Nmcli was deducing the port-type in some cases, obtaining it from
connection.type and/or connection.controller, then setting it in the
connection.
As nmcli processes the connection properties in order, a user provided
port-type was only used correctly if it was placed before
connection.type and connection.controller. If placed after, as in OVS
might be more than one interface with the same name, the wrong port-type
could be deduced from connection.controller because an interface of the
wrong type might be found first.
It is not needed at all to set the port-type from the client, the daemon
already does the right validation and adds it when possible. This patch
removes this logic from the client.
There are several places where nmcli/connections.c checks the port-type,
so they could be potentially affected, but:
1. They should already work without a port-type because the connection
didn't have one until type, controller or port-type was processed.
2. They are mostly code for autocomplete or generating the profile name,
so not something critical.
3. Anyway, if anything breaks due to this change, it should be fixed in
a different way, without setting a deduced port-type in the side of
the client, as this is a task for the daemon.
This change has an additional possitive side effect: for OVS, where the
port-type can be fully deduced from connection.type, it is not needed to
manually specify the port-type anymore. The daemon was capable to deduce
it, unlike nmcli.
Fixes: c5324ed285 ('nmcli: streamline connection addition')
Prevents NetworkManager from trying to determine the
transient hostname via DHCP or other means if "localhost"
is already configured as a static hostname, as the transient
hostname will be ignored by hostnamed if a static hostname
has already been set.
They cannot be used yet as Dracut need changes. They are not really
tested, neither, so better not to install them as they might cause
confusion about their purpose.
This allows us to drag in packages that replace something in the base
image. Notably, this fixes installing pppd on Fedora 42 that requires
systemd (whereas the image has conflicting systemd-standalone-sysusers).
(cherry picked from commit 979717df4e)
Attempting to install these takes time and they are no longer there
since ancient times. CentOS 9 and later are fine.
(cherry picked from commit 39efc8a394)
More fallout from set -e patch, encountered during a release.
Fixes: 64f8ec6099 ('rpm/build*.sh: trap failures')
(cherry picked from commit aea231f489)
The idea is to create a pair of VLAN and MACVLAN with AddAndActivate if
they are not present, and otherwise follow the ordinary (GetApplied &
Reapply) procedure if the devices are already present.
This reverts commit 87d7ccc580.
(cherry picked from commit 7eee669217)
This will be useful for updating configuration of Vlans and MacVlans,
some of having same MAC addresses as devices of other type.
This reverts commit cee0515f13.
(cherry picked from commit 6cedd594b6)
Baremetal instances in Oracle Cloud require special VLAN config. Parse
the metadata related to it.
This reverts commit 5eefd2d59c.
(cherry picked from commit 5c3efeef15)
This is done so that AddAndActivate() will return sensible errors in a
future patch that makes it support creating virtual devices.
In effect, all errors are logged in one place, therefore the log levels
are different. I don't think we're losing anything of value by being
a little less verbose here.
(cherry picked from commit 45d82f720c)
The idea is to create a pair of VLAN and MACVLAN with AddAndActivate if
they are not present, and otherwise follow the ordinary (GetApplied &
Reapply) procedure if the devices are already present.
This reverts commit 87d7ccc580.
This will be useful for updating configuration of Vlans and MacVlans,
some of having same MAC addresses as devices of other type.
This reverts commit cee0515f13.
This is done so that AddAndActivate() will return sensible errors in a
future patch that makes it support creating virtual devices.
In effect, all errors are logged in one place, therefore the log levels
are different. I don't think we're losing anything of value by being
a little less verbose here.
Decouple from a specific initrd generator (dracut) the systemd services that
provide networking in the initrd using NM, thus allowing other systemd-based
initrd generators to take advantage of it.
These new services are:
- `NetworkManager-config-initrd.service`: it starts very early at boot, parses
the kernel command line using `nm-initrd-generator` and sets the
`/run/NetworkManager/initrd/neednet` flag to activate the other initrd services,
and also sets the hostname if needed.
- `NetworkManager-initrd.service`: it basically does the same job as the
`NetworkManager.service`, but in the initrd.
- `NetworkManager-wait-online-initrd.service`: ordered before
`network-online.target`, it will allow other services that require networking to
delay their start until NM has finished.
This allows us to drag in packages that replace something in the base
image. Notably, this fixes installing pppd on Fedora 42 that requires
systemd (whereas the image has conflicting systemd-standalone-sysusers).
Domains are exported via D-Bus and so they must be valid UTF-8.
RFC 1035 specifies that domain labels can contain any 8 bit values,
but also recommends that they follow the "preferred syntax" which only
allows letters, digits and hypens.
Don't introduce a strict validation of the preferred syntax, but at
least discard non UTF-8 search domains, as they will cause assertion
failures later when they are sent over D-Bus.
The variable 'change' may be used uninitialized.
Fixes: 7acc66699a ('policy: always reset retries when unblocking children or ports')
(cherry picked from commit af6aca3527)
