Commit graph

264 commits

Author SHA1 Message Date
Maria Matejka
1757a6fce5 Filter: Stacks moved to thread-local storage if available. 2019-05-29 21:03:52 +02:00
Jan Maria Matejka
6479e403ef Filters: If somebody doesn't like _Thread_local, don't fail for now, just be a little slower.
When the parallel execution comes into place, we'll likely enforce this
C11 feature. It's much simpler and also faster than pthread_[sg]etspecific().
2019-05-23 11:27:24 +00:00
Jan Maria Matejka
96d757c13f Filter: Store variables and function arguments on stack 2019-05-21 16:33:37 +00:00
Jan Maria Matejka
20c6ea70cc Filter: Making the filter state thread local.
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.
2019-05-20 17:53:10 +00:00
Maria Matejka
f249d0b84c Filters: comparison of functions and filters caching 2019-02-26 16:44:24 +01:00
Maria Matejka
d1039926f5 Filter: Interpreter merged into the common m4 generator.
The config-time partial evaluation of constant expressions in filters is nearby.
2019-02-20 22:30:55 +01:00
Maria Matejka
ea4f55e3dc Filter: More cleanup -- customized structures also in struct f_line_item 2019-02-20 22:30:55 +01:00
Maria Matejka
0b39b1cbb7 Conf: Symbol implementation converted from void pointers to union
... and consted some declarations.
2019-02-20 22:30:55 +01:00
Maria Matejka
132529ce89 Filter: merged filter compare functions into common M4 file 2019-02-20 22:30:55 +01:00
Maria Matejka
dd4d409551 Filter: Merged postfixify routine 2019-02-20 22:30:54 +01:00
Maria Matejka
de12cd18fb Filter: Merged filter line item dumpers into common generated source 2019-02-20 22:30:54 +01:00
Maria Matejka
b256f24145 Filter: auto-generating enum-to-string 2019-02-20 22:30:54 +01:00
Maria Matejka
5289304519 Filter data manipulation functions separated to their file 2019-02-20 22:30:54 +01:00
Maria Matejka
4f082dfa89 Filter: merged filter instruction constructors, counting line size on instruction construct 2019-02-20 22:30:54 +01:00
Maria Matejka
8bdb05edb2 Filters: split the large filter.h file to smaller files.
This should be revised, there are still ugly things in the filter API.
2019-02-20 22:30:54 +01:00
Maria Matejka
c1e97169cd Filter: M4 convertors polished a bit. 2019-02-20 22:30:54 +01:00
Maria Matejka
9b46748d5b Filter: refactoring of instruction constructors 2019-02-20 22:30:54 +01:00
Maria Matejka
4c553c5a5b 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.
2019-02-20 22:30:54 +01:00
Maria Matejka
967b88d938 Filter refactoring: The instructions are converted to the switch body by M4 2019-02-20 22:30:54 +01:00
Maria Matejka
8436040735 Filter refactoring: Drop the roa check specific f_inst 2019-02-20 22:30:54 +01:00
Maria Matejka
7f0ac73724 Filter refactoring: Changed arguments from separate unions to an array 2019-02-20 22:30:54 +01:00
Maria Matejka
224b77d4f7 Filter refactoring: Converted condition to three-args instruction 2019-02-20 22:30:54 +01:00
Maria Matejka
8e8b1fe48c Filter refactoring: Some instructions eat up excessively much space. 2019-02-20 22:30:54 +01:00
Maria Matejka
a8740d6c09 Filter refactoring: indentation fix 2019-02-20 22:30:54 +01:00
Jan Maria Matejka
fc8df41ec6 Filter refactoring: The values are now saved on a custom stack.
This shall help with performance.
2019-02-20 22:30:54 +01:00
Jan Maria Matejka
7afa143886 Filter refactoring: Passing the resulting struct f_val as a pointer.
This also drops the multiplexing of errors with the f_val itself
together with the T_RETURN f_val type flag.
2019-02-20 22:30:54 +01:00
Jan Maria Matejka
f62a369fb4 Filter refactoring: Moved filter instruction definition to a separate file 2019-02-20 22:30:54 +01:00
Jan Maria Matejka
25566c6810 Filter refactoring: Moved the bitfield bit position formula to route.h 2019-02-20 22:30:54 +01:00
Jan Maria Matejka
aca8263926 Filter refactoring: Moved the interpret macros inside the block 2019-02-20 22:30:54 +01:00
Jan Maria Matejka
a946317fab 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.
2019-02-20 22:30:53 +01:00
Maria Matejka
265419a369 Custom route attributes
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.
2018-12-06 09:55:21 +01:00
Ondrej Zajicek (work)
f2d8e6801e Filter: Make ifname attribute modifiable
Allow to change an interface associated with a route by setting
ifname attribute. It will also change the route to a direct one.
2018-11-05 22:03:21 +01:00
Ondrej Zajicek (work)
41b83e52f7 Filter: Fix minor bug in accessing bgp_path
Not relevant for regular BGP paths, just for BGP paths added by filters
to e.g. static routes.
2018-10-25 11:23:15 +02:00
Ondrej Zajicek (work)
586c1800c4 Nest: Neighbor cache cleanups
Simplify neighbor cache code, fix several minor bugs, and improve
handling of ONLINK flag.
2018-06-27 16:57:07 +02:00
Jan Maria Matejka
1771f70d74 Filter: fixed eattr cached pointer
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.
2018-06-19 16:51:40 +02:00
Jan Maria Matejka
13c0be19d3 Nest: Removing separate tmpa from route propagation
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().
2018-05-30 17:08:49 +02:00
Jan Maria Matejka
ee7e2ffd26 Protocol: Introducing an enum protocol_class
This supersedes the EAP_* constants.
2018-05-29 12:35:06 +02:00
Jan Maria Matejka
c3becfe193 Filter: macro for recursive interpretation of instructions 2018-05-29 12:35:06 +02:00
Jan Maria Matejka
0ec6b5ecd3 Filter: Simple type checks converted to ARG() macro 2018-05-29 12:35:06 +02:00
Jan Maria Matejka
478f9babed Filter: Removing the third argument hack
Just to make the code a bit more clean and easier to maintain.
2018-05-29 11:53:51 +02:00
Ondrej Zajicek (work)
b24b781117 Filter: Add support for src filter op to access SADR source prefix
The patch allows to use 'net.src' to access SADR source prefix
from filters.

Thanks to Toke Hoiland-Jorgensen for the original patch for srclen.
2018-05-16 11:19:29 +02:00
Jan Maria Matejka
823ad12191 Filter: Added missing instruction comparators.
These instructions caused SIGABORTs on reconfiguration.
2018-04-27 14:38:41 +02:00
Jan Maria Matejka
8a871e890a Merge branch 'master' into int-new 2018-03-14 12:57:16 +01:00
Jan Maria Matejka
e8bc64e308 Filter: make bgpmask literals real constructors
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.
2018-03-14 11:34:29 +01:00
Jan Maria Matejka
e95705f00c Merge branch 'master' into int-new 2018-03-13 17:02:49 +01:00
Jan Maria Matejka
74bfd2f97c Filters: Removed FI_COMMA, not used for 19 years.
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
2018-03-13 17:01:37 +01:00
Jan Maria Matejka
d1ba927b36 Merge branch 'master' into int-new 2018-03-13 16:51:04 +01:00
Jan Maria Matejka
7c601e6b7b Filter: recursion to loop
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.
2018-03-13 16:29:33 +01:00
Maria Jan Matejka
5a14df3950 Filter: Instruction codes named as enum
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.
2018-03-13 16:29:33 +01:00
Ondrej Zajicek (work)
1561ee799c Handle properly enums for extended attributes 2018-02-13 19:52:22 +01:00