Claude review: drm/panel: Add panel driver for ChipWealth CH13726A based panels

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

Patch Review

**Issue 1 (bug): Incorrect Kconfig selects**

```
+config DRM_PANEL_CHIPWEALTH_CH13726A
+	tristate "CHIPWEALTH CH13726A-based DSI panel"
+	depends on OF
+	depends on DRM_MIPI_DSI
+	depends on BACKLIGHT_CLASS_DEVICE
+	select DRM_DISPLAY_DP_HELPER
+	select DRM_DISPLAY_HELPER
```

`DRM_DISPLAY_DP_HELPER` is for DisplayPort/eDP panels only. In the entire panel Kconfig, the only entries selecting it are `DRM_PANEL_SAMSUNG_ATNA33XC20` (an eDP panel) and `DRM_PANEL_EDP` (the generic eDP panel driver). This DSI panel driver does not use any DP or display helpers — it only uses `drm_mipi_dsi.h`, `drm_panel.h`, and `drm_modes.h`. Both `select` lines should be removed.

**Issue 2 (minor): `thor_bottom_desc` should be `const`**

```c
+static struct ch13726a_desc thor_bottom_desc = {
```

This should be `static const struct ch13726a_desc thor_bottom_desc`. The data is never modified at runtime, and the `of_match_table` `.data` field is `const void *`. This also ties into Issue 3 below.

**Issue 3 (minor): const-correctness for `desc` field**

```c
+struct ch13726a_panel {
+	...
+	struct ch13726a_desc *desc;
+	...
+};
```

The `desc` pointer should be `const struct ch13726a_desc *desc`, since the descriptor data pointed to (from the match table) should never be modified. The assignment in probe:

```c
+	ctx->desc = (struct ch13726a_desc *)of_device_get_match_data(dev);
```

currently casts away the `const` returned by `of_device_get_match_data()` (which returns `const void *`). With a `const` field, the explicit cast would be unnecessary:

```c
	ctx->desc = of_device_get_match_data(dev);
```

**Issue 4 (style): `uint8_t` loop variable**

```c
+	for (uint8_t i = 0; i < ctx->desc->num_modes; i++) {
```

The kernel convention is to use `unsigned int` for loop counters, not fixed-width types like `uint8_t`. A `uint8_t` counter is also fragile — it would silently wrap if `num_modes` ever exceeded 255, though that's admittedly unlikely for display modes.

**Issue 5 (nit): `of_device_get_match_data` vs `device_get_match_data`**

```c
+	ctx->desc = (struct ch13726a_desc *)of_device_get_match_data(dev);
```

Newer panel drivers are moving to the bus-agnostic `device_get_match_data(dev)`. While `of_device_get_match_data` still works, reviewers may prefer the newer API since the driver already `depends on OF` anyway and the function is equivalent for OF-based matching.

**Issue 6 (observation): LPM toggling in disable**

```c
+static int ch13726a_disable(struct drm_panel *panel)
+{
+	...
+	ctx->dsi->mode_flags &= ~MIPI_DSI_MODE_LPM;
+
+	mipi_dsi_dcs_set_display_off_multi(&dsi_ctx);
+	mipi_dsi_msleep(&dsi_ctx, 50);
+	mipi_dsi_dcs_enter_sleep_mode_multi(&dsi_ctx);
```

The disable path clears `MIPI_DSI_MODE_LPM` before sending DCS commands, meaning display-off and enter-sleep are sent in HS mode. Meanwhile `ch13726a_on()` sets `MIPI_DSI_MODE_LPM` to send init commands in LP mode. This asymmetry is present in some other drivers (e.g., `panel-jdi-fhd-r63452.c`), so it may be correct per the panel datasheet, but it's worth confirming that the CH13726A actually requires HS-mode for shutdown commands. If LP mode works for both init and shutdown, the mode_flags manipulation could be simplified.

**Non-issues / things done well:**
- Correct use of `devm_drm_panel_alloc()` (modern allocation pattern)
- Correct use of `devm_regulator_bulk_get_const()` with a static const supply array
- Proper use of `mipi_dsi_multi_context` for init sequence error accumulation
- `prepare_prev_first = true` is appropriate for a DSI video mode panel
- The 120Hz/60Hz mode clock calculations look correct: `(1080+28+4+36) * (1240+16+4+8) * rate / 1000` = `1148 * 1268 * 120 / 1000` = 174,614 for 120Hz
- Backlight via DCS brightness is standard
- Error cleanup paths in prepare and probe are correct

---
Generated by Claude Code Patch Reviewer