Claude review: drm/bridge: add drm_bridge_clear_and_put()

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

Patch Review

**Code review**:

The implementation at `drivers/gpu/drm/drm_bridge.c:340-346`:
```c
void drm_bridge_clear_and_put(struct drm_bridge **bridge_pp)
{
	struct drm_bridge *bridge = *bridge_pp;

	*bridge_pp = NULL;
	drm_bridge_put(bridge);
}
```

- **Concurrency concern**: As noted above, both the read of `*bridge_pp` and the write of `NULL` use plain accesses. If this is meant to be safe against concurrent readers (as the commit message strongly implies), the write should be `WRITE_ONCE(*bridge_pp, NULL)` and the initial read should be a corresponding `READ_ONCE`. The kernel has established patterns for this — e.g., `rcu_assign_pointer()` or at minimum `WRITE_ONCE()` / `smp_store_release()`. Without these, the compiler is free to reorder the NULL store after the `drm_bridge_put()` call or even optimize it in unexpected ways.

- **Documentation quality**: The kerneldoc is well-written, with clear examples showing the equivalent manual pattern. The cross-reference added to `drm_bridge_put()` is a nice touch.

- **Naming**: `clear_and_put` follows the existing kernel convention (similar to `fput_and_clear`, though that one is actually named `fput` not `put_and_clear`). The kernel has `kfree_and_null` patterns too. The name is fine.

- **Minor**: The function doesn't explicitly document thread-safety guarantees. If it's not intended to be called concurrently with readers, that should be stated. If it is, it needs the memory barriers.

---
Generated by Claude Code Patch Reviewer