Removing and readding the entry to a potentially different row/segment,
while driving out waiting and new threads, could prevent threads from
acquiring the SA even if they were waiting to check it out by unique
ID (which doesn't change), or if they were just trying to enumerate it.
With this change the row and segment doesn't change anymore and waiting
threads may acquire the SA. However, those looking for an IKE_SA by SPIs
might get one back that has a different SPI (but that's probably not
something that happens very often this early).
This was noticed because we check out SAs by unique ID in the Android
app to terminate them after failed retransmits if we are not reestablishing
the SA (otherwise we continue), and this sometimes failed.
Fixes: eaedcf8c00 ("ike-sa-manager: Add method to change the initiator SPI of an IKE_SA")
Much like in commit a68454b, we now use a global atomic counter to keep
track of the number of IKE_SAs currently registered. This should improve
scalability for a large number of segments even more.
Signed-off-by: Thomas Egerer <[email protected]>
Memory is allocated with calloc, hence set to zero, thus assigning the
numerical value 0 is not required.
Signed-off-by: Thomas Egerer <[email protected]>
This makes handling such IKE_SAs more specifically compared to keeping them
in state IKE_CONNECTING or IKE_ESTABLISHED (which we did when we lost a
collision - even triggering the ike_updown event), or using IKE_REKEYING for
them, which would also be ambiguous.
For instance, we can now reject anything but DELETES for such SAs.
A thread might check out a new IKE_SA via checkout_new() or
checkout_by_config() and start initiating it while the daemon is
terminating and the IKE_SA manager is flushed by the main thread.
That SA is not tracked yet so the main thread is not waiting for it and
the other thread is able to check it in and creating an entry after flush()
already terminated causing a memory leak.
Fixes#1348.
This could be helpful when debugging deadlocks that manifest around
wait_for_entry(), as it helps identifying other involved threads (the
thread object is seen in the thread_main() call in each thread's backtrace).
With IKEv1 we have to reuse IKE_SAs as otherwise the responder might
detect the new SA as reauthentication and will "adopt" the CHILD_SAs of
the original IKE_SA, while the initiator will not do so. This could
cause CHILD_SA rekeying to fail later.
Fixes#1236.
If the retransmit of an initial message is processed concurrently with the
original message it might not have been handled as intended as the
thread processing the retransmit might not have seen the correct value
of entry->processing set by the thread handling the original request.
For IKEv1, i.e. without proper message IDs, there might still be races e.g.
when receiving a retransmit of the initial IKE message while processing the
initiator's second request.
Fixes#1269.
We already did so during Phase 1 but because all three Quick Mode
message have the same message ID we occasionally dropped the third
message as retransmit, so we do it there too. For INFORMATIONAL
and TRANSACTION exchanges we don't expect more than one inbound message
with the same message ID so we still use them there.
Fixes#1198.
In some cases we call wait_for_entry() but don't actually check out the
entry afterwards (e.g. because it doesn't match certain criteria). So
there won't be a call to checkin() for such entries causing waiting
threads to get signaled. Instead, such threads would be blocked until
another thread properly checks out/in the entry (or does a blocking
enumeration).
Previously, the current segment was held while checking for duplicate
SAs, which requires acquiring all segments. If multiple threads did this
concurrently this resulted in a deadlock as they couldn't acquire the
segments held by the other threads attempting to do the same. With the
default configuration only one segment is used, which prevents the problem
as only one thread can check in an IKE SA concurrently.
Fixes: a064eaa8a6 ("Handling of initial contact")
Threads might still be allocating SPIs (e.g. triggered by an acquire or
an inbound message) while the main thread calls flush(). If there is a
context switch right after such a thread successfully checked this->rng
in get_spi() and the main thread destroys the RNG instance right then,
that worker thread will cause a segmentation fault when it continues and
attempts to call get_bytes().
Fixes#1014.
Some clients like iOS/Mac OS X don't do a mode config exchange on the
new SA during re-authentication. If we don't adopt the previous virtual
IP Quick Mode rekeying will later fail.
If a client does do Mode Config we directly reassign the VIPs we migrated
from the old SA, without querying the attributes framework.
Fixes#807, #810.
It is mandated by the RFCs and it is expected by the task managers.
Initial messages with invalid MID will be treated like regular messages,
so no IKE_SA will be created for them. Instead, if the responder SPI is 0
no SA will be found and the message is rejected with ALERT_INVALID_IKE_SPI.
If an SPI is set and we do find an SA, then we either ignore the message
because the MID is unexpected, or because we don't allow initial messages
on established connections.
There is one exception, though, if an attacker can slip in an IKE_SA_INIT
with both SPIs set before the client's IKE_AUTH is handled by the server,
it does get processed (see next commit).
References #816.
While the comment is rather clear that we should not adopt live CHILD_SAs
during reauthentication in IKEv2, the code does nonetheless. Add an additional
version check to fix reauthentication if the reauth responder has a replace
uniqueids policy.
Fixes#871.
To check if a received IKE_SA_INIT request is a new request or a
retransmit, charon maintains hashes of the pending IKE_SA_INIT
exchanges.
However, the hash calculation is not reentrant because a single hasher
is used for the whole IKE SA manager. It leads to bogus calculations
under high load and hence dropped messages on responder
(IkeInInvalidSpi incremented).
Don't share a single hasher in the IKE SA manager, create a transient
one whenever a message must be hashed.
Signed-off-by: Christophe Gouault <[email protected]>
This patch is based on one by Christoph Gouault.
Currently, to count the total number of half_open IKE_SAs,
get_half_open_count sums up the count of each segment in the SA hash
table (acquiring a lock for each segment). This procedure does not scale
well when the number of segments increases, as the method is called for
each new negotiation.
Instead, lets maintain a global atomic counter.
This optimization allows the use of big values for charon.ikesa_table_size
and charon.ikesa_table_segments.
When we have a "keep" unique policy in place, we have to be less strict in
rejecting Main/Aggressive Modes to enforce it. If the host/port equals to
that of an existing ISAKMP SA, we assume it is a reauthentication attempt
and accept the new SA (to replace the old).
Sending a DELETE for the replaced SA immediately is problematic during
reauthentication, as the peer might have associated the Quick Modes to the
old SA, and also delete them.
With this change the delete for the old ISAKMP SA is usually omitted, as it
is gets implicitly deleted by the reauth.
Fixes retransmit of the last Main Mode or IKE_AUTH message, and correctly
queues Main Mode messages when processing of the last message is still in
progress.
Tobias Brunner