[PATCH v2 04/12] drm/amdgpu: Introduce funcs for generating cper record
Xiang Liu
xiang.liu at amd.com
Fri Feb 14 08:07:11 UTC 2025
From: Hawking Zhang <Hawking.Zhang at amd.com>
Introduce new functions that are used to generate
cper ue or ce records.
v2: return -ENOMEM instead of false
v2: check return value of fill section function
Signed-off-by: Hawking Zhang <Hawking.Zhang at amd.com>
Signed-off-by: Xiang Liu <xiang.liu at amd.com>
Reviewed-by: Yang Wang <keivnyang.wang at amd.com>
Reviewed-by: Tao Zhou <tao.zhou1 at amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_aca.c | 12 +--
drivers/gpu/drm/amd/amdgpu/amdgpu_aca.h | 12 ++-
drivers/gpu/drm/amd/amdgpu/amdgpu_cper.c | 108 +++++++++++++++++++++++
drivers/gpu/drm/amd/amdgpu/amdgpu_cper.h | 9 +-
4 files changed, 128 insertions(+), 13 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_aca.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_aca.c
index 1a26b8ad14cb..ed1c20bd8114 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_aca.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_aca.c
@@ -30,16 +30,6 @@
typedef int bank_handler_t(struct aca_handle *handle, struct aca_bank *bank, enum aca_smu_type type, void *data);
-struct aca_banks {
- int nr_banks;
- struct list_head list;
-};
-
-struct aca_hwip {
- int hwid;
- int mcatype;
-};
-
static struct aca_hwip aca_hwid_mcatypes[ACA_HWIP_TYPE_COUNT] = {
ACA_BANK_HWID(SMU, 0x01, 0x01),
ACA_BANK_HWID(PCS_XGMI, 0x50, 0x00),
@@ -111,7 +101,7 @@ static struct aca_regs_dump {
{"STATUS", ACA_REG_IDX_STATUS},
{"ADDR", ACA_REG_IDX_ADDR},
{"MISC", ACA_REG_IDX_MISC0},
- {"CONFIG", ACA_REG_IDX_CONFG},
+ {"CONFIG", ACA_REG_IDX_CONFIG},
{"IPID", ACA_REG_IDX_IPID},
{"SYND", ACA_REG_IDX_SYND},
{"DESTAT", ACA_REG_IDX_DESTAT},
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_aca.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_aca.h
index 3cd0115b0244..b84a3489b116 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_aca.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_aca.h
@@ -81,7 +81,7 @@ enum aca_reg_idx {
ACA_REG_IDX_STATUS = 1,
ACA_REG_IDX_ADDR = 2,
ACA_REG_IDX_MISC0 = 3,
- ACA_REG_IDX_CONFG = 4,
+ ACA_REG_IDX_CONFIG = 4,
ACA_REG_IDX_IPID = 5,
ACA_REG_IDX_SYND = 6,
ACA_REG_IDX_DESTAT = 8,
@@ -114,6 +114,11 @@ enum aca_smu_type {
ACA_SMU_TYPE_COUNT,
};
+struct aca_hwip {
+ int hwid;
+ int mcatype;
+};
+
struct aca_bank {
enum aca_error_type aca_err_type;
enum aca_smu_type smu_err_type;
@@ -125,6 +130,11 @@ struct aca_bank_node {
struct list_head node;
};
+struct aca_banks {
+ int nr_banks;
+ struct list_head list;
+};
+
struct aca_bank_info {
int die_id;
int socket_id;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cper.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cper.c
index 8ce5dc6efcf9..f82aa12a88f4 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cper.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cper.c
@@ -21,6 +21,7 @@
* OTHER DEALINGS IN THE SOFTWARE.
*
*/
+#include <linux/list.h>
#include "amdgpu.h"
static const guid_t MCE = CPER_NOTIFY_MCE;
@@ -257,6 +258,113 @@ struct cper_hdr *amdgpu_cper_alloc_entry(struct amdgpu_device *adev,
return hdr;
}
+int amdgpu_cper_generate_ue_record(struct amdgpu_device *adev,
+ struct aca_bank *bank)
+{
+ struct cper_hdr *fatal = NULL;
+ struct cper_sec_crashdump_reg_data reg_data = { 0 };
+ int ret;
+
+ fatal = amdgpu_cper_alloc_entry(adev, AMDGPU_CPER_TYPE_FATAL, 1);
+ if (!fatal) {
+ dev_err(adev->dev, "fail to alloc cper entry for ue record\n");
+ return -ENOMEM;
+ }
+
+ reg_data.status_lo = lower_32_bits(bank->regs[ACA_REG_IDX_STATUS]);
+ reg_data.status_hi = upper_32_bits(bank->regs[ACA_REG_IDX_STATUS]);
+ reg_data.addr_lo = lower_32_bits(bank->regs[ACA_REG_IDX_ADDR]);
+ reg_data.addr_hi = upper_32_bits(bank->regs[ACA_REG_IDX_ADDR]);
+ reg_data.ipid_lo = lower_32_bits(bank->regs[ACA_REG_IDX_IPID]);
+ reg_data.ipid_hi = upper_32_bits(bank->regs[ACA_REG_IDX_IPID]);
+ reg_data.synd_lo = lower_32_bits(bank->regs[ACA_REG_IDX_SYND]);
+ reg_data.synd_hi = upper_32_bits(bank->regs[ACA_REG_IDX_SYND]);
+
+ amdgpu_cper_entry_fill_hdr(adev, fatal, AMDGPU_CPER_TYPE_FATAL, CPER_SEV_FATAL);
+ ret = amdgpu_cper_entry_fill_fatal_section(adev, fatal, 0, reg_data);
+ if (ret)
+ return ret;
+
+ /*TODO: commit the cper entry to cper ring */
+
+ return 0;
+}
+
+static enum cper_error_severity amdgpu_aca_err_type_to_cper_sev(struct amdgpu_device *adev,
+ enum aca_error_type aca_err_type)
+{
+ switch (aca_err_type) {
+ case ACA_ERROR_TYPE_UE:
+ return CPER_SEV_FATAL;
+ case ACA_ERROR_TYPE_CE:
+ return CPER_SEV_NON_FATAL_CORRECTED;
+ case ACA_ERROR_TYPE_DEFERRED:
+ return CPER_SEV_NON_FATAL_UNCORRECTED;
+ default:
+ dev_err(adev->dev, "Unknown ACA error type!\n");
+ return CPER_SEV_FATAL;
+ }
+}
+
+int amdgpu_cper_generate_ce_records(struct amdgpu_device *adev,
+ struct aca_banks *banks,
+ uint16_t bank_count)
+{
+ struct cper_hdr *corrected = NULL;
+ enum cper_error_severity sev = CPER_SEV_NON_FATAL_CORRECTED;
+ uint32_t reg_data[CPER_ACA_REG_COUNT] = { 0 };
+ struct aca_bank_node *node;
+ struct aca_bank *bank;
+ uint32_t i;
+ int ret;
+
+ corrected = amdgpu_cper_alloc_entry(adev, AMDGPU_CPER_TYPE_RUNTIME, bank_count);
+ if (!corrected) {
+ dev_err(adev->dev, "fail to allocate cper entry for ce records\n");
+ return -ENOMEM;
+ }
+
+ /* Raise severity if any DE is detected in the ACA bank list */
+ list_for_each_entry(node, &banks->list, node) {
+ bank = &node->bank;
+ if (bank->aca_err_type == ACA_ERROR_TYPE_DEFERRED) {
+ sev = CPER_SEV_NON_FATAL_UNCORRECTED;
+ break;
+ }
+ }
+
+ amdgpu_cper_entry_fill_hdr(adev, corrected, AMDGPU_CPER_TYPE_RUNTIME, sev);
+
+ /* Combine CE and UE in cper record */
+ list_for_each_entry(node, &banks->list, node) {
+ bank = &node->bank;
+ reg_data[CPER_ACA_REG_CTL_LO] = lower_32_bits(bank->regs[ACA_REG_IDX_CTL]);
+ reg_data[CPER_ACA_REG_CTL_HI] = upper_32_bits(bank->regs[ACA_REG_IDX_CTL]);
+ reg_data[CPER_ACA_REG_STATUS_LO] = lower_32_bits(bank->regs[ACA_REG_IDX_STATUS]);
+ reg_data[CPER_ACA_REG_STATUS_HI] = upper_32_bits(bank->regs[ACA_REG_IDX_STATUS]);
+ reg_data[CPER_ACA_REG_ADDR_LO] = lower_32_bits(bank->regs[ACA_REG_IDX_ADDR]);
+ reg_data[CPER_ACA_REG_ADDR_HI] = upper_32_bits(bank->regs[ACA_REG_IDX_ADDR]);
+ reg_data[CPER_ACA_REG_MISC0_LO] = lower_32_bits(bank->regs[ACA_REG_IDX_MISC0]);
+ reg_data[CPER_ACA_REG_MISC0_HI] = upper_32_bits(bank->regs[ACA_REG_IDX_MISC0]);
+ reg_data[CPER_ACA_REG_CONFIG_LO] = lower_32_bits(bank->regs[ACA_REG_IDX_CONFIG]);
+ reg_data[CPER_ACA_REG_CONFIG_HI] = upper_32_bits(bank->regs[ACA_REG_IDX_CONFIG]);
+ reg_data[CPER_ACA_REG_IPID_LO] = lower_32_bits(bank->regs[ACA_REG_IDX_IPID]);
+ reg_data[CPER_ACA_REG_IPID_HI] = upper_32_bits(bank->regs[ACA_REG_IDX_IPID]);
+ reg_data[CPER_ACA_REG_SYND_LO] = lower_32_bits(bank->regs[ACA_REG_IDX_SYND]);
+ reg_data[CPER_ACA_REG_SYND_HI] = upper_32_bits(bank->regs[ACA_REG_IDX_SYND]);
+
+ ret = amdgpu_cper_entry_fill_runtime_section(adev, corrected, i,
+ amdgpu_aca_err_type_to_cper_sev(adev, bank->aca_err_type),
+ reg_data, CPER_ACA_REG_COUNT);
+ if (ret)
+ return ret;
+ }
+
+ /*TODO: commit the cper entry to cper ring */
+
+ return 0;
+}
+
int amdgpu_cper_init(struct amdgpu_device *adev)
{
mutex_init(&adev->cper.cper_lock);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cper.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_cper.h
index 0ae845420983..6860a809f2f5 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cper.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cper.h
@@ -26,6 +26,7 @@
#define __AMDGPU_CPER_H__
#include "amd_cper.h"
+#include "amdgpu_aca.h"
#define CPER_MAX_ALLOWED_COUNT 0x1000
#define HDR_LEN (sizeof(struct cper_hdr))
@@ -84,7 +85,13 @@ int amdgpu_cper_entry_fill_bad_page_threshold_section(struct amdgpu_device *adev
struct cper_hdr *amdgpu_cper_alloc_entry(struct amdgpu_device *adev,
enum amdgpu_cper_type type,
uint16_t section_count);
-
+/* UE must be encoded into separated cper entries, 1 UE 1 cper */
+int amdgpu_cper_generate_ue_record(struct amdgpu_device *adev,
+ struct aca_bank *bank);
+/* CEs and DEs are combined into 1 cper entry */
+int amdgpu_cper_generate_ce_records(struct amdgpu_device *adev,
+ struct aca_banks *banks,
+ uint16_t bank_count);
int amdgpu_cper_init(struct amdgpu_device *adev);
int amdgpu_cper_fini(struct amdgpu_device *adev);
--
2.34.1
More information about the amd-gfx
mailing list