WIP: Trying to edit kernel config #60
Loading…
Reference in New Issue
Block a user
No description provided.
Delete Branch "lee.miller/openwrt:kernel"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Hi!
As far as I can test this doesn't work. But you can check this image, maybe it works for rpi4.
I wasn't able to test it directly, but I tested it indirectly, by comparing the
kernel8.img
with an old one, and they are the same (size and checksum). So it didn't work.I realized that Image Builder puts the stock kernel into the image. Investigating.
c60e33bc8f
toeb998042f1
The build timed out. I don't want to fiddle around with buildbot config at the moment, so how about you enable parallel builds (see #30)? That may help.
The timeout happens if there is nothing on stdout/stderr for 20 minutes, so if you have multiple
make
children running in parallel it may avoid it.I tested if OpenWRT preserves a third partition, i.e. if it even makes sense to boot from NVMe. The third partition was preserved, so I assume as long as the new image doesn't have a bigger partition assigned in the partition table than the old one, all should be fine.
I need to check it locally more thoroughly. It seems using a wrong config and building modules that we don't need.
In this edition of the
kernel-custom
I just added the parts needed to build the kernel into thepackages/kernel/linux
. Previously I removed parts, that seemed irrelevant and the build failed. After readinginclude/target.mk
I thought it can be done byTARGET_BUILD=1 make
, but it also failed.We'll see, but I'm not sure. BuildKernel in the
include/kernel-build.mk
has.NOTPARALLEL:
We'll see.
Seems to run in a single thread :-(
I bumped the timeout to 2 hours, now I'm cleaning the cache, then I'll rerun the job.
Didn't deploy timeout change correctly, now rerunning the job again, with timeout.
Ok the timout helped to get debug data on the build.
90115e0b47
to80c019c0b3
80c019c0b3
toc68f4ed285
Tried to boot from NVMe, didn't work, did a md5sum on
/boot/kernel8.img
and it's still the same as before.With the new kernel, the OS doesn't boot. Pi is stuck at the rainbow screen and ACT LED flashes three times, then a pause and then it repeats. The flashes are very fast, maybe 100ms apart, then about one second pause.
d9524b19d5
to66393f6aa1
Do you want me to test the latest change?
Yes, if you have spare time. Currently I'm building a similar image for my rpi. The goal is to check if the vanilla kernel is going to boot, because I see a difference between the
.config
generated in SDK byBuildKernel
and that found in the imagebuilder dir.66393f6aa1
to82b5291a27
Mine boots, so it's probably a config issue.
Tried the latest one, still doesn't work. I tried to boot it from an SD card as well and the result is basically the same, there is just an additional reboot in the cycle.
Oh, I see, the kernel for rpi is not in
build_dir/target-${PKG_ARCH}/linux-${OPENWRT_TARGET}_${OPENWRT_DEVICE}
. For packagesPKG_ARCH=arm_arm1176jzf-s_vfp
, but in this sentence it should've beenPKG_ARCH=arm_arm1176jzf-s+vfp_musl_eabi
. So I used the stock kernel for my rpi builds :(82b5291a27
toe533892391
This is strange. The latest image should contain a stock kernel, because kernel-custom failed to build: https://buildbot.bitmessage.org/#/builders/33/builds/18817/steps/3/logs/stdio (line 5670).
When I build the kernel for my rpi with the NVME settings, I'm getting this:
Hmm, it looks like the kernel was built despite of the error, because there is no error after the line
cp /var/lib/buildbot/workers/default/multibuild_child/build/openwrt-sdk-22.03.2-bcm27xx-bcm2711_gcc-11.2.0_musl.Linux-x86_64/build_dir/target-aarch64_cortex-a72_musl/linux-bcm27xx_bcm2711/vmlinux build_dir/target-aarch64_cortex-a72_musl/linux-bcm27xx_bcm2711/rpi-4-kernel.bin
Check if this recent post helps: https://forum.openwrt.org/t/guide-how-to-tweak-the-kernel-configuration-to-optimize-it-for-a-specific-device-sub-target-without-breaking-the-kernel-full-build-script-included/181051
This uses the full source - the environment where the kernel is supposed to be built. I had no problem building kernel from the full source.
Checkout
From your project repository, check out a new branch and test the changes.