Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <20230607034331.576623-1-acelan.kao@canonical.com>
 <20230607042032.GA28835@srcf.ucam.org> <CAFv23QmDNUFcPwvSQt5aUxtmHasfr8wrF72ObvcO-X19gfn=LA@mail.gmail.com>
 <20230607052724.GA29834@srcf.ucam.org> <CAFv23QkEdGnEz1q7vbyFCa9S9Dqh-zec72nRGyZ3wAz-8wpbvA@mail.gmail.com>
 <20230607062341.GA30618@srcf.ucam.org> <20230607065604.yaivqbbd3dkawxo4@pali>
In-Reply-To: <20230607065604.yaivqbbd3dkawxo4@pali>
From:   AceLan Kao <acelan.kao@canonical.com>
Date:   Wed, 7 Jun 2023 15:39:33 +0800
Message-ID: <CAFv23Q==r4newMXE3OWavRSRt-bEi5-qR0Vo-5HGLw4r9J36MA@mail.gmail.com>
Subject: Re: [PATCH] platform/x86: dell-laptop: Add drm module soft dependency
To:     =?UTF-8?Q?Pali_Roh=C3=A1r?= <pali@kernel.org>
Cc:     Matthew Garrett <mjg59@srcf.ucam.org>,
        Hans de Goede <hdegoede@redhat.com>,
        Mark Gross <markgross@kernel.org>,
        platform-driver-x86@vger.kernel.org, linux-kernel@vger.kernel.org,
        Kai-Heng Feng <kai.heng.feng@canonical.com>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

Pali Roh=C3=A1r <pali@kernel.org> =E6=96=BC 2023=E5=B9=B46=E6=9C=887=E6=97=
=A5 =E9=80=B1=E4=B8=89 =E4=B8=8B=E5=8D=882:56=E5=AF=AB=E9=81=93=EF=BC=9A
>
> On Wednesday 07 June 2023 07:23:41 Matthew Garrett wrote:
> > On Wed, Jun 07, 2023 at 02:13:31PM +0800, AceLan Kao wrote:
> > > Matthew Garrett <mjg59@srcf.ucam.org> =E6=96=BC 2023=E5=B9=B46=E6=9C=
=887=E6=97=A5 =E9=80=B1=E4=B8=89 =E4=B8=8B=E5=8D=881:27=E5=AF=AB=E9=81=93=
=EF=BC=9A
> > > >
> > > > On Wed, Jun 07, 2023 at 01:19:40PM +0800, AceLan Kao wrote:
> > > > > Gfx drivers(i915/amdgpu/nvidia) depend on the drm driver, so dela=
ying
> > > > > the loading of dell_laptop after drm can ease the issue the most.
> > > > > Right, it's still possible to encounter the issue, unfortunately,=
 we
> > > > > do not have a better solution for it at the moment.
> > > >
> > > > We could unregister inappropriate backlight drivers when a more
> > > > appropriate one is loaded, or the policy decision around which driv=
er to
> > > > use could be made in userland?
> > > It's hard to decide which backlight driver is redundant, and it's kin=
d of ugly
> > > to unregister the backlight driver which is registered by other drive=
r and maybe
> > > problematic.
> >
> > But you're relying on registering the working backlight first, which is
> > an inherently racy thing? We shouldn't be relying on order of
> > initialisation to make this work, either we should only export a workin=
g
> > interface or we should expose enough information for whatever is using
> > the interfaces to make an appropriate policy decision itself.
>
> IIRC, drm drivers unregister redundant fbcon drivers (vesafb), so cannot
> drm drivers use similar strategy also for backlight drivers and
> unregister the redundant? If every backlight driver would have some
> "flag" which say if it should be unregistered by drm then maybe it could
> work? Or are there some other pitfalls?
Matthew,

What do you think if we unregister backlight devices if the backlight type
is larger than the current registered one.
Do this check in backlight_device_register() and unregister backlight
devices by the order raw(1) > platform(2) > firmware(3)
And maybe introduce a sticky bit into the backlight device if the backlight
driver doesn't want to be removed.

Pali,

No, it doesn't work by doing this in the drm driver if the backlight driver
is registered after the drm driver has been loaded.