KPTI bug figured out

Well, this is the end.
twitter.com/brainsmoke/status/948561799875502080
Apparently it's caused by branch prediction on not-taken branches.

Other urls found in this thread:

digitaljournal.com/tech-and-science/technology/intel-processor-flaw-lets-hackers-bypass-security-mechanisms/article/477620
stackoverflow.com/a/11227902
security.googleblog.com/2018/01/todays-cpu-vulnerability-what-you-need.html
forbes.com/sites/thomasbrewster/2018/01/03/intel-meltdown-spectre-vulnerabilities-leave-millions-open-to-cyber-attack/
lewd.pics/p/?i=YsFT.mp4
cnbc.com/2018/01/03/intel-ceo-brian-krzanich-on-security-flaw.html
twitter.com/SFWRedditImages

Man I wish I knew what the fuck you are taking about.

This was implied with the statement from AMD about whether it affected them or not.

>AMD processors are not subject to the types of attacks that the kernel page table isolation feature protects against. The AMD microarchitecture does not allow memory references, including speculative references, that access higher privileged data when running in a lesser privileged mode when that access would result in a page fault.

explain this my friend :)

Can somebody explain this gibberish please?

It allows ring-3-level user code to read ring-0-level kernel data.

tyvm

Is this the same as the bug in this article from 2016?

digitaljournal.com/tech-and-science/technology/intel-processor-flaw-lets-hackers-bypass-security-mechanisms/article/477620

stackoverflow.com/a/11227902

What exploits does this enable?
I've read about the risks for cpus hosting multiple vms where memory can be read and written from one to another.
What's the risk for the layman?

English, doc.

What about negative rings

Imagine a rootkit that you cannot really patch because it's present in the actual hardware. That's what the flaw basically is. If software knows about the problem it can use it to instantly bypass every OS level security feature and effectively become root and then do whatever the fuck it wants to wreck your shit.

God why do we even have computers fuck this

How can a piece of software become root solely by reading kernel memory?

what does this mean? can Intel hardware bug be fixed without penalty loss?

Aslr bypass i think

If we have so much fucking access now why can't we just make a simple tool to remove Intel ME on every single Intel CPU

Yeah I don't know what the fuck am I talking about but still

H-how can this exploit be delivered?

Anything on your computer can read your secret data like passwords, etc.

Remember b9s for the 3ds its like that. It cant be fixed unless you get new hardware wuth it fixed. Its in silicon

LOL

Interesting read. I've been thinking for a long time, is there no way to branch out code without if statements?

It's not the flaw itself but what the flaw allows for. It can read and write at the kernel's privilege level which allows anyone to do anything the root user can do. It's not quite as intuitive as an actual "root" account but it can do all of the same things given enough time and effort.

Worst case anything running on the CPU can make use of the flaw, something as simple as javascript that happens to be interpreted the "correct" way can cause your system to be compromised.

It affects AMD too.

What's a computer?

The Windows insider preview seems to just show a 0.5-5% performance loss in gaming and most applications. Not sure about SQL and stuff yet.

5% performance loss in gaming is major to some people, though. Often Intel is only 5% faster.

Uh sure, like switches, but those still use the branch predictor.

I think the exploit is read-only, it allows reading from kernel memory, not modifying it.

A woman who likes doing math.

nope

so it has nothing to do with the kernel issues that macos and windows patched out a month ago then?

How would you take advantage of this? A program with a bazillionfuck branches that attempts to read kernel addresses on the branches that are never executed to avoid segfaulting and then somehow reading them anyways?

>cat
Why is he using cat in this case? Why should I trust a literal retard?

then why are most people today, including Google who claim to have discovered this last year, saying AMD is affected?

security.googleblog.com/2018/01/todays-cpu-vulnerability-what-you-need.html

>These vulnerabilities affect many CPUs, including those from AMD, ARM, and Intel, as well as the devices and operating systems running them.

Leave

> Often Intel is only 5% faster
keep on dreaming AMD poorshit

>switches
What, the switch-case thing? I was thinking more of using function pointers.

Nope,

the current fixes affect AMD because loonix kernel guys have just applied the fix to all x86.

You're wrong.

AMD x86/64 is apparently not affected, but some niche AMD processors are.

so its a scam by Intel, as noone cares about some custom 0.0001% of total AMD cpus

>How can a piece of software become root solely by reading kernel memory?
read memory where root password is
login as root

Because they're being paid by intel to shut it down. You do know intel is the same company that literally spent billions of dollars to get OEMs to not buy AMD CPUs even for free, right?

Intel: Say that also is AMD and we give you súper discount, and a house
Goigle mánager: ok after all is just a couple of words

did you even read the Google blog? there are multiple attack methods here, AMD x64 is affected by some, fuck off shill

They don't explain in any respectable detail how AMD and ARM are affected

I would love to see your sources factually backing up that statement.

>hahaha I spent more money on my hardware rootkit you fucking poorfag!

No, its actually worse than this. There is no rootkit because rootkits on intel processes are redundant now. Why bother with a rootkit when regular javascript code can just pluck sensitive stuff out protected kernel memory. In order to stop that you are going to need to take a 30% performance hit on every system call. It doesn't mean much for the home user but Amazon, Microsoft and all the other cloud service people are just looking at the data centers knowing that every xenon CPU they spent millions on will get 30% slower when the software patches hit.

Lisa Su can basically put in order for a gold plated super yacht now. The only limit to AMD's server CPU sales is how many they can make.

Did you make this one? Because out of all of them this one really shines like brain cancer on an MRI scan. I love it, saving it rn.

t. butthurt Google employee who went with Intel instead of AMD for the Google Cloud Platform server upgrade

>doesn't mean much
Excuse you, dickhead. I paid cash money for this processor.

AMD does also make ARM processors using reference cores, maybe they're talking about those?

>Meltdown wasn't the only problem uncovered by the researchers, however. They detailed a related issue dubbed Spectre, which they believe is harder to address than Meltdown and for which there aren't yet patches available. As noted in a whitepaper, which contains the full technical details, Spectre attacks induce a victim application to carry out the speculative execution "that would not occur during correct program execution and which leak the victim’s confidential information via a side channel to the adversary." Worryingly, it's not just Intel systems that are affected, but computers running AMD and ARM too, the researchers claimed. That would amount to not millions, but billions of machines.
forbes.com/sites/thomasbrewster/2018/01/03/intel-meltdown-spectre-vulnerabilities-leave-millions-open-to-cyber-attack/

>custom x86-64s are shit
No fucking shit you mong.

my cool video about this made with free software!

lewd.pics/p/?i=YsFT.mp4

Someone quickly check if Sandy Vaginas are affected by this.

...

of course they are lol

This is Intel's death sentence

Every Intel cpu

is sandry bridge 10 years old? if no, you're fuk

Thanks faggots.

Now I'm 100% certain this is an NSA backdoor.

Are you running a database or running a bunch of virtual machines? Those seem to be the use cases that are taking a 20-30% performance hit. I think people were reporting performance hits for gaming at under 10%.

Hillary just installed the patch on her server, and her house burned down.

If it's not 0%, then it's an objective loss.

>Sup Forums is only for kernel devs

How does it affect AMD CPUs? Don't they have this things that checks if code is supposed to run alongside protected code or something?

So if I write code that the branch predictor predicts will lead to data from kernel memory being loaded into a register it will pre-emptively load this data to an accessible location which can then be read without special privileges? And the error lies in the part that the predictor ignores memory protection? I only have generic CPU knowledge from microcontrollers so I'm unsure how registers are handled in x86

...

No that's a different but equally alarming intel flaw.

It wasn't for the IC, we know that for sure.

ARM64 designs are affected, and AMD happens to manufacture some. Doesn't affect AMD x64 arch.

what software you use to make it?

How to check if my shit is vulnerable>

If it's Intel it's vulnerable

Reminder Chinese CPU don't have branch predictions.

this is why mainframes use external cryptographic processors for ssh / OLTP

Does your processor say "Intel" on it and is it newer than a 486?

look up your processor manufacturer, if it has "Intel" as the manufacturer name, it is affected by the hardware security flaw

Pack it up folks, we're done here...

>He said he's "relatively confident" that a leaked security issue has not been exploited, and that the industry has been working together for a couple of months to address it.

CEO of intel says no one exploited their backdoor.
cnbc.com/2018/01/03/intel-ceo-brian-krzanich-on-security-flaw.html

fml. what do?

how would they know though? I thought that the hardware flaw allows anyone to bypass any security meassures, whether they are software or hardware security preventions

buy AMD Ryzen next time you are looking to buy new computer

Buy AMD or wait for a year for Intel to release the new CPUs that don't have this flaw, and give them more money despite fucking you in the ass.

>Intel Core i5 CPU 520
Poorfag, no h8.
fml. what do?

Intel just exploited everyone elses backdoors with their poorly engineered cock

>you seemed to have mistyped the captcha
>still posts
By then I'll be rich, so alright, thanks Sup Forumsuys!

It can be used to bypass ring 0 security measures but ring -1 measures like management engine can still observe, intercept, and disrupt those things.

So what are they going to do for my company that will have to replace devices in the field with all new hardware, because they can't afford a 30% performance hit? Surely Intel will be writing a check to cover all my device replacements?

There is nothing to disrupt. The kernel executes the read before anything else and dumps the kernel memory.
The CEO saying he's sure no one exploited it is akin to hillary clinton saying no one hacked her server,

Linkie detected, lmao keep dreaming you intelcuck.

>Atoms breaking shit?
>"Deal with it goyim, its not our problem"

So they'll just buy 20-30% more cpus to run the servers. What's the big deal?

nice kuumin

I was thinking confidential things like passwords can be encrypted using management engine which could disrupt password stealing malware making use of the CPU bug. I wonder if management engine is also susceptible to the hardware bug?

If you've spent $200m on processors thats another $40-$60m you need to spend to maintain similar performance. Consider what that does to profitability, cashflow and debt burden.

this

...

the bug has been there for a decade and noone took advantage. does this mean i can skip the windowns update and keep better performance?

and 20-30% more motherboards, ram, power supplies etc

they could buy epyc