Maria Matejka [Fri, 13 Oct 2023 08:22:09 +0000 (10:22 +0200)]
Protocol: better granularity of pool management
There are now 3 different pools with specific lifetime. All of these are
available since protocol start, anyway they get freed in different
moments.
First, pool_up gets freed immediately after announcing PS_STOP, to e.g.
stop all timers and events regularly updating the routing table when the
imports are already flushing.
Then, pool_inloop gets freed just before the protocol loop is finally
stopped, after all channels, imports and exports and other hooks are
cleaned up.
And finally, the pool itself is freed the last. Unless you explicitly
need the early free, use this pool.
Ondrej Zajicek [Thu, 27 Apr 2023 16:20:49 +0000 (18:20 +0200)]
Conf: Improve handling of keywords
For whatever reason, parser allocated a symbol for every parsed keyword
in each scope. That wasted time and memory. The effect is worsened with
recent changes allowing local scopes, so keywords often promote soft
scopes (with no symbols) to real scopes.
Do not allocate a symbol for a keyword. Take care of keywords that could
be promoted to symbols (kw_sym) and do it explicitly.
Maria Matejka [Fri, 29 Sep 2023 14:24:50 +0000 (16:24 +0200)]
Channel: Refeeding by an auxiliary request if needed.
If the protocol supports route refresh on export, we keep the stop-start
method of route refeed. This applies for BGP with ERR or with export
table on, for OSPF, Babel, RIP or Pipe.
For BGP without ERR or for future selective ROA reloads, we're adding an
auxiliary export request, doing the refeed while the main export request
is running, somehow resembling the original method of BIRD 2 refeed.
There is also a refeed request queue to keep track of different refeed
requests.
Maria Matejka [Sun, 24 Sep 2023 21:22:43 +0000 (23:22 +0200)]
Protocols have now assigned startup phases
For now, there are 4 phases: Necessary (device), Connector (kernel, pipe), Generator (static, rpki) and Regular.
Started and reconfigured are from Necessary to Regular, shutdown backwards.
This way, kernel can flush routes before actually being shutdown.
Maria Matejka [Sun, 24 Sep 2023 09:47:24 +0000 (11:47 +0200)]
Table: Fixed feed race condition
The problem happened like this:
1. Single route for the given net in table
2. A feed is started
3. The route is deleted (from another thread)
4. The feed finds an empty net, exports nothing, ignores journal (here is bug)
5. The route is added
6. The export transitions from FEEDING to READY
7. While processing the journal, the route deletion and addition combines into noop.
This way routes mysteriously disappeared in specific cases of link instability.
Problem fixed by explicitly marking the empty-net journal entries as processed in step 4.
Maria Matejka [Thu, 24 Aug 2023 15:00:54 +0000 (17:00 +0200)]
Logging: fixed size logfiles behaving as mmapped ringbuffers
This variant of logging avoids calling write() for every log line,
allowing for waitless logging. This makes heavy logging less heavy
and more useful for race condition debugging.
Maria Matejka [Mon, 21 Aug 2023 16:44:10 +0000 (18:44 +0200)]
Logging now doesn't lock with each message
The original logging routines were locking a common mutex. This led to
massive underperformance and unwanted serialization when heavily logging
due to lock contention. Now the logging is lockless, though still
serializing on write() syscalls to the same filedescriptor.
This change also brings in a persistent logging channel structures and
thus avoids writing into active configuration data structures during
regular run.
Ondrej Zajicek [Wed, 13 Sep 2023 04:21:26 +0000 (06:21 +0200)]
Filter: Use common initializer for undefined variables and eattrs.
Undefined paths and clists should use typed f_val with empty adata
instead of just void f_val. Use common initializer to handle both
variables and eattrs.
Ondrej Zajicek [Wed, 12 Jul 2023 18:01:03 +0000 (20:01 +0200)]
Filter: Move argument list reversal from function_call to var_list
List of arguments for function calls is constructed in reverse and then
reverted. This was done in function_call grammar rule. Do the reverse
directly in var_list grammar rule. This fixes reverse order of arguments
in method calls.
Ondrej Zajicek [Mon, 3 Jul 2023 15:00:58 +0000 (17:00 +0200)]
Filter: Implement multiple dispatch for methods
- Extend method descriptors with type signature
- Daisy chain method descriptors for the same symbol
- Dispatch methods for same symbol based on type signature
- Split add/delete/filter operations to multiple methods
- Replace ad-hoc dispatch of old-style syntax with scope-based dispatch
- Also change method->arg_num to count initial arg
It still needs some improvements, like better handling of untyped
expressions and better error reporting when no dispatch can be done.
The multiple dispatch could also be extended to dispatch regular
function-like expressions in a uniform way.
Maria Matejka [Fri, 16 Jun 2023 15:35:37 +0000 (17:35 +0200)]
Filter: Methods rework
Methods can now be called as x.m(y), as long as x can have its type
inferred in config time. If used as a command, it modifies the object,
if used as a value, it keeps the original object intact.
Also functions add(x,y), delete(x,y), filter(x,y) and prepend(x,y) now
spit a warning and are considered deprecated.
It's also possible to call a method on a constant, see filter/test.conf
for examples like bgp_path = +empty+.prepend(1).
Inside instruction definitions (filter/f-inst.c), a METHOD_CONSTRUCTOR()
call is added, which registers the instruction as a method for the type
of its first argument. Each type has its own method symbol table and
filter parser switches between them based on the inferred type of the
object calling the method.
Also FI_CLIST_(ADD|DELETE|FILTER) instructions have been split to allow
for this method dispatch. With type inference, it's now possible.
Ondrej Zajicek [Tue, 29 Aug 2023 16:23:29 +0000 (18:23 +0200)]
BFD: Improve handling of AdminDown
According to RFC 5882, system should not interpret the local or remote
session state transition to AdminDown as failure. We followed that for
the local session state but not for the remote session state (which
just triggered a transition of the local state to Down). The patch
fixes that.
We do not properly generate AdminDown on our side, so the patch is
relevant just for interoperability with other systems.