tapv2: multiple improvements
- change interface naming scheme
- rework netlink code
- add option to set link address, namespace
Change-Id: Icf667babb3077a07617b0b87c45c957e345cb4d1
Signed-off-by: Damjan Marion <damarion@cisco.com>
diff --git a/src/vat/api_format.c b/src/vat/api_format.c
index d39a61a..4b2602d 100644
--- a/src/vat/api_format.c
+++ b/src/vat/api_format.c
@@ -7810,8 +7810,11 @@
vl_api_tap_create_v2_t *mp;
u8 mac_address[6];
u8 random_mac = 1;
- u8 *tap_name = 0;
- u8 *host_namespace = 0;
+ u32 id = ~0;
+ u8 *host_if_name = 0;
+ u8 *host_ns = 0;
+ u8 host_mac_addr[6];
+ u8 host_mac_addr_set = 0;
u8 *host_bridge = 0;
ip4_address_t host_ip4_addr;
u32 host_ip4_prefix_len = 0;
@@ -7829,10 +7832,15 @@
{
random_mac = 0;
}
- else if (unformat (i, "name %s", &tap_name))
+ else if (unformat (i, "id %s", &id))
;
- else if (unformat (i, "host-ns %s", &host_namespace))
+ else if (unformat (i, "host-if-name %s", &host_if_name))
;
+ else if (unformat (i, "host-ns %s", &host_ns))
+ ;
+ else if (unformat (i, "host-mac-addr %U", unformat_ethernet_address,
+ host_mac_addr))
+ host_mac_addr_set = 1;
else if (unformat (i, "host-bridge %s", &host_bridge))
;
else if (unformat (i, "host-ip4-addr %U/%d", unformat_ip4_address,
@@ -7849,17 +7857,12 @@
break;
}
- if (tap_name == 0)
- {
- errmsg ("missing tap name. ");
- return -99;
- }
- if (vec_len (tap_name) > 63)
+ if (vec_len (host_if_name) > 63)
{
errmsg ("tap name too long. ");
return -99;
}
- if (vec_len (host_namespace) > 63)
+ if (vec_len (host_ns) > 63)
{
errmsg ("host name space too long. ");
return -99;
@@ -7900,23 +7903,27 @@
return -99;
}
- vec_add1 (tap_name, 0);
-
/* Construct the API message */
M (TAP_CREATE_V2, mp);
mp->use_random_mac = random_mac;
- clib_memcpy (mp->mac_address, mac_address, 6);
- clib_memcpy (mp->tap_name, tap_name, vec_len (tap_name));
- mp->host_namespace_set = host_namespace != 0;
+
+ mp->id = id;
+ mp->host_namespace_set = host_ns != 0;
mp->host_bridge_set = host_bridge != 0;
mp->host_ip4_addr_set = host_ip4_prefix_len != 0;
mp->host_ip6_addr_set = host_ip6_prefix_len != 0;
mp->rx_ring_sz = rx_ring_sz;
mp->tx_ring_sz = tx_ring_sz;
- if (host_namespace)
- clib_memcpy (mp->host_namespace, host_namespace,
- vec_len (host_namespace));
+
+ if (random_mac)
+ clib_memcpy (mp->mac_address, mac_address, 6);
+ if (host_mac_addr_set)
+ clib_memcpy (mp->host_mac_addr, host_mac_addr, 6);
+ if (host_if_name)
+ clib_memcpy (mp->host_if_name, host_if_name, vec_len (host_if_name));
+ if (host_ns)
+ clib_memcpy (mp->host_namespace, host_ns, vec_len (host_ns));
if (host_bridge)
clib_memcpy (mp->host_bridge, host_bridge, vec_len (host_bridge));
if (host_ip4_prefix_len)
@@ -7925,7 +7932,9 @@
clib_memcpy (mp->host_ip6_addr, &host_ip6_addr, 16);
- vec_free (tap_name);
+ vec_free (host_ns);
+ vec_free (host_if_name);
+ vec_free (host_bridge);
/* send it... */
S (mp);
diff --git a/src/vnet/devices/netlink.c b/src/vnet/devices/netlink.c
index b3330dd..5994366 100644
--- a/src/vnet/devices/netlink.c
+++ b/src/vnet/devices/netlink.c
@@ -19,36 +19,31 @@
#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 <linux/netlink.h>
#include <linux/rtnetlink.h>
#include <vlib/vlib.h>
#include <vlib/unix/unix.h>
+#include <vnet/devices/netlink.h>
typedef struct
{
u8 *data;
} vnet_netlink_msg_t;
-void
+static void
vnet_netlink_msg_init (vnet_netlink_msg_t * m, u16 type, u16 flags,
void *msg_data, int msg_len)
{
struct nlmsghdr *nh;
u8 *p;
- int len = NLMSG_LENGTH (msg_len);
memset (m, 0, sizeof (vnet_netlink_msg_t));
- vec_add2 (m->data, p, len);
+ vec_add2 (m->data, p, NLMSG_SPACE (msg_len));
ASSERT (m->data == p);
nh = (struct nlmsghdr *) p;
- nh->nlmsg_flags = flags;
+ nh->nlmsg_flags = flags | NLM_F_ACK;
nh->nlmsg_type = type;
clib_memcpy (m->data + sizeof (struct nlmsghdr), msg_data, msg_len);
}
@@ -60,7 +55,7 @@
struct rtattr *rta;
u8 *p;
- vec_add2 (m->data, p, RTA_LENGTH (rta_data_len));
+ vec_add2 (m->data, p, RTA_SPACE (rta_data_len));
rta = (struct rtattr *) p;
rta->rta_type = rta_type;
rta->rta_len = RTA_LENGTH (rta_data_len);
@@ -72,9 +67,10 @@
{
clib_error_t *err = 0;
struct sockaddr_nl ra = { 0 };
- int sock;
+ int len, sock;
struct nlmsghdr *nh = (struct nlmsghdr *) m->data;
nh->nlmsg_len = vec_len (m->data);
+ char buf[4096];
if ((sock = socket (AF_NETLINK, SOCK_RAW, NETLINK_ROUTE)) == -1)
return clib_error_return_unix (0, "socket(AF_NETLINK)");
@@ -85,87 +81,126 @@
if ((bind (sock, (struct sockaddr *) &ra, sizeof (ra))) == -1)
{
err = clib_error_return_unix (0, "bind");
- goto error;
+ goto done;
}
if ((send (sock, m->data, vec_len (m->data), 0)) == -1)
err = clib_error_return_unix (0, "send");
-error:
+ if ((len = recv (sock, buf, sizeof (buf), 0)) == -1)
+ err = clib_error_return_unix (0, "recv");
+
+ for (nh = (struct nlmsghdr *) buf; NLMSG_OK (nh, len);
+ nh = NLMSG_NEXT (nh, len))
+ {
+ if (nh->nlmsg_type == NLMSG_DONE)
+ goto done;
+
+ if (nh->nlmsg_type == NLMSG_ERROR)
+ {
+ struct nlmsgerr *e = (struct nlmsgerr *) NLMSG_DATA (nh);
+ if (e->error)
+ err = clib_error_return (0, "netlink error %d", e->error);
+ goto done;
+ }
+ }
+
+done:
close (sock);
vec_free (m->data);
return err;
}
clib_error_t *
-vnet_netlink_set_if_namespace (int ifindex, char *net_ns)
+vnet_netlink_set_link_name (int ifindex, char *new_ifname)
{
vnet_netlink_msg_t m;
struct ifinfomsg ifmsg = { 0 };
- clib_error_t *err;
- int data;
- u16 type;
- u8 *s;
-
- if (strncmp (net_ns, "pid:", 4) == 0)
- {
- data = atoi (net_ns + 4);
- type = IFLA_NET_NS_PID;
- }
- else
- {
- if (net_ns[0] == '/')
- s = format (0, "%s%c", net_ns, 0);
- else
- s = format (0, "/var/run/netns/%s%c", net_ns, 0);
-
- data = open ((char *) s, O_RDONLY);
- type = IFLA_NET_NS_FD;
- vec_free (s);
- if (data == -1)
- return clib_error_return (0, "namespace '%s' doesn't exist", net_ns);
- }
-
- ifmsg.ifi_family = AF_UNSPEC;
ifmsg.ifi_index = ifindex;
- ifmsg.ifi_change = 0xffffffff;
vnet_netlink_msg_init (&m, RTM_SETLINK, NLM_F_REQUEST,
&ifmsg, sizeof (struct ifinfomsg));
- vnet_netlink_msg_add_rtattr (&m, type, &data, sizeof (int));
- err = vnet_netlink_msg_send (&m);
+ vnet_netlink_msg_add_rtattr (&m, IFLA_IFNAME, new_ifname,
+ strlen (new_ifname) + 1);
- if (type == IFLA_NET_NS_FD)
- close (data);
- return err;
-}
-
-clib_error_t *
-vnet_netlink_set_if_master (int ifindex, int master_ifindex)
-{
- vnet_netlink_msg_t m;
- struct ifinfomsg ifmsg = { 0 };
-
- ifmsg.ifi_family = AF_UNSPEC;
- ifmsg.ifi_index = ifindex;
- ifmsg.ifi_change = 0xffffffff;
- vnet_netlink_msg_init (&m, RTM_SETLINK, NLM_F_REQUEST,
- &ifmsg, sizeof (struct ifinfomsg));
- vnet_netlink_msg_add_rtattr (&m, IFLA_MASTER, &master_ifindex,
- sizeof (int));
return vnet_netlink_msg_send (&m);
}
clib_error_t *
-vnet_netlink_set_if_mtu (int ifindex, int mtu)
+vnet_netlink_set_link_netns (int ifindex, int netns_fd, char *new_ifname)
{
vnet_netlink_msg_t m;
struct ifinfomsg ifmsg = { 0 };
- ifmsg.ifi_family = AF_UNSPEC;
ifmsg.ifi_index = ifindex;
- ifmsg.ifi_change = 0xffffffff;
+ vnet_netlink_msg_init (&m, RTM_SETLINK, NLM_F_REQUEST,
+ &ifmsg, sizeof (struct ifinfomsg));
+
+ vnet_netlink_msg_add_rtattr (&m, IFLA_NET_NS_FD, &netns_fd, sizeof (int));
+ if (new_ifname)
+ vnet_netlink_msg_add_rtattr (&m, IFLA_IFNAME, new_ifname,
+ strlen (new_ifname) + 1);
+
+ return vnet_netlink_msg_send (&m);
+}
+
+clib_error_t *
+vnet_netlink_set_link_master (int ifindex, char *master_ifname)
+{
+ vnet_netlink_msg_t m;
+ struct ifinfomsg ifmsg = { 0 };
+ int i;
+
+ ifmsg.ifi_index = ifindex;
+
+ if ((i = if_nametoindex (master_ifname)) == 0)
+ clib_error_return_unix (0, "unknown master interface '%s'",
+ master_ifname);
+
+ vnet_netlink_msg_init (&m, RTM_SETLINK, NLM_F_REQUEST,
+ &ifmsg, sizeof (struct ifinfomsg));
+ vnet_netlink_msg_add_rtattr (&m, IFLA_MASTER, &i, sizeof (int));
+ return vnet_netlink_msg_send (&m);
+}
+
+clib_error_t *
+vnet_netlink_set_link_addr (int ifindex, u8 * mac)
+{
+ vnet_netlink_msg_t m;
+ struct ifinfomsg ifmsg = { 0 };
+
+ ifmsg.ifi_index = ifindex;
+
+ vnet_netlink_msg_init (&m, RTM_SETLINK, NLM_F_REQUEST,
+ &ifmsg, sizeof (struct ifinfomsg));
+ vnet_netlink_msg_add_rtattr (&m, IFLA_ADDRESS, mac, 6);
+ return vnet_netlink_msg_send (&m);
+}
+
+clib_error_t *
+vnet_netlink_set_link_state (int ifindex, int up)
+{
+ vnet_netlink_msg_t m;
+ struct ifinfomsg ifmsg = { 0 };
+
+ ifmsg.ifi_flags = IFF_UP;
+ ifmsg.ifi_change = IFF_UP;
+ ifmsg.ifi_index = ifindex;
+
+ vnet_netlink_msg_init (&m, RTM_SETLINK, NLM_F_REQUEST,
+ &ifmsg, sizeof (struct ifinfomsg));
+ return vnet_netlink_msg_send (&m);
+}
+
+clib_error_t *
+vnet_netlink_set_link_mtu (int ifindex, int mtu)
+{
+ vnet_netlink_msg_t m;
+ struct ifinfomsg ifmsg = { 0 };
+
+ ifmsg.ifi_index = ifindex;
+
vnet_netlink_msg_init (&m, RTM_SETLINK, NLM_F_REQUEST,
&ifmsg, sizeof (struct ifinfomsg));
vnet_netlink_msg_add_rtattr (&m, IFLA_MTU, &mtu, sizeof (int));
diff --git a/src/vnet/devices/netlink.h b/src/vnet/devices/netlink.h
index e61b827..e850196 100644
--- a/src/vnet/devices/netlink.h
+++ b/src/vnet/devices/netlink.h
@@ -16,9 +16,12 @@
#ifndef included_vnet_device_netlink_h
#define included_vnet_device_netlink_h
-clib_error_t *vnet_netlink_set_if_mtu (int ifindex, int mtu);
-clib_error_t *vnet_netlink_set_if_namespace (int ifindex, char *net_ns);
-clib_error_t *vnet_netlink_set_if_master (int ifindex, int master_ifindex);
+clib_error_t *vnet_netlink_set_link_name (int ifindex, char *new_ifname);
+clib_error_t *vnet_netlink_set_link_netns (int ifindex, int netns_fd,
+ char *new_ifname);
+clib_error_t *vnet_netlink_set_link_master (int ifindex, char *master_ifname);
+clib_error_t *vnet_netlink_set_link_addr (int ifindex, u8 * addr);
+clib_error_t *vnet_netlink_set_link_state (int ifindex, int up);
clib_error_t *vnet_netlink_add_ip4_addr (int ifindex, void *addr,
int pfx_len);
clib_error_t *vnet_netlink_add_ip6_addr (int ifindex, void *addr,
diff --git a/src/vnet/devices/tap/cli.c b/src/vnet/devices/tap/cli.c
index f7fc1e6..c86995c 100644
--- a/src/vnet/devices/tap/cli.c
+++ b/src/vnet/devices/tap/cli.c
@@ -38,38 +38,47 @@
tap_create_if_args_t args = { 0 };
int ip_addr_set = 0;
- /* Get a line of input. */
- if (!unformat_user (input, unformat_line_input, line_input))
- return clib_error_return (0, "Missing name <interface>");
+ args.id = ~0;
- while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT)
+ /* Get a line of input. */
+ if (unformat_user (input, unformat_line_input, line_input))
{
- if (unformat (line_input, "name %s", &args.name))
- ;
- else if (unformat (line_input, "host-ns %s", &args.host_namespace))
- ;
- else if (unformat (line_input, "host-bridge %s", &args.host_bridge))
- ;
- else if (unformat (line_input, "host-ip4-addr %U/%d",
- unformat_ip4_address, &args.host_ip4_addr,
- &args.host_ip4_prefix_len))
- ip_addr_set = 1;
- else if (unformat (line_input, "host-ip6-addr %U/%d",
- unformat_ip6_address, &args.host_ip6_addr,
- &args.host_ip6_prefix_len))
- ip_addr_set = 1;
- else if (unformat (line_input, "rx-ring-size %d", &args.rx_ring_sz))
- ;
- else if (unformat (line_input, "tx-ring-size %d", &args.tx_ring_sz))
- ;
- else if (unformat (line_input, "hw-addr %U",
- unformat_ethernet_address, args.hw_addr))
- args.hw_addr_set = 1;
- else
- return clib_error_return (0, "unknown input `%U'",
- format_unformat_error, input);
+
+ while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT)
+ {
+ if (unformat (line_input, "id %u", &args.id))
+ ;
+ else
+ if (unformat (line_input, "host-if-name %s", &args.host_if_name))
+ ;
+ else if (unformat (line_input, "host-ns %s", &args.host_namespace))
+ ;
+ else if (unformat (line_input, "host-mac-addr %U",
+ unformat_ethernet_address, args.host_mac_addr))
+ ;
+ else if (unformat (line_input, "host-bridge %s", &args.host_bridge))
+ ;
+ else if (unformat (line_input, "host-ip4-addr %U/%d",
+ unformat_ip4_address, &args.host_ip4_addr,
+ &args.host_ip4_prefix_len))
+ ip_addr_set = 1;
+ else if (unformat (line_input, "host-ip6-addr %U/%d",
+ unformat_ip6_address, &args.host_ip6_addr,
+ &args.host_ip6_prefix_len))
+ ip_addr_set = 1;
+ else if (unformat (line_input, "rx-ring-size %d", &args.rx_ring_sz))
+ ;
+ else if (unformat (line_input, "tx-ring-size %d", &args.tx_ring_sz))
+ ;
+ else if (unformat (line_input, "hw-addr %U",
+ unformat_ethernet_address, args.mac_addr))
+ args.mac_addr_set = 1;
+ else
+ return clib_error_return (0, "unknown input `%U'",
+ format_unformat_error, input);
+ }
+ unformat_free (line_input);
}
- unformat_free (line_input);
if (ip_addr_set && args.host_bridge)
return clib_error_return (0, "Please specify either host ip address or "
@@ -77,7 +86,7 @@
tap_create_if (vm, &args);
- vec_free (args.name);
+ vec_free (args.host_if_name);
vec_free (args.host_namespace);
vec_free (args.host_bridge);
@@ -88,10 +97,10 @@
/* *INDENT-OFF* */
VLIB_CLI_COMMAND (tap_create_command, static) = {
.path = "create tap",
- .short_help = "create tap {name <if-name>} [hw-addr <mac-address>] "
+ .short_help = "create tap {id <if-id>} [hw-addr <mac-address>] "
"[rx-ring-size <size>] [tx-ring-size <size>] [host-ns <netns>] "
"[host-bridge <bridge-name>] [host-ip4-addr <ip4addr/mask>] "
- "[host-ip6-addr <ip6-addr]",
+ "[host-ip6-addr <ip6-addr] [host-if-name <name>]",
.function = tap_create_command_fn,
};
/* *INDENT-ON* */
@@ -209,8 +218,8 @@
vif = pool_elt_at_index (mm->interfaces, hi->dev_instance);
vlib_cli_output (vm, "interface %U", format_vnet_sw_if_index_name,
vnm, vif->sw_if_index);
- if (vif->name)
- vlib_cli_output (vm, " name \"%s\"", vif->name);
+ if (vif->host_if_name)
+ vlib_cli_output (vm, " name \"%s\"", vif->host_if_name);
if (vif->net_ns)
vlib_cli_output (vm, " host-ns \"%s\"", vif->net_ns);
vlib_cli_output (vm, " flags 0x%x", vif->flags);
diff --git a/src/vnet/devices/tap/tap.c b/src/vnet/devices/tap/tap.c
index b4004f7..f31548c 100644
--- a/src/vnet/devices/tap/tap.c
+++ b/src/vnet/devices/tap/tap.c
@@ -15,6 +15,7 @@
*------------------------------------------------------------------
*/
+#define _GNU_SOURCE
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
@@ -24,6 +25,7 @@
#include <linux/virtio_net.h>
#include <linux/vhost.h>
#include <sys/eventfd.h>
+#include <sched.h>
#include <linux/netlink.h>
#include <linux/rtnetlink.h>
@@ -37,6 +39,8 @@
#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) \
{ \
@@ -53,24 +57,79 @@
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;
+}
+
+
void
tap_create_if (vlib_main_t * vm, tap_create_if_args_t * args)
{
vnet_main_t *vnm = vnet_get_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, fd = -1;
+ int old_netns_fd = -1;
struct ifreq ifr;
size_t hdrsz;
struct vhost_memory *vhost_mem = 0;
virtio_if_t *vif = 0;
clib_error_t *err = 0;
+ uword *p;
+
+ if (args->id != ~0)
+ {
+ p = hash_get (tm->dev_instance_by_interface_id, args->id);
+ if (p)
+ {
+ args->rv = VNET_API_ERROR_INVALID_INTERFACE;
+ args->error = clib_error_return (0, "interface already exists");
+ return;
+ }
+ }
+ else
+ {
+ int tries = 1000;
+ while (--tries)
+ {
+ args->id = tm->last_used_interface_id++;
+ p = hash_get (tm->dev_instance_by_interface_id, args->id);
+ if (!p)
+ break;
+ }
+
+ if (!tries)
+ {
+ args->rv = VNET_API_ERROR_UNSPECIFIED;
+ args->error =
+ clib_error_return (0, "cannot find free interface id");
+ return;
+ }
+ }
memset (&ifr, 0, sizeof (ifr));
pool_get (vim->interfaces, vif);
vif->dev_instance = vif - vim->interfaces;
vif->tap_fd = -1;
+ vif->id = args->id;
+
+ hash_set (tm->dev_instance_by_interface_id, vif->id, vif->dev_instance);
if ((vif->fd = open ("/dev/vhost-net", O_RDWR | O_NONBLOCK)) < 0)
{
@@ -119,10 +178,8 @@
}
ifr.ifr_flags = IFF_TAP | IFF_NO_PI | IFF_ONE_QUEUE | IFF_VNET_HDR;
- strncpy (ifr.ifr_ifrn.ifrn_name, (char *) args->name, IF_NAMESIZE - 1);
_IOCTL (vif->tap_fd, TUNSETIFF, (void *) &ifr);
-
- vif->ifindex = if_nametoindex ((char *) args->name);
+ vif->ifindex = if_nametoindex (ifr.ifr_ifrn.ifrn_name);
unsigned int offload = 0;
hdrsz = sizeof (struct virtio_net_hdr_v1);
@@ -130,10 +187,61 @@
_IOCTL (vif->tap_fd, TUNSETVNETHDRSZ, &hdrsz);
_IOCTL (vif->fd, VHOST_SET_OWNER, 0);
- if (args->host_bridge)
+ /* if namespace is specified, all further netlink messages should be excuted
+ after we change our net namespace */
+ if (args->host_namespace)
{
- int master_ifindex = if_nametoindex ((char *) args->host_bridge);
- args->error = vnet_netlink_set_if_master (vif->ifindex, master_ifindex);
+ int fd;
+ 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;
+ }
+ close (fd);
+ 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;
@@ -141,11 +249,10 @@
}
}
- if (args->host_namespace)
+ if (args->host_bridge)
{
- args->error = vnet_netlink_set_if_namespace (vif->ifindex,
- (char *)
- args->host_namespace);
+ args->error = vnet_netlink_set_link_master (vif->ifindex,
+ (char *) args->host_bridge);
if (args->error)
{
args->rv = VNET_API_ERROR_NETLINK_ERROR;
@@ -153,6 +260,7 @@
}
}
+
if (args->host_ip4_prefix_len)
{
args->error = vnet_netlink_add_ip4_addr (vif->ifindex,
@@ -177,6 +285,25 @@
}
}
+ args->error = vnet_netlink_set_link_state (vif->ifindex, 1 /* UP */ );
+ 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);
@@ -197,33 +324,24 @@
goto error;
}
- /* set host side up */
- if ((fd = socket (AF_INET, SOCK_STREAM, 0)) > 0)
- {
- memset (&ifr, 0, sizeof (struct ifreq));
- strncpy (ifr.ifr_name, (char *) args->name, sizeof (ifr.ifr_name) - 1);
- _IOCTL (fd, SIOCGIFFLAGS, (void *) &ifr);
- ifr.ifr_flags |= IFF_UP | IFF_RUNNING;
- _IOCTL (fd, SIOCSIFFLAGS, (void *) &ifr);
- }
-
- if (!args->hw_addr_set)
+ if (!args->mac_addr_set)
{
f64 now = vlib_time_now (vm);
u32 rnd;
rnd = (u32) (now * 1e6);
rnd = random_u32 (&rnd);
- memcpy (args->hw_addr + 2, &rnd, sizeof (rnd));
- args->hw_addr[0] = 2;
- args->hw_addr[1] = 0xfe;
+ memcpy (args->mac_addr + 2, &rnd, sizeof (rnd));
+ args->mac_addr[0] = 2;
+ args->mac_addr[1] = 0xfe;
}
- vif->name = args->name;
- args->name = 0;
+ vif->host_if_name = args->host_if_name;
+ args->host_if_name = 0;
vif->net_ns = args->host_namespace;
args->host_namespace = 0;
args->error = ethernet_register_interface (vnm, virtio_device_class.index,
- vif->dev_instance, args->hw_addr,
+ vif->dev_instance,
+ args->mac_addr,
&vif->hw_if_index,
virtio_eth_flag_change);
if (args->error)
@@ -276,6 +394,7 @@
{
vnet_main_t *vnm = vnet_get_main ();
virtio_main_t *mm = &virtio_main;
+ tap_main_t *tm = &tap_main;
int i;
virtio_if_t *vif;
vnet_hw_interface_t *hw;
@@ -301,6 +420,7 @@
vec_foreach_index (i, vif->vrings) virtio_vring_free (vif, i);
vec_free (vif->vrings);
+ hash_unset (tm->dev_instance_by_interface_id, vif->id);
memset (vif, 0, sizeof (*vif));
pool_put (mm->interfaces, vif);
@@ -337,7 +457,8 @@
static clib_error_t *
tap_init (vlib_main_t * vm)
{
-
+ tap_main_t *tm = &tap_main;
+ tm->dev_instance_by_interface_id = hash_create (0, sizeof (uword));
return 0;
}
diff --git a/src/vnet/devices/tap/tap.h b/src/vnet/devices/tap/tap.h
index 0e0f8cb..7d07ffb 100644
--- a/src/vnet/devices/tap/tap.h
+++ b/src/vnet/devices/tap/tap.h
@@ -24,12 +24,14 @@
typedef struct
{
- u8 *name;
- u8 hw_addr_set;
- u8 hw_addr[6];
+ u32 id;
+ u8 mac_addr_set;
+ u8 mac_addr[6];
u16 rx_ring_sz;
u16 tx_ring_sz;
u8 *host_namespace;
+ u8 *host_if_name;
+ u8 host_mac_addr[6];
u8 *host_bridge;
ip4_address_t host_ip4_addr;
u32 host_ip4_prefix_len;
@@ -48,6 +50,12 @@
u8 dev_name[64];
} tap_interface_details_t;
+typedef struct
+{
+ u32 last_used_interface_id;
+ uword *dev_instance_by_interface_id;
+} tap_main_t;
+
void tap_create_if (vlib_main_t * vm, tap_create_if_args_t * args);
int tap_delete_if (vlib_main_t * vm, u32 sw_if_index);
int tap_dump_ifs (tap_interface_details_t ** out_tapids);
diff --git a/src/vnet/devices/tap/tapv2.api b/src/vnet/devices/tap/tapv2.api
index 0378860..a206269 100644
--- a/src/vnet/devices/tap/tapv2.api
+++ b/src/vnet/devices/tap/tapv2.api
@@ -24,11 +24,15 @@
/** \brief Initialize a new tap interface with the given paramters
@param client_index - opaque cookie to identify the sender
@param context - sender context, to match reply w/ request
+ @param id - interface id, 0xffff means auto
@param use_random_mac - let the system generate a unique mac address
- @param tap_name - name to associate with the new interface
@param mac_address - mac addr to assign to the interface if use_radom not set
@param tx_ring_sz - the number of entries of TX ring
@param rx_ring_sz - the number of entries of RX ring
+ @param host_mac_addr_set - host side interface mac address should be set
+ @param host_mac_addr - host side interface mac address
+ @param host_if_name_set - host side interface name should be set
+ @param host_if_name - host side interface name
@param host_namespace_set - host namespece should be set
@param host_namespace - host namespace to attach interface to
@param host_bridge_set - host bridge should be set
@@ -44,13 +48,17 @@
{
u32 client_index;
u32 context;
+ u32 id;
u8 use_random_mac;
- u8 tap_name[64];
u8 mac_address[6];
u16 tx_ring_sz; /* optional, default is 256 entries, must be power of 2 */
u16 rx_ring_sz; /* optional, default is 256 entries, must be power of 2 */
u8 host_namespace_set;
u8 host_namespace[64];
+ u8 host_mac_addr_set;
+ u8 host_mac_addr[6];
+ u8 host_if_name_set;
+ u8 host_if_name[64];
u8 host_bridge_set;
u8 host_bridge[64];
u8 host_ip4_addr_set;
diff --git a/src/vnet/devices/tap/tapv2_api.c b/src/vnet/devices/tap/tapv2_api.c
index 2b324d6..3cededb 100644
--- a/src/vnet/devices/tap/tapv2_api.c
+++ b/src/vnet/devices/tap/tapv2_api.c
@@ -59,16 +59,25 @@
memset (ap, 0, sizeof (*ap));
- ap->name = mp->tap_name;
+ ap->id = mp->id;
if (!mp->use_random_mac)
{
- clib_memcpy (ap->hw_addr, mp->mac_address, 6);
- ap->hw_addr_set = 1;
+ clib_memcpy (ap->mac_addr, mp->mac_address, 6);
+ ap->mac_addr_set = 1;
}
ap->rx_ring_sz = ntohs (mp->rx_ring_sz);
ap->tx_ring_sz = ntohs (mp->tx_ring_sz);
ap->sw_if_index = (u32) ~ 0;
+ if (mp->host_if_name_set)
+ ap->host_if_name = mp->host_if_name;
+
+ if (mp->host_mac_addr_set)
+ {
+ clib_memcpy (ap->host_mac_addr, mp->host_mac_addr, 6);
+ ap->mac_addr_set = 1;
+ }
+
if (mp->host_namespace_set)
ap->host_namespace = mp->host_namespace;
diff --git a/src/vnet/devices/virtio/device.c b/src/vnet/devices/virtio/device.c
index 275a3c7..d4ef6d8 100644
--- a/src/vnet/devices/virtio/device.c
+++ b/src/vnet/devices/virtio/device.c
@@ -55,7 +55,7 @@
if (vif->type == VIRTIO_IF_TYPE_TAP)
{
- s = format (s, "tap-%s", vif->name);
+ s = format (s, "tap%u", vif->id);
}
else
s = format (s, "virtio%lu", vif->dev_instance);
diff --git a/src/vnet/devices/virtio/virtio.h b/src/vnet/devices/virtio/virtio.h
index 7dcd90a..cb97df8 100644
--- a/src/vnet/devices/virtio/virtio.h
+++ b/src/vnet/devices/virtio/virtio.h
@@ -91,6 +91,7 @@
typedef struct
{
u32 flags;
+ u32 id;
u32 dev_instance;
u32 hw_if_index;
u32 sw_if_index;
@@ -102,7 +103,7 @@
u64 features, remote_features;
virtio_if_type_t type;
- u8 *name;
+ u8 *host_if_name;
u8 *net_ns;
int ifindex;
} virtio_if_t;
diff --git a/src/vnet/ethernet/ethernet.h b/src/vnet/ethernet/ethernet.h
index 2a3383a..9a97817 100644
--- a/src/vnet/ethernet/ethernet.h
+++ b/src/vnet/ethernet/ethernet.h
@@ -61,6 +61,12 @@
return (a & (1ULL << (5 * 8))) != 0;
}
+static inline int
+ethernet_mac_address_is_zero (u8 * mac)
+{
+ return ((*((u32 *) mac) == 0) && (*((u16 *) (mac + 4)) == 0));
+}
+
static_always_inline int
ethernet_frame_is_tagged (u16 type)
{
diff --git a/src/vpp/api/custom_dump.c b/src/vpp/api/custom_dump.c
index a4d5870..c26d65a 100644
--- a/src/vpp/api/custom_dump.c
+++ b/src/vpp/api/custom_dump.c
@@ -556,9 +556,15 @@
memset (null_mac, 0, sizeof (null_mac));
s = format (0, "SCRIPT: tap_create_v2 ");
- s = format (s, "name %s ", mp->tap_name);
+ s = format (s, "id %s ", mp->id);
if (memcmp (mp->mac_address, null_mac, 6))
- s = format (s, "hw-addr %U ", format_ethernet_address, mp->mac_address);
+ s = format (s, "mac-address %U ",
+ format_ethernet_address, mp->mac_address);
+ if (memcmp (mp->host_mac_addr, null_mac, 6))
+ s = format (s, "host-mac-addr %U ",
+ format_ethernet_address, mp->host_mac_addr);
+ if (mp->host_if_name_set)
+ s = format (s, "host-if-name %s ", mp->host_if_name);
if (mp->host_namespace_set)
s = format (s, "host-ns %s ", mp->host_namespace);
if (mp->host_bridge_set)