Comment 36 for bug 633983

This was the closest bug I could find to an issue I have been having with maverick built kernels and efi. some background first:

I have been building kernels on a linuxmint 9 (lucid) based system, for a joggler device target (efi based device - see http://en.wikipedia.org/wiki/O2_Joggler).

The bug:

Everything was working fine, until I upgraded my system to the linuxmint 10 RC (maverick based). I understand these are not specifically ubuntu, but the build and base system is the same.

After I had upgraded my kernel building machine to linuxmint 10 RC (Maverick based), my kernels have failed to boot on my efi based device with a failure mentioned earlier here "RAMDISK: couldn't find valid RAM disk image starting at 0." the initrd is loaded, but then this failure has occurred. I spent some 7+ hours thinking I had made some error somewhere, and modifying my kernel options, yet even using the config from a previous kernel which worked fine, it would always fail to use the initrd.

Having worked thought every possible "human error" cause I could think of, I could only fall back on the differences between the old and new system. For example GCC/binutils etc. I had already tried a few versions of GCC and it didn't seem to help.

I have however solved the issue now, and it points to an issue with the maverick binutils. After my many failures I downgraded binutils to the lucid release, and rebuild the kernel as usual. It worked!. to make sure it wasn't some fluke, i then upgraded binutils again to the maverick release and rebuilt. again the same failure.

I have no idea what would be causing this specifically with for example LD or some other part of binutils, but I am completely unable to build a kernel that will work with an initrd with the maverick binutils. downgrading them to the lucid version solves the problem completely. This points to what I think may be some significant problem with the maverick binutils, however due to my test case, it is hard to debug further.