Broadcom MIPS kernel bit-rot

[Bruce M. Simpson]

M. Warner Losh wrote:
> I think this may be a change in the SGI style ELF binaries to a
> Traditional SYS V style that we did before we imported the code, but
> after you did the original work on the sentry...
>   

I've committed a fix to SVN now. It seems that the required fix was to 
enforce the order of the ELF program segments using a GNU ld PHDRS 
config section. I also minimized the diff with the existing 
ldscript.mips file.

At the same time I fixed the dynamic hash load -- it turns out CFE will 
not load PT_DYNAMIC segments, however it will blindly load PT_LOAD 
segments containing dynamic ELF sections. Now we get symbol lookup in 
backtraces, however it also seems DDB isn't always smart enough to find 
the nearest symbol.