static int
kni_sock_release(struct socket *sock)
{
struct kni_vhost_queue *q =
container_of(sock->sk, struct kni_vhost_queue, sk);
struct kni_dev *kni;
if (q == NULL)
return 0;
if (NULL != (kni = q->kni)) {
kni->vq_status = BE_STOP;
KNI_VHOST_WAIT_WQ_SAFE();
kni->vhost_queue = NULL;
q->kni = NULL;
}
if (q->sockfd != -1)
q->sockfd = -1;
sk_set_socket(&q->sk, NULL);
sock->sk = NULL;
sock_put(&q->sk);
KNI_DBG("dummy sock release done\n");
return 0;
}
static unsigned int
kni_sock_poll(struct file *file, struct socket *sock, poll_table * wait)
{
struct kni_vhost_queue *q =
container_of(sock->sk, struct kni_vhost_queue, sk);
struct kni_dev *kni;
unsigned int mask = 0;
if (unlikely(q == NULL || q->kni == NULL))
return POLLERR;
kni = q->kni;
KNI_DBG("start kni_poll on group %d, wq 0x%16llx\n",
kni->group_id, (uint64_t)sock->wq);
poll_wait(file, &sock->wq->wait, wait);
if (kni_fifo_count(kni->rx_q) > 0)
mask |= POLLIN | POLLRDNORM;
if (sock_writeable(&q->sk) ||
(!test_and_set_bit(SOCK_ASYNC_NOSPACE, &q->sock->flags) &&
sock_writeable(&q->sk)))
mask |= POLLOUT | POLLWRNORM;
return mask;
}
/*
* Ioctl commands
*/
static int
kni_net_ioctl(struct net_device *dev, struct ifreq *rq, int cmd)
{
KNI_DBG("kni_net_ioctl %d\n",
((struct kni_dev *)netdev_priv(dev))->group_id);
return 0;
}
int
kni_sock_getname (struct socket *sock,
struct sockaddr *addr,
int *sockaddr_len, int peer)
{
KNI_DBG("dummy sock getname\n");
((struct sockaddr_ll*)addr)->sll_family = AF_PACKET;
return 0;
}
/*
* Deal with a transmit timeout.
*/
static void
kni_net_tx_timeout (struct net_device *dev)
{
struct kni_dev *kni = netdev_priv(dev);
KNI_DBG("Transmit timeout at %ld, latency %ld\n", jiffies,
jiffies - dev->trans_start);
kni->stats.tx_errors++;
netif_wake_queue(dev);
return;
}
void
kni_net_lb_init(struct net_device *dev)
{
struct kni_dev *kni = netdev_priv(dev);
KNI_DBG("kni_net_init\n");
init_waitqueue_head(&kni->wq);
mutex_init(&kni->sync_lock);
kni->dev_addr = NULL;
kni_lb_setup(dev);
dev->netdev_ops = &kni_net_netdev_ops;
dev->watchdog_timeo = WD_TIMEOUT;
}
/*
* Ioctl commands
*/
static int
kni_net_ioctl(struct net_device *dev, struct ifreq *rq, int cmd)
{
struct kni_dev *kni = netdev_priv(dev);
if (kni->kni_released) {
return -EAGAIN;
}
KNI_DBG("kni_net_ioctl %d\n",
((struct kni_dev *)netdev_priv(dev))->group_id);
return 0;
}
int
kni_vhost_init(struct kni_dev *kni)
{
struct net_device *dev = kni->net_dev;
if (sysfs_create_group(&dev->dev.kobj, &dev_attr_grp))
sysfs_remove_group(&dev->dev.kobj, &dev_attr_grp);
kni->vq_status = BE_STOP;
KNI_DBG("kni_vhost_init done\n");
return 0;
}
void
kni_net_init(struct net_device *dev)
{
struct kni_dev *kni = netdev_priv(dev);
KNI_DBG("kni_net_init\n");
init_waitqueue_head(&kni->wq);
mutex_init(&kni->sync_lock);
ether_setup(dev); /* assign some of the fields */
dev->netdev_ops = &kni_net_netdev_ops;
dev->header_ops = &kni_net_header_ops;
dev->watchdog_timeo = WD_TIMEOUT;
}
int
kni_vhost_backend_release(struct kni_dev *kni)
{
struct kni_vhost_queue *q = kni->vhost_queue;
if (q == NULL)
return 0;
/* dettach from kni */
q->kni = NULL;
KNI_DBG("release backend done\n");
return 0;
}
static int
kni_net_change_mtu(struct net_device *dev, int new_mtu)
{
int ret;
struct rte_kni_request req;
struct kni_dev *kni = netdev_priv(dev);
KNI_DBG("kni_net_change_mtu new mtu %d to be set\n", new_mtu);
memset(&req, 0, sizeof(req));
req.req_id = RTE_KNI_REQ_CHANGE_MTU;
req.new_mtu = new_mtu;
ret = kni_net_process_request(kni, &req);
if (ret == 0 && req.result == 0)
dev->mtu = new_mtu;
return (ret == 0 ? req.result : ret);
}
static int
kni_vhost_backend_init(struct kni_dev *kni)
{
struct kni_vhost_queue *q;
struct net *net = current->nsproxy->net_ns;
int err, i, sockfd;
struct rte_kni_fifo *fifo;
struct sk_buff *elem;
if (kni->vhost_queue != NULL)
return -1;
if (!(q = (struct kni_vhost_queue *)sk_alloc(
net, AF_UNSPEC, GFP_KERNEL, &kni_raw_proto)))
return -ENOMEM;
err = sock_create_lite(AF_UNSPEC, SOCK_RAW, IPPROTO_RAW, &q->sock);
if (err)
goto free_sk;
sockfd = kni_sock_map_fd(q->sock);
if (sockfd < 0) {
err = sockfd;
goto free_sock;
}
/* cache init */
q->cache = (struct sk_buff*)
kzalloc(RTE_KNI_VHOST_MAX_CACHE_SIZE * sizeof(struct sk_buff),
GFP_KERNEL);
if (!q->cache)
goto free_fd;
fifo = (struct rte_kni_fifo*)
kzalloc(RTE_KNI_VHOST_MAX_CACHE_SIZE * sizeof(void *)
+ sizeof(struct rte_kni_fifo), GFP_KERNEL);
if (!fifo)
goto free_cache;
kni_fifo_init(fifo, RTE_KNI_VHOST_MAX_CACHE_SIZE);
for (i = 0; i < RTE_KNI_VHOST_MAX_CACHE_SIZE; i++) {
elem = &q->cache[i];
kni_fifo_put(fifo, (void**)&elem, 1);
}
q->fifo = fifo;
/* store sockfd in vhost_queue */
q->sockfd = sockfd;
/* init socket */
q->sock->type = SOCK_RAW;
q->sock->state = SS_CONNECTED;
q->sock->ops = &kni_socket_ops;
sock_init_data(q->sock, &q->sk);
/* init sock data */
q->sk.sk_write_space = kni_sk_write_space;
q->sk.sk_destruct = kni_sk_destruct;
q->flags = IFF_NO_PI | IFF_TAP;
q->vnet_hdr_sz = sizeof(struct virtio_net_hdr);
#ifdef RTE_KNI_VHOST_VNET_HDR_EN
q->flags |= IFF_VNET_HDR;
#endif
/* bind kni_dev with vhost_queue */
q->kni = kni;
kni->vhost_queue = q;
wmb();
kni->vq_status = BE_START;
#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,35)
KNI_DBG("backend init sockfd=%d, sock->wq=0x%16llx,"
"sk->sk_wq=0x%16llx",
q->sockfd, (uint64_t)q->sock->wq,
(uint64_t)q->sk.sk_wq);
#else
KNI_DBG("backend init sockfd=%d, sock->wait at 0x%16llx,"
"sk->sk_sleep=0x%16llx",
q->sockfd, (uint64_t)&q->sock->wait,
(uint64_t)q->sk.sk_sleep);
#endif
return 0;
free_cache:
kfree(q->cache);
q->cache = NULL;
free_fd:
put_unused_fd(sockfd);
free_sock:
q->kni = NULL;
kni->vhost_queue = NULL;
kni->vq_status |= BE_FINISH;
sock_release(q->sock);
q->sock->ops = NULL;
q->sock = NULL;
free_sk:
sk_free((struct sock*)q);
return err;
}
kni_sock_rcvmsg(struct socket *sock,
struct msghdr *m, size_t len, int flags)
#endif /* HAVE_KIOCB_MSG_PARAM */
{
int vnet_hdr_len = 0;
int pkt_len = 0;
struct kni_vhost_queue *q =
container_of(sock->sk, struct kni_vhost_queue, sk);
static struct virtio_net_hdr
__attribute__ ((unused)) vnet_hdr = {
.flags = 0,
.gso_type = VIRTIO_NET_HDR_GSO_NONE
};
if (unlikely(q == NULL || q->kni == NULL))
return 0;
#ifdef RTE_KNI_VHOST_VNET_HDR_EN
if (likely(q->flags & IFF_VNET_HDR)) {
vnet_hdr_len = q->vnet_hdr_sz;
if ((len -= vnet_hdr_len) < 0)
return -EINVAL;
}
#endif
if (unlikely(0 == (pkt_len = kni_vhost_net_rx(q->kni,
m, vnet_hdr_len, len))))
return 0;
#ifdef RTE_KNI_VHOST_VNET_HDR_EN
/* no need to copy hdr when no pkt received */
#ifdef HAVE_IOV_ITER_MSGHDR
if (unlikely(copy_to_iter((void *)&vnet_hdr, vnet_hdr_len,
&m->msg_iter)))
#else
if (unlikely(memcpy_toiovecend(m->msg_iov,
(void *)&vnet_hdr, 0, vnet_hdr_len)))
#endif /* HAVE_IOV_ITER_MSGHDR */
return -EFAULT;
#endif /* RTE_KNI_VHOST_VNET_HDR_EN */
KNI_DBG_RX("kni_rcvmsg expect_len %ld, flags 0x%08x, pkt_len %d\n",
(unsigned long)len, q->flags, pkt_len);
return (pkt_len + vnet_hdr_len);
}
/* dummy tap like ioctl */
static int
kni_sock_ioctl(struct socket *sock, unsigned int cmd,
unsigned long arg)
{
void __user *argp = (void __user *)arg;
struct ifreq __user *ifr = argp;
unsigned int __user *up = argp;
struct kni_vhost_queue *q =
container_of(sock->sk, struct kni_vhost_queue, sk);
struct kni_dev *kni;
unsigned int u;
int __user *sp = argp;
int s;
int ret;
KNI_DBG("tap ioctl cmd 0x%08x\n", cmd);
switch (cmd) {
case TUNSETIFF:
KNI_DBG("TUNSETIFF\n");
/* ignore the name, just look at flags */
if (get_user(u, &ifr->ifr_flags))
return -EFAULT;
ret = 0;
if ((u & ~IFF_VNET_HDR) != (IFF_NO_PI | IFF_TAP))
ret = -EINVAL;
else
q->flags = u;
return ret;
case TUNGETIFF:
KNI_DBG("TUNGETIFF\n");
rcu_read_lock_bh();
kni = rcu_dereference_bh(q->kni);
if (kni)
dev_hold(kni->net_dev);
rcu_read_unlock_bh();
if (!kni)
return -ENOLINK;
ret = 0;
if (copy_to_user(&ifr->ifr_name, kni->net_dev->name, IFNAMSIZ) ||
put_user(q->flags, &ifr->ifr_flags))
ret = -EFAULT;
dev_put(kni->net_dev);
return ret;
case TUNGETFEATURES:
KNI_DBG("TUNGETFEATURES\n");
u = IFF_TAP | IFF_NO_PI;
#ifdef RTE_KNI_VHOST_VNET_HDR_EN
u |= IFF_VNET_HDR;
#endif
if (put_user(u, up))
return -EFAULT;
return 0;
case TUNSETSNDBUF:
KNI_DBG("TUNSETSNDBUF\n");
if (get_user(u, up))
return -EFAULT;
q->sk.sk_sndbuf = u;
return 0;
case TUNGETVNETHDRSZ:
s = q->vnet_hdr_sz;
if (put_user(s, sp))
return -EFAULT;
KNI_DBG("TUNGETVNETHDRSZ %d\n", s);
return 0;
case TUNSETVNETHDRSZ:
if (get_user(s, sp))
return -EFAULT;
if (s < (int)sizeof(struct virtio_net_hdr))
return -EINVAL;
KNI_DBG("TUNSETVNETHDRSZ %d\n", s);
q->vnet_hdr_sz = s;
return 0;
case TUNSETOFFLOAD:
KNI_DBG("TUNSETOFFLOAD %lx\n", arg);
#ifdef RTE_KNI_VHOST_VNET_HDR_EN
/* not support any offload yet */
if (!(q->flags & IFF_VNET_HDR))
return -EINVAL;
return 0;
#else
return -EINVAL;
#endif
default:
KNI_DBG("NOT SUPPORT\n");
return -EINVAL;
}
}
