Hyperthreading under scrutiny with new TLBleed crypto key leak

[wyattzx]

Quote

Last week, developers on OpenBSD—the open source operating system that prioritizes security—disabled hyperthreading on Intel processors. Project leader Theo de Raadt said that a research paper due to be presented at Black Hat in August prompted the change, but he would not elaborate further.

 

The situation has since become a little clearer. The Register reported on Friday that researchers at Vrije Universiteit Amsterdam in the Netherlands have found a new side-channel vulnerability on hyperthreaded processors that's been dubbed TLBleed. The vulnerability means that processes that share a physical core—but which are using different logical cores—can inadvertently leak information to each other.

Source: arsTECHNICA

 

Honestly, how many more flaws/bugs are we going to find in modern CPUs? Between this, Spectre and Meltdown, I'm beginning to think it's only a matter of time before the entire modern CPU architecture as a whole is filled with holes from left to right. I know that's an exaggeration, but it seems this year has not been too kind to Intel or AMD, and Intel is just getting hit again.

 

Thoughts?

[Guest]

I still have my AMD Athlon XP to fix these bugs. I expect AMD will be restocking shelves with them soon. 

[GOTSpectrum]

4 minutes ago, NonaHexa said:

Source: arsTECHNICA

 

Honestly, how many more flaws/bugs are we going to find in modern CPUs? Between this, Spectre and Meltdown, I'm beginning to think it's only a matter of time before the entire modern CPU architecture as a whole is filled with holes from left to right. I know that's an exaggeration, but it seems this year has not been too kind to Intel or AMD, and Intel is just getting hit again.

 

Thoughts?

I wonder how/if this effects SMT on AMD CPUs? 

[mr moose]

5 minutes ago, NonaHexa said:

Honestly, how many more flaws/bugs are we going to find in modern CPUs? B

 

Lot's, they are incredible complex machines now and humans no longer have the computational capacity to get their head around most of the process let alone have the ability to pre-empt security flaws.

[cj09beira]

 

3 minutes ago, Ben Quigley said:

I wonder how/if this effects SMT on AMD CPUs? 

i just read the article, the response is we don't know yet

[GOTSpectrum]

2 minutes ago, cj09beira said:

 

i just read the article, the response is we don't know yet

Oh I know, thats why I asked. PS does AMD even have a name for its implementation of SMT?

[mr moose]

2 minutes ago, Ben Quigley said:

Oh I know, thats why I asked. PS does AMD even have a name for its implementation of SMT?

Fuck, I always thought SMT was their name.

[cj09beira]

17 minutes ago, Ben Quigley said:

Oh I know, thats why I asked. PS does AMD even have a name for its implementation of SMT?

yes its SMT, amd has the tendency to not name things that dont need to be named, like DMI (aka the pcie 3.0 x4 that connects the cpu to chipset), amd is more straight forward, they just call it a pcie 3.0 x4 connection

[GOTSpectrum]

13 minutes ago, cj09beira said:

yes its SMT, amd has the tendency to not name things that font need to be named, like DMI (aka the pcie 3.0 x4 that connects the cpu to chipset), amd is more straight forward, they just call it a pcie 3.0 x4 connection

Makes sense, thanks for that.

[exetras]

If you take an earlier i7 and turn off Hyperthreading you basically have an i5. Would they become equals at that point? (if GHz were the same)

[cj09beira]

46 minutes ago, exetras said:

If you take an earlier i7 and turn off Hyperthreading you basically have an i5. Would they become equals at that point? (if GHz were the same)

almost the same the i7, as the i7 has slightly more cache

[asus killer]

this is starting to look a lot like if everyone started to inspect how sausages are made.

Flaws are a part of any tech, software or hardware.

[NMS]

Let's just stop. I honestly don't want to hear about ANY other flaw. Because at this point I don't think it can get any worse (yes it can)...

[Mira Yurizaki]

All of these features were conceived and developed anywhere from 15-20+ years ago when security threats basically boiled down to viruses and worms. I know that Hyperthreading at least had a report about its security concerns back in the mid 2000s, but I'm not sure if speculative execution had the same treatment. Though at the time it was just that: a threat on paper.

 

There's also the fact that some threats are not really the result of the idea of whatever feature it is being implemented, but some unexpected physical phenomena at play. Like that one time security researchers managed to grab an RSA key based on listening to the computer.

[captain_to_fire]

4 hours ago, NonaHexa said:

he Register reported on Friday that researchers at Vrije Universiteit Amsterdam in the Netherlands have found a new side-channel vulnerability on hyperthreaded processors that's been dubbed TLBleed

Do they call it variant 5 now?

 

[BuckGup]

This is sorta a big issue because a lot of programmers now days are just copying things from stack overflow and so on. Most don’t 100% know everything they are doing as it would take an immense amount of time. 

[LAwLz]

5 hours ago, Ben Quigley said:

I wonder how/if this effects SMT on AMD CPUs? 

It does. 

The developers disabled it on Intel processors now, and are working on disabling it on AMD processors right now. 

 

It's worth noting that:

1) This is an optional switch. OpenBSD is extremely safe on default settings and doesn't take any risks. Users can enable it again. 

2) This is implemented because apparently some motherboards don't let users disable hyperthreading in BIOS. So the OpenBSD developers want to give users that choice. 

[Christophe Corazza]

Am I to understand that every single performance enchancement made by Intel in the last 20 (?) years is flawed and prone to disaster-bugs?

TLBs, speculative execution, hyperthreading

[Mira Yurizaki]

1 hour ago, Christophe Corazza said:

Am I to understand that every single performance enchancement made by Intel in the last 20 (?) years is flawed and prone to disaster-bugs?

TLBs, speculative execution, hyperthreading

Intel isn't responsible for all of those, and multiple companies have used those.

 

The problem is that none of them were developed with security in mind by design because either the threat was theoretical and not considered in the realm of possibility for a foresable future or not considered because either because it's an extreme corner case or us fleshbags can't think of that many ways to break a computer.

 

It's worth noting that simultaneous multithreading was a thing IBM researched in the mid 60s and was first implemented commercially in an DEC Alpha processor.

[Bananasplit_00]

I... I just don't care enough anymore. I feel like I should just go back to my moms K6 laptop because that at least won't be targeted 

[Arika]

5 hours ago, Christophe Corazza said:

Am I to understand that every single performance enchancement made by Intel in the last 20 (?) years is flawed and prone to disaster-bugs?

TLBs, speculative execution, hyperthreading

With how fast technology is progressing, anything done by any company in the last 20 years is flawed and prone to disaster-bugs

[vorticalbox]

15 hours ago, BuckGup said:

This is sorta a big issue because a lot of programmers now days are just copying things from stack overflow and so on. Most don’t 100% know everything they are doing as it would take an immense amount of time. 

Only an incompetent programmer copies code without first understanding how the code works.

[LAwLz]

17 hours ago, BuckGup said:

This is sorta a big issue because a lot of programmers now days are just copying things from stack overflow and so on. Most don’t 100% know everything they are doing as it would take an immense amount of time. 

This has nothing to do with that. This is another variant of speculative execution that uses SMT as the attack vector. It's not some that happens because developers are copying code.

 

1 hour ago, vorticalbox said:

Only an incompetent programmer copies code without first understanding how the code works.

By that logic, everyone except asm developers are incompetent.

It is perfectly fine to not understand all of the code your program uses. Development would be waaaaay too time consuming otherwise. It's the entire reason we have libraries.

[vorticalbox]

3 hours ago, LAwLz said:

This has nothing to do with that. This is another variant of speculative execution that uses SMT as the attack vector. It's not some that happens because developers are copying code.

 

By that logic, everyone except asm developers are incompetent.

It is perfectly fine to not understand all of the code your program uses. Development would be waaaaay too time consuming otherwise. It's the entire reason we have libraries.

I think dropping in code you don't understand is far worse, at to be clear I'm not talking about frameworks or modules I'm talking taking a snip of code and adding to a project.

[ScratchCat]

55 minutes ago, vorticalbox said:

I think dropping in code you don't understand is far worse, at to be clear I'm not talking about frameworks or modules I'm talking taking a snip of code and adding to a project.

However most Stack Overflow posts have been written by competent people and have been scrutinized by peers, most comments result in an editing of the code to improve it.