MINIX: Intel's Hidden In-chip Operating System (zdnet.com)

Steven J. Vaughan-Nichols, writing for ZDNet: Matthew Garrett, the well-known Linux and security developer who works for Google, explained recently that, "Intel chipsets for some years have included a Management Engine [ME], a small microprocessor that runs independently of the main CPU and operating system. Various pieces of software run on the ME, ranging from code to handle media DRM to an implementation of a TPM. AMT [Active Management Technology] is another piece of software running on the ME." [...] At a presentation at Embedded Linux Conference Europe, Ronald Minnich, a Google software engineer reported that systems using Intel chips that have AMT, are running MINIX. So, what's it doing in Intel chips? A lot. These processors are running a closed-source variation of the open-source MINIX 3. We don't know exactly what version or how it's been modified since we don't have the source code. In addition, thanks to Minnich and his fellow researchers' work, MINIX is running on three separate x86 cores on modern chips. There, it's running: TCP/IP networking stacks (4 and 6), file systems, drivers (disk, net, USB, mouse), web servers. MINIX also has access to your passwords. It can also reimage your computer's firmware even if it's powered off. Let me repeat that. If your computer is "off" but still plugged in, MINIX can still potentially change your computer's fundamental settings. And, for even more fun, it "can implement self-modifying code that can persist across power cycles." So, if an exploit happens here, even if you unplug your server in one last desperate attempt to save it, the attack will still be there waiting for you when you plug it back in. How? MINIX can do all this because it runs at a fundamentally lower level. [...] According to Minnich, "there are big giant holes that people can drive exploits through." He continued, "Are you scared yet? If you're not scared yet, maybe I didn't explain it very well, because I sure am scared." Also read: Andrew S. Tanenbaum's (a professor of Computer Science at Vrije Universiteit) open letter to Intel.

Comments

Three questions

[Anonymous Coward]

1) Do AMD processors have similar vulnerabilities or is this an Intel issue only?

2) Why isn't Intel being held responsible to fix this, either by action of lawmakers or through lawsuits for providing a faulty product?

3) Shouldn't Intel either have to patch the vulnerabilities or issue a recall?

The years of the Minux desktop

[sinij]

Apparently, we have been having years of Minux desktop all this time and never knew.

Tanenbaum: a professor of Computer Science...?

[Barnoid]

Kids these days...

Andrew S. Tanenbaum is the original creator of MINIX, not just "a professor" at Vrije Universiteit.

Re:So it's a backdoor///

[swillden]

Let's call this what it is: A variation of the "clipper chip" like the government tried to do years ago, except this is more powerful and way worse.

That's a mischaracterization so egregious it could be called a lie.

The ME (and AMD's analogous PSP) have nothing to do with government, and nothing to do with cryptography (though they make heavy use of it). Clipper was about enforcing a standardized encryption mechanism with a built-in backdoor specifically for law enforcement. Completely different thing.

ME and PSP are remote system management tools. Their purpose is to enable enterprises to remotely administer systems, including not only being able to remotely install a new operating system, but to strongly verify the installation from the running OS. The reason it's in all systems, not just systems targeted at enterprise use, is that it's more economical to have a single solution

That said... you are absolutely correct that these tools *could* be used by malicious parties, whether for corporate espionage, government intrusion or anything else, and they are incredibly powerful, and not understood nearly well enough outside of the teams at Intel and AMD who build them. I know some of the people at Intel who work on this stuff and I'm pretty confident that they're doing good work, and doing the right things... but the lack of transparency makes me really nervous.

Remote management tools make sense, but it should be possible for end users to disable them, or to take ownership of them and use them for their own ends. The details of exactly how they work, including their source code, should be published. Indeed, I think government should mandate the publication of low-level system management tools and firmware. We need a lot more academic research into the security and operation of these systems.

wow the FUD is strong in that one

[kamakazi]

We have a couple facts here, and a whole bunch of conclusions.
The facts are that there is a general purpose OS running a microkernel in a management layer on unspecified Intel CPUs. This general purpose OS provides at least network accessible management interfaces.
The conclusions are this general purpose OS is infinitely exploitable to steal all your top secret information and redirect all you web requests to the mind control platform of the month.
This Minnich character (I enjoyed that similarity, Minnich/Minix) then jumps to a call to neuter everything below the user installed OS including UEFI. He then juts off on a side tangent and says trust me (He is a Google engineer) to always install good safe firmware on your Chromebook. That was a nice subtle bit of astroturfing there. He also blames Minix for slow boot time on an Open Compute server, not sure where minix plays into that or what axe he is grinding.

Let's look at it a little more objectively. Why do these processor companies keep putting general purpose OSs at a level which was traditionally all hardware/firmware, and why do systems makers use an accesible programming layer to configure hardware like UEFI? Well, whe we were running 386s and 486s we really were running microprocessors. Hardware was relatively static, device support was locked at time of manufacture, processors did processing (with maybe a coprocessor for math) and accessory cards did a single function each. In that time frame supers, like the first Crays, couldn't even boot themselves. They used a completely separate computer to boot and for time scheduling and such. Now today, we have computers which are powerful on the level of the early supers. Our processing no longer all happens on the CPU, but also in the GPU(s) and other pieces in the system. We no longer have external memory and bus controllers, they are built into the processor or the mandatory northbridge, and are much more capable and adaptive. There are hosts of sensors built into modern processors. All of these pieces need to be managed. There is an absolute necessity for a relatively capable computer in there to manage all these pieces.
It used to be done with static logic arrays, controlled by registers, and we called it BIOS, and it had a little interface that could usurp the monitor output and keybpoard and chirp the speaker, later got so fancy it could hijack a mouse on some systems. It was very limited, in fact, on the earliest PCs it didn't have a UI at all, it had dip switches or jumpers on the system board.

Now with the advent of negotiated buses (even memory buses, back in the day I never would have conceived of a CPU being able to ask a memory module what capabilities it possessed and automatically configure timing parameters to best talk to it) the management processor has a lot to do. On high end machines they even do this negotiation on the fly with the advent of hot plug PCI buses and on the fly memory error compensation. By the nature of the beast this management engine has to be able to see all the data buses, otherwise every single connection interface would need an out of band management channel.

I suppose you could make this management engine like a FPGA, configure it once and burn your bridges, no further interraction possible, but then what happens when you need to add or change something?

Likewise it often doesn't need a network interface, but if it doesn't have one then we have to do wake on LAN with yet another baby management computer. How about physical intrusion detection? again, not often needed, but sometimes...

Basically what a general purpose OS in the management layer does is give nearly infinite flexibility. This technology is a big part of the reason so much of our stuff just works.

Now, I am not really a drink the cool-aid from the benevolent overlords kind of guy, I am not at all in favor of secret OSs underpinning our hardware without our knowledge, but let's not throw out the baby too. That capability is in most cases useful