Jens, On Wednesday 03 November 2004 14:28, Jens Schmalzing wrote: > Alessandro Amici writes: > > Severity: important > > [...] > > > nice black magic trick: > > Well, I'm not sure if a bit of editing the kernel binary qualifies as > "major effect on the usability" or black magic, but anyway. As long > as you don't make it RC.
it took me a couple of days of searching through the net and on irc before i found the correct 'bit of editing' ;-) and i could not even start the installation process before that. so yes, i stand that it had a "major effect on the usability" of the package for me (and other people with chrp). anyhow, i will not complain if you lower the priority of the bug (at least the info on a possible work-around is now in the BTS). BTW. i filed the bug to kernel-image-2.6.8-power3 as suggested by Sven, but the problem is actually found in power3/vmlinuz.initrd, i don't know who is really in charge of that file. > > - with an hexeditor look for the values '00 0c 00 00' and '00 00 40 00' > > You mean '00 c0 00 00'. you are right, sorry. > Both values are coming into the compressed > kernel from the note file, which is created by mknote, which is built > from arch/ppc/boot/utils/mknote.c, which is shipped with the upstream > kernel source and has these values hardcoded. > > > in the first few bytes of the vmlinuz.initrd file and change both them > > of them to 'ff ff ff ff'. > > We could add a patch that hardcodes these values instead, meaning > real-base and load-base will be unspecified in the kernel, just like > all the other OF variables. This affects CHRP only, since it is the > only subarch that uses the note in the first place, so it would be > great to hear people who actually have the hardware... that would certainly help people trying the debian-installer on chrp, even if one still needs to manually increase the real-base value from the OF prompt, which needs to be documented somewhere. > > leighbb knows more than me > > ... like Leigh and Sven. cheers, alessandro
