Re: [PATCH 7/8] mm: vmscan: split shrink_node() into node part and memcgs part

[Johannes Weiner]

On Tue, Oct 22, 2019 at 08:08:23PM +0000, Roman Gushchin wrote:
> On Tue, Oct 22, 2019 at 10:48:02AM -0400, Johannes Weiner wrote:
> > This function is getting long and unwieldy, split out the memcg bits.
> > 
> > The updated shrink_node() handles the generic (node) reclaim aspects:
> >   - global vmpressure notifications
> >   - writeback and congestion throttling
> >   - reclaim/compaction management
> >   - kswapd giving up on unreclaimable nodes
> > 
> > It then calls a new shrink_node_memcgs() which handles cgroup specifics:
> >   - the cgroup tree traversal
> >   - memory.low considerations
> >   - per-cgroup slab shrinking callbacks
> >   - per-cgroup vmpressure notifications
> > 
> > Signed-off-by: Johannes Weiner <hannes@xxxxxxxxxxx>
> > ---
> >  mm/vmscan.c | 28 ++++++++++++++++++----------
> >  1 file changed, 18 insertions(+), 10 deletions(-)
> > 
> > diff --git a/mm/vmscan.c b/mm/vmscan.c
> > index db073b40c432..65baa89740dd 100644
> > --- a/mm/vmscan.c
> > +++ b/mm/vmscan.c
> > @@ -2722,18 +2722,10 @@ static bool pgdat_memcg_congested(pg_data_t *pgdat, struct mem_cgroup *memcg)
> >  		(memcg && memcg_congested(pgdat, memcg));
> >  }
> >  
> > -static bool shrink_node(pg_data_t *pgdat, struct scan_control *sc)
> > +static void shrink_node_memcgs(pg_data_t *pgdat, struct scan_control *sc)
> >  {
> > -	struct reclaim_state *reclaim_state = current->reclaim_state;
> >  	struct mem_cgroup *root = sc->target_mem_cgroup;
> > -	unsigned long nr_reclaimed, nr_scanned;
> > -	bool reclaimable = false;
> >  	struct mem_cgroup *memcg;
> > -again:
> > -	memset(&sc->nr, 0, sizeof(sc->nr));
> > -
> > -	nr_reclaimed = sc->nr_reclaimed;
> > -	nr_scanned = sc->nr_scanned;
> >  
> >  	memcg = mem_cgroup_iter(root, NULL, NULL);
> >  	do {
> > @@ -2786,6 +2778,22 @@ static bool shrink_node(pg_data_t *pgdat, struct scan_control *sc)
> >  			   sc->nr_reclaimed - reclaimed);
> >  
> >  	} while ((memcg = mem_cgroup_iter(root, memcg, NULL)));
> > +}
> > +
> > +static bool shrink_node(pg_data_t *pgdat, struct scan_control *sc)
> > +{
> > +	struct reclaim_state *reclaim_state = current->reclaim_state;
> > +	struct mem_cgroup *root = sc->target_mem_cgroup;
> > +	unsigned long nr_reclaimed, nr_scanned;
> > +	bool reclaimable = false;
> > +
> > +again:
> > +	memset(&sc->nr, 0, sizeof(sc->nr));
> > +
> > +	nr_reclaimed = sc->nr_reclaimed;
> > +	nr_scanned = sc->nr_scanned;
> > +
> > +	shrink_node_memcgs(pgdat, sc);
> >  
> >  	if (reclaim_state) {
> >  		sc->nr_reclaimed += reclaim_state->reclaimed_slab;
> > @@ -2793,7 +2801,7 @@ static bool shrink_node(pg_data_t *pgdat, struct scan_control *sc)
> >  	}
> >  
> >  	/* Record the subtree's reclaim efficiency */
> > -	vmpressure(sc->gfp_mask, sc->target_mem_cgroup, true,
> > +	vmpressure(sc->gfp_mask, root, true,
> 
> Maybe target? Or target_memcg? The word root is associated with the root cgroup.
>
> Other than root the patch looks good to me:
> 
> Reviewed-by: Roman Gushchin <guro@xxxxxx>

Thanks!

I agree, target_memcg is better than root. The next patch also
replaces some of these with target_lruvec.

This on top?