Claude review: cgroup/dmem: return error when failing to set dmem.max

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

Patch Review

**Correctness of the core fix:** The key change at line 121 of the diff is correct:

```c
-	page_counter_set_max(&pool->cnt, val);
+	return page_counter_set_max(&pool->cnt, val);
```

`page_counter_set_max()` returns `int` (confirmed at `include/linux/page_counter.h:86`), and can return `-EBUSY`. Propagating this is the right thing to do.

**Signature change is appropriate:** Changing the `apply` callback from `void (*apply)(...)` to `int (*apply)(...)` and updating `set_resource_min`/`set_resource_low` to return `0` is a clean way to handle this uniformly.

**Potential issue — partial application on multi-region writes:** Looking at the loop in `dmemcg_limit_write()` (line 734 of the existing code: `while (buf && !err)`), when a user writes multiple region limits in a single write (multiple lines), an error from `apply()` on one region will stop processing subsequent regions. The commit message acknowledges this ("Return as soon as setting one of the regions max limit fails"). However, consider that **regions processed before the failing one have already been committed** — the function does not roll back previously applied limits. This is a pre-existing pattern (same happens for parse errors), but worth noting that the semantics are "apply until first failure" with no rollback.

**Minor note — `err` after `apply` vs `goto out_put`:** Looking at the patched code:

```c
		err = apply(pool, new_limit);
		dmemcg_pool_put(pool);

out_put:
		kref_put(&region->ref, dmemcg_free_region);
```

When `apply()` returns an error, the code still correctly calls `dmemcg_pool_put(pool)` and `kref_put()` before the `while` loop condition `!err` terminates the loop. This is correct — no resource leak.

**Nit — `set_resource_min` and `set_resource_low`:** `page_counter_set_min()` and `page_counter_set_low()` are `void` functions, so wrapping them with `return 0` is fine. If these ever gain error returns in the future, this patch has already set up the plumbing to handle it. No issue here.

**Verdict:** The patch looks correct and ready to apply. The fix is minimal and appropriate. 

Reviewed-by quality: **Good to go**, no blocking issues.

---
Generated by Claude Code Patch Reviewer