Hello Maxime,
On 2024-07-10 09:13, Maxime Ripard wrote:
On Tue, Jul 09, 2024 at 06:36:08PM GMT, Dragan Simic wrote:
> > > > As I already wrote earlier, and as the above-linked discussions
> > > > conclude, solving these issues doesn't belong to any specific driver.
> > > > It should be resolved within the kernel's firmware loading mechanism
> > > > instead, and no driver should be specific in that regard.
> > >
> > > IT would be good if it can be resolved within the kernel's firmware
> > > loading mechanism.
> >
> > ... we'll need this as a systemic solution.
>
> The general policy has been to put drivers that need a firmware as a
> module, and just never build them statically.
I totally agree, but if Buildroot builds them statically and provides
no initial ramdisk, we need a better solution than having various
drivers
attempt to implement their own workarounds.
Buildroot typically allows custom kernel configurations, so it's not
really "enforcing" anything like another distro does.
It is definitely targetted towards very stripped down systems, so I
guess building the drivers statically is a natural choice, but it works
fine with modules too.
It all leads to a conclusion that we need better in-kernel support
for delayed firmware loading, instead of drivers implementing various
workarounds, for the layouts in which drivers are built statically
into the kernel image, but the required firmware blobs reside on the
root filesystem.
I'll start working on it, hopefully today. :)
