AMD Will Replace AGESA With Open Source Initialization Library 'openSIL' (phoronix.com) 9
Phoronix shares some overlooked news from AMD's openSIL presentation at the OCP Regional Summit in April. Specifically, that AMD openSIL — their open-source x86 silicon initialization library — "is planned to eventually replace AMD's well known AGESA [BIOS utility]" around 2026, and "it will be supported across AMD's entire processor stack — just not limited to EPYC server processors as some were initially concerned..."
Raj Kapoor, AMD Fellow and AMD's Chief Firmware Architect, in fact began the AMD openSIL presentation by talking about the challenges they've had with AGESA in adapting it to Coreboot for Chromebook purposes with Ryzen SoCs... With AMD openSIL not expected to be production ready until around 2026, this puts it roughly inline for an AMD Zen 6 or Zen 7 introduction. The proof of concept code for AMD Genoa is expected to come soon...
The presentation also noted that beyond AMD openSIL code being open-source, the openSIL specification will also be open. AMD "invites every silicon vendor" to participate in this open-source system firmware endeavor.
The presentation also noted that beyond AMD openSIL code being open-source, the openSIL specification will also be open. AMD "invites every silicon vendor" to participate in this open-source system firmware endeavor.
Shuuuure (Score:1)
They promised coreboot support before... then backtracked.
I'll believe it when I've compiled coreboot from scratch without any blobs, installed my own firmware, and booted from it, several times, lasting a couple months each run.
Since the PSP has this nasty habit of kicking you out after half an hour. Who's to say there's not more of that snuck into the silicon somewhere? Oh, right, source and same recipe for everything else that has a controller built in, including the southbridge these days.
And not jus
Re: (Score:3)
They will still require blobs. The CPU microcode is proprietary and I think it's also digitally signed for security reasons (i.e. so malware can't change the behaviour of your CPU).
Is it a big deal? They don't give you the schematics for the CPU either, they are a trade secret. Before such things were loaded by the firmware, they were just hard coded into the CPU anyway and you couldn't access them.
Supporting Coreboot means you can control the boot process, completely remove things like Intel Management Eng
Re: (Score:3)
The thing with IME and others is that it's required to manage a large and complex CPU. Sure you can do without it, but it's generally not reocmmended.
Modern processors, yes, including the one in your smartphone, are huge and complex and often now have boot architectures similar to supercomputers - you boot
Awesome. (Score:4, Insightful)
It's good to see a company working to help the open source community. I have no illusions about AMD's reasons but it's still far better than the Intel approach of intentionally or indifferently locking out open source projects.
What could possibly take 3 fucking years? (Score:2)
This sounds more like a management decision to massively delay it than an announcement to release it.
Re: (Score:2)
Ah, no. Getting something to work with your own people and turning it into professional level FOSS are to pretty different things, especially for FOSS than can easily kill your hardware. 3 years is quite reasonable.
Re: (Score:3)
If you look at how long it took for AMD to properly open source their GPU drivers (after initially making the decision to do so) 3 years to get this done seems reasonable.
That is really good news (Score:4, Interesting)
Once again, AMD decides to go the "lets to do good engineering to make money" path and not the "screw our customers, lets get rich" path, that Intel (among others) likes so much.
This also means they had requests from customers that do or plan to build their own hardware. And very likely from customers that want to be able to fix security problems in the BIOS themselves. This definitely is the future.
Re: (Score:2)