LinuxBIOS: the future of booting PCs?

[aidanjt]

Today we have 2 main computer booting systems, namly PC-BIOS and EFI, the former mostly driven by Phoenix, AMI and Award, and used in practically every PC sold since the late 80's, and the latter primarily used by Apple in their Macs.

The problems with PC-BIOS are numerous, firstly, it's all done in assembly code, while this keeps the code size compact, it's hard to read, port to other processor architectures, and thus debug, it's more difficult to keep it in a modularised form, and it has no true central code base, the PC-BIOS is basically in a constant state of forking.

Intel created EFI to address these problems, however it has roots in DRM/Treacherous computing, and has a very messy API, some even refer to it as a "bad OS".

There has been a project in the works for a number of years called LinuxBIOS, some bright sparks thought "Instead of replacing the BIOS with a bad OS, why not use a good OS that has drivers already written for it". While great in theory, LinuxBIOS has had a number of implementation problems they had to get around. For those of you that don't know, LinuxBIOS has been running for a few years now on things from TV sets to Supercomputing clusters

However recently an AMD engineer released a LinuxBIOS that works with Gigabyte GA-M57SLI-S4, the first AM2 motherboard to work with the advanced BIOS. AMD has even commented on the efficenty of the LinuxBIOS Hypertransport code.

Unfortunately Intel hasn't been forthcoming with their chipset documentation, so we wont be seeing LinuxBIOS on Intel boards for some time, however, with nVidia's interest peeked, we may see LinuxBIOS for 680i MCPs sometime in the future.

Here's a list of motherboards known to work with LinuxBIOS:
http://linuxbios.org/Supported_Motherboards

And here's a video of Ronald Minnich's presentation of LinuxBIOS at this years FOSDEM meeting at Brussels encoded in Ogg/Theora:
http://ftp.belnet.be/mirrors/FOSDEM/...-LinuxBios.ogg

So what do you think folks, is LinuxBIOS the way of the future?

[shaithis]

I cannot see Intel throwing EFI out, especially with the new Intel macs using it in anger already.

Perhaps there will be YET ANOTHER split? Intel + EFI or AMD/ATI + LinuxBIOS?

Seems about the only things left to tie to a CPU is the gfx card....although even that isn't currently an option for dual-gfx card users

[aidanjt]

You're absolutely right, I can't see Intel letting go of EFI either, but if AMD throws their weight behind LinuxBIOS, we could see AMD/ATI and nVidia distancing themselves from Intel, and that would be bad for Intel.

But the applications for LinuxBIOS is huge, it could make bootloaders and the crappy PC-BIOS partition tables completely redundant real fast.

[aidanjt]

I'm surprised there aren't more people interested in seeing the nasty old BIOS got rid of.
Just imagine, 3 second boots, no more ugly VGA res post screens, a more interactive configuration menu, no more partition restrictions, no more windows stealing the MBR for dual-booters,the possibilities are endless.

[chrisg22]

It looks like a very interesting idea, its just a matter of the big corp's being willing to help out and give away the chipset documents as you said, aidanjt.

From what I read, and I don't have alot of time as I have a exam in 15 minutes, its basically an OS instead of a BIOS which is hugely configurable and its created by aload of open-source techs?

[this_is_gav]

The sort of thing Microsoft would be interested in, had it not been Linux

It wasn't too many years ago (4-6 perhaps) when MS wanted the total boot time down to I think it was 30 seconds or something, POST included. I think since then, discounting the old IDE detection times, POST times have increased. Vista has done a lot to bring down overall boot times for Windows on the OS side, but POST times are still extremely long, especially with additional SATA/RAID/SCSI controllers.

[directhex]

technically a number of controllers have "wait" periods as part of the specification - i know SCSI does

[aidanjt]

It looks like a very interesting idea, its just a matter of the big corp's being willing to help out and give away the chipset documents as you said, aidanjt.

From what I read, and I don't have alot of time as I have a exam in 15 minutes, its basically an OS instead of a BIOS which is hugely configurable and its created by aload of open-source techs?

Basically yes.
Not only is LinuxBIOS highly configurable, it's more importantly end-user configurable because of the opensource nature of Linux and the GPL license.

Anyone could download the LinuxBIOS sources, modify it, and stick it on their motherboard and get it to do new funky stuff. Of course you'd probably want a separate EEPROM loader so if you hose your board you can fix it lol.

[aidanjt]

LinuxBIOS powering the One Laptop Per Child project:
YouTube Video

[Sinizter]

This LinuxBIOS thing has been floating around for a couple of years now hasnt it ?

[aidanjt]

Yup, it sure has, Cray has been using it on their XT3/4 series supercomputers, presumingly for it's efficient Hypertransport code and scalability. And it has been used on numerous other devices, including embedded stuff like TVs, PVRs and so on.

According to the LinuxBIOS website it's adoption has increased to over 1 million devices already, but it's still struggling to break into the motherboard market.

[Funkstar]

I seem to remember a lot of discussion of LinuxBIOS in relation to EPIA front end clients for media center applications like MythTV.

I haven't followed Myth or MiniMyth (http://www.linpvr.org/) for ages now, but there is some discussion on the topic on the LinPVR forums.

I really like the idea. Perfect for using as an advanced boot manager, minimal thin client, or as a media player using uPnP or RTSP. Combine this with loading apps off a flash drive or from a network share and things get really interesting.

[aidanjt]

Yup, the only real limit it has is that of space, but it squeezes a hell of a lot into a 2M image.

[kopite]

It sounds brilliant and if all the big companies got involved it could make Pc`s work faster.

Only problem I can see is bring opensource finding exploitations and ways to hack ppls PC`s could be easier...

[Funkstar]

Also, having a proper OS that isn't part of the storage media on a PC would have distinct advantages. For example, see this thread started the other day by Trig:

http://forums.hexus.net/showthread.php?t=100789

[aidanjt]

It sounds brilliant and if all the big companies got involved it could make Pc`s work faster.

Only problem I can see is bring opensource finding exploitations and ways to hack ppls PC`s could be easier...

That's FUD generated by propriatorists, Linux has no viruses, Windows has millions.

The 'many eyes' approach actually has the opposite effect with exploit turnarounds ranging from hours to a few days, compared to weeks to wontfixs plauging closed source operating systems, they just don't have the same kind of man power so they use 'security' through obscurity, which of course doesn't work.