On Wed, Apr 29, 2026 at 11:01:33PM +0200, Florian Bezdeka wrote: > > Which use case are you actually aiming to support? While dynamic > > reconfiguration would be ideal, the amount of work to get there is > > significant. I won't be signing up for it. > > The use case at hand is a RT enabled platform where the concrete RT > workload is not known at boot time. RT applications are deployed "on- > the-fly", nowadays using the existing container runtimes with some > extended resource management on top. > > Applications can request certain resources like isolated CPU cores, > special IRQ affinities, PCI devices to pass through, ..., so that the > resource management on the system can take care of proper system > configuration.
This is where I really question this use case. Currently, it takes quite a lot of time to tune a system to work properly for RT workloads. Between memory channel interference, GPU interference, and shared transports everywhere, you end up with a fixed split: a set of CPUs suitable for RT work and a set for housekeeping. This partitioning generally does not change during runtime, even if the way you utilize those two sets remains dynamic. Furthermore, reconfiguring a system while running an active RT workload is asking for trouble. I wouldn't be surprised if doing so triggered a wide range of unpredictable side effects.
