delta/NetworkManager.git, branch th/pr/1287

fix: dont return NM_DEVICE_CAP_IS_SOFTWARE for loopback

2022-07-26T15:20:00+00:00

You cannot create or delete loopback devices. They are always there.
Sure, they are purely in software, but still.

If we return that loopback has NM_DEVICE_CAP_IS_SOFTWARE, there are
probably other places that needs to be fixed, where we need to special
case loopback for this reason.

It might be easier to just claim that loopback is not a software device.

fix: choose very low route metric for loopback devices

2022-07-26T15:16:03+00:00

It's local. It seems if the user adds a route to that destination,
they would want it preferred over most other routes. Choose a small
metric.

fix: don't add loopback addr during _l3_commit_one()

2022-07-26T15:11:59+00:00

It's already added to the combined NML3ConfigData. Why here again?

fix: set a_no_auto_noprefixroute for loopback addr

2022-07-26T15:09:18+00:00

The default value is already FALSE. That cannot be correct, because
we need a non-default behavior here.

style: leave blank line after variable declarations

2022-07-26T15:07:39+00:00

fix: version macros/comment

2022-07-26T15:06:11+00:00

fix: usage of klass->allow_autoconnect_on_external is not right

2022-07-26T15:04:13+00:00

This needs a different solution. nm_device_autoconnect_allowed() cannot
force autoconnect to happen, it can only prevent it. There is only one caller
of this, NMPolicy.

That one already does:

    if (nm_device_get_act_request(device))
        return;

    if (!nm_device_autoconnect_allowed(device))
        return;

meaning, it won't autoconnect if we have an activation request
already.

However, the entire point of the exercies is to allow autoconnect,
even if the device has an (external) activation request.

fix: drop unused @NM_UNMANAGED_BY_TYPE

2022-07-26T14:57:00+00:00

fix: is_loopback() needs adjustment

2022-07-26T14:55:50+00:00

loopback: preserve the IP addresses when deactivating

2022-07-26T14:53:16+00:00

Signed-off-by: Wen Liang