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PATCH Superseded Ce Sun 2 participants 2 messages Jul 15

Subsystems (2)

v2 of 4 Superseded Latest is v4 →

lore
Share & tools
lore

Ce Sun

author
#0 lore raw

PATCH v2 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, including both ACA and CE error counter kernel messages.

---
 drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c | 12 ++++++++++++
 drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h |  1 +
 drivers/gpu/drm/amd/ras/rascore/ras.h            |  3 +++
 drivers/gpu/drm/amd/ras/rascore/ras_aca.c        |  7 +++++++
 drivers/gpu/drm/amd/ras/rascore/ras_core.c       | 10 ++++++++++
 5 files changed, 33 insertions(+)
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c +12 -0
index ea95e0f93d1c..bdc9d9583bdf 100644
@@ -437,6 +437,7 @@ static int amdgpu_ras_mgr_hw_init(struct amdgpu_ip_block *ip_block)
 	ras_mgr->ras_is_ready = true;
 
 	amdgpu_enable_uniras(adev, true);
+	amdgpu_set_ce_log_state(adev);
 
 	RAS_DEV_INFO(adev, "AMDGPU RAS Is Ready.\n");
 	return 0;
@@ -510,6 +511,17 @@ const struct amdgpu_ip_block_version ras_v1_0_ip_block = {
 	.funcs = &ras_v1_0_ip_funcs,
 };
 
+void amdgpu_set_ce_log_state(struct amdgpu_device *adev)
+{
+	struct amdgpu_ras_mgr *ras_mgr = amdgpu_ras_mgr_get_context(adev);
+	bool disable_ce_log = !!adev->debug_disable_ce_logs;
+
+	if (!ras_mgr || !ras_mgr->ras_core)
+		return;
+
+	ras_core_disable_ce_log(ras_mgr->ras_core, disable_ce_log);
+}
+
 int amdgpu_enable_uniras(struct amdgpu_device *adev, bool enable)
 {
 	struct amdgpu_ras_mgr *ras_mgr = amdgpu_ras_mgr_get_context(adev);
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h +1 -0
index a20bb8fdce87..efbdb25241a3 100644
@@ -63,6 +63,7 @@ extern const struct amdgpu_ip_block_version ras_v1_0_ip_block;
 struct amdgpu_ras_mgr *amdgpu_ras_mgr_get_context(
 			struct amdgpu_device *adev);
 int amdgpu_enable_uniras(struct amdgpu_device *adev, bool enable);
+void amdgpu_set_ce_log_state(struct amdgpu_device *adev);
 bool amdgpu_uniras_enabled(struct amdgpu_device *adev);
 int amdgpu_ras_mgr_handle_fatal_interrupt(struct amdgpu_device *adev, void *data);
 int amdgpu_ras_mgr_handle_controller_interrupt(struct amdgpu_device *adev, void *data);
drivers/gpu/drm/amd/ras/rascore/ras.h +3 -0
index 878dfdfcb18a..0c362f9e25d6 100644
@@ -346,6 +346,7 @@ struct ras_core_context {
 	spinlock_t seqno_lock;
 
 	bool ras_core_enabled;
+	bool ras_core_ce_log_disabled;
 
 	u64 ras_fw_features;
 };
@@ -403,4 +404,6 @@ 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);
+void ras_core_disable_ce_log(struct ras_core_context *ras_core, bool disable);
+bool ras_core_is_ce_log_disabled(struct ras_core_context *ras_core);
 #endif
drivers/gpu/drm/amd/ras/rascore/ras_aca.c +7 -0
index 67a35409ff0e..7219a1008dea 100644
@@ -99,6 +99,8 @@ static void aca_report_ecc_info(struct ras_core_context *ras_core,
 	}
 
 	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 +116,11 @@ 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_core.c +10 -0
index 08e17a83ad5b..76cc98e0abc5 100644
@@ -545,6 +545,16 @@ bool ras_core_is_enabled(struct ras_core_context *ras_core)
 	return ras_core->ras_core_enabled;
 }
 
+void ras_core_disable_ce_log(struct ras_core_context *ras_core, bool disable)
+{
+	ras_core->ras_core_ce_log_disabled = disable;
+}
+
+bool ras_core_is_ce_log_disabled(struct ras_core_context *ras_core)
+{
+	return ras_core->ras_core_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

Quoted 103 lines

-----Original Message-----
From: Sun, Ce(Overlord) <[email protected]>
Sent: Wednesday, July 15, 2026 3:06 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 v2] drm/amdgpu/ras: Add debug mask to disable CE logs for uniras

Add debug mask to disable kernel logs of RAS correctable errors, including both ACA and CE error counter kernel messages.

---
 drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c | 12 ++++++++++++
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h |  1 +
 drivers/gpu/drm/amd/ras/rascore/ras.h            |  3 +++
 drivers/gpu/drm/amd/ras/rascore/ras_aca.c        |  7 +++++++
 drivers/gpu/drm/amd/ras/rascore/ras_core.c       | 10 ++++++++++
 5 files changed, 33 insertions(+)
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c +1 -0
index ea95e0f93d1c..bdc9d9583bdf 100644
@@ -437,6 +437,7 @@ static int amdgpu_ras_mgr_hw_init(struct amdgpu_ip_block
 *ip_block)
      ras_mgr->ras_is_ready = true;
 
      amdgpu_enable_uniras(adev, true);
+     amdgpu_set_ce_log_state(adev);
 
      RAS_DEV_INFO(adev, "AMDGPU RAS Is Ready.\n");
      return 0;
drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.c +9 -1
@@ -510,6 +511,17 @@ const struct amdgpu_ip_block_version ras_v1_0_ip_block
 = {
      .funcs = &ras_v1_0_ip_funcs,
 };
 
+void amdgpu_set_ce_log_state(struct amdgpu_device *adev) {
+     struct amdgpu_ras_mgr *ras_mgr = amdgpu_ras_mgr_get_context(adev);
+     bool disable_ce_log = !!adev->debug_disable_ce_logs;
+
+     if (!ras_mgr || !ras_mgr->ras_core)
+             return;
+
+     ras_core_disable_ce_log(ras_mgr->ras_core, disable_ce_log); }
+
 int amdgpu_enable_uniras(struct amdgpu_device *adev, bool enable)  {
      struct amdgpu_ras_mgr *ras_mgr = amdgpu_ras_mgr_get_context(adev); diff
--git a/drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h

b/drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h index a20bb8fdce87..efbdb25241a3 100644

drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_mgr.h +0 -0
@@ -63,6 +63,7 @@ extern const struct amdgpu_ip_block_version

ras_v1_0_ip_block; struct amdgpu_ras_mgr *amdgpu_ras_mgr_get_context(

                      struct amdgpu_device *adev);

int amdgpu_enable_uniras(struct amdgpu_device *adev, bool enable);
+void amdgpu_set_ce_log_state(struct amdgpu_device *adev);
bool amdgpu_uniras_enabled(struct amdgpu_device *adev); int
amdgpu_ras_mgr_handle_fatal_interrupt(struct amdgpu_device *adev, void *data); int amdgpu_ras_mgr_handle_controller_interrupt(struct amdgpu_device *adev, void *data); diff --git a/drivers/gpu/drm/amd/ras/rascore/ras.h b/drivers/gpu/drm/amd/ras/rascore/ras.h
index 878dfdfcb18a..0c362f9e25d6 100644

drivers/gpu/drm/amd/ras/rascore/ras.h +1 -0
@@ -346,6 +346,7 @@ struct ras_core_context {
      spinlock_t seqno_lock;
 
      bool ras_core_enabled;
+     bool ras_core_ce_log_disabled;
 
      u64 ras_fw_features;
 };
@@ -403,4 +404,6 @@ 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);

drivers/gpu/drm/amd/ras/rascore/ras.h +3 -0
+void ras_core_disable_ce_log(struct ras_core_context *ras_core, bool
+disable); bool ras_core_is_ce_log_disabled(struct ras_core_context
+*ras_core);
 #endif
drivers/gpu/drm/amd/ras/rascore/ras_aca.c +2 -0
index 67a35409ff0e..7219a1008dea 100644
@@ -99,6 +99,8 @@ static void aca_report_ecc_info(struct ras_core_context
 *ras_core,
      }
 
      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));

drivers/gpu/drm/amd/ras/rascore/ras_aca.c static void aca_bank_log(struct ras_core_context · line 121

Quoted 6 lines
drivers/gpu/drm/amd/ras/rascore/ras_aca.c +3 -0
@@ -114,6 +116,11 @@ 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)

[Tao] the de_count is always 0 for non-umc blocks.

Quoted 28 lines
drivers/gpu/drm/amd/ras/rascore/ras_aca.c +2 -0
+             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_core.c +8 -1
index 08e17a83ad5b..76cc98e0abc5 100644
@@ -545,6 +545,16 @@ bool ras_core_is_enabled(struct ras_core_context
 *ras_core)
      return ras_core->ras_core_enabled;
 }
 
+void ras_core_disable_ce_log(struct ras_core_context *ras_core, bool
+disable) {
+     ras_core->ras_core_ce_log_disabled = disable; }
+
+bool ras_core_is_ce_log_disabled(struct ras_core_context *ras_core) {
+     return ras_core->ras_core_ce_log_disabled;
+}
+
 uint64_t ras_core_get_utc_second_timestamp(struct ras_core_context *ras_core)  {
      if (!ras_core)
--

2.34.1