On Fri, 24 Apr 2020 10:12:31 +0800
Hillf Danton <[email protected]> wrote:
> Yes and if you agree, we send it home during the 5.8 cycle, or not before
> fifo is reclaimed from modules.
>
> In the spin it now looks like
>
> --- a/include/linux/sched/prio.h
> +++ b/include/linux/sched/prio.h
> @@ -12,15 +12,12 @@
> * tasks are in the range MAX_RT_PRIO..MAX_PRIO-1. Priority
> * values are inverted: lower p->prio value means higher priority.
> *
> - * The MAX_USER_RT_PRIO value allows the actual maximum
> - * RT priority to be separate from the value exported to
> - * user-space. This allows kernel threads to set their
> - * priority to a value higher than any user task. Note:
> - * MAX_RT_PRIO must not be smaller than MAX_USER_RT_PRIO.
> + * Note: MAX_USER_RT_PRIO will be removed as early as 5.8,
> + * don't use it in new code.
> */
>
> -#define MAX_USER_RT_PRIO 100
> -#define MAX_RT_PRIO MAX_USER_RT_PRIO
> +#define MAX_RT_PRIO 100
> +#define MAX_USER_RT_PRIO MAX_RT_PRIO
>
> #define MAX_PRIO (MAX_RT_PRIO + NICE_WIDTH)
> #define DEFAULT_PRIO (MAX_RT_PRIO + NICE_WIDTH / 2)
No one has used it in years, I don't think we need this change. Just delete
it in one go.
Is making p->prio independent from p->mm needed for other changes? If not,
then we can hold off this change until we do so, otherwise, I would keep
your original patch as is, and then remove the extra check when we remove
MAX_USER_RT_PRIO.
-- Steve