KVM: PPC: Book3S HV P9: Avoid cpu_in_guest atomics on entry and exit

cpu_in_guest is set to determine if a CPU needs to be IPI'ed to exit
the guest and notice the need_tlb_flush bit.

This can be implemented as a global per-CPU pointer to the currently
running guest instead of per-guest cpumasks, saving 2 atomics per
entry/exit. P7/8 doesn't require cpu_in_guest, nor does a nested HV
(only the L0 does), so move it to the P9 HV path.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20211123095231.1036501-50-npiggin@gmail.com

arch/powerpc/include/asm/kvm_book3s_64.h

@@ -44,7 +44,6 @@ struct kvm_nested_guest {
 	struct mutex tlb_lock;		/* serialize page faults and tlbies */
 	struct kvm_nested_guest *next;
 	cpumask_t need_tlb_flush;
-	cpumask_t cpu_in_guest;
 	short prev_cpu[NR_CPUS];
 	u8 radix;			/* is this nested guest radix */
 };

arch/powerpc/include/asm/kvm_host.h

@@ -287,7 +287,6 @@ struct kvm_arch {
 	u32 online_vcores;
 	atomic_t hpte_mod_interest;
 	cpumask_t need_tlb_flush;
-	cpumask_t cpu_in_guest;
 	u8 radix;
 	u8 fwnmi_enabled;
 	u8 secure_guest;

arch/powerpc/kvm/book3s_hv.c

@@ -3009,19 +3009,18 @@ static void kvmppc_release_hwthread(int cpu)
 	tpaca->kvm_hstate.kvm_split_mode = NULL;
 }
 
+static DEFINE_PER_CPU(struct kvm *, cpu_in_guest);
+
 static void radix_flush_cpu(struct kvm *kvm, int cpu, struct kvm_vcpu *vcpu)
 {
 	struct kvm_nested_guest *nested = vcpu->arch.nested;
-	cpumask_t *cpu_in_guest, *need_tlb_flush;
+	cpumask_t *need_tlb_flush;
 	int i;
 
-	if (nested) {
+	if (nested)
 		need_tlb_flush = &nested->need_tlb_flush;
-		cpu_in_guest = &nested->cpu_in_guest;
-	} else {
+	else
 		need_tlb_flush = &kvm->arch.need_tlb_flush;
-		cpu_in_guest = &kvm->arch.cpu_in_guest;
-	}
 
 	cpu = cpu_first_tlb_thread_sibling(cpu);
 	for (i = cpu; i <= cpu_last_tlb_thread_sibling(cpu);
@@ -3029,16 +3028,21 @@ static void radix_flush_cpu(struct kvm *kvm, int cpu, struct kvm_vcpu *vcpu)
 		cpumask_set_cpu(i, need_tlb_flush);
 
 	/*
-	 * Make sure setting of bit in need_tlb_flush precedes
-	 * testing of cpu_in_guest bits.  The matching barrier on
-	 * the other side is the first smp_mb() in kvmppc_run_core().
+	 * Make sure setting of bit in need_tlb_flush precedes testing of
+	 * cpu_in_guest. The matching barrier on the other side is hwsync
+	 * when switching to guest MMU mode, which happens between
+	 * cpu_in_guest being set to the guest kvm, and need_tlb_flush bit
+	 * being tested.
 	 */
 	smp_mb();
 
 	for (i = cpu; i <= cpu_last_tlb_thread_sibling(cpu);
-					i += cpu_tlb_thread_sibling_step())
-		if (cpumask_test_cpu(i, cpu_in_guest))
+					i += cpu_tlb_thread_sibling_step()) {
+		struct kvm *running = *per_cpu_ptr(&cpu_in_guest, i);
+
+		if (running == kvm)
 			smp_call_function_single(i, do_nothing, NULL, 1);
+	}
 }
 
 static void do_migrate_away_vcpu(void *arg)
@@ -3105,7 +3109,6 @@ static void kvmppc_start_thread(struct kvm_vcpu *vcpu, struct kvmppc_vcore *vc)
 {
 	int cpu;
 	struct paca_struct *tpaca;
-	struct kvm *kvm = vc->kvm;
 
 	cpu = vc->pcpu;
 	if (vcpu) {
@@ -3116,7 +3119,6 @@ static void kvmppc_start_thread(struct kvm_vcpu *vcpu, struct kvmppc_vcore *vc)
 		cpu += vcpu->arch.ptid;
 		vcpu->cpu = vc->pcpu;
 		vcpu->arch.thread_cpu = cpu;
-		cpumask_set_cpu(cpu, &kvm->arch.cpu_in_guest);
 	}
 	tpaca = paca_ptrs[cpu];
 	tpaca->kvm_hstate.kvm_vcpu = vcpu;
@@ -3847,7 +3849,6 @@ static noinline void kvmppc_run_core(struct kvmppc_vcore *vc)
 		kvmppc_release_hwthread(pcpu + i);
 		if (sip && sip->napped[i])
 			kvmppc_ipi_thread(pcpu + i);
-		cpumask_clear_cpu(pcpu + i, &vc->kvm->arch.cpu_in_guest);
 	}
 
 	spin_unlock(&vc->lock);
@@ -4015,8 +4016,14 @@ static int kvmhv_p9_guest_entry(struct kvm_vcpu *vcpu, u64 time_limit,
 		}
 
 	} else {
+		struct kvm *kvm = vcpu->kvm;
+
 		kvmppc_xive_push_vcpu(vcpu);
+
+		__this_cpu_write(cpu_in_guest, kvm);
 		trap = kvmhv_vcpu_entry_p9(vcpu, time_limit, lpcr, tb);
+		__this_cpu_write(cpu_in_guest, NULL);
+
 		if (trap == BOOK3S_INTERRUPT_SYSCALL && !vcpu->arch.nested &&
 		    !(vcpu->arch.shregs.msr & MSR_PR)) {
 			unsigned long req = kvmppc_get_gpr(vcpu, 3);
@@ -4041,7 +4048,7 @@ static int kvmhv_p9_guest_entry(struct kvm_vcpu *vcpu, u64 time_limit,
 		}
 		kvmppc_xive_pull_vcpu(vcpu);
 
-		if (kvm_is_radix(vcpu->kvm))
+		if (kvm_is_radix(kvm))
 			vcpu->arch.slb_max = 0;
 	}
 
@@ -4531,8 +4538,6 @@ int kvmhv_run_single_vcpu(struct kvm_vcpu *vcpu, u64 time_limit,
 
 	powerpc_local_irq_pmu_restore(flags);
 
-	cpumask_clear_cpu(pcpu, &kvm->arch.cpu_in_guest);
-
 	preempt_enable();
 
 	/*