About a week after announcing the CentOS 7 alpha build for AArch64 hardware, I got an email from the fine folks at AMD asking for my address. Three days later, a pair of AMD's Seattle AArch64 development boards showed up at my front door. Hardware really is the best sort of gift, especially when you're building for a new platform. I needed to make a couple modifications to my lab (okay fine, my home) network in order to make use of the new AMD gear due to the board layout. The boards I have don't include USB by default, and I usually cheat and do USB based installs. Instead, I set up a proper pxe environment and loaded the OS up the right way.
Because the hardware was shipped with UEFI by default, there was very little to do in the way of configuration. I simply pointed the serial console at the pxe option and then did a normal vnc based install when the installer loaded. I don't want to spend too much time here talking about the installation, because it was exceptionally simple. If you don't have or want PXE, a pcie usb card would work fine. You can even simply yank the drive and dd the disk image directly to it, then power up. The only downside to the last approach is the need to run efibootmgr manually to set up the UEFI boot entry. In the coming days, I'll be updating our AArch64 documentation to reflect the added hardware and providing more detailed directions for configuration and setup for the hardware.
Apart from being able to report that the AMD boards work just fine with the CentOS 7 AArch64 beta, I can also happily report that they're proving quite useful. Since I have multiple machines, I'm now able to run a number of test cases and multiple configurations without the need to cannibalize my build system or boot from multiple drives. This has given me the ability to run our t_functional testing suite against the AArch64 beta, and identify a few minor build issues that I need to address before we can call the release GA. Thanks AMD!
One thought on "AMD Seattle and CentOS on AArch64"
We are trying CentOS in a AMD Seattle box. After installing it using serial line (the ASUS graphics card AMD told us to use did not work) we are having problems enabling the network.
When using DHCP settings the last kernel message is "[ 22.509670] amd-xgbe AMDI8001:00 eth0: Link is Up - 1Gbps/Full - flow control off
" but no IP is assigned.
Running "nmcli c down eth0; nmcli c up eth0" ends up with:
[root@seattle ~]# nmcli c down eth0; nmcli c up eth0
Error: 'eth0' is not an active connection.
Error: no active connection provided.
Error: Connection activation failed.
and when using static settings all seems to be correct (ifconfig output and kernel routing tables) but packets are always lost.
Have you been able to get the network running in this box using CentOS?
Thanks in advance!,