[PATCH 1/4] drm/amdgpu: add amdgpu smu mca dump feature support

Yang Wang kevinyang.wang at amd.com
Wed Sep 13 09:31:57 UTC 2023


add amdgpu smu mca dump feature support.

Signed-off-by: Yang Wang <kevinyang.wang at amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang at amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_mca.c | 68 +++++++++++++++++++++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_mca.h | 57 +++++++++++++++++++++
 2 files changed, 125 insertions(+)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.c
index 8d9ff9e151de..9fa88ae81b12 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.c
@@ -142,3 +142,71 @@ int amdgpu_mca_mpio_ras_sw_init(struct amdgpu_device *adev)
 
 	return 0;
 }
+
+void amdgpu_mca_smu_init_funcs(struct amdgpu_device *adev, const struct amdgpu_mca_smu_funcs *mca_funcs)
+{
+	struct amdgpu_mca *mca = &adev->mca;
+
+	mca->mca_funcs = mca_funcs;
+}
+
+int amdgpu_mca_smu_set_debug_mode(struct amdgpu_device *adev, bool enable)
+{
+	const struct amdgpu_mca_smu_funcs *mca_funcs = adev->mca.mca_funcs;
+
+	if (mca_funcs && mca_funcs->mca_set_debug_mode)
+		return mca_funcs->mca_set_debug_mode(adev, enable);
+
+	return -ENOTSUPP;
+}
+
+int amdgpu_mca_smu_get_valid_mca_count(struct amdgpu_device *adev, enum amdgpu_mca_error_type type, uint32_t *count)
+{
+	const struct amdgpu_mca_smu_funcs *mca_funcs = adev->mca.mca_funcs;
+
+	if (!count)
+		return -EINVAL;
+
+	if (mca_funcs && mca_funcs->mca_get_valid_mca_count)
+		return mca_funcs->mca_get_valid_mca_count(adev, type, count);
+
+	return -ENOTSUPP;
+}
+
+int amdgpu_mca_smu_get_error_count(struct amdgpu_device *adev, enum amdgpu_ras_block blk, enum amdgpu_mca_error_type type, uint32_t *count)
+{
+	const struct amdgpu_mca_smu_funcs *mca_funcs = adev->mca.mca_funcs;
+	if (!count)
+		return -EINVAL;
+
+	if (mca_funcs && mca_funcs->mca_get_error_count)
+		return mca_funcs->mca_get_error_count(adev, blk, type, count);
+
+	return -ENOTSUPP;
+}
+
+int amdgpu_mca_smu_get_mca_entry(struct amdgpu_device *adev,enum amdgpu_mca_error_type type, int idx, struct mca_bank_entry *entry)
+{
+	const struct amdgpu_mca_smu_funcs *mca_funcs = adev->mca.mca_funcs;
+	int count;
+
+	switch (type) {
+	case AMDGPU_MCA_ERROR_TYPE_UE:
+		count = mca_funcs->max_ue_count;
+		break;
+	case AMDGPU_MCA_ERROR_TYPE_CE:
+		count = mca_funcs->max_ce_count;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	if (idx >= count)
+		return -EINVAL;
+
+	if (mca_funcs && mca_funcs->mca_get_mca_entry)
+		return mca_funcs->mca_get_mca_entry(adev, type, idx, entry);
+
+	return -ENOTSUPP;
+}
+
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.h
index 997a073e2409..6915ae0d5b92 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.h
@@ -21,6 +21,26 @@
 #ifndef __AMDGPU_MCA_H__
 #define __AMDGPU_MCA_H__
 
+#include "amdgpu_ras.h"
+
+#define MCA_MAX_REGS_COUNT	(16)
+
+enum amdgpu_mca_ip {
+	AMDGPU_MCA_IP_UNKNOW = -1,
+	AMDGPU_MCA_IP_PSP = 0,
+	AMDGPU_MCA_IP_SDMA,
+	AMDGPU_MCA_IP_GC,
+	AMDGPU_MCA_IP_SMU,
+	AMDGPU_MCA_IP_MP5,
+	AMDGPU_MCA_IP_UMC,
+	AMDGPU_MCA_IP_COUNT,
+};
+
+enum amdgpu_mca_error_type {
+	AMDGPU_MCA_ERROR_TYPE_UE = 0,
+	AMDGPU_MCA_ERROR_TYPE_CE,
+};
+
 struct amdgpu_mca_ras_block {
 	struct amdgpu_ras_block_object ras_block;
 };
@@ -34,6 +54,36 @@ struct amdgpu_mca {
 	struct amdgpu_mca_ras mp0;
 	struct amdgpu_mca_ras mp1;
 	struct amdgpu_mca_ras mpio;
+	const struct amdgpu_mca_smu_funcs *mca_funcs;
+};
+
+struct mca_bank_info {
+	int socket_id;
+	int aid;
+	int hwid;
+	int mcatype;
+};
+
+struct mca_bank_entry {
+	int idx;
+	enum amdgpu_mca_error_type type;
+	enum amdgpu_mca_ip ip;
+	struct mca_bank_info info;
+	uint64_t regs[MCA_MAX_REGS_COUNT];
+};
+
+struct amdgpu_mca_smu_funcs {
+	int max_ue_count;
+	int max_ce_count;
+	int (*mca_set_debug_mode)(struct amdgpu_device *adev, bool enable);
+	int (*mca_get_error_count)(struct amdgpu_device *adev, enum amdgpu_ras_block blk,
+				   enum amdgpu_mca_error_type type, uint32_t *count);
+	int (*mca_get_valid_mca_count)(struct amdgpu_device *adev, enum amdgpu_mca_error_type type,
+				       uint32_t *count);
+	int (*mca_get_mca_entry)(struct amdgpu_device *adev, enum amdgpu_mca_error_type type,
+				 int idx, struct mca_bank_entry *entry);
+	int (*mca_get_ras_mca_idx_array)(struct amdgpu_device *adev, enum amdgpu_ras_block blk,
+					 enum amdgpu_mca_error_type type, int *idx_array, int *idx_array_size);
 };
 
 void amdgpu_mca_query_correctable_error_count(struct amdgpu_device *adev,
@@ -53,4 +103,11 @@ void amdgpu_mca_query_ras_error_count(struct amdgpu_device *adev,
 int amdgpu_mca_mp0_ras_sw_init(struct amdgpu_device *adev);
 int amdgpu_mca_mp1_ras_sw_init(struct amdgpu_device *adev);
 int amdgpu_mca_mpio_ras_sw_init(struct amdgpu_device *adev);
+
+void amdgpu_mca_smu_init_funcs(struct amdgpu_device *adev, const struct amdgpu_mca_smu_funcs *mca_funcs);
+int amdgpu_mca_smu_set_debug_mode(struct amdgpu_device *adev, bool enable);
+int amdgpu_mca_smu_get_valid_mca_count(struct amdgpu_device *adev, enum amdgpu_mca_error_type type, uint32_t *count);
+int amdgpu_mca_smu_get_error_count(struct amdgpu_device *adev, enum amdgpu_ras_block blk, enum amdgpu_mca_error_type type, uint32_t *count);
+int amdgpu_mca_smu_get_mca_entry(struct amdgpu_device *adev,enum amdgpu_mca_error_type type, int idx, struct mca_bank_entry *entry);
+
 #endif
-- 
2.34.1



More information about the amd-gfx mailing list