File 0139-virtio-net-byteswap-virtio-net-head.patch of Package qemu

From 5b635d5fef6e9a97e2201b30d50ffdb9cb2f3d33 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?C=C3=A9dric=20Le=20Goater?= <clg@fr.ibm.com>
Date: Tue, 24 Jun 2014 19:11:32 +0200
Subject: [PATCH] virtio-net: byteswap virtio-net header
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

TCP connectivity fails when the guest has a different endianness.
The packets are silently dropped on the host by the tap backend
when they are read from user space because the endianness of the
virtio-net header is in the wrong order. These lines may appear
in the guest console:

[  454.709327] skbuff: bad partial csum: csum=8704/4096 len=74
[  455.702554] skbuff: bad partial csum: csum=8704/4096 len=74

The issue that got first spotted with a ppc64le PowerKVM guest,
but it also exists for the less common case of a x86_64 guest run
by a big-endian ppc64 TCG hypervisor.

Signed-off-by: Cédric Le Goater <clg@fr.ibm.com>
[ Ported from PowerKVM,
  Greg Kurz <gkurz@linux.vnet.ibm.com> ]
Signed-off-by: Greg Kurz <gkurz@linux.vnet.ibm.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

(cherry picked from commit 032a74a1c0fcdd5fd1c69e56126b4c857ee36611)
Signed-off-by: Alexander Graf <agraf@suse.de>
---
 hw/net/virtio-net.c | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)

diff --git a/hw/net/virtio-net.c b/hw/net/virtio-net.c
index 940a7cf..2ac6ce5 100644
--- a/hw/net/virtio-net.c
+++ b/hw/net/virtio-net.c
@@ -863,6 +863,14 @@ static int virtio_net_has_buffers(VirtIONetQueue *q, int bufsize)
     return 1;
 }
 
+static void virtio_net_hdr_swap(struct virtio_net_hdr *hdr)
+{
+    tswap16s(&hdr->hdr_len);
+    tswap16s(&hdr->gso_size);
+    tswap16s(&hdr->csum_start);
+    tswap16s(&hdr->csum_offset);
+}
+
 /* dhclient uses AF_PACKET but doesn't pass auxdata to the kernel so
  * it never finds out that the packets don't have valid checksums.  This
  * causes dhclient to get upset.  Fedora's carried a patch for ages to
@@ -898,6 +906,7 @@ static void receive_header(VirtIONet *n, const struct iovec *iov, int iov_cnt,
         void *wbuf = (void *)buf;
         work_around_broken_dhclient(wbuf, wbuf + n->host_hdr_len,
                                     size - n->host_hdr_len);
+        virtio_net_hdr_swap(wbuf);
         iov_from_buf(iov, iov_cnt, 0, buf, sizeof(struct virtio_net_hdr));
     } else {
         struct virtio_net_hdr hdr = {
@@ -1106,6 +1115,14 @@ static int32_t virtio_net_flush_tx(VirtIONetQueue *q)
             exit(1);
         }
 
+        if (n->has_vnet_hdr) {
+            if (out_sg[0].iov_len < n->guest_hdr_len) {
+                error_report("virtio-net header incorrect");
+                exit(1);
+            }
+            virtio_net_hdr_swap((void *) out_sg[0].iov_base);
+        }
+
         /*
          * If host wants to see the guest header as is, we can
          * pass it on unchanged. Otherwise, copy just the parts