<ellyq>
jenneron[m]: Angelo from Collabora has MT8195, not sure if I had mentioned that
<ellyq>
Nicolas NRA Prado has MT8192
<jenneron[m]>
maybe there is a problem with our kernel config
<weirdtreething[m]>
I'm having a look at devicetrees
<janrinze>
my setup is almost how i want it, I can boot from nvme, that will make it easier to swap from debian to pmOS. making pmOS USB sticks can then be done from the laptop too.
<jenneron[m]>
i don't use defconfig from the fork because it is big af, kernel doesn't even build in CI because of CI timeout
<ellyq>
janrinze: i'm planning to add support for MT8186, MT8192, MT8195 in submarine
<jenneron[m]>
i think janrinze got even smaller kernel than me, but i'm trying to support all mt8186/mt8192/mt8195 while he needs only mt8195, so i'm not sure we can get that merged in pmOS
<ellyq>
we simply package it as ChromeOS kernel image, so depthcharge jumps to it
<jenneron[m]>
that's cool
<weirdtreething[m]>
it needs to be more friendly to devices without a keyboard first
<ellyq>
we can pass custom DTB to target kernel, patch ACPI on x86, change cmdline on the fly, switch between distros/kernels
<ellyq>
i.e i'm dual-booting pmOS and Fedora on KRANE (MT8183) using it
<jenneron[m]>
ellyq: it would be cool to make it working just as initramfs, so maintainers/people can supply their own kernel
<jenneron[m]>
for example, we have linux-postmarketos-exynos5 kernel package in pmOS
<weirdtreething[m]>
well it pretty much does work like that
<jenneron[m]>
it would be nice to re-use this kernel
<janrinze>
jenneron[m]: exynos5, hope to be testing that one when i have my laptop ready.
<weirdtreething[m]>
submarine just builds the initramfs into the kernel to make it easier to use on depthcharge
<ellyq>
yup
<jenneron[m]>
so we can install 2 packages: submarine and linux-postmarketos-exynos which will result in /boot/boot.img or /boot/boot.kpart or whatever
<jenneron[m]>
and we could re-use submarine with other kernel packages
<janrinze>
jenneron[m]: the size of the kernel for me was just so I don't have to fiddle with the chromeos partitions.
<ellyq>
yuo can use any kernel source really
<ellyq>
we wanted to keep it under 32MB
<janrinze>
does depthcharge support gzipped kernels?
<weirdtreething[m]>
yes
<jenneron[m]>
well, maybe building it on kernel upgrade is not a very good idea, idk yet
<ellyq>
yup
<janrinze>
then just use image.gz that's going to be less than 32MB for sure.
<weirdtreething[m]>
submarine on arm64 is huge because of all the DTBs
<jenneron[m]>
ellyq: for armv7 chromebooks you would want it under 8 or 16
<janrinze>
armv7 is armhf?
<weirdtreething[m]>
realistically you would need one build per board on armv7
<janrinze>
armhf has self-decompressing kernels.
<janrinze>
(or did they remove that too?)
<jenneron[m]>
janrinze: yes, but keep on mind 8M is a limit for kernel + dtb + initramfs
<jenneron[m]>
so you have to keep everything under 8M
<ellyq>
yup
<janrinze>
Ooh.. that's not much.
<jenneron[m]>
yeah
<jenneron[m]>
i really like pmOS solution to this problem
<jenneron[m]>
we even have FDE working on these poor devices
<ellyq>
you'd need heavy compression and bare minimum
<weirdtreething[m]>
8mb for the tegra boards, right?
<janrinze>
32MB for kernel + dtb + initramfs could be okay if kernel and initramfs are both compressed.
<weirdtreething[m]>
veyron is 16mb
<jenneron[m]>
janrinze: we have prebuilt images for exynos5 chromebooks, so you can do it without pmbootstrap
<ellyq>
they indeed are
<janrinze>
Tegra K1 is also here, so i can test that perhaps next weekend.
<jenneron[m]>
weirdtreething: 8M is on exynos 5
<weirdtreething[m]>
ah
<jenneron[m]>
on tegra it is 16, but there is a problem that kernel overrides initramfs in memory on decompressing
<jenneron[m]>
so initramfs has to be placed with an offset
<janrinze>
well, the exynos5 has a very old environment.
<jenneron[m]>
this makes it smaller than 16M
<jenneron[m]>
janrinze: what do you mean?
<janrinze>
chromeos on exynos 5 has stopped with 3.18 kernels..
<jenneron[m]>
yeah, chrome os on it is garbage
<jenneron[m]>
no android apps etc
<weirdtreething[m]>
well to be fair, it is the first arm chromebook
<janrinze>
was very early back then.
<janrinze>
that's what i meant
<jenneron[m]>
weirdtreething[m]: they had android apps on e.g. mt8173 with armv7 userspace, so they could get it working on exynos 5 i guess as well?
<jenneron[m]>
but anyway, i don't really care about chrome os
<jenneron[m]>
these devices are well supported in linux
<weirdtreething[m]>
janrinze: im surprised that google updated veyron to 4.14, for a while it was stuck on 3.18
<janrinze>
my exynos5 laptop has been running debian for almost a decade :-D
<weirdtreething[m]>
jenneron[m]: google doesnt really bother with old devices
<janrinze>
indeed, abandonware..
<weirdtreething[m]>
at least we can use regular linux on these devices
<janrinze>
NVidia also is ditching the Tegra and friends. Yes they still run but L4T but up to 21.4 i.r.r.c.
<jenneron[m]>
janrinze: with GPU?
<janrinze>
anything older than the Tegra AGX
<janrinze>
yes with GPU
<janrinze>
the exynos5 has the blob from chromeos.. ;-)
<jenneron[m]>
janrinze: anything downstream goes this way
<jenneron[m]>
janrinze: meeh
<jenneron[m]>
but that gives you OpenGL 3.1 i guess
<jenneron[m]>
we have only gl 2.1 and gles 2.0 with mesa
<janrinze>
There should be some recent/latest patches for Tegra K1 GPU in the kernel and in mesa
<jenneron[m]>
janrinze: nouveau is just broken
<jenneron[m]>
i hope for nvk
<janrinze>
have not tested either.. just the L4T stuff which is very fast on the K1
<jenneron[m]>
there is probably no wayland support, so just wait for a new xorg api incompatible with those drivers and that's it
<janrinze>
--force ;-)
<jenneron[m]>
yeah
<jenneron[m]>
on tegra 3 you would get a segfault from xorg
<jenneron[m]>
tegra 4 as well. tried it a while ago
<janrinze>
never got linux on my tegra 3. that one is on a shelf in a dark corner of the house.
<jenneron[m]>
janrinze: what device?
<janrinze>
some chunky old acer or asus.
<jenneron[m]>
well, specify
<jenneron[m]>
we have many of these tegra devices supported in U-Boot and pmOS
<jenneron[m]>
we got pcie working since then though
<weirdtreething[m]>
oh ok so it doesnt give any more details
<janrinze>
weirdtreething[m]: I'm still busy setting up the laptop to make things easier and to be able to quickly swap and such.
<ellyq>
janrinze: can you try re-modprobing audio modules once it reaches userspace?
<janrinze>
sure
<ellyq>
i think it might be related to AltD
<ellyq>
AltDP mixer*
<janrinze>
my rootfs is >64GB so it is all a bit slow back and forth to the USB stick..
<janrinze>
AltDP crashed somewhere i.i.r.c. from dmesg
<ellyq>
yup
<janrinze>
actually that probably is AltDP sound
<janrinze>
AltDP used to work okay, last time I tested it.
<ellyq>
if you have a pointer to AltDP bridge in devicetree, try commenting it out
<janrinze>
That's DVI over USB-C, right?
<ellyq>
we've had issues with that on KRANE
<weirdtreething[m]>
ellyq: cros has it, collabora doesnt
<ellyq>
well, DP/HDMI :P
<janrinze>
yup, it worked fine. I think the sound is part of the 'real' HDMI standard and as such probably requires some 'magic' or encrypted stuff.
<ellyq>
not really, I2S
<weirdtreething[m]>
does this kernel have dynamic debug enabled?
<janrinze>
no idea. it does have debug enabled.
<weirdtreething[m]>
getting debug messages from the module may help
<weirdtreething[m]>
the module is mt8195_mt6359
<janrinze>
The backup has finished, now copying and will unpack on the nvme. That wil take a bit of time but unpacking is faster than zipping ;-)
robclark_ has quit []
robclark has joined #linux-cros-arm
hexdump01 has quit []
hexdump0815 has joined #linux-cros-arm
<hexdump0815>
janrinze: to get smallest kernel for depthcharge oh 64bit arm the best is to use something like "lzma -9 -z -f -k -v Image" for the kernel and xz for the initramfs
<hexdump0815>
ellyq: does submarine work well if the submarine kernel and the finally booted one have different dtb files? when i tried that maybe about two years ago it did not work as there was a bug preventing this to work
<hexdump0815>
maybe it has been fixed meanwhile, which would be very nice as this would really help to solve the booting on depthcharge