[Xen-devel] [PATCH] xen/sched: avoid cpumasks on stack in sched/core.c

[Xen-devel] [PATCH] xen/sched: avoid cpumasks on stack in sched/core.c

[Juergen Gross]

There are still several instances of cpumask_t on the stack in
scheduling code. Avoid them as far as possible.

Signed-off-by: Juergen Gross <jgross@suse.com>
---
 xen/common/sched/core.c | 14 +++++---------
 1 file changed, 5 insertions(+), 9 deletions(-)

diff --git a/xen/common/sched/core.c b/xen/common/sched/core.c
index 944164d78a..73799c2508 100644
--- a/xen/common/sched/core.c
+++ b/xen/common/sched/core.c
@@ -1178,7 +1178,6 @@ int cpu_disable_scheduler(unsigned int cpu)
 {
     struct domain *d;
     const struct cpupool *c;
-    cpumask_t online_affinity;
     int ret = 0;
 
     rcu_read_lock(&sched_res_rculock);
@@ -1196,8 +1195,7 @@ int cpu_disable_scheduler(unsigned int cpu)
             unsigned long flags;
             spinlock_t *lock = unit_schedule_lock_irqsave(unit, &flags);
 
-            cpumask_and(&online_affinity, unit->cpu_hard_affinity, c->cpu_valid);
-            if ( cpumask_empty(&online_affinity) &&
+            if ( !cpumask_intersects(unit->cpu_hard_affinity, c->cpu_valid) &&
                  cpumask_test_cpu(cpu, unit->cpu_hard_affinity) )
             {
                 if ( sched_check_affinity_broken(unit) )
@@ -1336,12 +1334,10 @@ static int vcpu_set_affinity(
 
 int vcpu_set_hard_affinity(struct vcpu *v, const cpumask_t *affinity)
 {
-    cpumask_t online_affinity;
     cpumask_t *online;
 
     online = VCPU2ONLINE(v);
-    cpumask_and(&online_affinity, affinity, online);
-    if ( cpumask_empty(&online_affinity) )
+    if ( !cpumask_intersects(online, affinity) )
         return -EINVAL;
 
     return vcpu_set_affinity(v, affinity, v->sched_unit->cpu_hard_affinity);
@@ -2586,11 +2582,11 @@ static void schedule(void)
 
     if ( gran > 1 )
     {
-        cpumask_t mask;
+        cpumask_t *mask = cpumask_scratch_cpu(cpu);
 
         prev->rendezvous_in_cnt = gran;
-        cpumask_andnot(&mask, sr->cpus, cpumask_of(cpu));
-        cpumask_raise_softirq(&mask, SCHED_SLAVE_SOFTIRQ);
+        cpumask_andnot(mask, sr->cpus, cpumask_of(cpu));
+        cpumask_raise_softirq(mask, SCHED_SLAVE_SOFTIRQ);
         next = sched_wait_rendezvous_in(prev, &lock, cpu, now);
         if ( !next )
             return;
-- 
2.16.4


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel

Re: [Xen-devel] [PATCH] xen/sched: avoid cpumasks on stack in sched/core.c

[Dario Faggioli]

[-- Attachment #1.1: Type: text/plain, Size: 1195 bytes --]

On Thu, 2020-01-23 at 10:03 +0100, Juergen Gross wrote:
> There are still several instances of cpumask_t on the stack in
> scheduling code. Avoid them as far as possible.
> 
> Signed-off-by: Juergen Gross <jgross@suse.com>
>
Reviewed-by: Dario Faggioli <dfaggioli@suse.com>

Just curious...

> --- a/xen/common/sched/core.c
> +++ b/xen/common/sched/core.c
> @@ -2586,11 +2582,11 @@ static void schedule(void)
>  
>      if ( gran > 1 )
>      {
> -        cpumask_t mask;
> +        cpumask_t *mask = cpumask_scratch_cpu(cpu);
>  
>          prev->rendezvous_in_cnt = gran;
> -        cpumask_andnot(&mask, sr->cpus, cpumask_of(cpu));
> -        cpumask_raise_softirq(&mask, SCHED_SLAVE_SOFTIRQ);
> +        cpumask_andnot(mask, sr->cpus, cpumask_of(cpu));
> +        cpumask_raise_softirq(mask, SCHED_SLAVE_SOFTIRQ);
>
... why are we keeping the temporary variable (mask) ?

Thanks and Regards
-- 
Dario Faggioli, Ph.D
http://about.me/dario.faggioli
Virtualization Software Engineer
SUSE Labs, SUSE https://www.suse.com/
-------------------------------------------------------------------
<<This happens because _I_ choose it to happen!>> (Raistlin Majere)


[-- Attachment #1.2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

[-- Attachment #2: Type: text/plain, Size: 157 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel

Re: [Xen-devel] [PATCH] xen/sched: avoid cpumasks on stack in sched/core.c

[Jürgen Groß]

On 24.01.20 01:01, Dario Faggioli wrote:
> On Thu, 2020-01-23 at 10:03 +0100, Juergen Gross wrote:
>> There are still several instances of cpumask_t on the stack in
>> scheduling code. Avoid them as far as possible.
>>
>> Signed-off-by: Juergen Gross <jgross@suse.com>
>>
> Reviewed-by: Dario Faggioli <dfaggioli@suse.com>
> 
> Just curious...
> 
>> --- a/xen/common/sched/core.c
>> +++ b/xen/common/sched/core.c
>> @@ -2586,11 +2582,11 @@ static void schedule(void)
>>   
>>       if ( gran > 1 )
>>       {
>> -        cpumask_t mask;
>> +        cpumask_t *mask = cpumask_scratch_cpu(cpu);
>>   
>>           prev->rendezvous_in_cnt = gran;
>> -        cpumask_andnot(&mask, sr->cpus, cpumask_of(cpu));
>> -        cpumask_raise_softirq(&mask, SCHED_SLAVE_SOFTIRQ);
>> +        cpumask_andnot(mask, sr->cpus, cpumask_of(cpu));
>> +        cpumask_raise_softirq(mask, SCHED_SLAVE_SOFTIRQ);
>>
> ... why are we keeping the temporary variable (mask) ?

per_cpu accesses are more expensive than those to local variables.
mask is used twice.


Juergen

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel