Claude review: Avoid oops on illegal VGA register access

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

Overall Series Review

Subject: Avoid oops on illegal VGA register access
Author: Simon Richter <Simon.Richter@hogyros.de>
Patches: 2
Reviewed: 2026-02-18T06:51:00.223518

---

This is a single-patch series that adds error handling for `vga_get_uninterruptible()` in the i915 VGA code. The stated motivation is to avoid oops on non-x86 platforms where VGA registers cannot be mapped, which would cause `vga_get()` to return `-ENODEV`. The patch modifies `intel_vga_get()` to propagate error codes, and `intel_vga_disable()` to handle them by skipping VGA register access and jumping to the MMIO-based VGA plane disable.

The patch is based on a tree that contains a refactoring of `intel_vga.c` (adding `intel_vga_get()`, `intel_vga_put()`, `intel_pci_set_io_decode()`, etc.) that has not yet landed in drm-next. The base commit `15658979e64a` is not present in drm-next. This means the patch cannot be applied on its own — it depends on a prerequisite refactoring series. The cover letter acknowledges this indirectly, mentioning that "complementary changes in vgaarb" are also needed, but the dependency on the i915 VGA refactoring series is not stated.

There are a few coding style issues and one incomplete aspect: the error handling is only added for `intel_vga_get()` within `intel_vga_disable()`, but if the underlying problem is that `vga_get_uninterruptible()` can fail on non-x86, this should also be considered for `intel_vga_reset_io_mem()` which similarly calls `vga_get_uninterruptible()` without checking the return value.

---
Generated by Claude Code Patch Reviewer