Death by Fedora

Or “How I Learned to Stop Worrying and Love SELinux”

Timeline

Date Activities
2022-04-07 Reform DIY kit w/ NVME SSD and ath9k card arrives from CS, unit quickly assembled, boots
2022-04-08 LPC and Keyboard firmware updated from source, LPC status indicates missing cells (last two cells are reported as 0.3v and 5.3v respectively)
2022-04-09 Investigation reveals that the right battery connector was not fully seated, battery status now reported as normal
2022-04-10 Attempts to boot Fedora’s stock aarch64 image (loaded on NVME) result in failure
2022-04-12 through 2022-04-15 Various attempts to build a custom kernel, finally resulting in a working config, with slight modifications to reform-init (using the SD card’s /lib/modules as a bind mount on the Fedora chroot) and mkkernel (to build and install modules).
2022-04-15 After a successful test boot with SELinux set to permissive, a full relabel is attempted.
+30min I elevated the screen-side edge of the Reform to help cool it. Having been entirely on battery during this day’s tests, I plug in the charger. While no progress is reported, the kernel is still running as determined by plugging in a USB device and observing the console log.
+10min The reform appears powered down, and hot to the touch in numerous places. The LPC is unresponsive to pressing the circle key. A smell of overheated electronics is present. A survey with an IR thermometer reveals multiple parts of the motherboard are above 71°C/160°F. The batteries and power supply are hastily disconnected.
2022-04-16 The Reform is relocated to a less flammable area for testing. The right battery connector is measured via IR as nearing 87°C/190°F within 10s of being reconnected. Peripherals are gradually disconnected to eliminate possible causes and potentially reduce damage. Removal of F2 fuse does not impact overheating. Testing with batteries is discontinued due to the difficulty of ensuring a quick disconnection in the event of trouble. Testing continued via provided power supply connected to a surge protector with an easy-to-access off switch. Q17 is measured at similar temperatures to the right battery connector after an equally short time on.

Is there any update on this? Or is running Fedora on the MNT Reform in a usable fashion not possible at the moment? What about your custom kernel? Do you have a .rpm or a way to install it?
Would you recommend trying your current set up at all?

Is there any update on this?

The motherboard is going to be shipped back to MNT for diagnostics and repair.

Or is running Fedora on the MNT Reform in a usable fashion not possible at the moment?

I don’t know. The process seemed to be working smoothly until the Reform imploded. I’ve been told that it’s very unlikely that the damage was caused by the OS, but I can’t shake the feeling that something is amiss.

What about your custom kernel? Do you have a .rpm or a way to install it?

The kernel booted off the SD card and chroot’ed in to the Fedora partition on NVMe with a bind mount for the modules. Fedora’s kernel was never used, and RPM wouldn’t be able to install or update it.

You could try building from my config, but I wouldn’t recommend it until the board is checked out and it’s verifiably confirmed that the damage was from some sort of mishandling on my part, or hardware failure rather than some sort of weird kernel interaction.

Would you recommend trying your current set up at all?

My current setup involves a short circuited board and figuring out shipping to Germany, so no, I can’t say I would recommend it. :sweat_smile:

Thanks for your reply.
Would you mind posting your kernel config? If anything it might be useful for reference, I’d like to see Fedora running on the MNT Reform at least in some capacity. I have a separate I.MX8 board I got from an old employer laying around, if it got fried it wouldn’t hurt my feelings too bad. Maybe I can experiment with it.

1 Like