--- crowd-funded eco-conscious hardware: https://www.crowdsupply.com/eoma68
On Sun, Feb 12, 2017 at 9:23 AM, Tzafrir Cohen tzafrir@cohens.org.il wrote:
On Sat, Feb 11, 2017 at 12:39:46PM +0000, Luke Kenneth Casson Leighton wrote:
this is a decision that is easily justifiable based on the fact that it's going to have to be distributed with the sunxi 3.4 kernel as that's the only one which supports the full hardware.
That's a proper technical argument, indeed. Anybody here managed to get the Jessie systemd running on the 3.4 sunxi kernel?
Also: I figure that currently mainline kernel support for those is still not good enough for the image. But what are the chances for me to later have it running with mainline kernel and get a decent hardware support?
you'll have to hunt for the patch that was somewhere around 4.7rc0 to rc4 which causes the a20 cards to go unstable and crash arbitrarily within 30 to 200 seconds. i compiled about 100 versions of the linux kernel source trying to track down exactly where it was, but the above was as far as i could get. i had to stop as it was simply absorbing too much time right during the middle of the campaign.
i documented this on the updates.
[side-note directly to you tzafrir: btw i assume you'll be reading this, and that you've seen my response about libselinux1. i'm mentioning it because i have a vague recollection of answering your question multiple times, and each time you raising the exact same question demonstrating that you haven't seen my response. acknowledgement greatly appreciated so it doesn't keep on happening].
there appeared to be some significant chances to devicetree around that time and the cubieboard2 (which i was using as the basis for testing) was not kept up-to-date around that time.
once that bug which was introduced around that time has been found and fixed use of mainline kernel support should be absolutely fine...
but bear in mind that there is still a library needed to be written (which must go into u-boot as well) which reads the EOMA68 EEPROM at address 0x51 and loads the required "overlay" (devicetree fragment).
i just mention this just in case you were considering upstreaming an arbitrary devicetree file for the EOMA68-A20 - please don't, not without consulting me first.
l.