Re: is avoiding compat ioctls possible?

[Arnd Bergmann <arndbergmann@googlemail.com>]

On Wednesday 28 October 2009, Andi Kleen wrote:
> > > However some architectures need special operations on compat pointers
> > > (s390 iirc), but if you don't support those it might be reasonable
> > > to not support that.
> > 
> > s390 has to sign extend all 32-bit compat process pointers when
> > processing them in the 64-bit s390 kernel.  I think one other
> > architecture has this kind of situation too.
> 
> Which other architure? I reviewed all the definitions in tree
> and don't see any other than s390 doing magic there.

I'm also pretty sure that s390 is the only one needing this, I
added the compat_ptr stuff initially.

Note that a cast from pointer to unsigned long to u64 and back
in C does the correct 31 to 64 bit extension, which btw is not
a sign-extend but a unsigned extend clearing the upper 33 bits.

The easier rule to remember should be to always to compat_ptr()
on any pointer coming from user space, and to avoid pointers in
data structures where possible, as DaveM pointed out.

	Arnd <><