djurny

Testing on Nanopi R2S revealed that U-Boot there does not have setexpr built-in. I expect it will be the same on the 'Rock64' board you have? To make sure - as I do not have your hardware here - can you check if the following commands work on your U-Boot monitor commandline: setenv b setenv c setenv a "a/b" setexpr b sub "a/" "" a echo ${b} setexpr c 1 + 1 echo ${c} fdt Also, to give a workaround (hopefully), the output of the following on the U-Boot monitor commandline: echo ${fdtfile} echo ${fdt_addr_r} echo ${kernel_addr_r} echo ${ramdisk_addr_r} ver Thx, Groetjes,

Hi there, Ok, after some walking back and forth, typos, reboots, rewriting entire images, multiple armbianEnv.txt files, looking into U-Boot sources, googling (of course) I managed to find some things and got U-Boot on my Helios64 to boot properly again. In short, the aarch64 requirements to the linux kernel, the way U-Boot sometimes will move the loaded kernel image in memory might overwrite the starting point of the inital ramdisk. The load addresses for kernel and initial ramdisk are mostly hardcoded in U-Boot but can be overridden by armbianEnv.txt or the bootscript. As Helios64 uses Rockchip rk3399 from the aarch64 family, the attached boot.scr might also work for you. This bootscript will only work on aarch64 systems, as it assumes the kernel is uncompressed! If you test this, make sure to make a backup of your existing boot.scr or read up on how to regenerate the boot.scr from your boot.cmd. Make sure you have serial console available Copy the attached bootscript to /boot Add verbosity=1 (if not already there) to your armbianEnv.txt Reboot Would be great that if you test, to share the serial console output here to see it it also works for you. I will also test and verify on one of my Nanopi R2Ss somewhere today or tomorrow. Will put all the details into a pull request. Groetjes, Update: Removed boot.scr as testing on Nanopi R2S showed some misbehavior.

Hi there, I'm seeing the same while trying to update some U-Boot bootscripts on my Helios64 (rockchip aarch64). Seems there needs to be more space between $kernel_load_addr_r and $ramdisk_addr_r. How much is a bit puzzling, as for other archs it's usually enough to count the kernel's image size (and some alignment). If i come across something interesting, I will share here too. Groetjes,

I also have some oom issues when doing an aptitude upgrade in my opizero w/256 MiB ram. Guess even more on a 128MiB board like yours. Best to add a swap file to make sure upgrades go without issue. Grt,

Little odd that it fails to load the fix up script there; In the old boot.cmd it checks if it can find the file and loads it if it can. Can you share a serial console log of the 21.x release that did work? Incl aembianmonitor -u output? Thx, Grt,

Hi @Zaf9670, Your situation actually helped find an OBOE in the U-Boot image overlap test, so thank you for that! Can you try with an older image and share the U-Boot output? Also, can you add the following in your armbianEnv.txt: verbosity=8 earlycon=on console=serial I've had better experience with setting console to serial only and earlycon might help the kernel spew out some things earlier on. Groetjes,

Hi @Ed van den Enden, I'm still pressing "X" to doubt that dtoverlay=... will work for sunxi boards running Armbian like the Banana Pi M2 zero, as your points do not work on my Orange Pi zero (which is running sunxi Allwinner H2+ just like the Bananana Pi M2 zero). Also, your provided logging clearly indicates that the kernel did not configure any additional RTC, just the H2+ built-in one. But then again, I don't mind being wrong every now and then in life. Can you share the output of the following commands: ls -Failh /dev/rtc* ls -Failh /sys/class/rtc/ egrep -- . /sys/class/rtc/*/name dmesg | egrep -i -- 'rtc' And to read the actual clocks: sudo hwclock --rtc=/dev/rtc0 # per default this is the H2+ built-in RTC sudo hwclock --rtc=/dev/rtc1 # per default this is going to be the I2C RTC Posting the link to @IgorS's topic that shows exactly what to do here again: Also, for reference, here is a link to the Allwinner H2+ datasheet, where it also mentions clearly (chapter 4.8 RTC on page 171) that the SoC itself also has an on-board RTC (this is the RTC detected by the kernel as sun6i-rtc and the one you actually are reading in your case): https://wiki.friendlyelec.com/wiki/images/0/08/Allwinner_H2+_Datasheet_V1.2.pdf Groetjes,

Hi @Zaf9670, Are you sure you did the setenv fdt_addr_r 41000000 U-Boot claims DT is still loaded/applied to the original load address: Working FDT set to 41900000 Can you share your armbianEnv.txt and boot.cmd? I'll have a look at the defailt bootscript in the meantime, perhaps some parts of the bootscript is using fdt_addr_r and other parts might still use loadaddr or scriptaddr... Thx, Gr,

Hi @Zaf9670, Can you try to reset the U-Boot monitor, press any key to stop the autobooting, then enter the following on the commandline: setenv fdt_addr_r 41000000 setenv kernel_addr_r 41080000 setenv ramdisk_addr_r 42080000 These load addresses are based on the sizes of the DT (438272 bytes) and the kernel image (11013984 bytes). Then to kick off the boot: run bootcmd Do share the results please. Groetjes, p.s. That device tree is huge! 428 KiB??

Hi there, Now two pull requests are awaiting review: - https://github.com/armbian/build/pull/8166#issuecomment-2867147049 Update the boot.scr script to calculate load addresses in case `setexpr` is available on the U-Boot monitor. - https://github.com/armbian/build/pull/8170#issuecomment-2867915659 Enable the `setexpr` command on the U-Boot monitor, to unlock load address calculation in combination with the boot.scr update. Both are now tested OK using a built armbian 'minimal' image based on Bookworm. Groetjes,

Hi @Igor, all, Made some updates, see here (including some minor corrections). Tested it with a built image (bookworm-minimal) and it seems to work: I will see how i can change U-Boot configuration to allow for `setexpr` to work, as using `setexpr` will make sure this does not need constant changing whenever kernel grows in size. As this is my first pull request, not sure what to do next? Please have a look-see, Groetjes,

Hi @Zzzz, @Zaf9670, After it fails to boot, can you type the following in the U-Boot monitor prompt and share the ouput: echo $fdt_addr_r echo $kernel_addr_r echo $ramdisk_addr_r setexpr abc 1 + 1 echo "abc=x${abc}x" fdt header fdt header get def totalsize echo "def=x${def}x" Thanks, Groetjes,

Hi @Vaclav Rada, You can also disable the systemd 'predictable naming convention' of interface naming by adding: net.ifnames=0 to your kernel commandline. This can be done in armbianEnv.txt by adding them to the extraargs variable as follows: extraargs=net.ifnames=0 Would be better to adjust your armbianEnv.txt instead of changing on-disk systemd service unit files, as they will be overwritten whenever an update comes along. I have not seen a situation myself where using the 'new' naming convention for network interfaces would work out better. But perhaps my use cases are not as flashy as other's 🙂 Groetjes,

Hi @Igor, Seems that U-Boot 2019.04-armbian-2019.04-S3c99-P6351-H9530-V0854-Bb703-R448a does not have the `setexpr` command configured. I will do more testing and figure out how the change can handle `setexpr` missing and configuring U-Boot to have `setexpr` available. Thanks! Groetjes,

Hi @Heisath, I think i can make a pull request from here: https://github.com/armbian/build/compare/main...djurny:build:patch-1, but would like to see someone else besides me test and scrutinize change. If positive results are in, I would very happy to finally contribute something! Groetjes,