3 Installation on AArch64 #

SUSE Linux Enterprise Server is normally installed as an independent operating system. With virtualization it is also possible to run multiple instances of SUSE Linux Enterprise Server on the same hardware. However, the installation of the VM Host Server is performed like a typical installation with some additional packages. The installation of virtual guests is described in Chapter 8, Guest Installation.

Most installations are to a local hard disk. Therefore, it is necessary for the hard disk controllers to be available to the installation system. If a special controller (like a RAID controller) needs an extra kernel module, provide a kernel module update disk to the installation system.

Other installation targets may be various types of block devices that provide sufficient disk space and speed to run an operating system. This includes network block devices like iSCSI or SAN. It is also possible to install on network file systems that offer the standard Unix permissions. However, it may be problematic to boot these, because they must be supported by the initramfs before the actual system can start. Such installations can be useful when you need to start the same system in different locations or you plan to use virtualization features like domain migration.

Boot problems may prevent the YaST installer from starting on your system. Another symptom is failure to boot after the installation has been completed.

If an unexpected problem occurs during installation, information is needed to determine the cause of the problem. Use the following directions to help with troubleshooting:

To simplify the installation process and avoid accidental installations, the default setting on the installation DVD for SUSE Linux Enterprise Server is that your system is booted from the first hard disk. At this point, an installed boot loader normally takes over control of the system. This means that the boot DVD can stay in the drive during an installation. To start the installation, choose one of the installation possibilities in the boot menu of the media.

The primary processor on the Raspberry Pi's System-on-Chip (SoC) is the Broadcom VideoCore Graphics Processing Unit (GPU), not the Arm Central Processing Unit (CPU). It is the GPU which starts initializing the hardware from a first-stage boot loader in the on-chip Boot Read-Only Memory (Boot ROM). Only a few configuration options can influence the Boot ROM; see Section 3.6.1.2, “OTP Memory”.

The Raspberry Pi 3 hardware does not have any built-in firmware. Instead, its second-stage boot loader firmware bootcode.bin is loaded from the boot medium every time the machine is powered on. It in turn loads the third-stage boot loader start.elf.

The Raspberry Pi 4 hardware has a small Electrically Erasable Programmable Read-Only Memory (EEPROM) for the second-stage boot loader. Apart from that, its boot sequence is similar to that of the Raspberry Pi 3, loading the third-stage boot loader start4.elf from the boot medium.

Warning: EEPROM Update on Raspberry Pi 4

An update of the second-stage boot loader can be performed by booting from a specially prepared microSD card.

Only insert boot media that you trust, and verify that no file called recovery.bin is unintentionally present.

If an armstub8.bin file is present, it will be loaded as a fourth-stage boot loader at AArch64 Exception Level 3 (EL3). Otherwise, a minimal integrated stub will be used.

Note: EL3 Security Considerations

Code loaded for EL3 (often called BL31) will reside in memory, and Linux may attempt hypercalls into EL3 throughout its runtime.

Verify that your boot media have no armstub8.bin file unintentionally present. SUSE Linux Enterprise Server 15 SP2 does not include it.

Beware that the Raspberry Pi's SoC does not provide TrustZone secure memory. Both the OS on the CPU and any software on the GPU may access its RAM. It is therefore unsuited for cryptographic EL0-s applications. SUSE Linux Enterprise Server does not provide an EL1-s Trusted Execution Environment (TEE) for that reason.

SUSE Linux Enterprise Server for the Raspberry Pi is configured to load a fifth-stage boot loader called Das U-Boot.

3.6.1.2 OTP Memory #

  

The SoC also has a very small amount of One-Time Programmable Memory (OTP memory). This can be used to configure some settings, such as whether the Boot ROM should attempt to boot from USB devices or over Ethernet.

This OTP memory is described on the Raspberry Pi Foundation Web site: https://www.raspberrypi.org/documentation/hardware/raspberrypi/otpbits.md

Warning: One-Time Programmable Only

Configuration settings written into OTP memory cannot be reversed.

The most common use case for OTP memory will be enabling USB boot on Raspberry Pi 3 Model B or Compute Module 3.

3.6.1.2.1 Enabling USB Boot Mode for Raspberry Pi 3 Model B #

  

To permanently allow booting from connected USB mass storage devices on Raspberry Pi 3 Model B, and from its on-board USB Ethernet, prepare a microSD card as described in Section 3.6.3, “Deploying an Appliance Image”. Before unmounting or ejecting the card and booting from it, add to its FAT partition a text file extraconfig.txt (Section 3.6.1.1, “Config.txt”) with the following setting:

program_usb_boot_mode=1

Then continue to boot from the modified microSD card as usual. Once you see output from the U-Boot or GRUB boot loaders or the Linux kernel, you can remove power and then the microSD card. Your device should now be able to boot from USB (Section 3.6.4, “Installation from USB Media”).

Note that once USB boot mode has been enabled for Raspberry Pi 3 Model B, USB boot mode cannot be disabled again (Section 3.6.1.2, “OTP Memory”).

For more details, refer to the Raspberry Pi Foundation Web site: https://www.raspberrypi.org/documentation/hardware/raspberrypi/bootmodes/msd.md

For the Raspberry Pi Compute Module 3, the setting required is the same, but the deployment of the modified image is a little more complicated.

There is no battery-backed Real-Time Clock (RTC) on the Raspberry Pi itself.

Note: Time Synchronization

The lack of a Real-Time Clock means that Raspberry Pi devices need to be configured to fetch the time from a network server by Network Time Protocol.

However, base boards for the Raspberry Pi Compute Modules may feature an RTC.

It is also possible to connect an RTC via the GPIO connector, using Hardware Attached on Top (HATs) or other expansion boards.

Either way, check whether the respective RTC chipset is supported by SUSE Linux Enterprise Server. The connected RTC will need to be described to the operating system via a Device Tree Overlay (Section 3.6.1.1, “Config.txt”). For example, the MyPi base board might use:

dtparam=i2c1=on
dtoverlay=i2c-rtc,ds1307

The most common method to deploy an operating system onto Raspberry Pi hardware is to copy a pre-installed system image onto a boot medium, usually a microSD card. This is the simplest and easiest method.

SUSE provides a preconfigured bootable image of SUSE Linux Enterprise Server for Raspberry Pi hardware. This comes with the Btrfs file system, with compression enabled to improve performance and reduce wear on microSD media.

A microSD card with a minimum size of 8 GB is recommended. Faster cards will give better system performance. On the first boot, the operating system automatically expands the file system to fill the card. This means that the first boot will be substantially slower than subsequent boots.

The process of writing the card image onto microSD media is described in the Raspberry Pi Quick Start.

Some models of Raspberry Pi allow booting from USB mass storage devices. This will then allow deploying SUSE Linux Enterprise Server on Raspberry Pi similar to server platforms.

Installation can be performed from a removable USB medium, such as a memory stick, onto a microSD card in the machine's internal slot. Alternatively, it can be performed from a removable USB medium onto another USB medium, such as a USB-connected hard disk.

Note: USB Bandwidth Limitations

Note that the Ethernet controller on the Raspberry Pi 3 is connected to the device's on-board USB 2.0 bus.

Therefore an operating system running from a disk attached via USB must share the total 480 Mbps bandwidth of the USB 2.0 controller. This will limit performance, and could significantly impact network performance.

This limitation does not apply to the Raspberry Pi 4.

Newer models of Raspberry Pi 3 with BCM2837 B0 silicon (silver instead of black chip), including Raspberry Pi 3 Model B+ and Compute Module 3+, allow booting from USB-connected storage devices by default.

On older models, such as Raspberry Pi 3 Model B or Compute Module 3, USB boot can be enabled by booting from a specially prepared microSD card once. See Section 3.6.1.2, “OTP Memory” for instructions.

Because of the hardware's lack of on-board firmware (Section 3.6.1, “Boot Process”), network-booting the Raspberry Pi using PXE is more complex than with more conventional computers.

The process of setting up a PXE boot server for x86 and Arm is described in the SUSE Best Practices document How to Set Up a Multi-PXE Installation Server.

The Raspberry Pi Foundation publishes information on how to boot using PXE one Raspberry Pi from another Raspberry Pi: https://www.raspberrypi.org/documentation/hardware/raspberrypi/bootmodes/net_tutorial.md

For more information, consult the following resources: