On 10/29/07, michael.firth@bt.com michael.firth@bt.com wrote:
Yes, I saw that too with sometimes with the 83xx tree. Sometimes it
machine checks, sometimes it goes in an infinite loop.
This looks like the same issue I'm seeing trying to update my MPC834x based system to u-boot-1.3.0-rcx. Under a BDI2000 it seems that several of the malloc related pointers have addresses in flash, while in my older u-boot version they are correctly pointing to addresses in RAM.
From this thread, it seems that three fixes have been recommended:
- Disable "CONFIG_RELOC_FIXUP_WORKS" in the CPU header files
- Backout some changes which affect the fixup operations (?)
- Migrate to a newer compiler version
Actually, #3 is migrate to an *older* compiler version. :-( Unfortunately EABI relocation seems to be a rather undocumented thing.
If you can use an older compiler, #3 is the easiest workaround. I use ELDK 4.1 personally which include GCC 4.0.0. You can backout the enabling of CONFIG_RELOC_FIXUP_WORKS for 82xx, but I don't want to apply that change to mainline if at all possible. I'd rather get relocation working properly for all compilers once and for all.
Which of these three options is the best to take, or has this thread not yet reached that conclusion?
Thanks
Michael