Claude review: accel/ivpu: Add support for limiting NPU frequency

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

Overall Series Review

Subject: accel/ivpu: Add support for limiting NPU frequency
Author: Andrzej Kacprowski <andrzej.kacprowski@linux.intel.com>
Patches: 1
Reviewed: 2026-04-12T12:19:40.349981

---

This is a single-patch series (v3) adding configurable NPU frequency limits to the Intel IVPU accelerator driver. The feature introduces a new `freq/` sysfs subdirectory with hardware info and configurable min/max frequency attributes, a new JSM firmware message (`VPU_JSM_MSG_FREQ_CONFIG`), and legacy compatibility aliases for the existing top-level sysfs frequency attributes.

The overall design is reasonable: a `freq/` group for organized frequency attributes, firmware communication via JSM messages, and conditional exposure of writable attributes only on 50XX+ devices. The code is generally clean and follows existing driver patterns. However, there are several issues ranging from a type mismatch bug to missing access controls and cleanup concerns.

**Key issues:**
1. **Type mismatch between function signature and internal callers** (`u16` in declaration vs `u8` at call sites and `u32` in payload struct)
2. **No permission check on writable sysfs attributes** — any user can change NPU frequency
3. **Missing concurrency protection** for `cfg_min_ratio`/`cfg_max_ratio`
4. **Misleading comment** in `set_max_freq_store`
5. **`dpu_mhz_to_pll_ratio_lnl` used unconditionally** for MTL devices too
6. **Sysfs cleanup not handled** for manually-added files

---

---
Generated by Claude Code Patch Reviewer