[PATCH 108/109] drm/amd/ras: assign BERT platform record ownership

From: Xiang Liu <[email protected]>

IA processor and Platform Memory BERT sections are system-wide records
and must not be copied into every GPU RAS ring. Per-socket GPU records
still need to be offered to each matching GPU.

Assign platform sections to the first eligible RAS instance and claim
each GPU socket before parsing its cached BERT data. Serialize ownership
with the shared mutex and retain the state until the final RAS device
exits.

Signed-off-by: Xiang Liu <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
---
 .../gpu/drm/amd/ras/ras_mgr/amdgpu_ras_bert.c | 20 +++++++++++++++----
 1 file changed, 16 insertions(+), 4 deletions(-)

diff --git a/drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_bert.c b/drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_bert.c
index 04490f5173c41..71e9c8d3c8476 100644
--- a/drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_bert.c
+++ b/drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_bert.c
@@ -34,7 +34,9 @@
 struct amdgpu_ras_bert_boot_err_state {
 	struct mutex lock; /* protects cached boot error state */
 	bool polled;
+	bool platform_written;
 	bool written[MAX_GPU_INSTANCE];
+	u32 users;
 	int poll_result;
 	u8 *raw_data;
 	u32 raw_data_len;
@@ -60,6 +62,7 @@ static int amdgpu_ras_bert_process_boot_errors(struct amdgpu_device *adev)
 		return -EPERM;
 
 	mutex_lock(&state->lock);
+	state->users++;
 
 	if (!state->polled) {
 		state->polled = true;
@@ -135,10 +138,13 @@ static int amdgpu_ras_bert_process_boot_errors(struct amdgpu_device *adev)
 		goto out_unlock;
 	}
 
+	ras_mgr->ras_core->bert_platform_owner = !state->platform_written;
+	if (ras_mgr->ras_core->bert_platform_owner)
+		state->platform_written = true;
+	state->written[socket_id] = true;
 	ret = ras_bert_process_records(ras_mgr->ras_core,
 				       state->raw_data, state->raw_data_len);
-	if (!ret)
-		state->written[socket_id] = true;
+	ras_mgr->ras_core->bert_platform_owner = false;
 
 out_unlock:
 	mutex_unlock(&state->lock);
@@ -146,17 +152,23 @@ static int amdgpu_ras_bert_process_boot_errors(struct amdgpu_device *adev)
 	return ret;
 }
 
-static void amdgpu_ras_bert_reset_boot_errors(void)
+static void amdgpu_ras_bert_release_boot_errors(void)
 {
 	struct amdgpu_ras_bert_boot_err_state *state = &boot_err_state;
 
 	mutex_lock(&state->lock);
+	if (!state->users || --state->users)
+		goto out_unlock;
+
 	memset(state->written, 0, sizeof(state->written));
+	state->platform_written = false;
 	kfree(state->raw_data);
 	state->raw_data = NULL;
 	state->raw_data_len = 0;
 	state->poll_result = 0;
 	state->polled = false;
+
+out_unlock:
 	mutex_unlock(&state->lock);
 }
 #endif
@@ -173,7 +185,7 @@ int amdgpu_ras_bert_sw_init(struct amdgpu_device *adev)
 int amdgpu_ras_bert_sw_fini(struct amdgpu_device *adev)
 {
 #if defined(CONFIG_X86_MCE_AMD) && defined(CONFIG_ACPI_APEI)
-	amdgpu_ras_bert_reset_boot_errors();
+	amdgpu_ras_bert_release_boot_errors();
 #endif
 	return 0;
 }
-- 
2.55.0