Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
drm/amdgpu: add RAS page retirement functions for MCA
Define page retirement functions for MCA platform.
v2: remove page retirement handling from MCA poison handler,
let MCA notifier do page retirement.
v3: remove specific poison handler for MCA to simplify code.
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
cbe4d43e
25135748
+55
+53
drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c
+53
drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c
···
22
22
*/
23
23
24
24
#include "amdgpu.h"
25
+
#include "umc_v6_7.h"
26
+
27
+
static int amdgpu_umc_convert_error_address(struct amdgpu_device *adev,
28
+
struct ras_err_data *err_data, uint64_t err_addr,
29
+
uint32_t ch_inst, uint32_t umc_inst)
30
+
{
31
+
switch (adev->ip_versions[UMC_HWIP][0]) {
32
+
case IP_VERSION(6, 7, 0):
33
+
umc_v6_7_convert_error_address(adev,
34
+
err_data, err_addr, ch_inst, umc_inst);
35
+
break;
36
+
default:
37
+
dev_warn(adev->dev,
38
+
"UMC address to Physical address translation is not supported\n");
39
+
return AMDGPU_RAS_FAIL;
40
+
}
41
+
42
+
return AMDGPU_RAS_SUCCESS;
43
+
}
44
+
45
+
int amdgpu_umc_page_retirement_mca(struct amdgpu_device *adev,
46
+
uint64_t err_addr, uint32_t ch_inst, uint32_t umc_inst)
47
+
{
48
+
struct ras_err_data err_data = {0, 0, 0, NULL};
49
+
int ret = AMDGPU_RAS_FAIL;
50
+
51
+
err_data.err_addr =
52
+
kcalloc(adev->umc.max_ras_err_cnt_per_query,
53
+
sizeof(struct eeprom_table_record), GFP_KERNEL);
54
+
if (!err_data.err_addr) {
55
+
dev_warn(adev->dev,
56
+
"Failed to alloc memory for umc error record in MCA notifier!\n");
57
+
return AMDGPU_RAS_FAIL;
58
+
}
59
+
60
+
/*
61
+
* Translate UMC channel address to Physical address
62
+
*/
63
+
ret = amdgpu_umc_convert_error_address(adev, &err_data, err_addr,
64
+
ch_inst, umc_inst);
65
+
if (ret)
66
+
goto out;
67
+
68
+
if (amdgpu_bad_page_threshold != 0) {
69
+
amdgpu_ras_add_bad_pages(adev, err_data.err_addr,
70
+
err_data.err_addr_cnt);
71
+
amdgpu_ras_save_bad_pages(adev);
72
+
}
73
+
74
+
out:
75
+
kfree(err_data.err_addr);
76
+
return ret;
77
+
}
25
78
26
79
static int amdgpu_umc_do_page_retirement(struct amdgpu_device *adev,
27
80
void *ras_error_status,
+2
drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h
+2
drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h