Messages in this thread |  | | | Date | Tue, 1 Jul 2008 08:26:04 +0200 | | From | Ingo Molnar <> | | Subject | Re: divide error trying to set rt_period to zero | |
* Raistlin <raistlin@linux.it> wrote:
> Ops! :-O
>
> My fault, moving the 'if' back and forth from one function to another
> I finally sent the wrong (and broken!) version of the patch... Very
> sorry about it!
>
> I know, I would have noticed just looking at the code... But I missed
> it. :-(
hey, no problem.
> Anyway, if still interested in, the following one build and work...
> I've just tried it and I'm quite sure this time. :-P
applied to tip/sched/urgent - thanks Dario. We are interested in it of
course, you fixed a real bug in the scheduler :-)
small sidenote:
> +++ b/kernel/sched.c
> @@ -8501,6 +8501,9 @@ int sched_group_set_rt_period(struct task_group
> *tg, long rt_period_us)
> rt_period = (u64)rt_period_us * NSEC_PER_USEC;
> rt_runtime = tg->rt_bandwidth.rt_runtime;
>
> + if (rt_period == 0)
> + return -EINVAL;
> +
> return tg_set_bandwidth(tg, rt_period, rt_runtime);
> }
that patch was whitespace damaged (lines wrapped and tabs converted to
spaces). I fixed it up by hand - you can avoid such problems with future
patches by sending patches a'la Documentation/email-clients.txt.
Ingo
|  |