]> git.proxmox.com Git - mirror_qemu.git/commitdiff
spapr: Disallow unsupported kernel-irqchip settings
authorGreg Kurz <groug@kaod.org>
Fri, 8 Feb 2019 18:17:47 +0000 (19:17 +0100)
committerDavid Gibson <david@gibson.dropbear.id.au>
Sun, 17 Feb 2019 10:54:02 +0000 (21:54 +1100)
Split mode doesn't make sense on pseries, neither with XICS nor XIVE. But
passing kernel-irqchip=split silently behaves like kernel-irqchip=on.
Other architectures that support kernel-irqchip do terminate QEMU when
split mode is requested but not available though. Do the same with pseries
for consistency.

Similarly, passing kernel-irqchip=on,accel=tcg starts the machine with the
emulated interrupt controller, ie, behaves like kernel-irqchip=off. However,
when passing  kernel-irqchip=on,accel=kvm, if we can't initialize the KVM
XICS for some reason, ie, xics_kvm_init() fails, then QEMU is terminated.
This is inconsistent. Terminate QEMU all the same when requesting the
in-kernel interrupt controller without KVM.

Signed-off-by: Greg Kurz <groug@kaod.org>
Message-Id: <154964986747.291716.2679312373018476920.stgit@bahia.lan>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
hw/ppc/spapr_irq.c

index 2d7a7c163876c73e38d54449a35e1ec18ec8e7ce..80b0083b8e383c6cc185fef73d4eb68820aea609 100644 (file)
@@ -600,6 +600,19 @@ sPAPRIrq spapr_irq_dual = {
  */
 void spapr_irq_init(sPAPRMachineState *spapr, Error **errp)
 {
+    MachineState *machine = MACHINE(spapr);
+
+    if (machine_kernel_irqchip_split(machine)) {
+        error_setg(errp, "kernel_irqchip split mode not supported on pseries");
+        return;
+    }
+
+    if (!kvm_enabled() && machine_kernel_irqchip_required(machine)) {
+        error_setg(errp,
+                   "kernel_irqchip requested but only available with KVM");
+        return;
+    }
+
     /* Initialize the MSI IRQ allocator. */
     if (!SPAPR_MACHINE_GET_CLASS(spapr)->legacy_irq_allocation) {
         spapr_irq_msi_init(spapr, spapr->irq->nr_msis);