eselarm

Providing logs with armbianmonitor -u helps with troubleshooting and significantly raises chances that issue gets addressed. and/or lspci output it might also be a power issue or just that you specific nvme does not work with non-vndor kernel also kernel is already at 6.18.35 for current rockchip, maybe see if that fixes something w,r,t, yiour nvme

If 2 images don't work, I start thinking that the images are OK, but something goes wrong after extracting/decompressing, writing to SD-card, and reading at power by the bootROM/maskROM and U-Boot. I have various images downloaded and checked the past months, only for topics like this and I remember only 1 being wrong (there was no rootfs, only bootloader). I do not have a BPI-M2S, so cannot really test on hardware. And hardware might be the issue. Quite often nowadays it is fake or counterfeit SD-card. Those should work OK for writing the image and even verifying, but after that all sorts of strange errors and corruption can happen. It might be something with MBR or GPT failing, but this is rare. You simply need to check much more, check the SD-card in other Linux computer with fdisk etc. And connect serial console cable, so you can seen U-Boot and kernel logs and also interrupt U-Boot and do manual scan commands. Ultimately, you might need to build an image yourself and use Btrfs instead of Ext4. Might also make U-Boot understand Btrfs. Then any block-level corruption will be detected as Btrfs has checksums on all storage blocks. I use it for more than a decade and it really helps catching issues, even after years the same Btrfs root filesystem. And bad SD-cards are easily detected. Note that there is also btrfs-convert, allows converting existing images or SD-cards, but is a bit tricky as you need to do a few other things w.r.t. bootloader.

There is only this overlay for vendor kernel: /usr/lib/linux-image-6.1.115-vendor-rk35xx/rockchip/overlay/radxa-zero3-rpi-camera-v1.3.dtbo The mainline based kernel, what you use, there is none. So you would need to port it, from this radxa03 SBC or look at other devicetree overlay files, maybe from RadxaOS, maybe from RPiOS See 'man dtc' as a start, it might be a complete renaming of a lot of the .dtso file is needed, maybe it is just an enable line.

I have a few of those and when connected to RPi5 PSU and set to 12V, it can power my Rock5B with NVME SSD and some other peripherals as well.

The (standard) Armbian images for Rock5b use boot.scr, not efi/boot/bootaa64.efi from an extra FAT/ESP partition. So no surprise. If you use the EDK2 UEFIv1.1 in the SPI-flash, you should use a generic UEFI ARM64 image. 1 such is Armbian for UEFI ARM64, look at the download options. I don't know, but make sure it is one with latest mainline kernel 7.0.x. What I have used is a Debian Sid nocloud image just as a quick test (testing that image on a NanoPi-R6C, should also work on Rock5b, not the UEFI/bootloader behavior. See https://cloud.debian.org/images/cloud/sid/daily/latest/debian-sid-nocloud-arm64-daily.tar.xz As I mentoined earlier, I use fixed 12V for ROCK5B, else no success. For NanoPi-R6C it works with 5V 3A (27W RPI5 PSU).

You will have to make a choice which bootloader firmware you use. And then also which boot method and/or bootmanager. Armbian images for Rockchip SBC's are as simple as possible, so only 1 OS, no ad-hoc kernel selection. That means 1 partition for rootfs, no others. Different methods have different advantages and disadvantages. If you want other boot method and/or bootmanager, you will need to add a boot partition yourself, usually FAT formatted and type 0xEF00 (called ESP). The Armbian rockchip64 supports it all, no problem. Especially the Rock5B, which is more or less the blue-print of high-end ARM64 SBC. But all this own custom action also make it not Armbian anymore, but just a generic UEFI ARM64 computer. Where you could also boot from a CD-ROM .iso image and run a traditional Linux distro installer like used on x86 PC's.

Another option is to wipe the U-Boot from the SD-card and just use the SD-card only, no USB-adaptor needed. Then the only U-Boot in the system is the one in SPI-flash. I have done that once on an SBC with no SPI-flash but eMMC where the bootloader is stored. And that also involves GPT partitions and so on, not independent like SPI-flash. But as Werner says, backup is easy. For old 32-bit Armbians, I made some function in my unattended backup scripting to do that, so that after a year(s) or so, I can restore to get exactly the same system (when SD-card gets bad/broken for example).

Providing logs with armbianmonitor -u helps with troubleshooting and significantly raises chances that issue gets addressed.

OK, good that you see the higher voltage selected. I have also seen it, but was on a NanoPi-R6C, although it did not work with RPi5 27W PSU, but OK with HP tablet 45W PSU. The rest is then a matter of kernel and DTB. My ROCK5B runs fine with rolling release/latest Linux kernels (Debian Sid, Opensuse Tumbleweed), currently 7.0.10 version. Don't know about NixOS (on aarch64), but maybe first get some basic stability with standard Armbian Trixie image (mainline based). It gets tricky when you mix vendor and mainline/distro based. Also vendor (6.1.115) can show various issues if you try/use various external HW. Like an mt9721u USB stick of mine causes a failure in paging on ROCK3A, so need reboot, then it works, but don't know for how long (weeks or months)

this means bootloader in SPI-flash is used if armbian is booted from sd-card slot, the armbian bootloader from sd-card it used, at least that is what i think so you can wipe spi-flash or wriie a working bootloader in there

I was curious and looked into your script, I see parted uses sizes like 32MiB instead of sector numbers I tend to do when using gdisk or fdisk It is only the function write_uboot_platform_ufs() in the image that deals with 512/4k and then it is no issue for Btrfs as this is not related to CPU page-size issues I hinted earlier. So also older kernel is no issue and also RK3576 cannot do e.g. 16k page-size.

this I cannot remember I have experienced on my ROCK5B, it was indeed boot-loop when power 'was not good enough'; with fixed 12V power everything worked, even 3.5mm audio I remember. Now I have EDK2 UEFIv1.1 for the ROCK5B in the SPI-flash and it is like a PC, so even stores boot entries for several OSses (what is on the ESP). My ROCK3A showed similar behavior as your ROCK5B, that went away when fixed 12V power and Armbian legacy U-Boot and 6.1.115 (I use SATA overlay) At sector 64 the binary u-boot.bin is written, the terminology I am not sure of. You can look up all addresses the ROM uses at rock-chip.com or so. You might want to build/compile an as pure as possible and as latest as possible mainline U-Boot for ROCK5B, I saw kernel 6.17 or later should have fusb302 support, but it might be too late in the power-on process. EDK2 UEFIv1.1 has no fusb302 support.

root@nanopi-m5:~# grep BTRFS /usr/lib/u-boot/nanopi-m5-rk3576_defconfig # CONFIG_CMD_BTRFS is not set # CONFIG_FS_BTRFS is not set So no direct Btrfs rootfs support. It is image Armbian_26.5.1_Nanopi-m5_trixie_current_6.18.33_minimal.img run as container quick test. # gdisk -l Armbian_26.5.1_Nanopi-m5_trixie_current_6.18.33_minimal.img GPT fdisk (gdisk) version 1.0.10 Partition table scan: MBR: protective BSD: not present APM: not present GPT: present Found valid GPT with protective MBR; using GPT. Disk armbian.img: 3252224 sectors, 1.6 GiB Sector size (logical): 512 bytes Disk identifier (GUID): 1F37D403-5DE7-4F81-AB22-D9825AE1EFEB Partition table holds up to 128 entries Main partition table begins at sector 2 and ends at sector 33 First usable sector is 2048, last usable sector is 3252190 Partitions will be aligned on 2048-sector boundaries Total free space is 32735 sectors (16.0 MiB) Number Start (sector) End (sector) Size Code Name 1 32768 3250175 1.5 GiB 8305 rootfs There is no 4K sector 'awareness', that is the whole reason I ran this testing; If that is an issue with the M5, I don't own and don't know, but I have encountered quite some issues in Linux in general w.r.t. that, so I would for sure scan armbian-install script to see what is done. Risk with 6.1.115 kernel is that is messes up 512/4k things when Btrfs as kernel 6.7 or later is needed to handle it transparently. Check all yourself, as is long time ago I checked it and I use mostly latest Linux, so kernel > 6.18 currently.

My ROCK5B was ordered as a 'blue one' on Aliexpress a year ago; I got a 'green one' and as with included case the price was > 150 euro, it got stuck at customs for quite some time. It was with blank/empty SPI flash and endless boot-loop. That was already known by me, so I had fixed 12V soldered to a male USB-C connector. Same story for ROCK3A, although that one booted most of the time but then while running al sorts of strange errors and or crashes. So I think you have to assume USB-C PD negotiation is not working with the Armbian images you tested. Only the U-Boot starting at sector 64 does matter, rest (Linux) is don't care. If a stable U-Boot or UEFIv1.1 form SPI-flash or SD-card, you have already a very extensive commandline interface or text menu's in the EDK2. But lock up might simply happen because the PSU does not do 65W, instead max 15 W (only up to 3A) and the 5V might drop way too low, just very short dip that is not measurable without good oscilloscope or so. The Radxa image might have a bootloader variant that does do USB-C PD, although I haven't seen proof anywhere that it correctly handles and does operate the FUSB302, but that might simply be because my ROCK3A and ROCK5B came out of the box with empty SPI and for ROCK5B I have ignored Radxa images anyway after the big troubles with ROCK3A, so do not really know what happens with those. So I have no other advice then use fixed higher voltage, formally it is >= 9V according to Radxa wiki/docs AFAIR, but I remember the powertree design actually better than those quick facts and my conclusion was : forget about 5V, use 12V and it was easy decision for me as I use 12V UPS and 3.5inch HDD that also needs 12V anyway.

Here your debugging activity will need to start. There is an awful lot of things unknown, to others on this forum but more important, also to yourself. It could already be that your server (Armbian) has nothing to do with it, but that it is the info in your router (even caching stuff) that has refreshed, simply because you probably restarted, so something was not 'reset proof'. I have seen many of such things myself in my house/home networking and in conjunction with various ISP's (also mobile4G). Which made me use my own (open-source) router instance for example, so I can at least check all what is needed in such case as you have. But it might be simple, you can use wget (do 'man wget' first to see all options) to mimic browser behavior and that should already say something. If docker etc is involved, make sure you understand all (virtual) networking, it can get very complex. I haven't seen potentially disturbing things from Armbian 26.2.1, but I removed/blocked also several things, like no netplan.io, that at least works on Debian Trixie based installs. There have been lots of security related updates in Debian recently, so make sure you log and see changes to packages (edit listchanges.conf)

This topic is 2 years old and HA formally does not support supervised anymore. And doing 2 such modifications and then hoping all will be fine is false hope IMO. If you don't have a serial console cable to watch kernel en journal log etc to see what your system is doing, maybe it is time now to get one. I have an old supervised inatsllation (not use for many months, but it is plain Debian and a VM so I have free/easy CLI console.

What U-Boot/bootloader is in the MTD?

I just checked at my ROCK5B running Armbian userspace: # sudo apt update # sudo apt list -a linux-headers-*current-bcm2711 linux-headers-current-bcm2711/trixie 26.2.1 arm64 linux-headers-current-bcm2711/trixie 25.11.2 arm64 linux-headers-current-bcm2711/trixie 25.8.2 arm64 linux-headers-current-bcm2711/trixie 25.8.1 arm64 linux-headers-current-bcm2711/trixie 25.5.1 arm64 linux-headers-current-bcm2711/trixie 25.2.3 arm64 linux-headers-current-bcm2711/trixie 25.2.2 arm64 linux-headers-current-bcm2711/trixie 24.11.1 arm64 linux-headers-current-bcm2711/trixie 24.8.2 arm64 linux-headers-current-bcm2711/trixie 24.5.1 arm64 linux-headers-current-bcm2711/trixie 24.2.1 arm64 So you can install them with: # sudo apt install linux-headers-*current-bcm2711 You will get the latest if you do not select explicit version 6.18.9, I got 6.18.10 as that seems to be the latest now. Armbian also has edge and legacy, but for normal release all 64-bit Raspberry Pis is named 'current-bcm2711', which should be the equivalent of 'rpi-v8', which is the normal 4k pages downstream kernel in Raspberry Pi OS 64-bit. It might be you originally had the dedicated rpi5b installation, so you simply had not gotten the headers maybe. Armbian has no 16k pages kernel, if you want that, you need to build yourself, but note that this comes with quite some of issues, many people are not aware and cannot fix issues due to that. See 2+ years of trouble w.r.t. that on RPL forums. I do not use Ubuntu based Armbian, but Debian based, but should not matter for that kernel packages as those are Armbian on top of either distro.

I think most important is that people make sure they can fix their own issues if a HW/computer fails. I found high-availability interesting, but for just my house (or even 2 places/countries at the same time) I found it too much to make it all work. The thing I could maybe use is DRDB, however I see v9 is under development since 2011 and it still is not in mainline kernel. Compared to that I have some script to transfer latest differential Btrfs snapshot from 1 computer to the other on-the-fly, I doubt I can really benefit from DRDB, but maybe I set up a test and see what it does. I use only ARM64 for 24/7 servers, so no x86_64 <=> ARM64 incompatibility. So same as for ZFS for example, it is external to Linux distro and my experience is that complicated issues/failures always happen at the wrong moment, e.g. also no internet and/or mains power failure etc. So I try to minimize the amount of 'external' HW/SW modules, certainly if the backing company is commercial and in whatever country far away from where I am. Proxmox is nice, but also 'external'. Same for even Docker, so I have actually no such containers. But it all depends on how much you are involved in various HW and SW. As you can see in my earlier message, I already forgot HA was available as generic aarch64 image, so that says enough. I use several custom (own) HW (like 'changed' solar inverter) and HomeAssistent does not support it, so I actually do not really use it. It is mostly C-code and Node-RED in conjunction with various micro-controllers. A good test is to disconnect internet (power-off fiber-RJ45 box in my case) and/or also do an ad-hoc power cut (no cheating with UPS). And then see if you can get it all running again within a certain time frame (what you think is acceptable, like before temperature in freezer gets > -10 or so).

I see it is this board, quite new: http://www.orangepi.org/html/hardWare/computerAndMicrocontrollers/details/Orange-Pi-4A.html You can see at https://github.com/armbian/build/commits/main/config/boards/orangepi-4a.csc that no manintainer is defined, so up to yourself to figure out what is wrong. You need to start with attaching a serial debug cable (see 3 pins next to power key) Also I see this board has SPI-flash, there might be an old U-Boot variant in there that does not work together with mainline based 7.0.x kernel.

I have just cleaned-up / changed (again) things w.r.t. network names for my Nanopi-R6C. Same as Nanopi-R4S, it has 'wan' and 'lan', where wan is on-chip ethernet and lan is external PCIe connected. Double check for the R4S, but I am almost sure. So a mainline kernel in principle just numbers them eth<x> depending on timing and other things. New kernel behavior is that the on-chip gets named end0 and for my Nanopi-R6C the external gets named enP3p49s0, as that represents how it is HW connected via PCI-e. That is 7.0.<x-flavor> generic distro kernel. Then Armbian has 2 udev rules to rename to wan? and lan?. That is where the problem was and is for me as I use various different bootloader/firmware. I removed those udev rules as I run standard main distros as well and I need the names for bridges and vlans etc future proof and I don't care about names wan and lan (not using it as router). But if you want to keep that naming, check bootloader version (U-Boot I guess for you) and also udev rules and maybe other patches for RK3399 and/or R4S.

You know there is a fruit company in Cambridge UK that is proud to tell they have sold 75 million pieces and 75% is to non-noobs. They provide passwordless sudo by default and a self-modifying rpi-update script by default in their OS. So why so complicated with this copy-fail exploit, every script kiddy had and has an easy task of keeping the noobs on a leash for more than a decade.

I see it like who is motivated to check and do the work w.r.t U-Boot. Especially while EDK2 UEFIv1.1 is available already a year and at least for my ROCK5B and NanoPi-R6C with both Armbian rockchip edge kernels as well as standard Debian arm64 kernels works OK. We call those things 'firmware' but the 'firm' selling this OP5 HW is not delivering this. Good thing is that it is all open source at least. For the ROCK5B I also used it in 'vendor' mode, so 6.1.115 works and so video encoders as well (for Jellyfin). For NanoPi-R6C (RK3588S, same as OP5) it is in the on-board eMMC, that is some disadvantage compared to SPI and more as SD-card. So not sure how OP5 will deal with it.

There will be no fix if it is not clear what exactly needs to be fixed I would say. I guess you need to dig deeper yourself. You can build U-Boot independent of Armbian if you want, maybe less reading initially, still the testing/debugging is what matters.

Yes that is what it seems. But strange that mainline kernel+DTB gets it right and 2026.04 U-Boot not. You need to look into the build-config of this U-Boot for the OP5 then I think. All PCIe NVME needed for boot might be there, but still other factors might determine fail or success. Maybe timing at power on, some power line or other signal not enabled. You could flash https://github.com/edk2-porting/edk2-rk3588 into SPI and see if you can boot mainline based kernel with that with just SPI and NVME. Maybe other things don't work then, but depends on what the board should do. You need to add or already have an extra EFI type FAT formatted partition with that, usually done manually and also installing grub-efi package and installing all that manually. Also you may want to add loglevel=7 to kernel command line, then you see what is happening after 'Starting kernel'