Claude review: drm/hisilicon/hibmc: fix no showing when no connectors connected

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

Patch Review

This is the most complex patch in the series and has several issues:

**1. Comment style violation:** The multi-line comment in `hibmc_vdac_detect()` uses the wrong style for the kernel:
```c
	/* If the DP connectors are disconnected, the hibmc_vdac_detect function
	 * must return a connected state to ensure KVM display functionality.
```
This should start with `/*` on its own line per kernel coding style (though this is a minor nit — some subsystems accept this style).

**2. Race / ordering dependency between DP and VDAC detect:** `hibmc_vdac_detect()` reads `priv->dp.phys_state` to decide whether to force VGA connected. But there's no guarantee about the order in which connectors are polled. If the VGA connector is polled/detected *before* the DP connector, `dp.phys_state` could still hold a stale value from a previous cycle (or the initial `connector_status_disconnected`). On initial boot with DP connected, `dp.phys_state` starts as `connector_status_disconnected`, so the first VGA detect would force-return connected even if DP is actually plugged in but not yet detected.

**3. `drm_connector_helper_get_modes()` usage:** The `get_modes` callback now calls `drm_connector_helper_get_modes()` which reads the EDID via the connector's DDC adapter. The old code explicitly used `drm_edid_read_ddc(connector, &vdac->adapter)`. This should work since the connector was initialized with `drm_connector_init_with_ddc(..., &vdac->adapter)`, but it's worth confirming that `drm_connector_helper_get_modes` will use this DDC adapter path.

**4. Clearing EDID when falling back:** The `drm_edid_connector_update(connector, NULL)` call when falling back to noedid modes is good — it clears stale EDID data.

**5. `drm_dp_read_sink_count` return value reuse:** In `hibmc_dp_detect()`, the variable `ret` is reused to hold both the `drm_dp_read_sink_count()` return value (which can be a negative error or a positive count) and the connector status enum. The `goto exit` path then assigns this to `dp->phys_state`. If `drm_dp_read_sink_count()` returns a negative error, the code falls through to `exit:` with `ret` set to a negative errno, which gets stored in `dp->phys_state`. While the original code had the same issue (it would return `connector_status_disconnected` by falling through), the new code stores a negative errno in `phys_state`, which is neither `connector_status_connected` nor `connector_status_disconnected`. This is a bug — the error case from `drm_dp_read_sink_count` should explicitly set `ret = connector_status_disconnected` before the goto.

---
Generated by Claude Code Patch Reviewer