vmw_pvrdma: Avoid double unlock on qp->sq.lock

[ Upstream commit d8a052e ]

qp->sq.lock is currently unlocked twice when the ring state is invalid.
Fix this by removing the goto, and also set bad_wr correctly.

Fixes: 36756a6 ("libpvrdma: Add queue pair functions")
Cc: stable@linux-rdma.org
Reviewed-by: Vishnu Dasa <vdasa@vmware.com>
Reviewed-by: Adit Ranadive <aditr@vmware.com>
Reviewed-by: Rajesh Jalisatgi <rjalisatgi@vmware.com>
Signed-off-by: Bryan Tan <bryantan@vmware.com>
Signed-off-by: Nicolas Morey-Chaisemartin <nmoreychaisemartin@suse.com>

providers/vmw_pvrdma/qp.c

@@ -459,11 +459,12 @@ int pvrdma_post_send(struct ibv_qp *ibqp, struct ibv_send_wr *wr,
 	}
 
 	pthread_spin_lock(&qp->sq.lock);
+
 	ind = pvrdma_idx(&(qp->sq.ring_state->prod_tail), qp->sq.wqe_cnt);
 	if (ind < 0) {
 		pthread_spin_unlock(&qp->sq.lock);
-		ret = EINVAL;
-		goto out;
+		*bad_wr = wr;
+		return EINVAL;
 	}
 
 	for (nreq = 0; wr; ++nreq, wr = wr->next) {