Re: kernel firmwares: GR proposal
On Fri, Sep 08, 2006 at 11:20:43AM -0400, Raul Miller wrote:
> On 9/8/06, Sven Luther <firstname.lastname@example.org> wrote:
> >On Thu, Sep 07, 2006 at 05:08:28PM -0500, Bill Allombert wrote:
> >> On Fri, Sep 01, 2006 at 02:42:26PM -0400, Raul Miller wrote:
> >> > Perhaps we should start addressing the CD distributor problem (perhaps
> >> > tagging CD distributable software, and providing a simple mechanism to
> >> > pull only CD distributable software for
> >> > contrib/non-free).
> >> It will not work for firmware, be could be done for GFDL documentations:
> >Could you perhaps give us some insight as to why it will not work for
> >firmwares ?
> I think he means that for firmware there are additional technical
> As I understand it, currently the kernel incorporates the
> firmware image in the kernel module that implements
> the driver.
> The reason for this is that these drivers are used at the
> abstraction level where we are implementing file systems --
> the file system generally does not exist until after they
> are loaded. [There are exceptions, of course -- ram disk,
> for example.]
> This often means that that firmware needs to be included
> in the kernel source package, and that the module is built
> at the same time as the rest of the kernel.
> But note that I've not dealt with this in any detail, so I might
> have some details wrong.
The initramfs ramdisk is loaded right after cpu initialization, and the
firmware loading code can grab firmware from it, so there is no technical
reason for it not to work, apart that it needs some work, and people wanting
to do it.