From d3c04a3a6870a4b96ee213d99b4f4ca723e64025 Mon Sep 17 00:00:00 2001
From: Vijay Immanuel
Date: Thu, 5 Jul 2018 18:43:47 -0700
Subject: IB/rxe: vary the source udp port for receive scaling

Select the source udp port number for a QP based on the
source QPN. This provides a better spread of traffic
across NIC RX queues for RC/UC QPs.

Signed-off-by: Vijay Immanuel <vijayi@attalasystems.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
---
 drivers/infiniband/sw/rxe/rxe_net.c   |  8 ++++----
 drivers/infiniband/sw/rxe/rxe_qp.c    | 10 ++++++++++
 drivers/infiniband/sw/rxe/rxe_verbs.h |  1 +
 3 files changed, 15 insertions(+), 4 deletions(-)

(limited to 'drivers/infiniband/sw')

diff --git a/drivers/infiniband/sw/rxe/rxe_net.c b/drivers/infiniband/sw/rxe/rxe_net.c
index 8094cbaa54a9..4ddb6b4c6a46 100644
--- a/drivers/infiniband/sw/rxe/rxe_net.c
+++ b/drivers/infiniband/sw/rxe/rxe_net.c
@@ -396,8 +396,8 @@ static int prepare4(struct rxe_dev *rxe, struct rxe_pkt_info *pkt,
 	if (!memcmp(saddr, daddr, sizeof(*daddr)))
 		pkt->mask |= RXE_LOOPBACK_MASK;
 
-	prepare_udp_hdr(skb, htons(RXE_ROCE_V2_SPORT),
-			htons(ROCE_V2_UDP_DPORT));
+	prepare_udp_hdr(skb, cpu_to_be16(qp->src_port),
+			cpu_to_be16(ROCE_V2_UDP_DPORT));
 
 	prepare_ipv4_hdr(dst, skb, saddr->s_addr, daddr->s_addr, IPPROTO_UDP,
 			 av->grh.traffic_class, av->grh.hop_limit, df, xnet);
@@ -423,8 +423,8 @@ static int prepare6(struct rxe_dev *rxe, struct rxe_pkt_info *pkt,
 	if (!memcmp(saddr, daddr, sizeof(*daddr)))
 		pkt->mask |= RXE_LOOPBACK_MASK;
 
-	prepare_udp_hdr(skb, htons(RXE_ROCE_V2_SPORT),
-			htons(ROCE_V2_UDP_DPORT));
+	prepare_udp_hdr(skb, cpu_to_be16(qp->src_port),
+			cpu_to_be16(ROCE_V2_UDP_DPORT));
 
 	prepare_ipv6_hdr(dst, skb, saddr, daddr, IPPROTO_UDP,
 			 av->grh.traffic_class,
diff --git a/drivers/infiniband/sw/rxe/rxe_qp.c b/drivers/infiniband/sw/rxe/rxe_qp.c
index c58452daffc7..6ff88c8250f6 100644
--- a/drivers/infiniband/sw/rxe/rxe_qp.c
+++ b/drivers/infiniband/sw/rxe/rxe_qp.c
@@ -227,6 +227,16 @@ static int rxe_qp_init_req(struct rxe_dev *rxe, struct rxe_qp *qp,
 		return err;
 	qp->sk->sk->sk_user_data = qp;
 
+	/* pick a source UDP port number for this QP based on
+	 * the source QPN. this spreads traffic for different QPs
+	 * across different NIC RX queues (while using a single
+	 * flow for a given QP to maintain packet order).
+	 * the port number must be in the Dynamic Ports range
+	 * (0xc000 - 0xffff).
+	 */
+	qp->src_port = RXE_ROCE_V2_SPORT +
+		(hash_32_generic(qp_num(qp), 14) & 0x3fff);
+
 	qp->sq.max_wr		= init->cap.max_send_wr;
 	qp->sq.max_sge		= init->cap.max_send_sge;
 	qp->sq.max_inline	= init->cap.max_inline_data;
diff --git a/drivers/infiniband/sw/rxe/rxe_verbs.h b/drivers/infiniband/sw/rxe/rxe_verbs.h
index af1470d29391..0d920087811a 100644
--- a/drivers/infiniband/sw/rxe/rxe_verbs.h
+++ b/drivers/infiniband/sw/rxe/rxe_verbs.h
@@ -248,6 +248,7 @@ struct rxe_qp {
 
 	struct socket		*sk;
 	u32			dst_cookie;
+	u16			src_port;
 
 	struct rxe_av		pri_av;
 	struct rxe_av		alt_av;
-- 
cgit v1.2.3-55-g7522


From 8b7b59d030cc0d34b83b0c556bb6365c9ae6ba77 Mon Sep 17 00:00:00 2001
From: Zhu Yanjun
Date: Sun, 19 Aug 2018 15:04:01 +0800
Subject: IB/rxe: remove redudant qpn check

In the commit 536ca245c512 ("IB/rxe: Drop QP0 silently"), if qpn is
zero, the function directly returns. So in the following function,
it is not necessary to check qpn. The qpn check in the function
check_keys is removed.

Fixes: 536ca245c512 ("IB/rxe: Drop QP0 silently")
CC: Srinivas Eeda <srinivas.eeda@oracle.com>
CC: Junxiao Bi <junxiao.bi@oracle.com>
Signed-off-by: Zhu Yanjun <yanjun.zhu@oracle.com>
Reviewed-by: Yuval Shaia <yuval.shaia@oracle.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
---
 drivers/infiniband/sw/rxe/rxe_recv.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

(limited to 'drivers/infiniband/sw')

diff --git a/drivers/infiniband/sw/rxe/rxe_recv.c b/drivers/infiniband/sw/rxe/rxe_recv.c
index d30dbac24583..5c29a1bb575a 100644
--- a/drivers/infiniband/sw/rxe/rxe_recv.c
+++ b/drivers/infiniband/sw/rxe/rxe_recv.c
@@ -122,7 +122,7 @@ static int check_keys(struct rxe_dev *rxe, struct rxe_pkt_info *pkt,
 			set_bad_pkey_cntr(port);
 			goto err1;
 		}
-	} else if (qpn != 0) {
+	} else {
 		if (unlikely(!pkey_match(pkey,
 					 port->pkey_tbl[qp->attr.pkey_index]
 					))) {
@@ -134,7 +134,7 @@ static int check_keys(struct rxe_dev *rxe, struct rxe_pkt_info *pkt,
 	}
 
 	if ((qp_type(qp) == IB_QPT_UD || qp_type(qp) == IB_QPT_GSI) &&
-	    qpn != 0 && pkt->mask) {
+	    pkt->mask) {
 		u32 qkey = (qpn == 1) ? GSI_QKEY : qp->attr.qkey;
 
 		if (unlikely(deth_qkey(pkt) != qkey)) {
-- 
cgit v1.2.3-55-g7522


From 3db2bceb29fd9a2f31d5df573f337dbba577f8e9 Mon Sep 17 00:00:00 2001
From: Parav Pandit
Date: Tue, 28 Aug 2018 13:51:37 +0300
Subject: IB/rxe: Simplify rxe_find_route() to avoid GID query for netdev

rxe_prepare() is called on an skb which has ndev already initialized by
rxe_init_packet().
Therefore avoid querying the GID attribute again and use the available
netdevice from the skb->dev.

Signed-off-by: Parav Pandit <parav@mellanox.com>
Reviewed-by: Daniel Jurgens <danielj@mellanox.com>
Reviewed-by: Yuval Shaia <yuval.shaia@oracle.com>
Tested-by: Yuval Shaia <yuval.shaia@oracle.com>
Signed-off-by: Zhu Yanjun <yanjun.zhu@oracle.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
---
 drivers/infiniband/sw/rxe/rxe_loc.h  |  3 +--
 drivers/infiniband/sw/rxe/rxe_net.c  | 30 ++++++++++--------------------
 drivers/infiniband/sw/rxe/rxe_req.c  |  2 +-
 drivers/infiniband/sw/rxe/rxe_resp.c |  2 +-
 4 files changed, 13 insertions(+), 24 deletions(-)

(limited to 'drivers/infiniband/sw')

diff --git a/drivers/infiniband/sw/rxe/rxe_loc.h b/drivers/infiniband/sw/rxe/rxe_loc.h
index 87d14f7ef21b..8e305422adbb 100644
--- a/drivers/infiniband/sw/rxe/rxe_loc.h
+++ b/drivers/infiniband/sw/rxe/rxe_loc.h
@@ -144,8 +144,7 @@ void rxe_loopback(struct sk_buff *skb);
 int rxe_send(struct rxe_pkt_info *pkt, struct sk_buff *skb);
 struct sk_buff *rxe_init_packet(struct rxe_dev *rxe, struct rxe_av *av,
 				int paylen, struct rxe_pkt_info *pkt);
-int rxe_prepare(struct rxe_dev *rxe, struct rxe_pkt_info *pkt,
-		struct sk_buff *skb, u32 *crc);
+int rxe_prepare(struct rxe_pkt_info *pkt, struct sk_buff *skb, u32 *crc);
 enum rdma_link_layer rxe_link_layer(struct rxe_dev *rxe, unsigned int port_num);
 const char *rxe_parent_name(struct rxe_dev *rxe, unsigned int port_num);
 struct device *rxe_dma_device(struct rxe_dev *rxe);
diff --git a/drivers/infiniband/sw/rxe/rxe_net.c b/drivers/infiniband/sw/rxe/rxe_net.c
index 4ddb6b4c6a46..ed823cccca48 100644
--- a/drivers/infiniband/sw/rxe/rxe_net.c
+++ b/drivers/infiniband/sw/rxe/rxe_net.c
@@ -182,19 +182,11 @@ static struct dst_entry *rxe_find_route6(struct net_device *ndev,
 
 #endif
 
-static struct dst_entry *rxe_find_route(struct rxe_dev *rxe,
+static struct dst_entry *rxe_find_route(struct net_device *ndev,
 					struct rxe_qp *qp,
 					struct rxe_av *av)
 {
-	const struct ib_gid_attr *attr;
 	struct dst_entry *dst = NULL;
-	struct net_device *ndev;
-
-	attr = rdma_get_gid_attr(&rxe->ib_dev, qp->attr.port_num,
-				 av->grh.sgid_index);
-	if (IS_ERR(attr))
-		return NULL;
-	ndev = attr->ndev;
 
 	if (qp_type(qp) == IB_QPT_RC)
 		dst = sk_dst_get(qp->sk->sk);
@@ -229,7 +221,6 @@ static struct dst_entry *rxe_find_route(struct rxe_dev *rxe,
 			sk_dst_set(qp->sk->sk, dst);
 		}
 	}
-	rdma_put_gid_attr(attr);
 	return dst;
 }
 
@@ -377,8 +368,8 @@ static void prepare_ipv6_hdr(struct dst_entry *dst, struct sk_buff *skb,
 	ip6h->payload_len = htons(skb->len - sizeof(*ip6h));
 }
 
-static int prepare4(struct rxe_dev *rxe, struct rxe_pkt_info *pkt,
-		    struct sk_buff *skb, struct rxe_av *av)
+static int prepare4(struct rxe_pkt_info *pkt, struct sk_buff *skb,
+		    struct rxe_av *av)
 {
 	struct rxe_qp *qp = pkt->qp;
 	struct dst_entry *dst;
@@ -387,7 +378,7 @@ static int prepare4(struct rxe_dev *rxe, struct rxe_pkt_info *pkt,
 	struct in_addr *saddr = &av->sgid_addr._sockaddr_in.sin_addr;
 	struct in_addr *daddr = &av->dgid_addr._sockaddr_in.sin_addr;
 
-	dst = rxe_find_route(rxe, qp, av);
+	dst = rxe_find_route(skb->dev, qp, av);
 	if (!dst) {
 		pr_err("Host not reachable\n");
 		return -EHOSTUNREACH;
@@ -406,15 +397,15 @@ static int prepare4(struct rxe_dev *rxe, struct rxe_pkt_info *pkt,
 	return 0;
 }
 
-static int prepare6(struct rxe_dev *rxe, struct rxe_pkt_info *pkt,
-		    struct sk_buff *skb, struct rxe_av *av)
+static int prepare6(struct rxe_pkt_info *pkt, struct sk_buff *skb,
+		    struct rxe_av *av)
 {
 	struct rxe_qp *qp = pkt->qp;
 	struct dst_entry *dst;
 	struct in6_addr *saddr = &av->sgid_addr._sockaddr_in6.sin6_addr;
 	struct in6_addr *daddr = &av->dgid_addr._sockaddr_in6.sin6_addr;
 
-	dst = rxe_find_route(rxe, qp, av);
+	dst = rxe_find_route(skb->dev, qp, av);
 	if (!dst) {
 		pr_err("Host not reachable\n");
 		return -EHOSTUNREACH;
@@ -434,16 +425,15 @@ static int prepare6(struct rxe_dev *rxe, struct rxe_pkt_info *pkt,
 	return 0;
 }
 
-int rxe_prepare(struct rxe_dev *rxe, struct rxe_pkt_info *pkt,
-		struct sk_buff *skb, u32 *crc)
+int rxe_prepare(struct rxe_pkt_info *pkt, struct sk_buff *skb, u32 *crc)
 {
 	int err = 0;
 	struct rxe_av *av = rxe_get_av(pkt);
 
 	if (av->network_type == RDMA_NETWORK_IPV4)
-		err = prepare4(rxe, pkt, skb, av);
+		err = prepare4(pkt, skb, av);
 	else if (av->network_type == RDMA_NETWORK_IPV6)
-		err = prepare6(rxe, pkt, skb, av);
+		err = prepare6(pkt, skb, av);
 
 	*crc = rxe_icrc_hdr(pkt, skb);
 
diff --git a/drivers/infiniband/sw/rxe/rxe_req.c b/drivers/infiniband/sw/rxe/rxe_req.c
index 8be27238a86e..84d35520781c 100644
--- a/drivers/infiniband/sw/rxe/rxe_req.c
+++ b/drivers/infiniband/sw/rxe/rxe_req.c
@@ -476,7 +476,7 @@ static int fill_packet(struct rxe_qp *qp, struct rxe_send_wqe *wqe,
 	u32 *p;
 	int err;
 
-	err = rxe_prepare(rxe, pkt, skb, &crc);
+	err = rxe_prepare(pkt, skb, &crc);
 	if (err)
 		return err;
 
diff --git a/drivers/infiniband/sw/rxe/rxe_resp.c b/drivers/infiniband/sw/rxe/rxe_resp.c
index aa5833318372..c1948d406528 100644
--- a/drivers/infiniband/sw/rxe/rxe_resp.c
+++ b/drivers/infiniband/sw/rxe/rxe_resp.c
@@ -637,7 +637,7 @@ static struct sk_buff *prepare_ack_packet(struct rxe_qp *qp,
 	if (ack->mask & RXE_ATMACK_MASK)
 		atmack_set_orig(ack, qp->resp.atomic_orig);
 
-	err = rxe_prepare(rxe, ack, skb, &crc);
+	err = rxe_prepare(ack, skb, &crc);
 	if (err) {
 		kfree_skb(skb);
 		return NULL;
-- 
cgit v1.2.3-55-g7522


From 66d0f207dbf0e983c8006eb2fc5060163646210f Mon Sep 17 00:00:00 2001
From: Parav Pandit
Date: Mon, 27 Aug 2018 08:44:14 +0300
Subject: IB/rxe: Replace spinlock with rwlock

Concurrent readers which read rb tree are protected using read lock.
Concurrent writers which add element to pool are protected
using write lock.

Signed-off-by: Parav Pandit <parav@mellanox.com>
Reviewed-by: Daniel Jurgens <danielj@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Reviewed-by: Yuval Shaia <yuval.shaia@oracle.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
---
 drivers/infiniband/sw/rxe/rxe_pool.c | 36 ++++++++++++++++++------------------
 drivers/infiniband/sw/rxe/rxe_pool.h |  2 +-
 2 files changed, 19 insertions(+), 19 deletions(-)

(limited to 'drivers/infiniband/sw')

diff --git a/drivers/infiniband/sw/rxe/rxe_pool.c b/drivers/infiniband/sw/rxe/rxe_pool.c
index b4a8acc7bb7d..be076d818891 100644
--- a/drivers/infiniband/sw/rxe/rxe_pool.c
+++ b/drivers/infiniband/sw/rxe/rxe_pool.c
@@ -207,7 +207,7 @@ int rxe_pool_init(
 
 	kref_init(&pool->ref_cnt);
 
-	spin_lock_init(&pool->pool_lock);
+	rwlock_init(&pool->pool_lock);
 
 	if (rxe_type_info[type].flags & RXE_POOL_INDEX) {
 		err = rxe_pool_init_index(pool,
@@ -245,12 +245,12 @@ int rxe_pool_cleanup(struct rxe_pool *pool)
 {
 	unsigned long flags;
 
-	spin_lock_irqsave(&pool->pool_lock, flags);
+	write_lock_irqsave(&pool->pool_lock, flags);
 	pool->state = rxe_pool_invalid;
 	if (atomic_read(&pool->num_elem) > 0)
 		pr_warn("%s pool destroyed with unfree'd elem\n",
 			pool_name(pool));
-	spin_unlock_irqrestore(&pool->pool_lock, flags);
+	write_unlock_irqrestore(&pool->pool_lock, flags);
 
 	rxe_pool_put(pool);
 
@@ -336,10 +336,10 @@ void rxe_add_key(void *arg, void *key)
 	struct rxe_pool *pool = elem->pool;
 	unsigned long flags;
 
-	spin_lock_irqsave(&pool->pool_lock, flags);
+	write_lock_irqsave(&pool->pool_lock, flags);
 	memcpy((u8 *)elem + pool->key_offset, key, pool->key_size);
 	insert_key(pool, elem);
-	spin_unlock_irqrestore(&pool->pool_lock, flags);
+	write_unlock_irqrestore(&pool->pool_lock, flags);
 }
 
 void rxe_drop_key(void *arg)
@@ -348,9 +348,9 @@ void rxe_drop_key(void *arg)
 	struct rxe_pool *pool = elem->pool;
 	unsigned long flags;
 
-	spin_lock_irqsave(&pool->pool_lock, flags);
+	write_lock_irqsave(&pool->pool_lock, flags);
 	rb_erase(&elem->node, &pool->tree);
-	spin_unlock_irqrestore(&pool->pool_lock, flags);
+	write_unlock_irqrestore(&pool->pool_lock, flags);
 }
 
 void rxe_add_index(void *arg)
@@ -359,10 +359,10 @@ void rxe_add_index(void *arg)
 	struct rxe_pool *pool = elem->pool;
 	unsigned long flags;
 
-	spin_lock_irqsave(&pool->pool_lock, flags);
+	write_lock_irqsave(&pool->pool_lock, flags);
 	elem->index = alloc_index(pool);
 	insert_index(pool, elem);
-	spin_unlock_irqrestore(&pool->pool_lock, flags);
+	write_unlock_irqrestore(&pool->pool_lock, flags);
 }
 
 void rxe_drop_index(void *arg)
@@ -371,10 +371,10 @@ void rxe_drop_index(void *arg)
 	struct rxe_pool *pool = elem->pool;
 	unsigned long flags;
 
-	spin_lock_irqsave(&pool->pool_lock, flags);
+	write_lock_irqsave(&pool->pool_lock, flags);
 	clear_bit(elem->index - pool->min_index, pool->table);
 	rb_erase(&elem->node, &pool->tree);
-	spin_unlock_irqrestore(&pool->pool_lock, flags);
+	write_unlock_irqrestore(&pool->pool_lock, flags);
 }
 
 void *rxe_alloc(struct rxe_pool *pool)
@@ -384,13 +384,13 @@ void *rxe_alloc(struct rxe_pool *pool)
 
 	might_sleep_if(!(pool->flags & RXE_POOL_ATOMIC));
 
-	spin_lock_irqsave(&pool->pool_lock, flags);
+	read_lock_irqsave(&pool->pool_lock, flags);
 	if (pool->state != rxe_pool_valid) {
-		spin_unlock_irqrestore(&pool->pool_lock, flags);
+		read_unlock_irqrestore(&pool->pool_lock, flags);
 		return NULL;
 	}
 	kref_get(&pool->ref_cnt);
-	spin_unlock_irqrestore(&pool->pool_lock, flags);
+	read_unlock_irqrestore(&pool->pool_lock, flags);
 
 	kref_get(&pool->rxe->ref_cnt);
 
@@ -436,7 +436,7 @@ void *rxe_pool_get_index(struct rxe_pool *pool, u32 index)
 	struct rxe_pool_entry *elem = NULL;
 	unsigned long flags;
 
-	spin_lock_irqsave(&pool->pool_lock, flags);
+	read_lock_irqsave(&pool->pool_lock, flags);
 
 	if (pool->state != rxe_pool_valid)
 		goto out;
@@ -458,7 +458,7 @@ void *rxe_pool_get_index(struct rxe_pool *pool, u32 index)
 		kref_get(&elem->ref_cnt);
 
 out:
-	spin_unlock_irqrestore(&pool->pool_lock, flags);
+	read_unlock_irqrestore(&pool->pool_lock, flags);
 	return node ? elem : NULL;
 }
 
@@ -469,7 +469,7 @@ void *rxe_pool_get_key(struct rxe_pool *pool, void *key)
 	int cmp;
 	unsigned long flags;
 
-	spin_lock_irqsave(&pool->pool_lock, flags);
+	read_lock_irqsave(&pool->pool_lock, flags);
 
 	if (pool->state != rxe_pool_valid)
 		goto out;
@@ -494,6 +494,6 @@ void *rxe_pool_get_key(struct rxe_pool *pool, void *key)
 		kref_get(&elem->ref_cnt);
 
 out:
-	spin_unlock_irqrestore(&pool->pool_lock, flags);
+	read_unlock_irqrestore(&pool->pool_lock, flags);
 	return node ? elem : NULL;
 }
diff --git a/drivers/infiniband/sw/rxe/rxe_pool.h b/drivers/infiniband/sw/rxe/rxe_pool.h
index 47df28e43acf..6eb344236afc 100644
--- a/drivers/infiniband/sw/rxe/rxe_pool.h
+++ b/drivers/infiniband/sw/rxe/rxe_pool.h
@@ -90,7 +90,7 @@ struct rxe_pool_entry {
 
 struct rxe_pool {
 	struct rxe_dev		*rxe;
-	spinlock_t              pool_lock; /* pool spinlock */
+	rwlock_t		pool_lock; /* protects pool add/del/search */
 	size_t			elem_size;
 	struct kref		ref_cnt;
 	void			(*cleanup)(struct rxe_pool_entry *obj);
-- 
cgit v1.2.3-55-g7522


From 3ccf19e25adf2d3b9c5f1e65b950b6d223e1c4c0 Mon Sep 17 00:00:00 2001
From: Parav Pandit
Date: Mon, 27 Aug 2018 08:44:15 +0300
Subject: IB/rxe: Change pool state enums to capital letters

Normal practice is to have enum defines in capital letters.

Signed-off-by: Parav Pandit <parav@mellanox.com>
Reviewed-by: Daniel Jurgens <danielj@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Reviewed-by: Yuval Shaia <yuval.shaia@oracle.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
---
 drivers/infiniband/sw/rxe/rxe_pool.c | 12 ++++++------
 drivers/infiniband/sw/rxe/rxe_pool.h |  4 ++--
 2 files changed, 8 insertions(+), 8 deletions(-)

(limited to 'drivers/infiniband/sw')

diff --git a/drivers/infiniband/sw/rxe/rxe_pool.c b/drivers/infiniband/sw/rxe/rxe_pool.c
index be076d818891..c8598b91e454 100644
--- a/drivers/infiniband/sw/rxe/rxe_pool.c
+++ b/drivers/infiniband/sw/rxe/rxe_pool.c
@@ -222,7 +222,7 @@ int rxe_pool_init(
 		pool->key_size = rxe_type_info[type].key_size;
 	}
 
-	pool->state = rxe_pool_valid;
+	pool->state = RXE_POOL_STATE_VALID;
 
 out:
 	return err;
@@ -232,7 +232,7 @@ static void rxe_pool_release(struct kref *kref)
 {
 	struct rxe_pool *pool = container_of(kref, struct rxe_pool, ref_cnt);
 
-	pool->state = rxe_pool_invalid;
+	pool->state = RXE_POOL_STATE_INVALID;
 	kfree(pool->table);
 }
 
@@ -246,7 +246,7 @@ int rxe_pool_cleanup(struct rxe_pool *pool)
 	unsigned long flags;
 
 	write_lock_irqsave(&pool->pool_lock, flags);
-	pool->state = rxe_pool_invalid;
+	pool->state = RXE_POOL_STATE_INVALID;
 	if (atomic_read(&pool->num_elem) > 0)
 		pr_warn("%s pool destroyed with unfree'd elem\n",
 			pool_name(pool));
@@ -385,7 +385,7 @@ void *rxe_alloc(struct rxe_pool *pool)
 	might_sleep_if(!(pool->flags & RXE_POOL_ATOMIC));
 
 	read_lock_irqsave(&pool->pool_lock, flags);
-	if (pool->state != rxe_pool_valid) {
+	if (pool->state != RXE_POOL_STATE_VALID) {
 		read_unlock_irqrestore(&pool->pool_lock, flags);
 		return NULL;
 	}
@@ -438,7 +438,7 @@ void *rxe_pool_get_index(struct rxe_pool *pool, u32 index)
 
 	read_lock_irqsave(&pool->pool_lock, flags);
 
-	if (pool->state != rxe_pool_valid)
+	if (pool->state != RXE_POOL_STATE_VALID)
 		goto out;
 
 	node = pool->tree.rb_node;
@@ -471,7 +471,7 @@ void *rxe_pool_get_key(struct rxe_pool *pool, void *key)
 
 	read_lock_irqsave(&pool->pool_lock, flags);
 
-	if (pool->state != rxe_pool_valid)
+	if (pool->state != RXE_POOL_STATE_VALID)
 		goto out;
 
 	node = pool->tree.rb_node;
diff --git a/drivers/infiniband/sw/rxe/rxe_pool.h b/drivers/infiniband/sw/rxe/rxe_pool.h
index 6eb344236afc..aa4ba307097b 100644
--- a/drivers/infiniband/sw/rxe/rxe_pool.h
+++ b/drivers/infiniband/sw/rxe/rxe_pool.h
@@ -74,8 +74,8 @@ struct rxe_type_info {
 extern struct rxe_type_info rxe_type_info[];
 
 enum rxe_pool_state {
-	rxe_pool_invalid,
-	rxe_pool_valid,
+	RXE_POOL_STATE_INVALID,
+	RXE_POOL_STATE_VALID,
 };
 
 struct rxe_pool_entry {
-- 
cgit v1.2.3-55-g7522


From 536a631d1ec1de6ed51453954407068c10473eea Mon Sep 17 00:00:00 2001
From: Parav Pandit
Date: Mon, 27 Aug 2018 08:44:16 +0300
Subject: IB/rxe: Avoid NULL check when search is successful

While performing lookup in a pool, if entry is found, take the
reference right there, instead of checking again outside the loop and
save one branch.

Signed-off-by: Parav Pandit <parav@mellanox.com>
Reviewed-by: Daniel Jurgens <danielj@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Reviewed-by: Yuval Shaia <yuval.shaia@oracle.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
---
 drivers/infiniband/sw/rxe/rxe_pool.c | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)

(limited to 'drivers/infiniband/sw')

diff --git a/drivers/infiniband/sw/rxe/rxe_pool.c b/drivers/infiniband/sw/rxe/rxe_pool.c
index c8598b91e454..36b53fb94a49 100644
--- a/drivers/infiniband/sw/rxe/rxe_pool.c
+++ b/drivers/infiniband/sw/rxe/rxe_pool.c
@@ -450,13 +450,12 @@ void *rxe_pool_get_index(struct rxe_pool *pool, u32 index)
 			node = node->rb_left;
 		else if (elem->index < index)
 			node = node->rb_right;
-		else
+		else {
+			kref_get(&elem->ref_cnt);
 			break;
+		}
 	}
 
-	if (node)
-		kref_get(&elem->ref_cnt);
-
 out:
 	read_unlock_irqrestore(&pool->pool_lock, flags);
 	return node ? elem : NULL;
-- 
cgit v1.2.3-55-g7522


From 1703129ed2c12c112889f471bddf242012791b47 Mon Sep 17 00:00:00 2001
From: Parav Pandit
Date: Mon, 27 Aug 2018 08:44:17 +0300
Subject: IB/rxe: Refactor lookup memory function

Consolidate all error checks under single if() condition and use helper
unlikely() macro for them, in addition drop unneeded goto labels.

rxe_pool_get_index() already provides RB tree based efficient lookup.
Avoid doing extra checks for error cases which are rare and already
covered by rxe_pool_get_index().

Signed-off-by: Parav Pandit <parav@mellanox.com>
Reviewed-by: Daniel Jurgens <danielj@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Reviewed-by: Yuval Shaia <yuval.shaia@oracle.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
---
 drivers/infiniband/sw/rxe/rxe_mr.c | 35 +++++++++++------------------------
 1 file changed, 11 insertions(+), 24 deletions(-)

(limited to 'drivers/infiniband/sw')

diff --git a/drivers/infiniband/sw/rxe/rxe_mr.c b/drivers/infiniband/sw/rxe/rxe_mr.c
index dff605fdf60f..9d3916b93f23 100644
--- a/drivers/infiniband/sw/rxe/rxe_mr.c
+++ b/drivers/infiniband/sw/rxe/rxe_mr.c
@@ -573,33 +573,20 @@ struct rxe_mem *lookup_mem(struct rxe_pd *pd, int access, u32 key,
 	struct rxe_dev *rxe = to_rdev(pd->ibpd.device);
 	int index = key >> 8;
 
-	if (index >= RXE_MIN_MR_INDEX && index <= RXE_MAX_MR_INDEX) {
-		mem = rxe_pool_get_index(&rxe->mr_pool, index);
-		if (!mem)
-			goto err1;
-	} else {
-		goto err1;
+	mem = rxe_pool_get_index(&rxe->mr_pool, index);
+	if (!mem)
+		return NULL;
+
+	if (unlikely((type == lookup_local && mem->lkey != key) ||
+		     (type == lookup_remote && mem->rkey != key) ||
+		     mem->pd != pd ||
+		     (access && !(access & mem->access)) ||
+		     mem->state != RXE_MEM_STATE_VALID)) {
+		rxe_drop_ref(mem);
+		mem = NULL;
 	}
 
-	if ((type == lookup_local && mem->lkey != key) ||
-	    (type == lookup_remote && mem->rkey != key))
-		goto err2;
-
-	if (mem->pd != pd)
-		goto err2;
-
-	if (access && !(access & mem->access))
-		goto err2;
-
-	if (mem->state != RXE_MEM_STATE_VALID)
-		goto err2;
-
 	return mem;
-
-err2:
-	rxe_drop_ref(mem);
-err1:
-	return NULL;
 }
 
 int rxe_mem_map_pages(struct rxe_dev *rxe, struct rxe_mem *mem,
-- 
cgit v1.2.3-55-g7522


From b97db58557f4aa6d9903f8e1deea6b3d1ed0ba43 Mon Sep 17 00:00:00 2001
From: Vijay Immanuel
Date: Tue, 12 Jun 2018 18:20:49 -0700
Subject: IB/rxe: fix for duplicate request processing and ack psns

Don't reset the resp opcode for a replayed read response.
The resp opcode could be in the middle of a write or send
sequence, when the duplicate read request was received.
An example sequence is as follows:
- Receive read request for 12KB PSN 20. Transmit read response
  first, middle and last with PSNs 20,21,22.
- Receive write first PSN 23.
  At this point the resp psn is 24 and resp opcode is write first.
- The sender notices that PSN 20 is dropped and retransmits.
  Receive read request for 12KB PSN 20. Transmit read response
  first, middle and last with PSNs 20,21,22. The resp opcode is
  set to -1, the resp psn remains 24.
- Receive write first PSN 23. This is processed by duplicate_request().
  The resp opcode remains -1 and resp psn remains 24.
- Receive write middle PSN 24. check_op_seq() reports a missing
  first error since the resp opcode is -1.

When sending an ack for a duplicate send or write request,
use the psn of the previous ack sent. Do not use the psn
of a read response for the ack.
An example sequence is as follows:
- Receive write PSN 30. Transmit ACK for PSN 30.
- Receive read request 4KB PSN 31. Transmit read response with
  PSN 31. The resp psn is now 32.
- The sender notices that PSN 30 is dropped and retransmits.
  Receive write PSN 30. duplicate_request() sends an ACK with
  PSN 31. That is incorrect since PSN 31 was a read request.

Signed-off-by: Vijay Immanuel <vijayi@attalasystems.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
---
 drivers/infiniband/sw/rxe/rxe_resp.c  | 8 ++++++--
 drivers/infiniband/sw/rxe/rxe_verbs.h | 2 ++
 2 files changed, 8 insertions(+), 2 deletions(-)

(limited to 'drivers/infiniband/sw')

diff --git a/drivers/infiniband/sw/rxe/rxe_resp.c b/drivers/infiniband/sw/rxe/rxe_resp.c
index c1948d406528..c962160292f4 100644
--- a/drivers/infiniband/sw/rxe/rxe_resp.c
+++ b/drivers/infiniband/sw/rxe/rxe_resp.c
@@ -682,6 +682,7 @@ static enum resp_states read_reply(struct rxe_qp *qp,
 		rxe_advance_resp_resource(qp);
 
 		res->type		= RXE_READ_MASK;
+		res->replay		= 0;
 
 		res->read.va		= qp->resp.va;
 		res->read.va_org	= qp->resp.va;
@@ -752,7 +753,8 @@ static enum resp_states read_reply(struct rxe_qp *qp,
 		state = RESPST_DONE;
 	} else {
 		qp->resp.res = NULL;
-		qp->resp.opcode = -1;
+		if (!res->replay)
+			qp->resp.opcode = -1;
 		if (psn_compare(res->cur_psn, qp->resp.psn) >= 0)
 			qp->resp.psn = res->cur_psn;
 		state = RESPST_CLEANUP;
@@ -814,6 +816,7 @@ static enum resp_states execute(struct rxe_qp *qp, struct rxe_pkt_info *pkt)
 
 	/* next expected psn, read handles this separately */
 	qp->resp.psn = (pkt->psn + 1) & BTH_PSN_MASK;
+	qp->resp.ack_psn = qp->resp.psn;
 
 	qp->resp.opcode = pkt->opcode;
 	qp->resp.status = IB_WC_SUCCESS;
@@ -1065,7 +1068,7 @@ static enum resp_states duplicate_request(struct rxe_qp *qp,
 					  struct rxe_pkt_info *pkt)
 {
 	enum resp_states rc;
-	u32 prev_psn = (qp->resp.psn - 1) & BTH_PSN_MASK;
+	u32 prev_psn = (qp->resp.ack_psn - 1) & BTH_PSN_MASK;
 
 	if (pkt->mask & RXE_SEND_MASK ||
 	    pkt->mask & RXE_WRITE_MASK) {
@@ -1108,6 +1111,7 @@ static enum resp_states duplicate_request(struct rxe_qp *qp,
 			res->state = (pkt->psn == res->first_psn) ?
 					rdatm_res_state_new :
 					rdatm_res_state_replay;
+			res->replay = 1;
 
 			/* Reset the resource, except length. */
 			res->read.va_org = iova;
diff --git a/drivers/infiniband/sw/rxe/rxe_verbs.h b/drivers/infiniband/sw/rxe/rxe_verbs.h
index 0d920087811a..979e987e3c46 100644
--- a/drivers/infiniband/sw/rxe/rxe_verbs.h
+++ b/drivers/infiniband/sw/rxe/rxe_verbs.h
@@ -171,6 +171,7 @@ enum rdatm_res_state {
 
 struct resp_res {
 	int			type;
+	int			replay;
 	u32			first_psn;
 	u32			last_psn;
 	u32			cur_psn;
@@ -195,6 +196,7 @@ struct rxe_resp_info {
 	enum rxe_qp_state	state;
 	u32			msn;
 	u32			psn;
+	u32			ack_psn;
 	int			opcode;
 	int			drop_msg;
 	int			goto_error;
-- 
cgit v1.2.3-55-g7522


From 4e4c53df567714b3d08b2b5d8ccb1d175fc9be01 Mon Sep 17 00:00:00 2001
From: Vijay Immanuel
Date: Wed, 13 Jun 2018 18:47:30 -0700
Subject: IB/rxe: avoid back-to-back retries

Error retries can occur due to timeouts, NAKs or receiving
packets beyond the current read request. Avoid back-to-back
retries due to packet processing, by only retrying the initial
attempt immediately. Subsequent retries must be due to timeouts.

Continue to process completion packets after scheduling a retry.

Signed-off-by: Vijay Immanuel <vijayi@attalasystems.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
---
 drivers/infiniband/sw/rxe/rxe_comp.c  | 18 +++++++++++++++++-
 drivers/infiniband/sw/rxe/rxe_verbs.h |  1 +
 2 files changed, 18 insertions(+), 1 deletion(-)

(limited to 'drivers/infiniband/sw')

diff --git a/drivers/infiniband/sw/rxe/rxe_comp.c b/drivers/infiniband/sw/rxe/rxe_comp.c
index 83311dd07019..ed96441595d8 100644
--- a/drivers/infiniband/sw/rxe/rxe_comp.c
+++ b/drivers/infiniband/sw/rxe/rxe_comp.c
@@ -191,6 +191,7 @@ static inline void reset_retry_counters(struct rxe_qp *qp)
 {
 	qp->comp.retry_cnt = qp->attr.retry_cnt;
 	qp->comp.rnr_retry = qp->attr.rnr_retry;
+	qp->comp.started_retry = 0;
 }
 
 static inline enum comp_state check_psn(struct rxe_qp *qp,
@@ -676,6 +677,20 @@ int rxe_completer(void *arg)
 				goto exit;
 			}
 
+			/* if we've started a retry, don't start another
+			 * retry sequence, unless this is a timeout.
+			 */
+			if (qp->comp.started_retry &&
+			    !qp->comp.timeout_retry) {
+				if (pkt) {
+					rxe_drop_ref(pkt->qp);
+					kfree_skb(skb);
+					skb = NULL;
+				}
+
+				goto done;
+			}
+
 			if (qp->comp.retry_cnt > 0) {
 				if (qp->comp.retry_cnt != 7)
 					qp->comp.retry_cnt--;
@@ -692,6 +707,7 @@ int rxe_completer(void *arg)
 					rxe_counter_inc(rxe,
 							RXE_CNT_COMP_RETRY);
 					qp->req.need_retry = 1;
+					qp->comp.started_retry = 1;
 					rxe_run_task(&qp->req.task, 1);
 				}
 
@@ -701,7 +717,7 @@ int rxe_completer(void *arg)
 					skb = NULL;
 				}
 
-				goto exit;
+				goto done;
 
 			} else {
 				rxe_counter_inc(rxe, RXE_CNT_RETRY_EXCEEDED);
diff --git a/drivers/infiniband/sw/rxe/rxe_verbs.h b/drivers/infiniband/sw/rxe/rxe_verbs.h
index 979e987e3c46..82e670d6eeea 100644
--- a/drivers/infiniband/sw/rxe/rxe_verbs.h
+++ b/drivers/infiniband/sw/rxe/rxe_verbs.h
@@ -158,6 +158,7 @@ struct rxe_comp_info {
 	int			opcode;
 	int			timeout;
 	int			timeout_retry;
+	int			started_retry;
 	u32			retry_cnt;
 	u32			rnr_retry;
 	struct rxe_task		task;
-- 
cgit v1.2.3-55-g7522


From 030e46e495af855a13964a0aab9753ea82a96edc Mon Sep 17 00:00:00 2001
From: Vijay Immanuel
Date: Wed, 13 Jun 2018 18:48:07 -0700
Subject: IB/rxe: fixes for rdma read retry

When a read request is retried for the remaining partial
data, the response may restart from read response first
or read response only. So support those cases.

Do not advance the comp psn beyond the current wqe's last_psn
as that could skip over an entire read wqe and will cause the
req_retry() logic to set an incorrect req psn.
An example sequence is as follows:
Write        PSN 40 -- this is the current WQE.
Read request PSN 41
Write        PSN 42
Receive ACK  PSN 42 -- this will complete the current WQE
for PSN 40, and set the comp psn to 42 which is a problem
because the read request at PSN 41 has been skipped over.
So when req_retry() tries to retransmit the read request,
it sets the req psn to 42 which is incorrect.

When retrying a read request, calculate the number of psns
completed based on the dma resid instead of the wqe first_psn.
The wqe first_psn could have moved if the read request was
retried multiple times.

Set the reth length to the dma resid to handle read retries for
the remaining partial data.

Signed-off-by: Vijay Immanuel <vijayi@attalasystems.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
---
 drivers/infiniband/sw/rxe/rxe_comp.c | 21 ++++++++++++++++-----
 drivers/infiniband/sw/rxe/rxe_req.c  | 15 +++++++++------
 2 files changed, 25 insertions(+), 11 deletions(-)

(limited to 'drivers/infiniband/sw')

diff --git a/drivers/infiniband/sw/rxe/rxe_comp.c b/drivers/infiniband/sw/rxe/rxe_comp.c
index ed96441595d8..ea089cb091ad 100644
--- a/drivers/infiniband/sw/rxe/rxe_comp.c
+++ b/drivers/infiniband/sw/rxe/rxe_comp.c
@@ -254,6 +254,17 @@ static inline enum comp_state check_ack(struct rxe_qp *qp,
 	case IB_OPCODE_RC_RDMA_READ_RESPONSE_MIDDLE:
 		if (pkt->opcode != IB_OPCODE_RC_RDMA_READ_RESPONSE_MIDDLE &&
 		    pkt->opcode != IB_OPCODE_RC_RDMA_READ_RESPONSE_LAST) {
+			/* read retries of partial data may restart from
+			 * read response first or response only.
+			 */
+			if ((pkt->psn == wqe->first_psn &&
+			     pkt->opcode ==
+			     IB_OPCODE_RC_RDMA_READ_RESPONSE_FIRST) ||
+			    (wqe->first_psn == wqe->last_psn &&
+			     pkt->opcode ==
+			     IB_OPCODE_RC_RDMA_READ_RESPONSE_ONLY))
+				break;
+
 			return COMPST_ERROR;
 		}
 		break;
@@ -500,11 +511,11 @@ static inline enum comp_state complete_wqe(struct rxe_qp *qp,
 					   struct rxe_pkt_info *pkt,
 					   struct rxe_send_wqe *wqe)
 {
-	qp->comp.opcode = -1;
-
-	if (pkt) {
-		if (psn_compare(pkt->psn, qp->comp.psn) >= 0)
-			qp->comp.psn = (pkt->psn + 1) & BTH_PSN_MASK;
+	if (pkt && wqe->state == wqe_state_pending) {
+		if (psn_compare(wqe->last_psn, qp->comp.psn) >= 0) {
+			qp->comp.psn = (wqe->last_psn + 1) & BTH_PSN_MASK;
+			qp->comp.opcode = -1;
+		}
 
 		if (qp->req.wait_psn) {
 			qp->req.wait_psn = 0;
diff --git a/drivers/infiniband/sw/rxe/rxe_req.c b/drivers/infiniband/sw/rxe/rxe_req.c
index 84d35520781c..6c361d70d7cd 100644
--- a/drivers/infiniband/sw/rxe/rxe_req.c
+++ b/drivers/infiniband/sw/rxe/rxe_req.c
@@ -73,9 +73,6 @@ static void req_retry(struct rxe_qp *qp)
 	int npsn;
 	int first = 1;
 
-	wqe = queue_head(qp->sq.queue);
-	npsn = (qp->comp.psn - wqe->first_psn) & BTH_PSN_MASK;
-
 	qp->req.wqe_index	= consumer_index(qp->sq.queue);
 	qp->req.psn		= qp->comp.psn;
 	qp->req.opcode		= -1;
@@ -107,11 +104,17 @@ static void req_retry(struct rxe_qp *qp)
 		if (first) {
 			first = 0;
 
-			if (mask & WR_WRITE_OR_SEND_MASK)
+			if (mask & WR_WRITE_OR_SEND_MASK) {
+				npsn = (qp->comp.psn - wqe->first_psn) &
+					BTH_PSN_MASK;
 				retry_first_write_send(qp, wqe, mask, npsn);
+			}
 
-			if (mask & WR_READ_MASK)
+			if (mask & WR_READ_MASK) {
+				npsn = (wqe->dma.length - wqe->dma.resid) /
+					qp->mtu;
 				wqe->iova += npsn * qp->mtu;
+			}
 		}
 
 		wqe->state = wqe_state_posted;
@@ -435,7 +438,7 @@ static struct sk_buff *init_req_packet(struct rxe_qp *qp,
 	if (pkt->mask & RXE_RETH_MASK) {
 		reth_set_rkey(pkt, ibwr->wr.rdma.rkey);
 		reth_set_va(pkt, wqe->iova);
-		reth_set_len(pkt, wqe->dma.length);
+		reth_set_len(pkt, wqe->dma.resid);
 	}
 
 	if (pkt->mask & RXE_IMMDT_MASK)
-- 
cgit v1.2.3-55-g7522


From 0b79b27748cbec221e1ceabf63578198602bf01d Mon Sep 17 00:00:00 2001
From: Michael J. Ruhl
Date: Mon, 10 Sep 2018 09:49:27 -0700
Subject: IB/{hfi1, qib, rdmavt}: Schedule multi RC/UC packets instead of
 posting

The post_send() path determines if it should post directly or, schedule
the post for later.  The current logic is:

  if the swqe ring is empty or (for hfi1) wqe->length <= piothreshold
    post the send
  else
    schedule

This can allow large requests to call the send engine directly.  Large
requests can potentially produce a large number of packets prior to
returning to the caller, blocking the caller from posting more requests,
and allowing better parallel processing.

Allow the driver(s) more say in this logic (pass call_send to the driver,
rather than examining a return value).

Update hfi1/qib logic to schedule the send engine if an RC or UC message
is larger than the QP MTU size.

Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Reviewed-by: Ira Weiny <ira.weiny@intel.com>
Signed-off-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
---
 drivers/infiniband/hw/hfi1/qp.c       | 12 +++++-------
 drivers/infiniband/hw/hfi1/verbs.h    |  3 ++-
 drivers/infiniband/hw/qib/qib_qp.c    | 17 +++++++----------
 drivers/infiniband/hw/qib/qib_verbs.h |  3 ++-
 drivers/infiniband/sw/rdmavt/qp.c     | 14 ++++++++------
 include/rdma/rdma_vt.h                | 10 ++++++++--
 6 files changed, 32 insertions(+), 27 deletions(-)

(limited to 'drivers/infiniband/sw')

diff --git a/drivers/infiniband/hw/hfi1/qp.c b/drivers/infiniband/hw/hfi1/qp.c
index 9b1e84a6b1cc..54d9ff171059 100644
--- a/drivers/infiniband/hw/hfi1/qp.c
+++ b/drivers/infiniband/hw/hfi1/qp.c
@@ -285,17 +285,13 @@ void hfi1_modify_qp(struct rvt_qp *qp, struct ib_qp_attr *attr,
  * hfi1_check_send_wqe - validate wqe
  * @qp - The qp
  * @wqe - The built wqe
- *
- * validate wqe.  This is called
- * prior to inserting the wqe into
- * the ring but after the wqe has been
- * setup.
+ * @call_send - Determine if the send should be posted or scheduled.
  *
  * Returns 0 on success, -EINVAL on failure
  *
  */
 int hfi1_check_send_wqe(struct rvt_qp *qp,
-			struct rvt_swqe *wqe)
+			struct rvt_swqe *wqe, bool *call_send)
 {
 	struct hfi1_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
 	struct rvt_ah *ah;
@@ -305,6 +301,8 @@ int hfi1_check_send_wqe(struct rvt_qp *qp,
 	case IB_QPT_UC:
 		if (wqe->length > 0x80000000U)
 			return -EINVAL;
+		if (wqe->length > qp->pmtu)
+			*call_send = false;
 		break;
 	case IB_QPT_SMI:
 		ah = ibah_to_rvtah(wqe->ud_wr.ah);
@@ -321,7 +319,7 @@ int hfi1_check_send_wqe(struct rvt_qp *qp,
 	default:
 		break;
 	}
-	return wqe->length <= piothreshold;
+	return 0;
 }
 
 /**
diff --git a/drivers/infiniband/hw/hfi1/verbs.h b/drivers/infiniband/hw/hfi1/verbs.h
index a4d06502f06d..269ec338581b 100644
--- a/drivers/infiniband/hw/hfi1/verbs.h
+++ b/drivers/infiniband/hw/hfi1/verbs.h
@@ -343,7 +343,8 @@ int hfi1_check_modify_qp(struct rvt_qp *qp, struct ib_qp_attr *attr,
 void hfi1_modify_qp(struct rvt_qp *qp, struct ib_qp_attr *attr,
 		    int attr_mask, struct ib_udata *udata);
 void hfi1_restart_rc(struct rvt_qp *qp, u32 psn, int wait);
-int hfi1_check_send_wqe(struct rvt_qp *qp, struct rvt_swqe *wqe);
+int hfi1_check_send_wqe(struct rvt_qp *qp, struct rvt_swqe *wqe,
+			bool *call_send);
 
 extern const u32 rc_only_opcode;
 extern const u32 uc_only_opcode;
diff --git a/drivers/infiniband/hw/qib/qib_qp.c b/drivers/infiniband/hw/qib/qib_qp.c
index 344e401915f7..a81905df2d0f 100644
--- a/drivers/infiniband/hw/qib/qib_qp.c
+++ b/drivers/infiniband/hw/qib/qib_qp.c
@@ -378,25 +378,22 @@ void qib_flush_qp_waiters(struct rvt_qp *qp)
  * qib_check_send_wqe - validate wr/wqe
  * @qp - The qp
  * @wqe - The built wqe
+ * @call_send - Determine if the send should be posted or scheduled
  *
- * validate wr/wqe.  This is called
- * prior to inserting the wqe into
- * the ring but after the wqe has been
- * setup.
- *
- * Returns 1 to force direct progress, 0 otherwise, -EINVAL on failure
+ * Returns 0 on success, -EINVAL on failure
  */
 int qib_check_send_wqe(struct rvt_qp *qp,
-		       struct rvt_swqe *wqe)
+		       struct rvt_swqe *wqe, bool *call_send)
 {
 	struct rvt_ah *ah;
-	int ret = 0;
 
 	switch (qp->ibqp.qp_type) {
 	case IB_QPT_RC:
 	case IB_QPT_UC:
 		if (wqe->length > 0x80000000U)
 			return -EINVAL;
+		if (wqe->length > qp->pmtu)
+			*call_send = false;
 		break;
 	case IB_QPT_SMI:
 	case IB_QPT_GSI:
@@ -405,12 +402,12 @@ int qib_check_send_wqe(struct rvt_qp *qp,
 		if (wqe->length > (1 << ah->log_pmtu))
 			return -EINVAL;
 		/* progress hint */
-		ret = 1;
+		*call_send = true;
 		break;
 	default:
 		break;
 	}
-	return ret;
+	return 0;
 }
 
 #ifdef CONFIG_DEBUG_FS
diff --git a/drivers/infiniband/hw/qib/qib_verbs.h b/drivers/infiniband/hw/qib/qib_verbs.h
index 666613eef88f..3d7b744ae8fb 100644
--- a/drivers/infiniband/hw/qib/qib_verbs.h
+++ b/drivers/infiniband/hw/qib/qib_verbs.h
@@ -303,7 +303,8 @@ void qib_rc_rcv(struct qib_ctxtdata *rcd, struct ib_header *hdr,
 
 int qib_check_ah(struct ib_device *ibdev, struct rdma_ah_attr *ah_attr);
 
-int qib_check_send_wqe(struct rvt_qp *qp, struct rvt_swqe *wqe);
+int qib_check_send_wqe(struct rvt_qp *qp, struct rvt_swqe *wqe,
+		       bool *call_send);
 
 struct ib_ah *qib_create_qp0_ah(struct qib_ibport *ibp, u16 dlid);
 
diff --git a/drivers/infiniband/sw/rdmavt/qp.c b/drivers/infiniband/sw/rdmavt/qp.c
index 5ce403c6cddb..a9b7d7ff32ee 100644
--- a/drivers/infiniband/sw/rdmavt/qp.c
+++ b/drivers/infiniband/sw/rdmavt/qp.c
@@ -1718,7 +1718,7 @@ static inline int rvt_qp_is_avail(
  */
 static int rvt_post_one_wr(struct rvt_qp *qp,
 			   const struct ib_send_wr *wr,
-			   int *call_send)
+			   bool *call_send)
 {
 	struct rvt_swqe *wqe;
 	u32 next;
@@ -1825,11 +1825,9 @@ static int rvt_post_one_wr(struct rvt_qp *qp,
 
 	/* general part of wqe valid - allow for driver checks */
 	if (rdi->driver_f.check_send_wqe) {
-		ret = rdi->driver_f.check_send_wqe(qp, wqe);
+		ret = rdi->driver_f.check_send_wqe(qp, wqe, call_send);
 		if (ret < 0)
 			goto bail_inval_free;
-		if (ret)
-			*call_send = ret;
 	}
 
 	log_pmtu = qp->log_pmtu;
@@ -1897,7 +1895,7 @@ int rvt_post_send(struct ib_qp *ibqp, const struct ib_send_wr *wr,
 	struct rvt_qp *qp = ibqp_to_rvtqp(ibqp);
 	struct rvt_dev_info *rdi = ib_to_rvt(ibqp->device);
 	unsigned long flags = 0;
-	int call_send;
+	bool call_send;
 	unsigned nreq = 0;
 	int err = 0;
 
@@ -1930,7 +1928,11 @@ int rvt_post_send(struct ib_qp *ibqp, const struct ib_send_wr *wr,
 bail:
 	spin_unlock_irqrestore(&qp->s_hlock, flags);
 	if (nreq) {
-		if (call_send)
+		/*
+		 * Only call do_send if there is exactly one packet, and the
+		 * driver said it was ok.
+		 */
+		if (nreq == 1 && call_send)
 			rdi->driver_f.do_send(qp);
 		else
 			rdi->driver_f.schedule_send_no_lock(qp);
diff --git a/include/rdma/rdma_vt.h b/include/rdma/rdma_vt.h
index e79229a0cf01..e32facdd9fd3 100644
--- a/include/rdma/rdma_vt.h
+++ b/include/rdma/rdma_vt.h
@@ -214,8 +214,14 @@ struct rvt_driver_provided {
 	void (*schedule_send)(struct rvt_qp *qp);
 	void (*schedule_send_no_lock)(struct rvt_qp *qp);
 
-	/* Driver specific work request checking */
-	int (*check_send_wqe)(struct rvt_qp *qp, struct rvt_swqe *wqe);
+	/*
+	 * Validate the wqe.  This needs to be done prior to inserting the
+	 * wqe into the ring, but after the wqe has been set up.  Allow for
+	 * driver specific work request checking by providing a callback.
+	 * call_send indicates if the wqe should be posted or scheduled.
+	 */
+	int (*check_send_wqe)(struct rvt_qp *qp, struct rvt_swqe *wqe,
+			      bool *call_send);
 
 	/*
 	 * Sometimes rdmavt needs to kick the driver's send progress. That is
-- 
cgit v1.2.3-55-g7522


From e349f858d29f300ad9ad327fd57735a1d15e147f Mon Sep 17 00:00:00 2001
From: Jason Gunthorpe
Date: Tue, 25 Sep 2018 16:58:09 -0600
Subject: RDMA: Fully setup the device name in ib_register_device

The current code has two copies of the device name, ibdev->dev and
dev_name(&ibdev->dev), and they are setup at different times, which is
very confusing.

Set them both up at the same time and make dev_name() the lead name, which
is the proper use of the driver core APIs. To make it very clear that the
name is not valid until registration pass it in to the
ib_register_device() call rather than messing with ibdev->name directly.

Also the reorganization now checks that dev_name is unique even if it does
not contain a %.

Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Acked-by: Adit Ranadive <aditr@vmware.com>
Reviewed-by: Steve Wise <swise@opengridcomputing.com>
Acked-by: Devesh Sharma <devesh.sharma@broadcom.com>
Reviewed-by: Shiraz Saleem <shiraz.saleem@intel.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Reviewed-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
---
 drivers/infiniband/core/device.c               | 35 +++++++++++++++-----------
 drivers/infiniband/core/sysfs.c                |  4 ---
 drivers/infiniband/hw/bnxt_re/main.c           |  3 +--
 drivers/infiniband/hw/cxgb3/iwch_provider.c    |  3 +--
 drivers/infiniband/hw/cxgb4/provider.c         |  3 +--
 drivers/infiniband/hw/hns/hns_roce_main.c      |  3 +--
 drivers/infiniband/hw/i40iw/i40iw_verbs.c      |  3 +--
 drivers/infiniband/hw/mlx4/main.c              |  3 +--
 drivers/infiniband/hw/mlx5/main.c              | 15 ++++++-----
 drivers/infiniband/hw/mthca/mthca_provider.c   |  3 +--
 drivers/infiniband/hw/nes/nes_verbs.c          |  3 +--
 drivers/infiniband/hw/ocrdma/ocrdma_main.c     |  3 +--
 drivers/infiniband/hw/qedr/main.c              |  4 +--
 drivers/infiniband/hw/usnic/usnic_ib_main.c    |  3 +--
 drivers/infiniband/hw/vmw_pvrdma/pvrdma_main.c |  3 +--
 drivers/infiniband/sw/rdmavt/vt.c              |  3 ++-
 drivers/infiniband/sw/rxe/rxe_verbs.c          |  3 +--
 include/rdma/ib_verbs.h                        |  6 ++---
 include/rdma/rdma_vt.h                         |  9 ++++++-
 19 files changed, 53 insertions(+), 59 deletions(-)

(limited to 'drivers/infiniband/sw')

diff --git a/drivers/infiniband/core/device.c b/drivers/infiniband/core/device.c
index 5a680a88aa87..faacf95699d7 100644
--- a/drivers/infiniband/core/device.c
+++ b/drivers/infiniband/core/device.c
@@ -170,10 +170,9 @@ static struct ib_device *__ib_device_get_by_name(const char *name)
 	return NULL;
 }
 
-static int alloc_name(char *name)
+static int alloc_name(struct ib_device *ibdev, const char *name)
 {
 	unsigned long *inuse;
-	char buf[IB_DEVICE_NAME_MAX];
 	struct ib_device *device;
 	int i;
 
@@ -182,24 +181,21 @@ static int alloc_name(char *name)
 		return -ENOMEM;
 
 	list_for_each_entry(device, &device_list, core_list) {
-		if (!sscanf(device->name, name, &i))
+		char buf[IB_DEVICE_NAME_MAX];
+
+		if (sscanf(device->name, name, &i) != 1)
 			continue;
 		if (i < 0 || i >= PAGE_SIZE * 8)
 			continue;
 		snprintf(buf, sizeof buf, name, i);
-		if (!strncmp(buf, device->name, IB_DEVICE_NAME_MAX))
+		if (!strcmp(buf, dev_name(&device->dev)))
 			set_bit(i, inuse);
 	}
 
 	i = find_first_zero_bit(inuse, PAGE_SIZE * 8);
 	free_page((unsigned long) inuse);
-	snprintf(buf, sizeof buf, name, i);
-
-	if (__ib_device_get_by_name(buf))
-		return -ENFILE;
 
-	strlcpy(name, buf, IB_DEVICE_NAME_MAX);
-	return 0;
+	return dev_set_name(&ibdev->dev, name, i);
 }
 
 static void ib_device_release(struct device *device)
@@ -454,9 +450,9 @@ static u32 __dev_new_index(void)
  * callback for each device that is added. @device must be allocated
  * with ib_alloc_device().
  */
-int ib_register_device(struct ib_device *device,
-		       int (*port_callback)(struct ib_device *,
-					    u8, struct kobject *))
+int ib_register_device(struct ib_device *device, const char *name,
+		       int (*port_callback)(struct ib_device *, u8,
+					    struct kobject *))
 {
 	int ret;
 	struct ib_client *client;
@@ -495,11 +491,20 @@ int ib_register_device(struct ib_device *device,
 
 	mutex_lock(&device_mutex);
 
-	if (strchr(device->name, '%')) {
-		ret = alloc_name(device->name);
+	if (strchr(name, '%')) {
+		ret = alloc_name(device, name);
+		if (ret)
+			goto out;
+	} else {
+		ret = dev_set_name(&device->dev, name);
 		if (ret)
 			goto out;
 	}
+	if (__ib_device_get_by_name(dev_name(&device->dev))) {
+		ret = -ENFILE;
+		goto out;
+	}
+	strlcpy(device->name, dev_name(&device->dev), IB_DEVICE_NAME_MAX);
 
 	if (ib_device_check_mandatory(device)) {
 		ret = -EINVAL;
diff --git a/drivers/infiniband/core/sysfs.c b/drivers/infiniband/core/sysfs.c
index 0b04dbff884f..bc947a863b34 100644
--- a/drivers/infiniband/core/sysfs.c
+++ b/drivers/infiniband/core/sysfs.c
@@ -1311,10 +1311,6 @@ int ib_device_register_sysfs(struct ib_device *device,
 	int ret;
 	int i;
 
-	ret = dev_set_name(class_dev, "%s", device->name);
-	if (ret)
-		return ret;
-
 	device->groups[0] = &dev_attr_group;
 	class_dev->groups = device->groups;
 
diff --git a/drivers/infiniband/hw/bnxt_re/main.c b/drivers/infiniband/hw/bnxt_re/main.c
index 20b9f31052bf..73632e5b819f 100644
--- a/drivers/infiniband/hw/bnxt_re/main.c
+++ b/drivers/infiniband/hw/bnxt_re/main.c
@@ -579,7 +579,6 @@ static int bnxt_re_register_ib(struct bnxt_re_dev *rdev)
 	/* ib device init */
 	ibdev->owner = THIS_MODULE;
 	ibdev->node_type = RDMA_NODE_IB_CA;
-	strlcpy(ibdev->name, "bnxt_re%d", IB_DEVICE_NAME_MAX);
 	strlcpy(ibdev->node_desc, BNXT_RE_DESC " HCA",
 		strlen(BNXT_RE_DESC) + 5);
 	ibdev->phys_port_cnt = 1;
@@ -672,7 +671,7 @@ static int bnxt_re_register_ib(struct bnxt_re_dev *rdev)
 	ibdev->alloc_hw_stats           = bnxt_re_ib_alloc_hw_stats;
 
 	ibdev->driver_id = RDMA_DRIVER_BNXT_RE;
-	return ib_register_device(ibdev, NULL);
+	return ib_register_device(ibdev, "bnxt_re%d", NULL);
 }
 
 static ssize_t show_rev(struct device *device, struct device_attribute *attr,
diff --git a/drivers/infiniband/hw/cxgb3/iwch_provider.c b/drivers/infiniband/hw/cxgb3/iwch_provider.c
index 1b9ff21aa1d5..39530cc15f95 100644
--- a/drivers/infiniband/hw/cxgb3/iwch_provider.c
+++ b/drivers/infiniband/hw/cxgb3/iwch_provider.c
@@ -1319,7 +1319,6 @@ int iwch_register_device(struct iwch_dev *dev)
 	int i;
 
 	pr_debug("%s iwch_dev %p\n", __func__, dev);
-	strlcpy(dev->ibdev.name, "cxgb3_%d", IB_DEVICE_NAME_MAX);
 	memset(&dev->ibdev.node_guid, 0, sizeof(dev->ibdev.node_guid));
 	memcpy(&dev->ibdev.node_guid, dev->rdev.t3cdev_p->lldev->dev_addr, 6);
 	dev->ibdev.owner = THIS_MODULE;
@@ -1402,7 +1401,7 @@ int iwch_register_device(struct iwch_dev *dev)
 	       sizeof(dev->ibdev.iwcm->ifname));
 
 	dev->ibdev.driver_id = RDMA_DRIVER_CXGB3;
-	ret = ib_register_device(&dev->ibdev, NULL);
+	ret = ib_register_device(&dev->ibdev, "cxgb3_%d", NULL);
 	if (ret)
 		goto bail1;
 
diff --git a/drivers/infiniband/hw/cxgb4/provider.c b/drivers/infiniband/hw/cxgb4/provider.c
index 4eda6872e617..416f8d1af610 100644
--- a/drivers/infiniband/hw/cxgb4/provider.c
+++ b/drivers/infiniband/hw/cxgb4/provider.c
@@ -535,7 +535,6 @@ void c4iw_register_device(struct work_struct *work)
 	struct c4iw_dev *dev = ctx->dev;
 
 	pr_debug("c4iw_dev %p\n", dev);
-	strlcpy(dev->ibdev.name, "cxgb4_%d", IB_DEVICE_NAME_MAX);
 	memset(&dev->ibdev.node_guid, 0, sizeof(dev->ibdev.node_guid));
 	memcpy(&dev->ibdev.node_guid, dev->rdev.lldi.ports[0]->dev_addr, 6);
 	dev->ibdev.owner = THIS_MODULE;
@@ -627,7 +626,7 @@ void c4iw_register_device(struct work_struct *work)
 	       sizeof(dev->ibdev.iwcm->ifname));
 
 	dev->ibdev.driver_id = RDMA_DRIVER_CXGB4;
-	ret = ib_register_device(&dev->ibdev, NULL);
+	ret = ib_register_device(&dev->ibdev, "cxgb4_%d", NULL);
 	if (ret)
 		goto err_kfree_iwcm;
 
diff --git a/drivers/infiniband/hw/hns/hns_roce_main.c b/drivers/infiniband/hw/hns/hns_roce_main.c
index 6edb547baee8..5a86a48cba13 100644
--- a/drivers/infiniband/hw/hns/hns_roce_main.c
+++ b/drivers/infiniband/hw/hns/hns_roce_main.c
@@ -449,7 +449,6 @@ static int hns_roce_register_device(struct hns_roce_dev *hr_dev)
 	spin_lock_init(&iboe->lock);
 
 	ib_dev = &hr_dev->ib_dev;
-	strlcpy(ib_dev->name, "hns_%d", IB_DEVICE_NAME_MAX);
 
 	ib_dev->owner			= THIS_MODULE;
 	ib_dev->node_type		= RDMA_NODE_IB_CA;
@@ -530,7 +529,7 @@ static int hns_roce_register_device(struct hns_roce_dev *hr_dev)
 	ib_dev->disassociate_ucontext	= hns_roce_disassociate_ucontext;
 
 	ib_dev->driver_id = RDMA_DRIVER_HNS;
-	ret = ib_register_device(ib_dev, NULL);
+	ret = ib_register_device(ib_dev, "hns_%d", NULL);
 	if (ret) {
 		dev_err(dev, "ib_register_device failed!\n");
 		return ret;
diff --git a/drivers/infiniband/hw/i40iw/i40iw_verbs.c b/drivers/infiniband/hw/i40iw/i40iw_verbs.c
index e2e6c74a7452..cb2aef874ca8 100644
--- a/drivers/infiniband/hw/i40iw/i40iw_verbs.c
+++ b/drivers/infiniband/hw/i40iw/i40iw_verbs.c
@@ -2752,7 +2752,6 @@ static struct i40iw_ib_device *i40iw_init_rdma_device(struct i40iw_device *iwdev
 		i40iw_pr_err("iwdev == NULL\n");
 		return NULL;
 	}
-	strlcpy(iwibdev->ibdev.name, "i40iw%d", IB_DEVICE_NAME_MAX);
 	iwibdev->ibdev.owner = THIS_MODULE;
 	iwdev->iwibdev = iwibdev;
 	iwibdev->iwdev = iwdev;
@@ -2897,7 +2896,7 @@ int i40iw_register_rdma_device(struct i40iw_device *iwdev)
 	iwibdev = iwdev->iwibdev;
 
 	iwibdev->ibdev.driver_id = RDMA_DRIVER_I40IW;
-	ret = ib_register_device(&iwibdev->ibdev, NULL);
+	ret = ib_register_device(&iwibdev->ibdev, "i40iw%d", NULL);
 	if (ret)
 		goto error;
 
diff --git a/drivers/infiniband/hw/mlx4/main.c b/drivers/infiniband/hw/mlx4/main.c
index bf3cdb88aaf5..fa5d20eccc21 100644
--- a/drivers/infiniband/hw/mlx4/main.c
+++ b/drivers/infiniband/hw/mlx4/main.c
@@ -2540,7 +2540,6 @@ static void *mlx4_ib_add(struct mlx4_dev *dev)
 	ibdev->dev = dev;
 	ibdev->bond_next_port	= 0;
 
-	strlcpy(ibdev->ib_dev.name, "mlx4_%d", IB_DEVICE_NAME_MAX);
 	ibdev->ib_dev.owner		= THIS_MODULE;
 	ibdev->ib_dev.node_type		= RDMA_NODE_IB_CA;
 	ibdev->ib_dev.local_dma_lkey	= dev->caps.reserved_lkey;
@@ -2803,7 +2802,7 @@ static void *mlx4_ib_add(struct mlx4_dev *dev)
 		goto err_steer_free_bitmap;
 
 	ibdev->ib_dev.driver_id = RDMA_DRIVER_MLX4;
-	if (ib_register_device(&ibdev->ib_dev, NULL))
+	if (ib_register_device(&ibdev->ib_dev, "mlx4_%d", NULL))
 		goto err_diag_counters;
 
 	if (mlx4_ib_mad_init(ibdev))
diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c
index fb1e3c546826..597cd3c171c9 100644
--- a/drivers/infiniband/hw/mlx5/main.c
+++ b/drivers/infiniband/hw/mlx5/main.c
@@ -5671,7 +5671,6 @@ void mlx5_ib_stage_init_cleanup(struct mlx5_ib_dev *dev)
 int mlx5_ib_stage_init_init(struct mlx5_ib_dev *dev)
 {
 	struct mlx5_core_dev *mdev = dev->mdev;
-	const char *name;
 	int err;
 	int i;
 
@@ -5704,12 +5703,6 @@ int mlx5_ib_stage_init_init(struct mlx5_ib_dev *dev)
 	if (mlx5_use_mad_ifc(dev))
 		get_ext_port_caps(dev);
 
-	if (!mlx5_lag_is_active(mdev))
-		name = "mlx5_%d";
-	else
-		name = "mlx5_bond_%d";
-
-	strlcpy(dev->ib_dev.name, name, IB_DEVICE_NAME_MAX);
 	dev->ib_dev.owner		= THIS_MODULE;
 	dev->ib_dev.node_type		= RDMA_NODE_IB_CA;
 	dev->ib_dev.local_dma_lkey	= 0 /* not supported for now */;
@@ -6122,7 +6115,13 @@ static int mlx5_ib_stage_populate_specs(struct mlx5_ib_dev *dev)
 
 int mlx5_ib_stage_ib_reg_init(struct mlx5_ib_dev *dev)
 {
-	return ib_register_device(&dev->ib_dev, NULL);
+	const char *name;
+
+	if (!mlx5_lag_is_active(dev->mdev))
+		name = "mlx5_%d";
+	else
+		name = "mlx5_bond_%d";
+	return ib_register_device(&dev->ib_dev, name, NULL);
 }
 
 void mlx5_ib_stage_pre_ib_reg_umr_cleanup(struct mlx5_ib_dev *dev)
diff --git a/drivers/infiniband/hw/mthca/mthca_provider.c b/drivers/infiniband/hw/mthca/mthca_provider.c
index 0d3473b4596e..7bd7e2ad17e4 100644
--- a/drivers/infiniband/hw/mthca/mthca_provider.c
+++ b/drivers/infiniband/hw/mthca/mthca_provider.c
@@ -1198,7 +1198,6 @@ int mthca_register_device(struct mthca_dev *dev)
 	if (ret)
 		return ret;
 
-	strlcpy(dev->ib_dev.name, "mthca%d", IB_DEVICE_NAME_MAX);
 	dev->ib_dev.owner                = THIS_MODULE;
 
 	dev->ib_dev.uverbs_abi_ver	 = MTHCA_UVERBS_ABI_VERSION;
@@ -1297,7 +1296,7 @@ int mthca_register_device(struct mthca_dev *dev)
 	mutex_init(&dev->cap_mask_mutex);
 
 	dev->ib_dev.driver_id = RDMA_DRIVER_MTHCA;
-	ret = ib_register_device(&dev->ib_dev, NULL);
+	ret = ib_register_device(&dev->ib_dev, "mthca%d", NULL);
 	if (ret)
 		return ret;
 
diff --git a/drivers/infiniband/hw/nes/nes_verbs.c b/drivers/infiniband/hw/nes/nes_verbs.c
index 6940c7215961..2127cd2f4bec 100644
--- a/drivers/infiniband/hw/nes/nes_verbs.c
+++ b/drivers/infiniband/hw/nes/nes_verbs.c
@@ -3640,7 +3640,6 @@ struct nes_ib_device *nes_init_ofa_device(struct net_device *netdev)
 	if (nesibdev == NULL) {
 		return NULL;
 	}
-	strlcpy(nesibdev->ibdev.name, "nes%d", IB_DEVICE_NAME_MAX);
 	nesibdev->ibdev.owner = THIS_MODULE;
 
 	nesibdev->ibdev.node_type = RDMA_NODE_RNIC;
@@ -3798,7 +3797,7 @@ int nes_register_ofa_device(struct nes_ib_device *nesibdev)
 	int i, ret;
 
 	nesvnic->nesibdev->ibdev.driver_id = RDMA_DRIVER_NES;
-	ret = ib_register_device(&nesvnic->nesibdev->ibdev, NULL);
+	ret = ib_register_device(&nesvnic->nesibdev->ibdev, "nes%d", NULL);
 	if (ret) {
 		return ret;
 	}
diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_main.c b/drivers/infiniband/hw/ocrdma/ocrdma_main.c
index 7832ee3e0c84..4d3c27613351 100644
--- a/drivers/infiniband/hw/ocrdma/ocrdma_main.c
+++ b/drivers/infiniband/hw/ocrdma/ocrdma_main.c
@@ -116,7 +116,6 @@ static void get_dev_fw_str(struct ib_device *device, char *str)
 
 static int ocrdma_register_device(struct ocrdma_dev *dev)
 {
-	strlcpy(dev->ibdev.name, "ocrdma%d", IB_DEVICE_NAME_MAX);
 	ocrdma_get_guid(dev, (u8 *)&dev->ibdev.node_guid);
 	BUILD_BUG_ON(sizeof(OCRDMA_NODE_DESC) > IB_DEVICE_NODE_DESC_MAX);
 	memcpy(dev->ibdev.node_desc, OCRDMA_NODE_DESC,
@@ -214,7 +213,7 @@ static int ocrdma_register_device(struct ocrdma_dev *dev)
 		dev->ibdev.post_srq_recv = ocrdma_post_srq_recv;
 	}
 	dev->ibdev.driver_id = RDMA_DRIVER_OCRDMA;
-	return ib_register_device(&dev->ibdev, NULL);
+	return ib_register_device(&dev->ibdev, "ocrdma%d", NULL);
 }
 
 static int ocrdma_alloc_resources(struct ocrdma_dev *dev)
diff --git a/drivers/infiniband/hw/qedr/main.c b/drivers/infiniband/hw/qedr/main.c
index a0af6d424aed..cd7b8b39a129 100644
--- a/drivers/infiniband/hw/qedr/main.c
+++ b/drivers/infiniband/hw/qedr/main.c
@@ -170,8 +170,6 @@ static int qedr_register_device(struct qedr_dev *dev)
 {
 	int rc;
 
-	strlcpy(dev->ibdev.name, "qedr%d", IB_DEVICE_NAME_MAX);
-
 	dev->ibdev.node_guid = dev->attr.node_guid;
 	memcpy(dev->ibdev.node_desc, QEDR_NODE_DESC, sizeof(QEDR_NODE_DESC));
 	dev->ibdev.owner = THIS_MODULE;
@@ -264,7 +262,7 @@ static int qedr_register_device(struct qedr_dev *dev)
 	dev->ibdev.get_dev_fw_str = qedr_get_dev_fw_str;
 
 	dev->ibdev.driver_id = RDMA_DRIVER_QEDR;
-	return ib_register_device(&dev->ibdev, NULL);
+	return ib_register_device(&dev->ibdev, "qedr%d", NULL);
 }
 
 /* This function allocates fast-path status block memory */
diff --git a/drivers/infiniband/hw/usnic/usnic_ib_main.c b/drivers/infiniband/hw/usnic/usnic_ib_main.c
index f0538a460328..3b9f12928314 100644
--- a/drivers/infiniband/hw/usnic/usnic_ib_main.c
+++ b/drivers/infiniband/hw/usnic/usnic_ib_main.c
@@ -364,7 +364,6 @@ static void *usnic_ib_device_add(struct pci_dev *dev)
 	us_ibdev->ib_dev.num_comp_vectors = USNIC_IB_NUM_COMP_VECTORS;
 	us_ibdev->ib_dev.dev.parent = &dev->dev;
 	us_ibdev->ib_dev.uverbs_abi_ver = USNIC_UVERBS_ABI_VERSION;
-	strlcpy(us_ibdev->ib_dev.name, "usnic_%d", IB_DEVICE_NAME_MAX);
 
 	us_ibdev->ib_dev.uverbs_cmd_mask =
 		(1ull << IB_USER_VERBS_CMD_GET_CONTEXT) |
@@ -416,7 +415,7 @@ static void *usnic_ib_device_add(struct pci_dev *dev)
 
 
 	us_ibdev->ib_dev.driver_id = RDMA_DRIVER_USNIC;
-	if (ib_register_device(&us_ibdev->ib_dev, NULL))
+	if (ib_register_device(&us_ibdev->ib_dev, "usnic_%d", NULL))
 		goto err_fwd_dealloc;
 
 	usnic_fwd_set_mtu(us_ibdev->ufdev, us_ibdev->netdev->mtu);
diff --git a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_main.c b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_main.c
index a5719899f49a..6878107fc637 100644
--- a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_main.c
+++ b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_main.c
@@ -162,7 +162,6 @@ static int pvrdma_register_device(struct pvrdma_dev *dev)
 	int ret = -1;
 	int i = 0;
 
-	strlcpy(dev->ib_dev.name, "vmw_pvrdma%d", IB_DEVICE_NAME_MAX);
 	dev->ib_dev.node_guid = dev->dsr->caps.node_guid;
 	dev->sys_image_guid = dev->dsr->caps.sys_image_guid;
 	dev->flags = 0;
@@ -267,7 +266,7 @@ static int pvrdma_register_device(struct pvrdma_dev *dev)
 	dev->ib_dev.driver_id = RDMA_DRIVER_VMW_PVRDMA;
 	spin_lock_init(&dev->srq_tbl_lock);
 
-	ret = ib_register_device(&dev->ib_dev, NULL);
+	ret = ib_register_device(&dev->ib_dev, "vmw_pvrdma%d", NULL);
 	if (ret)
 		goto err_srq_free;
 
diff --git a/drivers/infiniband/sw/rdmavt/vt.c b/drivers/infiniband/sw/rdmavt/vt.c
index 17e4abc067af..e3249d46bcef 100644
--- a/drivers/infiniband/sw/rdmavt/vt.c
+++ b/drivers/infiniband/sw/rdmavt/vt.c
@@ -828,7 +828,8 @@ int rvt_register_device(struct rvt_dev_info *rdi, u32 driver_id)
 
 	rdi->ibdev.driver_id = driver_id;
 	/* We are now good to announce we exist */
-	ret =  ib_register_device(&rdi->ibdev, rdi->driver_f.port_callback);
+	ret = ib_register_device(&rdi->ibdev, dev_name(&rdi->ibdev.dev),
+				 rdi->driver_f.port_callback);
 	if (ret) {
 		rvt_pr_err(rdi, "Failed to register driver with ib core.\n");
 		goto bail_mr;
diff --git a/drivers/infiniband/sw/rxe/rxe_verbs.c b/drivers/infiniband/sw/rxe/rxe_verbs.c
index f5b1e0ad6142..e4da5b671e4a 100644
--- a/drivers/infiniband/sw/rxe/rxe_verbs.c
+++ b/drivers/infiniband/sw/rxe/rxe_verbs.c
@@ -1159,7 +1159,6 @@ int rxe_register_device(struct rxe_dev *rxe)
 	struct ib_device *dev = &rxe->ib_dev;
 	struct crypto_shash *tfm;
 
-	strlcpy(dev->name, "rxe%d", IB_DEVICE_NAME_MAX);
 	strlcpy(dev->node_desc, "rxe", sizeof(dev->node_desc));
 
 	dev->owner = THIS_MODULE;
@@ -1261,7 +1260,7 @@ int rxe_register_device(struct rxe_dev *rxe)
 	rxe->tfm = tfm;
 
 	dev->driver_id = RDMA_DRIVER_RXE;
-	err = ib_register_device(dev, NULL);
+	err = ib_register_device(dev, "rxe%d", NULL);
 	if (err) {
 		pr_warn("%s failed with error %d\n", __func__, err);
 		goto err1;
diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h
index 0d822a9db300..9897d2329f2c 100644
--- a/include/rdma/ib_verbs.h
+++ b/include/rdma/ib_verbs.h
@@ -2625,9 +2625,9 @@ void ib_dealloc_device(struct ib_device *device);
 
 void ib_get_device_fw_str(struct ib_device *device, char *str);
 
-int ib_register_device(struct ib_device *device,
-		       int (*port_callback)(struct ib_device *,
-					    u8, struct kobject *));
+int ib_register_device(struct ib_device *device, const char *name,
+		       int (*port_callback)(struct ib_device *, u8,
+					    struct kobject *));
 void ib_unregister_device(struct ib_device *device);
 
 int ib_register_client   (struct ib_client *client);
diff --git a/include/rdma/rdma_vt.h b/include/rdma/rdma_vt.h
index e32facdd9fd3..065c9fbe6589 100644
--- a/include/rdma/rdma_vt.h
+++ b/include/rdma/rdma_vt.h
@@ -429,7 +429,14 @@ static inline void rvt_set_ibdev_name(struct rvt_dev_info *rdi,
 				      const char *fmt, const char *name,
 				      const int unit)
 {
-	snprintf(rdi->ibdev.name, sizeof(rdi->ibdev.name), fmt, name, unit);
+	/*
+	 * FIXME: rvt and its users want to touch the ibdev before
+	 * registration and have things like the name work. We don't have the
+	 * infrastructure in the core to support this directly today, hack it
+	 * to work by setting the name manually here.
+	 */
+	dev_set_name(&rdi->ibdev.dev, fmt, name, unit);
+	strlcpy(rdi->ibdev.name, dev_name(&rdi->ibdev.dev), IB_DEVICE_NAME_MAX);
 }
 
 /**
-- 
cgit v1.2.3-55-g7522


From 5a738b5d47050b77ac8aa90bd79429940533ef6a Mon Sep 17 00:00:00 2001
From: Jason Gunthorpe
Date: Thu, 20 Sep 2018 16:42:24 -0600
Subject: RDMA/drivers: Use dev_err/dbg/etc instead of pr_* + ibdev->name

Kernel convention is that a driver for a subsystem will print using
dev_* on the subsystem's struct device, or with dev_* on the physical
device. Drivers should rarely use a pr_* function.

Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
---
 drivers/infiniband/hw/mlx5/mlx5_ib.h     | 18 +++++++++---------
 drivers/infiniband/hw/mlx5/qp.c          | 10 ++++++----
 drivers/infiniband/hw/mthca/mthca_mad.c  |  5 +++--
 drivers/infiniband/hw/ocrdma/ocrdma_hw.c |  2 +-
 drivers/infiniband/sw/rxe/rxe_net.c      |  4 ++--
 drivers/infiniband/sw/rxe/rxe_sysfs.c    |  2 +-
 6 files changed, 22 insertions(+), 19 deletions(-)

(limited to 'drivers/infiniband/sw')

diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h
index 94cf83bd1716..5ada28d6e5e9 100644
--- a/drivers/infiniband/hw/mlx5/mlx5_ib.h
+++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h
@@ -50,17 +50,17 @@
 #include <rdma/uverbs_ioctl.h>
 #include <rdma/mlx5_user_ioctl_cmds.h>
 
-#define mlx5_ib_dbg(dev, format, arg...)				\
-pr_debug("%s:%s:%d:(pid %d): " format, (dev)->ib_dev.name, __func__,	\
-	 __LINE__, current->pid, ##arg)
+#define mlx5_ib_dbg(_dev, format, arg...)                                      \
+	dev_dbg(&(_dev)->ib_dev.dev, "%s:%d:(pid %d): " format, __func__,      \
+		__LINE__, current->pid, ##arg)
 
-#define mlx5_ib_err(dev, format, arg...)				\
-pr_err("%s:%s:%d:(pid %d): " format, (dev)->ib_dev.name, __func__,	\
-	__LINE__, current->pid, ##arg)
+#define mlx5_ib_err(_dev, format, arg...)                                      \
+	dev_err(&(_dev)->ib_dev.dev, "%s:%d:(pid %d): " format, __func__,      \
+		__LINE__, current->pid, ##arg)
 
-#define mlx5_ib_warn(dev, format, arg...)				\
-pr_warn("%s:%s:%d:(pid %d): " format, (dev)->ib_dev.name, __func__,	\
-	__LINE__, current->pid, ##arg)
+#define mlx5_ib_warn(_dev, format, arg...)                                     \
+	dev_warn(&(_dev)->ib_dev.dev, "%s:%d:(pid %d): " format, __func__,     \
+		 __LINE__, current->pid, ##arg)
 
 #define field_avail(type, fld, sz) (offsetof(type, fld) +		\
 				    sizeof(((type *)0)->fld) <= (sz))
diff --git a/drivers/infiniband/hw/mlx5/qp.c b/drivers/infiniband/hw/mlx5/qp.c
index a0cb260a271f..3455b50705cd 100644
--- a/drivers/infiniband/hw/mlx5/qp.c
+++ b/drivers/infiniband/hw/mlx5/qp.c
@@ -2826,8 +2826,9 @@ static int modify_raw_packet_qp_rq(
 				   MLX5_MODIFY_RQ_IN_MODIFY_BITMASK_RQ_COUNTER_SET_ID);
 			MLX5_SET(rqc, rqc, counter_set_id, raw_qp_param->rq_q_ctr_id);
 		} else
-			pr_info_once("%s: RAW PACKET QP counters are not supported on current FW\n",
-				     dev->ib_dev.name);
+			dev_info_once(
+				&dev->ib_dev.dev,
+				"RAW PACKET QP counters are not supported on current FW\n");
 	}
 
 	err = mlx5_core_modify_rq(dev->mdev, rq->base.mqp.qpn, in, inlen);
@@ -5798,8 +5799,9 @@ int mlx5_ib_modify_wq(struct ib_wq *wq, struct ib_wq_attr *wq_attr,
 			MLX5_SET(rqc, rqc, counter_set_id,
 				 dev->port->cnts.set_id);
 		} else
-			pr_info_once("%s: Receive WQ counters are not supported on current FW\n",
-				     dev->ib_dev.name);
+			dev_info_once(
+				&dev->ib_dev.dev,
+				"Receive WQ counters are not supported on current FW\n");
 	}
 
 	err = mlx5_core_modify_rq(dev->mdev, rwq->core_qp.qpn, in, inlen);
diff --git a/drivers/infiniband/hw/mthca/mthca_mad.c b/drivers/infiniband/hw/mthca/mthca_mad.c
index 093f7755c843..2e5dc0a67cfc 100644
--- a/drivers/infiniband/hw/mthca/mthca_mad.c
+++ b/drivers/infiniband/hw/mthca/mthca_mad.c
@@ -58,8 +58,9 @@ static int mthca_update_rate(struct mthca_dev *dev, u8 port_num)
 
 	ret = ib_query_port(&dev->ib_dev, port_num, tprops);
 	if (ret) {
-		printk(KERN_WARNING "ib_query_port failed (%d) for %s port %d\n",
-		       ret, dev->ib_dev.name, port_num);
+		dev_warn(&dev->ib_dev.dev,
+			 "ib_query_port failed (%d) forport %d\n", ret,
+			 port_num);
 		goto out;
 	}
 
diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_hw.c b/drivers/infiniband/hw/ocrdma/ocrdma_hw.c
index e578281471af..241a57a07485 100644
--- a/drivers/infiniband/hw/ocrdma/ocrdma_hw.c
+++ b/drivers/infiniband/hw/ocrdma/ocrdma_hw.c
@@ -792,7 +792,7 @@ static void ocrdma_dispatch_ibevent(struct ocrdma_dev *dev,
 						     qp->srq->ibsrq.
 						     srq_context);
 	} else if (dev_event) {
-		pr_err("%s: Fatal event received\n", dev->ibdev.name);
+		dev_err(&dev->ibdev.dev, "Fatal event received\n");
 		ib_dispatch_event(&ib_evt);
 	}
 
diff --git a/drivers/infiniband/sw/rxe/rxe_net.c b/drivers/infiniband/sw/rxe/rxe_net.c
index ed823cccca48..8e658ce439b4 100644
--- a/drivers/infiniband/sw/rxe/rxe_net.c
+++ b/drivers/infiniband/sw/rxe/rxe_net.c
@@ -615,7 +615,7 @@ void rxe_port_up(struct rxe_dev *rxe)
 	port->attr.phys_state = IB_PHYS_STATE_LINK_UP;
 
 	rxe_port_event(rxe, IB_EVENT_PORT_ACTIVE);
-	pr_info("set %s active\n", rxe->ib_dev.name);
+	dev_info(&rxe->ib_dev.dev, "set active\n");
 }
 
 /* Caller must hold net_info_lock */
@@ -628,7 +628,7 @@ void rxe_port_down(struct rxe_dev *rxe)
 	port->attr.phys_state = IB_PHYS_STATE_LINK_DOWN;
 
 	rxe_port_event(rxe, IB_EVENT_PORT_ERR);
-	pr_info("set %s down\n", rxe->ib_dev.name);
+	dev_info(&rxe->ib_dev.dev, "set down\n");
 }
 
 static int rxe_notify(struct notifier_block *not_blk,
diff --git a/drivers/infiniband/sw/rxe/rxe_sysfs.c b/drivers/infiniband/sw/rxe/rxe_sysfs.c
index d5ed7571128f..73a19f808e1b 100644
--- a/drivers/infiniband/sw/rxe/rxe_sysfs.c
+++ b/drivers/infiniband/sw/rxe/rxe_sysfs.c
@@ -105,7 +105,7 @@ static int rxe_param_set_add(const char *val, const struct kernel_param *kp)
 	}
 
 	rxe_set_port_state(ndev);
-	pr_info("added %s to %s\n", rxe->ib_dev.name, intf);
+	dev_info(&rxe->ib_dev.dev, "added %s\n", intf);
 err:
 	if (ndev)
 		dev_put(ndev);
-- 
cgit v1.2.3-55-g7522


From 9de6986148360bdc954e8621cf024f4c830ddb13 Mon Sep 17 00:00:00 2001
From: Jason Gunthorpe
Date: Thu, 20 Sep 2018 16:42:26 -0600
Subject: RDMA/drivers: Use dev_name instead of ibdev->name

These return the same thing but dev_name is a more conventional use of the
kernel API.

Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
---
 drivers/infiniband/hw/nes/nes_verbs.c          |  2 +-
 drivers/infiniband/hw/ocrdma/ocrdma_stats.c    |  3 ++-
 drivers/infiniband/hw/qedr/qedr.h              |  2 +-
 drivers/infiniband/hw/usnic/usnic_ib_main.c    | 34 ++++++++++++++------------
 drivers/infiniband/hw/usnic/usnic_ib_sysfs.c   |  6 ++---
 drivers/infiniband/hw/usnic/usnic_ib_verbs.c   | 16 ++++++------
 drivers/infiniband/hw/vmw_pvrdma/pvrdma_main.c |  2 +-
 drivers/infiniband/sw/rxe/rxe_net.c            |  2 +-
 8 files changed, 36 insertions(+), 31 deletions(-)

(limited to 'drivers/infiniband/sw')

diff --git a/drivers/infiniband/hw/nes/nes_verbs.c b/drivers/infiniband/hw/nes/nes_verbs.c
index 2127cd2f4bec..94054bc611bd 100644
--- a/drivers/infiniband/hw/nes/nes_verbs.c
+++ b/drivers/infiniband/hw/nes/nes_verbs.c
@@ -687,7 +687,7 @@ static struct ib_pd *nes_alloc_pd(struct ib_device *ibdev,
 	}
 
 	nes_debug(NES_DBG_PD, "Allocating PD (%p) for ib device %s\n",
-			nespd, nesvnic->nesibdev->ibdev.name);
+			nespd, dev_name(&nesvnic->nesibdev->ibdev.dev));
 
 	nespd->pd_id = (pd_num << (PAGE_SHIFT-12)) + nesadapter->base_pd;
 
diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_stats.c b/drivers/infiniband/hw/ocrdma/ocrdma_stats.c
index 24d20a4aa262..290d776edf48 100644
--- a/drivers/infiniband/hw/ocrdma/ocrdma_stats.c
+++ b/drivers/infiniband/hw/ocrdma/ocrdma_stats.c
@@ -764,7 +764,8 @@ void ocrdma_add_port_stats(struct ocrdma_dev *dev)
 		return;
 
 	/* Create post stats base dir */
-	dev->dir = debugfs_create_dir(dev->ibdev.name, ocrdma_dbgfs_dir);
+	dev->dir =
+		debugfs_create_dir(dev_name(&dev->ibdev.dev), ocrdma_dbgfs_dir);
 	if (!dev->dir)
 		goto err;
 
diff --git a/drivers/infiniband/hw/qedr/qedr.h b/drivers/infiniband/hw/qedr/qedr.h
index a2d708dceb8d..53bbe6b4e6e6 100644
--- a/drivers/infiniband/hw/qedr/qedr.h
+++ b/drivers/infiniband/hw/qedr/qedr.h
@@ -43,7 +43,7 @@
 #include "qedr_hsi_rdma.h"
 
 #define QEDR_NODE_DESC "QLogic 579xx RoCE HCA"
-#define DP_NAME(dev) ((dev)->ibdev.name)
+#define DP_NAME(_dev) dev_name(&(_dev)->ibdev.dev)
 #define IS_IWARP(_dev) ((_dev)->rdma_type == QED_RDMA_TYPE_IWARP)
 #define IS_ROCE(_dev) ((_dev)->rdma_type == QED_RDMA_TYPE_ROCE)
 
diff --git a/drivers/infiniband/hw/usnic/usnic_ib_main.c b/drivers/infiniband/hw/usnic/usnic_ib_main.c
index 3b9f12928314..e3f98fe2f1e6 100644
--- a/drivers/infiniband/hw/usnic/usnic_ib_main.c
+++ b/drivers/infiniband/hw/usnic/usnic_ib_main.c
@@ -76,7 +76,7 @@ static LIST_HEAD(usnic_ib_ibdev_list);
 static int usnic_ib_dump_vf_hdr(void *obj, char *buf, int buf_sz)
 {
 	struct usnic_ib_vf *vf = obj;
-	return scnprintf(buf, buf_sz, "PF: %s ", vf->pf->ib_dev.name);
+	return scnprintf(buf, buf_sz, "PF: %s ", dev_name(&vf->pf->ib_dev.dev));
 }
 /* End callback dump funcs */
 
@@ -138,7 +138,7 @@ static void usnic_ib_handle_usdev_event(struct usnic_ib_dev *us_ibdev,
 	netdev = us_ibdev->netdev;
 	switch (event) {
 	case NETDEV_REBOOT:
-		usnic_info("PF Reset on %s\n", us_ibdev->ib_dev.name);
+		usnic_info("PF Reset on %s\n", dev_name(&us_ibdev->ib_dev.dev));
 		usnic_ib_qp_grp_modify_active_to_err(us_ibdev);
 		ib_event.event = IB_EVENT_PORT_ERR;
 		ib_event.device = &us_ibdev->ib_dev;
@@ -151,7 +151,8 @@ static void usnic_ib_handle_usdev_event(struct usnic_ib_dev *us_ibdev,
 		if (!us_ibdev->ufdev->link_up &&
 				netif_carrier_ok(netdev)) {
 			usnic_fwd_carrier_up(us_ibdev->ufdev);
-			usnic_info("Link UP on %s\n", us_ibdev->ib_dev.name);
+			usnic_info("Link UP on %s\n",
+				   dev_name(&us_ibdev->ib_dev.dev));
 			ib_event.event = IB_EVENT_PORT_ACTIVE;
 			ib_event.device = &us_ibdev->ib_dev;
 			ib_event.element.port_num = 1;
@@ -159,7 +160,8 @@ static void usnic_ib_handle_usdev_event(struct usnic_ib_dev *us_ibdev,
 		} else if (us_ibdev->ufdev->link_up &&
 				!netif_carrier_ok(netdev)) {
 			usnic_fwd_carrier_down(us_ibdev->ufdev);
-			usnic_info("Link DOWN on %s\n", us_ibdev->ib_dev.name);
+			usnic_info("Link DOWN on %s\n",
+				   dev_name(&us_ibdev->ib_dev.dev));
 			usnic_ib_qp_grp_modify_active_to_err(us_ibdev);
 			ib_event.event = IB_EVENT_PORT_ERR;
 			ib_event.device = &us_ibdev->ib_dev;
@@ -168,17 +170,17 @@ static void usnic_ib_handle_usdev_event(struct usnic_ib_dev *us_ibdev,
 		} else {
 			usnic_dbg("Ignoring %s on %s\n",
 					netdev_cmd_to_name(event),
-					us_ibdev->ib_dev.name);
+					dev_name(&us_ibdev->ib_dev.dev));
 		}
 		break;
 	case NETDEV_CHANGEADDR:
 		if (!memcmp(us_ibdev->ufdev->mac, netdev->dev_addr,
 				sizeof(us_ibdev->ufdev->mac))) {
 			usnic_dbg("Ignoring addr change on %s\n",
-					us_ibdev->ib_dev.name);
+				  dev_name(&us_ibdev->ib_dev.dev));
 		} else {
 			usnic_info(" %s old mac: %pM new mac: %pM\n",
-					us_ibdev->ib_dev.name,
+					dev_name(&us_ibdev->ib_dev.dev),
 					us_ibdev->ufdev->mac,
 					netdev->dev_addr);
 			usnic_fwd_set_mac(us_ibdev->ufdev, netdev->dev_addr);
@@ -193,19 +195,19 @@ static void usnic_ib_handle_usdev_event(struct usnic_ib_dev *us_ibdev,
 	case NETDEV_CHANGEMTU:
 		if (us_ibdev->ufdev->mtu != netdev->mtu) {
 			usnic_info("MTU Change on %s old: %u new: %u\n",
-					us_ibdev->ib_dev.name,
+					dev_name(&us_ibdev->ib_dev.dev),
 					us_ibdev->ufdev->mtu, netdev->mtu);
 			usnic_fwd_set_mtu(us_ibdev->ufdev, netdev->mtu);
 			usnic_ib_qp_grp_modify_active_to_err(us_ibdev);
 		} else {
 			usnic_dbg("Ignoring MTU change on %s\n",
-					us_ibdev->ib_dev.name);
+				  dev_name(&us_ibdev->ib_dev.dev));
 		}
 		break;
 	default:
 		usnic_dbg("Ignoring event %s on %s",
 				netdev_cmd_to_name(event),
-				us_ibdev->ib_dev.name);
+				dev_name(&us_ibdev->ib_dev.dev));
 	}
 	mutex_unlock(&us_ibdev->usdev_lock);
 }
@@ -267,7 +269,7 @@ static int usnic_ib_handle_inet_event(struct usnic_ib_dev *us_ibdev,
 	default:
 		usnic_info("Ignoring event %s on %s",
 				netdev_cmd_to_name(event),
-				us_ibdev->ib_dev.name);
+				dev_name(&us_ibdev->ib_dev.dev));
 	}
 	mutex_unlock(&us_ibdev->usdev_lock);
 
@@ -436,9 +438,9 @@ static void *usnic_ib_device_add(struct pci_dev *dev)
 	kref_init(&us_ibdev->vf_cnt);
 
 	usnic_info("Added ibdev: %s netdev: %s with mac %pM Link: %u MTU: %u\n",
-			us_ibdev->ib_dev.name, netdev_name(us_ibdev->netdev),
-			us_ibdev->ufdev->mac, us_ibdev->ufdev->link_up,
-			us_ibdev->ufdev->mtu);
+		   dev_name(&us_ibdev->ib_dev.dev),
+		   netdev_name(us_ibdev->netdev), us_ibdev->ufdev->mac,
+		   us_ibdev->ufdev->link_up, us_ibdev->ufdev->mtu);
 	return us_ibdev;
 
 err_fwd_dealloc:
@@ -451,7 +453,7 @@ err_dealloc:
 
 static void usnic_ib_device_remove(struct usnic_ib_dev *us_ibdev)
 {
-	usnic_info("Unregistering %s\n", us_ibdev->ib_dev.name);
+	usnic_info("Unregistering %s\n", dev_name(&us_ibdev->ib_dev.dev));
 	usnic_ib_sysfs_unregister_usdev(us_ibdev);
 	usnic_fwd_dev_free(us_ibdev->ufdev);
 	ib_unregister_device(&us_ibdev->ib_dev);
@@ -590,7 +592,7 @@ static int usnic_ib_pci_probe(struct pci_dev *pdev,
 	mutex_unlock(&pf->usdev_lock);
 
 	usnic_info("Registering usnic VF %s into PF %s\n", pci_name(pdev),
-			pf->ib_dev.name);
+		   dev_name(&pf->ib_dev.dev));
 	usnic_ib_log_vf(vf);
 	return 0;
 
diff --git a/drivers/infiniband/hw/usnic/usnic_ib_sysfs.c b/drivers/infiniband/hw/usnic/usnic_ib_sysfs.c
index 4210ca14014d..fab4cb780122 100644
--- a/drivers/infiniband/hw/usnic/usnic_ib_sysfs.c
+++ b/drivers/infiniband/hw/usnic/usnic_ib_sysfs.c
@@ -94,7 +94,7 @@ usnic_ib_show_config(struct device *device, struct device_attribute *attr,
 
 		n = scnprintf(ptr, left,
 			"%s: %s:%d.%d, %s, %pM, %u VFs\n Per VF:",
-			us_ibdev->ib_dev.name,
+			dev_name(&us_ibdev->ib_dev.dev),
 			busname,
 			PCI_SLOT(us_ibdev->pdev->devfn),
 			PCI_FUNC(us_ibdev->pdev->devfn),
@@ -119,7 +119,7 @@ usnic_ib_show_config(struct device *device, struct device_attribute *attr,
 		UPDATE_PTR_LEFT(n, ptr, left);
 	} else {
 		n = scnprintf(ptr, left, "%s: no VFs\n",
-				us_ibdev->ib_dev.name);
+				dev_name(&us_ibdev->ib_dev.dev));
 		UPDATE_PTR_LEFT(n, ptr, left);
 	}
 	mutex_unlock(&us_ibdev->usdev_lock);
@@ -285,7 +285,7 @@ int usnic_ib_sysfs_register_usdev(struct usnic_ib_dev *us_ibdev)
 						usnic_class_attributes[i]);
 		if (err) {
 			usnic_err("Failed to create device file %d for %s eith err %d",
-				i, us_ibdev->ib_dev.name, err);
+				i, dev_name(&us_ibdev->ib_dev.dev), err);
 			return -EINVAL;
 		}
 	}
diff --git a/drivers/infiniband/hw/usnic/usnic_ib_verbs.c b/drivers/infiniband/hw/usnic/usnic_ib_verbs.c
index 9973ac893635..0b91ff36768a 100644
--- a/drivers/infiniband/hw/usnic/usnic_ib_verbs.c
+++ b/drivers/infiniband/hw/usnic/usnic_ib_verbs.c
@@ -159,7 +159,8 @@ static int usnic_ib_fill_create_qp_resp(struct usnic_ib_qp_grp *qp_grp,
 
 	err = ib_copy_to_udata(udata, &resp, sizeof(resp));
 	if (err) {
-		usnic_err("Failed to copy udata for %s", us_ibdev->ib_dev.name);
+		usnic_err("Failed to copy udata for %s",
+			  dev_name(&us_ibdev->ib_dev.dev));
 		return err;
 	}
 
@@ -197,7 +198,7 @@ find_free_vf_and_create_qp_grp(struct usnic_ib_dev *us_ibdev,
 			vnic = vf->vnic;
 			if (!usnic_vnic_check_room(vnic, res_spec)) {
 				usnic_dbg("Found used vnic %s from %s\n",
-						us_ibdev->ib_dev.name,
+						dev_name(&us_ibdev->ib_dev.dev),
 						pci_name(usnic_vnic_get_pdev(
 									vnic)));
 				qp_grp = usnic_ib_qp_grp_create(us_ibdev->ufdev,
@@ -230,7 +231,8 @@ find_free_vf_and_create_qp_grp(struct usnic_ib_dev *us_ibdev,
 		spin_unlock(&vf->lock);
 	}
 
-	usnic_info("No free qp grp found on %s\n", us_ibdev->ib_dev.name);
+	usnic_info("No free qp grp found on %s\n",
+		   dev_name(&us_ibdev->ib_dev.dev));
 	return ERR_PTR(-ENOMEM);
 
 qp_grp_check:
@@ -471,7 +473,7 @@ struct ib_pd *usnic_ib_alloc_pd(struct ib_device *ibdev,
 	}
 
 	usnic_info("domain 0x%p allocated for context 0x%p and device %s\n",
-			pd, context, ibdev->name);
+		   pd, context, dev_name(&ibdev->dev));
 	return &pd->ibpd;
 }
 
@@ -508,20 +510,20 @@ struct ib_qp *usnic_ib_create_qp(struct ib_pd *pd,
 	err = ib_copy_from_udata(&cmd, udata, sizeof(cmd));
 	if (err) {
 		usnic_err("%s: cannot copy udata for create_qp\n",
-				us_ibdev->ib_dev.name);
+			  dev_name(&us_ibdev->ib_dev.dev));
 		return ERR_PTR(-EINVAL);
 	}
 
 	err = create_qp_validate_user_data(cmd);
 	if (err) {
 		usnic_err("%s: Failed to validate user data\n",
-				us_ibdev->ib_dev.name);
+			  dev_name(&us_ibdev->ib_dev.dev));
 		return ERR_PTR(-EINVAL);
 	}
 
 	if (init_attr->qp_type != IB_QPT_UD) {
 		usnic_err("%s asked to make a non-UD QP: %d\n",
-				us_ibdev->ib_dev.name, init_attr->qp_type);
+			  dev_name(&us_ibdev->ib_dev.dev), init_attr->qp_type);
 		return ERR_PTR(-EINVAL);
 	}
 
diff --git a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_main.c b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_main.c
index 6878107fc637..c1e31985b11c 100644
--- a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_main.c
+++ b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_main.c
@@ -734,7 +734,7 @@ static void pvrdma_netdevice_event_handle(struct pvrdma_dev *dev,
 
 	default:
 		dev_dbg(&dev->pdev->dev, "ignore netdevice event %ld on %s\n",
-			event, dev->ib_dev.name);
+			event, dev_name(&dev->ib_dev.dev));
 		break;
 	}
 }
diff --git a/drivers/infiniband/sw/rxe/rxe_net.c b/drivers/infiniband/sw/rxe/rxe_net.c
index 8e658ce439b4..fb06f94f33d8 100644
--- a/drivers/infiniband/sw/rxe/rxe_net.c
+++ b/drivers/infiniband/sw/rxe/rxe_net.c
@@ -72,7 +72,7 @@ struct rxe_dev *get_rxe_by_name(const char *name)
 
 	spin_lock_bh(&dev_list_lock);
 	list_for_each_entry(rxe, &rxe_dev_list, list) {
-		if (!strcmp(name, rxe->ib_dev.name)) {
+		if (!strcmp(name, dev_name(&rxe->ib_dev.dev))) {
 			found = rxe;
 			break;
 		}
-- 
cgit v1.2.3-55-g7522


From 721ad7e643f7002efa398838693f90284ea216d1 Mon Sep 17 00:00:00 2001
From: Zhu Yanjun
Date: Sun, 30 Sep 2018 01:57:42 -0400
Subject: IB/rxe: replace kvfree with vfree

The buf is allocated by vmalloc_user in the function rxe_queue_init.
So it is better to free it by vfree.

Fixes: 8700e3e7c485 ("Soft RoCE driver")
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Zhu Yanjun <yanjun.zhu@oracle.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
---
 drivers/infiniband/sw/rxe/rxe_cq.c | 4 ++--
 drivers/infiniband/sw/rxe/rxe_qp.c | 5 +++--
 2 files changed, 5 insertions(+), 4 deletions(-)

(limited to 'drivers/infiniband/sw')

diff --git a/drivers/infiniband/sw/rxe/rxe_cq.c b/drivers/infiniband/sw/rxe/rxe_cq.c
index 2ee4b08b00ea..a57276f2cb84 100644
--- a/drivers/infiniband/sw/rxe/rxe_cq.c
+++ b/drivers/infiniband/sw/rxe/rxe_cq.c
@@ -30,7 +30,7 @@
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
  */
-
+#include <linux/vmalloc.h>
 #include "rxe.h"
 #include "rxe_loc.h"
 #include "rxe_queue.h"
@@ -97,7 +97,7 @@ int rxe_cq_from_init(struct rxe_dev *rxe, struct rxe_cq *cq, int cqe,
 	err = do_mmap_info(rxe, uresp ? &uresp->mi : NULL, context,
 			   cq->queue->buf, cq->queue->buf_size, &cq->queue->ip);
 	if (err) {
-		kvfree(cq->queue->buf);
+		vfree(cq->queue->buf);
 		kfree(cq->queue);
 		return err;
 	}
diff --git a/drivers/infiniband/sw/rxe/rxe_qp.c b/drivers/infiniband/sw/rxe/rxe_qp.c
index 6ff88c8250f6..45b392b7342f 100644
--- a/drivers/infiniband/sw/rxe/rxe_qp.c
+++ b/drivers/infiniband/sw/rxe/rxe_qp.c
@@ -34,6 +34,7 @@
 #include <linux/skbuff.h>
 #include <linux/delay.h>
 #include <linux/sched.h>
+#include <linux/vmalloc.h>
 
 #include "rxe.h"
 #include "rxe_loc.h"
@@ -257,7 +258,7 @@ static int rxe_qp_init_req(struct rxe_dev *rxe, struct rxe_qp *qp,
 			   &qp->sq.queue->ip);
 
 	if (err) {
-		kvfree(qp->sq.queue->buf);
+		vfree(qp->sq.queue->buf);
 		kfree(qp->sq.queue);
 		return err;
 	}
@@ -310,7 +311,7 @@ static int rxe_qp_init_resp(struct rxe_dev *rxe, struct rxe_qp *qp,
 				   qp->rq.queue->buf, qp->rq.queue->buf_size,
 				   &qp->rq.queue->ip);
 		if (err) {
-			kvfree(qp->rq.queue->buf);
+			vfree(qp->rq.queue->buf);
 			kfree(qp->rq.queue);
 			return err;
 		}
-- 
cgit v1.2.3-55-g7522


From d205a06a14796a24b3447bc5d27b7dedff4479d5 Mon Sep 17 00:00:00 2001
From: Kaike Wan
Date: Wed, 26 Sep 2018 10:26:44 -0700
Subject: IB/rdmavt: Rename check_send_wqe as setup_wqe

The driver-provided function check_send_wqe allows the hardware driver to
check and set up the incoming send wqe before it is inserted into the swqe
ring. This patch will rename it as setup_wqe to better reflect its
usage. In addition, this function is only called when all setup is
complete in rdmavt.

Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Kaike Wan <kaike.wan@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
---
 drivers/infiniband/hw/hfi1/qp.c       | 11 ++++++++---
 drivers/infiniband/hw/hfi1/verbs.c    |  2 +-
 drivers/infiniband/hw/hfi1/verbs.h    |  4 ++--
 drivers/infiniband/hw/qib/qib_verbs.c |  2 +-
 drivers/infiniband/sw/rdmavt/qp.c     | 28 ++++++++++++++++++++--------
 include/rdma/rdma_vt.h                | 13 +++++++------
 6 files changed, 39 insertions(+), 21 deletions(-)

(limited to 'drivers/infiniband/sw')

diff --git a/drivers/infiniband/hw/hfi1/qp.c b/drivers/infiniband/hw/hfi1/qp.c
index 54d9ff171059..b1044a205ab6 100644
--- a/drivers/infiniband/hw/hfi1/qp.c
+++ b/drivers/infiniband/hw/hfi1/qp.c
@@ -282,16 +282,21 @@ void hfi1_modify_qp(struct rvt_qp *qp, struct ib_qp_attr *attr,
 }
 
 /**
- * hfi1_check_send_wqe - validate wqe
+ * hfi1_setup_wqe - set up the wqe
  * @qp - The qp
  * @wqe - The built wqe
  * @call_send - Determine if the send should be posted or scheduled.
  *
+ * Perform setup of the wqe.  This is called
+ * prior to inserting the wqe into the ring but after
+ * the wqe has been setup by RDMAVT. This function
+ * allows the driver the opportunity to perform
+ * validation and additional setup of the wqe.
+ *
  * Returns 0 on success, -EINVAL on failure
  *
  */
-int hfi1_check_send_wqe(struct rvt_qp *qp,
-			struct rvt_swqe *wqe, bool *call_send)
+int hfi1_setup_wqe(struct rvt_qp *qp, struct rvt_swqe *wqe, bool *call_send)
 {
 	struct hfi1_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
 	struct rvt_ah *ah;
diff --git a/drivers/infiniband/hw/hfi1/verbs.c b/drivers/infiniband/hw/hfi1/verbs.c
index 13374c727b14..bbee0cb77ff8 100644
--- a/drivers/infiniband/hw/hfi1/verbs.c
+++ b/drivers/infiniband/hw/hfi1/verbs.c
@@ -1937,7 +1937,7 @@ int hfi1_register_ib_device(struct hfi1_devdata *dd)
 	dd->verbs_dev.rdi.driver_f.check_modify_qp = hfi1_check_modify_qp;
 	dd->verbs_dev.rdi.driver_f.modify_qp = hfi1_modify_qp;
 	dd->verbs_dev.rdi.driver_f.notify_restart_rc = hfi1_restart_rc;
-	dd->verbs_dev.rdi.driver_f.check_send_wqe = hfi1_check_send_wqe;
+	dd->verbs_dev.rdi.driver_f.setup_wqe = hfi1_setup_wqe;
 	dd->verbs_dev.rdi.driver_f.comp_vect_cpu_lookup =
 						hfi1_comp_vect_mappings_lookup;
 
diff --git a/drivers/infiniband/hw/hfi1/verbs.h b/drivers/infiniband/hw/hfi1/verbs.h
index 269ec338581b..bc77ffec51ce 100644
--- a/drivers/infiniband/hw/hfi1/verbs.h
+++ b/drivers/infiniband/hw/hfi1/verbs.h
@@ -343,8 +343,8 @@ int hfi1_check_modify_qp(struct rvt_qp *qp, struct ib_qp_attr *attr,
 void hfi1_modify_qp(struct rvt_qp *qp, struct ib_qp_attr *attr,
 		    int attr_mask, struct ib_udata *udata);
 void hfi1_restart_rc(struct rvt_qp *qp, u32 psn, int wait);
-int hfi1_check_send_wqe(struct rvt_qp *qp, struct rvt_swqe *wqe,
-			bool *call_send);
+int hfi1_setup_wqe(struct rvt_qp *qp, struct rvt_swqe *wqe,
+		   bool *call_send);
 
 extern const u32 rc_only_opcode;
 extern const u32 uc_only_opcode;
diff --git a/drivers/infiniband/hw/qib/qib_verbs.c b/drivers/infiniband/hw/qib/qib_verbs.c
index 41babbc0db58..ad9093d33cb2 100644
--- a/drivers/infiniband/hw/qib/qib_verbs.c
+++ b/drivers/infiniband/hw/qib/qib_verbs.c
@@ -1588,7 +1588,7 @@ int qib_register_ib_device(struct qib_devdata *dd)
 	dd->verbs_dev.rdi.driver_f.port_callback = qib_create_port_files;
 	dd->verbs_dev.rdi.driver_f.get_pci_dev = qib_get_pci_dev;
 	dd->verbs_dev.rdi.driver_f.check_ah = qib_check_ah;
-	dd->verbs_dev.rdi.driver_f.check_send_wqe = qib_check_send_wqe;
+	dd->verbs_dev.rdi.driver_f.setup_wqe = qib_check_send_wqe;
 	dd->verbs_dev.rdi.driver_f.notify_new_ah = qib_notify_new_ah;
 	dd->verbs_dev.rdi.driver_f.alloc_qpn = qib_alloc_qpn;
 	dd->verbs_dev.rdi.driver_f.qp_priv_alloc = qib_qp_priv_alloc;
diff --git a/drivers/infiniband/sw/rdmavt/qp.c b/drivers/infiniband/sw/rdmavt/qp.c
index a9b7d7ff32ee..2db71e956d02 100644
--- a/drivers/infiniband/sw/rdmavt/qp.c
+++ b/drivers/infiniband/sw/rdmavt/qp.c
@@ -1823,13 +1823,11 @@ static int rvt_post_one_wr(struct rvt_qp *qp,
 		wqe->wr.num_sge = j;
 	}
 
-	/* general part of wqe valid - allow for driver checks */
-	if (rdi->driver_f.check_send_wqe) {
-		ret = rdi->driver_f.check_send_wqe(qp, wqe, call_send);
-		if (ret < 0)
-			goto bail_inval_free;
-	}
-
+	/*
+	 * Calculate and set SWQE PSN values prior to handing it off
+	 * to the driver's check routine. This give the driver the
+	 * opportunity to adjust PSN values based on internal checks.
+	 */
 	log_pmtu = qp->log_pmtu;
 	if (qp->ibqp.qp_type != IB_QPT_UC &&
 	    qp->ibqp.qp_type != IB_QPT_RC) {
@@ -1854,8 +1852,18 @@ static int rvt_post_one_wr(struct rvt_qp *qp,
 				(wqe->length ?
 					((wqe->length - 1) >> log_pmtu) :
 					0);
-		qp->s_next_psn = wqe->lpsn + 1;
 	}
+
+	/* general part of wqe valid - allow for driver checks */
+	if (rdi->driver_f.setup_wqe) {
+		ret = rdi->driver_f.setup_wqe(qp, wqe, call_send);
+		if (ret < 0)
+			goto bail_inval_free_ref;
+	}
+
+	if (!(rdi->post_parms[wr->opcode].flags & RVT_OPERATION_LOCAL))
+		qp->s_next_psn = wqe->lpsn + 1;
+
 	if (unlikely(reserved_op)) {
 		wqe->wr.send_flags |= RVT_SEND_RESERVE_USED;
 		rvt_qp_wqe_reserve(qp, wqe);
@@ -1869,6 +1877,10 @@ static int rvt_post_one_wr(struct rvt_qp *qp,
 
 	return 0;
 
+bail_inval_free_ref:
+	if (qp->ibqp.qp_type != IB_QPT_UC &&
+	    qp->ibqp.qp_type != IB_QPT_RC)
+		atomic_dec(&ibah_to_rvtah(ud_wr(wr)->ah)->refcount);
 bail_inval_free:
 	/* release mr holds */
 	while (j) {
diff --git a/include/rdma/rdma_vt.h b/include/rdma/rdma_vt.h
index cc08de3570b4..52907204afcd 100644
--- a/include/rdma/rdma_vt.h
+++ b/include/rdma/rdma_vt.h
@@ -215,13 +215,14 @@ struct rvt_driver_provided {
 	void (*schedule_send_no_lock)(struct rvt_qp *qp);
 
 	/*
-	 * Validate the wqe.  This needs to be done prior to inserting the
-	 * wqe into the ring, but after the wqe has been set up.  Allow for
-	 * driver specific work request checking by providing a callback.
-	 * call_send indicates if the wqe should be posted or scheduled.
+	 * Driver specific work request setup and checking.
+	 * This function is allowed to perform any setup, checks, or
+	 * adjustments required to the SWQE in order to be usable by
+	 * underlying protocols. This includes private data structure
+	 * allocations.
 	 */
-	int (*check_send_wqe)(struct rvt_qp *qp, struct rvt_swqe *wqe,
-			      bool *call_send);
+	int (*setup_wqe)(struct rvt_qp *qp, struct rvt_swqe *wqe,
+			 bool *call_send);
 
 	/*
 	 * Sometimes rdmavt needs to kick the driver's send progress. That is
-- 
cgit v1.2.3-55-g7522


From aae0484e15f062ad2c2502e68e15dfb8b8f84608 Mon Sep 17 00:00:00 2001
From: Zhu Yanjun
Date: Sun, 30 Sep 2018 02:27:16 -0400
Subject: IB/rxe: avoid srq memory leak

In rxe_queue_init, q and q->buf are allocated. In do_mmap_info, q->ip is
allocated. When error occurs, rxe_srq_from_init and the later error
handler do not free these allocated memories.  This will make memory leak.

Signed-off-by: Zhu Yanjun <yanjun.zhu@oracle.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
---
 drivers/infiniband/sw/rxe/rxe_srq.c | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)

(limited to 'drivers/infiniband/sw')

diff --git a/drivers/infiniband/sw/rxe/rxe_srq.c b/drivers/infiniband/sw/rxe/rxe_srq.c
index 0d6c04ba7fc3..c41a5fee81f7 100644
--- a/drivers/infiniband/sw/rxe/rxe_srq.c
+++ b/drivers/infiniband/sw/rxe/rxe_srq.c
@@ -31,6 +31,7 @@
  * SOFTWARE.
  */
 
+#include <linux/vmalloc.h>
 #include "rxe.h"
 #include "rxe_loc.h"
 #include "rxe_queue.h"
@@ -129,13 +130,18 @@ int rxe_srq_from_init(struct rxe_dev *rxe, struct rxe_srq *srq,
 
 	err = do_mmap_info(rxe, uresp ? &uresp->mi : NULL, context, q->buf,
 			   q->buf_size, &q->ip);
-	if (err)
+	if (err) {
+		vfree(q->buf);
+		kfree(q);
 		return err;
+	}
 
 	if (uresp) {
 		if (copy_to_user(&uresp->srq_num, &srq->srq_num,
-				 sizeof(uresp->srq_num)))
+				 sizeof(uresp->srq_num))) {
+			rxe_queue_cleanup(q);
 			return -EFAULT;
+		}
 	}
 
 	return 0;
-- 
cgit v1.2.3-55-g7522


From 03241627b222427917999c533f1a2f9d5beda2be Mon Sep 17 00:00:00 2001
From: Kamal Heib
Date: Tue, 2 Oct 2018 11:03:10 +0300
Subject: RDMA/rxe: Remove unused addr_same()

This function is not in use - delete it.

Signed-off-by: Kamal Heib <kamalheib1@gmail.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
---
 drivers/infiniband/sw/rxe/rxe_net.c | 5 -----
 1 file changed, 5 deletions(-)

(limited to 'drivers/infiniband/sw')

diff --git a/drivers/infiniband/sw/rxe/rxe_net.c b/drivers/infiniband/sw/rxe/rxe_net.c
index fb06f94f33d8..40e82e0f6c2d 100644
--- a/drivers/infiniband/sw/rxe/rxe_net.c
+++ b/drivers/infiniband/sw/rxe/rxe_net.c
@@ -491,11 +491,6 @@ void rxe_loopback(struct sk_buff *skb)
 	rxe_rcv(skb);
 }
 
-static inline int addr_same(struct rxe_dev *rxe, struct rxe_av *av)
-{
-	return rxe->port.port_guid == av->grh.dgid.global.interface_id;
-}
-
 struct sk_buff *rxe_init_packet(struct rxe_dev *rxe, struct rxe_av *av,
 				int paylen, struct rxe_pkt_info *pkt)
 {
-- 
cgit v1.2.3-55-g7522


From d31131bba5a1630304c55ea775c48cc84912ab59 Mon Sep 17 00:00:00 2001
From: Kamal Heib
Date: Tue, 2 Oct 2018 16:11:21 +0300
Subject: RDMA: Remove unused parameter from ib_modify_qp_is_ok()

The ll parameter is not used in ib_modify_qp_is_ok(), so remove it.

Signed-off-by: Kamal Heib <kamalheib1@gmail.com>
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
---
 drivers/infiniband/core/verbs.c              | 3 +--
 drivers/infiniband/hw/bnxt_re/ib_verbs.c     | 3 +--
 drivers/infiniband/hw/hns/hns_roce_qp.c      | 4 ++--
 drivers/infiniband/hw/mlx4/qp.c              | 8 +-------
 drivers/infiniband/hw/mlx5/qp.c              | 5 ++---
 drivers/infiniband/hw/mthca/mthca_qp.c       | 4 ++--
 drivers/infiniband/hw/ocrdma/ocrdma_verbs.c  | 3 +--
 drivers/infiniband/hw/qedr/verbs.c           | 3 +--
 drivers/infiniband/hw/vmw_pvrdma/pvrdma_qp.c | 2 +-
 drivers/infiniband/sw/rdmavt/qp.c            | 5 +----
 drivers/infiniband/sw/rxe/rxe_qp.c           | 3 +--
 include/rdma/ib_verbs.h                      | 4 +---
 12 files changed, 15 insertions(+), 32 deletions(-)

(limited to 'drivers/infiniband/sw')

diff --git a/drivers/infiniband/core/verbs.c b/drivers/infiniband/core/verbs.c
index ee5fc8408add..1e7ad5e0a46e 100644
--- a/drivers/infiniband/core/verbs.c
+++ b/drivers/infiniband/core/verbs.c
@@ -1509,8 +1509,7 @@ static const struct {
 };
 
 bool ib_modify_qp_is_ok(enum ib_qp_state cur_state, enum ib_qp_state next_state,
-			enum ib_qp_type type, enum ib_qp_attr_mask mask,
-			enum rdma_link_layer ll)
+			enum ib_qp_type type, enum ib_qp_attr_mask mask)
 {
 	enum ib_qp_attr_mask req_param, opt_param;
 
diff --git a/drivers/infiniband/hw/bnxt_re/ib_verbs.c b/drivers/infiniband/hw/bnxt_re/ib_verbs.c
index bc2b9e038439..9d7c48466f10 100644
--- a/drivers/infiniband/hw/bnxt_re/ib_verbs.c
+++ b/drivers/infiniband/hw/bnxt_re/ib_verbs.c
@@ -1598,8 +1598,7 @@ int bnxt_re_modify_qp(struct ib_qp *ib_qp, struct ib_qp_attr *qp_attr,
 		curr_qp_state = __to_ib_qp_state(qp->qplib_qp.cur_qp_state);
 		new_qp_state = qp_attr->qp_state;
 		if (!ib_modify_qp_is_ok(curr_qp_state, new_qp_state,
-					ib_qp->qp_type, qp_attr_mask,
-					IB_LINK_LAYER_ETHERNET)) {
+					ib_qp->qp_type, qp_attr_mask)) {
 			dev_err(rdev_to_dev(rdev),
 				"Invalid attribute mask: %#x specified ",
 				qp_attr_mask);
diff --git a/drivers/infiniband/hw/hns/hns_roce_qp.c b/drivers/infiniband/hw/hns/hns_roce_qp.c
index efb7e961ca65..0378fc41fcfa 100644
--- a/drivers/infiniband/hw/hns/hns_roce_qp.c
+++ b/drivers/infiniband/hw/hns/hns_roce_qp.c
@@ -952,8 +952,8 @@ int hns_roce_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
 		}
 	}
 
-	if (!ib_modify_qp_is_ok(cur_state, new_state, ibqp->qp_type, attr_mask,
-				IB_LINK_LAYER_ETHERNET)) {
+	if (!ib_modify_qp_is_ok(cur_state, new_state, ibqp->qp_type,
+				attr_mask)) {
 		dev_err(dev, "ib_modify_qp_is_ok failed\n");
 		goto out;
 	}
diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c
index 6dd3cd2c2f80..0711ca1dfb8f 100644
--- a/drivers/infiniband/hw/mlx4/qp.c
+++ b/drivers/infiniband/hw/mlx4/qp.c
@@ -2629,7 +2629,6 @@ enum {
 static int _mlx4_ib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
 			      int attr_mask, struct ib_udata *udata)
 {
-	enum rdma_link_layer ll = IB_LINK_LAYER_UNSPECIFIED;
 	struct mlx4_ib_dev *dev = to_mdev(ibqp->device);
 	struct mlx4_ib_qp *qp = to_mqp(ibqp);
 	enum ib_qp_state cur_state, new_state;
@@ -2639,13 +2638,8 @@ static int _mlx4_ib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
 	cur_state = attr_mask & IB_QP_CUR_STATE ? attr->cur_qp_state : qp->state;
 	new_state = attr_mask & IB_QP_STATE ? attr->qp_state : cur_state;
 
-	if (cur_state != new_state || cur_state != IB_QPS_RESET) {
-		int port = attr_mask & IB_QP_PORT ? attr->port_num : qp->port;
-		ll = rdma_port_get_link_layer(&dev->ib_dev, port);
-	}
-
 	if (!ib_modify_qp_is_ok(cur_state, new_state, ibqp->qp_type,
-				attr_mask, ll)) {
+				attr_mask)) {
 		pr_debug("qpn 0x%x: invalid attribute mask specified "
 			 "for transition %d to %d. qp_type %d,"
 			 " attr_mask 0x%x\n",
diff --git a/drivers/infiniband/hw/mlx5/qp.c b/drivers/infiniband/hw/mlx5/qp.c
index c49a0815a12b..fa8e5dc65cb4 100644
--- a/drivers/infiniband/hw/mlx5/qp.c
+++ b/drivers/infiniband/hw/mlx5/qp.c
@@ -3509,7 +3509,6 @@ int mlx5_ib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
 	size_t required_cmd_sz;
 	int err = -EINVAL;
 	int port;
-	enum rdma_link_layer ll = IB_LINK_LAYER_UNSPECIFIED;
 
 	if (ibqp->rwq_ind_tbl)
 		return -ENOSYS;
@@ -3555,7 +3554,6 @@ int mlx5_ib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
 
 	if (!(cur_state == new_state && cur_state == IB_QPS_RESET)) {
 		port = attr_mask & IB_QP_PORT ? attr->port_num : qp->port;
-		ll = dev->ib_dev.get_link_layer(&dev->ib_dev, port);
 	}
 
 	if (qp->flags & MLX5_IB_QP_UNDERLAY) {
@@ -3566,7 +3564,8 @@ int mlx5_ib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
 		}
 	} else if (qp_type != MLX5_IB_QPT_REG_UMR &&
 		   qp_type != MLX5_IB_QPT_DCI &&
-		   !ib_modify_qp_is_ok(cur_state, new_state, qp_type, attr_mask, ll)) {
+		   !ib_modify_qp_is_ok(cur_state, new_state, qp_type,
+				       attr_mask)) {
 		mlx5_ib_dbg(dev, "invalid QP state transition from %d to %d, qp_type %d, attr_mask 0x%x\n",
 			    cur_state, new_state, ibqp->qp_type, attr_mask);
 		goto out;
diff --git a/drivers/infiniband/hw/mthca/mthca_qp.c b/drivers/infiniband/hw/mthca/mthca_qp.c
index 3d37f2373d63..9d178ee3c96a 100644
--- a/drivers/infiniband/hw/mthca/mthca_qp.c
+++ b/drivers/infiniband/hw/mthca/mthca_qp.c
@@ -872,8 +872,8 @@ int mthca_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, int attr_mask,
 
 	new_state = attr_mask & IB_QP_STATE ? attr->qp_state : cur_state;
 
-	if (!ib_modify_qp_is_ok(cur_state, new_state, ibqp->qp_type, attr_mask,
-				IB_LINK_LAYER_UNSPECIFIED)) {
+	if (!ib_modify_qp_is_ok(cur_state, new_state, ibqp->qp_type,
+				attr_mask)) {
 		mthca_dbg(dev, "Bad QP transition (transport %d) "
 			  "%d->%d with attr 0x%08x\n",
 			  qp->transport, cur_state, new_state,
diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_verbs.c b/drivers/infiniband/hw/ocrdma/ocrdma_verbs.c
index c158ca9fde6d..06d2a7f3304c 100644
--- a/drivers/infiniband/hw/ocrdma/ocrdma_verbs.c
+++ b/drivers/infiniband/hw/ocrdma/ocrdma_verbs.c
@@ -1480,8 +1480,7 @@ int ocrdma_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
 		new_qps = old_qps;
 	spin_unlock_irqrestore(&qp->q_lock, flags);
 
-	if (!ib_modify_qp_is_ok(old_qps, new_qps, ibqp->qp_type, attr_mask,
-				IB_LINK_LAYER_ETHERNET)) {
+	if (!ib_modify_qp_is_ok(old_qps, new_qps, ibqp->qp_type, attr_mask)) {
 		pr_err("%s(%d) invalid attribute mask=0x%x specified for\n"
 		       "qpn=0x%x of type=0x%x old_qps=0x%x, new_qps=0x%x\n",
 		       __func__, dev->id, attr_mask, qp->id, ibqp->qp_type,
diff --git a/drivers/infiniband/hw/qedr/verbs.c b/drivers/infiniband/hw/qedr/verbs.c
index 9d4d165014d9..82ee4b4a7084 100644
--- a/drivers/infiniband/hw/qedr/verbs.c
+++ b/drivers/infiniband/hw/qedr/verbs.c
@@ -2238,8 +2238,7 @@ int qedr_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
 
 	if (rdma_protocol_roce(&dev->ibdev, 1)) {
 		if (!ib_modify_qp_is_ok(old_qp_state, new_qp_state,
-					ibqp->qp_type, attr_mask,
-					IB_LINK_LAYER_ETHERNET)) {
+					ibqp->qp_type, attr_mask)) {
 			DP_ERR(dev,
 			       "modify qp: invalid attribute mask=0x%x specified for\n"
 			       "qpn=0x%x of type=0x%x old_qp_state=0x%x, new_qp_state=0x%x\n",
diff --git a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_qp.c b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_qp.c
index 60083c0363a5..cf22f57a9f0d 100644
--- a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_qp.c
+++ b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_qp.c
@@ -499,7 +499,7 @@ int pvrdma_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
 	next_state = (attr_mask & IB_QP_STATE) ? attr->qp_state : cur_state;
 
 	if (!ib_modify_qp_is_ok(cur_state, next_state, ibqp->qp_type,
-				attr_mask, IB_LINK_LAYER_ETHERNET)) {
+				attr_mask)) {
 		ret = -EINVAL;
 		goto out;
 	}
diff --git a/drivers/infiniband/sw/rdmavt/qp.c b/drivers/infiniband/sw/rdmavt/qp.c
index 2db71e956d02..a036a5368103 100644
--- a/drivers/infiniband/sw/rdmavt/qp.c
+++ b/drivers/infiniband/sw/rdmavt/qp.c
@@ -1164,11 +1164,8 @@ int rvt_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
 	int lastwqe = 0;
 	int mig = 0;
 	int pmtu = 0; /* for gcc warning only */
-	enum rdma_link_layer link;
 	int opa_ah;
 
-	link = rdma_port_get_link_layer(ibqp->device, qp->port_num);
-
 	spin_lock_irq(&qp->r_lock);
 	spin_lock(&qp->s_hlock);
 	spin_lock(&qp->s_lock);
@@ -1179,7 +1176,7 @@ int rvt_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
 	opa_ah = rdma_cap_opa_ah(ibqp->device, qp->port_num);
 
 	if (!ib_modify_qp_is_ok(cur_state, new_state, ibqp->qp_type,
-				attr_mask, link))
+				attr_mask))
 		goto inval;
 
 	if (rdi->driver_f.check_modify_qp &&
diff --git a/drivers/infiniband/sw/rxe/rxe_qp.c b/drivers/infiniband/sw/rxe/rxe_qp.c
index 45b392b7342f..b9710907dac2 100644
--- a/drivers/infiniband/sw/rxe/rxe_qp.c
+++ b/drivers/infiniband/sw/rxe/rxe_qp.c
@@ -419,8 +419,7 @@ int rxe_qp_chk_attr(struct rxe_dev *rxe, struct rxe_qp *qp,
 	enum ib_qp_state new_state = (mask & IB_QP_STATE) ?
 					attr->qp_state : cur_state;
 
-	if (!ib_modify_qp_is_ok(cur_state, new_state, qp_type(qp), mask,
-				IB_LINK_LAYER_ETHERNET)) {
+	if (!ib_modify_qp_is_ok(cur_state, new_state, qp_type(qp), mask)) {
 		pr_warn("invalid mask or state for qp\n");
 		goto err1;
 	}
diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h
index 9897d2329f2c..f88c1071413a 100644
--- a/include/rdma/ib_verbs.h
+++ b/include/rdma/ib_verbs.h
@@ -2742,7 +2742,6 @@ static inline int ib_destroy_usecnt(atomic_t *usecnt,
  * @next_state: Next QP state
  * @type: QP type
  * @mask: Mask of supplied QP attributes
- * @ll : link layer of port
  *
  * This function is a helper function that a low-level driver's
  * modify_qp method can use to validate the consumer's input.  It
@@ -2751,8 +2750,7 @@ static inline int ib_destroy_usecnt(atomic_t *usecnt,
  * and that the attribute mask supplied is allowed for the transition.
  */
 bool ib_modify_qp_is_ok(enum ib_qp_state cur_state, enum ib_qp_state next_state,
-			enum ib_qp_type type, enum ib_qp_attr_mask mask,
-			enum rdma_link_layer ll);
+			enum ib_qp_type type, enum ib_qp_attr_mask mask);
 
 void ib_register_event_handler(struct ib_event_handler *event_handler);
 void ib_unregister_event_handler(struct ib_event_handler *event_handler);
-- 
cgit v1.2.3-55-g7522


From 019f118b94c895294debfaa394b267638fe2f648 Mon Sep 17 00:00:00 2001
From: Brian Welty
Date: Wed, 26 Sep 2018 10:44:33 -0700
Subject: IB/{hfi1, qib, rdmavt}: Move copy SGE logic into rdmavt

This patch moves hfi1_copy_sge() into rdmavt for sharing with qib.
This patch also moves all the wss_*() functions into rdmavt as
several wss_*() functions are called from hfi1_copy_sge()

When SGE copy mode is adaptive, cacheless copy may be done in some cases
for performance reasons. In those cases, X86 cacheless copy function
is called since the drivers that use rdmavt and may set SGE copy mode
to adaptive are X86 only. For this reason, this patch adds
"depends on X86_64" to rdmavt/Kconfig.

Reviewed-by: Ashutosh Dixit <ashutosh.dixit@intel.com>
Reviewed-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Brian Welty <brian.welty@intel.com>
Signed-off-by: Harish Chegondi <harish.chegondi@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
---
 drivers/infiniband/hw/hfi1/init.c     |   6 -
 drivers/infiniband/hw/hfi1/rc.c       |  10 +-
 drivers/infiniband/hw/hfi1/ruc.c      |   3 +-
 drivers/infiniband/hw/hfi1/uc.c       |  10 +-
 drivers/infiniband/hw/hfi1/ud.c       |  18 +--
 drivers/infiniband/hw/hfi1/verbs.c    | 226 +----------------------------
 drivers/infiniband/hw/hfi1/verbs.h    |  25 ----
 drivers/infiniband/hw/qib/qib_rc.c    |  10 +-
 drivers/infiniband/hw/qib/qib_ruc.c   |   2 +-
 drivers/infiniband/hw/qib/qib_uc.c    |  10 +-
 drivers/infiniband/hw/qib/qib_ud.c    |  13 +-
 drivers/infiniband/hw/qib/qib_verbs.c |  22 +--
 drivers/infiniband/hw/qib/qib_verbs.h |   3 -
 drivers/infiniband/sw/rdmavt/Kconfig  |   2 +-
 drivers/infiniband/sw/rdmavt/qp.c     | 258 ++++++++++++++++++++++++++++++++++
 drivers/infiniband/sw/rdmavt/qp.h     |   2 +
 drivers/infiniband/sw/rdmavt/vt.c     |  12 +-
 include/rdma/rdma_vt.h                |  22 +++
 include/rdma/rdmavt_qp.h              |   4 +
 19 files changed, 344 insertions(+), 314 deletions(-)

(limited to 'drivers/infiniband/sw')

diff --git a/drivers/infiniband/hw/hfi1/init.c b/drivers/infiniband/hw/hfi1/init.c
index 1e770a133779..09044905284f 100644
--- a/drivers/infiniband/hw/hfi1/init.c
+++ b/drivers/infiniband/hw/hfi1/init.c
@@ -1504,9 +1504,6 @@ static int __init hfi1_mod_init(void)
 	idr_init(&hfi1_unit_table);
 
 	hfi1_dbg_init();
-	ret = hfi1_wss_init();
-	if (ret < 0)
-		goto bail_wss;
 	ret = pci_register_driver(&hfi1_pci_driver);
 	if (ret < 0) {
 		pr_err("Unable to register driver: error %d\n", -ret);
@@ -1515,8 +1512,6 @@ static int __init hfi1_mod_init(void)
 	goto bail; /* all OK */
 
 bail_dev:
-	hfi1_wss_exit();
-bail_wss:
 	hfi1_dbg_exit();
 	idr_destroy(&hfi1_unit_table);
 	dev_cleanup();
@@ -1533,7 +1528,6 @@ static void __exit hfi1_mod_cleanup(void)
 {
 	pci_unregister_driver(&hfi1_pci_driver);
 	node_affinity_destroy_all();
-	hfi1_wss_exit();
 	hfi1_dbg_exit();
 
 	idr_destroy(&hfi1_unit_table);
diff --git a/drivers/infiniband/hw/hfi1/rc.c b/drivers/infiniband/hw/hfi1/rc.c
index 9bd63abb2dfe..673b31ebf0ac 100644
--- a/drivers/infiniband/hw/hfi1/rc.c
+++ b/drivers/infiniband/hw/hfi1/rc.c
@@ -1644,7 +1644,8 @@ read_middle:
 		qp->s_rdma_read_len -= pmtu;
 		update_last_psn(qp, psn);
 		spin_unlock_irqrestore(&qp->s_lock, flags);
-		hfi1_copy_sge(&qp->s_rdma_read_sge, data, pmtu, false, false);
+		rvt_copy_sge(qp, &qp->s_rdma_read_sge,
+			     data, pmtu, false, false);
 		goto bail;
 
 	case OP(RDMA_READ_RESPONSE_ONLY):
@@ -1684,7 +1685,8 @@ read_last:
 		if (unlikely(tlen != qp->s_rdma_read_len))
 			goto ack_len_err;
 		aeth = be32_to_cpu(ohdr->u.aeth);
-		hfi1_copy_sge(&qp->s_rdma_read_sge, data, tlen, false, false);
+		rvt_copy_sge(qp, &qp->s_rdma_read_sge,
+			     data, tlen, false, false);
 		WARN_ON(qp->s_rdma_read_sge.num_sge);
 		(void)do_rc_ack(qp, aeth, psn,
 				 OP(RDMA_READ_RESPONSE_LAST), 0, rcd);
@@ -2144,7 +2146,7 @@ send_middle:
 		qp->r_rcv_len += pmtu;
 		if (unlikely(qp->r_rcv_len > qp->r_len))
 			goto nack_inv;
-		hfi1_copy_sge(&qp->r_sge, data, pmtu, true, false);
+		rvt_copy_sge(qp, &qp->r_sge, data, pmtu, true, false);
 		break;
 
 	case OP(RDMA_WRITE_LAST_WITH_IMMEDIATE):
@@ -2200,7 +2202,7 @@ send_last:
 		wc.byte_len = tlen + qp->r_rcv_len;
 		if (unlikely(wc.byte_len > qp->r_len))
 			goto nack_inv;
-		hfi1_copy_sge(&qp->r_sge, data, tlen, true, copy_last);
+		rvt_copy_sge(qp, &qp->r_sge, data, tlen, true, copy_last);
 		rvt_put_ss(&qp->r_sge);
 		qp->r_msn++;
 		if (!__test_and_clear_bit(RVT_R_WRID_VALID, &qp->r_aflags))
diff --git a/drivers/infiniband/hw/hfi1/ruc.c b/drivers/infiniband/hw/hfi1/ruc.c
index 17b49b4309a3..223eaf184934 100644
--- a/drivers/infiniband/hw/hfi1/ruc.c
+++ b/drivers/infiniband/hw/hfi1/ruc.c
@@ -361,7 +361,8 @@ do_write:
 		if (len > sge->sge_length)
 			len = sge->sge_length;
 		WARN_ON_ONCE(len == 0);
-		hfi1_copy_sge(&qp->r_sge, sge->vaddr, len, release, copy_last);
+		rvt_copy_sge(qp, &qp->r_sge, sge->vaddr,
+			     len, release, copy_last);
 		sge->vaddr += len;
 		sge->length -= len;
 		sge->sge_length -= len;
diff --git a/drivers/infiniband/hw/hfi1/uc.c b/drivers/infiniband/hw/hfi1/uc.c
index e254dcec6f64..48a320c01552 100644
--- a/drivers/infiniband/hw/hfi1/uc.c
+++ b/drivers/infiniband/hw/hfi1/uc.c
@@ -426,7 +426,7 @@ send_first:
 		qp->r_rcv_len += pmtu;
 		if (unlikely(qp->r_rcv_len > qp->r_len))
 			goto rewind;
-		hfi1_copy_sge(&qp->r_sge, data, pmtu, false, false);
+		rvt_copy_sge(qp, &qp->r_sge, data, pmtu, false, false);
 		break;
 
 	case OP(SEND_LAST_WITH_IMMEDIATE):
@@ -449,7 +449,7 @@ send_last:
 		if (unlikely(wc.byte_len > qp->r_len))
 			goto rewind;
 		wc.opcode = IB_WC_RECV;
-		hfi1_copy_sge(&qp->r_sge, data, tlen, false, false);
+		rvt_copy_sge(qp, &qp->r_sge, data, tlen, false, false);
 		rvt_put_ss(&qp->s_rdma_read_sge);
 last_imm:
 		wc.wr_id = qp->r_wr_id;
@@ -523,7 +523,7 @@ rdma_first:
 		qp->r_rcv_len += pmtu;
 		if (unlikely(qp->r_rcv_len > qp->r_len))
 			goto drop;
-		hfi1_copy_sge(&qp->r_sge, data, pmtu, true, false);
+		rvt_copy_sge(qp, &qp->r_sge, data, pmtu, true, false);
 		break;
 
 	case OP(RDMA_WRITE_LAST_WITH_IMMEDIATE):
@@ -550,7 +550,7 @@ rdma_last_imm:
 		}
 		wc.byte_len = qp->r_len;
 		wc.opcode = IB_WC_RECV_RDMA_WITH_IMM;
-		hfi1_copy_sge(&qp->r_sge, data, tlen, true, false);
+		rvt_copy_sge(qp, &qp->r_sge, data, tlen, true, false);
 		rvt_put_ss(&qp->r_sge);
 		goto last_imm;
 
@@ -564,7 +564,7 @@ rdma_last:
 		tlen -= (hdrsize + extra_bytes);
 		if (unlikely(tlen + qp->r_rcv_len != qp->r_len))
 			goto drop;
-		hfi1_copy_sge(&qp->r_sge, data, tlen, true, false);
+		rvt_copy_sge(qp, &qp->r_sge, data, tlen, true, false);
 		rvt_put_ss(&qp->r_sge);
 		break;
 
diff --git a/drivers/infiniband/hw/hfi1/ud.c b/drivers/infiniband/hw/hfi1/ud.c
index 70d39fc450a1..e55bc4280d58 100644
--- a/drivers/infiniband/hw/hfi1/ud.c
+++ b/drivers/infiniband/hw/hfi1/ud.c
@@ -210,8 +210,8 @@ static void ud_loopback(struct rvt_qp *sqp, struct rvt_swqe *swqe)
 		}
 
 		hfi1_make_grh(ibp, &grh, &grd, 0, 0);
-		hfi1_copy_sge(&qp->r_sge, &grh,
-			      sizeof(grh), true, false);
+		rvt_copy_sge(qp, &qp->r_sge, &grh,
+			     sizeof(grh), true, false);
 		wc.wc_flags |= IB_WC_GRH;
 	} else {
 		rvt_skip_sge(&qp->r_sge, sizeof(struct ib_grh), true);
@@ -228,7 +228,7 @@ static void ud_loopback(struct rvt_qp *sqp, struct rvt_swqe *swqe)
 		if (len > sge->sge_length)
 			len = sge->sge_length;
 		WARN_ON_ONCE(len == 0);
-		hfi1_copy_sge(&qp->r_sge, sge->vaddr, len, true, false);
+		rvt_copy_sge(qp, &qp->r_sge, sge->vaddr, len, true, false);
 		sge->vaddr += len;
 		sge->length -= len;
 		sge->sge_length -= len;
@@ -1019,8 +1019,8 @@ void hfi1_ud_rcv(struct hfi1_packet *packet)
 		goto drop;
 	}
 	if (packet->grh) {
-		hfi1_copy_sge(&qp->r_sge, packet->grh,
-			      sizeof(struct ib_grh), true, false);
+		rvt_copy_sge(qp, &qp->r_sge, packet->grh,
+			     sizeof(struct ib_grh), true, false);
 		wc.wc_flags |= IB_WC_GRH;
 	} else if (packet->etype == RHF_RCV_TYPE_BYPASS) {
 		struct ib_grh grh;
@@ -1030,14 +1030,14 @@ void hfi1_ud_rcv(struct hfi1_packet *packet)
 		 * out when creating 16B, add back the GRH here.
 		 */
 		hfi1_make_ext_grh(packet, &grh, slid, dlid);
-		hfi1_copy_sge(&qp->r_sge, &grh,
-			      sizeof(struct ib_grh), true, false);
+		rvt_copy_sge(qp, &qp->r_sge, &grh,
+			     sizeof(struct ib_grh), true, false);
 		wc.wc_flags |= IB_WC_GRH;
 	} else {
 		rvt_skip_sge(&qp->r_sge, sizeof(struct ib_grh), true);
 	}
-	hfi1_copy_sge(&qp->r_sge, data, wc.byte_len - sizeof(struct ib_grh),
-		      true, false);
+	rvt_copy_sge(qp, &qp->r_sge, data, wc.byte_len - sizeof(struct ib_grh),
+		     true, false);
 	rvt_put_ss(&qp->r_sge);
 	if (!test_and_clear_bit(RVT_R_WRID_VALID, &qp->r_aflags))
 		return;
diff --git a/drivers/infiniband/hw/hfi1/verbs.c b/drivers/infiniband/hw/hfi1/verbs.c
index 16b88948383b..0a47b46f979e 100644
--- a/drivers/infiniband/hw/hfi1/verbs.c
+++ b/drivers/infiniband/hw/hfi1/verbs.c
@@ -129,8 +129,6 @@ unsigned short piothreshold = 256;
 module_param(piothreshold, ushort, S_IRUGO);
 MODULE_PARM_DESC(piothreshold, "size used to determine sdma vs. pio");
 
-#define COPY_CACHELESS 1
-#define COPY_ADAPTIVE  2
 static unsigned int sge_copy_mode;
 module_param(sge_copy_mode, uint, S_IRUGO);
 MODULE_PARM_DESC(sge_copy_mode,
@@ -151,159 +149,13 @@ static int pio_wait(struct rvt_qp *qp,
 /* 16B trailing buffer */
 static const u8 trail_buf[MAX_16B_PADDING];
 
-static uint wss_threshold;
+static uint wss_threshold = 80;
 module_param(wss_threshold, uint, S_IRUGO);
 MODULE_PARM_DESC(wss_threshold, "Percentage (1-100) of LLC to use as a threshold for a cacheless copy");
 static uint wss_clean_period = 256;
 module_param(wss_clean_period, uint, S_IRUGO);
 MODULE_PARM_DESC(wss_clean_period, "Count of verbs copies before an entry in the page copy table is cleaned");
 
-/* memory working set size */
-struct hfi1_wss {
-	unsigned long *entries;
-	atomic_t total_count;
-	atomic_t clean_counter;
-	atomic_t clean_entry;
-
-	int threshold;
-	int num_entries;
-	long pages_mask;
-};
-
-static struct hfi1_wss wss;
-
-int hfi1_wss_init(void)
-{
-	long llc_size;
-	long llc_bits;
-	long table_size;
-	long table_bits;
-
-	/* check for a valid percent range - default to 80 if none or invalid */
-	if (wss_threshold < 1 || wss_threshold > 100)
-		wss_threshold = 80;
-	/* reject a wildly large period */
-	if (wss_clean_period > 1000000)
-		wss_clean_period = 256;
-	/* reject a zero period */
-	if (wss_clean_period == 0)
-		wss_clean_period = 1;
-
-	/*
-	 * Calculate the table size - the next power of 2 larger than the
-	 * LLC size.  LLC size is in KiB.
-	 */
-	llc_size = wss_llc_size() * 1024;
-	table_size = roundup_pow_of_two(llc_size);
-
-	/* one bit per page in rounded up table */
-	llc_bits = llc_size / PAGE_SIZE;
-	table_bits = table_size / PAGE_SIZE;
-	wss.pages_mask = table_bits - 1;
-	wss.num_entries = table_bits / BITS_PER_LONG;
-
-	wss.threshold = (llc_bits * wss_threshold) / 100;
-	if (wss.threshold == 0)
-		wss.threshold = 1;
-
-	atomic_set(&wss.clean_counter, wss_clean_period);
-
-	wss.entries = kcalloc(wss.num_entries, sizeof(*wss.entries),
-			      GFP_KERNEL);
-	if (!wss.entries) {
-		hfi1_wss_exit();
-		return -ENOMEM;
-	}
-
-	return 0;
-}
-
-void hfi1_wss_exit(void)
-{
-	/* coded to handle partially initialized and repeat callers */
-	kfree(wss.entries);
-	wss.entries = NULL;
-}
-
-/*
- * Advance the clean counter.  When the clean period has expired,
- * clean an entry.
- *
- * This is implemented in atomics to avoid locking.  Because multiple
- * variables are involved, it can be racy which can lead to slightly
- * inaccurate information.  Since this is only a heuristic, this is
- * OK.  Any innaccuracies will clean themselves out as the counter
- * advances.  That said, it is unlikely the entry clean operation will
- * race - the next possible racer will not start until the next clean
- * period.
- *
- * The clean counter is implemented as a decrement to zero.  When zero
- * is reached an entry is cleaned.
- */
-static void wss_advance_clean_counter(void)
-{
-	int entry;
-	int weight;
-	unsigned long bits;
-
-	/* become the cleaner if we decrement the counter to zero */
-	if (atomic_dec_and_test(&wss.clean_counter)) {
-		/*
-		 * Set, not add, the clean period.  This avoids an issue
-		 * where the counter could decrement below the clean period.
-		 * Doing a set can result in lost decrements, slowing the
-		 * clean advance.  Since this a heuristic, this possible
-		 * slowdown is OK.
-		 *
-		 * An alternative is to loop, advancing the counter by a
-		 * clean period until the result is > 0. However, this could
-		 * lead to several threads keeping another in the clean loop.
-		 * This could be mitigated by limiting the number of times
-		 * we stay in the loop.
-		 */
-		atomic_set(&wss.clean_counter, wss_clean_period);
-
-		/*
-		 * Uniquely grab the entry to clean and move to next.
-		 * The current entry is always the lower bits of
-		 * wss.clean_entry.  The table size, wss.num_entries,
-		 * is always a power-of-2.
-		 */
-		entry = (atomic_inc_return(&wss.clean_entry) - 1)
-			& (wss.num_entries - 1);
-
-		/* clear the entry and count the bits */
-		bits = xchg(&wss.entries[entry], 0);
-		weight = hweight64((u64)bits);
-		/* only adjust the contended total count if needed */
-		if (weight)
-			atomic_sub(weight, &wss.total_count);
-	}
-}
-
-/*
- * Insert the given address into the working set array.
- */
-static void wss_insert(void *address)
-{
-	u32 page = ((unsigned long)address >> PAGE_SHIFT) & wss.pages_mask;
-	u32 entry = page / BITS_PER_LONG; /* assumes this ends up a shift */
-	u32 nr = page & (BITS_PER_LONG - 1);
-
-	if (!test_and_set_bit(nr, &wss.entries[entry]))
-		atomic_inc(&wss.total_count);
-
-	wss_advance_clean_counter();
-}
-
-/*
- * Is the working set larger than the threshold?
- */
-static inline bool wss_exceeds_threshold(void)
-{
-	return atomic_read(&wss.total_count) >= wss.threshold;
-}
-
 /*
  * Translate ib_wr_opcode into ib_wc_opcode.
  */
@@ -438,79 +290,6 @@ static const u32 pio_opmask[BIT(3)] = {
  */
 __be64 ib_hfi1_sys_image_guid;
 
-/**
- * hfi1_copy_sge - copy data to SGE memory
- * @ss: the SGE state
- * @data: the data to copy
- * @length: the length of the data
- * @release: boolean to release MR
- * @copy_last: do a separate copy of the last 8 bytes
- */
-void hfi1_copy_sge(
-	struct rvt_sge_state *ss,
-	void *data, u32 length,
-	bool release,
-	bool copy_last)
-{
-	struct rvt_sge *sge = &ss->sge;
-	int i;
-	bool in_last = false;
-	bool cacheless_copy = false;
-
-	if (sge_copy_mode == COPY_CACHELESS) {
-		cacheless_copy = length >= PAGE_SIZE;
-	} else if (sge_copy_mode == COPY_ADAPTIVE) {
-		if (length >= PAGE_SIZE) {
-			/*
-			 * NOTE: this *assumes*:
-			 * o The first vaddr is the dest.
-			 * o If multiple pages, then vaddr is sequential.
-			 */
-			wss_insert(sge->vaddr);
-			if (length >= (2 * PAGE_SIZE))
-				wss_insert(sge->vaddr + PAGE_SIZE);
-
-			cacheless_copy = wss_exceeds_threshold();
-		} else {
-			wss_advance_clean_counter();
-		}
-	}
-	if (copy_last) {
-		if (length > 8) {
-			length -= 8;
-		} else {
-			copy_last = false;
-			in_last = true;
-		}
-	}
-
-again:
-	while (length) {
-		u32 len = rvt_get_sge_length(sge, length);
-
-		WARN_ON_ONCE(len == 0);
-		if (unlikely(in_last)) {
-			/* enforce byte transfer ordering */
-			for (i = 0; i < len; i++)
-				((u8 *)sge->vaddr)[i] = ((u8 *)data)[i];
-		} else if (cacheless_copy) {
-			cacheless_memcpy(sge->vaddr, data, len);
-		} else {
-			memcpy(sge->vaddr, data, len);
-		}
-		rvt_update_sge(ss, len, release);
-		data += len;
-		length -= len;
-	}
-
-	if (copy_last) {
-		copy_last = false;
-		in_last = true;
-		length = 8;
-		goto again;
-	}
-}
-
 /*
  * Make sure the QP is ready and able to accept the given opcode.
  */
@@ -1949,6 +1728,9 @@ int hfi1_register_ib_device(struct hfi1_devdata *dd)
 	dd->verbs_dev.rdi.dparms.lkey_table_size = hfi1_lkey_table_size;
 	dd->verbs_dev.rdi.dparms.nports = dd->num_pports;
 	dd->verbs_dev.rdi.dparms.npkeys = hfi1_get_npkeys(dd);
+	dd->verbs_dev.rdi.dparms.sge_copy_mode = sge_copy_mode;
+	dd->verbs_dev.rdi.dparms.wss_threshold = wss_threshold;
+	dd->verbs_dev.rdi.dparms.wss_clean_period = wss_clean_period;
 
 	/* post send table */
 	dd->verbs_dev.rdi.post_parms = hfi1_post_parms;
diff --git a/drivers/infiniband/hw/hfi1/verbs.h b/drivers/infiniband/hw/hfi1/verbs.h
index d4164114396e..eb99e8df6251 100644
--- a/drivers/infiniband/hw/hfi1/verbs.h
+++ b/drivers/infiniband/hw/hfi1/verbs.h
@@ -315,9 +315,6 @@ void hfi1_put_txreq(struct verbs_txreq *tx);
 
 int hfi1_verbs_send(struct rvt_qp *qp, struct hfi1_pkt_state *ps);
 
-void hfi1_copy_sge(struct rvt_sge_state *ss, void *data, u32 length,
-		   bool release, bool copy_last);
-
 void hfi1_cnp_rcv(struct hfi1_packet *packet);
 
 void hfi1_uc_rcv(struct hfi1_packet *packet);
@@ -393,28 +390,6 @@ int hfi1_verbs_send_dma(struct rvt_qp *qp, struct hfi1_pkt_state *ps,
 int hfi1_verbs_send_pio(struct rvt_qp *qp, struct hfi1_pkt_state *ps,
 			u64 pbc);
 
-int hfi1_wss_init(void);
-void hfi1_wss_exit(void);
-
-/* platform specific: return the lowest level cache (llc) size, in KiB */
-static inline int wss_llc_size(void)
-{
-	/* assume that the boot CPU value is universal for all CPUs */
-	return boot_cpu_data.x86_cache_size;
-}
-
-/* platform specific: cacheless copy */
-static inline void cacheless_memcpy(void *dst, void *src, size_t n)
-{
-	/*
-	 * Use the only available X64 cacheless copy.  Add a __user cast
-	 * to quiet sparse.  The src agument is already in the kernel so
-	 * there are no security issues.  The extra fault recovery machinery
-	 * is not invoked.
-	 */
-	__copy_user_nocache(dst, (void __user *)src, n, 0);
-}
-
 static inline bool opa_bth_is_migration(struct ib_other_headers *ohdr)
 {
 	return ohdr->bth[1] & cpu_to_be32(OPA_BTH_MIG_REQ);
diff --git a/drivers/infiniband/hw/qib/qib_rc.c b/drivers/infiniband/hw/qib/qib_rc.c
index f35fdeb14347..034b9729f991 100644
--- a/drivers/infiniband/hw/qib/qib_rc.c
+++ b/drivers/infiniband/hw/qib/qib_rc.c
@@ -1425,7 +1425,8 @@ read_middle:
 		qp->s_rdma_read_len -= pmtu;
 		update_last_psn(qp, psn);
 		spin_unlock_irqrestore(&qp->s_lock, flags);
-		qib_copy_sge(&qp->s_rdma_read_sge, data, pmtu, 0);
+		rvt_copy_sge(qp, &qp->s_rdma_read_sge,
+			     data, pmtu, false, false);
 		goto bail;
 
 	case OP(RDMA_READ_RESPONSE_ONLY):
@@ -1471,7 +1472,8 @@ read_last:
 		if (unlikely(tlen != qp->s_rdma_read_len))
 			goto ack_len_err;
 		aeth = be32_to_cpu(ohdr->u.aeth);
-		qib_copy_sge(&qp->s_rdma_read_sge, data, tlen, 0);
+		rvt_copy_sge(qp, &qp->s_rdma_read_sge,
+			     data, tlen, false, false);
 		WARN_ON(qp->s_rdma_read_sge.num_sge);
 		(void) do_rc_ack(qp, aeth, psn,
 				 OP(RDMA_READ_RESPONSE_LAST), 0, rcd);
@@ -1844,7 +1846,7 @@ send_middle:
 		qp->r_rcv_len += pmtu;
 		if (unlikely(qp->r_rcv_len > qp->r_len))
 			goto nack_inv;
-		qib_copy_sge(&qp->r_sge, data, pmtu, 1);
+		rvt_copy_sge(qp, &qp->r_sge, data, pmtu, true, false);
 		break;
 
 	case OP(RDMA_WRITE_LAST_WITH_IMMEDIATE):
@@ -1890,7 +1892,7 @@ send_last:
 		wc.byte_len = tlen + qp->r_rcv_len;
 		if (unlikely(wc.byte_len > qp->r_len))
 			goto nack_inv;
-		qib_copy_sge(&qp->r_sge, data, tlen, 1);
+		rvt_copy_sge(qp, &qp->r_sge, data, tlen, true, false);
 		rvt_put_ss(&qp->r_sge);
 		qp->r_msn++;
 		if (!test_and_clear_bit(RVT_R_WRID_VALID, &qp->r_aflags))
diff --git a/drivers/infiniband/hw/qib/qib_ruc.c b/drivers/infiniband/hw/qib/qib_ruc.c
index f8a7de795beb..bc2a9e208d18 100644
--- a/drivers/infiniband/hw/qib/qib_ruc.c
+++ b/drivers/infiniband/hw/qib/qib_ruc.c
@@ -354,7 +354,7 @@ again:
 		if (len > sge->sge_length)
 			len = sge->sge_length;
 		BUG_ON(len == 0);
-		qib_copy_sge(&qp->r_sge, sge->vaddr, len, release);
+		rvt_copy_sge(qp, &qp->r_sge, sge->vaddr, len, release, false);
 		sge->vaddr += len;
 		sge->length -= len;
 		sge->sge_length -= len;
diff --git a/drivers/infiniband/hw/qib/qib_uc.c b/drivers/infiniband/hw/qib/qib_uc.c
index 3e54bc11e0ae..0a090569148c 100644
--- a/drivers/infiniband/hw/qib/qib_uc.c
+++ b/drivers/infiniband/hw/qib/qib_uc.c
@@ -359,7 +359,7 @@ send_first:
 		qp->r_rcv_len += pmtu;
 		if (unlikely(qp->r_rcv_len > qp->r_len))
 			goto rewind;
-		qib_copy_sge(&qp->r_sge, data, pmtu, 0);
+		rvt_copy_sge(qp, &qp->r_sge, data, pmtu, false, false);
 		break;
 
 	case OP(SEND_LAST_WITH_IMMEDIATE):
@@ -385,7 +385,7 @@ send_last:
 		if (unlikely(wc.byte_len > qp->r_len))
 			goto rewind;
 		wc.opcode = IB_WC_RECV;
-		qib_copy_sge(&qp->r_sge, data, tlen, 0);
+		rvt_copy_sge(qp, &qp->r_sge, data, tlen, false, false);
 		rvt_put_ss(&qp->s_rdma_read_sge);
 last_imm:
 		wc.wr_id = qp->r_wr_id;
@@ -449,7 +449,7 @@ rdma_first:
 		qp->r_rcv_len += pmtu;
 		if (unlikely(qp->r_rcv_len > qp->r_len))
 			goto drop;
-		qib_copy_sge(&qp->r_sge, data, pmtu, 1);
+		rvt_copy_sge(qp, &qp->r_sge, data, pmtu, true, false);
 		break;
 
 	case OP(RDMA_WRITE_LAST_WITH_IMMEDIATE):
@@ -479,7 +479,7 @@ rdma_last_imm:
 		}
 		wc.byte_len = qp->r_len;
 		wc.opcode = IB_WC_RECV_RDMA_WITH_IMM;
-		qib_copy_sge(&qp->r_sge, data, tlen, 1);
+		rvt_copy_sge(qp, &qp->r_sge, data, tlen, true, false);
 		rvt_put_ss(&qp->r_sge);
 		goto last_imm;
 
@@ -495,7 +495,7 @@ rdma_last:
 		tlen -= (hdrsize + pad + 4);
 		if (unlikely(tlen + qp->r_rcv_len != qp->r_len))
 			goto drop;
-		qib_copy_sge(&qp->r_sge, data, tlen, 1);
+		rvt_copy_sge(qp, &qp->r_sge, data, tlen, true, false);
 		rvt_put_ss(&qp->r_sge);
 		break;
 
diff --git a/drivers/infiniband/hw/qib/qib_ud.c b/drivers/infiniband/hw/qib/qib_ud.c
index f8d029a2390f..b12b9c3a6b5c 100644
--- a/drivers/infiniband/hw/qib/qib_ud.c
+++ b/drivers/infiniband/hw/qib/qib_ud.c
@@ -162,8 +162,8 @@ static void qib_ud_loopback(struct rvt_qp *sqp, struct rvt_swqe *swqe)
 		const struct ib_global_route *grd = rdma_ah_read_grh(ah_attr);
 
 		qib_make_grh(ibp, &grh, grd, 0, 0);
-		qib_copy_sge(&qp->r_sge, &grh,
-			     sizeof(grh), 1);
+		rvt_copy_sge(qp, &qp->r_sge, &grh,
+			     sizeof(grh), true, false);
 		wc.wc_flags |= IB_WC_GRH;
 	} else
 		rvt_skip_sge(&qp->r_sge, sizeof(struct ib_grh), true);
@@ -179,7 +179,7 @@ static void qib_ud_loopback(struct rvt_qp *sqp, struct rvt_swqe *swqe)
 		if (len > sge->sge_length)
 			len = sge->sge_length;
 		BUG_ON(len == 0);
-		qib_copy_sge(&qp->r_sge, sge->vaddr, len, 1);
+		rvt_copy_sge(qp, &qp->r_sge, sge->vaddr, len, true, false);
 		sge->vaddr += len;
 		sge->length -= len;
 		sge->sge_length -= len;
@@ -551,12 +551,13 @@ void qib_ud_rcv(struct qib_ibport *ibp, struct ib_header *hdr,
 		goto drop;
 	}
 	if (has_grh) {
-		qib_copy_sge(&qp->r_sge, &hdr->u.l.grh,
-			     sizeof(struct ib_grh), 1);
+		rvt_copy_sge(qp, &qp->r_sge, &hdr->u.l.grh,
+			     sizeof(struct ib_grh), true, false);
 		wc.wc_flags |= IB_WC_GRH;
 	} else
 		rvt_skip_sge(&qp->r_sge, sizeof(struct ib_grh), true);
-	qib_copy_sge(&qp->r_sge, data, wc.byte_len - sizeof(struct ib_grh), 1);
+	rvt_copy_sge(qp, &qp->r_sge, data, wc.byte_len - sizeof(struct ib_grh),
+		     true, false);
 	rvt_put_ss(&qp->r_sge);
 	if (!test_and_clear_bit(RVT_R_WRID_VALID, &qp->r_aflags))
 		return;
diff --git a/drivers/infiniband/hw/qib/qib_verbs.c b/drivers/infiniband/hw/qib/qib_verbs.c
index 26ab78e5aaa7..ae6d42cc9651 100644
--- a/drivers/infiniband/hw/qib/qib_verbs.c
+++ b/drivers/infiniband/hw/qib/qib_verbs.c
@@ -131,27 +131,6 @@ const enum ib_wc_opcode ib_qib_wc_opcode[] = {
  */
 __be64 ib_qib_sys_image_guid;
 
-/**
- * qib_copy_sge - copy data to SGE memory
- * @ss: the SGE state
- * @data: the data to copy
- * @length: the length of the data
- */
-void qib_copy_sge(struct rvt_sge_state *ss, void *data, u32 length, int release)
-{
-	struct rvt_sge *sge = &ss->sge;
-
-	while (length) {
-		u32 len = rvt_get_sge_length(sge, length);
-
-		WARN_ON_ONCE(len == 0);
-		memcpy(sge->vaddr, data, len);
-		rvt_update_sge(ss, len, release);
-		data += len;
-		length -= len;
-	}
-}
-
 /*
  * Count the number of DMA descriptors needed to send length bytes of data.
  * Don't modify the qib_sge_state to get the count.
@@ -1631,6 +1610,7 @@ int qib_register_ib_device(struct qib_devdata *dd)
 	dd->verbs_dev.rdi.dparms.node = dd->assigned_node_id;
 	dd->verbs_dev.rdi.dparms.core_cap_flags = RDMA_CORE_PORT_IBA_IB;
 	dd->verbs_dev.rdi.dparms.max_mad_size = IB_MGMT_MAD_SIZE;
+	dd->verbs_dev.rdi.dparms.sge_copy_mode = RVT_SGE_COPY_MEMCPY;
 
 	qib_fill_device_attr(dd);
 
diff --git a/drivers/infiniband/hw/qib/qib_verbs.h b/drivers/infiniband/hw/qib/qib_verbs.h
index df90a7a41534..0c5e623ec70c 100644
--- a/drivers/infiniband/hw/qib/qib_verbs.h
+++ b/drivers/infiniband/hw/qib/qib_verbs.h
@@ -292,9 +292,6 @@ void qib_put_txreq(struct qib_verbs_txreq *tx);
 int qib_verbs_send(struct rvt_qp *qp, struct ib_header *hdr,
 		   u32 hdrwords, struct rvt_sge_state *ss, u32 len);
 
-void qib_copy_sge(struct rvt_sge_state *ss, void *data, u32 length,
-		  int release);
-
 void qib_uc_rcv(struct qib_ibport *ibp, struct ib_header *hdr,
 		int has_grh, void *data, u32 tlen, struct rvt_qp *qp);
 
diff --git a/drivers/infiniband/sw/rdmavt/Kconfig b/drivers/infiniband/sw/rdmavt/Kconfig
index 98e798007f75..7df896a18d38 100644
--- a/drivers/infiniband/sw/rdmavt/Kconfig
+++ b/drivers/infiniband/sw/rdmavt/Kconfig
@@ -1,6 +1,6 @@
 config INFINIBAND_RDMAVT
 	tristate "RDMA verbs transport library"
-	depends on 64BIT && ARCH_DMA_ADDR_T_64BIT
+	depends on X86_64 && ARCH_DMA_ADDR_T_64BIT
 	depends on PCI
 	select DMA_VIRT_OPS
 	---help---
diff --git a/drivers/infiniband/sw/rdmavt/qp.c b/drivers/infiniband/sw/rdmavt/qp.c
index a036a5368103..d969b0803e6f 100644
--- a/drivers/infiniband/sw/rdmavt/qp.c
+++ b/drivers/infiniband/sw/rdmavt/qp.c
@@ -118,6 +118,187 @@ const int ib_rvt_state_ops[IB_QPS_ERR + 1] = {
 };
 EXPORT_SYMBOL(ib_rvt_state_ops);
 
+/* platform specific: return the last level cache (llc) size, in KiB */
+static int rvt_wss_llc_size(void)
+{
+	/* assume that the boot CPU value is universal for all CPUs */
+	return boot_cpu_data.x86_cache_size;
+}
+
+/* platform specific: cacheless copy */
+static void cacheless_memcpy(void *dst, void *src, size_t n)
+{
+	/*
+	 * Use the only available X64 cacheless copy.  Add a __user cast
+	 * to quiet sparse.  The src agument is already in the kernel so
+	 * there are no security issues.  The extra fault recovery machinery
+	 * is not invoked.
+	 */
+	__copy_user_nocache(dst, (void __user *)src, n, 0);
+}
+
+void rvt_wss_exit(struct rvt_dev_info *rdi)
+{
+	struct rvt_wss *wss = rdi->wss;
+
+	if (!wss)
+		return;
+
+	/* coded to handle partially initialized and repeat callers */
+	kfree(wss->entries);
+	wss->entries = NULL;
+	kfree(rdi->wss);
+	rdi->wss = NULL;
+}
+
+/**
+ * rvt_wss_init - Init wss data structures
+ *
+ * Return: 0 on success
+ */
+int rvt_wss_init(struct rvt_dev_info *rdi)
+{
+	unsigned int sge_copy_mode = rdi->dparms.sge_copy_mode;
+	unsigned int wss_threshold = rdi->dparms.wss_threshold;
+	unsigned int wss_clean_period = rdi->dparms.wss_clean_period;
+	long llc_size;
+	long llc_bits;
+	long table_size;
+	long table_bits;
+	struct rvt_wss *wss;
+	int node = rdi->dparms.node;
+
+	if (sge_copy_mode != RVT_SGE_COPY_ADAPTIVE) {
+		rdi->wss = NULL;
+		return 0;
+	}
+
+	rdi->wss = kzalloc_node(sizeof(*rdi->wss), GFP_KERNEL, node);
+	if (!rdi->wss)
+		return -ENOMEM;
+	wss = rdi->wss;
+
+	/* check for a valid percent range - default to 80 if none or invalid */
+	if (wss_threshold < 1 || wss_threshold > 100)
+		wss_threshold = 80;
+
+	/* reject a wildly large period */
+	if (wss_clean_period > 1000000)
+		wss_clean_period = 256;
+
+	/* reject a zero period */
+	if (wss_clean_period == 0)
+		wss_clean_period = 1;
+
+	/*
+	 * Calculate the table size - the next power of 2 larger than the
+	 * LLC size.  LLC size is in KiB.
+	 */
+	llc_size = rvt_wss_llc_size() * 1024;
+	table_size = roundup_pow_of_two(llc_size);
+
+	/* one bit per page in rounded up table */
+	llc_bits = llc_size / PAGE_SIZE;
+	table_bits = table_size / PAGE_SIZE;
+	wss->pages_mask = table_bits - 1;
+	wss->num_entries = table_bits / BITS_PER_LONG;
+
+	wss->threshold = (llc_bits * wss_threshold) / 100;
+	if (wss->threshold == 0)
+		wss->threshold = 1;
+
+	wss->clean_period = wss_clean_period;
+	atomic_set(&wss->clean_counter, wss_clean_period);
+
+	wss->entries = kcalloc_node(wss->num_entries, sizeof(*wss->entries),
+				    GFP_KERNEL, node);
+	if (!wss->entries) {
+		rvt_wss_exit(rdi);
+		return -ENOMEM;
+	}
+
+	return 0;
+}
+
+/*
+ * Advance the clean counter.  When the clean period has expired,
+ * clean an entry.
+ *
+ * This is implemented in atomics to avoid locking.  Because multiple
+ * variables are involved, it can be racy which can lead to slightly
+ * inaccurate information.  Since this is only a heuristic, this is
+ * OK.  Any innaccuracies will clean themselves out as the counter
+ * advances.  That said, it is unlikely the entry clean operation will
+ * race - the next possible racer will not start until the next clean
+ * period.
+ *
+ * The clean counter is implemented as a decrement to zero.  When zero
+ * is reached an entry is cleaned.
+ */
+static void wss_advance_clean_counter(struct rvt_wss *wss)
+{
+	int entry;
+	int weight;
+	unsigned long bits;
+
+	/* become the cleaner if we decrement the counter to zero */
+	if (atomic_dec_and_test(&wss->clean_counter)) {
+		/*
+		 * Set, not add, the clean period.  This avoids an issue
+		 * where the counter could decrement below the clean period.
+		 * Doing a set can result in lost decrements, slowing the
+		 * clean advance.  Since this a heuristic, this possible
+		 * slowdown is OK.
+		 *
+		 * An alternative is to loop, advancing the counter by a
+		 * clean period until the result is > 0. However, this could
+		 * lead to several threads keeping another in the clean loop.
+		 * This could be mitigated by limiting the number of times
+		 * we stay in the loop.
+		 */
+		atomic_set(&wss->clean_counter, wss->clean_period);
+
+		/*
+		 * Uniquely grab the entry to clean and move to next.
+		 * The current entry is always the lower bits of
+		 * wss.clean_entry.  The table size, wss.num_entries,
+		 * is always a power-of-2.
+		 */
+		entry = (atomic_inc_return(&wss->clean_entry) - 1)
+			& (wss->num_entries - 1);
+
+		/* clear the entry and count the bits */
+		bits = xchg(&wss->entries[entry], 0);
+		weight = hweight64((u64)bits);
+		/* only adjust the contended total count if needed */
+		if (weight)
+			atomic_sub(weight, &wss->total_count);
+	}
+}
+
+/*
+ * Insert the given address into the working set array.
+ */
+static void wss_insert(struct rvt_wss *wss, void *address)
+{
+	u32 page = ((unsigned long)address >> PAGE_SHIFT) & wss->pages_mask;
+	u32 entry = page / BITS_PER_LONG; /* assumes this ends up a shift */
+	u32 nr = page & (BITS_PER_LONG - 1);
+
+	if (!test_and_set_bit(nr, &wss->entries[entry]))
+		atomic_inc(&wss->total_count);
+
+	wss_advance_clean_counter(wss);
+}
+
+/*
+ * Is the working set larger than the threshold?
+ */
+static inline bool wss_exceeds_threshold(struct rvt_wss *wss)
+{
+	return atomic_read(&wss->total_count) >= wss->threshold;
+}
+
 static void get_map_page(struct rvt_qpn_table *qpt,
 			 struct rvt_qpn_map *map)
 {
@@ -2476,3 +2657,80 @@ void rvt_qp_iter(struct rvt_dev_info *rdi,
 	rcu_read_unlock();
 }
 EXPORT_SYMBOL(rvt_qp_iter);
+
+/**
+ * rvt_copy_sge - copy data to SGE memory
+ * @qp: associated QP
+ * @ss: the SGE state
+ * @data: the data to copy
+ * @length: the length of the data
+ * @release: boolean to release MR
+ * @copy_last: do a separate copy of the last 8 bytes
+ */
+void rvt_copy_sge(struct rvt_qp *qp, struct rvt_sge_state *ss,
+		  void *data, u32 length,
+		  bool release, bool copy_last)
+{
+	struct rvt_sge *sge = &ss->sge;
+	int i;
+	bool in_last = false;
+	bool cacheless_copy = false;
+	struct rvt_dev_info *rdi = ib_to_rvt(qp->ibqp.device);
+	struct rvt_wss *wss = rdi->wss;
+	unsigned int sge_copy_mode = rdi->dparms.sge_copy_mode;
+
+	if (sge_copy_mode == RVT_SGE_COPY_CACHELESS) {
+		cacheless_copy = length >= PAGE_SIZE;
+	} else if (sge_copy_mode == RVT_SGE_COPY_ADAPTIVE) {
+		if (length >= PAGE_SIZE) {
+			/*
+			 * NOTE: this *assumes*:
+			 * o The first vaddr is the dest.
+			 * o If multiple pages, then vaddr is sequential.
+			 */
+			wss_insert(wss, sge->vaddr);
+			if (length >= (2 * PAGE_SIZE))
+				wss_insert(wss, (sge->vaddr + PAGE_SIZE));
+
+			cacheless_copy = wss_exceeds_threshold(wss);
+		} else {
+			wss_advance_clean_counter(wss);
+		}
+	}
+
+	if (copy_last) {
+		if (length > 8) {
+			length -= 8;
+		} else {
+			copy_last = false;
+			in_last = true;
+		}
+	}
+
+again:
+	while (length) {
+		u32 len = rvt_get_sge_length(sge, length);
+
+		WARN_ON_ONCE(len == 0);
+		if (unlikely(in_last)) {
+			/* enforce byte transfer ordering */
+			for (i = 0; i < len; i++)
+				((u8 *)sge->vaddr)[i] = ((u8 *)data)[i];
+		} else if (cacheless_copy) {
+			cacheless_memcpy(sge->vaddr, data, len);
+		} else {
+			memcpy(sge->vaddr, data, len);
+		}
+		rvt_update_sge(ss, len, release);
+		data += len;
+		length -= len;
+	}
+
+	if (copy_last) {
+		copy_last = false;
+		in_last = true;
+		length = 8;
+		goto again;
+	}
+}
+EXPORT_SYMBOL(rvt_copy_sge);
diff --git a/drivers/infiniband/sw/rdmavt/qp.h b/drivers/infiniband/sw/rdmavt/qp.h
index 264811fdc530..6d883972e0b8 100644
--- a/drivers/infiniband/sw/rdmavt/qp.h
+++ b/drivers/infiniband/sw/rdmavt/qp.h
@@ -66,4 +66,6 @@ int rvt_post_send(struct ib_qp *ibqp, const struct ib_send_wr *wr,
 		  const struct ib_send_wr **bad_wr);
 int rvt_post_srq_recv(struct ib_srq *ibsrq, const struct ib_recv_wr *wr,
 		      const struct ib_recv_wr **bad_wr);
+int rvt_wss_init(struct rvt_dev_info *rdi);
+void rvt_wss_exit(struct rvt_dev_info *rdi);
 #endif          /* DEF_RVTQP_H */
diff --git a/drivers/infiniband/sw/rdmavt/vt.c b/drivers/infiniband/sw/rdmavt/vt.c
index e3249d46bcef..723d3daf2eba 100644
--- a/drivers/infiniband/sw/rdmavt/vt.c
+++ b/drivers/infiniband/sw/rdmavt/vt.c
@@ -774,6 +774,13 @@ int rvt_register_device(struct rvt_dev_info *rdi, u32 driver_id)
 		goto bail_no_mr;
 	}
 
+	/* Memory Working Set Size */
+	ret = rvt_wss_init(rdi);
+	if (ret) {
+		rvt_pr_err(rdi, "Error in WSS init.\n");
+		goto bail_mr;
+	}
+
 	/* Completion queues */
 	spin_lock_init(&rdi->n_cqs_lock);
 
@@ -832,7 +839,7 @@ int rvt_register_device(struct rvt_dev_info *rdi, u32 driver_id)
 				 rdi->driver_f.port_callback);
 	if (ret) {
 		rvt_pr_err(rdi, "Failed to register driver with ib core.\n");
-		goto bail_mr;
+		goto bail_wss;
 	}
 
 	rvt_create_mad_agents(rdi);
@@ -840,6 +847,8 @@ int rvt_register_device(struct rvt_dev_info *rdi, u32 driver_id)
 	rvt_pr_info(rdi, "Registration with rdmavt done.\n");
 	return ret;
 
+bail_wss:
+	rvt_wss_exit(rdi);
 bail_mr:
 	rvt_mr_exit(rdi);
 
@@ -863,6 +872,7 @@ void rvt_unregister_device(struct rvt_dev_info *rdi)
 	rvt_free_mad_agents(rdi);
 
 	ib_unregister_device(&rdi->ibdev);
+	rvt_wss_exit(rdi);
 	rvt_mr_exit(rdi);
 	rvt_qp_exit(rdi);
 }
diff --git a/include/rdma/rdma_vt.h b/include/rdma/rdma_vt.h
index 0a888a9ecc96..7fa2f2d46a3c 100644
--- a/include/rdma/rdma_vt.h
+++ b/include/rdma/rdma_vt.h
@@ -149,6 +149,10 @@ struct rvt_ibport {
 
 #define RVT_CQN_MAX 16 /* maximum length of cq name */
 
+#define RVT_SGE_COPY_MEMCPY	0
+#define RVT_SGE_COPY_CACHELESS	1
+#define RVT_SGE_COPY_ADAPTIVE	2
+
 /*
  * Things that are driver specific, module parameters in hfi1 and qib
  */
@@ -161,6 +165,9 @@ struct rvt_driver_params {
 	 */
 	unsigned int lkey_table_size;
 	unsigned int qp_table_size;
+	unsigned int sge_copy_mode;
+	unsigned int wss_threshold;
+	unsigned int wss_clean_period;
 	int qpn_start;
 	int qpn_inc;
 	int qpn_res_start;
@@ -193,6 +200,19 @@ struct rvt_ah {
 	u8 log_pmtu;
 };
 
+/* memory working set size */
+struct rvt_wss {
+	unsigned long *entries;
+	atomic_t total_count;
+	atomic_t clean_counter;
+	atomic_t clean_entry;
+
+	int threshold;
+	int num_entries;
+	long pages_mask;
+	unsigned int clean_period;
+};
+
 struct rvt_dev_info;
 struct rvt_swqe;
 struct rvt_driver_provided {
@@ -418,6 +438,8 @@ struct rvt_dev_info {
 	u32 n_mcast_grps_allocated; /* number of mcast groups allocated */
 	spinlock_t n_mcast_grps_lock;
 
+	/* Memory Working Set Size */
+	struct rvt_wss *wss;
 };
 
 /**
diff --git a/include/rdma/rdmavt_qp.h b/include/rdma/rdmavt_qp.h
index 927f6d5b6d0f..eaf2593ca822 100644
--- a/include/rdma/rdmavt_qp.h
+++ b/include/rdma/rdmavt_qp.h
@@ -678,6 +678,10 @@ void rvt_del_timers_sync(struct rvt_qp *qp);
 void rvt_stop_rc_timers(struct rvt_qp *qp);
 void rvt_add_retry_timer(struct rvt_qp *qp);
 
+void rvt_copy_sge(struct rvt_qp *qp, struct rvt_sge_state *ss,
+		  void *data, u32 length,
+		  bool release, bool copy_last);
+
 /**
  * struct rvt_qp_iter - the iterator for QPs
  * @qp - the current QP
-- 
cgit v1.2.3-55-g7522


From 116aa0330ec71b9872dfb3a3cc5202a72b5a1666 Mon Sep 17 00:00:00 2001
From: Venkata Sandeep Dhanalakota
Date: Wed, 26 Sep 2018 10:44:42 -0700
Subject: IB/{hfi1, qib, rdmavt}: Move send completion logic to rdmavt

Moving send completion code into rdmavt in order to have shared logic
between qib and hfi1 drivers.

Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Reviewed-by: Brian Welty <brian.welty@intel.com>
Signed-off-by: Venkata Sandeep Dhanalakota <venkata.s.dhanalakota@intel.com>
Signed-off-by: Harish Chegondi <harish.chegondi@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
---
 drivers/infiniband/hw/hfi1/rc.c         | 14 +++++-----
 drivers/infiniband/hw/hfi1/ruc.c        | 45 ++-------------------------------
 drivers/infiniband/hw/hfi1/uc.c         |  4 +--
 drivers/infiniband/hw/hfi1/ud.c         |  4 +--
 drivers/infiniband/hw/hfi1/verbs.c      |  9 ++++---
 drivers/infiniband/hw/hfi1/verbs.h      |  3 ---
 drivers/infiniband/hw/qib/qib_rc.c      |  8 +++---
 drivers/infiniband/hw/qib/qib_ruc.c     | 43 ++-----------------------------
 drivers/infiniband/hw/qib/qib_sdma.c    |  2 +-
 drivers/infiniband/hw/qib/qib_uc.c      |  2 +-
 drivers/infiniband/hw/qib/qib_ud.c      |  4 +--
 drivers/infiniband/hw/qib/qib_verbs.c   |  7 +++--
 drivers/infiniband/hw/qib/qib_verbs.h   |  3 ---
 drivers/infiniband/sw/rdmavt/qp.c       | 43 +++++++++++++++++++++++++++++++
 drivers/infiniband/sw/rdmavt/trace_tx.h | 42 ++++++++++++++++++++++++++++++
 include/rdma/rdma_vt.h                  |  3 +++
 include/rdma/rdmavt_qp.h                |  2 ++
 17 files changed, 124 insertions(+), 114 deletions(-)

(limited to 'drivers/infiniband/sw')

diff --git a/drivers/infiniband/hw/hfi1/rc.c b/drivers/infiniband/hw/hfi1/rc.c
index 673b31ebf0ac..188aa4f686a0 100644
--- a/drivers/infiniband/hw/hfi1/rc.c
+++ b/drivers/infiniband/hw/hfi1/rc.c
@@ -309,7 +309,7 @@ int hfi1_make_rc_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps)
 		}
 		clear_ahg(qp);
 		wqe = rvt_get_swqe_ptr(qp, qp->s_last);
-		hfi1_send_complete(qp, wqe, qp->s_last != qp->s_acked ?
+		rvt_send_complete(qp, wqe, qp->s_last != qp->s_acked ?
 			IB_WC_SUCCESS : IB_WC_WR_FLUSH_ERR);
 		/* will get called again */
 		goto done_free_tx;
@@ -378,9 +378,9 @@ int hfi1_make_rc_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps)
 						wqe->wr.ex.invalidate_rkey);
 					local_ops = 1;
 				}
-				hfi1_send_complete(qp, wqe,
-						   err ? IB_WC_LOC_PROT_ERR
-						       : IB_WC_SUCCESS);
+				rvt_send_complete(qp, wqe,
+						  err ? IB_WC_LOC_PROT_ERR
+						      : IB_WC_SUCCESS);
 				if (local_ops)
 					atomic_dec(&qp->local_ops_pending);
 				goto done_free_tx;
@@ -1043,7 +1043,7 @@ void hfi1_restart_rc(struct rvt_qp *qp, u32 psn, int wait)
 			hfi1_migrate_qp(qp);
 			qp->s_retry = qp->s_retry_cnt;
 		} else if (qp->s_last == qp->s_acked) {
-			hfi1_send_complete(qp, wqe, IB_WC_RETRY_EXC_ERR);
+			rvt_send_complete(qp, wqe, IB_WC_RETRY_EXC_ERR);
 			rvt_error_qp(qp, IB_WC_WR_FLUSH_ERR);
 			return;
 		} else { /* need to handle delayed completion */
@@ -1468,7 +1468,7 @@ static int do_rc_ack(struct rvt_qp *qp, u32 aeth, u32 psn, int opcode,
 			ibp->rvp.n_other_naks++;
 class_b:
 			if (qp->s_last == qp->s_acked) {
-				hfi1_send_complete(qp, wqe, status);
+				rvt_send_complete(qp, wqe, status);
 				rvt_error_qp(qp, IB_WC_WR_FLUSH_ERR);
 			}
 			break;
@@ -1706,7 +1706,7 @@ ack_len_err:
 	status = IB_WC_LOC_LEN_ERR;
 ack_err:
 	if (qp->s_last == qp->s_acked) {
-		hfi1_send_complete(qp, wqe, status);
+		rvt_send_complete(qp, wqe, status);
 		rvt_error_qp(qp, IB_WC_WR_FLUSH_ERR);
 	}
 ack_done:
diff --git a/drivers/infiniband/hw/hfi1/ruc.c b/drivers/infiniband/hw/hfi1/ruc.c
index 223eaf184934..db1d0d8a04a5 100644
--- a/drivers/infiniband/hw/hfi1/ruc.c
+++ b/drivers/infiniband/hw/hfi1/ruc.c
@@ -411,7 +411,7 @@ send_comp:
 	ibp->rvp.n_loop_pkts++;
 flush_send:
 	sqp->s_rnr_retry = sqp->s_rnr_retry_cnt;
-	hfi1_send_complete(sqp, wqe, send_status);
+	rvt_send_complete(sqp, wqe, send_status);
 	if (local_ops) {
 		atomic_dec(&sqp->local_ops_pending);
 		local_ops = 0;
@@ -459,7 +459,7 @@ err:
 
 serr:
 	spin_lock_irqsave(&sqp->s_lock, flags);
-	hfi1_send_complete(sqp, wqe, send_status);
+	rvt_send_complete(sqp, wqe, send_status);
 	if (sqp->ibqp.qp_type == IB_QPT_RC) {
 		int lastwqe = rvt_error_qp(sqp, IB_WC_WR_FLUSH_ERR);
 
@@ -922,44 +922,3 @@ void hfi1_do_send(struct rvt_qp *qp, bool in_thread)
 	iowait_starve_clear(ps.pkts_sent, &priv->s_iowait);
 	spin_unlock_irqrestore(&qp->s_lock, ps.flags);
 }
-
-/*
- * This should be called with s_lock held.
- */
-void hfi1_send_complete(struct rvt_qp *qp, struct rvt_swqe *wqe,
-			enum ib_wc_status status)
-{
-	u32 old_last, last;
-
-	if (!(ib_rvt_state_ops[qp->state] & RVT_PROCESS_OR_FLUSH_SEND))
-		return;
-
-	last = qp->s_last;
-	old_last = last;
-	trace_hfi1_qp_send_completion(qp, wqe, last);
-	if (++last >= qp->s_size)
-		last = 0;
-	trace_hfi1_qp_send_completion(qp, wqe, last);
-	qp->s_last = last;
-	/* See post_send() */
-	barrier();
-	rvt_put_swqe(wqe);
-	if (qp->ibqp.qp_type == IB_QPT_UD ||
-	    qp->ibqp.qp_type == IB_QPT_SMI ||
-	    qp->ibqp.qp_type == IB_QPT_GSI)
-		atomic_dec(&ibah_to_rvtah(wqe->ud_wr.ah)->refcount);
-
-	rvt_qp_swqe_complete(qp,
-			     wqe,
-			     ib_hfi1_wc_opcode[wqe->wr.opcode],
-			     status);
-
-	if (qp->s_acked == old_last)
-		qp->s_acked = last;
-	if (qp->s_cur == old_last)
-		qp->s_cur = last;
-	if (qp->s_tail == old_last)
-		qp->s_tail = last;
-	if (qp->state == IB_QPS_SQD && last == qp->s_cur)
-		qp->s_draining = 0;
-}
diff --git a/drivers/infiniband/hw/hfi1/uc.c b/drivers/infiniband/hw/hfi1/uc.c
index 48a320c01552..6aca0c5a7f97 100644
--- a/drivers/infiniband/hw/hfi1/uc.c
+++ b/drivers/infiniband/hw/hfi1/uc.c
@@ -88,7 +88,7 @@ int hfi1_make_uc_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps)
 		}
 		clear_ahg(qp);
 		wqe = rvt_get_swqe_ptr(qp, qp->s_last);
-		hfi1_send_complete(qp, wqe, IB_WC_WR_FLUSH_ERR);
+		rvt_send_complete(qp, wqe, IB_WC_WR_FLUSH_ERR);
 		goto done_free_tx;
 	}
 
@@ -140,7 +140,7 @@ int hfi1_make_uc_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps)
 					qp, wqe->wr.ex.invalidate_rkey);
 				local_ops = 1;
 			}
-			hfi1_send_complete(qp, wqe, err ? IB_WC_LOC_PROT_ERR
+			rvt_send_complete(qp, wqe, err ? IB_WC_LOC_PROT_ERR
 							: IB_WC_SUCCESS);
 			if (local_ops)
 				atomic_dec(&qp->local_ops_pending);
diff --git a/drivers/infiniband/hw/hfi1/ud.c b/drivers/infiniband/hw/hfi1/ud.c
index e55bc4280d58..4baa8f4d49de 100644
--- a/drivers/infiniband/hw/hfi1/ud.c
+++ b/drivers/infiniband/hw/hfi1/ud.c
@@ -518,7 +518,7 @@ int hfi1_make_ud_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps)
 			goto bail;
 		}
 		wqe = rvt_get_swqe_ptr(qp, qp->s_last);
-		hfi1_send_complete(qp, wqe, IB_WC_WR_FLUSH_ERR);
+		rvt_send_complete(qp, wqe, IB_WC_WR_FLUSH_ERR);
 		goto done_free_tx;
 	}
 
@@ -560,7 +560,7 @@ int hfi1_make_ud_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps)
 			ud_loopback(qp, wqe);
 			spin_lock_irqsave(&qp->s_lock, tflags);
 			ps->flags = tflags;
-			hfi1_send_complete(qp, wqe, IB_WC_SUCCESS);
+			rvt_send_complete(qp, wqe, IB_WC_SUCCESS);
 			goto done_free_tx;
 		}
 	}
diff --git a/drivers/infiniband/hw/hfi1/verbs.c b/drivers/infiniband/hw/hfi1/verbs.c
index 0a47b46f979e..bc7f00ba1988 100644
--- a/drivers/infiniband/hw/hfi1/verbs.c
+++ b/drivers/infiniband/hw/hfi1/verbs.c
@@ -492,7 +492,7 @@ static void verbs_sdma_complete(
 
 	spin_lock(&qp->s_lock);
 	if (tx->wqe) {
-		hfi1_send_complete(qp, tx->wqe, IB_WC_SUCCESS);
+		rvt_send_complete(qp, tx->wqe, IB_WC_SUCCESS);
 	} else if (qp->ibqp.qp_type == IB_QPT_RC) {
 		struct hfi1_opa_header *hdr;
 
@@ -938,7 +938,7 @@ int hfi1_verbs_send_pio(struct rvt_qp *qp, struct hfi1_pkt_state *ps,
 pio_bail:
 	if (qp->s_wqe) {
 		spin_lock_irqsave(&qp->s_lock, flags);
-		hfi1_send_complete(qp, qp->s_wqe, wc_status);
+		rvt_send_complete(qp, qp->s_wqe, wc_status);
 		spin_unlock_irqrestore(&qp->s_lock, flags);
 	} else if (qp->ibqp.qp_type == IB_QPT_RC) {
 		spin_lock_irqsave(&qp->s_lock, flags);
@@ -1145,7 +1145,7 @@ int hfi1_verbs_send(struct rvt_qp *qp, struct hfi1_pkt_state *ps)
 			hfi1_cdbg(PIO, "%s() Failed. Completing with err",
 				  __func__);
 			spin_lock_irqsave(&qp->s_lock, flags);
-			hfi1_send_complete(qp, qp->s_wqe, IB_WC_GENERAL_ERR);
+			rvt_send_complete(qp, qp->s_wqe, IB_WC_GENERAL_ERR);
 			spin_unlock_irqrestore(&qp->s_lock, flags);
 		}
 		return -EINVAL;
@@ -1735,6 +1735,9 @@ int hfi1_register_ib_device(struct hfi1_devdata *dd)
 	/* post send table */
 	dd->verbs_dev.rdi.post_parms = hfi1_post_parms;
 
+	/* opcode translation table */
+	dd->verbs_dev.rdi.wc_opcode = ib_hfi1_wc_opcode;
+
 	ppd = dd->pport;
 	for (i = 0; i < dd->num_pports; i++, ppd++)
 		rvt_init_port(&dd->verbs_dev.rdi,
diff --git a/drivers/infiniband/hw/hfi1/verbs.h b/drivers/infiniband/hw/hfi1/verbs.h
index eb99e8df6251..64c9054db5f3 100644
--- a/drivers/infiniband/hw/hfi1/verbs.h
+++ b/drivers/infiniband/hw/hfi1/verbs.h
@@ -363,9 +363,6 @@ void hfi1_do_send_from_rvt(struct rvt_qp *qp);
 
 void hfi1_do_send(struct rvt_qp *qp, bool in_thread);
 
-void hfi1_send_complete(struct rvt_qp *qp, struct rvt_swqe *wqe,
-			enum ib_wc_status status);
-
 void hfi1_send_rc_ack(struct hfi1_packet *packet, bool is_fecn);
 
 int hfi1_make_rc_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps);
diff --git a/drivers/infiniband/hw/qib/qib_rc.c b/drivers/infiniband/hw/qib/qib_rc.c
index 034b9729f991..6fa002940451 100644
--- a/drivers/infiniband/hw/qib/qib_rc.c
+++ b/drivers/infiniband/hw/qib/qib_rc.c
@@ -254,7 +254,7 @@ int qib_make_rc_req(struct rvt_qp *qp, unsigned long *flags)
 			goto bail;
 		}
 		wqe = rvt_get_swqe_ptr(qp, qp->s_last);
-		qib_send_complete(qp, wqe, qp->s_last != qp->s_acked ?
+		rvt_send_complete(qp, wqe, qp->s_last != qp->s_acked ?
 			IB_WC_SUCCESS : IB_WC_WR_FLUSH_ERR);
 		/* will get called again */
 		goto done;
@@ -838,7 +838,7 @@ void qib_restart_rc(struct rvt_qp *qp, u32 psn, int wait)
 			qib_migrate_qp(qp);
 			qp->s_retry = qp->s_retry_cnt;
 		} else if (qp->s_last == qp->s_acked) {
-			qib_send_complete(qp, wqe, IB_WC_RETRY_EXC_ERR);
+			rvt_send_complete(qp, wqe, IB_WC_RETRY_EXC_ERR);
 			rvt_error_qp(qp, IB_WC_WR_FLUSH_ERR);
 			return;
 		} else /* XXX need to handle delayed completion */
@@ -1221,7 +1221,7 @@ static int do_rc_ack(struct rvt_qp *qp, u32 aeth, u32 psn, int opcode,
 			ibp->rvp.n_other_naks++;
 class_b:
 			if (qp->s_last == qp->s_acked) {
-				qib_send_complete(qp, wqe, status);
+				rvt_send_complete(qp, wqe, status);
 				rvt_error_qp(qp, IB_WC_WR_FLUSH_ERR);
 			}
 			break;
@@ -1492,7 +1492,7 @@ ack_len_err:
 	status = IB_WC_LOC_LEN_ERR;
 ack_err:
 	if (qp->s_last == qp->s_acked) {
-		qib_send_complete(qp, wqe, status);
+		rvt_send_complete(qp, wqe, status);
 		rvt_error_qp(qp, IB_WC_WR_FLUSH_ERR);
 	}
 ack_done:
diff --git a/drivers/infiniband/hw/qib/qib_ruc.c b/drivers/infiniband/hw/qib/qib_ruc.c
index bc2a9e208d18..c5627baf5dbf 100644
--- a/drivers/infiniband/hw/qib/qib_ruc.c
+++ b/drivers/infiniband/hw/qib/qib_ruc.c
@@ -403,7 +403,7 @@ send_comp:
 	ibp->rvp.n_loop_pkts++;
 flush_send:
 	sqp->s_rnr_retry = sqp->s_rnr_retry_cnt;
-	qib_send_complete(sqp, wqe, send_status);
+	rvt_send_complete(sqp, wqe, send_status);
 	goto again;
 
 rnr_nak:
@@ -447,7 +447,7 @@ err:
 
 serr:
 	spin_lock_irqsave(&sqp->s_lock, flags);
-	qib_send_complete(sqp, wqe, send_status);
+	rvt_send_complete(sqp, wqe, send_status);
 	if (sqp->ibqp.qp_type == IB_QPT_RC) {
 		int lastwqe = rvt_error_qp(sqp, IB_WC_WR_FLUSH_ERR);
 
@@ -613,42 +613,3 @@ void qib_do_send(struct rvt_qp *qp)
 
 	spin_unlock_irqrestore(&qp->s_lock, flags);
 }
-
-/*
- * This should be called with s_lock held.
- */
-void qib_send_complete(struct rvt_qp *qp, struct rvt_swqe *wqe,
-		       enum ib_wc_status status)
-{
-	u32 old_last, last;
-
-	if (!(ib_rvt_state_ops[qp->state] & RVT_PROCESS_OR_FLUSH_SEND))
-		return;
-
-	last = qp->s_last;
-	old_last = last;
-	if (++last >= qp->s_size)
-		last = 0;
-	qp->s_last = last;
-	/* See post_send() */
-	barrier();
-	rvt_put_swqe(wqe);
-	if (qp->ibqp.qp_type == IB_QPT_UD ||
-	    qp->ibqp.qp_type == IB_QPT_SMI ||
-	    qp->ibqp.qp_type == IB_QPT_GSI)
-		atomic_dec(&ibah_to_rvtah(wqe->ud_wr.ah)->refcount);
-
-	rvt_qp_swqe_complete(qp,
-			     wqe,
-			     ib_qib_wc_opcode[wqe->wr.opcode],
-			     status);
-
-	if (qp->s_acked == old_last)
-		qp->s_acked = last;
-	if (qp->s_cur == old_last)
-		qp->s_cur = last;
-	if (qp->s_tail == old_last)
-		qp->s_tail = last;
-	if (qp->state == IB_QPS_SQD && last == qp->s_cur)
-		qp->s_draining = 0;
-}
diff --git a/drivers/infiniband/hw/qib/qib_sdma.c b/drivers/infiniband/hw/qib/qib_sdma.c
index d0723d4aef5c..757d4c9d713d 100644
--- a/drivers/infiniband/hw/qib/qib_sdma.c
+++ b/drivers/infiniband/hw/qib/qib_sdma.c
@@ -651,7 +651,7 @@ unmap:
 		if (ib_rvt_state_ops[qp->state] & RVT_PROCESS_RECV_OK)
 			rvt_error_qp(qp, IB_WC_GENERAL_ERR);
 	} else if (qp->s_wqe)
-		qib_send_complete(qp, qp->s_wqe, IB_WC_GENERAL_ERR);
+		rvt_send_complete(qp, qp->s_wqe, IB_WC_GENERAL_ERR);
 	spin_unlock(&qp->s_lock);
 	spin_unlock(&qp->r_lock);
 	/* return zero to process the next send work request */
diff --git a/drivers/infiniband/hw/qib/qib_uc.c b/drivers/infiniband/hw/qib/qib_uc.c
index 0a090569148c..30c70ad0f4bf 100644
--- a/drivers/infiniband/hw/qib/qib_uc.c
+++ b/drivers/infiniband/hw/qib/qib_uc.c
@@ -68,7 +68,7 @@ int qib_make_uc_req(struct rvt_qp *qp, unsigned long *flags)
 			goto bail;
 		}
 		wqe = rvt_get_swqe_ptr(qp, qp->s_last);
-		qib_send_complete(qp, wqe, IB_WC_WR_FLUSH_ERR);
+		rvt_send_complete(qp, wqe, IB_WC_WR_FLUSH_ERR);
 		goto done;
 	}
 
diff --git a/drivers/infiniband/hw/qib/qib_ud.c b/drivers/infiniband/hw/qib/qib_ud.c
index b12b9c3a6b5c..4d4c31ea4e2d 100644
--- a/drivers/infiniband/hw/qib/qib_ud.c
+++ b/drivers/infiniband/hw/qib/qib_ud.c
@@ -260,7 +260,7 @@ int qib_make_ud_req(struct rvt_qp *qp, unsigned long *flags)
 			goto bail;
 		}
 		wqe = rvt_get_swqe_ptr(qp, qp->s_last);
-		qib_send_complete(qp, wqe, IB_WC_WR_FLUSH_ERR);
+		rvt_send_complete(qp, wqe, IB_WC_WR_FLUSH_ERR);
 		goto done;
 	}
 
@@ -304,7 +304,7 @@ int qib_make_ud_req(struct rvt_qp *qp, unsigned long *flags)
 			qib_ud_loopback(qp, wqe);
 			spin_lock_irqsave(&qp->s_lock, tflags);
 			*flags = tflags;
-			qib_send_complete(qp, wqe, IB_WC_SUCCESS);
+			rvt_send_complete(qp, wqe, IB_WC_SUCCESS);
 			goto done;
 		}
 	}
diff --git a/drivers/infiniband/hw/qib/qib_verbs.c b/drivers/infiniband/hw/qib/qib_verbs.c
index ae6d42cc9651..8a45964c4700 100644
--- a/drivers/infiniband/hw/qib/qib_verbs.c
+++ b/drivers/infiniband/hw/qib/qib_verbs.c
@@ -731,7 +731,7 @@ static void sdma_complete(struct qib_sdma_txreq *cookie, int status)
 
 	spin_lock(&qp->s_lock);
 	if (tx->wqe)
-		qib_send_complete(qp, tx->wqe, IB_WC_SUCCESS);
+		rvt_send_complete(qp, tx->wqe, IB_WC_SUCCESS);
 	else if (qp->ibqp.qp_type == IB_QPT_RC) {
 		struct ib_header *hdr;
 
@@ -1004,7 +1004,7 @@ done:
 	}
 	if (qp->s_wqe) {
 		spin_lock_irqsave(&qp->s_lock, flags);
-		qib_send_complete(qp, qp->s_wqe, IB_WC_SUCCESS);
+		rvt_send_complete(qp, qp->s_wqe, IB_WC_SUCCESS);
 		spin_unlock_irqrestore(&qp->s_lock, flags);
 	} else if (qp->ibqp.qp_type == IB_QPT_RC) {
 		spin_lock_irqsave(&qp->s_lock, flags);
@@ -1491,6 +1491,9 @@ static void qib_fill_device_attr(struct qib_devdata *dd)
 					rdi->dparms.props.max_mcast_grp;
 	/* post send table */
 	dd->verbs_dev.rdi.post_parms = qib_post_parms;
+
+	/* opcode translation table */
+	dd->verbs_dev.rdi.wc_opcode = ib_qib_wc_opcode;
 }
 
 /**
diff --git a/drivers/infiniband/hw/qib/qib_verbs.h b/drivers/infiniband/hw/qib/qib_verbs.h
index 0c5e623ec70c..a4426c24b0d1 100644
--- a/drivers/infiniband/hw/qib/qib_verbs.h
+++ b/drivers/infiniband/hw/qib/qib_verbs.h
@@ -331,9 +331,6 @@ void _qib_do_send(struct work_struct *work);
 
 void qib_do_send(struct rvt_qp *qp);
 
-void qib_send_complete(struct rvt_qp *qp, struct rvt_swqe *wqe,
-		       enum ib_wc_status status);
-
 void qib_send_rc_ack(struct rvt_qp *qp);
 
 int qib_make_rc_req(struct rvt_qp *qp, unsigned long *flags);
diff --git a/drivers/infiniband/sw/rdmavt/qp.c b/drivers/infiniband/sw/rdmavt/qp.c
index d969b0803e6f..7e3ec6674cf7 100644
--- a/drivers/infiniband/sw/rdmavt/qp.c
+++ b/drivers/infiniband/sw/rdmavt/qp.c
@@ -2658,6 +2658,49 @@ void rvt_qp_iter(struct rvt_dev_info *rdi,
 }
 EXPORT_SYMBOL(rvt_qp_iter);
 
+/*
+ * This should be called with s_lock held.
+ */
+void rvt_send_complete(struct rvt_qp *qp, struct rvt_swqe *wqe,
+		       enum ib_wc_status status)
+{
+	u32 old_last, last;
+	struct rvt_dev_info *rdi = ib_to_rvt(qp->ibqp.device);
+
+	if (!(ib_rvt_state_ops[qp->state] & RVT_PROCESS_OR_FLUSH_SEND))
+		return;
+
+	last = qp->s_last;
+	old_last = last;
+	trace_rvt_qp_send_completion(qp, wqe, last);
+	if (++last >= qp->s_size)
+		last = 0;
+	trace_rvt_qp_send_completion(qp, wqe, last);
+	qp->s_last = last;
+	/* See post_send() */
+	barrier();
+	rvt_put_swqe(wqe);
+	if (qp->ibqp.qp_type == IB_QPT_UD ||
+	    qp->ibqp.qp_type == IB_QPT_SMI ||
+	    qp->ibqp.qp_type == IB_QPT_GSI)
+		atomic_dec(&ibah_to_rvtah(wqe->ud_wr.ah)->refcount);
+
+	rvt_qp_swqe_complete(qp,
+			     wqe,
+			     rdi->wc_opcode[wqe->wr.opcode],
+			     status);
+
+	if (qp->s_acked == old_last)
+		qp->s_acked = last;
+	if (qp->s_cur == old_last)
+		qp->s_cur = last;
+	if (qp->s_tail == old_last)
+		qp->s_tail = last;
+	if (qp->state == IB_QPS_SQD && last == qp->s_cur)
+		qp->s_draining = 0;
+}
+EXPORT_SYMBOL(rvt_send_complete);
+
 /**
  * rvt_copy_sge - copy data to SGE memory
  * @qp: associated QP
diff --git a/drivers/infiniband/sw/rdmavt/trace_tx.h b/drivers/infiniband/sw/rdmavt/trace_tx.h
index 0ef25fc49f25..d5df352eadb1 100644
--- a/drivers/infiniband/sw/rdmavt/trace_tx.h
+++ b/drivers/infiniband/sw/rdmavt/trace_tx.h
@@ -153,6 +153,48 @@ TRACE_EVENT(
 	)
 );
 
+TRACE_EVENT(
+	rvt_qp_send_completion,
+	TP_PROTO(struct rvt_qp *qp, struct rvt_swqe *wqe, u32 idx),
+	TP_ARGS(qp, wqe, idx),
+	TP_STRUCT__entry(
+		RDI_DEV_ENTRY(ib_to_rvt(qp->ibqp.device))
+		__field(struct rvt_swqe *, wqe)
+		__field(u64, wr_id)
+		__field(u32, qpn)
+		__field(u32, qpt)
+		__field(u32, length)
+		__field(u32, idx)
+		__field(u32, ssn)
+		__field(enum ib_wr_opcode, opcode)
+		__field(int, send_flags)
+	),
+	TP_fast_assign(
+		RDI_DEV_ASSIGN(ib_to_rvt(qp->ibqp.device))
+		__entry->wqe = wqe;
+		__entry->wr_id = wqe->wr.wr_id;
+		__entry->qpn = qp->ibqp.qp_num;
+		__entry->qpt = qp->ibqp.qp_type;
+		__entry->length = wqe->length;
+		__entry->idx = idx;
+		__entry->ssn = wqe->ssn;
+		__entry->opcode = wqe->wr.opcode;
+		__entry->send_flags = wqe->wr.send_flags;
+	),
+	TP_printk(
+		"[%s] qpn 0x%x qpt %u wqe %p idx %u wr_id %llx length %u ssn %u opcode %x send_flags %x",
+		__get_str(dev),
+		__entry->qpn,
+		__entry->qpt,
+		__entry->wqe,
+		__entry->idx,
+		__entry->wr_id,
+		__entry->length,
+		__entry->ssn,
+		__entry->opcode,
+		__entry->send_flags
+	)
+);
 #endif /* __RVT_TRACE_TX_H */
 
 #undef TRACE_INCLUDE_PATH
diff --git a/include/rdma/rdma_vt.h b/include/rdma/rdma_vt.h
index 7fa2f2d46a3c..3584d0816fcd 100644
--- a/include/rdma/rdma_vt.h
+++ b/include/rdma/rdma_vt.h
@@ -398,6 +398,9 @@ struct rvt_dev_info {
 	/* post send table */
 	const struct rvt_operation_params *post_parms;
 
+	/* opcode translation table */
+	const enum ib_wc_opcode *wc_opcode;
+
 	/* Driver specific helper functions */
 	struct rvt_driver_provided driver_f;
 
diff --git a/include/rdma/rdmavt_qp.h b/include/rdma/rdmavt_qp.h
index eaf2593ca822..6fd6f2ad9c0f 100644
--- a/include/rdma/rdmavt_qp.h
+++ b/include/rdma/rdmavt_qp.h
@@ -681,6 +681,8 @@ void rvt_add_retry_timer(struct rvt_qp *qp);
 void rvt_copy_sge(struct rvt_qp *qp, struct rvt_sge_state *ss,
 		  void *data, u32 length,
 		  bool release, bool copy_last);
+void rvt_send_complete(struct rvt_qp *qp, struct rvt_swqe *wqe,
+		       enum ib_wc_status status);
 
 /**
  * struct rvt_qp_iter - the iterator for QPs
-- 
cgit v1.2.3-55-g7522


From 15703461533a5ffd775722390431625daaae7618 Mon Sep 17 00:00:00 2001
From: Venkata Sandeep Dhanalakota
Date: Wed, 26 Sep 2018 10:44:52 -0700
Subject: IB/{hfi1, qib, rdmavt}: Move ruc_loopback to rdmavt

This patch moves ruc_loopback() from hfi1 into rdmavt for code sharing
with the qib driver.

Reviewed-by: Brian Welty <brian.welty@intel.com>
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Venkata Sandeep Dhanalakota <venkata.s.dhanalakota@intel.com>
Signed-off-by: Harish Chegondi <harish.chegondi@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
---
 drivers/infiniband/hw/hfi1/ruc.c    | 332 +-----------------------------------
 drivers/infiniband/hw/qib/qib_ruc.c | 303 +-------------------------------
 drivers/infiniband/sw/rdmavt/qp.c   | 331 +++++++++++++++++++++++++++++++++++
 include/rdma/rdmavt_qp.h            |   1 +
 4 files changed, 335 insertions(+), 632 deletions(-)

(limited to 'drivers/infiniband/sw')

diff --git a/drivers/infiniband/hw/hfi1/ruc.c b/drivers/infiniband/hw/hfi1/ruc.c
index db1d0d8a04a5..7fb317c711df 100644
--- a/drivers/infiniband/hw/hfi1/ruc.c
+++ b/drivers/infiniband/hw/hfi1/ruc.c
@@ -155,334 +155,6 @@ int hfi1_ruc_check_hdr(struct hfi1_ibport *ibp, struct hfi1_packet *packet)
 	return 0;
 }
 
-/**
- * ruc_loopback - handle UC and RC loopback requests
- * @sqp: the sending QP
- *
- * This is called from hfi1_do_send() to
- * forward a WQE addressed to the same HFI.
- * Note that although we are single threaded due to the send engine, we still
- * have to protect against post_send().  We don't have to worry about
- * receive interrupts since this is a connected protocol and all packets
- * will pass through here.
- */
-static void ruc_loopback(struct rvt_qp *sqp)
-{
-	struct hfi1_ibport *ibp = to_iport(sqp->ibqp.device, sqp->port_num);
-	struct rvt_qp *qp;
-	struct rvt_swqe *wqe;
-	struct rvt_sge *sge;
-	unsigned long flags;
-	struct ib_wc wc;
-	u64 sdata;
-	atomic64_t *maddr;
-	enum ib_wc_status send_status;
-	bool release;
-	int ret;
-	bool copy_last = false;
-	int local_ops = 0;
-
-	rcu_read_lock();
-
-	/*
-	 * Note that we check the responder QP state after
-	 * checking the requester's state.
-	 */
-	qp = rvt_lookup_qpn(ib_to_rvt(sqp->ibqp.device), &ibp->rvp,
-			    sqp->remote_qpn);
-
-	spin_lock_irqsave(&sqp->s_lock, flags);
-
-	/* Return if we are already busy processing a work request. */
-	if ((sqp->s_flags & (RVT_S_BUSY | HFI1_S_ANY_WAIT)) ||
-	    !(ib_rvt_state_ops[sqp->state] & RVT_PROCESS_OR_FLUSH_SEND))
-		goto unlock;
-
-	sqp->s_flags |= RVT_S_BUSY;
-
-again:
-	if (sqp->s_last == READ_ONCE(sqp->s_head))
-		goto clr_busy;
-	wqe = rvt_get_swqe_ptr(sqp, sqp->s_last);
-
-	/* Return if it is not OK to start a new work request. */
-	if (!(ib_rvt_state_ops[sqp->state] & RVT_PROCESS_NEXT_SEND_OK)) {
-		if (!(ib_rvt_state_ops[sqp->state] & RVT_FLUSH_SEND))
-			goto clr_busy;
-		/* We are in the error state, flush the work request. */
-		send_status = IB_WC_WR_FLUSH_ERR;
-		goto flush_send;
-	}
-
-	/*
-	 * We can rely on the entry not changing without the s_lock
-	 * being held until we update s_last.
-	 * We increment s_cur to indicate s_last is in progress.
-	 */
-	if (sqp->s_last == sqp->s_cur) {
-		if (++sqp->s_cur >= sqp->s_size)
-			sqp->s_cur = 0;
-	}
-	spin_unlock_irqrestore(&sqp->s_lock, flags);
-
-	if (!qp || !(ib_rvt_state_ops[qp->state] & RVT_PROCESS_RECV_OK) ||
-	    qp->ibqp.qp_type != sqp->ibqp.qp_type) {
-		ibp->rvp.n_pkt_drops++;
-		/*
-		 * For RC, the requester would timeout and retry so
-		 * shortcut the timeouts and just signal too many retries.
-		 */
-		if (sqp->ibqp.qp_type == IB_QPT_RC)
-			send_status = IB_WC_RETRY_EXC_ERR;
-		else
-			send_status = IB_WC_SUCCESS;
-		goto serr;
-	}
-
-	memset(&wc, 0, sizeof(wc));
-	send_status = IB_WC_SUCCESS;
-
-	release = true;
-	sqp->s_sge.sge = wqe->sg_list[0];
-	sqp->s_sge.sg_list = wqe->sg_list + 1;
-	sqp->s_sge.num_sge = wqe->wr.num_sge;
-	sqp->s_len = wqe->length;
-	switch (wqe->wr.opcode) {
-	case IB_WR_REG_MR:
-		goto send_comp;
-
-	case IB_WR_LOCAL_INV:
-		if (!(wqe->wr.send_flags & RVT_SEND_COMPLETION_ONLY)) {
-			if (rvt_invalidate_rkey(sqp,
-						wqe->wr.ex.invalidate_rkey))
-				send_status = IB_WC_LOC_PROT_ERR;
-			local_ops = 1;
-		}
-		goto send_comp;
-
-	case IB_WR_SEND_WITH_INV:
-		if (!rvt_invalidate_rkey(qp, wqe->wr.ex.invalidate_rkey)) {
-			wc.wc_flags = IB_WC_WITH_INVALIDATE;
-			wc.ex.invalidate_rkey = wqe->wr.ex.invalidate_rkey;
-		}
-		goto send;
-
-	case IB_WR_SEND_WITH_IMM:
-		wc.wc_flags = IB_WC_WITH_IMM;
-		wc.ex.imm_data = wqe->wr.ex.imm_data;
-		/* FALLTHROUGH */
-	case IB_WR_SEND:
-send:
-		ret = rvt_get_rwqe(qp, false);
-		if (ret < 0)
-			goto op_err;
-		if (!ret)
-			goto rnr_nak;
-		break;
-
-	case IB_WR_RDMA_WRITE_WITH_IMM:
-		if (unlikely(!(qp->qp_access_flags & IB_ACCESS_REMOTE_WRITE)))
-			goto inv_err;
-		wc.wc_flags = IB_WC_WITH_IMM;
-		wc.ex.imm_data = wqe->wr.ex.imm_data;
-		ret = rvt_get_rwqe(qp, true);
-		if (ret < 0)
-			goto op_err;
-		if (!ret)
-			goto rnr_nak;
-		/* skip copy_last set and qp_access_flags recheck */
-		goto do_write;
-	case IB_WR_RDMA_WRITE:
-		copy_last = rvt_is_user_qp(qp);
-		if (unlikely(!(qp->qp_access_flags & IB_ACCESS_REMOTE_WRITE)))
-			goto inv_err;
-do_write:
-		if (wqe->length == 0)
-			break;
-		if (unlikely(!rvt_rkey_ok(qp, &qp->r_sge.sge, wqe->length,
-					  wqe->rdma_wr.remote_addr,
-					  wqe->rdma_wr.rkey,
-					  IB_ACCESS_REMOTE_WRITE)))
-			goto acc_err;
-		qp->r_sge.sg_list = NULL;
-		qp->r_sge.num_sge = 1;
-		qp->r_sge.total_len = wqe->length;
-		break;
-
-	case IB_WR_RDMA_READ:
-		if (unlikely(!(qp->qp_access_flags & IB_ACCESS_REMOTE_READ)))
-			goto inv_err;
-		if (unlikely(!rvt_rkey_ok(qp, &sqp->s_sge.sge, wqe->length,
-					  wqe->rdma_wr.remote_addr,
-					  wqe->rdma_wr.rkey,
-					  IB_ACCESS_REMOTE_READ)))
-			goto acc_err;
-		release = false;
-		sqp->s_sge.sg_list = NULL;
-		sqp->s_sge.num_sge = 1;
-		qp->r_sge.sge = wqe->sg_list[0];
-		qp->r_sge.sg_list = wqe->sg_list + 1;
-		qp->r_sge.num_sge = wqe->wr.num_sge;
-		qp->r_sge.total_len = wqe->length;
-		break;
-
-	case IB_WR_ATOMIC_CMP_AND_SWP:
-	case IB_WR_ATOMIC_FETCH_AND_ADD:
-		if (unlikely(!(qp->qp_access_flags & IB_ACCESS_REMOTE_ATOMIC)))
-			goto inv_err;
-		if (unlikely(!rvt_rkey_ok(qp, &qp->r_sge.sge, sizeof(u64),
-					  wqe->atomic_wr.remote_addr,
-					  wqe->atomic_wr.rkey,
-					  IB_ACCESS_REMOTE_ATOMIC)))
-			goto acc_err;
-		/* Perform atomic OP and save result. */
-		maddr = (atomic64_t *)qp->r_sge.sge.vaddr;
-		sdata = wqe->atomic_wr.compare_add;
-		*(u64 *)sqp->s_sge.sge.vaddr =
-			(wqe->wr.opcode == IB_WR_ATOMIC_FETCH_AND_ADD) ?
-			(u64)atomic64_add_return(sdata, maddr) - sdata :
-			(u64)cmpxchg((u64 *)qp->r_sge.sge.vaddr,
-				      sdata, wqe->atomic_wr.swap);
-		rvt_put_mr(qp->r_sge.sge.mr);
-		qp->r_sge.num_sge = 0;
-		goto send_comp;
-
-	default:
-		send_status = IB_WC_LOC_QP_OP_ERR;
-		goto serr;
-	}
-
-	sge = &sqp->s_sge.sge;
-	while (sqp->s_len) {
-		u32 len = sqp->s_len;
-
-		if (len > sge->length)
-			len = sge->length;
-		if (len > sge->sge_length)
-			len = sge->sge_length;
-		WARN_ON_ONCE(len == 0);
-		rvt_copy_sge(qp, &qp->r_sge, sge->vaddr,
-			     len, release, copy_last);
-		sge->vaddr += len;
-		sge->length -= len;
-		sge->sge_length -= len;
-		if (sge->sge_length == 0) {
-			if (!release)
-				rvt_put_mr(sge->mr);
-			if (--sqp->s_sge.num_sge)
-				*sge = *sqp->s_sge.sg_list++;
-		} else if (sge->length == 0 && sge->mr->lkey) {
-			if (++sge->n >= RVT_SEGSZ) {
-				if (++sge->m >= sge->mr->mapsz)
-					break;
-				sge->n = 0;
-			}
-			sge->vaddr =
-				sge->mr->map[sge->m]->segs[sge->n].vaddr;
-			sge->length =
-				sge->mr->map[sge->m]->segs[sge->n].length;
-		}
-		sqp->s_len -= len;
-	}
-	if (release)
-		rvt_put_ss(&qp->r_sge);
-
-	if (!test_and_clear_bit(RVT_R_WRID_VALID, &qp->r_aflags))
-		goto send_comp;
-
-	if (wqe->wr.opcode == IB_WR_RDMA_WRITE_WITH_IMM)
-		wc.opcode = IB_WC_RECV_RDMA_WITH_IMM;
-	else
-		wc.opcode = IB_WC_RECV;
-	wc.wr_id = qp->r_wr_id;
-	wc.status = IB_WC_SUCCESS;
-	wc.byte_len = wqe->length;
-	wc.qp = &qp->ibqp;
-	wc.src_qp = qp->remote_qpn;
-	wc.slid = rdma_ah_get_dlid(&qp->remote_ah_attr) & U16_MAX;
-	wc.sl = rdma_ah_get_sl(&qp->remote_ah_attr);
-	wc.port_num = 1;
-	/* Signal completion event if the solicited bit is set. */
-	rvt_cq_enter(ibcq_to_rvtcq(qp->ibqp.recv_cq), &wc,
-		     wqe->wr.send_flags & IB_SEND_SOLICITED);
-
-send_comp:
-	spin_lock_irqsave(&sqp->s_lock, flags);
-	ibp->rvp.n_loop_pkts++;
-flush_send:
-	sqp->s_rnr_retry = sqp->s_rnr_retry_cnt;
-	rvt_send_complete(sqp, wqe, send_status);
-	if (local_ops) {
-		atomic_dec(&sqp->local_ops_pending);
-		local_ops = 0;
-	}
-	goto again;
-
-rnr_nak:
-	/* Handle RNR NAK */
-	if (qp->ibqp.qp_type == IB_QPT_UC)
-		goto send_comp;
-	ibp->rvp.n_rnr_naks++;
-	/*
-	 * Note: we don't need the s_lock held since the BUSY flag
-	 * makes this single threaded.
-	 */
-	if (sqp->s_rnr_retry == 0) {
-		send_status = IB_WC_RNR_RETRY_EXC_ERR;
-		goto serr;
-	}
-	if (sqp->s_rnr_retry_cnt < 7)
-		sqp->s_rnr_retry--;
-	spin_lock_irqsave(&sqp->s_lock, flags);
-	if (!(ib_rvt_state_ops[sqp->state] & RVT_PROCESS_RECV_OK))
-		goto clr_busy;
-	rvt_add_rnr_timer(sqp, qp->r_min_rnr_timer <<
-				IB_AETH_CREDIT_SHIFT);
-	goto clr_busy;
-
-op_err:
-	send_status = IB_WC_REM_OP_ERR;
-	wc.status = IB_WC_LOC_QP_OP_ERR;
-	goto err;
-
-inv_err:
-	send_status = IB_WC_REM_INV_REQ_ERR;
-	wc.status = IB_WC_LOC_QP_OP_ERR;
-	goto err;
-
-acc_err:
-	send_status = IB_WC_REM_ACCESS_ERR;
-	wc.status = IB_WC_LOC_PROT_ERR;
-err:
-	/* responder goes to error state */
-	rvt_rc_error(qp, wc.status);
-
-serr:
-	spin_lock_irqsave(&sqp->s_lock, flags);
-	rvt_send_complete(sqp, wqe, send_status);
-	if (sqp->ibqp.qp_type == IB_QPT_RC) {
-		int lastwqe = rvt_error_qp(sqp, IB_WC_WR_FLUSH_ERR);
-
-		sqp->s_flags &= ~RVT_S_BUSY;
-		spin_unlock_irqrestore(&sqp->s_lock, flags);
-		if (lastwqe) {
-			struct ib_event ev;
-
-			ev.device = sqp->ibqp.device;
-			ev.element.qp = &sqp->ibqp;
-			ev.event = IB_EVENT_QP_LAST_WQE_REACHED;
-			sqp->ibqp.event_handler(&ev, sqp->ibqp.qp_context);
-		}
-		goto done;
-	}
-clr_busy:
-	sqp->s_flags &= ~RVT_S_BUSY;
-unlock:
-	spin_unlock_irqrestore(&sqp->s_lock, flags);
-done:
-	rcu_read_unlock();
-}
-
 /**
  * hfi1_make_grh - construct a GRH header
  * @ibp: a pointer to the IB port
@@ -860,7 +532,7 @@ void hfi1_do_send(struct rvt_qp *qp, bool in_thread)
 		if (!loopback && ((rdma_ah_get_dlid(&qp->remote_ah_attr) &
 				   ~((1 << ps.ppd->lmc) - 1)) ==
 				  ps.ppd->lid)) {
-			ruc_loopback(qp);
+			rvt_ruc_loopback(qp);
 			return;
 		}
 		make_req = hfi1_make_rc_req;
@@ -870,7 +542,7 @@ void hfi1_do_send(struct rvt_qp *qp, bool in_thread)
 		if (!loopback && ((rdma_ah_get_dlid(&qp->remote_ah_attr) &
 				   ~((1 << ps.ppd->lmc) - 1)) ==
 				  ps.ppd->lid)) {
-			ruc_loopback(qp);
+			rvt_ruc_loopback(qp);
 			return;
 		}
 		make_req = hfi1_make_uc_req;
diff --git a/drivers/infiniband/hw/qib/qib_ruc.c b/drivers/infiniband/hw/qib/qib_ruc.c
index c5627baf5dbf..1fa21938f310 100644
--- a/drivers/infiniband/hw/qib/qib_ruc.c
+++ b/drivers/infiniband/hw/qib/qib_ruc.c
@@ -170,307 +170,6 @@ err:
 	return 1;
 }
 
-/**
- * qib_ruc_loopback - handle UC and RC lookback requests
- * @sqp: the sending QP
- *
- * This is called from qib_do_send() to
- * forward a WQE addressed to the same HCA.
- * Note that although we are single threaded due to the tasklet, we still
- * have to protect against post_send().  We don't have to worry about
- * receive interrupts since this is a connected protocol and all packets
- * will pass through here.
- */
-static void qib_ruc_loopback(struct rvt_qp *sqp)
-{
-	struct qib_ibport *ibp = to_iport(sqp->ibqp.device, sqp->port_num);
-	struct qib_pportdata *ppd = ppd_from_ibp(ibp);
-	struct qib_devdata *dd = ppd->dd;
-	struct rvt_dev_info *rdi = &dd->verbs_dev.rdi;
-	struct rvt_qp *qp;
-	struct rvt_swqe *wqe;
-	struct rvt_sge *sge;
-	unsigned long flags;
-	struct ib_wc wc;
-	u64 sdata;
-	atomic64_t *maddr;
-	enum ib_wc_status send_status;
-	int release;
-	int ret;
-
-	rcu_read_lock();
-	/*
-	 * Note that we check the responder QP state after
-	 * checking the requester's state.
-	 */
-	qp = rvt_lookup_qpn(rdi, &ibp->rvp, sqp->remote_qpn);
-	if (!qp)
-		goto done;
-
-	spin_lock_irqsave(&sqp->s_lock, flags);
-
-	/* Return if we are already busy processing a work request. */
-	if ((sqp->s_flags & (RVT_S_BUSY | RVT_S_ANY_WAIT)) ||
-	    !(ib_rvt_state_ops[sqp->state] & RVT_PROCESS_OR_FLUSH_SEND))
-		goto unlock;
-
-	sqp->s_flags |= RVT_S_BUSY;
-
-again:
-	if (sqp->s_last == READ_ONCE(sqp->s_head))
-		goto clr_busy;
-	wqe = rvt_get_swqe_ptr(sqp, sqp->s_last);
-
-	/* Return if it is not OK to start a new work reqeust. */
-	if (!(ib_rvt_state_ops[sqp->state] & RVT_PROCESS_NEXT_SEND_OK)) {
-		if (!(ib_rvt_state_ops[sqp->state] & RVT_FLUSH_SEND))
-			goto clr_busy;
-		/* We are in the error state, flush the work request. */
-		send_status = IB_WC_WR_FLUSH_ERR;
-		goto flush_send;
-	}
-
-	/*
-	 * We can rely on the entry not changing without the s_lock
-	 * being held until we update s_last.
-	 * We increment s_cur to indicate s_last is in progress.
-	 */
-	if (sqp->s_last == sqp->s_cur) {
-		if (++sqp->s_cur >= sqp->s_size)
-			sqp->s_cur = 0;
-	}
-	spin_unlock_irqrestore(&sqp->s_lock, flags);
-
-	if (!qp || !(ib_rvt_state_ops[qp->state] & RVT_PROCESS_RECV_OK) ||
-	    qp->ibqp.qp_type != sqp->ibqp.qp_type) {
-		ibp->rvp.n_pkt_drops++;
-		/*
-		 * For RC, the requester would timeout and retry so
-		 * shortcut the timeouts and just signal too many retries.
-		 */
-		if (sqp->ibqp.qp_type == IB_QPT_RC)
-			send_status = IB_WC_RETRY_EXC_ERR;
-		else
-			send_status = IB_WC_SUCCESS;
-		goto serr;
-	}
-
-	memset(&wc, 0, sizeof(wc));
-	send_status = IB_WC_SUCCESS;
-
-	release = 1;
-	sqp->s_sge.sge = wqe->sg_list[0];
-	sqp->s_sge.sg_list = wqe->sg_list + 1;
-	sqp->s_sge.num_sge = wqe->wr.num_sge;
-	sqp->s_len = wqe->length;
-	switch (wqe->wr.opcode) {
-	case IB_WR_SEND_WITH_IMM:
-		wc.wc_flags = IB_WC_WITH_IMM;
-		wc.ex.imm_data = wqe->wr.ex.imm_data;
-		/* FALLTHROUGH */
-	case IB_WR_SEND:
-		ret = rvt_get_rwqe(qp, false);
-		if (ret < 0)
-			goto op_err;
-		if (!ret)
-			goto rnr_nak;
-		break;
-
-	case IB_WR_RDMA_WRITE_WITH_IMM:
-		if (unlikely(!(qp->qp_access_flags & IB_ACCESS_REMOTE_WRITE)))
-			goto inv_err;
-		wc.wc_flags = IB_WC_WITH_IMM;
-		wc.ex.imm_data = wqe->wr.ex.imm_data;
-		ret = rvt_get_rwqe(qp, true);
-		if (ret < 0)
-			goto op_err;
-		if (!ret)
-			goto rnr_nak;
-		/* FALLTHROUGH */
-	case IB_WR_RDMA_WRITE:
-		if (unlikely(!(qp->qp_access_flags & IB_ACCESS_REMOTE_WRITE)))
-			goto inv_err;
-		if (wqe->length == 0)
-			break;
-		if (unlikely(!rvt_rkey_ok(qp, &qp->r_sge.sge, wqe->length,
-					  wqe->rdma_wr.remote_addr,
-					  wqe->rdma_wr.rkey,
-					  IB_ACCESS_REMOTE_WRITE)))
-			goto acc_err;
-		qp->r_sge.sg_list = NULL;
-		qp->r_sge.num_sge = 1;
-		qp->r_sge.total_len = wqe->length;
-		break;
-
-	case IB_WR_RDMA_READ:
-		if (unlikely(!(qp->qp_access_flags & IB_ACCESS_REMOTE_READ)))
-			goto inv_err;
-		if (unlikely(!rvt_rkey_ok(qp, &sqp->s_sge.sge, wqe->length,
-					  wqe->rdma_wr.remote_addr,
-					  wqe->rdma_wr.rkey,
-					  IB_ACCESS_REMOTE_READ)))
-			goto acc_err;
-		release = 0;
-		sqp->s_sge.sg_list = NULL;
-		sqp->s_sge.num_sge = 1;
-		qp->r_sge.sge = wqe->sg_list[0];
-		qp->r_sge.sg_list = wqe->sg_list + 1;
-		qp->r_sge.num_sge = wqe->wr.num_sge;
-		qp->r_sge.total_len = wqe->length;
-		break;
-
-	case IB_WR_ATOMIC_CMP_AND_SWP:
-	case IB_WR_ATOMIC_FETCH_AND_ADD:
-		if (unlikely(!(qp->qp_access_flags & IB_ACCESS_REMOTE_ATOMIC)))
-			goto inv_err;
-		if (unlikely(!rvt_rkey_ok(qp, &qp->r_sge.sge, sizeof(u64),
-					  wqe->atomic_wr.remote_addr,
-					  wqe->atomic_wr.rkey,
-					  IB_ACCESS_REMOTE_ATOMIC)))
-			goto acc_err;
-		/* Perform atomic OP and save result. */
-		maddr = (atomic64_t *) qp->r_sge.sge.vaddr;
-		sdata = wqe->atomic_wr.compare_add;
-		*(u64 *) sqp->s_sge.sge.vaddr =
-			(wqe->atomic_wr.wr.opcode == IB_WR_ATOMIC_FETCH_AND_ADD) ?
-			(u64) atomic64_add_return(sdata, maddr) - sdata :
-			(u64) cmpxchg((u64 *) qp->r_sge.sge.vaddr,
-				      sdata, wqe->atomic_wr.swap);
-		rvt_put_mr(qp->r_sge.sge.mr);
-		qp->r_sge.num_sge = 0;
-		goto send_comp;
-
-	default:
-		send_status = IB_WC_LOC_QP_OP_ERR;
-		goto serr;
-	}
-
-	sge = &sqp->s_sge.sge;
-	while (sqp->s_len) {
-		u32 len = sqp->s_len;
-
-		if (len > sge->length)
-			len = sge->length;
-		if (len > sge->sge_length)
-			len = sge->sge_length;
-		BUG_ON(len == 0);
-		rvt_copy_sge(qp, &qp->r_sge, sge->vaddr, len, release, false);
-		sge->vaddr += len;
-		sge->length -= len;
-		sge->sge_length -= len;
-		if (sge->sge_length == 0) {
-			if (!release)
-				rvt_put_mr(sge->mr);
-			if (--sqp->s_sge.num_sge)
-				*sge = *sqp->s_sge.sg_list++;
-		} else if (sge->length == 0 && sge->mr->lkey) {
-			if (++sge->n >= RVT_SEGSZ) {
-				if (++sge->m >= sge->mr->mapsz)
-					break;
-				sge->n = 0;
-			}
-			sge->vaddr =
-				sge->mr->map[sge->m]->segs[sge->n].vaddr;
-			sge->length =
-				sge->mr->map[sge->m]->segs[sge->n].length;
-		}
-		sqp->s_len -= len;
-	}
-	if (release)
-		rvt_put_ss(&qp->r_sge);
-
-	if (!test_and_clear_bit(RVT_R_WRID_VALID, &qp->r_aflags))
-		goto send_comp;
-
-	if (wqe->wr.opcode == IB_WR_RDMA_WRITE_WITH_IMM)
-		wc.opcode = IB_WC_RECV_RDMA_WITH_IMM;
-	else
-		wc.opcode = IB_WC_RECV;
-	wc.wr_id = qp->r_wr_id;
-	wc.status = IB_WC_SUCCESS;
-	wc.byte_len = wqe->length;
-	wc.qp = &qp->ibqp;
-	wc.src_qp = qp->remote_qpn;
-	wc.slid = rdma_ah_get_dlid(&qp->remote_ah_attr);
-	wc.sl = rdma_ah_get_sl(&qp->remote_ah_attr);
-	wc.port_num = 1;
-	/* Signal completion event if the solicited bit is set. */
-	rvt_cq_enter(ibcq_to_rvtcq(qp->ibqp.recv_cq), &wc,
-		     wqe->wr.send_flags & IB_SEND_SOLICITED);
-
-send_comp:
-	spin_lock_irqsave(&sqp->s_lock, flags);
-	ibp->rvp.n_loop_pkts++;
-flush_send:
-	sqp->s_rnr_retry = sqp->s_rnr_retry_cnt;
-	rvt_send_complete(sqp, wqe, send_status);
-	goto again;
-
-rnr_nak:
-	/* Handle RNR NAK */
-	if (qp->ibqp.qp_type == IB_QPT_UC)
-		goto send_comp;
-	ibp->rvp.n_rnr_naks++;
-	/*
-	 * Note: we don't need the s_lock held since the BUSY flag
-	 * makes this single threaded.
-	 */
-	if (sqp->s_rnr_retry == 0) {
-		send_status = IB_WC_RNR_RETRY_EXC_ERR;
-		goto serr;
-	}
-	if (sqp->s_rnr_retry_cnt < 7)
-		sqp->s_rnr_retry--;
-	spin_lock_irqsave(&sqp->s_lock, flags);
-	if (!(ib_rvt_state_ops[sqp->state] & RVT_PROCESS_RECV_OK))
-		goto clr_busy;
-	rvt_add_rnr_timer(sqp, qp->r_min_rnr_timer <<
-				IB_AETH_CREDIT_SHIFT);
-	goto clr_busy;
-
-op_err:
-	send_status = IB_WC_REM_OP_ERR;
-	wc.status = IB_WC_LOC_QP_OP_ERR;
-	goto err;
-
-inv_err:
-	send_status = IB_WC_REM_INV_REQ_ERR;
-	wc.status = IB_WC_LOC_QP_OP_ERR;
-	goto err;
-
-acc_err:
-	send_status = IB_WC_REM_ACCESS_ERR;
-	wc.status = IB_WC_LOC_PROT_ERR;
-err:
-	/* responder goes to error state */
-	rvt_rc_error(qp, wc.status);
-
-serr:
-	spin_lock_irqsave(&sqp->s_lock, flags);
-	rvt_send_complete(sqp, wqe, send_status);
-	if (sqp->ibqp.qp_type == IB_QPT_RC) {
-		int lastwqe = rvt_error_qp(sqp, IB_WC_WR_FLUSH_ERR);
-
-		sqp->s_flags &= ~RVT_S_BUSY;
-		spin_unlock_irqrestore(&sqp->s_lock, flags);
-		if (lastwqe) {
-			struct ib_event ev;
-
-			ev.device = sqp->ibqp.device;
-			ev.element.qp = &sqp->ibqp;
-			ev.event = IB_EVENT_QP_LAST_WQE_REACHED;
-			sqp->ibqp.event_handler(&ev, sqp->ibqp.qp_context);
-		}
-		goto done;
-	}
-clr_busy:
-	sqp->s_flags &= ~RVT_S_BUSY;
-unlock:
-	spin_unlock_irqrestore(&sqp->s_lock, flags);
-done:
-	rcu_read_unlock();
-}
-
 /**
  * qib_make_grh - construct a GRH header
  * @ibp: a pointer to the IB port
@@ -573,7 +272,7 @@ void qib_do_send(struct rvt_qp *qp)
 	     qp->ibqp.qp_type == IB_QPT_UC) &&
 	    (rdma_ah_get_dlid(&qp->remote_ah_attr) &
 	     ~((1 << ppd->lmc) - 1)) == ppd->lid) {
-		qib_ruc_loopback(qp);
+		rvt_ruc_loopback(qp);
 		return;
 	}
 
diff --git a/drivers/infiniband/sw/rdmavt/qp.c b/drivers/infiniband/sw/rdmavt/qp.c
index 7e3ec6674cf7..1735deb1a9d4 100644
--- a/drivers/infiniband/sw/rdmavt/qp.c
+++ b/drivers/infiniband/sw/rdmavt/qp.c
@@ -2777,3 +2777,334 @@ again:
 	}
 }
 EXPORT_SYMBOL(rvt_copy_sge);
+
+/**
+ * ruc_loopback - handle UC and RC loopback requests
+ * @sqp: the sending QP
+ *
+ * This is called from rvt_do_send() to forward a WQE addressed to the same HFI
+ * Note that although we are single threaded due to the send engine, we still
+ * have to protect against post_send().  We don't have to worry about
+ * receive interrupts since this is a connected protocol and all packets
+ * will pass through here.
+ */
+void rvt_ruc_loopback(struct rvt_qp *sqp)
+{
+	struct rvt_ibport *rvp =  NULL;
+	struct rvt_dev_info *rdi = ib_to_rvt(sqp->ibqp.device);
+	struct rvt_qp *qp;
+	struct rvt_swqe *wqe;
+	struct rvt_sge *sge;
+	unsigned long flags;
+	struct ib_wc wc;
+	u64 sdata;
+	atomic64_t *maddr;
+	enum ib_wc_status send_status;
+	bool release;
+	int ret;
+	bool copy_last = false;
+	int local_ops = 0;
+
+	rcu_read_lock();
+	rvp = rdi->ports[sqp->port_num - 1];
+
+	/*
+	 * Note that we check the responder QP state after
+	 * checking the requester's state.
+	 */
+
+	qp = rvt_lookup_qpn(ib_to_rvt(sqp->ibqp.device), rvp,
+			    sqp->remote_qpn);
+
+	spin_lock_irqsave(&sqp->s_lock, flags);
+
+	/* Return if we are already busy processing a work request. */
+	if ((sqp->s_flags & (RVT_S_BUSY | RVT_S_ANY_WAIT)) ||
+	    !(ib_rvt_state_ops[sqp->state] & RVT_PROCESS_OR_FLUSH_SEND))
+		goto unlock;
+
+	sqp->s_flags |= RVT_S_BUSY;
+
+again:
+	if (sqp->s_last == READ_ONCE(sqp->s_head))
+		goto clr_busy;
+	wqe = rvt_get_swqe_ptr(sqp, sqp->s_last);
+
+	/* Return if it is not OK to start a new work request. */
+	if (!(ib_rvt_state_ops[sqp->state] & RVT_PROCESS_NEXT_SEND_OK)) {
+		if (!(ib_rvt_state_ops[sqp->state] & RVT_FLUSH_SEND))
+			goto clr_busy;
+		/* We are in the error state, flush the work request. */
+		send_status = IB_WC_WR_FLUSH_ERR;
+		goto flush_send;
+	}
+
+	/*
+	 * We can rely on the entry not changing without the s_lock
+	 * being held until we update s_last.
+	 * We increment s_cur to indicate s_last is in progress.
+	 */
+	if (sqp->s_last == sqp->s_cur) {
+		if (++sqp->s_cur >= sqp->s_size)
+			sqp->s_cur = 0;
+	}
+	spin_unlock_irqrestore(&sqp->s_lock, flags);
+
+	if (!qp || !(ib_rvt_state_ops[qp->state] & RVT_PROCESS_RECV_OK) ||
+	    qp->ibqp.qp_type != sqp->ibqp.qp_type) {
+		rvp->n_pkt_drops++;
+		/*
+		 * For RC, the requester would timeout and retry so
+		 * shortcut the timeouts and just signal too many retries.
+		 */
+		if (sqp->ibqp.qp_type == IB_QPT_RC)
+			send_status = IB_WC_RETRY_EXC_ERR;
+		else
+			send_status = IB_WC_SUCCESS;
+		goto serr;
+	}
+
+	memset(&wc, 0, sizeof(wc));
+	send_status = IB_WC_SUCCESS;
+
+	release = true;
+	sqp->s_sge.sge = wqe->sg_list[0];
+	sqp->s_sge.sg_list = wqe->sg_list + 1;
+	sqp->s_sge.num_sge = wqe->wr.num_sge;
+	sqp->s_len = wqe->length;
+	switch (wqe->wr.opcode) {
+	case IB_WR_REG_MR:
+		goto send_comp;
+
+	case IB_WR_LOCAL_INV:
+		if (!(wqe->wr.send_flags & RVT_SEND_COMPLETION_ONLY)) {
+			if (rvt_invalidate_rkey(sqp,
+						wqe->wr.ex.invalidate_rkey))
+				send_status = IB_WC_LOC_PROT_ERR;
+			local_ops = 1;
+		}
+		goto send_comp;
+
+	case IB_WR_SEND_WITH_INV:
+		if (!rvt_invalidate_rkey(qp, wqe->wr.ex.invalidate_rkey)) {
+			wc.wc_flags = IB_WC_WITH_INVALIDATE;
+			wc.ex.invalidate_rkey = wqe->wr.ex.invalidate_rkey;
+		}
+		goto send;
+
+	case IB_WR_SEND_WITH_IMM:
+		wc.wc_flags = IB_WC_WITH_IMM;
+		wc.ex.imm_data = wqe->wr.ex.imm_data;
+		/* FALLTHROUGH */
+	case IB_WR_SEND:
+send:
+		ret = rvt_get_rwqe(qp, false);
+		if (ret < 0)
+			goto op_err;
+		if (!ret)
+			goto rnr_nak;
+		break;
+
+	case IB_WR_RDMA_WRITE_WITH_IMM:
+		if (unlikely(!(qp->qp_access_flags & IB_ACCESS_REMOTE_WRITE)))
+			goto inv_err;
+		wc.wc_flags = IB_WC_WITH_IMM;
+		wc.ex.imm_data = wqe->wr.ex.imm_data;
+		ret = rvt_get_rwqe(qp, true);
+		if (ret < 0)
+			goto op_err;
+		if (!ret)
+			goto rnr_nak;
+		/* skip copy_last set and qp_access_flags recheck */
+		goto do_write;
+	case IB_WR_RDMA_WRITE:
+		copy_last = rvt_is_user_qp(qp);
+		if (unlikely(!(qp->qp_access_flags & IB_ACCESS_REMOTE_WRITE)))
+			goto inv_err;
+do_write:
+		if (wqe->length == 0)
+			break;
+		if (unlikely(!rvt_rkey_ok(qp, &qp->r_sge.sge, wqe->length,
+					  wqe->rdma_wr.remote_addr,
+					  wqe->rdma_wr.rkey,
+					  IB_ACCESS_REMOTE_WRITE)))
+			goto acc_err;
+		qp->r_sge.sg_list = NULL;
+		qp->r_sge.num_sge = 1;
+		qp->r_sge.total_len = wqe->length;
+		break;
+
+	case IB_WR_RDMA_READ:
+		if (unlikely(!(qp->qp_access_flags & IB_ACCESS_REMOTE_READ)))
+			goto inv_err;
+		if (unlikely(!rvt_rkey_ok(qp, &sqp->s_sge.sge, wqe->length,
+					  wqe->rdma_wr.remote_addr,
+					  wqe->rdma_wr.rkey,
+					  IB_ACCESS_REMOTE_READ)))
+			goto acc_err;
+		release = false;
+		sqp->s_sge.sg_list = NULL;
+		sqp->s_sge.num_sge = 1;
+		qp->r_sge.sge = wqe->sg_list[0];
+		qp->r_sge.sg_list = wqe->sg_list + 1;
+		qp->r_sge.num_sge = wqe->wr.num_sge;
+		qp->r_sge.total_len = wqe->length;
+		break;
+
+	case IB_WR_ATOMIC_CMP_AND_SWP:
+	case IB_WR_ATOMIC_FETCH_AND_ADD:
+		if (unlikely(!(qp->qp_access_flags & IB_ACCESS_REMOTE_ATOMIC)))
+			goto inv_err;
+		if (unlikely(!rvt_rkey_ok(qp, &qp->r_sge.sge, sizeof(u64),
+					  wqe->atomic_wr.remote_addr,
+					  wqe->atomic_wr.rkey,
+					  IB_ACCESS_REMOTE_ATOMIC)))
+			goto acc_err;
+		/* Perform atomic OP and save result. */
+		maddr = (atomic64_t *)qp->r_sge.sge.vaddr;
+		sdata = wqe->atomic_wr.compare_add;
+		*(u64 *)sqp->s_sge.sge.vaddr =
+			(wqe->wr.opcode == IB_WR_ATOMIC_FETCH_AND_ADD) ?
+			(u64)atomic64_add_return(sdata, maddr) - sdata :
+			(u64)cmpxchg((u64 *)qp->r_sge.sge.vaddr,
+				      sdata, wqe->atomic_wr.swap);
+		rvt_put_mr(qp->r_sge.sge.mr);
+		qp->r_sge.num_sge = 0;
+		goto send_comp;
+
+	default:
+		send_status = IB_WC_LOC_QP_OP_ERR;
+		goto serr;
+	}
+
+	sge = &sqp->s_sge.sge;
+	while (sqp->s_len) {
+		u32 len = sqp->s_len;
+
+		if (len > sge->length)
+			len = sge->length;
+		if (len > sge->sge_length)
+			len = sge->sge_length;
+		WARN_ON_ONCE(len == 0);
+		rvt_copy_sge(qp, &qp->r_sge, sge->vaddr,
+			     len, release, copy_last);
+		sge->vaddr += len;
+		sge->length -= len;
+		sge->sge_length -= len;
+		if (sge->sge_length == 0) {
+			if (!release)
+				rvt_put_mr(sge->mr);
+			if (--sqp->s_sge.num_sge)
+				*sge = *sqp->s_sge.sg_list++;
+		} else if (sge->length == 0 && sge->mr->lkey) {
+			if (++sge->n >= RVT_SEGSZ) {
+				if (++sge->m >= sge->mr->mapsz)
+					break;
+				sge->n = 0;
+			}
+			sge->vaddr =
+				sge->mr->map[sge->m]->segs[sge->n].vaddr;
+			sge->length =
+				sge->mr->map[sge->m]->segs[sge->n].length;
+		}
+		sqp->s_len -= len;
+	}
+	if (release)
+		rvt_put_ss(&qp->r_sge);
+
+	if (!test_and_clear_bit(RVT_R_WRID_VALID, &qp->r_aflags))
+		goto send_comp;
+
+	if (wqe->wr.opcode == IB_WR_RDMA_WRITE_WITH_IMM)
+		wc.opcode = IB_WC_RECV_RDMA_WITH_IMM;
+	else
+		wc.opcode = IB_WC_RECV;
+	wc.wr_id = qp->r_wr_id;
+	wc.status = IB_WC_SUCCESS;
+	wc.byte_len = wqe->length;
+	wc.qp = &qp->ibqp;
+	wc.src_qp = qp->remote_qpn;
+	wc.slid = rdma_ah_get_dlid(&qp->remote_ah_attr) & U16_MAX;
+	wc.sl = rdma_ah_get_sl(&qp->remote_ah_attr);
+	wc.port_num = 1;
+	/* Signal completion event if the solicited bit is set. */
+	rvt_cq_enter(ibcq_to_rvtcq(qp->ibqp.recv_cq), &wc,
+		     wqe->wr.send_flags & IB_SEND_SOLICITED);
+
+send_comp:
+	spin_lock_irqsave(&sqp->s_lock, flags);
+	rvp->n_loop_pkts++;
+flush_send:
+	sqp->s_rnr_retry = sqp->s_rnr_retry_cnt;
+	rvt_send_complete(sqp, wqe, send_status);
+	if (local_ops) {
+		atomic_dec(&sqp->local_ops_pending);
+		local_ops = 0;
+	}
+	goto again;
+
+rnr_nak:
+	/* Handle RNR NAK */
+	if (qp->ibqp.qp_type == IB_QPT_UC)
+		goto send_comp;
+	rvp->n_rnr_naks++;
+	/*
+	 * Note: we don't need the s_lock held since the BUSY flag
+	 * makes this single threaded.
+	 */
+	if (sqp->s_rnr_retry == 0) {
+		send_status = IB_WC_RNR_RETRY_EXC_ERR;
+		goto serr;
+	}
+	if (sqp->s_rnr_retry_cnt < 7)
+		sqp->s_rnr_retry--;
+	spin_lock_irqsave(&sqp->s_lock, flags);
+	if (!(ib_rvt_state_ops[sqp->state] & RVT_PROCESS_RECV_OK))
+		goto clr_busy;
+	rvt_add_rnr_timer(sqp, qp->r_min_rnr_timer <<
+				IB_AETH_CREDIT_SHIFT);
+	goto clr_busy;
+
+op_err:
+	send_status = IB_WC_REM_OP_ERR;
+	wc.status = IB_WC_LOC_QP_OP_ERR;
+	goto err;
+
+inv_err:
+	send_status = IB_WC_REM_INV_REQ_ERR;
+	wc.status = IB_WC_LOC_QP_OP_ERR;
+	goto err;
+
+acc_err:
+	send_status = IB_WC_REM_ACCESS_ERR;
+	wc.status = IB_WC_LOC_PROT_ERR;
+err:
+	/* responder goes to error state */
+	rvt_rc_error(qp, wc.status);
+
+serr:
+	spin_lock_irqsave(&sqp->s_lock, flags);
+	rvt_send_complete(sqp, wqe, send_status);
+	if (sqp->ibqp.qp_type == IB_QPT_RC) {
+		int lastwqe = rvt_error_qp(sqp, IB_WC_WR_FLUSH_ERR);
+
+		sqp->s_flags &= ~RVT_S_BUSY;
+		spin_unlock_irqrestore(&sqp->s_lock, flags);
+		if (lastwqe) {
+			struct ib_event ev;
+
+			ev.device = sqp->ibqp.device;
+			ev.element.qp = &sqp->ibqp;
+			ev.event = IB_EVENT_QP_LAST_WQE_REACHED;
+			sqp->ibqp.event_handler(&ev, sqp->ibqp.qp_context);
+		}
+		goto done;
+	}
+clr_busy:
+	sqp->s_flags &= ~RVT_S_BUSY;
+unlock:
+	spin_unlock_irqrestore(&sqp->s_lock, flags);
+done:
+	rcu_read_unlock();
+}
+EXPORT_SYMBOL(rvt_ruc_loopback);
diff --git a/include/rdma/rdmavt_qp.h b/include/rdma/rdmavt_qp.h
index 6fd6f2ad9c0f..cbafb1878669 100644
--- a/include/rdma/rdmavt_qp.h
+++ b/include/rdma/rdmavt_qp.h
@@ -683,6 +683,7 @@ void rvt_copy_sge(struct rvt_qp *qp, struct rvt_sge_state *ss,
 		  bool release, bool copy_last);
 void rvt_send_complete(struct rvt_qp *qp, struct rvt_swqe *wqe,
 		       enum ib_wc_status status);
+void rvt_ruc_loopback(struct rvt_qp *qp);
 
 /**
  * struct rvt_qp_iter - the iterator for QPs
-- 
cgit v1.2.3-55-g7522


From 0797e6f1a85872ab48cb4986a06ebdcae22ecf1a Mon Sep 17 00:00:00 2001
From: Nathan Chancellor
Date: Wed, 26 Sep 2018 22:12:23 -0700
Subject: IB/rxe: Remove unnecessary enum values

Clang warns when an emumerated type is implicitly converted to another.

drivers/infiniband/sw/rxe/rxe.c:106:27: warning: implicit conversion
from enumeration type 'enum rxe_device_param' to different enumeration
type 'enum ib_atomic_cap' [-Wenum-conversion]
        rxe->attr.atomic_cap                    = RXE_ATOMIC_CAP;
                                                ~ ^~~~~~~~~~~~~~
drivers/infiniband/sw/rxe/rxe.c:131:22: warning: implicit conversion
from enumeration type 'enum rxe_port_param' to different enumeration
type 'enum ib_port_state' [-Wenum-conversion]
        port->attr.state                = RXE_PORT_STATE;
                                        ~ ^~~~~~~~~~~~~~
drivers/infiniband/sw/rxe/rxe.c:132:24: warning: implicit conversion
from enumeration type 'enum rxe_port_param' to different enumeration
type 'enum ib_mtu' [-Wenum-conversion]
        port->attr.max_mtu              = RXE_PORT_MAX_MTU;
                                        ~ ^~~~~~~~~~~~~~~~
drivers/infiniband/sw/rxe/rxe.c:133:27: warning: implicit conversion
from enumeration type 'enum rxe_port_param' to different enumeration
type 'enum ib_mtu' [-Wenum-conversion]
        port->attr.active_mtu           = RXE_PORT_ACTIVE_MTU;
                                        ~ ^~~~~~~~~~~~~~~~~~~
drivers/infiniband/sw/rxe/rxe.c:151:24: warning: implicit conversion
from enumeration type 'enum rxe_port_param' to different enumeration
type 'enum ib_mtu' [-Wenum-conversion]
                                ib_mtu_enum_to_int(RXE_PORT_ACTIVE_MTU);
                                ~~~~~~~~~~~~~~~~~~ ^~~~~~~~~~~~~~~~~~~
5 warnings generated.

Use the appropriate values from the expected enumerated type so no
conversion needs to happen then remove the unneeded definitions.

Reported-by: Nick Desaulniers <ndesaulniers@google.com>
Signed-off-by: Nathan Chancellor <natechancellor@gmail.com>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
---
 drivers/infiniband/sw/rxe/rxe.c       | 13 ++++++-------
 drivers/infiniband/sw/rxe/rxe_loc.h   |  2 +-
 drivers/infiniband/sw/rxe/rxe_param.h |  4 ----
 3 files changed, 7 insertions(+), 12 deletions(-)

(limited to 'drivers/infiniband/sw')

diff --git a/drivers/infiniband/sw/rxe/rxe.c b/drivers/infiniband/sw/rxe/rxe.c
index 10999fa69281..383e65c7bbc0 100644
--- a/drivers/infiniband/sw/rxe/rxe.c
+++ b/drivers/infiniband/sw/rxe/rxe.c
@@ -103,7 +103,7 @@ static void rxe_init_device_param(struct rxe_dev *rxe)
 	rxe->attr.max_res_rd_atom		= RXE_MAX_RES_RD_ATOM;
 	rxe->attr.max_qp_init_rd_atom		= RXE_MAX_QP_INIT_RD_ATOM;
 	rxe->attr.max_ee_init_rd_atom		= RXE_MAX_EE_INIT_RD_ATOM;
-	rxe->attr.atomic_cap			= RXE_ATOMIC_CAP;
+	rxe->attr.atomic_cap			= IB_ATOMIC_HCA;
 	rxe->attr.max_ee			= RXE_MAX_EE;
 	rxe->attr.max_rdd			= RXE_MAX_RDD;
 	rxe->attr.max_mw			= RXE_MAX_MW;
@@ -128,9 +128,9 @@ static void rxe_init_device_param(struct rxe_dev *rxe)
 /* initialize port attributes */
 static int rxe_init_port_param(struct rxe_port *port)
 {
-	port->attr.state		= RXE_PORT_STATE;
-	port->attr.max_mtu		= RXE_PORT_MAX_MTU;
-	port->attr.active_mtu		= RXE_PORT_ACTIVE_MTU;
+	port->attr.state		= IB_PORT_DOWN;
+	port->attr.max_mtu		= IB_MTU_4096;
+	port->attr.active_mtu		= IB_MTU_256;
 	port->attr.gid_tbl_len		= RXE_PORT_GID_TBL_LEN;
 	port->attr.port_cap_flags	= RXE_PORT_PORT_CAP_FLAGS;
 	port->attr.max_msg_sz		= RXE_PORT_MAX_MSG_SZ;
@@ -147,8 +147,7 @@ static int rxe_init_port_param(struct rxe_port *port)
 	port->attr.active_width		= RXE_PORT_ACTIVE_WIDTH;
 	port->attr.active_speed		= RXE_PORT_ACTIVE_SPEED;
 	port->attr.phys_state		= RXE_PORT_PHYS_STATE;
-	port->mtu_cap			=
-				ib_mtu_enum_to_int(RXE_PORT_ACTIVE_MTU);
+	port->mtu_cap			= ib_mtu_enum_to_int(IB_MTU_256);
 	port->subnet_prefix		= cpu_to_be64(RXE_PORT_SUBNET_PREFIX);
 
 	return 0;
@@ -300,7 +299,7 @@ void rxe_set_mtu(struct rxe_dev *rxe, unsigned int ndev_mtu)
 	mtu = eth_mtu_int_to_enum(ndev_mtu);
 
 	/* Make sure that new MTU in range */
-	mtu = mtu ? min_t(enum ib_mtu, mtu, RXE_PORT_MAX_MTU) : IB_MTU_256;
+	mtu = mtu ? min_t(enum ib_mtu, mtu, IB_MTU_4096) : IB_MTU_256;
 
 	port->attr.active_mtu = mtu;
 	port->mtu_cap = ib_mtu_enum_to_int(mtu);
diff --git a/drivers/infiniband/sw/rxe/rxe_loc.h b/drivers/infiniband/sw/rxe/rxe_loc.h
index 8e305422adbb..afd53f57a62b 100644
--- a/drivers/infiniband/sw/rxe/rxe_loc.h
+++ b/drivers/infiniband/sw/rxe/rxe_loc.h
@@ -195,7 +195,7 @@ static inline int qp_mtu(struct rxe_qp *qp)
 	if (qp->ibqp.qp_type == IB_QPT_RC || qp->ibqp.qp_type == IB_QPT_UC)
 		return qp->attr.path_mtu;
 	else
-		return RXE_PORT_MAX_MTU;
+		return IB_MTU_4096;
 }
 
 static inline int rcv_wqe_size(int max_sge)
diff --git a/drivers/infiniband/sw/rxe/rxe_param.h b/drivers/infiniband/sw/rxe/rxe_param.h
index 4555510d86c4..bdea899a58ac 100644
--- a/drivers/infiniband/sw/rxe/rxe_param.h
+++ b/drivers/infiniband/sw/rxe/rxe_param.h
@@ -90,7 +90,6 @@ enum rxe_device_param {
 	RXE_MAX_RES_RD_ATOM		= 0x3f000,
 	RXE_MAX_QP_INIT_RD_ATOM		= 128,
 	RXE_MAX_EE_INIT_RD_ATOM		= 0,
-	RXE_ATOMIC_CAP			= 1,
 	RXE_MAX_EE			= 0,
 	RXE_MAX_RDD			= 0,
 	RXE_MAX_MW			= 0,
@@ -139,9 +138,6 @@ enum rxe_device_param {
 
 /* default/initial rxe port parameters */
 enum rxe_port_param {
-	RXE_PORT_STATE			= IB_PORT_DOWN,
-	RXE_PORT_MAX_MTU		= IB_MTU_4096,
-	RXE_PORT_ACTIVE_MTU		= IB_MTU_256,
 	RXE_PORT_GID_TBL_LEN		= 1024,
 	RXE_PORT_PORT_CAP_FLAGS		= RDMA_CORE_CAP_PROT_ROCE_UDP_ENCAP,
 	RXE_PORT_MAX_MSG_SZ		= 0x800000,
-- 
cgit v1.2.3-55-g7522


From 508a523f6bc6cdfbf7031d66559d4ad24956b741 Mon Sep 17 00:00:00 2001
From: Parav Pandit
Date: Thu, 11 Oct 2018 22:31:54 +0300
Subject: RDMA/drivers: Use core provided API for registering device attributes

Use rdma_set_device_sysfs_group() to register device attributes and
simplify the driver.

Signed-off-by: Parav Pandit <parav@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
---
 drivers/infiniband/hw/bnxt_re/main.c           |  74 ++++++++----------
 drivers/infiniband/hw/cxgb3/iwch_provider.c    |  52 +++++--------
 drivers/infiniband/hw/cxgb4/provider.c         |  47 +++++-------
 drivers/infiniband/hw/hfi1/hfi.h               |   1 +
 drivers/infiniband/hw/hfi1/sysfs.c             |  69 ++++++++---------
 drivers/infiniband/hw/hfi1/verbs.c             |   3 +
 drivers/infiniband/hw/i40iw/i40iw_verbs.c      |  70 ++++++-----------
 drivers/infiniband/hw/mlx4/main.c              |  37 +++++----
 drivers/infiniband/hw/mlx5/ib_rep.c            |   3 -
 drivers/infiniband/hw/mlx5/main.c              |  66 ++++++----------
 drivers/infiniband/hw/mlx5/mlx5_ib.h           |   1 -
 drivers/infiniband/hw/mthca/mthca_provider.c   |  41 +++++-----
 drivers/infiniband/hw/nes/nes_verbs.c          |  58 +++++---------
 drivers/infiniband/hw/ocrdma/ocrdma_main.c     |  71 ++++++++---------
 drivers/infiniband/hw/qedr/main.c              |  69 +++++++----------
 drivers/infiniband/hw/qib/qib.h                |   2 +-
 drivers/infiniband/hw/qib/qib_sysfs.c          | 101 ++++++++++---------------
 drivers/infiniband/hw/qib/qib_verbs.c          |   7 +-
 drivers/infiniband/hw/usnic/usnic_ib_main.c    |   2 +
 drivers/infiniband/hw/usnic/usnic_ib_sysfs.c   |  70 ++++++-----------
 drivers/infiniband/hw/usnic/usnic_ib_sysfs.h   |   2 +
 drivers/infiniband/hw/vmw_pvrdma/pvrdma_main.c |  41 +++++-----
 drivers/infiniband/sw/rxe/rxe_verbs.c          |  26 ++-----
 23 files changed, 367 insertions(+), 546 deletions(-)

(limited to 'drivers/infiniband/sw')

diff --git a/drivers/infiniband/hw/bnxt_re/main.c b/drivers/infiniband/hw/bnxt_re/main.c
index e5d4b12f7bf5..cf2282654210 100644
--- a/drivers/infiniband/hw/bnxt_re/main.c
+++ b/drivers/infiniband/hw/bnxt_re/main.c
@@ -535,6 +535,34 @@ static struct bnxt_en_dev *bnxt_re_dev_probe(struct net_device *netdev)
 	return en_dev;
 }
 
+static ssize_t hw_rev_show(struct device *device, struct device_attribute *attr,
+			   char *buf)
+{
+	struct bnxt_re_dev *rdev = to_bnxt_re_dev(device, ibdev.dev);
+
+	return scnprintf(buf, PAGE_SIZE, "0x%x\n", rdev->en_dev->pdev->vendor);
+}
+static DEVICE_ATTR_RO(hw_rev);
+
+static ssize_t hca_type_show(struct device *device,
+			     struct device_attribute *attr, char *buf)
+{
+	struct bnxt_re_dev *rdev = to_bnxt_re_dev(device, ibdev.dev);
+
+	return scnprintf(buf, PAGE_SIZE, "%s\n", rdev->ibdev.node_desc);
+}
+static DEVICE_ATTR_RO(hca_type);
+
+static struct attribute *bnxt_re_attributes[] = {
+	&dev_attr_hw_rev.attr,
+	&dev_attr_hca_type.attr,
+	NULL
+};
+
+static const struct attribute_group bnxt_re_dev_attr_group = {
+	.attrs = bnxt_re_attributes,
+};
+
 static void bnxt_re_unregister_ib(struct bnxt_re_dev *rdev)
 {
 	ib_unregister_device(&rdev->ibdev);
@@ -638,34 +666,11 @@ static int bnxt_re_register_ib(struct bnxt_re_dev *rdev)
 	ibdev->get_hw_stats             = bnxt_re_ib_get_hw_stats;
 	ibdev->alloc_hw_stats           = bnxt_re_ib_alloc_hw_stats;
 
+	rdma_set_device_sysfs_group(ibdev, &bnxt_re_dev_attr_group);
 	ibdev->driver_id = RDMA_DRIVER_BNXT_RE;
 	return ib_register_device(ibdev, "bnxt_re%d", NULL);
 }
 
-static ssize_t show_rev(struct device *device, struct device_attribute *attr,
-			char *buf)
-{
-	struct bnxt_re_dev *rdev = to_bnxt_re_dev(device, ibdev.dev);
-
-	return scnprintf(buf, PAGE_SIZE, "0x%x\n", rdev->en_dev->pdev->vendor);
-}
-
-static ssize_t show_hca(struct device *device, struct device_attribute *attr,
-			char *buf)
-{
-	struct bnxt_re_dev *rdev = to_bnxt_re_dev(device, ibdev.dev);
-
-	return scnprintf(buf, PAGE_SIZE, "%s\n", rdev->ibdev.node_desc);
-}
-
-static DEVICE_ATTR(hw_rev, 0444, show_rev, NULL);
-static DEVICE_ATTR(hca_type, 0444, show_hca, NULL);
-
-static struct device_attribute *bnxt_re_attributes[] = {
-	&dev_attr_hw_rev,
-	&dev_attr_hca_type
-};
-
 static void bnxt_re_dev_remove(struct bnxt_re_dev *rdev)
 {
 	dev_put(rdev->netdev);
@@ -1200,12 +1205,9 @@ static int bnxt_re_setup_qos(struct bnxt_re_dev *rdev)
 
 static void bnxt_re_ib_unreg(struct bnxt_re_dev *rdev)
 {
-	int i, rc;
+	int rc;
 
 	if (test_and_clear_bit(BNXT_RE_FLAG_IBDEV_REGISTERED, &rdev->flags)) {
-		for (i = 0; i < ARRAY_SIZE(bnxt_re_attributes); i++)
-			device_remove_file(&rdev->ibdev.dev,
-					   bnxt_re_attributes[i]);
 		/* Cleanup ib dev */
 		bnxt_re_unregister_ib(rdev);
 	}
@@ -1255,7 +1257,7 @@ static void bnxt_re_worker(struct work_struct *work)
 
 static int bnxt_re_ib_reg(struct bnxt_re_dev *rdev)
 {
-	int i, j, rc;
+	int rc;
 
 	bool locked;
 
@@ -1375,20 +1377,6 @@ static int bnxt_re_ib_reg(struct bnxt_re_dev *rdev)
 	}
 	set_bit(BNXT_RE_FLAG_IBDEV_REGISTERED, &rdev->flags);
 	dev_info(rdev_to_dev(rdev), "Device registered successfully");
-	for (i = 0; i < ARRAY_SIZE(bnxt_re_attributes); i++) {
-		rc = device_create_file(&rdev->ibdev.dev,
-					bnxt_re_attributes[i]);
-		if (rc) {
-			dev_err(rdev_to_dev(rdev),
-				"Failed to create IB sysfs: %#x", rc);
-			/* Must clean up all created device files */
-			for (j = 0; j < i; j++)
-				device_remove_file(&rdev->ibdev.dev,
-						   bnxt_re_attributes[j]);
-			bnxt_re_unregister_ib(rdev);
-			goto fail;
-		}
-	}
 	ib_get_eth_speed(&rdev->ibdev, 1, &rdev->active_speed,
 			 &rdev->active_width);
 	set_bit(BNXT_RE_FLAG_ISSUE_ROCE_STATS, &rdev->flags);
diff --git a/drivers/infiniband/hw/cxgb3/iwch_provider.c b/drivers/infiniband/hw/cxgb3/iwch_provider.c
index 39530cc15f95..ebbec02cebe0 100644
--- a/drivers/infiniband/hw/cxgb3/iwch_provider.c
+++ b/drivers/infiniband/hw/cxgb3/iwch_provider.c
@@ -1127,17 +1127,18 @@ static int iwch_query_port(struct ib_device *ibdev,
 	return 0;
 }
 
-static ssize_t show_rev(struct device *dev, struct device_attribute *attr,
-			char *buf)
+static ssize_t hw_rev_show(struct device *dev,
+			   struct device_attribute *attr, char *buf)
 {
 	struct iwch_dev *iwch_dev = container_of(dev, struct iwch_dev,
 						 ibdev.dev);
 	pr_debug("%s dev 0x%p\n", __func__, dev);
 	return sprintf(buf, "%d\n", iwch_dev->rdev.t3cdev_p->type);
 }
+static DEVICE_ATTR_RO(hw_rev);
 
-static ssize_t show_hca(struct device *dev, struct device_attribute *attr,
-			char *buf)
+static ssize_t hca_type_show(struct device *dev,
+			     struct device_attribute *attr, char *buf)
 {
 	struct iwch_dev *iwch_dev = container_of(dev, struct iwch_dev,
 						 ibdev.dev);
@@ -1148,9 +1149,10 @@ static ssize_t show_hca(struct device *dev, struct device_attribute *attr,
 	lldev->ethtool_ops->get_drvinfo(lldev, &info);
 	return sprintf(buf, "%s\n", info.driver);
 }
+static DEVICE_ATTR_RO(hca_type);
 
-static ssize_t show_board(struct device *dev, struct device_attribute *attr,
-			  char *buf)
+static ssize_t board_id_show(struct device *dev,
+			     struct device_attribute *attr, char *buf)
 {
 	struct iwch_dev *iwch_dev = container_of(dev, struct iwch_dev,
 						 ibdev.dev);
@@ -1158,6 +1160,7 @@ static ssize_t show_board(struct device *dev, struct device_attribute *attr,
 	return sprintf(buf, "%x.%x\n", iwch_dev->rdev.rnic_info.pdev->vendor,
 		       iwch_dev->rdev.rnic_info.pdev->device);
 }
+static DEVICE_ATTR_RO(board_id);
 
 enum counters {
 	IPINRECEIVES,
@@ -1274,14 +1277,15 @@ static int iwch_get_mib(struct ib_device *ibdev, struct rdma_hw_stats *stats,
 	return stats->num_counters;
 }
 
-static DEVICE_ATTR(hw_rev, S_IRUGO, show_rev, NULL);
-static DEVICE_ATTR(hca_type, S_IRUGO, show_hca, NULL);
-static DEVICE_ATTR(board_id, S_IRUGO, show_board, NULL);
+static struct attribute *iwch_class_attributes[] = {
+	&dev_attr_hw_rev.attr,
+	&dev_attr_hca_type.attr,
+	&dev_attr_board_id.attr,
+	NULL
+};
 
-static struct device_attribute *iwch_class_attributes[] = {
-	&dev_attr_hw_rev,
-	&dev_attr_hca_type,
-	&dev_attr_board_id,
+static const struct attribute_group iwch_attr_group = {
+	.attrs = iwch_class_attributes,
 };
 
 static int iwch_port_immutable(struct ib_device *ibdev, u8 port_num,
@@ -1316,7 +1320,6 @@ static void get_dev_fw_ver_str(struct ib_device *ibdev, char *str)
 int iwch_register_device(struct iwch_dev *dev)
 {
 	int ret;
-	int i;
 
 	pr_debug("%s iwch_dev %p\n", __func__, dev);
 	memset(&dev->ibdev.node_guid, 0, sizeof(dev->ibdev.node_guid));
@@ -1401,33 +1404,16 @@ int iwch_register_device(struct iwch_dev *dev)
 	       sizeof(dev->ibdev.iwcm->ifname));
 
 	dev->ibdev.driver_id = RDMA_DRIVER_CXGB3;
+	rdma_set_device_sysfs_group(&dev->ibdev, &iwch_attr_group);
 	ret = ib_register_device(&dev->ibdev, "cxgb3_%d", NULL);
 	if (ret)
-		goto bail1;
-
-	for (i = 0; i < ARRAY_SIZE(iwch_class_attributes); ++i) {
-		ret = device_create_file(&dev->ibdev.dev,
-					 iwch_class_attributes[i]);
-		if (ret) {
-			goto bail2;
-		}
-	}
-	return 0;
-bail2:
-	ib_unregister_device(&dev->ibdev);
-bail1:
-	kfree(dev->ibdev.iwcm);
+		kfree(dev->ibdev.iwcm);
 	return ret;
 }
 
 void iwch_unregister_device(struct iwch_dev *dev)
 {
-	int i;
-
 	pr_debug("%s iwch_dev %p\n", __func__, dev);
-	for (i = 0; i < ARRAY_SIZE(iwch_class_attributes); ++i)
-		device_remove_file(&dev->ibdev.dev,
-				   iwch_class_attributes[i]);
 	ib_unregister_device(&dev->ibdev);
 	kfree(dev->ibdev.iwcm);
 	return;
diff --git a/drivers/infiniband/hw/cxgb4/provider.c b/drivers/infiniband/hw/cxgb4/provider.c
index 416f8d1af610..cbb3c0ddd990 100644
--- a/drivers/infiniband/hw/cxgb4/provider.c
+++ b/drivers/infiniband/hw/cxgb4/provider.c
@@ -373,8 +373,8 @@ static int c4iw_query_port(struct ib_device *ibdev, u8 port,
 	return 0;
 }
 
-static ssize_t show_rev(struct device *dev, struct device_attribute *attr,
-			char *buf)
+static ssize_t hw_rev_show(struct device *dev,
+			   struct device_attribute *attr, char *buf)
 {
 	struct c4iw_dev *c4iw_dev = container_of(dev, struct c4iw_dev,
 						 ibdev.dev);
@@ -382,9 +382,10 @@ static ssize_t show_rev(struct device *dev, struct device_attribute *attr,
 	return sprintf(buf, "%d\n",
 		       CHELSIO_CHIP_RELEASE(c4iw_dev->rdev.lldi.adapter_type));
 }
+static DEVICE_ATTR_RO(hw_rev);
 
-static ssize_t show_hca(struct device *dev, struct device_attribute *attr,
-			char *buf)
+static ssize_t hca_type_show(struct device *dev,
+			     struct device_attribute *attr, char *buf)
 {
 	struct c4iw_dev *c4iw_dev = container_of(dev, struct c4iw_dev,
 						 ibdev.dev);
@@ -395,9 +396,10 @@ static ssize_t show_hca(struct device *dev, struct device_attribute *attr,
 	lldev->ethtool_ops->get_drvinfo(lldev, &info);
 	return sprintf(buf, "%s\n", info.driver);
 }
+static DEVICE_ATTR_RO(hca_type);
 
-static ssize_t show_board(struct device *dev, struct device_attribute *attr,
-			  char *buf)
+static ssize_t board_id_show(struct device *dev, struct device_attribute *attr,
+			     char *buf)
 {
 	struct c4iw_dev *c4iw_dev = container_of(dev, struct c4iw_dev,
 						 ibdev.dev);
@@ -405,6 +407,7 @@ static ssize_t show_board(struct device *dev, struct device_attribute *attr,
 	return sprintf(buf, "%x.%x\n", c4iw_dev->rdev.lldi.pdev->vendor,
 		       c4iw_dev->rdev.lldi.pdev->device);
 }
+static DEVICE_ATTR_RO(board_id);
 
 enum counters {
 	IP4INSEGS,
@@ -461,14 +464,15 @@ static int c4iw_get_mib(struct ib_device *ibdev,
 	return stats->num_counters;
 }
 
-static DEVICE_ATTR(hw_rev, S_IRUGO, show_rev, NULL);
-static DEVICE_ATTR(hca_type, S_IRUGO, show_hca, NULL);
-static DEVICE_ATTR(board_id, S_IRUGO, show_board, NULL);
+static struct attribute *c4iw_class_attributes[] = {
+	&dev_attr_hw_rev.attr,
+	&dev_attr_hca_type.attr,
+	&dev_attr_board_id.attr,
+	NULL
+};
 
-static struct device_attribute *c4iw_class_attributes[] = {
-	&dev_attr_hw_rev,
-	&dev_attr_hca_type,
-	&dev_attr_board_id,
+static const struct attribute_group c4iw_attr_group = {
+	.attrs = c4iw_class_attributes,
 };
 
 static int c4iw_port_immutable(struct ib_device *ibdev, u8 port_num,
@@ -530,7 +534,6 @@ static int fill_res_entry(struct sk_buff *msg, struct rdma_restrack_entry *res)
 void c4iw_register_device(struct work_struct *work)
 {
 	int ret;
-	int i;
 	struct uld_ctx *ctx = container_of(work, struct uld_ctx, reg_work);
 	struct c4iw_dev *dev = ctx->dev;
 
@@ -625,20 +628,13 @@ void c4iw_register_device(struct work_struct *work)
 	memcpy(dev->ibdev.iwcm->ifname, dev->rdev.lldi.ports[0]->name,
 	       sizeof(dev->ibdev.iwcm->ifname));
 
+	rdma_set_device_sysfs_group(&dev->ibdev, &c4iw_attr_group);
 	dev->ibdev.driver_id = RDMA_DRIVER_CXGB4;
 	ret = ib_register_device(&dev->ibdev, "cxgb4_%d", NULL);
 	if (ret)
 		goto err_kfree_iwcm;
-
-	for (i = 0; i < ARRAY_SIZE(c4iw_class_attributes); ++i) {
-		ret = device_create_file(&dev->ibdev.dev,
-					 c4iw_class_attributes[i]);
-		if (ret)
-			goto err_unregister_device;
-	}
 	return;
-err_unregister_device:
-	ib_unregister_device(&dev->ibdev);
+
 err_kfree_iwcm:
 	kfree(dev->ibdev.iwcm);
 err_dealloc_ctx:
@@ -650,12 +646,7 @@ err_dealloc_ctx:
 
 void c4iw_unregister_device(struct c4iw_dev *dev)
 {
-	int i;
-
 	pr_debug("c4iw_dev %p\n", dev);
-	for (i = 0; i < ARRAY_SIZE(c4iw_class_attributes); ++i)
-		device_remove_file(&dev->ibdev.dev,
-				   c4iw_class_attributes[i]);
 	ib_unregister_device(&dev->ibdev);
 	kfree(dev->ibdev.iwcm);
 	return;
diff --git a/drivers/infiniband/hw/hfi1/hfi.h b/drivers/infiniband/hw/hfi1/hfi.h
index d3d5717650e9..1401b6ea4a28 100644
--- a/drivers/infiniband/hw/hfi1/hfi.h
+++ b/drivers/infiniband/hw/hfi1/hfi.h
@@ -1968,6 +1968,7 @@ static inline u32 get_rcvhdrtail(const struct hfi1_ctxtdata *rcd)
  */
 
 extern const char ib_hfi1_version[];
+extern const struct attribute_group ib_hfi1_attr_group;
 
 int hfi1_device_create(struct hfi1_devdata *dd);
 void hfi1_device_remove(struct hfi1_devdata *dd);
diff --git a/drivers/infiniband/hw/hfi1/sysfs.c b/drivers/infiniband/hw/hfi1/sysfs.c
index 25e867393463..2be513d4c9da 100644
--- a/drivers/infiniband/hw/hfi1/sysfs.c
+++ b/drivers/infiniband/hw/hfi1/sysfs.c
@@ -494,17 +494,18 @@ static struct kobj_type hfi1_vl2mtu_ktype = {
  * Start of per-unit (or driver, in some cases, but replicated
  * per unit) functions (these get a device *)
  */
-static ssize_t show_rev(struct device *device, struct device_attribute *attr,
-			char *buf)
+static ssize_t hw_rev_show(struct device *device, struct device_attribute *attr,
+			   char *buf)
 {
 	struct hfi1_ibdev *dev =
 		container_of(device, struct hfi1_ibdev, rdi.ibdev.dev);
 
 	return sprintf(buf, "%x\n", dd_from_dev(dev)->minrev);
 }
+static DEVICE_ATTR_RO(hw_rev);
 
-static ssize_t show_hfi(struct device *device, struct device_attribute *attr,
-			char *buf)
+static ssize_t board_id_show(struct device *device,
+			     struct device_attribute *attr, char *buf)
 {
 	struct hfi1_ibdev *dev =
 		container_of(device, struct hfi1_ibdev, rdi.ibdev.dev);
@@ -517,8 +518,9 @@ static ssize_t show_hfi(struct device *device, struct device_attribute *attr,
 		ret = scnprintf(buf, PAGE_SIZE, "%s\n", dd->boardname);
 	return ret;
 }
+static DEVICE_ATTR_RO(board_id);
 
-static ssize_t show_boardversion(struct device *device,
+static ssize_t boardversion_show(struct device *device,
 				 struct device_attribute *attr, char *buf)
 {
 	struct hfi1_ibdev *dev =
@@ -528,8 +530,9 @@ static ssize_t show_boardversion(struct device *device,
 	/* The string printed here is already newline-terminated. */
 	return scnprintf(buf, PAGE_SIZE, "%s", dd->boardversion);
 }
+static DEVICE_ATTR_RO(boardversion);
 
-static ssize_t show_nctxts(struct device *device,
+static ssize_t nctxts_show(struct device *device,
 			   struct device_attribute *attr, char *buf)
 {
 	struct hfi1_ibdev *dev =
@@ -546,8 +549,9 @@ static ssize_t show_nctxts(struct device *device,
 			 min(dd->num_user_contexts,
 			     (u32)dd->sc_sizes[SC_USER].count));
 }
+static DEVICE_ATTR_RO(nctxts);
 
-static ssize_t show_nfreectxts(struct device *device,
+static ssize_t nfreectxts_show(struct device *device,
 			       struct device_attribute *attr, char *buf)
 {
 	struct hfi1_ibdev *dev =
@@ -557,8 +561,9 @@ static ssize_t show_nfreectxts(struct device *device,
 	/* Return the number of free user ports (contexts) available. */
 	return scnprintf(buf, PAGE_SIZE, "%u\n", dd->freectxts);
 }
+static DEVICE_ATTR_RO(nfreectxts);
 
-static ssize_t show_serial(struct device *device,
+static ssize_t serial_show(struct device *device,
 			   struct device_attribute *attr, char *buf)
 {
 	struct hfi1_ibdev *dev =
@@ -567,8 +572,9 @@ static ssize_t show_serial(struct device *device,
 
 	return scnprintf(buf, PAGE_SIZE, "%s", dd->serial);
 }
+static DEVICE_ATTR_RO(serial);
 
-static ssize_t store_chip_reset(struct device *device,
+static ssize_t chip_reset_store(struct device *device,
 				struct device_attribute *attr, const char *buf,
 				size_t count)
 {
@@ -586,6 +592,7 @@ static ssize_t store_chip_reset(struct device *device,
 bail:
 	return ret < 0 ? ret : count;
 }
+static DEVICE_ATTR_WO(chip_reset);
 
 /*
  * Convert the reported temperature from an integer (reported in
@@ -598,7 +605,7 @@ bail:
 /*
  * Dump tempsense values, in decimal, to ease shell-scripts.
  */
-static ssize_t show_tempsense(struct device *device,
+static ssize_t tempsense_show(struct device *device,
 			      struct device_attribute *attr, char *buf)
 {
 	struct hfi1_ibdev *dev =
@@ -622,6 +629,7 @@ static ssize_t show_tempsense(struct device *device,
 	}
 	return ret;
 }
+static DEVICE_ATTR_RO(tempsense);
 
 /*
  * end of per-unit (or driver, in some cases, but replicated
@@ -629,24 +637,20 @@ static ssize_t show_tempsense(struct device *device,
  */
 
 /* start of per-unit file structures and support code */
-static DEVICE_ATTR(hw_rev, S_IRUGO, show_rev, NULL);
-static DEVICE_ATTR(board_id, S_IRUGO, show_hfi, NULL);
-static DEVICE_ATTR(nctxts, S_IRUGO, show_nctxts, NULL);
-static DEVICE_ATTR(nfreectxts, S_IRUGO, show_nfreectxts, NULL);
-static DEVICE_ATTR(serial, S_IRUGO, show_serial, NULL);
-static DEVICE_ATTR(boardversion, S_IRUGO, show_boardversion, NULL);
-static DEVICE_ATTR(tempsense, S_IRUGO, show_tempsense, NULL);
-static DEVICE_ATTR(chip_reset, S_IWUSR, NULL, store_chip_reset);
-
-static struct device_attribute *hfi1_attributes[] = {
-	&dev_attr_hw_rev,
-	&dev_attr_board_id,
-	&dev_attr_nctxts,
-	&dev_attr_nfreectxts,
-	&dev_attr_serial,
-	&dev_attr_boardversion,
-	&dev_attr_tempsense,
-	&dev_attr_chip_reset,
+static struct attribute *hfi1_attributes[] = {
+	&dev_attr_hw_rev.attr,
+	&dev_attr_board_id.attr,
+	&dev_attr_nctxts.attr,
+	&dev_attr_nfreectxts.attr,
+	&dev_attr_serial.attr,
+	&dev_attr_boardversion.attr,
+	&dev_attr_tempsense.attr,
+	&dev_attr_chip_reset.attr,
+	NULL,
+};
+
+const struct attribute_group ib_hfi1_attr_group = {
+	.attrs = hfi1_attributes,
 };
 
 int hfi1_create_port_files(struct ib_device *ibdev, u8 port_num,
@@ -832,12 +836,6 @@ int hfi1_verbs_register_sysfs(struct hfi1_devdata *dd)
 	struct device *class_dev = &dev->dev;
 	int i, j, ret;
 
-	for (i = 0; i < ARRAY_SIZE(hfi1_attributes); ++i) {
-		ret = device_create_file(&dev->dev, hfi1_attributes[i]);
-		if (ret)
-			goto bail;
-	}
-
 	for (i = 0; i < dd->num_sdma; i++) {
 		ret = kobject_init_and_add(&dd->per_sdma[i].kobj,
 					   &sde_ktype, &class_dev->kobj,
@@ -855,9 +853,6 @@ int hfi1_verbs_register_sysfs(struct hfi1_devdata *dd)
 
 	return 0;
 bail:
-	for (i = 0; i < ARRAY_SIZE(hfi1_attributes); ++i)
-		device_remove_file(&dev->dev, hfi1_attributes[i]);
-
 	for (i = 0; i < dd->num_sdma; i++)
 		kobject_del(&dd->per_sdma[i].kobj);
 
diff --git a/drivers/infiniband/hw/hfi1/verbs.c b/drivers/infiniband/hw/hfi1/verbs.c
index 5fc27a94b4f0..48e11e510358 100644
--- a/drivers/infiniband/hw/hfi1/verbs.c
+++ b/drivers/infiniband/hw/hfi1/verbs.c
@@ -1751,6 +1751,9 @@ int hfi1_register_ib_device(struct hfi1_devdata *dd)
 			      i,
 			      ppd->pkeys);
 
+	rdma_set_device_sysfs_group(&dd->verbs_dev.rdi.ibdev,
+				    &ib_hfi1_attr_group);
+
 	ret = rvt_register_device(&dd->verbs_dev.rdi, RDMA_DRIVER_HFI1);
 	if (ret)
 		goto err_verbs_txreq;
diff --git a/drivers/infiniband/hw/i40iw/i40iw_verbs.c b/drivers/infiniband/hw/i40iw/i40iw_verbs.c
index cb2aef874ca8..102875872bea 100644
--- a/drivers/infiniband/hw/i40iw/i40iw_verbs.c
+++ b/drivers/infiniband/hw/i40iw/i40iw_verbs.c
@@ -2135,10 +2135,10 @@ static int i40iw_dereg_mr(struct ib_mr *ib_mr)
 }
 
 /**
- * i40iw_show_rev
+ * hw_rev_show
  */
-static ssize_t i40iw_show_rev(struct device *dev,
-			      struct device_attribute *attr, char *buf)
+static ssize_t hw_rev_show(struct device *dev,
+			   struct device_attribute *attr, char *buf)
 {
 	struct i40iw_ib_device *iwibdev = container_of(dev,
 						       struct i40iw_ib_device,
@@ -2147,34 +2147,37 @@ static ssize_t i40iw_show_rev(struct device *dev,
 
 	return sprintf(buf, "%x\n", hw_rev);
 }
+static DEVICE_ATTR_RO(hw_rev);
 
 /**
- * i40iw_show_hca
+ * hca_type_show
  */
-static ssize_t i40iw_show_hca(struct device *dev,
-			      struct device_attribute *attr, char *buf)
+static ssize_t hca_type_show(struct device *dev,
+			     struct device_attribute *attr, char *buf)
 {
 	return sprintf(buf, "I40IW\n");
 }
+static DEVICE_ATTR_RO(hca_type);
 
 /**
- * i40iw_show_board
+ * board_id_show
  */
-static ssize_t i40iw_show_board(struct device *dev,
-				struct device_attribute *attr,
-				char *buf)
+static ssize_t board_id_show(struct device *dev,
+			     struct device_attribute *attr, char *buf)
 {
 	return sprintf(buf, "%.*s\n", 32, "I40IW Board ID");
 }
+static DEVICE_ATTR_RO(board_id);
 
-static DEVICE_ATTR(hw_rev, S_IRUGO, i40iw_show_rev, NULL);
-static DEVICE_ATTR(hca_type, S_IRUGO, i40iw_show_hca, NULL);
-static DEVICE_ATTR(board_id, S_IRUGO, i40iw_show_board, NULL);
+static struct attribute *i40iw_dev_attributes[] = {
+	&dev_attr_hw_rev.attr,
+	&dev_attr_hca_type.attr,
+	&dev_attr_board_id.attr,
+	NULL
+};
 
-static struct device_attribute *i40iw_dev_attributes[] = {
-	&dev_attr_hw_rev,
-	&dev_attr_hca_type,
-	&dev_attr_board_id
+static const struct attribute_group i40iw_attr_group = {
+	.attrs = i40iw_dev_attributes,
 };
 
 /**
@@ -2849,20 +2852,6 @@ void i40iw_port_ibevent(struct i40iw_device *iwdev)
 	ib_dispatch_event(&event);
 }
 
-/**
- * i40iw_unregister_rdma_device - unregister of iwarp from IB
- * @iwibdev: rdma device ptr
- */
-static void i40iw_unregister_rdma_device(struct i40iw_ib_device *iwibdev)
-{
-	int i;
-
-	for (i = 0; i < ARRAY_SIZE(i40iw_dev_attributes); ++i)
-		device_remove_file(&iwibdev->ibdev.dev,
-				   i40iw_dev_attributes[i]);
-	ib_unregister_device(&iwibdev->ibdev);
-}
-
 /**
  * i40iw_destroy_rdma_device - destroy rdma device and free resources
  * @iwibdev: IB device ptr
@@ -2872,7 +2861,7 @@ void i40iw_destroy_rdma_device(struct i40iw_ib_device *iwibdev)
 	if (!iwibdev)
 		return;
 
-	i40iw_unregister_rdma_device(iwibdev);
+	ib_unregister_device(&iwibdev->ibdev);
 	kfree(iwibdev->ibdev.iwcm);
 	iwibdev->ibdev.iwcm = NULL;
 	wait_event_timeout(iwibdev->iwdev->close_wq,
@@ -2887,32 +2876,19 @@ void i40iw_destroy_rdma_device(struct i40iw_ib_device *iwibdev)
  */
 int i40iw_register_rdma_device(struct i40iw_device *iwdev)
 {
-	int i, ret;
+	int ret;
 	struct i40iw_ib_device *iwibdev;
 
 	iwdev->iwibdev = i40iw_init_rdma_device(iwdev);
 	if (!iwdev->iwibdev)
 		return -ENOMEM;
 	iwibdev = iwdev->iwibdev;
-
+	rdma_set_device_sysfs_group(&iwibdev->ibdev, &i40iw_attr_group);
 	iwibdev->ibdev.driver_id = RDMA_DRIVER_I40IW;
 	ret = ib_register_device(&iwibdev->ibdev, "i40iw%d", NULL);
 	if (ret)
 		goto error;
 
-	for (i = 0; i < ARRAY_SIZE(i40iw_dev_attributes); ++i) {
-		ret =
-		    device_create_file(&iwibdev->ibdev.dev,
-				       i40iw_dev_attributes[i]);
-		if (ret) {
-			while (i > 0) {
-				i--;
-				device_remove_file(&iwibdev->ibdev.dev, i40iw_dev_attributes[i]);
-			}
-			ib_unregister_device(&iwibdev->ibdev);
-			goto error;
-		}
-	}
 	return 0;
 error:
 	kfree(iwdev->iwibdev->ibdev.iwcm);
diff --git a/drivers/infiniband/hw/mlx4/main.c b/drivers/infiniband/hw/mlx4/main.c
index 81126f18f1fd..0def2323459c 100644
--- a/drivers/infiniband/hw/mlx4/main.c
+++ b/drivers/infiniband/hw/mlx4/main.c
@@ -2039,39 +2039,43 @@ out:
 	return err;
 }
 
-static ssize_t show_hca(struct device *device, struct device_attribute *attr,
-			char *buf)
+static ssize_t hca_type_show(struct device *device,
+			     struct device_attribute *attr, char *buf)
 {
 	struct mlx4_ib_dev *dev =
 		container_of(device, struct mlx4_ib_dev, ib_dev.dev);
 	return sprintf(buf, "MT%d\n", dev->dev->persist->pdev->device);
 }
+static DEVICE_ATTR_RO(hca_type);
 
-static ssize_t show_rev(struct device *device, struct device_attribute *attr,
-			char *buf)
+static ssize_t hw_rev_show(struct device *device,
+			   struct device_attribute *attr, char *buf)
 {
 	struct mlx4_ib_dev *dev =
 		container_of(device, struct mlx4_ib_dev, ib_dev.dev);
 	return sprintf(buf, "%x\n", dev->dev->rev_id);
 }
+static DEVICE_ATTR_RO(hw_rev);
 
-static ssize_t show_board(struct device *device, struct device_attribute *attr,
-			  char *buf)
+static ssize_t board_id_show(struct device *device,
+			     struct device_attribute *attr, char *buf)
 {
 	struct mlx4_ib_dev *dev =
 		container_of(device, struct mlx4_ib_dev, ib_dev.dev);
 	return sprintf(buf, "%.*s\n", MLX4_BOARD_ID_LEN,
 		       dev->dev->board_id);
 }
+static DEVICE_ATTR_RO(board_id);
 
-static DEVICE_ATTR(hw_rev,   S_IRUGO, show_rev,    NULL);
-static DEVICE_ATTR(hca_type, S_IRUGO, show_hca,    NULL);
-static DEVICE_ATTR(board_id, S_IRUGO, show_board,  NULL);
+static struct attribute *mlx4_class_attributes[] = {
+	&dev_attr_hw_rev.attr,
+	&dev_attr_hca_type.attr,
+	&dev_attr_board_id.attr,
+	NULL
+};
 
-static struct device_attribute *mlx4_class_attributes[] = {
-	&dev_attr_hw_rev,
-	&dev_attr_hca_type,
-	&dev_attr_board_id
+static const struct attribute_group mlx4_attr_group = {
+	.attrs = mlx4_class_attributes,
 };
 
 struct diag_counter {
@@ -2803,6 +2807,7 @@ static void *mlx4_ib_add(struct mlx4_dev *dev)
 	if (mlx4_ib_alloc_diag_counters(ibdev))
 		goto err_steer_free_bitmap;
 
+	rdma_set_device_sysfs_group(&ibdev->ib_dev, &mlx4_attr_group);
 	ibdev->ib_dev.driver_id = RDMA_DRIVER_MLX4;
 	if (ib_register_device(&ibdev->ib_dev, "mlx4_%d", NULL))
 		goto err_diag_counters;
@@ -2827,12 +2832,6 @@ static void *mlx4_ib_add(struct mlx4_dev *dev)
 			goto err_notif;
 	}
 
-	for (j = 0; j < ARRAY_SIZE(mlx4_class_attributes); ++j) {
-		if (device_create_file(&ibdev->ib_dev.dev,
-				       mlx4_class_attributes[j]))
-			goto err_notif;
-	}
-
 	ibdev->ib_active = true;
 	mlx4_foreach_port(i, dev, MLX4_PORT_TYPE_IB)
 		devlink_port_type_ib_set(mlx4_get_devlink_port(dev, i),
diff --git a/drivers/infiniband/hw/mlx5/ib_rep.c b/drivers/infiniband/hw/mlx5/ib_rep.c
index 35a0e04c38f2..584ff2ea7810 100644
--- a/drivers/infiniband/hw/mlx5/ib_rep.c
+++ b/drivers/infiniband/hw/mlx5/ib_rep.c
@@ -39,9 +39,6 @@ static const struct mlx5_ib_profile rep_profile = {
 	STAGE_CREATE(MLX5_IB_STAGE_POST_IB_REG_UMR,
 		     mlx5_ib_stage_post_ib_reg_umr_init,
 		     NULL),
-	STAGE_CREATE(MLX5_IB_STAGE_CLASS_ATTR,
-		     mlx5_ib_stage_class_attr_init,
-		     NULL),
 };
 
 static int
diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c
index b3294a7e3ff9..be701d40289e 100644
--- a/drivers/infiniband/hw/mlx5/main.c
+++ b/drivers/infiniband/hw/mlx5/main.c
@@ -4079,16 +4079,17 @@ static int init_node_data(struct mlx5_ib_dev *dev)
 	return mlx5_query_node_guid(dev, &dev->ib_dev.node_guid);
 }
 
-static ssize_t show_fw_pages(struct device *device, struct device_attribute *attr,
-			     char *buf)
+static ssize_t fw_pages_show(struct device *device,
+			     struct device_attribute *attr, char *buf)
 {
 	struct mlx5_ib_dev *dev =
 		container_of(device, struct mlx5_ib_dev, ib_dev.dev);
 
 	return sprintf(buf, "%d\n", dev->mdev->priv.fw_pages);
 }
+static DEVICE_ATTR_RO(fw_pages);
 
-static ssize_t show_reg_pages(struct device *device,
+static ssize_t reg_pages_show(struct device *device,
 			      struct device_attribute *attr, char *buf)
 {
 	struct mlx5_ib_dev *dev =
@@ -4096,44 +4097,47 @@ static ssize_t show_reg_pages(struct device *device,
 
 	return sprintf(buf, "%d\n", atomic_read(&dev->mdev->priv.reg_pages));
 }
+static DEVICE_ATTR_RO(reg_pages);
 
-static ssize_t show_hca(struct device *device, struct device_attribute *attr,
-			char *buf)
+static ssize_t hca_type_show(struct device *device,
+			     struct device_attribute *attr, char *buf)
 {
 	struct mlx5_ib_dev *dev =
 		container_of(device, struct mlx5_ib_dev, ib_dev.dev);
 	return sprintf(buf, "MT%d\n", dev->mdev->pdev->device);
 }
+static DEVICE_ATTR_RO(hca_type);
 
-static ssize_t show_rev(struct device *device, struct device_attribute *attr,
-			char *buf)
+static ssize_t hw_rev_show(struct device *device,
+			   struct device_attribute *attr, char *buf)
 {
 	struct mlx5_ib_dev *dev =
 		container_of(device, struct mlx5_ib_dev, ib_dev.dev);
 	return sprintf(buf, "%x\n", dev->mdev->rev_id);
 }
+static DEVICE_ATTR_RO(hw_rev);
 
-static ssize_t show_board(struct device *device, struct device_attribute *attr,
-			  char *buf)
+static ssize_t board_id_show(struct device *device,
+			     struct device_attribute *attr, char *buf)
 {
 	struct mlx5_ib_dev *dev =
 		container_of(device, struct mlx5_ib_dev, ib_dev.dev);
 	return sprintf(buf, "%.*s\n", MLX5_BOARD_ID_LEN,
 		       dev->mdev->board_id);
 }
+static DEVICE_ATTR_RO(board_id);
 
-static DEVICE_ATTR(hw_rev,   S_IRUGO, show_rev,    NULL);
-static DEVICE_ATTR(hca_type, S_IRUGO, show_hca,    NULL);
-static DEVICE_ATTR(board_id, S_IRUGO, show_board,  NULL);
-static DEVICE_ATTR(fw_pages, S_IRUGO, show_fw_pages, NULL);
-static DEVICE_ATTR(reg_pages, S_IRUGO, show_reg_pages, NULL);
+static struct attribute *mlx5_class_attributes[] = {
+	&dev_attr_hw_rev.attr,
+	&dev_attr_hca_type.attr,
+	&dev_attr_board_id.attr,
+	&dev_attr_fw_pages.attr,
+	&dev_attr_reg_pages.attr,
+	NULL,
+};
 
-static struct device_attribute *mlx5_class_attributes[] = {
-	&dev_attr_hw_rev,
-	&dev_attr_hca_type,
-	&dev_attr_board_id,
-	&dev_attr_fw_pages,
-	&dev_attr_reg_pages,
+static const struct attribute_group mlx5_attr_group = {
+	.attrs = mlx5_class_attributes,
 };
 
 static void pkey_change_handler(struct work_struct *work)
@@ -6112,6 +6116,7 @@ int mlx5_ib_stage_ib_reg_init(struct mlx5_ib_dev *dev)
 {
 	const char *name;
 
+	rdma_set_device_sysfs_group(&dev->ib_dev, &mlx5_attr_group);
 	if (!mlx5_lag_is_active(dev->mdev))
 		name = "mlx5_%d";
 	else
@@ -6146,21 +6151,6 @@ static void mlx5_ib_stage_delay_drop_cleanup(struct mlx5_ib_dev *dev)
 	cancel_delay_drop(dev);
 }
 
-int mlx5_ib_stage_class_attr_init(struct mlx5_ib_dev *dev)
-{
-	int err;
-	int i;
-
-	for (i = 0; i < ARRAY_SIZE(mlx5_class_attributes); i++) {
-		err = device_create_file(&dev->ib_dev.dev,
-					 mlx5_class_attributes[i]);
-		if (err)
-			return err;
-	}
-
-	return 0;
-}
-
 static int mlx5_ib_stage_rep_reg_init(struct mlx5_ib_dev *dev)
 {
 	mlx5_ib_register_vport_reps(dev);
@@ -6268,9 +6258,6 @@ static const struct mlx5_ib_profile pf_profile = {
 	STAGE_CREATE(MLX5_IB_STAGE_DELAY_DROP,
 		     mlx5_ib_stage_delay_drop_init,
 		     mlx5_ib_stage_delay_drop_cleanup),
-	STAGE_CREATE(MLX5_IB_STAGE_CLASS_ATTR,
-		     mlx5_ib_stage_class_attr_init,
-		     NULL),
 };
 
 static const struct mlx5_ib_profile nic_rep_profile = {
@@ -6313,9 +6300,6 @@ static const struct mlx5_ib_profile nic_rep_profile = {
 	STAGE_CREATE(MLX5_IB_STAGE_POST_IB_REG_UMR,
 		     mlx5_ib_stage_post_ib_reg_umr_init,
 		     NULL),
-	STAGE_CREATE(MLX5_IB_STAGE_CLASS_ATTR,
-		     mlx5_ib_stage_class_attr_init,
-		     NULL),
 	STAGE_CREATE(MLX5_IB_STAGE_REP_REG,
 		     mlx5_ib_stage_rep_reg_init,
 		     mlx5_ib_stage_rep_reg_cleanup),
diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h
index e5ec3fdaa4d5..8444ea78229a 100644
--- a/drivers/infiniband/hw/mlx5/mlx5_ib.h
+++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h
@@ -1201,7 +1201,6 @@ void mlx5_ib_stage_pre_ib_reg_umr_cleanup(struct mlx5_ib_dev *dev);
 int mlx5_ib_stage_ib_reg_init(struct mlx5_ib_dev *dev);
 void mlx5_ib_stage_ib_reg_cleanup(struct mlx5_ib_dev *dev);
 int mlx5_ib_stage_post_ib_reg_umr_init(struct mlx5_ib_dev *dev);
-int mlx5_ib_stage_class_attr_init(struct mlx5_ib_dev *dev);
 void __mlx5_ib_remove(struct mlx5_ib_dev *dev,
 		      const struct mlx5_ib_profile *profile,
 		      int stage);
diff --git a/drivers/infiniband/hw/mthca/mthca_provider.c b/drivers/infiniband/hw/mthca/mthca_provider.c
index 7bd7e2ad17e4..691c6f048938 100644
--- a/drivers/infiniband/hw/mthca/mthca_provider.c
+++ b/drivers/infiniband/hw/mthca/mthca_provider.c
@@ -1076,16 +1076,17 @@ static int mthca_unmap_fmr(struct list_head *fmr_list)
 	return err;
 }
 
-static ssize_t show_rev(struct device *device, struct device_attribute *attr,
-			char *buf)
+static ssize_t hw_rev_show(struct device *device,
+			   struct device_attribute *attr, char *buf)
 {
 	struct mthca_dev *dev =
 		container_of(device, struct mthca_dev, ib_dev.dev);
 	return sprintf(buf, "%x\n", dev->rev_id);
 }
+static DEVICE_ATTR_RO(hw_rev);
 
-static ssize_t show_hca(struct device *device, struct device_attribute *attr,
-			char *buf)
+static ssize_t hca_type_show(struct device *device,
+			     struct device_attribute *attr, char *buf)
 {
 	struct mthca_dev *dev =
 		container_of(device, struct mthca_dev, ib_dev.dev);
@@ -1103,23 +1104,26 @@ static ssize_t show_hca(struct device *device, struct device_attribute *attr,
 		return sprintf(buf, "unknown\n");
 	}
 }
+static DEVICE_ATTR_RO(hca_type);
 
-static ssize_t show_board(struct device *device, struct device_attribute *attr,
-			  char *buf)
+static ssize_t board_id_show(struct device *device,
+			     struct device_attribute *attr, char *buf)
 {
 	struct mthca_dev *dev =
 		container_of(device, struct mthca_dev, ib_dev.dev);
 	return sprintf(buf, "%.*s\n", MTHCA_BOARD_ID_LEN, dev->board_id);
 }
+static DEVICE_ATTR_RO(board_id);
 
-static DEVICE_ATTR(hw_rev,   S_IRUGO, show_rev,    NULL);
-static DEVICE_ATTR(hca_type, S_IRUGO, show_hca,    NULL);
-static DEVICE_ATTR(board_id, S_IRUGO, show_board,  NULL);
+static struct attribute *mthca_dev_attributes[] = {
+	&dev_attr_hw_rev.attr,
+	&dev_attr_hca_type.attr,
+	&dev_attr_board_id.attr,
+	NULL
+};
 
-static struct device_attribute *mthca_dev_attributes[] = {
-	&dev_attr_hw_rev,
-	&dev_attr_hca_type,
-	&dev_attr_board_id
+static const struct attribute_group mthca_attr_group = {
+	.attrs = mthca_dev_attributes,
 };
 
 static int mthca_init_node_data(struct mthca_dev *dev)
@@ -1192,7 +1196,6 @@ static void get_dev_fw_str(struct ib_device *device, char *str)
 int mthca_register_device(struct mthca_dev *dev)
 {
 	int ret;
-	int i;
 
 	ret = mthca_init_node_data(dev);
 	if (ret)
@@ -1295,20 +1298,12 @@ int mthca_register_device(struct mthca_dev *dev)
 
 	mutex_init(&dev->cap_mask_mutex);
 
+	rdma_set_device_sysfs_group(&dev->ib_dev, &mthca_attr_group);
 	dev->ib_dev.driver_id = RDMA_DRIVER_MTHCA;
 	ret = ib_register_device(&dev->ib_dev, "mthca%d", NULL);
 	if (ret)
 		return ret;
 
-	for (i = 0; i < ARRAY_SIZE(mthca_dev_attributes); ++i) {
-		ret = device_create_file(&dev->ib_dev.dev,
-					 mthca_dev_attributes[i]);
-		if (ret) {
-			ib_unregister_device(&dev->ib_dev);
-			return ret;
-		}
-	}
-
 	mthca_start_catas_poll(dev);
 
 	return 0;
diff --git a/drivers/infiniband/hw/nes/nes_verbs.c b/drivers/infiniband/hw/nes/nes_verbs.c
index 94054bc611bd..92d1cadd4cfd 100644
--- a/drivers/infiniband/hw/nes/nes_verbs.c
+++ b/drivers/infiniband/hw/nes/nes_verbs.c
@@ -2556,8 +2556,8 @@ static int nes_dereg_mr(struct ib_mr *ib_mr)
 /**
  * show_rev
  */
-static ssize_t show_rev(struct device *dev, struct device_attribute *attr,
-			char *buf)
+static ssize_t hw_rev_show(struct device *dev,
+			   struct device_attribute *attr, char *buf)
 {
 	struct nes_ib_device *nesibdev =
 			container_of(dev, struct nes_ib_device, ibdev.dev);
@@ -2566,40 +2566,40 @@ static ssize_t show_rev(struct device *dev, struct device_attribute *attr,
 	nes_debug(NES_DBG_INIT, "\n");
 	return sprintf(buf, "%x\n", nesvnic->nesdev->nesadapter->hw_rev);
 }
-
+static DEVICE_ATTR_RO(hw_rev);
 
 /**
  * show_hca
  */
-static ssize_t show_hca(struct device *dev, struct device_attribute *attr,
-		        char *buf)
+static ssize_t hca_type_show(struct device *dev,
+			     struct device_attribute *attr, char *buf)
 {
 	nes_debug(NES_DBG_INIT, "\n");
 	return sprintf(buf, "NES020\n");
 }
-
+static DEVICE_ATTR_RO(hca_type);
 
 /**
  * show_board
  */
-static ssize_t show_board(struct device *dev, struct device_attribute *attr,
-			  char *buf)
+static ssize_t board_id_show(struct device *dev,
+			     struct device_attribute *attr, char *buf)
 {
 	nes_debug(NES_DBG_INIT, "\n");
 	return sprintf(buf, "%.*s\n", 32, "NES020 Board ID");
 }
+static DEVICE_ATTR_RO(board_id);
 
-
-static DEVICE_ATTR(hw_rev, S_IRUGO, show_rev, NULL);
-static DEVICE_ATTR(hca_type, S_IRUGO, show_hca, NULL);
-static DEVICE_ATTR(board_id, S_IRUGO, show_board, NULL);
-
-static struct device_attribute *nes_dev_attributes[] = {
-	&dev_attr_hw_rev,
-	&dev_attr_hca_type,
-	&dev_attr_board_id
+static struct attribute *nes_dev_attributes[] = {
+	&dev_attr_hw_rev.attr,
+	&dev_attr_hca_type.attr,
+	&dev_attr_board_id.attr,
+	NULL
 };
 
+static const struct attribute_group nes_attr_group = {
+	.attrs = nes_dev_attributes,
+};
 
 /**
  * nes_query_qp
@@ -3794,8 +3794,9 @@ int nes_register_ofa_device(struct nes_ib_device *nesibdev)
 	struct nes_vnic *nesvnic = nesibdev->nesvnic;
 	struct nes_device *nesdev = nesvnic->nesdev;
 	struct nes_adapter *nesadapter = nesdev->nesadapter;
-	int i, ret;
+	int ret;
 
+	rdma_set_device_sysfs_group(&nesvnic->nesibdev->ibdev, &nes_attr_group);
 	nesvnic->nesibdev->ibdev.driver_id = RDMA_DRIVER_NES;
 	ret = ib_register_device(&nesvnic->nesibdev->ibdev, "nes%d", NULL);
 	if (ret) {
@@ -3808,19 +3809,6 @@ int nes_register_ofa_device(struct nes_ib_device *nesibdev)
 	nesibdev->max_qp = (nesadapter->max_qp-NES_FIRST_QPN) / nesadapter->port_count;
 	nesibdev->max_pd = nesadapter->max_pd / nesadapter->port_count;
 
-	for (i = 0; i < ARRAY_SIZE(nes_dev_attributes); ++i) {
-		ret = device_create_file(&nesibdev->ibdev.dev, nes_dev_attributes[i]);
-		if (ret) {
-			while (i > 0) {
-				i--;
-				device_remove_file(&nesibdev->ibdev.dev,
-						   nes_dev_attributes[i]);
-			}
-			ib_unregister_device(&nesibdev->ibdev);
-			return ret;
-		}
-	}
-
 	nesvnic->of_device_registered = 1;
 
 	return 0;
@@ -3833,15 +3821,9 @@ int nes_register_ofa_device(struct nes_ib_device *nesibdev)
 static void nes_unregister_ofa_device(struct nes_ib_device *nesibdev)
 {
 	struct nes_vnic *nesvnic = nesibdev->nesvnic;
-	int i;
 
-	for (i = 0; i < ARRAY_SIZE(nes_dev_attributes); ++i) {
-		device_remove_file(&nesibdev->ibdev.dev, nes_dev_attributes[i]);
-	}
-
-	if (nesvnic->of_device_registered) {
+	if (nesvnic->of_device_registered)
 		ib_unregister_device(&nesibdev->ibdev);
-	}
 
 	nesvnic->of_device_registered = 0;
 }
diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_main.c b/drivers/infiniband/hw/ocrdma/ocrdma_main.c
index 4d3c27613351..873cc7f6fe61 100644
--- a/drivers/infiniband/hw/ocrdma/ocrdma_main.c
+++ b/drivers/infiniband/hw/ocrdma/ocrdma_main.c
@@ -114,6 +114,35 @@ static void get_dev_fw_str(struct ib_device *device, char *str)
 	snprintf(str, IB_FW_VERSION_NAME_MAX, "%s", &dev->attr.fw_ver[0]);
 }
 
+/* OCRDMA sysfs interface */
+static ssize_t hw_rev_show(struct device *device,
+			   struct device_attribute *attr, char *buf)
+{
+	struct ocrdma_dev *dev = dev_get_drvdata(device);
+
+	return scnprintf(buf, PAGE_SIZE, "0x%x\n", dev->nic_info.pdev->vendor);
+}
+static DEVICE_ATTR_RO(hw_rev);
+
+static ssize_t hca_type_show(struct device *device,
+			     struct device_attribute *attr, char *buf)
+{
+	struct ocrdma_dev *dev = dev_get_drvdata(device);
+
+	return scnprintf(buf, PAGE_SIZE, "%s\n", &dev->model_number[0]);
+}
+static DEVICE_ATTR_RO(hca_type);
+
+static struct attribute *ocrdma_attributes[] = {
+	&dev_attr_hw_rev.attr,
+	&dev_attr_hca_type.attr,
+	NULL
+};
+
+static const struct attribute_group ocrdma_attr_group = {
+	.attrs = ocrdma_attributes,
+};
+
 static int ocrdma_register_device(struct ocrdma_dev *dev)
 {
 	ocrdma_get_guid(dev, (u8 *)&dev->ibdev.node_guid);
@@ -212,6 +241,7 @@ static int ocrdma_register_device(struct ocrdma_dev *dev)
 		dev->ibdev.destroy_srq = ocrdma_destroy_srq;
 		dev->ibdev.post_srq_recv = ocrdma_post_srq_recv;
 	}
+	rdma_set_device_sysfs_group(&dev->ibdev, &ocrdma_attr_group);
 	dev->ibdev.driver_id = RDMA_DRIVER_OCRDMA;
 	return ib_register_device(&dev->ibdev, "ocrdma%d", NULL);
 }
@@ -259,42 +289,9 @@ static void ocrdma_free_resources(struct ocrdma_dev *dev)
 	kfree(dev->cq_tbl);
 }
 
-/* OCRDMA sysfs interface */
-static ssize_t show_rev(struct device *device, struct device_attribute *attr,
-			char *buf)
-{
-	struct ocrdma_dev *dev = dev_get_drvdata(device);
-
-	return scnprintf(buf, PAGE_SIZE, "0x%x\n", dev->nic_info.pdev->vendor);
-}
-
-static ssize_t show_hca_type(struct device *device,
-			     struct device_attribute *attr, char *buf)
-{
-	struct ocrdma_dev *dev = dev_get_drvdata(device);
-
-	return scnprintf(buf, PAGE_SIZE, "%s\n", &dev->model_number[0]);
-}
-
-static DEVICE_ATTR(hw_rev, S_IRUGO, show_rev, NULL);
-static DEVICE_ATTR(hca_type, S_IRUGO, show_hca_type, NULL);
-
-static struct device_attribute *ocrdma_attributes[] = {
-	&dev_attr_hw_rev,
-	&dev_attr_hca_type
-};
-
-static void ocrdma_remove_sysfiles(struct ocrdma_dev *dev)
-{
-	int i;
-
-	for (i = 0; i < ARRAY_SIZE(ocrdma_attributes); i++)
-		device_remove_file(&dev->ibdev.dev, ocrdma_attributes[i]);
-}
-
 static struct ocrdma_dev *ocrdma_add(struct be_dev_info *dev_info)
 {
-	int status = 0, i;
+	int status = 0;
 	u8 lstate = 0;
 	struct ocrdma_dev *dev;
 
@@ -330,9 +327,6 @@ static struct ocrdma_dev *ocrdma_add(struct be_dev_info *dev_info)
 	if (!status)
 		ocrdma_update_link_state(dev, lstate);
 
-	for (i = 0; i < ARRAY_SIZE(ocrdma_attributes); i++)
-		if (device_create_file(&dev->ibdev.dev, ocrdma_attributes[i]))
-			goto sysfs_err;
 	/* Init stats */
 	ocrdma_add_port_stats(dev);
 	/* Interrupt Moderation */
@@ -347,8 +341,6 @@ static struct ocrdma_dev *ocrdma_add(struct be_dev_info *dev_info)
 		dev_name(&dev->nic_info.pdev->dev), dev->id);
 	return dev;
 
-sysfs_err:
-	ocrdma_remove_sysfiles(dev);
 alloc_err:
 	ocrdma_free_resources(dev);
 	ocrdma_cleanup_hw(dev);
@@ -375,7 +367,6 @@ static void ocrdma_remove(struct ocrdma_dev *dev)
 	 * of the registered clients.
 	 */
 	cancel_delayed_work_sync(&dev->eqd_work);
-	ocrdma_remove_sysfiles(dev);
 	ib_unregister_device(&dev->ibdev);
 
 	ocrdma_rem_port_stats(dev);
diff --git a/drivers/infiniband/hw/qedr/main.c b/drivers/infiniband/hw/qedr/main.c
index cd7b8b39a129..8d6ff9df49fe 100644
--- a/drivers/infiniband/hw/qedr/main.c
+++ b/drivers/infiniband/hw/qedr/main.c
@@ -133,6 +133,33 @@ static int qedr_iw_port_immutable(struct ib_device *ibdev, u8 port_num,
 	return 0;
 }
 
+/* QEDR sysfs interface */
+static ssize_t hw_rev_show(struct device *device, struct device_attribute *attr,
+			   char *buf)
+{
+	struct qedr_dev *dev = dev_get_drvdata(device);
+
+	return scnprintf(buf, PAGE_SIZE, "0x%x\n", dev->pdev->vendor);
+}
+static DEVICE_ATTR_RO(hw_rev);
+
+static ssize_t hca_type_show(struct device *device,
+			     struct device_attribute *attr, char *buf)
+{
+	return scnprintf(buf, PAGE_SIZE, "%s\n", "HCA_TYPE_TO_SET");
+}
+static DEVICE_ATTR_RO(hca_type);
+
+static struct attribute *qedr_attributes[] = {
+	&dev_attr_hw_rev.attr,
+	&dev_attr_hca_type.attr,
+	NULL
+};
+
+static const struct attribute_group qedr_attr_group = {
+	.attrs = qedr_attributes,
+};
+
 static int qedr_iw_register_device(struct qedr_dev *dev)
 {
 	dev->ibdev.node_type = RDMA_NODE_RNIC;
@@ -260,7 +287,7 @@ static int qedr_register_device(struct qedr_dev *dev)
 
 	dev->ibdev.get_link_layer = qedr_link_layer;
 	dev->ibdev.get_dev_fw_str = qedr_get_dev_fw_str;
-
+	rdma_set_device_sysfs_group(&dev->ibdev, &qedr_attr_group);
 	dev->ibdev.driver_id = RDMA_DRIVER_QEDR;
 	return ib_register_device(&dev->ibdev, "qedr%d", NULL);
 }
@@ -402,37 +429,6 @@ err1:
 	return rc;
 }
 
-/* QEDR sysfs interface */
-static ssize_t show_rev(struct device *device, struct device_attribute *attr,
-			char *buf)
-{
-	struct qedr_dev *dev = dev_get_drvdata(device);
-
-	return scnprintf(buf, PAGE_SIZE, "0x%x\n", dev->pdev->vendor);
-}
-
-static ssize_t show_hca_type(struct device *device,
-			     struct device_attribute *attr, char *buf)
-{
-	return scnprintf(buf, PAGE_SIZE, "%s\n", "HCA_TYPE_TO_SET");
-}
-
-static DEVICE_ATTR(hw_rev, S_IRUGO, show_rev, NULL);
-static DEVICE_ATTR(hca_type, S_IRUGO, show_hca_type, NULL);
-
-static struct device_attribute *qedr_attributes[] = {
-	&dev_attr_hw_rev,
-	&dev_attr_hca_type
-};
-
-static void qedr_remove_sysfiles(struct qedr_dev *dev)
-{
-	int i;
-
-	for (i = 0; i < ARRAY_SIZE(qedr_attributes); i++)
-		device_remove_file(&dev->ibdev.dev, qedr_attributes[i]);
-}
-
 static void qedr_pci_set_atomic(struct qedr_dev *dev, struct pci_dev *pdev)
 {
 	int rc = pci_enable_atomic_ops_to_root(pdev,
@@ -853,7 +849,7 @@ static struct qedr_dev *qedr_add(struct qed_dev *cdev, struct pci_dev *pdev,
 {
 	struct qed_dev_rdma_info dev_info;
 	struct qedr_dev *dev;
-	int rc = 0, i;
+	int rc = 0;
 
 	dev = (struct qedr_dev *)ib_alloc_device(sizeof(*dev));
 	if (!dev) {
@@ -912,18 +908,12 @@ static struct qedr_dev *qedr_add(struct qed_dev *cdev, struct pci_dev *pdev,
 		goto reg_err;
 	}
 
-	for (i = 0; i < ARRAY_SIZE(qedr_attributes); i++)
-		if (device_create_file(&dev->ibdev.dev, qedr_attributes[i]))
-			goto sysfs_err;
-
 	if (!test_and_set_bit(QEDR_ENET_STATE_BIT, &dev->enet_state))
 		qedr_ib_dispatch_event(dev, QEDR_PORT, IB_EVENT_PORT_ACTIVE);
 
 	DP_DEBUG(dev, QEDR_MSG_INIT, "qedr driver loaded successfully\n");
 	return dev;
 
-sysfs_err:
-	ib_unregister_device(&dev->ibdev);
 reg_err:
 	qedr_sync_free_irqs(dev);
 irq_err:
@@ -942,7 +932,6 @@ static void qedr_remove(struct qedr_dev *dev)
 	/* First unregister with stack to stop all the active traffic
 	 * of the registered clients.
 	 */
-	qedr_remove_sysfiles(dev);
 	ib_unregister_device(&dev->ibdev);
 
 	qedr_stop_hw(dev);
diff --git a/drivers/infiniband/hw/qib/qib.h b/drivers/infiniband/hw/qib/qib.h
index 3461df002f81..83d2349188db 100644
--- a/drivers/infiniband/hw/qib/qib.h
+++ b/drivers/infiniband/hw/qib/qib.h
@@ -1390,13 +1390,13 @@ static inline u32 qib_get_hdrqtail(const struct qib_ctxtdata *rcd)
  */
 
 extern const char ib_qib_version[];
+extern const struct attribute_group qib_attr_group;
 
 int qib_device_create(struct qib_devdata *);
 void qib_device_remove(struct qib_devdata *);
 
 int qib_create_port_files(struct ib_device *ibdev, u8 port_num,
 			  struct kobject *kobj);
-int qib_verbs_register_sysfs(struct qib_devdata *);
 void qib_verbs_unregister_sysfs(struct qib_devdata *);
 /* Hook for sysfs read of QSFP */
 extern int qib_qsfp_dump(struct qib_pportdata *ppd, char *buf, int len);
diff --git a/drivers/infiniband/hw/qib/qib_sysfs.c b/drivers/infiniband/hw/qib/qib_sysfs.c
index ca2638d8f35e..1cf4ca3f23e3 100644
--- a/drivers/infiniband/hw/qib/qib_sysfs.c
+++ b/drivers/infiniband/hw/qib/qib_sysfs.c
@@ -551,17 +551,18 @@ static struct kobj_type qib_diagc_ktype = {
  * Start of per-unit (or driver, in some cases, but replicated
  * per unit) functions (these get a device *)
  */
-static ssize_t show_rev(struct device *device, struct device_attribute *attr,
-			char *buf)
+static ssize_t hw_rev_show(struct device *device, struct device_attribute *attr,
+			   char *buf)
 {
 	struct qib_ibdev *dev =
 		container_of(device, struct qib_ibdev, rdi.ibdev.dev);
 
 	return sprintf(buf, "%x\n", dd_from_dev(dev)->minrev);
 }
+static DEVICE_ATTR_RO(hw_rev);
 
-static ssize_t show_hca(struct device *device, struct device_attribute *attr,
-			char *buf)
+static ssize_t hca_type_show(struct device *device,
+			     struct device_attribute *attr, char *buf)
 {
 	struct qib_ibdev *dev =
 		container_of(device, struct qib_ibdev, rdi.ibdev.dev);
@@ -574,15 +575,18 @@ static ssize_t show_hca(struct device *device, struct device_attribute *attr,
 		ret = scnprintf(buf, PAGE_SIZE, "%s\n", dd->boardname);
 	return ret;
 }
+static DEVICE_ATTR_RO(hca_type);
+static DEVICE_ATTR(board_id, 0444, hca_type_show, NULL);
 
-static ssize_t show_version(struct device *device,
+static ssize_t version_show(struct device *device,
 			    struct device_attribute *attr, char *buf)
 {
 	/* The string printed here is already newline-terminated. */
 	return scnprintf(buf, PAGE_SIZE, "%s", (char *)ib_qib_version);
 }
+static DEVICE_ATTR_RO(version);
 
-static ssize_t show_boardversion(struct device *device,
+static ssize_t boardversion_show(struct device *device,
 				 struct device_attribute *attr, char *buf)
 {
 	struct qib_ibdev *dev =
@@ -592,9 +596,9 @@ static ssize_t show_boardversion(struct device *device,
 	/* The string printed here is already newline-terminated. */
 	return scnprintf(buf, PAGE_SIZE, "%s", dd->boardversion);
 }
+static DEVICE_ATTR_RO(boardversion);
 
-
-static ssize_t show_localbus_info(struct device *device,
+static ssize_t localbus_info_show(struct device *device,
 				  struct device_attribute *attr, char *buf)
 {
 	struct qib_ibdev *dev =
@@ -604,9 +608,9 @@ static ssize_t show_localbus_info(struct device *device,
 	/* The string printed here is already newline-terminated. */
 	return scnprintf(buf, PAGE_SIZE, "%s", dd->lbus_info);
 }
+static DEVICE_ATTR_RO(localbus_info);
 
-
-static ssize_t show_nctxts(struct device *device,
+static ssize_t nctxts_show(struct device *device,
 			   struct device_attribute *attr, char *buf)
 {
 	struct qib_ibdev *dev =
@@ -620,9 +624,10 @@ static ssize_t show_nctxts(struct device *device,
 			(dd->first_user_ctxt > dd->cfgctxts) ? 0 :
 			(dd->cfgctxts - dd->first_user_ctxt));
 }
+static DEVICE_ATTR_RO(nctxts);
 
-static ssize_t show_nfreectxts(struct device *device,
-			   struct device_attribute *attr, char *buf)
+static ssize_t nfreectxts_show(struct device *device,
+			       struct device_attribute *attr, char *buf)
 {
 	struct qib_ibdev *dev =
 		container_of(device, struct qib_ibdev, rdi.ibdev.dev);
@@ -631,8 +636,9 @@ static ssize_t show_nfreectxts(struct device *device,
 	/* Return the number of free user ports (contexts) available. */
 	return scnprintf(buf, PAGE_SIZE, "%u\n", dd->freectxts);
 }
+static DEVICE_ATTR_RO(nfreectxts);
 
-static ssize_t show_serial(struct device *device,
+static ssize_t serial_show(struct device *device,
 			   struct device_attribute *attr, char *buf)
 {
 	struct qib_ibdev *dev =
@@ -644,8 +650,9 @@ static ssize_t show_serial(struct device *device,
 	strcat(buf, "\n");
 	return strlen(buf);
 }
+static DEVICE_ATTR_RO(serial);
 
-static ssize_t store_chip_reset(struct device *device,
+static ssize_t chip_reset_store(struct device *device,
 				struct device_attribute *attr, const char *buf,
 				size_t count)
 {
@@ -663,11 +670,12 @@ static ssize_t store_chip_reset(struct device *device,
 bail:
 	return ret < 0 ? ret : count;
 }
+static DEVICE_ATTR_WO(chip_reset);
 
 /*
  * Dump tempsense regs. in decimal, to ease shell-scripts.
  */
-static ssize_t show_tempsense(struct device *device,
+static ssize_t tempsense_show(struct device *device,
 			      struct device_attribute *attr, char *buf)
 {
 	struct qib_ibdev *dev =
@@ -695,6 +703,7 @@ static ssize_t show_tempsense(struct device *device,
 				*(signed char *)(regvals + 7));
 	return ret;
 }
+static DEVICE_ATTR_RO(tempsense);
 
 /*
  * end of per-unit (or driver, in some cases, but replicated
@@ -702,30 +711,23 @@ static ssize_t show_tempsense(struct device *device,
  */
 
 /* start of per-unit file structures and support code */
-static DEVICE_ATTR(hw_rev, S_IRUGO, show_rev, NULL);
-static DEVICE_ATTR(hca_type, S_IRUGO, show_hca, NULL);
-static DEVICE_ATTR(board_id, S_IRUGO, show_hca, NULL);
-static DEVICE_ATTR(version, S_IRUGO, show_version, NULL);
-static DEVICE_ATTR(nctxts, S_IRUGO, show_nctxts, NULL);
-static DEVICE_ATTR(nfreectxts, S_IRUGO, show_nfreectxts, NULL);
-static DEVICE_ATTR(serial, S_IRUGO, show_serial, NULL);
-static DEVICE_ATTR(boardversion, S_IRUGO, show_boardversion, NULL);
-static DEVICE_ATTR(tempsense, S_IRUGO, show_tempsense, NULL);
-static DEVICE_ATTR(localbus_info, S_IRUGO, show_localbus_info, NULL);
-static DEVICE_ATTR(chip_reset, S_IWUSR, NULL, store_chip_reset);
-
-static struct device_attribute *qib_attributes[] = {
-	&dev_attr_hw_rev,
-	&dev_attr_hca_type,
-	&dev_attr_board_id,
-	&dev_attr_version,
-	&dev_attr_nctxts,
-	&dev_attr_nfreectxts,
-	&dev_attr_serial,
-	&dev_attr_boardversion,
-	&dev_attr_tempsense,
-	&dev_attr_localbus_info,
-	&dev_attr_chip_reset,
+static struct attribute *qib_attributes[] = {
+	&dev_attr_hw_rev.attr,
+	&dev_attr_hca_type.attr,
+	&dev_attr_board_id.attr,
+	&dev_attr_version.attr,
+	&dev_attr_nctxts.attr,
+	&dev_attr_nfreectxts.attr,
+	&dev_attr_serial.attr,
+	&dev_attr_boardversion.attr,
+	&dev_attr_tempsense.attr,
+	&dev_attr_localbus_info.attr,
+	&dev_attr_chip_reset.attr,
+	NULL,
+};
+
+const struct attribute_group qib_attr_group = {
+	.attrs = qib_attributes,
 };
 
 int qib_create_port_files(struct ib_device *ibdev, u8 port_num,
@@ -826,27 +828,6 @@ bail:
 	return ret;
 }
 
-/*
- * Register and create our files in /sys/class/infiniband.
- */
-int qib_verbs_register_sysfs(struct qib_devdata *dd)
-{
-	struct ib_device *dev = &dd->verbs_dev.rdi.ibdev;
-	int i, ret;
-
-	for (i = 0; i < ARRAY_SIZE(qib_attributes); ++i) {
-		ret = device_create_file(&dev->dev, qib_attributes[i]);
-		if (ret)
-			goto bail;
-	}
-
-	return 0;
-bail:
-	for (i = 0; i < ARRAY_SIZE(qib_attributes); ++i)
-		device_remove_file(&dev->dev, qib_attributes[i]);
-	return ret;
-}
-
 /*
  * Unregister and remove our files in /sys/class/infiniband.
  */
diff --git a/drivers/infiniband/hw/qib/qib_verbs.c b/drivers/infiniband/hw/qib/qib_verbs.c
index 8a45964c4700..4b0f5761a646 100644
--- a/drivers/infiniband/hw/qib/qib_verbs.c
+++ b/drivers/infiniband/hw/qib/qib_verbs.c
@@ -1625,19 +1625,14 @@ int qib_register_ib_device(struct qib_devdata *dd)
 			      i,
 			      dd->rcd[ctxt]->pkeys);
 	}
+	rdma_set_device_sysfs_group(&dd->verbs_dev.rdi.ibdev, &qib_attr_group);
 
 	ret = rvt_register_device(&dd->verbs_dev.rdi, RDMA_DRIVER_QIB);
 	if (ret)
 		goto err_tx;
 
-	ret = qib_verbs_register_sysfs(dd);
-	if (ret)
-		goto err_class;
-
 	return ret;
 
-err_class:
-	rvt_unregister_device(&dd->verbs_dev.rdi);
 err_tx:
 	while (!list_empty(&dev->txreq_free)) {
 		struct list_head *l = dev->txreq_free.next;
diff --git a/drivers/infiniband/hw/usnic/usnic_ib_main.c b/drivers/infiniband/hw/usnic/usnic_ib_main.c
index e3f98fe2f1e6..73bd00f8d2c8 100644
--- a/drivers/infiniband/hw/usnic/usnic_ib_main.c
+++ b/drivers/infiniband/hw/usnic/usnic_ib_main.c
@@ -417,6 +417,8 @@ static void *usnic_ib_device_add(struct pci_dev *dev)
 
 
 	us_ibdev->ib_dev.driver_id = RDMA_DRIVER_USNIC;
+	rdma_set_device_sysfs_group(&us_ibdev->ib_dev, &usnic_attr_group);
+
 	if (ib_register_device(&us_ibdev->ib_dev, "usnic_%d", NULL))
 		goto err_fwd_dealloc;
 
diff --git a/drivers/infiniband/hw/usnic/usnic_ib_sysfs.c b/drivers/infiniband/hw/usnic/usnic_ib_sysfs.c
index fab4cb780122..a7e4b2ccfaf8 100644
--- a/drivers/infiniband/hw/usnic/usnic_ib_sysfs.c
+++ b/drivers/infiniband/hw/usnic/usnic_ib_sysfs.c
@@ -46,9 +46,8 @@
 #include "usnic_ib_sysfs.h"
 #include "usnic_log.h"
 
-static ssize_t usnic_ib_show_board(struct device *device,
-					struct device_attribute *attr,
-					char *buf)
+static ssize_t board_id_show(struct device *device,
+			     struct device_attribute *attr, char *buf)
 {
 	struct usnic_ib_dev *us_ibdev =
 		container_of(device, struct usnic_ib_dev, ib_dev.dev);
@@ -60,13 +59,13 @@ static ssize_t usnic_ib_show_board(struct device *device,
 
 	return scnprintf(buf, PAGE_SIZE, "%hu\n", subsystem_device_id);
 }
+static DEVICE_ATTR_RO(board_id);
 
 /*
  * Report the configuration for this PF
  */
 static ssize_t
-usnic_ib_show_config(struct device *device, struct device_attribute *attr,
-			char *buf)
+config_show(struct device *device, struct device_attribute *attr, char *buf)
 {
 	struct usnic_ib_dev *us_ibdev;
 	char *ptr;
@@ -126,10 +125,10 @@ usnic_ib_show_config(struct device *device, struct device_attribute *attr,
 
 	return ptr - buf;
 }
+static DEVICE_ATTR_RO(config);
 
 static ssize_t
-usnic_ib_show_iface(struct device *device, struct device_attribute *attr,
-			char *buf)
+iface_show(struct device *device, struct device_attribute *attr, char *buf)
 {
 	struct usnic_ib_dev *us_ibdev;
 
@@ -138,10 +137,10 @@ usnic_ib_show_iface(struct device *device, struct device_attribute *attr,
 	return scnprintf(buf, PAGE_SIZE, "%s\n",
 			netdev_name(us_ibdev->netdev));
 }
+static DEVICE_ATTR_RO(iface);
 
 static ssize_t
-usnic_ib_show_max_vf(struct device *device, struct device_attribute *attr,
-			char *buf)
+max_vf_show(struct device *device, struct device_attribute *attr, char *buf)
 {
 	struct usnic_ib_dev *us_ibdev;
 
@@ -150,10 +149,10 @@ usnic_ib_show_max_vf(struct device *device, struct device_attribute *attr,
 	return scnprintf(buf, PAGE_SIZE, "%u\n",
 			kref_read(&us_ibdev->vf_cnt));
 }
+static DEVICE_ATTR_RO(max_vf);
 
 static ssize_t
-usnic_ib_show_qp_per_vf(struct device *device, struct device_attribute *attr,
-			char *buf)
+qp_per_vf_show(struct device *device, struct device_attribute *attr, char *buf)
 {
 	struct usnic_ib_dev *us_ibdev;
 	int qp_per_vf;
@@ -165,10 +164,10 @@ usnic_ib_show_qp_per_vf(struct device *device, struct device_attribute *attr,
 	return scnprintf(buf, PAGE_SIZE,
 				"%d\n", qp_per_vf);
 }
+static DEVICE_ATTR_RO(qp_per_vf);
 
 static ssize_t
-usnic_ib_show_cq_per_vf(struct device *device, struct device_attribute *attr,
-			char *buf)
+cq_per_vf_show(struct device *device, struct device_attribute *attr, char *buf)
 {
 	struct usnic_ib_dev *us_ibdev;
 
@@ -177,21 +176,20 @@ usnic_ib_show_cq_per_vf(struct device *device, struct device_attribute *attr,
 	return scnprintf(buf, PAGE_SIZE, "%d\n",
 			us_ibdev->vf_res_cnt[USNIC_VNIC_RES_TYPE_CQ]);
 }
+static DEVICE_ATTR_RO(cq_per_vf);
+
+static struct attribute *usnic_class_attributes[] = {
+	&dev_attr_board_id.attr,
+	&dev_attr_config.attr,
+	&dev_attr_iface.attr,
+	&dev_attr_max_vf.attr,
+	&dev_attr_qp_per_vf.attr,
+	&dev_attr_cq_per_vf.attr,
+	NULL
+};
 
-static DEVICE_ATTR(board_id, S_IRUGO, usnic_ib_show_board, NULL);
-static DEVICE_ATTR(config, S_IRUGO, usnic_ib_show_config, NULL);
-static DEVICE_ATTR(iface, S_IRUGO, usnic_ib_show_iface, NULL);
-static DEVICE_ATTR(max_vf, S_IRUGO, usnic_ib_show_max_vf, NULL);
-static DEVICE_ATTR(qp_per_vf, S_IRUGO, usnic_ib_show_qp_per_vf, NULL);
-static DEVICE_ATTR(cq_per_vf, S_IRUGO, usnic_ib_show_cq_per_vf, NULL);
-
-static struct device_attribute *usnic_class_attributes[] = {
-	&dev_attr_board_id,
-	&dev_attr_config,
-	&dev_attr_iface,
-	&dev_attr_max_vf,
-	&dev_attr_qp_per_vf,
-	&dev_attr_cq_per_vf,
+const struct attribute_group usnic_attr_group = {
+	.attrs = usnic_class_attributes,
 };
 
 struct qpn_attribute {
@@ -278,18 +276,6 @@ static struct kobj_type usnic_ib_qpn_type = {
 
 int usnic_ib_sysfs_register_usdev(struct usnic_ib_dev *us_ibdev)
 {
-	int i;
-	int err;
-	for (i = 0; i < ARRAY_SIZE(usnic_class_attributes); ++i) {
-		err = device_create_file(&us_ibdev->ib_dev.dev,
-						usnic_class_attributes[i]);
-		if (err) {
-			usnic_err("Failed to create device file %d for %s eith err %d",
-				i, dev_name(&us_ibdev->ib_dev.dev), err);
-			return -EINVAL;
-		}
-	}
-
 	/* create kernel object for looking at individual QPs */
 	kobject_get(&us_ibdev->ib_dev.dev.kobj);
 	us_ibdev->qpn_kobj = kobject_create_and_add("qpn",
@@ -304,12 +290,6 @@ int usnic_ib_sysfs_register_usdev(struct usnic_ib_dev *us_ibdev)
 
 void usnic_ib_sysfs_unregister_usdev(struct usnic_ib_dev *us_ibdev)
 {
-	int i;
-	for (i = 0; i < ARRAY_SIZE(usnic_class_attributes); ++i) {
-		device_remove_file(&us_ibdev->ib_dev.dev,
-					usnic_class_attributes[i]);
-	}
-
 	kobject_put(us_ibdev->qpn_kobj);
 }
 
diff --git a/drivers/infiniband/hw/usnic/usnic_ib_sysfs.h b/drivers/infiniband/hw/usnic/usnic_ib_sysfs.h
index 3d98e16cfeaf..b1f064cec850 100644
--- a/drivers/infiniband/hw/usnic/usnic_ib_sysfs.h
+++ b/drivers/infiniband/hw/usnic/usnic_ib_sysfs.h
@@ -41,4 +41,6 @@ void usnic_ib_sysfs_unregister_usdev(struct usnic_ib_dev *us_ibdev);
 void usnic_ib_sysfs_qpn_add(struct usnic_ib_qp_grp *qp_grp);
 void usnic_ib_sysfs_qpn_remove(struct usnic_ib_qp_grp *qp_grp);
 
+extern const struct attribute_group usnic_attr_group;
+
 #endif /* !USNIC_IB_SYSFS_H_ */
diff --git a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_main.c b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_main.c
index c1e31985b11c..398443f43dc3 100644
--- a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_main.c
+++ b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_main.c
@@ -65,32 +65,36 @@ static struct workqueue_struct *event_wq;
 static int pvrdma_add_gid(const struct ib_gid_attr *attr, void **context);
 static int pvrdma_del_gid(const struct ib_gid_attr *attr, void **context);
 
-static ssize_t show_hca(struct device *device, struct device_attribute *attr,
-			char *buf)
+static ssize_t hca_type_show(struct device *device,
+			     struct device_attribute *attr, char *buf)
 {
 	return sprintf(buf, "VMW_PVRDMA-%s\n", DRV_VERSION);
 }
+static DEVICE_ATTR_RO(hca_type);
 
-static ssize_t show_rev(struct device *device, struct device_attribute *attr,
-			char *buf)
+static ssize_t hw_rev_show(struct device *device,
+			   struct device_attribute *attr, char *buf)
 {
 	return sprintf(buf, "%d\n", PVRDMA_REV_ID);
 }
+static DEVICE_ATTR_RO(hw_rev);
 
-static ssize_t show_board(struct device *device, struct device_attribute *attr,
-			  char *buf)
+static ssize_t board_id_show(struct device *device,
+			     struct device_attribute *attr, char *buf)
 {
 	return sprintf(buf, "%d\n", PVRDMA_BOARD_ID);
 }
+static DEVICE_ATTR_RO(board_id);
 
-static DEVICE_ATTR(hw_rev,   S_IRUGO, show_rev,	   NULL);
-static DEVICE_ATTR(hca_type, S_IRUGO, show_hca,	   NULL);
-static DEVICE_ATTR(board_id, S_IRUGO, show_board,  NULL);
+static struct attribute *pvrdma_class_attributes[] = {
+	&dev_attr_hw_rev.attr,
+	&dev_attr_hca_type.attr,
+	&dev_attr_board_id.attr,
+	NULL,
+};
 
-static struct device_attribute *pvrdma_class_attributes[] = {
-	&dev_attr_hw_rev,
-	&dev_attr_hca_type,
-	&dev_attr_board_id
+static const struct attribute_group pvrdma_attr_group = {
+	.attrs = pvrdma_class_attributes,
 };
 
 static void pvrdma_get_fw_ver_str(struct ib_device *device, char *str)
@@ -160,7 +164,6 @@ static struct net_device *pvrdma_get_netdev(struct ib_device *ibdev,
 static int pvrdma_register_device(struct pvrdma_dev *dev)
 {
 	int ret = -1;
-	int i = 0;
 
 	dev->ib_dev.node_guid = dev->dsr->caps.node_guid;
 	dev->sys_image_guid = dev->dsr->caps.sys_image_guid;
@@ -265,24 +268,16 @@ static int pvrdma_register_device(struct pvrdma_dev *dev)
 	}
 	dev->ib_dev.driver_id = RDMA_DRIVER_VMW_PVRDMA;
 	spin_lock_init(&dev->srq_tbl_lock);
+	rdma_set_device_sysfs_group(&dev->ib_dev, &pvrdma_attr_group);
 
 	ret = ib_register_device(&dev->ib_dev, "vmw_pvrdma%d", NULL);
 	if (ret)
 		goto err_srq_free;
 
-	for (i = 0; i < ARRAY_SIZE(pvrdma_class_attributes); ++i) {
-		ret = device_create_file(&dev->ib_dev.dev,
-					 pvrdma_class_attributes[i]);
-		if (ret)
-			goto err_class;
-	}
-
 	dev->ib_active = true;
 
 	return 0;
 
-err_class:
-	ib_unregister_device(&dev->ib_dev);
 err_srq_free:
 	kfree(dev->srq_tbl);
 err_qp_free:
diff --git a/drivers/infiniband/sw/rxe/rxe_verbs.c b/drivers/infiniband/sw/rxe/rxe_verbs.c
index e4da5b671e4a..9c19f2027511 100644
--- a/drivers/infiniband/sw/rxe/rxe_verbs.c
+++ b/drivers/infiniband/sw/rxe/rxe_verbs.c
@@ -1148,14 +1148,18 @@ static ssize_t parent_show(struct device *device,
 
 static DEVICE_ATTR_RO(parent);
 
-static struct device_attribute *rxe_dev_attributes[] = {
-	&dev_attr_parent,
+static struct attribute *rxe_dev_attributes[] = {
+	&dev_attr_parent.attr,
+	NULL
+};
+
+static const struct attribute_group rxe_attr_group = {
+	.attrs = rxe_dev_attributes,
 };
 
 int rxe_register_device(struct rxe_dev *rxe)
 {
 	int err;
-	int i;
 	struct ib_device *dev = &rxe->ib_dev;
 	struct crypto_shash *tfm;
 
@@ -1259,6 +1263,7 @@ int rxe_register_device(struct rxe_dev *rxe)
 	}
 	rxe->tfm = tfm;
 
+	rdma_set_device_sysfs_group(dev, &rxe_attr_group);
 	dev->driver_id = RDMA_DRIVER_RXE;
 	err = ib_register_device(dev, "rxe%d", NULL);
 	if (err) {
@@ -1266,19 +1271,8 @@ int rxe_register_device(struct rxe_dev *rxe)
 		goto err1;
 	}
 
-	for (i = 0; i < ARRAY_SIZE(rxe_dev_attributes); ++i) {
-		err = device_create_file(&dev->dev, rxe_dev_attributes[i]);
-		if (err) {
-			pr_warn("%s failed with error %d for attr number %d\n",
-				__func__, err, i);
-			goto err2;
-		}
-	}
-
 	return 0;
 
-err2:
-	ib_unregister_device(dev);
 err1:
 	crypto_free_shash(rxe->tfm);
 
@@ -1287,12 +1281,8 @@ err1:
 
 int rxe_unregister_device(struct rxe_dev *rxe)
 {
-	int i;
 	struct ib_device *dev = &rxe->ib_dev;
 
-	for (i = 0; i < ARRAY_SIZE(rxe_dev_attributes); ++i)
-		device_remove_file(&dev->dev, rxe_dev_attributes[i]);
-
 	ib_unregister_device(dev);
 
 	return 0;
-- 
cgit v1.2.3-55-g7522