Re: [PATCH V4] drm: edid: add support for E-DDC

[Jean Delvare <jdelvare@suse.de>]

Hi all,

Sorry for breaking message threading but I was not included in
iterations 3 and 4 of this patch.

Random comments about v4:

> --- a/drivers/gpu/drm/drm_edid.c
> +++ b/drivers/gpu/drm/drm_edid.c
> @@ -254,6 +254,8 @@ drm_do_probe_ddc_edid(struct i2c_adapter *adapter, unsigned 
> char *buf,
>                       int block, int len)
>  {
>         unsigned char start = block * EDID_LENGTH;
> +       unsigned char segment = block >> 1;
> +       unsigned char xfers = segment ? 3 : 2;
>         int ret, retries = 5;
>  
>         /* The core i2c driver will automatically retry the transfer if the
> @@ -264,7 +266,12 @@ drm_do_probe_ddc_edid(struct i2c_adapter *adapter, 
> unsigned char *buf,
>          */
>         do {
>                 struct i2c_msg msgs[] = {
> -                       {
> +                       { /*set segment pointer */
> +                               .addr   = DDC_SEGMENT_ADDR,
> +                               .flags  = segment ? 0 : I2C_M_IGNORE_NAK,

I don't get the idea. If segment == 0, this message is never sent, so the
value of field flags doesn't matter. So flags will always be 0 when this
message is sent, so it can be hard-coded.

But from previous discussions my understanding was an agreement on always
using I2C_M_IGNORE_NAK for improved compatibility. So I2C_M_IGNORE_NAK
should be hard-coded, not 0?

> +                               .len    = 1,
> +                               .buf    = &segment,
> +                       }, {
>                                 .addr   = DDC_ADDR,
>                                 .flags  = 0,
>                                 .len    = 1,
> @@ -276,15 +283,22 @@ drm_do_probe_ddc_edid(struct i2c_adapter *adapter, 
> unsigned char *buf,
>                                 .buf    = buf,
>                         }
>                 };
> -               ret = i2c_transfer(adapter, msgs, 2);
> +       /* Avoid sending the segment addr to not upset non-compliant ddc
> +        * monitors.
> +        */

s/segment addr/segment/, plus it's abot E-DCC compliance as I understand it,
not DDC.

> +               if (!segment)
> +                       ret = i2c_transfer(adapter, &msgs[1], xfers);
> +               else
> +                       ret = i2c_transfer(adapter, msgs, xfers);
> +

This can be written:

		ret = i2c_transfer(adapter, &msgs[3 - xfers], xfers);

Which is more compact and, I suspect, faster.

>                 if (ret == -ENXIO) {
>                         DRM_DEBUG_KMS("drm: skipping non-existent adapter %s\n",
>                                         adapter->name);
>                         break;
>                 }
> -       } while (ret != 2 && --retries);
> +       } while (ret != xfers && --retries);
>  
> -       return ret == 2 ? 0 : -1;
> +       return ret == xfers ? 0 : -1;
>  }
>  
>  static bool drm_edid_is_zero(u8 *in_edid, int length)

Other than this, your code looks reasonable, not so different from what
I submitted 8 months ago actually. But ISTU you can test the code with
real hardware while I couldn't.

With the changes above applied, you can add:

Reviewed-by: Jean Delvare <jdelvare@suse.de>

-- 
Jean Delvare
Suse L3