RDMA/rxe: Fix qp reference counting for atomic ops

[ Upstream commit 15ae137 ]

Currently the rdma_rxe driver attempts to protect atomic responder
resources by taking a reference to the qp which is only freed when the
resource is recycled for a new read or atomic operation. This means that
in normal circumstances there is almost always an extra qp reference once
an atomic operation has been executed which prevents cleaning up the qp
and associated pd and cqs when the qp is destroyed.

This patch removes the call to rxe_add_ref() in send_atomic_ack() and the
call to rxe_drop_ref() in free_rd_atomic_resource(). If the qp is
destroyed while a peer is retrying an atomic op it will cause the
operation to fail which is acceptable.

Link: https://lore.kernel.org/r/20210604230558.4812-1-rpearsonhpe@gmail.com
Reported-by: Zhu Yanjun <zyjzyj2000@gmail.com>
Fixes: 86af617 ("IB/rxe: remove unnecessary skb_clone")
Signed-off-by: Bob Pearson <rpearsonhpe@gmail.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drivers/infiniband/sw/rxe/rxe_qp.c

@@ -125,7 +125,6 @@ static void free_rd_atomic_resources(struct rxe_qp *qp)
 void free_rd_atomic_resource(struct rxe_qp *qp, struct resp_res *res)
 {
 	if (res->type == RXE_ATOMIC_MASK) {
-		rxe_drop_ref(qp);
 		kfree_skb(res->atomic.skb);
 	} else if (res->type == RXE_READ_MASK) {
 		if (res->read.mr)

drivers/infiniband/sw/rxe/rxe_resp.c

@@ -966,8 +966,6 @@ static int send_atomic_ack(struct rxe_qp *qp, struct rxe_pkt_info *pkt,
 		goto out;
 	}
 
-	rxe_add_ref(qp);
-
 	res = &qp->resp.resources[qp->resp.res_head];
 	free_rd_atomic_resource(qp, res);
 	rxe_advance_resp_resource(qp);