simonb@wasabisystems.com said:
> The p_vaddr check is just something approximating "is this a low
> address".  It's not really MIPS specific. 

That's still a hack for a mips specific problem, and only
for the COMPAT_16 case. We shouldn't bloat other kernels
with this, and also visually separate it in the source code.
An #ifdef ELF_INTERP_NON_RELOCATABLE would be fine imo.

> I'd like
> to propose that we use the current patch for a couple of reasons
> [...]
> It's been well tested

I concur that it is better to use your tested patch for 3.x.
In theory it would still be conceivable that some linker change
causes a non-PT_LOAD section with a higher (junk) vaddr, undermining
your check, but this is unlikely in the release branch.
For HEAD, we should make sure to find a cleaner fix.

best regards
Matthias