xics/spapr: Detect old KVM XICS on POWER9 hosts

Older KVMs on POWER9 don't support destroying/recreating a KVM XICS
device, which is required by 'dual' interrupt controller mode. This
causes QEMU to emit a warning when the guest is rebooted and to fall
back on XICS emulation:

qemu-system-ppc64: warning: kernel_irqchip allowed but unavailable:
 Error on KVM_CREATE_DEVICE for XICS: File exists

If kernel irqchip is required, QEMU will thus exit when the guest is
first rebooted. Failing QEMU this late may be a painful experience
for the user.

Detect that and exit at machine init instead.

Signed-off-by: Greg Kurz <groug@kaod.org>
Message-Id: <156044430517.125694.6207865998817342638.stgit@bahia.lab.toulouse-stg.fr.ibm.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

docs/specs/ppc-spapr-xive.rst

@@ -142,8 +142,8 @@ xics            XICS KVM       XICS emul.     XICS KVM
 (3) QEMU fails at CAS with ``Guest requested unavailable interrupt
     mode (XICS), either don't set the ic-mode machine property or try
     ic-mode=xics or ic-mode=dual``
-(4) QEMU/KVM incompatibility due to device destruction in reset. This
-    needs to be addressed more cleanly with an error.
+(4) QEMU/KVM incompatibility due to device destruction in reset. QEMU fails
+    with ``KVM is too old to support ic-mode=dual,kernel-irqchip=on``
 
 
 XIVE Device tree properties

hw/intc/xics_kvm.c

@@ -452,3 +452,33 @@ void xics_kvm_disconnect(SpaprMachineState *spapr, Error **errp)
     /* Clear the presenter from the VCPUs */
     kvm_disable_icps();
 }
+
+/*
+ * This is a heuristic to detect older KVMs on POWER9 hosts that don't
+ * support destruction of a KVM XICS device while the VM is running.
+ * Required to start a spapr machine with ic-mode=dual,kernel-irqchip=on.
+ */
+bool xics_kvm_has_broken_disconnect(SpaprMachineState *spapr)
+{
+    int rc;
+
+    rc = kvm_create_device(kvm_state, KVM_DEV_TYPE_XICS, false);
+    if (rc < 0) {
+        /*
+         * The error is ignored on purpose. The KVM XICS setup code
+         * will catch it again anyway. The goal here is to see if
+         * close() actually destroys the device or not.
+         */
+        return false;
+    }
+
+    close(rc);
+
+    rc = kvm_create_device(kvm_state, KVM_DEV_TYPE_XICS, false);
+    if (rc >= 0) {
+        close(rc);
+        return false;
+    }
+
+    return errno == EEXIST;
+}

hw/ppc/spapr_irq.c

@@ -669,6 +669,19 @@ static void spapr_irq_check(SpaprMachineState *spapr, Error **errp)
             return;
         }
     }
+
+    /*
+     * On a POWER9 host, some older KVM XICS devices cannot be destroyed and
+     * re-created. Detect that early to avoid QEMU to exit later when the
+     * guest reboots.
+     */
+    if (kvm_enabled() &&
+        spapr->irq == &spapr_irq_dual &&
+        machine_kernel_irqchip_required(machine) &&
+        xics_kvm_has_broken_disconnect(spapr)) {
+        error_setg(errp, "KVM is too old to support ic-mode=dual,kernel-irqchip=on");
+        return;
+    }
 }
 
 /*

include/hw/ppc/xics_spapr.h

@@ -35,6 +35,7 @@ void spapr_dt_xics(SpaprMachineState *spapr, uint32_t nr_servers, void *fdt,
                    uint32_t phandle);
 int xics_kvm_init(SpaprMachineState *spapr, Error **errp);
 void xics_kvm_disconnect(SpaprMachineState *spapr, Error **errp);
+bool xics_kvm_has_broken_disconnect(SpaprMachineState *spapr);
 void xics_spapr_init(SpaprMachineState *spapr);
 void xics_spapr_connect(SpaprMachineState *spapr);