Re: kvm-84 + virtio Ubuntu Hardy guests

[Dustin Kirkland <kirkland@canonical.com>]

On Tue, Mar 31, 2009 at 5:28 PM, Dustin Kirkland <kirkland@canonical.com> wrote:
> I'm receiving a heavy volume of Ubuntu Jaunty Beta users reporting
> that Jaunty hosts running kvm-84 (userspace and kernel) are not able
> to boot previously-working Hardy guests (2.6.24 kernel) if virtio
> networking is enabled [1].  Users report that if e1000 is used
> instead, the guest is able to boot (with degraded network performance,
> obviously).  Users are also reporting that this was not a problem when
> kvm-82 was used in Jaunty (though we also merged libvirt 0.5.1 up to
> 0.6.0 in roughly the same timeframe).
...
> [1] https://bugs.launchpad.net/ubuntu/+source/kvm/+bug/331128

Howdy-

Just a follow-up...

Anthony was able to confirm this issue, and create a patch for KVM,
which we're carrying in Ubuntu.  It's a bit of a special-case hack,
but I'm dropping it here for the sake of completeness.

Basically, Hardy guests do not have working GSO (general segment
offload) support.  Some changes in kvm/libvirt appear to be exposing
this, and breaking some guests when running virtio.

This patch from Anthony basically disables this support in KVM
userspace (until we have a better solution for auto-detecting GSO
support or lack thereof).

:-Dustin

Work around broken virtio drivers in 2.6.26

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

diff --git a/qemu/hw/virtio-net.c b/qemu/hw/virtio-net.c
index 9bce3a0..5b615f9 100644
--- a/qemu/hw/virtio-net.c
+++ b/qemu/hw/virtio-net.c
@@ -120,6 +120,9 @@ static uint32_t virtio_net_get_features(VirtIODevice *vdev)

     if (tap_has_vnet_hdr(host)) {
         tap_using_vnet_hdr(host, 1);
+#if 0
+        /* Stop advertising advanced features until we work around the fact
+         * that this is totally broken in 2.6.26 kernels */
         features |= (1 << VIRTIO_NET_F_CSUM);
         features |= (1 << VIRTIO_NET_F_GUEST_CSUM);
         features |= (1 << VIRTIO_NET_F_GUEST_TSO4);
@@ -130,6 +133,7 @@ static uint32_t virtio_net_get_features(VirtIODevice *vdev)
         features |= (1 << VIRTIO_NET_F_HOST_ECN);
         features |= (1 << VIRTIO_NET_F_MRG_RXBUF);
         /* Kernel can't actually handle UFO in software currently. */
+#endif
     }
 #endif

@@ -374,8 +378,14 @@ static int receive_header(VirtIONet *n, struct
iovec *iov, int iovcnt,
     struct virtio_net_hdr *hdr = iov[0].iov_base;
     int offset = 0;

+#if 0
     hdr->flags = 0;
     hdr->gso_type = VIRTIO_NET_HDR_GSO_NONE;
+#else
+    /* we need to clear out the whole header, including any garbage that may be
+     */
+    memset(hdr, 0, sizeof(*hdr));
+#endif

 #ifdef TAP_VNET_HDR
     if (tap_has_vnet_hdr(n->vc->vlan->first_client)) {