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

Pull infiniband/rdma updates from Roland Dreier: - Re-enable flow steering verbs with new improved userspace ABI - Fixes for slow connection due to GID lookup scalability - IPoIB fixes - Many fixes to HW drivers including mlx4, mlx5, ocrdma and qib - Further improvements to SRP error handling - Add new transport type for Cisco usNIC * tag 'rdma-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband: (66 commits) IB/core: Re-enable create_flow/destroy_flow uverbs IB/core: extended command: an improved infrastructure for uverbs commands IB/core: Remove ib_uverbs_flow_spec structure from userspace IB/core: Use a common header for uverbs flow_specs IB/core: Make uverbs flow structure use names like verbs ones IB/core: Rename 'flow' structs to match other uverbs structs IB/core: clarify overflow/underflow checks on ib_create/destroy_flow IB/ucma: Convert use of typedef ctl_table to struct ctl_table IB/cm: Convert to using idr_alloc_cyclic() IB/mlx5: Fix page shift in create CQ for userspace IB/mlx4: Fix device max capabilities check IB/mlx5: Fix list_del of empty list IB/mlx5: Remove dead code IB/core: Encorce MR access rights rules on kernel consumers IB/mlx4: Fix endless loop in resize CQ RDMA/cma: Remove unused argument and minor dead code RDMA/ucma: Discard events for IDs not yet claimed by user space IB/core: Add Cisco usNIC rdma node and transport types RDMA/nes: Remove self-assignment from nes_query_qp() IB/srp: Report receive errors correctly ...
2026-05-01 15:00:59 -07:00 · 2013-11-18 15:36:04 -08:00
parent a709bd585f b4fdf52b3f
commit 1ea406c0e0
52 changed files with 1988 additions and 610 deletions
@@ -61,6 +61,12 @@ Description:	Interface for making ib_srp connect to a new target.
 		  interrupt is handled by a different CPU then the comp_vector
 		  parameter can be used to spread the SRP completion workload
 		  over multiple CPU's.
+		* tl_retry_count, a number in the range 2..7 specifying the
+		  IB RC retry count.
+		* queue_size, the maximum number of commands that the
+		  initiator is allowed to queue per SCSI host. The default
+		  value for this parameter is 62. The lowest supported value
+		  is 2.

 What:		/sys/class/infiniband_srp/srp-<hca>-<port_number>/ibdev
 Date:		January 2, 2006
@@ -153,6 +159,13 @@ Contact:	linux-rdma@vger.kernel.org
 Description:	InfiniBand service ID used for establishing communication with
 		the SRP	target.

+What:		/sys/class/scsi_host/host<n>/sgid
+Date:		February 1, 2014
+KernelVersion:	3.13
+Contact:	linux-rdma@vger.kernel.org
+Description:	InfiniBand GID of the source port used for communication with
+		the SRP target.
+
 What:		/sys/class/scsi_host/host<n>/zero_req_lim
 Date:		September 20, 2006
 KernelVersion:	2.6.18
@@ -5,6 +5,24 @@ Contact:	linux-scsi@vger.kernel.org, linux-rdma@vger.kernel.org
 Description:	Instructs an SRP initiator to disconnect from a target and to
 		remove all LUNs imported from that target.

+What:		/sys/class/srp_remote_ports/port-<h>:<n>/dev_loss_tmo
+Date:		February 1, 2014
+KernelVersion:	3.13
+Contact:	linux-scsi@vger.kernel.org, linux-rdma@vger.kernel.org
+Description:	Number of seconds the SCSI layer will wait after a transport
+		layer error has been observed before removing a target port.
+		Zero means immediate removal. Setting this attribute to "off"
+		will disable the dev_loss timer.
+
+What:		/sys/class/srp_remote_ports/port-<h>:<n>/fast_io_fail_tmo
+Date:		February 1, 2014
+KernelVersion:	3.13
+Contact:	linux-scsi@vger.kernel.org, linux-rdma@vger.kernel.org
+Description:	Number of seconds the SCSI layer will wait after a transport
+		layer error has been observed before failing I/O. Zero means
+		failing I/O immediately. Setting this attribute to "off" will
+		disable the fast_io_fail timer.
+
 What:		/sys/class/srp_remote_ports/port-<h>:<n>/port_id
 Date:		June 27, 2007
 KernelVersion:	2.6.24
@@ -12,8 +30,29 @@ Contact:	linux-scsi@vger.kernel.org
 Description:	16-byte local SRP port identifier in hexadecimal format. An
 		example: 4c:49:4e:55:58:20:56:49:4f:00:00:00:00:00:00:00.

+What:		/sys/class/srp_remote_ports/port-<h>:<n>/reconnect_delay
+Date:		February 1, 2014
+KernelVersion:	3.13
+Contact:	linux-scsi@vger.kernel.org, linux-rdma@vger.kernel.org
+Description:	Number of seconds the SCSI layer will wait after a reconnect
+		attempt failed before retrying. Setting this attribute to
+		"off" will disable time-based reconnecting.
+
 What:		/sys/class/srp_remote_ports/port-<h>:<n>/roles
 Date:		June 27, 2007
 KernelVersion:	2.6.24
 Contact:	linux-scsi@vger.kernel.org
 Description:	Role of the remote port. Either "SRP Initiator" or "SRP Target".
+
+What:		/sys/class/srp_remote_ports/port-<h>:<n>/state
+Date:		February 1, 2014
+KernelVersion:	3.13
+Contact:	linux-scsi@vger.kernel.org, linux-rdma@vger.kernel.org
+Description:	State of the transport layer used for communication with the
+		remote port. "running" if the transport layer is operational;
+		"blocked" if a transport layer error has been encountered but
+		the fast_io_fail_tmo timer has not yet fired; "fail-fast"
+		after the fast_io_fail_tmo timer has fired and before the
+		"dev_loss_tmo" timer has fired; "lost" after the
+		"dev_loss_tmo" timer has fired and before the port is finally
+		removed.
@@ -31,17 +31,6 @@ config INFINIBAND_USER_ACCESS
 	  libibverbs, libibcm and a hardware driver library from
 	  <http://www.openfabrics.org/git/>.

-config INFINIBAND_EXPERIMENTAL_UVERBS_FLOW_STEERING
-       bool "Experimental and unstable ABI for userspace access to flow steering verbs"
-       depends on INFINIBAND_USER_ACCESS
-       depends on STAGING
-	---help---
-	  The final ABI for userspace access to flow steering verbs
-	  has not been defined.  To use the current ABI, *WHICH WILL
-	  CHANGE IN THE FUTURE*, say Y here.
-
-	  If unsure, say N.
-
 config INFINIBAND_USER_MEM
 	bool
 	depends on INFINIBAND_USER_ACCESS != n
@@ -383,14 +383,11 @@ static int cm_alloc_id(struct cm_id_private *cm_id_priv)
 {
 	unsigned long flags;
 	int id;
-	static int next_id;

 	idr_preload(GFP_KERNEL);
 	spin_lock_irqsave(&cm.lock, flags);

-	id = idr_alloc(&cm.local_id_table, cm_id_priv, next_id, 0, GFP_NOWAIT);
-	if (id >= 0)
-		next_id = max(id + 1, 0);
+	id = idr_alloc_cyclic(&cm.local_id_table, cm_id_priv, 0, 0, GFP_NOWAIT);

 	spin_unlock_irqrestore(&cm.lock, flags);
 	idr_preload_end();
@@ -328,28 +328,6 @@ static int cma_set_qkey(struct rdma_id_private *id_priv, u32 qkey)
 	return ret;
 }

-static int find_gid_port(struct ib_device *device, union ib_gid *gid, u8 port_num)
-{
-	int i;
-	int err;
-	struct ib_port_attr props;
-	union ib_gid tmp;
-
-	err = ib_query_port(device, port_num, &props);
-	if (err)
-		return err;
-
-	for (i = 0; i < props.gid_tbl_len; ++i) {
-		err = ib_query_gid(device, port_num, i, &tmp);
-		if (err)
-			return err;
-		if (!memcmp(&tmp, gid, sizeof tmp))
-			return 0;
-	}
-
-	return -EADDRNOTAVAIL;
-}
-
 static void cma_translate_ib(struct sockaddr_ib *sib, struct rdma_dev_addr *dev_addr)
 {
 	dev_addr->dev_type = ARPHRD_INFINIBAND;
@@ -371,13 +349,14 @@ static int cma_translate_addr(struct sockaddr *addr, struct rdma_dev_addr *dev_a
 	return ret;
 }

-static int cma_acquire_dev(struct rdma_id_private *id_priv)
+static int cma_acquire_dev(struct rdma_id_private *id_priv,
+			   struct rdma_id_private *listen_id_priv)
 {
 	struct rdma_dev_addr *dev_addr = &id_priv->id.route.addr.dev_addr;
 	struct cma_device *cma_dev;
 	union ib_gid gid, iboe_gid;
 	int ret = -ENODEV;
-	u8 port;
+	u8 port, found_port;
 	enum rdma_link_layer dev_ll = dev_addr->dev_type == ARPHRD_INFINIBAND ?
 		IB_LINK_LAYER_INFINIBAND : IB_LINK_LAYER_ETHERNET;

@@ -389,17 +368,39 @@ static int cma_acquire_dev(struct rdma_id_private *id_priv)
 	iboe_addr_get_sgid(dev_addr, &iboe_gid);
 	memcpy(&gid, dev_addr->src_dev_addr +
 	       rdma_addr_gid_offset(dev_addr), sizeof gid);
+	if (listen_id_priv &&
+	    rdma_port_get_link_layer(listen_id_priv->id.device,
+				     listen_id_priv->id.port_num) == dev_ll) {
+		cma_dev = listen_id_priv->cma_dev;
+		port = listen_id_priv->id.port_num;
+		if (rdma_node_get_transport(cma_dev->device->node_type) == RDMA_TRANSPORT_IB &&
+		    rdma_port_get_link_layer(cma_dev->device, port) == IB_LINK_LAYER_ETHERNET)
+			ret = ib_find_cached_gid(cma_dev->device, &iboe_gid,
+						 &found_port, NULL);
+		else
+			ret = ib_find_cached_gid(cma_dev->device, &gid,
+						 &found_port, NULL);
+
+		if (!ret && (port  == found_port)) {
+			id_priv->id.port_num = found_port;
+			goto out;
+		}
+	}
 	list_for_each_entry(cma_dev, &dev_list, list) {
 		for (port = 1; port <= cma_dev->device->phys_port_cnt; ++port) {
+			if (listen_id_priv &&
+			    listen_id_priv->cma_dev == cma_dev &&
+			    listen_id_priv->id.port_num == port)
+				continue;
 			if (rdma_port_get_link_layer(cma_dev->device, port) == dev_ll) {
 				if (rdma_node_get_transport(cma_dev->device->node_type) == RDMA_TRANSPORT_IB &&
 				    rdma_port_get_link_layer(cma_dev->device, port) == IB_LINK_LAYER_ETHERNET)
-					ret = find_gid_port(cma_dev->device, &iboe_gid, port);
+					ret = ib_find_cached_gid(cma_dev->device, &iboe_gid, &found_port, NULL);
 				else
-					ret = find_gid_port(cma_dev->device, &gid, port);
+					ret = ib_find_cached_gid(cma_dev->device, &gid, &found_port, NULL);

-				if (!ret) {
-					id_priv->id.port_num = port;
+				if (!ret && (port == found_port)) {
+					id_priv->id.port_num = found_port;
 					goto out;
 				}
 			}
@@ -1292,7 +1293,7 @@ static int cma_req_handler(struct ib_cm_id *cm_id, struct ib_cm_event *ib_event)
 	}

 	mutex_lock_nested(&conn_id->handler_mutex, SINGLE_DEPTH_NESTING);
-	ret = cma_acquire_dev(conn_id);
+	ret = cma_acquire_dev(conn_id, listen_id);
 	if (ret)
 		goto err2;

@@ -1451,7 +1452,6 @@ static int iw_conn_req_handler(struct iw_cm_id *cm_id,
 {
 	struct rdma_cm_id *new_cm_id;
 	struct rdma_id_private *listen_id, *conn_id;
-	struct net_device *dev = NULL;
 	struct rdma_cm_event event;
 	int ret;
 	struct ib_device_attr attr;
@@ -1481,7 +1481,7 @@ static int iw_conn_req_handler(struct iw_cm_id *cm_id,
 		goto out;
 	}

-	ret = cma_acquire_dev(conn_id);
+	ret = cma_acquire_dev(conn_id, listen_id);
 	if (ret) {
 		mutex_unlock(&conn_id->handler_mutex);
 		rdma_destroy_id(new_cm_id);
@@ -1529,8 +1529,6 @@ static int iw_conn_req_handler(struct iw_cm_id *cm_id,
 	cma_deref_id(conn_id);

 out:
-	if (dev)
-		dev_put(dev);
 	mutex_unlock(&listen_id->handler_mutex);
 	return ret;
 }
@@ -2066,7 +2064,7 @@ static void addr_handler(int status, struct sockaddr *src_addr,
 		goto out;

 	if (!status && !id_priv->cma_dev)
-		status = cma_acquire_dev(id_priv);
+		status = cma_acquire_dev(id_priv, NULL);

 	if (status) {
 		if (!cma_comp_exch(id_priv, RDMA_CM_ADDR_RESOLVED,
@@ -2563,7 +2561,7 @@ int rdma_bind_addr(struct rdma_cm_id *id, struct sockaddr *addr)
 		if (ret)
 			goto err1;

-		ret = cma_acquire_dev(id_priv);
+		ret = cma_acquire_dev(id_priv, NULL);
 		if (ret)
 			goto err1;
 	}
@@ -148,7 +148,7 @@ static int ibnl_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
 	list_for_each_entry(client, &client_list, list) {
 		if (client->index == index) {
 			if (op < 0 || op >= client->nops ||
-			    !client->cb_table[RDMA_NL_GET_OP(op)].dump)
+			    !client->cb_table[op].dump)
 				return -EINVAL;

 			{
@@ -612,6 +612,7 @@ static ssize_t show_node_type(struct device *device,
 	switch (dev->node_type) {
 	case RDMA_NODE_IB_CA:	  return sprintf(buf, "%d: CA\n", dev->node_type);
 	case RDMA_NODE_RNIC:	  return sprintf(buf, "%d: RNIC\n", dev->node_type);
+	case RDMA_NODE_USNIC:	  return sprintf(buf, "%d: usNIC\n", dev->node_type);
 	case RDMA_NODE_IB_SWITCH: return sprintf(buf, "%d: switch\n", dev->node_type);
 	case RDMA_NODE_IB_ROUTER: return sprintf(buf, "%d: router\n", dev->node_type);
 	default:		  return sprintf(buf, "%d: <unknown>\n", dev->node_type);
@@ -57,7 +57,7 @@ MODULE_LICENSE("Dual BSD/GPL");
 static unsigned int max_backlog = 1024;

 static struct ctl_table_header *ucma_ctl_table_hdr;
-static ctl_table ucma_ctl_table[] = {
+static struct ctl_table ucma_ctl_table[] = {
 	{
 		.procname	= "max_backlog",
 		.data		= &max_backlog,
@@ -271,7 +271,7 @@ static int ucma_event_handler(struct rdma_cm_id *cm_id,
 			goto out;
 		}
 		ctx->backlog--;
-	} else if (!ctx->uid) {
+	} else if (!ctx->uid || ctx->cm_id != cm_id) {
 		/*
 		 * We ignore events for new connections until userspace has set
 		 * their context.  This can only happen if an error occurs on a
@@ -47,6 +47,14 @@
 #include <rdma/ib_umem.h>
 #include <rdma/ib_user_verbs.h>

+#define INIT_UDATA(udata, ibuf, obuf, ilen, olen)			\
+	do {								\
+		(udata)->inbuf  = (void __user *) (ibuf);		\
+		(udata)->outbuf = (void __user *) (obuf);		\
+		(udata)->inlen  = (ilen);				\
+		(udata)->outlen = (olen);				\
+	} while (0)
+
 /*
 * Our lifetime rules for these structs are the following:
 *
@@ -178,6 +186,22 @@ void ib_uverbs_event_handler(struct ib_event_handler *handler,
 			     struct ib_event *event);
 void ib_uverbs_dealloc_xrcd(struct ib_uverbs_device *dev, struct ib_xrcd *xrcd);

+struct ib_uverbs_flow_spec {
+	union {
+		union {
+			struct ib_uverbs_flow_spec_hdr hdr;
+			struct {
+				__u32 type;
+				__u16 size;
+				__u16 reserved;
+			};
+		};
+		struct ib_uverbs_flow_spec_eth     eth;
+		struct ib_uverbs_flow_spec_ipv4    ipv4;
+		struct ib_uverbs_flow_spec_tcp_udp tcp_udp;
+	};
+};
+
 #define IB_UVERBS_DECLARE_CMD(name)					\
 	ssize_t ib_uverbs_##name(struct ib_uverbs_file *file,		\
 				 const char __user *buf, int in_len,	\
@@ -217,9 +241,13 @@ IB_UVERBS_DECLARE_CMD(destroy_srq);
 IB_UVERBS_DECLARE_CMD(create_xsrq);
 IB_UVERBS_DECLARE_CMD(open_xrcd);
 IB_UVERBS_DECLARE_CMD(close_xrcd);
-#ifdef CONFIG_INFINIBAND_EXPERIMENTAL_UVERBS_FLOW_STEERING
-IB_UVERBS_DECLARE_CMD(create_flow);
-IB_UVERBS_DECLARE_CMD(destroy_flow);
-#endif /* CONFIG_INFINIBAND_EXPERIMENTAL_UVERBS_FLOW_STEERING */
+
+#define IB_UVERBS_DECLARE_EX_CMD(name)				\
+	int ib_uverbs_ex_##name(struct ib_uverbs_file *file,	\
+				struct ib_udata *ucore,		\
+				struct ib_udata *uhw)
+
+IB_UVERBS_DECLARE_EX_CMD(create_flow);
+IB_UVERBS_DECLARE_EX_CMD(destroy_flow);

 #endif /* UVERBS_H */
@@ -54,17 +54,7 @@ static struct uverbs_lock_class qp_lock_class	= { .name = "QP-uobj" };
 static struct uverbs_lock_class ah_lock_class	= { .name = "AH-uobj" };
 static struct uverbs_lock_class srq_lock_class	= { .name = "SRQ-uobj" };
 static struct uverbs_lock_class xrcd_lock_class = { .name = "XRCD-uobj" };
-#ifdef CONFIG_INFINIBAND_EXPERIMENTAL_UVERBS_FLOW_STEERING
 static struct uverbs_lock_class rule_lock_class = { .name = "RULE-uobj" };
-#endif /* CONFIG_INFINIBAND_EXPERIMENTAL_UVERBS_FLOW_STEERING */
-
-#define INIT_UDATA(udata, ibuf, obuf, ilen, olen)			\
-	do {								\
-		(udata)->inbuf  = (void __user *) (ibuf);		\
-		(udata)->outbuf = (void __user *) (obuf);		\
-		(udata)->inlen  = (ilen);				\
-		(udata)->outlen = (olen);				\
-	} while (0)

 /*
 * The ib_uobject locking scheme is as follows:
@@ -939,13 +929,9 @@ ssize_t ib_uverbs_reg_mr(struct ib_uverbs_file *file,
 	if ((cmd.start & ~PAGE_MASK) != (cmd.hca_va & ~PAGE_MASK))
 		return -EINVAL;

-	/*
-	 * Local write permission is required if remote write or
-	 * remote atomic permission is also requested.
-	 */
-	if (cmd.access_flags & (IB_ACCESS_REMOTE_ATOMIC | IB_ACCESS_REMOTE_WRITE) &&
-	    !(cmd.access_flags & IB_ACCESS_LOCAL_WRITE))
-		return -EINVAL;
+	ret = ib_check_mr_access(cmd.access_flags);
+	if (ret)
+		return ret;

 	uobj = kmalloc(sizeof *uobj, GFP_KERNEL);
 	if (!uobj)
@@ -2128,6 +2114,9 @@ ssize_t ib_uverbs_post_send(struct ib_uverbs_file *file,
 			}
 			next->wr.ud.remote_qpn  = user_wr->wr.ud.remote_qpn;
 			next->wr.ud.remote_qkey = user_wr->wr.ud.remote_qkey;
+			if (next->opcode == IB_WR_SEND_WITH_IMM)
+				next->ex.imm_data =
+					(__be32 __force) user_wr->ex.imm_data;
 		} else {
 			switch (next->opcode) {
 			case IB_WR_RDMA_WRITE_WITH_IMM:
@@ -2601,8 +2590,7 @@ out_put:
 	return ret ? ret : in_len;
 }

-#ifdef CONFIG_INFINIBAND_EXPERIMENTAL_UVERBS_FLOW_STEERING
-static int kern_spec_to_ib_spec(struct ib_kern_spec *kern_spec,
+static int kern_spec_to_ib_spec(struct ib_uverbs_flow_spec *kern_spec,
 				union ib_flow_spec *ib_spec)
 {
 	ib_spec->type = kern_spec->type;
@@ -2642,28 +2630,31 @@ static int kern_spec_to_ib_spec(struct ib_kern_spec *kern_spec,
 	return 0;
 }

-ssize_t ib_uverbs_create_flow(struct ib_uverbs_file *file,
-			      const char __user *buf, int in_len,
-			      int out_len)
+int ib_uverbs_ex_create_flow(struct ib_uverbs_file *file,
+			     struct ib_udata *ucore,
+			     struct ib_udata *uhw)
 {
 	struct ib_uverbs_create_flow	  cmd;
 	struct ib_uverbs_create_flow_resp resp;
 	struct ib_uobject		  *uobj;
 	struct ib_flow			  *flow_id;
-	struct ib_kern_flow_attr	  *kern_flow_attr;
+	struct ib_uverbs_flow_attr	  *kern_flow_attr;
 	struct ib_flow_attr		  *flow_attr;
 	struct ib_qp			  *qp;
 	int err = 0;
 	void *kern_spec;
 	void *ib_spec;
 	int i;
-	int kern_attr_size;

-	if (out_len < sizeof(resp))
+	if (ucore->outlen < sizeof(resp))
 		return -ENOSPC;

-	if (copy_from_user(&cmd, buf, sizeof(cmd)))
-		return -EFAULT;
+	err = ib_copy_from_udata(&cmd, ucore, sizeof(cmd));
+	if (err)
+		return err;
+
+	ucore->inbuf += sizeof(cmd);
+	ucore->inlen -= sizeof(cmd);

 	if (cmd.comp_mask)
 		return -EINVAL;
@@ -2672,32 +2663,27 @@ ssize_t ib_uverbs_create_flow(struct ib_uverbs_file *file,
 	     !capable(CAP_NET_ADMIN)) || !capable(CAP_NET_RAW))
 		return -EPERM;

-	if (cmd.flow_attr.num_of_specs < 0 ||
-	    cmd.flow_attr.num_of_specs > IB_FLOW_SPEC_SUPPORT_LAYERS)
+	if (cmd.flow_attr.num_of_specs > IB_FLOW_SPEC_SUPPORT_LAYERS)
 		return -EINVAL;

-	kern_attr_size = cmd.flow_attr.size - sizeof(cmd) -
-			 sizeof(struct ib_uverbs_cmd_hdr_ex);
-
-	if (cmd.flow_attr.size < 0 || cmd.flow_attr.size > in_len ||
-	    kern_attr_size < 0 || kern_attr_size >
-	    (cmd.flow_attr.num_of_specs * sizeof(struct ib_kern_spec)))
+	if (cmd.flow_attr.size > ucore->inlen ||
+	    cmd.flow_attr.size >
+	    (cmd.flow_attr.num_of_specs * sizeof(struct ib_uverbs_flow_spec)))
 		return -EINVAL;

 	if (cmd.flow_attr.num_of_specs) {
-		kern_flow_attr = kmalloc(cmd.flow_attr.size, GFP_KERNEL);
+		kern_flow_attr = kmalloc(sizeof(*kern_flow_attr) + cmd.flow_attr.size,
+					 GFP_KERNEL);
 		if (!kern_flow_attr)
 			return -ENOMEM;

 		memcpy(kern_flow_attr, &cmd.flow_attr, sizeof(*kern_flow_attr));
-		if (copy_from_user(kern_flow_attr + 1, buf + sizeof(cmd),
-				   kern_attr_size)) {
-			err = -EFAULT;
+		err = ib_copy_from_udata(kern_flow_attr + 1, ucore,
+					 cmd.flow_attr.size);
+		if (err)
 			goto err_free_attr;
-		}
 	} else {
 		kern_flow_attr = &cmd.flow_attr;
-		kern_attr_size = sizeof(cmd.flow_attr);
 	}

 	uobj = kmalloc(sizeof(*uobj), GFP_KERNEL);
@@ -2714,7 +2700,7 @@ ssize_t ib_uverbs_create_flow(struct ib_uverbs_file *file,
 		goto err_uobj;
 	}

-	flow_attr = kmalloc(cmd.flow_attr.size, GFP_KERNEL);
+	flow_attr = kmalloc(sizeof(*flow_attr) + cmd.flow_attr.size, GFP_KERNEL);
 	if (!flow_attr) {
 		err = -ENOMEM;
 		goto err_put;
@@ -2729,19 +2715,22 @@ ssize_t ib_uverbs_create_flow(struct ib_uverbs_file *file,

 	kern_spec = kern_flow_attr + 1;
 	ib_spec = flow_attr + 1;
-	for (i = 0; i < flow_attr->num_of_specs && kern_attr_size > 0; i++) {
+	for (i = 0; i < flow_attr->num_of_specs &&
+	     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);
 		if (err)
 			goto err_free;
 		flow_attr->size +=
 			((union ib_flow_spec *) ib_spec)->size;
-		kern_attr_size -= ((struct ib_kern_spec *) kern_spec)->size;
-		kern_spec += ((struct ib_kern_spec *) kern_spec)->size;
+		cmd.flow_attr.size -= ((struct ib_uverbs_flow_spec *)kern_spec)->size;
+		kern_spec += ((struct ib_uverbs_flow_spec *) kern_spec)->size;
 		ib_spec += ((union ib_flow_spec *) ib_spec)->size;
 	}
-	if (kern_attr_size) {
-		pr_warn("create flow failed, %d bytes left from uverb cmd\n",
-			kern_attr_size);
+	if (cmd.flow_attr.size || (i != flow_attr->num_of_specs)) {
+		pr_warn("create flow failed, flow %d: %d bytes left from uverb cmd\n",
+			i, cmd.flow_attr.size);
 		goto err_free;
 	}
 	flow_id = ib_create_flow(qp, flow_attr, IB_FLOW_DOMAIN_USER);
@@ -2760,11 +2749,10 @@ ssize_t ib_uverbs_create_flow(struct ib_uverbs_file *file,
 	memset(&resp, 0, sizeof(resp));
 	resp.flow_handle = uobj->id;

-	if (copy_to_user((void __user *)(unsigned long) cmd.response,
-			 &resp, sizeof(resp))) {
-		err = -EFAULT;
+	err = ib_copy_to_udata(ucore,
+			       &resp, sizeof(resp));
+	if (err)
 		goto err_copy;
-	}

 	put_qp_read(qp);
 	mutex_lock(&file->mutex);
@@ -2777,7 +2765,7 @@ ssize_t ib_uverbs_create_flow(struct ib_uverbs_file *file,
 	kfree(flow_attr);
 	if (cmd.flow_attr.num_of_specs)
 		kfree(kern_flow_attr);
-	return in_len;
+	return 0;
 err_copy:
 	idr_remove_uobj(&ib_uverbs_rule_idr, uobj);
 destroy_flow:
@@ -2794,16 +2782,18 @@ err_free_attr:
 	return err;
 }

-ssize_t ib_uverbs_destroy_flow(struct ib_uverbs_file *file,
-			       const char __user *buf, int in_len,
-			       int out_len) {
+int ib_uverbs_ex_destroy_flow(struct ib_uverbs_file *file,
+			      struct ib_udata *ucore,
+			      struct ib_udata *uhw)
+{
 	struct ib_uverbs_destroy_flow	cmd;
 	struct ib_flow			*flow_id;
 	struct ib_uobject		*uobj;
 	int				ret;

-	if (copy_from_user(&cmd, buf, sizeof(cmd)))
-		return -EFAULT;
+	ret = ib_copy_from_udata(&cmd, ucore, sizeof(cmd));
+	if (ret)
+		return ret;

 	uobj = idr_write_uobj(&ib_uverbs_rule_idr, cmd.flow_handle,
 			      file->ucontext);
@@ -2825,9 +2815,8 @@ ssize_t ib_uverbs_destroy_flow(struct ib_uverbs_file *file,

 	put_uobj(uobj);

-	return ret ? ret : in_len;
+	return ret;
 }
-#endif /* CONFIG_INFINIBAND_EXPERIMENTAL_UVERBS_FLOW_STEERING */

 static int __uverbs_create_xsrq(struct ib_uverbs_file *file,
 				struct ib_uverbs_create_xsrq *cmd,
@@ -115,10 +115,13 @@ static ssize_t (*uverbs_cmd_table[])(struct ib_uverbs_file *file,
 	[IB_USER_VERBS_CMD_CLOSE_XRCD]		= ib_uverbs_close_xrcd,
 	[IB_USER_VERBS_CMD_CREATE_XSRQ]		= ib_uverbs_create_xsrq,
 	[IB_USER_VERBS_CMD_OPEN_QP]		= ib_uverbs_open_qp,
-#ifdef CONFIG_INFINIBAND_EXPERIMENTAL_UVERBS_FLOW_STEERING
-	[IB_USER_VERBS_CMD_CREATE_FLOW]		= ib_uverbs_create_flow,
-	[IB_USER_VERBS_CMD_DESTROY_FLOW]	= ib_uverbs_destroy_flow
-#endif /* CONFIG_INFINIBAND_EXPERIMENTAL_UVERBS_FLOW_STEERING */
+};
+
+static int (*uverbs_ex_cmd_table[])(struct ib_uverbs_file *file,
+				    struct ib_udata *ucore,
+				    struct ib_udata *uhw) = {
+	[IB_USER_VERBS_EX_CMD_CREATE_FLOW]	= ib_uverbs_ex_create_flow,
+	[IB_USER_VERBS_EX_CMD_DESTROY_FLOW]	= ib_uverbs_ex_destroy_flow
 };

 static void ib_uverbs_add_one(struct ib_device *device);
@@ -589,6 +592,7 @@ static ssize_t ib_uverbs_write(struct file *filp, const char __user *buf,
 {
 	struct ib_uverbs_file *file = filp->private_data;
 	struct ib_uverbs_cmd_hdr hdr;
+	__u32 flags;

 	if (count < sizeof hdr)
 		return -EINVAL;
@@ -596,45 +600,105 @@ static ssize_t ib_uverbs_write(struct file *filp, const char __user *buf,
 	if (copy_from_user(&hdr, buf, sizeof hdr))
 		return -EFAULT;

-	if (hdr.command >= ARRAY_SIZE(uverbs_cmd_table) ||
-	    !uverbs_cmd_table[hdr.command])
-		return -EINVAL;
+	flags = (hdr.command &
+		 IB_USER_VERBS_CMD_FLAGS_MASK) >> IB_USER_VERBS_CMD_FLAGS_SHIFT;

-	if (!file->ucontext &&
-	    hdr.command != IB_USER_VERBS_CMD_GET_CONTEXT)
-		return -EINVAL;
+	if (!flags) {
+		__u32 command;

-	if (!(file->device->ib_dev->uverbs_cmd_mask & (1ull << hdr.command)))
-		return -ENOSYS;
-
-#ifdef CONFIG_INFINIBAND_EXPERIMENTAL_UVERBS_FLOW_STEERING
-	if (hdr.command >= IB_USER_VERBS_CMD_THRESHOLD) {
-		struct ib_uverbs_cmd_hdr_ex hdr_ex;
-
-		if (copy_from_user(&hdr_ex, buf, sizeof(hdr_ex)))
-			return -EFAULT;
-
-		if (((hdr_ex.in_words + hdr_ex.provider_in_words) * 4) != count)
+		if (hdr.command & ~(__u32)(IB_USER_VERBS_CMD_FLAGS_MASK |
+					   IB_USER_VERBS_CMD_COMMAND_MASK))
 			return -EINVAL;

-		return uverbs_cmd_table[hdr.command](file,
-						     buf + sizeof(hdr_ex),
-						     (hdr_ex.in_words +
-						      hdr_ex.provider_in_words) * 4,
-						     (hdr_ex.out_words +
-						      hdr_ex.provider_out_words) * 4);
-	} else {
-#endif /* CONFIG_INFINIBAND_EXPERIMENTAL_UVERBS_FLOW_STEERING */
+		command = hdr.command & IB_USER_VERBS_CMD_COMMAND_MASK;
+
+		if (command >= ARRAY_SIZE(uverbs_cmd_table) ||
+		    !uverbs_cmd_table[command])
+			return -EINVAL;
+
+		if (!file->ucontext &&
+		    command != IB_USER_VERBS_CMD_GET_CONTEXT)
+			return -EINVAL;
+
+		if (!(file->device->ib_dev->uverbs_cmd_mask & (1ull << command)))
+			return -ENOSYS;
+
 		if (hdr.in_words * 4 != count)
 			return -EINVAL;

-		return uverbs_cmd_table[hdr.command](file,
-						     buf + sizeof(hdr),
-						     hdr.in_words * 4,
-						     hdr.out_words * 4);
-#ifdef CONFIG_INFINIBAND_EXPERIMENTAL_UVERBS_FLOW_STEERING
+		return uverbs_cmd_table[command](file,
+						 buf + sizeof(hdr),
+						 hdr.in_words * 4,
+						 hdr.out_words * 4);
+
+	} else if (flags == IB_USER_VERBS_CMD_FLAG_EXTENDED) {
+		__u32 command;
+
+		struct ib_uverbs_ex_cmd_hdr ex_hdr;
+		struct ib_udata ucore;
+		struct ib_udata uhw;
+		int err;
+		size_t written_count = count;
+
+		if (hdr.command & ~(__u32)(IB_USER_VERBS_CMD_FLAGS_MASK |
+					   IB_USER_VERBS_CMD_COMMAND_MASK))
+			return -EINVAL;
+
+		command = hdr.command & IB_USER_VERBS_CMD_COMMAND_MASK;
+
+		if (command >= ARRAY_SIZE(uverbs_ex_cmd_table) ||
+		    !uverbs_ex_cmd_table[command])
+			return -ENOSYS;
+
+		if (!file->ucontext)
+			return -EINVAL;
+
+		if (!(file->device->ib_dev->uverbs_ex_cmd_mask & (1ull << command)))
+			return -ENOSYS;
+
+		if (count < (sizeof(hdr) + sizeof(ex_hdr)))
+			return -EINVAL;
+
+		if (copy_from_user(&ex_hdr, buf + sizeof(hdr), sizeof(ex_hdr)))
+			return -EFAULT;
+
+		count -= sizeof(hdr) + sizeof(ex_hdr);
+		buf += sizeof(hdr) + sizeof(ex_hdr);
+
+		if ((hdr.in_words + ex_hdr.provider_in_words) * 8 != count)
+			return -EINVAL;
+
+		if (ex_hdr.response) {
+			if (!hdr.out_words && !ex_hdr.provider_out_words)
+				return -EINVAL;
+		} else {
+			if (hdr.out_words || ex_hdr.provider_out_words)
+				return -EINVAL;
+		}
+
+		INIT_UDATA(&ucore,
+			   (hdr.in_words) ? buf : 0,
+			   (unsigned long)ex_hdr.response,
+			   hdr.in_words * 8,
+			   hdr.out_words * 8);
+
+		INIT_UDATA(&uhw,
+			   (ex_hdr.provider_in_words) ? buf + ucore.inlen : 0,
+			   (ex_hdr.provider_out_words) ? (unsigned long)ex_hdr.response + ucore.outlen : 0,
+			   ex_hdr.provider_in_words * 8,
+			   ex_hdr.provider_out_words * 8);
+
+		err = uverbs_ex_cmd_table[command](file,
+						   &ucore,
+						   &uhw);
+
+		if (err)
+			return err;
+
+		return written_count;
 	}
-#endif /* CONFIG_INFINIBAND_EXPERIMENTAL_UVERBS_FLOW_STEERING */
+
+	return -ENOSYS;
 }

 static int ib_uverbs_mmap(struct file *filp, struct vm_area_struct *vma)
@@ -114,6 +114,8 @@ rdma_node_get_transport(enum rdma_node_type node_type)
 		return RDMA_TRANSPORT_IB;
 	case RDMA_NODE_RNIC:
 		return RDMA_TRANSPORT_IWARP;
+	case RDMA_NODE_USNIC:
+		return RDMA_TRANSPORT_USNIC;
 	default:
 		BUG();
 		return 0;
@@ -130,6 +132,7 @@ enum rdma_link_layer rdma_port_get_link_layer(struct ib_device *device, u8 port_
 	case RDMA_TRANSPORT_IB:
 		return IB_LINK_LAYER_INFINIBAND;
 	case RDMA_TRANSPORT_IWARP:
+	case RDMA_TRANSPORT_USNIC:
 		return IB_LINK_LAYER_ETHERNET;
 	default:
 		return IB_LINK_LAYER_UNSPECIFIED;
@@ -958,6 +961,11 @@ EXPORT_SYMBOL(ib_resize_cq);
 struct ib_mr *ib_get_dma_mr(struct ib_pd *pd, int mr_access_flags)
 {
 	struct ib_mr *mr;
+	int err;
+
+	err = ib_check_mr_access(mr_access_flags);
+	if (err)
+		return ERR_PTR(err);

 	mr = pd->device->get_dma_mr(pd, mr_access_flags);

@@ -980,6 +988,11 @@ struct ib_mr *ib_reg_phys_mr(struct ib_pd *pd,
 			     u64 *iova_start)
 {
 	struct ib_mr *mr;
+	int err;
+
+	err = ib_check_mr_access(mr_access_flags);
+	if (err)
+		return ERR_PTR(err);

 	if (!pd->device->reg_phys_mr)
 		return ERR_PTR(-ENOSYS);
@@ -1010,6 +1023,10 @@ int ib_rereg_phys_mr(struct ib_mr *mr,
 	struct ib_pd *old_pd;
 	int ret;

+	ret = ib_check_mr_access(mr_access_flags);
+	if (ret)
+		return ret;
+
 	if (!mr->device->rereg_phys_mr)
 		return -ENOSYS;

@@ -602,10 +602,10 @@ static int c4iw_rdev_open(struct c4iw_rdev *rdev)
 	     rdev->lldi.vr->qp.size,
 	     rdev->lldi.vr->cq.start,
 	     rdev->lldi.vr->cq.size);
-	PDBG("udb len 0x%x udb base %p db_reg %p gts_reg %p qpshift %lu "
+	PDBG("udb len 0x%x udb base %llx db_reg %p gts_reg %p qpshift %lu "
 	     "qpmask 0x%x cqshift %lu cqmask 0x%x\n",
 	     (unsigned)pci_resource_len(rdev->lldi.pdev, 2),
-	     (void *)(unsigned long)pci_resource_start(rdev->lldi.pdev, 2),
+	     (u64)pci_resource_start(rdev->lldi.pdev, 2),
 	     rdev->lldi.db_reg,
 	     rdev->lldi.gts_reg,
 	     rdev->qpshift, rdev->qpmask,
@@ -280,9 +280,7 @@ static int ipath_user_sdma_pin_pages(const struct ipath_devdata *dd,
 	int j;
 	int ret;

-	ret = get_user_pages(current, current->mm, addr,
-			     npages, 0, 1, pages, NULL);
-
+	ret = get_user_pages_fast(addr, npages, 0, pages);
 	if (ret != npages) {
 		int i;

@@ -811,10 +809,7 @@ int ipath_user_sdma_writev(struct ipath_devdata *dd,
 	while (dim) {
 		const int mxp = 8;

-		down_write(&current->mm->mmap_sem);
 		ret = ipath_user_sdma_queue_pkts(dd, pq, &list, iov, dim, mxp);
-		up_write(&current->mm->mmap_sem);
-
 		if (ret <= 0)
 			goto done_unlock;
 		else {
@@ -324,7 +324,7 @@ static int mlx4_ib_get_outstanding_cqes(struct mlx4_ib_cq *cq)
 	u32 i;

 	i = cq->mcq.cons_index;
-	while (get_sw_cqe(cq, i & cq->ibcq.cqe))
+	while (get_sw_cqe(cq, i))
 		++i;

 	return i - cq->mcq.cons_index;
@@ -365,7 +365,7 @@ int mlx4_ib_resize_cq(struct ib_cq *ibcq, int entries, struct ib_udata *udata)

 	mutex_lock(&cq->resize_mutex);

-	if (entries < 1 || entries > dev->dev->caps.max_cqes) {
+	if (entries < 1) {
 		err = -EINVAL;
 		goto out;
 	}
@@ -376,6 +376,11 @@ int mlx4_ib_resize_cq(struct ib_cq *ibcq, int entries, struct ib_udata *udata)
 		goto out;
 	}

+	if (entries > dev->dev->caps.max_cqes) {
+		err = -EINVAL;
+		goto out;
+	}
+
 	if (ibcq->uobject) {
 		err = mlx4_alloc_resize_umem(dev, cq, entries, udata);
 		if (err)
@@ -1685,11 +1685,9 @@ static void *mlx4_ib_add(struct mlx4_dev *dev)
 		ibdev->ib_dev.create_flow	= mlx4_ib_create_flow;
 		ibdev->ib_dev.destroy_flow	= mlx4_ib_destroy_flow;

-#ifdef CONFIG_INFINIBAND_EXPERIMENTAL_UVERBS_FLOW_STEERING
-		ibdev->ib_dev.uverbs_cmd_mask	|=
-			(1ull << IB_USER_VERBS_CMD_CREATE_FLOW) |
-			(1ull << IB_USER_VERBS_CMD_DESTROY_FLOW);
-#endif /* CONFIG_INFINIBAND_EXPERIMENTAL_UVERBS_FLOW_STEERING */
+		ibdev->ib_dev.uverbs_ex_cmd_mask	|=
+			(1ull << IB_USER_VERBS_EX_CMD_CREATE_FLOW) |
+			(1ull << IB_USER_VERBS_EX_CMD_DESTROY_FLOW);
 	}

 	mlx4_ib_alloc_eqs(dev, ibdev);
@@ -556,7 +556,7 @@ static int create_cq_user(struct mlx5_ib_dev *dev, struct ib_udata *udata,
 		goto err_db;
 	}
 	mlx5_ib_populate_pas(dev, cq->buf.umem, page_shift, (*cqb)->pas, 0);
-	(*cqb)->ctx.log_pg_sz = page_shift - PAGE_SHIFT;
+	(*cqb)->ctx.log_pg_sz = page_shift - MLX5_ADAPTER_PAGE_SHIFT;

 	*index = to_mucontext(context)->uuari.uars[0].index;

@@ -620,7 +620,7 @@ static int create_cq_kernel(struct mlx5_ib_dev *dev, struct mlx5_ib_cq *cq,
 	}
 	mlx5_fill_page_array(&cq->buf.buf, (*cqb)->pas);

-	(*cqb)->ctx.log_pg_sz = cq->buf.buf.page_shift - PAGE_SHIFT;
+	(*cqb)->ctx.log_pg_sz = cq->buf.buf.page_shift - MLX5_ADAPTER_PAGE_SHIFT;
 	*index = dev->mdev.priv.uuari.uars[0].index;

 	return 0;
@@ -653,8 +653,11 @@ struct ib_cq *mlx5_ib_create_cq(struct ib_device *ibdev, int entries,
 	int eqn;
 	int err;

+	if (entries < 0)
+		return ERR_PTR(-EINVAL);
+
 	entries = roundup_pow_of_two(entries + 1);
-	if (entries < 1 || entries > dev->mdev.caps.max_cqes)
+	if (entries > dev->mdev.caps.max_cqes)
 		return ERR_PTR(-EINVAL);

 	cq = kzalloc(sizeof(*cq), GFP_KERNEL);
@@ -747,17 +750,9 @@ int mlx5_ib_destroy_cq(struct ib_cq *cq)
 	return 0;
 }

-static int is_equal_rsn(struct mlx5_cqe64 *cqe64, struct mlx5_ib_srq *srq,
-			u32 rsn)
+static int is_equal_rsn(struct mlx5_cqe64 *cqe64, u32 rsn)
 {
-	u32 lrsn;
-
-	if (srq)
-		lrsn = be32_to_cpu(cqe64->srqn) & 0xffffff;
-	else
-		lrsn = be32_to_cpu(cqe64->sop_drop_qpn) & 0xffffff;
-
-	return rsn == lrsn;
+	return rsn == (ntohl(cqe64->sop_drop_qpn) & 0xffffff);
 }

 void __mlx5_ib_cq_clean(struct mlx5_ib_cq *cq, u32 rsn, struct mlx5_ib_srq *srq)
@@ -787,8 +782,8 @@ void __mlx5_ib_cq_clean(struct mlx5_ib_cq *cq, u32 rsn, struct mlx5_ib_srq *srq)
 	while ((int) --prod_index - (int) cq->mcq.cons_index >= 0) {
 		cqe = get_cqe(cq, prod_index & cq->ibcq.cqe);
 		cqe64 = (cq->mcq.cqe_sz == 64) ? cqe : cqe + 64;
-		if (is_equal_rsn(cqe64, srq, rsn)) {
-			if (srq)
+		if (is_equal_rsn(cqe64, rsn)) {
+			if (srq && (ntohl(cqe64->srqn) & 0xffffff))
 				mlx5_ib_free_srq_wqe(srq, be16_to_cpu(cqe64->wqe_counter));
 			++nfreed;
 		} else if (nfreed) {
@@ -745,7 +745,8 @@ static int alloc_pa_mkey(struct mlx5_ib_dev *dev, u32 *key, u32 pdn)
 	seg->qpn_mkey7_0 = cpu_to_be32(0xffffff << 8);
 	seg->start_addr = 0;

-	err = mlx5_core_create_mkey(&dev->mdev, &mr, in, sizeof(*in));
+	err = mlx5_core_create_mkey(&dev->mdev, &mr, in, sizeof(*in),
+				    NULL, NULL, NULL);
 	if (err) {
 		mlx5_ib_warn(dev, "failed to create mkey, %d\n", err);
 		goto err_in;
@@ -262,6 +262,9 @@ struct mlx5_ib_mr {
 	int			npages;
 	struct completion	done;
 	enum ib_wc_status	status;
+	struct mlx5_ib_dev     *dev;
+	struct mlx5_create_mkey_mbox_out out;
+	unsigned long		start;
 };

 struct mlx5_ib_fast_reg_page_list {
@@ -323,6 +326,7 @@ struct mlx5_cache_ent {
 	struct mlx5_ib_dev     *dev;
 	struct work_struct	work;
 	struct delayed_work	dwork;
+	int			pending;
 };

 struct mlx5_mr_cache {
@@ -358,6 +362,8 @@ struct mlx5_ib_dev {
 	spinlock_t			mr_lock;
 	struct mlx5_ib_resources	devr;
 	struct mlx5_mr_cache		cache;
+	struct timer_list		delay_timer;
+	int				fill_delay;
 };

 static inline struct mlx5_ib_cq *to_mibcq(struct mlx5_core_cq *mcq)
@@ -35,11 +35,12 @@
 #include <linux/random.h>
 #include <linux/debugfs.h>
 #include <linux/export.h>
+#include <linux/delay.h>
 #include <rdma/ib_umem.h>
 #include "mlx5_ib.h"

 enum {
-	DEF_CACHE_SIZE	= 10,
+	MAX_PENDING_REG_MR = 8,
 };

 enum {
@@ -63,6 +64,51 @@ static int order2idx(struct mlx5_ib_dev *dev, int order)
 		return order - cache->ent[0].order;
 }

+static void reg_mr_callback(int status, void *context)
+{
+	struct mlx5_ib_mr *mr = context;
+	struct mlx5_ib_dev *dev = mr->dev;
+	struct mlx5_mr_cache *cache = &dev->cache;
+	int c = order2idx(dev, mr->order);
+	struct mlx5_cache_ent *ent = &cache->ent[c];
+	u8 key;
+	unsigned long flags;
+
+	spin_lock_irqsave(&ent->lock, flags);
+	ent->pending--;
+	spin_unlock_irqrestore(&ent->lock, flags);
+	if (status) {
+		mlx5_ib_warn(dev, "async reg mr failed. status %d\n", status);
+		kfree(mr);
+		dev->fill_delay = 1;
+		mod_timer(&dev->delay_timer, jiffies + HZ);
+		return;
+	}
+
+	if (mr->out.hdr.status) {
+		mlx5_ib_warn(dev, "failed - status %d, syndorme 0x%x\n",
+			     mr->out.hdr.status,
+			     be32_to_cpu(mr->out.hdr.syndrome));
+		kfree(mr);
+		dev->fill_delay = 1;
+		mod_timer(&dev->delay_timer, jiffies + HZ);
+		return;
+	}
+
+	spin_lock_irqsave(&dev->mdev.priv.mkey_lock, flags);
+	key = dev->mdev.priv.mkey_key++;
+	spin_unlock_irqrestore(&dev->mdev.priv.mkey_lock, flags);
+	mr->mmr.key = mlx5_idx_to_mkey(be32_to_cpu(mr->out.mkey) & 0xffffff) | key;
+
+	cache->last_add = jiffies;
+
+	spin_lock_irqsave(&ent->lock, flags);
+	list_add_tail(&mr->list, &ent->head);
+	ent->cur++;
+	ent->size++;
+	spin_unlock_irqrestore(&ent->lock, flags);
+}
+
 static int add_keys(struct mlx5_ib_dev *dev, int c, int num)
 {
 	struct mlx5_mr_cache *cache = &dev->cache;
@@ -78,36 +124,39 @@ static int add_keys(struct mlx5_ib_dev *dev, int c, int num)
 		return -ENOMEM;

 	for (i = 0; i < num; i++) {
+		if (ent->pending >= MAX_PENDING_REG_MR) {
+			err = -EAGAIN;
+			break;
+		}
+
 		mr = kzalloc(sizeof(*mr), GFP_KERNEL);
 		if (!mr) {
 			err = -ENOMEM;
-			goto out;
+			break;
 		}
 		mr->order = ent->order;
 		mr->umred = 1;
+		mr->dev = dev;
 		in->seg.status = 1 << 6;
 		in->seg.xlt_oct_size = cpu_to_be32((npages + 1) / 2);
 		in->seg.qpn_mkey7_0 = cpu_to_be32(0xffffff << 8);
 		in->seg.flags = MLX5_ACCESS_MODE_MTT | MLX5_PERM_UMR_EN;
 		in->seg.log2_page_size = 12;

+		spin_lock_irq(&ent->lock);
+		ent->pending++;
+		spin_unlock_irq(&ent->lock);
+		mr->start = jiffies;
 		err = mlx5_core_create_mkey(&dev->mdev, &mr->mmr, in,
-					    sizeof(*in));
+					    sizeof(*in), reg_mr_callback,
+					    mr, &mr->out);
 		if (err) {
 			mlx5_ib_warn(dev, "create mkey failed %d\n", err);
 			kfree(mr);
-			goto out;
+			break;
 		}
-		cache->last_add = jiffies;
-
-		spin_lock(&ent->lock);
-		list_add_tail(&mr->list, &ent->head);
-		ent->cur++;
-		ent->size++;
-		spin_unlock(&ent->lock);
 	}

-out:
 	kfree(in);
 	return err;
 }
@@ -121,16 +170,16 @@ static void remove_keys(struct mlx5_ib_dev *dev, int c, int num)
 	int i;

 	for (i = 0; i < num; i++) {
-		spin_lock(&ent->lock);
+		spin_lock_irq(&ent->lock);
 		if (list_empty(&ent->head)) {
-			spin_unlock(&ent->lock);
+			spin_unlock_irq(&ent->lock);
 			return;
 		}
 		mr = list_first_entry(&ent->head, struct mlx5_ib_mr, list);
 		list_del(&mr->list);
 		ent->cur--;
 		ent->size--;
-		spin_unlock(&ent->lock);
+		spin_unlock_irq(&ent->lock);
 		err = mlx5_core_destroy_mkey(&dev->mdev, &mr->mmr);
 		if (err)
 			mlx5_ib_warn(dev, "failed destroy mkey\n");
@@ -162,9 +211,13 @@ static ssize_t size_write(struct file *filp, const char __user *buf,
 		return -EINVAL;

 	if (var > ent->size) {
-		err = add_keys(dev, c, var - ent->size);
-		if (err)
-			return err;
+		do {
+			err = add_keys(dev, c, var - ent->size);
+			if (err && err != -EAGAIN)
+				return err;
+
+			usleep_range(3000, 5000);
+		} while (err);
 	} else if (var < ent->size) {
 		remove_keys(dev, c, ent->size - var);
 	}
@@ -280,23 +333,37 @@ static void __cache_work_func(struct mlx5_cache_ent *ent)
 	struct mlx5_ib_dev *dev = ent->dev;
 	struct mlx5_mr_cache *cache = &dev->cache;
 	int i = order2idx(dev, ent->order);
+	int err;

 	if (cache->stopped)
 		return;

 	ent = &dev->cache.ent[i];
-	if (ent->cur < 2 * ent->limit) {
-		add_keys(dev, i, 1);
-		if (ent->cur < 2 * ent->limit)
-			queue_work(cache->wq, &ent->work);
+	if (ent->cur < 2 * ent->limit && !dev->fill_delay) {
+		err = add_keys(dev, i, 1);
+		if (ent->cur < 2 * ent->limit) {
+			if (err == -EAGAIN) {
+				mlx5_ib_dbg(dev, "returned eagain, order %d\n",
+					    i + 2);
+				queue_delayed_work(cache->wq, &ent->dwork,
+						   msecs_to_jiffies(3));
+			} else if (err) {
+				mlx5_ib_warn(dev, "command failed order %d, err %d\n",
+					     i + 2, err);
+				queue_delayed_work(cache->wq, &ent->dwork,
+						   msecs_to_jiffies(1000));
+			} else {
+				queue_work(cache->wq, &ent->work);
+			}
+		}
 	} else if (ent->cur > 2 * ent->limit) {
 		if (!someone_adding(cache) &&
-		    time_after(jiffies, cache->last_add + 60 * HZ)) {
+		    time_after(jiffies, cache->last_add + 300 * HZ)) {
 			remove_keys(dev, i, 1);
 			if (ent->cur > ent->limit)
 				queue_work(cache->wq, &ent->work);
 		} else {
-			queue_delayed_work(cache->wq, &ent->dwork, 60 * HZ);
+			queue_delayed_work(cache->wq, &ent->dwork, 300 * HZ);
 		}
 	}
 }
@@ -336,18 +403,18 @@ static struct mlx5_ib_mr *alloc_cached_mr(struct mlx5_ib_dev *dev, int order)

 		mlx5_ib_dbg(dev, "order %d, cache index %d\n", ent->order, i);

-		spin_lock(&ent->lock);
+		spin_lock_irq(&ent->lock);
 		if (!list_empty(&ent->head)) {
 			mr = list_first_entry(&ent->head, struct mlx5_ib_mr,
 					      list);
 			list_del(&mr->list);
 			ent->cur--;
-			spin_unlock(&ent->lock);
+			spin_unlock_irq(&ent->lock);
 			if (ent->cur < ent->limit)
 				queue_work(cache->wq, &ent->work);
 			break;
 		}
-		spin_unlock(&ent->lock);
+		spin_unlock_irq(&ent->lock);

 		queue_work(cache->wq, &ent->work);

@@ -374,12 +441,12 @@ static void free_cached_mr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr)
 		return;
 	}
 	ent = &cache->ent[c];
-	spin_lock(&ent->lock);
+	spin_lock_irq(&ent->lock);
 	list_add_tail(&mr->list, &ent->head);
 	ent->cur++;
 	if (ent->cur > 2 * ent->limit)
 		shrink = 1;
-	spin_unlock(&ent->lock);
+	spin_unlock_irq(&ent->lock);

 	if (shrink)
 		queue_work(cache->wq, &ent->work);
@@ -394,16 +461,16 @@ static void clean_keys(struct mlx5_ib_dev *dev, int c)

 	cancel_delayed_work(&ent->dwork);
 	while (1) {
-		spin_lock(&ent->lock);
+		spin_lock_irq(&ent->lock);
 		if (list_empty(&ent->head)) {
-			spin_unlock(&ent->lock);
+			spin_unlock_irq(&ent->lock);
 			return;
 		}
 		mr = list_first_entry(&ent->head, struct mlx5_ib_mr, list);
 		list_del(&mr->list);
 		ent->cur--;
 		ent->size--;
-		spin_unlock(&ent->lock);
+		spin_unlock_irq(&ent->lock);
 		err = mlx5_core_destroy_mkey(&dev->mdev, &mr->mmr);
 		if (err)
 			mlx5_ib_warn(dev, "failed destroy mkey\n");
@@ -464,12 +531,18 @@ static void mlx5_mr_cache_debugfs_cleanup(struct mlx5_ib_dev *dev)
 	debugfs_remove_recursive(dev->cache.root);
 }

+static void delay_time_func(unsigned long ctx)
+{
+	struct mlx5_ib_dev *dev = (struct mlx5_ib_dev *)ctx;
+
+	dev->fill_delay = 0;
+}
+
 int mlx5_mr_cache_init(struct mlx5_ib_dev *dev)
 {
 	struct mlx5_mr_cache *cache = &dev->cache;
 	struct mlx5_cache_ent *ent;
 	int limit;
-	int size;
 	int err;
 	int i;

@@ -479,6 +552,7 @@ int mlx5_mr_cache_init(struct mlx5_ib_dev *dev)
 		return -ENOMEM;
 	}

+	setup_timer(&dev->delay_timer, delay_time_func, (unsigned long)dev);
 	for (i = 0; i < MAX_MR_CACHE_ENTRIES; i++) {
 		INIT_LIST_HEAD(&cache->ent[i].head);
 		spin_lock_init(&cache->ent[i].lock);
@@ -489,13 +563,11 @@ int mlx5_mr_cache_init(struct mlx5_ib_dev *dev)
 		ent->order = i + 2;
 		ent->dev = dev;

-		if (dev->mdev.profile->mask & MLX5_PROF_MASK_MR_CACHE) {
-			size = dev->mdev.profile->mr_cache[i].size;
+		if (dev->mdev.profile->mask & MLX5_PROF_MASK_MR_CACHE)
 			limit = dev->mdev.profile->mr_cache[i].limit;
-		} else {
-			size = DEF_CACHE_SIZE;
+		else
 			limit = 0;
-		}
+
 		INIT_WORK(&ent->work, cache_work_func);
 		INIT_DELAYED_WORK(&ent->dwork, delayed_cache_work_func);
 		ent->limit = limit;
@@ -522,6 +594,7 @@ int mlx5_mr_cache_cleanup(struct mlx5_ib_dev *dev)
 		clean_keys(dev, i);

 	destroy_workqueue(dev->cache.wq);
+	del_timer_sync(&dev->delay_timer);

 	return 0;
 }
@@ -551,7 +624,8 @@ struct ib_mr *mlx5_ib_get_dma_mr(struct ib_pd *pd, int acc)
 	seg->qpn_mkey7_0 = cpu_to_be32(0xffffff << 8);
 	seg->start_addr = 0;

-	err = mlx5_core_create_mkey(mdev, &mr->mmr, in, sizeof(*in));
+	err = mlx5_core_create_mkey(mdev, &mr->mmr, in, sizeof(*in), NULL, NULL,
+				    NULL);
 	if (err)
 		goto err_in;

@@ -660,14 +734,14 @@ static struct mlx5_ib_mr *reg_umr(struct ib_pd *pd, struct ib_umem *umem,
 	int err;
 	int i;

-	for (i = 0; i < 10; i++) {
+	for (i = 0; i < 1; i++) {
 		mr = alloc_cached_mr(dev, order);
 		if (mr)
 			break;

 		err = add_keys(dev, order2idx(dev, order), 1);
-		if (err) {
-			mlx5_ib_warn(dev, "add_keys failed\n");
+		if (err && err != -EAGAIN) {
+			mlx5_ib_warn(dev, "add_keys failed, err %d\n", err);
 			break;
 		}
 	}
@@ -759,8 +833,10 @@ static struct mlx5_ib_mr *reg_create(struct ib_pd *pd, u64 virt_addr,
 	in->seg.xlt_oct_size = cpu_to_be32(get_octo_len(virt_addr, length, 1 << page_shift));
 	in->seg.log2_page_size = page_shift;
 	in->seg.qpn_mkey7_0 = cpu_to_be32(0xffffff << 8);
-	in->xlat_oct_act_size = cpu_to_be32(get_octo_len(virt_addr, length, 1 << page_shift));
-	err = mlx5_core_create_mkey(&dev->mdev, &mr->mmr, in, inlen);
+	in->xlat_oct_act_size = cpu_to_be32(get_octo_len(virt_addr, length,
+							 1 << page_shift));
+	err = mlx5_core_create_mkey(&dev->mdev, &mr->mmr, in, inlen, NULL,
+				    NULL, NULL);
 	if (err) {
 		mlx5_ib_warn(dev, "create mkey failed\n");
 		goto err_2;
@@ -944,7 +1020,8 @@ struct ib_mr *mlx5_ib_alloc_fast_reg_mr(struct ib_pd *pd,
 	 * TBD not needed - issue 197292 */
 	in->seg.log2_page_size = PAGE_SHIFT;

-	err = mlx5_core_create_mkey(&dev->mdev, &mr->mmr, in, sizeof(*in));
+	err = mlx5_core_create_mkey(&dev->mdev, &mr->mmr, in, sizeof(*in), NULL,
+				    NULL, NULL);
 	kfree(in);
 	if (err)
 		goto err_free;
--- a/Show More
+++ b/Show More