Guide to Set Up Debugging Environment for Total Compute Software Stack

Arm Total Compute Solutions (TCS) is the platform for mobile computing, with double double-digit gains in CPU performance and efficiency to power the next generation of AI and immersive mobile gaming experiences.

Total Compute (TC) software stack is a reference software stack for the Arm Total Compute Fixed Virtual Platform (FVP). The TC software consists of firmware, kernel and file system components that can run on the associated FVP.

Some of its main components are as follows:

RSS firmware, which provides Hardware Root of Trust
SCP firmware, which is responsible for system initialization, clock and power control;
AP firmware, which is Trusted Firmware-A (TF-A)
Secure Partition Manager - Hafnium
Secure Partitions:
- OP-TEE Trusted OS in Buildroot
- Trusted Services in Buildroot
- Trusty Trusted OS in Android
U-Boot, which loads and verifies kernel and rootfs
Linux Kernel

This blog shows you how to set up the debugging environment step by step, including how to configure the debug environments for RSS, SCP, and AP.

Prerequisites

This blog assumes that you already set up the following environment:

Host PC (Ubuntu Linux 20.04)
TC2 Fast Model platform (FVP) (version 11.23.17)
Arm Development Studio (Latest version)
Download and build TC2 software stack, by following the Total Compute User Guide.

Overview

This blog describes the following steps to set up debugging environment for Total Compute Software Stack:

Step 1: Import TC FVP Model into Arm DS.

Step 2: Start FVP and launch the Iris debug server.

Step 3: Establish debug connections in Arm DS, including:

Application Processor debug connection.
System Control Processor debug connection.
Runtime Security Subsystem debug connection.

Step 4: Load corresponding symbol tables file and start debugging.

Step 1: Import TC FVP Model into Arm DS

Arm DS supports some built-in FVP models, but TC2 FVP is not included in the built-in models of Arm DS. Before you start debugging TC2, you must use the Model Configuration wizard in Arm DS to create debug configurations for new models:

Open the Model Configuration wizard
From the Main menu, select File -> New -> Other -> Configuration Database -> Model Configuration and click Next, as follows:
Create the configuration database where you want to include your TC2 model, as follows:
Choose Configuration Database and click Next, as follows:
Select Iris as the model interface for connecting to your model, as follows:
Select the Launch and connect to specific to your model, as follows:
In Model Selection from File System dialog box, click File to browse for a model and select it, as follows:
Click the Import button to import the TC2 FVP model, as follows:
Verify whether the import is successful, by checking for the imported model in the Project Explorer, as follows:

Note: If you encounter a smiliar error message, as shown in the following figure, during the import process, it is because the current version of Arm DS cannot recognize the latest CPU type. In this case, you must update Arm DS to the latest version to support this CPU type.

Step 2: Start FVP and launch the Iris debug server

By default, to run the software stack for TC, the following command is used to launch the FVP.

./run-scripts/tc2/run_model.sh -m <model binary path> -d <distro>

For debugging TC software with Arm DS, it is necessary to start the Iris debug server when starting the FVP. This allows Arm DS to connect to the debug server later. To initiate the Iris debug server during FVP startup, include an additional option -I. When connecting Arm DS to the Iris debug server, the server port is required. To print this port during startup, include an additional parameter -p. This way, the FVP can be launched by using the following command:

./run-scripts/tc2/run_model.sh -m <model binary path> -d <distro> -- -I -p

By default, FVP uses xterm to initialize the serial port output and input for TC platform. In certain situations, the xterm font might appear relatively small, so copying and pasting operations might not be as convenient. For example:

If you want to replace it with the default terminal of Ubuntu, you must add additional parameters as follows:

./run-scripts/tc2/run_model.sh -m <model binary path> -d <distro> -- -I -p \
-C css.terminal_uart1_ap.terminal_command="gnome-terminal -t %title -- telnet localhost %port" \
-C css.terminal_uart_ap.terminal_command="gnome-terminal -t %title -- telnet localhost %port"  \
-C soc.terminal_s0.terminal_command="gnome-terminal -t %title -- telnet localhost %port"  \
-C soc.terminal_s1.terminal_command="gnome-terminal -t %title -- telnet localhost %port "

The serial ports of the TC FVP are as follows:

After running this command, the output is as follows:

Iris server started listening to port 7100
INFO: GPU Model initialized.
INFO: TC2.css.cmn: CI-700 r1p0 model, periphbase=0x0000000050000000, mesh_config_file='TC2_ci700.yml'

terminal_uart_ap: Listening for serial connection on port 5006
terminal_uart1_ap: Listening for serial connection on port 5007
terminal_s0: Listening for serial connection on port 5008
terminal_s1: Listening for serial connection on port 5009
terminal_0: Listening for serial connection on port 5010
terminal_1: Listening for serial connection on port 5011

You can observe the port of the Iris debug server, which can be used later.

Step 3: Establish debug connections in Arm DS

The TC platform has:

The Application process
The System control processor
The Runtime Security Subsystem Processor

These three systems run different software stacks. To debug them separately or simultaneously, you must create debug connections for each of these three systems in Arm DS.

To create debug connections for these three systems, perform the following steps:

From the main menu, select Run-> Debug Configuration -> Generic Arm C/C++ Application and Click New Launch Configuration, as follows:
Select the previously imported model and choose the CPU core you want to debug. This example uses Cortex-A520X4 SMP Cluster 0 as an application processor.
Within the Connections section below, choose Connect to an already running model. The address is 127.0.0.1, and the port is the one printed when launching the FVP.
Click the Apply button. Once applied successfully, you can see the following in the Debug Control on the left:

For the System Control Processor and Runtime Security Subsystem, you can repeat the above process. The only difference is that when selecting the CPU, choose Cortex-M3 for the System Control Processor and Cortex-M55 for the Runtime Security Subsystem.

After completing these steps, you have three debug connections as follows: