[Qemu-devel] [PATCH 4/6] virtio network driver

[Anthony Liguori <aliguori@us.ibm.com>]

This patch implements the virtio network driver backend.  In KVM, this driver
can achieve 1gbit tx/rx performance.  More patches are required to improve the
network IO infrastructure to achieve better performance in QEMU.

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

diff --git a/Makefile.target b/Makefile.target
index 6815ba8..3ea40d1 100644
--- a/Makefile.target
+++ b/Makefile.target
@@ -535,7 +535,7 @@ OBJS += rtl8139.o
 OBJS += e1000.o
 
 # virtio devices
-OBJS += virtio.o
+OBJS += virtio.o virtio-net.o
 
 ifeq ($(TARGET_BASE_ARCH), i386)
 # Hardware support
diff --git a/hw/pci.c b/hw/pci.c
index 99c206f..8dca481 100644
--- a/hw/pci.c
+++ b/hw/pci.c
@@ -654,9 +654,11 @@ void pci_nic_init(PCIBus *bus, NICInfo *nd, int devfn)
         pci_e1000_init(bus, nd, devfn);
     } else if (strcmp(nd->model, "pcnet") == 0) {
         pci_pcnet_init(bus, nd, devfn);
+    } else if (strcmp(nd->model, "virtio") == 0) {
+	virtio_net_init(bus, nd, devfn);
     } else if (strcmp(nd->model, "?") == 0) {
         fprintf(stderr, "qemu: Supported PCI NICs: i82551 i82557b i82559er"
-                        " ne2k_pci pcnet rtl8139 e1000\n");
+                        " ne2k_pci pcnet rtl8139 e1000 virtio\n");
         exit (1);
     } else {
         fprintf(stderr, "qemu: Unsupported NIC: %s\n", nd->model);
diff --git a/hw/pci.h b/hw/pci.h
index b965919..f4db366 100644
--- a/hw/pci.h
+++ b/hw/pci.h
@@ -143,4 +143,7 @@ PCIBus *pci_prep_init(qemu_irq *pic);
 PCIBus *pci_apb_init(target_phys_addr_t special_base, target_phys_addr_t mem_base,
                      qemu_irq *pic);
 
+/* virtio.c */
+PCIDevice *virtio_net_init(PCIBus *bus, NICInfo *nd, int devfn);
+
 #endif
diff --git a/hw/virtio-net.c b/hw/virtio-net.c
new file mode 100644
index 0000000..de4723a
--- /dev/null
+++ b/hw/virtio-net.c
@@ -0,0 +1,180 @@
+/*
+ * Virtio Network Device
+ *
+ * Copyright IBM, Corp. 2007
+ *
+ * Authors:
+ *  Anthony Liguori   <aliguori@us.ibm.com>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2.  See
+ * the COPYING file in the top-level directory.
+ *
+ */
+
+#include "virtio.h"
+#include "net.h"
+#include "pc.h"
+#include "qemu-timer.h"
+#include "virtio-net.h"
+
+#define TX_TIMER_INTERVAL (1000 / 500)
+
+typedef struct VirtIONet
+{
+    VirtIODevice vdev;
+    uint8_t mac[6];
+    VirtQueue *rx_vq;
+    VirtQueue *tx_vq;
+    VLANClientState *vc;
+    int can_receive;
+    QEMUTimer *tx_timer;
+    int tx_timer_active;
+} VirtIONet;
+
+static VirtIONet *to_virtio_net(VirtIODevice *vdev)
+{
+    return (VirtIONet *)vdev;
+}
+
+static void virtio_net_get_config(VirtIODevice *vdev, uint8_t *config)
+{
+    VirtIONet *n = to_virtio_net(vdev);
+    struct virtio_net_config netcfg;
+
+    memcpy(netcfg.mac, n->mac, 6);
+    memcpy(config, &netcfg, sizeof(netcfg));
+}
+
+static uint32_t virtio_net_get_features(VirtIODevice *vdev)
+{
+    return (1 << VIRTIO_NET_F_MAC);
+}
+
+/* RX */
+
+static void virtio_net_handle_rx(VirtIODevice *vdev, VirtQueue *vq)
+{
+    VirtIONet *n = to_virtio_net(vdev);
+    n->can_receive = 1;
+}
+
+static int virtio_net_can_receive(void *opaque)
+{
+    VirtIONet *n = opaque;
+
+    return (n->vdev.status & VIRTIO_CONFIG_S_DRIVER_OK) && n->can_receive;
+}
+
+static void virtio_net_receive(void *opaque, const uint8_t *buf, int size)
+{
+    VirtIONet *n = opaque;
+    VirtQueueElement *elem;
+    struct virtio_net_hdr hdr;
+
+    /* FIXME: the drivers really need to set their status better */
+    if (!virtio_ring_inited(n->rx_vq)) {
+	n->can_receive = 0;
+	return;
+    }
+
+    if ((elem = virtqueue_pop(n->rx_vq)) == NULL) {
+	/* wait until the guest adds some rx bufs */
+	n->can_receive = 0;
+	return;
+    }
+
+    memset(&hdr, 0, sizeof(hdr));
+    hdr.flags = 0;
+    hdr.gso_type = VIRTIO_NET_HDR_GSO_NONE;
+
+    memcpy_to_iovector(&hdr, 0, sizeof(hdr), elem->virt_in);
+    memcpy_to_iovector(buf, sizeof(hdr), size, elem->virt_in);
+
+    /* signal other side */
+    virtqueue_push(n->rx_vq, elem, sizeof(hdr) + size);
+    virtio_notify(&n->vdev, n->rx_vq);
+}
+
+/* TX */
+static void virtio_net_flush_tx(VirtIONet *n, VirtQueue *vq)
+{
+    VirtQueueElement *elem;
+
+    if (!(n->vdev.status & VIRTIO_CONFIG_S_DRIVER_OK))
+        return;
+
+    while ((elem = virtqueue_pop(vq))) {
+	IOVector *sg;
+	size_t out_size;
+	int i;
+
+	/* ignore the header for now */
+	out_size = iovector_size(elem->virt_out);
+
+	sg = iovector_trim(elem->virt_out, sizeof(struct virtio_net_hdr),
+			   out_size - sizeof(struct virtio_net_hdr));
+
+	for (i = 0; i < sg->num; i++)
+	    qemu_send_packet(n->vc, sg->sg[i].base, sg->sg[i].len);
+
+	qemu_free(sg);
+
+	virtqueue_push(vq, elem, out_size);
+	virtio_notify(&n->vdev, vq);
+    }
+}
+
+static void virtio_net_handle_tx(VirtIODevice *vdev, VirtQueue *vq)
+{
+    VirtIONet *n = to_virtio_net(vdev);
+
+    if (n->tx_timer_active &&
+	virtio_ring_avail_size(vq) == 64) {
+	virtio_ring_set_used_notify(vq, 0);
+	qemu_del_timer(n->tx_timer);
+	n->tx_timer_active = 0;
+	virtio_net_flush_tx(n, vq);
+    } else {
+	qemu_mod_timer(n->tx_timer,
+		       qemu_get_clock(vm_clock) + TX_TIMER_INTERVAL);
+	n->tx_timer_active = 1;
+	virtio_ring_set_used_notify(vq, 1);
+    }
+}
+
+static void virtio_net_tx_timer(void *opaque)
+{
+    VirtIONet *n = opaque;
+
+    n->tx_timer_active = 0;
+
+    /* Just in case the driver is not ready on more */
+    if (!(n->vdev.status & VIRTIO_CONFIG_S_DRIVER_OK))
+        return;
+
+    virtio_ring_set_used_notify(n->tx_vq, 0);
+    virtio_net_flush_tx(n, n->tx_vq);
+}
+
+PCIDevice *virtio_net_init(PCIBus *bus, NICInfo *nd, int devfn)
+{
+    VirtIONet *n;
+
+    n = (VirtIONet *)virtio_init_pci(bus, "virtio-net", 6900, 0x1000,
+				     0, VIRTIO_ID_NET,
+				     0x02, 0x00, 0x00,
+				     6, sizeof(VirtIONet));
+
+    n->vdev.get_config = virtio_net_get_config;
+    n->vdev.get_features = virtio_net_get_features;
+    n->rx_vq = virtio_add_queue(&n->vdev, 512, virtio_net_handle_rx);
+    n->tx_vq = virtio_add_queue(&n->vdev, 128, virtio_net_handle_tx);
+    n->can_receive = 0;
+    memcpy(n->mac, nd->macaddr, 6);
+    n->vc = qemu_new_vlan_client(nd->vlan, virtio_net_receive,
+                                 virtio_net_can_receive, n);
+    n->tx_timer = qemu_new_timer(vm_clock, virtio_net_tx_timer, n);
+    n->tx_timer_active = 0;
+
+    return (PCIDevice *)n;
+}
diff --git a/hw/virtio-net.h b/hw/virtio-net.h
new file mode 100644
index 0000000..2959198
--- /dev/null
+++ b/hw/virtio-net.h
@@ -0,0 +1,54 @@
+/*
+ * Virtio-net Support
+ *
+ * Copyright IBM, Corp. 2007-2008
+ *
+ * Authors:
+ *  Anthony Liguori   <aliguori@us.ibm.com>
+ *  Rusty Russell     <rusty@rustcorp.com.au>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2.  See
+ * the COPYING file in the top-level directory.
+ *
+ */
+
+#ifndef _QEMU_VIRTIO_NET_H
+#define _QEMU_VIRTIO_NET_H
+
+/* from Linux's virtio_net.h */
+
+/* The ID for virtio_net */
+#define VIRTIO_ID_NET	1
+
+/* The feature bitmap for virtio net */
+#define VIRTIO_NET_F_NO_CSUM	0
+#define VIRTIO_NET_F_MAC	5
+#define VIRTIO_NET_F_GS0	6
+
+/* The config defining mac address (6 bytes) */
+struct virtio_net_config
+{
+    uint8_t mac[6];
+} __attribute__((packed));
+
+/* This is the first element of the scatter-gather list.  If you don't
+ * specify GSO or CSUM features, you can simply ignore the header. */
+struct virtio_net_hdr
+{
+#define VIRTIO_NET_HDR_F_NEEDS_CSUM	1	// Use csum_start, csum_offset
+    uint8_t flags;
+#define VIRTIO_NET_HDR_GSO_NONE		0	// Not a GSO frame
+#define VIRTIO_NET_HDR_GSO_TCPV4	1	// GSO frame, IPv4 TCP (TSO)
+/* FIXME: Do we need this?  If they said they can handle ECN, do they care? */
+#define VIRTIO_NET_HDR_GSO_TCPV4_ECN	2	// GSO frame, IPv4 TCP w/ ECN
+#define VIRTIO_NET_HDR_GSO_UDP		3	// GSO frame, IPv4 UDP (UFO)
+#define VIRTIO_NET_HDR_GSO_TCPV6	4	// GSO frame, IPv6 TCP
+#define VIRTIO_NET_HDR_GSO_ECN		0x80	// TCP has ECN set
+    uint8_t gso_type;
+    uint16_t hdr_len;
+    uint16_t gso_size;
+    uint16_t csum_start;
+    uint16_t csum_offset;
+};
+
+#endif