// SPDX-License-Identifier: GPL-2.0 OR MIT
// LILYGO T-Display-K230.
//
// Derived from canaan/k230-canmv-v3-lcd.dts, which is the reference this
// board is a variant of -- U-Boot on our hardware prints
// "Model: kendryte k230 canmv v3.0". Every divergence from that reference
// is marked DIVERGENCE below and listed in
// docs/evidence/dts-divergence.md.
//
// UNVERIFIED: this has never been booted.

/dts-v1/;

#include "k230.dtsi"
#include "display-rm69a10-568x1232.dtsi"

/ {
	model = "LILYGO T-Display-K230";
	compatible = "lilygo,t-display-k230", "canaan,canmv-k230", "canaan,kendryte-k230";

	aliases {
		serial3 = &uart3;
		// DIVERGENCE (naming only): on this board mmc0 carries the
		// RTL8189FTV over SDIO and mmc1 carries the TF card. The
		// reference names them the same way; recorded because
		// docs/findings.md originally had them swapped.
		mmc0 = &mmc_sd0;
		mmc1 = &mmc_sd1;
		i2c0 = &i2c4;
		i2c1 = &i2c3;
	};

	chosen {
		// Linux console. Note U-Boot's own console is ttyS1 per the
		// SDK environment; the two legitimately differ.
		stdout-path = "serial0:115200n8";
	};

	ddr: memory@0 {
		device_type = "memory";
		reg = <0x0 0x0 0x0 0x40000000>;
	};

	// DIVERGENCE: stage 1 scans this XRGB buffer until Linux owns the panel.
	// The values come from nix/boot-splash.nix through device-tree.nix's
	// preprocessor definitions, shared with U-Boot's Kconfig fragment.
	reserved-memory {
		#address-cells = <2>;
		#size-cells = <2>;
		ranges;

		splash_framebuffer: framebuffer@K230_SPLASH_FRAMEBUFFER_UNIT_ADDRESS {
			reg = <0x0 K230_SPLASH_FRAMEBUFFER_ADDRESS
			       0x0 K230_SPLASH_FRAMEBUFFER_SIZE>;
			no-map;
		};
	};

	// PMU INT0 power key. The kernel input driver reports edges; userspace
	// decides short press vs. hold. Physical-key proof is still pending.
	soc {
		pmu_pwrkey: pmu-pwrkey@91000000 {
			compatible = "canaan,k230-pmu-pwrkey";
			reg = <0x0 0x91000000 0x0 0xc00>,
			      <0x0 0x91103000 0x0 0x1000>;
			reg-names = "pmu", "pwr";
			interrupts = <175 IRQ_TYPE_LEVEL_HIGH>;
			wakeup-source;
		};
	};

	sound {
		status                      = "okay";
		compatible                  = "canaan,k230-audio-inno";
		canaan,model                = "K230_I2S_INNO";
		canaan,k230-i2s-controller  = <&i2s>;
		canaan,k230-audio-codec     = <&inno_codec>;

		// Deliberately absent: canaan,external-i2s-output-default. Leaving
		// it unset keeps this board booting through the Inno codec
		// (headphone/line-out) exactly as before; the external route below
		// is opt-in at runtime via the "External I2S Output Switch" ALSA
		// control (nix/patches/canaan-audio-external-i2s-switch.patch).
	};
};

// MAX98357A external I2S amplifier -- the-handheld-plays-through-its-speaker.
//
// LILYGO's own BSP (Xinyuan-LilyGO/T-Display-K230,
// k230_bsp/docs/HARDWARE_PINMAP.md) places this part on the OPTIONAL
// "nRF52840 BLE / Audio / Sensor" add-on board, not the main T-Display-K230
// board and not the nRF9151 keyboard base -- confirmed by that same file's
// three-way board grouping and by CHANGELOG.MD listing it alongside the
// nRF52840/nRF9151/LoRa presets under "Included Hardware Support". Whether
// THIS board has that add-on attached is UNVERIFIED; nothing in this repo
// has recorded one before this change.
//
//   BCLK   GPIO32   I2S bit clock, shared with the Inno codec's I2S output
//   LRCK   GPIO33   I2S word-select, shared with the Inno codec's I2S output
//   DOUT   GPIO35   I2S data, shared with the Inno codec's I2S output
//   SDMODE GPIO34   plain GPIO, high enables the amp, low shuts it down
//
// Grounding for all four: k230_bsp/docs/HARDWARE_PINMAP.md's "nRF52840 BLE /
// Audio / Sensor Base Board" table, and
// k230_bsp/overlay/buildroot-overlay/linux/0058-riscv-dts-rm69a10-add-audio-fan-sensor-pins.patch
// (commit 6b5d4c3a, same repo). GPIO34 is a plain GPIO, not behind the
// XL9555 expander -- that expander is documented in the same file under the
// SEPARATE nRF9151 keyboard base, for its own LEDs and keyboard matrix only.
//
// UNVERIFIED: whether GPIO32/33/35 are actually muxed to their I2S
// alternate function on THIS board. This SoC's FPIOA pin mux is set by
// U-Boot, not by the Linux device tree -- there is no pinctrl driver
// anywhere in the pinned kernel tree (see docs/dts-evidence.md, "There is
// no pinctrl in the Linux device tree"), and LILYGO's own pinctrl-based
// patch (0058, above) targets a pinctrl-k230 driver that patch series does
// not otherwise add here and that this pinned tree does not carry. Setting
// the ALSA route to "external" only tells the SAI peripheral to stop
// feeding the Inno codec and drive these pads directly; it says nothing
// about whether U-Boot already muxed them for I2S rather than plain GPIO.
// See docs/evidence/max98357a-speaker.md.
//
// GPIO34 needs no device-tree node to be usable: it is a plain,
// unclaimed line on the existing always-on gpio1_ports controller
// (k230.dtsi:415-421), driven directly by libgpiod from userspace --
// exactly how LILYGO's own launcher does it (ui_hardware.c:1655-1731,
// AMP_SHUTDOWN_GPIO 34). A "maxim,max98357a" ALSA codec DT node was
// deliberately NOT added: that driver claims its sdmode-gpios exclusively
// via devm_gpiod_get, which would conflict with a userspace helper toggling
// the same line, and nothing here ever wires such a node into a dai-link
// for its DAPM power sequencing to have anything to switch. Naming the pin
// below is documentation only; it claims nothing.
&gpio1_ports {
	gpio-line-names =
		"MAX98357A_I2S_BCLK",
		"MAX98357A_I2S_LRCK",
		"MAX98357A_SDMODE",
		"MAX98357A_I2S_DOUT";
};

&uart0 {
	status = "okay";
};

&uart3 {
	status = "okay";
};

// The TF card.
&mmc_sd1 {
	status = "okay";
	no-1-8-v;
	cap-sd-highspeed;
	mshc_ctrl_r = <0x00>;
};

// The RTL8189FTV, over SDIO.
&mmc_sd0 {
	status = "okay";
	no-1-8-v;
	rx_delay_line = <0x0d>;
	tx_delay_line = <0xc0>;
};

&usb0 {
	status = "okay";
};

&usb1 {
	status = "okay";
};

&i2c3 {
	status = "okay";

	// DIVERGENCE: the reference fits an edt,edt-ft5306 at 0x38. This
	// board has a Goodix GT9895 at 0x5d -- a Berlin-generation part,
	// which the pinned 6.6 tree does not support. The driver arrives as
	// a backport in this change; without it this node simply does not
	// probe and the panel still works.
	//
	// The GPIOs are the same as the reference's touch controller:
	// RST on 24, INT on 23. Confirmed against the schematic.
	//
	// Two compatibles, most-specific first, which is how a fallback is
	// meant to be written. The part really is a GT9895, and that string
	// is what a driver gaining real GT9895 support would match on. But
	// the backported goodix_berlin matches ONLY "goodix,gt9916" --
	// nix/patches/goodix-berlin/goodix_berlin_i2c.c:57 -- so with the
	// specific string alone nothing binds and the experiment task 4.1
	// asks for cannot even run. The fallback lets the Berlin driver
	// attach while the device tree keeps telling the truth about the
	// silicon.
	//
	// It is close enough: the backport binds, reads the firmware version
	// and the IC-info blob without error, and derives touch_data_addr
	// from the part at runtime. See docs/evidence/gt9895-port.md.
	touchscreen@5d {
		compatible = "goodix,gt9895", "goodix,gt9916";
		reg = <0x5d>;
		reset-gpios = <&gpio0_ports 24 GPIO_ACTIVE_LOW>;
		interrupt-parent = <&gpio0_ports>;

		// LEVEL_LOW, not EDGE_FALLING.
		//
		// EDGE_FALLING was inherited from the reference board's
		// edt-ft5306 (k230-canmv-v3-lcd.dts:66), a different part,
		// and it is why touch bound but reported nothing: four
		// interrupts on IRQ 87 since boot and then silence, with a
		// live controller answering i2c. A Goodix Berlin part holds
		// INT asserted low until the host clears the status byte at
		// touch_data_addr; goodix_berlin_irq() returns early WITHOUT
		// clearing whenever it reads status == 0
		// (nix/patches/goodix-berlin/goodix_berlin_core.c:566), so
		// one such early return leaves the line low forever and an
		// edge detector never sees another falling edge. A level
		// detector re-asserts immediately and the handler clears it.
		//
		// caveman99/k230-linux runs this same backported driver on
		// this same 6.6.36 tree on this same board with
		// IRQ_TYPE_LEVEL_LOW and reports working multitouch; its
		// /proc/interrupts line reads "gpio-k230 23 Level
		// goodix-berlin" where ours reads "Edge". gpio-k230 supports
		// both (drivers/gpio/gpio-k230.c:333).
		interrupts = <23 IRQ_TYPE_LEVEL_LOW>;

		// The digitizer grid, not the panel. goodix_berlin declares
		// ABS_MT_POSITION_X/Y max = SZ_64K-1 and then lets
		// touchscreen_parse_properties() override it
		// (goodix_berlin_core.c:621-627); with nothing to read, every
		// touch lands in the top-left 1% of any client that scales by
		// the declared range.
		//
		// 1024 x 2400 has two independent sources: LILYGO's own
		// RT-Smart driver, which divides raw coordinates by exactly
		// these constants
		// (repo/canmv_k230/src/rtsmart/rtsmart/kernel/bsp/maix3/drivers/extdrv/touch/gt9895.c:44-45),
		// and caveman99's Linux port. The panel is 568x1232, so the
		// sensor grid is roughly 1.8x the pixel grid in both axes --
		// same orientation, hence no swapped-x-y or inverted-* here.
		/* The controller reports X 0..1079: a finger at the
		 * long edge hit raw 1079 on the second board
		 * (docs/evidence/panel-landscape/touch-range.md), and
		 * LILYGO's own tree gives panel-max-x 1060. 1024 lost
		 * the last 5% of the glass past the panel edge. */
		touchscreen-size-x = <1080>;
		touchscreen-size-y = <2400>;
	};
};

&i2c4 {
	status = "okay";
};

&mipi0 {
	// set to csi2
	id = <2>;
	reg = <0x0 0x9000a800 0x0 0x800>;
	interrupts = <121 IRQ_TYPE_LEVEL_HIGH>;
	resets = <&sysctl_reset K230_RESET_CSI2_REG_OFFSET K230_RESET_CSI2_TYPE
			  K230_RESET_CSI2_DONE_BIT K230_RESET_CSI2_ASSERT_BIT>,
	<&sysctl_reset K230_RESET_M2_REG_OFFSET K230_RESET_M2_TYPE
		 K230_RESET_M2_DONE_BIT K230_RESET_M2_ASSERT_BIT>;
};

// Reset and backlight for the panel. `lcd` is the label on the panel node
// in display-rm69a10-568x1232.dtsi; the panel dtsi deliberately carries no
// GPIOs of its own, matching how the reference splits panel from board.
//
// GPIO22 reset is ACTIVE_HIGH and GPIO25 is the backlight -- both taken
// from k230-canmv-v3-lcd.dts, and both were unestablished before that.
&lcd {
	dsi_reset-gpios = <&gpio0_ports 22 GPIO_ACTIVE_HIGH>;
	backlight_gpio-gpios = <&gpio0_ports 25 GPIO_ACTIVE_HIGH>;
};