2014-11-13 08:48:00

by Wanpeng Li

Subject: [PATCH] sched/rt: fix rt runtime corrupt when dl refuse a smaller bandwidth

Dl class will refuse the bandwidth being set to some value smaller
than the currently allocated bandwidth in any of the root_domains
through sched_rt_runtime_us and sched_rt_period_us. RT runtime will
be set according to sched_rt_runtime_us before dl class verify if
the new bandwidth is suitable in the case of !CONFIG_RT_GROUP_SCHED.

However, rt runtime will be corrupt if dl refuse the new bandwidth
since there is no undo to reset the rt runtime to the old value.

This patch fix it by setting rt runtime after all kinds of rational
checking in the case of !CONFIG_RT_GROUP_SCHED.

Signed-off-by: Wanpeng Li <[email protected]>
---
kernel/sched/core.c | 30 ++++++++++++++++--------------
1 file changed, 16 insertions(+), 14 deletions(-)

diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 2e7578a..355dde3 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -7795,20 +7795,7 @@ static int sched_rt_can_attach(struct task_group *tg, struct task_struct *tsk)
#else /* !CONFIG_RT_GROUP_SCHED */
static int sched_rt_global_constraints(void)
{
- unsigned long flags;
- int i, ret = 0;
-
- raw_spin_lock_irqsave(&def_rt_bandwidth.rt_runtime_lock, flags);
- for_each_possible_cpu(i) {
- struct rt_rq *rt_rq = &cpu_rq(i)->rt;
-
- raw_spin_lock(&rt_rq->rt_runtime_lock);
- rt_rq->rt_runtime = global_rt_runtime();
- raw_spin_unlock(&rt_rq->rt_runtime_lock);
- }
- raw_spin_unlock_irqrestore(&def_rt_bandwidth.rt_runtime_lock, flags);
-
- return ret;
+ return 0;
}
#endif /* CONFIG_RT_GROUP_SCHED */

@@ -7890,6 +7877,21 @@ static int sched_rt_global_validate(void)

static void sched_rt_do_global(void)
{
+#ifndef CONFIG_RT_GROUP_SCHED
+ unsigned long flags;
+ int i;
+
+ raw_spin_lock_irqsave(&def_rt_bandwidth.rt_runtime_lock, flags);
+ for_each_possible_cpu(i) {
+ struct rt_rq *rt_rq = &cpu_rq(i)->rt;
+
+ raw_spin_lock(&rt_rq->rt_runtime_lock);
+ rt_rq->rt_runtime = global_rt_runtime();
+ raw_spin_unlock(&rt_rq->rt_runtime_lock);
+ }
+ raw_spin_unlock_irqrestore(&def_rt_bandwidth.rt_runtime_lock, flags);
+#endif
+
def_rt_bandwidth.rt_runtime = global_rt_runtime();
def_rt_bandwidth.rt_period = ns_to_ktime(global_rt_period());
}
--
1.9.1

2014-11-18 02:41:42

by Wanpeng Li

Subject: Re: [PATCH] sched/rt: fix rt runtime corrupt when dl refuse a smaller bandwidth

Ping Juri, ;-)
On Thu, Nov 13, 2014 at 04:47:39PM +0800, Wanpeng Li wrote:
>Dl class will refuse the bandwidth being set to some value smaller
>than the currently allocated bandwidth in any of the root_domains
>through sched_rt_runtime_us and sched_rt_period_us. RT runtime will
>be set according to sched_rt_runtime_us before dl class verify if
>the new bandwidth is suitable in the case of !CONFIG_RT_GROUP_SCHED.
>
>However, rt runtime will be corrupt if dl refuse the new bandwidth
>since there is no undo to reset the rt runtime to the old value.
>
>This patch fix it by setting rt runtime after all kinds of rational
>checking in the case of !CONFIG_RT_GROUP_SCHED.
>
>Signed-off-by: Wanpeng Li <[email protected]>
>---
> kernel/sched/core.c | 30 ++++++++++++++++--------------
> 1 file changed, 16 insertions(+), 14 deletions(-)
>
>diff --git a/kernel/sched/core.c b/kernel/sched/core.c
>index 2e7578a..355dde3 100644
>--- a/kernel/sched/core.c
>+++ b/kernel/sched/core.c
>@@ -7795,20 +7795,7 @@ static int sched_rt_can_attach(struct task_group *tg, struct task_struct *tsk)
> #else /* !CONFIG_RT_GROUP_SCHED */
> static int sched_rt_global_constraints(void)
> {
>- unsigned long flags;
>- int i, ret = 0;
>-
>- raw_spin_lock_irqsave(&def_rt_bandwidth.rt_runtime_lock, flags);
>- for_each_possible_cpu(i) {
>- struct rt_rq *rt_rq = &cpu_rq(i)->rt;
>-
>- raw_spin_lock(&rt_rq->rt_runtime_lock);
>- rt_rq->rt_runtime = global_rt_runtime();
>- raw_spin_unlock(&rt_rq->rt_runtime_lock);
>- }
>- raw_spin_unlock_irqrestore(&def_rt_bandwidth.rt_runtime_lock, flags);
>-
>- return ret;
>+ return 0;
> }
> #endif /* CONFIG_RT_GROUP_SCHED */
>
>@@ -7890,6 +7877,21 @@ static int sched_rt_global_validate(void)
>
> static void sched_rt_do_global(void)
> {
>+#ifndef CONFIG_RT_GROUP_SCHED
>+ unsigned long flags;
>+ int i;
>+
>+ raw_spin_lock_irqsave(&def_rt_bandwidth.rt_runtime_lock, flags);
>+ for_each_possible_cpu(i) {
>+ struct rt_rq *rt_rq = &cpu_rq(i)->rt;
>+
>+ raw_spin_lock(&rt_rq->rt_runtime_lock);
>+ rt_rq->rt_runtime = global_rt_runtime();
>+ raw_spin_unlock(&rt_rq->rt_runtime_lock);
>+ }
>+ raw_spin_unlock_irqrestore(&def_rt_bandwidth.rt_runtime_lock, flags);
>+#endif
>+
> def_rt_bandwidth.rt_runtime = global_rt_runtime();
> def_rt_bandwidth.rt_period = ns_to_ktime(global_rt_period());
> }
>--
>1.9.1

2014-11-18 18:00:04

by Juri Lelli

Subject: Re: [PATCH] sched/rt: fix rt runtime corrupt when dl refuse a smaller bandwidth

Hi,

On 13/11/14 08:47, Wanpeng Li wrote:
> Dl class will refuse the bandwidth being set to some value smaller
> than the currently allocated bandwidth in any of the root_domains
> through sched_rt_runtime_us and sched_rt_period_us. RT runtime will
> be set according to sched_rt_runtime_us before dl class verify if
> the new bandwidth is suitable in the case of !CONFIG_RT_GROUP_SCHED.
>
> However, rt runtime will be corrupt if dl refuse the new bandwidth
> since there is no undo to reset the rt runtime to the old value.
>

Can't we just move sched_dl_global_constraints() before
sched_rt_global_constraints(), and change the name of the
former to sched_dl_global_validate() ?

Thanks,

- Juri

> This patch fix it by setting rt runtime after all kinds of rational
> checking in the case of !CONFIG_RT_GROUP_SCHED.
>
> Signed-off-by: Wanpeng Li <[email protected]>
> ---
> kernel/sched/core.c | 30 ++++++++++++++++--------------
> 1 file changed, 16 insertions(+), 14 deletions(-)
>
> diff --git a/kernel/sched/core.c b/kernel/sched/core.c
> index 2e7578a..355dde3 100644
> --- a/kernel/sched/core.c
> +++ b/kernel/sched/core.c
> @@ -7795,20 +7795,7 @@ static int sched_rt_can_attach(struct task_group *tg, struct task_struct *tsk)
> #else /* !CONFIG_RT_GROUP_SCHED */
> static int sched_rt_global_constraints(void)
> {
> - unsigned long flags;
> - int i, ret = 0;
> -
> - raw_spin_lock_irqsave(&def_rt_bandwidth.rt_runtime_lock, flags);
> - for_each_possible_cpu(i) {
> - struct rt_rq *rt_rq = &cpu_rq(i)->rt;
> -
> - raw_spin_lock(&rt_rq->rt_runtime_lock);
> - rt_rq->rt_runtime = global_rt_runtime();
> - raw_spin_unlock(&rt_rq->rt_runtime_lock);
> - }
> - raw_spin_unlock_irqrestore(&def_rt_bandwidth.rt_runtime_lock, flags);
> -
> - return ret;
> + return 0;
> }
> #endif /* CONFIG_RT_GROUP_SCHED */
>
> @@ -7890,6 +7877,21 @@ static int sched_rt_global_validate(void)
>
> static void sched_rt_do_global(void)
> {
> +#ifndef CONFIG_RT_GROUP_SCHED
> + unsigned long flags;
> + int i;
> +
> + raw_spin_lock_irqsave(&def_rt_bandwidth.rt_runtime_lock, flags);
> + for_each_possible_cpu(i) {
> + struct rt_rq *rt_rq = &cpu_rq(i)->rt;
> +
> + raw_spin_lock(&rt_rq->rt_runtime_lock);
> + rt_rq->rt_runtime = global_rt_runtime();
> + raw_spin_unlock(&rt_rq->rt_runtime_lock);
> + }
> + raw_spin_unlock_irqrestore(&def_rt_bandwidth.rt_runtime_lock, flags);
> +#endif
> +
> def_rt_bandwidth.rt_runtime = global_rt_runtime();
> def_rt_bandwidth.rt_period = ns_to_ktime(global_rt_period());
> }
>

2014-11-18 22:42:09

by Wanpeng Li

Subject: Re: [PATCH] sched/rt: fix rt runtime corrupt when dl refuse a smaller bandwidth

Hi Juri,
On Tue, Nov 18, 2014 at 06:00:20PM +0000, Juri Lelli wrote:
>Hi,
>
>On 13/11/14 08:47, Wanpeng Li wrote:
>> Dl class will refuse the bandwidth being set to some value smaller
>> than the currently allocated bandwidth in any of the root_domains
>> through sched_rt_runtime_us and sched_rt_period_us. RT runtime will
>> be set according to sched_rt_runtime_us before dl class verify if
>> the new bandwidth is suitable in the case of !CONFIG_RT_GROUP_SCHED.
>>
>> However, rt runtime will be corrupt if dl refuse the new bandwidth
>> since there is no undo to reset the rt runtime to the old value.
>>
>
>Can't we just move sched_dl_global_constraints() before
>sched_rt_global_constraints(), and change the name of the
>former to sched_dl_global_validate() ?

Good idea, I will send out the next version later. ;-)

Regards,
Wanpeng Li

>
>Thanks,
>
>- Juri
>
>> This patch fix it by setting rt runtime after all kinds of rational
>> checking in the case of !CONFIG_RT_GROUP_SCHED.
>>
>> Signed-off-by: Wanpeng Li <[email protected]>
>> ---
>> kernel/sched/core.c | 30 ++++++++++++++++--------------
>> 1 file changed, 16 insertions(+), 14 deletions(-)
>>
>> diff --git a/kernel/sched/core.c b/kernel/sched/core.c
>> index 2e7578a..355dde3 100644
>> --- a/kernel/sched/core.c
>> +++ b/kernel/sched/core.c
>> @@ -7795,20 +7795,7 @@ static int sched_rt_can_attach(struct task_group *tg, struct task_struct *tsk)
>> #else /* !CONFIG_RT_GROUP_SCHED */
>> static int sched_rt_global_constraints(void)
>> {
>> - unsigned long flags;
>> - int i, ret = 0;
>> -
>> - raw_spin_lock_irqsave(&def_rt_bandwidth.rt_runtime_lock, flags);
>> - for_each_possible_cpu(i) {
>> - struct rt_rq *rt_rq = &cpu_rq(i)->rt;
>> -
>> - raw_spin_lock(&rt_rq->rt_runtime_lock);
>> - rt_rq->rt_runtime = global_rt_runtime();
>> - raw_spin_unlock(&rt_rq->rt_runtime_lock);
>> - }
>> - raw_spin_unlock_irqrestore(&def_rt_bandwidth.rt_runtime_lock, flags);
>> -
>> - return ret;
>> + return 0;
>> }
>> #endif /* CONFIG_RT_GROUP_SCHED */
>>
>> @@ -7890,6 +7877,21 @@ static int sched_rt_global_validate(void)
>>
>> static void sched_rt_do_global(void)
>> {
>> +#ifndef CONFIG_RT_GROUP_SCHED
>> + unsigned long flags;
>> + int i;
>> +
>> + raw_spin_lock_irqsave(&def_rt_bandwidth.rt_runtime_lock, flags);
>> + for_each_possible_cpu(i) {
>> + struct rt_rq *rt_rq = &cpu_rq(i)->rt;
>> +
>> + raw_spin_lock(&rt_rq->rt_runtime_lock);
>> + rt_rq->rt_runtime = global_rt_runtime();
>> + raw_spin_unlock(&rt_rq->rt_runtime_lock);
>> + }
>> + raw_spin_unlock_irqrestore(&def_rt_bandwidth.rt_runtime_lock, flags);
>> +#endif
>> +
>> def_rt_bandwidth.rt_runtime = global_rt_runtime();
>> def_rt_bandwidth.rt_period = ns_to_ktime(global_rt_period());
>> }
>>