Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
drm/amdgpu: add RAS poison handling for MCA
For MCA poison, if unmap queue fails, only gpu reset should be
triggered without page retirement handling, MCA notifier will do it.
v2: handle MCA poison consumption in umc_poison_handler directly.
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
ae45a18b
24b82292
+20
-11
+20
-11
drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c
+20
-11
drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c
···
169
169
void *ras_error_status,
170
170
bool reset)
171
171
{
172
-
int ret;
173
-
struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status;
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
+
int ret = AMDGPU_RAS_SUCCESS;
178
173
179
-
ret =
180
-
amdgpu_umc_do_page_retirement(adev, ras_error_status, NULL, reset);
174
+
if (!adev->gmc.xgmi.connected_to_cpu) {
175
+
struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status;
176
+
struct ras_common_if head = {
177
+
.block = AMDGPU_RAS_BLOCK__UMC,
178
+
};
179
+
struct ras_manager *obj = amdgpu_ras_find_obj(adev, &head);
181
180
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;
181
+
ret =
182
+
amdgpu_umc_do_page_retirement(adev, ras_error_status, NULL, reset);
183
+
184
+
if (ret == AMDGPU_RAS_SUCCESS && obj) {
185
+
obj->err_data.ue_count += err_data->ue_count;
186
+
obj->err_data.ce_count += err_data->ce_count;
187
+
}
188
+
} else if (reset) {
189
+
/* MCA poison handler is only responsible for GPU reset,
190
+
* let MCA notifier do page retirement.
191
+
*/
192
+
kgd2kfd_set_sram_ecc_flag(adev->kfd.dev);
193
+
amdgpu_ras_reset_gpu(adev);
185
194
}
186
195
187
196
return ret;