Claude review: dma-fence: Clarify external lock use case in dma_fence_init() docs

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

Patch Review

**Correctness of the new wording:**

The replacement text for `dma_fence_init()`:

```
+ * External locks are a relic of legacy use cases that needed a shared lock
+ * to serialize signaling when no out-of-order signaling was possible through
+ * &dma_fence_ops.signaled. Drivers have abandoned this concept since the
+ * introduction of the callback, but the external lock is still around. New
+ * users MUST NOT use external locks, as they force the issuer to outlive all
+ * fences that reference the lock.
```

This is accurate. The old comment implied the shared lock *prevented* out-of-order signaling, but in reality signaling order is the driver's responsibility. The new text correctly identifies the actual purpose (serializing signaling) and the actual problem with external locks (lifetime coupling — the issuer must outlive all fences). The reference to `&dma_fence_ops.signaled` as the mechanism that made external locks unnecessary is technically sound, as the `.signaled` callback allows the fence framework to peek at fence state without holding the shared lock.

**De-duplication in `dma_fence_init64()`:**

```
+ * New users MUST NOT use external locks. Check the documentation in
+ * dma_fence_init() to understand the motives behind the legacy use cases.
```

This is a reasonable approach. One very minor nit: kerneldoc cross-references typically use `&function()` or are more explicit. The plain text "dma_fence_init()" will render fine in the generated docs, but using `dma_fence_init()` as-is is conventional enough and readable. Not worth a respin.

**Minor observation:** The phrase "when no out-of-order signaling was possible through `&dma_fence_ops.signaled`" could be read two ways — either "when the `.signaled` callback couldn't handle out-of-order signaling" or "when out-of-order signaling was not possible [and they relied on] `.signaled`". In context, the intended meaning (the former) is clear enough, especially for the audience of dma-fence developers.

**No issues found.** This is a clean documentation improvement with correct technical content. 

Reviewed-by recommendation: **Accept as-is.**

---
Generated by Claude Code Patch Reviewer