From: Ian Kelling <iank@fsf.org>
Date: Wed, 27 Mar 2024 20:17:30 +0000 (-0400)
Subject: Improvements, including compatibility issue.
X-Git-Url: https://iankelling.org/git/?a=commitdiff_plain;h=4c177e45b9aa0c9b26aefaabbc8acac767ac294f;p=newns

Improvements, including compatibility issue.

We now save mount namespace info in a temporary directory. If you
upgrade from here, cleanup old directory with rm -rf
/root/mount_namespaces.

We add a show command for debugging, and fixup some minor bugs in
stop. Adds a check for silent failure of adding default route failure
which I might be the cause of an intermittent issue I've seen. I will
have to wait and see if it gets detected.
---

diff --git a/newns b/newns
index 315c786..34e4ca5 100755
--- a/newns
+++ b/newns
@@ -19,18 +19,38 @@
 # https://savannah.nongnu.org/projects/bash-bear-trap/
 set -e; . /usr/local/lib/bash-bear; set +e
 
+m() {
+  local out
+  printf "newns: %s\n" "$*"
+  if ! out=$("$@" 2>&1); then
+    echo "newns: WARNING: last command exit code: $?"
+  elif [[ ! $out ]]; then
+    echo "newns: WARNING: no output from last command"
+  fi
+}
 
 usage() {
   cat <<EOF
-usage: ${0##*/} [OPTS] start|stop NS_NAME
+usage: ${0##*/} [OPTS] start|stop|show NS_NAME
 Nat a network namespace. systemd friendly
 
 Also creates a mount namespace with a cloned /run/resolvconf.
 
+Arguments:
+
+start|stop: these do what they say.
+
+show: Show the state we expected to be there or not there based on
+start/stop. This is useful for debugging.
+
+NS_NAME: We use this to name the interfaces we create, the mount
+namespace, and if we are creating a named network space, that too.
+
 -c, --create    Create or destroy a named network namespace. When running from
                 the same network namespace as pid 1, this is set automatically.
-                A systemd created private network is in a network namespace
-                different than pid 1.
+                A systemd created private network is in an unnamed network namespace
+                different than pid 1. I haven't found a need for a named network
+                namespace in that case.
 -n NETWORK      x.x.x /24 private network to use. If not specified, uses
                 the first unused one starting at 10.173.1
 -h, --help      Show this help and exit.
@@ -38,7 +58,7 @@ Also creates a mount namespace with a cloned /run/resolvconf.
 From a normal shell:
 
 If we do create the netns, to join it with a shell, we can do (as root)
-/usr/bin/nsenter --mount=/root/mount_namespaces/NAME --net=/var/run/netns/NAME bash
+/usr/bin/nsenter --mount=/run/mount-namespaces/NAME --net=/var/run/netns/NAME bash
 
 If you dont care about the mount namespace, you can leave that option off.
 
@@ -50,10 +70,10 @@ would be called from ExecStartPre, and or subsequent units called with
 JoinsNamespaceOf= and PrivateNetwork=true.
 
 If resolvconf is installed, we create a named mount namespace under
-/root/mount_namespaces, so we can alter some system config for this
+/run/mount-namespaces, so we can alter some system config for this
 namespace. systemd command lines would be prefixed with:
 
-/usr/bin/nsenter --mount=/root/mount_namespaces/NS_NAME
+/usr/bin/nsenter --mount=/run/mount-namespaces/NS_NAME
 
 Note, this means that they can't run as unpriveledged users, but once
 systemd 233 comes out, it will have a bind mount option from within unit
@@ -85,6 +105,8 @@ copy /run/resolvconf somehwere then bind mount it on top of
 
 Note: for debugging, adding set -x is a pretty good option.
 
+TODO: make "start" be idempotent.
+
 Please email me if you have a patches, bugs, feedback, or republish this
 somewhere else: Ian Kelling <ian@iankelling.org>.
 EOF
@@ -94,7 +116,7 @@ EOF
 
 #### begin arg parsing ####
 create=false
-temp=$(getopt -l help,create hcn: "$@") || usage 1
+temp=$(getopt -l help,create hcdn: "$@") || usage 1
 eval set -- "$temp"
 while true; do
   case $1 in
@@ -132,37 +154,39 @@ v0=veth0-$nn
 v1=veth1-$nn
 ip_base=10.173
 
-if ! $create && [[ $(readlink /proc/self/ns/net) == "$(readlink /proc/1/ns/net)" ]]; then
-  create=true
-fi
-
-# make the default network namespace be named
 
+### begin make the default network namespace be named "default" ###
 mkdir -p /run/netns
 target=/run/netns/default
 if [[ ! -e $target && ! -L $target ]]; then
   # -f to avoid a race condition with running twice
   ln -sf /proc/1/ns/net $target
 fi
+### end make the default network namespace be named "default" ###
 
-ipd() { ip -n default "$@"; }
-
-
+if ! $create && [[ $(readlink /proc/self/ns/net) == "$(readlink /proc/1/ns/net)" ]]; then
+  create=true
+fi
 # otherwise we are already in the network namespace and it's unnamed.
 if $create; then
   ipnnargs="-n $nn"
 fi
+
+
+ipd() { ip -n default "$@"; }
+
 # run ip in the network namespace
 ipnn() { ip $ipnnargs "$@"; }
 
 # default network namespace exec
 dexec() { ip netns exec default "$@"; }
 # mount namespace exec
-mexec() { /usr/bin/nsenter --mount=/root/mount_namespaces/$nn "$@"; }
+mexec() { /usr/bin/nsenter --mount=/run/mount-namespaces/$nn "$@"; }
 
 
 nat() {
-  # note, in a previous commit i specified the output interface with -o,
+  # Note: duplicated in show()
+  # Note, in a previous commit i specified the output interface with -o,
   # but that broke things when my gateway interface changed, and I can't
   # see any advantage to it, so I removed it.
   dexec iptables -t nat $1 POSTROUTING -s $network.0/24 -j MASQUERADE \
@@ -177,44 +201,41 @@ diptables-add() {
 
 }
 
-find_network() {
+find-network() {
   if [[ $network ]]; then
     return
   fi
   found=false
-  existing=false
   ips="$(ipd addr show | awk '$1 == "inet" {print $2}')"
   for ((i=1; i <= 254; i++)); do
     network=$ip_base.$i
-    if printf "%s\n" "$ips" | grep "^${network//./\\.}" >/dev/null; then
-      existing=true
-    else
+    if ! printf "%s\n" "$ips" | grep "^${network//./\\.}" >/dev/null; then
       found=true
       break
     fi
   done
-}
-
-start() {
-  find_network
   if ! $found; then
     echo "$0: error: no open network found"
     exit 1
   fi
+}
+
+start() {
+  find-network
 
   #### begin mount namespace setup ####
-  mkdir -p /root/mount_namespaces
-  if ! mountpoint /root/mount_namespaces >/dev/null; then
-    mount --bind /root/mount_namespaces /root/mount_namespaces
+  mkdir -p /run/mount-namespaces
+  if ! mountpoint /run/mount-namespaces >/dev/null; then
+    mount --bind /run/mount-namespaces /run/mount-namespaces
   fi
   # note: This is outside the mount condition because I've mysteriously
   # had this become shared instead of private, perhaps it
   # got remounted somehow and lost the setting.
-  mount --make-private /root/mount_namespaces
-  if [[ ! -e /root/mount_namespaces/$nn ]]; then
-    touch /root/mount_namespaces/$nn
+  mount --make-private /run/mount-namespaces
+  if [[ ! -e /run/mount-namespaces/$nn ]]; then
+    touch /run/mount-namespaces/$nn
   fi
-  if ! mountpoint /root/mount_namespaces/$nn >/dev/null; then
+  if ! mountpoint /run/mount-namespaces/$nn >/dev/null; then
     # Here, we specify that we only want mount changes changes under
     # this mountpoint to be propagated into the bind, but changes
     # from within the bind do not propagate to outside the bind.
@@ -224,12 +245,12 @@ start() {
     # documentation on propagation is a bit weird because it
     # confusingly talks about binds, namespaces, and mirrors (which
     # seems to be just another name for bind), shared subtrees
-    # (which seems to a term for binds and namespaces), and does not
+    # (which seems to be a term for binds and namespaces), and does not
     # properly specify whether the documentation applies to binds,
     # namespaces, or both. Notably, propagation for binds is marked
     # on the original mount point, and propagation for a mount
     # namespace is marked on mounts within the namespace.
-    unshare --propagation slave --mount=/root/mount_namespaces/$nn /bin/true
+    unshare --propagation slave --mount=/run/mount-namespaces/$nn /bin/true
   fi
 
   ####   end mount namespace setup ####
@@ -255,7 +276,34 @@ start() {
   nat -C &>/dev/null || nat -A
   ipnn addr add $network.2/24 dev $v1
   ipnn link set $v1 up
-  ipnn route add default via $network.1
+  cmd="ipnn route add default via $network.1"
+  $cmd
+  fails=0
+  max_fails=2
+  # I've had adding the default route mysteriously fail on boot, so
+  # here we check that it succeeded, do a sleep and a retry.
+  while true; do
+    default_route=$(ipnn route show default | sed -r 's,^[[:space:]]+|[[:space:]]+$,,')
+    if [[ $default_route != "default via $network.1 dev $v1" ]]; then
+      fails=$((fails + 1))
+    else
+      break
+    fi
+    if (( fails >= max_fails )); then
+      echo "$0: ERROR: default route added but not found, retried $max_fails. expected route: 'default via $network.1 dev $v1', found: '$default_route'"
+      # Note: for debugging, if you have a systemd unit which tears down
+      # the newns upon failure, you may want to uncomment the break so
+      # that we proceed and can inspect the system.  break
+      exit 1
+    else
+      sleep 1
+      $cmd
+    fi
+  done
+  if (( fails >= 1 )); then
+    echo "$0: WARNING: route added but not found until retried $max_fails times: $cmd"
+  fi
+
 
   ###### begin setup resolvconf
   if [[ -e /run/resolvconf ]]; then # resolvconf probably installed
@@ -291,19 +339,17 @@ start() {
 
   fi # end if [[ -e /run/resolvconf ]]
   ######  end setup resolvconf
-
-
 }
 
 stop() {
+  if [[ ! $network ]]; then
+    network=$(ipd -f inet a show dev $v0 2>/dev/null | awk '/inet / {print $2}' | sed -r 's,\.[0-9]+/.*,,' ||:)
+  fi
   if ipd link list $v0 &>/dev/null; then
     # this also deletes $v1 and the route we added.
     ipd link del $v0
   fi
-  find_network
-  if ! $existing; then
-    if nat -C &>/dev/null; then nat -D; fi
-  fi
+  if [[ $network ]] && nat -C &>/dev/null; then nat -D; fi
   dexec iptables -D FORWARD -i $v0 -j ACCEPT &>/dev/null ||:
   if $create && [[ -e /var/run/netns/$nn ]]; then
     ip netns del $nn
@@ -314,13 +360,22 @@ stop() {
     mexec umount /run/resolvconf
   fi
 
-  if mountpoint /root/mount_namespaces/$nn >/dev/null; then
-    umount /root/mount_namespaces/$nn
+  if mountpoint /run/mount-namespaces/$nn >/dev/null; then
+    umount /run/mount-namespaces/$nn
   fi
 }
 
+show() {
+  m ipd link list $v0
+  m dexec iptables -t nat -C POSTROUTING -s $network.0/24 -j MASQUERADE \
+    -m comment --comment "systemd network namespace nat" ||:
+  m dexec iptables -C FORWARD -i $v0 -j ACCEPT
+  m mexec mountpoint /run/resolvconf
+  m mountpoint /run/mount-namespaces/$nn
+}
+
 case $action in
-  start|stop)
+  start|stop|show)
     $action
     ;;
   *)