Claude review: drm/xe/xe_ras: Move xe drm_ras registration

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

Patch Review

Moves `xe_drm_ras_init()` call from `xe_hw_error_init()` into a new `xe_ras_init()` function, and reorders `xe_soc_remapper_init()` and `xe_sysctrl_init()` earlier in the probe sequence.

**Issue: Commit message has typos.**

```
gove xe drm_ras registration to RAS initialization flow and keep
gardware error initialization for processing errors reported
via irq.
```

"gove" should be "Move", "gardware" should be "hardware". These look like keyboard transposition errors and should be fixed before merge.

**Concern: Probe ordering change risk.** Moving `xe_soc_remapper_init()` and `xe_sysctrl_init()` earlier — before `xe_display_init_early()` — is a significant probe ordering change. The original code had them much later (after GT init, IRQ install, etc.). The commit message says this is because "ras init is dependent on both", which is true, but the move also changes when these subsystems initialize relative to display, IRQ, and GT init. If `xe_soc_remapper_init` or `xe_sysctrl_init` had any implicit dependencies on GT or IRQ being set up, this could break. The series should confirm that these subsystems have no such dependencies.

**Note: Error handling difference.** The old code used `drm_err` with `%pe` format specifier (ERR_PTR). The new code uses `%d`:

```c
// Old (removed from xe_hw_error.c):
drm_err(&xe->drm, "Failed to initialize XE DRM RAS (%pe)\n", ERR_PTR(ret));

// New (in xe_ras.c):
drm_err(&xe->drm, "Failed to initialize xe_drm_ras %d\n", ret);
```

Using `%pe` is the preferred kernel style for printing errno values as human-readable strings. The new code should use `%pe` for consistency.

---

---
Generated by Claude Code Patch Reviewer