While having the filter code still reentrant if we really need,
the compiler can now do constant propagation and address the
thread local storage directly to save some computation time.
The temporary atttributes are no longer removed by ea_do_prune(), but
they are undefined by store_tmp_attrs() protocol hooks. This fixes
several bugs where temporary attributes were removed when they should
not or not removed when they should be. The flag EAF_TEMP is no longer
needed and was removed.
Update all protocol make_tmp_attrs() / store_tmp_attrs() hooks to use
helper functions and to handle unset attributes properly.
Also fix some related bugs like improper handling of empty eattr list.
Keep track of whether OSPF tmpattrs are actually defined for given route
(using flags in rte->pflags). That makes them behave more like real
eattrs so a protocol can define just a subset of them or they can be
undefined by filters.
Do not set ospf_metric2 for other than type 2 external OSPF routes and do
not set ospf_tag for non-external OSPF routes. That also fixes a bug
where internal/inter-area route propagated from one OSPF instance to
another is initiated with infinity ospf_metric2.
Nest: Do not compare rte.flags during rte_update()
Route flags are mosty internal state of rtable, they are not significant
to whether a route has changed. With the old code, all routes received as
a part of enhanced route refresh are always re-announced to other peers
due to change in REF_STALE.
Maria Matejka [Thu, 27 Dec 2018 13:26:11 +0000 (14:26 +0100)]
Filter refactoring: dropped the recursion from the interpreter
This is a major change of how the filters are interpreted. If everything
works how it should, it should not affect you unless you are hacking the
filters themselves.
Anyway, this change should make a huge improvement in the filter performance
as previous benchmarks showed that our major problem lies in the
recursion itself.
There are also some changes in nest and protocols, related mostly to
spreading const declarations throughout the whole BIRD and also to
refactored dynamic attribute definitions. The need of these came up
during the whole work and it is too difficult to split out these
not-so-related changes.
Maria Matejka [Thu, 20 Dec 2018 14:25:04 +0000 (15:25 +0100)]
Filter refactoring: The constant f_val is simply included inside the instruction
With 32 bits, size of the args is 12 bytes, the f_val is 20 bytes.
With 64 bits, size of the args is 24 bytes, the f_val the same.
This is not so nice on 32 bits, anyway the f_inst itself is
24 vs. 32 bytes and the overall size of filters must be 32k of
instructions to get to one megabyte of RAM eaten by f_inst.
Therefore it seems to be improbable for common user to get into
problems with this change.
Filter: Converted static global variables to a filter_state struct.
The static filter state was messy and blocked the planned parallel
execution of filters. Anyway, this will be also slower as the state
structure must be passed almost everywhere with us.
When area is reconfigured to a different type, we need to flush LSAs as
they may not be valid (e.g. NSSA-LSA for non-NSSA area). Also, when we
have have just one OSPF area and that changes type, we could restart OSPF
as there is no state to keep anyway. That solves issue with different
handling of external routes exported to OSPF based of main area type.
OSPF: Basic support for DN-bit handling (RFC 4576)
External LSAs originated by OSPF routers with VPN-PE behavior enabled are
marked by DN flag and they are ignored by other OSPF routers with VPN-PE
enabled.
One of previous workarounds for phantom route avoidance breaks export
counters by expanding sending of spurious withdraws, which are send when
we are not sure whether we have advertised that routes in the past.
If not, then export counter is decreased, but it was not increased
before, so it overflows under zero.
The patch fixes that by sendung spurious withdraws, but not counting them
on export counter. That may lead to error in the other direction, but that
happens only as a race condition (i.e., in normal operation filters
return proper values about old route export state).
Direct acknowledgements should be send as unicast to a corresponding
neighbor. Only delayed acks should be send as multicast to all/designated
routers.
OSPF: Reject duplicate DBDES packets after dead interval
Master may free last DBDES packet immediately. Slave must wait dead
interval before freeing last DBDES packet and then reject duplicate
DBDES packets with SeqNumberMismatch.
Nest: Prevent withdraws from propagation back to source protocol
The earlier fix loosen conditions for not running filters on old
route when deciding about route propagation to a protocol to avoid
issues with ghost routes in some race conditions.
Unfortunately, the fix also caused back-propagation of withdraws. For
regular updates, back-propagation is prevented in import_control hooks,
but these are not called on withdraws. For them, import_control hooks
are called on old routes instead, changing (old, NULL) notification
to (NULL, NULL), which is ignored. By not calling export processing
in some cases, the withdraw is not ignored and is back-propagated.
This patch fixes that by contract conditions so the earlier fix is not
applied to back-propagated updates.
Since v2 we have multiple listening BGP sockets, and each BGP protocol
has associated one of them. Use listening socket that accepted the
incoming connection as a key in the dispatch process so only BGP
protocols assocaited with that listening socket can be selected.
This is necesary for proper dispatch when VRFs are used.
FreeBSD silently changes TTL to 1 when MSG_DONTROUTE is used, even when
it is explicitly set to another value. That breaks TTL security sockets,
including BFD which always uses TTL 255. Bad FreeBSD!
Maria Matějka [Fri, 26 Oct 2018 07:32:35 +0000 (09:32 +0200)]
Perf: Protocol to measure BIRD performance internally
This protocol is highly experimental and nobody should use it in
production. Anyway it may help you getting some insight into what eats
so much time in filter processing.
In some circumstances (old LSA flushed but not acknowledged and not
removed) origination of a new LSA may wrongly triggers LSA collision
code. The patch fixes that.
Thanks to Asbjorn Mikkelsen for the bugreport and @mdelagueronniere
for the original patch.
BGP: Extend 'next hop keep' and 'next hop self' options
Extend 'next hop keep' and 'next hop self' options to have boolean values
(enabled / disabled) and also values 'ibgp'/ 'ebgp' to restrict it to
routes received from IBGP / EBGP. This allows to have it enabled by
default in some cases, matches features of other implementations, and
allows to handle some strange cases like EBGP border router with 'next
hop self' also doing IBGP route reflecting.
Change default of 'next hop keep' to enabled for route servers, and
'ibgp' for route reflectors.