| Message ID | 20190809145833.1020-25-jgross@suse.com (mailing list archive) |
|---|---|
| State | Superseded |
| Headers | show |
| Series | xen: add core scheduling support | expand |
On 09.08.2019 16:58, Juergen Gross wrote: > @@ -504,22 +511,21 @@ int sched_move_domain(struct domain *d, struct cpupool *c) > if ( IS_ERR(domdata) ) > return PTR_ERR(domdata); > > - vcpu_priv = xzalloc_array(void *, d->max_vcpus); > - if ( vcpu_priv == NULL ) > + unit_priv = xzalloc_array(void *, d->max_vcpus); I find it confusing that an array of units (as per the use below) is dimensioned by the domain's vCPU count. Isn't there a correlation between vCPU IDs and units IDs, perhaps along the lines of CPU APIC (thread), core, and socket IDs? If so, the array size could be bounded here by a smaller (down the road) value. > @@ -880,18 +889,36 @@ void vcpu_force_reschedule(struct vcpu *v) > vcpu_migrate_finish(v); > } > > +static bool sched_check_affinity_broken(struct sched_unit *unit) const > +{ > + struct vcpu *v; const > @@ -910,18 +937,20 @@ void restore_vcpu_affinity(struct domain *d) > cpupool_domain_cpumask(d)); > if ( cpumask_empty(cpumask_scratch_cpu(cpu)) ) > { > - if ( v->affinity_broken ) > + if ( sched_check_affinity_broken(unit) ) > { > - sched_set_affinity(v, unit->cpu_hard_affinity_saved, NULL); > - v->affinity_broken = 0; > + sched_set_affinity(unit->vcpu_list, > + unit->cpu_hard_affinity_saved, NULL); > + sched_reset_affinity_broken(unit); > cpumask_and(cpumask_scratch_cpu(cpu), unit->cpu_hard_affinity, > cpupool_domain_cpumask(d)); > } > > if ( cpumask_empty(cpumask_scratch_cpu(cpu)) ) > { > - printk(XENLOG_DEBUG "Breaking affinity for %pv\n", v); > - sched_set_affinity(v, &cpumask_all, NULL); > + printk(XENLOG_DEBUG "Breaking affinity for %pv\n", > + unit->vcpu_list); > + sched_set_affinity(unit->vcpu_list, &cpumask_all, NULL); > cpumask_and(cpumask_scratch_cpu(cpu), unit->cpu_hard_affinity, > cpupool_domain_cpumask(d)); > } >[...]> @@ -964,17 +992,18 @@ int cpu_disable_scheduler(unsigned int cpu) > > for_each_domain_in_cpupool ( d, c ) > { > - for_each_vcpu ( d, v ) > + struct sched_unit *unit; > + > + for_each_sched_unit ( d, unit ) > { > unsigned long flags; > - struct sched_unit *unit = v->sched_unit; > spinlock_t *lock = unit_schedule_lock_irqsave(unit, &flags); > > cpumask_and(&online_affinity, unit->cpu_hard_affinity, c->cpu_valid); > if ( cpumask_empty(&online_affinity) && > cpumask_test_cpu(cpu, unit->cpu_hard_affinity) ) > { > - if ( v->affinity_broken ) > + if ( unit->vcpu_list->affinity_broken ) > { > /* The vcpu is temporarily pinned, can't move it. */ > unit_schedule_unlock_irqrestore(lock, flags, unit); > @@ -982,14 +1011,15 @@ int cpu_disable_scheduler(unsigned int cpu) > break; > } > > - printk(XENLOG_DEBUG "Breaking affinity for %pv\n", v); > + printk(XENLOG_DEBUG "Breaking affinity for %pv\n", > + unit->vcpu_list); > > - sched_set_affinity(v, &cpumask_all, NULL); > + sched_set_affinity(unit->vcpu_list, &cpumask_all, NULL); > } > > - if ( v->processor != cpu ) > + if ( sched_unit_cpu(unit) != sched_get_resource_cpu(cpu) ) > { > - /* The vcpu is not on this cpu, so we can move on. */ > + /* The unit is not on this cpu, so we can move on. */ > unit_schedule_unlock_irqrestore(lock, flags, unit); > continue; > } > @@ -1002,17 +1032,17 @@ int cpu_disable_scheduler(unsigned int cpu) > * * the scheduler will always find a suitable solution, or > * things would have failed before getting in here. > */ > - vcpu_migrate_start(v); > + vcpu_migrate_start(unit->vcpu_list); > unit_schedule_unlock_irqrestore(lock, flags, unit); > > - vcpu_migrate_finish(v); > + vcpu_migrate_finish(unit->vcpu_list); All the ->vcpu_list references look bogus considering where you're moving, but I can only guess that all of this will need touching again later in the series. I wonder though whether these wouldn't better change into for-each-vCPU-in-unit loops right away. > /* > * The only caveat, in this case, is that if a vcpu active in > * the hypervisor isn't migratable. In this case, the caller > * should try again after releasing and reaquiring all locks. > */ > - if ( v->processor == cpu ) > + if ( sched_unit_cpu(unit) == sched_get_resource_cpu(cpu) ) Is comparing the (pseudo) CPU values here the most efficient approach generated code wise? Can't there be some pointer comparison that's cheaper? > @@ -1023,8 +1053,8 @@ int cpu_disable_scheduler(unsigned int cpu) > static int cpu_disable_scheduler_check(unsigned int cpu) > { > struct domain *d; > - struct vcpu *v; > struct cpupool *c; > + struct vcpu *v; > > c = per_cpu(cpupool, cpu); > if ( c == NULL ) Stray change? Jan
On 09.09.19 17:14, Jan Beulich wrote: > On 09.08.2019 16:58, Juergen Gross wrote: >> @@ -504,22 +511,21 @@ int sched_move_domain(struct domain *d, struct cpupool *c) >> if ( IS_ERR(domdata) ) >> return PTR_ERR(domdata); >> >> - vcpu_priv = xzalloc_array(void *, d->max_vcpus); >> - if ( vcpu_priv == NULL ) >> + unit_priv = xzalloc_array(void *, d->max_vcpus); > > I find it confusing that an array of units (as per the use below) > is dimensioned by the domain's vCPU count. Isn't there a correlation > between vCPU IDs and units IDs, perhaps along the lines of CPU > APIC (thread), core, and socket IDs? If so, the array size could > be bounded here by a smaller (down the road) value. I'll add a comment in this regard in this patch and when the number of vcpus per unit gets added (patch 31) I'll modify the array size. > >> @@ -880,18 +889,36 @@ void vcpu_force_reschedule(struct vcpu *v) >> vcpu_migrate_finish(v); >> } >> >> +static bool sched_check_affinity_broken(struct sched_unit *unit) > > const Okay. > >> +{ >> + struct vcpu *v; > > const Okay. > >> @@ -910,18 +937,20 @@ void restore_vcpu_affinity(struct domain *d) >> cpupool_domain_cpumask(d)); >> if ( cpumask_empty(cpumask_scratch_cpu(cpu)) ) >> { >> - if ( v->affinity_broken ) >> + if ( sched_check_affinity_broken(unit) ) >> { >> - sched_set_affinity(v, unit->cpu_hard_affinity_saved, NULL); >> - v->affinity_broken = 0; >> + sched_set_affinity(unit->vcpu_list, >> + unit->cpu_hard_affinity_saved, NULL); >> + sched_reset_affinity_broken(unit); >> cpumask_and(cpumask_scratch_cpu(cpu), unit->cpu_hard_affinity, >> cpupool_domain_cpumask(d)); >> } >> >> if ( cpumask_empty(cpumask_scratch_cpu(cpu)) ) >> { >> - printk(XENLOG_DEBUG "Breaking affinity for %pv\n", v); >> - sched_set_affinity(v, &cpumask_all, NULL); >> + printk(XENLOG_DEBUG "Breaking affinity for %pv\n", >> + unit->vcpu_list); >> + sched_set_affinity(unit->vcpu_list, &cpumask_all, NULL); >> cpumask_and(cpumask_scratch_cpu(cpu), unit->cpu_hard_affinity, >> cpupool_domain_cpumask(d)); >> } >> [...]> @@ -964,17 +992,18 @@ int cpu_disable_scheduler(unsigned int cpu) >> >> for_each_domain_in_cpupool ( d, c ) >> { >> - for_each_vcpu ( d, v ) >> + struct sched_unit *unit; >> + >> + for_each_sched_unit ( d, unit ) >> { >> unsigned long flags; >> - struct sched_unit *unit = v->sched_unit; >> spinlock_t *lock = unit_schedule_lock_irqsave(unit, &flags); >> >> cpumask_and(&online_affinity, unit->cpu_hard_affinity, c->cpu_valid); >> if ( cpumask_empty(&online_affinity) && >> cpumask_test_cpu(cpu, unit->cpu_hard_affinity) ) >> { >> - if ( v->affinity_broken ) >> + if ( unit->vcpu_list->affinity_broken ) >> { >> /* The vcpu is temporarily pinned, can't move it. */ >> unit_schedule_unlock_irqrestore(lock, flags, unit); >> @@ -982,14 +1011,15 @@ int cpu_disable_scheduler(unsigned int cpu) >> break; >> } >> >> - printk(XENLOG_DEBUG "Breaking affinity for %pv\n", v); >> + printk(XENLOG_DEBUG "Breaking affinity for %pv\n", >> + unit->vcpu_list); >> >> - sched_set_affinity(v, &cpumask_all, NULL); >> + sched_set_affinity(unit->vcpu_list, &cpumask_all, NULL); >> } >> >> - if ( v->processor != cpu ) >> + if ( sched_unit_cpu(unit) != sched_get_resource_cpu(cpu) ) >> { >> - /* The vcpu is not on this cpu, so we can move on. */ >> + /* The unit is not on this cpu, so we can move on. */ >> unit_schedule_unlock_irqrestore(lock, flags, unit); >> continue; >> } >> @@ -1002,17 +1032,17 @@ int cpu_disable_scheduler(unsigned int cpu) >> * * the scheduler will always find a suitable solution, or >> * things would have failed before getting in here. >> */ >> - vcpu_migrate_start(v); >> + vcpu_migrate_start(unit->vcpu_list); >> unit_schedule_unlock_irqrestore(lock, flags, unit); >> >> - vcpu_migrate_finish(v); >> + vcpu_migrate_finish(unit->vcpu_list); > > All the ->vcpu_list references look bogus considering where you're > moving, but I can only guess that all of this will need touching > again later in the series. I wonder though whether these wouldn't > better change into for-each-vCPU-in-unit loops right away. Especially the vcpu_migrate part is more complicated. I think it is much easier to review with the more mechanical changes split from the logical changes. > >> /* >> * The only caveat, in this case, is that if a vcpu active in >> * the hypervisor isn't migratable. In this case, the caller >> * should try again after releasing and reaquiring all locks. >> */ >> - if ( v->processor == cpu ) >> + if ( sched_unit_cpu(unit) == sched_get_resource_cpu(cpu) ) > > Is comparing the (pseudo) CPU values here the most efficient approach > generated code wise? Can't there be some pointer comparison that's > cheaper? Yes, you are right. unit->res == get_sched_res(cpu) is equivalent. > >> @@ -1023,8 +1053,8 @@ int cpu_disable_scheduler(unsigned int cpu) >> static int cpu_disable_scheduler_check(unsigned int cpu) >> { >> struct domain *d; >> - struct vcpu *v; >> struct cpupool *c; >> + struct vcpu *v; >> >> c = per_cpu(cpupool, cpu); >> if ( c == NULL ) > > Stray change? Yes. Juergen
On 12.09.2019 16:02, Juergen Gross wrote: > On 09.09.19 17:14, Jan Beulich wrote: >> On 09.08.2019 16:58, Juergen Gross wrote: >>> @@ -1002,17 +1032,17 @@ int cpu_disable_scheduler(unsigned int cpu) >>> * * the scheduler will always find a suitable solution, or >>> * things would have failed before getting in here. >>> */ >>> - vcpu_migrate_start(v); >>> + vcpu_migrate_start(unit->vcpu_list); >>> unit_schedule_unlock_irqrestore(lock, flags, unit); >>> >>> - vcpu_migrate_finish(v); >>> + vcpu_migrate_finish(unit->vcpu_list); >> >> All the ->vcpu_list references look bogus considering where you're >> moving, but I can only guess that all of this will need touching >> again later in the series. I wonder though whether these wouldn't >> better change into for-each-vCPU-in-unit loops right away. > > Especially the vcpu_migrate part is more complicated. I think it is > much easier to review with the more mechanical changes split from the > logical changes. Yes, and I appreciate you separating mechanical from logical changes. However, as already pointed out in the context where I had convinced you of using "vcpu_list" as a name, individual actions on vcpu_list now look bogus throughout the series. They should really (almost?) all be loops over the entire list; I have a hard time imagining possible exceptions, but I'm not going to exclude there may be one or a few. Introducing such loops should, as long as there's only ever on vCPU on such a list, too be a mostly mechanical step, which imo should have happened before (or with) changes like this one. Jan
On 12.09.19 16:40, Jan Beulich wrote: > On 12.09.2019 16:02, Juergen Gross wrote: >> On 09.09.19 17:14, Jan Beulich wrote: >>> On 09.08.2019 16:58, Juergen Gross wrote: >>>> @@ -1002,17 +1032,17 @@ int cpu_disable_scheduler(unsigned int cpu) >>>> * * the scheduler will always find a suitable solution, or >>>> * things would have failed before getting in here. >>>> */ >>>> - vcpu_migrate_start(v); >>>> + vcpu_migrate_start(unit->vcpu_list); >>>> unit_schedule_unlock_irqrestore(lock, flags, unit); >>>> >>>> - vcpu_migrate_finish(v); >>>> + vcpu_migrate_finish(unit->vcpu_list); >>> >>> All the ->vcpu_list references look bogus considering where you're >>> moving, but I can only guess that all of this will need touching >>> again later in the series. I wonder though whether these wouldn't >>> better change into for-each-vCPU-in-unit loops right away. >> >> Especially the vcpu_migrate part is more complicated. I think it is >> much easier to review with the more mechanical changes split from the >> logical changes. > > Yes, and I appreciate you separating mechanical from logical changes. > However, as already pointed out in the context where I had convinced > you of using "vcpu_list" as a name, individual actions on vcpu_list > now look bogus throughout the series. They should really (almost?) > all be loops over the entire list; I have a hard time imagining > possible exceptions, but I'm not going to exclude there may be one > or a few. Introducing such loops should, as long as there's only > ever on vCPU on such a list, too be a mostly mechanical step, which > imo should have happened before (or with) changes like this one. I think the easiest way to handle that is to add a comment like: /* TODO: switch to for_each_sched_unit_vcpu() */ This will show the need for the loop without having to do logic changes. Juergen
diff --git a/xen/common/domain.c b/xen/common/domain.c index 19d881b0c3..91b01c220e 100644 --- a/xen/common/domain.c +++ b/xen/common/domain.c @@ -550,7 +550,7 @@ void domain_update_node_affinity(struct domain *d) cpumask_var_t dom_cpumask, dom_cpumask_soft; cpumask_t *dom_affinity; const cpumask_t *online; - struct vcpu *v; + struct sched_unit *unit; unsigned int cpu; /* Do we have vcpus already? If not, no need to update node-affinity. */ @@ -583,12 +583,11 @@ void domain_update_node_affinity(struct domain *d) * and the full mask of where it would prefer to run (the union of * the soft affinity of all its various vcpus). Let's build them. */ - for_each_vcpu ( d, v ) + for_each_sched_unit ( d, unit ) { - cpumask_or(dom_cpumask, dom_cpumask, - v->sched_unit->cpu_hard_affinity); + cpumask_or(dom_cpumask, dom_cpumask, unit->cpu_hard_affinity); cpumask_or(dom_cpumask_soft, dom_cpumask_soft, - v->sched_unit->cpu_soft_affinity); + unit->cpu_soft_affinity); } /* Filter out non-online cpus */ cpumask_and(dom_cpumask, dom_cpumask, online); diff --git a/xen/common/schedule.c b/xen/common/schedule.c index fcd083d31d..7b0ff8395e 100644 --- a/xen/common/schedule.c +++ b/xen/common/schedule.c @@ -147,26 +147,32 @@ static inline struct scheduler *dom_scheduler(const struct domain *d) return &ops; } -static inline struct scheduler *vcpu_scheduler(const struct vcpu *v) +static inline struct scheduler *unit_scheduler(const struct sched_unit *unit) { - struct domain *d = v->domain; + struct domain *d = unit->domain; if ( likely(d->cpupool != NULL) ) return d->cpupool->sched; /* - * If d->cpupool is NULL, this is a vCPU of the idle domain. And this + * If d->cpupool is NULL, this is a unit of the idle domain. And this * case is special because the idle domain does not really belong to * a cpupool and, hence, doesn't really have a scheduler). In fact, its - * vCPUs (may) run on pCPUs which are in different pools, with different + * units (may) run on pCPUs which are in different pools, with different * schedulers. * * What we want, in this case, is the scheduler of the pCPU where this - * particular idle vCPU is running. And, since v->processor never changes - * for idle vCPUs, it is safe to use it, with no locks, to figure that out. + * particular idle unit is running. And, since unit->res never changes + * for idle units, it is safe to use it, with no locks, to figure that out. */ + ASSERT(is_idle_domain(d)); - return per_cpu(scheduler, v->processor); + return per_cpu(scheduler, unit->res->processor); +} + +static inline struct scheduler *vcpu_scheduler(const struct vcpu *v) +{ + return unit_scheduler(v->sched_unit); } #define VCPU2ONLINE(_v) cpupool_domain_cpumask((_v)->domain) @@ -487,10 +493,11 @@ static void sched_move_irqs(struct sched_unit *unit) int sched_move_domain(struct domain *d, struct cpupool *c) { struct vcpu *v; + struct sched_unit *unit; unsigned int new_p; - void **vcpu_priv; + void **unit_priv; void *domdata; - void *vcpudata; + void *unitdata; struct scheduler *old_ops; void *old_domdata; @@ -504,22 +511,21 @@ int sched_move_domain(struct domain *d, struct cpupool *c) if ( IS_ERR(domdata) ) return PTR_ERR(domdata); - vcpu_priv = xzalloc_array(void *, d->max_vcpus); - if ( vcpu_priv == NULL ) + unit_priv = xzalloc_array(void *, d->max_vcpus); + if ( unit_priv == NULL ) { sched_free_domdata(c->sched, domdata); return -ENOMEM; } - for_each_vcpu ( d, v ) + for_each_sched_unit ( d, unit ) { - vcpu_priv[v->vcpu_id] = sched_alloc_vdata(c->sched, v->sched_unit, - domdata); - if ( vcpu_priv[v->vcpu_id] == NULL ) + unit_priv[unit->unit_id] = sched_alloc_vdata(c->sched, unit, domdata); + if ( unit_priv[unit->unit_id] == NULL ) { - for_each_vcpu ( d, v ) - xfree(vcpu_priv[v->vcpu_id]); - xfree(vcpu_priv); + for_each_sched_unit ( d, unit ) + xfree(unit_priv[unit->unit_id]); + xfree(unit_priv); sched_free_domdata(c->sched, domdata); return -ENOMEM; } @@ -530,30 +536,35 @@ int sched_move_domain(struct domain *d, struct cpupool *c) old_ops = dom_scheduler(d); old_domdata = d->sched_priv; - for_each_vcpu ( d, v ) + for_each_sched_unit ( d, unit ) { - sched_remove_unit(old_ops, v->sched_unit); + sched_remove_unit(old_ops, unit); } d->cpupool = c; d->sched_priv = domdata; new_p = cpumask_first(c->cpu_valid); - for_each_vcpu ( d, v ) + for_each_sched_unit ( d, unit ) { spinlock_t *lock; + unsigned int unit_p = new_p; - vcpudata = v->sched_unit->priv; + unitdata = unit->priv; - migrate_timer(&v->periodic_timer, new_p); - migrate_timer(&v->singleshot_timer, new_p); - migrate_timer(&v->poll_timer, new_p); + for_each_sched_unit_vcpu ( unit, v ) + { + migrate_timer(&v->periodic_timer, new_p); + migrate_timer(&v->singleshot_timer, new_p); + migrate_timer(&v->poll_timer, new_p); + new_p = cpumask_cycle(new_p, c->cpu_valid); + } - lock = unit_schedule_lock_irq(v->sched_unit); + lock = unit_schedule_lock_irq(unit); - sched_set_affinity(v, &cpumask_all, &cpumask_all); + sched_set_affinity(unit->vcpu_list, &cpumask_all, &cpumask_all); - sched_set_res(v->sched_unit, get_sched_res(new_p)); + sched_set_res(unit, get_sched_res(unit_p)); /* * With v->processor modified we must not * - make any further changes assuming we hold the scheduler lock, @@ -561,15 +572,13 @@ int sched_move_domain(struct domain *d, struct cpupool *c) */ spin_unlock_irq(lock); - v->sched_unit->priv = vcpu_priv[v->vcpu_id]; + unit->priv = unit_priv[unit->unit_id]; if ( !d->is_dying ) - sched_move_irqs(v->sched_unit); - - new_p = cpumask_cycle(new_p, c->cpu_valid); + sched_move_irqs(unit); - sched_insert_unit(c->sched, v->sched_unit); + sched_insert_unit(c->sched, unit); - sched_free_vdata(old_ops, vcpudata); + sched_free_vdata(old_ops, unitdata); } domain_update_node_affinity(d); @@ -578,7 +587,7 @@ int sched_move_domain(struct domain *d, struct cpupool *c) sched_free_domdata(old_ops, old_domdata); - xfree(vcpu_priv); + xfree(unit_priv); return 0; } @@ -880,18 +889,36 @@ void vcpu_force_reschedule(struct vcpu *v) vcpu_migrate_finish(v); } +static bool sched_check_affinity_broken(struct sched_unit *unit) +{ + struct vcpu *v; + + for_each_sched_unit_vcpu ( unit, v ) + if ( v->affinity_broken ) + return true; + + return false; +} + +static void sched_reset_affinity_broken(struct sched_unit *unit) +{ + struct vcpu *v; + + for_each_sched_unit_vcpu ( unit, v ) + v->affinity_broken = false; +} + void restore_vcpu_affinity(struct domain *d) { unsigned int cpu = smp_processor_id(); - struct vcpu *v; + struct sched_unit *unit; ASSERT(system_state == SYS_STATE_resume); - for_each_vcpu ( d, v ) + for_each_sched_unit ( d, unit ) { spinlock_t *lock; - unsigned int old_cpu = v->processor; - struct sched_unit *unit = v->sched_unit; + unsigned int old_cpu = sched_unit_cpu(unit); struct sched_resource *res; ASSERT(!unit_runnable(unit)); @@ -910,18 +937,20 @@ void restore_vcpu_affinity(struct domain *d) cpupool_domain_cpumask(d)); if ( cpumask_empty(cpumask_scratch_cpu(cpu)) ) { - if ( v->affinity_broken ) + if ( sched_check_affinity_broken(unit) ) { - sched_set_affinity(v, unit->cpu_hard_affinity_saved, NULL); - v->affinity_broken = 0; + sched_set_affinity(unit->vcpu_list, + unit->cpu_hard_affinity_saved, NULL); + sched_reset_affinity_broken(unit); cpumask_and(cpumask_scratch_cpu(cpu), unit->cpu_hard_affinity, cpupool_domain_cpumask(d)); } if ( cpumask_empty(cpumask_scratch_cpu(cpu)) ) { - printk(XENLOG_DEBUG "Breaking affinity for %pv\n", v); - sched_set_affinity(v, &cpumask_all, NULL); + printk(XENLOG_DEBUG "Breaking affinity for %pv\n", + unit->vcpu_list); + sched_set_affinity(unit->vcpu_list, &cpumask_all, NULL); cpumask_and(cpumask_scratch_cpu(cpu), unit->cpu_hard_affinity, cpupool_domain_cpumask(d)); } @@ -934,12 +963,12 @@ void restore_vcpu_affinity(struct domain *d) /* v->processor might have changed, so reacquire the lock. */ lock = unit_schedule_lock_irq(unit); - res = sched_pick_resource(vcpu_scheduler(v), unit); + res = sched_pick_resource(unit_scheduler(unit), unit); sched_set_res(unit, res); spin_unlock_irq(lock); - if ( old_cpu != v->processor ) - sched_move_irqs(v->sched_unit); + if ( old_cpu != sched_unit_cpu(unit) ) + sched_move_irqs(unit); } domain_update_node_affinity(d); @@ -953,7 +982,6 @@ void restore_vcpu_affinity(struct domain *d) int cpu_disable_scheduler(unsigned int cpu) { struct domain *d; - struct vcpu *v; struct cpupool *c; cpumask_t online_affinity; int ret = 0; @@ -964,17 +992,18 @@ int cpu_disable_scheduler(unsigned int cpu) for_each_domain_in_cpupool ( d, c ) { - for_each_vcpu ( d, v ) + struct sched_unit *unit; + + for_each_sched_unit ( d, unit ) { unsigned long flags; - struct sched_unit *unit = v->sched_unit; spinlock_t *lock = unit_schedule_lock_irqsave(unit, &flags); cpumask_and(&online_affinity, unit->cpu_hard_affinity, c->cpu_valid); if ( cpumask_empty(&online_affinity) && cpumask_test_cpu(cpu, unit->cpu_hard_affinity) ) { - if ( v->affinity_broken ) + if ( unit->vcpu_list->affinity_broken ) { /* The vcpu is temporarily pinned, can't move it. */ unit_schedule_unlock_irqrestore(lock, flags, unit); @@ -982,14 +1011,15 @@ int cpu_disable_scheduler(unsigned int cpu) break; } - printk(XENLOG_DEBUG "Breaking affinity for %pv\n", v); + printk(XENLOG_DEBUG "Breaking affinity for %pv\n", + unit->vcpu_list); - sched_set_affinity(v, &cpumask_all, NULL); + sched_set_affinity(unit->vcpu_list, &cpumask_all, NULL); } - if ( v->processor != cpu ) + if ( sched_unit_cpu(unit) != sched_get_resource_cpu(cpu) ) { - /* The vcpu is not on this cpu, so we can move on. */ + /* The unit is not on this cpu, so we can move on. */ unit_schedule_unlock_irqrestore(lock, flags, unit); continue; } @@ -1002,17 +1032,17 @@ int cpu_disable_scheduler(unsigned int cpu) * * the scheduler will always find a suitable solution, or * things would have failed before getting in here. */ - vcpu_migrate_start(v); + vcpu_migrate_start(unit->vcpu_list); unit_schedule_unlock_irqrestore(lock, flags, unit); - vcpu_migrate_finish(v); + vcpu_migrate_finish(unit->vcpu_list); /* * The only caveat, in this case, is that if a vcpu active in * the hypervisor isn't migratable. In this case, the caller * should try again after releasing and reaquiring all locks. */ - if ( v->processor == cpu ) + if ( sched_unit_cpu(unit) == sched_get_resource_cpu(cpu) ) ret = -EAGAIN; } } @@ -1023,8 +1053,8 @@ int cpu_disable_scheduler(unsigned int cpu) static int cpu_disable_scheduler_check(unsigned int cpu) { struct domain *d; - struct vcpu *v; struct cpupool *c; + struct vcpu *v; c = per_cpu(cpupool, cpu); if ( c == NULL )
Where appropriate switch from for_each_vcpu() to for_each_sched_unit() in order to prepare core scheduling. As it is beneficial once here and for sure in future add a unit_scheduler() helper and let vcpu_scheduler() use it. Signed-off-by: Juergen Gross <jgross@suse.com> --- V2: - handle affinity_broken correctly (Jan Beulich) - add unit_scheduler() (Jan Beulich) --- xen/common/domain.c | 9 ++- xen/common/schedule.c | 148 ++++++++++++++++++++++++++++++-------------------- 2 files changed, 93 insertions(+), 64 deletions(-)