lkml.org 
[lkml]   [2007]   [Oct]   [9]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
    Patch in this message
    /
    From
    Subject[PATCH 5/5] RT - Track which CPUs should get IPI'd on rt-overload
    Date
    The code currently blindly fires IPIs out whenever an overload occurs.
    However, there are strict events that govern when a rt-overload exists
    (e.g. RT task added to a RQ, or an RT task preempted). Therefore, we
    attempt to efficiently track which CPUs are eligible for rebalancing, and we
    only IPI those affected units.

    Signed-off-by: Gregory Haskins <ghaskins@novell.com>
    CC: Peter W. Morreale <pmorreale@novell.com>
    ---

    kernel/sched.c | 15 +++++++++++++--
    1 files changed, 13 insertions(+), 2 deletions(-)

    diff --git a/kernel/sched.c b/kernel/sched.c
    index a28ca9d..6ca5f4f 100644
    --- a/kernel/sched.c
    +++ b/kernel/sched.c
    @@ -359,6 +359,8 @@ struct rq {
    unsigned long rto_pulled;
    #endif
    struct lock_class_key rq_lock_key;
    +
    + cpumask_t rto_resched; /* Which of our peers needs rescheduling */
    };

    static DEFINE_PER_CPU_SHARED_ALIGNED(struct rq, runqueues);
    @@ -645,6 +647,9 @@ static inline void inc_rt_tasks(struct task_struct *p, struct rq *rq)
    smp_wmb();
    atomic_inc(&rt_overload);
    }
    +
    + cpus_or(rq->rto_resched, rq->rto_resched, p->cpus_allowed);
    + cpu_clear(rq->cpu, rq->rto_resched);
    # endif
    }
    #endif
    @@ -2213,9 +2218,15 @@ static inline void finish_task_switch(struct rq *rq, struct task_struct *prev)
    * If we pushed an RT task off the runqueue,
    * then kick other CPUs, they might run it:
    */
    - if (unlikely(rt_task(current) && rq->rt_nr_running > 1)) {
    + if (unlikely(rt_task(current) && prev->se.on_rq && rt_task(prev))) {
    + cpus_or(rq->rto_resched, rq->rto_resched, prev->cpus_allowed);
    + cpu_clear(rq->cpu, rq->rto_resched);
    + }
    +
    + if (unlikely(rq->rt_nr_running > 1 && !cpus_empty(rq->rto_resched))) {
    schedstat_inc(rq, rto_schedule);
    - smp_send_reschedule_allbutself();
    + smp_send_reschedule_allbutself_cpumask(rq->rto_resched);
    + cpus_clear(rq->rto_resched);
    }
    #endif
    prev_state = prev->state;
    -
    To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
    the body of a message to majordomo@vger.kernel.org
    More majordomo info at http://vger.kernel.org/majordomo-info.html
    Please read the FAQ at http://www.tux.org/lkml/

    \
     
     \ /
      Last update: 2007-10-09 16:43    [W:0.021 / U:0.508 seconds]
    ©2003-2016 Jasper Spaans. hosted at Digital OceanAdvertise on this site