fraz0815 Posted October 7, 2020 Posted October 7, 2020 hey, this change https://github.com/armbian/build/pull/2010/files made my rockpi-4b (with 4GB ram) finally stable some time ago DDR_BLOB='rk33/rk3399_ddr_933MHz_v1.20.bin' #instead of 1.24 After messing with rx/tx offloading, this was the solution, no more failed apt and download errors, great! It seems that this change got lost in the transition to the new u-boot / different logical structure or is not needed anymore for the 1GB models it was mentioned originally, see here: https://github.com/armbian/build/commit/f2db96252530d2f5585755b518e7fd020b5f0392#diff-18c1f509b638f24dc14854e6af5173bb Any chance to get it back for rockpi-4b without creating trouble, I am not sure where this change should go: /build/config/sources/families/rk3399.conf or build/config/sources/families/include/rockchip64_common.inc ? Thanks
piter75 Posted October 7, 2020 Posted October 7, 2020 1 hour ago, fraz0815 said: Any chance to get it back for rockpi-4b without creating trouble, I am not sure where this change should go Did you verify if your troubles have returned with the mentioned commit? The commit is switching boot loading for RockPi4 (and RockPro64) to a different process - it is using mainline's u-boot TPL/SPL combination instead of binary DDR blob from Rockchip. That's why the DDR_BLOB reference is gone.
fraz0815 Posted October 8, 2020 Author Posted October 8, 2020 Ok thanks for clarification, I did not really understand whats going on now, or why sometimes BLOBS are used or what TPL/SPL does instead. But I will post detailed tomorrow what I did and its results, spoiler so far: downloaded (and also updated) Armbian_20.08.1_Rockpi-4b_buster_current_5.8.6_desktop.img is not able to run simple 'apt update' without errors, or download a larger file with correct checksum., e.g. 'wget someubuntu.iso'. It is exactly the same behavior prior first commit when it was changed to 1.20, which made the board nearly useless if using 1GB ethernet. 100MB helps, playing with offloading may help a bit, but it never gets stable. Right next to the rockpi-4b (was not running last months) is a nanopc-t4 running 24/7 perfectly fine. Compiling and logs follow...
piter75 Posted October 8, 2020 Posted October 8, 2020 6 hours ago, fraz0815 said: so far: downloaded (and also updated) Armbian_20.08.1_Rockpi-4b_buster_current_5.8.6_desktop.img is not able to run simple 'apt update' without errors, or download a larger file with correct checksum., e.g. 'wget someubuntu.iso'. This is most probably not caused by the commit you have mentioned as this commit is not part of any released artefacts yet (neither images not debs). Please provide some more details about the behaviour (link to armbianmonitor -u results would also help).
fraz0815 Posted October 9, 2020 Author Posted October 9, 2020 Ok, it looked like an obvious correlation to me, sorry - it was not. Long story short: I don't know what it was, sometimes the rockpi-4b gives me headaches. Fresh install of Armbian_20.08.1_Rockpi-4b_buster_current_5.8.6_desktop.img (ckecksum checked) to sdcard with etcher (which also validates) results in: To double check I also disconnected emmc and nvme extension (even though they were not used or mounted for anything) armbianmonitor -u : http://ix.io/2AeJ and it runs fine - should have done this earlier -.- Let's see when the errors come back, plugged back in the emmc and installed boot+system to it via armbian-config, verything running fine: http://ix.io/2AeP Now the nvme: http://ix.io/2AeT seems fine, no checksum errors or whatsoever. Thanks for your time, my best guess is either emmc module gets corrupted over time with no power attached or the nvme extension made trouble, or using sd+emmc+nvme together at the same time is a bad idea. Next time this happens, I will try to better investigate instead of guessing about some commits.
piter75 Posted October 9, 2020 Posted October 9, 2020 I am glad it works! 1 hour ago, fraz0815 said: or the nvme extension made trouble, or using sd+emmc+nvme together at the same time is a bad idea What's the power supply you are using? Is it a PD one? This combination (accompanied by cpu load) is most probably too much for most of 5V dumb usb power supplies.
fraz0815 Posted October 11, 2020 Author Posted October 11, 2020 It is a RockPi labeled power supply with output listed as 5V@3A / 9V@2A /12V@1.5A. I switched it always to some other PSU's laying around when having difficulties, with no success. I will see what happens this time and report back eventually (and keep it powered for now).
piter75 Posted October 11, 2020 Posted October 11, 2020 14 minutes ago, fraz0815 said: It is a RockPi labeled power supply with output listed as 5V@3A / 9V@2A /12V@1.5A. This should be good enough. It's not a dumb 5V one.
Recommended Posts