drm/amdgpu: add mutex to protect ras shared memory

Add mutex to protect ras shared memory.

v2:
  Add TA_RAS_COMMAND__TRIGGER_ERROR command call
  status check.

Signed-off-by: YiPeng Chai <YiPeng.Chai@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

3 files changed, 86 insertions, 40 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
index 800cc7a148b2..7cdff355cedb 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
@@ -1591,6 +1591,68 @@ static void psp_ras_ta_check_status(struct psp_context *psp)
 	}
 }
 
+static int psp_ras_send_cmd(struct psp_context *psp,
+		enum ras_command cmd_id, void *in, void *out)
+{
+	struct ta_ras_shared_memory *ras_cmd;
+	uint32_t cmd = cmd_id;
+	int ret = 0;
+
+	if (!in)
+		return -EINVAL;
+
+	mutex_lock(&psp->ras_context.mutex);
+	ras_cmd = (struct ta_ras_shared_memory *)psp->ras_context.context.mem_context.shared_buf;
+	memset(ras_cmd, 0, sizeof(struct ta_ras_shared_memory));
+
+	switch (cmd) {
+	case TA_RAS_COMMAND__ENABLE_FEATURES:
+	case TA_RAS_COMMAND__DISABLE_FEATURES:
+		memcpy(&ras_cmd->ras_in_message,
+			in, sizeof(ras_cmd->ras_in_message));
+		break;
+	case TA_RAS_COMMAND__TRIGGER_ERROR:
+		memcpy(&ras_cmd->ras_in_message.trigger_error,
+			in, sizeof(ras_cmd->ras_in_message.trigger_error));
+		break;
+	case TA_RAS_COMMAND__QUERY_ADDRESS:
+		memcpy(&ras_cmd->ras_in_message.address,
+			in, sizeof(ras_cmd->ras_in_message.address));
+		break;
+	default:
+		dev_err(psp->adev->dev, "Invalid ras cmd id: %u\n", cmd);
+		ret = -EINVAL;
+		goto err_out;
+	}
+
+	ras_cmd->cmd_id = cmd;
+	ret = psp_ras_invoke(psp, ras_cmd->cmd_id);
+
+	switch (cmd) {
+	case TA_RAS_COMMAND__TRIGGER_ERROR:
+		if (ret || psp->cmd_buf_mem->resp.status)
+			ret = -EINVAL;
+		else if (out)
+			memcpy(out, &ras_cmd->ras_status, sizeof(ras_cmd->ras_status));
+		break;
+	case TA_RAS_COMMAND__QUERY_ADDRESS:
+		if (ret || ras_cmd->ras_status || psp->cmd_buf_mem->resp.status)
+			ret = -EINVAL;
+		else if (out)
+			memcpy(out,
+				&ras_cmd->ras_out_message.address,
+				sizeof(ras_cmd->ras_out_message.address));
+		break;
+	default:
+		break;
+	}
+
+err_out:
+	mutex_unlock(&psp->ras_context.mutex);
+
+	return ret;
+}
+
 int psp_ras_invoke(struct psp_context *psp, uint32_t ta_cmd_id)
 {
 	struct ta_ras_shared_memory *ras_cmd;
@@ -1632,23 +1694,15 @@ int psp_ras_invoke(struct psp_context *psp, uint32_t ta_cmd_id)
 int psp_ras_enable_features(struct psp_context *psp,
 		union ta_ras_cmd_input *info, bool enable)
 {
-	struct ta_ras_shared_memory *ras_cmd;
+	enum ras_command cmd_id;
 	int ret;
 
-	if (!psp->ras_context.context.initialized)
+	if (!psp->ras_context.context.initialized || !info)
 		return -EINVAL;
 
-	ras_cmd = (struct ta_ras_shared_memory *)psp->ras_context.context.mem_context.shared_buf;
-	memset(ras_cmd, 0, sizeof(struct ta_ras_shared_memory));
-
-	if (enable)
-		ras_cmd->cmd_id = TA_RAS_COMMAND__ENABLE_FEATURES;
-	else
-		ras_cmd->cmd_id = TA_RAS_COMMAND__DISABLE_FEATURES;
-
-	ras_cmd->ras_in_message = *info;
-
-	ret = psp_ras_invoke(psp, ras_cmd->cmd_id);
+	cmd_id = enable ?
+		TA_RAS_COMMAND__ENABLE_FEATURES : TA_RAS_COMMAND__DISABLE_FEATURES;
+	ret = psp_ras_send_cmd(psp, cmd_id, info, NULL);
 	if (ret)
 		return -EINVAL;
 
@@ -1672,6 +1726,8 @@ int psp_ras_terminate(struct psp_context *psp)
 
 	psp->ras_context.context.initialized = false;
 
+	mutex_destroy(&psp->ras_context.mutex);
+
 	return ret;
 }
 
@@ -1756,9 +1812,10 @@ int psp_ras_initialize(struct psp_context *psp)
 
 	ret = psp_ta_load(psp, &psp->ras_context.context);
 
-	if (!ret && !ras_cmd->ras_status)
+	if (!ret && !ras_cmd->ras_status) {
 		psp->ras_context.context.initialized = true;
-	else {
+		mutex_init(&psp->ras_context.mutex);
+	} else {
 		if (ras_cmd->ras_status)
 			dev_warn(adev->dev, "RAS Init Status: 0x%X\n", ras_cmd->ras_status);
 
@@ -1772,12 +1829,12 @@ int psp_ras_initialize(struct psp_context *psp)
 int psp_ras_trigger_error(struct psp_context *psp,
 			  struct ta_ras_trigger_error_input *info, uint32_t instance_mask)
 {
-	struct ta_ras_shared_memory *ras_cmd;
 	struct amdgpu_device *adev = psp->adev;
 	int ret;
 	uint32_t dev_mask;
+	uint32_t ras_status = 0;
 
-	if (!psp->ras_context.context.initialized)
+	if (!psp->ras_context.context.initialized || !info)
 		return -EINVAL;
 
 	switch (info->block_id) {
@@ -1801,13 +1858,8 @@ int psp_ras_trigger_error(struct psp_context *psp,
 	dev_mask &= AMDGPU_RAS_INST_MASK;
 	info->sub_block_index |= dev_mask;
 
-	ras_cmd = (struct ta_ras_shared_memory *)psp->ras_context.context.mem_context.shared_buf;
-	memset(ras_cmd, 0, sizeof(struct ta_ras_shared_memory));
-
-	ras_cmd->cmd_id = TA_RAS_COMMAND__TRIGGER_ERROR;
-	ras_cmd->ras_in_message.trigger_error = *info;
-
-	ret = psp_ras_invoke(psp, ras_cmd->cmd_id);
+	ret = psp_ras_send_cmd(psp,
+			TA_RAS_COMMAND__TRIGGER_ERROR, info, &ras_status);
 	if (ret)
 		return -EINVAL;
 
@@ -1817,9 +1869,9 @@ int psp_ras_trigger_error(struct psp_context *psp,
 	if (amdgpu_ras_intr_triggered())
 		return 0;
 
-	if (ras_cmd->ras_status == TA_RAS_STATUS__TEE_ERROR_ACCESS_DENIED)
+	if (ras_status == TA_RAS_STATUS__TEE_ERROR_ACCESS_DENIED)
 		return -EACCES;
-	else if (ras_cmd->ras_status)
+	else if (ras_status)
 		return -EINVAL;
 
 	return 0;
@@ -1829,25 +1881,16 @@ int psp_ras_query_address(struct psp_context *psp,
 			  struct ta_ras_query_address_input *addr_in,
 			  struct ta_ras_query_address_output *addr_out)
 {
-	struct ta_ras_shared_memory *ras_cmd;
 	int ret;
 
-	if (!psp->ras_context.context.initialized)
-		return -EINVAL;
-
-	ras_cmd = (struct ta_ras_shared_memory *)psp->ras_context.context.mem_context.shared_buf;
-	memset(ras_cmd, 0, sizeof(struct ta_ras_shared_memory));
-
-	ras_cmd->cmd_id = TA_RAS_COMMAND__QUERY_ADDRESS;
-	ras_cmd->ras_in_message.address = *addr_in;
-
-	ret = psp_ras_invoke(psp, ras_cmd->cmd_id);
-	if (ret || ras_cmd->ras_status || psp->cmd_buf_mem->resp.status)
+	if (!psp->ras_context.context.initialized ||
+		!addr_in || !addr_out)
 		return -EINVAL;
 
-	*addr_out = ras_cmd->ras_out_message.address;
+	ret = psp_ras_send_cmd(psp,
+			TA_RAS_COMMAND__QUERY_ADDRESS, addr_in, addr_out);
 
-	return 0;
+	return ret;
 }
 // ras end
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h
index 3635303e6548..74a96516c913 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h
@@ -200,6 +200,7 @@ struct psp_xgmi_context {
 struct psp_ras_context {
 	struct ta_context		context;
 	struct amdgpu_ras		*ras;
+	struct mutex			mutex;
 };
 
 #define MEM_TRAIN_SYSTEM_SIGNATURE		0x54534942
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp_ta.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp_ta.c
index 8e8afbd237bc..0c856005df6b 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp_ta.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp_ta.c
@@ -348,6 +348,7 @@ static ssize_t ta_if_invoke_debugfs_write(struct file *fp, const char *buf, size
 
 	context->session_id = ta_id;
 
+	mutex_lock(&psp->ras_context.mutex);
 	ret = prep_ta_mem_context(&context->mem_context, shared_buf, shared_buf_len);
 	if (ret)
 		goto err_free_shared_buf;
@@ -366,6 +367,7 @@ static ssize_t ta_if_invoke_debugfs_write(struct file *fp, const char *buf, size
 		ret = -EFAULT;
 
 err_free_shared_buf:
+	mutex_unlock(&psp->ras_context.mutex);
 	kfree(shared_buf);
 
 	return ret;