[PATCH v8 0/2] Make find_later_rq() choose a closer cpu in topology

[PATCH v8 0/2] Make find_later_rq() choose a closer cpu in topology

[Byungchul Park]

Thanks to Joel, I could fix a typo and simplify code more.

-----8<-----

When cpudl_find() returns any among free_cpus, the cpu might not be
closer than others, considering sched domain. For example:

   this_cpu: 15
   free_cpus: 0, 1,..., 14 (== later_mask)
   best_cpu: 0

   topology:

   0 --+
       +--+
   1 --+  |
          +-- ... --+
   2 --+  |         |
       +--+         |
   3 --+            |

   ...             ...

   12 --+           |
        +--+        |
   13 --+  |        |
           +-- ... -+
   14 --+  |
        +--+
   15 --+

In this case, it would be best to select 14 since it's a free cpu and
closest to 15(this_cpu). However, currently the code select 0(best_cpu)
even though that's just any among free_cpus. Fix it.

Change from v7
   -. fix a trivial typo
   -. modify commit messages to explain what it does more clearly
   -. simplify code with an existing macro

Change from v6
   -. add a comment about selection of fallback_cpu incase more than one exist
   -. modify a comment explaining what we do wrt PREFER_SIBLING

Change from v5
   -. exclude two patches already picked up by peterz
      (sched/deadline: Make find_later_rq() choose a closer cpu in topology)
      (sched/deadline: Change return value of cpudl_find())
   -. apply what peterz fixed for 'prefer sibling', into deadline and rt

Change from v4
   -. remove a patch that might cause huge lock contention
      (by spin lock(&cpudl.lock) in a hot path of scheduler)

Change from v3
   -. rename closest_cpu to best_cpu so that it align with rt
   -. protect referring cpudl.elements with cpudl.lock
   -. change return value of cpudl_find() to bool

Change from v2
   -. add support for SD_PREFER_SIBLING

Change from v1
   -. clean up the patch

Byungchul Park (2):
  sched/deadline: Add support for SD_PREFER_SIBLING on find_later_rq()
  sched/rt: Add support for SD_PREFER_SIBLING on find_lowest_rq()

 kernel/sched/deadline.c | 55 +++++++++++++++++++++++++++++++++++++++++++++---
 kernel/sched/rt.c       | 56 ++++++++++++++++++++++++++++++++++++++++++++++---
 2 files changed, 105 insertions(+), 6 deletions(-)

-- 
1.9.1

[PATCH v8 1/2] sched/deadline: Add support for SD_PREFER_SIBLING on find_later_rq()

[Byungchul Park]

It would be better to try to check other siblings first if
SD_PREFER_SIBLING is flaged when pushing tasks - migration.

Signed-off-by: Byungchul Park <byungchul.park@lge.com>
---
 kernel/sched/deadline.c | 55 ++++++++++++++++++++++++++++++++++++++++++++++---
 1 file changed, 52 insertions(+), 3 deletions(-)

diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c
index 0223694..115250b 100644
--- a/kernel/sched/deadline.c
+++ b/kernel/sched/deadline.c
@@ -1319,12 +1319,35 @@ static struct task_struct *pick_earliest_pushable_dl_task(struct rq *rq, int cpu
 
 static DEFINE_PER_CPU(cpumask_var_t, local_cpu_mask_dl);
 
+/*
+ * Find the first cpu in: mask & sd & ~prefer
+ */
+static int find_cpu(const struct cpumask *mask,
+		    const struct sched_domain *sd,
+		    const struct sched_domain *prefer)
+{
+	const struct cpumask *sds = sched_domain_span(sd);
+	const struct cpumask *ps  = prefer ? sched_domain_span(prefer) : NULL;
+	int cpu;
+
+	for_each_cpu(cpu, mask) {
+		if (!cpumask_test_cpu(cpu, sds))
+			continue;
+		if (ps && cpumask_test_cpu(cpu, ps))
+			continue;
+		break;
+	}
+
+	return cpu;
+}
+
 static int find_later_rq(struct task_struct *task)
 {
-	struct sched_domain *sd;
+	struct sched_domain *sd, *prefer = NULL;
 	struct cpumask *later_mask = this_cpu_cpumask_var_ptr(local_cpu_mask_dl);
 	int this_cpu = smp_processor_id();
 	int cpu = task_cpu(task);
+	int fallback_cpu = -1;
 
 	/* Make sure the mask is initialized first */
 	if (unlikely(!later_mask))
@@ -1376,8 +1399,7 @@ static int find_later_rq(struct task_struct *task)
 				return this_cpu;
 			}
 
-			best_cpu = cpumask_first_and(later_mask,
-							sched_domain_span(sd));
+			best_cpu = find_cpu(later_mask, sd, prefer);
 			/*
 			 * Last chance: if a cpu being in both later_mask
 			 * and current sd span is valid, that becomes our
@@ -1385,6 +1407,26 @@ static int find_later_rq(struct task_struct *task)
 			 * already under consideration through later_mask.
 			 */
 			if (best_cpu < nr_cpu_ids) {
+				/*
+				 * If current domain is SD_PREFER_SIBLING
+				 * flaged, we have to try to check other
+				 * siblings first.
+				 */
+				if (sd->flags & SD_PREFER_SIBLING) {
+					prefer = sd;
+
+					/*
+					 * fallback_cpu should be one
+					 * in the closest domain among
+					 * SD_PREFER_SIBLING domains,
+					 * in case that more than one
+					 * SD_PREFER_SIBLING domains
+					 * exist in the hierachy.
+					 */
+					if (fallback_cpu == -1)
+						fallback_cpu = best_cpu;
+					continue;
+				}
 				rcu_read_unlock();
 				return best_cpu;
 			}
@@ -1393,6 +1435,13 @@ static int find_later_rq(struct task_struct *task)
 	rcu_read_unlock();
 
 	/*
+	 * If fallback_cpu is valid, all our guesses failed *except* for
+	 * SD_PREFER_SIBLING domain. Now, we can return the fallback cpu.
+	 */
+	if (fallback_cpu != -1)
+		return fallback_cpu;
+
+	/*
 	 * At this point, all our guesses failed, we just return
 	 * 'something', and let the caller sort the things out.
 	 */
-- 
1.9.1

[PATCH v8 2/2] sched/rt: Add support for SD_PREFER_SIBLING on find_lowest_rq()

[Byungchul Park]

It would be better to try to check other siblings first if
SD_PREFER_SIBLING is flaged when pushing tasks - migration.

Signed-off-by: Byungchul Park <byungchul.park@lge.com>
---
 kernel/sched/rt.c | 56 ++++++++++++++++++++++++++++++++++++++++++++++++++++---
 1 file changed, 53 insertions(+), 3 deletions(-)

diff --git a/kernel/sched/rt.c b/kernel/sched/rt.c
index 979b734..b6d8ba9 100644
--- a/kernel/sched/rt.c
+++ b/kernel/sched/rt.c
@@ -1618,12 +1618,35 @@ static struct task_struct *pick_highest_pushable_task(struct rq *rq, int cpu)
 
 static DEFINE_PER_CPU(cpumask_var_t, local_cpu_mask);
 
+/*
+ * Find the first cpu in: mask & sd & ~prefer
+ */
+static int find_cpu(const struct cpumask *mask,
+		    const struct sched_domain *sd,
+		    const struct sched_domain *prefer)
+{
+	const struct cpumask *sds = sched_domain_span(sd);
+	const struct cpumask *ps  = prefer ? sched_domain_span(prefer) : NULL;
+	int cpu;
+
+	for_each_cpu(cpu, mask) {
+		if (!cpumask_test_cpu(cpu, sds))
+			continue;
+		if (ps && cpumask_test_cpu(cpu, ps))
+			continue;
+		break;
+	}
+
+	return cpu;
+}
+
 static int find_lowest_rq(struct task_struct *task)
 {
-	struct sched_domain *sd;
+	struct sched_domain *sd, *prefer = NULL;
 	struct cpumask *lowest_mask = this_cpu_cpumask_var_ptr(local_cpu_mask);
 	int this_cpu = smp_processor_id();
 	int cpu      = task_cpu(task);
+	int fallback_cpu = -1;
 
 	/* Make sure the mask is initialized first */
 	if (unlikely(!lowest_mask))
@@ -1668,9 +1691,29 @@ static int find_lowest_rq(struct task_struct *task)
 				return this_cpu;
 			}
 
-			best_cpu = cpumask_first_and(lowest_mask,
-						     sched_domain_span(sd));
+			best_cpu = find_cpu(lowest_mask, sd, prefer);
+
 			if (best_cpu < nr_cpu_ids) {
+				/*
+				 * If current domain is SD_PREFER_SIBLING
+				 * flaged, we have to try to check other
+				 * siblings first.
+				 */
+				if (sd->flags & SD_PREFER_SIBLING) {
+					prefer = sd;
+
+					/*
+					 * fallback_cpu should be one
+					 * in the closest domain among
+					 * SD_PREFER_SIBLING domains,
+					 * in case that more than one
+					 * SD_PREFER_SIBLING domains
+					 * exist in the hierachy.
+					 */
+					if (fallback_cpu == -1)
+						fallback_cpu = best_cpu;
+					continue;
+				}
 				rcu_read_unlock();
 				return best_cpu;
 			}
@@ -1679,6 +1722,13 @@ static int find_lowest_rq(struct task_struct *task)
 	rcu_read_unlock();
 
 	/*
+	 * If fallback_cpu is valid, all our guesses failed *except* for
+	 * SD_PREFER_SIBLING domain. Now, we can return the fallback cpu.
+	 */
+	if (fallback_cpu != -1)
+		return fallback_cpu;
+
+	/*
 	 * And finally, if there were no matches within the domains
 	 * just give the caller *something* to work with from the compatible
 	 * locations.
-- 
1.9.1

Re: [PATCH v8 2/2] sched/rt: Add support for SD_PREFER_SIBLING on find_lowest_rq()

[Steven Rostedt]

On Fri, 18 Aug 2017 17:21:59 +0900
Byungchul Park <byungchul.park@lge.com> wrote:

> It would be better to try to check other siblings first if
> SD_PREFER_SIBLING is flaged when pushing tasks - migration.
> 
> Signed-off-by: Byungchul Park <byungchul.park@lge.com>

Looks good.

Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>

-- Steve

Re: [PATCH v8 1/2] sched/deadline: Add support for SD_PREFER_SIBLING on find_later_rq()

[Juri Lelli]

Hi,

On 18/08/17 17:21, Byungchul Park wrote:
> It would be better to try to check other siblings first if
> SD_PREFER_SIBLING is flaged when pushing tasks - migration.
> 
> Signed-off-by: Byungchul Park <byungchul.park@lge.com>

Mmm, this looks like Peter's proposed patch, maybe add (at least) a
Suggested-by: him ?

https://marc.info/?l=linux-kernel&m=150176183807073

Also, I'm not sure what Peter meant with

"But still this isn't quite right, because when we consider this for SMT
(as was the intent here) we'll happily occupy a full sibling core over
finding an empty one."

since we are still using the later_mask, which should not include full
cores (unless it is the one with the lates deadline)?

> ---
>  kernel/sched/deadline.c | 55 ++++++++++++++++++++++++++++++++++++++++++++++---
>  1 file changed, 52 insertions(+), 3 deletions(-)
> 
> diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c
> index 0223694..115250b 100644
> --- a/kernel/sched/deadline.c
> +++ b/kernel/sched/deadline.c
> @@ -1319,12 +1319,35 @@ static struct task_struct *pick_earliest_pushable_dl_task(struct rq *rq, int cpu
>  
>  static DEFINE_PER_CPU(cpumask_var_t, local_cpu_mask_dl);
>  
> +/*
> + * Find the first cpu in: mask & sd & ~prefer
> + */
> +static int find_cpu(const struct cpumask *mask,
> +		    const struct sched_domain *sd,
> +		    const struct sched_domain *prefer)
> +{
> +	const struct cpumask *sds = sched_domain_span(sd);
> +	const struct cpumask *ps  = prefer ? sched_domain_span(prefer) : NULL;
> +	int cpu;
> +
> +	for_each_cpu(cpu, mask) {
> +		if (!cpumask_test_cpu(cpu, sds))
> +			continue;
> +		if (ps && cpumask_test_cpu(cpu, ps))
> +			continue;
> +		break;
> +	}
> +
> +	return cpu;
> +}
> +
>  static int find_later_rq(struct task_struct *task)
>  {
> -	struct sched_domain *sd;
> +	struct sched_domain *sd, *prefer = NULL;
>  	struct cpumask *later_mask = this_cpu_cpumask_var_ptr(local_cpu_mask_dl);
>  	int this_cpu = smp_processor_id();
>  	int cpu = task_cpu(task);
> +	int fallback_cpu = -1;
>  
>  	/* Make sure the mask is initialized first */
>  	if (unlikely(!later_mask))
> @@ -1376,8 +1399,7 @@ static int find_later_rq(struct task_struct *task)
>  				return this_cpu;
>  			}
>  
> -			best_cpu = cpumask_first_and(later_mask,
> -							sched_domain_span(sd));
> +			best_cpu = find_cpu(later_mask, sd, prefer);
>  			/*
>  			 * Last chance: if a cpu being in both later_mask
>  			 * and current sd span is valid, that becomes our
> @@ -1385,6 +1407,26 @@ static int find_later_rq(struct task_struct *task)
>  			 * already under consideration through later_mask.
>  			 */

It seems that the comment above should be updated as well.

Thanks,

- Juri

Re: [PATCH v8 1/2] sched/deadline: Add support for SD_PREFER_SIBLING on find_later_rq()

[Peter Zijlstra]

On Mon, Aug 21, 2017 at 02:44:58PM +0100, Juri Lelli wrote:

> Also, I'm not sure what Peter meant with
> 
> "But still this isn't quite right, because when we consider this for SMT
> (as was the intent here) we'll happily occupy a full sibling core over
> finding an empty one."

Consider a 4 core, SMT2 system:

LLC	[0         -         7]

SMT	[0,1] [2,3] [4,5] [6,7]

If we do a wake-up on CPU0, we'll find CPU1, mark that as fallback,
continue up the domain tree, exclude 0,1 from 0-7 and find CPU2.

A next wakeup on CPU0 does the same and will find CPU3, fully loading
that core, instead of considering CPU4 first.

Doing this 'right' is difficult and expensive :-/

Re: [PATCH v8 1/2] sched/deadline: Add support for SD_PREFER_SIBLING on find_later_rq()

[Juri Lelli]

On 21/08/17 15:56, Peter Zijlstra wrote:
> On Mon, Aug 21, 2017 at 02:44:58PM +0100, Juri Lelli wrote:
> 
> > Also, I'm not sure what Peter meant with
> > 
> > "But still this isn't quite right, because when we consider this for SMT
> > (as was the intent here) we'll happily occupy a full sibling core over
> > finding an empty one."
> 
> Consider a 4 core, SMT2 system:
> 
> LLC	[0         -         7]
> 
> SMT	[0,1] [2,3] [4,5] [6,7]
> 
> If we do a wake-up on CPU0, we'll find CPU1, mark that as fallback,
> continue up the domain tree, exclude 0,1 from 0-7 and find CPU2.
> 
> A next wakeup on CPU0 does the same and will find CPU3, fully loading
> that core, instead of considering CPU4 first.
> 

Ah, right, I see. Thanks for explaining.

Byungchul, maybe you could add this explanation as a comment?

> Doing this 'right' is difficult and expensive :-/
> 

Re: [PATCH v8 1/2] sched/deadline: Add support for SD_PREFER_SIBLING on find_later_rq()

[Byungchul Park]

On Mon, Aug 21, 2017 at 02:44:58PM +0100, Juri Lelli wrote:
> Hi,
> On 18/08/17 17:21, Byungchul Park wrote:
> > It would be better to try to check other siblings first if
> > SD_PREFER_SIBLING is flaged when pushing tasks - migration.
> > 
> > Signed-off-by: Byungchul Park <byungchul.park@lge.com>
> 
> Mmm, this looks like Peter's proposed patch, maybe add (at least) a
> Suggested-by: him ?

Hi Juri,

Why not. I will add it from the next spin.

BTW, is it enough? I don't know the way I should do, whenever I got
thankful suggestions. I really want to add them as a separate patch
which can be stacked on my patches _if possible_. But in case that
it's better to merge them into one like this, I don't know how.

I mean I will add 'Suggested-by' from now on - I learned what I should
do (at least) in this case thanks to Juri, but I'm still not sure if
it's enough.

Speaking of which, I have something to ask Peterz and Ingo for. I really
want to interact with maintainers actively e.g. asking ways they prefer.
But it takes too much long to get responses from them e.g. at most 2
monthes in case rushing them. I should have decided and done what the
best I think is, than asking.

It would be very appriciated if you pay more attention.

> > @@ -1376,8 +1399,7 @@ static int find_later_rq(struct task_struct *task)
> >  				return this_cpu;
> >  			}
> >  
> > -			best_cpu = cpumask_first_and(later_mask,
> > -							sched_domain_span(sd));
> > +			best_cpu = find_cpu(later_mask, sd, prefer);
> >  			/*
> >  			 * Last chance: if a cpu being in both later_mask
> >  			 * and current sd span is valid, that becomes our
> > @@ -1385,6 +1407,26 @@ static int find_later_rq(struct task_struct *task)
> >  			 * already under consideration through later_mask.
> >  			 */
> 
> It seems that the comment above should be updated as well.

How? Could you explain it more?

Thanks,
Byungchul

Re: [PATCH v8 1/2] sched/deadline: Add support for SD_PREFER_SIBLING on find_later_rq()

[Byungchul Park]

On Mon, Aug 21, 2017 at 03:07:57PM +0100, Juri Lelli wrote:
> > Consider a 4 core, SMT2 system:
> > 
> > LLC	[0         -         7]
> > 
> > SMT	[0,1] [2,3] [4,5] [6,7]
> > 
> > If we do a wake-up on CPU0, we'll find CPU1, mark that as fallback,
> > continue up the domain tree, exclude 0,1 from 0-7 and find CPU2.
> > 
> > A next wakeup on CPU0 does the same and will find CPU3, fully loading
> > that core, instead of considering CPU4 first.
> > 
> 
> Ah, right, I see. Thanks for explaining.
> 
> Byungchul, maybe you could add this explanation as a comment?

Yes. Good idea. I will add it.

Thank you,
Byungchul

Re: [PATCH v8 1/2] sched/deadline: Add support for SD_PREFER_SIBLING on find_later_rq()

[Byungchul Park]

On Tue, Aug 22, 2017 at 02:53:25PM +0900, Byungchul Park wrote:
> On Mon, Aug 21, 2017 at 02:44:58PM +0100, Juri Lelli wrote:
> > Hi,
> > On 18/08/17 17:21, Byungchul Park wrote:
> > > It would be better to try to check other siblings first if
> > > SD_PREFER_SIBLING is flaged when pushing tasks - migration.
> > > 
> > > Signed-off-by: Byungchul Park <byungchul.park@lge.com>
> > 
> > Mmm, this looks like Peter's proposed patch, maybe add (at least) a
> > Suggested-by: him ?
> 
> Hi Juri,
> 
> Why not. I will add it from the next spin.
> 
> BTW, is it enough? I don't know the way I should do, whenever I got
> thankful suggestions. I really want to add them as a separate patch
> which can be stacked on my patches _if possible_. But in case that
> it's better to merge them into one like this, I don't know how.
> 
> I mean I will add 'Suggested-by' from now on - I learned what I should
> do (at least) in this case thanks to Juri, but I'm still not sure if
> it's enough.
> 
> Speaking of which, I have something to ask Peterz and Ingo for. I really
> want to interact with maintainers actively e.g. asking ways they prefer.
> But it takes too much long to get responses from them e.g. at most 2
> monthes in case rushing them. I should have decided and done what the
> best I think is, than asking.
> 
> It would be very appriciated if you pay more attention.
> 
> > > @@ -1376,8 +1399,7 @@ static int find_later_rq(struct task_struct *task)
> > >  				return this_cpu;
> > >  			}
> > >  
> > > -			best_cpu = cpumask_first_and(later_mask,
> > > -							sched_domain_span(sd));
> > > +			best_cpu = find_cpu(later_mask, sd, prefer);
> > >  			/*
> > >  			 * Last chance: if a cpu being in both later_mask
> > >  			 * and current sd span is valid, that becomes our
> > > @@ -1385,6 +1407,26 @@ static int find_later_rq(struct task_struct *task)
> > >  			 * already under consideration through later_mask.
> > >  			 */
> > 
> > It seems that the comment above should be updated as well.
> 
> How? Could you explain it more?

Let me try it by myself.. Please fix me at the next spin if needed.

Thank you,
Byungchul

Re: [PATCH v8 1/2] sched/deadline: Add support for SD_PREFER_SIBLING on find_later_rq()

[Juri Lelli]

Hi,

On 22/08/17 14:53, Byungchul Park wrote:
> On Mon, Aug 21, 2017 at 02:44:58PM +0100, Juri Lelli wrote:
> > Hi,
> > On 18/08/17 17:21, Byungchul Park wrote:
> > > It would be better to try to check other siblings first if
> > > SD_PREFER_SIBLING is flaged when pushing tasks - migration.
> > > 
> > > Signed-off-by: Byungchul Park <byungchul.park@lge.com>
> > 
> > Mmm, this looks like Peter's proposed patch, maybe add (at least) a
> > Suggested-by: him ?
> 
> Hi Juri,
> 
> Why not. I will add it from the next spin.
> 
> BTW, is it enough? I don't know the way I should do, whenever I got
> thankful suggestions. I really want to add them as a separate patch
> which can be stacked on my patches _if possible_. But in case that
> it's better to merge them into one like this, I don't know how.
> 

It usually depends on the entity of the suggestion (is it expressed with
a sentence, actual code or a proper patch?) and what the person
suggesting it is fine with. I usually simply ask. :)

> I mean I will add 'Suggested-by' from now on - I learned what I should
> do (at least) in this case thanks to Juri, but I'm still not sure if
> it's enough.
> 
> Speaking of which, I have something to ask Peterz and Ingo for. I really
> want to interact with maintainers actively e.g. asking ways they prefer.
> But it takes too much long to get responses from them e.g. at most 2
> monthes in case rushing them. I should have decided and done what the
> best I think is, than asking.
> 
> It would be very appriciated if you pay more attention.
> 

I try to be as responsive as I can (I think this applies to everyone)
and I apologize if from time to time it takes too much to reply. Balance
between upstream and product work means that there are times when one of
the two gets a bit delayed, I'm afraid.

Please keep asking questions, propose solutions and chase people! :)

> > > @@ -1376,8 +1399,7 @@ static int find_later_rq(struct task_struct *task)
> > >  				return this_cpu;
> > >  			}
> > >  
> > > -			best_cpu = cpumask_first_and(later_mask,
> > > -							sched_domain_span(sd));
> > > +			best_cpu = find_cpu(later_mask, sd, prefer);
> > >  			/*
> > >  			 * Last chance: if a cpu being in both later_mask
> > >  			 * and current sd span is valid, that becomes our
> > > @@ -1385,6 +1407,26 @@ static int find_later_rq(struct task_struct *task)
> > >  			 * already under consideration through later_mask.
> > >  			 */
> > 
> > It seems that the comment above should be updated as well.
> 
> How? Could you explain it more?
> 

Simply removing it might just work.

Thanks,

- Juri