|
John Regehr
@johnregehr
|
19. ruj |
|
memory tagging should be a game changer for C and C++; get with the program, @intel and @apple!! pic.twitter.com/z2vXAtTb7z
|
||
|
||
|
Rich Felker
@RichFelker
|
19. ruj |
|
MTE requires relatively little work to adopt (I say this as someone who's adamant about not adopting invasive "hardening" snakeoil like CET) and completely closes off huge classes of vulns.
|
||
|
|
||
|
Gok
@Gok
|
19. ruj |
|
I keep hearing this and I feel like I'm missing something. What classes does it completely close off? The 4-bit tags seem ridiculously easy to bypass to a motivated attacker.
|
||
|
|
||
|
The Doge Mocenigo
@DogeMocenigo
|
19. ruj |
|
It is not (only) a (weak) mitigation - it is a tool to discover bugs and vulnerabilities in the field. Testing - including smart fuzzing - does not find all memory access defects. But if MTE information is collected IN THE FIELD you find these defects and with absolute precision
|
||
|
|
||
|
|
Gok
@Gok
|
19. ruj |
|
If that was really the goal you could just opportunistically deploy (HW)Asan, which is more precise and doesn’t require replacing billions of devices.
|
||
|
|
||
|
|
The Doge Mocenigo
@DogeMocenigo
|
19. ruj |
|
And re: a lot of work to adopt. Link to a different version of glibc and use a compiler flag is really too much?
|
||
|
|
||
|
|
Gok
@Gok
|
19. ruj |
|
Every binary needs to be updated. And everyone that wrote their own memory allocator (i.e. every sufficiently large C project) has to do work.
|
||
|
|
||