Skip to contentSkip to conversation

drm/amdgpu/ras: Add debug mask to disable CE logs for uniras

PATCH Superseded Ce Sun 2 participants 2 messages Jul 16

Subsystems (2)

v3 of 4 Superseded Latest is v4 →

lore
Share & tools
lore

Ce Sun

author
#0 lore raw

PATCH v3 1/1 drm/amdgpu/ras: Add debug mask to disable CE logs for uniras

Add debug mask to disable kernel logs of RAS correctable errors

---
 drivers/gpu/drm/amd/amdgpu/amdgpu.h           |  1 +
 .../gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c  |  1 +
 drivers/gpu/drm/amd/ras/rascore/ras.h         |  4 ++++
 drivers/gpu/drm/amd/ras/rascore/ras_aca.c     | 17 +++++++++-------
 .../gpu/drm/amd/ras/rascore/ras_aca_v1_0.c    |  2 +-
 drivers/gpu/drm/amd/ras/rascore/ras_core.c    | 20 +++++++++++++++++++
 6 files changed, 37 insertions(+), 8 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu.h +1 -0
index a3dafdca7eb3..bc6b7da3914f 100644
@@ -226,6 +226,7 @@ extern int amdgpu_use_xgmi_p2p;
 extern bool pcie_p2p;
 extern int amdgpu_mtype_local;
 extern int amdgpu_enforce_isolation;
+extern uint amdgpu_debug_mask;
 #ifdef CONFIG_HSA_AMD
 extern int sched_policy;
 extern bool debug_evictions;
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c +1 -0
index ea95e0f93d1c..0db755ba7037 100644
@@ -309,6 +309,7 @@ static struct ras_core_context *amdgpu_ras_mgr_create_ras_core(struct amdgpu_dev
 		amdgpu_ras_mgr_eeprom_is_supported(adev);
 	init_config.poison_supported =
 		amdgpu_ras_is_poison_mode_supported(adev);
+	init_config.ras_debug_mask = amdgpu_debug_mask;
 
 	amdgpu_ras_mgr_init_aca_config(adev, &init_config);
 	amdgpu_ras_mgr_init_eeprom_config(adev, &init_config);
drivers/gpu/drm/amd/ras/rascore/ras.h +4 -0
index 878dfdfcb18a..53abf1a8600b 100644
@@ -301,6 +301,7 @@ struct ras_core_config {
 
 	bool poison_supported;
 	bool ras_eeprom_supported;
+	uint ras_debug_mask;
 	const struct ras_sys_func *sys_fn;
 
 	struct ras_aca_config aca_cfg;
@@ -346,6 +347,8 @@ struct ras_core_context {
 	spinlock_t seqno_lock;
 
 	bool ras_core_enabled;
+	bool ras_ce_log_disabled;
+	uint ras_debug_mask;
 
 	u64 ras_fw_features;
 };
@@ -403,4 +406,5 @@ int ras_core_convert_soc_pa_to_cur_nps_pages(struct ras_core_context *ras_core,
 int ras_core_check_address_sanity(struct ras_core_context *ras_core, uint64_t addr);
 
 int ras_core_set_debug_mode(struct ras_core_context *ras_core, bool enable);
+bool ras_core_is_ce_log_disabled(struct ras_core_context *ras_core);
 #endif
drivers/gpu/drm/amd/ras/rascore/ras_aca.c +10 -7
index 67a35409ff0e..a1fb9a5f72cf 100644
@@ -83,22 +83,20 @@ static void aca_report_ecc_info(struct ras_core_context *ras_core,
 			seq_no, skt, aid, ecc_count.total_ue_count, blk_name(blk));
 	}
 
-	if (ecc_count.new_de_count) {
+	if (ecc_count.new_de_count && blk == RAS_BLOCK_ID__UMC) {
 		RAS_DEV_INFO(ras_core->dev,
-		"{%llu} socket: %d, die: %d, %u new %s detected in %s block\n",
+		"{%llu} socket: %d, die: %d, %u new deferred hardware errors detected in %s block\n",
 			seq_no, skt, aid, ecc_count.new_de_count,
-			(blk == RAS_BLOCK_ID__UMC) ?
-				"deferred hardware errors" : "poison consumption",
 			blk_name(blk));
 		RAS_DEV_INFO(ras_core->dev,
-		"{%llu} socket: %d, die: %d, %u %s detected in total in %s block\n",
+		"{%llu} socket: %d, die: %d, %u deferred hardware errors detected in total in %s block\n",
 			seq_no, skt, aid, ecc_count.total_de_count,
-			(blk == RAS_BLOCK_ID__UMC) ?
-				"deferred hardware errors" : "poison consumption",
 			blk_name(blk));
 	}
 
 	if (ecc_count.new_ce_count) {
+		if (ras_core_is_ce_log_disabled(ras_core))
+			return;
 		RAS_DEV_INFO(ras_core->dev,
 		"{%llu} socket: %d, die: %d, %u new correctable hardware errors detected in %s block\n",
 			seq_no, skt, aid, ecc_count.new_ce_count, blk_name(blk));
@@ -114,6 +112,11 @@ static void aca_bank_log(struct ras_core_context *ras_core,
 {
 	int i;
 
+	if(ras_core_is_ce_log_disabled(ras_core) &&
+	   bank->ecc_type == RAS_ERR_TYPE__CE &&
+	   !bank_ecc->de_count)
+		return;
+
 	RAS_DEV_INFO(ras_core->dev,
 		"{%llu}" RAS_HW_ERR "Accelerator Check Architecture events logged\n",
 		bank->seq_no);
drivers/gpu/drm/amd/ras/rascore/ras_aca_v1_0.c +1 -1
index 840610538c1f..288fad0862a2 100644
@@ -265,7 +265,7 @@ static int aca_parse_bank_default(struct ras_core_context *ras_core,
 	ecc->bank_info.addr = bank->regs[ACA_REG_IDX__ADDR];
 
 	if (aca_check_bank_is_de(ras_core, status)) {
-		ecc->de_count = 0;
+		ecc->de_count = 1;
 	} else {
 		if (bank->ecc_type == RAS_ERR_TYPE__UE)
 			ecc->ue_count = 1;
drivers/gpu/drm/amd/ras/rascore/ras_core.c +20 -0
index 08e17a83ad5b..f3c6cbc7e05f 100644
@@ -28,6 +28,10 @@
 
 #define IS_LEAP_YEAR(x) ((x % 4 == 0 && x % 100 != 0) || x % 400 == 0)
 
+enum RAS_DEBUG_MASK {
+	RAS_DEBUG_DISABLE_RAS_CE_LOG = BIT(9),
+};
+
 static const char * const ras_block_name[] = {
 	"umc",
 	"sdma",
@@ -374,6 +378,14 @@ int ras_core_sw_fini(struct ras_core_context *ras_core)
 	return 0;
 }
 
+static void ras_init_debug_options(struct ras_core_context *ras_core)
+{
+	if (ras_core->ras_debug_mask & RAS_DEBUG_DISABLE_RAS_CE_LOG) {
+		RAS_DEV_INFO(ras_core->dev, "debug: disable kernel logs of correctable errors\n");
+		ras_core->ras_ce_log_disabled = true;
+	}
+}
+
 int ras_core_hw_init(struct ras_core_context *ras_core)
 {
 	int ret;
@@ -382,6 +394,7 @@ int ras_core_hw_init(struct ras_core_context *ras_core)
 			ras_core->config->ras_eeprom_supported;
 
 	ras_core->poison_supported = ras_core->config->poison_supported;
+	ras_core->ras_debug_mask = ras_core->config->ras_debug_mask;
 
 	ret = ras_psp_hw_init(ras_core);
 	if (ret)
@@ -434,6 +447,8 @@ int ras_core_hw_init(struct ras_core_context *ras_core)
 	if (ret)
 		goto init_err7;
 
+	ras_init_debug_options(ras_core);
+
 	ras_core->is_initialized = true;
 
 	return 0;
@@ -545,6 +560,11 @@ bool ras_core_is_enabled(struct ras_core_context *ras_core)
 	return ras_core->ras_core_enabled;
 }
 
+bool ras_core_is_ce_log_disabled(struct ras_core_context *ras_core)
+{
+	return ras_core->ras_ce_log_disabled;
+}
+
 uint64_t ras_core_get_utc_second_timestamp(struct ras_core_context *ras_core)
 {
 	if (!ras_core)
Signature · 1 line
2.34.1
#1 lore raw

AMD General

Best Regards,
Thomas
-----Original Message-----
From: Sun, Ce(Overlord) <[email protected]>
Sent: Thursday, July 16, 2026 4:23 PM
To: [email protected]
Cc: Zhang, Hawking <[email protected]>; Chai, Thomas <[email protected]>; Zhou1, Tao <[email protected]>; Yang, Stanley <[email protected]>; Sun, Ce(Overlord) <[email protected]>
Subject: [PATCH V3] drm/amdgpu/ras: Add debug mask to disable CE logs for uniras

Add debug mask to disable kernel logs of RAS correctable errors

---
 drivers/gpu/drm/amd/amdgpu/amdgpu.h           |  1 +
 .../gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c  |  1 +
 drivers/gpu/drm/amd/ras/rascore/ras.h         |  4 ++++
 drivers/gpu/drm/amd/ras/rascore/ras_aca.c     | 17 +++++++++-------
 .../gpu/drm/amd/ras/rascore/ras_aca_v1_0.c    |  2 +-
 drivers/gpu/drm/amd/ras/rascore/ras_core.c    | 20 +++++++++++++++++++
 6 files changed, 37 insertions(+), 8 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu.h +1 -0
index a3dafdca7eb3..bc6b7da3914f 100644
@@ -226,6 +226,7 @@ extern int amdgpu_use_xgmi_p2p;  extern bool pcie_p2p;  extern int amdgpu_mtype_local;  extern int amdgpu_enforce_isolation;
+extern uint amdgpu_debug_mask;
 #ifdef CONFIG_HSA_AMD
 extern int sched_policy;
 extern bool debug_evictions;
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c +1 -0
index ea95e0f93d1c..0db755ba7037 100644
@@ -309,6 +309,7 @@ static struct ras_core_context *amdgpu_ras_mgr_create_ras_core(struct amdgpu_dev
                amdgpu_ras_mgr_eeprom_is_supported(adev);
        init_config.poison_supported =
                amdgpu_ras_is_poison_mode_supported(adev);
+       init_config.ras_debug_mask = amdgpu_debug_mask;
 
        amdgpu_ras_mgr_init_aca_config(adev, &init_config);
        amdgpu_ras_mgr_init_eeprom_config(adev, &init_config); diff --git a/drivers/gpu/drm/amd/ras/rascore/ras.h b/drivers/gpu/drm/amd/ras/rascore/ras.h

index 878dfdfcb18a..53abf1a8600b 100644

drivers/gpu/drm/amd/ras/rascore/ras.h +4 -0
@@ -301,6 +301,7 @@ struct ras_core_config {
 
        bool poison_supported;
        bool ras_eeprom_supported;
+       uint ras_debug_mask;
        const struct ras_sys_func *sys_fn;
 
        struct ras_aca_config aca_cfg;
@@ -346,6 +347,8 @@ struct ras_core_context {
        spinlock_t seqno_lock;
 
        bool ras_core_enabled;
+       bool ras_ce_log_disabled;
+       uint ras_debug_mask;
 
        u64 ras_fw_features;
 };
@@ -403,4 +406,5 @@ int ras_core_convert_soc_pa_to_cur_nps_pages(struct ras_core_context *ras_core,  int ras_core_check_address_sanity(struct ras_core_context *ras_core, uint64_t addr);
 
 int ras_core_set_debug_mode(struct ras_core_context *ras_core, bool enable);
+bool ras_core_is_ce_log_disabled(struct ras_core_context *ras_core);
 #endif
drivers/gpu/drm/amd/ras/rascore/ras_aca.c +7 -7
index 67a35409ff0e..a1fb9a5f72cf 100644
@@ -83,22 +83,20 @@ static void aca_report_ecc_info(struct ras_core_context *ras_core,
                        seq_no, skt, aid, ecc_count.total_ue_count, blk_name(blk));
        }
 
-       if (ecc_count.new_de_count) {
+       if (ecc_count.new_de_count && blk == RAS_BLOCK_ID__UMC) {
                RAS_DEV_INFO(ras_core->dev,
-               "{%llu} socket: %d, die: %d, %u new %s detected in %s block\n",
+               "{%llu} socket: %d, die: %d, %u new deferred hardware errors detected
+in %s block\n",
                        seq_no, skt, aid, ecc_count.new_de_count,
-                       (blk == RAS_BLOCK_ID__UMC) ?
-                               "deferred hardware errors" : "poison consumption",
                        blk_name(blk));
                RAS_DEV_INFO(ras_core->dev,
-               "{%llu} socket: %d, die: %d, %u %s detected in total in %s block\n",
+               "{%llu} socket: %d, die: %d, %u deferred hardware errors detected in
+total in %s block\n",
                        seq_no, skt, aid, ecc_count.total_de_count,
-                       (blk == RAS_BLOCK_ID__UMC) ?
-                               "deferred hardware errors" : "poison consumption",
                        blk_name(blk));
        }
 
        if (ecc_count.new_ce_count) {
+               if (ras_core_is_ce_log_disabled(ras_core))
+                       return;
                RAS_DEV_INFO(ras_core->dev,
                "{%llu} socket: %d, die: %d, %u new correctable hardware errors detected in %s block\n",
                        seq_no, skt, aid, ecc_count.new_ce_count, blk_name(blk)); @@ -114,6 +112,11 @@ static void aca_bank_log(struct ras_core_context *ras_core,  {
        int i;

+ if(ras_core_is_ce_log_disabled(ras_core) &&
+ bank->ecc_type == RAS_ERR_TYPE__CE &&
+ !bank_ecc->de_count)
+ return;
+

        RAS_DEV_INFO(ras_core->dev,
                "{%llu}" RAS_HW_ERR "Accelerator Check Architecture events logged\n",
                bank->seq_no);
drivers/gpu/drm/amd/ras/rascore/ras_aca_v1_0.c +1 -1
index 840610538c1f..288fad0862a2 100644
@@ -265,7 +265,7 @@ static int aca_parse_bank_default(struct ras_core_context *ras_core,
        ecc->bank_info.addr = bank->regs[ACA_REG_IDX__ADDR];
 
        if (aca_check_bank_is_de(ras_core, status)) {
-               ecc->de_count = 0;
+               ecc->de_count = 1;
        } else {
                if (bank->ecc_type == RAS_ERR_TYPE__UE)
                        ecc->ue_count = 1;
drivers/gpu/drm/amd/ras/rascore/ras_core.c +17 -0
index 08e17a83ad5b..f3c6cbc7e05f 100644
@@ -28,6 +28,10 @@
 
 #define IS_LEAP_YEAR(x) ((x % 4 == 0 && x % 100 != 0) || x % 400 == 0)
 
+enum RAS_DEBUG_MASK {
+       RAS_DEBUG_DISABLE_RAS_CE_LOG = BIT(9), };
+
 static const char * const ras_block_name[] = {
        "umc",
        "sdma",
@@ -374,6 +378,14 @@ int ras_core_sw_fini(struct ras_core_context *ras_core)
        return 0;
 }
 
+static void ras_init_debug_options(struct ras_core_context *ras_core) {
+       if (ras_core->ras_debug_mask & RAS_DEBUG_DISABLE_RAS_CE_LOG) {
+               RAS_DEV_INFO(ras_core->dev, "debug: disable kernel logs of correctable errors\n");
+               ras_core->ras_ce_log_disabled = true;
+       }
+}
+
 int ras_core_hw_init(struct ras_core_context *ras_core)  {
        int ret;
@@ -382,6 +394,7 @@ int ras_core_hw_init(struct ras_core_context *ras_core)
                        ras_core->config->ras_eeprom_supported;
 
        ras_core->poison_supported = ras_core->config->poison_supported;
+       ras_core->ras_debug_mask = ras_core->config->ras_debug_mask;
 
        ret = ras_psp_hw_init(ras_core);
        if (ret)
@@ -434,6 +447,8 @@ int ras_core_hw_init(struct ras_core_context *ras_core)
        if (ret)
                goto init_err7;
 
+       ras_init_debug_options(ras_core);
+
        ras_core->is_initialized = true;
 
        return 0;
@@ -545,6 +560,11 @@ bool ras_core_is_enabled(struct ras_core_context *ras_core)
        return ras_core->ras_core_enabled;
 }
 
+bool ras_core_is_ce_log_disabled(struct ras_core_context *ras_core) {
+       return ras_core->ras_ce_log_disabled;
+}
+
 

[Thomas] Since ras_debug_mask is a read-only parameter, set RAS_DEBUG_DISABLE_RAS_CE_LOG = 9 and directly return ras_core->config->ras_debug_mask & BIT(RAS_DEBUG_DISABLE_RAS_CE_LOG),This way, perhaps there is no need to introduce the additional ras_core->ras_ce_log_disabled and ras_core->ras_debug_mask fields?

 uint64_t ras_core_get_utc_second_timestamp(struct ras_core_context *ras_core)  {
        if (!ras_core)

--
2.34.1