qemu/vhost-net: mergeable buffers support

index 9f94358..727fcf5 100644
@@ -52,4 +52,10 @@ struct virtio_net_hdr
     uint16_t csum_offset;
 };
 
+struct virtio_net_hdr_mrg_rxbuf
+{
+    struct virtio_net_hdr hdr;
+    uint16_t num_buffers;   /* Number of merged rx buffers */
+};
+
 #endif /* QEMU_TAP_H */

index 0147dab..2866ff4 100644
@@ -57,10 +57,10 @@ typedef struct TAPState {
     uint8_t buf[TAP_BUFSIZE];
     unsigned int read_poll : 1;
     unsigned int write_poll : 1;
-    unsigned int has_vnet_hdr : 1;
     unsigned int using_vnet_hdr : 1;
     unsigned int has_ufo: 1;
     VHostNetState *vhost_net;
+    unsigned host_vnet_hdr_len;
 } TAPState;
 
 static int launch_script(const char *setup_script, const char *ifname, int fd);

@@ -121,11 +121,11 @@ static ssize_t tap_receive_iov(VLANClientState *nc, const struct iovec *iov,
     TAPState *s = DO_UPCAST(TAPState, nc, nc);
     const struct iovec *iovp = iov;
     struct iovec iov_copy[iovcnt + 1];
-    struct virtio_net_hdr hdr = { 0, };
+    struct virtio_net_hdr_mrg_rxbuf hdr = { };
 
-    if (s->has_vnet_hdr && !s->using_vnet_hdr) {
+    if (s->host_vnet_hdr_len && !s->using_vnet_hdr) {
         iov_copy[0].iov_base = &hdr;
-        iov_copy[0].iov_len =  sizeof(hdr);
+        iov_copy[0].iov_len =  s->host_vnet_hdr_len;
         memcpy(&iov_copy[1], iov, iovcnt * sizeof(*iov));
         iovp = iov_copy;
         iovcnt++;

@@ -139,11 +139,11 @@ static ssize_t tap_receive_raw(VLANClientState *nc, const uint8_t *buf, size_t s
     TAPState *s = DO_UPCAST(TAPState, nc, nc);
     struct iovec iov[2];
     int iovcnt = 0;
-    struct virtio_net_hdr hdr = { 0, };
+    struct virtio_net_hdr_mrg_rxbuf hdr = { };
 
-    if (s->has_vnet_hdr) {
+    if (s->host_vnet_hdr_len) {
         iov[iovcnt].iov_base = &hdr;
-        iov[iovcnt].iov_len  = sizeof(hdr);
+        iov[iovcnt].iov_len  = s->host_vnet_hdr_len;
         iovcnt++;
     }
 

@@ -159,7 +159,7 @@ static ssize_t tap_receive(VLANClientState *nc, const uint8_t *buf, size_t size)
     TAPState *s = DO_UPCAST(TAPState, nc, nc);
     struct iovec iov[1];
 
-    if (s->has_vnet_hdr && !s->using_vnet_hdr) {
+    if (s->host_vnet_hdr_len && !s->using_vnet_hdr) {
         return tap_receive_raw(nc, buf, size);
     }
 

@@ -202,9 +202,9 @@ static void tap_send(void *opaque)
             break;
         }
 
-        if (s->has_vnet_hdr && !s->using_vnet_hdr) {
-            buf  += sizeof(struct virtio_net_hdr);
-            size -= sizeof(struct virtio_net_hdr);
+        if (s->host_vnet_hdr_len && !s->using_vnet_hdr) {
+            buf  += s->host_vnet_hdr_len;
+            size -= s->host_vnet_hdr_len;
         }
 
         size = qemu_send_packet_async(&s->nc, buf, size, tap_send_completed);

@@ -229,7 +229,7 @@ int tap_has_vnet_hdr(VLANClientState *nc)
 
     assert(nc->info->type == NET_CLIENT_TYPE_TAP);
 
-    return s->has_vnet_hdr;
+    return !!s->host_vnet_hdr_len;
 }
 
 void tap_using_vnet_hdr(VLANClientState *nc, int using_vnet_hdr)

@@ -239,7 +239,7 @@ void tap_using_vnet_hdr(VLANClientState *nc, int using_vnet_hdr)
     using_vnet_hdr = using_vnet_hdr != 0;
 
     assert(nc->info->type == NET_CLIENT_TYPE_TAP);
-    assert(s->has_vnet_hdr == using_vnet_hdr);
+    assert(!!s->host_vnet_hdr_len == using_vnet_hdr);
 
     s->using_vnet_hdr = using_vnet_hdr;
 }

@@ -310,7 +310,7 @@ static TAPState *net_tap_fd_init(VLANState *vlan,
     s = DO_UPCAST(TAPState, nc, nc);
 
     s->fd = fd;
-    s->has_vnet_hdr = vnet_hdr != 0;
+    s->host_vnet_hdr_len = vnet_hdr ? sizeof(struct virtio_net_hdr) : 0;
     s->using_vnet_hdr = 0;
     s->has_ufo = tap_probe_has_ufo(s->fd);
     tap_set_offload(&s->nc, 0, 0, 0, 0, 0);

index 4bc9f2f..e19aaba 100644
@@ -46,6 +46,15 @@ int tap_probe_has_ufo(int fd)
     return 0;
 }
 
+int tap_probe_vnet_hdr_len(int fd, int len)
+{
+    return 0;
+}
+
+void tap_fd_set_vnet_hdr_len(int fd, int len)
+{
+}
+
 void tap_fd_set_offload(int fd, int csum, int tso4,
                         int tso6, int ecn, int ufo)
 {

index 3773d5d..3513075 100644
@@ -116,6 +116,15 @@ int tap_probe_has_ufo(int fd)
     return 0;
 }
 
+int tap_probe_vnet_hdr_len(int fd, int len)
+{
+    return 0;
+}
+
+void tap_fd_set_vnet_hdr_len(int fd, int len)
+{
+}
+
 void tap_fd_set_offload(int fd, int csum, int tso4,
                         int tso6, int ecn, int ufo)
 {

index c92983c..f7aa904 100644
@@ -129,6 +129,35 @@ int tap_probe_has_ufo(int fd)
     return 1;
 }
 
+/* Verify that we can assign given length */
+int tap_probe_vnet_hdr_len(int fd, int len)
+{
+    int orig;
+    if (ioctl(fd, TUNGETVNETHDRSZ, &orig) == -1) {
+        return 0;
+    }
+    if (ioctl(fd, TUNSETVNETHDRSZ, &len) == -1) {
+        return 0;
+    }
+    /* Restore original length: we can't handle failure. */
+    if (ioctl(fd, TUNSETVNETHDRSZ, &orig) == -1) {
+        fprintf(stderr, "TUNGETVNETHDRSZ ioctl() failed: %s. Exiting.\n",
+                strerror(errno));
+        assert(0);
+        return -errno;
+    }
+    return 1;
+}
+
+void tap_fd_set_vnet_hdr_len(int fd, int len)
+{
+    if (ioctl(fd, TUNSETVNETHDRSZ, &len) == -1) {
+        fprintf(stderr, "TUNSETVNETHDRSZ ioctl() failed: %s. Exiting.\n",
+                strerror(errno));
+        assert(0);
+    }
+}
+
 void tap_fd_set_offload(int fd, int csum, int tso4,
                         int tso6, int ecn, int ufo)
 {

index 727fcf5..659e981 100644
@@ -27,6 +27,8 @@
 #define TUNSETOFFLOAD  _IOW('T', 208, unsigned int)
 #define TUNGETIFF      _IOR('T', 210, unsigned int)
 #define TUNSETSNDBUF   _IOW('T', 212, int)
+#define TUNGETVNETHDRSZ _IOR('T', 215, int)
+#define TUNSETVNETHDRSZ _IOW('T', 216, int)
 
 #endif
 

index 50d127a..ca1d496 100644
@@ -211,6 +211,15 @@ int tap_probe_has_ufo(int fd)
     return 0;
 }
 
+int tap_probe_vnet_hdr_len(int fd, int len)
+{
+    return 0;
+}
+
+void tap_fd_set_vnet_hdr_len(int fd, int len)
+{
+}
+
 void tap_fd_set_offload(int fd, int csum, int tso4,
                         int tso6, int ecn, int ufo)
 {

index 74348da..9fe4fcd 100644
@@ -728,6 +728,15 @@ int tap_has_vnet_hdr(VLANClientState *vc)
     return 0;
 }
 
+int tap_probe_vnet_hdr_len(int fd, int len)
+{
+    return 0;
+}
+
+void tap_fd_set_vnet_hdr_len(int fd, int len)
+{
+}
+
 void tap_using_vnet_hdr(VLANClientState *vc, int using_vnet_hdr)
 {
 }

index 2866ff4..e0f673c 100644
@@ -232,6 +232,27 @@ int tap_has_vnet_hdr(VLANClientState *nc)
     return !!s->host_vnet_hdr_len;
 }
 
+int tap_has_vnet_hdr_len(VLANClientState *nc, int len)
+{
+    TAPState *s = DO_UPCAST(TAPState, nc, nc);
+
+    assert(nc->info->type == NET_CLIENT_TYPE_TAP);
+
+    return tap_probe_vnet_hdr_len(s->fd, len);
+}
+
+void tap_set_vnet_hdr_len(VLANClientState *nc, int len)
+{
+    TAPState *s = DO_UPCAST(TAPState, nc, nc);
+
+    assert(nc->info->type == NET_CLIENT_TYPE_TAP);
+    assert(len == sizeof(virtio_net_hdr_mrg) ||
+           len == sizeof(virtio_net_hdr));
+
+    tap_fd_set_vnet_hdr_len(s->fd, len);
+    s->host_vnet_hdr_len = len;
+}
+
 void tap_using_vnet_hdr(VLANClientState *nc, int using_vnet_hdr)
 {
     TAPState *s = DO_UPCAST(TAPState, nc, nc);

index b8cec83..e44bd2b 100644
@@ -40,13 +40,17 @@ ssize_t tap_read_packet(int tapfd, uint8_t *buf, int maxlen);
 
 int tap_has_ufo(VLANClientState *vc);
 int tap_has_vnet_hdr(VLANClientState *vc);
+int tap_has_vnet_hdr_len(VLANClientState *vc, int len);
 void tap_using_vnet_hdr(VLANClientState *vc, int using_vnet_hdr);
 void tap_set_offload(VLANClientState *vc, int csum, int tso4, int tso6, int ecn, int ufo);
+void tap_set_vnet_hdr_len(VLANClientState *vc, int len);
 
 int tap_set_sndbuf(int fd, QemuOpts *opts);
 int tap_probe_vnet_hdr(int fd);
+int tap_probe_vnet_hdr_len(int fd, int len);
 int tap_probe_has_ufo(int fd);
 void tap_fd_set_offload(int fd, int csum, int tso4, int tso6, int ecn, int ufo);
+void tap_fd_set_vnet_hdr_len(int fd, int len);
 
 int tap_get_fd(VLANClientState *vc);
 

index 606aa0c..1693090 100644
@@ -51,7 +51,9 @@ unsigned vhost_net_get_features(struct vhost_net *net, unsigned features)
     if (!(net->dev.features & (1 << VIRTIO_RING_F_INDIRECT_DESC))) {
         features &= ~(1 << VIRTIO_RING_F_INDIRECT_DESC);
     }
-    features &= ~(1 << VIRTIO_NET_F_MRG_RXBUF);
+    if (!(net->dev.features & (1 << VIRTIO_NET_F_MRG_RXBUF))) {
+        features &= ~(1 << VIRTIO_NET_F_MRG_RXBUF);
+    }
     return features;
 }
 

@@ -64,6 +66,9 @@ void vhost_net_ack_features(struct vhost_net *net, unsigned features)
     if (features & (1 << VIRTIO_RING_F_INDIRECT_DESC)) {
         net->dev.acked_features |= (1 << VIRTIO_RING_F_INDIRECT_DESC);
     }
+    if (features & (1 << VIRTIO_NET_F_MRG_RXBUF)) {
+        net->dev.acked_features |= (1 << VIRTIO_NET_F_MRG_RXBUF);
+    }
 }
 
 static int vhost_net_get_fd(VLANClientState *backend)

@@ -98,6 +103,10 @@ struct vhost_net *vhost_net_init(VLANClientState *backend, int devfd)
     if (r < 0) {
         goto fail;
     }
+    if (!tap_has_vnet_hdr_len(backend,
+                              sizeof(struct virtio_net_hdr_mrg_rxbuf))) {
+        net->dev.features &= ~(1 << VIRTIO_NET_F_MRG_RXBUF);
+    }
     if (~net->dev.features & net->dev.backend_features) {
         fprintf(stderr, "vhost lacks feature mask %" PRIu64 " for backend\n",
                 (uint64_t)(~net->dev.features & net->dev.backend_features));

@@ -118,6 +127,9 @@ int vhost_net_start(struct vhost_net *net,
 {
     struct vhost_vring_file file = { };
     int r;
+    if (net->dev.acked_features & VIRTIO_NET_F_MRG_RXBUF) {
+        tap_set_vnet_hdr_len(net->backend, sizeof(virtio_net_hdr_mrg_rxbuf));
+    }
 
     net->dev.nvqs = 2;
     net->dev.vqs = net->vqs;

@@ -145,6 +157,9 @@ fail:
     }
     net->vc->info->poll(net->vc, true);
     vhost_dev_stop(&net->dev, dev);
+    if (net->dev.acked_features & VIRTIO_NET_F_MRG_RXBUF) {
+        tap_set_vnet_hdr_len(net->backend, sizeof(virtio_net_hdr));
+    }
     return r;
 }
 

@@ -159,11 +174,17 @@ void vhost_net_stop(struct vhost_net *net,
     }
     net->vc->info->poll(net->vc, true);
     vhost_dev_stop(&net->dev, dev);
+    if (net->dev.acked_features & VIRTIO_NET_F_MRG_RXBUF) {
+        tap_set_vnet_hdr_len(net->backend, sizeof(virtio_net_hdr));
+    }
 }
 
 void vhost_net_cleanup(struct vhost_net *net)
 {
     vhost_dev_cleanup(&net->dev);
+    if (net->dev.acked_features & VIRTIO_NET_F_MRG_RXBUF) {
+        tap_set_vnet_hdr_len(net->backend, sizeof(virtio_net_hdr));
+    }
     qemu_free(net);
 }
 #else