Nouveau Open-Source NVIDIA Driver Going Through A Big Rework

[duby229]

I don't have a problem with firmware that runs on a manufacturer's hardware. The alternative is to put the firmware into immutable ROM on the device, making it part of the hardware. There would be no blob to load, and all the software on the system would be open source. The downside, of course, is that if you find a bug you have to live with it, or buy new hardware. Even many CPUs have firmware these days.

Whether the firmware get's loaded from a ROM or a folder, it's still proprietary code. But it's what allows AMD to present hardware features in a way that can be used appropriately. There is absolutely nothing wrong with that. If it was open firmware that might be better, but at least our hardware is usable.

[Ancurio]

As far as I know, this firmware is proprietary, which means there are basically no fully free/libre graphics with latest-generation hardware (AMD, Intel [Skylake+], NVIDIA [Maxwell+] all require some kind of blobs on your PC).

But most FOSS advocates don't care too much because the blob doesn't run on their CPU, just the GPU. I think there was an effort underway to write open firmware for some card (not sure if radeon or nvidia), but it got abandoned due to lack of motivation. And really, it wouldn't offer you much anyway. Although it would be really interesting for learning how GPUs work internally.

[marek]

BTW, it's not an NVIDIA driver, it's a Red Hat driver for NVIDIA hardware

[chithanh]

But most FOSS advocates don't care too much because the blob doesn't run on their CPU, just the GPU.

Interestingly the line between those that reject/accept proprietary firmware runs largely along the "Free Software" vs. "Open Source" camps.

I think there was an effort underway to write open firmware for some card (not sure if radeon or nvidia), but it got abandoned due to lack of motivation.

There are several projects to replace the proprietary software that powers AMD graphics
https://github.com/alterapraxisptyltd/openatom OpenAtom, a free replacement for the video BIOS on Radeon cards
There is also a project to reverse engineer the Adreno microcode which turned out to be very similar to the R600 microcode.

[Nille_kungen]

Whether the firmware get's loaded from a ROM or a folder, it's still proprietary code. But it's what allows AMD to present hardware features in a way that can be used appropriately. There is absolutely nothing wrong with that. If it was open firmware that might be better, but at least our hardware is usable.

Yes the AMD microcode is rather small, this is for my system.

$ ls -lh /lib/firmware/radeon/{R600_*,hainan_*,ARUBA_*}
-rw-r--r-- 1 root root 8,5K aug 20 22:34 /lib/firmware/radeon/ARUBA_me.bin
-rw-r--r-- 1 root root 8,5K aug 20 22:34 /lib/firmware/radeon/ARUBA_pfp.bin
-rw-r--r-- 1 root root 6,0K aug 20 22:34 /lib/firmware/radeon/ARUBA_rlc.bin
-rw-r--r-- 1 root root  21K aug 20 22:34 /lib/firmware/radeon/R600_me.bin
-rw-r--r-- 1 root root 2,3K aug 20 22:34 /lib/firmware/radeon/R600_pfp.bin
-rw-r--r-- 1 root root 3,0K aug 20 22:34 /lib/firmware/radeon/R600_rlc.bin
-rw-r--r-- 1 root root  73K aug 20 22:34 /lib/firmware/radeon/R600_uvd.bin
-rw-r--r-- 1 root root 8,7K aug 20 22:34 /lib/firmware/radeon/hainan_ce.bin
-rw-r--r-- 1 root root  32K aug 20 22:34 /lib/firmware/radeon/hainan_mc.bin
-rw-r--r-- 1 root root 8,7K aug 20 22:34 /lib/firmware/radeon/hainan_me.bin
-rw-r--r-- 1 root root 8,7K aug 20 22:34 /lib/firmware/radeon/hainan_pfp.bin
-rw-r--r-- 1 root root 8,3K aug 20 22:34 /lib/firmware/radeon/hainan_rlc.bin
-rw-r--r-- 1 root root  61K aug 20 22:34 /lib/firmware/radeon/hainan_smc.bin

[Delgarde]

Interestingly the line between those that reject/accept proprietary firmware runs largely along the "Free Software" vs. "Open Source" camps.

Not all that surprising. In both cases, the distinction is essentially between pragmatism and idealism - both groups want both free software and working systems, but one group puts functional concerns ahead of strict principals, while the latter hold to principals even when it makes life difficult for them. Different priorities, basically.

[chithanh]

Yeah, that's why I said "interestingly" and not "suprisingly"
The free software camp rejecting all proprietary software (including firmware) for ethical reasons, even saying that stealing food is preferable to earning money with proprietary software.
Open source camp having no problem with firmware blobs in critical components, or TiVo-style locking up of hardware that the user supposedly owns.

[Delgarde]

The free software camp rejecting all proprietary software (including firmware) for ethical reasons, even saying that stealing food is preferable to earning money with proprietary software.
Open source camp having no problem with firmware blobs in critical components, or TiVo-style locking up of hardware that the user supposedly owns.

Though it's a spectrum, of course - most in the FSF camp aren't quite so zealous in their views as RMS. And most in the OS camp generally *do* have a problem with firmware blobs, but they'd rather live with that problem than with non-working hardware.

[kernelOfTruth]

Looking forward to the Re-Clocking changes,

hope they'll make it in time for 4.3 Cycle