Re: [PATCH] sched/core: fix sched_rt_global_validate
From: Juri Lelli
Date: Mon Feb 03 2014 - 05:14:20 EST
On 02/03/2014 10:53 AM, Henrik Austad wrote:
> On Fri, Jan 31, 2014 at 05:43:27PM +0100, Juri Lelli wrote:
>> Don't compare sysctl_sched_rt_runtime against sysctl_sched_rt_period if
>> the former is equal to RUNTIME_INF, otherwise disabling -rt bandwidth
>> management always fails.
>>
>> Cc: Ingo Molnar <mingo@xxxxxxxxxx>
>> Cc: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
>> Signed-off-by: Juri Lelli <juri.lelli@xxxxxxxxx>
>> ---
>> kernel/sched/core.c | 3 ++-
>> 1 file changed, 2 insertions(+), 1 deletion(-)
>>
>> diff --git a/kernel/sched/core.c b/kernel/sched/core.c
>> index 210a12a..5c0a304 100644
>> --- a/kernel/sched/core.c
>> +++ b/kernel/sched/core.c
>> @@ -7477,7 +7477,8 @@ static int sched_rt_global_validate(void)
>> if (sysctl_sched_rt_period <= 0)
>> return -EINVAL;
>>
>> - if (sysctl_sched_rt_runtime > sysctl_sched_rt_period)
>> + if ((sysctl_sched_rt_runtime != RUNTIME_INF) &&
>> + (sysctl_sched_rt_runtime > sysctl_sched_rt_period))
>> return -EINVAL;
>
> Won't this be caught by the test above?
>
> #define RUNTIME_INF ((u64)~0ULL)
>
> which means that if sysctl_sched_rt_runtime is set to RUNTIME_INF, it will
> trigger on the previous test, and the first part of this test will always
> be true.
>
> Or have I suffered catastrophic monday-morning braindamage?
>
As I understand it. When you do
echo -1 > /proc/sys/kernel/sched_rt_runtime_us
sysctl_sched_rt_runtime is actually set to -1 (being an int).
But then you compare it against and unsigned int, so the cast converts it to
actually be RUNTIME_INF, and thus greater than sysctl_sched_rt_period (so the
function returns -EINVAL, while you'd want it to return 0, as you are disabling
-rt throttling).
Makes sense?
Thanks,
- Juri
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/