[SCHED] any modern replacement for RT_RUNTIME_GREED in the mainline kernel?

[SCHED] any modern replacement for RT_RUNTIME_GREED in the mainline kernel?

[Chris Friesen]

Hi all,

Some time back 
(https://lore.kernel.org/lkml/fa5b1b55d8934c6a0e02e04a7ad6afdf4012c2e0.1478506194.git.bristot@redhat.com/) 
a patch was proposed that would allow high-CPU usage RT tasks to use 
100% of the CPU as long as there were no non-RT tasks that wanted to 
run, but if a non-RT task did want to run the RT-throttling would kick 
in to prevent starvation and deadlocks.

It was rejected for philosophical reasons, with a mention of a 
hypothetical deadline server being the "proper" solution.

Has anything been done in this area in the years since then?

Thanks,
Chris

(please CC me on replies, I'm not subscribed to the list)

Re: [SCHED] any modern replacement for RT_RUNTIME_GREED in the mainline kernel?

[Christian Loehle]

On 12/6/24 18:12, Chris Friesen wrote:
> Hi all,
> 
> Some time back (https://lore.kernel.org/lkml/fa5b1b55d8934c6a0e02e04a7ad6afdf4012c2e0.1478506194.git.bristot@redhat.com/) a patch was proposed that would allow high-CPU usage RT tasks to use 100% of the CPU as long as there were no non-RT tasks that wanted to run, but if a non-RT task did want to run the RT-throttling would kick in to prevent starvation and deadlocks.
> 
> It was rejected for philosophical reasons, with a mention of a hypothetical deadline server being the "proper" solution.
> 
> Has anything been done in this area in the years since then?

There has!
The 'hypothetical deadline server' was implemented and is part
of 6.12
https://lore.kernel.org/lkml/cover.1716811043.git.bristot@kernel.org/