Re: [PATCH v3 03/20] drm/dp: Move i2c init to drm_dp_aux_init, add __must_check and fini

From: Lyude Paul <lyude@redhat.com>
To: Thierry Reding <treding@nvidia.com>
Cc: "Neil Armstrong" <narmstrong@baylibre.com>,
	"David Airlie" <airlied@linux.ie>,
	nouveau@lists.freedesktop.org, "Guido Günther" <agx@sigxcpu.org>,
	"Joonas Lahtinen" <joonas.lahtinen@linux.intel.com>,
	"Oleg Vasilev" <oleg.vasilev@intel.com>,
	dri-devel@lists.freedesktop.org,
	"Meenakshikumar Somasundaram"
	<meenakshikumar.somasundaram@amd.com>,
	"Andrzej Hajda" <a.hajda@samsung.com>,
	"Aurabindo Pillai" <aurabindo.pillai@amd.com>,
	"Thierry Reding" <thierry.reding@gmail.com>,
	"Laurent Pinchart" <Laurent.pinchart@ideasonboard.com>,
	"Sam Ravnborg" <sam@ravnborg.org>,
	"Ville Syrjälä" <ville.syrjala@linux.intel.com>,
	"Tomi Valkeinen" <tomi.valkeinen@ti.com>,
	"Parshuram Thombare" <pthombar@cadence.com>,
	"Rodrigo Siqueira" <Rodrigo.Siqueira@amd.com>,
	"Michal Simek" <michal.simek@xilinx.com>,
	amd-gfx@lists.freedesktop.org,
	"Jonathan Hunter" <jonathanh@nvidia.com>,
	"Luben Tuikov" <luben.tuikov@amd.com>,
	"Ben Skeggs" <bskeggs@redhat.com>,
	"Swapnil Jakhade" <sjakhade@cadence.com>,
	"Harry Wentland" <harry.wentland@amd.com>,
	"Imre Deak" <imre.deak@intel.com>,
	"Thomas Zimmermann" <tzimmermann@suse.de>,
	"Jonas Karlman" <jonas@kwiboo.se>, "Leo Li" <sunpeng.li@amd.com>,
	intel-gfx@lists.freedesktop.org,
	"Maarten Lankhorst" <maarten.lankhorst@linux.intel.com>,
	"Jani Nikula" <jani.nikula@linux.intel.com>,
	"Joe Perches" <joe@perches.com>,
	"Yuti Amonkar" <yamonkar@cadence.com>,
	"Maxime Ripard" <mripard@kernel.org>,
	"Rodrigo Vivi" <rodrigo.vivi@intel.com>,
	"Mikita Lipski" <mikita.lipski@amd.com>,
	"Sean Paul" <sean@poorly.run>,
	"Navid Emamdoost" <navid.emamdoost@gmail.com>,
	"Jernej Skrabec" <jernej.skrabec@siol.net>,
	"Chris Park" <Chris.Park@amd.com>,
	"Eryk Brol" <eryk.brol@amd.com>,
	"Hyun Kwon" <hyun.kwon@xilinx.com>,
	"Robert Foss" <robert.foss@linaro.org>,
	"Julia Lawall" <Julia.Lawall@inria.fr>,
	"Rob Clark" <robdclark@gmail.com>,
	"Boris Brezillon" <boris.brezillon@collabora.com>,
	"Daniel Vetter" <daniel@ffwll.ch>,
	"Alex Deucher" <alexander.deucher@amd.com>,
	"Christian König" <christian.koenig@amd.com>
Subject: Re: [PATCH v3 03/20] drm/dp: Move i2c init to drm_dp_aux_init, add __must_check and fini
Date: Fri, 23 Apr 2021 13:53:33 -0400	[thread overview]
Message-ID: <cb7339fe172f31c615d824b56a709d2df6b79bb7.camel@redhat.com> (raw)
In-Reply-To: <YILABCrcMb5MksxD@orome.fritz.box>

On Fri, 2021-04-23 at 14:39 +0200, Thierry Reding wrote:
> 
> I'm curious: how is a DP AUX adapter reference count going to solve the
> issue of potentially registering devices too early (i.e. before the DRM
> is registered)?
> 
> Is it because registering too early could cause a reference count
> problem if somebody get a hold of the DP AUX adapter before the parent
> DRM device is around?

Well currently the problem is that we kind of want to avoid setting up the i2c
adapter before the DRM driver is registered with userspace, but it's not
really possible to do that if we need the core device struct for the ddc
adapter initialized so that tegra can call get_device() on it in
drivers/gpu/drm/tegra/sor.c. So my thought is instead of calling get_device()
on the ddc adapter that the AUX channel provides, why not just call it on the
actual platform device that implements the AUX channel instead? I think this
should work pretty nicely while still preventing the platform device for the
AUX channel from disappearing before the SOR has disappeared.

> 
> Thierry

-- 
Cheers,
 Lyude Paul (she/her)
 Software Engineer at Red Hat

_______________________________________________
amd-gfx mailing list
amd-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/amd-gfx