News HyperThreading has serious security flaws

http://www.bit-tech.net/news/2005/05/14/hyperthreading_security_flaw/

 

yay! thar it is!
anyways, i wonder if AMD is going to pay this guy for this. it's sure to hurt Intel sales a bit once it hits the mainstream press.

 

Wouldn't count on it, it really doesn't mean jack squat for the average user. In fact if I read the snippet right, it won't even mean a whole lot to many businesses, but I could certainly be wrong there. But as the fix takes two seconds in the BIOS, I couldn't imagine it affecting sales much.

 

Agreed...but I love my HT so much... It's so bad to hear its name tarnished.

 

I read the article linked on slashdot posted by the guy who found the vulnerablility, but isn't "the fix" switching off HT? Hardly seems like a solution to remove one of the main selling points of a product. The server/business market is huge, and they take advantage of hyperthreading more than anyone, correct? Sounds like a threat to sales to me.

 

Okay, I tried to read the article... but it was a bit beyond me, technically...

However, I would be fascinated to know if the same security issue existed with the new AMD X2 processors, or if they're safe because they have separate L2 cache for each processor...

 

I'm interested in that too, is the flaw present in the A64 X2 processors, because they share the same memory controller? Or is this only a cache related security flaw?

Interesting...

 

My bet would be that it is much harder, but still possible. Though the separate cache prevents a lot of access, the same controller is like having a door between the two. A lot more narrow than an open room, but still an opening.

 

Intel's response http://www.xbitlabs.com/news/cpu/display/20050515030244.html

 

skim read the paper, the first thing he mentions, about trojen and spy program communication. Erm, not really worried about that myself, as what your in effect been able to see, is if a page was accessed, granted its more than you might want to know, but its not reliable.

i'm GUESSING from the intel reply that you need to be in a ring lower than 3 to take advantage of this anyway for some memory models.

Windows users, well it shouldn't matter because they need to know certain information that they wouldn't know (memory locations to determine caching of). They'd have to have a pretty damn genious code to get round it.

Linux users, yes more dangerous as the OpenSSL example shows, but its one of those things, it might take a good few months to be able to read enough to grab a 512 RSA key. But hey linux kernel isn't really interproccess secure now is it.

other running proccesses will effect it, if you've got a security thing thats using that CPU perminantly, then yes, you might be in for a problem.

but updates are been made for windows + linux.

as to wether this problem could come up on the AMD, its more about the timing of the seg fault, now from my knowledge of the A64 X2 i couldn't say, but the OS'es will be ready to work around it, thats for sure.