IB/ehca: Discard double CQE for one WR

Under rare circumstances, the ehca hardware might erroneously generate
two CQEs for the same WQE, which is not compliant to the IB spec and
will cause unpredictable errors like memory being freed twice. To
avoid this problem, the driver needs to detect the second CQE and
discard it.

For this purpose, introduce an array holding as many elements as the
SQ of the QP, called sq_map. Each sq_map entry stores a "reported"
flag for one WQE in the SQ. When a work request is posted to the SQ,
the respective "reported" flag is set to zero. After the arrival of a
CQE, the flag is set to 1, which allows to detect the occurence of a
second CQE.

The mapping between WQE / CQE and the corresponding sq_map element is
implemented by replacing the lowest 16 Bits of the wr_id with the
index in the queue map. The original 16 Bits are stored in the sq_map
entry and are restored when the CQE is passed to the application.

Signed-off-by: Alexander Schmidt <alexs@linux.vnet.ibm.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

1 files changed, 25 insertions, 9 deletions

diff --git a/drivers/infiniband/hw/ehca/ehca_qp.c b/drivers/infiniband/hw/ehca/ehca_qp.c
index c58fd4eead1..b6bcee03673 100644
--- a/drivers/infiniband/hw/ehca/ehca_qp.c
+++ b/drivers/infiniband/hw/ehca/ehca_qp.c
@@ -412,6 +412,7 @@ static struct ehca_qp *internal_create_qp(
 	struct ehca_shca *shca = container_of(pd->device, struct ehca_shca,
 					      ib_device);
 	struct ib_ucontext *context = NULL;
+	u32 nr_qes;
 	u64 h_ret;
 	int is_llqp = 0, has_srq = 0;
 	int qp_type, max_send_sge, max_recv_sge, ret;
@@ -715,6 +716,15 @@ static struct ehca_qp *internal_create_qp(
 				 "and pages ret=%i", ret);
 			goto create_qp_exit2;
 		}
+		nr_qes = my_qp->ipz_squeue.queue_length /
+			 my_qp->ipz_squeue.qe_size;
+		my_qp->sq_map = vmalloc(nr_qes *
+					sizeof(struct ehca_qmap_entry));
+		if (!my_qp->sq_map) {
+			ehca_err(pd->device, "Couldn't allocate squeue "
+				 "map ret=%i", ret);
+			goto create_qp_exit3;
+		}
 	}
 
 	if (HAS_RQ(my_qp)) {
@@ -724,7 +734,7 @@ static struct ehca_qp *internal_create_qp(
 		if (ret) {
 			ehca_err(pd->device, "Couldn't initialize rqueue "
 				 "and pages ret=%i", ret);
-			goto create_qp_exit3;
+			goto create_qp_exit4;
 		}
 	}
 
@@ -770,7 +780,7 @@ static struct ehca_qp *internal_create_qp(
 			if (!my_qp->mod_qp_parm) {
 				ehca_err(pd->device,
 					 "Could not alloc mod_qp_parm");
-				goto create_qp_exit4;
+				goto create_qp_exit5;
 			}
 		}
 	}
@@ -780,7 +790,7 @@ static struct ehca_qp *internal_create_qp(
 		h_ret = ehca_define_sqp(shca, my_qp, init_attr);
 		if (h_ret != H_SUCCESS) {
 			ret = ehca2ib_return_code(h_ret);
-			goto create_qp_exit5;
+			goto create_qp_exit6;
 		}
 	}
 
@@ -789,7 +799,7 @@ static struct ehca_qp *internal_create_qp(
 		if (ret) {
 			ehca_err(pd->device,
 				 "Couldn't assign qp to send_cq ret=%i", ret);
-			goto create_qp_exit5;
+			goto create_qp_exit6;
 		}
 	}
 
@@ -815,22 +825,26 @@ static struct ehca_qp *internal_create_qp(
 		if (ib_copy_to_udata(udata, &resp, sizeof resp)) {
 			ehca_err(pd->device, "Copy to udata failed");
 			ret = -EINVAL;
-			goto create_qp_exit6;
+			goto create_qp_exit7;
 		}
 	}
 
 	return my_qp;
 
-create_qp_exit6:
+create_qp_exit7:
 	ehca_cq_unassign_qp(my_qp->send_cq, my_qp->real_qp_num);
 
-create_qp_exit5:
+create_qp_exit6:
 	kfree(my_qp->mod_qp_parm);
 
-create_qp_exit4:
+create_qp_exit5:
 	if (HAS_RQ(my_qp))
 		ipz_queue_dtor(my_pd, &my_qp->ipz_rqueue);
 
+create_qp_exit4:
+	if (HAS_SQ(my_qp))
+		vfree(my_qp->sq_map);
+
 create_qp_exit3:
 	if (HAS_SQ(my_qp))
 		ipz_queue_dtor(my_pd, &my_qp->ipz_squeue);
@@ -1979,8 +1993,10 @@ static int internal_destroy_qp(struct ib_device *dev, struct ehca_qp *my_qp,
 
 	if (HAS_RQ(my_qp))
 		ipz_queue_dtor(my_pd, &my_qp->ipz_rqueue);
-	if (HAS_SQ(my_qp))
+	if (HAS_SQ(my_qp)) {
 		ipz_queue_dtor(my_pd, &my_qp->ipz_squeue);
+		vfree(my_qp->sq_map);
+	}
 	kmem_cache_free(qp_cache, my_qp);
 	atomic_dec(&shca->num_qps);
 	return 0;