[PATCH v1] drm/bridge: inno-hdmi: reset DDC segment pointer per I2C transfer

[PATCH v1] drm/bridge: inno-hdmi: reset DDC segment pointer per I2C transfer

[Dominique Belhachemi]

The segment_addr and ddc_addr fields need to be reset per I2C transfer
to avoid stale values. The reset wasn't required on simple displays with
only one segment, but modern displays with two or more segments end up
with corrupt EDIDs.

This was found on a StarFive VisionFive 2 SBC.

Signed-off-by: Dominique Belhachemi <domibel@debian.org>
---
 drivers/gpu/drm/bridge/inno-hdmi.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/drivers/gpu/drm/bridge/inno-hdmi.c b/drivers/gpu/drm/bridge/inno-hdmi.c
index 5fa533a4eb342..297be02bbd58d 100644
--- a/drivers/gpu/drm/bridge/inno-hdmi.c
+++ b/drivers/gpu/drm/bridge/inno-hdmi.c
@@ -989,6 +989,10 @@ static int inno_hdmi_i2c_xfer(struct i2c_adapter *adap,
 
 	mutex_lock(&i2c->lock);
 
+	/* Reset segment_addr and ddc_addr at the start of every I2C transfer */
+	i2c->ddc_addr = 0;
+	i2c->segment_addr = 0;
+
 	/* Clear the EDID interrupt flag and unmute the interrupt */
 	hdmi_writeb(hdmi, HDMI_INTERRUPT_MASK1, m_INT_EDID_READY);
 	hdmi_writeb(hdmi, HDMI_INTERRUPT_STATUS1, m_INT_EDID_READY);
-- 
2.53.0

Re: [PATCH v1] drm/bridge: inno-hdmi: reset DDC segment pointer per I2C transfer

[Dominique Belhachemi]

On Thu, May 28, 2026 at 2:41 PM Dominique Belhachemi <domibel@debian.org> wrote:
>
> The segment_addr and ddc_addr fields need to be reset per I2C transfer
> to avoid stale values. The reset wasn't required on simple displays with
> only one segment, but modern displays with two or more segments end up
> with corrupt EDIDs.
>
> This was found on a StarFive VisionFive 2 SBC.
>
> Signed-off-by: Dominique Belhachemi <domibel@debian.org>
> ---
>  drivers/gpu/drm/bridge/inno-hdmi.c | 4 ++++
>  1 file changed, 4 insertions(+)
>
> diff --git a/drivers/gpu/drm/bridge/inno-hdmi.c b/drivers/gpu/drm/bridge/inno-hdmi.c
> index 5fa533a4eb342..297be02bbd58d 100644
> --- a/drivers/gpu/drm/bridge/inno-hdmi.c
> +++ b/drivers/gpu/drm/bridge/inno-hdmi.c
> @@ -989,6 +989,10 @@ static int inno_hdmi_i2c_xfer(struct i2c_adapter *adap,
>
>         mutex_lock(&i2c->lock);
>
> +       /* Reset segment_addr and ddc_addr at the start of every I2C transfer */
> +       i2c->ddc_addr = 0;
> +       i2c->segment_addr = 0;
> +
>         /* Clear the EDID interrupt flag and unmute the interrupt */
>         hdmi_writeb(hdmi, HDMI_INTERRUPT_MASK1, m_INT_EDID_READY);
>         hdmi_writeb(hdmi, HDMI_INTERRUPT_STATUS1, m_INT_EDID_READY);
> --

Hi, Is there anything else I can provide to get this merged?

Thanks
-Dominique

Re: [PATCH v1] drm/bridge: inno-hdmi: reset DDC segment pointer per I2C transfer

[Michal Wilczynski]

On 5/28/26 20:41, Dominique Belhachemi wrote:
> The segment_addr and ddc_addr fields need to be reset per I2C transfer
> to avoid stale values. The reset wasn't required on simple displays with
> only one segment, but modern displays with two or more segments end up
> with corrupt EDIDs.
> 
> This was found on a StarFive VisionFive 2 SBC.
> 
> Signed-off-by: Dominique Belhachemi <domibel@debian.org>
> ---
>  drivers/gpu/drm/bridge/inno-hdmi.c | 4 ++++
>  1 file changed, 4 insertions(+)
> 
> diff --git a/drivers/gpu/drm/bridge/inno-hdmi.c b/drivers/gpu/drm/bridge/inno-hdmi.c
> index 5fa533a4eb342..297be02bbd58d 100644
> --- a/drivers/gpu/drm/bridge/inno-hdmi.c
> +++ b/drivers/gpu/drm/bridge/inno-hdmi.c
> @@ -989,6 +989,10 @@ static int inno_hdmi_i2c_xfer(struct i2c_adapter *adap,
>  
>  	mutex_lock(&i2c->lock);
>  
> +	/* Reset segment_addr and ddc_addr at the start of every I2C transfer */
> +	i2c->ddc_addr = 0;
> +	i2c->segment_addr = 0;
> +
>  	/* Clear the EDID interrupt flag and unmute the interrupt */
>  	hdmi_writeb(hdmi, HDMI_INTERRUPT_MASK1, m_INT_EDID_READY);
>  	hdmi_writeb(hdmi, HDMI_INTERRUPT_STATUS1, m_INT_EDID_READY);

Reviewed-by: Michal Wilczynski <m.wilczynski@samsung.com>

Re: [PATCH v1] drm/bridge: inno-hdmi: reset DDC segment pointer per I2C transfer

[Michal Wilczynski]

On 9/27/26 10:51, Michal Wilczynski wrote:
> 
> 
> On 5/28/26 20:41, Dominique Belhachemi wrote:
>> The segment_addr and ddc_addr fields need to be reset per I2C transfer
>> to avoid stale values. The reset wasn't required on simple displays with
>> only one segment, but modern displays with two or more segments end up
>> with corrupt EDIDs.
>>
>> This was found on a StarFive VisionFive 2 SBC.
>>
>> Signed-off-by: Dominique Belhachemi <domibel@debian.org>
>> ---
>>  drivers/gpu/drm/bridge/inno-hdmi.c | 4 ++++
>>  1 file changed, 4 insertions(+)
>>
>> diff --git a/drivers/gpu/drm/bridge/inno-hdmi.c b/drivers/gpu/drm/bridge/inno-hdmi.c
>> index 5fa533a4eb342..297be02bbd58d 100644
>> --- a/drivers/gpu/drm/bridge/inno-hdmi.c
>> +++ b/drivers/gpu/drm/bridge/inno-hdmi.c
>> @@ -989,6 +989,10 @@ static int inno_hdmi_i2c_xfer(struct i2c_adapter *adap,
>>  
>>  	mutex_lock(&i2c->lock);
>>  
>> +	/* Reset segment_addr and ddc_addr at the start of every I2C transfer */
>> +	i2c->ddc_addr = 0;
>> +	i2c->segment_addr = 0;
>> +
>>  	/* Clear the EDID interrupt flag and unmute the interrupt */
>>  	hdmi_writeb(hdmi, HDMI_INTERRUPT_MASK1, m_INT_EDID_READY);
>>  	hdmi_writeb(hdmi, HDMI_INTERRUPT_STATUS1, m_INT_EDID_READY);
> 
> Reviewed-by: Michal Wilczynski <m.wilczynski@samsung.com>

I was also told that there were some changes in the MAINTAINERS file
regarding this patch and some core people may be missing e.g Luca and
Maxime.

It would be best to resend with updated list of maintainers.

Best regards,
-- 
Michal Wilczynski <m.wilczynski@samsung.com>

Re: [PATCH v1] drm/bridge: inno-hdmi: reset DDC segment pointer per I2C transfer

[Dominique Belhachemi]

On Mon, Sep 28, 2026 at 7:15 AM Michal Wilczynski
<m.wilczynski@samsung.com> wrote:
>
>
>
> On 9/27/26 10:51, Michal Wilczynski wrote:
> >
> >
> > On 5/28/26 20:41, Dominique Belhachemi wrote:
> >> The segment_addr and ddc_addr fields need to be reset per I2C transfer
> >> to avoid stale values. The reset wasn't required on simple displays with
> >> only one segment, but modern displays with two or more segments end up
> >> with corrupt EDIDs.
> >>
> >> This was found on a StarFive VisionFive 2 SBC.
> >>
> >> Signed-off-by: Dominique Belhachemi <domibel@debian.org>
> >> ---
> >>  drivers/gpu/drm/bridge/inno-hdmi.c | 4 ++++
> >>  1 file changed, 4 insertions(+)
> >>
> >> diff --git a/drivers/gpu/drm/bridge/inno-hdmi.c b/drivers/gpu/drm/bridge/inno-hdmi.c
> >> index 5fa533a4eb342..297be02bbd58d 100644
> >> --- a/drivers/gpu/drm/bridge/inno-hdmi.c
> >> +++ b/drivers/gpu/drm/bridge/inno-hdmi.c
> >> @@ -989,6 +989,10 @@ static int inno_hdmi_i2c_xfer(struct i2c_adapter *adap,
> >>
> >>      mutex_lock(&i2c->lock);
> >>
> >> +    /* Reset segment_addr and ddc_addr at the start of every I2C transfer */
> >> +    i2c->ddc_addr = 0;
> >> +    i2c->segment_addr = 0;
> >> +
> >>      /* Clear the EDID interrupt flag and unmute the interrupt */
> >>      hdmi_writeb(hdmi, HDMI_INTERRUPT_MASK1, m_INT_EDID_READY);
> >>      hdmi_writeb(hdmi, HDMI_INTERRUPT_STATUS1, m_INT_EDID_READY);
> >
> > Reviewed-by: Michal Wilczynski <m.wilczynski@samsung.com>
>
> I was also told that there were some changes in the MAINTAINERS file
> regarding this patch and some core people may be missing e.g Luca and
> Maxime.
>
> It would be best to resend with updated list of maintainers.
>
Thanks Michal, I am adding them to Cc.

Best
-Dominique