<samueldr>
not sure there's anything useful in there
<samueldr>
there are no SD cards plugged in, u-boot was able to start
<samueldr>
from what I understand, there may be multiple "mass storage" kinds, could it be that the one drive I'm using is fine for the raspberry pi itself, but not for u-boot?
<Dezgeg>
probably the real problem is the 'unable to get device descriptor (error -110)' error
<samueldr>
(you're the expert)
<samueldr>
common/usb.c
<Dezgeg>
I would try another usb stick or playing with the usb commands
<samueldr>
well, fun time ahead, with 2017.03 it boots, I'll try to narrow down something later
<samueldr>
at least you telling me it should or did work made me confident in trying
<samueldr>
(right now it was easier than getting the serial interface to fiddle with u-boot commands)
<Dezgeg>
no keyboard?
<samueldr>
it doesn't work?
<samueldr>
well, at least, not with the build
<samueldr>
oh, but seems to work for 2017.03
<Dezgeg>
sounds like more brokenness in the usb drivers then
<samueldr>
:/
<Dezgeg>
I do recall it working for me in latest versions though
<samueldr>
though with the new bootloader booted from mmc the keyboard works
<Dezgeg>
probably it wil also boot from the stick now if you remove the card
<samueldr>
hmm, no, but maybe I removed it too soon?
<samueldr>
I got the bad descriptor
<Dezgeg>
or it's an intermittent problem possibly
<samueldr>
as long as that usb drive is plugged, 2018.07 will fail usb, apparently
<samueldr>
a couple tries (maybe not statistically significant) and it looks that way
<samueldr>
using "usb reset" then on 2018.07, plugging in the usb drive gives me error=-110
<samueldr>
and everything stops working
<samueldr>
ah, let me rephrase, booting the sd card without the usb drive, removing the sd card, the u-boot command line works. Plugging in the drive, "usb reset", and then everything fails
<samueldr>
(no need to remove the sd card)
<samueldr>
and usb reset, without the drive seems to always succeed :/
<Dezgeg>
sounds "fun"
<samueldr>
at least the problem seems isolated enough, while I don't have the expertise (yet) to figure out good means to debug, I can probably bisect it enough to get something useful
<samueldr>
100% repro helps
<samueldr>
thanks for insights
<Dezgeg>
yes, bisecting should give clues
<samueldr>
(just nothing, 2017.11 shows the issue, but that's not an invitaiton to anything, just sharing the first step: checking versions already available in nixpkgs)
<samueldr>
(just noting*)
jtojnar has joined #nixos-aarch64
Enzime has quit [*.net *.split]
Enzime has joined #nixos-aarch64
jtojnar has quit [Quit: jtojnar]
jtojnar has joined #nixos-aarch64
jtojnar has quit [Quit: jtojnar]
<thefloweringash>
Re: scrollback, can confirm, running Nixos on a rk3399 "kevin" chromebook. I haven't got any GPU acceleration going, just `services.xserver.videoDrivers = [ "modesetting" "fbdev" ];` set.
<thefloweringash>
I'm not sure what exactly people mean by "mali blob". My understanding is that there's a kernel side interface that's gpl, and a matching libmali.so which is not open in any way. If I could, I would use the one from ChromeOS, but it's 32-bit, and my nixos is 64-bit, which I believe means it's a no go.
orivej has joined #nixos-aarch64
orivej has quit [Ping timeout: 244 seconds]
<andi->
thefloweringash: have you tried using etnaviv with just wayland instead of X?
<thefloweringash>
I hadn't heard of etnaviv. Skimming their top page, sounds like it's for a different GPU? rk3399 is Mali-T864
<thefloweringash>
Unfortunately not. The versions between the kernel driver and the libmali have to match, and Google's kernel version isn't the same one as Rockchip's libmali release. (or at least wasn't when I last tried)
<sphalerite>
right
<sphalerite>
oh so you're using the chromeos kernel?
<thefloweringash>
In theory it should be possible to update the kernel with the matching open source mali drivers, but I didn't manage it
<thefloweringash>
I am, yeah
<sphalerite>
ok
<sphalerite>
I'm getting a similar chromebook quite soon
<sphalerite>
but bob, not kevin, because kevin isn't availbel in Europe apparently
<sphalerite>
thefloweringash: any reason you repackage vboot_utils rather than using the one in nixpkgs?
<thefloweringash>
I think the only reason was that I didn't notice it in nixpkgs, and the full build didn't work without cross compiling (fixed by #37127). I made a branch to switch to upstream a few months ago, but forgot to merge it. Probably becuase I was trying to keep the repo compatible with release-18.03 and nixos-unstable.