Claude review: phy: freescale: Add DisplayPort/HDMI Combo-PHY driver for i.MX8MQ

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

Patch Review

Large PHY driver (1359 lines) with extensive PLL configuration tables.

**Issues:**

1. **`cdns_phy->base` pointer lifetime**: The PHY obtains the `cdns_mhdp_base` pointer from its parent via `dev_get_drvdata(cdns_phy->dev->parent)` in `cdns_hdptx_phy_init()`. This pointer is valid as long as the parent bridge device is alive, but there's no reference counting or deferred probe protection. If the parent driver is unbound first, `cdns_phy->base` becomes a dangling pointer. The `if (!cdns_phy->base)` checks in power_on/off/set_mode only protect against init not being called yet, not against parent unbind.

2. **`hdptx_clk_enable()` uses `devm_clk_get_enabled()` at probe time**: This means clocks are enabled from probe until device removal. For a PHY that might not always be in use, this wastes power. Consider using `devm_clk_get()` at probe and `clk_prepare_enable()`/`clk_disable_unprepare()` in `power_on`/`power_off`.

3. **`ndelay(150)` usage in `hdptx_dp_aux_cfg()`**: Many sequential `ndelay(150)` calls between register writes. While the hardware may require these delays, `ndelay` is busy-wait and can be inaccurate on some platforms. The existing approach is functional but worth noting.

4. **Error accumulator pattern**: The `int *err` parameter on `cdns_phy_reg_read/write` is a nice pattern for batching error checks, but the accumulated error could mask which specific register access failed during debugging.

5. **Multi-line comment style**:
```c
/* Configure PHY in A0 mode (PHY must be in the A0 power
 * state in order to transmit data)
 */
```
Should follow kernel style: `/*` on its own line for multi-line comments, or use single-line.

---
Generated by Claude Code Patch Reviewer