On Tue, 15 Nov 2022 at 15:34, Peter Böhm <peter.bo@xxxxxx> wrote: > > Dear Ard Biesheuvel, > > this is my first kernel bug report ever (and I am not a native english > speaker), please be patient with me ;-) > Hello Peter, Thanks for the report. > I have read > https://www.kernel.org/doc/html/latest/admin-guide/reporting-issues.html > and found your email address here: > https://www.kernel.org/doc/html/latest/process/maintainers.html#maintainers > > I configure my kernels since 20 years but I am not a kernel hacker. I have a > working LTS kernel (5.15.78) in a Gentoo system. > So if I am understanding this correctly, you are reporting that the config below simply does not work at all, and this 5.15 kernel does not use a built-in initramfs, right? Or did something get broken after 5.15 was released? > > A) Bug Summary: > > Kbuild allows me to add an embedded (=built-in) initramfs to my working efi- > stub-kernel (signed because of secureboot). This combination does not boot: > BIOS boot screen starts with: "EFI Stub: UEFI Secure Boot is enabled"; then > blank screen; then two beeps; then trying to boot again; endless loop can only > be stopped by pressing <F12> and entering UEFI BIOS (and selecting a working > kernel). > > Note: Kernel builds without any error message (of course I have beside the > initramfs_list also a init in /usr/src/initramfs) > How much does the size of the bzImage increase due to the initramfs being incorporated? Given the below, it should be tiny, right? > > B) Difference between working kernel and failed: > > CONFIG_BLK_DEV_INITRD=y > CONFIG_INITRAMFS_SOURCE="/usr/src/initramfs/initramfs_list" > CONFIG_INITRAMFS_ROOT_UID=0 > CONFIG_INITRAMFS_ROOT_GID=0 > CONFIG_RD_GZIP=y > CONFIG_INITRAMFS_COMPRESSION_GZIP=y > > I also know a solution called "UKI" (unified kernel image) in other > distributions; but why building such scrap if my beloved kernel has all > necessary pieces also ? Kernel hast efi-stub, built-in kernel command line and > embedded initramfs ... > Agreed. You shouldn't need UKI for this. However, I tested this locally with a rootfs.cpio file I use for testing, and it happily boots in EFI mode and going through the stub etc. Can you check and see if earlycon=efifb produces some more output?
