Am 2014-08-28 23:38, schrieb Todd Weaver:
Comparing the blob-free versions from Intel (which are apparently signed, so according to http://www.coreboot.org/Binary_situation are at a 9000+ panic level), would it be possible to have blob-free (probably through RE) that would work (meaning does not require signed binaries) on an AMD board?
Newer chipsets (as in yet to be released) come with signed parts, but it seems the scope of the signature is configurable somehow by AMD. There were some mails about that a couple of days ago ("AMD PSP").
We will do what we can here. The issue is even with immense leverage, having the source released (from AMI, or AMD, (or Intel for that matter)) would undermine tremendous profit that these companies make by keeping this proprietary.
AMD claims that they stopped working on open sourcing their initialization code because it's lots of work (ie. money) with limited return on investment. How much work that is isn't here or there (most of that is because their internal development process is less than optimal and, like most processes in most organizations, hard to change). But it means that someone could make it worth their while given the right kind of project.
What they don't provide sources for is CPU microcode updates (no one does since it's of limited value without the microcode development toolchain and the microcode itself that is getting updated) and various smaller firmware (USB3 which is a licensed core, IMC, an embedded controller, and the SMU). For IMC and SMU there's some reverse engineering effort, partially documented in the wiki, so by asking the right questions to the right people in this community, plus some development, it might be possible to get them opened up even without AMD's help.
Patrick