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

Pull rdma updates from Jason Gunthorpe:
 "Seveal fixes scattered across the drivers and a few new features:

   - Minor updates and bug fixes to hfi1, efa, iopob, bnxt, hns

   - Force disassociate the userspace FD when hns does an async reset

   - bnxt new features for optimized modify QP to skip certain stayes,
     CQ coalescing, better debug dumping

   - mlx5 new data placement ordering feature

   - Faster destruction of mlx5 devx HW objects

   - Improvements to RDMA CM mad handling"

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma: (51 commits)
  RDMA/bnxt_re: Correct the sequence of device suspend
  RDMA/bnxt_re: Use the default mode of congestion control
  RDMA/bnxt_re: Support different traffic class
  IB/cm: Rework sending DREQ when destroying a cm_id
  IB/cm: Do not hold reference on cm_id unless needed
  IB/cm: Explicitly mark if a response MAD is a retransmission
  RDMA/mlx5: Move events notifier registration to be after device registration
  RDMA/bnxt_re: Cache MSIx info to a local structure
  RDMA/bnxt_re: Refurbish CQ to NQ hash calculation
  RDMA/bnxt_re: Refactor NQ allocation
  RDMA/bnxt_re: Fail probe early when not enough MSI-x vectors are reserved
  RDMA/hns: Fix different dgids mapping to the same dip_idx
  RDMA/bnxt_re: Add set_func_resources support for P5/P7 adapters
  RDMA/bnxt_re: Enhance RoCE SRIOV resource configuration design
  bnxt_en: Add support for RoCE sriov configuration
  RDMA/hns: Fix NULL pointer derefernce in hns_roce_map_mr_sg()
  RDMA/hns: Fix out-of-order issue of requester when setting FENCE
  RDMA/nldev: Add IB device and net device rename events
  RDMA/mlx5: Add implementation for ufile_hw_cleanup device operation
  RDMA/core: Move ib_uverbs_file struct to uverbs_types.h
  ...

Author: torvalds

drivers/infiniband/core/cm.c

@@ -437,6 +437,7 @@ int ib_device_rename(struct ib_device *ibdev, const char *name)
 		client->rename(ibdev, client_data);
 	}
 	up_read(&ibdev->client_data_rwsem);
+	rdma_nl_notify_event(ibdev, 0, RDMA_RENAME_EVENT);
 	up_read(&devices_rwsem);
 	return 0;
 }
@@ -2759,6 +2760,7 @@ void ib_set_device_ops(struct ib_device *dev, const struct ib_device_ops *ops)
 	SET_DEVICE_OP(dev_ops, resize_cq);
 	SET_DEVICE_OP(dev_ops, set_vf_guid);
 	SET_DEVICE_OP(dev_ops, set_vf_link_state);
+	SET_DEVICE_OP(dev_ops, ufile_hw_cleanup);
 
 	SET_OBJ_SIZE(dev_ops, ib_ah);
 	SET_OBJ_SIZE(dev_ops, ib_counters);
@@ -2852,6 +2854,40 @@ static const struct rdma_nl_cbs ibnl_ls_cb_table[RDMA_NL_LS_NUM_OPS] = {
 	},
 };
 
+static int ib_netdevice_event(struct notifier_block *this,
+			      unsigned long event, void *ptr)
+{
+	struct net_device *ndev = netdev_notifier_info_to_dev(ptr);
+	struct net_device *ib_ndev;
+	struct ib_device *ibdev;
+	u32 port;
+
+	switch (event) {
+	case NETDEV_CHANGENAME:
+		ibdev = ib_device_get_by_netdev(ndev, RDMA_DRIVER_UNKNOWN);
+		if (!ibdev)
+			return NOTIFY_DONE;
+
+		rdma_for_each_port(ibdev, port) {
+			ib_ndev = ib_device_get_netdev(ibdev, port);
+			if (ndev == ib_ndev)
+				rdma_nl_notify_event(ibdev, port,
+						     RDMA_NETDEV_RENAME_EVENT);
+			dev_put(ib_ndev);
+		}
+		ib_device_put(ibdev);
+		break;
+	default:
+		break;
+	}
+
+	return NOTIFY_DONE;
+}
+
+static struct notifier_block nb_netdevice = {
+	.notifier_call = ib_netdevice_event,
+};
+
 static int __init ib_core_init(void)
 {
 	int ret = -ENOMEM;
@@ -2923,6 +2959,8 @@ static int __init ib_core_init(void)
 		goto err_parent;
 	}
 
+	register_netdevice_notifier(&nb_netdevice);
+
 	return 0;
 
 err_parent:
@@ -2952,6 +2990,7 @@ static int __init ib_core_init(void)
 
 static void __exit ib_core_cleanup(void)
 {
+	unregister_netdevice_notifier(&nb_netdevice);
 	roce_gid_mgmt_cleanup();
 	rdma_nl_unregister(RDMA_NL_LS);
 	nldev_exit();

drivers/infiniband/core/device.c

@@ -2729,6 +2729,25 @@ static const struct rdma_nl_cbs nldev_cb_table[RDMA_NLDEV_NUM_OPS] = {
 	},
 };
 
+static int fill_mon_netdev_rename(struct sk_buff *msg,
+				  struct ib_device *device, u32 port,
+				  const struct net *net)
+{
+	struct net_device *netdev = ib_device_get_netdev(device, port);
+	int ret = 0;
+
+	if (!netdev || !net_eq(dev_net(netdev), net))
+		goto out;
+
+	ret = nla_put_u32(msg, RDMA_NLDEV_ATTR_NDEV_INDEX, netdev->ifindex);
+	if (ret)
+		goto out;
+	ret = nla_put_string(msg, RDMA_NLDEV_ATTR_NDEV_NAME, netdev->name);
+out:
+	dev_put(netdev);
+	return ret;
+}
+
 static int fill_mon_netdev_association(struct sk_buff *msg,
 				       struct ib_device *device, u32 port,
 				       const struct net *net)
@@ -2793,6 +2812,18 @@ static void rdma_nl_notify_err_msg(struct ib_device *device, u32 port_num,
 				     "Failed to send RDMA monitor netdev detach event: port %d\n",
 				     port_num);
 		break;
+	case RDMA_RENAME_EVENT:
+		dev_warn_ratelimited(&device->dev,
+				     "Failed to send RDMA monitor rename device event\n");
+		break;
+
+	case RDMA_NETDEV_RENAME_EVENT:
+		netdev = ib_device_get_netdev(device, port_num);
+		dev_warn_ratelimited(&device->dev,
+				     "Failed to send RDMA monitor netdev rename event: port %d netdev %d\n",
+				     port_num, netdev->ifindex);
+		dev_put(netdev);
+		break;
 	default:
 		break;
 	}
@@ -2822,14 +2853,19 @@ int rdma_nl_notify_event(struct ib_device *device, u32 port_num,
 	switch (type) {
 	case RDMA_REGISTER_EVENT:
 	case RDMA_UNREGISTER_EVENT:
+	case RDMA_RENAME_EVENT:
 		ret = fill_nldev_handle(skb, device);
 		if (ret)
 			goto err_free;
 		break;
 	case RDMA_NETDEV_ATTACH_EVENT:
 	case RDMA_NETDEV_DETACH_EVENT:
-		ret = fill_mon_netdev_association(skb, device,
-						  port_num, net);
+		ret = fill_mon_netdev_association(skb, device, port_num, net);
+		if (ret)
+			goto err_free;
+		break;
+	case RDMA_NETDEV_RENAME_EVENT:
+		ret = fill_mon_netdev_rename(skb, device, port_num, net);
 		if (ret)
 			goto err_free;
 		break;

drivers/infiniband/core/nldev.c

@@ -58,8 +58,8 @@ void uverbs_uobject_put(struct ib_uobject *uobject)
 }
 EXPORT_SYMBOL(uverbs_uobject_put);
 
-static int uverbs_try_lock_object(struct ib_uobject *uobj,
-				  enum rdma_lookup_mode mode)
+int uverbs_try_lock_object(struct ib_uobject *uobj,
+			   enum rdma_lookup_mode mode)
 {
 	/*
 	 * When a shared access is required, we use a positive counter. Each
@@ -84,6 +84,7 @@ static int uverbs_try_lock_object(struct ib_uobject *uobj,
 	}
 	return 0;
 }
+EXPORT_SYMBOL(uverbs_try_lock_object);
 
 static void assert_uverbs_usecnt(struct ib_uobject *uobj,
 				 enum rdma_lookup_mode mode)
@@ -880,9 +881,14 @@ static void ufile_destroy_ucontext(struct ib_uverbs_file *ufile,
 static int __uverbs_cleanup_ufile(struct ib_uverbs_file *ufile,
 				  enum rdma_remove_reason reason)
 {
+	struct uverbs_attr_bundle attrs = { .ufile = ufile };
+	struct ib_ucontext *ucontext = ufile->ucontext;
+	struct ib_device *ib_dev = ucontext->device;
 	struct ib_uobject *obj, *next_obj;
 	int ret = -EINVAL;
-	struct uverbs_attr_bundle attrs = { .ufile = ufile };
+
+	if (ib_dev->ops.ufile_hw_cleanup)
+		ib_dev->ops.ufile_hw_cleanup(ufile);
 
 	/*
 	 * This shouldn't run while executing other commands on this

drivers/infiniband/core/rdma_core.c

@@ -515,6 +515,27 @@ void rdma_roce_rescan_device(struct ib_device *ib_dev)
 }
 EXPORT_SYMBOL(rdma_roce_rescan_device);
 
+/**
+ * rdma_roce_rescan_port - Rescan all of the network devices in the system
+ * and add their gids if relevant to the port of the RoCE device.
+ *
+ * @ib_dev: IB device
+ * @port: Port number
+ */
+void rdma_roce_rescan_port(struct ib_device *ib_dev, u32 port)
+{
+	struct net_device *ndev = NULL;
+
+	if (rdma_protocol_roce(ib_dev, port)) {
+		ndev = ib_device_get_netdev(ib_dev, port);
+		if (!ndev)
+			return;
+		enum_all_gids_of_dev_cb(ib_dev, port, ndev, ndev);
+		dev_put(ndev);
+	}
+}
+EXPORT_SYMBOL(rdma_roce_rescan_port);
+
 static void callback_for_addr_gid_device_scan(struct ib_device *device,
 					      u32 port,
 					      struct net_device *rdma_ndev,
@@ -575,16 +596,17 @@ static void handle_netdev_upper(struct ib_device *ib_dev, u32 port,
 	}
 }
 
-static void _roce_del_all_netdev_gids(struct ib_device *ib_dev, u32 port,
-				      struct net_device *event_ndev)
+void roce_del_all_netdev_gids(struct ib_device *ib_dev,
+			      u32 port, struct net_device *ndev)
 {
-	ib_cache_gid_del_all_netdev_gids(ib_dev, port, event_ndev);
+	ib_cache_gid_del_all_netdev_gids(ib_dev, port, ndev);
 }
+EXPORT_SYMBOL(roce_del_all_netdev_gids);
 
 static void del_netdev_upper_ips(struct ib_device *ib_dev, u32 port,
 				 struct net_device *rdma_ndev, void *cookie)
 {
-	handle_netdev_upper(ib_dev, port, cookie, _roce_del_all_netdev_gids);
+	handle_netdev_upper(ib_dev, port, cookie, roce_del_all_netdev_gids);
 }
 
 static void add_netdev_upper_ips(struct ib_device *ib_dev, u32 port,

drivers/infiniband/core/roce_gid_mgmt.c

@@ -133,35 +133,6 @@ struct ib_uverbs_completion_event_file {
 	struct ib_uverbs_event_queue		ev_queue;
 };
 
-struct ib_uverbs_file {
-	struct kref				ref;
-	struct ib_uverbs_device		       *device;
-	struct mutex				ucontext_lock;
-	/*
-	 * ucontext must be accessed via ib_uverbs_get_ucontext() or with
-	 * ucontext_lock held
-	 */
-	struct ib_ucontext		       *ucontext;
-	struct ib_uverbs_async_event_file      *default_async_file;
-	struct list_head			list;
-
-	/*
-	 * To access the uobjects list hw_destroy_rwsem must be held for write
-	 * OR hw_destroy_rwsem held for read AND uobjects_lock held.
-	 * hw_destroy_rwsem should be called across any destruction of the HW
-	 * object of an associated uobject.
-	 */
-	struct rw_semaphore	hw_destroy_rwsem;
-	spinlock_t		uobjects_lock;
-	struct list_head	uobjects;
-
-	struct mutex umap_lock;
-	struct list_head umaps;
-	struct page *disassociate_page;
-
-	struct xarray		idr;
-};
-
 struct ib_uverbs_event {
 	union {
 		struct ib_uverbs_async_event_desc	async;

drivers/infiniband/core/uverbs.h

@@ -76,6 +76,7 @@ static dev_t dynamic_uverbs_dev;
 static DEFINE_IDA(uverbs_ida);
 static int ib_uverbs_add_one(struct ib_device *device);
 static void ib_uverbs_remove_one(struct ib_device *device, void *client_data);
+static struct ib_client uverbs_client;
 
 static char *uverbs_devnode(const struct device *dev, umode_t *mode)
 {
@@ -217,6 +218,7 @@ void ib_uverbs_release_file(struct kref *ref)
 
 	if (file->disassociate_page)
 		__free_pages(file->disassociate_page, 0);
+	mutex_destroy(&file->disassociation_lock);
 	mutex_destroy(&file->umap_lock);
 	mutex_destroy(&file->ucontext_lock);
 	kfree(file);
@@ -698,8 +700,13 @@ static int ib_uverbs_mmap(struct file *filp, struct vm_area_struct *vma)
 		ret = PTR_ERR(ucontext);
 		goto out;
 	}
+
+	mutex_lock(&file->disassociation_lock);
+
 	vma->vm_ops = &rdma_umap_ops;
 	ret = ucontext->device->ops.mmap(ucontext, vma);
+
+	mutex_unlock(&file->disassociation_lock);
 out:
 	srcu_read_unlock(&file->device->disassociate_srcu, srcu_key);
 	return ret;
@@ -721,6 +728,8 @@ static void rdma_umap_open(struct vm_area_struct *vma)
 	/* We are racing with disassociation */
 	if (!down_read_trylock(&ufile->hw_destroy_rwsem))
 		goto out_zap;
+	mutex_lock(&ufile->disassociation_lock);
+
 	/*
 	 * Disassociation already completed, the VMA should already be zapped.
 	 */
@@ -732,10 +741,12 @@ static void rdma_umap_open(struct vm_area_struct *vma)
 		goto out_unlock;
 	rdma_umap_priv_init(priv, vma, opriv->entry);
 
+	mutex_unlock(&ufile->disassociation_lock);
 	up_read(&ufile->hw_destroy_rwsem);
 	return;
 
 out_unlock:
+	mutex_unlock(&ufile->disassociation_lock);
 	up_read(&ufile->hw_destroy_rwsem);
 out_zap:
 	/*
@@ -819,7 +830,7 @@ void uverbs_user_mmap_disassociate(struct ib_uverbs_file *ufile)
 {
 	struct rdma_umap_priv *priv, *next_priv;
 
-	lockdep_assert_held(&ufile->hw_destroy_rwsem);
+	mutex_lock(&ufile->disassociation_lock);
 
 	while (1) {
 		struct mm_struct *mm = NULL;
@@ -845,8 +856,10 @@ void uverbs_user_mmap_disassociate(struct ib_uverbs_file *ufile)
 			break;
 		}
 		mutex_unlock(&ufile->umap_lock);
-		if (!mm)
+		if (!mm) {
+			mutex_unlock(&ufile->disassociation_lock);
 			return;
+		}
 
 		/*
 		 * The umap_lock is nested under mmap_lock since it used within
@@ -876,7 +889,31 @@ void uverbs_user_mmap_disassociate(struct ib_uverbs_file *ufile)
 		mmap_read_unlock(mm);
 		mmput(mm);
 	}
+
+	mutex_unlock(&ufile->disassociation_lock);
+}
+
+/**
+ * rdma_user_mmap_disassociate() - Revoke mmaps for a device
+ * @device: device to revoke
+ *
+ * This function should be called by drivers that need to disable mmaps for the
+ * device, for instance because it is going to be reset.
+ */
+void rdma_user_mmap_disassociate(struct ib_device *device)
+{
+	struct ib_uverbs_device *uverbs_dev =
+		ib_get_client_data(device, &uverbs_client);
+	struct ib_uverbs_file *ufile;
+
+	mutex_lock(&uverbs_dev->lists_mutex);
+	list_for_each_entry(ufile, &uverbs_dev->uverbs_file_list, list) {
+		if (ufile->ucontext)
+			uverbs_user_mmap_disassociate(ufile);
+	}
+	mutex_unlock(&uverbs_dev->lists_mutex);
 }
+EXPORT_SYMBOL(rdma_user_mmap_disassociate);
 
 /*
  * ib_uverbs_open() does not need the BKL:
@@ -947,6 +984,8 @@ static int ib_uverbs_open(struct inode *inode, struct file *filp)
 	mutex_init(&file->umap_lock);
 	INIT_LIST_HEAD(&file->umaps);
 
+	mutex_init(&file->disassociation_lock);
+
 	filp->private_data = file;
 	list_add_tail(&file->list, &dev->uverbs_file_list);
 	mutex_unlock(&dev->lists_mutex);

drivers/infiniband/core/uverbs_main.c

@@ -4,4 +4,5 @@ ccflags-y := -I $(srctree)/drivers/net/ethernet/broadcom/bnxt
 obj-$(CONFIG_INFINIBAND_BNXT_RE) += bnxt_re.o
 bnxt_re-y := main.o ib_verbs.o \
 	     qplib_res.o qplib_rcfw.o	\
-	     qplib_sp.o qplib_fp.o  hw_counters.o
+	     qplib_sp.o qplib_fp.o  hw_counters.o	\
+	     debugfs.o