AMD Publishes New SB Register, Programming Docs

[phoronix]

Phoronix: AMD Publishes New SB Register, Programming Docs

While most of the time getting new documentation out of AMD is for their ATI graphics processors, today they have pushed out four documents that amount to several hundred pages of information covering their latest Southbridges. The AMD SB700/710/750 chipsets are now well documented in these NDA-free programming guides that also cover the registers for this hardware.Over at AMD Developer Central is the AMD SB700/710/750 Register Reference Guide, AMD SB700/710/750 BIOS Developerâ??s Guide, AMD SB700/710/750 Register Programming Requirements, and lastly is the AMD SB710 Databook.Sure, programming guides and register descriptions for a southbridge is not nearly as exciting as for the latest ATI GPU, but this information is what the CoreBoot developers (the same ones working on Flashrom) have been after for a while...

http://www.phoronix.com/vr.php?view=NzM3Ng

[mirv]

cheers AMD!

[Pfanne]

now we only need a company that focuses on tinkering a complete oss experience arround amd hardware...
something like the open source apple with hard+software that works together perfectly.
so you can get your pear (my imaginary company's name) computer and you know itll work perfectly with linux...
unlike other companies like acer or asus they dont have to sell crippled hardware with their linux computers, because they cant lose any microsoft partnership contracts....
haaaaaaa, in my imaginary world everything would be sooooooooooo fine...

[DanL]

so you can get your pear (my imaginary company's name) computer

The Pear Computers name has already been used in multiple places as a way to imply "Apple" without copyright infringement. Try again.

[snogglethorpe]

Thanks AMD!

It's so nice to finally see more major hardware companies actually getting the point that widespread knowledge of how to use your hardware is good, and helps you sell more chips...

I work for a big semiconductor company, and have met many older managers who simply cannot come to terms with idea of giving anything away without strict conditions and renumeration, even if restricting information in the end hurts the company far more than any advantage it confers. There's a sort of ingrained conservatism and downright control-freakery that seems very hard to get rid of except though generational change.

[bridgman]

Here it's mostly us old guys driving the open source support :D

[Louise]

This is so cool!

CoreBoot really is an important project that many don't care much about. Apparently also Michael ;)

But it really is important to have an open source BIOS, as it is possible for the BIOS vendors to include a secret hyper visor that in "best case" "only" monitors what you do.

In "worst" case, it can forbid you from doing some things.

So CoreBoot really is important, and anything that helps that is exceeding good news :)

Also with CoreBoot you would be able to encrypt your harddrive, and have CoreBoot ask for the passphrase.

[deanjo]

But it really is important to have an open source BIOS, as it is possible for the BIOS vendors to include a secret hyper visor that in "best case" "only" monitors what you do.

In "worst" case, it can forbid you from doing some things.


Heh, that's just plain paranoia talking.

[Kano]

Working Coreboot on latest AMD boards would be very interesting for me - those docs will certainly help that. Intel docs seem to be very outdated, only very old chipsets are supported. As you can add custom payloads it should be really fun to use, hopefully the bios chip is on a socket ;)

[unix_epoch]

The most annoying thing about proprietary BIOSes in my opinion is that they are completely unpredictable in their use of SMIs (the so-called ring -2 that people get paranoid about). My old Sony laptop had random apparent hangs for as long as 10-20ms (fatal to the audio project I was trying to use it for). Downgrading its BIOS fixed the problem, which was caused by system management interrupts taking too long to give control back to the OS.

SMIs can be used for good, of course, like updating ACPI data, monitoring CPU temperature, and adjusting fan speed, but in a real operating system, I'd much rather do that using a low-priority preemptible thread than a realtime priority non-preemptible SMI.

[mirv]

The most annoying thing about proprietary BIOSes in my opinion is that they are completely unpredictable in their use of SMIs (the so-called ring -2 that people get paranoid about). My old Sony laptop had random apparent hangs for as long as 10-20ms (fatal to the audio project I was trying to use it for). Downgrading its BIOS fixed the problem, which was caused by system management interrupts taking too long to give control back to the OS.

SMIs can be used for good, of course, like updating ACPI data, monitoring CPU temperature, and adjusting fan speed, but in a real operating system, I'd much rather do that using a low-priority preemptible thread than a realtime priority non-preemptible SMI.

I know this is really off-topic, but monitoring CPU temperature should always be possible, independent of the OS - don't want a crashing OS to stop temperature monitoring and other things like that.
Back on-topic, it's doubtful I'll ever get into using these docs directly, but it's great to see AMD pushing out this sort of thing for those who will be using it!

[RoboJ1M]

Here it's mostly us old guys driving the open source support :D

Gandalf? 8D

Yay! I bought an SB750 board! With onboard Radeon 3300! OSS everyrthing!

J1M.

[Louise]

Heh, that's just plain paranoia talking.

You are probably right. Microsoft would never want to forbid Linux from running on PC's.

http://epic.org/privacy/consumer/microsoft/palladium.html

http://en.wikipedia.org/wiki/Trusted_Computing

[hakl]

I would definitely get an AMD based motherboard if it meant a free BIOS. Found this on Google: http://www.coreboot.org/pipermail/coreboot/2008-September/039094.html

[rohcQaH]

But it really is important to have an open source BIOS, as it is possible for the BIOS vendors to include a secret hyper visor that in "best case" "only" monitors what you do.
If someone wanted to do something like that really bad, they'd just implement it in hardware. (Palladium relies on custom TPM chips anyway).

Your gfx card most likely already supports HDCP, a technology designed to restrict your actions. Is an open source BIOS going to help you against that? No.

You'd need "Open Source Hardware" for that. But while everyone can get a compiler, not everyone can get his own fab, so that's a pretty pointless discussion.

[Louise]

You'd need "Open Source Hardware" for that. But while everyone can get a compiler, not everyone can get his own fab, so that's a pretty pointless discussion.
Like any hardware, the evil doers need to be able to upgrade the DRM once in a while, as hackers will find vulnerabilities.

Making it in silicon means that they can't upgrade it, and they would end up with millions of useless chips every time a vulnerabilities was found.

This is not something I invent to be right. Check out the Coreboot presentation from 25c3:

http://events.ccc.de/congress/2008/Fahrplan/events/2970.en.html

Can't remember the link for the download site, but search and you will find :)

[Fixxer_Linux]

coreboot is a nice project, but too much in darkness. I really wish I could switch to coreboot.
I introduced a customer question in gigabyte customer feedback service, to request them publishing some details to coreboot team for my motherboard (an EP45-DS4 model).
They replied kindly that this would studied, which is the polite way of saying fuck.
That's a pity as gigabyte as already pushed out one (not two, just one) motherboard (an Athlon X2 motherboard), and that now coreboot can handle perfectly this mobo.
I don't know why they didn't did it again. Probably Gigabyte has done that just to pressure BIOS vendors for some price drop or something else. Since, the closed-source BIOS manufacturer have corrected the problems or have dropped their prices and therefore, Gigabyte doesn't feel the need of throwing another mobo to coreboot devs. But this only supposition from myself.

Does intel publish some documentation as AMD has given today ? I guess so.

[Kano]

I guess that's just NOT the case when you look at supported Coreboot chipsets.

[jeffro-tull]

I was planning on going AMD when I went AM3 anyway (nevermind that they're practically the only game in town for that), but this is icing on the cake.

Keep up the good work!

[lbcoder]

Though it may support it, that doesn't mean that you need to USE it. The frightening thing about mobo-based TC is that it might violate your security before the OS can even boot -- it *does* have a network card right? Coreboot won't do this, and at least with a display card, you can simply IGNORE the TC/DRM crap (current versions anyways) and it'll keep working as normal, just won't work with the other TC/DRM crap that you might try to feed into it. It won't be able to mess with your network or do any other kinds of weird things since it simply doesn't have access -- its a one way thing, from the mobo into the vid card and out to the TV.

The bad news is that there are some extremely frightening things in the pipes that are sure to enforce TC crap by shutting down or self-destructing if it finds itself in a non-compliant environment, and the even scarier part is that most of the MS-lemmings will accept it out of ignorance, which leaves everyone interested in their own personal rights and freedoms out in the cold.

You bring up a valid point about open sourced hardware and everyone having their own personal fab's. The funny thing is that there are certainly enough people out there interested in freedom that at some point, we MUST band together to build it. I just hope that by the time we need it, that the legislation doesn't enforce the use of TC crap. Hint: AMD DO THIS! You don't need to do it on everything, but make some things with this option where there is no TC crap built in (or even built in and disabled) and provide FULL design specifications and open source it. Keep selling things to the MS-lemmings, but when the s**t hits the fans and there is no turning back for anyone else, you'll already have the ability to provide the freedom that is sure to be in high demand.

It really makes you think... maybe the terrorists have it right. Don't get me wrong, their methods are definitely wrong, but they seem to have picked the right enemy to fight. The US government, with their DMCA "treat EVERYONE as a terrorist/criminal" crap NEEDS to be seriously rethought. Back in the old days the rules were that you would be "assumed innocent until proven guilty", now its "assumed potential criminal".


If someone wanted to do something like that really bad, they'd just implement it in hardware. (Palladium relies on custom TPM chips anyway).

Your gfx card most likely already supports HDCP, a technology designed to restrict your actions. Is an open source BIOS going to help you against that? No.

You'd need "Open Source Hardware" for that. But while everyone can get a compiler, not everyone can get his own fab, so that's a pretty pointless discussion.

[unix_epoch]

...DMCA "treat EVERYONE as a terrorist/criminal" crap NEEDS to be seriously rethought

Thanks for your interesting message, and I agree with you. However, placing the quotation after your response was rather odd, and made understanding the context of the message slightly more difficult. Top posting can make sense in a corporate e-mail thread, but on a web-based forum that records all messages for everyone, it hinders readability. (Of course, if most people here disagree with me, I'll gladly accept the majority will)

[nanonyme]

Imo the neatest part about open BIOS's is that BIOS's sometimes have bugs which the vendor will never fix. Then you need to have kernel hacks to workaround. If the BIOS is open, developers can keep working on it until it Just Works (tm).

[lbcoder]

Response should ALWAYS be above the quote, otherwise you have to scroll through a mile of crap to get to the new stuff.

Thanks for your interesting message, and I agree with you. However, placing the quotation after your response was rather odd, and made understanding the context of the message slightly more difficult. Top posting can make sense in a corporate e-mail thread, but on a web-based forum that records all messages for everyone, it hinders readability. (Of course, if most people here disagree with me, I'll gladly accept the majority will)

[nanonyme]

Response should ALWAYS be above the quote, otherwise you have to scroll through a mile of crap to get to the new stuff.Gods, please don't start another flamewar. Top-posting vs bottom-posting is a bit similar to vim vs. Emacs. (Personally I think they're both wrong, you should only be replying to the sentences you want to reply to, not the whole damn thread (which happens often with newsletters) As in, the "quote, answer, quote answer, quote answer" style where you pull the quotes from arbitrary depths from the thread. No sense quoting if you're not even going to comment on everything that was said! ;) )

[Louise]

Just watched the latest video from coreboot.org... He got to be the worst speaker of all time. Boredom can't kill you, but sometimes you'd wish it could :D

Anyway, I was surprised to hear why the name change from LinuxBIOS to CoreBoot -- and I will spare you the horror to watch it yourself:

Original the goal was to fit the entire Linux kernel in the flash, but as BIOS flash didn't get any larger, that could not be done.

So Core Boot will initialize the hardware, and boot a payload. http://www.coreboot.org/Payloads

So CoreBoot is not a BIOS, where you can do over clocking, setting the clock and stuff like that.

So I am wondering, can that be done from the OS?

[curaga]

I asked about overclocking in #coreboot IRC - they say it will be a part of coreboot at some point.

[Louise]

I asked about overclocking in #coreboot IRC - they say it will be a part of coreboot at some point.

Okay, pretty cool! :)

I was fairly sure that CoreBoot would never be main stream on desktop mainboards.

Just looked in my BIOS, and the only thing I need and can't live without is being able to set Cool'n'Quite to begin at 45 degress, so the CPU fan spins at the lowest below 45 degrees.

Besides PXE, all the other stuff in the BIOS: Don't need it. :)

[Kano]

At least you find bios editors for standard systems to add gpxe. Did that successfully on 2 Asrock SiS based socket A boards. The default PXE loader did not work at all for me, not even with gpxelinux.

[Louise]

At least you find bios editors for standard systems to add gpxe. Did that successfully on 2 Asrock SiS based socket A boards. The default PXE loader did not work at all for me, not even with gpxelinux.

What is PXELinux? A BIOS replacement?

[Kano]

No, pxelinux.0 is part of syslinux. Newer syslinux versions also have got gpxelinux.0 which is basically the pxe core of gpxe added in front of pxelinux.0 - that can help partly broken pxe loaders, as you only have to replace that file. With that i could boot from a skge card which is only supported in an old private branch of gpxe (which somebody merged into mainline for me for testing). Without that the loader was too broken to boot.

[Adarion]

Ah, nice to read. Since I use a SB7xx. Very nice.

[alazyworkaholic]

I have an AMD SB700. Does this news matter to anyone besides coreboot users? When I ran Ubuntu 8.04 my computer had no problems. Now with 9.04 I get "softreset error"s while the computer boots, but they don't hurt except for a 10 second pause before things continue as normal.
Will this documentation improve the way Linux works with my motherboard?

[bugmenot]

This is AWESOME!

Best news for a long time. Thanks AMD! :)

http://www.coreboot.org/pipermail/coreboot/2009-July/050625.html

[hargut]

I have an AMD SB700. Does this news matter to anyone besides coreboot users? When I ran Ubuntu 8.04 my computer had no problems. Now with 9.04 I get "softreset error"s while the computer boots, but they don't hurt except for a 10 second pause before things continue as normal.
Will this documentation improve the way Linux works with my motherboard?

The "softreset errors" are mostly caused by some ATA device and often just mean "right now the disk is not ready".
After a few seconds or less than a second, when the drive is ready, the boot continues.
Nothing to worry about.

Regards,
Harald

[storma]

Remove the sata controller Port Multiplier (PMP) support from the kernel to fix this.