[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <aDW6EckuCFTZfPZ8@jlelli-thinkpadt14gen4.remote.csb>
Date: Tue, 27 May 2025 15:11:45 +0200
From: Juri Lelli <juri.lelli@...hat.com>
To: Shashank Balaji <shashank.mahadasyam@...y.com>
Cc: Jonathan Corbet <corbet@....net>, Peter Zijlstra <peterz@...radead.org>,
linux-doc@...r.kernel.org, linux-kernel@...r.kernel.org,
Shinya Takumi <shinya.takumi@...y.com>
Subject: Re: [PATCH] sched_deadline, docs: add affinity setting with cgroup2
cpuset controller
Hello!
On 22/05/25 17:06, Shashank Balaji wrote:
> Setting the cpu affinity mask of a SCHED_DEADLINE process using the cgroup v1
> cpuset controller is already detailed. Add similar information for cgroup v2's
> cpuset controller.
Thanks a lot for working on this. Really appreaciated.
> Signed-off-by: Shashank Balaji <shashank.mahadasyam@...y.com>
> ---
> Documentation/scheduler/sched-deadline.rst | 29 +++++++++++++++++++++++------
> 1 file changed, 23 insertions(+), 6 deletions(-)
>
> diff --git a/Documentation/scheduler/sched-deadline.rst b/Documentation/scheduler/sched-deadline.rst
> index a727827b8dd52710f880c2b92d3a8224c259873c..e3d7968ff6c3c43f87e249dbcb309619dbd616bf 100644
> --- a/Documentation/scheduler/sched-deadline.rst
> +++ b/Documentation/scheduler/sched-deadline.rst
> @@ -20,7 +20,8 @@ Deadline Task Scheduling
> 4.3 Default behavior
> 4.4 Behavior of sched_yield()
> 5. Tasks CPU affinity
> - 5.1 SCHED_DEADLINE and cpusets HOWTO
> + 5.1 Using cgroup v1 cpuset controller
> + 5.2 Using cgroup v2 cpuset controller
> 6. Future plans
> A. Test suite
> B. Minimal main()
> @@ -671,12 +672,15 @@ Deadline Task Scheduling
> 5. Tasks CPU affinity
> =====================
>
> - -deadline tasks cannot have an affinity mask smaller that the entire
> - root_domain they are created on. However, affinities can be specified
> - through the cpuset facility (Documentation/admin-guide/cgroup-v1/cpusets.rst).
> + Deadline tasks cannot have a cpu affinity mask smaller than the root domain they
> + are created on. So, using ``sched_setaffinity(2)`` won't work. Instead, the
> + the deadline task should be created in a restricted root domain. This can be
> + done using the cpuset controller of either cgroup v1 (deprecated) or cgroup v2.
> + See :ref:`Documentation/admin-guide/cgroup-v1/cpusets.rst <cpusets>` and
> + :ref:`Documentation/admin-guide/cgroup-v2.rst <cgroup-v2>` for more information.
>
> -5.1 SCHED_DEADLINE and cpusets HOWTO
> -------------------------------------
> +5.1 Using cgroup v1 cpuset controller
> +-------------------------------------
>
> An example of a simple configuration (pin a -deadline task to CPU0)
> follows (rt-app is used to create a -deadline task)::
> @@ -695,6 +699,19 @@ Deadline Task Scheduling
> rt-app -t 100000:10000:d:0 -D5 # it is now actually superfluous to specify
> # task affinity
>
> +5.2 Using cgroup v2 cpuset controller
> +-------------------------------------
> +
> + Assuming the cgroup v2 root is mounted at ``/sys/fs/cgroup``.
> +
> + cd /sys/fs/cgroup
> + echo '+cpuset' > cgroup.subtree_control
> + mkdir deadline_group
> + echo 0 > deadline_group/cpuset.cpus
> + echo 'root' > deadline_group/cpuset.cpus.partition
> + echo $$ > deadline_group/cgroup.procs
> + rt-app -t 100000:10000:d:0 -D5
Sadly, the example with cgroup v1 was made at a time when rt-app still
supported command line parameters like the above. I believe nowadays
that is not the case anymore and one needs to create a json file
describing the task to run it with rt-app.
I would say we should update both examples to use something a little
more generic, e.g.,
# chrt --deadline --sched-runtime 10000000 --sched-period 100000000 0 yes >/dev/null &
What do you think?
Best,
Juri
Powered by blists - more mailing lists