sched: reset loop counters if all tasks are pinned and we need to redo load balance
PATCH 1/1 sched: reset loop counters if all tasks are pinned and we need to redo load balance
While load balancing, if all tasks on the source runqueue are pinned, we retry after excluding the corresponding source cpu. However, loop counters env.loop and env.loop_break are not reset before retrying, which can lead to failure in moving the tasks. In this patch we reset env.loop and env.loop_break to their inital values before we retry.
---
kernel/sched/fair.c +4 -1
@@ -4520,8 +4520,11 @@ more_balance:/* All tasks on this runqueue were pinned by CPU affinity */ if (unlikely (env.flags &LBF_ALL_PINNED )) {cpumask_clear_cpu (cpu_of (busiest), cpus); -if (!cpumask_empty (cpus)) +if (!cpumask_empty (cpus)) { + env.loop =0 ; + env.loop_break = sched_nr_migrate_break; goto redo; + } goto out_balanced; } }
[tip:sched/core] sched: Reset loop counters if all tasks are pinned and we need to redo load balance
Commit-ID: 4d44f7090f8233f6994e7397679a32c81187506d Gitweb: http://git.kernel.org/tip/4d44f7090f8233f6994e7397679a32c81187506d Author: Prashanth Nageshappa <[email protected]> AuthorDate: Tue, 19 Jun 2012 17:52:07 +0530 Committer: Ingo Molnar <[email protected]> CommitDate: Thu, 5 Jul 2012 21:09:10 +0200
sched: Reset loop counters if all tasks are pinned and we need to redo load balance
While load balancing, if all tasks on the source runqueue are pinned, we retry after excluding the corresponding source cpu. However, loop counters env.loop and env.loop_break are not reset before retrying, which can lead to failure in moving the tasks. In this patch we reset env.loop and env.loop_break to their inital values before we retry.
--- kernel/sched/fair.c | 5 ++++- 1 files changed, 4 insertions(+), 1 deletions(-)
kernel/sched/fair.c +4 -1
@@ -4497,8 +4497,11 @@ more_balance:/* All tasks on this runqueue were pinned by CPU affinity */ if (unlikely (env.flags &LBF_ALL_PINNED )) {cpumask_clear_cpu (cpu_of (busiest), cpus); -if (!cpumask_empty (cpus)) +if (!cpumask_empty (cpus)) { + env.loop =0 ; + env.loop_break = sched_nr_migrate_break; goto redo; + } goto out_balanced; } }
[tip:sched/core] sched: Reset loop counters if all tasks are pinned and we need to redo load balance
In reply to Prashanth NageshappaCommit-ID: bbf18b19495942cc730e8ff11fc3ffadf20cbfe1 Gitweb: http://git.kernel.org/tip/bbf18b19495942cc730e8ff11fc3ffadf20cbfe1 Author: Prashanth Nageshappa <[email protected]> AuthorDate: Tue, 19 Jun 2012 17:52:07 +0530 Committer: Ingo Molnar <[email protected]> CommitDate: Tue, 24 Jul 2012 13:55:37 +0200
sched: Reset loop counters if all tasks are pinned and we need to redo load balance
While load balancing, if all tasks on the source runqueue are pinned, we retry after excluding the corresponding source cpu. However, loop counters env.loop and env.loop_break are not reset before retrying, which can lead to failure in moving the tasks. In this patch we reset env.loop and env.loop_break to their inital values before we retry.
--- kernel/sched/fair.c | 5 ++++- 1 files changed, 4 insertions(+), 1 deletions(-)
kernel/sched/fair.c +4 -1
@@ -4288,8 +4288,11 @@ more_balance:/* All tasks on this runqueue were pinned by CPU affinity */ if (unlikely (env.flags &LBF_ALL_PINNED )) {cpumask_clear_cpu (cpu_of (busiest), cpus); -if (!cpumask_empty (cpus)) +if (!cpumask_empty (cpus)) { + env.loop =0 ; + env.loop_break = sched_nr_migrate_break; goto redo; + } goto out_balanced; } }