On 2019/2/13 15:01, Dan Carpenter wrote:
> Hello Wei Hu (Xavier),
>
> The patch 6a04aed6afae: "RDMA/hns: Fix the chip hanging caused by
> sending mailbox&CMQ during reset" from Feb 3, 2019, leads to the
> following static checker warning:
>
> drivers/infiniband/hw/hns/hns_roce_qp.c:940 hns_roce_create_qp()
> error: passing non negative 2 to ERR_PTR
>
> drivers/infiniband/hw/hns/hns_roce_qp.c
> 930 hr_qp->ibqp.qp_num = HNS_ROCE_MAX_PORTS +
> 931 hr_dev->iboe.phy_port[hr_qp->port];
> 932 else
> 933 hr_qp->ibqp.qp_num = 1;
> 934
> 935 ret = hns_roce_create_qp_common(hr_dev, pd, init_attr, udata,
> 936 hr_qp->ibqp.qp_num, hr_qp);
> 937 if (ret) {
> 938 dev_err(dev, "Create GSI QP failed!\n");
> 939 kfree(hr_sqp);
> --> 940 return ERR_PTR(ret);
> 941 }
> 942
> 943 break;
> 944 }
> 945 default:{
> 946 dev_err(dev, "not support QP type %d\n", init_attr->qp_type);
> 947 return ERR_PTR(-EINVAL);
> 948 }
> 949 }
> 950
> 951 return &hr_qp->ibqp;
> 952 }
>
> The number 2 comes from hns_roce_cmq_send().
>
>
> drivers/infiniband/hw/hns/hns_roce_hw_v2
> 1089 int hns_roce_cmq_send(struct hns_roce_dev *hr_dev,
> 1090 struct hns_roce_cmq_desc *desc, int num)
> 1091 {
> 1092 int retval;
> 1093 int ret;
> 1094
> 1095 ret = hns_roce_v2_rst_process_cmd(hr_dev);
> 1096 if (ret == CMD_RST_PRC_SUCCESS)
> 1097 return 0;
> 1098 if (ret == CMD_RST_PRC_EBUSY)
> 1099 return ret;
> ^^^^^^^^^^
>
> 1100
> 1101 ret = __hns_roce_cmq_send(hr_dev, desc, num);
> 1102 if (ret) {
> 1103 retval = hns_roce_v2_rst_process_cmd(hr_dev);
> 1104 if (retval == CMD_RST_PRC_SUCCESS)
> 1105 return 0;
> 1106 else if (retval == CMD_RST_PRC_EBUSY)
> 1107 return retval;
> ^^^^^^^^^^^^^
>
> These should probably be -EBUSY.
>
> 1108 }
> 1109
> 1110 return ret;
> 1111 }
>
> regards,
> dan carpenter
Hi, Dan carpenter
Thanks for you comments.
We will carefully analyze this and amend it as soon as possible.
Regards
Xaiver
>
> .
>
