Signed-off-by: Gregory Haskins <[EMAIL PROTECTED]>
---
drivers/kvm/Kconfig | 5
drivers/kvm/Makefile | 2
drivers/kvm/ioqnet_host.c | 556 +++++++++++++++++++++++++++++++++++++++++++++
3 files changed, 563 insertions(+), 0 deletions(-)
diff --git a/drivers/kvm/Kconfig b/drivers/kvm/Kconfig
index cb674bb..1c884c5 100644
--- a/drivers/kvm/Kconfig
+++ b/drivers/kvm/Kconfig
@@ -62,6 +62,11 @@ config KVM_PVBUS_HOST
of the hypervisor itself. You only need this option if you plan to
run PVBUS based PV guests in KVM.
+config KVM_IOQNET
+ boolean "IOQNET host support"
+ depends on KVM
+ select KVM_PVBUS_HOST
+
config KVM_NET_HOST
tristate "Para virtual network host device"
depends on KVM
diff --git a/drivers/kvm/Makefile b/drivers/kvm/Makefile
index 8926fa9..66e5272 100644
--- a/drivers/kvm/Makefile
+++ b/drivers/kvm/Makefile
@@ -22,3 +22,5 @@ kvm-net-host-objs = kvm_net_host.o
obj-$(CONFIG_KVM_NET_HOST) += kvm_net_host.o
kvm-pvbus-objs := ioq_guest.o pvbus_guest.o
obj-$(CONFIG_KVM_PVBUS_GUEST) += kvm-pvbus.o
+kvm-ioqnet-objs := ioqnet_host.o
+obj-$(CONFIG_KVM_IOQNET) += kvm-ioqnet.o
\ No newline at end of file
diff --git a/drivers/kvm/ioqnet_host.c b/drivers/kvm/ioqnet_host.c
new file mode 100644
index 0000000..aff7e5c
--- /dev/null
+++ b/drivers/kvm/ioqnet_host.c
@@ -0,0 +1,556 @@
+/*
+ * Copyright 2007 Novell. All Rights Reserved.
+ *
+ * ioqnet - A paravirtualized network device based on the IOQ interface.
+ *
+ * This module represents the backend driver for an IOQNET driver on the KVM
+ * platform.
+ *
+ * Author:
+ * Gregory Haskins <[EMAIL PROTECTED]>
+ *
+ * Derived in part from the SNULL example from the book "Linux Device
+ * Drivers" by Alessandro Rubini and Jonathan Corbet, published
+ * by O'Reilly & Associates.
+ *
+ * This file is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License
+ * as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA.
+ */
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/moduleparam.h>
+
+#include <linux/sched.h>
+#include <linux/kernel.h> /* printk() */
+#include <linux/slab.h> /* kmalloc() */
+#include <linux/errno.h> /* error codes */
+#include <linux/types.h> /* size_t */
+#include <linux/interrupt.h> /* mark_bh */
+
+#include <linux/in.h>
+#include <linux/netdevice.h> /* struct device, and other headers */
+#include <linux/etherdevice.h> /* eth_type_trans */
+#include <linux/ip.h> /* struct iphdr */
+#include <linux/tcp.h> /* struct tcphdr */
+#include <linux/skbuff.h>
+#include <linux/ioq.h>
+#include <linux/pvbus.h>
+
+#include <linux/in6.h>
+#include <asm/checksum.h>
+#include <linux/ioq.h>
+#include <linux/ioqnet.h>
+#include <linux/highmem.h>
+
+#include "pvbus_host.h"
+#include "kvm.h"
+
+MODULE_AUTHOR("Gregory Haskins");
+MODULE_LICENSE("GPL");
+
+#define IOQNET_NAME "ioqnet"
+
+/*
+ * FIXME: Any "BUG_ON" code that can be triggered by a malicious guest must
+ * be turned into an inject_gp()
+ */
+
+struct ioqnet_queue {
+ struct ioq *queue;
+ struct ioq_notifier notifier;
+};
+
+struct ioqnet_priv {
+ spinlock_t lock;
+ struct kvm *kvm;
+ struct kvm_pv_device pvdev;
+ struct net_device *netdev;
+ struct net_device_stats stats;
+ struct ioqnet_queue rxq;
+ struct ioqnet_queue txq;
+ struct tasklet_struct txtask;
+ int connected;
+ int opened;
+};
+
+#undef PDEBUG /* undef it, just in case */
+#ifdef IOQNET_DEBUG
+# define PDEBUG(fmt, args...) printk( KERN_DEBUG "ioqnet: " fmt, ## args)
+#else
+# define PDEBUG(fmt, args...) /* not debugging: nothing */
+#endif
+
+/*
+ * Enable and disable receive interrupts.
+ */
+static void ioqnet_rx_ints(struct net_device *dev, int enable)
+{
+ struct ioqnet_priv *priv = netdev_priv(dev);
+ struct ioq *ioq = priv->rxq.queue;
+
+ if (priv->connected) {
+ if (enable)
+ ioq_start(ioq, 0);
+ else
+ ioq_stop(ioq, 0);
+ }
+}
+
+/*
+ * Open and close
+ */
+
+int ioqnet_open(struct net_device *dev)
+{
+ struct ioqnet_priv *priv = netdev_priv(dev);
+
+ priv->opened = 1;
+ netif_start_queue(dev);
+
+ return 0;
+}
+
+int ioqnet_release(struct net_device *dev)
+{
+ struct ioqnet_priv *priv = netdev_priv(dev);
+
+ priv->opened = 0;
+ netif_stop_queue(dev);
+
+ return 0;
+}
+
+/*
+ * Configuration changes (passed on by ifconfig)
+ */
+int ioqnet_config(struct net_device *dev, struct ifmap *map)
+{
+ if (dev->flags & IFF_UP) /* can't act on a running interface */
+ return -EBUSY;
+
+ /* Don't allow changing the I/O address */
+ if (map->base_addr != dev->base_addr) {
+ printk(KERN_WARNING "ioqnet: Can't change I/O address\n");
+ return -EOPNOTSUPP;
+ }
+
+ /* ignore other fields */
+ return 0;
+}
+
+/*
+ * The poll implementation.
+ */
+static int ioqnet_poll(struct net_device *dev, int *budget)
+{
+ int npackets = 0, quota = min(dev->quota, *budget);
+ struct ioqnet_priv *priv = netdev_priv(dev);
+ struct ioq_iterator iter;
+ unsigned long flags;
+ int ret;
+
+ if (!priv->connected)
+ return 0;
+
+ spin_lock_irqsave(&priv->lock, flags);
+
+ /* We want to iterate on the tail of the in-use index */
+ ret = ioq_iter(priv->rxq.queue, &iter, ioq_idxtype_inuse, 0);
+ BUG_ON(ret < 0);
+
+ ret = ioq_iter_seek(&iter, ioq_seek_tail, 0, 0);
+ BUG_ON(ret < 0);
+
+ /*
+ * We stop if we have met the quota or there are no more packets.
+ * The EOM is indicated by finding a packet that is still owned by
+ * the north side
+ */
+ while ((npackets < quota) && iter.desc->sown) {
+ struct ioq_ring_desc *desc = iter.desc;
+ struct ioqnet_tx_ptr *ptr = gpa_to_hva(priv->kvm, desc->ptr);
+ struct sk_buff *skb;
+ int i;
+ size_t len = 0;
+
+ /* First figure out how much of an skb we need */
+ for (i = 0; i < desc->alen; ++i) {
+ len += ptr[i].len;
+ }
+
+ skb = dev_alloc_skb(len + 2);
+ if (!skb) {
+ /* FIXME: This leaks... */
+ printk(KERN_ERR "FATAL: Out of memory on IOQNET\n");
+ netif_stop_queue(dev);
+ return -ENOMEM;
+ }
+
+ skb_reserve(skb, 2);
+
+ /* Then copy the data out to our fresh SKB */
+ for (i = 0; i < desc->alen; ++i) {
+ struct ioqnet_tx_ptr *p = &ptr[i];
+ void *d = gpa_to_hva(priv->kvm,
+ p->data);
+
+ memcpy(skb_push(skb, p->len), d, p->len);
+ kunmap(d);
+ }
+
+ /* Maintain stats */
+ npackets++;
+ priv->stats.rx_packets++;
+ priv->stats.rx_bytes += len;
+
+ /* Pass the buffer up to the stack */
+ skb->dev = dev;
+ skb->protocol = eth_type_trans(skb, dev);
+ netif_receive_skb(skb);
+
+ /* Advance the in-use tail */
+ desc->sown = 0;
+ ret = ioq_iter_pop(&iter, 0);
+ BUG_ON(ret < 0);
+
+ /* Toggle the lock */
+ spin_unlock_irqrestore(&priv->lock, flags);
+ spin_lock_irqsave(&priv->lock, flags);
+ }
+
+ /*
+ * If we processed all packets, we're done; tell the kernel and
+ * reenable ints
+ */
+ *budget -= npackets;
+ dev->quota -= npackets;
+ if (ioq_empty(priv->rxq.queue, ioq_idxtype_inuse)) {
+ /* FIXME: there is a race with enabling interrupts */
+ netif_rx_complete(dev);
+ ioqnet_rx_ints(dev, 1);
+ ret = 0;
+ } else
+ /* We couldn't process everything. */
+ ret = 1;
+
+ spin_unlock_irqrestore(&priv->lock, flags);
+
+ /* And let the north side know that we changed the rx-queue */
+ ioq_signal(priv->rxq.queue, 0);
+
+ return ret;
+}
+
+/*
+ * Transmit a packet (called by the kernel)
+ */
+int ioqnet_tx(struct sk_buff *skb, struct net_device *dev)
+{
+ struct ioqnet_priv *priv = netdev_priv(dev);
+ struct ioq_iterator iter;
+ int ret;
+ unsigned long flags;
+ char *data;
+
+ if (skb->len < ETH_ZLEN)
+ return -EINVAL;
+
+ if (!priv->connected)
+ return 0;
+
+ spin_lock_irqsave(&priv->lock, flags);
+
+ if (ioq_full(priv->txq.queue, ioq_idxtype_valid)) {
+ /*
+ * We must flow-control the kernel by disabling the queue
+ */
+ spin_unlock_irqrestore(&priv->lock, flags);
+ netif_stop_queue(dev);
+ return 0;
+ }
+
+ /*
+ * We want to iterate on the head of the "inuse" index
+ */
+ ret = ioq_iter(priv->txq.queue, &iter, ioq_idxtype_inuse, 0);
+ BUG_ON(ret < 0);
+
+ ret = ioq_iter_seek(&iter, ioq_seek_head, 0, 0);
+ BUG_ON(ret < 0);
+
+ if (skb->len > iter.desc->len)
+ return -EINVAL;
+
+ dev->trans_start = jiffies; /* save the timestamp */
+
+ /* Copy the data to the north-side buffer */
+ data = (char*)gpa_to_hva(priv->kvm, iter.desc->ptr);
+ memcpy(data, skb->data, skb->len);
+ kunmap(data);
+
+ /* Give ownership back to the north */
+ iter.desc->sown = 0;
+
+ /* Advance the index */
+ ret = ioq_iter_push(&iter, 0);
+ BUG_ON(ret < 0);
+
+ /*
+ * This will signal the north side to consume the packet
+ */
+ ioq_signal(priv->txq.queue, 0);
+
+ spin_unlock_irqrestore(&priv->lock, flags);
+
+ return 0;
+}
+
+void ioqnet_tx_intr(unsigned long data)
+{
+ struct ioqnet_priv *priv = (struct ioqnet_priv*)data;
+ unsigned long flags;
+
+ spin_lock_irqsave(&priv->lock, flags);
+
+ /*
+ * If we were previously stopped due to flow control, restart the
+ * processing
+ */
+ if (netif_queue_stopped(priv->netdev)
+ && !ioq_full(priv->txq.queue, ioq_idxtype_inuse)) {
+
+ netif_wake_queue(priv->netdev);
+ }
+
+ spin_unlock_irqrestore(&priv->lock, flags);
+}
+
+/*
+ * Ioctl commands
+ */
+int ioqnet_ioctl(struct net_device *dev, struct ifreq *rq, int cmd)
+{
+ PDEBUG("ioctl\n");
+ return 0;
+}
+
+/*
+ * Return statistics to the caller
+ */
+struct net_device_stats *ioqnet_stats(struct net_device *dev)
+{
+ struct ioqnet_priv *priv = netdev_priv(dev);
+ return &priv->stats;
+}
+
+static void ioq_rx_notify(struct ioq_notifier *notifier)
+{
+ struct ioqnet_priv *priv;
+ struct net_device *dev;
+
+ priv = container_of(notifier, struct ioqnet_priv, rxq.notifier);
+ dev = priv->netdev;
+
+ ioqnet_rx_ints(dev, 0); /* Disable further interrupts */
+ netif_rx_schedule(dev);
+}
+
+static void ioq_tx_notify(struct ioq_notifier *notifier)
+{
+ struct ioqnet_priv *priv;
+
+ priv = container_of(notifier, struct ioqnet_priv, txq.notifier);
+
+ tasklet_schedule(&priv->txtask);
+}
+
+/*
+ * The init function (sometimes called probe).
+ * It is invoked by register_netdev()
+ */
+void ioqnet_init(struct net_device *dev)
+{
+ ether_setup(dev); /* assign some of the fields */
+
+ dev->open = ioqnet_open;
+ dev->stop = ioqnet_release;
+ dev->set_config = ioqnet_config;
+ dev->hard_start_xmit = ioqnet_tx;
+ dev->do_ioctl = ioqnet_ioctl;
+ dev->get_stats = ioqnet_stats;
+ dev->poll = ioqnet_poll;
+ dev->weight = 2;
+ dev->hard_header_cache = NULL; /* Disable caching */
+
+ /* We go "link down" until the guest connects to us */
+ netif_carrier_off(dev);
+
+}
+
+
+/* -------------------------------------------------------------- */
+
+static inline struct ioqnet_priv* to_priv(struct kvm_pv_device *t)
+{
+ return container_of(t, struct ioqnet_priv, pvdev);
+}
+
+
+static int ioqnet_connect(struct ioqnet_priv *priv,
+ ioq_id_t id,
+ struct ioqnet_queue *q,
+ void (*func)(struct ioq_notifier*))
+{
+ int ret;
+ struct ioq_mgr *ioqmgr = priv->kvm->ioqmgr;
+
+ ret = ioqmgr->connect(ioqmgr, id, &q->queue, 0);
+ if (ret < 0)
+ return ret;
+
+ q->notifier.signal = func;
+
+ return 0;
+}
+
+static int ioqnet_pvbus_connect(struct ioqnet_priv *priv,
+ void *data, size_t len)
+{
+ struct ioqnet_connect *cnct = (struct ioqnet_connect*)data;
+ int ret;
+
+ /* We connect the north's rxq to our txq */
+ ret = ioqnet_connect(priv, cnct->rxq, &priv->txq, ioq_tx_notify);
+ if (ret < 0)
+ return ret;
+
+ /* And vice-versa */
+ ret = ioqnet_connect(priv, cnct->txq, &priv->rxq, ioq_rx_notify);
+ if (ret < 0)
+ return ret;
+
+ /*
+ * So now that the guest has connected we can send a "link up" event
+ * to the kernel.
+ */
+ netif_carrier_on(priv->netdev);
+
+ priv->connected = 1;
+
+ return 0;
+}
+
+static int ioqnet_pvbus_query_mac(struct ioqnet_priv *priv,
+ void *data, size_t len)
+{
+ if (len != ETH_ALEN)
+ return -EINVAL;
+
+ memcpy(data, priv->netdev->dev_addr, ETH_ALEN);
+
+ return 0;
+}
+
+/*
+ * This function is invoked whenever a guest calls pvbus_ops->call() against
+ * our instance ID
+ */
+static int ioqnet_pvbus_device_call(struct kvm_pv_device *t, u32 func,
+ void *data, size_t len)
+{
+ struct ioqnet_priv *priv = to_priv(t);
+ int ret;
+
+ switch (func) {
+ case IOQNET_CONNECT:
+ ret = ioqnet_pvbus_connect(priv, data, len);
+ break;
+ case IOQNET_QUERY_MAC:
+ ret = ioqnet_pvbus_query_mac(priv, data, len);
+ break;
+ }
+
+ return ret;
+}
+
+static void ioqnet_pvbus_device_destroy(struct kvm_pv_device *t)
+{
+ return 0;
+}
+
+/*
+ * This function is invoked whenever someone instantiates an IOQNET object
+ */
+static int ioqnet_pvbus_devtype_create(struct kvm *kvm,
+ struct kvm_pv_devtype *t, u64 id,
+ const char *cfg,
+ struct kvm_pv_device **pvdev)
+{
+ struct net_device *dev;
+ struct ioqnet_priv *priv;
+
+ dev = alloc_netdev(sizeof(struct ioqnet_priv), "ioq%d",
+ ioqnet_init);
+ if (!dev)
+ return -ENOMEM;
+
+ priv = netdev_priv(dev);
+
+ memset(priv, 0, sizeof(*priv));
+
+ priv->pvdev.call = ioqnet_pvbus_device_call;
+ priv->pvdev.destroy = ioqnet_pvbus_device_destroy;
+ priv->pvdev.id = id;
+ priv->pvdev.ver = IOQNET_VERSION;
+
+ spin_lock_init(&priv->lock);
+ priv->kvm = kvm;
+ priv->netdev = dev;
+ tasklet_init(&priv->txtask, ioqnet_tx_intr, (unsigned long)priv);
+
+ ret = register_netdev(dev);
+ if (ret < 0) {
+ printk("ioqnet: error %i registering device \"%s\"\n",
+ ret, dev->name);
+ free_netdev(dev);
+ }
+
+ *pvdev = &priv->pvdev;
+
+ return 0;
+}
+
+static int ioqnet_pvbus_devtype_destroy(struct kvm_pv_devtype *t)
+{
+ return -ENOSYS;
+}
+
+static struct kvm_pv_devtype ioqnet_devtype = {
+ .create = ioqnet_pvbus_devtype_create,
+ .destroy = ioqnet_pvbus_devtype_destroy,
+ .name = IOQNET_NAME,
+};
+
+static int __init ioqnet_init_module(void)
+{
+ return kvm_pvbus_registertype(&ioqnet_devtype);
+}
+
+static void __exit ioqnet_cleanup_module(void)
+{
+ kvm_pvbus_unregistertype(IOQNET_NAME);
+}
+
+module_init(ioqnet_init_module);
+module_exit(ioqnet_cleanup_module);
-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems? Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >> http://get.splunk.com/
_______________________________________________
kvm-devel mailing list
kvm-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/kvm-devel
