[PATCH v3 0/3] Take the scheduling domain into account in numa balancin

[Chuyi Zhou]

This patchset tries to fix the logic of handling isolate cpus in numa balancing.

patch#1: Clean up for task_numa_migrate().

patch#2: Introduce numa_balance_mask which will be used in
task_numa_migrate() as a temporary cpumask.

patch#3: Consider src_rq->rd->span to avoid gathering CPU's load not in the
current scheduling domain in update_numa_state(), and ensure that an
isolated CPU is not selected as best_cpu.

v3 ->v2: https://lore.kernel.org/lkml/20250103065930.413464-1-zhouchuyi@xxxxxxxxxxxxx/#r
 - Use rq->rd->span in task_numa_migrate. (Peter)
 - Introduce per cpu mask as a temporary variable in task_numa_migrate.

v1 -> v2:
 - Collect Prateek's reviewed-by and add the history of
   task_numa_migrate() to commit log.

Chuyi Zhou (3):
  sched/fair: Remove unused task_numa_migrate return value
  sched/fair: Introduce per cpu numa_balance_mask
  sched/fair: Take sched_domain into account in task_numa_migrate

 kernel/sched/fair.c | 58 ++++++++++++++++++++++++++++++++-------------
 1 file changed, 41 insertions(+), 17 deletions(-)

-- 
2.20.1