Adding the edge kernel to Rock 2F/2A

[humanus]

Hello,

 

These boards (Radxa Rock 2F and Rock 2A) have standard support so there should probably be a forum section for them. In its absence, I created this topic in the section for the most "similar" board.

 

Currently, we only have the `vendor` kernel available for both. It seems that both of them have been mainlined in 6.18, as seen here for example: https://www.cnx-software.com/2025/12/01/linux-6-18-release-main-changes-arm-risc-v-and-mips-architectures/ 

 

The config file for the board currently looks like this:

 

# Rockchip RK3528 quad core 1-4GB SoC WIFI/BT 0-32GB eMMC
BOARD_NAME="ROCK 2F"
BOARDFAMILY="rk35xx"
BOOTCONFIG="rock-2-rk3528_defconfig"
BOARD_MAINTAINER="CodeChenL"
KERNEL_TARGET="vendor"
KERNEL_TEST_TARGET="vendor"
FULL_DESKTOP="yes"
BOOT_LOGO="desktop"
BOOT_FDT_FILE="rockchip/rk3528-rock-2f.dtb"
BOOT_SCENARIO="spl-blobs"
IMAGE_PARTITION_TABLE="gpt"
enable_extension "radxa-aic8800"
AIC8800_TYPE="usb"

 

I tried just adding "edge" in the KERNEL_TARGET line and building an image but this won't be enough to make it boot, will it? Any tutorials on how to get them to work on mainline?

[Werner]

moved to staging

[Igor]

7 hours ago, humanus said:

but this won't be enough to make it boot, will it?

You can try, worse case it won't boot. What's the problem with that ?  

 

7 hours ago, humanus said:

Any tutorials

 

Only general.

 

Switch to nightly repository and then switch to edge kernel:
https://docs.armbian.com/User-Guide_Armbian-Config/System/#rolling

https://docs.armbian.com/User-Guide_Armbian-Config/System/#alternative-kernels

but if it won't work, you are on your own to find out why it doesn't work. Also some features won't work. We never supported EDGE kernels due to extreme costs associated with such operation.

[humanus]

There's no problem of course, I just did not have time for testing the built image yet, and was asking if it's not pointless. Thanks for the response so far, I'll try... 

As far as I understand it, the device tree from the vendor kernel won't work on mainline and I have no clue how to point armbian to the proper one. Maybe I'll just try removing the line about fdtfile from armbianEnv.txt lol

Edited December 3 by humanus
Clarification

[Werner]

IIRC stuff for rock2a/f was sent to upstream uboot too. Try adding a condition to use mainline uboot as well. You can check various rockchip based boards which also have this in their board config file.

[humanus]

I'll try. So far, just booting the board did not work. I suppose it does need mainline uboot. Need to figure out this fdt stuff somehow too... 

[humanus]

With the help of Claude I generated a config with both. Compiling now. Hopefully it boots this time. If it does, I may then do a PR.

[humanus]

The kernel seems to be starting but there is no further output from the serial console. Here are some attemps (forst one with FDT commented out, which prevented the kernel from loading, the last ones also, maybe due to SD card corruption). The green light is on and the blue light is pulsing.

This is the config it came up with with my guidance:

# Rockchip RK3528 quad core 1-4GB SoC WIFI/BT 0-32GB eMMC
BOARD_NAME="ROCK 2F"
BOARD_VENDOR="radxa"
BOARDFAMILY="rk35xx"
BOOTCONFIG="rock-2-rk3528_defconfig"
BOARD_MAINTAINER="CodeChenL"
KERNEL_TARGET="vendor,edge"
KERNEL_TEST_TARGET="vendor"
FULL_DESKTOP="yes"
BOOT_LOGO="desktop"
BOOT_FDT_FILE="rockchip/rk3528-rock-2f.dtb"
BOOT_SCENARIO="spl-blobs"
IMAGE_PARTITION_TABLE="gpt"
enable_extension "radxa-aic8800"
AIC8800_TYPE="usb"

function post_family_config__rock-2f_use_mainline_uboot() {
	# use mainline uboot for _edge_
	[[ "${BRANCH}" == "vendor" ]] && return 0

	display_alert "$BOARD" "Mainline U-Boot overrides for $BOARD - $BRANCH" "info"
	declare -g BOOTCONFIG="generic-rk3528_defconfig"  # Use generic RK3528 config since rock-2f specific config doesn't exist yet in mainline
	declare -g BOOTDELAY=1
	declare -g BOOTSOURCE="https://github.com/u-boot/u-boot"
	declare -g BOOTBRANCH="tag:v2025.10"
	declare -g BOOTPATCHDIR="v2025.10"
	declare -g BOOTDIR="u-boot-${BOARD}"
	declare -g UBOOT_TARGET_MAP="BL31=${RKBIN_DIR}/${BL31_BLOB} ROCKCHIP_TPL=${RKBIN_DIR}/${DDR_BLOB};;u-boot-rockchip.bin"
	unset uboot_custom_postprocess write_uboot_platform write_uboot_platform_mtd # disable stuff from rockchip64_common; we're using binman here which does all the work already

	# Just use the binman-provided u-boot-rockchip.bin, which is ready-to-go
	function write_uboot_platform() {
		dd "if=$1/u-boot-rockchip.bin" "of=$2" bs=32k seek=1 conv=notrunc status=none
	}
}

Here is the log: pastebin
 

Any ideas? I observed that with some boards, the line "BOOT_FDT_FILE=..." is absent altogether. is removing it possibly the way to go?

[Werner]

Quote

Could not find a valid device tree

Make sure the file is there and name is correct

[humanus]

Yup, there are numerous boot attempts if you take a look. There are two that are successful but not much happens afterwards, look for "Starting kernel". So when it finds the tree, it "boots" but there is no further output.

Edited yesterday at 05:43 PM by humanus
typo

[Werner]

Set verbosity to 7 in armbianEnv.txt to get an idea about what happens after "Starting kernel..."

[humanus]

Weird. I set verbosity to 7 but still nothing.

 

See this https://paste.armbian.eu/zubunigowi.vbnet

 

Nothing afterwards.

 

Claude suggested adding the parameter 

SERIALCON="ttyS0"

because apparently the default is ttyS2, which does not match rk3528. Let's see if it's just a hallucination.
EDIT: nope. Still no luck.

 

EDIT2. Something is wrong with the serial console on the vendor image too. I cannot type in it. It is connected correctly.

Edited yesterday at 11:12 PM by humanus
update

[Werner]

Wrong serial console set in kernel parameter could be a reason for having not output afterwards, so the suggestion isnt completely off. Though no clue what might be correct here.

[humanus]

What is happening with the board maintainer in general? I think they have been incommunicado pretty much since the first release and now is the time to do something (especially since the board has been "upgraded" now). 

 

One more detail. The filesystem does get expanded to fill the whole SD. Which part of the booting process does this happen at? 

Is there any way to set journal or dmesg logging to a file while not being able to log in (ie by editing a config file, or maybe crontab but externally?) 

Edited 4 hours ago by humanus
Typo/update