X-Git-Url: https://iankelling.org/git/?a=blobdiff_plain;f=newns;h=713c29de8cd7945314d7f9e009a6cbe364e6b0e2;hb=8d1e271a9757b6c405860effff82e9e28bd89827;hp=a2233fb700b077a6933381f812ed9b5ce89919c7;hpb=fedb02a73e1171c91cca9bc91bc357b1db8477ec;p=newns

diff --git a/newns b/newns
index a2233fb..713c29d 100755
--- a/newns
+++ b/newns
@@ -38,21 +38,22 @@ fi
 usage() {
     cat <<EOF
 usage: ${0##*/} [OPTS] start|stop NS_NAME
-Setup new or systemd created network namespace with nat and mount namespace
+Nat a network namespace. systemd friendly
 
--c, --create    Create network namespace. For running outside systemd private net.
+Also creates a mount namespace with a cloned /etc/resolv.conf.
+
+-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 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.
 
-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.
-
 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:
@@ -74,10 +75,10 @@ in.
 Background:
 
 This script does not make the namespace be named like ip does, because
-the naming is not necessary, although it could have with some more
-work. For debugging and joining the namespace with a bash shell, I use
-nsenter -n -m -t $(pgrep PROCESS_IN_NAMESPACE).  Note: if I knew how to
-easily ask systemd what pid a unit has, i would do that.
+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,
@@ -95,11 +96,12 @@ EOF
 
 #### 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 "$0: Internal error!" ; exit 1 ;;
@@ -157,15 +159,18 @@ dexec() { ip netns exec default "$@"; }
 
 # 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 "^${network//./\\.}" >/dev/null; then
             existing=true
@@ -187,13 +192,26 @@ start() {
     mkdir -p /root/mount_namespaces
     if ! mountpoint /root/mount_namespaces >/dev/null; then
         mount --bind /root/mount_namespaces /root/mount_namespaces
-        mount --make-private /root/mount_namespaces
     fi
+    # note: This is outside the mount condition because I've mysteriously
+    # had this become shared instead of private, perhaps it
+    # got remounted somehow and lost the setting.
+    mount --make-private /root/mount_namespaces
     if [[ ! -e /root/mount_namespaces/$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 ####