Tips & Tricks for Certified Ubuntu AMD-Xilinx Devices

This page collects various tips & tricks for getting the most out of the Certified Ubuntu on Xilinx Devices release.

Unless otherwise noted, the information on this page applies to both 20.04 and 22.04 images.

Table of Contents

Installing a Graphical Package Manager

The core of the Ubuntu ecosystem is 3rd-party packages. Ubuntu 20.04 LTS provides some packages as tradition Debian .deb packages in apt repositories. Sandboxed modern applications are provided as Snaps (such as xlnx-config). By default, the Certified Ubuntu on Xilinx Devices image does not provide a graphical package manager. If you would like to add one, there are several available in the existing Ubuntu package repository ecosystem.

A common lightweight package manager for apt packages is called synaptic. It can be installed with the following command:

$ sudo apt install synaptic

The Ubuntu snap ecosystem provides a separate GUI for managing and installing snaps. The Ubuntu Snap Store application is called snap-store. It can be installed with the following command:

$ sudo snap install snap-store

Stopping the Ubuntu Graphical Desktop Environment

If you wish to stop the Ubuntu graphical desktop environment and return to a strictly command line environment, you can do so by using the systemd interfaces. The systemd “multi-user” target is equivalent to runlevels 3, 4, or 5 in traditional systems based on SysVInit. This can be used to save system memory which would otherwise be used for running the rich desktop experience.

To switch to the “multi-user” target temporarily:

$ sudo systemctl isolate

To return to the graphical desktop environment:

To make this change permanent (eg, the graphical desktop environment will no longer start automatically):

Switching Desktop Environments

The default graphical environment in Ubuntu 20.04 LTS and 22.04 LTS is based on GNOME3. Other desktop environments are available and can be installed to suit personal preference.

Replacement desktop environments can be installed from the command line. You can connect to the system via direct USB UART connection, SSH, or by switching the monitor output from graphical multi-user mode to terminal-only multi-user mode. For details on switching the monitor output to terminal multi-user mode, see the Stopping the Ubuntu Graphical Desktop Environment section.

From the terminal first, make sure that the apt package cache is up-to-date:

Next, install the new desktop environment using the apt-get tools:

For a list of desktop environment packages, see the table below.

Desktop Environment Name

Package Name


Desktop Environment Name

Package Name


Ubuntu Desktop


This is the default Ubuntu desktop environment. This can be installed on Ubuntu Server images to provide the standard Ubuntu Desktop GUI experience on server images.

Mate (from ubuntu-mate remix distribution)


Tested and works. May take a long time to install depending on the speed of your SD card.

Unity (traditional Ubuntu desktop environment)



KDE (from kubuntu remix distribution)



Xfce (from xubuntu remix distribution)


Tested and works.

LXDE (from lubuntu remix distribution)


Tested and works.

Cinnamon (from ubuntu-cinnamon remix distribution)



NOTE: The default Ubuntu display manager is gdm. Some desktop environments require the lightdm display manager. You may need to switch the default display manager when switching desktop environments

NOTE: To configure Ubuntu to start the in different modes, issue the following commands:

For terminal-based multi-user mode (default)
$ sudo systemctl set-default multi-user

For graphical multi-user mode (GUI desktop)

$ sudo systemctl set-default graphical

Screen Sharing and VNC

You can share the desktop by enabling Screen Sharing in the GNOME desktop settings. You first need to install the Vino package. More details can be found at the following page:

If you’d rather run a more traditional VNC server with support for multiple virtual desktops instead of sharing the primary desktop, you can refer to the following guide:

Understanding CMA Usage

Many of the Xilinx designs, particularly those that leverage the Xilinx Video Codec Unit (VCU), make extensive use of Contiguous Memory Allocator (CMA) in the Linux kernel.

There are many resources that explain how CMA works:

The Xilinx VCU User Guide (PG252) has details on how the VCU device drivers leverage CMA and contains references on CMA sizes for various applications.

The default Certified Ubuntu on Xilinx Devices image specifies a CMA size that is best suited for the official out-of-the-box demonstrations. Depending on your workloads and usage, you may find that you need to adjust the default CMA size to meet your needs.

The following Bash line can be used to monitor the current CMA usage:

If you would like to alter CMA value to free up memory for other applications, please see the Getting Started with Certified Ubuntu on Xilinx Devices page in the Changing the Kernel bootargs Used By U-Boot section.

Updating the Board-Level Metadata EEPROM

Xilinx evaluation boards have metadata about the board stored in an I2C EEPROM soldered to the board. Among this data is the serial number of the board, name of the board (eg, ZCU102), board revision, and Ethernet MAC address.
Xilinx declares how & where this information is stored in the device tree DTS file for the ZCU102 here:

For some boards (especially boards early in the production run), this data may be incomplete. To check this data on your board, stop the boot process during the U-Boot startup phase by pressing CTRL-C when you see the following messages on the USB UART terminal:

This will stop the boot process at the U-Boot terminal. U-Boot provides an I2C interface which can read and write data from the meta-data EEPROM. Xilinx has populated the ZCU10x evaluation boards with multiple I2C devices, each attached to the output of an I2C mux. To get the topology of the I2C chain use the following command:

Cross-referencing with the DTS file, we see that the device of interest is on Bus 5, at address 54. First, switch to Bus 5 with the command:

Next, check the contents of the EEPROM at the addresses shown in DTS file:

Most of the data is informational only, but two pieces are critical:

  1. Board Name (address 0xd0)

  2. MAC address (address 0x20)

If the board name field is not correct, the Certified Ubuntu on Xilinx Devices image will not boot properly. The Certified Ubuntu release uses this data during its initial boot phase (the ImgSel tool) documented by Xilinx here: . The ImgSel tool reads the board name and then uses this information to boot the correct BOOT.BIN file.

The MAC address field is probed during boot. If it is found populated with a valid set of values, it will use this data for the remainder of the boot process. If it doesn't, it will randomly create a MAC address which will change from boot to boot.

To fix the data, use the U-Boot I2C commands to write data back to EEPROM. For example, to fix the MAC address, use a sequence of commands to write the MAC address octet values:

The above example is intentionally incomplete. The valid addresses for a MAC address are from 0x20 through 0x25. Repeat the i2c mw commands to write all of the octets. To read them back, simply use the i2c md command sequence shown above. Once you have updated the MAC address and verified it, you can reboot the board with:

This time just let the boot process proceed like normal. Once Ubuntu starts, you can verify the MAC address at the terminal with the following command:

You should see the MAC address value set in U-Boot. In the data shown above, the actual MAC address is [REDACTED].

The same basic process can be used to update the Board Name, Board Serial Number, or Board Version fields. Keep in mind that when writing the values for the Board Name, Board Version Number, and Board Serial Number to use the ASCII values in hexadecimal. The MAC address is written in native hexadecimal values.

Ubuntu Internal Error Messages

From time to time, the Ubuntu desktop may display error messages similar to Sorry, Ubuntu 20.04 has experienced an internal error.

These messages are generated by the Ubuntu error checking tool called Apport. You can learn more about Apport here. If this is the first time you’ve seen the message, you can learn more details by clicking the Show Details button to learn more.

Many types of scenarios can generate Apport errors and not all of them indicate an actual problem in the system. If you wish, you can click the boxes Remember this in future and Ignore future problems of this type to prevent these general types of warnings from being generated. Please be sure to check the details of the error messages before suppressing them.

Apport runs as a background service. If you are confident that the errors you see are not representative of larger errors in your system, you can disable the Apport service by issuing the command:

Rebuilding the Image Selector Utility

The core of the Certified Ubuntu for Xilinx Devices boot process on ZCU10x boards is the Image Selector tool. This utility performs the initial boot of the board, reads board-identifying information from the attached EEPROM, and then reboots into a customized set of boot collateral for the specific board it is running on. For more detail on this process, see the page.This is what allows the same Certified Ubuntu for Xilinx Devices SD card image to boot on multiple boards. Typically, there is no need to modify this code. If a user determines that they would like to modify the way that Image Selector operates, it can be rebuilt inside the Vitis environment. The source code of Image Selector is part of the Xilinx embeddedsw repository on the Xilinx GitHub.

The Tcl code below automates the building of the Image Selector tool inside of Vitis. Be sure to pass the symbol XIS_UART_ENABLE to see the STDOUT info (as seen below).


When building Image Selector manually, be sure to include at least the minimal hardware definition found below:

Minimum Processing System (PS) configuration:

Minimum Board Support configuration:

Using a Custom Device Tree

To temporarily override the DTB without modifying the FIT image (Kria) or bootxxxx.bin (ZCU10x), you can create a new dtb named user-override.dtb.  Copy user-override.dtb to the FAT partition and reboot. If user-override.dtb is found on the FAT partition by U-boot, it will use this DTB instead of the board-specific DTB found in the FIT image when Linux is started.

Understanding the Configuring flash-kernel Prompts When Updating the Linux Kernel

When updating the Linux kernel via packages available in the Ubuntu package feed, the following prompt may appear:

In general, it is safe to choose either the local version or the maintainer’s version. This message is an artifact how the Linux kernel packages are configured for the Certified Ubuntu image. Typically, the Ubuntu version of the Linux kernel includes both the splash and quietcommand line arguments. The splash argument is typically used in desktop Linux distributions and instructs the kernel to display a splash screen (eg, a screen displaying the distribution's logo) if one is available. The quiet argument puts the kernel into non-verbose mode. Omitting this option will produce much more output on the terminal during Linux kernel boot. In most Linux distributions (including Ubuntu), the quiet argument is used to produce a more aesthetically pleasing boot experience. The AMD-Xilinx Certified Ubuntu image omits the quiet argument in order to produce more debug messages that are useful when prototyping. This change in the default boot arguments is applied using a meta-package overlay. This overlay causes a mis-match in the kernel package configuration when updating it via normal package management methods.

When updating the kernel, the options presented correspond to:

  1. install the package maintainer's version -- This option will add the default "quiet splash" options back to the Linux kernel command line

  2. keep the local version currently installed -- This option will keep the AMD-Xilinx configuration which omits the quiet option.


© Copyright 2019 - 2022 Xilinx Inc. Privacy Policy