| Age | Commit message (Collapse) | Author |
|---|
|
Local-address should be checked/executed only if it exists in the
openvpn configuration, dictionary, jinja2 template
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Add support for OpenVPN client mode with only the CA certificate of the server
installed.
|
|
|
|
|
|
|
|
|
|
Custom OpenVPN options moved back to the command line from a
configuration file. This should keep full compatibility with the
`crux` branch, and allows to avoid mistakes with parsing options
that contain `--` in the middle.
The only smart part of this - handling a `push` option. Because
of internal changes in OpenVPN, previously it did not require an
argument in the double-quotes, but after version update in
`equuleus` and `sagitta` old syntax became invalid. So, all the
`push` options are processed to add quotes. The solution is still
not complete, because if a single config line contains `push` with
other options, it will not work, but it is better than nothing.
|
|
|
|
Add support for OpenVPN client mode with only the CA certificate of the server
installed.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
This reverts commit c414479fdf1d5ad77170f977481fb9197c9559ae.
This commit broke the smoketests and also OpenVPN complains:
Options error: You must define certificate file (--cert) or PKCS#12 file (--pkcs12)
|
|
This makes the tls cert-file and key-file optional and allows for more
advanced configurations via "openvpn-option", such as pkcs11 or pkcs12
options.
|
|
|
|
Interface.get_config() was always a pure helper which exposed a "per interface
type" dictionary which was then fed by the caller to create interfaces by
iproute2 which required additional options during creation time.
Such interfaces had been:
* tunnel
* vxlan
* geneve
* macsec
* wifi
* macvlan / pseudo-ethernet
The code was always duplicated to convert from the VyOS CLI based get_config_dict()
to a dict which can be used to feed iproute2.
This path has been removed and we now always feed in the entire dictionary
retrieved by get_config_dict() or in the interfaces case, it's high-level wrapper
get_interface_dict() to the interface we wan't to create.
This also adds the - personally long awaited - possibility to get rid of the
derived tunnel classes for e.g. GRE, IPIP, IPIP6 and so on.
|
|
|
|
|
|
|
|
When rendering the configs "ifconfig" statement wrong IP addresses have been
used for the "tun" operating mode. This has been corrected.
|
|
|
|
|
|
We had two places were the is_ip, is_ipv4 and is_ipv6 helpers had been defined.
All places now have been converged into vyos.template as they are used both
in the Jinja2 templates and also in our scripts.
|
|
|
|
This commit partially reverts commit eb1ed5e518 ("openvpn: T2994: re-add
ifconfig-pool statement in server config").
|
|
Re-organize the template code and add addtitional Jinja2 filters for processing
the ifconfig-pool statement. This reverts the changes from commit 7e546be9
("openvpn: T2994: temporary revert to 1.2 crux behavior for client pools").
|
|
Remove workaround which split (local|remote)_address and also subnet keys into
individual keys for the assigned IP address family (4/6).
During template rendering check IP version by introducing new ipv4 and ipv6
Jinja2 filters {% if foo | ipv4 %} or {% if bar | ipv6 %} options.
|
|
|
|
|
|
|
|
A lot of VyOS code requires the Kernel interface to be present in order to
properly work and adjust the interface to the users CLI intends (alias, ipv6,
vrf - just to name a few).
OpenVPN - when run in client mode - only creates the interface (e.g. vtun1) when
the connection to the OpenVPN server was successful. This can't be always the
case due to e.g. software-updates or routing issues to the remote side. This
will in the end result in a zombie OpenVPN client interface where some config
items might not have been set when the interface finally comes up - imagine a
wrong assigned VRF instance.
By always creating the OpenVPN interface manuall we ensure that all the CLI
settings are properly configured in the OS kernel.
|
|
The current implementation for bridge based interfaces has an issue which is
caused by priority inheritance. We always assumed that the bridge interface will
be created last, but this may not be true in all cases, where some interfaces
will be created "on demand" - e.g. OpenVPN or late (VXLAN, GENEVE).
As we already have a bunch of verify steps in place we should not see a bridge
interface leak to the underlaying infrastructure code. This means, whenever an
interface will be member of a bridge, and the bridge does yet not exist, we will
create it in advance in the interface context, as the bridge code will be run
in the same commit but maybe sooner or later.
This will also be the solution for T2924.
|
|
|
