Bug 1512316

when you add a software-device (like dummy, bridge, bond) outside of NM, NM recognizes that it was added outside of NetworkManager. As long as it has no IP addresses, NM marks it as unmanaged.

Once you externally up the interface, and add IP addresses, NM will generate an external connection to represent that something is active on the device. That does not involve touching the device or interfering with it.

When you remove the IP address, NM will cleanup  the generated connection, and mark the device as unmanaged again.

To avoid that, configure the device as explicitly unmanaged by one of the various means:
 - ENV{NM_UNMANAGED}="1" in udev
 - device*.managed = `man NetworkManager.conf`
 - nmcli device set "$DEV" managed no (this way does not persist across 
     reboot, but across restart of the service)
 - NM_CONTROLLED=no in an ifcfg file
 - keyfile.unmanaged-devices in `man NetworkManager.conf` (deprecated)
Unless you do that, NM behaves as explained above.


Does that cause a problem for you?

Yay, sorry, my reproducer does not reflect the problem. It looks like NM does set the IP back. Here is a new reproducer. With NM running, it fails in a second (found ipv6 address after ipv6 del). With NM stopped, it runs without an error.

#!/usr/bin/bash

DUMMY=dummy_111

function set_up {
    ip l add $DUMMY type dummy
    ip l set up dev $DUMMY
}

function tear_down {
    ip l del $DUMMY
}

function test_add_del_addr {
    tear_down
    set_up

    ip a add 192.168.212.212/24 dev $DUMMY
    ip a | grep -q '192.168.212.212/24' || { echo missing ipv4 address after ipv4 add; exit 1; }

    ip a add 2002:99::1/64 dev $DUMMY
    ip a | grep -q '192.168.212.212/24' || { echo missing ipv4 address after ipv6 add; exit 1; }
    ip a | grep -q '2002:99' || { echo missing ipv6 address after ipv6 add; exit 1; }

    ip a del 2002:99::1/64 dev $DUMMY
    ip a | grep -q '192.168.212.212/24' || { echo missing ipv4 address after ipv6 del; exit 1; }
    ip a | grep -q '2002:99' && { echo found ipv6 address after ipv6 del; exit 1; }

    ip a del 192.168.212.212/24 dev $DUMMY
    ip a | grep -q '192.168.212.212/24' && { echo found ipv4 address after ipv4 del; exit 1; }
    ip a | grep -q '2002:99' && { echo found ipv6 address after ipv4 del; exit 1; }
}

for i in `seq 1 100` :; do
    test_add_del_addr
done

Created attachment 1353316 [details]
device: start managing external devices on reapply

In the next commit we will modify ipX_config_merge_and_apply to never
touch external devices. When a "reapply" call is issued on an external
device we are no longer simply tracking its state but we are actively
managing it and so its sys-iface-state must be promoted to managed.

Created attachment 1353317 [details]
device: don't touch external devices

If a device is 'external' (which means that NM generated an in-memory
connection to only to track the device state) we should not change its
IP configuration.

both lgtm.

Applied to master:

https://cgit.freedesktop.org/NetworkManager/NetworkManager/commit/?id=9e41ed446176acc3f82cf23c3b6d7692ffa7b5b3

https://cgit.freedesktop.org/NetworkManager/NetworkManager/commit/?id=60334a2893fe27241151acb4d022bf685c664c62

and nm-1-10.

*** Bug 1526021 has been marked as a duplicate of this bug. ***

automated, test running and passing

*** Bug 1523572 has been marked as a duplicate of this bug. ***

Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.

For information on the advisory, and where to find the updated
files, follow the link below.

If the solution does not work for you, open a new bug report.

https://access.redhat.com/errata/RHBA-2018:0778