Re: [PATCH 1/3] memcg: Add memcg_stat_mod()

[Johannes Weiner <hannes@cmpxchg.org>]

On Wed, Feb 25, 2026 at 04:22:15PM +0000, Matthew Wilcox (Oracle) wrote:
> This function lets the caller find the memcg somewhere other than
> page->memcg_data.
> 
> Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>

> @@ -787,24 +787,27 @@ void mod_lruvec_state(struct lruvec *lruvec, enum node_stat_item idx,
>  		mod_memcg_lruvec_state(lruvec, idx, val);
>  }
>  
> +void memcg_stat_mod(struct mem_cgroup *memcg, pg_data_t *pgdat,
> +		enum node_stat_item idx, long val)
> +{
> +	/* Untracked pages have no memcg, no lruvec. Update only the node */
> +	if (!memcg) {
> +		mod_node_page_state(pgdat, idx, val);
> +	} else {
> +		struct lruvec *lruvec = mem_cgroup_lruvec(memcg, pgdat);
> +		mod_lruvec_state(lruvec, idx, val);
> +	}
> +}

The refactor (and the one in the next patch) looks good to me.

But we already have a mod_memcg_state(), which genuinely just updates
the memcg counters, and memcg_stat_mod() makes it a bit non-obvious
that this is a "core" stat accounting function (that happens to do
memcg when compiled in).

Can we go with this instead?

	void mod_node_memcg_state(pg_data_t *, struct mem_cgroup *, ...)