RDMA 5.12 first rc pull request

- Fix corner cases in the rxe reference counting cleanup that are causing
   regressions in blktests for SRP
 
 - Two kdoc fixes so W=1 is clean
 
 - Missing error return in error unwind for mlx5
 
 - Wrong lock type nesting in IB CM
 -----BEGIN PGP SIGNATURE-----
 
 iQIzBAABCgAdFiEEfB7FMLh+8QxL+6i3OG33FX4gmxoFAmBCeEwACgkQOG33FX4g
 mxohmQ/7Bnf52Ay59dqdQzc8nDPXMwQ5ejZke+MnYZG3j79UqNQrbEnLJ8Y2BlQU
 IkpdTi8zUHm69GH7Ix9m/FsW52XFUuq2OEfGLvZNRUhi7hGEmc6llvfCK9y6a4Mt
 VEpyo7vC4QvA1OvMBF3Zv91jylKPJt5U51OuRtbh+6eoS6oVm7ajjOay+USYT/kX
 3zftTd6DUg6vpUP5j2WvjocgKUVGsh8GiyA3xC7u7AUTemcxvZ40wB535fLpn1JC
 ObU8aiRKw0Ib5JbRukCFFBV13kmA3lnAEy0SktPVhIoeUnF0m1vNtAJrdnHsWPxW
 lxtAXA3bN3oINhsEhnBllONKRxlqsUz7KT5bVAehsl8LENNLHhJWp0hwKawKK74K
 5MAQ+P6CIW8TLMKeOcYMXBD2UiElc9sdhYsDliM7leclftwOO5gFI4WgTWrQwxvl
 m+KewiByT6pzK7oyGvg9w/dl3y7atzy8y/7uMAfI2L6oL43z9yt24hHQTo3Z7UM7
 /bbSas/yRbZujB7qE45c6NvwRFQk6mYTK0Is0huMWOLn0Zf0fVcy3um0q+OKQ7At
 VS1Z7Fhy+ev1Lk764n5rmVUx8NWDlXkzwX7qk7SnQk7dWQ5lNFbziFDKmkiKgcAa
 //oN0I/FbAM4FjhKImRzbXJiLdULxJCqpfQ5VXbUXlfXdpzjabY=
 =j5ZU
 -----END PGP SIGNATURE-----

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

Pull rdma fixes from Jason Gunthorpe:
 "Nothing special here, though Bob's regression fixes for rxe would have
  made it before the rc cycle had there not been such strong winter
  weather!

   - Fix corner cases in the rxe reference counting cleanup that are
     causing regressions in blktests for SRP

   - Two kdoc fixes so W=1 is clean

   - Missing error return in error unwind for mlx5

   - Wrong lock type nesting in IB CM"

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma:
  RDMA/rxe: Fix errant WARN_ONCE in rxe_completer()
  RDMA/rxe: Fix extra deref in rxe_rcv_mcast_pkt()
  RDMA/rxe: Fix missed IB reference counting in loopback
  RDMA/uverbs: Fix kernel-doc warning of _uverbs_alloc
  RDMA/mlx5: Set correct kernel-doc identifier
  IB/mlx5: Add missing error code
  RDMA/rxe: Fix missing kconfig dependency on CRYPTO
  RDMA/cm: Fix IRQ restore in ib_send_cm_sidr_rep

drivers/infiniband/core/cm.c

@@ -3651,6 +3651,7 @@ static int cm_send_sidr_rep_locked(struct cm_id_private *cm_id_priv,
 				   struct ib_cm_sidr_rep_param *param)
 {
 	struct ib_mad_send_buf *msg;
+	unsigned long flags;
 	int ret;
 
 	lockdep_assert_held(&cm_id_priv->lock);
@@ -3676,12 +3677,12 @@ static int cm_send_sidr_rep_locked(struct cm_id_private *cm_id_priv,
 		return ret;
 	}
 	cm_id_priv->id.state = IB_CM_IDLE;
-	spin_lock_irq(&cm.lock);
+	spin_lock_irqsave(&cm.lock, flags);
 	if (!RB_EMPTY_NODE(&cm_id_priv->sidr_id_node)) {
 		rb_erase(&cm_id_priv->sidr_id_node, &cm.remote_sidr_table);
 		RB_CLEAR_NODE(&cm_id_priv->sidr_id_node);
 	}
-	spin_unlock_irq(&cm.lock);
+	spin_unlock_irqrestore(&cm.lock, flags);
 	return 0;
 }
 

drivers/infiniband/core/uverbs_ioctl.c

@@ -91,7 +91,7 @@ void uapi_compute_bundle_size(struct uverbs_api_ioctl_method *method_elm,
 }
 
 /**
- * uverbs_alloc() - Quickly allocate memory for use with a bundle
+ * _uverbs_alloc() - Quickly allocate memory for use with a bundle
  * @bundle: The bundle
  * @size: Number of bytes to allocate
  * @flags: Allocator flags

drivers/infiniband/hw/mlx5/devx.c

@@ -2073,8 +2073,10 @@ static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_SUBSCRIBE_EVENT)(
 
 		num_alloc_xa_entries++;
 		event_sub = kzalloc(sizeof(*event_sub), GFP_KERNEL);
-		if (!event_sub)
+		if (!event_sub) {
+			err = -ENOMEM;
 			goto err;
+		}
 
 		list_add_tail(&event_sub->event_list, &sub_list);
 		uverbs_uobject_get(&ev_file->uobj);

drivers/infiniband/hw/mlx5/odp.c

@@ -1082,7 +1082,7 @@ end:
 	return ret ? ret : npages;
 }
 
-/**
+/*
  * Parse a series of data segments for page fault handling.
  *
  * @dev:  Pointer to mlx5 IB device

drivers/infiniband/sw/rxe/Kconfig

@@ -4,6 +4,7 @@ config RDMA_RXE
 	depends on INET && PCI && INFINIBAND
 	depends on INFINIBAND_VIRT_DMA
 	select NET_UDP_TUNNEL
+	select CRYPTO
 	select CRYPTO_CRC32
 	help
 	This driver implements the InfiniBand RDMA transport over

drivers/infiniband/sw/rxe/rxe_comp.c

@@ -547,6 +547,7 @@ int rxe_completer(void *arg)
 	struct sk_buff *skb = NULL;
 	struct rxe_pkt_info *pkt = NULL;
 	enum comp_state state;
+	int ret = 0;
 
 	rxe_add_ref(qp);
 
@@ -554,7 +555,8 @@ int rxe_completer(void *arg)
 	    qp->req.state == QP_STATE_RESET) {
 		rxe_drain_resp_pkts(qp, qp->valid &&
 				    qp->req.state == QP_STATE_ERROR);
-		goto exit;
+		ret = -EAGAIN;
+		goto done;
 	}
 
 	if (qp->comp.timeout) {
@@ -564,8 +566,10 @@ int rxe_completer(void *arg)
 		qp->comp.timeout_retry = 0;
 	}
 
-	if (qp->req.need_retry)
-		goto exit;
+	if (qp->req.need_retry) {
+		ret = -EAGAIN;
+		goto done;
+	}
 
 	state = COMPST_GET_ACK;
 
@@ -636,8 +640,6 @@ int rxe_completer(void *arg)
 			break;
 
 		case COMPST_DONE:
-			if (pkt)
-				free_pkt(pkt);
 			goto done;
 
 		case COMPST_EXIT:
@@ -660,7 +662,8 @@ int rxe_completer(void *arg)
 			    qp->qp_timeout_jiffies)
 				mod_timer(&qp->retrans_timer,
 					  jiffies + qp->qp_timeout_jiffies);
-			goto exit;
+			ret = -EAGAIN;
+			goto done;
 
 		case COMPST_ERROR_RETRY:
 			/* we come here if the retry timer fired and we did
@@ -672,18 +675,18 @@ int rxe_completer(void *arg)
 			 */
 
 			/* there is nothing to retry in this case */
-			if (!wqe || (wqe->state == wqe_state_posted))
-				goto exit;
+			if (!wqe || (wqe->state == wqe_state_posted)) {
+				pr_warn("Retry attempted without a valid wqe\n");
+				ret = -EAGAIN;
+				goto done;
+			}
 
 			/* if we've started a retry, don't start another
 			 * retry sequence, unless this is a timeout.
 			 */
 			if (qp->comp.started_retry &&
-			    !qp->comp.timeout_retry) {
-				if (pkt)
-					free_pkt(pkt);
+			    !qp->comp.timeout_retry)
 				goto done;
-			}
 
 			if (qp->comp.retry_cnt > 0) {
 				if (qp->comp.retry_cnt != 7)
@@ -704,8 +707,6 @@ int rxe_completer(void *arg)
 					qp->comp.started_retry = 1;
 					rxe_run_task(&qp->req.task, 0);
 				}
-				if (pkt)
-					free_pkt(pkt);
 				goto done;
 
 			} else {
@@ -726,8 +727,8 @@ int rxe_completer(void *arg)
 				mod_timer(&qp->rnr_nak_timer,
 					  jiffies + rnrnak_jiffies(aeth_syn(pkt)
 						& ~AETH_TYPE_MASK));
-				free_pkt(pkt);
-				goto exit;
+				ret = -EAGAIN;
+				goto done;
 			} else {
 				rxe_counter_inc(rxe,
 						RXE_CNT_RNR_RETRY_EXCEEDED);
@@ -740,25 +741,15 @@ int rxe_completer(void *arg)
 			WARN_ON_ONCE(wqe->status == IB_WC_SUCCESS);
 			do_complete(qp, wqe);
 			rxe_qp_error(qp);
-			if (pkt)
-				free_pkt(pkt);
-			goto exit;
+			ret = -EAGAIN;
+			goto done;
 		}
 	}
 
-exit:
-	/* we come here if we are done with processing and want the task to
-	 * exit from the loop calling us
-	 */
-	WARN_ON_ONCE(skb);
-	rxe_drop_ref(qp);
-	return -EAGAIN;
-
 done:
-	/* we come here if we have processed a packet we want the task to call
-	 * us again to see if there is anything else to do
-	 */
-	WARN_ON_ONCE(skb);
+	if (pkt)
+		free_pkt(pkt);
 	rxe_drop_ref(qp);
-	return 0;
+
+	return ret;
 }

drivers/infiniband/sw/rxe/rxe_net.c

@@ -407,13 +407,21 @@ int rxe_send(struct rxe_pkt_info *pkt, struct sk_buff *skb)
 	return 0;
 }
 
+/* fix up a send packet to match the packets
+ * received from UDP before looping them back
+ */
 void rxe_loopback(struct sk_buff *skb)
 {
+	struct rxe_pkt_info *pkt = SKB_TO_PKT(skb);
+
 	if (skb->protocol == htons(ETH_P_IP))
 		skb_pull(skb, sizeof(struct iphdr));
 	else
 		skb_pull(skb, sizeof(struct ipv6hdr));
 
+	if (WARN_ON(!ib_device_try_get(&pkt->rxe->ib_dev)))
+		kfree_skb(skb);
+	else
 		rxe_rcv(skb);
 }
 

drivers/infiniband/sw/rxe/rxe_recv.c

@@ -237,8 +237,6 @@ static void rxe_rcv_mcast_pkt(struct rxe_dev *rxe, struct sk_buff *skb)
 	struct rxe_mc_elem *mce;
 	struct rxe_qp *qp;
 	union ib_gid dgid;
-	struct sk_buff *per_qp_skb;
-	struct rxe_pkt_info *per_qp_pkt;
 	int err;
 
 	if (skb->protocol == htons(ETH_P_IP))
@@ -250,10 +248,15 @@ static void rxe_rcv_mcast_pkt(struct rxe_dev *rxe, struct sk_buff *skb)
 	/* lookup mcast group corresponding to mgid, takes a ref */
 	mcg = rxe_pool_get_key(&rxe->mc_grp_pool, &dgid);
 	if (!mcg)
-		goto err1;	/* mcast group not registered */
+		goto drop;	/* mcast group not registered */
 
 	spin_lock_bh(&mcg->mcg_lock);
 
+	/* this is unreliable datagram service so we let
+	 * failures to deliver a multicast packet to a
+	 * single QP happen and just move on and try
+	 * the rest of them on the list
+	 */
 	list_for_each_entry(mce, &mcg->qp_list, qp_list) {
 		qp = mce->qp;
 
@@ -266,39 +269,47 @@ static void rxe_rcv_mcast_pkt(struct rxe_dev *rxe, struct sk_buff *skb)
 		if (err)
 			continue;
 
-		/* for all but the last qp create a new clone of the
-		 * skb and pass to the qp. If an error occurs in the
-		 * checks for the last qp in the list we need to
-		 * free the skb since it hasn't been passed on to
-		 * rxe_rcv_pkt() which would free it later.
+		/* for all but the last QP create a new clone of the
+		 * skb and pass to the QP. Pass the original skb to
+		 * the last QP in the list.
 		 */
 		if (mce->qp_list.next != &mcg->qp_list) {
-			per_qp_skb = skb_clone(skb, GFP_ATOMIC);
+			struct sk_buff *cskb;
+			struct rxe_pkt_info *cpkt;
+
+			cskb = skb_clone(skb, GFP_ATOMIC);
+			if (unlikely(!cskb))
+				continue;
+
 			if (WARN_ON(!ib_device_try_get(&rxe->ib_dev))) {
-				kfree_skb(per_qp_skb);
-				continue;
-			}
-		} else {
-			per_qp_skb = skb;
-			/* show we have consumed the skb */
-			skb = NULL;
+				kfree_skb(cskb);
+				break;
 			}
 
-		if (unlikely(!per_qp_skb))
-			continue;
-
-		per_qp_pkt = SKB_TO_PKT(per_qp_skb);
-		per_qp_pkt->qp = qp;
+			cpkt = SKB_TO_PKT(cskb);
+			cpkt->qp = qp;
 			rxe_add_ref(qp);
-		rxe_rcv_pkt(per_qp_pkt, per_qp_skb);
+			rxe_rcv_pkt(cpkt, cskb);
+		} else {
+			pkt->qp = qp;
+			rxe_add_ref(qp);
+			rxe_rcv_pkt(pkt, skb);
+			skb = NULL;	/* mark consumed */
+		}
 	}
 
 	spin_unlock_bh(&mcg->mcg_lock);
 
 	rxe_drop_ref(mcg);	/* drop ref from rxe_pool_get_key. */
 
-err1:
-	/* free skb if not consumed */
+	if (likely(!skb))
+		return;
+
+	/* This only occurs if one of the checks fails on the last
+	 * QP in the list above
+	 */
+
+drop:
 	kfree_skb(skb);
 	ib_device_put(&rxe->ib_dev);
 }