Пример #1
0
int
zfs_zget(zfs_sb_t *zsb, uint64_t obj_num, znode_t **zpp)
{
	dmu_object_info_t doi;
	dmu_buf_t	*db;
	znode_t		*zp;
	int err;
	sa_handle_t	*hdl;

	*zpp = NULL;

again:
	ZFS_OBJ_HOLD_ENTER(zsb, obj_num);

	err = sa_buf_hold(zsb->z_os, obj_num, NULL, &db);
	if (err) {
		ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
		return (err);
	}

	dmu_object_info_from_db(db, &doi);
	if (doi.doi_bonus_type != DMU_OT_SA &&
	    (doi.doi_bonus_type != DMU_OT_ZNODE ||
	    (doi.doi_bonus_type == DMU_OT_ZNODE &&
	    doi.doi_bonus_size < sizeof (znode_phys_t)))) {
		sa_buf_rele(db, NULL);
		ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
		return (SET_ERROR(EINVAL));
	}

	hdl = dmu_buf_get_user(db);
	if (hdl != NULL) {
		zp = sa_get_userdata(hdl);


		/*
		 * Since "SA" does immediate eviction we
		 * should never find a sa handle that doesn't
		 * know about the znode.
		 */

		ASSERT3P(zp, !=, NULL);

		mutex_enter(&zp->z_lock);
		ASSERT3U(zp->z_id, ==, obj_num);
		if (zp->z_unlinked) {
			err = SET_ERROR(ENOENT);
		} else {
			/*
			 * If igrab() returns NULL the VFS has independently
			 * determined the inode should be evicted and has
			 * called iput_final() to start the eviction process.
			 * The SA handle is still valid but because the VFS
			 * requires that the eviction succeed we must drop
			 * our locks and references to allow the eviction to
			 * complete.  The zfs_zget() may then be retried.
			 *
			 * This unlikely case could be optimized by registering
			 * a sops->drop_inode() callback.  The callback would
			 * need to detect the active SA hold thereby informing
			 * the VFS that this inode should not be evicted.
			 */
			if (igrab(ZTOI(zp)) == NULL) {
				mutex_exit(&zp->z_lock);
				sa_buf_rele(db, NULL);
				ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
				goto again;
			}
			*zpp = zp;
			err = 0;
		}
		mutex_exit(&zp->z_lock);
		sa_buf_rele(db, NULL);
		ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
		return (err);
	}
Пример #2
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);
}
Пример #3
0
int
zfs_zget(zfs_sb_t *zsb, uint64_t obj_num, znode_t **zpp)
{
	dmu_object_info_t doi;
	dmu_buf_t	*db;
	znode_t		*zp;
	int err;
	sa_handle_t	*hdl;
	struct inode	*ip;

	*zpp = NULL;

again:
	ip = ilookup(zsb->z_sb, obj_num);

	ZFS_OBJ_HOLD_ENTER(zsb, obj_num);

	err = sa_buf_hold(zsb->z_os, obj_num, NULL, &db);
	if (err) {
		ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
		iput(ip);
		return (err);
	}

	dmu_object_info_from_db(db, &doi);
	if (doi.doi_bonus_type != DMU_OT_SA &&
	    (doi.doi_bonus_type != DMU_OT_ZNODE ||
	    (doi.doi_bonus_type == DMU_OT_ZNODE &&
	    doi.doi_bonus_size < sizeof (znode_phys_t)))) {
		sa_buf_rele(db, NULL);
		ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
		iput(ip);
		return (SET_ERROR(EINVAL));
	}

	hdl = dmu_buf_get_user(db);
	if (hdl != NULL) {
		if (ip == NULL) {
			/*
			 * ilookup returned NULL, which means
			 * the znode is dying - but the SA handle isn't
			 * quite dead yet, we need to drop any locks
			 * we're holding, re-schedule the task and try again.
			 */
			sa_buf_rele(db, NULL);
			ZFS_OBJ_HOLD_EXIT(zsb, obj_num);

			schedule();
			goto again;
		}

		zp = sa_get_userdata(hdl);

		/*
		 * Since "SA" does immediate eviction we
		 * should never find a sa handle that doesn't
		 * know about the znode.
		 */

		ASSERT3P(zp, !=, NULL);

		mutex_enter(&zp->z_lock);
		ASSERT3U(zp->z_id, ==, obj_num);
		if (zp->z_unlinked) {
			err = SET_ERROR(ENOENT);
		} else {
			igrab(ZTOI(zp));
			*zpp = zp;
			err = 0;
		}
		sa_buf_rele(db, NULL);
		mutex_exit(&zp->z_lock);
		ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
		iput(ip);
		return (err);
	}

	ASSERT3P(ip, ==, NULL);

	/*
	 * Not found create new znode/vnode but only if file exists.
	 *
	 * There is a small window where zfs_vget() could
	 * find this object while a file create is still in
	 * progress.  This is checked for in zfs_znode_alloc()
	 *
	 * if zfs_znode_alloc() fails it will drop the hold on the
	 * bonus buffer.
	 */
	zp = zfs_znode_alloc(zsb, db, doi.doi_data_block_size,
	    doi.doi_bonus_type, obj_num, NULL, NULL);
	if (zp == NULL) {
		err = SET_ERROR(ENOENT);
	} else {
		*zpp = zp;
	}
	ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
	return (err);
}