The previous code had some issues because it handled each child config
separately. Not only was this quite inefficient because all IKE_SAs had
to be enumerated for every config, it also caused problems with the check
for other CHILD_SAs in order to decide whether to delete the IKE_SA or
not. Because CHILD_SAs are deleted with an INFORMATIONAL exchange, they
are not immediately gone. This caused a race condition and with more
than one child config and SAs the IKE_SA could be kept because it
could appear as if other, unrelated CHILD_SAs were still there.
Another race condition, which is fixed by the previous commit, occurred
when only changing child configs. Then it could happen that the code
deemed the IKE_SA empty and a delete for it was queued. If that happened
while the IKE_SA was deleting one of the CHILD_SAs (or was busy with some
other exchange), the IKE_SA was not switched to IKE_DELETING. So it
looked usable and create-child tasks for the updated configs might have
gotten queued. Unfortunately, once the ike-delete task is eventually
executed, these tasks would be gone and the replacement CHILD_SAs never
created. This commit additionally avoids actually deleting the IKE_SA
even if all child configs change or get removed if any new CHILD_SAs are
to be initiated.
The previous option caused such requests to be enabled if not explicitly
disabled, which only the vici plugin did, for all other backends requests
would have been sent.
References strongswan/strongswan#2016
socket.connect() takes a single argument with the address that depends
on the address family, for TCP it's a tuple with IP and port.
Fixes: 00a75e332f ("vici: Create default TCP socket on Windows in Python bindings")
Closesstrongswan/strongswan#1874
As long as any `child*` selector is received, only CHILD_SAs will be
terminated or rekeyed. Any passed `ike*` selectors will only be used to
filter the IKE_SAs when looking for matching CHILD_SAs. However, the
previous log messages seemed to indicate that IKE_SAs will also be
terminated/rekeyed.
References strongswan/strongswan#1655
Previously, the logger installed by the controller always announced
LEVEL_PRIVATE(4), which produced completely useless logging calls with
the common clients (vici/stroke) whose default log level is LEVEL_CTRL(1).
This can produce quite some overhead if there are e.g. a lot of concurrent
initiate() calls.
This should prevent a deadlock that could previously be caused when a
control-log event was raised. The deadlock looked something like this:
* Thread A holds the read lock on bus_t and raises the control-log event.
This requires acquiring the connection entry in write mode to queue the
outgoing message. If it is already held by another thread, this blocks
on a condvar.
* Thread B is registering the on_write() callback on the same connection's
stream due to a previous log message. Before this change, the code
acquired the entry in write mode as well, thus, blocking thread A. To
remove/add the stream, the mutex in watcher_t needs to be acquired.
* Thread C is in watcher_t's watch() and holds the mutex while logging on
level 2 or 3. The latter requires the read lock on bus_t, which should
usually be acquirable even if thread A holds it. Unless writers are
concurrently waiting on the lock and the implementation is blocking
new readers to prevent writer starvation.
* Thread D is removing a logger from the bus (e.g. after an initiate()
call) and is waiting to acquire the write lock on bus_t and is thereby
blocking thread C.
With this change, thread B should not block thread A anymore. So thread D
and thread C should eventually be able to proceed as well.
Thread A could be held up a bit if there is a thread already sending
messages for the same connection, but that should only cause a delay, no
deadlock, as on_write() and do_write() don't log (or lock) anything while
keeping the entry locked in write mode.
Closesstrongswan/strongswan#566
Also adds error reporting via VICI for issues with mediation settings.
Fixes: 229cdf6bc8 ("vici: Order auth rounds by optional `round` parameter instead of by position in the request")
References strongswan/strongswan#1569
Since 3dd5dc5011 ("Merge branch 'vici-python-timeout'"), any timeout
set directly on the socket is reset by `vici.Transport.receive()` unless
called by `vici.Session.listen()`. This prevents configuring a default
timeout directly on the socket.
However, setting a timeout directly on the socket also had drawbacks
since it can cause `vici.Transport.receive()` to raise a timeout error
while a subset of the data have been received, with no way to recover.
This commit merges both approaches by considering the timeout configured
on the socket by default (when no timeout is explicitly set) and
keeping the switch to blocking receive once the first byte has been
received. When the full expected data have been received, the timeout
configured on the socket is restored.
Co-authored-by: Tobias Brunner <tobias@strongswan.org>
Signed-off-by: Jean-Tiare Le Bigot <jt@yadutaf.fr>
Closesstrongswan/strongswan#1562
When issuing a deregistration message, some last events may still be sent
by the daemon before deregistration completes. Ignore such event messages
silently during deregistration, avoiding exceptions like:
vici.exception.SessionException:
Unexpected response type 7, expected '5' (EVENT_CONFIRM)
The architecture of the Python client is completely blocking, which is fine
for many simple scripts. For more complex applications that do other I/O
and listen for vici events, the most feasible way to integrate the client is
to use a dedicated thread.
Unfortunately, Python has no simple support for thread cancellation. And
having that thread in a blocking recv() does not allow to terminate the
thread gracefully with an Event or the like.
As a way out, add a timeout to the listen() call, so the thread can
periodically do other things, like checking for termination Event and
react on it. Returning from listen() on timeout can be suboptimal, though,
as it involves registration/deregistration for events, including the risk
for missing events while not registered. So return a (None, None) tuple
instead on timeout, allowing the caller to periodically do other things
while staying registered for the events and continue in listen().
The timeout applies to the socket recv() for the start of the header, only,
so a message is either read in full or times out, avoiding the risk
of breaking message framing on the stream with partial reads.
Depending on the configured hard lifetime the default soft lifetime
might not make sense and could even cause rekeying to get disabled.
To avoid that, derive the soft lifetime from the hard lifetime so it's
10% higher than the soft lifetime.
References strongswan/strongswan#1414
There are a lot of calls like this:
this->dispatcher->raise_event(this->dispatcher, "...", 0,
b->finalize(b));
However, if finalize() fails, e.g. because a previous call to add()
failed due to the size limit, it returns NULL. This then caused a
segmentation fault in raise_event() when it interacted with that value.
Closesstrongswan/strongswan#1278
mallinfo() is deprecated because it uses `int` for the members of the
returned struct, whereas mallinfo2() uses `size_t`. It's available
since glibc 2.33.
While combining the actions could cause duplicates (while the SA is
initiated, traffic might trigger the trap and the initiation of another
CHILD_SA), the previous commit should avoid most duplicates. If reuse_ikesa
is disabled, duplicates can't be prevented, though.
If two threads are waiting in find_entry() and remove_entry(),
respectively, and the former is woken first, the latter remains stuck
as it won't get signaled.
If there are threads waiting in find_entry() and one in remove_entry()
and the latter is woken first by a thread calling put_entry(), the
former threads would remain stuck as they get never signaled.
Tobias Brunner