Care to elaborate a bit on that, is there more information somewhere, a bug report, anything ?
*i* haven't... because the linux-sunxi community operate off of non-free infrastructure.
You could also report that kind of things to LKML, as a generic ARM problem, it'll probably reach a significant part of the same people...
if you'd like to report a bug using the non-free resource known as github, or would like to join their mailing list using the proprietary web interface, and are happy to have your email address and your copyrighted words treated as "advertising fodder" by google, please feel free to do so! :)
There are alternatives, and I don't understand why *I* should report something I don't have experienced myself, I don't have the HW to reproduce and is too vague... But I understand your time is precious these days.
there's clearly a bug somewhere around 4.0rc5 to 4.0rc6 which, if found, fixed, and patched, would make mainline perfectly acceptable for ongoing usage (with the EOMA68-A20).
That should be easy enough to bissect, if you have a reproducer.
i tried: it was an absolute bitch. as in, i tried for THREE DAYS, did approximately 100 kernel compiles, and still couldn't isolate it. the problem stems from the way that "git bisect" works, compounded by the fact that other kernel errors interfered with the assessment of whether a given kernel was "good" or "bad".
Yes, sometimes there are multiple bugs that fight against your bisection, I already have endured one of those (in intel's drm driver), but eventually found the culprit... It was time consuming though...