Age | Commit message (Collapse) | Author |
|
|
|
|
|
|
|
In the past whenever a change happened to any interface and it was configured
as a DHCP client, VyOS always had a breif outage as DHCP released the old lease
and re-aquired a new one - bad!
This commit changes the behavior that DHCP client is only restarted if any one
of the possible options one can set for DHCP client under the "dhcp-options"
node is altered.
|
|
* 'firewall' of https://github.com/sarthurdev/vyos-1x:
zone_policy: T3873: Implement intra-zone-filtering
policy: T2199: Migrate policy route op-mode to XML/Python
policy: T2199: Migrate policy route to XML/Python
zone-policy: T2199: Migrate zone-policy op-mode to XML/Python
zone-policy: T2199: Migrate zone-policy to XML/Python
firewall: T2199: Migrate firewall op-mode to XML/Python
firewall: T2199: Migrate firewall to XML/Python
|
|
Traffic-policy rules are generated by old Perl code
This commit prevents to unexpected override this code by python.
|
|
|
|
|
|
|
|
There are not any reason to enable only DHCP or only static address
on interface at the same time
It is possible to have both.
|
|
It seems not all systems have eth0 - get a list of all available Ethernet
interfaces on the system (without VLAN subinterfaces) and then take the
first one.
|
|
We can not pass None as VRF name, this raises an exception.
OSError: [Errno 255] failed to run command: ip link set dev eth2 master None
(cherry picked from commit e687502b1cf4a3e15c562a3662afcbe0776b1fe7)
|
|
Commit 081e23996f (vyos.ifconfig: get_mac_synthetic() must generate a stable
"MAC") calculated a "stable" synthetic MAC address per the interface based on
UUID and the interface name. The problem is that this calculation is too stable
when run on multiple instances of VyOS on different hosts/hypervisors.
Having R1 and R2 setup a connection both via "tun10" interface will become the
same "synthetic" MAC address manifesting in the same link-local IPv6 address.
This e.g. breaks OSPFv3 badly as both neighbors communicate using the same
link-local address.
As workaround one can:
set interfaces tunnel tun1337 address 'fe80::1:1337/64'
set interfaces tunnel tun1337 ipv6 address no-default-link-local
This commit changes the way in how the synthetic MAC address is generated. It's
based on the first 48 bits of a sha256 sum build from a CPU ID retrieved via
DMI, the MAC address of eth0 and the interface name as used before. This should
add enough entropy to get a stable pseudo MAC address.
|
|
|
|
Commit dd2eb5e5686655 ("dhcp: T3300: add DHCP default route distance") changed
the logic on how the DHCP process is going to be started. The systemd unit was
always "started" even if it was already running. It should rather be re-started
to track changes in e.g. the DHCP hostname setting.
|
|
|
|
Commit b7d30137b1 ("vyos.ifconfig: provide generic get_mac_synthetic() method")
provided a common helper to generate MAC addresses used by EUI64 addresses for
interfaces not having a layer2 interface (WireGuard or ip tunnel).
The problem is that every call to the helper always yielded a new MAC address.
This becomes problematic when IPv6 link-local addresses are generated and
modified on the interface as multiple link-local (fe80::/64) addresses can
easily be added to the interface leaving ... a mess.
This commit changes the way how the "synthetic" MAC is generated, we generate a
UUID which is stable as it is based on the interface name. We take out the last
48 bits of the UUID and form the "MAC" address.
|
|
When using VRRP on any given interface and performing an action against that
interface - be it even only changing the alias - will trigger a removal of the
VRRP IP address.
The issue is caused by:
# determine IP addresses which are assigned to the interface and build a
# list of addresses which are no longer in the dict so they can be removed
cur_addr = self.get_addr()
for addr in list_diff(cur_addr, new_addr):
When the script calls into the library - we will drop all IP addresses set on
the adapter but not available in the config dict.
We should only remove the IP addresses marked by the CLI to be deleted!
|
|
There is no need to alter interface parameters if they have not changed at all.
|
|
Some tc qdisc rules are generated by old perl code
It prevent to unexpected override this code by python.
|
|
Check eui64_old value before deleting
It can be empty or not ipv6 address.
(cherry picked from commit 0de23064b9d575ce0569839e3b4453a0c2e9dc1c)
|
|
WireGuard, Tunnel and also PPPoE all need a ways to calculate a synthetic MAC
address used for the EUI64 link-local addresses. Instead of copying the code
from Tunnel to WireGuard to PPPoE, use a generic implementation.
|
|
level
Getting rid of "set firewall options" and move it from:
set firewall options interface ethX adjust-mss 1400
set firewall options interface ethX adjust-mss6 1400
to:
set interfaces ethernet ethX ip adjust-mss 1400
set interfaces ethernet ethX ipv6 adjust-mss 1400
In addition add an extra option called clamp-mss-to-pmtu instead of a value.
|
|
Currently, all VRFs share the same connection tracking table, which can
lead to problems:
- traffic leaks to a wrong VRF
- improper NAT rules handling when multiple VRFs contain the same IP
networks
- stateful firewall rules issues
The commit implements connection tracking zones support. Each VRF
utilizes its own zone, so connections will never mix up.
It also adds some restrictions to VRF names and assigned table numbers,
because of nftables and conntrack requirements:
- VRF name should always start from a letter (interfaces that start from
numbers are not supported in nftables rules)
- table number must be in the 100-65535 range because conntrack supports
only 65535 zones
|
|
|
|
|
|
It is not possible to change the VLAN encapsulation protocol "on-the-fly". For
this "quirk" we need to actively delete and re-create the VIF-S interface.
|
|
|
|
|
|
Removing a VLAN (VIF) interface from the CLI always deleted all interfaces the
kernel listed as "upper" in the /sys/class/net folder. This had the drawback
that when deleting a VIF, also the VRF interface was simply deleted - killing
all VRF related services.
|
|
Private VLAN, also known as port isolation, is a technique in computer
networking where a VLAN contains switch ports that are restricted such that
they can only communicate with a given "uplink". The restricted ports are
called "private ports".
Each private VLAN typically contains many private ports, and a single uplink.
The uplink will typically be a port (or link aggregation group) connected to a
router, firewall, server, provider network, or similar central resource.
Q: https://en.wikipedia.org/wiki/Private_VLAN
|
|
dhcp: T3300: add DHCP default route distance
|
|
When a VIF/VLAN interface is placed in admin down state but the lower
interface, serving the vlan, is moved from admin down -> admin up, all its
vlan interfaces will be placed in admin up state, too.
This is bad as a VLAN interface will become admin up even if its specified as
admin down after a reboot.
To reproduce:
set interfaces ethernet eth1 vif 20 disable
set interfaces ethernet eth1 disable
commit
delete interfaces ethernet eth1 disable
commit
Now check the interface state and it returns UP,LOWER_UP
7: eth1.20@eth1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP group default qlen 1000
link/ether 00:50:56:b3:09:07 brd ff:ff:ff:ff:ff:ff
inet6 fe80::250:56ff:feb3:907/64 scope link
valid_lft forever preferred_lft forever
|
|
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.
|
|
|
|
|
|
|
|
If dhcpv6-options is configured without requesting a DHCPv6 address or PD, the
dhcpv6pd variable is assigned an empty dict.
|
|
|
|
|
|
|
|
equipment
According to the consensus, the specific behavior of a VLAN aware bridge should conform
to the behavior of professional equipment. This commit makes a significant change to the
behavior of VLAN aware bridge, and has the following behaviors:
1. Disable `vif 1` configuration
2. When the VLAN aware bridge is enabled, the parent interface is always VLAN 1
3. When `native-vlan` is not configured, the default behavior of the device is `native-vlan 1`
4. The VLAN ids forwarded by the bridge are determined by `vif`
5. It has an `enable-vlan` node to enable VLAN awareness
6. VLAN configuration is allowed only when VLAN aware bridge is activated
|
|
After switching to iproute2 in commit 92f36735 ("ifconfig: T2653: use iproute2
commands for alias, mac and mtu set()/get()" it is necessary to return an empty
string as iproute2 returns None.
|
|
|
|
|
|
|
|
- remove redundant code paths apply_mirror() / apply_mirror_of_monitor()
- have single source available
|
|
is wrong
In e8957b5, we used json to parse the `tc qdisc` filter to determine whether it needs
to be deleted (reduction of exception mechanism), but now we find that the json output
by this command will output unparsed json in some cases,
so We have to go back to the processing of the exception mechanism
|
|
|
|
This reverts commit 9541355433e202fade4692851bffa33ba9d48f44.
|