/*
*------------------------------------------------------------------
* Copyright (c) 2017 Cisco and/or its affiliates.
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at:
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*------------------------------------------------------------------
*/
#define _GNU_SOURCE
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <net/if.h>
#include <linux/if_tun.h>
#include <sys/ioctl.h>
#include <linux/virtio_net.h>
#include <linux/vhost.h>
#include <sys/eventfd.h>
#include <sched.h>
#include <linux/netlink.h>
#include <linux/rtnetlink.h>
#include <vlib/vlib.h>
#include <vlib/unix/unix.h>
#include <vnet/ethernet/ethernet.h>
#include <vnet/ip/ip4_packet.h>
#include <vnet/ip/ip6_packet.h>
#include <vnet/devices/netlink.h>
#include <vnet/devices/virtio/virtio.h>
#include <vnet/devices/tap/tap.h>
tap_main_t tap_main;
#define _IOCTL(fd,a,...) \
if (ioctl (fd, a, __VA_ARGS__) < 0) \
{ \
err = clib_error_return_unix (0, "ioctl(" #a ")"); \
goto error; \
}
static u32
virtio_eth_flag_change (vnet_main_t * vnm, vnet_hw_interface_t * hi,
u32 flags)
{
/* nothing for now */
//TODO On MTU change call vnet_netlink_set_if_mtu
return 0;
}
void vl_api_rpc_call_main_thread (void *fp, u8 * data, u32 data_length);
static clib_error_t *
call_tap_read_ready (clib_file_t * uf)
{
/* nothing to do */
return 0;
}
static void
tap_delete_if_cp (u32 * sw_if_index)
{
vlib_main_t *vm = vlib_get_main ();
tap_delete_if (vm, *sw_if_index);
}
/*
* Tap clean-up routine:
* Linux side of tap interface can be deleted i.e. tap is
* attached to container and if someone will delete this
* container, will also removes tap interface. While VPP
* will have other side of tap. This function will RPC
* main thread to call the tap_delete_if to cleanup tap.
*/
static clib_error_t *
call_tap_error_ready (clib_file_t * uf)
{
vl_api_rpc_call_main_thread (tap_delete_if_cp, (u8 *) & uf->private_data,
sizeof (uf->private_data));
return 0;
}
static int
open_netns_fd (char *netns)
{
u8 *s = 0;
int fd;
if (strncmp (netns, "pid:", 4) == 0)
s = format (0, "/proc/%u/ns/net%c", atoi (netns + 4), 0);
else if (netns[0] == '/')
s = format (0, "%s%c", netns, 0);
else
s = format (0, "/var/run/netns/%s%c", netns, 0);
fd = open ((char *) s, O_RDONLY);
vec_free (s);
return fd;
}
#define TAP_MAX_INSTANCE 1024
void
tap_create_if (vlib_main_t * vm, tap_create_if_args_t * args)
{
vnet_main_t *vnm = vnet_get_main ();
vlib_thread_main_t *thm = vlib_get_thread_main ();
virtio_main_t *vim = &virtio_main;
tap_main_t *tm = &tap_main;
vnet_sw_interface_t *sw;
vnet_hw_interface_t *hw;
int i;
int old_netns_fd = -1;
struct ifreq ifr;
size_t hdrsz;
struct vhost_memory *vhost_mem = 0;
virtio_if_t *vif = 0;
clib_file_t t = { 0 };
clib_error_t *err = 0;
int fd = -1;
if (args->id != ~0)
{
if (clib_bitmap_get (tm->tap_ids, args->id))
{
args->rv = VNET_API_ERROR_INVALID_INTERFACE;
args->error = clib_error_return (0, "interface already exists");
return;
}
}
else
{
args->id = clib_bitmap_first_clear (tm->tap_ids);
}
if (args->id > TAP_MAX_INSTANCE)
{
args->rv = VNET_API_ERROR_UNSPECIFIED;
args->error = clib_error_return (0, "cannot find free interface id");
return;
}
clib_memset (&ifr, 0, sizeof (ifr));
pool_get (vim->interfaces, vif);
vif->dev_instance = vif - vim->interfaces;
vif->tap_fd = -1;
vif->id = args->id;
if ((vif->fd = open ("/dev/vhost-net", O_RDWR | O_NONBLOCK)) < 0)
{
args->rv = VNET_API_ERROR_SYSCALL_ERROR_1;
args->error = clib_error_return_unix (0, "open '/dev/vhost-net'");
goto error;
}
_IOCTL (vif->fd, VHOST_GET_FEATURES, &vif->remote_features);
if ((vif->remote_features & VIRTIO_FEATURE (VIRTIO_NET_F_MRG_RXBUF)) == 0)
{
args->rv = VNET_API_ERROR_UNSUPPORTED;
args->error = clib_error_return (0, "vhost-net backend doesn't support "
"VIRTIO_NET_F_MRG_RXBUF feature");
goto error;
}
if ((vif->remote_features & VIRTIO_FEATURE (VIRTIO_RING_F_INDIRECT_DESC)) ==
0)
{
args->rv = VNET_API_ERROR_UNSUPPORTED;
args->error = clib_error_return (0, "vhost-net backend doesn't support "
"VIRTIO_RING_F_INDIRECT_DESC feature");
goto error;
}
if ((vif->remote_features & VIRTIO_FEATURE (VIRTIO_F_VERSION_1)) == 0)
{
args->rv = VNET_API_ERROR_UNSUPPORTED;
args->error = clib_error_return (0, "vhost-net backend doesn't support "
"VIRTIO_F_VERSION_1 features");
goto error;
}
vif->features |= VIRTIO_FEATURE (VIRTIO_NET_F_MRG_RXBUF);
vif->features |= VIRTIO_FEATURE (VIRTIO_F_VERSION_1);
vif->features |= VIRTIO_FEATURE (VIRTIO_RING_F_INDIRECT_DESC);
virtio_set_net_hdr_size (vif);
_IOCTL (vif->fd, VHOST_SET_FEATURES, &vif->features);
if ((vif->tap_fd = open ("/dev/net/tun", O_RDWR | O_NONBLOCK)) < 0)
{
args->rv = VNET_API_ERROR_SYSCALL_ERROR_2;
args->error = clib_error_return_unix (0, "open '/dev/net/tun'");
goto error;
}
ifr.ifr_flags = IFF_TAP | IFF_NO_PI | IFF_ONE_QUEUE | IFF_VNET_HDR;
_IOCTL (vif->tap_fd, TUNSETIFF, (void *) &ifr);
vif->ifindex = if_nametoindex (ifr.ifr_ifrn.ifrn_name);
if (!args->host_if_name)
args->host_if_name = (u8 *) ifr.ifr_ifrn.ifrn_name;
unsigned int offload = 0;
hdrsz = sizeof (struct virtio_net_hdr_v1);
if (args->tap_flags & TAP_FLAG_GSO)
{
offload = TUN_F_CSUM | TUN_F_TSO4 | TUN_F_TSO6;
vif->gso_enabled = 1;
}
else
{
vif->gso_enabled = 0;
}
_IOCTL (vif->tap_fd, TUNSETOFFLOAD, offload);
_IOCTL (vif->tap_fd, TUNSETVNETHDRSZ, &hdrsz);
_IOCTL (vif->fd, VHOST_SET_OWNER, 0);
/* if namespace is specified, all further netlink messages should be excuted
after we change our net namespace */
if (args->host_namespace)
{
old_netns_fd = open ("/proc/self/ns/net", O_RDONLY);
if ((fd = open_netns_fd ((char *) args->host_namespace)) == -1)
{
args->rv = VNET_API_ERROR_SYSCALL_ERROR_2;
args->error = clib_error_return_unix (0, "open_netns_fd '%s'",
args->host_namespace);
goto error;
}
args->error = vnet_netlink_set_link_netns (vif->ifindex, fd,
(char *) args->host_if_name);
if (args->error)
{
args->rv = VNET_API_ERROR_NETLINK_ERROR;
goto error;
}
if (setns (fd, CLONE_NEWNET) == -1)
{
args->rv = VNET_API_ERROR_SYSCALL_ERROR_3;
args->error = clib_error_return_unix (0, "setns '%s'",
args->host_namespace);
goto error;
}
if ((vif->ifindex = if_nametoindex ((char *) args->host_if_name)) == 0)
{
args->rv = VNET_API_ERROR_SYSCALL_ERROR_3;
args->error = clib_error_return_unix (0, "if_nametoindex '%s'",
args->host_if_name);
goto error;
}
}
else
{
if (args->host_if_name)
{
args->error = vnet_netlink_set_link_name (vif->ifindex,
(char *)
args->host_if_name);
if (args->error)
{
args->rv = VNET_API_ERROR_NETLINK_ERROR;
goto error;
}
}
}
if (!ethernet_mac_address_is_zero (args->host_mac_addr))
{
args->error = vnet_netlink_set_link_addr (vif->ifindex,
args->host_mac_addr);
if (args->error)
{
args->rv = VNET_API_ERROR_NETLINK_ERROR;
goto error;
}
}
if (args->host_bridge)
{
args->error = vnet_netlink_set_link_master (vif->ifindex,
(char *) args->host_bridge);
if (args->error)
{
args->rv = VNET_API_ERROR_NETLINK_ERROR;
goto error;
}
}
if (args->host_ip4_prefix_len)
{
args->error = vnet_netlink_add_ip4_addr (vif->ifindex,
&args->host_ip4_addr,
args->host_ip4_prefix_len);
if (args->error)
{
args->rv = VNET_API_ERROR_NETLINK_ERROR;
goto error;
}
}
if (args->host_ip6_prefix_len)
{
args->error = vnet_netlink_add_ip6_addr (vif->ifindex,
&args->host_ip6_addr,
args->host_ip6_prefix_len);
if (args->error)
{
args->rv = VNET_API_ERROR_NETLINK_ERROR;
goto error;
}
}
args->error = vnet_netlink_set_link_state (vif->ifindex, 1 /* UP */ );
if (args->error)
{
args->rv = VNET_API_ERROR_NETLINK_ERROR;
goto error;
}
if (args->host_ip4_gw_set)
{
args->error = vnet_netlink_add_ip4_route (0, 0, &args->host_ip4_gw);
if (args->error)
{
args->rv = VNET_API_ERROR_NETLINK_ERROR;
goto error;
}
}
if (args->host_ip6_gw_set)
{
args->error = vnet_netlink_add_ip6_route (0, 0, &args->host_ip6_gw);
if (args->error)
{
args->rv = VNET_API_ERROR_NETLINK_ERROR;
goto error;
}
}
/* switch back to old net namespace */
if (args->host_namespace)
{
if (setns (old_netns_fd, CLONE_NEWNET) == -1)
{
args->rv = VNET_API_ERROR_SYSCALL_ERROR_2;
args->error = clib_error_return_unix (0, "setns '%s'",
args->host_namespace);
goto error;
}
}
/* Set vhost memory table */
i = sizeof (struct vhost_memory) + sizeof (struct vhost_memory_region);
vhost_mem = clib_mem_alloc (i);
clib_memset (vhost_mem, 0, i);
vhost_mem->nregions = 1;
vhost_mem->regions[0].memory_size = (1ULL << 47) - 4096;
_IOCTL (vif->fd, VHOST_SET_MEM_TABLE, vhost_mem);
if ((args->error =
virtio_vring_init (vm, vif, RX_QUEUE (0), args->rx_ring_sz)))
{
args->rv = VNET_API_ERROR_INIT_FAILED;
goto error;
}
vif->num_rxqs = 1;
if ((args->error =
virtio_vring_init (vm, vif, TX_QUEUE (0), args->tx_ring_sz)))
{
args->rv = VNET_API_ERROR_INIT_FAILED;
goto error;
}
vif->num_txqs = 1;
if (!args->mac_addr_set)
ethernet_mac_address_generate (args->mac_addr);
clib_memcpy (vif->mac_addr, args->