From d2773e59de3dd970d861e9f996bc48de20ef4314 Mon Sep 17 00:00:00 2001
From: Lennart Poettering <lennart@poettering.net>
Date: Mon, 25 Apr 2016 17:36:51 +0200
Subject: [PATCH] nspawn: explicitly remove veth links after use (#3111)

* sd-netlink: permit RTM_DELLINK messages with no ifindex

This is useful for removing network interfaces by name.

* nspawn: explicitly remove veth links we created after use

Sometimes the kernel keeps veth links pinned after the namespace they have been
joined to died. Let's hence explicitly remove veth links after use.

Fixes: #2173
---
 src/libsystemd/sd-netlink/rtnl-message.c |  1 -
 src/nspawn/nspawn-network.c              | 47 ++++++++++++++++++++++++
 src/nspawn/nspawn-network.h              |  2 +
 src/nspawn/nspawn.c                      |  2 +
 4 files changed, 51 insertions(+), 1 deletion(-)

diff --git a/src/libsystemd/sd-netlink/rtnl-message.c b/src/libsystemd/sd-netlink/rtnl-message.c
index 255526bf32d..f251536a891 100644
--- a/src/libsystemd/sd-netlink/rtnl-message.c
+++ b/src/libsystemd/sd-netlink/rtnl-message.c
@@ -402,7 +402,6 @@ int sd_rtnl_message_new_link(sd_netlink *rtnl, sd_netlink_message **ret,
         int r;
 
         assert_return(rtnl_message_type_is_link(nlmsg_type), -EINVAL);
-        assert_return(nlmsg_type != RTM_DELLINK || index > 0, -EINVAL);
         assert_return(ret, -EINVAL);
 
         r = message_new(rtnl, ret, nlmsg_type);
diff --git a/src/nspawn/nspawn-network.c b/src/nspawn/nspawn-network.c
index 74a0ae865b4..f2b7e4dd797 100644
--- a/src/nspawn/nspawn-network.c
+++ b/src/nspawn/nspawn-network.c
@@ -538,3 +538,50 @@ int veth_extra_parse(char ***l, const char *p) {
         a = b = NULL;
         return 0;
 }
+
+static int remove_one_veth_link(sd_netlink *rtnl, const char *name) {
+        _cleanup_(sd_netlink_message_unrefp) sd_netlink_message *m = NULL;
+        int r;
+
+        if (isempty(name))
+                return 0;
+
+        r = sd_rtnl_message_new_link(rtnl, &m, RTM_DELLINK, 0);
+        if (r < 0)
+                return log_error_errno(r, "Failed to allocate netlink message: %m");
+
+        r = sd_netlink_message_append_string(m, IFLA_IFNAME, name);
+        if (r < 0)
+                return log_error_errno(r, "Failed to add netlink interface name: %m");
+
+        r = sd_netlink_call(rtnl, m, 0, NULL);
+        if (r == -ENODEV) /* Already gone */
+                return 0;
+        if (r < 0)
+                return log_error_errno(r, "Failed to remove veth interface %s: %m", name);
+
+        return 1;
+}
+
+int remove_veth_links(const char *primary, char **pairs) {
+        _cleanup_(sd_netlink_unrefp) sd_netlink *rtnl = NULL;
+        char **a, **b;
+        int r;
+
+        /* In some cases the kernel might pin the veth links between host and container even after the namespace
+         * died. Hence, let's better remove them explicitly too. */
+
+        if (isempty(primary) && strv_isempty(pairs))
+                return 0;
+
+        r = sd_netlink_open(&rtnl);
+        if (r < 0)
+                return log_error_errno(r, "Failed to connect to netlink: %m");
+
+        remove_one_veth_link(rtnl, primary);
+
+        STRV_FOREACH_PAIR(a, b, pairs)
+                remove_one_veth_link(rtnl, *a);
+
+        return 0;
+}
diff --git a/src/nspawn/nspawn-network.h b/src/nspawn/nspawn-network.h
index 9ab1606d1c5..c5036ab470e 100644
--- a/src/nspawn/nspawn-network.h
+++ b/src/nspawn/nspawn-network.h
@@ -34,3 +34,5 @@ int setup_ipvlan(const char *machine_name, pid_t pid, char **ifaces);
 int move_network_interfaces(pid_t pid, char **ifaces);
 
 int veth_extra_parse(char ***l, const char *p);
+
+int remove_veth_links(const char *primary, char **pairs);
diff --git a/src/nspawn/nspawn.c b/src/nspawn/nspawn.c
index e1d37d383a2..d687df8a091 100644
--- a/src/nspawn/nspawn.c
+++ b/src/nspawn/nspawn.c
@@ -3713,6 +3713,7 @@ int main(int argc, char *argv[]) {
                 }
 
                 expose_port_flush(arg_expose_ports, &exposed);
+                (void) remove_veth_links(veth_name, arg_network_veth_extra);
         }
 
 finish:
@@ -3745,6 +3746,7 @@ finish:
         }
 
         expose_port_flush(arg_expose_ports, &exposed);
+        (void) remove_veth_links(veth_name, arg_network_veth_extra);
 
         free(arg_directory);
         free(arg_template);