interfaces: T3871: name interfaces from a slot-anchored mapping store

Names used to be derived from 'hw-id' in config.boot. A node carrying an
address but no hw-id had to be matched to some NIC, and nothing available at
boot could say which one - every heuristic could bind a configured address to
the wrong physical wire.

Naming authority moves to a mapping store binding a name to the slot the
hardware occupies. The slot is the identity, so a replacement card in that
slot is that interface again. The store renders systemd .link files, leaving
udev to apply names as devices appear.

1 files changed, 36 insertions, 1 deletions

diff --git a/src/init/vyos-router b/src/init/vyos-router
index c0d12e2a7..ff1de0584 100755
--- a/src/init/vyos-router
+++ b/src/init/vyos-router
@@ -357,11 +357,39 @@ clear_or_override_config_files ()
 
 update_interface_config ()
 {
-    if [ -d /run/udev/vyos ]; then
+    # add a configuration node for any interface the name store knows about
+    # that does not have one yet
+    if [ -f /opt/vyatta/etc/config/interface-mapping.json ]; then
         $vyos_libexec_dir/vyos-interface-rescan.py $BOOTFILE
     fi
 }
 
+# warn (but do not fail boot) about configured hw-id interfaces whose
+# hardware was not found by vyos-net-name-resolve.service
+warn_missing_interface_hardware ()
+{
+    local status_file="/run/vyos-net-name-resolve.json"
+    [ -f "${status_file}" ] || return 0
+
+    jq -r '.missing // {} | to_entries[] | "\(.key) \(.value)"' "${status_file}" \
+        2>/dev/null | while read -r intf hwid; do
+        [ -n "${intf}" ] || continue
+        log_failure_msg "interface '${intf}' (${hwid}) did not appear during boot - check cabling, hardware and driver/firmware. Its name stays reserved."
+    done
+
+    # A replacement NIC inheriting a slot's name also inherits that
+    # interface's addresses, so say so rather than doing it silently.
+    jq -r '.matched_by_path // {} | to_entries[] | "\(.key) \(.value)"' "${status_file}" \
+        2>/dev/null | while read -r intf hwid; do
+        [ -n "${intf}" ] || continue
+        log_warning_msg "interface '${intf}' was matched to new hardware ${hwid} by slot - verify it is the port you expect before relying on its configuration"
+    done
+
+    if [ "$(jq -r '.persisted // true' "${status_file}" 2>/dev/null)" = "false" ]; then
+        log_failure_msg "interface names could not be persisted - the configuration volume is not mounted"
+    fi
+}
+
 cleanup_post_commit_hooks () {
     # Remove links from the post-commit hooks directory.
     # note that this approach only supports hooks that are "configured",
@@ -606,6 +634,12 @@ start ()
 
     disabled migrate || migrate_bootfile || overall_status=1
 
+    # config.boot is only guaranteed to exist/be current from this point on
+    # (tmpfs mount + encrypted config unlock + syntax migration have all
+    # already happened above) - hw-id based interface naming must be
+    # resolved against it before anything below applies interface config.
+    systemctl start vyos-net-name-resolve.service || overall_status=1
+
     restore_if_missing_bootup_script pre
 
     run_preconfig_script
@@ -613,6 +647,7 @@ start ()
     run_postupgrade_script
 
     update_interface_config || overall_status=1
+    warn_missing_interface_hardware
 
     disabled system_activate || system_activate || overall_status=1