On 2023-05-26 09:32:45, Neil Armstrong wrote:
<snip>
> >>>>> +static int visionox_r66451_bl_update_status(struct backlight_device *bl)
> >>>>> +{
> >>>>> + struct mipi_dsi_device *dsi = bl_get_data(bl);
> >>>>> + u16 brightness = backlight_get_brightness(bl);
> >>>>> +
> >>>>> + return mipi_dsi_dcs_set_display_brightness(dsi, cpu_to_le16(brightness));
> >>>>
> >>>> mipi_dsi_dcs_set_display_brightness() already converts the brightness,
> >>>> so you don't need cpu_to_le16 here.
> >>>
> >>> Tread carefully here: we've had the same issue and conversation on our
> >>> Sony panels where this extra inversion is required.
> >>> set_display_brightness() sends the bytes as little-endian to the panel
> >>> (and it even assumes little-endian in get_display_brightness()) but the
> >>> spec for 16-bit brightness values states that they have to be sent in
> >>> big-endian. This is why c9d27c6be518b ("drm/mipi-dsi: Fix byte order of
> >>> 16-bit DCS set/get brightness") added
> >>> mipi_dsi_dcs_set_display_brightness_large().
> >>>
> >>> Jessica, if you need to have the endian swap here (should be very easy
> >>> to test with a real panel, but it should be given the max_brightness
> >>> value being over 8 bits) please switch to the _large() variant.
> >>
> >> Got it, thanks for the heads up!
> >
> > Hi Marijn,
> >
> > Just wanted to update this thread -- I've checked the backlight brightness values in the sysfs and it matches the value being given in the panel driver (255), so I think it should be fine to use *_set_display_brightness() without the _large() variant.
>
> Sure, I was also misleaded by you using cpu_to_le16() but as Dmitry said it's already
> done in mipi_dsi_dcs_set_display_brightness() and a no-op on LE arm64 platforms anyway.
Yuck, right, it's cpu_to_le16 here and not cpu_to_be16. @Jessica, can
you please remove this misleading conversion?
mipi_dsi_dcs_set_display_brightness() takes a native u16, not a specific
__le16.
- Marijn
