[bug report] drm/amdgpu : Generate XGMI topology info from driver level
Liu, Shaoyun
Shaoyun.Liu at amd.com
Fri Jan 4 16:52:23 UTC 2019
It seems patch d492c564 revert to this original code otherwise we
shouldn't get the warning . I can add ret = -EINVAL for this.
Regards
shaoyun.liu
On 2019-01-04 6:53 a.m., Dan Carpenter wrote:
> Hello Shaoyun Liu,
>
> The patch fb30fc59a245: "drm/amdgpu : Generate XGMI topology info
> from driver level" from Jun 27, 2018, leads to the following static
> checker warning:
>
> drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c:117 amdgpu_xgmi_add_device()
> warn: missing error code here? 'amdgpu_get_xgmi_hive()' failed. 'ret' = '0'
>
> drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c
> 88 int amdgpu_xgmi_add_device(struct amdgpu_device *adev)
> 89 {
> 90 struct psp_xgmi_topology_info *hive_topology;
> 91 struct amdgpu_hive_info *hive;
> 92 struct amdgpu_xgmi *entry;
> 93 struct amdgpu_device *tmp_adev = NULL;
> 94
> 95 int count = 0, ret = -EINVAL;
> 96
> 97 if (!adev->gmc.xgmi.supported)
> 98 return 0;
> 99
> 100 ret = psp_xgmi_get_node_id(&adev->psp, &adev->gmc.xgmi.node_id);
> 101 if (ret) {
> 102 dev_err(adev->dev,
> 103 "XGMI: Failed to get node id\n");
> 104 return ret;
> 105 }
> 106
> 107 ret = psp_xgmi_get_hive_id(&adev->psp, &adev->gmc.xgmi.hive_id);
> 108 if (ret) {
> 109 dev_err(adev->dev,
> 110 "XGMI: Failed to get hive id\n");
> 111 return ret;
> 112 }
> 113
> 114 mutex_lock(&xgmi_mutex);
> 115 hive = amdgpu_get_xgmi_hive(adev);
> 116 if (!hive)
> --> 117 goto exit;
>
> It does look like we should set "ret = -ESOMETHING;"... Not sure what.
>
> 118
> 119 hive_topology = &hive->topology_info;
> 120
> 121 list_add_tail(&adev->gmc.xgmi.head, &hive->device_list);
> 122 list_for_each_entry(entry, &hive->device_list, head)
> 123 hive_topology->nodes[count++].node_id = entry->node_id;
> 124 hive->number_devices = count;
> 125
> 126 /* Each psp need to get the latest topology */
> 127 list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head) {
> 128 ret = psp_xgmi_get_topology_info(&tmp_adev->psp, count, hive_topology);
> 129 if (ret) {
> 130 dev_err(tmp_adev->dev,
> 131 "XGMI: Get topology failure on device %llx, hive %llx, ret %d",
> 132 tmp_adev->gmc.xgmi.node_id,
> 133 tmp_adev->gmc.xgmi.hive_id, ret);
> 134 /* To do : continue with some node failed or disable the whole hive */
> 135 break;
> 136 }
> 137 }
> 138
> 139 list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head) {
> 140 ret = amdgpu_xgmi_update_topology(hive, tmp_adev);
> 141 if (ret)
> 142 break;
> 143 }
> 144
> 145 exit:
> 146 mutex_unlock(&xgmi_mutex);
> 147 return ret;
> 148 }
>
> regards,
> dan carpenter
More information about the amd-gfx
mailing list