Add poison mode check error condition for umc v12_0.
Signed-off-by: YiPeng Chai <yipeng.c...@amd.com>
---
drivers/gpu/drm/amd/amdgpu/umc_v12_0.c | 20 ++++++++++++++-----
drivers/gpu/drm/amd/amdgpu/umc_v12_0.h | 4 ++--
.../drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c | 4 ++--
3 files changed, 19 insertions(+), 9 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v12_0.c
b/drivers/gpu/drm/amd/amdgpu/umc_v12_0.c
index 8d60c39ae1c5..8430888760ba 100644
--- a/drivers/gpu/drm/amd/amdgpu/umc_v12_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/umc_v12_0.c
@@ -88,16 +88,26 @@ static void umc_v12_0_reset_error_count(struct
amdgpu_device *adev)
umc_v12_0_reset_error_count_per_channel, NULL);
}
-bool umc_v12_0_is_uncorrectable_error(uint64_t mc_umc_status)
+bool umc_v12_0_is_uncorrectable_error(struct amdgpu_device *adev, uint64_t
mc_umc_status)
{
+ if (amdgpu_ras_is_poison_mode_supported(adev) &&
+ (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Val) ==
1) &&
+ (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0,
Deferred) == 1))
+ return true;
+
return ((REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Val)
== 1) &&
(REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, PCC)
== 1 ||
REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UC)
== 1 ||
REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, TCC)
== 1));
}
-bool umc_v12_0_is_correctable_error(uint64_t mc_umc_status)
+bool umc_v12_0_is_correctable_error(struct amdgpu_device *adev, uint64_t
mc_umc_status)
{
+ if (amdgpu_ras_is_poison_mode_supported(adev) &&
+ (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Val) ==
1) &&
+ (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0,
Deferred) == 1))
+ return false;
+
return (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Val)
== 1 &&
(REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0,
CECC) == 1 ||
(REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0,
UECC) == 1 &&
@@ -105,7 +115,7 @@ bool umc_v12_0_is_correctable_error(uint64_t mc_umc_status)
/* Identify data parity error in replay mode */
((REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0,
ErrorCodeExt) == 0x5 ||
REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0,
ErrorCodeExt) == 0xb) &&
- !(umc_v12_0_is_uncorrectable_error(mc_umc_status)))));
+ !(umc_v12_0_is_uncorrectable_error(adev, mc_umc_status)))));
}
static void umc_v12_0_query_correctable_error_count(struct amdgpu_device *adev,
@@ -124,7 +134,7 @@ static void umc_v12_0_query_correctable_error_count(struct
amdgpu_device *adev,
mc_umc_status =
RREG64_PCIE_EXT((mc_umc_status_addr + umc_reg_offset) * 4);
- if (umc_v12_0_is_correctable_error(mc_umc_status))
+ if (umc_v12_0_is_correctable_error(adev, mc_umc_status))
*error_count += 1;
}
@@ -142,7 +152,7 @@ static void
umc_v12_0_query_uncorrectable_error_count(struct amdgpu_device *adev
mc_umc_status =
RREG64_PCIE_EXT((mc_umc_status_addr + umc_reg_offset) * 4);
- if (umc_v12_0_is_uncorrectable_error(mc_umc_status))
+ if (umc_v12_0_is_uncorrectable_error(adev, mc_umc_status))
*error_count += 1;
}
diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v12_0.h
b/drivers/gpu/drm/amd/amdgpu/umc_v12_0.h
index b34b1e358f8b..17b4b52d6f13 100644
--- a/drivers/gpu/drm/amd/amdgpu/umc_v12_0.h
+++ b/drivers/gpu/drm/amd/amdgpu/umc_v12_0.h
@@ -117,8 +117,8 @@
(pa) |= (UMC_V12_0_CHANNEL_HASH_CH6(channel_idx, pa) <<
UMC_V12_0_PA_CH6_BIT); \
} while (0)
-bool umc_v12_0_is_uncorrectable_error(uint64_t mc_umc_status);
-bool umc_v12_0_is_correctable_error(uint64_t mc_umc_status);
+bool umc_v12_0_is_uncorrectable_error(struct amdgpu_device *adev, uint64_t
mc_umc_status);
+bool umc_v12_0_is_correctable_error(struct amdgpu_device *adev, uint64_t
mc_umc_status);
extern const uint32_t
umc_v12_0_channel_idx_tbl[]
diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c
b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c
index ddd782fbee7a..bb3e953bfed5 100644
--- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c
+++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c
@@ -2524,9 +2524,9 @@ static int mca_umc_mca_get_err_count(const struct
mca_ras_info *mca_ras, struct
return 0;
}
- if (type == AMDGPU_MCA_ERROR_TYPE_UE &&
umc_v12_0_is_uncorrectable_error(status0))
+ if (type == AMDGPU_MCA_ERROR_TYPE_UE &&
umc_v12_0_is_uncorrectable_error(adev, status0))
*count = 1;
- else if (type == AMDGPU_MCA_ERROR_TYPE_CE &&
umc_v12_0_is_correctable_error(status0))
+ else if (type == AMDGPU_MCA_ERROR_TYPE_CE &&
umc_v12_0_is_correctable_error(adev, status0))
*count = 1;
return 0;
--
2.34.1
