Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
drm/amdgpu: add RAS poison consumption handler for SRIOV
Send message to PF if VF receives RAS poison consumption interrupt.
Signed-off-by: Tao Zhou <tao.zhou1@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
authored by
Tao Zhou
and committed by
Alex Deucher
e643823d
ae844dd7
+24
-16
+24
-16
drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c
+24
-16
drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c
···
169
169
{
170
170
int ret = AMDGPU_RAS_SUCCESS;
171
171
172
-
if (!adev->gmc.xgmi.connected_to_cpu) {
173
-
struct ras_err_data err_data = {0, 0, 0, NULL};
174
-
struct ras_common_if head = {
175
-
.block = AMDGPU_RAS_BLOCK__UMC,
176
-
};
177
-
struct ras_manager *obj = amdgpu_ras_find_obj(adev, &head);
172
+
if (!amdgpu_sriov_vf(adev)) {
173
+
if (!adev->gmc.xgmi.connected_to_cpu) {
174
+
struct ras_err_data err_data = {0, 0, 0, NULL};
175
+
struct ras_common_if head = {
176
+
.block = AMDGPU_RAS_BLOCK__UMC,
177
+
};
178
+
struct ras_manager *obj = amdgpu_ras_find_obj(adev, &head);
178
179
179
-
ret = amdgpu_umc_do_page_retirement(adev, &err_data, NULL, reset);
180
+
ret = amdgpu_umc_do_page_retirement(adev, &err_data, NULL, reset);
180
181
181
-
if (ret == AMDGPU_RAS_SUCCESS && obj) {
182
-
obj->err_data.ue_count += err_data.ue_count;
183
-
obj->err_data.ce_count += err_data.ce_count;
182
+
if (ret == AMDGPU_RAS_SUCCESS && obj) {
183
+
obj->err_data.ue_count += err_data.ue_count;
184
+
obj->err_data.ce_count += err_data.ce_count;
185
+
}
186
+
} else if (reset) {
187
+
/* MCA poison handler is only responsible for GPU reset,
188
+
* let MCA notifier do page retirement.
189
+
*/
190
+
kgd2kfd_set_sram_ecc_flag(adev->kfd.dev);
191
+
amdgpu_ras_reset_gpu(adev);
184
192
}
185
-
} else if (reset) {
186
-
/* MCA poison handler is only responsible for GPU reset,
187
-
* let MCA notifier do page retirement.
188
-
*/
189
-
kgd2kfd_set_sram_ecc_flag(adev->kfd.dev);
190
-
amdgpu_ras_reset_gpu(adev);
193
+
} else {
194
+
if (adev->virt.ops && adev->virt.ops->ras_poison_handler)
195
+
adev->virt.ops->ras_poison_handler(adev);
196
+
else
197
+
dev_warn(adev->dev,
198
+
"No ras_poison_handler interface in SRIOV!\n");
191
199
}
192
200
193
201
return ret;