Intel MPX Skylake ISA

Carsten Mattner carstenmattner at gmail.com
Sat Sep 13 12:58:23 PDT 2014


On Sat, Sep 13, 2014 at 7:48 PM, Matthew Dillon wrote:
> Really unlikely.  A better solution is to eventually migrate away from

Eventually that will happen but isn't -fmpx useful if available
like the NX bit even though GRsec and OpenBSD have many more
unassisted mitigations going. I mean there's going to be C code
- especially random C code in the ports tree - running for quite
some time. Niklaus Wirth showed us how to write whole operating
systems on custom hardware in a child of Pascal that didn't make
the same mistakes as the AT&T creation but nobody cared
or others had bigger marketing budgets. I digress but yes
we will move away from C some day but until then have to
use debugging tools and extensions like this and other contortions
to make up for it.

> languages like C which do not do bounds-checking of pointers.  These
> hardware solutions are non-portable and Intel in particular tends to create
> a real mess for backend implementations with their ad-hoc 'solutions'.

fwiw this the patchset for Linux https://lkml.org/lkml/2014/9/11/182

If -fmpx has low overhead then the chances of adoption in peoples
CFLAGS are higher than -fstack-protector.

> On Sat, Sep 13, 2014 at 3:38 AM, Carsten Mattner <carstenmattner at gmail.com>
> wrote:
>>
>> Any plans to include the necessary support for Intel MPX
>> (https://en.wikipedia.org/wiki/Intel_MPX)?



More information about the Users mailing list