Re: [Qemu-devel] Re: [RFC][PATCH] x86: Optional segment type and limit checks - v2

[Jamie Lokier]

Jan Kiszka wrote:
> Don't want to stop all your creativity, but just like Paul I'm also a
> bit skeptical about the TLB way of achieving range and type safety.

Sure, that's understandable.  There has to be some "show me the code"
- I wish I had the time.  These are not simple optimisations, but they
could make segment checks free for most guests, so it's worth
considering them even if they're discarded.

> My major concern once was that the TLB works on a global scope so that
> you cannot tell the original segments behind some address apart.

Yes.  The only way around it is context - either having multiple TLBs,
or inlining segments into translated code, so you tell the original
segments apart at the call site.

> And extending the virtual address space for this is a no-go on
> 32-bit hosts (which I unfortunately had and still have to support
> here :->).

Definitely not increasing the virtual address space - I'm still using
32-bit hosts too :-)  I'm not sure that would be safe even on 64-bit
hosts - what if the guest has its own uses for the whole address
space?

Expanding the TLB key space (or equivalently, multiple TLBs) is
different from expanding the virtual address space, though.

-- Jamie