Bpi-cm4 A311D with encrypted NVME - questions?

MNT Reform
1

The good news!

I have managed to install the CPU upgrade and boot off an SD card.

Wifi works! Performance is improved! I can (manually) mount the encrypted NVME drive!

Using reform-emmc-bootstrap seemed to work, and I haven’t soft-locked anything as I can still boot from SD card, but without the SD card the screen stays black and the system is just kind of there – HDMI and eDP screens are both black. The oled control panel is still responsive, as always.

Based on other threads, I’m not actually sure I can successfully boot off of emmc with the A311D board, yet? And there is a risk of soft-lock if I try flashing/reflashing things (which I thankfully haven’t run into yet).

Failing that, I think what I want is a dedicated SD card that’s got my boot partition on it, which mounts my encrypted NVME as the root filesystem (and swap).

Right?

Are these instructions still current?

Is there anything needed to convince the system to always boot using the internal eDP screen, not the HDMI output? It seems to intermittently decide to use the HDMI output instead, and I haven’t figured out the criteria yet…

Thanks for all the help already!

2

That is odd. When you ran reform-emmc-bootstrap it should’ve printed a warning but otherwise should’ve offered to flash u-boot on eMMC. Did you take this offer or did you decline it?

Yes, the risk exists. But unless you flash an unknown u-boot binary or loose power during flashing at a very bad moment, all should be good. Flashing u-boot to eMMC using reform-flash-uboot has been successfully confirmed to work by multiple parties (including myself). I am booting my A311D classic Reform with u-boot and /boot on eMMC and do not need an SD-card anymore. In fact, this has already saved my butt because I temporarily had my sd-card reader broken and had that persisted, I would not’ve been able to boot my machine at all (luckily I was able to fix this issue with the help from minute by bending some of the pins back again).

That is possible and I have been doing that for a long time before I switched to having u-boot and /boot on emmc.

No, they are not. They are as outdated as the instructions in Howto: Upgrade an existing MNT Reform system on encrypted NVMe for RK3588 So I now added the same disclaimer to that article. Though in principle, if you want to do things manually, then yes, the instructions are still okay in general.

Are you booting with a HDMI display attached? I haven’t seen this problem with my A311D classic reform yet but I’m also still on kernel 6.12…

1 Like
3

I did accept flashing uboot! And I’ve manually verified that there is a boot partition on the emmc. But boot without an sd card is currently just giving me a blank screen, kind of forever, with or without an HDMI screen attached. I can try again. I have tried both with an without an HDMI screen, and I get no output on either device.

Additional good news – I followed those instructions with two small changes:

  1. I used the linux-image package that matched what my encrypted NVME drive already had, not the one in that post and
  2. I used a blank sd card in an adapter (presenting as /dev/sdb) instead of /dev/mmcblk0p1 because that had the rescue image and I didn’t want to clobber that. I created a boot partition manually on this blank card and mounted it on /mnt/boot and did the reinstall that way.

I shut everything down, ejected all the cards, put the new boot partition card in, and was able to boot just fine, and got prompted to unlock the NVME card.

I am still oscillating between the boot chosing the HDMI port or the internal screen, and I don’t know why that would happen. It’s possible I kind of squished my flex cable funny? Maybe there’s some sensor in there? Or there’s some timing thing with 6.14?

I haven’t looked in journalctl yet, so maybe there’s a hint.

I would like to boot off the emmc, so I’ll keep poking at it. Let me know if there’s anything diagnostic I can run!

I do seem to see both the /dev/mmcblk1pX and the /dev/mmcblk1bootX partitions. Is that expected?

4

Do you have a USB UART adapter that you can use to look at what gets displayed via serial in the very early boot stages?

Yes, the bootX partitions are normal on eMMC.

5

RE UART adapter – I think I have one somewhere, but I haven’t used it much/ever. I can definitely go hunt it up, or if worse comes to worse buy one.

However, it’s now working (well, I don’t know if the spurious HDMI issue is still happening, but I am currently booted off the emmc).

I manually ran reform-flash-uboot emmc, and also did the manual kernel image reinstall step with the emmc partition instead of the sd card. All the relevant files appeared to be there anyway, but I decided to do it just in case. I also manually made the mmcblk0p1 <->mmcblk1p1 change (I know the previous runs of the reform-emmc-bootstrap script had already done this, but I’d edited it to put it back to the SD card so I could boot from there, so I undid that change).

I’m confused, but happy. I’ll keep monitoring, and I’ll see if I can find my uart cable/get a uart cable. I am curious what’s going on with the HDMI flapping. I can also open it back up and seeing if my flex cable is pinched/awkwardly strained.

Thanks again for all your help, both now and virtually in the past via old forum threads!

I am really enjoying the tinkering!

6

Additional data point!

My first boot on the emmc went to the internal monitor. The first shutdown/reboot after that success and after using the unit for several hours, went to the HDMI output (nothing on the screen, plugged in an HDMI monitor and the nvme encryption password prompt was displaying there).

Subsequent reboots – one after the other, both selecting reboot from the three dot menu in waybar (under wayfire) and selecting shutdown then rebooting from the oled menu) resulted in the display staying on the internal monitor. No interference from me in terms of where the output should be (I didn’t use wlr-randr after getting the HDMI output, because I had a sneaking suspicion it was alternating – and it wasn’t). I feel like this is either a kernel timing thing, or I’ve missed something or done something weird.

I’ll keep trying to figure out what’s going on.

7

Things seem mostly stable – most reboots, the internal panel is used as expected, but at least once since this thread (earlier this evening) I rebooted and got bumped to the HDMI port.

I’ve located a 3.3v USB A to 4 dupont pins FTDI cable, which I think I could use on the reform’s UART pins, but just to make sure I’ve also ordered a 3.3v USB A to 3 dupont pins UART cable (I think the only difference is a dupont connector labeled Vcc, which would only be needed if the device being connected to needed external power, which is not the case here).

One weirdness – when I ran wlr-randr to switch from the HDMI screen to the eDP screen, the eDP screen never woke up, and I ended up using the oled menu to reboot, which selected the internal screen as desired.

I examined both the dmesg and journalctl logs and the only difference I can see is that on the second boot, with an HDMI monitor still plugged in, but where the boot activated the eDP screen, the journalctl output had a couple lines like:

May 14 19:10:29 reform dunst[1858]: WARNING: Screen '0' not found, using focused monitor

I think this is a red herring, because previous boots, which went to the eDP screen did not have this, and the boot which went to the HDMI port also didn’t have it.

I will keep poking at it. On the boots where things are pushed to the HDMI port, zero output goes to the eDP, and this seems like it has to be happening pretty early – maybe during uboot? I have zero experience looking/using/debugging uboot for real (other than using SBCs which had it installed), so this is a good opportunity for me to learn, maybe.

Sorry I’m being so talkative. I am having a good time with all this.

2 Likes