On 03/07/2025 14:22, Suzuki K Poulose wrote:

On 11/06/2025 11:48, Steven Price wrote:

quoted

The RMM provides emulation of a VGIC to the realm guest but delegates
much of the handling to the host. Implement support in KVM for
saving/restoring state to/from the REC structure.

Signed-off-by: Steven Price <steven.price@arm.com>
---
Changes from v8:
  * Propagate gicv3_hcr to from the RMM.
Changes from v5:
  * Handle RMM providing fewer GIC LRs than the hardware supports.
---
  arch/arm64/include/asm/kvm_rme.h |  1 +
  arch/arm64/kvm/arm.c             | 16 +++++++++--
  arch/arm64/kvm/rme.c             |  5 ++++
  arch/arm64/kvm/vgic/vgic-init.c  |  2 +-
  arch/arm64/kvm/vgic/vgic-v3.c    |  6 +++-
  arch/arm64/kvm/vgic/vgic.c       | 49 ++++++++++++++++++++++++++++++--
  6 files changed, 72 insertions(+), 7 deletions(-)

...

quoted

diff --git a/arch/arm64/kvm/vgic/vgic-init.c b/arch/arm64/kvm/vgic/

vgic-init.c
index eb1205654ac8..77b4962ebfb6 100644

--- a/arch/arm64/kvm/vgic/vgic-init.c
+++ b/arch/arm64/kvm/vgic/vgic-init.c

@@ -81,7 +81,7 @@ int kvm_vgic_create(struct kvm *kvm, u32 type)

       * the proper checks already.
       */
      if (type == KVM_DEV_TYPE_ARM_VGIC_V2 &&
-        !kvm_vgic_global_state.can_emulate_gicv2)
+        (!kvm_vgic_global_state.can_emulate_gicv2 || kvm_is_realm(kvm)))
          return -ENODEV;
        /*

diff --git a/arch/arm64/kvm/vgic/vgic-v3.c b/arch/arm64/kvm/vgic/vgic-

v3.c
index b9ad7c42c5b0..c10ad817030d 100644

--- a/arch/arm64/kvm/vgic/vgic-v3.c
+++ b/arch/arm64/kvm/vgic/vgic-v3.c

@@ -8,9 +8,11 @@

  #include <linux/kvm_host.h>
  #include <linux/string_choices.h>
  #include <kvm/arm_vgic.h>
+#include <asm/kvm_emulate.h>
  #include <asm/kvm_hyp.h>
  #include <asm/kvm_mmu.h>
  #include <asm/kvm_asm.h>
+#include <asm/rmi_smc.h>
    #include "vgic.h"
  @@ -758,7 +760,9 @@ void vgic_v3_put(struct kvm_vcpu *vcpu)
          return;
      }
  -    if (likely(!is_protected_kvm_enabled()))
+    if (vcpu_is_rec(vcpu))
+        cpu_if->vgic_vmcr = vcpu->arch.rec.run->exit.gicv3_vmcr;
+    else if (likely(!is_protected_kvm_enabled()))

This function is never reached for Realms, this should be folded in
vgic_rmm_save_state(). See below.

Good catch - I can drop the changes in this file and handle VMCR in
vgic_rmm_save_state().

Thanks,
Steve

quoted

          kvm_call_hyp(__vgic_v3_save_vmcr_aprs, cpu_if);
      WARN_ON(vgic_v4_put(vcpu));
  diff --git a/arch/arm64/kvm/vgic/vgic.c b/arch/arm64/kvm/vgic/vgic.c
index c7aed48c5668..2908b4610c4e 100644

--- a/arch/arm64/kvm/vgic/vgic.c
+++ b/arch/arm64/kvm/vgic/vgic.c

...

quoted

  void kvm_vgic_put(struct kvm_vcpu *vcpu)
  {
-    if (unlikely(!irqchip_in_kernel(vcpu->kvm) || !
vgic_initialized(vcpu->kvm))) {
+    if (unlikely(vcpu_is_rec(vcpu)))
+        return;

^^^

quoted

+    if (unlikely(!irqchip_in_kernel(vcpu->kvm) ||
+             !vgic_initialized(vcpu->kvm))) {
          if (has_vhe() &&
static_branch_unlikely(&kvm_vgic_global_state.gicv3_cpuif))
              __vgic_v3_deactivate_traps(&vcpu->arch.vgic_cpu.vgic_v3);
          return;

Suzuki