Freedom and security issues on x86 platforms

Given that chip development has been hitting diminishing returns for a few years it might be time for Open Source to eat the world of processors as well.

It feels like the sort of opportune market that server operating systems, databases and web servers occupied: less of a visual aesthetic and more of a better-design-wins market.

It's not going to be easy - I'd guess that it would take at least 10 years for a project to get any sort of traction outside of a very small niche group.

Re: Freedom and security issues on x86 platforms

#5

Post by speeder » Mon, Apr 04, 2016, 3:58 PM UTC

I've been looking into this recently.

Basically, the things mentioned on the text, made free software bios and firmwares impossible, some of the free software projects that exist now are mostly "binary blobs loaders", having more binary blob than free software code running.

There is some good analysis on why even Intel can't fix this if they wanted to, unless they stopped shipping some features entirely, their Intel ME system rely on a couple of proprietary third party code, that has on contract with Intel explicit prohibitions of Intel ever letting anyone seeing their source, or the keys needed to sign them.

Also, Intel ME can't be really trusted, the code is not really "reverse-engineeringable", and it works as a full second OS of sorts, it even has its own JVM running, if someone somehow decide to inject spy software into it, you will never know, also I assume that the first destructive virus to latch into that stuff, will take the world truly by surprise depending on when it triggers (for example if it spreads silently but triggers the destructive payload on a specific date).

Also, these features can be abused to abuse the market itself, for example by intentionally making the hardware underperform, and then sell "superior" hardware that has the only difference some software.

Re: Freedom and security issues on x86 platforms

#6

Post by ashitlerferad » Mon, Apr 04, 2016, 4:00 PM UTC

odinduty wrote: post #2

> Secure Boot [...] requires FOSS users to purchase a license from Microsoft Nope.

Requires purchase of a cerficate from one of the authorities Microsoft recognises (Verisign/Digicert/...) and then the signature of Microsoft on compiled bootloader code. Either way, you have to pay and you have to get Microsoft's permission.

Re: Freedom and security issues on x86 platforms

#7

Post by maaku » Mon, Apr 04, 2016, 4:00 PM UTC

mmastrac wrote: post #4

Given that chip development has been hitting diminishing returns for a few years it might be time for Open Source to eat the world of processors as well. It feels like the sort of opportune market that server operating systems, databases and web servers occupied: less of a visual aesthetic and more of a better-design-wins market. It's not going to be easy - I'd guess that it would take at least 10 years for a project…

People are working on that, and it might take less time than you think:

http://riscv.org/

Re: Freedom and security issues on x86 platforms

#8

Post by darpa_escapee » Mon, Apr 04, 2016, 4:00 PM UTC

ARM architectures also suffer from this. You'll be hard pressed to find a board that doesn't require a propriety board support package somewhere in the stack.

Ironically, it is usually the bootloader that is/requires a blob or it is the DTB.

I remember being in middle school and reading Stallman's articles on the dangers of a TPM-oriented push by manufacturers. As cliche as it is, Stallman was right.

The push for platform security is also a push for platform ownership. Tinkering/hacking/your ability as a hardware owner is at ends with corporate security needs and that is a shame.

Re: Freedom and security issues on x86 platforms

#9

Post by ashitlerferad » Mon, Apr 04, 2016, 4:02 PM UTC

maaku wrote: post #7

mmastrac wrote: post #4

Given that chip development has been hitting diminishing returns for a few years it might be time for Open Source to eat the world of processors as well. It feels like the sort of opportune market that server operating systems, databases and web servers occupied: less of a visual aesthetic and more of a better-design-wins market. It's not going to be easy - I'd guess that it would take at least 10 years for a project…

People are working on that, and it might take less time than you think: http://riscv.org/

and some of those are working on physical chip production:

http://www.lowrisc.org/

Re: Freedom and security issues on x86 platforms

#10

Post by ashitlerferad » Mon, Apr 04, 2016, 4:04 PM UTC

darpa_escapee wrote: post #8

ARM architectures also suffer from this. You'll be hard pressed to find a board that doesn't require a propriety board support package somewhere in the stack. Ironically, it is usually the bootloader that is/requires a blob or it is the DTB. I remember being in middle school and reading Stallman's articles on the dangers of a TPM-oriented push by manufacturers. As cliche as it is, Stallman was right. The push for pla…

There are plenty of boards that run using mainline u-boot and mainline Linux, you can often ignore the vendor BSP.

You are right about ARM though, TrustZone is another DRM-oriented aspect of ARM-based SoCs.