Claude review: drm/nouveau: handle nouveau_hw_get_clock() return value

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

Overall Series Review

Subject: drm/nouveau: handle nouveau_hw_get_clock() return value
Author: Bulatov Mark <ms.bulatov7@gmail.com>
Patches: 1
Reviewed: 2026-05-16T10:30:37.685317

---

This is a single-patch series addressing a potential division-by-zero in the nouveau NV04 display arbitration code. The problem is real: `nouveau_hw_get_clock()` can return 0 (when `nouveau_hw_pllvals_to_clk()` gets zero M1/M2 divisors) or a negative error code (when `nouveau_hw_get_pllvals()` fails with `-ENOENT`). These values are then used as divisors throughout `nv04_calc_arb()` and `nv10_calc_arb()`, so the defensive check is warranted.

However, the fix has a significant problem: it introduces a **silent early return that leaves the caller's output parameters (`*burst` and `*lwm`) uninitialized**. The caller in `crtc.c:893` declares `arb_burst` and `arb_lwm` as bare `int` locals (line 848) without initialization, then immediately writes them to CRTC registers after the call. A silent return turns a division-by-zero into use-of-uninitialized-stack-data written to hardware registers — arguably trading one bug for another.

**Verdict: Needs revision.** The issue being fixed is real, but the fix needs to either (a) initialize the output parameters to safe defaults before returning, or (b) propagate an error code so the caller can skip the register writes.

---
Generated by Claude Code Patch Reviewer