It's currently broken.  Let's use the slow path to at least make it
functional.

Tested-by: Eric Auger <eric.auger@redhat.com>
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Signed-off-by: Peter Xu <peterx@redhat.com>
---
 hw/vfio/pci.c | 12 ++++++++++++
 1 file changed, 12 insertions(+)

diff --git a/hw/vfio/pci.c b/hw/vfio/pci.c
index 5e75a95129..98e0e0c994 100644
--- a/hw/vfio/pci.c
+++ b/hw/vfio/pci.c
@@ -128,6 +128,18 @@ static void vfio_intx_enable_kvm(VFIOPCIDevice *vdev, Error **errp)
         return;
     }
 
+    if (kvm_irqchip_is_split()) {
+        /*
+         * VFIO INTx is currently not working with split kernel
+         * irqchip for level triggered interrupts.  Go the slow path
+         * as long as split is enabled so we can be at least
+         * functional (even with poor performance).
+         *
+         * TODO: Remove this after all things fixed up.
+         */
+        return;
+    }
+
     /* Get to a known interrupt state */
     qemu_set_fd_handler(irqfd.fd, NULL, NULL, vdev);
     vfio_mask_single_irqindex(&vdev->vbasedev, VFIO_PCI_INTX_IRQ_INDEX);
-- 
2.24.1

On Tue, 17 Mar 2020 15:50:38 -0400
Peter Xu <peterx@redhat.com> wrote:

> It's currently broken.  Let's use the slow path to at least make it
> functional.
> 
> Tested-by: Eric Auger <eric.auger@redhat.com>
> Reviewed-by: Eric Auger <eric.auger@redhat.com>
> Signed-off-by: Peter Xu <peterx@redhat.com>
> ---
>  hw/vfio/pci.c | 12 ++++++++++++
>  1 file changed, 12 insertions(+)

Reviewed-by: Alex Williamson <alex.williamson@redhat.com>
Acked-by: Alex Williamson <alex.williamson@redhat.com>
 
> diff --git a/hw/vfio/pci.c b/hw/vfio/pci.c
> index 5e75a95129..98e0e0c994 100644
> --- a/hw/vfio/pci.c
> +++ b/hw/vfio/pci.c
> @@ -128,6 +128,18 @@ static void vfio_intx_enable_kvm(VFIOPCIDevice *vdev, Error **errp)
>          return;
>      }
>  
> +    if (kvm_irqchip_is_split()) {
> +        /*
> +         * VFIO INTx is currently not working with split kernel
> +         * irqchip for level triggered interrupts.  Go the slow path
> +         * as long as split is enabled so we can be at least
> +         * functional (even with poor performance).
> +         *
> +         * TODO: Remove this after all things fixed up.
> +         */
> +        return;
> +    }
> +
>      /* Get to a known interrupt state */
>      qemu_set_fd_handler(irqfd.fd, NULL, NULL, vdev);
>      vfio_mask_single_irqindex(&vdev->vbasedev, VFIO_PCI_INTX_IRQ_INDEX);