[PATCH 0/2] drm/panel: add the Novatek NT51021 panel in the Amazon Fire HD 10

[PATCH 0/2] drm/panel: add the Novatek NT51021 panel in the Amazon Fire HD 10

[Ryan Brue via B4 Relay]

Add the 1200x1920 DSI panel fitted to the Amazon Fire HD 10 (2017), an
MT8173 tablet. The controller is a Novatek NT51021; the module carries
BOE glass.

The controller has no readable panel identity. MediaTek's LCM framework
identifies panels through a .compare_id callback and the vendor's driver
for this one leaves it unset, and the glass part number is not recorded
anywhere in the vendor tree -- it names the module only "BOE". I had the
compatible name the module after the board it ships in, the way
xiaomi,elish-boe-nt36523 and lenovo,j606f-boe-nt36523w do for this
situation.

The vendor's bootloader and kernel both carry a second register sequence
for Innolux glass, selected from two board ID straps, so a second module
exists as far as the vendor firmware is concerned. I didn't have an
Innolux unit to test with, and I don't know its part number, so it is not
described here. Naming the glass in the compatible leaves room for it as
a second compatible should someone have the hardware and the need.

This controller accepts commands either over I2C or in-band over DSI, and
on the shipping boards it does not default to DSI. Writing 0xa5 to
register 0x8f opens a window in which DSI commands are accepted and
writing 0x00 closes it, so every command the driver sends, including
sleep in/out and display on/off, has to sit inside such a window. That is
why those are not simply issued from enable()/disable() as most panel
drivers do.

The vendor's timings are used verbatim.

The board DTS that uses this is not in this series -- it's a whole new
board file that I plan on upstreaming over time after getting its
components upstreamed.

Tested on the Amazon Fire HD 10 (2017) driven by mediatek-drm: 1200x1920
at 62.5 Hz across a cold boot, three display off and on cycles and a
suspend/resume, with the MIPI TX PLL read back at the intended rate.

Signed-off-by: Ryan Brue <ryanbrue.dev@gmail.com>
---
Ryan Brue (2):
      dt-bindings: display: panel: add Novatek NT51021
      drm/panel: add Novatek NT51021 1200x1920 DSI panel driver

 .../bindings/display/panel/novatek,nt51021.yaml    |  65 ++++
 MAINTAINERS                                        |   7 +
 drivers/gpu/drm/panel/Kconfig                      |  11 +
 drivers/gpu/drm/panel/Makefile                     |   1 +
 drivers/gpu/drm/panel/panel-novatek-nt51021.c      | 329 +++++++++++++++++++++
 5 files changed, 413 insertions(+)
---
base-commit: cee9395acd8043be0644b25c34bfa86623f2b935
change-id: 20260905-rbrue-drm-panel-add-novatek-nt51021-b286a67e4de3

Best regards,
--  
Ryan Brue <ryanbrue.dev@gmail.com>

[PATCH 1/2] dt-bindings: display: panel: add Novatek NT51021

[Ryan Brue via B4 Relay]

From: Ryan Brue <ryanbrue.dev@gmail.com>

The Novatek NT51021 is a DSI panel controller driving 1200x1920 WUXGA
panels over four lanes in video mode. It is fitted in the Amazon Fire HD 10
(2017), on a module carrying BOE glass.

The controller has no readable panel identity. MediaTek's LCM framework
identifies panels through a .compare_id callback and the vendor's driver
for this one leaves it unset, and the glass part number is not recorded
anywhere in the vendor tree -- it names the module only "BOE". I had the
compatible name the module after the board it ships in, the way
xiaomi,elish-boe-nt36523 and lenovo,j606f-boe-nt36523w do for this
situation.

The vendor's bootloader and kernel both carry a second register sequence
for Innolux glass, selected from two board ID straps, so a second module
exists as far as the vendor firmware is concerned. I didn't have an
Innolux unit to test with, and I don't know its part number, so it is not
described here. Naming the glass in the compatible leaves room for it as
a second compatible should someone have the hardware and the need.

The rest is panel-common: a power supply, a reset GPIO, an optional
backlight and the DSI port.

Assisted-by: LLM
Signed-off-by: Ryan Brue <ryanbrue.dev@gmail.com>
---
 .../bindings/display/panel/novatek,nt51021.yaml    | 65 ++++++++++++++++++++++
 1 file changed, 65 insertions(+)

diff --git a/Documentation/devicetree/bindings/display/panel/novatek,nt51021.yaml b/Documentation/devicetree/bindings/display/panel/novatek,nt51021.yaml
new file mode 100644
index 000000000000..ee3adc88ec9b
--- /dev/null
+++ b/Documentation/devicetree/bindings/display/panel/novatek,nt51021.yaml
@@ -0,0 +1,65 @@
+# SPDX-License-Identifier: GPL-2.0-only OR BSD-2-Clause
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/display/panel/novatek,nt51021.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: Novatek NT51021 based 1200x1920 DSI display panels
+
+maintainers:
+  - Ryan Brue <ryanbrue.dev@gmail.com>
+
+description:
+  The Novatek NT51021 is a DSI panel controller driving 1200x1920 WUXGA
+  panels over four lanes in video mode. The controller carries no readable
+  panel identity, so each panel module it is fitted to needs its own
+  compatible naming the module.
+
+allOf:
+  - $ref: panel-common.yaml#
+
+properties:
+  compatible:
+    items:
+      - const: amazon,suez-boe-nt51021
+      - const: novatek,nt51021
+
+  reg:
+    maxItems: 1
+
+  power-supply: true
+  reset-gpios: true
+  backlight: true
+  port: true
+
+required:
+  - compatible
+  - reg
+  - power-supply
+  - reset-gpios
+  - port
+
+unevaluatedProperties: false
+
+examples:
+  - |
+    #include <dt-bindings/gpio/gpio.h>
+
+    dsi {
+        #address-cells = <1>;
+        #size-cells = <0>;
+
+        panel@0 {
+            compatible = "amazon,suez-boe-nt51021", "novatek,nt51021";
+            reg = <0>;
+            backlight = <&backlight_lcd>;
+            power-supply = <&vgp6>;
+            reset-gpios = <&pio 105 GPIO_ACTIVE_LOW>;
+
+            port {
+                endpoint {
+                    remote-endpoint = <&dsi_out>;
+                };
+            };
+        };
+    };

-- 
2.55.0

[PATCH 2/2] drm/panel: add Novatek NT51021 1200x1920 DSI panel driver

[Ryan Brue via B4 Relay]

From: Ryan Brue <ryanbrue.dev@gmail.com>

Add a driver for the BOE module built around the Novatek NT51021, as
fitted to the Amazon Fire HD 10 (2017). The panel is 1200x1920, four
lanes, RGB888, sync-event video mode with a continuous HS clock,
136x221 mm. The timings are those of the vendor's LCM driver
(drivers/misc/mediatek/lcm/nt51021_wuxga_dsi_vdo/ in the Fire OS 3.18
tree): hsync 1, hbp 32, hfp 110, vsync 1, vbp 14, vfp 11, so htotal
1343 and vtotal 1946. They are used verbatim.

The pixel clock is the vendor's PLL_CLOCK = 490, which after DDR is 980
Mbps per lane and over four lanes at 24bpp is 163.33 MHz, so 62.5 Hz.
That is also what the bootloader leaves running: read back before the
kernel touches it, this board's MIPI TX PLL decodes to exactly 980 Mbps
per lane, so the panel is already at this rate when the kernel takes
over.

This controller accepts commands either over I2C or in-band over DSI, and
on the shipping boards it does not default to DSI. Writing 0xa5 to
register 0x8f opens a window in which DSI commands are accepted and
writing 0x00 closes it, so every command the driver sends, including
sleep in/out and display on/off, has to sit inside such a window. That is
why those are not simply issued from enable()/disable() as most panel
drivers do.

The vendor's bootloader and kernel both carry a second register sequence
for Innolux glass, chosen from two board ID straps. No Innolux unit was
available, so that sequence is left out rather than carried untested;
adding it later is one compatible, one table and one desc.

Tested on the Amazon Fire HD 10 (2017) driven by mediatek-drm: 1200x1920
at 62.5 Hz across a cold boot, three display off and on cycles and a
suspend/resume, with the MIPI TX PLL reading back at the intended rate.

Assisted-by: LLM
Signed-off-by: Ryan Brue <ryanbrue.dev@gmail.com>
---
 MAINTAINERS                                   |   7 +
 drivers/gpu/drm/panel/Kconfig                 |  11 +
 drivers/gpu/drm/panel/Makefile                |   1 +
 drivers/gpu/drm/panel/panel-novatek-nt51021.c | 329 ++++++++++++++++++++++++++
 4 files changed, 348 insertions(+)

diff --git a/MAINTAINERS b/MAINTAINERS
index 3a19da74d00c..1cef4dd31d03 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -8353,6 +8353,13 @@ S:	Maintained
 F:	Documentation/devicetree/bindings/display/panel/novatek,nt37801.yaml
 F:	drivers/gpu/drm/panel/panel-novatek-nt37801.c
 
+DRM DRIVER FOR NOVATEK NT51021 PANELS
+M:	Ryan Brue <ryanbrue.dev@gmail.com>
+S:	Maintained
+T:	git https://gitlab.freedesktop.org/drm/misc/kernel.git
+F:	Documentation/devicetree/bindings/display/panel/novatek,nt51021.yaml
+F:	drivers/gpu/drm/panel/panel-novatek-nt51021.c
+
 DRM DRIVER FOR NVIDIA GEFORCE/QUADRO GPUS
 M:	Lyude Paul <lyude@redhat.com>
 M:	Danilo Krummrich <dakr@kernel.org>
diff --git a/drivers/gpu/drm/panel/Kconfig b/drivers/gpu/drm/panel/Kconfig
index cbdf7b8f7f7a..50a5dc473138 100644
--- a/drivers/gpu/drm/panel/Kconfig
+++ b/drivers/gpu/drm/panel/Kconfig
@@ -727,6 +727,17 @@ config DRM_PANEL_NOVATEK_NT39016
 	  Say Y here if you want to enable support for the panels built
 	  around the Novatek NT39016 display controller.
 
+config DRM_PANEL_NOVATEK_NT51021
+	tristate "Novatek NT51021 1200x1920 DSI video mode panel"
+	depends on OF
+	depends on DRM_MIPI_DSI
+	depends on BACKLIGHT_CLASS_DEVICE
+	help
+	  Say Y here if you want to enable support for the Novatek NT51021
+	  1200x1920 DSI video mode panel, as found in the Amazon Fire HD 10
+	  (2017). The controller carries no readable panel identity, so the
+	  module fitted to a board is named by its compatible string.
+
 config DRM_PANEL_OLIMEX_LCD_OLINUXINO
 	tristate "Olimex LCD-OLinuXino panel"
 	depends on OF
diff --git a/drivers/gpu/drm/panel/Makefile b/drivers/gpu/drm/panel/Makefile
index 3b523cf37833..3fe7c2e579d3 100644
--- a/drivers/gpu/drm/panel/Makefile
+++ b/drivers/gpu/drm/panel/Makefile
@@ -70,6 +70,7 @@ obj-$(CONFIG_DRM_PANEL_NOVATEK_NT36672E) += panel-novatek-nt36672e.o
 obj-$(CONFIG_DRM_PANEL_NOVATEK_NT37700F) += panel-novatek-nt37700f.o
 obj-$(CONFIG_DRM_PANEL_NOVATEK_NT37801) += panel-novatek-nt37801.o
 obj-$(CONFIG_DRM_PANEL_NOVATEK_NT39016) += panel-novatek-nt39016.o
+obj-$(CONFIG_DRM_PANEL_NOVATEK_NT51021) += panel-novatek-nt51021.o
 obj-$(CONFIG_DRM_PANEL_MANTIX_MLAF057WE51) += panel-mantix-mlaf057we51.o
 obj-$(CONFIG_DRM_PANEL_OLIMEX_LCD_OLINUXINO) += panel-olimex-lcd-olinuxino.o
 obj-$(CONFIG_DRM_PANEL_ORISETECH_OTA5601A) += panel-orisetech-ota5601a.o
diff --git a/drivers/gpu/drm/panel/panel-novatek-nt51021.c b/drivers/gpu/drm/panel/panel-novatek-nt51021.c
new file mode 100644
index 000000000000..d92f3e18f716
--- /dev/null
+++ b/drivers/gpu/drm/panel/panel-novatek-nt51021.c
@@ -0,0 +1,329 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Novatek NT51021 1200x1920 DSI video-mode panel
+ *
+ * Derived from the vendor LCM driver in the Amazon 3.18 tree,
+ * drivers/misc/mediatek/lcm/nt51021_wuxga_dsi_vdo/.
+ *
+ * Note on the command interface: this panel can take its control commands
+ * either over I2C or in-band over MIPI, and on the shipping ("dvt") boards it
+ * does NOT default to MIPI. Writing 0xa5 to register 0x8f forces commands to
+ * be accepted over DSI, and writing 0x00 hands the interface back. Every
+ * command this driver sends therefore has to sit inside such a window --
+ * including sleep in/out and display on/off, which is why they are not simply
+ * issued from enable()/disable() the way most panel drivers do it.
+ */
+
+#include <linux/delay.h>
+#include <linux/gpio/consumer.h>
+#include <linux/module.h>
+#include <linux/of.h>
+#include <linux/regulator/consumer.h>
+
+#include <drm/drm_mipi_dsi.h>
+#include <drm/drm_modes.h>
+#include <drm/drm_panel.h>
+#include <drm/drm_probe_helper.h>
+
+/* Register 0x8f selects where the panel listens for commands. */
+#define NT51021_CMD_IF			0x8f
+#define NT51021_CMD_IF_FORCE_MIPI	0xa5
+#define NT51021_CMD_IF_RELEASE		0x00
+
+struct nt51021_cmd {
+	u8 reg;
+	u8 val;
+	u16 delay_ms;
+};
+
+/*
+ * Taken verbatim from the vendor's init_lcm_registers() BOE branch, minus the
+ * command-window writes, which this driver manages itself.
+ */
+static const struct nt51021_cmd nt51021_init_suez_boe[] = {
+	{ 0x01, 0x00, 20 },	/* software reset */
+	{ 0x83, 0x00, 0 },	/* page select */
+	{ 0x84, 0x00, 0 },
+	{ 0x8c, 0x80, 0 },	/* GOP setting */
+	{ 0xcd, 0x6c, 0 },	/* 3 dummy */
+	{ 0xc0, 0x8b, 0 },	/* GCH */
+	{ 0xc8, 0xf0, 0 },	/* GCH */
+	{ 0x97, 0x00, 0 },	/* resistor setting, 100 ohm */
+	{ 0x8b, 0x10, 0 },
+	{ 0xa9, 0x20, 0 },	/* enable TP_SYNC */
+	{ 0x83, 0xaa, 0 },	/* page select */
+	{ 0x84, 0x11, 0 },
+	{ 0xa9, 0x4b, 0 },	/* MIPI Rx drive strength, 85% */
+	{ 0x85, 0x04, 0 },	/* test mode 1 */
+	{ 0x86, 0x08, 0 },	/* test mode 2 */
+	{ 0x9c, 0x10, 0 },	/* test mode 3 */
+};
+
+struct nt51021_desc {
+	const struct nt51021_cmd *cmds;
+	unsigned int num_cmds;
+};
+
+static const struct nt51021_desc suez_boe_desc = {
+	.cmds = nt51021_init_suez_boe,
+	.num_cmds = ARRAY_SIZE(nt51021_init_suez_boe),
+};
+
+struct nt51021 {
+	struct drm_panel panel;
+	struct mipi_dsi_device *dsi;
+	struct regulator *supply;
+	struct gpio_desc *reset_gpio;
+	const struct nt51021_desc *desc;
+};
+
+static inline struct nt51021 *to_nt51021(struct drm_panel *panel)
+{
+	return container_of(panel, struct nt51021, panel);
+}
+
+/*
+ * A plain "write this value to that register" DCS write. The
+ * mipi_dsi_dcs_write_seq_multi() macro cannot be used for these: it stashes its
+ * payload in a static const array, so every byte has to be a compile-time
+ * constant, and both the register and the value here come from a table.
+ */
+static void nt51021_write(struct mipi_dsi_multi_context *ctx, u8 reg, u8 val)
+{
+	const u8 d[] = { reg, val };
+
+	mipi_dsi_dcs_write_buffer_multi(ctx, d, sizeof(d));
+}
+
+/*
+ * Open or close the in-band command window. Outside it the panel ignores DSI
+ * commands entirely, so every caller that wants to be heard must wrap its
+ * writes in a pair of these.
+ */
+static void nt51021_cmd_window(struct mipi_dsi_multi_context *ctx, bool open)
+{
+	nt51021_write(ctx, NT51021_CMD_IF,
+		      open ? NT51021_CMD_IF_FORCE_MIPI : NT51021_CMD_IF_RELEASE);
+	if (open)
+		mipi_dsi_msleep(ctx, 1);
+}
+
+static int nt51021_prepare(struct drm_panel *panel)
+{
+	struct nt51021 *ctx = to_nt51021(panel);
+	struct mipi_dsi_multi_context dsi_ctx = { .dsi = ctx->dsi };
+	unsigned int i;
+	int ret;
+
+	ret = regulator_enable(ctx->supply);
+	if (ret)
+		return ret;
+
+	/*
+	 * Vendor timing: reset held low for 20 ms, then 5 ms of settling before
+	 * the first command. The GPIO is active low in the device tree, so
+	 * "1" here means asserted.
+	 */
+	gpiod_set_value_cansleep(ctx->reset_gpio, 1);
+	msleep(20);
+	gpiod_set_value_cansleep(ctx->reset_gpio, 0);
+	usleep_range(5000, 6000);
+
+	nt51021_cmd_window(&dsi_ctx, true);
+
+	for (i = 0; i < ctx->desc->num_cmds; i++) {
+		const struct nt51021_cmd *cmd = &ctx->desc->cmds[i];
+
+		nt51021_write(&dsi_ctx, cmd->reg, cmd->val);
+		if (cmd->delay_ms)
+			mipi_dsi_msleep(&dsi_ctx, cmd->delay_ms);
+	}
+
+	/* Vendor init_lcm_registers(): Sleep Out, then 120 ms. */
+	mipi_dsi_dcs_exit_sleep_mode_multi(&dsi_ctx);
+	mipi_dsi_msleep(&dsi_ctx, 120);
+
+	nt51021_cmd_window(&dsi_ctx, false);
+	mipi_dsi_msleep(&dsi_ctx, 5);
+
+	if (dsi_ctx.accum_err) {
+		gpiod_set_value_cansleep(ctx->reset_gpio, 1);
+		regulator_disable(ctx->supply);
+	}
+
+	return dsi_ctx.accum_err;
+}
+
+static int nt51021_enable(struct drm_panel *panel)
+{
+	struct nt51021 *ctx = to_nt51021(panel);
+	struct mipi_dsi_multi_context dsi_ctx = { .dsi = ctx->dsi };
+
+	nt51021_cmd_window(&dsi_ctx, true);
+	mipi_dsi_dcs_set_display_on_multi(&dsi_ctx);
+	nt51021_cmd_window(&dsi_ctx, false);
+	mipi_dsi_msleep(&dsi_ctx, 20);
+
+	return dsi_ctx.accum_err;
+}
+
+static int nt51021_disable(struct drm_panel *panel)
+{
+	struct nt51021 *ctx = to_nt51021(panel);
+	struct mipi_dsi_multi_context dsi_ctx = { .dsi = ctx->dsi };
+
+	nt51021_cmd_window(&dsi_ctx, true);
+	mipi_dsi_dcs_set_display_off_multi(&dsi_ctx);
+	nt51021_cmd_window(&dsi_ctx, false);
+
+	return dsi_ctx.accum_err;
+}
+
+static int nt51021_unprepare(struct drm_panel *panel)
+{
+	struct nt51021 *ctx = to_nt51021(panel);
+	struct mipi_dsi_multi_context dsi_ctx = { .dsi = ctx->dsi };
+
+	/* Vendor lcm_suspend(): force MIPI, Sleep In, 50 ms. */
+	nt51021_cmd_window(&dsi_ctx, true);
+	mipi_dsi_dcs_enter_sleep_mode_multi(&dsi_ctx);
+	mipi_dsi_msleep(&dsi_ctx, 50);
+
+	gpiod_set_value_cansleep(ctx->reset_gpio, 1);
+	regulator_disable(ctx->supply);
+
+	return dsi_ctx.accum_err;
+}
+
+/*
+ * Timings from the vendor's lcm_get_params(): 1200x1920, hsync 1 / hbp 32 /
+ * hfp 110, vsync 1 / vbp 14 / vfp 11. That gives htotal 1343 and vtotal 1946.
+ *
+ * The pixel clock follows the vendor's PLL_CLOCK = 490, which after DDR is
+ * 980 Mbps per lane, and over four lanes at 24bpp is 163.33 MHz, giving
+ * 62.5 Hz. That is also what the bootloader leaves behind: read back before
+ * the kernel touches it, the MIPI TX PLL on this board decodes to exactly
+ * 980 Mbps per lane, so the panel is already running at this rate when the
+ * kernel takes over.
+ *
+ * Note for anyone tempted by the "fps=6025" the bootloader puts on the kernel
+ * command line: it is not the DSI rate and does not describe this PLL. It also
+ * is not stable across boots.
+ */
+static const struct drm_display_mode nt51021_mode = {
+	.clock = 163333,
+	.hdisplay = 1200,
+	.hsync_start = 1200 + 110,
+	.hsync_end = 1200 + 110 + 1,
+	.htotal = 1200 + 110 + 1 + 32,
+	.vdisplay = 1920,
+	.vsync_start = 1920 + 11,
+	.vsync_end = 1920 + 11 + 1,
+	.vtotal = 1920 + 11 + 1 + 14,
+	.width_mm = 136,
+	.height_mm = 221,
+	.type = DRM_MODE_TYPE_DRIVER | DRM_MODE_TYPE_PREFERRED,
+};
+
+static int nt51021_get_modes(struct drm_panel *panel,
+			     struct drm_connector *connector)
+{
+	return drm_connector_helper_get_modes_fixed(connector, &nt51021_mode);
+}
+
+static const struct drm_panel_funcs nt51021_panel_funcs = {
+	.prepare = nt51021_prepare,
+	.enable = nt51021_enable,
+	.disable = nt51021_disable,
+	.unprepare = nt51021_unprepare,
+	.get_modes = nt51021_get_modes,
+};
+
+static int nt51021_probe(struct mipi_dsi_device *dsi)
+{
+	struct device *dev = &dsi->dev;
+	struct nt51021 *ctx;
+	int ret;
+
+	ctx = devm_drm_panel_alloc(dev, struct nt51021, panel,
+				   &nt51021_panel_funcs,
+				   DRM_MODE_CONNECTOR_DSI);
+	if (IS_ERR(ctx))
+		return PTR_ERR(ctx);
+
+	ctx->supply = devm_regulator_get(dev, "power");
+	if (IS_ERR(ctx->supply))
+		return dev_err_probe(dev, PTR_ERR(ctx->supply),
+				     "failed to get power supply\n");
+
+	/* Active low in the device tree, so OUT_HIGH leaves reset asserted. */
+	ctx->reset_gpio = devm_gpiod_get(dev, "reset", GPIOD_OUT_HIGH);
+	if (IS_ERR(ctx->reset_gpio))
+		return dev_err_probe(dev, PTR_ERR(ctx->reset_gpio),
+				     "failed to get reset GPIO\n");
+
+	ctx->desc = of_device_get_match_data(dev);
+	if (!ctx->desc)
+		return -ENODEV;
+
+	ctx->dsi = dsi;
+	mipi_dsi_set_drvdata(dsi, ctx);
+
+	dsi->lanes = 4;
+	dsi->format = MIPI_DSI_FMT_RGB888;
+	/*
+	 * Video mode with sync events rather than sync pulses
+	 * (vendor SYNC_EVENT_VDO_MODE), and a continuously running HS clock
+	 * (vendor cont_clock = 1), so MIPI_DSI_CLOCK_NON_CONTINUOUS is
+	 * deliberately absent. Commands go out in LP.
+	 */
+	dsi->mode_flags = MIPI_DSI_MODE_VIDEO | MIPI_DSI_MODE_LPM;
+
+	ret = drm_panel_of_backlight(&ctx->panel);
+	if (ret)
+		return dev_err_probe(dev, ret, "failed to get backlight\n");
+
+	/* The DSI host must be running before we can send the init sequence. */
+	ctx->panel.prepare_prev_first = true;
+
+	drm_panel_add(&ctx->panel);
+
+	ret = mipi_dsi_attach(dsi);
+	if (ret < 0) {
+		drm_panel_remove(&ctx->panel);
+		return dev_err_probe(dev, ret, "failed to attach to DSI host\n");
+	}
+
+	return 0;
+}
+
+static void nt51021_remove(struct mipi_dsi_device *dsi)
+{
+	struct nt51021 *ctx = mipi_dsi_get_drvdata(dsi);
+	int ret;
+
+	ret = mipi_dsi_detach(dsi);
+	if (ret < 0)
+		dev_err(&dsi->dev, "failed to detach from DSI host: %d\n", ret);
+
+	drm_panel_remove(&ctx->panel);
+}
+
+static const struct of_device_id nt51021_of_match[] = {
+	{ .compatible = "amazon,suez-boe-nt51021", .data = &suez_boe_desc },
+	{ }
+};
+MODULE_DEVICE_TABLE(of, nt51021_of_match);
+
+static struct mipi_dsi_driver nt51021_driver = {
+	.probe = nt51021_probe,
+	.remove = nt51021_remove,
+	.driver = {
+		.name = "panel-novatek-nt51021",
+		.of_match_table = nt51021_of_match,
+	},
+};
+module_mipi_dsi_driver(nt51021_driver);
+
+MODULE_DESCRIPTION("Novatek NT51021 1200x1920 DSI panel driver");
+MODULE_LICENSE("GPL");

-- 
2.55.0

Re: [PATCH 0/2] drm/panel: add Novatek NT51021 1200x1920 DSI panel driver

[Ryan Brue]

Please disregard this series.

Nickolay Goppen posted an NT51021 driver a week before my patch, and I
didn't catch it.

https://lore.kernel.org/all/20260829-qcom-novatek-nt51021-panels-v1-0-cc1ca61bd261@mainlining.org/

I checked that there were no patches on the mailing list prior to
working on it, and then forgot to check prior to sending mine. What are
the odds, for an 8+ year old panel? :D

His covers more panels, but my panel seems to be slightly different
from his, based on the vendor kernel I referenced for my bringup. I
tested his driver on my hardware and it works; I'm going to make a
reply on that thread, and I will set up the Amazon Fire HD 10 panel as
a variant on top of his series instead.

Sorry for the noise, I'm new to this and I hope to learn quickly.

Ryan