Go Back  Bike Forums > The Lounge > Foo
Reload this Page >

Arrrrg. Massive processor flaw leaves both PCs and Macs vulnerable

Foo Off-Topic chit chat with no general subject.

Arrrrg. Massive processor flaw leaves both PCs and Macs vulnerable

Reply

Old 01-08-18, 09:07 AM
  #51  
himespau 
Senior Member
 
himespau's Avatar
 
Join Date: Jun 2008
Location: Louisville, KY
Posts: 10,857
Mentioned: 14 Post(s)
Tagged: 0 Thread(s)
Quoted: 1113 Post(s)
Originally Posted by UmneyDurak View Post
No, or micro opcode update. This is Kernel/Browser/firmware update stuff. Depending if it's Meltdown or Spectre.
So updating the BIOS won't do anything to fix it? My computer has been giving me notifications that there is an update available for the BIOS, but the last time I did that, I wasn't ever able to get the computer to restart and ended up having to get a new motherboard/CPU (everything else stayed the same and worked), so I'm a bit leery of doing any more BIOS updates - especially at the start of the semester on my work computer.
himespau is offline  
Reply With Quote
Old 01-08-18, 10:37 AM
  #52  
UmneyDurak
RacingBear
 
UmneyDurak's Avatar
 
Join Date: Dec 2004
Location: NorCal
Posts: 8,870
Mentioned: 2 Post(s)
Tagged: 0 Thread(s)
Quoted: 179 Post(s)
It took software to patch this long, why do you expect HW to be faster? That is even more involved. RTL freezes happen months in advance. This isn't just one bug there are three under two general terms of Spectre and Meltdown. Spectre effects ARM, AMD, Apple, Intel. Meltdown originally was reported to effect Intel, but now looks like might effect ARM also. Problem is it's fundamental issue with how modern processors work.
Some articles I read said there is firmware patch along side with kernel updates, others didn't mention it. Then there are others like this saying there are potential microopcodes update that help to mitigate: https://arstechnica.com/gadgets/2018...oing-about-it/
UmneyDurak is offline  
Reply With Quote
Old 01-08-18, 01:40 PM
  #53  
CliffordK
Senior Member
 
CliffordK's Avatar
 
Join Date: Nov 2014
Location: Eugene, Oregon, USA
Posts: 20,111
Mentioned: 124 Post(s)
Tagged: 0 Thread(s)
Quoted: 7077 Post(s)
I noticed these articles today.

https://www.fool.com/investing/2017/...-of-stock.aspx

Intel CEO's massive stock dump raises eyebrows - Jan. 4, 2018

https://arstechnica.com/information-...ses-questions/

Because of the way this flaw was released... 6 months ago to the businesses involved, and just recently to the public, this massive stock sale has to raise eyebrows for insider trading.

In all fairness, the value did bump up in late December, so the sale hasn't saved much yet, but those December gains have already been 100% wiped out.
CliffordK is offline  
Reply With Quote
Old 01-08-18, 03:19 PM
  #54  
Shimagnolo
Senior Member
 
Shimagnolo's Avatar
 
Join Date: May 2008
Location: Zang's Spur, CO
Posts: 8,569
Mentioned: 7 Post(s)
Tagged: 0 Thread(s)
Quoted: 1351 Post(s)
Apple released new updates today for iOS and MacOS
Shimagnolo is online now  
Reply With Quote
Old 01-09-18, 09:11 AM
  #55  
ptempel
Senior Member
 
Join Date: Feb 2005
Location: New Jersry the beautiful Garden State
Posts: 1,914

Bikes: 2007 Ridley Excalibur, 2003 Orbea Orca, 199? Cannondale Headshock MTB hardtail

Mentioned: 29 Post(s)
Tagged: 0 Thread(s)
Quoted: 514 Post(s)
Originally Posted by bigbenaugust View Post
I also have one Intel machine that is too old for this bug, I think.
Reminds me of an article talking about a guy that installed the Microsoft patch on a system with a 10 year old Athlon processor. The patch hosed up his Windows installaton:

https://betanews.com/2018/01/08/micr...ricks-amd-pcs/

Microsoft needs to add a sanity check on whether the processor is supported, I guess. And I thought that my Core Duo laptops were old. Sheesh.
ptempel is offline  
Reply With Quote
Old 01-09-18, 09:14 AM
  #56  
ptempel
Senior Member
 
Join Date: Feb 2005
Location: New Jersry the beautiful Garden State
Posts: 1,914

Bikes: 2007 Ridley Excalibur, 2003 Orbea Orca, 199? Cannondale Headshock MTB hardtail

Mentioned: 29 Post(s)
Tagged: 0 Thread(s)
Quoted: 514 Post(s)
Originally Posted by himespau View Post
So updating the BIOS won't do anything to fix it? My computer has been giving me notifications that there is an update available for the BIOS, but the last time I did that, I wasn't ever able to get the computer to restart and ended up having to get a new motherboard/CPU (everything else stayed the same and worked), so I'm a bit leery of doing any more BIOS updates - especially at the start of the semester on my work computer.
No, since one appears to be a problem in the memory management unit (MMU) logic. It read like Intel (and maybe some others) might need to revise their MMUs and come out with a new chip. Well, its on the processor nowadays so they have to cut a new die and manufacture a whole new processor.

This is a nice higher level explanation of the two exploits:

https://linuxaria.com/article/spectr...down-explained

Last edited by ptempel; 01-09-18 at 09:17 AM.
ptempel is offline  
Reply With Quote
Old 01-09-18, 08:18 PM
  #57  
Marcus_Ti 
Only Slightly Bent
 
Marcus_Ti's Avatar
 
Join Date: Sep 2014
Location: Lincoln, Nebraska
Posts: 4,017

Bikes: Roadie: Seven Axiom Race Ti w/Chorus 11s. CX/Adventure: Carver Gravel Grinder w/ Di2

Mentioned: 24 Post(s)
Tagged: 0 Thread(s)
Quoted: 1525 Post(s)
Appears the Spectre patch hosed the ancient AMD6400+ K9N Neo system running Win7x64 in my family.
Marcus_Ti is online now  
Reply With Quote
Old 01-09-18, 10:17 PM
  #58  
CliffordK
Senior Member
 
CliffordK's Avatar
 
Join Date: Nov 2014
Location: Eugene, Oregon, USA
Posts: 20,111
Mentioned: 124 Post(s)
Tagged: 0 Thread(s)
Quoted: 7077 Post(s)
Originally Posted by ptempel View Post
No, since one appears to be a problem in the memory management unit (MMU) logic. It read like Intel (and maybe some others) might need to revise their MMUs and come out with a new chip. Well, its on the processor nowadays so they have to cut a new die and manufacture a whole new processor.

This is a nice higher level explanation of the two exploits:

https://linuxaria.com/article/spectr...down-explained
Thanks for the link.

It is a little bizarre, retrieving memory out of the cache that isn't the program's own memory. One would think that would inherently generate errors. And perhaps it does. Maybe one has to quickly dump the dirty memory into a clean location, somewhat like laundering money.

I still think that this might be something that could be patched with BIOS. Although perhaps there are enough different versions of BIOS that it is easier to patch in the OS and Browsers. And there may be issues with running redundant patches.

Hopefully there will be some continuing support for the older versions of software. Last summer we had some major software hacks due to Microsoft initially refusing to patch their older software for released CIA/FBI hacks.

I still think Intel needs to be leading the fixes. CPU's are rarely upgraded, but I would consider dumping say $100 to upgrade the CPU (faster, more cache, etc) that also fixed this bug on an older computer.

It is unclear what all CPUs, GPUs, and FPUs are involved. Apparently NVidia is already releasing new GPU drivers for their hardware. There are a lot of potential attack points.
CliffordK is offline  
Reply With Quote
Old 01-09-18, 10:26 PM
  #59  
CliffordK
Senior Member
 
CliffordK's Avatar
 
Join Date: Nov 2014
Location: Eugene, Oregon, USA
Posts: 20,111
Mentioned: 124 Post(s)
Tagged: 0 Thread(s)
Quoted: 7077 Post(s)
With this Google Think Tank...

I wonder how many Google vulnerabilities there are?

Bike Forums apparently has had issues with Google pass-through ads doing unwanted stuff.
CliffordK is offline  
Reply With Quote
Old 01-09-18, 10:34 PM
  #60  
Marcus_Ti 
Only Slightly Bent
 
Marcus_Ti's Avatar
 
Join Date: Sep 2014
Location: Lincoln, Nebraska
Posts: 4,017

Bikes: Roadie: Seven Axiom Race Ti w/Chorus 11s. CX/Adventure: Carver Gravel Grinder w/ Di2

Mentioned: 24 Post(s)
Tagged: 0 Thread(s)
Quoted: 1525 Post(s)
Originally Posted by CliffordK View Post
With this Google Think Tank...

I wonder how many Google vulnerabilities there are?

Bike Forums apparently has had issues with Google pass-through ads doing unwanted stuff.

Relative to Microsoft or any other major company....Google is constantly rewriting and rebuilding code. The odds of a decades long bug buried are low...because, infuriatingly, Google products don't last that long before being re-created ground up.
Marcus_Ti is online now  
Reply With Quote
Old 01-09-18, 10:44 PM
  #61  
CliffordK
Senior Member
 
CliffordK's Avatar
 
Join Date: Nov 2014
Location: Eugene, Oregon, USA
Posts: 20,111
Mentioned: 124 Post(s)
Tagged: 0 Thread(s)
Quoted: 7077 Post(s)
Originally Posted by Marcus_Ti View Post
Relative to Microsoft or any other major company....Google is constantly rewriting and rebuilding code. The odds of a decades long bug buried are low...because, infuriatingly, Google products don't last that long before being re-created ground up.
True, although I use the old version of the Google E-Mail system because I think the new version is a memory/resource hog.

But the risk, of course, is a snatch and grab attack. Say I convince Google that I own a bicycle company with too much excess cash which I want to give to Google. So, I buy an ad with them.

How much of my own code is used when they push my Ad up to BikeForums?

Now, some BikeForum member inadvertently clicks on my WonderBike ad, and it sends them directly to my website with software that either hasn't been vetted by Google, or that I have control of changing after purchasing the ad.

Of course, not unlike having a website that pops up in a search that requires the use of Javascript to do or see anything on the page.

Microsoft, and many Linux vendors go through major re-writes of their code every year or so. So, it is not uncommon for something that previously worked to suddenly become broke. But, they also tow along a lot of really old baggage.

Many old DOS commands still work under all versions of Windows. Presumably some have been rewritten, but they may still have some associated code back to the CPM days.
CliffordK is offline  
Reply With Quote
Old 01-09-18, 11:29 PM
  #62  
Shimagnolo
Senior Member
 
Shimagnolo's Avatar
 
Join Date: May 2008
Location: Zang's Spur, CO
Posts: 8,569
Mentioned: 7 Post(s)
Tagged: 0 Thread(s)
Quoted: 1351 Post(s)
Originally Posted by CliffordK View Post
T

I still think that this might be something that could be patched with BIOS. Although perhaps there are enough different versions of BIOS that it is easier to patch in the OS and Browsers.
The BIOS is only involved in the booting of the computer.
Once booted, the OS is talking to the hardware, and the BIOS is no longer involved.

Furthermore, the exploits utilize the cache which inside the CPU package.
The BIOS is on separate chip on the motherboard, and has zero influence on what is happening inside the CPU.
Shimagnolo is online now  
Reply With Quote
Old 01-10-18, 12:46 AM
  #63  
CliffordK
Senior Member
 
CliffordK's Avatar
 
Join Date: Nov 2014
Location: Eugene, Oregon, USA
Posts: 20,111
Mentioned: 124 Post(s)
Tagged: 0 Thread(s)
Quoted: 7077 Post(s)
Originally Posted by Shimagnolo View Post
The BIOS is only involved in the booting of the computer.
Once booted, the OS is talking to the hardware, and the BIOS is no longer involved.

Furthermore, the exploits utilize the cache which inside the CPU package.
The BIOS is on separate chip on the motherboard, and has zero influence on what is happening inside the CPU.
I thought the BIOS (Basic Input/Output System) contained code for the execution of certain routines, for example hard drive access might be handled through a bios call.

However, it is possible that the PC design has allowed these bios calls to be bypassed, or there would be some operations that would be handled directly without going through BIOS. I haven't dealt with the machine/assembly level much.

I'm also surprised that this kind of manipulation is allowed through what are supposed to be interpreted languages.

Perhaps it is time to toss the old PCs and build a new "secure" version from scratch.
CliffordK is offline  
Reply With Quote
Old 01-10-18, 01:11 AM
  #64  
Shimagnolo
Senior Member
 
Shimagnolo's Avatar
 
Join Date: May 2008
Location: Zang's Spur, CO
Posts: 8,569
Mentioned: 7 Post(s)
Tagged: 0 Thread(s)
Quoted: 1351 Post(s)
Originally Posted by CliffordK View Post
I thought the BIOS (Basic Input/Output System) contained code for the execution of certain routines, for example hard drive access might be handled through a bios call.
That was true back in the days of DOS.
Shimagnolo is online now  
Reply With Quote
Old 01-10-18, 01:34 AM
  #65  
CliffordK
Senior Member
 
CliffordK's Avatar
 
Join Date: Nov 2014
Location: Eugene, Oregon, USA
Posts: 20,111
Mentioned: 124 Post(s)
Tagged: 0 Thread(s)
Quoted: 7077 Post(s)
Originally Posted by Shimagnolo View Post
That was true back in the days of DOS.
You mean I can't use my MFM drives anymore?

I suppose that is why so many devices come with drivers, although I would think some basic system functions would still be handled by BIOS, but maybe not.

I have to look back at the concept of levels of abstraction, and I suppose why they aren't enforced.
CliffordK is offline  
Reply With Quote
Old 01-10-18, 02:28 PM
  #66  
CliffordK
Senior Member
 
CliffordK's Avatar
 
Join Date: Nov 2014
Location: Eugene, Oregon, USA
Posts: 20,111
Mentioned: 124 Post(s)
Tagged: 0 Thread(s)
Quoted: 7077 Post(s)
It looks like there is about to be a lot more pressure on an insider trading case against Intel's CEO. I wonder how many other insiders did the same thing.

https://www.bloomberg.com/news/artic...-s-share-sales

Shareholders mulling suit against Intel over CEO's stock sale - Business Insider

The truth is that the stocks had gained some value in December, so he might have been OK to schedule the sale after the bug announcement, although the share prices continue to fall now.

Of course, the CEO could claim that he is a bad manager, and was just clueless about one of the most major bugs in chip design to hit his company in the last two decades.
CliffordK is offline  
Reply With Quote
Old 01-10-18, 02:37 PM
  #67  
Marcus_Ti 
Only Slightly Bent
 
Marcus_Ti's Avatar
 
Join Date: Sep 2014
Location: Lincoln, Nebraska
Posts: 4,017

Bikes: Roadie: Seven Axiom Race Ti w/Chorus 11s. CX/Adventure: Carver Gravel Grinder w/ Di2

Mentioned: 24 Post(s)
Tagged: 0 Thread(s)
Quoted: 1525 Post(s)
Originally Posted by Shimagnolo View Post
The BIOS is only involved in the booting of the computer.
Once booted, the OS is talking to the hardware, and the BIOS is no longer involved.

Furthermore, the exploits utilize the cache which inside the CPU package.
The BIOS is on separate chip on the motherboard, and has zero influence on what is happening inside the CPU.
Most systems don't use BIOS anymore--they're UEFI....which how much a difference that makes WRT mitigating/patching these exploits I don't know, if any.
Marcus_Ti is online now  
Reply With Quote
Old 01-10-18, 02:45 PM
  #68  
genec
genec
Thread Starter
 
genec's Avatar
 
Join Date: Sep 2004
Location: West Coast
Posts: 26,221

Bikes: custom built, sannino, beachbike, giant trance x2

Mentioned: 21 Post(s)
Tagged: 0 Thread(s)
Quoted: 5354 Post(s)
Originally Posted by CliffordK View Post
It looks like there is about to be a lot more pressure on an insider trading case against Intel's CEO. I wonder how many other insiders did the same thing.

https://www.bloomberg.com/news/artic...-s-share-sales

Shareholders mulling suit against Intel over CEO's stock sale - Business Insider

The truth is that the stocks had gained some value in December, so he might have been OK to schedule the sale after the bug announcement, although the share prices continue to fall now.

Of course, the CEO could claim that he is a bad manager, and was just clueless about one of the most major bugs in chip design to hit his company in the last two decades.
Is it really a bug though? The chips have been designed in this manner for some time... primarily to keep operations fast in the CPU. It's just that "recently" it was discovered that there is a way to access that buffer and read it in a way to get information from the computer in a clandestine manner.

What if we discover a new way to gain access to computers in the future... some other normal design feature that can be exploited to gain access to your info... maybe the pulsing of the powersupply as read through the power lines, or something as innocuous as that. How are companies supposed to anticipate that later tech being able to dive into older designs in that manner. In a way, it is like busting a past Tour d France winner for using a drug that wasn't even being looked for, 10 years ago. (cycling forum... I get to use such metaphors... )

I'm just not sure this was a "design flaw" like the "divide error" was, in Pentiums, years ago.
genec is offline  
Reply With Quote
Old 01-10-18, 02:56 PM
  #69  
Juan Foote
LBKA (formerly punkncat)
 
Juan Foote's Avatar
 
Join Date: Jan 2010
Location: Jawja
Posts: 3,377

Bikes: Spec Roubaix SL4, GT Traffic 1.0

Mentioned: 9 Post(s)
Tagged: 0 Thread(s)
Quoted: 875 Post(s)
^ I tend to agree with you. Black hat is always looking for a new way to exploit systems and came across this after what? 20 years of design?

Even in spite of this, reading indicates that the risk is fairly minimal in many situations and there are Intel recommends not to even do this in certain server systems that aren't "on the web", so to speak.

I have one of the older systems where 'the word' is that my slowdown would be notable. I think I did the first day or so but have become used to it to a degree that I just don't see it, and could EASILY be mistaken for internet latency which I experience more than system slowdown. I have one computer, and older AMD, that is truly being hampered by it but it's use as an HTPC doesn't make this critical in any kind of way.
I have been looking to replace a few of the boxes we work from and now have a GREAT excuse to wait for the new risk mitigated chip design to come out. This whole situation just saved me money, really.
Juan Foote is offline  
Reply With Quote
Old 01-10-18, 03:04 PM
  #70  
Shimagnolo
Senior Member
 
Shimagnolo's Avatar
 
Join Date: May 2008
Location: Zang's Spur, CO
Posts: 8,569
Mentioned: 7 Post(s)
Tagged: 0 Thread(s)
Quoted: 1351 Post(s)
Originally Posted by genec View Post
Is it really a bug though? The chips have been designed in this manner for some time... primarily to keep operations fast in the CPU. It's just that "recently" it was discovered that there is a way to access that buffer and read it in a way to get information from the computer in a clandestine manner.

What if we discover a new way to gain access to computers in the future... some other normal design feature that can be exploited to gain access to your info... maybe the pulsing of the powersupply as read through the power lines, or something as innocuous as that. How are companies supposed to anticipate that later tech being able to dive into older designs in that manner. In a way, it is like busting a past Tour d France winner for using a drug that wasn't even being looked for, 10 years ago. (cycling forum... I get to use such metaphors... )

I'm just not sure this was a "design flaw" like the "divide error" was, in Pentiums, years ago.
It flat out is a design flaw.
One process should never be able to access the memory of a different process.
We are not talking about clandestine snooping techniques;
It is just a matter of feeding some instructions to the CPU.
And not all CPUs are susceptible; Just some of those running speculative execution.
This is analogous to a software vendor blindly adding new features to a product with no consideration for how they might be maliciously exploited.
Shimagnolo is online now  
Reply With Quote
Old 01-10-18, 03:06 PM
  #71  
Marcus_Ti 
Only Slightly Bent
 
Marcus_Ti's Avatar
 
Join Date: Sep 2014
Location: Lincoln, Nebraska
Posts: 4,017

Bikes: Roadie: Seven Axiom Race Ti w/Chorus 11s. CX/Adventure: Carver Gravel Grinder w/ Di2

Mentioned: 24 Post(s)
Tagged: 0 Thread(s)
Quoted: 1525 Post(s)
Originally Posted by Shimagnolo View Post
It flat out is a design flaw.
One process should never be able to access the memory of a different process.
We are not talking about clandestine snooping techniques;
It is just a matter of feeding some instructions to the CPU.
And not all CPUs are susceptible; Just some of those running speculative execution.
This is analogous to a software vendor blindly adding new features to a product with no consideration for how they might be maliciously exploited.
I would think in software debugging this would be desirable. But for a consumer, yea--aside from fringe things like video game trainers/cheats etc that is all I can think of.
Marcus_Ti is online now  
Reply With Quote
Old 01-10-18, 03:12 PM
  #72  
Shimagnolo
Senior Member
 
Shimagnolo's Avatar
 
Join Date: May 2008
Location: Zang's Spur, CO
Posts: 8,569
Mentioned: 7 Post(s)
Tagged: 0 Thread(s)
Quoted: 1351 Post(s)
BTW Here is an interesting article with interesting info about a couple solutions (retpoline and PCID): https://www.theregister.co.uk/2018/0...ectre_slowdown
Shimagnolo is online now  
Reply With Quote
Old 01-10-18, 03:13 PM
  #73  
genec
genec
Thread Starter
 
genec's Avatar
 
Join Date: Sep 2004
Location: West Coast
Posts: 26,221

Bikes: custom built, sannino, beachbike, giant trance x2

Mentioned: 21 Post(s)
Tagged: 0 Thread(s)
Quoted: 5354 Post(s)
Originally Posted by Shimagnolo View Post
It flat out is a design flaw.
One process should never be able to access the memory of a different process.
We are not talking about clandestine snooping techniques;
It is just a matter of feeding some instructions to the CPU.
And not all CPUs are susceptible; Just some of those running speculative execution.
This is analogous to a software vendor blindly adding new features to a product with no consideration for how they might be maliciously exploited.
But doesn't it make sense to share the contents of said memory when doing (attempting) "parallel processing?"

In some respects, this reminds me of different applications sharing .dll files.
genec is offline  
Reply With Quote
Old 01-10-18, 03:16 PM
  #74  
Juan Foote
LBKA (formerly punkncat)
 
Juan Foote's Avatar
 
Join Date: Jan 2010
Location: Jawja
Posts: 3,377

Bikes: Spec Roubaix SL4, GT Traffic 1.0

Mentioned: 9 Post(s)
Tagged: 0 Thread(s)
Quoted: 875 Post(s)
Keeping in mind, that even if you refuse the update you will create a huge headache within the 10 enviro, and cause issue with future updating wanting to auto add it anyway. IMO, suck it up and deal with the update.
Juan Foote is offline  
Reply With Quote
Old 01-10-18, 03:18 PM
  #75  
Marcus_Ti 
Only Slightly Bent
 
Marcus_Ti's Avatar
 
Join Date: Sep 2014
Location: Lincoln, Nebraska
Posts: 4,017

Bikes: Roadie: Seven Axiom Race Ti w/Chorus 11s. CX/Adventure: Carver Gravel Grinder w/ Di2

Mentioned: 24 Post(s)
Tagged: 0 Thread(s)
Quoted: 1525 Post(s)
Originally Posted by genec View Post
But doesn't it make sense to share the contents of said memory when doing (attempting) "parallel processing?"

In some respects, this reminds me of different applications sharing .dll files.
Which can make things much simpler...or can set the entire neighborhood on fire.

About a decade ago now was Arch Linux was one of the first to push an update to the shared library LibPNG and LibJPEG. No biggie right? Just another update to another package....Funny thing, it was a massive overhaul of both...and, well, everything uses both....and everything can and should have been recompiled against the new libs, except they weren't. Everything, because basically everything with a GUI uses either PNGs or JPEGs.


Long story short, too late, EVERYTHING with a GUI was broken by the update...and lots of Arch users early to -Syu lost Xorg as well as their DE....and had to revert via command line. That was a fun few days on Arch Forums, and Allen really did break it.
Marcus_Ti is online now  
Reply With Quote

Thread Tools
Search this Thread

Contact Us Archive Advertising Cookie Policy Privacy Statement Terms of Service