Memory management bug in Intel CPUs threatens massive performance hits.

[Ampera]

https://www.theregister.co.uk/2018/01/02/intel_cpu_design_flaw/

Lovely for people like me who run an i7-4790k.

Curious to know what this crippling big actually is, but from what was described, I'm about ready to join with the rest of the Intel users here in giving Intel a collective backhand slap to the head.

So the question is now what sort of performance hits could we be seeing...

[stj]

it's a memory management "bug" that should stop regular code seeing kernel workings - in simple terms.

if fixed in the way they are sugesting the hit will be absolutly huge.
they are talking about flushing the cpu cache every time a user-space thread makes a system call.

keep in mind the cache is what makes a difference between a celeron and a xeon!!!

[bd139]

This is a pretty big one and they can't fix it with microcode either.

Interested to see real world load changes before everyone shits the bed however. Either way it's going to cost us a percentage more, particularly on AWS.

Intel CEO knew something was going down as well: https://www.fool.com/investing/2017/12/19/intels-ceo-just-sold-a-lot-of-stock.aspx

[Jeroen3]

As fix, they reset the Translation lookaside buffer each context switch.

The linux devs called it FUCKWIT 

[wraper]

As fix, they reset the Translation lookaside buffer each context switch.

The linux devs called it FUCKWIT 

Forcefully Unmap Complete Kernel With Interrupt Trampolines

[Avacee]

Short-term can expect Intel's share price to drop - especially when the class action law suits start.
But I can't help but wonder if Intel's share price will go up in the mid-term as lots of people replace/upgrade their CPU's :p

[dr.diesel]

Looks like a fix might take a branch prediction rework.  If so, not trivial, and if a fix wasn't already in the works, something like this could take quite some time to fix.

[bd139]

Looks like it has been exploited already to read kernel memory:

https://twitter.com/brainsmoke/status/948561799875502080

Shit is now officially fucked.

[JoeO]

And to think that just today I was reading a post here on the EEVBLOG about how great Intel's processors are compared to AMD's.

[bd139]

They're both as shit as each other. AMD has had a fair number of problems too. If that makes you feel better

Really this has been on the cards for a number of years. All current Intel (and AMD) CPUs are pretty much emulators. They are actually a crazy hyper-pipelined RISC microcoded virtual machines that happens to run x86 and x86-64 instructions. The problem here is that most of the bugs you can fix by changing the virtual machine implementation (microcode) but this one is actually in the physical virtual machine implementation at the bottom of the pile of turds. They hire hoards of design verification engineers to make sure that there are no holes in the native and virtual execution environments but this one slipped through. Actually quite a few have slipped through causing everything from random process crashes to big security holes like ASLR bypass.

And this is what happens people when you layer abstractions so deep and so complicated that you require several volumes of books to just explain the ISA and to maintain backwards compatibility to what is fundamentally some crack smoke inspired architecture from the late 1970s.

I hope the hell POWER wins some fans out of this.

[Ampera]

It's still a case of sit back and see how bad shit gets. Hope for the best, expect the worst.

[dr.diesel]

Ars just posted an excellent write-up:

https://arstechnica.com/gadgets/2018/01/whats-behind-the-intel-design-flaw-forcing-numerous-patches/

Might also include ARM.

[bd139]

To be clear the problem doesn’t affect ARM as far as anyone knows but the architectural change in Linux is being applied as a “defence in depth” strategy.

[dr.diesel]

To be clear the problem doesn’t affect ARM as far as anyone knows but the architectural change in Linux is being applied as a “defence in depth” strategy.

AMD patch already submitted:

https://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git/commit/?h=x86/pti&id=694d99d40972f12e59a3696effee8a376b79d7c8

[Mr. Scram]

This is a pretty big one and they can't fix it with microcode either.

Interested to see real world load changes before everyone shits the bed however. Either way it's going to cost us a percentage more, particularly on AWS.

Intel CEO knew something was going down as well: https://www.fool.com/investing/2017/12/19/intels-ceo-just-sold-a-lot-of-stock.aspx

He wouldn't be that stupid, right? That's how you get torn apart by investigators or even go to jail.

[bd139]

To be clear the problem doesn’t affect ARM as far as anyone knows but the architectural change in Linux is being applied as a “defence in depth” strategy.

AMD patch already submitted:

https://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git/commit/?h=x86/pti&id=694d99d40972f12e59a3696effee8a376b79d7c8

That patch is why AMD ticker shot up today. Whoever fought that deserves a bonus.

This is a pretty big one and they can't fix it with microcode either.

Interested to see real world load changes before everyone shits the bed however. Either way it's going to cost us a percentage more, particularly on AWS.

Intel CEO knew something was going down as well: https://www.fool.com/investing/2017/12/19/intels-ceo-just-sold-a-lot-of-stock.aspx

He wouldn't be that stupid, right? That's how you get torn apart by investigators or even go to jail.

When it comes to CXO level staff, asking for forgiveness gets you a long way.

[Lightages]

So do I start shopping for a Threadripper right now? Do I disable W7 updates until I get something that isn't going to pulled back to 2010 levels of performance?

[bd139]

I would sit down and do nothing for now and see what happens. Most of the embargoes are only lifted tomorrow with patches as well so time will tell.

[Gyro]

This is a pretty big one and they can't fix it with microcode either.

Interested to see real world load changes before everyone shits the bed however. Either way it's going to cost us a percentage more, particularly on AWS.

Intel CEO knew something was going down as well: https://www.fool.com/investing/2017/12/19/intels-ceo-just-sold-a-lot-of-stock.aspx

He wouldn't be that stupid, right? That's how you get torn apart by investigators or even go to jail.

Combined with this quote from the link that dr.diesel posted...

Microsoft has been testing the Windows updates in the Insider program since November,

It does look dangerously close to insider trading. 

[Ampera]

I am hearing anecdotal claims that the effect isn't as bad in 3D workloads as claimed, but it's still yet to be seen.

I don't think it will affect the consumer or even generic power user as much as people who work with hypervizors.

It's definitely a dancing day for AMD, though. With AMD back in the game, who knows if this is going to sink Intel's 5-6 year strong lead.

[Mr. Scram]

I am hearing anecdotal claims that the effect isn't as bad in 3D workloads as claimed, but it's still yet to be seen.

I don't think it will affect the consumer or even generic power user as much as people who work with hypervizors.

It's definitely a dancing day for AMD, though. With AMD back in the game, who knows if this is going to sink Intel's 5-6 year strong lead.

AMD was lagging a single digit percentage in performance, but if these percentages turn out to be correct AMD might very well lead by the same margin. I loathe to think what discussions this will cause amongst the fanboys on either side.

[tszaboo]

Are you people crazy? It affects Virtual machines that can read from each other. It only affects you, if you are running more than 1 virtual machines on your PC server, and one would run malicious code, specifically designed to attack the other virtual machine. This is only an issue for cloud providers.
99.9999% of PC users are not affected.

[Monkeh]

Are you people crazy? It affects Virtual machines that can read from each other. It only affects you, if you are running more than 1 virtual machines on your PC server, and one would run malicious code, specifically designed to attack the other virtual machine. This is only an issue for cloud providers.
99.9999% of PC users are not affected.

.. no, no, that isn't it.

This is an issue which can potentially allow an unprivileged user-mode process to read kernel memory.

[pigrew]

Do VM hypervisors normally allow multiple VMs to execute simultaneously (by dividing up cores)?

[Monkeh]

Do VM hypervisors normally allow multiple VMs to execute simultaneously (by dividing up cores)?

Sure. Or you'd have a VM with one core assigned blocking the whole shebang.