Re: [RFC PATCH bpf-next 1/3] sched: add helpers for numa balancing

[Yafang Shao <laoar.shao@gmail.com>]

On Tue, Jan 13, 2026 at 8:42 PM <bot+bpf-ci@kernel.org> wrote:
>
> > diff --git a/include/linux/sched/numa_balancing.h b/include/linux/sched/numa_balancing.h
> > index 52b22c5c396d..792b6665f476 100644
> > --- a/include/linux/sched/numa_balancing.h
> > +++ b/include/linux/sched/numa_balancing.h
>
> [ ... ]
>
> > @@ -32,6 +33,28 @@ extern void set_numabalancing_state(bool enabled);
> > extern void task_numa_free(struct task_struct *p, bool final);
> > bool should_numa_migrate_memory(struct task_struct *p, struct folio *folio,
> >                               int src_nid, int dst_cpu);
> > +
> > +extern struct static_key_false sched_numa_balancing;
> > +static inline bool task_numab_enabled(struct task_struct *p)
> > +{
> > +     if (static_branch_unlikely(&sched_numa_balancing))
> > +             return true;
> > +     return false;
> > +}
> > +
> > +static inline bool task_numab_mode_normal(void)
> > +{
> > +     if (sysctl_numa_balancing_mode & NUMA_BALANCING_NORMAL)
> > +             return true;
> > +     return false;
> > +}
> > +
> > +static inline bool task_numab_mode_tiering(void)
> > +{
> > +     if (sysctl_numa_balancing_mode & NUMA_BALANCING_MEMORY_TIERING)
> > +             return true;
> > +     return false;
> > +}
> >  #else
> >  static inline void task_numa_fault(int last_node, int node, int pages,
> >                                  int flags)
>
> [ ... ]
>
> > @@ -52,6 +75,10 @@ static inline bool should_numa_migrate_memory(struct task_struct *p,
> >  {
> >       return true;
> >  }
> > +static inline bool task_numab_enabled(struct task_struct *p)
> > +{
> > +     return false;
> > +}
> >  #endif
>
> The #else block adds a stub for task_numab_enabled(), but task_numab_mode_normal()
> and task_numab_mode_tiering() are missing from this section. These helpers are
> defined when CONFIG_NUMA_BALANCING is enabled, but have no stubs when it is
> disabled.
>
> > diff --git a/mm/vmscan.c b/mm/vmscan.c
> > index 670fe9fae5ba..7ee5695326e3 100644
> > --- a/mm/vmscan.c
> > +++ b/mm/vmscan.c
>
> [ ... ]
>
> > @@ -4843,9 +4844,7 @@ static bool should_abort_scan(struct lruvec *lruvec, struct scan_control *sc)
> >       if (!current_is_kswapd() || sc->order)
> >               return false;
> >
> > -     mark = sysctl_numa_balancing_mode & NUMA_BALANCING_MEMORY_TIERING ?
> > -            WMARK_PROMO : WMARK_HIGH;
> > -
> > +     mark = task_numab_mode_tiering() ? WMARK_PROMO : WMARK_HIGH;
>
> [ ... ]
>
> > @@ -6774,7 +6773,7 @@ static bool pgdat_balanced(pg_data_t *pgdat, int order, int highest_zoneidx)
> >               enum zone_stat_item item;
> >               unsigned long free_pages;
> >
> > -             if (sysctl_numa_balancing_mode & NUMA_BALANCING_MEMORY_TIERING)
> > +             if (task_numab_mode_tiering())
> >                       mark = promo_wmark_pages(zone);
>
> The original sysctl_numa_balancing_mode is defined to 0 when CONFIG_NUMA_BALANCING
> is disabled (see include/linux/sched/sysctl.h), allowing this code to compile and
> work correctly in both configurations. The new task_numab_mode_tiering() helper
> lacks a stub in the #else block, so should_abort_scan() and pgdat_balanced() in
> mm/vmscan.c will fail to compile when CONFIG_NUMA_BALANCING is not enabled.
>
> Would it make sense to add stubs for task_numab_mode_normal() and
> task_numab_mode_tiering() to the #else block, returning false to match the
> original behavior when sysctl_numa_balancing_mode is defined as 0?

Right.
I missed it.  Thanks for the review.

-- 
Regards
Yafang