[v3,1/2] dt-bindings: panel: Add Samsung S6E3FA2 panel

The Samsung S6E3FA2 AMOLED cmd LCD panel is used on the Samsung Galaxy
S5 (klte).

Signed-off-by: Iskren Chernev <iskren.chernev@gmail.com>
---
Add a simple generated panel driver that supports on/off and the corresponding
binding documentation.

Changes in v3:
- fix dt_binding_check issue with missing include
- fix panel type (cmd) in kconfig description

Changes in v2:
- move bindings to separate file, add 2 regulators
- add standalone panel driver

v1: https://lkml.org/lkml/2020/12/30/293
v2: https://lkml.org/lkml/2021/2/1/313

 .../display/panel/samsung,s6e3fa2.yaml        | 64 +++++++++++++++++++
 1 file changed, 64 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/display/panel/samsung,s6e3fa2.yaml


base-commit: fd821bf0ed9a7db09d2e007df697f4d9ecfda99a
--
2.30.0

Hi Iskren,

thanks for your patch!

On Mon, Feb 1, 2021 at 5:56 PM Iskren Chernev <iskren.chernev@gmail.com> wrote:

> +required:
> +  - compatible
> +  - reset-gpios
> +  - iovdd-supply
> +  - vddr-supply
> +  - port

I do not think port should be required because the DSI
framework allows panels to be put as children right inside
the DSI controller node without any port linking the two.

Other than that it looks good!

Yours,
Linus Walleij

Hi Iskren,

thanks for your patch!

On Mon, Feb 1, 2021 at 5:56 PM Iskren Chernev <iskren.chernev@gmail.com> wrote:

> The Samsung Galaxy S5 uses the samsung s6e3fa2 AMOLED cmd LCD panel.
>
> This driver was generated with [1], with the addition of
> mipi_dsi_dcs_set_display_on at the end of the on method.
>
> [1] https://github.com/msm8916-mainline/linux-mdss-dsi-panel-driver-generator
>
> Signed-off-by: Iskren Chernev <iskren.chernev@gmail.com>

(...)

> +config DRM_PANEL_SAMSUNG_S6E3FA2
> +       tristate "Samsung S6E3FA2 DSI cmd mode panel"

Spell it out "command mode" no need to abbreviate.

> +struct samsung_s6e3fa2 {
> +       struct drm_panel panel;
> +       struct mipi_dsi_device *dsi;
> +       struct regulator_bulk_data supplies[2];
> +       struct gpio_desc *reset_gpio;
> +       bool prepared;

Why do you need to keep track of prepared?
The framework already does that and that is why it keeps
different callbacks for prepare and enable.

> +#define dsi_generic_write_seq(dsi, seq...) do {                                \
> +               static const u8 d[] = { seq };                          \
> +               int ret;                                                \
> +               ret = mipi_dsi_generic_write(dsi, d, ARRAY_SIZE(d));    \
> +               if (ret < 0)                                            \
> +                       return ret;                                     \
> +       } while (0)
> +
> +#define dsi_dcs_write_seq(dsi, seq...) do {                            \
> +               static const u8 d[] = { seq };                          \
> +               int ret;                                                \
> +               ret = mipi_dsi_dcs_write_buffer(dsi, d, ARRAY_SIZE(d)); \
> +               if (ret < 0)                                            \
> +                       return ret;                                     \
> +       } while (0)

Maybe this is something we should add to
include/drm/drm_mipi_dsi.h?
It looks very generic and helpful.

> +static void samsung_s6e3fa2_reset(struct samsung_s6e3fa2 *ctx)
> +{
> +       gpiod_set_value_cansleep(ctx->reset_gpio, 0);
> +       usleep_range(5000, 6000);

There is really no need to do this, the next statement asserts
reset. It looks like a copy/paste thing from a less informed
framework.

> +       gpiod_set_value_cansleep(ctx->reset_gpio, 1);
> +       usleep_range(5000, 6000);
> +       gpiod_set_value_cansleep(ctx->reset_gpio, 0);
> +       usleep_range(7000, 8000);

These two look good.

> +static int samsung_s6e3fa2_on(struct samsung_s6e3fa2 *ctx)
> +{
> +       struct mipi_dsi_device *dsi = ctx->dsi;
> +       struct device *dev = &dsi->dev;
> +       int ret;
> +
> +       dsi->mode_flags |= MIPI_DSI_MODE_LPM;
> +
> +       dsi_generic_write_seq(dsi, 0xf0, 0x5a, 0x5a);
> +       dsi_generic_write_seq(dsi, 0xfc, 0x5a, 0x5a);

These are very similar to other Samsung panels so you can definately
add this to your driver:

#define S6E3FA2_LEVEL_2_KEY         0xf0

> +       dsi_generic_write_seq(dsi, 0xf0, 0xa5, 0xa5);
> +       dsi_generic_write_seq(dsi, 0xfc, 0xa5, 0xa5);

Add a comment that this locks level 2 control again.

> +static int samsung_s6e3fa2_prepare(struct drm_panel *panel)
> +{
> +       struct samsung_s6e3fa2 *ctx = to_samsung_s6e3fa2(panel);
> +       struct device *dev = &ctx->dsi->dev;
> +       int ret;
> +
> +       if (ctx->prepared)
> +               return 0;

Looks unnecessary. The framework should keep track of this.

> +       dsi->lanes = 4;
> +       dsi->format = MIPI_DSI_FMT_RGB888;
> +       dsi->mode_flags = MIPI_DSI_MODE_VIDEO_BURST |
> +                         MIPI_DSI_CLOCK_NON_CONTINUOUS;

You are not setting dsi->hs_rate and dsi->lp_rate.
The panel definitely has these max timings so when you
don't set them the DSI host will just go with some defaults
(some of them don't even allow you to configure this).

For reusability it would be great if you know roughly what these
clocks are, sometimes board files in vendor trees give a hint.

Worst case drop a comment that HS and LP rates are unknown.

I also wonder if the panel can report an MTP ID?

If you add some code like this and call from prepare(), what
happens?

#define READ_ID1                 0xda /* Read panel ID 1 */
#define READ_ID2                 0xdb /* Read panel ID 2 */
#define READ_ID3                 0xdc /* Read panel ID 3 */

static int s6e3fa2_read_id(struct samsung_s6e3fa2 *ctx)
{
        struct mipi_dsi_device *dsi = ctx->dsi;
        struct device *dev = &dsi->dev;
        u8 id1, id2, id3;
        int ret;

        ret = mipi_dsi_dcs_read(dsi, READ_ID1, &id1, 1);
        if (ret < 0) {
                dev_err(dev, "could not read MTP ID1\n");
                return ret;
        }
        ret = mipi_dsi_dcs_read(dsi, READ_ID2, &id2, 1);
        if (ret < 0) {
                dev_err(dev, "could not read MTP ID2\n");
                return ret;
        }
        ret = mipi_dsi_dcs_read(dsi, READ_ID3, &id3, 1);
        if (ret < 0) {
                dev_err(dev, "could not read MTP ID3\n");
                return ret;
        }

        dev_info(dev, "MTP ID manufacturer: %02x version: %02x driver:
%02x\n", id1, id2, id3);

        return 0;
}

(Dry coded, but you get the idea.)

Yours,
Linus Walleij

On 7/26/21 10:34 AM, Linus Walleij wrote:
> Hi Iskren,

> 

> thanks for your patch!


Thanks a lot for reviewing this. Alexey wrote a new version of the patch [1]
that was recently reviewed by Sam, and it also includes support for brightness
and another panel found on the S5. My patch is pretty much untouched out of the
panel driver generator, so we can definitely improve the generated code based
on your comments.

[1] https://lists.freedesktop.org/archives/dri-devel/2021-July/316804.html

I guess Alexey will submit a new version with the two panels split, and he
might incorporate your suggestions (as his code comes from the generator as
well).

Regards,
Iskren