On 2025-09-17, Uwe Kleine-König wrote: > on a system that relies on flash-kernel (here: arm64 machine with an > rockchip SoC) dracut isn't usable. With the intention to make dracut the > default initramfs builder, having it work on systems with flash-kernel > would be beneficial. > > I think adding support is easy but didn't look into the details. > Technically the integration can live in dracut itself or flash-kernel, > but I think the latter is the more sensible option, so this report is > against flash-kernel.
I am not sure this should really block the transition. I believe the initramfs-tools/hooks/flash_kernel_set_root script is mostly used on old armel systems; most armhf and arm64 systems have root= set appropriately, in my experience. The initramfs-hook/flash-kernel arguably should have a correlary, but again, in most cases I am aware of, unless the initramfs generated by dracut changes the initramfs filename every time, flash-kernel does not usually do anything with that filename on most armhf/arm64 systems (and the flash-kernel scripts should get triggered by the corresponding kernel update for the parts that are needed). The initramfs integration for flash-kernel has been a source of headaches and bugs for quite some time (running flash-kernel for both kernel updates and for initramfs regeneration), and I am not positive it is even necessary for a switch to dracut... and would welcome the opportunity to drop or at least not add more code to flash-kernel... which at best is reluctantly maintained. Though some of this may be my lack of understanding of dracut ... does it not support the root= value passed to the kernel? Are there other factors to consider? live well, vagrant
signature.asc
Description: PGP signature
