ScottP

I ran into this issue a while back, I just downgraded back to the 5.10 kernel as it was a production server and I didn't have time to troubleshoot it. Are the current builds in the repo no fixed? I want to upgrade again to get the newer NPU driver to try out some LLMs on it Edit: I installed the latest image: linux-image-vendor-rk35xx 24.11.2 and it appears that this issue is resolved. Many thanks!

The other issue with the vendor kernels is that the DTB seems to contain every rockchip dtb known - hundreds of them not just rk35xx - the default 256MB boot partition fills up as there is no room to have two kernel/overlay versions

Same - this kernel seems to limit number of processes - fork failed, resource temporarily unavailable. Show your user limit: ulimit -u Show what limits other processes has: grep process /proc/*/limits

I installed this kernel on a Rock 5B. Although the above fixed the issue for the original poster the failure of configure stage is not addressed. I think there is something wrong with this kernel. On my Home Assistant server some containers were failing to start and the errors pointed towards some resource/limits issue even for processes running as root. The errors were something like - fork failed, resource temporarily not available. I think update-initramfs ran into this issue producing the errors in the post. I don't have any logs since it has hard to do anything at this point, commands failing when trying to diagnose. I rolled back to 5.10 manually to get my server up and running. I think the maximum number of processes was being limited artificially low despite nothing being in /etc/security/limits* - I did see for not root users this limited to around 2k processes via ulimit -u - cant remember the exact number. If anyone is running this kernel you can check the limits for various processes - user and root alike by looking at /proc/<PID>/limits On my rolled back 5.10 legacy kernel this is 61790

I Enabled overlay for USB0. Had to comment out g_serial too otherwise get hangs and usb disconnects. Odd decision to make the only USB port this board has into client mode instead of host so g_serial can be used as another console?

Seems like a bug, this is an old post so maybe if anyone has the same issue, do what I did: I got same error so I created two partitions for boot and root. Of course /boot was not mounted and used at all so it ended up as part of root and my board would not boot. I fixed this by mounting it and copying the files myself.

I using an intel ax210 on armbian with 6.1.x kernel. Armbian did not have firmware for the card, i added them manually. Whether intel cards do interrupts differently idk uname -r 6.1.7-rk35xx scott@3dp:/boot/dtb/rockchip/overlay$ dmesg|grep iwl [ 14.462615] iwlwifi 0000:01:00.0: enabling device (0000 -> 0002) [ 14.480578] iwlwifi 0000:01:00.0: api flags index 2 larger than supported by driver [ 14.480668] iwlwifi 0000:01:00.0: TLV_FW_FSEQ_VERSION: FSEQ Version: 0.0.2.36 [ 14.482525] iwlwifi 0000:01:00.0: loaded firmware version 72.daa05125.0 ty-a0-gf-a0-72.ucode op_mode iwlmvm [ 14.648881] iwlwifi 0000:01:00.0: Detected Intel(R) Wi-Fi 6 AX210 160MHz, REV=0x420 [ 14.845179] iwlwifi 0000:01:00.0: loaded PNVM version 64acdc51 [ 14.861028] iwlwifi 0000:01:00.0: Detected RF GF, rfid=0x10d000 [ 14.930488] iwlwifi 0000:01:00.0: base HW address: 3c:e9:f7:5f:84:da [ 15.199855] iwlwifi 0000:01:00.0 wlp1s0: renamed from wlan0

Is the only difference device tree? Could I use h68k image on rock 3a but replace the dtb before booting it? If so get dtb from one of radxa images of 5.10?

You have too many zeros on the speed its 1500000 ie 1.5Mbps

I just ran apt upgrade and ended up with 5.19.5 - Can it be taken out of the repo if its known to be broken? My root filesystem is now read only - so there are other issues than USB. Thanks