3 new extreme 'L1TF' exploits found on Intel

>vulnerabilities don't matter!

Everytime this happens, the web is totally full of """executive summaries""" of these vulnerabilities and it's totally impossible to find actual technical details. Does anyone know what it's actually about?

Actually, I take that back. The RedHat explanation was actually technical enough to describe what's actually going on. Thanks for linking to it, senpai!

Literally who cares

Anyone running untrusted VMs would care very much.

I'll start with the logo.

>you actually need to fully disable hyper-threading to not be vulnerable to L1TF lvl 3 on your server
I don't think that's the case. Hypervisors/kernels just need to ensure that sibling threads aren't used to run VMs/processes that don't trust each other. Which quite arguably they should be doing anyway.

given it's yet another, intel specific exploit that requires disabling SMT and flushing L1 cache periodically or going back to trapping vm paging operations, anyone who depends on VMs

This parade of exploits is going to permanently assfuck Intel out of the hypervisor market sooner or later. I can't wait.

DELET

Attached: 1525554923397.jpg (691x771, 112K)

My machine works fine and my virtual machines work fine. It's fucking nothing.

t. Pajeet Goldstein

that's not the point idiot. this is literally a meltdown tier exploit that lets you fucking read memory in l1 cache because the speculative executor is too fucking lazy to fucking read one stupid ass bit on a page

But meltdown was never even used and even people who didn't update with any of the patches for it are fine.

"Intel always blah blah" published 90 days after no response from the careless money lovers!

DELID DIS

Attached: 1527629778452.jpg (679x758, 54K)

Every competent intelligence agencies in the world are milking this for all it's worth, don't even kid yourself. They definitely have dedicated teams for these exploits.

100% guaranteed.

hello, AMDshill, how do you do?

Can you provide a source so it doesn't look like you're talking out of your ass?

CIA Vault 7 leaks confirmed they stockpile 0days and pressure companies to add backdoors. That's where all these "exploits" are coming from - targeted leaks to crush Intel.

Can Intel chips even add 2+2 properly?

Attached: what-is-good.jpg (480x270, 27K)

Not if you fuck with the speculative execution lol.

Frankly, if they haven't been exploiting it, they aren't doing their jobs.

>no 10 nm for more than another year, and even then it's worse than expected
>suddenly, vulnerabilities left and right, particularly critical for cloud virtualization, one of Intel's biggest cash cows
>amd returns to excellence with server cpus specifically aimed at cloud virtualization
It really is the perfect storm for Intel, isn't it?

Attached: intel-2.png (660x421, 473K)

>the ride never ends

Attached: Screenshot_20180814_225638.png (1400x730, 871K)

>5 to 15% in performance hit for servers running VMs (everyone)
>already had a 5 to 30% hit from spectre/meltdown

Compounding those two together, worst case you're looking at a 40.5% total performance hit from pre-Meltdown.
>Intel chips now half as fast
>Linux (and probably ESXi) twice as fast as Windows on TR2/EBYN due to NUMA and sheer core count
Wintel is literally finished.

no

intel lmao

> tfw hyperthreading axed from i7 line finally starts making sense

Attached: andrei.gif (304x256, 1.82M)

the people who knew/know about spectre and still buy/bought current intel are absolute retards

Fuck's sake!

>Side-channel attacks
>Speculative Execution
There's nothing new here, and it'll work on everything non-intel as well once you calibrate it.

>There's nothing new here, and it'll work on everything non-intel as well once you calibrate it.

most of these sidechannels are due to a pervasive design philosophy Intel maintained of aggressively following speculation and waiting until the last possible moment to do validation checks. other chip makers (but we'll just assume we're discussing x86 and thus only AMD here) have not gotten hit nearly as badly, and it has little to do with the attacks not be "calibrated" to them yet.

AMD engineers explicitly said "we are not vulnerable to this" when submitting code for a Meltdown patch to the Linux kernel, and suggested defining a CPU_INSECURE flag if the chip was Intel at all. In retrospect they were right.

As opposed to buying amd which is still also affected by spectre?
The only hardware methods to prevent side-channel memory timing attacks is to insert enable buffers at the caches and/or TLB that effectively temporarily disables a memory region that was loaded in from a execution violation, or to insert a tapered semi-random delay for memory access after an execution violation.

>most of these sidechannels are due to a pervasive design philosophy Intel maintained of aggressively following speculation and waiting until the last possible moment to do validation checks.
No. Most of these side channel attacks have nothing to do with that and will work on a majority of chip maker's cards. A small minority of the side channel attacks make use of Intel specific technologies or features; of course those won't affect others because it's literally tailor-made for the architecture. Any speculative execution or memory-sharing feature on other boards will also have their own attacks specific to that technology. The reason you don't "see" as many of those is because you have to spend your time targeting and testing that specific architectural feature and less people are using non-intel as a baseline.

>last possible moment to do validation checks
You have absolutely no idea how the fetch-decode-execute cycle works, do you? This is entirely the fault of the OS, not the hardware.

>all these lies and damage control
Isn't it pretty late in Tel Aviv?

Attached: intel.jpg (1845x1923, 1.25M)

So basically all the intended NSA & Israel backdoors are getting exposed by some vigilante?

The absolute state of x86. When will you give up on this shit?

Daily reminder that hating popular things doesn't make you an interesting person.

>turns out building in Mossad backdoors also leaves you vulnerable for other attackers.

Attached: Intel Israel 1976.jpg (800x495, 249K)

Clearly the only reason to hate x86 is popularity, even if the most popular processors are ARM.

>Lies
If you're not an ignorant twat, you're in denial.

>Damage control
I'm not claiming that Intel is safe. I'm pointing out that everyone gets fucked by these. I cannot fathom how people ignored Spectre and tunnel visioned on Meltdown when the public found out about all these vulnerabilities. You think amd-specific performance enhancement architectures related to speculative execution and memory management are safe? You think any of this is new? This shit has been possible for years. Anyone in the same room as your computer could use the mic on their fucking cellphone to break any (most?) encryption happening on your computer in real time. If you don't have a faraday cage and a cardboard sound muffler on your laptop it's easily more vulnerable than the difference between having an intel chip and an amd chip would be.

To continue >You think any of this is new? This shit has been possible for years.
Have some papers and sources to read up on.
cs.unc.edu/~reiter/papers/2012/CCS.pdf >Cross-VM Side Channels and Their Use to Extract Private Keys
tau.ac.il/~tromer/acoustic/ cs.tau.ac.il/~tromer/papers/acoustic-20131218.pdf >RSA Key Extraction via Low-Bandwidth Acoustic Cryptanalysis
The Program Counter Security Model: Automatic Detection and Removal of Control-Flow Side Channel Attacks - USENIX
Defending against side-channel attacks - Gilbert Goodwill, Cryptography Research, Inc
spectrum.ieee.org/computing/it/encryption-chip-fights-off-sneak-attacks
Timing Attacks on Implementations of Diffie-Hellman, RSA, DSS, and Other Systems - Paul C. Kocher

If you want to be paranoid about Intel being unsafe due to side channel attacks, have fun with this instead: Lookup "hardware trojans" and then worry about if Intel, AMD, Nvidia, or even just any of the manufacturers who actually fabricated the chips have sneaked some in.

so does your english teacher

is core 2 duo, too?

What is Intel's next realistic step?

Crashing this industry, WITH NO SECURITY

Same goes for TLBleed. Even if you are magically not affected by L1TF V3, you will STILL be affected by TLBleed.

Bring back Itanium

It's a feature!

Release chips with all of the affected components disabled as a stopgap solution until like ~2020 when they release chips with the proper fixes implemented. It's gonna be a rocky couple of years for Intel.

kek