[PATCH v4 06/13] drm/bridge: lvds-codec: Add "lvds-decoder" support

[Laurent Pinchart]

Hello,

On Tue, Dec 17, 2019 at 01:38:51PM +0100, Geert Uytterhoeven wrote:
> On Tue, Dec 17, 2019 at 1:31 PM Fabrizio Castro wrote:
> > On 17 December 2019 12:21, Geert Uytterhoeven wrote:
> >> On Tue, Dec 17, 2019 at 12:03 PM Fabrizio Castro wrote:
> >>> On 13 December 2019 17:11, Laurent Pinchart wrote:
> >>>> On Wed, Nov 13, 2019 at 03:51:25PM +0000, Fabrizio Castro wrote:
> >>>>> Add support for transparent LVDS decoders by adding a new
> >>>>> compatible string ("lvds-decoder") to the driver.
> >>>>> This patch also adds member connector_type to struct lvds_codec,
> >>>>> and that's because LVDS decoders have a different connector type
> >>>>> from LVDS encoders. We fill this new member up with the data
> >>>>> matching the compatible string.
> >>>>>
> >>>>> Signed-off-by: Fabrizio Castro <fabrizio.castro at bp.renesas.com>
> >>>>>
> >>>>> --- a/drivers/gpu/drm/bridge/lvds-codec.c
> >>>>> +++ b/drivers/gpu/drm/bridge/lvds-codec.c
> >>>>>
> >>>>> @@ -65,6 +67,7 @@ static int lvds_codec_probe(struct platform_device *pdev)
> >>>>>     if (!lvds_codec)
> >>>>>             return -ENOMEM;
> >>>>>
> >>>>> +   lvds_codec->connector_type = (u32)of_device_get_match_data(&pdev->dev);
> >>>>
> >>>> I'm now getting a compilation failure here:
> >>>>
> >>>> drivers/gpu/drm/bridge/lvds-codec.c: In function ‘lvds_codec_probe’:
> >>>> drivers/gpu/drm/bridge/lvds-codec.c:68:31: error: cast from pointer to integer of different size [-Werror=pointer-to-int-cast]
> >>>>   lvds_codec->connector_type = (u32)of_device_get_match_data(&pdev->dev);
> >>>>                                ^
> >>>>
> >>>> The fix should be simple:
> >>>>
> >>>>       lvds_codec->connector_type = (uintptr_t)of_device_get_match_data(dev);
> >>>>
> >>>> I'm bothered by the fact that I've compiled this before without any
> >>>> issue, so this really puzzles me. Do you get the same warning ?
> >>>
> >>> The warning appears when compiling for arm64, understandably so.
> >>> We must have compiled this for arm only the first time around.
> >>>
> >>> I think the right way to solve this is to either cast to (u32)(uintptr_t) or (u32)(unsigned long).
> >>
> >> Just casting to uintptr_t should be sufficient.
> >
> > It should be sufficient for the compiler, but I have seen examples where people
> > preferred to be explicit, like in:
> > drivers/mailbox/mtk-cmdq-mailbox.c
> > drivers/leds/leds-pm8058.c
> >
> > Since the kernel is increasing its tightness with respect to warnings, I personally prefer
> > (u32)(uintptr_t), even though not strictly necessary, but I am fine with (uintptr_t) if you
> > don't like (u32)(uintptr_t).
> 
> It depends. I try to have as few casts as possible ("casts are evil").
> 
> While adding the extra (u32) cast makes it clearer that the intended
> result is a u32 (for now), it will cause silent truncation on 64-bit if
> connector_type is ever enlarged to unsigned long, and larger values are
> used.
> 
> In this particular case this is unlikely, though, as unsigned long would
> still be 32-bit on 32-bit platforms, so the larger values cannot be
> used.

I also try to add as few casts as possible, so (uintptr_t) would be my
preference.

Fabrizio, could you submit a new version of this patch with the problem
fixed (and with the casts you decide to use, but using uintptr_t instead
of unsigned long in any case) ?

-- 
Regards,

Laurent Pinchart