| Age | Commit message (Collapse) | Author |
|---|
|
Adversarial review findings on the previous head:
* get_process_cmdline() could re-resolve a NEW PID after the
'ip vrf pids' membership assertion had already run against the old
one, so the two assertions could validate different processes. The
helper now returns the (pid, cmdline) pair it actually validated and
the VRF checks moved after it, targeting the returned PID.
* The one-shot retry left a residual double-race window. The helper
now polls PID discovery and the /proc read together under the same
PROCESS_WAIT_TIMEOUT deadline.
🤖 Generated by [robots](https://vyos.io)
|
|
The re-resolve path read /proc/<pid>/cmdline without defaultonfailure,
so a repeated PID race raised an unhandled exception (test ERROR)
instead of failing with a clear assertion message.
🤖 Generated by [robots](https://vyos.io)
|
|
Interface smoketests fail intermittently in CI on DHCP-related
assertions (test_interfaces_cli job pass rate 42-55% over the last 30
workflow runs). Three timing hazards in the interface test base class:
* process_named_running() was polled with a 10 second window at all
dhclient/dhcp6c call sites - too short on a loaded runner. Raise to a
shared PROCESS_WAIT_TIMEOUT of 60 seconds; the poll returns as soon
as the process appears, so this only delays the failure path.
* /proc/<pid>/cmdline was read unguarded after PID discovery. dhclient
re-executes itself while daemonizing, so the discovered PID can be
gone by the time /proc is read - read_file() then raises and the test
errors out instead of failing cleanly. New get_process_cmdline()
helper re-resolves the PID once when the read fails.
* tearDown() asserted daemon absence immediately after the config was
removed, reporting daemons still in their shutdown path as leaks.
Use the existing wait_for_result() shim helper with the same 60s
bound to grant a grace period before declaring a leak; the poll
returns on first observation of a clean state, so passing runs pay
no extra wall clock.
🤖 Generated by [robots](https://vyos.io)
|
|
Replace two-word "can not" / "Can not" with "cannot" across comments,
ConfigError messages, CLI help text, and op-mode output.
Standard SNMP MIB files under mibs/ are left unchanged.
|
|
Fix typos and mistakes
No functional changes
|
|
When removing the mirror CLI node to stop mirroring or redirecting traffic
to another interface, the egress configuration was not cleared. This caused
traffic to continue being sent out the SPAN port even after the node was
removed.
Fix by properly clearing all tc(8) qdiscs.
Update smoketests to verify nothing remains after mirror deletion.
|
|
Some drivers are limited to an MTU of 1500 byte - e.g. when VyOS runs on
PROXMOX with default bridge settings. We use lower values for the testcase
test_vif_8021q_mtu_limits which will fit for almost every NIC, but keep the
testcase logic intact.
|
|
While working on task T7664 (FRR 10.4 upgrade), I identified the need for
additional validation and safeguards around the FRR management daemon. The
most appropriate place for this logic is in the setUp() and tearDown() methods
of the smoketest base class, VyOSUnitTestSHIM.
However, during implementation, it became apparent that test cases do not
consistently invoke the base class's setup and teardown methods. This
inconsistency complicates the process of capturing the FRR mgmtd PID at the
start of a test and verifying that it remains unchanged by the end - a key step
in detecting crashes or unexpected terminations (e.g., SIGSEGV) of the FRR
management daemon during tests.
|
|
|
|
|
|
The current template has an overly optimistic logic for sla-len calculation,
relying on an assumption that a server always provides a prefix with the exact
length requested by the client. This is incorrect. According to RFC8415 and
RFC7084, the prefix length in a request is only a hint for the server, which
may decide to provide a different prefix size.
The big issue here is that wide-dhcpv6-client uses the sla-len value from the
configuration regardless of the received prefix size. This seems to be the
known issue.
The good news is that we have already inherited a patch from Debian to mitigate
the issue [1]. It accomplishes exactly what we are doing in the configuration
template, but with the advantage of using the actual prefix length from the
received prefix, rather than the one configured in the configuration file.
If we simply remove sla-len from the template, everything appears to function
normally.
Before - server sending /56
set interfaces ethernet eth1 address 'dhcpv6'
set interfaces ethernet eth1 dhcpv6-options pd 1 interface eth2 address '1'
set interfaces ethernet eth1 dhcpv6-options pd 1 interface eth2 sla-id '0'
set interfaces ethernet eth1 dhcpv6-options pd 1 length '60'
Resulted in:
vyos@vyos# run show interfaces
Codes: S - State, L - Link, u - Up, D - Down, A - Admin Down
Interface IP Address MAC VRF MTU S/L Description
----------- ------------------- ----------------- ------- ----- ----- -------------
eth1 fc00:0:0:1::200/128 0c:67:94:67:00:01 default 1500 u/u
eth2 fc00:0:2:ff00::1/60 0c:67:94:67:00:02 default 1500 u/u
Whereas IPv6 PD should always use a /64 prefix on the interface we assign a
dynamic DHCPv6 prefix to.
After the fix:
vyos@vyos:~$ show interfaces
Codes: S - State, L - Link, u - Up, D - Down, A - Admin Down
Interface IP Address MAC VRF MTU S/L Description
----------- ------------------- ----------------- ------- ----- ----- -------------
eth1 fc00:0:0:1::200/128 0c:67:94:67:00:01 default 1500 u/u
eth2 fc00:0:2:ff00::1/64 0c:67:94:67:00:02 default 1500 u/u
If a DHCPv6 server (e.g. VyOS itself) even provides multiple PD prefixes, the
right one is choosen. Assume the following server configuration:
set service dhcpv6-server shared-network-name dhcpv6-1 subnet fc00:0:0:1::/64 address-range start fc00:0:0:1::100 stop 'fc00:0:0:1::200'
set service dhcpv6-server shared-network-name dhcpv6-1 subnet fc00:0:0:1::/64 lease-time default '120'
set service dhcpv6-server shared-network-name dhcpv6-1 subnet fc00:0:0:1::/64 prefix-delegation start fc00:0:2:: prefix-length '56'
set service dhcpv6-server shared-network-name dhcpv6-1 subnet fc00:0:0:1::/64 prefix-delegation start fc00:0:2:: stop 'fc00:0:2:ff00::'
set service dhcpv6-server shared-network-name dhcpv6-1 subnet fc00:0:0:1::/64 prefix-delegation start fc00:0:3:: prefix-length '60'
set service dhcpv6-server shared-network-name dhcpv6-1 subnet fc00:0:0:1::/64 prefix-delegation start fc00:0:3:: stop 'fc00:0:3:ff00::'
If a /56 is requested ("dhcpv6-options pd 1 length 56") - we will get a per
interface /64 from the original /56 prefix-delegation from the DHCPv6 server.
If the user suddently requests a /60 ("dhcpv6-options pd 1 length 60") the
delegated prefix will be from the above fc00:0:3:: pool.
1: https://salsa.debian.org/debian/wide-dhcpv6/-/blob/debian/20080615-23/debian/patches/0021-Make-sla-len-config-optional.patch
|
|
The legal team says years are not necessary so we can go ahead with it, since
it will simplify backporting.
Automatically removed using: git ls-files | grep -v libvyosconfig | xargs sed -i -E \
's/^# Copyright (19|20)[0-9]{2}(-[0-9]{4})? VyOS maintainers.*/# Copyright VyOS maintainers and contributors <maintainers@vyos.io>/g'
In addition we will error-out during "make" if someone re-adds a legacy
copyright notice
|
|
Add common IPv6 CLI option (use ethernet as example):
set interfaces ethernet eth0 ipv6 address interface-identifier
Co-authored-by: Christian Breunig <christian@breunig.cc>
|
|
|
|
|
|
T7050 revealed an issue in the CI system where VyOS CLI was rendering a
configuration for wide-dhcpv6-client that was not working due to a missing
patch of the wide-dhcpv6-client source code.
This can be prevented by checking the daemon logs after startup for any config
linguistic (Yacc/Bison) issues.
|
|
|
|
Previously the DHCPv6 client was restarted on any change to the interface,
including changes only to the interface description. Re-use pattern from IPv4
DHCP to only restart the DHCP client if necessary.
|
|
Moving an interface in, out or between VRFs will not re-install the received
default route. This is because the dhclient binary is not restarted in the new
VRF. Dhclient itself will report an error like: "receive_packet failed on
eth0.10: Network is down".
Take the return value of vyos.ifconfig.Interface().set_vrf() into account to
forcefully restart the DHCP client process and optain a proper lease.
|
|
* smoketest: T6747: call wait after commit() only for FRR related tests
Commit 702a60a8de28 ("smoketest: T6746: wait after commit() until frr-reload
is no longer running") added a guard timeout for every commit executed via CLI
smoketests. This commit changes the bahavior to only add the guard timeout
for FRR related testscases.
This improves the overall smoketest time.
* configd: T6747: use one long-lived instance of FRRender
Previously there was one FRRender() instance per config session. This resulted
in re-rendering the FRR configuration every time a new config session was
created.
Example:
vyos@vyos:~$ configure
vyos@vyos# set interfaces dummy dum0 description foo
vyos@vyos# commit
vyos@vyos# exit
vyos@vyos:~$ configure
vyos@vyos# set interfaces dummy dum0 description bar
vyos@vyos# commit
vyos@vyos# exit
In the past this caused a re-render of the FRR configuration as the delta check
added in commit ec80c75d6776 ("frrender: T6746: only re-render FRR config if
config_dict did change") evaluated to false, as it operated on a new instance
of the FRRender class.
With this change there is no FRR re-render, as there is nothing to update
in FRR.
|
|
|
|
Instead of having EAPoL (Extensible Authentication Protocol over Local Area
Network) support only available for ethernet interfaces, move this to common
ground at vyos.ifconfig.interface making it available for all sorts of
interfaces by simply including the XML portion
#include <include/interface/eapol.xml.i>
|
|
interface: T6592: remove interface from conntrack ct_iface_map on deletion
|
|
deletion
Now that interfaces are deleted from ct_iface_map during deletion it's time to
also add a smoketest ensuring there is no entry in the ct_iface_map once an
interface was deleted from the CLI.
|
|
Only some (e.g. ethernet or wireguard) interfaces validate if the supplied VRF
actually exists. If this is not validated, one can pass an invalid VRF to the
system which generates an OSError exception.
To reproduce
set interfaces vxlan vxlan1 vni 1000
set interfaces vxlan vxlan1 remote 1.2.3.4
set interfaces vxlan vxlan1 vrf smoketest
results in
OSError: [Errno 255] failed to run command: ip link set dev vxlan1 master smoketest_mgmt
This commit adds the missing verify_vrf() call to the missing interface types
and an appropriate smoketest for all interfaces supporting VRF assignment.
|
|
To reproduce:
set vrf name mgmt table '150'
set vrf name no-mgmt table '151'
set interfaces ethernet eth2 vrf 'mgmt'
commit
set interfaces ethernet eth2 vrf no-mgmt
commit
This resulted in an error while interacting with nftables:
[Errno 1] failed to run command: nft add element inet vrf_zones ct_iface_map { "eth2" : 151 }
The reason is that the old mapping entry still exists and was not removed.
This commit adds a new utility function get_vrf_tableid() and compares the
current and new VRF table IDs assigned to an interface. If the IDs do not
match, the nftables ct_iface_map entry is removed before the new entry is added.
|
|
There are cloud environments available where the maximum supported ethernet
MTU is e.g. 1450 bytes, thus we clamp this to the adapters maximum MTU value
or 1500 bytes - whatever is lower.
|
|
found using "git ls-files *.py | xargs pylint | grep W0611"
|
|
e.g. Linux Kernel only supports 255 and not 256 characters for the ifalias field.
|
|
Commit ad9bdfc24 ("T5195: add timeout argument to process_named_running()")
added a 2*10 seconds penalty for every interface test (dhcp and dhcpv6).
This leads to long runs of "make test" after an ISO build.
There is no need to wait 10 seconds for a test that checks for a process
not running. The timeout is there to give the process some time to startup.
|
|
Smoketests heavily rely on process_named_running() so in order to "relax"
system constraints during a test we will add a timeout of 10 seconds for
every testcase provided by base_interfaces_test.py
|
|
Found an odd behavior on Linux and the VyOS CLI implementation. If adding VXLAN
interfaces using iproute2 the MTU differs depending on the creation syntax:
ip -4 link add vxlan100 type vxlan dstport 4789 external df unset tos inherit \
ttl 16 nolearning vnifilter local 172.16.33.201
ip -4 link add vxlan200 type vxlan id 200 dstport 4789 local 172.16.33.201 dev eth0
ip -6 link add vxlan300 type vxlan id 300 dstport 4789 local 2001:db8:1::1 dev eth0
132: vxlan300: <BROADCAST,MULTICAST> mtu 1430 qdisc noop state DOWN group default qlen 1000
link/ether 4e:fb:e3:f5:d9:59 brd ff:ff:ff:ff:ff:ff
133: vxlan200: <BROADCAST,MULTICAST> mtu 1450 qdisc noop state DOWN group default qlen 1000
link/ether 0e:4e:f4:76:59:3f brd ff:ff:ff:ff:ff:ff
134: vxlan100: <BROADCAST,MULTICAST> mtu 1500 qdisc noop state DOWN group default qlen 1000
link/ether ba:b6:b7:0c:b1:37 brd ff:ff:ff:ff:ff:ff
VyOS always sets a default MTU of 1450 bytes which is correct for IPv4 p2p links
or multicast, but invalid for IPv6 p2p. Also this will break EVPN deployments
as ethernet bridges with MTU < 1500 bytes are less fun.
Increase default MTU to 1500 bytes. Migrate old configurations to use 1450
bytes if not specified otherwise on the CLI.
|
|
Example:
set interfaces ethernet eth0 dhcp-options user-class VyOS
or
set interfaces ethernet eth0 dhcp-options user-class 56:79:4f:53
|
|
The string data type specifies either an NVT ASCII string enclosed in double
quotes, or a series of octets specified in hexadecimal, separated by colons.
For example:
set interfaces ethernet eth0 dhcp-options client-id CLIENT-FOO
or
set interfaces ethernet eth0 dhcp-options client-id 43:4c:49:45:54:2d:46:4f:4f
As of now there was no input validation performed.
|
|
Migrate CLI configuration retrival to common get_config_dict(). In addition
add new functionality to VyOS that is PIM related and already available in FRR.
|
|
- Migrate IPv4 source-validation to nftables
- Interface source-validation value takes priority, fallback to global value
|
|
|
|
setting
|
|
|
|
Start IPv4/IPv6 DHCP clients on an interface bound to a given VRF. Verify that
the client process runs in the VRF context.
|
|
|
|
|
|
|
|
* T5195: move run, cmd, call, rc_cmd helper to vyos.utils.process
* T5195: use read_file and write_file implementation from vyos.utils.file
Changed code automatically using:
find . -type f -not -path '*/\.*' -exec sed -i 's/^from vyos.util import read_file$/from vyos.utils.file import read_file/g' {} +
find . -type f -not -path '*/\.*' -exec sed -i 's/^from vyos.util import write_file$/from vyos.utils.file import write_file/g' {} +
* T5195: move chmod* helpers to vyos.utils.permission
* T5195: use colon_separated_to_dict from vyos.utils.dict
* T5195: move is_systemd_service_* to vyos.utils.process
* T5195: fix boot issues with missing imports
* T5195: move dict_search_* helpers to vyos.utils.dict
* T5195: move network helpers to vyos.utils.network
* T5195: move commit_* helpers to vyos.utils.commit
* T5195: move user I/O helpers to vyos.utils.io
|
|
|
|
|
|
Current state of vyos-1x smoketests have hardcoded features to test. The
feature support is inside the base class BasicInterfaceTest
class BasicInterfaceTest:
class TestCase(VyOSUnitTestSHIM.TestCase):
_test_dhcp = False
_test_ip = False
_test_mtu = False
_test_vlan = False
_test_qinq = False
_test_ipv6 = False
_test_ipv6_pd = False
_test_ipv6_dhcpc6 = False
_test_mirror = False
All derived classes need to enable the tests that are supported on this
interface type. Adding new feature to a given interface (like vif support in
T5237) require manually enabling those tests for the given interface.
It would make much more sense, if we can query the config backend for supported
interface options - or in other words - is there a CLI node available that
corresponds to set interfaces ethernet <name> vif - if that's the case,
_test_vlan = True.
|
|
|
|
Replace links to the phabricator site from https://phabricator.vyos.net to
https://vyos.dev
|
|
Directed broadcast is described in rfc1812#section-5.3.5.2 and rfc2644.
By default Linux kernel doesn't forward directed broadcast
packets unless both of `/proc/sys/net/ipv4/conf/all/bc_forwarding`
and `/proc/sys/net/ipv4/conf/$iface/bc_forwarding` are set to 1.
|
