Claude review: driver core: platform: provide platform_device_set_of_node_from_dev()

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

Patch Review

```c
void platform_device_set_of_node_from_dev(struct platform_device *pdev,
					  const struct device *dev2)
{
	device_set_of_node_from_dev(&pdev->dev, dev2);
	pdev->dev.fwnode = of_fwnode_handle(pdev->dev.of_node);
}
```

This wraps `device_set_of_node_from_dev()` (which does `of_node_get`/`of_node_put` and sets `of_node_reused = true`) and then syncs the fwnode pointer.

**Minor concern for post-patch-23**: After patch 23 changes release to `fwnode_handle_put()`, this function sets `pdev->dev.fwnode` without calling `fwnode_handle_put()` on the old value first. If the old fwnode was a non-OF fwnode that was set via `platform_device_set_fwnode()` (which after patch 23 takes a ref via `fwnode_handle_get()`), this would leak. In practice this is extremely unlikely since these are called once on freshly allocated pdevs, but for symmetry with how `platform_device_set_fwnode()` works post-patch-23, consider implementing this as:
```c
platform_device_set_of_node(pdev, dev2->of_node);
pdev->dev.of_node_reused = true;
```
This would route through the helper chain and handle put/get correctly.

### PATCHES 9-22: Driver conversions

These are all mechanical conversions from direct `pdev->dev.of_node = ...` or `pdev->dev.fwnode = ...` assignments to using the new helpers. Each one is straightforward:

- **Patch 9** (of/platform): Replaces `device_set_node(&dev->dev, of_fwnode_handle(of_node_get(np)))` with `platform_device_set_of_node(dev, np)`. Clean simplification.

- **Patch 10** (powerpc/powermac): Converts the fix from patch 5 to use the helper. Removes the explicit `of_node_get()` since the helper does it internally.

- **Patch 11** (i2c/pxa-pci): Straightforward conversion.

- **Patch 12** (iommu/fsl): This one is interesting — in addition to the conversion, it restructures the error handling:
```c
-	pdev->dev.of_node = of_node_get(np);
+	platform_device_set_of_node(pdev, np);
 ...
+	of_node_put(np);
 	return 0;
 
 error_device_add:
-	of_node_put(pdev->dev.of_node);
-	pdev->dev.of_node = NULL;
-
 	platform_device_put(pdev);
```
The error path simplification is correct: `platform_device_put()` → `platform_device_release()` → `of_node_put()` handles the pdev's reference. The new `of_node_put(np)` on the success path drops the original search reference. The old error path code (`of_node_put` + NULL assignment) was being done manually because the caller wanted to avoid the double-put in `platform_device_release()`, but with the helper taking its own ref, `platform_device_put()` cleanly handles it. **However**, this patch is doing a subtle bug fix (adding the missing `of_node_put(np)` on the success path) mixed in with the conversion. It might be worth splitting or at least noting in the commit message.

- **Patch 13** (net/bcmgenet): Nice use of `__free(device_node)` cleanup attribute:
```c
+		struct device_node *np __free(device_node) = bcmgenet_mii_of_find_mdio(priv);
+		platform_device_set_of_node(ppdev, np);
```
The `__free(device_node)` ensures the search reference is dropped when `np` goes out of scope, while the helper takes its own ref. Clean.

- **Patches 14-17**: Straightforward conversions of the fixup patches (1-5 sites) to use helpers.

- **Patch 18** (drm/xe/i2c): The DRM-relevant patch. Trivial conversion:
```c
-	pdev->dev.fwnode = fwnode;
+	platform_device_set_fwnode(pdev, fwnode);
```
Since `fwnode` here is a software node (not an OF node), in the pre-patch-23 world this is a no-op behavioral change (the `else` branch of `platform_device_set_fwnode` just does `pdev->dev.fwnode = fwnode`). After patch 23, it will correctly take a fwnode reference. **No issues for the DRM subsystem**.

- **Patch 19** (platform/surface): Same pattern as patch 18 with non-OF fwnode.

- **Patches 20-22**: Convert `device_set_of_node_from_dev()` calls to `platform_device_set_of_node_from_dev()`. The key addition is that fwnode is now synced alongside of_node.

---
Generated by Claude Code Patch Reviewer