1 files changed, 74 insertions, 103 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v6_7.c b/drivers/gpu/drm/amd/amdgpu/umc_v6_7.c
index 5d5d031c9e7d..a3ee3c4c650f 100644
--- a/drivers/gpu/drm/amd/amdgpu/umc_v6_7.c
+++ b/drivers/gpu/drm/amd/amdgpu/umc_v6_7.c
@@ -57,13 +57,6 @@ static inline uint32_t get_umc_v6_7_reg_offset(struct amdgpu_device *adev,
 	return adev->umc.channel_offs * ch_inst + UMC_V6_7_INST_DIST * umc_inst;
 }
 
-static inline uint32_t get_umc_v6_7_channel_index(struct amdgpu_device *adev,
-					      uint32_t umc_inst,
-					      uint32_t ch_inst)
-{
-	return adev->umc.channel_idx_tbl[umc_inst * adev->umc.channel_inst_num + ch_inst];
-}
-
 static void umc_v6_7_query_error_status_helper(struct amdgpu_device *adev,
 						  uint64_t mc_umc_status, uint32_t umc_reg_offset)
 {
@@ -71,7 +64,7 @@ static void umc_v6_7_query_error_status_helper(struct amdgpu_device *adev,
 	uint64_t reg_value;
 
 	if (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Deferred) == 1)
-		dev_info(adev->dev, "Deferred error, no user action is needed.\n");
+		dev_info(adev->dev, "Deferred error\n");
 
 	if (mc_umc_status)
 		dev_info(adev->dev, "MCA STATUS 0x%llx, umc_reg_offset 0x%x\n", mc_umc_status, umc_reg_offset);
@@ -167,29 +160,33 @@ static void umc_v6_7_ecc_info_querry_uncorrectable_error_count(struct amdgpu_dev
 	}
 }
 
+static int umc_v6_7_ecc_info_querry_ecc_error_count(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;
+
+	umc_v6_7_ecc_info_query_correctable_error_count(adev,
+		umc_inst, ch_inst,
+		&(err_data->ce_count));
+
+	umc_v6_7_ecc_info_querry_uncorrectable_error_count(adev,
+		umc_inst, ch_inst,
+		&(err_data->ue_count));
+
+	return 0;
+}
+
 static void umc_v6_7_ecc_info_query_ras_error_count(struct amdgpu_device *adev,
 					   void *ras_error_status)
 {
-	struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status;
-
-	uint32_t umc_inst        = 0;
-	uint32_t ch_inst         = 0;
-
-	/*TODO: driver needs to toggle DF Cstate to ensure
-	 * safe access of UMC registers. Will add the protection */
-	LOOP_UMC_INST_AND_CH(umc_inst, ch_inst) {
-		umc_v6_7_ecc_info_query_correctable_error_count(adev,
-						      umc_inst, ch_inst,
-						      &(err_data->ce_count));
-		umc_v6_7_ecc_info_querry_uncorrectable_error_count(adev,
-						      umc_inst, ch_inst,
-							  &(err_data->ue_count));
-	}
+	amdgpu_umc_loop_channels(adev,
+		umc_v6_7_ecc_info_querry_ecc_error_count, ras_error_status);
 }
 
-static void umc_v6_7_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)
+void umc_v6_7_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 channel_index;
 	uint64_t soc_pa, retired_page, column;
@@ -222,23 +219,23 @@ static void umc_v6_7_convert_error_address(struct amdgpu_device *adev,
 	}
 }
 
-static void umc_v6_7_ecc_info_query_error_address(struct amdgpu_device *adev,
-					 struct ras_err_data *err_data,
-					 uint32_t ch_inst,
-					 uint32_t umc_inst)
+static int umc_v6_7_ecc_info_query_error_address(struct amdgpu_device *adev,
+					uint32_t node_inst, uint32_t umc_inst,
+					uint32_t ch_inst, void *data)
 {
 	uint64_t mc_umc_status, err_addr;
 	uint32_t eccinfo_table_idx;
 	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
+	struct ras_err_data *err_data = (struct ras_err_data *)data;
 
 	eccinfo_table_idx = umc_inst * adev->umc.channel_inst_num + ch_inst;
 	mc_umc_status = ras->umc_ecc.ecc[eccinfo_table_idx].mca_umc_status;
 
 	if (mc_umc_status == 0)
-		return;
+		return 0;
 
 	if (!err_data->err_addr)
-		return;
+		return 0;
 
 	/* calculate error address if ue error is detected */
 	if (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Val) == 1 &&
@@ -250,25 +247,15 @@ static void umc_v6_7_ecc_info_query_error_address(struct amdgpu_device *adev,
 		umc_v6_7_convert_error_address(adev, err_data, err_addr,
 					ch_inst, umc_inst);
 	}
+
+	return 0;
 }
 
 static void umc_v6_7_ecc_info_query_ras_error_address(struct amdgpu_device *adev,
 					     void *ras_error_status)
 {
-	struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status;
-
-	uint32_t umc_inst        = 0;
-	uint32_t ch_inst         = 0;
-
-	/*TODO: driver needs to toggle DF Cstate to ensure
-	 * safe access of UMC resgisters. Will add the protection
-	 * when firmware interface is ready */
-	LOOP_UMC_INST_AND_CH(umc_inst, ch_inst) {
-		umc_v6_7_ecc_info_query_error_address(adev,
-					     err_data,
-					     ch_inst,
-					     umc_inst);
-	}
+	amdgpu_umc_loop_channels(adev,
+	    umc_v6_7_ecc_info_query_error_address, ras_error_status);
 }
 
 static void umc_v6_7_query_correctable_error_count(struct amdgpu_device *adev,
@@ -371,11 +358,14 @@ static void umc_v6_7_querry_uncorrectable_error_count(struct amdgpu_device *adev
 	}
 }
 
-static void umc_v6_7_reset_error_count_per_channel(struct amdgpu_device *adev,
-						   uint32_t umc_reg_offset)
+static int umc_v6_7_reset_error_count_per_channel(struct amdgpu_device *adev,
+					uint32_t node_inst, uint32_t umc_inst,
+					uint32_t ch_inst, void *data)
 {
 	uint32_t ecc_err_cnt_addr;
 	uint32_t ecc_err_cnt_sel, ecc_err_cnt_sel_addr;
+	uint32_t umc_reg_offset =
+		get_umc_v6_7_reg_offset(adev, umc_inst, ch_inst);
 
 	ecc_err_cnt_sel_addr =
 		SOC15_REG_OFFSET(UMC, 0,
@@ -409,58 +399,54 @@ static void umc_v6_7_reset_error_count_per_channel(struct amdgpu_device *adev,
 	/* clear higher chip error count */
 	WREG32_PCIE((ecc_err_cnt_addr + umc_reg_offset) * 4,
 			UMC_V6_7_CE_CNT_INIT);
+
+	return 0;
 }
 
 static void umc_v6_7_reset_error_count(struct amdgpu_device *adev)
 {
-	uint32_t umc_inst        = 0;
-	uint32_t ch_inst         = 0;
-	uint32_t umc_reg_offset  = 0;
+	amdgpu_umc_loop_channels(adev,
+		umc_v6_7_reset_error_count_per_channel, NULL);
+}
 
-	LOOP_UMC_INST_AND_CH(umc_inst, ch_inst) {
-		umc_reg_offset = get_umc_v6_7_reg_offset(adev,
-							 umc_inst,
-							 ch_inst);
+static int umc_v6_7_query_ecc_error_count(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;
+	uint32_t umc_reg_offset =
+		get_umc_v6_7_reg_offset(adev, umc_inst, ch_inst);
 
-		umc_v6_7_reset_error_count_per_channel(adev,
-						       umc_reg_offset);
-	}
+	umc_v6_7_query_correctable_error_count(adev,
+					umc_reg_offset,
+					&(err_data->ce_count),
+					ch_inst, umc_inst);
+
+	umc_v6_7_querry_uncorrectable_error_count(adev,
+					umc_reg_offset,
+					&(err_data->ue_count));
+
+	return 0;
 }
 
 static void umc_v6_7_query_ras_error_count(struct amdgpu_device *adev,
 					   void *ras_error_status)
 {
-	struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status;
-
-	uint32_t umc_inst        = 0;
-	uint32_t ch_inst         = 0;
-	uint32_t umc_reg_offset  = 0;
-
-	/*TODO: driver needs to toggle DF Cstate to ensure
-	 * safe access of UMC registers. Will add the protection */
-	LOOP_UMC_INST_AND_CH(umc_inst, ch_inst) {
-		umc_reg_offset = get_umc_v6_7_reg_offset(adev,
-							 umc_inst,
-							 ch_inst);
-		umc_v6_7_query_correctable_error_count(adev,
-						       umc_reg_offset,
-						       &(err_data->ce_count),
-						       ch_inst, umc_inst);
-		umc_v6_7_querry_uncorrectable_error_count(adev,
-							  umc_reg_offset,
-							  &(err_data->ue_count));
-	}
+	amdgpu_umc_loop_channels(adev,
+		umc_v6_7_query_ecc_error_count, ras_error_status);
 
 	umc_v6_7_reset_error_count(adev);
 }
 
-static void umc_v6_7_query_error_address(struct amdgpu_device *adev,
-					 struct ras_err_data *err_data,
-					 uint32_t umc_reg_offset, uint32_t ch_inst,
-					 uint32_t umc_inst)
+static int umc_v6_7_query_error_address(struct amdgpu_device *adev,
+					uint32_t node_inst, uint32_t umc_inst,
+					uint32_t ch_inst, void *data)
 {
 	uint32_t mc_umc_status_addr;
 	uint64_t mc_umc_status = 0, mc_umc_addrt0, err_addr;
+	struct ras_err_data *err_data = (struct ras_err_data *)data;
+	uint32_t umc_reg_offset =
+		get_umc_v6_7_reg_offset(adev, umc_inst, ch_inst);
 
 	mc_umc_status_addr =
 		SOC15_REG_OFFSET(UMC, 0, regMCA_UMC_UMC0_MCUMC_STATUST0);
@@ -470,12 +456,12 @@ static void umc_v6_7_query_error_address(struct amdgpu_device *adev,
 	mc_umc_status = RREG64_PCIE((mc_umc_status_addr + umc_reg_offset) * 4);
 
 	if (mc_umc_status == 0)
-		return;
+		return 0;
 
 	if (!err_data->err_addr) {
 		/* clear umc status */
 		WREG64_PCIE((mc_umc_status_addr + umc_reg_offset) * 4, 0x0ULL);
-		return;
+		return 0;
 	}
 
 	/* calculate error address if ue error is detected */
@@ -491,29 +477,15 @@ static void umc_v6_7_query_error_address(struct amdgpu_device *adev,
 
 	/* clear umc status */
 	WREG64_PCIE((mc_umc_status_addr + umc_reg_offset) * 4, 0x0ULL);
+
+	return 0;
 }
 
 static void umc_v6_7_query_ras_error_address(struct amdgpu_device *adev,
 					     void *ras_error_status)
 {
-	struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status;
-
-	uint32_t umc_inst        = 0;
-	uint32_t ch_inst         = 0;
-	uint32_t umc_reg_offset  = 0;
-
-	/*TODO: driver needs to toggle DF Cstate to ensure
-	 * safe access of UMC resgisters. Will add the protection
-	 * when firmware interface is ready */
-	LOOP_UMC_INST_AND_CH(umc_inst, ch_inst) {
-		umc_reg_offset = get_umc_v6_7_reg_offset(adev,
-							 umc_inst,
-							 ch_inst);
-		umc_v6_7_query_error_address(adev,
-					     err_data,
-					     umc_reg_offset, ch_inst,
-					     umc_inst);
-	}
+	amdgpu_umc_loop_channels(adev,
+		umc_v6_7_query_error_address, ras_error_status);
 }
 
 static uint32_t umc_v6_7_query_ras_poison_mode_per_channel(
@@ -553,5 +525,4 @@ struct amdgpu_umc_ras umc_v6_7_ras = {
 	.query_ras_poison_mode = umc_v6_7_query_ras_poison_mode,
 	.ecc_info_query_ras_error_count = umc_v6_7_ecc_info_query_ras_error_count,
 	.ecc_info_query_ras_error_address = umc_v6_7_ecc_info_query_ras_error_address,
-	.convert_ras_error_address = umc_v6_7_convert_error_address,
 };