Intel Skylake & Broxton To Require Graphics Firmware Blobs

[PuckPoltergeist]

It's an hindrance to those who want to run exclusively free/libre software.

They have to discard their x86 based computers. It always run some kind of binary only firmware, called microcode.

[moilami]

<insert some rant about human rights and freedom followed up with some RMS toe cheese and wine>

I was about to laugh out so loud that I had to keep cool, else people around might had thought I am crazy

[schmidtbag]

It's an hindrance to those who want to run exclusively free/libre software.
Also, remember that free/libre software is mostly a matter of “what if”, not of practical usage. Look, it's cool not to have handcuffs, you can drive and shift gears at the same time. Do you get the point now?

No, I legitimately don't get the point. Aside from legal reasons (such as redistribution) or just simply "hey check out what I can do with entirely free+open source software", what possible need do you have for an exclusively free/libre system? Proprietary software isn't handcuffing you. It's more like not having a passport, which for the vast majority of people, isn't going to ruin your experience. Being handcuffed is when you're stuck with closed source drivers that can't do what you need them to and the developers behind them refuse to help. For example, nvidia's drivers with wayland support or proper Optimus support. But, intel's drivers don't have such restrictions.

GPU drivers are not easy things to work with. If something goes wrong, you as a user are probably going to have no ability to fix the source code yourself. So who do you go to? The only obvious choice is Intel, because nobody else is going to know what to do either. But if Intel is doing all the work, why does it matter if they open source anything? Remember too - the only systems with intel graphics are intel processors or motherboards of the x86, x86-64, and MAYBE Itanium. Almost all of the benefits of having open source intel drivers gets thrown out the window because the hardware platform itself is so locked down.

Intel is usually the first company to meet with the demands of other open source software, such as Wayland or Mesa. If that weren't the case, maybe you'd be onto something. But really, since there is no actual problem with intel close-sourcing things, you're just whining because of the principle, and that's not a good reason.


Y'know what really bothers me? People who say they want a fully libre system and stop at free and open software. You're a hypocrite if you decide to go for intel at all. Go for an architecture like SPARC where even the CPU schematics are open source, and THEN you've got a legitimately FOSS system.

[amehaye]

As people said before me, there are several problems with closed-source blobs.

First, you can't fix bugs. You depend on the goodwill of the hardware manufacturer, and these tend to stop supporting 'old' hardware, where old could mean even hardware only 2-years old, as is common on many Android devices.

Second, and related to the first point, bugs could mean not just missing functionality but security vulnerabilities as well. For example there are known BIOS and even GPU-based rootkits based on closed-source firmware exploits. Remember that closed-source doesn't mean secure. Example: Windows.

Third, you can't add new functionality. For example your hardware might support OpenGLESv3 but the firmware shipped only supports GLESv2 and the product has reached it's end-of-life status.

Fourth, and actually not that important from my point of view, is the fear from 'NSA' or whatever surveillance.

All in all this is a step in a very wrong direction. I hope that Intel reverse this poor decision.

[MoonMoon]

As people said before me, there are several problems with closed-source blobs.

First, you can't fix bugs. You depend on the goodwill of the hardware manufacturer, and these tend to stop supporting 'old' hardware, where old could mean even hardware only 2-years old, as is common on many Android devices.

Second, and related to the first point, bugs could mean not just missing functionality but security vulnerabilities as well. For example there are known BIOS and even GPU-based rootkits based on closed-source firmware exploits. Remember that closed-source doesn't mean secure. Example: Windows.

Third, you can't add new functionality. For example your hardware might support OpenGLESv3 but the firmware shipped only supports GLESv2 and the product has reached it's end-of-life status.

Fourth, and actually not that important from my point of view, is the fear from 'NSA' or whatever surveillance.

All in all this is a step in a very wrong direction. I hope that Intel reverse this poor decision.

Exactly zero of these problems would have gone away if Intel had decided to put the blob in hardware (Flash, ROM, whatever), but nobody would be here complaining about that.

As I see it, this whole thread (with few exceptions) is nothing more than "whining because being able to whine".

[Gusar]

I hope that Intel reverse this poor decision.

Do you also hope AMD will reverse their decision to need blobs, which BTW has been the case for years? Do you also hope Nvidia will reverse their decision to use *signed* blobs, which makes it impossible to use open firmware even if one was created (like it was for earlier generation Nvidia GPUs)?

As I see it, this whole thread (with few exceptions) is nothing more than "whining because being able to whine".

Pretty much.

[Del_]

And for the person mentioning Qualcomm going to the "dark side" for requiring blobs for the ath10k driver... The ath9k and ath5k drivers aren't blob-free either. Sure, there's no file in the /lib/firmware directory. But instead, the source code contains tables and tables filled with magic numbers. Just check the driver source for "initvals" and you'll find these huge tables.

Luis Rodriguez gave us years of hard work to provide solid ath9k and ath5k open drivers, even ridding us of the dreaded blobs. Fast forward, he left Atheros and blobs are back. You are showing disrespect and ignorance. Both you and schmidtbag should read through the thread again and try to absorb all the information you have been provided.

[Zapitron]

what possible need do you have for an exclusively free/libre system?

Security. If the code is audit-able then it may be audited, so that people can determine whether their computer serves its owner or another party. If it's not auditable, then nobody has a way of knowing. Unintentional exploits have been found in drivers before, so we already had a pretty serious security risk with them, and there is a growing tension between people and governments about who should be in control of computers and communication.

Yes, a growing tension. Previously theoretical attacks continue to become practical and real attacks. (For example, a couple decades ago we used to "worry" (in a purely academic sense) about a worst-case scenario of SSL MitM being a possible thing that could happen, but in 2015 we know for sure that it really happens and that if you're not careful, it will happen.) We need to be making our computers more secure, not less secure, and that means a larger fraction of it should be auditable. Ideally 100% but if we can't get to 100% we should at least be inching closer that. A move in the less secure direction should be viewed with suspicion and loathing. Even if you tolerate it and reluctantly hand over your dollars, you should be unhappy about it.

And so..

Do you seriously think the GPU is a good place to add incriminating backdoors to your system?

Yes! Yes, I do. From an adversary's PoV, it's a good idea. We don't know they're doing it, but they should be trying to do that. It would be irrational of them to not make everyone's machines' exploitable. If they're not trying, then they're either not doing their jobs, or they're incompetent. Now, maybe they are incompetent but I don't know why we would want to gamble on that.

And: "incriminating?" Ok, we don't know it's happening, but let's ask "what if?" Just what consequences do you think there are, should someone be caught? Suppose they did get caught, and it turns out there's a deal (or more likely: coercion) between the US government and Intel. Right off the bat, you'd know that AMD and Nvidia would surely have been under the same pressure, so it wouldn't put Intel at a competitive disadvantage. We know from the AT&T retroactive immunity case, that Intel (and the other manufacturers) wouldn't suffer a liability disadvantage. So please, go on, tell me how it would be "incriminating." Were it to happen, what would be the downside for either Intel or a government pressuring them?

No, I'm not saying there's a backdoor in the blob. But it's a real risk and there's little reason there shouldn't be a backdoor. And a few years from now, it'll be even more likely, and a few years after that, even moreso. Running more unauditable blobs is the wrong direction for users to be going.

[robclark]

Why does this have to be in an blob when it could be open source.
Why trying to forbid "no reverse engineering, decompilation, or disassembly of this software is permitted." when it's allowed in EU by law and they can't forbid it?

jfyi, the license text for intel blobs (last I looked) was basically identical to radeon blobs.. not saying I like it, but seems to be a fact of life that as hw gets more complex and product refresh cycles stay the same, the hw vendors want to leave themselves some room to continue the dev/debug/fix cycle without having to respin the chips due to bugs..

[FuturePilot]

For example, nvidia's drivers with wayland support

They're working on it