summaryrefslogtreecommitdiffstats
path: root/kernel
diff options
context:
space:
mode:
authorMel Gorman <mgorman@techsingularity.net>2020-01-28 15:40:06 +0000
committerIngo Molnar <mingo@kernel.org>2020-02-10 11:24:37 +0100
commit52262ee567ad14c9606be25f3caddcefa3c514e4 (patch)
treeaaf21f3202c57034932f0e09e18bd611369a62a4 /kernel
parent2a4b03ffc69f2dedc6388e9a6438b5f4c133a40d (diff)
downloadlinux-52262ee567ad14c9606be25f3caddcefa3c514e4.tar.bz2
sched/fair: Allow a per-CPU kthread waking a task to stack on the same CPU, to fix XFS performance regression
The following XFS commit: 8ab39f11d974 ("xfs: prevent CIL push holdoff in log recovery") changed the logic from using bound workqueues to using unbound workqueues. Functionally this makes sense but it was observed at the time that the dbench performance dropped quite a lot and CPU migrations were increased. The current pattern of the task migration is straight-forward. With XFS, an IO issuer delegates work to xlog_cil_push_work ()on an unbound kworker. This runs on a nearby CPU and on completion, dbench wakes up on its old CPU as it is still idle and no migration occurs. dbench then queues the real IO on the blk_mq_requeue_work() work item which runs on a bound kworker which is forced to run on the same CPU as dbench. When IO completes, the bound kworker wakes dbench but as the kworker is a bound but, real task, the CPU is not considered idle and dbench gets migrated by select_idle_sibling() to a new CPU. dbench may ping-pong between two CPUs for a while but ultimately it starts a round-robin of all CPUs sharing the same LLC. High-frequency migration on each IO completion has poor performance overall. It has negative implications both in commication costs and power management. mpstat confirmed that at low thread counts that all CPUs sharing an LLC has low level of activity. Note that even if the CIL patch was reverted, there still would be migrations but the impact is less noticeable. It turns out that individually the scheduler, XFS, blk-mq and workqueues all made sensible decisions but in combination, the overall effect was sub-optimal. This patch special cases the IO issue/completion pattern and allows a bound kworker waker and a task wakee to stack on the same CPU if there is a strong chance they are directly related. The expectation is that the kworker is likely going back to sleep shortly. This is not guaranteed as the IO could be queued asynchronously but there is a very strong relationship between the task and kworker in this case that would justify stacking on the same CPU instead of migrating. There should be few concerns about kworker starvation given that the special casing is only when the kworker is the waker. DBench on XFS MMTests config: io-dbench4-async modified to run on a fresh XFS filesystem UMA machine with 8 cores sharing LLC 5.5.0-rc7 5.5.0-rc7 tipsched-20200124 kworkerstack Amean 1 22.63 ( 0.00%) 20.54 * 9.23%* Amean 2 25.56 ( 0.00%) 23.40 * 8.44%* Amean 4 28.63 ( 0.00%) 27.85 * 2.70%* Amean 8 37.66 ( 0.00%) 37.68 ( -0.05%) Amean 64 469.47 ( 0.00%) 468.26 ( 0.26%) Stddev 1 1.00 ( 0.00%) 0.72 ( 28.12%) Stddev 2 1.62 ( 0.00%) 1.97 ( -21.54%) Stddev 4 2.53 ( 0.00%) 3.58 ( -41.19%) Stddev 8 5.30 ( 0.00%) 5.20 ( 1.92%) Stddev 64 86.36 ( 0.00%) 94.53 ( -9.46%) NUMA machine, 48 CPUs total, 24 CPUs share cache 5.5.0-rc7 5.5.0-rc7 tipsched-20200124 kworkerstack-v1r2 Amean 1 58.69 ( 0.00%) 30.21 * 48.53%* Amean 2 60.90 ( 0.00%) 35.29 * 42.05%* Amean 4 66.77 ( 0.00%) 46.55 * 30.28%* Amean 8 81.41 ( 0.00%) 68.46 * 15.91%* Amean 16 113.29 ( 0.00%) 107.79 * 4.85%* Amean 32 199.10 ( 0.00%) 198.22 * 0.44%* Amean 64 478.99 ( 0.00%) 477.06 * 0.40%* Amean 128 1345.26 ( 0.00%) 1372.64 * -2.04%* Stddev 1 2.64 ( 0.00%) 4.17 ( -58.08%) Stddev 2 4.35 ( 0.00%) 5.38 ( -23.73%) Stddev 4 6.77 ( 0.00%) 6.56 ( 3.00%) Stddev 8 11.61 ( 0.00%) 10.91 ( 6.04%) Stddev 16 18.63 ( 0.00%) 19.19 ( -3.01%) Stddev 32 38.71 ( 0.00%) 38.30 ( 1.06%) Stddev 64 100.28 ( 0.00%) 91.24 ( 9.02%) Stddev 128 186.87 ( 0.00%) 160.34 ( 14.20%) Dbench has been modified to report the time to complete a single "load file". This is a more meaningful metric for dbench that a throughput metric as the benchmark makes many different system calls that are not throughput-related Patch shows a 9.23% and 48.53% reduction in the time to process a load file with the difference partially explained by the number of CPUs sharing a LLC. In a separate run, task migrations were almost eliminated by the patch for low client counts. In case people have issue with the metric used for the benchmark, this is a comparison of the throughputs as reported by dbench on the NUMA machine. dbench4 Throughput (misleading but traditional) 5.5.0-rc7 5.5.0-rc7 tipsched-20200124 kworkerstack-v1r2 Hmean 1 321.41 ( 0.00%) 617.82 * 92.22%* Hmean 2 622.87 ( 0.00%) 1066.80 * 71.27%* Hmean 4 1134.56 ( 0.00%) 1623.74 * 43.12%* Hmean 8 1869.96 ( 0.00%) 2212.67 * 18.33%* Hmean 16 2673.11 ( 0.00%) 2806.13 * 4.98%* Hmean 32 3032.74 ( 0.00%) 3039.54 ( 0.22%) Hmean 64 2514.25 ( 0.00%) 2498.96 * -0.61%* Hmean 128 1778.49 ( 0.00%) 1746.05 * -1.82%* Note that this is somewhat specific to XFS and ext4 shows no performance difference as it does not rely on kworkers in the same way. No major problem was observed running other workloads on different machines although not all tests have completed yet. Signed-off-by: Mel Gorman <mgorman@techsingularity.net> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Link: https://lkml.kernel.org/r/20200128154006.GD3466@techsingularity.net Signed-off-by: Ingo Molnar <mingo@kernel.org>
Diffstat (limited to 'kernel')
-rw-r--r--kernel/sched/core.c11
-rw-r--r--kernel/sched/fair.c14
-rw-r--r--kernel/sched/sched.h13
3 files changed, 27 insertions, 11 deletions
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 89e54f3ed571..1a9983da4408 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -1447,17 +1447,6 @@ void check_preempt_curr(struct rq *rq, struct task_struct *p, int flags)
#ifdef CONFIG_SMP
-static inline bool is_per_cpu_kthread(struct task_struct *p)
-{
- if (!(p->flags & PF_KTHREAD))
- return false;
-
- if (p->nr_cpus_allowed != 1)
- return false;
-
- return true;
-}
-
/*
* Per-CPU kthreads are allowed to run on !active && online CPUs, see
* __set_cpus_allowed_ptr() and select_fallback_rq().
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 25dffc03f0f6..94c3b8469cf6 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -5912,6 +5912,20 @@ static int select_idle_sibling(struct task_struct *p, int prev, int target)
(available_idle_cpu(prev) || sched_idle_cpu(prev)))
return prev;
+ /*
+ * Allow a per-cpu kthread to stack with the wakee if the
+ * kworker thread and the tasks previous CPUs are the same.
+ * The assumption is that the wakee queued work for the
+ * per-cpu kthread that is now complete and the wakeup is
+ * essentially a sync wakeup. An obvious example of this
+ * pattern is IO completions.
+ */
+ if (is_per_cpu_kthread(current) &&
+ prev == smp_processor_id() &&
+ this_rq()->nr_running <= 1) {
+ return prev;
+ }
+
/* Check a recently used CPU as a potential idle candidate: */
recent_used_cpu = p->recent_used_cpu;
if (recent_used_cpu != prev &&
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index 1a88dc8ad11b..5876e6ba5903 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -2479,3 +2479,16 @@ static inline void membarrier_switch_mm(struct rq *rq,
{
}
#endif
+
+#ifdef CONFIG_SMP
+static inline bool is_per_cpu_kthread(struct task_struct *p)
+{
+ if (!(p->flags & PF_KTHREAD))
+ return false;
+
+ if (p->nr_cpus_allowed != 1)
+ return false;
+
+ return true;
+}
+#endif