Initial benchmarks for Linux's performance impacting security update are out

it is most likely that developers are taking the "better safe than sorry" approach to security. AMD should be able to demonstrate that their products are unaffected by the bug and should be able to get the kernel patched with exceptions in the near future, leaving this mess solely at Intel's feet.

they (at least microsoft) work on this patch since october.

they should be able to confirm if AMD is affected or not... :mellow:
 
Linux are just being cautious with the inclusion of ALL x86 processors being tagged as insecure. We all know the flaw in Intels and there are yet to be any confirmation that AMD is affected. Still I think it is right to err on the side of caution just in case
 
new patch arrived already here it is: :)
https://patchwork.kernel.org/patch/10133447/
hopefully the OP can update the article

I have updated the article to link the patch and the known workaround. The problem is that the patch doesn't seem to be merged with the mainline Kernel.

Latest patch notes:
https://cdn.kernel.org/pub/linux/kernel/v4.x/ChangeLog-4.14.11

Many x86 CPUs leak information to user space due to missing isolation of
user space and kernel space page tables. There are many well documented
ways to exploit that.

The upcoming software migitation of isolating the user and kernel space
page tables needs a misfeature flag so code can be made runtime
conditional.

Add the BUG bits which indicates that the CPU is affected and add a feature
bit which indicates that the software migitation is enabled.

Assume for now that _ALL_ x86 CPUs are affected by this. Exceptions can be
made later.
 
With a drop in performance of up to 30% for Intel CPUs, that's going to make AMD even more attractive, and in some cases a complete no brainer.
 
Spoke to my mate about this. He reckons it won't affect gaming *but* will absolutely destroy things like Handbrake. Looking forward to seeing it tested, and would still like to see gaming benchmarks before and after the Windows patch *pokes Mark* :D

Yeah this is crazy though man. No wonder that dude from Intel sold his shares lol.
 
Spoke to my mate about this. He reckons it won't affect gaming *but* will absolutely destroy things like Handbrake. Looking forward to seeing it tested, and would still like to see gaming benchmarks before and after the Windows patch *pokes Mark* :D

Yeah this is crazy though man. No wonder that dude from Intel sold his shares lol.

Some gaming tests already done and show little to no impact for gaming
 
Given the apparent gravity of this design flaw this performance hit will likely be a price that users are willing to pay in the name of security.

No, they are not willing but simply lacking a choice.
Time will tell if this affects AMD or not.

This next question will be where to pin the blame.
Currently looking like Intel, but that is very dependant on whether the problem is Intel exclusive or not.
Last time it was a flaw in a useful feature, though a flaw Intel were aware of for a long time, and didn't care about fixing.
I am not seeing anything to this but simply a security risk.
 
Last edited:
The problem is Intel exclusive, it is a flaw in the design. There is an area of the CPU that allows direct access to the kernel. This means hackers can basically walk straight in.

I explained how you fix it. Well, no, I didn't, Gareth Halfacree did with the "You can't turn left" theory. IE - Intel need to patch the kernel so that it uses a different path than the one it does now, meaning it is much slower because it has to go the long way. IE - try and plan in your head getting to a destination without turning left..

They will only fix it if and when they completely change the technology. Obviously Jim Keller did not make this "mistake" nor did he do it possibly just to boost performance and make the competition look bad. That is one question we will never know the answer to, but yes, until Intel change their tech (and not just a shrink) the "error" or "flaw" will continue to exist.

It's a flaw with the entire architecture. Now imagine having to explain to a huge data centre that their productivity is about to nose dive by 30% pmsl. I shouldn't laugh, but yeah, cue Epyc....
 
The problem is Intel exclusive, it is a flaw in the design. There is an area of the CPU that allows direct access to the kernel. This means hackers can basically walk straight in.

I explained how you fix it. Well, no, I didn't, Gareth Halfacree did with the "You can't turn left" theory. IE - Intel need to patch the kernel so that it uses a different path than the one it does now, meaning it is much slower because it has to go the long way. IE - try and plan in your head getting to a destination without turning left..

They will only fix it if and when they completely change the technology. Obviously Jim Keller did not make this "mistake" nor did he do it possibly just to boost performance and make the competition look bad. That is one question we will never know the answer to, but yes, until Intel change their tech (and not just a shrink) the "error" or "flaw" will continue to exist.

It's a flaw with the entire architecture. Now imagine having to explain to a huge data centre that their productivity is about to nose dive by 30% pmsl. I shouldn't laugh, but yeah, cue Epyc....

Summed up nicely
 
Well the bench in the main article says h264 doesn't take a hit.
And again. Everywhere they mention x86, so is it even a problem in x64?
 
Back
Top