Age | Commit message (Collapse) | Author |
|
The initial implementation in commit 9fb9e5cade ("ethernet: T3171: add CLI
option to enable RPS (Receive Packet Steering)" only changed the CPU affinity
for RX queue 0.
This commit takes all RX queues into account.
(cherry picked from commit 13645bc2cfd31f1525078469f23e89491987e0ea)
|
|
T4630: disallow same source-interface for macsec and pseudo-ethernet
|
|
In the OpenVPN site-to-site config we can use IPv6 peers
without IPv4 configurations but "verify()" checks also local and
remote IPv4 addresses that in this case will be empty lists
For example:
set interfaces openvpn vtun2 local-address 2001:db8::1
set interfaces openvpn vtun2 remote-address 2001:db8::2
Check in the commit (v4loAddr == v4remAddr) <= both empty lists
commit
DEBUG: [] == [] or ['2001:db8::2'] == []
So we should also check v4loAddr, v4remAddr, v6loAddr, v6remAddr
are not empty
|
|
A macsec interface requires a dedicated source interface, it can not be
shared with another macsec or a pseudo-ethernet interface.
set interfaces macsec macsec10 address '192.168.2.1/30'
set interfaces macsec macsec10 security cipher 'gcm-aes-256'
set interfaces macsec macsec10 security encrypt
set interfaces macsec macsec10 security mka cak '232e44b7fda6f8e2d88a07bf78a7aff4232e44b7fda6f8e2d88a07bf78a7aff4'
set interfaces macsec macsec10 security mka ckn '09924585a6f3010208cf5222ef24c821405b0e34f4b4f63b1f0ced474b9bb6e6'
set interfaces macsec macsec10 source-interface 'eth1'
commit
set interfaces pseudo-ethernet peth0 source-interface eth1
commit
Reuslts in
FileNotFoundError: [Errno 2] failed to run command: ip link add peth0 link eth1 type macvlan mode private
returned:
exit code: 2
noteworthy:
cmd 'ip link add peth0 link eth1 type macvlan mode private'
returned (out):
returned (err):
RTNETLINK answers: Device or resource busy
[[interfaces pseudo-ethernet peth0]] failed
Commit failed
(cherry picked from commit eb4a7ee3afc0765671ce0fa379ab5e3518e9e49e)
|
|
Fixes several bugs around bonding member interface states not matching
the committed configuration, including:
- Disabled removed interfaces coming back up
- Newly added disabled interfaces not staying down
- Newly added interfaces not showing up in the bond
|
|
Commit 31169fa8a763e ("vyos.ifconfig: T3619: only set offloading options if
supported by NIC") added the new implementation which handles NIC offloading.
Unfortunately every single implementation was copied from "gro" which resulted
in a change to gro for each offloading option - thus options like lro, sg, tso
had no effect at all.
It all comes down to copy/paste errors ... one way or another.
(cherry picked from commit b01f27b3bb3f4cbc6096011856d83009d0440313)
|
|
bridge: T4632: vlan aware bridge lacks CPU forwarding
|
|
The VLAN aware bridge was forwarding traffic between member ports, but traffic
destined torwards the CPU was dropped. This resulted in a gateway not reachable
or DHCP leases that could not be handed out.
Tested via:
VyOS
set interfaces bridge br0 enable-vlan
set interfaces bridge br0 member interface eth1 allowed-vlan '10'
set interfaces bridge br0 member interface eth1 allowed-vlan '20'
set interfaces bridge br0 member interface eth1 allowed-vlan '30'
set interfaces bridge br0 member interface eth1 allowed-vlan '40'
set interfaces bridge br0 member interface eth1 native-vlan '40'
set interfaces bridge br0 member interface eth2 allowed-vlan '30'
set interfaces bridge br0 member interface eth2 allowed-vlan '20'
set interfaces bridge br0 member interface eth2 allowed-vlan '10'
set interfaces bridge br0 member interface eth2 allowed-vlan '40'
set interfaces bridge br0 vif 10 address '10.0.10.1/24'
set interfaces bridge br0 vif 20 address '10.0.20.1/24'
set interfaces bridge br0 vif 30 address '10.0.30.1/24'
set interfaces bridge br0 vif 40 address '10.0.40.1/24'
Arista vEOS
vlan 10,20,30,40
interface Ethernet1
switchport trunk allowed vlan 10,20,30,40
interface Vlan10
ip address 10.0.10.2/24
interface Vlan20
ip address 10.0.20.2/24
interface Vlan30
ip address 10.0.30.2/24
interface Vlan40
ip address 10.0.40.2/24
interface Ethernet1
switchport trunk allowed vlan 10,20,30,40
switchport mode trunk
spanning-tree portfast
Cisco vIOS
interface GigabitEthernet0/0
ip address 10.0.40.3 255.255.255.0
duplex auto
speed auto
media-type rj45
!
interface GigabitEthernet0/0.10
encapsulation dot1Q 10
ip address 10.0.10.3 255.255.255.0
!
interface GigabitEthernet0/0.20
encapsulation dot1Q 20
ip address 10.0.20.3 255.255.255.0
!
interface GigabitEthernet0/0.30
encapsulation dot1Q 30
ip address 10.0.30.3 255.255.255.0
!
(cherry picked from commit f60d0e1ce029925b843f635b36154c90049b9577)
|
|
interface dict on error
(cherry picked from commit 475fbb785dca76868715827833dc44115635c4a6)
|
|
mtu: T4572: Add DHCP-option MTU to get values from DHCP-server
|
|
Ability to get MTU from DHCP-server and don't touch it per
any interface change if interface 'dhcp-options mtu' is
configured
(cherry picked from commit 29b0ee30bf2622a40ca3d17e3f6b9e94e5b62072)
|
|
We should not print the entire dictionary - we only need the bridge interface
name:
Bug:
Cannot assign address to interface "eth1" as it is a member of bridge
"{'br0': {'allowed_vlan': ['5-50', '101'], 'native_vlan': '101'}}"!
Fixed:
Cannot assign address to interface "eth1" as it is a member of bridge
"br0"!
(cherry picked from commit 71f2f2d789bd3452ca7b5306f39df5f6537bfef0)
|
|
... otherwise functionality like bridge VLANs will loose configuration
on membe rinterface update (e.g. description)
(cherry picked from commit a295ec1fc4cb9cdbb0bd5efd7c8e72060ce036f5)
|
|
(cherry picked from commit 54227591a0eb3c7aa8c896c6ec8b1826ce070ddf)
|
|
Interface() base class already takes care about VLAN creation/removal of
newly added or no longer required interfaces. No need to code this logic again.
(cherry picked from commit 41477cc85208507be55f8db4e412ad78eae764eb)
|
|
(cherry picked from commit 99ff0ac7d29a65466a0fe4a12ef1a42cce90ece0)
|
|
equuleus: Bond and Bridge interface fixes + new smoketests
|
|
(cherry picked from commit 81e0f4a8dece85da7169ba05448e870206aaf57b)
|
|
When removing a VRF from an ethernet interface and adding the interface to a
bond in the same commit led to an OSError: [Errno 16] Device or resource busy!
(cherry picked from commit 3592f56a8deb6c44dcdd7a44ef54fc2c39eb1a3b)
|
|
This extends commit 39157912 ("vyos.configdict(): T4228: is_member() must use
the "real" hardware interface") and returns the config dict of the used member
interfaces.
(cherry picked from commit 5b4f76429989a6ab8ca64aeed5a1fae09fe7c6ca)
|
|
It makes no sense to enslave an interface to a bond or a bridge device if it is
bound to a given VRF. If VRFs should be used - the encapuslating/master
interface should be part of the VRF.
Error out if the member interface is part of a VRF.
(cherry picked from commit 87d2dff241d9ab4de9f3a2c7fbf9852934557aef)
|
|
Commit 39157912 ("vyos.configdict(): T4228: is_member() must use the "real"
hardware interface") added a bugfix on calling is_member() to retrieve the real
physical information about an interface. It did not include a code path to also
split up VLAN interfaces.
This has been fixed.
(cherry picked from commit fdeae251431cb747e8f60d96269b4365b7401807)
|
|
Linux Kernel supports to specify the MII link monitoring frequency in
milliseconds. This determines how often the link state of each slave is
inspected for link failures. A value of zero disables MII link monitoring.
A value of 100 is a good starting point.
The default value is 100.
set interfaces bonding bond0 mii-mon-interval <n>
(cherry picked from commit 4315c8fa5bb090e2b7edd6bda205041623e2511d)
|
|
When is_member() is inspecting the bridge/Bond member interfaces it must work
with the real interface (e.g. eth1) under the "ethernet" node and not work on
the "member interface eth1" CLI tree, that makes no sense at all.
(cherry picked from commit 3915791216998a18bf6831450df68ee199e2e4f8)
|
|
Fix for IPv6 netflow_plugin name
When we use IPv6 uacctd.conf doesnt expect coluns in the plugin
name. Replace coluns to dash. Place IPv6 address into [] brackets
|
|
(cherry picked from commit cabfd006bed9cd2d1512cb313616a8e97fe29b9e)
|
|
The code path for changing the interval is never executed.
(cherry picked from commit daddb0ad7812843f64a7ae0bf4b5e15db7b1758e)
|
|
Add action 'reset' (op-mode) for HTTP-API
http://localhost/reset
curl --unix-socket /run/api.sock -X POST -Fkey=mykey \
-Fdata='{"op": "reset", "path": ["ip", "bgp", "192.0.2.14"]}' \
http://localhost/reset
|
|
(cherry-picked from commit 63fee7ab0c082f3c614bb140de1e7482399dff2b)
|
|
(cherry picked from commit 714346e2ee0c61a08a9d17fdb962f2fbea9f73c9)
|
|
Add support for the configtree diff algorithm. A new function
ConfigDiff().is_node_changed(path) -> bool
is added to recursively detect changes in the tree below the node at
path; existing functions take the keyword argument 'recursive: bool' to
apply the algorithm in place of the existing, non-recursive, comparison.
(cherry picked from commit e5d04b20be0ef270d20f1d5ac9203b3a03649135)
|
|
The DiffTree class maintains both the 'sub'(-tract) configtree,
containing all paths in the LHS of the comparison that are not in the
RHS, and the 'delete' configtree: the delete tree is the minimal subtree
containing only the first node of a path not present in the RHS. It is
the delete tree that is needed to produce 'delete' commands for config
mode, whereas the 'sub' tree contains full information, needed for
recursively detecting changes to a node.
(cherry picked from commit 193cbd15ba39a41614c63b997e6a62254589158a)
|
|
(cherry picked from commit 4625fd41f99ddf77c104a657cd90a1ddf5449dd8)
|
|
The return value of diff_tree is now a single config_tree, with initial
children of names: ["add", "delete", "inter"] containing the config
sub-trees of added paths; deleted paths; and intersection, respectively.
The simplifies dumping to json, and checking existence of paths, hence,
of node changes.
(cherry picked from commit 3d28528ff84b4e874faf80028709bd08b2956933)
|
|
(cherry picked from commit ff7e2cd622cf3679cd9265b2cb766395a1830f50)
|
|
(cherry picked from commit 2ee94418ce24429dbf6a52c2a327ed08a1935958)
|
|
T4324, T4338, T4339 WWAN interface bugfixes
|
|
T4331: IPv6 link local addresses are not configured when an interface is in a VRF (equuleus)
|
|
(cherry picked from commit 957f73ed8c2c22afd5e56adc36b4d032b3f1a5f1)
|
|
If we have link-local static address and vrf, for example:
set interfaces ethernet eth2 address 'fe80::5200:ff:fe55:222/64'
set interfaces ethernet eth2 vrf 'foo'
This IPv6 address was assigned before vrf, as result after
attaching the intreface to vrf we lose this static linklocal
address
DEBUG/IFCONFIG cmd 'ip addr add fe80::5200:ff:fe55:222/64 dev eth2'
DEBUG/IFCONFIG cmd 'ip link set dev eth2 master foo'
DEBUG/IFCONFIG cmd 'ip addr add fe80::5208:ff:fe13:2/64 dev eth2'
This commit fixes this, the address is assigned after vrf assign
(cherry picked from commit d6e22b28887c7a3f7d2f8b955c2e90bcadaeeeba)
|
|
(cherry picked from commit 15c94a8706622927850eba8c22fcff2df32978b4)
|
|
This extends the fix from 53e20097 ("vyos.ifconfig: T4330: bugfix changing MTU
when IPv6 is disabled") by ordering the execution in a way the Kernel does not
complain.
(cherry picked from commit 601ab19fd8c81a998b3c966cc83b85ed60ac5ae0)
|
|
Commit f8b3d8999c ("ipv6: T4319: do not configure IPv6 related settings if it's
disabled") moved the MTU configuration part under the code path which is only
run if IPv6 is enabled on the system.
This prevented MTU changes on IPv6 disabled systems.
(cherry picked from commit 53e20097d227ebf4bdb4dc6c85427ec9c5ec3982)
|
|
(cherry picked from commit 60f009defadb9d36bf84def1e839cb11a0b3d619)
|
|
(cherry picked from commit df0fbfeedce0f163e9d10be21d58ad4dc797a28a)
|
|
|
|
(cherry picked from commit 9d3acc2b55f2d1c563f1941e59c98c159211dc58)
|
|
(cherry picked from commit 364009e4317fb5c6732635726b511613aa2ed519)
|
|
(cherry picked from commit f8b3d8999cbea988ce8e7d303957558308ddc1bc)
|
|
(cherry picked from commit 52cb6185a4a51ffa92f10e0ded55a943bc21bc60)
|