void
zfs_rmnode(znode_t *zp)
{
zfsvfs_t *zfsvfs = zp->z_zfsvfs;
objset_t *os = zfsvfs->z_os;
znode_t *xzp = NULL;
dmu_tx_t *tx;
uint64_t acl_obj;
uint64_t xattr_obj;
int error;
ASSERT(zp->z_links == 0);
ASSERT_VOP_ELOCKED(ZTOV(zp), __func__);
/*
* If this is an attribute directory, purge its contents.
*/
if (ZTOV(zp) != NULL && ZTOV(zp)->v_type == VDIR &&
(zp->z_pflags & ZFS_XATTR)) {
if (zfs_purgedir(zp) != 0) {
/*
* Not enough space to delete some xattrs.
* Leave it in the unlinked set.
*/
zfs_znode_dmu_fini(zp);
zfs_znode_free(zp);
return;
}
} else {
/*
* Free up all the data in the file. We don't do this for
* XATTR directories because we need truncate and remove to be
* in the same tx, like in zfs_znode_delete(). Otherwise, if
* we crash here we'll end up with an inconsistent truncated
* zap object in the delete queue. Note a truncated file is
* harmless since it only contains user data.
*/
error = dmu_free_long_range(os, zp->z_id, 0, DMU_OBJECT_END);
if (error) {
/*
* Not enough space. Leave the file in the unlinked
* set.
*/
zfs_znode_dmu_fini(zp);
zfs_znode_free(zp);
return;
}
}
/*
* If the file has extended attributes, we're going to unlink
* the xattr dir.
*/
error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs),
&xattr_obj, sizeof (xattr_obj));
if (error == 0 && xattr_obj) {
error = zfs_zget(zfsvfs, xattr_obj, &xzp);
ASSERT3S(error, ==, 0);
vn_lock(ZTOV(xzp), LK_EXCLUSIVE | LK_RETRY);
}
void getfinderinfo(znode_t *zp, cred_t *cr, finderinfo_t *fip)
{
vnode_t *xdvp = NULLVP;
vnode_t *xvp = NULLVP;
struct uio *auio = NULL;
struct componentname cn;
int error;
uint64_t xattr = 0;
if (sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zp->z_zfsvfs),
&xattr, sizeof(xattr)) ||
(xattr == 0)) {
goto nodata;
}
auio = uio_create(1, 0, UIO_SYSSPACE, UIO_READ);
if (auio == NULL) {
goto nodata;
}
uio_addiov(auio, CAST_USER_ADDR_T(fip), sizeof (finderinfo_t));
/*
* Grab the hidden attribute directory vnode.
*
* XXX - switch to embedded Finder Info when it becomes available
*/
if ((error = zfs_get_xattrdir(zp, &xdvp, cr, 0))) {
goto out;
}
bzero(&cn, sizeof (cn));
cn.cn_nameiop = LOOKUP;
cn.cn_flags = ISLASTCN;
cn.cn_nameptr = XATTR_FINDERINFO_NAME;
cn.cn_namelen = strlen(cn.cn_nameptr);
if ((error = zfs_dirlook(VTOZ(xdvp), cn.cn_nameptr, &xvp, 0, NULL, &cn))) {
goto out;
}
error = dmu_read_uio(zp->z_zfsvfs->z_os, VTOZ(xvp)->z_id, auio,
sizeof (finderinfo_t));
out:
if (auio)
uio_free(auio);
if (xvp)
vnode_put(xvp);
if (xdvp)
vnode_put(xdvp);
if (error == 0)
return;
nodata:
bzero(fip, sizeof (finderinfo_t));
}
static int
osd_object_sa_init(struct osd_object *obj, udmu_objset_t *uos)
{
int rc;
LASSERT(obj->oo_sa_hdl == NULL);
LASSERT(obj->oo_db != NULL);
rc = -sa_handle_get(uos->os, obj->oo_db->db_object, obj,
SA_HDL_PRIVATE, &obj->oo_sa_hdl);
if (rc)
return rc;
/* Cache the xattr object id, valid for the life of the object */
rc = -sa_lookup(obj->oo_sa_hdl, SA_ZPL_XATTR(uos), &obj->oo_xattr, 8);
if (rc == -ENOENT) {
obj->oo_xattr = ZFS_NO_OBJECT;
rc = 0;
} else if (rc) {
osd_object_sa_fini(obj);
}
return rc;
}
/*
* 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);
}
int
__osd_xattr_set(const struct lu_env *env, struct osd_object *obj,
const struct lu_buf *buf, const char *name, int fl,
struct osd_thandle *oh)
{
struct osd_device *osd = osd_obj2dev(obj);
dmu_buf_t *xa_zap_db = NULL;
dmu_buf_t *xa_data_db = NULL;
uint64_t xa_data_obj;
sa_handle_t *sa_hdl = NULL;
dmu_tx_t *tx = oh->ot_tx;
uint64_t size;
int rc;
LASSERT(obj->oo_sa_hdl);
if (obj->oo_xattr == ZFS_NO_OBJECT) {
struct lu_attr *la = &osd_oti_get(env)->oti_la;
la->la_valid = LA_MODE;
la->la_mode = S_IFDIR | S_IRUGO | S_IWUSR | S_IXUGO;
rc = __osd_zap_create(env, osd, &xa_zap_db, tx, la,
obj->oo_db->db_object, 0);
if (rc)
return rc;
obj->oo_xattr = xa_zap_db->db_object;
rc = osd_object_sa_update(obj, SA_ZPL_XATTR(osd),
&obj->oo_xattr, 8, oh);
if (rc)
goto out;
}
rc = -zap_lookup(osd->od_os, obj->oo_xattr, name, sizeof(uint64_t), 1,
&xa_data_obj);
if (rc == 0) {
if (fl & LU_XATTR_CREATE) {
rc = -EEXIST;
goto out;
}
/*
* Entry already exists.
* We'll truncate the existing object.
*/
rc = __osd_obj2dbuf(env, osd->od_os, xa_data_obj,
&xa_data_db);
if (rc)
goto out;
rc = -sa_handle_get(osd->od_os, xa_data_obj, NULL,
SA_HDL_PRIVATE, &sa_hdl);
if (rc)
goto out;
rc = -sa_lookup(sa_hdl, SA_ZPL_SIZE(osd), &size, 8);
if (rc)
goto out_sa;
rc = -dmu_free_range(osd->od_os, xa_data_db->db_object,
0, DMU_OBJECT_END, tx);
if (rc)
goto out_sa;
} else if (rc == -ENOENT) {
struct lu_attr *la = &osd_oti_get(env)->oti_la;
/*
* Entry doesn't exist, we need to create a new one and a new
* object to store the value.
*/
if (fl & LU_XATTR_REPLACE) {
/* should be ENOATTR according to the
* man, but that is undefined here */
rc = -ENODATA;
goto out;
}
la->la_valid = LA_MODE;
la->la_mode = S_IFREG | S_IRUGO | S_IWUSR;
rc = __osd_object_create(env, obj, &xa_data_db, tx, la,
obj->oo_xattr);
if (rc)
goto out;
xa_data_obj = xa_data_db->db_object;
rc = -sa_handle_get(osd->od_os, xa_data_obj, NULL,
SA_HDL_PRIVATE, &sa_hdl);
if (rc)
goto out;
rc = -zap_add(osd->od_os, obj->oo_xattr, name, sizeof(uint64_t),
1, &xa_data_obj, tx);
if (rc)
goto out_sa;
} else {
/* There was an error looking up the xattr name */
goto out;
}
/* Finally write the xattr value */
dmu_write(osd->od_os, xa_data_obj, 0, buf->lb_len, buf->lb_buf, tx);
size = buf->lb_len;
rc = -sa_update(sa_hdl, SA_ZPL_SIZE(osd), &size, 8, tx);
out_sa:
sa_handle_destroy(sa_hdl);
out:
if (xa_data_db != NULL)
dmu_buf_rele(xa_data_db, FTAG);
if (xa_zap_db != NULL)
dmu_buf_rele(xa_zap_db, FTAG);
return rc;
}
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);
}
void
zfs_rmnode(znode_t *zp)
{
zfsvfs_t *zfsvfs = ZTOZSB(zp);
objset_t *os = zfsvfs->z_os;
znode_t *xzp = NULL;
dmu_tx_t *tx;
uint64_t acl_obj;
uint64_t xattr_obj;
uint64_t links;
int error;
ASSERT(ZTOI(zp)->i_nlink == 0);
ASSERT(atomic_read(&ZTOI(zp)->i_count) == 0);
/*
* If this is an attribute directory, purge its contents.
*/
if (S_ISDIR(ZTOI(zp)->i_mode) && (zp->z_pflags & ZFS_XATTR)) {
if (zfs_purgedir(zp) != 0) {
/*
* Not enough space to delete some xattrs.
* Leave it in the unlinked set.
*/
zfs_znode_dmu_fini(zp);
return;
}
}
/*
* Free up all the data in the file. We don't do this for directories
* because we need truncate and remove to be in the same tx, like in
* zfs_znode_delete(). Otherwise, if we crash here we'll end up with
* an inconsistent truncated zap object in the delete queue. Note a
* truncated file is harmless since it only contains user data.
*/
if (S_ISREG(ZTOI(zp)->i_mode)) {
error = dmu_free_long_range(os, zp->z_id, 0, DMU_OBJECT_END);
if (error) {
/*
* Not enough space or we were interrupted by unmount.
* Leave the file in the unlinked set.
*/
zfs_znode_dmu_fini(zp);
return;
}
}
/*
* If the file has extended attributes, we're going to unlink
* the xattr dir.
*/
error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs),
&xattr_obj, sizeof (xattr_obj));
if (error == 0 && xattr_obj) {
error = zfs_zget(zfsvfs, xattr_obj, &xzp);
ASSERT(error == 0);
}
acl_obj = zfs_external_acl(zp);
/*
* Set up the final transaction.
*/
tx = dmu_tx_create(os);
dmu_tx_hold_free(tx, zp->z_id, 0, DMU_OBJECT_END);
dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
if (xzp) {
dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, TRUE, NULL);
dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE);
}
if (acl_obj)
dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END);
zfs_sa_upgrade_txholds(tx, zp);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
/*
* Not enough space to delete the file. Leave it in the
* unlinked set, leaking it until the fs is remounted (at
* which point we'll call zfs_unlinked_drain() to process it).
*/
dmu_tx_abort(tx);
zfs_znode_dmu_fini(zp);
goto out;
}
if (xzp) {
ASSERT(error == 0);
mutex_enter(&xzp->z_lock);
xzp->z_unlinked = B_TRUE; /* mark xzp for deletion */
clear_nlink(ZTOI(xzp)); /* no more links to it */
links = 0;
VERIFY(0 == sa_update(xzp->z_sa_hdl, SA_ZPL_LINKS(zfsvfs),
&links, sizeof (links), tx));
mutex_exit(&xzp->z_lock);
zfs_unlinked_add(xzp, tx);
}
/* Remove this znode from the unlinked set */
VERIFY3U(0, ==,
zap_remove_int(zfsvfs->z_os, zfsvfs->z_unlinkedobj, zp->z_id, tx));
zfs_znode_delete(zp, tx);
dmu_tx_commit(tx);
out:
if (xzp)
zfs_iput_async(ZTOI(xzp));
}
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);
}
zfs_dirent_lock(zfs_dirlock_t **dlpp, znode_t *dzp, char *name, znode_t **zpp,
int flag)
#endif
{
zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
zfs_dirlock_t *dl;
uint64_t zoid;
int error;
vnode_t *vp;
#ifdef __APPLE__
char *name;
u_int8_t *nfc_name = NULL; /* NFC form of name */
int nfc_namesize = 0;
#endif
*zpp = NULL;
*dlpp = NULL;
#ifdef __APPLE__
/* Note: cnp will be NULL for ZXATTR case */
name = cnp ? cnp->cn_nameptr : "";
if (cnp)
ASSERT(name[cnp->cn_namelen] == '\0');
#endif
/*
* Verify that we are not trying to lock '.', '..', or '.zfs'
*/
if ((name[0] == '.') &&
((name[1] == '\0') || ((name[1] == '.') && (name[2] == '\0'))) ||
zfs_has_ctldir(dzp) && strcmp(name, ZFS_CTLDIR_NAME) == 0)
return (EEXIST);
#ifdef __APPLE__
/*
* Mac OS X: store non-ascii names in UTF-8 NFC (pre-composed) on disk.
*
* The NFC name ptr is stored in dl->dl_name (allocated here)
* and its freed by zfs_dirent_unlock (since dl_namesize != 0).
*
* Since NFC size will not expand, we can allocate the same sized buffer.
*/
if (!is_ascii_str(name)) {
size_t outlen;
nfc_namesize = strlen(name) + 1;
nfc_name = kmem_alloc(nfc_namesize, KM_SLEEP);
if (utf8_normalizestr((const u_int8_t *)name, nfc_namesize, nfc_name,
&outlen, nfc_namesize, UTF_PRECOMPOSED) == 0) {
/* Normalization succeeded, switch to NFC name. */
name = (char *)nfc_name;
} else {
/* Normalization failed, just use input name as-is. */
kmem_free(nfc_name, nfc_namesize);
nfc_name = NULL;
}
}
#endif
/*
* Wait until there are no locks on this name.
*/
rw_enter(&dzp->z_name_lock, RW_READER);
mutex_enter(&dzp->z_lock);
for (;;) {
if (dzp->z_unlinked) {
mutex_exit(&dzp->z_lock);
rw_exit(&dzp->z_name_lock);
#ifdef __APPLE__
/* Release any unused NFC name before returning */
if (nfc_name) {
kmem_free(nfc_name, nfc_namesize);
}
#endif
return (ENOENT);
}
for (dl = dzp->z_dirlocks; dl != NULL; dl = dl->dl_next)
if (strcmp(name, dl->dl_name) == 0)
break;
if (dl == NULL) {
/*
* Allocate a new dirlock and add it to the list.
*/
dl = kmem_alloc(sizeof (zfs_dirlock_t), KM_SLEEP);
cv_init(&dl->dl_cv, NULL, CV_DEFAULT, NULL);
dl->dl_name = name;
dl->dl_sharecnt = 0;
dl->dl_namesize = 0;
dl->dl_dzp = dzp;
dl->dl_next = dzp->z_dirlocks;
dzp->z_dirlocks = dl;
#ifdef __APPLE__
/*
* Keep the NFC name around in dir lock by tagging it
* (setting nfc_namesize).
*/
if (nfc_name) {
dl->dl_namesize = nfc_namesize;
nfc_name = NULL; /* its now part of the dir lock */
}
#endif
break;
}
if ((flag & ZSHARED) && dl->dl_sharecnt != 0)
break;
cv_wait(&dl->dl_cv, &dzp->z_lock);
dl=NULL;
}
#ifdef __APPLE__
/*
* Release any unused NFC name (ie if we found a pre-existing lock entry)
*/
if (nfc_name) {
kmem_free(nfc_name, nfc_namesize);
nfc_name = NULL;
}
#endif
if ((flag & ZSHARED) && ++dl->dl_sharecnt > 1 && dl->dl_namesize == 0) {
/*
* We're the second shared reference to dl. Make a copy of
* dl_name in case the first thread goes away before we do.
* Note that we initialize the new name before storing its
* pointer into dl_name, because the first thread may load
* dl->dl_name at any time. He'll either see the old value,
* which is his, or the new shared copy; either is OK.
*/
dl->dl_namesize = strlen(dl->dl_name) + 1;
name = kmem_alloc(dl->dl_namesize, KM_SLEEP);
bcopy(dl->dl_name, name, dl->dl_namesize);
dl->dl_name = name;
}
mutex_exit(&dzp->z_lock);
/*
* We have a dirlock on the name. (Note that it is the dirlock,
* not the dzp's z_lock, that protects the name in the zap object.)
* See if there's an object by this name; if so, put a hold on it.
*/
if (flag & ZXATTR) {
error = sa_lookup(dzp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), &zoid,
sizeof (zoid));
if (error == 0)
error = (zoid == 0 ? ENOENT : 0);
} else {
#ifdef __APPLE__
/*
* Lookup an entry in the vnode name cache
*
* If the lookup succeeds, the vnode is returned in *vpp,
* and a status of -1 is returned.
*
* If the lookup determines that the name does not exist
* (negative caching), a status of ENOENT is returned.
*
* If the lookup fails, a status of zero is returned.
*/
switch ( cache_lookup(ZTOV(dzp), &vp, cnp) ) {
case -1:
break;
case ENOENT:
vp = DNLC_NO_VNODE;
break;
default:
vp = NULLVP;
}
#else
vp = dnlc_lookup(ZTOV(dzp), name);
#endif /* __APPLE__ */
if (vp == DNLC_NO_VNODE) {
VN_RELE(vp);
error = ENOENT;
} else if (vp) {
if (flag & ZNEW) {
zfs_dirent_unlock(dl);
VN_RELE(vp);
return (EEXIST);
}
*dlpp = dl;
*zpp = VTOZ(vp);
return (0);
} else {
error = zap_lookup(zfsvfs->z_os, dzp->z_id, name,
8, 1, &zoid);
zoid = ZFS_DIRENT_OBJ(zoid);
if (error == ENOENT)
#ifdef __APPLE__
/*
* Add a negative entry into the VFS name cache
*/
if ((flag & ZNEW) == 0 &&
(dzp->z_pflags & ZFS_XATTR) == 0 &&
(cnp) &&
(cnp->cn_flags & MAKEENTRY) &&
(cnp->cn_nameiop != CREATE) &&
(cnp->cn_nameiop != RENAME)) {
cache_enter(ZTOV(dzp), NULLVP, cnp);
}
#else
dnlc_update(ZTOV(dzp), name, DNLC_NO_VNODE);
#endif /* __APPLE__ */
}
}
if (error) {
if (error != ENOENT || (flag & ZEXISTS)) {
zfs_dirent_unlock(dl);
return (error);
}
} else {
if (flag & ZNEW) {
zfs_dirent_unlock(dl);
return (EEXIST);
}
//error = zfs_zget_sans_vnode(zfsvfs, zoid, zpp);
error = zfs_zget(zfsvfs, zoid, zpp);
if (error) {
zfs_dirent_unlock(dl);
return (error);
} else {
// Should this be here?
//printf("zfs_dir attach 1\n");
//zfs_attach_vnode(*zpp);
}
if (!(flag & ZXATTR))
#ifdef __APPLE__
if (cnp && cnp->cn_flags & MAKEENTRY)
cache_enter(ZTOV(dzp), ZTOV(*zpp), cnp);
#else
dnlc_update(ZTOV(dzp), name, ZTOV(*zpp));
#endif /* __APPLE__ */
}
*dlpp = dl;
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;
uint64_t xoid;
int error;
zfs_acl_ids_t acl_ids;
boolean_t fuid_dirtied;
*xvpp = NULL;
#ifndef __APPLE__
/* In Mac OS X access preflighting is done above the file system. */
if (error = zfs_zaccess(zp, ACE_WRITE_NAMED_ATTRS, cr))
return (error);
#endif /*!__APPLE__*/
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);
if (fuid_dirtied)
zfs_fuid_sync(zfsvfs, tx);
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);
/* Cleanup any znode we consumed during zfs_mknode() */
printf("ZFS_POSTPROCESS_ZP(xzp);\n");
printf("zfs_dir attach 2\n");
zfs_attach_vnode(xzp);
*xvpp = ZTOV(xzp);
return (0);
}
void
zfs_rmnode(znode_t *zp)
{
zfs_sb_t *zsb = ZTOZSB(zp);
objset_t *os = zsb->z_os;
znode_t *xzp = NULL;
dmu_tx_t *tx;
uint64_t acl_obj;
uint64_t xattr_obj;
uint64_t count;
int error;
ASSERT(zp->z_links == 0);
ASSERT(atomic_read(&ZTOI(zp)->i_count) == 0);
/*
* If this is an attribute directory, purge its contents.
*/
if (S_ISDIR(ZTOI(zp)->i_mode) && (zp->z_pflags & ZFS_XATTR)) {
error = zap_count(os, zp->z_id, &count);
if (error) {
zfs_znode_dmu_fini(zp);
return;
}
if (count > 0) {
taskq_t *taskq;
/*
* There are still directory entries in this xattr
* directory. Let zfs_unlinked_drain() deal with
* them to avoid deadlocking this process in the
* zfs_purgedir()->zfs_zget()->ilookup() callpath
* on the xattr inode's I_FREEING bit.
*/
taskq = dsl_pool_iput_taskq(dmu_objset_pool(os));
taskq_dispatch(taskq, (task_func_t *)
zfs_unlinked_drain, zsb, TQ_SLEEP);
zfs_znode_dmu_fini(zp);
return;
}
}
/*
* Free up all the data in the file.
*/
error = dmu_free_long_range(os, zp->z_id, 0, DMU_OBJECT_END);
if (error) {
/*
* Not enough space. Leave the file in the unlinked set.
*/
zfs_znode_dmu_fini(zp);
return;
}
/*
* If the file has extended attributes, we're going to unlink
* the xattr dir.
*/
error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zsb),
&xattr_obj, sizeof (xattr_obj));
if (error == 0 && xattr_obj) {
error = zfs_zget(zsb, xattr_obj, &xzp);
ASSERT(error == 0);
}
acl_obj = zfs_external_acl(zp);
/*
* Set up the final transaction.
*/
tx = dmu_tx_create(os);
dmu_tx_hold_free(tx, zp->z_id, 0, DMU_OBJECT_END);
dmu_tx_hold_zap(tx, zsb->z_unlinkedobj, FALSE, NULL);
if (xzp) {
dmu_tx_hold_zap(tx, zsb->z_unlinkedobj, TRUE, NULL);
dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE);
}
if (acl_obj)
dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END);
zfs_sa_upgrade_txholds(tx, zp);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
/*
* Not enough space to delete the file. Leave it in the
* unlinked set, leaking it until the fs is remounted (at
* which point we'll call zfs_unlinked_drain() to process it).
*/
dmu_tx_abort(tx);
zfs_znode_dmu_fini(zp);
goto out;
}
if (xzp) {
ASSERT(error == 0);
mutex_enter(&xzp->z_lock);
xzp->z_unlinked = B_TRUE; /* mark xzp for deletion */
xzp->z_links = 0; /* no more links to it */
VERIFY(0 == sa_update(xzp->z_sa_hdl, SA_ZPL_LINKS(zsb),
&xzp->z_links, sizeof (xzp->z_links), tx));
mutex_exit(&xzp->z_lock);
zfs_unlinked_add(xzp, tx);
}
/* Remove this znode from the unlinked set */
VERIFY3U(0, ==,
zap_remove_int(zsb->z_os, zsb->z_unlinkedobj, zp->z_id, tx));
zfs_znode_delete(zp, tx);
dmu_tx_commit(tx);
out:
if (xzp)
iput(ZTOI(xzp));
}
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);
}
void fileattrpack(attrinfo_t *aip, zfsvfs_t *zfsvfs, znode_t *zp)
{
attrgroup_t fileattr = aip->ai_attrlist->fileattr;
void *attrbufptr = *aip->ai_attrbufpp;
void *varbufptr = *aip->ai_varbufpp;
uint64_t allocsize = 0;
cred_t *cr = (cred_t *)vfs_context_ucred(aip->ai_context);
if ((ATTR_FILE_ALLOCSIZE | ATTR_FILE_DATAALLOCSIZE) & fileattr && zp) {
uint32_t blksize;
u_longlong_t nblks;
sa_object_size(zp->z_sa_hdl, &blksize, &nblks);
allocsize = (uint64_t)512LL * (uint64_t)nblks;
}
if (ATTR_FILE_LINKCOUNT & fileattr) {
uint64_t val;
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_LINKS(zfsvfs),
&val, sizeof(val)) == 0);
*((u_int32_t *)attrbufptr) = val;
attrbufptr = ((u_int32_t *)attrbufptr) + 1;
}
if (ATTR_FILE_TOTALSIZE & fileattr) {
uint64_t val;
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs),
&val, sizeof(val)) == 0);
*((off_t *)attrbufptr) = val;
attrbufptr = ((off_t *)attrbufptr) + 1;
}
if (ATTR_FILE_ALLOCSIZE & fileattr) {
*((off_t *)attrbufptr) = allocsize;
attrbufptr = ((off_t *)attrbufptr) + 1;
}
if (ATTR_FILE_IOBLOCKSIZE & fileattr && zp) {
*((u_int32_t *)attrbufptr) =
zp->z_blksz ? zp->z_blksz : zfsvfs->z_max_blksz;
attrbufptr = ((u_int32_t *)attrbufptr) + 1;
}
if (ATTR_FILE_DEVTYPE & fileattr) {
uint64_t mode, val=0;
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs),
&mode, sizeof(mode)) == 0);
sa_lookup(zp->z_sa_hdl, SA_ZPL_RDEV(zfsvfs),
&val, sizeof(val));
if (S_ISBLK(mode) || S_ISCHR(mode))
*((u_int32_t *)attrbufptr) = (u_int32_t)val;
else
*((u_int32_t *)attrbufptr) = 0;
attrbufptr = ((u_int32_t *)attrbufptr) + 1;
}
if (ATTR_FILE_DATALENGTH & fileattr) {
uint64_t val;
VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs),
&val, sizeof(val)) == 0);
*((off_t *)attrbufptr) = val;
attrbufptr = ((off_t *)attrbufptr) + 1;
}
if (ATTR_FILE_DATAALLOCSIZE & fileattr) {
*((off_t *)attrbufptr) = allocsize;
attrbufptr = ((off_t *)attrbufptr) + 1;
}
if ((ATTR_FILE_RSRCLENGTH | ATTR_FILE_RSRCALLOCSIZE) & fileattr) {
uint64_t rsrcsize = 0;
uint64_t xattr;
if (!sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs),
&xattr, sizeof(xattr)) &&
xattr) {
vnode_t *xdvp = NULLVP;
vnode_t *xvp = NULLVP;
struct componentname cn;
bzero(&cn, sizeof (cn));
cn.cn_nameiop = LOOKUP;
cn.cn_flags = ISLASTCN;
cn.cn_nameptr = XATTR_RESOURCEFORK_NAME;
cn.cn_namelen = strlen(cn.cn_nameptr);
/* Grab the hidden attribute directory vnode. */
if (zfs_get_xattrdir(zp, &xdvp, cr, 0) == 0 &&
zfs_dirlook(VTOZ(xdvp), cn.cn_nameptr, &xvp, 0, NULL,
&cn) == 0) {
rsrcsize = VTOZ(xvp)->z_size;
}
if (xvp)
vnode_put(xvp);
if (xdvp)
vnode_put(xdvp);
}
if (ATTR_FILE_RSRCLENGTH & fileattr) {
*((off_t *)attrbufptr) = rsrcsize;
attrbufptr = ((off_t *)attrbufptr) + 1;
}
if (ATTR_FILE_RSRCALLOCSIZE & fileattr) {
*((off_t *)attrbufptr) = roundup(rsrcsize, 512);
attrbufptr = ((off_t *)attrbufptr) + 1;
}
}
*aip->ai_attrbufpp = attrbufptr;
*aip->ai_varbufpp = varbufptr;
}
/*
* 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);
}
/*
* Lock a directory entry. A dirlock on <dzp, name> protects that name
* in dzp's directory zap object. As long as you hold a dirlock, you can
* assume two things: (1) dzp cannot be reaped, and (2) no other thread
* can change the zap entry for (i.e. link or unlink) this name.
*
* Input arguments:
* dzp - znode for directory
* name - name of entry to lock
* flag - ZNEW: if the entry already exists, fail with EEXIST.
* ZEXISTS: if the entry does not exist, fail with ENOENT.
* ZSHARED: allow concurrent access with other ZSHARED callers.
* ZXATTR: we want dzp's xattr directory
* ZCILOOK: On a mixed sensitivity file system,
* this lookup should be case-insensitive.
* ZCIEXACT: On a purely case-insensitive file system,
* this lookup should be case-sensitive.
* ZRENAMING: we are locking for renaming, force narrow locks
* ZHAVELOCK: Don't grab the z_name_lock for this call. The
* current thread already holds it.
*
* Output arguments:
* zpp - pointer to the znode for the entry (NULL if there isn't one)
* dlpp - pointer to the dirlock for this entry (NULL on error)
* direntflags - (case-insensitive lookup only)
* flags if multiple case-sensitive matches exist in directory
* realpnp - (case-insensitive lookup only)
* actual name matched within the directory
*
* Return value: 0 on success or errno on failure.
*
* NOTE: Always checks for, and rejects, '.' and '..'.
* NOTE: For case-insensitive file systems we take wide locks (see below),
* but return znode pointers to a single match.
*/
int
zfs_dirent_lock(zfs_dirlock_t **dlpp, znode_t *dzp, char *name, znode_t **zpp,
int flag, int *direntflags, pathname_t *realpnp)
{
zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
zfs_dirlock_t *dl;
boolean_t update;
boolean_t exact;
uint64_t zoid;
vnode_t *vp = NULL;
int error = 0;
int cmpflags;
*zpp = NULL;
*dlpp = NULL;
/*
* Verify that we are not trying to lock '.', '..', or '.zfs'
*/
if ((name[0] == '.' &&
(name[1] == '\0' || (name[1] == '.' && name[2] == '\0'))) ||
(zfs_has_ctldir(dzp) && strcmp(name, ZFS_CTLDIR_NAME) == 0))
return (EEXIST);
/*
* Case sensitivity and normalization preferences are set when
* the file system is created. These are stored in the
* zfsvfs->z_case and zfsvfs->z_norm fields. These choices
* affect what vnodes can be cached in the DNLC, how we
* perform zap lookups, and the "width" of our dirlocks.
*
* A normal dirlock locks a single name. Note that with
* normalization a name can be composed multiple ways, but
* when normalized, these names all compare equal. A wide
* dirlock locks multiple names. We need these when the file
* system is supporting mixed-mode access. It is sometimes
* necessary to lock all case permutations of file name at
* once so that simultaneous case-insensitive/case-sensitive
* behaves as rationally as possible.
*/
/*
* Decide if exact matches should be requested when performing
* a zap lookup on file systems supporting case-insensitive
* access.
*/
exact =
((zfsvfs->z_case == ZFS_CASE_INSENSITIVE) && (flag & ZCIEXACT)) ||
((zfsvfs->z_case == ZFS_CASE_MIXED) && !(flag & ZCILOOK));
/*
* Only look in or update the DNLC if we are looking for the
* name on a file system that does not require normalization
* or case folding. We can also look there if we happen to be
* on a non-normalizing, mixed sensitivity file system IF we
* are looking for the exact name.
*
* Maybe can add TO-UPPERed version of name to dnlc in ci-only
* case for performance improvement?
*/
update = !zfsvfs->z_norm ||
((zfsvfs->z_case == ZFS_CASE_MIXED) &&
!(zfsvfs->z_norm & ~U8_TEXTPREP_TOUPPER) && !(flag & ZCILOOK));
/*
* ZRENAMING indicates we are in a situation where we should
* take narrow locks regardless of the file system's
* preferences for normalizing and case folding. This will
* prevent us deadlocking trying to grab the same wide lock
* twice if the two names happen to be case-insensitive
* matches.
*/
if (flag & ZRENAMING)
cmpflags = 0;
else
cmpflags = zfsvfs->z_norm;
/*
* Wait until there are no locks on this name.
*
* Don't grab the the lock if it is already held. However, cannot
* have both ZSHARED and ZHAVELOCK together.
*/
ASSERT(!(flag & ZSHARED) || !(flag & ZHAVELOCK));
if (!(flag & ZHAVELOCK))
rw_enter(&dzp->z_name_lock, RW_READER);
mutex_enter(&dzp->z_lock);
for (;;) {
if (dzp->z_unlinked) {
mutex_exit(&dzp->z_lock);
if (!(flag & ZHAVELOCK))
rw_exit(&dzp->z_name_lock);
return (ENOENT);
}
for (dl = dzp->z_dirlocks; dl != NULL; dl = dl->dl_next) {
if ((u8_strcmp(name, dl->dl_name, 0, cmpflags,
U8_UNICODE_LATEST, &error) == 0) || error != 0)
break;
}
if (error != 0) {
mutex_exit(&dzp->z_lock);
if (!(flag & ZHAVELOCK))
rw_exit(&dzp->z_name_lock);
return (ENOENT);
}
if (dl == NULL) {
/*
* Allocate a new dirlock and add it to the list.
*/
dl = kmem_alloc(sizeof (zfs_dirlock_t), KM_SLEEP);
cv_init(&dl->dl_cv, NULL, CV_DEFAULT, NULL);
dl->dl_name = name;
dl->dl_sharecnt = 0;
dl->dl_namelock = 0;
dl->dl_namesize = 0;
dl->dl_dzp = dzp;
dl->dl_next = dzp->z_dirlocks;
dzp->z_dirlocks = dl;
break;
}
if ((flag & ZSHARED) && dl->dl_sharecnt != 0)
break;
cv_wait(&dl->dl_cv, &dzp->z_lock);
}
/*
* If the z_name_lock was NOT held for this dirlock record it.
*/
if (flag & ZHAVELOCK)
dl->dl_namelock = 1;
if ((flag & ZSHARED) && ++dl->dl_sharecnt > 1 && dl->dl_namesize == 0) {
/*
* We're the second shared reference to dl. Make a copy of
* dl_name in case the first thread goes away before we do.
* Note that we initialize the new name before storing its
* pointer into dl_name, because the first thread may load
* dl->dl_name at any time. He'll either see the old value,
* which is his, or the new shared copy; either is OK.
*/
dl->dl_namesize = strlen(dl->dl_name) + 1;
name = kmem_alloc(dl->dl_namesize, KM_SLEEP);
bcopy(dl->dl_name, name, dl->dl_namesize);
dl->dl_name = name;
}
mutex_exit(&dzp->z_lock);
/*
* We have a dirlock on the name. (Note that it is the dirlock,
* not the dzp's z_lock, that protects the name in the zap object.)
* See if there's an object by this name; if so, put a hold on it.
*/
if (flag & ZXATTR) {
error = sa_lookup(dzp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), &zoid,
sizeof (zoid));
if (error == 0)
error = (zoid == 0 ? ENOENT : 0);
} else {
if (update)
vp = dnlc_lookup(ZTOV(dzp), name);
if (vp == DNLC_NO_VNODE) {
VN_RELE(vp);
error = ENOENT;
} else if (vp) {
if (flag & ZNEW) {
zfs_dirent_unlock(dl);
VN_RELE(vp);
return (EEXIST);
}
*dlpp = dl;
*zpp = VTOZ(vp);
return (0);
} else {
error = zfs_match_find(zfsvfs, dzp, name, exact,
update, direntflags, realpnp, &zoid);
}
}
if (error) {
if (error != ENOENT || (flag & ZEXISTS)) {
zfs_dirent_unlock(dl);
return (error);
}
} else {
if (flag & ZNEW) {
zfs_dirent_unlock(dl);
return (EEXIST);
}
error = zfs_zget(zfsvfs, zoid, zpp);
if (error) {
zfs_dirent_unlock(dl);
return (error);
}
if (!(flag & ZXATTR) && update)
dnlc_update(ZTOV(dzp), name, ZTOV(*zpp));
}
*dlpp = dl;
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);
}
void
zfs_rmnode(znode_t *zp)
{
zfsvfs_t *zfsvfs = zp->z_zfsvfs;
objset_t *os = zfsvfs->z_os;
znode_t *xzp = NULL;
dmu_tx_t *tx;
uint64_t acl_obj;
uint64_t xattr_obj;
int error;
ASSERT(zp->z_links == 0);
ASSERT(ZTOV(zp)->v_count == 0);
/*
* If this is an attribute directory, purge its contents.
*/
if (ZTOV(zp)->v_type == VDIR && (zp->z_pflags & ZFS_XATTR)) {
if (zfs_purgedir(zp) != 0) {
/*
* Not enough space to delete some xattrs.
* Leave it in the unlinked set.
*/
zfs_znode_dmu_fini(zp);
zfs_znode_free(zp);
return;
}
}
/*
* Free up all the data in the file.
*/
error = dmu_free_long_range(os, zp->z_id, 0, DMU_OBJECT_END);
if (error) {
/*
* Not enough space. Leave the file in the unlinked set.
*/
zfs_znode_dmu_fini(zp);
zfs_znode_free(zp);
return;
}
/*
* If the file has extended attributes, we're going to unlink
* the xattr dir.
*/
error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs),
&xattr_obj, sizeof (xattr_obj));
if (error == 0 && xattr_obj) {
error = zfs_zget(zfsvfs, xattr_obj, &xzp);
ASSERT(error == 0);
}
acl_obj = zfs_external_acl(zp);
/*
* Set up the final transaction.
*/
tx = dmu_tx_create(os);
dmu_tx_hold_free(tx, zp->z_id, 0, DMU_OBJECT_END);
dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
if (xzp) {
dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, TRUE, NULL);
dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE);
}
if (acl_obj)
dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END);
zfs_sa_upgrade_txholds(tx, zp);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
/*
* Not enough space to delete the file. Leave it in the
* unlinked set, leaking it until the fs is remounted (at
* which point we'll call zfs_unlinked_drain() to process it).
*/
dmu_tx_abort(tx);
zfs_znode_dmu_fini(zp);
zfs_znode_free(zp);
goto out;
}
if (xzp) {
ASSERT(error == 0);
mutex_enter(&xzp->z_lock);
xzp->z_unlinked = B_TRUE; /* mark xzp for deletion */
xzp->z_links = 0; /* no more links to it */
VERIFY(0 == sa_update(xzp->z_sa_hdl, SA_ZPL_LINKS(zfsvfs),
&xzp->z_links, sizeof (xzp->z_links), tx));
mutex_exit(&xzp->z_lock);
zfs_unlinked_add(xzp, tx);
}
/* Remove this znode from the unlinked set */
VERIFY3U(0, ==,
zap_remove_int(zfsvfs->z_os, zfsvfs->z_unlinkedobj, zp->z_id, tx));
zfs_znode_delete(zp, tx);
dmu_tx_commit(tx);
out:
if (xzp)
VN_RELE(ZTOV(xzp));
}
/*
* Look up a directory entry under a locked vnode.
* dvp being locked gives us a guarantee that there are no concurrent
* modification of the directory and, thus, if a node can be found in
* the directory, then it must not be unlinked.
*
* Input arguments:
* dzp - znode for directory
* name - name of entry to lock
* flag - ZNEW: if the entry already exists, fail with EEXIST.
* ZEXISTS: if the entry does not exist, fail with ENOENT.
* ZXATTR: we want dzp's xattr directory
*
* Output arguments:
* zpp - pointer to the znode for the entry (NULL if there isn't one)
*
* Return value: 0 on success or errno on failure.
*
* NOTE: Always checks for, and rejects, '.' and '..'.
*/
int
zfs_dirent_lookup(znode_t *dzp, const char *name, znode_t **zpp, int flag)
{
zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
boolean_t exact;
uint64_t zoid;
vnode_t *vp = NULL;
int error = 0;
ASSERT_VOP_LOCKED(ZTOV(dzp), __func__);
*zpp = NULL;
/*
* Verify that we are not trying to lock '.', '..', or '.zfs'
*/
if (name[0] == '.' &&
(name[1] == '\0' || (name[1] == '.' && name[2] == '\0')) ||
zfs_has_ctldir(dzp) && strcmp(name, ZFS_CTLDIR_NAME) == 0)
return (SET_ERROR(EEXIST));
/*
* Case sensitivity and normalization preferences are set when
* the file system is created. These are stored in the
* zfsvfs->z_case and zfsvfs->z_norm fields. These choices
* affect how we perform zap lookups.
*
* Decide if exact matches should be requested when performing
* a zap lookup on file systems supporting case-insensitive
* access.
*
* NB: we do not need to worry about this flag for ZFS_CASE_SENSITIVE
* because in that case MT_EXACT and MT_FIRST should produce exactly
* the same result.
*/
exact = zfsvfs->z_case == ZFS_CASE_MIXED;
if (dzp->z_unlinked && !(flag & ZXATTR))
return (ENOENT);
if (flag & ZXATTR) {
error = sa_lookup(dzp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), &zoid,
sizeof (zoid));
if (error == 0)
error = (zoid == 0 ? ENOENT : 0);
} else {
error = zfs_match_find(zfsvfs, dzp, name, exact, &zoid);
}
if (error) {
if (error != ENOENT || (flag & ZEXISTS)) {
return (error);
}
} else {
if (flag & ZNEW) {
return (SET_ERROR(EEXIST));
}
error = zfs_zget(zfsvfs, zoid, zpp);
if (error)
return (error);
ASSERT(!(*zpp)->z_unlinked);
}
return (0);
}
