cma.c

linux 内核源代码

{
	struct ib_qp_attr qp_attr;
	int qp_attr_mask, ret;

	mutex_lock(&id_priv->qp_mutex);
	if (!id_priv->id.qp) {
		ret = 0;
		goto out;
	}

	qp_attr.qp_state = IB_QPS_RTS;
	ret = rdma_init_qp_attr(&id_priv->id, &qp_attr, &qp_attr_mask);
	if (ret)
		goto out;

	ret = ib_modify_qp(id_priv->id.qp, &qp_attr, qp_attr_mask);
out:
	mutex_unlock(&id_priv->qp_mutex);
	return ret;
}

static int cma_modify_qp_err(struct rdma_id_private *id_priv)
{
	struct ib_qp_attr qp_attr;
	int ret;

	mutex_lock(&id_priv->qp_mutex);
	if (!id_priv->id.qp) {
		ret = 0;
		goto out;
	}

	qp_attr.qp_state = IB_QPS_ERR;
	ret = ib_modify_qp(id_priv->id.qp, &qp_attr, IB_QP_STATE);
out:
	mutex_unlock(&id_priv->qp_mutex);
	return ret;
}

static int cma_ib_init_qp_attr(struct rdma_id_private *id_priv,
			       struct ib_qp_attr *qp_attr, int *qp_attr_mask)
{
	struct rdma_dev_addr *dev_addr = &id_priv->id.route.addr.dev_addr;
	int ret;

	ret = ib_find_cached_pkey(id_priv->id.device, id_priv->id.port_num,
				  ib_addr_get_pkey(dev_addr),
				  &qp_attr->pkey_index);
	if (ret)
		return ret;

	qp_attr->port_num = id_priv->id.port_num;
	*qp_attr_mask = IB_QP_STATE | IB_QP_PKEY_INDEX | IB_QP_PORT;

	if (cma_is_ud_ps(id_priv->id.ps)) {
		qp_attr->qkey = id_priv->qkey;
		*qp_attr_mask |= IB_QP_QKEY;
	} else {
		qp_attr->qp_access_flags = 0;
		*qp_attr_mask |= IB_QP_ACCESS_FLAGS;
	}
	return 0;
}

int rdma_init_qp_attr(struct rdma_cm_id *id, struct ib_qp_attr *qp_attr,
		       int *qp_attr_mask)
{
	struct rdma_id_private *id_priv;
	int ret = 0;

	id_priv = container_of(id, struct rdma_id_private, id);
	switch (rdma_node_get_transport(id_priv->id.device->node_type)) {
	case RDMA_TRANSPORT_IB:
		if (!id_priv->cm_id.ib || cma_is_ud_ps(id_priv->id.ps))
			ret = cma_ib_init_qp_attr(id_priv, qp_attr, qp_attr_mask);
		else
			ret = ib_cm_init_qp_attr(id_priv->cm_id.ib, qp_attr,
						 qp_attr_mask);
		if (qp_attr->qp_state == IB_QPS_RTR)
			qp_attr->rq_psn = id_priv->seq_num;
		break;
	case RDMA_TRANSPORT_IWARP:
		if (!id_priv->cm_id.iw) {
			qp_attr->qp_access_flags = 0;
			*qp_attr_mask = IB_QP_STATE | IB_QP_ACCESS_FLAGS;
		} else
			ret = iw_cm_init_qp_attr(id_priv->cm_id.iw, qp_attr,
						 qp_attr_mask);
		break;
	default:
		ret = -ENOSYS;
		break;
	}

	return ret;
}
EXPORT_SYMBOL(rdma_init_qp_attr);

static inline int cma_zero_addr(struct sockaddr *addr)
{
	struct in6_addr *ip6;

	if (addr->sa_family == AF_INET)
		return ZERONET(((struct sockaddr_in *) addr)->sin_addr.s_addr);
	else {
		ip6 = &((struct sockaddr_in6 *) addr)->sin6_addr;
		return (ip6->s6_addr32[0] | ip6->s6_addr32[1] |
			ip6->s6_addr32[2] | ip6->s6_addr32[3]) == 0;
	}
}

static inline int cma_loopback_addr(struct sockaddr *addr)
{
	return LOOPBACK(((struct sockaddr_in *) addr)->sin_addr.s_addr);
}

static inline int cma_any_addr(struct sockaddr *addr)
{
	return cma_zero_addr(addr) || cma_loopback_addr(addr);
}

static inline __be16 cma_port(struct sockaddr *addr)
{
	if (addr->sa_family == AF_INET)
		return ((struct sockaddr_in *) addr)->sin_port;
	else
		return ((struct sockaddr_in6 *) addr)->sin6_port;
}

static inline int cma_any_port(struct sockaddr *addr)
{
	return !cma_port(addr);
}

static int cma_get_net_info(void *hdr, enum rdma_port_space ps,
			    u8 *ip_ver, __u16 *port,
			    union cma_ip_addr **src, union cma_ip_addr **dst)
{
	switch (ps) {
	case RDMA_PS_SDP:
		if (sdp_get_majv(((struct sdp_hh *) hdr)->sdp_version) !=
		    SDP_MAJ_VERSION)
			return -EINVAL;

		*ip_ver	= sdp_get_ip_ver(hdr);
		*port	= ((struct sdp_hh *) hdr)->port;
		*src	= &((struct sdp_hh *) hdr)->src_addr;
		*dst	= &((struct sdp_hh *) hdr)->dst_addr;
		break;
	default:
		if (((struct cma_hdr *) hdr)->cma_version != CMA_VERSION)
			return -EINVAL;

		*ip_ver	= cma_get_ip_ver(hdr);
		*port	= ((struct cma_hdr *) hdr)->port;
		*src	= &((struct cma_hdr *) hdr)->src_addr;
		*dst	= &((struct cma_hdr *) hdr)->dst_addr;
		break;
	}

	if (*ip_ver != 4 && *ip_ver != 6)
		return -EINVAL;
	return 0;
}

static void cma_save_net_info(struct rdma_addr *addr,
			      struct rdma_addr *listen_addr,
			      u8 ip_ver, __u16 port,
			      union cma_ip_addr *src, union cma_ip_addr *dst)
{
	struct sockaddr_in *listen4, *ip4;
	struct sockaddr_in6 *listen6, *ip6;

	switch (ip_ver) {
	case 4:
		listen4 = (struct sockaddr_in *) &listen_addr->src_addr;
		ip4 = (struct sockaddr_in *) &addr->src_addr;
		ip4->sin_family = listen4->sin_family;
		ip4->sin_addr.s_addr = dst->ip4.addr;
		ip4->sin_port = listen4->sin_port;

		ip4 = (struct sockaddr_in *) &addr->dst_addr;
		ip4->sin_family = listen4->sin_family;
		ip4->sin_addr.s_addr = src->ip4.addr;
		ip4->sin_port = port;
		break;
	case 6:
		listen6 = (struct sockaddr_in6 *) &listen_addr->src_addr;
		ip6 = (struct sockaddr_in6 *) &addr->src_addr;
		ip6->sin6_family = listen6->sin6_family;
		ip6->sin6_addr = dst->ip6;
		ip6->sin6_port = listen6->sin6_port;

		ip6 = (struct sockaddr_in6 *) &addr->dst_addr;
		ip6->sin6_family = listen6->sin6_family;
		ip6->sin6_addr = src->ip6;
		ip6->sin6_port = port;
		break;
	default:
		break;
	}
}

static inline int cma_user_data_offset(enum rdma_port_space ps)
{
	switch (ps) {
	case RDMA_PS_SDP:
		return 0;
	default:
		return sizeof(struct cma_hdr);
	}
}

static void cma_cancel_route(struct rdma_id_private *id_priv)
{
	switch (rdma_node_get_transport(id_priv->id.device->node_type)) {
	case RDMA_TRANSPORT_IB:
		if (id_priv->query)
			ib_sa_cancel_query(id_priv->query_id, id_priv->query);
		break;
	default:
		break;
	}
}

static void cma_cancel_listens(struct rdma_id_private *id_priv)
{
	struct rdma_id_private *dev_id_priv;

	/*
	 * Remove from listen_any_list to prevent added devices from spawning
	 * additional listen requests.
	 */
	mutex_lock(&lock);
	list_del(&id_priv->list);

	while (!list_empty(&id_priv->listen_list)) {
		dev_id_priv = list_entry(id_priv->listen_list.next,
					 struct rdma_id_private, listen_list);
		/* sync with device removal to avoid duplicate destruction */
		list_del_init(&dev_id_priv->list);
		list_del(&dev_id_priv->listen_list);
		mutex_unlock(&lock);

		rdma_destroy_id(&dev_id_priv->id);
		mutex_lock(&lock);
	}
	mutex_unlock(&lock);
}

static void cma_cancel_operation(struct rdma_id_private *id_priv,
				 enum cma_state state)
{
	switch (state) {
	case CMA_ADDR_QUERY:
		rdma_addr_cancel(&id_priv->id.route.addr.dev_addr);
		break;
	case CMA_ROUTE_QUERY:
		cma_cancel_route(id_priv);
		break;
	case CMA_LISTEN:
		if (cma_any_addr(&id_priv->id.route.addr.src_addr) &&
		    !id_priv->cma_dev)
			cma_cancel_listens(id_priv);
		break;
	default:
		break;
	}
}

static void cma_release_port(struct rdma_id_private *id_priv)
{
	struct rdma_bind_list *bind_list = id_priv->bind_list;

	if (!bind_list)
		return;

	mutex_lock(&lock);
	hlist_del(&id_priv->node);
	if (hlist_empty(&bind_list->owners)) {
		idr_remove(bind_list->ps, bind_list->port);
		kfree(bind_list);
	}
	mutex_unlock(&lock);
}

static void cma_leave_mc_groups(struct rdma_id_private *id_priv)
{
	struct cma_multicast *mc;

	while (!list_empty(&id_priv->mc_list)) {
		mc = container_of(id_priv->mc_list.next,
				  struct cma_multicast, list);
		list_del(&mc->list);
		ib_sa_free_multicast(mc->multicast.ib);
		kfree(mc);
	}
}

void rdma_destroy_id(struct rdma_cm_id *id)
{
	struct rdma_id_private *id_priv;
	enum cma_state state;

	id_priv = container_of(id, struct rdma_id_private, id);
	state = cma_exch(id_priv, CMA_DESTROYING);
	cma_cancel_operation(id_priv, state);

	mutex_lock(&lock);
	if (id_priv->cma_dev) {
		mutex_unlock(&lock);
		switch (rdma_node_get_transport(id->device->node_type)) {
		case RDMA_TRANSPORT_IB:
			if (id_priv->cm_id.ib && !IS_ERR(id_priv->cm_id.ib))
				ib_destroy_cm_id(id_priv->cm_id.ib);
			break;
		case RDMA_TRANSPORT_IWARP:
			if (id_priv->cm_id.iw && !IS_ERR(id_priv->cm_id.iw))
				iw_destroy_cm_id(id_priv->cm_id.iw);
			break;
		default:
			break;
		}
		cma_leave_mc_groups(id_priv);
		mutex_lock(&lock);
		cma_detach_from_dev(id_priv);
	}
	mutex_unlock(&lock);

	cma_release_port(id_priv);
	cma_deref_id(id_priv);
	wait_for_completion(&id_priv->comp);

	if (id_priv->internal_id)
		cma_deref_id(id_priv->id.context);

	kfree(id_priv->id.route.path_rec);
	kfree(id_priv);
}
EXPORT_SYMBOL(rdma_destroy_id);

static int cma_rep_recv(struct rdma_id_private *id_priv)
{
	int ret;

	ret = cma_modify_qp_rtr(id_priv);
	if (ret)
		goto reject;

	ret = cma_modify_qp_rts(id_priv);
	if (ret)
		goto reject;

	ret = ib_send_cm_rtu(id_priv->cm_id.ib, NULL, 0);
	if (ret)
		goto reject;

	return 0;
reject:
	cma_modify_qp_err(id_priv);
	ib_send_cm_rej(id_priv->cm_id.ib, IB_CM_REJ_CONSUMER_DEFINED,
		       NULL, 0, NULL, 0);
	return ret;
}

static int cma_verify_rep(struct rdma_id_private *id_priv, void *data)
{
	if (id_priv->id.ps == RDMA_PS_SDP &&
	    sdp_get_majv(((struct sdp_hah *) data)->sdp_version) !=
	    SDP_MAJ_VERSION)
		return -EINVAL;

	return 0;
}

static void cma_set_rep_event_data(struct rdma_cm_event *event,
				   struct ib_cm_rep_event_param *rep_data,
				   void *private_data)
{
	event->param.conn.private_data = private_data;
	event->param.conn.private_data_len = IB_CM_REP_PRIVATE_DATA_SIZE;
	event->param.conn.responder_resources = rep_data->responder_resources;
	event->param.conn.initiator_depth = rep_data->initiator_depth;
	event->param.conn.flow_control = rep_data->flow_control;
	event->param.conn.rnr_retry_count = rep_data->rnr_retry_count;
	event->param.conn.srq = rep_data->srq;
	event->param.conn.qp_num = rep_data->remote_qpn;
}

static int cma_ib_handler(struct ib_cm_id *cm_id, struct ib_cm_event *ib_event)
{
	struct rdma_id_private *id_priv = cm_id->context;
	struct rdma_cm_event event;
	int ret = 0;

	if (cma_disable_remove(id_priv, CMA_CONNECT))
		return 0;

	memset(&event, 0, sizeof event);
	switch (ib_event->event) {
	case IB_CM_REQ_ERROR:
	case IB_CM_REP_ERROR:
		event.event = RDMA_CM_EVENT_UNREACHABLE;
		event.status = -ETIMEDOUT;
		break;
	case IB_CM_REP_RECEIVED:
		event.status = cma_verify_rep(id_priv, ib_event->private_data);
		if (event.status)
			event.event = RDMA_CM_EVENT_CONNECT_ERROR;
		else if (id_priv->id.qp && id_priv->id.ps != RDMA_PS_SDP) {
			event.status = cma_rep_recv(id_priv);
			event.event = event.status ? RDMA_CM_EVENT_CONNECT_ERROR :
						     RDMA_CM_EVENT_ESTABLISHED;
		} else
			event.event = RDMA_CM_EVENT_CONNECT_RESPONSE;
		cma_set_rep_event_data(&event, &ib_event->param.rep_rcvd,
				       ib_event->private_data);
		break;
	case IB_CM_RTU_RECEIVED:
	case IB_CM_USER_ESTABLISHED:
		event.event = RDMA_CM_EVENT_ESTABLISHED;
		break;
	case IB_CM_DREQ_ERROR:
		event.status = -ETIMEDOUT; /* fall through */
	case IB_CM_DREQ_RECEIVED:
	case IB_CM_DREP_RECEIVED:
		if (!cma_comp_exch(id_priv, CMA_CONNECT, CMA_DISCONNECT))
			goto out;
		event.event = RDMA_CM_EVENT_DISCONNECTED;
		break;
	case IB_CM_TIMEWAIT_EXIT:
	case IB_CM_MRA_RECEIVED:
		/* ignore event */
		goto out;
	case IB_CM_REJ_RECEIVED:
		cma_modify_qp_err(id_priv);
		event.status = ib_event->param.rej_rcvd.reason;
		event.event = RDMA_CM_EVENT_REJECTED;
		event.param.conn.private_data = ib_event->private_data;
		event.param.conn.private_data_len = IB_CM_REJ_PRIVATE_DATA_SIZE;
		break;
	default:
		printk(KERN_ERR "RDMA CMA: unexpected IB CM event: %d",
		       ib_event->event);
		goto out;
	}

	ret = id_priv->id.event_handler(&id_priv->id, &event);
	if (ret) {
		/* Destroy the CM ID by returning a non-zero value. */
		id_priv->cm_id.ib = NULL;
		cma_exch(id_priv, CMA_DESTROYING);
		cma_enable_remove(id_priv);
		rdma_destroy_id(&id_priv->id);
		return ret;
	}
out:
	cma_enable_remove(id_priv);
	return ret;
}

static struct rdma_id_private *cma_new_conn_id(struct rdma_cm_id *listen_id,
					       struct ib_cm_event *ib_event)
{
	struct rdma_id_private *id_priv;
	struct rdma_cm_id *id;
	struct rdma_route *rt;
	union cma_ip_addr *src, *dst;
	__u16 port;
	u8 ip_ver;

	if (cma_get_net_info(ib_event->private_data, listen_id->ps,
			     &ip_ver, &port, &src, &dst))
		goto err;

	id = rdma_create_id(listen_id->event_handler, listen_id->context,
			    listen_id->ps);
	if (IS_ERR(id))
		goto err;

	cma_save_net_info(&id->route.addr, &listen_id->route.addr,
			  ip_ver, port, src, dst);

	rt = &id->route;
	rt->num_paths = ib_event->param.req_rcvd.alternate_path ? 2 : 1;
	rt->path_rec = kmalloc(sizeof *rt->path_rec * rt->num_paths,
			       GFP_KERNEL);
	if (!rt->path_rec)
		goto destroy_id;

	rt->path_rec[0] = *ib_event->param.req_rcvd.primary_path;
	if (rt->num_paths == 2)
		rt->path_rec[1] = *ib_event->param.req_rcvd.alternate_path;

	ib_addr_set_sgid(&rt->addr.dev_addr, &rt->path_rec[0].sgid);
	ib_addr_set_dgid(&rt->addr.dev_addr, &rt->path_rec[0].dgid);
	ib_addr_set_pkey(&rt->addr.dev_addr, be16_to_cpu(rt->path_rec[0].pkey));
	rt->addr.dev_addr.dev_type = RDMA_NODE_IB_CA;

	id_priv = container_of(id, struct rdma_id_private, id);
	id_priv->state = CMA_CONNECT;
	return id_priv;

destroy_id:
	rdma_destroy_id(id);
err:
	return NULL;
}

static struct rdma_id_private *cma_new_udp_id(struct rdma_cm_id *listen_id,
					      struct ib_cm_event *ib_event)
{
	struct rdma_id_private *id_priv;
	struct rdma_cm_id *id;
	union cma_ip_addr *src, *dst;
	__u16 port;
	u8 ip_ver;
	int ret;

	id = rdma_create_id(listen_id->event_handler, listen_id->context,
			    listen_id->ps);
	if (IS_ERR(id))
		return NULL;