[PATCH] drm: bridge: adv7511: fix reading edid segments

[Emil Abildgaard Svendsen]

On Thu, Oct 26, 2023 at 09:11:53PM +0200, Frieder Schrempf wrote:
> [You don't often get email from frieder.schrempf at kontron.de. Learn why this is important at https://aka.ms/LearnAboutSenderIdentification ]
> 
> On 26.10.23 13:30, Emil Abildgaard Svendsen wrote:
> > [Sie erhalten nicht häufig E-Mails von emas at bang-olufsen.dk. Weitere Informationen, warum dies wichtig ist, finden Sie unter https://aka.ms/LearnAboutSenderIdentification ]
> >
> > Currently reading EDID only works because usually only two EDID blocks
> > of 128 bytes is used. Where an EDID segment holds 256 bytes or two EDID
> > blocks. And the first EDID segment read works fine but E-EDID specifies
> > up to 128 segments.
> >
> > The logic is broken so change EDID segment index to multiple of 256
> > bytes and not 128 (block size).
> >
> > Also change check of DDC status. Instead of silently not reading EDID
> > when in "IDLE" state [1]. Check if the DDC controller is in reset
> > instead (no HPD).
> >
> > [1]
> > ADV7511 Programming Guide: Table 11: DDCController Status:
> >
> >   0xC8 [3:0]  DDC Controller State
> >
> >   0000        In Reset (No Hot Plug Detected)
> >   0001        Reading EDID
> >   0010        IDLE (Waiting for HDCP Requested)
> >   0011        Initializing HDCP
> >   0100        HDCP Enabled
> >   0101        Initializing HDCP Repeater
> >
> > Signed-off-by: Emil Svendsen <emas at bang-olufsen.dk>
> 
> This patch somehow seems to break the reported display modes with my
> setup (i.MX8MM DSI -> ADV7535 -> FullHD screen) when I test on current
> linux-next.
> 
> Without this patch I get 30 valid modes reported by modetest.
> 
> With this patch applied I only get 5 valid modes. The screen still comes
> up with the 1080p default mode though, that is now not even in the list
> anymore.

Okay, I believe it's because I return when DDC controller is in reset. I
believe you hit this because HPD override is enabled for ADV7535 which
in my experiments isn't needed. But I will see if I can address that in
another commit.

For this commit I will simply remove the return. Reading EDID clearly
still works when DDC is in reset. Maybe because of the 200 ms wait?