If we change the the MTU of an ovs interface only through netlink, the
change could be overridden by ovs-vswitchd at any time when other
interfaces change. Set the MTU also in the ovsdb to prevent such
changes.
Note that if the MTU comes from the connection, we already set the
ovsdb MTU at creation time and so this other update becomes
useless. But it is needed when changing the MTU at runtime (reapply)
or when the MTU comes from a different source (e.g. DHCP).
The ovs-vswitchd.conf.db(5) man page says about the the mtu_request
column in the Interface table:
"Requested MTU (Maximum Transmission Unit) for the interface. A
client can fill this column to change the MTU of an
interface [...] If this is not set and if the interface has
internal type, Open vSwitch will change the MTU to match the
minimum of the other interfaces in the bridge."
Therefore, if the connection specifies a MTU, set it early when adding
the interface to the ovsdb so that it will not be changed to the
minimum of other interfaces.
g_clear_pointer() would always cast the destroy notify function
pointer to GDestroyNotify. That means, it lost some type safety, like
GPtrArray *ptr_arr = ...
g_clear_pointer (&ptr_arr, g_array_unref);
Since glib 2.58 ([1]), g_clear_pointer() is also more type safe. But
this is not used by NetworkManager, because we don't set
GLIB_VERSION_MIN_REQUIRED to 2.58.
[1] f9a9902aac
We have nm_clear_pointer() to avoid this issue for a long time (pre
1.12.0). Possibly we should redefine in our source tree g_clear_pointer()
as nm_clear_pointer(). However, I don't like to patch glib functions
with our own variant. Arguably, we do patch g_clear_error() in
such a manner. But there the point is to make the function inlinable.
Also, nm_clear_pointer() returns a boolean that indicates whether
anything was cleared. That is sometimes useful. I think we should
just consistently use nm_clear_pointer() instead, which does always
the preferable thing.
Replace:
sed 's/\<g_clear_pointer *(\([^;]*\), *\([a-z_A-Z0-9]\+\) *)/nm_clear_pointer (\1, \2)/g' $(git grep -l g_clear_pointer) -i
I think it's preferable to use nm_clear_g_free() instead of
g_clear_pointer(, g_free). The reasons are not very strong,
but I think it is overall preferable to have a shorthand for this
frequently used functionality.
sed 's/\<g_clear_pointer *(\([^;]*\), *\(g_free\) *)/nm_clear_g_free (\1)/g' $(git grep -l g_clear_pointer) -i
Configuration stages like act_stage2_config() can postpone progressing
to the next stage. Currently, when the condition that we wait for gets
satisfied, the code schedules the next stage from there.
I think that is wrong, because when we postpone from act_stage2_config(),
follow up steps of stage2 get skipped. Thus, when we are ready to progress,
the class should enter stage 2 again.
This requires that stage2 becomes reentrant and that the code reenters the
same stage.
We usually want to schedule stage2 when we just completed with the previous
stage (or, if we are currently in stage2, and want to re-enter it).
In those cases, the conditions are often right to just proceed right away.
No need to schedule the stage on an idle handler. Allow to invoke stage2
right away.
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.
Add '_nm_setting_bond_get_option_or_default()' and move all the custom
policies applied by NM for bond options in there.
One such example of a custom policy is to set 'miimon' to 0 (instead of its
default value of 100) if 'arp_interval' is explicitly enabled
and 'miimon' is not.
This means removing every piece of logic from
nm_setting_bond_add_option() which used to clear out 'arp_interval' and
'arp_ip_target' if 'miimon' was set or clear out 'miimon' along with
'downdelay', 'updelay' and 'miimon' if 'arp_interval' was set.
This behaviour is a bug since the kernel allow setting any combination
of this options for bonds and NetworkManager should not limit the user
to do so.
Also use 'set_bond_attr_or_default()' instead of 'set_bond_attr()' as
the former calls '_nm_setting_bond_get_option_or_default()' to implement
the right logic to retrieve bond options according to current bond
configuration.
Fail the enslavement of the ovs port if the bridge device is not
found, instead of generating assertions and potentially crash later.
https://bugzilla.redhat.com/show_bug.cgi?id=1797696
Fixes: 101e65d2bb ('ovs: allow changing mac address of bridges and interfaces')
The previous code tried to get the bridge active connection and it
used the port active connection instead in case of failure. This
doesn't seem right, as in nm-ovsdb.c the bridge AC is used to get the
bridge settings (including the uuid, interface name, and cloned mac).
In case of failure getting the bridge AC we should just fail.
Fixes: 830a5a14cb ('device: add support for OpenVSwitch devices')
The advantage is that the API is now the same for IPv4 and IPv6: it's
all nm_dhcp_config_*() and we can (easier) treat the address family
generically.
We still need two distinct GObject types, mainly because of the
glue code for exposing the object on D-Bus as NMDBusObject. Of course,
that could be solved differently, but as it is, it's quite nice.
When the ovs interface gets deactivated, it is released from the
master port and we call nm_device_update_from_platform_link (dev,
NULL) to ignore any later event for the interface. This is important
especially because it sets a zero ifindex on the interface and so,
later when the link disappears, we don't unmanage the device but
directly remove it.
However, since ovs commands are queued, the link could appear during
the deactivation and we need to ignore such events. Add a new device
method can_update_from_platform_link() for such purpose.
Tracking the deletion of link by ifindex is difficult because the
ifindex of the device is updated through delayed (idle) calls in
NMDevice and so there is the possibility that at a certain time the
device ifindex is not in sync with platform state. It seems simpler to
watch instead the interface name. The ugly thing is that the interface
name can be changed externally, but if users do that on an activating
device they are looking for trouble.
Also change the deactivate code to deal with the scenario where we
already created the interface in the ovsdb but the link didn't show up
yet. To ensure a proper cleanup we must wait that the link appears and
then goes away; however the link may never appear if vswitchd sees
only the last state in ovsdb, and so we must use a ugly timeout to
avoid waiting forever.
https://bugzilla.redhat.com/show_bug.cgi?id=1787989
At this point, just move the fields in their respective address-family
specific structure. We don't use it generically yet, but instead always
explicitly select IPv4 or IPv6. But this would allow to access those
fields by address-family in the future.
Often, the code paths for IPv4 and IPv6 are very similar. We should try
to unify those code paths. The main advantage of doing that, is that
we don't unintentionally end up doing different things. And of course,
it removes duplicate code.
In a first step, unify handling of the grace timeout for DHCPv4 and
DHCPv6.
We will add a property NM_NDISC_RA_TIMEOUT for which this name is better
suited. The problem is really that our convention for object properties
and signals defines have no prefix to indicate whether it's a property
or a signal.
Rename.
This is C, we have almost no IDE support. And ctags/cscope is much more
helpful if we use unique names.
Don't use the get_dhcp_timeout() name, because that is already used in
"src/devices/nm-device.c" already. Rename.
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.
Several macros are used to define function. They had a "_STATIC" variant,
to define the function as static.
I think those macros should not try to abstract entirely what they do.
They should not accept the function scope as argument (or have two
variants per scope). This also because it might make sense to add
additional __attribute__(()) to the function. That only works, if
the macro does not pretend to *not* define a plain function.
Instead, embrace what the function does and let the users place the
function scope as they see fit.
This also follows what is already done with
static NM_CACHED_QUARK_FCN ("autoconnect-root", autoconnect_root_quark)
