linux-kernel - Re: [PATCH] sched: Document that RT task priorities are 1…99

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Date:   Mon, 17 Jun 2019 14:24:48 +0200
From:   Peter Zijlstra <peterz@...radead.org>
To:     Sebastian Andrzej Siewior <bigeasy@...utronix.de>
Cc:     linux-kernel@...r.kernel.org, Ingo Molnar <mingo@...nel.org>,
        tglx@...utronix.de, John Ogness <john.ogness@...utronix.de>,
        richard@....at
Subject: Re: [PATCH] sched: Document that RT task priorities are 1…99

On Wed, Apr 03, 2019 at 11:08:21PM +0200, Sebastian Andrzej Siewior wrote:
> John identified three files which claim that RT task priorities start at
> zero. As far as I understand, 0 is used for DL and has nothing to do
> wihich RT priorities as identified by the RT policy.
> 
> Correct the comment, valid RT priorities are in the range from 1 to 99.

It all depends on what view I'm afraid. User priorities go from 1-99, as
per sched_get_priority_{min,max}(), but Kernel priority is:

 kernel_prio := MAX_RT_PRIO-1 - user_prio

and that then gives [0-98], where 0 is max and 98 is min (see
__sched_setscheduler() and normal_prio()).

And DL uses kernel prio -1 (there is no user prio equivalent).

Nice maps to:

 kernel_prio := MAX_RT_PRIO + (MAX_NICE - MIN_NICE + 1) / 2 + user_nice

which is: [100-139].

Which then, as derRichard says, leaves (kernel) 99 unaccounted for.


> Reported-by: John Ogness <john.ogness@...utronix.de>
> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@...utronix.de>
> ---
>  Documentation/scheduler/sched-rt-group.txt | 2 +-
>  include/linux/sched/prio.h                 | 2 +-
>  kernel/sched/cpupri.h                      | 2 +-
>  3 files changed, 3 insertions(+), 3 deletions(-)
> 
> diff --git a/Documentation/scheduler/sched-rt-group.txt b/Documentation/scheduler/sched-rt-group.txt
> index d8fce3e784574..23f8f8465a775 100644
> --- a/Documentation/scheduler/sched-rt-group.txt
> +++ b/Documentation/scheduler/sched-rt-group.txt
> @@ -175,7 +175,7 @@ get their allocated time.
>  
>  Implementing SCHED_EDF might take a while to complete. Priority Inheritance is
>  the biggest challenge as the current linux PI infrastructure is geared towards
> -the limited static priority levels 0-99. With deadline scheduling you need to
> +the limited static priority levels 1-99. With deadline scheduling you need to
>  do deadline inheritance (since priority is inversely proportional to the
>  deadline delta (deadline - now)).
>  

This might be correct,.. but I feel we should strive to delete
everything rt groups.

> diff --git a/include/linux/sched/prio.h b/include/linux/sched/prio.h
> index 7d64feafc408e..6986c32356842 100644
> --- a/include/linux/sched/prio.h
> +++ b/include/linux/sched/prio.h
> @@ -8,7 +8,7 @@
>  
>  /*
>   * Priority of a process goes from 0..MAX_PRIO-1, valid RT
> - * priority is 0..MAX_RT_PRIO-1, and SCHED_NORMAL/SCHED_BATCH
> + * priority is 1..MAX_RT_PRIO-1, and SCHED_NORMAL/SCHED_BATCH
>   * tasks are in the range MAX_RT_PRIO..MAX_PRIO-1. Priority
>   * values are inverted: lower p->prio value means higher priority.
>   *

So that comment talks about kernel prio, and there, as we've shown, 0 is
an actual valid RR/FIFO priority (in fact, the highest).

> diff --git a/kernel/sched/cpupri.h b/kernel/sched/cpupri.h
> index 7dc20a3232e72..40257a97fb8f2 100644
> --- a/kernel/sched/cpupri.h
> +++ b/kernel/sched/cpupri.h
> @@ -5,7 +5,7 @@
>  #define CPUPRI_INVALID		-1
>  #define CPUPRI_IDLE		 0
>  #define CPUPRI_NORMAL		 1
> -/* values 2-101 are RT priorities 0-99 */
> +/* values 2-101 are RT priorities 1-99 */

Again, this is kernel prios, not user prios.

>  struct cpupri_vec {
>  	atomic_t		count;
> -- 
> 2.20.1
>