Package: grml-debootstrap Version: 0.88 Severity: important We noticed that the network configuration generated by grml-debootstrap is unreliable for the predictable network naming schema (when net.ifname=0 isn't set).
The underlying code in grml-debootstrap currently iterates over all
network devices via:
for interface in $(udevadm info -e | sed -n -e 's/E:
ID_NET_NAME_PATH=\([^$*]\)/\1/p'); do
[...]
This means that we get something like the following for a (Proxmox)
VM with a virtio network device:
# udevadm info -e | grep 'ID_NET_NAME_PATH'
E: ID_NET_NAME_PATH=enp0s18
While on actual boot the device is named ens18 then. This results in
the VM having a broking network configuration.
I managed to track this down already. In systemd's link_config_apply
function (defined in src/udev/net/link-config.c) we have:
| switch (policy) {
| case NAMEPOLICY_KERNEL:
| if (name_type != NET_NAME_PREDICTABLE)
| continue;
|
| /* The kernel claims to have given a predictable name, keep
it. */
| log_device_debug(device, "Policy *%s*: keeping predictable
kernel name",
| name_policy_to_string(policy));
| goto no_rename;
| case NAMEPOLICY_KEEP:
| if (!IN_SET(name_type, NET_NAME_USER, NET_NAME_RENAMED))
| continue;
|
| log_device_debug(device, "Policy *%s*: keeping existing
userspace name",
| name_policy_to_string(policy));
| goto no_rename;
| case NAMEPOLICY_DATABASE:
| (void) sd_device_get_property_value(device,
"ID_NET_NAME_FROM_DATABASE", &new_name);
| break;
| case NAMEPOLICY_ONBOARD:
| (void) sd_device_get_property_value(device,
"ID_NET_NAME_ONBOARD", &new_name);
| break;
| case NAMEPOLICY_SLOT:
| (void) sd_device_get_property_value(device,
"ID_NET_NAME_SLOT", &new_name);
| break;
| case NAMEPOLICY_PATH:
| (void) sd_device_get_property_value(device,
"ID_NET_NAME_PATH", &new_name);
| break;
| case NAMEPOLICY_MAC:
| (void) sd_device_get_property_value(device,
"ID_NET_NAME_MAC", &new_name);
| break;
| default:
| assert_not_reached("invalid policy");
| }
So if a device has the propery ID_NET_NAME_PATH set (e.g. enp0s18),
but also ID_NET_NAME_SLOT (ens18), we can't just use the value from
ID_NET_NAME_PATH but have to follow the precedence.
I tend to call this a RC bug, so we should get this fixed for
Debian/buster.
regards
-mika-
signature.asc
Description: Digital signature
