[Intel-gfx] [PATCH 1/1] firmware/dmc/icl: load v1.07 on icelake.

Joonas Lahtinen joonas.lahtinen at linux.intel.com
Fri Sep 7 07:47:08 UTC 2018


Quoting Rodrigo Vivi (2018-09-06 20:35:19)
> On Thu, Sep 06, 2018 at 09:46:13AM +0300, Jani Nikula wrote:
> > On Wed, 05 Sep 2018, Rodrigo Vivi <rodrigo.vivi at intel.com> wrote:
> > > On Wed, Sep 05, 2018 at 12:07:43PM +0300, Joonas Lahtinen wrote:
> > >> Was not the decision that we only gate the MODULE_FIRMWARE line until
> > >> the firmware is in linux-firmware.git?
> > >> 
> > >> So it should be no harm to support loading firmwares that are available
> > >> from drm-firmware.git as long as we don't add the MODULE_FIRMWARE which
> > >> would trigger false positives for distro packagers.
> > >
> > > As far as I can remember we had agreed on changing this and adding
> > > MODULE_FIRMWARE from the beginning.
> > >
> > > 1. the process gets complex
> > > 2. we forgot many times to add it afterwards
> > > 3. module_firmware changes nothing... only the fact that initrd generation
> > >    wont complain if firmware is not there yet.
> > > 4. The old issue was that patches were merged without the pull request
> > >    being sent... We fixed that by only accepting patches after pull request
> > >    is sent.
> > > 5. By the time that all lands to distros linux-firmware.git will have
> > > the firmware because of 4.
> > > 6. We have now drm-firmware to mirror what official linux-firmware.git will
> > > be in few weeks from now.
> > >
> > > So I don't see a reason anymore why to keep with complicated process with
> > > split MODULE_FIRMWARE.
> > 
> > I've changed my mind, I agree with *not* splitting MODULE_FIRMWARE
> > changes to another patch anymore.
> > 
> > However, with that I think we need to redefine when we can add
> > request_firmware and MODULE_FIRMWARE for a specific firmware
> > blob. Here's my proposal.
> > 
> > - Before merging a patch adding or changing a firmware blob in dinq,
> >   said blob must be publicly available at a known location, and no known
> >   blockers for linux-firmware.git merge may exist.
> > 
> > - We take care to get the blob to linux-firmware.git before the changes
> >   hit Linus' -rc1 after the merge window. With our -rc5 feature
> >   deadline, this should give us at least 4-5 weeks to get the blob
> >   merged to linux-firmware.git.
> > 
> > - It follows that linux-next and drm-tip may contain MODULE_FIRMWARE for
> >   blobs that don't exist in linux-firmware.git.
> 
> Agreed. And we should document this somewhere.

Well, there are still quite a few conditionals included, if you ask me.

But if you see the MODULE_FIRMWARE as a separate patch as such a burden
that we should make the assumptions, I can live with it. At least I know
who to call in case there was one too many conditional and we get a
distro complaint again :)

Regards, Joonas

> 
> > 
> > BR,
> > Jani.
> > 
> > -- 
> > Jani Nikula, Intel Open Source Graphics Center


More information about the Intel-gfx mailing list