[virtio] Add legacy driver for KVM virtio-net

This patch adds support for the virtio-net adapter provided by KVM.

Written by Laurent Vivier <Laurent.Vivier@bull.net> for Etherboot.
Wrapped as legacy driver for gPXE by Stefan Hajnoczi
<stefanha@gmail.com>.

src/drivers/net/virtio-net.c

@@ -0,0 +1,492 @@
+/* virtio-net.c - etherboot driver for virtio network interface
+ *
+ * (c) Copyright 2008 Bull S.A.S.
+ *
+ *  Author: Laurent Vivier <Laurent.Vivier@bull.net>
+ *
+ * some parts from Linux Virtio PCI driver
+ *
+ *  Copyright IBM Corp. 2007
+ *  Authors: Anthony Liguori  <aliguori@us.ibm.com>
+ *
+ *  some parts from Linux Virtio Ring
+ *
+ *  Copyright Rusty Russell IBM Corporation 2007
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ *
+ *
+ */
+
+#include "etherboot.h"
+#include "nic.h"
+#include "virtio-ring.h"
+#include "virtio-pci.h"
+#include "virtio-net.h"
+
+#define BUG() do { \
+   printf("BUG: failure at %s:%d/%s()!\n", \
+          __FILE__, __LINE__, __FUNCTION__); \
+   while(1); \
+} while (0)
+#define BUG_ON(condition) do { if (condition) BUG(); } while (0)
+
+/* Ethernet header */
+
+struct eth_hdr {
+   unsigned char dst_addr[ETH_ALEN];
+   unsigned char src_addr[ETH_ALEN];
+   unsigned short type;
+};
+
+struct eth_frame {
+   struct eth_hdr hdr;
+   unsigned char data[ETH_FRAME_LEN];
+};
+
+typedef unsigned char virtio_queue_t[PAGE_MASK + vring_size(MAX_QUEUE_NUM)];
+
+/* TX: virtio header and eth buffer */
+
+static struct virtio_net_hdr tx_virtio_hdr;
+static struct eth_frame tx_eth_frame;
+
+/* RX: virtio headers and buffers */
+
+#define RX_BUF_NB  6
+static struct virtio_net_hdr rx_hdr[RX_BUF_NB];
+static unsigned char rx_buffer[RX_BUF_NB][ETH_FRAME_LEN];
+
+/* virtio queues and vrings */
+
+enum {
+   RX_INDEX = 0,
+   TX_INDEX,
+   QUEUE_NB
+};
+
+static virtio_queue_t queue[QUEUE_NB];
+static struct vring vring[QUEUE_NB];
+static u16 free_head[QUEUE_NB];
+static u16 last_used_idx[QUEUE_NB];
+static u16 vdata[QUEUE_NB][MAX_QUEUE_NUM];
+
+/*
+ * Virtio PCI interface
+ *
+ */
+
+static int vp_find_vq(struct nic *nic, int queue_index)
+{
+   struct vring * vr = &vring[queue_index];
+   u16 num;
+
+   /* select the queue */
+
+   outw(queue_index, nic->ioaddr + VIRTIO_PCI_QUEUE_SEL);
+
+   /* check if the queue is available */
+
+   num = inw(nic->ioaddr + VIRTIO_PCI_QUEUE_NUM);
+   if (!num) {
+           printf("ERROR: queue size is 0\n");
+           return -1;
+   }
+
+   if (num > MAX_QUEUE_NUM) {
+           printf("ERROR: queue size %d > %d\n", num, MAX_QUEUE_NUM);
+           return -1;
+   }
+
+   /* check if the queue is already active */
+
+   if (inl(nic->ioaddr + VIRTIO_PCI_QUEUE_PFN)) {
+           printf("ERROR: queue already active\n");
+           return -1;
+   }
+
+   /* initialize the queue */
+
+   vring_init(vr, num, (unsigned char*)&queue[queue_index]);
+
+   /* activate the queue
+    *
+    * NOTE: vr->desc is initialized by vring_init()
+    */
+
+   outl((unsigned long)virt_to_phys(vr->desc) >> PAGE_SHIFT,
+        nic->ioaddr + VIRTIO_PCI_QUEUE_PFN);
+
+   return num;
+}
+
+/*
+ * Virtual ring management
+ *
+ */
+
+static void vring_enable_cb(int queue_index)
+{
+   vring[queue_index].avail->flags &= ~VRING_AVAIL_F_NO_INTERRUPT;
+}
+
+static void vring_disable_cb(int queue_index)
+{
+   vring[queue_index].avail->flags |= VRING_AVAIL_F_NO_INTERRUPT;
+}
+
+/*
+ * vring_free
+ *
+ * put at the begin of the free list the current desc[head]
+ */
+
+static void vring_detach(int queue_index, unsigned int head)
+{
+   struct vring *vr = &vring[queue_index];
+        unsigned int i;
+
+        /* find end of given descriptor */
+
+   i = head;
+   while (vr->desc[i].flags & VRING_DESC_F_NEXT)
+           i = vr->desc[i].next;
+
+   /* link it with free list and point to it */
+
+   vr->desc[i].next = free_head[queue_index];
+   wmb();
+   free_head[queue_index] = head;
+}
+
+/*
+ * vring_more_used
+ *
+ * is there some used buffers ?
+ *
+ */
+
+static inline int vring_more_used(int queue_index)
+{
+   wmb();
+   return last_used_idx[queue_index] != vring[queue_index].used->idx;
+}
+
+/*
+ * vring_get_buf
+ *
+ * get a buffer from the used list
+ *
+ */
+
+static int vring_get_buf(int queue_index, unsigned int *len)
+{
+   struct vring *vr = &vring[queue_index];
+   struct vring_used_elem *elem;
+   u32 id;
+   int ret;
+
+   elem = &vr->used->ring[last_used_idx[queue_index] % vr->num];
+   wmb();
+   id = elem->id;
+   if (len != NULL)
+           *len = elem->len;
+
+   ret = vdata[queue_index][id];
+
+   vring_detach(queue_index, id);
+
+   last_used_idx[queue_index]++;
+
+   return ret;
+}
+
+static void vring_add_buf(int queue_index, int index, int num_added)
+{
+   struct vring *vr = &vring[queue_index];
+   int i, avail, head;
+
+   BUG_ON(queue_index >= QUEUE_NB);
+
+   head = free_head[queue_index];
+   i = head;
+
+   if (queue_index == TX_INDEX) {
+
+           BUG_ON(index != 0);
+
+           /* add header into vring */
+
+           vr->desc[i].flags = VRING_DESC_F_NEXT;
+           vr->desc[i].addr = (u64)virt_to_phys(&tx_virtio_hdr);
+           vr->desc[i].len = sizeof(struct virtio_net_hdr);
+           i = vr->desc[i].next;
+
+           /* add frame buffer into vring */
+
+           vr->desc[i].flags = 0;
+           vr->desc[i].addr = (u64)virt_to_phys(&tx_eth_frame);
+           vr->desc[i].len = ETH_FRAME_LEN;
+           i = vr->desc[i].next;
+
+   } else if (queue_index == RX_INDEX) {
+
+           BUG_ON(index >= RX_BUF_NB);
+
+           /* add header into vring */
+
+           vr->desc[i].flags = VRING_DESC_F_NEXT|VRING_DESC_F_WRITE;
+           vr->desc[i].addr = (u64)virt_to_phys(&rx_hdr[index]);
+           vr->desc[i].len = sizeof(struct virtio_net_hdr);
+           i = vr->desc[i].next;
+
+           /* add frame buffer into vring */
+
+           vr->desc[i].flags = VRING_DESC_F_WRITE;
+           vr->desc[i].addr = (u64)virt_to_phys(&rx_buffer[index]);
+           vr->desc[i].len = ETH_FRAME_LEN;
+           i = vr->desc[i].next;
+   }
+
+   free_head[queue_index] = i;
+
+   vdata[queue_index][head] = index;
+
+   avail = (vr->avail->idx + num_added) % vr->num;
+   vr->avail->ring[avail] = head;
+   wmb();
+}
+
+static void vring_kick(struct nic *nic, int queue_index, int num_added)
+{
+   struct vring *vr = &vring[queue_index];
+
+   wmb();
+   vr->avail->idx += num_added;
+
+   mb();
+   if (!(vr->used->flags & VRING_USED_F_NO_NOTIFY))
+           vp_notify(nic, queue_index);
+}
+
+/*
+ * virtnet_disable
+ *
+ * Turn off ethernet interface
+ *
+ */
+
+static void virtnet_disable(struct nic *nic)
+{
+   int i;
+
+   for (i = 0; i < QUEUE_NB; i++) {
+           vring_disable_cb(i);
+           vp_del_vq(nic, i);
+   }
+   vp_reset(nic);
+}
+
+/*
+ * virtnet_poll
+ *
+ * Wait for a frame
+ *
+ * return true if there is a packet ready to read
+ *
+ * nic->packet should contain data on return
+ * nic->packetlen should contain length of data
+ *
+ */
+static int virtnet_poll(struct nic *nic, int retrieve)
+{
+   unsigned int len;
+   u16 token;
+   struct virtio_net_hdr *hdr;
+
+   if (!vring_more_used(RX_INDEX))
+           return 0;
+
+   if (!retrieve)
+           return 1;
+
+   token = vring_get_buf(RX_INDEX, &len);
+
+   BUG_ON(len > sizeof(struct virtio_net_hdr) + ETH_FRAME_LEN);
+
+   hdr = &rx_hdr[token];   /* FIXME: check flags */
+   len -= sizeof(struct virtio_net_hdr);
+
+        nic->packetlen = len;
+   memcpy(nic->packet, (char *)rx_buffer[token], nic->packetlen);
+
+   /* add buffer to desc */
+
+   vring_add_buf(RX_INDEX, token, 0);
+   vring_kick(nic, RX_INDEX, 1);
+
+   return 1;
+}
+
+/*
+ *
+ * virtnet_transmit
+ *
+ * Transmit a frame
+ *
+ */
+
+static void virtnet_transmit(struct nic *nic, const char *destaddr,
+        unsigned int type, unsigned int len, const char *data)
+{
+   /*
+    * from http://www.etherboot.org/wiki/dev/devmanual :
+    *     "You do not need more than one transmit buffer."
+    */
+
+   /* FIXME: initialize header according to vp_get_features() */
+
+   tx_virtio_hdr.flags = 0;
+   tx_virtio_hdr.csum_offset = 0;
+   tx_virtio_hdr.csum_start = 0;
+   tx_virtio_hdr.gso_type = VIRTIO_NET_HDR_GSO_NONE;
+   tx_virtio_hdr.gso_size = 0;
+   tx_virtio_hdr.hdr_len = 0;
+
+   /* add ethernet frame into vring */
+
+   BUG_ON(len > sizeof(tx_eth_frame.data));
+
+   memcpy(tx_eth_frame.hdr.dst_addr, destaddr, ETH_ALEN);
+   memcpy(tx_eth_frame.hdr.src_addr, nic->node_addr, ETH_ALEN);
+   tx_eth_frame.hdr.type = htons(type);
+   memcpy(tx_eth_frame.data, data, len);
+
+   vring_add_buf(TX_INDEX, 0, 0);
+
+   /*
+    * http://www.etherboot.org/wiki/dev/devmanual
+    *
+    *   "You should ensure the packet is fully transmitted
+    *    before returning from this routine"
+    */
+
+   while (vring_more_used(TX_INDEX)) {
+           mb();
+           udelay(10);
+   }
+
+   vring_kick(nic, TX_INDEX, 1);
+
+   /* free desc */
+
+   (void)vring_get_buf(TX_INDEX, NULL);
+}
+
+static void virtnet_irq(struct nic *nic __unused, irq_action_t action)
+{
+   switch ( action ) {
+   case DISABLE :
+           vring_disable_cb(RX_INDEX);
+           vring_disable_cb(TX_INDEX);
+           break;
+   case ENABLE :
+           vring_enable_cb(RX_INDEX);
+           vring_enable_cb(TX_INDEX);
+           break;
+   case FORCE :
+           break;
+   }
+}
+
+static void provide_buffers(struct nic *nic)
+{
+   int i;
+
+   for (i = 0; i < RX_BUF_NB; i++)
+           vring_add_buf(RX_INDEX, i, i);
+
+   /* nofify */
+
+   vring_kick(nic, RX_INDEX, i);
+}
+
+static struct nic_operations virtnet_operations = {
+	.connect = dummy_connect,
+	.poll = virtnet_poll,
+	.transmit = virtnet_transmit,
+	.irq = virtnet_irq,
+};
+
+/*
+ * virtnet_probe
+ *
+ * Look for a virtio network adapter
+ *
+ */
+
+static int virtnet_probe(struct nic *nic, struct pci_device *pci)
+{
+   u32 features;
+   int i;
+
+   /* Mask the bit that says "this is an io addr" */
+
+   nic->ioaddr = pci->ioaddr & ~3;
+
+   /* Copy IRQ from PCI information */
+
+   nic->irqno = pci->irq;
+
+   printf("I/O address 0x%08x, IRQ #%d\n", nic->ioaddr, nic->irqno);
+
+   adjust_pci_device(pci);
+
+   vp_reset(nic);
+
+   features = vp_get_features(nic);
+   if (features & (1 << VIRTIO_NET_F_MAC)) {
+           vp_get(nic, offsetof(struct virtio_net_config, mac),
+                  nic->node_addr, ETH_ALEN);
+           printf("MAC address ");
+	   for (i = 0; i < ETH_ALEN; i++) {
+                   printf("%02x%c", nic->node_addr[i],
+                          (i == ETH_ALEN - 1) ? '\n' : ':');
+           }
+   }
+
+   /* initialize emit/receive queue */
+
+   for (i = 0; i < QUEUE_NB; i++) {
+           free_head[i] = 0;
+           last_used_idx[i] = 0;
+           memset((char*)&queue[i], 0, sizeof(queue[i]));
+           if (vp_find_vq(nic, i) == -1)
+                   printf("Cannot register queue #%d\n", i);
+   }
+
+   /* provide some receive buffers */
+
+    provide_buffers(nic);
+
+   /* define NIC interface */
+
+    nic->nic_op = &virtnet_operations;
+
+   /* driver is ready */
+
+   vp_set_features(nic, features & (1 << VIRTIO_NET_F_MAC));
+   vp_set_status(nic, VIRTIO_CONFIG_S_DRIVER | VIRTIO_CONFIG_S_DRIVER_OK);
+
+   return 1;
+}
+
+static struct pci_device_id virtnet_nics[] = {
+PCI_ROM(0x1af4, 0x1000, "virtio-net",              "Virtio Network Interface"),
+};
+
+PCI_DRIVER ( virtnet_driver, virtnet_nics, PCI_NO_CLASS );
+
+DRIVER ( "VIRTIO-NET", nic_driver, pci_driver, virtnet_driver,
+	 virtnet_probe, virtnet_disable );

src/drivers/net/virtio-net.h

@@ -0,0 +1,44 @@
+#ifndef _VIRTIO_NET_H_
+# define _VIRTIO_NET_H_
+
+/* The feature bitmap for virtio net */
+#define VIRTIO_NET_F_CSUM       0       /* Host handles pkts w/ partial csum */
+#define VIRTIO_NET_F_GUEST_CSUM 1       /* Guest handles pkts w/ partial csum */
+#define VIRTIO_NET_F_MAC        5       /* Host has given MAC address. */
+#define VIRTIO_NET_F_GSO        6       /* Host handles pkts w/ any GSO type */
+#define VIRTIO_NET_F_GUEST_TSO4 7       /* Guest can handle TSOv4 in. */
+#define VIRTIO_NET_F_GUEST_TSO6 8       /* Guest can handle TSOv6 in. */
+#define VIRTIO_NET_F_GUEST_ECN  9       /* Guest can handle TSO[6] w/ ECN in. */
+#define VIRTIO_NET_F_GUEST_UFO  10      /* Guest can handle UFO in. */
+#define VIRTIO_NET_F_HOST_TSO4  11      /* Host can handle TSOv4 in. */
+#define VIRTIO_NET_F_HOST_TSO6  12      /* Host can handle TSOv6 in. */
+#define VIRTIO_NET_F_HOST_ECN   13      /* Host can handle TSO[6] w/ ECN in. */
+#define VIRTIO_NET_F_HOST_UFO   14      /* Host can handle UFO in. */
+
+struct virtio_net_config
+{
+   /* The config defining mac address (if VIRTIO_NET_F_MAC) */
+   u8 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 /* _VIRTIO_NET_H_ */

src/drivers/net/virtio-pci.h

@@ -0,0 +1,94 @@
+#ifndef _VIRTIO_PCI_H_
+# define _VIRTIO_PCI_H_
+
+/* A 32-bit r/o bitmask of the features supported by the host */
+#define VIRTIO_PCI_HOST_FEATURES        0
+
+/* A 32-bit r/w bitmask of features activated by the guest */
+#define VIRTIO_PCI_GUEST_FEATURES       4
+
+/* A 32-bit r/w PFN for the currently selected queue */
+#define VIRTIO_PCI_QUEUE_PFN            8
+
+/* A 16-bit r/o queue size for the currently selected queue */
+#define VIRTIO_PCI_QUEUE_NUM            12
+
+/* A 16-bit r/w queue selector */
+#define VIRTIO_PCI_QUEUE_SEL            14
+
+/* A 16-bit r/w queue notifier */
+#define VIRTIO_PCI_QUEUE_NOTIFY         16
+
+/* An 8-bit device status register.  */
+#define VIRTIO_PCI_STATUS               18
+
+/* An 8-bit r/o interrupt status register.  Reading the value will return the
+ * current contents of the ISR and will also clear it.  This is effectively
+ * a read-and-acknowledge. */
+#define VIRTIO_PCI_ISR                  19
+
+/* The bit of the ISR which indicates a device configuration change. */
+#define VIRTIO_PCI_ISR_CONFIG           0x2
+
+/* The remaining space is defined by each driver as the per-driver
+ * configuration space */
+#define VIRTIO_PCI_CONFIG               20
+
+/* Virtio ABI version, this must match exactly */
+#define VIRTIO_PCI_ABI_VERSION          0
+
+static inline u32 vp_get_features(struct nic *nic)
+{
+   return inl(nic->ioaddr + VIRTIO_PCI_HOST_FEATURES);
+}
+
+static inline void vp_set_features(struct nic *nic, u32 features)
+{
+        outl(features, nic->ioaddr + VIRTIO_PCI_GUEST_FEATURES);
+}
+
+static inline void vp_get(struct nic *nic, unsigned offset,
+                     void *buf, unsigned len)
+{
+   u8 *ptr = buf;
+   unsigned i;
+
+   for (i = 0; i < len; i++)
+           ptr[i] = inb(nic->ioaddr + VIRTIO_PCI_CONFIG + offset + i);
+}
+
+static inline u8 vp_get_status(struct nic *nic)
+{
+   return inb(nic->ioaddr + VIRTIO_PCI_STATUS);
+}
+
+static inline void vp_set_status(struct nic *nic, u8 status)
+{
+   if (status == 0)        /* reset */
+           return;
+        outb(status, nic->ioaddr + VIRTIO_PCI_STATUS);
+}
+
+
+static inline void vp_reset(struct nic *nic)
+{
+   outb(0, nic->ioaddr + VIRTIO_PCI_STATUS);
+   (void)inb(nic->ioaddr + VIRTIO_PCI_ISR);
+}
+
+static inline void vp_notify(struct nic *nic, int queue_index)
+{
+   outw(queue_index, nic->ioaddr + VIRTIO_PCI_QUEUE_NOTIFY);
+}
+
+static inline void vp_del_vq(struct nic *nic, int queue_index)
+{
+   /* select the queue */
+
+   outw(queue_index, nic->ioaddr + VIRTIO_PCI_QUEUE_SEL);
+
+   /* deactivate the queue */
+
+   outl(0, nic->ioaddr + VIRTIO_PCI_QUEUE_PFN);
+}
+#endif /* _VIRTIO_PCI_H_ */

src/drivers/net/virtio-ring.h

@@ -0,0 +1,93 @@
+#ifndef _VIRTIO_RING_H_
+# define _VIRTIO_RING_H_
+#define PAGE_SHIFT (12)
+#define PAGE_SIZE  (1<<PAGE_SHIFT)
+#define PAGE_MASK  (PAGE_SIZE-1)
+
+/* Status byte for guest to report progress, and synchronize features. */
+/* We have seen device and processed generic fields (VIRTIO_CONFIG_F_VIRTIO) */
+#define VIRTIO_CONFIG_S_ACKNOWLEDGE     1
+/* We have found a driver for the device. */
+#define VIRTIO_CONFIG_S_DRIVER          2
+/* Driver has used its parts of the config, and is happy */
+#define VIRTIO_CONFIG_S_DRIVER_OK       4
+/* We've given up on this device. */
+#define VIRTIO_CONFIG_S_FAILED          0x80
+
+#define MAX_QUEUE_NUM      (512)
+
+#define VRING_DESC_F_NEXT  1
+#define VRING_DESC_F_WRITE 2
+
+#define VRING_AVAIL_F_NO_INTERRUPT 1
+
+#define VRING_USED_F_NO_NOTIFY     1
+
+struct vring_desc
+{
+   u64 addr;
+   u32 len;
+   u16 flags;
+   u16 next;
+};
+
+struct vring_avail
+{
+   u16 flags;
+   u16 idx;
+   u16 ring[0];
+};
+
+struct vring_used_elem
+{
+   u32 id;
+   u32 len;
+};
+
+struct vring_used
+{
+   u16 flags;
+   u16 idx;
+   struct vring_used_elem ring[];
+};
+
+struct vring {
+   unsigned int num;
+   struct vring_desc *desc;
+   struct vring_avail *avail;
+   struct vring_used *used;
+};
+
+static inline void vring_init(struct vring *vr,
+                         unsigned int num, unsigned char *queue)
+{
+   unsigned int i;
+   unsigned long pa;
+
+        vr->num = num;
+
+   /* physical address of desc must be page aligned */
+
+   pa = virt_to_phys(queue);
+   pa = (pa + PAGE_MASK) & ~PAGE_MASK;
+   vr->desc = phys_to_virt(pa);
+
+        vr->avail = (struct vring_avail *)&vr->desc[num];
+
+   /* physical address of used must be page aligned */
+
+   pa = virt_to_phys(&vr->avail->ring[num]);
+   pa = (pa + PAGE_MASK) & ~PAGE_MASK;
+        vr->used = phys_to_virt(pa);
+
+   for (i = 0; i < num - 1; i++)
+           vr->desc[i].next = i + 1;
+   vr->desc[i].next = 0;
+}
+
+#define vring_size(num) \
+   (((((sizeof(struct vring_desc) * num) + \
+      (sizeof(struct vring_avail) + sizeof(u16) * num)) \
+         + PAGE_MASK) & ~PAGE_MASK) + \
+         (sizeof(struct vring_used) + sizeof(struct vring_used_elem) * num))
+#endif /* _VIRTIO_RING_H_ */