Merge tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dledford/rdma

Pull rdma fixes from Doug Ledford:
 "Round three of 4.8 rc fixes.

  This is likely the last rdma pull request this cycle.  The new rxe
  driver had a few issues (you probably saw the boot bot bug report) and
  they should be addressed now.  There are a couple other fixes here,
  mainly mlx4.  There are still two outstanding issues that need
  resolved but I don't think their fix will make this kernel cycle.

  Summary:

   - Various fixes to rdmavt, ipoib, mlx5, mlx4, rxe"

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dledford/rdma:
  IB/rdmavt: Don't vfree a kzalloc'ed memory region
  IB/rxe: Fix kmem_cache leak
  IB/rxe: Fix race condition between requester and completer
  IB/rxe: Fix duplicate atomic request handling
  IB/rxe: Fix kernel panic in udp_setup_tunnel
  IB/mlx5: Set source mac address in FTE
  IB/mlx5: Enable MAD_IFC commands for IB ports only
  IB/mlx4: Diagnostic HW counters are not supported in slave mode
  IB/mlx4: Use correct subnet-prefix in QP1 mads under SR-IOV
  IB/mlx4: Fix code indentation in QP1 MAD flow
  IB/mlx4: Fix incorrect MC join state bit-masking on SR-IOV
  IB/ipoib: Don't allow MC joins during light MC flush
  IB/rxe: fix GFP_KERNEL in spinlock context

drivers/infiniband/hw/mlx4/mad.c

@@ -1128,6 +1128,27 @@ void handle_port_mgmt_change_event(struct work_struct *work)
 
 		/* Generate GUID changed event */
 		if (changed_attr & MLX4_EQ_PORT_INFO_GID_PFX_CHANGE_MASK) {
+			if (mlx4_is_master(dev->dev)) {
+				union ib_gid gid;
+				int err = 0;
+
+				if (!eqe->event.port_mgmt_change.params.port_info.gid_prefix)
+					err = __mlx4_ib_query_gid(&dev->ib_dev, port, 0, &gid, 1);
+				else
+					gid.global.subnet_prefix =
+						eqe->event.port_mgmt_change.params.port_info.gid_prefix;
+				if (err) {
+					pr_warn("Could not change QP1 subnet prefix for port %d: query_gid error (%d)\n",
+						port, err);
+				} else {
+					pr_debug("Changing QP1 subnet prefix for port %d. old=0x%llx. new=0x%llx\n",
+						 port,
+						 (u64)atomic64_read(&dev->sriov.demux[port - 1].subnet_prefix),
+						 be64_to_cpu(gid.global.subnet_prefix));
+					atomic64_set(&dev->sriov.demux[port - 1].subnet_prefix,
+						     be64_to_cpu(gid.global.subnet_prefix));
+				}
+			}
 			mlx4_ib_dispatch_event(dev, port, IB_EVENT_GID_CHANGE);
 			/*if master, notify all slaves*/
 			if (mlx4_is_master(dev->dev))
@@ -2202,6 +2223,8 @@ int mlx4_ib_init_sriov(struct mlx4_ib_dev *dev)
 		if (err)
 			goto demux_err;
 		dev->sriov.demux[i].guid_cache[0] = gid.global.interface_id;
+		atomic64_set(&dev->sriov.demux[i].subnet_prefix,
+			     be64_to_cpu(gid.global.subnet_prefix));
 		err = alloc_pv_object(dev, mlx4_master_func_num(dev->dev), i + 1,
 				      &dev->sriov.sqps[i]);
 		if (err)

drivers/infiniband/hw/mlx4/main.c

@@ -2202,6 +2202,9 @@ static int mlx4_ib_alloc_diag_counters(struct mlx4_ib_dev *ibdev)
 	bool per_port = !!(ibdev->dev->caps.flags2 &
 		MLX4_DEV_CAP_FLAG2_DIAG_PER_PORT);
 
+	if (mlx4_is_slave(ibdev->dev))
+		return 0;
+
 	for (i = 0; i < MLX4_DIAG_COUNTERS_TYPES; i++) {
 		/* i == 1 means we are building port counters */
 		if (i && !per_port)

drivers/infiniband/hw/mlx4/mcg.c

@@ -489,7 +489,7 @@ static u8 get_leave_state(struct mcast_group *group)
 		if (!group->members[i])
 			leave_state |= (1 << i);
 
-	return leave_state & (group->rec.scope_join_state & 7);
+	return leave_state & (group->rec.scope_join_state & 0xf);
 }
 
 static int join_group(struct mcast_group *group, int slave, u8 join_mask)
@@ -564,8 +564,8 @@ static void mlx4_ib_mcg_timeout_handler(struct work_struct *work)
 		} else
 			mcg_warn_group(group, "DRIVER BUG\n");
 	} else if (group->state == MCAST_LEAVE_SENT) {
-		if (group->rec.scope_join_state & 7)
-			group->rec.scope_join_state &= 0xf8;
+		if (group->rec.scope_join_state & 0xf)
+			group->rec.scope_join_state &= 0xf0;
 		group->state = MCAST_IDLE;
 		mutex_unlock(&group->lock);
 		if (release_group(group, 1))
@@ -605,7 +605,7 @@ static int handle_leave_req(struct mcast_group *group, u8 leave_mask,
 static int handle_join_req(struct mcast_group *group, u8 join_mask,
 			   struct mcast_req *req)
 {
-	u8 group_join_state = group->rec.scope_join_state & 7;
+	u8 group_join_state = group->rec.scope_join_state & 0xf;
 	int ref = 0;
 	u16 status;
 	struct ib_sa_mcmember_data *sa_data = (struct ib_sa_mcmember_data *)req->sa_mad.data;
@@ -690,8 +690,8 @@ static void mlx4_ib_mcg_work_handler(struct work_struct *work)
 			u8 cur_join_state;
 
 			resp_join_state = ((struct ib_sa_mcmember_data *)
-						group->response_sa_mad.data)->scope_join_state & 7;
-			cur_join_state = group->rec.scope_join_state & 7;
+						group->response_sa_mad.data)->scope_join_state & 0xf;
+			cur_join_state = group->rec.scope_join_state & 0xf;
 
 			if (method == IB_MGMT_METHOD_GET_RESP) {
 				/* successfull join */
@@ -710,7 +710,7 @@ static void mlx4_ib_mcg_work_handler(struct work_struct *work)
 		req = list_first_entry(&group->pending_list, struct mcast_req,
 				       group_list);
 		sa_data = (struct ib_sa_mcmember_data *)req->sa_mad.data;
-		req_join_state = sa_data->scope_join_state & 0x7;
+		req_join_state = sa_data->scope_join_state & 0xf;
 
 		/* For a leave request, we will immediately answer the VF, and
 		 * update our internal counters. The actual leave will be sent

drivers/infiniband/hw/mlx4/mlx4_ib.h

@@ -448,7 +448,7 @@ struct mlx4_ib_demux_ctx {
 	struct workqueue_struct *wq;
 	struct workqueue_struct *ud_wq;
 	spinlock_t ud_lock;
-	__be64 subnet_prefix;
+	atomic64_t subnet_prefix;
 	__be64 guid_cache[128];
 	struct mlx4_ib_dev *dev;
 	/* the following lock protects both mcg_table and mcg_mgid0_list */

drivers/infiniband/hw/mlx4/qp.c

@@ -2493,25 +2493,28 @@ static int build_mlx_header(struct mlx4_ib_sqp *sqp, struct ib_ud_wr *wr,
 		sqp->ud_header.grh.flow_label    =
 			ah->av.ib.sl_tclass_flowlabel & cpu_to_be32(0xfffff);
 		sqp->ud_header.grh.hop_limit     = ah->av.ib.hop_limit;
-		if (is_eth)
+		if (is_eth) {
 			memcpy(sqp->ud_header.grh.source_gid.raw, sgid.raw, 16);
-		else {
+		} else {
 			if (mlx4_is_mfunc(to_mdev(ib_dev)->dev)) {
 				/* When multi-function is enabled, the ib_core gid
 				 * indexes don't necessarily match the hw ones, so
-			 * we must use our own cache */
+				 * we must use our own cache
+				 */
 				sqp->ud_header.grh.source_gid.global.subnet_prefix =
-				to_mdev(ib_dev)->sriov.demux[sqp->qp.port - 1].
-						       subnet_prefix;
+					cpu_to_be64(atomic64_read(&(to_mdev(ib_dev)->sriov.
+								    demux[sqp->qp.port - 1].
+								    subnet_prefix)));
 				sqp->ud_header.grh.source_gid.global.interface_id =
 					to_mdev(ib_dev)->sriov.demux[sqp->qp.port - 1].
 						       guid_cache[ah->av.ib.gid_index];
-		} else
+			} else {
 				ib_get_cached_gid(ib_dev,
 						  be32_to_cpu(ah->av.ib.port_pd) >> 24,
 						  ah->av.ib.gid_index,
 						  &sqp->ud_header.grh.source_gid, NULL);
 			}
+		}
 		memcpy(sqp->ud_header.grh.destination_gid.raw,
 		       ah->av.ib.dgid, 16);
 	}

drivers/infiniband/hw/mlx5/main.c

@@ -288,7 +288,9 @@ __be16 mlx5_get_roce_udp_sport(struct mlx5_ib_dev *dev, u8 port_num,
 
 static int mlx5_use_mad_ifc(struct mlx5_ib_dev *dev)
 {
+	if (MLX5_CAP_GEN(dev->mdev, port_type) == MLX5_CAP_PORT_TYPE_IB)
 		return !MLX5_CAP_GEN(dev->mdev, ib_virt);
+	return 0;
 }
 
 enum {
@@ -1428,6 +1430,13 @@ static int parse_flow_attr(u32 *match_c, u32 *match_v,
 					     dmac_47_16),
 				ib_spec->eth.val.dst_mac);
 
+		ether_addr_copy(MLX5_ADDR_OF(fte_match_set_lyr_2_4, outer_headers_c,
+					     smac_47_16),
+				ib_spec->eth.mask.src_mac);
+		ether_addr_copy(MLX5_ADDR_OF(fte_match_set_lyr_2_4, outer_headers_v,
+					     smac_47_16),
+				ib_spec->eth.val.src_mac);
+
 		if (ib_spec->eth.mask.vlan_tag) {
 			MLX5_SET(fte_match_set_lyr_2_4, outer_headers_c,
 				 vlan_tag, 1);

drivers/infiniband/sw/rdmavt/mr.c

@@ -294,7 +294,7 @@ static void __rvt_free_mr(struct rvt_mr *mr)
 {
 	rvt_deinit_mregion(&mr->mr);
 	rvt_free_lkey(&mr->mr);
-	vfree(mr);
+	kfree(mr);
 }
 
 /**

drivers/infiniband/sw/rxe/rxe.c

@@ -362,15 +362,34 @@ static int __init rxe_module_init(void)
 		return err;
 	}
 
-	err = rxe_net_init();
+	err = rxe_net_ipv4_init();
 	if (err) {
-		pr_err("rxe: unable to init\n");
+		pr_err("rxe: unable to init ipv4 tunnel\n");
 		rxe_cache_exit();
-		return err;
+		goto exit;
+	}
+
+	err = rxe_net_ipv6_init();
+	if (err) {
+		pr_err("rxe: unable to init ipv6 tunnel\n");
+		rxe_cache_exit();
+		goto exit;
 	}
+
+	err = register_netdevice_notifier(&rxe_net_notifier);
+	if (err) {
+		pr_err("rxe: Failed to rigister netdev notifier\n");
+		goto exit;
+	}
+
 	pr_info("rxe: loaded\n");
 
 	return 0;
+
+exit:
+	rxe_release_udp_tunnel(recv_sockets.sk4);
+	rxe_release_udp_tunnel(recv_sockets.sk6);
+	return err;
 }
 
 static void __exit rxe_module_exit(void)

drivers/infiniband/sw/rxe/rxe_comp.c

@@ -689,7 +689,14 @@ int rxe_completer(void *arg)
 					qp->req.need_retry = 1;
 					rxe_run_task(&qp->req.task, 1);
 				}
+
+				if (pkt) {
+					rxe_drop_ref(pkt->qp);
+					kfree_skb(skb);
+				}
+
 				goto exit;
+
 			} else {
 				wqe->status = IB_WC_RETRY_EXC_ERR;
 				state = COMPST_ERROR;
@@ -716,6 +723,12 @@ int rxe_completer(void *arg)
 		case COMPST_ERROR:
 			do_complete(qp, wqe);
 			rxe_qp_error(qp);
+
+			if (pkt) {
+				rxe_drop_ref(pkt->qp);
+				kfree_skb(skb);
+			}
+
 			goto exit;
 		}
 	}

drivers/infiniband/sw/rxe/rxe_net.c

@@ -275,8 +275,9 @@ static struct socket *rxe_setup_udp_tunnel(struct net *net, __be16 port,
 	return sock;
 }
 
-static void rxe_release_udp_tunnel(struct socket *sk)
+void rxe_release_udp_tunnel(struct socket *sk)
 {
+	if (sk)
 		udp_tunnel_sock_release(sk);
 }
 
@@ -658,51 +659,45 @@ static int rxe_notify(struct notifier_block *not_blk,
 	return NOTIFY_OK;
 }
 
-static struct notifier_block rxe_net_notifier = {
+struct notifier_block rxe_net_notifier = {
 	.notifier_call = rxe_notify,
 };
 
-int rxe_net_init(void)
+int rxe_net_ipv4_init(void)
 {
-	int err;
-
 	spin_lock_init(&dev_list_lock);
 
-	recv_sockets.sk6 = rxe_setup_udp_tunnel(&init_net,
-			htons(ROCE_V2_UDP_DPORT), true);
-	if (IS_ERR(recv_sockets.sk6)) {
-		recv_sockets.sk6 = NULL;
-		pr_err("rxe: Failed to create IPv6 UDP tunnel\n");
-		return -1;
-	}
-
 	recv_sockets.sk4 = rxe_setup_udp_tunnel(&init_net,
 				htons(ROCE_V2_UDP_DPORT), false);
 	if (IS_ERR(recv_sockets.sk4)) {
-		rxe_release_udp_tunnel(recv_sockets.sk6);
 		recv_sockets.sk4 = NULL;
-		recv_sockets.sk6 = NULL;
 		pr_err("rxe: Failed to create IPv4 UDP tunnel\n");
 		return -1;
 	}
 
-	err = register_netdevice_notifier(&rxe_net_notifier);
-	if (err) {
-		rxe_release_udp_tunnel(recv_sockets.sk6);
-		rxe_release_udp_tunnel(recv_sockets.sk4);
-		pr_err("rxe: Failed to rigister netdev notifier\n");
-	}
+	return 0;
+}
 
-	return err;
+int rxe_net_ipv6_init(void)
+{
+#if IS_ENABLED(CONFIG_IPV6)
+
+	spin_lock_init(&dev_list_lock);
+
+	recv_sockets.sk6 = rxe_setup_udp_tunnel(&init_net,
+						htons(ROCE_V2_UDP_DPORT), true);
+	if (IS_ERR(recv_sockets.sk6)) {
+		recv_sockets.sk6 = NULL;
+		pr_err("rxe: Failed to create IPv6 UDP tunnel\n");
+		return -1;
+	}
+#endif
+	return 0;
 }
 
 void rxe_net_exit(void)
 {
-	if (recv_sockets.sk6)
 	rxe_release_udp_tunnel(recv_sockets.sk6);
-
-	if (recv_sockets.sk4)
 	rxe_release_udp_tunnel(recv_sockets.sk4);
-
 	unregister_netdevice_notifier(&rxe_net_notifier);
 }

drivers/infiniband/sw/rxe/rxe_net.h

@@ -44,10 +44,13 @@ struct rxe_recv_sockets {
 };
 
 extern struct rxe_recv_sockets recv_sockets;
+extern struct notifier_block rxe_net_notifier;
+void rxe_release_udp_tunnel(struct socket *sk);
 
 struct rxe_dev *rxe_net_add(struct net_device *ndev);
 
-int rxe_net_init(void);
+int rxe_net_ipv4_init(void);
+int rxe_net_ipv6_init(void);
 void rxe_net_exit(void);
 
 #endif /* RXE_NET_H */

drivers/infiniband/sw/rxe/rxe_recv.c

@@ -312,7 +312,7 @@ static void rxe_rcv_mcast_pkt(struct rxe_dev *rxe, struct sk_buff *skb)
 		 * make a copy of the skb to post to the next qp
 		 */
 		skb_copy = (mce->qp_list.next != &mcg->qp_list) ?
-				skb_clone(skb, GFP_KERNEL) : NULL;
+				skb_clone(skb, GFP_ATOMIC) : NULL;
 
 		pkt->qp = qp;
 		rxe_add_ref(qp);

drivers/infiniband/sw/rxe/rxe_req.c

@@ -512,23 +512,20 @@ static int fill_packet(struct rxe_qp *qp, struct rxe_send_wqe *wqe,
 
 static void update_wqe_state(struct rxe_qp *qp,
 		struct rxe_send_wqe *wqe,
-			     struct rxe_pkt_info *pkt,
-			     enum wqe_state *prev_state)
+		struct rxe_pkt_info *pkt)
 {
-	enum wqe_state prev_state_ = wqe->state;
-
 	if (pkt->mask & RXE_END_MASK) {
 		if (qp_type(qp) == IB_QPT_RC)
 			wqe->state = wqe_state_pending;
 	} else {
 		wqe->state = wqe_state_processing;
 	}
-
-	*prev_state = prev_state_;
 }
 
-static void update_state(struct rxe_qp *qp, struct rxe_send_wqe *wqe,
-			 struct rxe_pkt_info *pkt, int payload)
+static void update_wqe_psn(struct rxe_qp *qp,
+			   struct rxe_send_wqe *wqe,
+			   struct rxe_pkt_info *pkt,
+			   int payload)
 {
 	/* number of packets left to send including current one */
 	int num_pkt = (wqe->dma.resid + payload + qp->mtu - 1) / qp->mtu;
@@ -546,9 +543,34 @@ static void update_state(struct rxe_qp *qp, struct rxe_send_wqe *wqe,
 		qp->req.psn = (wqe->first_psn + num_pkt) & BTH_PSN_MASK;
 	else
 		qp->req.psn = (qp->req.psn + 1) & BTH_PSN_MASK;
+}
 
-	qp->req.opcode = pkt->opcode;
+static void save_state(struct rxe_send_wqe *wqe,
+		       struct rxe_qp *qp,
+		       struct rxe_send_wqe *rollback_wqe,
+		       struct rxe_qp *rollback_qp)
+{
+	rollback_wqe->state     = wqe->state;
+	rollback_wqe->first_psn = wqe->first_psn;
+	rollback_wqe->last_psn  = wqe->last_psn;
+	rollback_qp->req.psn    = qp->req.psn;
+}
+
+static void rollback_state(struct rxe_send_wqe *wqe,
+			   struct rxe_qp *qp,
+			   struct rxe_send_wqe *rollback_wqe,
+			   struct rxe_qp *rollback_qp)
+{
+	wqe->state     = rollback_wqe->state;
+	wqe->first_psn = rollback_wqe->first_psn;
+	wqe->last_psn  = rollback_wqe->last_psn;
+	qp->req.psn    = rollback_qp->req.psn;
+}
 
+static void update_state(struct rxe_qp *qp, struct rxe_send_wqe *wqe,
+			 struct rxe_pkt_info *pkt, int payload)
+{
+	qp->req.opcode = pkt->opcode;
 
 	if (pkt->mask & RXE_END_MASK)
 		qp->req.wqe_index = next_index(qp->sq.queue, qp->req.wqe_index);
@@ -571,7 +593,8 @@ int rxe_requester(void *arg)
 	int mtu;
 	int opcode;
 	int ret;
-	enum wqe_state prev_state;
+	struct rxe_qp rollback_qp;
+	struct rxe_send_wqe rollback_wqe;
 
 next_wqe:
 	if (unlikely(!qp->valid || qp->req.state == QP_STATE_ERROR))
@@ -688,13 +711,21 @@ int rxe_requester(void *arg)
 		goto err;
 	}
 
-	update_wqe_state(qp, wqe, &pkt, &prev_state);
+	/*
+	 * To prevent a race on wqe access between requester and completer,
+	 * wqe members state and psn need to be set before calling
+	 * rxe_xmit_packet().
+	 * Otherwise, completer might initiate an unjustified retry flow.
+	 */
+	save_state(wqe, qp, &rollback_wqe, &rollback_qp);
+	update_wqe_state(qp, wqe, &pkt);
+	update_wqe_psn(qp, wqe, &pkt, payload);
 	ret = rxe_xmit_packet(to_rdev(qp->ibqp.device), qp, &pkt, skb);
 	if (ret) {
 		qp->need_req_skb = 1;
 		kfree_skb(skb);
 
-		wqe->state = prev_state;
+		rollback_state(wqe, qp, &rollback_wqe, &rollback_qp);
 
 		if (ret == -EAGAIN) {
 			rxe_run_task(&qp->req.task, 1);

drivers/infiniband/sw/rxe/rxe_resp.c

@@ -972,11 +972,13 @@ static int send_atomic_ack(struct rxe_qp *qp, struct rxe_pkt_info *pkt,
 	free_rd_atomic_resource(qp, res);
 	rxe_advance_resp_resource(qp);
 
+	memcpy(SKB_TO_PKT(skb), &ack_pkt, sizeof(skb->cb));
+
 	res->type = RXE_ATOMIC_MASK;
 	res->atomic.skb = skb;
-	res->first_psn = qp->resp.psn;
-	res->last_psn = qp->resp.psn;
-	res->cur_psn = qp->resp.psn;
+	res->first_psn = ack_pkt.psn;
+	res->last_psn  = ack_pkt.psn;
+	res->cur_psn   = ack_pkt.psn;
 
 	rc = rxe_xmit_packet(rxe, qp, &ack_pkt, skb_copy);
 	if (rc) {
@@ -1116,8 +1118,7 @@ static enum resp_states duplicate_request(struct rxe_qp *qp,
 				rc = RESPST_CLEANUP;
 				goto out;
 			}
-			bth_set_psn(SKB_TO_PKT(skb_copy),
-				    qp->resp.psn - 1);
+
 			/* Resend the result. */
 			rc = rxe_xmit_packet(to_rdev(qp->ibqp.device), qp,
 					     pkt, skb_copy);

drivers/infiniband/ulp/ipoib/ipoib_ib.c

@@ -1161,8 +1161,17 @@ static void __ipoib_ib_dev_flush(struct ipoib_dev_priv *priv,
 	}
 
 	if (level == IPOIB_FLUSH_LIGHT) {
+		int oper_up;
 		ipoib_mark_paths_invalid(dev);
+		/* Set IPoIB operation as down to prevent races between:
+		 * the flush flow which leaves MCG and on the fly joins
+		 * which can happen during that time. mcast restart task
+		 * should deal with join requests we missed.
+		 */
+		oper_up = test_and_clear_bit(IPOIB_FLAG_OPER_UP, &priv->flags);
 		ipoib_mcast_dev_flush(dev);
+		if (oper_up)
+			set_bit(IPOIB_FLAG_OPER_UP, &priv->flags);
 		ipoib_flush_ah(dev);
 	}