Claude review: Let userspace know about swapped out panthor GEM objects

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

Overall Series Review

Subject: Let userspace know about swapped out panthor GEM objects
Author: Nicolas Frattaroli <nicolas.frattaroli@collabora.com>
Patches: 6
Reviewed: 2026-04-23T09:29:11.567506

---

This is a clean, small, 2-patch series that adds "evicted" (swapped-out) memory tracking to the DRM fdinfo infrastructure and wires it up in the panthor driver. The motivation is sound: without this, userspace cannot distinguish between a BO that was never backed by pages and one that was backed but later evicted to swap. The implementation is straightforward and follows the established patterns for `RESIDENT`, `PURGEABLE`, and `ACTIVE` statuses.

**Key observations:**

1. The series depends on a not-yet-landed panthor GEM shrinker series (the patches reference `bo->backing.pages` while the current drm-next tree uses `bo->base.pages`, and `panthor_gem_evict_locked` doesn't exist yet). This is fine as a dependency chain, but should be noted.

2. The `atomic_t reclaimed_count` wrap-around-avoidance logic is clever but over-engineered for what is effectively a boolean "has ever been evicted" flag. The counter approach tries to serve dual duty — providing the count for potential future use while also acting as a boolean — and the wrap-around handling adds complexity for a scenario that is practically impossible (2^31 evictions of a single BO).

3. The fdinfo integration in patch 1 is clean and follows established patterns correctly. The `supported_status` bitmask mechanism means the "evicted" line only appears in fdinfo output when at least one object has actually been evicted, which is the right behavior.

**Overall: Series looks good. A few minor issues and design questions below.**

---

---
Generated by Claude Code Patch Reviewer