There is a simple universal IO loop, taking care of events, timers and
sockets. Primarily, one instance of a protocol should use exactly one IO
loop to do all its work, as is now done in BFD.
Contrary to previous versions, the loop is now launched and cleaned by
the nest/proto.c code, allowing for a protocol to just request its own
loop by setting the loop's lock order in config higher than the_bird.
It is not supported nor checked if any protocol changed the requested
lock order in reconfigure. No protocol should do it at all.
Channels have now included rt_import_req and rt_export_req to hook into
the table instead of just one list node. This will (in future) allow for:
* channel import and export bound to different tables
* more efficient pipe code (dropping most of the channel code)
* conversion of 'show route' to a special kind of export
* temporary static routes from CLI
The import / export states are also updated to the new algorithms.
Trie walking allows enumeration of prefixes in a trie in the usual
lexicographic order. Optionally, trie enumeration can be restricted
to a chosen subnet (and its descendants).
Routes are now allocated only when they are just to be inserted to the
table. Updating a route needs a locally allocated route structure.
Ownership of the attributes is also now not transfered from protocols to
tables and vice versa but just borrowed which should be easier to handle
in a multithreaded environment.
Add trie tests intended as benchmarks that use external datasets
instead of generated prefixes. As datasets are not included, they
are commented out by default.
Use 16-way (4bit) branching in prefix trie instead of basic binary
branching. The change makes IPv4 prefix sets almost 3x faster, but
with more memory consumption and much more complicated algorithm.
Together with a previous filter change, it makes IPv4 prefix sets
about ~4.3x faster and slightly smaller (on my test data).
For numeric operators, comma is used for disjunction in expressions like
"10, 20, 30..40". But for bitmask operators, comma is used for
conjunction in a way that does not really make much sense. Use always
explicit logical operators (&& and ||) to connect bitmask operators.
Thanks to Matt Corallo for the bugreport.
Add support to set or read outgoing MPLS labels using filters. Currently
this supports the addition of one label per route for the first next hop.
Minor changes by committer.
Broken detection of top-level case caused crash when return was called
from top-of-stack position. It should behave as reject/accept.
Thanks to Damian Zaremba for the bugreport.
Add 'weight' route attribute that allows to get and set ECMP weight of
nexthops. Similar to 'gw' attribute, it is limited to the first nexthop,
but it is useful for handling BGP multipath, where an ECMP route is
merged from multiple regular routes.
Compare the content of PM_ASN_SET in path masks. A reconfiguration
was not properly triggering a reload of affected protocols when the
members of a set in a path mask change.
Also, update the printing code to so that it can display sets in a path
mask.
Add a missing return statement. Path masks with the same length were all
considered the same. Comparing two with different length would cause
out-of-bounds memory access.
Implement regex-like '+' operator in BGP path masks to match previous
path mask item multiple times. This is useful as ASNs may appear
multiple times in paths due to path prepending for traffic engineering
purposes.
Do not apply dynamic type check for second argument of AND/OR, as it is
not evaluated immediately like regular argument would be.
Thanks to Mikael for the bugreport.
Initial parsing of test.conf must be done directly in filter_test main,
while reconfiguration is handled as a regular test. Also fix several
minor issues in test code.
We use constant promotion from IPv4 to Router-ID values, as they have
same literals. Instead of ad-hoc code in filter instructions, add
constant promotion code to parse-time typecheck code.
Most expressions can be type-validated in parse time. It is not
strong enough to eliminate runtime checks, but at least one gets
errors immediately during reconfigure.
Function bodies were compared in post-parse time, yet the result was not
used and the functions were incorrectly considered the same as before.
Now the result is used to reload affected protocols.
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.
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.
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.
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.
For local route marking purposes, local custom route attributes may be
defined. These attributes are seamlessly stripped after export filter to
every real protocol like Kernel, BGP or OSPF, they however pass through
pipes. We currently allow at most 256 custom attributes.
This should be much faster than currently used bgp communities
for marking routes.
Use ACCESS_RTE to guard **f_rte, use ACCESS_EATTRS to guard **f_eattrs.
Use f_rta_cow() before writing to rta or eattrs, use f_rte_cow() before
writing preference (stored in rte).
Do not access eattrs indirectly through (*f_rte)->attrs->eattrs, it is
way too slow. The cached pointer is faster.
This is a fundamental change of an original (1999) concept of route
processing inside BIRD. During import/export, there was a temporary
ea_list created which was to be used instead of the another one inside
the route itself.
This led to some confusion, quirks, and strange filter code that handled
extended route attributes. Dropping it now.
The protocol interface has changed in an uniform way -- the
`struct ea_list *attrs` argument has been removed from store_tmp_attrs(),
import_control(), rt_notify() and get_route_info().
The bgpmask literals can include expressions. This is OK but they have
to be interpreted as soon as the code is run, not in the time the code
is used as value.
This led to strange behavior like rewriting bgpmasks when they shan't
be rewritten:
function mask_generator(int as)
{
return [= * as * =];
}
function another()
bgpmask m1;
bgpmask m2;
{
m1 = mask_generator(10);
m2 = mask_generator(20);
if (m1 == m2) {
print("strange"); # this would happen
}
}
Moreover, sending this to CLI would cause stack overflow and knock down the
whole BIRD, as soon as there is at least one route to execute the given
filter on.
show route filter bgpmask mmm; bgppath ppp; { ppp = +empty+; mmm = [= (ppp ~ mmm) =]; print(mmm); accept; }
The magic match operator (~) inside the bgpmask literal would try to
resolve mmm, which points to the same bgpmask so it would resolve
itself, call the magic match operator and vice versa.
After this patch, the bgpmask literal will get resolved as soon as it's
assigned to mmm and it also will return a type error as bool is not
convertible to ASN in BIRD.
This instruction was removed in the commit linked below
and never used ever again. Rest in peace.
commit 84c7e1943f
Author: Pavel Machek <pavel@ucw.cz>
Date: Tue Mar 2 19:49:28 1999 +0000
It was supposed to do tail-recursion in interpret() but it didn't
compile as such. Converting it to loop makes a significant filter
performance improvement for flat filters.
The two-letter instructions were quite messy but they could be easily
read from memory dumps. Now GDB (since 2012) supports pretty printing
enum values and GCC checks the switch construction for missing enum
values so we are converting the nice two-byte values to enums.
Anyway, the enum still keeps the old two-byte values to be able to read
the instruction codes even without GDB from plain memory dump.
A filter should log messages only if executed explicitly (e.g., during
route export or route import). When a filter is executed for technical
reasons (e.g., to establish whether a route was exported before), it
should run silently.
The patch implements Default Router Preferences and More-Specific Routes
(RFC 4191) for RAdv protocol, allowing to announce router preference and
more specific routes in router advertisements. Routes can be exported to
RAdv like to regular routing protocols.
Some cleanups, bugfixes and other changes done by Ondrej Zajicek.
Dropped struct mpnh and mpnh_*()
Now struct nexthop exists, nexthop_*(), and also included struct nexthop
into struct rta.
Also converted RTD_DEVICE and RTD_ROUTER to RTD_UNICAST. If it is needed
to distinguish between these two cases, RTD_DEVICE is equivalent to
IPA_ZERO(a->nh.gw), RTD_ROUTER is then IPA_NONZERO(a->nh.gw).
From now on, we also explicitely want C99 compatible compiler. We assume
that this 20-year norm should be known almost everywhere.
- Unit Testing Framework (BirdTest)
- Integration of BirdTest into the BIRD build system
- Tests for several BIRD modules
Based on squashed Pavel Tvrdik's int-test branch, updated for
current int-new branch.
We set buffer->pos to buffer->end in function buffer_print() when
bvsnprintf() failed, so there would be uninitialized memory between
the old buffer->pos and the current buffer->pos.
Add support for large communities (draft-ietf-idr-large-community),
96bit alternative to RFC 1997 communities.
Thanks to Matt Griswold for the original patch.
Also removed the lib-dir merging with sysdep. Updated #include's
accordingly.
Fixed make doc on recent Debian together with moving generated doc into
objdir.
Moved Makefile.in into root dir
Retired all.o and birdlib.a
Linking the final binaries directly from all the .o files.
Explicit setting of AF_INET(6|) in IP socket creation. BFD set to listen
on v6, without setting the V6ONLY flag to catch both v4 and v6 traffic.
Squashing and minor changes by Ondrej Santiago Zajicek
New data types net_addr and variants (in lib/net.h) describing
network addresses (prefix/pxlen). Modifications of FIB structures
to handle these data types and changing everything to use these
data types instead of prefix/pxlen pairs where possible.
The commit is WiP, some protocols are not yet updated (BGP, Kernel),
and the code contains some temporary scaffolding.
Comments are welcome.
The patch adds suport for specifying route attributes together with
static routes, e.g.:
route 10.1.1.0/24 via 10.0.0.1 { krt_advmss = 1200; ospf_metric1 = 100; };