linux-cp: Add VPP->Linux synchronization

Part 1 -- notes in https://ipng.ch/s/articles/2021/08/13/vpp-2.html
Add the ability for VPP to copy out (sync) its state from the dataplane
to Linux Interface Pairs, when they exist. Gated by a configuration
flag (linux-cp { lcp-sync }), and by a CLI option to toggle on/off,
synchronize the following events:

- Interface state changes
- Interface MTU changes
- Interface IPv4/IPv6 address add/deletion

In VPP, subints can have any link state and MTU, orthogonal to their
phy. In Linux, setting admin-down on a phy forces its children to be
down as well. Also, in Linux, MTU of children must not exceed that of
the phy. Add a state synchronizer which walks over phy+subints to
ensure Linux and VPP end up in the same consistent state.

Part 2 -- notes in https://ipng.ch/s/articles/2021/08/15/vpp-3.html
Add the ability for VPP to autocreate sub-interfaces of existing Linux
Interface pairs. Gated by a configuration flag
(linux-cp { lcp-auto-subint }), and by a CLI option to toggle on/off,
synchronize the following event:

- Sub-interface creation (dot1q, dot1ad, QinQ and QinAD)

A few other changes:
- Add two functions into netlink.[ch] to delete ip4 and ip6 addresses.
- Remove a spurious logline (printing MTU) in netlink.c.
- Resolve a TODO around vnet_sw_interface_supports_addressing()

Type: improvement
Signed-off-by: Pim van Pelt <pim@ipng.nl>
Change-Id: I34fc070e80af4013be58d7a8cbf64296cc760e4e
Signed-off-by: Pim van Pelt <pim@ipng.nl>

1 files changed, 44 insertions, 1 deletions

diff --git a/src/vnet/devices/netlink.c b/src/vnet/devices/netlink.c
index 9aae205c54f..da21e9adea1 100644
--- a/src/vnet/devices/netlink.c
+++ b/src/vnet/devices/netlink.c
@@ -273,7 +273,6 @@ vnet_netlink_get_link_mtu (int ifindex, u32 *mtu)
 	    *mtu = clib_net_to_host_u32 (msg_mtu);
 	  else
 	    *mtu = msg_mtu;
-	  clib_warning ("mtu: %d", *mtu);
 	  goto done;
 	}
       offset = NLA_ALIGN (attr->nla_len);
@@ -409,6 +408,50 @@ vnet_netlink_add_ip6_route (void *dst, u8 dst_len, void *gw)
   return err;
 }
 
+clib_error_t *
+vnet_netlink_del_ip4_addr (int ifindex, void *addr, int pfx_len)
+{
+  vnet_netlink_msg_t m;
+  struct ifaddrmsg ifa = { 0 };
+  clib_error_t *err = 0;
+
+  ifa.ifa_family = AF_INET;
+  ifa.ifa_prefixlen = pfx_len;
+  ifa.ifa_index = ifindex;
+
+  vnet_netlink_msg_init (&m, RTM_DELADDR, NLM_F_REQUEST, &ifa,
+			 sizeof (struct ifaddrmsg));
+
+  vnet_netlink_msg_add_rtattr (&m, IFA_LOCAL, addr, 4);
+  vnet_netlink_msg_add_rtattr (&m, IFA_ADDRESS, addr, 4);
+  err = vnet_netlink_msg_send (&m, NULL);
+  if (err)
+    err = clib_error_return (0, "del ip4 addr %U", format_clib_error, err);
+  return err;
+}
+
+clib_error_t *
+vnet_netlink_del_ip6_addr (int ifindex, void *addr, int pfx_len)
+{
+  vnet_netlink_msg_t m;
+  struct ifaddrmsg ifa = { 0 };
+  clib_error_t *err = 0;
+
+  ifa.ifa_family = AF_INET6;
+  ifa.ifa_prefixlen = pfx_len;
+  ifa.ifa_index = ifindex;
+
+  vnet_netlink_msg_init (&m, RTM_DELADDR, NLM_F_REQUEST, &ifa,
+			 sizeof (struct ifaddrmsg));
+
+  vnet_netlink_msg_add_rtattr (&m, IFA_LOCAL, addr, 16);
+  vnet_netlink_msg_add_rtattr (&m, IFA_ADDRESS, addr, 16);
+  err = vnet_netlink_msg_send (&m, NULL);
+  if (err)
+    err = clib_error_return (0, "del ip6 addr %U", format_clib_error, err);
+  return err;
+}
+
 /*
  * fd.io coding-style-patch-verification: ON
  *