geninitrd + initramfs sucks more

[Szymon Siwek]

I use grub. Kernel cmdline looks like "root=foo 5".

geninitrd-10000-2 invoked with --initrdfs initramfs.
/sbin/init enters runlevel = 3 (initdefault). Wrong.

geninitrd-10000-2 invoked with --initrdfs rom.
/sbin/init enters runlevel = 5. OK

Absolutely no idea WTF.

[Patryk Zawadzki (patrys)]

romfs is launched and the real init is executed by kernel after romfs exits.

initramfs has to manually start init and that's why init does not get any command line arguments.

[Szymon Siwek (slsiwek)]

Oh, it's so obvious. Looks like my geninitrd-analyzer worn off.
I think about patch, but it doesn't look trivial.

[Elan Ruusamäe (glen666)]

On Sunday 23 November 2008 23:04, Patryk Zawadzki wrote:
> romfs is launched and the real init is executed by kernel after romfs
> exits.
>
> initramfs has to manually start init and that's why init does not get
> any command line arguments.

how about patch to init to parse /proc/cmdline itself?

--
glen

[Patryk Zawadzki (patrys)]

If we want SystemV compatibility, the patch would likely consist of parsing the useful bits from /proc/cmdline and passing them to init.

If we choose to move to upstart then it's trivial to implement it entirely in the event (the same way "noX" is handled to force runlevel 3).