Subject: [PATCH] drm/mediatek: hdmi: Perform NULL pointer check for mtk_hdmi_conf

In commit 41ca9caaae0b ("drm/mediatek: hdmi: Add check for CEA modes only")
a check for CEA modes was added to function mtk_hdmi_bridge_mode_valid()
in order to address possible issues on MT8167; moreover, with commit
c91026a938c2 ("drm/mediatek: hdmi: Add optional limit on maximal HDMI mode clock")
another similar check was introduced.

Unfortunately though, at the time of writing, MT8173 does not provide
any mtk_hdmi_conf structure and this is crashing the kernel with NULL
pointer upon entering mtk_hdmi_bridge_mode_valid(), which happens as
soon as a HDMI cable gets plugged in.

To fix this regression, add a NULL pointer check for hdmi->conf in the
said function, restoring HDMI functionality and avoiding NULL pointer
kernel panics.

Fixes: 41ca9caaae0b ("drm/mediatek: hdmi: Add check for CEA modes only")
Fixes: c91026a938c2 ("drm/mediatek: hdmi: Add optional limit on maximal HDMI mode clock")
Signed-off-by: AngeloGioacchino Del Regno <[email protected]>
---
drivers/gpu/drm/mediatek/mtk_hdmi.c | 12 +++++++-----
1 file changed, 7 insertions(+), 5 deletions(-)

diff --git a/drivers/gpu/drm/mediatek/mtk_hdmi.c b/drivers/gpu/drm/mediatek/mtk_hdmi.c
index 5838c44cbf6f..3196189429bc 100644
--- a/drivers/gpu/drm/mediatek/mtk_hdmi.c
+++ b/drivers/gpu/drm/mediatek/mtk_hdmi.c
@@ -1224,12 +1224,14 @@ static int mtk_hdmi_bridge_mode_valid(struct drm_bridge *bridge,
return MODE_BAD;
}

- if (hdmi->conf->cea_modes_only && !drm_match_cea_mode(mode))
- return MODE_BAD;
+ if (hdmi->conf) {
+ if (hdmi->conf->cea_modes_only && !drm_match_cea_mode(mode))
+ return MODE_BAD;

- if (hdmi->conf->max_mode_clock &&
- mode->clock > hdmi->conf->max_mode_clock)
- return MODE_CLOCK_HIGH;
+ if (hdmi->conf->max_mode_clock &&
+ mode->clock > hdmi->conf->max_mode_clock)
+ return MODE_CLOCK_HIGH;
+ }

if (mode->clock < 27000)
return MODE_CLOCK_LOW;
--
2.33.0


Subject: Re: [PATCH] drm/mediatek: hdmi: Perform NULL pointer check for mtk_hdmi_conf

Il 28/10/21 09:43, AngeloGioacchino Del Regno ha scritto:
> In commit 41ca9caaae0b ("drm/mediatek: hdmi: Add check for CEA modes only")
> a check for CEA modes was added to function mtk_hdmi_bridge_mode_valid()
> in order to address possible issues on MT8167; moreover, with commit
> c91026a938c2 ("drm/mediatek: hdmi: Add optional limit on maximal HDMI mode clock")
> another similar check was introduced.
>
> Unfortunately though, at the time of writing, MT8173 does not provide
> any mtk_hdmi_conf structure and this is crashing the kernel with NULL
> pointer upon entering mtk_hdmi_bridge_mode_valid(), which happens as
> soon as a HDMI cable gets plugged in.
>
> To fix this regression, add a NULL pointer check for hdmi->conf in the
> said function, restoring HDMI functionality and avoiding NULL pointer
> kernel panics.
>
> Fixes: 41ca9caaae0b ("drm/mediatek: hdmi: Add check for CEA modes only")
> Fixes: c91026a938c2 ("drm/mediatek: hdmi: Add optional limit on maximal HDMI mode clock")
> Signed-off-by: AngeloGioacchino Del Regno <[email protected]>
> ---
> drivers/gpu/drm/mediatek/mtk_hdmi.c | 12 +++++++-----
> 1 file changed, 7 insertions(+), 5 deletions(-)
>
> diff --git a/drivers/gpu/drm/mediatek/mtk_hdmi.c b/drivers/gpu/drm/mediatek/mtk_hdmi.c
> index 5838c44cbf6f..3196189429bc 100644
> --- a/drivers/gpu/drm/mediatek/mtk_hdmi.c
> +++ b/drivers/gpu/drm/mediatek/mtk_hdmi.c
> @@ -1224,12 +1224,14 @@ static int mtk_hdmi_bridge_mode_valid(struct drm_bridge *bridge,
> return MODE_BAD;
> }
>
> - if (hdmi->conf->cea_modes_only && !drm_match_cea_mode(mode))
> - return MODE_BAD;
> + if (hdmi->conf) {
> + if (hdmi->conf->cea_modes_only && !drm_match_cea_mode(mode))
> + return MODE_BAD;
>
> - if (hdmi->conf->max_mode_clock &&
> - mode->clock > hdmi->conf->max_mode_clock)
> - return MODE_CLOCK_HIGH;
> + if (hdmi->conf->max_mode_clock &&
> + mode->clock > hdmi->conf->max_mode_clock)
> + return MODE_CLOCK_HIGH;
> + }
>
> if (mode->clock < 27000)
> return MODE_CLOCK_LOW;
>

Hello, friendly ping!

Can I please get a review (or merge) of this commit?

This is an important fix for a null pointer KP and restores HDMI functionality.

Thanks,
- Angelo

2021-12-13 13:08:33

by Chun-Kuang Hu

[permalink] [raw]
Subject: Re: [PATCH] drm/mediatek: hdmi: Perform NULL pointer check for mtk_hdmi_conf

Hi, Angelo:

AngeloGioacchino Del Regno <[email protected]> 於
2021年12月13日 週一 下午6:02寫道：
>
> Il 28/10/21 09:43, AngeloGioacchino Del Regno ha scritto:
> > In commit 41ca9caaae0b ("drm/mediatek: hdmi: Add check for CEA modes only")
> > a check for CEA modes was added to function mtk_hdmi_bridge_mode_valid()
> > in order to address possible issues on MT8167; moreover, with commit
> > c91026a938c2 ("drm/mediatek: hdmi: Add optional limit on maximal HDMI mode clock")
> > another similar check was introduced.
> >
> > Unfortunately though, at the time of writing, MT8173 does not provide
> > any mtk_hdmi_conf structure and this is crashing the kernel with NULL
> > pointer upon entering mtk_hdmi_bridge_mode_valid(), which happens as
> > soon as a HDMI cable gets plugged in.
> >
> > To fix this regression, add a NULL pointer check for hdmi->conf in the
> > said function, restoring HDMI functionality and avoiding NULL pointer
> > kernel panics.
> >
> > Fixes: 41ca9caaae0b ("drm/mediatek: hdmi: Add check for CEA modes only")
> > Fixes: c91026a938c2 ("drm/mediatek: hdmi: Add optional limit on maximal HDMI mode clock")
> > Signed-off-by: AngeloGioacchino Del Regno <[email protected]>
> > ---
> > drivers/gpu/drm/mediatek/mtk_hdmi.c | 12 +++++++-----
> > 1 file changed, 7 insertions(+), 5 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/mediatek/mtk_hdmi.c b/drivers/gpu/drm/mediatek/mtk_hdmi.c
> > index 5838c44cbf6f..3196189429bc 100644
> > --- a/drivers/gpu/drm/mediatek/mtk_hdmi.c
> > +++ b/drivers/gpu/drm/mediatek/mtk_hdmi.c
> > @@ -1224,12 +1224,14 @@ static int mtk_hdmi_bridge_mode_valid(struct drm_bridge *bridge,
> > return MODE_BAD;
> > }
> >
> > - if (hdmi->conf->cea_modes_only && !drm_match_cea_mode(mode))
> > - return MODE_BAD;
> > + if (hdmi->conf) {
> > + if (hdmi->conf->cea_modes_only && !drm_match_cea_mode(mode))
> > + return MODE_BAD;
> >
> > - if (hdmi->conf->max_mode_clock &&
> > - mode->clock > hdmi->conf->max_mode_clock)
> > - return MODE_CLOCK_HIGH;
> > + if (hdmi->conf->max_mode_clock &&
> > + mode->clock > hdmi->conf->max_mode_clock)
> > + return MODE_CLOCK_HIGH;
> > + }
> >
> > if (mode->clock < 27000)
> > return MODE_CLOCK_LOW;
> >
>
> Hello, friendly ping!
>
> Can I please get a review (or merge) of this commit?
>
> This is an important fix for a null pointer KP and restores HDMI functionality.

Applied to mediatek-drm-fixes [1], thanks.

[1] https://git.kernel.org/pub/scm/linux/kernel/git/chunkuang.hu/linux.git/log/?h=mediatek-drm-fixes

Regards,
Chun-Kuang.

>
> Thanks,
> - Angelo