Specter and Meltdown: - How Serious? Taking it on the Chip.

TLS Guy

TLS Guy

Audioholic Jedi
We seem to have another computer threat to worry about. It seems all CPUs made on the last 20 years, including those unsold and in production are vulnerable to Specter and Meltdown hacks.

These hacks go right for the CPU and allow the hackers to mine memory for information. Worse these hacks are hard to detect.

All this has received greater attention in the UK press than here. Mainly I think because UK manufacturer ARM makes Apple CPUs. It seems all ARM, Intel and AMD CPUs are vulnerable.

It seems hard to come up with a fix. Specter is the most serious and as yet there is no solution. Many experts are skeptical this back door can be closed with software patches.

Apple and Microsoft were supposed to issue patches Jan 4 for Meltdown, but did not. Rumor has it they were only partially effective. There is no patch on the horizon for the more serious Specter.

Now this is a very difficult vulnerability to exploit, discovered by Google. It was supposed to stay secret, but some idiot could not keep his mouth shut and released the info the Reddit. So the info had to be made public.

It seems the easiest route for hackers is adverts. So I strong ad blocker is recommended. I have used U-Block Origin on all our devices for some time, and would give it a strong recommendation to you all.
 
NINaudio

NINaudio

Audioholic Samurai
Thanks TLS guy, I use adblock plus on my pcs, and ad-block browser on my phone. I've heard of U-block before, but never looked much into it. I'll have to check it out.
 
Irvrobinson

Irvrobinson

Audioholic Spartan
We seem to have another computer threat to worry about. It seems all CPUs made on the last 20 years, including those unsold and in production are vulnerable to Specter and Meltdown hacks.

These hacks go right for the CPU and allow the hackers to mine memory for information. Worse these hacks are hard to detect.

All this has received greater attention in the UK press than here. Mainly I think because UK manufacturer ARM makes Apple CPUs. It seems all ARM, Intel and AMD CPUs are vulnerable.

It seems hard to come up with a fix. Specter is the most serious and as yet there is no solution. Many experts are skeptical this back door can be closed with software patches.

Apple and Microsoft were supposed to issue patches Jan 4 for Meltdown, but did not. Rumor has it they were only partially effective. There is no patch on the horizon for the more serious Specter.

Now this is a very difficult vulnerability to exploit, discovered by Google. It was supposed to stay secret, but some idiot could not keep his mouth shut and released the info the Reddit. So the info had to be made public.

It seems the easiest route for hackers is adverts. So I strong ad blocker is recommended. I have used U-Block Origin on all our devices for some time, and would give it a strong recommendation to you all.
First of all, just to be precise, ARM does not make Apple CPUs. In fact, ARM doesn't design or make any CPUs at all. ARM is a provider of different kinds of CPU design "intellectual property", and other companies, like Apple, Broadcom, Qualcomm, and numerous others, use ARM's IP to design their CPUs. That's why they're technically called "ARM-based CPUs". MIPS is another example of this type of IP provider.

(As an aside, in modern parlance, a CPU is one or more CMOS dies that are fitted into a chip package, that equates to a single "socket" that plugs into a computer motherboard. Each "CPU" includes multiple "cores", where each core has certain hardware modules, lately called execution units, that process the instructions as defined by the software. There are numerous different types of execution units in each CPU design. Each core can have separately addressable register sets that allow the sharing of these execution units between multiple instruction streams, which can make each core look like multiple cores to the operating system. These logical instances of physical cores are called "hardware threads". Typically, CPUs have cores that support between one and eight hardware threads, so an 8-core CPU with dual threads per core looks like 16 actual "CPUs" to the operating system.)

So, Apple, for example, uses ARM IP to design their A-series CPUs, and I suspect Apple has the type of license from ARM that allows them to modify the processor logic. Of course, Apple doesn't make CPUs either, in a physical sense, nor does Broadcom or Qualcomm - these companies are called fabless semiconductor companies. They send their designs to fabricators like TSMC, Samsung, and Intel to manufacture the chips for them. So all ARM-based CPUs are different, though they have the same basic instruction set to adhere to the IP license agreement.

Obviously, Apple uses Intel CPUs for Macs.

So, how bad are these "flaws"? Well, anyone can do exhaustive internet searches to try to understand what Meltdown and Spectre are, but the information available looks pretty skimpy. CPUs are so fast compared to memory that CPU designers use tricks to keep the cores busy and keep them from waiting on memory accesses, which slows performance. (CPUs process multiple instructions per clock cycle, and their clock speeds are less than a nanosecond. DRAM DIMM access times are typically 50-90 nanoseconds, so we're talking differences of multiple orders of magnitude.) Two strategies used by CPUs, which are completely invisible to the operating systems, are speculative execution and CPU-local static RAM caches. Speculative execution "looks ahead" in the software path and does some level of instruction processing in advance. CPU-local caches operate at speeds that are reasonable multiples of the CPU clock speed, making the CPU wait much less for instructions and data. Unlike memory, which is managed by the operating system, caches are largely managed by the CPU hardware or microcode.

It looks like Spectre and Meltdown involve exploiting characteristics of these mechanisms, which due to their CPU-specific nature circumvent operating system security architectures, to allow malware to directly access or infer information owned by other applications on the same computer system. This would be extraordinarily sophisticated malware, on the order of Stuxnet. NSA-class development, not your typical malware hacker. The risk seems to be that lots of not-so-friendly countries have NSA-class capabilities, and potentially could use these so-called flaws to develop some very nefarious malware that could affect basic infrastructure, just guessing.

How much of a threat is it really to a home system? I have no idea, but on a general level the indicator I use for risk is how much money the industry is spending on avoidance, and they're spending a lot to mitigate these two vulnerabilities. I'm not sure what to do about it though other than apply OS or BIOS updates as they become available, which we should always do. No matter how well you design a lock or a barrier, someone will eventually figure how to break it or go around it. Malware designers have more time to think about ways to break things than product developers have time to analyze their designs. And the bad guys out-number the designers too.
 
Last edited:
TLS Guy

TLS Guy

Audioholic Jedi
Thanks Irv for that detailed post.

I posted about this as I'm totally unsure of the risk. I agree information is sketchy, but that is probably a good thing.

My program developer son seems very worried about this development. Like you he is especially concerned about unfriendly enemy states.

I have to say I sense this is an unending "arms race," that in the long run has a good chance of ending very badly and rapidly change the world as we have come to know it.

My late father was always very concerned about the possibility of a total crash and destruction of the digital universe. He always advised to stock up on quill pens and ink! That is not to say that he did not embrace these changes, and actually used computers and advance digital AV systems into very old age. He was just concerned about the total dependency on what are really very new technologies still. I think his concerns about a major melt down were valid unfortunately.
 
TheWarrior

TheWarrior

Audioholic Ninja
I'm far more concerned about 'blue-jacking' as an immediate threat. The ease of intrusion through bluetooth connections is staggering. I do not leave any of my devices' BT connection on.

Irv got pretty in depth with it. Going a bit deeper, my understanding is the kernel process that separates internal processes from user inputs is the way in which intrusion can occur. During this process an attacker is able to trick your CPU in to thinking the attackers computer should be receiving those internal processes.

For this CPU issue, I have read numerous times that the issue is repairable via firmware updates. I don't store passwords on my computer, and our credit union's security has already been tested by my brother in law, so I'm too worried.....

What is concerning is that the CEO of Intel cashed out his shares to the exact minimum he is required to hold prior to this news coming to light. Sounds very familiar to me after my TDI was bought back, and Winterkorn was given a $30M Euro severance.
 
Irvrobinson

Irvrobinson

Audioholic Spartan
my understanding is the kernel process that separates internal processes from user inputs is the way in which intrusion can occur. During this process an attacker is able to trick your CPU in to thinking the attackers computer should be receiving those internal processes.
Operating system kernels are not a process, per se, they are a special CPU execution mode that runs in the software thread context of every user process. When an application executes certain functions, like I/O functions, file system functions, network sends and receives, stuff like that, the user process or thread enters kernel execution mode, which is a different hardware-enforced protection level. The kernel can impact user processes, but user processes are prevented from directly addressing kernel data areas. In modern virtual memory operating systems, user processes each get a separate address space, but the kernel address space is common across the entire operating system. So an OS kernel looks like a god of sorts to the user mode applications.

To make matters more complicated, Virtual Machine Managers (VMMs), are a type of operating system that runs an entire operating system, which is then called a "Guest", as a process, allowing multiple different operating system instances to share the same CPU. CPUs have special hardware support for VMMs, such as for page tables and protection levels. The VMM is like a god to the Guest OS, and the Guest OS is still like a god to the user-mode applications. And if that isn't complicated enough, some CPUs have hardware support for nested VMMs...

The vulnerabilities with Spectre and Meltdown still seem rooted in low-level CPU capabilities that allow the clever hacker to circumvent some of this VMM and OS security and protection. I don't know for sure, but it smells like it.
 
Irvrobinson

Irvrobinson

Audioholic Spartan
My late father was always very concerned about the possibility of a total crash and destruction of the digital universe.
Fortunately, due to the diversity of software in the "digital universe" that seems very unlikely, but people really do worry about malware affecting things like the electrical grid, air traffic control, weapon systems, railroads, etc. I suspect whatever damage happens will be painful, but localized to specific systems. I do wonder about the electrical grid more than is probably healthy though. ;-)
 
Last edited:
Irvrobinson

Irvrobinson

Audioholic Spartan
This is the best semi-technical article I've seen on the subject so far:

https://www.bloomberg.com/news/articles/2018-01-08/-it-can-t-be-true-inside-the-semiconductor-industry-s-meltdown

I find this quote especially interesting, as it supports my suspicion that the CPUs trying to second-guess the software to enhance performance is the multi-headed root of these problems:

"It makes you shudder," said Paul Kocher, who helped find Spectre and started studying trade-offs between security and performance after leaving chip company Rambus Inc. last year. "The processor people were looking at performance and not looking at security."
 
Irvrobinson

Irvrobinson

Audioholic Spartan
slipperybidness

slipperybidness

Audioholic Warlord
@Irvrobinson
Very informative posts!

One item to mention on a side topic--Qualcomm will not be a fabless semiconductor company for much longer. They are in the process of purchasing NXP semiconductor now. Once that deal goes through, they get all of the NXP fabs in their pocket, including the former-FSL fabs (that were Motorola fabs before FSL) in TX and AZ.
 
Irvrobinson

Irvrobinson

Audioholic Spartan
@Irvrobinson
Very informative posts!

One item to mention on a side topic--Qualcomm will not be a fabless semiconductor company for much longer. They are in the process of purchasing NXP semiconductor now. Once that deal goes through, they get all of the NXP fabs in their pocket, including the former-FSL fabs (that were Motorola fabs before FSL) in TX and AZ.
True - I had forgotten about the old Motorola fabs. For enterprise-class CPUs they're not using a sufficiently sophisticated fab process. I'd be really surprised if Qualcomm used them to fab the Centriq chips, which are the ones claimed to be vulnerable in this discussion.
 
slipperybidness

slipperybidness

Audioholic Warlord
True - I had forgotten about the old Motorola fabs. For enterprise-class CPUs they're not using a sufficiently sophisticated fab process. I'd be really surprised if Qualcomm used them to fab the Centriq chips, which are the ones claimed to be vulnerable in this discussion.
Oh yeah, the cutting edge stuff will definitely have to go to foundries. The US fabs are in a nice spot to spit out die at their respective technology nodes on the mature technologies. I'm guessing Global Foundries and TSMC will be the big foundries for Qualcomm.
 
BoredSysAdmin

BoredSysAdmin

Audioholic Slumlord
"A bit" anti-Intel? He's also full of baloney, and twisting some information. I'm with him on the Intel obscurity stance and how frustrating it can be, but as for it "devastating their market share", well he must be long AMD or something along those lines, and just very hopeful. :)
Charlie is known to over state few things here and there. But in overall take, Intel PR, the ones which have most to loose, yet again has very poor PR stance on this. AMD is not bulletproof, but they did dodge one of three known vunerabilies, the same one which could be significant performance hit on Intel CPUs to implement proper work around.

To be honest I'm not that concerned about desktops, but our virtual environment.
VMware already released statement, but I am not yet sure if full patch for esxi released yet.
 
Irvrobinson

Irvrobinson

Audioholic Spartan
Charlie is known to over state few things here and there. But in overall take, Intel PR, the ones which have most to loose, yet again has very poor PR stance on this. AMD is not bulletproof, but they did dodge one of three known vunerabilies, the same one which could be significant performance hit on Intel CPUs to implement proper work around.

To be honest I'm not that concerned about desktops, but our virtual environment.
VMware already released statement, but I am not yet sure if full patch for esxi released yet.
The performance hit is definitely in the "it depends" category, so we'll see once production patches are out there. The first clumsy versions of patches usually have performance hits, and then refinement sometimes is discovered. I'm withholding judgment rather than piling on. As for AMD, all they appeared to do is respect protection ring restrictions and not speculate so deeply. In theory, when the Intel fixes are implemented in HW (and if the current explanations are accurate), Intel's performance advantage from speculative execution falls to comparable with AMD's.

As for the virtual machine issues, I'm not concerned about that for on-prem at all. It's a cloud computing problem, and it could easily be mitigated by restricting VMs on specific hardware to specific users. That would increase cost to increase security, but let's face facts. This discussion is about identified vulnerabilities. You know as well as anyone that the really dangerous vulnerabilities are the ones you don't know about yet. The full flower of cloud computing tests the isolation characteristics of every component of a computing system, and we'll never know if we get it all right.

These problems are actually an interesting argument for using skinnier hardware and not virtualizing anything.
 
newsletter

  • RBHsound.com
  • BlueJeansCable.com
  • SVS Sound Subwoofers
  • Experience the Martin Logan Montis
Top