Skip to content
Snippets Groups Projects
  1. Oct 09, 2023
  2. Aug 17, 2023
  3. Aug 09, 2023
  4. Jul 14, 2023
  5. Jul 12, 2023
  6. Jun 26, 2023
  7. Jun 22, 2023
  8. Jun 21, 2023
  9. Jun 20, 2023
  10. Jun 13, 2023
  11. Jun 10, 2023
    • Sebastian Krzyszkowiak's avatar
      defaults: systemd: Force poweroff instead of halt · 3bde7ec5
      Sebastian Krzyszkowiak authored
      Some systems power off after executing "halt". Users with a habit
      of "halt" usage may end up surprised when the Librem 5 appears
      completely off after using it, but actually is not - after halting,
      the SoC remains on, consumes about 1W of power and does not respond
      to any input (including low battery level!) until the power button
      is held for ~5 seconds or power is cut.
      
      There's little reason to ever halt a mobile phone without powering
      it off, so don't let the users shoot themselves in their foot
      (especially given that this can easily end up with battery's
      undercharge protection circuit being involved) and force poweroff
      whenever halt is invoked.
      3bde7ec5
  12. May 04, 2023
    • Sebastian Krzyszkowiak's avatar
    • Sebastian Krzyszkowiak's avatar
      d/control: Update dependencies to support MM's quick suspend resume · e26d7ad2
      Sebastian Krzyszkowiak authored
      We need a recent enough kernel that keeps USB link active during
      suspend and recent enough ModemManager that can handle incoming
      messages arriving during suspend.
      e26d7ad2
    • Sebastian Krzyszkowiak's avatar
    • Sebastian Krzyszkowiak's avatar
      defaults: librem5-pm: Enable simple_ondemand governor for DRAM · e6259ad8
      Sebastian Krzyszkowiak authored
      I did a simple test with a USB power meter to see the power draw over
      5 minutes since running gnome-software in service mode using both
      `powersave` and `simple_ondemand` governor with screen off.
      
      With `powersave`, packagekitd was keeping one CPU busy at 100% for
      slightly above 4 minutes. At the end of the 5 minutes period the power
      meter reading was 125mWh.
      
      With `simple_ondemand`, packagekitd was done after slightly above one
      minute already, so the phone was idling rest of the time. Power meter
      reading was 109mWh (it was at about 45mWh at the moment when packagekitd
      finished its work).
      
      I've also recorded two minutes of idling (nothing but kgx running):
      the result was the same for both governors - 28mWh; and two minutes of
      keeping all CPU cores busy with compressing random data: 61mWh for
      `powersave` and 99mWh for `simple_ondemand`.
      
      This is consistent with my anecdata - `simple_ondemand` appears to save
      battery and makes the phone cooler. It also makes the UX much better when
      using the phone over ssh. The one case where it makes things
      worse - a rogue process constantly consuming 100% CPU in
      the background - is already a pathological case with poor results on both
      governors. When the process actually does some sensible and finite work,
      it's going to finish it *much faster* with simple_ondemand, saving plenty
      of power. Therefore, I'm going to merge this.
      
      To go back to previous behavior, users can always use this udev rule:
      ```
      SUBSYSTEM=="devfreq", TEST=="governor", ATTR{governor}="powersave"
      ```
      e6259ad8
  13. May 03, 2023
  14. May 02, 2023
  15. Apr 14, 2023
  16. Apr 07, 2023