Re: [PATCH 1/2] memcg: move mem_cgroup_swappiness to memcontrol.h

On Tue, Jul 14, 2026 at 9:20 AM Ridong Chen <[email protected]> wrote:
>
>
>
> On 7/13/2026 11:08 PM, Barry Song wrote:
> > On Sat, Jul 11, 2026 at 5:12 PM Ridong Chen <[email protected]> wrote:
> >>
> >> From: Ridong Chen <[email protected]>
> >>
> >> The per-memcg swappiness knob is v1-only; v2 always uses global
> >> vm_swappiness and ignores the per-cgroup field.
> >>
> >> Guard memcg->swappiness with CONFIG_MEMCG_V1, and move the helper
> >> to memcontrol.h where it belongs.
> >>
> >> No functional change for v1; v2-only kernels drop the unused field.
> >>
> >> Signed-off-by: Ridong Chen <[email protected]>
> >> Acked-by: Johannes Weiner <[email protected]>
> >
> > Reviewed-by: Barry Song <[email protected]>
> >
> > With some nits.
> >
> >> ---
> > [...]
> >>          struct mem_cgroup_per_node *nodeinfo[];
> >> @@ -365,6 +366,9 @@ enum objext_flags {
> >>
> >>   #define OBJEXTS_FLAGS_MASK (__NR_OBJEXTS_FLAGS - 1)
> >>
> >> +/* Defined in mm/vmscan.c; used by mem_cgroup_swappiness(). */
> >> +extern int vm_swappiness;
> >
> > This is a bit unusual. I'm not sure whether mm/swap.h would be
> > a more appropriate place for this.
> >
> Thank you for your reply.
>
> The vm_swappiness variable is not utilized within mm/swap.c.
> Furthermore, since memcontrol.h does not include swap.h, retaining the
> extern int vm_swappiness declaration in mm/swap.h will result in a
> compilation failure.

If this is the case, it still seems better to keep
extern int vm_swappiness in include/linux/swap.h.

Then we don't need the comment:
/* Defined in mm/vmscan.c; used by mem_cgroup_swappiness(). */

It also makes it clearer that vm_swappiness is an extern variable
belonging to the swap module, rather than the memcontrol module.

Thanks
Barry