Messages in this thread | | | Date | Tue, 16 Jun 2015 11:00:18 -0400 | From | Rik van Riel <> | Subject | Re: [PATCH v2 4/4] sched:Use correct nid while evaluating task weights |
| |
On 06/16/2015 07:56 AM, Srikar Dronamraju wrote: > In task_numa_migrate(), while evaluating other nodes for group > consolidation, env.dst_nid is used instead of using the iterator nid. > Using env.dst_nid would mean dist is always the same. Infact the same > dist was calculated above while evaluating the preferred node. > > Fix the above to use the iterator nid.
Good catch.
> Also the task/group weights from the src_nid should be calculated > irrespective of numa topology type.
If you look at score_nearby_nodes(), you will see that maxdist is only used when the topology is NUMA_BACKPLANE.
The source score never changes when having a directly connected or NUMA_GLUELESS_MESH type system, so the source score does not need to be recalculated unless we are actually dealing with a NUMA_BACKPLANE topology.
Looking forward to a v2 with just the first fix.
> Signed-off-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com> > --- > kernel/sched/fair.c | 5 ++--- > 1 file changed, 2 insertions(+), 3 deletions(-) > > diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c > index d1aa374..e1b3393 100644 > --- a/kernel/sched/fair.c > +++ b/kernel/sched/fair.c > @@ -1493,9 +1493,8 @@ static int task_numa_migrate(struct task_struct *p) > if (nid == env.src_nid || nid == p->numa_preferred_nid) > continue; > > - dist = node_distance(env.src_nid, env.dst_nid); > - if (sched_numa_topology_type == NUMA_BACKPLANE && > - dist != env.dist) { > + dist = node_distance(env.src_nid, nid); > + if (dist != env.dist) { > taskweight = task_weight(p, env.src_nid, dist); > groupweight = group_weight(p, env.src_nid, dist); > } >
| |