Linux vanilla 4.4.x + mainline buildroot

Options
Jaap Crezee
Jaap Crezee New Member Posts: 19 ✭
in UP Board Linux
Hello fellow UP-ers, community,

We have succeeded into (partly) porting our product to the UP-board. The performance is very nice, although in idle the board still draws quite a lot of power.
I can confirm that using non-vendor software works, but not everything is supported by mainline/vanilla Linux yet, but hopefully will (Intel?!).

We are using git.linux.org / linux-stable/linux-4.4.y with a recent (git) version of buildroot.
Oh and btw, the board can only be booted using EFI. We are directly booting the kernel as EFI-stub without the use of a bootloader. One should call the kernel EFI/BOOT/BOOTX64.efi in that case.

Have fun!

Jaap Crezee
ZUPR BV - Netherlands
· Share on FacebookShare on Twitter

Comments

  • Javier Arteaga
    Javier Arteaga Emutex Posts: 163 mod
    Options
    Hi Jaap,

    Thanks for sharing! Glad to hear you've had success in applying it to your product. We definitely want to upstream the Linux support, but it will take some time.

    On the current SoC for UP, power consumption will be higher overall than ARM-based equivalents (on the plus side, the higher performance enables new usecases). Regarding idle power specifically, did you measure from ubilinux or from your custom-built kernel? I ask this because, in ubilinux, we set the intel_idle.max_cstate=2 parameter by default, to reduce instances of this hardware bug:
    https://bugzilla.kernel.org/show_bug.cgi?id=109051
    (recently acknowledged for Cherry Trail as CHT45 in http://www.intel.com/content/www/us/en/processors/atom/atom-z8000-spec-update.html)
    This stability workaround, however, negatively impacts power performance, as it disallows deep sleep states.

    Regards
    · Share on FacebookShare on Twitter
  • Jaap Crezee
    Jaap Crezee New Member Posts: 19 ✭
    Options
    Hello Javier (?),

    I did not run into any stability issues with our own software running on the vanilla kernel without any intel_idle.max.. parameter.
    I measured the idle dissipation with both an Archlinux liveiso, Ubilinux and our own firmware loaded. I can produce some defined numbers later this weekend.
    What is curious though, is that with Ubilinux I do experience random crashes (and not with all other software). Not sure why yet, because the board hangs with no output and haven't had the time to connect Uart0 yet.

    Jaap
    · Share on FacebookShare on Twitter

Categories