Re: [Nouveau] [PATCH 03/30] drm/tegra: Don't register DP AUX channels before connectors

From: Thierry Reding <thierry.reding@gmail.com>
To: Lyude Paul <lyude@redhat.com>
Cc: dri-devel@lists.freedesktop.org,
	"David Airlie" <airlied@linux.ie>,
	nouveau@lists.freedesktop.org, intel-gfx@lists.freedesktop.org,
	"open list" <linux-kernel@vger.kernel.org>,
	"Jani Nikula" <jani.nikula@linux.intel.com>,
	amd-gfx@lists.freedesktop.org,
	"Rodrigo Vivi" <rodrigo.vivi@intel.com>,
	"open list:DRM DRIVERS FOR NVIDIA TEGRA"
	<linux-tegra@vger.kernel.org>, "Daniel Vetter" <daniel@ffwll.ch>,
	"Ville Syrjälä" <ville.syrjala@linux.intel.com>
Subject: Re: [Nouveau] [PATCH 03/30] drm/tegra: Don't register DP AUX channels before connectors
Date: Wed, 14 Apr 2021 18:49:12 +0200	[thread overview]
Message-ID: <YHcdCAQ7Dlb2vSzn@orome.fritz.box> (raw)
In-Reply-To: <20210219215326.2227596-4-lyude@redhat.com>

[-- Attachment #1.1: Type: text/plain, Size: 2612 bytes --]

On Fri, Feb 19, 2021 at 04:52:59PM -0500, Lyude Paul wrote:
> As pointed out by the documentation for drm_dp_aux_register(),
> drm_dp_aux_init() should be used in situations where the AUX channel for a
> display driver can potentially be registered before it's respective DRM
> driver. This is the case with Tegra, since the DP aux channel exists as a
> platform device instead of being a grandchild of the DRM device.
> 
> Since we're about to add a backpointer to a DP AUX channel's respective DRM
> device, let's fix this so that we don't potentially allow userspace to use
> the AUX channel before we've associated it with it's DRM connector.
> 
> Signed-off-by: Lyude Paul <lyude@redhat.com>
> ---
>  drivers/gpu/drm/tegra/dpaux.c | 11 ++++++-----
>  1 file changed, 6 insertions(+), 5 deletions(-)
> 
> diff --git a/drivers/gpu/drm/tegra/dpaux.c b/drivers/gpu/drm/tegra/dpaux.c
> index 105fb9cdbb3b..ea56c6ec25e4 100644
> --- a/drivers/gpu/drm/tegra/dpaux.c
> +++ b/drivers/gpu/drm/tegra/dpaux.c
> @@ -534,9 +534,7 @@ static int tegra_dpaux_probe(struct platform_device *pdev)
>  	dpaux->aux.transfer = tegra_dpaux_transfer;
>  	dpaux->aux.dev = &pdev->dev;
>  
> -	err = drm_dp_aux_register(&dpaux->aux);
> -	if (err < 0)
> -		return err;
> +	drm_dp_aux_init(&dpaux->aux);

I just noticed that this change causes an error on some setups that I
haven't seen before. The problem is that the SOR driver tries to grab a
reference to the I2C device to make sure it doesn't go away while it has
a pointer to it.

However, since now the I2C adapter hasn't been registered yet, I get
this:

	[   15.013969] kobject: '(null)' (000000005c903e43): is not initialized, yet kobject_get() is being called.

I recall that you wanted to make this change so that a backpointer to
the DRM device could be added (I think that's patch 15 of the series),
but I didn't see that patch get merged, so it's a bit difficult to try
and fix this up.

Has the situation changed? Do we no longer need the backpointer? If we
still want it, what's the plan for merging the change? Should I work
under the assumption that patch will make it in sometime and try to fix
this on top of that?

I'm thinking that perhaps we can move the I2C adapter registration into
drm_dp_aux_init() since that's independent of the DRM device. It would
also make a bit more sense from the Tegra driver's point of view where
all devices would be created during the ->probe() path, and only during
the ->init() path would the connection between DRM device and DRM DP AUX
device be established.

Thierry

[-- Attachment #1.2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

[-- Attachment #2: Type: text/plain, Size: 154 bytes --]

_______________________________________________
Nouveau mailing list
Nouveau@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/nouveau