An unofficial Debian Installer for Pinebook Pro

[rillian]

Yes, the included uboot doesn't support compressed kernels. How long did you wait for the boot to complete? One of the issues danielt mentioned patching was to avoid scheduling anything on the slower cores until the cpu frequency manager is running; u-boot apparently hands execution over to the core with them scaled down to the point where booting takes minutes rather than seconds.

[as400]

Regarding cpus.
Can anyone point me to a working method of bringing big cores back online manually ?

If I do ”echo 1 >” I get input/output error.

[gillham]

Hey @danielt thanks for making this installer and for recently adding 'do_shell' as well.  It makes it easier to chroot in and run tasksel.  I select the bare minimum when the initial installer runs so I can add my apt-cacher-ng proxy information later and run tasksel.  That way it pulls all the packages from my local cache.
I added it to the installer as a mirror argument to debootstrap but then it wouldn't pull down the kernel from the suse link.  Something with my apt-cacher-ng not liking the url that I haven't looked into.
Yeah I shouldn't be installing over and over but have been running into a bunch of stuff that dumps core while trying to find a working firefox.
And when it is really broken it is somewhat easier to just reinstall.

[xmixahlx]

Be careful upgrading to unstable packages as I installed firefox from unstable (which works) but later did a 'dist-upgrade' that installed a number of new packages from unstable.  Then gdm3 wouldn't come up due to a segv somewhere.  I tested a few times and figured out a number of packages that needed to be held back.  I don't have the list handy right now, but I can post it later.  Basically it seems to me there is a compiler problem affecting us.

EDIT:  FWIW, here are the packages from unstable I had to hold to keep GDM3 from core dumping.  There might be more, this is just what I ran into as mentioned above.

echo "libgjs0g hold" | dpkg --set-selections
echo "libgirepository-1.0-1 hold" | dpkg --set-selections

echo "libglib2.0-0 hold" | dpkg --set-selections
echo "libglib2.0-bin hold" | dpkg --set-selections

echo "libwayland-client0 hold" | dpkg --set-selections
echo "libwayland-server0 hold" | dpkg --set-selections

you need to pin apt preferences and set unstable and experimental packages to lower than 100 priority.

I use 2 for experimental and 3 for unstable just to confirm the configs work.

[Jeremiah Cornelius]

@danielt is your kernel config shared anywhere here? I'd like to repro it exactly, with the addition to run armhf binaries when making.
Thanks!

[xmixahlx]

pass ARCH=armhf to the installer to build a complete armhf system

the kernel is arm64 and the config is in /boot on your system already.

could post the config later tonight if you are on a diff system...

[jpakkane]

Chromium has migrated finally to testing and can be installed with apt. Unfortunately it has no HW acceleration for video for some reason. It currently pegs all cores playing a 360px video on Youtube and even then the frame rate is low.

[danielt]

My kernel was half the size of Daniel's kernel. That made me realize that mine was compressed while his was not.

I assume that's why uboot refused to boot it and moved on to the next kernel in the extlinux.conf list.

Now that I uncompressed my kernel, a boot ends up in a black screen. I assume Daniel put some magic into his kernel deb.

Sort of. There is an environment variable that controls which kernel image goes into the deb and it needs to be set to point to an uncompressed image (as it happens I *did* have to add some code for get KBUILD_IMAGE to work on the openbuildservice but it is not needed for local builds).

Try something like:

Code:

make  KBUILD_IMAGE=arch/arm64/boot/Image bindeb-pkg

Note that if you have already built a kernel then you don't have to build the kernel again, you can simply uncompress it:

Code:

sudo mv vmlinuz-$ver vmlinuz-$ver.gz
sudo gunzip vmlinuz-$ver.gz

---

@danielt is your kernel config shared anywhere here? I'd like to repro it exactly, with the addition to run armhf binaries when making.
Thanks!

Depends how exact you want to be!

As mentioned in a previous post https://forum.pine64.org/showthread.php?...7#pid56887 I don't want my kernel tree to distract attention from Thomas' since it simply fragments the community (that post also shows how I build the kernel).

However neither do I want to be secretive, especially since the kernel config is included in the kernel itself (try: gunzip -c /proc/kconfig.gz ). Should you ever need it the exact sources can be found on the openbuildservice: https://build.opensuse.org/package/show/...elthompson .

---

Regarding cpus.
Can anyone point me to a working method of bringing big cores back online manually ?

If I do ”echo 1 >” I get input/output error.

Big cores are brought back online (as root) with:

Code:

echo 1 > /sys/devices/system/cpu/cpu4/online
echo 1 > /sys/devices/system/cpu/cpu5/online

If you get input/output errors when running the above as root then check the dmesg log to see if there are any extra details (especially if you are running a self-compiled kernel).

[as400]

Big cores are brought back online (as root) with:

Code:

echo 1 > /sys/devices/system/cpu/cpu4/online
echo 1 > /sys/devices/system/cpu/cpu5/online

If you get input/output errors when running the above as root then check the dmesg log to see if there are any extra details (especially if you are running a self-compiled kernel).

If I do "echo 1 > /sys/devices/system/cpu/cpu4/online" I get input/output error.
Dmesg shows the same - just "input/output error" - nothing else.

I use Manjaro 5.5 kernel.

[danielt]

If I do "echo 1 > /sys/devices/system/cpu/cpu4/online" I get input/output error.
Dmesg shows the same - just "input/output error" - nothing else.

I use Manjaro 5.5 kernel.

Can you explain in more detail what you mean by "shows the same" (perhaps post the dmesg output to a pastebin or github gist)?