free kernel and 3d graphics acceleration

Michał Masłowski mtjm at
Fri Sep 28 07:19:38 UTC 2012


> I do not run any non-free application software
> (for example, Java) on my desktop or servers, and I use a free
> (de-blobbed) kernel.

There are free implementations of Java using OpenJDK.

> The most difficult and frustrating part of
> following this choice is that, as far as I can tell, there is no way at
> all to have 3d graphics acceleration and use only free software. (I
> would be glad to be corrected if I am wrong!) I am unable to play any 3d
> video games, and I cannot do any practical 3d programming, both of which
> are things I very much want to do.

A popular way is to use Intel CPUs and GPUs, they have free drivers (and
a manufacturer opposed to free BIOSes).  Some Nvidia cards have some
support in Nouveau.

> As
> far as I know, the free software community has every component necessary
> to have 3d graphics acceleration using Radeon cards, except for one:
> free firmware.

It has also nonfree video BIOS needed for modesetting.  (Intel drivers
don't need these.)  This will be a problem with Lemote machines having
otherwise free boot firmware.

> However, the strange thing about this is that, in the
> Linux kernel, the binary blob has a free software license attached to
> it! (See firmware/WHENCE; it appears to be an MIT-style license.) Why
> did AMD release their driver under a FOSS-compatible license, but not
> include the source code? Was this simply overlooked?

It wasn't overlooked, they used this license probably since it was
originally a binary array in a MIT-style licensed source file.
Releasing their source wouldn't solve the problem since new firmware
(R6xx, R7xx RLC and anything newer than R7xx) is released under a
different, nonfree license (and is in the linux-firmware repository).

> Has the FSF, or anyone else, ever asked AMD to release the source code
> for the driver? Perhaps they would...? I could ask, but I am nobody and
> have no influence. Perhaps if an important organization like FSF were to
> demand it, then maybe they would release the source...?

They certainly have discussed this.  Arguments from AMD against
releasing it was that no other vendor does it, it would make breaking
their digital restrictions management easier, it is a part of the
hardware, legal checking of its documentation would be very expensive.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 835 bytes
Desc: not available
URL: <>

More information about the linux-libre mailing list