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

Pull rdma updates from Jason Gunthorpe:
 "This cycle we got a new RDMA driver "ERDMA" for the Alibaba cloud
  environment. Otherwise the changes are dominated by rxe fixes.

  There is another RDMA driver on the list that might get merged next
  cycle, 'MANA' for the Azure cloud environment.

  Summary:

   - Bug fixes and small features for irdma, hns, siw, qedr, hfi1, mlx5

   - General spelling/grammer fixes

   - rdma cm can follow changes in neighbours for control packets

   - Significant amounts of rxe fixes and spec compliance changes

   - Use the modern NAPI API

   - Use the bitmap API instead of open coding

   - Performance improvements for rtrs

   - Add the ERDMA driver for Alibaba cloud

   - Fix a use after free bug in SRP"

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma: (99 commits)
  RDMA/ib_srpt: Unify checking rdma_cm_id condition in srpt_cm_req_recv()
  RDMA/rxe: Fix error unwind in rxe_create_qp()
  RDMA/mlx5: Add missing check for return value in get namespace flow
  RDMA/rxe: Split qp state for requester and completer
  RDMA/rxe: Generate error completion for error requester QP state
  RDMA/rxe: Update wqe_index for each wqe error completion
  RDMA/srpt: Fix a use-after-free
  RDMA/srpt: Introduce a reference count in struct srpt_device
  RDMA/srpt: Duplicate port name members
  IB/qib: Fix repeated "in" within comments
  RDMA/erdma: Add driver to kernel build environment
  RDMA/erdma: Add the ABI definitions
  RDMA/erdma: Add the erdma module
  RDMA/erdma: Add connection management (CM) support
  RDMA/erdma: Add verbs implementation
  RDMA/erdma: Add verbs header file
  RDMA/erdma: Add event queue implementation
  RDMA/erdma: Add cmdq implementation
  RDMA/erdma: Add main include file
  RDMA/erdma: Add the hardware related definitions
  ...

12 files changed, 215 insertions, 139 deletions

diff --git a/drivers/infiniband/ulp/ipoib/ipoib_ib.c b/drivers/infiniband/ulp/ipoib/ipoib_ib.c
index f7995519bbc8..ed25061fac62 100644
--- a/drivers/infiniband/ulp/ipoib/ipoib_ib.c
+++ b/drivers/infiniband/ulp/ipoib/ipoib_ib.c
@@ -1109,7 +1109,7 @@ static bool ipoib_dev_addr_changed_valid(struct ipoib_dev_priv *priv)
 	 * if he sets the device address back to be based on GID index 0,
 	 * he no longer wishs to control it.
 	 *
-	 * If the user doesn't control the the device address,
+	 * If the user doesn't control the device address,
 	 * IPOIB_FLAG_DEV_ADDR_SET is set and ib_find_gid failed it means
 	 * the port GUID has changed and GID at index 0 has changed
 	 * so we need to change priv->local_gid and priv->dev->dev_addr
diff --git a/drivers/infiniband/ulp/ipoib/ipoib_main.c b/drivers/infiniband/ulp/ipoib/ipoib_main.c
index 2a8961b685c2..a4904371e2db 100644
--- a/drivers/infiniband/ulp/ipoib/ipoib_main.c
+++ b/drivers/infiniband/ulp/ipoib/ipoib_main.c
@@ -1664,8 +1664,10 @@ static void ipoib_napi_add(struct net_device *dev)
 {
 	struct ipoib_dev_priv *priv = ipoib_priv(dev);
 
-	netif_napi_add(dev, &priv->recv_napi, ipoib_rx_poll, IPOIB_NUM_WC);
-	netif_napi_add(dev, &priv->send_napi, ipoib_tx_poll, MAX_SEND_CQE);
+	netif_napi_add_weight(dev, &priv->recv_napi, ipoib_rx_poll,
+			      IPOIB_NUM_WC);
+	netif_napi_add_weight(dev, &priv->send_napi, ipoib_tx_poll,
+			      MAX_SEND_CQE);
 }
 
 static void ipoib_napi_del(struct net_device *dev)
diff --git a/drivers/infiniband/ulp/iser/iser_verbs.c b/drivers/infiniband/ulp/iser/iser_verbs.c
index c08f2d9133b6..a00ca117303a 100644
--- a/drivers/infiniband/ulp/iser/iser_verbs.c
+++ b/drivers/infiniband/ulp/iser/iser_verbs.c
@@ -246,6 +246,7 @@ static int iser_create_ib_conn_res(struct ib_conn *ib_conn)
 	device = ib_conn->device;
 	ib_dev = device->ib_device;
 
+	/* +1 for drain */
 	if (ib_conn->pi_support)
 		max_send_wr = ISER_QP_SIG_MAX_REQ_DTOS + 1;
 	else
@@ -267,7 +268,8 @@ static int iser_create_ib_conn_res(struct ib_conn *ib_conn)
 	init_attr.qp_context = (void *)ib_conn;
 	init_attr.send_cq = ib_conn->cq;
 	init_attr.recv_cq = ib_conn->cq;
-	init_attr.cap.max_recv_wr = ISER_QP_MAX_RECV_DTOS;
+	/* +1 for drain */
+	init_attr.cap.max_recv_wr = ISER_QP_MAX_RECV_DTOS + 1;
 	init_attr.cap.max_send_sge = 2;
 	init_attr.cap.max_recv_sge = 1;
 	init_attr.sq_sig_type = IB_SIGNAL_REQ_WR;
@@ -485,7 +487,7 @@ int iser_conn_terminate(struct iser_conn *iser_conn)
 				 iser_conn, err);
 
 		/* block until all flush errors are consumed */
-		ib_drain_sq(ib_conn->qp);
+		ib_drain_qp(ib_conn->qp);
 	}
 
 	return 1;
diff --git a/drivers/infiniband/ulp/rtrs/rtrs-clt-stats.c b/drivers/infiniband/ulp/rtrs/rtrs-clt-stats.c
index 385a19846c24..1e6ffafa2db3 100644
--- a/drivers/infiniband/ulp/rtrs/rtrs-clt-stats.c
+++ b/drivers/infiniband/ulp/rtrs/rtrs-clt-stats.c
@@ -32,11 +32,7 @@ void rtrs_clt_update_wc_stats(struct rtrs_clt_con *con)
 
 void rtrs_clt_inc_failover_cnt(struct rtrs_clt_stats *stats)
 {
-	struct rtrs_clt_stats_pcpu *s;
-
-	s = get_cpu_ptr(stats->pcpu_stats);
-	s->rdma.failover_cnt++;
-	put_cpu_ptr(stats->pcpu_stats);
+	this_cpu_inc(stats->pcpu_stats->rdma.failover_cnt);
 }
 
 int rtrs_clt_stats_migration_from_cnt_to_str(struct rtrs_clt_stats *stats, char *buf)
@@ -169,12 +165,8 @@ int rtrs_clt_reset_all_stats(struct rtrs_clt_stats *s, bool enable)
 static inline void rtrs_clt_update_rdma_stats(struct rtrs_clt_stats *stats,
 					       size_t size, int d)
 {
-	struct rtrs_clt_stats_pcpu *s;
-
-	s = get_cpu_ptr(stats->pcpu_stats);
-	s->rdma.dir[d].cnt++;
-	s->rdma.dir[d].size_total += size;
-	put_cpu_ptr(stats->pcpu_stats);
+	this_cpu_inc(stats->pcpu_stats->rdma.dir[d].cnt);
+	this_cpu_add(stats->pcpu_stats->rdma.dir[d].size_total, size);
 }
 
 void rtrs_clt_update_all_stats(struct rtrs_clt_io_req *req, int dir)
diff --git a/drivers/infiniband/ulp/rtrs/rtrs-clt.c b/drivers/infiniband/ulp/rtrs/rtrs-clt.c
index 9809c3883979..baecde41d126 100644
--- a/drivers/infiniband/ulp/rtrs/rtrs-clt.c
+++ b/drivers/infiniband/ulp/rtrs/rtrs-clt.c
@@ -740,25 +740,25 @@ struct path_it {
 	struct rtrs_clt_path *(*next_path)(struct path_it *it);
 };
 
-/**
- * list_next_or_null_rr_rcu - get next list element in round-robin fashion.
+/*
+ * rtrs_clt_get_next_path_or_null - get clt path from the list or return NULL
  * @head:	the head for the list.
- * @ptr:        the list head to take the next element from.
- * @type:       the type of the struct this is embedded in.
- * @memb:       the name of the list_head within the struct.
+ * @clt_path:	The element to take the next clt_path from.
  *
- * Next element returned in round-robin fashion, i.e. head will be skipped,
+ * Next clt path returned in round-robin fashion, i.e. head will be skipped,
  * but if list is observed as empty, NULL will be returned.
  *
- * This primitive may safely run concurrently with the _rcu list-mutation
+ * This function may safely run concurrently with the _rcu list-mutation
  * primitives such as list_add_rcu() as long as it's guarded by rcu_read_lock().
  */
-#define list_next_or_null_rr_rcu(head, ptr, type, memb) \
-({ \
-	list_next_or_null_rcu(head, ptr, type, memb) ?: \
-		list_next_or_null_rcu(head, READ_ONCE((ptr)->next), \
-				      type, memb); \
-})
+static inline struct rtrs_clt_path *
+rtrs_clt_get_next_path_or_null(struct list_head *head, struct rtrs_clt_path *clt_path)
+{
+	return list_next_or_null_rcu(head, &clt_path->s.entry, typeof(*clt_path), s.entry) ?:
+				     list_next_or_null_rcu(head,
+							   READ_ONCE((&clt_path->s.entry)->next),
+							   typeof(*clt_path), s.entry);
+}
 
 /**
  * get_next_path_rr() - Returns path in round-robin fashion.
@@ -789,10 +789,8 @@ static struct rtrs_clt_path *get_next_path_rr(struct path_it *it)
 		path = list_first_or_null_rcu(&clt->paths_list,
 					      typeof(*path), s.entry);
 	else
-		path = list_next_or_null_rr_rcu(&clt->paths_list,
-						&path->s.entry,
-						typeof(*path),
-						s.entry);
+		path = rtrs_clt_get_next_path_or_null(&clt->paths_list, path);
+
 	rcu_assign_pointer(*ppcpu_path, path);
 
 	return path;
@@ -1403,8 +1401,7 @@ static int alloc_permits(struct rtrs_clt_sess *clt)
 	unsigned int chunk_bits;
 	int err, i;
 
-	clt->permits_map = kcalloc(BITS_TO_LONGS(clt->queue_depth),
-				   sizeof(long), GFP_KERNEL);
+	clt->permits_map = bitmap_zalloc(clt->queue_depth, GFP_KERNEL);
 	if (!clt->permits_map) {
 		err = -ENOMEM;
 		goto out_err;
@@ -1426,7 +1423,7 @@ static int alloc_permits(struct rtrs_clt_sess *clt)
 	return 0;
 
 err_map:
-	kfree(clt->permits_map);
+	bitmap_free(clt->permits_map);
 	clt->permits_map = NULL;
 out_err:
 	return err;
@@ -1434,13 +1431,11 @@ out_err:
 
 static void free_permits(struct rtrs_clt_sess *clt)
 {
-	if (clt->permits_map) {
-		size_t sz = clt->queue_depth;
-
+	if (clt->permits_map)
 		wait_event(clt->permits_wait,
-			   find_first_bit(clt->permits_map, sz) >= sz);
-	}
-	kfree(clt->permits_map);
+			   bitmap_empty(clt->permits_map, clt->queue_depth));
+
+	bitmap_free(clt->permits_map);
 	clt->permits_map = NULL;
 	kfree(clt->permits);
 	clt->permits = NULL;
@@ -2277,8 +2272,7 @@ static void rtrs_clt_remove_path_from_arr(struct rtrs_clt_path *clt_path)
 	 * removed.  If @sess is the last element, then @next is NULL.
 	 */
 	rcu_read_lock();
-	next = list_next_or_null_rr_rcu(&clt->paths_list, &clt_path->s.entry,
-					typeof(*next), s.entry);
+	next = rtrs_clt_get_next_path_or_null(&clt->paths_list, clt_path);
 	rcu_read_unlock();
 
 	/*
diff --git a/drivers/infiniband/ulp/rtrs/rtrs-pri.h b/drivers/infiniband/ulp/rtrs/rtrs-pri.h
index 9a1e5c2ae55c..ac0df734eba8 100644
--- a/drivers/infiniband/ulp/rtrs/rtrs-pri.h
+++ b/drivers/infiniband/ulp/rtrs/rtrs-pri.h
@@ -23,6 +23,17 @@
 #define RTRS_PROTO_VER_STRING __stringify(RTRS_PROTO_VER_MAJOR) "." \
 			       __stringify(RTRS_PROTO_VER_MINOR)
 
+/*
+ * Max IB immediate data size is 2^28 (MAX_IMM_PAYL_BITS)
+ * and the minimum chunk size is 4096 (2^12).
+ * So the maximum sess_queue_depth is 65536 (2^16) in theory.
+ * But mempool_create, create_qp and ib_post_send fail with
+ * "cannot allocate memory" error if sess_queue_depth is too big.
+ * Therefore the pratical max value of sess_queue_depth is
+ * somewhere between 1 and 65534 and it depends on the system.
+ */
+#define MAX_SESS_QUEUE_DEPTH 65535
+
 enum rtrs_imm_const {
 	MAX_IMM_TYPE_BITS = 4,
 	MAX_IMM_TYPE_MASK = ((1 << MAX_IMM_TYPE_BITS) - 1),
@@ -46,16 +57,6 @@ enum {
 
 	MAX_PATHS_NUM = 128,
 
-	/*
-	 * Max IB immediate data size is 2^28 (MAX_IMM_PAYL_BITS)
-	 * and the minimum chunk size is 4096 (2^12).
-	 * So the maximum sess_queue_depth is 65536 (2^16) in theory.
-	 * But mempool_create, create_qp and ib_post_send fail with
-	 * "cannot allocate memory" error if sess_queue_depth is too big.
-	 * Therefore the pratical max value of sess_queue_depth is
-	 * somewhere between 1 and 65534 and it depends on the system.
-	 */
-	MAX_SESS_QUEUE_DEPTH = 65535,
 	MIN_CHUNK_SIZE = 8192,
 
 	RTRS_HB_INTERVAL_MS = 5000,
diff --git a/drivers/infiniband/ulp/rtrs/rtrs-srv-stats.c b/drivers/infiniband/ulp/rtrs/rtrs-srv-stats.c
index 44b1c1652131..2aff1213a19d 100644
--- a/drivers/infiniband/ulp/rtrs/rtrs-srv-stats.c
+++ b/drivers/infiniband/ulp/rtrs/rtrs-srv-stats.c
@@ -14,9 +14,14 @@
 int rtrs_srv_reset_rdma_stats(struct rtrs_srv_stats *stats, bool enable)
 {
 	if (enable) {
-		struct rtrs_srv_stats_rdma_stats *r = &stats->rdma_stats;
+		int cpu;
+		struct rtrs_srv_stats_rdma_stats *r;
+
+		for_each_possible_cpu(cpu) {
+			r = per_cpu_ptr(stats->rdma_stats, cpu);
+			memset(r, 0, sizeof(*r));
+		}
 
-		memset(r, 0, sizeof(*r));
 		return 0;
 	}
 
@@ -25,11 +30,22 @@ int rtrs_srv_reset_rdma_stats(struct rtrs_srv_stats *stats, bool enable)
 
 ssize_t rtrs_srv_stats_rdma_to_str(struct rtrs_srv_stats *stats, char *page)
 {
-	struct rtrs_srv_stats_rdma_stats *r = &stats->rdma_stats;
+	int cpu;
+	struct rtrs_srv_stats_rdma_stats sum;
+	struct rtrs_srv_stats_rdma_stats *r;
+
+	memset(&sum, 0, sizeof(sum));
+
+	for_each_possible_cpu(cpu) {
+		r = per_cpu_ptr(stats->rdma_stats, cpu);
+
+		sum.dir[READ].cnt	  += r->dir[READ].cnt;
+		sum.dir[READ].size_total  += r->dir[READ].size_total;
+		sum.dir[WRITE].cnt	  += r->dir[WRITE].cnt;
+		sum.dir[WRITE].size_total += r->dir[WRITE].size_total;
+	}
 
-	return sysfs_emit(page, "%lld %lld %lld %lldn %u\n",
-			  (s64)atomic64_read(&r->dir[READ].cnt),
-			  (s64)atomic64_read(&r->dir[READ].size_total),
-			  (s64)atomic64_read(&r->dir[WRITE].cnt),
-			  (s64)atomic64_read(&r->dir[WRITE].size_total), 0);
+	return sysfs_emit(page, "%llu %llu %llu %llu\n",
+			  sum.dir[READ].cnt, sum.dir[READ].size_total,
+			  sum.dir[WRITE].cnt, sum.dir[WRITE].size_total);
 }
diff --git a/drivers/infiniband/ulp/rtrs/rtrs-srv-sysfs.c b/drivers/infiniband/ulp/rtrs/rtrs-srv-sysfs.c
index b94ae12c2795..2a3c9ac64a42 100644
--- a/drivers/infiniband/ulp/rtrs/rtrs-srv-sysfs.c
+++ b/drivers/infiniband/ulp/rtrs/rtrs-srv-sysfs.c
@@ -220,6 +220,8 @@ static void rtrs_srv_path_stats_release(struct kobject *kobj)
 
 	stats = container_of(kobj, struct rtrs_srv_stats, kobj_stats);
 
+	free_percpu(stats->rdma_stats);
+
 	kfree(stats);
 }
 
diff --git a/drivers/infiniband/ulp/rtrs/rtrs-srv.c b/drivers/infiniband/ulp/rtrs/rtrs-srv.c
index 24024bce2566..34c03bde5064 100644
--- a/drivers/infiniband/ulp/rtrs/rtrs-srv.c
+++ b/drivers/infiniband/ulp/rtrs/rtrs-srv.c
@@ -11,7 +11,6 @@
 #define pr_fmt(fmt) KBUILD_MODNAME " L" __stringify(__LINE__) ": " fmt
 
 #include <linux/module.h>
-#include <linux/mempool.h>
 
 #include "rtrs-srv.h"
 #include "rtrs-log.h"
@@ -26,11 +25,7 @@ MODULE_LICENSE("GPL");
 #define DEFAULT_SESS_QUEUE_DEPTH 512
 #define MAX_HDR_SIZE PAGE_SIZE
 
-/* We guarantee to serve 10 paths at least */
-#define CHUNK_POOL_SZ 10
-
 static struct rtrs_rdma_dev_pd dev_pd;
-static mempool_t *chunk_pool;
 struct class *rtrs_dev_class;
 static struct rtrs_srv_ib_ctx ib_ctx;
 
@@ -1358,7 +1353,7 @@ static void free_srv(struct rtrs_srv_sess *srv)
 
 	WARN_ON(refcount_read(&srv->refcount));
 	for (i = 0; i < srv->queue_depth; i++)
-		mempool_free(srv->chunks[i], chunk_pool);
+		__free_pages(srv->chunks[i], get_order(max_chunk_size));
 	kfree(srv->chunks);
 	mutex_destroy(&srv->paths_mutex);
 	mutex_destroy(&srv->paths_ev_mutex);
@@ -1411,7 +1406,8 @@ static struct rtrs_srv_sess *get_or_create_srv(struct rtrs_srv_ctx *ctx,
 		goto err_free_srv;
 
 	for (i = 0; i < srv->queue_depth; i++) {
-		srv->chunks[i] = mempool_alloc(chunk_pool, GFP_KERNEL);
+		srv->chunks[i] = alloc_pages(GFP_KERNEL,
+					     get_order(max_chunk_size));
 		if (!srv->chunks[i])
 			goto err_free_chunks;
 	}
@@ -1424,7 +1420,7 @@ static struct rtrs_srv_sess *get_or_create_srv(struct rtrs_srv_ctx *ctx,
 
 err_free_chunks:
 	while (i--)
-		mempool_free(srv->chunks[i], chunk_pool);
+		__free_pages(srv->chunks[i], get_order(max_chunk_size));
 	kfree(srv->chunks);
 
 err_free_srv:
@@ -1513,6 +1509,7 @@ static void free_path(struct rtrs_srv_path *srv_path)
 		kobject_del(&srv_path->kobj);
 		kobject_put(&srv_path->kobj);
 	} else {
+		free_percpu(srv_path->stats->rdma_stats);
 		kfree(srv_path->stats);
 		kfree(srv_path);
 	}
@@ -1755,13 +1752,17 @@ static struct rtrs_srv_path *__alloc_path(struct rtrs_srv_sess *srv,
 	if (!srv_path->stats)
 		goto err_free_sess;
 
+	srv_path->stats->rdma_stats = alloc_percpu(struct rtrs_srv_stats_rdma_stats);
+	if (!srv_path->stats->rdma_stats)
+		goto err_free_stats;
+
 	srv_path->stats->srv_path = srv_path;
 
 	srv_path->dma_addr = kcalloc(srv->queue_depth,
 				     sizeof(*srv_path->dma_addr),
 				     GFP_KERNEL);
 	if (!srv_path->dma_addr)
-		goto err_free_stats;
+		goto err_free_percpu;
 
 	srv_path->s.con = kcalloc(con_num, sizeof(*srv_path->s.con),
 				  GFP_KERNEL);
@@ -1813,6 +1814,8 @@ err_free_con:
 	kfree(srv_path->s.con);
 err_free_dma_addr:
 	kfree(srv_path->dma_addr);
+err_free_percpu:
+	free_percpu(srv_path->stats->rdma_stats);
 err_free_stats:
 	kfree(srv_path->stats);
 err_free_sess:
@@ -2266,14 +2269,10 @@ static int __init rtrs_server_init(void)
 		       err);
 		return err;
 	}
-	chunk_pool = mempool_create_page_pool(sess_queue_depth * CHUNK_POOL_SZ,
-					      get_order(max_chunk_size));
-	if (!chunk_pool)
-		return -ENOMEM;
 	rtrs_dev_class = class_create(THIS_MODULE, "rtrs-server");
 	if (IS_ERR(rtrs_dev_class)) {
 		err = PTR_ERR(rtrs_dev_class);
-		goto out_chunk_pool;
+		goto out_err;
 	}
 	rtrs_wq = alloc_workqueue("rtrs_server_wq", 0, 0);
 	if (!rtrs_wq) {
@@ -2285,9 +2284,7 @@ static int __init rtrs_server_init(void)
 
 out_dev_class:
 	class_destroy(rtrs_dev_class);
-out_chunk_pool:
-	mempool_destroy(chunk_pool);
-
+out_err:
 	return err;
 }
 
@@ -2295,7 +2292,6 @@ static void __exit rtrs_server_exit(void)
 {
 	destroy_workqueue(rtrs_wq);
 	class_destroy(rtrs_dev_class);
-	mempool_destroy(chunk_pool);
 	rtrs_rdma_dev_pd_deinit(&dev_pd);
 }
 
diff --git a/drivers/infiniband/ulp/rtrs/rtrs-srv.h b/drivers/infiniband/ulp/rtrs/rtrs-srv.h
index 6292e87f6afd..186a63c217df 100644
--- a/drivers/infiniband/ulp/rtrs/rtrs-srv.h
+++ b/drivers/infiniband/ulp/rtrs/rtrs-srv.h
@@ -12,6 +12,7 @@
 
 #include <linux/device.h>
 #include <linux/refcount.h>
+#include <linux/percpu.h>
 #include "rtrs-pri.h"
 
 /*
@@ -29,15 +30,15 @@ enum rtrs_srv_state {
  */
 struct rtrs_srv_stats_rdma_stats {
 	struct {
-		atomic64_t	cnt;
-		atomic64_t	size_total;
+		u64 cnt;
+		u64 size_total;
 	} dir[2];
 };
 
 struct rtrs_srv_stats {
-	struct kobject				kobj_stats;
-	struct rtrs_srv_stats_rdma_stats	rdma_stats;
-	struct rtrs_srv_path			*srv_path;
+	struct kobject					kobj_stats;
+	struct rtrs_srv_stats_rdma_stats __percpu	*rdma_stats;
+	struct rtrs_srv_path				*srv_path;
 };
 
 struct rtrs_srv_con {
@@ -130,8 +131,8 @@ void close_path(struct rtrs_srv_path *srv_path);
 static inline void rtrs_srv_update_rdma_stats(struct rtrs_srv_stats *s,
 					      size_t size, int d)
 {
-	atomic64_inc(&s->rdma_stats.dir[d].cnt);
-	atomic64_add(size, &s->rdma_stats.dir[d].size_total);
+	this_cpu_inc(s->rdma_stats->dir[d].cnt);
+	this_cpu_add(s->rdma_stats->dir[d].size_total, size);
 }
 
 /* functions which are implemented in rtrs-srv-stats.c */
diff --git a/drivers/infiniband/ulp/srpt/ib_srpt.c b/drivers/infiniband/ulp/srpt/ib_srpt.c
index f86ee1c4b970..21cbe30d526f 100644
--- a/drivers/infiniband/ulp/srpt/ib_srpt.c
+++ b/drivers/infiniband/ulp/srpt/ib_srpt.c
@@ -565,12 +565,9 @@ static int srpt_refresh_port(struct srpt_port *sport)
 	if (ret)
 		return ret;
 
-	sport->port_guid_id.wwn.priv = sport;
-	srpt_format_guid(sport->port_guid_id.name,
-			 sizeof(sport->port_guid_id.name),
+	srpt_format_guid(sport->guid_name, ARRAY_SIZE(sport->guid_name),
 			 &sport->gid.global.interface_id);
-	sport->port_gid_id.wwn.priv = sport;
-	snprintf(sport->port_gid_id.name, sizeof(sport->port_gid_id.name),
+	snprintf(sport->gid_name, ARRAY_SIZE(sport->gid_name),
 		 "0x%016llx%016llx",
 		 be64_to_cpu(sport->gid.global.subnet_prefix),
 		 be64_to_cpu(sport->gid.global.interface_id));
@@ -2221,13 +2218,13 @@ static int srpt_cm_req_recv(struct srpt_device *const sdev,
 	ch->zw_cqe.done = srpt_zerolength_write_done;
 	INIT_WORK(&ch->release_work, srpt_release_channel_work);
 	ch->sport = sport;
-	if (ib_cm_id) {
-		ch->ib_cm.cm_id = ib_cm_id;
-		ib_cm_id->context = ch;
-	} else {
+	if (rdma_cm_id) {
 		ch->using_rdma_cm = true;
 		ch->rdma_cm.cm_id = rdma_cm_id;
 		rdma_cm_id->context = ch;
+	} else {
+		ch->ib_cm.cm_id = ib_cm_id;
+		ib_cm_id->context = ch;
 	}
 	/*
 	 * ch->rq_size should be at least as large as the initiator queue
@@ -2314,31 +2311,35 @@ static int srpt_cm_req_recv(struct srpt_device *const sdev,
 	tag_num = ch->rq_size;
 	tag_size = 1; /* ib_srpt does not use se_sess->sess_cmd_map */
 
-	mutex_lock(&sport->port_guid_id.mutex);
-	list_for_each_entry(stpg, &sport->port_guid_id.tpg_list, entry) {
-		if (!IS_ERR_OR_NULL(ch->sess))
-			break;
-		ch->sess = target_setup_session(&stpg->tpg, tag_num,
+	if (sport->guid_id) {
+		mutex_lock(&sport->guid_id->mutex);
+		list_for_each_entry(stpg, &sport->guid_id->tpg_list, entry) {
+			if (!IS_ERR_OR_NULL(ch->sess))
+				break;
+			ch->sess = target_setup_session(&stpg->tpg, tag_num,
 						tag_size, TARGET_PROT_NORMAL,
 						ch->sess_name, ch, NULL);
+		}
+		mutex_unlock(&sport->guid_id->mutex);
 	}
-	mutex_unlock(&sport->port_guid_id.mutex);
 
-	mutex_lock(&sport->port_gid_id.mutex);
-	list_for_each_entry(stpg, &sport->port_gid_id.tpg_list, entry) {
-		if (!IS_ERR_OR_NULL(ch->sess))
-			break;
-		ch->sess = target_setup_session(&stpg->tpg, tag_num,
+	if (sport->gid_id) {
+		mutex_lock(&sport->gid_id->mutex);
+		list_for_each_entry(stpg, &sport->gid_id->tpg_list, entry) {
+			if (!IS_ERR_OR_NULL(ch->sess))
+				break;
+			ch->sess = target_setup_session(&stpg->tpg, tag_num,
 					tag_size, TARGET_PROT_NORMAL, i_port_id,
 					ch, NULL);
-		if (!IS_ERR_OR_NULL(ch->sess))
-			break;
-		/* Retry without leading "0x" */
-		ch->sess = target_setup_session(&stpg->tpg, tag_num,
+			if (!IS_ERR_OR_NULL(ch->sess))
+				break;
+			/* Retry without leading "0x" */
+			ch->sess = target_setup_session(&stpg->tpg, tag_num,
 						tag_size, TARGET_PROT_NORMAL,
 						i_port_id + 2, ch, NULL);
+		}
+		mutex_unlock(&sport->gid_id->mutex);
 	}
-	mutex_unlock(&sport->port_gid_id.mutex);
 
 	if (IS_ERR_OR_NULL(ch->sess)) {
 		WARN_ON_ONCE(ch->sess == NULL);
@@ -2983,7 +2984,12 @@ static int srpt_release_sport(struct srpt_port *sport)
 	return 0;
 }
 
-static struct se_wwn *__srpt_lookup_wwn(const char *name)
+struct port_and_port_id {
+	struct srpt_port *sport;
+	struct srpt_port_id **port_id;
+};
+
+static struct port_and_port_id __srpt_lookup_port(const char *name)
 {
 	struct ib_device *dev;
 	struct srpt_device *sdev;
@@ -2998,25 +3004,38 @@ static struct se_wwn *__srpt_lookup_wwn(const char *name)
 		for (i = 0; i < dev->phys_port_cnt; i++) {
 			sport = &sdev->port[i];
 
-			if (strcmp(sport->port_guid_id.name, name) == 0)
-				return &sport->port_guid_id.wwn;
-			if (strcmp(sport->port_gid_id.name, name) == 0)
-				return &sport->port_gid_id.wwn;
+			if (strcmp(sport->guid_name, name) == 0) {
+				kref_get(&sdev->refcnt);
+				return (struct port_and_port_id){
+					sport, &sport->guid_id};
+			}
+			if (strcmp(sport->gid_name, name) == 0) {
+				kref_get(&sdev->refcnt);
+				return (struct port_and_port_id){
+					sport, &sport->gid_id};
+			}
 		}
 	}
 
-	return NULL;
+	return (struct port_and_port_id){};
 }
 
-static struct se_wwn *srpt_lookup_wwn(const char *name)
+/**
+ * srpt_lookup_port() - Look up an RDMA port by name
+ * @name: ASCII port name
+ *
+ * Increments the RDMA port reference count if an RDMA port pointer is returned.
+ * The caller must drop that reference count by calling srpt_port_put_ref().
+ */
+static struct port_and_port_id srpt_lookup_port(const char *name)
 {
-	struct se_wwn *wwn;
+	struct port_and_port_id papi;
 
 	spin_lock(&srpt_dev_lock);
-	wwn = __srpt_lookup_wwn(name);
+	papi = __srpt_lookup_port(name);
 	spin_unlock(&srpt_dev_lock);
 
-	return wwn;
+	return papi;
 }
 
 static void srpt_free_srq(struct srpt_device *sdev)
@@ -3101,6 +3120,18 @@ static int srpt_use_srq(struct srpt_device *sdev, bool use_srq)
 	return ret;
 }
 
+static void srpt_free_sdev(struct kref *refcnt)
+{
+	struct srpt_device *sdev = container_of(refcnt, typeof(*sdev), refcnt);
+
+	kfree(sdev);
+}
+
+static void srpt_sdev_put(struct srpt_device *sdev)
+{
+	kref_put(&sdev->refcnt, srpt_free_sdev);
+}
+
 /**
  * srpt_add_one - InfiniBand device addition callback function
  * @device: Describes a HCA.
@@ -3119,6 +3150,7 @@ static int srpt_add_one(struct ib_device *device)
 	if (!sdev)
 		return -ENOMEM;
 
+	kref_init(&sdev->refcnt);
 	sdev->device = device;
 	mutex_init(&sdev->sdev_mutex);
 
@@ -3182,10 +3214,6 @@ static int srpt_add_one(struct ib_device *device)
 		sport->port_attrib.srp_sq_size = DEF_SRPT_SQ_SIZE;
 		sport->port_attrib.use_srq = false;
 		INIT_WORK(&sport->work, srpt_refresh_port_work);
-		mutex_init(&sport->port_guid_id.mutex);
-		INIT_LIST_HEAD(&sport->port_guid_id.tpg_list);
-		mutex_init(&sport->port_gid_id.mutex);
-		INIT_LIST_HEAD(&sport->port_gid_id.tpg_list);
 
 		ret = srpt_refresh_port(sport);
 		if (ret) {
@@ -3214,7 +3242,7 @@ err_ring:
 	srpt_free_srq(sdev);
 	ib_dealloc_pd(sdev->pd);
 free_dev:
-	kfree(sdev);
+	srpt_sdev_put(sdev);
 	pr_info("%s(%s) failed.\n", __func__, dev_name(&device->dev));
 	return ret;
 }
@@ -3258,7 +3286,7 @@ static void srpt_remove_one(struct ib_device *device, void *client_data)
 
 	ib_dealloc_pd(sdev->pd);
 
-	kfree(sdev);
+	srpt_sdev_put(sdev);
 }
 
 static struct ib_client srpt_client = {
@@ -3286,10 +3314,10 @@ static struct srpt_port_id *srpt_wwn_to_sport_id(struct se_wwn *wwn)
 {
 	struct srpt_port *sport = wwn->priv;
 
-	if (wwn == &sport->port_guid_id.wwn)
-		return &sport->port_guid_id;
-	if (wwn == &sport->port_gid_id.wwn)
-		return &sport->port_gid_id;
+	if (sport->guid_id && &sport->guid_id->wwn == wwn)
+		return sport->guid_id;
+	if (sport->gid_id && &sport->gid_id->wwn == wwn)
+		return sport->gid_id;
 	WARN_ON_ONCE(true);
 	return NULL;
 }
@@ -3774,7 +3802,31 @@ static struct se_wwn *srpt_make_tport(struct target_fabric_configfs *tf,
 				      struct config_group *group,
 				      const char *name)
 {
-	return srpt_lookup_wwn(name) ? : ERR_PTR(-EINVAL);
+	struct port_and_port_id papi = srpt_lookup_port(name);
+	struct srpt_port *sport = papi.sport;
+	struct srpt_port_id *port_id;
+
+	if (!papi.port_id)
+		return ERR_PTR(-EINVAL);
+	if (*papi.port_id) {
+		/* Attempt to create a directory that already exists. */
+		WARN_ON_ONCE(true);
+		return &(*papi.port_id)->wwn;
+	}
+	port_id = kzalloc(sizeof(*port_id), GFP_KERNEL);
+	if (!port_id) {
+		srpt_sdev_put(sport->sdev);
+		return ERR_PTR(-ENOMEM);
+	}
+	mutex_init(&port_id->mutex);
+	INIT_LIST_HEAD(&port_id->tpg_list);
+	port_id->wwn.priv = sport;
+	memcpy(port_id->name, port_id == sport->guid_id ? sport->guid_name :
+	       sport->gid_name, ARRAY_SIZE(port_id->name));
+
+	*papi.port_id = port_id;
+
+	return &port_id->wwn;
 }
 
 /**
@@ -3783,6 +3835,18 @@ static struct se_wwn *srpt_make_tport(struct target_fabric_configfs *tf,
  */
 static void srpt_drop_tport(struct se_wwn *wwn)
 {
+	struct srpt_port_id *port_id = container_of(wwn, typeof(*port_id), wwn);
+	struct srpt_port *sport = wwn->priv;
+
+	if (sport->guid_id == port_id)
+		sport->guid_id = NULL;
+	else if (sport->gid_id == port_id)
+		sport->gid_id = NULL;
+	else
+		WARN_ON_ONCE(true);
+
+	srpt_sdev_put(sport->sdev);
+	kfree(port_id);
 }
 
 static ssize_t srpt_wwn_version_show(struct config_item *item, char *buf)
diff --git a/drivers/infiniband/ulp/srpt/ib_srpt.h b/drivers/infiniband/ulp/srpt/ib_srpt.h
index 76e66f630c17..4c46b301eea1 100644
--- a/drivers/infiniband/ulp/srpt/ib_srpt.h
+++ b/drivers/infiniband/ulp/srpt/ib_srpt.h
@@ -376,7 +376,7 @@ struct srpt_tpg {
 };
 
 /**
- * struct srpt_port_id - information about an RDMA port name
+ * struct srpt_port_id - LIO RDMA port information
  * @mutex:	Protects @tpg_list changes.
  * @tpg_list:	TPGs associated with the RDMA port name.
  * @wwn:	WWN associated with the RDMA port name.
@@ -393,7 +393,7 @@ struct srpt_port_id {
 };
 
 /**
- * struct srpt_port - information associated by SRPT with a single IB port
+ * struct srpt_port - SRPT RDMA port information
  * @sdev:      backpointer to the HCA information.
  * @mad_agent: per-port management datagram processing information.
  * @enabled:   Whether or not this target port is enabled.
@@ -402,8 +402,10 @@ struct srpt_port_id {
  * @lid:       cached value of the port's lid.
  * @gid:       cached value of the port's gid.
  * @work:      work structure for refreshing the aforementioned cached values.
- * @port_guid_id: target port GUID
- * @port_gid_id: target port GID
+ * @guid_name: port name in GUID format.
+ * @guid_id:   LIO target port information for the port name in GUID format.
+ * @gid_name:  port name in GID format.
+ * @gid_id:    LIO target port information for the port name in GID format.
  * @port_attrib:   Port attributes that can be accessed through configfs.
  * @refcount:	   Number of objects associated with this port.
  * @freed_channels: Completion that will be signaled once @refcount becomes 0.
@@ -419,8 +421,10 @@ struct srpt_port {
 	u32			lid;
 	union ib_gid		gid;
 	struct work_struct	work;
-	struct srpt_port_id	port_guid_id;
-	struct srpt_port_id	port_gid_id;
+	char			guid_name[64];
+	struct srpt_port_id	*guid_id;
+	char			gid_name[64];
+	struct srpt_port_id	*gid_id;
 	struct srpt_port_attrib port_attrib;
 	atomic_t		refcount;
 	struct completion	*freed_channels;
@@ -430,6 +434,7 @@ struct srpt_port {
 
 /**
  * struct srpt_device - information associated by SRPT with a single HCA
+ * @refcnt:	   Reference count for this device.
  * @device:        Backpointer to the struct ib_device managed by the IB core.
  * @pd:            IB protection domain.
  * @lkey:          L_Key (local key) with write access to all local memory.
@@ -445,6 +450,7 @@ struct srpt_port {
  * @port:          Information about the ports owned by this HCA.
  */
 struct srpt_device {
+	struct kref		refcnt;
 	struct ib_device	*device;
 	struct ib_pd		*pd;
 	u32			lkey;