Re: [PATCH] Documentation: real-time: Add boot parameters and task isolation section

From: sashiko-bot

Date: Fri Nov 20 2026 - 00:49:48 EST


Thank you for your contribution! Sashiko AI review found 1 potential
issue to consider:
- [Low] docs: real-time: outdated sched_rt_runtime_us default and
throttling behavior
--

commit ae882b9844760c78627ba7cec7ca4d357c476313
Author: Zixin Wang <w1355457260@xxxxxxxxx>

Documentation: real-time: Add boot parameters and task isolation section

This commit adds a section to the real-time kernel configuration guide
documenting boot parameters and runtime tunables for CPU isolation,
nohz_full, RCU callback offloading, interrupt affinity, idle states, and
real-time throttling.

Link: https://bugzilla.kernel.org/show_bug.cgi?id=220633
Link: https://github.com/elisa-tech/wg-aerospace/issues/222

> diff --git a/Documentation/core-api/real-time/kernel-configuration.rst b/Documentation/core-api/real-time/kernel-configuration.rst
> index d7f08e7b87604..6553249646f10 100644
> --- a/Documentation/core-api/real-time/kernel-configuration.rst
> +++ b/Documentation/core-api/real-time/kernel-configuration.rst
> @@ -216,6 +216,140 @@ just by being enabled.

[ ... ]

> +Real-time throttling
> +--------------------
> +
> +``kernel.sched_rt_runtime_us``
> + This sysctl, normally set at boot through a ``sysctl.d``
> + configuration file, limits the CPU time available to real-time
> + scheduling during ``kernel.sched_rt_period_us``. The default of
> + 950000 us per 1000000 us reserves 5% of the CPU for non-real-time
> + tasks. A value of -1 removes the limit, allowing real-time tasks to
> + use the full period; do this only after validating the workload,
> + because a runaway real-time task can then starve the rest of the
> + system. Recent kernels require the runtime/period ratio to stay
> + above 0.05 to preserve bandwidth for the fair deadline server. See
> + Documentation/scheduler/sched-rt-group.rst.

[Severity: Low]
Is this description of kernel.sched_rt_runtime_us up to date?

Since c2e390197ad1 ("sched/rt: Update default bandwidth for real-time tasks
to ONE") and 5f6bd380c7bd ("sched/rt: Remove default bandwidth control"),
sysctl_sched_rt_runtime defaults to 1000000 us rather than 950000 us, and no
longer throttles SCHED_FIFO or SCHED_RR tasks when CONFIG_RT_GROUP_SCHED is
disabled. The default 5% CPU reservation for non-real-time tasks is instead
enforced per CPU by rq->fair_server (50 ms per 1000 ms in
sched_init_dl_servers(), tuned via
/sys/kernel/debug/sched/fair_server/cpuX/runtime), which is not disabled by
setting kernel.sched_rt_runtime_us to -1.

--
Sashiko AI review · https://sashiko.dev/#/patchset/20261120054242.523606-1-w1355457260@xxxxxxxxx?part=1