[PATCH 1/2] sched_rt: don't start timer when rt bandwidth disabled

[PATCH 1/2] sched_rt: don't start timer when rt bandwidth disabled

[Hiroshi Shimamoto]

From: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>

Impact: fix incorrect condition check

No need to start rt bandwidth timer when rt bandwidth is disabled.

Signed-off-by: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
---
 kernel/sched.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/kernel/sched.c b/kernel/sched.c
index 1e965b7..4c03c88 100644
--- a/kernel/sched.c
+++ b/kernel/sched.c
@@ -223,7 +223,7 @@ static void start_rt_bandwidth(struct rt_bandwidth *rt_b)
 {
 	ktime_t now;
 
-	if (rt_bandwidth_enabled() && rt_b->rt_runtime == RUNTIME_INF)
+	if (!rt_bandwidth_enabled() || rt_b->rt_runtime == RUNTIME_INF)
 		return;
 
 	if (hrtimer_active(&rt_b->rt_period_timer))
-- 
1.6.1.2

[PATCH 2/2] sched_rt: protect rt_rq->rt_time by rt_runtime_lock

[Hiroshi Shimamoto]

From: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>

Impact: fix possible race condition

rt_rq->rt_time should be protected.

Signed-off-by: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
---
 kernel/sched_rt.c |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/kernel/sched_rt.c b/kernel/sched_rt.c
index 299d012..d7654a3 100644
--- a/kernel/sched_rt.c
+++ b/kernel/sched_rt.c
@@ -496,10 +496,10 @@ static int do_sched_rt_period_timer(struct rt_bandwidth *rt_b, int overrun)
 		struct rq *rq = rq_of_rt_rq(rt_rq);
 
 		spin_lock(&rq->lock);
+		spin_lock(&rt_rq->rt_runtime_lock);
 		if (rt_rq->rt_time) {
 			u64 runtime;
 
-			spin_lock(&rt_rq->rt_runtime_lock);
 			if (rt_rq->rt_throttled)
 				balance_runtime(rt_rq);
 			runtime = rt_rq->rt_runtime;
@@ -510,9 +510,9 @@ static int do_sched_rt_period_timer(struct rt_bandwidth *rt_b, int overrun)
 			}
 			if (rt_rq->rt_time || rt_rq->rt_nr_running)
 				idle = 0;
-			spin_unlock(&rt_rq->rt_runtime_lock);
 		} else if (rt_rq->rt_nr_running)
 			idle = 0;
+		spin_unlock(&rt_rq->rt_runtime_lock);
 
 		if (enqueue)
 			sched_rt_rq_enqueue(rt_rq);
-- 
1.6.1.2

Re: [PATCH 2/2] sched_rt: protect rt_rq->rt_time by rt_runtime_lock

[Peter Zijlstra]

On Thu, 2009-02-05 at 15:38 -0800, Hiroshi Shimamoto wrote:
> From: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
> 
> Impact: fix possible race condition
> 
> rt_rq->rt_time should be protected.

Don't think so, all we do is check the value for non-zero outside the
lock, that should be ok.

If there's still a race, your changelog utterly fails to mention the
race scenario.

> Signed-off-by: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
> ---
>  kernel/sched_rt.c |    4 ++--
>  1 files changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/kernel/sched_rt.c b/kernel/sched_rt.c
> index 299d012..d7654a3 100644
> --- a/kernel/sched_rt.c
> +++ b/kernel/sched_rt.c
> @@ -496,10 +496,10 @@ static int do_sched_rt_period_timer(struct rt_bandwidth *rt_b, int overrun)
>  		struct rq *rq = rq_of_rt_rq(rt_rq);
>  
>  		spin_lock(&rq->lock);
> +		spin_lock(&rt_rq->rt_runtime_lock);
>  		if (rt_rq->rt_time) {
>  			u64 runtime;
>  
> -			spin_lock(&rt_rq->rt_runtime_lock);
>  			if (rt_rq->rt_throttled)
>  				balance_runtime(rt_rq);
>  			runtime = rt_rq->rt_runtime;
> @@ -510,9 +510,9 @@ static int do_sched_rt_period_timer(struct rt_bandwidth *rt_b, int overrun)
>  			}
>  			if (rt_rq->rt_time || rt_rq->rt_nr_running)
>  				idle = 0;
> -			spin_unlock(&rt_rq->rt_runtime_lock);
>  		} else if (rt_rq->rt_nr_running)
>  			idle = 0;
> +		spin_unlock(&rt_rq->rt_runtime_lock);
>  
>  		if (enqueue)
>  			sched_rt_rq_enqueue(rt_rq);

Re: [PATCH 2/2] sched_rt: protect rt_rq->rt_time by rt_runtime_lock

[Hiroshi Shimamoto]

Peter Zijlstra wrote:
> On Thu, 2009-02-05 at 15:38 -0800, Hiroshi Shimamoto wrote:
>> From: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
>>
>> Impact: fix possible race condition
>>
>> rt_rq->rt_time should be protected.
> 
> Don't think so, all we do is check the value for non-zero outside the
> lock, that should be ok.
> 
> If there's still a race, your changelog utterly fails to mention the
> race scenario.

OK. I just noticed rt_time is accessed with rt_runtime_lock held except
this point. Will look at this again.

Thanks,
Hiroshi

> 
>> Signed-off-by: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
>> ---
>>  kernel/sched_rt.c |    4 ++--
>>  1 files changed, 2 insertions(+), 2 deletions(-)
>>
>> diff --git a/kernel/sched_rt.c b/kernel/sched_rt.c
>> index 299d012..d7654a3 100644
>> --- a/kernel/sched_rt.c
>> +++ b/kernel/sched_rt.c
>> @@ -496,10 +496,10 @@ static int do_sched_rt_period_timer(struct rt_bandwidth *rt_b, int overrun)
>>  		struct rq *rq = rq_of_rt_rq(rt_rq);
>>  
>>  		spin_lock(&rq->lock);
>> +		spin_lock(&rt_rq->rt_runtime_lock);
>>  		if (rt_rq->rt_time) {
>>  			u64 runtime;
>>  
>> -			spin_lock(&rt_rq->rt_runtime_lock);
>>  			if (rt_rq->rt_throttled)
>>  				balance_runtime(rt_rq);
>>  			runtime = rt_rq->rt_runtime;
>> @@ -510,9 +510,9 @@ static int do_sched_rt_period_timer(struct rt_bandwidth *rt_b, int overrun)
>>  			}
>>  			if (rt_rq->rt_time || rt_rq->rt_nr_running)
>>  				idle = 0;
>> -			spin_unlock(&rt_rq->rt_runtime_lock);
>>  		} else if (rt_rq->rt_nr_running)
>>  			idle = 0;
>> +		spin_unlock(&rt_rq->rt_runtime_lock);
>>  
>>  		if (enqueue)
>>  			sched_rt_rq_enqueue(rt_rq);
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
> 

Re: [PATCH 2/2] sched_rt: protect rt_rq->rt_time by rt_runtime_lock

[Hiroshi Shimamoto]

Hiroshi Shimamoto wrote:
> Peter Zijlstra wrote:
>> On Thu, 2009-02-05 at 15:38 -0800, Hiroshi Shimamoto wrote:
>>> From: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
>>>
>>> Impact: fix possible race condition
>>>
>>> rt_rq->rt_time should be protected.
>> Don't think so, all we do is check the value for non-zero outside the
>> lock, that should be ok.

OK, now I see. I didn't find any real race.

Thanks,
Hiroshi