[PATCH 2/2] drm/amdgpu: simplify convert_error_address interface for UMC v12
Tao Zhou
tao.zhou1 at amd.com
Thu Mar 21 03:30:07 UTC 2024
Replace separate parameters with struct ta_ras_query_address_input.
Signed-off-by: Tao Zhou <tao.zhou1 at amd.com>
---
drivers/gpu/drm/amd/amdgpu/umc_v12_0.c | 57 ++++++++++++++------------
1 file changed, 30 insertions(+), 27 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v12_0.c b/drivers/gpu/drm/amd/amdgpu/umc_v12_0.c
index 0a9cc87e98d0..d0fcfcb3404f 100644
--- a/drivers/gpu/drm/amd/amdgpu/umc_v12_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/umc_v12_0.c
@@ -266,26 +266,19 @@ static void umc_v12_0_mca_addr_to_pa(struct amdgpu_device *adev,
}
static void umc_v12_0_convert_error_address(struct amdgpu_device *adev,
- struct ras_err_data *err_data, uint64_t err_addr,
- uint32_t ch_inst, uint32_t umc_inst,
- uint32_t node_inst, uint32_t socket_id)
+ struct ras_err_data *err_data,
+ struct ta_ras_query_address_input *addr_in)
{
uint32_t col, row, row_xor, bank, channel_index;
- uint64_t soc_pa, retired_page, column;
- struct ta_ras_query_address_input addr_in;
+ uint64_t soc_pa, retired_page, column, err_addr;
struct ta_ras_query_address_output addr_out;
- addr_in.addr_type = TA_RAS_MCA_TO_PA;
- addr_in.ma.err_addr = err_addr;
- addr_in.ma.ch_inst = ch_inst;
- addr_in.ma.umc_inst = umc_inst;
- addr_in.ma.node_inst = node_inst;
- addr_in.ma.socket_id = socket_id;
-
- if (psp_ras_query_address(&adev->psp, &addr_in, &addr_out))
+ err_addr = addr_in->ma.err_addr;
+ addr_in->addr_type = TA_RAS_MCA_TO_PA;
+ if (psp_ras_query_address(&adev->psp, addr_in, &addr_out))
/* fallback to old path if fail to get pa from psp */
- umc_v12_0_mca_addr_to_pa(adev, err_addr, ch_inst, umc_inst,
- node_inst, &addr_out);
+ umc_v12_0_mca_addr_to_pa(adev, err_addr, addr_in->ma.ch_inst,
+ addr_in->ma.umc_inst, addr_in->ma.node_inst, &addr_out);
soc_pa = addr_out.pa.pa;
bank = addr_out.pa.bank;
@@ -310,7 +303,7 @@ static void umc_v12_0_convert_error_address(struct amdgpu_device *adev,
"Error Address(PA):0x%-10llx Row:0x%-4x Col:0x%-2x Bank:0x%x Channel:0x%x\n",
retired_page, row, col, bank, channel_index);
amdgpu_umc_fill_error_record(err_data, err_addr,
- retired_page, channel_index, umc_inst);
+ retired_page, channel_index, addr_in->ma.umc_inst);
/* shift R13 bit */
retired_page ^= (0x1ULL << UMC_V12_0_PA_R13_BIT);
@@ -318,7 +311,7 @@ static void umc_v12_0_convert_error_address(struct amdgpu_device *adev,
"Error Address(PA):0x%-10llx Row:0x%-4x Col:0x%-2x Bank:0x%x Channel:0x%x\n",
retired_page, row_xor, col, bank, channel_index);
amdgpu_umc_fill_error_record(err_data, err_addr,
- retired_page, channel_index, umc_inst);
+ retired_page, channel_index, addr_in->ma.umc_inst);
}
}
@@ -326,13 +319,13 @@ static int umc_v12_0_query_error_address(struct amdgpu_device *adev,
uint32_t node_inst, uint32_t umc_inst,
uint32_t ch_inst, void *data)
{
+ struct ras_err_data *err_data = (struct ras_err_data *)data;
+ struct ta_ras_query_address_input addr_in;
uint64_t mc_umc_status_addr;
uint64_t mc_umc_status, err_addr;
uint64_t mc_umc_addrt0;
- struct ras_err_data *err_data = (struct ras_err_data *)data;
uint64_t umc_reg_offset =
get_umc_v12_0_reg_offset(adev, node_inst, umc_inst, ch_inst);
- uint32_t socket_id = 0;
mc_umc_status_addr =
SOC15_REG_OFFSET(UMC, 0, regMCA_UMC_UMC0_MCUMC_STATUST0);
@@ -362,10 +355,16 @@ static int umc_v12_0_query_error_address(struct amdgpu_device *adev,
if (!adev->aid_mask &&
adev->smuio.funcs &&
adev->smuio.funcs->get_socket_id)
- socket_id = adev->smuio.funcs->get_socket_id(adev);
+ addr_in.ma.socket_id = adev->smuio.funcs->get_socket_id(adev);
+ else
+ addr_in.ma.socket_id = 0;
+
+ addr_in.ma.err_addr = err_addr;
+ addr_in.ma.ch_inst = ch_inst;
+ addr_in.ma.umc_inst = umc_inst;
+ addr_in.ma.node_inst = node_inst;
- umc_v12_0_convert_error_address(adev, err_data, err_addr,
- ch_inst, umc_inst, node_inst, socket_id);
+ umc_v12_0_convert_error_address(adev, err_data, &addr_in);
}
/* clear umc status */
@@ -425,12 +424,16 @@ static void umc_v12_0_ecc_info_query_ras_error_address(struct amdgpu_device *ade
struct ras_err_info *err_info;
struct ras_err_addr *mca_err_addr, *tmp;
struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status;
+ struct ta_ras_query_address_input addr_in;
for_each_ras_error(err_node, err_data) {
err_info = &err_node->err_info;
if (list_empty(&err_info->err_addr_list))
continue;
+ addr_in.ma.node_inst = err_info->mcm_info.die_id;
+ addr_in.ma.socket_id = err_info->mcm_info.socket_id;
+
list_for_each_entry_safe(mca_err_addr, tmp, &err_info->err_addr_list, node) {
mc_umc_status = mca_err_addr->err_status;
if (mc_umc_status &&
@@ -446,6 +449,10 @@ static void umc_v12_0_ecc_info_query_ras_error_address(struct amdgpu_device *ade
MCA_UMC_UMC0_MCUMC_ADDRT0, ErrorAddr);
InstanceIdLo = REG_GET_FIELD(mca_ipid, MCMP1_IPIDT0, InstanceIdLo);
+ addr_in.ma.err_addr = err_addr;
+ addr_in.ma.ch_inst = MCA_IPID_LO_2_UMC_CH(InstanceIdLo);
+ addr_in.ma.umc_inst = MCA_IPID_LO_2_UMC_INST(InstanceIdLo);
+
dev_info(adev->dev, "UMC:IPID:0x%llx, aid:%d, inst:%d, ch:%d, err_addr:0x%llx\n",
mca_ipid,
err_info->mcm_info.die_id,
@@ -454,11 +461,7 @@ static void umc_v12_0_ecc_info_query_ras_error_address(struct amdgpu_device *ade
err_addr);
umc_v12_0_convert_error_address(adev,
- err_data, err_addr,
- MCA_IPID_LO_2_UMC_CH(InstanceIdLo),
- MCA_IPID_LO_2_UMC_INST(InstanceIdLo),
- err_info->mcm_info.die_id,
- err_info->mcm_info.socket_id);
+ err_data, &addr_in);
}
/* Delete error address node from list and free memory */
--
2.34.1
More information about the amd-gfx
mailing list