[OS] Debian Bullseye for ODROID N1 (WIP)

[meveric]

This is my version of Debian Bullseye (11) with upstream Kernel support for the ODROID N1.

Due to the fact, that the Rockchip RK3399 hardware is one of the best supported ARM hardware in Linux Kernel at the moment I wanted to test what is currently available and working and was surprised to find that nearly everything worked out of the box.

This image is based on Debian Bullseye.

It's a headless server image only with user root. But it has all my repositories included, which allows for easy installation and updates of packages such as Kernel and Headers and other packages.

The image has my usual setup: means on first boot it's resizing the rootfs partition and configures SSH. It will automatically reboot after the initial setup after which this image is ready to use.
Kernel and headers are already installed if you need to build your own drivers.
A few basic tools such as htop, mc, vim and bash-completion are already installed for your convenience.

Download: https://oph.mdrjr.net/meveric/images/Bu ... -N1.img.xz md5 sha512 sig

Size compressed: 165MB
Size uncompressed: 1.3GB

Default Login: root
Default Password: odroid

It's ready to be used as a server image.

First things first:
Since there's constantly new development for this image and ODROIDs in general the First thing you should do after the image is up and running is to install all updates:

Code: Select all

apt update && apt upgrade && apt dist-upgrade

How to convert the image into a Desktop image:
The image can easily converted into a Desktop image looking similar to this:


I created a setup tool that's already preinstalled:

Code: Select all

setup-odroid

Which should easily allow you to install a Desktop Environment.

You can even install Gnom3 under wayland:


Hints:
Changing language:
If you install a Desktop environment you should already have keybord-configuration installed, it will ask upon the first installation what keyboard layout it should use.
This configuration is for your X11 Desktop so you have your desired keyboard layout.
For console you need to install console-setup to do the same, so just in case:

Code: Select all

apt-get install console-setup keyboard-configuration
dpkg-reconfigure keyboard-configuration

should be all that is needed to change your keyboard layout.
You probably want to set timezone as well

Code: Select all

dpkg-reconfigure tzdata

Setting up WLAN from command line
Edit the file /etc/network/interfaces or create a new one under /etc/network/interfaces.d/ and add the following lines

Code: Select all

auto wlan0
iface wlan0 inet dhcp
    wpa-essid <NAME-OF-YOUR-WLAN>
    wpa-psk <WLAN-PASSWORD>

If you're using a "hidden" network, please also add wpa-scan-ssid 1

Webbrowser:
Both Chromium and Firefox should work fine with hardware accelerated browsing and WebGL support.

Kodi (X11):
Kodi will work "fine". You can use setup-odroid to install Kodi for X11 (should work under wayland as well).
It will not use hardware accelerated decoding, but will use GPU support and the CPU is fast enough to even show H265 1080p videos.

Known Issues:
Vulkan support will not be available MESA developers dropped Vulkan support for Midgard Mali GPU architecture.
If you run the image from SD card reboot is not working.

My backported MESA drivers currently support both OpenGL ES and OpenGL up to version 3.1

[AreaScout]

This is very good, now I can successfully test the vendor with the best Panfrost support

RG

[mad_ady]

meveric, whatever possesed you to create an image for an EOL board that has maybe 30 boards in the wild and is loved by all who have it?
Though I'm not a regular Debian user I feel compelled to steal the kernel and use it on my in-house Ubuntu. Could you share your kernel config so I can validate it has docker support and gpio-ir-tx?
Also - are you running mainline uboot?
Thanks!

[meveric]

meveric, whatever possesed you to create an image for an EOL board that has maybe 30 boards in the wild and is loved by all who have it?
Though I'm not a regular Debian user I feel compelled to steal the kernel and use it on my in-house Ubuntu.

As you said it's kinda a favorite of mine as well and I wanted to see how Panfrost performs on a SoC that is "fully" supported.
It was also good training to maybe get other boards up and running with mainline as well.
I learned a lot at least

Could you share your kernel config so I can validate it has docker support and gpio-ir-tx?
Also - are you running mainline uboot?
Thanks!

I only used the stuff that is in the Trying to get mainline kernel working on N1 thread.
Including the config for the Kernel and the U-Boot

[meveric]

I backported MESA 20.2.1 from Debian Sid to Debian Bullseye which can be installed from my repository.
Panfrost now has OpenGL ES 3.0 support as well as more OpenGL features in total.
Funny though, Wayland performance went down from over 1700 FPS tops to 1100 FPS
But X11 performance went up a little.

[brad]

Thanks @meveric will move my board over soon I'm getting sick of seeing this "Linux odroid 4.4.114-21 #1 SMP Mon Mar 19 01:05:12 UTC 2018 aarch64 GNU/Linux"

[rooted]

Thanks for doing this.

[meveric]

Update:
I've updated the base image with newer version of u-boot.
I also fixed SD card support which is now working (without UHS support at the moment it seems).

[rooted]

Don't use it anyway but it's nice to know it's working.

[meveric]

well for me it was annoying, I was using a 16GB eMMC originally as most of the software was on the network.
But since I now start with an entire new image I first have to test the software directly on the system and probably start over and over again several times.
And 16GB doesn't hold much data, but I still had a 128 GB SD laying around from my GO2 testing so that is something that's very handy for testing
But not being able to boot from is was really annoying

I wanna test some Windows games on it with wine and so on therefore I need quite a bit of space

[joerg]

@meveric: Thank you very much for this. I am on the way to rebuild my N1 NAS with two new harddisks and with your Debian image. I installed it on a 64GB Sandisk Extreme as I don't want yet overwrite the emmc that has a working system. So far it has booted, now I will continue to install OMV and Nextcloud, I will report after.

[edit]
It turns out that OMV5 needs php7.3 that is not installable on Bullseye. Currently I am compiling php7.3. Don't know if the decencies will be met to install OMV5...

[edit]
It stucks at

Code: Select all

The following packages have unmet dependencies:
 php-pam : Depends: phpapi-20180731 but it is not installable

I wonder how I can make this virtual package phpapi by myself. I haven't found so far...

[meveric]

php7.3 is Debian Buster, since this is bullseye it's a little ahead, although it's odd that it's not downward compatible.
Instead of compiling it yourself you can also try adding:
https://packages.sury.org/php/

Code: Select all

apt install wget
wget -O- https://packages.sury.org/php/apt.gpg | apt-key add -
echo "deb https://packages.sury.org/php/ buster main" > /etc/apt/sources.list.d/php.list
apt update

although it's for buster, it should be compatible with bullseye as well (I hope) and it has different versions of php available that you can install.
It will also give you security updates and fixes, which is not the case if you compile it yourself.

It should provide packages for php 5.6, 7.0. 7.1 7.2 7.3 7.4 and 8.0 so even if the requirements of OMV change you should be covered for some time

[joerg]

I will try it, thank you.
But later this week, because I installed back the emmc to have my NAS online.

[joerg]

@meveric: I was able to install php7.3 with the buster sources.

Then I installed OMV5 as is written here for buster: https://forum.openmediavault.org/index. ... 10-buster/
I had to disable apache2.
And I had to install an additional php package php7.3-xml as there was an error Uncaught Error: Class 'DOMDocument' not found

OMV5 now is coming up. Let's see if I can adjust it with all options i have on the old system.

[meveric]

glad to see it's coming together.. Hope it works out

[joerg]

It was as much work as I expected.
I had to export the Nextcloud database from the old system and import it to the new system, to have all the contacts and the calendar entries restored.
But I had to learn step by step to come to this. On the new system (Bullseye with OMV5) I had to use the old Nextcloud 15 first. I was not able to update Nextcloud from within the gui, the button was missing although I use php7.3. So I had to rename the whole Nextcloud folder and to unzip the next higher update to 16 and to restore the config and the data folder from 15 to 16. The same with 17. The same with 18. The same with 19....Stop! No, I wasn't able to fix a server error with 19. I must stay with Nextcloud 18.
It would be possible to install Nextcloud 20, but I would loose all as the Nextcloud database would be created new.
Then I uninstalled OMV5 by mistake and I had to learn how to reinstall it.
I had some trouble with apt-get update as the system time didn't synchronize after reboot. I learned that with OMV the package chrony is needed. But chrony service was masked. I enabled the service and put one additional line into the service:

Code: Select all

ExecStartPost=/usr/bin/chronyc -a makestep

This makes the system time synchronized after start up. (It seems to work )
Next, I had also trouble with OMV logon errors. I found at the end, that the openmediavault-engined.service was also masked and didn't start after reboot.
I unmasked and enabled it with:

Code: Select all

sudo systemctl unmask openmediavault-engined
sudo systemctl enable openmediavault-engined

BTW, @meveric, you know that the N1 does not make reboot? It does start after power cycle or with press PWR button. (Not so important for me as it should stay on 24/7)

[mad_ady]

Does it reboot correctly with reboot -f? My N1, on kernel 5.7.0 rc5 reboots correctly, but a while ago reboot was problematic. There might be some patch missing (in u-boot?)

[meveric]

Does it reboot correctly with reboot -f? My N1, on kernel 5.7.0 rc5 reboots correctly, but a while ago reboot was problematic. There might be some patch missing (in u-boot?)

For me, reboot from eMMC works fine, but reboot from SD always hangs and you need to press the reset button.

[mad_ady]

You may be right - I only use emmc with N1. So mine may behave the same way.

[joerg]

Funny, now the Updater button was present in Nextcloud and I could do the update from 18.0.4 to 18.0.11 to 19.0.5 to 20.0.2.
Sometimes it's good to wait...

[brad]

I will take a look at sd card, UHS mode and sd reboot issue once I get my new drives for the HC4 and migrate my data (hopefully by the end of the week).

I suspect the sd card reset line is not working so kernel cannot reset to UHS modes and also the reboot can not reset to standard mode for trusted firmware to find the initial bootloader code.

[joerg]

Thank you @brad. But do this work only if you find it important. For me it's not so important, as I hope that the N1 with Bullseye and mainline kernel runs as stable that I don't need to reboot for really long time.

[meveric]

Just to show off:
System details, to show it's running on the N1 using experimental OpenGL 3.3 backend for Panfrost:
https://oph.mdrjr.net/meveric/other/Vid ... tor2-1.mp4

In game graphics:
https://oph.mdrjr.net/meveric/other/Vid ... tor2-2.mp4

The game seems to run at full speed, I did not experience any lagging.
It's the x86 version of KotoR II which requires OpenGL 3.x to work.
In fact it even works with OpenGL 2.1 (default for Panfrost) but videos won't work.

Since this is an x86 game it runs on box86 from @ptitSeb

It's amazing, that is two out of three games that I want to run on ODROIDs
And the third game already work but is just too slow.

[mad_ady]

Impressive!

[mad_ady]

Ok, just to show everyone what a leach I can be, I tried to steal meveric's kernel and install it on my Ubuntu 20.04 running on N1 (was running kernel 5.8 previously). And I failed to make it boot - not sure exactly why.
First I copied the kernel, dtb, modules to the right places (for me the kernel lives in /media/boot/boot along with boot.scr). I even took meveric's original boot.scr and adapted it to my needs (looks like this after recompilation):

Code: Select all

$ strings boot.scr
# flash-kernel: bootscr.odroid-n1
##!/bin/bash
#mkimage -A arm64 -T script -O linux -d boot.txt boot.scr
# Bootscript using the new unified bootcmd handling
# Expects to be called with the following environment variables set:
#  devtype              e.g. mmc/scsi etc
#  devnum               The device number of the given type
#  bootpart             The partition containing the boot files
#  distro_bootpart      The partition containing the boot files
#                       (introduced in u-boot mainline 2016.01)
#  prefix               Prefix within the boot partiion to the boot files
#  kernel_addr_r        Address to load the kernel to
#  fdt_addr_r           Address to load the FDT to
#  ramdisk_addr_r       Address to load the initrd to.
# The uboot must support the booti and generic filesystem load commands.
#setenv bootargs " ${bootargs} root=UUID=a977a8c4-0492-48d9-9876-8f53ac858a41 rootwait ro"
setenv bootargs " earlyprintk root=/dev/mmcblk1p2 net.ifnames=0 biosdevname=0 fsck.mode=auto fsck.repair=yes mitigations=off rootwait ro console=ttyS2,115200n8 cma=800M clk_ignore_unused usb-storage.quirks=0x174c:0x1153:u,0x152d:0x0578:u"
setenv fk_kvers "5.10.0-rc7-arm64"
setenv fdtfile "rk3399-odroid-n1.dtb"
setenv prefix "/boot/"
setenv fdt_addr_r "0x20000000"
if test -z "${fk_kvers}"; then
   setenv fk_kvers '5.10.0-rc1-arm64'
if test -z "${distro_bootpart}"; then
  setenv partition ${bootpart}
else
  setenv partition ${distro_bootpart}
load ${devtype} ${devnum}:${partition} ${kernel_addr_r} ${prefix}vmlinuz-${fk_kvers}
load ${devtype} ${devnum}:${partition} ${fdt_addr_r} ${prefix}${fdtfile}
load ${devtype} ${devnum}:${partition} ${ramdisk_addr_r} ${prefix}initrd.img-${fk_kvers}
echo "Booting Linux ${fk_kvers} from ${devtype} ${devnum}:${partition}..."
booti ${kernel_addr_r} ${ramdisk_addr_r}:${filesize} ${fdt_addr_r}

The kernel, dtb and initrd are as follows:

Code: Select all

-rwxr-xr-x 1 root root  8759572 Dec 26 18:57 initrd.img-5.10.0-rc7-arm64
-rwxr-xr-x 1 root root    57639 Dec 26 17:31 rk3399-odroid-n1.dtb
-rwxr-xr-x 1 root root 11919230 Dec 11 15:37 vmlinuz-5.10.0-rc7-arm64

However, uboot complains that:

Code: Select all

Hit any key to stop autoboot:  0
switch to partitions #0, OK
mmc0(part 0) is current device
Scanning mmc 0:1...
Found U-Boot script /boot/boot.scr
1948 bytes read in 19 ms (99.6 KiB/s)
## Executing script at 00500000
syntax error
SCRIPT FAILED: continuing...

Not sure where the syntax error is... When I get my next manintenance window I'll try to execute each line manually to see what happens....

[meveric]

You if statement for partition is not closed.
Also I'm not sure if you use prefix correctly.
That one would try to find everything in a subfolder called boot

[mad_ady]

Thanks, I missed the ending if statement while .... borrowing your code.
The /boot subfolder is correct for my case.

[mad_ady]

Ok, I figured out why the "fi" statements were missing - I ran strings on your boot.scr and "fi" by itself was too short of a string and didn't get printed out. I should have used cat instead. Lesson learned!

I fixed boot.scr and it still failed to boot. Here are the boot commands ran manually and the error I get:

Code: Select all

odroidn1 => 
odroidn1 => setenv bootargs " earlyprintk root=/dev/mmcblk1p2 net.ifnames=0 biosdevname=0 fsck.mode=auto fsck.repair=yes mitigations=off rootwait ro console=ttyS2,115200n8 cma=800M clk_ignore_unused usb-storage.quirks=0x174c:0x1153:u,0x152d:0x0578:u"
odroidn1 => setenv fk_kvers "5.10.0-rc7-arm64"
odroidn1 => setenv fdtfile "rk3399-odroid-n1.dtb"
odroidn1 => setenv prefix "/boot/"
odroidn1 => setenv fdt_addr_r "0x20000000"
odroidn1 => if test -z "${fk_kvers}"; then
>    setenv fk_kvers '5.10.0-rc1-arm64'
> fi
odroidn1 => printenv fk_kvers
fk_kvers=5.10.0-rc7-arm64
odroidn1 => if test -z "${distro_bootpart}"; then
>   setenv partition ${bootpart}
> else
>   setenv partition ${distro_bootpart}
> fi
odroidn1 => 
odroidn1 => printenv partition
partition=1 
odroidn1 => printenv devtype
## Error: "devtype" not defined
odroidn1 => printenv devnum
## Error: "devnum" not defined
odroidn1 => printenv $devtype
## Error: "mmc" not defined
odroidn1 => printenv ${devtype}
## Error: "mmc" not defined
odroidn1 => printenv $devnum
## Error: "0" not defined
odroidn1 => printenv $kernel_addr_r
## Error: "0x02080000" not defined
odroidn1 => printenv $fdt_addr_r
## Error: "0x20000000" not defined
odroidn1 => printenv $ramdisk_addr_r
## Error: "0x06000000" not defined
odroidn1 => load ${devtype} ${devnum}:${partition} ${kernel_addr_r} ${prefix}vmlinuz-${fk_kvers}
11919230 bytes read in 301 ms (37.8 MiB/s)
odroidn1 => load ${devtype} ${devnum}:${partition} ${fdt_addr_r} ${prefix}${fdtfile}
57639 bytes read in 26 ms (2.1 MiB/s)
odroidn1 => load ${devtype} ${devnum}:${partition} ${ramdisk_addr_r} ${prefix}initrd.img-${fk_kvers}
8759572 bytes read in 245 ms (34.1 MiB/s)
odroidn1 => echo "Booting Linux ${fk_kvers} from ${devtype} ${devnum}:${partition}..."
Booting Linux 5.10.0-rc7-arm64 from mmc 0:1...
odroidn1 => echo $filesize
85a914
odroidn1 => booti ${kernel_addr_r} ${ramdisk_addr_r}:${filesize} ${fdt_addr_r}
kernel_comp_addr_r or kernel_comp_size is not provided!

I'm guessing the difference is in u-boot version (or compile options). My uboot is probably older than yours:

Code: Select all

U-Boot TPL 2020.07-rc3-00048-ga08e0a28eb-dirty (May 30 2020 - 21:42:06)
Channel 0: DDR3, 800MHz
BW=32 Col=10 Bk=8 CS0 Row=15 CS1 Row=15 CS=2 Die BW=16 Size=2048MB
Channel 1: DDR3, 800MHz
BW=32 Col=10 Bk=8 CS0 Row=15 CS1 Row=15 CS=2 Die BW=16 Size=2048MB
256B stride
256B stride
Trying to boot from BOOTROM
Returning to boot ROM...

U-Boot SPL 2020.07-rc3-00048-ga08e0a28eb-dirty (May 30 2020 - 21:42:06 -0400)
Trying to boot from MMC2
NOTICE:  BL31: v2.3(release):v2.3-107-gdd1eb34aa
NOTICE:  BL31: Built : 21:42:00, May 30 2020


U-Boot 2020.07-rc3-00048-ga08e0a28eb-dirty (May 30 2020 - 21:42:06 -0400)

Next I'll try to steal your uboot as well and see how that goes... I'll have to be careful not to nuke my partition table...

[meveric]

there was an error I did on the last kernel.. which apparently didn't matter with my uboot but may be an issue with yours.
The vmlinuz-5.10.0-rc7-arm64 is actually compressed but you want an uncompressed one.
change the name to vmlinuz-5.10.0-rc7-arm64.gz use gunzip to uncompress it and try again.
My next Kernel has this fixed again.

[mad_ady]

Thanks, I'll try that

[mad_ady]

Ok, this time - success! I can confirm that your 5.10.3-arm64 runs fine on my Ubuntu N1
I still had issues:
- emmc became mmcblk2 instead of mmcblk1 (and my rootfs was not mounted by UUID, fixed it on my side)
- the initrd - the one copied from your image fails to boot because it can't find :

Code: Select all

Begin: Running /scripts/local-block ... done.
Begin: Running /scripts/local-block ... done.
Begin: Running /scripts/local-block ... done.
done.
Gave up waiting for root file system device.  Common problems:
 - Boot args (cat /proc/cmdline)
   - Check rootdelay= (did the system wait long enough?)
 - Missing modules (cat /proc/modules; ls /dev)
ALERT!  /dev/disk/by-uuid/52662db4-d4fa-434f-b6ff-75eab420603b does not exist.  Dropping to a shell!


BusyBox v1.30.1 (Debian 1:1.30.1-6) built-in shell (ash)
Enter 'help' for a list of built-in commands.

(initramfs) cat /proc/cmdline
 earlyprintk root=UUID=a130ca78-0841-400a-8823-06a304a00850 net.ifnames=0 biosdevname=0 fsck.mode=auto fsck.repair=yes mitigations=off rootwait ro console=ttyS2,115200n8 cma=800M clk_ignore_unused usb-storage.quirks=0x174c:0x1153:u,0x152d:0x0578:u
(initramfs) blkid
/dev/mmcblk2p1: SEC_TYPE="msdos" LABEL_FATBOOT="boot" LABEL="boot" UUID="AAAA-BBBB" BLOCK_SIZE="512" TYPE="vfat" PARTUUID="630bb630-01"
/dev/mmcblk2p2: LABEL="rootfs" UUID="a130ca78-0841-400a-8823-06a304a00850" BLOCK_SIZE="4096" TYPE="ext4" PARTUUID="630bb630-02"
/dev/sdb1: UUID="585826ea-f84f-578c-5a00-8477c557061a" TYPE="linux_raid_member" PARTLABEL="primary1" PARTUUID="0f4e09a0-0b94-4e84-ad81-27bcdac76162"
/dev/sdb2: UUID="fe63ac7f-2a4b-5df6-5a00-8477c557061a" TYPE="linux_raid_member" PARTLABEL="primary2" PARTUUID="cffcdeda-63cd-4a0f-8dea-57ddd5871c4a"
/dev/sdb3: UUID="b0fd946f-37d1-aec0-5a00-8477c557061a" TYPE="linux_raid_member" PARTLABEL="primary3" PARTUUID="6e1322cf-f880-4714-ad2c-751cba6d711a"
/dev/sdb4: UUID="f722c7ae-496c-6306-bb26-17a5f4338677" UUID_SUB="c7eb2f4b-5781-47db-8be4-9e1c78a28d62" LABEL="MyBookWorld:2" TYPE="linux_raid_member" PARTLABEL="primary4" PARTUUID="e523e666-b50e-4e85-9c45-9f37b49909b2"
/dev/sda1: LABEL="ssd480_venus" UUID="a21c689b-dc6a-46cb-b650-f79d4bf46e9a" BLOCK_SIZE="4096" TYPE="ext4" PARTUUID="000ef0d0-01"
/dev/sdc1: LABEL="4TB" UUID="5c4d416f-5bce-46e5-b628-2f822cff6d59" BLOCK_SIZE="4096" TYPE="xfs" PARTLABEL="primary" PARTUUID="f605e7fe-df4c-4ddc-acd4-ea3a71b7f496"
(initramfs) ls -l /dev/disk/by-uuid/
total 0
lrwxrwxrwx    1 0        0               10 Jan  1 00:00 5c4d416f-5bce-46e5-b628-2f822cff6d59 -> ../../sdc1
lrwxrwxrwx    1 0        0               15 Jan  1 00:00 AAAA-BBBB -> ../../mmcblk2p1
lrwxrwxrwx    1 0        0               15 Jan  1 00:00 a130ca78-0841-400a-8823-06a304a00850 -> ../../mmcblk2p2
lrwxrwxrwx    1 0        0               10 Jan  1 00:00 a21c689b-dc6a-46cb-b650-f79d4bf46e9a -> ../../sda1

Presumably that's the uuid of your image's rootfs. So I ended up using my old initrd and the system booted just fine.
One thing I've noticed is the fan doesn't start. I haven't changed the trip points, but it seems to enter thermal throttling at 70C and fan doesn't start. I set the fan speed via dtb:

Code: Select all

sudo fdtput -t x rk3399-odroid-n1.dtb /pwm-fan cooling-levels 0x0 0x50 0x96 0xff

Docker and modules were all there, so thanks for your work!

[meveric]

I have issues with the fan myself. I actually put the powerlevels to 255 and even then it's just like a slight breeze.
I think the power regulator is not working correctly, but it's already at 5.5V I think kinda the highest that exists or else I'd have tried to use a higher power regulator for the fan.

[mad_ady]

Hmm, odd... It works on 5.8.0 (built from tobetter's branch (I think). Do you think the difference may be in the dtb? I'll check when I get a bit of time...

[meveric]

I used the dtb/dts that was in the other thread: viewtopic.php?f=155&t=33373&start=132

So haven't changed anything here, still I'm not sure.
If you find out what could be the problem, let me know, cause I'd really like to have that working as well.

[mad_ady]

You're right, it behaves the same way with the old 5.8 DTB. What 5.10 branch did you use to build your kernel? The official one (mainline), or some other one? I want to look for differences in PWM driver, see if I can spot something.

[meveric]

You're right, it behaves the same way with the old 5.8 DTB. What 5.10 branch did you use to build your kernel? The official one (mainline), or some other one? I want to look for differences in PWM driver, see if I can spot something.

The "Master" package (up to rc7) were directly from Linus.
The new 5.10.3 comes from the linux stable kernel tree: https://git.kernel.org/pub/scm/linux/ke ... /linux.git

[mad_ady]

@meveric - is the newest kernel you're building/using on N1 this one? https://oph.mdrjr.net/meveric/pool/n1/l ... rc7-arm64/ ? Or is it a different package name/location that hosts a more recent kernel?
Thanks!

Edit: Should I use this one instead?
https://oph.mdrjr.net/meveric/pool/main ... 64-odroid/

[meveric]

https://oph.mdrjr.net/meveric/pool/main ... -upstream/
The 5.17.5 is the current one.. the N1 specific Kernel I dropped after I confirmed that for arm64 one kernel is enough for all boards

If you use my repository:
https://oph.mdrjr.net/meveric/sources.l ... -main.list

you can install it via:

Code: Select all

apt install linux-image-arm64-odroid linux-headers-arm64-odroid

The one that you linked is the meta package that is installed for that.

[mad_ady]

Thanks!

Not sure it's safe to add your repo on top of my ubuntu, but I can install the kernel from deb instead...

[meveric]

should be fine the all/main package list should only have packages that are OS independent.. that's why I have different repos for wheezy, jessie, strech, buster, bullseye and of cause "all"
Kernel is a kernel independent packages... doesn't matter which OS. which is why it's in all/main.

GPU drivers (MESA) for example is under bullseye/backports, as it depends on the OS version and as it replaces original packages with packages from a different version (backport from Debian Sid or Experimental).

Also you can always put a option in your /etc/apt/preferences, which lowers the priority of that repository.
Which means it will never be installed automatically, and always require specific option via apt to install packages from that repository.

[mad_ady]

Ok, I tried to boot kernel 5.17, but my uboot had other plans:

Code: Select all

U-Boot TPL 2020.07-rc3-00048-ga08e0a28eb-dirty (May 30 2020 - 21:42:06)
Channel 0: DDR3, 800MHz
BW=32 Col=10 Bk=8 CS0 Row=15 CS1 Row=15 CS=2 Die BW=16 Size=2048MB
Channel 1: DDR3, 800MHz
BW=32 Col=10 Bk=8 CS0 Row=15 CS1 Row=15 CS=2 Die BW=16 Size=2048MB
256B stride
256B stride
Trying to boot from BOOTROM
Returning to boot ROM...

U-Boot SPL 2020.07-rc3-00048-ga08e0a28eb-dirty (May 30 2020 - 21:42:06 -0400)
Trying to boot from MMC2
NOTICE:  BL31: v2.3(release):v2.3-107-gdd1eb34aa
NOTICE:  BL31: Built : 21:42:00, May 30 2020


U-Boot 2020.07-rc3-00048-ga08e0a28eb-dirty (May 30 2020 - 21:42:06 -0400)

Model: Hardkernel ODROID-N1
DRAM:  3.9 GiB
PMIC:  RK808
MMC:   mmc@fe320000: 1, sdhci@fe330000: 0
Loading Environment from MMC... Card did not respond to voltage select!
*** Warning - No block device, using default environment

In:    serial
Out:   serial
Err:   serial
Model: Hardkernel ODROID-N1
Net:
Error: ethernet@fe300000 address not set.
No ethernet found.

Hit any key to stop autoboot:  0
switch to partitions #0, OK
mmc0(part 0) is current device
Scanning mmc 0:1...
Found U-Boot script /boot/boot.scr
1975 bytes read in 19 ms (100.6 KiB/s)
## Executing script at 00500000
14399368 bytes read in 395 ms (34.8 MiB/s)
57313 bytes read in 20 ms (2.7 MiB/s)
23072474 bytes read in 592 ms (37.2 MiB/s)
Booting Linux 5.17.5-arm64 from mmc 0:1...
kernel_comp_addr_r or kernel_comp_size is not provided!
SCRIPT FAILED: continuing...

My boot.scr is built from this:

Code: Select all

#
# flash-kernel: bootscr.odroid-n1
#
#
##!/bin/bash
#mkimage -A arm64 -T script -O linux -d boot.txt boot.scr
#
#

# Bootscript using the new unified bootcmd handling
#
# Expects to be called with the following environment variables set:
#
#  devtype              e.g. mmc/scsi etc
#  devnum               The device number of the given type
#  bootpart             The partition containing the boot files
#  distro_bootpart      The partition containing the boot files
#                       (introduced in u-boot mainline 2016.01)
#  prefix               Prefix within the boot partiion to the boot files
#  kernel_addr_r        Address to load the kernel to
#  fdt_addr_r           Address to load the FDT to
#  ramdisk_addr_r       Address to load the initrd to.
#
# The uboot must support the booti and generic filesystem load commands.

#setenv bootargs " ${bootargs} root=UUID=a977a8c4-0492-48d9-9876-8f53ac858a41 rootwait ro"
setenv bootargs " earlyprintk root=UUID=a130ca78-0841-400a-8823-06a304a00850 net.ifnames=0 biosdevname=0 fsck.mode=auto fsck.repair=yes mitigations=off rootwait ro console=ttyS2,115200n8 cma=800M clk_ignore_unused usb-storage.quirks=0x174c:0x1153:u,0x152d:0x0578:u"
setenv fk_kvers "5.17.5-arm64"
setenv fdtfile "rk3399-odroid-n1.dtb"
setenv prefix "/boot/"
setenv fdt_addr_r "0x20000000"
if test -z "${fk_kvers}"; then
   setenv fk_kvers '5.17.5-arm64'
fi

if test -z "${distro_bootpart}"; then
  setenv partition ${bootpart}
else
  setenv partition ${distro_bootpart}
fi

load ${devtype} ${devnum}:${partition} ${kernel_addr_r} ${prefix}vmlinuz-${fk_kvers}
load ${devtype} ${devnum}:${partition} ${fdt_addr_r} ${prefix}${fdtfile}
load ${devtype} ${devnum}:${partition} ${ramdisk_addr_r} ${prefix}initrd.img-${fk_kvers}
echo "Booting Linux ${fk_kvers} from ${devtype} ${devnum}:${partition}..."
booti ${kernel_addr_r} ${ramdisk_addr_r}:${filesize} ${fdt_addr_r}

You had promissed an uncompressed vmlinuz... . After it was ungzipped, it booted just fine, thanks!

Update: The fan seems to have been fixed. My Noctua fan is spinning nicely when running stress and the board doesn't go over 60C. Great! I will check other things as well (such as my ir-blaster), but it looks like it's solid!

[meveric]

my u-boot is slightly newer 2020.10 and already has configuration for compressed Kernel files.
I guess you can configure it manually:

Code: Select all

kernel_comp_addr_r=0x08000000
kernel_comp_size=0x2000000

sadly my fan is still not fixed, I even switched out the fan since I suspected the fan might actually be damaged, but guess that's not the case.

[DarkBahamut]

sadly my fan is still not fixed, I even switched out the fan since I suspected the fan might actually be damaged, but guess that's not the case.

I recall around 5.10 RC5 they made a change to SPI handling that stopped my N1 fan working too. At the time I made this change and It's seemed to work normally ever since

https://github.com/Emerald-Phoenix/odro ... 548107bb79

[mad_ady]

@DarkBahamut are you sure that patch affects fan? It's a change under SPI flash (which the N1 doesn't have on-board).

I've added the change to my dtb:

Code: Select all

fdtput rk3399-odroid-n1.dtb /spi@ff1d0000/w25q128@0 spi-cs-high

I'll give it a try later today, when I am allowed to make a scheduled downtime

[mad_ady]

Update - I made the change to the DTB and rebooted and yes, it seems to have made a difference. The fan seems to spin faster - at least I can hear it slowing down and starting up again - previously it wasn't noticeable. Also, temperature is 3-4 degrees lower under load. Most likely it hits 60C and uses a different fan speed and lowers the temperature quickly, starts to spin lower, but then the temperature rises and can't spin down.
Anyway, thank you for the fix!

@meveric - I've noticed one problem compared to my old 5.8 kernel - the eth0 NIC gets a random MAC address on boot. I couldn't see what causes it - no obvious e-fuse errors in dmesg:

Code: Select all

[    3.406317] Run /init as init process
[    3.410426]   with arguments:
[    3.410429]     /init
[    3.410431]     earlyprintk
[    3.410433]   with environment:
[    3.410435]     HOME=/
[    3.410437]     TERM=linux
[    3.410439]     biosdevname=0
[    3.740633] rk_gmac-dwmac fe300000.ethernet: IRQ eth_wake_irq not found
[    3.748110] rk_gmac-dwmac fe300000.ethernet: IRQ eth_lpi not found
[    3.755224] rk_gmac-dwmac fe300000.ethernet: PTP uses main clock
[    3.762169] rk_gmac-dwmac fe300000.ethernet: clock input or output? (input).
[    3.770122] rk_gmac-dwmac fe300000.ethernet: TX delay(0x28).
[    3.776476] rk_gmac-dwmac fe300000.ethernet: RX delay(0x11).
[    3.782878] rk_gmac-dwmac fe300000.ethernet: integrated PHY? (no).
[    3.790306] rk_gmac-dwmac fe300000.ethernet: cannot get clock clk_mac_speed
[    3.800989] rk_gmac-dwmac fe300000.ethernet: clock input from PHY
[    3.813261] rk_gmac-dwmac fe300000.ethernet: init for RGMII
[    3.823311] rk_gmac-dwmac fe300000.ethernet: User ID: 0x10, Synopsys ID: 0x35
[    3.831382] rk_gmac-dwmac fe300000.ethernet:         DWMAC1000
[    3.837278] rk_gmac-dwmac fe300000.ethernet: DMA HW capability register supported
[    3.845705] rk_gmac-dwmac fe300000.ethernet: RX Checksum Offload Engine supported
[    3.854130] rk_gmac-dwmac fe300000.ethernet: COE Type 2
[    3.860017] rk_gmac-dwmac fe300000.ethernet: TX Checksum insertion supported
[    3.863060] rockchip-vop ff8f0000.vop: Adding to iommu group 3
[    3.867941] rk_gmac-dwmac fe300000.ethernet: Wake-Up On Lan supported
[    3.879903] panfrost ff9a0000.gpu: clock rate = 500000000
[    3.881931] rk_gmac-dwmac fe300000.ethernet: Normal descriptors
[    3.894366] rk_gmac-dwmac fe300000.ethernet: Ring mode enabled
[    3.896526] rk808-rtc rk808-rtc: registered as rtc0
[    3.900897] rk_gmac-dwmac fe300000.ethernet: Enable RX Mitigation via HW Watchdog Timer
[    3.900906] rk_gmac-dwmac fe300000.ethernet: device MAC address 12:bb:7d:ee:67:02

I wonder if it happens for you too, or if it's caused by using a slightly older u-boot.
The problem is the box gets a different link-local IPv6 address, and I advertise the old one as my DNS server. I guess time to either force a MAC on boot, or set up Unique local IPv6 addresses instead (https://datatracker.ietf.org/doc/html/rfc4193).

[meveric]

First thanks for the DTB patch I'm gonna try this later today as well.

And yes the change with the MAC I noticed as well.
That's why I have this as an option in my config.ini to set a fixed IP.

[meveric]

I recall around 5.10 RC5 they made a change to SPI handling that stopped my N1 fan working too. At the time I made this change and It's seemed to work normally ever since

https://github.com/Emerald-Phoenix/odro ... 548107bb79

Nice git repo.. I probably gonna take over some of your changes in my Kernel build
Thanks for keeping the N1 alive

[meveric]

Update:
Updated the base image to my unified Debian-Bullseye arm64 image with Kernel 6.1.0
U-Boot has been update to version 2022.10 and overlay support is activated (not sure if it can be used for anything good yet).