In lppaca region, add a new attribute idle_hint which can allow guest scheduler
for
better cpu selection. Hypervisor can update idle_hint attribute based on
the prediction that if vCPU needs to be scheduled then can it be scheduled
instantly or not.
Signed-off-by: Parth Shah <pa...@linux.ibm.com>
---
arch/powerpc/include/asm/idle_hint.h | 28 +++++++++++++++++++++++
arch/powerpc/include/asm/lppaca.h | 3 ++-
arch/powerpc/kvm/book3s.h | 2 ++
arch/powerpc/kvm/book3s_hv.c | 34 ++++++++++++++++++++++++++++
4 files changed, 66 insertions(+), 1 deletion(-)
create mode 100644 arch/powerpc/include/asm/idle_hint.h
diff --git a/arch/powerpc/include/asm/idle_hint.h
b/arch/powerpc/include/asm/idle_hint.h
new file mode 100644
index 000000000000..165d65c0275b
--- /dev/null
+++ b/arch/powerpc/include/asm/idle_hint.h
@@ -0,0 +1,28 @@
+#ifndef _ASM_POWERPC_IDLEHINT_H
+#define _ASM_POWERPC_IDLEHINT_H
+
+#include <linux/kvm_host.h>
+
+extern void kvmppc_idle_hint_set(struct kvm_vcpu *vcpu, int idle_hint);
+
+extern int idle_hint_is_active;
+
+extern void set_idle_hint(int cpu, int value);
+
+static inline int prev_cpu_of_kvm(struct kvm_vcpu *vcpu)
+{
+ struct pid *pid;
+ struct task_struct *task = NULL;
+
+ rcu_read_lock();
+ pid = rcu_dereference(vcpu->pid);
+ if (pid)
+ task = get_pid_task(pid, PIDTYPE_PID);
+ rcu_read_unlock();
+
+ if (!task)
+ return -1;
+
+ return task_cpu(task);
+}
+#endif
diff --git a/arch/powerpc/include/asm/lppaca.h
b/arch/powerpc/include/asm/lppaca.h
index c390ec377bae..ee790a566036 100644
--- a/arch/powerpc/include/asm/lppaca.h
+++ b/arch/powerpc/include/asm/lppaca.h
@@ -111,7 +111,8 @@ struct lppaca {
__be32 page_ins; /* CMO Hint - # page ins by OS */
u8 reserved11[148];
volatile __be64 dtl_idx; /* Dispatch Trace Log head index */
- u8 reserved12[96];
+ volatile __be32 idle_hint; /* Can vCPU be scheduled instantly? */
+ u8 reserved12[92];
} ____cacheline_aligned;
#define lppaca_of(cpu) (*paca_ptrs[cpu]->lppaca_ptr)
diff --git a/arch/powerpc/kvm/book3s.h b/arch/powerpc/kvm/book3s.h
index 740e51def5a5..61b0741c139a 100644
--- a/arch/powerpc/kvm/book3s.h
+++ b/arch/powerpc/kvm/book3s.h
@@ -7,6 +7,8 @@
#ifndef __POWERPC_KVM_BOOK3S_H__
#define __POWERPC_KVM_BOOK3S_H__
+#include <asm/idle_hint.h>
+
extern void kvmppc_core_flush_memslot_hv(struct kvm *kvm,
struct kvm_memory_slot *memslot);
extern bool kvm_unmap_gfn_range_hv(struct kvm *kvm, struct kvm_gfn_range
*range);
diff --git a/arch/powerpc/kvm/book3s_hv.c b/arch/powerpc/kvm/book3s_hv.c
index bc0813644666..c008be20294d 100644
--- a/arch/powerpc/kvm/book3s_hv.c
+++ b/arch/powerpc/kvm/book3s_hv.c
@@ -447,6 +447,7 @@ static void init_vpa(struct kvm_vcpu *vcpu, struct lppaca
*vpa)
{
vpa->__old_status |= LPPACA_OLD_SHARED_PROC;
vpa->yield_count = cpu_to_be32(1);
+ vpa->idle_hint = cpu_to_be32(0);
}
static int set_vpa(struct kvm_vcpu *vcpu, struct kvmppc_vpa *v,
@@ -911,6 +912,17 @@ static int kvm_arch_vcpu_yield_to(struct kvm_vcpu *target)
return kvm_vcpu_yield_to(target);
}
+void kvmppc_idle_hint_set(struct kvm_vcpu *vcpu, int idle_hint)
+{
+ struct lppaca *lppaca;
+
+ if (!vcpu) return;
+
+ lppaca = (struct lppaca *)vcpu->arch.vpa.pinned_addr;
+ if (lppaca)
+ lppaca->idle_hint = cpu_to_be32(idle_hint);
+}
+
static int kvmppc_get_yield_count(struct kvm_vcpu *vcpu)
{
int yield_count = 0;
@@ -2803,6 +2815,28 @@ static int on_primary_thread(void)
return 1;
}
+void set_idle_hint_for_kvm(struct kvm *kvm, int cpu, int value)
+{
+ int i;
+ struct kvm_vcpu *vcpu;
+
+ kvm_for_each_vcpu(i, vcpu, kvm) {
+ if (cpu == prev_cpu_of_kvm(vcpu)) {
+ kvmppc_idle_hint_set(vcpu, value);
+ }
+ }
+}
+
+void set_idle_hint(int cpu, int value)
+{
+ struct kvm *kvm;
+ struct kvm *tmp;
+
+ list_for_each_entry_safe(kvm, tmp, &vm_list, vm_list) {
+ set_idle_hint_for_kvm(kvm, cpu, value);
+ }
+}
+
/*
* A list of virtual cores for each physical CPU.
* These are vcores that could run but their runner VCPU tasks are
--
2.26.3
