Claude review: drm/pagemap: Unlock and put folios when possible

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

Patch Review

**Good:** Clean conversion from page-level to folio-level unlock/put. The loop change from `++i` to stepping by `NR_PAGES(order)` correctly skips tail pages of large folios.

**Concern — uninitialized `order` on zero migrate_pfn path:**

```c
for (i = 0; i < npages;) {
    struct page *page;
    struct folio *folio;
    unsigned int order = 0;

    if (!migrate_pfn[i])
        goto next;
    ...
next:
    i += NR_PAGES(order);
}
```

When `migrate_pfn[i]` is zero, `order` is 0 and `NR_PAGES(0)` is 1, so the loop advances by 1. This is correct. However, it means for null entries interleaved within a large folio's PFN slots, each null entry is iterated individually. This is fine for correctness — the null entries simply indicate pages that weren't migrated — but it's worth noting the semantics are consistent.

**Minor:** The `migrate_pfn[i] = 0` clear on line 273 only clears the head page's entry, not tail entries. This appears correct since `drm_pagemap_migration_unlock_put_pages` is cleaning up after errors and only the head PFN would have been set during setup.

Looks good.

---
Generated by Claude Code Patch Reviewer