Tomash

Members
  • Content Count

    9
  • Joined

  • Last visited

  1. The problem: I want to run my Banana Pi M2 Zero from a battery / powerbank setup. It works like a charm from a beefy powerbank capable of consistent 2A output [1], even with added 5" LCD screen to power at the same time and wireless keyboard transmitter. BUT it does hang up hard, I guess due to undervoltage, when powered from a smaller, weaker powerbank with a single 18650 cell and max output of about 500mA (theoretically 1A but let's underestimate, I'll try to measure it with an oscilloscope). Armbian with 4.19 kernel and Ubuntu 18.04 (but booting in text mode, without any GUI). The hangup occurs during services startup, usually around networking ones (quite often the last line is about starting OpenSSH server), so systemd is already on. And here's my question: can I throttle CPU during boot time, to curb down on power spikes and keep the current requirements without measureable limits, to prevent undervoltage? I know that there are no undervoltage-mitigating mechanisms in Armbian for Allwinner SoC's. Also: what's the best way to measure momentary current consumption, with an oscilloscope? Via voltage drop on shunt resistor having 0.1 Ohm? [1]
  2. Thanks, here it is: https://github.com/armbian/build/pull/1503 Although it's a bit premature, I haven't tested DEV build yet.
  3. Spot on, disabling this patch (by touch-ing an empty file in userpatches) fixed the issue, I've got cpufreq information and the frequency does change. stress results in dropping the frequency to 800Mhz when the temperature reaches 80 C, perfect. At least with NEXT, I'll build and test with DEV kernel soon. Should my entire PR be just a removal of this patch file altogether?
  4. It's the same in 4.19. /sys/devices/system/cpu/cpufreq/ is an empty directory. How can I help in debugging and fixing it?
  5. Good tip, thanks, I'll go ahead with 4.19 for now.
  6. Build on new kernel (5.1), Ubuntu 18.04. No cpufreq information (0 MHz in htop etc.) also means there's no working frequency scaling. Which for Allwinner H2+ (Banana Pi M2 Zero) means quick hard-hang due to overheating whenever I compile anything non trivial, unless I point a USB Fan on the CPU (which already got a heatsink). Is there any chance of getting cpufreq and dvfs to work?
  7. Turns our dr_mode was "otg", changed it to host and usb peripherals work now, great!
  8. Can I only disable it at image build time, or can I remove this module from a working image and have a working OTG? Thanks for the CSC. I'm never again buying any SBC from Banana brand, though
  9. After some frustration with poor support for Banana Pi M2 Zero board by the manufacturer and not a very active community on Banana Pi forums, I stumbled upon Armbian and found out there's a CSC for building Armbian for this board. And because I'm having issues with wifi on the images found on Banana forums, with all images being old and based on kernel 3.4 -- wifi disconnects with some weird firmware errors in dmesg after about 3 minutes of running, requiring reboot to get back up again -- I was happy to read Armbian runs fine on BPI-M2Z for some folks here. So I did the build process, twice, so that I have two images to test, one with kernel 4.19 and one with 5.1, both running Ubuntu 18.04 as I'm more familiar with this than with Debian. It runs good. Great work, I'm really in awe. There is one issue, though: no usb device (and I'd love at least a keyboard) connected to "OTG" microusb port on the board works. The hardware stack is fine since on Raspbian 9.4 with 3.4 kernel I have it working with USB keyboard, mice, usb hubs, even a wifi dongle is detected and I can modprobe a required module. But in Armbian, regardless of which of the two kernels I try, there's no trace in dmesg that any USB device has been connected (the board does report as UART/serial device when I use same USB port to connect it with my Ubuntu PC). lsusb shows three hubs but nothing beyond that. I read Vincent's blogpost [1]. I checked dr_mode in dtb/dts file and it's "host", as I guess it should be. Can't compare it with the usb-working Raspbian since it has no dtb files in /boot. [1] https://vjordan.info/log/fpga/nanopi-neo-usb-otg-with-armbian-mainline-4x-kernel.html