X-Git-Url: https://iankelling.org/git/?p=newns;a=blobdiff_plain;f=newns;h=823d184b5b07b323d17acbe91a46b5b2a3ccd694;hp=31f1b528c80bdab325c17631adf5c6245c745895;hb=6c826f3613b0c3b86257b111b3078013fdf34b78;hpb=914cdb74031df6ae068ed87376c47944fbfaf934

diff --git a/newns b/newns
index 31f1b52..823d184 100755
--- a/newns
+++ b/newns
@@ -37,24 +37,26 @@ fi
 
 usage() {
     cat <<EOF
-usage: ${0##*/} [OPTS] start|stop NETNS_NAME
-Setup new or systemd created network namespace with nat and mount namespace
-
--c, --create    Create network namespace. For running outside systemd private net.
+usage: ${0##*/} [OPTS] start|stop NS_NAME
+Nat a network namespace. create a mount ns. systemd friendly
+
+-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.
+-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.
 
-From within systemd network namespace, nat it to the outside. If given
--c, or if in the default network namespace, create a named network
-namepace natted to the current netns.
-
-Uses /24 network, finding the first locally unused one starting at
-10.173.0.
+From within a systemd network namespace, 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:
 
-/usr/bin/nsenter --mount=/root/mount_namespaces/NETNS_NAME
+/usr/bin/nsenter --mount=/root/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
@@ -68,8 +70,16 @@ 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: "ip netns new ..." also does a mount namespace, then bind
-mounts each file/dir in /etc/netns/NETNS_NAME to /etc/NETNS_NAME. Note,
+Background:
+
+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.
+
+"ip netns new ..." also does a mount namespace, then bind
+mounts each file/dir in /etc/netns/NS_NAME to /etc/NS_NAME. Note,
 for openvpn having it's own resolv.conf by using it's user script which
 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
@@ -82,13 +92,14 @@ EOF
 }
 
 
-## begin arg parsing ##
+#### begin arg parsing ####
 create=false
-temp=$(getopt -l help,create hc "$@") || usage 1
+temp=$(getopt -l help,create hcn: "$@") || 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 ;;
@@ -99,11 +110,10 @@ if (( $# != 2 )); then
 fi
 
 action=$1
-nn=$2 # network namespace / namespace name
-## end arg parsing ##
-
-## begin sanity checking ##
+nn=$2 # namespace name
+#### end arg parsing ####
 
+#### begin sanity checking ####
 install_error=false
 if ! type -p ip &>/dev/null; then
     echo "please install the iproute2 package"
@@ -116,8 +126,7 @@ fi
 if $install_error; then
     exit 1
 fi
-
-##   end sanity checking ##
+####   end sanity checking ####
 
 
 v0=veth0-$nn
@@ -128,10 +137,10 @@ if ! $create && [[ $(readlink /proc/self/ns/net) == "$(readlink /proc/1/ns/net)"
     create=true
 fi
 
+# make the default network namespace be named
 target=/run/netns/default
 if [[ ! -e $target && ! -L $target ]]; then
     mkdir -p /run/netns
-    # make the default network namespace be named
     ln -s /proc/1/ns/net $target
 fi
 
@@ -146,17 +155,20 @@ fi
 dexec() { ip netns exec default "$@"; }
 
 
-# head -n1 is defensive. Not sure if there is some weird feature
+# background: head -n1 is defensive. Not sure if there is some weird feature
 # for 2 routes to be 0/0.
-gateway_if=$(ipd route list exact 0/0 | head -n1| sed -r 's/.*\s(\S+)\s*$/\1/')
+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 $2}')"
-    for ((i=0; i <= 254; i++)); do
+    for ((i=1; i <= 254; i++)); do
         network=$ip_base.$i
         if printf "%s\n" "$ips" | grep "^${network//./\\.}" >/dev/null; then
             existing=true
@@ -168,24 +180,38 @@ find_network() {
 }
 
 start() {
-
     find_network
     if ! $found; then
         echo "$0: error: no open network found"
         exit 1
     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
-        mount --make-private /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/$nn >/dev/null; then
-        unshare --mount=/root/mount_namespaces/$nn
+        # 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
     fi
+    ####   end mount namespace setup ####
 
 
     if $create; then
@@ -193,8 +219,6 @@ start() {
         ip -n $nn link set dev lo up
     fi
 
-
-
     echo 1 | dexec dd of=/proc/sys/net/ipv4/ip_forward 2>/dev/null
 
     _errcatch_cleanup=stop