs;
int cnt = 0;
zfs_acl_locator_cb_t locate = { 0 };
if (zsb->z_replay) {
obj = vap->va_nodeid;
now = vap->va_ctime; /* see zfs_replay_create() */
gen = vap->va_nblocks; /* ditto */
} else {
obj = 0;
gethrestime(&now);
gen = dmu_tx_get_txg(tx);
}
obj_type = zsb->z_use_sa ? DMU_OT_SA : DMU_OT_ZNODE;
bonuslen = (obj_type == DMU_OT_SA) ?
DN_MAX_BONUSLEN : ZFS_OLD_ZNODE_PHYS_SIZE;
/*
* Create a new DMU object.
*/
/*
* There's currently no mechanism for pre-reading the blocks that will
* be needed to allocate a new object, so we accept the small chance
* that there will be an i/o error and we will fail one of the
* assertions below.
*/
if (S_ISDIR(vap->va_mode)) {
if (zsb->z_replay) {
VERIFY0(zap_create_claim_norm(zsb->z_os, obj,
zsb->z_norm, DMU_OT_DIRECTORY_CONTENTS,
obj_type, bonuslen, tx));
} else {
obj = zap_create_norm(zsb->z_os,
zsb->z_norm, DMU_OT_DIRECTORY_CONTENTS,
obj_type, bonuslen, tx);
}
} else {
if (zsb->z_replay) {
VERIFY0(dmu_object_claim(zsb->z_os, obj,
DMU_OT_PLAIN_FILE_CONTENTS, 0,
obj_type, bonuslen, tx));
} else {
obj = dmu_object_alloc(zsb->z_os,
DMU_OT_PLAIN_FILE_CONTENTS, 0,
obj_type, bonuslen, tx);
}
}
ZFS_OBJ_HOLD_ENTER(zsb, obj);
VERIFY(0 == sa_buf_hold(zsb->z_os, obj, NULL, &db));
/*
* If this is the root, fix up the half-initialized parent pointer
* to reference the just-allocated physical data area.
*/
if (flag & IS_ROOT_NODE) {
dzp->z_id = obj;
} else {
dzp_pflags = dzp->z_pflags;
}
/*
* If parent is an xattr, so am I.
*/
if (dzp_pflags & ZFS_XATTR) {
flag |= IS_XATTR;
}
if (zsb->z_use_fuids)
pflags = ZFS_ARCHIVE | ZFS_AV_MODIFIED;
else
pflags = 0;
if (S_ISDIR(vap->va_mode)) {
size = 2; /* contents ("." and "..") */
links = (flag & (IS_ROOT_NODE | IS_XATTR)) ? 2 : 1;
} else {
size = links = 0;
}
if (S_ISBLK(vap->va_mode) || S_ISCHR(vap->va_mode))
rdev = vap->va_rdev;
parent = dzp->z_id;
mode = acl_ids->z_mode;
if (flag & IS_XATTR)
pflags |= ZFS_XATTR;
/*
* No execs denied will be deterimed when zfs_mode_compute() is called.
*/
pflags |= acl_ids->z_aclp->z_hints &
(ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|ZFS_ACL_AUTO_INHERIT|
ZFS_ACL_DEFAULTED|ZFS_ACL_PROTECTED);
ZFS_TIME_ENCODE(&now, crtime);
ZFS_TIME_ENCODE(&now, ctime);
if (vap->va_mask & ATTR_ATIME) {
ZFS_TIME_ENCODE(&vap->va_atime, atime);
} else {
ZFS_TIME_ENCODE(&now, atime);
}
if (vap->va_mask & ATTR_MTIME) {
ZFS_TIME_ENCODE(&vap->va_mtime, mtime);
} else {
ZFS_TIME_ENCODE(&now, mtime);
}
/* Now add in all of the "SA" attributes */
VERIFY(0 == sa_handle_get_from_db(zsb->z_os, db, NULL, SA_HDL_SHARED,
&sa_hdl));
/*
* Setup the array of attributes to be replaced/set on the new file
*
* order for DMU_OT_ZNODE is critical since it needs to be constructed
* in the old znode_phys_t format. Don't change this ordering
*/
sa_attrs = kmem_alloc(sizeof (sa_bulk_attr_t) * ZPL_END, KM_SLEEP);
if (obj_type == DMU_OT_ZNODE) {
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zsb),
NULL, &atime, 16);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zsb),
NULL, &mtime, 16);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zsb),
NULL, &ctime, 16);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zsb),
NULL, &crtime, 16);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zsb),
NULL, &gen, 8);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zsb),
NULL, &mode, 8);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zsb),
NULL, &size, 8);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zsb),
NULL, &parent, 8);
} else {
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zsb),
NULL, &mode, 8);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zsb),
NULL, &size, 8);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zsb),
NULL, &gen, 8);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zsb),
NULL, &acl_ids->z_fuid, 8);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zsb),
NULL, &acl_ids->z_fgid, 8);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zsb),
NULL, &parent, 8);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zsb),
NULL, &pflags, 8);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zsb),
NULL, &atime, 16);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zsb),
NULL, &mtime, 16);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zsb),
NULL, &ctime, 16);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zsb),
NULL, &crtime, 16);
}
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_LINKS(zsb), NULL, &links, 8);
if (obj_type == DMU_OT_ZNODE) {
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_XATTR(zsb), NULL,
&empty_xattr, 8);
}
if (obj_type == DMU_OT_ZNODE ||
(S_ISBLK(vap->va_mode) || S_ISCHR(vap->va_mode))) {
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_RDEV(zsb),
NULL, &rdev, 8);
}
if (obj_type == DMU_OT_ZNODE) {
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zsb),
NULL, &pflags, 8);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zsb), NULL,
&acl_ids->z_fuid, 8);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zsb), NULL,
&acl_ids->z_fgid, 8);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PAD(zsb), NULL, pad,
sizeof (uint64_t) * 4);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ZNODE_ACL(zsb), NULL,
&acl_phys, sizeof (zfs_acl_phys_t));
} else if (acl_ids->z_aclp->z_version >= ZFS_ACL_VERSION_FUID) {
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_COUNT(zsb), NULL,
&acl_ids->z_aclp->z_acl_count, 8);
locate.cb_aclp = acl_ids->z_aclp;
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_ACES(zsb),
zfs_acl_data_locator, &locate,
acl_ids->z_aclp->z_acl_bytes);
mode = zfs_mode_compute(mode, acl_ids->z_aclp, &pflags,
acl_ids->z_fuid, acl_ids->z_fgid);
}
VERIFY(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx) == 0);
if (!(flag & IS_ROOT_NODE)) {
*zpp = zfs_znode_alloc(zsb, db, 0, obj_type, obj, sa_hdl,
ZTOI(dzp));
VERIFY(*zpp != NULL);
VERIFY(dzp != NULL);
} else {
/*
* If we are creating the root node, the "parent" we
* passed in is the znode for the root.
*/
*zpp = dzp;
(*zpp)->z_sa_hdl = sa_hdl;
}
(*zpp)->z_pflags = pflags;
(*zpp)->z_mode = mode;
if (obj_type == DMU_OT_ZNODE ||
acl_ids->z_aclp->z_version < ZFS_ACL_VERSION_FUID) {
VERIFY0(zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx));
}
kmem_free(sa_attrs, sizeof (sa_bulk_attr_t) * ZPL_END);
ZFS_OBJ_HOLD_EXIT(zsb, obj);
}
/*
* Construct a znode+inode and initialize.
*
* This does not do a call to dmu_set_user() that is
* up to the caller to do, in case you don't want to
* return the znode
*/
static znode_t *
zfs_znode_alloc(zfs_sb_t *zsb, dmu_buf_t *db, int blksz,
dmu_object_type_t obj_type, uint64_t obj, sa_handle_t *hdl,
struct inode *dip)
{
znode_t *zp;
struct inode *ip;
uint64_t mode;
uint64_t parent;
sa_bulk_attr_t bulk[9];
int count = 0;
ASSERT(zsb != NULL);
ip = new_inode(zsb->z_sb);
if (ip == NULL)
return (NULL);
zp = ITOZ(ip);
ASSERT(zp->z_dirlocks == NULL);
ASSERT3P(zp->z_acl_cached, ==, NULL);
ASSERT3P(zp->z_xattr_cached, ==, NULL);
ASSERT3P(zp->z_xattr_parent, ==, NULL);
zp->z_moved = 0;
zp->z_sa_hdl = NULL;
zp->z_unlinked = 0;
zp->z_atime_dirty = 0;
zp->z_mapcnt = 0;
zp->z_id = db->db_object;
zp->z_blksz = blksz;
zp->z_seq = 0x7A4653;
zp->z_sync_cnt = 0;
zp->z_is_zvol = B_FALSE;
zp->z_is_mapped = B_FALSE;
zp->z_is_ctldir = B_FALSE;
zp->z_is_stale = B_FALSE;
zfs_znode_sa_init(zsb, zp, db, obj_type, hdl);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL, &mode, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zsb), NULL, &zp->z_gen, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), NULL, &zp->z_size, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zsb), NULL, &zp->z_links, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL,
&zp->z_pflags, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zsb), NULL,
&parent, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zsb), NULL,
&zp->z_atime, 16);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL, &zp->z_uid, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), NULL, &zp->z_gid, 8);
if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || zp->z_gen == 0) {
if (hdl == NULL)
sa_handle_destroy(zp->z_sa_hdl);
goto error;
}
zp->z_mode = mode;
/*
* xattr znodes hold a reference on their unique parent
*/
if (dip && zp->z_pflags & ZFS_XATTR) {
igrab(dip);
zp->z_xattr_parent = ITOZ(dip);
}
ip->i_ino = obj;
zfs_inode_update(zp);
zfs_inode_set_ops(zsb, ip);
/*
* The only way insert_inode_locked() can fail is if the ip->i_ino
* number is already hashed for this super block. This can never
* happen because the inode numbers map 1:1 with the object numbers.
*
* The one exception is rolling back a mounted file system, but in
* this case all the active inode are unhashed during the rollback.
*/
VERIFY3S(insert_inode_locked(ip), ==, 0);
mutex_enter(&zsb->z_znodes_lock);
list_insert_tail(&zsb->z_all_znodes, zp);
zsb->z_nr_znodes++;
membar_producer();
mutex_exit(&zsb->z_znodes_lock);
unlock_new_inode(ip);
return (zp);
error:
unlock_new_inode(ip);
iput(ip);
return (NULL);
}
/*
* Link zp into dl. Can only fail if zp has been unlinked.
*/
int
zfs_link_create(zfs_dirlock_t *dl, znode_t *zp, dmu_tx_t *tx, int flag)
{
znode_t *dzp = dl->dl_dzp;
zfsvfs_t *zfsvfs = zp->z_zfsvfs;
#ifdef __APPLE__
uint64_t value;
/* OSX - don't access the vnode here since it might not be attached yet. */
//int zp_is_dir = S_ISDIR(zp->z_phys->zp_mode);
int zp_is_dir = S_ISDIR(zp->z_mode);
#else
vnode_t *vp = ZTOV(zp);
uint64_t value;
int zp_is_dir = (vp->v_type == VDIR);
#endif
sa_bulk_attr_t bulk[5];
uint64_t mtime[2], ctime[2];
int count = 0;
int error;
mutex_enter(&zp->z_lock);
if (!(flag & ZRENAMING)) {
if (zp->z_unlinked) { /* no new links to unlinked zp */
ASSERT(!(flag & (ZNEW | ZEXISTS)));
mutex_exit(&zp->z_lock);
return (ENOENT);
}
zp->z_links++;
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
&zp->z_links, sizeof (zp->z_links));
}
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL,
&dzp->z_id, sizeof (dzp->z_id));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
&zp->z_pflags, sizeof (zp->z_pflags));
if (!(flag & ZNEW)) {
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
ctime, sizeof (ctime));
zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime,
ctime, B_TRUE);
}
error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
// Needed?
#ifdef __APPLE__
zp->z_parent = dzp->z_id;
#endif
mutex_exit(&zp->z_lock);
mutex_enter(&dzp->z_lock);
dzp->z_size++;
dzp->z_links += zp_is_dir;
count = 0;
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
&dzp->z_size, sizeof (dzp->z_size));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
&dzp->z_links, sizeof (dzp->z_links));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL,
mtime, sizeof (mtime));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
ctime, sizeof (ctime));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
&dzp->z_pflags, sizeof (dzp->z_pflags));
zfs_tstamp_update_setup(dzp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
error = sa_bulk_update(dzp->z_sa_hdl, bulk, count, tx);
mutex_exit(&dzp->z_lock);
value = zfs_dirent(zp, zp->z_mode);
error = zap_add(zp->z_zfsvfs->z_os, dzp->z_id, dl->dl_name,
8, 1, &value, tx);
ASSERT(error == 0);
#ifndef __APPLE__
/* On Mac OS X, this is done up in VFS layer. */
dnlc_update(ZTOV(dzp), dl->dl_name, vp);
#endif
return (0);
}
int
zfs_make_xattrdir(znode_t *zp, vattr_t *vap, vnode_t **xvpp, cred_t *cr)
{
zfsvfs_t *zfsvfs = zp->z_zfsvfs;
znode_t *xzp;
dmu_tx_t *tx;
int error;
zfs_acl_ids_t acl_ids;
boolean_t fuid_dirtied;
uint64_t parent;
*xvpp = NULL;
/*
* In FreeBSD, access checking for creating an EA is being done
* in zfs_setextattr(),
*/
#ifndef __FreeBSD_kernel__
if (error = zfs_zaccess(zp, ACE_WRITE_NAMED_ATTRS, 0, B_FALSE, cr))
return (error);
#endif
if ((error = zfs_acl_ids_create(zp, IS_XATTR, vap, cr, NULL,
&acl_ids)) != 0)
return (error);
if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) {
zfs_acl_ids_free(&acl_ids);
return (SET_ERROR(EDQUOT));
}
getnewvnode_reserve(1);
tx = dmu_tx_create(zfsvfs->z_os);
dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes +
ZFS_SA_BASE_ATTR_SIZE);
dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
fuid_dirtied = zfsvfs->z_fuid_dirty;
if (fuid_dirtied)
zfs_fuid_txhold(zfsvfs, tx);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
zfs_acl_ids_free(&acl_ids);
dmu_tx_abort(tx);
return (error);
}
zfs_mknode(zp, vap, tx, cr, IS_XATTR, &xzp, &acl_ids);
if (fuid_dirtied)
zfs_fuid_sync(zfsvfs, tx);
#ifdef DEBUG
error = sa_lookup(xzp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs),
&parent, sizeof (parent));
ASSERT(error == 0 && parent == zp->z_id);
#endif
VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), &xzp->z_id,
sizeof (xzp->z_id), tx));
(void) zfs_log_create(zfsvfs->z_log, tx, TX_MKXATTR, zp,
xzp, "", NULL, acl_ids.z_fuidp, vap);
zfs_acl_ids_free(&acl_ids);
dmu_tx_commit(tx);
getnewvnode_drop_reserve();
*xvpp = ZTOV(xzp);
return (0);
}
/*
* Look up an entry in a directory.
*
* NOTE: '.' and '..' are handled as special cases because
* no directory entries are actually stored for them. If this is
* the root of a filesystem, then '.zfs' is also treated as a
* special pseudo-directory.
*/
int
zfs_dirlook(znode_t *dzp, char *name, vnode_t **vpp, int flags,
int *deflg, pathname_t *rpnp)
{
zfs_dirlock_t *dl;
znode_t *zp;
int error = 0;
uint64_t parent;
if (name[0] == 0 || (name[0] == '.' && name[1] == 0)) {
vnode_t *vn = ZTOV(dzp);
if (VN_HOLD(vn) == NULL)
error = EAGAIN;
else
*vpp = vn;
} else if (name[0] == '.' && name[1] == '.' && name[2] == 0) {
zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
/*
* If we are a snapshot mounted under .zfs, return
* the vp for the snapshot directory.
*/
if ((error = sa_lookup(dzp->z_sa_hdl,
SA_ZPL_PARENT(zfsvfs), &parent, sizeof (parent))) != 0)
return (error);
if (parent == dzp->z_id && zfsvfs->z_parent != zfsvfs) {
error = zfsctl_root_lookup(zfsvfs->z_parent->z_ctldir,
"snapshot", vpp, NULL, 0, NULL, kcred,
NULL, NULL, NULL);
return (error);
}
rw_enter(&dzp->z_parent_lock, RW_READER);
error = zfs_zget(zfsvfs, parent, &zp);
if (error == 0)
*vpp = ZTOV(zp);
rw_exit(&dzp->z_parent_lock);
} else if (zfs_has_ctldir(dzp) && strcmp(name, ZFS_CTLDIR_NAME) == 0) {
vnode_t *vn = zfsctl_root(dzp);
if (vn)
*vpp = vn;
else
error = EAGAIN;
} else {
int zf;
zf = ZEXISTS | ZSHARED;
if (flags & FIGNORECASE)
zf |= ZCILOOK;
error = zfs_dirent_lock(&dl, dzp, name, &zp, zf, deflg, rpnp);
if (error == 0) {
*vpp = ZTOV(zp);
zfs_dirent_unlock(dl);
dzp->z_zn_prefetch = B_TRUE; /* enable prefetching */
}
rpnp = NULL;
}
if ((flags & FIGNORECASE) && rpnp && !error)
(void) strlcpy(rpnp->pn_buf, name, rpnp->pn_bufsize);
return (error);
}
void commonattrpack(attrinfo_t *aip, zfsvfs_t *zfsvfs, znode_t *zp,
const char *name, ino64_t objnum, enum vtype vtype,
boolean_t user64)
{
attrgroup_t commonattr = aip->ai_attrlist->commonattr;
void *attrbufptr = *aip->ai_attrbufpp;
void *varbufptr = *aip->ai_varbufpp;
struct mount *mp = zfsvfs->z_vfs;
cred_t *cr = (cred_t *)vfs_context_ucred(aip->ai_context);
finderinfo_t finderinfo;
/*
* We should probably combine all the sa_lookup into a bulk
* lookup operand.
*/
finderinfo.fi_flags = 0;
if (ATTR_CMN_NAME & commonattr) {
nameattrpack(aip, name, strlen(name));
attrbufptr = *aip->ai_attrbufpp;
varbufptr = *aip->ai_varbufpp;
}
if (ATTR_CMN_DEVID & commonattr) {
*((dev_t *)attrbufptr) = vfs_statfs(mp)->f_fsid.val[0];
attrbufptr = ((dev_t *)attrbufptr) + 1;
}
if (ATTR_CMN_FSID & commonattr) {
*((fsid_t *)attrbufptr) = vfs_statfs(mp)->f_fsid;
attrbufptr = ((fsid_t *)attrbufptr) + 1;
}
if (ATTR_CMN_OBJTYPE & commonattr) {
*((fsobj_type_t *)attrbufptr) = vtype;
attrbufptr = ((fsobj_type_t *)attrbufptr) + 1;
}
if (ATTR_CMN_OBJTAG & commonattr) {
*((fsobj_tag_t *)attrbufptr) = VT_ZFS;
attrbufptr = ((fsobj_tag_t *)attrbufptr) + 1;
}
/*
* Note: ATTR_CMN_OBJID is lossy (only 32 bits).
*/
if ((ATTR_CMN_OBJID | ATTR_CMN_OBJPERMANENTID) & commonattr) {
u_int32_t fileid;
/*
* On Mac OS X we always export the root directory id as 2
*/
fileid = (objnum == zfsvfs->z_root) ? 2 : objnum;
if (ATTR_CMN_OBJID & commonattr) {
((fsobj_id_t *)attrbufptr)->fid_objno = fileid;
((fsobj_id_t *)attrbufptr)->fid_generation = 0;
attrbufptr = ((fsobj_id_t *)attrbufptr) + 1;
}
if (ATTR_CMN_OBJPERMANENTID & commonattr) {
((fsobj_id_t *)attrbufptr)->fid_objno = fileid;
((fsobj_id_t *)attrbufptr)->fid_generation = 0;
attrbufptr = ((fsobj_id_t *)attrbufptr) + 1;
}
}
/*
* Note: ATTR_CMN_PAROBJID is lossy (only 32 bits).
*/
if (ATTR_CMN_PAROBJID & commonattr) {
uint64_t parentid;
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs),
&parentid, sizeof (parentid)) == 0);
/*
* On Mac OS X we always export the root
* directory id as 2 and its parent as 1
*/
if (zp && zp->z_id == zfsvfs->z_root)
parentid = 1;
else if (parentid == zfsvfs->z_root)
parentid = 2;
ASSERT(parentid != 0);
((fsobj_id_t *)attrbufptr)->fid_objno = (uint32_t)parentid;
((fsobj_id_t *)attrbufptr)->fid_generation = 0;
attrbufptr = ((fsobj_id_t *)attrbufptr) + 1;
}
if (ATTR_CMN_SCRIPT & commonattr) {
*((text_encoding_t *)attrbufptr) = kTextEncodingMacUnicode;
attrbufptr = ((text_encoding_t *)attrbufptr) + 1;
}
if (ATTR_CMN_CRTIME & commonattr) {
uint64_t times[2];
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(zfsvfs),
times, sizeof(times)) == 0);
if (user64) {
ZFS_TIME_DECODE((timespec_user64_t *)attrbufptr,
times);
attrbufptr = ((timespec_user64_t *)attrbufptr) + 1;
} else {
ZFS_TIME_DECODE((timespec_user32_t *)attrbufptr,
times);
attrbufptr = ((timespec_user32_t *)attrbufptr) + 1;
}
}
if (ATTR_CMN_MODTIME & commonattr) {
uint64_t times[2];
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_MTIME(zfsvfs),
times, sizeof(times)) == 0);
if (user64) {
ZFS_TIME_DECODE((timespec_user64_t *)attrbufptr,
times);
attrbufptr = ((timespec_user64_t *)attrbufptr) + 1;
} else {
ZFS_TIME_DECODE((timespec_user32_t *)attrbufptr,
times);
attrbufptr = ((timespec_user32_t *)attrbufptr) + 1;
}
}
if (ATTR_CMN_CHGTIME & commonattr) {
uint64_t times[2];
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_CTIME(zfsvfs),
times, sizeof(times)) == 0);
if (user64) {
ZFS_TIME_DECODE((timespec_user64_t *)attrbufptr,
times);
attrbufptr = ((timespec_user64_t *)attrbufptr) + 1;
} else {
ZFS_TIME_DECODE((timespec_user32_t *)attrbufptr,
times);
attrbufptr = ((timespec_user32_t *)attrbufptr) + 1;
}
}
if (ATTR_CMN_ACCTIME & commonattr) {
uint64_t times[2];
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_ATIME(zfsvfs),
times, sizeof(times)) == 0);
if (user64) {
ZFS_TIME_DECODE((timespec_user64_t *)attrbufptr,
times);
attrbufptr = ((timespec_user64_t *)attrbufptr) + 1;
} else {
ZFS_TIME_DECODE((timespec_user32_t *)attrbufptr,
times);
attrbufptr = ((timespec_user32_t *)attrbufptr) + 1;
}
}
if (ATTR_CMN_BKUPTIME & commonattr) {
/* legacy attribute -- just pass zero */
if (user64) {
((timespec_user64_t *)attrbufptr)->tv_sec = 0;
((timespec_user64_t *)attrbufptr)->tv_nsec = 0;
attrbufptr = ((timespec_user64_t *)attrbufptr) + 1;
} else {
((timespec_user32_t *)attrbufptr)->tv_sec = 0;
((timespec_user32_t *)attrbufptr)->tv_nsec = 0;
attrbufptr = ((timespec_user32_t *)attrbufptr) + 1;
}
}
if (ATTR_CMN_FNDRINFO & commonattr) {
uint64_t val;
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_FLAGS(zfsvfs),
&val, sizeof(val)) == 0);
getfinderinfo(zp, cr, &finderinfo);
/* Shadow ZFS_HIDDEN to Finder Info's invisible bit */
if (val & ZFS_HIDDEN) {
finderinfo.fi_flags |=
OSSwapHostToBigConstInt16(kIsInvisible);
}
bcopy(&finderinfo, attrbufptr, sizeof (finderinfo));
attrbufptr = (char *)attrbufptr + 32;
}
if (ATTR_CMN_OWNERID & commonattr) {
uint64_t val;
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_UID(zfsvfs),
&val, sizeof(val)) == 0);
*((uid_t *)attrbufptr) = val;
attrbufptr = ((uid_t *)attrbufptr) + 1;
}
if (ATTR_CMN_GRPID & commonattr) {
uint64_t val;
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_GID(zfsvfs),
&val, sizeof(val)) == 0);
*((gid_t *)attrbufptr) = val;
attrbufptr = ((gid_t *)attrbufptr) + 1;
}
if (ATTR_CMN_ACCESSMASK & commonattr) {
uint64_t val;
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs),
&val, sizeof(val)) == 0);
*((u_int32_t *)attrbufptr) = val;
attrbufptr = ((u_int32_t *)attrbufptr) + 1;
}
if (ATTR_CMN_FLAGS & commonattr) {
u_int32_t flags = zfs_getbsdflags(zp);
/* Shadow Finder Info's invisible bit to UF_HIDDEN */
if ((ATTR_CMN_FNDRINFO & commonattr) &&
(OSSwapBigToHostInt16(finderinfo.fi_flags) & kIsInvisible))
flags |= UF_HIDDEN;
*((u_int32_t *)attrbufptr) = flags;
attrbufptr = ((u_int32_t *)attrbufptr) + 1;
}
if (ATTR_CMN_USERACCESS & commonattr) {
u_int32_t user_access = 0;
uint64_t val;
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_FLAGS(zfsvfs),
&val, sizeof(val)) == 0);
user_access = getuseraccess(zp, aip->ai_context);
/* Also consider READ-ONLY file system. */
if (vfs_flags(mp) & MNT_RDONLY) {
user_access &= ~W_OK;
}
/* Locked objects are not writable either */
if ((val & ZFS_IMMUTABLE) &&
(vfs_context_suser(aip->ai_context) != 0)) {
user_access &= ~W_OK;
}
*((u_int32_t *)attrbufptr) = user_access;
attrbufptr = ((u_int32_t *)attrbufptr) + 1;
}
if (ATTR_CMN_FILEID & commonattr) {
/*
* On Mac OS X we always export the root directory id as 2
*/
if (objnum == zfsvfs->z_root)
objnum = 2;
*((u_int64_t *)attrbufptr) = objnum;
attrbufptr = ((u_int64_t *)attrbufptr) + 1;
}
if (ATTR_CMN_PARENTID & commonattr) {
uint64_t parentid;
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs),
&parentid, sizeof (parentid)) == 0);
/*
* On Mac OS X we always export the root
* directory id as 2 and its parent as 1
*/
if (zp && zp->z_id == zfsvfs->z_root)
parentid = 1;
else if (parentid == zfsvfs->z_root)
parentid = 2;
ASSERT(parentid != 0);
*((u_int64_t *)attrbufptr) = parentid;
attrbufptr = ((u_int64_t *)attrbufptr) + 1;
}
*aip->ai_attrbufpp = attrbufptr;
*aip->ai_varbufpp = varbufptr;
}
/* For part 1 of zfs_getattr() */
int
zfs_getattr_znode_locked(vattr_t *vap, znode_t *zp, cred_t *cr)
{
zfsvfs_t *zfsvfs = zp->z_zfsvfs;
int error;
uint64_t times[2];
uint64_t val;
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs),
&val, sizeof (val)) == 0);
vap->va_mode = val & MODEMASK;
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_UID(zfsvfs),
&val, sizeof (val)) == 0);
vap->va_uid = val;
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_GID(zfsvfs),
&val, sizeof (val)) == 0);
vap->va_gid = val;
//vap->va_fsid = zp->z_zfsvfs->z_vfs->vfs_dev;
/* On OS X, the root directory id is always 2 */
vap->va_fileid = (zp->z_id == zfsvfs->z_root) ? 2 : zp->z_id;
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_LINKS(zfsvfs),
&val, sizeof (val)) == 0);
vap->va_nlink = val;
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs),
&val, sizeof (val)) == 0);
vap->va_data_size = val;
vap->va_total_size = val;
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_RDEV(zfsvfs),
&val, sizeof (val)) == 0);
vap->va_rdev = val;
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zfsvfs),
&val, sizeof (val)) == 0);
vap->va_gen = val;
(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(zfsvfs),
times, sizeof (times));
ZFS_TIME_DECODE(&vap->va_create_time, times);
(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_ATIME(zfsvfs),
times, sizeof (times));
ZFS_TIME_DECODE(&vap->va_access_time, times);
(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_MTIME(zfsvfs),
times, sizeof (times));
ZFS_TIME_DECODE(&vap->va_modify_time, times);
(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CTIME(zfsvfs),
times, sizeof (times));
ZFS_TIME_DECODE(&vap->va_change_time, times);
if (VATTR_IS_ACTIVE(vap, va_backup_time)) {
vap->va_backup_time.tv_sec = 0;
vap->va_backup_time.tv_nsec = 0;
VATTR_SET_SUPPORTED(vap, va_backup_time);
}
vap->va_flags = zfs_getbsdflags(zp);
/* On OS X, the root directory id is always 2 and its parent is 1 */
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs),
&val, sizeof (val)) == 0);
if (zp->z_id == zfsvfs->z_root)
vap->va_parentid = 1;
else if (val == zfsvfs->z_root)
vap->va_parentid = 2;
else
vap->va_parentid = val;
vap->va_iosize = zp->z_blksz ? zp->z_blksz : zfsvfs->z_max_blksz;
VATTR_SET_SUPPORTED(vap, va_iosize);
printf("stat blksize set to %d\n", vap->va_iosize);
vap->va_supported |= ZFS_SUPPORTED_VATTRS;
if (VATTR_IS_ACTIVE(vap, va_nchildren) && vnode_isdir(ZTOV(zp)))
VATTR_RETURN(vap, va_nchildren, vap->va_nlink - 2);
if (VATTR_IS_ACTIVE(vap, va_acl)) {
if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(zfsvfs),
times, sizeof (times)))) {
// if (zp->z_phys->zp_acl.z_acl_count == 0) {
vap->va_acl = (kauth_acl_t) KAUTH_FILESEC_NONE;
} else {
error = zfs_getacl(zp, &vap->va_acl, B_TRUE, cr);
if (error)
return (error);
VATTR_SET_SUPPORTED(vap, va_acl);
/*
* va_acl implies that va_uuuid and va_guuid are
* also supported.
*/
VATTR_RETURN(vap, va_uuuid, kauth_null_guid);
VATTR_RETURN(vap, va_guuid, kauth_null_guid);
}
}
return (0);
}
int
zfs_getattr_znode_unlocked(struct vnode *vp, vattr_t *vap)
{
znode_t *zp = VTOZ(vp);
zfsvfs_t *zfsvfs = zp->z_zfsvfs;
int error = 0;
uint64_t parent;
//printf("getattr_osx\n");
ZFS_ENTER(zfsvfs);
/*
* On Mac OS X we always export the root directory id as 2
*/
vap->va_fileid = (zp->z_id == zfsvfs->z_root) ? 2 : zp->z_id;
//vap->va_fileid = (zp->z_id == zfsvfs->z_root) ? 2 : zp->z_vid;
vap->va_nlink = zp->z_links;
vap->va_data_size = zp->z_size;
vap->va_total_size = zp->z_size;
vap->va_gen = zp->z_gen;
/*
* For Carbon compatibility,pretend to support this legacy/unused attribute
*/
if (VATTR_IS_ACTIVE(vap, va_backup_time)) {
vap->va_backup_time.tv_sec = 0;
vap->va_backup_time.tv_nsec = 0;
VATTR_SET_SUPPORTED(vap, va_backup_time);
}
vap->va_flags = zfs_getbsdflags(zp);
/*
* On Mac OS X we always export the root directory id as 2
* and its parent as 1
*/
error = sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs),
&parent, sizeof (parent));
if (!error) {
if (zp->z_id == zfsvfs->z_root)
vap->va_parentid = 1;
else if (parent == zfsvfs->z_root)
vap->va_parentid = 2;
else
vap->va_parentid = parent;
}
vap->va_iosize = zp->z_blksz ? zp->z_blksz : zfsvfs->z_max_blksz;
//vap->va_iosize = 512;
VATTR_SET_SUPPORTED(vap, va_iosize);
/* Don't include '.' and '..' in the number of entries */
if (VATTR_IS_ACTIVE(vap, va_nchildren) && vnode_isdir(vp)) {
VATTR_RETURN(vap, va_nchildren, vap->va_nlink - 2);
}
/*
* va_dirlinkcount is the count of directory hard links. When a file
* system does not support ATTR_DIR_LINKCOUNT, xnu will default to 1.
* Since we claim to support ATTR_DIR_LINKCOUNT both as valid and as
* native, we'll just return 1. We set 1 for this value in dirattrpack
* as well. If in the future ZFS actually supports directory hard links,
* we can return a real value.
*/
if (VATTR_IS_ACTIVE(vap, va_dirlinkcount) && vnode_isdir(vp)) {
VATTR_RETURN(vap, va_dirlinkcount, 1);
}
if (VATTR_IS_ACTIVE(vap, va_acl)) {
//printf("want acl\n");
#if 0
zfs_acl_phys_t acl;
if (sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(zfsvfs),
&acl, sizeof (zfs_acl_phys_t))) {
//if (zp->z_acl.z_acl_count == 0) {
vap->va_acl = (kauth_acl_t) KAUTH_FILESEC_NONE;
} else {
if ((error = zfs_getacl(zp, &vap->va_acl, B_TRUE, NULL))) {
dprintf("zfs_getacl returned error %d\n", error);
error = 0;
//ZFS_EXIT(zfsvfs);
//return (error);
}
}
#endif
//VATTR_SET_SUPPORTED(vap, va_acl);
VATTR_RETURN(vap, va_uuuid, kauth_null_guid);
VATTR_RETURN(vap, va_guuid, kauth_null_guid);
dprintf("Calling getacl\n");
if ((error = zfs_getacl(zp, &vap->va_acl, B_FALSE, NULL))) {
dprintf("zfs_getacl returned error %d\n", error);
error = 0;
} else {
VATTR_SET_SUPPORTED(vap, va_acl);
/* va_acl implies that va_uuuid and va_guuid are also supported. */
VATTR_RETURN(vap, va_uuuid, kauth_null_guid);
VATTR_RETURN(vap, va_guuid, kauth_null_guid);
}
}
if (VATTR_IS_ACTIVE(vap, va_data_alloc) || VATTR_IS_ACTIVE(vap, va_total_alloc)) {
uint32_t blksize;
u_longlong_t nblks;
sa_object_size(zp->z_sa_hdl, &blksize, &nblks);
vap->va_data_alloc = (uint64_t)512LL * (uint64_t)nblks;
vap->va_total_alloc = vap->va_data_alloc;
vap->va_supported |= VNODE_ATTR_va_data_alloc |
VNODE_ATTR_va_total_alloc;
}
if (VATTR_IS_ACTIVE(vap, va_name)) {
vap->va_name[0] = 0;
if (!vnode_isvroot(vp)) {
/* Lets not supply name as zap_cursor can cause panic */
#if 0
if (zap_value_search(zfsvfs->z_os, parent, zp->z_id,
ZFS_DIRENT_OBJ(-1ULL), vap->va_name) == 0)
VATTR_SET_SUPPORTED(vap, va_name);
#endif
} else {
/*
* The vroot objects must return a unique name for Finder to
* be able to distringuish between mounts. For this reason
* we simply return the fullname, from the statfs mountedfrom
*/
strlcpy(vap->va_name,
vfs_statfs(vnode_mount(vp))->f_mntfromname,
MAXPATHLEN);
VATTR_SET_SUPPORTED(vap, va_name);
}
}
if (VATTR_IS_ACTIVE(vap, va_filerev)) {
VATTR_RETURN(vap, va_filerev, 0);
}
if (VATTR_IS_ACTIVE(vap, va_linkid)) {
VATTR_RETURN(vap, va_linkid, vap->va_fileid);
}
if (VATTR_IS_ACTIVE(vap, va_fsid)) {
VATTR_RETURN(vap, va_fsid, vfs_statfs(zfsvfs->z_vfs)->f_fsid.val[0]);
}
if (VATTR_IS_ACTIVE(vap, va_type)) {
VATTR_RETURN(vap, va_type, vnode_vtype(ZTOV(zp)));
}
if (VATTR_IS_ACTIVE(vap, va_encoding)) {
VATTR_RETURN(vap, va_encoding, kTextEncodingMacUnicode);
}
#ifdef VNODE_ATTR_va_addedtime
if (VATTR_IS_ACTIVE(vap, va_addedtime)) {
VATTR_RETURN(vap, va_addedtime, vap->va_ctime);
}
#endif
if (VATTR_IS_ACTIVE(vap, va_uuuid)) {
kauth_cred_uid2guid(zp->z_uid, &vap->va_uuuid);
}
if (VATTR_IS_ACTIVE(vap, va_guuid)) {
kauth_cred_uid2guid(zp->z_gid, &vap->va_guuid);
}
vap->va_supported |= ZFS_SUPPORTED_VATTRS;
ZFS_EXIT(zfsvfs);
return (error);
}
/*
* as we don't know FID, we can't use LU object, so this function
* partially duplicate __osd_xattr_get() which is built around
* LU-object and uses it to cache data like regular EA dnode, etc
*/
static int osd_find_parent_by_dnode(const struct lu_env *env,
struct dt_object *o,
struct lu_fid *fid)
{
struct lustre_mdt_attrs *lma;
udmu_objset_t *uos = &osd_obj2dev(osd_dt_obj(o))->od_objset;
struct lu_buf buf;
sa_handle_t *sa_hdl;
nvlist_t *nvbuf = NULL;
uchar_t *value;
uint64_t dnode;
int rc, size;
ENTRY;
/* first of all, get parent dnode from own attributes */
LASSERT(osd_dt_obj(o)->oo_db);
rc = -sa_handle_get(uos->os, osd_dt_obj(o)->oo_db->db_object,
NULL, SA_HDL_PRIVATE, &sa_hdl);
if (rc)
RETURN(rc);
dnode = ZFS_NO_OBJECT;
rc = -sa_lookup(sa_hdl, SA_ZPL_PARENT(uos), &dnode, 8);
sa_handle_destroy(sa_hdl);
if (rc)
RETURN(rc);
/* now get EA buffer */
rc = __osd_xattr_load(uos, dnode, &nvbuf);
if (rc)
GOTO(regular, rc);
/* XXX: if we get that far.. should we cache the result? */
/* try to find LMA attribute */
LASSERT(nvbuf != NULL);
rc = -nvlist_lookup_byte_array(nvbuf, XATTR_NAME_LMA, &value, &size);
if (rc == 0 && size >= sizeof(*lma)) {
lma = (struct lustre_mdt_attrs *)value;
lustre_lma_swab(lma);
*fid = lma->lma_self_fid;
GOTO(out, rc = 0);
}
regular:
/* no LMA attribute in SA, let's try regular EA */
/* first of all, get parent dnode storing regular EA */
rc = -sa_handle_get(uos->os, dnode, NULL, SA_HDL_PRIVATE, &sa_hdl);
if (rc)
GOTO(out, rc);
dnode = ZFS_NO_OBJECT;
rc = -sa_lookup(sa_hdl, SA_ZPL_XATTR(uos), &dnode, 8);
sa_handle_destroy(sa_hdl);
if (rc)
GOTO(out, rc);
CLASSERT(sizeof(*lma) <= sizeof(osd_oti_get(env)->oti_buf));
buf.lb_buf = osd_oti_get(env)->oti_buf;
buf.lb_len = sizeof(osd_oti_get(env)->oti_buf);
/* now try to find LMA */
rc = __osd_xattr_get_large(env, uos, dnode, &buf,
XATTR_NAME_LMA, &size);
if (rc == 0 && size >= sizeof(*lma)) {
lma = buf.lb_buf;
lustre_lma_swab(lma);
*fid = lma->lma_self_fid;
GOTO(out, rc = 0);
} else if (rc < 0) {
GOTO(out, rc);
} else {
GOTO(out, rc = -EIO);
}
out:
if (nvbuf != NULL)
nvlist_free(nvbuf);
RETURN(rc);
}
/**
* Inserts (key, value) pair in \a directory object.
*
* \param dt osd index object
* \param key key for index
* \param rec record reference
* \param th transaction handler
* \param capa capability descriptor
* \param ignore_quota update should not affect quota
*
* \retval 0 success
* \retval -ve failure
*/
static int osd_dir_insert(const struct lu_env *env, struct dt_object *dt,
const struct dt_rec *rec, const struct dt_key *key,
struct thandle *th, struct lustre_capa *capa,
int ignore_quota)
{
struct osd_thread_info *oti = osd_oti_get(env);
struct osd_object *parent = osd_dt_obj(dt);
struct osd_device *osd = osd_obj2dev(parent);
struct lu_fid *fid = (struct lu_fid *)rec;
struct osd_thandle *oh;
struct osd_object *child = NULL;
__u32 attr;
char *name = (char *)key;
int rc;
ENTRY;
LASSERT(parent->oo_db);
LASSERT(udmu_object_is_zap(parent->oo_db));
LASSERT(dt_object_exists(dt));
LASSERT(osd_invariant(parent));
LASSERT(th != NULL);
oh = container_of0(th, struct osd_thandle, ot_super);
rc = osd_remote_fid(env, osd, fid);
if (rc < 0) {
CERROR("%s: Can not find object "DFID": rc = %d\n",
osd->od_svname, PFID(fid), rc);
RETURN(rc);
}
if (unlikely(rc == 1)) {
/* Insert remote entry */
memset(&oti->oti_zde.lzd_reg, 0, sizeof(oti->oti_zde.lzd_reg));
oti->oti_zde.lzd_reg.zde_type = IFTODT(S_IFDIR & S_IFMT);
} else {
/*
* To simulate old Orion setups with ./.. stored in the
* directories
*/
/* Insert local entry */
child = osd_object_find(env, dt, fid);
if (IS_ERR(child))
RETURN(PTR_ERR(child));
LASSERT(child->oo_db);
if (name[0] == '.') {
if (name[1] == 0) {
/* do not store ".", instead generate it
* during iteration */
GOTO(out, rc = 0);
} else if (name[1] == '.' && name[2] == 0) {
/* update parent dnode in the child.
* later it will be used to generate ".." */
udmu_objset_t *uos = &osd->od_objset;
rc = osd_object_sa_update(parent,
SA_ZPL_PARENT(uos),
&child->oo_db->db_object,
8, oh);
GOTO(out, rc);
}
}
CLASSERT(sizeof(oti->oti_zde.lzd_reg) == 8);
CLASSERT(sizeof(oti->oti_zde) % 8 == 0);
attr = child->oo_dt.do_lu.lo_header ->loh_attr;
oti->oti_zde.lzd_reg.zde_type = IFTODT(attr & S_IFMT);
oti->oti_zde.lzd_reg.zde_dnode = child->oo_db->db_object;
}
oti->oti_zde.lzd_fid = *fid;
/* Insert (key,oid) into ZAP */
rc = -zap_add(osd->od_objset.os, parent->oo_db->db_object,
(char *)key, 8, sizeof(oti->oti_zde) / 8,
(void *)&oti->oti_zde, oh->ot_tx);
out:
if (child != NULL)
osd_object_put(env, child);
RETURN(rc);
}
/*
* Link zp into dl. Can only fail if zp has been unlinked.
*/
int
zfs_link_create(zfs_dirlock_t *dl, znode_t *zp, dmu_tx_t *tx, int flag)
{
znode_t *dzp = dl->dl_dzp;
zfsvfs_t *zfsvfs = zp->z_zfsvfs;
vnode_t *vp = ZTOV(zp);
uint64_t value;
int zp_is_dir = (vp->v_type == VDIR);
sa_bulk_attr_t bulk[5];
uint64_t mtime[2], ctime[2];
int count = 0;
int error;
mutex_enter(&zp->z_lock);
if (!(flag & ZRENAMING)) {
if (zp->z_unlinked) { /* no new links to unlinked zp */
ASSERT(!(flag & (ZNEW | ZEXISTS)));
mutex_exit(&zp->z_lock);
return (ENOENT);
}
zp->z_links++;
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
&zp->z_links, sizeof (zp->z_links));
}
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL,
&dzp->z_id, sizeof (dzp->z_id));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
&zp->z_pflags, sizeof (zp->z_pflags));
if (!(flag & ZNEW)) {
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
ctime, sizeof (ctime));
zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime,
ctime, B_TRUE);
}
error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
ASSERT(error == 0);
mutex_exit(&zp->z_lock);
mutex_enter(&dzp->z_lock);
dzp->z_size++;
dzp->z_links += zp_is_dir;
count = 0;
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
&dzp->z_size, sizeof (dzp->z_size));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
&dzp->z_links, sizeof (dzp->z_links));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL,
mtime, sizeof (mtime));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
ctime, sizeof (ctime));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
&dzp->z_pflags, sizeof (dzp->z_pflags));
zfs_tstamp_update_setup(dzp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
error = sa_bulk_update(dzp->z_sa_hdl, bulk, count, tx);
ASSERT(error == 0);
mutex_exit(&dzp->z_lock);
value = zfs_dirent(zp, zp->z_mode);
error = zap_add(zp->z_zfsvfs->z_os, dzp->z_id, dl->dl_name,
8, 1, &value, tx);
ASSERT(error == 0);
dnlc_update(ZTOV(dzp), dl->dl_name, vp);
return (0);
}
/*
* Construct a znode+inode and initialize.
*
* This does not do a call to dmu_set_user() that is
* up to the caller to do, in case you don't want to
* return the znode
*/
static znode_t *
zfs_znode_alloc(zfs_sb_t *zsb, dmu_buf_t *db, int blksz,
dmu_object_type_t obj_type, uint64_t obj, sa_handle_t *hdl,
struct dentry *dentry, struct inode *dip)
{
znode_t *zp;
struct inode *ip;
uint64_t parent;
sa_bulk_attr_t bulk[9];
int count = 0;
ASSERT(zsb != NULL);
ip = new_inode(zsb->z_sb);
if (ip == NULL)
return (NULL);
zp = ITOZ(ip);
ASSERT(zp->z_dirlocks == NULL);
ASSERT3P(zp->z_acl_cached, ==, NULL);
ASSERT3P(zp->z_xattr_cached, ==, NULL);
zp->z_moved = 0;
zp->z_sa_hdl = NULL;
zp->z_unlinked = 0;
zp->z_atime_dirty = 0;
zp->z_mapcnt = 0;
zp->z_id = db->db_object;
zp->z_blksz = blksz;
zp->z_seq = 0x7A4653;
zp->z_sync_cnt = 0;
zp->z_is_zvol = B_FALSE;
zp->z_is_mapped = B_FALSE;
zp->z_is_ctldir = B_FALSE;
zfs_znode_sa_init(zsb, zp, db, obj_type, hdl);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL, &zp->z_mode, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zsb), NULL, &zp->z_gen, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), NULL, &zp->z_size, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zsb), NULL, &zp->z_links, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL,
&zp->z_pflags, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zsb), NULL,
&parent, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zsb), NULL,
&zp->z_atime, 16);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL, &zp->z_uid, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), NULL, &zp->z_gid, 8);
if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || zp->z_gen == 0) {
if (hdl == NULL)
sa_handle_destroy(zp->z_sa_hdl);
goto error;
}
ip->i_ino = obj;
zfs_inode_update(zp);
zfs_inode_set_ops(zsb, ip);
if (insert_inode_locked(ip))
goto error;
if (dentry) {
if (zpl_xattr_security_init(ip, dip, &dentry->d_name))
goto error;
d_instantiate(dentry, ip);
}
mutex_enter(&zsb->z_znodes_lock);
list_insert_tail(&zsb->z_all_znodes, zp);
zsb->z_nr_znodes++;
membar_producer();
mutex_exit(&zsb->z_znodes_lock);
unlock_new_inode(ip);
return (zp);
error:
unlock_new_inode(ip);
iput(ip);
return NULL;
}
