1 files changed, 150 insertions, 237 deletions

diff --git a/net/9p/trans_rdma.c b/net/9p/trans_rdma.c
index 928f2bb9bf8d..4d406479f83b 100644
--- a/net/9p/trans_rdma.c
+++ b/net/9p/trans_rdma.c
@@ -1,6 +1,5 @@
+// SPDX-License-Identifier: GPL-2.0-only
 /*
- * linux/fs/9p/trans_rdma.c
- *
  * RDMA transport layer based on the trans_fd.c implementation.
  *
  *  Copyright (C) 2008 by Tom Tucker <tom@opengridcomputing.com>
@@ -8,22 +7,6 @@
  *  Copyright (C) 2004-2005 by Latchesar Ionkov <lucho@ionkov.net>
  *  Copyright (C) 2004-2008 by Eric Van Hensbergen <ericvh@gmail.com>
  *  Copyright (C) 1997-2002 by Ron Minnich <rminnich@sarnoff.com>
- *
- *  This program is free software; you can redistribute it and/or modify
- *  it under the terms of the GNU General Public License version 2
- *  as published by the Free Software Foundation.
- *
- *  This program is distributed in the hope that it will be useful,
- *  but WITHOUT ANY WARRANTY; without even the implied warranty of
- *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
- *  GNU General Public License for more details.
- *
- *  You should have received a copy of the GNU General Public License
- *  along with this program; if not, write to:
- *  Free Software Foundation
- *  51 Franklin Street, Fifth Floor
- *  Boston, MA  02111-1301  USA
- *
  */
 
 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
@@ -38,25 +21,21 @@
 #include <linux/un.h>
 #include <linux/uaccess.h>
 #include <linux/inet.h>
-#include <linux/idr.h>
 #include <linux/file.h>
-#include <linux/parser.h>
+#include <linux/fs_context.h>
 #include <linux/semaphore.h>
 #include <linux/slab.h>
+#include <linux/seq_file.h>
 #include <net/9p/9p.h>
 #include <net/9p/client.h>
 #include <net/9p/transport.h>
 #include <rdma/ib_verbs.h>
 #include <rdma/rdma_cm.h>
 
-#define P9_PORT			5640
-#define P9_RDMA_SQ_DEPTH	32
-#define P9_RDMA_RQ_DEPTH	32
 #define P9_RDMA_SEND_SGE	4
 #define P9_RDMA_RECV_SGE	4
 #define P9_RDMA_IRD		0
 #define P9_RDMA_ORD		0
-#define P9_RDMA_TIMEOUT		30000		/* 30 seconds */
 #define P9_RDMA_MAXSIZE		(1024*1024)	/* 1MB */
 
 /**
@@ -67,9 +46,9 @@
  * @pd: Protection Domain pointer
  * @qp: Queue Pair pointer
  * @cq: Completion Queue pointer
- * @dm_mr: DMA Memory Region pointer
- * @lkey: The local access only memory region key
  * @timeout: Number of uSecs to wait for connection management events
+ * @privport: Whether a privileged port may be used
+ * @port: The port to use
  * @sq_depth: The depth of the Send Queue
  * @sq_sem: Semaphore for the SQ
  * @rq_depth: The depth of the Receive Queue.
@@ -94,9 +73,9 @@ struct p9_trans_rdma {
 	struct ib_pd *pd;
 	struct ib_qp *qp;
 	struct ib_cq *cq;
-	struct ib_mr *dma_mr;
-	u32 lkey;
 	long timeout;
+	bool privport;
+	u16 port;
 	int sq_depth;
 	struct semaphore sq_sem;
 	int rq_depth;
@@ -108,117 +87,39 @@ struct p9_trans_rdma {
 	struct completion cm_done;
 };
 
+struct p9_rdma_req;
+
 /**
- * p9_rdma_context - Keeps track of in-process WR
+ * struct p9_rdma_context - Keeps track of in-process WR
  *
- * @wc_op: The original WR op for when the CQE completes in error.
+ * @cqe: completion queue entry
  * @busa: Bus address to unmap when the WR completes
  * @req: Keeps track of requests (send)
  * @rc: Keepts track of replies (receive)
  */
-struct p9_rdma_req;
 struct p9_rdma_context {
-	enum ib_wc_opcode wc_op;
+	struct ib_cqe cqe;
 	dma_addr_t busa;
 	union {
 		struct p9_req_t *req;
-		struct p9_fcall *rc;
+		struct p9_fcall rc;
 	};
 };
 
-/**
- * p9_rdma_opts - Collection of mount options
- * @port: port of connection
- * @sq_depth: The requested depth of the SQ. This really doesn't need
- * to be any deeper than the number of threads used in the client
- * @rq_depth: The depth of the RQ. Should be greater than or equal to SQ depth
- * @timeout: Time to wait in msecs for CM events
- */
-struct p9_rdma_opts {
-	short port;
-	int sq_depth;
-	int rq_depth;
-	long timeout;
-};
-
-/*
- * Option Parsing (code inspired by NFS code)
- */
-enum {
-	/* Options that take integer arguments */
-	Opt_port, Opt_rq_depth, Opt_sq_depth, Opt_timeout, Opt_err,
-};
-
-static match_table_t tokens = {
-	{Opt_port, "port=%u"},
-	{Opt_sq_depth, "sq=%u"},
-	{Opt_rq_depth, "rq=%u"},
-	{Opt_timeout, "timeout=%u"},
-	{Opt_err, NULL},
-};
-
-/**
- * parse_opts - parse mount options into rdma options structure
- * @params: options string passed from mount
- * @opts: rdma transport-specific structure to parse options into
- *
- * Returns 0 upon success, -ERRNO upon failure
- */
-static int parse_opts(char *params, struct p9_rdma_opts *opts)
+static int p9_rdma_show_options(struct seq_file *m, struct p9_client *clnt)
 {
-	char *p;
-	substring_t args[MAX_OPT_ARGS];
-	int option;
-	char *options, *tmp_options;
-
-	opts->port = P9_PORT;
-	opts->sq_depth = P9_RDMA_SQ_DEPTH;
-	opts->rq_depth = P9_RDMA_RQ_DEPTH;
-	opts->timeout = P9_RDMA_TIMEOUT;
-
-	if (!params)
-		return 0;
-
-	tmp_options = kstrdup(params, GFP_KERNEL);
-	if (!tmp_options) {
-		p9_debug(P9_DEBUG_ERROR,
-			 "failed to allocate copy of option string\n");
-		return -ENOMEM;
-	}
-	options = tmp_options;
-
-	while ((p = strsep(&options, ",")) != NULL) {
-		int token;
-		int r;
-		if (!*p)
-			continue;
-		token = match_token(p, tokens, args);
-		r = match_int(&args[0], &option);
-		if (r < 0) {
-			p9_debug(P9_DEBUG_ERROR,
-				 "integer field, but no integer?\n");
-			continue;
-		}
-		switch (token) {
-		case Opt_port:
-			opts->port = option;
-			break;
-		case Opt_sq_depth:
-			opts->sq_depth = option;
-			break;
-		case Opt_rq_depth:
-			opts->rq_depth = option;
-			break;
-		case Opt_timeout:
-			opts->timeout = option;
-			break;
-		default:
-			continue;
-		}
-	}
-	/* RQ must be at least as large as the SQ */
-	opts->rq_depth = max(opts->rq_depth, opts->sq_depth);
-	kfree(tmp_options);
+	struct p9_trans_rdma *rdma = clnt->trans;
+
+	if (rdma->port != P9_RDMA_PORT)
+		seq_printf(m, ",port=%u", rdma->port);
+	if (rdma->sq_depth != P9_RDMA_SQ_DEPTH)
+		seq_printf(m, ",sq=%u", rdma->sq_depth);
+	if (rdma->rq_depth != P9_RDMA_RQ_DEPTH)
+		seq_printf(m, ",rq=%u", rdma->rq_depth);
+	if (rdma->timeout != P9_RDMA_TIMEOUT)
+		seq_printf(m, ",timeout=%lu", rdma->timeout);
+	if (rdma->privport)
+		seq_puts(m, ",privport");
 	return 0;
 }
 
@@ -246,8 +147,7 @@ p9_cm_event_handler(struct rdma_cm_id *id, struct rdma_cm_event *event)
 	case RDMA_CM_EVENT_DISCONNECTED:
 		if (rdma)
 			rdma->state = P9_RDMA_CLOSED;
-		if (c)
-			c->status = Disconnected;
+		c->status = Disconnected;
 		break;
 
 	case RDMA_CM_EVENT_TIMEWAIT_EXIT:
@@ -275,9 +175,12 @@ p9_cm_event_handler(struct rdma_cm_id *id, struct rdma_cm_event *event)
 }
 
 static void
-handle_recv(struct p9_client *client, struct p9_trans_rdma *rdma,
-	    struct p9_rdma_context *c, enum ib_wc_status status, u32 byte_len)
+recv_done(struct ib_cq *cq, struct ib_wc *wc)
 {
+	struct p9_client *client = cq->cq_context;
+	struct p9_trans_rdma *rdma = client->trans;
+	struct p9_rdma_context *c =
+		container_of(wc->wr_cqe, struct p9_rdma_context, cqe);
 	struct p9_req_t *req;
 	int err = 0;
 	int16_t tag;
@@ -286,10 +189,11 @@ handle_recv(struct p9_client *client, struct p9_trans_rdma *rdma,
 	ib_dma_unmap_single(rdma->cm_id->device, c->busa, client->msize,
 							 DMA_FROM_DEVICE);
 
-	if (status != IB_WC_SUCCESS)
+	if (wc->status != IB_WC_SUCCESS)
 		goto err_out;
 
-	err = p9_parse_header(c->rc, NULL, NULL, &tag, 1);
+	c->rc.size = wc->byte_len;
+	err = p9_parse_header(&c->rc, NULL, NULL, &tag, 1);
 	if (err)
 		goto err_out;
 
@@ -299,30 +203,42 @@ handle_recv(struct p9_client *client, struct p9_trans_rdma *rdma,
 
 	/* Check that we have not yet received a reply for this request.
 	 */
-	if (unlikely(req->rc)) {
+	if (unlikely(req->rc.sdata)) {
 		pr_err("Duplicate reply for request %d", tag);
 		goto err_out;
 	}
 
-	req->rc = c->rc;
-	req->status = REQ_STATUS_RCVD;
-	p9_client_cb(client, req);
+	req->rc.size = c->rc.size;
+	req->rc.sdata = c->rc.sdata;
+	p9_client_cb(client, req, REQ_STATUS_RCVD);
 
+ out:
+	up(&rdma->rq_sem);
+	kfree(c);
 	return;
 
  err_out:
-	p9_debug(P9_DEBUG_ERROR, "req %p err %d status %d\n", req, err, status);
+	p9_debug(P9_DEBUG_ERROR, "req %p err %d status %d\n",
+			req, err, wc->status);
 	rdma->state = P9_RDMA_FLUSHING;
 	client->status = Disconnected;
+	goto out;
 }
 
 static void
-handle_send(struct p9_client *client, struct p9_trans_rdma *rdma,
-	    struct p9_rdma_context *c, enum ib_wc_status status, u32 byte_len)
+send_done(struct ib_cq *cq, struct ib_wc *wc)
 {
+	struct p9_client *client = cq->cq_context;
+	struct p9_trans_rdma *rdma = client->trans;
+	struct p9_rdma_context *c =
+		container_of(wc->wr_cqe, struct p9_rdma_context, cqe);
+
 	ib_dma_unmap_single(rdma->cm_id->device,
-			    c->busa, c->req->tc->size,
+			    c->busa, c->req->tc.size,
 			    DMA_TO_DEVICE);
+	up(&rdma->sq_sem);
+	p9_req_put(client, c->req);
+	kfree(c);
 }
 
 static void qp_event_handler(struct ib_event *event, void *context)
@@ -331,50 +247,11 @@ static void qp_event_handler(struct ib_event *event, void *context)
 		 event->event, context);
 }
 
-static void cq_comp_handler(struct ib_cq *cq, void *cq_context)
-{
-	struct p9_client *client = cq_context;
-	struct p9_trans_rdma *rdma = client->trans;
-	int ret;
-	struct ib_wc wc;
-
-	ib_req_notify_cq(rdma->cq, IB_CQ_NEXT_COMP);
-	while ((ret = ib_poll_cq(cq, 1, &wc)) > 0) {
-		struct p9_rdma_context *c = (void *) (unsigned long) wc.wr_id;
-
-		switch (c->wc_op) {
-		case IB_WC_RECV:
-			handle_recv(client, rdma, c, wc.status, wc.byte_len);
-			up(&rdma->rq_sem);
-			break;
-
-		case IB_WC_SEND:
-			handle_send(client, rdma, c, wc.status, wc.byte_len);
-			up(&rdma->sq_sem);
-			break;
-
-		default:
-			pr_err("unexpected completion type, c->wc_op=%d, wc.opcode=%d, status=%d\n",
-			       c->wc_op, wc.opcode, wc.status);
-			break;
-		}
-		kfree(c);
-	}
-}
-
-static void cq_event_handler(struct ib_event *e, void *v)
-{
-	p9_debug(P9_DEBUG_ERROR, "CQ event %d context %p\n", e->event, v);
-}
-
 static void rdma_destroy_trans(struct p9_trans_rdma *rdma)
 {
 	if (!rdma)
 		return;
 
-	if (rdma->dma_mr && !IS_ERR(rdma->dma_mr))
-		ib_dereg_mr(rdma->dma_mr);
-
 	if (rdma->qp && !IS_ERR(rdma->qp))
 		ib_destroy_qp(rdma->qp);
 
@@ -382,7 +259,7 @@ static void rdma_destroy_trans(struct p9_trans_rdma *rdma)
 		ib_dealloc_pd(rdma->pd);
 
 	if (rdma->cq && !IS_ERR(rdma->cq))
-		ib_destroy_cq(rdma->cq);
+		ib_free_cq(rdma->cq);
 
 	if (rdma->cm_id && !IS_ERR(rdma->cm_id))
 		rdma_destroy_id(rdma->cm_id);
@@ -394,25 +271,32 @@ static int
 post_recv(struct p9_client *client, struct p9_rdma_context *c)
 {
 	struct p9_trans_rdma *rdma = client->trans;
-	struct ib_recv_wr wr, *bad_wr;
+	struct ib_recv_wr wr;
 	struct ib_sge sge;
+	int ret;
 
 	c->busa = ib_dma_map_single(rdma->cm_id->device,
-				    c->rc->sdata, client->msize,
+				    c->rc.sdata, client->msize,
 				    DMA_FROM_DEVICE);
 	if (ib_dma_mapping_error(rdma->cm_id->device, c->busa))
 		goto error;
 
+	c->cqe.done = recv_done;
+
 	sge.addr = c->busa;
 	sge.length = client->msize;
-	sge.lkey = rdma->lkey;
+	sge.lkey = rdma->pd->local_dma_lkey;
 
 	wr.next = NULL;
-	c->wc_op = IB_WC_RECV;
-	wr.wr_id = (unsigned long) c;
+	wr.wr_cqe = &c->cqe;
 	wr.sg_list = &sge;
 	wr.num_sge = 1;
-	return ib_post_recv(rdma->qp, &wr, &bad_wr);
+
+	ret = ib_post_recv(rdma->qp, &wr, NULL);
+	if (ret)
+		ib_dma_unmap_single(rdma->cm_id->device, c->busa,
+				    client->msize, DMA_FROM_DEVICE);
+	return ret;
 
  error:
 	p9_debug(P9_DEBUG_ERROR, "EIO\n");
@@ -422,7 +306,7 @@ post_recv(struct p9_client *client, struct p9_rdma_context *c)
 static int rdma_request(struct p9_client *client, struct p9_req_t *req)
 {
 	struct p9_trans_rdma *rdma = client->trans;
-	struct ib_send_wr wr, *bad_wr;
+	struct ib_send_wr wr;
 	struct ib_sge sge;
 	int err = 0;
 	unsigned long flags;
@@ -439,9 +323,9 @@ static int rdma_request(struct p9_client *client, struct p9_req_t *req)
 	 **/
 	if (unlikely(atomic_read(&rdma->excess_rc) > 0)) {
 		if ((atomic_sub_return(1, &rdma->excess_rc) >= 0)) {
-			/* Got one ! */
-			kfree(req->rc);
-			req->rc = NULL;
+			/* Got one! */
+			p9_fcall_fini(&req->rc);
+			req->rc.sdata = NULL;
 			goto dont_need_post_recv;
 		} else {
 			/* We raced and lost. */
@@ -455,7 +339,7 @@ static int rdma_request(struct p9_client *client, struct p9_req_t *req)
 		err = -ENOMEM;
 		goto recv_error;
 	}
-	rpl_context->rc = req->rc;
+	rpl_context->rc.sdata = req->rc.sdata;
 
 	/*
 	 * Post a receive buffer for this request. We need to ensure
@@ -471,11 +355,11 @@ static int rdma_request(struct p9_client *client, struct p9_req_t *req)
 
 	err = post_recv(client, rpl_context);
 	if (err) {
-		p9_debug(P9_DEBUG_FCALL, "POST RECV failed\n");
+		p9_debug(P9_DEBUG_ERROR, "POST RECV failed: %d\n", err);
 		goto recv_error;
 	}
 	/* remove posted receive buffer from request structure */
-	req->rc = NULL;
+	req->rc.sdata = NULL;
 
 dont_need_post_recv:
 	/* Post the request */
@@ -487,20 +371,21 @@ dont_need_post_recv:
 	c->req = req;
 
 	c->busa = ib_dma_map_single(rdma->cm_id->device,
-				    c->req->tc->sdata, c->req->tc->size,
+				    c->req->tc.sdata, c->req->tc.size,
 				    DMA_TO_DEVICE);
 	if (ib_dma_mapping_error(rdma->cm_id->device, c->busa)) {
 		err = -EIO;
 		goto send_error;
 	}
 
+	c->cqe.done = send_done;
+
 	sge.addr = c->busa;
-	sge.length = c->req->tc->size;
-	sge.lkey = rdma->lkey;
+	sge.length = c->req->tc.size;
+	sge.lkey = rdma->pd->local_dma_lkey;
 
 	wr.next = NULL;
-	c->wc_op = IB_WC_SEND;
-	wr.wr_id = (unsigned long) c;
+	wr.wr_cqe = &c->cqe;
 	wr.opcode = IB_WR_SEND;
 	wr.send_flags = IB_SEND_SIGNALED;
 	wr.sg_list = &sge;
@@ -508,18 +393,27 @@ dont_need_post_recv:
 
 	if (down_interruptible(&rdma->sq_sem)) {
 		err = -EINTR;
-		goto send_error;
+		goto dma_unmap;
 	}
 
-	err = ib_post_send(rdma->qp, &wr, &bad_wr);
+	/* Mark request as `sent' *before* we actually send it,
+	 * because doing if after could erase the REQ_STATUS_RCVD
+	 * status in case of a very fast reply.
+	 */
+	WRITE_ONCE(req->status, REQ_STATUS_SENT);
+	err = ib_post_send(rdma->qp, &wr, NULL);
 	if (err)
-		goto send_error;
+		goto dma_unmap;
 
 	/* Success */
 	return 0;
 
+dma_unmap:
+	ib_dma_unmap_single(rdma->cm_id->device, c->busa,
+			    c->req->tc.size, DMA_TO_DEVICE);
  /* Handle errors that happened during or while preparing the send: */
  send_error:
+	WRITE_ONCE(req->status, REQ_STATUS_ERROR);
 	kfree(c);
 	p9_debug(P9_DEBUG_ERROR, "Error %d in rdma_request()\n", err);
 
@@ -533,7 +427,7 @@ dont_need_post_recv:
  recv_error:
 	kfree(rpl_context);
 	spin_lock_irqsave(&rdma->req_lock, flags);
-	if (rdma->state < P9_RDMA_CLOSING) {
+	if (err != -EINTR && rdma->state < P9_RDMA_CLOSING) {
 		rdma->state = P9_RDMA_CLOSING;
 		spin_unlock_irqrestore(&rdma->req_lock, flags);
 		rdma_disconnect(rdma->cm_id);
@@ -570,6 +464,8 @@ static struct p9_trans_rdma *alloc_rdma(struct p9_rdma_opts *opts)
 	if (!rdma)
 		return NULL;
 
+	rdma->port = opts->port;
+	rdma->privport = opts->privport;
 	rdma->sq_depth = opts->sq_depth;
 	rdma->rq_depth = opts->rq_depth;
 	rdma->timeout = opts->timeout;
@@ -582,9 +478,11 @@ static struct p9_trans_rdma *alloc_rdma(struct p9_rdma_opts *opts)
 	return rdma;
 }
 
-/* its not clear to me we can do anything after send has been posted */
 static int rdma_cancel(struct p9_client *client, struct p9_req_t *req)
 {
+	/* Nothing to do here.
+	 * We will take care of it (if we have to) in rdma_cancelled()
+	 */
 	return 1;
 }
 
@@ -594,31 +492,48 @@ static int rdma_cancel(struct p9_client *client, struct p9_req_t *req)
 static int rdma_cancelled(struct p9_client *client, struct p9_req_t *req)
 {
 	struct p9_trans_rdma *rdma = client->trans;
-
 	atomic_inc(&rdma->excess_rc);
 	return 0;
 }
 
+static int p9_rdma_bind_privport(struct p9_trans_rdma *rdma)
+{
+	struct sockaddr_in cl = {
+		.sin_family = AF_INET,
+		.sin_addr.s_addr = htonl(INADDR_ANY),
+	};
+	int port, err = -EINVAL;
+
+	for (port = P9_DEF_MAX_RESVPORT; port >= P9_DEF_MIN_RESVPORT; port--) {
+		cl.sin_port = htons((ushort)port);
+		err = rdma_bind_addr(rdma->cm_id, (struct sockaddr *)&cl);
+		if (err != -EADDRINUSE)
+			break;
+	}
+	return err;
+}
+
 /**
- * trans_create_rdma - Transport method for creating atransport instance
+ * rdma_create_trans - Transport method for creating a transport instance
  * @client: client instance
- * @addr: IP address string
- * @args: Mount options string
+ * @fc: The filesystem context
  */
 static int
-rdma_create_trans(struct p9_client *client, const char *addr, char *args)
+rdma_create_trans(struct p9_client *client, struct fs_context *fc)
 {
+	const char *addr = fc->source;
+	struct v9fs_context *ctx = fc->fs_private;
+	struct p9_rdma_opts opts = ctx->rdma_opts;
 	int err;
-	struct p9_rdma_opts opts;
 	struct p9_trans_rdma *rdma;
 	struct rdma_conn_param conn_param;
 	struct ib_qp_init_attr qp_attr;
-	struct ib_device_attr devattr;
 
-	/* Parse the transport specific mount options */
-	err = parse_opts(args, &opts);
-	if (err < 0)
-		return err;
+	if (addr == NULL)
+		return -EINVAL;
+
+	/* options are already parsed, in the fs context */
+	opts = ctx->rdma_opts;
 
 	/* Create and initialize the RDMA transport structure */
 	rdma = alloc_rdma(&opts);
@@ -626,14 +541,24 @@ rdma_create_trans(struct p9_client *client, const char *addr, char *args)
 		return -ENOMEM;
 
 	/* Create the RDMA CM ID */
-	rdma->cm_id = rdma_create_id(p9_cm_event_handler, client, RDMA_PS_TCP,
-				     IB_QPT_RC);
+	rdma->cm_id = rdma_create_id(&init_net, p9_cm_event_handler, client,
+				     RDMA_PS_TCP, IB_QPT_RC);
 	if (IS_ERR(rdma->cm_id))
 		goto error;
 
 	/* Associate the client with the transport */
 	client->trans = rdma;
 
+	/* Bind to a privileged port if we need to */
+	if (opts.privport) {
+		err = p9_rdma_bind_privport(rdma);
+		if (err < 0) {
+			pr_err("%s (%d): problem binding to privport: %d\n",
+			       __func__, task_pid_nr(current), -err);
+			goto error;
+		}
+	}
+
 	/* Resolve the server's address */
 	rdma->addr.sin_family = AF_INET;
 	rdma->addr.sin_addr.s_addr = in_aton(addr);
@@ -655,35 +580,18 @@ rdma_create_trans(struct p9_client *client, const char *addr, char *args)
 	if (err || (rdma->state != P9_RDMA_ROUTE_RESOLVED))
 		goto error;
 
-	/* Query the device attributes */
-	err = ib_query_device(rdma->cm_id->device, &devattr);
-	if (err)
-		goto error;
-
 	/* Create the Completion Queue */
-	rdma->cq = ib_create_cq(rdma->cm_id->device, cq_comp_handler,
-				cq_event_handler, client,
-				opts.sq_depth + opts.rq_depth + 1, 0);
+	rdma->cq = ib_alloc_cq_any(rdma->cm_id->device, client,
+				   opts.sq_depth + opts.rq_depth + 1,
+				   IB_POLL_SOFTIRQ);
 	if (IS_ERR(rdma->cq))
 		goto error;
-	ib_req_notify_cq(rdma->cq, IB_CQ_NEXT_COMP);
 
 	/* Create the Protection Domain */
-	rdma->pd = ib_alloc_pd(rdma->cm_id->device);
+	rdma->pd = ib_alloc_pd(rdma->cm_id->device, 0);
 	if (IS_ERR(rdma->pd))
 		goto error;
 
-	/* Cache the DMA lkey in the transport */
-	rdma->dma_mr = NULL;
-	if (devattr.device_cap_flags & IB_DEVICE_LOCAL_DMA_LKEY)
-		rdma->lkey = rdma->cm_id->device->local_dma_lkey;
-	else {
-		rdma->dma_mr = ib_get_dma_mr(rdma->pd, IB_ACCESS_LOCAL_WRITE);
-		if (IS_ERR(rdma->dma_mr))
-			goto error;
-		rdma->lkey = rdma->dma_mr->lkey;
-	}
-
 	/* Create the Queue Pair */
 	memset(&qp_attr, 0, sizeof qp_attr);
 	qp_attr.event_handler = qp_event_handler;
@@ -726,12 +634,16 @@ error:
 static struct p9_trans_module p9_rdma_trans = {
 	.name = "rdma",
 	.maxsize = P9_RDMA_MAXSIZE,
-	.def = 0,
+	.pooled_rbuffers = true,
+	.def = false,
+	.supports_vmalloc = false,
 	.owner = THIS_MODULE,
 	.create = rdma_create_trans,
 	.close = rdma_close,
 	.request = rdma_request,
 	.cancel = rdma_cancel,
+	.cancelled = rdma_cancelled,
+	.show_options = p9_rdma_show_options,
 };
 
 /**
@@ -750,6 +662,7 @@ static void __exit p9_trans_rdma_exit(void)
 
 module_init(p9_trans_rdma_init);
 module_exit(p9_trans_rdma_exit);
+MODULE_ALIAS_9P("rdma");
 
 MODULE_AUTHOR("Tom Tucker <tom@opengridcomputing.com>");
 MODULE_DESCRIPTION("RDMA Transport for 9P");