Skip to contentSkip to conversation

drm/amdgpu/pm: support mca_ceumc_addr in ecctable

PATCH In review Stanley.Yang 3 participants 7 messages May 25, 2022

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH v3 1/2 drm/amdgpu/pm: support mca_ceumc_addr in ecctable

SMU add a new variable mca_ceumc_addr to record umc correctable error address in EccInfo table, driver side add EccInfo_V2_t to support this feature

Changed from V1:
remove ecc_table_v2 and unnecessary table id, define union struct include EccInfo_t and EccInfo_V2_t.

Changed from V2:
sync patch verion

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h       |  1 +
 .../inc/pmfw_if/smu13_driver_if_aldebaran.h   | 16 +++++-
 .../drm/amd/pm/swsmu/smu13/aldebaran_ppt.c    | 53 ++++++++++++++-----
 3 files changed, 57 insertions(+), 13 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +1 -0
index b9a6fac2b8b2..28e603243b67 100644
@@ -328,6 +328,7 @@ struct ecc_info_per_ch {
 	uint16_t ce_count_hi_chip;
 	uint64_t mca_umc_status;
 	uint64_t mca_umc_addr;
+	uint64_t mca_ceumc_addr;
 };
 
 struct umc_ecc_info {
drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu13_driver_if_aldebaran.h +15 -1
index 0f67c56c2863..6f92038470ec 100644
@@ -519,7 +519,21 @@ typedef struct {
 } EccInfo_t;
 
 typedef struct {
-	EccInfo_t  EccInfo[ALDEBARAN_UMC_CHANNEL_NUM];
+	uint64_t mca_umc_status;
+	uint64_t mca_umc_addr;
+	uint64_t mca_ceumc_addr;
+
+	uint16_t ce_count_lo_chip;
+	uint16_t ce_count_hi_chip;
+
+	uint32_t eccPadding;
+} EccInfo_V2_t;
+
+typedef struct {
+	union {
+		EccInfo_t  EccInfo[ALDEBARAN_UMC_CHANNEL_NUM];
+		EccInfo_V2_t EccInfo_V2[ALDEBARAN_UMC_CHANNEL_NUM];
+	};
 } EccInfoTable_t;
 
 // These defines are used with the following messages:
drivers/gpu/drm/amd/pm/swsmu/smu13/aldebaran_ppt.c +41 -12

85 lines not on this page · raw message

Signature · 1 line
2.17.1
#1 lore raw

PATCH v3 2/2 drm/amdgpu: print umc correctable error address

Changed from V1:
remove unnecessary same row physical address calculation

Changed from V2:
move record_ce_addr_supported to umc_ecc_info struct

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h       |  5 ++
 drivers/gpu/drm/amd/amdgpu/umc_v6_7.c         | 50 ++++++++++++++++++-
 .../drm/amd/pm/swsmu/smu13/aldebaran_ppt.c    |  1 +
 3 files changed, 54 insertions(+), 2 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +5 -0
index 28e603243b67..bf5a95104ec1 100644
@@ -333,6 +333,11 @@ struct ecc_info_per_ch {
 
 struct umc_ecc_info {
 	struct ecc_info_per_ch ecc[MAX_UMC_CHANNEL_NUM];
+
+	/* Determine smu ecctable whether support
+	 * record correctable error address
+	 */
+	int record_ce_addr_supported;
 };
 
 struct amdgpu_ras {
drivers/gpu/drm/amd/amdgpu/umc_v6_7.c +48 -2

74 lines not on this page · raw message

drivers/gpu/drm/amd/pm/swsmu/smu13/aldebaran_ppt.c +1 -0
index 9cdfeea58085..c7e0fec614ea 100644
@@ -1883,6 +1883,7 @@ static ssize_t aldebaran_get_ecc_info(struct smu_context *smu,
 			ecc_info_per_channel->mca_ceumc_addr =
 				ecc_table->EccInfo_V2[i].mca_ceumc_addr;
 		}
+		eccinfo->record_ce_addr_supported =1;
 	}
 
 	return ret;
Signature · 1 line
2.17.1
#2 lore raw

Re: [PATCH Review v3 2/2] drm/amdgpu: print umc correctable error address

[AMD Official Use Only - General]

From: amd-gfx <[email protected]> on behalf of Stanley.Yang <[email protected]>
Sent: Wednesday, May 25, 2022 2:10 PM
To: [email protected] <[email protected]>; Zhang, Hawking <[email protected]>; Zhou1, Tao <[email protected]>; Quan, Evan <[email protected]>; Lazar, Lijo <[email protected]>
Cc: Yang, Stanley <[email protected]>
Subject: [PATCH Review v3 2/2] drm/amdgpu: print umc correctable error address

Changed from V1:
remove unnecessary same row physical address calculation

Changed from V2:
move record_ce_addr_supported to umc_ecc_info struct

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h       |  5 ++
 drivers/gpu/drm/amd/amdgpu/umc_v6_7.c         | 50 ++++++++++++++++++-
 .../drm/amd/pm/swsmu/smu13/aldebaran_ppt.c    |  1 +
 3 files changed, 54 insertions(+), 2 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +5 -0
index 28e603243b67..bf5a95104ec1 100644
@@ -333,6 +333,11 @@ struct ecc_info_per_ch {
 
 struct umc_ecc_info {
         struct ecc_info_per_ch ecc[MAX_UMC_CHANNEL_NUM];
+
+       /* Determine smu ecctable whether support
+        * record correctable error address
+        */
+       int record_ce_addr_supported;
 };
 

[kevin]:

1. the new field of record_ce_addr_supported is not set on sienna_cichlid chip.

1. and this field is better to renamed to others when this ecc table(pmfw side) update again in the furture. .e.g: ecc_table_version

Best Regards
Kevin

struct amdgpu_ras {

drivers/gpu/drm/amd/amdgpu/umc_v6_7.c +48 -2

74 lines not on this page · raw message

drivers/gpu/drm/amd/pm/swsmu/smu13/aldebaran_ppt.c +1 -0
index 9cdfeea58085..c7e0fec614ea 100644
@@ -1883,6 +1883,7 @@ static ssize_t aldebaran_get_ecc_info(struct smu_context *smu,
                         ecc_info_per_channel->mca_ceumc_addr =
                                 ecc_table->EccInfo_V2[i].mca_ceumc_addr;
                 }
+               eccinfo->record_ce_addr_supported =1;
         }
 
         return ret;

--
2.17.1

#3 lore raw

答复: [PATCH Review v3 2/2] drm/amdgpu: print umc correctable error address

[AMD Official Use Only - General]

[AMD Official Use Only - General]

发件人: Wang, Yang(Kevin) <[email protected]>
日期: 星期三, 2022年5月25日 下午2:52
收件人: Yang, Stanley <[email protected]>, [email protected] <[email protected]>, Zhang, Hawking <[email protected]>, Zhou1, Tao <[email protected]>, Quan, Evan <[email protected]>, Lazar, Lijo <[email protected]> 主题: Re: [PATCH Review v3 2/2] drm/amdgpu: print umc correctable error address

[AMD Official Use Only - General]

From: amd-gfx <[email protected]> on behalf of Stanley.Yang <[email protected]>
Sent: Wednesday, May 25, 2022 2:10 PM
To: [email protected] <[email protected]>; Zhang, Hawking <[email protected]>; Zhou1, Tao <[email protected]>; Quan, Evan <[email protected]>; Lazar, Lijo <[email protected]>
Cc: Yang, Stanley <[email protected]>
Subject: [PATCH Review v3 2/2] drm/amdgpu: print umc correctable error address

Changed from V1:
remove unnecessary same row physical address calculation

Changed from V2:
move record_ce_addr_supported to umc_ecc_info struct

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h       |  5 ++
 drivers/gpu/drm/amd/amdgpu/umc_v6_7.c         | 50 ++++++++++++++++++-
 .../drm/amd/pm/swsmu/smu13/aldebaran_ppt.c    |  1 +
 3 files changed, 54 insertions(+), 2 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +5 -0
index 28e603243b67..bf5a95104ec1 100644
@@ -333,6 +333,11 @@ struct ecc_info_per_ch {
 
 struct umc_ecc_info {
         struct ecc_info_per_ch ecc[MAX_UMC_CHANNEL_NUM];
+
+       /* Determine smu ecctable whether support
+        * record correctable error address
+        */
+       int record_ce_addr_supported;
 };
 

[kevin]:

1. the new field of record_ce_addr_supported is not set on sienna_cichlid chip.
Stanley: Sienna_cichild not support this feature, so do not set record_ce_addr_supported.

1. and this field is better to renamed to others when this ecc table(pmfw side) update again in the furture. .e.g: ecc_table_version

Stanley: To name record_ce_addr_supported is more intuitive then using ecc_table_version or others.

Best Regards
Kevin

struct amdgpu_ras {

drivers/gpu/drm/amd/amdgpu/umc_v6_7.c +48 -2

74 lines not on this page · raw message

drivers/gpu/drm/amd/pm/swsmu/smu13/aldebaran_ppt.c +1 -0
index 9cdfeea58085..c7e0fec614ea 100644
@@ -1883,6 +1883,7 @@ static ssize_t aldebaran_get_ecc_info(struct smu_context *smu,
                         ecc_info_per_channel->mca_ceumc_addr =
                                 ecc_table->EccInfo_V2[i].mca_ceumc_addr;
                 }
+               eccinfo->record_ce_addr_supported =1;
         }
 
         return ret;

--
2.17.1

Acked-by on v3 2/2
#4 lore raw

Re: [PATCH Review v3 2/2] drm/amdgpu: print umc correctable error address

In reply to Stanley.Yang
Quoted 43 lines · Stanley.Yang

Changed from V1:
remove unnecessary same row physical address calculation

Changed from V2:
move record_ce_addr_supported to umc_ecc_info struct

---
  drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h       |  5 ++
  drivers/gpu/drm/amd/amdgpu/umc_v6_7.c         | 50 ++++++++++++++++++-
  .../drm/amd/pm/swsmu/smu13/aldebaran_ppt.c    |  1 +
  3 files changed, 54 insertions(+), 2 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +5 -0
index 28e603243b67..bf5a95104ec1 100644
@@ -333,6 +333,11 @@ struct ecc_info_per_ch {
 
  struct umc_ecc_info {
  	struct ecc_info_per_ch ecc[MAX_UMC_CHANNEL_NUM];
+
+	/* Determine smu ecctable whether support
+	 * record correctable error address
+	 */
+	int record_ce_addr_supported;
  };
 
  struct amdgpu_ras {
drivers/gpu/drm/amd/amdgpu/umc_v6_7.c +6 -0
index 606892dbea1c..bf7524f16b66 100644
@@ -119,6 +119,24 @@ static void umc_v6_7_ecc_info_query_correctable_error_count(struct amdgpu_device
  		*error_count += 1;
 
  		umc_v6_7_query_error_status_helper(adev, mc_umc_status, umc_reg_offset);
+
+		if (ras->umc_ecc.record_ce_addr_supported)	{
+			uint64_t err_addr, soc_pa;
+			uint32_t channel_index =
+				adev->umc.channel_idx_tbl[umc_inst * adev->umc.channel_inst_num + ch_inst];
+

drivers/gpu/drm/amd/amdgpu/umc_v6_7.c static void umc_v6_7_ecc_info_query_correctable_error_count(struct amdgpu_device · line 137

Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/umc_v6_7.c +10 -0
+			err_addr = ras->umc_ecc.ecc[eccinfo_table_idx].mca_ceumc_addr;
+			err_addr = REG_GET_FIELD(err_addr, MCA_UMC_UMC0_MCUMC_ADDRT0, ErrorAddr);
+			/* translate umc channel address to soc pa, 3 parts are included */
+			soc_pa = ADDR_OF_8KB_BLOCK(err_addr) |
+					ADDR_OF_256B_BLOCK(channel_index) |
+					OFFSET_IN_256B_BLOCK(err_addr);
+
+			/* The umc channel bits are not original values, they are hashed */
+			SET_CHANNEL_HASH(channel_index, soc_pa);
+

UMC address to PA conversion is common regardless of UE/CE error addresses. You may want to pack it in a small function.

Regardless,

Thanks,
Lijo

Quoted 71 lines
drivers/gpu/drm/amd/amdgpu/umc_v6_7.c +32 -2
+			dev_info(adev->dev, "Error Address(PA): 0x%llx\n", soc_pa);
+		}
  	}
  }
 
@@ -251,7 +269,9 @@ static void umc_v6_7_ecc_info_query_ras_error_address(struct amdgpu_device *adev
 
  static void umc_v6_7_query_correctable_error_count(struct amdgpu_device *adev,
  						   uint32_t umc_reg_offset,
-						   unsigned long *error_count)
+						   unsigned long *error_count,
+						   uint32_t ch_inst,
+						   uint32_t umc_inst)
  {
  	uint32_t ecc_err_cnt_sel, ecc_err_cnt_sel_addr;
  	uint32_t ecc_err_cnt, ecc_err_cnt_addr;
@@ -295,6 +315,31 @@ static void umc_v6_7_query_correctable_error_count(struct amdgpu_device *adev,
  		*error_count += 1;
 
  		umc_v6_7_query_error_status_helper(adev, mc_umc_status, umc_reg_offset);
+
+		{
+			uint64_t err_addr, soc_pa;
+			uint32_t mc_umc_addrt0;
+			uint32_t channel_index;
+
+			mc_umc_addrt0 =
+				SOC15_REG_OFFSET(UMC, 0, regMCA_UMC_UMC0_MCUMC_ADDRT0);
+
+			channel_index =
+				adev->umc.channel_idx_tbl[umc_inst * adev->umc.channel_inst_num + ch_inst];
+
+			err_addr = RREG64_PCIE((mc_umc_addrt0 + umc_reg_offset) * 4);
+			err_addr = REG_GET_FIELD(err_addr, MCA_UMC_UMC0_MCUMC_ADDRT0, ErrorAddr);
+
+			/* translate umc channel address to soc pa, 3 parts are included */
+			soc_pa = ADDR_OF_8KB_BLOCK(err_addr) |
+					ADDR_OF_256B_BLOCK(channel_index) |
+					OFFSET_IN_256B_BLOCK(err_addr);
+
+			/* The umc channel bits are not original values, they are hashed */
+			SET_CHANNEL_HASH(channel_index, soc_pa);
+
+			dev_info(adev->dev, "Error Address(PA): 0x%llx\n", soc_pa);
+		}
  	}
  }
 
@@ -395,7 +440,8 @@ static void umc_v6_7_query_ras_error_count(struct amdgpu_device *adev,
  							 ch_inst);
  		umc_v6_7_query_correctable_error_count(adev,
  						       umc_reg_offset,
-						       &(err_data->ce_count));
+						       &(err_data->ce_count),
+						       ch_inst, umc_inst);
  		umc_v6_7_querry_uncorrectable_error_count(adev,
  							  umc_reg_offset,
  							  &(err_data->ue_count));
drivers/gpu/drm/amd/pm/swsmu/smu13/aldebaran_ppt.c +1 -0
index 9cdfeea58085..c7e0fec614ea 100644
@@ -1883,6 +1883,7 @@ static ssize_t aldebaran_get_ecc_info(struct smu_context *smu,
  			ecc_info_per_channel->mca_ceumc_addr =
  				ecc_table->EccInfo_V2[i].mca_ceumc_addr;
  		}
+		eccinfo->record_ce_addr_supported =1;
  	}
 
  	return ret;
author Acked-by Lijo Lazar
#5 lore raw

答复: [PATCH Review v3 2/2] drm/amdgpu: print umc correctable error address

[AMD Official Use Only - General]

[AMD Official Use Only - General]

发件人: Lazar, Lijo <[email protected]>
日期: 星期三, 2022年5月25日 下午8:38
收件人: Yang, Stanley <[email protected]>, [email protected] <[email protected]>, Zhang, Hawking <[email protected]>, Zhou1, Tao <[email protected]>, Quan, Evan <[email protected]> 主题: Re: [PATCH Review v3 2/2] drm/amdgpu: print umc correctable error address

Quoted 43 lines · Stanley.Yang

Changed from V1:
remove unnecessary same row physical address calculation

Changed from V2:
move record_ce_addr_supported to umc_ecc_info struct

---
  drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h       |  5 ++
  drivers/gpu/drm/amd/amdgpu/umc_v6_7.c         | 50 ++++++++++++++++++-
  .../drm/amd/pm/swsmu/smu13/aldebaran_ppt.c    |  1 +
  3 files changed, 54 insertions(+), 2 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +5 -0
index 28e603243b67..bf5a95104ec1 100644
@@ -333,6 +333,11 @@ struct ecc_info_per_ch {
 
  struct umc_ecc_info {
       struct ecc_info_per_ch ecc[MAX_UMC_CHANNEL_NUM];
+
+     /* Determine smu ecctable whether support
+      * record correctable error address
+      */
+     int record_ce_addr_supported;
  };
 
  struct amdgpu_ras {
drivers/gpu/drm/amd/amdgpu/umc_v6_7.c +6 -0
index 606892dbea1c..bf7524f16b66 100644
@@ -119,6 +119,24 @@ static void umc_v6_7_ecc_info_query_correctable_error_count(struct amdgpu_device
               *error_count += 1;
 
               umc_v6_7_query_error_status_helper(adev, mc_umc_status, umc_reg_offset);
+
+             if (ras->umc_ecc.record_ce_addr_supported)      {
+                     uint64_t err_addr, soc_pa;
+                     uint32_t channel_index =
+                             adev->umc.channel_idx_tbl[umc_inst * adev->umc.channel_inst_num + ch_inst];
+

drivers/gpu/drm/amd/amdgpu/umc_v6_7.c static void umc_v6_7_ecc_info_query_correctable_error_count(struct amdgpu_device · line 137

Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/umc_v6_7.c +10 -0
+                     err_addr = ras->umc_ecc.ecc[eccinfo_table_idx].mca_ceumc_addr;
+                     err_addr = REG_GET_FIELD(err_addr, MCA_UMC_UMC0_MCUMC_ADDRT0, ErrorAddr);
+                     /* translate umc channel address to soc pa, 3 parts are included */
+                     soc_pa = ADDR_OF_8KB_BLOCK(err_addr) |
+                                     ADDR_OF_256B_BLOCK(channel_index) |
+                                     OFFSET_IN_256B_BLOCK(err_addr);
+
+                     /* The umc channel bits are not original values, they are hashed */
+                     SET_CHANNEL_HASH(channel_index, soc_pa);
+

UMC address to PA conversion is common regardless of UE/CE error addresses. You may want to pack it in a small function.

Regardless,

Thanks,
Lijo
Stanley: These lines are indeed redundant. I'll make a patch to simplify it.

Reagards,
Stanley

Quoted 71 lines
drivers/gpu/drm/amd/amdgpu/umc_v6_7.c +32 -2
+                     dev_info(adev->dev, "Error Address(PA): 0x%llx\n", soc_pa);
+             }
       }
  }
 
@@ -251,7 +269,9 @@ static void umc_v6_7_ecc_info_query_ras_error_address(struct amdgpu_device *adev
 
  static void umc_v6_7_query_correctable_error_count(struct amdgpu_device *adev,
                                                  uint32_t umc_reg_offset,
-                                                unsigned long *error_count)
+                                                unsigned long *error_count,
+                                                uint32_t ch_inst,
+                                                uint32_t umc_inst)
  {
       uint32_t ecc_err_cnt_sel, ecc_err_cnt_sel_addr;
       uint32_t ecc_err_cnt, ecc_err_cnt_addr;
@@ -295,6 +315,31 @@ static void umc_v6_7_query_correctable_error_count(struct amdgpu_device *adev,
               *error_count += 1;
 
               umc_v6_7_query_error_status_helper(adev, mc_umc_status, umc_reg_offset);
+
+             {
+                     uint64_t err_addr, soc_pa;
+                     uint32_t mc_umc_addrt0;
+                     uint32_t channel_index;
+
+                     mc_umc_addrt0 =
+                             SOC15_REG_OFFSET(UMC, 0, regMCA_UMC_UMC0_MCUMC_ADDRT0);
+
+                     channel_index =
+                             adev->umc.channel_idx_tbl[umc_inst * adev->umc.channel_inst_num + ch_inst];
+
+                     err_addr = RREG64_PCIE((mc_umc_addrt0 + umc_reg_offset) * 4);
+                     err_addr = REG_GET_FIELD(err_addr, MCA_UMC_UMC0_MCUMC_ADDRT0, ErrorAddr);
+
+                     /* translate umc channel address to soc pa, 3 parts are included */
+                     soc_pa = ADDR_OF_8KB_BLOCK(err_addr) |
+                                     ADDR_OF_256B_BLOCK(channel_index) |
+                                     OFFSET_IN_256B_BLOCK(err_addr);
+
+                     /* The umc channel bits are not original values, they are hashed */
+                     SET_CHANNEL_HASH(channel_index, soc_pa);
+
+                     dev_info(adev->dev, "Error Address(PA): 0x%llx\n", soc_pa);
+             }
       }
  }
 
@@ -395,7 +440,8 @@ static void umc_v6_7_query_ras_error_count(struct amdgpu_device *adev,
                                                        ch_inst);
               umc_v6_7_query_correctable_error_count(adev,
                                                      umc_reg_offset,
-                                                    &(err_data->ce_count));
+                                                    &(err_data->ce_count),
+                                                    ch_inst, umc_inst);
               umc_v6_7_querry_uncorrectable_error_count(adev,
                                                         umc_reg_offset,
                                                         &(err_data->ue_count));
drivers/gpu/drm/amd/pm/swsmu/smu13/aldebaran_ppt.c +1 -0
index 9cdfeea58085..c7e0fec614ea 100644
@@ -1883,6 +1883,7 @@ static ssize_t aldebaran_get_ecc_info(struct smu_context *smu,
                       ecc_info_per_channel->mca_ceumc_addr =
                               ecc_table->EccInfo_V2[i].mca_ceumc_addr;
               }
+             eccinfo->record_ce_addr_supported =1;
       }
 
       return ret;
Reviewed-by on v3 1/2
#6 lore raw
In reply to Stanley.Yang
Quoted 4 lines · Stanley.Yang

SMU add a new variable mca_ceumc_addr to record umc correctable error address in EccInfo table, driver side add EccInfo_V2_t to support this feature

Quoted 8 lines

Changed from V1:
remove ecc_table_v2 and unnecessary table id, define union struct include EccInfo_t and EccInfo_V2_t.

Changed from V2:
sync patch verion

Thanks,
Lijo

Quoted 140 lines
---
  drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h       |  1 +
  .../inc/pmfw_if/smu13_driver_if_aldebaran.h   | 16 +++++-
  .../drm/amd/pm/swsmu/smu13/aldebaran_ppt.c    | 53 ++++++++++++++-----
  3 files changed, 57 insertions(+), 13 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +1 -0
index b9a6fac2b8b2..28e603243b67 100644
@@ -328,6 +328,7 @@ struct ecc_info_per_ch {
  	uint16_t ce_count_hi_chip;
  	uint64_t mca_umc_status;
  	uint64_t mca_umc_addr;
+	uint64_t mca_ceumc_addr;
  };
 
  struct umc_ecc_info {
drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu13_driver_if_aldebaran.h +15 -1
index 0f67c56c2863..6f92038470ec 100644
@@ -519,7 +519,21 @@ typedef struct {
  } EccInfo_t;
 
  typedef struct {
-	EccInfo_t  EccInfo[ALDEBARAN_UMC_CHANNEL_NUM];
+	uint64_t mca_umc_status;
+	uint64_t mca_umc_addr;
+	uint64_t mca_ceumc_addr;
+
+	uint16_t ce_count_lo_chip;
+	uint16_t ce_count_hi_chip;
+
+	uint32_t eccPadding;
+} EccInfo_V2_t;
+
+typedef struct {
+	union {
+		EccInfo_t  EccInfo[ALDEBARAN_UMC_CHANNEL_NUM];
+		EccInfo_V2_t EccInfo_V2[ALDEBARAN_UMC_CHANNEL_NUM];
+	};
  } EccInfoTable_t;
 
  // These defines are used with the following messages:
drivers/gpu/drm/amd/pm/swsmu/smu13/aldebaran_ppt.c +41 -12
index 38af648cb857..9cdfeea58085 100644
@@ -82,6 +82,12 @@
   */
  #define SUPPORT_ECCTABLE_SMU_VERSION 0x00442a00
 
+/*
+ * SMU support mca_ceumc_addr in ECCTABLE since version 68.55.0,
+ * use this to check mca_ceumc_addr record whether support
+ */
+#define SUPPORT_ECCTABLE_V2_SMU_VERSION 0x00443700
+
  /*
   * SMU support BAD CHENNEL info MSG since version 68.51.00,
   * use this to check ECCTALE feature whether support
@@ -1802,7 +1808,8 @@ static ssize_t aldebaran_get_gpu_metrics(struct smu_context *smu,
  	return sizeof(struct gpu_metrics_v1_3);
  }
 
-static int aldebaran_check_ecc_table_support(struct smu_context *smu)
+static int aldebaran_check_ecc_table_support(struct smu_context *smu,
+		int *ecctable_version)
  {
  	uint32_t if_version = 0xff, smu_version = 0xff;
  	int ret = 0;
@@ -1815,6 +1822,11 @@ static int aldebaran_check_ecc_table_support(struct smu_context *smu)
 
  	if (smu_version < SUPPORT_ECCTABLE_SMU_VERSION)
  		ret = -EOPNOTSUPP;
+	else if (smu_version >= SUPPORT_ECCTABLE_SMU_VERSION &&
+			smu_version < SUPPORT_ECCTABLE_V2_SMU_VERSION)
+		*ecctable_version = 1;
+	else
+		*ecctable_version = 2;
 
  	return ret;
  }
@@ -1826,9 +1838,10 @@ static ssize_t aldebaran_get_ecc_info(struct smu_context *smu,
  	EccInfoTable_t *ecc_table = NULL;
  	struct ecc_info_per_ch *ecc_info_per_channel = NULL;
  	int i, ret = 0;
+	int table_version = 0;
  	struct umc_ecc_info *eccinfo = (struct umc_ecc_info *)table;
 
-	ret = aldebaran_check_ecc_table_support(smu);
+	ret = aldebaran_check_ecc_table_support(smu, &table_version);
  	if (ret)
  		return ret;
 
@@ -1844,16 +1857,32 @@ static ssize_t aldebaran_get_ecc_info(struct smu_context *smu,
 
  	ecc_table = (EccInfoTable_t *)smu_table->ecc_table;
 
-	for (i = 0; i < ALDEBARAN_UMC_CHANNEL_NUM; i++) {
-		ecc_info_per_channel = &(eccinfo->ecc[i]);
-		ecc_info_per_channel->ce_count_lo_chip =
-			ecc_table->EccInfo[i].ce_count_lo_chip;
-		ecc_info_per_channel->ce_count_hi_chip =
-			ecc_table->EccInfo[i].ce_count_hi_chip;
-		ecc_info_per_channel->mca_umc_status =
-			ecc_table->EccInfo[i].mca_umc_status;
-		ecc_info_per_channel->mca_umc_addr =
-			ecc_table->EccInfo[i].mca_umc_addr;
+	if (table_version == 1) {
+		for (i = 0; i < ALDEBARAN_UMC_CHANNEL_NUM; i++) {
+			ecc_info_per_channel = &(eccinfo->ecc[i]);
+			ecc_info_per_channel->ce_count_lo_chip =
+				ecc_table->EccInfo[i].ce_count_lo_chip;
+			ecc_info_per_channel->ce_count_hi_chip =
+				ecc_table->EccInfo[i].ce_count_hi_chip;
+			ecc_info_per_channel->mca_umc_status =
+				ecc_table->EccInfo[i].mca_umc_status;
+			ecc_info_per_channel->mca_umc_addr =
+				ecc_table->EccInfo[i].mca_umc_addr;
+		}
+	} else if (table_version == 2) {
+		for (i = 0; i < ALDEBARAN_UMC_CHANNEL_NUM; i++) {
+			ecc_info_per_channel = &(eccinfo->ecc[i]);
+			ecc_info_per_channel->ce_count_lo_chip =
+				ecc_table->EccInfo_V2[i].ce_count_lo_chip;
+			ecc_info_per_channel->ce_count_hi_chip =
+				ecc_table->EccInfo_V2[i].ce_count_hi_chip;
+			ecc_info_per_channel->mca_umc_status =
+				ecc_table->EccInfo_V2[i].mca_umc_status;
+			ecc_info_per_channel->mca_umc_addr =
+				ecc_table->EccInfo_V2[i].mca_umc_addr;
+			ecc_info_per_channel->mca_ceumc_addr =
+				ecc_table->EccInfo_V2[i].mca_ceumc_addr;
+		}
  	}
 
  	return ret;