From 8448af56da3cdaf6e50b075b322031e8edeb9760 Mon Sep 17 00:00:00 2001
From: Ian Kelling <iank@fsf.org>
Date: Sun, 20 Aug 2017 22:08:39 -0700
Subject: [PATCH] add resolvconf mount namespace

---
 newns | 55 +++++++++++++++++++++++++++++++++++++++++++++++++++----
 1 file changed, 51 insertions(+), 4 deletions(-)

diff --git a/newns b/newns
index 713c29d..e4d1f86 100755
--- a/newns
+++ b/newns
@@ -40,12 +40,12 @@ usage() {
 usage: ${0##*/} [OPTS] start|stop NS_NAME
 Nat a network namespace. systemd friendly
 
-Also creates a mount namespace with a cloned /etc/resolv.conf.
+Also creates a mount namespace with a cloned /run/resolvconf.
 
 -c, --create    Create a named network namespace. When running from
                 the same network namespace as pid 1, this is set automatically.
-                This is the case when running outside a systemd created
-                private network.
+                A systemd created private network is in a network namespace
+                different than pid 1.
 -n NETWORK      x.x.x /24 private network to use. If not specified, uses
                 the first one starting at 10.173.1
 -h, --help      Show this help and exit.
@@ -87,6 +87,8 @@ calls resolvconf, this doesn't help much. What we actually want to do is
 copy /run/resolvconf somehwere then bind mount it on top of
 /run/resolvconf.
 
+Note: for debugging, adding set -x is a pretty good option.
+
 Please email me if you have a patches, bugs, feedback, or republish this
 somewhere else: Ian Kelling <ian@iankelling.org>.
 EOF
@@ -149,12 +151,17 @@ fi
 
 ipd() { ip -n default "$@"; }
 if $create; then
+    # run ip in the network namespace
     ipnn() { ip -n $nn "$@"; }
 else
     # we are already in the network namespace and it's unnamed.
+    # run ip in the network namespace
     ipnn() { ip "$@"; }
 fi
+# default network namespace exec
 dexec() { ip netns exec default "$@"; }
+# mount namespace exec
+mexec() { /usr/bin/nsenter --mount=/root/mount_namespaces/$nn "$@"; }
 
 
 # background: head -n1 is defensive. Not sure if there is some weird feature
@@ -211,8 +218,9 @@ start() {
         # namespace is marked on mounts within the namespace.  Here, we
         # specify that we want mount changes propagated to us, but not
         # back.
-        unshare --propagation slave --mount=/root/mount_namespaces/$nn
+        unshare --propagation slave --mount=/root/mount_namespaces/$nn /bin/true
     fi
+
     ####   end mount namespace setup ####
 
 
@@ -223,6 +231,11 @@ start() {
 
     echo 1 | dexec dd of=/proc/sys/net/ipv4/ip_forward 2>/dev/null
 
+    # docker helpfully changes the default FORWARD to drop...
+    if ! dexec iptables -C FORWARD -i $v0 -j ACCEPT &>/dev/null; then
+        dexec iptables -A FORWARD -i $v0 -j ACCEPT
+    fi
+
     _errcatch_cleanup=stop
     ipnn link add $v0 type veth peer name $v1
     ipnn link set $v0 netns default
@@ -233,6 +246,33 @@ start() {
     ipnn link set $v1 up
     ipnn route add default via $network.1
 
+    ###### begin setup resolvconf
+    resolv_copy=/root/resolvconf-$nn
+
+    # this condition should never happen, just coding defensively
+    if mexec mountpoint /run/resolvconf &>/dev/null; then
+        mexec umount /run/resolvconf
+    fi
+    cp -aT /run/resolvconf $resolv_copy
+    if ! mexec mount -o bind $resolv_copy /run/resolvconf; then
+        echo "error: resolv-conf bindmount failed"
+        exit 1
+    fi
+    # if running dnsmasq, we have 127.0.0.1 for dns, but it can't listen on the loopback
+    # in the network namespace, so adjust the address.
+    if mexec [ -s /run/resolvconf/interface/lo.dnsmasq ]; then
+        mexec sed --follow-symlinks -i "s/nameserver 127\..*/nameserver $network.1/" /run/resolvconf/interface/lo.dnsmasq
+        mexec resolvconf -u
+    fi
+    # background: if we did this in openvpn's resolv-conf script, we could guard it in
+    # if capsh --print|grep '\bcap_sys_admin\b' &>/dev/null
+    # and we could get $nn by
+    # config_basename=${config%%.*}
+    # config_basename=${config_basename##*/}
+    # but dnsmasq forces us to do it earlier.
+    ######  end setup resolvconf
+
+
 }
 
 stop() {
@@ -244,9 +284,16 @@ stop() {
     if ! $existing; then
         if nat -C &>/dev/null; then nat -D; fi
     fi
+    dexec iptables -D FORWARD -i $v0 -j ACCEPT ||:
     if $create; then
         ip netns del $nn
     fi
+
+    # not sure this is necessary since we are tearing down the mount namespace
+    if mexec mountpoint /run/resolvconf &>/dev/null; then
+        mexec umount /run/resolvconf
+    fi
+
     if mountpoint /root/mount_namespaces/$nn >/dev/null; then
         umount /root/mount_namespaces/$nn
     fi
-- 
2.30.2