X-Git-Url: https://iankelling.org/git/?p=newns;a=blobdiff_plain;f=newns;h=5075181714569a1d6e8980bb27af27e0b3eb4ddc;hp=e4d1f8622c971fe7a6882f54b3783fd9057ca5e8;hb=HEAD;hpb=8448af56da3cdaf6e50b075b322031e8edeb9760

diff --git a/newns b/newns
index e4d1f86..9a7a6ff 100755
--- a/newns
+++ b/newns
@@ -1,5 +1,12 @@
 #!/bin/bash
-# Copyright (C) 2017 Ian Kelling
+# I, Ian Kelling, follow the GNU license recommendations at
+# https://www.gnu.org/licenses/license-recommendations.en.html. They
+# recommend that small programs, < 300 lines, be licensed under the
+# Apache License 2.0. This file contains or is part of one or more small
+# programs. If a small program grows beyond 300 lines, I plan to switch
+# its license to GPL.
+
+# Copyright 2024 Ian Kelling
 
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -16,49 +23,64 @@
 
 [[ $EUID == 0 ]] || exec sudo -E "$BASH_SOURCE" "$@"
 
-if [[ ! $ERRHANDLE_PATH ]]; then
-    ERRHANDLE_PATH=$(readlink -f "${BASH_SOURCE}")
-    ERRHANDLE_PATH=$(readlink -f ${ERRHANDLE_PATH%/*}/../errhandle)
-fi
-err_sourced=true
-for p in $ERRHANDLE_PATH/{errcatch-function,bash-trace-function}; do
-    if [[ -e $p ]]; then
-        source $p
-    else
-        err_sourced=false
-    fi
-done
-if $err_sourced; then
-    errcatch
-else
-    set -eE -o pipefail
-    trap 'echo "$0:$LINENO:error: \"$BASH_COMMAND\" returned $?" >&2' ERR
-fi
+# https://savannah.nongnu.org/projects/bash-bear-trap/
+set -e; . /usr/local/lib/bash-bear; set +e
+
+m() {
+  local out
+  printf "newns: %s\n" "$*"
+  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
+  cat <<EOF
+usage: ${0##*/} [OPTS] start|stop|show NS_NAME
 Nat a network namespace. systemd friendly
 
 Also creates a mount namespace with a cloned /run/resolvconf.
 
--c, --create    Create a named network namespace. When running from
+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 one starting at 10.173.1
+                the first unused one starting at 10.173.1
 -h, --help      Show this help and exit.
 
-From within a systemd network namespace, nat it to the outside. This
+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=/run/mount-namespaces/NAME --net=/var/run/netns/NAME bash
+
+If you dont care about the mount namespace, you can leave that option off.
+
+
+For systemd:
+
+From within a systemd network namespace, we nat it to the outside. This
 would be called from ExecStartPre, and or subsequent units called with
 JoinsNamespaceOf= and PrivateNetwork=true.
 
-Also create a named mount namespace under /root/mount_namespaces, so we
-can alter some system config for this namespace. Subsequent systemd
-command lines would be prefixed with:
+If resolvconf is installed, we create a named mount namespace under
+/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
@@ -66,19 +88,19 @@ files, so the mount namespace won't be needed for most use cases, and I
 will update the script to that the mount namespace not created unless a
 flag is passed in. Patch welcome to add that flag before then.
 
-A recommmended dependency of this script is my other repo named "errhandle",
-which prints stack trace on error, and calls a cleanup function:
-https://iankelling.org/git/?p=errhandle, set ERRHANDLE_PATH, or put it
-in a directory adjacent to the absolute, resolved directory this file is
-in.
 
-Background:
+This script has a dependency
+https://savannah.nongnu.org/projects/bash-bear-trap/ . Search the script for "source" to see where to install or modify the  installed location.
 
-This script does not make the namespace be named like ip does, because
-the naming is not necessary, although it could have been done with some
-more work. For debugging and joining the namespace with a bash shell, I
-use nsenter -n -m -t \$(pgrep PROCESS_IN_NAMESPACE) bash.  Note: if I
-knew how to easily ask systemd what pid a unit has, i would do that.
+
+Background on this project (you can skip if you like):
+
+If we aren't creating a named network namespace, to join the namespace
+with a shell, I use:
+nsenter -n -m -t \$(pgrep PROCESS_IN_NAMESPACE) bash
+
+Note: if I knew how to easily ask systemd what pid a unit has, i would
+do that.
 
 "ip netns new ..." also does a mount namespace, then bind
 mounts each file/dir in /etc/netns/NS_NAME to /etc/NS_NAME. Note,
@@ -87,30 +109,33 @@ 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.
 
+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
-    exit ${1:-0}
+  exit ${1:-0}
 }
 
 
 #### 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
-        -c|--create) create=true; shift ;;
-        -n) network=$2; shift 2 ;;
-        -h|--help) usage ;;
-        --) shift; break ;;
-        *) echo "$0: Internal error!" ; exit 1 ;;
-    esac
+  case $1 in
+    -c|--create) create=true; shift ;;
+    -n) network=$2; shift 2 ;;
+    -h|--help) usage ;;
+    --) shift; break ;;
+    *) echo "$0: Internal error!" ; exit 1 ;;
+  esac
 done
 if (( $# != 2 )); then
-    usage 1
+  usage 1
 fi
 
 action=$1
@@ -120,191 +145,262 @@ nn=$2 # namespace name
 #### begin sanity checking ####
 install_error=false
 if ! type -p ip &>/dev/null; then
-    echo "please install the iproute2 package"
-    install_error=true
+  echo "please install the iproute2 package"
+  install_error=true
 fi
 if ! type -p iptables &>/dev/null; then
-    echo "please install the iptables package"
-    install_error=true
+  echo "please install the iptables package"
+  install_error=true
 fi
 if $install_error; then
-    exit 1
+  exit 1
 fi
 ####   end sanity checking ####
 
-
 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
-    mkdir -p /run/netns
-    ln -s /proc/1/ns/net $target
+  # -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
-    # 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 "$@"; }
+  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 "$@"; }
 
 
-# background: head -n1 is defensive. Not sure if there is some weird feature
-# for 2 routes to be 0/0.
-gateway_if=$(ipd route list exact 0/0 | head -n1| sed -r 's/.*dev\s+(\S+).*/\1/')
-nat() { dexec iptables -t nat $1 POSTROUTING -o $gateway_if -j MASQUERADE \
-              -m comment --comment "systemd network namespace nat"; }
-
-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
-            found=true
-            break
-        fi
-    done
+nat() {
+  # 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 \
+        -m comment --comment "systemd network namespace nat"
 }
 
-start() {
-    find_network
-    if ! $found; then
-        echo "$0: error: no open network found"
-        exit 1
-    fi
+# d = default
+diptables-add() {
+  if ! dexec iptables -C "$@" &>/dev/null; then
+    dexec iptables -I "$@"
+  fi
 
-    #### 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
-    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
-    fi
-    if ! mountpoint /root/mount_namespaces/$nn >/dev/null; then
-        # documentation on propagation is a bit weird because it
-        # confusingly talks about binds, namespaces, and mirrors (which
-        # seems to be just another name for bind), shared subtrees
-        # (which seems to a term for binds and namespaces), and does not
-        # properly specify whether the documentation applies to binds,
-        # namespaces, or both. Notably, propagation for binds is marked
-        # on the original mount point, and propagation for a mount
-        # namespace is marked on mounts within the namespace.  Here, we
-        # specify that we want mount changes propagated to us, but not
-        # back.
-        unshare --propagation slave --mount=/root/mount_namespaces/$nn /bin/true
+}
+
+find-network() {
+  if [[ $network ]]; then
+    return
+  fi
+  found=false
+  ips="$(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
+      found=true
+      break
     fi
+  done
+  if ! $found; then
+    echo "$0: error: no open network found"
+    exit 1
+  fi
+}
 
-    ####   end mount namespace setup ####
+# ip add idempotent (if it doesn't exist already)
+ip-add() {
+  local cmd net dev
+  cmd=$1
+  net=$2
+  dev=$3
+  if ! $cmd addr show dev $dev | sed 's/^ *//;s/ *$//' | grep -xF "inet $net scope global $dev"; then
+    $cmd addr add $net dev $dev
+  fi
 
+}
 
-    if $create; then
-        ip netns add $nn
-        ip -n $nn link set dev lo up
+start() {
+  find-network
+
+  #### begin mount namespace setup ####
+  mkdir -p /run/mount-namespaces
+  if ! mountpoint /run/mount-namespaces >/dev/null; then
+    mount --bind /run/mount-namespaces /run/mount-namespaces
+  fi
+  # note: This is outside the mount condition because I've mysteriously
+  # had this become shared instead of private, perhaps it
+  # got remounted somehow and lost the setting.
+  mount --make-private /run/mount-namespaces
+  if [[ ! -e /run/mount-namespaces/$nn ]]; then
+    touch /run/mount-namespaces/$nn
+  fi
+  if ! mountpoint /run/mount-namespaces/$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.
+    #
+    # slave is documented in.
+    # /usr/share/doc/linux-doc-4.9/Documentation/filesystems/sharedsubtree.txt.gz
+    # 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 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=/run/mount-namespaces/$nn /bin/true
+  fi
+
+  ####   end mount namespace setup ####
+
+
+  if $create; then
+    if ! ip netns | grep -xF $nn &>/dev/null; then
+      ip netns add $nn
     fi
-
-    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
+    ip -n $nn link set dev lo up
+  fi
+
+  echo 1 | dexec dd of=/proc/sys/net/ipv4/ip_forward status=none
+
+  # docker helpfully changes the default FORWARD to drop...
+  diptables-add FORWARD -i $v0 -j ACCEPT
+  diptables-add FORWARD -o $v0 -j ACCEPT
+
+
+  err-cleanup() { stop; }
+  ipnn link add $v0 type veth peer name $v1
+  ipnn link set $v0 netns default
+  ip-add ipd $network.1/24 $v0
+  ipd link set $v0 up
+  nat -C &>/dev/null || nat -A
+  ip-add ipnn $network.2/24 $v1
+  ipnn link set $v1 up
+  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
+
 
-    _errcatch_cleanup=stop
-    ipnn link add $v0 type veth peer name $v1
-    ipnn link set $v0 netns default
-    ipd addr add $network.1/24 dev $v0
-    ipd link set $v0 up
-    nat -C &>/dev/null || nat -A
-    ipnn addr add $network.2/24 dev $v1
-    ipnn link set $v1 up
-    ipnn route add default via $network.1
-
-    ###### begin setup resolvconf
+  ###### begin setup resolvconf
+  if [[ -e /run/resolvconf ]]; then # resolvconf probably installed
     resolv_copy=/root/resolvconf-$nn
 
     # this condition should never happen, just coding defensively
     if mexec mountpoint /run/resolvconf &>/dev/null; then
-        mexec umount /run/resolvconf
+      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
+      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
+      mexec sed --follow-symlinks -i "s/nameserver 127\..*/nameserver $network.1/" /run/resolvconf/interface/lo.dnsmasq
+      mexec resolvconf -u
     fi
+    # and in debian based distros at least, it runs with --local-service, and needs a restart
+    # to know about the new local network
+    if [[ $(systemctl --no-pager show -p ActiveState dnsmasq ) == ActiveState=active ]]; then
+      systemctl restart dnsmasq
+    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
-
 
+  fi # end if [[ -e /run/resolvconf ]]
+  ######  end setup resolvconf
 }
 
 stop() {
-    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
-    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 [[ ! $network ]]; then
+    network=$(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
+  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 $nn
+  fi
+
+  # not sure this is necessary since we are tearing down the mount namespace
+  if mexec mountpoint /run/resolvconf &>/dev/null; then
+    mexec umount /run/resolvconf
+  fi
+
+  if mountpoint /run/mount-namespaces/$nn >/dev/null; then
+    umount /run/mount-namespaces/$nn
+  fi
+}
 
-    if mountpoint /root/mount_namespaces/$nn >/dev/null; then
-        umount /root/mount_namespaces/$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)
-        $action
-        ;;
-    *)
-        echo "$0: error: unsupported action"
-        exit 1
-        ;;
+  start|stop|show)
+    $action
+    ;;
+  *)
+    echo "$0: error: unsupported action"
+    exit 1
+    ;;
 esac