Re: [PATCH v2 0/4] sched/fair: cpu.max.runtime and idle refill for CPU credits

[Tejun Heo <tj@kernel.org>]

Hello, Fernand.

On Wed, Nov 11, 2026 at 02:20:39PM +0200, Fernand Sieber wrote:
> The minimal viable bar for us is to never lose track of CPU credits.
...
> The one case it cannot cover is an unplanned exit, a BPF scheduler
> erroring out or killed by the watchdog and falling back to fair, where
> there is no window for userspace to hand the level back.

Where you set your minimum viable bar isn't an argument for a design
decision. Stating that this is the bar doesn't make it so. It has to be
justifiable in a generic manner.

This is a crashing scheduler. The kernel is going to be scanning all the
CPUs and tasks to abort the in-flight scheduler while pushing all tasks to
temporary dumb FIFO scheduling to make forward progress. Accounting wise,
the credits may not be lost, but how meaningful is that across such a
transition? If a machine is doing this regularly to the point where
accurate accounting on that particular transition matters, credit
accounting is the least of its problems.

Thanks.

--
tejun