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

Pull rdma updates from Jason Gunthorpe:
 "This has been a smaller cycle with many of the commits being smallish
  code fixes and improvements across the drivers.

   - Driver updates for bnxt_re, cxgb4, hfi1, hns, mlx5, nes, qedr, and
     rxe

   - Memory window support in hns

   - mlx5 user API 'flow mutate/steering' allows accessing the full
     packet mangling and matching machinery from user space

   - Support inter-working with verbs API calls in the 'devx' mlx5 user
     API, and provide options to use devx with less privilege

   - Modernize the use of syfs and the device interface to use attribute
     groups and cdev properly for uverbs, and clean up some of the core
     code's device list management

   - More progress on net namespaces for RDMA devices

   - Consolidate driver BAR mmapping support into core code helpers and
     rework how RDMA holds poitners to mm_struct for get_user_pages
     cases

   - First pass to use 'dev_name' instead of ib_device->name

   - Device renaming for RDMA devices"

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma: (242 commits)
  IB/mlx5: Add support for extended atomic operations
  RDMA/core: Fix comment for hw stats init for port == 0
  RDMA/core: Refactor ib_register_device() function
  RDMA/core: Fix unwinding flow in case of error to register device
  ib_srp: Remove WARN_ON in srp_terminate_io()
  IB/mlx5: Allow scatter to CQE without global signaled WRs
  IB/mlx5: Verify that driver supports user flags
  IB/mlx5: Support scatter to CQE for DC transport type
  RDMA/drivers: Use core provided API for registering device attributes
  RDMA/core: Allow existing drivers to set one sysfs group per device
  IB/rxe: Remove unnecessary enum values
  RDMA/umad: Use kernel API to allocate umad indexes
  RDMA/uverbs: Use kernel API to allocate uverbs indexes
  RDMA/core: Increase total number of RDMA ports across all devices
  IB/mlx4: Add port and TID to MAD debug print
  IB/mlx4: Enable debug print of SMPs
  RDMA/core: Rename ports_parent to ports_kobj
  RDMA/core: Do not expose unsupported counters
  IB/mlx4: Refer to the device kobject instead of ports_parent
  RDMA/nldev: Allow IB device rename through RDMA netlink
  ...

31 files changed, 1678 insertions, 1093 deletions

diff --git a/drivers/infiniband/core/addr.c b/drivers/infiniband/core/addr.c
index 46b855a42884..0dce94e3c495 100644
--- a/drivers/infiniband/core/addr.c
+++ b/drivers/infiniband/core/addr.c
@@ -45,6 +45,7 @@
 #include <net/addrconf.h>
 #include <net/ip6_route.h>
 #include <rdma/ib_addr.h>
+#include <rdma/ib_sa.h>
 #include <rdma/ib.h>
 #include <rdma/rdma_netlink.h>
 #include <net/netlink.h>
@@ -61,6 +62,7 @@ struct addr_req {
 			 struct rdma_dev_addr *addr, void *context);
 	unsigned long timeout;
 	struct delayed_work work;
+	bool resolve_by_gid_attr;	/* Consider gid attr in resolve phase */
 	int status;
 	u32 seq;
 };
@@ -219,60 +221,75 @@ int rdma_addr_size_kss(struct __kernel_sockaddr_storage *addr)
 }
 EXPORT_SYMBOL(rdma_addr_size_kss);
 
-void rdma_copy_addr(struct rdma_dev_addr *dev_addr,
-		    const struct net_device *dev,
-		    const unsigned char *dst_dev_addr)
+/**
+ * rdma_copy_src_l2_addr - Copy netdevice source addresses
+ * @dev_addr:	Destination address pointer where to copy the addresses
+ * @dev:	Netdevice whose source addresses to copy
+ *
+ * rdma_copy_src_l2_addr() copies source addresses from the specified netdevice.
+ * This includes unicast address, broadcast address, device type and
+ * interface index.
+ */
+void rdma_copy_src_l2_addr(struct rdma_dev_addr *dev_addr,
+			   const struct net_device *dev)
 {
 	dev_addr->dev_type = dev->type;
 	memcpy(dev_addr->src_dev_addr, dev->dev_addr, MAX_ADDR_LEN);
 	memcpy(dev_addr->broadcast, dev->broadcast, MAX_ADDR_LEN);
-	if (dst_dev_addr)
-		memcpy(dev_addr->dst_dev_addr, dst_dev_addr, MAX_ADDR_LEN);
 	dev_addr->bound_dev_if = dev->ifindex;
 }
-EXPORT_SYMBOL(rdma_copy_addr);
+EXPORT_SYMBOL(rdma_copy_src_l2_addr);
 
-int rdma_translate_ip(const struct sockaddr *addr,
-		      struct rdma_dev_addr *dev_addr)
+static struct net_device *
+rdma_find_ndev_for_src_ip_rcu(struct net *net, const struct sockaddr *src_in)
 {
-	struct net_device *dev;
+	struct net_device *dev = NULL;
+	int ret = -EADDRNOTAVAIL;
 
-	if (dev_addr->bound_dev_if) {
-		dev = dev_get_by_index(dev_addr->net, dev_addr->bound_dev_if);
-		if (!dev)
-			return -ENODEV;
-		rdma_copy_addr(dev_addr, dev, NULL);
-		dev_put(dev);
-		return 0;
-	}
-
-	switch (addr->sa_family) {
+	switch (src_in->sa_family) {
 	case AF_INET:
-		dev = ip_dev_find(dev_addr->net,
-			((const struct sockaddr_in *)addr)->sin_addr.s_addr);
-
-		if (!dev)
-			return -EADDRNOTAVAIL;
-
-		rdma_copy_addr(dev_addr, dev, NULL);
-		dev_put(dev);
+		dev = __ip_dev_find(net,
+				    ((const struct sockaddr_in *)src_in)->sin_addr.s_addr,
+				    false);
+		if (dev)
+			ret = 0;
 		break;
 #if IS_ENABLED(CONFIG_IPV6)
 	case AF_INET6:
-		rcu_read_lock();
-		for_each_netdev_rcu(dev_addr->net, dev) {
-			if (ipv6_chk_addr(dev_addr->net,
-					  &((const struct sockaddr_in6 *)addr)->sin6_addr,
+		for_each_netdev_rcu(net, dev) {
+			if (ipv6_chk_addr(net,
+					  &((const struct sockaddr_in6 *)src_in)->sin6_addr,
 					  dev, 1)) {
-				rdma_copy_addr(dev_addr, dev, NULL);
+				ret = 0;
 				break;
 			}
 		}
-		rcu_read_unlock();
 		break;
 #endif
 	}
-	return 0;
+	return ret ? ERR_PTR(ret) : dev;
+}
+
+int rdma_translate_ip(const struct sockaddr *addr,
+		      struct rdma_dev_addr *dev_addr)
+{
+	struct net_device *dev;
+
+	if (dev_addr->bound_dev_if) {
+		dev = dev_get_by_index(dev_addr->net, dev_addr->bound_dev_if);
+		if (!dev)
+			return -ENODEV;
+		rdma_copy_src_l2_addr(dev_addr, dev);
+		dev_put(dev);
+		return 0;
+	}
+
+	rcu_read_lock();
+	dev = rdma_find_ndev_for_src_ip_rcu(dev_addr->net, addr);
+	if (!IS_ERR(dev))
+		rdma_copy_src_l2_addr(dev_addr, dev);
+	rcu_read_unlock();
+	return PTR_ERR_OR_ZERO(dev);
 }
 EXPORT_SYMBOL(rdma_translate_ip);
 
@@ -295,15 +312,12 @@ static void queue_req(struct addr_req *req)
 	spin_unlock_bh(&lock);
 }
 
-static int ib_nl_fetch_ha(const struct dst_entry *dst,
-			  struct rdma_dev_addr *dev_addr,
+static int ib_nl_fetch_ha(struct rdma_dev_addr *dev_addr,
 			  const void *daddr, u32 seq, u16 family)
 {
-	if (rdma_nl_chk_listeners(RDMA_NL_GROUP_LS))
+	if (!rdma_nl_chk_listeners(RDMA_NL_GROUP_LS))
 		return -EADDRNOTAVAIL;
 
-	/* We fill in what we can, the response will fill the rest */
-	rdma_copy_addr(dev_addr, dst->dev, NULL);
 	return ib_nl_ip_send_msg(dev_addr, daddr, seq, family);
 }
 
@@ -322,7 +336,7 @@ static int dst_fetch_ha(const struct dst_entry *dst,
 		neigh_event_send(n, NULL);
 		ret = -ENODATA;
 	} else {
-		rdma_copy_addr(dev_addr, dst->dev, n->ha);
+		memcpy(dev_addr->dst_dev_addr, n->ha, MAX_ADDR_LEN);
 	}
 
 	neigh_release(n);
@@ -356,18 +370,22 @@ static int fetch_ha(const struct dst_entry *dst, struct rdma_dev_addr *dev_addr,
 		(const void *)&dst_in6->sin6_addr;
 	sa_family_t family = dst_in->sa_family;
 
-	/* Gateway + ARPHRD_INFINIBAND -> IB router */
-	if (has_gateway(dst, family) && dst->dev->type == ARPHRD_INFINIBAND)
-		return ib_nl_fetch_ha(dst, dev_addr, daddr, seq, family);
+	/* If we have a gateway in IB mode then it must be an IB network */
+	if (has_gateway(dst, family) && dev_addr->network == RDMA_NETWORK_IB)
+		return ib_nl_fetch_ha(dev_addr, daddr, seq, family);
 	else
 		return dst_fetch_ha(dst, dev_addr, daddr);
 }
 
-static int addr4_resolve(struct sockaddr_in *src_in,
-			 const struct sockaddr_in *dst_in,
+static int addr4_resolve(struct sockaddr *src_sock,
+			 const struct sockaddr *dst_sock,
 			 struct rdma_dev_addr *addr,
 			 struct rtable **prt)
 {
+	struct sockaddr_in *src_in = (struct sockaddr_in *)src_sock;
+	const struct sockaddr_in *dst_in =
+			(const struct sockaddr_in *)dst_sock;
+
 	__be32 src_ip = src_in->sin_addr.s_addr;
 	__be32 dst_ip = dst_in->sin_addr.s_addr;
 	struct rtable *rt;
@@ -383,16 +401,8 @@ static int addr4_resolve(struct sockaddr_in *src_in,
 	if (ret)
 		return ret;
 
-	src_in->sin_family = AF_INET;
 	src_in->sin_addr.s_addr = fl4.saddr;
 
-	/* If there's a gateway and type of device not ARPHRD_INFINIBAND, we're
-	 * definitely in RoCE v2 (as RoCE v1 isn't routable) set the network
-	 * type accordingly.
-	 */
-	if (rt->rt_uses_gateway && rt->dst.dev->type != ARPHRD_INFINIBAND)
-		addr->network = RDMA_NETWORK_IPV4;
-
 	addr->hoplimit = ip4_dst_hoplimit(&rt->dst);
 
 	*prt = rt;
@@ -400,14 +410,16 @@ static int addr4_resolve(struct sockaddr_in *src_in,
 }
 
 #if IS_ENABLED(CONFIG_IPV6)
-static int addr6_resolve(struct sockaddr_in6 *src_in,
-			 const struct sockaddr_in6 *dst_in,
+static int addr6_resolve(struct sockaddr *src_sock,
+			 const struct sockaddr *dst_sock,
 			 struct rdma_dev_addr *addr,
 			 struct dst_entry **pdst)
 {
+	struct sockaddr_in6 *src_in = (struct sockaddr_in6 *)src_sock;
+	const struct sockaddr_in6 *dst_in =
+				(const struct sockaddr_in6 *)dst_sock;
 	struct flowi6 fl6;
 	struct dst_entry *dst;
-	struct rt6_info *rt;
 	int ret;
 
 	memset(&fl6, 0, sizeof fl6);
@@ -419,19 +431,8 @@ static int addr6_resolve(struct sockaddr_in6 *src_in,
 	if (ret < 0)
 		return ret;
 
-	rt = (struct rt6_info *)dst;
-	if (ipv6_addr_any(&src_in->sin6_addr)) {
-		src_in->sin6_family = AF_INET6;
+	if (ipv6_addr_any(&src_in->sin6_addr))
 		src_in->sin6_addr = fl6.saddr;
-	}
-
-	/* If there's a gateway and type of device not ARPHRD_INFINIBAND, we're
-	 * definitely in RoCE v2 (as RoCE v1 isn't routable) set the network
-	 * type accordingly.
-	 */
-	if (rt->rt6i_flags & RTF_GATEWAY &&
-	    ip6_dst_idev(dst)->dev->type != ARPHRD_INFINIBAND)
-		addr->network = RDMA_NETWORK_IPV6;
 
 	addr->hoplimit = ip6_dst_hoplimit(dst);
 
@@ -439,8 +440,8 @@ static int addr6_resolve(struct sockaddr_in6 *src_in,
 	return 0;
 }
 #else
-static int addr6_resolve(struct sockaddr_in6 *src_in,
-			 const struct sockaddr_in6 *dst_in,
+static int addr6_resolve(struct sockaddr *src_sock,
+			 const struct sockaddr *dst_sock,
 			 struct rdma_dev_addr *addr,
 			 struct dst_entry **pdst)
 {
@@ -451,36 +452,110 @@ static int addr6_resolve(struct sockaddr_in6 *src_in,
 static int addr_resolve_neigh(const struct dst_entry *dst,
 			      const struct sockaddr *dst_in,
 			      struct rdma_dev_addr *addr,
+			      unsigned int ndev_flags,
 			      u32 seq)
 {
-	if (dst->dev->flags & IFF_LOOPBACK) {
-		int ret;
+	int ret = 0;
 
-		ret = rdma_translate_ip(dst_in, addr);
-		if (!ret)
-			memcpy(addr->dst_dev_addr, addr->src_dev_addr,
-			       MAX_ADDR_LEN);
+	if (ndev_flags & IFF_LOOPBACK) {
+		memcpy(addr->dst_dev_addr, addr->src_dev_addr, MAX_ADDR_LEN);
+	} else {
+		if (!(ndev_flags & IFF_NOARP)) {
+			/* If the device doesn't do ARP internally */
+			ret = fetch_ha(dst, addr, dst_in, seq);
+		}
+	}
+	return ret;
+}
 
-		return ret;
+static int copy_src_l2_addr(struct rdma_dev_addr *dev_addr,
+			    const struct sockaddr *dst_in,
+			    const struct dst_entry *dst,
+			    const struct net_device *ndev)
+{
+	int ret = 0;
+
+	if (dst->dev->flags & IFF_LOOPBACK)
+		ret = rdma_translate_ip(dst_in, dev_addr);
+	else
+		rdma_copy_src_l2_addr(dev_addr, dst->dev);
+
+	/*
+	 * If there's a gateway and type of device not ARPHRD_INFINIBAND,
+	 * we're definitely in RoCE v2 (as RoCE v1 isn't routable) set the
+	 * network type accordingly.
+	 */
+	if (has_gateway(dst, dst_in->sa_family) &&
+	    ndev->type != ARPHRD_INFINIBAND)
+		dev_addr->network = dst_in->sa_family == AF_INET ?
+						RDMA_NETWORK_IPV4 :
+						RDMA_NETWORK_IPV6;
+	else
+		dev_addr->network = RDMA_NETWORK_IB;
+
+	return ret;
+}
+
+static int rdma_set_src_addr_rcu(struct rdma_dev_addr *dev_addr,
+				 unsigned int *ndev_flags,
+				 const struct sockaddr *dst_in,
+				 const struct dst_entry *dst)
+{
+	struct net_device *ndev = READ_ONCE(dst->dev);
+
+	*ndev_flags = ndev->flags;
+	/* A physical device must be the RDMA device to use */
+	if (ndev->flags & IFF_LOOPBACK) {
+		/*
+		 * RDMA (IB/RoCE, iWarp) doesn't run on lo interface or
+		 * loopback IP address. So if route is resolved to loopback
+		 * interface, translate that to a real ndev based on non
+		 * loopback IP address.
+		 */
+		ndev = rdma_find_ndev_for_src_ip_rcu(dev_net(ndev), dst_in);
+		if (IS_ERR(ndev))
+			return -ENODEV;
 	}
 
-	/* If the device doesn't do ARP internally */
-	if (!(dst->dev->flags & IFF_NOARP))
-		return fetch_ha(dst, addr, dst_in, seq);
+	return copy_src_l2_addr(dev_addr, dst_in, dst, ndev);
+}
+
+static int set_addr_netns_by_gid_rcu(struct rdma_dev_addr *addr)
+{
+	struct net_device *ndev;
 
-	rdma_copy_addr(addr, dst->dev, NULL);
+	ndev = rdma_read_gid_attr_ndev_rcu(addr->sgid_attr);
+	if (IS_ERR(ndev))
+		return PTR_ERR(ndev);
 
+	/*
+	 * Since we are holding the rcu, reading net and ifindex
+	 * are safe without any additional reference; because
+	 * change_net_namespace() in net/core/dev.c does rcu sync
+	 * after it changes the state to IFF_DOWN and before
+	 * updating netdev fields {net, ifindex}.
+	 */
+	addr->net = dev_net(ndev);
+	addr->bound_dev_if = ndev->ifindex;
 	return 0;
 }
 
+static void rdma_addr_set_net_defaults(struct rdma_dev_addr *addr)
+{
+	addr->net = &init_net;
+	addr->bound_dev_if = 0;
+}
+
 static int addr_resolve(struct sockaddr *src_in,
 			const struct sockaddr *dst_in,
 			struct rdma_dev_addr *addr,
 			bool resolve_neigh,
+			bool resolve_by_gid_attr,
 			u32 seq)
 {
-	struct net_device *ndev;
-	struct dst_entry *dst;
+	struct dst_entry *dst = NULL;
+	unsigned int ndev_flags = 0;
+	struct rtable *rt = NULL;
 	int ret;
 
 	if (!addr->net) {
@@ -488,58 +563,55 @@ static int addr_resolve(struct sockaddr *src_in,
 		return -EINVAL;
 	}
 
-	if (src_in->sa_family == AF_INET) {
-		struct rtable *rt = NULL;
-		const struct sockaddr_in *dst_in4 =
-			(const struct sockaddr_in *)dst_in;
-
-		ret = addr4_resolve((struct sockaddr_in *)src_in,
-				    dst_in4, addr, &rt);
-		if (ret)
-			return ret;
-
-		if (resolve_neigh)
-			ret = addr_resolve_neigh(&rt->dst, dst_in, addr, seq);
-
-		if (addr->bound_dev_if) {
-			ndev = dev_get_by_index(addr->net, addr->bound_dev_if);
-		} else {
-			ndev = rt->dst.dev;
-			dev_hold(ndev);
+	rcu_read_lock();
+	if (resolve_by_gid_attr) {
+		if (!addr->sgid_attr) {
+			rcu_read_unlock();
+			pr_warn_ratelimited("%s: missing gid_attr\n", __func__);
+			return -EINVAL;
 		}
-
-		ip_rt_put(rt);
-	} else {
-		const struct sockaddr_in6 *dst_in6 =
-			(const struct sockaddr_in6 *)dst_in;
-
-		ret = addr6_resolve((struct sockaddr_in6 *)src_in,
-				    dst_in6, addr,
-				    &dst);
-		if (ret)
+		/*
+		 * If the request is for a specific gid attribute of the
+		 * rdma_dev_addr, derive net from the netdevice of the
+		 * GID attribute.
+		 */
+		ret = set_addr_netns_by_gid_rcu(addr);
+		if (ret) {
+			rcu_read_unlock();
 			return ret;
-
-		if (resolve_neigh)
-			ret = addr_resolve_neigh(dst, dst_in, addr, seq);
-
-		if (addr->bound_dev_if) {
-			ndev = dev_get_by_index(addr->net, addr->bound_dev_if);
-		} else {
-			ndev = dst->dev;
-			dev_hold(ndev);
 		}
-
-		dst_release(dst);
 	}
-
-	if (ndev) {
-		if (ndev->flags & IFF_LOOPBACK)
-			ret = rdma_translate_ip(dst_in, addr);
-		else
-			addr->bound_dev_if = ndev->ifindex;
-		dev_put(ndev);
+	if (src_in->sa_family == AF_INET) {
+		ret = addr4_resolve(src_in, dst_in, addr, &rt);
+		dst = &rt->dst;
+	} else {
+		ret = addr6_resolve(src_in, dst_in, addr, &dst);
 	}
+	if (ret) {
+		rcu_read_unlock();
+		goto done;
+	}
+	ret = rdma_set_src_addr_rcu(addr, &ndev_flags, dst_in, dst);
+	rcu_read_unlock();
+
+	/*
+	 * Resolve neighbor destination address if requested and
+	 * only if src addr translation didn't fail.
+	 */
+	if (!ret && resolve_neigh)
+		ret = addr_resolve_neigh(dst, dst_in, addr, ndev_flags, seq);
 
+	if (src_in->sa_family == AF_INET)
+		ip_rt_put(rt);
+	else
+		dst_release(dst);
+done:
+	/*
+	 * Clear the addr net to go back to its original state, only if it was
+	 * derived from GID attribute in this context.
+	 */
+	if (resolve_by_gid_attr)
+		rdma_addr_set_net_defaults(addr);
 	return ret;
 }
 
@@ -554,7 +626,8 @@ static void process_one_req(struct work_struct *_work)
 		src_in = (struct sockaddr *)&req->src_addr;
 		dst_in = (struct sockaddr *)&req->dst_addr;
 		req->status = addr_resolve(src_in, dst_in, req->addr,
-					   true, req->seq);
+					   true, req->resolve_by_gid_attr,
+					   req->seq);
 		if (req->status && time_after_eq(jiffies, req->timeout)) {
 			req->status = -ETIMEDOUT;
 		} else if (req->status == -ENODATA) {
@@ -586,10 +659,10 @@ static void process_one_req(struct work_struct *_work)
 }
 
 int rdma_resolve_ip(struct sockaddr *src_addr, const struct sockaddr *dst_addr,
-		    struct rdma_dev_addr *addr, int timeout_ms,
+		    struct rdma_dev_addr *addr, unsigned long timeout_ms,
 		    void (*callback)(int status, struct sockaddr *src_addr,
 				     struct rdma_dev_addr *addr, void *context),
-		    void *context)
+		    bool resolve_by_gid_attr, void *context)
 {
 	struct sockaddr *src_in, *dst_in;
 	struct addr_req *req;
@@ -617,10 +690,12 @@ int rdma_resolve_ip(struct sockaddr *src_addr, const struct sockaddr *dst_addr,
 	req->addr = addr;
 	req->callback = callback;
 	req->context = context;
+	req->resolve_by_gid_attr = resolve_by_gid_attr;
 	INIT_DELAYED_WORK(&req->work, process_one_req);
 	req->seq = (u32)atomic_inc_return(&ib_nl_addr_request_seq);
 
-	req->status = addr_resolve(src_in, dst_in, addr, true, req->seq);
+	req->status = addr_resolve(src_in, dst_in, addr, true,
+				   req->resolve_by_gid_attr, req->seq);
 	switch (req->status) {
 	case 0:
 		req->timeout = jiffies;
@@ -641,25 +716,53 @@ err:
 }
 EXPORT_SYMBOL(rdma_resolve_ip);
 
-int rdma_resolve_ip_route(struct sockaddr *src_addr,
-			  const struct sockaddr *dst_addr,
-			  struct rdma_dev_addr *addr)
+int roce_resolve_route_from_path(struct sa_path_rec *rec,
+				 const struct ib_gid_attr *attr)
 {
-	struct sockaddr_storage ssrc_addr = {};
-	struct sockaddr *src_in = (struct sockaddr *)&ssrc_addr;
+	union {
+		struct sockaddr     _sockaddr;
+		struct sockaddr_in  _sockaddr_in;
+		struct sockaddr_in6 _sockaddr_in6;
+	} sgid, dgid;
+	struct rdma_dev_addr dev_addr = {};
+	int ret;
 
-	if (src_addr) {
-		if (src_addr->sa_family != dst_addr->sa_family)
-			return -EINVAL;
+	if (rec->roce.route_resolved)
+		return 0;
 
-		memcpy(src_in, src_addr, rdma_addr_size(src_addr));
-	} else {
-		src_in->sa_family = dst_addr->sa_family;
-	}
+	rdma_gid2ip(&sgid._sockaddr, &rec->sgid);
+	rdma_gid2ip(&dgid._sockaddr, &rec->dgid);
+
+	if (sgid._sockaddr.sa_family != dgid._sockaddr.sa_family)
+		return -EINVAL;
+
+	if (!attr || !attr->ndev)
+		return -EINVAL;
+
+	dev_addr.net = &init_net;
+	dev_addr.sgid_attr = attr;
+
+	ret = addr_resolve(&sgid._sockaddr, &dgid._sockaddr,
+			   &dev_addr, false, true, 0);
+	if (ret)
+		return ret;
+
+	if ((dev_addr.network == RDMA_NETWORK_IPV4 ||
+	     dev_addr.network == RDMA_NETWORK_IPV6) &&
+	    rec->rec_type != SA_PATH_REC_TYPE_ROCE_V2)
+		return -EINVAL;
 
-	return addr_resolve(src_in, dst_addr, addr, false, 0);
+	rec->roce.route_resolved = true;
+	return 0;
 }
 
+/**
+ * rdma_addr_cancel - Cancel resolve ip request
+ * @addr:	Pointer to address structure given previously
+ *		during rdma_resolve_ip().
+ * rdma_addr_cancel() is synchronous function which cancels any pending
+ * request if there is any.
+ */
 void rdma_addr_cancel(struct rdma_dev_addr *addr)
 {
 	struct addr_req *req, *temp_req;
@@ -687,11 +790,6 @@ void rdma_addr_cancel(struct rdma_dev_addr *addr)
 	 * guarentees no work is running and none will be started.
 	 */
 	cancel_delayed_work_sync(&found->work);
-
-	if (found->callback)
-		found->callback(-ECANCELED, (struct sockaddr *)&found->src_addr,
-			      found->addr, found->context);
-
 	kfree(found);
 }
 EXPORT_SYMBOL(rdma_addr_cancel);
@@ -710,7 +808,7 @@ static void resolve_cb(int status, struct sockaddr *src_addr,
 
 int rdma_addr_find_l2_eth_by_grh(const union ib_gid *sgid,
 				 const union ib_gid *dgid,
-				 u8 *dmac, const struct net_device *ndev,
+				 u8 *dmac, const struct ib_gid_attr *sgid_attr,
 				 int *hoplimit)
 {
 	struct rdma_dev_addr dev_addr;
@@ -726,12 +824,12 @@ int rdma_addr_find_l2_eth_by_grh(const union ib_gid *sgid,
 	rdma_gid2ip(&dgid_addr._sockaddr, dgid);
 
 	memset(&dev_addr, 0, sizeof(dev_addr));
-	dev_addr.bound_dev_if = ndev->ifindex;
 	dev_addr.net = &init_net;
+	dev_addr.sgid_attr = sgid_attr;
 
 	init_completion(&ctx.comp);
 	ret = rdma_resolve_ip(&sgid_addr._sockaddr, &dgid_addr._sockaddr,
-			      &dev_addr, 1000, resolve_cb, &ctx);
+			      &dev_addr, 1000, resolve_cb, true, &ctx);
 	if (ret)
 		return ret;
 
diff --git a/drivers/infiniband/core/cache.c b/drivers/infiniband/core/cache.c
index 3208ad6ad540..5b2fce4a7091 100644
--- a/drivers/infiniband/core/cache.c
+++ b/drivers/infiniband/core/cache.c
@@ -212,9 +212,8 @@ static void free_gid_entry_locked(struct ib_gid_table_entry *entry)
 	u8 port_num = entry->attr.port_num;
 	struct ib_gid_table *table = rdma_gid_table(device, port_num);
 
-	pr_debug("%s device=%s port=%d index=%d gid %pI6\n", __func__,
-		 device->name, port_num, entry->attr.index,
-		 entry->attr.gid.raw);
+	dev_dbg(&device->dev, "%s port=%d index=%d gid %pI6\n", __func__,
+		port_num, entry->attr.index, entry->attr.gid.raw);
 
 	if (rdma_cap_roce_gid_table(device, port_num) &&
 	    entry->state != GID_TABLE_ENTRY_INVALID)
@@ -289,9 +288,9 @@ static void store_gid_entry(struct ib_gid_table *table,
 {
 	entry->state = GID_TABLE_ENTRY_VALID;
 
-	pr_debug("%s device=%s port=%d index=%d gid %pI6\n", __func__,
-		 entry->attr.device->name, entry->attr.port_num,
-		 entry->attr.index, entry->attr.gid.raw);
+	dev_dbg(&entry->attr.device->dev, "%s port=%d index=%d gid %pI6\n",
+		__func__, entry->attr.port_num, entry->attr.index,
+		entry->attr.gid.raw);
 
 	lockdep_assert_held(&table->lock);
 	write_lock_irq(&table->rwlock);
@@ -320,17 +319,16 @@ static int add_roce_gid(struct ib_gid_table_entry *entry)
 	int ret;
 
 	if (!attr->ndev) {
-		pr_err("%s NULL netdev device=%s port=%d index=%d\n",
-		       __func__, attr->device->name, attr->port_num,
-		       attr->index);
+		dev_err(&attr->device->dev, "%s NULL netdev port=%d index=%d\n",
+			__func__, attr->port_num, attr->index);
 		return -EINVAL;
 	}
 	if (rdma_cap_roce_gid_table(attr->device, attr->port_num)) {
 		ret = attr->device->add_gid(attr, &entry->context);
 		if (ret) {
-			pr_err("%s GID add failed device=%s port=%d index=%d\n",
-			       __func__, attr->device->name, attr->port_num,
-			       attr->index);
+			dev_err(&attr->device->dev,
+				"%s GID add failed port=%d index=%d\n",
+				__func__, attr->port_num, attr->index);
 			return ret;
 		}
 	}
@@ -353,9 +351,8 @@ static void del_gid(struct ib_device *ib_dev, u8 port,
 
 	lockdep_assert_held(&table->lock);
 
-	pr_debug("%s device=%s port=%d index=%d gid %pI6\n", __func__,
-		 ib_dev->name, port, ix,
-		 table->data_vec[ix]->attr.gid.raw);
+	dev_dbg(&ib_dev->dev, "%s port=%d index=%d gid %pI6\n", __func__, port,
+		ix, table->data_vec[ix]->attr.gid.raw);
 
 	write_lock_irq(&table->rwlock);
 	entry = table->data_vec[ix];
@@ -782,9 +779,9 @@ static void release_gid_table(struct ib_device *device, u8 port,
 		if (is_gid_entry_free(table->data_vec[i]))
 			continue;
 		if (kref_read(&table->data_vec[i]->kref) > 1) {
-			pr_err("GID entry ref leak for %s (index %d) ref=%d\n",
-			       device->name, i,
-			       kref_read(&table->data_vec[i]->kref));
+			dev_err(&device->dev,
+				"GID entry ref leak for index %d ref=%d\n", i,
+				kref_read(&table->data_vec[i]->kref));
 			leak = true;
 		}
 	}
@@ -1252,6 +1249,39 @@ void rdma_hold_gid_attr(const struct ib_gid_attr *attr)
 }
 EXPORT_SYMBOL(rdma_hold_gid_attr);
 
+/**
+ * rdma_read_gid_attr_ndev_rcu - Read GID attribute netdevice
+ * which must be in UP state.
+ *
+ * @attr:Pointer to the GID attribute
+ *
+ * Returns pointer to netdevice if the netdevice was attached to GID and
+ * netdevice is in UP state. Caller must hold RCU lock as this API
+ * reads the netdev flags which can change while netdevice migrates to
+ * different net namespace. Returns ERR_PTR with error code otherwise.
+ *
+ */
+struct net_device *rdma_read_gid_attr_ndev_rcu(const struct ib_gid_attr *attr)
+{
+	struct ib_gid_table_entry *entry =
+			container_of(attr, struct ib_gid_table_entry, attr);
+	struct ib_device *device = entry->attr.device;
+	struct net_device *ndev = ERR_PTR(-ENODEV);
+	u8 port_num = entry->attr.port_num;
+	struct ib_gid_table *table;
+	unsigned long flags;
+	bool valid;
+
+	table = rdma_gid_table(device, port_num);
+
+	read_lock_irqsave(&table->rwlock, flags);
+	valid = is_gid_entry_valid(table->data_vec[attr->index]);
+	if (valid && attr->ndev && (READ_ONCE(attr->ndev->flags) & IFF_UP))
+		ndev = attr->ndev;
+	read_unlock_irqrestore(&table->rwlock, flags);
+	return ndev;
+}
+
 static int config_non_roce_gid_cache(struct ib_device *device,
 				     u8 port, int gid_tbl_len)
 {
@@ -1270,8 +1300,9 @@ static int config_non_roce_gid_cache(struct ib_device *device,
 			continue;
 		ret = device->query_gid(device, port, i, &gid_attr.gid);
 		if (ret) {
-			pr_warn("query_gid failed (%d) for %s (index %d)\n",
-				ret, device->name, i);
+			dev_warn(&device->dev,
+				 "query_gid failed (%d) for index %d\n", ret,
+				 i);
 			goto err;
 		}
 		gid_attr.index = i;
@@ -1300,8 +1331,7 @@ static void ib_cache_update(struct ib_device *device,
 
 	ret = ib_query_port(device, port, tprops);
 	if (ret) {
-		pr_warn("ib_query_port failed (%d) for %s\n",
-			ret, device->name);
+		dev_warn(&device->dev, "ib_query_port failed (%d)\n", ret);
 		goto err;
 	}
 
@@ -1323,8 +1353,9 @@ static void ib_cache_update(struct ib_device *device,
 	for (i = 0; i < pkey_cache->table_len; ++i) {
 		ret = ib_query_pkey(device, port, i, pkey_cache->table + i);
 		if (ret) {
-			pr_warn("ib_query_pkey failed (%d) for %s (index %d)\n",
-				ret, device->name, i);
+			dev_warn(&device->dev,
+				 "ib_query_pkey failed (%d) for index %d\n",
+				 ret, i);
 			goto err;
 		}
 	}
diff --git a/drivers/infiniband/core/cm.c b/drivers/infiniband/core/cm.c
index 6e39c27dca8e..edb2cb758be7 100644
--- a/drivers/infiniband/core/cm.c
+++ b/drivers/infiniband/core/cm.c
@@ -3292,8 +3292,11 @@ static int cm_lap_handler(struct cm_work *work)
 	if (ret)
 		goto unlock;
 
-	cm_init_av_by_path(param->alternate_path, NULL, &cm_id_priv->alt_av,
-			   cm_id_priv);
+	ret = cm_init_av_by_path(param->alternate_path, NULL,
+				 &cm_id_priv->alt_av, cm_id_priv);
+	if (ret)
+		goto unlock;
+
 	cm_id_priv->id.lap_state = IB_CM_LAP_RCVD;
 	cm_id_priv->tid = lap_msg->hdr.tid;
 	ret = atomic_inc_and_test(&cm_id_priv->work_count);
@@ -4367,7 +4370,7 @@ static void cm_add_one(struct ib_device *ib_device)
 	cm_dev->going_down = 0;
 	cm_dev->device = device_create(&cm_class, &ib_device->dev,
 				       MKDEV(0, 0), NULL,
-				       "%s", ib_device->name);
+				       "%s", dev_name(&ib_device->dev));
 	if (IS_ERR(cm_dev->device)) {
 		kfree(cm_dev);
 		return;
diff --git a/drivers/infiniband/core/cma.c b/drivers/infiniband/core/cma.c
index a36c94930c31..15d5bb7bf6bb 100644
--- a/drivers/infiniband/core/cma.c
+++ b/drivers/infiniband/core/cma.c
@@ -639,13 +639,21 @@ static void cma_bind_sgid_attr(struct rdma_id_private *id_priv,
 	id_priv->id.route.addr.dev_addr.sgid_attr = sgid_attr;
 }
 
-static int cma_acquire_dev(struct rdma_id_private *id_priv,
-			   const struct rdma_id_private *listen_id_priv)
+/**
+ * cma_acquire_dev_by_src_ip - Acquire cma device, port, gid attribute
+ * based on source ip address.
+ * @id_priv:	cm_id which should be bound to cma device
+ *
+ * cma_acquire_dev_by_src_ip() binds cm id to cma device, port and GID attribute
+ * based on source IP address. It returns 0 on success or error code otherwise.
+ * It is applicable to active and passive side cm_id.
+ */
+static int cma_acquire_dev_by_src_ip(struct rdma_id_private *id_priv)
 {
 	struct rdma_dev_addr *dev_addr = &id_priv->id.route.addr.dev_addr;
 	const struct ib_gid_attr *sgid_attr;
-	struct cma_device *cma_dev;
 	union ib_gid gid, iboe_gid, *gidp;
+	struct cma_device *cma_dev;
 	enum ib_gid_type gid_type;
 	int ret = -ENODEV;
 	u8 port;
@@ -654,41 +662,125 @@ static int cma_acquire_dev(struct rdma_id_private *id_priv,
 	    id_priv->id.ps == RDMA_PS_IPOIB)
 		return -EINVAL;
 
-	mutex_lock(&lock);
 	rdma_ip2gid((struct sockaddr *)&id_priv->id.route.addr.src_addr,
 		    &iboe_gid);
 
 	memcpy(&gid, dev_addr->src_dev_addr +
-	       rdma_addr_gid_offset(dev_addr), sizeof gid);
-
-	if (listen_id_priv) {
-		cma_dev = listen_id_priv->cma_dev;
-		port = listen_id_priv->id.port_num;
-		gidp = rdma_protocol_roce(cma_dev->device, port) ?
-		       &iboe_gid : &gid;
-		gid_type = listen_id_priv->gid_type;
-		sgid_attr = cma_validate_port(cma_dev->device, port,
-					      gid_type, gidp, id_priv);
-		if (!IS_ERR(sgid_attr)) {
-			id_priv->id.port_num = port;
-			cma_bind_sgid_attr(id_priv, sgid_attr);
-			ret = 0;
-			goto out;
+	       rdma_addr_gid_offset(dev_addr), sizeof(gid));
+
+	mutex_lock(&lock);
+	list_for_each_entry(cma_dev, &dev_list, list) {
+		for (port = rdma_start_port(cma_dev->device);
+		     port <= rdma_end_port(cma_dev->device); port++) {
+			gidp = rdma_protocol_roce(cma_dev->device, port) ?
+			       &iboe_gid : &gid;
+			gid_type = cma_dev->default_gid_type[port - 1];
+			sgid_attr = cma_validate_port(cma_dev->device, port,
+						      gid_type, gidp, id_priv);
+			if (!IS_ERR(sgid_attr)) {
+				id_priv->id.port_num = port;
+				cma_bind_sgid_attr(id_priv, sgid_attr);
+				cma_attach_to_dev(id_priv, cma_dev);
+				ret = 0;
+				goto out;
+			}
 		}
 	}
+out:
+	mutex_unlock(&lock);
+	return ret;
+}
+
+/**
+ * cma_ib_acquire_dev - Acquire cma device, port and SGID attribute
+ * @id_priv:		cm id to bind to cma device
+ * @listen_id_priv:	listener cm id to match against
+ * @req:		Pointer to req structure containaining incoming
+ *			request information
+ * cma_ib_acquire_dev() acquires cma device, port and SGID attribute when
+ * rdma device matches for listen_id and incoming request. It also verifies
+ * that a GID table entry is present for the source address.
+ * Returns 0 on success, or returns error code otherwise.
+ */
+static int cma_ib_acquire_dev(struct rdma_id_private *id_priv,
+			      const struct rdma_id_private *listen_id_priv,
+			      struct cma_req_info *req)
+{
+	struct rdma_dev_addr *dev_addr = &id_priv->id.route.addr.dev_addr;
+	const struct ib_gid_attr *sgid_attr;
+	enum ib_gid_type gid_type;
+	union ib_gid gid;
+
+	if (dev_addr->dev_type != ARPHRD_INFINIBAND &&
+	    id_priv->id.ps == RDMA_PS_IPOIB)
+		return -EINVAL;
+
+	if (rdma_protocol_roce(req->device, req->port))
+		rdma_ip2gid((struct sockaddr *)&id_priv->id.route.addr.src_addr,
+			    &gid);
+	else
+		memcpy(&gid, dev_addr->src_dev_addr +
+		       rdma_addr_gid_offset(dev_addr), sizeof(gid));
+
+	gid_type = listen_id_priv->cma_dev->default_gid_type[req->port - 1];
+	sgid_attr = cma_validate_port(req->device, req->port,
+				      gid_type, &gid, id_priv);
+	if (IS_ERR(sgid_attr))
+		return PTR_ERR(sgid_attr);
+
+	id_priv->id.port_num = req->port;
+	cma_bind_sgid_attr(id_priv, sgid_attr);
+	/* Need to acquire lock to protect against reader
+	 * of cma_dev->id_list such as cma_netdev_callback() and
+	 * cma_process_remove().
+	 */
+	mutex_lock(&lock);
+	cma_attach_to_dev(id_priv, listen_id_priv->cma_dev);
+	mutex_unlock(&lock);
+	return 0;
+}
+
+static int cma_iw_acquire_dev(struct rdma_id_private *id_priv,
+			      const struct rdma_id_private *listen_id_priv)
+{
+	struct rdma_dev_addr *dev_addr = &id_priv->id.route.addr.dev_addr;
+	const struct ib_gid_attr *sgid_attr;
+	struct cma_device *cma_dev;
+	enum ib_gid_type gid_type;
+	int ret = -ENODEV;
+	union ib_gid gid;
+	u8 port;
+
+	if (dev_addr->dev_type != ARPHRD_INFINIBAND &&
+	    id_priv->id.ps == RDMA_PS_IPOIB)
+		return -EINVAL;
+
+	memcpy(&gid, dev_addr->src_dev_addr +
+	       rdma_addr_gid_offset(dev_addr), sizeof(gid));
+
+	mutex_lock(&lock);
+
+	cma_dev = listen_id_priv->cma_dev;
+	port = listen_id_priv->id.port_num;
+	gid_type = listen_id_priv->gid_type;
+	sgid_attr = cma_validate_port(cma_dev->device, port,
+				      gid_type, &gid, id_priv);
+	if (!IS_ERR(sgid_attr)) {
+		id_priv->id.port_num = port;
+		cma_bind_sgid_attr(id_priv, sgid_attr);
+		ret = 0;
+		goto out;
+	}
 
 	list_for_each_entry(cma_dev, &dev_list, list) {
 		for (port = 1; port <= cma_dev->device->phys_port_cnt; ++port) {
-			if (listen_id_priv &&
-			    listen_id_priv->cma_dev == cma_dev &&
+			if (listen_id_priv->cma_dev == cma_dev &&
 			    listen_id_priv->id.port_num == port)
 				continue;
 
-			gidp = rdma_protocol_roce(cma_dev->device, port) ?
-			       &iboe_gid : &gid;
 			gid_type = cma_dev->default_gid_type[port - 1];
 			sgid_attr = cma_validate_port(cma_dev->device, port,
-						      gid_type, gidp, id_priv);
+						      gid_type, &gid, id_priv);
 			if (!IS_ERR(sgid_attr)) {
 				id_priv->id.port_num = port;
 				cma_bind_sgid_attr(id_priv, sgid_attr);
@@ -785,10 +877,7 @@ struct rdma_cm_id *__rdma_create_id(struct net *net,
 	if (!id_priv)
 		return ERR_PTR(-ENOMEM);
 
-	if (caller)
-		id_priv->res.kern_name = caller;
-	else
-		rdma_restrack_set_task(&id_priv->res, current);
+	rdma_restrack_set_task(&id_priv->res, caller);
 	id_priv->res.type = RDMA_RESTRACK_CM_ID;
 	id_priv->state = RDMA_CM_IDLE;
 	id_priv->id.context = context;
@@ -1462,18 +1551,35 @@ static bool cma_protocol_roce(const struct rdma_cm_id *id)
 	return rdma_protocol_roce(device, port_num);
 }
 
+static bool cma_is_req_ipv6_ll(const struct cma_req_info *req)
+{
+	const struct sockaddr *daddr =
+			(const struct sockaddr *)&req->listen_addr_storage;
+	const struct sockaddr_in6 *daddr6 = (const struct sockaddr_in6 *)daddr;
+
+	/* Returns true if the req is for IPv6 link local */
+	return (daddr->sa_family == AF_INET6 &&
+		(ipv6_addr_type(&daddr6->sin6_addr) & IPV6_ADDR_LINKLOCAL));
+}
+
 static bool cma_match_net_dev(const struct rdma_cm_id *id,
 			      const struct net_device *net_dev,
-			      u8 port_num)
+			      const struct cma_req_info *req)
 {
 	const struct rdma_addr *addr = &id->route.addr;
 
 	if (!net_dev)
 		/* This request is an AF_IB request */
-		return (!id->port_num || id->port_num == port_num) &&
+		return (!id->port_num || id->port_num == req->port) &&
 		       (addr->src_addr.ss_family == AF_IB);
 
 	/*
+	 * If the request is not for IPv6 link local, allow matching
+	 * request to any netdevice of the one or multiport rdma device.
+	 */
+	if (!cma_is_req_ipv6_ll(req))
+		return true;
+	/*
 	 * Net namespaces must match, and if the listner is listening
 	 * on a specific netdevice than netdevice must match as well.
 	 */
@@ -1500,13 +1606,14 @@ static struct rdma_id_private *cma_find_listener(
 	hlist_for_each_entry(id_priv, &bind_list->owners, node) {
 		if (cma_match_private_data(id_priv, ib_event->private_data)) {
 			if (id_priv->id.device == cm_id->device &&
-			    cma_match_net_dev(&id_priv->id, net_dev, req->port))
+			    cma_match_net_dev(&id_priv->id, net_dev, req))
 				return id_priv;
 			list_for_each_entry(id_priv_dev,
 					    &id_priv->listen_list,
 					    listen_list) {
 				if (id_priv_dev->id.device == cm_id->device &&
-				    cma_match_net_dev(&id_priv_dev->id, net_dev, req->port))
+				    cma_match_net_dev(&id_priv_dev->id,
+						      net_dev, req))
 					return id_priv_dev;
 			}
 		}
@@ -1518,18 +1625,18 @@ static struct rdma_id_private *cma_find_listener(
 static struct rdma_id_private *
 cma_ib_id_from_event(struct ib_cm_id *cm_id,
 		     const struct ib_cm_event *ib_event,
+		     struct cma_req_info *req,
 		     struct net_device **net_dev)
 {
-	struct cma_req_info req;
 	struct rdma_bind_list *bind_list;
 	struct rdma_id_private *id_priv;
 	int err;
 
-	err = cma_save_req_info(ib_event, &req);
+	err = cma_save_req_info(ib_event, req);
 	if (err)
 		return ERR_PTR(err);
 
-	*net_dev = cma_get_net_dev(ib_event, &req);
+	*net_dev = cma_get_net_dev(ib_event, req);
 	if (IS_ERR(*net_dev)) {
 		if (PTR_ERR(*net_dev) == -EAFNOSUPPORT) {
 			/* Assuming the protocol is AF_IB */
@@ -1567,17 +1674,17 @@ cma_ib_id_from_event(struct ib_cm_id *cm_id,
 		}
 
 		if (!validate_net_dev(*net_dev,
-				 (struct sockaddr *)&req.listen_addr_storage,
-				 (struct sockaddr *)&req.src_addr_storage)) {
+				 (struct sockaddr *)&req->listen_addr_storage,
+				 (struct sockaddr *)&req->src_addr_storage)) {
 			id_priv = ERR_PTR(-EHOSTUNREACH);
 			goto err;
 		}
 	}
 
 	bind_list = cma_ps_find(*net_dev ? dev_net(*net_dev) : &init_net,
-				rdma_ps_from_service_id(req.service_id),
-				cma_port_from_service_id(req.service_id));
-	id_priv = cma_find_listener(bind_list, cm_id, ib_event, &req, *net_dev);
+				rdma_ps_from_service_id(req->service_id),
+				cma_port_from_service_id(req->service_id));
+	id_priv = cma_find_listener(bind_list, cm_id, ib_event, req, *net_dev);
 err:
 	rcu_read_unlock();
 	if (IS_ERR(id_priv) && *net_dev) {
@@ -1710,8 +1817,8 @@ void rdma_destroy_id(struct rdma_cm_id *id)
 	mutex_lock(&id_priv->handler_mutex);
 	mutex_unlock(&id_priv->handler_mutex);
 
+	rdma_restrack_del(&id_priv->res);
 	if (id_priv->cma_dev) {
-		rdma_restrack_del(&id_priv->res);
 		if (rdma_cap_ib_cm(id_priv->id.device, 1)) {
 			if (id_priv->cm_id.ib)
 				ib_destroy_cm_id(id_priv->cm_id.ib);
@@ -1902,7 +2009,7 @@ cma_ib_new_conn_id(const struct rdma_cm_id *listen_id,
 		rt->path_rec[1] = *ib_event->param.req_rcvd.alternate_path;
 
 	if (net_dev) {
-		rdma_copy_addr(&rt->addr.dev_addr, net_dev, NULL);
+		rdma_copy_src_l2_addr(&rt->addr.dev_addr, net_dev);
 	} else {
 		if (!cma_protocol_roce(listen_id) &&
 		    cma_any_addr(cma_src_addr(id_priv))) {
@@ -1952,7 +2059,7 @@ cma_ib_new_udp_id(const struct rdma_cm_id *listen_id,
 		goto err;
 
 	if (net_dev) {
-		rdma_copy_addr(&id->route.addr.dev_addr, net_dev, NULL);
+		rdma_copy_src_l2_addr(&id->route.addr.dev_addr, net_dev);
 	} else {
 		if (!cma_any_addr(cma_src_addr(id_priv))) {
 			ret = cma_translate_addr(cma_src_addr(id_priv),
@@ -1999,11 +2106,12 @@ static int cma_ib_req_handler(struct ib_cm_id *cm_id,
 {
 	struct rdma_id_private *listen_id, *conn_id = NULL;
 	struct rdma_cm_event event = {};
+	struct cma_req_info req = {};
 	struct net_device *net_dev;
 	u8 offset;
 	int ret;
 
-	listen_id = cma_ib_id_from_event(cm_id, ib_event, &net_dev);
+	listen_id = cma_ib_id_from_event(cm_id, ib_event, &req, &net_dev);
 	if (IS_ERR(listen_id))
 		return PTR_ERR(listen_id);
 
@@ -2036,7 +2144,7 @@ static int cma_ib_req_handler(struct ib_cm_id *cm_id,
 	}
 
 	mutex_lock_nested(&conn_id->handler_mutex, SINGLE_DEPTH_NESTING);
-	ret = cma_acquire_dev(conn_id, listen_id);
+	ret = cma_ib_acquire_dev(conn_id, listen_id, &req);
 	if (ret)
 		goto err2;
 
@@ -2232,7 +2340,7 @@ static int iw_conn_req_handler(struct iw_cm_id *cm_id,
 		goto out;
 	}
 
-	ret = cma_acquire_dev(conn_id, listen_id);
+	ret = cma_iw_acquire_dev(conn_id, listen_id);
 	if (ret) {
 		mutex_unlock(&conn_id->handler_mutex);
 		rdma_destroy_id(new_cm_id);
@@ -2354,8 +2462,8 @@ static void cma_listen_on_dev(struct rdma_id_private *id_priv,
 
 	ret = rdma_listen(id, id_priv->backlog);
 	if (ret)
-		pr_warn("RDMA CMA: cma_listen_on_dev, error %d, listening on device %s\n",
-			ret, cma_dev->device->name);
+		dev_warn(&cma_dev->device->dev,
+			 "RDMA CMA: cma_listen_on_dev, error %d\n", ret);
 }
 
 static void cma_listen_on_all(struct rdma_id_private *id_priv)
@@ -2402,8 +2510,8 @@ static void cma_query_handler(int status, struct sa_path_rec *path_rec,
 	queue_work(cma_wq, &work->work);
 }
 
-static int cma_query_ib_route(struct rdma_id_private *id_priv, int timeout_ms,
-			      struct cma_work *work)
+static int cma_query_ib_route(struct rdma_id_private *id_priv,
+			      unsigned long timeout_ms, struct cma_work *work)
 {
 	struct rdma_dev_addr *dev_addr = &id_priv->id.route.addr.dev_addr;
 	struct sa_path_rec path_rec;
@@ -2521,7 +2629,8 @@ static void cma_init_resolve_addr_work(struct cma_work *work,
 	work->event.event = RDMA_CM_EVENT_ADDR_RESOLVED;
 }
 
-static int cma_resolve_ib_route(struct rdma_id_private *id_priv, int timeout_ms)
+static int cma_resolve_ib_route(struct rdma_id_private *id_priv,
+				unsigned long timeout_ms)
 {
 	struct rdma_route *route = &id_priv->id.route;
 	struct cma_work *work;
@@ -2643,7 +2752,7 @@ err:
 }
 EXPORT_SYMBOL(rdma_set_ib_path);
 
-static int cma_resolve_iw_route(struct rdma_id_private *id_priv, int timeout_ms)
+static int cma_resolve_iw_route(struct rdma_id_private *id_priv)
 {
 	struct cma_work *work;
 
@@ -2744,7 +2853,7 @@ err1:
 	return ret;
 }
 
-int rdma_resolve_route(struct rdma_cm_id *id, int timeout_ms)
+int rdma_resolve_route(struct rdma_cm_id *id, unsigned long timeout_ms)
 {
 	struct rdma_id_private *id_priv;
 	int ret;
@@ -2759,7 +2868,7 @@ int rdma_resolve_route(struct rdma_cm_id *id, int timeout_ms)
 	else if (rdma_protocol_roce(id->device, id->port_num))
 		ret = cma_resolve_iboe_route(id_priv);
 	else if (rdma_protocol_iwarp(id->device, id->port_num))
-		ret = cma_resolve_iw_route(id_priv, timeout_ms);
+		ret = cma_resolve_iw_route(id_priv);
 	else
 		ret = -ENOSYS;
 
@@ -2862,7 +2971,7 @@ static void addr_handler(int status, struct sockaddr *src_addr,
 
 	memcpy(cma_src_addr(id_priv), src_addr, rdma_addr_size(src_addr));
 	if (!status && !id_priv->cma_dev) {
-		status = cma_acquire_dev(id_priv, NULL);
+		status = cma_acquire_dev_by_src_ip(id_priv);
 		if (status)
 			pr_debug_ratelimited("RDMA CM: ADDR_ERROR: failed to acquire device. status %d\n",
 					     status);
@@ -2882,13 +2991,11 @@ static void addr_handler(int status, struct sockaddr *src_addr,
 	if (id_priv->id.event_handler(&id_priv->id, &event)) {
 		cma_exch(id_priv, RDMA_CM_DESTROYING);
 		mutex_unlock(&id_priv->handler_mutex);
-		cma_deref_id(id_priv);
 		rdma_destroy_id(&id_priv->id);
 		return;
 	}
 out:
 	mutex_unlock(&id_priv->handler_mutex);
-	cma_deref_id(id_priv);
 }
 
 static int cma_resolve_loopback(struct rdma_id_private *id_priv)
@@ -2966,7 +3073,7 @@ static int cma_bind_addr(struct rdma_cm_id *id, struct sockaddr *src_addr,
 }
 
 int rdma_resolve_addr(struct rdma_cm_id *id, struct sockaddr *src_addr,
-		      const struct sockaddr *dst_addr, int timeout_ms)
+		      const struct sockaddr *dst_addr, unsigned long timeout_ms)
 {
 	struct rdma_id_private *id_priv;
 	int ret;
@@ -2985,16 +3092,16 @@ int rdma_resolve_addr(struct rdma_cm_id *id, struct sockaddr *src_addr,
 		return -EINVAL;
 
 	memcpy(cma_dst_addr(id_priv), dst_addr, rdma_addr_size(dst_addr));
-	atomic_inc(&id_priv->refcount);
 	if (cma_any_addr(dst_addr)) {
 		ret = cma_resolve_loopback(id_priv);
 	} else {
 		if (dst_addr->sa_family == AF_IB) {
 			ret = cma_resolve_ib_addr(id_priv);
 		} else {
-			ret = rdma_resolve_ip(cma_src_addr(id_priv),
-					      dst_addr, &id->route.addr.dev_addr,
-					      timeout_ms, addr_handler, id_priv);
+			ret = rdma_resolve_ip(cma_src_addr(id_priv), dst_addr,
+					      &id->route.addr.dev_addr,
+					      timeout_ms, addr_handler,
+					      false, id_priv);
 		}
 	}
 	if (ret)
@@ -3003,7 +3110,6 @@ int rdma_resolve_addr(struct rdma_cm_id *id, struct sockaddr *src_addr,
 	return 0;
 err:
 	cma_comp_exch(id_priv, RDMA_CM_ADDR_QUERY, RDMA_CM_ADDR_BOUND);
-	cma_deref_id(id_priv);
 	return ret;
 }
 EXPORT_SYMBOL(rdma_resolve_addr);
@@ -3414,7 +3520,7 @@ int rdma_bind_addr(struct rdma_cm_id *id, struct sockaddr *addr)
 		if (ret)
 			goto err1;
 
-		ret = cma_acquire_dev(id_priv, NULL);
+		ret = cma_acquire_dev_by_src_ip(id_priv);
 		if (ret)
 			goto err1;
 	}
@@ -3439,10 +3545,9 @@ int rdma_bind_addr(struct rdma_cm_id *id, struct sockaddr *addr)
 
 	return 0;
 err2:
-	if (id_priv->cma_dev) {
-		rdma_restrack_del(&id_priv->res);
+	rdma_restrack_del(&id_priv->res);
+	if (id_priv->cma_dev)
 		cma_release_dev(id_priv);
-	}
 err1:
 	cma_comp_exch(id_priv, RDMA_CM_ADDR_BOUND, RDMA_CM_IDLE);
 	return ret;
@@ -3839,10 +3944,7 @@ int __rdma_accept(struct rdma_cm_id *id, struct rdma_conn_param *conn_param,
 
 	id_priv = container_of(id, struct rdma_id_private, id);
 
-	if (caller)
-		id_priv->res.kern_name = caller;
-	else
-		rdma_restrack_set_task(&id_priv->res, current);
+	rdma_restrack_set_task(&id_priv->res, caller);
 
 	if (!cma_comp(id_priv, RDMA_CM_CONNECT))
 		return -EINVAL;
@@ -4087,9 +4189,10 @@ static int cma_join_ib_multicast(struct rdma_id_private *id_priv,
 	    (!ib_sa_sendonly_fullmem_support(&sa_client,
 					     id_priv->id.device,
 					     id_priv->id.port_num))) {
-		pr_warn("RDMA CM: %s port %u Unable to multicast join\n"
-			"RDMA CM: SM doesn't support Send Only Full Member option\n",
-			id_priv->id.device->name, id_priv->id.port_num);
+		dev_warn(
+			&id_priv->id.device->dev,
+			"RDMA CM: port %u Unable to multicast join: SM doesn't support Send Only Full Member option\n",
+			id_priv->id.port_num);
 		return -EOPNOTSUPP;
 	}
 
diff --git a/drivers/infiniband/core/cma_configfs.c b/drivers/infiniband/core/cma_configfs.c
index eee38b40be99..8c2dfb3e294e 100644
--- a/drivers/infiniband/core/cma_configfs.c
+++ b/drivers/infiniband/core/cma_configfs.c
@@ -65,7 +65,7 @@ static struct cma_dev_port_group *to_dev_port_group(struct config_item *item)
 
 static bool filter_by_name(struct ib_device *ib_dev, void *cookie)
 {
-	return !strcmp(ib_dev->name, cookie);
+	return !strcmp(dev_name(&ib_dev->dev), cookie);
 }
 
 static int cma_configfs_params_get(struct config_item *item,
diff --git a/drivers/infiniband/core/core_priv.h b/drivers/infiniband/core/core_priv.h
index 77c7005c396c..bb9007a0cca7 100644
--- a/drivers/infiniband/core/core_priv.h
+++ b/drivers/infiniband/core/core_priv.h
@@ -44,7 +44,7 @@
 #include "mad_priv.h"
 
 /* Total number of ports combined across all struct ib_devices's */
-#define RDMA_MAX_PORTS 1024
+#define RDMA_MAX_PORTS 8192
 
 struct pkey_index_qp_list {
 	struct list_head    pkey_index_list;
@@ -87,6 +87,7 @@ int  ib_device_register_sysfs(struct ib_device *device,
 			      int (*port_callback)(struct ib_device *,
 						   u8, struct kobject *));
 void ib_device_unregister_sysfs(struct ib_device *device);
+int ib_device_rename(struct ib_device *ibdev, const char *name);
 
 typedef void (*roce_netdev_callback)(struct ib_device *device, u8 port,
 	      struct net_device *idev, void *cookie);
@@ -338,7 +339,14 @@ int rdma_resolve_ip_route(struct sockaddr *src_addr,
 
 int rdma_addr_find_l2_eth_by_grh(const union ib_gid *sgid,
 				 const union ib_gid *dgid,
-				 u8 *dmac, const struct net_device *ndev,
+				 u8 *dmac, const struct ib_gid_attr *sgid_attr,
 				 int *hoplimit);
+void rdma_copy_src_l2_addr(struct rdma_dev_addr *dev_addr,
+			   const struct net_device *dev);
 
+struct sa_path_rec;
+int roce_resolve_route_from_path(struct sa_path_rec *rec,
+				 const struct ib_gid_attr *attr);
+
+struct net_device *rdma_read_gid_attr_ndev_rcu(const struct ib_gid_attr *attr);
 #endif /* _CORE_PRIV_H */
diff --git a/drivers/infiniband/core/cq.c b/drivers/infiniband/core/cq.c
index af5ad6a56ae4..b1e5365ddafa 100644
--- a/drivers/infiniband/core/cq.c
+++ b/drivers/infiniband/core/cq.c
@@ -112,12 +112,12 @@ static void ib_cq_poll_work(struct work_struct *work)
 				    IB_POLL_BATCH);
 	if (completed >= IB_POLL_BUDGET_WORKQUEUE ||
 	    ib_req_notify_cq(cq, IB_POLL_FLAGS) > 0)
-		queue_work(ib_comp_wq, &cq->work);
+		queue_work(cq->comp_wq, &cq->work);
 }
 
 static void ib_cq_completion_workqueue(struct ib_cq *cq, void *private)
 {
-	queue_work(ib_comp_wq, &cq->work);
+	queue_work(cq->comp_wq, &cq->work);
 }
 
 /**
@@ -161,7 +161,7 @@ struct ib_cq *__ib_alloc_cq(struct ib_device *dev, void *private,
 		goto out_destroy_cq;
 
 	cq->res.type = RDMA_RESTRACK_CQ;
-	cq->res.kern_name = caller;
+	rdma_restrack_set_task(&cq->res, caller);
 	rdma_restrack_add(&cq->res);
 
 	switch (cq->poll_ctx) {
@@ -175,9 +175,12 @@ struct ib_cq *__ib_alloc_cq(struct ib_device *dev, void *private,
 		ib_req_notify_cq(cq, IB_CQ_NEXT_COMP);
 		break;
 	case IB_POLL_WORKQUEUE:
+	case IB_POLL_UNBOUND_WORKQUEUE:
 		cq->comp_handler = ib_cq_completion_workqueue;
 		INIT_WORK(&cq->work, ib_cq_poll_work);
 		ib_req_notify_cq(cq, IB_CQ_NEXT_COMP);
+		cq->comp_wq = (cq->poll_ctx == IB_POLL_WORKQUEUE) ?
+				ib_comp_wq : ib_comp_unbound_wq;
 		break;
 	default:
 		ret = -EINVAL;
@@ -213,6 +216,7 @@ void ib_free_cq(struct ib_cq *cq)
 		irq_poll_disable(&cq->iop);
 		break;
 	case IB_POLL_WORKQUEUE:
+	case IB_POLL_UNBOUND_WORKQUEUE:
 		cancel_work_sync(&cq->work);
 		break;
 	default:
diff --git a/drivers/infiniband/core/device.c b/drivers/infiniband/core/device.c
index db3b6271f09d..87eb4f2cdd7d 100644
--- a/drivers/infiniband/core/device.c
+++ b/drivers/infiniband/core/device.c
@@ -61,6 +61,7 @@ struct ib_client_data {
 };
 
 struct workqueue_struct *ib_comp_wq;
+struct workqueue_struct *ib_comp_unbound_wq;
 struct workqueue_struct *ib_wq;
 EXPORT_SYMBOL_GPL(ib_wq);
 
@@ -122,8 +123,9 @@ static int ib_device_check_mandatory(struct ib_device *device)
 
 	for (i = 0; i < ARRAY_SIZE(mandatory_table); ++i) {
 		if (!*(void **) ((void *) device + mandatory_table[i].offset)) {
-			pr_warn("Device %s is missing mandatory function %s\n",
-				device->name, mandatory_table[i].name);
+			dev_warn(&device->dev,
+				 "Device is missing mandatory function %s\n",
+				 mandatory_table[i].name);
 			return -EINVAL;
 		}
 	}
@@ -163,16 +165,40 @@ static struct ib_device *__ib_device_get_by_name(const char *name)
 	struct ib_device *device;
 
 	list_for_each_entry(device, &device_list, core_list)
-		if (!strncmp(name, device->name, IB_DEVICE_NAME_MAX))
+		if (!strcmp(name, dev_name(&device->dev)))
 			return device;
 
 	return NULL;
 }
 
-static int alloc_name(char *name)
+int ib_device_rename(struct ib_device *ibdev, const char *name)
+{
+	struct ib_device *device;
+	int ret = 0;
+
+	if (!strcmp(name, dev_name(&ibdev->dev)))
+		return ret;
+
+	mutex_lock(&device_mutex);
+	list_for_each_entry(device, &device_list, core_list) {
+		if (!strcmp(name, dev_name(&device->dev))) {
+			ret = -EEXIST;
+			goto out;
+		}
+	}
+
+	ret = device_rename(&ibdev->dev, name);
+	if (ret)
+		goto out;
+	strlcpy(ibdev->name, name, IB_DEVICE_NAME_MAX);
+out:
+	mutex_unlock(&device_mutex);
+	return ret;
+}
+
+static int alloc_name(struct ib_device *ibdev, const char *name)
 {
 	unsigned long *inuse;
-	char buf[IB_DEVICE_NAME_MAX];
 	struct ib_device *device;
 	int i;
 
@@ -181,24 +207,21 @@ static int alloc_name(char *name)
 		return -ENOMEM;
 
 	list_for_each_entry(device, &device_list, core_list) {
-		if (!sscanf(device->name, name, &i))
+		char buf[IB_DEVICE_NAME_MAX];
+
+		if (sscanf(dev_name(&device->dev), name, &i) != 1)
 			continue;
 		if (i < 0 || i >= PAGE_SIZE * 8)
 			continue;
 		snprintf(buf, sizeof buf, name, i);
-		if (!strncmp(buf, device->name, IB_DEVICE_NAME_MAX))
+		if (!strcmp(buf, dev_name(&device->dev)))
 			set_bit(i, inuse);
 	}
 
 	i = find_first_zero_bit(inuse, PAGE_SIZE * 8);
 	free_page((unsigned long) inuse);
-	snprintf(buf, sizeof buf, name, i);
 
-	if (__ib_device_get_by_name(buf))
-		return -ENFILE;
-
-	strlcpy(name, buf, IB_DEVICE_NAME_MAX);
-	return 0;
+	return dev_set_name(&ibdev->dev, name, i);
 }
 
 static void ib_device_release(struct device *device)
@@ -221,9 +244,7 @@ static void ib_device_release(struct device *device)
 static int ib_device_uevent(struct device *device,
 			    struct kobj_uevent_env *env)
 {
-	struct ib_device *dev = container_of(device, struct ib_device, dev);
-
-	if (add_uevent_var(env, "NAME=%s", dev->name))
+	if (add_uevent_var(env, "NAME=%s", dev_name(device)))
 		return -ENOMEM;
 
 	/*
@@ -269,7 +290,7 @@ struct ib_device *ib_alloc_device(size_t size)
 
 	INIT_LIST_HEAD(&device->event_handler_list);
 	spin_lock_init(&device->event_handler_lock);
-	spin_lock_init(&device->client_data_lock);
+	rwlock_init(&device->client_data_lock);
 	INIT_LIST_HEAD(&device->client_data_list);
 	INIT_LIST_HEAD(&device->port_list);
 
@@ -285,6 +306,7 @@ EXPORT_SYMBOL(ib_alloc_device);
  */
 void ib_dealloc_device(struct ib_device *device)
 {
+	WARN_ON(!list_empty(&device->client_data_list));
 	WARN_ON(device->reg_state != IB_DEV_UNREGISTERED &&
 		device->reg_state != IB_DEV_UNINITIALIZED);
 	rdma_restrack_clean(&device->res);
@@ -295,9 +317,8 @@ EXPORT_SYMBOL(ib_dealloc_device);
 static int add_client_context(struct ib_device *device, struct ib_client *client)
 {
 	struct ib_client_data *context;
-	unsigned long flags;
 
-	context = kmalloc(sizeof *context, GFP_KERNEL);
+	context = kmalloc(sizeof(*context), GFP_KERNEL);
 	if (!context)
 		return -ENOMEM;
 
@@ -306,9 +327,9 @@ static int add_client_context(struct ib_device *device, struct ib_client *client
 	context->going_down = false;
 
 	down_write(&lists_rwsem);
-	spin_lock_irqsave(&device->client_data_lock, flags);
+	write_lock_irq(&device->client_data_lock);
 	list_add(&context->list, &device->client_data_list);
-	spin_unlock_irqrestore(&device->client_data_lock, flags);
+	write_unlock_irq(&device->client_data_lock);
 	up_write(&lists_rwsem);
 
 	return 0;
@@ -444,22 +465,8 @@ static u32 __dev_new_index(void)
 	}
 }
 
-/**
- * ib_register_device - Register an IB device with IB core
- * @device:Device to register
- *
- * Low-level drivers use ib_register_device() to register their
- * devices with the IB core.  All registered clients will receive a
- * callback for each device that is added. @device must be allocated
- * with ib_alloc_device().
- */
-int ib_register_device(struct ib_device *device,
-		       int (*port_callback)(struct ib_device *,
-					    u8, struct kobject *))
+static void setup_dma_device(struct ib_device *device)
 {
-	int ret;
-	struct ib_client *client;
-	struct ib_udata uhw = {.outlen = 0, .inlen = 0};
 	struct device *parent = device->dev.parent;
 
 	WARN_ON_ONCE(device->dma_device);
@@ -491,56 +498,113 @@ int ib_register_device(struct ib_device *device,
 		WARN_ON_ONCE(!parent);
 		device->dma_device = parent;
 	}
+}
 
-	mutex_lock(&device_mutex);
+static void cleanup_device(struct ib_device *device)
+{
+	ib_cache_cleanup_one(device);
+	ib_cache_release_one(device);
+	kfree(device->port_pkey_list);
+	kfree(device->port_immutable);
+}
 
-	if (strchr(device->name, '%')) {
-		ret = alloc_name(device->name);
-		if (ret)
-			goto out;
-	}
+static int setup_device(struct ib_device *device)
+{
+	struct ib_udata uhw = {.outlen = 0, .inlen = 0};
+	int ret;
 
-	if (ib_device_check_mandatory(device)) {
-		ret = -EINVAL;
-		goto out;
-	}
+	ret = ib_device_check_mandatory(device);
+	if (ret)
+		return ret;
 
 	ret = read_port_immutable(device);
 	if (ret) {
-		pr_warn("Couldn't create per port immutable data %s\n",
-			device->name);
-		goto out;
+		dev_warn(&device->dev,
+			 "Couldn't create per port immutable data\n");
+		return ret;
 	}
 
-	ret = setup_port_pkey_list(device);
+	memset(&device->attrs, 0, sizeof(device->attrs));
+	ret = device->query_device(device, &device->attrs, &uhw);
 	if (ret) {
-		pr_warn("Couldn't create per port_pkey_list\n");
-		goto out;
+		dev_warn(&device->dev,
+			 "Couldn't query the device attributes\n");
+		goto port_cleanup;
 	}
 
-	ret = ib_cache_setup_one(device);
+	ret = setup_port_pkey_list(device);
 	if (ret) {
-		pr_warn("Couldn't set up InfiniBand P_Key/GID cache\n");
+		dev_warn(&device->dev, "Couldn't create per port_pkey_list\n");
 		goto port_cleanup;
 	}
 
-	ret = ib_device_register_rdmacg(device);
+	ret = ib_cache_setup_one(device);
 	if (ret) {
-		pr_warn("Couldn't register device with rdma cgroup\n");
-		goto cache_cleanup;
+		dev_warn(&device->dev,
+			 "Couldn't set up InfiniBand P_Key/GID cache\n");
+		goto pkey_cleanup;
+	}
+	return 0;
+
+pkey_cleanup:
+	kfree(device->port_pkey_list);
+port_cleanup:
+	kfree(device->port_immutable);
+	return ret;
+}
+
+/**
+ * ib_register_device - Register an IB device with IB core
+ * @device:Device to register
+ *
+ * Low-level drivers use ib_register_device() to register their
+ * devices with the IB core.  All registered clients will receive a
+ * callback for each device that is added. @device must be allocated
+ * with ib_alloc_device().
+ */
+int ib_register_device(struct ib_device *device, const char *name,
+		       int (*port_callback)(struct ib_device *, u8,
+					    struct kobject *))
+{
+	int ret;
+	struct ib_client *client;
+
+	setup_dma_device(device);
+
+	mutex_lock(&device_mutex);
+
+	if (strchr(name, '%')) {
+		ret = alloc_name(device, name);
+		if (ret)
+			goto out;
+	} else {
+		ret = dev_set_name(&device->dev, name);
+		if (ret)
+			goto out;
+	}
+	if (__ib_device_get_by_name(dev_name(&device->dev))) {
+		ret = -ENFILE;
+		goto out;
 	}
+	strlcpy(device->name, dev_name(&device->dev), IB_DEVICE_NAME_MAX);
 
-	memset(&device->attrs, 0, sizeof(device->attrs));
-	ret = device->query_device(device, &device->attrs, &uhw);
+	ret = setup_device(device);
+	if (ret)
+		goto out;
+
+	device->index = __dev_new_index();
+
+	ret = ib_device_register_rdmacg(device);
 	if (ret) {
-		pr_warn("Couldn't query the device attributes\n");
-		goto cg_cleanup;
+		dev_warn(&device->dev,
+			 "Couldn't register device with rdma cgroup\n");
+		goto dev_cleanup;
 	}
 
 	ret = ib_device_register_sysfs(device, port_callback);
 	if (ret) {
-		pr_warn("Couldn't register device %s with driver model\n",
-			device->name);
+		dev_warn(&device->dev,
+			 "Couldn't register device with driver model\n");
 		goto cg_cleanup;
 	}
 
@@ -550,7 +614,6 @@ int ib_register_device(struct ib_device *device,
 		if (!add_client_context(device, client) && client->add)
 			client->add(device);
 
-	device->index = __dev_new_index();
 	down_write(&lists_rwsem);
 	list_add_tail(&device->core_list, &device_list);
 	up_write(&lists_rwsem);
@@ -559,11 +622,8 @@ int ib_register_device(struct ib_device *device,
 
 cg_cleanup:
 	ib_device_unregister_rdmacg(device);
-cache_cleanup:
-	ib_cache_cleanup_one(device);
-	ib_cache_release_one(device);
-port_cleanup:
-	kfree(device->port_immutable);
+dev_cleanup:
+	cleanup_device(device);
 out:
 	mutex_unlock(&device_mutex);
 	return ret;
@@ -585,21 +645,20 @@ void ib_unregister_device(struct ib_device *device)
 
 	down_write(&lists_rwsem);
 	list_del(&device->core_list);
-	spin_lock_irqsave(&device->client_data_lock, flags);
-	list_for_each_entry_safe(context, tmp, &device->client_data_list, list)
+	write_lock_irq(&device->client_data_lock);
+	list_for_each_entry(context, &device->client_data_list, list)
 		context->going_down = true;
-	spin_unlock_irqrestore(&device->client_data_lock, flags);
+	write_unlock_irq(&device->client_data_lock);
 	downgrade_write(&lists_rwsem);
 
-	list_for_each_entry_safe(context, tmp, &device->client_data_list,
-				 list) {
+	list_for_each_entry(context, &device->client_data_list, list) {
 		if (context->client->remove)
 			context->client->remove(device, context->data);
 	}
 	up_read(&lists_rwsem);
 
-	ib_device_unregister_rdmacg(device);
 	ib_device_unregister_sysfs(device);
+	ib_device_unregister_rdmacg(device);
 
 	mutex_unlock(&device_mutex);
 
@@ -609,10 +668,13 @@ void ib_unregister_device(struct ib_device *device)
 	kfree(device->port_pkey_list);
 
 	down_write(&lists_rwsem);
-	spin_lock_irqsave(&device->client_data_lock, flags);
-	list_for_each_entry_safe(context, tmp, &device->client_data_list, list)
+	write_lock_irqsave(&device->client_data_lock, flags);
+	list_for_each_entry_safe(context, tmp, &device->client_data_list,
+				 list) {
+		list_del(&context->list);
 		kfree(context);
-	spin_unlock_irqrestore(&device->client_data_lock, flags);
+	}
+	write_unlock_irqrestore(&device->client_data_lock, flags);
 	up_write(&lists_rwsem);
 
 	device->reg_state = IB_DEV_UNREGISTERED;
@@ -662,9 +724,8 @@ EXPORT_SYMBOL(ib_register_client);
  */
 void ib_unregister_client(struct ib_client *client)
 {
-	struct ib_client_data *context, *tmp;
+	struct ib_client_data *context;
 	struct ib_device *device;
-	unsigned long flags;
 
 	mutex_lock(&device_mutex);
 
@@ -676,14 +737,14 @@ void ib_unregister_client(struct ib_client *client)
 		struct ib_client_data *found_context = NULL;
 
 		down_write(&lists_rwsem);
-		spin_lock_irqsave(&device->client_data_lock, flags);
-		list_for_each_entry_safe(context, tmp, &device->client_data_list, list)
+		write_lock_irq(&device->client_data_lock);
+		list_for_each_entry(context, &device->client_data_list, list)
 			if (context->client == client) {
 				context->going_down = true;
 				found_context = context;
 				break;
 			}
-		spin_unlock_irqrestore(&device->client_data_lock, flags);
+		write_unlock_irq(&device->client_data_lock);
 		up_write(&lists_rwsem);
 
 		if (client->remove)
@@ -691,17 +752,18 @@ void ib_unregister_client(struct ib_client *client)
 					       found_context->data : NULL);
 
 		if (!found_context) {
-			pr_warn("No client context found for %s/%s\n",
-				device->name, client->name);
+			dev_warn(&device->dev,
+				 "No client context found for %s\n",
+				 client->name);
 			continue;
 		}
 
 		down_write(&lists_rwsem);
-		spin_lock_irqsave(&device->client_data_lock, flags);
+		write_lock_irq(&device->client_data_lock);
 		list_del(&found_context->list);
-		kfree(found_context);
-		spin_unlock_irqrestore(&device->client_data_lock, flags);
+		write_unlock_irq(&device->client_data_lock);
 		up_write(&lists_rwsem);
+		kfree(found_context);
 	}
 
 	mutex_unlock(&device_mutex);
@@ -722,13 +784,13 @@ void *ib_get_client_data(struct ib_device *device, struct ib_client *client)
 	void *ret = NULL;
 	unsigned long flags;
 
-	spin_lock_irqsave(&device->client_data_lock, flags);
+	read_lock_irqsave(&device->client_data_lock, flags);
 	list_for_each_entry(context, &device->client_data_list, list)
 		if (context->client == client) {
 			ret = context->data;
 			break;
 		}
-	spin_unlock_irqrestore(&device->client_data_lock, flags);
+	read_unlock_irqrestore(&device->client_data_lock, flags);
 
 	return ret;
 }
@@ -749,18 +811,18 @@ void ib_set_client_data(struct ib_device *device, struct ib_client *client,
 	struct ib_client_data *context;
 	unsigned long flags;
 
-	spin_lock_irqsave(&device->client_data_lock, flags);
+	write_lock_irqsave(&device->client_data_lock, flags);
 	list_for_each_entry(context, &device->client_data_list, list)
 		if (context->client == client) {
 			context->data = data;
 			goto out;
 		}
 
-	pr_warn("No client context found for %s/%s\n",
-		device->name, client->name);
+	dev_warn(&device->dev, "No client context found for %s\n",
+		 client->name);
 
 out:
-	spin_unlock_irqrestore(&device->client_data_lock, flags);
+	write_unlock_irqrestore(&device->client_data_lock, flags);
 }
 EXPORT_SYMBOL(ib_set_client_data);
 
@@ -1166,10 +1228,19 @@ static int __init ib_core_init(void)
 		goto err;
 	}
 
+	ib_comp_unbound_wq =
+		alloc_workqueue("ib-comp-unb-wq",
+				WQ_UNBOUND | WQ_HIGHPRI | WQ_MEM_RECLAIM |
+				WQ_SYSFS, WQ_UNBOUND_MAX_ACTIVE);
+	if (!ib_comp_unbound_wq) {
+		ret = -ENOMEM;
+		goto err_comp;
+	}
+
 	ret = class_register(&ib_class);
 	if (ret) {
 		pr_warn("Couldn't create InfiniBand device class\n");
-		goto err_comp;
+		goto err_comp_unbound;
 	}
 
 	ret = rdma_nl_init();
@@ -1218,6 +1289,8 @@ err_ibnl:
 	rdma_nl_exit();
 err_sysfs:
 	class_unregister(&ib_class);
+err_comp_unbound:
+	destroy_workqueue(ib_comp_unbound_wq);
 err_comp:
 	destroy_workqueue(ib_comp_wq);
 err:
@@ -1236,6 +1309,7 @@ static void __exit ib_core_cleanup(void)
 	addr_cleanup();
 	rdma_nl_exit();
 	class_unregister(&ib_class);
+	destroy_workqueue(ib_comp_unbound_wq);
 	destroy_workqueue(ib_comp_wq);
 	/* Make sure that any pending umem accounting work is done. */
 	destroy_workqueue(ib_wq);
diff --git a/drivers/infiniband/core/fmr_pool.c b/drivers/infiniband/core/fmr_pool.c
index a077500f7f32..83ba0068e8bb 100644
--- a/drivers/infiniband/core/fmr_pool.c
+++ b/drivers/infiniband/core/fmr_pool.c
@@ -213,7 +213,7 @@ struct ib_fmr_pool *ib_create_fmr_pool(struct ib_pd             *pd,
 	device = pd->device;
 	if (!device->alloc_fmr    || !device->dealloc_fmr  ||
 	    !device->map_phys_fmr || !device->unmap_fmr) {
-		pr_info(PFX "Device %s does not support FMRs\n", device->name);
+		dev_info(&device->dev, "Device does not support FMRs\n");
 		return ERR_PTR(-ENOSYS);
 	}
 
@@ -257,7 +257,8 @@ struct ib_fmr_pool *ib_create_fmr_pool(struct ib_pd             *pd,
 	atomic_set(&pool->flush_ser, 0);
 	init_waitqueue_head(&pool->force_wait);
 
-	pool->worker = kthread_create_worker(0, "ib_fmr(%s)", device->name);
+	pool->worker =
+		kthread_create_worker(0, "ib_fmr(%s)", dev_name(&device->dev));
 	if (IS_ERR(pool->worker)) {
 		pr_warn(PFX "couldn't start cleanup kthread worker\n");
 		ret = PTR_ERR(pool->worker);
diff --git a/drivers/infiniband/core/iwcm.c b/drivers/infiniband/core/iwcm.c
index 5d676cff41f4..ba668d49c751 100644
--- a/drivers/infiniband/core/iwcm.c
+++ b/drivers/infiniband/core/iwcm.c
@@ -509,7 +509,7 @@ static int iw_cm_map(struct iw_cm_id *cm_id, bool active)
 	cm_id->m_local_addr = cm_id->local_addr;
 	cm_id->m_remote_addr = cm_id->remote_addr;
 
-	memcpy(pm_reg_msg.dev_name, cm_id->device->name,
+	memcpy(pm_reg_msg.dev_name, dev_name(&cm_id->device->dev),
 	       sizeof(pm_reg_msg.dev_name));
 	memcpy(pm_reg_msg.if_name, cm_id->device->iwcm->ifname,
 	       sizeof(pm_reg_msg.if_name));
diff --git a/drivers/infiniband/core/mad.c b/drivers/infiniband/core/mad.c
index ef459f2f2eeb..d7025cd5be28 100644
--- a/drivers/infiniband/core/mad.c
+++ b/drivers/infiniband/core/mad.c
@@ -220,33 +220,37 @@ struct ib_mad_agent *ib_register_mad_agent(struct ib_device *device,
 	int ret2, qpn;
 	u8 mgmt_class, vclass;
 
+	if ((qp_type == IB_QPT_SMI && !rdma_cap_ib_smi(device, port_num)) ||
+	    (qp_type == IB_QPT_GSI && !rdma_cap_ib_cm(device, port_num)))
+		return ERR_PTR(-EPROTONOSUPPORT);
+
 	/* Validate parameters */
 	qpn = get_spl_qp_index(qp_type);
 	if (qpn == -1) {
-		dev_notice(&device->dev,
-			   "ib_register_mad_agent: invalid QP Type %d\n",
-			   qp_type);
+		dev_dbg_ratelimited(&device->dev, "%s: invalid QP Type %d\n",
+				    __func__, qp_type);
 		goto error1;
 	}
 
 	if (rmpp_version && rmpp_version != IB_MGMT_RMPP_VERSION) {
-		dev_notice(&device->dev,
-			   "ib_register_mad_agent: invalid RMPP Version %u\n",
-			   rmpp_version);
+		dev_dbg_ratelimited(&device->dev,
+				    "%s: invalid RMPP Version %u\n",
+				    __func__, rmpp_version);
 		goto error1;
 	}
 
 	/* Validate MAD registration request if supplied */
 	if (mad_reg_req) {
 		if (mad_reg_req->mgmt_class_version >= MAX_MGMT_VERSION) {
-			dev_notice(&device->dev,
-				   "ib_register_mad_agent: invalid Class Version %u\n",
-				   mad_reg_req->mgmt_class_version);
+			dev_dbg_ratelimited(&device->dev,
+					    "%s: invalid Class Version %u\n",
+					    __func__,
+					    mad_reg_req->mgmt_class_version);
 			goto error1;
 		}
 		if (!recv_handler) {
-			dev_notice(&device->dev,
-				   "ib_register_mad_agent: no recv_handler\n");
+			dev_dbg_ratelimited(&device->dev,
+					    "%s: no recv_handler\n", __func__);
 			goto error1;
 		}
 		if (mad_reg_req->mgmt_class >= MAX_MGMT_CLASS) {
@@ -256,9 +260,9 @@ struct ib_mad_agent *ib_register_mad_agent(struct ib_device *device,
 			 */
 			if (mad_reg_req->mgmt_class !=
 			    IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE) {
-				dev_notice(&device->dev,
-					   "ib_register_mad_agent: Invalid Mgmt Class 0x%x\n",
-					   mad_reg_req->mgmt_class);
+				dev_dbg_ratelimited(&device->dev,
+					"%s: Invalid Mgmt Class 0x%x\n",
+					__func__, mad_reg_req->mgmt_class);
 				goto error1;
 			}
 		} else if (mad_reg_req->mgmt_class == 0) {
@@ -266,8 +270,9 @@ struct ib_mad_agent *ib_register_mad_agent(struct ib_device *device,
 			 * Class 0 is reserved in IBA and is used for
 			 * aliasing of IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE
 			 */
-			dev_notice(&device->dev,
-				   "ib_register_mad_agent: Invalid Mgmt Class 0\n");
+			dev_dbg_ratelimited(&device->dev,
+					    "%s: Invalid Mgmt Class 0\n",
+					    __func__);
 			goto error1;
 		} else if (is_vendor_class(mad_reg_req->mgmt_class)) {
 			/*
@@ -275,18 +280,19 @@ struct ib_mad_agent *ib_register_mad_agent(struct ib_device *device,
 			 * ensure supplied OUI is not zero
 			 */
 			if (!is_vendor_oui(mad_reg_req->oui)) {
-				dev_notice(&device->dev,
-					   "ib_register_mad_agent: No OUI specified for class 0x%x\n",
-					   mad_reg_req->mgmt_class);
+				dev_dbg_ratelimited(&device->dev,
+					"%s: No OUI specified for class 0x%x\n",
+					__func__,
+					mad_reg_req->mgmt_class);
 				goto error1;
 			}
 		}
 		/* Make sure class supplied is consistent with RMPP */
 		if (!ib_is_mad_class_rmpp(mad_reg_req->mgmt_class)) {
 			if (rmpp_version) {
-				dev_notice(&device->dev,
-					   "ib_register_mad_agent: RMPP version for non-RMPP class 0x%x\n",
-					   mad_reg_req->mgmt_class);
+				dev_dbg_ratelimited(&device->dev,
+					"%s: RMPP version for non-RMPP class 0x%x\n",
+					__func__, mad_reg_req->mgmt_class);
 				goto error1;
 			}
 		}
@@ -297,9 +303,9 @@ struct ib_mad_agent *ib_register_mad_agent(struct ib_device *device,
 					IB_MGMT_CLASS_SUBN_LID_ROUTED) &&
 			    (mad_reg_req->mgmt_class !=
 					IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE)) {
-				dev_notice(&device->dev,
-					   "ib_register_mad_agent: Invalid SM QP type: class 0x%x\n",
-					   mad_reg_req->mgmt_class);
+				dev_dbg_ratelimited(&device->dev,
+					"%s: Invalid SM QP type: class 0x%x\n",
+					__func__, mad_reg_req->mgmt_class);
 				goto error1;
 			}
 		} else {
@@ -307,9 +313,9 @@ struct ib_mad_agent *ib_register_mad_agent(struct ib_device *device,
 					IB_MGMT_CLASS_SUBN_LID_ROUTED) ||
 			    (mad_reg_req->mgmt_class ==
 					IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE)) {
-				dev_notice(&device->dev,
-					   "ib_register_mad_agent: Invalid GS QP type: class 0x%x\n",
-					   mad_reg_req->mgmt_class);
+				dev_dbg_ratelimited(&device->dev,
+					"%s: Invalid GS QP type: class 0x%x\n",
+					__func__, mad_reg_req->mgmt_class);
 				goto error1;
 			}
 		}
@@ -324,18 +330,18 @@ struct ib_mad_agent *ib_register_mad_agent(struct ib_device *device,
 	/* Validate device and port */
 	port_priv = ib_get_mad_port(device, port_num);
 	if (!port_priv) {
-		dev_notice(&device->dev,
-			   "ib_register_mad_agent: Invalid port %d\n",
-			   port_num);
+		dev_dbg_ratelimited(&device->dev, "%s: Invalid port %d\n",
+				    __func__, port_num);
 		ret = ERR_PTR(-ENODEV);
 		goto error1;
 	}
 
-	/* Verify the QP requested is supported.  For example, Ethernet devices
-	 * will not have QP0 */
+	/* Verify the QP requested is supported. For example, Ethernet devices
+	 * will not have QP0.
+	 */
 	if (!port_priv->qp_info[qpn].qp) {
-		dev_notice(&device->dev,
-			   "ib_register_mad_agent: QP %d not supported\n", qpn);
+		dev_dbg_ratelimited(&device->dev, "%s: QP %d not supported\n",
+				    __func__, qpn);
 		ret = ERR_PTR(-EPROTONOSUPPORT);
 		goto error1;
 	}
@@ -2408,7 +2414,7 @@ static void wait_for_response(struct ib_mad_send_wr_private *mad_send_wr)
 }
 
 void ib_reset_mad_timeout(struct ib_mad_send_wr_private *mad_send_wr,
-			  int timeout_ms)
+			  unsigned long timeout_ms)
 {
 	mad_send_wr->timeout = msecs_to_jiffies(timeout_ms);
 	wait_for_response(mad_send_wr);
@@ -3183,7 +3189,7 @@ static int ib_mad_port_open(struct ib_device *device,
 		cq_size *= 2;
 
 	port_priv->cq = ib_alloc_cq(port_priv->device, port_priv, cq_size, 0,
-			IB_POLL_WORKQUEUE);
+			IB_POLL_UNBOUND_WORKQUEUE);
 	if (IS_ERR(port_priv->cq)) {
 		dev_err(&device->dev, "Couldn't create ib_mad CQ\n");
 		ret = PTR_ERR(port_priv->cq);
diff --git a/drivers/infiniband/core/mad_priv.h b/drivers/infiniband/core/mad_priv.h
index d84ae1671898..216509036aa8 100644
--- a/drivers/infiniband/core/mad_priv.h
+++ b/drivers/infiniband/core/mad_priv.h
@@ -221,6 +221,6 @@ void ib_mad_complete_send_wr(struct ib_mad_send_wr_private *mad_send_wr,
 void ib_mark_mad_done(struct ib_mad_send_wr_private *mad_send_wr);
 
 void ib_reset_mad_timeout(struct ib_mad_send_wr_private *mad_send_wr,
-			  int timeout_ms);
+			  unsigned long timeout_ms);
 
 #endif	/* __IB_MAD_PRIV_H__ */
diff --git a/drivers/infiniband/core/netlink.c b/drivers/infiniband/core/netlink.c
index 3ccaae18ad75..724f5a62e82f 100644
--- a/drivers/infiniband/core/netlink.c
+++ b/drivers/infiniband/core/netlink.c
@@ -47,9 +47,9 @@ static struct {
 	const struct rdma_nl_cbs   *cb_table;
 } rdma_nl_types[RDMA_NL_NUM_CLIENTS];
 
-int rdma_nl_chk_listeners(unsigned int group)
+bool rdma_nl_chk_listeners(unsigned int group)
 {
-	return (netlink_has_listeners(nls, group)) ? 0 : -1;
+	return netlink_has_listeners(nls, group);
 }
 EXPORT_SYMBOL(rdma_nl_chk_listeners);
 
diff --git a/drivers/infiniband/core/nldev.c b/drivers/infiniband/core/nldev.c
index 0385ab438320..573399e3ccc1 100644
--- a/drivers/infiniband/core/nldev.c
+++ b/drivers/infiniband/core/nldev.c
@@ -179,7 +179,8 @@ static int fill_nldev_handle(struct sk_buff *msg, struct ib_device *device)
 {
 	if (nla_put_u32(msg, RDMA_NLDEV_ATTR_DEV_INDEX, device->index))
 		return -EMSGSIZE;
-	if (nla_put_string(msg, RDMA_NLDEV_ATTR_DEV_NAME, device->name))
+	if (nla_put_string(msg, RDMA_NLDEV_ATTR_DEV_NAME,
+			   dev_name(&device->dev)))
 		return -EMSGSIZE;
 
 	return 0;
@@ -645,6 +646,36 @@ err:
 	return err;
 }
 
+static int nldev_set_doit(struct sk_buff *skb, struct nlmsghdr *nlh,
+			  struct netlink_ext_ack *extack)
+{
+	struct nlattr *tb[RDMA_NLDEV_ATTR_MAX];
+	struct ib_device *device;
+	u32 index;
+	int err;
+
+	err = nlmsg_parse(nlh, 0, tb, RDMA_NLDEV_ATTR_MAX - 1, nldev_policy,
+			  extack);
+	if (err || !tb[RDMA_NLDEV_ATTR_DEV_INDEX])
+		return -EINVAL;
+
+	index = nla_get_u32(tb[RDMA_NLDEV_ATTR_DEV_INDEX]);
+	device = ib_device_get_by_index(index);
+	if (!device)
+		return -EINVAL;
+
+	if (tb[RDMA_NLDEV_ATTR_DEV_NAME]) {
+		char name[IB_DEVICE_NAME_MAX] = {};
+
+		nla_strlcpy(name, tb[RDMA_NLDEV_ATTR_DEV_NAME],
+			    IB_DEVICE_NAME_MAX);
+		err = ib_device_rename(device, name);
+	}
+
+	put_device(&device->dev);
+	return err;
+}
+
 static int _nldev_get_dumpit(struct ib_device *device,
 			     struct sk_buff *skb,
 			     struct netlink_callback *cb,
@@ -1077,6 +1108,10 @@ static const struct rdma_nl_cbs nldev_cb_table[RDMA_NLDEV_NUM_OPS] = {
 		.doit = nldev_get_doit,
 		.dump = nldev_get_dumpit,
 	},
+	[RDMA_NLDEV_CMD_SET] = {
+		.doit = nldev_set_doit,
+		.flags = RDMA_NL_ADMIN_PERM,
+	},
 	[RDMA_NLDEV_CMD_PORT_GET] = {
 		.doit = nldev_port_get_doit,
 		.dump = nldev_port_get_dumpit,
diff --git a/drivers/infiniband/core/rdma_core.c b/drivers/infiniband/core/rdma_core.c
index c4118bcd5103..752a55c6bdce 100644
--- a/drivers/infiniband/core/rdma_core.c
+++ b/drivers/infiniband/core/rdma_core.c
@@ -794,44 +794,6 @@ void uverbs_close_fd(struct file *f)
 	uverbs_uobject_put(uobj);
 }
 
-static void ufile_disassociate_ucontext(struct ib_ucontext *ibcontext)
-{
-	struct ib_device *ib_dev = ibcontext->device;
-	struct task_struct *owning_process  = NULL;
-	struct mm_struct   *owning_mm       = NULL;
-
-	owning_process = get_pid_task(ibcontext->tgid, PIDTYPE_PID);
-	if (!owning_process)
-		return;
-
-	owning_mm = get_task_mm(owning_process);
-	if (!owning_mm) {
-		pr_info("no mm, disassociate ucontext is pending task termination\n");
-		while (1) {
-			put_task_struct(owning_process);
-			usleep_range(1000, 2000);
-			owning_process = get_pid_task(ibcontext->tgid,
-						      PIDTYPE_PID);
-			if (!owning_process ||
-			    owning_process->state == TASK_DEAD) {
-				pr_info("disassociate ucontext done, task was terminated\n");
-				/* in case task was dead need to release the
-				 * task struct.
-				 */
-				if (owning_process)
-					put_task_struct(owning_process);
-				return;
-			}
-		}
-	}
-
-	down_write(&owning_mm->mmap_sem);
-	ib_dev->disassociate_ucontext(ibcontext);
-	up_write(&owning_mm->mmap_sem);
-	mmput(owning_mm);
-	put_task_struct(owning_process);
-}
-
 /*
  * Drop the ucontext off the ufile and completely disconnect it from the
  * ib_device
@@ -840,20 +802,28 @@ static void ufile_destroy_ucontext(struct ib_uverbs_file *ufile,
 				   enum rdma_remove_reason reason)
 {
 	struct ib_ucontext *ucontext = ufile->ucontext;
+	struct ib_device *ib_dev = ucontext->device;
 	int ret;
 
-	if (reason == RDMA_REMOVE_DRIVER_REMOVE)
-		ufile_disassociate_ucontext(ucontext);
+	/*
+	 * If we are closing the FD then the user mmap VMAs must have
+	 * already been destroyed as they hold on to the filep, otherwise
+	 * they need to be zap'd.
+	 */
+	if (reason == RDMA_REMOVE_DRIVER_REMOVE) {
+		uverbs_user_mmap_disassociate(ufile);
+		if (ib_dev->disassociate_ucontext)
+			ib_dev->disassociate_ucontext(ucontext);
+	}
 
-	put_pid(ucontext->tgid);
-	ib_rdmacg_uncharge(&ucontext->cg_obj, ucontext->device,
+	ib_rdmacg_uncharge(&ucontext->cg_obj, ib_dev,
 			   RDMACG_RESOURCE_HCA_HANDLE);
 
 	/*
 	 * FIXME: Drivers are not permitted to fail dealloc_ucontext, remove
 	 * the error return.
 	 */
-	ret = ucontext->device->dealloc_ucontext(ucontext);
+	ret = ib_dev->dealloc_ucontext(ucontext);
 	WARN_ON(ret);
 
 	ufile->ucontext = NULL;
diff --git a/drivers/infiniband/core/rdma_core.h b/drivers/infiniband/core/rdma_core.h
index f962f2a593ba..4886d2bba7c7 100644
--- a/drivers/infiniband/core/rdma_core.h
+++ b/drivers/infiniband/core/rdma_core.h
@@ -160,5 +160,6 @@ void uverbs_disassociate_api(struct uverbs_api *uapi);
 void uverbs_destroy_api(struct uverbs_api *uapi);
 void uapi_compute_bundle_size(struct uverbs_api_ioctl_method *method_elm,
 			      unsigned int num_attrs);
+void uverbs_user_mmap_disassociate(struct ib_uverbs_file *ufile);
 
 #endif /* RDMA_CORE_H */
diff --git a/drivers/infiniband/core/restrack.c b/drivers/infiniband/core/restrack.c
index 3b7fa0ccaa08..06d8657ce583 100644
--- a/drivers/infiniband/core/restrack.c
+++ b/drivers/infiniband/core/restrack.c
@@ -50,8 +50,7 @@ void rdma_restrack_clean(struct rdma_restrack_root *res)
 
 	dev = container_of(res, struct ib_device, res);
 	pr_err("restrack: %s", CUT_HERE);
-	pr_err("restrack: BUG: RESTRACK detected leak of resources on %s\n",
-	       dev->name);
+	dev_err(&dev->dev, "BUG: RESTRACK detected leak of resources\n");
 	hash_for_each(res->hash, bkt, e, node) {
 		if (rdma_is_kernel_res(e)) {
 			owner = e->kern_name;
@@ -156,6 +155,21 @@ static bool res_is_user(struct rdma_restrack_entry *res)
 	}
 }
 
+void rdma_restrack_set_task(struct rdma_restrack_entry *res,
+			    const char *caller)
+{
+	if (caller) {
+		res->kern_name = caller;
+		return;
+	}
+
+	if (res->task)
+		put_task_struct(res->task);
+	get_task_struct(current);
+	res->task = current;
+}
+EXPORT_SYMBOL(rdma_restrack_set_task);
+
 void rdma_restrack_add(struct rdma_restrack_entry *res)
 {
 	struct ib_device *dev = res_to_dev(res);
@@ -168,7 +182,7 @@ void rdma_restrack_add(struct rdma_restrack_entry *res)
 
 	if (res_is_user(res)) {
 		if (!res->task)
-			rdma_restrack_set_task(res, current);
+			rdma_restrack_set_task(res, NULL);
 		res->kern_name = NULL;
 	} else {
 		set_kern_name(res);
@@ -209,7 +223,7 @@ void rdma_restrack_del(struct rdma_restrack_entry *res)
 	struct ib_device *dev;
 
 	if (!res->valid)
-		return;
+		goto out;
 
 	dev = res_to_dev(res);
 	if (!dev)
@@ -222,8 +236,12 @@ void rdma_restrack_del(struct rdma_restrack_entry *res)
 	down_write(&dev->res.rwsem);
 	hash_del(&res->node);
 	res->valid = false;
-	if (res->task)
-		put_task_struct(res->task);
 	up_write(&dev->res.rwsem);
+
+out:
+	if (res->task) {
+		put_task_struct(res->task);
+		res->task = NULL;
+	}
 }
 EXPORT_SYMBOL(rdma_restrack_del);
diff --git a/drivers/infiniband/core/sa.h b/drivers/infiniband/core/sa.h
index b1d4bbf4ce5c..cbaaaa92fff3 100644
--- a/drivers/infiniband/core/sa.h
+++ b/drivers/infiniband/core/sa.h
@@ -49,16 +49,14 @@ static inline void ib_sa_client_put(struct ib_sa_client *client)
 }
 
 int ib_sa_mcmember_rec_query(struct ib_sa_client *client,
-			     struct ib_device *device, u8 port_num,
-			     u8 method,
+			     struct ib_device *device, u8 port_num, u8 method,
 			     struct ib_sa_mcmember_rec *rec,
 			     ib_sa_comp_mask comp_mask,
-			     int timeout_ms, gfp_t gfp_mask,
+			     unsigned long timeout_ms, gfp_t gfp_mask,
 			     void (*callback)(int status,
 					      struct ib_sa_mcmember_rec *resp,
 					      void *context),
-			     void *context,
-			     struct ib_sa_query **sa_query);
+			     void *context, struct ib_sa_query **sa_query);
 
 int mcast_init(void);
 void mcast_cleanup(void);
diff --git a/drivers/infiniband/core/sa_query.c b/drivers/infiniband/core/sa_query.c
index 7b794a14d6e8..be5ba5e15496 100644
--- a/drivers/infiniband/core/sa_query.c
+++ b/drivers/infiniband/core/sa_query.c
@@ -761,7 +761,7 @@ static void ib_nl_set_path_rec_attrs(struct sk_buff *skb,
 
 	/* Construct the family header first */
 	header = skb_put(skb, NLMSG_ALIGN(sizeof(*header)));
-	memcpy(header->device_name, query->port->agent->device->name,
+	memcpy(header->device_name, dev_name(&query->port->agent->device->dev),
 	       LS_DEVICE_NAME_MAX);
 	header->port_num = query->port->port_num;
 
@@ -835,7 +835,6 @@ static int ib_nl_send_msg(struct ib_sa_query *query, gfp_t gfp_mask)
 	struct sk_buff *skb = NULL;
 	struct nlmsghdr *nlh;
 	void *data;
-	int ret = 0;
 	struct ib_sa_mad *mad;
 	int len;
 
@@ -862,13 +861,7 @@ static int ib_nl_send_msg(struct ib_sa_query *query, gfp_t gfp_mask)
 	/* Repair the nlmsg header length */
 	nlmsg_end(skb, nlh);
 
-	ret = rdma_nl_multicast(skb, RDMA_NL_GROUP_LS, gfp_mask);
-	if (!ret)
-		ret = len;
-	else
-		ret = 0;
-
-	return ret;
+	return rdma_nl_multicast(skb, RDMA_NL_GROUP_LS, gfp_mask);
 }
 
 static int ib_nl_make_request(struct ib_sa_query *query, gfp_t gfp_mask)
@@ -891,14 +884,12 @@ static int ib_nl_make_request(struct ib_sa_query *query, gfp_t gfp_mask)
 	spin_unlock_irqrestore(&ib_nl_request_lock, flags);
 
 	ret = ib_nl_send_msg(query, gfp_mask);
-	if (ret <= 0) {
+	if (ret) {
 		ret = -EIO;
 		/* Remove the request */
 		spin_lock_irqsave(&ib_nl_request_lock, flags);
 		list_del(&query->list);
 		spin_unlock_irqrestore(&ib_nl_request_lock, flags);
-	} else {
-		ret = 0;
 	}
 
 	return ret;
@@ -1227,46 +1218,6 @@ static u8 get_src_path_mask(struct ib_device *device, u8 port_num)
 	return src_path_mask;
 }
 
-static int roce_resolve_route_from_path(struct sa_path_rec *rec,
-					const struct ib_gid_attr *attr)
-{
-	struct rdma_dev_addr dev_addr = {};
-	union {
-		struct sockaddr     _sockaddr;
-		struct sockaddr_in  _sockaddr_in;
-		struct sockaddr_in6 _sockaddr_in6;
-	} sgid_addr, dgid_addr;
-	int ret;
-
-	if (rec->roce.route_resolved)
-		return 0;
-	if (!attr || !attr->ndev)
-		return -EINVAL;
-
-	dev_addr.bound_dev_if = attr->ndev->ifindex;
-	/* TODO: Use net from the ib_gid_attr once it is added to it,
-	 * until than, limit itself to init_net.
-	 */
-	dev_addr.net = &init_net;
-
-	rdma_gid2ip(&sgid_addr._sockaddr, &rec->sgid);
-	rdma_gid2ip(&dgid_addr._sockaddr, &rec->dgid);
-
-	/* validate the route */
-	ret = rdma_resolve_ip_route(&sgid_addr._sockaddr,
-				    &dgid_addr._sockaddr, &dev_addr);
-	if (ret)
-		return ret;
-
-	if ((dev_addr.network == RDMA_NETWORK_IPV4 ||
-	     dev_addr.network == RDMA_NETWORK_IPV6) &&
-	    rec->rec_type != SA_PATH_REC_TYPE_ROCE_V2)
-		return -EINVAL;
-
-	rec->roce.route_resolved = true;
-	return 0;
-}
-
 static int init_ah_attr_grh_fields(struct ib_device *device, u8 port_num,
 				   struct sa_path_rec *rec,
 				   struct rdma_ah_attr *ah_attr,
@@ -1409,7 +1360,8 @@ static void init_mad(struct ib_sa_query *query, struct ib_mad_agent *agent)
 	spin_unlock_irqrestore(&tid_lock, flags);
 }
 
-static int send_mad(struct ib_sa_query *query, int timeout_ms, gfp_t gfp_mask)
+static int send_mad(struct ib_sa_query *query, unsigned long timeout_ms,
+		    gfp_t gfp_mask)
 {
 	bool preload = gfpflags_allow_blocking(gfp_mask);
 	unsigned long flags;
@@ -1433,7 +1385,7 @@ static int send_mad(struct ib_sa_query *query, int timeout_ms, gfp_t gfp_mask)
 
 	if ((query->flags & IB_SA_ENABLE_LOCAL_SERVICE) &&
 	    (!(query->flags & IB_SA_QUERY_OPA))) {
-		if (!rdma_nl_chk_listeners(RDMA_NL_GROUP_LS)) {
+		if (rdma_nl_chk_listeners(RDMA_NL_GROUP_LS)) {
 			if (!ib_nl_make_request(query, gfp_mask))
 				return id;
 		}
@@ -1599,7 +1551,7 @@ int ib_sa_path_rec_get(struct ib_sa_client *client,
 		       struct ib_device *device, u8 port_num,
 		       struct sa_path_rec *rec,
 		       ib_sa_comp_mask comp_mask,
-		       int timeout_ms, gfp_t gfp_mask,
+		       unsigned long timeout_ms, gfp_t gfp_mask,
 		       void (*callback)(int status,
 					struct sa_path_rec *resp,
 					void *context),
@@ -1753,7 +1705,7 @@ int ib_sa_service_rec_query(struct ib_sa_client *client,
 			    struct ib_device *device, u8 port_num, u8 method,
 			    struct ib_sa_service_rec *rec,
 			    ib_sa_comp_mask comp_mask,
-			    int timeout_ms, gfp_t gfp_mask,
+			    unsigned long timeout_ms, gfp_t gfp_mask,
 			    void (*callback)(int status,
 					     struct ib_sa_service_rec *resp,
 					     void *context),
@@ -1850,7 +1802,7 @@ int ib_sa_mcmember_rec_query(struct ib_sa_client *client,
 			     u8 method,
 			     struct ib_sa_mcmember_rec *rec,
 			     ib_sa_comp_mask comp_mask,
-			     int timeout_ms, gfp_t gfp_mask,
+			     unsigned long timeout_ms, gfp_t gfp_mask,
 			     void (*callback)(int status,
 					      struct ib_sa_mcmember_rec *resp,
 					      void *context),
@@ -1941,7 +1893,7 @@ int ib_sa_guid_info_rec_query(struct ib_sa_client *client,
 			      struct ib_device *device, u8 port_num,
 			      struct ib_sa_guidinfo_rec *rec,
 			      ib_sa_comp_mask comp_mask, u8 method,
-			      int timeout_ms, gfp_t gfp_mask,
+			      unsigned long timeout_ms, gfp_t gfp_mask,
 			      void (*callback)(int status,
 					       struct ib_sa_guidinfo_rec *resp,
 					       void *context),
@@ -2108,7 +2060,7 @@ static void ib_sa_classport_info_rec_release(struct ib_sa_query *sa_query)
 }
 
 static int ib_sa_classport_info_rec_query(struct ib_sa_port *port,
-					  int timeout_ms,
+					  unsigned long timeout_ms,
 					  void (*callback)(void *context),
 					  void *context,
 					  struct ib_sa_query **sa_query)
diff --git a/drivers/infiniband/core/security.c b/drivers/infiniband/core/security.c
index 9b0bea8303e0..1143c0448666 100644
--- a/drivers/infiniband/core/security.c
+++ b/drivers/infiniband/core/security.c
@@ -685,9 +685,8 @@ static int ib_mad_agent_security_change(struct notifier_block *nb,
 	if (event != LSM_POLICY_CHANGE)
 		return NOTIFY_DONE;
 
-	ag->smp_allowed = !security_ib_endport_manage_subnet(ag->security,
-							     ag->device->name,
-							     ag->port_num);
+	ag->smp_allowed = !security_ib_endport_manage_subnet(
+		ag->security, dev_name(&ag->device->dev), ag->port_num);
 
 	return NOTIFY_OK;
 }
@@ -708,7 +707,7 @@ int ib_mad_agent_security_setup(struct ib_mad_agent *agent,
 		return 0;
 
 	ret = security_ib_endport_manage_subnet(agent->security,
-						agent->device->name,
+						dev_name(&agent->device->dev),
 						agent->port_num);
 	if (ret)
 		return ret;
diff --git a/drivers/infiniband/core/sysfs.c b/drivers/infiniband/core/sysfs.c
index 7fd14ead7b37..6fcce2c206c6 100644
--- a/drivers/infiniband/core/sysfs.c
+++ b/drivers/infiniband/core/sysfs.c
@@ -512,7 +512,7 @@ static ssize_t show_pma_counter(struct ib_port *p, struct port_attribute *attr,
 	ret = get_perf_mad(p->ibdev, p->port_num, tab_attr->attr_id, &data,
 			40 + offset / 8, sizeof(data));
 	if (ret < 0)
-		return sprintf(buf, "N/A (no PMA)\n");
+		return ret;
 
 	switch (width) {
 	case 4:
@@ -1036,7 +1036,7 @@ static int add_port(struct ib_device *device, int port_num,
 	p->port_num   = port_num;
 
 	ret = kobject_init_and_add(&p->kobj, &port_type,
-				   device->ports_parent,
+				   device->ports_kobj,
 				   "%d", port_num);
 	if (ret) {
 		kfree(p);
@@ -1057,10 +1057,12 @@ static int add_port(struct ib_device *device, int port_num,
 		goto err_put;
 	}
 
-	p->pma_table = get_counter_table(device, port_num);
-	ret = sysfs_create_group(&p->kobj, p->pma_table);
-	if (ret)
-		goto err_put_gid_attrs;
+	if (device->process_mad) {
+		p->pma_table = get_counter_table(device, port_num);
+		ret = sysfs_create_group(&p->kobj, p->pma_table);
+		if (ret)
+			goto err_put_gid_attrs;
+	}
 
 	p->gid_group.name  = "gids";
 	p->gid_group.attrs = alloc_group_attrs(show_port_gid, attr.gid_tbl_len);
@@ -1118,9 +1120,9 @@ static int add_port(struct ib_device *device, int port_num,
 	}
 
 	/*
-	 * If port == 0, it means we have only one port and the parent
-	 * device, not this port device, should be the holder of the
-	 * hw_counters
+	 * If port == 0, it means hw_counters are per device and not per
+	 * port, so holder should be device. Therefore skip per port conunter
+	 * initialization.
 	 */
 	if (device->alloc_hw_stats && port_num)
 		setup_hw_stats(device, p, port_num);
@@ -1173,7 +1175,8 @@ err_free_gid:
 	p->gid_group.attrs = NULL;
 
 err_remove_pma:
-	sysfs_remove_group(&p->kobj, p->pma_table);
+	if (p->pma_table)
+		sysfs_remove_group(&p->kobj, p->pma_table);
 
 err_put_gid_attrs:
 	kobject_put(&p->gid_attr_group->kobj);
@@ -1183,7 +1186,7 @@ err_put:
 	return ret;
 }
 
-static ssize_t show_node_type(struct device *device,
+static ssize_t node_type_show(struct device *device,
 			      struct device_attribute *attr, char *buf)
 {
 	struct ib_device *dev = container_of(device, struct ib_device, dev);
@@ -1198,8 +1201,9 @@ static ssize_t show_node_type(struct device *device,
 	default:		  return sprintf(buf, "%d: <unknown>\n", dev->node_type);
 	}
 }
+static DEVICE_ATTR_RO(node_type);
 
-static ssize_t show_sys_image_guid(struct device *device,
+static ssize_t sys_image_guid_show(struct device *device,
 				   struct device_attribute *dev_attr, char *buf)
 {
 	struct ib_device *dev = container_of(device, struct ib_device, dev);
@@ -1210,8 +1214,9 @@ static ssize_t show_sys_image_guid(struct device *device,
 		       be16_to_cpu(((__be16 *) &dev->attrs.sys_image_guid)[2]),
 		       be16_to_cpu(((__be16 *) &dev->attrs.sys_image_guid)[3]));
 }
+static DEVICE_ATTR_RO(sys_image_guid);
 
-static ssize_t show_node_guid(struct device *device,
+static ssize_t node_guid_show(struct device *device,
 			      struct device_attribute *attr, char *buf)
 {
 	struct ib_device *dev = container_of(device, struct ib_device, dev);
@@ -1222,8 +1227,9 @@ static ssize_t show_node_guid(struct device *device,
 		       be16_to_cpu(((__be16 *) &dev->node_guid)[2]),
 		       be16_to_cpu(((__be16 *) &dev->node_guid)[3]));
 }
+static DEVICE_ATTR_RO(node_guid);
 
-static ssize_t show_node_desc(struct device *device,
+static ssize_t node_desc_show(struct device *device,
 			      struct device_attribute *attr, char *buf)
 {
 	struct ib_device *dev = container_of(device, struct ib_device, dev);
@@ -1231,9 +1237,9 @@ static ssize_t show_node_desc(struct device *device,
 	return sprintf(buf, "%.64s\n", dev->node_desc);
 }
 
-static ssize_t set_node_desc(struct device *device,
-			     struct device_attribute *attr,
-			     const char *buf, size_t count)
+static ssize_t node_desc_store(struct device *device,
+			       struct device_attribute *attr,
+			       const char *buf, size_t count)
 {
 	struct ib_device *dev = container_of(device, struct ib_device, dev);
 	struct ib_device_modify desc = {};
@@ -1249,8 +1255,9 @@ static ssize_t set_node_desc(struct device *device,
 
 	return count;
 }
+static DEVICE_ATTR_RW(node_desc);
 
-static ssize_t show_fw_ver(struct device *device, struct device_attribute *attr,
+static ssize_t fw_ver_show(struct device *device, struct device_attribute *attr,
 			   char *buf)
 {
 	struct ib_device *dev = container_of(device, struct ib_device, dev);
@@ -1259,19 +1266,19 @@ static ssize_t show_fw_ver(struct device *device, struct device_attribute *attr,
 	strlcat(buf, "\n", IB_FW_VERSION_NAME_MAX);
 	return strlen(buf);
 }
+static DEVICE_ATTR_RO(fw_ver);
+
+static struct attribute *ib_dev_attrs[] = {
+	&dev_attr_node_type.attr,
+	&dev_attr_node_guid.attr,
+	&dev_attr_sys_image_guid.attr,
+	&dev_attr_fw_ver.attr,
+	&dev_attr_node_desc.attr,
+	NULL,
+};
 
-static DEVICE_ATTR(node_type, S_IRUGO, show_node_type, NULL);
-static DEVICE_ATTR(sys_image_guid, S_IRUGO, show_sys_image_guid, NULL);
-static DEVICE_ATTR(node_guid, S_IRUGO, show_node_guid, NULL);
-static DEVICE_ATTR(node_desc, S_IRUGO | S_IWUSR, show_node_desc, set_node_desc);
-static DEVICE_ATTR(fw_ver, S_IRUGO, show_fw_ver, NULL);
-
-static struct device_attribute *ib_class_attributes[] = {
-	&dev_attr_node_type,
-	&dev_attr_sys_image_guid,
-	&dev_attr_node_guid,
-	&dev_attr_node_desc,
-	&dev_attr_fw_ver,
+static const struct attribute_group dev_attr_group = {
+	.attrs = ib_dev_attrs,
 };
 
 static void free_port_list_attributes(struct ib_device *device)
@@ -1285,7 +1292,9 @@ static void free_port_list_attributes(struct ib_device *device)
 			kfree(port->hw_stats);
 			free_hsag(&port->kobj, port->hw_stats_ag);
 		}
-		sysfs_remove_group(p, port->pma_table);
+
+		if (port->pma_table)
+			sysfs_remove_group(p, port->pma_table);
 		sysfs_remove_group(p, &port->pkey_group);
 		sysfs_remove_group(p, &port->gid_group);
 		sysfs_remove_group(&port->gid_attr_group->kobj,
@@ -1296,7 +1305,7 @@ static void free_port_list_attributes(struct ib_device *device)
 		kobject_put(p);
 	}
 
-	kobject_put(device->ports_parent);
+	kobject_put(device->ports_kobj);
 }
 
 int ib_device_register_sysfs(struct ib_device *device,
@@ -1307,23 +1316,15 @@ int ib_device_register_sysfs(struct ib_device *device,
 	int ret;
 	int i;
 
-	ret = dev_set_name(class_dev, "%s", device->name);
-	if (ret)
-		return ret;
+	device->groups[0] = &dev_attr_group;
+	class_dev->groups = device->groups;
 
 	ret = device_add(class_dev);
 	if (ret)
 		goto err;
 
-	for (i = 0; i < ARRAY_SIZE(ib_class_attributes); ++i) {
-		ret = device_create_file(class_dev, ib_class_attributes[i]);
-		if (ret)
-			goto err_unregister;
-	}
-
-	device->ports_parent = kobject_create_and_add("ports",
-						      &class_dev->kobj);
-	if (!device->ports_parent) {
+	device->ports_kobj = kobject_create_and_add("ports", &class_dev->kobj);
+	if (!device->ports_kobj) {
 		ret = -ENOMEM;
 		goto err_put;
 	}
@@ -1347,20 +1348,15 @@ int ib_device_register_sysfs(struct ib_device *device,
 
 err_put:
 	free_port_list_attributes(device);
-
-err_unregister:
 	device_del(class_dev);
-
 err:
 	return ret;
 }
 
 void ib_device_unregister_sysfs(struct ib_device *device)
 {
-	int i;
-
-	/* Hold kobject until ib_dealloc_device() */
-	kobject_get(&device->dev.kobj);
+	/* Hold device until ib_dealloc_device() */
+	get_device(&device->dev);
 
 	free_port_list_attributes(device);
 
@@ -1369,8 +1365,5 @@ void ib_device_unregister_sysfs(struct ib_device *device)
 		free_hsag(&device->dev.kobj, device->hw_stats_ag);
 	}
 
-	for (i = 0; i < ARRAY_SIZE(ib_class_attributes); ++i)
-		device_remove_file(&device->dev, ib_class_attributes[i]);
-
 	device_unregister(&device->dev);
 }
diff --git a/drivers/infiniband/core/umem.c b/drivers/infiniband/core/umem.c
index a41792dbae1f..c6144df47ea4 100644
--- a/drivers/infiniband/core/umem.c
+++ b/drivers/infiniband/core/umem.c
@@ -85,7 +85,9 @@ struct ib_umem *ib_umem_get(struct ib_ucontext *context, unsigned long addr,
 	struct page **page_list;
 	struct vm_area_struct **vma_list;
 	unsigned long lock_limit;
+	unsigned long new_pinned;
 	unsigned long cur_base;
+	struct mm_struct *mm;
 	unsigned long npages;
 	int ret;
 	int i;
@@ -107,25 +109,32 @@ struct ib_umem *ib_umem_get(struct ib_ucontext *context, unsigned long addr,
 	if (!can_do_mlock())
 		return ERR_PTR(-EPERM);
 
-	umem = kzalloc(sizeof *umem, GFP_KERNEL);
-	if (!umem)
-		return ERR_PTR(-ENOMEM);
+	if (access & IB_ACCESS_ON_DEMAND) {
+		umem = kzalloc(sizeof(struct ib_umem_odp), GFP_KERNEL);
+		if (!umem)
+			return ERR_PTR(-ENOMEM);
+		umem->is_odp = 1;
+	} else {
+		umem = kzalloc(sizeof(*umem), GFP_KERNEL);
+		if (!umem)
+			return ERR_PTR(-ENOMEM);
+	}
 
 	umem->context    = context;
 	umem->length     = size;
 	umem->address    = addr;
 	umem->page_shift = PAGE_SHIFT;
 	umem->writable   = ib_access_writable(access);
+	umem->owning_mm = mm = current->mm;
+	mmgrab(mm);
 
 	if (access & IB_ACCESS_ON_DEMAND) {
-		ret = ib_umem_odp_get(context, umem, access);
+		ret = ib_umem_odp_get(to_ib_umem_odp(umem), access);
 		if (ret)
 			goto umem_kfree;
 		return umem;
 	}
 
-	umem->odp_data = NULL;
-
 	/* We assume the memory is from hugetlb until proved otherwise */
 	umem->hugetlb   = 1;
 
@@ -144,25 +153,25 @@ struct ib_umem *ib_umem_get(struct ib_ucontext *context, unsigned long addr,
 		umem->hugetlb = 0;
 
 	npages = ib_umem_num_pages(umem);
+	if (npages == 0 || npages > UINT_MAX) {
+		ret = -EINVAL;
+		goto out;
+	}
 
 	lock_limit = rlimit(RLIMIT_MEMLOCK) >> PAGE_SHIFT;
 
-	down_write(&current->mm->mmap_sem);
-	current->mm->pinned_vm += npages;
-	if ((current->mm->pinned_vm > lock_limit) && !capable(CAP_IPC_LOCK)) {
-		up_write(&current->mm->mmap_sem);
+	down_write(&mm->mmap_sem);
+	if (check_add_overflow(mm->pinned_vm, npages, &new_pinned) ||
+	    (new_pinned > lock_limit && !capable(CAP_IPC_LOCK))) {
+		up_write(&mm->mmap_sem);
 		ret = -ENOMEM;
-		goto vma;
+		goto out;
 	}
-	up_write(&current->mm->mmap_sem);
+	mm->pinned_vm = new_pinned;
+	up_write(&mm->mmap_sem);
 
 	cur_base = addr & PAGE_MASK;
 
-	if (npages == 0 || npages > UINT_MAX) {
-		ret = -EINVAL;
-		goto vma;
-	}
-
 	ret = sg_alloc_table(&umem->sg_head, npages, GFP_KERNEL);
 	if (ret)
 		goto vma;
@@ -172,14 +181,14 @@ struct ib_umem *ib_umem_get(struct ib_ucontext *context, unsigned long addr,
 
 	sg_list_start = umem->sg_head.sgl;
 
-	down_read(&current->mm->mmap_sem);
 	while (npages) {
+		down_read(&mm->mmap_sem);
 		ret = get_user_pages_longterm(cur_base,
 				     min_t(unsigned long, npages,
 					   PAGE_SIZE / sizeof (struct page *)),
 				     gup_flags, page_list, vma_list);
 		if (ret < 0) {
-			up_read(&current->mm->mmap_sem);
+			up_read(&mm->mmap_sem);
 			goto umem_release;
 		}
 
@@ -187,17 +196,20 @@ struct ib_umem *ib_umem_get(struct ib_ucontext *context, unsigned long addr,
 		cur_base += ret * PAGE_SIZE;
 		npages   -= ret;
 
+		/* Continue to hold the mmap_sem as vma_list access
+		 * needs to be protected.
+		 */
 		for_each_sg(sg_list_start, sg, ret, i) {
 			if (vma_list && !is_vm_hugetlb_page(vma_list[i]))
 				umem->hugetlb = 0;
 
 			sg_set_page(sg, page_list[i], PAGE_SIZE, 0);
 		}
+		up_read(&mm->mmap_sem);
 
 		/* preparing for next loop */
 		sg_list_start = sg;
 	}
-	up_read(&current->mm->mmap_sem);
 
 	umem->nmap = ib_dma_map_sg_attrs(context->device,
 				  umem->sg_head.sgl,
@@ -216,29 +228,40 @@ struct ib_umem *ib_umem_get(struct ib_ucontext *context, unsigned long addr,
 umem_release:
 	__ib_umem_release(context->device, umem, 0);
 vma:
-	down_write(&current->mm->mmap_sem);
-	current->mm->pinned_vm -= ib_umem_num_pages(umem);
-	up_write(&current->mm->mmap_sem);
+	down_write(&mm->mmap_sem);
+	mm->pinned_vm -= ib_umem_num_pages(umem);
+	up_write(&mm->mmap_sem);
 out:
 	if (vma_list)
 		free_page((unsigned long) vma_list);
 	free_page((unsigned long) page_list);
 umem_kfree:
-	if (ret)
+	if (ret) {
+		mmdrop(umem->owning_mm);
 		kfree(umem);
+	}
 	return ret ? ERR_PTR(ret) : umem;
 }
 EXPORT_SYMBOL(ib_umem_get);
 
-static void ib_umem_account(struct work_struct *work)
+static void __ib_umem_release_tail(struct ib_umem *umem)
+{
+	mmdrop(umem->owning_mm);
+	if (umem->is_odp)
+		kfree(to_ib_umem_odp(umem));
+	else
+		kfree(umem);
+}
+
+static void ib_umem_release_defer(struct work_struct *work)
 {
 	struct ib_umem *umem = container_of(work, struct ib_umem, work);
 
-	down_write(&umem->mm->mmap_sem);
-	umem->mm->pinned_vm -= umem->diff;
-	up_write(&umem->mm->mmap_sem);
-	mmput(umem->mm);
-	kfree(umem);
+	down_write(&umem->owning_mm->mmap_sem);
+	umem->owning_mm->pinned_vm -= ib_umem_num_pages(umem);
+	up_write(&umem->owning_mm->mmap_sem);
+
+	__ib_umem_release_tail(umem);
 }
 
 /**
@@ -248,52 +271,36 @@ static void ib_umem_account(struct work_struct *work)
 void ib_umem_release(struct ib_umem *umem)
 {
 	struct ib_ucontext *context = umem->context;
-	struct mm_struct *mm;
-	struct task_struct *task;
-	unsigned long diff;
 
-	if (umem->odp_data) {
-		ib_umem_odp_release(umem);
+	if (umem->is_odp) {
+		ib_umem_odp_release(to_ib_umem_odp(umem));
+		__ib_umem_release_tail(umem);
 		return;
 	}
 
 	__ib_umem_release(umem->context->device, umem, 1);
 
-	task = get_pid_task(umem->context->tgid, PIDTYPE_PID);
-	if (!task)
-		goto out;
-	mm = get_task_mm(task);
-	put_task_struct(task);
-	if (!mm)
-		goto out;
-
-	diff = ib_umem_num_pages(umem);
-
 	/*
 	 * We may be called with the mm's mmap_sem already held.  This
 	 * can happen when a userspace munmap() is the call that drops
 	 * the last reference to our file and calls our release
 	 * method.  If there are memory regions to destroy, we'll end
 	 * up here and not be able to take the mmap_sem.  In that case
-	 * we defer the vm_locked accounting to the system workqueue.
+	 * we defer the vm_locked accounting a workqueue.
 	 */
 	if (context->closing) {
-		if (!down_write_trylock(&mm->mmap_sem)) {
-			INIT_WORK(&umem->work, ib_umem_account);
-			umem->mm   = mm;
-			umem->diff = diff;
-
+		if (!down_write_trylock(&umem->owning_mm->mmap_sem)) {
+			INIT_WORK(&umem->work, ib_umem_release_defer);
 			queue_work(ib_wq, &umem->work);
 			return;
 		}
-	} else
-		down_write(&mm->mmap_sem);
+	} else {
+		down_write(&umem->owning_mm->mmap_sem);
+	}
+	umem->owning_mm->pinned_vm -= ib_umem_num_pages(umem);
+	up_write(&umem->owning_mm->mmap_sem);
 
-	mm->pinned_vm -= diff;
-	up_write(&mm->mmap_sem);
-	mmput(mm);
-out:
-	kfree(umem);
+	__ib_umem_release_tail(umem);
 }
 EXPORT_SYMBOL(ib_umem_release);
 
@@ -303,7 +310,7 @@ int ib_umem_page_count(struct ib_umem *umem)
 	int n;
 	struct scatterlist *sg;
 
-	if (umem->odp_data)
+	if (umem->is_odp)
 		return ib_umem_num_pages(umem);
 
 	n = 0;
diff --git a/drivers/infiniband/core/umem_odp.c b/drivers/infiniband/core/umem_odp.c
index 6ec748eccff7..2b4c5e7dd5a1 100644
--- a/drivers/infiniband/core/umem_odp.c
+++ b/drivers/infiniband/core/umem_odp.c
@@ -58,7 +58,7 @@ static u64 node_start(struct umem_odp_node *n)
 	struct ib_umem_odp *umem_odp =
 			container_of(n, struct ib_umem_odp, interval_tree);
 
-	return ib_umem_start(umem_odp->umem);
+	return ib_umem_start(&umem_odp->umem);
 }
 
 /* Note that the representation of the intervals in the interval tree
@@ -71,140 +71,86 @@ static u64 node_last(struct umem_odp_node *n)
 	struct ib_umem_odp *umem_odp =
 			container_of(n, struct ib_umem_odp, interval_tree);
 
-	return ib_umem_end(umem_odp->umem) - 1;
+	return ib_umem_end(&umem_odp->umem) - 1;
 }
 
 INTERVAL_TREE_DEFINE(struct umem_odp_node, rb, u64, __subtree_last,
 		     node_start, node_last, static, rbt_ib_umem)
 
-static void ib_umem_notifier_start_account(struct ib_umem *item)
+static void ib_umem_notifier_start_account(struct ib_umem_odp *umem_odp)
 {
-	mutex_lock(&item->odp_data->umem_mutex);
-
-	/* Only update private counters for this umem if it has them.
-	 * Otherwise skip it. All page faults will be delayed for this umem. */
-	if (item->odp_data->mn_counters_active) {
-		int notifiers_count = item->odp_data->notifiers_count++;
-
-		if (notifiers_count == 0)
-			/* Initialize the completion object for waiting on
-			 * notifiers. Since notifier_count is zero, no one
-			 * should be waiting right now. */
-			reinit_completion(&item->odp_data->notifier_completion);
-	}
-	mutex_unlock(&item->odp_data->umem_mutex);
-}
-
-static void ib_umem_notifier_end_account(struct ib_umem *item)
-{
-	mutex_lock(&item->odp_data->umem_mutex);
-
-	/* Only update private counters for this umem if it has them.
-	 * Otherwise skip it. All page faults will be delayed for this umem. */
-	if (item->odp_data->mn_counters_active) {
+	mutex_lock(&umem_odp->umem_mutex);
+	if (umem_odp->notifiers_count++ == 0)
 		/*
-		 * This sequence increase will notify the QP page fault that
-		 * the page that is going to be mapped in the spte could have
-		 * been freed.
+		 * Initialize the completion object for waiting on
+		 * notifiers. Since notifier_count is zero, no one should be
+		 * waiting right now.
 		 */
-		++item->odp_data->notifiers_seq;
-		if (--item->odp_data->notifiers_count == 0)
-			complete_all(&item->odp_data->notifier_completion);
-	}
-	mutex_unlock(&item->odp_data->umem_mutex);
+		reinit_completion(&umem_odp->notifier_completion);
+	mutex_unlock(&umem_odp->umem_mutex);
 }
 
-/* Account for a new mmu notifier in an ib_ucontext. */
-static void ib_ucontext_notifier_start_account(struct ib_ucontext *context)
+static void ib_umem_notifier_end_account(struct ib_umem_odp *umem_odp)
 {
-	atomic_inc(&context->notifier_count);
+	mutex_lock(&umem_odp->umem_mutex);
+	/*
+	 * This sequence increase will notify the QP page fault that the page
+	 * that is going to be mapped in the spte could have been freed.
+	 */
+	++umem_odp->notifiers_seq;
+	if (--umem_odp->notifiers_count == 0)
+		complete_all(&umem_odp->notifier_completion);
+	mutex_unlock(&umem_odp->umem_mutex);
 }
 
-/* Account for a terminating mmu notifier in an ib_ucontext.
- *
- * Must be called with the ib_ucontext->umem_rwsem semaphore unlocked, since
- * the function takes the semaphore itself. */
-static void ib_ucontext_notifier_end_account(struct ib_ucontext *context)
+static int ib_umem_notifier_release_trampoline(struct ib_umem_odp *umem_odp,
+					       u64 start, u64 end, void *cookie)
 {
-	int zero_notifiers = atomic_dec_and_test(&context->notifier_count);
-
-	if (zero_notifiers &&
-	    !list_empty(&context->no_private_counters)) {
-		/* No currently running mmu notifiers. Now is the chance to
-		 * add private accounting to all previously added umems. */
-		struct ib_umem_odp *odp_data, *next;
-
-		/* Prevent concurrent mmu notifiers from working on the
-		 * no_private_counters list. */
-		down_write(&context->umem_rwsem);
-
-		/* Read the notifier_count again, with the umem_rwsem
-		 * semaphore taken for write. */
-		if (!atomic_read(&context->notifier_count)) {
-			list_for_each_entry_safe(odp_data, next,
-						 &context->no_private_counters,
-						 no_private_counters) {
-				mutex_lock(&odp_data->umem_mutex);
-				odp_data->mn_counters_active = true;
-				list_del(&odp_data->no_private_counters);
-				complete_all(&odp_data->notifier_completion);
-				mutex_unlock(&odp_data->umem_mutex);
-			}
-		}
-
-		up_write(&context->umem_rwsem);
-	}
-}
+	struct ib_umem *umem = &umem_odp->umem;
 
-static int ib_umem_notifier_release_trampoline(struct ib_umem *item, u64 start,
-					       u64 end, void *cookie) {
 	/*
 	 * Increase the number of notifiers running, to
 	 * prevent any further fault handling on this MR.
 	 */
-	ib_umem_notifier_start_account(item);
-	item->odp_data->dying = 1;
+	ib_umem_notifier_start_account(umem_odp);
+	umem_odp->dying = 1;
 	/* Make sure that the fact the umem is dying is out before we release
 	 * all pending page faults. */
 	smp_wmb();
-	complete_all(&item->odp_data->notifier_completion);
-	item->context->invalidate_range(item, ib_umem_start(item),
-					ib_umem_end(item));
+	complete_all(&umem_odp->notifier_completion);
+	umem->context->invalidate_range(umem_odp, ib_umem_start(umem),
+					ib_umem_end(umem));
 	return 0;
 }
 
 static void ib_umem_notifier_release(struct mmu_notifier *mn,
 				     struct mm_struct *mm)
 {
-	struct ib_ucontext *context = container_of(mn, struct ib_ucontext, mn);
-
-	if (!context->invalidate_range)
-		return;
-
-	ib_ucontext_notifier_start_account(context);
-	down_read(&context->umem_rwsem);
-	rbt_ib_umem_for_each_in_range(&context->umem_tree, 0,
-				      ULLONG_MAX,
-				      ib_umem_notifier_release_trampoline,
-				      true,
-				      NULL);
-	up_read(&context->umem_rwsem);
+	struct ib_ucontext_per_mm *per_mm =
+		container_of(mn, struct ib_ucontext_per_mm, mn);
+
+	down_read(&per_mm->umem_rwsem);
+	if (per_mm->active)
+		rbt_ib_umem_for_each_in_range(
+			&per_mm->umem_tree, 0, ULLONG_MAX,
+			ib_umem_notifier_release_trampoline, true, NULL);
+	up_read(&per_mm->umem_rwsem);
 }
 
-static int invalidate_page_trampoline(struct ib_umem *item, u64 start,
+static int invalidate_page_trampoline(struct ib_umem_odp *item, u64 start,
 				      u64 end, void *cookie)
 {
 	ib_umem_notifier_start_account(item);
-	item->context->invalidate_range(item, start, start + PAGE_SIZE);
+	item->umem.context->invalidate_range(item, start, start + PAGE_SIZE);
 	ib_umem_notifier_end_account(item);
 	return 0;
 }
 
-static int invalidate_range_start_trampoline(struct ib_umem *item, u64 start,
-					     u64 end, void *cookie)
+static int invalidate_range_start_trampoline(struct ib_umem_odp *item,
+					     u64 start, u64 end, void *cookie)
 {
 	ib_umem_notifier_start_account(item);
-	item->context->invalidate_range(item, start, end);
+	item->umem.context->invalidate_range(item, start, end);
 	return 0;
 }
 
@@ -214,28 +160,30 @@ static int ib_umem_notifier_invalidate_range_start(struct mmu_notifier *mn,
 						    unsigned long end,
 						    bool blockable)
 {
-	struct ib_ucontext *context = container_of(mn, struct ib_ucontext, mn);
-	int ret;
-
-	if (!context->invalidate_range)
-		return 0;
+	struct ib_ucontext_per_mm *per_mm =
+		container_of(mn, struct ib_ucontext_per_mm, mn);
 
 	if (blockable)
-		down_read(&context->umem_rwsem);
-	else if (!down_read_trylock(&context->umem_rwsem))
+		down_read(&per_mm->umem_rwsem);
+	else if (!down_read_trylock(&per_mm->umem_rwsem))
 		return -EAGAIN;
 
-	ib_ucontext_notifier_start_account(context);
-	ret = rbt_ib_umem_for_each_in_range(&context->umem_tree, start,
-				      end,
-				      invalidate_range_start_trampoline,
-				      blockable, NULL);
-	up_read(&context->umem_rwsem);
+	if (!per_mm->active) {
+		up_read(&per_mm->umem_rwsem);
+		/*
+		 * At this point active is permanently set and visible to this
+		 * CPU without a lock, that fact is relied on to skip the unlock
+		 * in range_end.
+		 */
+		return 0;
+	}
 
-	return ret;
+	return rbt_ib_umem_for_each_in_range(&per_mm->umem_tree, start, end,
+					     invalidate_range_start_trampoline,
+					     blockable, NULL);
 }
 
-static int invalidate_range_end_trampoline(struct ib_umem *item, u64 start,
+static int invalidate_range_end_trampoline(struct ib_umem_odp *item, u64 start,
 					   u64 end, void *cookie)
 {
 	ib_umem_notifier_end_account(item);
@@ -247,22 +195,16 @@ static void ib_umem_notifier_invalidate_range_end(struct mmu_notifier *mn,
 						  unsigned long start,
 						  unsigned long end)
 {
-	struct ib_ucontext *context = container_of(mn, struct ib_ucontext, mn);
+	struct ib_ucontext_per_mm *per_mm =
+		container_of(mn, struct ib_ucontext_per_mm, mn);
 
-	if (!context->invalidate_range)
+	if (unlikely(!per_mm->active))
 		return;
 
-	/*
-	 * TODO: we currently bail out if there is any sleepable work to be done
-	 * in ib_umem_notifier_invalidate_range_start so we shouldn't really block
-	 * here. But this is ugly and fragile.
-	 */
-	down_read(&context->umem_rwsem);
-	rbt_ib_umem_for_each_in_range(&context->umem_tree, start,
+	rbt_ib_umem_for_each_in_range(&per_mm->umem_tree, start,
 				      end,
 				      invalidate_range_end_trampoline, true, NULL);
-	up_read(&context->umem_rwsem);
-	ib_ucontext_notifier_end_account(context);
+	up_read(&per_mm->umem_rwsem);
 }
 
 static const struct mmu_notifier_ops ib_umem_notifiers = {
@@ -271,31 +213,158 @@ static const struct mmu_notifier_ops ib_umem_notifiers = {
 	.invalidate_range_end       = ib_umem_notifier_invalidate_range_end,
 };
 
-struct ib_umem *ib_alloc_odp_umem(struct ib_ucontext *context,
-				  unsigned long addr,
-				  size_t size)
+static void add_umem_to_per_mm(struct ib_umem_odp *umem_odp)
 {
-	struct ib_umem *umem;
+	struct ib_ucontext_per_mm *per_mm = umem_odp->per_mm;
+	struct ib_umem *umem = &umem_odp->umem;
+
+	down_write(&per_mm->umem_rwsem);
+	if (likely(ib_umem_start(umem) != ib_umem_end(umem)))
+		rbt_ib_umem_insert(&umem_odp->interval_tree,
+				   &per_mm->umem_tree);
+	up_write(&per_mm->umem_rwsem);
+}
+
+static void remove_umem_from_per_mm(struct ib_umem_odp *umem_odp)
+{
+	struct ib_ucontext_per_mm *per_mm = umem_odp->per_mm;
+	struct ib_umem *umem = &umem_odp->umem;
+
+	down_write(&per_mm->umem_rwsem);
+	if (likely(ib_umem_start(umem) != ib_umem_end(umem)))
+		rbt_ib_umem_remove(&umem_odp->interval_tree,
+				   &per_mm->umem_tree);
+	complete_all(&umem_odp->notifier_completion);
+
+	up_write(&per_mm->umem_rwsem);
+}
+
+static struct ib_ucontext_per_mm *alloc_per_mm(struct ib_ucontext *ctx,
+					       struct mm_struct *mm)
+{
+	struct ib_ucontext_per_mm *per_mm;
+	int ret;
+
+	per_mm = kzalloc(sizeof(*per_mm), GFP_KERNEL);
+	if (!per_mm)
+		return ERR_PTR(-ENOMEM);
+
+	per_mm->context = ctx;
+	per_mm->mm = mm;
+	per_mm->umem_tree = RB_ROOT_CACHED;
+	init_rwsem(&per_mm->umem_rwsem);
+	per_mm->active = ctx->invalidate_range;
+
+	rcu_read_lock();
+	per_mm->tgid = get_task_pid(current->group_leader, PIDTYPE_PID);
+	rcu_read_unlock();
+
+	WARN_ON(mm != current->mm);
+
+	per_mm->mn.ops = &ib_umem_notifiers;
+	ret = mmu_notifier_register(&per_mm->mn, per_mm->mm);
+	if (ret) {
+		dev_err(&ctx->device->dev,
+			"Failed to register mmu_notifier %d\n", ret);
+		goto out_pid;
+	}
+
+	list_add(&per_mm->ucontext_list, &ctx->per_mm_list);
+	return per_mm;
+
+out_pid:
+	put_pid(per_mm->tgid);
+	kfree(per_mm);
+	return ERR_PTR(ret);
+}
+
+static int get_per_mm(struct ib_umem_odp *umem_odp)
+{
+	struct ib_ucontext *ctx = umem_odp->umem.context;
+	struct ib_ucontext_per_mm *per_mm;
+
+	/*
+	 * Generally speaking we expect only one or two per_mm in this list,
+	 * so no reason to optimize this search today.
+	 */
+	mutex_lock(&ctx->per_mm_list_lock);
+	list_for_each_entry(per_mm, &ctx->per_mm_list, ucontext_list) {
+		if (per_mm->mm == umem_odp->umem.owning_mm)
+			goto found;
+	}
+
+	per_mm = alloc_per_mm(ctx, umem_odp->umem.owning_mm);
+	if (IS_ERR(per_mm)) {
+		mutex_unlock(&ctx->per_mm_list_lock);
+		return PTR_ERR(per_mm);
+	}
+
+found:
+	umem_odp->per_mm = per_mm;
+	per_mm->odp_mrs_count++;
+	mutex_unlock(&ctx->per_mm_list_lock);
+
+	return 0;
+}
+
+static void free_per_mm(struct rcu_head *rcu)
+{
+	kfree(container_of(rcu, struct ib_ucontext_per_mm, rcu));
+}
+
+void put_per_mm(struct ib_umem_odp *umem_odp)
+{
+	struct ib_ucontext_per_mm *per_mm = umem_odp->per_mm;
+	struct ib_ucontext *ctx = umem_odp->umem.context;
+	bool need_free;
+
+	mutex_lock(&ctx->per_mm_list_lock);
+	umem_odp->per_mm = NULL;
+	per_mm->odp_mrs_count--;
+	need_free = per_mm->odp_mrs_count == 0;
+	if (need_free)
+		list_del(&per_mm->ucontext_list);
+	mutex_unlock(&ctx->per_mm_list_lock);
+
+	if (!need_free)
+		return;
+
+	/*
+	 * NOTE! mmu_notifier_unregister() can happen between a start/end
+	 * callback, resulting in an start/end, and thus an unbalanced
+	 * lock. This doesn't really matter to us since we are about to kfree
+	 * the memory that holds the lock, however LOCKDEP doesn't like this.
+	 */
+	down_write(&per_mm->umem_rwsem);
+	per_mm->active = false;
+	up_write(&per_mm->umem_rwsem);
+
+	WARN_ON(!RB_EMPTY_ROOT(&per_mm->umem_tree.rb_root));
+	mmu_notifier_unregister_no_release(&per_mm->mn, per_mm->mm);
+	put_pid(per_mm->tgid);
+	mmu_notifier_call_srcu(&per_mm->rcu, free_per_mm);
+}
+
+struct ib_umem_odp *ib_alloc_odp_umem(struct ib_ucontext_per_mm *per_mm,
+				      unsigned long addr, size_t size)
+{
+	struct ib_ucontext *ctx = per_mm->context;
 	struct ib_umem_odp *odp_data;
+	struct ib_umem *umem;
 	int pages = size >> PAGE_SHIFT;
 	int ret;
 
-	umem = kzalloc(sizeof(*umem), GFP_KERNEL);
-	if (!umem)
+	odp_data = kzalloc(sizeof(*odp_data), GFP_KERNEL);
+	if (!odp_data)
 		return ERR_PTR(-ENOMEM);
-
-	umem->context    = context;
+	umem = &odp_data->umem;
+	umem->context    = ctx;
 	umem->length     = size;
 	umem->address    = addr;
 	umem->page_shift = PAGE_SHIFT;
 	umem->writable   = 1;
-
-	odp_data = kzalloc(sizeof(*odp_data), GFP_KERNEL);
-	if (!odp_data) {
-		ret = -ENOMEM;
-		goto out_umem;
-	}
-	odp_data->umem = umem;
+	umem->is_odp = 1;
+	odp_data->per_mm = per_mm;
 
 	mutex_init(&odp_data->umem_mutex);
 	init_completion(&odp_data->notifier_completion);
@@ -314,39 +383,34 @@ struct ib_umem *ib_alloc_odp_umem(struct ib_ucontext *context,
 		goto out_page_list;
 	}
 
-	down_write(&context->umem_rwsem);
-	context->odp_mrs_count++;
-	rbt_ib_umem_insert(&odp_data->interval_tree, &context->umem_tree);
-	if (likely(!atomic_read(&context->notifier_count)))
-		odp_data->mn_counters_active = true;
-	else
-		list_add(&odp_data->no_private_counters,
-			 &context->no_private_counters);
-	up_write(&context->umem_rwsem);
-
-	umem->odp_data = odp_data;
+	/*
+	 * Caller must ensure that the umem_odp that the per_mm came from
+	 * cannot be freed during the call to ib_alloc_odp_umem.
+	 */
+	mutex_lock(&ctx->per_mm_list_lock);
+	per_mm->odp_mrs_count++;
+	mutex_unlock(&ctx->per_mm_list_lock);
+	add_umem_to_per_mm(odp_data);
 
-	return umem;
+	return odp_data;
 
 out_page_list:
 	vfree(odp_data->page_list);
 out_odp_data:
 	kfree(odp_data);
-out_umem:
-	kfree(umem);
 	return ERR_PTR(ret);
 }
 EXPORT_SYMBOL(ib_alloc_odp_umem);
 
-int ib_umem_odp_get(struct ib_ucontext *context, struct ib_umem *umem,
-		    int access)
+int ib_umem_odp_get(struct ib_umem_odp *umem_odp, int access)
 {
+	struct ib_umem *umem = &umem_odp->umem;
+	/*
+	 * NOTE: This must called in a process context where umem->owning_mm
+	 * == current->mm
+	 */
+	struct mm_struct *mm = umem->owning_mm;
 	int ret_val;
-	struct pid *our_pid;
-	struct mm_struct *mm = get_task_mm(current);
-
-	if (!mm)
-		return -EINVAL;
 
 	if (access & IB_ACCESS_HUGETLB) {
 		struct vm_area_struct *vma;
@@ -366,111 +430,43 @@ int ib_umem_odp_get(struct ib_ucontext *context, struct ib_umem *umem,
 		umem->hugetlb = 0;
 	}
 
-	/* Prevent creating ODP MRs in child processes */
-	rcu_read_lock();
-	our_pid = get_task_pid(current->group_leader, PIDTYPE_PID);
-	rcu_read_unlock();
-	put_pid(our_pid);
-	if (context->tgid != our_pid) {
-		ret_val = -EINVAL;
-		goto out_mm;
-	}
-
-	umem->odp_data = kzalloc(sizeof(*umem->odp_data), GFP_KERNEL);
-	if (!umem->odp_data) {
-		ret_val = -ENOMEM;
-		goto out_mm;
-	}
-	umem->odp_data->umem = umem;
-
-	mutex_init(&umem->odp_data->umem_mutex);
+	mutex_init(&umem_odp->umem_mutex);
 
-	init_completion(&umem->odp_data->notifier_completion);
+	init_completion(&umem_odp->notifier_completion);
 
 	if (ib_umem_num_pages(umem)) {
-		umem->odp_data->page_list =
-			vzalloc(array_size(sizeof(*umem->odp_data->page_list),
+		umem_odp->page_list =
+			vzalloc(array_size(sizeof(*umem_odp->page_list),
 					   ib_umem_num_pages(umem)));
-		if (!umem->odp_data->page_list) {
-			ret_val = -ENOMEM;
-			goto out_odp_data;
-		}
+		if (!umem_odp->page_list)
+			return -ENOMEM;
 
-		umem->odp_data->dma_list =
-			vzalloc(array_size(sizeof(*umem->odp_data->dma_list),
+		umem_odp->dma_list =
+			vzalloc(array_size(sizeof(*umem_odp->dma_list),
 					   ib_umem_num_pages(umem)));
-		if (!umem->odp_data->dma_list) {
+		if (!umem_odp->dma_list) {
 			ret_val = -ENOMEM;
 			goto out_page_list;
 		}
 	}
 
-	/*
-	 * When using MMU notifiers, we will get a
-	 * notification before the "current" task (and MM) is
-	 * destroyed. We use the umem_rwsem semaphore to synchronize.
-	 */
-	down_write(&context->umem_rwsem);
-	context->odp_mrs_count++;
-	if (likely(ib_umem_start(umem) != ib_umem_end(umem)))
-		rbt_ib_umem_insert(&umem->odp_data->interval_tree,
-				   &context->umem_tree);
-	if (likely(!atomic_read(&context->notifier_count)) ||
-	    context->odp_mrs_count == 1)
-		umem->odp_data->mn_counters_active = true;
-	else
-		list_add(&umem->odp_data->no_private_counters,
-			 &context->no_private_counters);
-	downgrade_write(&context->umem_rwsem);
-
-	if (context->odp_mrs_count == 1) {
-		/*
-		 * Note that at this point, no MMU notifier is running
-		 * for this context!
-		 */
-		atomic_set(&context->notifier_count, 0);
-		INIT_HLIST_NODE(&context->mn.hlist);
-		context->mn.ops = &ib_umem_notifiers;
-		/*
-		 * Lock-dep detects a false positive for mmap_sem vs.
-		 * umem_rwsem, due to not grasping downgrade_write correctly.
-		 */
-		lockdep_off();
-		ret_val = mmu_notifier_register(&context->mn, mm);
-		lockdep_on();
-		if (ret_val) {
-			pr_err("Failed to register mmu_notifier %d\n", ret_val);
-			ret_val = -EBUSY;
-			goto out_mutex;
-		}
-	}
-
-	up_read(&context->umem_rwsem);
+	ret_val = get_per_mm(umem_odp);
+	if (ret_val)
+		goto out_dma_list;
+	add_umem_to_per_mm(umem_odp);
 
-	/*
-	 * Note that doing an mmput can cause a notifier for the relevant mm.
-	 * If the notifier is called while we hold the umem_rwsem, this will
-	 * cause a deadlock. Therefore, we release the reference only after we
-	 * released the semaphore.
-	 */
-	mmput(mm);
 	return 0;
 
-out_mutex:
-	up_read(&context->umem_rwsem);
-	vfree(umem->odp_data->dma_list);
+out_dma_list:
+	vfree(umem_odp->dma_list);
 out_page_list:
-	vfree(umem->odp_data->page_list);
-out_odp_data:
-	kfree(umem->odp_data);
-out_mm:
-	mmput(mm);
+	vfree(umem_odp->page_list);
 	return ret_val;
 }
 
-void ib_umem_odp_release(struct ib_umem *umem)
+void ib_umem_odp_release(struct ib_umem_odp *umem_odp)
 {
-	struct ib_ucontext *context = umem->context;
+	struct ib_umem *umem = &umem_odp->umem;
 
 	/*
 	 * Ensure that no more pages are mapped in the umem.
@@ -478,61 +474,13 @@ void ib_umem_odp_release(struct ib_umem *umem)
 	 * It is the driver's responsibility to ensure, before calling us,
 	 * that the hardware will not attempt to access the MR any more.
 	 */
-	ib_umem_odp_unmap_dma_pages(umem, ib_umem_start(umem),
+	ib_umem_odp_unmap_dma_pages(umem_odp, ib_umem_start(umem),
 				    ib_umem_end(umem));
 
-	down_write(&context->umem_rwsem);
-	if (likely(ib_umem_start(umem) != ib_umem_end(umem)))
-		rbt_ib_umem_remove(&umem->odp_data->interval_tree,
-				   &context->umem_tree);
-	context->odp_mrs_count--;
-	if (!umem->odp_data->mn_counters_active) {
-		list_del(&umem->odp_data->no_private_counters);
-		complete_all(&umem->odp_data->notifier_completion);
-	}
-
-	/*
-	 * Downgrade the lock to a read lock. This ensures that the notifiers
-	 * (who lock the mutex for reading) will be able to finish, and we
-	 * will be able to enventually obtain the mmu notifiers SRCU. Note
-	 * that since we are doing it atomically, no other user could register
-	 * and unregister while we do the check.
-	 */
-	downgrade_write(&context->umem_rwsem);
-	if (!context->odp_mrs_count) {
-		struct task_struct *owning_process = NULL;
-		struct mm_struct *owning_mm        = NULL;
-
-		owning_process = get_pid_task(context->tgid,
-					      PIDTYPE_PID);
-		if (owning_process == NULL)
-			/*
-			 * The process is already dead, notifier were removed
-			 * already.
-			 */
-			goto out;
-
-		owning_mm = get_task_mm(owning_process);
-		if (owning_mm == NULL)
-			/*
-			 * The process' mm is already dead, notifier were
-			 * removed already.
-			 */
-			goto out_put_task;
-		mmu_notifier_unregister(&context->mn, owning_mm);
-
-		mmput(owning_mm);
-
-out_put_task:
-		put_task_struct(owning_process);
-	}
-out:
-	up_read(&context->umem_rwsem);
-
-	vfree(umem->odp_data->dma_list);
-	vfree(umem->odp_data->page_list);
-	kfree(umem->odp_data);
-	kfree(umem);
+	remove_umem_from_per_mm(umem_odp);
+	put_per_mm(umem_odp);
+	vfree(umem_odp->dma_list);
+	vfree(umem_odp->page_list);
 }
 
 /*
@@ -544,7 +492,7 @@ out:
  * @access_mask: access permissions needed for this page.
  * @current_seq: sequence number for synchronization with invalidations.
  *               the sequence number is taken from
- *               umem->odp_data->notifiers_seq.
+ *               umem_odp->notifiers_seq.
  *
  * The function returns -EFAULT if the DMA mapping operation fails. It returns
  * -EAGAIN if a concurrent invalidation prevents us from updating the page.
@@ -554,12 +502,13 @@ out:
  * umem.
  */
 static int ib_umem_odp_map_dma_single_page(
-		struct ib_umem *umem,
+		struct ib_umem_odp *umem_odp,
 		int page_index,
 		struct page *page,
 		u64 access_mask,
 		unsigned long current_seq)
 {
+	struct ib_umem *umem = &umem_odp->umem;
 	struct ib_device *dev = umem->context->device;
 	dma_addr_t dma_addr;
 	int stored_page = 0;
@@ -571,11 +520,11 @@ static int ib_umem_odp_map_dma_single_page(
 	 * handle case of a racing notifier. This check also allows us to bail
 	 * early if we have a notifier running in parallel with us.
 	 */
-	if (ib_umem_mmu_notifier_retry(umem, current_seq)) {
+	if (ib_umem_mmu_notifier_retry(umem_odp, current_seq)) {
 		ret = -EAGAIN;
 		goto out;
 	}
-	if (!(umem->odp_data->dma_list[page_index])) {
+	if (!(umem_odp->dma_list[page_index])) {
 		dma_addr = ib_dma_map_page(dev,
 					   page,
 					   0, BIT(umem->page_shift),
@@ -584,15 +533,15 @@ static int ib_umem_odp_map_dma_single_page(
 			ret = -EFAULT;
 			goto out;
 		}
-		umem->odp_data->dma_list[page_index] = dma_addr | access_mask;
-		umem->odp_data->page_list[page_index] = page;
+		umem_odp->dma_list[page_index] = dma_addr | access_mask;
+		umem_odp->page_list[page_index] = page;
 		umem->npages++;
 		stored_page = 1;
-	} else if (umem->odp_data->page_list[page_index] == page) {
-		umem->odp_data->dma_list[page_index] |= access_mask;
+	} else if (umem_odp->page_list[page_index] == page) {
+		umem_odp->dma_list[page_index] |= access_mask;
 	} else {
 		pr_err("error: got different pages in IB device and from get_user_pages. IB device page: %p, gup page: %p\n",
-		       umem->odp_data->page_list[page_index], page);
+		       umem_odp->page_list[page_index], page);
 		/* Better remove the mapping now, to prevent any further
 		 * damage. */
 		remove_existing_mapping = 1;
@@ -605,7 +554,7 @@ out:
 
 	if (remove_existing_mapping && umem->context->invalidate_range) {
 		invalidate_page_trampoline(
-			umem,
+			umem_odp,
 			ib_umem_start(umem) + (page_index >> umem->page_shift),
 			ib_umem_start(umem) + ((page_index + 1) >>
 					       umem->page_shift),
@@ -621,7 +570,7 @@ out:
  *
  * Pins the range of pages passed in the argument, and maps them to
  * DMA addresses. The DMA addresses of the mapped pages is updated in
- * umem->odp_data->dma_list.
+ * umem_odp->dma_list.
  *
  * Returns the number of pages mapped in success, negative error code
  * for failure.
@@ -629,7 +578,7 @@ out:
  * the function from completing its task.
  * An -ENOENT error code indicates that userspace process is being terminated
  * and mm was already destroyed.
- * @umem: the umem to map and pin
+ * @umem_odp: the umem to map and pin
  * @user_virt: the address from which we need to map.
  * @bcnt: the minimal number of bytes to pin and map. The mapping might be
  *        bigger due to alignment, and may also be smaller in case of an error
@@ -639,13 +588,15 @@ out:
  *               range.
  * @current_seq: the MMU notifiers sequance value for synchronization with
  *               invalidations. the sequance number is read from
- *               umem->odp_data->notifiers_seq before calling this function
+ *               umem_odp->notifiers_seq before calling this function
  */
-int ib_umem_odp_map_dma_pages(struct ib_umem *umem, u64 user_virt, u64 bcnt,
-			      u64 access_mask, unsigned long current_seq)
+int ib_umem_odp_map_dma_pages(struct ib_umem_odp *umem_odp, u64 user_virt,
+			      u64 bcnt, u64 access_mask,
+			      unsigned long current_seq)
 {
+	struct ib_umem *umem = &umem_odp->umem;
 	struct task_struct *owning_process  = NULL;
-	struct mm_struct   *owning_mm       = NULL;
+	struct mm_struct *owning_mm = umem_odp->umem.owning_mm;
 	struct page       **local_page_list = NULL;
 	u64 page_mask, off;
 	int j, k, ret = 0, start_idx, npages = 0, page_shift;
@@ -669,15 +620,14 @@ int ib_umem_odp_map_dma_pages(struct ib_umem *umem, u64 user_virt, u64 bcnt,
 	user_virt = user_virt & page_mask;
 	bcnt += off; /* Charge for the first page offset as well. */
 
-	owning_process = get_pid_task(umem->context->tgid, PIDTYPE_PID);
-	if (owning_process == NULL) {
+	/*
+	 * owning_process is allowed to be NULL, this means somehow the mm is
+	 * existing beyond the lifetime of the originating process.. Presumably
+	 * mmget_not_zero will fail in this case.
+	 */
+	owning_process = get_pid_task(umem_odp->per_mm->tgid, PIDTYPE_PID);
+	if (WARN_ON(!mmget_not_zero(umem_odp->umem.owning_mm))) {
 		ret = -EINVAL;
-		goto out_no_task;
-	}
-
-	owning_mm = get_task_mm(owning_process);
-	if (owning_mm == NULL) {
-		ret = -ENOENT;
 		goto out_put_task;
 	}
 
@@ -709,7 +659,7 @@ int ib_umem_odp_map_dma_pages(struct ib_umem *umem, u64 user_virt, u64 bcnt,
 			break;
 
 		bcnt -= min_t(size_t, npages << PAGE_SHIFT, bcnt);
-		mutex_lock(&umem->odp_data->umem_mutex);
+		mutex_lock(&umem_odp->umem_mutex);
 		for (j = 0; j < npages; j++, user_virt += PAGE_SIZE) {
 			if (user_virt & ~page_mask) {
 				p += PAGE_SIZE;
@@ -722,7 +672,7 @@ int ib_umem_odp_map_dma_pages(struct ib_umem *umem, u64 user_virt, u64 bcnt,
 			}
 
 			ret = ib_umem_odp_map_dma_single_page(
-					umem, k, local_page_list[j],
+					umem_odp, k, local_page_list[j],
 					access_mask, current_seq);
 			if (ret < 0)
 				break;
@@ -730,7 +680,7 @@ int ib_umem_odp_map_dma_pages(struct ib_umem *umem, u64 user_virt, u64 bcnt,
 			p = page_to_phys(local_page_list[j]);
 			k++;
 		}
-		mutex_unlock(&umem->odp_data->umem_mutex);
+		mutex_unlock(&umem_odp->umem_mutex);
 
 		if (ret < 0) {
 			/* Release left over pages when handling errors. */
@@ -749,16 +699,17 @@ int ib_umem_odp_map_dma_pages(struct ib_umem *umem, u64 user_virt, u64 bcnt,
 
 	mmput(owning_mm);
 out_put_task:
-	put_task_struct(owning_process);
-out_no_task:
+	if (owning_process)
+		put_task_struct(owning_process);
 	free_page((unsigned long)local_page_list);
 	return ret;
 }
 EXPORT_SYMBOL(ib_umem_odp_map_dma_pages);
 
-void ib_umem_odp_unmap_dma_pages(struct ib_umem *umem, u64 virt,
+void ib_umem_odp_unmap_dma_pages(struct ib_umem_odp *umem_odp, u64 virt,
 				 u64 bound)
 {
+	struct ib_umem *umem = &umem_odp->umem;
 	int idx;
 	u64 addr;
 	struct ib_device *dev = umem->context->device;
@@ -770,12 +721,12 @@ void ib_umem_odp_unmap_dma_pages(struct ib_umem *umem, u64 virt,
 	 * faults from completion. We might be racing with other
 	 * invalidations, so we must make sure we free each page only
 	 * once. */
-	mutex_lock(&umem->odp_data->umem_mutex);
+	mutex_lock(&umem_odp->umem_mutex);
 	for (addr = virt; addr < bound; addr += BIT(umem->page_shift)) {
 		idx = (addr - ib_umem_start(umem)) >> umem->page_shift;
-		if (umem->odp_data->page_list[idx]) {
-			struct page *page = umem->odp_data->page_list[idx];
-			dma_addr_t dma = umem->odp_data->dma_list[idx];
+		if (umem_odp->page_list[idx]) {
+			struct page *page = umem_odp->page_list[idx];
+			dma_addr_t dma = umem_odp->dma_list[idx];
 			dma_addr_t dma_addr = dma & ODP_DMA_ADDR_MASK;
 
 			WARN_ON(!dma_addr);
@@ -798,12 +749,12 @@ void ib_umem_odp_unmap_dma_pages(struct ib_umem *umem, u64 virt,
 			/* on demand pinning support */
 			if (!umem->context->invalidate_range)
 				put_page(page);
-			umem->odp_data->page_list[idx] = NULL;
-			umem->odp_data->dma_list[idx] = 0;
+			umem_odp->page_list[idx] = NULL;
+			umem_odp->dma_list[idx] = 0;
 			umem->npages--;
 		}
 	}
-	mutex_unlock(&umem->odp_data->umem_mutex);
+	mutex_unlock(&umem_odp->umem_mutex);
 }
 EXPORT_SYMBOL(ib_umem_odp_unmap_dma_pages);
 
@@ -830,7 +781,7 @@ int rbt_ib_umem_for_each_in_range(struct rb_root_cached *root,
 			return -EAGAIN;
 		next = rbt_ib_umem_iter_next(node, start, last - 1);
 		umem = container_of(node, struct ib_umem_odp, interval_tree);
-		ret_val = cb(umem->umem, start, last, cookie) || ret_val;
+		ret_val = cb(umem, start, last, cookie) || ret_val;
 	}
 
 	return ret_val;
diff --git a/drivers/infiniband/core/user_mad.c b/drivers/infiniband/core/user_mad.c
index c34a6852d691..f55f48f6b272 100644
--- a/drivers/infiniband/core/user_mad.c
+++ b/drivers/infiniband/core/user_mad.c
@@ -138,7 +138,7 @@ static const dev_t base_issm_dev = MKDEV(IB_UMAD_MAJOR, IB_UMAD_MINOR_BASE) +
 static dev_t dynamic_umad_dev;
 static dev_t dynamic_issm_dev;
 
-static DECLARE_BITMAP(dev_map, IB_UMAD_MAX_PORTS);
+static DEFINE_IDA(umad_ida);
 
 static void ib_umad_add_one(struct ib_device *device);
 static void ib_umad_remove_one(struct ib_device *device, void *client_data);
@@ -1132,7 +1132,7 @@ static ssize_t show_ibdev(struct device *dev, struct device_attribute *attr,
 	if (!port)
 		return -ENODEV;
 
-	return sprintf(buf, "%s\n", port->ib_dev->name);
+	return sprintf(buf, "%s\n", dev_name(&port->ib_dev->dev));
 }
 static DEVICE_ATTR(ibdev, S_IRUGO, show_ibdev, NULL);
 
@@ -1159,11 +1159,10 @@ static int ib_umad_init_port(struct ib_device *device, int port_num,
 	dev_t base_umad;
 	dev_t base_issm;
 
-	devnum = find_first_zero_bit(dev_map, IB_UMAD_MAX_PORTS);
-	if (devnum >= IB_UMAD_MAX_PORTS)
+	devnum = ida_alloc_max(&umad_ida, IB_UMAD_MAX_PORTS - 1, GFP_KERNEL);
+	if (devnum < 0)
 		return -1;
 	port->dev_num = devnum;
-	set_bit(devnum, dev_map);
 	if (devnum >= IB_UMAD_NUM_FIXED_MINOR) {
 		base_umad = dynamic_umad_dev + devnum - IB_UMAD_NUM_FIXED_MINOR;
 		base_issm = dynamic_issm_dev + devnum - IB_UMAD_NUM_FIXED_MINOR;
@@ -1227,7 +1226,7 @@ err_dev:
 
 err_cdev:
 	cdev_del(&port->cdev);
-	clear_bit(devnum, dev_map);
+	ida_free(&umad_ida, devnum);
 
 	return -1;
 }
@@ -1261,7 +1260,7 @@ static void ib_umad_kill_port(struct ib_umad_port *port)
 	}
 
 	mutex_unlock(&port->file_mutex);
-	clear_bit(port->dev_num, dev_map);
+	ida_free(&umad_ida, port->dev_num);
 }
 
 static void ib_umad_add_one(struct ib_device *device)
diff --git a/drivers/infiniband/core/uverbs.h b/drivers/infiniband/core/uverbs.h
index 5df8e548cc14..c97935a0c7c6 100644
--- a/drivers/infiniband/core/uverbs.h
+++ b/drivers/infiniband/core/uverbs.h
@@ -100,13 +100,14 @@ struct ib_uverbs_device {
 	atomic_t				refcount;
 	int					num_comp_vectors;
 	struct completion			comp;
-	struct device			       *dev;
+	struct device				dev;
+	/* First group for device attributes, NULL terminated array */
+	const struct attribute_group		*groups[2];
 	struct ib_device	__rcu	       *ib_dev;
 	int					devnum;
 	struct cdev			        cdev;
 	struct rb_root				xrcd_tree;
 	struct mutex				xrcd_tree_mutex;
-	struct kobject				kobj;
 	struct srcu_struct			disassociate_srcu;
 	struct mutex				lists_mutex; /* protect lists */
 	struct list_head			uverbs_file_list;
@@ -146,7 +147,6 @@ struct ib_uverbs_file {
 	struct ib_event_handler			event_handler;
 	struct ib_uverbs_async_event_file       *async_file;
 	struct list_head			list;
-	int					is_closed;
 
 	/*
 	 * To access the uobjects list hw_destroy_rwsem must be held for write
@@ -158,6 +158,9 @@ struct ib_uverbs_file {
 	spinlock_t		uobjects_lock;
 	struct list_head	uobjects;
 
+	struct mutex umap_lock;
+	struct list_head umaps;
+
 	u64 uverbs_cmd_mask;
 	u64 uverbs_ex_cmd_mask;
 
@@ -218,12 +221,6 @@ struct ib_ucq_object {
 	u32			async_events_reported;
 };
 
-struct ib_uflow_resources;
-struct ib_uflow_object {
-	struct ib_uobject		uobject;
-	struct ib_uflow_resources	*resources;
-};
-
 extern const struct file_operations uverbs_event_fops;
 void ib_uverbs_init_event_queue(struct ib_uverbs_event_queue *ev_queue);
 struct file *ib_uverbs_alloc_async_event_file(struct ib_uverbs_file *uverbs_file,
diff --git a/drivers/infiniband/core/uverbs_cmd.c b/drivers/infiniband/core/uverbs_cmd.c
index e012ca80f9d1..a93853770e3c 100644
--- a/drivers/infiniband/core/uverbs_cmd.c
+++ b/drivers/infiniband/core/uverbs_cmd.c
@@ -117,18 +117,12 @@ ssize_t ib_uverbs_get_context(struct ib_uverbs_file *file,
 	/* ufile is required when some objects are released */
 	ucontext->ufile = file;
 
-	rcu_read_lock();
-	ucontext->tgid = get_task_pid(current->group_leader, PIDTYPE_PID);
-	rcu_read_unlock();
-	ucontext->closing = 0;
+	ucontext->closing = false;
 	ucontext->cleanup_retryable = false;
 
 #ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
-	ucontext->umem_tree = RB_ROOT_CACHED;
-	init_rwsem(&ucontext->umem_rwsem);
-	ucontext->odp_mrs_count = 0;
-	INIT_LIST_HEAD(&ucontext->no_private_counters);
-
+	mutex_init(&ucontext->per_mm_list_lock);
+	INIT_LIST_HEAD(&ucontext->per_mm_list);
 	if (!(ib_dev->attrs.device_cap_flags & IB_DEVICE_ON_DEMAND_PAGING))
 		ucontext->invalidate_range = NULL;
 
@@ -172,7 +166,6 @@ err_fd:
 	put_unused_fd(resp.async_fd);
 
 err_free:
-	put_pid(ucontext->tgid);
 	ib_dev->dealloc_ucontext(ucontext);
 
 err_alloc:
@@ -2769,16 +2762,7 @@ out_put:
 	return ret ? ret : in_len;
 }
 
-struct ib_uflow_resources {
-	size_t			max;
-	size_t			num;
-	size_t			collection_num;
-	size_t			counters_num;
-	struct ib_counters	**counters;
-	struct ib_flow_action	**collection;
-};
-
-static struct ib_uflow_resources *flow_resources_alloc(size_t num_specs)
+struct ib_uflow_resources *flow_resources_alloc(size_t num_specs)
 {
 	struct ib_uflow_resources *resources;
 
@@ -2808,6 +2792,7 @@ err:
 
 	return NULL;
 }
+EXPORT_SYMBOL(flow_resources_alloc);
 
 void ib_uverbs_flow_resources_free(struct ib_uflow_resources *uflow_res)
 {
@@ -2826,10 +2811,11 @@ void ib_uverbs_flow_resources_free(struct ib_uflow_resources *uflow_res)
 	kfree(uflow_res->counters);
 	kfree(uflow_res);
 }
+EXPORT_SYMBOL(ib_uverbs_flow_resources_free);
 
-static void flow_resources_add(struct ib_uflow_resources *uflow_res,
-			       enum ib_flow_spec_type type,
-			       void *ibobj)
+void flow_resources_add(struct ib_uflow_resources *uflow_res,
+			enum ib_flow_spec_type type,
+			void *ibobj)
 {
 	WARN_ON(uflow_res->num >= uflow_res->max);
 
@@ -2850,6 +2836,7 @@ static void flow_resources_add(struct ib_uflow_resources *uflow_res,
 
 	uflow_res->num++;
 }
+EXPORT_SYMBOL(flow_resources_add);
 
 static int kern_spec_to_ib_spec_action(struct ib_uverbs_file *ufile,
 				       struct ib_uverbs_flow_spec *kern_spec,
@@ -3484,7 +3471,6 @@ int ib_uverbs_ex_create_flow(struct ib_uverbs_file *file,
 	struct ib_uverbs_create_flow	  cmd;
 	struct ib_uverbs_create_flow_resp resp;
 	struct ib_uobject		  *uobj;
-	struct ib_uflow_object		  *uflow;
 	struct ib_flow			  *flow_id;
 	struct ib_uverbs_flow_attr	  *kern_flow_attr;
 	struct ib_flow_attr		  *flow_attr;
@@ -3623,13 +3609,8 @@ int ib_uverbs_ex_create_flow(struct ib_uverbs_file *file,
 		err = PTR_ERR(flow_id);
 		goto err_free;
 	}
-	atomic_inc(&qp->usecnt);
-	flow_id->qp = qp;
-	flow_id->device = qp->device;
-	flow_id->uobject = uobj;
-	uobj->object = flow_id;
-	uflow = container_of(uobj, typeof(*uflow), uobject);
-	uflow->resources = uflow_res;
+
+	ib_set_flow(uobj, flow_id, qp, qp->device, uflow_res);
 
 	memset(&resp, 0, sizeof(resp));
 	resp.flow_handle = uobj->id;
diff --git a/drivers/infiniband/core/uverbs_ioctl.c b/drivers/infiniband/core/uverbs_ioctl.c
index 1a6b229e3db3..b0e493e8d860 100644
--- a/drivers/infiniband/core/uverbs_ioctl.c
+++ b/drivers/infiniband/core/uverbs_ioctl.c
@@ -57,6 +57,7 @@ struct bundle_priv {
 	struct ib_uverbs_attr *uattrs;
 
 	DECLARE_BITMAP(uobj_finalize, UVERBS_API_ATTR_BKEY_LEN);
+	DECLARE_BITMAP(spec_finalize, UVERBS_API_ATTR_BKEY_LEN);
 
 	/*
 	 * Must be last. bundle ends in a flex array which overlaps
@@ -143,6 +144,86 @@ static bool uverbs_is_attr_cleared(const struct ib_uverbs_attr *uattr,
 			   0, uattr->len - len);
 }
 
+static int uverbs_process_idrs_array(struct bundle_priv *pbundle,
+				     const struct uverbs_api_attr *attr_uapi,
+				     struct uverbs_objs_arr_attr *attr,
+				     struct ib_uverbs_attr *uattr,
+				     u32 attr_bkey)
+{
+	const struct uverbs_attr_spec *spec = &attr_uapi->spec;
+	size_t array_len;
+	u32 *idr_vals;
+	int ret = 0;
+	size_t i;
+
+	if (uattr->attr_data.reserved)
+		return -EINVAL;
+
+	if (uattr->len % sizeof(u32))
+		return -EINVAL;
+
+	array_len = uattr->len / sizeof(u32);
+	if (array_len < spec->u2.objs_arr.min_len ||
+	    array_len > spec->u2.objs_arr.max_len)
+		return -EINVAL;
+
+	attr->uobjects =
+		uverbs_alloc(&pbundle->bundle,
+			     array_size(array_len, sizeof(*attr->uobjects)));
+	if (IS_ERR(attr->uobjects))
+		return PTR_ERR(attr->uobjects);
+
+	/*
+	 * Since idr is 4B and *uobjects is >= 4B, we can use attr->uobjects
+	 * to store idrs array and avoid additional memory allocation. The
+	 * idrs array is offset to the end of the uobjects array so we will be
+	 * able to read idr and replace with a pointer.
+	 */
+	idr_vals = (u32 *)(attr->uobjects + array_len) - array_len;
+
+	if (uattr->len > sizeof(uattr->data)) {
+		ret = copy_from_user(idr_vals, u64_to_user_ptr(uattr->data),
+				     uattr->len);
+		if (ret)
+			return -EFAULT;
+	} else {
+		memcpy(idr_vals, &uattr->data, uattr->len);
+	}
+
+	for (i = 0; i != array_len; i++) {
+		attr->uobjects[i] = uverbs_get_uobject_from_file(
+			spec->u2.objs_arr.obj_type, pbundle->bundle.ufile,
+			spec->u2.objs_arr.access, idr_vals[i]);
+		if (IS_ERR(attr->uobjects[i])) {
+			ret = PTR_ERR(attr->uobjects[i]);
+			break;
+		}
+	}
+
+	attr->len = i;
+	__set_bit(attr_bkey, pbundle->spec_finalize);
+	return ret;
+}
+
+static int uverbs_free_idrs_array(const struct uverbs_api_attr *attr_uapi,
+				  struct uverbs_objs_arr_attr *attr,
+				  bool commit)
+{
+	const struct uverbs_attr_spec *spec = &attr_uapi->spec;
+	int current_ret;
+	int ret = 0;
+	size_t i;
+
+	for (i = 0; i != attr->len; i++) {
+		current_ret = uverbs_finalize_object(
+			attr->uobjects[i], spec->u2.objs_arr.access, commit);
+		if (!ret)
+			ret = current_ret;
+	}
+
+	return ret;
+}
+
 static int uverbs_process_attr(struct bundle_priv *pbundle,
 			       const struct uverbs_api_attr *attr_uapi,
 			       struct ib_uverbs_attr *uattr, u32 attr_bkey)
@@ -246,6 +327,11 @@ static int uverbs_process_attr(struct bundle_priv *pbundle,
 		}
 
 		break;
+
+	case UVERBS_ATTR_TYPE_IDRS_ARRAY:
+		return uverbs_process_idrs_array(pbundle, attr_uapi,
+						 &e->objs_arr_attr, uattr,
+						 attr_bkey);
 	default:
 		return -EOPNOTSUPP;
 	}
@@ -300,8 +386,7 @@ static int uverbs_set_attr(struct bundle_priv *pbundle,
 			return -EPROTONOSUPPORT;
 		return 0;
 	}
-	attr = srcu_dereference(
-		*slot, &pbundle->bundle.ufile->device->disassociate_srcu);
+	attr = rcu_dereference_protected(*slot, true);
 
 	/* Reject duplicate attributes from user-space */
 	if (test_bit(attr_bkey, pbundle->bundle.attr_present))
@@ -384,6 +469,7 @@ static int bundle_destroy(struct bundle_priv *pbundle, bool commit)
 	unsigned int i;
 	int ret = 0;
 
+	/* fast path for simple uobjects */
 	i = -1;
 	while ((i = find_next_bit(pbundle->uobj_finalize, key_bitmap_len,
 				  i + 1)) < key_bitmap_len) {
@@ -397,6 +483,30 @@ static int bundle_destroy(struct bundle_priv *pbundle, bool commit)
 			ret = current_ret;
 	}
 
+	i = -1;
+	while ((i = find_next_bit(pbundle->spec_finalize, key_bitmap_len,
+				  i + 1)) < key_bitmap_len) {
+		struct uverbs_attr *attr = &pbundle->bundle.attrs[i];
+		const struct uverbs_api_attr *attr_uapi;
+		void __rcu **slot;
+		int current_ret;
+
+		slot = uapi_get_attr_for_method(
+			pbundle,
+			pbundle->method_key | uapi_bkey_to_key_attr(i));
+		if (WARN_ON(!slot))
+			continue;
+
+		attr_uapi = rcu_dereference_protected(*slot, true);
+
+		if (attr_uapi->spec.type == UVERBS_ATTR_TYPE_IDRS_ARRAY) {
+			current_ret = uverbs_free_idrs_array(
+				attr_uapi, &attr->objs_arr_attr, commit);
+			if (!ret)
+				ret = current_ret;
+		}
+	}
+
 	for (memblock = pbundle->allocated_mem; memblock;) {
 		struct bundle_alloc_head *tmp = memblock;
 
@@ -429,7 +539,7 @@ static int ib_uverbs_cmd_verbs(struct ib_uverbs_file *ufile,
 			uapi_key_ioctl_method(hdr->method_id));
 	if (unlikely(!slot))
 		return -EPROTONOSUPPORT;
-	method_elm = srcu_dereference(*slot, &ufile->device->disassociate_srcu);
+	method_elm = rcu_dereference_protected(*slot, true);
 
 	if (!method_elm->use_stack) {
 		pbundle = kmalloc(method_elm->bundle_size, GFP_KERNEL);
@@ -461,6 +571,7 @@ static int ib_uverbs_cmd_verbs(struct ib_uverbs_file *ufile,
 	memset(pbundle->bundle.attr_present, 0,
 	       sizeof(pbundle->bundle.attr_present));
 	memset(pbundle->uobj_finalize, 0, sizeof(pbundle->uobj_finalize));
+	memset(pbundle->spec_finalize, 0, sizeof(pbundle->spec_finalize));
 
 	ret = ib_uverbs_run_method(pbundle, hdr->num_attrs);
 	destroy_ret = bundle_destroy(pbundle, ret == 0);
@@ -611,3 +722,26 @@ int uverbs_copy_to(const struct uverbs_attr_bundle *bundle, size_t idx,
 	return 0;
 }
 EXPORT_SYMBOL(uverbs_copy_to);
+
+int _uverbs_get_const(s64 *to, const struct uverbs_attr_bundle *attrs_bundle,
+		      size_t idx, s64 lower_bound, u64 upper_bound,
+		      s64  *def_val)
+{
+	const struct uverbs_attr *attr;
+
+	attr = uverbs_attr_get(attrs_bundle, idx);
+	if (IS_ERR(attr)) {
+		if ((PTR_ERR(attr) != -ENOENT) || !def_val)
+			return PTR_ERR(attr);
+
+		*to = *def_val;
+	} else {
+		*to = attr->ptr_attr.data;
+	}
+
+	if (*to < lower_bound || (*to > 0 && (u64)*to > upper_bound))
+		return -EINVAL;
+
+	return 0;
+}
+EXPORT_SYMBOL(_uverbs_get_const);
diff --git a/drivers/infiniband/core/uverbs_main.c b/drivers/infiniband/core/uverbs_main.c
index 50152c1b1004..6d373f5515b7 100644
--- a/drivers/infiniband/core/uverbs_main.c
+++ b/drivers/infiniband/core/uverbs_main.c
@@ -45,6 +45,7 @@
 #include <linux/cdev.h>
 #include <linux/anon_inodes.h>
 #include <linux/slab.h>
+#include <linux/sched/mm.h>
 
 #include <linux/uaccess.h>
 
@@ -72,7 +73,7 @@ enum {
 static dev_t dynamic_uverbs_dev;
 static struct class *uverbs_class;
 
-static DECLARE_BITMAP(dev_map, IB_UVERBS_MAX_DEVICES);
+static DEFINE_IDA(uverbs_ida);
 
 static ssize_t (*uverbs_cmd_table[])(struct ib_uverbs_file *file,
 				     const char __user *buf, int in_len,
@@ -169,20 +170,16 @@ int uverbs_dealloc_mw(struct ib_mw *mw)
 	return ret;
 }
 
-static void ib_uverbs_release_dev(struct kobject *kobj)
+static void ib_uverbs_release_dev(struct device *device)
 {
 	struct ib_uverbs_device *dev =
-		container_of(kobj, struct ib_uverbs_device, kobj);
+			container_of(device, struct ib_uverbs_device, dev);
 
 	uverbs_destroy_api(dev->uapi);
 	cleanup_srcu_struct(&dev->disassociate_srcu);
 	kfree(dev);
 }
 
-static struct kobj_type ib_uverbs_dev_ktype = {
-	.release = ib_uverbs_release_dev,
-};
-
 static void ib_uverbs_release_async_event_file(struct kref *ref)
 {
 	struct ib_uverbs_async_event_file *file =
@@ -265,7 +262,7 @@ void ib_uverbs_release_file(struct kref *ref)
 	if (atomic_dec_and_test(&file->device->refcount))
 		ib_uverbs_comp_dev(file->device);
 
-	kobject_put(&file->device->kobj);
+	put_device(&file->device->dev);
 	kfree(file);
 }
 
@@ -817,6 +814,226 @@ out:
 }
 
 /*
+ * Each time we map IO memory into user space this keeps track of the mapping.
+ * When the device is hot-unplugged we 'zap' the mmaps in user space to point
+ * to the zero page and allow the hot unplug to proceed.
+ *
+ * This is necessary for cases like PCI physical hot unplug as the actual BAR
+ * memory may vanish after this and access to it from userspace could MCE.
+ *
+ * RDMA drivers supporting disassociation must have their user space designed
+ * to cope in some way with their IO pages going to the zero page.
+ */
+struct rdma_umap_priv {
+	struct vm_area_struct *vma;
+	struct list_head list;
+};
+
+static const struct vm_operations_struct rdma_umap_ops;
+
+static void rdma_umap_priv_init(struct rdma_umap_priv *priv,
+				struct vm_area_struct *vma)
+{
+	struct ib_uverbs_file *ufile = vma->vm_file->private_data;
+
+	priv->vma = vma;
+	vma->vm_private_data = priv;
+	vma->vm_ops = &rdma_umap_ops;
+
+	mutex_lock(&ufile->umap_lock);
+	list_add(&priv->list, &ufile->umaps);
+	mutex_unlock(&ufile->umap_lock);
+}
+
+/*
+ * The VMA has been dup'd, initialize the vm_private_data with a new tracking
+ * struct
+ */
+static void rdma_umap_open(struct vm_area_struct *vma)
+{
+	struct ib_uverbs_file *ufile = vma->vm_file->private_data;
+	struct rdma_umap_priv *opriv = vma->vm_private_data;
+	struct rdma_umap_priv *priv;
+
+	if (!opriv)
+		return;
+
+	/* We are racing with disassociation */
+	if (!down_read_trylock(&ufile->hw_destroy_rwsem))
+		goto out_zap;
+	/*
+	 * Disassociation already completed, the VMA should already be zapped.
+	 */
+	if (!ufile->ucontext)
+		goto out_unlock;
+
+	priv = kzalloc(sizeof(*priv), GFP_KERNEL);
+	if (!priv)
+		goto out_unlock;
+	rdma_umap_priv_init(priv, vma);
+
+	up_read(&ufile->hw_destroy_rwsem);
+	return;
+
+out_unlock:
+	up_read(&ufile->hw_destroy_rwsem);
+out_zap:
+	/*
+	 * We can't allow the VMA to be created with the actual IO pages, that
+	 * would break our API contract, and it can't be stopped at this
+	 * point, so zap it.
+	 */
+	vma->vm_private_data = NULL;
+	zap_vma_ptes(vma, vma->vm_start, vma->vm_end - vma->vm_start);
+}
+
+static void rdma_umap_close(struct vm_area_struct *vma)
+{
+	struct ib_uverbs_file *ufile = vma->vm_file->private_data;
+	struct rdma_umap_priv *priv = vma->vm_private_data;
+
+	if (!priv)
+		return;
+
+	/*
+	 * The vma holds a reference on the struct file that created it, which
+	 * in turn means that the ib_uverbs_file is guaranteed to exist at
+	 * this point.
+	 */
+	mutex_lock(&ufile->umap_lock);
+	list_del(&priv->list);
+	mutex_unlock(&ufile->umap_lock);
+	kfree(priv);
+}
+
+static const struct vm_operations_struct rdma_umap_ops = {
+	.open = rdma_umap_open,
+	.close = rdma_umap_close,
+};
+
+static struct rdma_umap_priv *rdma_user_mmap_pre(struct ib_ucontext *ucontext,
+						 struct vm_area_struct *vma,
+						 unsigned long size)
+{
+	struct ib_uverbs_file *ufile = ucontext->ufile;
+	struct rdma_umap_priv *priv;
+
+	if (vma->vm_end - vma->vm_start != size)
+		return ERR_PTR(-EINVAL);
+
+	/* Driver is using this wrong, must be called by ib_uverbs_mmap */
+	if (WARN_ON(!vma->vm_file ||
+		    vma->vm_file->private_data != ufile))
+		return ERR_PTR(-EINVAL);
+	lockdep_assert_held(&ufile->device->disassociate_srcu);
+
+	priv = kzalloc(sizeof(*priv), GFP_KERNEL);
+	if (!priv)
+		return ERR_PTR(-ENOMEM);
+	return priv;
+}
+
+/*
+ * Map IO memory into a process. This is to be called by drivers as part of
+ * their mmap() functions if they wish to send something like PCI-E BAR memory
+ * to userspace.
+ */
+int rdma_user_mmap_io(struct ib_ucontext *ucontext, struct vm_area_struct *vma,
+		      unsigned long pfn, unsigned long size, pgprot_t prot)
+{
+	struct rdma_umap_priv *priv = rdma_user_mmap_pre(ucontext, vma, size);
+
+	if (IS_ERR(priv))
+		return PTR_ERR(priv);
+
+	vma->vm_page_prot = prot;
+	if (io_remap_pfn_range(vma, vma->vm_start, pfn, size, prot)) {
+		kfree(priv);
+		return -EAGAIN;
+	}
+
+	rdma_umap_priv_init(priv, vma);
+	return 0;
+}
+EXPORT_SYMBOL(rdma_user_mmap_io);
+
+/*
+ * The page case is here for a slightly different reason, the driver expects
+ * to be able to free the page it is sharing to user space when it destroys
+ * its ucontext, which means we need to zap the user space references.
+ *
+ * We could handle this differently by providing an API to allocate a shared
+ * page and then only freeing the shared page when the last ufile is
+ * destroyed.
+ */
+int rdma_user_mmap_page(struct ib_ucontext *ucontext,
+			struct vm_area_struct *vma, struct page *page,
+			unsigned long size)
+{
+	struct rdma_umap_priv *priv = rdma_user_mmap_pre(ucontext, vma, size);
+
+	if (IS_ERR(priv))
+		return PTR_ERR(priv);
+
+	if (remap_pfn_range(vma, vma->vm_start, page_to_pfn(page), size,
+			    vma->vm_page_prot)) {
+		kfree(priv);
+		return -EAGAIN;
+	}
+
+	rdma_umap_priv_init(priv, vma);
+	return 0;
+}
+EXPORT_SYMBOL(rdma_user_mmap_page);
+
+void uverbs_user_mmap_disassociate(struct ib_uverbs_file *ufile)
+{
+	struct rdma_umap_priv *priv, *next_priv;
+
+	lockdep_assert_held(&ufile->hw_destroy_rwsem);
+
+	while (1) {
+		struct mm_struct *mm = NULL;
+
+		/* Get an arbitrary mm pointer that hasn't been cleaned yet */
+		mutex_lock(&ufile->umap_lock);
+		if (!list_empty(&ufile->umaps)) {
+			mm = list_first_entry(&ufile->umaps,
+					      struct rdma_umap_priv, list)
+				     ->vma->vm_mm;
+			mmget(mm);
+		}
+		mutex_unlock(&ufile->umap_lock);
+		if (!mm)
+			return;
+
+		/*
+		 * The umap_lock is nested under mmap_sem since it used within
+		 * the vma_ops callbacks, so we have to clean the list one mm
+		 * at a time to get the lock ordering right. Typically there
+		 * will only be one mm, so no big deal.
+		 */
+		down_write(&mm->mmap_sem);
+		mutex_lock(&ufile->umap_lock);
+		list_for_each_entry_safe (priv, next_priv, &ufile->umaps,
+					  list) {
+			struct vm_area_struct *vma = priv->vma;
+
+			if (vma->vm_mm != mm)
+				continue;
+			list_del_init(&priv->list);
+
+			zap_vma_ptes(vma, vma->vm_start,
+				     vma->vm_end - vma->vm_start);
+			vma->vm_flags &= ~(VM_SHARED | VM_MAYSHARE);
+		}
+		mutex_unlock(&ufile->umap_lock);
+		up_write(&mm->mmap_sem);
+		mmput(mm);
+	}
+}
+
+/*
  * ib_uverbs_open() does not need the BKL:
  *
  *  - the ib_uverbs_device structures are properly reference counted and
@@ -839,6 +1056,7 @@ static int ib_uverbs_open(struct inode *inode, struct file *filp)
 	if (!atomic_inc_not_zero(&dev->refcount))
 		return -ENXIO;
 
+	get_device(&dev->dev);
 	srcu_key = srcu_read_lock(&dev->disassociate_srcu);
 	mutex_lock(&dev->lists_mutex);
 	ib_dev = srcu_dereference(dev->ib_dev,
@@ -876,9 +1094,10 @@ static int ib_uverbs_open(struct inode *inode, struct file *filp)
 	spin_lock_init(&file->uobjects_lock);
 	INIT_LIST_HEAD(&file->uobjects);
 	init_rwsem(&file->hw_destroy_rwsem);
+	mutex_init(&file->umap_lock);
+	INIT_LIST_HEAD(&file->umaps);
 
 	filp->private_data = file;
-	kobject_get(&dev->kobj);
 	list_add_tail(&file->list, &dev->uverbs_file_list);
 	mutex_unlock(&dev->lists_mutex);
 	srcu_read_unlock(&dev->disassociate_srcu, srcu_key);
@@ -899,6 +1118,7 @@ err:
 	if (atomic_dec_and_test(&dev->refcount))
 		ib_uverbs_comp_dev(dev);
 
+	put_device(&dev->dev);
 	return ret;
 }
 
@@ -909,10 +1129,7 @@ static int ib_uverbs_close(struct inode *inode, struct file *filp)
 	uverbs_destroy_ufile_hw(file, RDMA_REMOVE_CLOSE);
 
 	mutex_lock(&file->device->lists_mutex);
-	if (!file->is_closed) {
-		list_del(&file->list);
-		file->is_closed = 1;
-	}
+	list_del_init(&file->list);
 	mutex_unlock(&file->device->lists_mutex);
 
 	if (file->async_file)
@@ -951,37 +1168,34 @@ static struct ib_client uverbs_client = {
 	.remove = ib_uverbs_remove_one
 };
 
-static ssize_t show_ibdev(struct device *device, struct device_attribute *attr,
+static ssize_t ibdev_show(struct device *device, struct device_attribute *attr,
 			  char *buf)
 {
+	struct ib_uverbs_device *dev =
+			container_of(device, struct ib_uverbs_device, dev);
 	int ret = -ENODEV;
 	int srcu_key;
-	struct ib_uverbs_device *dev = dev_get_drvdata(device);
 	struct ib_device *ib_dev;
 
-	if (!dev)
-		return -ENODEV;
-
 	srcu_key = srcu_read_lock(&dev->disassociate_srcu);
 	ib_dev = srcu_dereference(dev->ib_dev, &dev->disassociate_srcu);
 	if (ib_dev)
-		ret = sprintf(buf, "%s\n", ib_dev->name);
+		ret = sprintf(buf, "%s\n", dev_name(&ib_dev->dev));
 	srcu_read_unlock(&dev->disassociate_srcu, srcu_key);
 
 	return ret;
 }
-static DEVICE_ATTR(ibdev, S_IRUGO, show_ibdev, NULL);
+static DEVICE_ATTR_RO(ibdev);
 
-static ssize_t show_dev_abi_version(struct device *device,
-				    struct device_attribute *attr, char *buf)
+static ssize_t abi_version_show(struct device *device,
+				struct device_attribute *attr, char *buf)
 {
-	struct ib_uverbs_device *dev = dev_get_drvdata(device);
+	struct ib_uverbs_device *dev =
+			container_of(device, struct ib_uverbs_device, dev);
 	int ret = -ENODEV;
 	int srcu_key;
 	struct ib_device *ib_dev;
 
-	if (!dev)
-		return -ENODEV;
 	srcu_key = srcu_read_lock(&dev->disassociate_srcu);
 	ib_dev = srcu_dereference(dev->ib_dev, &dev->disassociate_srcu);
 	if (ib_dev)
@@ -990,7 +1204,17 @@ static ssize_t show_dev_abi_version(struct device *device,
 
 	return ret;
 }
-static DEVICE_ATTR(abi_version, S_IRUGO, show_dev_abi_version, NULL);
+static DEVICE_ATTR_RO(abi_version);
+
+static struct attribute *ib_dev_attrs[] = {
+	&dev_attr_abi_version.attr,
+	&dev_attr_ibdev.attr,
+	NULL,
+};
+
+static const struct attribute_group dev_attr_group = {
+	.attrs = ib_dev_attrs,
+};
 
 static CLASS_ATTR_STRING(abi_version, S_IRUGO,
 			 __stringify(IB_USER_VERBS_ABI_VERSION));
@@ -1028,65 +1252,56 @@ static void ib_uverbs_add_one(struct ib_device *device)
 		return;
 	}
 
+	device_initialize(&uverbs_dev->dev);
+	uverbs_dev->dev.class = uverbs_class;
+	uverbs_dev->dev.parent = device->dev.parent;
+	uverbs_dev->dev.release = ib_uverbs_release_dev;
+	uverbs_dev->groups[0] = &dev_attr_group;
+	uverbs_dev->dev.groups = uverbs_dev->groups;
 	atomic_set(&uverbs_dev->refcount, 1);
 	init_completion(&uverbs_dev->comp);
 	uverbs_dev->xrcd_tree = RB_ROOT;
 	mutex_init(&uverbs_dev->xrcd_tree_mutex);
-	kobject_init(&uverbs_dev->kobj, &ib_uverbs_dev_ktype);
 	mutex_init(&uverbs_dev->lists_mutex);
 	INIT_LIST_HEAD(&uverbs_dev->uverbs_file_list);
 	INIT_LIST_HEAD(&uverbs_dev->uverbs_events_file_list);
+	rcu_assign_pointer(uverbs_dev->ib_dev, device);
+	uverbs_dev->num_comp_vectors = device->num_comp_vectors;
 
-	devnum = find_first_zero_bit(dev_map, IB_UVERBS_MAX_DEVICES);
-	if (devnum >= IB_UVERBS_MAX_DEVICES)
+	devnum = ida_alloc_max(&uverbs_ida, IB_UVERBS_MAX_DEVICES - 1,
+			       GFP_KERNEL);
+	if (devnum < 0)
 		goto err;
 	uverbs_dev->devnum = devnum;
-	set_bit(devnum, dev_map);
 	if (devnum >= IB_UVERBS_NUM_FIXED_MINOR)
 		base = dynamic_uverbs_dev + devnum - IB_UVERBS_NUM_FIXED_MINOR;
 	else
 		base = IB_UVERBS_BASE_DEV + devnum;
 
-	rcu_assign_pointer(uverbs_dev->ib_dev, device);
-	uverbs_dev->num_comp_vectors = device->num_comp_vectors;
-
 	if (ib_uverbs_create_uapi(device, uverbs_dev))
 		goto err_uapi;
 
-	cdev_init(&uverbs_dev->cdev, NULL);
+	uverbs_dev->dev.devt = base;
+	dev_set_name(&uverbs_dev->dev, "uverbs%d", uverbs_dev->devnum);
+
+	cdev_init(&uverbs_dev->cdev,
+		  device->mmap ? &uverbs_mmap_fops : &uverbs_fops);
 	uverbs_dev->cdev.owner = THIS_MODULE;
-	uverbs_dev->cdev.ops = device->mmap ? &uverbs_mmap_fops : &uverbs_fops;
-	cdev_set_parent(&uverbs_dev->cdev, &uverbs_dev->kobj);
-	kobject_set_name(&uverbs_dev->cdev.kobj, "uverbs%d", uverbs_dev->devnum);
-	if (cdev_add(&uverbs_dev->cdev, base, 1))
-		goto err_cdev;
-
-	uverbs_dev->dev = device_create(uverbs_class, device->dev.parent,
-					uverbs_dev->cdev.dev, uverbs_dev,
-					"uverbs%d", uverbs_dev->devnum);
-	if (IS_ERR(uverbs_dev->dev))
-		goto err_cdev;
-
-	if (device_create_file(uverbs_dev->dev, &dev_attr_ibdev))
-		goto err_class;
-	if (device_create_file(uverbs_dev->dev, &dev_attr_abi_version))
-		goto err_class;
 
-	ib_set_client_data(device, &uverbs_client, uverbs_dev);
+	ret = cdev_device_add(&uverbs_dev->cdev, &uverbs_dev->dev);
+	if (ret)
+		goto err_uapi;
 
+	ib_set_client_data(device, &uverbs_client, uverbs_dev);
 	return;
 
-err_class:
-	device_destroy(uverbs_class, uverbs_dev->cdev.dev);
-err_cdev:
-	cdev_del(&uverbs_dev->cdev);
 err_uapi:
-	clear_bit(devnum, dev_map);
+	ida_free(&uverbs_ida, devnum);
 err:
 	if (atomic_dec_and_test(&uverbs_dev->refcount))
 		ib_uverbs_comp_dev(uverbs_dev);
 	wait_for_completion(&uverbs_dev->comp);
-	kobject_put(&uverbs_dev->kobj);
+	put_device(&uverbs_dev->dev);
 	return;
 }
 
@@ -1107,8 +1322,7 @@ static void ib_uverbs_free_hw_resources(struct ib_uverbs_device *uverbs_dev,
 	while (!list_empty(&uverbs_dev->uverbs_file_list)) {
 		file = list_first_entry(&uverbs_dev->uverbs_file_list,
 					struct ib_uverbs_file, list);
-		file->is_closed = 1;
-		list_del(&file->list);
+		list_del_init(&file->list);
 		kref_get(&file->ref);
 
 		/* We must release the mutex before going ahead and calling
@@ -1156,10 +1370,8 @@ static void ib_uverbs_remove_one(struct ib_device *device, void *client_data)
 	if (!uverbs_dev)
 		return;
 
-	dev_set_drvdata(uverbs_dev->dev, NULL);
-	device_destroy(uverbs_class, uverbs_dev->cdev.dev);
-	cdev_del(&uverbs_dev->cdev);
-	clear_bit(uverbs_dev->devnum, dev_map);
+	cdev_device_del(&uverbs_dev->cdev, &uverbs_dev->dev);
+	ida_free(&uverbs_ida, uverbs_dev->devnum);
 
 	if (device->disassociate_ucontext) {
 		/* We disassociate HW resources and immediately return.
@@ -1182,7 +1394,7 @@ static void ib_uverbs_remove_one(struct ib_device *device, void *client_data)
 	if (wait_clients)
 		wait_for_completion(&uverbs_dev->comp);
 
-	kobject_put(&uverbs_dev->kobj);
+	put_device(&uverbs_dev->dev);
 }
 
 static char *uverbs_devnode(struct device *dev, umode_t *mode)
diff --git a/drivers/infiniband/core/uverbs_std_types_flow_action.c b/drivers/infiniband/core/uverbs_std_types_flow_action.c
index d8cfafe23bd9..cb9486ad5c67 100644
--- a/drivers/infiniband/core/uverbs_std_types_flow_action.c
+++ b/drivers/infiniband/core/uverbs_std_types_flow_action.c
@@ -326,11 +326,8 @@ static int UVERBS_HANDLER(UVERBS_METHOD_FLOW_ACTION_ESP_CREATE)(
 	if (IS_ERR(action))
 		return PTR_ERR(action);
 
-	atomic_set(&action->usecnt, 0);
-	action->device = ib_dev;
-	action->type = IB_FLOW_ACTION_ESP;
-	action->uobject = uobj;
-	uobj->object = action;
+	uverbs_flow_action_fill_action(action, uobj, ib_dev,
+				       IB_FLOW_ACTION_ESP);
 
 	return 0;
 }
diff --git a/drivers/infiniband/core/uverbs_uapi.c b/drivers/infiniband/core/uverbs_uapi.c
index be854628a7c6..86f3fc5e04b4 100644
--- a/drivers/infiniband/core/uverbs_uapi.c
+++ b/drivers/infiniband/core/uverbs_uapi.c
@@ -73,6 +73,18 @@ static int uapi_merge_method(struct uverbs_api *uapi,
 		if (attr->attr.type == UVERBS_ATTR_TYPE_ENUM_IN)
 			method_elm->driver_method |= is_driver;
 
+		/*
+		 * Like other uobject based things we only support a single
+		 * uobject being NEW'd or DESTROY'd
+		 */
+		if (attr->attr.type == UVERBS_ATTR_TYPE_IDRS_ARRAY) {
+			u8 access = attr->attr.u2.objs_arr.access;
+
+			if (WARN_ON(access == UVERBS_ACCESS_NEW ||
+				    access == UVERBS_ACCESS_DESTROY))
+				return -EINVAL;
+		}
+
 		attr_slot =
 			uapi_add_elm(uapi, method_key | uapi_key_attr(attr->id),
 				     sizeof(*attr_slot));
diff --git a/drivers/infiniband/core/verbs.c b/drivers/infiniband/core/verbs.c
index 8ec7418e99f0..178899e3ce73 100644
--- a/drivers/infiniband/core/verbs.c
+++ b/drivers/infiniband/core/verbs.c
@@ -264,7 +264,7 @@ struct ib_pd *__ib_alloc_pd(struct ib_device *device, unsigned int flags,
 	}
 
 	pd->res.type = RDMA_RESTRACK_PD;
-	pd->res.kern_name = caller;
+	rdma_restrack_set_task(&pd->res, caller);
 	rdma_restrack_add(&pd->res);
 
 	if (mr_access_flags) {
@@ -710,7 +710,7 @@ static int ib_resolve_unicast_gid_dmac(struct ib_device *device,
 
 	ret = rdma_addr_find_l2_eth_by_grh(&sgid_attr->gid, &grh->dgid,
 					   ah_attr->roce.dmac,
-					   sgid_attr->ndev, &hop_limit);
+					   sgid_attr, &hop_limit);
 
 	grh->hop_limit = hop_limit;
 	return ret;
@@ -1509,8 +1509,7 @@ static const struct {
 };
 
 bool ib_modify_qp_is_ok(enum ib_qp_state cur_state, enum ib_qp_state next_state,
-			enum ib_qp_type type, enum ib_qp_attr_mask mask,
-			enum rdma_link_layer ll)
+			enum ib_qp_type type, enum ib_qp_attr_mask mask)
 {
 	enum ib_qp_attr_mask req_param, opt_param;
 
@@ -1629,14 +1628,16 @@ static int _ib_modify_qp(struct ib_qp *qp, struct ib_qp_attr *attr,
 
 	if (rdma_ib_or_roce(qp->device, port)) {
 		if (attr_mask & IB_QP_RQ_PSN && attr->rq_psn & ~0xffffff) {
-			pr_warn("%s: %s rq_psn overflow, masking to 24 bits\n",
-				__func__, qp->device->name);
+			dev_warn(&qp->device->dev,
+				 "%s rq_psn overflow, masking to 24 bits\n",
+				 __func__);
 			attr->rq_psn &= 0xffffff;
 		}
 
 		if (attr_mask & IB_QP_SQ_PSN && attr->sq_psn & ~0xffffff) {
-			pr_warn("%s: %s sq_psn overflow, masking to 24 bits\n",
-				__func__, qp->device->name);
+			dev_warn(&qp->device->dev,
+				 " %s sq_psn overflow, masking to 24 bits\n",
+				 __func__);
 			attr->sq_psn &= 0xffffff;
 		}
 	}
@@ -1888,7 +1889,7 @@ struct ib_cq *__ib_create_cq(struct ib_device *device,
 		cq->cq_context    = cq_context;
 		atomic_set(&cq->usecnt, 0);
 		cq->res.type = RDMA_RESTRACK_CQ;
-		cq->res.kern_name = caller;
+		rdma_restrack_set_task(&cq->res, caller);
 		rdma_restrack_add(&cq->res);
 	}