Follow-up for 77451f654a.
Now, gateway for routes to DNS or NTP servers should be correctly picked,
hence it is not necessary to adjust the gateway address in
dhcp4_request_route_auto() again.
Also, similar for classless static routes, let's always honor
gateway address specified in (non-classless) static routes.
And if not found, refuse to configure the route.
If a DHCP server provides classless static or static routes, then we
should use the gateway for accessing a node in the range specified in
the route. E.g. if a DHCP server provides the default gateway is
192.168.0.1, and classless static route for 8.0.0.0/8 with gateway
192.168.0.2, then we should access 8.8.8.8 through 192.168.0.2 rather
than 192.168.0.1, but should use 192.168.0.1 for 9.9.9.9.
Fixes#28358.
This fixes regression introduced in 5a0c810462 with which all requests
for link properties ended up with EINVAL as we kept hitting
the signature_is_single() assert in sd_bus_get_property().
We create the udevadm -> systemd-udevd symlink during the install phase,
so it doesn't exist in the just compiled tree. This worked in CI since
the symlink is manually created there post-build.
From `ip-link(8)`:
> [no]ignore-df - enables/disables IPv4 DF
suppression on this tunnel. Normally datagrams
that exceed the MTU will be fragmented; the
presence of the DF flag inhibits this, resulting
instead in an ICMP Unreachable (Fragmentation
Required) message. Enabling this attribute causes
the DF flag to be ignored.
If this option is enabled for a GRE/GRETAP tunnel, the `DF` flag in the outer IP header
will not inherit the inner IP header's `DF` flag.
This is useful to transfer packets that exceed the MTU of the underlay
network.
This makes a bond or bridge interface in the degraded-carrier state but has a routable address
handled as routable operational state.
If the carrier is degraded but the address state is routable then the operational state should be
seen as routable and not degraded because that may be the case for bonds if some of the links are down,
but when that happens the bond as whole is still routable.
This also makes operational state to degraded if address state is degraded even if the link state is
degraded-carrier.
Fixes#22713.
Fixes a regression caused by ab3aed4a03.
I thought the commit does not cause any severe regression. However,
drivers for network interfaces may be loaded later. So, we should wait
if no network interface is found.
Fixes#27822.
The setting IPv6AcceptRA= is defaults to yes, hence, even if a .network
file for an interface has static IP address configuration, it may takes
few seconds for the interface being configured state, as NDisc for the
interface needs to be finished. That makes wait-online.service
needlessly slow. Typically, such delay is not necessary for statically
configured networks.
Let's make the required condition slightly relaxed; if a .network file
has static IP address configurations, then let's make the matching
interface enter the 'configured' state soon after the static addresses
configured on the interface.
Note, this does not change the default for IPv6AcceptRA=, hence, NDisc
still runs on interfaces by default. So, addresses, routes, DNS servers,
and so on based on RA will be assigned on interfaces later.
Strictly speaking, this breaks backward compatibility, but the previous
behavior is not clearly documented. If a user requested both static
IPv4 address and IPv6 SLAAC address configured before an interface being
entered to the 'configured' state, then '--ipv6' for wait-online can be
used. So, the behavior change should not cause severe regression.
Closes#27779.
The commit 0f707207b9 enables strong
warning about missing prefix length in Address= setting.
The change was done in v241, and was about 4 years ago.
Let's drop the legacy assumption and make the parser consistent with
'ip address' command.
C.f. #11307.
Closes#26102.
If another bound interface (dummy98) will be removed before that dummy99
is processed by udevd, then removing dummy98 in the next step makes the
target interface (test1) bring down.
Follow-up for 3e2f7c46da.
The outstanding kernel panic should be already fixed in recent enough
kernels by [0]. To make the test safe to run anywhere, let's implement
a simple kernel version check and run the test only if we're running
with at least kernel 6.x. The patch might be in some 5.x kernels as
well, but let's be on the safe side and use 6.x as a baseline here
(which is currently the case for Arch and Fedora Rawhide anyway).
[0] https://lore.kernel.org/netdev/7b3fd03e1a46047e5ffe2a389fe74501f0a93206.1656519221.git.sd@queasysnail.net/T/#u
(s) is just ugly with a vibe of DOS. In most cases just using the normal plural
form is more natural and gramatically correct.
There are some log_debug() statements left, and texts in foreign licenses or
headers. Those are not touched on purpose.
That's not necessary. Moreover, if the socket units are stopped in
`setUpModule()`, then there exists a short timespan that we cannot call
`udevadm control`, as the control socket may not be opened yet.
If we run whole tests, then the first test is
NetworkctlTests.test_altname, and it calls `udevadm control` in `setUp()`.
Hence, the test may fail.
Fixes https://github.com/systemd/systemd-centos-ci/pull/512#issuecomment-1191591008.
Several DHCP client tests change the system timezone.
Let's save the current timezone at the beginning, and restore it with
the saved value at the end.
- introduce several helper functions
- do not list unit files, but remove the runtime unit directory in
tearDown().
- do not list used interfaces, but remove all interfaces previously not
exists in tearDown().
- save routes and routing policy rules before running tests, and flush
unnecessary routes and rules in each tearDown() calls.
- drop many time.sleep() calls.
- call tearDown() after each sub tests.
- shorten code.
- several coding style fixes.
- etc, etc...
Hopefully, this improves performance of the test.
`global` is needed only when assigning a new value to the global
variable; it's not necessary when modifying a mutable object (in our
case we just append items to the global list).
Otherwise, this easily trigger another exception:
```
======================================================================
ERROR: test_erspan_tunnel_v0 (__main__.NetworkdNetDevTests)
----------------------------------------------------------------------
Traceback (most recent call last):
File "./test/test-network/systemd-networkd-tests.py", line 686, in wait_online
check_output(*args, env=env)
File "./test/test-network/systemd-networkd-tests.py", line 65, in check_output
return subprocess.check_output(command, universal_newlines=True, **kwargs).rstrip()
File "/usr/lib64/python3.6/subprocess.py", line 356, in check_output
**kwargs).stdout
File "/usr/lib64/python3.6/subprocess.py", line 438, in run
output=stdout, stderr=stderr)
subprocess.CalledProcessError: Command '['/usr/lib/systemd/systemd-networkd-wait-online', '--timeout=20s', '--interface=erspan99:routable', '--interface=erspan98:routable', '--interface=dummy98:degraded']' returned non-zero exit status 1.
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "./test/test-network/systemd-networkd-tests.py", line 1808, in test_erspan_tunnel_v0
self.wait_online(['erspan99:routable', 'erspan98:routable', 'dummy98:degraded'])
File "./test/test-network/systemd-networkd-tests.py", line 689, in wait_online
output = check_output(*networkctl_cmd, '-n', '0', 'status', link.split(':')[0], env=env)
File "./test/test-network/systemd-networkd-tests.py", line 65, in check_output
return subprocess.check_output(command, universal_newlines=True, **kwargs).rstrip()
File "/usr/lib64/python3.6/subprocess.py", line 356, in check_output
**kwargs).stdout
File "/usr/lib64/python3.6/subprocess.py", line 438, in run
output=stdout, stderr=stderr)
subprocess.CalledProcessError: Command '['/usr/bin/networkctl', '-n', '0', 'status', 'erspan99']' returned non-zero exit status 1.
```
Not sure when the issue was fixed.
- kernel-3.10 on CentOS 7 has the issue,
- kernel-4.18 on CentOS 8 works fine.
Note, the workaround dropped by the commit is not incomplete:
with an old kernel which has the issue, all non-prefix routes are
configured on the specified route table, but the prefix route is
configured on the main table. That should not work for most cases,
hence, the workaround is mostly meaningless.
There is no reason networkd refuses that. Especially, when multiple
downstream interfaces are connected to the same network, it is natural to
assign the same subnet prefix to them.
Prompted by #22571.
If we assign our own test runner, passing arguments stops working
as unittest won't instantiate its own test runner after it parses
the arguments from sys.argv.
Consequence is that the tests will write to stderr now instead of
stdout since it doesn't seem possible to configure the stream that
unittest.main() will instantiate its test runner with so it'll
default to sys.stderr.
Add the "Isolated" parameter in the *.network file, e.g.,
[Bridge]
Isolated=true|false
When the Isolated parameter is true, traffic coming out of this port
will only be forward to other ports whose Isolated parameter is false.
When Isolated is not specified, the port uses the kernel default
setting (false).
The "Isolated" parameter was introduced in Linux 4.19.
See man bridge(8) for more details.
But even though the kernel and bridge/iproute2 recognize the "Isolated"
parameter, systemd-networkd did not have a way to set it.
The warning is correct, since we don't inherit the necessary
unittest.TestCase class, but that's on purpose, since the Utilities
class is not supposed to be instantiated on its own, but should
complement other classes' definitions which do inherit from the
unittest.TestCase class.
It currently works because `\(` and `\)` are not valid escape sequences,
so they're not treated differently. Using raw strings (or double
backslashes) is a more correct solution.
Multiline comments are converted to docstrings only when they're the
first statement in a function/method. Even though they're still a no-op
otherwise, let's use "true" comments to make pylint happy.
If wait_operstate() is called super quickly after ip command, then the
up/down state may not be changed and propagated to networkd, and
wait_operstate() mistakenly pass with the previous state.
To avoid such situation, wait for a while to make networkd actually
detect the interface brought up/down.
This fixes the following race:
1. when a dummy interface is created, it is initially down state,
2. hence, wait_operstate() may pass before the link is activated,
3. and the ip command bring up the interface before the activation,
4. and networkd activates, that is, brings down the interface,
5. thus, next wait_operstate() timedout, as it waits for the interface up.
To fix the race, let's wait the link is activated, before enter the loop
of wait_operstate().
Fixes#22267.
Strictly speaking, this breaks the backward compatibility, but I guess
in most cases people already sets Scope=link for such routes.
This behavior matches with how 'ip route' command adds such route by
default.
Prompted by https://twitter.com/jplitza/status/1480500562391179270.
People often assigns the MAC address of the enslaved interface to e.g.
bridge interface. So, the local assignment bit should not be adjusted.
Fixes#21649.
