Claude review: drm/amd/display: Fix a benign uninitialized variable warning

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

Patch Review

**Status: Looks good**

The original code was:

```c
adev->dm.cached_state = drm_atomic_helper_suspend(adev_to_drm(adev));
if (IS_ERR(adev->dm.cached_state)) {
    r = PTR_ERR(adev->dm.cached_state);
    adev->dm.cached_state = NULL;
}

return adev->dm.cached_state ? 0 : r;
```

GCC correctly identifies that if `drm_atomic_helper_suspend()` returns a valid (non-NULL, non-ERR) pointer, the `IS_ERR` branch is not taken and `r` remains uninitialized — yet the final `return` expression `adev->dm.cached_state ? 0 : r` would use `r` if `cached_state` were NULL. In practice this path is unreachable (a valid pointer is not NULL), but the compiler can't prove that.

The fix replaces this with:

```c
r = PTR_ERR_OR_ZERO(adev->dm.cached_state);
if (r)
    adev->dm.cached_state = NULL;

return r;
```

This is logically equivalent and cleaner:
- `PTR_ERR_OR_ZERO()` returns 0 for valid pointers and the error code for `IS_ERR` pointers, so `r` is always initialized.
- The `if (r)` branch correctly NULLs out the cached state on error.
- `return r` returns 0 on success, the negative error code on failure — same behavior as before.

**Minor observations:**
- The commit message is well-written and honest about the cosmetic nature of the fix.
- The `int r` declaration could now be combined with initialization (`int r = PTR_ERR_OR_ZERO(...)`) to further simplify, but keeping the declaration separate is fine and matches kernel style in this file.
- No functional change, no risk of regression.

**Reviewed-by worthy.**

---
Generated by Claude Code Patch Reviewer