Claude review: drm/msm/hdmi: switch to generic PHY subsystem

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

Patch Review

This is the core patch — it removes the ad-hoc HDMI PHY code from the DRM driver and creates new PHY drivers under `drivers/phy/qualcomm/`. The DRM side now uses the standard `devm_phy_get()`, `phy_init()`, `phy_configure()`, `phy_power_on()` / `phy_power_off()`, `phy_exit()` API. This is clean.

**Bug: Missing `qcom,hdmi-phy-8998` in OF match table**

The old driver had entries for both `qcom,hdmi-phy-8996` and `qcom,hdmi-phy-8998`. The new `phy-qcom-qmp-hdmi-base.c` only has:

```c
static const struct of_device_id qmp_hdmi_of_match_table[] = {
	{
		.compatible = "qcom,hdmi-phy-8996", .data = &qmp_hdmi_8996_cfg,
	},
	{ },
};
```

The MSM8998 code is fully implemented in `phy-qcom-qmp-hdmi-msm8998.c` with `qmp_hdmi_8998_cfg` exported, but it's never wired into the match table. This means MSM8998 HDMI will be broken. Add:
```c
	{ .compatible = "qcom,hdmi-phy-8998", .data = &qmp_hdmi_8998_cfg, },
```

**Bug: Wrong pointer in MSM8996 PHY `set_rate`**

In `phy-qcom-qmp-hdmi-msm8996.c`, line 5926 of the mbox:
```c
		writel(11000000 / (parent_rate / 20),
		       hdmi_phy + QSERDES_COM_CP_CTRL_MODE0);
```

This writes to `hdmi_phy + offset` (a `struct qmp_hdmi_phy *` plus an integer offset), not `hdmi_phy->serdes + offset`. This is a pointer arithmetic bug that will corrupt memory or cause a crash. Should be:
```c
		       hdmi_phy->serdes + QSERDES_COM_CP_CTRL_MODE0);
```

**Minor: `pm_runtime_put_noidle` in phy_exit**

Both `qcom_hdmi_preqmp_phy_exit()` and `qmp_hdmi_phy_exit()` use `pm_runtime_put_noidle()`, which decrements the refcount without actually triggering runtime suspend. The matching `phy_init` uses `pm_runtime_resume_and_get()`. Consider using `pm_runtime_put` or `pm_runtime_put_sync` instead to actually suspend the device when no longer in use.

**Minor: `devm_phy_get` error message**

In `hdmi.c`:
```c
	hdmi->phy = devm_phy_get(&pdev->dev, NULL);
	if (IS_ERR(hdmi->phy)) {
		DRM_DEV_ERROR(&pdev->dev, "failed to get phy\n");
		return PTR_ERR(hdmi->phy);
	}
```

`devm_phy_get` already returns `-EPROBE_DEFER` etc. Consider using `dev_err_probe()` instead of `DRM_DEV_ERROR` to avoid spamming the log during deferred probing.

**Minor: Duplicate `write16`/`write24`/`read24` helpers**

The `write16()`, `write24()`, `read24()` helper functions are defined identically in three files: `phy-qcom-hdmi-28lpm.c`, `phy-qcom-qmp-hdmi-msm8996.c`, and `phy-qcom-qmp-hdmi-msm8998.c`. These could be shared through a common header (e.g., `phy-qcom-qmp-hdmi.h` or a utility header).

**Minor: `__maybe_unused` on pm ops**

The runtime PM callbacks use `__maybe_unused` annotations. Since these are always used (referenced by `DEFINE_RUNTIME_DEV_PM_OPS`), the annotation is unnecessary but harmless.

**Minor: `of_device_get_match_data` deprecation**

Both `phy-qcom-hdmi-preqmp.c` and `phy-qcom-qmp-hdmi-base.c` use `of_device_get_match_data()`. Modern kernel code prefers `device_get_match_data()`.

**Minor: Typo in Kconfig**

```
+	  Enable this to support the Qualcomm HDMI PHY presend on 32-bit platforms:
```
"presend" should be "present".

---
Generated by Claude Code Patch Reviewer