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

Pull rdma updates from Jason Gunthorpe:
 "Doug and I are at a conference next week so if another PR is sent I
  expect it to only be bug fixes. Parav noted yesterday that there are
  some fringe case behavior changes in his work that he would like to
  fix, and I see that Intel has a number of rc looking patches for HFI1
  they posted yesterday.

  Parav is again the biggest contributor by patch count with his ongoing
  work to enable container support in the RDMA stack, followed by Leon
  doing syzkaller inspired cleanups, though most of the actual fixing
  went to RC.

  There is one uncomfortable series here fixing the user ABI to actually
  work as intended in 32 bit mode. There are lots of notes in the commit
  messages, but the basic summary is we don't think there is an actual
  32 bit kernel user of drivers/infiniband for several good reasons.

  However we are seeing people want to use a 32 bit user space with 64
  bit kernel, which didn't completely work today. So in fixing it we
  required a 32 bit rxe user to upgrade their userspace. rxe users are
  still already quite rare and we think a 32 bit one is non-existing.

   - Fix RDMA uapi headers to actually compile in userspace and be more
     complete

   - Three shared with netdev pull requests from Mellanox:

      * 7 patches, mostly to net with 1 IB related one at the back).
        This series addresses an IRQ performance issue (patch 1),
        cleanups related to the fix for the IRQ performance problem
        (patches 2-6), and then extends the fragmented completion queue
        support that already exists in the net side of the driver to the
        ib side of the driver (patch 7).

      * Mostly IB, with 5 patches to net that are needed to support the
        remaining 10 patches to the IB subsystem. This series extends
        the current 'representor' framework when the mlx5 driver is in
        switchdev mode from being a netdev only construct to being a
        netdev/IB dev construct. The IB dev is limited to raw Eth queue
        pairs only, but by having an IB dev of this type attached to the
        representor for a switchdev port, it enables DPDK to work on the
        switchdev device.

      * All net related, but needed as infrastructure for the rdma
        driver

   - Updates for the hns, i40iw, bnxt_re, cxgb3, cxgb4, hns drivers

   - SRP performance updates

   - IB uverbs write path cleanup patch series from Leon

   - Add RDMA_CM support to ib_srpt. This is disabled by default. Users
     need to set the port for ib_srpt to listen on in configfs in order
     for it to be enabled
     (/sys/kernel/config/target/srpt/discovery_auth/rdma_cm_port)

   - TSO and Scatter FCS support in mlx4

   - Refactor of modify_qp routine to resolve problems seen while
     working on new code that is forthcoming

   - More refactoring and updates of RDMA CM for containers support from
     Parav

   - mlx5 'fine grained packet pacing', 'ipsec offload' and 'device
     memory' user API features

   - Infrastructure updates for the new IOCTL interface, based on
     increased usage

   - ABI compatibility bug fixes to fully support 32 bit userspace on 64
     bit kernel as was originally intended. See the commit messages for
     extensive details

   - Syzkaller bugs and code cleanups motivated by them"

* tag 'for-linus-unmerged' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma: (199 commits)
  IB/rxe: Fix for oops in rxe_register_device on ppc64le arch
  IB/mlx5: Device memory mr registration support
  net/mlx5: Mkey creation command adjustments
  IB/mlx5: Device memory support in mlx5_ib
  net/mlx5: Query device memory capabilities
  IB/uverbs: Add device memory registration ioctl support
  IB/uverbs: Add alloc/free dm uverbs ioctl support
  IB/uverbs: Add device memory capabilities reporting
  IB/uverbs: Expose device memory capabilities to user
  RDMA/qedr: Fix wmb usage in qedr
  IB/rxe: Removed GID add/del dummy routines
  RDMA/qedr: Zero stack memory before copying to user space
  IB/mlx5: Add ability to hash by IPSEC_SPI when creating a TIR
  IB/mlx5: Add information for querying IPsec capabilities
  IB/mlx5: Add IPsec support for egress and ingress
  {net,IB}/mlx5: Add ipsec helper
  IB/mlx5: Add modify_flow_action_esp verb
  IB/mlx5: Add implementation for create and destroy action_xfrm
  IB/uverbs: Introduce ESP steering match filter
  IB/uverbs: Add modify ESP flow_action
  ...

.mailmap

@@ -102,6 +102,8 @@ Koushik <raghavendra.koushik@neterion.com>
 Krzysztof Kozlowski <krzk@kernel.org> <k.kozlowski@samsung.com>
 Krzysztof Kozlowski <krzk@kernel.org> <k.kozlowski.k@gmail.com>
 Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
+Leon Romanovsky <leon@kernel.org> <leon@leon.nu>
+Leon Romanovsky <leon@kernel.org> <leonro@mellanox.com>
 Leonid I Ananiev <leonid.i.ananiev@intel.com>
 Linas Vepstas <linas@austin.ibm.com>
 Linus Lüssing <linus.luessing@c0d3.blue> <linus.luessing@web.de>

MAINTAINERS

@@ -7214,6 +7214,7 @@ M:	Shiraz Saleem <shiraz.saleem@intel.com>
 L:	linux-rdma@vger.kernel.org
 S:	Supported
 F:	drivers/infiniband/hw/i40iw/
+F:	include/uapi/rdma/i40iw-abi.h
 
 INTEL SHA MULTIBUFFER DRIVER
 M:	Megha Dey <megha.dey@linux.intel.com>

drivers/infiniband/Kconfig

@@ -35,14 +35,13 @@ config INFINIBAND_USER_ACCESS
 	  libibverbs, libibcm and a hardware driver library from
 	  rdma-core <https://github.com/linux-rdma/rdma-core>.
 
-config INFINIBAND_EXP_USER_ACCESS
-	bool "Enable the full uverbs ioctl interface (EXPERIMENTAL)"
+config INFINIBAND_EXP_LEGACY_VERBS_NEW_UAPI
+	bool "Allow experimental legacy verbs in new ioctl uAPI  (EXPERIMENTAL)"
 	depends on INFINIBAND_USER_ACCESS
 	---help---
-	  IOCTL based ABI support for Infiniband. This allows userspace
-	  to invoke the experimental IOCTL based ABI.
-	  These commands are parsed via per-device parsing tree and
-	  enables per-device features.
+	  IOCTL based uAPI support for Infiniband is enabled by default for
+	  new verbs only. This allows userspace to invoke the IOCTL based uAPI
+	  for current legacy verbs too.
 
 config INFINIBAND_USER_MEM
 	bool

drivers/infiniband/core/Makefile

@@ -34,4 +34,6 @@ ib_ucm-y :=			ucm.o
 
 ib_uverbs-y :=			uverbs_main.o uverbs_cmd.o uverbs_marshall.o \
 				rdma_core.o uverbs_std_types.o uverbs_ioctl.o \
-				uverbs_ioctl_merge.o
+				uverbs_ioctl_merge.o uverbs_std_types_cq.o \
+				uverbs_std_types_flow_action.o uverbs_std_types_dm.o \
+				uverbs_std_types_mr.o

drivers/infiniband/core/addr.c

@@ -329,7 +329,8 @@ static void queue_req(struct addr_req *req)
 	mutex_unlock(&lock);
 }
 
-static int ib_nl_fetch_ha(struct dst_entry *dst, struct rdma_dev_addr *dev_addr,
+static int ib_nl_fetch_ha(const struct dst_entry *dst,
+			  struct rdma_dev_addr *dev_addr,
 			  const void *daddr, u32 seq, u16 family)
 {
 	if (rdma_nl_chk_listeners(RDMA_NL_GROUP_LS))
@@ -340,7 +341,8 @@ static int ib_nl_fetch_ha(struct dst_entry *dst, struct rdma_dev_addr *dev_addr,
 	return ib_nl_ip_send_msg(dev_addr, daddr, seq, family);
 }
 
-static int dst_fetch_ha(struct dst_entry *dst, struct rdma_dev_addr *dev_addr,
+static int dst_fetch_ha(const struct dst_entry *dst,
+			struct rdma_dev_addr *dev_addr,
 			const void *daddr)
 {
 	struct neighbour *n;
@@ -364,7 +366,7 @@ static int dst_fetch_ha(struct dst_entry *dst, struct rdma_dev_addr *dev_addr,
 	return ret;
 }
 
-static bool has_gateway(struct dst_entry *dst, sa_family_t family)
+static bool has_gateway(const struct dst_entry *dst, sa_family_t family)
 {
 	struct rtable *rt;
 	struct rt6_info *rt6;
@@ -378,7 +380,7 @@ static bool has_gateway(struct dst_entry *dst, sa_family_t family)
 	return rt6->rt6i_flags & RTF_GATEWAY;
 }
 
-static int fetch_ha(struct dst_entry *dst, struct rdma_dev_addr *dev_addr,
+static int fetch_ha(const struct dst_entry *dst, struct rdma_dev_addr *dev_addr,
 		    const struct sockaddr *dst_in, u32 seq)
 {
 	const struct sockaddr_in *dst_in4 =
@@ -482,7 +484,7 @@ static int addr6_resolve(struct sockaddr_in6 *src_in,
 }
 #endif
 
-static int addr_resolve_neigh(struct dst_entry *dst,
+static int addr_resolve_neigh(const struct dst_entry *dst,
 			      const struct sockaddr *dst_in,
 			      struct rdma_dev_addr *addr,
 			      u32 seq)
@@ -736,7 +738,6 @@ int rdma_resolve_ip_route(struct sockaddr *src_addr,
 
 	return addr_resolve(src_in, dst_addr, addr, false, 0);
 }
-EXPORT_SYMBOL(rdma_resolve_ip_route);
 
 void rdma_addr_cancel(struct rdma_dev_addr *addr)
 {

drivers/infiniband/core/cache.c

@@ -59,8 +59,6 @@ struct ib_update_work {
 union ib_gid zgid;
 EXPORT_SYMBOL(zgid);
 
-static const struct ib_gid_attr zattr;
-
 enum gid_attr_find_mask {
 	GID_ATTR_FIND_MASK_GID          = 1UL << 0,
 	GID_ATTR_FIND_MASK_NETDEV	= 1UL << 1,
@@ -73,15 +71,6 @@ enum gid_table_entry_props {
 	GID_TABLE_ENTRY_DEFAULT		= 1UL << 1,
 };
 
-enum gid_table_write_action {
-	GID_TABLE_WRITE_ACTION_ADD,
-	GID_TABLE_WRITE_ACTION_DEL,
-	/* MODIFY only updates the GID table. Currently only used by
-	 * ib_cache_update.
-	 */
-	GID_TABLE_WRITE_ACTION_MODIFY
-};
-
 struct ib_gid_table_entry {
 	unsigned long	    props;
 	union ib_gid        gid;
@@ -100,31 +89,26 @@ struct ib_gid_table {
 	 * (a) Find the GID
 	 * (b) Delete it.
 	 *
-	 * Add/delete should be carried out atomically.
-	 * This is done by locking this mutex from multiple
-	 * writers. We don't need this lock for IB, as the MAD
-	 * layer replaces all entries. All data_vec entries
-	 * are locked by this lock.
 	 **/
-	struct mutex         lock;
-	/* This lock protects the table entries from being
-	 * read and written simultaneously.
+	/* Any writer to data_vec must hold this lock and the write side of
+	 * rwlock. readers must hold only rwlock. All writers must be in a
+	 * sleepable context.
 	 */
+	struct mutex         lock;
+	/* rwlock protects data_vec[ix]->props. */
 	rwlock_t	     rwlock;
 	struct ib_gid_table_entry *data_vec;
 };
 
 static void dispatch_gid_change_event(struct ib_device *ib_dev, u8 port)
 {
-	if (rdma_cap_roce_gid_table(ib_dev, port)) {
-		struct ib_event event;
+	struct ib_event event;
 
-		event.device		= ib_dev;
-		event.element.port_num	= port;
-		event.event		= IB_EVENT_GID_CHANGE;
+	event.device		= ib_dev;
+	event.element.port_num	= port;
+	event.event		= IB_EVENT_GID_CHANGE;
 
-		ib_dispatch_event(&event);
-	}
+	ib_dispatch_event(&event);
 }
 
 static const char * const gid_type_str[] = {
@@ -165,94 +149,127 @@ int ib_cache_gid_parse_type_str(const char *buf)
 }
 EXPORT_SYMBOL(ib_cache_gid_parse_type_str);
 
-/* This function expects that rwlock will be write locked in all
- * scenarios and that lock will be locked in sleep-able (RoCE)
- * scenarios.
- */
-static int write_gid(struct ib_device *ib_dev, u8 port,
-		     struct ib_gid_table *table, int ix,
-		     const union ib_gid *gid,
-		     const struct ib_gid_attr *attr,
-		     enum gid_table_write_action action,
-		     bool  default_gid)
-	__releases(&table->rwlock) __acquires(&table->rwlock)
+static void del_roce_gid(struct ib_device *device, u8 port_num,
+			 struct ib_gid_table *table, int ix)
 {
-	int ret = 0;
-	struct net_device *old_net_dev;
-	enum ib_gid_type old_gid_type;
+	pr_debug("%s device=%s port=%d index=%d gid %pI6\n", __func__,
+		 device->name, port_num, ix,
+		 table->data_vec[ix].gid.raw);
+
+	if (rdma_cap_roce_gid_table(device, port_num))
+		device->del_gid(&table->data_vec[ix].attr,
+				&table->data_vec[ix].context);
+	dev_put(table->data_vec[ix].attr.ndev);
+}
 
-	/* in rdma_cap_roce_gid_table, this funciton should be protected by a
-	 * sleep-able lock.
-	 */
+static int add_roce_gid(struct ib_gid_table *table,
+			const union ib_gid *gid,
+			const struct ib_gid_attr *attr)
+{
+	struct ib_gid_table_entry *entry;
+	int ix = attr->index;
+	int ret = 0;
 
-	if (rdma_cap_roce_gid_table(ib_dev, port)) {
-		table->data_vec[ix].props |= GID_TABLE_ENTRY_INVALID;
-		write_unlock_irq(&table->rwlock);
-		/* GID_TABLE_WRITE_ACTION_MODIFY currently isn't supported by
-		 * RoCE providers and thus only updates the cache.
-		 */
-		if (action == GID_TABLE_WRITE_ACTION_ADD)
-			ret = ib_dev->add_gid(ib_dev, port, ix, gid, attr,
-					      &table->data_vec[ix].context);
-		else if (action == GID_TABLE_WRITE_ACTION_DEL)
-			ret = ib_dev->del_gid(ib_dev, port, ix,
-					      &table->data_vec[ix].context);
-		write_lock_irq(&table->rwlock);
+	if (!attr->ndev) {
+		pr_err("%s NULL netdev device=%s port=%d index=%d\n",
+		       __func__, attr->device->name, attr->port_num,
+		       attr->index);
+		return -EINVAL;
 	}
 
-	old_net_dev = table->data_vec[ix].attr.ndev;
-	old_gid_type = table->data_vec[ix].attr.gid_type;
-	if (old_net_dev && old_net_dev != attr->ndev)
-		dev_put(old_net_dev);
-	/* if modify_gid failed, just delete the old gid */
-	if (ret || action == GID_TABLE_WRITE_ACTION_DEL) {
-		gid = &zgid;
-		attr = &zattr;
-		table->data_vec[ix].context = NULL;
+	entry = &table->data_vec[ix];
+	if ((entry->props & GID_TABLE_ENTRY_INVALID) == 0) {
+		WARN(1, "GID table corruption device=%s port=%d index=%d\n",
+		     attr->device->name, attr->port_num,
+		     attr->index);
+		return -EINVAL;
 	}
 
-	memcpy(&table->data_vec[ix].gid, gid, sizeof(*gid));
-	memcpy(&table->data_vec[ix].attr, attr, sizeof(*attr));
-	if (default_gid) {
-		table->data_vec[ix].props |= GID_TABLE_ENTRY_DEFAULT;
-		if (action == GID_TABLE_WRITE_ACTION_DEL)
-			table->data_vec[ix].attr.gid_type = old_gid_type;
+	if (rdma_cap_roce_gid_table(attr->device, attr->port_num)) {
+		ret = attr->device->add_gid(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);
+			goto add_err;
+		}
 	}
-	if (table->data_vec[ix].attr.ndev &&
-	    table->data_vec[ix].attr.ndev != old_net_dev)
-		dev_hold(table->data_vec[ix].attr.ndev);
-
-	table->data_vec[ix].props &= ~GID_TABLE_ENTRY_INVALID;
+	dev_hold(attr->ndev);
 
+add_err:
+	if (!ret)
+		pr_debug("%s device=%s port=%d index=%d gid %pI6\n", __func__,
+			 attr->device->name, attr->port_num, ix, gid->raw);
 	return ret;
 }
 
-static int add_gid(struct ib_device *ib_dev, u8 port,
-		   struct ib_gid_table *table, int ix,
-		   const union ib_gid *gid,
-		   const struct ib_gid_attr *attr,
-		   bool  default_gid) {
-	return write_gid(ib_dev, port, table, ix, gid, attr,
-			 GID_TABLE_WRITE_ACTION_ADD, default_gid);
-}
+/**
+ * add_modify_gid - Add or modify GID table entry
+ *
+ * @table:	GID table in which GID to be added or modified
+ * @gid:	GID content
+ * @attr:	Attributes of the GID
+ *
+ * Returns 0 on success or appropriate error code. It accepts zero
+ * GID addition for non RoCE ports for HCA's who report them as valid
+ * GID. However such zero GIDs are not added to the cache.
+ */
+static int add_modify_gid(struct ib_gid_table *table,
+			  const union ib_gid *gid,
+			  const struct ib_gid_attr *attr)
+{
+	int ret;
+
+	if (rdma_protocol_roce(attr->device, attr->port_num)) {
+		ret = add_roce_gid(table, gid, attr);
+		if (ret)
+			return ret;
+	} else {
+		/*
+		 * Some HCA's report multiple GID entries with only one
+		 * valid GID, but remaining as zero GID.
+		 * So ignore such behavior for IB link layer and don't
+		 * fail the call, but don't add such entry to GID cache.
+		 */
+		if (!memcmp(gid, &zgid, sizeof(*gid)))
+			return 0;
+	}
+
+	lockdep_assert_held(&table->lock);
+	memcpy(&table->data_vec[attr->index].gid, gid, sizeof(*gid));
+	memcpy(&table->data_vec[attr->index].attr, attr, sizeof(*attr));
 
-static int modify_gid(struct ib_device *ib_dev, u8 port,
-		      struct ib_gid_table *table, int ix,
-		      const union ib_gid *gid,
-		      const struct ib_gid_attr *attr,
-		      bool  default_gid) {
-	return write_gid(ib_dev, port, table, ix, gid, attr,
-			 GID_TABLE_WRITE_ACTION_MODIFY, default_gid);
+	write_lock_irq(&table->rwlock);
+	table->data_vec[attr->index].props &= ~GID_TABLE_ENTRY_INVALID;
+	write_unlock_irq(&table->rwlock);
+	return 0;
 }
 
-static int del_gid(struct ib_device *ib_dev, u8 port,
-		   struct ib_gid_table *table, int ix,
-		   bool  default_gid) {
-	return write_gid(ib_dev, port, table, ix, &zgid, &zattr,
-			 GID_TABLE_WRITE_ACTION_DEL, default_gid);
+/**
+ * del_gid - Delete GID table entry
+ *
+ * @ib_dev:	IB device whose GID entry to be deleted
+ * @port:	Port number of the IB device
+ * @table:	GID table of the IB device for a port
+ * @ix:		GID entry index to delete
+ *
+ */
+static void del_gid(struct ib_device *ib_dev, u8 port,
+		    struct ib_gid_table *table, int ix)
+{
+	lockdep_assert_held(&table->lock);
+	write_lock_irq(&table->rwlock);
+	table->data_vec[ix].props |= GID_TABLE_ENTRY_INVALID;
+	write_unlock_irq(&table->rwlock);
+
+	if (rdma_protocol_roce(ib_dev, port))
+		del_roce_gid(ib_dev, port, table, ix);
+	memcpy(&table->data_vec[ix].gid, &zgid, sizeof(zgid));
+	memset(&table->data_vec[ix].attr, 0, sizeof(table->data_vec[ix].attr));
+	table->data_vec[ix].context = NULL;
 }
 
-/* rwlock should be read locked */
+/* rwlock should be read locked, or lock should be held */
 static int find_gid(struct ib_gid_table *table, const union ib_gid *gid,
 		    const struct ib_gid_attr *val, bool default_gid,
 		    unsigned long mask, int *pempty)
@@ -268,15 +285,32 @@ static int find_gid(struct ib_gid_table *table, const union ib_gid *gid,
 
 		i++;
 
+		/* find_gid() is used during GID addition where it is expected
+		 * to return a free entry slot which is not duplicate.
+		 * Free entry slot is requested and returned if pempty is set,
+		 * so lookup free slot only if requested.
+		 */
+		if (pempty && empty < 0) {
+			if (data->props & GID_TABLE_ENTRY_INVALID) {
+				/* Found an invalid (free) entry; allocate it */
+				if (data->props & GID_TABLE_ENTRY_DEFAULT) {
+					if (default_gid)
+						empty = curr_index;
+				} else {
+					empty = curr_index;
+				}
+			}
+		}
+
+		/*
+		 * Additionally find_gid() is used to find valid entry during
+		 * lookup operation, where validity needs to be checked. So
+		 * find the empty entry first to continue to search for a free
+		 * slot and ignore its INVALID flag.
+		 */
 		if (data->props & GID_TABLE_ENTRY_INVALID)
 			continue;
 
-		if (empty < 0)
-			if (!memcmp(&data->gid, &zgid, sizeof(*gid)) &&
-			    !memcmp(attr, &zattr, sizeof(*attr)) &&
-			    !data->props)
-				empty = curr_index;
-
 		if (found >= 0)
 			continue;
 
@@ -312,20 +346,56 @@ static void make_default_gid(struct  net_device *dev, union ib_gid *gid)
 	addrconf_ifid_eui48(&gid->raw[8], dev);
 }
 
-int ib_cache_gid_add(struct ib_device *ib_dev, u8 port,
-		     union ib_gid *gid, struct ib_gid_attr *attr)
+static int __ib_cache_gid_add(struct ib_device *ib_dev, u8 port,
+			      union ib_gid *gid, struct ib_gid_attr *attr,
+			      unsigned long mask, bool default_gid)
 {
 	struct ib_gid_table *table;
-	int ix;
 	int ret = 0;
-	struct net_device *idev;
 	int empty;
+	int ix;
 
-	table = ib_dev->cache.ports[port - rdma_start_port(ib_dev)].gid;
-
+	/* Do not allow adding zero GID in support of
+	 * IB spec version 1.3 section 4.1.1 point (6) and
+	 * section 12.7.10 and section 12.7.20
+	 */
 	if (!memcmp(gid, &zgid, sizeof(*gid)))
 		return -EINVAL;
 
+	table = ib_dev->cache.ports[port - rdma_start_port(ib_dev)].gid;
+
+	mutex_lock(&table->lock);
+
+	ix = find_gid(table, gid, attr, default_gid, mask, &empty);
+	if (ix >= 0)
+		goto out_unlock;
+
+	if (empty < 0) {
+		ret = -ENOSPC;
+		goto out_unlock;
+	}
+	attr->device = ib_dev;
+	attr->index = empty;
+	attr->port_num = port;
+	ret = add_modify_gid(table, gid, attr);
+	if (!ret)
+		dispatch_gid_change_event(ib_dev, port);
+
+out_unlock:
+	mutex_unlock(&table->lock);
+	if (ret)
+		pr_warn("%s: unable to add gid %pI6 error=%d\n",
+			__func__, gid->raw, ret);
+	return ret;
+}
+
+int ib_cache_gid_add(struct ib_device *ib_dev, u8 port,
+		     union ib_gid *gid, struct ib_gid_attr *attr)
+{
+	struct net_device *idev;
+	unsigned long mask;
+	int ret;
+
 	if (ib_dev->get_netdev) {
 		idev = ib_dev->get_netdev(ib_dev, port);
 		if (idev && attr->ndev != idev) {
@@ -342,27 +412,11 @@ int ib_cache_gid_add(struct ib_device *ib_dev, u8 port,
 			dev_put(idev);
 	}
 
-	mutex_lock(&table->lock);
-	write_lock_irq(&table->rwlock);
-
-	ix = find_gid(table, gid, attr, false, GID_ATTR_FIND_MASK_GID |
-		      GID_ATTR_FIND_MASK_GID_TYPE |
-		      GID_ATTR_FIND_MASK_NETDEV, &empty);
-	if (ix >= 0)
-		goto out_unlock;
-
-	if (empty < 0) {
-		ret = -ENOSPC;
-		goto out_unlock;
-	}
-
-	ret = add_gid(ib_dev, port, table, empty, gid, attr, false);
-	if (!ret)
-		dispatch_gid_change_event(ib_dev, port);
+	mask = GID_ATTR_FIND_MASK_GID |
+	       GID_ATTR_FIND_MASK_GID_TYPE |
+	       GID_ATTR_FIND_MASK_NETDEV;
 
-out_unlock:
-	write_unlock_irq(&table->rwlock);
-	mutex_unlock(&table->lock);
+	ret = __ib_cache_gid_add(ib_dev, port, gid, attr, mask, false);
 	return ret;
 }
 
@@ -370,29 +424,32 @@ int ib_cache_gid_del(struct ib_device *ib_dev, u8 port,
 		     union ib_gid *gid, struct ib_gid_attr *attr)
 {
 	struct ib_gid_table *table;
+	int ret = 0;
 	int ix;
 
 	table = ib_dev->cache.ports[port - rdma_start_port(ib_dev)].gid;
 
 	mutex_lock(&table->lock);
-	write_lock_irq(&table->rwlock);
 
 	ix = find_gid(table, gid, attr, false,
 		      GID_ATTR_FIND_MASK_GID	  |
 		      GID_ATTR_FIND_MASK_GID_TYPE |
-		      GID_ATTR_FIND_MASK_NETDEV	  |
-		      GID_ATTR_FIND_MASK_DEFAULT,
+		      GID_ATTR_FIND_MASK_NETDEV,
 		      NULL);
-	if (ix < 0)
+	if (ix < 0) {
+		ret = -EINVAL;
 		goto out_unlock;
+	}
 
-	if (!del_gid(ib_dev, port, table, ix, false))
-		dispatch_gid_change_event(ib_dev, port);
+	del_gid(ib_dev, port, table, ix);
+	dispatch_gid_change_event(ib_dev, port);
 
 out_unlock:
-	write_unlock_irq(&table->rwlock);
 	mutex_unlock(&table->lock);
-	return 0;
+	if (ret)
+		pr_debug("%s: can't delete gid %pI6 error=%d\n",
+			 __func__, gid->raw, ret);
+	return ret;
 }
 
 int ib_cache_gid_del_all_netdev_gids(struct ib_device *ib_dev, u8 port,
@@ -405,16 +462,14 @@ int ib_cache_gid_del_all_netdev_gids(struct ib_device *ib_dev, u8 port,
 	table = ib_dev->cache.ports[port - rdma_start_port(ib_dev)].gid;
 
 	mutex_lock(&table->lock);
-	write_lock_irq(&table->rwlock);
 
-	for (ix = 0; ix < table->sz; ix++)
-		if (table->data_vec[ix].attr.ndev == ndev)
-			if (!del_gid(ib_dev, port, table, ix,
-				     !!(table->data_vec[ix].props &
-					GID_TABLE_ENTRY_DEFAULT)))
-				deleted = true;
+	for (ix = 0; ix < table->sz; ix++) {
+		if (table->data_vec[ix].attr.ndev == ndev) {
+			del_gid(ib_dev, port, table, ix);
+			deleted = true;
+		}
+	}
 
-	write_unlock_irq(&table->rwlock);
 	mutex_unlock(&table->lock);
 
 	if (deleted)
@@ -492,6 +547,19 @@ static int ib_cache_gid_find(struct ib_device *ib_dev,
 					mask, port, index);
 }
 
+/**
+ * ib_find_cached_gid_by_port - Returns the GID table index where a specified
+ * GID value occurs. It searches for the specified GID value in the local
+ * software cache.
+ * @device: The device to query.
+ * @gid: The GID value to search for.
+ * @gid_type: The GID type to search for.
+ * @port_num: The port number of the device where the GID value should be
+ *   searched.
+ * @ndev: In RoCE, the net device of the device. Null means ignore.
+ * @index: The index into the cached GID table where the GID was found. This
+ *   parameter may be NULL.
+ */
 int ib_find_cached_gid_by_port(struct ib_device *ib_dev,
 			       const union ib_gid *gid,
 			       enum ib_gid_type gid_type,
@@ -528,7 +596,7 @@ int ib_find_cached_gid_by_port(struct ib_device *ib_dev,
 EXPORT_SYMBOL(ib_find_cached_gid_by_port);
 
 /**
- * ib_find_gid_by_filter - Returns the GID table index where a specified
+ * ib_cache_gid_find_by_filter - Returns the GID table index where a specified
  * GID value occurs
  * @device: The device to query.
  * @gid: The GID value to search for.
@@ -539,7 +607,7 @@ EXPORT_SYMBOL(ib_find_cached_gid_by_port);
  *   otherwise, we continue searching the GID table. It's guaranteed that
  *   while filter is executed, ndev field is valid and the structure won't
  *   change. filter is executed in an atomic context. filter must not be NULL.
- * @index: The index into the cached GID table where the GID was found.  This
+ * @index: The index into the cached GID table where the GID was found. This
  *   parameter may be NULL.
  *
  * ib_cache_gid_find_by_filter() searches for the specified GID value
@@ -598,6 +666,7 @@ static struct ib_gid_table *alloc_gid_table(int sz)
 {
 	struct ib_gid_table *table =
 		kzalloc(sizeof(struct ib_gid_table), GFP_KERNEL);
+	int i;
 
 	if (!table)
 		return NULL;
@@ -611,6 +680,11 @@ static struct ib_gid_table *alloc_gid_table(int sz)
 	table->sz = sz;
 	rwlock_init(&table->rwlock);
 
+	/* Mark all entries as invalid so that allocator can allocate
+	 * one of the invalid (free) entry.
+	 */
+	for (i = 0; i < sz; i++)
+		table->data_vec[i].props |= GID_TABLE_ENTRY_INVALID;
 	return table;
 
 err_free_table:
@@ -635,16 +709,15 @@ static void cleanup_gid_table_port(struct ib_device *ib_dev, u8 port,
 	if (!table)
 		return;
 
-	write_lock_irq(&table->rwlock);
+	mutex_lock(&table->lock);
 	for (i = 0; i < table->sz; ++i) {
 		if (memcmp(&table->data_vec[i].gid, &zgid,
-			   sizeof(table->data_vec[i].gid)))
-			if (!del_gid(ib_dev, port, table, i,
-				     table->data_vec[i].props &
-				     GID_ATTR_FIND_MASK_DEFAULT))
-				deleted = true;
+			   sizeof(table->data_vec[i].gid))) {
+			del_gid(ib_dev, port, table, i);
+			deleted = true;
+		}
 	}
-	write_unlock_irq(&table->rwlock);
+	mutex_unlock(&table->lock);
 
 	if (deleted)
 		dispatch_gid_change_event(ib_dev, port);
@@ -657,9 +730,9 @@ void ib_cache_gid_set_default_gid(struct ib_device *ib_dev, u8 port,
 {
 	union ib_gid gid;
 	struct ib_gid_attr gid_attr;
-	struct ib_gid_attr zattr_type = zattr;
 	struct ib_gid_table *table;
 	unsigned int gid_type;
+	unsigned long mask;
 
 	table = ib_dev->cache.ports[port - rdma_start_port(ib_dev)].gid;
 
@@ -668,60 +741,19 @@ void ib_cache_gid_set_default_gid(struct ib_device *ib_dev, u8 port,
 	gid_attr.ndev = ndev;
 
 	for (gid_type = 0; gid_type < IB_GID_TYPE_SIZE; ++gid_type) {
-		int ix;
-		union ib_gid current_gid;
-		struct ib_gid_attr current_gid_attr = {};
-
 		if (1UL << gid_type & ~gid_type_mask)
 			continue;
 
 		gid_attr.gid_type = gid_type;
 
-		mutex_lock(&table->lock);
-		write_lock_irq(&table->rwlock);
-		ix = find_gid(table, NULL, &gid_attr, true,
-			      GID_ATTR_FIND_MASK_GID_TYPE |
-			      GID_ATTR_FIND_MASK_DEFAULT,
-			      NULL);
-
-		/* Coudn't find default GID location */
-		if (WARN_ON(ix < 0))
-			goto release;
-
-		zattr_type.gid_type = gid_type;
-
-		if (!__ib_cache_gid_get(ib_dev, port, ix,
-					&current_gid, &current_gid_attr) &&
-		    mode == IB_CACHE_GID_DEFAULT_MODE_SET &&
-		    !memcmp(&gid, &current_gid, sizeof(gid)) &&
-		    !memcmp(&gid_attr, &current_gid_attr, sizeof(gid_attr)))
-			goto release;
-
-		if (memcmp(&current_gid, &zgid, sizeof(current_gid)) ||
-		    memcmp(&current_gid_attr, &zattr_type,
-			   sizeof(current_gid_attr))) {
-			if (del_gid(ib_dev, port, table, ix, true)) {
-				pr_warn("ib_cache_gid: can't delete index %d for default gid %pI6\n",
-					ix, gid.raw);
-				goto release;
-			} else {
-				dispatch_gid_change_event(ib_dev, port);
-			}
-		}
-
 		if (mode == IB_CACHE_GID_DEFAULT_MODE_SET) {
-			if (add_gid(ib_dev, port, table, ix, &gid, &gid_attr, true))
-				pr_warn("ib_cache_gid: unable to add default gid %pI6\n",
-					gid.raw);
-			else
-				dispatch_gid_change_event(ib_dev, port);
+			mask = GID_ATTR_FIND_MASK_GID_TYPE |
+			       GID_ATTR_FIND_MASK_DEFAULT;
+			__ib_cache_gid_add(ib_dev, port, &gid,
+					   &gid_attr, mask, true);
+		} else if (mode == IB_CACHE_GID_DEFAULT_MODE_DELETE) {
+			ib_cache_gid_del(ib_dev, port, &gid, &gid_attr);
 		}
-
-release:
-		if (current_gid_attr.ndev)
-			dev_put(current_gid_attr.ndev);
-		write_unlock_irq(&table->rwlock);
-		mutex_unlock(&table->lock);
 	}
 }
 
@@ -848,6 +880,20 @@ int ib_get_cached_gid(struct ib_device *device,
 }
 EXPORT_SYMBOL(ib_get_cached_gid);
 
+/**
+ * ib_find_cached_gid - Returns the port number and GID table index where
+ *   a specified GID value occurs.
+ * @device: The device to query.
+ * @gid: The GID value to search for.
+ * @gid_type: The GID type to search for.
+ * @ndev: In RoCE, the net device of the device. NULL means ignore.
+ * @port_num: The port number of the device where the GID value was found.
+ * @index: The index into the cached GID table where the GID was found.  This
+ *   parameter may be NULL.
+ *
+ * ib_find_cached_gid() searches for the specified GID value in
+ * the local software cache.
+ */
 int ib_find_cached_gid(struct ib_device *device,
 		       const union ib_gid *gid,
 		       enum ib_gid_type gid_type,
@@ -868,7 +914,7 @@ int ib_find_gid_by_filter(struct ib_device *device,
 			  void *context, u16 *index)
 {
 	/* Only RoCE GID table supports filter function */
-	if (!rdma_cap_roce_gid_table(device, port_num) && filter)
+	if (!rdma_protocol_roce(device, port_num) && filter)
 		return -EPROTONOSUPPORT;
 
 	return ib_cache_gid_find_by_filter(device, gid,
@@ -910,8 +956,7 @@ int ib_get_cached_subnet_prefix(struct ib_device *device,
 	unsigned long flags;
 	int p;
 
-	if (port_num < rdma_start_port(device) ||
-	    port_num > rdma_end_port(device))
+	if (!rdma_is_port_valid(device, port_num))
 		return -EINVAL;
 
 	p = port_num - rdma_start_port(device);
@@ -1021,7 +1066,7 @@ int ib_get_cached_port_state(struct ib_device   *device,
 	unsigned long flags;
 	int ret = 0;
 
-	if (port_num < rdma_start_port(device) || port_num > rdma_end_port(device))
+	if (!rdma_is_port_valid(device, port_num))
 		return -EINVAL;
 
 	read_lock_irqsave(&device->cache.lock, flags);
@@ -1033,21 +1078,46 @@ int ib_get_cached_port_state(struct ib_device   *device,
 }
 EXPORT_SYMBOL(ib_get_cached_port_state);
 
+static int config_non_roce_gid_cache(struct ib_device *device,
+				     u8 port, int gid_tbl_len)
+{
+	struct ib_gid_attr gid_attr = {};
+	struct ib_gid_table *table;
+	union ib_gid gid;
+	int ret = 0;
+	int i;
+
+	gid_attr.device = device;
+	gid_attr.port_num = port;
+	table = device->cache.ports[port - rdma_start_port(device)].gid;
+
+	mutex_lock(&table->lock);
+	for (i = 0; i < gid_tbl_len; ++i) {
+		if (!device->query_gid)
+			continue;
+		ret = device->query_gid(device, port, i, &gid);
+		if (ret) {
+			pr_warn("query_gid failed (%d) for %s (index %d)\n",
+				ret, device->name, i);
+			goto err;
+		}
+		gid_attr.index = i;
+		add_modify_gid(table, &gid, &gid_attr);
+	}
+err:
+	mutex_unlock(&table->lock);
+	return ret;
+}
+
 static void ib_cache_update(struct ib_device *device,
 			    u8                port,
 			    bool	      enforce_security)
 {
 	struct ib_port_attr       *tprops = NULL;
 	struct ib_pkey_cache      *pkey_cache = NULL, *old_pkey_cache;
-	struct ib_gid_cache {
-		int             table_len;
-		union ib_gid    table[0];
-	}			  *gid_cache = NULL;
 	int                        i;
 	int                        ret;
 	struct ib_gid_table	  *table;
-	bool			   use_roce_gid_table =
-					rdma_cap_roce_gid_table(device, port);
 
 	if (!rdma_is_port_valid(device, port))
 		return;
@@ -1065,6 +1135,13 @@ static void ib_cache_update(struct ib_device *device,
 		goto err;
 	}
 
+	if (!rdma_protocol_roce(device, port)) {
+		ret = config_non_roce_gid_cache(device, port,
+						tprops->gid_tbl_len);
+		if (ret)
+			goto err;
+	}
+
 	pkey_cache = kmalloc(sizeof *pkey_cache + tprops->pkey_tbl_len *
 			     sizeof *pkey_cache->table, GFP_KERNEL);
 	if (!pkey_cache)
@@ -1072,15 +1149,6 @@ static void ib_cache_update(struct ib_device *device,
 
 	pkey_cache->table_len = tprops->pkey_tbl_len;
 
-	if (!use_roce_gid_table) {
-		gid_cache = kmalloc(sizeof(*gid_cache) + tprops->gid_tbl_len *
-			    sizeof(*gid_cache->table), GFP_KERNEL);
-		if (!gid_cache)
-			goto err;
-
-		gid_cache->table_len = tprops->gid_tbl_len;
-	}
-
 	for (i = 0; i < pkey_cache->table_len; ++i) {
 		ret = ib_query_pkey(device, port, i, pkey_cache->table + i);
 		if (ret) {
@@ -1090,33 +1158,12 @@ static void ib_cache_update(struct ib_device *device,
 		}
 	}
 
-	if (!use_roce_gid_table) {
-		for (i = 0;  i < gid_cache->table_len; ++i) {
-			ret = ib_query_gid(device, port, i,
-					   gid_cache->table + i, NULL);
-			if (ret) {
-				pr_warn("ib_query_gid failed (%d) for %s (index %d)\n",
-					ret, device->name, i);
-				goto err;
-			}
-		}
-	}
-
 	write_lock_irq(&device->cache.lock);
 
 	old_pkey_cache = device->cache.ports[port -
 		rdma_start_port(device)].pkey;
 
 	device->cache.ports[port - rdma_start_port(device)].pkey = pkey_cache;
-	if (!use_roce_gid_table) {
-		write_lock(&table->rwlock);
-		for (i = 0; i < gid_cache->table_len; i++) {
-			modify_gid(device, port, table, i, gid_cache->table + i,
-				   &zattr, false);
-		}
-		write_unlock(&table->rwlock);
-	}
-
 	device->cache.ports[port - rdma_start_port(device)].lmc = tprops->lmc;
 	device->cache.ports[port - rdma_start_port(device)].port_state =
 		tprops->state;
@@ -1130,14 +1177,12 @@ static void ib_cache_update(struct ib_device *device,
 					 port,
 					 tprops->subnet_prefix);
 
-	kfree(gid_cache);
 	kfree(old_pkey_cache);
 	kfree(tprops);
 	return;
 
 err:
 	kfree(pkey_cache);
-	kfree(gid_cache);
 	kfree(tprops);
 }
 

drivers/infiniband/core/cm.c

@@ -462,13 +462,31 @@ static int cm_init_av_for_response(struct cm_port *port, struct ib_wc *wc,
 				       grh, &av->ah_attr);
 }
 
-static int cm_init_av_by_path(struct sa_path_rec *path, struct cm_av *av,
-			      struct cm_id_private *cm_id_priv)
+static int add_cm_id_to_port_list(struct cm_id_private *cm_id_priv,
+				  struct cm_av *av,
+				  struct cm_port *port)
+{
+	unsigned long flags;
+	int ret = 0;
+
+	spin_lock_irqsave(&cm.lock, flags);
+
+	if (&cm_id_priv->av == av)
+		list_add_tail(&cm_id_priv->prim_list, &port->cm_priv_prim_list);
+	else if (&cm_id_priv->alt_av == av)
+		list_add_tail(&cm_id_priv->altr_list, &port->cm_priv_altr_list);
+	else
+		ret = -EINVAL;
+
+	spin_unlock_irqrestore(&cm.lock, flags);
+	return ret;
+}
+
+static struct cm_port *get_cm_port_from_path(struct sa_path_rec *path)
 {
 	struct cm_device *cm_dev;
 	struct cm_port *port = NULL;
 	unsigned long flags;
-	int ret;
 	u8 p;
 	struct net_device *ndev = ib_get_ndev_from_path(path);
 
@@ -477,7 +495,7 @@ static int cm_init_av_by_path(struct sa_path_rec *path, struct cm_av *av,
 		if (!ib_find_cached_gid(cm_dev->ib_device, &path->sgid,
 					sa_conv_pathrec_to_gid_type(path),
 					ndev, &p, NULL)) {
-			port = cm_dev->port[p-1];
+			port = cm_dev->port[p - 1];
 			break;
 		}
 	}
@@ -485,9 +503,20 @@ static int cm_init_av_by_path(struct sa_path_rec *path, struct cm_av *av,
 
 	if (ndev)
 		dev_put(ndev);
+	return port;
+}
 
+static int cm_init_av_by_path(struct sa_path_rec *path, struct cm_av *av,
+			      struct cm_id_private *cm_id_priv)
+{
+	struct cm_device *cm_dev;
+	struct cm_port *port;
+	int ret;
+
+	port = get_cm_port_from_path(path);
 	if (!port)
 		return -EINVAL;
+	cm_dev = port->cm_dev;
 
 	ret = ib_find_cached_pkey(cm_dev->ib_device, port->port_num,
 				  be16_to_cpu(path->pkey), &av->pkey_index);
@@ -502,16 +531,7 @@ static int cm_init_av_by_path(struct sa_path_rec *path, struct cm_av *av,
 
 	av->timeout = path->packet_life_time + 1;
 
-	spin_lock_irqsave(&cm.lock, flags);
-	if (&cm_id_priv->av == av)
-		list_add_tail(&cm_id_priv->prim_list, &port->cm_priv_prim_list);
-	else if (&cm_id_priv->alt_av == av)
-		list_add_tail(&cm_id_priv->altr_list, &port->cm_priv_altr_list);
-	else
-		ret = -EINVAL;
-
-	spin_unlock_irqrestore(&cm.lock, flags);
-
+	ret = add_cm_id_to_port_list(cm_id_priv, av, port);
 	return ret;
 }
 
@@ -1523,6 +1543,8 @@ static void cm_format_paths_from_req(struct cm_req_msg *req_msg,
 		cm_req_get_primary_local_ack_timeout(req_msg);
 	primary_path->packet_life_time -= (primary_path->packet_life_time > 0);
 	primary_path->service_id = req_msg->service_id;
+	if (sa_path_is_roce(primary_path))
+		primary_path->roce.route_resolved = false;
 
 	if (cm_req_has_alt_path(req_msg)) {
 		alt_path->dgid = req_msg->alt_local_gid;
@@ -1542,6 +1564,9 @@ static void cm_format_paths_from_req(struct cm_req_msg *req_msg,
 			cm_req_get_alt_local_ack_timeout(req_msg);
 		alt_path->packet_life_time -= (alt_path->packet_life_time > 0);
 		alt_path->service_id = req_msg->service_id;
+
+		if (sa_path_is_roce(alt_path))
+			alt_path->roce.route_resolved = false;
 	}
 	cm_format_path_lid_from_req(req_msg, primary_path, alt_path);
 }
@@ -3150,6 +3175,13 @@ static int cm_lap_handler(struct cm_work *work)
 	struct ib_mad_send_buf *msg = NULL;
 	int ret;
 
+	/* Currently Alternate path messages are not supported for
+	 * RoCE link layer.
+	 */
+	if (rdma_protocol_roce(work->port->cm_dev->ib_device,
+			       work->port->port_num))
+		return -EINVAL;
+
 	/* todo: verify LAP request and send reject APR if invalid. */
 	lap_msg = (struct cm_lap_msg *)work->mad_recv_wc->recv_buf.mad;
 	cm_id_priv = cm_acquire_id(lap_msg->remote_comm_id,
@@ -3299,6 +3331,13 @@ static int cm_apr_handler(struct cm_work *work)
 	struct cm_apr_msg *apr_msg;
 	int ret;
 
+	/* Currently Alternate path messages are not supported for
+	 * RoCE link layer.
+	 */
+	if (rdma_protocol_roce(work->port->cm_dev->ib_device,
+			       work->port->port_num))
+		return -EINVAL;
+
 	apr_msg = (struct cm_apr_msg *)work->mad_recv_wc->recv_buf.mad;
 	cm_id_priv = cm_acquire_id(apr_msg->remote_comm_id,
 				   apr_msg->local_comm_id);

drivers/infiniband/core/cma.c

@@ -62,6 +62,7 @@
 #include <rdma/iw_cm.h>
 
 #include "core_priv.h"
+#include "cma_priv.h"
 
 MODULE_AUTHOR("Sean Hefty");
 MODULE_DESCRIPTION("Generic RDMA CM Agent");
@@ -174,7 +175,7 @@ static struct cma_pernet *cma_pernet(struct net *net)
 	return net_generic(net, cma_pernet_id);
 }
 
-static struct idr *cma_pernet_idr(struct net *net, enum rdma_port_space ps)
+static struct idr *cma_pernet_idr(struct net *net, enum rdma_ucm_port_space ps)
 {
 	struct cma_pernet *pernet = cma_pernet(net);
 
@@ -203,7 +204,7 @@ struct cma_device {
 };
 
 struct rdma_bind_list {
-	enum rdma_port_space	ps;
+	enum rdma_ucm_port_space ps;
 	struct hlist_head	owners;
 	unsigned short		port;
 };
@@ -216,7 +217,7 @@ struct class_port_info_context {
 	u8				port_num;
 };
 
-static int cma_ps_alloc(struct net *net, enum rdma_port_space ps,
+static int cma_ps_alloc(struct net *net, enum rdma_ucm_port_space ps,
 			struct rdma_bind_list *bind_list, int snum)
 {
 	struct idr *idr = cma_pernet_idr(net, ps);
@@ -225,14 +226,15 @@ static int cma_ps_alloc(struct net *net, enum rdma_port_space ps,
 }
 
 static struct rdma_bind_list *cma_ps_find(struct net *net,
-					  enum rdma_port_space ps, int snum)
+					  enum rdma_ucm_port_space ps, int snum)
 {
 	struct idr *idr = cma_pernet_idr(net, ps);
 
 	return idr_find(idr, snum);
 }
 
-static void cma_ps_remove(struct net *net, enum rdma_port_space ps, int snum)
+static void cma_ps_remove(struct net *net, enum rdma_ucm_port_space ps,
+			  int snum)
 {
 	struct idr *idr = cma_pernet_idr(net, ps);
 
@@ -327,46 +329,6 @@ struct ib_device *cma_get_ib_dev(struct cma_device *cma_dev)
  * We do this by disabling removal notification while a callback is in process,
  * and reporting it after the callback completes.
  */
-struct rdma_id_private {
-	struct rdma_cm_id	id;
-
-	struct rdma_bind_list	*bind_list;
-	struct hlist_node	node;
-	struct list_head	list; /* listen_any_list or cma_device.list */
-	struct list_head	listen_list; /* per device listens */
-	struct cma_device	*cma_dev;
-	struct list_head	mc_list;
-
-	int			internal_id;
-	enum rdma_cm_state	state;
-	spinlock_t		lock;
-	struct mutex		qp_mutex;
-
-	struct completion	comp;
-	atomic_t		refcount;
-	struct mutex		handler_mutex;
-
-	int			backlog;
-	int			timeout_ms;
-	struct ib_sa_query	*query;
-	int			query_id;
-	union {
-		struct ib_cm_id	*ib;
-		struct iw_cm_id	*iw;
-	} cm_id;
-
-	u32			seq_num;
-	u32			qkey;
-	u32			qp_num;
-	pid_t			owner;
-	u32			options;
-	u8			srq;
-	u8			tos;
-	bool			tos_set;
-	u8			reuseaddr;
-	u8			afonly;
-	enum ib_gid_type	gid_type;
-};
 
 struct cma_multicast {
 	struct rdma_id_private *id_priv;
@@ -505,6 +467,8 @@ static void _cma_attach_to_dev(struct rdma_id_private *id_priv,
 	id_priv->id.route.addr.dev_addr.transport =
 		rdma_node_get_transport(cma_dev->device->node_type);
 	list_add_tail(&id_priv->list, &cma_dev->id_list);
+	id_priv->res.type = RDMA_RESTRACK_CM_ID;
+	rdma_restrack_add(&id_priv->res);
 }
 
 static void cma_attach_to_dev(struct rdma_id_private *id_priv,
@@ -777,10 +741,10 @@ static void cma_deref_id(struct rdma_id_private *id_priv)
 		complete(&id_priv->comp);
 }
 
-struct rdma_cm_id *rdma_create_id(struct net *net,
-				  rdma_cm_event_handler event_handler,
-				  void *context, enum rdma_port_space ps,
-				  enum ib_qp_type qp_type)
+struct rdma_cm_id *__rdma_create_id(struct net *net,
+				    rdma_cm_event_handler event_handler,
+				    void *context, enum rdma_ucm_port_space ps,
+				    enum ib_qp_type qp_type, const char *caller)
 {
 	struct rdma_id_private *id_priv;
 
@@ -788,7 +752,10 @@ struct rdma_cm_id *rdma_create_id(struct net *net,
 	if (!id_priv)
 		return ERR_PTR(-ENOMEM);
 
-	id_priv->owner = task_pid_nr(current);
+	if (caller)
+		id_priv->res.kern_name = caller;
+	else
+		rdma_restrack_set_task(&id_priv->res, current);
 	id_priv->state = RDMA_CM_IDLE;
 	id_priv->id.context = context;
 	id_priv->id.event_handler = event_handler;
@@ -808,7 +775,7 @@ struct rdma_cm_id *rdma_create_id(struct net *net,
 
 	return &id_priv->id;
 }
-EXPORT_SYMBOL(rdma_create_id);
+EXPORT_SYMBOL(__rdma_create_id);
 
 static int cma_init_ud_qp(struct rdma_id_private *id_priv, struct ib_qp *qp)
 {
@@ -1400,7 +1367,7 @@ static struct net_device *cma_get_net_dev(struct ib_cm_event *ib_event,
 	return net_dev;
 }
 
-static enum rdma_port_space rdma_ps_from_service_id(__be64 service_id)
+static enum rdma_ucm_port_space rdma_ps_from_service_id(__be64 service_id)
 {
 	return (be64_to_cpu(service_id) >> 16) & 0xffff;
 }
@@ -1441,21 +1408,12 @@ static bool cma_match_private_data(struct rdma_id_private *id_priv,
 	return true;
 }
 
-static bool cma_protocol_roce_dev_port(struct ib_device *device, int port_num)
-{
-	enum rdma_link_layer ll = rdma_port_get_link_layer(device, port_num);
-	enum rdma_transport_type transport =
-		rdma_node_get_transport(device->node_type);
-
-	return ll == IB_LINK_LAYER_ETHERNET && transport == RDMA_TRANSPORT_IB;
-}
-
 static bool cma_protocol_roce(const struct rdma_cm_id *id)
 {
 	struct ib_device *device = id->device;
 	const int port_num = id->port_num ?: rdma_start_port(device);
 
-	return cma_protocol_roce_dev_port(device, port_num);
+	return rdma_protocol_roce(device, port_num);
 }
 
 static bool cma_match_net_dev(const struct rdma_cm_id *id,
@@ -1468,7 +1426,7 @@ static bool cma_match_net_dev(const struct rdma_cm_id *id,
 		/* This request is an AF_IB request or a RoCE request */
 		return (!id->port_num || id->port_num == port_num) &&
 		       (addr->src_addr.ss_family == AF_IB ||
-			cma_protocol_roce_dev_port(id->device, port_num));
+			rdma_protocol_roce(id->device, port_num));
 
 	return !addr->dev_addr.bound_dev_if ||
 	       (net_eq(dev_net(net_dev), addr->dev_addr.net) &&
@@ -1523,7 +1481,7 @@ static struct rdma_id_private *cma_id_from_event(struct ib_cm_id *cm_id,
 		if (PTR_ERR(*net_dev) == -EAFNOSUPPORT) {
 			/* Assuming the protocol is AF_IB */
 			*net_dev = NULL;
-		} else if (cma_protocol_roce_dev_port(req.device, req.port)) {
+		} else if (rdma_protocol_roce(req.device, req.port)) {
 			/* TODO find the net dev matching the request parameters
 			 * through the RoCE GID table */
 			*net_dev = NULL;
@@ -1668,6 +1626,7 @@ void rdma_destroy_id(struct rdma_cm_id *id)
 	mutex_unlock(&id_priv->handler_mutex);
 
 	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);
@@ -1817,6 +1776,7 @@ static struct rdma_id_private *cma_new_conn_id(struct rdma_cm_id *listen_id,
 					       struct ib_cm_event *ib_event,
 					       struct net_device *net_dev)
 {
+	struct rdma_id_private *listen_id_priv;
 	struct rdma_id_private *id_priv;
 	struct rdma_cm_id *id;
 	struct rdma_route *rt;
@@ -1826,9 +1786,11 @@ static struct rdma_id_private *cma_new_conn_id(struct rdma_cm_id *listen_id,
 		ib_event->param.req_rcvd.primary_path->service_id;
 	int ret;
 
-	id = rdma_create_id(listen_id->route.addr.dev_addr.net,
+	listen_id_priv = container_of(listen_id, struct rdma_id_private, id);
+	id = __rdma_create_id(listen_id->route.addr.dev_addr.net,
 			    listen_id->event_handler, listen_id->context,
-			    listen_id->ps, ib_event->param.req_rcvd.qp_type);
+			    listen_id->ps, ib_event->param.req_rcvd.qp_type,
+			    listen_id_priv->res.kern_name);
 	if (IS_ERR(id))
 		return NULL;
 
@@ -1877,14 +1839,17 @@ static struct rdma_id_private *cma_new_udp_id(struct rdma_cm_id *listen_id,
 					      struct ib_cm_event *ib_event,
 					      struct net_device *net_dev)
 {
+	struct rdma_id_private *listen_id_priv;
 	struct rdma_id_private *id_priv;
 	struct rdma_cm_id *id;
 	const sa_family_t ss_family = listen_id->route.addr.src_addr.ss_family;
 	struct net *net = listen_id->route.addr.dev_addr.net;
 	int ret;
 
-	id = rdma_create_id(net, listen_id->event_handler, listen_id->context,
-			    listen_id->ps, IB_QPT_UD);
+	listen_id_priv = container_of(listen_id, struct rdma_id_private, id);
+	id = __rdma_create_id(net, listen_id->event_handler, listen_id->context,
+			      listen_id->ps, IB_QPT_UD,
+			      listen_id_priv->res.kern_name);
 	if (IS_ERR(id))
 		return NULL;
 
@@ -2150,10 +2115,11 @@ static int iw_conn_req_handler(struct iw_cm_id *cm_id,
 		goto out;
 
 	/* Create a new RDMA id for the new IW CM ID */
-	new_cm_id = rdma_create_id(listen_id->id.route.addr.dev_addr.net,
-				   listen_id->id.event_handler,
-				   listen_id->id.context,
-				   RDMA_PS_TCP, IB_QPT_RC);
+	new_cm_id = __rdma_create_id(listen_id->id.route.addr.dev_addr.net,
+				     listen_id->id.event_handler,
+				     listen_id->id.context,
+				     RDMA_PS_TCP, IB_QPT_RC,
+				     listen_id->res.kern_name);
 	if (IS_ERR(new_cm_id)) {
 		ret = -ENOMEM;
 		goto out;
@@ -2278,8 +2244,8 @@ static void cma_listen_on_dev(struct rdma_id_private *id_priv,
 	if (cma_family(id_priv) == AF_IB && !rdma_cap_ib_cm(cma_dev->device, 1))
 		return;
 
-	id = rdma_create_id(net, cma_listen_handler, id_priv, id_priv->id.ps,
-			    id_priv->id.qp_type);
+	id = __rdma_create_id(net, cma_listen_handler, id_priv, id_priv->id.ps,
+			      id_priv->id.qp_type, id_priv->res.kern_name);
 	if (IS_ERR(id))
 		return;
 
@@ -2541,6 +2507,7 @@ cma_iboe_set_path_rec_l2_fields(struct rdma_id_private *id_priv)
 		gid_type = ib_network_to_gid_type(addr->dev_addr.network);
 	route->path_rec->rec_type = sa_conv_gid_to_pathrec_type(gid_type);
 
+	route->path_rec->roce.route_resolved = true;
 	sa_path_set_ndev(route->path_rec, addr->dev_addr.net);
 	sa_path_set_ifindex(route->path_rec, ndev->ifindex);
 	sa_path_set_dmac(route->path_rec, addr->dev_addr.dst_dev_addr);
@@ -3028,7 +2995,7 @@ static void cma_bind_port(struct rdma_bind_list *bind_list,
 	hlist_add_head(&id_priv->node, &bind_list->owners);
 }
 
-static int cma_alloc_port(enum rdma_port_space ps,
+static int cma_alloc_port(enum rdma_ucm_port_space ps,
 			  struct rdma_id_private *id_priv, unsigned short snum)
 {
 	struct rdma_bind_list *bind_list;
@@ -3091,7 +3058,7 @@ static int cma_port_is_unique(struct rdma_bind_list *bind_list,
 	return 0;
 }
 
-static int cma_alloc_any_port(enum rdma_port_space ps,
+static int cma_alloc_any_port(enum rdma_ucm_port_space ps,
 			      struct rdma_id_private *id_priv)
 {
 	static unsigned int last_used_port;
@@ -3169,7 +3136,7 @@ static int cma_check_port(struct rdma_bind_list *bind_list,
 	return 0;
 }
 
-static int cma_use_port(enum rdma_port_space ps,
+static int cma_use_port(enum rdma_ucm_port_space ps,
 			struct rdma_id_private *id_priv)
 {
 	struct rdma_bind_list *bind_list;
@@ -3203,8 +3170,8 @@ static int cma_bind_listen(struct rdma_id_private *id_priv)
 	return ret;
 }
 
-static enum rdma_port_space cma_select_inet_ps(
-		struct rdma_id_private *id_priv)
+static enum rdma_ucm_port_space
+cma_select_inet_ps(struct rdma_id_private *id_priv)
 {
 	switch (id_priv->id.ps) {
 	case RDMA_PS_TCP:
@@ -3218,9 +3185,10 @@ static enum rdma_port_space cma_select_inet_ps(
 	}
 }
 
-static enum rdma_port_space cma_select_ib_ps(struct rdma_id_private *id_priv)
+static enum rdma_ucm_port_space
+cma_select_ib_ps(struct rdma_id_private *id_priv)
 {
-	enum rdma_port_space ps = 0;
+	enum rdma_ucm_port_space ps = 0;
 	struct sockaddr_ib *sib;
 	u64 sid_ps, mask, sid;
 
@@ -3251,7 +3219,7 @@ static enum rdma_port_space cma_select_ib_ps(struct rdma_id_private *id_priv)
 
 static int cma_get_port(struct rdma_id_private *id_priv)
 {
-	enum rdma_port_space ps;
+	enum rdma_ucm_port_space ps;
 	int ret;
 
 	if (cma_family(id_priv) != AF_IB)
@@ -3389,8 +3357,10 @@ int rdma_bind_addr(struct rdma_cm_id *id, struct sockaddr *addr)
 
 	return 0;
 err2:
-	if (id_priv->cma_dev)
+	if (id_priv->cma_dev) {
+		rdma_restrack_del(&id_priv->res);
 		cma_release_dev(id_priv);
+	}
 err1:
 	cma_comp_exch(id_priv, RDMA_CM_ADDR_BOUND, RDMA_CM_IDLE);
 	return ret;
@@ -3773,14 +3743,18 @@ static int cma_send_sidr_rep(struct rdma_id_private *id_priv,
 	return ib_send_cm_sidr_rep(id_priv->cm_id.ib, &rep);
 }
 
-int rdma_accept(struct rdma_cm_id *id, struct rdma_conn_param *conn_param)
+int __rdma_accept(struct rdma_cm_id *id, struct rdma_conn_param *conn_param,
+		  const char *caller)
 {
 	struct rdma_id_private *id_priv;
 	int ret;
 
 	id_priv = container_of(id, struct rdma_id_private, id);
 
-	id_priv->owner = task_pid_nr(current);
+	if (caller)
+		id_priv->res.kern_name = caller;
+	else
+		rdma_restrack_set_task(&id_priv->res, current);
 
 	if (!cma_comp(id_priv, RDMA_CM_CONNECT))
 		return -EINVAL;
@@ -3820,7 +3794,7 @@ reject:
 	rdma_reject(id, NULL, 0);
 	return ret;
 }
-EXPORT_SYMBOL(rdma_accept);
+EXPORT_SYMBOL(__rdma_accept);
 
 int rdma_notify(struct rdma_cm_id *id, enum ib_event_type event)
 {
@@ -3938,10 +3912,14 @@ static int cma_ib_mc_handler(int status, struct ib_sa_multicast *multicast)
 			rdma_start_port(id_priv->cma_dev->device)];
 
 		event.event = RDMA_CM_EVENT_MULTICAST_JOIN;
-		ib_init_ah_from_mcmember(id_priv->id.device,
-					 id_priv->id.port_num, &multicast->rec,
-					 ndev, gid_type,
-					 &event.param.ud.ah_attr);
+		ret = ib_init_ah_from_mcmember(id_priv->id.device,
+					       id_priv->id.port_num,
+					       &multicast->rec,
+					       ndev, gid_type,
+					       &event.param.ud.ah_attr);
+		if (ret)
+			event.event = RDMA_CM_EVENT_MULTICAST_ERROR;
+
 		event.param.ud.qp_num = 0xFFFFFF;
 		event.param.ud.qkey = be32_to_cpu(multicast->rec.qkey);
 		if (ndev)
@@ -4501,7 +4479,7 @@ static int cma_get_id_stats(struct sk_buff *skb, struct netlink_callback *cb)
 					  RDMA_NL_RDMA_CM_ATTR_DST_ADDR))
 				goto out;
 
-			id_stats->pid		= id_priv->owner;
+			id_stats->pid	= task_pid_vnr(id_priv->res.task);
 			id_stats->port_space	= id->ps;
 			id_stats->cm_state	= id_priv->state;
 			id_stats->qp_num	= id_priv->qp_num;

drivers/infiniband/core/cma_priv.h

@@ -0,0 +1,97 @@
+/*
+ * Copyright (c) 2005 Voltaire Inc.  All rights reserved.
+ * Copyright (c) 2002-2005, Network Appliance, Inc. All rights reserved.
+ * Copyright (c) 1999-2005, Mellanox Technologies, Inc. All rights reserved.
+ * Copyright (c) 2005-2006 Intel Corporation.  All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef _CMA_PRIV_H
+#define _CMA_PRIV_H
+
+enum rdma_cm_state {
+	RDMA_CM_IDLE,
+	RDMA_CM_ADDR_QUERY,
+	RDMA_CM_ADDR_RESOLVED,
+	RDMA_CM_ROUTE_QUERY,
+	RDMA_CM_ROUTE_RESOLVED,
+	RDMA_CM_CONNECT,
+	RDMA_CM_DISCONNECT,
+	RDMA_CM_ADDR_BOUND,
+	RDMA_CM_LISTEN,
+	RDMA_CM_DEVICE_REMOVAL,
+	RDMA_CM_DESTROYING
+};
+
+struct rdma_id_private {
+	struct rdma_cm_id	id;
+
+	struct rdma_bind_list	*bind_list;
+	struct hlist_node	node;
+	struct list_head	list; /* listen_any_list or cma_device.list */
+	struct list_head	listen_list; /* per device listens */
+	struct cma_device	*cma_dev;
+	struct list_head	mc_list;
+
+	int			internal_id;
+	enum rdma_cm_state	state;
+	spinlock_t		lock;
+	struct mutex		qp_mutex;
+
+	struct completion	comp;
+	atomic_t		refcount;
+	struct mutex		handler_mutex;
+
+	int			backlog;
+	int			timeout_ms;
+	struct ib_sa_query	*query;
+	int			query_id;
+	union {
+		struct ib_cm_id	*ib;
+		struct iw_cm_id	*iw;
+	} cm_id;
+
+	u32			seq_num;
+	u32			qkey;
+	u32			qp_num;
+	u32			options;
+	u8			srq;
+	u8			tos;
+	bool			tos_set;
+	u8			reuseaddr;
+	u8			afonly;
+	enum ib_gid_type	gid_type;
+
+	/*
+	 * Internal to RDMA/core, don't use in the drivers
+	 */
+	struct rdma_restrack_entry     res;
+};
+#endif /* _CMA_PRIV_H */

drivers/infiniband/core/core_priv.h

@@ -333,4 +333,15 @@ static inline struct ib_qp *_ib_create_qp(struct ib_device *dev,
 
 	return qp;
 }
+
+struct rdma_dev_addr;
+int rdma_resolve_ip_route(struct sockaddr *src_addr,
+			  const struct sockaddr *dst_addr,
+			  struct rdma_dev_addr *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,
+				 int *hoplimit);
+
 #endif /* _CORE_PRIV_H */

drivers/infiniband/core/device.c

@@ -103,7 +103,6 @@ static int ib_device_check_mandatory(struct ib_device *device)
 		IB_MANDATORY_FUNC(query_device),
 		IB_MANDATORY_FUNC(query_port),
 		IB_MANDATORY_FUNC(query_pkey),
-		IB_MANDATORY_FUNC(query_gid),
 		IB_MANDATORY_FUNC(alloc_pd),
 		IB_MANDATORY_FUNC(dealloc_pd),
 		IB_MANDATORY_FUNC(create_ah),
@@ -853,7 +852,7 @@ int ib_query_port(struct ib_device *device,
 	if (rdma_port_get_link_layer(device, port_num) != IB_LINK_LAYER_INFINIBAND)
 		return 0;
 
-	err = ib_query_gid(device, port_num, 0, &gid, NULL);
+	err = device->query_gid(device, port_num, 0, &gid);
 	if (err)
 		return err;
 
@@ -871,19 +870,13 @@ EXPORT_SYMBOL(ib_query_port);
  * @attr: Returned GID attributes related to this GID index (only in RoCE).
  *   NULL means ignore.
  *
- * ib_query_gid() fetches the specified GID table entry.
+ * ib_query_gid() fetches the specified GID table entry from the cache.
  */
 int ib_query_gid(struct ib_device *device,
 		 u8 port_num, int index, union ib_gid *gid,
 		 struct ib_gid_attr *attr)
 {
-	if (rdma_cap_roce_gid_table(device, port_num))
-		return ib_get_cached_gid(device, port_num, index, gid, attr);
-
-	if (attr)
-		return -EINVAL;
-
-	return device->query_gid(device, port_num, index, gid);
+	return ib_get_cached_gid(device, port_num, index, gid, attr);
 }
 EXPORT_SYMBOL(ib_query_gid);
 
@@ -1049,19 +1042,18 @@ EXPORT_SYMBOL(ib_modify_port);
  *   a specified GID value occurs. Its searches only for IB link layer.
  * @device: The device to query.
  * @gid: The GID value to search for.
- * @ndev: The ndev related to the GID to search for.
  * @port_num: The port number of the device where the GID value was found.
  * @index: The index into the GID table where the GID was found.  This
  *   parameter may be NULL.
  */
 int ib_find_gid(struct ib_device *device, union ib_gid *gid,
-		struct net_device *ndev, u8 *port_num, u16 *index)
+		u8 *port_num, u16 *index)
 {
 	union ib_gid tmp_gid;
 	int ret, port, i;
 
 	for (port = rdma_start_port(device); port <= rdma_end_port(device); ++port) {
-		if (rdma_cap_roce_gid_table(device, port))
+		if (!rdma_protocol_ib(device, port))
 			continue;
 
 		for (i = 0; i < device->port_immutable[port].gid_tbl_len; ++i) {

drivers/infiniband/core/iwpm_util.c

@@ -439,10 +439,9 @@ struct sk_buff *iwpm_create_nlmsg(u32 nl_op, struct nlmsghdr **nlh,
 	struct sk_buff *skb = NULL;
 
 	skb = dev_alloc_skb(IWPM_MSG_SIZE);
-	if (!skb) {
-		pr_err("%s Unable to allocate skb\n", __func__);
+	if (!skb)
 		goto create_nlmsg_exit;
-	}
+
 	if (!(ibnl_put_msg(skb, nlh, 0, 0, nl_client, nl_op,
 			   NLM_F_REQUEST))) {
 		pr_warn("%s: Unable to put the nlmsg header\n", __func__);

drivers/infiniband/core/multicast.c

@@ -724,21 +724,19 @@ int ib_init_ah_from_mcmember(struct ib_device *device, u8 port_num,
 {
 	int ret;
 	u16 gid_index;
-	u8 p;
-
-	if (rdma_protocol_roce(device, port_num)) {
-		ret = ib_find_cached_gid_by_port(device, &rec->port_gid,
-						 gid_type, port_num,
-						 ndev,
-						 &gid_index);
-	} else if (rdma_protocol_ib(device, port_num)) {
-		ret = ib_find_cached_gid(device, &rec->port_gid,
-					 IB_GID_TYPE_IB, NULL, &p,
-					 &gid_index);
-	} else {
-		ret = -EINVAL;
-	}
 
+	/* GID table is not based on the netdevice for IB link layer,
+	 * so ignore ndev during search.
+	 */
+	if (rdma_protocol_ib(device, port_num))
+		ndev = NULL;
+	else if (!rdma_protocol_roce(device, port_num))
+		return -EINVAL;
+
+	ret = ib_find_cached_gid_by_port(device, &rec->port_gid,
+					 gid_type, port_num,
+					 ndev,
+					 &gid_index);
 	if (ret)
 		return ret;
 

drivers/infiniband/core/nldev.c

@@ -34,9 +34,11 @@
 #include <linux/pid.h>
 #include <linux/pid_namespace.h>
 #include <net/netlink.h>
+#include <rdma/rdma_cm.h>
 #include <rdma/rdma_netlink.h>
 
 #include "core_priv.h"
+#include "cma_priv.h"
 
 static const struct nla_policy nldev_policy[RDMA_NLDEV_ATTR_MAX] = {
 	[RDMA_NLDEV_ATTR_DEV_INDEX]     = { .type = NLA_U32 },
@@ -71,6 +73,31 @@ static const struct nla_policy nldev_policy[RDMA_NLDEV_ATTR_MAX] = {
 	[RDMA_NLDEV_ATTR_RES_PID]		= { .type = NLA_U32 },
 	[RDMA_NLDEV_ATTR_RES_KERN_NAME]		= { .type = NLA_NUL_STRING,
 						    .len = TASK_COMM_LEN },
+	[RDMA_NLDEV_ATTR_RES_CM_ID]		= { .type = NLA_NESTED },
+	[RDMA_NLDEV_ATTR_RES_CM_ID_ENTRY]	= { .type = NLA_NESTED },
+	[RDMA_NLDEV_ATTR_RES_PS]		= { .type = NLA_U32 },
+	[RDMA_NLDEV_ATTR_RES_SRC_ADDR]	= {
+			.len = sizeof(struct __kernel_sockaddr_storage) },
+	[RDMA_NLDEV_ATTR_RES_DST_ADDR]	= {
+			.len = sizeof(struct __kernel_sockaddr_storage) },
+	[RDMA_NLDEV_ATTR_RES_CQ]		= { .type = NLA_NESTED },
+	[RDMA_NLDEV_ATTR_RES_CQ_ENTRY]		= { .type = NLA_NESTED },
+	[RDMA_NLDEV_ATTR_RES_CQE]		= { .type = NLA_U32 },
+	[RDMA_NLDEV_ATTR_RES_USECNT]		= { .type = NLA_U64 },
+	[RDMA_NLDEV_ATTR_RES_POLL_CTX]		= { .type = NLA_U8 },
+	[RDMA_NLDEV_ATTR_RES_MR]		= { .type = NLA_NESTED },
+	[RDMA_NLDEV_ATTR_RES_MR_ENTRY]		= { .type = NLA_NESTED },
+	[RDMA_NLDEV_ATTR_RES_RKEY]		= { .type = NLA_U32 },
+	[RDMA_NLDEV_ATTR_RES_LKEY]		= { .type = NLA_U32 },
+	[RDMA_NLDEV_ATTR_RES_IOVA]		= { .type = NLA_U64 },
+	[RDMA_NLDEV_ATTR_RES_MRLEN]		= { .type = NLA_U64 },
+	[RDMA_NLDEV_ATTR_RES_PD]		= { .type = NLA_NESTED },
+	[RDMA_NLDEV_ATTR_RES_PD_ENTRY]		= { .type = NLA_NESTED },
+	[RDMA_NLDEV_ATTR_RES_LOCAL_DMA_LKEY]	= { .type = NLA_U32 },
+	[RDMA_NLDEV_ATTR_RES_UNSAFE_GLOBAL_RKEY] = { .type = NLA_U32 },
+	[RDMA_NLDEV_ATTR_NDEV_INDEX]		= { .type = NLA_U32 },
+	[RDMA_NLDEV_ATTR_NDEV_NAME]		= { .type = NLA_NUL_STRING,
+						    .len = IFNAMSIZ },
 };
 
 static int fill_nldev_handle(struct sk_buff *msg, struct ib_device *device)
@@ -99,7 +126,7 @@ static int fill_dev_info(struct sk_buff *msg, struct ib_device *device)
 		return -EMSGSIZE;
 
 	ib_get_device_fw_str(device, fw);
-	/* Device without FW has strlen(fw) */
+	/* Device without FW has strlen(fw) = 0 */
 	if (strlen(fw) && nla_put_string(msg, RDMA_NLDEV_ATTR_FW_VERSION, fw))
 		return -EMSGSIZE;
 
@@ -115,8 +142,10 @@ static int fill_dev_info(struct sk_buff *msg, struct ib_device *device)
 }
 
 static int fill_port_info(struct sk_buff *msg,
-			  struct ib_device *device, u32 port)
+			  struct ib_device *device, u32 port,
+			  const struct net *net)
 {
+	struct net_device *netdev = NULL;
 	struct ib_port_attr attr;
 	int ret;
 
@@ -150,7 +179,23 @@ static int fill_port_info(struct sk_buff *msg,
 		return -EMSGSIZE;
 	if (nla_put_u8(msg, RDMA_NLDEV_ATTR_PORT_PHYS_STATE, attr.phys_state))
 		return -EMSGSIZE;
-	return 0;
+
+	if (device->get_netdev)
+		netdev = device->get_netdev(device, port);
+
+	if (netdev && net_eq(dev_net(netdev), net)) {
+		ret = nla_put_u32(msg,
+				  RDMA_NLDEV_ATTR_NDEV_INDEX, netdev->ifindex);
+		if (ret)
+			goto out;
+		ret = nla_put_string(msg,
+				     RDMA_NLDEV_ATTR_NDEV_NAME, netdev->name);
+	}
+
+out:
+	if (netdev)
+		dev_put(netdev);
+	return ret;
 }
 
 static int fill_res_info_entry(struct sk_buff *msg,
@@ -182,6 +227,8 @@ static int fill_res_info(struct sk_buff *msg, struct ib_device *device)
 		[RDMA_RESTRACK_PD] = "pd",
 		[RDMA_RESTRACK_CQ] = "cq",
 		[RDMA_RESTRACK_QP] = "qp",
+		[RDMA_RESTRACK_CM_ID] = "cm_id",
+		[RDMA_RESTRACK_MR] = "mr",
 	};
 
 	struct rdma_restrack_root *res = &device->res;
@@ -212,10 +259,29 @@ err:
 	return ret;
 }
 
-static int fill_res_qp_entry(struct sk_buff *msg,
-			     struct ib_qp *qp, uint32_t port)
+static int fill_res_name_pid(struct sk_buff *msg,
+			     struct rdma_restrack_entry *res)
+{
+	/*
+	 * For user resources, user is should read /proc/PID/comm to get the
+	 * name of the task file.
+	 */
+	if (rdma_is_kernel_res(res)) {
+		if (nla_put_string(msg, RDMA_NLDEV_ATTR_RES_KERN_NAME,
+		    res->kern_name))
+			return -EMSGSIZE;
+	} else {
+		if (nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_PID,
+		    task_pid_vnr(res->task)))
+			return -EMSGSIZE;
+	}
+	return 0;
+}
+
+static int fill_res_qp_entry(struct sk_buff *msg, struct netlink_callback *cb,
+			     struct rdma_restrack_entry *res, uint32_t port)
 {
-	struct rdma_restrack_entry *res = &qp->res;
+	struct ib_qp *qp = container_of(res, struct ib_qp, res);
 	struct ib_qp_init_attr qp_init_attr;
 	struct nlattr *entry_attr;
 	struct ib_qp_attr qp_attr;
@@ -262,19 +328,172 @@ static int fill_res_qp_entry(struct sk_buff *msg,
 	if (nla_put_u8(msg, RDMA_NLDEV_ATTR_RES_STATE, qp_attr.qp_state))
 		goto err;
 
-	/*
-	 * Existence of task means that it is user QP and netlink
-	 * user is invited to go and read /proc/PID/comm to get name
-	 * of the task file and res->task_com should be NULL.
-	 */
-	if (rdma_is_kernel_res(res)) {
-		if (nla_put_string(msg, RDMA_NLDEV_ATTR_RES_KERN_NAME, res->kern_name))
+	if (fill_res_name_pid(msg, res))
+		goto err;
+
+	nla_nest_end(msg, entry_attr);
+	return 0;
+
+err:
+	nla_nest_cancel(msg, entry_attr);
+out:
+	return -EMSGSIZE;
+}
+
+static int fill_res_cm_id_entry(struct sk_buff *msg,
+				struct netlink_callback *cb,
+				struct rdma_restrack_entry *res, uint32_t port)
+{
+	struct rdma_id_private *id_priv =
+				container_of(res, struct rdma_id_private, res);
+	struct rdma_cm_id *cm_id = &id_priv->id;
+	struct nlattr *entry_attr;
+
+	if (port && port != cm_id->port_num)
+		return 0;
+
+	entry_attr = nla_nest_start(msg, RDMA_NLDEV_ATTR_RES_CM_ID_ENTRY);
+	if (!entry_attr)
+		goto out;
+
+	if (cm_id->port_num &&
+	    nla_put_u32(msg, RDMA_NLDEV_ATTR_PORT_INDEX, cm_id->port_num))
+		goto err;
+
+	if (id_priv->qp_num) {
+		if (nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_LQPN, id_priv->qp_num))
 			goto err;
-	} else {
-		if (nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_PID, task_pid_vnr(res->task)))
+		if (nla_put_u8(msg, RDMA_NLDEV_ATTR_RES_TYPE, cm_id->qp_type))
 			goto err;
 	}
 
+	if (nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_PS, cm_id->ps))
+		goto err;
+
+	if (nla_put_u8(msg, RDMA_NLDEV_ATTR_RES_STATE, id_priv->state))
+		goto err;
+
+	if (cm_id->route.addr.src_addr.ss_family &&
+	    nla_put(msg, RDMA_NLDEV_ATTR_RES_SRC_ADDR,
+		    sizeof(cm_id->route.addr.src_addr),
+		    &cm_id->route.addr.src_addr))
+		goto err;
+	if (cm_id->route.addr.dst_addr.ss_family &&
+	    nla_put(msg, RDMA_NLDEV_ATTR_RES_DST_ADDR,
+		    sizeof(cm_id->route.addr.dst_addr),
+		    &cm_id->route.addr.dst_addr))
+		goto err;
+
+	if (fill_res_name_pid(msg, res))
+		goto err;
+
+	nla_nest_end(msg, entry_attr);
+	return 0;
+
+err:
+	nla_nest_cancel(msg, entry_attr);
+out:
+	return -EMSGSIZE;
+}
+
+static int fill_res_cq_entry(struct sk_buff *msg, struct netlink_callback *cb,
+			     struct rdma_restrack_entry *res, uint32_t port)
+{
+	struct ib_cq *cq = container_of(res, struct ib_cq, res);
+	struct nlattr *entry_attr;
+
+	entry_attr = nla_nest_start(msg, RDMA_NLDEV_ATTR_RES_CQ_ENTRY);
+	if (!entry_attr)
+		goto out;
+
+	if (nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_CQE, cq->cqe))
+		goto err;
+	if (nla_put_u64_64bit(msg, RDMA_NLDEV_ATTR_RES_USECNT,
+			      atomic_read(&cq->usecnt), 0))
+		goto err;
+
+	/* Poll context is only valid for kernel CQs */
+	if (rdma_is_kernel_res(res) &&
+	    nla_put_u8(msg, RDMA_NLDEV_ATTR_RES_POLL_CTX, cq->poll_ctx))
+		goto err;
+
+	if (fill_res_name_pid(msg, res))
+		goto err;
+
+	nla_nest_end(msg, entry_attr);
+	return 0;
+
+err:
+	nla_nest_cancel(msg, entry_attr);
+out:
+	return -EMSGSIZE;
+}
+
+static int fill_res_mr_entry(struct sk_buff *msg, struct netlink_callback *cb,
+			     struct rdma_restrack_entry *res, uint32_t port)
+{
+	struct ib_mr *mr = container_of(res, struct ib_mr, res);
+	struct nlattr *entry_attr;
+
+	entry_attr = nla_nest_start(msg, RDMA_NLDEV_ATTR_RES_MR_ENTRY);
+	if (!entry_attr)
+		goto out;
+
+	if (netlink_capable(cb->skb, CAP_NET_ADMIN)) {
+		if (nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_RKEY, mr->rkey))
+			goto err;
+		if (nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_LKEY, mr->lkey))
+			goto err;
+		if (nla_put_u64_64bit(msg, RDMA_NLDEV_ATTR_RES_IOVA,
+				      mr->iova, 0))
+			goto err;
+	}
+
+	if (nla_put_u64_64bit(msg, RDMA_NLDEV_ATTR_RES_MRLEN, mr->length, 0))
+		goto err;
+
+	if (fill_res_name_pid(msg, res))
+		goto err;
+
+	nla_nest_end(msg, entry_attr);
+	return 0;
+
+err:
+	nla_nest_cancel(msg, entry_attr);
+out:
+	return -EMSGSIZE;
+}
+
+static int fill_res_pd_entry(struct sk_buff *msg, struct netlink_callback *cb,
+			     struct rdma_restrack_entry *res, uint32_t port)
+{
+	struct ib_pd *pd = container_of(res, struct ib_pd, res);
+	struct nlattr *entry_attr;
+
+	entry_attr = nla_nest_start(msg, RDMA_NLDEV_ATTR_RES_PD_ENTRY);
+	if (!entry_attr)
+		goto out;
+
+	if (netlink_capable(cb->skb, CAP_NET_ADMIN)) {
+		if (nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_LOCAL_DMA_LKEY,
+				pd->local_dma_lkey))
+			goto err;
+		if ((pd->flags & IB_PD_UNSAFE_GLOBAL_RKEY) &&
+		    nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_UNSAFE_GLOBAL_RKEY,
+				pd->unsafe_global_rkey))
+			goto err;
+	}
+	if (nla_put_u64_64bit(msg, RDMA_NLDEV_ATTR_RES_USECNT,
+			      atomic_read(&pd->usecnt), 0))
+		goto err;
+	if ((pd->flags & IB_PD_UNSAFE_GLOBAL_RKEY) &&
+	    nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_UNSAFE_GLOBAL_RKEY,
+			pd->unsafe_global_rkey))
+		goto err;
+
+	if (fill_res_name_pid(msg, res))
+		goto err;
+
 	nla_nest_end(msg, entry_attr);
 	return 0;
 
@@ -405,7 +624,7 @@ static int nldev_port_get_doit(struct sk_buff *skb, struct nlmsghdr *nlh,
 			RDMA_NL_GET_TYPE(RDMA_NL_NLDEV, RDMA_NLDEV_CMD_GET),
 			0, 0);
 
-	err = fill_port_info(msg, device, port);
+	err = fill_port_info(msg, device, port, sock_net(skb->sk));
 	if (err)
 		goto err_free;
 
@@ -465,7 +684,7 @@ static int nldev_port_get_dumpit(struct sk_buff *skb,
 						 RDMA_NLDEV_CMD_PORT_GET),
 				0, NLM_F_MULTI);
 
-		if (fill_port_info(skb, device, p)) {
+		if (fill_port_info(skb, device, p, sock_net(skb->sk))) {
 			nlmsg_cancel(skb, nlh);
 			goto out;
 		}
@@ -558,23 +777,60 @@ static int nldev_res_get_dumpit(struct sk_buff *skb,
 	return ib_enum_all_devs(_nldev_res_get_dumpit, skb, cb);
 }
 
-static int nldev_res_get_qp_dumpit(struct sk_buff *skb,
-				   struct netlink_callback *cb)
+struct nldev_fill_res_entry {
+	int (*fill_res_func)(struct sk_buff *msg, struct netlink_callback *cb,
+			     struct rdma_restrack_entry *res, u32 port);
+	enum rdma_nldev_attr nldev_attr;
+	enum rdma_nldev_command nldev_cmd;
+};
+
+static const struct nldev_fill_res_entry fill_entries[RDMA_RESTRACK_MAX] = {
+	[RDMA_RESTRACK_QP] = {
+		.fill_res_func = fill_res_qp_entry,
+		.nldev_cmd = RDMA_NLDEV_CMD_RES_QP_GET,
+		.nldev_attr = RDMA_NLDEV_ATTR_RES_QP,
+	},
+	[RDMA_RESTRACK_CM_ID] = {
+		.fill_res_func = fill_res_cm_id_entry,
+		.nldev_cmd = RDMA_NLDEV_CMD_RES_CM_ID_GET,
+		.nldev_attr = RDMA_NLDEV_ATTR_RES_CM_ID,
+	},
+	[RDMA_RESTRACK_CQ] = {
+		.fill_res_func = fill_res_cq_entry,
+		.nldev_cmd = RDMA_NLDEV_CMD_RES_CQ_GET,
+		.nldev_attr = RDMA_NLDEV_ATTR_RES_CQ,
+	},
+	[RDMA_RESTRACK_MR] = {
+		.fill_res_func = fill_res_mr_entry,
+		.nldev_cmd = RDMA_NLDEV_CMD_RES_MR_GET,
+		.nldev_attr = RDMA_NLDEV_ATTR_RES_MR,
+	},
+	[RDMA_RESTRACK_PD] = {
+		.fill_res_func = fill_res_pd_entry,
+		.nldev_cmd = RDMA_NLDEV_CMD_RES_PD_GET,
+		.nldev_attr = RDMA_NLDEV_ATTR_RES_PD,
+	},
+};
+
+static int res_get_common_dumpit(struct sk_buff *skb,
+				 struct netlink_callback *cb,
+				 enum rdma_restrack_type res_type)
 {
+	const struct nldev_fill_res_entry *fe = &fill_entries[res_type];
 	struct nlattr *tb[RDMA_NLDEV_ATTR_MAX];
 	struct rdma_restrack_entry *res;
 	int err, ret = 0, idx = 0;
 	struct nlattr *table_attr;
 	struct ib_device *device;
 	int start = cb->args[0];
-	struct ib_qp *qp = NULL;
 	struct nlmsghdr *nlh;
 	u32 index, port = 0;
+	bool filled = false;
 
 	err = nlmsg_parse(cb->nlh, 0, tb, RDMA_NLDEV_ATTR_MAX - 1,
 			  nldev_policy, NULL);
 	/*
-	 * Right now, we are expecting the device index to get QP information,
+	 * Right now, we are expecting the device index to get res information,
 	 * but it is possible to extend this code to return all devices in
 	 * one shot by checking the existence of RDMA_NLDEV_ATTR_DEV_INDEX.
 	 * if it doesn't exist, we will iterate over all devices.
@@ -601,7 +857,7 @@ static int nldev_res_get_qp_dumpit(struct sk_buff *skb,
 	}
 
 	nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
-			RDMA_NL_GET_TYPE(RDMA_NL_NLDEV, RDMA_NLDEV_CMD_RES_QP_GET),
+			RDMA_NL_GET_TYPE(RDMA_NL_NLDEV, fe->nldev_cmd),
 			0, NLM_F_MULTI);
 
 	if (fill_nldev_handle(skb, device)) {
@@ -609,24 +865,26 @@ static int nldev_res_get_qp_dumpit(struct sk_buff *skb,
 		goto err;
 	}
 
-	table_attr = nla_nest_start(skb, RDMA_NLDEV_ATTR_RES_QP);
+	table_attr = nla_nest_start(skb, fe->nldev_attr);
 	if (!table_attr) {
 		ret = -EMSGSIZE;
 		goto err;
 	}
 
 	down_read(&device->res.rwsem);
-	hash_for_each_possible(device->res.hash, res, node, RDMA_RESTRACK_QP) {
+	hash_for_each_possible(device->res.hash, res, node, res_type) {
 		if (idx < start)
 			goto next;
 
 		if ((rdma_is_kernel_res(res) &&
 		     task_active_pid_ns(current) != &init_pid_ns) ||
-		    (!rdma_is_kernel_res(res) &&
-		     task_active_pid_ns(current) != task_active_pid_ns(res->task)))
+		    (!rdma_is_kernel_res(res) && task_active_pid_ns(current) !=
+		     task_active_pid_ns(res->task)))
 			/*
-			 * 1. Kernel QPs should be visible in init namspace only
-			 * 2. Present only QPs visible in the current namespace
+			 * 1. Kern resources should be visible in init
+			 *    namspace only
+			 * 2. Present only resources visible in the current
+			 *    namespace
 			 */
 			goto next;
 
@@ -638,10 +896,10 @@ static int nldev_res_get_qp_dumpit(struct sk_buff *skb,
 			 */
 			goto next;
 
-		qp = container_of(res, struct ib_qp, res);
+		filled = true;
 
 		up_read(&device->res.rwsem);
-		ret = fill_res_qp_entry(skb, qp, port);
+		ret = fe->fill_res_func(skb, cb, res, port);
 		down_read(&device->res.rwsem);
 		/*
 		 * Return resource back, but it won't be released till
@@ -667,10 +925,10 @@ next:		idx++;
 	cb->args[0] = idx;
 
 	/*
-	 * No more QPs to fill, cancel the message and
+	 * No more entries to fill, cancel the message and
 	 * return 0 to mark end of dumpit.
 	 */
-	if (!qp)
+	if (!filled)
 		goto err;
 
 	put_device(&device->dev);
@@ -688,6 +946,36 @@ err_index:
 	return ret;
 }
 
+static int nldev_res_get_qp_dumpit(struct sk_buff *skb,
+				   struct netlink_callback *cb)
+{
+	return res_get_common_dumpit(skb, cb, RDMA_RESTRACK_QP);
+}
+
+static int nldev_res_get_cm_id_dumpit(struct sk_buff *skb,
+				      struct netlink_callback *cb)
+{
+	return res_get_common_dumpit(skb, cb, RDMA_RESTRACK_CM_ID);
+}
+
+static int nldev_res_get_cq_dumpit(struct sk_buff *skb,
+				   struct netlink_callback *cb)
+{
+	return res_get_common_dumpit(skb, cb, RDMA_RESTRACK_CQ);
+}
+
+static int nldev_res_get_mr_dumpit(struct sk_buff *skb,
+				   struct netlink_callback *cb)
+{
+	return res_get_common_dumpit(skb, cb, RDMA_RESTRACK_MR);
+}
+
+static int nldev_res_get_pd_dumpit(struct sk_buff *skb,
+				   struct netlink_callback *cb)
+{
+	return res_get_common_dumpit(skb, cb, RDMA_RESTRACK_PD);
+}
+
 static const struct rdma_nl_cbs nldev_cb_table[RDMA_NLDEV_NUM_OPS] = {
 	[RDMA_NLDEV_CMD_GET] = {
 		.doit = nldev_get_doit,
@@ -714,6 +1002,18 @@ static const struct rdma_nl_cbs nldev_cb_table[RDMA_NLDEV_NUM_OPS] = {
 		 * too.
 		 */
 	},
+	[RDMA_NLDEV_CMD_RES_CM_ID_GET] = {
+		.dump = nldev_res_get_cm_id_dumpit,
+	},
+	[RDMA_NLDEV_CMD_RES_CQ_GET] = {
+		.dump = nldev_res_get_cq_dumpit,
+	},
+	[RDMA_NLDEV_CMD_RES_MR_GET] = {
+		.dump = nldev_res_get_mr_dumpit,
+	},
+	[RDMA_NLDEV_CMD_RES_PD_GET] = {
+		.dump = nldev_res_get_pd_dumpit,
+	},
 };
 
 void __init nldev_init(void)

drivers/infiniband/core/rdma_core.c

@@ -350,13 +350,6 @@ struct ib_uobject *rdma_alloc_begin_uobject(const struct uverbs_obj_type *type,
 	return type->type_class->alloc_begin(type, ucontext);
 }
 
-static void uverbs_uobject_add(struct ib_uobject *uobject)
-{
-	mutex_lock(&uobject->context->uobjects_lock);
-	list_add(&uobject->list, &uobject->context->uobjects);
-	mutex_unlock(&uobject->context->uobjects_lock);
-}
-
 static int __must_check remove_commit_idr_uobject(struct ib_uobject *uobj,
 						  enum rdma_remove_reason why)
 {
@@ -502,7 +495,6 @@ out:
 
 static void alloc_commit_idr_uobject(struct ib_uobject *uobj)
 {
-	uverbs_uobject_add(uobj);
 	spin_lock(&uobj->context->ufile->idr_lock);
 	/*
 	 * We already allocated this IDR with a NULL object, so
@@ -518,7 +510,6 @@ static void alloc_commit_fd_uobject(struct ib_uobject *uobj)
 	struct ib_uobject_file *uobj_file =
 		container_of(uobj, struct ib_uobject_file, uobj);
 
-	uverbs_uobject_add(&uobj_file->uobj);
 	fd_install(uobj_file->uobj.id, uobj->object);
 	/* This shouldn't be used anymore. Use the file object instead */
 	uobj_file->uobj.id = 0;
@@ -545,6 +536,10 @@ int rdma_alloc_commit_uobject(struct ib_uobject *uobj)
 	assert_uverbs_usecnt(uobj, true);
 	atomic_set(&uobj->usecnt, 0);
 
+	mutex_lock(&uobj->context->uobjects_lock);
+	list_add(&uobj->list, &uobj->context->uobjects);
+	mutex_unlock(&uobj->context->uobjects_lock);
+
 	uobj->type->type_class->alloc_commit(uobj);
 	up_read(&uobj->context->cleanup_rwsem);
 

drivers/infiniband/core/restrack.c

@@ -3,20 +3,66 @@
  * Copyright (c) 2017-2018 Mellanox Technologies. All rights reserved.
  */
 
+#include <rdma/rdma_cm.h>
 #include <rdma/ib_verbs.h>
 #include <rdma/restrack.h>
 #include <linux/mutex.h>
 #include <linux/sched/task.h>
 #include <linux/pid_namespace.h>
 
+#include "cma_priv.h"
+
 void rdma_restrack_init(struct rdma_restrack_root *res)
 {
 	init_rwsem(&res->rwsem);
 }
 
+static const char *type2str(enum rdma_restrack_type type)
+{
+	static const char * const names[RDMA_RESTRACK_MAX] = {
+		[RDMA_RESTRACK_PD] = "PD",
+		[RDMA_RESTRACK_CQ] = "CQ",
+		[RDMA_RESTRACK_QP] = "QP",
+		[RDMA_RESTRACK_CM_ID] = "CM_ID",
+		[RDMA_RESTRACK_MR] = "MR",
+	};
+
+	return names[type];
+};
+
 void rdma_restrack_clean(struct rdma_restrack_root *res)
 {
-	WARN_ON_ONCE(!hash_empty(res->hash));
+	struct rdma_restrack_entry *e;
+	char buf[TASK_COMM_LEN];
+	struct ib_device *dev;
+	const char *owner;
+	int bkt;
+
+	if (hash_empty(res->hash))
+		return;
+
+	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);
+	hash_for_each(res->hash, bkt, e, node) {
+		if (rdma_is_kernel_res(e)) {
+			owner = e->kern_name;
+		} else {
+			/*
+			 * There is no need to call get_task_struct here,
+			 * because we can be here only if there are more
+			 * get_task_struct() call than put_task_struct().
+			 */
+			get_task_comm(buf, e->task);
+			owner = buf;
+		}
+
+		pr_err("restrack: %s %s object allocated by %s is not freed\n",
+		       rdma_is_kernel_res(e) ? "Kernel" : "User",
+		       type2str(e->type), owner);
+	}
+	pr_err("restrack: %s", CUT_HERE);
 }
 
 int rdma_restrack_count(struct rdma_restrack_root *res,
@@ -40,51 +86,48 @@ EXPORT_SYMBOL(rdma_restrack_count);
 
 static void set_kern_name(struct rdma_restrack_entry *res)
 {
-	enum rdma_restrack_type type = res->type;
-	struct ib_qp *qp;
-
-	if (type != RDMA_RESTRACK_QP)
-		/* PD and CQ types already have this name embedded in */
-		return;
+	struct ib_pd *pd;
 
-	qp = container_of(res, struct ib_qp, res);
-	if (!qp->pd) {
-		WARN_ONCE(true, "XRC QPs are not supported\n");
-		/* Survive, despite the programmer's error */
-		res->kern_name = " ";
-		return;
+	switch (res->type) {
+	case RDMA_RESTRACK_QP:
+		pd = container_of(res, struct ib_qp, res)->pd;
+		if (!pd) {
+			WARN_ONCE(true, "XRC QPs are not supported\n");
+			/* Survive, despite the programmer's error */
+			res->kern_name = " ";
+		}
+		break;
+	case RDMA_RESTRACK_MR:
+		pd = container_of(res, struct ib_mr, res)->pd;
+		break;
+	default:
+		/* Other types set kern_name directly */
+		pd = NULL;
+		break;
 	}
 
-	res->kern_name = qp->pd->res.kern_name;
+	if (pd)
+		res->kern_name = pd->res.kern_name;
 }
 
 static struct ib_device *res_to_dev(struct rdma_restrack_entry *res)
 {
-	enum rdma_restrack_type type = res->type;
-	struct ib_device *dev;
-	struct ib_pd *pd;
-	struct ib_cq *cq;
-	struct ib_qp *qp;
-
-	switch (type) {
+	switch (res->type) {
 	case RDMA_RESTRACK_PD:
-		pd = container_of(res, struct ib_pd, res);
-		dev = pd->device;
-		break;
+		return container_of(res, struct ib_pd, res)->device;
 	case RDMA_RESTRACK_CQ:
-		cq = container_of(res, struct ib_cq, res);
-		dev = cq->device;
-		break;
+		return container_of(res, struct ib_cq, res)->device;
 	case RDMA_RESTRACK_QP:
-		qp = container_of(res, struct ib_qp, res);
-		dev = qp->device;
-		break;
+		return container_of(res, struct ib_qp, res)->device;
+	case RDMA_RESTRACK_CM_ID:
+		return container_of(res, struct rdma_id_private,
+				    res)->id.device;
+	case RDMA_RESTRACK_MR:
+		return container_of(res, struct ib_mr, res)->device;
 	default:
-		WARN_ONCE(true, "Wrong resource tracking type %u\n", type);
+		WARN_ONCE(true, "Wrong resource tracking type %u\n", res->type);
 		return NULL;
 	}
-
-	return dev;
 }
 
 static bool res_is_user(struct rdma_restrack_entry *res)
@@ -96,6 +139,10 @@ static bool res_is_user(struct rdma_restrack_entry *res)
 		return container_of(res, struct ib_cq, res)->uobject;
 	case RDMA_RESTRACK_QP:
 		return container_of(res, struct ib_qp, res)->uobject;
+	case RDMA_RESTRACK_CM_ID:
+		return !res->kern_name;
+	case RDMA_RESTRACK_MR:
+		return container_of(res, struct ib_mr, res)->pd->uobject;
 	default:
 		WARN_ONCE(true, "Wrong resource tracking type %u\n", res->type);
 		return false;
@@ -109,13 +156,15 @@ void rdma_restrack_add(struct rdma_restrack_entry *res)
 	if (!dev)
 		return;
 
+	if (res->type != RDMA_RESTRACK_CM_ID || !res_is_user(res))
+		res->task = NULL;
+
 	if (res_is_user(res)) {
-		get_task_struct(current);
-		res->task = current;
+		if (!res->task)
+			rdma_restrack_set_task(res, current);
 		res->kern_name = NULL;
 	} else {
 		set_kern_name(res);
-		res->task = NULL;
 	}
 
 	kref_init(&res->kref);

drivers/infiniband/core/sa_query.c

@@ -1227,118 +1227,130 @@ static u8 get_src_path_mask(struct ib_device *device, u8 port_num)
 	return src_path_mask;
 }
 
-int ib_init_ah_attr_from_path(struct ib_device *device, u8 port_num,
-			      struct sa_path_rec *rec,
-			      struct rdma_ah_attr *ah_attr)
+static int
+roce_resolve_route_from_path(struct ib_device *device, u8 port_num,
+			     struct sa_path_rec *rec)
 {
+	struct net_device *resolved_dev;
+	struct net_device *ndev;
+	struct net_device *idev;
+	struct rdma_dev_addr dev_addr = {
+		.bound_dev_if = ((sa_path_get_ifindex(rec) >= 0) ?
+				 sa_path_get_ifindex(rec) : 0),
+		.net = sa_path_get_ndev(rec) ?
+			sa_path_get_ndev(rec) :
+			&init_net
+	};
+	union {
+		struct sockaddr     _sockaddr;
+		struct sockaddr_in  _sockaddr_in;
+		struct sockaddr_in6 _sockaddr_in6;
+	} sgid_addr, dgid_addr;
 	int ret;
-	u16 gid_index;
-	int use_roce;
-	struct net_device *ndev = NULL;
 
-	memset(ah_attr, 0, sizeof *ah_attr);
-	ah_attr->type = rdma_ah_find_type(device, port_num);
+	if (rec->roce.route_resolved)
+		return 0;
 
-	rdma_ah_set_dlid(ah_attr, be32_to_cpu(sa_path_get_dlid(rec)));
+	if (!device->get_netdev)
+		return -EOPNOTSUPP;
 
-	if ((ah_attr->type == RDMA_AH_ATTR_TYPE_OPA) &&
-	    (rdma_ah_get_dlid(ah_attr) == be16_to_cpu(IB_LID_PERMISSIVE)))
-		rdma_ah_set_make_grd(ah_attr, true);
+	rdma_gid2ip(&sgid_addr._sockaddr, &rec->sgid);
+	rdma_gid2ip(&dgid_addr._sockaddr, &rec->dgid);
 
-	rdma_ah_set_sl(ah_attr, rec->sl);
-	rdma_ah_set_path_bits(ah_attr, be32_to_cpu(sa_path_get_slid(rec)) &
-			      get_src_path_mask(device, port_num));
-	rdma_ah_set_port_num(ah_attr, port_num);
-	rdma_ah_set_static_rate(ah_attr, rec->rate);
-	use_roce = rdma_cap_eth_ah(device, port_num);
-
-	if (use_roce) {
-		struct net_device *idev;
-		struct net_device *resolved_dev;
-		struct rdma_dev_addr dev_addr = {
-			.bound_dev_if = ((sa_path_get_ifindex(rec) >= 0) ?
-					 sa_path_get_ifindex(rec) : 0),
-			.net = sa_path_get_ndev(rec) ?
-				sa_path_get_ndev(rec) :
-				&init_net
-		};
-		union {
-			struct sockaddr     _sockaddr;
-			struct sockaddr_in  _sockaddr_in;
-			struct sockaddr_in6 _sockaddr_in6;
-		} sgid_addr, dgid_addr;
-
-		if (!device->get_netdev)
-			return -EOPNOTSUPP;
-
-		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;
+	/* 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;
+	if ((dev_addr.network == RDMA_NETWORK_IPV4 ||
+	     dev_addr.network == RDMA_NETWORK_IPV6) &&
+	    rec->rec_type != SA_PATH_REC_TYPE_ROCE_V2)
+		return -EINVAL;
 
-		idev = device->get_netdev(device, port_num);
-		if (!idev)
-			return -ENODEV;
+	idev = device->get_netdev(device, port_num);
+	if (!idev)
+		return -ENODEV;
 
-		resolved_dev = dev_get_by_index(dev_addr.net,
-						dev_addr.bound_dev_if);
-		if (!resolved_dev) {
-			dev_put(idev);
-			return -ENODEV;
-		}
-		ndev = ib_get_ndev_from_path(rec);
-		rcu_read_lock();
-		if ((ndev && ndev != resolved_dev) ||
-		    (resolved_dev != idev &&
-		     !rdma_is_upper_dev_rcu(idev, resolved_dev)))
-			ret = -EHOSTUNREACH;
-		rcu_read_unlock();
-		dev_put(idev);
-		dev_put(resolved_dev);
-		if (ret) {
-			if (ndev)
-				dev_put(ndev);
-			return ret;
-		}
+	resolved_dev = dev_get_by_index(dev_addr.net,
+					dev_addr.bound_dev_if);
+	if (!resolved_dev) {
+		ret = -ENODEV;
+		goto done;
 	}
+	ndev = ib_get_ndev_from_path(rec);
+	rcu_read_lock();
+	if ((ndev && ndev != resolved_dev) ||
+	    (resolved_dev != idev &&
+	     !rdma_is_upper_dev_rcu(idev, resolved_dev)))
+		ret = -EHOSTUNREACH;
+	rcu_read_unlock();
+	dev_put(resolved_dev);
+	if (ndev)
+		dev_put(ndev);
+done:
+	dev_put(idev);
+	if (!ret)
+		rec->roce.route_resolved = true;
+	return ret;
+}
 
-	if (rec->hop_limit > 0 || use_roce) {
-		enum ib_gid_type type = sa_conv_pathrec_to_gid_type(rec);
+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)
+{
+	enum ib_gid_type type = sa_conv_pathrec_to_gid_type(rec);
+	struct net_device *ndev;
+	u16 gid_index;
+	int ret;
 
-		ret = ib_find_cached_gid_by_port(device, &rec->sgid, type,
-						 port_num, ndev, &gid_index);
-		if (ret) {
-			if (ndev)
-				dev_put(ndev);
-			return ret;
-		}
+	ndev = ib_get_ndev_from_path(rec);
+	ret = ib_find_cached_gid_by_port(device, &rec->sgid, type,
+					 port_num, ndev, &gid_index);
+	if (ndev)
+		dev_put(ndev);
+	if (ret)
+		return ret;
 
-		rdma_ah_set_grh(ah_attr, &rec->dgid,
-				be32_to_cpu(rec->flow_label),
-				gid_index, rec->hop_limit,
-				rec->traffic_class);
-		if (ndev)
-			dev_put(ndev);
-	}
+	rdma_ah_set_grh(ah_attr, &rec->dgid,
+			be32_to_cpu(rec->flow_label),
+			gid_index, rec->hop_limit,
+			rec->traffic_class);
+	return 0;
+}
 
-	if (use_roce) {
-		u8 *dmac = sa_path_get_dmac(rec);
+int ib_init_ah_attr_from_path(struct ib_device *device, u8 port_num,
+			      struct sa_path_rec *rec,
+			      struct rdma_ah_attr *ah_attr)
+{
+	int ret = 0;
+
+	memset(ah_attr, 0, sizeof(*ah_attr));
+	ah_attr->type = rdma_ah_find_type(device, port_num);
+	rdma_ah_set_sl(ah_attr, rec->sl);
+	rdma_ah_set_port_num(ah_attr, port_num);
+	rdma_ah_set_static_rate(ah_attr, rec->rate);
 
-		if (!dmac)
-			return -EINVAL;
-		memcpy(ah_attr->roce.dmac, dmac, ETH_ALEN);
+	if (sa_path_is_roce(rec)) {
+		ret = roce_resolve_route_from_path(device, port_num, rec);
+		if (ret)
+			return ret;
+
+		memcpy(ah_attr->roce.dmac, sa_path_get_dmac(rec), ETH_ALEN);
+	} else {
+		rdma_ah_set_dlid(ah_attr, be32_to_cpu(sa_path_get_dlid(rec)));
+		if (sa_path_is_opa(rec) &&
+		    rdma_ah_get_dlid(ah_attr) == be16_to_cpu(IB_LID_PERMISSIVE))
+			rdma_ah_set_make_grd(ah_attr, true);
+
+		rdma_ah_set_path_bits(ah_attr,
+				      be32_to_cpu(sa_path_get_slid(rec)) &
+				      get_src_path_mask(device, port_num));
 	}
 
-	return 0;
+	if (rec->hop_limit > 0 || sa_path_is_roce(rec))
+		ret = init_ah_attr_grh_fields(device, port_num, rec, ah_attr);
+	return ret;
 }
 EXPORT_SYMBOL(ib_init_ah_attr_from_path);
 

drivers/infiniband/core/sysfs.c

@@ -273,6 +273,7 @@ static ssize_t rate_show(struct ib_port *p, struct port_attribute *unused,
 		break;
 	case IB_SPEED_SDR:
 	default:		/* default to SDR for invalid rates */
+		speed = " SDR";
 		rate = 25;
 		break;
 	}
@@ -388,14 +389,26 @@ static ssize_t show_port_gid(struct ib_port *p, struct port_attribute *attr,
 {
 	struct port_table_attribute *tab_attr =
 		container_of(attr, struct port_table_attribute, attr);
+	union ib_gid *pgid;
 	union ib_gid gid;
 	ssize_t ret;
 
 	ret = ib_query_gid(p->ibdev, p->port_num, tab_attr->index, &gid, NULL);
-	if (ret)
-		return ret;
 
-	return sprintf(buf, "%pI6\n", gid.raw);
+	/* If reading GID fails, it is likely due to GID entry being empty
+	 * (invalid) or reserved GID in the table.
+	 * User space expects to read GID table entries as long as it given
+	 * index is within GID table size.
+	 * Administrative/debugging tool fails to query rest of the GID entries
+	 * if it hits error while querying a GID of the given index.
+	 * To avoid user space throwing such error on fail to read gid, return
+	 * zero GID as before. This maintains backward compatibility.
+	 */
+	if (ret)
+		pgid = &zgid;
+	else
+		pgid = &gid;
+	return sprintf(buf, "%pI6\n", pgid->raw);
 }
 
 static ssize_t show_port_gid_attr_ndev(struct ib_port *p,
@@ -810,10 +823,15 @@ static ssize_t show_hw_stats(struct kobject *kobj, struct attribute *attr,
 		dev = port->ibdev;
 		stats = port->hw_stats;
 	}
+	mutex_lock(&stats->lock);
 	ret = update_hw_stats(dev, stats, hsa->port_num, hsa->index);
 	if (ret)
-		return ret;
-	return print_hw_stat(stats, hsa->index, buf);
+		goto unlock;
+	ret = print_hw_stat(stats, hsa->index, buf);
+unlock:
+	mutex_unlock(&stats->lock);
+
+	return ret;
 }
 
 static ssize_t show_stats_lifespan(struct kobject *kobj,
@@ -821,17 +839,25 @@ static ssize_t show_stats_lifespan(struct kobject *kobj,
 				   char *buf)
 {
 	struct hw_stats_attribute *hsa;
+	struct rdma_hw_stats *stats;
 	int msecs;
 
 	hsa = container_of(attr, struct hw_stats_attribute, attr);
 	if (!hsa->port_num) {
 		struct ib_device *dev = container_of((struct device *)kobj,
 						     struct ib_device, dev);
-		msecs = jiffies_to_msecs(dev->hw_stats->lifespan);
+
+		stats = dev->hw_stats;
 	} else {
 		struct ib_port *p = container_of(kobj, struct ib_port, kobj);
-		msecs = jiffies_to_msecs(p->hw_stats->lifespan);
+
+		stats = p->hw_stats;
 	}
+
+	mutex_lock(&stats->lock);
+	msecs = jiffies_to_msecs(stats->lifespan);
+	mutex_unlock(&stats->lock);
+
 	return sprintf(buf, "%d\n", msecs);
 }
 
@@ -840,6 +866,7 @@ static ssize_t set_stats_lifespan(struct kobject *kobj,
 				  const char *buf, size_t count)
 {
 	struct hw_stats_attribute *hsa;
+	struct rdma_hw_stats *stats;
 	int msecs;
 	int jiffies;
 	int ret;
@@ -854,11 +881,18 @@ static ssize_t set_stats_lifespan(struct kobject *kobj,
 	if (!hsa->port_num) {
 		struct ib_device *dev = container_of((struct device *)kobj,
 						     struct ib_device, dev);
-		dev->hw_stats->lifespan = jiffies;
+
+		stats = dev->hw_stats;
 	} else {
 		struct ib_port *p = container_of(kobj, struct ib_port, kobj);
-		p->hw_stats->lifespan = jiffies;
+
+		stats = p->hw_stats;
 	}
+
+	mutex_lock(&stats->lock);
+	stats->lifespan = jiffies;
+	mutex_unlock(&stats->lock);
+
 	return count;
 }
 
@@ -951,6 +985,7 @@ static void setup_hw_stats(struct ib_device *device, struct ib_port *port,
 		sysfs_attr_init(hsag->attrs[i]);
 	}
 
+	mutex_init(&stats->lock);
 	/* treat an error here as non-fatal */
 	hsag->attrs[i] = alloc_hsa_lifespan("lifespan", port_num);
 	if (hsag->attrs[i])

drivers/infiniband/core/ucm.c

@@ -430,7 +430,7 @@ static ssize_t ib_ucm_event(struct ib_ucm_file *file,
 		uevent->resp.id = ctx->id;
 	}
 
-	if (copy_to_user((void __user *)(unsigned long)cmd.response,
+	if (copy_to_user(u64_to_user_ptr(cmd.response),
 			 &uevent->resp, sizeof(uevent->resp))) {
 		result = -EFAULT;
 		goto done;
@@ -441,7 +441,7 @@ static ssize_t ib_ucm_event(struct ib_ucm_file *file,
 			result = -ENOMEM;
 			goto done;
 		}
-		if (copy_to_user((void __user *)(unsigned long)cmd.data,
+		if (copy_to_user(u64_to_user_ptr(cmd.data),
 				 uevent->data, uevent->data_len)) {
 			result = -EFAULT;
 			goto done;
@@ -453,7 +453,7 @@ static ssize_t ib_ucm_event(struct ib_ucm_file *file,
 			result = -ENOMEM;
 			goto done;
 		}
-		if (copy_to_user((void __user *)(unsigned long)cmd.info,
+		if (copy_to_user(u64_to_user_ptr(cmd.info),
 				 uevent->info, uevent->info_len)) {
 			result = -EFAULT;
 			goto done;
@@ -502,7 +502,7 @@ static ssize_t ib_ucm_create_id(struct ib_ucm_file *file,
 	}
 
 	resp.id = ctx->id;
-	if (copy_to_user((void __user *)(unsigned long)cmd.response,
+	if (copy_to_user(u64_to_user_ptr(cmd.response),
 			 &resp, sizeof(resp))) {
 		result = -EFAULT;
 		goto err2;
@@ -556,7 +556,7 @@ static ssize_t ib_ucm_destroy_id(struct ib_ucm_file *file,
 	ib_ucm_cleanup_events(ctx);
 
 	resp.events_reported = ctx->events_reported;
-	if (copy_to_user((void __user *)(unsigned long)cmd.response,
+	if (copy_to_user(u64_to_user_ptr(cmd.response),
 			 &resp, sizeof(resp)))
 		result = -EFAULT;
 
@@ -588,7 +588,7 @@ static ssize_t ib_ucm_attr_id(struct ib_ucm_file *file,
 	resp.local_id     = ctx->cm_id->local_id;
 	resp.remote_id    = ctx->cm_id->remote_id;
 
-	if (copy_to_user((void __user *)(unsigned long)cmd.response,
+	if (copy_to_user(u64_to_user_ptr(cmd.response),
 			 &resp, sizeof(resp)))
 		result = -EFAULT;
 
@@ -625,7 +625,7 @@ static ssize_t ib_ucm_init_qp_attr(struct ib_ucm_file *file,
 
 	ib_copy_qp_attr_to_user(ctx->cm_id->device, &resp, &qp_attr);
 
-	if (copy_to_user((void __user *)(unsigned long)cmd.response,
+	if (copy_to_user(u64_to_user_ptr(cmd.response),
 			 &resp, sizeof(resp)))
 		result = -EFAULT;
 
@@ -699,7 +699,7 @@ static int ib_ucm_alloc_data(const void **dest, u64 src, u32 len)
 	if (!len)
 		return 0;
 
-	data = memdup_user((void __user *)(unsigned long)src, len);
+	data = memdup_user(u64_to_user_ptr(src), len);
 	if (IS_ERR(data))
 		return PTR_ERR(data);
 
@@ -721,7 +721,7 @@ static int ib_ucm_path_get(struct sa_path_rec **path, u64 src)
 	if (!sa_path)
 		return -ENOMEM;
 
-	if (copy_from_user(&upath, (void __user *)(unsigned long)src,
+	if (copy_from_user(&upath, u64_to_user_ptr(src),
 			   sizeof(upath))) {
 
 		kfree(sa_path);

drivers/infiniband/core/ucma.c

@@ -382,7 +382,11 @@ static ssize_t ucma_get_event(struct ucma_file *file, const char __user *inbuf,
 	struct ucma_event *uevent;
 	int ret = 0;
 
-	if (out_len < sizeof uevent->resp)
+	/*
+	 * Old 32 bit user space does not send the 4 byte padding in the
+	 * reserved field. We don't care, allow it to keep working.
+	 */
+	if (out_len < sizeof(uevent->resp) - sizeof(uevent->resp.reserved))
 		return -ENOSPC;
 
 	if (copy_from_user(&cmd, inbuf, sizeof(cmd)))
@@ -416,8 +420,9 @@ static ssize_t ucma_get_event(struct ucma_file *file, const char __user *inbuf,
 		uevent->resp.id = ctx->id;
 	}
 
-	if (copy_to_user((void __user *)(unsigned long)cmd.response,
-			 &uevent->resp, sizeof uevent->resp)) {
+	if (copy_to_user(u64_to_user_ptr(cmd.response),
+			 &uevent->resp,
+			 min_t(size_t, out_len, sizeof(uevent->resp)))) {
 		ret = -EFAULT;
 		goto done;
 	}
@@ -477,15 +482,15 @@ static ssize_t ucma_create_id(struct ucma_file *file, const char __user *inbuf,
 		return -ENOMEM;
 
 	ctx->uid = cmd.uid;
-	cm_id = rdma_create_id(current->nsproxy->net_ns,
-			       ucma_event_handler, ctx, cmd.ps, qp_type);
+	cm_id = __rdma_create_id(current->nsproxy->net_ns,
+				 ucma_event_handler, ctx, cmd.ps, qp_type, NULL);
 	if (IS_ERR(cm_id)) {
 		ret = PTR_ERR(cm_id);
 		goto err1;
 	}
 
 	resp.id = ctx->id;
-	if (copy_to_user((void __user *)(unsigned long)cmd.response,
+	if (copy_to_user(u64_to_user_ptr(cmd.response),
 			 &resp, sizeof(resp))) {
 		ret = -EFAULT;
 		goto err2;
@@ -615,7 +620,7 @@ static ssize_t ucma_destroy_id(struct ucma_file *file, const char __user *inbuf,
 	}
 
 	resp.events_reported = ucma_free_ctx(ctx);
-	if (copy_to_user((void __user *)(unsigned long)cmd.response,
+	if (copy_to_user(u64_to_user_ptr(cmd.response),
 			 &resp, sizeof(resp)))
 		ret = -EFAULT;
 
@@ -845,7 +850,7 @@ static ssize_t ucma_query_route(struct ucma_file *file,
 		ucma_copy_iw_route(&resp, &ctx->cm_id->route);
 
 out:
-	if (copy_to_user((void __user *)(unsigned long)cmd.response,
+	if (copy_to_user(u64_to_user_ptr(cmd.response),
 			 &resp, sizeof(resp)))
 		ret = -EFAULT;
 
@@ -991,7 +996,7 @@ static ssize_t ucma_query(struct ucma_file *file,
 	if (copy_from_user(&cmd, inbuf, sizeof(cmd)))
 		return -EFAULT;
 
-	response = (void __user *)(unsigned long) cmd.response;
+	response = u64_to_user_ptr(cmd.response);
 	ctx = ucma_get_ctx(file, cmd.id);
 	if (IS_ERR(ctx))
 		return PTR_ERR(ctx);
@@ -1094,12 +1099,12 @@ static ssize_t ucma_accept(struct ucma_file *file, const char __user *inbuf,
 	if (cmd.conn_param.valid) {
 		ucma_copy_conn_param(ctx->cm_id, &conn_param, &cmd.conn_param);
 		mutex_lock(&file->mut);
-		ret = rdma_accept(ctx->cm_id, &conn_param);
+		ret = __rdma_accept(ctx->cm_id, &conn_param, NULL);
 		if (!ret)
 			ctx->uid = cmd.uid;
 		mutex_unlock(&file->mut);
 	} else
-		ret = rdma_accept(ctx->cm_id, NULL);
+		ret = __rdma_accept(ctx->cm_id, NULL, NULL);
 
 	ucma_put_ctx(ctx);
 	return ret;
@@ -1179,7 +1184,7 @@ static ssize_t ucma_init_qp_attr(struct ucma_file *file,
 		goto out;
 
 	ib_copy_qp_attr_to_user(ctx->cm_id->device, &resp, &qp_attr);
-	if (copy_to_user((void __user *)(unsigned long)cmd.response,
+	if (copy_to_user(u64_to_user_ptr(cmd.response),
 			 &resp, sizeof(resp)))
 		ret = -EFAULT;
 
@@ -1241,6 +1246,9 @@ static int ucma_set_ib_path(struct ucma_context *ctx,
 	if (!optlen)
 		return -EINVAL;
 
+	if (!ctx->cm_id->device)
+		return -EINVAL;
+
 	memset(&sa_path, 0, sizeof(sa_path));
 
 	sa_path.rec_type = SA_PATH_REC_TYPE_IB;
@@ -1315,7 +1323,7 @@ static ssize_t ucma_set_option(struct ucma_file *file, const char __user *inbuf,
 	if (unlikely(cmd.optlen > KMALLOC_MAX_SIZE))
 		return -EINVAL;
 
-	optval = memdup_user((void __user *) (unsigned long) cmd.optval,
+	optval = memdup_user(u64_to_user_ptr(cmd.optval),
 			     cmd.optlen);
 	if (IS_ERR(optval)) {
 		ret = PTR_ERR(optval);
@@ -1395,7 +1403,7 @@ static ssize_t ucma_process_join(struct ucma_file *file,
 		goto err2;
 
 	resp.id = mc->id;
-	if (copy_to_user((void __user *)(unsigned long) cmd->response,
+	if (copy_to_user(u64_to_user_ptr(cmd->response),
 			 &resp, sizeof(resp))) {
 		ret = -EFAULT;
 		goto err3;
@@ -1500,7 +1508,7 @@ static ssize_t ucma_leave_multicast(struct ucma_file *file,
 	resp.events_reported = mc->events_reported;
 	kfree(mc);
 
-	if (copy_to_user((void __user *)(unsigned long)cmd.response,
+	if (copy_to_user(u64_to_user_ptr(cmd.response),
 			 &resp, sizeof(resp)))
 		ret = -EFAULT;
 out:
@@ -1587,7 +1595,7 @@ static ssize_t ucma_migrate_id(struct ucma_file *new_file,
 	ucma_unlock_files(cur_file, new_file);
 
 response:
-	if (copy_to_user((void __user *)(unsigned long)cmd.response,
+	if (copy_to_user(u64_to_user_ptr(cmd.response),
 			 &resp, sizeof(resp)))
 		ret = -EFAULT;
 

drivers/infiniband/core/uverbs.h

@@ -46,6 +46,10 @@
 #include <rdma/ib_verbs.h>
 #include <rdma/ib_umem.h>
 #include <rdma/ib_user_verbs.h>
+#include <rdma/uverbs_std_types.h>
+
+#define UVERBS_MODULE_NAME ib_uverbs
+#include <rdma/uverbs_named_ioctl.h>
 
 static inline void
 ib_uverbs_init_udata(struct ib_udata *udata,
@@ -199,11 +203,18 @@ 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,
 					      struct ib_device *ib_dev);
 void ib_uverbs_free_async_event_file(struct ib_uverbs_file *uverbs_file);
+void ib_uverbs_flow_resources_free(struct ib_uflow_resources *uflow_res);
 
 void ib_uverbs_release_ucq(struct ib_uverbs_file *file,
 			   struct ib_uverbs_completion_event_file *ev_file,
@@ -226,7 +237,13 @@ int uverbs_dealloc_mw(struct ib_mw *mw);
 void ib_uverbs_detach_umcast(struct ib_qp *qp,
 			     struct ib_uqp_object *uobj);
 
+void create_udata(struct uverbs_attr_bundle *ctx, struct ib_udata *udata);
+extern const struct uverbs_attr_def uverbs_uhw_compat_in;
+extern const struct uverbs_attr_def uverbs_uhw_compat_out;
 long ib_uverbs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg);
+int uverbs_destroy_def_handler(struct ib_device *ib_dev,
+			       struct ib_uverbs_file *file,
+			       struct uverbs_attr_bundle *attrs);
 
 struct ib_uverbs_flow_spec {
 	union {
@@ -240,13 +257,37 @@ struct ib_uverbs_flow_spec {
 		};
 		struct ib_uverbs_flow_spec_eth     eth;
 		struct ib_uverbs_flow_spec_ipv4    ipv4;
+		struct ib_uverbs_flow_spec_esp     esp;
 		struct ib_uverbs_flow_spec_tcp_udp tcp_udp;
 		struct ib_uverbs_flow_spec_ipv6    ipv6;
 		struct ib_uverbs_flow_spec_action_tag	flow_tag;
 		struct ib_uverbs_flow_spec_action_drop	drop;
+		struct ib_uverbs_flow_spec_action_handle action;
 	};
 };
 
+int ib_uverbs_kern_spec_to_ib_spec_filter(enum ib_flow_spec_type type,
+					  const void *kern_spec_mask,
+					  const void *kern_spec_val,
+					  size_t kern_filter_sz,
+					  union ib_flow_spec *ib_spec);
+
+extern const struct uverbs_object_def UVERBS_OBJECT(UVERBS_OBJECT_DEVICE);
+extern const struct uverbs_object_def UVERBS_OBJECT(UVERBS_OBJECT_PD);
+extern const struct uverbs_object_def UVERBS_OBJECT(UVERBS_OBJECT_MR);
+extern const struct uverbs_object_def UVERBS_OBJECT(UVERBS_OBJECT_COMP_CHANNEL);
+extern const struct uverbs_object_def UVERBS_OBJECT(UVERBS_OBJECT_CQ);
+extern const struct uverbs_object_def UVERBS_OBJECT(UVERBS_OBJECT_QP);
+extern const struct uverbs_object_def UVERBS_OBJECT(UVERBS_OBJECT_AH);
+extern const struct uverbs_object_def UVERBS_OBJECT(UVERBS_OBJECT_MW);
+extern const struct uverbs_object_def UVERBS_OBJECT(UVERBS_OBJECT_SRQ);
+extern const struct uverbs_object_def UVERBS_OBJECT(UVERBS_OBJECT_FLOW);
+extern const struct uverbs_object_def UVERBS_OBJECT(UVERBS_OBJECT_WQ);
+extern const struct uverbs_object_def UVERBS_OBJECT(UVERBS_OBJECT_RWQ_IND_TBL);
+extern const struct uverbs_object_def UVERBS_OBJECT(UVERBS_OBJECT_XRCD);
+extern const struct uverbs_object_def UVERBS_OBJECT(UVERBS_OBJECT_FLOW_ACTION);
+extern const struct uverbs_object_def UVERBS_OBJECT(UVERBS_OBJECT_DM);
+
 #define IB_UVERBS_DECLARE_CMD(name)					\
 	ssize_t ib_uverbs_##name(struct ib_uverbs_file *file,		\
 				 struct ib_device *ib_dev,              \

drivers/infiniband/core/uverbs_cmd.c

@@ -50,7 +50,7 @@
 static struct ib_uverbs_completion_event_file *
 ib_uverbs_lookup_comp_file(int fd, struct ib_ucontext *context)
 {
-	struct ib_uobject *uobj = uobj_get_read(uobj_get_type(comp_channel),
+	struct ib_uobject *uobj = uobj_get_read(UVERBS_OBJECT_COMP_CHANNEL,
 						fd, context);
 	struct ib_uobject_file *uobj_file;
 
@@ -322,7 +322,7 @@ ssize_t ib_uverbs_alloc_pd(struct ib_uverbs_file *file,
                    in_len - sizeof(cmd) - sizeof(struct ib_uverbs_cmd_hdr),
                    out_len - sizeof(resp));
 
-	uobj  = uobj_alloc(uobj_get_type(pd), file->ucontext);
+	uobj  = uobj_alloc(UVERBS_OBJECT_PD, file->ucontext);
 	if (IS_ERR(uobj))
 		return PTR_ERR(uobj);
 
@@ -372,7 +372,7 @@ ssize_t ib_uverbs_dealloc_pd(struct ib_uverbs_file *file,
 	if (copy_from_user(&cmd, buf, sizeof cmd))
 		return -EFAULT;
 
-	uobj  = uobj_get_write(uobj_get_type(pd), cmd.pd_handle,
+	uobj  = uobj_get_write(UVERBS_OBJECT_PD, cmd.pd_handle,
 			       file->ucontext);
 	if (IS_ERR(uobj))
 		return PTR_ERR(uobj);
@@ -517,7 +517,7 @@ ssize_t ib_uverbs_open_xrcd(struct ib_uverbs_file *file,
 		}
 	}
 
-	obj  = (struct ib_uxrcd_object *)uobj_alloc(uobj_get_type(xrcd),
+	obj  = (struct ib_uxrcd_object *)uobj_alloc(UVERBS_OBJECT_XRCD,
 						    file->ucontext);
 	if (IS_ERR(obj)) {
 		ret = PTR_ERR(obj);
@@ -602,7 +602,7 @@ ssize_t ib_uverbs_close_xrcd(struct ib_uverbs_file *file,
 	if (copy_from_user(&cmd, buf, sizeof cmd))
 		return -EFAULT;
 
-	uobj  = uobj_get_write(uobj_get_type(xrcd), cmd.xrcd_handle,
+	uobj  = uobj_get_write(UVERBS_OBJECT_XRCD, cmd.xrcd_handle,
 			       file->ucontext);
 	if (IS_ERR(uobj))
 		return PTR_ERR(uobj);
@@ -663,11 +663,11 @@ ssize_t ib_uverbs_reg_mr(struct ib_uverbs_file *file,
 	if (ret)
 		return ret;
 
-	uobj  = uobj_alloc(uobj_get_type(mr), file->ucontext);
+	uobj  = uobj_alloc(UVERBS_OBJECT_MR, file->ucontext);
 	if (IS_ERR(uobj))
 		return PTR_ERR(uobj);
 
-	pd = uobj_get_obj_read(pd, cmd.pd_handle, file->ucontext);
+	pd = uobj_get_obj_read(pd, UVERBS_OBJECT_PD, cmd.pd_handle, file->ucontext);
 	if (!pd) {
 		ret = -EINVAL;
 		goto err_free;
@@ -693,6 +693,8 @@ ssize_t ib_uverbs_reg_mr(struct ib_uverbs_file *file,
 	mr->pd      = pd;
 	mr->uobject = uobj;
 	atomic_inc(&pd->usecnt);
+	mr->res.type = RDMA_RESTRACK_MR;
+	rdma_restrack_add(&mr->res);
 
 	uobj->object = mr;
 
@@ -756,7 +758,7 @@ ssize_t ib_uverbs_rereg_mr(struct ib_uverbs_file *file,
 	     (cmd.start & ~PAGE_MASK) != (cmd.hca_va & ~PAGE_MASK)))
 			return -EINVAL;
 
-	uobj  = uobj_get_write(uobj_get_type(mr), cmd.mr_handle,
+	uobj  = uobj_get_write(UVERBS_OBJECT_MR, cmd.mr_handle,
 			       file->ucontext);
 	if (IS_ERR(uobj))
 		return PTR_ERR(uobj);
@@ -770,7 +772,7 @@ ssize_t ib_uverbs_rereg_mr(struct ib_uverbs_file *file,
 	}
 
 	if (cmd.flags & IB_MR_REREG_PD) {
-		pd = uobj_get_obj_read(pd, cmd.pd_handle, file->ucontext);
+		pd = uobj_get_obj_read(pd, UVERBS_OBJECT_PD, cmd.pd_handle, file->ucontext);
 		if (!pd) {
 			ret = -EINVAL;
 			goto put_uobjs;
@@ -822,7 +824,7 @@ ssize_t ib_uverbs_dereg_mr(struct ib_uverbs_file *file,
 	if (copy_from_user(&cmd, buf, sizeof cmd))
 		return -EFAULT;
 
-	uobj  = uobj_get_write(uobj_get_type(mr), cmd.mr_handle,
+	uobj  = uobj_get_write(UVERBS_OBJECT_MR, cmd.mr_handle,
 			       file->ucontext);
 	if (IS_ERR(uobj))
 		return PTR_ERR(uobj);
@@ -851,11 +853,11 @@ ssize_t ib_uverbs_alloc_mw(struct ib_uverbs_file *file,
 	if (copy_from_user(&cmd, buf, sizeof(cmd)))
 		return -EFAULT;
 
-	uobj  = uobj_alloc(uobj_get_type(mw), file->ucontext);
+	uobj  = uobj_alloc(UVERBS_OBJECT_MW, file->ucontext);
 	if (IS_ERR(uobj))
 		return PTR_ERR(uobj);
 
-	pd = uobj_get_obj_read(pd, cmd.pd_handle, file->ucontext);
+	pd = uobj_get_obj_read(pd, UVERBS_OBJECT_PD, cmd.pd_handle, file->ucontext);
 	if (!pd) {
 		ret = -EINVAL;
 		goto err_free;
@@ -914,7 +916,7 @@ ssize_t ib_uverbs_dealloc_mw(struct ib_uverbs_file *file,
 	if (copy_from_user(&cmd, buf, sizeof(cmd)))
 		return -EFAULT;
 
-	uobj  = uobj_get_write(uobj_get_type(mw), cmd.mw_handle,
+	uobj  = uobj_get_write(UVERBS_OBJECT_MW, cmd.mw_handle,
 			       file->ucontext);
 	if (IS_ERR(uobj))
 		return PTR_ERR(uobj);
@@ -939,7 +941,7 @@ ssize_t ib_uverbs_create_comp_channel(struct ib_uverbs_file *file,
 	if (copy_from_user(&cmd, buf, sizeof cmd))
 		return -EFAULT;
 
-	uobj = uobj_alloc(uobj_get_type(comp_channel), file->ucontext);
+	uobj = uobj_alloc(UVERBS_OBJECT_COMP_CHANNEL, file->ucontext);
 	if (IS_ERR(uobj))
 		return PTR_ERR(uobj);
 
@@ -984,7 +986,7 @@ static struct ib_ucq_object *create_cq(struct ib_uverbs_file *file,
 	if (cmd->comp_vector >= file->device->num_comp_vectors)
 		return ERR_PTR(-EINVAL);
 
-	obj  = (struct ib_ucq_object *)uobj_alloc(uobj_get_type(cq),
+	obj  = (struct ib_ucq_object *)uobj_alloc(UVERBS_OBJECT_CQ,
 						  file->ucontext);
 	if (IS_ERR(obj))
 		return obj;
@@ -1173,7 +1175,7 @@ ssize_t ib_uverbs_resize_cq(struct ib_uverbs_file *file,
 		   in_len - sizeof(cmd) - sizeof(struct ib_uverbs_cmd_hdr),
 		   out_len - sizeof(resp));
 
-	cq = uobj_get_obj_read(cq, cmd.cq_handle, file->ucontext);
+	cq = uobj_get_obj_read(cq, UVERBS_OBJECT_CQ, cmd.cq_handle, file->ucontext);
 	if (!cq)
 		return -EINVAL;
 
@@ -1238,7 +1240,7 @@ ssize_t ib_uverbs_poll_cq(struct ib_uverbs_file *file,
 	if (copy_from_user(&cmd, buf, sizeof cmd))
 		return -EFAULT;
 
-	cq = uobj_get_obj_read(cq, cmd.cq_handle, file->ucontext);
+	cq = uobj_get_obj_read(cq, UVERBS_OBJECT_CQ, cmd.cq_handle, file->ucontext);
 	if (!cq)
 		return -EINVAL;
 
@@ -1285,7 +1287,7 @@ ssize_t ib_uverbs_req_notify_cq(struct ib_uverbs_file *file,
 	if (copy_from_user(&cmd, buf, sizeof cmd))
 		return -EFAULT;
 
-	cq = uobj_get_obj_read(cq, cmd.cq_handle, file->ucontext);
+	cq = uobj_get_obj_read(cq, UVERBS_OBJECT_CQ, cmd.cq_handle, file->ucontext);
 	if (!cq)
 		return -EINVAL;
 
@@ -1312,7 +1314,7 @@ ssize_t ib_uverbs_destroy_cq(struct ib_uverbs_file *file,
 	if (copy_from_user(&cmd, buf, sizeof cmd))
 		return -EFAULT;
 
-	uobj  = uobj_get_write(uobj_get_type(cq), cmd.cq_handle,
+	uobj  = uobj_get_write(UVERBS_OBJECT_CQ, cmd.cq_handle,
 			       file->ucontext);
 	if (IS_ERR(uobj))
 		return PTR_ERR(uobj);
@@ -1371,7 +1373,7 @@ static int create_qp(struct ib_uverbs_file *file,
 	if (cmd->qp_type == IB_QPT_RAW_PACKET && !capable(CAP_NET_RAW))
 		return -EPERM;
 
-	obj  = (struct ib_uqp_object *)uobj_alloc(uobj_get_type(qp),
+	obj  = (struct ib_uqp_object *)uobj_alloc(UVERBS_OBJECT_QP,
 						  file->ucontext);
 	if (IS_ERR(obj))
 		return PTR_ERR(obj);
@@ -1382,7 +1384,7 @@ static int create_qp(struct ib_uverbs_file *file,
 	if (cmd_sz >= offsetof(typeof(*cmd), rwq_ind_tbl_handle) +
 		      sizeof(cmd->rwq_ind_tbl_handle) &&
 		      (cmd->comp_mask & IB_UVERBS_CREATE_QP_MASK_IND_TABLE)) {
-		ind_tbl = uobj_get_obj_read(rwq_ind_table,
+		ind_tbl = uobj_get_obj_read(rwq_ind_table, UVERBS_OBJECT_RWQ_IND_TBL,
 					    cmd->rwq_ind_tbl_handle,
 					    file->ucontext);
 		if (!ind_tbl) {
@@ -1409,7 +1411,7 @@ static int create_qp(struct ib_uverbs_file *file,
 		has_sq = false;
 
 	if (cmd->qp_type == IB_QPT_XRC_TGT) {
-		xrcd_uobj = uobj_get_read(uobj_get_type(xrcd), cmd->pd_handle,
+		xrcd_uobj = uobj_get_read(UVERBS_OBJECT_XRCD, cmd->pd_handle,
 					  file->ucontext);
 
 		if (IS_ERR(xrcd_uobj)) {
@@ -1429,7 +1431,7 @@ static int create_qp(struct ib_uverbs_file *file,
 			cmd->max_recv_sge = 0;
 		} else {
 			if (cmd->is_srq) {
-				srq = uobj_get_obj_read(srq, cmd->srq_handle,
+				srq = uobj_get_obj_read(srq, UVERBS_OBJECT_SRQ, cmd->srq_handle,
 							file->ucontext);
 				if (!srq || srq->srq_type == IB_SRQT_XRC) {
 					ret = -EINVAL;
@@ -1439,7 +1441,7 @@ static int create_qp(struct ib_uverbs_file *file,
 
 			if (!ind_tbl) {
 				if (cmd->recv_cq_handle != cmd->send_cq_handle) {
-					rcq = uobj_get_obj_read(cq, cmd->recv_cq_handle,
+					rcq = uobj_get_obj_read(cq, UVERBS_OBJECT_CQ, cmd->recv_cq_handle,
 								file->ucontext);
 					if (!rcq) {
 						ret = -EINVAL;
@@ -1450,11 +1452,11 @@ static int create_qp(struct ib_uverbs_file *file,
 		}
 
 		if (has_sq)
-			scq = uobj_get_obj_read(cq, cmd->send_cq_handle,
+			scq = uobj_get_obj_read(cq, UVERBS_OBJECT_CQ, cmd->send_cq_handle,
 						file->ucontext);
 		if (!ind_tbl)
 			rcq = rcq ?: scq;
-		pd  = uobj_get_obj_read(pd, cmd->pd_handle, file->ucontext);
+		pd  = uobj_get_obj_read(pd, UVERBS_OBJECT_PD, cmd->pd_handle, file->ucontext);
 		if (!pd || (!scq && has_sq)) {
 			ret = -EINVAL;
 			goto err_put;
@@ -1751,12 +1753,12 @@ ssize_t ib_uverbs_open_qp(struct ib_uverbs_file *file,
 		   in_len - sizeof(cmd) - sizeof(struct ib_uverbs_cmd_hdr),
 		   out_len - sizeof(resp));
 
-	obj  = (struct ib_uqp_object *)uobj_alloc(uobj_get_type(qp),
+	obj  = (struct ib_uqp_object *)uobj_alloc(UVERBS_OBJECT_QP,
 						  file->ucontext);
 	if (IS_ERR(obj))
 		return PTR_ERR(obj);
 
-	xrcd_uobj = uobj_get_read(uobj_get_type(xrcd), cmd.pd_handle,
+	xrcd_uobj = uobj_get_read(UVERBS_OBJECT_XRCD, cmd.pd_handle,
 				  file->ucontext);
 	if (IS_ERR(xrcd_uobj)) {
 		ret = -EINVAL;
@@ -1859,7 +1861,7 @@ ssize_t ib_uverbs_query_qp(struct ib_uverbs_file *file,
 		goto out;
 	}
 
-	qp = uobj_get_obj_read(qp, cmd.qp_handle, file->ucontext);
+	qp = uobj_get_obj_read(qp, UVERBS_OBJECT_QP, cmd.qp_handle, file->ucontext);
 	if (!qp) {
 		ret = -EINVAL;
 		goto out;
@@ -1964,7 +1966,7 @@ static int modify_qp(struct ib_uverbs_file *file,
 	if (!attr)
 		return -ENOMEM;
 
-	qp = uobj_get_obj_read(qp, cmd->base.qp_handle, file->ucontext);
+	qp = uobj_get_obj_read(qp, UVERBS_OBJECT_QP, cmd->base.qp_handle, file->ucontext);
 	if (!qp) {
 		ret = -EINVAL;
 		goto out;
@@ -1989,6 +1991,13 @@ static int modify_qp(struct ib_uverbs_file *file,
 		goto release_qp;
 	}
 
+	if ((cmd->base.attr_mask & IB_QP_CUR_STATE &&
+	    cmd->base.cur_qp_state > IB_QPS_ERR) ||
+	    cmd->base.qp_state > IB_QPS_ERR) {
+		ret = -EINVAL;
+		goto release_qp;
+	}
+
 	attr->qp_state		  = cmd->base.qp_state;
 	attr->cur_qp_state	  = cmd->base.cur_qp_state;
 	attr->path_mtu		  = cmd->base.path_mtu;
@@ -2112,7 +2121,7 @@ ssize_t ib_uverbs_destroy_qp(struct ib_uverbs_file *file,
 
 	memset(&resp, 0, sizeof resp);
 
-	uobj  = uobj_get_write(uobj_get_type(qp), cmd.qp_handle,
+	uobj  = uobj_get_write(UVERBS_OBJECT_QP, cmd.qp_handle,
 			       file->ucontext);
 	if (IS_ERR(uobj))
 		return PTR_ERR(uobj);
@@ -2178,7 +2187,7 @@ ssize_t ib_uverbs_post_send(struct ib_uverbs_file *file,
 	if (!user_wr)
 		return -ENOMEM;
 
-	qp = uobj_get_obj_read(qp, cmd.qp_handle, file->ucontext);
+	qp = uobj_get_obj_read(qp, UVERBS_OBJECT_QP, cmd.qp_handle, file->ucontext);
 	if (!qp)
 		goto out;
 
@@ -2214,7 +2223,7 @@ ssize_t ib_uverbs_post_send(struct ib_uverbs_file *file,
 				goto out_put;
 			}
 
-			ud->ah = uobj_get_obj_read(ah, user_wr->wr.ud.ah,
+			ud->ah = uobj_get_obj_read(ah, UVERBS_OBJECT_AH, user_wr->wr.ud.ah,
 						   file->ucontext);
 			if (!ud->ah) {
 				kfree(ud);
@@ -2449,7 +2458,7 @@ ssize_t ib_uverbs_post_recv(struct ib_uverbs_file *file,
 	if (IS_ERR(wr))
 		return PTR_ERR(wr);
 
-	qp = uobj_get_obj_read(qp, cmd.qp_handle, file->ucontext);
+	qp = uobj_get_obj_read(qp, UVERBS_OBJECT_QP, cmd.qp_handle, file->ucontext);
 	if (!qp)
 		goto out;
 
@@ -2498,7 +2507,7 @@ ssize_t ib_uverbs_post_srq_recv(struct ib_uverbs_file *file,
 	if (IS_ERR(wr))
 		return PTR_ERR(wr);
 
-	srq = uobj_get_obj_read(srq, cmd.srq_handle, file->ucontext);
+	srq = uobj_get_obj_read(srq, UVERBS_OBJECT_SRQ, cmd.srq_handle, file->ucontext);
 	if (!srq)
 		goto out;
 
@@ -2555,11 +2564,11 @@ ssize_t ib_uverbs_create_ah(struct ib_uverbs_file *file,
 		   in_len - sizeof(cmd) - sizeof(struct ib_uverbs_cmd_hdr),
 		   out_len - sizeof(resp));
 
-	uobj  = uobj_alloc(uobj_get_type(ah), file->ucontext);
+	uobj  = uobj_alloc(UVERBS_OBJECT_AH, file->ucontext);
 	if (IS_ERR(uobj))
 		return PTR_ERR(uobj);
 
-	pd = uobj_get_obj_read(pd, cmd.pd_handle, file->ucontext);
+	pd = uobj_get_obj_read(pd, UVERBS_OBJECT_PD, cmd.pd_handle, file->ucontext);
 	if (!pd) {
 		ret = -EINVAL;
 		goto err;
@@ -2627,7 +2636,7 @@ ssize_t ib_uverbs_destroy_ah(struct ib_uverbs_file *file,
 	if (copy_from_user(&cmd, buf, sizeof cmd))
 		return -EFAULT;
 
-	uobj  = uobj_get_write(uobj_get_type(ah), cmd.ah_handle,
+	uobj  = uobj_get_write(UVERBS_OBJECT_AH, cmd.ah_handle,
 			       file->ucontext);
 	if (IS_ERR(uobj))
 		return PTR_ERR(uobj);
@@ -2650,7 +2659,7 @@ ssize_t ib_uverbs_attach_mcast(struct ib_uverbs_file *file,
 	if (copy_from_user(&cmd, buf, sizeof cmd))
 		return -EFAULT;
 
-	qp = uobj_get_obj_read(qp, cmd.qp_handle, file->ucontext);
+	qp = uobj_get_obj_read(qp, UVERBS_OBJECT_QP, cmd.qp_handle, file->ucontext);
 	if (!qp)
 		return -EINVAL;
 
@@ -2701,7 +2710,7 @@ ssize_t ib_uverbs_detach_mcast(struct ib_uverbs_file *file,
 	if (copy_from_user(&cmd, buf, sizeof cmd))
 		return -EFAULT;
 
-	qp = uobj_get_obj_read(qp, cmd.qp_handle, file->ucontext);
+	qp = uobj_get_obj_read(qp, UVERBS_OBJECT_QP, cmd.qp_handle, file->ucontext);
 	if (!qp)
 		return -EINVAL;
 
@@ -2730,8 +2739,52 @@ out_put:
 	return ret ? ret : in_len;
 }
 
-static int kern_spec_to_ib_spec_action(struct ib_uverbs_flow_spec *kern_spec,
-				       union ib_flow_spec *ib_spec)
+struct ib_uflow_resources {
+	size_t			max;
+	size_t			num;
+	struct ib_flow_action	*collection[0];
+};
+
+static struct ib_uflow_resources *flow_resources_alloc(size_t num_specs)
+{
+	struct ib_uflow_resources *resources;
+
+	resources =
+		kmalloc(sizeof(*resources) +
+			num_specs * sizeof(*resources->collection), GFP_KERNEL);
+
+	if (!resources)
+		return NULL;
+
+	resources->num = 0;
+	resources->max = num_specs;
+
+	return resources;
+}
+
+void ib_uverbs_flow_resources_free(struct ib_uflow_resources *uflow_res)
+{
+	unsigned int i;
+
+	for (i = 0; i < uflow_res->num; i++)
+		atomic_dec(&uflow_res->collection[i]->usecnt);
+
+	kfree(uflow_res);
+}
+
+static void flow_resources_add(struct ib_uflow_resources *uflow_res,
+			       struct ib_flow_action *action)
+{
+	WARN_ON(uflow_res->num >= uflow_res->max);
+
+	atomic_inc(&action->usecnt);
+	uflow_res->collection[uflow_res->num++] = action;
+}
+
+static int kern_spec_to_ib_spec_action(struct ib_ucontext *ucontext,
+				       struct ib_uverbs_flow_spec *kern_spec,
+				       union ib_flow_spec *ib_spec,
+				       struct ib_uflow_resources *uflow_res)
 {
 	ib_spec->type = kern_spec->type;
 	switch (ib_spec->type) {
@@ -2750,19 +2803,34 @@ static int kern_spec_to_ib_spec_action(struct ib_uverbs_flow_spec *kern_spec,
 
 		ib_spec->drop.size = sizeof(struct ib_flow_spec_action_drop);
 		break;
+	case IB_FLOW_SPEC_ACTION_HANDLE:
+		if (kern_spec->action.size !=
+		    sizeof(struct ib_uverbs_flow_spec_action_handle))
+			return -EOPNOTSUPP;
+		ib_spec->action.act = uobj_get_obj_read(flow_action,
+							UVERBS_OBJECT_FLOW_ACTION,
+							kern_spec->action.handle,
+							ucontext);
+		if (!ib_spec->action.act)
+			return -EINVAL;
+		ib_spec->action.size =
+			sizeof(struct ib_flow_spec_action_handle);
+		flow_resources_add(uflow_res, ib_spec->action.act);
+		uobj_put_obj_read(ib_spec->action.act);
+		break;
 	default:
 		return -EINVAL;
 	}
 	return 0;
 }
 
-static size_t kern_spec_filter_sz(struct ib_uverbs_flow_spec_hdr *spec)
+static size_t kern_spec_filter_sz(const struct ib_uverbs_flow_spec_hdr *spec)
 {
 	/* Returns user space filter size, includes padding */
 	return (spec->size - sizeof(struct ib_uverbs_flow_spec_hdr)) / 2;
 }
 
-static ssize_t spec_filter_size(void *kern_spec_filter, u16 kern_filter_size,
+static ssize_t spec_filter_size(const void *kern_spec_filter, u16 kern_filter_size,
 				u16 ib_real_filter_sz)
 {
 	/*
@@ -2780,28 +2848,21 @@ static ssize_t spec_filter_size(void *kern_spec_filter, u16 kern_filter_size,
 	return kern_filter_size;
 }
 
-static int kern_spec_to_ib_spec_filter(struct ib_uverbs_flow_spec *kern_spec,
-				       union ib_flow_spec *ib_spec)
+int ib_uverbs_kern_spec_to_ib_spec_filter(enum ib_flow_spec_type type,
+					  const void *kern_spec_mask,
+					  const void *kern_spec_val,
+					  size_t kern_filter_sz,
+					  union ib_flow_spec *ib_spec)
 {
 	ssize_t actual_filter_sz;
-	ssize_t kern_filter_sz;
 	ssize_t ib_filter_sz;
-	void *kern_spec_mask;
-	void *kern_spec_val;
 
-	if (kern_spec->reserved)
-		return -EINVAL;
-
-	ib_spec->type = kern_spec->type;
-
-	kern_filter_sz = kern_spec_filter_sz(&kern_spec->hdr);
 	/* User flow spec size must be aligned to 4 bytes */
 	if (kern_filter_sz != ALIGN(kern_filter_sz, 4))
 		return -EINVAL;
 
-	kern_spec_val = (void *)kern_spec +
-		sizeof(struct ib_uverbs_flow_spec_hdr);
-	kern_spec_mask = kern_spec_val + kern_filter_sz;
+	ib_spec->type = type;
+
 	if (ib_spec->type == (IB_FLOW_SPEC_INNER | IB_FLOW_SPEC_VXLAN_TUNNEL))
 		return -EINVAL;
 
@@ -2870,20 +2931,56 @@ static int kern_spec_to_ib_spec_filter(struct ib_uverbs_flow_spec *kern_spec,
 		    (ntohl(ib_spec->tunnel.val.tunnel_id)) >= BIT(24))
 			return -EINVAL;
 		break;
+	case IB_FLOW_SPEC_ESP:
+		ib_filter_sz = offsetof(struct ib_flow_esp_filter, real_sz);
+		actual_filter_sz = spec_filter_size(kern_spec_mask,
+						    kern_filter_sz,
+						    ib_filter_sz);
+		if (actual_filter_sz <= 0)
+			return -EINVAL;
+		ib_spec->esp.size = sizeof(struct ib_flow_spec_esp);
+		memcpy(&ib_spec->esp.val, kern_spec_val, actual_filter_sz);
+		memcpy(&ib_spec->esp.mask, kern_spec_mask, actual_filter_sz);
+		break;
 	default:
 		return -EINVAL;
 	}
 	return 0;
 }
 
-static int kern_spec_to_ib_spec(struct ib_uverbs_flow_spec *kern_spec,
-				union ib_flow_spec *ib_spec)
+static int kern_spec_to_ib_spec_filter(struct ib_uverbs_flow_spec *kern_spec,
+				       union ib_flow_spec *ib_spec)
+{
+	ssize_t kern_filter_sz;
+	void *kern_spec_mask;
+	void *kern_spec_val;
+
+	if (kern_spec->reserved)
+		return -EINVAL;
+
+	kern_filter_sz = kern_spec_filter_sz(&kern_spec->hdr);
+
+	kern_spec_val = (void *)kern_spec +
+		sizeof(struct ib_uverbs_flow_spec_hdr);
+	kern_spec_mask = kern_spec_val + kern_filter_sz;
+
+	return ib_uverbs_kern_spec_to_ib_spec_filter(kern_spec->type,
+						     kern_spec_mask,
+						     kern_spec_val,
+						     kern_filter_sz, ib_spec);
+}
+
+static int kern_spec_to_ib_spec(struct ib_ucontext *ucontext,
+				struct ib_uverbs_flow_spec *kern_spec,
+				union ib_flow_spec *ib_spec,
+				struct ib_uflow_resources *uflow_res)
 {
 	if (kern_spec->reserved)
 		return -EINVAL;
 
 	if (kern_spec->type >= IB_FLOW_SPEC_ACTION_TAG)
-		return kern_spec_to_ib_spec_action(kern_spec, ib_spec);
+		return kern_spec_to_ib_spec_action(ucontext, kern_spec, ib_spec,
+						   uflow_res);
 	else
 		return kern_spec_to_ib_spec_filter(kern_spec, ib_spec);
 }
@@ -2925,18 +3022,18 @@ int ib_uverbs_ex_create_wq(struct ib_uverbs_file *file,
 	if (cmd.comp_mask)
 		return -EOPNOTSUPP;
 
-	obj  = (struct ib_uwq_object *)uobj_alloc(uobj_get_type(wq),
+	obj  = (struct ib_uwq_object *)uobj_alloc(UVERBS_OBJECT_WQ,
 						  file->ucontext);
 	if (IS_ERR(obj))
 		return PTR_ERR(obj);
 
-	pd  = uobj_get_obj_read(pd, cmd.pd_handle, file->ucontext);
+	pd  = uobj_get_obj_read(pd, UVERBS_OBJECT_PD, cmd.pd_handle, file->ucontext);
 	if (!pd) {
 		err = -EINVAL;
 		goto err_uobj;
 	}
 
-	cq = uobj_get_obj_read(cq, cmd.cq_handle, file->ucontext);
+	cq = uobj_get_obj_read(cq, UVERBS_OBJECT_CQ, cmd.cq_handle, file->ucontext);
 	if (!cq) {
 		err = -EINVAL;
 		goto err_put_pd;
@@ -3040,7 +3137,7 @@ int ib_uverbs_ex_destroy_wq(struct ib_uverbs_file *file,
 		return -EOPNOTSUPP;
 
 	resp.response_length = required_resp_len;
-	uobj  = uobj_get_write(uobj_get_type(wq), cmd.wq_handle,
+	uobj  = uobj_get_write(UVERBS_OBJECT_WQ, cmd.wq_handle,
 			       file->ucontext);
 	if (IS_ERR(uobj))
 		return PTR_ERR(uobj);
@@ -3091,7 +3188,7 @@ int ib_uverbs_ex_modify_wq(struct ib_uverbs_file *file,
 	if (cmd.attr_mask > (IB_WQ_STATE | IB_WQ_CUR_STATE | IB_WQ_FLAGS))
 		return -EINVAL;
 
-	wq = uobj_get_obj_read(wq, cmd.wq_handle, file->ucontext);
+	wq = uobj_get_obj_read(wq, UVERBS_OBJECT_WQ, cmd.wq_handle, file->ucontext);
 	if (!wq)
 		return -EINVAL;
 
@@ -3185,7 +3282,7 @@ int ib_uverbs_ex_create_rwq_ind_table(struct ib_uverbs_file *file,
 
 	for (num_read_wqs = 0; num_read_wqs < num_wq_handles;
 			num_read_wqs++) {
-		wq = uobj_get_obj_read(wq, wqs_handles[num_read_wqs],
+		wq = uobj_get_obj_read(wq, UVERBS_OBJECT_WQ, wqs_handles[num_read_wqs],
 				       file->ucontext);
 		if (!wq) {
 			err = -EINVAL;
@@ -3195,7 +3292,7 @@ int ib_uverbs_ex_create_rwq_ind_table(struct ib_uverbs_file *file,
 		wqs[num_read_wqs] = wq;
 	}
 
-	uobj  = uobj_alloc(uobj_get_type(rwq_ind_table), file->ucontext);
+	uobj  = uobj_alloc(UVERBS_OBJECT_RWQ_IND_TBL, file->ucontext);
 	if (IS_ERR(uobj)) {
 		err = PTR_ERR(uobj);
 		goto put_wqs;
@@ -3282,7 +3379,7 @@ int ib_uverbs_ex_destroy_rwq_ind_table(struct ib_uverbs_file *file,
 	if (cmd.comp_mask)
 		return -EOPNOTSUPP;
 
-	uobj  = uobj_get_write(uobj_get_type(rwq_ind_table), cmd.ind_tbl_handle,
+	uobj  = uobj_get_write(UVERBS_OBJECT_RWQ_IND_TBL, cmd.ind_tbl_handle,
 			       file->ucontext);
 	if (IS_ERR(uobj))
 		return PTR_ERR(uobj);
@@ -3298,10 +3395,12 @@ 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;
 	struct ib_qp			  *qp;
+	struct ib_uflow_resources	  *uflow_res;
 	int err = 0;
 	void *kern_spec;
 	void *ib_spec;
@@ -3361,13 +3460,13 @@ int ib_uverbs_ex_create_flow(struct ib_uverbs_file *file,
 		kern_flow_attr = &cmd.flow_attr;
 	}
 
-	uobj  = uobj_alloc(uobj_get_type(flow), file->ucontext);
+	uobj  = uobj_alloc(UVERBS_OBJECT_FLOW, file->ucontext);
 	if (IS_ERR(uobj)) {
 		err = PTR_ERR(uobj);
 		goto err_free_attr;
 	}
 
-	qp = uobj_get_obj_read(qp, cmd.qp_handle, file->ucontext);
+	qp = uobj_get_obj_read(qp, UVERBS_OBJECT_QP, cmd.qp_handle, file->ucontext);
 	if (!qp) {
 		err = -EINVAL;
 		goto err_uobj;
@@ -3379,6 +3478,11 @@ int ib_uverbs_ex_create_flow(struct ib_uverbs_file *file,
 		err = -ENOMEM;
 		goto err_put;
 	}
+	uflow_res = flow_resources_alloc(cmd.flow_attr.num_of_specs);
+	if (!uflow_res) {
+		err = -ENOMEM;
+		goto err_free_flow_attr;
+	}
 
 	flow_attr->type = kern_flow_attr->type;
 	flow_attr->priority = kern_flow_attr->priority;
@@ -3393,7 +3497,8 @@ int ib_uverbs_ex_create_flow(struct ib_uverbs_file *file,
 	     cmd.flow_attr.size > offsetof(struct ib_uverbs_flow_spec, reserved) &&
 	     cmd.flow_attr.size >=
 	     ((struct ib_uverbs_flow_spec *)kern_spec)->size; i++) {
-		err = kern_spec_to_ib_spec(kern_spec, ib_spec);
+		err = kern_spec_to_ib_spec(file->ucontext, kern_spec, ib_spec,
+					   uflow_res);
 		if (err)
 			goto err_free;
 		flow_attr->size +=
@@ -3415,6 +3520,8 @@ int ib_uverbs_ex_create_flow(struct ib_uverbs_file *file,
 	}
 	flow_id->uobject = uobj;
 	uobj->object = flow_id;
+	uflow = container_of(uobj, typeof(*uflow), uobject);
+	uflow->resources = uflow_res;
 
 	memset(&resp, 0, sizeof(resp));
 	resp.flow_handle = uobj->id;
@@ -3433,6 +3540,8 @@ int ib_uverbs_ex_create_flow(struct ib_uverbs_file *file,
 err_copy:
 	ib_destroy_flow(flow_id);
 err_free:
+	ib_uverbs_flow_resources_free(uflow_res);
+err_free_flow_attr:
 	kfree(flow_attr);
 err_put:
 	uobj_put_obj_read(qp);
@@ -3463,7 +3572,7 @@ int ib_uverbs_ex_destroy_flow(struct ib_uverbs_file *file,
 	if (cmd.comp_mask)
 		return -EINVAL;
 
-	uobj  = uobj_get_write(uobj_get_type(flow), cmd.flow_handle,
+	uobj  = uobj_get_write(UVERBS_OBJECT_FLOW, cmd.flow_handle,
 			       file->ucontext);
 	if (IS_ERR(uobj))
 		return PTR_ERR(uobj);
@@ -3485,7 +3594,7 @@ static int __uverbs_create_xsrq(struct ib_uverbs_file *file,
 	struct ib_srq_init_attr          attr;
 	int ret;
 
-	obj  = (struct ib_usrq_object *)uobj_alloc(uobj_get_type(srq),
+	obj  = (struct ib_usrq_object *)uobj_alloc(UVERBS_OBJECT_SRQ,
 						   file->ucontext);
 	if (IS_ERR(obj))
 		return PTR_ERR(obj);
@@ -3494,7 +3603,7 @@ static int __uverbs_create_xsrq(struct ib_uverbs_file *file,
 		attr.ext.tag_matching.max_num_tags = cmd->max_num_tags;
 
 	if (cmd->srq_type == IB_SRQT_XRC) {
-		xrcd_uobj = uobj_get_read(uobj_get_type(xrcd), cmd->xrcd_handle,
+		xrcd_uobj = uobj_get_read(UVERBS_OBJECT_XRCD, cmd->xrcd_handle,
 					  file->ucontext);
 		if (IS_ERR(xrcd_uobj)) {
 			ret = -EINVAL;
@@ -3512,7 +3621,7 @@ static int __uverbs_create_xsrq(struct ib_uverbs_file *file,
 	}
 
 	if (ib_srq_has_cq(cmd->srq_type)) {
-		attr.ext.cq  = uobj_get_obj_read(cq, cmd->cq_handle,
+		attr.ext.cq  = uobj_get_obj_read(cq, UVERBS_OBJECT_CQ, cmd->cq_handle,
 						 file->ucontext);
 		if (!attr.ext.cq) {
 			ret = -EINVAL;
@@ -3520,7 +3629,7 @@ static int __uverbs_create_xsrq(struct ib_uverbs_file *file,
 		}
 	}
 
-	pd  = uobj_get_obj_read(pd, cmd->pd_handle, file->ucontext);
+	pd  = uobj_get_obj_read(pd, UVERBS_OBJECT_PD, cmd->pd_handle, file->ucontext);
 	if (!pd) {
 		ret = -EINVAL;
 		goto err_put_cq;
@@ -3572,7 +3681,7 @@ static int __uverbs_create_xsrq(struct ib_uverbs_file *file,
 	if (cmd->srq_type == IB_SRQT_XRC)
 		resp.srqn = srq->ext.xrc.srq_num;
 
-	if (copy_to_user((void __user *) (unsigned long) cmd->response,
+	if (copy_to_user(u64_to_user_ptr(cmd->response),
 			 &resp, sizeof resp)) {
 		ret = -EFAULT;
 		goto err_copy;
@@ -3692,7 +3801,7 @@ ssize_t ib_uverbs_modify_srq(struct ib_uverbs_file *file,
 	ib_uverbs_init_udata(&udata, buf + sizeof cmd, NULL, in_len - sizeof cmd,
 		   out_len);
 
-	srq = uobj_get_obj_read(srq, cmd.srq_handle, file->ucontext);
+	srq = uobj_get_obj_read(srq, UVERBS_OBJECT_SRQ, cmd.srq_handle, file->ucontext);
 	if (!srq)
 		return -EINVAL;
 
@@ -3723,7 +3832,7 @@ ssize_t ib_uverbs_query_srq(struct ib_uverbs_file *file,
 	if (copy_from_user(&cmd, buf, sizeof cmd))
 		return -EFAULT;
 
-	srq = uobj_get_obj_read(srq, cmd.srq_handle, file->ucontext);
+	srq = uobj_get_obj_read(srq, UVERBS_OBJECT_SRQ, cmd.srq_handle, file->ucontext);
 	if (!srq)
 		return -EINVAL;
 
@@ -3760,7 +3869,7 @@ ssize_t ib_uverbs_destroy_srq(struct ib_uverbs_file *file,
 	if (copy_from_user(&cmd, buf, sizeof cmd))
 		return -EFAULT;
 
-	uobj  = uobj_get_write(uobj_get_type(srq), cmd.srq_handle,
+	uobj  = uobj_get_write(UVERBS_OBJECT_SRQ, cmd.srq_handle,
 			       file->ucontext);
 	if (IS_ERR(uobj))
 		return PTR_ERR(uobj);
@@ -3897,6 +4006,12 @@ int ib_uverbs_ex_query_device(struct ib_uverbs_file *file,
 	resp.cq_moderation_caps.max_cq_moderation_period =
 		attr.cq_caps.max_cq_moderation_period;
 	resp.response_length += sizeof(resp.cq_moderation_caps);
+
+	if (ucore->outlen < resp.response_length + sizeof(resp.max_dm_size))
+		goto end;
+
+	resp.max_dm_size = attr.max_dm_size;
+	resp.response_length += sizeof(resp.max_dm_size);
 end:
 	err = ib_copy_to_udata(ucore, &resp, resp.response_length);
 	return err;
@@ -3933,7 +4048,7 @@ int ib_uverbs_ex_modify_cq(struct ib_uverbs_file *file,
 	if (cmd.attr_mask > IB_CQ_MODERATE)
 		return -EOPNOTSUPP;
 
-	cq = uobj_get_obj_read(cq, cmd.cq_handle, file->ucontext);
+	cq = uobj_get_obj_read(cq, UVERBS_OBJECT_CQ, cmd.cq_handle, file->ucontext);
 	if (!cq)
 		return -EINVAL;
 

drivers/infiniband/core/uverbs_ioctl.c

@@ -35,6 +35,17 @@
 #include "rdma_core.h"
 #include "uverbs.h"
 
+static bool uverbs_is_attr_cleared(const struct ib_uverbs_attr *uattr,
+				   u16 len)
+{
+	if (uattr->len > sizeof(((struct ib_uverbs_attr *)0)->data))
+		return ib_is_buffer_cleared(u64_to_user_ptr(uattr->data) + len,
+					    uattr->len - len);
+
+	return !memchr_inv((const void *)&uattr->data + len,
+			   0, uattr->len - len);
+}
+
 static int uverbs_process_attr(struct ib_device *ibdev,
 			       struct ib_ucontext *ucontext,
 			       const struct ib_uverbs_attr *uattr,
@@ -44,14 +55,12 @@ static int uverbs_process_attr(struct ib_device *ibdev,
 			       struct ib_uverbs_attr __user *uattr_ptr)
 {
 	const struct uverbs_attr_spec *spec;
+	const struct uverbs_attr_spec *val_spec;
 	struct uverbs_attr *e;
 	const struct uverbs_object_spec *object;
 	struct uverbs_obj_attr *o_attr;
 	struct uverbs_attr *elements = attr_bundle_h->attrs;
 
-	if (uattr->reserved)
-		return -EINVAL;
-
 	if (attr_id >= attr_spec_bucket->num_attrs) {
 		if (uattr->flags & UVERBS_ATTR_F_MANDATORY)
 			return -EINVAL;
@@ -63,15 +72,46 @@ static int uverbs_process_attr(struct ib_device *ibdev,
 		return -EINVAL;
 
 	spec = &attr_spec_bucket->attrs[attr_id];
+	val_spec = spec;
 	e = &elements[attr_id];
 	e->uattr = uattr_ptr;
 
 	switch (spec->type) {
+	case UVERBS_ATTR_TYPE_ENUM_IN:
+		if (uattr->attr_data.enum_data.elem_id >= spec->enum_def.num_elems)
+			return -EOPNOTSUPP;
+
+		if (uattr->attr_data.enum_data.reserved)
+			return -EINVAL;
+
+		val_spec = &spec->enum_def.ids[uattr->attr_data.enum_data.elem_id];
+
+		/* Currently we only support PTR_IN based enums */
+		if (val_spec->type != UVERBS_ATTR_TYPE_PTR_IN)
+			return -EOPNOTSUPP;
+
+		e->ptr_attr.enum_id = uattr->attr_data.enum_data.elem_id;
+	/* fall through */
 	case UVERBS_ATTR_TYPE_PTR_IN:
+		/* Ensure that any data provided by userspace beyond the known
+		 * struct is zero. Userspace that knows how to use some future
+		 * longer struct will fail here if used with an old kernel and
+		 * non-zero content, making ABI compat/discovery simpler.
+		 */
+		if (uattr->len > val_spec->ptr.len &&
+		    val_spec->flags & UVERBS_ATTR_SPEC_F_MIN_SZ_OR_ZERO &&
+		    !uverbs_is_attr_cleared(uattr, val_spec->ptr.len))
+			return -EOPNOTSUPP;
+
+	/* fall through */
 	case UVERBS_ATTR_TYPE_PTR_OUT:
-		if (uattr->len < spec->len ||
-		    (!(spec->flags & UVERBS_ATTR_SPEC_F_MIN_SZ) &&
-		     uattr->len > spec->len))
+		if (uattr->len < val_spec->ptr.min_len ||
+		    (!(val_spec->flags & UVERBS_ATTR_SPEC_F_MIN_SZ_OR_ZERO) &&
+		     uattr->len > val_spec->ptr.len))
+			return -EINVAL;
+
+		if (spec->type != UVERBS_ATTR_TYPE_ENUM_IN &&
+		    uattr->attr_data.reserved)
 			return -EINVAL;
 
 		e->ptr_attr.data = uattr->data;
@@ -84,6 +124,9 @@ static int uverbs_process_attr(struct ib_device *ibdev,
 			return -EINVAL;
 	/* fall through */
 	case UVERBS_ATTR_TYPE_FD:
+		if (uattr->attr_data.reserved)
+			return -EINVAL;
+
 		if (uattr->len != 0 || !ucontext || uattr->data > INT_MAX)
 			return -EINVAL;
 
@@ -246,6 +289,9 @@ static long ib_uverbs_cmd_verbs(struct ib_device *ib_dev,
 	size_t ctx_size;
 	uintptr_t data[UVERBS_OPTIMIZE_USING_STACK_SZ / sizeof(uintptr_t)];
 
+	if (hdr->driver_id != ib_dev->driver_id)
+		return -EINVAL;
+
 	object_spec = uverbs_get_object(ib_dev, hdr->object_id);
 	if (!object_spec)
 		return -EPROTONOSUPPORT;
@@ -350,7 +396,7 @@ long ib_uverbs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
 			goto out;
 		}
 
-		if (hdr.reserved) {
+		if (hdr.reserved1 || hdr.reserved2) {
 			err = -EPROTONOSUPPORT;
 			goto out;
 		}

drivers/infiniband/core/uverbs_ioctl_merge.c

@@ -379,7 +379,7 @@ static struct uverbs_method_spec *build_method_with_attrs(const struct uverbs_me
 				 "ib_uverbs: Tried to merge attr (%d) but it's an object with new/destroy access but isn't mandatory\n",
 				 min_id) ||
 			    WARN(IS_ATTR_OBJECT(attr) &&
-				 attr->flags & UVERBS_ATTR_SPEC_F_MIN_SZ,
+				 attr->flags & UVERBS_ATTR_SPEC_F_MIN_SZ_OR_ZERO,
 				 "ib_uverbs: Tried to merge attr (%d) but it's an object with min_sz flag\n",
 				 min_id)) {
 				res = -EINVAL;

drivers/infiniband/core/uverbs_main.c

@@ -468,7 +468,7 @@ void ib_uverbs_comp_handler(struct ib_cq *cq, void *cq_context)
 		return;
 	}
 
-	entry = kmalloc(sizeof *entry, GFP_ATOMIC);
+	entry = kmalloc(sizeof(*entry), GFP_ATOMIC);
 	if (!entry) {
 		spin_unlock_irqrestore(&ev_queue->lock, flags);
 		return;
@@ -501,7 +501,7 @@ static void ib_uverbs_async_handler(struct ib_uverbs_file *file,
 		return;
 	}
 
-	entry = kmalloc(sizeof *entry, GFP_ATOMIC);
+	entry = kmalloc(sizeof(*entry), GFP_ATOMIC);
 	if (!entry) {
 		spin_unlock_irqrestore(&file->async_file->ev_queue.lock, flags);
 		return;
@@ -635,39 +635,87 @@ err_put_refs:
 	return filp;
 }
 
-static int verify_command_mask(struct ib_device *ib_dev, __u32 command)
+static bool verify_command_mask(struct ib_device *ib_dev,
+				u32 command, bool extended)
 {
-	u64 mask;
+	if (!extended)
+		return ib_dev->uverbs_cmd_mask & BIT_ULL(command);
 
-	if (command <= IB_USER_VERBS_CMD_OPEN_QP)
-		mask = ib_dev->uverbs_cmd_mask;
-	else
-		mask = ib_dev->uverbs_ex_cmd_mask;
-
-	if (mask & ((u64)1 << command))
-		return 0;
-
-	return -1;
+	return ib_dev->uverbs_ex_cmd_mask & BIT_ULL(command);
 }
 
 static bool verify_command_idx(u32 command, bool extended)
 {
 	if (extended)
-		return command < ARRAY_SIZE(uverbs_ex_cmd_table);
+		return command < ARRAY_SIZE(uverbs_ex_cmd_table) &&
+		       uverbs_ex_cmd_table[command];
+
+	return command < ARRAY_SIZE(uverbs_cmd_table) &&
+	       uverbs_cmd_table[command];
+}
+
+static ssize_t process_hdr(struct ib_uverbs_cmd_hdr *hdr,
+			   u32 *command, bool *extended)
+{
+	if (hdr->command & ~(u32)(IB_USER_VERBS_CMD_FLAG_EXTENDED |
+				   IB_USER_VERBS_CMD_COMMAND_MASK))
+		return -EINVAL;
+
+	*command = hdr->command & IB_USER_VERBS_CMD_COMMAND_MASK;
+	*extended = hdr->command & IB_USER_VERBS_CMD_FLAG_EXTENDED;
+
+	if (!verify_command_idx(*command, *extended))
+		return -EOPNOTSUPP;
+
+	return 0;
+}
+
+static ssize_t verify_hdr(struct ib_uverbs_cmd_hdr *hdr,
+			  struct ib_uverbs_ex_cmd_hdr *ex_hdr,
+			  size_t count, bool extended)
+{
+	if (extended) {
+		count -= sizeof(*hdr) + sizeof(*ex_hdr);
+
+		if ((hdr->in_words + ex_hdr->provider_in_words) * 8 != count)
+			return -EINVAL;
+
+		if (ex_hdr->cmd_hdr_reserved)
+			return -EINVAL;
 
-	return command < ARRAY_SIZE(uverbs_cmd_table);
+		if (ex_hdr->response) {
+			if (!hdr->out_words && !ex_hdr->provider_out_words)
+				return -EINVAL;
+
+			if (!access_ok(VERIFY_WRITE,
+				       u64_to_user_ptr(ex_hdr->response),
+				       (hdr->out_words + ex_hdr->provider_out_words) * 8))
+				return -EFAULT;
+		} else {
+			if (hdr->out_words || ex_hdr->provider_out_words)
+				return -EINVAL;
+		}
+
+		return 0;
+	}
+
+	/* not extended command */
+	if (hdr->in_words * 4 != count)
+		return -EINVAL;
+
+	return 0;
 }
 
 static ssize_t ib_uverbs_write(struct file *filp, const char __user *buf,
 			     size_t count, loff_t *pos)
 {
 	struct ib_uverbs_file *file = filp->private_data;
+	struct ib_uverbs_ex_cmd_hdr ex_hdr;
 	struct ib_device *ib_dev;
 	struct ib_uverbs_cmd_hdr hdr;
-	bool extended_command;
-	__u32 command;
-	__u32 flags;
+	bool extended;
 	int srcu_key;
+	u32 command;
 	ssize_t ret;
 
 	if (!ib_safe_file_access(filp)) {
@@ -676,12 +724,31 @@ static ssize_t ib_uverbs_write(struct file *filp, const char __user *buf,
 		return -EACCES;
 	}
 
-	if (count < sizeof hdr)
+	if (count < sizeof(hdr))
 		return -EINVAL;
 
-	if (copy_from_user(&hdr, buf, sizeof hdr))
+	if (copy_from_user(&hdr, buf, sizeof(hdr)))
 		return -EFAULT;
 
+	ret = process_hdr(&hdr, &command, &extended);
+	if (ret)
+		return ret;
+
+	if (!file->ucontext &&
+	    (command != IB_USER_VERBS_CMD_GET_CONTEXT || extended))
+		return -EINVAL;
+
+	if (extended) {
+		if (count < (sizeof(hdr) + sizeof(ex_hdr)))
+			return -EINVAL;
+		if (copy_from_user(&ex_hdr, buf + sizeof(hdr), sizeof(ex_hdr)))
+			return -EFAULT;
+	}
+
+	ret = verify_hdr(&hdr, &ex_hdr, count, extended);
+	if (ret)
+		return ret;
+
 	srcu_key = srcu_read_lock(&file->device->disassociate_srcu);
 	ib_dev = srcu_dereference(file->device->ib_dev,
 				  &file->device->disassociate_srcu);
@@ -690,106 +757,22 @@ static ssize_t ib_uverbs_write(struct file *filp, const char __user *buf,
 		goto out;
 	}
 
-	if (hdr.command & ~(__u32)(IB_USER_VERBS_CMD_FLAGS_MASK |
-				   IB_USER_VERBS_CMD_COMMAND_MASK)) {
-		ret = -EINVAL;
-		goto out;
-	}
-
-	command = hdr.command & IB_USER_VERBS_CMD_COMMAND_MASK;
-	flags = (hdr.command &
-		 IB_USER_VERBS_CMD_FLAGS_MASK) >> IB_USER_VERBS_CMD_FLAGS_SHIFT;
-
-	extended_command = flags & IB_USER_VERBS_CMD_FLAG_EXTENDED;
-	if (!verify_command_idx(command, extended_command)) {
-		ret = -EINVAL;
-		goto out;
-	}
-
-	if (verify_command_mask(ib_dev, command)) {
+	if (!verify_command_mask(ib_dev, command, extended)) {
 		ret = -EOPNOTSUPP;
 		goto out;
 	}
 
-	if (!file->ucontext &&
-	    command != IB_USER_VERBS_CMD_GET_CONTEXT) {
-		ret = -EINVAL;
-		goto out;
-	}
-
-	if (!flags) {
-		if (!uverbs_cmd_table[command]) {
-			ret = -EINVAL;
-			goto out;
-		}
-
-		if (hdr.in_words * 4 != count) {
-			ret = -EINVAL;
-			goto out;
-		}
+	buf += sizeof(hdr);
 
-		ret = uverbs_cmd_table[command](file, ib_dev,
-						 buf + sizeof(hdr),
-						 hdr.in_words * 4,
-						 hdr.out_words * 4);
-
-	} else if (flags == IB_USER_VERBS_CMD_FLAG_EXTENDED) {
-		struct ib_uverbs_ex_cmd_hdr ex_hdr;
+	if (!extended) {
+		ret = uverbs_cmd_table[command](file, ib_dev, buf,
+						hdr.in_words * 4,
+						hdr.out_words * 4);
+	} else {
 		struct ib_udata ucore;
 		struct ib_udata uhw;
-		size_t written_count = count;
 
-		if (!uverbs_ex_cmd_table[command]) {
-			ret = -ENOSYS;
-			goto out;
-		}
-
-		if (!file->ucontext) {
-			ret = -EINVAL;
-			goto out;
-		}
-
-		if (count < (sizeof(hdr) + sizeof(ex_hdr))) {
-			ret = -EINVAL;
-			goto out;
-		}
-
-		if (copy_from_user(&ex_hdr, buf + sizeof(hdr), sizeof(ex_hdr))) {
-			ret = -EFAULT;
-			goto out;
-		}
-
-		count -= sizeof(hdr) + sizeof(ex_hdr);
-		buf += sizeof(hdr) + sizeof(ex_hdr);
-
-		if ((hdr.in_words + ex_hdr.provider_in_words) * 8 != count) {
-			ret = -EINVAL;
-			goto out;
-		}
-
-		if (ex_hdr.cmd_hdr_reserved) {
-			ret = -EINVAL;
-			goto out;
-		}
-
-		if (ex_hdr.response) {
-			if (!hdr.out_words && !ex_hdr.provider_out_words) {
-				ret = -EINVAL;
-				goto out;
-			}
-
-			if (!access_ok(VERIFY_WRITE,
-				       u64_to_user_ptr(ex_hdr.response),
-				       (hdr.out_words + ex_hdr.provider_out_words) * 8)) {
-				ret = -EFAULT;
-				goto out;
-			}
-		} else {
-			if (hdr.out_words || ex_hdr.provider_out_words) {
-				ret = -EINVAL;
-				goto out;
-			}
-		}
+		buf += sizeof(ex_hdr);
 
 		ib_uverbs_init_udata_buf_or_null(&ucore, buf,
 					u64_to_user_ptr(ex_hdr.response),
@@ -802,10 +785,7 @@ static ssize_t ib_uverbs_write(struct file *filp, const char __user *buf,
 					ex_hdr.provider_out_words * 8);
 
 		ret = uverbs_ex_cmd_table[command](file, ib_dev, &ucore, &uhw);
-		if (!ret)
-			ret = written_count;
-	} else {
-		ret = -ENOSYS;
+		ret = (ret) ? : count;
 	}
 
 out:
@@ -953,10 +933,8 @@ static const struct file_operations uverbs_fops = {
 	.open	 = ib_uverbs_open,
 	.release = ib_uverbs_close,
 	.llseek	 = no_llseek,
-#if IS_ENABLED(CONFIG_INFINIBAND_EXP_USER_ACCESS)
 	.unlocked_ioctl = ib_uverbs_ioctl,
 	.compat_ioctl = ib_uverbs_ioctl,
-#endif
 };
 
 static const struct file_operations uverbs_mmap_fops = {
@@ -966,10 +944,8 @@ static const struct file_operations uverbs_mmap_fops = {
 	.open	 = ib_uverbs_open,
 	.release = ib_uverbs_close,
 	.llseek	 = no_llseek,
-#if IS_ENABLED(CONFIG_INFINIBAND_EXP_USER_ACCESS)
 	.unlocked_ioctl = ib_uverbs_ioctl,
 	.compat_ioctl = ib_uverbs_ioctl,
-#endif
 };
 
 static struct ib_client uverbs_client = {
@@ -1032,7 +1008,7 @@ static void ib_uverbs_add_one(struct ib_device *device)
 	if (!device->alloc_ucontext)
 		return;
 
-	uverbs_dev = kzalloc(sizeof *uverbs_dev, GFP_KERNEL);
+	uverbs_dev = kzalloc(sizeof(*uverbs_dev), GFP_KERNEL);
 	if (!uverbs_dev)
 		return;
 

drivers/infiniband/core/uverbs_std_types.c

@@ -48,7 +48,16 @@ static int uverbs_free_ah(struct ib_uobject *uobject,
 static int uverbs_free_flow(struct ib_uobject *uobject,
 			    enum rdma_remove_reason why)
 {
-	return ib_destroy_flow((struct ib_flow *)uobject->object);
+	int ret;
+	struct ib_flow *flow = (struct ib_flow *)uobject->object;
+	struct ib_uflow_object *uflow =
+		container_of(uobject, struct ib_uflow_object, uobject);
+
+	ret = ib_destroy_flow(flow);
+	if (!ret)
+		ib_uverbs_flow_resources_free(uflow->resources);
+
+	return ret;
 }
 
 static int uverbs_free_mw(struct ib_uobject *uobject,
@@ -135,31 +144,6 @@ static int uverbs_free_srq(struct ib_uobject *uobject,
 	return ret;
 }
 
-static int uverbs_free_cq(struct ib_uobject *uobject,
-			  enum rdma_remove_reason why)
-{
-	struct ib_cq *cq = uobject->object;
-	struct ib_uverbs_event_queue *ev_queue = cq->cq_context;
-	struct ib_ucq_object *ucq =
-		container_of(uobject, struct ib_ucq_object, uobject);
-	int ret;
-
-	ret = ib_destroy_cq(cq);
-	if (!ret || why != RDMA_REMOVE_DESTROY)
-		ib_uverbs_release_ucq(uobject->context->ufile, ev_queue ?
-				      container_of(ev_queue,
-						   struct ib_uverbs_completion_event_file,
-						   ev_queue) : NULL,
-				      ucq);
-	return ret;
-}
-
-static int uverbs_free_mr(struct ib_uobject *uobject,
-			  enum rdma_remove_reason why)
-{
-	return ib_dereg_mr((struct ib_mr *)uobject->object);
-}
-
 static int uverbs_free_xrcd(struct ib_uobject *uobject,
 			    enum rdma_remove_reason why)
 {
@@ -210,18 +194,26 @@ static int uverbs_hot_unplug_completion_event_file(struct ib_uobject_file *uobj_
 	return 0;
 };
 
+int uverbs_destroy_def_handler(struct ib_device *ib_dev,
+			       struct ib_uverbs_file *file,
+			       struct uverbs_attr_bundle *attrs)
+{
+	return 0;
+}
+
 /*
  * This spec is used in order to pass information to the hardware driver in a
  * legacy way. Every verb that could get driver specific data should get this
  * spec.
  */
-static const struct uverbs_attr_def uverbs_uhw_compat_in =
-	UVERBS_ATTR_PTR_IN_SZ(UVERBS_UHW_IN, 0, UA_FLAGS(UVERBS_ATTR_SPEC_F_MIN_SZ));
-static const struct uverbs_attr_def uverbs_uhw_compat_out =
-	UVERBS_ATTR_PTR_OUT_SZ(UVERBS_UHW_OUT, 0, UA_FLAGS(UVERBS_ATTR_SPEC_F_MIN_SZ));
-
-static void create_udata(struct uverbs_attr_bundle *ctx,
-			 struct ib_udata *udata)
+const struct uverbs_attr_def uverbs_uhw_compat_in =
+	UVERBS_ATTR_PTR_IN_SZ(UVERBS_ATTR_UHW_IN, UVERBS_ATTR_SIZE(0, USHRT_MAX),
+			      UA_FLAGS(UVERBS_ATTR_SPEC_F_MIN_SZ_OR_ZERO));
+const struct uverbs_attr_def uverbs_uhw_compat_out =
+	UVERBS_ATTR_PTR_OUT_SZ(UVERBS_ATTR_UHW_OUT, UVERBS_ATTR_SIZE(0, USHRT_MAX),
+			       UA_FLAGS(UVERBS_ATTR_SPEC_F_MIN_SZ_OR_ZERO));
+
+void create_udata(struct uverbs_attr_bundle *ctx, struct ib_udata *udata)
 {
 	/*
 	 * This is for ease of conversion. The purpose is to convert all drivers
@@ -229,9 +221,9 @@ static void create_udata(struct uverbs_attr_bundle *ctx,
 	 * Assume attr == 0 is input and attr == 1 is output.
 	 */
 	const struct uverbs_attr *uhw_in =
-		uverbs_attr_get(ctx, UVERBS_UHW_IN);
+		uverbs_attr_get(ctx, UVERBS_ATTR_UHW_IN);
 	const struct uverbs_attr *uhw_out =
-		uverbs_attr_get(ctx, UVERBS_UHW_OUT);
+		uverbs_attr_get(ctx, UVERBS_ATTR_UHW_OUT);
 
 	if (!IS_ERR(uhw_in)) {
 		udata->inlen = uhw_in->ptr_attr.len;
@@ -253,207 +245,67 @@ static void create_udata(struct uverbs_attr_bundle *ctx,
 	}
 }
 
-static int uverbs_create_cq_handler(struct ib_device *ib_dev,
-				    struct ib_uverbs_file *file,
-				    struct uverbs_attr_bundle *attrs)
+DECLARE_UVERBS_NAMED_OBJECT(UVERBS_OBJECT_COMP_CHANNEL,
+			    &UVERBS_TYPE_ALLOC_FD(0,
+						  sizeof(struct ib_uverbs_completion_event_file),
+						  uverbs_hot_unplug_completion_event_file,
+						  &uverbs_event_fops,
+						  "[infinibandevent]", O_RDONLY));
+
+DECLARE_UVERBS_NAMED_OBJECT(UVERBS_OBJECT_QP,
+			    &UVERBS_TYPE_ALLOC_IDR_SZ(sizeof(struct ib_uqp_object), 0,
+						      uverbs_free_qp));
+
+DECLARE_UVERBS_NAMED_OBJECT(UVERBS_OBJECT_MW,
+			    &UVERBS_TYPE_ALLOC_IDR(0, uverbs_free_mw));
+
+DECLARE_UVERBS_NAMED_OBJECT(UVERBS_OBJECT_SRQ,
+			    &UVERBS_TYPE_ALLOC_IDR_SZ(sizeof(struct ib_usrq_object), 0,
+						      uverbs_free_srq));
+
+DECLARE_UVERBS_NAMED_OBJECT(UVERBS_OBJECT_AH,
+			    &UVERBS_TYPE_ALLOC_IDR(0, uverbs_free_ah));
+
+DECLARE_UVERBS_NAMED_OBJECT(UVERBS_OBJECT_FLOW,
+			    &UVERBS_TYPE_ALLOC_IDR_SZ(sizeof(struct ib_uflow_object),
+						      0, uverbs_free_flow));
+
+DECLARE_UVERBS_NAMED_OBJECT(UVERBS_OBJECT_WQ,
+			    &UVERBS_TYPE_ALLOC_IDR_SZ(sizeof(struct ib_uwq_object), 0,
+						      uverbs_free_wq));
+
+DECLARE_UVERBS_NAMED_OBJECT(UVERBS_OBJECT_RWQ_IND_TBL,
+			    &UVERBS_TYPE_ALLOC_IDR(0, uverbs_free_rwq_ind_tbl));
+
+DECLARE_UVERBS_NAMED_OBJECT(UVERBS_OBJECT_XRCD,
+			    &UVERBS_TYPE_ALLOC_IDR_SZ(sizeof(struct ib_uxrcd_object), 0,
+						      uverbs_free_xrcd));
+
+DECLARE_UVERBS_NAMED_OBJECT(UVERBS_OBJECT_PD,
+			    /* 2 is used in order to free the PD after MRs */
+			    &UVERBS_TYPE_ALLOC_IDR(2, uverbs_free_pd));
+
+DECLARE_UVERBS_NAMED_OBJECT(UVERBS_OBJECT_DEVICE, NULL);
+
+static DECLARE_UVERBS_OBJECT_TREE(uverbs_default_objects,
+				  &UVERBS_OBJECT(UVERBS_OBJECT_DEVICE),
+				  &UVERBS_OBJECT(UVERBS_OBJECT_PD),
+				  &UVERBS_OBJECT(UVERBS_OBJECT_MR),
+				  &UVERBS_OBJECT(UVERBS_OBJECT_COMP_CHANNEL),
+				  &UVERBS_OBJECT(UVERBS_OBJECT_CQ),
+				  &UVERBS_OBJECT(UVERBS_OBJECT_QP),
+				  &UVERBS_OBJECT(UVERBS_OBJECT_AH),
+				  &UVERBS_OBJECT(UVERBS_OBJECT_MW),
+				  &UVERBS_OBJECT(UVERBS_OBJECT_SRQ),
+				  &UVERBS_OBJECT(UVERBS_OBJECT_FLOW),
+				  &UVERBS_OBJECT(UVERBS_OBJECT_WQ),
+				  &UVERBS_OBJECT(UVERBS_OBJECT_RWQ_IND_TBL),
+				  &UVERBS_OBJECT(UVERBS_OBJECT_XRCD),
+				  &UVERBS_OBJECT(UVERBS_OBJECT_FLOW_ACTION),
+				  &UVERBS_OBJECT(UVERBS_OBJECT_DM));
+
+const struct uverbs_object_tree_def *uverbs_default_get_objects(void)
 {
-	struct ib_ucontext *ucontext = file->ucontext;
-	struct ib_ucq_object           *obj;
-	struct ib_udata uhw;
-	int ret;
-	u64 user_handle;
-	struct ib_cq_init_attr attr = {};
-	struct ib_cq                   *cq;
-	struct ib_uverbs_completion_event_file    *ev_file = NULL;
-	const struct uverbs_attr *ev_file_attr;
-	struct ib_uobject *ev_file_uobj;
-
-	if (!(ib_dev->uverbs_cmd_mask & 1ULL << IB_USER_VERBS_CMD_CREATE_CQ))
-		return -EOPNOTSUPP;
-
-	ret = uverbs_copy_from(&attr.comp_vector, attrs, CREATE_CQ_COMP_VECTOR);
-	if (!ret)
-		ret = uverbs_copy_from(&attr.cqe, attrs, CREATE_CQ_CQE);
-	if (!ret)
-		ret = uverbs_copy_from(&user_handle, attrs, CREATE_CQ_USER_HANDLE);
-	if (ret)
-		return ret;
-
-	/* Optional param, if it doesn't exist, we get -ENOENT and skip it */
-	if (uverbs_copy_from(&attr.flags, attrs, CREATE_CQ_FLAGS) == -EFAULT)
-		return -EFAULT;
-
-	ev_file_attr = uverbs_attr_get(attrs, CREATE_CQ_COMP_CHANNEL);
-	if (!IS_ERR(ev_file_attr)) {
-		ev_file_uobj = ev_file_attr->obj_attr.uobject;
-
-		ev_file = container_of(ev_file_uobj,
-				       struct ib_uverbs_completion_event_file,
-				       uobj_file.uobj);
-		uverbs_uobject_get(ev_file_uobj);
-	}
-
-	if (attr.comp_vector >= ucontext->ufile->device->num_comp_vectors) {
-		ret = -EINVAL;
-		goto err_event_file;
-	}
-
-	obj = container_of(uverbs_attr_get(attrs, CREATE_CQ_HANDLE)->obj_attr.uobject,
-			   typeof(*obj), uobject);
-	obj->uverbs_file	   = ucontext->ufile;
-	obj->comp_events_reported  = 0;
-	obj->async_events_reported = 0;
-	INIT_LIST_HEAD(&obj->comp_list);
-	INIT_LIST_HEAD(&obj->async_list);
-
-	/* Temporary, only until drivers get the new uverbs_attr_bundle */
-	create_udata(attrs, &uhw);
-
-	cq = ib_dev->create_cq(ib_dev, &attr, ucontext, &uhw);
-	if (IS_ERR(cq)) {
-		ret = PTR_ERR(cq);
-		goto err_event_file;
-	}
-
-	cq->device        = ib_dev;
-	cq->uobject       = &obj->uobject;
-	cq->comp_handler  = ib_uverbs_comp_handler;
-	cq->event_handler = ib_uverbs_cq_event_handler;
-	cq->cq_context    = ev_file ? &ev_file->ev_queue : NULL;
-	obj->uobject.object = cq;
-	obj->uobject.user_handle = user_handle;
-	atomic_set(&cq->usecnt, 0);
-	cq->res.type = RDMA_RESTRACK_CQ;
-	rdma_restrack_add(&cq->res);
-
-	ret = uverbs_copy_to(attrs, CREATE_CQ_RESP_CQE, &cq->cqe,
-			     sizeof(cq->cqe));
-	if (ret)
-		goto err_cq;
-
-	return 0;
-err_cq:
-	ib_destroy_cq(cq);
-
-err_event_file:
-	if (ev_file)
-		uverbs_uobject_put(ev_file_uobj);
-	return ret;
-};
-
-static DECLARE_UVERBS_METHOD(
-	uverbs_method_cq_create, UVERBS_CQ_CREATE, uverbs_create_cq_handler,
-	&UVERBS_ATTR_IDR(CREATE_CQ_HANDLE, UVERBS_OBJECT_CQ, UVERBS_ACCESS_NEW,
-			 UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)),
-	&UVERBS_ATTR_PTR_IN(CREATE_CQ_CQE, u32,
-			    UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)),
-	&UVERBS_ATTR_PTR_IN(CREATE_CQ_USER_HANDLE, u64,
-			    UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)),
-	&UVERBS_ATTR_FD(CREATE_CQ_COMP_CHANNEL, UVERBS_OBJECT_COMP_CHANNEL,
-			UVERBS_ACCESS_READ),
-	&UVERBS_ATTR_PTR_IN(CREATE_CQ_COMP_VECTOR, u32,
-			    UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)),
-	&UVERBS_ATTR_PTR_IN(CREATE_CQ_FLAGS, u32),
-	&UVERBS_ATTR_PTR_OUT(CREATE_CQ_RESP_CQE, u32,
-			     UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)),
-	&uverbs_uhw_compat_in, &uverbs_uhw_compat_out);
-
-static int uverbs_destroy_cq_handler(struct ib_device *ib_dev,
-				     struct ib_uverbs_file *file,
-				     struct uverbs_attr_bundle *attrs)
-{
-	struct ib_uverbs_destroy_cq_resp resp;
-	struct ib_uobject *uobj =
-		uverbs_attr_get(attrs, DESTROY_CQ_HANDLE)->obj_attr.uobject;
-	struct ib_ucq_object *obj = container_of(uobj, struct ib_ucq_object,
-						 uobject);
-	int ret;
-
-	if (!(ib_dev->uverbs_cmd_mask & 1ULL << IB_USER_VERBS_CMD_DESTROY_CQ))
-		return -EOPNOTSUPP;
-
-	ret = rdma_explicit_destroy(uobj);
-	if (ret)
-		return ret;
-
-	resp.comp_events_reported  = obj->comp_events_reported;
-	resp.async_events_reported = obj->async_events_reported;
-
-	return uverbs_copy_to(attrs, DESTROY_CQ_RESP, &resp, sizeof(resp));
+	return &uverbs_default_objects;
 }
-
-static DECLARE_UVERBS_METHOD(
-	uverbs_method_cq_destroy, UVERBS_CQ_DESTROY, uverbs_destroy_cq_handler,
-	&UVERBS_ATTR_IDR(DESTROY_CQ_HANDLE, UVERBS_OBJECT_CQ,
-			 UVERBS_ACCESS_DESTROY,
-			 UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)),
-	&UVERBS_ATTR_PTR_OUT(DESTROY_CQ_RESP, struct ib_uverbs_destroy_cq_resp,
-			     UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)));
-
-DECLARE_UVERBS_OBJECT(uverbs_object_comp_channel,
-		      UVERBS_OBJECT_COMP_CHANNEL,
-		      &UVERBS_TYPE_ALLOC_FD(0,
-					      sizeof(struct ib_uverbs_completion_event_file),
-					      uverbs_hot_unplug_completion_event_file,
-					      &uverbs_event_fops,
-					      "[infinibandevent]", O_RDONLY));
-
-DECLARE_UVERBS_OBJECT(uverbs_object_cq, UVERBS_OBJECT_CQ,
-		      &UVERBS_TYPE_ALLOC_IDR_SZ(sizeof(struct ib_ucq_object), 0,
-						  uverbs_free_cq),
-		      &uverbs_method_cq_create,
-		      &uverbs_method_cq_destroy);
-
-DECLARE_UVERBS_OBJECT(uverbs_object_qp, UVERBS_OBJECT_QP,
-		      &UVERBS_TYPE_ALLOC_IDR_SZ(sizeof(struct ib_uqp_object), 0,
-						  uverbs_free_qp));
-
-DECLARE_UVERBS_OBJECT(uverbs_object_mw, UVERBS_OBJECT_MW,
-		      &UVERBS_TYPE_ALLOC_IDR(0, uverbs_free_mw));
-
-DECLARE_UVERBS_OBJECT(uverbs_object_mr, UVERBS_OBJECT_MR,
-		      /* 1 is used in order to free the MR after all the MWs */
-		      &UVERBS_TYPE_ALLOC_IDR(1, uverbs_free_mr));
-
-DECLARE_UVERBS_OBJECT(uverbs_object_srq, UVERBS_OBJECT_SRQ,
-		      &UVERBS_TYPE_ALLOC_IDR_SZ(sizeof(struct ib_usrq_object), 0,
-						  uverbs_free_srq));
-
-DECLARE_UVERBS_OBJECT(uverbs_object_ah, UVERBS_OBJECT_AH,
-		      &UVERBS_TYPE_ALLOC_IDR(0, uverbs_free_ah));
-
-DECLARE_UVERBS_OBJECT(uverbs_object_flow, UVERBS_OBJECT_FLOW,
-		      &UVERBS_TYPE_ALLOC_IDR(0, uverbs_free_flow));
-
-DECLARE_UVERBS_OBJECT(uverbs_object_wq, UVERBS_OBJECT_WQ,
-		      &UVERBS_TYPE_ALLOC_IDR_SZ(sizeof(struct ib_uwq_object), 0,
-						  uverbs_free_wq));
-
-DECLARE_UVERBS_OBJECT(uverbs_object_rwq_ind_table,
-		      UVERBS_OBJECT_RWQ_IND_TBL,
-		      &UVERBS_TYPE_ALLOC_IDR(0, uverbs_free_rwq_ind_tbl));
-
-DECLARE_UVERBS_OBJECT(uverbs_object_xrcd, UVERBS_OBJECT_XRCD,
-		      &UVERBS_TYPE_ALLOC_IDR_SZ(sizeof(struct ib_uxrcd_object), 0,
-						  uverbs_free_xrcd));
-
-DECLARE_UVERBS_OBJECT(uverbs_object_pd, UVERBS_OBJECT_PD,
-		      /* 2 is used in order to free the PD after MRs */
-		      &UVERBS_TYPE_ALLOC_IDR(2, uverbs_free_pd));
-
-DECLARE_UVERBS_OBJECT(uverbs_object_device, UVERBS_OBJECT_DEVICE, NULL);
-
-DECLARE_UVERBS_OBJECT_TREE(uverbs_default_objects,
-			   &uverbs_object_device,
-			   &uverbs_object_pd,
-			   &uverbs_object_mr,
-			   &uverbs_object_comp_channel,
-			   &uverbs_object_cq,
-			   &uverbs_object_qp,
-			   &uverbs_object_ah,
-			   &uverbs_object_mw,
-			   &uverbs_object_srq,
-			   &uverbs_object_flow,
-			   &uverbs_object_wq,
-			   &uverbs_object_rwq_ind_table,
-			   &uverbs_object_xrcd);
+EXPORT_SYMBOL_GPL(uverbs_default_get_objects);

drivers/infiniband/core/uverbs_std_types_cq.c

@@ -0,0 +1,210 @@
+/*
+ * Copyright (c) 2017, Mellanox Technologies inc.  All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <rdma/uverbs_std_types.h>
+#include "rdma_core.h"
+#include "uverbs.h"
+
+static int uverbs_free_cq(struct ib_uobject *uobject,
+			  enum rdma_remove_reason why)
+{
+	struct ib_cq *cq = uobject->object;
+	struct ib_uverbs_event_queue *ev_queue = cq->cq_context;
+	struct ib_ucq_object *ucq =
+		container_of(uobject, struct ib_ucq_object, uobject);
+	int ret;
+
+	ret = ib_destroy_cq(cq);
+	if (!ret || why != RDMA_REMOVE_DESTROY)
+		ib_uverbs_release_ucq(uobject->context->ufile, ev_queue ?
+				      container_of(ev_queue,
+						   struct ib_uverbs_completion_event_file,
+						   ev_queue) : NULL,
+				      ucq);
+	return ret;
+}
+
+static int UVERBS_HANDLER(UVERBS_METHOD_CQ_CREATE)(struct ib_device *ib_dev,
+						   struct ib_uverbs_file *file,
+						   struct uverbs_attr_bundle *attrs)
+{
+	struct ib_ucontext *ucontext = file->ucontext;
+	struct ib_ucq_object           *obj;
+	struct ib_udata uhw;
+	int ret;
+	u64 user_handle;
+	struct ib_cq_init_attr attr = {};
+	struct ib_cq                   *cq;
+	struct ib_uverbs_completion_event_file    *ev_file = NULL;
+	const struct uverbs_attr *ev_file_attr;
+	struct ib_uobject *ev_file_uobj;
+
+	if (!(ib_dev->uverbs_cmd_mask & 1ULL << IB_USER_VERBS_CMD_CREATE_CQ))
+		return -EOPNOTSUPP;
+
+	ret = uverbs_copy_from(&attr.comp_vector, attrs,
+			       UVERBS_ATTR_CREATE_CQ_COMP_VECTOR);
+	if (!ret)
+		ret = uverbs_copy_from(&attr.cqe, attrs,
+				       UVERBS_ATTR_CREATE_CQ_CQE);
+	if (!ret)
+		ret = uverbs_copy_from(&user_handle, attrs,
+				       UVERBS_ATTR_CREATE_CQ_USER_HANDLE);
+	if (ret)
+		return ret;
+
+	/* Optional param, if it doesn't exist, we get -ENOENT and skip it */
+	if (IS_UVERBS_COPY_ERR(uverbs_copy_from(&attr.flags, attrs,
+						UVERBS_ATTR_CREATE_CQ_FLAGS)))
+		return -EFAULT;
+
+	ev_file_attr = uverbs_attr_get(attrs, UVERBS_ATTR_CREATE_CQ_COMP_CHANNEL);
+	if (!IS_ERR(ev_file_attr)) {
+		ev_file_uobj = ev_file_attr->obj_attr.uobject;
+
+		ev_file = container_of(ev_file_uobj,
+				       struct ib_uverbs_completion_event_file,
+				       uobj_file.uobj);
+		uverbs_uobject_get(ev_file_uobj);
+	}
+
+	if (attr.comp_vector >= ucontext->ufile->device->num_comp_vectors) {
+		ret = -EINVAL;
+		goto err_event_file;
+	}
+
+	obj = container_of(uverbs_attr_get(attrs,
+					   UVERBS_ATTR_CREATE_CQ_HANDLE)->obj_attr.uobject,
+			   typeof(*obj), uobject);
+	obj->uverbs_file	   = ucontext->ufile;
+	obj->comp_events_reported  = 0;
+	obj->async_events_reported = 0;
+	INIT_LIST_HEAD(&obj->comp_list);
+	INIT_LIST_HEAD(&obj->async_list);
+
+	/* Temporary, only until drivers get the new uverbs_attr_bundle */
+	create_udata(attrs, &uhw);
+
+	cq = ib_dev->create_cq(ib_dev, &attr, ucontext, &uhw);
+	if (IS_ERR(cq)) {
+		ret = PTR_ERR(cq);
+		goto err_event_file;
+	}
+
+	cq->device        = ib_dev;
+	cq->uobject       = &obj->uobject;
+	cq->comp_handler  = ib_uverbs_comp_handler;
+	cq->event_handler = ib_uverbs_cq_event_handler;
+	cq->cq_context    = ev_file ? &ev_file->ev_queue : NULL;
+	obj->uobject.object = cq;
+	obj->uobject.user_handle = user_handle;
+	atomic_set(&cq->usecnt, 0);
+	cq->res.type = RDMA_RESTRACK_CQ;
+	rdma_restrack_add(&cq->res);
+
+	ret = uverbs_copy_to(attrs, UVERBS_ATTR_CREATE_CQ_RESP_CQE, &cq->cqe,
+			     sizeof(cq->cqe));
+	if (ret)
+		goto err_cq;
+
+	return 0;
+err_cq:
+	ib_destroy_cq(cq);
+
+err_event_file:
+	if (ev_file)
+		uverbs_uobject_put(ev_file_uobj);
+	return ret;
+};
+
+static DECLARE_UVERBS_NAMED_METHOD(UVERBS_METHOD_CQ_CREATE,
+	&UVERBS_ATTR_IDR(UVERBS_ATTR_CREATE_CQ_HANDLE, UVERBS_OBJECT_CQ,
+			 UVERBS_ACCESS_NEW,
+			 UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)),
+	&UVERBS_ATTR_PTR_IN(UVERBS_ATTR_CREATE_CQ_CQE,
+			    UVERBS_ATTR_TYPE(u32),
+			    UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)),
+	&UVERBS_ATTR_PTR_IN(UVERBS_ATTR_CREATE_CQ_USER_HANDLE,
+			    UVERBS_ATTR_TYPE(u64),
+			    UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)),
+	&UVERBS_ATTR_FD(UVERBS_ATTR_CREATE_CQ_COMP_CHANNEL,
+			UVERBS_OBJECT_COMP_CHANNEL,
+			UVERBS_ACCESS_READ),
+	&UVERBS_ATTR_PTR_IN(UVERBS_ATTR_CREATE_CQ_COMP_VECTOR, UVERBS_ATTR_TYPE(u32),
+			    UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)),
+	&UVERBS_ATTR_PTR_IN(UVERBS_ATTR_CREATE_CQ_FLAGS, UVERBS_ATTR_TYPE(u32)),
+	&UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_CREATE_CQ_RESP_CQE, UVERBS_ATTR_TYPE(u32),
+			     UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)),
+	&uverbs_uhw_compat_in, &uverbs_uhw_compat_out);
+
+static int UVERBS_HANDLER(UVERBS_METHOD_CQ_DESTROY)(struct ib_device *ib_dev,
+						    struct ib_uverbs_file *file,
+						    struct uverbs_attr_bundle *attrs)
+{
+	struct ib_uverbs_destroy_cq_resp resp;
+	struct ib_uobject *uobj =
+		uverbs_attr_get(attrs, UVERBS_ATTR_DESTROY_CQ_HANDLE)->obj_attr.uobject;
+	struct ib_ucq_object *obj = container_of(uobj, struct ib_ucq_object,
+						 uobject);
+	int ret;
+
+	if (!(ib_dev->uverbs_cmd_mask & 1ULL << IB_USER_VERBS_CMD_DESTROY_CQ))
+		return -EOPNOTSUPP;
+
+	ret = rdma_explicit_destroy(uobj);
+	if (ret)
+		return ret;
+
+	resp.comp_events_reported  = obj->comp_events_reported;
+	resp.async_events_reported = obj->async_events_reported;
+
+	return uverbs_copy_to(attrs, UVERBS_ATTR_DESTROY_CQ_RESP, &resp,
+			      sizeof(resp));
+}
+
+static DECLARE_UVERBS_NAMED_METHOD(UVERBS_METHOD_CQ_DESTROY,
+	&UVERBS_ATTR_IDR(UVERBS_ATTR_DESTROY_CQ_HANDLE, UVERBS_OBJECT_CQ,
+			 UVERBS_ACCESS_DESTROY,
+			 UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)),
+	&UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_DESTROY_CQ_RESP,
+			     UVERBS_ATTR_TYPE(struct ib_uverbs_destroy_cq_resp),
+			     UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)));
+
+DECLARE_UVERBS_NAMED_OBJECT(UVERBS_OBJECT_CQ,
+			    &UVERBS_TYPE_ALLOC_IDR_SZ(sizeof(struct ib_ucq_object), 0,
+						      uverbs_free_cq),
+#if IS_ENABLED(CONFIG_INFINIBAND_EXP_LEGACY_VERBS_NEW_UAPI)
+			    &UVERBS_METHOD(UVERBS_METHOD_CQ_CREATE),
+			    &UVERBS_METHOD(UVERBS_METHOD_CQ_DESTROY)
+#endif
+			   );
+

drivers/infiniband/core/uverbs_std_types_dm.c

@@ -0,0 +1,108 @@
+/*
+ * Copyright (c) 2018, Mellanox Technologies inc.  All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include "uverbs.h"
+#include <rdma/uverbs_std_types.h>
+
+static int uverbs_free_dm(struct ib_uobject *uobject,
+			  enum rdma_remove_reason why)
+{
+	struct ib_dm *dm = uobject->object;
+
+	if (why == RDMA_REMOVE_DESTROY && atomic_read(&dm->usecnt))
+		return -EBUSY;
+
+	return dm->device->dealloc_dm(dm);
+}
+
+static int UVERBS_HANDLER(UVERBS_METHOD_DM_ALLOC)(struct ib_device *ib_dev,
+						  struct ib_uverbs_file *file,
+						  struct uverbs_attr_bundle *attrs)
+{
+	struct ib_ucontext *ucontext = file->ucontext;
+	struct ib_dm_alloc_attr attr = {};
+	struct ib_uobject *uobj;
+	struct ib_dm *dm;
+	int ret;
+
+	if (!ib_dev->alloc_dm)
+		return -EOPNOTSUPP;
+
+	ret = uverbs_copy_from(&attr.length, attrs,
+			       UVERBS_ATTR_ALLOC_DM_LENGTH);
+	if (ret)
+		return ret;
+
+	ret = uverbs_copy_from(&attr.alignment, attrs,
+			       UVERBS_ATTR_ALLOC_DM_ALIGNMENT);
+	if (ret)
+		return ret;
+
+	uobj = uverbs_attr_get(attrs, UVERBS_ATTR_ALLOC_DM_HANDLE)->obj_attr.uobject;
+
+	dm = ib_dev->alloc_dm(ib_dev, ucontext, &attr, attrs);
+	if (IS_ERR(dm))
+		return PTR_ERR(dm);
+
+	dm->device  = ib_dev;
+	dm->length  = attr.length;
+	dm->uobject = uobj;
+	atomic_set(&dm->usecnt, 0);
+
+	uobj->object = dm;
+
+	return 0;
+}
+
+static DECLARE_UVERBS_NAMED_METHOD(UVERBS_METHOD_DM_ALLOC,
+	&UVERBS_ATTR_IDR(UVERBS_ATTR_ALLOC_DM_HANDLE, UVERBS_OBJECT_DM,
+			 UVERBS_ACCESS_NEW,
+			 UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)),
+	&UVERBS_ATTR_PTR_IN(UVERBS_ATTR_ALLOC_DM_LENGTH,
+			    UVERBS_ATTR_TYPE(u64),
+			    UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)),
+	&UVERBS_ATTR_PTR_IN(UVERBS_ATTR_ALLOC_DM_ALIGNMENT,
+			    UVERBS_ATTR_TYPE(u32),
+			    UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)));
+
+static DECLARE_UVERBS_NAMED_METHOD_WITH_HANDLER(UVERBS_METHOD_DM_FREE,
+	uverbs_destroy_def_handler,
+	&UVERBS_ATTR_IDR(UVERBS_ATTR_FREE_DM_HANDLE,
+			 UVERBS_OBJECT_DM,
+			 UVERBS_ACCESS_DESTROY,
+			 UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)));
+
+DECLARE_UVERBS_NAMED_OBJECT(UVERBS_OBJECT_DM,
+			    /* 1 is used in order to free the DM after MRs */
+			    &UVERBS_TYPE_ALLOC_IDR(1, uverbs_free_dm),
+			    &UVERBS_METHOD(UVERBS_METHOD_DM_ALLOC),
+			    &UVERBS_METHOD(UVERBS_METHOD_DM_FREE));

drivers/infiniband/core/uverbs_std_types_flow_action.c

@@ -0,0 +1,435 @@
+/*
+ * Copyright (c) 2018, Mellanox Technologies inc.  All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include "uverbs.h"
+#include <rdma/uverbs_std_types.h>
+
+static int uverbs_free_flow_action(struct ib_uobject *uobject,
+				   enum rdma_remove_reason why)
+{
+	struct ib_flow_action *action = uobject->object;
+
+	if (why == RDMA_REMOVE_DESTROY &&
+	    atomic_read(&action->usecnt))
+		return -EBUSY;
+
+	return action->device->destroy_flow_action(action);
+}
+
+static u64 esp_flags_uverbs_to_verbs(struct uverbs_attr_bundle *attrs,
+				     u32 flags, bool is_modify)
+{
+	u64 verbs_flags = flags;
+
+	if (uverbs_attr_is_valid(attrs, UVERBS_ATTR_FLOW_ACTION_ESP_ESN))
+		verbs_flags |= IB_FLOW_ACTION_ESP_FLAGS_ESN_TRIGGERED;
+
+	if (is_modify && uverbs_attr_is_valid(attrs,
+					      UVERBS_ATTR_FLOW_ACTION_ESP_ATTRS))
+		verbs_flags |= IB_FLOW_ACTION_ESP_FLAGS_MOD_ESP_ATTRS;
+
+	return verbs_flags;
+};
+
+static int validate_flow_action_esp_keymat_aes_gcm(struct ib_flow_action_attrs_esp_keymats *keymat)
+{
+	struct ib_uverbs_flow_action_esp_keymat_aes_gcm *aes_gcm =
+		&keymat->keymat.aes_gcm;
+
+	if (aes_gcm->iv_algo > IB_UVERBS_FLOW_ACTION_IV_ALGO_SEQ)
+		return -EOPNOTSUPP;
+
+	if (aes_gcm->key_len != 32 &&
+	    aes_gcm->key_len != 24 &&
+	    aes_gcm->key_len != 16)
+		return -EINVAL;
+
+	if (aes_gcm->icv_len != 16 &&
+	    aes_gcm->icv_len != 8 &&
+	    aes_gcm->icv_len != 12)
+		return -EINVAL;
+
+	return 0;
+}
+
+static int (* const flow_action_esp_keymat_validate[])(struct ib_flow_action_attrs_esp_keymats *keymat) = {
+	[IB_UVERBS_FLOW_ACTION_ESP_KEYMAT_AES_GCM] = validate_flow_action_esp_keymat_aes_gcm,
+};
+
+static int flow_action_esp_replay_none(struct ib_flow_action_attrs_esp_replays *replay,
+				       bool is_modify)
+{
+	/* This is used in order to modify an esp flow action with an enabled
+	 * replay protection to a disabled one. This is only supported via
+	 * modify, as in create verb we can simply drop the REPLAY attribute and
+	 * achieve the same thing.
+	 */
+	return is_modify ? 0 : -EINVAL;
+}
+
+static int flow_action_esp_replay_def_ok(struct ib_flow_action_attrs_esp_replays *replay,
+					 bool is_modify)
+{
+	/* Some replay protections could always be enabled without validating
+	 * anything.
+	 */
+	return 0;
+}
+
+static int (* const flow_action_esp_replay_validate[])(struct ib_flow_action_attrs_esp_replays *replay,
+						       bool is_modify) = {
+	[IB_UVERBS_FLOW_ACTION_ESP_REPLAY_NONE] = flow_action_esp_replay_none,
+	[IB_UVERBS_FLOW_ACTION_ESP_REPLAY_BMP] = flow_action_esp_replay_def_ok,
+};
+
+static int parse_esp_ip(enum ib_flow_spec_type proto,
+			const void __user *val_ptr,
+			size_t len, union ib_flow_spec *out)
+{
+	int ret;
+	const struct ib_uverbs_flow_ipv4_filter ipv4 = {
+		.src_ip = cpu_to_be32(0xffffffffUL),
+		.dst_ip = cpu_to_be32(0xffffffffUL),
+		.proto = 0xff,
+		.tos = 0xff,
+		.ttl = 0xff,
+		.flags = 0xff,
+	};
+	const struct ib_uverbs_flow_ipv6_filter ipv6 = {
+		.src_ip = {0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
+			   0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff},
+		.dst_ip = {0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
+			   0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff},
+		.flow_label = cpu_to_be32(0xffffffffUL),
+		.next_hdr = 0xff,
+		.traffic_class = 0xff,
+		.hop_limit = 0xff,
+	};
+	union {
+		struct ib_uverbs_flow_ipv4_filter ipv4;
+		struct ib_uverbs_flow_ipv6_filter ipv6;
+	} user_val = {};
+	const void *user_pmask;
+	size_t val_len;
+
+	/* If the flow IPv4/IPv6 flow specifications are extended, the mask
+	 * should be changed as well.
+	 */
+	BUILD_BUG_ON(offsetof(struct ib_uverbs_flow_ipv4_filter, flags) +
+		     sizeof(ipv4.flags) != sizeof(ipv4));
+	BUILD_BUG_ON(offsetof(struct ib_uverbs_flow_ipv6_filter, reserved) +
+		     sizeof(ipv6.reserved) != sizeof(ipv6));
+
+	switch (proto) {
+	case IB_FLOW_SPEC_IPV4:
+		if (len > sizeof(user_val.ipv4) &&
+		    !ib_is_buffer_cleared(val_ptr + sizeof(user_val.ipv4),
+					  len - sizeof(user_val.ipv4)))
+			return -EOPNOTSUPP;
+
+		val_len = min_t(size_t, len, sizeof(user_val.ipv4));
+		ret = copy_from_user(&user_val.ipv4, val_ptr,
+				     val_len);
+		if (ret)
+			return -EFAULT;
+
+		user_pmask = &ipv4;
+		break;
+	case IB_FLOW_SPEC_IPV6:
+		if (len > sizeof(user_val.ipv6) &&
+		    !ib_is_buffer_cleared(val_ptr + sizeof(user_val.ipv6),
+					  len - sizeof(user_val.ipv6)))
+			return -EOPNOTSUPP;
+
+		val_len = min_t(size_t, len, sizeof(user_val.ipv6));
+		ret = copy_from_user(&user_val.ipv6, val_ptr,
+				     val_len);
+		if (ret)
+			return -EFAULT;
+
+		user_pmask = &ipv6;
+		break;
+	default:
+		return -EOPNOTSUPP;
+	}
+
+	return ib_uverbs_kern_spec_to_ib_spec_filter(proto, user_pmask,
+						     &user_val,
+						     val_len, out);
+}
+
+static int flow_action_esp_get_encap(struct ib_flow_spec_list *out,
+				     struct uverbs_attr_bundle *attrs)
+{
+	struct ib_uverbs_flow_action_esp_encap uverbs_encap;
+	int ret;
+
+	ret = uverbs_copy_from(&uverbs_encap, attrs,
+			       UVERBS_ATTR_FLOW_ACTION_ESP_ENCAP);
+	if (ret)
+		return ret;
+
+	/* We currently support only one encap */
+	if (uverbs_encap.next_ptr)
+		return -EOPNOTSUPP;
+
+	if (uverbs_encap.type != IB_FLOW_SPEC_IPV4 &&
+	    uverbs_encap.type != IB_FLOW_SPEC_IPV6)
+		return -EOPNOTSUPP;
+
+	return parse_esp_ip(uverbs_encap.type,
+			    u64_to_user_ptr(uverbs_encap.val_ptr),
+			    uverbs_encap.len,
+			    &out->spec);
+}
+
+struct ib_flow_action_esp_attr {
+	struct	ib_flow_action_attrs_esp		hdr;
+	struct	ib_flow_action_attrs_esp_keymats	keymat;
+	struct	ib_flow_action_attrs_esp_replays	replay;
+	/* We currently support only one spec */
+	struct	ib_flow_spec_list			encap;
+};
+
+#define ESP_LAST_SUPPORTED_FLAG		IB_UVERBS_FLOW_ACTION_ESP_FLAGS_ESN_NEW_WINDOW
+static int parse_flow_action_esp(struct ib_device *ib_dev,
+				 struct ib_uverbs_file *file,
+				 struct uverbs_attr_bundle *attrs,
+				 struct ib_flow_action_esp_attr *esp_attr,
+				 bool is_modify)
+{
+	struct ib_uverbs_flow_action_esp uverbs_esp = {};
+	int ret;
+
+	/* Optional param, if it doesn't exist, we get -ENOENT and skip it */
+	ret = uverbs_copy_from(&esp_attr->hdr.esn, attrs,
+			       UVERBS_ATTR_FLOW_ACTION_ESP_ESN);
+	if (IS_UVERBS_COPY_ERR(ret))
+		return ret;
+
+	/* This can be called from FLOW_ACTION_ESP_MODIFY where
+	 * UVERBS_ATTR_FLOW_ACTION_ESP_ATTRS is optional
+	 */
+	if (uverbs_attr_is_valid(attrs, UVERBS_ATTR_FLOW_ACTION_ESP_ATTRS)) {
+		ret = uverbs_copy_from_or_zero(&uverbs_esp, attrs,
+					       UVERBS_ATTR_FLOW_ACTION_ESP_ATTRS);
+		if (ret)
+			return ret;
+
+		if (uverbs_esp.flags & ~((ESP_LAST_SUPPORTED_FLAG << 1) - 1))
+			return -EOPNOTSUPP;
+
+		esp_attr->hdr.spi = uverbs_esp.spi;
+		esp_attr->hdr.seq = uverbs_esp.seq;
+		esp_attr->hdr.tfc_pad = uverbs_esp.tfc_pad;
+		esp_attr->hdr.hard_limit_pkts = uverbs_esp.hard_limit_pkts;
+	}
+	esp_attr->hdr.flags = esp_flags_uverbs_to_verbs(attrs, uverbs_esp.flags,
+							is_modify);
+
+	if (uverbs_attr_is_valid(attrs, UVERBS_ATTR_FLOW_ACTION_ESP_KEYMAT)) {
+		esp_attr->keymat.protocol =
+			uverbs_attr_get_enum_id(attrs,
+						UVERBS_ATTR_FLOW_ACTION_ESP_KEYMAT);
+		ret = uverbs_copy_from_or_zero(&esp_attr->keymat.keymat,
+					       attrs,
+					       UVERBS_ATTR_FLOW_ACTION_ESP_KEYMAT);
+		if (ret)
+			return ret;
+
+		ret = flow_action_esp_keymat_validate[esp_attr->keymat.protocol](&esp_attr->keymat);
+		if (ret)
+			return ret;
+
+		esp_attr->hdr.keymat = &esp_attr->keymat;
+	}
+
+	if (uverbs_attr_is_valid(attrs, UVERBS_ATTR_FLOW_ACTION_ESP_REPLAY)) {
+		esp_attr->replay.protocol =
+			uverbs_attr_get_enum_id(attrs,
+						UVERBS_ATTR_FLOW_ACTION_ESP_REPLAY);
+
+		ret = uverbs_copy_from_or_zero(&esp_attr->replay.replay,
+					       attrs,
+					       UVERBS_ATTR_FLOW_ACTION_ESP_REPLAY);
+		if (ret)
+			return ret;
+
+		ret = flow_action_esp_replay_validate[esp_attr->replay.protocol](&esp_attr->replay,
+										 is_modify);
+		if (ret)
+			return ret;
+
+		esp_attr->hdr.replay = &esp_attr->replay;
+	}
+
+	if (uverbs_attr_is_valid(attrs, UVERBS_ATTR_FLOW_ACTION_ESP_ENCAP)) {
+		ret = flow_action_esp_get_encap(&esp_attr->encap, attrs);
+		if (ret)
+			return ret;
+
+		esp_attr->hdr.encap = &esp_attr->encap;
+	}
+
+	return 0;
+}
+
+static int UVERBS_HANDLER(UVERBS_METHOD_FLOW_ACTION_ESP_CREATE)(struct ib_device *ib_dev,
+								struct ib_uverbs_file *file,
+								struct uverbs_attr_bundle *attrs)
+{
+	int				  ret;
+	struct ib_uobject		  *uobj;
+	struct ib_flow_action		  *action;
+	struct ib_flow_action_esp_attr	  esp_attr = {};
+
+	if (!ib_dev->create_flow_action_esp)
+		return -EOPNOTSUPP;
+
+	ret = parse_flow_action_esp(ib_dev, file, attrs, &esp_attr, false);
+	if (ret)
+		return ret;
+
+	/* No need to check as this attribute is marked as MANDATORY */
+	uobj = uverbs_attr_get(attrs, UVERBS_ATTR_FLOW_ACTION_ESP_HANDLE)->obj_attr.uobject;
+	action = ib_dev->create_flow_action_esp(ib_dev, &esp_attr.hdr, attrs);
+	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;
+
+	return 0;
+}
+
+static int UVERBS_HANDLER(UVERBS_METHOD_FLOW_ACTION_ESP_MODIFY)(struct ib_device *ib_dev,
+								struct ib_uverbs_file *file,
+								struct uverbs_attr_bundle *attrs)
+{
+	int				  ret;
+	struct ib_uobject		  *uobj;
+	struct ib_flow_action		  *action;
+	struct ib_flow_action_esp_attr	  esp_attr = {};
+
+	if (!ib_dev->modify_flow_action_esp)
+		return -EOPNOTSUPP;
+
+	ret = parse_flow_action_esp(ib_dev, file, attrs, &esp_attr, true);
+	if (ret)
+		return ret;
+
+	uobj = uverbs_attr_get(attrs, UVERBS_ATTR_FLOW_ACTION_ESP_HANDLE)->obj_attr.uobject;
+	action = uobj->object;
+
+	if (action->type != IB_FLOW_ACTION_ESP)
+		return -EINVAL;
+
+	return ib_dev->modify_flow_action_esp(action,
+					      &esp_attr.hdr,
+					      attrs);
+}
+
+static const struct uverbs_attr_spec uverbs_flow_action_esp_keymat[] = {
+	[IB_UVERBS_FLOW_ACTION_ESP_KEYMAT_AES_GCM] = {
+		.ptr = {
+			.type = UVERBS_ATTR_TYPE_PTR_IN,
+			UVERBS_ATTR_TYPE(struct ib_uverbs_flow_action_esp_keymat_aes_gcm),
+			.flags = UVERBS_ATTR_SPEC_F_MIN_SZ_OR_ZERO,
+		},
+	},
+};
+
+static const struct uverbs_attr_spec uverbs_flow_action_esp_replay[] = {
+	[IB_UVERBS_FLOW_ACTION_ESP_REPLAY_NONE] = {
+		.ptr = {
+			.type = UVERBS_ATTR_TYPE_PTR_IN,
+			/* No need to specify any data */
+			.len = 0,
+		}
+	},
+	[IB_UVERBS_FLOW_ACTION_ESP_REPLAY_BMP] = {
+		.ptr = {
+			.type = UVERBS_ATTR_TYPE_PTR_IN,
+			UVERBS_ATTR_STRUCT(struct ib_uverbs_flow_action_esp_replay_bmp, size),
+			.flags = UVERBS_ATTR_SPEC_F_MIN_SZ_OR_ZERO,
+		}
+	},
+};
+
+static DECLARE_UVERBS_NAMED_METHOD(UVERBS_METHOD_FLOW_ACTION_ESP_CREATE,
+	&UVERBS_ATTR_IDR(UVERBS_ATTR_FLOW_ACTION_ESP_HANDLE, UVERBS_OBJECT_FLOW_ACTION,
+			 UVERBS_ACCESS_NEW,
+			 UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)),
+	&UVERBS_ATTR_PTR_IN(UVERBS_ATTR_FLOW_ACTION_ESP_ATTRS,
+			    UVERBS_ATTR_STRUCT(struct ib_uverbs_flow_action_esp, hard_limit_pkts),
+			    UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY |
+				     UVERBS_ATTR_SPEC_F_MIN_SZ_OR_ZERO)),
+	&UVERBS_ATTR_PTR_IN(UVERBS_ATTR_FLOW_ACTION_ESP_ESN, UVERBS_ATTR_TYPE(__u32)),
+	&UVERBS_ATTR_ENUM_IN(UVERBS_ATTR_FLOW_ACTION_ESP_KEYMAT,
+			     uverbs_flow_action_esp_keymat,
+			     UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)),
+	&UVERBS_ATTR_ENUM_IN(UVERBS_ATTR_FLOW_ACTION_ESP_REPLAY,
+			     uverbs_flow_action_esp_replay),
+	&UVERBS_ATTR_PTR_IN(UVERBS_ATTR_FLOW_ACTION_ESP_ENCAP,
+			    UVERBS_ATTR_STRUCT(struct ib_uverbs_flow_action_esp_encap, type)));
+
+static DECLARE_UVERBS_NAMED_METHOD(UVERBS_METHOD_FLOW_ACTION_ESP_MODIFY,
+	&UVERBS_ATTR_IDR(UVERBS_ATTR_FLOW_ACTION_ESP_HANDLE, UVERBS_OBJECT_FLOW_ACTION,
+			 UVERBS_ACCESS_WRITE,
+			 UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)),
+	&UVERBS_ATTR_PTR_IN(UVERBS_ATTR_FLOW_ACTION_ESP_ATTRS,
+			    UVERBS_ATTR_STRUCT(struct ib_uverbs_flow_action_esp, hard_limit_pkts),
+			    UA_FLAGS(UVERBS_ATTR_SPEC_F_MIN_SZ_OR_ZERO)),
+	&UVERBS_ATTR_PTR_IN(UVERBS_ATTR_FLOW_ACTION_ESP_ESN, UVERBS_ATTR_TYPE(__u32)),
+	&UVERBS_ATTR_ENUM_IN(UVERBS_ATTR_FLOW_ACTION_ESP_KEYMAT,
+			     uverbs_flow_action_esp_keymat),
+	&UVERBS_ATTR_ENUM_IN(UVERBS_ATTR_FLOW_ACTION_ESP_REPLAY,
+			     uverbs_flow_action_esp_replay),
+	&UVERBS_ATTR_PTR_IN(UVERBS_ATTR_FLOW_ACTION_ESP_ENCAP,
+			    UVERBS_ATTR_STRUCT(struct ib_uverbs_flow_action_esp_encap, type)));
+
+static DECLARE_UVERBS_NAMED_METHOD_WITH_HANDLER(UVERBS_METHOD_FLOW_ACTION_DESTROY,
+	uverbs_destroy_def_handler,
+	&UVERBS_ATTR_IDR(UVERBS_ATTR_DESTROY_FLOW_ACTION_HANDLE,
+			 UVERBS_OBJECT_FLOW_ACTION,
+			 UVERBS_ACCESS_DESTROY,
+			 UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)));
+
+DECLARE_UVERBS_NAMED_OBJECT(UVERBS_OBJECT_FLOW_ACTION,
+			    &UVERBS_TYPE_ALLOC_IDR(0, uverbs_free_flow_action),
+			    &UVERBS_METHOD(UVERBS_METHOD_FLOW_ACTION_ESP_CREATE),
+			    &UVERBS_METHOD(UVERBS_METHOD_FLOW_ACTION_DESTROY),
+			    &UVERBS_METHOD(UVERBS_METHOD_FLOW_ACTION_ESP_MODIFY));
+

drivers/infiniband/core/uverbs_std_types_mr.c

@@ -0,0 +1,147 @@
+/*
+ * Copyright (c) 2018, Mellanox Technologies inc.  All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include "uverbs.h"
+#include <rdma/uverbs_std_types.h>
+
+static int uverbs_free_mr(struct ib_uobject *uobject,
+			  enum rdma_remove_reason why)
+{
+	return ib_dereg_mr((struct ib_mr *)uobject->object);
+}
+
+static int UVERBS_HANDLER(UVERBS_METHOD_DM_MR_REG)(struct ib_device *ib_dev,
+						   struct ib_uverbs_file *file,
+						   struct uverbs_attr_bundle *attrs)
+{
+	struct ib_dm_mr_attr attr = {};
+	struct ib_uobject *uobj;
+	struct ib_dm *dm;
+	struct ib_pd *pd;
+	struct ib_mr *mr;
+	int ret;
+
+	if (!ib_dev->reg_dm_mr)
+		return -EOPNOTSUPP;
+
+	ret = uverbs_copy_from(&attr.offset, attrs, UVERBS_ATTR_REG_DM_MR_OFFSET);
+	if (ret)
+		return ret;
+
+	ret = uverbs_copy_from(&attr.length, attrs,
+			       UVERBS_ATTR_REG_DM_MR_LENGTH);
+	if (ret)
+		return ret;
+
+	ret = uverbs_copy_from(&attr.access_flags, attrs,
+			       UVERBS_ATTR_REG_DM_MR_ACCESS_FLAGS);
+	if (ret)
+		return ret;
+
+	if (!(attr.access_flags & IB_ZERO_BASED))
+		return -EINVAL;
+
+	ret = ib_check_mr_access(attr.access_flags);
+	if (ret)
+		return ret;
+
+	pd = uverbs_attr_get_obj(attrs, UVERBS_ATTR_REG_DM_MR_PD_HANDLE);
+
+	dm = uverbs_attr_get_obj(attrs, UVERBS_ATTR_REG_DM_MR_DM_HANDLE);
+
+	uobj = uverbs_attr_get(attrs, UVERBS_ATTR_REG_DM_MR_HANDLE)->obj_attr.uobject;
+
+	if (attr.offset > dm->length || attr.length > dm->length ||
+	    attr.length > dm->length - attr.offset)
+		return -EINVAL;
+
+	mr = pd->device->reg_dm_mr(pd, dm, &attr, attrs);
+	if (IS_ERR(mr))
+		return PTR_ERR(mr);
+
+	mr->device  = pd->device;
+	mr->pd      = pd;
+	mr->dm      = dm;
+	mr->uobject = uobj;
+	atomic_inc(&pd->usecnt);
+	atomic_inc(&dm->usecnt);
+
+	uobj->object = mr;
+
+	ret = uverbs_copy_to(attrs, UVERBS_ATTR_REG_DM_MR_RESP_LKEY, &mr->lkey,
+			     sizeof(mr->lkey));
+	if (ret)
+		goto err_dereg;
+
+	ret = uverbs_copy_to(attrs, UVERBS_ATTR_REG_DM_MR_RESP_RKEY,
+			     &mr->rkey, sizeof(mr->rkey));
+	if (ret)
+		goto err_dereg;
+
+	return 0;
+
+err_dereg:
+	ib_dereg_mr(mr);
+
+	return ret;
+}
+
+static DECLARE_UVERBS_NAMED_METHOD(UVERBS_METHOD_DM_MR_REG,
+	&UVERBS_ATTR_IDR(UVERBS_ATTR_REG_DM_MR_HANDLE, UVERBS_OBJECT_MR,
+			 UVERBS_ACCESS_NEW,
+			 UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)),
+	&UVERBS_ATTR_PTR_IN(UVERBS_ATTR_REG_DM_MR_OFFSET,
+			    UVERBS_ATTR_TYPE(u64),
+			    UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)),
+	&UVERBS_ATTR_PTR_IN(UVERBS_ATTR_REG_DM_MR_LENGTH,
+			    UVERBS_ATTR_TYPE(u64),
+			    UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)),
+	&UVERBS_ATTR_IDR(UVERBS_ATTR_REG_DM_MR_PD_HANDLE, UVERBS_OBJECT_PD,
+			 UVERBS_ACCESS_READ,
+			 UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)),
+	&UVERBS_ATTR_PTR_IN(UVERBS_ATTR_REG_DM_MR_ACCESS_FLAGS,
+			    UVERBS_ATTR_TYPE(u32),
+			    UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)),
+	&UVERBS_ATTR_IDR(UVERBS_ATTR_REG_DM_MR_DM_HANDLE, UVERBS_OBJECT_DM,
+			 UVERBS_ACCESS_READ,
+			 UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)),
+	&UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_REG_DM_MR_RESP_LKEY,
+			     UVERBS_ATTR_TYPE(u32),
+			     UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)),
+	&UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_REG_DM_MR_RESP_RKEY,
+			     UVERBS_ATTR_TYPE(u32),
+			     UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)));
+
+DECLARE_UVERBS_NAMED_OBJECT(UVERBS_OBJECT_MR,
+			    /* 1 is used in order to free the MR after all the MWs */
+			    &UVERBS_TYPE_ALLOC_IDR(1, uverbs_free_mr),
+			    &UVERBS_METHOD(UVERBS_METHOD_DM_MR_REG));

drivers/infiniband/core/verbs.c

@@ -655,7 +655,7 @@ int rdma_modify_ah(struct ib_ah *ah, struct rdma_ah_attr *ah_attr)
 
 	return ah->device->modify_ah ?
 		ah->device->modify_ah(ah, ah_attr) :
-		-ENOSYS;
+		-EOPNOTSUPP;
 }
 EXPORT_SYMBOL(rdma_modify_ah);
 
@@ -663,7 +663,7 @@ int rdma_query_ah(struct ib_ah *ah, struct rdma_ah_attr *ah_attr)
 {
 	return ah->device->query_ah ?
 		ah->device->query_ah(ah, ah_attr) :
-		-ENOSYS;
+		-EOPNOTSUPP;
 }
 EXPORT_SYMBOL(rdma_query_ah);
 
@@ -689,7 +689,7 @@ struct ib_srq *ib_create_srq(struct ib_pd *pd,
 	struct ib_srq *srq;
 
 	if (!pd->device->create_srq)
-		return ERR_PTR(-ENOSYS);
+		return ERR_PTR(-EOPNOTSUPP);
 
 	srq = pd->device->create_srq(pd, srq_init_attr, NULL);
 
@@ -722,7 +722,7 @@ int ib_modify_srq(struct ib_srq *srq,
 {
 	return srq->device->modify_srq ?
 		srq->device->modify_srq(srq, srq_attr, srq_attr_mask, NULL) :
-		-ENOSYS;
+		-EOPNOTSUPP;
 }
 EXPORT_SYMBOL(ib_modify_srq);
 
@@ -730,7 +730,7 @@ int ib_query_srq(struct ib_srq *srq,
 		 struct ib_srq_attr *srq_attr)
 {
 	return srq->device->query_srq ?
-		srq->device->query_srq(srq, srq_attr) : -ENOSYS;
+		srq->device->query_srq(srq, srq_attr) : -EOPNOTSUPP;
 }
 EXPORT_SYMBOL(ib_query_srq);
 
@@ -1263,34 +1263,30 @@ static const struct {
 	}
 };
 
-int 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)
+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_attr_mask req_param, opt_param;
 
-	if (cur_state  < 0 || cur_state  > IB_QPS_ERR ||
-	    next_state < 0 || next_state > IB_QPS_ERR)
-		return 0;
-
 	if (mask & IB_QP_CUR_STATE  &&
 	    cur_state != IB_QPS_RTR && cur_state != IB_QPS_RTS &&
 	    cur_state != IB_QPS_SQD && cur_state != IB_QPS_SQE)
-		return 0;
+		return false;
 
 	if (!qp_state_table[cur_state][next_state].valid)
-		return 0;
+		return false;
 
 	req_param = qp_state_table[cur_state][next_state].req_param[type];
 	opt_param = qp_state_table[cur_state][next_state].opt_param[type];
 
 	if ((mask & req_param) != req_param)
-		return 0;
+		return false;
 
 	if (mask & ~(req_param | opt_param | IB_QP_STATE))
-		return 0;
+		return false;
 
-	return 1;
+	return true;
 }
 EXPORT_SYMBOL(ib_modify_qp_is_ok);
 
@@ -1457,7 +1453,7 @@ int ib_query_qp(struct ib_qp *qp,
 {
 	return qp->device->query_qp ?
 		qp->device->query_qp(qp->real_qp, qp_attr, qp_attr_mask, qp_init_attr) :
-		-ENOSYS;
+		-EOPNOTSUPP;
 }
 EXPORT_SYMBOL(ib_query_qp);
 
@@ -1594,7 +1590,7 @@ EXPORT_SYMBOL(ib_create_cq);
 int rdma_set_cq_moderation(struct ib_cq *cq, u16 cq_count, u16 cq_period)
 {
 	return cq->device->modify_cq ?
-		cq->device->modify_cq(cq, cq_count, cq_period) : -ENOSYS;
+		cq->device->modify_cq(cq, cq_count, cq_period) : -EOPNOTSUPP;
 }
 EXPORT_SYMBOL(rdma_set_cq_moderation);
 
@@ -1611,7 +1607,7 @@ EXPORT_SYMBOL(ib_destroy_cq);
 int ib_resize_cq(struct ib_cq *cq, int cqe)
 {
 	return cq->device->resize_cq ?
-		cq->device->resize_cq(cq, cqe, NULL) : -ENOSYS;
+		cq->device->resize_cq(cq, cqe, NULL) : -EOPNOTSUPP;
 }
 EXPORT_SYMBOL(ib_resize_cq);
 
@@ -1620,11 +1616,16 @@ EXPORT_SYMBOL(ib_resize_cq);
 int ib_dereg_mr(struct ib_mr *mr)
 {
 	struct ib_pd *pd = mr->pd;
+	struct ib_dm *dm = mr->dm;
 	int ret;
 
+	rdma_restrack_del(&mr->res);
 	ret = mr->device->dereg_mr(mr);
-	if (!ret)
+	if (!ret) {
 		atomic_dec(&pd->usecnt);
+		if (dm)
+			atomic_dec(&dm->usecnt);
+	}
 
 	return ret;
 }
@@ -1649,7 +1650,7 @@ struct ib_mr *ib_alloc_mr(struct ib_pd *pd,
 	struct ib_mr *mr;
 
 	if (!pd->device->alloc_mr)
-		return ERR_PTR(-ENOSYS);
+		return ERR_PTR(-EOPNOTSUPP);
 
 	mr = pd->device->alloc_mr(pd, mr_type, max_num_sg);
 	if (!IS_ERR(mr)) {
@@ -1658,6 +1659,8 @@ struct ib_mr *ib_alloc_mr(struct ib_pd *pd,
 		mr->uobject = NULL;
 		atomic_inc(&pd->usecnt);
 		mr->need_inval = false;
+		mr->res.type = RDMA_RESTRACK_MR;
+		rdma_restrack_add(&mr->res);
 	}
 
 	return mr;
@@ -1673,7 +1676,7 @@ struct ib_fmr *ib_alloc_fmr(struct ib_pd *pd,
 	struct ib_fmr *fmr;
 
 	if (!pd->device->alloc_fmr)
-		return ERR_PTR(-ENOSYS);
+		return ERR_PTR(-EOPNOTSUPP);
 
 	fmr = pd->device->alloc_fmr(pd, mr_access_flags, fmr_attr);
 	if (!IS_ERR(fmr)) {
@@ -1757,7 +1760,7 @@ int ib_attach_mcast(struct ib_qp *qp, union ib_gid *gid, u16 lid)
 	int ret;
 
 	if (!qp->device->attach_mcast)
-		return -ENOSYS;
+		return -EOPNOTSUPP;
 
 	if (!rdma_is_multicast_addr((struct in6_addr *)gid->raw) ||
 	    qp->qp_type != IB_QPT_UD || !is_valid_mcast_lid(qp, lid))
@@ -1775,7 +1778,7 @@ int ib_detach_mcast(struct ib_qp *qp, union ib_gid *gid, u16 lid)
 	int ret;
 
 	if (!qp->device->detach_mcast)
-		return -ENOSYS;
+		return -EOPNOTSUPP;
 
 	if (!rdma_is_multicast_addr((struct in6_addr *)gid->raw) ||
 	    qp->qp_type != IB_QPT_UD || !is_valid_mcast_lid(qp, lid))
@@ -1793,7 +1796,7 @@ struct ib_xrcd *__ib_alloc_xrcd(struct ib_device *device, const char *caller)
 	struct ib_xrcd *xrcd;
 
 	if (!device->alloc_xrcd)
-		return ERR_PTR(-ENOSYS);
+		return ERR_PTR(-EOPNOTSUPP);
 
 	xrcd = device->alloc_xrcd(device, NULL, NULL);
 	if (!IS_ERR(xrcd)) {
@@ -1847,7 +1850,7 @@ struct ib_wq *ib_create_wq(struct ib_pd *pd,
 	struct ib_wq *wq;
 
 	if (!pd->device->create_wq)
-		return ERR_PTR(-ENOSYS);
+		return ERR_PTR(-EOPNOTSUPP);
 
 	wq = pd->device->create_wq(pd, wq_attr, NULL);
 	if (!IS_ERR(wq)) {
@@ -1902,7 +1905,7 @@ int ib_modify_wq(struct ib_wq *wq, struct ib_wq_attr *wq_attr,
 	int err;
 
 	if (!wq->device->modify_wq)
-		return -ENOSYS;
+		return -EOPNOTSUPP;
 
 	err = wq->device->modify_wq(wq, wq_attr, wq_attr_mask, NULL);
 	return err;
@@ -1927,7 +1930,7 @@ struct ib_rwq_ind_table *ib_create_rwq_ind_table(struct ib_device *device,
 	u32 table_size;
 
 	if (!device->create_rwq_ind_table)
-		return ERR_PTR(-ENOSYS);
+		return ERR_PTR(-EOPNOTSUPP);
 
 	table_size = (1 << init_attr->log_ind_tbl_size);
 	rwq_ind_table = device->create_rwq_ind_table(device,
@@ -1977,7 +1980,7 @@ struct ib_flow *ib_create_flow(struct ib_qp *qp,
 {
 	struct ib_flow *flow_id;
 	if (!qp->device->create_flow)
-		return ERR_PTR(-ENOSYS);
+		return ERR_PTR(-EOPNOTSUPP);
 
 	flow_id = qp->device->create_flow(qp, flow_attr, domain);
 	if (!IS_ERR(flow_id)) {
@@ -2004,7 +2007,7 @@ int ib_check_mr_status(struct ib_mr *mr, u32 check_mask,
 		       struct ib_mr_status *mr_status)
 {
 	return mr->device->check_mr_status ?
-		mr->device->check_mr_status(mr, check_mask, mr_status) : -ENOSYS;
+		mr->device->check_mr_status(mr, check_mask, mr_status) : -EOPNOTSUPP;
 }
 EXPORT_SYMBOL(ib_check_mr_status);
 
@@ -2012,7 +2015,7 @@ int ib_set_vf_link_state(struct ib_device *device, int vf, u8 port,
 			 int state)
 {
 	if (!device->set_vf_link_state)
-		return -ENOSYS;
+		return -EOPNOTSUPP;
 
 	return device->set_vf_link_state(device, vf, port, state);
 }
@@ -2022,7 +2025,7 @@ int ib_get_vf_config(struct ib_device *device, int vf, u8 port,
 		     struct ifla_vf_info *info)
 {
 	if (!device->get_vf_config)
-		return -ENOSYS;
+		return -EOPNOTSUPP;
 
 	return device->get_vf_config(device, vf, port, info);
 }
@@ -2032,7 +2035,7 @@ int ib_get_vf_stats(struct ib_device *device, int vf, u8 port,
 		    struct ifla_vf_stats *stats)
 {
 	if (!device->get_vf_stats)
-		return -ENOSYS;
+		return -EOPNOTSUPP;
 
 	return device->get_vf_stats(device, vf, port, stats);
 }
@@ -2042,7 +2045,7 @@ int ib_set_vf_guid(struct ib_device *device, int vf, u8 port, u64 guid,
 		   int type)
 {
 	if (!device->set_vf_guid)
-		return -ENOSYS;
+		return -EOPNOTSUPP;
 
 	return device->set_vf_guid(device, vf, port, guid, type);
 }
@@ -2077,7 +2080,7 @@ int ib_map_mr_sg(struct ib_mr *mr, struct scatterlist *sg, int sg_nents,
 		 unsigned int *sg_offset, unsigned int page_size)
 {
 	if (unlikely(!mr->device->map_mr_sg))
-		return -ENOSYS;
+		return -EOPNOTSUPP;
 
 	mr->page_size = page_size;
 
@@ -2194,7 +2197,14 @@ static void __ib_drain_sq(struct ib_qp *qp)
 	struct ib_cq *cq = qp->send_cq;
 	struct ib_qp_attr attr = { .qp_state = IB_QPS_ERR };
 	struct ib_drain_cqe sdrain;
-	struct ib_send_wr swr = {}, *bad_swr;
+	struct ib_send_wr *bad_swr;
+	struct ib_rdma_wr swr = {
+		.wr = {
+			.next = NULL,
+			{ .wr_cqe	= &sdrain.cqe, },
+			.opcode	= IB_WR_RDMA_WRITE,
+		},
+	};
 	int ret;
 
 	ret = ib_modify_qp(qp, &attr, IB_QP_STATE);
@@ -2203,11 +2213,10 @@ static void __ib_drain_sq(struct ib_qp *qp)
 		return;
 	}
 
-	swr.wr_cqe = &sdrain.cqe;
 	sdrain.cqe.done = ib_drain_qp_done;
 	init_completion(&sdrain.done);
 
-	ret = ib_post_send(qp, &swr, &bad_swr);
+	ret = ib_post_send(qp, &swr.wr, &bad_swr);
 	if (ret) {
 		WARN_ONCE(ret, "failed to drain send queue: %d\n", ret);
 		return;

drivers/infiniband/hw/bnxt_re/ib_verbs.c

@@ -314,12 +314,11 @@ int bnxt_re_query_gid(struct ib_device *ibdev, u8 port_num,
 	return rc;
 }
 
-int bnxt_re_del_gid(struct ib_device *ibdev, u8 port_num,
-		    unsigned int index, void **context)
+int bnxt_re_del_gid(const struct ib_gid_attr *attr, void **context)
 {
 	int rc = 0;
 	struct bnxt_re_gid_ctx *ctx, **ctx_tbl;
-	struct bnxt_re_dev *rdev = to_bnxt_re_dev(ibdev, ibdev);
+	struct bnxt_re_dev *rdev = to_bnxt_re_dev(attr->device, ibdev);
 	struct bnxt_qplib_sgid_tbl *sgid_tbl = &rdev->qplib_res.sgid_tbl;
 	struct bnxt_qplib_gid *gid_to_del;
 
@@ -365,15 +364,14 @@ int bnxt_re_del_gid(struct ib_device *ibdev, u8 port_num,
 	return rc;
 }
 
-int bnxt_re_add_gid(struct ib_device *ibdev, u8 port_num,
-		    unsigned int index, const union ib_gid *gid,
+int bnxt_re_add_gid(const union ib_gid *gid,
 		    const struct ib_gid_attr *attr, void **context)
 {
 	int rc;
 	u32 tbl_idx = 0;
 	u16 vlan_id = 0xFFFF;
 	struct bnxt_re_gid_ctx *ctx, **ctx_tbl;
-	struct bnxt_re_dev *rdev = to_bnxt_re_dev(ibdev, ibdev);
+	struct bnxt_re_dev *rdev = to_bnxt_re_dev(attr->device, ibdev);
 	struct bnxt_qplib_sgid_tbl *sgid_tbl = &rdev->qplib_res.sgid_tbl;
 
 	if ((attr->ndev) && is_vlan_dev(attr->ndev))
@@ -718,8 +716,7 @@ struct ib_ah *bnxt_re_create_ah(struct ib_pd *ib_pd,
 				grh->sgid_index);
 			goto fail;
 		}
-		if (sgid_attr.ndev)
-			dev_put(sgid_attr.ndev);
+		dev_put(sgid_attr.ndev);
 		/* Get network header type for this GID */
 		nw_type = ib_gid_to_network_type(sgid_attr.gid_type, &sgid);
 		switch (nw_type) {
@@ -1540,14 +1537,13 @@ int bnxt_re_post_srq_recv(struct ib_srq *ib_srq, struct ib_recv_wr *wr,
 					       ib_srq);
 	struct bnxt_qplib_swqe wqe;
 	unsigned long flags;
-	int rc = 0, payload_sz = 0;
+	int rc = 0;
 
 	spin_lock_irqsave(&srq->lock, flags);
 	while (wr) {
 		/* Transcribe each ib_recv_wr to qplib_swqe */
 		wqe.num_sge = wr->num_sge;
-		payload_sz = bnxt_re_build_sgl(wr->sg_list, wqe.sg_list,
-					       wr->num_sge);
+		bnxt_re_build_sgl(wr->sg_list, wqe.sg_list, wr->num_sge);
 		wqe.wr_id = wr->wr_id;
 		wqe.type = BNXT_QPLIB_SWQE_TYPE_RECV;
 
@@ -1698,7 +1694,7 @@ int bnxt_re_modify_qp(struct ib_qp *ib_qp, struct ib_qp_attr *qp_attr,
 		status = ib_get_cached_gid(&rdev->ibdev, 1,
 					   grh->sgid_index,
 					   &sgid, &sgid_attr);
-		if (!status && sgid_attr.ndev) {
+		if (!status) {
 			memcpy(qp->qplib_qp.smac, sgid_attr.ndev->dev_addr,
 			       ETH_ALEN);
 			dev_put(sgid_attr.ndev);

drivers/infiniband/hw/bnxt_re/ib_verbs.h

@@ -157,10 +157,8 @@ int bnxt_re_get_port_immutable(struct ib_device *ibdev, u8 port_num,
 void bnxt_re_query_fw_str(struct ib_device *ibdev, char *str);
 int bnxt_re_query_pkey(struct ib_device *ibdev, u8 port_num,
 		       u16 index, u16 *pkey);
-int bnxt_re_del_gid(struct ib_device *ibdev, u8 port_num,
-		    unsigned int index, void **context);
-int bnxt_re_add_gid(struct ib_device *ibdev, u8 port_num,
-		    unsigned int index, const union ib_gid *gid,
+int bnxt_re_del_gid(const struct ib_gid_attr *attr, void **context);
+int bnxt_re_add_gid(const union ib_gid *gid,
 		    const struct ib_gid_attr *attr, void **context);
 int bnxt_re_query_gid(struct ib_device *ibdev, u8 port_num,
 		      int index, union ib_gid *gid);

drivers/infiniband/hw/bnxt_re/main.c

@@ -574,7 +574,6 @@ static int bnxt_re_register_ib(struct bnxt_re_dev *rdev)
 	ibdev->get_port_immutable	= bnxt_re_get_port_immutable;
 	ibdev->get_dev_fw_str           = bnxt_re_query_fw_str;
 	ibdev->query_pkey		= bnxt_re_query_pkey;
-	ibdev->query_gid		= bnxt_re_query_gid;
 	ibdev->get_netdev		= bnxt_re_get_netdev;
 	ibdev->add_gid			= bnxt_re_add_gid;
 	ibdev->del_gid			= bnxt_re_del_gid;
@@ -619,6 +618,7 @@ static int bnxt_re_register_ib(struct bnxt_re_dev *rdev)
 	ibdev->get_hw_stats             = bnxt_re_ib_get_hw_stats;
 	ibdev->alloc_hw_stats           = bnxt_re_ib_alloc_hw_stats;
 
+	ibdev->driver_id = RDMA_DRIVER_BNXT_RE;
 	return ib_register_device(ibdev, NULL);
 }
 

drivers/infiniband/hw/bnxt_re/qplib_sp.c

@@ -154,7 +154,7 @@ int bnxt_qplib_get_dev_attr(struct bnxt_qplib_rcfw *rcfw,
 		attr->tqm_alloc_reqs[i * 4 + 3] = *(++tqm_alloc);
 	}
 
-	attr->is_atomic = 0;
+	attr->is_atomic = false;
 bail:
 	bnxt_qplib_rcfw_free_sbuf(rcfw, sbuf);
 	return rc;

drivers/infiniband/hw/cxgb3/Kconfig

@@ -16,12 +16,3 @@ config INFINIBAND_CXGB3
 
 	  To compile this driver as a module, choose M here: the module
 	  will be called iw_cxgb3.
-
-config INFINIBAND_CXGB3_DEBUG
-	bool "Verbose debugging output"
-	depends on INFINIBAND_CXGB3
-	default n
-	---help---
-	  This option causes the Chelsio RDMA driver to produce copious
-	  amounts of debug messages.  Select this if you are developing
-	  the driver or trying to diagnose a problem.

drivers/infiniband/hw/cxgb3/Makefile

@@ -5,5 +5,3 @@ obj-$(CONFIG_INFINIBAND_CXGB3) += iw_cxgb3.o
 
 iw_cxgb3-y :=  iwch_cm.o iwch_ev.o iwch_cq.o iwch_qp.o iwch_mem.o \
 	       iwch_provider.o iwch.o cxio_hal.o cxio_resource.o
-
-ccflags-$(CONFIG_INFINIBAND_CXGB3_DEBUG) += -DDEBUG

drivers/infiniband/hw/cxgb3/cxio_dbg.c

@@ -1,206 +0,0 @@
-/*
- * Copyright (c) 2006 Chelsio, Inc. All rights reserved.
- *
- * This software is available to you under a choice of one of two
- * licenses.  You may choose to be licensed under the terms of the GNU
- * General Public License (GPL) Version 2, available from the file
- * COPYING in the main directory of this source tree, or the
- * OpenIB.org BSD license below:
- *
- *     Redistribution and use in source and binary forms, with or
- *     without modification, are permitted provided that the following
- *     conditions are met:
- *
- *      - Redistributions of source code must retain the above
- *        copyright notice, this list of conditions and the following
- *        disclaimer.
- *
- *      - Redistributions in binary form must reproduce the above
- *        copyright notice, this list of conditions and the following
- *        disclaimer in the documentation and/or other materials
- *        provided with the distribution.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
- * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
- * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
- * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
- * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
- * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#ifdef DEBUG
-#include <linux/types.h>
-#include <linux/slab.h>
-#include "common.h"
-#include "cxgb3_ioctl.h"
-#include "cxio_hal.h"
-#include "cxio_wr.h"
-
-void cxio_dump_tpt(struct cxio_rdev *rdev, u32 stag)
-{
-	struct ch_mem_range *m;
-	u64 *data;
-	int rc;
-	int size = 32;
-
-	m = kmalloc(sizeof(*m) + size, GFP_ATOMIC);
-	if (!m)
-		return;
-
-	m->mem_id = MEM_PMRX;
-	m->addr = (stag>>8) * 32 + rdev->rnic_info.tpt_base;
-	m->len = size;
-	pr_debug("%s TPT addr 0x%x len %d\n", __func__, m->addr, m->len);
-	rc = rdev->t3cdev_p->ctl(rdev->t3cdev_p, RDMA_GET_MEM, m);
-	if (rc) {
-		pr_debug("%s toectl returned error %d\n", __func__, rc);
-		kfree(m);
-		return;
-	}
-
-	data = (u64 *)m->buf;
-	while (size > 0) {
-		pr_debug("TPT %08x: %016llx\n",
-			 m->addr, (unsigned long long)*data);
-		size -= 8;
-		data++;
-		m->addr += 8;
-	}
-	kfree(m);
-}
-
-void cxio_dump_pbl(struct cxio_rdev *rdev, u32 pbl_addr, uint len, u8 shift)
-{
-	struct ch_mem_range *m;
-	u64 *data;
-	int rc;
-	int size, npages;
-
-	shift += 12;
-	npages = (len + (1ULL << shift) - 1) >> shift;
-	size = npages * sizeof(u64);
-
-	m = kmalloc(sizeof(*m) + size, GFP_ATOMIC);
-	if (!m)
-		return;
-
-	m->mem_id = MEM_PMRX;
-	m->addr = pbl_addr;
-	m->len = size;
-	pr_debug("%s PBL addr 0x%x len %d depth %d\n",
-		 __func__, m->addr, m->len, npages);
-	rc = rdev->t3cdev_p->ctl(rdev->t3cdev_p, RDMA_GET_MEM, m);
-	if (rc) {
-		pr_debug("%s toectl returned error %d\n", __func__, rc);
-		kfree(m);
-		return;
-	}
-
-	data = (u64 *)m->buf;
-	while (size > 0) {
-		pr_debug("PBL %08x: %016llx\n",
-			 m->addr, (unsigned long long)*data);
-		size -= 8;
-		data++;
-		m->addr += 8;
-	}
-	kfree(m);
-}
-
-void cxio_dump_wqe(union t3_wr *wqe)
-{
-	__be64 *data = (__be64 *)wqe;
-	uint size = (uint)(be64_to_cpu(*data) & 0xff);
-
-	if (size == 0)
-		size = 8;
-	while (size > 0) {
-		pr_debug("WQE %p: %016llx\n",
-			 data, (unsigned long long)be64_to_cpu(*data));
-		size--;
-		data++;
-	}
-}
-
-void cxio_dump_wce(struct t3_cqe *wce)
-{
-	__be64 *data = (__be64 *)wce;
-	int size = sizeof(*wce);
-
-	while (size > 0) {
-		pr_debug("WCE %p: %016llx\n",
-			 data, (unsigned long long)be64_to_cpu(*data));
-		size -= 8;
-		data++;
-	}
-}
-
-void cxio_dump_rqt(struct cxio_rdev *rdev, u32 hwtid, int nents)
-{
-	struct ch_mem_range *m;
-	int size = nents * 64;
-	u64 *data;
-	int rc;
-
-	m = kmalloc(sizeof(*m) + size, GFP_ATOMIC);
-	if (!m)
-		return;
-
-	m->mem_id = MEM_PMRX;
-	m->addr = ((hwtid)<<10) + rdev->rnic_info.rqt_base;
-	m->len = size;
-	pr_debug("%s RQT addr 0x%x len %d\n", __func__, m->addr, m->len);
-	rc = rdev->t3cdev_p->ctl(rdev->t3cdev_p, RDMA_GET_MEM, m);
-	if (rc) {
-		pr_debug("%s toectl returned error %d\n", __func__, rc);
-		kfree(m);
-		return;
-	}
-
-	data = (u64 *)m->buf;
-	while (size > 0) {
-		pr_debug("RQT %08x: %016llx\n",
-			 m->addr, (unsigned long long)*data);
-		size -= 8;
-		data++;
-		m->addr += 8;
-	}
-	kfree(m);
-}
-
-void cxio_dump_tcb(struct cxio_rdev *rdev, u32 hwtid)
-{
-	struct ch_mem_range *m;
-	int size = TCB_SIZE;
-	u32 *data;
-	int rc;
-
-	m = kmalloc(sizeof(*m) + size, GFP_ATOMIC);
-	if (!m)
-		return;
-
-	m->mem_id = MEM_CM;
-	m->addr = hwtid * size;
-	m->len = size;
-	pr_debug("%s TCB %d len %d\n", __func__, m->addr, m->len);
-	rc = rdev->t3cdev_p->ctl(rdev->t3cdev_p, RDMA_GET_MEM, m);
-	if (rc) {
-		pr_debug("%s toectl returned error %d\n", __func__, rc);
-		kfree(m);
-		return;
-	}
-
-	data = (u32 *)m->buf;
-	while (size > 0) {
-		printk("%2u: %08x %08x %08x %08x %08x %08x %08x %08x\n",
-			m->addr,
-			*(data+2), *(data+3), *(data),*(data+1),
-			*(data+6), *(data+7), *(data+4), *(data+5));
-		size -= 32;
-		data += 8;
-		m->addr += 32;
-	}
-	kfree(m);
-}
-#endif

drivers/infiniband/hw/cxgb3/cxio_hal.h

@@ -202,13 +202,4 @@ int iwch_cxgb3_ofld_send(struct t3cdev *tdev, struct sk_buff *skb);
 
 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
 
-#ifdef DEBUG
-void cxio_dump_tpt(struct cxio_rdev *rev, u32 stag);
-void cxio_dump_pbl(struct cxio_rdev *rev, u32 pbl_addr, uint len, u8 shift);
-void cxio_dump_wqe(union t3_wr *wqe);
-void cxio_dump_wce(struct t3_cqe *wce);
-void cxio_dump_rqt(struct cxio_rdev *rdev, u32 hwtid, int nents);
-void cxio_dump_tcb(struct cxio_rdev *rdev, u32 hwtid);
-#endif
-
 #endif

drivers/infiniband/hw/cxgb3/iwch_cq.c

@@ -200,9 +200,6 @@ int iwch_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *wc)
 
 	spin_lock_irqsave(&chp->lock, flags);
 	for (npolled = 0; npolled < num_entries; ++npolled) {
-#ifdef DEBUG
-		int i=0;
-#endif
 
 		/*
 		 * Because T3 can post CQEs that are _not_ associated
@@ -211,9 +208,6 @@ int iwch_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *wc)
 		 */
 		do {
 			err = iwch_poll_cq_one(rhp, chp, wc + npolled);
-#ifdef DEBUG
-			BUG_ON(++i > 1000);
-#endif
 		} while (err == -EAGAIN);
 		if (err <= 0)
 			break;

drivers/infiniband/hw/cxgb3/iwch_provider.c

@@ -440,7 +440,9 @@ static struct ib_pd *iwch_allocate_pd(struct ib_device *ibdev,
 	php->pdid = pdid;
 	php->rhp = rhp;
 	if (context) {
-		if (ib_copy_to_udata(udata, &php->pdid, sizeof (__u32))) {
+		struct iwch_alloc_pd_resp resp = {.pdid = php->pdid};
+
+		if (ib_copy_to_udata(udata, &resp, sizeof(resp))) {
 			iwch_deallocate_pd(&php->ibpd);
 			return ERR_PTR(-EFAULT);
 		}
@@ -1439,6 +1441,7 @@ int iwch_register_device(struct iwch_dev *dev)
 	memcpy(dev->ibdev.iwcm->ifname, dev->rdev.t3cdev_p->lldev->name,
 	       sizeof(dev->ibdev.iwcm->ifname));
 
+	dev->ibdev.driver_id = RDMA_DRIVER_CXGB3;
 	ret = ib_register_device(&dev->ibdev, NULL);
 	if (ret)
 		goto bail1;

drivers/infiniband/hw/cxgb4/device.c

@@ -220,14 +220,14 @@ static void set_ep_sin_addrs(struct c4iw_ep *ep,
 {
 	struct iw_cm_id *id = ep->com.cm_id;
 
-	*lsin = (struct sockaddr_in *)&ep->com.local_addr;
-	*rsin = (struct sockaddr_in *)&ep->com.remote_addr;
+	*m_lsin = (struct sockaddr_in *)&ep->com.local_addr;
+	*m_rsin = (struct sockaddr_in *)&ep->com.remote_addr;
 	if (id) {
-		*m_lsin = (struct sockaddr_in *)&id->m_local_addr;
-		*m_rsin = (struct sockaddr_in *)&id->m_remote_addr;
+		*lsin = (struct sockaddr_in *)&id->local_addr;
+		*rsin = (struct sockaddr_in *)&id->remote_addr;
 	} else {
-		*m_lsin = &zero_sin;
-		*m_rsin = &zero_sin;
+		*lsin = &zero_sin;
+		*rsin = &zero_sin;
 	}
 }
 
@@ -239,14 +239,14 @@ static void set_ep_sin6_addrs(struct c4iw_ep *ep,
 {
 	struct iw_cm_id *id = ep->com.cm_id;
 
-	*lsin6 = (struct sockaddr_in6 *)&ep->com.local_addr;
-	*rsin6 = (struct sockaddr_in6 *)&ep->com.remote_addr;
+	*m_lsin6 = (struct sockaddr_in6 *)&ep->com.local_addr;
+	*m_rsin6 = (struct sockaddr_in6 *)&ep->com.remote_addr;
 	if (id) {
-		*m_lsin6 = (struct sockaddr_in6 *)&id->m_local_addr;
-		*m_rsin6 = (struct sockaddr_in6 *)&id->m_remote_addr;
+		*lsin6 = (struct sockaddr_in6 *)&id->local_addr;
+		*rsin6 = (struct sockaddr_in6 *)&id->remote_addr;
 	} else {
-		*m_lsin6 = &zero_sin6;
-		*m_rsin6 = &zero_sin6;
+		*lsin6 = &zero_sin6;
+		*rsin6 = &zero_sin6;
 	}
 }
 

drivers/infiniband/hw/cxgb4/mem.c

@@ -391,6 +391,9 @@ static int finish_mem_reg(struct c4iw_mr *mhp, u32 stag)
 	mhp->attr.stag = stag;
 	mmid = stag >> 8;
 	mhp->ibmr.rkey = mhp->ibmr.lkey = stag;
+	mhp->ibmr.length = mhp->attr.len;
+	mhp->ibmr.iova = mhp->attr.va_fbo;
+	mhp->ibmr.page_size = 1U << (mhp->attr.page_size + 12);
 	pr_debug("mmid 0x%x mhp %p\n", mmid, mhp);
 	return insert_handle(mhp->rhp, &mhp->rhp->mmidr, mhp, mmid);
 }

drivers/infiniband/hw/cxgb4/provider.c

@@ -281,7 +281,9 @@ static struct ib_pd *c4iw_allocate_pd(struct ib_device *ibdev,
 	php->pdid = pdid;
 	php->rhp = rhp;
 	if (context) {
-		if (ib_copy_to_udata(udata, &php->pdid, sizeof(u32))) {
+		struct c4iw_alloc_pd_resp uresp = {.pdid = php->pdid};
+
+		if (ib_copy_to_udata(udata, &uresp, sizeof(uresp))) {
 			c4iw_deallocate_pd(&php->ibpd);
 			return ERR_PTR(-EFAULT);
 		}
@@ -531,6 +533,24 @@ static void get_dev_fw_str(struct ib_device *dev, char *str)
 		 FW_HDR_FW_VER_BUILD_G(c4iw_dev->rdev.lldi.fw_vers));
 }
 
+static struct net_device *get_netdev(struct ib_device *dev, u8 port)
+{
+	struct c4iw_dev *c4iw_dev = container_of(dev, struct c4iw_dev, ibdev);
+	struct c4iw_rdev *rdev = &c4iw_dev->rdev;
+	struct net_device *ndev;
+
+	if (!port || port > rdev->lldi.nports)
+		return NULL;
+
+	rcu_read_lock();
+	ndev = rdev->lldi.ports[port - 1];
+	if (ndev)
+		dev_hold(ndev);
+	rcu_read_unlock();
+
+	return ndev;
+}
+
 void c4iw_register_device(struct work_struct *work)
 {
 	int ret;
@@ -609,6 +629,7 @@ void c4iw_register_device(struct work_struct *work)
 	dev->ibdev.uverbs_abi_ver = C4IW_UVERBS_ABI_VERSION;
 	dev->ibdev.get_port_immutable = c4iw_port_immutable;
 	dev->ibdev.get_dev_fw_str = get_dev_fw_str;
+	dev->ibdev.get_netdev = get_netdev;
 
 	dev->ibdev.iwcm = kmalloc(sizeof(struct iw_cm_verbs), GFP_KERNEL);
 	if (!dev->ibdev.iwcm) {
@@ -627,6 +648,7 @@ void c4iw_register_device(struct work_struct *work)
 	memcpy(dev->ibdev.iwcm->ifname, dev->rdev.lldi.ports[0]->name,
 	       sizeof(dev->ibdev.iwcm->ifname));
 
+	dev->ibdev.driver_id = RDMA_DRIVER_CXGB4;
 	ret = ib_register_device(&dev->ibdev, NULL);
 	if (ret)
 		goto err_kfree_iwcm;

drivers/infiniband/hw/hfi1/driver.c

@@ -390,6 +390,7 @@ static void rcv_hdrerr(struct hfi1_ctxtdata *rcd, struct hfi1_pportdata *ppd,
 				svc_type = IB_CC_SVCTYPE_UC;
 				break;
 			default:
+				rcu_read_unlock();
 				goto drop;
 			}
 

drivers/infiniband/hw/hfi1/hfi.h

@@ -70,7 +70,6 @@
 #include <linux/rhashtable.h>
 #include <linux/netdevice.h>
 #include <rdma/rdma_vt.h>
-#include <rdma/opa_addr.h>
 
 #include "chip_registers.h"
 #include "common.h"

drivers/infiniband/hw/hfi1/qp.c

@@ -481,7 +481,6 @@ static void iowait_sdma_drained(struct iowait *wait)
 }
 
 /**
- *
  * qp_to_sdma_engine - map a qp to a send engine
  * @qp: the QP
  * @sc5: the 5 bit sc

drivers/infiniband/hw/hfi1/user_exp_rcv.c

@@ -473,7 +473,7 @@ nomem:
 		tinfo->tidcnt = tididx;
 		tinfo->length = mapped_pages * PAGE_SIZE;
 
-		if (copy_to_user((void __user *)(unsigned long)tinfo->tidlist,
+		if (copy_to_user(u64_to_user_ptr(tinfo->tidlist),
 				 tidlist, sizeof(tidlist[0]) * tididx)) {
 			/*
 			 * On failure to copy to the user level, we need to undo
@@ -513,7 +513,7 @@ int hfi1_user_exp_rcv_clear(struct hfi1_filedata *fd,
 	if (unlikely(tinfo->tidcnt > fd->tid_used))
 		return -EINVAL;
 
-	tidinfo = memdup_user((void __user *)(unsigned long)tinfo->tidlist,
+	tidinfo = memdup_user(u64_to_user_ptr(tinfo->tidlist),
 			      sizeof(tidinfo[0]) * tinfo->tidcnt);
 	if (IS_ERR(tidinfo))
 		return PTR_ERR(tidinfo);

drivers/infiniband/hw/hfi1/verbs.c

@@ -1960,7 +1960,7 @@ int hfi1_register_ib_device(struct hfi1_devdata *dd)
 			      i,
 			      ppd->pkeys);
 
-	ret = rvt_register_device(&dd->verbs_dev.rdi);
+	ret = rvt_register_device(&dd->verbs_dev.rdi, RDMA_DRIVER_HFI1);
 	if (ret)
 		goto err_verbs_txreq;
 

drivers/infiniband/hw/hns/Makefile

@@ -7,7 +7,7 @@ ccflags-y :=  -Idrivers/net/ethernet/hisilicon/hns3
 obj-$(CONFIG_INFINIBAND_HNS) += hns-roce.o
 hns-roce-objs := hns_roce_main.o hns_roce_cmd.o hns_roce_pd.o \
 	hns_roce_ah.o hns_roce_hem.o hns_roce_mr.o hns_roce_qp.o \
-	hns_roce_cq.o hns_roce_alloc.o
+	hns_roce_cq.o hns_roce_alloc.o hns_roce_db.o
 obj-$(CONFIG_INFINIBAND_HNS_HIP06) += hns-roce-hw-v1.o
 hns-roce-hw-v1-objs := hns_roce_hw_v1.o
 obj-$(CONFIG_INFINIBAND_HNS_HIP08) += hns-roce-hw-v2.o

drivers/infiniband/hw/hns/hns_roce_ah.c

@@ -68,11 +68,9 @@ struct ib_ah *hns_roce_create_ah(struct ib_pd *ibpd,
 		return ERR_PTR(ret);
 	}
 
-	if (gid_attr.ndev) {
-		if (is_vlan_dev(gid_attr.ndev))
-			vlan_tag = vlan_dev_vlan_id(gid_attr.ndev);
-		dev_put(gid_attr.ndev);
-	}
+	if (is_vlan_dev(gid_attr.ndev))
+		vlan_tag = vlan_dev_vlan_id(gid_attr.ndev);
+	dev_put(gid_attr.ndev);
 
 	if (vlan_tag < 0x1000)
 		vlan_tag |= (rdma_ah_get_sl(ah_attr) &

drivers/infiniband/hw/hns/hns_roce_cq.c

@@ -315,6 +315,7 @@ struct ib_cq *hns_roce_ib_create_cq(struct ib_device *ib_dev,
 	struct hns_roce_dev *hr_dev = to_hr_dev(ib_dev);
 	struct device *dev = hr_dev->dev;
 	struct hns_roce_ib_create_cq ucmd;
+	struct hns_roce_ib_create_cq_resp resp = {};
 	struct hns_roce_cq *hr_cq = NULL;
 	struct hns_roce_uar *uar = NULL;
 	int vector = attr->comp_vector;
@@ -354,15 +355,36 @@ struct ib_cq *hns_roce_ib_create_cq(struct ib_device *ib_dev,
 			goto err_cq;
 		}
 
+		if ((hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_RECORD_DB) &&
+		    (udata->outlen >= sizeof(resp))) {
+			ret = hns_roce_db_map_user(to_hr_ucontext(context),
+						   ucmd.db_addr, &hr_cq->db);
+			if (ret) {
+				dev_err(dev, "cq record doorbell map failed!\n");
+				goto err_mtt;
+			}
+			hr_cq->db_en = 1;
+			resp.cap_flags |= HNS_ROCE_SUPPORT_CQ_RECORD_DB;
+		}
+
 		/* Get user space parameters */
 		uar = &to_hr_ucontext(context)->uar;
 	} else {
+		if (hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_RECORD_DB) {
+			ret = hns_roce_alloc_db(hr_dev, &hr_cq->db, 1);
+			if (ret)
+				goto err_cq;
+
+			hr_cq->set_ci_db = hr_cq->db.db_record;
+			*hr_cq->set_ci_db = 0;
+		}
+
 		/* Init mmt table and write buff address to mtt table */
 		ret = hns_roce_ib_alloc_cq_buf(hr_dev, &hr_cq->hr_buf,
 					       cq_entries);
 		if (ret) {
 			dev_err(dev, "Failed to alloc_cq_buf.\n");
-			goto err_cq;
+			goto err_db;
 		}
 
 		uar = &hr_dev->priv_uar;
@@ -375,7 +397,7 @@ struct ib_cq *hns_roce_ib_create_cq(struct ib_device *ib_dev,
 				hr_cq, vector);
 	if (ret) {
 		dev_err(dev, "Creat CQ .Failed to cq_alloc.\n");
-		goto err_mtt;
+		goto err_dbmap;
 	}
 
 	/*
@@ -393,10 +415,10 @@ struct ib_cq *hns_roce_ib_create_cq(struct ib_device *ib_dev,
 	hr_cq->cq_depth = cq_entries;
 
 	if (context) {
-		if (ib_copy_to_udata(udata, &hr_cq->cqn, sizeof(u64))) {
-			ret = -EFAULT;
+		resp.cqn = hr_cq->cqn;
+		ret = ib_copy_to_udata(udata, &resp, sizeof(resp));
+		if (ret)
 			goto err_cqc;
-		}
 	}
 
 	return &hr_cq->ib_cq;
@@ -404,6 +426,12 @@ struct ib_cq *hns_roce_ib_create_cq(struct ib_device *ib_dev,
 err_cqc:
 	hns_roce_free_cq(hr_dev, hr_cq);
 
+err_dbmap:
+	if (context && (hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_RECORD_DB) &&
+	    (udata->outlen >= sizeof(resp)))
+		hns_roce_db_unmap_user(to_hr_ucontext(context),
+				       &hr_cq->db);
+
 err_mtt:
 	hns_roce_mtt_cleanup(hr_dev, &hr_cq->hr_buf.hr_mtt);
 	if (context)
@@ -412,6 +440,10 @@ err_mtt:
 		hns_roce_ib_free_cq_buf(hr_dev, &hr_cq->hr_buf,
 					hr_cq->ib_cq.cqe);
 
+err_db:
+	if (!context && (hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_RECORD_DB))
+		hns_roce_free_db(hr_dev, &hr_cq->db);
+
 err_cq:
 	kfree(hr_cq);
 	return ERR_PTR(ret);
@@ -430,12 +462,20 @@ int hns_roce_ib_destroy_cq(struct ib_cq *ib_cq)
 		hns_roce_free_cq(hr_dev, hr_cq);
 		hns_roce_mtt_cleanup(hr_dev, &hr_cq->hr_buf.hr_mtt);
 
-		if (ib_cq->uobject)
+		if (ib_cq->uobject) {
 			ib_umem_release(hr_cq->umem);
-		else
+
+			if (hr_cq->db_en == 1)
+				hns_roce_db_unmap_user(
+					to_hr_ucontext(ib_cq->uobject->context),
+					&hr_cq->db);
+		} else {
 			/* Free the buff of stored cq */
 			hns_roce_ib_free_cq_buf(hr_dev, &hr_cq->hr_buf,
 						ib_cq->cqe);
+			if (hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_RECORD_DB)
+				hns_roce_free_db(hr_dev, &hr_cq->db);
+		}
 
 		kfree(hr_cq);
 	}

drivers/infiniband/hw/hns/hns_roce_db.c

@@ -0,0 +1,180 @@
+/* SPDX-License-Identifier: (GPL-2.0 OR BSD-2-Clause) */
+/*
+ * Copyright (c) 2017 Hisilicon Limited.
+ * Copyright (c) 2007, 2008 Mellanox Technologies. All rights reserved.
+ */
+
+#include <linux/platform_device.h>
+#include <rdma/ib_umem.h>
+#include "hns_roce_device.h"
+
+int hns_roce_db_map_user(struct hns_roce_ucontext *context, unsigned long virt,
+			 struct hns_roce_db *db)
+{
+	struct hns_roce_user_db_page *page;
+	int ret = 0;
+
+	mutex_lock(&context->page_mutex);
+
+	list_for_each_entry(page, &context->page_list, list)
+		if (page->user_virt == (virt & PAGE_MASK))
+			goto found;
+
+	page = kmalloc(sizeof(*page), GFP_KERNEL);
+	if (!page) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	refcount_set(&page->refcount, 1);
+	page->user_virt = (virt & PAGE_MASK);
+	page->umem = ib_umem_get(&context->ibucontext, virt & PAGE_MASK,
+				 PAGE_SIZE, 0, 0);
+	if (IS_ERR(page->umem)) {
+		ret = PTR_ERR(page->umem);
+		kfree(page);
+		goto out;
+	}
+
+	list_add(&page->list, &context->page_list);
+
+found:
+	db->dma = sg_dma_address(page->umem->sg_head.sgl) +
+		  (virt & ~PAGE_MASK);
+	db->u.user_page = page;
+	refcount_inc(&page->refcount);
+
+out:
+	mutex_unlock(&context->page_mutex);
+
+	return ret;
+}
+EXPORT_SYMBOL(hns_roce_db_map_user);
+
+void hns_roce_db_unmap_user(struct hns_roce_ucontext *context,
+			    struct hns_roce_db *db)
+{
+	mutex_lock(&context->page_mutex);
+
+	refcount_dec(&db->u.user_page->refcount);
+	if (refcount_dec_if_one(&db->u.user_page->refcount)) {
+		list_del(&db->u.user_page->list);
+		ib_umem_release(db->u.user_page->umem);
+		kfree(db->u.user_page);
+	}
+
+	mutex_unlock(&context->page_mutex);
+}
+EXPORT_SYMBOL(hns_roce_db_unmap_user);
+
+static struct hns_roce_db_pgdir *hns_roce_alloc_db_pgdir(
+					struct device *dma_device)
+{
+	struct hns_roce_db_pgdir *pgdir;
+
+	pgdir = kzalloc(sizeof(*pgdir), GFP_KERNEL);
+	if (!pgdir)
+		return NULL;
+
+	bitmap_fill(pgdir->order1, HNS_ROCE_DB_PER_PAGE / 2);
+	pgdir->bits[0] = pgdir->order0;
+	pgdir->bits[1] = pgdir->order1;
+	pgdir->page = dma_alloc_coherent(dma_device, PAGE_SIZE,
+					 &pgdir->db_dma, GFP_KERNEL);
+	if (!pgdir->page) {
+		kfree(pgdir);
+		return NULL;
+	}
+
+	return pgdir;
+}
+
+static int hns_roce_alloc_db_from_pgdir(struct hns_roce_db_pgdir *pgdir,
+					struct hns_roce_db *db, int order)
+{
+	int o;
+	int i;
+
+	for (o = order; o <= 1; ++o) {
+		i = find_first_bit(pgdir->bits[o], HNS_ROCE_DB_PER_PAGE >> o);
+		if (i < HNS_ROCE_DB_PER_PAGE >> o)
+			goto found;
+	}
+
+	return -ENOMEM;
+
+found:
+	clear_bit(i, pgdir->bits[o]);
+
+	i <<= o;
+
+	if (o > order)
+		set_bit(i ^ 1, pgdir->bits[order]);
+
+	db->u.pgdir	= pgdir;
+	db->index	= i;
+	db->db_record	= pgdir->page + db->index;
+	db->dma		= pgdir->db_dma  + db->index * 4;
+	db->order	= order;
+
+	return 0;
+}
+
+int hns_roce_alloc_db(struct hns_roce_dev *hr_dev, struct hns_roce_db *db,
+		      int order)
+{
+	struct hns_roce_db_pgdir *pgdir;
+	int ret = 0;
+
+	mutex_lock(&hr_dev->pgdir_mutex);
+
+	list_for_each_entry(pgdir, &hr_dev->pgdir_list, list)
+		if (!hns_roce_alloc_db_from_pgdir(pgdir, db, order))
+			goto out;
+
+	pgdir = hns_roce_alloc_db_pgdir(hr_dev->dev);
+	if (!pgdir) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	list_add(&pgdir->list, &hr_dev->pgdir_list);
+
+	/* This should never fail -- we just allocated an empty page: */
+	WARN_ON(hns_roce_alloc_db_from_pgdir(pgdir, db, order));
+
+out:
+	mutex_unlock(&hr_dev->pgdir_mutex);
+
+	return ret;
+}
+EXPORT_SYMBOL_GPL(hns_roce_alloc_db);
+
+void hns_roce_free_db(struct hns_roce_dev *hr_dev, struct hns_roce_db *db)
+{
+	int o;
+	int i;
+
+	mutex_lock(&hr_dev->pgdir_mutex);
+
+	o = db->order;
+	i = db->index;
+
+	if (db->order == 0 && test_bit(i ^ 1, db->u.pgdir->order0)) {
+		clear_bit(i ^ 1, db->u.pgdir->order0);
+		++o;
+	}
+
+	i >>= o;
+	set_bit(i, db->u.pgdir->bits[o]);
+
+	if (bitmap_full(db->u.pgdir->order1, HNS_ROCE_DB_PER_PAGE / 2)) {
+		dma_free_coherent(hr_dev->dev, PAGE_SIZE, db->u.pgdir->page,
+				  db->u.pgdir->db_dma);
+		list_del(&db->u.pgdir->list);
+		kfree(db->u.pgdir);
+	}
+
+	mutex_unlock(&hr_dev->pgdir_mutex);
+}
+EXPORT_SYMBOL_GPL(hns_roce_free_db);

drivers/infiniband/hw/hns/hns_roce_device.h

@@ -105,6 +105,14 @@
 #define PAGES_SHIFT_24				24
 #define PAGES_SHIFT_32				32
 
+enum {
+	HNS_ROCE_SUPPORT_RQ_RECORD_DB = 1 << 0,
+};
+
+enum {
+	HNS_ROCE_SUPPORT_CQ_RECORD_DB = 1 << 0,
+};
+
 enum hns_roce_qp_state {
 	HNS_ROCE_QP_STATE_RST,
 	HNS_ROCE_QP_STATE_INIT,
@@ -178,7 +186,8 @@ enum {
 enum {
 	HNS_ROCE_CAP_FLAG_REREG_MR		= BIT(0),
 	HNS_ROCE_CAP_FLAG_ROCE_V1_V2		= BIT(1),
-	HNS_ROCE_CAP_FLAG_RQ_INLINE		= BIT(2)
+	HNS_ROCE_CAP_FLAG_RQ_INLINE		= BIT(2),
+	HNS_ROCE_CAP_FLAG_RECORD_DB		= BIT(3)
 };
 
 enum hns_roce_mtt_type {
@@ -186,6 +195,10 @@ enum hns_roce_mtt_type {
 	MTT_TYPE_CQE,
 };
 
+enum {
+	HNS_ROCE_DB_PER_PAGE = PAGE_SIZE / 4
+};
+
 #define HNS_ROCE_CMD_SUCCESS			1
 
 #define HNS_ROCE_PORT_DOWN			0
@@ -203,6 +216,8 @@ struct hns_roce_uar {
 struct hns_roce_ucontext {
 	struct ib_ucontext	ibucontext;
 	struct hns_roce_uar	uar;
+	struct list_head	page_list;
+	struct mutex		page_mutex;
 };
 
 struct hns_roce_pd {
@@ -335,6 +350,33 @@ struct hns_roce_buf {
 	int				page_shift;
 };
 
+struct hns_roce_db_pgdir {
+	struct list_head	list;
+	DECLARE_BITMAP(order0, HNS_ROCE_DB_PER_PAGE);
+	DECLARE_BITMAP(order1, HNS_ROCE_DB_PER_PAGE / 2);
+	unsigned long		*bits[2];
+	u32			*page;
+	dma_addr_t		db_dma;
+};
+
+struct hns_roce_user_db_page {
+	struct list_head	list;
+	struct ib_umem		*umem;
+	unsigned long		user_virt;
+	refcount_t		refcount;
+};
+
+struct hns_roce_db {
+	u32		*db_record;
+	union {
+		struct hns_roce_db_pgdir *pgdir;
+		struct hns_roce_user_db_page *user_page;
+	} u;
+	dma_addr_t	dma;
+	int		index;
+	int		order;
+};
+
 struct hns_roce_cq_buf {
 	struct hns_roce_buf hr_buf;
 	struct hns_roce_mtt hr_mtt;
@@ -343,6 +385,8 @@ struct hns_roce_cq_buf {
 struct hns_roce_cq {
 	struct ib_cq			ib_cq;
 	struct hns_roce_cq_buf		hr_buf;
+	struct hns_roce_db		db;
+	u8				db_en;
 	spinlock_t			lock;
 	struct ib_umem			*umem;
 	void (*comp)(struct hns_roce_cq *cq);
@@ -351,6 +395,7 @@ struct hns_roce_cq {
 	struct hns_roce_uar		*uar;
 	u32				cq_depth;
 	u32				cons_index;
+	u32				*set_ci_db;
 	void __iomem			*cq_db_l;
 	u16				*tptr_addr;
 	int				arm_sn;
@@ -466,6 +511,8 @@ struct hns_roce_qp {
 	struct ib_qp		ibqp;
 	struct hns_roce_buf	hr_buf;
 	struct hns_roce_wq	rq;
+	struct hns_roce_db	rdb;
+	u8			rdb_en;
 	u32			doorbell_qpn;
 	__le32			sq_signal_bits;
 	u32			sq_next_wqe;
@@ -725,6 +772,8 @@ struct hns_roce_dev {
 	spinlock_t		bt_cmd_lock;
 	struct hns_roce_ib_iboe iboe;
 
+	struct list_head        pgdir_list;
+	struct mutex            pgdir_mutex;
 	int			irq[HNS_ROCE_MAX_IRQ_NUM];
 	u8 __iomem		*reg_base;
 	struct hns_roce_caps	caps;
@@ -930,6 +979,14 @@ struct ib_cq *hns_roce_ib_create_cq(struct ib_device *ib_dev,
 int hns_roce_ib_destroy_cq(struct ib_cq *ib_cq);
 void hns_roce_free_cq(struct hns_roce_dev *hr_dev, struct hns_roce_cq *hr_cq);
 
+int hns_roce_db_map_user(struct hns_roce_ucontext *context, unsigned long virt,
+			 struct hns_roce_db *db);
+void hns_roce_db_unmap_user(struct hns_roce_ucontext *context,
+			    struct hns_roce_db *db);
+int hns_roce_alloc_db(struct hns_roce_dev *hr_dev, struct hns_roce_db *db,
+		      int order);
+void hns_roce_free_db(struct hns_roce_dev *hr_dev, struct hns_roce_db *db);
+
 void hns_roce_cq_completion(struct hns_roce_dev *hr_dev, u32 cqn);
 void hns_roce_cq_event(struct hns_roce_dev *hr_dev, u32 cqn, int event_type);
 void hns_roce_qp_event(struct hns_roce_dev *hr_dev, u32 qpn, int event_type);

drivers/infiniband/hw/hns/hns_roce_hw_v1.c

@@ -1687,13 +1687,13 @@ static int hns_roce_v1_post_mbox(struct hns_roce_dev *hr_dev, u64 in_param,
 	roce_set_field(val, ROCEE_MB6_ROCEE_MB_TOKEN_M,
 		       ROCEE_MB6_ROCEE_MB_TOKEN_S, token);
 
-	__raw_writeq(cpu_to_le64(in_param), hcr + 0);
-	__raw_writeq(cpu_to_le64(out_param), hcr + 2);
-	__raw_writel(cpu_to_le32(in_modifier), hcr + 4);
+	writeq(in_param, hcr + 0);
+	writeq(out_param, hcr + 2);
+	writel(in_modifier, hcr + 4);
 	/* Memory barrier */
 	wmb();
 
-	__raw_writel(cpu_to_le32(val), hcr + 5);
+	writel(val, hcr + 5);
 
 	mmiowb();
 

drivers/infiniband/hw/hns/hns_roce_hw_v2.c

@@ -498,7 +498,6 @@ static int hns_roce_v2_post_recv(struct ib_qp *ibqp, struct ib_recv_wr *wr,
 	struct hns_roce_v2_wqe_data_seg *dseg;
 	struct hns_roce_rinl_sge *sge_list;
 	struct device *dev = hr_dev->dev;
-	struct hns_roce_v2_db rq_db;
 	unsigned long flags;
 	void *wqe = NULL;
 	int ret = 0;
@@ -509,7 +508,7 @@ static int hns_roce_v2_post_recv(struct ib_qp *ibqp, struct ib_recv_wr *wr,
 	spin_lock_irqsave(&hr_qp->rq.lock, flags);
 	ind = hr_qp->rq.head & (hr_qp->rq.wqe_cnt - 1);
 
-	if (hr_qp->state == IB_QPS_RESET || hr_qp->state == IB_QPS_ERR) {
+	if (hr_qp->state == IB_QPS_RESET) {
 		spin_unlock_irqrestore(&hr_qp->rq.lock, flags);
 		*bad_wr = wr;
 		return -EINVAL;
@@ -564,17 +563,7 @@ out:
 		/* Memory barrier */
 		wmb();
 
-		rq_db.byte_4 = 0;
-		rq_db.parameter = 0;
-
-		roce_set_field(rq_db.byte_4, V2_DB_BYTE_4_TAG_M,
-			       V2_DB_BYTE_4_TAG_S, hr_qp->qpn);
-		roce_set_field(rq_db.byte_4, V2_DB_BYTE_4_CMD_M,
-			       V2_DB_BYTE_4_CMD_S, HNS_ROCE_V2_RQ_DB);
-		roce_set_field(rq_db.parameter, V2_DB_PARAMETER_CONS_IDX_M,
-			       V2_DB_PARAMETER_CONS_IDX_S, hr_qp->rq.head);
-
-		hns_roce_write64_k((__le32 *)&rq_db, hr_qp->rq.db_reg_l);
+		*hr_qp->rdb.db_record = hr_qp->rq.head & 0xffff;
 	}
 	spin_unlock_irqrestore(&hr_qp->rq.lock, flags);
 
@@ -1168,7 +1157,8 @@ static int hns_roce_v2_profile(struct hns_roce_dev *hr_dev)
 
 	caps->flags		= HNS_ROCE_CAP_FLAG_REREG_MR |
 				  HNS_ROCE_CAP_FLAG_ROCE_V1_V2 |
-				  HNS_ROCE_CAP_FLAG_RQ_INLINE;
+				  HNS_ROCE_CAP_FLAG_RQ_INLINE |
+				  HNS_ROCE_CAP_FLAG_RECORD_DB;
 	caps->pkey_table_len[0] = 1;
 	caps->gid_table_len[0] = HNS_ROCE_V2_GID_INDEX_NUM;
 	caps->ceqe_depth	= HNS_ROCE_V2_COMP_EQE_NUM;
@@ -1228,14 +1218,14 @@ static int hns_roce_v2_post_mbox(struct hns_roce_dev *hr_dev, u64 in_param,
 	roce_set_field(val1, HNS_ROCE_VF_MB5_TOKEN_MASK,
 		       HNS_ROCE_VF_MB5_TOKEN_SHIFT, token);
 
-	__raw_writeq(cpu_to_le64(in_param), hcr + 0);
-	__raw_writeq(cpu_to_le64(out_param), hcr + 2);
+	writeq(in_param, hcr + 0);
+	writeq(out_param, hcr + 2);
 
 	/* Memory barrier */
 	wmb();
 
-	__raw_writel(cpu_to_le32(val0), hcr + 4);
-	__raw_writel(cpu_to_le32(val1), hcr + 5);
+	writel(val0, hcr + 4);
+	writel(val1, hcr + 5);
 
 	mmiowb();
 
@@ -1507,24 +1497,7 @@ static struct hns_roce_v2_cqe *next_cqe_sw_v2(struct hns_roce_cq *hr_cq)
 
 static void hns_roce_v2_cq_set_ci(struct hns_roce_cq *hr_cq, u32 cons_index)
 {
-	struct hns_roce_v2_cq_db cq_db;
-
-	cq_db.byte_4 = 0;
-	cq_db.parameter = 0;
-
-	roce_set_field(cq_db.byte_4, V2_CQ_DB_BYTE_4_TAG_M,
-		       V2_CQ_DB_BYTE_4_TAG_S, hr_cq->cqn);
-	roce_set_field(cq_db.byte_4, V2_CQ_DB_BYTE_4_CMD_M,
-		       V2_CQ_DB_BYTE_4_CMD_S, HNS_ROCE_V2_CQ_DB_PTR);
-
-	roce_set_field(cq_db.parameter, V2_CQ_DB_PARAMETER_CONS_IDX_M,
-		       V2_CQ_DB_PARAMETER_CONS_IDX_S,
-		       cons_index & ((hr_cq->cq_depth << 1) - 1));
-	roce_set_field(cq_db.parameter, V2_CQ_DB_PARAMETER_CMD_SN_M,
-		       V2_CQ_DB_PARAMETER_CMD_SN_S, 1);
-
-	hns_roce_write64_k((__be32 *)&cq_db, hr_cq->cq_db_l);
-
+	*hr_cq->set_ci_db = cons_index & 0xffffff;
 }
 
 static void __hns_roce_v2_cq_clean(struct hns_roce_cq *hr_cq, u32 qpn,
@@ -1637,6 +1610,16 @@ static void hns_roce_v2_write_cqc(struct hns_roce_dev *hr_dev,
 	roce_set_field(cq_context->byte_40_cqe_ba, V2_CQC_BYTE_40_CQE_BA_M,
 		       V2_CQC_BYTE_40_CQE_BA_S, (dma_handle >> (32 + 3)));
 
+	if (hr_cq->db_en)
+		roce_set_bit(cq_context->byte_44_db_record,
+			     V2_CQC_BYTE_44_DB_RECORD_EN_S, 1);
+
+	roce_set_field(cq_context->byte_44_db_record,
+		       V2_CQC_BYTE_44_DB_RECORD_ADDR_M,
+		       V2_CQC_BYTE_44_DB_RECORD_ADDR_S,
+		       ((u32)hr_cq->db.dma) >> 1);
+	cq_context->db_record_addr = hr_cq->db.dma >> 32;
+
 	roce_set_field(cq_context->byte_56_cqe_period_maxcnt,
 		       V2_CQC_BYTE_56_CQ_MAX_CNT_M,
 		       V2_CQC_BYTE_56_CQ_MAX_CNT_S,
@@ -2274,6 +2257,23 @@ static void modify_qp_reset_to_init(struct ib_qp *ibqp,
 		hr_qp->qkey = attr->qkey;
 	}
 
+	if (hr_qp->rdb_en) {
+		roce_set_bit(context->byte_68_rq_db,
+			     V2_QPC_BYTE_68_RQ_RECORD_EN_S, 1);
+		roce_set_bit(qpc_mask->byte_68_rq_db,
+			     V2_QPC_BYTE_68_RQ_RECORD_EN_S, 0);
+	}
+
+	roce_set_field(context->byte_68_rq_db,
+		       V2_QPC_BYTE_68_RQ_DB_RECORD_ADDR_M,
+		       V2_QPC_BYTE_68_RQ_DB_RECORD_ADDR_S,
+		       ((u32)hr_qp->rdb.dma) >> 1);
+	roce_set_field(qpc_mask->byte_68_rq_db,
+		       V2_QPC_BYTE_68_RQ_DB_RECORD_ADDR_M,
+		       V2_QPC_BYTE_68_RQ_DB_RECORD_ADDR_S, 0);
+	context->rq_db_record_addr = hr_qp->rdb.dma >> 32;
+	qpc_mask->rq_db_record_addr = 0;
+
 	roce_set_bit(context->byte_76_srqn_op_en, V2_QPC_BYTE_76_RQIE_S, 1);
 	roce_set_bit(qpc_mask->byte_76_srqn_op_en, V2_QPC_BYTE_76_RQIE_S, 0);
 
@@ -3211,6 +3211,8 @@ static int hns_roce_v2_modify_qp(struct ib_qp *ibqp,
 		hr_qp->sq.tail = 0;
 		hr_qp->sq_next_wqe = 0;
 		hr_qp->next_sge = 0;
+		if (hr_qp->rq.wqe_cnt)
+			*hr_qp->rdb.db_record = 0;
 	}
 
 out:
@@ -3437,11 +3439,17 @@ static int hns_roce_v2_destroy_qp_common(struct hns_roce_dev *hr_dev,
 	hns_roce_mtt_cleanup(hr_dev, &hr_qp->mtt);
 
 	if (is_user) {
+		if (hr_qp->rq.wqe_cnt && (hr_qp->rdb_en == 1))
+			hns_roce_db_unmap_user(
+				to_hr_ucontext(hr_qp->ibqp.uobject->context),
+				&hr_qp->rdb);
 		ib_umem_release(hr_qp->umem);
 	} else {
 		kfree(hr_qp->sq.wrid);
 		kfree(hr_qp->rq.wrid);
 		hns_roce_buf_free(hr_dev, hr_qp->buff_size, &hr_qp->hr_buf);
+		if (hr_qp->rq.wqe_cnt)
+			hns_roce_free_db(hr_dev, &hr_qp->rdb);
 	}
 
 	if (hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_RQ_INLINE) {

drivers/infiniband/hw/hns/hns_roce_hw_v2.h

@@ -299,6 +299,9 @@ struct hns_roce_v2_cq_context {
 
 #define	V2_CQC_BYTE_44_DB_RECORD_EN_S 0
 
+#define	V2_CQC_BYTE_44_DB_RECORD_ADDR_S 1
+#define	V2_CQC_BYTE_44_DB_RECORD_ADDR_M GENMASK(31, 1)
+
 #define	V2_CQC_BYTE_52_CQE_CNT_S 0
 #define	V2_CQC_BYTE_52_CQE_CNT_M GENMASK(23, 0)
 

drivers/infiniband/hw/hns/hns_roce_main.c

@@ -74,12 +74,11 @@ static int hns_roce_set_mac(struct hns_roce_dev *hr_dev, u8 port, u8 *addr)
 	return hr_dev->hw->set_mac(hr_dev, phy_port, addr);
 }
 
-static int hns_roce_add_gid(struct ib_device *device, u8 port_num,
-			    unsigned int index, const union ib_gid *gid,
+static int hns_roce_add_gid(const union ib_gid *gid,
 			    const struct ib_gid_attr *attr, void **context)
 {
-	struct hns_roce_dev *hr_dev = to_hr_dev(device);
-	u8 port = port_num - 1;
+	struct hns_roce_dev *hr_dev = to_hr_dev(attr->device);
+	u8 port = attr->port_num - 1;
 	unsigned long flags;
 	int ret;
 
@@ -88,20 +87,20 @@ static int hns_roce_add_gid(struct ib_device *device, u8 port_num,
 
 	spin_lock_irqsave(&hr_dev->iboe.lock, flags);
 
-	ret = hr_dev->hw->set_gid(hr_dev, port, index, (union ib_gid *)gid,
-				   attr);
+	ret = hr_dev->hw->set_gid(hr_dev, port, attr->index,
+				  (union ib_gid *)gid, attr);
 
 	spin_unlock_irqrestore(&hr_dev->iboe.lock, flags);
 
 	return ret;
 }
 
-static int hns_roce_del_gid(struct ib_device *device, u8 port_num,
-			    unsigned int index, void **context)
+static int hns_roce_del_gid(const struct ib_gid_attr *attr, void **context)
 {
-	struct hns_roce_dev *hr_dev = to_hr_dev(device);
+	struct hns_roce_dev *hr_dev = to_hr_dev(attr->device);
+	struct ib_gid_attr zattr = { };
 	union ib_gid zgid = { {0} };
-	u8 port = port_num - 1;
+	u8 port = attr->port_num - 1;
 	unsigned long flags;
 	int ret;
 
@@ -110,7 +109,7 @@ static int hns_roce_del_gid(struct ib_device *device, u8 port_num,
 
 	spin_lock_irqsave(&hr_dev->iboe.lock, flags);
 
-	ret = hr_dev->hw->set_gid(hr_dev, port, index, &zgid, NULL);
+	ret = hr_dev->hw->set_gid(hr_dev, port, attr->index, &zgid, &zattr);
 
 	spin_unlock_irqrestore(&hr_dev->iboe.lock, flags);
 
@@ -295,12 +294,6 @@ static enum rdma_link_layer hns_roce_get_link_layer(struct ib_device *device,
 	return IB_LINK_LAYER_ETHERNET;
 }
 
-static int hns_roce_query_gid(struct ib_device *ib_dev, u8 port_num, int index,
-			      union ib_gid *gid)
-{
-	return 0;
-}
-
 static int hns_roce_query_pkey(struct ib_device *ib_dev, u8 port, u16 index,
 			       u16 *pkey)
 {
@@ -337,7 +330,7 @@ static struct ib_ucontext *hns_roce_alloc_ucontext(struct ib_device *ib_dev,
 {
 	int ret = 0;
 	struct hns_roce_ucontext *context;
-	struct hns_roce_ib_alloc_ucontext_resp resp;
+	struct hns_roce_ib_alloc_ucontext_resp resp = {};
 	struct hns_roce_dev *hr_dev = to_hr_dev(ib_dev);
 
 	resp.qp_tab_size = hr_dev->caps.num_qps;
@@ -350,6 +343,11 @@ static struct ib_ucontext *hns_roce_alloc_ucontext(struct ib_device *ib_dev,
 	if (ret)
 		goto error_fail_uar_alloc;
 
+	if (hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_RECORD_DB) {
+		INIT_LIST_HEAD(&context->page_list);
+		mutex_init(&context->page_mutex);
+	}
+
 	ret = ib_copy_to_udata(udata, &resp, sizeof(resp));
 	if (ret)
 		goto error_fail_copy_to_udata;
@@ -476,7 +474,6 @@ static int hns_roce_register_device(struct hns_roce_dev *hr_dev)
 	ib_dev->modify_port		= hns_roce_modify_port;
 	ib_dev->get_link_layer		= hns_roce_get_link_layer;
 	ib_dev->get_netdev		= hns_roce_get_netdev;
-	ib_dev->query_gid		= hns_roce_query_gid;
 	ib_dev->add_gid			= hns_roce_add_gid;
 	ib_dev->del_gid			= hns_roce_del_gid;
 	ib_dev->query_pkey		= hns_roce_query_pkey;
@@ -520,6 +517,7 @@ static int hns_roce_register_device(struct hns_roce_dev *hr_dev)
 	/* OTHERS */
 	ib_dev->get_port_immutable	= hns_roce_port_immutable;
 
+	ib_dev->driver_id = RDMA_DRIVER_HNS;
 	ret = ib_register_device(ib_dev, NULL);
 	if (ret) {
 		dev_err(dev, "ib_register_device failed!\n");
@@ -659,6 +657,11 @@ static int hns_roce_setup_hca(struct hns_roce_dev *hr_dev)
 	spin_lock_init(&hr_dev->sm_lock);
 	spin_lock_init(&hr_dev->bt_cmd_lock);
 
+	if (hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_RECORD_DB) {
+		INIT_LIST_HEAD(&hr_dev->pgdir_list);
+		mutex_init(&hr_dev->pgdir_mutex);
+	}
+
 	ret = hns_roce_init_uar_table(hr_dev);
 	if (ret) {
 		dev_err(dev, "Failed to initialize uar table. aborting\n");

drivers/infiniband/hw/hns/hns_roce_mr.c

@@ -933,7 +933,7 @@ int hns_roce_ib_umem_write_mtt(struct hns_roce_dev *hr_dev,
 		ret = hns_roce_write_mtt(hr_dev, mtt, n, i, pages);
 
 out:
-	free_page((unsigned long) pages);
+	free_pages((unsigned long) pages, order);
 	return ret;
 }
 

drivers/infiniband/hw/hns/hns_roce_pd.c

@@ -32,6 +32,7 @@
 
 #include <linux/platform_device.h>
 #include <linux/pci.h>
+#include <uapi/rdma/hns-abi.h>
 #include "hns_roce_device.h"
 
 static int hns_roce_pd_alloc(struct hns_roce_dev *hr_dev, unsigned long *pdn)
@@ -77,7 +78,9 @@ struct ib_pd *hns_roce_alloc_pd(struct ib_device *ib_dev,
 	}
 
 	if (context) {
-		if (ib_copy_to_udata(udata, &pd->pdn, sizeof(u64))) {
+		struct hns_roce_ib_alloc_pd_resp uresp = {.pdn = pd->pdn};
+
+		if (ib_copy_to_udata(udata, &uresp, sizeof(uresp))) {
 			hns_roce_pd_free(to_hr_dev(ib_dev), pd->pdn);
 			dev_err(dev, "[alloc_pd]ib_copy_to_udata failed!\n");
 			kfree(pd);

drivers/infiniband/hw/hns/hns_roce_qp.c

@@ -489,6 +489,15 @@ static int hns_roce_set_kernel_sq_size(struct hns_roce_dev *hr_dev,
 	return 0;
 }
 
+static int hns_roce_qp_has_rq(struct ib_qp_init_attr *attr)
+{
+	if (attr->qp_type == IB_QPT_XRC_INI ||
+	    attr->qp_type == IB_QPT_XRC_TGT || attr->srq)
+		return 0;
+
+	return 1;
+}
+
 static int hns_roce_create_qp_common(struct hns_roce_dev *hr_dev,
 				     struct ib_pd *ib_pd,
 				     struct ib_qp_init_attr *init_attr,
@@ -497,6 +506,7 @@ static int hns_roce_create_qp_common(struct hns_roce_dev *hr_dev,
 {
 	struct device *dev = hr_dev->dev;
 	struct hns_roce_ib_create_qp ucmd;
+	struct hns_roce_ib_create_qp_resp resp = {};
 	unsigned long qpn = 0;
 	int ret = 0;
 	u32 page_shift;
@@ -602,6 +612,18 @@ static int hns_roce_create_qp_common(struct hns_roce_dev *hr_dev,
 			dev_err(dev, "hns_roce_ib_umem_write_mtt error for create qp\n");
 			goto err_mtt;
 		}
+
+		if ((hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_RECORD_DB) &&
+		    (udata->outlen >= sizeof(resp)) &&
+		    hns_roce_qp_has_rq(init_attr)) {
+			ret = hns_roce_db_map_user(
+					to_hr_ucontext(ib_pd->uobject->context),
+					ucmd.db_addr, &hr_qp->rdb);
+			if (ret) {
+				dev_err(dev, "rp record doorbell map failed!\n");
+				goto err_mtt;
+			}
+		}
 	} else {
 		if (init_attr->create_flags &
 		    IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK) {
@@ -630,6 +652,16 @@ static int hns_roce_create_qp_common(struct hns_roce_dev *hr_dev,
 		hr_qp->rq.db_reg_l = hr_dev->reg_base + hr_dev->odb_offset +
 				     DB_REG_OFFSET * hr_dev->priv_uar.index;
 
+		if ((hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_RECORD_DB) &&
+		    hns_roce_qp_has_rq(init_attr)) {
+			ret = hns_roce_alloc_db(hr_dev, &hr_qp->rdb, 0);
+			if (ret) {
+				dev_err(dev, "rq record doorbell alloc failed!\n");
+				goto err_rq_sge_list;
+			}
+			*hr_qp->rdb.db_record = 0;
+		}
+
 		/* Allocate QP buf */
 		page_shift = PAGE_SHIFT + hr_dev->caps.mtt_buf_pg_sz;
 		if (hns_roce_buf_alloc(hr_dev, hr_qp->buff_size,
@@ -637,7 +669,7 @@ static int hns_roce_create_qp_common(struct hns_roce_dev *hr_dev,
 				       &hr_qp->hr_buf, page_shift)) {
 			dev_err(dev, "hns_roce_buf_alloc error!\n");
 			ret = -ENOMEM;
-			goto err_rq_sge_list;
+			goto err_db;
 		}
 
 		hr_qp->mtt.mtt_type = MTT_TYPE_WQE;
@@ -698,17 +730,44 @@ static int hns_roce_create_qp_common(struct hns_roce_dev *hr_dev,
 	else
 		hr_qp->doorbell_qpn = cpu_to_le64(hr_qp->qpn);
 
+	if (ib_pd->uobject && (udata->outlen >= sizeof(resp)) &&
+		(hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_RECORD_DB)) {
+
+		/* indicate kernel supports record db */
+		resp.cap_flags |= HNS_ROCE_SUPPORT_RQ_RECORD_DB;
+		ret = ib_copy_to_udata(udata, &resp, sizeof(resp));
+		if (ret)
+			goto err_qp;
+
+		hr_qp->rdb_en = 1;
+	}
 	hr_qp->event = hns_roce_ib_qp_event;
 
 	return 0;
 
+err_qp:
+	if (init_attr->qp_type == IB_QPT_GSI &&
+		hr_dev->hw_rev == HNS_ROCE_HW_VER1)
+		hns_roce_qp_remove(hr_dev, hr_qp);
+	else
+		hns_roce_qp_free(hr_dev, hr_qp);
+
 err_qpn:
 	if (!sqpn)
 		hns_roce_release_range_qp(hr_dev, qpn, 1);
 
 err_wrid:
-	kfree(hr_qp->sq.wrid);
-	kfree(hr_qp->rq.wrid);
+	if (ib_pd->uobject) {
+		if ((hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_RECORD_DB) &&
+		    (udata->outlen >= sizeof(resp)) &&
+		    hns_roce_qp_has_rq(init_attr))
+			hns_roce_db_unmap_user(
+					to_hr_ucontext(ib_pd->uobject->context),
+					&hr_qp->rdb);
+	} else {
+		kfree(hr_qp->sq.wrid);
+		kfree(hr_qp->rq.wrid);
+	}
 
 err_mtt:
 	hns_roce_mtt_cleanup(hr_dev, &hr_qp->mtt);
@@ -719,6 +778,11 @@ err_buf:
 	else
 		hns_roce_buf_free(hr_dev, hr_qp->buff_size, &hr_qp->hr_buf);
 
+err_db:
+	if (!ib_pd->uobject && hns_roce_qp_has_rq(init_attr) &&
+	    (hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_RECORD_DB))
+		hns_roce_free_db(hr_dev, &hr_qp->rdb);
+
 err_rq_sge_list:
 	if (hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_RQ_INLINE)
 		kfree(hr_qp->rq_inl_buf.wqe_list[0].sg_list);

drivers/infiniband/hw/i40iw/i40iw.h

@@ -60,7 +60,7 @@
 #include <i40e_client.h>
 #include "i40iw_type.h"
 #include "i40iw_p.h"
-#include "i40iw_ucontext.h"
+#include <rdma/i40iw-abi.h>
 #include "i40iw_pble.h"
 #include "i40iw_verbs.h"
 #include "i40iw_cm.h"
@@ -559,18 +559,25 @@ void i40iw_next_iw_state(struct i40iw_qp *iwqp,
 			 u8 state, u8 del_hash,
 			 u8 term, u8 term_len);
 int i40iw_send_syn(struct i40iw_cm_node *cm_node, u32 sendack);
+int i40iw_send_reset(struct i40iw_cm_node *cm_node);
 struct i40iw_cm_node *i40iw_find_node(struct i40iw_cm_core *cm_core,
 				      u16 rem_port,
 				      u32 *rem_addr,
 				      u16 loc_port,
 				      u32 *loc_addr,
-				      bool add_refcnt);
+				      bool add_refcnt,
+				      bool accelerated_list);
 
 enum i40iw_status_code i40iw_hw_flush_wqes(struct i40iw_device *iwdev,
 					   struct i40iw_sc_qp *qp,
 					   struct i40iw_qp_flush_info *info,
 					   bool wait);
 
+void i40iw_gen_ae(struct i40iw_device *iwdev,
+		  struct i40iw_sc_qp *qp,
+		  struct i40iw_gen_ae_info *info,
+		  bool wait);
+
 void i40iw_copy_ip_ntohl(u32 *dst, __be32 *src);
 struct ib_mr *i40iw_reg_phys_mr(struct ib_pd *ib_pd,
 				u64 addr,

drivers/infiniband/hw/i40iw/i40iw_cm.c

@@ -539,7 +539,7 @@ static struct i40iw_puda_buf *i40iw_form_cm_frame(struct i40iw_cm_node *cm_node,
  * i40iw_send_reset - Send RST packet
  * @cm_node: connection's node
  */
-static int i40iw_send_reset(struct i40iw_cm_node *cm_node)
+int i40iw_send_reset(struct i40iw_cm_node *cm_node)
 {
 	struct i40iw_puda_buf *sqbuf;
 	int flags = SET_RST | SET_ACK;
@@ -1182,6 +1182,26 @@ static void i40iw_handle_close_entry(struct i40iw_cm_node *cm_node, u32 rem_node
 	cm_node->close_entry = NULL;
 }
 
+/**
+ * i40iw_build_timer_list - Add cm_nodes to timer list
+ * @timer_list: ptr to timer list
+ * @hte: ptr to accelerated or non-accelerated list
+ */
+static void i40iw_build_timer_list(struct list_head *timer_list,
+				   struct list_head *hte)
+{
+	struct i40iw_cm_node *cm_node;
+	struct list_head *list_core_temp, *list_node;
+
+	list_for_each_safe(list_node, list_core_temp, hte) {
+		cm_node = container_of(list_node, struct i40iw_cm_node, list);
+		if (cm_node->close_entry || cm_node->send_entry) {
+			atomic_inc(&cm_node->ref_count);
+			list_add(&cm_node->timer_entry, timer_list);
+		}
+	}
+}
+
 /**
  * i40iw_cm_timer_tick - system's timer expired callback
  * @pass: Pointing to cm_core
@@ -1202,15 +1222,10 @@ static void i40iw_cm_timer_tick(struct timer_list *t)
 	struct list_head timer_list;
 
 	INIT_LIST_HEAD(&timer_list);
-	spin_lock_irqsave(&cm_core->ht_lock, flags);
 
-	list_for_each_safe(list_node, list_core_temp, &cm_core->connected_nodes) {
-		cm_node = container_of(list_node, struct i40iw_cm_node, list);
-		if (cm_node->close_entry || cm_node->send_entry) {
-			atomic_inc(&cm_node->ref_count);
-			list_add(&cm_node->timer_entry, &timer_list);
-		}
-	}
+	spin_lock_irqsave(&cm_core->ht_lock, flags);
+	i40iw_build_timer_list(&timer_list, &cm_core->non_accelerated_list);
+	i40iw_build_timer_list(&timer_list, &cm_core->accelerated_list);
 	spin_unlock_irqrestore(&cm_core->ht_lock, flags);
 
 	list_for_each_safe(list_node, list_core_temp, &timer_list) {
@@ -1406,19 +1421,22 @@ static int i40iw_send_fin(struct i40iw_cm_node *cm_node)
  * @loc_port: local tcp port num
  * @loc_addr: loc ip addr
  * @add_refcnt: flag to increment refcount of cm_node
+ * @accelerated_list: flag for accelerated vs non-accelerated list to search
  */
 struct i40iw_cm_node *i40iw_find_node(struct i40iw_cm_core *cm_core,
 				      u16 rem_port,
 				      u32 *rem_addr,
 				      u16 loc_port,
 				      u32 *loc_addr,
-				      bool add_refcnt)
+				      bool add_refcnt,
+				      bool accelerated_list)
 {
 	struct list_head *hte;
 	struct i40iw_cm_node *cm_node;
 	unsigned long flags;
 
-	hte = &cm_core->connected_nodes;
+	hte = accelerated_list ?
+	      &cm_core->accelerated_list : &cm_core->non_accelerated_list;
 
 	/* walk list and find cm_node associated with this session ID */
 	spin_lock_irqsave(&cm_core->ht_lock, flags);
@@ -1487,21 +1505,39 @@ static struct i40iw_cm_listener *i40iw_find_listener(
 static void i40iw_add_hte_node(struct i40iw_cm_core *cm_core,
 			       struct i40iw_cm_node *cm_node)
 {
-	struct list_head *hte;
 	unsigned long flags;
 
 	if (!cm_node || !cm_core) {
 		i40iw_pr_err("cm_node or cm_core == NULL\n");
 		return;
 	}
-	spin_lock_irqsave(&cm_core->ht_lock, flags);
 
-	/* get a handle on the hash table element (list head for this slot) */
-	hte = &cm_core->connected_nodes;
-	list_add_tail(&cm_node->list, hte);
+	spin_lock_irqsave(&cm_core->ht_lock, flags);
+	list_add_tail(&cm_node->list, &cm_core->non_accelerated_list);
 	spin_unlock_irqrestore(&cm_core->ht_lock, flags);
 }
 
+/**
+ * i40iw_find_port - find port that matches reference port
+ * @port: port number
+ * @accelerated_list: flag for accelerated vs non-accelerated list
+ */
+static bool i40iw_find_port(struct i40iw_cm_core *cm_core, u16 port,
+			    bool accelerated_list)
+{
+	struct list_head *hte;
+	struct i40iw_cm_node *cm_node;
+
+	hte = accelerated_list ?
+	      &cm_core->accelerated_list : &cm_core->non_accelerated_list;
+
+	list_for_each_entry(cm_node, hte, list) {
+		if (cm_node->loc_port == port)
+			return true;
+	}
+	return false;
+}
+
 /**
  * i40iw_port_in_use - determine if port is in use
  * @port: port number
@@ -1510,19 +1546,14 @@ static void i40iw_add_hte_node(struct i40iw_cm_core *cm_core,
 static bool i40iw_port_in_use(struct i40iw_cm_core *cm_core, u16 port, bool active_side)
 {
 	struct i40iw_cm_listener *listen_node;
-	struct i40iw_cm_node *cm_node;
 	unsigned long flags;
 	bool ret = false;
 
 	if (active_side) {
-		/* search connected node list */
 		spin_lock_irqsave(&cm_core->ht_lock, flags);
-		list_for_each_entry(cm_node, &cm_core->connected_nodes, list) {
-			if (cm_node->loc_port == port) {
-				ret = true;
-				break;
-			}
-		}
+		ret = i40iw_find_port(cm_core, port, true);
+		if (!ret)
+			ret = i40iw_find_port(cm_core, port, false);
 		if (!ret)
 			clear_bit(port, cm_core->active_side_ports);
 		spin_unlock_irqrestore(&cm_core->ht_lock, flags);
@@ -1829,9 +1860,11 @@ static int i40iw_dec_refcnt_listen(struct i40iw_cm_core *cm_core,
 	INIT_LIST_HEAD(&reset_list);
 	if (free_hanging_nodes) {
 		spin_lock_irqsave(&cm_core->ht_lock, flags);
-		list_for_each_safe(list_pos, list_temp, &cm_core->connected_nodes) {
+		list_for_each_safe(list_pos,
+				   list_temp, &cm_core->non_accelerated_list) {
 			cm_node = container_of(list_pos, struct i40iw_cm_node, list);
-			if ((cm_node->listener == listener) && !cm_node->accelerated) {
+			if ((cm_node->listener == listener) &&
+			    !cm_node->accelerated) {
 				atomic_inc(&cm_node->ref_count);
 				list_add(&cm_node->reset_entry, &reset_list);
 			}
@@ -3144,7 +3177,8 @@ void i40iw_receive_ilq(struct i40iw_sc_vsi *vsi, struct i40iw_puda_buf *rbuf)
 				  cm_info.rem_addr,
 				  cm_info.loc_port,
 				  cm_info.loc_addr,
-				  true);
+				  true,
+				  false);
 
 	if (!cm_node) {
 		/* Only type of packet accepted are for */
@@ -3202,7 +3236,8 @@ void i40iw_setup_cm_core(struct i40iw_device *iwdev)
 	cm_core->iwdev = iwdev;
 	cm_core->dev = &iwdev->sc_dev;
 
-	INIT_LIST_HEAD(&cm_core->connected_nodes);
+	INIT_LIST_HEAD(&cm_core->accelerated_list);
+	INIT_LIST_HEAD(&cm_core->non_accelerated_list);
 	INIT_LIST_HEAD(&cm_core->listen_nodes);
 
 	timer_setup(&cm_core->tcp_timer, i40iw_cm_timer_tick, 0);
@@ -3585,6 +3620,7 @@ int i40iw_accept(struct iw_cm_id *cm_id, struct iw_cm_conn_param *conn_param)
 	struct i40iw_qp *iwqp;
 	struct i40iw_device *iwdev;
 	struct i40iw_sc_dev *dev;
+	struct i40iw_cm_core *cm_core;
 	struct i40iw_cm_node *cm_node;
 	struct ib_qp_attr attr;
 	int passive_state;
@@ -3594,6 +3630,7 @@ int i40iw_accept(struct iw_cm_id *cm_id, struct iw_cm_conn_param *conn_param)
 	struct i40iw_kmem_info accept;
 	enum i40iw_status_code status;
 	u64 tagged_offset;
+	unsigned long flags;
 
 	memset(&attr, 0, sizeof(attr));
 	ibqp = i40iw_get_qp(cm_id->device, conn_param->qpn);
@@ -3603,6 +3640,7 @@ int i40iw_accept(struct iw_cm_id *cm_id, struct iw_cm_conn_param *conn_param)
 	iwqp = to_iwqp(ibqp);
 	iwdev = iwqp->iwdev;
 	dev = &iwdev->sc_dev;
+	cm_core = &iwdev->cm_core;
 	cm_node = (struct i40iw_cm_node *)cm_id->provider_data;
 
 	if (((struct sockaddr_in *)&cm_id->local_addr)->sin_family == AF_INET) {
@@ -3697,6 +3735,10 @@ int i40iw_accept(struct iw_cm_id *cm_id, struct iw_cm_conn_param *conn_param)
 	i40iw_modify_qp(&iwqp->ibqp, &attr, IB_QP_STATE, NULL);
 
 	cm_node->accelerated = true;
+	spin_lock_irqsave(&cm_core->ht_lock, flags);
+	list_move_tail(&cm_node->list, &cm_core->accelerated_list);
+	spin_unlock_irqrestore(&cm_core->ht_lock, flags);
+
 	status =
 		i40iw_send_cm_event(cm_node, cm_id, IW_CM_EVENT_ESTABLISHED, 0);
 	if (status)
@@ -4026,10 +4068,12 @@ static void i40iw_cm_event_connected(struct i40iw_cm_event *event)
 {
 	struct i40iw_qp *iwqp;
 	struct i40iw_device *iwdev;
+	struct i40iw_cm_core *cm_core;
 	struct i40iw_cm_node *cm_node;
 	struct i40iw_sc_dev *dev;
 	struct ib_qp_attr attr;
 	struct iw_cm_id *cm_id;
+	unsigned long flags;
 	int status;
 	bool read0;
 
@@ -4038,6 +4082,7 @@ static void i40iw_cm_event_connected(struct i40iw_cm_event *event)
 	iwqp = (struct i40iw_qp *)cm_id->provider_data;
 	iwdev = to_iwdev(iwqp->ibqp.device);
 	dev = &iwdev->sc_dev;
+	cm_core = &iwdev->cm_core;
 
 	if (iwqp->destroyed) {
 		status = -ETIMEDOUT;
@@ -4057,6 +4102,9 @@ static void i40iw_cm_event_connected(struct i40iw_cm_event *event)
 	i40iw_modify_qp(&iwqp->ibqp, &attr, IB_QP_STATE, NULL);
 
 	cm_node->accelerated = true;
+	spin_lock_irqsave(&cm_core->ht_lock, flags);
+	list_move_tail(&cm_node->list, &cm_core->accelerated_list);
+	spin_unlock_irqrestore(&cm_core->ht_lock, flags);
 	status = i40iw_send_cm_event(cm_node, cm_id, IW_CM_EVENT_CONNECT_REPLY,
 				     0);
 	if (status)
@@ -4256,25 +4304,38 @@ void i40iw_cm_teardown_connections(struct i40iw_device *iwdev, u32 *ipaddr,
 	struct list_head *list_node;
 	struct i40iw_cm_node *cm_node;
 	unsigned long flags;
-	struct list_head connected_list;
+	struct list_head teardown_list;
 	struct ib_qp_attr attr;
 
-	INIT_LIST_HEAD(&connected_list);
+	INIT_LIST_HEAD(&teardown_list);
 	spin_lock_irqsave(&cm_core->ht_lock, flags);
-	list_for_each_safe(list_node, list_core_temp, &cm_core->connected_nodes) {
+	list_for_each_safe(list_node, list_core_temp,
+			   &cm_core->accelerated_list) {
+		cm_node = container_of(list_node, struct i40iw_cm_node, list);
+		if (disconnect_all ||
+		    (nfo->vlan_id == cm_node->vlan_id &&
+		    (!memcmp(cm_node->loc_addr, ipaddr, nfo->ipv4 ? 4 : 16) ||
+		     !memcmp(cm_node->rem_addr, ipaddr, nfo->ipv4 ? 4 : 16)))) {
+			atomic_inc(&cm_node->ref_count);
+			list_add(&cm_node->teardown_entry, &teardown_list);
+		}
+	}
+	list_for_each_safe(list_node, list_core_temp,
+			   &cm_core->non_accelerated_list) {
 		cm_node = container_of(list_node, struct i40iw_cm_node, list);
 		if (disconnect_all ||
 		    (nfo->vlan_id == cm_node->vlan_id &&
 		    (!memcmp(cm_node->loc_addr, ipaddr, nfo->ipv4 ? 4 : 16) ||
 		     !memcmp(cm_node->rem_addr, ipaddr, nfo->ipv4 ? 4 : 16)))) {
 			atomic_inc(&cm_node->ref_count);
-			list_add(&cm_node->connected_entry, &connected_list);
+			list_add(&cm_node->teardown_entry, &teardown_list);
 		}
 	}
 	spin_unlock_irqrestore(&cm_core->ht_lock, flags);
 
-	list_for_each_safe(list_node, list_core_temp, &connected_list) {
-		cm_node = container_of(list_node, struct i40iw_cm_node, connected_entry);
+	list_for_each_safe(list_node, list_core_temp, &teardown_list) {
+		cm_node = container_of(list_node, struct i40iw_cm_node,
+				       teardown_entry);
 		attr.qp_state = IB_QPS_ERR;
 		i40iw_modify_qp(&cm_node->iwqp->ibqp, &attr, IB_QP_STATE, NULL);
 		if (iwdev->reset)

drivers/infiniband/hw/i40iw/i40iw_cm.h

@@ -341,7 +341,7 @@ struct i40iw_cm_node {
 	int accept_pend;
 	struct list_head timer_entry;
 	struct list_head reset_entry;
-	struct list_head connected_entry;
+	struct list_head teardown_entry;
 	atomic_t passive_state;
 	bool qhash_set;
 	u8 user_pri;
@@ -403,7 +403,8 @@ struct i40iw_cm_core {
 	struct i40iw_sc_dev *dev;
 
 	struct list_head listen_nodes;
-	struct list_head connected_nodes;
+	struct list_head accelerated_list;
+	struct list_head non_accelerated_list;
 
 	struct timer_list tcp_timer;
 

drivers/infiniband/hw/i40iw/i40iw_ctrl.c

@@ -2614,10 +2614,8 @@ static enum i40iw_status_code i40iw_sc_qp_flush_wqes(
 
 	qp->flush_sq |= flush_sq;
 	qp->flush_rq |= flush_rq;
-	if (!flush_sq && !flush_rq) {
-		if (info->ae_code != I40IW_AE_LLP_RECEIVED_MPA_CRC_ERROR)
-			return 0;
-	}
+	if (!flush_sq && !flush_rq)
+		return 0;
 
 	cqp = qp->pd->dev->cqp;
 	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
@@ -2658,6 +2656,49 @@ static enum i40iw_status_code i40iw_sc_qp_flush_wqes(
 	return 0;
 }
 
+/**
+ * i40iw_sc_gen_ae - generate AE, currently uses flush WQE CQP OP
+ * @qp: sc qp
+ * @info: gen ae information
+ * @scratch: u64 saved to be used during cqp completion
+ * @post_sq: flag for cqp db to ring
+ */
+static enum i40iw_status_code i40iw_sc_gen_ae(
+				struct i40iw_sc_qp *qp,
+				struct i40iw_gen_ae_info *info,
+				u64 scratch,
+				bool post_sq)
+{
+	u64 temp;
+	u64 *wqe;
+	struct i40iw_sc_cqp *cqp;
+	u64 header;
+
+	cqp = qp->pd->dev->cqp;
+	wqe = i40iw_sc_cqp_get_next_send_wqe(cqp, scratch);
+	if (!wqe)
+		return I40IW_ERR_RING_FULL;
+
+	temp = info->ae_code |
+	       LS_64(info->ae_source, I40IW_CQPSQ_FWQE_AESOURCE);
+
+	set_64bit_val(wqe, 8, temp);
+
+	header = qp->qp_uk.qp_id |
+		 LS_64(I40IW_CQP_OP_GEN_AE, I40IW_CQPSQ_OPCODE) |
+		 LS_64(1, I40IW_CQPSQ_FWQE_GENERATE_AE) |
+		 LS_64(cqp->polarity, I40IW_CQPSQ_WQEVALID);
+
+	i40iw_insert_wqe_hdr(wqe, header);
+
+	i40iw_debug_buf(cqp->dev, I40IW_DEBUG_WQE, "GEN_AE WQE",
+			wqe, I40IW_CQP_WQE_SIZE * 8);
+
+	if (post_sq)
+		i40iw_sc_cqp_post_sq(cqp);
+	return 0;
+}
+
 /**
  * i40iw_sc_qp_upload_context - upload qp's context
  * @dev: sc device struct
@@ -4148,6 +4189,13 @@ static enum i40iw_status_code i40iw_exec_cqp_cmd(struct i40iw_sc_dev *dev,
 				pcmdinfo->in.u.qp_flush_wqes.
 				scratch, pcmdinfo->post_sq);
 		break;
+	case OP_GEN_AE:
+		status = i40iw_sc_gen_ae(
+				pcmdinfo->in.u.gen_ae.qp,
+				&pcmdinfo->in.u.gen_ae.info,
+				pcmdinfo->in.u.gen_ae.scratch,
+				pcmdinfo->post_sq);
+		break;
 	case OP_ADD_ARP_CACHE_ENTRY:
 		status = i40iw_sc_add_arp_cache_entry(
 				pcmdinfo->in.u.add_arp_cache_entry.cqp,

drivers/infiniband/hw/i40iw/i40iw_d.h

@@ -418,6 +418,8 @@
 #define I40IW_CQP_OP_QUERY_FPM_VALUES           0x20
 #define I40IW_CQP_OP_COMMIT_FPM_VALUES          0x21
 #define I40IW_CQP_OP_FLUSH_WQES                 0x22
+/* I40IW_CQP_OP_GEN_AE is the same value as I40IW_CQP_OP_FLUSH_WQES */
+#define I40IW_CQP_OP_GEN_AE                     0x22
 #define I40IW_CQP_OP_MANAGE_APBVT               0x23
 #define I40IW_CQP_OP_NOP                        0x24
 #define I40IW_CQP_OP_MANAGE_QUAD_HASH_TABLE_ENTRY 0x25
@@ -1729,6 +1731,7 @@ enum i40iw_alignment {
 #define OP_COMMIT_FPM_VALUES                    30
 #define OP_REQUESTED_COMMANDS                   31
 #define OP_COMPLETED_COMMANDS                   32
-#define OP_SIZE_CQP_STAT_ARRAY                  33
+#define OP_GEN_AE                               33
+#define OP_SIZE_CQP_STAT_ARRAY                  34
 
 #endif

drivers/infiniband/hw/i40iw/i40iw_hw.c

@@ -352,6 +352,8 @@ void i40iw_process_aeq(struct i40iw_device *iwdev)
 			else
 				i40iw_cm_disconn(iwqp);
 			break;
+		case I40IW_AE_BAD_CLOSE:
+			/* fall through */
 		case I40IW_AE_RESET_SENT:
 			i40iw_next_iw_state(iwqp, I40IW_QP_STATE_ERROR, 1, 0, 0);
 			i40iw_cm_disconn(iwqp);
@@ -667,6 +669,39 @@ enum i40iw_status_code i40iw_hw_flush_wqes(struct i40iw_device *iwdev,
 	return 0;
 }
 
+/**
+ * i40iw_gen_ae - generate AE
+ * @iwdev: iwarp device
+ * @qp: qp associated with AE
+ * @info: info for ae
+ * @wait: wait for completion
+ */
+void i40iw_gen_ae(struct i40iw_device *iwdev,
+		  struct i40iw_sc_qp *qp,
+		  struct i40iw_gen_ae_info *info,
+		  bool wait)
+{
+	struct i40iw_gen_ae_info *ae_info;
+	struct i40iw_cqp_request *cqp_request;
+	struct cqp_commands_info *cqp_info;
+
+	cqp_request = i40iw_get_cqp_request(&iwdev->cqp, wait);
+	if (!cqp_request)
+		return;
+
+	cqp_info = &cqp_request->info;
+	ae_info = &cqp_request->info.in.u.gen_ae.info;
+	memcpy(ae_info, info, sizeof(*ae_info));
+
+	cqp_info->cqp_cmd = OP_GEN_AE;
+	cqp_info->post_sq = 1;
+	cqp_info->in.u.gen_ae.qp = qp;
+	cqp_info->in.u.gen_ae.scratch = (uintptr_t)cqp_request;
+	if (i40iw_handle_cqp_op(iwdev, cqp_request))
+		i40iw_pr_err("CQP OP failed attempting to generate ae_code=0x%x\n",
+			     info->ae_code);
+}
+
 /**
  * i40iw_hw_manage_vf_pble_bp - manage vf pbles
  * @iwdev: iwarp device

drivers/infiniband/hw/i40iw/i40iw_main.c

@@ -1560,8 +1560,6 @@ static enum i40iw_status_code i40iw_setup_init_state(struct i40iw_handler *hdl,
 	enum i40iw_status_code status;
 
 	memcpy(&hdl->ldev, ldev, sizeof(*ldev));
-	if (resource_profile == 1)
-		resource_profile = 2;
 
 	iwdev->mpa_version = mpa_version;
 	iwdev->resource_profile = (resource_profile < I40IW_HMC_PROFILE_EQUAL) ?

drivers/infiniband/hw/i40iw/i40iw_puda.c

@@ -348,8 +348,8 @@ enum i40iw_status_code i40iw_puda_poll_completion(struct i40iw_sc_dev *dev,
 		spin_lock_irqsave(&rsrc->bufpool_lock, flags);
 		rsrc->tx_wqe_avail_cnt++;
 		spin_unlock_irqrestore(&rsrc->bufpool_lock, flags);
-		if (!list_empty(&rsrc->vsi->ilq->txpend))
-			i40iw_puda_send_buf(rsrc->vsi->ilq, NULL);
+		if (!list_empty(&rsrc->txpend))
+			i40iw_puda_send_buf(rsrc, NULL);
 	}
 
 done:
@@ -1471,10 +1471,6 @@ static void i40iw_ieq_tx_compl(struct i40iw_sc_vsi *vsi, void *sqwrid)
 	struct i40iw_puda_buf *buf = (struct i40iw_puda_buf *)sqwrid;
 
 	i40iw_puda_ret_bufpool(ieq, buf);
-	if (!list_empty(&ieq->txpend)) {
-		buf = i40iw_puda_get_listbuf(&ieq->txpend);
-		i40iw_puda_send_buf(ieq, buf);
-	}
 }
 
 /**

drivers/infiniband/hw/i40iw/i40iw_type.h

@@ -1004,6 +1004,11 @@ struct i40iw_cqp_query_fpm_values {
 	u32 pbl_max;
 };
 
+struct i40iw_gen_ae_info {
+	u16 ae_code;
+	u8 ae_source;
+};
+
 struct i40iw_cqp_ops {
 	enum i40iw_status_code (*cqp_init)(struct i40iw_sc_cqp *,
 					   struct i40iw_cqp_init_info *);
@@ -1290,6 +1295,12 @@ struct cqp_info {
 			u64 scratch;
 		} qp_flush_wqes;
 
+		struct {
+			struct i40iw_sc_qp *qp;
+			struct i40iw_gen_ae_info info;
+			u64 scratch;
+		} gen_ae;
+
 		struct {
 			struct i40iw_sc_cqp *cqp;
 			void *fpm_values_va;

drivers/infiniband/hw/i40iw/i40iw_utils.c

@@ -1284,15 +1284,13 @@ void i40iw_cqp_qp_destroy_cmd(struct i40iw_sc_dev *dev, struct i40iw_sc_qp *qp)
  */
 void i40iw_ieq_mpa_crc_ae(struct i40iw_sc_dev *dev, struct i40iw_sc_qp *qp)
 {
-	struct i40iw_qp_flush_info info;
+	struct i40iw_gen_ae_info info;
 	struct i40iw_device *iwdev = (struct i40iw_device *)dev->back_dev;
 
 	i40iw_debug(dev, I40IW_DEBUG_AEQ, "%s entered\n", __func__);
-	memset(&info, 0, sizeof(info));
 	info.ae_code = I40IW_AE_LLP_RECEIVED_MPA_CRC_ERROR;
-	info.generate_ae = true;
-	info.ae_source = 0x3;
-	(void)i40iw_hw_flush_wqes(iwdev, qp, &info, false);
+	info.ae_source = I40IW_AE_SOURCE_RQ;
+	i40iw_gen_ae(iwdev, qp, &info, false);
 }
 
 /**
@@ -1407,7 +1405,7 @@ struct i40iw_sc_qp *i40iw_ieq_get_qp(struct i40iw_sc_dev *dev,
 	rem_port = ntohs(tcph->source);
 
 	cm_node = i40iw_find_node(&iwdev->cm_core, rem_port, rem_addr, loc_port,
-				  loc_addr, false);
+				  loc_addr, false, true);
 	if (!cm_node)
 		return NULL;
 	iwqp = cm_node->iwqp;

drivers/infiniband/hw/i40iw/i40iw_verbs.c

@@ -38,6 +38,7 @@
 #include <linux/highmem.h>
 #include <linux/time.h>
 #include <linux/hugetlb.h>
+#include <linux/irq.h>
 #include <asm/byteorder.h>
 #include <net/ip.h>
 #include <rdma/ib_verbs.h>
@@ -830,10 +831,10 @@ static int i40iw_query_qp(struct ib_qp *ibqp,
 void i40iw_hw_modify_qp(struct i40iw_device *iwdev, struct i40iw_qp *iwqp,
 			struct i40iw_modify_qp_info *info, bool wait)
 {
-	enum i40iw_status_code status;
 	struct i40iw_cqp_request *cqp_request;
 	struct cqp_commands_info *cqp_info;
 	struct i40iw_modify_qp_info *m_info;
+	struct i40iw_gen_ae_info ae_info;
 
 	cqp_request = i40iw_get_cqp_request(&iwdev->cqp, wait);
 	if (!cqp_request)
@@ -846,9 +847,25 @@ void i40iw_hw_modify_qp(struct i40iw_device *iwdev, struct i40iw_qp *iwqp,
 	cqp_info->post_sq = 1;
 	cqp_info->in.u.qp_modify.qp = &iwqp->sc_qp;
 	cqp_info->in.u.qp_modify.scratch = (uintptr_t)cqp_request;
-	status = i40iw_handle_cqp_op(iwdev, cqp_request);
-	if (status)
-		i40iw_pr_err("CQP-OP Modify QP fail");
+	if (!i40iw_handle_cqp_op(iwdev, cqp_request))
+		return;
+
+	switch (m_info->next_iwarp_state) {
+	case I40IW_QP_STATE_RTS:
+		if (iwqp->iwarp_state == I40IW_QP_STATE_IDLE)
+			i40iw_send_reset(iwqp->cm_node);
+		/* fall through */
+	case I40IW_QP_STATE_IDLE:
+	case I40IW_QP_STATE_TERMINATE:
+	case I40IW_QP_STATE_CLOSING:
+		ae_info.ae_code = I40IW_AE_BAD_CLOSE;
+		ae_info.ae_source = 0;
+		i40iw_gen_ae(iwdev, &iwqp->sc_qp, &ae_info, false);
+		break;
+	case I40IW_QP_STATE_ERROR:
+	default:
+		break;
+	}
 }
 
 /**
@@ -961,10 +978,6 @@ int i40iw_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
 
 		iwqp->ibqp_state = attr->qp_state;
 
-		if (issue_modify_qp)
-			iwqp->iwarp_state = info.next_iwarp_state;
-		else
-			info.next_iwarp_state = iwqp->iwarp_state;
 	}
 	if (attr_mask & IB_QP_ACCESS_FLAGS) {
 		ctx_info->iwarp_info_valid = true;
@@ -1002,9 +1015,14 @@ int i40iw_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
 
 	spin_unlock_irqrestore(&iwqp->lock, flags);
 
-	if (issue_modify_qp)
+	if (issue_modify_qp) {
 		i40iw_hw_modify_qp(iwdev, iwqp, &info, true);
 
+		spin_lock_irqsave(&iwqp->lock, flags);
+		iwqp->iwarp_state = info.next_iwarp_state;
+		spin_unlock_irqrestore(&iwqp->lock, flags);
+	}
+
 	if (issue_modify_qp && (iwqp->ibqp_state > IB_QPS_RTS)) {
 		if (dont_wait) {
 			if (iwqp->cm_id && iwqp->hw_tcp_state) {
@@ -2728,6 +2746,25 @@ static int i40iw_destroy_ah(struct ib_ah *ah)
 	return -ENOSYS;
 }
 
+/**
+ * i40iw_get_vector_affinity - report IRQ affinity mask
+ * @ibdev: IB device
+ * @comp_vector: completion vector index
+ */
+static const struct cpumask *i40iw_get_vector_affinity(struct ib_device *ibdev,
+						       int comp_vector)
+{
+	struct i40iw_device *iwdev = to_iwdev(ibdev);
+	struct i40iw_msix_vector *msix_vec;
+
+	if (iwdev->msix_shared)
+		msix_vec = &iwdev->iw_msixtbl[comp_vector];
+	else
+		msix_vec = &iwdev->iw_msixtbl[comp_vector + 1];
+
+	return irq_get_affinity_mask(msix_vec->irq);
+}
+
 /**
  * i40iw_init_rdma_device - initialization of iwarp device
  * @iwdev: iwarp device
@@ -2824,6 +2861,7 @@ static struct i40iw_ib_device *i40iw_init_rdma_device(struct i40iw_device *iwdev
 	iwibdev->ibdev.req_notify_cq = i40iw_req_notify_cq;
 	iwibdev->ibdev.post_send = i40iw_post_send;
 	iwibdev->ibdev.post_recv = i40iw_post_recv;
+	iwibdev->ibdev.get_vector_affinity = i40iw_get_vector_affinity;
 
 	return iwibdev;
 }
@@ -2889,6 +2927,7 @@ int i40iw_register_rdma_device(struct i40iw_device *iwdev)
 		return -ENOMEM;
 	iwibdev = iwdev->iwibdev;
 
+	iwibdev->ibdev.driver_id = RDMA_DRIVER_I40IW;
 	ret = ib_register_device(&iwibdev->ibdev, NULL);
 	if (ret)
 		goto error;

drivers/infiniband/hw/mlx4/ah.c

@@ -101,12 +101,10 @@ static struct ib_ah *create_iboe_ah(struct ib_pd *pd,
 	if (ret)
 		return ERR_PTR(ret);
 	eth_zero_addr(ah->av.eth.s_mac);
-	if (gid_attr.ndev) {
-		if (is_vlan_dev(gid_attr.ndev))
-			vlan_tag = vlan_dev_vlan_id(gid_attr.ndev);
-		memcpy(ah->av.eth.s_mac, gid_attr.ndev->dev_addr, ETH_ALEN);
-		dev_put(gid_attr.ndev);
-	}
+	if (is_vlan_dev(gid_attr.ndev))
+		vlan_tag = vlan_dev_vlan_id(gid_attr.ndev);
+	memcpy(ah->av.eth.s_mac, gid_attr.ndev->dev_addr, ETH_ALEN);
+	dev_put(gid_attr.ndev);
 	if (vlan_tag < 0x1000)
 		vlan_tag |= (rdma_ah_get_sl(ah_attr) & 7) << 13;
 	ah->av.eth.port_pd = cpu_to_be32(to_mpd(pd)->pdn |

drivers/infiniband/hw/mlx4/main.c

@@ -246,14 +246,11 @@ static int mlx4_ib_update_gids(struct gid_entry *gids,
 	return mlx4_ib_update_gids_v1(gids, ibdev, port_num);
 }
 
-static int mlx4_ib_add_gid(struct ib_device *device,
-			   u8 port_num,
-			   unsigned int index,
-			   const union ib_gid *gid,
+static int mlx4_ib_add_gid(const union ib_gid *gid,
 			   const struct ib_gid_attr *attr,
 			   void **context)
 {
-	struct mlx4_ib_dev *ibdev = to_mdev(device);
+	struct mlx4_ib_dev *ibdev = to_mdev(attr->device);
 	struct mlx4_ib_iboe *iboe = &ibdev->iboe;
 	struct mlx4_port_gid_table   *port_gid_table;
 	int free = -1, found = -1;
@@ -262,16 +259,16 @@ static int mlx4_ib_add_gid(struct ib_device *device,
 	int i;
 	struct gid_entry *gids = NULL;
 
-	if (!rdma_cap_roce_gid_table(device, port_num))
+	if (!rdma_cap_roce_gid_table(attr->device, attr->port_num))
 		return -EINVAL;
 
-	if (port_num > MLX4_MAX_PORTS)
+	if (attr->port_num > MLX4_MAX_PORTS)
 		return -EINVAL;
 
 	if (!context)
 		return -EINVAL;
 
-	port_gid_table = &iboe->gids[port_num - 1];
+	port_gid_table = &iboe->gids[attr->port_num - 1];
 	spin_lock_bh(&iboe->lock);
 	for (i = 0; i < MLX4_MAX_PORT_GIDS; ++i) {
 		if (!memcmp(&port_gid_table->gids[i].gid, gid, sizeof(*gid)) &&
@@ -318,33 +315,30 @@ static int mlx4_ib_add_gid(struct ib_device *device,
 	spin_unlock_bh(&iboe->lock);
 
 	if (!ret && hw_update) {
-		ret = mlx4_ib_update_gids(gids, ibdev, port_num);
+		ret = mlx4_ib_update_gids(gids, ibdev, attr->port_num);
 		kfree(gids);
 	}
 
 	return ret;
 }
 
-static int mlx4_ib_del_gid(struct ib_device *device,
-			   u8 port_num,
-			   unsigned int index,
-			   void **context)
+static int mlx4_ib_del_gid(const struct ib_gid_attr *attr, void **context)
 {
 	struct gid_cache_context *ctx = *context;
-	struct mlx4_ib_dev *ibdev = to_mdev(device);
+	struct mlx4_ib_dev *ibdev = to_mdev(attr->device);
 	struct mlx4_ib_iboe *iboe = &ibdev->iboe;
 	struct mlx4_port_gid_table   *port_gid_table;
 	int ret = 0;
 	int hw_update = 0;
 	struct gid_entry *gids = NULL;
 
-	if (!rdma_cap_roce_gid_table(device, port_num))
+	if (!rdma_cap_roce_gid_table(attr->device, attr->port_num))
 		return -EINVAL;
 
-	if (port_num > MLX4_MAX_PORTS)
+	if (attr->port_num > MLX4_MAX_PORTS)
 		return -EINVAL;
 
-	port_gid_table = &iboe->gids[port_num - 1];
+	port_gid_table = &iboe->gids[attr->port_num - 1];
 	spin_lock_bh(&iboe->lock);
 	if (ctx) {
 		ctx->refcount--;
@@ -376,7 +370,7 @@ static int mlx4_ib_del_gid(struct ib_device *device,
 	spin_unlock_bh(&iboe->lock);
 
 	if (!ret && hw_update) {
-		ret = mlx4_ib_update_gids(gids, ibdev, port_num);
+		ret = mlx4_ib_update_gids(gids, ibdev, attr->port_num);
 		kfree(gids);
 	}
 	return ret;
@@ -411,9 +405,6 @@ int mlx4_ib_gid_index_to_real_index(struct mlx4_ib_dev *ibdev,
 	if (attr.ndev)
 		dev_put(attr.ndev);
 
-	if (!memcmp(&gid, &zgid, sizeof(gid)))
-		return -EINVAL;
-
 	spin_lock_irqsave(&iboe->lock, flags);
 	port_gid_table = &iboe->gids[port_num - 1];
 
@@ -429,6 +420,9 @@ int mlx4_ib_gid_index_to_real_index(struct mlx4_ib_dev *ibdev,
 	return real_index;
 }
 
+#define field_avail(type, fld, sz) (offsetof(type, fld) + \
+				    sizeof(((type *)0)->fld) <= (sz))
+
 static int mlx4_ib_query_device(struct ib_device *ibdev,
 				struct ib_device_attr *props,
 				struct ib_udata *uhw)
@@ -556,14 +550,19 @@ static int mlx4_ib_query_device(struct ib_device *ibdev,
 	props->timestamp_mask = 0xFFFFFFFFFFFFULL;
 	props->max_ah = INT_MAX;
 
-	if ((dev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_RSS) &&
-	    (mlx4_ib_port_link_layer(ibdev, 1) == IB_LINK_LAYER_ETHERNET ||
-	     mlx4_ib_port_link_layer(ibdev, 2) == IB_LINK_LAYER_ETHERNET)) {
-		props->rss_caps.max_rwq_indirection_tables = props->max_qp;
-		props->rss_caps.max_rwq_indirection_table_size =
-			dev->dev->caps.max_rss_tbl_sz;
-		props->rss_caps.supported_qpts = 1 << IB_QPT_RAW_PACKET;
-		props->max_wq_type_rq = props->max_qp;
+	if (mlx4_ib_port_link_layer(ibdev, 1) == IB_LINK_LAYER_ETHERNET ||
+	    mlx4_ib_port_link_layer(ibdev, 2) == IB_LINK_LAYER_ETHERNET) {
+		if (dev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_RSS) {
+			props->rss_caps.max_rwq_indirection_tables =
+				props->max_qp;
+			props->rss_caps.max_rwq_indirection_table_size =
+				dev->dev->caps.max_rss_tbl_sz;
+			props->rss_caps.supported_qpts = 1 << IB_QPT_RAW_PACKET;
+			props->max_wq_type_rq = props->max_qp;
+		}
+
+		if (dev->dev->caps.flags & MLX4_DEV_CAP_FLAG_FCS_KEEP)
+			props->raw_packet_caps |= IB_RAW_PACKET_CAP_SCATTER_FCS;
 	}
 
 	props->cq_caps.max_cq_moderation_count = MLX4_MAX_CQ_COUNT;
@@ -575,7 +574,7 @@ static int mlx4_ib_query_device(struct ib_device *ibdev,
 	if (uhw->outlen >= resp.response_length + sizeof(resp.hca_core_clock_offset)) {
 		resp.response_length += sizeof(resp.hca_core_clock_offset);
 		if (!err && !mlx4_is_slave(dev->dev)) {
-			resp.comp_mask |= QUERY_DEVICE_RESP_MASK_TIMESTAMP;
+			resp.comp_mask |= MLX4_IB_QUERY_DEV_RESP_MASK_CORE_CLOCK_OFFSET;
 			resp.hca_core_clock_offset = clock_params.offset % PAGE_SIZE;
 		}
 	}
@@ -587,8 +586,7 @@ static int mlx4_ib_query_device(struct ib_device *ibdev,
 			sizeof(struct mlx4_wqe_data_seg);
 	}
 
-	if (uhw->outlen >= resp.response_length + sizeof(resp.rss_caps)) {
-		resp.response_length += sizeof(resp.rss_caps);
+	if (field_avail(typeof(resp), rss_caps, uhw->outlen)) {
 		if (props->rss_caps.supported_qpts) {
 			resp.rss_caps.rx_hash_function =
 				MLX4_IB_RX_HASH_FUNC_TOEPLITZ;
@@ -608,6 +606,22 @@ static int mlx4_ib_query_device(struct ib_device *ibdev,
 				resp.rss_caps.rx_hash_fields_mask |=
 					MLX4_IB_RX_HASH_INNER;
 		}
+		resp.response_length = offsetof(typeof(resp), rss_caps) +
+				       sizeof(resp.rss_caps);
+	}
+
+	if (field_avail(typeof(resp), tso_caps, uhw->outlen)) {
+		if (dev->dev->caps.max_gso_sz &&
+		    ((mlx4_ib_port_link_layer(ibdev, 1) ==
+		    IB_LINK_LAYER_ETHERNET) ||
+		    (mlx4_ib_port_link_layer(ibdev, 2) ==
+		    IB_LINK_LAYER_ETHERNET))) {
+			resp.tso_caps.max_tso = dev->dev->caps.max_gso_sz;
+			resp.tso_caps.supported_qpts |=
+				1 << IB_QPT_RAW_PACKET;
+		}
+		resp.response_length = offsetof(typeof(resp), tso_caps) +
+				       sizeof(resp.tso_caps);
 	}
 
 	if (uhw->outlen) {
@@ -865,24 +879,9 @@ out:
 static int mlx4_ib_query_gid(struct ib_device *ibdev, u8 port, int index,
 			     union ib_gid *gid)
 {
-	int ret;
-
 	if (rdma_protocol_ib(ibdev, port))
 		return __mlx4_ib_query_gid(ibdev, port, index, gid, 0);
-
-	if (!rdma_protocol_roce(ibdev, port))
-		return -ENODEV;
-
-	if (!rdma_cap_roce_gid_table(ibdev, port))
-		return -ENODEV;
-
-	ret = ib_get_cached_gid(ibdev, port, index, gid, NULL);
-	if (ret == -EAGAIN) {
-		memcpy(gid, &zgid, sizeof(*gid));
-		return 0;
-	}
-
-	return ret;
+	return 0;
 }
 
 static int mlx4_ib_query_sl2vl(struct ib_device *ibdev, u8 port, u64 *sl2vl_tbl)
@@ -1330,7 +1329,7 @@ static struct ib_pd *mlx4_ib_alloc_pd(struct ib_device *ibdev,
 	struct mlx4_ib_pd *pd;
 	int err;
 
-	pd = kmalloc(sizeof *pd, GFP_KERNEL);
+	pd = kzalloc(sizeof(*pd), GFP_KERNEL);
 	if (!pd)
 		return ERR_PTR(-ENOMEM);
 
@@ -1346,7 +1345,6 @@ static struct ib_pd *mlx4_ib_alloc_pd(struct ib_device *ibdev,
 			kfree(pd);
 			return ERR_PTR(-EFAULT);
 		}
-
 	return &pd->ibpd;
 }
 
@@ -1860,6 +1858,9 @@ static struct ib_flow *mlx4_ib_create_flow(struct ib_qp *qp,
 	if (flow_attr->port < 1 || flow_attr->port > qp->device->phys_port_cnt)
 		return ERR_PTR(-EINVAL);
 
+	if (flow_attr->flags & ~IB_FLOW_ATTR_FLAGS_DONT_TRAP)
+		return ERR_PTR(-EOPNOTSUPP);
+
 	if ((flow_attr->flags & IB_FLOW_ATTR_FLAGS_DONT_TRAP) &&
 	    (flow_attr->type != IB_FLOW_ATTR_NORMAL))
 		return ERR_PTR(-EOPNOTSUPP);
@@ -2933,6 +2934,7 @@ static void *mlx4_ib_add(struct mlx4_dev *dev)
 	if (mlx4_ib_alloc_diag_counters(ibdev))
 		goto err_steer_free_bitmap;
 
+	ibdev->ib_dev.driver_id = RDMA_DRIVER_MLX4;
 	if (ib_register_device(&ibdev->ib_dev, NULL))
 		goto err_diag_counters;
 

drivers/infiniband/hw/mlx4/mlx4_ib.h

@@ -189,6 +189,7 @@ enum mlx4_ib_qp_flags {
 	MLX4_IB_QP_LSO = IB_QP_CREATE_IPOIB_UD_LSO,
 	MLX4_IB_QP_BLOCK_MULTICAST_LOOPBACK = IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK,
 	MLX4_IB_QP_NETIF = IB_QP_CREATE_NETIF_QP,
+	MLX4_IB_QP_SCATTER_FCS = IB_QP_CREATE_SCATTER_FCS,
 
 	/* Mellanox specific flags start from IB_QP_CREATE_RESERVED_START */
 	MLX4_IB_ROCE_V2_GSI_QP = MLX4_IB_QP_CREATE_ROCE_V2_GSI,
@@ -641,24 +642,6 @@ struct mlx4_uverbs_ex_query_device {
 	__u32 reserved;
 };
 
-enum query_device_resp_mask {
-	QUERY_DEVICE_RESP_MASK_TIMESTAMP = 1UL << 0,
-};
-
-struct mlx4_ib_rss_caps {
-	__u64 rx_hash_fields_mask; /* enum mlx4_rx_hash_fields */
-	__u8 rx_hash_function; /* enum mlx4_rx_hash_function_flags */
-	__u8 reserved[7];
-};
-
-struct mlx4_uverbs_ex_query_device_resp {
-	__u32			comp_mask;
-	__u32			response_length;
-	__u64			hca_core_clock_offset;
-	__u32			max_inl_recv_sz;
-	struct mlx4_ib_rss_caps	rss_caps;
-};
-
 static inline struct mlx4_ib_dev *to_mdev(struct ib_device *ibdev)
 {
 	return container_of(ibdev, struct mlx4_ib_dev, ib_dev);

drivers/infiniband/hw/mlx4/mr.c

@@ -407,6 +407,9 @@ struct ib_mr *mlx4_ib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
 		goto err_mr;
 
 	mr->ibmr.rkey = mr->ibmr.lkey = mr->mmr.key;
+	mr->ibmr.length = length;
+	mr->ibmr.iova = virt_addr;
+	mr->ibmr.page_size = 1U << shift;
 
 	return &mr->ibmr;
 

drivers/infiniband/hw/mlx4/qp.c

@@ -1096,6 +1096,17 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd,
 			qp->inl_recv_sz = ucmd.qp.inl_recv_sz;
 		}
 
+		if (init_attr->create_flags & IB_QP_CREATE_SCATTER_FCS) {
+			if (!(dev->dev->caps.flags &
+			      MLX4_DEV_CAP_FLAG_FCS_KEEP)) {
+				pr_debug("scatter FCS is unsupported\n");
+				err = -EOPNOTSUPP;
+				goto err;
+			}
+
+			qp->flags |= MLX4_IB_QP_SCATTER_FCS;
+		}
+
 		err = set_rq_size(dev, &init_attr->cap, !!pd->uobject,
 				  qp_has_rq(init_attr), qp, qp->inl_recv_sz);
 		if (err)
@@ -2234,6 +2245,9 @@ static int __mlx4_ib_modify_qp(void *src, enum mlx4_ib_source_type src_type,
 	if (qp->inl_recv_sz)
 		context->param3 |= cpu_to_be32(1 << 25);
 
+	if (qp->flags & MLX4_IB_QP_SCATTER_FCS)
+		context->param3 |= cpu_to_be32(1 << 29);
+
 	if (qp_type == IB_QPT_GSI || qp_type == IB_QPT_SMI)
 		context->mtu_msgmax = (IB_MTU_4096 << 5) | 11;
 	else if (qp_type == IB_QPT_RAW_PACKET)
@@ -2356,9 +2370,7 @@ static int __mlx4_ib_modify_qp(void *src, enum mlx4_ib_source_type src_type,
 
 			status = ib_get_cached_gid(&dev->ib_dev, port_num,
 						   index, &gid, &gid_attr);
-			if (!status && !memcmp(&gid, &zgid, sizeof(gid)))
-				status = -ENOENT;
-			if (!status && gid_attr.ndev) {
+			if (!status) {
 				vlan = rdma_vlan_dev_vlan_id(gid_attr.ndev);
 				memcpy(smac, gid_attr.ndev->dev_addr, ETH_ALEN);
 				dev_put(gid_attr.ndev);
@@ -3880,8 +3892,8 @@ out:
 		 */
 		wmb();
 
-		writel(qp->doorbell_qpn,
-		       to_mdev(ibqp->device)->uar_map + MLX4_SEND_DOORBELL);
+		writel_relaxed(qp->doorbell_qpn,
+			to_mdev(ibqp->device)->uar_map + MLX4_SEND_DOORBELL);
 
 		/*
 		 * Make sure doorbells don't leak out of SQ spinlock
@@ -4204,7 +4216,7 @@ struct ib_wq *mlx4_ib_create_wq(struct ib_pd *pd,
 		return ERR_PTR(-EOPNOTSUPP);
 	}
 
-	if (init_attr->create_flags) {
+	if (init_attr->create_flags & ~IB_WQ_FLAGS_SCATTER_FCS) {
 		pr_debug("unsupported create_flags %u\n",
 			 init_attr->create_flags);
 		return ERR_PTR(-EOPNOTSUPP);
@@ -4225,6 +4237,9 @@ struct ib_wq *mlx4_ib_create_wq(struct ib_pd *pd,
 	ib_qp_init_attr.recv_cq = init_attr->cq;
 	ib_qp_init_attr.send_cq = ib_qp_init_attr.recv_cq; /* Dummy CQ */
 
+	if (init_attr->create_flags & IB_WQ_FLAGS_SCATTER_FCS)
+		ib_qp_init_attr.create_flags |= IB_QP_CREATE_SCATTER_FCS;
+
 	err = create_qp_common(dev, pd, MLX4_IB_RWQ_SRC, &ib_qp_init_attr,
 			       udata, 0, &qp);
 	if (err) {

drivers/infiniband/hw/mlx5/ah.c

@@ -36,6 +36,9 @@ static struct ib_ah *create_ib_ah(struct mlx5_ib_dev *dev,
 				  struct mlx5_ib_ah *ah,
 				  struct rdma_ah_attr *ah_attr)
 {
+	enum ib_gid_type gid_type;
+	int err;
+
 	if (rdma_ah_get_ah_flags(ah_attr) & IB_AH_GRH) {
 		const struct ib_global_route *grh = rdma_ah_read_grh(ah_attr);
 
@@ -50,6 +53,12 @@ static struct ib_ah *create_ib_ah(struct mlx5_ib_dev *dev,
 	ah->av.stat_rate_sl = (rdma_ah_get_static_rate(ah_attr) << 4);
 
 	if (ah_attr->type == RDMA_AH_ATTR_TYPE_ROCE) {
+		err = mlx5_get_roce_gid_type(dev, ah_attr->port_num,
+					     ah_attr->grh.sgid_index,
+					     &gid_type);
+		if (err)
+			return ERR_PTR(err);
+
 		memcpy(ah->av.rmac, ah_attr->roce.dmac,
 		       sizeof(ah_attr->roce.dmac));
 		ah->av.udp_sport =
@@ -57,6 +66,9 @@ static struct ib_ah *create_ib_ah(struct mlx5_ib_dev *dev,
 					rdma_ah_get_port_num(ah_attr),
 					rdma_ah_read_grh(ah_attr)->sgid_index);
 		ah->av.stat_rate_sl |= (rdma_ah_get_sl(ah_attr) & 0x7) << 1;
+		if (gid_type == IB_GID_TYPE_ROCE_UDP_ENCAP)
+#define MLX5_ECN_ENABLED BIT(1)
+			ah->av.tclass |= MLX5_ECN_ENABLED;
 	} else {
 		ah->av.rlid = cpu_to_be16(rdma_ah_get_dlid(ah_attr));
 		ah->av.fl_mlid = rdma_ah_get_path_bits(ah_attr) & 0x7f;

drivers/infiniband/hw/mlx5/cmd.c

@@ -66,3 +66,107 @@ int mlx5_cmd_modify_cong_params(struct mlx5_core_dev *dev,
 
 	return mlx5_cmd_exec(dev, in, in_size, out, sizeof(out));
 }
+
+int mlx5_cmd_alloc_memic(struct mlx5_memic *memic, phys_addr_t *addr,
+			  u64 length, u32 alignment)
+{
+	struct mlx5_core_dev *dev = memic->dev;
+	u64 num_memic_hw_pages = MLX5_CAP_DEV_MEM(dev, memic_bar_size)
+					>> PAGE_SHIFT;
+	u64 hw_start_addr = MLX5_CAP64_DEV_MEM(dev, memic_bar_start_addr);
+	u32 max_alignment = MLX5_CAP_DEV_MEM(dev, log_max_memic_addr_alignment);
+	u32 num_pages = DIV_ROUND_UP(length, PAGE_SIZE);
+	u32 out[MLX5_ST_SZ_DW(alloc_memic_out)] = {};
+	u32 in[MLX5_ST_SZ_DW(alloc_memic_in)] = {};
+	u32 mlx5_alignment;
+	u64 page_idx = 0;
+	int ret = 0;
+
+	if (!length || (length & MLX5_MEMIC_ALLOC_SIZE_MASK))
+		return -EINVAL;
+
+	/* mlx5 device sets alignment as 64*2^driver_value
+	 * so normalizing is needed.
+	 */
+	mlx5_alignment = (alignment < MLX5_MEMIC_BASE_ALIGN) ? 0 :
+			 alignment - MLX5_MEMIC_BASE_ALIGN;
+	if (mlx5_alignment > max_alignment)
+		return -EINVAL;
+
+	MLX5_SET(alloc_memic_in, in, opcode, MLX5_CMD_OP_ALLOC_MEMIC);
+	MLX5_SET(alloc_memic_in, in, range_size, num_pages * PAGE_SIZE);
+	MLX5_SET(alloc_memic_in, in, memic_size, length);
+	MLX5_SET(alloc_memic_in, in, log_memic_addr_alignment,
+		 mlx5_alignment);
+
+	while (page_idx < num_memic_hw_pages) {
+		spin_lock(&memic->memic_lock);
+		page_idx = bitmap_find_next_zero_area(memic->memic_alloc_pages,
+						      num_memic_hw_pages,
+						      page_idx,
+						      num_pages, 0);
+
+		if (page_idx < num_memic_hw_pages)
+			bitmap_set(memic->memic_alloc_pages,
+				   page_idx, num_pages);
+
+		spin_unlock(&memic->memic_lock);
+
+		if (page_idx >= num_memic_hw_pages)
+			break;
+
+		MLX5_SET64(alloc_memic_in, in, range_start_addr,
+			   hw_start_addr + (page_idx * PAGE_SIZE));
+
+		ret = mlx5_cmd_exec(dev, in, sizeof(in), out, sizeof(out));
+		if (ret) {
+			spin_lock(&memic->memic_lock);
+			bitmap_clear(memic->memic_alloc_pages,
+				     page_idx, num_pages);
+			spin_unlock(&memic->memic_lock);
+
+			if (ret == -EAGAIN) {
+				page_idx++;
+				continue;
+			}
+
+			return ret;
+		}
+
+		*addr = pci_resource_start(dev->pdev, 0) +
+			MLX5_GET64(alloc_memic_out, out, memic_start_addr);
+
+		return 0;
+	}
+
+	return -ENOMEM;
+}
+
+int mlx5_cmd_dealloc_memic(struct mlx5_memic *memic, u64 addr, u64 length)
+{
+	struct mlx5_core_dev *dev = memic->dev;
+	u64 hw_start_addr = MLX5_CAP64_DEV_MEM(dev, memic_bar_start_addr);
+	u32 num_pages = DIV_ROUND_UP(length, PAGE_SIZE);
+	u32 out[MLX5_ST_SZ_DW(dealloc_memic_out)] = {0};
+	u32 in[MLX5_ST_SZ_DW(dealloc_memic_in)] = {0};
+	u64 start_page_idx;
+	int err;
+
+	addr -= pci_resource_start(dev->pdev, 0);
+	start_page_idx = (addr - hw_start_addr) >> PAGE_SHIFT;
+
+	MLX5_SET(dealloc_memic_in, in, opcode, MLX5_CMD_OP_DEALLOC_MEMIC);
+	MLX5_SET64(dealloc_memic_in, in, memic_start_addr, addr);
+	MLX5_SET(dealloc_memic_in, in, memic_size, length);
+
+	err =  mlx5_cmd_exec(dev, in, sizeof(in), out, sizeof(out));
+
+	if (!err) {
+		spin_lock(&memic->memic_lock);
+		bitmap_clear(memic->memic_alloc_pages,
+			     start_page_idx, num_pages);
+		spin_unlock(&memic->memic_lock);
+	}
+
+	return err;
+}

drivers/infiniband/hw/mlx5/cmd.h

@@ -33,6 +33,7 @@
 #ifndef MLX5_IB_CMD_H
 #define MLX5_IB_CMD_H
 
+#include "mlx5_ib.h"
 #include <linux/kernel.h>
 #include <linux/mlx5/driver.h>
 
@@ -41,4 +42,7 @@ int mlx5_cmd_query_cong_params(struct mlx5_core_dev *dev, int cong_point,
 			       void *out, int out_size);
 int mlx5_cmd_modify_cong_params(struct mlx5_core_dev *mdev,
 				void *in, int in_size);
+int mlx5_cmd_alloc_memic(struct mlx5_memic *memic, phys_addr_t *addr,
+			 u64 length, u32 alignment);
+int mlx5_cmd_dealloc_memic(struct mlx5_memic *memic, u64 addr, u64 length);
 #endif /* MLX5_IB_CMD_H */

drivers/infiniband/hw/mlx5/main.c

@@ -38,6 +38,7 @@
 #include <linux/pci.h>
 #include <linux/dma-mapping.h>
 #include <linux/slab.h>
+#include <linux/bitmap.h>
 #if defined(CONFIG_X86)
 #include <asm/pat.h>
 #endif
@@ -51,6 +52,7 @@
 #include <linux/mlx5/port.h>
 #include <linux/mlx5/vport.h>
 #include <linux/mlx5/fs.h>
+#include <linux/mlx5/fs_helpers.h>
 #include <linux/list.h>
 #include <rdma/ib_smi.h>
 #include <rdma/ib_umem.h>
@@ -60,6 +62,13 @@
 #include "ib_rep.h"
 #include "cmd.h"
 #include <linux/mlx5/fs_helpers.h>
+#include <linux/mlx5/accel.h>
+#include <rdma/uverbs_std_types.h>
+#include <rdma/mlx5_user_ioctl_verbs.h>
+#include <rdma/mlx5_user_ioctl_cmds.h>
+
+#define UVERBS_MODULE_NAME mlx5_ib
+#include <rdma/uverbs_named_ioctl.h>
 
 #define DRIVER_NAME "mlx5_ib"
 #define DRIVER_VERSION "5.0-0"
@@ -92,6 +101,12 @@ static LIST_HEAD(mlx5_ib_dev_list);
  */
 static DEFINE_MUTEX(mlx5_ib_multiport_mutex);
 
+/* We can't use an array for xlt_emergency_page because dma_map_single
+ * doesn't work on kernel modules memory
+ */
+static unsigned long xlt_emergency_page;
+static struct mutex xlt_emergency_page_mutex;
+
 struct mlx5_ib_dev *mlx5_ib_get_ibdev_from_mpi(struct mlx5_ib_multiport_info *mpi)
 {
 	struct mlx5_ib_dev *dev;
@@ -399,6 +414,9 @@ static int mlx5_query_port_roce(struct ib_device *device, u8 port_num,
 	if (err)
 		goto out;
 
+	props->active_width     = IB_WIDTH_4X;
+	props->active_speed     = IB_SPEED_QDR;
+
 	translate_eth_proto_oper(eth_prot_oper, &props->active_speed,
 				 &props->active_width);
 
@@ -493,18 +511,19 @@ static int set_roce_addr(struct mlx5_ib_dev *dev, u8 port_num,
 				      vlan_id, port_num);
 }
 
-static int mlx5_ib_add_gid(struct ib_device *device, u8 port_num,
-			   unsigned int index, const union ib_gid *gid,
+static int mlx5_ib_add_gid(const union ib_gid *gid,
 			   const struct ib_gid_attr *attr,
 			   __always_unused void **context)
 {
-	return set_roce_addr(to_mdev(device), port_num, index, gid, attr);
+	return set_roce_addr(to_mdev(attr->device), attr->port_num,
+			     attr->index, gid, attr);
 }
 
-static int mlx5_ib_del_gid(struct ib_device *device, u8 port_num,
-			   unsigned int index, __always_unused void **context)
+static int mlx5_ib_del_gid(const struct ib_gid_attr *attr,
+			   __always_unused void **context)
 {
-	return set_roce_addr(to_mdev(device), port_num, index, NULL, NULL);
+	return set_roce_addr(to_mdev(attr->device), attr->port_num,
+			     attr->index, NULL, NULL);
 }
 
 __be16 mlx5_get_roce_udp_sport(struct mlx5_ib_dev *dev, u8 port_num,
@@ -516,9 +535,6 @@ __be16 mlx5_get_roce_udp_sport(struct mlx5_ib_dev *dev, u8 port_num,
 	if (ib_get_cached_gid(&dev->ib_dev, port_num, index, &gid, &attr))
 		return 0;
 
-	if (!attr.ndev)
-		return 0;
-
 	dev_put(attr.ndev);
 
 	if (attr.gid_type != IB_GID_TYPE_ROCE_UDP_ENCAP)
@@ -538,9 +554,6 @@ int mlx5_get_roce_gid_type(struct mlx5_ib_dev *dev, u8 port_num,
 	if (ret)
 		return ret;
 
-	if (!attr.ndev)
-		return -ENODEV;
-
 	dev_put(attr.ndev);
 
 	*gid_type = attr.gid_type;
@@ -844,6 +857,10 @@ static int mlx5_ib_query_device(struct ib_device *ibdev,
 						MLX5_RX_HASH_SRC_PORT_UDP |
 						MLX5_RX_HASH_DST_PORT_UDP |
 						MLX5_RX_HASH_INNER;
+			if (mlx5_accel_ipsec_device_caps(dev->mdev) &
+			    MLX5_ACCEL_IPSEC_CAP_DEVICE)
+				resp.rss_caps.rx_hash_fields_mask |=
+					MLX5_RX_HASH_IPSEC_SPI;
 			resp.response_length += sizeof(resp.rss_caps);
 		}
 	} else {
@@ -875,6 +892,11 @@ static int mlx5_ib_query_device(struct ib_device *ibdev,
 		props->raw_packet_caps |= IB_RAW_PACKET_CAP_SCATTER_FCS;
 	}
 
+	if (MLX5_CAP_DEV_MEM(mdev, memic)) {
+		props->max_dm_size =
+			MLX5_CAP_DEV_MEM(mdev, max_memic_size);
+	}
+
 	if (mlx5_get_flow_namespace(dev->mdev, MLX5_FLOW_NAMESPACE_BYPASS))
 		props->device_cap_flags |= IB_DEVICE_MANAGED_FLOW_STEERING;
 
@@ -980,6 +1002,10 @@ static int mlx5_ib_query_device(struct ib_device *ibdev,
 				MLX5_CAP_QOS(mdev, packet_pacing_min_rate);
 			resp.packet_pacing_caps.supported_qpts |=
 				1 << IB_QPT_RAW_PACKET;
+			if (MLX5_CAP_QOS(mdev, packet_pacing_burst_bound) &&
+			    MLX5_CAP_QOS(mdev, packet_pacing_typical_size))
+				resp.packet_pacing_caps.cap_flags |=
+					MLX5_IB_PP_SUPPORT_BURST;
 		}
 		resp.response_length += sizeof(resp.packet_pacing_caps);
 	}
@@ -1665,6 +1691,18 @@ static struct ib_ucontext *mlx5_ib_alloc_ucontext(struct ib_device *ibdev,
 	resp.response_length = min(offsetof(typeof(resp), response_length) +
 				   sizeof(resp.response_length), udata->outlen);
 
+	if (mlx5_accel_ipsec_device_caps(dev->mdev) & MLX5_ACCEL_IPSEC_CAP_DEVICE) {
+		if (mlx5_get_flow_namespace(dev->mdev, MLX5_FLOW_NAMESPACE_EGRESS))
+			resp.flow_action_flags |= MLX5_USER_ALLOC_UCONTEXT_FLOW_ACTION_FLAGS_ESP_AES_GCM;
+		if (mlx5_accel_ipsec_device_caps(dev->mdev) & MLX5_ACCEL_IPSEC_CAP_REQUIRED_METADATA)
+			resp.flow_action_flags |= MLX5_USER_ALLOC_UCONTEXT_FLOW_ACTION_FLAGS_ESP_AES_GCM_REQ_METADATA;
+		if (MLX5_CAP_FLOWTABLE(dev->mdev, flow_table_properties_nic_receive.ft_field_support.outer_esp_spi))
+			resp.flow_action_flags |= MLX5_USER_ALLOC_UCONTEXT_FLOW_ACTION_FLAGS_ESP_AES_GCM_SPI_STEERING;
+		if (mlx5_accel_ipsec_device_caps(dev->mdev) & MLX5_ACCEL_IPSEC_CAP_TX_IV_IS_ESN)
+			resp.flow_action_flags |= MLX5_USER_ALLOC_UCONTEXT_FLOW_ACTION_FLAGS_ESP_AES_GCM_TX_IV_IS_ESN;
+		/* MLX5_USER_ALLOC_UCONTEXT_FLOW_ACTION_FLAGS_ESP_AES_GCM_FULL_OFFLOAD is currently always 0 */
+	}
+
 	context = kzalloc(sizeof(*context), GFP_KERNEL);
 	if (!context)
 		return ERR_PTR(-ENOMEM);
@@ -1702,17 +1740,10 @@ static struct ib_ucontext *mlx5_ib_alloc_ucontext(struct ib_device *ibdev,
 	context->ibucontext.invalidate_range = &mlx5_ib_invalidate_range;
 #endif
 
-	context->upd_xlt_page = __get_free_page(GFP_KERNEL);
-	if (!context->upd_xlt_page) {
-		err = -ENOMEM;
-		goto out_uars;
-	}
-	mutex_init(&context->upd_xlt_page_mutex);
-
 	if (MLX5_CAP_GEN(dev->mdev, log_max_transport_domain)) {
 		err = mlx5_ib_alloc_transport_domain(dev, &context->tdn);
 		if (err)
-			goto out_page;
+			goto out_uars;
 	}
 
 	INIT_LIST_HEAD(&context->vma_private_list);
@@ -1789,9 +1820,6 @@ out_td:
 	if (MLX5_CAP_GEN(dev->mdev, log_max_transport_domain))
 		mlx5_ib_dealloc_transport_domain(dev, context->tdn);
 
-out_page:
-	free_page(context->upd_xlt_page);
-
 out_uars:
 	deallocate_uars(dev, context);
 
@@ -1817,7 +1845,6 @@ static int mlx5_ib_dealloc_ucontext(struct ib_ucontext *ibcontext)
 	if (MLX5_CAP_GEN(dev->mdev, log_max_transport_domain))
 		mlx5_ib_dealloc_transport_domain(dev, context->tdn);
 
-	free_page(context->upd_xlt_page);
 	deallocate_uars(dev, context);
 	kfree(bfregi->sys_pages);
 	kfree(bfregi->count);
@@ -1993,6 +2020,8 @@ static inline char *mmap_cmd2str(enum mlx5_ib_mmap_cmd cmd)
 		return "best effort WC";
 	case MLX5_IB_MMAP_NC_PAGE:
 		return "NC";
+	case MLX5_IB_MMAP_DEVICE_MEM:
+		return "Device Memory";
 	default:
 		return NULL;
 	}
@@ -2151,6 +2180,34 @@ free_bfreg:
 	return err;
 }
 
+static int dm_mmap(struct ib_ucontext *context, struct vm_area_struct *vma)
+{
+	struct mlx5_ib_ucontext *mctx = to_mucontext(context);
+	struct mlx5_ib_dev *dev = to_mdev(context->device);
+	u16 page_idx = get_extended_index(vma->vm_pgoff);
+	size_t map_size = vma->vm_end - vma->vm_start;
+	u32 npages = map_size >> PAGE_SHIFT;
+	phys_addr_t pfn;
+	pgprot_t prot;
+
+	if (find_next_zero_bit(mctx->dm_pages, page_idx + npages, page_idx) !=
+	    page_idx + npages)
+		return -EINVAL;
+
+	pfn = ((pci_resource_start(dev->mdev->pdev, 0) +
+	      MLX5_CAP64_DEV_MEM(dev->mdev, memic_bar_start_addr)) >>
+	      PAGE_SHIFT) +
+	      page_idx;
+	prot = pgprot_writecombine(vma->vm_page_prot);
+	vma->vm_page_prot = prot;
+
+	if (io_remap_pfn_range(vma, vma->vm_start, pfn, map_size,
+			       vma->vm_page_prot))
+		return -EAGAIN;
+
+	return mlx5_ib_set_vma_data(vma, mctx);
+}
+
 static int mlx5_ib_mmap(struct ib_ucontext *ibcontext, struct vm_area_struct *vma)
 {
 	struct mlx5_ib_ucontext *context = to_mucontext(ibcontext);
@@ -2195,6 +2252,9 @@ static int mlx5_ib_mmap(struct ib_ucontext *ibcontext, struct vm_area_struct *vm
 	case MLX5_IB_MMAP_CLOCK_INFO:
 		return mlx5_ib_mmap_clock_info_page(dev, vma, context);
 
+	case MLX5_IB_MMAP_DEVICE_MEM:
+		return dm_mmap(ibcontext, vma);
+
 	default:
 		return -EINVAL;
 	}
@@ -2202,6 +2262,87 @@ static int mlx5_ib_mmap(struct ib_ucontext *ibcontext, struct vm_area_struct *vm
 	return 0;
 }
 
+struct ib_dm *mlx5_ib_alloc_dm(struct ib_device *ibdev,
+			       struct ib_ucontext *context,
+			       struct ib_dm_alloc_attr *attr,
+			       struct uverbs_attr_bundle *attrs)
+{
+	u64 act_size = roundup(attr->length, MLX5_MEMIC_BASE_SIZE);
+	struct mlx5_memic *memic = &to_mdev(ibdev)->memic;
+	phys_addr_t memic_addr;
+	struct mlx5_ib_dm *dm;
+	u64 start_offset;
+	u32 page_idx;
+	int err;
+
+	dm = kzalloc(sizeof(*dm), GFP_KERNEL);
+	if (!dm)
+		return ERR_PTR(-ENOMEM);
+
+	mlx5_ib_dbg(to_mdev(ibdev), "alloc_memic req: user_length=0x%llx act_length=0x%llx log_alignment=%d\n",
+		    attr->length, act_size, attr->alignment);
+
+	err = mlx5_cmd_alloc_memic(memic, &memic_addr,
+				   act_size, attr->alignment);
+	if (err)
+		goto err_free;
+
+	start_offset = memic_addr & ~PAGE_MASK;
+	page_idx = (memic_addr - pci_resource_start(memic->dev->pdev, 0) -
+		    MLX5_CAP64_DEV_MEM(memic->dev, memic_bar_start_addr)) >>
+		    PAGE_SHIFT;
+
+	err = uverbs_copy_to(attrs,
+			     MLX5_IB_ATTR_ALLOC_DM_RESP_START_OFFSET,
+			     &start_offset, sizeof(start_offset));
+	if (err)
+		goto err_dealloc;
+
+	err = uverbs_copy_to(attrs,
+			     MLX5_IB_ATTR_ALLOC_DM_RESP_PAGE_INDEX,
+			     &page_idx, sizeof(page_idx));
+	if (err)
+		goto err_dealloc;
+
+	bitmap_set(to_mucontext(context)->dm_pages, page_idx,
+		   DIV_ROUND_UP(act_size, PAGE_SIZE));
+
+	dm->dev_addr = memic_addr;
+
+	return &dm->ibdm;
+
+err_dealloc:
+	mlx5_cmd_dealloc_memic(memic, memic_addr,
+			       act_size);
+err_free:
+	kfree(dm);
+	return ERR_PTR(err);
+}
+
+int mlx5_ib_dealloc_dm(struct ib_dm *ibdm)
+{
+	struct mlx5_memic *memic = &to_mdev(ibdm->device)->memic;
+	struct mlx5_ib_dm *dm = to_mdm(ibdm);
+	u64 act_size = roundup(dm->ibdm.length, MLX5_MEMIC_BASE_SIZE);
+	u32 page_idx;
+	int ret;
+
+	ret = mlx5_cmd_dealloc_memic(memic, dm->dev_addr, act_size);
+	if (ret)
+		return ret;
+
+	page_idx = (dm->dev_addr - pci_resource_start(memic->dev->pdev, 0) -
+		    MLX5_CAP64_DEV_MEM(memic->dev, memic_bar_start_addr)) >>
+		    PAGE_SHIFT;
+	bitmap_clear(to_mucontext(ibdm->uobject->context)->dm_pages,
+		     page_idx,
+		     DIV_ROUND_UP(act_size, PAGE_SIZE));
+
+	kfree(dm);
+
+	return 0;
+}
+
 static struct ib_pd *mlx5_ib_alloc_pd(struct ib_device *ibdev,
 				      struct ib_ucontext *context,
 				      struct ib_udata *udata)
@@ -2317,8 +2458,28 @@ static void set_tos(void *outer_c, void *outer_v, u8 mask, u8 val)
 		   offsetof(typeof(filter), field) -\
 		   sizeof(filter.field))
 
+static int parse_flow_flow_action(const union ib_flow_spec *ib_spec,
+				  const struct ib_flow_attr *flow_attr,
+				  struct mlx5_flow_act *action)
+{
+	struct mlx5_ib_flow_action *maction = to_mflow_act(ib_spec->action.act);
+
+	switch (maction->ib_action.type) {
+	case IB_FLOW_ACTION_ESP:
+		/* Currently only AES_GCM keymat is supported by the driver */
+		action->esp_id = (uintptr_t)maction->esp_aes_gcm.ctx;
+		action->action |= flow_attr->flags & IB_FLOW_ATTR_FLAGS_EGRESS ?
+			MLX5_FLOW_CONTEXT_ACTION_ENCRYPT :
+			MLX5_FLOW_CONTEXT_ACTION_DECRYPT;
+		return 0;
+	default:
+		return -EOPNOTSUPP;
+	}
+}
+
 static int parse_flow_attr(struct mlx5_core_dev *mdev, u32 *match_c,
 			   u32 *match_v, const union ib_flow_spec *ib_spec,
+			   const struct ib_flow_attr *flow_attr,
 			   struct mlx5_flow_act *action)
 {
 	void *misc_params_c = MLX5_ADDR_OF(fte_match_param, match_c,
@@ -2328,6 +2489,7 @@ static int parse_flow_attr(struct mlx5_core_dev *mdev, u32 *match_c,
 	void *headers_c;
 	void *headers_v;
 	int match_ipv;
+	int ret;
 
 	if (ib_spec->type & IB_FLOW_SPEC_INNER) {
 		headers_c = MLX5_ADDR_OF(fte_match_param, match_c,
@@ -2478,7 +2640,15 @@ static int parse_flow_attr(struct mlx5_core_dev *mdev, u32 *match_c,
 			       ntohl(ib_spec->ipv6.mask.flow_label),
 			       ntohl(ib_spec->ipv6.val.flow_label),
 			       ib_spec->type & IB_FLOW_SPEC_INNER);
+		break;
+	case IB_FLOW_SPEC_ESP:
+		if (ib_spec->esp.mask.seq)
+			return -EOPNOTSUPP;
 
+		MLX5_SET(fte_match_set_misc, misc_params_c, outer_esp_spi,
+			 ntohl(ib_spec->esp.mask.spi));
+		MLX5_SET(fte_match_set_misc, misc_params_v, outer_esp_spi,
+			 ntohl(ib_spec->esp.val.spi));
 		break;
 	case IB_FLOW_SPEC_TCP:
 		if (FIELDS_NOT_SUPPORTED(ib_spec->tcp_udp.mask,
@@ -2546,6 +2716,11 @@ static int parse_flow_attr(struct mlx5_core_dev *mdev, u32 *match_c,
 			return -EOPNOTSUPP;
 		action->action |= MLX5_FLOW_CONTEXT_ACTION_DROP;
 		break;
+	case IB_FLOW_SPEC_ACTION_HANDLE:
+		ret = parse_flow_flow_action(ib_spec, flow_attr, action);
+		if (ret)
+			return ret;
+		break;
 	default:
 		return -EINVAL;
 	}
@@ -2587,6 +2762,46 @@ static bool flow_is_multicast_only(const struct ib_flow_attr *ib_attr)
 	return false;
 }
 
+enum valid_spec {
+	VALID_SPEC_INVALID,
+	VALID_SPEC_VALID,
+	VALID_SPEC_NA,
+};
+
+static enum valid_spec
+is_valid_esp_aes_gcm(struct mlx5_core_dev *mdev,
+		     const struct mlx5_flow_spec *spec,
+		     const struct mlx5_flow_act *flow_act,
+		     bool egress)
+{
+	const u32 *match_c = spec->match_criteria;
+	bool is_crypto =
+		(flow_act->action & (MLX5_FLOW_CONTEXT_ACTION_ENCRYPT |
+				     MLX5_FLOW_CONTEXT_ACTION_DECRYPT));
+	bool is_ipsec = mlx5_fs_is_ipsec_flow(match_c);
+	bool is_drop = flow_act->action & MLX5_FLOW_CONTEXT_ACTION_DROP;
+
+	/*
+	 * Currently only crypto is supported in egress, when regular egress
+	 * rules would be supported, always return VALID_SPEC_NA.
+	 */
+	if (!is_crypto)
+		return egress ? VALID_SPEC_INVALID : VALID_SPEC_NA;
+
+	return is_crypto && is_ipsec &&
+		(!egress || (!is_drop && !flow_act->has_flow_tag)) ?
+		VALID_SPEC_VALID : VALID_SPEC_INVALID;
+}
+
+static bool is_valid_spec(struct mlx5_core_dev *mdev,
+			  const struct mlx5_flow_spec *spec,
+			  const struct mlx5_flow_act *flow_act,
+			  bool egress)
+{
+	/* We curretly only support ipsec egress flow */
+	return is_valid_esp_aes_gcm(mdev, spec, flow_act, egress) != VALID_SPEC_INVALID;
+}
+
 static bool is_valid_ethertype(struct mlx5_core_dev *mdev,
 			       const struct ib_flow_attr *flow_attr,
 			       bool check_inner)
@@ -2711,13 +2926,17 @@ static struct mlx5_ib_flow_prio *get_flow_table(struct mlx5_ib_dev *dev,
 	max_table_size = BIT(MLX5_CAP_FLOWTABLE_NIC_RX(dev->mdev,
 						       log_max_ft_size));
 	if (flow_attr->type == IB_FLOW_ATTR_NORMAL) {
-		if (flow_is_multicast_only(flow_attr) &&
-		    !dont_trap)
+		if (ft_type == MLX5_IB_FT_TX)
+			priority = 0;
+		else if (flow_is_multicast_only(flow_attr) &&
+			 !dont_trap)
 			priority = MLX5_IB_FLOW_MCAST_PRIO;
 		else
 			priority = ib_prio_to_core_prio(flow_attr->priority,
 							dont_trap);
 		ns = mlx5_get_flow_namespace(dev->mdev,
+					     ft_type == MLX5_IB_FT_TX ?
+					     MLX5_FLOW_NAMESPACE_EGRESS :
 					     MLX5_FLOW_NAMESPACE_BYPASS);
 		num_entries = MLX5_FS_MAX_ENTRIES;
 		num_groups = MLX5_FS_MAX_TYPES;
@@ -2804,6 +3023,7 @@ static struct mlx5_ib_flow_handler *_create_flow_rule(struct mlx5_ib_dev *dev,
 	unsigned int spec_index;
 	int err = 0;
 	int dest_num = 1;
+	bool is_egress = flow_attr->flags & IB_FLOW_ATTR_FLAGS_EGRESS;
 
 	if (!is_valid_attr(dev->mdev, flow_attr))
 		return ERR_PTR(-EINVAL);
@@ -2820,7 +3040,7 @@ static struct mlx5_ib_flow_handler *_create_flow_rule(struct mlx5_ib_dev *dev,
 	for (spec_index = 0; spec_index < flow_attr->num_of_specs; spec_index++) {
 		err = parse_flow_attr(dev->mdev, spec->match_criteria,
 				      spec->match_value,
-				      ib_flow, &flow_act);
+				      ib_flow, flow_attr, &flow_act);
 		if (err < 0)
 			goto free;
 
@@ -2843,12 +3063,23 @@ static struct mlx5_ib_flow_handler *_create_flow_rule(struct mlx5_ib_dev *dev,
 	}
 
 	spec->match_criteria_enable = get_match_criteria_enable(spec->match_criteria);
+
+	if (is_egress &&
+	    !is_valid_spec(dev->mdev, spec, &flow_act, is_egress)) {
+		err = -EINVAL;
+		goto free;
+	}
+
 	if (flow_act.action & MLX5_FLOW_CONTEXT_ACTION_DROP) {
 		rule_dst = NULL;
 		dest_num = 0;
 	} else {
-		flow_act.action = dst ? MLX5_FLOW_CONTEXT_ACTION_FWD_DEST :
-		    MLX5_FLOW_CONTEXT_ACTION_FWD_NEXT_PRIO;
+		if (is_egress)
+			flow_act.action |= MLX5_FLOW_CONTEXT_ACTION_ALLOW;
+		else
+			flow_act.action |=
+				dst ? MLX5_FLOW_CONTEXT_ACTION_FWD_DEST :
+					MLX5_FLOW_CONTEXT_ACTION_FWD_NEXT_PRIO;
 	}
 
 	if (flow_act.has_flow_tag &&
@@ -3022,6 +3253,7 @@ static struct ib_flow *mlx5_ib_create_flow(struct ib_qp *qp,
 	struct mlx5_flow_destination *dst = NULL;
 	struct mlx5_ib_flow_prio *ft_prio_tx = NULL;
 	struct mlx5_ib_flow_prio *ft_prio;
+	bool is_egress = flow_attr->flags & IB_FLOW_ATTR_FLAGS_EGRESS;
 	int err;
 	int underlay_qpn;
 
@@ -3030,7 +3262,13 @@ static struct ib_flow *mlx5_ib_create_flow(struct ib_qp *qp,
 
 	if (domain != IB_FLOW_DOMAIN_USER ||
 	    flow_attr->port > dev->num_ports ||
-	    (flow_attr->flags & ~IB_FLOW_ATTR_FLAGS_DONT_TRAP))
+	    (flow_attr->flags & ~(IB_FLOW_ATTR_FLAGS_DONT_TRAP |
+				  IB_FLOW_ATTR_FLAGS_EGRESS)))
+		return ERR_PTR(-EINVAL);
+
+	if (is_egress &&
+	    (flow_attr->type == IB_FLOW_ATTR_ALL_DEFAULT ||
+	     flow_attr->type == IB_FLOW_ATTR_MC_DEFAULT))
 		return ERR_PTR(-EINVAL);
 
 	dst = kzalloc(sizeof(*dst), GFP_KERNEL);
@@ -3039,7 +3277,8 @@ static struct ib_flow *mlx5_ib_create_flow(struct ib_qp *qp,
 
 	mutex_lock(&dev->flow_db->lock);
 
-	ft_prio = get_flow_table(dev, flow_attr, MLX5_IB_FT_RX);
+	ft_prio = get_flow_table(dev, flow_attr,
+				 is_egress ? MLX5_IB_FT_TX : MLX5_IB_FT_RX);
 	if (IS_ERR(ft_prio)) {
 		err = PTR_ERR(ft_prio);
 		goto unlock;
@@ -3053,11 +3292,15 @@ static struct ib_flow *mlx5_ib_create_flow(struct ib_qp *qp,
 		}
 	}
 
-	dst->type = MLX5_FLOW_DESTINATION_TYPE_TIR;
-	if (mqp->flags & MLX5_IB_QP_RSS)
-		dst->tir_num = mqp->rss_qp.tirn;
-	else
-		dst->tir_num = mqp->raw_packet_qp.rq.tirn;
+	if (is_egress) {
+		dst->type = MLX5_FLOW_DESTINATION_TYPE_PORT;
+	} else {
+		dst->type = MLX5_FLOW_DESTINATION_TYPE_TIR;
+		if (mqp->flags & MLX5_IB_QP_RSS)
+			dst->tir_num = mqp->rss_qp.tirn;
+		else
+			dst->tir_num = mqp->raw_packet_qp.rq.tirn;
+	}
 
 	if (flow_attr->type == IB_FLOW_ATTR_NORMAL) {
 		if (flow_attr->flags & IB_FLOW_ATTR_FLAGS_DONT_TRAP)  {
@@ -3102,6 +3345,170 @@ unlock:
 	return ERR_PTR(err);
 }
 
+static u32 mlx5_ib_flow_action_flags_to_accel_xfrm_flags(u32 mlx5_flags)
+{
+	u32 flags = 0;
+
+	if (mlx5_flags & MLX5_IB_UAPI_FLOW_ACTION_FLAGS_REQUIRE_METADATA)
+		flags |= MLX5_ACCEL_XFRM_FLAG_REQUIRE_METADATA;
+
+	return flags;
+}
+
+#define MLX5_FLOW_ACTION_ESP_CREATE_LAST_SUPPORTED	MLX5_IB_UAPI_FLOW_ACTION_FLAGS_REQUIRE_METADATA
+static struct ib_flow_action *
+mlx5_ib_create_flow_action_esp(struct ib_device *device,
+			       const struct ib_flow_action_attrs_esp *attr,
+			       struct uverbs_attr_bundle *attrs)
+{
+	struct mlx5_ib_dev *mdev = to_mdev(device);
+	struct ib_uverbs_flow_action_esp_keymat_aes_gcm *aes_gcm;
+	struct mlx5_accel_esp_xfrm_attrs accel_attrs = {};
+	struct mlx5_ib_flow_action *action;
+	u64 action_flags;
+	u64 flags;
+	int err = 0;
+
+	if (IS_UVERBS_COPY_ERR(uverbs_copy_from(&action_flags, attrs,
+						MLX5_IB_ATTR_CREATE_FLOW_ACTION_FLAGS)))
+		return ERR_PTR(-EFAULT);
+
+	if (action_flags >= (MLX5_FLOW_ACTION_ESP_CREATE_LAST_SUPPORTED << 1))
+		return ERR_PTR(-EOPNOTSUPP);
+
+	flags = mlx5_ib_flow_action_flags_to_accel_xfrm_flags(action_flags);
+
+	/* We current only support a subset of the standard features. Only a
+	 * keymat of type AES_GCM, with icv_len == 16, iv_algo == SEQ and esn
+	 * (with overlap). Full offload mode isn't supported.
+	 */
+	if (!attr->keymat || attr->replay || attr->encap ||
+	    attr->spi || attr->seq || attr->tfc_pad ||
+	    attr->hard_limit_pkts ||
+	    (attr->flags & ~(IB_FLOW_ACTION_ESP_FLAGS_ESN_TRIGGERED |
+			     IB_UVERBS_FLOW_ACTION_ESP_FLAGS_ENCRYPT)))
+		return ERR_PTR(-EOPNOTSUPP);
+
+	if (attr->keymat->protocol !=
+	    IB_UVERBS_FLOW_ACTION_ESP_KEYMAT_AES_GCM)
+		return ERR_PTR(-EOPNOTSUPP);
+
+	aes_gcm = &attr->keymat->keymat.aes_gcm;
+
+	if (aes_gcm->icv_len != 16 ||
+	    aes_gcm->iv_algo != IB_UVERBS_FLOW_ACTION_IV_ALGO_SEQ)
+		return ERR_PTR(-EOPNOTSUPP);
+
+	action = kmalloc(sizeof(*action), GFP_KERNEL);
+	if (!action)
+		return ERR_PTR(-ENOMEM);
+
+	action->esp_aes_gcm.ib_flags = attr->flags;
+	memcpy(&accel_attrs.keymat.aes_gcm.aes_key, &aes_gcm->aes_key,
+	       sizeof(accel_attrs.keymat.aes_gcm.aes_key));
+	accel_attrs.keymat.aes_gcm.key_len = aes_gcm->key_len * 8;
+	memcpy(&accel_attrs.keymat.aes_gcm.salt, &aes_gcm->salt,
+	       sizeof(accel_attrs.keymat.aes_gcm.salt));
+	memcpy(&accel_attrs.keymat.aes_gcm.seq_iv, &aes_gcm->iv,
+	       sizeof(accel_attrs.keymat.aes_gcm.seq_iv));
+	accel_attrs.keymat.aes_gcm.icv_len = aes_gcm->icv_len * 8;
+	accel_attrs.keymat.aes_gcm.iv_algo = MLX5_ACCEL_ESP_AES_GCM_IV_ALGO_SEQ;
+	accel_attrs.keymat_type = MLX5_ACCEL_ESP_KEYMAT_AES_GCM;
+
+	accel_attrs.esn = attr->esn;
+	if (attr->flags & IB_FLOW_ACTION_ESP_FLAGS_ESN_TRIGGERED)
+		accel_attrs.flags |= MLX5_ACCEL_ESP_FLAGS_ESN_TRIGGERED;
+	if (attr->flags & IB_UVERBS_FLOW_ACTION_ESP_FLAGS_ESN_NEW_WINDOW)
+		accel_attrs.flags |= MLX5_ACCEL_ESP_FLAGS_ESN_STATE_OVERLAP;
+
+	if (attr->flags & IB_UVERBS_FLOW_ACTION_ESP_FLAGS_ENCRYPT)
+		accel_attrs.action |= MLX5_ACCEL_ESP_ACTION_ENCRYPT;
+
+	action->esp_aes_gcm.ctx =
+		mlx5_accel_esp_create_xfrm(mdev->mdev, &accel_attrs, flags);
+	if (IS_ERR(action->esp_aes_gcm.ctx)) {
+		err = PTR_ERR(action->esp_aes_gcm.ctx);
+		goto err_parse;
+	}
+
+	action->esp_aes_gcm.ib_flags = attr->flags;
+
+	return &action->ib_action;
+
+err_parse:
+	kfree(action);
+	return ERR_PTR(err);
+}
+
+static int
+mlx5_ib_modify_flow_action_esp(struct ib_flow_action *action,
+			       const struct ib_flow_action_attrs_esp *attr,
+			       struct uverbs_attr_bundle *attrs)
+{
+	struct mlx5_ib_flow_action *maction = to_mflow_act(action);
+	struct mlx5_accel_esp_xfrm_attrs accel_attrs;
+	int err = 0;
+
+	if (attr->keymat || attr->replay || attr->encap ||
+	    attr->spi || attr->seq || attr->tfc_pad ||
+	    attr->hard_limit_pkts ||
+	    (attr->flags & ~(IB_FLOW_ACTION_ESP_FLAGS_ESN_TRIGGERED |
+			     IB_FLOW_ACTION_ESP_FLAGS_MOD_ESP_ATTRS |
+			     IB_UVERBS_FLOW_ACTION_ESP_FLAGS_ESN_NEW_WINDOW)))
+		return -EOPNOTSUPP;
+
+	/* Only the ESN value or the MLX5_ACCEL_ESP_FLAGS_ESN_STATE_OVERLAP can
+	 * be modified.
+	 */
+	if (!(maction->esp_aes_gcm.ib_flags &
+	      IB_FLOW_ACTION_ESP_FLAGS_ESN_TRIGGERED) &&
+	    attr->flags & (IB_FLOW_ACTION_ESP_FLAGS_ESN_TRIGGERED |
+			   IB_UVERBS_FLOW_ACTION_ESP_FLAGS_ESN_NEW_WINDOW))
+		return -EINVAL;
+
+	memcpy(&accel_attrs, &maction->esp_aes_gcm.ctx->attrs,
+	       sizeof(accel_attrs));
+
+	accel_attrs.esn = attr->esn;
+	if (attr->flags & IB_UVERBS_FLOW_ACTION_ESP_FLAGS_ESN_NEW_WINDOW)
+		accel_attrs.flags |= MLX5_ACCEL_ESP_FLAGS_ESN_STATE_OVERLAP;
+	else
+		accel_attrs.flags &= ~MLX5_ACCEL_ESP_FLAGS_ESN_STATE_OVERLAP;
+
+	err = mlx5_accel_esp_modify_xfrm(maction->esp_aes_gcm.ctx,
+					 &accel_attrs);
+	if (err)
+		return err;
+
+	maction->esp_aes_gcm.ib_flags &=
+		~IB_UVERBS_FLOW_ACTION_ESP_FLAGS_ESN_NEW_WINDOW;
+	maction->esp_aes_gcm.ib_flags |=
+		attr->flags & IB_UVERBS_FLOW_ACTION_ESP_FLAGS_ESN_NEW_WINDOW;
+
+	return 0;
+}
+
+static int mlx5_ib_destroy_flow_action(struct ib_flow_action *action)
+{
+	struct mlx5_ib_flow_action *maction = to_mflow_act(action);
+
+	switch (action->type) {
+	case IB_FLOW_ACTION_ESP:
+		/*
+		 * We only support aes_gcm by now, so we implicitly know this is
+		 * the underline crypto.
+		 */
+		mlx5_accel_esp_destroy_xfrm(maction->esp_aes_gcm.ctx);
+		break;
+	default:
+		WARN_ON(true);
+		break;
+	}
+
+	kfree(maction);
+	return 0;
+}
+
 static int mlx5_ib_mcg_attach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid)
 {
 	struct mlx5_ib_dev *dev = to_mdev(ibqp->device);
@@ -4553,6 +4960,47 @@ static void mlx5_ib_cleanup_multiport_master(struct mlx5_ib_dev *dev)
 	mlx5_nic_vport_disable_roce(dev->mdev);
 }
 
+ADD_UVERBS_ATTRIBUTES_SIMPLE(mlx5_ib_dm, UVERBS_OBJECT_DM,
+			     UVERBS_METHOD_DM_ALLOC,
+			     &UVERBS_ATTR_PTR_OUT(MLX5_IB_ATTR_ALLOC_DM_RESP_START_OFFSET,
+						  UVERBS_ATTR_TYPE(u64),
+						  UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)),
+			     &UVERBS_ATTR_PTR_OUT(MLX5_IB_ATTR_ALLOC_DM_RESP_PAGE_INDEX,
+						  UVERBS_ATTR_TYPE(u16),
+						  UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)));
+
+ADD_UVERBS_ATTRIBUTES_SIMPLE(mlx5_ib_flow_action, UVERBS_OBJECT_FLOW_ACTION,
+			     UVERBS_METHOD_FLOW_ACTION_ESP_CREATE,
+			     &UVERBS_ATTR_PTR_IN(MLX5_IB_ATTR_CREATE_FLOW_ACTION_FLAGS,
+						 UVERBS_ATTR_TYPE(u64),
+						 UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)));
+
+#define NUM_TREES	2
+static int populate_specs_root(struct mlx5_ib_dev *dev)
+{
+	const struct uverbs_object_tree_def *default_root[NUM_TREES + 1] = {
+		uverbs_default_get_objects()};
+	size_t num_trees = 1;
+
+	if (mlx5_accel_ipsec_device_caps(dev->mdev) & MLX5_ACCEL_IPSEC_CAP_DEVICE &&
+	    !WARN_ON(num_trees >= ARRAY_SIZE(default_root)))
+		default_root[num_trees++] = &mlx5_ib_flow_action;
+
+	if (MLX5_CAP_DEV_MEM(dev->mdev, memic) &&
+	    !WARN_ON(num_trees >= ARRAY_SIZE(default_root)))
+		default_root[num_trees++] = &mlx5_ib_dm;
+
+	dev->ib_dev.specs_root =
+		uverbs_alloc_spec_tree(num_trees, default_root);
+
+	return PTR_ERR_OR_ZERO(dev->ib_dev.specs_root);
+}
+
+static void depopulate_specs_root(struct mlx5_ib_dev *dev)
+{
+	uverbs_free_spec_tree(dev->ib_dev.specs_root);
+}
+
 void mlx5_ib_stage_init_cleanup(struct mlx5_ib_dev *dev)
 {
 	mlx5_ib_cleanup_multiport_master(dev);
@@ -4616,6 +5064,9 @@ int mlx5_ib_stage_init_init(struct mlx5_ib_dev *dev)
 	INIT_LIST_HEAD(&dev->qp_list);
 	spin_lock_init(&dev->reset_flow_resource_lock);
 
+	spin_lock_init(&dev->memic.memic_lock);
+	dev->memic.dev = mdev;
+
 #ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
 	err = init_srcu_struct(&dev->mr_srcu);
 	if (err)
@@ -4778,11 +5229,21 @@ int mlx5_ib_stage_caps_init(struct mlx5_ib_dev *dev)
 			(1ull << IB_USER_VERBS_CMD_CLOSE_XRCD);
 	}
 
+	if (MLX5_CAP_DEV_MEM(mdev, memic)) {
+		dev->ib_dev.alloc_dm = mlx5_ib_alloc_dm;
+		dev->ib_dev.dealloc_dm = mlx5_ib_dealloc_dm;
+		dev->ib_dev.reg_dm_mr = mlx5_ib_reg_dm_mr;
+	}
+
 	dev->ib_dev.create_flow	= mlx5_ib_create_flow;
 	dev->ib_dev.destroy_flow = mlx5_ib_destroy_flow;
 	dev->ib_dev.uverbs_ex_cmd_mask |=
 			(1ull << IB_USER_VERBS_EX_CMD_CREATE_FLOW) |
 			(1ull << IB_USER_VERBS_EX_CMD_DESTROY_FLOW);
+	dev->ib_dev.create_flow_action_esp = mlx5_ib_create_flow_action_esp;
+	dev->ib_dev.destroy_flow_action = mlx5_ib_destroy_flow_action;
+	dev->ib_dev.modify_flow_action_esp = mlx5_ib_modify_flow_action_esp;
+	dev->ib_dev.driver_id = RDMA_DRIVER_MLX5;
 
 	err = init_node_data(dev);
 	if (err)
@@ -4997,11 +5458,21 @@ void mlx5_ib_stage_bfrag_cleanup(struct mlx5_ib_dev *dev)
 	mlx5_free_bfreg(dev->mdev, &dev->bfreg);
 }
 
+static int mlx5_ib_stage_populate_specs(struct mlx5_ib_dev *dev)
+{
+	return populate_specs_root(dev);
+}
+
 int mlx5_ib_stage_ib_reg_init(struct mlx5_ib_dev *dev)
 {
 	return ib_register_device(&dev->ib_dev, NULL);
 }
 
+static void mlx5_ib_stage_depopulate_specs(struct mlx5_ib_dev *dev)
+{
+	depopulate_specs_root(dev);
+}
+
 void mlx5_ib_stage_pre_ib_reg_umr_cleanup(struct mlx5_ib_dev *dev)
 {
 	destroy_umrc_res(dev);
@@ -5136,6 +5607,9 @@ static const struct mlx5_ib_profile pf_profile = {
 	STAGE_CREATE(MLX5_IB_STAGE_PRE_IB_REG_UMR,
 		     NULL,
 		     mlx5_ib_stage_pre_ib_reg_umr_cleanup),
+	STAGE_CREATE(MLX5_IB_STAGE_SPECS,
+		     mlx5_ib_stage_populate_specs,
+		     mlx5_ib_stage_depopulate_specs),
 	STAGE_CREATE(MLX5_IB_STAGE_IB_REG,
 		     mlx5_ib_stage_ib_reg_init,
 		     mlx5_ib_stage_ib_reg_cleanup),
@@ -5181,6 +5655,9 @@ static const struct mlx5_ib_profile nic_rep_profile = {
 	STAGE_CREATE(MLX5_IB_STAGE_PRE_IB_REG_UMR,
 		     NULL,
 		     mlx5_ib_stage_pre_ib_reg_umr_cleanup),
+	STAGE_CREATE(MLX5_IB_STAGE_SPECS,
+		     mlx5_ib_stage_populate_specs,
+		     mlx5_ib_stage_depopulate_specs),
 	STAGE_CREATE(MLX5_IB_STAGE_IB_REG,
 		     mlx5_ib_stage_ib_reg_init,
 		     mlx5_ib_stage_ib_reg_cleanup),
@@ -5301,13 +5778,32 @@ static struct mlx5_interface mlx5_ib_interface = {
 	.protocol	= MLX5_INTERFACE_PROTOCOL_IB,
 };
 
+unsigned long mlx5_ib_get_xlt_emergency_page(void)
+{
+	mutex_lock(&xlt_emergency_page_mutex);
+	return xlt_emergency_page;
+}
+
+void mlx5_ib_put_xlt_emergency_page(void)
+{
+	mutex_unlock(&xlt_emergency_page_mutex);
+}
+
 static int __init mlx5_ib_init(void)
 {
 	int err;
 
+	xlt_emergency_page = __get_free_page(GFP_KERNEL);
+	if (!xlt_emergency_page)
+		return -ENOMEM;
+
+	mutex_init(&xlt_emergency_page_mutex);
+
 	mlx5_ib_event_wq = alloc_ordered_workqueue("mlx5_ib_event_wq", 0);
-	if (!mlx5_ib_event_wq)
+	if (!mlx5_ib_event_wq) {
+		free_page(xlt_emergency_page);
 		return -ENOMEM;
+	}
 
 	mlx5_ib_odp_init();
 
@@ -5320,6 +5816,8 @@ static void __exit mlx5_ib_cleanup(void)
 {
 	mlx5_unregister_interface(&mlx5_ib_interface);
 	destroy_workqueue(mlx5_ib_event_wq);
+	mutex_destroy(&xlt_emergency_page_mutex);
+	free_page(xlt_emergency_page);
 }
 
 module_init(mlx5_ib_init);

drivers/infiniband/hw/mlx5/mlx5_ib.h

@@ -45,6 +45,7 @@
 #include <linux/mlx5/transobj.h>
 #include <rdma/ib_user_verbs.h>
 #include <rdma/mlx5-abi.h>
+#include <rdma/uverbs_ioctl.h>
 
 #define mlx5_ib_dbg(dev, format, arg...)				\
 pr_debug("%s:%s:%d:(pid %d): " format, (dev)->ib_dev.name, __func__,	\
@@ -108,6 +109,16 @@ enum {
 	MLX5_IB_INVALID_BFREG		= BIT(31),
 };
 
+enum {
+	MLX5_MAX_MEMIC_PAGES = 0x100,
+	MLX5_MEMIC_ALLOC_SIZE_MASK = 0x3f,
+};
+
+enum {
+	MLX5_MEMIC_BASE_ALIGN	= 6,
+	MLX5_MEMIC_BASE_SIZE	= 1 << MLX5_MEMIC_BASE_ALIGN,
+};
+
 struct mlx5_ib_vma_private_data {
 	struct list_head list;
 	struct vm_area_struct *vma;
@@ -130,10 +141,8 @@ struct mlx5_ib_ucontext {
 	/* protect vma_private_list add/del */
 	struct mutex		vma_private_list_mutex;
 
-	unsigned long		upd_xlt_page;
-	/* protect ODP/KSM */
-	struct mutex		upd_xlt_page_mutex;
 	u64			lib_caps;
+	DECLARE_BITMAP(dm_pages, MLX5_MAX_MEMIC_PAGES);
 };
 
 static inline struct mlx5_ib_ucontext *to_mucontext(struct ib_ucontext *ibucontext)
@@ -155,6 +164,7 @@ struct mlx5_ib_pd {
 
 #define MLX5_IB_NUM_FLOW_FT		(MLX5_IB_FLOW_LEFTOVERS_PRIO + 1)
 #define MLX5_IB_NUM_SNIFFER_FTS		2
+#define MLX5_IB_NUM_EGRESS_FTS		1
 struct mlx5_ib_flow_prio {
 	struct mlx5_flow_table		*flow_table;
 	unsigned int			refcount;
@@ -170,6 +180,7 @@ struct mlx5_ib_flow_handler {
 struct mlx5_ib_flow_db {
 	struct mlx5_ib_flow_prio	prios[MLX5_IB_NUM_FLOW_FT];
 	struct mlx5_ib_flow_prio	sniffer[MLX5_IB_NUM_SNIFFER_FTS];
+	struct mlx5_ib_flow_prio	egress[MLX5_IB_NUM_EGRESS_FTS];
 	struct mlx5_flow_table		*lag_demux_ft;
 	/* Protect flow steering bypass flow tables
 	 * when add/del flow rules.
@@ -406,7 +417,7 @@ struct mlx5_ib_qp {
 	struct list_head	qps_list;
 	struct list_head	cq_recv_list;
 	struct list_head	cq_send_list;
-	u32			rate_limit;
+	struct mlx5_rate_limit	rl;
 	u32                     underlay_qpn;
 	bool			tunnel_offload_en;
 	/* storage for qp sub type when core qp type is IB_QPT_DRIVER */
@@ -522,8 +533,19 @@ enum mlx5_ib_mtt_access_flags {
 	MLX5_IB_MTT_WRITE = (1 << 1),
 };
 
+struct mlx5_ib_dm {
+	struct ib_dm		ibdm;
+	phys_addr_t		dev_addr;
+};
+
 #define MLX5_IB_MTT_PRESENT (MLX5_IB_MTT_READ | MLX5_IB_MTT_WRITE)
 
+#define MLX5_IB_DM_ALLOWED_ACCESS (IB_ACCESS_LOCAL_WRITE   |\
+				   IB_ACCESS_REMOTE_WRITE  |\
+				   IB_ACCESS_REMOTE_READ   |\
+				   IB_ACCESS_REMOTE_ATOMIC |\
+				   IB_ZERO_BASED)
+
 struct mlx5_ib_mr {
 	struct ib_mr		ibmr;
 	void			*descs;
@@ -743,6 +765,7 @@ enum mlx5_ib_stages {
 	MLX5_IB_STAGE_UAR,
 	MLX5_IB_STAGE_BFREG,
 	MLX5_IB_STAGE_PRE_IB_REG_UMR,
+	MLX5_IB_STAGE_SPECS,
 	MLX5_IB_STAGE_IB_REG,
 	MLX5_IB_STAGE_POST_IB_REG_UMR,
 	MLX5_IB_STAGE_DELAY_DROP,
@@ -774,6 +797,22 @@ struct mlx5_ib_multiport_info {
 	bool unaffiliate;
 };
 
+struct mlx5_ib_flow_action {
+	struct ib_flow_action		ib_action;
+	union {
+		struct {
+			u64			    ib_flags;
+			struct mlx5_accel_esp_xfrm *ctx;
+		} esp_aes_gcm;
+	};
+};
+
+struct mlx5_memic {
+	struct mlx5_core_dev *dev;
+	spinlock_t		memic_lock;
+	DECLARE_BITMAP(memic_alloc_pages, MLX5_MAX_MEMIC_PAGES);
+};
+
 struct mlx5_ib_dev {
 	struct ib_device		ib_dev;
 	struct mlx5_core_dev		*mdev;
@@ -820,6 +859,7 @@ struct mlx5_ib_dev {
 	u8			umr_fence;
 	struct list_head	ib_dev_list;
 	u64			sys_image_guid;
+	struct mlx5_memic	memic;
 };
 
 static inline struct mlx5_ib_cq *to_mibcq(struct mlx5_core_cq *mcq)
@@ -887,6 +927,11 @@ static inline struct mlx5_ib_srq *to_mibsrq(struct mlx5_core_srq *msrq)
 	return container_of(msrq, struct mlx5_ib_srq, msrq);
 }
 
+static inline struct mlx5_ib_dm *to_mdm(struct ib_dm *ibdm)
+{
+	return container_of(ibdm, struct mlx5_ib_dm, ibdm);
+}
+
 static inline struct mlx5_ib_mr *to_mmr(struct ib_mr *ibmr)
 {
 	return container_of(ibmr, struct mlx5_ib_mr, ibmr);
@@ -897,6 +942,12 @@ static inline struct mlx5_ib_mw *to_mmw(struct ib_mw *ibmw)
 	return container_of(ibmw, struct mlx5_ib_mw, ibmw);
 }
 
+static inline struct mlx5_ib_flow_action *
+to_mflow_act(struct ib_flow_action *ibact)
+{
+	return container_of(ibact, struct mlx5_ib_flow_action, ib_action);
+}
+
 int mlx5_ib_db_map_user(struct mlx5_ib_ucontext *context, unsigned long virt,
 			struct mlx5_db *db);
 void mlx5_ib_db_unmap_user(struct mlx5_ib_ucontext *context, struct mlx5_db *db);
@@ -1025,7 +1076,14 @@ struct ib_rwq_ind_table *mlx5_ib_create_rwq_ind_table(struct ib_device *device,
 						      struct ib_udata *udata);
 int mlx5_ib_destroy_rwq_ind_table(struct ib_rwq_ind_table *wq_ind_table);
 bool mlx5_ib_dc_atomic_is_supported(struct mlx5_ib_dev *dev);
-
+struct ib_dm *mlx5_ib_alloc_dm(struct ib_device *ibdev,
+			       struct ib_ucontext *context,
+			       struct ib_dm_alloc_attr *attr,
+			       struct uverbs_attr_bundle *attrs);
+int mlx5_ib_dealloc_dm(struct ib_dm *ibdm);
+struct ib_mr *mlx5_ib_reg_dm_mr(struct ib_pd *pd, struct ib_dm *dm,
+				struct ib_dm_mr_attr *attr,
+				struct uverbs_attr_bundle *attrs);
 
 #ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
 void mlx5_ib_internal_fill_odp_caps(struct mlx5_ib_dev *dev);
@@ -1221,4 +1279,7 @@ static inline int get_num_static_uars(struct mlx5_ib_dev *dev,
 	return get_uars_per_sys_page(dev, bfregi->lib_uar_4k) * bfregi->num_static_sys_pages;
 }
 
+unsigned long mlx5_ib_get_xlt_emergency_page(void);
+void mlx5_ib_put_xlt_emergency_page(void);
+
 #endif /* MLX5_IB_H */

drivers/infiniband/hw/mlx5/mr.c

@@ -47,10 +47,25 @@ enum {
 
 #define MLX5_UMR_ALIGN 2048
 
-static int clean_mr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr);
-static int dereg_mr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr);
+static void clean_mr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr);
+static void dereg_mr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr);
 static int mr_cache_max_order(struct mlx5_ib_dev *dev);
 static int unreg_umr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr);
+static bool umr_can_modify_entity_size(struct mlx5_ib_dev *dev)
+{
+	return !MLX5_CAP_GEN(dev->mdev, umr_modify_entity_size_disabled);
+}
+
+static bool umr_can_use_indirect_mkey(struct mlx5_ib_dev *dev)
+{
+	return !MLX5_CAP_GEN(dev->mdev, umr_indirect_mkey_disabled);
+}
+
+static bool use_umr(struct mlx5_ib_dev *dev, int order)
+{
+	return order <= mr_cache_max_order(dev) &&
+		umr_can_modify_entity_size(dev);
+}
 
 static int destroy_mkey(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr)
 {
@@ -189,7 +204,9 @@ static int add_keys(struct mlx5_ib_dev *dev, int c, int num)
 
 		MLX5_SET(mkc, mkc, free, 1);
 		MLX5_SET(mkc, mkc, umr_en, 1);
-		MLX5_SET(mkc, mkc, access_mode, ent->access_mode);
+		MLX5_SET(mkc, mkc, access_mode_1_0, ent->access_mode & 0x3);
+		MLX5_SET(mkc, mkc, access_mode_4_2,
+			 (ent->access_mode >> 2) & 0x7);
 
 		MLX5_SET(mkc, mkc, qpn, 0xffffff);
 		MLX5_SET(mkc, mkc, translations_octword_size, ent->xlt);
@@ -220,26 +237,32 @@ static void remove_keys(struct mlx5_ib_dev *dev, int c, int num)
 {
 	struct mlx5_mr_cache *cache = &dev->cache;
 	struct mlx5_cache_ent *ent = &cache->ent[c];
+	struct mlx5_ib_mr *tmp_mr;
 	struct mlx5_ib_mr *mr;
-	int err;
+	LIST_HEAD(del_list);
 	int i;
 
 	for (i = 0; i < num; i++) {
 		spin_lock_irq(&ent->lock);
 		if (list_empty(&ent->head)) {
 			spin_unlock_irq(&ent->lock);
-			return;
+			break;
 		}
 		mr = list_first_entry(&ent->head, struct mlx5_ib_mr, list);
-		list_del(&mr->list);
+		list_move(&mr->list, &del_list);
 		ent->cur--;
 		ent->size--;
 		spin_unlock_irq(&ent->lock);
-		err = destroy_mkey(dev, mr);
-		if (err)
-			mlx5_ib_warn(dev, "failed destroy mkey\n");
-		else
-			kfree(mr);
+		mlx5_core_destroy_mkey(dev->mdev, &mr->mmkey);
+	}
+
+#ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
+	synchronize_srcu(&dev->mr_srcu);
+#endif
+
+	list_for_each_entry_safe(mr, tmp_mr, &del_list, list) {
+		list_del(&mr->list);
+		kfree(mr);
 	}
 }
 
@@ -562,26 +585,32 @@ static void clean_keys(struct mlx5_ib_dev *dev, int c)
 {
 	struct mlx5_mr_cache *cache = &dev->cache;
 	struct mlx5_cache_ent *ent = &cache->ent[c];
+	struct mlx5_ib_mr *tmp_mr;
 	struct mlx5_ib_mr *mr;
-	int err;
+	LIST_HEAD(del_list);
 
 	cancel_delayed_work(&ent->dwork);
 	while (1) {
 		spin_lock_irq(&ent->lock);
 		if (list_empty(&ent->head)) {
 			spin_unlock_irq(&ent->lock);
-			return;
+			break;
 		}
 		mr = list_first_entry(&ent->head, struct mlx5_ib_mr, list);
-		list_del(&mr->list);
+		list_move(&mr->list, &del_list);
 		ent->cur--;
 		ent->size--;
 		spin_unlock_irq(&ent->lock);
-		err = destroy_mkey(dev, mr);
-		if (err)
-			mlx5_ib_warn(dev, "failed destroy mkey\n");
-		else
-			kfree(mr);
+		mlx5_core_destroy_mkey(dev->mdev, &mr->mmkey);
+	}
+
+#ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
+	synchronize_srcu(&dev->mr_srcu);
+#endif
+
+	list_for_each_entry_safe(mr, tmp_mr, &del_list, list) {
+		list_del(&mr->list);
+		kfree(mr);
 	}
 }
 
@@ -780,7 +809,7 @@ struct ib_mr *mlx5_ib_get_dma_mr(struct ib_pd *pd, int acc)
 
 	mkc = MLX5_ADDR_OF(create_mkey_in, in, memory_key_mkey_entry);
 
-	MLX5_SET(mkc, mkc, access_mode, MLX5_MKC_ACCESS_MODE_PA);
+	MLX5_SET(mkc, mkc, access_mode_1_0, MLX5_MKC_ACCESS_MODE_PA);
 	MLX5_SET(mkc, mkc, a, !!(acc & IB_ACCESS_REMOTE_ATOMIC));
 	MLX5_SET(mkc, mkc, rw, !!(acc & IB_ACCESS_REMOTE_WRITE));
 	MLX5_SET(mkc, mkc, rr, !!(acc & IB_ACCESS_REMOTE_READ));
@@ -947,7 +976,10 @@ static inline int populate_xlt(struct mlx5_ib_mr *mr, int idx, int npages,
 {
 	struct mlx5_ib_dev *dev = mr->dev;
 	struct ib_umem *umem = mr->umem;
+
 	if (flags & MLX5_IB_UPD_XLT_INDIRECT) {
+		if (!umr_can_use_indirect_mkey(dev))
+			return -EPERM;
 		mlx5_odp_populate_klm(xlt, idx, npages, mr, flags);
 		return npages;
 	}
@@ -977,7 +1009,6 @@ int mlx5_ib_update_xlt(struct mlx5_ib_mr *mr, u64 idx, int npages,
 {
 	struct mlx5_ib_dev *dev = mr->dev;
 	struct device *ddev = dev->ib_dev.dev.parent;
-	struct mlx5_ib_ucontext *uctx = NULL;
 	int size;
 	void *xlt;
 	dma_addr_t dma;
@@ -993,6 +1024,11 @@ int mlx5_ib_update_xlt(struct mlx5_ib_mr *mr, u64 idx, int npages,
 	size_t pages_to_map = 0;
 	size_t pages_iter = 0;
 	gfp_t gfp;
+	bool use_emergency_page = false;
+
+	if ((flags & MLX5_IB_UPD_XLT_INDIRECT) &&
+	    !umr_can_use_indirect_mkey(dev))
+		return -EPERM;
 
 	/* UMR copies MTTs in units of MLX5_UMR_MTT_ALIGNMENT bytes,
 	 * so we need to align the offset and length accordingly
@@ -1019,12 +1055,11 @@ int mlx5_ib_update_xlt(struct mlx5_ib_mr *mr, u64 idx, int npages,
 	}
 
 	if (!xlt) {
-		uctx = to_mucontext(mr->ibmr.pd->uobject->context);
 		mlx5_ib_warn(dev, "Using XLT emergency buffer\n");
+		xlt = (void *)mlx5_ib_get_xlt_emergency_page();
 		size = PAGE_SIZE;
-		xlt = (void *)uctx->upd_xlt_page;
-		mutex_lock(&uctx->upd_xlt_page_mutex);
 		memset(xlt, 0, size);
+		use_emergency_page = true;
 	}
 	pages_iter = size / desc_size;
 	dma = dma_map_single(ddev, xlt, size, DMA_TO_DEVICE);
@@ -1088,8 +1123,8 @@ int mlx5_ib_update_xlt(struct mlx5_ib_mr *mr, u64 idx, int npages,
 	dma_unmap_single(ddev, dma, size, DMA_TO_DEVICE);
 
 free_xlt:
-	if (uctx)
-		mutex_unlock(&uctx->upd_xlt_page_mutex);
+	if (use_emergency_page)
+		mlx5_ib_put_xlt_emergency_page();
 	else
 		free_pages((unsigned long)xlt, get_order(size));
 
@@ -1141,7 +1176,7 @@ static struct mlx5_ib_mr *reg_create(struct ib_mr *ibmr, struct ib_pd *pd,
 
 	mkc = MLX5_ADDR_OF(create_mkey_in, in, memory_key_mkey_entry);
 	MLX5_SET(mkc, mkc, free, !populate);
-	MLX5_SET(mkc, mkc, access_mode, MLX5_MKC_ACCESS_MODE_MTT);
+	MLX5_SET(mkc, mkc, access_mode_1_0, MLX5_MKC_ACCESS_MODE_MTT);
 	MLX5_SET(mkc, mkc, a, !!(access_flags & IB_ACCESS_REMOTE_ATOMIC));
 	MLX5_SET(mkc, mkc, rw, !!(access_flags & IB_ACCESS_REMOTE_WRITE));
 	MLX5_SET(mkc, mkc, rr, !!(access_flags & IB_ACCESS_REMOTE_READ));
@@ -1197,22 +1232,96 @@ static void set_mr_fileds(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr,
 	mr->access_flags = access_flags;
 }
 
+static struct ib_mr *mlx5_ib_get_memic_mr(struct ib_pd *pd, u64 memic_addr,
+					  u64 length, int acc)
+{
+	struct mlx5_ib_dev *dev = to_mdev(pd->device);
+	int inlen = MLX5_ST_SZ_BYTES(create_mkey_in);
+	struct mlx5_core_dev *mdev = dev->mdev;
+	struct mlx5_ib_mr *mr;
+	void *mkc;
+	u32 *in;
+	int err;
+
+	mr = kzalloc(sizeof(*mr), GFP_KERNEL);
+	if (!mr)
+		return ERR_PTR(-ENOMEM);
+
+	in = kzalloc(inlen, GFP_KERNEL);
+	if (!in) {
+		err = -ENOMEM;
+		goto err_free;
+	}
+
+	mkc = MLX5_ADDR_OF(create_mkey_in, in, memory_key_mkey_entry);
+
+	MLX5_SET(mkc, mkc, access_mode_1_0, MLX5_MKC_ACCESS_MODE_MEMIC & 0x3);
+	MLX5_SET(mkc, mkc, access_mode_4_2,
+		 (MLX5_MKC_ACCESS_MODE_MEMIC >> 2) & 0x7);
+	MLX5_SET(mkc, mkc, a, !!(acc & IB_ACCESS_REMOTE_ATOMIC));
+	MLX5_SET(mkc, mkc, rw, !!(acc & IB_ACCESS_REMOTE_WRITE));
+	MLX5_SET(mkc, mkc, rr, !!(acc & IB_ACCESS_REMOTE_READ));
+	MLX5_SET(mkc, mkc, lw, !!(acc & IB_ACCESS_LOCAL_WRITE));
+	MLX5_SET(mkc, mkc, lr, 1);
+
+	MLX5_SET64(mkc, mkc, len, length);
+	MLX5_SET(mkc, mkc, pd, to_mpd(pd)->pdn);
+	MLX5_SET(mkc, mkc, qpn, 0xffffff);
+	MLX5_SET64(mkc, mkc, start_addr,
+		   memic_addr - pci_resource_start(dev->mdev->pdev, 0));
+
+	err = mlx5_core_create_mkey(mdev, &mr->mmkey, in, inlen);
+	if (err)
+		goto err_in;
+
+	kfree(in);
+
+	mr->umem = NULL;
+	set_mr_fileds(dev, mr, 0, length, acc);
+
+	return &mr->ibmr;
+
+err_in:
+	kfree(in);
+
+err_free:
+	kfree(mr);
+
+	return ERR_PTR(err);
+}
+
+struct ib_mr *mlx5_ib_reg_dm_mr(struct ib_pd *pd, struct ib_dm *dm,
+				struct ib_dm_mr_attr *attr,
+				struct uverbs_attr_bundle *attrs)
+{
+	struct mlx5_ib_dm *mdm = to_mdm(dm);
+	u64 memic_addr;
+
+	if (attr->access_flags & ~MLX5_IB_DM_ALLOWED_ACCESS)
+		return ERR_PTR(-EINVAL);
+
+	memic_addr = mdm->dev_addr + attr->offset;
+
+	return mlx5_ib_get_memic_mr(pd, memic_addr, attr->length,
+				    attr->access_flags);
+}
+
 struct ib_mr *mlx5_ib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
 				  u64 virt_addr, int access_flags,
 				  struct ib_udata *udata)
 {
 	struct mlx5_ib_dev *dev = to_mdev(pd->device);
 	struct mlx5_ib_mr *mr = NULL;
+	bool populate_mtts = false;
 	struct ib_umem *umem;
 	int page_shift;
 	int npages;
 	int ncont;
 	int order;
 	int err;
-	bool use_umr = true;
 
 	if (!IS_ENABLED(CONFIG_INFINIBAND_USER_MEM))
-		return ERR_PTR(-EINVAL);
+		return ERR_PTR(-EOPNOTSUPP);
 
 	mlx5_ib_dbg(dev, "start 0x%llx, virt_addr 0x%llx, length 0x%llx, access_flags 0x%x\n",
 		    start, virt_addr, length, access_flags);
@@ -1224,6 +1333,8 @@ struct ib_mr *mlx5_ib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
 			return ERR_PTR(-EINVAL);
 
 		mr = mlx5_ib_alloc_implicit_mr(to_mpd(pd), access_flags);
+		if (IS_ERR(mr))
+			return ERR_CAST(mr);
 		return &mr->ibmr;
 	}
 #endif
@@ -1234,26 +1345,29 @@ struct ib_mr *mlx5_ib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
 	if (err < 0)
 		return ERR_PTR(err);
 
-	if (order <= mr_cache_max_order(dev)) {
+	if (use_umr(dev, order)) {
 		mr = alloc_mr_from_cache(pd, umem, virt_addr, length, ncont,
 					 page_shift, order, access_flags);
 		if (PTR_ERR(mr) == -EAGAIN) {
 			mlx5_ib_dbg(dev, "cache empty for order %d\n", order);
 			mr = NULL;
 		}
+		populate_mtts = false;
 	} else if (!MLX5_CAP_GEN(dev->mdev, umr_extended_translation_offset)) {
 		if (access_flags & IB_ACCESS_ON_DEMAND) {
 			err = -EINVAL;
 			pr_err("Got MR registration for ODP MR > 512MB, not supported for Connect-IB\n");
 			goto error;
 		}
-		use_umr = false;
+		populate_mtts = true;
 	}
 
 	if (!mr) {
+		if (!umr_can_modify_entity_size(dev))
+			populate_mtts = true;
 		mutex_lock(&dev->slow_path_mutex);
 		mr = reg_create(NULL, pd, virt_addr, length, umem, ncont,
-				page_shift, access_flags, !use_umr);
+				page_shift, access_flags, populate_mtts);
 		mutex_unlock(&dev->slow_path_mutex);
 	}
 
@@ -1271,7 +1385,7 @@ struct ib_mr *mlx5_ib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
 	update_odp_mr(mr);
 #endif
 
-	if (use_umr) {
+	if (!populate_mtts) {
 		int update_xlt_flags = MLX5_IB_UPD_XLT_ENABLE;
 
 		if (access_flags & IB_ACCESS_ON_DEMAND)
@@ -1286,7 +1400,9 @@ struct ib_mr *mlx5_ib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
 		}
 	}
 
+#ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
 	mr->live = 1;
+#endif
 	return &mr->ibmr;
 error:
 	ib_umem_release(umem);
@@ -1365,36 +1481,34 @@ int mlx5_ib_rereg_user_mr(struct ib_mr *ib_mr, int flags, u64 start,
 		ib_umem_release(mr->umem);
 		err = mr_umem_get(pd, addr, len, access_flags, &mr->umem,
 				  &npages, &page_shift, &ncont, &order);
-		if (err < 0) {
-			clean_mr(dev, mr);
-			return err;
-		}
+		if (err)
+			goto err;
 	}
 
 	if (flags & IB_MR_REREG_TRANS && !use_umr_mtt_update(mr, addr, len)) {
 		/*
 		 * UMR can't be used - MKey needs to be replaced.
 		 */
-		if (mr->allocated_from_cache) {
+		if (mr->allocated_from_cache)
 			err = unreg_umr(dev, mr);
-			if (err)
-				mlx5_ib_warn(dev, "Failed to unregister MR\n");
-		} else {
+		else
 			err = destroy_mkey(dev, mr);
-			if (err)
-				mlx5_ib_warn(dev, "Failed to destroy MKey\n");
-		}
 		if (err)
-			return err;
+			goto err;
 
 		mr = reg_create(ib_mr, pd, addr, len, mr->umem, ncont,
 				page_shift, access_flags, true);
 
-		if (IS_ERR(mr))
-			return PTR_ERR(mr);
+		if (IS_ERR(mr)) {
+			err = PTR_ERR(mr);
+			mr = to_mmr(ib_mr);
+			goto err;
+		}
 
 		mr->allocated_from_cache = 0;
+#ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
 		mr->live = 1;
+#endif
 	} else {
 		/*
 		 * Send a UMR WQE
@@ -1417,13 +1531,8 @@ int mlx5_ib_rereg_user_mr(struct ib_mr *ib_mr, int flags, u64 start,
 			err = rereg_umr(pd, mr, access_flags, flags);
 		}
 
-		if (err) {
-			mlx5_ib_warn(dev, "Failed to rereg UMR\n");
-			ib_umem_release(mr->umem);
-			mr->umem = NULL;
-			clean_mr(dev, mr);
-			return err;
-		}
+		if (err)
+			goto err;
 	}
 
 	set_mr_fileds(dev, mr, npages, len, access_flags);
@@ -1432,6 +1541,14 @@ int mlx5_ib_rereg_user_mr(struct ib_mr *ib_mr, int flags, u64 start,
 	update_odp_mr(mr);
 #endif
 	return 0;
+
+err:
+	if (mr->umem) {
+		ib_umem_release(mr->umem);
+		mr->umem = NULL;
+	}
+	clean_mr(dev, mr);
+	return err;
 }
 
 static int
@@ -1480,10 +1597,9 @@ mlx5_free_priv_descs(struct mlx5_ib_mr *mr)
 	}
 }
 
-static int clean_mr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr)
+static void clean_mr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr)
 {
 	int allocated_from_cache = mr->allocated_from_cache;
-	int err;
 
 	if (mr->sig) {
 		if (mlx5_core_destroy_psv(dev->mdev,
@@ -1500,21 +1616,11 @@ static int clean_mr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr)
 
 	mlx5_free_priv_descs(mr);
 
-	if (!allocated_from_cache) {
-		u32 key = mr->mmkey.key;
-
-		err = destroy_mkey(dev, mr);
-		if (err) {
-			mlx5_ib_warn(dev, "failed to destroy mkey 0x%x (%d)\n",
-				     key, err);
-			return err;
-		}
-	}
-
-	return 0;
+	if (!allocated_from_cache)
+		destroy_mkey(dev, mr);
 }
 
-static int dereg_mr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr)
+static void dereg_mr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr)
 {
 	int npages = mr->npages;
 	struct ib_umem *umem = mr->umem;
@@ -1555,16 +1661,12 @@ static int dereg_mr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr)
 		kfree(mr);
 	else
 		mlx5_mr_cache_free(dev, mr);
-
-	return 0;
 }
 
 int mlx5_ib_dereg_mr(struct ib_mr *ibmr)
 {
-	struct mlx5_ib_dev *dev = to_mdev(ibmr->device);
-	struct mlx5_ib_mr *mr = to_mmr(ibmr);
-
-	return dereg_mr(dev, mr);
+	dereg_mr(to_mdev(ibmr->device), to_mmr(ibmr));
+	return 0;
 }
 
 struct ib_mr *mlx5_ib_alloc_mr(struct ib_pd *pd,
@@ -1645,7 +1747,8 @@ struct ib_mr *mlx5_ib_alloc_mr(struct ib_pd *pd,
 		goto err_free_in;
 	}
 
-	MLX5_SET(mkc, mkc, access_mode, mr->access_mode);
+	MLX5_SET(mkc, mkc, access_mode_1_0, mr->access_mode & 0x3);
+	MLX5_SET(mkc, mkc, access_mode_4_2, (mr->access_mode >> 2) & 0x7);
 	MLX5_SET(mkc, mkc, umr_en, 1);
 
 	mr->ibmr.device = pd->device;
@@ -1726,7 +1829,7 @@ struct ib_mw *mlx5_ib_alloc_mw(struct ib_pd *pd, enum ib_mw_type type,
 	MLX5_SET(mkc, mkc, pd, to_mpd(pd)->pdn);
 	MLX5_SET(mkc, mkc, umr_en, 1);
 	MLX5_SET(mkc, mkc, lr, 1);
-	MLX5_SET(mkc, mkc, access_mode, MLX5_MKC_ACCESS_MODE_KLMS);
+	MLX5_SET(mkc, mkc, access_mode_1_0, MLX5_MKC_ACCESS_MODE_KLMS);
 	MLX5_SET(mkc, mkc, en_rinval, !!((type == IB_MW_TYPE_2)));
 	MLX5_SET(mkc, mkc, qpn, 0xffffff);
 

drivers/infiniband/hw/mlx5/qp.c

@@ -86,7 +86,9 @@ struct mlx5_modify_raw_qp_param {
 	u16 operation;
 
 	u32 set_mask; /* raw_qp_set_mask_map */
-	u32 rate_limit;
+
+	struct mlx5_rate_limit rl;
+
 	u8 rq_q_ctr_id;
 };
 
@@ -878,7 +880,7 @@ static int create_user_qp(struct mlx5_ib_dev *dev, struct ib_pd *pd,
 		goto err_free;
 	}
 
-	err = ib_copy_to_udata(udata, resp, sizeof(*resp));
+	err = ib_copy_to_udata(udata, resp, min(udata->outlen, sizeof(*resp)));
 	if (err) {
 		mlx5_ib_dbg(dev, "copy failed\n");
 		goto err_unmap;
@@ -1411,6 +1413,7 @@ static int create_rss_raw_qp_tir(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp,
 	void *tirc;
 	void *hfso;
 	u32 selected_fields = 0;
+	u32 outer_l4;
 	size_t min_resp_len;
 	u32 tdn = mucontext->tdn;
 	struct mlx5_ib_create_qp_rss ucmd = {};
@@ -1466,7 +1469,7 @@ static int create_rss_raw_qp_tir(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp,
 		return -EOPNOTSUPP;
 	}
 
-	err = ib_copy_to_udata(udata, &resp, min_resp_len);
+	err = ib_copy_to_udata(udata, &resp, min(udata->outlen, sizeof(resp)));
 	if (err) {
 		mlx5_ib_dbg(dev, "copy failed\n");
 		return -EINVAL;
@@ -1541,10 +1544,14 @@ static int create_rss_raw_qp_tir(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp,
 		MLX5_SET(rx_hash_field_select, hfso, l3_prot_type,
 			 MLX5_L3_PROT_TYPE_IPV6);
 
-	if (((ucmd.rx_hash_fields_mask & MLX5_RX_HASH_SRC_PORT_TCP) ||
-	     (ucmd.rx_hash_fields_mask & MLX5_RX_HASH_DST_PORT_TCP)) &&
-	     ((ucmd.rx_hash_fields_mask & MLX5_RX_HASH_SRC_PORT_UDP) ||
-	     (ucmd.rx_hash_fields_mask & MLX5_RX_HASH_DST_PORT_UDP))) {
+	outer_l4 = ((ucmd.rx_hash_fields_mask & MLX5_RX_HASH_SRC_PORT_TCP) ||
+		    (ucmd.rx_hash_fields_mask & MLX5_RX_HASH_DST_PORT_TCP)) << 0 |
+		   ((ucmd.rx_hash_fields_mask & MLX5_RX_HASH_SRC_PORT_UDP) ||
+		    (ucmd.rx_hash_fields_mask & MLX5_RX_HASH_DST_PORT_UDP)) << 1 |
+		   (ucmd.rx_hash_fields_mask & MLX5_RX_HASH_IPSEC_SPI) << 2;
+
+	/* Check that only one l4 protocol is set */
+	if (outer_l4 & (outer_l4 - 1)) {
 		err = -EINVAL;
 		goto err;
 	}
@@ -1575,6 +1582,9 @@ static int create_rss_raw_qp_tir(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp,
 	    (ucmd.rx_hash_fields_mask & MLX5_RX_HASH_DST_PORT_UDP))
 		selected_fields |= MLX5_HASH_FIELD_SEL_L4_DPORT;
 
+	if (ucmd.rx_hash_fields_mask & MLX5_RX_HASH_IPSEC_SPI)
+		selected_fields |= MLX5_HASH_FIELD_SEL_IPSEC_SPI;
+
 	MLX5_SET(rx_hash_field_select, hfso, selected_fields, selected_fields);
 
 create_tir:
@@ -2774,8 +2784,9 @@ static int modify_raw_packet_qp_sq(struct mlx5_core_dev *dev,
 				   const struct mlx5_modify_raw_qp_param *raw_qp_param)
 {
 	struct mlx5_ib_qp *ibqp = sq->base.container_mibqp;
-	u32 old_rate = ibqp->rate_limit;
-	u32 new_rate = old_rate;
+	struct mlx5_rate_limit old_rl = ibqp->rl;
+	struct mlx5_rate_limit new_rl = old_rl;
+	bool new_rate_added = false;
 	u16 rl_index = 0;
 	void *in;
 	void *sqc;
@@ -2797,39 +2808,43 @@ static int modify_raw_packet_qp_sq(struct mlx5_core_dev *dev,
 			pr_warn("%s: Rate limit can only be changed when SQ is moving to RDY\n",
 				__func__);
 		else
-			new_rate = raw_qp_param->rate_limit;
+			new_rl = raw_qp_param->rl;
 	}
 
-	if (old_rate != new_rate) {
-		if (new_rate) {
-			err = mlx5_rl_add_rate(dev, new_rate, &rl_index);
+	if (!mlx5_rl_are_equal(&old_rl, &new_rl)) {
+		if (new_rl.rate) {
+			err = mlx5_rl_add_rate(dev, &rl_index, &new_rl);
 			if (err) {
-				pr_err("Failed configuring rate %u: %d\n",
-				       new_rate, err);
+				pr_err("Failed configuring rate limit(err %d): \
+				       rate %u, max_burst_sz %u, typical_pkt_sz %u\n",
+				       err, new_rl.rate, new_rl.max_burst_sz,
+				       new_rl.typical_pkt_sz);
+
 				goto out;
 			}
+			new_rate_added = true;
 		}
 
 		MLX5_SET64(modify_sq_in, in, modify_bitmask, 1);
+		/* index 0 means no limit */
 		MLX5_SET(sqc, sqc, packet_pacing_rate_limit_index, rl_index);
 	}
 
 	err = mlx5_core_modify_sq(dev, sq->base.mqp.qpn, in, inlen);
 	if (err) {
 		/* Remove new rate from table if failed */
-		if (new_rate &&
-		    old_rate != new_rate)
-			mlx5_rl_remove_rate(dev, new_rate);
+		if (new_rate_added)
+			mlx5_rl_remove_rate(dev, &new_rl);
 		goto out;
 	}
 
 	/* Only remove the old rate after new rate was set */
-	if ((old_rate &&
-	    (old_rate != new_rate)) ||
+	if ((old_rl.rate &&
+	     !mlx5_rl_are_equal(&old_rl, &new_rl)) ||
 	    (new_state != MLX5_SQC_STATE_RDY))
-		mlx5_rl_remove_rate(dev, old_rate);
+		mlx5_rl_remove_rate(dev, &old_rl);
 
-	ibqp->rate_limit = new_rate;
+	ibqp->rl = new_rl;
 	sq->state = new_state;
 
 out:
@@ -2906,7 +2921,8 @@ static int modify_raw_packet_qp(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp,
 
 static int __mlx5_ib_modify_qp(struct ib_qp *ibqp,
 			       const struct ib_qp_attr *attr, int attr_mask,
-			       enum ib_qp_state cur_state, enum ib_qp_state new_state)
+			       enum ib_qp_state cur_state, enum ib_qp_state new_state,
+			       const struct mlx5_ib_modify_qp *ucmd)
 {
 	static const u16 optab[MLX5_QP_NUM_STATE][MLX5_QP_NUM_STATE] = {
 		[MLX5_QP_STATE_RST] = {
@@ -2959,18 +2975,16 @@ static int __mlx5_ib_modify_qp(struct ib_qp *ibqp,
 	u16 op;
 	u8 tx_affinity = 0;
 
+	mlx5_st = to_mlx5_st(ibqp->qp_type == IB_QPT_DRIVER ?
+			     qp->qp_sub_type : ibqp->qp_type);
+	if (mlx5_st < 0)
+		return -EINVAL;
+
 	context = kzalloc(sizeof(*context), GFP_KERNEL);
 	if (!context)
 		return -ENOMEM;
 
-	err = to_mlx5_st(ibqp->qp_type == IB_QPT_DRIVER ?
-			 qp->qp_sub_type : ibqp->qp_type);
-	if (err < 0) {
-		mlx5_ib_dbg(dev, "unsupported qp type %d\n", ibqp->qp_type);
-		goto out;
-	}
-
-	context->flags = cpu_to_be32(err << 16);
+	context->flags = cpu_to_be32(mlx5_st << 16);
 
 	if (!(attr_mask & IB_QP_PATH_MIG_STATE)) {
 		context->flags |= cpu_to_be32(MLX5_QP_PM_MIGRATED << 11);
@@ -3124,10 +3138,6 @@ static int __mlx5_ib_modify_qp(struct ib_qp *ibqp,
 
 	mlx5_cur = to_mlx5_state(cur_state);
 	mlx5_new = to_mlx5_state(new_state);
-	mlx5_st = to_mlx5_st(ibqp->qp_type == IB_QPT_DRIVER ?
-			     qp->qp_sub_type : ibqp->qp_type);
-	if (mlx5_st < 0)
-		goto out;
 
 	if (mlx5_cur >= MLX5_QP_NUM_STATE || mlx5_new >= MLX5_QP_NUM_STATE ||
 	    !optab[mlx5_cur][mlx5_new]) {
@@ -3150,7 +3160,30 @@ static int __mlx5_ib_modify_qp(struct ib_qp *ibqp,
 		}
 
 		if (attr_mask & IB_QP_RATE_LIMIT) {
-			raw_qp_param.rate_limit = attr->rate_limit;
+			raw_qp_param.rl.rate = attr->rate_limit;
+
+			if (ucmd->burst_info.max_burst_sz) {
+				if (attr->rate_limit &&
+				    MLX5_CAP_QOS(dev->mdev, packet_pacing_burst_bound)) {
+					raw_qp_param.rl.max_burst_sz =
+						ucmd->burst_info.max_burst_sz;
+				} else {
+					err = -EINVAL;
+					goto out;
+				}
+			}
+
+			if (ucmd->burst_info.typical_pkt_sz) {
+				if (attr->rate_limit &&
+				    MLX5_CAP_QOS(dev->mdev, packet_pacing_typical_size)) {
+					raw_qp_param.rl.typical_pkt_sz =
+						ucmd->burst_info.typical_pkt_sz;
+				} else {
+					err = -EINVAL;
+					goto out;
+				}
+			}
+
 			raw_qp_param.set_mask |= MLX5_RAW_QP_RATE_LIMIT;
 		}
 
@@ -3178,7 +3211,8 @@ static int __mlx5_ib_modify_qp(struct ib_qp *ibqp,
 	 * If we moved a kernel QP to RESET, clean up all old CQ
 	 * entries and reinitialize the QP.
 	 */
-	if (new_state == IB_QPS_RESET && !ibqp->uobject) {
+	if (new_state == IB_QPS_RESET &&
+	    !ibqp->uobject && ibqp->qp_type != IB_QPT_XRC_TGT) {
 		mlx5_ib_cq_clean(recv_cq, base->mqp.qpn,
 				 ibqp->srq ? to_msrq(ibqp->srq) : NULL);
 		if (send_cq != recv_cq)
@@ -3337,8 +3371,10 @@ int mlx5_ib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
 {
 	struct mlx5_ib_dev *dev = to_mdev(ibqp->device);
 	struct mlx5_ib_qp *qp = to_mqp(ibqp);
+	struct mlx5_ib_modify_qp ucmd = {};
 	enum ib_qp_type qp_type;
 	enum ib_qp_state cur_state, new_state;
+	size_t required_cmd_sz;
 	int err = -EINVAL;
 	int port;
 	enum rdma_link_layer ll = IB_LINK_LAYER_UNSPECIFIED;
@@ -3346,6 +3382,28 @@ int mlx5_ib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
 	if (ibqp->rwq_ind_tbl)
 		return -ENOSYS;
 
+	if (udata && udata->inlen) {
+		required_cmd_sz = offsetof(typeof(ucmd), reserved) +
+			sizeof(ucmd.reserved);
+		if (udata->inlen < required_cmd_sz)
+			return -EINVAL;
+
+		if (udata->inlen > sizeof(ucmd) &&
+		    !ib_is_udata_cleared(udata, sizeof(ucmd),
+					 udata->inlen - sizeof(ucmd)))
+			return -EOPNOTSUPP;
+
+		if (ib_copy_from_udata(&ucmd, udata,
+				       min(udata->inlen, sizeof(ucmd))))
+			return -EFAULT;
+
+		if (ucmd.comp_mask ||
+		    memchr_inv(&ucmd.reserved, 0, sizeof(ucmd.reserved)) ||
+		    memchr_inv(&ucmd.burst_info.reserved, 0,
+			       sizeof(ucmd.burst_info.reserved)))
+			return -EOPNOTSUPP;
+	}
+
 	if (unlikely(ibqp->qp_type == IB_QPT_GSI))
 		return mlx5_ib_gsi_modify_qp(ibqp, attr, attr_mask);
 
@@ -3426,7 +3484,8 @@ int mlx5_ib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
 		goto out;
 	}
 
-	err = __mlx5_ib_modify_qp(ibqp, attr, attr_mask, cur_state, new_state);
+	err = __mlx5_ib_modify_qp(ibqp, attr, attr_mask, cur_state,
+				  new_state, &ucmd);
 
 out:
 	mutex_unlock(&qp->mutex);
@@ -3646,8 +3705,19 @@ static __be64 get_umr_update_pd_mask(void)
 	return cpu_to_be64(result);
 }
 
-static void set_reg_umr_segment(struct mlx5_wqe_umr_ctrl_seg *umr,
-				struct ib_send_wr *wr, int atomic)
+static int umr_check_mkey_mask(struct mlx5_ib_dev *dev, u64 mask)
+{
+	if ((mask & MLX5_MKEY_MASK_PAGE_SIZE &&
+	     MLX5_CAP_GEN(dev->mdev, umr_modify_entity_size_disabled)) ||
+	    (mask & MLX5_MKEY_MASK_A &&
+	     MLX5_CAP_GEN(dev->mdev, umr_modify_atomic_disabled)))
+		return -EPERM;
+	return 0;
+}
+
+static int set_reg_umr_segment(struct mlx5_ib_dev *dev,
+			       struct mlx5_wqe_umr_ctrl_seg *umr,
+			       struct ib_send_wr *wr, int atomic)
 {
 	struct mlx5_umr_wr *umrwr = umr_wr(wr);
 
@@ -3679,6 +3749,8 @@ static void set_reg_umr_segment(struct mlx5_wqe_umr_ctrl_seg *umr,
 
 	if (!wr->num_sge)
 		umr->flags |= MLX5_UMR_INLINE;
+
+	return umr_check_mkey_mask(dev, be64_to_cpu(umr->mkey_mask));
 }
 
 static u8 get_umr_flags(int acc)
@@ -4501,7 +4573,9 @@ int mlx5_ib_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
 			}
 			qp->sq.wr_data[idx] = MLX5_IB_WR_UMR;
 			ctrl->imm = cpu_to_be32(umr_wr(wr)->mkey);
-			set_reg_umr_segment(seg, wr, !!(MLX5_CAP_GEN(mdev, atomic)));
+			err = set_reg_umr_segment(dev, seg, wr, !!(MLX5_CAP_GEN(mdev, atomic)));
+			if (unlikely(err))
+				goto out;
 			seg += sizeof(struct mlx5_wqe_umr_ctrl_seg);
 			size += sizeof(struct mlx5_wqe_umr_ctrl_seg) / 16;
 			if (unlikely((seg == qend)))

drivers/infiniband/hw/mthca/mthca_provider.c

@@ -1295,6 +1295,7 @@ int mthca_register_device(struct mthca_dev *dev)
 
 	mutex_init(&dev->cap_mask_mutex);
 
+	dev->ib_dev.driver_id = RDMA_DRIVER_MTHCA;
 	ret = ib_register_device(&dev->ib_dev, NULL);
 	if (ret)
 		return ret;

drivers/infiniband/hw/nes/nes_verbs.c

@@ -3854,6 +3854,7 @@ int nes_register_ofa_device(struct nes_ib_device *nesibdev)
 	struct nes_adapter *nesadapter = nesdev->nesadapter;
 	int i, ret;
 
+	nesvnic->nesibdev->ibdev.driver_id = RDMA_DRIVER_NES;
 	ret = ib_register_device(&nesvnic->nesibdev->ibdev, NULL);
 	if (ret) {
 		return ret;

drivers/infiniband/hw/ocrdma/ocrdma_ah.c

@@ -193,11 +193,9 @@ struct ib_ah *ocrdma_create_ah(struct ib_pd *ibpd, struct rdma_ah_attr *attr,
 		      __func__, status);
 		goto av_conf_err;
 	}
-	if (sgid_attr.ndev) {
-		if (is_vlan_dev(sgid_attr.ndev))
-			vlan_tag = vlan_dev_vlan_id(sgid_attr.ndev);
-		dev_put(sgid_attr.ndev);
-	}
+	if (is_vlan_dev(sgid_attr.ndev))
+		vlan_tag = vlan_dev_vlan_id(sgid_attr.ndev);
+	dev_put(sgid_attr.ndev);
 	/* Get network header type for this GID */
 	ah->hdr_type = ib_gid_to_network_type(sgid_attr.gid_type, &sgid);
 

drivers/infiniband/hw/ocrdma/ocrdma_hw.c

@@ -2014,7 +2014,7 @@ static int ocrdma_mbx_reg_mr_cont(struct ocrdma_dev *dev,
 				  struct ocrdma_hw_mr *hwmr, u32 pbl_cnt,
 				  u32 pbl_offset, u32 last)
 {
-	int status = -ENOMEM;
+	int status;
 	int i;
 	struct ocrdma_reg_nsmr_cont *cmd;
 
@@ -2033,9 +2033,7 @@ static int ocrdma_mbx_reg_mr_cont(struct ocrdma_dev *dev,
 		    upper_32_bits(hwmr->pbl_table[i + pbl_offset].pa);
 	}
 	status = ocrdma_mbx_cmd(dev, (struct ocrdma_mqe *)cmd);
-	if (status)
-		goto mbx_err;
-mbx_err:
+
 	kfree(cmd);
 	return status;
 }
@@ -2496,7 +2494,7 @@ static int ocrdma_set_av_params(struct ocrdma_qp *qp,
 {
 	int status;
 	struct rdma_ah_attr *ah_attr = &attrs->ah_attr;
-	union ib_gid sgid, zgid;
+	union ib_gid sgid;
 	struct ib_gid_attr sgid_attr;
 	u32 vlan_id = 0xFFFF;
 	u8 mac_addr[6], hdr_type;
@@ -2529,16 +2527,12 @@ static int ocrdma_set_av_params(struct ocrdma_qp *qp,
 
 	status = ib_get_cached_gid(&dev->ibdev, 1, grh->sgid_index,
 				   &sgid, &sgid_attr);
-	if (!status && sgid_attr.ndev) {
+	if (!status) {
 		vlan_id = rdma_vlan_dev_vlan_id(sgid_attr.ndev);
 		memcpy(mac_addr, sgid_attr.ndev->dev_addr, ETH_ALEN);
 		dev_put(sgid_attr.ndev);
 	}
 
-	memset(&zgid, 0, sizeof(zgid));
-	if (!memcmp(&sgid, &zgid, sizeof(zgid)))
-		return -EINVAL;
-
 	qp->sgid_idx = grh->sgid_index;
 	memcpy(&cmd->params.sgid[0], &sgid.raw[0], sizeof(cmd->params.sgid));
 	status = ocrdma_resolve_dmac(dev, ah_attr, &mac_addr[0]);
@@ -3133,12 +3127,12 @@ done:
 static int ocrdma_mbx_modify_eqd(struct ocrdma_dev *dev, struct ocrdma_eq *eq,
 				 int num)
 {
-	int i, status = -ENOMEM;
+	int i, status;
 	struct ocrdma_modify_eqd_req *cmd;
 
 	cmd = ocrdma_init_emb_mqe(OCRDMA_CMD_MODIFY_EQ_DELAY, sizeof(*cmd));
 	if (!cmd)
-		return status;
+		return -ENOMEM;
 
 	ocrdma_init_mch(&cmd->cmd.req, OCRDMA_CMD_MODIFY_EQ_DELAY,
 			OCRDMA_SUBSYS_COMMON, sizeof(*cmd));
@@ -3151,9 +3145,7 @@ static int ocrdma_mbx_modify_eqd(struct ocrdma_dev *dev, struct ocrdma_eq *eq,
 				(eq[i].aic_obj.prev_eqd * 65)/100;
 	}
 	status = ocrdma_mbx_cmd(dev, (struct ocrdma_mqe *)cmd);
-	if (status)
-		goto mbx_err;
-mbx_err:
+
 	kfree(cmd);
 	return status;
 }

drivers/infiniband/hw/ocrdma/ocrdma_main.c

@@ -158,10 +158,7 @@ static int ocrdma_register_device(struct ocrdma_dev *dev)
 	dev->ibdev.query_device = ocrdma_query_device;
 	dev->ibdev.query_port = ocrdma_query_port;
 	dev->ibdev.modify_port = ocrdma_modify_port;
-	dev->ibdev.query_gid = ocrdma_query_gid;
 	dev->ibdev.get_netdev = ocrdma_get_netdev;
-	dev->ibdev.add_gid = ocrdma_add_gid;
-	dev->ibdev.del_gid = ocrdma_del_gid;
 	dev->ibdev.get_link_layer = ocrdma_link_layer;
 	dev->ibdev.alloc_pd = ocrdma_alloc_pd;
 	dev->ibdev.dealloc_pd = ocrdma_dealloc_pd;
@@ -217,6 +214,7 @@ static int ocrdma_register_device(struct ocrdma_dev *dev)
 		dev->ibdev.destroy_srq = ocrdma_destroy_srq;
 		dev->ibdev.post_srq_recv = ocrdma_post_srq_recv;
 	}
+	dev->ibdev.driver_id = RDMA_DRIVER_OCRDMA;
 	return ib_register_device(&dev->ibdev, NULL);
 }
 

drivers/infiniband/hw/ocrdma/ocrdma_verbs.c

@@ -62,40 +62,6 @@ int ocrdma_query_pkey(struct ib_device *ibdev, u8 port, u16 index, u16 *pkey)
 	return 0;
 }
 
-int ocrdma_query_gid(struct ib_device *ibdev, u8 port,
-		     int index, union ib_gid *sgid)
-{
-	int ret;
-
-	memset(sgid, 0, sizeof(*sgid));
-	if (index >= OCRDMA_MAX_SGID)
-		return -EINVAL;
-
-	ret = ib_get_cached_gid(ibdev, port, index, sgid, NULL);
-	if (ret == -EAGAIN) {
-		memcpy(sgid, &zgid, sizeof(*sgid));
-		return 0;
-	}
-
-	return ret;
-}
-
-int ocrdma_add_gid(struct ib_device *device,
-		   u8 port_num,
-		   unsigned int index,
-		   const union ib_gid *gid,
-		   const struct ib_gid_attr *attr,
-		   void **context) {
-	return  0;
-}
-
-int  ocrdma_del_gid(struct ib_device *device,
-		    u8 port_num,
-		    unsigned int index,
-		    void **context) {
-	return 0;
-}
-
 int ocrdma_query_device(struct ib_device *ibdev, struct ib_device_attr *attr,
 			struct ib_udata *uhw)
 {

drivers/infiniband/hw/ocrdma/ocrdma_verbs.h

@@ -61,19 +61,7 @@ enum rdma_protocol_type
 ocrdma_query_protocol(struct ib_device *device, u8 port_num);
 
 void ocrdma_get_guid(struct ocrdma_dev *, u8 *guid);
-int ocrdma_query_gid(struct ib_device *, u8 port,
-		     int index, union ib_gid *gid);
 struct net_device *ocrdma_get_netdev(struct ib_device *device, u8 port_num);
-int ocrdma_add_gid(struct ib_device *device,
-		   u8 port_num,
-		   unsigned int index,
-		   const union ib_gid *gid,
-		   const struct ib_gid_attr *attr,
-		   void **context);
-int  ocrdma_del_gid(struct ib_device *device,
-		    u8 port_num,
-		    unsigned int index,
-		    void **context);
 int ocrdma_query_pkey(struct ib_device *, u8 port, u16 index, u16 *pkey);
 
 struct ib_ucontext *ocrdma_alloc_ucontext(struct ib_device *,

drivers/infiniband/hw/qedr/main.c

@@ -162,10 +162,6 @@ static int qedr_iw_register_device(struct qedr_dev *dev)
 static void qedr_roce_register_device(struct qedr_dev *dev)
 {
 	dev->ibdev.node_type = RDMA_NODE_IB_CA;
-	dev->ibdev.query_gid = qedr_query_gid;
-
-	dev->ibdev.add_gid = qedr_add_gid;
-	dev->ibdev.del_gid = qedr_del_gid;
 
 	dev->ibdev.get_port_immutable = qedr_roce_port_immutable;
 }
@@ -257,6 +253,7 @@ static int qedr_register_device(struct qedr_dev *dev)
 	dev->ibdev.get_link_layer = qedr_link_layer;
 	dev->ibdev.get_dev_fw_str = qedr_get_dev_fw_str;
 
+	dev->ibdev.driver_id = RDMA_DRIVER_QEDR;
 	return ib_register_device(&dev->ibdev, NULL);
 }
 
@@ -707,7 +704,7 @@ static void qedr_affiliated_event(void *context, u8 e_code, void *fw_handle)
 			     "Error: CQ event with NULL pointer ibcq. Handle=%llx\n",
 			     roce_handle64);
 		}
-		DP_ERR(dev, "CQ event %d on hanlde %p\n", e_code, cq);
+		DP_ERR(dev, "CQ event %d on handle %p\n", e_code, cq);
 		break;
 	case EVENT_TYPE_QP:
 		qp = (struct qedr_qp *)(uintptr_t)roce_handle64;
@@ -723,7 +720,7 @@ static void qedr_affiliated_event(void *context, u8 e_code, void *fw_handle)
 			     "Error: QP event with NULL pointer ibqp. Handle=%llx\n",
 			     roce_handle64);
 		}
-		DP_ERR(dev, "QP event %d on hanlde %p\n", e_code, qp);
+		DP_ERR(dev, "QP event %d on handle %p\n", e_code, qp);
 		break;
 	default:
 		break;

drivers/infiniband/hw/qedr/qedr_roce_cm.c

@@ -412,19 +412,11 @@ static inline int qedr_gsi_build_header(struct qedr_dev *dev,
 		return rc;
 	}
 
-	if (sgid_attr.ndev) {
-		vlan_id = rdma_vlan_dev_vlan_id(sgid_attr.ndev);
-		if (vlan_id < VLAN_CFI_MASK)
-			has_vlan = true;
+	vlan_id = rdma_vlan_dev_vlan_id(sgid_attr.ndev);
+	if (vlan_id < VLAN_CFI_MASK)
+		has_vlan = true;
 
-		dev_put(sgid_attr.ndev);
-	}
-
-	if (!memcmp(&sgid, &zgid, sizeof(sgid))) {
-		DP_ERR(dev, "gsi post send: GID not found GID index %d\n",
-		       grh->sgid_index);
-		return -ENOENT;
-	}
+	dev_put(sgid_attr.ndev);
 
 	has_udp = (sgid_attr.gid_type == IB_GID_TYPE_ROCE_UDP_ENCAP);
 	if (!has_udp) {

drivers/infiniband/hw/qedr/verbs.c

@@ -84,58 +84,6 @@ int qedr_iw_query_gid(struct ib_device *ibdev, u8 port,
 	return 0;
 }
 
-int qedr_query_gid(struct ib_device *ibdev, u8 port, int index,
-		   union ib_gid *sgid)
-{
-	struct qedr_dev *dev = get_qedr_dev(ibdev);
-	int rc = 0;
-
-	if (!rdma_cap_roce_gid_table(ibdev, port))
-		return -ENODEV;
-
-	rc = ib_get_cached_gid(ibdev, port, index, sgid, NULL);
-	if (rc == -EAGAIN) {
-		memcpy(sgid, &zgid, sizeof(*sgid));
-		return 0;
-	}
-
-	DP_DEBUG(dev, QEDR_MSG_INIT, "query gid: index=%d %llx:%llx\n", index,
-		 sgid->global.interface_id, sgid->global.subnet_prefix);
-
-	return rc;
-}
-
-int qedr_add_gid(struct ib_device *device, u8 port_num,
-		 unsigned int index, const union ib_gid *gid,
-		 const struct ib_gid_attr *attr, void **context)
-{
-	if (!rdma_cap_roce_gid_table(device, port_num))
-		return -EINVAL;
-
-	if (port_num > QEDR_MAX_PORT)
-		return -EINVAL;
-
-	if (!context)
-		return -EINVAL;
-
-	return 0;
-}
-
-int qedr_del_gid(struct ib_device *device, u8 port_num,
-		 unsigned int index, void **context)
-{
-	if (!rdma_cap_roce_gid_table(device, port_num))
-		return -EINVAL;
-
-	if (port_num > QEDR_MAX_PORT)
-		return -EINVAL;
-
-	if (!context)
-		return -EINVAL;
-
-	return 0;
-}
-
 int qedr_query_device(struct ib_device *ibdev,
 		      struct ib_device_attr *attr, struct ib_udata *udata)
 {
@@ -525,9 +473,9 @@ struct ib_pd *qedr_alloc_pd(struct ib_device *ibdev,
 	pd->pd_id = pd_id;
 
 	if (udata && context) {
-		struct qedr_alloc_pd_uresp uresp;
-
-		uresp.pd_id = pd_id;
+		struct qedr_alloc_pd_uresp uresp = {
+			.pd_id = pd_id,
+		};
 
 		rc = qedr_ib_copy_to_udata(udata, &uresp, sizeof(uresp));
 		if (rc) {
@@ -856,8 +804,6 @@ static inline void qedr_init_cq_params(struct qedr_cq *cq,
 
 static void doorbell_cq(struct qedr_cq *cq, u32 cons, u8 flags)
 {
-	/* Flush data before signalling doorbell */
-	wmb();
 	cq->db.data.agg_flags = flags;
 	cq->db.data.value = cpu_to_le32(cons);
 	writeq(cq->db.raw, cq->db_addr);
@@ -1145,46 +1091,41 @@ static inline int get_gid_info_from_table(struct ib_qp *ibqp,
 	if (rc)
 		return rc;
 
-	if (!memcmp(&gid, &zgid, sizeof(gid)))
-		return -ENOENT;
-
-	if (gid_attr.ndev) {
-		qp_params->vlan_id = rdma_vlan_dev_vlan_id(gid_attr.ndev);
-
-		dev_put(gid_attr.ndev);
-		nw_type = ib_gid_to_network_type(gid_attr.gid_type, &gid);
-		switch (nw_type) {
-		case RDMA_NETWORK_IPV6:
-			memcpy(&qp_params->sgid.bytes[0], &gid.raw[0],
-			       sizeof(qp_params->sgid));
-			memcpy(&qp_params->dgid.bytes[0],
-			       &grh->dgid,
-			       sizeof(qp_params->dgid));
-			qp_params->roce_mode = ROCE_V2_IPV6;
-			SET_FIELD(qp_params->modify_flags,
-				  QED_ROCE_MODIFY_QP_VALID_ROCE_MODE, 1);
-			break;
-		case RDMA_NETWORK_IB:
-			memcpy(&qp_params->sgid.bytes[0], &gid.raw[0],
-			       sizeof(qp_params->sgid));
-			memcpy(&qp_params->dgid.bytes[0],
-			       &grh->dgid,
-			       sizeof(qp_params->dgid));
-			qp_params->roce_mode = ROCE_V1;
-			break;
-		case RDMA_NETWORK_IPV4:
-			memset(&qp_params->sgid, 0, sizeof(qp_params->sgid));
-			memset(&qp_params->dgid, 0, sizeof(qp_params->dgid));
-			ipv4_addr = qedr_get_ipv4_from_gid(gid.raw);
-			qp_params->sgid.ipv4_addr = ipv4_addr;
-			ipv4_addr =
-			    qedr_get_ipv4_from_gid(grh->dgid.raw);
-			qp_params->dgid.ipv4_addr = ipv4_addr;
-			SET_FIELD(qp_params->modify_flags,
-				  QED_ROCE_MODIFY_QP_VALID_ROCE_MODE, 1);
-			qp_params->roce_mode = ROCE_V2_IPV4;
-			break;
-		}
+	qp_params->vlan_id = rdma_vlan_dev_vlan_id(gid_attr.ndev);
+
+	dev_put(gid_attr.ndev);
+	nw_type = ib_gid_to_network_type(gid_attr.gid_type, &gid);
+	switch (nw_type) {
+	case RDMA_NETWORK_IPV6:
+		memcpy(&qp_params->sgid.bytes[0], &gid.raw[0],
+		       sizeof(qp_params->sgid));
+		memcpy(&qp_params->dgid.bytes[0],
+		       &grh->dgid,
+		       sizeof(qp_params->dgid));
+		qp_params->roce_mode = ROCE_V2_IPV6;
+		SET_FIELD(qp_params->modify_flags,
+			  QED_ROCE_MODIFY_QP_VALID_ROCE_MODE, 1);
+		break;
+	case RDMA_NETWORK_IB:
+		memcpy(&qp_params->sgid.bytes[0], &gid.raw[0],
+		       sizeof(qp_params->sgid));
+		memcpy(&qp_params->dgid.bytes[0],
+		       &grh->dgid,
+		       sizeof(qp_params->dgid));
+		qp_params->roce_mode = ROCE_V1;
+		break;
+	case RDMA_NETWORK_IPV4:
+		memset(&qp_params->sgid, 0, sizeof(qp_params->sgid));
+		memset(&qp_params->dgid, 0, sizeof(qp_params->dgid));
+		ipv4_addr = qedr_get_ipv4_from_gid(gid.raw);
+		qp_params->sgid.ipv4_addr = ipv4_addr;
+		ipv4_addr =
+		    qedr_get_ipv4_from_gid(grh->dgid.raw);
+		qp_params->dgid.ipv4_addr = ipv4_addr;
+		SET_FIELD(qp_params->modify_flags,
+			  QED_ROCE_MODIFY_QP_VALID_ROCE_MODE, 1);
+		qp_params->roce_mode = ROCE_V2_IPV4;
+		break;
 	}
 
 	for (i = 0; i < 4; i++) {
@@ -1870,7 +1811,6 @@ static int qedr_update_qp_state(struct qedr_dev *dev,
 			 */
 
 			if (rdma_protocol_roce(&dev->ibdev, 1)) {
-				wmb();
 				writel(qp->rq.db_data.raw, qp->rq.db);
 				/* Make sure write takes effect */
 				mmiowb();
@@ -3274,8 +3214,15 @@ int qedr_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
 	 * vane. However this is not harmful (as long as the producer value is
 	 * unchanged). For performance reasons we avoid checking for this
 	 * redundant doorbell.
+	 *
+	 * qp->wqe_wr_id is accessed during qedr_poll_cq, as
+	 * soon as we give the doorbell, we could get a completion
+	 * for this wr, therefore we need to make sure that the
+	 * memory is updated before giving the doorbell.
+	 * During qedr_poll_cq, rmb is called before accessing the
+	 * cqe. This covers for the smp_rmb as well.
 	 */
-	wmb();
+	smp_wmb();
 	writel(qp->sq.db_data.raw, qp->sq.db);
 
 	/* Make sure write sticks */
@@ -3362,8 +3309,14 @@ int qedr_post_recv(struct ib_qp *ibqp, struct ib_recv_wr *wr,
 
 		qedr_inc_sw_prod(&qp->rq);
 
-		/* Flush all the writes before signalling doorbell */
-		wmb();
+		/* qp->rqe_wr_id is accessed during qedr_poll_cq, as
+		 * soon as we give the doorbell, we could get a completion
+		 * for this wr, therefore we need to make sure that the
+		 * memory is update before giving the doorbell.
+		 * During qedr_poll_cq, rmb is called before accessing the
+		 * cqe. This covers for the smp_rmb as well.
+		 */
+		smp_wmb();
 
 		qp->rq.db_data.data.value++;
 

drivers/infiniband/hw/qedr/verbs.h

@@ -38,7 +38,6 @@ int qedr_query_port(struct ib_device *, u8 port, struct ib_port_attr *props);
 int qedr_modify_port(struct ib_device *, u8 port, int mask,
 		     struct ib_port_modify *props);
 
-int qedr_query_gid(struct ib_device *, u8 port, int index, union ib_gid *gid);
 int qedr_iw_query_gid(struct ib_device *ibdev, u8 port,
 		      int index, union ib_gid *gid);
 
@@ -48,11 +47,6 @@ struct ib_ucontext *qedr_alloc_ucontext(struct ib_device *, struct ib_udata *);
 int qedr_dealloc_ucontext(struct ib_ucontext *);
 
 int qedr_mmap(struct ib_ucontext *, struct vm_area_struct *vma);
-int qedr_del_gid(struct ib_device *device, u8 port_num,
-		 unsigned int index, void **context);
-int qedr_add_gid(struct ib_device *device, u8 port_num,
-		 unsigned int index, const union ib_gid *gid,
-		 const struct ib_gid_attr *attr, void **context);
 struct ib_pd *qedr_alloc_pd(struct ib_device *,
 			    struct ib_ucontext *, struct ib_udata *);
 int qedr_dealloc_pd(struct ib_pd *pd);

drivers/infiniband/hw/qib/qib.h

@@ -472,9 +472,6 @@ enum qib_sdma_events {
 	qib_sdma_event_e90_timer_tick,
 };
 
-extern char *qib_sdma_state_names[];
-extern char *qib_sdma_event_names[];
-
 struct sdma_set_state_action {
 	unsigned op_enable:1;
 	unsigned op_intenable:1;

drivers/infiniband/hw/qib/qib_diag.c

@@ -614,7 +614,7 @@ static ssize_t qib_diagpkt_write(struct file *fp,
 	}
 
 	if (copy_from_user(tmpbuf,
-			   (const void __user *) (unsigned long) dp.data,
+			   u64_to_user_ptr(dp.data),
 			   dp.len)) {
 		ret = -EFAULT;
 		goto bail;

drivers/infiniband/hw/qib/qib_file_ops.c

@@ -443,7 +443,7 @@ cleanup:
 			ret = -EFAULT;
 			goto cleanup;
 		}
-		if (copy_to_user((void __user *) (unsigned long) ti->tidmap,
+		if (copy_to_user(u64_to_user_ptr(ti->tidmap),
 				 tidmap, sizeof(tidmap))) {
 			ret = -EFAULT;
 			goto cleanup;
@@ -490,7 +490,7 @@ static int qib_tid_free(struct qib_ctxtdata *rcd, unsigned subctxt,
 		goto done;
 	}
 
-	if (copy_from_user(tidmap, (void __user *)(unsigned long)ti->tidmap,
+	if (copy_from_user(tidmap, u64_to_user_ptr(ti->tidmap),
 			   sizeof(tidmap))) {
 		ret = -EFAULT;
 		goto done;
@@ -2168,8 +2168,8 @@ static ssize_t qib_write(struct file *fp, const char __user *data,
 		ret = qib_do_user_init(fp, &cmd.cmd.user_info);
 		if (ret)
 			goto bail;
-		ret = qib_get_base_info(fp, (void __user *) (unsigned long)
-					cmd.cmd.user_info.spu_base_info,
+		ret = qib_get_base_info(fp, u64_to_user_ptr(
+					  cmd.cmd.user_info.spu_base_info),
 					cmd.cmd.user_info.spu_base_info_size);
 		break;
 

drivers/infiniband/hw/qib/qib_iba7322.c

@@ -463,6 +463,16 @@ static u8 ib_rate_to_delay[IB_RATE_120_GBPS + 1] = {
 	[IB_RATE_40_GBPS] = 1
 };
 
+static const char * const qib_sdma_state_names[] = {
+	[qib_sdma_state_s00_hw_down]          = "s00_HwDown",
+	[qib_sdma_state_s10_hw_start_up_wait] = "s10_HwStartUpWait",
+	[qib_sdma_state_s20_idle]             = "s20_Idle",
+	[qib_sdma_state_s30_sw_clean_up_wait] = "s30_SwCleanUpWait",
+	[qib_sdma_state_s40_hw_clean_up_wait] = "s40_HwCleanUpWait",
+	[qib_sdma_state_s50_hw_halt_wait]     = "s50_HwHaltWait",
+	[qib_sdma_state_s99_running]          = "s99_Running",
+};
+
 #define IBA7322_LINKSPEED_SHIFT SYM_LSB(IBCStatusA_0, LinkSpeedActive)
 #define IBA7322_LINKWIDTH_SHIFT SYM_LSB(IBCStatusA_0, LinkWidthActive)
 

drivers/infiniband/hw/qib/qib_init.c

@@ -678,11 +678,9 @@ int qib_init(struct qib_devdata *dd, int reinit)
 		lastfail = qib_create_rcvhdrq(dd, rcd);
 		if (!lastfail)
 			lastfail = qib_setup_eagerbufs(rcd);
-		if (lastfail) {
+		if (lastfail)
 			qib_dev_err(dd,
 				"failed to allocate kernel ctxt's rcvhdrq and/or egr bufs\n");
-			continue;
-		}
 	}
 
 	for (pidx = 0; pidx < dd->num_pports; ++pidx) {