An Effort Making An Open-Source Radeon Video BIOS

[eisenhart]

Can you post an update? On which cards it runs? What's the status? Do you use a public repository tree to work on it?

There isnt much progress atm. Im debugging windows code in VM.
Wiki Page created https://sourceforge.net/p/openradeonbios/wiki/Home/
Do you mean write access to other developer?
Source code: http://sourceforge.net/p/openradeonb...i/master/tree/

[oibaf]

Can you post an update? On which cards it runs? What's the status? Do you use a public repository tree to work on it?

[eisenhart]

Are you still working on this?

Im working again on it.
Trying to get windows driver working on it.
Any wishes/questions?

[eisenhart]

Would be a better thing if there was an opensource VBIOS for NVIDIA's cards instead

I hadnt a nvidia card (ok, an old 8600gts somewhere and GT430 now) and GPU passthrough didnt worked on nvidia.
Maybe a nouveau developer can easier create a VBIOS.
VBIOS isnt a fullfeatured video driver. Its an API for legacy BIOS/OS.
It musts only init a gpu(without command processor, shaders, etc. ) and memory.
Then it musts be able to set different video modes.
Bochs VGA BIOS code can be used to serve legacy int 10h API.

[oibaf]

Are you still working on this?

[DeepDayze]

What's the point?

Would be a better thing if there was an opensource VBIOS for NVIDIA's cards instead

[Caledar]

If it was constantly sending out packets it would be too obvious. What you need is it to only respond when a request is made of sort. Probably some kind of key. I think the chances are its there, just really really well hidden. But then again, if its not there, they'd probably have a few zero day exploits up their back pockets..

[Luke]

Wireshark would bust an NSA backdoor in a graphics card

First it cant hurt. 2nd NSA Scandal. could somebody say me if you coulld implement a nsa backdoor into such a firmware and 2nd question, if not, could you do it in a driver like the proprietary nvidia driverblob or the one from amd?

and another question is it possible to hinder a bios to make tcp connections and send data?

I would love to have a coreboot patched system but they are very rare, at the moment.


And no I dont think the radeon firmware is a big problem. the one from intel is also closedsource. I dont get the big difference rms does about if its flashable or not... antifeatures could be in the initial firmware if its possible.

I would love to hear something about it. I am no driver developer so I am not shure, I think a firmware of a grafic card should not be able to make the kernel send data into the internet right? could a full blob do that? could it have direct hw-access to other cards or something?

Here's what would happen if say, Nvidia were dumb enough to use their binary blob to carry a backdoor for the NSA. Let's assume that big blob could more easily do this than the tiny Intel (on chip) or ATI (on disk) firmwares or video card BIOS blobs. Somebody, somwhere, running Wireshark would notice the packets always going out to one server or a small group of servers, possibly because they didn't trust the blob (or the firmware,or their BIOS, or whatever). They could compare with and without the blob(or coreboot, etc) and quickly see the suspicious activity. Open source or closed, concealing network activity from all users is rather like hiding a back door from all users of an open-source program. Instead of source code, someone is surely looking at packets.

Nvidia and ATI would shit bricks if the NSA asked them to risk installing a surveillance program that could be detected by a random hacker from Anonymous, then revealed in all it's glory to their potential customers. All this assumes the graphics card can even get access to the network, the only thing I ever worried about was passwords echoed to the screen, an unnecessary risk that is always blocked by just about everyone. If Intel does this, it won't be in the graphics but elsewhere-and they'd still have Wireshark, etc to worry about.

[Adarion]

Even if it isn't that much yet, good news for me.
Binary blobs are obsolete.

Besides all the secret services around the world, just think of SMM. Or TR069 to configure your router remotely. There are a lot of dangerous things in the firmware, some maybe once built with a good intention but now prone to security holes and generally dangerous to the user and system. Open Firmware is more trustworthy in the long term, and once grown also more stable.

[oliver]

So if I understand correctly, it replaces the bios on the card, exports the same tables just as atombios did, you extract them from atombios and we still require the propriatary firmware blob in /lib/firmware?

I would love to use it, I should find some old videocards to start testing it, but what's are the supported cards? Yeah a faq would help a lot here

I think it can be the starting point for something beautiful and yeah, it's definitely worth it. Also, one blob gone is still one blob gone.