Re: [PATCH] sched/fair: fix pick_eevdf to always find the correct se

[Peter Zijlstra]

On Fri, Oct 06, 2023 at 04:39:09PM +0200, Mike Galbraith wrote:
> On Fri, 2023-10-06 at 16:00 +0200, Peter Zijlstra wrote:
> > On Fri, Oct 06, 2023 at 12:31:28PM +0200, Mike Galbraith wrote:
> > > On Fri, 2023-10-06 at 10:35 +0200, Marek Szyprowski wrote:
> > > >
> > > >
> > > > Just to note, I've run into this issue on the QEmu's 'arm64/virt'
> > > > platform, not on the Samsung specific hardware. 
> > >
> > > It doesn't appear to be arch specific, all I have to do is enable
> > > autogroup, raspberry or x86_64 desktop box, the occasional failure
> > > tripping over task groups, leaving both best and best_left with
> > > identical but not what we're looking for ->min_deadline.
> >
> > OK, autogroups enabled and booted, /debug/sched/debug shows autogroups
> > are indeed in existence.
> >
> > I've ran hackbench and a kernel build, but no whoopsie yet :-(
> >
> > I suppose I'll kick some benchmarks and go make a cup 'o tea or
> > something.
> 
> Hm, just booting gets me a handful, and generic desktop activity
> produces a fairly regular supply.  This is virgin 6.6.0.ga9e6eb3-tip.

You're running that systemd thing, eh?

If I create two groups (or two users with autogroup on) and have them
both build a kernel I do indeed get splat.

And yeah, min_deadline is hosed somehow.

migration/28-185     [028] d..2.    70.264274: validate_cfs_rq: --- /
migration/28-185     [028] d..2.    70.264277: __print_se: ffff88845cf48080 w: 1024 ve: -58857638 lag: 870381 vd: -55861854 vmd: -66302085 E (11372/tr)
migration/28-185     [028] d..2.    70.264280: __print_se:   ffff88810d165800 w: 25 ve: -80323686 lag: 22336429 vd: -41496434 vmd: -66302085 E (-1//autogroup-31)
migration/28-185     [028] d..2.    70.264282: __print_se:   ffff888108379000 w: 25 ve: 0 lag: -57987257 vd: 114632828 vmd: 114632828 N (-1//autogroup-33)
migration/28-185     [028] d..2.    70.264283: validate_cfs_rq: min_deadline: -55861854 avg_vruntime: -62278313462 / 1074 = -57987256

I need to go make dinner (kids hungry), but I'll see if I can figure out
how this happens...