eselarm

I think I mis-interpreted the sentence with 'the computer'; I did not relate that to OPi5 (but 'normal' PC). Re-reading again, indeed the OPi5 is meant. Then the power is there, the question is then, how the RK3588S and surrounding hardware will handle things. I don't know exactly anymore what all sleep states for traditional PCs comprehend. I know what can be achieved in mobile platforms (very low power), but that also includes switching off parts of DRAM that don't contain needed data, and many more like a tiny Cortex-M0 or so somewhere that is always on. The question is also, how much power is saved? For a LiPo battery in a smartphone this is quite different, so lots of extra system design is done in HW (SOC, PMIC). RK3588 should support it I think, but I am not sure if all the various SBCs can use it, because certain HW lines etc might not be there. At least Broadcom/RPi platforms cannot do anything like that.

I have used/tested KDE-neon a while ago December 2024 I see from rootfs backup, that was on a ROCK3A and to enjoy KDE6. Maybe now if you want KDE6, Trixie has it, so as base Armbian (Debian) then upgrade and/or install KDE plasma (sudo tasksel to select and install). I have it on my ROCK5B, at least with edge kernel it works all as expected, compared with KDE6 on Opensuse Tumbleweed and Debian Trixie on N100 miniPC. In-place upgrade went fine, although the whole user environment from KDE5 is gone and you start from scratch new template KDE6 when upgrade from Bookworm->Trixie. If you want latest KDE, then Neon is now already newer than the 6.3 in Trixie AFAIR, but it seems to me that integration issues can be expected. I see Ubuntu Noble (Kubuntu) has KDE5 still.

This is great, I have done some tests with my NanoPi-R6C a year ago. I have not noted what kernel etc, it also has no dedicated power-button (but a 'user-button') so I think I tried (as root) systemctl suspend and logging via serial console cable. Also looked at wake-on-LAN. CPU is the same as OrangePi5 and your logs I recognize going into suspend, but then it stayed frozen. Maybe if I re-purpose the user-button and push that it will resume. Such a button can be easily connected to a pin on the PMIC or SoC. At least 1 of the RJ45 ports supports wake-on-LAN, I could also set in NetworkManager, but it gets stuck in PCI-E somewhere or on the SoC I think, can be very complex, I don't know. For keyboard that is external to the product (the ARM SBC), USB connected I assume, you need to keep the USB power (5V) alive. Also some extra hardware AFAIK (not the SoC) that detects activity on USB keyboard. Can be very simple, but I think it is all not there for SBCs. There is no keyboard controller chip like in traditional PCs. So I will maybe sometime look in schematics and see what would be possible. So far I only focused on power architecture in general, whether USB-C PD works or not, etc. My NanoPi-R6C is 1.5 Watt when idle and vendor kernel, measured with a USB power meter. I will try to reproduce your power-button suspend-resume sequence first and then also measure power. I use EDK2-UEFI v1.1 at the moment, set to mainline and 6.16-rc3 kernel. Is Armbian Bookworm, will upgrade to Trixie first and then see.

I don't know, but what you describe is that OP1 is essentially able to run from RAM (network I/O mainly) so not so much storage I/O which then lowers the change of corruption. For OP2, is quite heavy disk I/O it seems, also torrents writing is tricky for some filesystems. If you want to know more why and how w.r.t. corruption, use Btrfs instead of Ext4. It still can be difficult if HW caches do the wrong thing, but you can make metadata duplicate and also play with commit time. I have a Pi3B running from SD-card and 4T 3.5i HDD via USB attached. It runs Debian Testing (Trixie now) , FAT bootfs and Btrfs rootfs with hourly snapper and also manual extra snapshots if I do upgrades. Also HDD is Btrfs formatted, 'dup' profile for metadata 'single' profile for data. I gets hard power cut every day at least once as solar powered. I do not do any UPS like noticing, also the power loss might be during a btrfs send|receive transfer, that will then fail, but in my script I delete read-write snapshots first as they should be read-only (so were correctly finished). No data loss or corruption during years. I just have a very simple guess about the amount of sunlight, which might be almost zero for days in winter. On a Pi2B that seems to have damaged GPIO pin 4 and was using a bad brand SD-card, I could quite perfectly see when corruption started, which files you see when you do scrub. It was a logfile and some other file that were recently written, but unimportant, so I could still do a btrfs send of a snapshot of the rootfs to my local laptop new SD-card. I know how to create bootfs for Raspberries. For my NanoPi-NEOs running Armbian I changed rootfs Ext4 into bootfs Ext4 + rootfs Btrfs. You can also use Armbian build to generate that off-the-shelf, I would now use FAT for bootfs as that matches how UEFI computers are organized. You also need extra write of U-Boot, that is not needed for RPi or UEFI. An Ext4 filesystem can be turned into Btrfs in-place with tool btrfs-convert. If it is a rootfilesystem where the OS runs from, you still need to create an extra bootfs though and so also change fstab and also as a consequence the partition table and organisation.

sudo apt-mark hold linux-image-vendor-rk35xx sudo apt-mark hold linux-dtb-vendor-rk35xx

I never used Ext4 directly on a whole block device. Linux in general assumes partitions. Advice is to use partitions. So create one first wit fdisk.

mmcblk1boot0 mmcblk1boot0 are for booting via other boot methods, not used for U-Boot in Linux/Armbian. Don't use them I would say, otherwisie the ROM in the BPI might leave you with a bricked board. You can make a partition on mmcblk1, format it and mount it. Same as with other disks.

lsblk should show you 2 mmcblk devices; 1 is SD-card, the other is on-board eMMC. Which is which number can vary depending on what kernel and DTB. If you only see 1, you should change operating system maybe.

OK so it is Armbian. You can use KVM with the 6.1.115-vendor kernel, but you must set fixed allocation in the VM which CPU cores to use. You can chose 0 to 7, 0-3 are Cortex-A55, 4-7 are Cortex-A76. Read the post(s) I linked to earlier, I and others explain it there (even 2 methods). Use virt-manager GUI ane enable .xml editing. then change that line with the cpus, I my case I use 2 cores. You van also do it with virsh CLI tool, but you need to read through the options then.

gfs2-utils But it looks like you want to share storage. You do not need a GFS2 system for that, but a regular Linux filesystem and a NFS server or SMB server for sharing over the network. Most people will use Ext4 or Btrfs or XFS in Linux.

This is not a Orange Pi 5 Plus where this topic is about. Also you need to provide more info, which Armbian version, etc. Your error might have a different reason. In general you can use armbian-config to switch kernels. I never did it like that so cannot really say what you should do. KVM / virtualization you need to read about and understand good enough in order to fix things like this. It is not armbian specific.

You need to create a logical volume in that vgroup and then format that logical volume with mkfs, then mount that. See: https://docs.redhat.com/en/documentation/red_hat_enterprise_linux/6/html/logical_volume_manager_administration/lvm_examples

Maybe others know, but I stopped buying and using USB cabling and hubs for storage/NAS solutions more than a decade ago, except that unfortunately for RaspberryPi4 I had a Samsung 840 Evo 256GB reserved form an older PC. Raspberries only have USB so you cannot use dedicated normal storage I/O protocols like SATA and NVME. Only way to not let it crash every now and then is a USB adaptor cable with ASMedia chipset (not JMicron at least) and own dedicated externally powered USB3 hub. I use 4-port RSTECH with a 2A adaptor, it prevents back feeding 5V from the adaptor to the Pi at reboots and shutdown. If it is all only 5V, assume max 3A for the whole setup, else you need to study your power tree setup. And solder DC/DC converters etc yourself or indeed look for USB hubs that don't exist. USB2 is 0.5 Amps per port, 10x as much you only get when there is special chips and negotiation in the USB-C cables that allow switching from 3A to 5A and also switching from 5v to 9v or 12v or 15v or 20v. I went back to what I did 2 decades ago for PC NAS: SATA and 12V (for a large 3.5 inch HDD). I use a 12V 10A powerbrick, but also enough old and good unused PC PSU units that can do many Amps on 12V and I believe 1 PSU can do 40 Amps on 5V. They have a fan, that is why I do not use them. I use a car battery as UPS. Swapping /dev/sdX names is also often a sign of power problems so USB storage disappears but the kernel still uses a particular /dev/sdX. In mutli-device setups, it is a receipt for disaster mostly (corrupt filesystem).

Maybe have a look at

Have a look in /dev/disk/ You can use the 'by-id' symbolic links normally to identify the storage devices exactly.

Knowing this means your options are to return the case+NVMe back to shop. Or do a cross-check; case with other NVME and NVME in M.2 slot of some other computer. But this is a typical Raspberry problem; storage via USB (SATA or NVME nowadays) easily leads to trouble. Mostly power related, but also many just the chipset in the adaptor. In case of Pi5, you could get an NVMe adaptor board, but also that is not always working out-of-the box. Only if you buy RPi adaptor and RPI NVME it should work out-of-the-box. Or other SBC that has M.2 slot already on the board.

The more SW components you try to let work together in your own systems, the higher the risk something will break as no one else has the same setup. Also if that SW is commercial (and closed-source) things will get worse. Final killer might be external storage like OneDrive, it is not even related to SW actually, just is fundamentally out of your control. If POTUS wants, internet plug is pulled and no access to your data, that, in the mean time can be used for al sorts of purposes you won't benefit from yourself on the longer term. I keep a Windows10 VM around, I had one (upgraded from free Windows Internal test license AFAIR) in VirtualBox, but wanted to move to libvirt/QEMU/KVM as that also works nice for ARM64 and it turned out that the USB extension not always worked (and needed manual install). Also as you noticed, VirtualBox is not in Debian. It was/is? in Opensuse, but with custom kernel module (that is what VMware and VirtualBox need). Now can be based on KVM, but not their defaults and complex as you need to setup/compile? yourself. Windows10 sees another computer when going from VirtualBox to libvirt/QEMU/KVM (with virt-manager as GUI) so I had to buy a new digital license for 20 Euros. I read in september MS will stop delivering updates (unpaid). That day would come of course, so I have a multi-year plan to get rid of Windows (and Google). Windows is almost done. I also have only 1 Intel box running (N100, runs Opensuse Tumbleweed but came pre-installed with Windows11), rest is ARM (or RISC-V or Atmel or Xtensa). Google is more difficult to get rid of, but slowly progressing.

Maybe some further notes on w.r.t. running HAOS in a VM: As indicated, the process is described for Intel, see https://www.home-assistant.io/installation/alternative For ARM, people have selected RaspberryPi and then ask why it does not work. That is because RPi uses proprietary bootloader (not U-Boot, nor EFI). You need the haos_generic-aarch64 build/image, see https://github.com/home-assistant/operating-system/releases I do not use qcow2, but just a flat img. That is easy to mount as loop device in case some changes are needed to partitions or so. I extracted the .img.xz file and changed the storage (VirtIO Disk) of my existing HA VM to use the new native HAOS instead of my old Debian+supervised_HA. It started without issues and I get to the webpage from where the rest is a standard 'HA getting started', e.g. restore its backup so all devices stats etc are back again. I know many people ask for VMware or Proxmox (or some for VirtualBox), but Linux has anyway all the low-level stuff build-in and a GUI as well: root@rock5b:~# apt install virt-manager virt-manager is already the newest version (1:5.0.0-5). I have this running on ARM64 since Debian Buster 64-bit beta RPi (and 1st early 64-bit Ubuntu server image for RPi4 and also Opensuse Tumbleweed 64-bit) so the commercial ones I dropped long time ago. Above is latest Trixie. Kernel is Armbian vendor currently, but that is only because I want video HW accel (ffmpeg codecs) to work. Otherwise Armbian edge or current offers better stability IMO. Simple tests show that DRAM speed for RK3588 is same or better as for old Intel Core I7, so with also up to 32GB, is great KVM server and desktop same as that old Intel boxes, but 4 Watt idle instead of 40 Watt or so. 40W means about 40 * 9000 is 360 kWh per year, that is 108 Euros per year for my energy contract (30 cents per kWh).

It depends on what kind of- and if- HW I/O is used. USB can be redirected. PCI-E can be re-allocated from host to VM when proper IOMMU (HW) is there, that I only go working for Intel PCs, there is interest in it for ARM64, but so far haven't seen anyone got that working. Other typical SBC HW I/O like SPI, 1-wire, GPIO high-low toggle, etc, can not be virtualized, so not available to a VM. Same for video codecs HW accell blocks. Storage and networking, all things typically relevant for enterprise/servers, is no issue.

The way I do it is to run HA (their HAOS) as a virtual machine. The instructions how to do that are quite clear for Intel, so I did that long time ago on 16GB old Intel Atom server board. Standard 32GB image with partition8 Ext4 data AFAIR. Later on, I created a HA supervised VM for ARM64. I had already several VMs running on a RPI4 for other functions, same libvirtd/QEMU/KVM as on Intel boxes. If you don't know how to get that running, it might be easier to use a docker method, but I consider that as more maintenance (longer term) for myself. I have no real use-case for HA as most I do with own SW. Node-RED, influxDB, Grafana, etc, so currently not running/used. It seems supervised method will loose support sometime, so if I would have to set it up now, I would make a HAOS image running on my ROCK5B (16GB) or NanoPi-R6C (8GB) and import a backup of the database from last year. I can share my serial output of electricity meter now (HW and SW wise) point-to-multipoint, so I will maybe have a look. It is mainly to see how and if metering works for my country/contract out of the box. It is all about the money, we need to pay here for delivering solar energy back to the grid. You need to be a financial expert more or less to get it right. And I am still waiting for cost numbers for energy company, they should have informed me before 2025-07-01, but it seems is it impossible to send an email with 2 or 3 numbers (cents per kWh).

It all depends on what version or variant of OS you are using. I have a R6C and have done many fixes and workarounds for various OSses/distros to the point that I just maybe take what generic upstream kernel does. I anyhow change/fake MAC addresses as I use bridges and also VLANs. Is in Armbian Bookworm and Opensuse Tumbleweed. Ubuntu based has and extra netplan layer that is even more a PITA as it keeps .nmconnections file also in another place (generated) then /etc/NetworkManager/system-connections/ Recently a change in scripting has been done so that name 'wan' is/or has been changed into 'wan1', so at least this does not match what is printed on the metalcase. At least that is what I remember from a quick check on github; See what is in your udev rules scripts yourself. https://github.com/armbian/build/blob/main/config/boards/nanopi-r6s.conf

it all depends on what your plan is what the device should do; I have a R6C for about 18 months now and bought it as an alternative/better more integrated alternative for existing RPi4 (and also RPi5 that still has no M.2 socket). I see many people on for example RPi forum who want to create some router thing, so they buy HATs and USB stuff for an RPi, but this NanoPi series has even WRT preinstalled. I removed it as I wanted generic server box (virtual machines host mainly). I do not really need the 2x RJ45 currenlty, but the metal case is great. Is simple passive cooling and insect/animal proof. It is mainly bootloader and kernel (and DTB/overlays) that matter. The userspace can be taken from elsewhere if you know how to. I currently use the R6C as desktop as well in another cooler room (summer hot here). EDK2-UEFI v1.1 bootloader, Armbian edge 6.16-rc3 kernel, Opensuse Tumbleweed for Btrfs rootfs to drive a DVI monitor KDE plasma 6.4.x. It means various Rockchips HW blocks do not work like NPU and video codecs HW accel block, but that is a choice. If you need those, use vendor kernel and see what exctly you need/want. The 2x RJ45 might be a chalenge, see maybe also:

It is now half a year later, this SD-card was put back into the RPi4 and no issues there. The RPi4 has been upgraded from Bookworm to Trixie in-place (just latest pre-release) so gets quite some writes. It still has Bookworm Btrfs snapshot as well, total of about 9GB excluding 2GB swap partition and 512MB bootFAT. Avarage Btrfs scrub speed 41MB/s. The ROCK3A has no SD-card inserted anymore, running fine, except that in the last 40 days I see: root@rock3a:~# uname -a Linux rock3a 6.1.115-vendor-rk35xx #1 SMP Fri May 30 01:18:17 UTC 2025 aarch64 GNU/Linux root@rock3a:~# dmesg | grep "page allocation failure:" [ 527.016616] dmcrypt_write/2: page allocation failure: order:0, mode:0x800(GFP_NOWAIT), nodemask=(null),cpuset=/,mems_allowed=0 [2879228.441453] dec0:0:hevc_rkm: page allocation failure: order:5, mode:0x40dc0(GFP_KERNEL|__GFP_COMP|__GFP_ZERO), nodemask=(null),cpuset=user.slice,mems_allowed=0 [2879523.127829] dec0:0:hevc_rkm: page allocation failure: order:5, mode:0x40dc0(GFP_KERNEL|__GFP_COMP|__GFP_ZERO), nodemask=(null),cpuset=user.slice,mems_allowed=0 [3029341.482298] kworker/0:1: page allocation failure: order:0, mode:0x800(GFP_NOWAIT), nodemask=(null),cpuset=/,mems_allowed=0 The middle 2 is clear what source is, the others I do not know. I think I will first upgrade userspace to Trixie. I currently have no clue what the details w.r.t. page faults mean, so have to see and figure out what issue could be or maybe mitigate somehow.

Yes it is dead, that is why my comments between brackets. Start with http://beta.armbian.com then click with your mouse per sub object. Or wait until it is out of beta phase or modify your sources.list.d