Messages in this thread | | | From | Brendan Jackman <> | Subject | [PATCH v2 0/2] sched/fair: remote load updates for idle CPUs | Date | Fri, 1 Dec 2017 18:01:55 +0000 |
| |
Hi,
Here is the v2, which was just delayed by my holiday.
Until a NOHZ idle balance takes place on behalf of a CPU (which may never happen), the blocked load of of its root cfs_rq and its contributions to task group shares are updated only by that CPU. That means if a CPU goes suddenly from being busy to totally idle, its load and effect on shares may not be updated for a long time.
Schedutil works around this problem by ignoring the util of CPUs that were last updated more than a tick ago. However the stale load does impact task placement: elements that look at load and util (in particular the slow-path of select_task_rq_fair) can leave the idle CPUs un-used while other CPUs go unnecessarily overloaded. Furthermore the stale shares can impact CPU time allotment.
Two complementary solutions are proposed here: 1. When a task wakes up, if necessary an idle CPU is woken as if to perform a NOHZ idle balance, which is then aborted once the load of NOHZ idle CPUs has been updated. This solves the problem but brings with it extra CPU wakeups, which have an energy cost. 2. During newly-idle load balancing, the load of remote nohz-idle CPUs in the sched_domain is updated. When all of the idle CPUs were updated in that step, the nohz.next_update field is pushed further into the future. This field is used to determine the need for triggering the newly-added NOHZ kick. So if such newly-idle balances are happening often enough, no additional CPU wakeups are required to keep all the CPUs' loads and shares contributions updated.
Changes v1 -> v2:
- Vincent pointed out I broke !CONFIG_NO_HZ_COMMON, fixed that.
- Tood Kjos pointed out that a stats kick from CPU A can inhibit a full balance kick from CPU B. Reduced the scope for that by having CPU B convert the pending/ongoing stats kick to a proper balance by clearing the NOHZ_STATS_KICK bit in nohz_kick_needed.
Brendan Jackman (1): sched/fair: Update blocked load from newly idle balance
Vincent Guittot (1): sched: force update of blocked load of idle cpus
kernel/sched/core.c | 1 + kernel/sched/fair.c | 128 +++++++++++++++++++++++++++++++++++++++++++++------ kernel/sched/sched.h | 2 + 3 files changed, 116 insertions(+), 15 deletions(-)
-- 2.14.1
| |