Skip to contentSkip to conversation

RDMA/restrack: Store all special QPs in restrack DB

bug report Dan Carpenter 2 participants 2 messages Nov 26, 2020 InfiniBand

lore
Share & tools
lore
#0 lore raw

Hello Leon Romanovsky,

The patch 0413755c95e7: "RDMA/restrack: Store all special QPs in restrack DB" from Nov 4, 2020, leads to the following static checker warning:

drivers/infiniband/core/restrack.c:235 rdma_restrack_add()
warn: right shifting more than type allows 8 vs 8

drivers/infiniband/core/restrack.c
220 void rdma_restrack_add(struct rdma_restrack_entry *res) 221 {

   222          struct ib_device *dev = res_to_dev(res);
   223          struct rdma_restrack_root *rt;
   224          int ret;
   225  
   226          if (!dev)
   227                  return;
   228  
   229          rt = &dev->res[res->type];
   230  
   231          if (res->type == RDMA_RESTRACK_QP) {
   232                  /* Special case to ensure that LQPN points to right QP */
   233                  struct ib_qp *qp = container_of(res, struct ib_qp, res);
   234  
   235                  WARN_ONCE(qp->qp_num >> 24 || qp->port >> 8,
                                                      ^^^^^^^^^^^^^

qp->port is a u8 so this is always going to be zero.

   236                            "QP number 0x%0X and port 0x%0X", qp->qp_num,
   237                            qp->port);
   238                  res->id = qp->qp_num;
   239                  if (qp->qp_type == IB_QPT_SMI || qp->qp_type == IB_QPT_GSI)
   240                          res->id |= qp->port << 24;
   241                  ret = xa_insert(&rt->xa, res->id, res, GFP_KERNEL);
   242                  if (ret)
   243                          res->id = 0;
   244          } else if (res->type == RDMA_RESTRACK_COUNTER) {
   245                  /* Special case to ensure that cntn points to right counter */

regards,
dan carpenter

#1 lore raw
Quoted 7 lines · Dan Carpenter
   231          if (res->type == RDMA_RESTRACK_QP) {
   232                  /* Special case to ensure that LQPN points to right QP */
   233                  struct ib_qp *qp = container_of(res, struct ib_qp, res);
   234  
   235                  WARN_ONCE(qp->qp_num >> 24 || qp->port >> 8,
                                                      ^^^^^^^^^^^^^

qp->port is a u8 so this is always going to be zero.

I think we should ignore this one. I'm expecting the width of port to increase..

Jason