Claude review: dma-buf: Change st-dma-resv.c to use kunit

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

Patch Review

Good conversion. The change from `static struct spinlock fence_lock;` with a runtime `spin_lock_init()` call (formerly in the deleted `dma_resv()` function) to `static DEFINE_SPINLOCK(fence_lock);` is a correct improvement since the old code relied on `dma_resv()` being called before any test used the lock.

The `KUNIT_ARRAY_PARAM_DESC` usage for iterating over `DMA_RESV_USAGE_KERNEL` through `DMA_RESV_USAGE_BOOKKEEP` is clean and idiomatic kunit.

**Stray trailing `\n`:** In `test_sanitycheck`, there's an inconsistent trailing newline:
```c
		KUNIT_FAIL(test, "Resv locking failed\n");
```
All other `KUNIT_FAIL` calls in this patch (and the entire series) omit the `\n`. Kunit adds its own newline when printing, so this should be removed for consistency.

**Dead `r = -EINVAL` assignments:** In `test_for_each` and `test_for_each_unlocked`, some error paths still assign `r = -EINVAL` after `KUNIT_FAIL`:
```c
		if (f != fence) {
			KUNIT_FAIL(test, "Unexpected fence");
			r = -EINVAL;
			goto err_unlock;
		}
```
While the "More than one fence found" path had its `r = -EINVAL` removed. Since the functions are now void and `r` is no longer returned, these surviving assignments are dead code. They're harmless, but the inconsistency is a bit untidy. Consider removing them all, or leaving them all -- pick one.

---
Generated by Claude Code Patch Reviewer