> Date: Wed, 6 Apr 2016 11:03:29 +0200
> From: Martin Pieuchot <m...@openbsd.org>
>
> On 27/03/16(Sun) 12:00, Martin Pieuchot wrote:
> > Diff below moves resetpriority() out of updatepri() and resched_proc()
> > out of resetpriority() in order to expose a double resched_proc() call
> > in setrunnable().
> >
> > This happens when a thread has slept for more than 1 second. Since
> > p_usrpri is equal or bigger than p_priority, if the first call
> > triggers a need_resched() then the second will also do so.
> >
> > I'd like to commit this diff which does not introduce any behavior
> > change then remove the first resched_proc() in setrunnable().
>
> I'm still looking for oks for this diff.
>
> I'm currently trying to have a better view of the differences in
> voluntary vs involuntary context switches between SP and MP kernels
> for the same workload. So removing a resched_proc() will help.
I am not in favour of this diff simply because it increases the risk
that somebody forgets to call resched_proc() when it is needed. I
don't think the optimization you're talking about will make a
meaningful difference. But I don't feel too strongly about this. So
if there is a consensus that this diff is the way to go, I won't
object.
> > Index: kern/sched_bsd.c
> > ===================================================================
> > RCS file: /cvs/src/sys/kern/sched_bsd.c,v
> > retrieving revision 1.43
> > diff -u -p -r1.43 sched_bsd.c
> > --- kern/sched_bsd.c 9 Mar 2016 13:38:50 -0000 1.43
> > +++ kern/sched_bsd.c 24 Mar 2016 11:37:34 -0000
> > @@ -246,6 +246,7 @@ schedcpu(void *arg)
> > newcpu = (u_int) decay_cpu(loadfac, p->p_estcpu);
> > p->p_estcpu = newcpu;
> > resetpriority(p);
> > + resched_proc(p, p->p_usrpri);
> > if (p->p_priority >= PUSER) {
> > if (p->p_stat == SRUN &&
> > (p->p_priority / SCHED_PPQ) !=
> > @@ -284,7 +285,6 @@ updatepri(struct proc *p)
> > newcpu = (int) decay_cpu(loadfac, newcpu);
> > p->p_estcpu = newcpu;
> > }
> > - resetpriority(p);
> > }
> >
> > /*
> > @@ -454,7 +454,7 @@ mi_switch(void)
> > #endif
> > }
> >
> > -static __inline void
> > +void
> > resched_proc(struct proc *p, u_char pri)
> > {
> > struct cpu_info *ci;
> > @@ -509,8 +509,11 @@ setrunnable(struct proc *p)
> > p->p_stat = SRUN;
> > p->p_cpu = sched_choosecpu(p);
> > setrunqueue(p);
> > - if (p->p_slptime > 1)
> > + if (p->p_slptime > 1) {
> > updatepri(p);
> > + resetpriority(p);
> > + resched_proc(p, p->p_usrpri);
> > + }
> > p->p_slptime = 0;
> > resched_proc(p, p->p_priority);
> > }
> > @@ -531,7 +534,6 @@ resetpriority(struct proc *p)
> > NICE_WEIGHT * (p->p_p->ps_nice - NZERO);
> > newpriority = min(newpriority, MAXPRI);
> > p->p_usrpri = newpriority;
> > - resched_proc(p, p->p_usrpri);
> > }
> >
> > /*
> > @@ -556,6 +558,7 @@ schedclock(struct proc *p)
> > SCHED_LOCK(s);
> > p->p_estcpu = ESTCPULIM(p->p_estcpu + 1);
> > resetpriority(p);
> > + resched_proc(p, p->p_usrpri);
> > if (p->p_priority >= PUSER)
> > p->p_priority = p->p_usrpri;
> > SCHED_UNLOCK(s);
> > Index: kern/kern_resource.c
> > ===================================================================
> > RCS file: /cvs/src/sys/kern/kern_resource.c,v
> > retrieving revision 1.55
> > diff -u -p -r1.55 kern_resource.c
> > --- kern/kern_resource.c 5 Dec 2015 10:11:53 -0000 1.55
> > +++ kern/kern_resource.c 24 Mar 2016 11:30:32 -0000
> > @@ -194,8 +194,10 @@ donice(struct proc *curp, struct process
> > return (EACCES);
> > chgpr->ps_nice = n;
> > SCHED_LOCK(s);
> > - TAILQ_FOREACH(p, &chgpr->ps_threads, p_thr_link)
> > - (void)resetpriority(p);
> > + TAILQ_FOREACH(p, &chgpr->ps_threads, p_thr_link) {
> > + resetpriority(p);
> > + resched_proc(p, p->p_usrpri);
> > + }
> > SCHED_UNLOCK(s);
> > return (0);
> > }
> > Index: sys/sched.h
> > ===================================================================
> > RCS file: /cvs/src/sys/sys/sched.h,v
> > retrieving revision 1.41
> > diff -u -p -r1.41 sched.h
> > --- sys/sched.h 17 Mar 2016 13:18:47 -0000 1.41
> > +++ sys/sched.h 24 Mar 2016 11:35:38 -0000
> > @@ -166,6 +166,7 @@ void sched_stop_secondary_cpus(void);
> > #define cpu_is_idle(ci) ((ci)->ci_schedstate.spc_whichqs == 0)
> >
> > void sched_init_runqueues(void);
> > +void resched_proc(struct proc *p, u_char);
> > void setrunqueue(struct proc *);
> > void remrunqueue(struct proc *);
> >
> >
>
>
