Claude review: drm: property: use kzalloc_flex

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

Patch Review

**Correctness: Good**

- The conversion from a separate `uint64_t *values` pointer to `uint64_t values[] __counted_by(num_values)` flexible array member is valid. The struct is only ever allocated on the heap via `drm_property_create()`, never embedded in another struct or placed on the stack.

- The `kzalloc_flex(*property, values, num_values)` call correctly allocates `sizeof(struct drm_property) + num_values * sizeof(uint64_t)`.

- Moving `property->num_values = num_values` to immediately after allocation is required by `__counted_by` semantics — the count field must be set before any access to the flexible array. This is done correctly.

**Cleanup correctness: Good**

- Removing `kfree(property->values)` from both the error path in `drm_property_create()` and from `drm_property_destroy()` is correct since the values are now part of the same allocation as the struct.

- The `if (property->num_values)` guard before `kfree(property->values)` was unnecessary even before (kfree(NULL) is safe), but removing both lines is the right thing with the flexible array.

**Minor observations:**

- The original code had `if (num_values)` guarding the `kcalloc` — with `kzalloc_flex`, a `num_values` of 0 will allocate just `sizeof(struct drm_property)` with no trailing array space, which is fine. All callers that pass `num_values=0` never access `property->values[]`, so this is safe.

- No other in-tree code assigns to `property->values` as a pointer (verified via grep), so there are no missed conversion sites.

No issues found. The patch is ready to merge.

---
Generated by Claude Code Patch Reviewer