'Kernel memory leaking' Intel processor -- a serious cpu bug!?!

[fluffy_little_something]

"Microsoft is now preventing AMD PCs from receiving updates, to stop machines from getting into an unbootable state."

https://www.theverge.com/2018/1/9/16867 ... pcs-issues

[ghettosynth]

"Microsoft is now preventing AMD PCs from receiving updates, to stop machines from getting into an unbootable state."

https://www.theverge.com/2018/1/9/16867 ... pcs-issues

Good thing that it wasn't rushed out or anything.

[aciddose]

Meanwhile, pathetic attempt at misdirection aside

Awww bless, did I hurt your feelings. Come here

It's an attempt by Microsoft though. Obviously any business has its bottom line in mind whenever it says anything at all.

Re: rushed patches.

It wasn't necessarily "rushed", but rather they recently laid off the large portion of their QA staff. You know, like the sort of people running 100s of different PCs in various configurations to test the installation for problems like:



Or this recent issue with "something happening".

You can bet that Microsoft has fully capable QA staff on hand and all the hardware they need to ensure new drivers and kernels operate correctly on supported hardware. So the only thing that can explain this happening is that they didn't test it for QA at all.

[ghettosynth]

Cmon man you're sucking all of the air out of my funneh!

[aciddose]

The sad thing about win10 is Microsoft said themselves (I haven't found the quote yet) that officially all "home" and "business" licenses of win10 are equivalent to "beta testers" for their "enterprise" customers. So you can expect patches to be rolled out without any testing whatsoever because the one responsible for that testing is you, the "not actually a valuable customer, rather just a lab rat."

This is why like a rat, you can't choose whether or when to have your gonads removed or an ear stitched to your back.

[Zombie Queen]

Microsoft says older Windows versions will face greatest performance hits after Meltdown, Spectre patches. Windows 7 and Windows 8 users will notice the greatest decrease in system performance after the processor patches are applied

Says Microsoft... I wouldn't be surprised if they packed-in some deliberate slowdowns.

[Daags]

Microsoft says older Windows versions will face greatest performance hits after Meltdown, Spectre patches. Windows 7 and Windows 8 users will notice the greatest decrease in system performance after the processor patches are applied

conspiracy confirmed. intel and microsoft in cahoots to strong arm people from their beloved win 7 setups, into the telemetry & forced update nightmare that is win 10.

[fluffy_little_something]

Thought the same thing, just another opportunity for MS to force W10 on everyone.

[Touch The Universe]

Let me get this straight, this bug has been going on for 10 plus years and they only found out about it now?

[fluffy_little_something]

Who says it is a bug, maybe it has been a secret backdoor for secret services from the start...

[mcbpete]

Let me get this straight, this bug has been going on for 10 plus years and they only found out about it now?

There's been longer unspotted bugs in the wild - http://www.osnews.com/story/19731/The-2 ... d-UNIX-Bug

[ghettosynth]

Let me get this straight, this bug has been going on for 10 plus years and they only found out about it now?

The exploit was only recently discovered by researchers and this isn't an everyday kind of trivial exploit. However, you would be surprised at how long bugs can exist in normal code before they are discovered. That is, you can find many bugs in open source code where the defect origin in the source is retrospectively determined to be many years old. Finding bugs that can be traced back to the earliest versions of a system is not uncommon and there are more than a few systems that are more than a decade old.

To a certain extent, intel has a point, it's not a "bug" in the sense that we usually think of a bug. If your code isn't maliciously exploiting the system then the system does its job as expected. You cannot simply read data using conventional means, if you try, the system simply won't let you.

Complaining that this is a bug is a bit like complaining that a lock doesn't work when you use a credit card to jimmy the door. Sure, it's a weaker lock than it could be, but the door is still locked and if you just try to use the handle to get in it won't work. That's not a defense of Intel, per se. But, I can see their frustration with this.

From a probabilistic point of view, this exploit affecting you is probably like being killed by a terrorist, you really should worry more about your drive home tonight on the interstate.

[PurpleSunray]

Who says it is a bug, maybe it has been a secret backdoor for secret services from the start...

That would be pretty dump.
Why is this "backdoor" on Intel only?
Why so unreliable? It's easy to access memory that belongs to an other process, however it is difficult to actually ready the data you interested on. I don't know of any actual attack that was able to do that up not now, it is all just "in theory you could ..".

Also, looking at the problem itself it seems more like a design issue / some architect forgot to think about, rather then a "backdoor feature" that was explictily introduced.
Just read about the details of problem - it is very good example of a major design flaw that is not easily visible because of the complexity of the system. Every softwae dev that does multi-thread programming knows what I'm talking about. The special with this issue is that Intel introduced it as part of the hardware, so effects a lot of systems and is hard to fix.

If you are looking for NSA backdoors in Intel's, look at all the issues arround Intel Managment Engine.
Reading about it, you could easily come to the conclusion that all ME is about, is to allow 3rd-parties to access your system from remote via hidden/secret "bugs".

[jdnz]

since we have a LOT of older i7 machines in use at my work (up to 5 years old) and most of the older ones are running win7 I was somewhat alarmed by microsoft's warning that 'cpus prior to 2015 models and especially those running win7 will be most affected by slowdowns due to the kernel change' (with them citing win7/win8 kernel based font rendered affecting overall UI performance)

So rather than believe what you read online, I grabbed a desktop with an i7-3770 (so no PCID support) which is about the oldest build we have in active use here and ran PCmark10 on it before the kernel update was applied and again after (having verified that the patch was working and that KPTI was active using the powershell)

since the machine is very low spec'd (including a lowly gt430 video card!) I expected the worst - however the 2 runs showed no significant changes in either the PCmark10 'productivity' score (which would be the one most affected by the font rendering issue), nor in the other scores (the 'essentials' suite scored 2.6% lower post patch, but 'productivity' went UP 2.8% post patch and 'digital content creation' went up 7.6% though I suspect I'm getting weird reading on the DCC score as the video card has issues running it to completion)

the test machine is too old to receive an updated BIOS with the new intel microcode so there was NO spectre mitigation active (just the meltdown fix), I'll hunt out a newer machine with win7 on and re-run the tests with both the spectre and meltdown fixes active next

[jdnz]

since the machine is very low spec'd (including a lowly gt430 video card!) I expected the worst - however the 2 runs showed no significant changes in either the PCmark10 'productivity' score (which would be the one most affected by the font rendering issue), nor in the other scores (the 'essentials' suite scored 2.6% lower post patch, but 'productivity' went UP 2.8% post patch and 'digital content creation' went up 7.6% though I suspect I'm getting weird reading on the DCC score as the video card has issues running it to completion)

and here's the results

https://www.3dmark.com/compare/pcm10b/1 ... 10b/178119#

if you ignore the erroneous digital content scores a quick analysis of the detailed result figures shows basically a 3-4% slow down post patch which is no worse than people are reporting on current cpus running win10