Re: [PATCH 0/4] sched/rt: Distribute tasks in find_lowest_rq()

[Marc Zyngier]

On 2020-04-21 15:22, Qais Yousef wrote:
> On 04/21/20 15:09, Marc Zyngier wrote:
> > On 2020-04-21 14:18, Valentin Schneider wrote:
> > > On 21/04/20 13:13, Qais Yousef wrote:
> >
> > [...]
> >
> > > > I CCed Marc who's the maintainer of this file who can clarify better
> > > > if this
> > > > really breaks anything.
> > > >
> > > > If any interrupt expects to be affined to a specific CPU then this
> > > > must be
> > > > described in DT/driver. I think the GIC controller is free to
> > > > distribute them
> > > > to any cpu otherwise if !force. Which is usually done by
> > > > irq_balancer anyway
> > > > in userspace, IIUC.
> > > >
> > > > I don't see how cpumask_any_and() break anything here too. I
> > > > actually think it
> > > > improves on things by better distribute the irqs on the system by
> > > > default.
> >
> > That's a pretty bold statement. Unfortunately, it isn't universally 
> > true.
> > Some workload will be very happy with interrupts spread all over the 
> > map,
> > and some others will suffer from it because, well, it interrupts 
> > userspace.
> >
> > > As you say, if someone wants smarter IRQ affinity they can do
> > > irq_balancer
> > > and whatnot. The default kernel policy for now has been to shove
> > > everything
> > > on the lowest-numbered CPU, and I see no valid reason to change that.
> >
> > Exactly. I would like to keep the kernel policy as simple as possible 
> > for
> > non-managed interrupts (managed interrupts are another kettle of fish
> > entirely).
> > Userpace is in control to place things "intelligently", so let's not 
> > try and
> > make the kernel smarter than it strictly needs to be.
>
> Fair enough. But why is it asking for cpumask_any() in the first place?

Implementation detail. Turn it into cpumask_first_and() if you want.

        M.
--
Jazz is not dead. It just smells funny...