Claude review: backlight: Add SY7758 6-channel High Efficiency LED Driver support

[Claude Code Review Bot <claude-review@example.com>]

Patch Review

**Issue 1 (medium) — No backlight off control when brightness is 0:**

`sy7758_backlight_update_status()` only writes the brightness registers:

```c
static int sy7758_backlight_update_status(struct backlight_device *backlight_dev)
{
	struct sy7758 *sydev = bl_get_data(backlight_dev);
	unsigned int brightness = backlight_get_brightness(backlight_dev);
	int ret;

	ret = regmap_write(sydev->regmap, REG_BRT_12BIT_L,
			   FIELD_PREP(MSK_BRT_12BIT_L,
				      brightness & 0xff));
	...
	ret = regmap_write(sydev->regmap, REG_BRT_12BIT_H,
			   FIELD_PREP(MSK_BRT_12BIT_H,
				      (brightness >> 8) & 0xf));
	...
}
```

When brightness reaches 0, the LED channels remain powered (`BIT_CFG2_BL_ON` stays set). Compare with ktz8866 which toggles `BL_EN_BIT` on brightness 0, and aw99706 which toggles `BACKLIGHT_EN_MASK`. This matters for power consumption during suspend (the backlight core sets brightness=0 on suspend due to `BL_CORE_SUSPENDRESUME`). The `update_status` callback should also clear/set `BIT_CFG2_BL_ON` based on whether brightness is zero.

Similarly, `sy7758_remove()` calls `backlight_disable()` but that only writes brightness=0 — without toggling `BIT_CFG2_BL_ON`, the chip may still be driving the LEDs at minimum current.

**Issue 2 (minor) — Hardcoded hardware configuration limits reusability:**

`sy7758_init()` writes board-specific values directly:

```c
	ret = regmap_write(sydev->regmap, REG_OTP_CFG0,
			   FIELD_PREP(MSK_CFG0_CURRENT_LOW, 85));
	...
	ret = regmap_write(sydev->regmap, REG_OTP_CFG1,
			   BIT_CFG1_PDET_STDBY |
			   FIELD_PREP(MSK_CFG1_CURRENT_MAX, 1) |
			   FIELD_PREP(MSK_CFG1_CURRENT_HIGH, 10));
```

The LED current (85 + 10<<8 = 2645 in whatever unit), PWM frequency (mode 4), boost voltage max (mode 4), UVLO enable, etc. are all hardcoded for the Ayaneo Pocket S2. A second board using SY7758 would need to fork the init or add DT properties. This is acceptable for a v1 driver targeting a single board, but worth a comment acknowledging the limitation, or better yet, reading `default-brightness` and `max-brightness` from the DT as `common.yaml` provides them.

**Issue 3 (nit) — Redundant masking with FIELD_PREP:**

```c
	ret = regmap_write(sydev->regmap, REG_BRT_12BIT_L,
			   FIELD_PREP(MSK_BRT_12BIT_L,
				      brightness & 0xff));
```

`MSK_BRT_12BIT_L` is `GENMASK(7, 0)` = 0xFF, starting at bit 0. `FIELD_PREP` already masks the value to fit the field, so the explicit `& 0xff` is redundant. Same for `(brightness >> 8) & 0xf` with `MSK_BRT_12BIT_H` = `GENMASK(3, 0)`. This appears in both `update_status` and `sy7758_init`. Not a bug, just noise.

**Issue 4 (nit) — Unnecessary initialization of `ret`:**

```c
static int sy7758_init(struct sy7758 *sydev)
{
	int ret = 0;

	ret = regmap_write(sydev->regmap, REG_DEV_CTL, ...);
```

`ret` is initialized to 0 then immediately overwritten. Should be just `int ret;`.

**Issue 5 (nit) — Last return in `sy7758_init` could be simplified:**

```c
	ret = regmap_write(sydev->regmap, REG_OTP_CFG2,
			   BIT_CFG2_BL_ON | BIT_CFG2_UVLO_EN);
	if (ret)
		return ret;

	return 0;
```

This could be simply:

```c
	return regmap_write(sydev->regmap, REG_OTP_CFG2,
			    BIT_CFG2_BL_ON | BIT_CFG2_UVLO_EN);
```

**Positive notes:**

- Good use of `devm_regulator_get_enable()` and `devm_gpiod_get()` for resource management.
- Proper `dev_err_probe()` usage throughout.
- Device ID check is solid with a clear error message.
- The `fsleep(10000)` (10ms) after asserting the enable GPIO before attempting I2C communication is reasonable and was already iterated upon in earlier versions.
- Kconfig correctly uses `select REGMAP_I2C` and `depends on I2C`.
- Makefile entry maintains alphabetical order.

---
Generated by Claude Code Patch Reviewer