Re: [Qemu-devel] [PATCH v5 1/8] virtio_get_byteswap: function for endian-ambivalent targets using virtio

[Stefan Hajnoczi <stefanha@redhat.com>]

On Fri, Feb 21, 2014 at 12:28:11PM +0100, Greg Kurz wrote:
> diff --git a/hw/virtio/virtio.c b/hw/virtio/virtio.c
> index aeabf3a..4fd6ac2 100644
> --- a/hw/virtio/virtio.c
> +++ b/hw/virtio/virtio.c
> @@ -19,6 +19,9 @@
>  #include "hw/virtio/virtio.h"
>  #include "qemu/atomic.h"
>  #include "hw/virtio/virtio-bus.h"
> +#include "hw/virtio/virtio-access.h"
> +
> +bool virtio_byteswap;
>  
>  /*
>   * The alignment to use between consumer and producer parts of vring.
> @@ -546,6 +549,9 @@ void virtio_reset(void *opaque)
>  
>      virtio_set_status(vdev, 0);
>  
> +    /* We assume all devices are the same endian. */
> +    virtio_byteswap = virtio_get_byteswap();
> +
>      if (k->reset) {
>          k->reset(vdev);
>      }

In the previous version it was pointed out that the global is nasty.

A cleaner solution is a per-VirtIODevice enum device_endian value which
gets passed to lduw_phys_internal()/stl_phys_internal()/etc.

Now the same code can be used for:
1. Host == guest-endian
2. Host != guest-endian
3. "Everything is little-endian" (VIRTIO 1.0)

Keeping it per-device means we can have both legacy and VIRTIO 1.0
devices in a guest.

That's better than hacking in a global now and having to undo it when
VIRTIO 1.0 support gets merged.

Stefan