Claude review: platform/x86/intel/vsec: Switch exported helpers from pci_dev to device

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

Patch Review

**Overall: The largest patch, mostly mechanical. Has a potential safety concern.**

This is the core conversion: `struct intel_vsec_device` gains `struct device *dev` replacing `struct pci_dev *pcidev`. All consumers are updated.

**Concern — unchecked `to_pci_dev()` in downstream code after ACPI support is added:**

Several downstream consumers now do `to_pci_dev(ivdev->dev)` or `to_pci_dev(entry->ep->dev)` without first checking `dev_is_pci()`. While this is safe *today* because only PCI parents exist, patch 6 adds ACPI discovery support, and future ACPI-enumerated PMT endpoints could reach this code. Specifically:

1. In `class.c`, `intel_pmt_populate_entry()`:
   ```c
   struct pci_dev *pci_dev = to_pci_dev(ivdev->dev);
   ```
   This function uses `pci_resource_start()` which will crash on a non-PCI device.

2. In `telemetry.c`, `pmt_copy_region()`:
   ```c
   struct pci_dev *pdev = to_pci_dev(entry->ep->dev);
   ```

3. In `sdsi.c`, `sdsi_map_mbox_registers()`:
   ```c
   struct pci_dev *parent = to_pci_dev(dev);
   ```

4. In `vsec_tpmi.c`, `intel_vsec_tpmi_init()`:
   ```c
   struct pci_dev *pci_dev = to_pci_dev(vsec_dev->dev);
   ```

While the commit message says "same APIs work for PCI and ACPI parents", these unconverted sites will crash with non-PCI devices. It would be prudent to add `dev_is_pci()` guards or at minimum a comment at each `to_pci_dev()` site asserting that only PCI devices reach that path. The `intel_vsec_set_mapping()` function correctly adds such a guard:

```c
if (!dev_is_pci(vsec_dev->dev))
    return -ENODEV;
```

This pattern should be applied more consistently.

**Minor: `xe_pmt_telem_read` uses `kdev_to_xe_device()`** — this needs `kdev_to_xe_device()` to exist in the xe driver. It does (it's a standard xe helper), so this is fine.

---

---
Generated by Claude Code Patch Reviewer