@@ -372,6 +372,12 @@ struct kvm_vcpu_arch {
u64 last_steal;
gpa_t base;
} steal;
+
+ /* PV state of the VCPU (e.g. vcpu_is_preempted()) */
+ struct {
+ gpa_t base;
+ struct gfn_to_hva_cache ghc;
+ } pv_state;
};
/* Pointer to the vcpu's SVE FFR for sve_{save,load}_state() */
@@ -556,6 +562,23 @@ static inline bool kvm_arm_is_pvtime_enabled(struct kvm_vcpu_arch *vcpu_arch)
return (vcpu_arch->steal.base != GPA_INVALID);
}
+int kvm_init_vcpu_state(struct kvm_vcpu *vcpu, gfn_t addr);
+int kvm_release_vcpu_state(struct kvm_vcpu *vcpu);
+
+static inline void kvm_arm_vcpu_state_init(struct kvm_vcpu_arch *vcpu_arch)
+{
+ vcpu_arch->pv_state.base = GPA_INVALID;
+ memset(&vcpu_arch->pv_state.ghc, 0x00, sizeof(struct gfn_to_hva_cache));
+}
+
+static inline bool
+kvm_arm_is_vcpu_state_enabled(struct kvm_vcpu_arch *vcpu_arch)
+{
+ return (vcpu_arch->pv_state.base != GPA_INVALID);
+}
+
+void kvm_update_vcpu_preempted(struct kvm_vcpu *vcpu, bool preempted);
+
void kvm_set_sei_esr(struct kvm_vcpu *vcpu, u64 syndrome);
struct kvm_vcpu *kvm_mpidr_to_vcpu(struct kvm *kvm, unsigned long mpidr);
@@ -12,7 +12,7 @@ obj-$(CONFIG_KVM) += hyp/
kvm-y := $(KVM)/kvm_main.o $(KVM)/coalesced_mmio.o $(KVM)/eventfd.o \
$(KVM)/vfio.o $(KVM)/irqchip.o \
- arm.o mmu.o mmio.o psci.o perf.o hypercalls.o pvtime.o \
+ arm.o mmu.o mmio.o psci.o perf.o hypercalls.o pvtime.o pvstate.o \
inject_fault.o regmap.o va_layout.o hyp.o handle_exit.o \
guest.o debug.o reset.o sys_regs.o \
vgic-sys-reg-v3.o fpsimd.o pmu.o \
@@ -265,6 +265,8 @@ int kvm_arch_vcpu_create(struct kvm_vcpu *vcpu)
kvm_arm_pvtime_vcpu_init(&vcpu->arch);
+ kvm_arm_vcpu_state_init(&vcpu->arch);
+
vcpu->arch.hw_mmu = &vcpu->kvm->arch.mmu;
err = kvm_vgic_vcpu_init(vcpu);
@@ -355,10 +357,12 @@ void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
if (vcpu_has_ptrauth(vcpu))
vcpu_ptrauth_disable(vcpu);
+ kvm_update_vcpu_preempted(vcpu, false);
}
void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
{
+ kvm_update_vcpu_preempted(vcpu, true);
kvm_arch_vcpu_put_fp(vcpu);
if (has_vhe())
kvm_vcpu_put_sysregs_vhe(vcpu);
@@ -52,6 +52,9 @@ int kvm_hvc_call_handler(struct kvm_vcpu *vcpu)
case ARM_SMCCC_HV_PV_TIME_FEATURES:
val = SMCCC_RET_SUCCESS;
break;
+ case ARM_SMCCC_HV_PV_STATE_FEATURES:
+ val = SMCCC_RET_SUCCESS;
+ break;
}
break;
case ARM_SMCCC_HV_PV_TIME_FEATURES:
@@ -62,6 +65,14 @@ int kvm_hvc_call_handler(struct kvm_vcpu *vcpu)
if (gpa != GPA_INVALID)
val = gpa;
break;
+ case ARM_SMCCC_HV_PV_STATE_INIT:
+ if (kvm_init_vcpu_state(vcpu, smccc_get_arg1(vcpu)) == 0)
+ val = SMCCC_RET_SUCCESS;
+ break;
+ case ARM_SMCCC_HV_PV_STATE_RELEASE:
+ if (kvm_release_vcpu_state(vcpu) == 0)
+ val = SMCCC_RET_SUCCESS;
+ break;
default:
return kvm_psci_call(vcpu);
}
new file mode 100644
@@ -0,0 +1,58 @@
+// SPDX-License-Identifier: GPL-2.0-only
+
+#include <linux/arm-smccc.h>
+#include <linux/kvm_host.h>
+
+#include <asm/kvm_mmu.h>
+#include <asm/paravirt.h>
+
+#include <kvm/arm_hypercalls.h>
+
+int kvm_init_vcpu_state(struct kvm_vcpu *vcpu, gpa_t addr)
+{
+ if (kvm_arm_is_vcpu_state_enabled(&vcpu->arch))
+ return 0;
+
+ if (kvm_gfn_to_hva_cache_init(vcpu->kvm,
+ &vcpu->arch.pv_state.ghc,
+ addr,
+ sizeof(struct pvstate_vcpu_info)))
+ return -EINVAL;
+
+ vcpu->arch.pv_state.base = addr;
+ return 0;
+}
+
+int kvm_release_vcpu_state(struct kvm_vcpu *vcpu)
+{
+ if (!kvm_arm_is_vcpu_state_enabled(&vcpu->arch))
+ return 0;
+
+ vcpu->arch.pv_state.base = GPA_INVALID;
+ return 0;
+}
+
+void kvm_update_vcpu_preempted(struct kvm_vcpu *vcpu, bool preempted)
+{
+ struct kvm *kvm = vcpu->kvm;
+ u64 idx;
+
+ if (!kvm_arm_is_vcpu_state_enabled(&vcpu->arch))
+ return;
+
+ /*
+ * This function is called from atomic context, so we need to
+ * disable page faults. kvm_write_guest_cached() will call
+ * might_fault().
+ */
+ pagefault_disable();
+ /*
+ * Need to take the SRCU lock because kvm_write_guest_offset_cached()
+ * calls kvm_memslots();
+ */
+ idx = srcu_read_lock(&kvm->srcu);
+ kvm_write_guest_cached(kvm, &vcpu->arch.pv_state.ghc,
+ &preempted, sizeof(bool));
+ srcu_read_unlock(&kvm->srcu, idx);
+ pagefault_enable();
+}
Add PV-state support bits to the host. Host uses the guest PV-state per-CPU pointers to update the VCPU state each time it kvm_arch_vcpu_load() or kvm_arch_vcpu_put() the VCPU, so that guest scheduler can become aware of the fact that not all VCPUs are always available. Currently guest scheduler on amr64 always assumes that all CPUs are available because vcpu_is_preempted() is not implemented on arm64. Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> --- arch/arm64/include/asm/kvm_host.h | 23 ++++++++++++ arch/arm64/kvm/Makefile | 2 +- arch/arm64/kvm/arm.c | 4 +++ arch/arm64/kvm/hypercalls.c | 11 ++++++ arch/arm64/kvm/pvstate.c | 58 +++++++++++++++++++++++++++++++ 5 files changed, 97 insertions(+), 1 deletion(-) create mode 100644 arch/arm64/kvm/pvstate.c