Facebook + Intel Get Open-Source FSP Booting On Xeon Scalable

[phoronix]

Phoronix: Facebook + Intel Get Open-Source FSP Booting On Xeon Scalable

Facebook and Intel have been working on being able to enable Xeon Scalable Open Compute Project systems with an open-source FSP...

http://www.phoronix.com/scan.php?page=news_item&px=Facebook-Intel-FSP-OCP-OSF

[angrypie]

So, will AMD open up AGESA again to compete with this?

[Guest]

I very much doubt the general public will benefit from this, this is only going to be for server platforms. The source may also be available only to commercial entities.

Companies like Facebook love open source when they benefit from it, they don't like opening up the software on devices they sell to consumers (e.g Facebook Portal).

[kneekoo]

I feel like we're quite far from fully open firmware in mass-produced PCs, but it would be a great thing for Intel to lead the way, forcing others to do the same. I won't hold my breath until that happens, though.

[Spacefish]

Ok CoreBoot, but they will probably still load proprietary bin images for IME and so on to the CPU.. Step in the right direction, but the management engine should be free as well

[starshipeleven]

Companies like Facebook love open source when they benefit from it, they don't like opening up the software on devices they sell to consumers (e.g Facebook Portal).

It's more like none gives a fuck so why bother going th eextra mile. I mean it's a glorified network-connected webcam, you can do the same with a raspi since ages ago.

[starshipeleven]

Ok CoreBoot, but they will probably still load proprietary bin images for IME and so on to the CPU.. Step in the right direction, but the management engine should be free as well

I would say "optional" more than free. It can stay closed source, most people just don't need it at all.

[oiaohm]

I would say "optional" more than free. It can stay closed source, most people just don't need it at all.

Really it would be good if the code in the ME was in fact published for full peer review so exploits in there could be located and fixed. Being intel signed means you cannot replace it with anything you like but it would be good to know if it bug free or not.

Really you modern day intel systems don't start up without ME code either the default or the section extracted from the firmware. People got annoyed when Intel added the 30 min auto power down to the ME default code.

[starshipeleven]

Really it would be good if

What I meant is that if the ME becomes optional to the system, you could just remove it alltogether (or not install it) and therefore it would not need to be opensourced. WHat is not there cannot be bugged.

Even now it is "essential" only because Intel said so (only the ME can disable some additional watchdog that resets the system), not because it is really necessary to use the system.

For example, that watchdog is NOT available to be used for anything else, and it would really be nice to have an integrated watchdog like server boards have.

[oiaohm]

What I meant is that if the ME becomes optional to the system, you could just remove it alltogether (or not install it) and therefore it would not need to be opensourced. WHat is not there cannot be bugged.

Even now it is "essential" only because Intel said so (only the ME can disable some additional watchdog that resets the system), not because it is really necessary to use the system..

No you are wrong. ME cannot be fully disabled me_cleaner is not disabled instead cuts ME back to bare bones functionality.

Things that the ME must do and things like ME must do.

1) ME is you power management that turns on your x86/microcode cpu cores.
2) ME is the processor that loads the microcode in bios image into the x86/microcode cpu cores before you start processing bios image in x86.
3) ME loading updated firmware for it self from bios/EFI firmware because it at times need updates because of bugs/incompadiblity like might have the wrong way to train memory rommed into CPU.

This is the price of a microcode based cpu design that you can fix cpu instruction bugs and initialisation bugs after the cpu leaves the factory you need another cpu of some type to fire the CPU up this cpu will have very high level system access.

Basically no ME code no run most intel X86 chips. Amd has something equal so disable Amd version same problem no start. This lack of means to fix anything after they leave factory would equal higher CPU prices due to higher number of reject cpus. So we may not like ME or AMD equal but its a deal with the devil you kind of have todo for cost effective production. Of course knowing it a deal with devil does mean we should be asking for it to be demonstrated to be secure.

interesting enough the Soc's on the raspberry pis has something equal but its for the GPU not CPU but it starts the CPU as well on the raspberry pis again disable this part of silicon and chip will not start.

What you were just suggesting was car does not using alternator/starter motor all the time so lets remove it and then wonder why the car does not start. Intel ME was/is a starter motor/alternator with a hell load of optional extras even if it does not have a hell load of optional extras it would be good to be able to inspect the design to make sure its not going to do stupid things.

The issue here is particular design choices mandate you have particular things and those choices improve cost effectiveness of production so there is no way out of it. But even if there is no way to avoid doing something does not mean you cannot take due care in this case due care is wanting to see the source code to make sure what in there is not incorrectly coded.