Re: [PATCH v2 3/3] sched/fair: add lsub_positive and use it consistently

[Ingo Molnar]

* Patrick Bellasi  wrote:

> The following pattern:
> 
>var -= min_t(typeof(var), var, val);
> 
> is used multiple times in fair.c.
> 
> The existing sub_positive() already capture that pattern but it adds
> also explicit load-sotre to properly support lockless observations.
> In other cases, the patter above is used to update local, and/or not
> concurrently accessed, variables.

> Let's add a simpler version of sub_positive, targeted to local variables
> updates, which gives the same readability benefits at calling sites
> without enforcing {READ,WRITE}_ONCE barriers.

> +/*
> + * Remove and clamp on negative, from a local variable.
> + *
> + * A variant of sub_positive which do not use explicit load-store
> + * and thus optimized for local variable updates.

I fixed up the two typos ('load-sotre', 'patter'), and fixed eight 
grammar mistakes (!) in the changelog and in the code comments, but 
*please* read the changelogs and code you are writing, this is scheduler 
code after all ...

( Please also use the fn() notation in changelogs consistently in the 
  future: first you use 'sub_positive()' correctly then it becomes 
  'sub_positive'. )

Anyway, the fix looks sane so I've applied it to sched/urgent.

Thanks,

Ingo

[PATCH v2 3/3] sched/fair: add lsub_positive and use it consistently

[Patrick Bellasi]

The following pattern:

   var -= min_t(typeof(var), var, val);

is used multiple times in fair.c.

The existing sub_positive() already capture that pattern but it adds
also explicit load-sotre to properly support lockless observations.
In other cases, the patter above is used to update local, and/or not
concurrently accessed, variables.

Let's add a simpler version of sub_positive, targeted to local variables
updates, which gives the same readability benefits at calling sites
without enforcing {READ,WRITE}_ONCE barriers.

Signed-off-by: Patrick Bellasi 
Link: 
https://lore.kernel.org/lkml/20181031184527.ga3...@hirez.programming.kicks-ass.net
Cc: Ingo Molnar 
Cc: Peter Zijlstra 
Cc: linux-kernel@vger.kernel.org
---
 kernel/sched/fair.c | 24 +---
 1 file changed, 17 insertions(+), 7 deletions(-)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index aeb37fe4dbb1..d50c739127d6 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -2734,6 +2734,17 @@ account_entity_dequeue(struct cfs_rq *cfs_rq, struct 
sched_entity *se)
WRITE_ONCE(*ptr, res);  \
 } while (0)
 
+/*
+ * Remove and clamp on negative, from a local variable.
+ *
+ * A variant of sub_positive which do not use explicit load-store
+ * and thus optimized for local variable updates.
+ */
+#define lsub_positive(_ptr, _val) do { \
+   typeof(_ptr) ptr = (_ptr);  \
+   *ptr -= min_t(typeof(*ptr), *ptr, _val);\
+} while (0)
+
 #ifdef CONFIG_SMP
 static inline void
 enqueue_runnable_load_avg(struct cfs_rq *cfs_rq, struct sched_entity *se)
@@ -4639,7 +4650,7 @@ static int do_sched_cfs_period_timer(struct cfs_bandwidth 
*cfs_b, int overrun)
cfs_b->distribute_running = 0;
throttled = !list_empty(&cfs_b->throttled_cfs_rq);
 
-   cfs_b->runtime -= min(runtime, cfs_b->runtime);
+   lsub_positive(&cfs_b->runtime, runtime);
}
 
/*
@@ -4773,7 +4784,7 @@ static void do_sched_cfs_slack_timer(struct cfs_bandwidth 
*cfs_b)
 
raw_spin_lock(&cfs_b->lock);
if (expires == cfs_b->runtime_expires)
-   cfs_b->runtime -= min(runtime, cfs_b->runtime);
+   lsub_positive(&cfs_b->runtime, runtime);
cfs_b->distribute_running = 0;
raw_spin_unlock(&cfs_b->lock);
 }
@@ -6240,7 +6251,7 @@ static unsigned long cpu_util_without(int cpu, struct 
task_struct *p)
util = READ_ONCE(cfs_rq->avg.util_avg);
 
/* Discount task's util from CPU's util */
-   util -= min_t(unsigned int, util, task_util(p));
+   lsub_positive(&util, task_util(p));
 
/*
 * Covered cases:
@@ -6289,10 +6300,9 @@ static unsigned long cpu_util_without(int cpu, struct 
task_struct *p)
 * properly fix the execl regression and it helps in further
 * reducing the chances for the above race.
 */
-   if (unlikely(task_on_rq_queued(p) || current == p)) {
-   estimated -= min_t(unsigned int, estimated,
-  _task_util_est(p));
-   }
+   if (unlikely(task_on_rq_queued(p) || current == p))
+   lsub_positive(&estimated, _task_util_est(p));
+
util = max(util, estimated);
}
 
-- 
2.18.0