Currently src/platform depends on libnm-core. libnm-core is large
optimally we have a better separation between our code. That means
libnm-core does not depend on platform and vice versa.
However, nm-platform re-uses some enums from libnm-core for internal code.
To avoid that dependency, add _NMSettingWiredWakeOnLan as a duplicate to
nm-base/nm-base.h. nm-base can both be used by libnm-core, nm-platform
and src/platform.
The only problem is that NMSettingWiredWakeOnLan is also part of public
API of libnm. It means, we must duplicate the enum. But with several
static assertions in unit tests I think that is not a problem to do.
With LTO and gcc-10.2.1-9.fc33.s390x we get:
src/platform/nm-platform.c:3325:1: error: link_duplex may be used uninitialized in this function [-Werror=maybe-uninitialized]
3325 | NM_UTILS_LOOKUP_STR_DEFINE(nm_platform_link_duplex_type_to_string,
| ^
src/devices/nm-device-ethernet.c:899: note: link_duplex was declared here
899 | NMPlatformLinkDuplexType link_duplex;
|
Currently libnm headers include <linux/if_{ether,infiniband,vlan}.h>.
These are public headers, that means we drag in the linux header to all
users of <NetworkManager.h>.
Often the linux headers work badly together with certain headers from libc.
Depending on the libc version, you have to order linux headers in the right
order with respect to libc headers.
We should do better about libnm headers. As a first step, assume that
the linux headers don't get included by libnm, and explicitly include
them where they are needed.
NetworkManager is now able to configure veth interfaces throught the
NMSettingVeth. Veth interfaces only have "peer" property.
In order to support Veth interfaces in NetworkManager the design need
to pass the following requirements:
* Veth setting only has "peer" attribute.
* Ethernet profiles must be applicable to Veth interfaces.
* When creating a veth interface, the peer will be managed by
NetworkManager but will not have a profile.
* Veth connection can reapply only if the peer has not been modified.
* In order to modify the veth peer, NetworkManager must deactivate the
connection and create a new one with peer modified.
In general, it should support the basis of veth interfaces but without
breaking any existing feature or use case. The users that are using veth
interfaces as ethernet should not notice anything changed unless they
specified the veth peer setting.
Creating a Veth interface in NetworkManager is useful even without the
support for namespaces for some use cases, e.g "connecting one side of
the veth to an OVS bridge and the other side to a Linux bridge" this is
done when using OVN kubernetes [1][2]. In addition, it would provide
persistent configuration and rollback support for Veth interfaces.
[1] https://bugzilla.redhat.com/show_bug.cgi?id=1885605
[2] https://bugzilla.redhat.com/show_bug.cgi?id=1894139
Signed-off-by: Fernando Fernandez Mancera <ffmancera@riseup.net>
C casts unconditionally force the type, and as such they don't
necessarily improve type safety, but rather overcome restrictions
from the compiler when necessary.
Casting a void pointer is unnecessary (in C), it does not make the
code more readable nor more safe. In particular for g_object_new(),
which is known to return a void pointer of the right type.
Drop such casts.
sed 's/([A-Za-z_0-9]\+ *\* *) *g_object_new/g_object_new/g' $(git grep -l g_object_new) -i
./contrib/scripts/nm-code-format-container.sh
"nm-device-logging.h" defines logging macros for a NMDevice instance.
It also expects a "self" variable in the call environment, and that
variable had to be in the type of NMDevice or the NMDevice subclass.
Extend the macro foo, so that @self can be either a NMDevice* pointer
or a NMDevice$SUBTYPE.
Of course, that would have always been possible, if we would simply cast
to "(NMDevice *)" where we need it. The trick is that the macro only
works if @self is one of the two expected types, and not some arbitrary
unrelated type.
Run:
./contrib/scripts/nm-code-format.sh -i
./contrib/scripts/nm-code-format.sh -i
Yes, it needs to run twice because the first run doesn't yet produce the
final result.
Signed-off-by: Antonio Cardace <acardace@redhat.com>
The autoneg/speed ethtool settings are important. If they are wrong,
the device might not get any carrier. Having no carrier means that
you may be unable to activate a profile (because depending on
configuration, carrier is required to activate a profile).
Since activating profiles are the means to configure the link settings
in NetworkManager, and activating a profile can be hampered by wrong link
settings, it's important to reset the "correct" settings, when deactivating
a profile.
"Correct" in this case means to restore the settings that were present
before NM changed the settings. Presumably, these are the right once.
Beyond that, in the future it might make sense to support configuring
the default link settings per device. So that NM will always restore a
defined, configured, working state. The problem is that per-device
settings currently are only available via NetworkManager.conf, which
is rather inflexible.
Also, when you restart NetworkManager service, it leaves the interface
up but forgets the previous setting. That possibly could be fixed by
persisting the previous link state in /run. However, it's not
implemented yet.
https://gitlab.freedesktop.org/NetworkManager/NetworkManager/-/issues/356https://bugzilla.redhat.com/show_bug.cgi?id=1807171
There was only API to schedule the stage on an idle handler.
Sometimes, we are just in the right situation to schedule the stage
right away. It should be possibly to avoid going through the extra hop.
For now, none of the caller makes use of this. So, there isn't any
actual change in behavior. But by adding this possibility, we may do
use in the future.
Avoid GDBusProxy, instead use GDBusConnection directly. I very much
prefer this because that way we have explicit control over what happens
on D-Bus. With GDBusProxy this is hidden under another layer of complex
code. The hardest part when using a D-Bus interface is to manage the
state via an asynchronous medium. GDBusProxy contains state about the
D-Bus interface and duplicate the state that we track. This makes it hard
to reason about things.
Rework creation of NMSupplicantInterface. Previously, a NMSupplicantInterface
had multiple initialization states. In particular, the first state would not
yet tie the interface to a certain D-Bus object path. Instead, NMSupplicantInterface
would try and retry to create the D-Bus object.
Now, NMSupplicantManager has an asynchronous method to create interface
instances. The manager only creates an interface instance after the D-Bus
path is known. That means, a NMSupplicantInterface instance is now
strongly tied to a name-owner and D-Bus path.
It follows that the state of NMSupplicantInterface can only go from STARTING,
via the supplicant states, to DOWN. Never back. That was already previously
the case that the state from DOWN was final and once the 3 initial
states were passed, the interface's state would never go back to the initial
state. Now this is more strict and more formalized. The 3 initialization states
are combined.
I think the tighter state handling simplifies users of NMSupplicantInterface.
See for example "nm-device-ethernet.c". It's still complicated, because handling
state is fundamentally difficult.
NMSupplicantManager will take care to D-Bus activate wpa_supplicant only
when necessary (poke). Previously, creating the manager instance
would always start suppliant service. Now, it's started on demand.
The _GET_PRIVATE() macros are all implemented based on
_NM_GET_PRIVATE(). That macro tries to be more type safe and uses
_Generic() to do the right thing. Explicitly casting is not only
unnecessary, it defeats these (static) type checks.
Don't do that.
We keep adding capabilities. Tracking them individually via boolean (or
ternary) properties is cumbersome.
Instead, use an enum NMSupplCapType and a corresponding bitmask
NMSupplCapMask. The latter can track whether a capability is detected,
detected to be absent or not detected (unknown).
Most callers would pass FALSE to nm_utils_error_is_cancelled(). That's
not very useful. Split the two functions and have nm_utils_error_is_cancelled()
and nm_utils_error_is_cancelled_is_disposing().
The abbreviations "ns" and "ms" seem not very clear to me. Spell them
out to nsec/msec. Also, in parts we already used the longer abbreviations,
so it wasn't consistent.
After we set link parameters (auto-negotiation, speed, duplex) in
stage1, the carrier can go down for several seconds because the
Ethernet PHY needs to renegotiate the link. Wait that carrier goes up
before starting the supplicant or the EAPoL start packet can be lost
causing an authentication failure.
https://bugzilla.redhat.com/show_bug.cgi?id=1759797
Allow a reapply of the connection when the device is still activating
and ensure that each reapply action is performed only at a given
activation stage. For example, the IP configuration is not reactivated
if the device is in the prepare stage.
https://bugzilla.redhat.com/show_bug.cgi?id=1763062
If the 802.1X authentication fails and 802-1x.optional is set,
continue with activation. In this case, subscribe to the auth-state
supplicant property so that any dynamic IP method can be restarted
when the authentication succeeds. This is because upon authentication
the switch could have changed the VLAN we are connected to.
We try to set only one time the MTU from the connection to not
interfere with manual user changes.
If at some point the parent interface changes temporarily MTU to a
lower value (for example, because the connection was reactivated), the
kernel will also lower the MTU on child interface and we will not
update it ever again.
Add a workaround to this. If we detect that the MTU we want to set
from connection is higher that the allowed one, go into a state where
we follow the parent MTU until it is possible to set again the desired
MTU. This is a bit ugly, but I can't think of any nicer way to do it.
https://bugzilla.redhat.com/show_bug.cgi?id=1751079
I am about to change the when stage1 gets postponed, then the way to
proceed it is to schedule stage1 again (instead of scheduling stage2).
The reason is that stage1 handling should be reentrant and we should
keep entering it until there is no more reason to postpone it. If
a subclass postpones stage1 and then later progresses it by directly
scheduling stage2, then only the subclass is in control over postponing
stage 2.
Instead, anybody should be able to delay stage2 independently. That can
only work if everybody signals readyness to proceed by scheduling stage1
again.
NMDevice's act_stage1_prepare() now does nothing. Calling it is not
useful and has no effect.
In general, when a subclass overwrites a virtual function, it must be
defined whether the subclass must, may or must-not call the parents
implementation. Likewise, it must be clear when the parents
implementation should be chained: first, as last, does it matter?
In any case, that very much depends on how the parent is implemented
and this can only be solved by documentation and common conventions.
It's a forgiving approach to have a parents implementation do nothing,
then the subclass may call it at any time (or not call it at all).
This is especially useful if classes don't know their parent class well.
But in NetworkManager code the relationship between classes are known
at compile time, so every of these classes knows it derives directly
from NMDevice.
This forgingin approach was what NMDevice's act_stage1_prepare() was doing.
However, it also adds lines of code resulting in a different kind of complexity.
So, it's not clear that this forgiving approach is really better. Note
that it also has a (tiny) runtime and code-size overhead.
Change the expectation of how NMDevice's act_stage1_prepare() should be
called: it is no longer implemented, and subclasses *MUST* not chain up.
Only NMDeviceEthernet implements new_default_connection(). Anyway, it
makes only sense to do this precheck by the caller first, and not by
each implementation.
Changing "ipv4.route-table" and "ipv6.route-table" was not allowed
during reapply.
The main difficulty for supporting that is changing the sync-mode.
With route-table 0, we don't sync all tables but only the main table.
So, when reapply changes from full-sync to no-full-sync, it's slightly
more complicated.
But it's probably not too complicated either. The change from
no-full-sync to full-sync is simple: we just start doing a full-sync.
The reverse change is slightly more complicated, because we need to
do one last full-sync, to get rid of routes that we configured on those
other tables.
We no longer add these. If you use Emacs, configure it yourself.
Also, due to our "smart-tab" usage the editor anyway does a subpar
job handling our tabs. However, on the upside every user can choose
whatever tab-width he/she prefers. If "smart-tabs" are used properly
(like we do), every tab-width will work.
No manual changes, just ran commands:
F=($(git grep -l -e '-\*-'))
sed '1 { /\/\* *-\*- *[mM]ode.*\*\/$/d }' -i "${F[@]}"
sed '1,4 { /^\(#\|--\|dnl\) *-\*- [mM]ode/d }' -i "${F[@]}"
Check remaining lines with:
git grep -e '-\*-'
The ultimate purpose of this is to cleanup our files and eventually use
SPDX license identifiers. For that, first get rid of the boilerplate lines.
If the AddAndActivate() caller didn't explicitely a MAC address, default
to pinpointing the connection to the device by the means of an interface
name. This makes more sense than a MAC address with stable device names.
- the previous implementation of nm_setting_wired_get_s390_option()
returned the elements in an arbitrary order (because it just iterated
idx times over the unsorted hash table).
- the API for "s390-options" suggests both accessing by index and by
name. Storing the options in a hash-table is not optimal for lookup
by index. It also requires us to sort the elements over and over
again.
Use instead a sorted array. Note that add/remove of course requires to
move the elements (and has thus O(n)).
- "s390-options" are very seldomly set. We shouldn't pay the price in every
NMSettingWired to allocate a GHashTable and deal with it.
- don't assert in nm_setting_wired_add_s390_option() and
nm_setting_wired_remove_s390_option() that the key is valid.
ifcfg-rh reader understandably does not want to implement additional
logic to pre-validate the key, so any invalid keys would trigger an
assertion failure. We have verify() for this purpose.
We built (among others) two libraries from the sources in "shared/nm-utils":
"libnm-utils-base.la" and "libnm-utils-udev.la".
It's confusing. Instead use directories so there is a direct
correspondence between these internal libraries and the source files.
(cherry picked from commit 2973d68253)
It is preferable to treat IPv4 and IPv6 in a similar manner.
This moves the places where we differ down the call-stack.
It also make it clearer how IPv6 behaves differently. I think this
is a bug, but leave it for now.
+ /* If IP had previously failed, move it back to IP_CONF since we
+ * clearly now have configuration.
+ */
+ if (priv->ip6_state == IP_FAIL)
+ _set_ip_state (self, AF_INET6, IP_CONF);
(cherry picked from commit 1585eaf473)
