August 5, 2020

Building Armbian from sources

What I did, was to use armbian-config on my target board to download the sources. Then (on the target board, mind you): 
cd /usr/src/linux-source-5.4.43-sunxi64
make
And away it goes! The config has already been done appropriately (as near as I can tell). I did run "make config", which asks you about 1 million questions, but once I concluded that all the defaults looked right and the existing config was correct, I interrupted it and simply typed make.

Make is running on one core, which although slower than it would be using all cores, is not giving me any thermal issues. I am using "armbianmonitor -m" to keep an eye on CPU temperature and it sits between 40 and 50 C (104 - 122F). I have a heatsink on the chip and it never even feels warm. I am reading that Armbian does thermal throttling to keep the temperature in an acceptable range. However, armbianmonitor always tells me the CPU is running at 1368 Mhz.

After several hours (6 or so), I returned to my workshop to check on the build and the ssh connection had announced a broken pipe. The serial console was unresponsive. Pings failed. The system was clearly hung, so the power switch was the thing.

I used the serial console to log back in. My suspicion was that the disk had filled, but that was not the case: 

root@orangepipc2:~# df
Filesystem     1K-blocks    Used Available Use% Mounted on
udev              434060       0    434060   0% /dev
tmpfs             101172    3100     98072   4% /run
/dev/mmcblk0p1  15094960 5260328   9623980  36% /
tmpfs             505848       0    505848   0% /dev/shm
tmpfs               5120       4      5116   1% /run/lock
tmpfs             505848       0    505848   0% /sys/fs/cgroup
tmpfs             505848      12    505836   1% /tmp
/dev/zram0         49584    1816     44184   4% /var/log
tmpfs             101168       8    101160   1% /run/user/1000
tmpfs             101168       0    101168   0% /run/user/0
So, I typed "make" to resume the build. There is always the danger of a corrupt object file for the file that was being compiled when the system locked up. I will have to keep an eye out for that, and a clue will be the file and directory where the build resumes. The first file it decides to compile is: 
  CC [M]  drivers/net/wireless/rtl8811cu/hal/hal_halmac.o
I had meant to cut and paste the build history, but I foolishly closed the window with the ssh session, so that information is lost.

Here is the log, cut and pasted from the second half of the build:

All in all, this took on the order of 10 hours. I can't be sure exactly. I started it around 1:00 in the afternoon. It locked up sometime between 6:00 and 9:00. I restarted it and went to bed. It was done in the morning (and no more lockup).

Have any comments? Questions? Drop me a line!

Tom's electronics pages / tom@mmto.org