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
  ...

2 files changed, 335 insertions, 69 deletions

diff --git a/drivers/infiniband/ulp/srpt/ib_srpt.c b/drivers/infiniband/ulp/srpt/ib_srpt.c
index 0373b7c40902..dfec0e1fac29 100644
--- a/drivers/infiniband/ulp/srpt/ib_srpt.c
+++ b/drivers/infiniband/ulp/srpt/ib_srpt.c
@@ -41,6 +41,7 @@
 #include <linux/string.h>
 #include <linux/delay.h>
 #include <linux/atomic.h>
+#include <linux/inet.h>
 #include <rdma/ib_cache.h>
 #include <scsi/scsi_proto.h>
 #include <scsi/scsi_tcq.h>
@@ -92,6 +93,11 @@ MODULE_PARM_DESC(srpt_service_guid,
 		 " instead of using the node_guid of the first HCA.");
 
 static struct ib_client srpt_client;
+/* Protects both rdma_cm_port and rdma_cm_id. */
+static DEFINE_MUTEX(rdma_cm_mutex);
+/* Port number RDMA/CM will bind to. */
+static u16 rdma_cm_port;
+static struct rdma_cm_id *rdma_cm_id;
 static void srpt_release_cmd(struct se_cmd *se_cmd);
 static void srpt_free_ch(struct kref *kref);
 static int srpt_queue_status(struct se_cmd *cmd);
@@ -220,7 +226,10 @@ static void srpt_qp_event(struct ib_event *event, struct srpt_rdma_ch *ch)
 
 	switch (event->event) {
 	case IB_EVENT_COMM_EST:
-		ib_cm_notify(ch->ib_cm.cm_id, event->event);
+		if (ch->using_rdma_cm)
+			rdma_notify(ch->rdma_cm.cm_id, event->event);
+		else
+			ib_cm_notify(ch->ib_cm.cm_id, event->event);
 		break;
 	case IB_EVENT_QP_LAST_WQE_REACHED:
 		pr_debug("%s-%d, state %s: received Last WQE event.\n",
@@ -838,16 +847,20 @@ static int srpt_post_recv(struct srpt_device *sdev, struct srpt_rdma_ch *ch,
  */
 static int srpt_zerolength_write(struct srpt_rdma_ch *ch)
 {
-	struct ib_send_wr wr, *bad_wr;
+	struct ib_send_wr *bad_wr;
+	struct ib_rdma_wr wr = {
+		.wr = {
+			.next		= NULL,
+			{ .wr_cqe	= &ch->zw_cqe, },
+			.opcode		= IB_WR_RDMA_WRITE,
+			.send_flags	= IB_SEND_SIGNALED,
+		}
+	};
 
 	pr_debug("%s-%d: queued zerolength write\n", ch->sess_name,
 		 ch->qp->qp_num);
 
-	memset(&wr, 0, sizeof(wr));
-	wr.opcode = IB_WR_RDMA_WRITE;
-	wr.wr_cqe = &ch->zw_cqe;
-	wr.send_flags = IB_SEND_SIGNALED;
-	return ib_post_send(ch->qp, &wr, &bad_wr);
+	return ib_post_send(ch->qp, &wr.wr, &bad_wr);
 }
 
 static void srpt_zerolength_write_done(struct ib_cq *cq, struct ib_wc *wc)
@@ -1057,6 +1070,8 @@ static int srpt_init_ch_qp(struct srpt_rdma_ch *ch, struct ib_qp *qp)
 	struct ib_qp_attr *attr;
 	int ret;
 
+	WARN_ON_ONCE(ch->using_rdma_cm);
+
 	attr = kzalloc(sizeof(*attr), GFP_KERNEL);
 	if (!attr)
 		return -ENOMEM;
@@ -1096,6 +1111,8 @@ static int srpt_ch_qp_rtr(struct srpt_rdma_ch *ch, struct ib_qp *qp)
 	int attr_mask;
 	int ret;
 
+	WARN_ON_ONCE(ch->using_rdma_cm);
+
 	qp_attr.qp_state = IB_QPS_RTR;
 	ret = ib_cm_init_qp_attr(ch->ib_cm.cm_id, &qp_attr, &attr_mask);
 	if (ret)
@@ -1746,18 +1763,33 @@ retry:
 		qp_init->cap.max_recv_sge = qp_init->cap.max_send_sge;
 	}
 
-	ch->qp = ib_create_qp(sdev->pd, qp_init);
-	if (IS_ERR(ch->qp)) {
-		ret = PTR_ERR(ch->qp);
-		if (ret == -ENOMEM) {
-			sq_size /= 2;
-			if (sq_size >= MIN_SRPT_SQ_SIZE) {
-				ib_destroy_cq(ch->cq);
-				goto retry;
-			}
+	if (ch->using_rdma_cm) {
+		ret = rdma_create_qp(ch->rdma_cm.cm_id, sdev->pd, qp_init);
+		ch->qp = ch->rdma_cm.cm_id->qp;
+	} else {
+		ch->qp = ib_create_qp(sdev->pd, qp_init);
+		if (!IS_ERR(ch->qp)) {
+			ret = srpt_init_ch_qp(ch, ch->qp);
+			if (ret)
+				ib_destroy_qp(ch->qp);
+		} else {
+			ret = PTR_ERR(ch->qp);
+		}
+	}
+	if (ret) {
+		bool retry = sq_size > MIN_SRPT_SQ_SIZE;
+
+		if (retry) {
+			pr_debug("failed to create queue pair with sq_size = %d (%d) - retrying\n",
+				 sq_size, ret);
+			ib_free_cq(ch->cq);
+			sq_size = max(sq_size / 2, MIN_SRPT_SQ_SIZE);
+			goto retry;
+		} else {
+			pr_err("failed to create queue pair with sq_size = %d (%d)\n",
+			       sq_size, ret);
+			goto err_destroy_cq;
 		}
-		pr_err("failed to create_qp ret= %d\n", ret);
-		goto err_destroy_cq;
 	}
 
 	atomic_set(&ch->sq_wr_avail, qp_init->cap.max_send_wr);
@@ -1766,10 +1798,6 @@ retry:
 		 __func__, ch->cq->cqe, qp_init->cap.max_send_sge,
 		 qp_init->cap.max_send_wr, ch);
 
-	ret = srpt_init_ch_qp(ch, ch->qp);
-	if (ret)
-		goto err_destroy_qp;
-
 	if (!sdev->use_srq)
 		for (i = 0; i < ch->rq_size; i++)
 			srpt_post_recv(sdev, ch, ch->ioctx_recv_ring[i]);
@@ -1778,9 +1806,8 @@ out:
 	kfree(qp_init);
 	return ret;
 
-err_destroy_qp:
-	ib_destroy_qp(ch->qp);
 err_destroy_cq:
+	ch->qp = NULL;
 	ib_free_cq(ch->cq);
 	goto out;
 }
@@ -1849,9 +1876,13 @@ static int srpt_disconnect_ch(struct srpt_rdma_ch *ch)
 	if (!srpt_set_ch_state(ch, CH_DISCONNECTING))
 		return -ENOTCONN;
 
-	ret = ib_send_cm_dreq(ch->ib_cm.cm_id, NULL, 0);
-	if (ret < 0)
-		ret = ib_send_cm_drep(ch->ib_cm.cm_id, NULL, 0);
+	if (ch->using_rdma_cm) {
+		ret = rdma_disconnect(ch->rdma_cm.cm_id);
+	} else {
+		ret = ib_send_cm_dreq(ch->ib_cm.cm_id, NULL, 0);
+		if (ret < 0)
+			ret = ib_send_cm_drep(ch->ib_cm.cm_id, NULL, 0);
+	}
 
 	if (ret < 0 && srpt_close_ch(ch))
 		ret = 0;
@@ -2002,7 +2033,10 @@ static void srpt_release_channel_work(struct work_struct *w)
 	transport_deregister_session(se_sess);
 	ch->sess = NULL;
 
-	ib_destroy_cm_id(ch->ib_cm.cm_id);
+	if (ch->using_rdma_cm)
+		rdma_destroy_id(ch->rdma_cm.cm_id);
+	else
+		ib_destroy_cm_id(ch->ib_cm.cm_id);
 
 	srpt_destroy_ch_ib(ch);
 
@@ -2026,26 +2060,33 @@ static void srpt_release_channel_work(struct work_struct *w)
 
 /**
  * srpt_cm_req_recv - process the event IB_CM_REQ_RECEIVED
- * @cm_id: IB/CM connection identifier.
- * @port_num: Port through which the IB/CM REQ message was received.
+ * @sdev: HCA through which the login request was received.
+ * @ib_cm_id: IB/CM connection identifier in case of IB/CM.
+ * @rdma_cm_id: RDMA/CM connection identifier in case of RDMA/CM.
+ * @port_num: Port through which the REQ message was received.
  * @pkey: P_Key of the incoming connection.
  * @req: SRP login request.
- * @src_addr: GID of the port that submitted the login request.
+ * @src_addr: GID (IB/CM) or IP address (RDMA/CM) of the port that submitted
+ * the login request.
  *
  * Ownership of the cm_id is transferred to the target session if this
- * functions returns zero. Otherwise the caller remains the owner of cm_id.
+ * function returns zero. Otherwise the caller remains the owner of cm_id.
  */
-static int srpt_cm_req_recv(struct ib_cm_id *cm_id,
+static int srpt_cm_req_recv(struct srpt_device *const sdev,
+			    struct ib_cm_id *ib_cm_id,
+			    struct rdma_cm_id *rdma_cm_id,
 			    u8 port_num, __be16 pkey,
 			    const struct srp_login_req *req,
 			    const char *src_addr)
 {
-	struct srpt_device *sdev = cm_id->context;
 	struct srpt_port *sport = &sdev->port[port_num - 1];
 	struct srpt_nexus *nexus;
 	struct srp_login_rsp *rsp = NULL;
 	struct srp_login_rej *rej = NULL;
-	struct ib_cm_rep_param *rep_param = NULL;
+	union {
+		struct rdma_conn_param rdma_cm;
+		struct ib_cm_rep_param ib_cm;
+	} *rep_param = NULL;
 	struct srpt_rdma_ch *ch;
 	char i_port_id[36];
 	u32 it_iu_len;
@@ -2115,8 +2156,14 @@ static int srpt_cm_req_recv(struct ib_cm_id *cm_id,
 	ch->zw_cqe.done = srpt_zerolength_write_done;
 	INIT_WORK(&ch->release_work, srpt_release_channel_work);
 	ch->sport = sport;
-	ch->ib_cm.cm_id = cm_id;
-	cm_id->context = ch;
+	if (ib_cm_id) {
+		ch->ib_cm.cm_id = ib_cm_id;
+		ib_cm_id->context = ch;
+	} else {
+		ch->using_rdma_cm = true;
+		ch->rdma_cm.cm_id = rdma_cm_id;
+		rdma_cm_id->context = ch;
+	}
 	/*
 	 * ch->rq_size should be at least as large as the initiator queue
 	 * depth to avoid that the initiator driver has to report QUEUE_FULL
@@ -2227,7 +2274,7 @@ static int srpt_cm_req_recv(struct ib_cm_id *cm_id,
 
 	mutex_unlock(&sport->mutex);
 
-	ret = srpt_ch_qp_rtr(ch, ch->qp);
+	ret = ch->using_rdma_cm ? 0 : srpt_ch_qp_rtr(ch, ch->qp);
 	if (ret) {
 		rej->reason = cpu_to_be32(SRP_LOGIN_REJ_INSUFFICIENT_RESOURCES);
 		pr_err("rejected SRP_LOGIN_REQ because enabling RTR failed (error code = %d)\n",
@@ -2251,25 +2298,38 @@ static int srpt_cm_req_recv(struct ib_cm_id *cm_id,
 	atomic_set(&ch->req_lim_delta, 0);
 
 	/* create cm reply */
-	rep_param->qp_num = ch->qp->qp_num;
-	rep_param->private_data = (void *)rsp;
-	rep_param->private_data_len = sizeof(*rsp);
-	rep_param->rnr_retry_count = 7;
-	rep_param->flow_control = 1;
-	rep_param->failover_accepted = 0;
-	rep_param->srq = 1;
-	rep_param->responder_resources = 4;
-	rep_param->initiator_depth = 4;
+	if (ch->using_rdma_cm) {
+		rep_param->rdma_cm.private_data = (void *)rsp;
+		rep_param->rdma_cm.private_data_len = sizeof(*rsp);
+		rep_param->rdma_cm.rnr_retry_count = 7;
+		rep_param->rdma_cm.flow_control = 1;
+		rep_param->rdma_cm.responder_resources = 4;
+		rep_param->rdma_cm.initiator_depth = 4;
+	} else {
+		rep_param->ib_cm.qp_num = ch->qp->qp_num;
+		rep_param->ib_cm.private_data = (void *)rsp;
+		rep_param->ib_cm.private_data_len = sizeof(*rsp);
+		rep_param->ib_cm.rnr_retry_count = 7;
+		rep_param->ib_cm.flow_control = 1;
+		rep_param->ib_cm.failover_accepted = 0;
+		rep_param->ib_cm.srq = 1;
+		rep_param->ib_cm.responder_resources = 4;
+		rep_param->ib_cm.initiator_depth = 4;
+	}
 
 	/*
 	 * Hold the sport mutex while accepting a connection to avoid that
 	 * srpt_disconnect_ch() is invoked concurrently with this code.
 	 */
 	mutex_lock(&sport->mutex);
-	if (sport->enabled && ch->state == CH_CONNECTING)
-		ret = ib_send_cm_rep(cm_id, rep_param);
-	else
+	if (sport->enabled && ch->state == CH_CONNECTING) {
+		if (ch->using_rdma_cm)
+			ret = rdma_accept(rdma_cm_id, &rep_param->rdma_cm);
+		else
+			ret = ib_send_cm_rep(ib_cm_id, &rep_param->ib_cm);
+	} else {
 		ret = -EINVAL;
+	}
 	mutex_unlock(&sport->mutex);
 
 	switch (ret) {
@@ -2299,7 +2359,8 @@ free_ring:
 			     ch->sport->sdev, ch->rq_size,
 			     ch->max_rsp_size, DMA_TO_DEVICE);
 free_ch:
-	cm_id->context = NULL;
+	if (ib_cm_id)
+		ib_cm_id->context = NULL;
 	kfree(ch);
 	ch = NULL;
 
@@ -2312,8 +2373,11 @@ reject:
 	rej->buf_fmt = cpu_to_be16(SRP_BUF_FORMAT_DIRECT |
 				   SRP_BUF_FORMAT_INDIRECT);
 
-	ib_send_cm_rej(cm_id, IB_CM_REJ_CONSUMER_DEFINED, NULL, 0,
-			     (void *)rej, sizeof(*rej));
+	if (rdma_cm_id)
+		rdma_reject(rdma_cm_id, rej, sizeof(*rej));
+	else
+		ib_send_cm_rej(ib_cm_id, IB_CM_REJ_CONSUMER_DEFINED, NULL, 0,
+			       rej, sizeof(*rej));
 
 out:
 	kfree(rep_param);
@@ -2332,10 +2396,44 @@ static int srpt_ib_cm_req_recv(struct ib_cm_id *cm_id,
 	srpt_format_guid(sguid, sizeof(sguid),
 			 &param->primary_path->dgid.global.interface_id);
 
-	return srpt_cm_req_recv(cm_id, param->port, param->primary_path->pkey,
+	return srpt_cm_req_recv(cm_id->context, cm_id, NULL, param->port,
+				param->primary_path->pkey,
 				private_data, sguid);
 }
 
+static int srpt_rdma_cm_req_recv(struct rdma_cm_id *cm_id,
+				 struct rdma_cm_event *event)
+{
+	struct srpt_device *sdev;
+	struct srp_login_req req;
+	const struct srp_login_req_rdma *req_rdma;
+	char src_addr[40];
+
+	sdev = ib_get_client_data(cm_id->device, &srpt_client);
+	if (!sdev)
+		return -ECONNREFUSED;
+
+	if (event->param.conn.private_data_len < sizeof(*req_rdma))
+		return -EINVAL;
+
+	/* Transform srp_login_req_rdma into srp_login_req. */
+	req_rdma = event->param.conn.private_data;
+	memset(&req, 0, sizeof(req));
+	req.opcode		= req_rdma->opcode;
+	req.tag			= req_rdma->tag;
+	req.req_it_iu_len	= req_rdma->req_it_iu_len;
+	req.req_buf_fmt		= req_rdma->req_buf_fmt;
+	req.req_flags		= req_rdma->req_flags;
+	memcpy(req.initiator_port_id, req_rdma->initiator_port_id, 16);
+	memcpy(req.target_port_id, req_rdma->target_port_id, 16);
+
+	snprintf(src_addr, sizeof(src_addr), "%pIS",
+		 &cm_id->route.addr.src_addr);
+
+	return srpt_cm_req_recv(sdev, NULL, cm_id, cm_id->port_num,
+				cm_id->route.path_rec->pkey, &req, src_addr);
+}
+
 static void srpt_cm_rej_recv(struct srpt_rdma_ch *ch,
 			     enum ib_cm_rej_reason reason,
 			     const u8 *private_data,
@@ -2359,14 +2457,14 @@ static void srpt_cm_rej_recv(struct srpt_rdma_ch *ch,
  * srpt_cm_rtu_recv - process an IB_CM_RTU_RECEIVED or USER_ESTABLISHED event
  * @ch: SRPT RDMA channel.
  *
- * An IB_CM_RTU_RECEIVED message indicates that the connection is established
- * and that the recipient may begin transmitting (RTU = ready to use).
+ * An RTU (ready to use) message indicates that the connection has been
+ * established and that the recipient may begin transmitting.
  */
 static void srpt_cm_rtu_recv(struct srpt_rdma_ch *ch)
 {
 	int ret;
 
-	ret = srpt_ch_qp_rts(ch, ch->qp);
+	ret = ch->using_rdma_cm ? 0 : srpt_ch_qp_rts(ch, ch->qp);
 	if (ret < 0) {
 		pr_err("%s-%d: QP transition to RTS failed\n", ch->sess_name,
 		       ch->qp->qp_num);
@@ -2453,6 +2551,49 @@ static int srpt_cm_handler(struct ib_cm_id *cm_id, struct ib_cm_event *event)
 	return ret;
 }
 
+static int srpt_rdma_cm_handler(struct rdma_cm_id *cm_id,
+				struct rdma_cm_event *event)
+{
+	struct srpt_rdma_ch *ch = cm_id->context;
+	int ret = 0;
+
+	switch (event->event) {
+	case RDMA_CM_EVENT_CONNECT_REQUEST:
+		ret = srpt_rdma_cm_req_recv(cm_id, event);
+		break;
+	case RDMA_CM_EVENT_REJECTED:
+		srpt_cm_rej_recv(ch, event->status,
+				 event->param.conn.private_data,
+				 event->param.conn.private_data_len);
+		break;
+	case RDMA_CM_EVENT_ESTABLISHED:
+		srpt_cm_rtu_recv(ch);
+		break;
+	case RDMA_CM_EVENT_DISCONNECTED:
+		if (ch->state < CH_DISCONNECTING)
+			srpt_disconnect_ch(ch);
+		else
+			srpt_close_ch(ch);
+		break;
+	case RDMA_CM_EVENT_TIMEWAIT_EXIT:
+		srpt_close_ch(ch);
+		break;
+	case RDMA_CM_EVENT_UNREACHABLE:
+		pr_info("Received CM REP error for ch %s-%d.\n", ch->sess_name,
+			ch->qp->qp_num);
+		break;
+	case RDMA_CM_EVENT_DEVICE_REMOVAL:
+	case RDMA_CM_EVENT_ADDR_CHANGE:
+		break;
+	default:
+		pr_err("received unrecognized RDMA CM event %d\n",
+		       event->event);
+		break;
+	}
+
+	return ret;
+}
+
 static int srpt_write_pending_status(struct se_cmd *se_cmd)
 {
 	struct srpt_send_ioctx *ioctx;
@@ -2824,7 +2965,7 @@ static void srpt_add_one(struct ib_device *device)
 {
 	struct srpt_device *sdev;
 	struct srpt_port *sport;
-	int i;
+	int i, ret;
 
 	pr_debug("device = %p\n", device);
 
@@ -2848,9 +2989,15 @@ static void srpt_add_one(struct ib_device *device)
 	if (!srpt_service_guid)
 		srpt_service_guid = be64_to_cpu(device->node_guid);
 
-	sdev->cm_id = ib_create_cm_id(device, srpt_cm_handler, sdev);
-	if (IS_ERR(sdev->cm_id))
-		goto err_ring;
+	if (rdma_port_get_link_layer(device, 1) == IB_LINK_LAYER_INFINIBAND)
+		sdev->cm_id = ib_create_cm_id(device, srpt_cm_handler, sdev);
+	if (IS_ERR(sdev->cm_id)) {
+		pr_info("ib_create_cm_id() failed: %ld\n",
+			PTR_ERR(sdev->cm_id));
+		sdev->cm_id = NULL;
+		if (!rdma_cm_id)
+			goto err_ring;
+	}
 
 	/* print out target login information */
 	pr_debug("Target login info: id_ext=%016llx,ioc_guid=%016llx,"
@@ -2863,8 +3010,14 @@ static void srpt_add_one(struct ib_device *device)
 	 * in the system as service_id; therefore, the target_id will change
 	 * if this HCA is gone bad and replaced by different HCA
 	 */
-	if (ib_cm_listen(sdev->cm_id, cpu_to_be64(srpt_service_guid), 0))
+	ret = sdev->cm_id ?
+		ib_cm_listen(sdev->cm_id, cpu_to_be64(srpt_service_guid), 0) :
+		0;
+	if (ret < 0) {
+		pr_err("ib_cm_listen() failed: %d (cm_id state = %d)\n", ret,
+		       sdev->cm_id->state);
 		goto err_cm;
+	}
 
 	INIT_IB_EVENT_HANDLER(&sdev->event_handler, sdev->device,
 			      srpt_event_handler);
@@ -2904,7 +3057,8 @@ out:
 err_event:
 	ib_unregister_event_handler(&sdev->event_handler);
 err_cm:
-	ib_destroy_cm_id(sdev->cm_id);
+	if (sdev->cm_id)
+		ib_destroy_cm_id(sdev->cm_id);
 err_ring:
 	srpt_free_srq(sdev);
 	ib_dealloc_pd(sdev->pd);
@@ -2939,7 +3093,10 @@ static void srpt_remove_one(struct ib_device *device, void *client_data)
 	for (i = 0; i < sdev->device->phys_port_cnt; i++)
 		cancel_work_sync(&sdev->port[i].work);
 
-	ib_destroy_cm_id(sdev->cm_id);
+	if (sdev->cm_id)
+		ib_destroy_cm_id(sdev->cm_id);
+
+	ib_set_client_data(device, &srpt_client, NULL);
 
 	/*
 	 * Unregistering a target must happen after destroying sdev->cm_id
@@ -3103,18 +3260,26 @@ static int srpt_parse_i_port_id(u8 i_port_id[16], const char *name)
 	leading_zero_bytes = 16 - count;
 	memset(i_port_id, 0, leading_zero_bytes);
 	ret = hex2bin(i_port_id + leading_zero_bytes, p, count);
-	if (ret < 0)
-		pr_debug("hex2bin failed for srpt_parse_i_port_id: %d\n", ret);
+
 out:
 	return ret;
 }
 
 /*
- * configfs callback function invoked for
- * mkdir /sys/kernel/config/target/$driver/$port/$tpg/acls/$i_port_id
+ * configfs callback function invoked for mkdir
+ * /sys/kernel/config/target/$driver/$port/$tpg/acls/$i_port_id
+ *
+ * i_port_id must be an initiator port GUID, GID or IP address. See also the
+ * target_alloc_session() calls in this driver. Examples of valid initiator
+ * port IDs:
+ * 0x0000000000000000505400fffe4a0b7b
+ * 0000000000000000505400fffe4a0b7b
+ * 5054:00ff:fe4a:0b7b
+ * 192.168.122.76
  */
 static int srpt_init_nodeacl(struct se_node_acl *se_nacl, const char *name)
 {
+	struct sockaddr_storage sa;
 	u64 guid;
 	u8 i_port_id[16];
 	int ret;
@@ -3123,6 +3288,9 @@ static int srpt_init_nodeacl(struct se_node_acl *se_nacl, const char *name)
 	if (ret < 0)
 		ret = srpt_parse_i_port_id(i_port_id, name);
 	if (ret < 0)
+		ret = inet_pton_with_scope(&init_net, AF_UNSPEC, name, NULL,
+					   &sa);
+	if (ret < 0)
 		pr_err("invalid initiator port ID %s\n", name);
 	return ret;
 }
@@ -3296,6 +3464,95 @@ static struct configfs_attribute *srpt_tpg_attrib_attrs[] = {
 	NULL,
 };
 
+static struct rdma_cm_id *srpt_create_rdma_id(struct sockaddr *listen_addr)
+{
+	struct rdma_cm_id *rdma_cm_id;
+	int ret;
+
+	rdma_cm_id = rdma_create_id(&init_net, srpt_rdma_cm_handler,
+				    NULL, RDMA_PS_TCP, IB_QPT_RC);
+	if (IS_ERR(rdma_cm_id)) {
+		pr_err("RDMA/CM ID creation failed: %ld\n",
+		       PTR_ERR(rdma_cm_id));
+		goto out;
+	}
+
+	ret = rdma_bind_addr(rdma_cm_id, listen_addr);
+	if (ret) {
+		char addr_str[64];
+
+		snprintf(addr_str, sizeof(addr_str), "%pISp", listen_addr);
+		pr_err("Binding RDMA/CM ID to address %s failed: %d\n",
+		       addr_str, ret);
+		rdma_destroy_id(rdma_cm_id);
+		rdma_cm_id = ERR_PTR(ret);
+		goto out;
+	}
+
+	ret = rdma_listen(rdma_cm_id, 128);
+	if (ret) {
+		pr_err("rdma_listen() failed: %d\n", ret);
+		rdma_destroy_id(rdma_cm_id);
+		rdma_cm_id = ERR_PTR(ret);
+	}
+
+out:
+	return rdma_cm_id;
+}
+
+static ssize_t srpt_rdma_cm_port_show(struct config_item *item, char *page)
+{
+	return sprintf(page, "%d\n", rdma_cm_port);
+}
+
+static ssize_t srpt_rdma_cm_port_store(struct config_item *item,
+				       const char *page, size_t count)
+{
+	struct sockaddr_in  addr4 = { .sin_family  = AF_INET  };
+	struct sockaddr_in6 addr6 = { .sin6_family = AF_INET6 };
+	struct rdma_cm_id *new_id = NULL;
+	u16 val;
+	int ret;
+
+	ret = kstrtou16(page, 0, &val);
+	if (ret < 0)
+		return ret;
+	ret = count;
+	if (rdma_cm_port == val)
+		goto out;
+
+	if (val) {
+		addr6.sin6_port = cpu_to_be16(val);
+		new_id = srpt_create_rdma_id((struct sockaddr *)&addr6);
+		if (IS_ERR(new_id)) {
+			addr4.sin_port = cpu_to_be16(val);
+			new_id = srpt_create_rdma_id((struct sockaddr *)&addr4);
+			if (IS_ERR(new_id)) {
+				ret = PTR_ERR(new_id);
+				goto out;
+			}
+		}
+	}
+
+	mutex_lock(&rdma_cm_mutex);
+	rdma_cm_port = val;
+	swap(rdma_cm_id, new_id);
+	mutex_unlock(&rdma_cm_mutex);
+
+	if (new_id)
+		rdma_destroy_id(new_id);
+	ret = count;
+out:
+	return ret;
+}
+
+CONFIGFS_ATTR(srpt_, rdma_cm_port);
+
+static struct configfs_attribute *srpt_da_attrs[] = {
+	&srpt_attr_rdma_cm_port,
+	NULL,
+};
+
 static ssize_t srpt_tpg_enable_show(struct config_item *item, char *page)
 {
 	struct se_portal_group *se_tpg = to_tpg(item);
@@ -3441,6 +3698,7 @@ static const struct target_core_fabric_ops srpt_template = {
 	.fabric_drop_tpg		= srpt_drop_tpg,
 	.fabric_init_nodeacl		= srpt_init_nodeacl,
 
+	.tfc_discovery_attrs		= srpt_da_attrs,
 	.tfc_wwn_attrs			= srpt_wwn_attrs,
 	.tfc_tpg_base_attrs		= srpt_tpg_attrs,
 	.tfc_tpg_attrib_attrs		= srpt_tpg_attrib_attrs,
@@ -3494,6 +3752,8 @@ out:
 
 static void __exit srpt_cleanup_module(void)
 {
+	if (rdma_cm_id)
+		rdma_destroy_id(rdma_cm_id);
 	ib_unregister_client(&srpt_client);
 	target_unregister_template(&srpt_template);
 }
diff --git a/drivers/infiniband/ulp/srpt/ib_srpt.h b/drivers/infiniband/ulp/srpt/ib_srpt.h
index 4d9199fd00dc..2361483476a0 100644
--- a/drivers/infiniband/ulp/srpt/ib_srpt.h
+++ b/drivers/infiniband/ulp/srpt/ib_srpt.h
@@ -42,6 +42,7 @@
 #include <rdma/ib_verbs.h>
 #include <rdma/ib_sa.h>
 #include <rdma/ib_cm.h>
+#include <rdma/rdma_cm.h>
 #include <rdma/rw.h>
 
 #include <scsi/srp.h>
@@ -261,6 +262,7 @@ enum rdma_ch_state {
  * @spinlock:      Protects free_list and state.
  * @free_list:     Head of list with free send I/O contexts.
  * @state:         channel state. See also enum rdma_ch_state.
+ * @using_rdma_cm: Whether the RDMA/CM or IB/CM is used for this channel.
  * @processing_wait_list: Whether or not cmd_wait_list is being processed.
  * @ioctx_ring:    Send ring.
  * @ioctx_recv_ring: Receive I/O context ring.
@@ -280,6 +282,9 @@ struct srpt_rdma_ch {
 		struct {
 			struct ib_cm_id		*cm_id;
 		} ib_cm;
+		struct {
+			struct rdma_cm_id	*cm_id;
+		} rdma_cm;
 	};
 	struct ib_cq		*cq;
 	struct ib_cqe		zw_cqe;
@@ -300,9 +305,10 @@ struct srpt_rdma_ch {
 	struct list_head	list;
 	struct list_head	cmd_wait_list;
 	uint16_t		pkey;
+	bool			using_rdma_cm;
 	bool			processing_wait_list;
 	struct se_session	*sess;
-	u8			sess_name[24];
+	u8			sess_name[40];
 	struct work_struct	release_work;
 };