> Question. I am trying to load and elf file (sh, the first file the kernel
> tries to execute) and it's failing. It seems as if it's trying to load it
> at 0x400000 which is a physical address. The function do_load_elf_binary
> in fs/binfmt_elf.c returns no error, but when it tries to pad zeros or
> execute it starting at 0x400000 logical it dies. Is there some mapping
> that has to go on that I have failed to set up ? I am using version
> 2.0.21 of the kernel on the Algorithmics P4032 board with a 5230
> processor from QED.
That function tries to access the userspace using clear_user(). Doing
so it will also load TLB entries into the TLB. It might be that the
pipeline of the 5230 is a bit different such that this fails.
What exactly does happen? If you can catch the error - what's the TLB's
contents after your problem happend?
For testing purposes it is probably helpfull if you use a very small
program like the assembler written hello world program available on