[Qemu-devel] [PATCH] virtio-pci: pass real with_irqfd to virtio_pci_set_guest_notifier()

[Qemu-devel] [PATCH] virtio-pci: pass real with_irqfd to virtio_pci_set_guest_notifier()

[Stefan Hajnoczi]

virtio_pci_set_guest_notifiers() checks whether irqfd can be used and
whether MSI-X is enabled for the PCI adapter.  But then it calls
virtio_pci_set_guest_notifier() and passes kvm_msi_via_irqfd_enabled()
instead of with_irqfd.

When MSI-X is disabled but irqfd is allowed this means that
guest_notifier has neither irqfd nor a
virtio_queue_guest_notifier_read() handler.  Therefore the guest cannot
receive notifications.

This issue is triggered by a Windows 7 Professional 64-bit guest with
-device virtio-blk-pci,x-data-plane=on.  The guest driver does not
enable MSI-X and the guest gets stuck at the Windows boot screen since
it does not receive notifications.

Reported-by: 张磊强 (Leiqiang Zhang) <leiqzhang@gmail.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
---
 hw/virtio-pci.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/hw/virtio-pci.c b/hw/virtio-pci.c
index 736a9bf..84ece51 100644
--- a/hw/virtio-pci.c
+++ b/hw/virtio-pci.c
@@ -798,8 +798,7 @@ static int virtio_pci_set_guest_notifiers(DeviceState *d, int nvqs, bool assign)
             break;
         }
 
-        r = virtio_pci_set_guest_notifier(d, n, assign,
-                                          kvm_msi_via_irqfd_enabled());
+        r = virtio_pci_set_guest_notifier(d, n, assign, with_irqfd);
         if (r < 0) {
             goto assign_error;
         }
-- 
1.8.1.4