EFI Stub with embedded initramfs boot failure since 2.0.0.

[tietzc]

I boot via efi stub and have the initramfs embedded into my custom kernel (gentoo-sources). The only kernel command line option is: net.ifnames=0

Other than that I have only added cryptsetup_trim = true to /etc/ugrd/config.toml

I created the initramfs by ugrd –no-firmware, as all necessary firmware is already embedded into the kernel.

This worked like charm until ugrd-2.0.0.

Here is a screenshot at the point where everything stops. The initramfs is not being loaded at all.

Hope we can debug this :-)

[desultory]

I think this is solved in the dev branch currently: #320

Without that change, using the "ugrd.base.compat" module should do the same thing.

I think this is because 2.0.0 removed the console device node from the image.

[tietzc]

Awaiting push to master then. I will try 9999 on Gentoo soon.

Thanks alot.

[desultory]

Awaiting push to master then. I will try 9999 on Gentoo soon.

Thanks alot.

before I push it can you give it a test? There is a small chance it may also need /dev/null (another thing which would be in the devtmpfs)

You should be able to simply add ugrd.base.compat to your modules list in /etc/ugrd/config.toml

I haven't been able to test this in a VM, and I'm not sure how I can add this into automated testing without having the runners build a kernel.

[tietzc]

Will do so. Talk to you after rebooting :-) Hopefully :-)

[tietzc]

Same result.ugrd.base.compat did nothing

[desultory]

Same result.ugrd.base.compat did nothing

can you confirm the /dev/console node is in the image? It could be that it needs /dev/null and others as well.

If you decompress the cpio image, and run "pycpio -pi " you should see something like:

CPIO_CharDev [0] dev/console:
    Permissions: rw--r---r--- 0 0 (5, 1)

I'll try to test this in a VM later once I have the time. Could you try and confirm that the image works if you don't embed it? like using the initrd= line? That would confirm this has to do with packing it into the kernel

[desultory]

hmmm i just tested in a vm, and adding the console device node seems to fix it, i replicated the broken output without that.

[tietzc]

On a second attempt it does also work for me. I must have done something wrong earlier.

[desultory]

On a second attempt it does also work for me. I must have done something wrong earlier.

thanks for testing, it's real easy to miss a step packing. I'll merge the fix into the main branch :D

[tietzc]

This works fine on all of my machines with dev-python/zenlib-9999 , dev-python/pycpio-9999, and sys-kernel/ugrd-9999.

Would you mind making a new release?

[desultory]

This works fine on all of my machines with dev-python/zenlib-9999 , dev-python/pycpio-9999, and sys-kernel/ugrd-9999.

Would you mind making a new release?

sure, it's in progress now, thanks for helping find this bug

gentoo/gentoo#42787