diff mbox series

[PULL,24/37] hw/arm/virt: Enable NMI support in the GIC if the CPU has FEAT_NMI

Message ID 20240425103958.3237225-25-peter.maydell@linaro.org
State Accepted
Commit 5ae47f7aece19c5a6efbf99d000f389278e93c1d
Headers show
Series [PULL,01/37] target/arm: Handle HCR_EL2 accesses for bits introduced with FEAT_NMI | expand

Commit Message

Peter Maydell April 25, 2024, 10:39 a.m. UTC 
From: Jinjie Ruan <ruanjinjie@huawei.com>

If the CPU implements FEAT_NMI, then turn on the NMI support in the
GICv3 too.  It's permitted to have a configuration with FEAT_NMI in
the CPU (and thus NMI support in the CPU interfaces too) but no NMI
support in the distributor and redistributor, but this isn't a very
useful setup as it's close to having no NMI support at all.

We don't need to gate the enabling of NMI in the GIC behind a
machine version property, because none of our current CPUs
implement FEAT_NMI, and '-cpu max' is not something we maintain
migration compatibility across versions for. So we can always
enable the GIC NMI support when the CPU has it.

Neither hvf nor KVM support NMI in the GIC yet, so we don't enable
it unless we're using TCG.

Signed-off-by: Jinjie Ruan <ruanjinjie@huawei.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20240407081733.3231820-25-ruanjinjie@huawei.com
[PMM: Update comment and commit message]
Suggested-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
---
 hw/arm/virt.c | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)
diff mbox series

Patch

diff --git a/hw/arm/virt.c b/hw/arm/virt.c
index c4b03b09c27..3c93c0c0a61 100644
--- a/hw/arm/virt.c
+++ b/hw/arm/virt.c
@@ -729,6 +729,20 @@  static void create_v2m(VirtMachineState *vms)
     vms->msi_controller = VIRT_MSI_CTRL_GICV2M;
 }
 
+/*
+ * If the CPU has FEAT_NMI, then turn on the NMI support in the GICv3 too.
+ * It's permitted to have a configuration with NMI in the CPU (and thus the
+ * GICv3 CPU interface) but not in the distributor/redistributors, but it's
+ * not very useful.
+ */
+static bool gicv3_nmi_present(VirtMachineState *vms)
+{
+    ARMCPU *cpu = ARM_CPU(qemu_get_cpu(0));
+
+    return tcg_enabled() && cpu_isar_feature(aa64_nmi, cpu) &&
+           (vms->gic_version != VIRT_GIC_VERSION_2);
+}
+
 static void create_gic(VirtMachineState *vms, MemoryRegion *mem)
 {
     MachineState *ms = MACHINE(vms);
@@ -802,6 +816,11 @@  static void create_gic(VirtMachineState *vms, MemoryRegion *mem)
                               vms->virt);
         }
     }
+
+    if (gicv3_nmi_present(vms)) {
+        qdev_prop_set_bit(vms->gic, "has-nmi", true);
+    }
+
     gicbusdev = SYS_BUS_DEVICE(vms->gic);
     sysbus_realize_and_unref(gicbusdev, &error_fatal);
     sysbus_mmio_map(gicbusdev, 0, vms->memmap[VIRT_GIC_DIST].base);