Skip to content

L
linux
Commit 0ee51f1d authored by John Clements's avatar John Clements Committed by Alex Deucher
Browse files
Options

drm/amdgpu: resolved bug in UMC RAS CE query 

switch CE counter register access' to use SMN

disable UMC indexing mode
Reviewed-by: default avatarTao Zhou <tao.zhou1@amd.com>
Signed-off-by: default avatarJohn Clements <john.clements@amd.com>
Signed-off-by: default avatarAlex Deucher <alexander.deucher@amd.com>
parent a64c9e15
Hide whitespace changes
Inline Side-by-side
Showing with 20 additions and 12 deletions
drivers/gpu/drm/amd/amdgpu/umc_v6_1.c
View file @ 0ee51f1d
...@@ -52,6 +52,12 @@ const uint32_t ...@@ -52,6 +52,12 @@ const uint32_t
{9, 25, 0, 16}, {15, 31, 6, 22} {9, 25, 0, 16}, {15, 31, 6, 22}
}; };
static void umc_v6_1_disable_umc_index_mode(struct amdgpu_device *adev)
{
WREG32_FIELD15(RSMU, 0, RSMU_UMC_INDEX_REGISTER_NBIF_VG20_GPU,
RSMU_UMC_INDEX_MODE_EN, 0);
}
static inline uint32_t get_umc_6_reg_offset(struct amdgpu_device *adev, static inline uint32_t get_umc_6_reg_offset(struct amdgpu_device *adev,
uint32_t umc_inst, uint32_t umc_inst,
uint32_t ch_inst) uint32_t ch_inst)
...@@ -87,27 +93,27 @@ static void umc_v6_1_query_correctable_error_count(struct amdgpu_device *adev, ...@@ -87,27 +93,27 @@ static void umc_v6_1_query_correctable_error_count(struct amdgpu_device *adev,
} }
/* select the lower chip and check the error count */ /* select the lower chip and check the error count */
ecc_err_cnt_sel = RREG32(ecc_err_cnt_sel_addr + umc_reg_offset); ecc_err_cnt_sel = RREG32_PCIE((ecc_err_cnt_sel_addr + umc_reg_offset) * 4);
ecc_err_cnt_sel = REG_SET_FIELD(ecc_err_cnt_sel, UMCCH0_0_EccErrCntSel, ecc_err_cnt_sel = REG_SET_FIELD(ecc_err_cnt_sel, UMCCH0_0_EccErrCntSel,
EccErrCntCsSel, 0); EccErrCntCsSel, 0);
WREG32(ecc_err_cnt_sel_addr + umc_reg_offset, ecc_err_cnt_sel); WREG32_PCIE((ecc_err_cnt_sel_addr + umc_reg_offset) * 4, ecc_err_cnt_sel);
ecc_err_cnt = RREG32(ecc_err_cnt_addr + umc_reg_offset); ecc_err_cnt = RREG32_PCIE((ecc_err_cnt_addr + umc_reg_offset) * 4);
*error_count += *error_count +=
(REG_GET_FIELD(ecc_err_cnt, UMCCH0_0_EccErrCnt, EccErrCnt) - (REG_GET_FIELD(ecc_err_cnt, UMCCH0_0_EccErrCnt, EccErrCnt) -
UMC_V6_1_CE_CNT_INIT); UMC_V6_1_CE_CNT_INIT);
/* clear the lower chip err count */ /* clear the lower chip err count */
WREG32(ecc_err_cnt_addr + umc_reg_offset, UMC_V6_1_CE_CNT_INIT); WREG32_PCIE((ecc_err_cnt_addr + umc_reg_offset) * 4, UMC_V6_1_CE_CNT_INIT);
/* select the higher chip and check the err counter */ /* select the higher chip and check the err counter */
ecc_err_cnt_sel = REG_SET_FIELD(ecc_err_cnt_sel, UMCCH0_0_EccErrCntSel, ecc_err_cnt_sel = REG_SET_FIELD(ecc_err_cnt_sel, UMCCH0_0_EccErrCntSel,
EccErrCntCsSel, 1); EccErrCntCsSel, 1);
WREG32(ecc_err_cnt_sel_addr + umc_reg_offset, ecc_err_cnt_sel); WREG32_PCIE((ecc_err_cnt_sel_addr + umc_reg_offset) * 4, ecc_err_cnt_sel);
ecc_err_cnt = RREG32(ecc_err_cnt_addr + umc_reg_offset); ecc_err_cnt = RREG32_PCIE((ecc_err_cnt_addr + umc_reg_offset) * 4);
*error_count += *error_count +=
(REG_GET_FIELD(ecc_err_cnt, UMCCH0_0_EccErrCnt, EccErrCnt) - (REG_GET_FIELD(ecc_err_cnt, UMCCH0_0_EccErrCnt, EccErrCnt) -
UMC_V6_1_CE_CNT_INIT); UMC_V6_1_CE_CNT_INIT);
/* clear the higher chip err count */ /* clear the higher chip err count */
WREG32(ecc_err_cnt_addr + umc_reg_offset, UMC_V6_1_CE_CNT_INIT); WREG32_PCIE((ecc_err_cnt_addr + umc_reg_offset) * 4, UMC_V6_1_CE_CNT_INIT);
/* check for SRAM correctable error /* check for SRAM correctable error
MCUMC_STATUS is a 64 bit register */ MCUMC_STATUS is a 64 bit register */
...@@ -283,21 +289,21 @@ static void umc_v6_1_err_cnt_init_per_channel(struct amdgpu_device *adev, ...@@ -283,21 +289,21 @@ static void umc_v6_1_err_cnt_init_per_channel(struct amdgpu_device *adev,
} }
/* select the lower chip and check the error count */ /* select the lower chip and check the error count */
ecc_err_cnt_sel = RREG32(ecc_err_cnt_sel_addr + umc_reg_offset); ecc_err_cnt_sel = RREG32_PCIE((ecc_err_cnt_sel_addr + umc_reg_offset) * 4);
ecc_err_cnt_sel = REG_SET_FIELD(ecc_err_cnt_sel, UMCCH0_0_EccErrCntSel, ecc_err_cnt_sel = REG_SET_FIELD(ecc_err_cnt_sel, UMCCH0_0_EccErrCntSel,
EccErrCntCsSel, 0); EccErrCntCsSel, 0);
/* set ce error interrupt type to APIC based interrupt */ /* set ce error interrupt type to APIC based interrupt */
ecc_err_cnt_sel = REG_SET_FIELD(ecc_err_cnt_sel, UMCCH0_0_EccErrCntSel, ecc_err_cnt_sel = REG_SET_FIELD(ecc_err_cnt_sel, UMCCH0_0_EccErrCntSel,
EccErrInt, 0x1); EccErrInt, 0x1);
WREG32(ecc_err_cnt_sel_addr + umc_reg_offset, ecc_err_cnt_sel); WREG32_PCIE((ecc_err_cnt_sel_addr + umc_reg_offset) * 4, ecc_err_cnt_sel);
/* set error count to initial value */ /* set error count to initial value */
WREG32(ecc_err_cnt_addr + umc_reg_offset, UMC_V6_1_CE_CNT_INIT); WREG32_PCIE((ecc_err_cnt_addr + umc_reg_offset) * 4, UMC_V6_1_CE_CNT_INIT);
/* select the higher chip and check the err counter */ /* select the higher chip and check the err counter */
ecc_err_cnt_sel = REG_SET_FIELD(ecc_err_cnt_sel, UMCCH0_0_EccErrCntSel, ecc_err_cnt_sel = REG_SET_FIELD(ecc_err_cnt_sel, UMCCH0_0_EccErrCntSel,
EccErrCntCsSel, 1); EccErrCntCsSel, 1);
WREG32(ecc_err_cnt_sel_addr + umc_reg_offset, ecc_err_cnt_sel); WREG32_PCIE((ecc_err_cnt_sel_addr + umc_reg_offset) * 4, ecc_err_cnt_sel);
WREG32(ecc_err_cnt_addr + umc_reg_offset, UMC_V6_1_CE_CNT_INIT); WREG32_PCIE((ecc_err_cnt_addr + umc_reg_offset) * 4, UMC_V6_1_CE_CNT_INIT);
} }
static void umc_v6_1_err_cnt_init(struct amdgpu_device *adev) static void umc_v6_1_err_cnt_init(struct amdgpu_device *adev)
...@@ -306,6 +312,8 @@ static void umc_v6_1_err_cnt_init(struct amdgpu_device *adev) ...@@ -306,6 +312,8 @@ static void umc_v6_1_err_cnt_init(struct amdgpu_device *adev)
uint32_t ch_inst = 0; uint32_t ch_inst = 0;
uint32_t umc_reg_offset = 0; uint32_t umc_reg_offset = 0;
umc_v6_1_disable_umc_index_mode(adev);
for (umc_inst = 0; umc_inst < adev->umc.umc_inst_num; umc_inst++) { for (umc_inst = 0; umc_inst < adev->umc.umc_inst_num; umc_inst++) {
for (ch_inst = 0; ch_inst < adev->umc.channel_inst_num; ch_inst++) { for (ch_inst = 0; ch_inst < adev->umc.channel_inst_num; ch_inst++) {
umc_reg_offset = get_umc_6_reg_offset(adev, umc_reg_offset = get_umc_6_reg_offset(adev,
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment
GitLab Nexedi Edition | About GitLab | About Nexedi | 沪ICP备2021021310号-2 | 沪ICP备2021021310号-7