I’m trying something different for this post – it's spoken (a short podcast, if you like).
This is my first ‘week bits’, where I talk about something (hopefully interesting) that happened during the previous week. It’s an experiment, and I won’t do one every week (or possibly ever again).
This week, I’m talking about Ubuntu, the Linux-based operating system. On Friday, I turned my laptop on in the morning, as I do most weekdays, and there were some unusual errors displayed while Ubuntu was booting. They read: ‘blacklist: Problem blacklisting hash (-13)’. That same error was repeated 33 times. I’d never seen these errors before, so ‘What is that?’ was the main thought in my mind. Ubuntu continued booting without problem, so I ignored the errors for the moment.
Once I logged into Ubuntu, and opened Firefox, I ran into another problem: it wasn’t loading any web pages. In fact, I seemed to have no internet connectivity at all. I checked the Wi-Fi status in Ubuntu and it wasn’t even connected to a network. ‘That’s weird’, I thought. I saw one of my Wi-Fi networks listed and manually connected to it, and my internet was back.
That, however, wasn’t the end of my Friday woes: later that morning, when I was on a Teams call in Chrome, I flicked over to Slack and my display froze, flashed black and then remained frozen. I still had audio so I stayed on the call, and then forcibly rebooted my laptop after the call had ended.
When Ubuntu booted back up, once again it wasn’t connected to a Wi-Fi network. This time I paid a bit more attention: my laptop was no longer seeing my 6GHz Wi-Fi 6E network. It was still seeing my 2.4 and 5 GHz networks, and although those ones were also saved, Ubuntu wasn’t configured to automatically connect to them as it was for the 6 GHz network. I suspected some kernel and firmware updates Ubuntu had installed the day before were the cause of the Wi-Fi 6E problem. I had a quick look at the update logs, but I didn’t have time to investigate further, so I just made Ubuntu connect to my 5 GHz network automatically so I could get on with my day.
Then the afternoon came. Twice more my display froze. The first time I was using Teams and Slack again, the second time I was just using Slack.
I’d had enough at that point and dug through the system logs, looking for clues. And it didn’t take long to find some. There were loads of amdgpu ‘no-retry page fault’ errors, all mentioning Slack, at the time of the last display freeze. Bingo. I searched the Ubuntu bug tracker for that error, and quickly found an entry, but for Ubuntu 22.10 Kinetic. It looked like the same bug, but I was running Ubuntu 22.04, the latest long-term support release. There was a link to the upstream bug for the problem, and the bug was introduced in kernel version 5.19, and fixed in 6.1.
I was confused about why this would suddenly start affecting me. So, I looked at the Ubuntu update logs more closely: the kernel had been updated from version 5.15 to 5.19. That explained everything.
Based on what the amdgpu bug report said, I decided to update to the Ubuntu linux-oem-22.04c kernel, which is based on the upstream 6.1 kernel. I also manually updated some firmware images from the upstream linux-firmware Git repository to see if it helped with the Wi-Fi 6E problem.
The kernel update solved the display freezes, but the Wi-Fi 6E problem remained. I checked my router settings and noticed that the 6 GHz network was configured to use a non-preferred scanning channel (or non-PSC as they are called). I switched to a preferred scanning channel, and Ubuntu saw the Wi-Fi 6E network once again. I also checked the behaviour in Windows 11 and it had no problem connecting to the original channel I was using. Regardless of which one is correct, I don’t really expect to turn on my laptop one day and find it won’t connect to my Wi-Fi network. The polite thing to do would’ve been to warn about the problem with the channel for a period of time before goign on to break it.
The last problem remaining was the unhelpful blacklisting errors on boot. A quick search online showed those errors are also new in version 5.19 of the kernel, and are caused by a BIOS bug relating to secure boot keys. But the errors can, apparently, be safely ignored. So, ignore them I shall.
So, all-in-all, a turbulent kernel update, and I’m surprised Canonical is pushing the 5.19 kernel to Ubuntu 22.04 with such a severe AMD GPU bug. No doubt there are others running into the same problem, as I speak.
And that’s the end of my first ‘week bits’. Hope you enjoyed them.
Links to the AMD GPU bug mentioned:
The kernel I ended up installing: