Kernel does not all allow to configure a route via a gateway, if the
gateway is not directly reachable.
For non-manually added routes (e.g. from DHCP), we ignore them as a
server configuration errors. For manually added routes, we try to work
around them.
Note that if the user adds a manual route that references a gateway,
maybe he should be required to also add a matching onlink route for
the gateway (or an address that results in a device-route), otherwise
the configuration could be considered invalid. That was however not
done historically, and also, it seems a rather unhelpful behavior.
NetworkManage should just make it work, not not assume anything is
wrong with the configuration. Similarly, for IPv4, the user could
configure the route as onlink, however, that still requires extra
configuration of which the user might not be aware.
This would apply for example, when a connection has method=auto,
and would obtain the routes automatically. It seems sensible to
allow the user to add a route via the gateway, if he ~knows~ that
this particular network will provide such a configuration via DHCP.
In the past however, we tried not to automatically add a device route,
but instead see whether we will get a suitable route via DHCP. If we
wouldn't get such a route, we would however fail the connection.
However, this is really very hard to get right.
We call ip_config_merge_and_apply() possibly before receiving automatic
IP configuration (commit 7070d17ced, "device: reset
@con_ip6_config on failure before RA"). In this case, we could not yet
configure the route. Instead, we also cannot fail (yet), because we should
wait whether we will receive a route that makes this configuration
feasable.
That is hard to get right. How long should we wait? If we get a DHCP lease
and still cannot add the route, should we fail the IP configuration or wait
longer for another lease? Worse, if we decide to fail the IP configuration,
it might not fail the entire activation. Instead, we would only mark the
current address family as failed. If we later get a DHCP lease, should we
retry to add the route again? -- probably yes. If we still fail, we would
need to keep the IP configuration in failed state, regardless that DHCP
succeeded. Part of the problem is, that we are bad at tracking the
failed state per IP method. So, if manual configuration fails but DHCP
succeeds, we get the state wrong. That should be fixed separately, but it
just shows how hard it is to have this route that we currently cannot
add, and wanting to wait for something that might never come, but still
fail at some point.
Instead, if we cannot add a route due to a missing onlink gateway,
just retry and add the /32 or /128 direct route ourself.
Note that for IPv6 routes that have a "src" address which is still
TENTATIVE, we also cannot currently add the route and retry later.
However, that is fundamentally different, because:
- the configuration here is correct, it's only that the address
didn't yet pass IPv6 DAD and kernel is being unhelpful (rh#1457196).
- we only have to wait a few seconds for DAD to complete or fail.
So, it's easy to implement this sensibly.
The device must not directly add addresses or routes. Instead,
it must track the addresses/routes it wants to add in the NMIP6Config.
Otherwise, during reapply, the information is lost and the next
sync will remove them.
Fixes-test: @ipv6_preserve_cached_routes
We already have IP4LL and maybe should re-use that also for IPv6.
However, when adding the prefix route for IPv6 link local addresses,
we want to add it with protocol "kernel", unlike "user" for IPv4.
There is no strong reason for this. I don't think the protocol matters
much. But up to now kernel automatically adds this prefix route, so
as we are going to change that and let NetworkManager add it, keep the
protocol at "kernel".
The libnm API fir checkpoints was only introduced with 1.11. It
is not yet stable, so there is still time to adjust it. Note that
this changes API/ABI of the development branch.
Changes:
- we only add async variants of the checkpoint functions. I believe
that synchronous D-Bus methods are fundamentally flawed, because
they mess up the ordering of events.
Rename the async functions by removing the "_async" suffix. This
matches glib style, for which the async form is also not specially
marked.
- for function that refere to a particular checkpoint (rollback and
destroy), accept the D-Bus path as string, instead of an NMCheckpoint
instance. This form is more flexible, because it allows to use
the function without having a NMCheckpoint instance at hand. On the
other hand, if one has a NMCheckpoint instance, he can trivially
obtain the path to make the call.
- indent with spaces for wrapped line
- drop g_return_val_if_fail(). It really shouldn't happen,
and if it does, we can just continue and will hit g_critical()
warnings below, when using the variables.
- get the NMActRequest instance first, and from there the applied
and settings connection. Not the other way around.
nm_device_get_settings_connection() and nm_device_get_applied_connection()
are just convenience functions to get the fields from
act_request.
- make python2 and python3 compatible
- support Checkpoint create flags (requires recent libnm for GI
of flags)
- support adjust-rollback-timeout command
- print elapsed time in show output
We need common operations from the python scripts.
For example, to print the boot-time.
Move such utils to a separate nmex.py file ("ex" for
example). This file should contain helper functions that
are pure python (or, if the have requirements, load them
only on demand, so that examples that need those have
additional dependencies). It should also be simple to extract
individual helpers from nmex, so that the user can take an
example, merge parts of nmex.py in, and modify it to his needs.
This allows to adjust the timeout of an existing checkpoint.
The main usecase of checkpoints, is to have a fail-safe when
configuring the network remotely. By allowing to reset the timeout,
the user can perform a series of actions, and keep bumping the
timeout. That way, the entire series is still guarded by the same
checkpoint, but the user can start with short timeout, and
re-adjust the timeout as he goes along.
The libnm API only implements the async form (at least for now).
Sync methods are fundamentally wrong with D-Bus, and it's probably
not needed. Also, follow glib convenction, where the async form
doesn't have the _async name suffix. Also, accept a D-Bus path
as argument, not a NMCheckpoint instance. The libnm API should
not be more restricted than the underlying D-Bus API. It would
be cumbersome to require the user to lookup the NMCheckpoint
instance first, especially since libnm doesn't provide an efficient
or convenient lookup-by-path method. On the other hand, retrieving
the path from a NMCheckpoint instance is always possible.
Note that this changes API for checkpoint_create_async() in Python
via GIR. Previously it would require an integer argument, now a flags
argument. But this API is still unstable, it will be introduced with
1.12.
Introduce a new flag NM_CHECKPOINT_CREATE_FLAG_ALLOW_OVERLAPPING
that allows the creation of overlapping checkpoints. Before, and
by default, checkpoints that reference a same device conflict,
and creating such a checkpoint failed.
Now, allow this. But during rollback automatically destroy all
overlapping checkpoints that were created after the checkpoint
that is about to rollback.
With this, you can create a series of checkpoints, and rollback them
individually. With the restriction, that if you once rolled back to an
older checkpoint, you no longer can roll"forward" to a younger one.
What this implies and what is new here, is that the checkpoint might be
automatically destroyed by NetworkManager before the timeout expires. When
the user later would try to manually destroy/rollback such a checkpoint, it
would fail because the checkpoint no longer exists.
We already do error checking in nm_checkpoint_manager_create(). No need
to split it in two places. Let all error conditions be handled by
nm_checkpoint_manager_create() first, and then once we decide all is
good, nm_checkpoint_new() can no longer fail.
Instead of scheduling one timeout only, let each checkpoint instance
individually schedule a timeout. This has some overhead, but glib
is supposed to make scheduling many timers efficient. Otherwise,
glib should be fixed.
This simplifies in my opinion the code, because it's up to each
checkpoint to maintain its own timeout.
Later we will also add a AdjustRollbackTimeout operation, which
allow to reschedule the timeout. It also seems slightly simpler,
if scheduling of the timeout is done by the NMCheckpoint instance
itself.
If no device paths are given, we can take the devices directly.
We don't need to first create a list of paths, and then
look them up by path again to add them to the list.
NMCheckpointManager was added and is not ref-countable, because it
is not needed.
I still often like for such objects (that are not ref-countable),
that their destroy function is called "unref". Both for consistency,
and also if we would later add ref-counting to the object.
However, NMCheckpointManager keeps a pointer to NMManager. So, when
NMManager gets destroyed, it *MUST* destroy the NMCheckpointManager.
It cannot accept that the checkpoint manager outlives NMManager,
but the "unref" name suggests that somebody else might have still
a reference to this object keeping it alive. That is not the case.
Rename so that this is clear.
I would name it nm_checkpoint_manager_destroy(), but "destroy" already
has a meaning for NMCheckpoint instances, so use "free".
We don't need an external CheckpointItem, just to wrap the
CList instance. Embed it directly in NMCheckpoint.
Sure, that exposes the checkpoints_lst field in the (internal)
header file, hiding the private member less.
Now that the D-Bus signals in server are reordered, creating
a checkpoint in libnm crashes:
$ examples/python/gi/checkpoint.py create 4
#0 0x00007ffff6d011ee in __strcmp_sse2_unaligned () at /lib64/libc.so.6
#1 0x00007fffeb611c90 in find_checkpoint_info (manager=manager@entry=0x5555559e5110 [NMManager], path=0x7fffdc0092f0 "/org/freedesktop/NetworkManager/Checkpoint/6")
at libnm/nm-manager.c:153
#2 0x00007fffeb611d8f in checkpoint_added (manager=0x5555559e5110 [NMManager], checkpoint=checkpoint@entry=0x555555a122d0 [NMCheckpoint]) at libnm/nm-manager.c:1194
#3 0x00007fffef7db929 in g_cclosure_marshal_VOID__OBJECTv (closure=0x5555559e4b30, return_value=<optimized out>, instance=<optimized out>, args=<optimized out>, marshal_data=<optimized out>, n_params=<optimized out>, param_types=0x5555559e2fc0) at gmarshal.c:2102
#4 0x00007fffef7d8976 in _g_closure_invoke_va (closure=0x5555559e4b30, return_value=0x0, instance=0x5555559e5110, args=0x7fffffffc1c8, n_params=1, param_types=0x5555559e2fc0)
at gclosure.c:867
#5 0x00007fffef7f3ff4 in g_signal_emit_valist (instance=instance@entry=0x5555559e5110, signal_id=signal_id@entry=97, detail=0, var_args=var_args@entry=0x7fffffffc1c8) at gsignal.c:3300
#6 0x00007fffef7f4b48 in g_signal_emit_by_name (instance=instance@entry=0x5555559e5110, detailed_signal=detailed_signal@entry=0x7fffffffc310 "checkpoint-added") at gsignal.c:3487
#7 0x00007fffeb6156d1 in deferred_notify_cb (data=0x5555559e5110) at libnm/nm-object.c:219
#8 0x00007fffeb615ae7 in object_property_maybe_complete (self=0x5555559e5110 [NMManager]) at libnm/nm-object.c:555
#9 0x00007fffeb615e5d in object_created (obj=<optimized out>, path=<optimized out>, user_data=<optimized out>) at libnm/nm-object.c:576
#10 0x00007fffeb61648b in handle_object_array_property (pi=<optimized out>, value=0x7fffdc075070, property_name=0x7fffeb67f117 "checkpoints", self=0x5555559e5110 [NMManager])
at libnm/nm-object.c:671
#11 0x00007fffeb61648b in handle_property_changed (self=self@entry=0x5555559e5110 [NMManager], dbus_name=<optimized out>, value=<optimized out>) at libnm/nm-object.c:740
#12 0x00007fffeb6166e9 in properties_changed (proxy=<optimized out>, changed_properties=<optimized out>, invalidated_properties=<optimized out>, user_data=0x5555559e5110)
at libnm/nm-object.c:772
...
That is, because NetworkManager now first emits signals that the checkpoint
object was created, before answering the D-Bus request. That makes more
sense, but leads to this crash.
The ugliness of how libnm handles object visibility is considerable.
libnm hides objects until they are fully initialized. So, when
the async create-checkpoint operation returns, the object might not
yet be ready to be exposed. We need to delay the result. It would be
better if the API would simply return the created path.
Similar cases of such a field are named "manager". Also,
internal names shall not have an "nm" prefix, contrary
to names in a header file, which shall have such a prefix.
I find it slightly nicer and explict. Also, the list elements
are strictly speaking private, we should better not explicitly
use them outside of NMManager/NMDevice. The macro hides this.
Instead of generating a connection and checking whether it is
compatible with the connection indicated in the state file, just pick
the indicated connection after a basic check of compatibility with the
device.
https://bugzilla.redhat.com/show_bug.cgi?id=1551958
It generates the following warning:
libnm/nm-autoptr.h:25: Error: NM: identifier not found on the first line:
* Note that you might use this header with older versions of libnm
Fixes: ff8e563365
They were not (notably) touched in more than 3 years.
I doubt anybody is using them.
Also, nowadays we have contrib/rpm to build NetworkManager
packages for Fedora/RPM. We have copr, we have automated CI
in CentOS CI and beaker.
Also, nowadays it should be easy to spawn a a fedora image
in a container or tools like vagrant.
I think there are better alternatives. Drop the scripts.
Eventually, we should replace our uses of libgsystem's gsystem-local-alloc.h
by glib's g_auto*. As a first tiny step, add a compat implementation for g_autofree,
so that we could at least go ahead and use it instead of gs_free.
https://bugzilla.gnome.org/show_bug.cgi?id=794294
"nm-autoptr.h" is done in a way that allows you to copy the header
in your source tree to support older versions of libnm, that didn't
contain the header yet. For example, we might want to use it in
network-manager-applet, but we don't want to bump the libnm dependency
to 1.11.2+ only to get this functionality.
Note that G_DEFINE_AUTOPTR_CLEANUP_FUNC() was added in glib 2.43.4,
and requires compiler support for the cleanup attribute. The compiler
support is taken as given, because we rely on it already. However,
NetworkManager and network-manager-applet still don't depend on a glib
version recent enough to provide these macros. To actually use them
(*inside*) NetworkManager/network-manager-applet, we either would have
to bump the glib minimal dependency, or reimplement g_autoptr in
/shared/nm-utils/nm-glib.h compat header.
https://bugzilla.gnome.org/show_bug.cgi?id=794294
- consistently set options, searches, domains fields to %NULL,
if there are no values.
- in nm_global_dns_config_update_checksum(), ensure that we uniquely
hash values. E.g. a config with "searches[a], options=[b]" should
hash differently from "searches=[ab], options=[]".
- in nm_global_dns_config_to_dbus(), reuse the sorted domain list.
We already have it, and it guarantees a consistent ordering of
fields.
- in global_dns_domain_from_dbus(), fix memleaks if D-Bus strdict
contains duplicate entries.
