]> git.karo-electronics.de Git - linux-beck.git/commitdiff
sched/fair: Fix overly small weight for interactive group entities
authorYuyang Du <yuyang.du@intel.com>
Tue, 13 Oct 2015 01:18:22 +0000 (09:18 +0800)
committerIngo Molnar <mingo@kernel.org>
Tue, 20 Oct 2015 08:13:34 +0000 (10:13 +0200)
Commit:

  9d89c257dfb9 ("sched/fair: Rewrite runnable load and utilization average tracking")

led to an overly small weight for interactive group entities. The bad case
can be easily reproduced when a number of CPU hogs compete for the CPUs
at the same time (thanks to Mike). This is largly because the task group's
load average tracking cross CPUs lags behind the real changes.

To fix this we accelerate the group share distribution process by using
the load.weight of the cfs_rq. This may increase the entire group's
share, but we have to do so to protect the (fragile) interactive
tasks, especially from CPU hogs.

Reported-by: Mike Galbraith <umgwanakikbuti@gmail.com>
Tested-by: Dietmar Eggemann <dietmar.eggemann@arm.com>
Tested-by: Mike Galbraith <umgwanakikbuti@gmail.com>
Signed-off-by: Yuyang Du <yuyang.du@intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Dietmar Eggemann <dietmar.eggemann@arm.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/1444699103-20272-1-git-send-email-yuyang.du@intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
kernel/sched/fair.c

index 6e2e3483b1ecff588e76103e0b7b7d673f16d2a5..bc62c5096e5470ce34f96d41f62df18fa1fd8ac5 100644 (file)
@@ -2363,7 +2363,7 @@ static inline long calc_tg_weight(struct task_group *tg, struct cfs_rq *cfs_rq)
         */
        tg_weight = atomic_long_read(&tg->load_avg);
        tg_weight -= cfs_rq->tg_load_avg_contrib;
-       tg_weight += cfs_rq_load_avg(cfs_rq);
+       tg_weight += cfs_rq->load.weight;
 
        return tg_weight;
 }
@@ -2373,7 +2373,7 @@ static long calc_cfs_shares(struct cfs_rq *cfs_rq, struct task_group *tg)
        long tg_weight, load, shares;
 
        tg_weight = calc_tg_weight(tg, cfs_rq);
-       load = cfs_rq_load_avg(cfs_rq);
+       load = cfs_rq->load.weight;
 
        shares = (tg->shares * load);
        if (tg_weight)