Claude review: Introduce Xe Uncorrectable Error Handling

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

Patch Review

*Marked "DO NOT REVIEW. COMPILATION ONLY" — flagging issues for completeness since this code does compile and execute.*

**BUG: Uninitialized variable in `handle_page_offline()`**

```c
static int handle_page_offline(struct xe_device *xe, u64 page_address, bool send_offline_cmd)
{
	enum xe_ras_page_action action;
	int ret;
	...
	/*
	 * TODO: Call function to handle address fault
	 * ret = xe_ttm_vram_handle_addr_fault(xe, page_address);
	 */

	switch (ret) {
```

`ret` is never assigned before the `switch`. The TODO function call is commented out, but the code that consumes its return value is live. This is undefined behavior — whatever happens to be on the stack determines the control flow. The compiler should warn about this.

**Wedge API bypass in `xe_pci_error_handling()`**

```c
static void xe_pci_error_handling(struct pci_dev *pdev)
{
	...
	xe_device_set_in_reset(xe);
	atomic_set(&xe->wedged.flag, 1);

	for_each_gt(gt, xe, id)
		xe_gt_declare_wedged(gt);
```

This bypasses `xe_device_declare_wedged()` and directly sets the flag. `xe_gt_declare_wedged()` asserts `xe_gt_assert(gt, gt_to_xe(gt)->wedged.mode)` at `xe_gt.c:1190`. If `wedged.mode` is `XE_WEDGED_MODE_NEVER` (0), this assertion fires. The normal `xe_device_declare_wedged()` path checks for this and returns early. This error-handling path should respect that check or explicitly set `wedged.mode` before calling `xe_gt_declare_wedged()`.

**Duplicate macro definition**

`XE_SYSCTRL_FLOOD` is defined as `16` in both `xe_sysctrl_event_types.h` and `xe_sysctrl_mailbox_types.h`. It should be in one place only — likely `xe_sysctrl_mailbox_types.h` since that's where the mailbox infrastructure lives, and `xe_sysctrl_event.c` can include that header.

**Missing header declaration**

```c
extern const struct pci_error_handlers xe_pci_error_handlers;
```

This `extern` in `xe_pci.c` should be in a header (e.g., `xe_pci.h` or a new `xe_pci_error.h`) so the definition in `xe_pci_error.c` and the usage in `xe_pci.c` share one declaration. A bare `extern` in a `.c` file is fragile — if the type ever changes, the compiler can't catch the mismatch.

**`xe_pci_error_slot_reset()` — unchecked `pci_match_id()` return**

```c
const struct pci_device_id *ent = pci_match_id(pdev->driver->id_table, pdev);
```

`pci_match_id()` can return NULL. If it does, `pdev->driver->probe(pdev, ent)` passes NULL as the `ent` argument, which will likely crash. While this shouldn't happen during normal error recovery (the device already matched), defensive code should check.

---

---
Generated by Claude Code Patch Reviewer