> Andrew Ip <[EMAIL PROTECTED]> writes: > > > > Most likely what is caused by binutils? > > To be clear, the messages: > | Unknown bootloader class! > | type=0x000101C0 > | data=0x000103BE > | Firmware type: LinuxBIOS > Are printed by the code in mkelfImage when it is loaded under a bootloader > it does not recognize and are non-fatal. > > > I don't know what's exactly causing the problem. However, my > problem went away > > when I downgrade my binutils to 2.9.5 on my Debian system. > Similar problem > > happens on RH 7.2, but not on RH 7.3. > > None of the binutils problems I know of trigger the previous > symptom. If they > can I would like to know about it. > > As an additional data point mkelfImage-1.14 has been fixed so > that it works > on RH 7.2 and it has been tested there. Additionaly the fix was such that > mkelfImage is not pushing the limits of ld so binutils buts should go from > being common to being very, very, unlikely and rare. > > Given that I believe I have fixed the underling problem people > being confident > that any issue must be a binutils problem worry me. > > Eric Because of RH issues with compilers in the 7.X series, and after having broken stuff, I actually still use an old RH6.2 installation, but with some upgraded utilities (like modutils), however, I am still using binutils-2.9.5 from RH6.2.
In fact, Eric, as you said, the messages are printed by code in mkelfImage and it does in fact jump to the kernel - I have added debugging into the kernel, and it hangs up while trying to find the highest page frame number available in setup_arch() in arch/i386/kernel/setup.c. I have not had enough time to look fully into this, but it would appear that maybe the memory tables passed to the kernel are getting screwed up by something. I will try the latest mkelfImage.
