diff options
69 files changed, 6243 insertions, 835 deletions
diff --git a/.gitlab-ci.yml b/.gitlab-ci.yml index 65a0a05b..678a08f6 100644 --- a/.gitlab-ci.yml +++ b/.gitlab-ci.yml @@ -3,7 +3,7 @@ variables: LC_ALL: C.UTF-8 GIT_STRATEGY: fetch DOCKER_CMD: docker --config="$HOME/.docker/$CI_JOB_ID/" - IMG_BASE: registry.labs.nic.cz/labs/bird + IMG_BASE: registry.nic.cz/labs/bird TOOLS_DIR: /var/lib/gitlab-runner/bird-tools stages: @@ -16,7 +16,7 @@ stages: stage: image allow_failure: true script: - - $DOCKER_CMD login -u gitlab-ci-token -p $CI_BUILD_TOKEN registry.labs.nic.cz + - $DOCKER_CMD login -u gitlab-ci-token -p $CI_BUILD_TOKEN registry.nic.cz # Make sure we refresh the base image if it updates (eg. security updates, etc) # If we do just the build, cache is always reused and the freshness of the # base image is never checked. However, pull always asks and updates the @@ -164,9 +164,9 @@ docker_ubuntu-20_04-amd64: IMG_NAME: "ubuntu-20.04-amd64" <<: *docker_build -docker_ubuntu-20_10-amd64: +docker_ubuntu-21_10-amd64: variables: - IMG_NAME: "ubuntu-20.10-amd64" + IMG_NAME: "ubuntu-21.10-amd64" <<: *docker_build # GPG error @@ -234,131 +234,131 @@ docker_opensuse-15.3-amd64: build-debian-8-amd64: <<: *build-linux - image: registry.labs.nic.cz/labs/bird:debian-8-amd64 + image: registry.nic.cz/labs/bird:debian-8-amd64 build-debian-8-i386: <<: *build-linux - image: registry.labs.nic.cz/labs/bird:debian-8-i386 + image: registry.nic.cz/labs/bird:debian-8-i386 build-debian-9-amd64: <<: *build-linux - image: registry.labs.nic.cz/labs/bird:debian-9-amd64 + image: registry.nic.cz/labs/bird:debian-9-amd64 build-debian-9-i386: <<: *build-linux - image: registry.labs.nic.cz/labs/bird:debian-9-i386 + image: registry.nic.cz/labs/bird:debian-9-i386 build-debian-10-amd64: <<: *build-linux - image: registry.labs.nic.cz/labs/bird:debian-10-amd64 + image: registry.nic.cz/labs/bird:debian-10-amd64 build-debian-10-i386: <<: *build-linux - image: registry.labs.nic.cz/labs/bird:debian-10-i386 + image: registry.nic.cz/labs/bird:debian-10-i386 build-debian-11-amd64: <<: *build-linux - image: registry.labs.nic.cz/labs/bird:debian-11-amd64 + image: registry.nic.cz/labs/bird:debian-11-amd64 #build-debian-11-i386: # <<: *build-linux -# image: registry.labs.nic.cz/labs/bird:debian-11-i386 +# image: registry.nic.cz/labs/bird:debian-11-i386 build-debian-testing-amd64: <<: *build-linux - image: registry.labs.nic.cz/labs/bird:debian-testing-amd64 + image: registry.nic.cz/labs/bird:debian-testing-amd64 #build-debian-testing-i386: # <<: *build-linux -# image: registry.labs.nic.cz/labs/bird:debian-testing-i386 +# image: registry.nic.cz/labs/bird:debian-testing-i386 build-fedora-25-amd64: <<: *build-linux - image: registry.labs.nic.cz/labs/bird:fedora-25-amd64 + image: registry.nic.cz/labs/bird:fedora-25-amd64 build-fedora-26-amd64: <<: *build-linux - image: registry.labs.nic.cz/labs/bird:fedora-26-amd64 + image: registry.nic.cz/labs/bird:fedora-26-amd64 build-fedora-27-amd64: <<: *build-linux - image: registry.labs.nic.cz/labs/bird:fedora-27-amd64 + image: registry.nic.cz/labs/bird:fedora-27-amd64 build-fedora-28-amd64: <<: *build-linux - image: registry.labs.nic.cz/labs/bird:fedora-28-amd64 + image: registry.nic.cz/labs/bird:fedora-28-amd64 build-fedora-29-amd64: <<: *build-linux - image: registry.labs.nic.cz/labs/bird:fedora-29-amd64 + image: registry.nic.cz/labs/bird:fedora-29-amd64 build-fedora-30-amd64: <<: *build-linux - image: registry.labs.nic.cz/labs/bird:fedora-30-amd64 + image: registry.nic.cz/labs/bird:fedora-30-amd64 build-fedora-31-amd64: <<: *build-linux - image: registry.labs.nic.cz/labs/bird:fedora-31-amd64 + image: registry.nic.cz/labs/bird:fedora-31-amd64 build-fedora-32-amd64: <<: *build-linux - image: registry.labs.nic.cz/labs/bird:fedora-32-amd64 + image: registry.nic.cz/labs/bird:fedora-32-amd64 build-fedora-33-amd64: <<: *build-linux - image: registry.labs.nic.cz/labs/bird:fedora-33-amd64 + image: registry.nic.cz/labs/bird:fedora-33-amd64 build-fedora-34-amd64: <<: *build-linux - image: registry.labs.nic.cz/labs/bird:fedora-33-amd64 + image: registry.nic.cz/labs/bird:fedora-33-amd64 build-centos-7-amd64: <<: *build-linux - image: registry.labs.nic.cz/labs/bird:centos-7-amd64 + image: registry.nic.cz/labs/bird:centos-7-amd64 build-centos-8-amd64: <<: *build-linux - image: registry.labs.nic.cz/labs/bird:centos-8-amd64 + image: registry.nic.cz/labs/bird:centos-8-amd64 build-ubuntu-14_04-amd64: <<: *build-linux - image: registry.labs.nic.cz/labs/bird:ubuntu-14.04-amd64 + image: registry.nic.cz/labs/bird:ubuntu-14.04-amd64 build-ubuntu-16_04-amd64: <<: *build-linux - image: registry.labs.nic.cz/labs/bird:ubuntu-16.04-amd64 + image: registry.nic.cz/labs/bird:ubuntu-16.04-amd64 build-ubuntu-18_04-amd64: <<: *build-linux - image: registry.labs.nic.cz/labs/bird:ubuntu-18.04-amd64 + image: registry.nic.cz/labs/bird:ubuntu-18.04-amd64 build-ubuntu-20_04-amd64: <<: *build-linux - image: registry.labs.nic.cz/labs/bird:ubuntu-20.04-amd64 + image: registry.nic.cz/labs/bird:ubuntu-20.04-amd64 -build-ubuntu-20_10-amd64: +build-ubuntu-21_10-amd64: <<: *build-linux - image: registry.labs.nic.cz/labs/bird:ubuntu-20.10-amd64 + image: registry.nic.cz/labs/bird:ubuntu-21.10-amd64 #build-ubuntu-21_04-amd64: # <<: *build-linux -# image: registry.labs.nic.cz/labs/bird:ubuntu-21.04-amd64 +# image: registry.nic.cz/labs/bird:ubuntu-21.04-amd64 build-opensuse-15.0-amd64: <<: *build-linux - image: registry.labs.nic.cz/labs/bird:opensuse-15.0-amd64 + image: registry.nic.cz/labs/bird:opensuse-15.0-amd64 build-opensuse-15.1-amd64: <<: *build-linux - image: registry.labs.nic.cz/labs/bird:opensuse-15.1-amd64 + image: registry.nic.cz/labs/bird:opensuse-15.1-amd64 build-opensuse-15.2-amd64: <<: *build-linux - image: registry.labs.nic.cz/labs/bird:opensuse-15.2-amd64 + image: registry.nic.cz/labs/bird:opensuse-15.2-amd64 build-opensuse-15.3-amd64: <<: *build-linux - image: registry.labs.nic.cz/labs/bird:opensuse-15.3-amd64 + image: registry.nic.cz/labs/bird:opensuse-15.3-amd64 build-freebsd-11-amd64: <<: *build-base @@ -408,112 +408,113 @@ build-freebsd-11-i386: #pkg-debian-8-amd64: # <<: *pkg-deb # needs: [build-debian-8-amd64] -# image: registry.labs.nic.cz/labs/bird:debian-8-amd64 +# image: registry.nic.cz/labs/bird:debian-8-amd64 # Dpkg error: PATH is not set #pkg-debian-8-i386: # <<: *pkg-deb # needs: [build-debian-8-i386] -# image: registry.labs.nic.cz/labs/bird:debian-8-i386 +# image: registry.nic.cz/labs/bird:debian-8-i386 # Dpkg error: PATH is not set pkg-debian-9-amd64: <<: *pkg-deb needs: [build-debian-9-amd64] - image: registry.labs.nic.cz/labs/bird:debian-9-amd64 + image: registry.nic.cz/labs/bird:debian-9-amd64 # Dpkg error: PATH is not set pkg-debian-9-i386: <<: *pkg-deb needs: [build-debian-9-i386] - image: registry.labs.nic.cz/labs/bird:debian-9-i386 + image: registry.nic.cz/labs/bird:debian-9-i386 pkg-debian-10-amd64: <<: *pkg-deb needs: [build-debian-10-amd64] - image: registry.labs.nic.cz/labs/bird:debian-10-amd64 + image: registry.nic.cz/labs/bird:debian-10-amd64 pkg-debian-10-i386: <<: *pkg-deb needs: [build-debian-10-i386] - image: registry.labs.nic.cz/labs/bird:debian-10-i386 + image: registry.nic.cz/labs/bird:debian-10-i386 pkg-debian-11-amd64: <<: *pkg-deb needs: [build-debian-11-amd64] - image: registry.labs.nic.cz/labs/bird:debian-11-amd64 + image: registry.nic.cz/labs/bird:debian-11-amd64 pkg-fedora-30-amd64: <<: *pkg-rpm-wa needs: [build-fedora-30-amd64] - image: registry.labs.nic.cz/labs/bird:fedora-30-amd64 + image: registry.nic.cz/labs/bird:fedora-30-amd64 pkg-fedora-31-amd64: <<: *pkg-rpm-wa needs: [build-fedora-31-amd64] - image: registry.labs.nic.cz/labs/bird:fedora-31-amd64 + image: registry.nic.cz/labs/bird:fedora-31-amd64 pkg-fedora-32-amd64: <<: *pkg-rpm-wa needs: [build-fedora-32-amd64] - image: registry.labs.nic.cz/labs/bird:fedora-32-amd64 + image: registry.nic.cz/labs/bird:fedora-32-amd64 pkg-fedora-33-amd64: <<: *pkg-rpm-wa needs: [build-fedora-33-amd64] - image: registry.labs.nic.cz/labs/bird:fedora-33-amd64 + image: registry.nic.cz/labs/bird:fedora-33-amd64 pkg-fedora-34-amd64: <<: *pkg-rpm needs: [build-fedora-34-amd64] - image: registry.labs.nic.cz/labs/bird:fedora-34-amd64 + image: registry.nic.cz/labs/bird:fedora-34-amd64 pkg-centos-7-amd64: <<: *pkg-rpm-wa variables: LC_ALL: en_US.UTF-8 needs: [build-centos-7-amd64] - image: registry.labs.nic.cz/labs/bird:centos-7-amd64 + image: registry.nic.cz/labs/bird:centos-7-amd64 pkg-centos-8-amd64: <<: *pkg-rpm-wa needs: [build-centos-8-amd64] - image: registry.labs.nic.cz/labs/bird:centos-8-amd64 + image: registry.nic.cz/labs/bird:centos-8-amd64 pkg-ubuntu-18.04-amd64: <<: *pkg-deb needs: [build-ubuntu-18_04-amd64] - image: registry.labs.nic.cz/labs/bird:ubuntu-18.04-amd64 + image: registry.nic.cz/labs/bird:ubuntu-18.04-amd64 pkg-ubuntu-20.04-amd64: <<: *pkg-deb needs: [build-ubuntu-20_04-amd64] - image: registry.labs.nic.cz/labs/bird:ubuntu-20.04-amd64 + image: registry.nic.cz/labs/bird:ubuntu-20.04-amd64 -pkg-ubuntu-20.10-amd64: + +pkg-ubuntu-21.10-amd64: <<: *pkg-deb - needs: [build-ubuntu-20_10-amd64] - image: registry.labs.nic.cz/labs/bird:ubuntu-20.10-amd64 + needs: [build-ubuntu-21_10-amd64] + image: registry.nic.cz/labs/bird:ubuntu-21.10-amd64 #pkg-ubuntu-21.04-amd64: # <<: *pkg-deb # needs: [build-ubuntu-21_04-amd64] -# image: registry.labs.nic.cz/labs/bird:ubuntu-21.04-amd64 +# image: registry.nic.cz/labs/bird:ubuntu-21.04-amd64 pkg-opensuse-15.1-amd64: <<: *pkg-rpm-wa needs: [build-opensuse-15.1-amd64] - image: registry.labs.nic.cz/labs/bird:opensuse-15.1-amd64 + image: registry.nic.cz/labs/bird:opensuse-15.1-amd64 pkg-opensuse-15.2-amd64: <<: *pkg-rpm-wa needs: [build-opensuse-15.2-amd64] - image: registry.labs.nic.cz/labs/bird:opensuse-15.2-amd64 + image: registry.nic.cz/labs/bird:opensuse-15.2-amd64 pkg-opensuse-15.3-amd64: <<: *pkg-rpm-wa needs: [build-opensuse-15.3-amd64] - image: registry.labs.nic.cz/labs/bird:opensuse-15.3-amd64 + image: registry.nic.cz/labs/bird:opensuse-15.3-amd64 build-birdlab: @@ -582,6 +583,11 @@ test-ospf-custom: variables: TEST_NAME: cf-ospf-custom +test-ospf-area: + <<: *test-base + variables: + TEST_NAME: cf-ospf-area + test-ospf-vrf: <<: *test-base variables: @@ -636,3 +642,8 @@ test-babel-auth: <<: *test-base variables: TEST_NAME: cf-babel-auth + +test-rip-base: + <<: *test-base + variables: + TEST_NAME: cf-rip-base @@ -1,3 +1,38 @@ +Version 2.0.9 (2022-02-09) + o BGP: Flowspec validation procedure + o Babel: MAC authentication support + o Routing table configuration blocks + o Optional prefix trie in routing table for faster LPM/interval queries + o CLI: New 'show route in <prefix>' command + o Filter: Faster (16-way) prefix sets + o Filter: MPLS label route attribute + o Filter: Operators to pick community components + o Filter: Operators to find minimum and maximum element of lists + o BGP: New 'free bind' option + o BGP: Log route updates that were changed to withdraws + o BGP: Improved 'invalid next hop' error reporting + o OSPF: Allow ifaces with host address as unnumbered PtP or PtMP ifaces + o OSPF: All packets on PtP networks should be sent to AllSPFRouters address + o Scripts for apkg-powered upstream packaging for deb and rpm + o Support for Blake2s and Blake2b hash functions + o Security keys / passwords can be entered in hexadecimal digits + o Memory statistics split into Effective and Overhead + o Linux: New option 'netlink rx buffer' to specify netlink socket buffer size + o BSD: Assume onlink flag on ifaces with only host addresses + o Many bugfixes + + Notes: + + For OSPF on PtP network, BIRD now sends all packets to multicast AllSPFRouters + address (as required in RFC 2328 8.1). This likely breaks setups with multiple + neighbors on a network configured as PtP, which worked in previous versions. + Such links should be configured as PtMP. + + Since Linux 5.3, netlink socket can be flooded by route cache entries during + route table scan. This version mitigates that issue by using strict netlink + filtering. + + Version 2.0.8 (2021-03-18) o Automatic channel reloads based on RPKI changes o Multiple static routes with the same network diff --git a/conf/cf-lex.l b/conf/cf-lex.l index 704a1750..c9d2f5a5 100644 --- a/conf/cf-lex.l +++ b/conf/cf-lex.l @@ -255,7 +255,7 @@ WHITE [ \t] return IP4; } -{XIGIT}{2}(:{XIGIT}{2}|{XIGIT}{2}){15,} { +{XIGIT}{2}((:{XIGIT}{2}){15,}|({XIGIT}{2}){15,}) { char *s = yytext; size_t len = 0, i; struct bytestring *bytes; diff --git a/conf/conf.c b/conf/conf.c index 58abcde1..a2b01667 100644 --- a/conf/conf.c +++ b/conf/conf.c @@ -520,6 +520,9 @@ order_shutdown(int gr) memcpy(c, config, sizeof(struct config)); init_list(&c->protos); init_list(&c->tables); + init_list(&c->symbols); + memset(c->def_tables, 0, sizeof(c->def_tables)); + HASH_INIT(c->sym_hash, c->pool, 4); c->shutdown = 1; c->gr_down = gr; diff --git a/doc/bird.sgml b/doc/bird.sgml index a2138b55..4df1d94f 100644 --- a/doc/bird.sgml +++ b/doc/bird.sgml @@ -14,7 +14,7 @@ configuration - something in config which is not keyword. (set-fill-column 80) - Copyright 1999,2000 Pavel Machek <pavel@ucw.cz>, distribute under GPL version 2 or later. + Copyright 1999 - 2022 CZ.NIC, z.s.p.o , distribute under GPL version 2 or later. --> @@ -23,7 +23,6 @@ configuration - something in config which is not keyword. <title>BIRD 2.0 User's Guide <author> Ondrej Filip <it/<feela@network.cz>/, -Pavel Machek <it/<pavel@ucw.cz>/, Martin Mares <it/<mj@ucw.cz>/, Maria Matejka <it/<mq@jmq.cz>/, Ondrej Zajicek <it/<santiago@crfreenet.org>/ @@ -145,13 +144,6 @@ BIRD executable by configuring out routing protocols you don't use, and <p>You can pass several command-line options to bird: <descrip> - <tag><label id="argv-block">-B <m/exp/</tag> - allocate memory using 2^<cf/exp/ byte sized blocks; - if you're expecting high memory load, raise this to - reduce number of allocated memory pages. For a million routes - in one table, the recommended setting is 18. - Default is your system page size, typically 12 for 4096 bytes. - <tag><label id="argv-config">-c <m/config name/</tag> use given configuration file instead of <it/prefix/<file>/etc/bird.conf</file>. @@ -259,16 +251,9 @@ The global best route selection algorithm is (roughly) as follows: </itemize> <p><label id="dsc-table-sorted">Usually, a routing table just chooses a selected -route from a list of entries for one network. But if the <cf/sorted/ option is -activated, these lists of entries are kept completely sorted (according to -preference or some protocol-dependent metric). This is needed for some features -of some protocols (e.g. <cf/secondary/ option of BGP protocol, which allows to -accept not just a selected route, but the first route (in the sorted list) that -is accepted by filters), but it is incompatible with some other features (e.g. -<cf/deterministic med/ option of BGP protocol, which activates a way of choosing -selected route that cannot be described using comparison and ordering). Minor -advantage is that routes are shown sorted in <cf/show route/, minor disadvantage -is that it is slightly more computationally expensive. +route from a list of entries for one network. Optionally, these lists of entries +are kept completely sorted (according to preference or some protocol-dependent +metric). See <ref id="rtable-sorted" name="sorted"> table option for details. <sect>Routes and network types <label id="routes"> @@ -635,18 +620,73 @@ include "tablename.conf";; <cf/protocol/ times, and the <cf/iso long ms/ format for <cf/base/ and <cf/log/ times. - <tag><label id="opt-table"><m/nettype/ table <m/name/ [sorted]</tag> - Create a new routing table. The default routing tables <cf/master4/ and - <cf/master6/ are created implicitly, other routing tables have to be - added by this command. Option <cf/sorted/ can be used to enable sorting - of routes, see <ref id="dsc-table-sorted" name="sorted table"> - description for details. + <tag><label id="opt-table"><m/nettype/ table <m/name/ [ { <m/option/; [<m/.../] } ]</tag> + Define a new routing table. The default routing tables <cf/master4/ and + <cf/master6/ are defined implicitly, other routing tables have to be + defined by this option. See the <ref id="rtable-opts" + name="routing table configuration section"> for routing table options. <tag><label id="opt-eval">eval <m/expr/</tag> Evaluates given filter expression. It is used by the developers for testing of filters. </descrip> +<sect>Routing table options +<label id="rtable-opts"> + +<p>Most routing tables do not need any options and are defined without an option +block, but there are still some options to tweak routing table behavior. Note +that implicit tables (<cf/master4/ and <cf/master6/) can be redefined in order +to set options. + +<descrip> + <tag><label id="rtable-sorted">sorted <m/switch/</tag> + Usually, a routing table just chooses the selected (best) route from a + list of routes for each network, while keeping remaining routes unsorted. + If enabled, these lists of routes are kept completely sorted (according + to preference or some protocol-dependent metric). + + This is needed for some protocol features (e.g. <cf/secondary/ option of + BGP protocol, which allows to accept not just a selected route, but the + first route (in the sorted list) that is accepted by filters), but it is + incompatible with some other features (e.g. <cf/deterministic med/ + option of BGP protocol, which activates a way of choosing selected route + that cannot be described using comparison and ordering). Minor advantage + is that routes are shown sorted in <cf/show route/, minor disadvantage + is that it is slightly more computationally expensive. Default: off. + + <tag><label id="rtable-trie">trie <m/switch/</tag> + BIRD routing tables are implemented with hash tables, which is efficient + for exact-match lookups, but inconvenient for longest-match lookups or + interval lookups (finding superprefix or subprefixes). This option + activates additional trie structure that is used to accelerate these + lookups, while using the hash table for exact-match lookups. + + This has advantage for <ref id="rpki" name="RPKI"> (on ROA tables), + for <ref id="bgp-gateway" name="recursive next-hops"> (on IGP tables), + and is required for <ref id="bgp-validate" name="flowspec validation"> + (on base IP tables). Another advantage is that interval results (like + from <cf/show route in .../ command) are lexicographically sorted. The + disadvantage is that trie-enabled routing tables require more memory, + which may be an issue especially in multi-table setups. Default: off. + + <tag><label id="rtable-min-settle-time">min settle time <m/time/</tag> + Specify a minimum value of the settle time. When a ROA table changes, + automatic <ref id="proto-rpki-reload" name="RPKI reload"> may be + triggered, after a short settle time. Minimum settle time is a delay + from the last ROA table change to wait for more updates. Default: 1 s. + + + <tag><label id="rtable-max-settle-time">max settle time <m/time/</tag> + Specify a maximum value of the settle time. When a ROA table changes, + automatic <ref id="proto-rpki-reload" name="RPKI reload"> may be + triggered, after a short settle time. Maximum settle time is an upper + limit to the settle time from the initial ROA table change even if + there are consecutive updates gradually renewing the settle time. + Default: 20 s. +</descrip> + + <sect>Protocol options <label id="protocol-opts"> @@ -1299,6 +1339,9 @@ in the foot). The same syntax can also be used to construct a pair from two arbitrary integer expressions (for example <cf/(1+2,a)/). + Operators <cf/.asn/ and <cf/.data/ can be used to extract corresponding + components of a pair: <cf>(<m/asn/, <m/data/)</cf>. + <tag><label id="type-quad">quad</tag> This is a dotted quad of numbers used to represent router IDs (and others). Each component can have a value from 0 to 255. Literals of @@ -1389,6 +1432,10 @@ in the foot). pairs, LCs can be constructed using expressions for its parts, (e.g. <cf/(myas, 10+20, 3*10)/, where <cf/myas/ is an integer variable). + Operators <cf/.asn/, <cf/.data1/, and <cf/.data2/ can be used + to extract corresponding components of LCs: + <cf>(<m/asn/, <m/data1/, <m/data2/)</cf>. + <tag><label id="type-set">int|pair|quad|ip|prefix|ec|lc|enum set</tag> Filters recognize four types of sets. Sets are similar to strings: you can pass them around but you can't modify them. Literals of type <cf>int @@ -1532,7 +1579,7 @@ in the foot). Clist is similar to a set, except that unlike other sets, it can be modified. The type is used for community list (a set of pairs) and for cluster list (a set of quads). There exist no literals of this type. - There are three special operators on clists: + There are special operators on clists: <cf><m/C/.len</cf> returns the length of clist <m/C/. @@ -1559,6 +1606,15 @@ in the foot). <cf><m/C/.add(<m/P/);</cf> if <m/C/ is appropriate route attribute (for example <cf/bgp_community/). Similarly for <cf/delete/ and <cf/filter/. + <cf><m/C/.min</cf> returns the minimum element of clist <m/C/. + + <cf><m/C/.max</cf> returns the maximum element of clist <m/C/. + + Operators <cf/.min/, <cf/.max/ can be used together with <cf/filter/ + to extract the community from the specific subset of communities + (e.g. localpref or prepend) without the need to check every possible + value (e.g. <cf/filter(bgp_community, [(23456, 1000..1099)]).min/). + <tag><label id="type-eclist">eclist</tag> Eclist is a data type used for BGP extended community lists. Eclists are very similar to clists, but they are sets of ECs instead of pairs. @@ -2281,6 +2337,7 @@ avoid routing loops. <item> <rfc id="8092"> - BGP Large Communities Attribute <item> <rfc id="8203"> - BGP Administrative Shutdown Communication <item> <rfc id="8212"> - Default EBGP Route Propagation Behavior without Policies +<item> <rfc id="9117"> - Revised Validation Procedure for BGP Flow Specifications </itemize> <sect1>Route selection rules @@ -2403,6 +2460,12 @@ using the following configuration parameters: same address family and using the same local port) should have set <cf/strict bind/, or none of them. Default: disabled. + <tag><label id="bgp-free-bind">free bind <m/switch/</tag> + Use IP_FREEBIND socket option for the listening socket, which allows + binding to an IP address not (yet) assigned to an interface. Note that + all BGP instances that share a listening socket should have the same + value of the <cf/freebind/ option. Default: disabled. + <tag><label id="bgp-check-link">check link <M>switch</M></tag> BGP could use hardware link state into consideration. If enabled, BIRD tracks the link state of the associated interface and when link @@ -2614,13 +2677,6 @@ using the following configuration parameters: disabled. Default: on, with automatic fallback to off when received capability-related error. - <tag><label id="bgp-advertise-ipv4">advertise ipv4 <m/switch/</tag> - Advertise IPv4 multiprotocol capability. This is not a correct behavior - according to the strict interpretation of <rfc id="4760">, but it is - widespread and required by some BGP implementations (Cisco and Quagga). - This option is relevant to IPv4 mode with enabled capability - advertisement only. Default: on. - <tag><label id="bgp-advertise-hostname">advertise hostname <m/switch/</tag> Advertise hostname capability along with the hostname. Default: off. @@ -2666,7 +2722,7 @@ using the following configuration parameters: <tag><label id="bgp-error-wait-time">error wait time <m/number/,<m/number/</tag> Minimum and maximum delay in seconds between a protocol failure (either - local or reported by the peer) and automatic restart. Doesn't apply + local or reported by the peer) and automatic restart. Doesn not apply when <cf/disable after error/ is configured. If consecutive errors happen, the delay is increased exponentially until it reaches the maximum. Default: 60, 300. @@ -2844,6 +2900,31 @@ be used in explicit configuration. explicitly (to conserve memory). This option requires that the connected routing table is <ref id="dsc-table-sorted" name="sorted">. Default: off. + <tag><label id="bgp-validate">validate <m/switch/</tag> + Apply flowspec validation procedure as described in <rfc id="8955"> + section 6 and <rfc id="9117">. The Validation procedure enforces that + only routers in the forwarding path for a network can originate flowspec + rules for that network. The validation procedure should be used for EBGP + to prevent injection of malicious flowspec rules from outside, but it + should also be used for IBGP to ensure that selected flowspec rules are + consistent with selected IP routes. The validation procedure uses an IP + routing table (<ref id="bgp-base-table" name="base table">, see below) + against which flowspec rules are validated. This option is limited to + flowspec channels. Default: off (for compatibility reasons). + + Note that currently the flowspec validation does not work reliably + together with <ref id="bgp-import-table" name="import table"> option + enabled on flowspec channels. + + <tag><label id="bgp-base-table">base table <m/name/</tag> + Specifies an IP table used for the flowspec validation procedure. The + table must have enabled <cf/trie/ option, otherwise the validation + procedure would not work. The type of the table must be <cf/ipv4/ for + <cf/flow4/ channels and <cf/ipv6/ for <cf/flow6/ channels. This option + is limited to flowspec channels. Default: the main table of the + <cf/ipv4/ / <cf/ipv6/ channel of the same BGP instance, or the + <cf/master4/ / <cf/master6/ table if there is no such channel. + <tag><label id="bgp-extended-next-hop">extended next hop <m/switch/</tag> BGP expects that announced next hops have the same address family as associated network prefixes. This option provides an extension to use @@ -3240,6 +3321,12 @@ channels. allows to specify a limit on maximal number of nexthops in one route. By default, multipath merging is disabled. If enabled, default value of the limit is 16. + + <tag><label id="krt-netlink-rx-buffer">netlink rx buffer <m/number/</tag> (Linux) + Set kernel receive buffer size (in bytes) for the netlink socket. The default + value is OS-dependent (from the <file>/proc/sys/net/core/rmem_default</file> + file), If you get some "Kernel dropped some netlink message ..." warnings, + you may increase this value. </descrip> <sect1>Attributes @@ -5348,15 +5435,15 @@ name="atrey.karlin.mff.cuni.cz:/pub/rfc">). </book> <!-- -LocalWords: GPL IPv GateD BGPv RIPv OSPFv Linux sgml html dvi sgmltools Pavel +LocalWords: GPL IPv GateD BGPv RIPv OSPFv Linux sgml html dvi sgmltools LocalWords: linuxdoc dtd descrip config conf syslog stderr auth ospf bgp Mbps LocalWords: router's eval expr num birdc ctl UNIX if's enums bool int ip GCC LocalWords: len ipaddress pxlen netmask enum bgppath bgpmask clist gw md eth -LocalWords: RTS printn quitbird iBGP AS'es eBGP RFC multiprotocol IGP Machek +LocalWords: RTS printn quitbird iBGP AS'es eBGP RFC multiprotocol IGP LocalWords: EGP misconfigurations keepalive pref aggr aggregator BIRD's RTC LocalWords: OS'es AS's multicast nolisten misconfigured UID blackhole MRTD MTU LocalWords: uninstalls ethernets IP binutils ANYCAST anycast dest RTD ICMP rfc LocalWords: compat multicasts nonbroadcast pointopoint loopback sym stats LocalWords: Perl SIGHUP dd mm yy HH MM SS EXT IA UNICAST multihop Discriminator txt -LocalWords: proto wildcard Ondrej Filip +LocalWords: proto wildcard --> diff --git a/doc/prog-head.sgml b/doc/prog-head.sgml index 0eec367e..4daca3a3 100644 --- a/doc/prog-head.sgml +++ b/doc/prog-head.sgml @@ -12,9 +12,9 @@ <title>BIRD Programmer's Documentation <author> Ondrej Filip <it/<feela@network.cz>/, -Pavel Machek <it/<pavel@ucw.cz>/, Martin Mares <it/<mj@ucw.cz>/, Ondrej Zajicek <it/<santiago@crfreenet.org>/ +Maria Matejka <it/<mq@jmq.cz>/, </author> <abstract> diff --git a/filter/config.Y b/filter/config.Y index 8034b790..15b77761 100644 --- a/filter/config.Y +++ b/filter/config.Y @@ -279,9 +279,11 @@ CF_KEYWORDS(FUNCTION, PRINT, PRINTN, UNSET, RETURN, ROA_CHECK, ASN, SRC, DST, IS_V4, IS_V6, LEN, MAXLEN, + DATA, DATA1, DATA2, DEFINED, ADD, DELETE, CONTAINS, RESET, PREPEND, FIRST, LAST, LAST_NONAGGREGATED, MATCH, + MIN, MAX, EMPTY, FILTER, WHERE, EVAL, ATTRIBUTE, BT_ASSERT, BT_TEST_SUITE, BT_CHECK_ASSIGN, BT_TEST_SAME, FORMAT, STACKS) @@ -790,13 +792,18 @@ term: | term '.' RD { $$ = f_new_inst(FI_ROUTE_DISTINGUISHER, $1); } | term '.' LEN { $$ = f_new_inst(FI_LENGTH, $1); } | term '.' MAXLEN { $$ = f_new_inst(FI_ROA_MAXLEN, $1); } - | term '.' ASN { $$ = f_new_inst(FI_ROA_ASN, $1); } + | term '.' ASN { $$ = f_new_inst(FI_ASN, $1); } | term '.' SRC { $$ = f_new_inst(FI_NET_SRC, $1); } | term '.' DST { $$ = f_new_inst(FI_NET_DST, $1); } | term '.' MASK '(' term ')' { $$ = f_new_inst(FI_IP_MASK, $1, $5); } | term '.' FIRST { $$ = f_new_inst(FI_AS_PATH_FIRST, $1); } | term '.' LAST { $$ = f_new_inst(FI_AS_PATH_LAST, $1); } | term '.' LAST_NONAGGREGATED { $$ = f_new_inst(FI_AS_PATH_LAST_NAG, $1); } + | term '.' DATA { $$ = f_new_inst(FI_PAIR_DATA, $1); } + | term '.' DATA1 { $$ = f_new_inst(FI_LC_DATA1, $1); } + | term '.' DATA2 { $$ = f_new_inst(FI_LC_DATA2, $1); } + | term '.' MIN { $$ = f_new_inst(FI_MIN, $1); } + | term '.' MAX { $$ = f_new_inst(FI_MAX, $1); } /* Communities */ /* This causes one shift/reduce conflict diff --git a/filter/data.c b/filter/data.c index 7c33d2cb..56c1fb17 100644 --- a/filter/data.c +++ b/filter/data.c @@ -68,6 +68,17 @@ f_type_name(enum f_type t) return "?"; } +enum f_type +f_type_element_type(enum f_type t) +{ + switch(t) { + case T_CLIST: return T_PAIR; + case T_ECLIST: return T_EC; + case T_LCLIST: return T_LC; + default: return T_VOID; + }; +} + const struct f_val f_const_empty_path = { .type = T_PATH, .val.ad = &null_adata, diff --git a/filter/data.h b/filter/data.h index 45246f9f..6d588204 100644 --- a/filter/data.h +++ b/filter/data.h @@ -141,18 +141,23 @@ struct f_tree { void *data; }; +#define TRIE_STEP 4 +#define TRIE_STACK_LENGTH 33 + struct f_trie_node4 { ip4_addr addr, mask, accept; - uint plen; - struct f_trie_node4 *c[2]; + u16 plen; + u16 local; + struct f_trie_node4 *c[1 << TRIE_STEP]; }; struct f_trie_node6 { ip6_addr addr, mask, accept; - uint plen; - struct f_trie_node6 *c[2]; + u16 plen; + u16 local; + struct f_trie_node6 *c[1 << TRIE_STEP]; }; struct f_trie_node @@ -169,9 +174,20 @@ struct f_trie u8 zero; s8 ipv4; /* -1 for undefined / empty */ u16 data_size; /* Additional data for each trie node */ + u32 prefix_count; /* Works only for restricted tries (pxlen == l == h) */ struct f_trie_node root; /* Root trie node */ }; +struct f_trie_walk_state +{ + u8 ipv4; + u8 accept_length; /* Current inter-node prefix position */ + u8 start_pos; /* Initial prefix position in stack[0] */ + u8 local_pos; /* Current intra-node prefix position */ + u8 stack_pos; /* Current node in stack below */ + const struct f_trie_node *stack[TRIE_STACK_LENGTH]; +}; + struct f_tree *f_new_tree(void); struct f_tree *build_tree(struct f_tree *); const struct f_tree *find_tree(const struct f_tree *t, const struct f_val *val); @@ -182,13 +198,76 @@ void tree_walk(const struct f_tree *t, void (*hook)(const struct f_tree *, void struct f_trie *f_new_trie(linpool *lp, uint data_size); void *trie_add_prefix(struct f_trie *t, const net_addr *n, uint l, uint h); int trie_match_net(const struct f_trie *t, const net_addr *n); +int trie_match_longest_ip4(const struct f_trie *t, const net_addr_ip4 *net, net_addr_ip4 *dst, ip4_addr *found0); +int trie_match_longest_ip6(const struct f_trie *t, const net_addr_ip6 *net, net_addr_ip6 *dst, ip6_addr *found0); +void trie_walk_init(struct f_trie_walk_state *s, const struct f_trie *t, const net_addr *from); +int trie_walk_next(struct f_trie_walk_state *s, net_addr *net); int trie_same(const struct f_trie *t1, const struct f_trie *t2); void trie_format(const struct f_trie *t, buffer *buf); +static inline int +trie_match_next_longest_ip4(net_addr_ip4 *n, ip4_addr *found) +{ + while (n->pxlen) + { + n->pxlen--; + ip4_clrbit(&n->prefix, n->pxlen); + + if (ip4_getbit(*found, n->pxlen)) + return 1; + } + + return 0; +} + +static inline int +trie_match_next_longest_ip6(net_addr_ip6 *n, ip6_addr *found) +{ + while (n->pxlen) + { + n->pxlen--; + ip6_clrbit(&n->prefix, n->pxlen); + + if (ip6_getbit(*found, n->pxlen)) + return 1; + } + + return 0; +} + + +#define TRIE_WALK_TO_ROOT_IP4(trie, net, dst) ({ \ + net_addr_ip4 dst; \ + ip4_addr _found; \ + for (int _n = trie_match_longest_ip4(trie, net, &dst, &_found); \ + _n; \ + _n = trie_match_next_longest_ip4(&dst, &_found)) + +#define TRIE_WALK_TO_ROOT_IP6(trie, net, dst) ({ \ + net_addr_ip6 dst; \ + ip6_addr _found; \ + for (int _n = trie_match_longest_ip6(trie, net, &dst, &_found); \ + _n; \ + _n = trie_match_next_longest_ip6(&dst, &_found)) + +#define TRIE_WALK_TO_ROOT_END }) + + +#define TRIE_WALK(trie, net, from) ({ \ + net_addr net; \ + struct f_trie_walk_state tws_; \ + trie_walk_init(&tws_, trie, from); \ + while (trie_walk_next(&tws_, &net)) + +#define TRIE_WALK_END }) + + #define F_CMP_ERROR 999 const char *f_type_name(enum f_type t); +enum f_type f_type_element_type(enum f_type t); + int val_same(const struct f_val *v1, const struct f_val *v2); int val_compare(const struct f_val *v1, const struct f_val *v2); void val_format(const struct f_val *v, buffer *buf); diff --git a/filter/f-inst.c b/filter/f-inst.c index 00e22383..532c6989 100644 --- a/filter/f-inst.c +++ b/filter/f-inst.c @@ -901,14 +901,31 @@ ((net_addr_roa6 *) v1.val.net)->max_pxlen); } - INST(FI_ROA_ASN, 1, 1) { /* Get ROA ASN */ - ARG(1, T_NET); - if (!net_is_roa(v1.val.net)) - runtime( "ROA expected" ); + INST(FI_ASN, 1, 1) { /* Get ROA ASN or community ASN part */ + ARG_ANY(1); + RESULT_TYPE(T_INT); + switch(v1.type) + { + case T_NET: + if (!net_is_roa(v1.val.net)) + runtime( "ROA expected" ); - RESULT(T_INT, i, (v1.val.net->type == NET_ROA4) ? - ((net_addr_roa4 *) v1.val.net)->asn : - ((net_addr_roa6 *) v1.val.net)->asn); + RESULT_(T_INT, i, (v1.val.net->type == NET_ROA4) ? + ((net_addr_roa4 *) v1.val.net)->asn : + ((net_addr_roa6 *) v1.val.net)->asn); + break; + + case T_PAIR: + RESULT_(T_INT, i, v1.val.i >> 16); + break; + + case T_LC: + RESULT_(T_INT, i, v1.val.lc.asn); + break; + + default: + runtime( "Net, pair or lc expected" ); + } } INST(FI_IP, 1, 1) { /* Convert prefix to ... */ @@ -942,6 +959,89 @@ RESULT(T_INT, i, as_path_get_last_nonaggregated(v1.val.ad)); } + INST(FI_PAIR_DATA, 1, 1) { /* Get data part from the standard community */ + ARG(1, T_PAIR); + RESULT(T_INT, i, v1.val.i & 0xFFFF); + } + + INST(FI_LC_DATA1, 1, 1) { /* Get data1 part from the large community */ + ARG(1, T_LC); + RESULT(T_INT, i, v1.val.lc.ldp1); + } + + INST(FI_LC_DATA2, 1, 1) { /* Get data2 part from the large community */ + ARG(1, T_LC); + RESULT(T_INT, i, v1.val.lc.ldp2); + } + + INST(FI_MIN, 1, 1) { /* Get minimum element from set */ + ARG_ANY(1); + RESULT_TYPE(f_type_element_type(v1.type)); + switch(v1.type) + { + case T_CLIST: + { + u32 val = 0; + int_set_min(v1.val.ad, &val); + RESULT_(T_PAIR, i, val); + } + break; + + case T_ECLIST: + { + u64 val = 0; + ec_set_min(v1.val.ad, &val); + RESULT_(T_EC, ec, val); + } + break; + + case T_LCLIST: + { + lcomm val = { 0, 0, 0 }; + lc_set_min(v1.val.ad, &val); + RESULT_(T_LC, lc, val); + } + break; + + default: + runtime( "Clist or lclist expected" ); + } + } + + INST(FI_MAX, 1, 1) { /* Get maximum element from set */ + ARG_ANY(1); + RESULT_TYPE(f_type_element_type(v1.type)); + switch(v1.type) + { + case T_CLIST: + { + u32 val = 0; + int_set_max(v1.val.ad, &val); + RESULT_(T_PAIR, i, val); + } + break; + + case T_ECLIST: + { + u64 val = 0; + ec_set_max(v1.val.ad, &val); + RESULT_(T_EC, ec, val); + } + break; + + case T_LCLIST: + { + lcomm val = { 0, 0, 0 }; + lc_set_max(v1.val.ad, &val); + RESULT_(T_LC, lc, val); + } + break; + + default: + runtime( "Clist or lclist expected" ); + } + } + INST(FI_RETURN, 1, 1) { NEVER_CONSTANT; /* Acquire the return value */ diff --git a/filter/filter_test.c b/filter/filter_test.c index 7e4af092..e8e8b747 100644 --- a/filter/filter_test.c +++ b/filter/filter_test.c @@ -46,9 +46,7 @@ run_function(const void *arg) if (t->cmp) return t->result == f_same(t->fn, t->cmp); - linpool *tmp = lp_new_default(&root_pool); - enum filter_return fret = f_eval(t->fn, tmp, NULL); - rfree(tmp); + enum filter_return fret = f_eval(t->fn, tmp_linpool, NULL); return (fret < F_REJECT); } diff --git a/filter/test.conf b/filter/test.conf index 93e7a770..2a5a2d98 100644 --- a/filter/test.conf +++ b/filter/test.conf @@ -335,6 +335,26 @@ ip p; p = 1234:5678::; bt_assert(!p.is_v4); bt_assert(p.mask(24) = 1234:5600::); + + p = 1:2:3:4:5:6:7:8; + bt_assert(!p.is_v4); + bt_assert(format(p) = "1:2:3:4:5:6:7:8"); + bt_assert(p.mask(64) = 1:2:3:4::); + + p = 10:20:30:40:50:60:70:80; + bt_assert(!p.is_v4); + bt_assert(format(p) = "10:20:30:40:50:60:70:80"); + bt_assert(p.mask(64) = 10:20:30:40::); + + p = 1090:20a0:30b0:40c0:50d0:60e0:70f0:8000; + bt_assert(!p.is_v4); + bt_assert(format(p) = "1090:20a0:30b0:40c0:50d0:60e0:70f0:8000"); + bt_assert(p.mask(64) = 1090:20a0:30b0:40c0::); + + p = ::fffe:6:c0c:936d:88c7:35d3; + bt_assert(!p.is_v4); + bt_assert(format(p) = "::fffe:6:c0c:936d:88c7:35d3"); + bt_assert(p.mask(64) = 0:0:fffe:6::); } bt_test_suite(t_ip, "Testing ip address"); @@ -478,6 +498,33 @@ prefix set pxs; bt_assert(1.2.0.0/16 ~ [ 1.0.0.0/8{ 15 , 17 } ]); bt_assert([ 10.0.0.0/8{ 15 , 17 } ] != [ 11.0.0.0/8{ 15 , 17 } ]); + + /* Formatting of prefix sets, some cases are a bit strange */ + bt_assert(format([ 0.0.0.0/0 ]) = "[0.0.0.0/0]"); + bt_assert(format([ 10.10.0.0/32 ]) = "[10.10.0.0/32{0.0.0.1}]"); + bt_assert(format([ 10.10.0.0/17 ]) = "[10.10.0.0/17{0.0.128.0}]"); + bt_assert(format([ 10.10.0.0/17{17,19} ]) = "[10.10.0.0/17{0.0.224.0}]"); # 224 = 128+64+32 + bt_assert(format([ 10.10.128.0/17{18,19} ]) = "[10.10.128.0/18{0.0.96.0}, 10.10.192.0/18{0.0.96.0}]"); # 96 = 64+32 + bt_assert(format([ 10.10.64.0/18- ]) = "[0.0.0.0/0, 0.0.0.0/1{128.0.0.0}, 0.0.0.0/2{64.0.0.0}, 0.0.0.0/3{32.0.0.0}, 10.10.0.0/16{255.255.0.0}, 10.10.0.0/17{0.0.128.0}, 10.10.64.0/18{0.0.64.0}]"); + bt_assert(format([ 10.10.64.0/18+ ]) = "[10.10.64.0/18{0.0.96.0}, 10.10.64.0/20{0.0.31.255}, 10.10.80.0/20{0.0.31.255}, 10.10.96.0/20{0.0.31.255}, 10.10.112.0/20{0.0.31.255}]"); + + bt_assert(format([ 10.10.160.0/19 ]) = "[10.10.160.0/19{0.0.32.0}]"); + bt_assert(format([ 10.10.160.0/19{19,22} ]) = "[10.10.160.0/19{0.0.32.0}, 10.10.160.0/20{0.0.28.0}, 10.10.176.0/20{0.0.28.0}]"); # 28 = 16+8+4 + bt_assert(format([ 10.10.160.0/19+ ]) = "[10.10.160.0/19{0.0.32.0}, 10.10.160.0/20{0.0.31.255}, 10.10.176.0/20{0.0.31.255}]"); + + bt_assert(format([ ::/0 ]) = "[::/0]"); + bt_assert(format([ 11:22:33:44:55:66:77:88/128 ]) = "[11:22:33:44:55:66:77:88/128{::1}]"); + bt_assert(format([ 11:22:33:44::/64 ]) = "[11:22:33:44::/64{0:0:0:1::}]"); + bt_assert(format([ 11:22:33:44::/64+ ]) = "[11:22:33:44::/64{::1:ffff:ffff:ffff:ffff}]"); + + bt_assert(format([ 11:22:33:44::/65 ]) = "[11:22:33:44::/65{::8000:0:0:0}]"); + bt_assert(format([ 11:22:33:44::/65{65,67} ]) = "[11:22:33:44::/65{::e000:0:0:0}]"); # e = 8+4+2 + bt_assert(format([ 11:22:33:44:8000::/65{66,67} ]) = "[11:22:33:44:8000::/66{::6000:0:0:0}, 11:22:33:44:c000::/66{::6000:0:0:0}]"); # 6 = 4+2 + bt_assert(format([ 11:22:33:44:4000::/66- ]) = "[::/0, ::/1{8000::}, ::/2{4000::}, ::/3{2000::}, 11:22:33:44::/64{ffff:ffff:ffff:ffff::}, 11:22:33:44::/65{::8000:0:0:0}, 11:22:33:44:4000::/66{::4000:0:0:0}]"); + bt_assert(format([ 11:22:33:44:4000::/66+ ]) = "[11:22:33:44:4000::/66{::6000:0:0:0}, 11:22:33:44:4000::/68{::1fff:ffff:ffff:ffff}, 11:22:33:44:5000::/68{::1fff:ffff:ffff:ffff}, 11:22:33:44:6000::/68{::1fff:ffff:ffff:ffff}, 11:22:33:44:7000::/68{::1fff:ffff:ffff:ffff}]"); + bt_assert(format([ 11:22:33:44:c000::/67 ]) = "[11:22:33:44:c000::/67{::2000:0:0:0}]"); + bt_assert(format([ 11:22:33:44:c000::/67{67,71} ]) = "[11:22:33:44:c000::/67{::2000:0:0:0}, 11:22:33:44:c000::/68{::1e00:0:0:0}, 11:22:33:44:d000::/68{::1e00:0:0:0}]"); + bt_assert(format([ 11:22:33:44:c000::/67+ ]) = "[11:22:33:44:c000::/67{::2000:0:0:0}, 11:22:33:44:c000::/68{::1fff:ffff:ffff:ffff}, 11:22:33:44:d000::/68{::1fff:ffff:ffff:ffff}]"); } bt_test_suite(t_prefix_set, "Testing prefix sets"); @@ -557,6 +604,12 @@ prefix set pxs; bt_assert(2000::/29 !~ pxs); bt_assert(1100::/10 !~ pxs); bt_assert(2010::/26 !~ pxs); + + pxs = [ 52E0::/13{13,128} ]; + bt_assert(52E7:BE81:379B:E6FD:541F:B0D0::/93 ~ pxs); + + pxs = [ 41D8:8718::/30{0,30}, 413A:99A8:6C00::/38{38,128} ]; + bt_assert(4180::/9 ~ pxs); } bt_test_suite(t_prefix6_set, "Testing prefix IPv6 sets"); @@ -683,6 +736,11 @@ clist l; clist l2; clist r; { + bt_assert((10, 20).asn = 10); + bt_assert((10, 20).data = 20); + bt_assert(p23.asn = 2); + bt_assert(p23.data = 3); + l = - empty -; bt_assert(l !~ [(*,*)]); bt_assert((l ~ [(*,*)]) != (l !~ [(*,*)])); @@ -775,6 +833,12 @@ clist r; r = filter(l, [(3,1), (*,2)]); bt_assert(r = add(add(-empty-, (3,1)), (3,2))); bt_assert(format(r) = "(clist (3,1) (3,2))"); + + # minimim & maximum element + r = add(add(add(add(add(-empty-, (2,1)), (1,3)), (2,2)), (3,1)), (2,3)); + bt_assert(format(r) = "(clist (2,1) (1,3) (2,2) (3,1) (2,3))"); + bt_assert(r.min = (1,3)); + bt_assert(r.max = (3,1)); } bt_test_suite(t_clist, "Testing lists of communities"); @@ -880,6 +944,12 @@ eclist r; r = filter(el, [(rt, 10, 1), (rt, 10, 25..30), (ro, 10, 40)]); bt_assert(r = add(add(--empty--, (rt, 10, 1)), (rt, 10, 30))); bt_assert(format(r) = "(eclist (rt, 10, 1) (rt, 10, 30))"); + + # minimim & maximum element + r = add(add(add(add(add(--empty--, (rt, 2, 1)), (rt, 1, 3)), (rt, 2, 2)), (rt, 3, 1)), (rt, 2, 3)); + bt_assert(format(r) = "(eclist (rt, 2, 1) (rt, 1, 3) (rt, 2, 2) (rt, 3, 1) (rt, 2, 3))"); + bt_assert(r.min = (rt, 1, 3)); + bt_assert(r.max = (rt, 3, 1)); } bt_test_suite(t_eclist, "Testing lists of extended communities"); @@ -939,6 +1009,10 @@ lclist r; bt_assert(---empty--- = ---empty---); bt_assert((10, 20, 30) !~ ---empty---); + bt_assert((10, 20, 30).asn = 10); + bt_assert((10, 20, 30).data1 = 20); + bt_assert((10, 20, 30).data2 = 30); + ll = --- empty ---; ll = add(ll, (ten, 20, 30)); ll = add(ll, (1000, 2000, 3000)); @@ -985,6 +1059,12 @@ lclist r; r = filter(ll, [(5..15, *, *), (20, 15..25, *)]); bt_assert(r = add(add(---empty---, (10, 10, 10)), (20, 20, 20))); bt_assert(format(r) = "(lclist (10, 10, 10) (20, 20, 20))"); + + # minimim & maximum element + r = add(add(add(add(add(---empty---, (2, 3, 3)), (1, 2, 3)), (2, 3, 1)), (3, 1, 2)), (2, 1, 3)); + bt_assert(format(r) = "(lclist (2, 3, 3) (1, 2, 3) (2, 3, 1) (3, 1, 2) (2, 1, 3))"); + bt_assert(r.min = (1, 2, 3)); + bt_assert(r.max = (3, 1, 2)); } bt_test_suite(t_lclist, "Testing lists of large communities"); diff --git a/filter/tree_test.c b/filter/tree_test.c index 6472d17e..05702f81 100644 --- a/filter/tree_test.c +++ b/filter/tree_test.c @@ -19,10 +19,7 @@ static void start_conf_env(void) { bt_bird_init(); - - pool *p = rp_new(&root_pool, "helper_pool"); - linpool *l = lp_new_default(p); - cfg_mem = l; + cfg_mem = tmp_linpool; } static struct f_tree * diff --git a/filter/trie.c b/filter/trie.c index 1a4e1ac3..12ba0b82 100644 --- a/filter/trie.c +++ b/filter/trie.c @@ -1,7 +1,8 @@ /* * Filters: Trie for prefix sets * - * Copyright 2009 Ondrej Zajicek <santiago@crfreenet.org> + * (c) 2009--2021 Ondrej Zajicek <santiago@crfreenet.org> + * (c) 2009--2021 CZ.NIC z.s.p.o. * * Can be freely distributed and used under the terms of the GNU GPL. */ @@ -9,53 +10,68 @@ /** * DOC: Trie for prefix sets * - * We use a (compressed) trie to represent prefix sets. Every node - * in the trie represents one prefix (&addr/&plen) and &plen also - * indicates the index of the bit in the address that is used to - * branch at the node. If we need to represent just a set of - * prefixes, it would be simple, but we have to represent a - * set of prefix patterns. Each prefix pattern consists of - * &ppaddr/&pplen and two integers: &low and &high, and a prefix - * &paddr/&plen matches that pattern if the first MIN(&plen, &pplen) - * bits of &paddr and &ppaddr are the same and &low <= &plen <= &high. - * - * We use a bitmask (&accept) to represent accepted prefix lengths - * at a node. As there are 33 prefix lengths (0..32 for IPv4), but - * there is just one prefix of zero length in the whole trie so we - * have &zero flag in &f_trie (indicating whether the trie accepts - * prefix 0.0.0.0/0) as a special case, and &accept bitmask + * We use a (compressed) trie to represent prefix sets. Every node in the trie + * represents one prefix (&addr/&plen) and &plen also indicates the index of + * bits in the address that are used to branch at the node. Note that such + * prefix is not necessary a member of the prefix set, it is just a canonical + * prefix associated with a node. Prefix lengths of nodes are aligned to + * multiples of &TRIE_STEP (4) and there is 16-way branching in each + * node. Therefore, we say that a node is associated with a range of prefix + * lengths (&plen .. &plen + TRIE_STEP - 1). + * + * The prefix set is not just a set of prefixes, it is defined by a set of + * prefix patterns. Each prefix pattern consists of &ppaddr/&pplen and two + * integers: &low and &high. The tested prefix &paddr/&plen matches that pattern + * if the first MIN(&plen, &pplen) bits of &paddr and &ppaddr are the same and + * &low <= &plen <= &high. + * + * There are two ways to represent accepted prefixes for a node. First, there is + * a bitmask &local, which represents independently all 15 prefixes that extend + * the canonical prefix of the node and are within a range of prefix lengths + * associated with the node. E.g., for node 10.0.0.0/8 they are 10.0.0.0/8, + * 10.0.0.0/9, 10.128.0.0/9, .. 10.224.0.0/11. This order (first by length, then + * lexicographically) is used for indexing the bitmask &local, starting at + * position 1. I.e., index is 2^(plen - base) + offset within the same length, + * see function trie_local_mask6() for details. + * + * Second, we use a bitmask &accept to represent accepted prefix lengths at a + * node. The bit is set means that all prefixes of given length that are either + * subprefixes or superprefixes of the canonical prefix are accepted. As there + * are 33 prefix lengths (0..32 for IPv4), but there is just one prefix of zero + * length in the whole trie so we have &zero flag in &f_trie (indicating whether + * the trie accepts prefix 0.0.0.0/0) as a special case, and &accept bitmask * represents accepted prefix lengths from 1 to 32. * - * There are two cases in prefix matching - a match when the length - * of the prefix is smaller that the length of the prefix pattern, - * (&plen < &pplen) and otherwise. The second case is simple - we - * just walk through the trie and look at every visited node - * whether that prefix accepts our prefix length (&plen). The - * first case is tricky - we don't want to examine every descendant - * of a final node, so (when we create the trie) we have to propagate - * that information from nodes to their ascendants. - * - * Suppose that we have two masks (M1 and M2) for a node. Mask M1 - * represents accepted prefix lengths by just the node and mask M2 - * represents accepted prefix lengths by the node or any of its - * descendants. Therefore M2 is a bitwise or of M1 and children's - * M2 and this is a maintained invariant during trie building. - * Basically, when we want to match a prefix, we walk through the trie, - * check mask M1 for our prefix length and when we came to - * final node, we check mask M2. - * - * There are two differences in the real implementation. First, - * we use a compressed trie so there is a case that we skip our - * final node (if it is not in the trie) and we came to node that - * is either extension of our prefix, or completely out of path - * In the first case, we also have to check M2. - * - * Second, we really need not to maintain two separate bitmasks. - * Checks for mask M1 are always larger than &applen and we need - * just the first &pplen bits of mask M2 (if trie compression - * hadn't been used it would suffice to know just $applen-th bit), - * so we have to store them together in &accept mask - the first - * &pplen bits of mask M2 and then mask M1. + * One complication is handling of prefix patterns with unaligned prefix length. + * When such pattern is to be added, we add a primary node above (with rounded + * down prefix length &nlen) and a set of secondary nodes below (with rounded up + * prefix lengths &slen). Accepted prefix lengths of the original prefix pattern + * are then represented in different places based on their lengths. For prefixes + * shorter than &nlen, it is &accept bitmask of the primary node, for prefixes + * between &nlen and &slen - 1 it is &local bitmask of the primary node, and for + * prefixes longer of equal &slen it is &accept bitmasks of secondary nodes. + * + * There are two cases in prefix matching - a match when the length of the + * prefix is smaller that the length of the prefix pattern, (&plen < &pplen) and + * otherwise. The second case is simple - we just walk through the trie and look + * at every visited node whether that prefix accepts our prefix length (&plen). + * The first case is tricky - we do not want to examine every descendant of a + * final node, so (when we create the trie) we have to propagate that + * information from nodes to their ascendants. + * + * There are two kinds of propagations - propagation from child's &accept + * bitmask to parent's &accept bitmask, and propagation from child's &accept + * bitmask to parent's &local bitmask. The first kind is simple - as all + * superprefixes of a parent are also all superprefixes of appropriate length of + * a child, then we can just add (by bitwise or) a child &accept mask masked by + * parent prefix length mask to the parent &accept mask. This handles prefixes + * shorter than node &plen. + * + * The second kind of propagation is necessary to handle superprefixes of a + * child that are represented by parent &local mask - that are in the range of + * prefix lengths associated with the parent. For each accepted (by child + * &accept mask) prefix length from that range, we need to set appropriate bit + * in &local mask. See function trie_amask_to_local() for details. * * There are four cases when we walk through a trie: * @@ -65,8 +81,32 @@ * - we are beyond the end of path (node length > &plen) * - we are still on path and keep walking (node length < &plen) * - * The walking code in trie_match_prefix() is structured according to - * these cases. + * The walking code in trie_match_net() is structured according to these cases. + * + * Iteration over prefixes in a trie can be done using TRIE_WALK() macro, or + * directly using trie_walk_init() and trie_walk_next() functions. The second + * approach allows suspending the iteration and continuing in it later. + * Prefixes are enumerated in the usual lexicographic order and may be + * restricted to a subset of the trie (all subnets of a specified prefix). + * + * Note that the trie walk does not reliably enumerate `implicit' prefixes + * defined by &low and &high fields in prefix patterns, it is supposed to be + * used on tries constructed from `explicit' prefixes (&low == &plen == &high + * in call to trie_add_prefix()). + * + * The trie walk has three basic state variables stored in the struct + * &f_trie_walk_state -- the current node in &stack[stack_pos], &accept_length + * for iteration over inter-node prefixes (non-branching prefixes on compressed + * path between the current node and its parent node, stored in the bitmap + * &accept of the current node) and &local_pos for iteration over intra-node + * prefixes (stored in the bitmap &local). + * + * The trie also supports longest-prefix-match query by trie_match_longest_ip4() + * and it can be extended to iteration over all covering prefixes for a given + * prefix (from longest to shortest) using TRIE_WALK_TO_ROOT_IP4() macro. There + * are also IPv6 versions (for practical reasons, these functions and macros are + * separate for IPv4 and IPv6). There is the same limitation to enumeration of + * `implicit' prefixes like with the previous TRIE_WALK() macro. */ #include "nest/bird.h" @@ -86,7 +126,10 @@ #define ipa_mkmask(x) ip6_mkmask(x) #define ipa_masklen(x) ip6_masklen(&x) #define ipa_pxlen(x,y) ip6_pxlen(x,y) -#define ipa_getbit(x,n) ip6_getbit(x,n) +#define ipa_getbit(a,p) ip6_getbit(a,p) +#define ipa_getbits(a,p,n) ip6_getbits(a,p,n) +#define ipa_setbits(a,p,n) ip6_setbits(a,p,n) +#define trie_local_mask(a,b,c) trie_local_mask6(a,b,c) #define ipt_from_ip4(x) _MI6(_I(x), 0, 0, 0) #define ipt_to_ip4(x) _MI4(_I0(x)) @@ -109,10 +152,11 @@ f_new_trie(linpool *lp, uint data_size) } static inline struct f_trie_node4 * -new_node4(struct f_trie *t, int plen, ip4_addr paddr, ip4_addr pmask, ip4_addr amask) +new_node4(struct f_trie *t, uint plen, uint local, ip4_addr paddr, ip4_addr pmask, ip4_addr amask) { struct f_trie_node4 *n = lp_allocz(t->lp, sizeof(struct f_trie_node4) + t->data_size); n->plen = plen; + n->local = local; n->addr = paddr; n->mask = pmask; n->accept = amask; @@ -120,10 +164,11 @@ new_node4(struct f_trie *t, int plen, ip4_addr paddr, ip4_addr pmask, ip4_addr a } static inline struct f_trie_node6 * -new_node6(struct f_trie *t, int plen, ip6_addr paddr, ip6_addr pmask, ip6_addr amask) +new_node6(struct f_trie *t, uint plen, uint local, ip6_addr paddr, ip6_addr pmask, ip6_addr amask) { struct f_trie_node6 *n = lp_allocz(t->lp, sizeof(struct f_trie_node6) + t->data_size); n->plen = plen; + n->local = local; n->addr = paddr; n->mask = pmask; n->accept = amask; @@ -131,24 +176,24 @@ new_node6(struct f_trie *t, int plen, ip6_addr paddr, ip6_addr pmask, ip6_addr a } static inline struct f_trie_node * -new_node(struct f_trie *t, int plen, ip_addr paddr, ip_addr pmask, ip_addr amask) +new_node(struct f_trie *t, uint plen, uint local, ip_addr paddr, ip_addr pmask, ip_addr amask) { if (t->ipv4) - return (struct f_trie_node *) new_node4(t, plen, ipt_to_ip4(paddr), ipt_to_ip4(pmask), ipt_to_ip4(amask)); + return (struct f_trie_node *) new_node4(t, plen, local, ipt_to_ip4(paddr), ipt_to_ip4(pmask), ipt_to_ip4(amask)); else - return (struct f_trie_node *) new_node6(t, plen, ipa_to_ip6(paddr), ipa_to_ip6(pmask), ipa_to_ip6(amask)); + return (struct f_trie_node *) new_node6(t, plen, local, ipa_to_ip6(paddr), ipa_to_ip6(pmask), ipa_to_ip6(amask)); } static inline void attach_node4(struct f_trie_node4 *parent, struct f_trie_node4 *child) { - parent->c[ip4_getbit(child->addr, parent->plen) ? 1 : 0] = child; + parent->c[ip4_getbits(child->addr, parent->plen, TRIE_STEP)] = child; } static inline void attach_node6(struct f_trie_node6 *parent, struct f_trie_node6 *child) { - parent->c[ip6_getbit(child->addr, parent->plen) ? 1 : 0] = child; + parent->c[ip6_getbits(child->addr, parent->plen, TRIE_STEP)] = child; } static inline void @@ -160,63 +205,96 @@ attach_node(struct f_trie_node *parent, struct f_trie_node *child, int v4) attach_node6(&parent->v6, &child->v6); } -#define GET_ADDR(N,F,X) ((X) ? ipt_from_ip4((N)->v4.F) : ipa_from_ip6((N)->v6.F)) -#define SET_ADDR(N,F,X,V) ({ if (X) (N)->v4.F =ipt_to_ip4(V); else (N)->v6.F =ipa_to_ip6(V); }) -#define GET_CHILD(N,F,X,I) ((X) ? (struct f_trie_node *) (N)->v4.c[I] : (struct f_trie_node *) (N)->v6.c[I]) -/** - * trie_add_prefix - * @t: trie to add to - * @net: IP network prefix - * @l: prefix lower bound - * @h: prefix upper bound +/* + * Internal prefixes of a node a represented by the local bitmask, each bit for + * one prefix. Bit 0 is unused, Bit 1 is for the main prefix of the node, + * remaining bits correspond to subprefixes by this pattern: * - * Adds prefix (prefix pattern) @n to trie @t. @l and @h are lower - * and upper bounds on accepted prefix lengths, both inclusive. - * 0 <= l, h <= 32 (128 for IPv6). + * 1 + * 2 3 + * 4 5 6 7 + * 8 9 A B C D E F * - * Returns a pointer to the allocated node. The function can return a pointer to - * an existing node if @px and @plen are the same. If px/plen == 0/0 (or ::/0), - * a pointer to the root node is returned. Returns NULL when called with - * mismatched IPv4/IPv6 net type. + * E.g. for 10.0.0.0/8 node, the 10.64.0.0/10 would be position 5. */ -void * -trie_add_prefix(struct f_trie *t, const net_addr *net, uint l, uint h) +/* + * Compute appropriate mask representing prefix px/plen in local bitmask of node + * with prefix length nlen. Assuming that nlen <= plen < (nlen + TRIE_STEP). + */ +static inline uint +trie_local_mask4(ip4_addr px, uint plen, uint nlen) { - uint plen = net_pxlen(net); - ip_addr px; - int v4; + uint step = plen - nlen; + uint pos = (1u << step) + ip4_getbits(px, nlen, step); + return 1u << pos; +} - switch (net->type) - { - case NET_IP4: px = ipt_from_ip4(net4_prefix(net)); v4 = 1; break; - case NET_IP6: px = ipa_from_ip6(net6_prefix(net)); v4 = 0; break; - default: bug("invalid type"); - } +static inline uint +trie_local_mask6(ip6_addr px, uint plen, uint nlen) +{ + uint step = plen - nlen; + uint pos = (1u << step) + ip6_getbits(px, nlen, step); + return 1u << pos; +} - if (t->ipv4 != v4) - { - if (t->ipv4 < 0) - t->ipv4 = v4; - else - return NULL; - } +/* + * Compute an appropriate local mask (for a node with prefix length nlen) + * representing prefixes of px that are accepted by amask and fall within the + * range associated with that node. Used for propagation of child accept mask + * to parent local mask. + */ +static inline uint +trie_amask_to_local(ip_addr px, ip_addr amask, uint nlen) +{ + uint local = 0; - if (l == 0) - t->zero = 1; - else - l--; + for (uint plen = MAX(nlen, 1); plen < (nlen + TRIE_STEP); plen++) + if (ipa_getbit(amask, plen - 1)) + local |= trie_local_mask(px, plen, nlen); - if (h < plen) - plen = h; + return local; +} + +/* + * Compute a bitmask representing a level of subprefixes (of the same length), + * using specified position as a root. E.g., level 2 from root position 3 would + * be bit positions C-F, returned as bitmask 0xf000. + */ +static inline uint +trie_level_mask(uint pos, uint level) +{ + return ((1u << (1u << level)) - 1) << (pos << level); +} - ip_addr amask = ipa_xor(ipa_mkmask(l), ipa_mkmask(h)); + +#define GET_ADDR(N,F,X) ((X) ? ipt_from_ip4((N)->v4.F) : ipa_from_ip6((N)->v6.F)) +#define SET_ADDR(N,F,X,V) ({ if (X) (N)->v4.F =ipt_to_ip4(V); else (N)->v6.F =ipa_to_ip6(V); }) + +#define GET_LOCAL(N,X) ((X) ? (N)->v4.local : (N)->v6.local) +#define ADD_LOCAL(N,X,V) ({ uint v_ = (V); if (X) (N)->v4.local |= v_; else (N)->v6.local |= v_; }) + +#define GET_CHILD(N,X,I) ((X) ? (struct f_trie_node *) (N)->v4.c[I] : (struct f_trie_node *) (N)->v6.c[I]) + + +static void * +trie_add_node(struct f_trie *t, uint plen, ip_addr px, uint local, uint l, uint h) +{ + uint l_ = l ? (l - 1) : 0; + ip_addr amask = (l_ < h) ? ipa_xor(ipa_mkmask(l_), ipa_mkmask(h)) : IPA_NONE; ip_addr pmask = ipa_mkmask(plen); ip_addr paddr = ipa_and(px, pmask); struct f_trie_node *o = NULL; struct f_trie_node *n = &t->root; + int v4 = t->ipv4; + /* Add all bits for each active level (0x0002 0x000c 0x00f0 0xff00) */ + for (uint i = 0; i < TRIE_STEP; i++) + if ((l <= (plen + i)) && ((plen + i) <= h)) + local |= trie_level_mask(1, i); + + DBG("Insert node %I/%u (%I %x)\n", paddr, plen, amask, local); while (n) { ip_addr naddr = GET_ADDR(n, addr, v4); @@ -225,23 +303,31 @@ trie_add_prefix(struct f_trie *t, const net_addr *net, uint l, uint h) ip_addr cmask = ipa_and(nmask, pmask); uint nlen = v4 ? n->v4.plen : n->v6.plen; + DBG("Found node %I/%u (%I %x)\n", + naddr, nlen, accept, v4 ? n->v4.local : n->v6.local); + if (ipa_compare(ipa_and(paddr, cmask), ipa_and(naddr, cmask))) { /* We are out of path - we have to add branching node 'b' between node 'o' and node 'n', and attach new node 'a' as the other child of 'b'. */ - int blen = ipa_pxlen(paddr, naddr); + int blen = ROUND_DOWN_POW2(ipa_pxlen(paddr, naddr), TRIE_STEP); ip_addr bmask = ipa_mkmask(blen); ip_addr baddr = ipa_and(px, bmask); /* Merge accept masks from children to get accept mask for node 'b' */ ip_addr baccm = ipa_and(ipa_or(amask, accept), bmask); + uint bloc = trie_amask_to_local(naddr, accept, blen) | + trie_amask_to_local(paddr, amask, blen); - struct f_trie_node *a = new_node(t, plen, paddr, pmask, amask); - struct f_trie_node *b = new_node(t, blen, baddr, bmask, baccm); + struct f_trie_node *a = new_node(t, plen, local, paddr, pmask, amask); + struct f_trie_node *b = new_node(t, blen, bloc, baddr, bmask, baccm); attach_node(o, b, v4); attach_node(b, n, v4); attach_node(b, a, v4); + t->prefix_count++; + + DBG("Case 1\n"); return a; } @@ -249,66 +335,195 @@ trie_add_prefix(struct f_trie *t, const net_addr *net, uint l, uint h) { /* We add new node 'a' between node 'o' and node 'n' */ amask = ipa_or(amask, ipa_and(accept, pmask)); - struct f_trie_node *a = new_node(t, plen, paddr, pmask, amask); + local |= trie_amask_to_local(naddr, accept, plen); + struct f_trie_node *a = new_node(t, plen, local, paddr, pmask, amask); attach_node(o, a, v4); attach_node(a, n, v4); + t->prefix_count++; + + DBG("Case 2\n"); return a; } if (plen == nlen) { - /* We already found added node in trie. Just update accept mask */ + /* We already found added node in trie. Just update accept and local mask */ accept = ipa_or(accept, amask); SET_ADDR(n, accept, v4, accept); + + if ((GET_LOCAL(n, v4) & local) != local) + t->prefix_count++; + + ADD_LOCAL(n, v4, local); + + DBG("Case 3\n"); return n; } /* Update accept mask part M2 and go deeper */ accept = ipa_or(accept, ipa_and(amask, nmask)); SET_ADDR(n, accept, v4, accept); + ADD_LOCAL(n, v4, trie_amask_to_local(paddr, amask, nlen)); + + DBG("Step %u\n", ipa_getbits(paddr, nlen)); /* n->plen < plen and plen <= 32 (128) */ o = n; - n = GET_CHILD(n, c, v4, ipa_getbit(paddr, nlen) ? 1 : 0); + n = GET_CHILD(n, v4, ipa_getbits(paddr, nlen, TRIE_STEP)); } /* We add new tail node 'a' after node 'o' */ - struct f_trie_node *a = new_node(t, plen, paddr, pmask, amask); + struct f_trie_node *a = new_node(t, plen, local, paddr, pmask, amask); attach_node(o, a, v4); + t->prefix_count++; + DBG("Case 4\n"); return a; } +/** + * trie_add_prefix + * @t: trie to add to + * @net: IP network prefix + * @l: prefix lower bound + * @h: prefix upper bound + * + * Adds prefix (prefix pattern) @n to trie @t. @l and @h are lower + * and upper bounds on accepted prefix lengths, both inclusive. + * 0 <= l, h <= 32 (128 for IPv6). + * + * Returns a pointer to the allocated node. The function can return a pointer to + * an existing node if @px and @plen are the same. If px/plen == 0/0 (or ::/0), + * a pointer to the root node is returned. Returns NULL when called with + * mismatched IPv4/IPv6 net type. + */ +void * +trie_add_prefix(struct f_trie *t, const net_addr *net, uint l, uint h) +{ + uint plen = net_pxlen(net); + ip_addr px; + int v4; + + switch (net->type) + { + case NET_IP4: + case NET_VPN4: + case NET_ROA4: + px = ipt_from_ip4(net4_prefix(net)); + v4 = 1; + break; + + case NET_IP6: + case NET_VPN6: + case NET_ROA6: + case NET_IP6_SADR: + px = ipa_from_ip6(net6_prefix(net)); + v4 = 0; + break; + + default: + bug("invalid type"); + } + + if (t->ipv4 != v4) + { + if (t->ipv4 < 0) + t->ipv4 = v4; + else + return NULL; + } + + DBG("\nInsert net %N (%u-%u)\n", net, l, h); + + if (l == 0) + t->zero = 1; + + if (h < plen) + plen = h; + + /* Primary node length, plen rounded down */ + uint nlen = ROUND_DOWN_POW2(plen, TRIE_STEP); + + if (plen == nlen) + return trie_add_node(t, nlen, px, 0, l, h); + + /* Secondary node length, plen rouned up */ + uint slen = nlen + TRIE_STEP; + void *node = NULL; + + /* + * For unaligned prefix lengths it is more complicated. We need to encode + * matching prefixes of lengths from l to h. There are three cases of lengths: + * + * 1) 0..nlen are encoded by the accept mask of the primary node + * 2) nlen..(slen-1) are encoded by the local mask of the primary node + * 3) slen..max are encoded in secondary nodes + */ + + if (l < slen) + { + uint local = 0; + + /* Compute local bits for accepted nlen..(slen-1) prefixes */ + for (uint i = 0; i < TRIE_STEP; i++) + if ((l <= (nlen + i)) && ((nlen + i) <= h)) + { + uint pos = (1u << i) + ipa_getbits(px, nlen, i); + uint len = ((nlen + i) <= plen) ? 1 : (1u << (nlen + i - plen)); + + /* We need to fill 'len' bits starting at 'pos' position */ + local |= ((1u << len) - 1) << pos; + } + + /* Add the primary node */ + node = trie_add_node(t, nlen, px, local, l, nlen); + } + + if (slen <= h) + { + uint l2 = MAX(l, slen); + uint max = (1u << (slen - plen)); + + /* Add secondary nodes */ + for (uint i = 0; i < max; i++) + node = trie_add_node(t, slen, ipa_setbits(px, slen - 1, i), 0, l2, h); + } + + return node; +} + + static int trie_match_net4(const struct f_trie *t, ip4_addr px, uint plen) { - ip4_addr pmask = ip4_mkmask(plen); - ip4_addr paddr = ip4_and(px, pmask); - if (plen == 0) return t->zero; int plentest = plen - 1; + uint nlen = ROUND_DOWN_POW2(plen, TRIE_STEP); + uint local = trie_local_mask4(px, plen, nlen); const struct f_trie_node4 *n = &t->root.v4; while (n) { - ip4_addr cmask = ip4_and(n->mask, pmask); - /* We are out of path */ - if (ip4_compare(ip4_and(paddr, cmask), ip4_and(n->addr, cmask))) + if (!ip4_prefix_equal(px, n->addr, MIN(plen, n->plen))) return 0; + /* Check local mask */ + if ((n->plen == nlen) && (n->local & local)) + return 1; + /* Check accept mask */ if (ip4_getbit(n->accept, plentest)) return 1; /* We finished trie walk and still no match */ - if (plen <= n->plen) + if (nlen <= n->plen) return 0; /* Choose children */ - n = n->c[(ip4_getbit(paddr, n->plen)) ? 1 : 0]; + n = n->c[ip4_getbits(px, n->plen, TRIE_STEP)]; } return 0; @@ -317,33 +532,34 @@ trie_match_net4(const struct f_trie *t, ip4_addr px, uint plen) static int trie_match_net6(const struct f_trie *t, ip6_addr px, uint plen) { - ip6_addr pmask = ip6_mkmask(plen); - ip6_addr paddr = ip6_and(px, pmask); - if (plen == 0) return t->zero; int plentest = plen - 1; + uint nlen = ROUND_DOWN_POW2(plen, TRIE_STEP); + uint local = trie_local_mask6(px, plen, nlen); const struct f_trie_node6 *n = &t->root.v6; while (n) { - ip6_addr cmask = ip6_and(n->mask, pmask); - /* We are out of path */ - if (ip6_compare(ip6_and(paddr, cmask), ip6_and(n->addr, cmask))) + if (!ip6_prefix_equal(px, n->addr, MIN(plen, n->plen))) return 0; + /* Check local mask */ + if ((n->plen == nlen) && (n->local & local)) + return 1; + /* Check accept mask */ if (ip6_getbit(n->accept, plentest)) return 1; /* We finished trie walk and still no match */ - if (plen <= n->plen) + if (nlen <= n->plen) return 0; /* Choose children */ - n = n->c[(ip6_getbit(paddr, n->plen)) ? 1 : 0]; + n = n->c[ip6_getbits(px, n->plen, TRIE_STEP)]; } return 0; @@ -378,6 +594,412 @@ trie_match_net(const struct f_trie *t, const net_addr *n) } } + +/** + * trie_match_longest_ip4 + * @t: trie + * @net: net address + * @dst: return value + * @found0: optional returned bitmask of found nodes + * + * Perform longest prefix match for the address @net and return the resulting + * prefix in the buffer @dst. The bitmask @found0 is used to report lengths of + * prefixes on the path from the root to the resulting prefix. E.g., if there is + * also a /20 shorter matching prefix, then 20-th bit is set in @found0. This + * can be used to enumerate all matching prefixes for the network @net using + * function trie_match_next_longest_ip4() or macro TRIE_WALK_TO_ROOT_IP4(). + * + * This function assumes IPv4 trie, there is also an IPv6 variant. The @net + * argument is typed as net_addr_ip4, but would accept any IPv4-based net_addr, + * like net4_prefix(). Anyway, returned @dst is always net_addr_ip4. + * + * Result: 1 if a matching prefix was found, 0 if not. + */ +int +trie_match_longest_ip4(const struct f_trie *t, const net_addr_ip4 *net, net_addr_ip4 *dst, ip4_addr *found0) +{ + ASSERT(t->ipv4); + + const ip4_addr prefix = net->prefix; + const int pxlen = net->pxlen; + + const struct f_trie_node4 *n = &t->root.v4; + int len = 0; + + ip4_addr found = IP4_NONE; + int last = -1; + + while (n) + { + /* We are out of path */ + if (!ip4_prefix_equal(prefix, n->addr, MIN(pxlen, n->plen))) + goto done; + + /* Check accept mask */ + for (; len < n->plen; len++) + { + if (len > pxlen) + goto done; + + if (ip4_getbit(n->accept, len - 1)) + { + /* len is always < 32 due to len < n->plen */ + ip4_setbit(&found, len); + last = len; + } + } + + /* Special case for max length, there is only one valid local position */ + if (len == IP4_MAX_PREFIX_LENGTH) + { + if (n->local & (1u << 1)) + last = len; + + goto done; + } + + /* Check local mask */ + for (int pos = 1; pos < (1 << TRIE_STEP); pos = 2 * pos + ip4_getbit(prefix, len), len++) + { + if (len > pxlen) + goto done; + + if (n->local & (1u << pos)) + { + /* len is always < 32 due to special case above */ + ip4_setbit(&found, len); + last = len; + } + } + + /* Choose child */ + n = n->c[ip4_getbits(prefix, n->plen, TRIE_STEP)]; + } + +done: + if (last < 0) + return 0; + + *dst = NET_ADDR_IP4(ip4_and(prefix, ip4_mkmask(last)), last); + + if (found0) + *found0 = found; + + return 1; +} + + +/** + * trie_match_longest_ip6 + * @t: trie + * @net: net address + * @dst: return value + * @found0: optional returned bitmask of found nodes + * + * Perform longest prefix match for the address @net and return the resulting + * prefix in the buffer @dst. The bitmask @found0 is used to report lengths of + * prefixes on the path from the root to the resulting prefix. E.g., if there is + * also a /20 shorter matching prefix, then 20-th bit is set in @found0. This + * can be used to enumerate all matching prefixes for the network @net using + * function trie_match_next_longest_ip6() or macro TRIE_WALK_TO_ROOT_IP6(). + * + * This function assumes IPv6 trie, there is also an IPv4 variant. The @net + * argument is typed as net_addr_ip6, but would accept any IPv6-based net_addr, + * like net6_prefix(). Anyway, returned @dst is always net_addr_ip6. + * + * Result: 1 if a matching prefix was found, 0 if not. + */ +int +trie_match_longest_ip6(const struct f_trie *t, const net_addr_ip6 *net, net_addr_ip6 *dst, ip6_addr *found0) +{ + ASSERT(!t->ipv4); + + const ip6_addr prefix = net->prefix; + const int pxlen = net->pxlen; + + const struct f_trie_node6 *n = &t->root.v6; + int len = 0; + + ip6_addr found = IP6_NONE; + int last = -1; + + while (n) + { + /* We are out of path */ + if (!ip6_prefix_equal(prefix, n->addr, MIN(pxlen, n->plen))) + goto done; + + /* Check accept mask */ + for (; len < n->plen; len++) + { + if (len > pxlen) + goto done; + + if (ip6_getbit(n->accept, len - 1)) + { + /* len is always < 128 due to len < n->plen */ + ip6_setbit(&found, len); + last = len; + } + } + + /* Special case for max length, there is only one valid local position */ + if (len == IP6_MAX_PREFIX_LENGTH) + { + if (n->local & (1u << 1)) + last = len; + + goto done; + } + + /* Check local mask */ + for (int pos = 1; pos < (1 << TRIE_STEP); pos = 2 * pos + ip6_getbit(prefix, len), len++) + { + if (len > pxlen) + goto done; + + if (n->local & (1u << pos)) + { + /* len is always < 128 due to special case above */ + ip6_setbit(&found, len); + last = len; + } + } + + /* Choose child */ + n = n->c[ip6_getbits(prefix, n->plen, TRIE_STEP)]; + } + +done: + if (last < 0) + return 0; + + *dst = NET_ADDR_IP6(ip6_and(prefix, ip6_mkmask(last)), last); + + if (found0) + *found0 = found; + + return 1; +} + +#define SAME_PREFIX(A,B,X,L) ((X) ? ip4_prefix_equal((A)->v4.addr, net4_prefix(B), (L)) : ip6_prefix_equal((A)->v6.addr, net6_prefix(B), (L))) +#define GET_NET_BITS(N,X,A,B) ((X) ? ip4_getbits(net4_prefix(N), (A), (B)) : ip6_getbits(net6_prefix(N), (A), (B))) + +/** + * trie_walk_init + * @s: walk state + * @t: trie + * @net: optional subnet for walk + * + * Initialize walk state for subsequent walk through nodes of the trie @t by + * trie_walk_next(). The argument @net allows to restrict walk to given subnet, + * otherwise full walk over all nodes is used. This is done by finding node at + * or below @net and starting position in it. + */ +void +trie_walk_init(struct f_trie_walk_state *s, const struct f_trie *t, const net_addr *net) +{ + *s = (struct f_trie_walk_state) { + .ipv4 = t->ipv4, + .accept_length = 0, + .start_pos = 1, + .local_pos = 1, + .stack_pos = 0, + .stack[0] = &t->root + }; + + if (!net) + return; + + /* We want to find node of level at least plen */ + int plen = ROUND_DOWN_POW2(net->pxlen, TRIE_STEP); + const struct f_trie_node *n = &t->root; + const int v4 = t->ipv4; + + while (n) + { + int nlen = v4 ? n->v4.plen : n->v6.plen; + + /* We are out of path */ + if (!SAME_PREFIX(n, net, v4, MIN(net->pxlen, nlen))) + break; + + /* We found final node */ + if (nlen >= plen) + { + if (nlen == plen) + { + /* Find proper local_pos, while accept_length is not used */ + int step = net->pxlen - plen; + s->start_pos = s->local_pos = (1u << step) + GET_NET_BITS(net, v4, plen, step); + s->accept_length = plen; + } + else + { + /* Start from pos 1 in local node, but first try accept mask */ + s->accept_length = net->pxlen; + } + + s->stack[0] = n; + return; + } + + /* Choose child */ + n = GET_CHILD(n, v4, GET_NET_BITS(net, v4, nlen, TRIE_STEP)); + } + + s->stack[0] = NULL; + return; +} + +#define GET_ACCEPT_BIT(N,X,B) ((X) ? ip4_getbit((N)->v4.accept, (B)) : ip6_getbit((N)->v6.accept, (B))) +#define GET_LOCAL_BIT(N,X,B) (((X) ? (N)->v4.local : (N)->v6.local) & (1u << (B))) + +/** + * trie_walk_next + * @s: walk state + * @net: return value + * + * Find the next prefix in the trie walk and return it in the buffer @net. + * Prefixes are walked in the usual lexicographic order and may be restricted + * to a subset of the trie during walk setup by trie_walk_init(). Note that the + * trie walk does not iterate reliably over 'implicit' prefixes defined by &low + * and &high fields in prefix patterns, it is supposed to be used on tries + * constructed from 'explicit' prefixes (&low == &plen == &high in call to + * trie_add_prefix()). + * + * Result: 1 if the next prefix was found, 0 for the end of walk. + */ +int +trie_walk_next(struct f_trie_walk_state *s, net_addr *net) +{ + const struct f_trie_node *n = s->stack[s->stack_pos]; + int len = s->accept_length; + int pos = s->local_pos; + int v4 = s->ipv4; + + /* + * The walk has three basic state variables -- n, len and pos. In each node n, + * we first walk superprefixes (by len in &accept bitmask), and then we walk + * internal positions (by pos in &local bitmask). These positions are: + * + * 1 + * 2 3 + * 4 5 6 7 + * 8 9 A B C D E F + * + * We walk them depth-first, including virtual positions 10-1F that are + * equivalent of position 1 in child nodes 0-F. + */ + + if (!n) + { + memset(net, 0, v4 ? sizeof(net_addr_ip4) : sizeof(net_addr_ip6)); + return 0; + } + +next_node:; + /* Current node prefix length */ + int nlen = v4 ? n->v4.plen : n->v6.plen; + + /* First, check for accept prefix */ + for (; len < nlen; len++) + if (GET_ACCEPT_BIT(n, v4, len - 1)) + { + if (v4) + net_fill_ip4(net, ip4_and(n->v4.addr, ip4_mkmask(len)), len); + else + net_fill_ip6(net, ip6_and(n->v6.addr, ip6_mkmask(len)), len); + + s->local_pos = pos; + s->accept_length = len + 1; + return 1; + } + +next_pos: + /* Bottom of this node */ + if (pos >= (1 << TRIE_STEP)) + { + const struct f_trie_node *child = GET_CHILD(n, v4, pos - (1 << TRIE_STEP)); + int dir = 0; + + /* No child node */ + if (!child) + { + /* Step up until return from left child (pos is even) */ + do + { + /* Step up from start node */ + if ((s->stack_pos == 0) && (pos == s->start_pos)) + { + s->stack[0] = NULL; + memset(net, 0, v4 ? sizeof(net_addr_ip4) : sizeof(net_addr_ip6)); + return 0; + } + + /* Top of this node */ + if (pos == 1) + { + ASSERT(s->stack_pos); + const struct f_trie_node *old = n; + + /* Move to parent node */ + s->stack_pos--; + n = s->stack[s->stack_pos]; + nlen = v4 ? n->v4.plen : n->v6.plen; + + pos = v4 ? + ip4_getbits(old->v4.addr, nlen, TRIE_STEP) : + ip6_getbits(old->v6.addr, nlen, TRIE_STEP); + pos += (1 << TRIE_STEP); + len = nlen; + + ASSERT(GET_CHILD(n, v4, pos - (1 << TRIE_STEP)) == old); + } + + /* Step up */ + dir = pos % 2; + pos = pos / 2; + } + while (dir); + + /* Continue with step down to the right child */ + pos = 2 * pos + 1; + goto next_pos; + } + + /* Move to child node */ + pos = 1; + len = nlen + TRIE_STEP; + + s->stack_pos++; + n = s->stack[s->stack_pos] = child; + goto next_node; + } + + /* Check for local prefix */ + if (GET_LOCAL_BIT(n, v4, pos)) + { + /* Convert pos to address of local network */ + int x = (pos >= 2) + (pos >= 4) + (pos >= 8); + int y = pos & ((1u << x) - 1); + + if (v4) + net_fill_ip4(net, !x ? n->v4.addr : ip4_setbits(n->v4.addr, nlen + x - 1, y), nlen + x); + else + net_fill_ip6(net, !x ? n->v6.addr : ip6_setbits(n->v6.addr, nlen + x - 1, y), nlen + x); + + s->local_pos = 2 * pos; + s->accept_length = len; + return 1; + } + + /* Step down */ + pos = 2 * pos; + goto next_pos; +} + + static int trie_node_same4(const struct f_trie_node4 *t1, const struct f_trie_node4 *t2) { @@ -392,7 +1014,11 @@ trie_node_same4(const struct f_trie_node4 *t1, const struct f_trie_node4 *t2) (! ip4_equal(t1->accept, t2->accept))) return 0; - return trie_node_same4(t1->c[0], t2->c[0]) && trie_node_same4(t1->c[1], t2->c[1]); + for (uint i = 0; i < (1 << TRIE_STEP); i++) + if (! trie_node_same4(t1->c[i], t2->c[i])) + return 0; + + return 1; } static int @@ -409,7 +1035,11 @@ trie_node_same6(const struct f_trie_node6 *t1, const struct f_trie_node6 *t2) (! ip6_equal(t1->accept, t2->accept))) return 0; - return trie_node_same6(t1->c[0], t2->c[0]) && trie_node_same6(t1->c[1], t2->c[1]); + for (uint i = 0; i < (1 << TRIE_STEP); i++) + if (! trie_node_same6(t1->c[i], t2->c[i])) + return 0; + + return 1; } /** @@ -431,30 +1061,70 @@ trie_same(const struct f_trie *t1, const struct f_trie *t2) return trie_node_same6(&t1->root.v6, &t2->root.v6); } + +static const u8 log2[16] = {0, 0, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 3}; + static void -trie_node_format4(const struct f_trie_node4 *t, buffer *buf) +trie_node_format(const struct f_trie_node *n, buffer *buf, int v4) { - if (t == NULL) + if (n == NULL) return; - if (ip4_nonzero(t->accept)) - buffer_print(buf, "%I4/%d{%I4}, ", t->addr, t->plen, t->accept); + if (v4) + { + if (ip4_nonzero(n->v4.accept)) + buffer_print(buf, "%I4/%d{%I4}, ", n->v4.addr, n->v4.plen, n->v4.accept); + } + else + { + if (ip6_nonzero(n->v6.accept)) + buffer_print(buf, "%I6/%d{%I6}, ", n->v6.addr, n->v6.plen, n->v6.accept); + } - trie_node_format4(t->c[0], buf); - trie_node_format4(t->c[1], buf); -} + int nlen = v4 ? n->v4.plen : n->v6.plen; + uint local = v4 ? n->v4.local : n->v6.local; -static void -trie_node_format6(const struct f_trie_node6 *t, buffer *buf) -{ - if (t == NULL) - return; + for (int i = (nlen ? 0 : 1); i < TRIE_STEP; i++) + if (GET_ACCEPT_BIT(n, v4, nlen + i - 1)) + local &= ~trie_level_mask(1, i); - if (ip6_nonzero(t->accept)) - buffer_print(buf, "%I6/%d{%I6}, ", t->addr, t->plen, t->accept); + for (int pos = 2; local && (pos < (1 << TRIE_STEP)); pos++) + if (local & (1u << pos)) + { + int lvl = log2[pos]; + int plen = nlen + lvl; + + int i; + for (i = 0; lvl + i < TRIE_STEP; i++) + { + uint lmask = trie_level_mask(pos, i); + + if ((local & lmask) != lmask) + break; + + local &= ~lmask; + } + + uint addr_bits = pos & ((1u << lvl) - 1); + uint accept_bits = (1u << i) - 1; + int h = plen + i - 1; + + if (v4) + { + ip4_addr addr = ip4_setbits(n->v4.addr, plen - 1, addr_bits); + ip4_addr mask = ip4_setbits(IP4_NONE, h - 1, accept_bits); + buffer_print(buf, "%I4/%d{%I4}, ", addr, plen, mask); + } + else + { + ip6_addr addr = ip6_setbits(n->v6.addr, plen - 1, addr_bits); + ip6_addr mask = ip6_setbits(IP6_NONE, h - 1, accept_bits); + buffer_print(buf, "%I6/%d{%I6}, ", addr, plen, mask); + } + } - trie_node_format6(t->c[0], buf); - trie_node_format6(t->c[1], buf); + for (int i = 0; i < (1 << TRIE_STEP); i++) + trie_node_format(GET_CHILD(n, v4, i), buf, v4); } /** @@ -472,10 +1142,7 @@ trie_format(const struct f_trie *t, buffer *buf) if (t->zero) buffer_print(buf, "%I/%d, ", t->ipv4 ? IPA_NONE4 : IPA_NONE6, 0); - if (t->ipv4) - trie_node_format4(&t->root.v4, buf); - else - trie_node_format6(&t->root.v6, buf); + trie_node_format(&t->root, buf, t->ipv4); if (buf->pos == buf->end) return; diff --git a/filter/trie_test.c b/filter/trie_test.c index b2b36716..dc791280 100644 --- a/filter/trie_test.c +++ b/filter/trie_test.c @@ -14,9 +14,12 @@ #include "conf/conf.h" #define TESTS_NUM 10 -#define PREFIXES_NUM 10 +#define PREFIXES_NUM 32 #define PREFIX_TESTS_NUM 10000 +#define PREFIX_BENCH_NUM 100000000 +#define TRIE_BUFFER_SIZE 1024 +#define TEST_BUFFER_SIZE (1024*1024) #define BIG_BUFFER_SIZE 10000 /* Wrapping structure for storing f_prefixes structures in list */ @@ -31,146 +34,849 @@ xrandom(u32 max) return (bt_random() % max); } +static inline uint +get_exp_random(void) +{ + uint r, n = 0; + + for (r = bt_random(); r & 1; r = r >> 1) + n++; + + return n; +} + static int -is_prefix_included(list *prefixes, struct f_prefix *needle) +compare_prefixes(const void *a, const void *b) { - struct f_prefix_node *n; - WALK_LIST(n, *prefixes) - { - ip6_addr cmask = ip6_mkmask(MIN(n->prefix.net.pxlen, needle->net.pxlen)); + return net_compare(&((const struct f_prefix *) a)->net, + &((const struct f_prefix *) b)->net); +} + +static inline int +matching_ip4_nets(const net_addr_ip4 *a, const net_addr_ip4 *b) +{ + ip4_addr cmask = ip4_mkmask(MIN(a->pxlen, b->pxlen)); + return ip4_compare(ip4_and(a->prefix, cmask), ip4_and(b->prefix, cmask)) == 0; +} + +static inline int +matching_ip6_nets(const net_addr_ip6 *a, const net_addr_ip6 *b) +{ + ip6_addr cmask = ip6_mkmask(MIN(a->pxlen, b->pxlen)); + return ip6_compare(ip6_and(a->prefix, cmask), ip6_and(b->prefix, cmask)) == 0; +} - ip6_addr ip = net6_prefix(&n->prefix.net); - ip6_addr needle_ip = net6_prefix(&needle->net); +static inline int +matching_nets(const net_addr *a, const net_addr *b) +{ + if (a->type != b->type) + return 0; + + return (a->type == NET_IP4) ? + matching_ip4_nets((const net_addr_ip4 *) a, (const net_addr_ip4 *) b) : + matching_ip6_nets((const net_addr_ip6 *) a, (const net_addr_ip6 *) b); +} - if ((ipa_compare(ipa_and(ip, cmask), ipa_and(needle_ip, cmask)) == 0) && - (n->prefix.lo <= needle->net.pxlen) && (needle->net.pxlen <= n->prefix.hi)) +static int +is_prefix_included(list *prefixes, const net_addr *needle) +{ + struct f_prefix_node *n; + WALK_LIST(n, *prefixes) + if (matching_nets(&n->prefix.net, needle) && + (n->prefix.lo <= needle->pxlen) && (needle->pxlen <= n->prefix.hi)) { - bt_debug("FOUND\t" PRIip6 "/%d %d-%d\n", ARGip6(net6_prefix(&n->prefix.net)), n->prefix.net.pxlen, n->prefix.lo, n->prefix.hi); + char buf[64]; + bt_format_net(buf, 64, &n->prefix.net); + bt_debug("FOUND %s %d-%d\n", buf, n->prefix.lo, n->prefix.hi); + return 1; /* OK */ } - } + return 0; /* FAIL */ } -static struct f_prefix -get_random_ip6_prefix(void) +static void +get_random_net(net_addr *net, int v6) { - struct f_prefix p; - u8 pxlen = xrandom(120)+8; - ip6_addr ip6 = ip6_build(bt_random(),bt_random(),bt_random(),bt_random()); - net_addr_ip6 net6 = NET_ADDR_IP6(ip6, pxlen); + if (!v6) + { + uint pxlen = xrandom(24)+8; + ip4_addr ip4 = ip4_from_u32((u32) bt_random()); + net_fill_ip4(net, ip4_and(ip4, ip4_mkmask(pxlen)), pxlen); + } + else + { + uint pxlen = xrandom(120)+8; + ip6_addr ip6 = ip6_build(bt_random(), bt_random(), bt_random(), bt_random()); + net_fill_ip6(net, ip6_and(ip6, ip6_mkmask(pxlen)), pxlen); + } +} - p.net = *((net_addr*) &net6); +static void +get_random_prefix(struct f_prefix *px, int v6, int tight) +{ + get_random_net(&px->net, v6); + + if (tight) + { + px->lo = px->hi = px->net.pxlen; + } + else if (bt_random() % 2) + { + px->lo = 0; + px->hi = px->net.pxlen; + } + else + { + px->lo = px->net.pxlen; + px->hi = net_max_prefix_length[px->net.type]; + } +} + +static void +get_random_ip4_subnet(net_addr_ip4 *net, const net_addr_ip4 *src, int pxlen) +{ + *net = NET_ADDR_IP4(ip4_and(src->prefix, ip4_mkmask(pxlen)), pxlen); + + if (pxlen > src->pxlen) + { + ip4_addr rnd = ip4_from_u32((u32) bt_random()); + ip4_addr mask = ip4_xor(ip4_mkmask(src->pxlen), ip4_mkmask(pxlen)); + net->prefix = ip4_or(net->prefix, ip4_and(rnd, mask)); + } +} + +static void +get_random_ip6_subnet(net_addr_ip6 *net, const net_addr_ip6 *src, int pxlen) +{ + *net = NET_ADDR_IP6(ip6_and(src->prefix, ip6_mkmask(pxlen)), pxlen); + + if (pxlen > src->pxlen) + { + ip6_addr rnd = ip6_build(bt_random(), bt_random(), bt_random(), bt_random()); + ip6_addr mask = ip6_xor(ip6_mkmask(src->pxlen), ip6_mkmask(pxlen)); + net->prefix = ip6_or(net->prefix, ip6_and(rnd, mask)); + } +} + +static void +get_random_subnet(net_addr *net, const net_addr *src, int pxlen) +{ + if (src->type == NET_IP4) + get_random_ip4_subnet((net_addr_ip4 *) net, (const net_addr_ip4 *) src, pxlen); + else + get_random_ip6_subnet((net_addr_ip6 *) net, (const net_addr_ip6 *) src, pxlen); +} + +static void +get_inner_net(net_addr *net, const struct f_prefix *src) +{ + int pxlen, step; if (bt_random() % 2) { - p.lo = 0; - p.hi = p.net.pxlen; + step = get_exp_random(); + step = MIN(step, src->hi - src->lo); + pxlen = (bt_random() % 2) ? (src->lo + step) : (src->hi - step); } else + pxlen = src->lo + bt_random() % (src->hi - src->lo + 1); + + get_random_subnet(net, &src->net, pxlen); +} + +static void +swap_random_bits_ip4(net_addr_ip4 *net, int num) +{ + for (int i = 0; i < num; i++) { - p.lo = p.net.pxlen; - p.hi = net_max_prefix_length[p.net.type]; + ip4_addr swap = IP4_NONE; + ip4_setbit(&swap, bt_random() % net->pxlen); + net->prefix = ip4_xor(net->prefix, swap); } +} - return p; +static void +swap_random_bits_ip6(net_addr_ip6 *net, int num) +{ + for (int i = 0; i < num; i++) + { + ip6_addr swap = IP6_NONE; + ip6_setbit(&swap, bt_random() % net->pxlen); + net->prefix = ip6_xor(net->prefix, swap); + } } static void -generate_random_ipv6_prefixes(list *prefixes) +swap_random_bits(net_addr *net, int num) { - int i; - for (i = 0; i < PREFIXES_NUM; i++) + if (net->type == NET_IP4) + swap_random_bits_ip4((net_addr_ip4 *) net, num); + else + swap_random_bits_ip6((net_addr_ip6 *) net, num); +} + +static void +get_outer_net(net_addr *net, const struct f_prefix *src) +{ + int pxlen, step; + int inside = 0; + int max = net_max_prefix_length[src->net.type]; + + if ((src->lo > 0) && (bt_random() % 3)) + { + step = 1 + get_exp_random(); + step = MIN(step, src->lo); + pxlen = src->lo - step; + } + else if ((src->hi < max) && (bt_random() % 2)) { - struct f_prefix f = get_random_ip6_prefix(); + step = 1 + get_exp_random(); + step = MIN(step, max - src->hi); + pxlen = src->hi + step; + } + else + { + pxlen = src->lo + bt_random() % (src->hi - src->lo + 1); + inside = 1; + } - struct f_prefix_node *px = calloc(1, sizeof(struct f_prefix_node)); - px->prefix = f; + get_random_subnet(net, &src->net, pxlen); - bt_debug("ADD\t" PRIip6 "/%d %d-%d\n", ARGip6(net6_prefix(&px->prefix.net)), px->prefix.net.pxlen, px->prefix.lo, px->prefix.hi); + /* Perhaps swap some bits in prefix */ + if ((net->pxlen > 0) && (inside || (bt_random() % 4))) + swap_random_bits(net, 1 + get_exp_random()); +} + +static list * +make_random_prefix_list(int num, int v6, int tight) +{ + list *prefixes = lp_allocz(tmp_linpool, sizeof(struct f_prefix_node)); + init_list(prefixes); + + for (int i = 0; i < num; i++) + { + struct f_prefix_node *px = lp_allocz(tmp_linpool, sizeof(struct f_prefix_node)); + get_random_prefix(&px->prefix, v6, tight); add_tail(prefixes, &px->n); + + char buf[64]; + bt_format_net(buf, 64, &px->prefix.net); + bt_debug("ADD %s{%d,%d}\n", buf, px->prefix.lo, px->prefix.hi); + } + + return prefixes; +} + +static struct f_trie * +make_trie_from_prefix_list(list *prefixes) +{ + struct f_trie *trie = f_new_trie(tmp_linpool, 0); + + struct f_prefix_node *n; + WALK_LIST(n, *prefixes) + trie_add_prefix(trie, &n->prefix.net, n->prefix.lo, n->prefix.hi); + + return trie; +} + +/* + * Read sequence of prefixes from file handle and return prefix list. + * Each prefix is on one line, sequence terminated by empty line or eof. + * Arg @plus means prefix should include all longer ones. + */ +static list * +read_prefix_list(FILE *f, int v6, int plus) +{ + ASSERT(!v6); + + uint a0, a1, a2, a3, pl; + char s[32]; + int n; + + list *pxlist = lp_allocz(tmp_linpool, sizeof(struct f_prefix_node)); + init_list(pxlist); + + errno = 0; + while (fgets(s, 32, f)) + { + if (s[0] == '\n') + return pxlist; + + n = sscanf(s, "%u.%u.%u.%u/%u", &a0, &a1, &a2, &a3, &pl); + + if (n != 5) + bt_abort_msg("Invalid content of trie_data"); + + struct f_prefix_node *px = lp_allocz(tmp_linpool, sizeof(struct f_prefix_node)); + net_fill_ip4(&px->prefix.net, ip4_build(a0, a1, a2, a3), pl); + px->prefix.lo = pl; + px->prefix.hi = plus ? IP4_MAX_PREFIX_LENGTH : pl; + add_tail(pxlist, &px->n); + + char buf[64]; + bt_format_net(buf, 64, &px->prefix.net); + bt_debug("ADD %s{%d,%d}\n", buf, px->prefix.lo, px->prefix.hi); } + + bt_syscall(errno, "fgets()"); + return EMPTY_LIST(*pxlist) ? NULL : pxlist; } +/* + * Open file, read multiple sequences of prefixes from it. Fill @data with + * prefix lists and @trie with generated tries. Return number of sequences / + * tries. Use separate linpool @lp0 for prefix lists and @lp1 for tries. + * Arg @plus means prefix should include all longer ones. + */ static int -t_match_net(void) +read_prefix_file(const char *filename, int plus, + list *data[], struct f_trie *trie[]) +{ + FILE *f = fopen(filename, "r"); + bt_syscall(!f, "fopen(%s)", filename); + + int n = 0; + list *pxlist; + while (pxlist = read_prefix_list(f, 0, plus)) + { + data[n] = pxlist; + trie[n] = make_trie_from_prefix_list(pxlist); + bt_debug("NEXT\n"); + n++; + } + + fclose(f); + bt_debug("DONE reading %d tries\n", n); + + return n; +} + +/* + * Select random subset of @dn prefixes from prefix list @src of length @sn, + * and store them to buffer @dst (of size @dn). Prefixes may be chosen multiple + * times. Randomize order of prefixes in @dst buffer. + */ +static void +select_random_prefix_subset(list *src[], net_addr dst[], int sn, int dn) +{ + int pn = 0; + + if (!dn) + return; + + /* Compute total prefix number */ + for (int i = 0; i < sn; i++) + pn += list_length(src[i]); + + /* Change of selecting a prefix */ + int rnd = (pn / dn) + 10; + int n = 0; + + /* Iterate indefinitely over src array */ + for (int i = 0; 1; i++, i = (i < sn) ? i : 0) + { + struct f_prefix_node *px; + WALK_LIST(px, *src[i]) + { + if (xrandom(rnd) != 0) + continue; + + net_copy(&dst[n], &px->prefix.net); + n++; + + /* We have enough */ + if (n == dn) + goto done; + } + } + +done: + /* Shuffle networks */ + for (int i = 0; i < dn; i++) + { + int j = xrandom(dn); + + if (i == j) + continue; + + net_addr tmp; + net_copy(&tmp, &dst[i]); + net_copy(&dst[i], &dst[j]); + net_copy(&dst[j], &tmp); + } +} + +/* Fill @dst buffer with @dn randomly generated /32 prefixes */ +static void +make_random_addresses(net_addr dst[], int dn) +{ + for (int i = 0; i < dn; i++) + net_fill_ip4(&dst[i], ip4_from_u32((u32) bt_random()), IP4_MAX_PREFIX_LENGTH); +} + +static void +test_match_net(list *prefixes, struct f_trie *trie, const net_addr *net) +{ + char buf[64]; + bt_format_net(buf, 64, net); + bt_debug("TEST %s\n", buf); + + int should_be = is_prefix_included(prefixes, net); + int is_there = trie_match_net(trie, net); + + bt_assert_msg(should_be == is_there, "Prefix %s %s match", buf, + (should_be ? "should" : "should not")); +} + +static int +t_match_random_net(void) { bt_bird_init(); bt_config_parse(BT_CONFIG_SIMPLE); - uint round; - for (round = 0; round < TESTS_NUM; round++) + int v6 = 0; + for (int round = 0; round < TESTS_NUM; round++) { - list prefixes; /* of structs f_extended_prefix */ - init_list(&prefixes); - struct f_trie *trie = f_new_trie(config->mem, 0); + list *prefixes = make_random_prefix_list(PREFIXES_NUM, v6, 0); + struct f_trie *trie = make_trie_from_prefix_list(prefixes); - generate_random_ipv6_prefixes(&prefixes); - struct f_prefix_node *n; - WALK_LIST(n, prefixes) + for (int i = 0; i < PREFIX_TESTS_NUM; i++) { - trie_add_prefix(trie, &n->prefix.net, n->prefix.lo, n->prefix.hi); + net_addr net; + get_random_net(&net, v6); + test_match_net(prefixes, trie, &net); } - int i; - for (i = 0; i < PREFIX_TESTS_NUM; i++) + v6 = !v6; + tmp_flush(); + } + + bt_bird_cleanup(); + return 1; +} + +static int +t_match_inner_net(void) +{ + bt_bird_init(); + bt_config_parse(BT_CONFIG_SIMPLE); + + int v6 = 0; + for (int round = 0; round < TESTS_NUM; round++) + { + list *prefixes = make_random_prefix_list(PREFIXES_NUM, v6, 0); + struct f_trie *trie = make_trie_from_prefix_list(prefixes); + + struct f_prefix_node *n = HEAD(*prefixes); + for (int i = 0; i < PREFIX_TESTS_NUM; i++) { - struct f_prefix f = get_random_ip6_prefix(); - bt_debug("TEST\t" PRIip6 "/%d\n", ARGip6(net6_prefix(&f.net)), f.net.pxlen); + net_addr net; + get_inner_net(&net, &n->prefix); + test_match_net(prefixes, trie, &net); - int should_be = is_prefix_included(&prefixes, &f); - int is_there = trie_match_net(trie, &f.net); - bt_assert_msg(should_be == is_there, "Prefix " PRIip6 "/%d %s", ARGip6(net6_prefix(&f.net)), f.net.pxlen, (should_be ? "should be found in trie" : "should not be found in trie")); + n = NODE_VALID(NODE_NEXT(n)) ? NODE_NEXT(n) : HEAD(*prefixes); } - struct f_prefix_node *nxt; - WALK_LIST_DELSAFE(n, nxt, prefixes) + v6 = !v6; + tmp_flush(); + } + + bt_bird_cleanup(); + return 1; +} + +static int +t_match_outer_net(void) +{ + bt_bird_init(); + bt_config_parse(BT_CONFIG_SIMPLE); + + int v6 = 0; + for (int round = 0; round < TESTS_NUM; round++) + { + list *prefixes = make_random_prefix_list(PREFIXES_NUM, v6, 0); + struct f_trie *trie = make_trie_from_prefix_list(prefixes); + + struct f_prefix_node *n = HEAD(*prefixes); + for (int i = 0; i < PREFIX_TESTS_NUM; i++) { - free(n); + net_addr net; + get_outer_net(&net, &n->prefix); + test_match_net(prefixes, trie, &net); + + n = NODE_VALID(NODE_NEXT(n)) ? NODE_NEXT(n) : HEAD(*prefixes); } + + v6 = !v6; + tmp_flush(); } + v6 = !v6; + bt_bird_cleanup(); + return 1; +} + +/* + * Read prefixes from @filename, build set of tries, prepare test data and do + * PREFIX_BENCH_NUM trie lookups. With @plus = 0, use random subset of known + * prefixes as test data, with @plus = 1, use randomly generated /32 prefixes + * as test data. + */ +static int +benchmark_trie_dataset(const char *filename, int plus) +{ + int n = 0; + list *data[TRIE_BUFFER_SIZE]; + struct f_trie *trie[TRIE_BUFFER_SIZE]; + net_addr *nets; + + bt_reset_suite_case_timer(); + bt_log_suite_case_result(1, "Reading %s", filename, n); + n = read_prefix_file(filename, plus, data, trie); + bt_log_suite_case_result(1, "Read prefix data, %d lists, ", n); + + size_t trie_size = rmemsize(tmp_linpool).effective * 1000 / (1024*1024); + bt_log_suite_case_result(1, "Trie size %u.%03u MB", + (uint) (trie_size / 1000), (uint) (trie_size % 1000)); + + int t = PREFIX_BENCH_NUM / n; + int tb = MIN(t, TEST_BUFFER_SIZE); + nets = tmp_alloc(tb * sizeof(net_addr)); + + if (!plus) + select_random_prefix_subset(data, nets, n, tb); + else + make_random_addresses(nets, tb); + + bt_log_suite_case_result(1, "Make test data, %d (%d) tests", t, tb); + bt_reset_suite_case_timer(); + + /* + int match = 0; + for (int i = 0; i < t; i++) + for (int j = 0; j < n; j++) + test_match_net(data[j], trie[j], &nets[i]); + */ + + int match = 0; + for (int i = 0; i < t; i++) + for (int j = 0; j < n; j++) + if (trie_match_net(trie[j], &nets[i % TEST_BUFFER_SIZE])) + match++; + + bt_log_suite_case_result(1, "Matching done, %d / %d matches", match, t * n); + + tmp_flush(); + return 1; +} + +static int UNUSED +t_bench_trie_datasets_subset(void) +{ + bt_bird_init(); + bt_config_parse(BT_CONFIG_SIMPLE); + + /* Specific datasets, not included */ + benchmark_trie_dataset("trie-data-bgp-1", 0); + benchmark_trie_dataset("trie-data-bgp-10", 0); + benchmark_trie_dataset("trie-data-bgp-100", 0); + benchmark_trie_dataset("trie-data-bgp-1000", 0); + bt_bird_cleanup(); + return 1; } +static int UNUSED +t_bench_trie_datasets_random(void) +{ + bt_bird_init(); + bt_config_parse(BT_CONFIG_SIMPLE); + + /* Specific datasets, not included */ + benchmark_trie_dataset("trie-data-bgp-1", 1); + benchmark_trie_dataset("trie-data-bgp-10", 1); + benchmark_trie_dataset("trie-data-bgp-100", 1); + benchmark_trie_dataset("trie-data-bgp-1000", 1); + + bt_bird_cleanup(); + + return 1; +} + + static int t_trie_same(void) { bt_bird_init(); bt_config_parse(BT_CONFIG_SIMPLE); - int round; - for (round = 0; round < TESTS_NUM*4; round++) + int v6 = 0; + for (int round = 0; round < TESTS_NUM*4; round++) { - struct f_trie * trie1 = f_new_trie(config->mem, 0); - struct f_trie * trie2 = f_new_trie(config->mem, 0); + list *prefixes = make_random_prefix_list(100 * PREFIXES_NUM, v6, 0); + struct f_trie *trie1 = f_new_trie(tmp_linpool, 0); + struct f_trie *trie2 = f_new_trie(tmp_linpool, 0); - list prefixes; /* a list of f_extended_prefix structures */ - init_list(&prefixes); - int i; - for (i = 0; i < 100; i++) - generate_random_ipv6_prefixes(&prefixes); + struct f_prefix_node *n; + WALK_LIST(n, *prefixes) + trie_add_prefix(trie1, &n->prefix.net, n->prefix.lo, n->prefix.hi); + + WALK_LIST_BACKWARDS(n, *prefixes) + trie_add_prefix(trie2, &n->prefix.net, n->prefix.lo, n->prefix.hi); + + bt_assert(trie_same(trie1, trie2)); + + v6 = !v6; + tmp_flush(); + } + + bt_bird_cleanup(); + return 1; +} + +static inline void +log_networks(const net_addr *a, const net_addr *b) +{ + if (bt_verbose >= BT_VERBOSE_ABSOLUTELY_ALL) + { + char buf0[64]; + char buf1[64]; + bt_format_net(buf0, 64, a); + bt_format_net(buf1, 64, b); + bt_debug("Found %s expected %s\n", buf0, buf1); + } +} + +static int +t_trie_walk(void) +{ + bt_bird_init(); + bt_config_parse(BT_CONFIG_SIMPLE); + + for (int round = 0; round < TESTS_NUM*8; round++) + { + int level = round / TESTS_NUM; + int v6 = level % 2; + int num = PREFIXES_NUM * (int[]){1, 10, 100, 1000}[level / 2]; + int pos = 0, end = 0; + list *prefixes = make_random_prefix_list(num, v6, 1); + struct f_trie *trie = make_trie_from_prefix_list(prefixes); + struct f_prefix *pxset = malloc((num + 1) * sizeof(struct f_prefix)); struct f_prefix_node *n; - WALK_LIST(n, prefixes) + WALK_LIST(n, *prefixes) + pxset[pos++] = n->prefix; + memset(&pxset[pos], 0, sizeof (struct f_prefix)); + + qsort(pxset, num, sizeof(struct f_prefix), compare_prefixes); + + + /* Full walk */ + bt_debug("Full walk (round %d, %d nets)\n", round, num); + + pos = 0; + uint pxc = 0; + TRIE_WALK(trie, net, NULL) { - trie_add_prefix(trie1, &n->prefix.net, n->prefix.lo, n->prefix.hi); + log_networks(&net, &pxset[pos].net); + bt_assert(net_equal(&net, &pxset[pos].net)); + + /* Skip possible duplicates */ + while (net_equal(&pxset[pos].net, &pxset[pos + 1].net)) + pos++; + + pos++; + pxc++; } - WALK_LIST_BACKWARDS(n, prefixes) + TRIE_WALK_END; + + bt_assert(pos == num); + bt_assert(pxc == trie->prefix_count); + bt_debug("Full walk done\n"); + + + /* Prepare net for subnet walk - start with random prefix */ + pos = bt_random() % num; + end = pos + (int[]){2, 2, 3, 4}[level / 2]; + end = MIN(end, num); + + struct f_prefix from = pxset[pos]; + + /* Find a common superprefix to several subsequent prefixes */ + for (; pos < end; pos++) { - trie_add_prefix(trie2, &n->prefix.net, n->prefix.lo, n->prefix.hi); + if (net_equal(&from.net, &pxset[pos].net)) + continue; + + int common = !v6 ? + ip4_pxlen(net4_prefix(&from.net), net4_prefix(&pxset[pos].net)) : + ip6_pxlen(net6_prefix(&from.net), net6_prefix(&pxset[pos].net)); + from.net.pxlen = MIN(from.net.pxlen, common); + + if (!v6) + ((net_addr_ip4 *) &from.net)->prefix = + ip4_and(net4_prefix(&from.net), net4_prefix(&pxset[pos].net)); + else + ((net_addr_ip6 *) &from.net)->prefix = + ip6_and(net6_prefix(&from.net), net6_prefix(&pxset[pos].net)); } - bt_assert(trie_same(trie1, trie2)); + /* Fix irrelevant bits */ + if (!v6) + ((net_addr_ip4 *) &from.net)->prefix = + ip4_and(net4_prefix(&from.net), ip4_mkmask(net4_pxlen(&from.net))); + else + ((net_addr_ip6 *) &from.net)->prefix = + ip6_and(net6_prefix(&from.net), ip6_mkmask(net6_pxlen(&from.net))); + + + /* Find initial position for final prefix */ + for (pos = 0; pos < num; pos++) + if (compare_prefixes(&pxset[pos], &from) >= 0) + break; + + int p0 = pos; + char buf0[64]; + bt_format_net(buf0, 64, &from.net); + bt_debug("Subnet walk for %s (round %d, %d nets)\n", buf0, round, num); + + /* Subnet walk */ + TRIE_WALK(trie, net, &from.net) + { + log_networks(&net, &pxset[pos].net); + bt_assert(net_equal(&net, &pxset[pos].net)); + bt_assert(net_in_netX(&net, &from.net)); + + /* Skip possible duplicates */ + while (net_equal(&pxset[pos].net, &pxset[pos + 1].net)) + pos++; - struct f_prefix_node *nxt; - WALK_LIST_DELSAFE(n, nxt, prefixes) + pos++; + } + TRIE_WALK_END; + + bt_assert((pos == num) || !net_in_netX(&pxset[pos].net, &from.net)); + bt_debug("Subnet walk done for %s (found %d nets)\n", buf0, pos - p0); + + tmp_flush(); + } + + bt_bird_cleanup(); + return 1; +} + +static int +find_covering_nets(struct f_prefix *prefixes, int num, const net_addr *net, net_addr *found) +{ + struct f_prefix key; + net_addr *n = &key.net; + int found_num = 0; + + net_copy(n, net); + + while (1) + { + struct f_prefix *px = + bsearch(&key, prefixes, num, sizeof(struct f_prefix), compare_prefixes); + + if (px) + { + net_copy(&found[found_num], n); + found_num++; + } + + if (n->pxlen == 0) + return found_num; + + n->pxlen--; + + if (n->type == NET_IP4) + ip4_clrbit(&((net_addr_ip4 *) n)->prefix, n->pxlen); + else + ip6_clrbit(&((net_addr_ip6 *) n)->prefix, n->pxlen); + } +} + +static int +t_trie_walk_to_root(void) +{ + bt_bird_init(); + bt_config_parse(BT_CONFIG_SIMPLE); + + for (int round = 0; round < TESTS_NUM * 4; round++) + { + int level = round / TESTS_NUM; + int v6 = level % 2; + int num = PREFIXES_NUM * (int[]){32, 512}[level / 2]; + int pos = 0; + int st = 0, sn = 0, sm = 0; + + list *prefixes = make_random_prefix_list(num, v6, 1); + struct f_trie *trie = make_trie_from_prefix_list(prefixes); + struct f_prefix *pxset = malloc((num + 1) * sizeof(struct f_prefix)); + + struct f_prefix_node *pxn; + WALK_LIST(pxn, *prefixes) + pxset[pos++] = pxn->prefix; + memset(&pxset[pos], 0, sizeof (struct f_prefix)); + + qsort(pxset, num, sizeof(struct f_prefix), compare_prefixes); + + int i; + for (i = 0; i < (PREFIX_TESTS_NUM / 10); i++) { - free(n); + net_addr from; + get_random_net(&from, v6); + + net_addr found[129]; + int found_num = find_covering_nets(pxset, num, &from, found); + int n = 0; + + if (bt_verbose >= BT_VERBOSE_ABSOLUTELY_ALL) + { + char buf[64]; + bt_format_net(buf, 64, &from); + bt_debug("Lookup for %s (expect %d)\n", buf, found_num); + } + + /* Walk to root, separate for IPv4 and IPv6 */ + if (!v6) + { + TRIE_WALK_TO_ROOT_IP4(trie, (net_addr_ip4 *) &from, net) + { + log_networks((net_addr *) &net, &found[n]); + bt_assert((n < found_num) && net_equal((net_addr *) &net, &found[n])); + n++; + } + TRIE_WALK_TO_ROOT_END; + } + else + { + TRIE_WALK_TO_ROOT_IP6(trie, (net_addr_ip6 *) &from, net) + { + log_networks((net_addr *) &net, &found[n]); + bt_assert((n < found_num) && net_equal((net_addr *) &net, &found[n])); + n++; + } + TRIE_WALK_TO_ROOT_END; + } + + bt_assert(n == found_num); + + /* Stats */ + st += n; + sn += !!n; + sm = MAX(sm, n); } + + bt_debug("Success in %d / %d, sum %d, max %d\n", sn, i, st, sm); + + tmp_flush(); } + bt_bird_cleanup(); return 1; } @@ -179,8 +885,15 @@ main(int argc, char *argv[]) { bt_init(argc, argv); - bt_test_suite(t_match_net, "Testing random prefix matching"); + bt_test_suite(t_match_random_net, "Testing random prefix matching"); + bt_test_suite(t_match_inner_net, "Testing random inner prefix matching"); + bt_test_suite(t_match_outer_net, "Testing random outer prefix matching"); bt_test_suite(t_trie_same, "A trie filled forward should be same with a trie filled backward."); + bt_test_suite(t_trie_walk, "Testing TRIE_WALK() on random tries"); + bt_test_suite(t_trie_walk_to_root, "Testing TRIE_WALK_TO_ROOT() on random tries"); + + // bt_test_suite(t_bench_trie_datasets_subset, "Benchmark tries from datasets by random subset of nets"); + // bt_test_suite(t_bench_trie_datasets_random, "Benchmark tries from datasets by generated addresses"); return bt_exit_value(); } diff --git a/lib/birdlib.h b/lib/birdlib.h index 431b7c0d..81d4908a 100644 --- a/lib/birdlib.h +++ b/lib/birdlib.h @@ -32,6 +32,9 @@ struct align_probe { char x; long int y; }; #define MAX(a,b) MAX_(a,b) #endif +#define ROUND_DOWN_POW2(a,b) ((a) & ~((b)-1)) +#define ROUND_UP_POW2(a,b) (((a)+((b)-1)) & ~((b)-1)) + #define U64(c) UINT64_C(c) #define ABS(a) ((a)>=0 ? (a) : -(a)) #define DELTA(a,b) (((a)>=(b))?(a)-(b):(b)-(a)) diff --git a/lib/bitmap_test.c b/lib/bitmap_test.c index 0595a4d0..07860c94 100644 --- a/lib/bitmap_test.c +++ b/lib/bitmap_test.c @@ -24,7 +24,6 @@ t_bmap_set_clear_random(void) { struct bmap b; - resource_init(); bmap_init(&b, &root_pool, 1024); char expected[MAX_NUM] = {}; @@ -60,7 +59,6 @@ t_hmap_set_clear_random(void) { struct hmap b; - resource_init(); hmap_init(&b, &root_pool, 1024); char expected[MAX_NUM] = {}; @@ -119,7 +117,6 @@ t_hmap_set_clear_fill(void) { struct hmap b; - resource_init(); hmap_init(&b, &root_pool, 1024); char expected[MAX_NUM] = {}; diff --git a/lib/buffer_test.c b/lib/buffer_test.c index 5b7de330..0629e901 100644 --- a/lib/buffer_test.c +++ b/lib/buffer_test.c @@ -41,7 +41,6 @@ fill_expected_array(void) static void init_buffer(void) { - resource_init(); buffer_pool = &root_pool; BUFFER_INIT(buf, buffer_pool, MAX_NUM); } diff --git a/lib/event.c b/lib/event.c index 273447e0..33dc00b0 100644 --- a/lib/event.c +++ b/lib/event.c @@ -157,6 +157,7 @@ ev_run_list(event_list *l) io_log_event(e->hook, e->data); ev_run(e); + tmp_flush(); } return !EMPTY_LIST(*l); @@ -184,6 +185,7 @@ ev_run_list_limited(event_list *l, uint limit) io_log_event(e->hook, e->data); ev_run(e); + tmp_flush(); limit--; } diff --git a/lib/event_test.c b/lib/event_test.c index e1215bba..e1fbea8f 100644 --- a/lib/event_test.c +++ b/lib/event_test.c @@ -53,7 +53,6 @@ t_ev_run_list(void) { int i; - resource_init(); olock_init(); timer_init(); io_init(); diff --git a/lib/flowspec_test.c b/lib/flowspec_test.c index ed4afe51..03649b99 100644 --- a/lib/flowspec_test.c +++ b/lib/flowspec_test.c @@ -446,10 +446,7 @@ t_validation6(void) static int t_builder4(void) { - resource_init(); - struct flow_builder *fb = flow_builder_init(&root_pool); - linpool *lp = lp_new_default(&root_pool); /* Expectation */ @@ -492,7 +489,7 @@ t_builder4(void) flow_builder_set_type(fb, FLOW_TYPE_TCP_FLAGS); flow_builder_add_op_val(fb, 0, 0x55); - net_addr_flow4 *res = flow_builder4_finalize(fb, lp); + net_addr_flow4 *res = flow_builder4_finalize(fb, tmp_linpool); bt_assert(memcmp(res, expect, expect->length) == 0); @@ -529,8 +526,6 @@ t_builder6(void) { net_addr_ip6 ip; - resource_init(); - linpool *lp = lp_new_default(&root_pool); struct flow_builder *fb = flow_builder_init(&root_pool); fb->ipv6 = 1; @@ -574,7 +569,7 @@ t_builder6(void) flow_builder_set_type(fb, FLOW_TYPE_LABEL); flow_builder_add_op_val(fb, 0, 0x55); - net_addr_flow6 *res = flow_builder6_finalize(fb, lp); + net_addr_flow6 *res = flow_builder6_finalize(fb, tmp_linpool); bt_assert(memcmp(res, expect, expect->length) == 0); /* Reverse order */ @@ -601,7 +596,7 @@ t_builder6(void) flow_builder_set_type(fb, FLOW_TYPE_DST_PREFIX); flow_builder6_add_pfx(fb, &ip, 61); - res = flow_builder6_finalize(fb, lp); + res = flow_builder6_finalize(fb, tmp_linpool); bt_assert(memcmp(res, expect, expect->length) == 0); return 1; diff --git a/lib/hash_test.c b/lib/hash_test.c index 59beb7c0..4bce7017 100644 --- a/lib/hash_test.c +++ b/lib/hash_test.c @@ -61,7 +61,6 @@ dump_nodes(void) static void init_hash_(uint order) { - resource_init(); my_pool = rp_new(&root_pool, "Test pool"); HASH_INIT(hash, my_pool, order); @@ -279,11 +279,35 @@ static inline uint ip6_pxlen(ip6_addr a, ip6_addr b) return 32 * i + 31 - u32_log2(a.addr[i] ^ b.addr[i]); } +static inline int ip4_prefix_equal(ip4_addr a, ip4_addr b, uint n) +{ + return (_I(a) ^ _I(b)) < ((u64) 1 << (32 - n)); +} + +static inline int ip6_prefix_equal(ip6_addr a, ip6_addr b, uint n) +{ + uint n0 = n / 32; + uint n1 = n % 32; + + return + ((n0 <= 0) || (_I0(a) == _I0(b))) && + ((n0 <= 1) || (_I1(a) == _I1(b))) && + ((n0 <= 2) || (_I2(a) == _I2(b))) && + ((n0 <= 3) || (_I3(a) == _I3(b))) && + (!n1 || ((a.addr[n0] ^ b.addr[n0]) < (1u << (32 - n1)))); +} + static inline u32 ip4_getbit(ip4_addr a, uint pos) -{ return _I(a) & (0x80000000 >> pos); } +{ return (_I(a) >> (31 - pos)) & 1; } + +static inline u32 ip4_getbits(ip4_addr a, uint pos, uint n) +{ return (_I(a) >> ((32 - n) - pos)) & ((1u << n) - 1); } static inline u32 ip6_getbit(ip6_addr a, uint pos) -{ return a.addr[pos / 32] & (0x80000000 >> (pos % 32)); } +{ return (a.addr[pos / 32] >> (31 - (pos % 32))) & 0x1; } + +static inline u32 ip6_getbits(ip6_addr a, uint pos, uint n) +{ return (a.addr[pos / 32] >> ((32 - n) - (pos % 32))) & ((1u << n) - 1); } static inline u32 ip4_setbit(ip4_addr *a, uint pos) { return _I(*a) |= (0x80000000 >> pos); } @@ -297,6 +321,13 @@ static inline u32 ip4_clrbit(ip4_addr *a, uint pos) static inline u32 ip6_clrbit(ip6_addr *a, uint pos) { return a->addr[pos / 32] &= ~(0x80000000 >> (pos % 32)); } +static inline ip4_addr ip4_setbits(ip4_addr a, uint pos, uint val) +{ _I(a) |= val << (31 - pos); return a; } + +static inline ip6_addr ip6_setbits(ip6_addr a, uint pos, uint val) +{ a.addr[pos / 32] |= val << (31 - pos % 32); return a; } + + static inline ip4_addr ip4_opposite_m1(ip4_addr a) { return _MI4(_I(a) ^ 1); } diff --git a/lib/ip_test.c b/lib/ip_test.c index 36d10d68..eee0a427 100644 --- a/lib/ip_test.c +++ b/lib/ip_test.c @@ -167,6 +167,70 @@ t_ip6_ntop(void) return bt_assert_batch(test_vectors, test_ipa_ntop, bt_fmt_ipa, bt_fmt_str); } +static int +t_ip4_prefix_equal(void) +{ + bt_assert( ip4_prefix_equal(ip4_from_u32(0x12345678), ip4_from_u32(0x1234ffff), 16)); + bt_assert(!ip4_prefix_equal(ip4_from_u32(0x12345678), ip4_from_u32(0x1234ffff), 17)); + bt_assert( ip4_prefix_equal(ip4_from_u32(0x12345678), ip4_from_u32(0x12345000), 21)); + bt_assert(!ip4_prefix_equal(ip4_from_u32(0x12345678), ip4_from_u32(0x12345000), 22)); + + bt_assert( ip4_prefix_equal(ip4_from_u32(0x00000000), ip4_from_u32(0xffffffff), 0)); + bt_assert( ip4_prefix_equal(ip4_from_u32(0x12345678), ip4_from_u32(0x12345678), 0)); + + bt_assert( ip4_prefix_equal(ip4_from_u32(0x12345678), ip4_from_u32(0x12345678), 32)); + bt_assert(!ip4_prefix_equal(ip4_from_u32(0x12345678), ip4_from_u32(0x12345679), 32)); + bt_assert(!ip4_prefix_equal(ip4_from_u32(0x12345678), ip4_from_u32(0x92345678), 32)); + + return 1; +} + +static int +t_ip6_prefix_equal(void) +{ + bt_assert( ip6_prefix_equal(ip6_build(0x20010db8, 0x12345678, 0x10101010, 0x20202020), + ip6_build(0x20010db8, 0x1234ffff, 0xfefefefe, 0xdcdcdcdc), + 48)); + + bt_assert(!ip6_prefix_equal(ip6_build(0x20010db8, 0x12345678, 0x10101010, 0x20202020), + ip6_build(0x20010db8, 0x1234ffff, 0xfefefefe, 0xdcdcdcdc), + 49)); + + bt_assert(!ip6_prefix_equal(ip6_build(0x20010db8, 0x12345678, 0x10101010, 0x20202020), + ip6_build(0x20020db8, 0x12345678, 0xfefefefe, 0xdcdcdcdc), + 48)); + + bt_assert( ip6_prefix_equal(ip6_build(0x20010db8, 0x12345678, 0x10101010, 0x20202020), + ip6_build(0x20010db8, 0x12345678, 0xfefefefe, 0xdcdcdcdc), + 64)); + + bt_assert(!ip6_prefix_equal(ip6_build(0x20010db8, 0x12345678, 0x10101010, 0x20202020), + ip6_build(0x20010db8, 0x1234567e, 0xfefefefe, 0xdcdcdcdc), + 64)); + + bt_assert( ip6_prefix_equal(ip6_build(0x20010db8, 0x12345678, 0x10101010, 0x20002020), + ip6_build(0x20010db8, 0x12345678, 0x10101010, 0x20202020), + 106)); + + bt_assert(!ip6_prefix_equal(ip6_build(0x20010db8, 0x12345678, 0x10101010, 0x20002020), + ip6_build(0x20010db8, 0x12345678, 0x10101010, 0x20202020), + 107)); + + bt_assert( ip6_prefix_equal(ip6_build(0xfeef0db8, 0x87654321, 0x10101010, 0x20202020), + ip6_build(0x20010db8, 0x12345678, 0xfefefefe, 0xdcdcdcdc), + 0)); + + bt_assert( ip6_prefix_equal(ip6_build(0x20010db8, 0x12345678, 0x10101010, 0x20202020), + ip6_build(0x20010db8, 0x12345678, 0x10101010, 0x20202020), + 128)); + + bt_assert(!ip6_prefix_equal(ip6_build(0x20010db8, 0x12345678, 0x10101010, 0x20202020), + ip6_build(0x20010db8, 0x12345678, 0x10101010, 0x20202021), + 128)); + + return 1; +} + int main(int argc, char *argv[]) { @@ -176,6 +240,8 @@ main(int argc, char *argv[]) bt_test_suite(t_ip6_pton, "Converting IPv6 string to ip6_addr struct"); bt_test_suite(t_ip4_ntop, "Converting ip4_addr struct to IPv4 string"); bt_test_suite(t_ip6_ntop, "Converting ip6_addr struct to IPv6 string"); + bt_test_suite(t_ip4_prefix_equal, "Testing ip4_prefix_equal()"); + bt_test_suite(t_ip6_prefix_equal, "Testing ip6_prefix_equal()"); return bt_exit_value(); } diff --git a/lib/lists.h b/lib/lists.h index 479f4ed1..7e6d5467 100644 --- a/lib/lists.h +++ b/lib/lists.h @@ -42,6 +42,7 @@ typedef union list { /* In fact two overlayed nodes */ }; } list; +#define STATIC_LIST_INIT(name) name = { .head = &name.tail_node, .tail = &name.head_node, .null = NULL } #define NODE (node *) #define HEAD(list) ((void *)((list).head)) diff --git a/lib/mempool.c b/lib/mempool.c index 8f300b81..c75f1f5b 100644 --- a/lib/mempool.c +++ b/lib/mempool.c @@ -37,16 +37,17 @@ const int lp_chunk_size = sizeof(struct lp_chunk); struct linpool { resource r; byte *ptr, *end; - pool *p; struct lp_chunk *first, *current; /* Normal (reusable) chunks */ struct lp_chunk *first_large; /* Large chunks */ uint chunk_size, threshold, total:31, use_pages:1, total_large; }; +_Thread_local linpool *tmp_linpool; + static void lp_free(resource *); static void lp_dump(resource *); static resource *lp_lookup(resource *, unsigned long); -static size_t lp_memsize(resource *r); +static struct resmem lp_memsize(resource *r); static struct resclass lp_class = { "LinPool", @@ -70,7 +71,6 @@ linpool *lp_new(pool *p, uint blk) { linpool *m = ralloc(p, &lp_class); - m->p = p; if (!blk) { m->use_pages = 1; @@ -130,7 +130,7 @@ lp_alloc(linpool *m, uint size) { /* Need to allocate a new chunk */ if (m->use_pages) - c = alloc_page(m->p); + c = alloc_page(); else c = xmalloc(sizeof(struct lp_chunk) + m->chunk_size); @@ -271,7 +271,7 @@ lp_free(resource *r) { c = d->next; if (m->use_pages) - free_page(m->p, d); + free_page(d); else xfree(d); } @@ -302,21 +302,28 @@ lp_dump(resource *r) m->total_large); } -static size_t +static struct resmem lp_memsize(resource *r) { linpool *m = (linpool *) r; - struct lp_chunk *c; - int cnt = 0; + struct resmem sz = { + .overhead = sizeof(struct linpool) + ALLOC_OVERHEAD, + }; - for(c=m->first; c; c=c->next) - cnt++; - for(c=m->first_large; c; c=c->next) - cnt++; + for (struct lp_chunk *c = m->first_large; c; c = c->next) + { + sz.effective += c->size; + sz.overhead += lp_chunk_size + ALLOC_OVERHEAD; + } + + uint regular = 0; + for (struct lp_chunk *c = m->first; c; c = c->next) + regular++; + + sz.effective += m->chunk_size * regular; + sz.overhead += (lp_chunk_size + ALLOC_OVERHEAD) * regular; - return ALLOC_OVERHEAD + sizeof(struct linpool) + - cnt * (ALLOC_OVERHEAD + sizeof(struct lp_chunk)) + - m->total + m->total_large; + return sz; } @@ -38,6 +38,7 @@ #define NB_IP (NB_IP4 | NB_IP6) #define NB_VPN (NB_VPN4 | NB_VPN6) +#define NB_ROA (NB_ROA4 | NB_ROA6) #define NB_FLOW (NB_FLOW4 | NB_FLOW6) #define NB_DEST (NB_IP | NB_IP6_SADR | NB_VPN | NB_MPLS) #define NB_ANY 0xffffffff diff --git a/lib/resource.c b/lib/resource.c index e80b315b..a179afe3 100644 --- a/lib/resource.c +++ b/lib/resource.c @@ -2,6 +2,7 @@ * BIRD Resource Manager * * (c) 1998--2000 Martin Mares <mj@ucw.cz> + * (c) 2021 Maria Matejka <mq@jmq.cz> * * Can be freely distributed and used under the terms of the GNU GPL. */ @@ -31,22 +32,13 @@ struct pool { resource r; list inside; - struct pool_pages *pages; const char *name; }; -struct pool_pages { - uint free; - uint used; - void *ptr[0]; -}; - -#define POOL_PAGES_MAX ((page_size - sizeof(struct pool_pages)) / sizeof (void *)) - static void pool_dump(resource *); static void pool_free(resource *); static resource *pool_lookup(resource *, unsigned long); -static size_t pool_memsize(resource *P); +static struct resmem pool_memsize(resource *P); static struct resclass pool_class = { "Pool", @@ -59,9 +51,6 @@ static struct resclass pool_class = { pool root_pool; -void *alloc_sys_page(void); -void free_sys_page(void *); - static int indent; /** @@ -94,14 +83,6 @@ pool_free(resource *P) xfree(r); r = rr; } - - if (p->pages) - { - ASSERT_DIE(!p->pages->used); - for (uint i=0; i<p->pages->free; i++) - free_sys_page(p->pages->ptr[i]); - free_sys_page(p->pages); - } } static void @@ -117,18 +98,22 @@ pool_dump(resource *P) indent -= 3; } -static size_t +static struct resmem pool_memsize(resource *P) { pool *p = (pool *) P; resource *r; - size_t sum = sizeof(pool) + ALLOC_OVERHEAD; + struct resmem sum = { + .effective = 0, + .overhead = sizeof(pool) + ALLOC_OVERHEAD, + }; WALK_LIST(r, p->inside) - sum += rmemsize(r); - - if (p->pages) - sum += page_size * (p->pages->used + p->pages->free + 1); + { + struct resmem add = rmemsize(r); + sum.effective += add.effective; + sum.overhead += add.overhead; + } return sum; } @@ -216,14 +201,17 @@ rdump(void *res) debug("NULL\n"); } -size_t +struct resmem rmemsize(void *res) { resource *r = res; if (!r) - return 0; + return (struct resmem) {}; if (!r->class->memsize) - return r->class->size + ALLOC_OVERHEAD; + return (struct resmem) { + .effective = r->class->size - sizeof(resource), + .overhead = ALLOC_OVERHEAD + sizeof(resource), + }; return r->class->memsize(r); } @@ -282,9 +270,12 @@ rlookup(unsigned long a) void resource_init(void) { + resource_sys_init(); + root_pool.r.class = &pool_class; root_pool.name = "Root"; init_list(&root_pool.inside); + tmp_init(&root_pool); } /** @@ -328,11 +319,14 @@ mbl_lookup(resource *r, unsigned long a) return NULL; } -static size_t +static struct resmem mbl_memsize(resource *r) { struct mblock *m = (struct mblock *) r; - return ALLOC_OVERHEAD + sizeof(struct mblock) + m->size; + return (struct resmem) { + .effective = m->size, + .overhead = ALLOC_OVERHEAD + sizeof(struct mblock), + }; } static struct resclass mb_class = { @@ -448,39 +442,6 @@ mb_free(void *m) rfree(b); } -void * -alloc_page(pool *p) -{ - if (!p->pages) - { - p->pages = alloc_sys_page(); - p->pages->free = 0; - p->pages->used = 1; - } - else - p->pages->used++; - - if (p->pages->free) - { - void *ptr = p->pages->ptr[--p->pages->free]; - bzero(ptr, page_size); - return ptr; - } - else - return alloc_sys_page(); -} - -void -free_page(pool *p, void *ptr) -{ - ASSERT_DIE(p->pages); - p->pages->used--; - - if (p->pages->free >= POOL_PAGES_MAX) - return free_sys_page(ptr); - else - p->pages->ptr[p->pages->free++] = ptr; -} #define STEP_UP(x) ((x) + (x)/2 + 4) diff --git a/lib/resource.h b/lib/resource.h index 26030aea..313b01dc 100644 --- a/lib/resource.h +++ b/lib/resource.h @@ -2,6 +2,7 @@ * BIRD Resource Manager * * (c) 1998--1999 Martin Mares <mj@ucw.cz> + * (c) 2021 Maria Matejka <mq@jmq.cz> * * Can be freely distributed and used under the terms of the GNU GPL. */ @@ -11,6 +12,11 @@ #include "lib/lists.h" +struct resmem { + size_t effective; /* Memory actually used for data storage */ + size_t overhead; /* Overhead memory imposed by allocator strategies */ +}; + /* Resource */ typedef struct resource { @@ -26,11 +32,11 @@ struct resclass { void (*free)(resource *); /* Freeing function */ void (*dump)(resource *); /* Dump to debug output */ resource *(*lookup)(resource *, unsigned long); /* Look up address (only for debugging) */ - size_t (*memsize)(resource *); /* Return size of memory used by the resource, may be NULL */ + struct resmem (*memsize)(resource *); /* Return size of memory used by the resource, may be NULL */ }; /* Estimate of system allocator overhead per item, for memory consumtion stats */ -#define ALLOC_OVERHEAD 8 +#define ALLOC_OVERHEAD 16 /* Generic resource manipulation */ @@ -40,7 +46,7 @@ void resource_init(void); pool *rp_new(pool *, const char *); /* Create new pool */ void rfree(void *); /* Free single resource */ void rdump(void *); /* Dump to debug output */ -size_t rmemsize(void *res); /* Return size of memory used by the resource */ +struct resmem rmemsize(void *res); /* Return size of memory used by the resource */ void rlookup(unsigned long); /* Look up address (only for debugging) */ void rmove(void *, pool *); /* Move to a different pool */ @@ -73,6 +79,15 @@ void lp_flush(linpool *); /* Free everything, but leave linpool */ void lp_save(linpool *m, lp_state *p); /* Save state */ void lp_restore(linpool *m, lp_state *p); /* Restore state */ +extern _Thread_local linpool *tmp_linpool; /* Temporary linpool autoflushed regularily */ + +#define tmp_alloc(sz) lp_alloc(tmp_linpool, sz) +#define tmp_allocu(sz) lp_allocu(tmp_linpool, sz) +#define tmp_allocz(sz) lp_allocz(tmp_linpool, sz) + +#define tmp_init(p) tmp_linpool = lp_new_default(p) +#define tmp_flush() lp_flush(tmp_linpool) + extern const int lp_chunk_size; #define LP_GAS 1024 #define LP_GOOD_SIZE(x) (((x + LP_GAS - 1) & (~(LP_GAS - 1))) - lp_chunk_size) @@ -94,12 +109,12 @@ void sl_free(slab *, void *); void buffer_realloc(void **buf, unsigned *size, unsigned need, unsigned item_size); +/* Allocator of whole pages; for use in slabs and other high-level allocators. */ extern long page_size; +void *alloc_page(void); +void free_page(void *); -/* Allocator of whole pages; for use in slabs and other high-level allocators. */ -void *alloc_page(pool *); -void free_page(pool *, void *); -#define PAGE_HEAD(x) ((void *) (((intptr_t) (x)) & ~(page_size-1))) +void resource_sys_init(void); #ifdef HAVE_LIBDMALLOC /* @@ -42,7 +42,7 @@ static void slab_free(resource *r); static void slab_dump(resource *r); static resource *slab_lookup(resource *r, unsigned long addr); -static size_t slab_memsize(resource *r); +static struct resmem slab_memsize(resource *r); #ifdef FAKE_SLAB @@ -128,7 +128,7 @@ slab_dump(resource *r) debug("(%d objects per %d bytes)\n", cnt, s->size); } -static size_t +static struct resmem slab_memsize(resource *r) { slab *s = (slab *) r; @@ -138,7 +138,10 @@ slab_memsize(resource *r) WALK_LIST(o, s->objs) cnt++; - return ALLOC_OVERHEAD + sizeof(struct slab) + cnt * (ALLOC_OVERHEAD + s->size); + return (struct resmem) { + .effective = cnt * s->size, + .overhead = ALLOC_OVERHEAD + sizeof(struct slab) + cnt * ALLOC_OVERHEAD, + }; } @@ -152,7 +155,6 @@ slab_memsize(resource *r) struct slab { resource r; - pool *p; uint obj_size, head_size, head_bitfield_len; uint objs_per_slab, num_empty_heads, data_size; list empty_heads, partial_heads, full_heads; @@ -178,7 +180,7 @@ struct sl_alignment { /* Magic structure for testing of alignment */ int x[0]; }; -#define SL_GET_HEAD(x) ((struct sl_head *) PAGE_HEAD(x)) +#define SL_GET_HEAD(x) ((struct sl_head *) (((uintptr_t) (x)) & ~(page_size-1))) /** * sl_new - create a new Slab @@ -192,10 +194,9 @@ slab * sl_new(pool *p, uint size) { slab *s = ralloc(p, &sl_class); - s->p = p; uint align = sizeof(struct sl_alignment); - if (align < sizeof(int)) - align = sizeof(int); + if (align < sizeof(void *)) + align = sizeof(void *); s->data_size = size; size = (size + align - 1) / align * align; s->obj_size = size; @@ -269,7 +270,7 @@ no_partial: s->num_empty_heads--; goto okay; } - h = alloc_page(s->p); + h = alloc_page(); #ifdef POISON memset(h, 0xba, page_size); #endif @@ -332,7 +333,7 @@ sl_free(slab *s, void *oo) #ifdef POISON memset(h, 0xde, page_size); #endif - free_page(s->p, h); + free_page(h); } else { @@ -349,11 +350,11 @@ slab_free(resource *r) struct sl_head *h, *g; WALK_LIST_DELSAFE(h, g, s->empty_heads) - free_page(s->p, h); + free_page(h); WALK_LIST_DELSAFE(h, g, s->partial_heads) - free_page(s->p, h); + free_page(h); WALK_LIST_DELSAFE(h, g, s->full_heads) - free_page(s->p, h); + free_page(h); } static void @@ -372,22 +373,33 @@ slab_dump(resource *r) debug("(%de+%dp+%df blocks per %d objs per %d bytes)\n", ec, pc, fc, s->objs_per_slab, s->obj_size); } -static size_t +static struct resmem slab_memsize(resource *r) { slab *s = (slab *) r; size_t heads = 0; struct sl_head *h; - WALK_LIST(h, s->empty_heads) + WALK_LIST(h, s->full_heads) heads++; + + size_t items = heads * s->objs_per_slab; + WALK_LIST(h, s->partial_heads) + { heads++; - WALK_LIST(h, s->full_heads) + items += h->num_full; + } + + WALK_LIST(h, s->empty_heads) heads++; -// return ALLOC_OVERHEAD + sizeof(struct slab) + heads * (ALLOC_OVERHEAD + page_size); - return ALLOC_OVERHEAD + sizeof(struct slab); /* The page sizes are accounted for in the pool */ + size_t eff = items * s->obj_size; + + return (struct resmem) { + .effective = eff, + .overhead = ALLOC_OVERHEAD + sizeof(struct slab) + heads * page_size - eff, + }; } static resource * diff --git a/lib/socket.h b/lib/socket.h index 96fedeeb..0b6ac589 100644 --- a/lib/socket.h +++ b/lib/socket.h @@ -123,6 +123,7 @@ extern int sk_priority_control; /* Suggested priority for control traffic, shou #define SKF_TTL_RX 0x08 /* Report TTL / Hop Limit for RX packets */ #define SKF_BIND 0x10 /* Bind datagram socket to given source address */ #define SKF_HIGH_PORT 0x20 /* Choose port from high range if possible */ +#define SKF_FREEBIND 0x40 /* Allow socket to bind to a nonlocal address */ #define SKF_THREAD 0x100 /* Socked used in thread, Do not add to main loop */ #define SKF_TRUNCATED 0x200 /* Received packet was truncated, set by IO layer */ diff --git a/lib/timer.c b/lib/timer.c index 381163d0..c47e0bbc 100644 --- a/lib/timer.c +++ b/lib/timer.c @@ -233,6 +233,7 @@ timers_fire(struct timeloop *loop) io_log_event(t->hook, t->data); t->hook(t); + tmp_flush(); } } diff --git a/misc/bird.spec b/misc/bird.spec index 26b43011..425a6340 100644 --- a/misc/bird.spec +++ b/misc/bird.spec @@ -1,6 +1,6 @@ Summary: BIRD Internet Routing Daemon Name: bird -Version: 2.0.8 +Version: 2.0.9 Release: 1 Copyright: GPL Group: Networking/Daemons diff --git a/misc/docker/ubuntu-20.10-amd64/Dockerfile b/misc/docker/ubuntu-21.10-amd64/Dockerfile index 19cb1b85..aa0987b6 100644 --- a/misc/docker/ubuntu-20.10-amd64/Dockerfile +++ b/misc/docker/ubuntu-21.10-amd64/Dockerfile @@ -1,9 +1,10 @@ -FROM ubuntu:20.10 +FROM ubuntu:21.10 ENV DEBIAN_FRONTEND=noninteractive RUN sed -i 's/deb.debian.org/ftp.cz.debian.org/' /etc/apt/sources.list RUN apt-get -y update RUN apt-get -y upgrade -RUN apt-get -y install \ +RUN apt-get -y --no-install-recommends install \ + tzdata \ build-essential \ flex \ bison \ diff --git a/nest/a-path_test.c b/nest/a-path_test.c index 9ed0a786..e007a450 100644 --- a/nest/a-path_test.c +++ b/nest/a-path_test.c @@ -23,8 +23,6 @@ static int t_as_path_match(void) { - resource_init(); - int round; for (round = 0; round < TESTS_NUM; round++) { @@ -32,14 +30,13 @@ t_as_path_match(void) struct adata *as_path = &empty_as_path; u32 first_prepended, last_prepended; first_prepended = last_prepended = 0; - struct linpool *lp = lp_new_default(&root_pool); struct f_path_mask *mask = alloca(sizeof(struct f_path_mask) + AS_PATH_LENGTH * sizeof(struct f_path_mask_item)); mask->len = AS_PATH_LENGTH; for (int i = AS_PATH_LENGTH - 1; i >= 0; i--) { u32 val = bt_random(); - as_path = as_path_prepend(lp, as_path, val); + as_path = as_path_prepend(tmp_linpool, as_path, val); bt_debug("Prepending ASN: %10u \n", val); if (i == 0) @@ -61,7 +58,7 @@ t_as_path_match(void) bt_assert(as_path_get_last(as_path, &asn)); bt_assert_msg(asn == first_prepended, "as_path_get_last() should return the first prepended ASN"); - rfree(lp); + tmp_flush(); } return 1; @@ -70,16 +67,13 @@ t_as_path_match(void) static int t_path_format(void) { - resource_init(); - struct adata empty_as_path = {}; struct adata *as_path = &empty_as_path; - struct linpool *lp = lp_new_default(&root_pool); uint i; for (i = 4294967285; i <= 4294967294; i++) { - as_path = as_path_prepend(lp, as_path, i); + as_path = as_path_prepend(tmp_linpool, as_path, i); bt_debug("Prepending ASN: %10u \n", i); } @@ -97,7 +91,7 @@ t_path_format(void) as_path_format(as_path, buf2, SMALL_BUFFER_SIZE); bt_assert_msg(strcmp(buf2, "4294967294 42...") == 0, "Small Buffer(%zu): '%s'", strlen(buf2), buf2); - rfree(lp); + tmp_flush(); return 1; } @@ -116,11 +110,8 @@ count_asn_in_array(const u32 *array, u32 asn) static int t_path_include(void) { - resource_init(); - struct adata empty_as_path = {}; struct adata *as_path = &empty_as_path; - struct linpool *lp = lp_new_default(&root_pool); u32 as_nums[AS_PATH_LENGTH] = {}; int i; @@ -128,7 +119,7 @@ t_path_include(void) { u32 val = bt_random(); as_nums[i] = val; - as_path = as_path_prepend(lp, as_path, val); + as_path = as_path_prepend(tmp_linpool, as_path, val); } for (i = 0; i < AS_PATH_LENGTH; i++) @@ -136,8 +127,8 @@ t_path_include(void) int counts_of_contains = count_asn_in_array(as_nums, as_nums[i]); bt_assert_msg(as_path_contains(as_path, as_nums[i], counts_of_contains), "AS Path should contains %d-times number %d", counts_of_contains, as_nums[i]); - bt_assert(as_path_filter(lp, as_path, NULL, as_nums[i], 0) != NULL); - bt_assert(as_path_filter(lp, as_path, NULL, as_nums[i], 1) != NULL); + bt_assert(as_path_filter(tmp_linpool, as_path, NULL, as_nums[i], 0) != NULL); + bt_assert(as_path_filter(tmp_linpool, as_path, NULL, as_nums[i], 1) != NULL); } for (i = 0; i < 10000; i++) @@ -152,7 +143,7 @@ t_path_include(void) bt_assert_msg(result == 0, "As path should not contain the number %u", test_val); } - rfree(lp); + tmp_flush(); return 1; } @@ -161,16 +152,13 @@ t_path_include(void) static int t_as_path_converting(void) { - resource_init(); - struct adata empty_as_path = {}; struct adata *as_path = &empty_as_path; - struct linpool *lp = lp_new_default(&root_pool); #define AS_PATH_LENGTH_FOR_CONVERTING_TEST 10 int i; for (i = 0; i < AS_PATH_LENGTH_FOR_CONVERTING_TEST; i++) - as_path = as_path_prepend(lp, as_path, i); + as_path = as_path_prepend(tmp_linpool, as_path, i); bt_debug("data length: %u \n", as_path->length); diff --git a/nest/a-set.c b/nest/a-set.c index 1186eb56..71fbac94 100644 --- a/nest/a-set.c +++ b/nest/a-set.c @@ -516,6 +516,48 @@ int_set_sort(struct linpool *pool, const struct adata *src) return dst; } +int +int_set_min(const struct adata *list, u32 *val) +{ + if (!list) + return 0; + + u32 *l = (u32 *) list->data; + int len = int_set_get_size(list); + int i; + + if (len < 1) + return 0; + + *val = *l++; + for (i = 1; i < len; i++, l++) + if (int_set_cmp(val, l) > 0) + *val = *l; + + return 1; +} + +int +int_set_max(const struct adata *list, u32 *val) +{ + if (!list) + return 0; + + u32 *l = (u32 *) list->data; + int len = int_set_get_size(list); + int i; + + if (len < 1) + return 0; + + *val = *l++; + for (i = 1; i < len; i++, l++) + if (int_set_cmp(val, l) < 0) + *val = *l; + + return 1; +} + static int ec_set_cmp(const void *X, const void *Y) @@ -541,6 +583,50 @@ ec_set_sort_x(struct adata *set) qsort(set->data, set->length / 8, 8, ec_set_cmp); } +int +ec_set_min(const struct adata *list, u64 *val) +{ + if (!list) + return 0; + + u32 *l = int_set_get_data(list); + int len = int_set_get_size(list); + int i; + + if (len < 1) + return 0; + + u32 *res = l; l += 2; + for (i = 2; i < len; i += 2, l += 2) + if (ec_set_cmp(res, l) > 0) + res = l; + + *val = ec_generic(res[0], res[1]); + return 1; +} + +int +ec_set_max(const struct adata *list, u64 *val) +{ + if (!list) + return 0; + + u32 *l = int_set_get_data(list); + int len = int_set_get_size(list); + int i; + + if (len < 1) + return 0; + + u32 *res = l; l += 2; + for (i = 2; i < len; i += 2, l += 2) + if (ec_set_cmp(res, l) < 0) + res = l; + + *val = ec_generic(res[0], res[1]); + return 1; +} + static int lc_set_cmp(const void *X, const void *Y) @@ -563,3 +649,47 @@ lc_set_sort(struct linpool *pool, const struct adata *src) qsort(dst->data, dst->length / LCOMM_LENGTH, LCOMM_LENGTH, lc_set_cmp); return dst; } + +int +lc_set_min(const struct adata *list, lcomm *val) +{ + if (!list) + return 0; + + u32 *l = int_set_get_data(list); + int len = int_set_get_size(list); + int i; + + if (len < 1) + return 0; + + u32 *res = l; l += 3; + for (i = 3; i < len; i += 3, l += 3) + if (lc_set_cmp(res, l) > 0) + res = l; + + *val = (lcomm) { res[0], res[1], res[2] }; + return 1; +} + +int +lc_set_max(const struct adata *list, lcomm *val) +{ + if (!list) + return 0; + + u32 *l = int_set_get_data(list); + int len = int_set_get_size(list); + int i; + + if (len < 1) + return 0; + + u32 *res = l; l += 3; + for (i = 3; i < len; i += 3, l += 3) + if (lc_set_cmp(res, l) < 0) + res = l; + + *val = (lcomm) { res[0], res[1], res[2] }; + return 1; +} diff --git a/nest/a-set_test.c b/nest/a-set_test.c index 96b6a727..904e6764 100644 --- a/nest/a-set_test.c +++ b/nest/a-set_test.c @@ -25,8 +25,6 @@ static byte buf[BUFFER_SIZE] = {}; #define SET_SIZE_FOR_FORMAT_OUTPUT 10 -struct linpool *lp; - enum set_type { SET_TYPE_INT, @@ -38,24 +36,23 @@ generate_set_sequence(enum set_type type, int len) { struct adata empty_as_path = {}; set_sequence = set_sequence_same = set_sequence_higher = set_random = &empty_as_path; - lp = lp_new_default(&root_pool); int i; for (i = 0; i < len; i++) { if (type == SET_TYPE_INT) { - set_sequence = int_set_add(lp, set_sequence, i); - set_sequence_same = int_set_add(lp, set_sequence_same, i); - set_sequence_higher = int_set_add(lp, set_sequence_higher, i + SET_SIZE); - set_random = int_set_add(lp, set_random, bt_random()); + set_sequence = int_set_add(tmp_linpool, set_sequence, i); + set_sequence_same = int_set_add(tmp_linpool, set_sequence_same, i); + set_sequence_higher = int_set_add(tmp_linpool, set_sequence_higher, i + SET_SIZE); + set_random = int_set_add(tmp_linpool, set_random, bt_random()); } else if (type == SET_TYPE_EC) { - set_sequence = ec_set_add(lp, set_sequence, i); - set_sequence_same = ec_set_add(lp, set_sequence_same, i); - set_sequence_higher = ec_set_add(lp, set_sequence_higher, i + SET_SIZE); - set_random = ec_set_add(lp, set_random, (bt_random() << 32 | bt_random())); + set_sequence = ec_set_add(tmp_linpool, set_sequence, i); + set_sequence_same = ec_set_add(tmp_linpool, set_sequence_same, i); + set_sequence_higher = ec_set_add(tmp_linpool, set_sequence_higher, i + SET_SIZE); + set_random = ec_set_add(tmp_linpool, set_random, (bt_random() << 32 | bt_random())); } else bt_abort_msg("This should be unreachable"); @@ -71,7 +68,6 @@ t_set_int_contains(void) { int i; - resource_init(); generate_set_sequence(SET_TYPE_INT, SET_SIZE); bt_assert(int_set_get_size(set_sequence) == SET_SIZE); @@ -85,33 +81,29 @@ t_set_int_contains(void) for (i = 0; i < SET_SIZE; i++) bt_assert_msg(data[i] == i, "(data[i] = %d) == i = %d)", data[i], i); - rfree(lp); return 1; } static int t_set_int_union(void) { - resource_init(); generate_set_sequence(SET_TYPE_INT, SET_SIZE); const struct adata *set_union; - set_union = int_set_union(lp, set_sequence, set_sequence_same); + set_union = int_set_union(tmp_linpool, set_sequence, set_sequence_same); bt_assert(int_set_get_size(set_union) == SET_SIZE); bt_assert(int_set_format(set_union, 0, 2, buf, BUFFER_SIZE) == 0); - set_union = int_set_union(lp, set_sequence, set_sequence_higher); + set_union = int_set_union(tmp_linpool, set_sequence, set_sequence_higher); bt_assert_msg(int_set_get_size(set_union) == SET_SIZE*2, "int_set_get_size(set_union) %d, SET_SIZE*2 %d", int_set_get_size(set_union), SET_SIZE*2); bt_assert(int_set_format(set_union, 0, 2, buf, BUFFER_SIZE) == 0); - rfree(lp); return 1; } static int t_set_int_format(void) { - resource_init(); generate_set_sequence(SET_TYPE_INT, SET_SIZE_FOR_FORMAT_OUTPUT); bt_assert(int_set_format(set_sequence, 0, 0, buf, BUFFER_SIZE) == 0); @@ -125,21 +117,19 @@ t_set_int_format(void) bt_assert(int_set_format(set_sequence, 1, 0, buf, BUFFER_SIZE) == 0); bt_assert(strcmp(buf, "(0,0) (0,1) (0,2) (0,3) (0,4) (0,5) (0,6) (0,7) (0,8) (0,9)") == 0); - rfree(lp); return 1; } static int t_set_int_delete(void) { - resource_init(); generate_set_sequence(SET_TYPE_INT, SET_SIZE); const struct adata *deleting_sequence = set_sequence; u32 i; for (i = 0; i < SET_SIZE; i++) { - deleting_sequence = int_set_del(lp, deleting_sequence, i); + deleting_sequence = int_set_del(tmp_linpool, deleting_sequence, i); bt_assert_msg(int_set_get_size(deleting_sequence) == (int) (SET_SIZE-1-i), "int_set_get_size(deleting_sequence) %d == SET_SIZE-1-i %d", int_set_get_size(deleting_sequence), @@ -160,7 +150,6 @@ t_set_ec_contains(void) { u32 i; - resource_init(); generate_set_sequence(SET_TYPE_EC, SET_SIZE); bt_assert(ec_set_get_size(set_sequence) == SET_SIZE); @@ -174,62 +163,54 @@ t_set_ec_contains(void) // for (i = 0; i < SET_SIZE; i++) // bt_assert_msg(data[i] == (SET_SIZE-1-i), "(data[i] = %d) == ((SET_SIZE-1-i) = %d)", data[i], SET_SIZE-1-i); - rfree(lp); return 1; } static int t_set_ec_union(void) { - resource_init(); generate_set_sequence(SET_TYPE_EC, SET_SIZE); const struct adata *set_union; - set_union = ec_set_union(lp, set_sequence, set_sequence_same); + set_union = ec_set_union(tmp_linpool, set_sequence, set_sequence_same); bt_assert(ec_set_get_size(set_union) == SET_SIZE); bt_assert(ec_set_format(set_union, 0, buf, BUFFER_SIZE) == 0); - set_union = ec_set_union(lp, set_sequence, set_sequence_higher); + set_union = ec_set_union(tmp_linpool, set_sequence, set_sequence_higher); bt_assert_msg(ec_set_get_size(set_union) == SET_SIZE*2, "ec_set_get_size(set_union) %d, SET_SIZE*2 %d", ec_set_get_size(set_union), SET_SIZE*2); bt_assert(ec_set_format(set_union, 0, buf, BUFFER_SIZE) == 0); - rfree(lp); return 1; } static int t_set_ec_format(void) { - resource_init(); - const struct adata empty_as_path = {}; set_sequence = set_sequence_same = set_sequence_higher = set_random = &empty_as_path; - lp = lp_new_default(&root_pool); u64 i = 0; - set_sequence = ec_set_add(lp, set_sequence, i); + set_sequence = ec_set_add(tmp_linpool, set_sequence, i); for (i = 1; i < SET_SIZE_FOR_FORMAT_OUTPUT; i++) - set_sequence = ec_set_add(lp, set_sequence, i + ((i%2) ? ((u64)EC_RO << 48) : ((u64)EC_RT << 48))); + set_sequence = ec_set_add(tmp_linpool, set_sequence, i + ((i%2) ? ((u64)EC_RO << 48) : ((u64)EC_RT << 48))); bt_assert(ec_set_format(set_sequence, 0, buf, BUFFER_SIZE) == 0); bt_assert_msg(strcmp(buf, "(unknown 0x0, 0, 0) (ro, 0, 1) (rt, 0, 2) (ro, 0, 3) (rt, 0, 4) (ro, 0, 5) (rt, 0, 6) (ro, 0, 7) (rt, 0, 8) (ro, 0, 9)") == 0, "ec_set_format() returns '%s'", buf); - rfree(lp); return 1; } static int t_set_ec_delete(void) { - resource_init(); generate_set_sequence(SET_TYPE_EC, SET_SIZE); const struct adata *deleting_sequence = set_sequence; u32 i; for (i = 0; i < SET_SIZE; i++) { - deleting_sequence = ec_set_del(lp, deleting_sequence, i); + deleting_sequence = ec_set_del(tmp_linpool, deleting_sequence, i); bt_assert_msg(ec_set_get_size(deleting_sequence) == (int) (SET_SIZE-1-i), "ec_set_get_size(deleting_sequence) %d == SET_SIZE-1-i %d", ec_set_get_size(deleting_sequence), SET_SIZE-1-i); diff --git a/nest/attrs.h b/nest/attrs.h index 50da817b..ef2b95e6 100644 --- a/nest/attrs.h +++ b/nest/attrs.h @@ -218,6 +218,12 @@ struct adata *ec_set_del_nontrans(struct linpool *pool, const struct adata *set) struct adata *int_set_sort(struct linpool *pool, const struct adata *src); struct adata *ec_set_sort(struct linpool *pool, const struct adata *src); struct adata *lc_set_sort(struct linpool *pool, const struct adata *src); +int int_set_min(const struct adata *list, u32 *val); +int ec_set_min(const struct adata *list, u64 *val); +int lc_set_min(const struct adata *list, lcomm *val); +int int_set_max(const struct adata *list, u32 *val); +int ec_set_max(const struct adata *list, u64 *val); +int lc_set_max(const struct adata *list, lcomm *val); void ec_set_sort_x(struct adata *set); /* Sort in place */ diff --git a/nest/cmds.c b/nest/cmds.c index 18f39eb5..8481bf96 100644 --- a/nest/cmds.c +++ b/nest/cmds.c @@ -67,31 +67,63 @@ cmd_show_symbols(struct sym_show_data *sd) } } -static void -print_size(char *dsc, size_t val) +#define SIZE_SUFFIX " kMGT" +#define SIZE_FORMAT "% 4u.%1u % 1cB" +#define SIZE_ARGS(a) (a).val, (a).decimal, SIZE_SUFFIX[(a).magnitude] + +struct size_args { + u64 val:48; + u64 decimal:8; + u64 magnitude:8; +}; + +static struct size_args +get_size_args(u64 val) { - char *px = " kMG"; - int i = 0; - while ((val >= 10000) && (i < 3)) +#define VALDEC 10 /* One decimal place */ + val *= VALDEC; + + uint i = 0; + while ((val >= 10000 * VALDEC) && (i < 4)) { val = (val + 512) / 1024; i++; } - cli_msg(-1018, "%-17s %4u %cB", dsc, (unsigned) val, px[i]); + return (struct size_args) { + .val = (val / VALDEC), + .decimal = (val % VALDEC), + .magnitude = i, + }; +} + +static void +print_size(char *dsc, struct resmem vals) +{ + struct size_args effective = get_size_args(vals.effective); + struct size_args overhead = get_size_args(vals.overhead); + + cli_msg(-1018, "%-17s " SIZE_FORMAT " " SIZE_FORMAT, dsc, SIZE_ARGS(effective), SIZE_ARGS(overhead)); } extern pool *rt_table_pool; extern pool *rta_pool; +extern uint *pages_kept; void cmd_show_memory(void) { cli_msg(-1018, "BIRD memory usage"); + cli_msg(-1018, "%-17s Effective Overhead", ""); print_size("Routing tables:", rmemsize(rt_table_pool)); print_size("Route attributes:", rmemsize(rta_pool)); print_size("Protocols:", rmemsize(proto_pool)); - print_size("Total:", rmemsize(&root_pool)); + struct resmem total = rmemsize(&root_pool); +#ifdef HAVE_MMAP + print_size("Standby memory:", (struct resmem) { .overhead = page_size * *pages_kept }); + total.overhead += page_size * *pages_kept; +#endif + print_size("Total:", total); cli_msg(0, ""); } diff --git a/nest/config.Y b/nest/config.Y index ef3e27c0..92a80589 100644 --- a/nest/config.Y +++ b/nest/config.Y @@ -17,6 +17,7 @@ CF_HDR CF_DEFINES +static struct rtable_config *this_table; static struct proto_config *this_proto; static struct channel_config *this_channel; static struct iface_patt *this_ipatt; @@ -117,13 +118,14 @@ CF_KEYWORDS(IPV4, IPV6, VPN4, VPN6, ROA4, ROA6, FLOW4, FLOW6, SADR, MPLS) CF_KEYWORDS(RECEIVE, LIMIT, ACTION, WARN, BLOCK, RESTART, DISABLE, KEEP, FILTERED, RPKI) CF_KEYWORDS(PASSWORD, KEY, FROM, PASSIVE, TO, ID, EVENTS, PACKETS, PROTOCOLS, CHANNELS, INTERFACES) CF_KEYWORDS(ALGORITHM, KEYED, HMAC, MD5, SHA1, SHA256, SHA384, SHA512, BLAKE2S128, BLAKE2S256, BLAKE2B256, BLAKE2B512) -CF_KEYWORDS(PRIMARY, STATS, COUNT, BY, FOR, COMMANDS, PREEXPORT, NOEXPORT, EXPORTED, GENERATE) -CF_KEYWORDS(BGP, PASSWORDS, DESCRIPTION, SORTED) +CF_KEYWORDS(PRIMARY, STATS, COUNT, BY, FOR, IN, COMMANDS, PREEXPORT, NOEXPORT, EXPORTED, GENERATE) +CF_KEYWORDS(BGP, PASSWORDS, DESCRIPTION) CF_KEYWORDS(RELOAD, IN, OUT, MRTDUMP, MESSAGES, RESTRICT, MEMORY, IGP_METRIC, CLASS, DSCP) CF_KEYWORDS(TIMEFORMAT, ISO, SHORT, LONG, ROUTE, PROTOCOL, BASE, LOG, S, MS, US) CF_KEYWORDS(GRACEFUL, RESTART, WAIT, MAX, FLUSH, AS) CF_KEYWORDS(MIN, IDLE, RX, TX, INTERVAL, MULTIPLIER, PASSIVE) CF_KEYWORDS(CHECK, LINK) +CF_KEYWORDS(SORTED, TRIE, MIN, MAX, SETTLE, TIME) /* For r_args_channel */ CF_KEYWORDS(IPV4, IPV4_MC, IPV4_MPLS, IPV6, IPV6_MC, IPV6_MPLS, IPV6_SADR, VPN4, VPN4_MC, VPN4_MPLS, VPN6, VPN6_MC, VPN6_MPLS, ROA4, ROA6, FLOW4, FLOW6, MPLS, PRI, SEC) @@ -141,7 +143,7 @@ CF_ENUM_PX(T_ENUM_AF, AF_, AFI_, IPV4, IPV6) %type <s> optproto %type <ra> r_args %type <sd> sym_args -%type <i> proto_start echo_mask echo_size debug_mask debug_list debug_flag mrtdump_mask mrtdump_list mrtdump_flag export_mode limit_action net_type table_sorted tos password_algorithm +%type <i> proto_start echo_mask echo_size debug_mask debug_list debug_flag mrtdump_mask mrtdump_list mrtdump_flag export_mode limit_action net_type tos password_algorithm %type <ps> proto_patt proto_patt2 %type <cc> channel_start proto_channel %type <cl> limit_spec @@ -206,16 +208,37 @@ CF_ENUM(T_ENUM_NETTYPE, NET_, IP4, IP6, VPN4, VPN6, ROA4, ROA6, FLOW4, FLOW6, IP conf: table ; +table: table_start table_sorted table_opt_list ; + +table_start: net_type TABLE symbol { + this_table = rt_new_table($3, $1); + } + ; + table_sorted: - { $$ = 0; } - | SORTED { $$ = 1; } + /* empty */ + | SORTED { this_table->sorted = 1; } ; -table: net_type TABLE symbol table_sorted { - struct rtable_config *cf; - cf = rt_new_table($3, $1); - cf->sorted = $4; +table_opt: + SORTED bool { this_table->sorted = $2; } + | TRIE bool { + if (!net_val_match(this_table->addr_type, NB_IP | NB_VPN | NB_ROA | NB_IP6_SADR)) + cf_error("Trie option not supported for %s table", net_label[this_table->addr_type]); + this_table->trie_used = $2; } + | MIN SETTLE TIME expr_us { this_table->min_settle_time = $4; } + | MAX SETTLE TIME expr_us { this_table->max_settle_time = $4; } + ; + +table_opts: + /* empty */ + | table_opts table_opt ';' + ; + +table_opt_list: + /* empty */ + | '{' table_opts '}' ; @@ -621,14 +644,22 @@ r_args: $$ = $1; if ($$->addr) cf_error("Only one prefix expected"); $$->addr = $2; + $$->addr_mode = RSD_ADDR_EQUAL; } | r_args FOR r_args_for { $$ = $1; if ($$->addr) cf_error("Only one prefix expected"); - $$->show_for = 1; $$->addr = $3; + $$->addr_mode = RSD_ADDR_FOR; + } + | r_args IN net_any { + $$ = $1; + if ($$->addr) cf_error("Only one prefix expected"); + if (!net_type_match($3, NB_IP)) cf_error("Only IP networks accepted for 'in' argument"); + $$->addr = $3; + $$->addr_mode = RSD_ADDR_IN; } - | r_args TABLE CF_SYM_KNOWN { +| r_args TABLE CF_SYM_KNOWN { cf_assert_symbol($3, SYM_TABLE); $$ = $1; rt_show_add_table($$, $3->table->table); diff --git a/nest/iface.c b/nest/iface.c index 83a633a3..682340c5 100644 --- a/nest/iface.c +++ b/nest/iface.c @@ -591,7 +591,7 @@ ifa_update(struct ifa *a) if (ipa_equal(b->brd, a->brd) && ipa_equal(b->opposite, a->opposite) && b->scope == a->scope && - !((b->flags ^ a->flags) & IA_PEER)) + !((b->flags ^ a->flags) & (IA_SECONDARY | IA_PEER | IA_HOST))) { b->flags |= IA_UPDATED; return b; diff --git a/nest/route.h b/nest/route.h index 4944a854..c7fb67aa 100644 --- a/nest/route.h +++ b/nest/route.h @@ -20,7 +20,10 @@ struct proto; struct rte_src; struct symbol; struct timer; +struct fib; struct filter; +struct f_trie; +struct f_trie_walk_state; struct cli; /* @@ -49,7 +52,7 @@ struct fib_iterator { /* See lib/slists.h for an explanation */ uint hash; }; -typedef void (*fib_init_fn)(void *); +typedef void (*fib_init_fn)(struct fib *, void *); struct fib { pool *fib_pool; /* Pool holding all our data */ @@ -149,6 +152,7 @@ struct rtable_config { int gc_min_time; /* Minimum time between two consecutive GC runs */ byte sorted; /* Routes of network are sorted according to rte_better() */ byte internal; /* Internal table of a protocol */ + byte trie_used; /* Rtable has attached trie */ btime min_settle_time; /* Minimum settle time for notifications */ btime max_settle_time; /* Maximum settle time for notifications */ }; @@ -159,6 +163,7 @@ typedef struct rtable { pool *rp; /* Resource pool to allocate everything from, including itself */ struct slab *rte_slab; /* Slab to allocate route objects */ struct fib fib; + struct f_trie *trie; /* Trie of prefixes defined in fib */ char *name; /* Name of this table */ list channels; /* List of attached channels (struct channel) */ uint addr_type; /* Type of address data stored in table (NET_*) */ @@ -181,13 +186,20 @@ typedef struct rtable { btime gc_time; /* Time of last GC */ int gc_counter; /* Number of operations since last GC */ byte prune_state; /* Table prune state, 1 -> scheduled, 2-> running */ + byte prune_trie; /* Prune prefix trie during next table prune */ byte hcu_scheduled; /* Hostcache update is scheduled */ byte nhu_state; /* Next Hop Update state */ struct fib_iterator prune_fit; /* Rtable prune FIB iterator */ struct fib_iterator nhu_fit; /* Next Hop Update FIB iterator */ + struct f_trie *trie_new; /* New prefix trie defined during pruning */ + struct f_trie *trie_old; /* Old prefix trie waiting to be freed */ + u32 trie_lock_count; /* Prefix trie locked by walks */ + u32 trie_old_lock_count; /* Old prefix trie locked by walks */ list subscribers; /* Subscribers for notifications */ struct timer *settle_timer; /* Settle time for notifications */ + list flowspec_links; /* List of flowspec links, src for NET_IPx and dst for NET_FLOWx */ + struct f_trie *flowspec_trie; /* Trie for evaluation of flowspec notifications */ } rtable; struct rt_subscription { @@ -197,6 +209,13 @@ struct rt_subscription { void *data; }; +struct rt_flowspec_link { + node n; + rtable *src; + rtable *dst; + u32 uc; +}; + #define NHU_CLEAN 0 #define NHU_SCHEDULED 1 #define NHU_RUNNING 2 @@ -323,8 +342,12 @@ void rt_preconfig(struct config *); void rt_commit(struct config *new, struct config *old); void rt_lock_table(rtable *); void rt_unlock_table(rtable *); +struct f_trie * rt_lock_trie(rtable *tab); +void rt_unlock_trie(rtable *tab, struct f_trie *trie); void rt_subscribe(rtable *tab, struct rt_subscription *s); void rt_unsubscribe(struct rt_subscription *s); +void rt_flowspec_link(rtable *src, rtable *dst); +void rt_flowspec_unlink(rtable *src, rtable *dst); rtable *rt_setup(pool *, struct rtable_config *); static inline void rt_shutdown(rtable *r) { rfree(r->rp); } @@ -332,7 +355,8 @@ static inline net *net_find(rtable *tab, const net_addr *addr) { return (net *) static inline net *net_find_valid(rtable *tab, const net_addr *addr) { net *n = net_find(tab, addr); return (n && n->routes && rte_is_valid(&n->routes->rte)) ? n : NULL; } static inline net *net_get(rtable *tab, const net_addr *addr) { return (net *) fib_get(&tab->fib, addr); } -void *net_route(rtable *tab, const net_addr *n); +net *net_get(rtable *tab, const net_addr *addr); +net *net_route(rtable *tab, const net_addr *n); int net_roa_check(rtable *tab, const net_addr *n, u32 asn); int rt_examine(rtable *t, net_addr *a, struct channel *c, const struct filter *filter); rte *rt_export_merged(struct channel *c, net *net, linpool *pool, int silent); @@ -353,6 +377,18 @@ void rt_prune_sync(rtable *t, int all); int rte_update_out(struct channel *c, const net_addr *n, rte *new, rte *old, struct rte_storage **old_exported, int refeed); struct rtable_config *rt_new_table(struct symbol *s, uint addr_type); +static inline int rt_is_ip(rtable *tab) +{ return (tab->addr_type == NET_IP4) || (tab->addr_type == NET_IP6); } + +static inline int rt_is_vpn(rtable *tab) +{ return (tab->addr_type == NET_VPN4) || (tab->addr_type == NET_VPN6); } + +static inline int rt_is_roa(rtable *tab) +{ return (tab->addr_type == NET_ROA4) || (tab->addr_type == NET_ROA6); } + +static inline int rt_is_flow(rtable *tab) +{ return (tab->addr_type == NET_FLOW4) || (tab->addr_type == NET_FLOW6); } + /* Default limit for ECMP next hops, defined in sysdep code */ extern const int rt_default_ecmp; @@ -369,6 +405,8 @@ struct rt_show_data { struct rt_show_data_rtable *tab; /* Iterator over table list */ struct rt_show_data_rtable *last_table; /* Last table in output */ struct fib_iterator fit; /* Iterator over networks in table */ + struct f_trie_walk_state *walk_state; /* Iterator over networks in trie */ + struct f_trie *walk_lock; /* Locked trie for walking */ int verbose, tables_defined_by; const struct filter *filter; struct proto *show_protocol; @@ -376,9 +414,10 @@ struct rt_show_data { struct channel *export_channel; struct config *running_on_config; struct krt_proto *kernel; - int export_mode, primary_only, filtered, stats, show_for; + int export_mode, addr_mode, primary_only, filtered, stats; int table_open; /* Iteration (fit) is open */ + int trie_walk; /* Current table is iterated using trie */ int net_counter, rt_counter, show_counter, table_counter; int net_counter_last, rt_counter_last, show_counter_last; }; @@ -395,6 +434,11 @@ struct rt_show_data_rtable * rt_show_add_table(struct rt_show_data *d, rtable *t #define RSD_TDB_SET 0x1 /* internal: show empty tables */ #define RSD_TDB_NMN 0x2 /* internal: need matching net */ +/* Value of addr_mode */ +#define RSD_ADDR_EQUAL 1 /* Exact query - show route <addr> */ +#define RSD_ADDR_FOR 2 /* Longest prefix match - show route for <addr> */ +#define RSD_ADDR_IN 3 /* Interval query - show route in <addr> */ + /* Value of export_mode in struct rt_show_data */ #define RSEM_NONE 0 /* Export mode not used */ #define RSEM_PREEXPORT 1 /* Routes ready for export, before filtering */ @@ -706,6 +750,9 @@ rta_set_recursive_next_hop(rtable *dep, rta *a, rtable *tab, ip_addr gw, ip_addr static inline void rt_lock_hostentry(struct hostentry *he) { if (he) he->uc++; } static inline void rt_unlock_hostentry(struct hostentry *he) { if (he) he->uc--; } +int rt_flowspec_check(rtable *tab_ip, rtable *tab_flow, const net_addr *n, rta *a, int interior); + + /* * Default protocol preferences */ diff --git a/nest/rt-fib.c b/nest/rt-fib.c index a7f70371..1690a8f6 100644 --- a/nest/rt-fib.c +++ b/nest/rt-fib.c @@ -331,7 +331,7 @@ fib_get(struct fib *f, const net_addr *a) memset(b, 0, f->node_offset); if (f->init) - f->init(b); + f->init(f, b); if (f->entries++ > f->entries_max) fib_rehash(f, HASH_HI_STEP); diff --git a/nest/rt-show.c b/nest/rt-show.c index ae5000f5..7b05c64e 100644 --- a/nest/rt-show.c +++ b/nest/rt-show.c @@ -15,6 +15,7 @@ #include "nest/cli.h" #include "nest/iface.h" #include "filter/filter.h" +#include "filter/data.h" #include "sysdep/unix/krt.h" static void @@ -109,10 +110,9 @@ rt_show_net(struct cli *c, net *n, struct rt_show_data *d) ASSUME(!d->export_mode || ec); int first = 1; + int first_show = 1; int pass = 0; - bsnprintf(ia, sizeof(ia), "%N", n->n.addr); - for (struct rte_storage *er = n->routes; er; er = er->next) { if (rte_is_filtered(&er->rte) != d->filtered) @@ -185,10 +185,17 @@ rt_show_net(struct cli *c, net *n, struct rt_show_data *d) goto skip; if (d->stats < 2) + { + if (first_show) + net_format(n->n.addr, ia, sizeof(ia)); + else + ia[0] = 0; + rt_show_rte(c, ia, &e, d, (n->routes == er)); + first_show = 0; + } d->show_counter++; - ia[0] = 0; skip: lp_flush(c->show_pool); @@ -205,9 +212,12 @@ rt_show_cleanup(struct cli *c) struct rt_show_data_rtable *tab; /* Unlink the iterator */ - if (d->table_open) + if (d->table_open && !d->trie_walk) fit_get(&d->tab->table->fib, &d->fit); + if (d->walk_lock) + rt_unlock_trie(d->tab->table, d->walk_lock); + /* Unlock referenced tables */ WALK_LIST(tab, d->tables) rt_unlock_table(tab->table); @@ -217,12 +227,13 @@ static void rt_show_cont(struct cli *c) { struct rt_show_data *d = c->rover; + struct rtable *tab = d->tab->table; #ifdef DEBUGGING unsigned max = 4; #else unsigned max = 64; #endif - struct fib *fib = &d->tab->table->fib; + struct fib *fib = &tab->fib; struct fib_iterator *it = &d->fit; if (d->running_on_config && (d->running_on_config != config)) @@ -233,7 +244,22 @@ rt_show_cont(struct cli *c) if (!d->table_open) { - FIB_ITERATE_INIT(&d->fit, &d->tab->table->fib); + /* We use either trie-based walk or fib-based walk */ + d->trie_walk = tab->trie && + (d->addr_mode == RSD_ADDR_IN) && + net_val_match(tab->addr_type, NB_IP); + + if (d->trie_walk && !d->walk_state) + d->walk_state = lp_allocz(c->parser_pool, sizeof (struct f_trie_walk_state)); + + if (d->trie_walk) + { + d->walk_lock = rt_lock_trie(tab); + trie_walk_init(d->walk_state, tab->trie, d->addr); + } + else + FIB_ITERATE_INIT(&d->fit, &tab->fib); + d->table_open = 1; d->table_counter++; d->kernel = rt_show_get_kernel(d); @@ -246,16 +272,44 @@ rt_show_cont(struct cli *c) rt_show_table(c, d); } - FIB_ITERATE_START(fib, it, net, n) + if (d->trie_walk) + { + /* Trie-based walk */ + net_addr addr; + while (trie_walk_next(d->walk_state, &addr)) + { + net *n = net_find(tab, &addr); + if (!n) + continue; + + rt_show_net(c, n, d); + + if (!--max) + return; + } + + rt_unlock_trie(tab, d->walk_lock); + d->walk_lock = NULL; + } + else { - if (!max--) + /* fib-based walk */ + FIB_ITERATE_START(fib, it, net, n) { - FIB_ITERATE_PUT(it); - return; + if ((d->addr_mode == RSD_ADDR_IN) && (!net_in_netX(n->n.addr, d->addr))) + goto next; + + if (!max--) + { + FIB_ITERATE_PUT(it); + return; + } + rt_show_net(c, n, d); + + next:; } - rt_show_net(c, n, d); + FIB_ITERATE_END; } - FIB_ITERATE_END; if (d->stats) { @@ -264,7 +318,7 @@ rt_show_cont(struct cli *c) cli_printf(c, -1007, "%d of %d routes for %d networks in table %s", d->show_counter - d->show_counter_last, d->rt_counter - d->rt_counter_last, - d->net_counter - d->net_counter_last, d->tab->table->name); + d->net_counter - d->net_counter_last, tab->name); } d->kernel = NULL; @@ -395,7 +449,7 @@ rt_show(struct rt_show_data *d) rt_show_prepare_tables(d); - if (!d->addr) + if (!d->addr || (d->addr_mode == RSD_ADDR_IN)) { WALK_LIST(tab, d->tables) rt_lock_table(tab->table); @@ -413,7 +467,7 @@ rt_show(struct rt_show_data *d) d->tab = tab; d->kernel = rt_show_get_kernel(d); - if (d->show_for) + if (d->addr_mode == RSD_ADDR_FOR) n = net_route(tab->table, d->addr); else n = net_find(tab->table, d->addr); diff --git a/nest/rt-table.c b/nest/rt-table.c index c49400b7..9a9d57ad 100644 --- a/nest/rt-table.c +++ b/nest/rt-table.c @@ -26,6 +26,66 @@ * (see the route attribute module for a precise explanation) holding the * remaining route attributes which are expected to be shared by multiple * routes in order to conserve memory. + * + * There are several mechanisms that allow automatic update of routes in one + * routing table (dst) as a result of changes in another routing table (src). + * They handle issues of recursive next hop resolving, flowspec validation and + * RPKI validation. + * + * The first such mechanism is handling of recursive next hops. A route in the + * dst table has an indirect next hop address, which is resolved through a route + * in the src table (which may also be the same table) to get an immediate next + * hop. This is implemented using structure &hostcache attached to the src + * table, which contains &hostentry structures for each tracked next hop + * address. These structures are linked from recursive routes in dst tables, + * possibly multiple routes sharing one hostentry (as many routes may have the + * same indirect next hop). There is also a trie in the hostcache, which matches + * all prefixes that may influence resolving of tracked next hops. + * + * When a best route changes in the src table, the hostcache is notified using + * rt_notify_hostcache(), which immediately checks using the trie whether the + * change is relevant and if it is, then it schedules asynchronous hostcache + * recomputation. The recomputation is done by rt_update_hostcache() (called + * from rt_event() of src table), it walks through all hostentries and resolves + * them (by rt_update_hostentry()). It also updates the trie. If a change in + * hostentry resolution was found, then it schedules asynchronous nexthop + * recomputation of associated dst table. That is done by rt_next_hop_update() + * (called from rt_event() of dst table), it iterates over all routes in the dst + * table and re-examines their hostentries for changes. Note that in contrast to + * hostcache update, next hop update can be interrupted by main loop. These two + * full-table walks (over hostcache and dst table) are necessary due to absence + * of direct lookups (route -> affected nexthop, nexthop -> its route). + * + * The second mechanism is for flowspec validation, where validity of flowspec + * routes depends of resolving their network prefixes in IP routing tables. This + * is similar to the recursive next hop mechanism, but simpler as there are no + * intermediate hostcache and hostentries (because flows are less likely to + * share common net prefix than routes sharing a common next hop). In src table, + * there is a list of dst tables (list flowspec_links), this list is updated by + * flowpsec channels (by rt_flowspec_link() and rt_flowspec_unlink() during + * channel start/stop). Each dst table has its own trie of prefixes that may + * influence validation of flowspec routes in it (flowspec_trie). + * + * When a best route changes in the src table, rt_flowspec_notify() immediately + * checks all dst tables from the list using their tries to see whether the + * change is relevant for them. If it is, then an asynchronous re-validation of + * flowspec routes in the dst table is scheduled. That is also done by function + * rt_next_hop_update(), like nexthop recomputation above. It iterates over all + * flowspec routes and re-validates them. It also recalculates the trie. + * + * Note that in contrast to the hostcache update, here the trie is recalculated + * during the rt_next_hop_update(), which may be interleaved with IP route + * updates. The trie is flushed at the beginning of recalculation, which means + * that such updates may use partial trie to see if they are relevant. But it + * works anyway! Either affected flowspec was already re-validated and added to + * the trie, then IP route change would match the trie and trigger a next round + * of re-validation, or it was not yet re-validated and added to the trie, but + * will be re-validated later in this round anyway. + * + * The third mechanism is used for RPKI re-validation of IP routes and it is the + * simplest. It is just a list of subscribers in src table, who are notified + * when any change happened, but only after a settle time. Also, in RPKI case + * the dst is not a table, but a channel, who refeeds routes through a filter. */ #undef LOCAL_DEBUG @@ -44,6 +104,11 @@ #include "lib/hash.h" #include "lib/string.h" #include "lib/alloca.h" +#include "lib/flowspec.h" + +#ifdef CONFIG_BGP +#include "proto/bgp/bgp.h" +#endif pool *rt_table_pool; @@ -57,41 +122,184 @@ static void rt_update_hostcache(rtable *tab); static void rt_next_hop_update(rtable *tab); static inline void rt_prune_table(rtable *tab); static inline void rt_schedule_notify(rtable *tab); +static void rt_flowspec_notify(rtable *tab, net *net); + + +static void +net_init_with_trie(struct fib *f, void *N) +{ + rtable *tab = SKIP_BACK(rtable, fib, f); + net *n = N; + if (tab->trie) + trie_add_prefix(tab->trie, n->n.addr, n->n.addr->pxlen, n->n.addr->pxlen); + + if (tab->trie_new) + trie_add_prefix(tab->trie_new, n->n.addr, n->n.addr->pxlen, n->n.addr->pxlen); +} + +static inline net * +net_route_ip4_trie(rtable *t, const net_addr_ip4 *n0) +{ + TRIE_WALK_TO_ROOT_IP4(t->trie, n0, n) + { + net *r; + if (r = net_find_valid(t, (net_addr *) &n)) + return r; + } + TRIE_WALK_TO_ROOT_END; + + return NULL; +} + +static inline net * +net_route_vpn4_trie(rtable *t, const net_addr_vpn4 *n0) +{ + TRIE_WALK_TO_ROOT_IP4(t->trie, (const net_addr_ip4 *) n0, px) + { + net_addr_vpn4 n = NET_ADDR_VPN4(px.prefix, px.pxlen, n0->rd); + + net *r; + if (r = net_find_valid(t, (net_addr *) &n)) + return r; + } + TRIE_WALK_TO_ROOT_END; + + return NULL; +} + +static inline net * +net_route_ip6_trie(rtable *t, const net_addr_ip6 *n0) +{ + TRIE_WALK_TO_ROOT_IP6(t->trie, n0, n) + { + net *r; + if (r = net_find_valid(t, (net_addr *) &n)) + return r; + } + TRIE_WALK_TO_ROOT_END; + + return NULL; +} + +static inline net * +net_route_vpn6_trie(rtable *t, const net_addr_vpn6 *n0) +{ + TRIE_WALK_TO_ROOT_IP6(t->trie, (const net_addr_ip6 *) n0, px) + { + net_addr_vpn6 n = NET_ADDR_VPN6(px.prefix, px.pxlen, n0->rd); + + net *r; + if (r = net_find_valid(t, (net_addr *) &n)) + return r; + } + TRIE_WALK_TO_ROOT_END; + + return NULL; +} -/* Like fib_route(), but skips empty net entries */ static inline void * -net_route_ip4(rtable *t, net_addr_ip4 *n) +net_route_ip6_sadr_trie(rtable *t, const net_addr_ip6_sadr *n0) +{ + TRIE_WALK_TO_ROOT_IP6(t->trie, (const net_addr_ip6 *) n0, px) + { + net_addr_ip6_sadr n = NET_ADDR_IP6_SADR(px.prefix, px.pxlen, n0->src_prefix, n0->src_pxlen); + net *best = NULL; + int best_pxlen = 0; + + /* We need to do dst first matching. Since sadr addresses are hashed on dst + prefix only, find the hash table chain and go through it to find the + match with the longest matching src prefix. */ + for (struct fib_node *fn = fib_get_chain(&t->fib, (net_addr *) &n); fn; fn = fn->next) + { + net_addr_ip6_sadr *a = (void *) fn->addr; + + if (net_equal_dst_ip6_sadr(&n, a) && + net_in_net_src_ip6_sadr(&n, a) && + (a->src_pxlen >= best_pxlen)) + { + best = fib_node_to_user(&t->fib, fn); + best_pxlen = a->src_pxlen; + } + } + + if (best) + return best; + } + TRIE_WALK_TO_ROOT_END; + + return NULL; +} + +static inline net * +net_route_ip4_fib(rtable *t, const net_addr_ip4 *n0) { + net_addr_ip4 n; + net_copy_ip4(&n, n0); + net *r; + while (r = net_find_valid(t, (net_addr *) &n), (!r) && (n.pxlen > 0)) + { + n.pxlen--; + ip4_clrbit(&n.prefix, n.pxlen); + } - while (r = net_find_valid(t, (net_addr *) n), (!r) && (n->pxlen > 0)) + return r; +} + +static inline net * +net_route_vpn4_fib(rtable *t, const net_addr_vpn4 *n0) +{ + net_addr_vpn4 n; + net_copy_vpn4(&n, n0); + + net *r; + while (r = net_find_valid(t, (net_addr *) &n), (!r) && (n.pxlen > 0)) { - n->pxlen--; - ip4_clrbit(&n->prefix, n->pxlen); + n.pxlen--; + ip4_clrbit(&n.prefix, n.pxlen); } return r; } -static inline void * -net_route_ip6(rtable *t, net_addr_ip6 *n) +static inline net * +net_route_ip6_fib(rtable *t, const net_addr_ip6 *n0) { + net_addr_ip6 n; + net_copy_ip6(&n, n0); + net *r; + while (r = net_find_valid(t, (net_addr *) &n), (!r) && (n.pxlen > 0)) + { + n.pxlen--; + ip6_clrbit(&n.prefix, n.pxlen); + } + + return r; +} + +static inline net * +net_route_vpn6_fib(rtable *t, const net_addr_vpn6 *n0) +{ + net_addr_vpn6 n; + net_copy_vpn6(&n, n0); - while (r = net_find_valid(t, (net_addr *) n), (!r) && (n->pxlen > 0)) + net *r; + while (r = net_find_valid(t, (net_addr *) &n), (!r) && (n.pxlen > 0)) { - n->pxlen--; - ip6_clrbit(&n->prefix, n->pxlen); + n.pxlen--; + ip6_clrbit(&n.prefix, n.pxlen); } return r; } static inline void * -net_route_ip6_sadr(rtable *t, net_addr_ip6_sadr *n) +net_route_ip6_sadr_fib(rtable *t, const net_addr_ip6_sadr *n0) { - struct fib_node *fn; + net_addr_ip6_sadr n; + net_copy_ip6_sadr(&n, n0); while (1) { @@ -100,13 +308,13 @@ net_route_ip6_sadr(rtable *t, net_addr_ip6_sadr *n) /* We need to do dst first matching. Since sadr addresses are hashed on dst prefix only, find the hash table chain and go through it to find the - match with the smallest matching src prefix. */ - for (fn = fib_get_chain(&t->fib, (net_addr *) n); fn; fn = fn->next) + match with the longest matching src prefix. */ + for (struct fib_node *fn = fib_get_chain(&t->fib, (net_addr *) &n); fn; fn = fn->next) { net_addr_ip6_sadr *a = (void *) fn->addr; - if (net_equal_dst_ip6_sadr(n, a) && - net_in_net_src_ip6_sadr(n, a) && + if (net_equal_dst_ip6_sadr(&n, a) && + net_in_net_src_ip6_sadr(&n, a) && (a->src_pxlen >= best_pxlen)) { best = fib_node_to_user(&t->fib, fn); @@ -117,38 +325,52 @@ net_route_ip6_sadr(rtable *t, net_addr_ip6_sadr *n) if (best) return best; - if (!n->dst_pxlen) + if (!n.dst_pxlen) break; - n->dst_pxlen--; - ip6_clrbit(&n->dst_prefix, n->dst_pxlen); + n.dst_pxlen--; + ip6_clrbit(&n.dst_prefix, n.dst_pxlen); } return NULL; } -void * +net * net_route(rtable *tab, const net_addr *n) { ASSERT(tab->addr_type == n->type); - net_addr *n0 = alloca(n->length); - net_copy(n0, n); - switch (n->type) { case NET_IP4: + if (tab->trie) + return net_route_ip4_trie(tab, (net_addr_ip4 *) n); + else + return net_route_ip4_fib (tab, (net_addr_ip4 *) n); + case NET_VPN4: - case NET_ROA4: - return net_route_ip4(tab, (net_addr_ip4 *) n0); + if (tab->trie) + return net_route_vpn4_trie(tab, (net_addr_vpn4 *) n); + else + return net_route_vpn4_fib (tab, (net_addr_vpn4 *) n); case NET_IP6: + if (tab->trie) + return net_route_ip6_trie(tab, (net_addr_ip6 *) n); + else + return net_route_ip6_fib (tab, (net_addr_ip6 *) n); + case NET_VPN6: - case NET_ROA6: - return net_route_ip6(tab, (net_addr_ip6 *) n0); + if (tab->trie) + return net_route_vpn6_trie(tab, (net_addr_vpn6 *) n); + else + return net_route_vpn6_fib (tab, (net_addr_vpn6 *) n); case NET_IP6_SADR: - return net_route_ip6_sadr(tab, (net_addr_ip6_sadr *) n0); + if (tab->trie) + return net_route_ip6_sadr_trie(tab, (net_addr_ip6_sadr *) n); + else + return net_route_ip6_sadr_fib (tab, (net_addr_ip6_sadr *) n); default: return NULL; @@ -157,7 +379,35 @@ net_route(rtable *tab, const net_addr *n) static int -net_roa_check_ip4(rtable *tab, const net_addr_ip4 *px, u32 asn) +net_roa_check_ip4_trie(rtable *tab, const net_addr_ip4 *px, u32 asn) +{ + int anything = 0; + + TRIE_WALK_TO_ROOT_IP4(tab->trie, px, px0) + { + net_addr_roa4 roa0 = NET_ADDR_ROA4(px0.prefix, px0.pxlen, 0, 0); + + struct fib_node *fn; + for (fn = fib_get_chain(&tab->fib, (net_addr *) &roa0); fn; fn = fn->next) + { + net_addr_roa4 *roa = (void *) fn->addr; + net *r = fib_node_to_user(&tab->fib, fn); + + if (net_equal_prefix_roa4(roa, &roa0) && rte_is_valid(r->routes)) + { + anything = 1; + if (asn && (roa->asn == asn) && (roa->max_pxlen >= px->pxlen)) + return ROA_VALID; + } + } + } + TRIE_WALK_TO_ROOT_END; + + return anything ? ROA_INVALID : ROA_UNKNOWN; +} + +static int +net_roa_check_ip4_fib(rtable *tab, const net_addr_ip4 *px, u32 asn) { struct net_addr_roa4 n = NET_ADDR_ROA4(px->prefix, px->pxlen, 0, 0); struct fib_node *fn; @@ -189,7 +439,35 @@ net_roa_check_ip4(rtable *tab, const net_addr_ip4 *px, u32 asn) } static int -net_roa_check_ip6(rtable *tab, const net_addr_ip6 *px, u32 asn) +net_roa_check_ip6_trie(rtable *tab, const net_addr_ip6 *px, u32 asn) +{ + int anything = 0; + + TRIE_WALK_TO_ROOT_IP6(tab->trie, px, px0) + { + net_addr_roa6 roa0 = NET_ADDR_ROA6(px0.prefix, px0.pxlen, 0, 0); + + struct fib_node *fn; + for (fn = fib_get_chain(&tab->fib, (net_addr *) &roa0); fn; fn = fn->next) + { + net_addr_roa6 *roa = (void *) fn->addr; + net *r = fib_node_to_user(&tab->fib, fn); + + if (net_equal_prefix_roa6(roa, &roa0) && rte_is_valid(r->routes)) + { + anything = 1; + if (asn && (roa->asn == asn) && (roa->max_pxlen >= px->pxlen)) + return ROA_VALID; + } + } + } + TRIE_WALK_TO_ROOT_END; + + return anything ? ROA_INVALID : ROA_UNKNOWN; +} + +static int +net_roa_check_ip6_fib(rtable *tab, const net_addr_ip6 *px, u32 asn) { struct net_addr_roa6 n = NET_ADDR_ROA6(px->prefix, px->pxlen, 0, 0); struct fib_node *fn; @@ -239,9 +517,19 @@ int net_roa_check(rtable *tab, const net_addr *n, u32 asn) { if ((tab->addr_type == NET_ROA4) && (n->type == NET_IP4)) - return net_roa_check_ip4(tab, (const net_addr_ip4 *) n, asn); + { + if (tab->trie) + return net_roa_check_ip4_trie(tab, (const net_addr_ip4 *) n, asn); + else + return net_roa_check_ip4_fib (tab, (const net_addr_ip4 *) n, asn); + } else if ((tab->addr_type == NET_ROA6) && (n->type == NET_IP6)) - return net_roa_check_ip6(tab, (const net_addr_ip6 *) n, asn); + { + if (tab->trie) + return net_roa_check_ip6_trie(tab, (const net_addr_ip6 *) n, asn); + else + return net_roa_check_ip6_fib (tab, (const net_addr_ip6 *) n, asn); + } else return ROA_UNKNOWN; /* Should not happen */ } @@ -738,6 +1026,9 @@ rte_announce(rtable *tab, uint type, net *net, struct rte_storage *new, struct r if (tab->hostcache) rt_notify_hostcache(tab, net); + + if (!EMPTY_LIST(tab->flowspec_links)) + rt_flowspec_notify(tab, net); } rt_schedule_notify(tab); @@ -800,6 +1091,10 @@ rte_validate(rte *e) if (net_type_match(n, NB_DEST) == !e->attrs->dest) { + /* Exception for flowspec that failed validation */ + if (net_is_flow(n) && (e->attrs->dest == RTD_UNREACHABLE)) + return 1; + log(L_WARN "Ignoring route %N with invalid dest %d received via %s", n, e->attrs->dest, e->sender->proto->name); return 0; @@ -1515,6 +1810,90 @@ rt_unsubscribe(struct rt_subscription *s) rt_unlock_table(s->tab); } +static struct rt_flowspec_link * +rt_flowspec_find_link(rtable *src, rtable *dst) +{ + struct rt_flowspec_link *ln; + WALK_LIST(ln, src->flowspec_links) + if ((ln->src == src) && (ln->dst == dst)) + return ln; + + return NULL; +} + +void +rt_flowspec_link(rtable *src, rtable *dst) +{ + ASSERT(rt_is_ip(src)); + ASSERT(rt_is_flow(dst)); + + struct rt_flowspec_link *ln = rt_flowspec_find_link(src, dst); + + if (!ln) + { + rt_lock_table(src); + rt_lock_table(dst); + + ln = mb_allocz(src->rp, sizeof(struct rt_flowspec_link)); + ln->src = src; + ln->dst = dst; + add_tail(&src->flowspec_links, &ln->n); + } + + ln->uc++; +} + +void +rt_flowspec_unlink(rtable *src, rtable *dst) +{ + struct rt_flowspec_link *ln = rt_flowspec_find_link(src, dst); + + ASSERT(ln && (ln->uc > 0)); + + ln->uc--; + + if (!ln->uc) + { + rem_node(&ln->n); + mb_free(ln); + + rt_unlock_table(src); + rt_unlock_table(dst); + } +} + +static void +rt_flowspec_notify(rtable *src, net *net) +{ + /* Only IP tables are src links */ + ASSERT(rt_is_ip(src)); + + struct rt_flowspec_link *ln; + WALK_LIST(ln, src->flowspec_links) + { + rtable *dst = ln->dst; + ASSERT(rt_is_flow(dst)); + + /* No need to inspect it further if recalculation is already active */ + if ((dst->nhu_state == NHU_SCHEDULED) || (dst->nhu_state == NHU_DIRTY)) + continue; + + if (trie_match_net(dst->flowspec_trie, net->n.addr)) + rt_schedule_nhu(dst); + } +} + +static void +rt_flowspec_reset_trie(rtable *tab) +{ + linpool *lp = tab->flowspec_trie->lp; + int ipv4 = tab->flowspec_trie->ipv4; + + lp_flush(lp); + tab->flowspec_trie = f_new_trie(lp, 0); + tab->flowspec_trie->ipv4 = ipv4; +} + static void rt_free(resource *_r) { @@ -1579,16 +1958,31 @@ rt_setup(pool *pp, struct rtable_config *cf) fib_init(&t->fib, p, t->addr_type, sizeof(net), OFFSETOF(net, n), 0, NULL); + if (cf->trie_used) + { + t->trie = f_new_trie(lp_new_default(p), 0); + t->trie->ipv4 = net_val_match(t->addr_type, NB_IP4 | NB_VPN4 | NB_ROA4); + + t->fib.init = net_init_with_trie; + } + + init_list(&t->channels); + init_list(&t->flowspec_links); + init_list(&t->subscribers); + if (!(t->internal = cf->internal)) { - init_list(&t->channels); hmap_init(&t->id_map, p, 1024); hmap_set(&t->id_map, 0); - init_list(&t->subscribers); - t->rt_event = ev_new_init(p, rt_event, t); t->last_rt_change = t->gc_time = current_time(); + + if (rt_is_flow(t)) + { + t->flowspec_trie = f_new_trie(lp_new_default(p), 0); + t->flowspec_trie->ipv4 = (t->addr_type == NET_FLOW4); + } } return t; @@ -1650,23 +2044,30 @@ rt_prune_table(rtable *tab) FIB_ITERATE_INIT(fit, &tab->fib); tab->prune_state = 2; + + if (tab->prune_trie) + { + /* Init prefix trie pruning */ + tab->trie_new = f_new_trie(lp_new_default(tab->rp), 0); + tab->trie_new->ipv4 = tab->trie->ipv4; + } } again: FIB_ITERATE_START(&tab->fib, fit, net, n) { rescan: + if (limit <= 0) + { + FIB_ITERATE_PUT(fit); + ev_schedule(tab->rt_event); + return; + } + for (struct rte_storage *e=n->routes; e; e=e->next) { if (e->rte.sender->flush_active || (e->rte.flags & REF_DISCARD)) { - if (limit <= 0) - { - FIB_ITERATE_PUT(fit); - ev_schedule(tab->rt_event); - return; - } - rte_discard(n, &e->rte); limit--; @@ -1675,13 +2076,6 @@ again: if (e->rte.flags & REF_MODIFY) { - if (limit <= 0) - { - FIB_ITERATE_PUT(fit); - ev_schedule(tab->rt_event); - return; - } - rte_modify(n, &e->rte); limit--; @@ -1695,6 +2089,12 @@ again: fib_delete(&tab->fib, n); goto again; } + + if (tab->trie_new) + { + trie_add_prefix(tab->trie_new, n->n.addr, n->n.addr->pxlen, n->n.addr->pxlen); + limit--; + } } FIB_ITERATE_END; @@ -1708,6 +2108,37 @@ again: /* state change 2->0, 3->1 */ tab->prune_state &= 1; + if (tab->trie_new) + { + /* Finish prefix trie pruning */ + + if (!tab->trie_lock_count) + { + rfree(tab->trie->lp); + } + else + { + ASSERT(!tab->trie_old); + tab->trie_old = tab->trie; + tab->trie_old_lock_count = tab->trie_lock_count; + tab->trie_lock_count = 0; + } + + tab->trie = tab->trie_new; + tab->trie_new = NULL; + tab->prune_trie = 0; + } + else + { + /* Schedule prefix trie pruning */ + if (tab->trie && !tab->trie_old && (tab->trie->prefix_count > (2 * tab->fib.entries))) + { + /* state change 0->1, 2->3 */ + tab->prune_state |= 1; + tab->prune_trie = 1; + } + } + if (tab->prune_state > 0) ev_schedule(tab->rt_event); @@ -1725,6 +2156,72 @@ again: return; } +/** + * rt_lock_trie - lock a prefix trie of a routing table + * @tab: routing table with prefix trie to be locked + * + * The prune loop may rebuild the prefix trie and invalidate f_trie_walk_state + * structures. Therefore, asynchronous walks should lock the prefix trie using + * this function. That allows the prune loop to rebuild the trie, but postpones + * its freeing until all walks are done (unlocked by rt_unlock_trie()). + * + * Return a current trie that will be locked, the value should be passed back to + * rt_unlock_trie() for unlocking. + * + */ +struct f_trie * +rt_lock_trie(rtable *tab) +{ + ASSERT(tab->trie); + + tab->trie_lock_count++; + return tab->trie; +} + +/** + * rt_unlock_trie - unlock a prefix trie of a routing table + * @tab: routing table with prefix trie to be locked + * @trie: value returned by matching rt_lock_trie() + * + * Done for trie locked by rt_lock_trie() after walk over the trie is done. + * It may free the trie and schedule next trie pruning. + */ +void +rt_unlock_trie(rtable *tab, struct f_trie *trie) +{ + ASSERT(trie); + + if (trie == tab->trie) + { + /* Unlock the current prefix trie */ + ASSERT(tab->trie_lock_count); + tab->trie_lock_count--; + } + else if (trie == tab->trie_old) + { + /* Unlock the old prefix trie */ + ASSERT(tab->trie_old_lock_count); + tab->trie_old_lock_count--; + + /* Free old prefix trie that is no longer needed */ + if (!tab->trie_old_lock_count) + { + rfree(tab->trie_old->lp); + tab->trie_old = NULL; + + /* Kick prefix trie pruning that was postponed */ + if (tab->trie && (tab->trie->prefix_count > (2 * tab->fib.entries))) + { + tab->prune_trie = 1; + rt_schedule_prune(tab); + } + } + } + else + log(L_BUG "Invalid arg to rt_unlock_trie()"); +} + + void rt_preconfig(struct config *c) { @@ -1740,21 +2237,6 @@ rt_preconfig(struct config *c) * triggered by rt_schedule_nhu(). */ -static inline int -rta_next_hop_outdated(rta *a) -{ - struct hostentry *he = a->hostentry; - - if (!he) - return 0; - - if (!he->src) - return a->dest != RTD_UNREACHABLE; - - return (a->dest != he->dest) || (a->igp_metric != he->igp_metric) || - (!he->nexthop_linkable) || !nexthop_same(&(a->nh), &(he->src->nh)); -} - void rta_apply_hostentry(rta *a, struct hostentry *he, mpls_label_stack *mls) { @@ -1846,9 +2328,27 @@ no_nexthop: } } +static inline int +rta_next_hop_outdated(rta *a) +{ + struct hostentry *he = a->hostentry; + + if (!he) + return 0; + + if (!he->src) + return a->dest != RTD_UNREACHABLE; + + return (a->dest != he->dest) || (a->igp_metric != he->igp_metric) || + (!he->nexthop_linkable) || !nexthop_same(&(a->nh), &(he->src->nh)); +} + static inline struct rte_storage * rt_next_hop_update_rte(rtable *tab, net *n, rte *old) { + if (!rta_next_hop_outdated(old->attrs)) + return NULL; + rta *a = alloca(RTA_MAX_SIZE); memcpy(a, old->attrs, rta_size(old->attrs)); @@ -1864,6 +2364,155 @@ rt_next_hop_update_rte(rtable *tab, net *n, rte *old) return rte_store(&e0, n, tab); } + +#ifdef CONFIG_BGP + +static inline int +net_flow_has_dst_prefix(const net_addr *n) +{ + ASSUME(net_is_flow(n)); + + if (n->pxlen) + return 1; + + if (n->type == NET_FLOW4) + { + const net_addr_flow4 *n4 = (void *) n; + return (n4->length > sizeof(net_addr_flow4)) && (n4->data[0] == FLOW_TYPE_DST_PREFIX); + } + else + { + const net_addr_flow6 *n6 = (void *) n; + return (n6->length > sizeof(net_addr_flow6)) && (n6->data[0] == FLOW_TYPE_DST_PREFIX); + } +} + +static inline int +rta_as_path_is_empty(rta *a) +{ + eattr *e = ea_find(a->eattrs, EA_CODE(PROTOCOL_BGP, BA_AS_PATH)); + return !e || (as_path_getlen(e->u.ptr) == 0); +} + +static inline u32 +rta_get_first_asn(rta *a) +{ + eattr *e = ea_find(a->eattrs, EA_CODE(PROTOCOL_BGP, BA_AS_PATH)); + u32 asn; + + return (e && as_path_get_first_regular(e->u.ptr, &asn)) ? asn : 0; +} + +int +rt_flowspec_check(rtable *tab_ip, rtable *tab_flow, const net_addr *n, rta *a, int interior) +{ + ASSERT(rt_is_ip(tab_ip)); + ASSERT(rt_is_flow(tab_flow)); + ASSERT(tab_ip->trie); + + /* RFC 8955 6. a) Flowspec has defined dst prefix */ + if (!net_flow_has_dst_prefix(n)) + return 0; + + /* RFC 9117 4.1. Accept AS_PATH is empty (fr */ + if (interior && rta_as_path_is_empty(a)) + return 1; + + + /* RFC 8955 6. b) Flowspec and its best-match route have the same originator */ + + /* Find flowspec dst prefix */ + net_addr dst; + if (n->type == NET_FLOW4) + net_fill_ip4(&dst, net4_prefix(n), net4_pxlen(n)); + else + net_fill_ip6(&dst, net6_prefix(n), net6_pxlen(n)); + + /* Find best-match BGP unicast route for flowspec dst prefix */ + net *nb = net_route(tab_ip, &dst); + const rte *rb = nb ? &nb->routes->rte : NULL; + + /* Register prefix to trie for tracking further changes */ + int max_pxlen = (n->type == NET_FLOW4) ? IP4_MAX_PREFIX_LENGTH : IP6_MAX_PREFIX_LENGTH; + trie_add_prefix(tab_flow->flowspec_trie, &dst, (nb ? nb->n.addr->pxlen : 0), max_pxlen); + + /* No best-match BGP route -> no flowspec */ + if (!rb || (rb->attrs->source != RTS_BGP)) + return 0; + + /* Find ORIGINATOR_ID values */ + u32 orig_a = ea_get_int(a->eattrs, EA_CODE(PROTOCOL_BGP, BA_ORIGINATOR_ID), 0); + u32 orig_b = ea_get_int(rb->attrs->eattrs, EA_CODE(PROTOCOL_BGP, BA_ORIGINATOR_ID), 0); + + /* Originator is either ORIGINATOR_ID (if present), or BGP neighbor address (if not) */ + if ((orig_a != orig_b) || (!orig_a && !orig_b && !ipa_equal(a->from, rb->attrs->from))) + return 0; + + + /* Find ASN of the best-match route, for use in next checks */ + u32 asn_b = rta_get_first_asn(rb->attrs); + if (!asn_b) + return 0; + + /* RFC 9117 4.2. For EBGP, flowspec and its best-match route are from the same AS */ + if (!interior && (rta_get_first_asn(a) != asn_b)) + return 0; + + /* RFC 8955 6. c) More-specific routes are from the same AS as the best-match route */ + TRIE_WALK(tab_ip->trie, subnet, &dst) + { + net *nc = net_find_valid(tab_ip, &subnet); + if (!nc) + continue; + + const rte *rc = &nc->routes->rte; + if (rc->attrs->source != RTS_BGP) + return 0; + + if (rta_get_first_asn(rc->attrs) != asn_b) + return 0; + } + TRIE_WALK_END; + + return 1; +} + +#endif /* CONFIG_BGP */ + +static struct rte_storage * +rt_flowspec_update_rte(rtable *tab, net *n, rte *r) +{ +#ifdef CONFIG_BGP + if (r->attrs->source != RTS_BGP) + return NULL; + + struct bgp_channel *bc = (struct bgp_channel *) r->sender; + if (!bc->base_table) + return NULL; + + struct bgp_proto *p = (void *) r->src->proto; + int valid = rt_flowspec_check(bc->base_table, tab, n->n.addr, r->attrs, p->is_interior); + int dest = valid ? RTD_NONE : RTD_UNREACHABLE; + + if (dest == r->attrs->dest) + return NULL; + + rta *a = alloca(RTA_MAX_SIZE); + memcpy(a, r->attrs, rta_size(r->attrs)); + a->dest = dest; + a->cached = 0; + + rte new; + memcpy(&new, r, sizeof(rte)); + new.attrs = a; + + return rte_store(&new, n, tab); +#else + return NULL; +#endif +} + + static inline int rt_next_hop_update_net(rtable *tab, net *n) { @@ -1876,9 +2525,14 @@ rt_next_hop_update_net(rtable *tab, net *n) return 0; for (k = &n->routes; e = *k; k = &e->next) - if (rta_next_hop_outdated(e->rte.attrs)) + { + if (!net_is_flow(n->n.addr)) + new = rt_next_hop_update_rte(tab, n, &e->rte); + else + new = rt_flowspec_update_rte(tab, n, &e->rte); + + if (new) { - new = rt_next_hop_update_rte(tab, n, &e->rte); new->next = e->next; *k = new; @@ -1898,6 +2552,7 @@ rt_next_hop_update_net(rtable *tab, net *n) e = new; count++; } + } if (!count) return 0; @@ -1945,6 +2600,9 @@ rt_next_hop_update(rtable *tab) { FIB_ITERATE_INIT(fit, &tab->fib); tab->nhu_state = NHU_RUNNING; + + if (tab->flowspec_trie) + rt_flowspec_reset_trie(tab); } FIB_ITERATE_START(&tab->fib, fit, net, n) @@ -2033,6 +2691,22 @@ rt_unlock_table(rtable *r) } } +static int +rt_reconfigure(rtable *tab, struct rtable_config *new, struct rtable_config *old) +{ + if ((new->addr_type != old->addr_type) || + (new->sorted != old->sorted) || + (new->trie_used != old->trie_used)) + return 0; + + DBG("\t%s: same\n", new->name); + new->table = tab; + tab->name = new->name; + tab->config = new; + + return 1; +} + static struct rtable_config * rt_find_table_config(struct config *cf, char *name) { @@ -2062,28 +2736,19 @@ rt_commit(struct config *new, struct config *old) { WALK_LIST(o, old->tables) { - rtable *ot = o->table; - if (!ot->deleted) - { - r = rt_find_table_config(new, o->name); - if (r && (r->addr_type == o->addr_type) && !new->shutdown) - { - DBG("\t%s: same\n", o->name); - r->table = ot; - ot->name = r->name; - ot->config = r; - if (o->sorted != r->sorted) - log(L_WARN "Reconfiguration of rtable sorted flag not implemented"); - } - else - { - DBG("\t%s: deleted\n", o->name); - ot->deleted = old; - config_add_obstacle(old); - rt_lock_table(ot); - rt_unlock_table(ot); - } - } + rtable *tab = o->table; + if (tab->deleted) + continue; + + r = rt_find_table_config(new, o->name); + if (r && !new->shutdown && rt_reconfigure(tab, r, o)) + continue; + + DBG("\t%s: deleted\n", o->name); + tab->deleted = old; + config_add_obstacle(old); + rt_lock_table(tab); + rt_unlock_table(tab); } } diff --git a/proto/babel/babel.c b/proto/babel/babel.c index 03c85b7d..c7d78b33 100644 --- a/proto/babel/babel.c +++ b/proto/babel/babel.c @@ -63,7 +63,7 @@ static inline void babel_iface_kick_timer(struct babel_iface *ifa); */ static void -babel_init_entry(void *E) +babel_init_entry(struct fib *f UNUSED, void *E) { struct babel_entry *e = E; @@ -1919,7 +1919,7 @@ babel_reconfigure_ifaces(struct babel_proto *p, struct babel_config *cf) struct babel_iface *ifa = babel_find_iface(p, iface); struct babel_iface_config *ic = (void *) iface_patt_find(&cf->iface_list, iface, NULL); - if (ic && iface_is_valid(p, iface)) + if (ic && !iface_is_valid(p, iface)) ic = NULL; if (ifa && ic) diff --git a/proto/bfd/io.c b/proto/bfd/io.c index 1cd9365a..e696cc89 100644 --- a/proto/bfd/io.c +++ b/proto/bfd/io.c @@ -482,6 +482,8 @@ birdloop_main(void *arg) birdloop_set_current(loop); + tmp_init(loop->pool); + pthread_mutex_lock(&loop->mutex); while (1) { diff --git a/proto/bgp/attrs.c b/proto/bgp/attrs.c index 1d9b3630..1b36368f 100644 --- a/proto/bgp/attrs.c +++ b/proto/bgp/attrs.c @@ -45,9 +45,9 @@ * * export - Hook that validates and normalizes attribute during export phase. * Receives eattr, may modify it (e.g., sort community lists for canonical - * representation), UNSET() it (e.g., skip empty lists), or WITHDRAW() it if - * necessary. May assume that eattr has value valid w.r.t. its type, but may be - * invalid w.r.t. BGP constraints. Optional. + * representation), UNSET() it (e.g., skip empty lists), or REJECT() the route + * if necessary. May assume that eattr has value valid w.r.t. its type, but may + * be invalid w.r.t. BGP constraints. Optional. * * encode - Hook that converts internal representation to external one during * packet writing. Receives eattr and puts it in the buffer (including attribute @@ -108,6 +108,9 @@ bgp_set_attr(ea_list **attrs, struct linpool *pool, uint code, uint flags, uintp #define UNSET(a) \ ({ a->type = EAF_TYPE_UNDEF; return; }) +#define REJECT(msg, args...) \ + ({ log(L_ERR "%s: " msg, s->proto->p.name, ## args); s->err_reject = 1; return; }) + #define NEW_BGP "Discarding %s attribute received from AS4-aware neighbor" #define BAD_EBGP "Discarding %s attribute received from EBGP neighbor" #define BAD_LENGTH "Malformed %s attribute - invalid length (%u)" @@ -387,7 +390,7 @@ static void bgp_export_origin(struct bgp_export_state *s, eattr *a) { if (a->u.data > 2) - WITHDRAW(BAD_VALUE, "ORIGIN", a->u.data); + REJECT(BAD_VALUE, "ORIGIN", a->u.data); } static void @@ -909,20 +912,20 @@ bgp_export_mpls_label_stack(struct bgp_export_state *s, eattr *a) /* Perhaps we should just ignore it? */ if (!s->mpls) - WITHDRAW("Unexpected MPLS stack"); + REJECT("Unexpected MPLS stack"); /* Empty MPLS stack is not allowed */ if (!lnum) - WITHDRAW("Malformed MPLS stack - empty"); + REJECT("Malformed MPLS stack - empty"); /* This is ugly, but we must ensure that labels fit into NLRI field */ if ((24*lnum + (net_is_vpn(n) ? 64 : 0) + net_pxlen(n)) > 255) - WITHDRAW("Malformed MPLS stack - too many labels (%u)", lnum); + REJECT("Malformed MPLS stack - too many labels (%u)", lnum); for (uint i = 0; i < lnum; i++) { if (labels[i] > 0xfffff) - WITHDRAW("Malformed MPLS stack - invalid label (%u)", labels[i]); + REJECT("Malformed MPLS stack - invalid label (%u)", labels[i]); /* TODO: Check for special-purpose label values? */ } @@ -1203,7 +1206,7 @@ bgp_export_attrs(struct bgp_export_state *s, ea_list *attrs) for (i = 0; i < count; i++) bgp_export_attr(s, &new->attrs[i], new); - if (s->err_withdraw) + if (s->err_reject) return NULL; return new; @@ -1401,19 +1404,19 @@ bgp_decode_attrs(struct bgp_parse_state *s, byte *data, uint len) /* Reject routes with our ASN in AS_PATH attribute */ if (bgp_as_path_loopy(p, attrs, p->local_as)) - goto withdraw; + goto loop; /* Reject routes with our Confederation ID in AS_PATH attribute; RFC 5065 4.0 */ if ((p->public_as != p->local_as) && bgp_as_path_loopy(p, attrs, p->public_as)) - goto withdraw; + goto loop; /* Reject routes with our Router ID in ORIGINATOR_ID attribute; RFC 4456 8 */ if (p->is_internal && bgp_originator_id_loopy(p, attrs)) - goto withdraw; + goto loop; /* Reject routes with our Cluster ID in CLUSTER_LIST attribute; RFC 4456 8 */ if (p->rr_client && bgp_cluster_list_loopy(p, attrs)) - goto withdraw; + goto loop; /* If there is no local preference, define one */ if (!BIT32_TEST(s->attrs_seen, BA_LOCAL_PREF)) @@ -1434,6 +1437,10 @@ withdraw: s->err_withdraw = 1; return NULL; + +loop: + /* Loops are handled as withdraws, but ignored silently. Do not set err_withdraw. */ + return NULL; } void @@ -1682,6 +1689,10 @@ bgp_preexport(struct channel *c, rte *e) if (src == NULL) return 0; + /* Reject flowspec that failed validation */ + if ((e->attrs->dest == RTD_UNREACHABLE) && net_is_flow(e->net)) + return -1; + /* IBGP route reflection, RFC 4456 */ if (p->is_internal && src->is_internal && (p->local_as == src->local_as)) { @@ -1849,13 +1860,15 @@ bgp_rt_notify(struct proto *P, struct channel *C, const net_addr *n, rte *new, c if (new) { - struct ea_list *attrs = bgp_update_attrs(p, c, new, new->attrs->eattrs, bgp_linpool2); + struct ea_list *attrs = bgp_update_attrs(p, c, new, new->attrs->eattrs, tmp_linpool); + + /* Error during attribute processing */ + if (!attrs) + log(L_ERR "%s: Invalid route %N withdrawn", p->p.name, n); /* If attributes are invalid, we fail back to withdraw */ buck = attrs ? bgp_get_bucket(c, attrs) : bgp_get_withdraw_bucket(c); path = new->src->global_id; - - lp_flush(bgp_linpool2); } else { diff --git a/proto/bgp/bgp.c b/proto/bgp/bgp.c index 5c78bfa1..52400762 100644 --- a/proto/bgp/bgp.c +++ b/proto/bgp/bgp.c @@ -101,6 +101,7 @@ * RFC 8203 - BGP Administrative Shutdown Communication * RFC 8212 - Default EBGP Route Propagation Behavior without Policies * RFC 8654 - Extended Message Support for BGP + * RFC 9117 - Revised Validation Procedure for BGP Flow Specifications * draft-ietf-idr-ext-opt-param-07 * draft-uttaro-idr-bgp-persistence-04 * draft-walton-bgp-hostname-capability-02 @@ -125,9 +126,7 @@ #include "bgp.h" -struct linpool *bgp_linpool; /* Global temporary pool */ -struct linpool *bgp_linpool2; /* Global temporary pool for bgp_rt_notify() */ -static list bgp_sockets; /* Global list of listening sockets */ +static list STATIC_LIST_INIT(bgp_sockets); /* Global list of listening sockets */ static void bgp_connect(struct bgp_proto *p); @@ -157,16 +156,17 @@ bgp_open(struct bgp_proto *p) ip_addr addr = p->cf->strict_bind ? p->cf->local_ip : (p->ipv4 ? IPA_NONE4 : IPA_NONE6); uint port = p->cf->local_port; - - /* FIXME: Add some global init? */ - if (!bgp_linpool) - init_list(&bgp_sockets); + uint flags = p->cf->free_bind ? SKF_FREEBIND : 0; + uint flag_mask = SKF_FREEBIND; /* We assume that cf->iface is defined iff cf->local_ip is link-local */ WALK_LIST(bs, bgp_sockets) - if (ipa_equal(bs->sk->saddr, addr) && (bs->sk->sport == port) && - (bs->sk->iface == ifa) && (bs->sk->vrf == p->p.vrf)) + if (ipa_equal(bs->sk->saddr, addr) && + (bs->sk->sport == port) && + (bs->sk->iface == ifa) && + (bs->sk->vrf == p->p.vrf) && + ((bs->sk->flags & flag_mask) == flags)) { bs->uc++; p->sock = bs; @@ -180,7 +180,7 @@ bgp_open(struct bgp_proto *p) sk->sport = port; sk->iface = ifa; sk->vrf = p->p.vrf; - sk->flags = 0; + sk->flags = flags; sk->tos = IP_PREC_INTERNET_CONTROL; sk->rbsize = BGP_RX_BUFFER_SIZE; sk->tbsize = BGP_TX_BUFFER_SIZE; @@ -198,12 +198,6 @@ bgp_open(struct bgp_proto *p) add_tail(&bgp_sockets, &bs->n); - if (!bgp_linpool) - { - bgp_linpool = lp_new_default(proto_pool); - bgp_linpool2 = lp_new_default(proto_pool); - } - return 0; err: @@ -232,15 +226,6 @@ bgp_close(struct bgp_proto *p) rfree(bs->sk); rem_node(&bs->n); mb_free(bs); - - if (!EMPTY_LIST(bgp_sockets)) - return; - - rfree(bgp_linpool); - bgp_linpool = NULL; - - rfree(bgp_linpool2); - bgp_linpool2 = NULL; } static inline int @@ -1736,6 +1721,9 @@ bgp_channel_init(struct channel *C, struct channel_config *CF) if (cf->igp_table_ip6) c->igp_table_ip6 = cf->igp_table_ip6->table; + + if (cf->base_table) + c->base_table = cf->base_table->table; } static int @@ -1751,6 +1739,12 @@ bgp_channel_start(struct channel *C) if (c->igp_table_ip6) rt_lock_table(c->igp_table_ip6); + if (c->base_table) + { + rt_lock_table(c->base_table); + rt_flowspec_link(c->base_table, c->c.table); + } + c->pool = p->p.pool; // XXXX bgp_init_bucket_table(c); bgp_init_prefix_table(c); @@ -1835,6 +1829,12 @@ bgp_channel_cleanup(struct channel *C) if (c->igp_table_ip6) rt_unlock_table(c->igp_table_ip6); + if (c->base_table) + { + rt_flowspec_unlink(c->base_table, c->c.table); + rt_unlock_table(c->base_table); + } + c->index = 0; /* Cleanup rest of bgp_channel starting at pool field */ @@ -1882,6 +1882,25 @@ bgp_default_igp_table(struct bgp_config *cf, struct bgp_channel_config *cc, u32 cf_error("Undefined IGP table"); } +static struct rtable_config * +bgp_default_base_table(struct bgp_config *cf, struct bgp_channel_config *cc) +{ + /* Expected table type */ + u32 type = (cc->afi == BGP_AF_FLOW4) ? NET_IP4 : NET_IP6; + + /* First, try appropriate IP channel */ + u32 afi2 = BGP_AF(BGP_AFI(cc->afi), BGP_SAFI_UNICAST); + struct bgp_channel_config *cc2 = bgp_find_channel_config(cf, afi2); + if (cc2 && (cc2->c.table->addr_type == type)) + return cc2->c.table; + + /* Last, try default table of given type */ + struct rtable_config *tab = cf->c.global->def_tables[type]; + if (tab) + return tab; + + cf_error("Undefined base table"); +} void bgp_postconfig(struct proto_config *CF) @@ -2026,6 +2045,14 @@ bgp_postconfig(struct proto_config *CF) cf_error("Mismatched IGP table type"); } + /* Default value of base table */ + if ((BGP_SAFI(cc->afi) == BGP_SAFI_FLOW) && cc->validate && !cc->base_table) + cc->base_table = bgp_default_base_table(cf, cc); + + if (cc->base_table && !cc->base_table->trie_used) + cf_error("Flowspec validation requires base table (%s) with trie", + cc->base_table->name); + if (cf->multihop && (cc->gw_mode == GW_DIRECT)) cf_error("Multihop BGP cannot use direct gateway mode"); @@ -2094,7 +2121,7 @@ bgp_reconfigure(struct proto *P, struct proto_config *CF) return same; } -#define IGP_TABLE(cf, sym) ((cf)->igp_table_##sym ? (cf)->igp_table_##sym ->table : NULL ) +#define TABLE(cf, NAME) ((cf)->NAME ? (cf)->NAME->table : NULL ) static int bgp_channel_reconfigure(struct channel *C, struct channel_config *CC, int *import_changed, int *export_changed) @@ -2105,6 +2132,7 @@ bgp_channel_reconfigure(struct channel *C, struct channel_config *CC, int *impor struct bgp_channel_config *old = c->cf; if ((new->secondary != old->secondary) || + (new->validate != old->validate) || (new->gr_able != old->gr_able) || (new->llgr_able != old->llgr_able) || (new->llgr_time != old->llgr_time) || @@ -2112,8 +2140,9 @@ bgp_channel_reconfigure(struct channel *C, struct channel_config *CC, int *impor (new->add_path != old->add_path) || (new->import_table != old->import_table) || (new->export_table != old->export_table) || - (IGP_TABLE(new, ip4) != IGP_TABLE(old, ip4)) || - (IGP_TABLE(new, ip6) != IGP_TABLE(old, ip6))) + (TABLE(new, igp_table_ip4) != TABLE(old, igp_table_ip4)) || + (TABLE(new, igp_table_ip6) != TABLE(old, igp_table_ip6)) || + (TABLE(new, base_table) != TABLE(old, base_table))) return 0; if (new->mandatory && !old->mandatory && (C->channel_state != CS_UP)) @@ -2527,6 +2556,9 @@ bgp_show_proto_info(struct proto *P) if (c->igp_table_ip6) cli_msg(-1006, " IGP IPv6 table: %s", c->igp_table_ip6->name); + + if (c->base_table) + cli_msg(-1006, " Base table: %s", c->base_table->name); } } } diff --git a/proto/bgp/bgp.h b/proto/bgp/bgp.h index c79dd1b2..655b2636 100644 --- a/proto/bgp/bgp.h +++ b/proto/bgp/bgp.h @@ -20,7 +20,6 @@ #include "lib/hash.h" #include "lib/socket.h" -struct linpool; struct eattr; @@ -86,6 +85,7 @@ struct bgp_config { int peer_type; /* Internal or external BGP (BGP_PT_*, optional) */ int multihop; /* Number of hops if multihop */ int strict_bind; /* Bind listening socket to local address */ + int free_bind; /* Bind listening socket with SKF_FREEBIND */ int ttl_security; /* Enable TTL security [RFC 5082] */ int compare_path_lengths; /* Use path lengths when selecting best route */ int med_metric; /* Compare MULTI_EXIT_DISC even between routes from differen ASes */ @@ -146,6 +146,7 @@ struct bgp_channel_config { u8 mandatory; /* Channel is mandatory in capability negotiation */ u8 gw_mode; /* How we compute route gateway from next_hop attr, see GW_* */ u8 secondary; /* Accept also non-best routes (i.e. RA_ACCEPTED) */ + u8 validate; /* Validate Flowspec per RFC 8955 (6) */ u8 gr_able; /* Allow full graceful restart for the channel */ u8 llgr_able; /* Allow full long-lived GR for the channel */ uint llgr_time; /* Long-lived graceful restart stale time */ @@ -159,6 +160,7 @@ struct bgp_channel_config { struct rtable_config *igp_table_ip4; /* Table for recursive IPv4 next hop lookups */ struct rtable_config *igp_table_ip6; /* Table for recursive IPv6 next hop lookups */ + struct rtable_config *base_table; /* Base table for Flowspec validation */ }; #define BGP_PT_INTERNAL 1 @@ -344,6 +346,7 @@ struct bgp_channel { rtable *igp_table_ip4; /* Table for recursive IPv4 next hop lookups */ rtable *igp_table_ip6; /* Table for recursive IPv6 next hop lookups */ + rtable *base_table; /* Base table for Flowspec validation */ /* Rest are zeroed when down */ pool *pool; @@ -400,7 +403,7 @@ struct bgp_export_state { int mpls; u32 attrs_seen[1]; - uint err_withdraw; + uint err_reject; uint local_next_hop; }; @@ -426,6 +429,7 @@ struct bgp_parse_state { int as4_session; int add_path; int mpls; + int reach_nlri_step; u32 attrs_seen[256/32]; @@ -493,9 +497,6 @@ bgp_parse_error(struct bgp_parse_state *s, uint subcode) longjmp(s->err_jmpbuf, 1); } -extern struct linpool *bgp_linpool; -extern struct linpool *bgp_linpool2; - void bgp_start_timer(timer *t, uint value); void bgp_check_config(struct bgp_config *c); @@ -519,7 +520,7 @@ struct rte_source *bgp_get_source(struct bgp_proto *p, u32 path_id); static inline int rta_resolvable(rta *a) { - return a->dest == RTD_UNICAST; + return a->dest != RTD_UNREACHABLE; } diff --git a/proto/bgp/config.Y b/proto/bgp/config.Y index 2dfbdca9..241aa7c2 100644 --- a/proto/bgp/config.Y +++ b/proto/bgp/config.Y @@ -31,7 +31,7 @@ CF_KEYWORDS(BGP, LOCAL, NEIGHBOR, AS, HOLD, TIME, CONNECT, RETRY, KEEPALIVE, STRICT, BIND, CONFEDERATION, MEMBER, MULTICAST, FLOW4, FLOW6, LONG, LIVED, STALE, IMPORT, IBGP, EBGP, MANDATORY, INTERNAL, EXTERNAL, SETS, DYNAMIC, RANGE, NAME, DIGITS, BGP_AIGP, AIGP, ORIGINATE, COST, ENFORCE, - FIRST) + FIRST, FREE, VALIDATE, BASE) %type <i> bgp_nh %type <i32> bgp_afi @@ -155,6 +155,7 @@ bgp_proto: } | bgp_proto DYNAMIC NAME DIGITS expr ';' { BGP_CFG->dynamic_name_digits = $5; if ($5>10) cf_error("Dynamic name digits must be at most 10"); } | bgp_proto STRICT BIND bool ';' { BGP_CFG->strict_bind = $4; } + | bgp_proto FREE BIND bool ';' { BGP_CFG->free_bind = $4; } | bgp_proto PATH METRIC bool ';' { BGP_CFG->compare_path_lengths = $4; } | bgp_proto MED METRIC bool ';' { BGP_CFG->med_metric = $4; } | bgp_proto IGP METRIC bool ';' { BGP_CFG->igp_metric = $4; } @@ -255,6 +256,11 @@ bgp_channel_item: | GATEWAY DIRECT { BGP_CC->gw_mode = GW_DIRECT; } | GATEWAY RECURSIVE { BGP_CC->gw_mode = GW_RECURSIVE; } | SECONDARY bool { BGP_CC->secondary = $2; } + | VALIDATE bool { + BGP_CC->validate = $2; + if (BGP_SAFI(BGP_CC->afi) != BGP_SAFI_FLOW) + cf_error("Validate option limited to flowspec channels"); + } | GRACEFUL RESTART bool { BGP_CC->gr_able = $3; } | LONG LIVED GRACEFUL RESTART bool { BGP_CC->llgr_able = $5; } | LONG LIVED STALE TIME expr { BGP_CC->llgr_time = $5; } @@ -278,6 +284,16 @@ bgp_channel_item: else cf_error("Mismatched IGP table type"); } + | BASE TABLE rtable { + if (BGP_SAFI(BGP_CC->afi) != BGP_SAFI_FLOW) + cf_error("Base table option limited to flowspec channels"); + + if (((BGP_CC->afi == BGP_AF_FLOW4) && ($3->addr_type == NET_IP4)) || + ((BGP_CC->afi == BGP_AF_FLOW6) && ($3->addr_type == NET_IP6))) + BGP_CC->base_table = $3; + else + cf_error("Mismatched base table type"); + } ; bgp_channel_opts: diff --git a/proto/bgp/packets.c b/proto/bgp/packets.c index 647551e5..7be34677 100644 --- a/proto/bgp/packets.c +++ b/proto/bgp/packets.c @@ -932,11 +932,15 @@ bgp_rx_open(struct bgp_conn *conn, byte *pkt, uint len) #define WITHDRAW(msg, args...) \ ({ REPORT(msg, ## args); s->err_withdraw = 1; return; }) +#define REJECT(msg, args...) \ + ({ log(L_ERR "%s: " msg, s->proto->p.name, ## args); s->err_reject = 1; return; }) + #define BAD_AFI "Unexpected AF <%u/%u> in UPDATE" #define BAD_NEXT_HOP "Invalid NEXT_HOP attribute" #define NO_NEXT_HOP "Missing NEXT_HOP attribute" #define NO_LABEL_STACK "Missing MPLS stack" +#define MISMATCHED_AF " - mismatched address family (%I for %s)" static void bgp_apply_next_hop(struct bgp_parse_state *s, rta *a, ip_addr gw, ip_addr ll) @@ -949,13 +953,18 @@ bgp_apply_next_hop(struct bgp_parse_state *s, rta *a, ip_addr gw, ip_addr ll) neighbor *nbr = NULL; /* GW_DIRECT -> single_hop -> p->neigh != NULL */ - if (ipa_nonzero(gw)) + if (ipa_nonzero2(gw)) nbr = neigh_find(&p->p, gw, NULL, 0); else if (ipa_nonzero(ll)) nbr = neigh_find(&p->p, ll, p->neigh->iface, 0); + else + WITHDRAW(BAD_NEXT_HOP " - zero address"); - if (!nbr || (nbr->scope == SCOPE_HOST)) - WITHDRAW(BAD_NEXT_HOP); + if (!nbr) + WITHDRAW(BAD_NEXT_HOP " - address %I not directly reachable", ipa_nonzero(gw) ? gw : ll); + + if (nbr->scope == SCOPE_HOST) + WITHDRAW(BAD_NEXT_HOP " - address %I is local", nbr->addr); a->dest = RTD_UNICAST; a->nh.gw = nbr->addr; @@ -964,8 +973,8 @@ bgp_apply_next_hop(struct bgp_parse_state *s, rta *a, ip_addr gw, ip_addr ll) } else /* GW_RECURSIVE */ { - if (ipa_zero(gw)) - WITHDRAW(BAD_NEXT_HOP); + if (ipa_zero2(gw)) + WITHDRAW(BAD_NEXT_HOP " - zero address"); rtable *tab = ipa_is_ip4(gw) ? c->igp_table_ip4 : c->igp_table_ip6; s->hostentry = rt_get_hostentry(tab, gw, ll, c->c.table); @@ -1009,6 +1018,20 @@ bgp_apply_mpls_labels(struct bgp_parse_state *s, rta *a, u32 *labels, uint lnum) } } +static void +bgp_apply_flow_validation(struct bgp_parse_state *s, const net_addr *n, rta *a) +{ + struct bgp_channel *c = s->channel; + int valid = rt_flowspec_check(c->base_table, c->c.table, n, a, s->proto->is_interior); + a->dest = valid ? RTD_NONE : RTD_UNREACHABLE; + + /* Invalidate cached rta if dest changes */ + if (s->cached_rta && (s->cached_rta->dest != a->dest)) + { + rta_free(s->cached_rta); + s->cached_rta = NULL; + } +} static int bgp_match_src(struct bgp_export_state *s, int mode) @@ -1120,28 +1143,28 @@ bgp_update_next_hop_ip(struct bgp_export_state *s, eattr *a, ea_list **to) /* Check if next hop is valid */ a = bgp_find_attr(*to, BA_NEXT_HOP); if (!a) - WITHDRAW(NO_NEXT_HOP); + REJECT(NO_NEXT_HOP); ip_addr *nh = (void *) a->u.ptr->data; ip_addr peer = s->proto->remote_ip; uint len = a->u.ptr->length; /* Forbid zero next hop */ - if (ipa_zero(nh[0]) && ((len != 32) || ipa_zero(nh[1]))) - WITHDRAW(BAD_NEXT_HOP); + if (ipa_zero2(nh[0]) && ((len != 32) || ipa_zero(nh[1]))) + REJECT(BAD_NEXT_HOP " - zero address"); /* Forbid next hop equal to neighbor IP */ if (ipa_equal(peer, nh[0]) || ((len == 32) && ipa_equal(peer, nh[1]))) - WITHDRAW(BAD_NEXT_HOP); + REJECT(BAD_NEXT_HOP " - neighbor address %I", peer); /* Forbid next hop with non-matching AF */ if ((ipa_is_ip4(nh[0]) != bgp_channel_is_ipv4(s->channel)) && !s->channel->ext_next_hop) - WITHDRAW(BAD_NEXT_HOP); + REJECT(BAD_NEXT_HOP MISMATCHED_AF, nh[0], s->channel->desc->name); /* Just check if MPLS stack */ if (s->mpls && !bgp_find_attr(*to, BA_MPLS_LABEL_STACK)) - WITHDRAW(NO_LABEL_STACK); + REJECT(NO_LABEL_STACK); } static uint @@ -1212,7 +1235,7 @@ bgp_decode_next_hop_ip(struct bgp_parse_state *s, byte *data, uint len, rta *a) ad->length = 16; if ((bgp_channel_is_ipv4(c) != ipa_is_ip4(nh[0])) && !c->ext_next_hop) - WITHDRAW(BAD_NEXT_HOP); + WITHDRAW(BAD_NEXT_HOP MISMATCHED_AF, nh[0], c->desc->name); // XXXX validate next hop @@ -1293,7 +1316,7 @@ bgp_decode_next_hop_vpn(struct bgp_parse_state *s, byte *data, uint len, rta *a) bgp_parse_error(s, 9); if ((bgp_channel_is_ipv4(c) != ipa_is_ip4(nh[0])) && !c->ext_next_hop) - WITHDRAW(BAD_NEXT_HOP); + WITHDRAW(BAD_NEXT_HOP MISMATCHED_AF, nh[0], c->desc->name); // XXXX validate next hop @@ -1335,7 +1358,7 @@ bgp_update_next_hop_none(struct bgp_export_state *s, eattr *a, ea_list **to) */ static void -bgp_rte_update(struct bgp_parse_state *s, net_addr *n, u32 path_id, rta *a0) +bgp_rte_update(struct bgp_parse_state *s, const net_addr *n, u32 path_id, rta *a0) { if (path_id != s->last_id) { @@ -1348,6 +1371,10 @@ bgp_rte_update(struct bgp_parse_state *s, net_addr *n, u32 path_id, rta *a0) if (!a0) { + /* Route update was changed to withdraw */ + if (s->err_withdraw && s->reach_nlri_step) + REPORT("Invalid route %N withdrawn", n); + /* Route withdraw */ rte_update(&s->channel->c, n, NULL, s->last_src); return; @@ -1406,7 +1433,7 @@ bgp_decode_mpls_labels(struct bgp_parse_state *s, byte **pos, uint *len, uint *p /* RFC 8277 2.4 - withdraw does not have variable-size MPLS stack but fixed-size 24-bit Compatibility field, which MUST be ignored */ - if (!a && !s->err_withdraw) + if (!s->reach_nlri_step) return; } while (!(label & BGP_MPLS_BOS)); @@ -1881,6 +1908,10 @@ bgp_decode_nlri_flow4(struct bgp_parse_state *s, byte *pos, uint len, rta *a) net_fill_flow4(n, px, pxlen, pos, flen); ADVANCE(pos, len, flen); + /* Apply validation procedure per RFC 8955 (6) */ + if (a && s->channel->cf->validate) + bgp_apply_flow_validation(s, n, a); + bgp_rte_update(s, n, path_id, a); } } @@ -1969,6 +2000,10 @@ bgp_decode_nlri_flow6(struct bgp_parse_state *s, byte *pos, uint len, rta *a) net_fill_flow6(n, px, pxlen, pos, flen); ADVANCE(pos, len, flen); + /* Apply validation procedure per RFC 8955 (6) */ + if (a && s->channel->cf->validate) + bgp_apply_flow_validation(s, n, a); + bgp_rte_update(s, n, path_id, a); } } @@ -2293,7 +2328,7 @@ again: ; struct bgp_write_state s = { .proto = p, .channel = c, - .pool = bgp_linpool, + .pool = tmp_linpool, .mp_reach = (c->afi != BGP_AF_IPV4) || c->ext_next_hop, .as4_session = p->as4_session, .add_path = c->add_path_tx, @@ -2475,7 +2510,7 @@ bgp_rx_update(struct bgp_conn *conn, byte *pkt, uint len) /* Initialize parse state */ struct bgp_parse_state s = { .proto = p, - .pool = bgp_linpool, + .pool = tmp_linpool, .as4_session = p->as4_session, }; @@ -2541,6 +2576,8 @@ bgp_rx_update(struct bgp_conn *conn, byte *pkt, uint len) if (s.mp_unreach_len) bgp_decode_nlri(&s, s.mp_unreach_af, s.mp_unreach_nlri, s.mp_unreach_len, NULL, NULL, 0); + s.reach_nlri_step = 1; + if (s.ip_reach_len) bgp_decode_nlri(&s, BGP_AF_IPV4, s.ip_reach_nlri, s.ip_reach_len, ea, s.ip_next_hop_data, s.ip_next_hop_len); diff --git a/proto/rpki/packets.c b/proto/rpki/packets.c index dd11f997..d246dd50 100644 --- a/proto/rpki/packets.c +++ b/proto/rpki/packets.c @@ -737,6 +737,33 @@ rpki_handle_prefix_pdu(struct rpki_cache *cache, const struct pdu_header *pdu) net_addr_union addr = {}; rpki_prefix_pdu_2_net_addr(pdu, &addr); + if (type == IPV4_PREFIX) + { + if ((addr.roa4.pxlen > addr.roa4.max_pxlen) || + (addr.roa4.max_pxlen > IP4_MAX_PREFIX_LENGTH)) + { + RPKI_WARN(cache->p, "Received corrupt packet from RPKI cache server: invalid pxlen or max_pxlen"); + byte tmp[pdu->len]; + const struct pdu_header *hton_pdu = rpki_pdu_back_to_network_byte_order((void *) tmp, (const void *) pdu); + rpki_send_error_pdu(cache, CORRUPT_DATA, pdu->len, hton_pdu, "Corrupted PDU: invalid pxlen or max_pxlen"); + rpki_cache_change_state(cache, RPKI_CS_ERROR_FATAL); + return RPKI_ERROR; + } + } + else + { + if ((addr.roa6.pxlen > addr.roa6.max_pxlen) || + (addr.roa6.max_pxlen > IP6_MAX_PREFIX_LENGTH)) + { + RPKI_WARN(cache->p, "Received corrupt packet from RPKI cache server: invalid pxlen or max_pxlen"); + byte tmp[pdu->len]; + const struct pdu_header *hton_pdu = rpki_pdu_back_to_network_byte_order((void *) tmp, (const void *) pdu); + rpki_send_error_pdu(cache, CORRUPT_DATA, pdu->len, hton_pdu, "Corrupted PDU: invalid pxlen or max_pxlen"); + rpki_cache_change_state(cache, RPKI_CS_ERROR_FATAL); + return RPKI_ERROR; + } + } + if (cf->ignore_max_length) { if (type == IPV4_PREFIX) diff --git a/sysdep/bsd/krt-sock.c b/sysdep/bsd/krt-sock.c index 6f788ac2..a37e8786 100644 --- a/sysdep/bsd/krt-sock.c +++ b/sysdep/bsd/krt-sock.c @@ -366,6 +366,30 @@ krt_replace_rte(struct krt_proto *p, net *n, rte *new, rte *old) } } +/** + * krt_assume_onlink - check if routes on interface are considered onlink + * @iface: The interface of the next hop + * @ipv6: Switch to only consider IPv6 or IPv4 addresses. + * + * The BSD kernel does not support an onlink flag. If the interface has only + * host addresses configured, all routes should be considered as onlink and + * the function returns 1. + */ +static int +krt_assume_onlink(struct iface *iface, int ipv6) +{ + const u8 type = ipv6 ? NET_IP6 : NET_IP4; + + struct ifa *ifa; + WALK_LIST(ifa, iface->addrs) + { + if ((ifa->prefix.type == type) && !(ifa->flags & IA_HOST)) + return 0; + } + + return 1; +} + #define SKIP(ARG...) do { DBG("KRT: Ignoring route - " ARG); return; } while(0) static void @@ -526,15 +550,21 @@ krt_read_route(struct ks_msg *msg, struct krt_proto *p, int scan) a.dest = RTD_UNICAST; if (flags & RTF_GATEWAY) { - neighbor *ng; a.nh.gw = igate; /* Clean up embedded interface ID returned in link-local address */ if (ipa_is_link_local(a.nh.gw)) _I0(a.nh.gw) = 0xfe800000; - ng = neigh_find(&p->p, a.nh.gw, a.nh.iface, 0); - if (!ng || (ng->scope == SCOPE_HOST)) + /* The BSD kernel does not support an onlink flag. We heuristically + set the onlink flag, if the iface has only host addresses. */ + if (krt_assume_onlink(a.nh.iface, ipv6)) + a.nh.flags |= RNF_ONLINK; + + neighbor *nbr; + nbr = neigh_find(&p->p, a.nh.gw, a.nh.iface, + (a.nh.flags & RNF_ONLINK) ? NEF_ONLINK : 0); + if (!nbr || (nbr->scope == SCOPE_HOST)) { /* Ignore routes with next-hop 127.0.0.1, host routes with such next-hop appear on OpenBSD for address aliases. */ diff --git a/sysdep/bsd/sysio.h b/sysdep/bsd/sysio.h index c757960a..f1887fb4 100644 --- a/sysdep/bsd/sysio.h +++ b/sysdep/bsd/sysio.h @@ -271,3 +271,9 @@ sk_set_priority(sock *s, int prio UNUSED) { ERR_MSG("Socket priority not supported"); } + +static inline int +sk_set_freebind(sock *s) +{ + ERR_MSG("Freebind is not supported"); +} diff --git a/sysdep/config.h b/sysdep/config.h index 55be90f0..b0531844 100644 --- a/sysdep/config.h +++ b/sysdep/config.h @@ -13,7 +13,7 @@ #ifdef GIT_LABEL #define BIRD_VERSION XSTR1(GIT_LABEL) #else -#define BIRD_VERSION "2.0.8" +#define BIRD_VERSION "2.0.9" #endif /* Include parameters determined by configure script */ diff --git a/sysdep/linux/krt-sys.h b/sysdep/linux/krt-sys.h index a8af4c95..8897f889 100644 --- a/sysdep/linux/krt-sys.h +++ b/sysdep/linux/krt-sys.h @@ -69,6 +69,7 @@ static inline struct ifa * kif_get_primary_ip(struct iface *i UNUSED) { return N struct krt_params { u32 table_id; /* Kernel table ID we sync with */ u32 metric; /* Kernel metric used for all routes */ + uint netlink_rx_buffer; /* Rx buffer size for the netlink socket */ }; struct krt_state { diff --git a/sysdep/linux/netlink.Y b/sysdep/linux/netlink.Y index 7097f577..487ad1d8 100644 --- a/sysdep/linux/netlink.Y +++ b/sysdep/linux/netlink.Y @@ -10,7 +10,8 @@ CF_HDR CF_DECLS -CF_KEYWORDS(KERNEL, TABLE, METRIC, KRT_PREFSRC, KRT_REALM, KRT_SCOPE, KRT_MTU, KRT_WINDOW, +CF_KEYWORDS(KERNEL, TABLE, METRIC, NETLINK, RX, BUFFER, + KRT_PREFSRC, KRT_REALM, KRT_SCOPE, KRT_MTU, KRT_WINDOW, KRT_RTT, KRT_RTTVAR, KRT_SSTRESH, KRT_CWND, KRT_ADVMSS, KRT_REORDERING, KRT_HOPLIMIT, KRT_INITCWND, KRT_RTO_MIN, KRT_INITRWND, KRT_QUICKACK, KRT_LOCK_MTU, KRT_LOCK_WINDOW, KRT_LOCK_RTT, KRT_LOCK_RTTVAR, @@ -24,6 +25,7 @@ kern_proto: kern_proto kern_sys_item ';' ; kern_sys_item: KERNEL TABLE expr { THIS_KRT->sys.table_id = $3; } | METRIC expr { THIS_KRT->sys.metric = $2; } + | NETLINK RX BUFFER expr { THIS_KRT->sys.netlink_rx_buffer = $4; } ; dynamic_attr: KRT_PREFSRC { $$ = f_new_dynamic_attr(EAF_TYPE_IP_ADDRESS, T_IP, EA_KRT_PREFSRC); } ; diff --git a/sysdep/linux/netlink.c b/sysdep/linux/netlink.c index bff2d579..35ba116c 100644 --- a/sysdep/linux/netlink.c +++ b/sysdep/linux/netlink.c @@ -69,6 +69,10 @@ #define RTA_ENCAP 22 #endif +#ifndef NETLINK_GET_STRICT_CHK +#define NETLINK_GET_STRICT_CHK 12 +#endif + #define krt_ipv4(p) ((p)->af == AF_INET) #define krt_ecmp6(p) ((p)->af == AF_INET6) @@ -130,7 +134,7 @@ struct nl_sock uint last_size; }; -#define NL_RX_SIZE 8192 +#define NL_RX_SIZE 32768 #define NL_OP_DELETE 0 #define NL_OP_ADD (NLM_F_CREATE|NLM_F_EXCL) @@ -158,10 +162,46 @@ nl_open_sock(struct nl_sock *nl) } static void +nl_set_strict_dump(struct nl_sock *nl UNUSED, int strict UNUSED) +{ + /* + * Strict checking is not necessary, it improves behavior on newer kernels. + * If it is not available (missing SOL_NETLINK compile-time, or ENOPROTOOPT + * run-time), we can just ignore it. + */ +#ifdef SOL_NETLINK + setsockopt(nl->fd, SOL_NETLINK, NETLINK_GET_STRICT_CHK, &strict, sizeof(strict)); +#endif +} + +static void +nl_set_rcvbuf(int fd, uint val) +{ + if (setsockopt(fd, SOL_SOCKET, SO_RCVBUFFORCE, &val, sizeof(val)) < 0) + log(L_WARN "KRT: Cannot set netlink rx buffer size to %u: %m", val); +} + +static uint +nl_cfg_rx_buffer_size(struct config *cfg) +{ + uint bufsize = 0; + + struct proto_config *pc; + WALK_LIST(pc, cfg->protos) + if ((pc->protocol == &proto_unix_kernel) && !pc->disabled) + bufsize = MAX(bufsize, ((struct krt_config *) pc)->sys.netlink_rx_buffer); + + return bufsize; +} + + +static void nl_open(void) { nl_open_sock(&nl_scan); nl_open_sock(&nl_req); + + nl_set_strict_dump(&nl_scan, 1); } static void @@ -180,20 +220,60 @@ nl_send(struct nl_sock *nl, struct nlmsghdr *nh) } static void -nl_request_dump(int af, int cmd) +nl_request_dump_link(void) { struct { struct nlmsghdr nh; - struct rtgenmsg g; + struct ifinfomsg ifi; } req = { - .nh.nlmsg_type = cmd, - .nh.nlmsg_len = sizeof(req), + .nh.nlmsg_type = RTM_GETLINK, + .nh.nlmsg_len = NLMSG_LENGTH(sizeof(struct ifinfomsg)), .nh.nlmsg_flags = NLM_F_REQUEST | NLM_F_DUMP, - .g.rtgen_family = af + .nh.nlmsg_seq = ++(nl_scan.seq), + .ifi.ifi_family = AF_UNSPEC, }; - nl_send(&nl_scan, &req.nh); + + send(nl_scan.fd, &req, sizeof(req), 0); + nl_scan.last_hdr = NULL; } +static void +nl_request_dump_addr(int af) +{ + struct { + struct nlmsghdr nh; + struct ifaddrmsg ifa; + } req = { + .nh.nlmsg_type = RTM_GETADDR, + .nh.nlmsg_len = NLMSG_LENGTH(sizeof(struct ifaddrmsg)), + .nh.nlmsg_flags = NLM_F_REQUEST | NLM_F_DUMP, + .nh.nlmsg_seq = ++(nl_scan.seq), + .ifa.ifa_family = af, + }; + + send(nl_scan.fd, &req, sizeof(req), 0); + nl_scan.last_hdr = NULL; +} + +static void +nl_request_dump_route(int af) +{ + struct { + struct nlmsghdr nh; + struct rtmsg rtm; + } req = { + .nh.nlmsg_type = RTM_GETROUTE, + .nh.nlmsg_len = NLMSG_LENGTH(sizeof(struct rtmsg)), + .nh.nlmsg_flags = NLM_F_REQUEST | NLM_F_DUMP, + .nh.nlmsg_seq = ++(nl_scan.seq), + .rtm.rtm_family = af, + }; + + send(nl_scan.fd, &req, sizeof(req), 0); + nl_scan.last_hdr = NULL; +} + + static struct nlmsghdr * nl_get_reply(struct nl_sock *nl) { @@ -681,7 +761,7 @@ nl_add_multipath(struct nlmsghdr *h, uint bufsize, struct nexthop *nh, int af, e } static struct nexthop * -nl_parse_multipath(struct nl_parse_state *s, struct krt_proto *p, struct rtattr *ra, int af) +nl_parse_multipath(struct nl_parse_state *s, struct krt_proto *p, const net_addr *n, struct rtattr *ra, int af, int krt_src) { struct rtattr *a[BIRD_RTA_MAX]; struct rtnexthop *nh = RTA_DATA(ra); @@ -695,9 +775,9 @@ nl_parse_multipath(struct nl_parse_state *s, struct krt_proto *p, struct rtattr { /* Use RTNH_OK(nh,len) ?? */ if ((len < sizeof(*nh)) || (len < nh->rtnh_len)) - return NULL; + goto err; - if (nh->rtnh_flags & RTNH_F_DEAD) + if ((nh->rtnh_flags & RTNH_F_DEAD) && (krt_src != KRT_SRC_BIRD)) goto next; *last = rv = lp_allocz(s->pool, NEXTHOP_MAX_SIZE); @@ -706,7 +786,10 @@ nl_parse_multipath(struct nl_parse_state *s, struct krt_proto *p, struct rtattr rv->weight = nh->rtnh_hops; rv->iface = if_find_by_index(nh->rtnh_ifindex); if (!rv->iface) - return NULL; + { + log(L_ERR "KRT: Received route %N with unknown ifindex %u", n, nh->rtnh_ifindex); + return NULL; + } /* Nonexistent RTNH_PAYLOAD ?? */ nl_attr_len = nh->rtnh_len - RTNH_LENGTH(0); @@ -714,18 +797,18 @@ nl_parse_multipath(struct nl_parse_state *s, struct krt_proto *p, struct rtattr { case AF_INET: if (!nl_parse_attrs(RTNH_DATA(nh), nexthop_attr_want4, a, sizeof(a))) - return NULL; + goto err; break; case AF_INET6: if (!nl_parse_attrs(RTNH_DATA(nh), nexthop_attr_want6, a, sizeof(a))) - return NULL; + goto err; break; #ifdef HAVE_MPLS_KERNEL case AF_MPLS: if (!nl_parse_attrs(RTNH_DATA(nh), nexthop_attr_want_mpls, a, sizeof(a))) - return NULL; + goto err; if (a[RTA_NEWDST]) rv->labels = rta_get_mpls(a[RTA_NEWDST], rv->label); @@ -734,7 +817,7 @@ nl_parse_multipath(struct nl_parse_state *s, struct krt_proto *p, struct rtattr #endif default: - return NULL; + goto err; } if (a[RTA_GATEWAY]) @@ -757,14 +840,19 @@ nl_parse_multipath(struct nl_parse_state *s, struct krt_proto *p, struct rtattr nbr = neigh_find(&p->p, rv->gw, rv->iface, (rv->flags & RNF_ONLINK) ? NEF_ONLINK : 0); if (!nbr || (nbr->scope == SCOPE_HOST)) - return NULL; + { + log(L_ERR "KRT: Received route %N with strange next-hop %I", n, rv->gw); + return NULL; + } } #ifdef HAVE_MPLS_KERNEL if (a[RTA_ENCAP] && a[RTA_ENCAP_TYPE]) { - if (rta_get_u16(a[RTA_ENCAP_TYPE]) != LWTUNNEL_ENCAP_MPLS) { - log(L_WARN "KRT: Unknown encapsulation method %d in multipath", rta_get_u16(a[RTA_ENCAP_TYPE])); + if (rta_get_u16(a[RTA_ENCAP_TYPE]) != LWTUNNEL_ENCAP_MPLS) + { + log(L_WARN "KRT: Received route %N with unknown encapsulation method %d", + n, rta_get_u16(a[RTA_ENCAP_TYPE])); return NULL; } @@ -785,6 +873,10 @@ nl_parse_multipath(struct nl_parse_state *s, struct krt_proto *p, struct rtattr first = nexthop_sort(first); return first; + +err: + log(L_ERR "KRT: Received strange multipath route %N", n); + return NULL; } static void @@ -1139,7 +1231,7 @@ kif_do_scan(struct kif_proto *p UNUSED) if_start_update(); - nl_request_dump(AF_UNSPEC, RTM_GETLINK); + nl_request_dump_link(); while (h = nl_get_scan()) if (h->nlmsg_type == RTM_NEWLINK || h->nlmsg_type == RTM_DELLINK) nl_parse_link(h, 1); @@ -1166,14 +1258,14 @@ kif_do_scan(struct kif_proto *p UNUSED) } } - nl_request_dump(AF_INET, RTM_GETADDR); + nl_request_dump_addr(AF_INET); while (h = nl_get_scan()) if (h->nlmsg_type == RTM_NEWADDR || h->nlmsg_type == RTM_DELADDR) nl_parse_addr(h, 1); else log(L_DEBUG "nl_scan_ifaces: Unknown packet received (type=%d)", h->nlmsg_type); - nl_request_dump(AF_INET6, RTM_GETADDR); + nl_request_dump_addr(AF_INET6); while (h = nl_get_scan()) if (h->nlmsg_type == RTM_NEWADDR || h->nlmsg_type == RTM_DELADDR) nl_parse_addr(h, 1); @@ -1532,7 +1624,8 @@ nl_parse_end(struct nl_parse_state *s) } -#define SKIP(ARG...) do { DBG("KRT: Ignoring route - " ARG); return; } while(0) +#define SKIP0(ARG, ...) do { DBG("KRT: Ignoring route - " ARG, ##__VA_ARGS__); return; } while(0) +#define SKIP(ARG, ...) do { DBG("KRT: Ignoring route %N - " ARG, &dst, ##__VA_ARGS__); return; } while(0) static void nl_parse_route(struct nl_parse_state *s, struct nlmsghdr *h) @@ -1585,10 +1678,10 @@ nl_parse_route(struct nl_parse_state *s, struct nlmsghdr *h) return; if (!a[RTA_DST]) - SKIP("MPLS route without RTA_DST"); + SKIP0("MPLS route without RTA_DST\n"); if (rta_get_mpls(a[RTA_DST], rta_mpls_stack) != 1) - SKIP("MPLS route with multi-label RTA_DST"); + SKIP0("MPLS route with multi-label RTA_DST\n"); net_fill_mpls(&dst, rta_mpls_stack[0]); break; @@ -1606,6 +1699,9 @@ nl_parse_route(struct nl_parse_state *s, struct nlmsghdr *h) else table_id = i->rtm_table; + if (i->rtm_flags & RTM_F_CLONED) + SKIP("cloned\n"); + /* Do we know this table? */ p = HASH_FIND(nl_table_map, RTH, i->rtm_family, table_id); if (!p) @@ -1681,19 +1777,16 @@ nl_parse_route(struct nl_parse_state *s, struct nlmsghdr *h) if (a[RTA_MULTIPATH]) { - struct nexthop *nh = nl_parse_multipath(s, p, a[RTA_MULTIPATH], i->rtm_family); + struct nexthop *nh = nl_parse_multipath(s, p, net, a[RTA_MULTIPATH], i->rtm_family, krt_src); if (!nh) - { - log(L_ERR "KRT: Received strange multipath route %N", net); - return; - } + SKIP("strange RTA_MULTIPATH\n"); nexthop_link(ra, nh); break; } - if (i->rtm_flags & RTNH_F_DEAD) - return; + if ((i->rtm_flags & RTNH_F_DEAD) && (krt_src != KRT_SRC_BIRD)) + SKIP("ignore RTNH_F_DEAD\n"); ra->nh.iface = if_find_by_index(oif); if (!ra->nh.iface) @@ -1896,7 +1989,7 @@ krt_do_scan(struct krt_proto *p UNUSED) /* CONFIG_ALL_TABLES_AT_ONCE => p is NUL struct nl_parse_state s; nl_parse_begin(&s, 1); - nl_request_dump(AF_UNSPEC, RTM_GETROUTE); + nl_request_dump_route(AF_UNSPEC); while (h = nl_get_scan()) if (h->nlmsg_type == RTM_NEWROUTE || h->nlmsg_type == RTM_DELROUTE) nl_parse_route(&s, h); @@ -1911,6 +2004,8 @@ krt_do_scan(struct krt_proto *p UNUSED) /* CONFIG_ALL_TABLES_AT_ONCE => p is NUL static sock *nl_async_sk; /* BIRD socket for asynchronous notifications */ static byte *nl_async_rx_buffer; /* Receive buffer */ +static uint nl_async_bufsize; /* Kernel rx buffer size for the netlink socket */ +static struct config *nl_last_config; /* For tracking changes to nl_async_bufsize */ static void nl_async_msg(struct nlmsghdr *h) @@ -2046,6 +2141,32 @@ nl_open_async(void) bug("Netlink: sk_open failed"); } +static void +nl_update_async_bufsize(void) +{ + /* No async socket */ + if (!nl_async_sk) + return; + + /* Already reconfigured */ + if (nl_last_config == config) + return; + + /* Update netlink buffer size */ + uint bufsize = nl_cfg_rx_buffer_size(config); + if (bufsize && (bufsize != nl_async_bufsize)) + { + /* Log message for reconfigurations only */ + if (nl_last_config) + log(L_INFO "KRT: Changing netlink rx buffer size to %u", bufsize); + + nl_set_rcvbuf(nl_async_sk->fd, bufsize); + nl_async_bufsize = bufsize; + } + + nl_last_config = config; +} + /* * Interface to the UNIX krt module @@ -2074,6 +2195,7 @@ krt_sys_start(struct krt_proto *p) nl_open(); nl_open_async(); + nl_update_async_bufsize(); return 1; } @@ -2081,12 +2203,16 @@ krt_sys_start(struct krt_proto *p) void krt_sys_shutdown(struct krt_proto *p) { + nl_update_async_bufsize(); + HASH_REMOVE2(nl_table_map, RTH, krt_pool, p); } int krt_sys_reconfigure(struct krt_proto *p UNUSED, struct krt_config *n, struct krt_config *o) { + nl_update_async_bufsize(); + return (n->sys.table_id == o->sys.table_id) && (n->sys.metric == o->sys.metric); } diff --git a/sysdep/linux/netlink.c.orig b/sysdep/linux/netlink.c.orig new file mode 100644 index 00000000..7cea5322 --- /dev/null +++ b/sysdep/linux/netlink.c.orig @@ -0,0 +1,2179 @@ +/* + * BIRD -- Linux Netlink Interface + * + * (c) 1999--2000 Martin Mares <mj@ucw.cz> + * + * Can be freely distributed and used under the terms of the GNU GPL. + */ + +#include <alloca.h> +#include <stdio.h> +#include <unistd.h> +#include <fcntl.h> +#include <sys/socket.h> +#include <sys/uio.h> +#include <errno.h> + +#undef LOCAL_DEBUG + +#include "nest/bird.h" +#include "nest/route.h" +#include "nest/protocol.h" +#include "nest/iface.h" +#include "lib/alloca.h" +#include "sysdep/unix/unix.h" +#include "sysdep/unix/krt.h" +#include "lib/socket.h" +#include "lib/string.h" +#include "lib/hash.h" +#include "conf/conf.h" + +#include <asm/types.h> +#include <linux/if.h> +#include <linux/netlink.h> +#include <linux/rtnetlink.h> + +#ifdef HAVE_MPLS_KERNEL +#include <linux/lwtunnel.h> +#endif + +#ifndef MSG_TRUNC /* Hack: Several versions of glibc miss this one :( */ +#define MSG_TRUNC 0x20 +#endif + +#ifndef IFA_FLAGS +#define IFA_FLAGS 8 +#endif + +#ifndef IFF_LOWER_UP +#define IFF_LOWER_UP 0x10000 +#endif + +#ifndef RTA_TABLE +#define RTA_TABLE 15 +#endif + +#ifndef RTA_VIA +#define RTA_VIA 18 +#endif + +#ifndef RTA_NEWDST +#define RTA_NEWDST 19 +#endif + +#ifndef RTA_ENCAP_TYPE +#define RTA_ENCAP_TYPE 21 +#endif + +#ifndef RTA_ENCAP +#define RTA_ENCAP 22 +#endif + +#define krt_ipv4(p) ((p)->af == AF_INET) +#define krt_ecmp6(p) ((p)->af == AF_INET6) + +const int rt_default_ecmp = 16; + +/* + * Structure nl_parse_state keeps state of received route processing. Ideally, + * we could just independently parse received Netlink messages and immediately + * propagate received routes to the rest of BIRD, but older Linux kernel (before + * version 4.11) represents and announces IPv6 ECMP routes not as one route with + * multiple next hops (like RTA_MULTIPATH in IPv4 ECMP), but as a sequence of + * routes with the same prefix. More recent kernels work as with IPv4. + * + * Therefore, BIRD keeps currently processed route in nl_parse_state structure + * and postpones its propagation until we expect it to be final; i.e., when + * non-matching route is received or when the scan ends. When another matching + * route is received, it is merged with the already processed route to form an + * ECMP route. Note that merging is done only for IPv6 (merge == 1), but the + * postponing is done in both cases (for simplicity). All IPv4 routes or IPv6 + * routes with RTA_MULTIPATH set are just considered non-matching. + * + * This is ignored for asynchronous notifications (every notification is handled + * as a separate route). It is not an issue for our routes, as we ignore such + * notifications anyways. But importing alien IPv6 ECMP routes does not work + * properly with older kernels. + * + * Whatever the kernel version is, IPv6 ECMP routes are sent as multiple routes + * for the same prefix. + */ + +struct nl_parse_state +{ + struct linpool *pool; + int scan; + int merge; + + net *net; + rta *attrs; + struct krt_proto *proto; + s8 new; + s8 krt_src; + u8 krt_type; + u8 krt_proto; + u32 krt_metric; + + u32 rta_flow; /* Used during parsing */ +}; + +/* + * Synchronous Netlink interface + */ + +struct nl_sock +{ + int fd; + u32 seq; + byte *rx_buffer; /* Receive buffer */ + struct nlmsghdr *last_hdr; /* Recently received packet */ + uint last_size; +}; + +#define NL_RX_SIZE 8192 + +#define NL_OP_DELETE 0 +#define NL_OP_ADD (NLM_F_CREATE|NLM_F_EXCL) +#define NL_OP_REPLACE (NLM_F_CREATE|NLM_F_REPLACE) +#define NL_OP_APPEND (NLM_F_CREATE|NLM_F_APPEND) + +static linpool *nl_linpool; + +static struct nl_sock nl_scan = {.fd = -1}; /* Netlink socket for synchronous scan */ +static struct nl_sock nl_req = {.fd = -1}; /* Netlink socket for requests */ + +static void +nl_open_sock(struct nl_sock *nl) +{ + if (nl->fd < 0) + { + nl->fd = socket(PF_NETLINK, SOCK_RAW, NETLINK_ROUTE); + if (nl->fd < 0) + die("Unable to open rtnetlink socket: %m"); + nl->seq = (u32) (current_time() TO_S); /* Or perhaps random_u32() ? */ + nl->rx_buffer = xmalloc(NL_RX_SIZE); + nl->last_hdr = NULL; + nl->last_size = 0; + } +} + +static void +nl_open(void) +{ + nl_open_sock(&nl_scan); + nl_open_sock(&nl_req); +} + +static void +nl_send(struct nl_sock *nl, struct nlmsghdr *nh) +{ + struct sockaddr_nl sa; + + memset(&sa, 0, sizeof(sa)); + sa.nl_family = AF_NETLINK; + nh->nlmsg_pid = 0; + nh->nlmsg_seq = ++(nl->seq); + nh->nlmsg_len = NLMSG_ALIGN(nh->nlmsg_len); + if (sendto(nl->fd, nh, nh->nlmsg_len, 0, (struct sockaddr *)&sa, sizeof(sa)) < 0) + die("rtnetlink sendto: %m"); + nl->last_hdr = NULL; +} + +static void +nl_request_dump(int af, int cmd) +{ + struct { + struct nlmsghdr nh; + struct rtgenmsg g; + } req = { + .nh.nlmsg_type = cmd, + .nh.nlmsg_len = sizeof(req), + .nh.nlmsg_flags = NLM_F_REQUEST | NLM_F_DUMP, + .g.rtgen_family = af + }; + nl_send(&nl_scan, &req.nh); +} + +static struct nlmsghdr * +nl_get_reply(struct nl_sock *nl) +{ + for(;;) + { + if (!nl->last_hdr) + { + struct iovec iov = { nl->rx_buffer, NL_RX_SIZE }; + struct sockaddr_nl sa; + struct msghdr m = { + .msg_name = &sa, + .msg_namelen = sizeof(sa), + .msg_iov = &iov, + .msg_iovlen = 1, + }; + int x = recvmsg(nl->fd, &m, 0); + if (x < 0) + die("nl_get_reply: %m"); + if (sa.nl_pid) /* It isn't from the kernel */ + { + DBG("Non-kernel packet\n"); + continue; + } + nl->last_size = x; + nl->last_hdr = (void *) nl->rx_buffer; + if (m.msg_flags & MSG_TRUNC) + bug("nl_get_reply: got truncated reply which should be impossible"); + } + if (NLMSG_OK(nl->last_hdr, nl->last_size)) + { + struct nlmsghdr *h = nl->last_hdr; + nl->last_hdr = NLMSG_NEXT(h, nl->last_size); + if (h->nlmsg_seq != nl->seq) + { + log(L_WARN "nl_get_reply: Ignoring out of sequence netlink packet (%x != %x)", + h->nlmsg_seq, nl->seq); + continue; + } + return h; + } + if (nl->last_size) + log(L_WARN "nl_get_reply: Found packet remnant of size %d", nl->last_size); + nl->last_hdr = NULL; + } +} + +static struct tbf rl_netlink_err = TBF_DEFAULT_LOG_LIMITS; + +static int +nl_error(struct nlmsghdr *h, int ignore_esrch) +{ + struct nlmsgerr *e; + int ec; + + if (h->nlmsg_len < NLMSG_LENGTH(sizeof(struct nlmsgerr))) + { + log(L_WARN "Netlink: Truncated error message received"); + return ENOBUFS; + } + e = (struct nlmsgerr *) NLMSG_DATA(h); + ec = -e->error; + if (ec && !(ignore_esrch && (ec == ESRCH))) + log_rl(&rl_netlink_err, L_WARN "Netlink: %s", strerror(ec)); + return ec; +} + +static struct nlmsghdr * +nl_get_scan(void) +{ + struct nlmsghdr *h = nl_get_reply(&nl_scan); + + if (h->nlmsg_type == NLMSG_DONE) + return NULL; + if (h->nlmsg_type == NLMSG_ERROR) + { + nl_error(h, 0); + return NULL; + } + return h; +} + +static int +nl_exchange(struct nlmsghdr *pkt, int ignore_esrch) +{ + struct nlmsghdr *h; + + nl_send(&nl_req, pkt); + for(;;) + { + h = nl_get_reply(&nl_req); + if (h->nlmsg_type == NLMSG_ERROR) + break; + log(L_WARN "nl_exchange: Unexpected reply received"); + } + return nl_error(h, ignore_esrch) ? -1 : 0; +} + +/* + * Netlink attributes + */ + +static int nl_attr_len; + +static void * +nl_checkin(struct nlmsghdr *h, int lsize) +{ + nl_attr_len = h->nlmsg_len - NLMSG_LENGTH(lsize); + if (nl_attr_len < 0) + { + log(L_ERR "nl_checkin: underrun by %d bytes", -nl_attr_len); + return NULL; + } + return NLMSG_DATA(h); +} + +struct nl_want_attrs { + u8 defined:1; + u8 checksize:1; + u8 size; +}; + + +#define BIRD_IFLA_MAX (IFLA_WIRELESS+1) + +static struct nl_want_attrs ifla_attr_want[BIRD_IFLA_MAX] = { + [IFLA_IFNAME] = { 1, 0, 0 }, + [IFLA_MTU] = { 1, 1, sizeof(u32) }, + [IFLA_MASTER] = { 1, 1, sizeof(u32) }, + [IFLA_WIRELESS] = { 1, 0, 0 }, +}; + + +#define BIRD_IFA_MAX (IFA_FLAGS+1) + +static struct nl_want_attrs ifa_attr_want4[BIRD_IFA_MAX] = { + [IFA_ADDRESS] = { 1, 1, sizeof(ip4_addr) }, + [IFA_LOCAL] = { 1, 1, sizeof(ip4_addr) }, + [IFA_BROADCAST] = { 1, 1, sizeof(ip4_addr) }, + [IFA_FLAGS] = { 1, 1, sizeof(u32) }, +}; + +static struct nl_want_attrs ifa_attr_want6[BIRD_IFA_MAX] = { + [IFA_ADDRESS] = { 1, 1, sizeof(ip6_addr) }, + [IFA_LOCAL] = { 1, 1, sizeof(ip6_addr) }, + [IFA_FLAGS] = { 1, 1, sizeof(u32) }, +}; + + +#define BIRD_RTA_MAX (RTA_ENCAP+1) + +static struct nl_want_attrs nexthop_attr_want4[BIRD_RTA_MAX] = { + [RTA_GATEWAY] = { 1, 1, sizeof(ip4_addr) }, + [RTA_VIA] = { 1, 0, 0 }, + [RTA_FLOW] = { 1, 1, sizeof(u32) }, + [RTA_ENCAP_TYPE]= { 1, 1, sizeof(u16) }, + [RTA_ENCAP] = { 1, 0, 0 }, +}; + +static struct nl_want_attrs nexthop_attr_want6[BIRD_RTA_MAX] = { + [RTA_GATEWAY] = { 1, 1, sizeof(ip6_addr) }, + [RTA_VIA] = { 1, 0, 0 }, + [RTA_FLOW] = { 1, 1, sizeof(u32) }, + [RTA_ENCAP_TYPE]= { 1, 1, sizeof(u16) }, + [RTA_ENCAP] = { 1, 0, 0 }, +}; + +#ifdef HAVE_MPLS_KERNEL +static struct nl_want_attrs nexthop_attr_want_mpls[BIRD_RTA_MAX] = { + [RTA_VIA] = { 1, 0, 0 }, + [RTA_NEWDST] = { 1, 0, 0 }, +}; + +static struct nl_want_attrs encap_mpls_want[BIRD_RTA_MAX] = { + [RTA_DST] = { 1, 0, 0 }, +}; +#endif + +static struct nl_want_attrs rtm_attr_want4[BIRD_RTA_MAX] = { + [RTA_DST] = { 1, 1, sizeof(ip4_addr) }, + [RTA_OIF] = { 1, 1, sizeof(u32) }, + [RTA_GATEWAY] = { 1, 1, sizeof(ip4_addr) }, + [RTA_PRIORITY] = { 1, 1, sizeof(u32) }, + [RTA_PREFSRC] = { 1, 1, sizeof(ip4_addr) }, + [RTA_METRICS] = { 1, 0, 0 }, + [RTA_MULTIPATH] = { 1, 0, 0 }, + [RTA_FLOW] = { 1, 1, sizeof(u32) }, + [RTA_TABLE] = { 1, 1, sizeof(u32) }, + [RTA_VIA] = { 1, 0, 0 }, + [RTA_ENCAP_TYPE]= { 1, 1, sizeof(u16) }, + [RTA_ENCAP] = { 1, 0, 0 }, +}; + +static struct nl_want_attrs rtm_attr_want6[BIRD_RTA_MAX] = { + [RTA_DST] = { 1, 1, sizeof(ip6_addr) }, + [RTA_SRC] = { 1, 1, sizeof(ip6_addr) }, + [RTA_IIF] = { 1, 1, sizeof(u32) }, + [RTA_OIF] = { 1, 1, sizeof(u32) }, + [RTA_GATEWAY] = { 1, 1, sizeof(ip6_addr) }, + [RTA_PRIORITY] = { 1, 1, sizeof(u32) }, + [RTA_PREFSRC] = { 1, 1, sizeof(ip6_addr) }, + [RTA_METRICS] = { 1, 0, 0 }, + [RTA_MULTIPATH] = { 1, 0, 0 }, + [RTA_FLOW] = { 1, 1, sizeof(u32) }, + [RTA_TABLE] = { 1, 1, sizeof(u32) }, + [RTA_VIA] = { 1, 0, 0 }, + [RTA_ENCAP_TYPE]= { 1, 1, sizeof(u16) }, + [RTA_ENCAP] = { 1, 0, 0 }, +}; + +#ifdef HAVE_MPLS_KERNEL +static struct nl_want_attrs rtm_attr_want_mpls[BIRD_RTA_MAX] = { + [RTA_DST] = { 1, 1, sizeof(u32) }, + [RTA_IIF] = { 1, 1, sizeof(u32) }, + [RTA_OIF] = { 1, 1, sizeof(u32) }, + [RTA_PRIORITY] = { 1, 1, sizeof(u32) }, + [RTA_METRICS] = { 1, 0, 0 }, + [RTA_MULTIPATH] = { 1, 0, 0 }, + [RTA_FLOW] = { 1, 1, sizeof(u32) }, + [RTA_TABLE] = { 1, 1, sizeof(u32) }, + [RTA_VIA] = { 1, 0, 0 }, + [RTA_NEWDST] = { 1, 0, 0 }, +}; +#endif + + +static int +nl_parse_attrs(struct rtattr *a, struct nl_want_attrs *want, struct rtattr **k, int ksize) +{ + int max = ksize / sizeof(struct rtattr *); + bzero(k, ksize); + + for ( ; RTA_OK(a, nl_attr_len); a = RTA_NEXT(a, nl_attr_len)) + { + if ((a->rta_type >= max) || !want[a->rta_type].defined) + continue; + + if (want[a->rta_type].checksize && (RTA_PAYLOAD(a) != want[a->rta_type].size)) + { + log(L_ERR "nl_parse_attrs: Malformed attribute received"); + return 0; + } + + k[a->rta_type] = a; + } + + if (nl_attr_len) + { + log(L_ERR "nl_parse_attrs: remnant of size %d", nl_attr_len); + return 0; + } + + return 1; +} + +static inline u16 rta_get_u16(struct rtattr *a) +{ return *(u16 *) RTA_DATA(a); } + +static inline u32 rta_get_u32(struct rtattr *a) +{ return *(u32 *) RTA_DATA(a); } + +static inline ip4_addr rta_get_ip4(struct rtattr *a) +{ return ip4_ntoh(*(ip4_addr *) RTA_DATA(a)); } + +static inline ip6_addr rta_get_ip6(struct rtattr *a) +{ return ip6_ntoh(*(ip6_addr *) RTA_DATA(a)); } + +static inline ip_addr rta_get_ipa(struct rtattr *a) +{ + if (RTA_PAYLOAD(a) == sizeof(ip4_addr)) + return ipa_from_ip4(rta_get_ip4(a)); + else + return ipa_from_ip6(rta_get_ip6(a)); +} + +#ifdef HAVE_MPLS_KERNEL +static inline ip_addr rta_get_via(struct rtattr *a) +{ + struct rtvia *v = RTA_DATA(a); + switch(v->rtvia_family) { + case AF_INET: return ipa_from_ip4(ip4_ntoh(*(ip4_addr *) v->rtvia_addr)); + case AF_INET6: return ipa_from_ip6(ip6_ntoh(*(ip6_addr *) v->rtvia_addr)); + } + return IPA_NONE; +} + +static u32 rta_mpls_stack[MPLS_MAX_LABEL_STACK]; +static inline int rta_get_mpls(struct rtattr *a, u32 *stack) +{ + if (!a) + return 0; + + if (RTA_PAYLOAD(a) % 4) + log(L_WARN "KRT: Strange length of received MPLS stack: %u", RTA_PAYLOAD(a)); + + int labels = mpls_get(RTA_DATA(a), RTA_PAYLOAD(a) & ~0x3, stack); + + if (labels < 0) + { + log(L_WARN "KRT: Too long MPLS stack received, ignoring"); + labels = 0; + } + + return labels; +} +#endif + +struct rtattr * +nl_add_attr(struct nlmsghdr *h, uint bufsize, uint code, const void *data, uint dlen) +{ + uint pos = NLMSG_ALIGN(h->nlmsg_len); + uint len = RTA_LENGTH(dlen); + + if (pos + len > bufsize) + bug("nl_add_attr: packet buffer overflow"); + + struct rtattr *a = (struct rtattr *)((char *)h + pos); + a->rta_type = code; + a->rta_len = len; + h->nlmsg_len = pos + len; + + if (dlen > 0) + memcpy(RTA_DATA(a), data, dlen); + + return a; +} + +static inline struct rtattr * +nl_open_attr(struct nlmsghdr *h, uint bufsize, uint code) +{ + return nl_add_attr(h, bufsize, code, NULL, 0); +} + +static inline void +nl_close_attr(struct nlmsghdr *h, struct rtattr *a) +{ + a->rta_len = (void *)h + NLMSG_ALIGN(h->nlmsg_len) - (void *)a; +} + +static inline void +nl_add_attr_u16(struct nlmsghdr *h, uint bufsize, int code, u16 data) +{ + nl_add_attr(h, bufsize, code, &data, 2); +} + +static inline void +nl_add_attr_u32(struct nlmsghdr *h, uint bufsize, int code, u32 data) +{ + nl_add_attr(h, bufsize, code, &data, 4); +} + +static inline void +nl_add_attr_ip4(struct nlmsghdr *h, uint bufsize, int code, ip4_addr ip4) +{ + ip4 = ip4_hton(ip4); + nl_add_attr(h, bufsize, code, &ip4, sizeof(ip4)); +} + +static inline void +nl_add_attr_ip6(struct nlmsghdr *h, uint bufsize, int code, ip6_addr ip6) +{ + ip6 = ip6_hton(ip6); + nl_add_attr(h, bufsize, code, &ip6, sizeof(ip6)); +} + +static inline void +nl_add_attr_ipa(struct nlmsghdr *h, uint bufsize, int code, ip_addr ipa) +{ + if (ipa_is_ip4(ipa)) + nl_add_attr_ip4(h, bufsize, code, ipa_to_ip4(ipa)); + else + nl_add_attr_ip6(h, bufsize, code, ipa_to_ip6(ipa)); +} + +#ifdef HAVE_MPLS_KERNEL +static inline void +nl_add_attr_mpls(struct nlmsghdr *h, uint bufsize, int code, int len, u32 *stack) +{ + char buf[len*4]; + mpls_put(buf, len, stack); + nl_add_attr(h, bufsize, code, buf, len*4); +} + +static inline void +nl_add_attr_mpls_encap(struct nlmsghdr *h, uint bufsize, int len, u32 *stack) +{ + nl_add_attr_u16(h, bufsize, RTA_ENCAP_TYPE, LWTUNNEL_ENCAP_MPLS); + + struct rtattr *nest = nl_open_attr(h, bufsize, RTA_ENCAP); + nl_add_attr_mpls(h, bufsize, RTA_DST, len, stack); + nl_close_attr(h, nest); +} + +static inline void +nl_add_attr_via(struct nlmsghdr *h, uint bufsize, ip_addr ipa) +{ + struct rtvia *via = alloca(sizeof(struct rtvia) + 16); + + if (ipa_is_ip4(ipa)) + { + via->rtvia_family = AF_INET; + put_ip4(via->rtvia_addr, ipa_to_ip4(ipa)); + nl_add_attr(h, bufsize, RTA_VIA, via, sizeof(struct rtvia) + 4); + } + else + { + via->rtvia_family = AF_INET6; + put_ip6(via->rtvia_addr, ipa_to_ip6(ipa)); + nl_add_attr(h, bufsize, RTA_VIA, via, sizeof(struct rtvia) + 16); + } +} +#endif + +static inline struct rtnexthop * +nl_open_nexthop(struct nlmsghdr *h, uint bufsize) +{ + uint pos = NLMSG_ALIGN(h->nlmsg_len); + uint len = RTNH_LENGTH(0); + + if (pos + len > bufsize) + bug("nl_open_nexthop: packet buffer overflow"); + + h->nlmsg_len = pos + len; + + return (void *)h + pos; +} + +static inline void +nl_close_nexthop(struct nlmsghdr *h, struct rtnexthop *nh) +{ + nh->rtnh_len = (void *)h + NLMSG_ALIGN(h->nlmsg_len) - (void *)nh; +} + +static inline void +nl_add_nexthop(struct nlmsghdr *h, uint bufsize, struct nexthop *nh, int af UNUSED) +{ +#ifdef HAVE_MPLS_KERNEL + if (nh->labels > 0) + if (af == AF_MPLS) + nl_add_attr_mpls(h, bufsize, RTA_NEWDST, nh->labels, nh->label); + else + nl_add_attr_mpls_encap(h, bufsize, nh->labels, nh->label); + + if (ipa_nonzero(nh->gw)) + { + if (af == (ipa_is_ip4(nh->gw) ? AF_INET : AF_INET6)) + nl_add_attr_ipa(h, bufsize, RTA_GATEWAY, nh->gw); + else + nl_add_attr_via(h, bufsize, nh->gw); + } +#else + + if (ipa_nonzero(nh->gw)) + nl_add_attr_ipa(h, bufsize, RTA_GATEWAY, nh->gw); +#endif +} + +static void +nl_add_multipath(struct nlmsghdr *h, uint bufsize, struct nexthop *nh, int af, ea_list *eattrs) +{ + struct rtattr *a = nl_open_attr(h, bufsize, RTA_MULTIPATH); + eattr *flow = ea_find(eattrs, EA_KRT_REALM); + + for (; nh; nh = nh->next) + { + struct rtnexthop *rtnh = nl_open_nexthop(h, bufsize); + + rtnh->rtnh_flags = 0; + rtnh->rtnh_hops = nh->weight; + rtnh->rtnh_ifindex = nh->iface->index; + + nl_add_nexthop(h, bufsize, nh, af); + + if (nh->flags & RNF_ONLINK) + rtnh->rtnh_flags |= RTNH_F_ONLINK; + + /* Our KRT_REALM is per-route, but kernel RTA_FLOW is per-nexthop. + Therefore, we need to attach the same attribute to each nexthop. */ + if (flow) + nl_add_attr_u32(h, bufsize, RTA_FLOW, flow->u.data); + + nl_close_nexthop(h, rtnh); + } + + nl_close_attr(h, a); +} + +static struct nexthop * +nl_parse_multipath(struct nl_parse_state *s, struct krt_proto *p, const net_addr *n, struct rtattr *ra, int af, int krt_src) +{ + struct rtattr *a[BIRD_RTA_MAX]; + struct rtnexthop *nh = RTA_DATA(ra); + struct nexthop *rv, *first, **last; + unsigned len = RTA_PAYLOAD(ra); + + first = NULL; + last = &first; + + while (len) + { + /* Use RTNH_OK(nh,len) ?? */ + if ((len < sizeof(*nh)) || (len < nh->rtnh_len)) + goto err; + + if ((nh->rtnh_flags & RTNH_F_DEAD) && (krt_src != KRT_SRC_BIRD)) + goto next; + + *last = rv = lp_allocz(s->pool, NEXTHOP_MAX_SIZE); + last = &(rv->next); + + rv->weight = nh->rtnh_hops; + rv->iface = if_find_by_index(nh->rtnh_ifindex); + if (!rv->iface) + { + log(L_ERR "KRT: Received route %N with unknown ifindex %u", n, nh->rtnh_ifindex); + return NULL; + } + + /* Nonexistent RTNH_PAYLOAD ?? */ + nl_attr_len = nh->rtnh_len - RTNH_LENGTH(0); + switch (af) + { + case AF_INET: + if (!nl_parse_attrs(RTNH_DATA(nh), nexthop_attr_want4, a, sizeof(a))) + goto err; + break; + + case AF_INET6: + if (!nl_parse_attrs(RTNH_DATA(nh), nexthop_attr_want6, a, sizeof(a))) + goto err; + break; + +#ifdef HAVE_MPLS_KERNEL + case AF_MPLS: + if (!nl_parse_attrs(RTNH_DATA(nh), nexthop_attr_want_mpls, a, sizeof(a))) + goto err; + + if (a[RTA_NEWDST]) + rv->labels = rta_get_mpls(a[RTA_NEWDST], rv->label); + + break; +#endif + + default: + goto err; + } + + if (a[RTA_GATEWAY]) + rv->gw = rta_get_ipa(a[RTA_GATEWAY]); + + if (a[RTA_FLOW]) + s->rta_flow = rta_get_u32(a[RTA_FLOW]); + +#ifdef HAVE_MPLS_KERNEL + if (a[RTA_VIA]) + rv->gw = rta_get_via(a[RTA_VIA]); +#endif + + if (ipa_nonzero(rv->gw)) + { + if (nh->rtnh_flags & RTNH_F_ONLINK) + rv->flags |= RNF_ONLINK; + + neighbor *nbr; + nbr = neigh_find(&p->p, rv->gw, rv->iface, + (rv->flags & RNF_ONLINK) ? NEF_ONLINK : 0); + if (!nbr || (nbr->scope == SCOPE_HOST)) + { + log(L_ERR "KRT: Received route %N with strange next-hop %I", n, rv->gw); + return NULL; + } + } + +#ifdef HAVE_MPLS_KERNEL + if (a[RTA_ENCAP] && a[RTA_ENCAP_TYPE]) + { + if (rta_get_u16(a[RTA_ENCAP_TYPE]) != LWTUNNEL_ENCAP_MPLS) + { + log(L_WARN "KRT: Received route %N with unknown encapsulation method %d", + n, rta_get_u16(a[RTA_ENCAP_TYPE])); + return NULL; + } + + struct rtattr *enca[BIRD_RTA_MAX]; + nl_attr_len = RTA_PAYLOAD(a[RTA_ENCAP]); + nl_parse_attrs(RTA_DATA(a[RTA_ENCAP]), encap_mpls_want, enca, sizeof(enca)); + rv->labels = rta_get_mpls(enca[RTA_DST], rv->label); + } +#endif + + next: + len -= NLMSG_ALIGN(nh->rtnh_len); + nh = RTNH_NEXT(nh); + } + + /* Ensure nexthops are sorted to satisfy nest invariant */ + if (!nexthop_is_sorted(first)) + first = nexthop_sort(first); + + return first; + +err: + log(L_ERR "KRT: Received strange multipath route %N", n); + return NULL; +} + +static void +nl_add_metrics(struct nlmsghdr *h, uint bufsize, u32 *metrics, int max) +{ + struct rtattr *a = nl_open_attr(h, bufsize, RTA_METRICS); + int t; + + for (t = 1; t < max; t++) + if (metrics[0] & (1 << t)) + nl_add_attr_u32(h, bufsize, t, metrics[t]); + + nl_close_attr(h, a); +} + +static int +nl_parse_metrics(struct rtattr *hdr, u32 *metrics, int max) +{ + struct rtattr *a = RTA_DATA(hdr); + int len = RTA_PAYLOAD(hdr); + + metrics[0] = 0; + for (; RTA_OK(a, len); a = RTA_NEXT(a, len)) + { + if (a->rta_type == RTA_UNSPEC) + continue; + + if (a->rta_type >= max) + continue; + + if (RTA_PAYLOAD(a) != 4) + return -1; + + metrics[0] |= 1 << a->rta_type; + metrics[a->rta_type] = rta_get_u32(a); + } + + if (len > 0) + return -1; + + return 0; +} + + +/* + * Scanning of interfaces + */ + +static void +nl_parse_link(struct nlmsghdr *h, int scan) +{ + struct ifinfomsg *i; + struct rtattr *a[BIRD_IFLA_MAX]; + int new = h->nlmsg_type == RTM_NEWLINK; + struct iface f = {}; + struct iface *ifi; + char *name; + u32 mtu, master = 0; + uint fl; + + if (!(i = nl_checkin(h, sizeof(*i))) || !nl_parse_attrs(IFLA_RTA(i), ifla_attr_want, a, sizeof(a))) + return; + if (!a[IFLA_IFNAME] || (RTA_PAYLOAD(a[IFLA_IFNAME]) < 2) || !a[IFLA_MTU]) + { + /* + * IFLA_IFNAME and IFLA_MTU are required, in fact, but there may also come + * a message with IFLA_WIRELESS set, where (e.g.) no IFLA_IFNAME exists. + * We simply ignore all such messages with IFLA_WIRELESS without notice. + */ + + if (a[IFLA_WIRELESS]) + return; + + log(L_ERR "KIF: Malformed message received"); + return; + } + + name = RTA_DATA(a[IFLA_IFNAME]); + mtu = rta_get_u32(a[IFLA_MTU]); + + if (a[IFLA_MASTER]) + master = rta_get_u32(a[IFLA_MASTER]); + + ifi = if_find_by_index(i->ifi_index); + if (!new) + { + DBG("KIF: IF%d(%s) goes down\n", i->ifi_index, name); + if (!ifi) + return; + + if_delete(ifi); + } + else + { + DBG("KIF: IF%d(%s) goes up (mtu=%d,flg=%x)\n", i->ifi_index, name, mtu, i->ifi_flags); + if (ifi && strncmp(ifi->name, name, sizeof(ifi->name)-1)) + if_delete(ifi); + + strncpy(f.name, name, sizeof(f.name)-1); + f.index = i->ifi_index; + f.mtu = mtu; + + f.master_index = master; + f.master = if_find_by_index(master); + + fl = i->ifi_flags; + if (fl & IFF_UP) + f.flags |= IF_ADMIN_UP; + if (fl & IFF_LOWER_UP) + f.flags |= IF_LINK_UP; + if (fl & IFF_LOOPBACK) /* Loopback */ + f.flags |= IF_MULTIACCESS | IF_LOOPBACK | IF_IGNORE; + else if (fl & IFF_POINTOPOINT) /* PtP */ + f.flags |= IF_MULTICAST; + else if (fl & IFF_BROADCAST) /* Broadcast */ + f.flags |= IF_MULTIACCESS | IF_BROADCAST | IF_MULTICAST; + else + f.flags |= IF_MULTIACCESS; /* NBMA */ + + if (fl & IFF_MULTICAST) + f.flags |= IF_MULTICAST; + + ifi = if_update(&f); + + if (!scan) + if_end_partial_update(ifi); + } +} + +static void +nl_parse_addr4(struct ifaddrmsg *i, int scan, int new) +{ + struct rtattr *a[BIRD_IFA_MAX]; + struct iface *ifi; + u32 ifa_flags; + int scope; + + if (!nl_parse_attrs(IFA_RTA(i), ifa_attr_want4, a, sizeof(a))) + return; + + if (!a[IFA_LOCAL]) + { + log(L_ERR "KIF: Malformed message received (missing IFA_LOCAL)"); + return; + } + if (!a[IFA_ADDRESS]) + { + log(L_ERR "KIF: Malformed message received (missing IFA_ADDRESS)"); + return; + } + + ifi = if_find_by_index(i->ifa_index); + if (!ifi) + { + log(L_ERR "KIF: Received address message for unknown interface %d", i->ifa_index); + return; + } + + if (a[IFA_FLAGS]) + ifa_flags = rta_get_u32(a[IFA_FLAGS]); + else + ifa_flags = i->ifa_flags; + + struct ifa ifa; + bzero(&ifa, sizeof(ifa)); + ifa.iface = ifi; + if (ifa_flags & IFA_F_SECONDARY) + ifa.flags |= IA_SECONDARY; + + ifa.ip = rta_get_ipa(a[IFA_LOCAL]); + + if (i->ifa_prefixlen > IP4_MAX_PREFIX_LENGTH) + { + log(L_ERR "KIF: Invalid prefix length for interface %s: %d", ifi->name, i->ifa_prefixlen); + new = 0; + } + if (i->ifa_prefixlen == IP4_MAX_PREFIX_LENGTH) + { + ifa.brd = rta_get_ipa(a[IFA_ADDRESS]); + net_fill_ip4(&ifa.prefix, rta_get_ip4(a[IFA_ADDRESS]), i->ifa_prefixlen); + + /* It is either a host address or a peer address */ + if (ipa_equal(ifa.ip, ifa.brd)) + ifa.flags |= IA_HOST; + else + { + ifa.flags |= IA_PEER; + ifa.opposite = ifa.brd; + } + } + else + { + net_fill_ip4(&ifa.prefix, ipa_to_ip4(ifa.ip), i->ifa_prefixlen); + net_normalize(&ifa.prefix); + + if (i->ifa_prefixlen == IP4_MAX_PREFIX_LENGTH - 1) + ifa.opposite = ipa_opposite_m1(ifa.ip); + + if (i->ifa_prefixlen == IP4_MAX_PREFIX_LENGTH - 2) + ifa.opposite = ipa_opposite_m2(ifa.ip); + + if (ifi->flags & IF_BROADCAST) + { + /* If kernel offers us a broadcast address, we trust it */ + if (a[IFA_BROADCAST]) + ifa.brd = ipa_from_ip4(rta_get_ip4(a[IFA_BROADCAST])); + /* Otherwise we create one (except for /31) */ + else if (i->ifa_prefixlen < (IP4_MAX_PREFIX_LENGTH - 1)) + ifa.brd = ipa_from_ip4(ip4_or(ipa_to_ip4(ifa.ip), + ip4_not(ip4_mkmask(i->ifa_prefixlen)))); + } + } + + scope = ipa_classify(ifa.ip); + if (scope < 0) + { + log(L_ERR "KIF: Invalid interface address %I for %s", ifa.ip, ifi->name); + return; + } + ifa.scope = scope & IADDR_SCOPE_MASK; + + DBG("KIF: IF%d(%s): %s IPA %I, flg %x, net %N, brd %I, opp %I\n", + ifi->index, ifi->name, + new ? "added" : "removed", + ifa.ip, ifa.flags, &ifa.prefix, ifa.brd, ifa.opposite); + + if (new) + ifa_update(&ifa); + else + ifa_delete(&ifa); + + if (!scan) + if_end_partial_update(ifi); +} + +static void +nl_parse_addr6(struct ifaddrmsg *i, int scan, int new) +{ + struct rtattr *a[BIRD_IFA_MAX]; + struct iface *ifi; + u32 ifa_flags; + int scope; + + if (!nl_parse_attrs(IFA_RTA(i), ifa_attr_want6, a, sizeof(a))) + return; + + if (!a[IFA_ADDRESS]) + { + log(L_ERR "KIF: Malformed message received (missing IFA_ADDRESS)"); + return; + } + + ifi = if_find_by_index(i->ifa_index); + if (!ifi) + { + log(L_ERR "KIF: Received address message for unknown interface %d", i->ifa_index); + return; + } + + if (a[IFA_FLAGS]) + ifa_flags = rta_get_u32(a[IFA_FLAGS]); + else + ifa_flags = i->ifa_flags; + + struct ifa ifa; + bzero(&ifa, sizeof(ifa)); + ifa.iface = ifi; + if (ifa_flags & IFA_F_SECONDARY) + ifa.flags |= IA_SECONDARY; + + /* Ignore tentative addresses silently */ + if (ifa_flags & IFA_F_TENTATIVE) + return; + + /* IFA_LOCAL can be unset for IPv6 interfaces */ + ifa.ip = rta_get_ipa(a[IFA_LOCAL] ? : a[IFA_ADDRESS]); + + if (i->ifa_prefixlen > IP6_MAX_PREFIX_LENGTH) + { + log(L_ERR "KIF: Invalid prefix length for interface %s: %d", ifi->name, i->ifa_prefixlen); + new = 0; + } + if (i->ifa_prefixlen == IP6_MAX_PREFIX_LENGTH) + { + ifa.brd = rta_get_ipa(a[IFA_ADDRESS]); + net_fill_ip6(&ifa.prefix, rta_get_ip6(a[IFA_ADDRESS]), i->ifa_prefixlen); + + /* It is either a host address or a peer address */ + if (ipa_equal(ifa.ip, ifa.brd)) + ifa.flags |= IA_HOST; + else + { + ifa.flags |= IA_PEER; + ifa.opposite = ifa.brd; + } + } + else + { + net_fill_ip6(&ifa.prefix, ipa_to_ip6(ifa.ip), i->ifa_prefixlen); + net_normalize(&ifa.prefix); + + if (i->ifa_prefixlen == IP6_MAX_PREFIX_LENGTH - 1) + ifa.opposite = ipa_opposite_m1(ifa.ip); + } + + scope = ipa_classify(ifa.ip); + if (scope < 0) + { + log(L_ERR "KIF: Invalid interface address %I for %s", ifa.ip, ifi->name); + return; + } + ifa.scope = scope & IADDR_SCOPE_MASK; + + DBG("KIF: IF%d(%s): %s IPA %I, flg %x, net %N, brd %I, opp %I\n", + ifi->index, ifi->name, + new ? "added" : "removed", + ifa.ip, ifa.flags, &ifa.prefix, ifa.brd, ifa.opposite); + + if (new) + ifa_update(&ifa); + else + ifa_delete(&ifa); + + if (!scan) + if_end_partial_update(ifi); +} + +static void +nl_parse_addr(struct nlmsghdr *h, int scan) +{ + struct ifaddrmsg *i; + + if (!(i = nl_checkin(h, sizeof(*i)))) + return; + + int new = (h->nlmsg_type == RTM_NEWADDR); + + switch (i->ifa_family) + { + case AF_INET: + return nl_parse_addr4(i, scan, new); + + case AF_INET6: + return nl_parse_addr6(i, scan, new); + } +} + +void +kif_do_scan(struct kif_proto *p UNUSED) +{ + struct nlmsghdr *h; + + if_start_update(); + + nl_request_dump(AF_UNSPEC, RTM_GETLINK); + while (h = nl_get_scan()) + if (h->nlmsg_type == RTM_NEWLINK || h->nlmsg_type == RTM_DELLINK) + nl_parse_link(h, 1); + else + log(L_DEBUG "nl_scan_ifaces: Unknown packet received (type=%d)", h->nlmsg_type); + + /* Re-resolve master interface for slaves */ + struct iface *i; + WALK_LIST(i, iface_list) + if (i->master_index) + { + struct iface f = { + .flags = i->flags, + .mtu = i->mtu, + .index = i->index, + .master_index = i->master_index, + .master = if_find_by_index(i->master_index) + }; + + if (f.master != i->master) + { + memcpy(f.name, i->name, sizeof(f.name)); + if_update(&f); + } + } + + nl_request_dump(AF_INET, RTM_GETADDR); + while (h = nl_get_scan()) + if (h->nlmsg_type == RTM_NEWADDR || h->nlmsg_type == RTM_DELADDR) + nl_parse_addr(h, 1); + else + log(L_DEBUG "nl_scan_ifaces: Unknown packet received (type=%d)", h->nlmsg_type); + + nl_request_dump(AF_INET6, RTM_GETADDR); + while (h = nl_get_scan()) + if (h->nlmsg_type == RTM_NEWADDR || h->nlmsg_type == RTM_DELADDR) + nl_parse_addr(h, 1); + else + log(L_DEBUG "nl_scan_ifaces: Unknown packet received (type=%d)", h->nlmsg_type); + + if_end_update(); +} + +/* + * Routes + */ + +static inline u32 +krt_table_id(struct krt_proto *p) +{ + return KRT_CF->sys.table_id; +} + +static HASH(struct krt_proto) nl_table_map; + +#define RTH_KEY(p) p->af, krt_table_id(p) +#define RTH_NEXT(p) p->sys.hash_next +#define RTH_EQ(a1,i1,a2,i2) a1 == a2 && i1 == i2 +#define RTH_FN(a,i) a ^ u32_hash(i) + +#define RTH_REHASH rth_rehash +#define RTH_PARAMS /8, *2, 2, 2, 6, 20 + +HASH_DEFINE_REHASH_FN(RTH, struct krt_proto) + +int +krt_capable(rte *e) +{ + rta *a = e->attrs; + + switch (a->dest) + { + case RTD_UNICAST: + case RTD_BLACKHOLE: + case RTD_UNREACHABLE: + case RTD_PROHIBIT: + return 1; + + default: + return 0; + } +} + +static inline int +nh_bufsize(struct nexthop *nh) +{ + int rv = 0; + for (; nh != NULL; nh = nh->next) + rv += RTNH_LENGTH(RTA_LENGTH(sizeof(ip_addr))); + return rv; +} + +static int +nl_send_route(struct krt_proto *p, rte *e, int op, int dest, struct nexthop *nh) +{ + eattr *ea; + net *net = e->net; + rta *a = e->attrs; + ea_list *eattrs = a->eattrs; + int bufsize = 128 + KRT_METRICS_MAX*8 + nh_bufsize(&(a->nh)); + u32 priority = 0; + + struct { + struct nlmsghdr h; + struct rtmsg r; + char buf[0]; + } *r; + + int rsize = sizeof(*r) + bufsize; + r = alloca(rsize); + + DBG("nl_send_route(%N,op=%x)\n", net->n.addr, op); + + bzero(&r->h, sizeof(r->h)); + bzero(&r->r, sizeof(r->r)); + r->h.nlmsg_type = op ? RTM_NEWROUTE : RTM_DELROUTE; + r->h.nlmsg_len = NLMSG_LENGTH(sizeof(struct rtmsg)); + r->h.nlmsg_flags = op | NLM_F_REQUEST | NLM_F_ACK; + + r->r.rtm_family = p->af; + r->r.rtm_dst_len = net_pxlen(net->n.addr); + r->r.rtm_protocol = RTPROT_BIRD; + r->r.rtm_scope = RT_SCOPE_NOWHERE; +#ifdef HAVE_MPLS_KERNEL + if (p->af == AF_MPLS) + { + /* + * Kernel MPLS code is a bit picky. We must: + * 1) Always set RT_SCOPE_UNIVERSE and RTN_UNICAST (even for RTM_DELROUTE) + * 2) Never use RTA_PRIORITY + */ + + u32 label = net_mpls(net->n.addr); + nl_add_attr_mpls(&r->h, rsize, RTA_DST, 1, &label); + r->r.rtm_scope = RT_SCOPE_UNIVERSE; + r->r.rtm_type = RTN_UNICAST; + } + else +#endif + { + nl_add_attr_ipa(&r->h, rsize, RTA_DST, net_prefix(net->n.addr)); + + /* Add source address for IPv6 SADR routes */ + if (net->n.addr->type == NET_IP6_SADR) + { + net_addr_ip6_sadr *a = (void *) &net->n.addr; + nl_add_attr_ip6(&r->h, rsize, RTA_SRC, a->src_prefix); + r->r.rtm_src_len = a->src_pxlen; + } + } + + /* + * Strange behavior for RTM_DELROUTE: + * 1) rtm_family is ignored in IPv6, works for IPv4 + * 2) not setting RTA_PRIORITY is different from setting default value (on IPv6) + * 3) not setting RTA_PRIORITY is equivalent to setting 0, which is wildcard + */ + + if (krt_table_id(p) < 256) + r->r.rtm_table = krt_table_id(p); + else + nl_add_attr_u32(&r->h, rsize, RTA_TABLE, krt_table_id(p)); + + if (p->af == AF_MPLS) + priority = 0; + else if (a->source == RTS_DUMMY) + priority = e->u.krt.metric; + else if (KRT_CF->sys.metric) + priority = KRT_CF->sys.metric; + else if ((op != NL_OP_DELETE) && (ea = ea_find(eattrs, EA_KRT_METRIC))) + priority = ea->u.data; + + if (priority) + nl_add_attr_u32(&r->h, rsize, RTA_PRIORITY, priority); + + /* For route delete, we do not specify remaining route attributes */ + if (op == NL_OP_DELETE) + goto dest; + + /* Default scope is LINK for device routes, UNIVERSE otherwise */ + if (p->af == AF_MPLS) + r->r.rtm_scope = RT_SCOPE_UNIVERSE; + else if (ea = ea_find(eattrs, EA_KRT_SCOPE)) + r->r.rtm_scope = ea->u.data; + else + r->r.rtm_scope = (dest == RTD_UNICAST && ipa_zero(nh->gw)) ? RT_SCOPE_LINK : RT_SCOPE_UNIVERSE; + + if (ea = ea_find(eattrs, EA_KRT_PREFSRC)) + nl_add_attr_ipa(&r->h, rsize, RTA_PREFSRC, *(ip_addr *)ea->u.ptr->data); + + if (ea = ea_find(eattrs, EA_KRT_REALM)) + nl_add_attr_u32(&r->h, rsize, RTA_FLOW, ea->u.data); + + + u32 metrics[KRT_METRICS_MAX]; + metrics[0] = 0; + + struct ea_walk_state ews = { .eattrs = eattrs }; + while (ea = ea_walk(&ews, EA_KRT_METRICS, KRT_METRICS_MAX)) + { + int id = ea->id - EA_KRT_METRICS; + metrics[0] |= 1 << id; + metrics[id] = ea->u.data; + } + + if (metrics[0]) + nl_add_metrics(&r->h, rsize, metrics, KRT_METRICS_MAX); + + +dest: + switch (dest) + { + case RTD_UNICAST: + r->r.rtm_type = RTN_UNICAST; + if (nh->next && !krt_ecmp6(p)) + nl_add_multipath(&r->h, rsize, nh, p->af, eattrs); + else + { + nl_add_attr_u32(&r->h, rsize, RTA_OIF, nh->iface->index); + nl_add_nexthop(&r->h, rsize, nh, p->af); + + if (nh->flags & RNF_ONLINK) + r->r.rtm_flags |= RTNH_F_ONLINK; + } + break; + case RTD_BLACKHOLE: + r->r.rtm_type = RTN_BLACKHOLE; + break; + case RTD_UNREACHABLE: + r->r.rtm_type = RTN_UNREACHABLE; + break; + case RTD_PROHIBIT: + r->r.rtm_type = RTN_PROHIBIT; + break; + case RTD_NONE: + break; + default: + bug("krt_capable inconsistent with nl_send_route"); + } + + /* Ignore missing for DELETE */ + return nl_exchange(&r->h, (op == NL_OP_DELETE)); +} + +static inline int +nl_add_rte(struct krt_proto *p, rte *e) +{ + rta *a = e->attrs; + int err = 0; + + if (krt_ecmp6(p) && a->nh.next) + { + struct nexthop *nh = &(a->nh); + + err = nl_send_route(p, e, NL_OP_ADD, RTD_UNICAST, nh); + if (err < 0) + return err; + + for (nh = nh->next; nh; nh = nh->next) + err += nl_send_route(p, e, NL_OP_APPEND, RTD_UNICAST, nh); + + return err; + } + + return nl_send_route(p, e, NL_OP_ADD, a->dest, &(a->nh)); +} + +static inline int +nl_delete_rte(struct krt_proto *p, rte *e) +{ + int err = 0; + + /* For IPv6, we just repeatedly request DELETE until we get error */ + do + err = nl_send_route(p, e, NL_OP_DELETE, RTD_NONE, NULL); + while (krt_ecmp6(p) && !err); + + return err; +} + +static inline int +nl_replace_rte(struct krt_proto *p, rte *e) +{ + rta *a = e->attrs; + return nl_send_route(p, e, NL_OP_REPLACE, a->dest, &(a->nh)); +} + + +void +krt_replace_rte(struct krt_proto *p, net *n UNUSED, rte *new, rte *old) +{ + int err = 0; + + /* + * We use NL_OP_REPLACE for IPv4, it has an issue with not checking for + * matching rtm_protocol, but that is OK when dedicated priority is used. + * + * We do not use NL_OP_REPLACE for IPv6, as it has broken semantics for ECMP + * and with some kernel versions ECMP replace crashes kernel. Would need more + * testing and checks for kernel versions. + * + * For IPv6, we use NL_OP_DELETE and then NL_OP_ADD. We also do not trust the + * old route value, so we do not try to optimize IPv6 ECMP reconfigurations. + */ + + if (krt_ipv4(p) && old && new) + { + err = nl_replace_rte(p, new); + } + else + { + if (old) + nl_delete_rte(p, old); + + if (new) + err = nl_add_rte(p, new); + } + + if (new) + { + if (err < 0) + bmap_clear(&p->sync_map, new->id); + else + bmap_set(&p->sync_map, new->id); + } +} + +static int +nl_mergable_route(struct nl_parse_state *s, net *net, struct krt_proto *p, uint priority, uint krt_type, uint rtm_family) +{ + /* Route merging is used for IPv6 scans */ + if (!s->scan || (rtm_family != AF_INET6)) + return 0; + + /* Saved and new route must have same network, proto/table, and priority */ + if ((s->net != net) || (s->proto != p) || (s->krt_metric != priority)) + return 0; + + /* Both must be regular unicast routes */ + if ((s->krt_type != RTN_UNICAST) || (krt_type != RTN_UNICAST)) + return 0; + + return 1; +} + +static void +nl_announce_route(struct nl_parse_state *s) +{ + rte *e = rte_get_temp(s->attrs); + e->net = s->net; + e->u.krt.src = s->krt_src; + e->u.krt.proto = s->krt_proto; + e->u.krt.seen = 0; + e->u.krt.best = 0; + e->u.krt.metric = s->krt_metric; + + if (s->scan) + krt_got_route(s->proto, e); + else + krt_got_route_async(s->proto, e, s->new); + + s->net = NULL; + s->attrs = NULL; + s->proto = NULL; + lp_flush(s->pool); +} + +static inline void +nl_parse_begin(struct nl_parse_state *s, int scan) +{ + memset(s, 0, sizeof (struct nl_parse_state)); + s->pool = nl_linpool; + s->scan = scan; +} + +static inline void +nl_parse_end(struct nl_parse_state *s) +{ + if (s->net) + nl_announce_route(s); +} + + +#define SKIP0(ARG, ...) do { DBG("KRT: Ignoring route - " ARG, ##__VA_ARGS__); return; } while(0) +#define SKIP(ARG, ...) do { DBG("KRT: Ignoring route %N - " ARG, &dst, ##__VA_ARGS__); return; } while(0) + +static void +nl_parse_route(struct nl_parse_state *s, struct nlmsghdr *h) +{ + struct krt_proto *p; + struct rtmsg *i; + struct rtattr *a[BIRD_RTA_MAX]; + int new = h->nlmsg_type == RTM_NEWROUTE; + + net_addr dst, src = {}; + u32 oif = ~0; + u32 table_id; + u32 priority = 0; + u32 def_scope = RT_SCOPE_UNIVERSE; + int krt_src; + + if (!(i = nl_checkin(h, sizeof(*i)))) + return; + + switch (i->rtm_family) + { + case AF_INET: + if (!nl_parse_attrs(RTM_RTA(i), rtm_attr_want4, a, sizeof(a))) + return; + + if (a[RTA_DST]) + net_fill_ip4(&dst, rta_get_ip4(a[RTA_DST]), i->rtm_dst_len); + else + net_fill_ip4(&dst, IP4_NONE, 0); + break; + + case AF_INET6: + if (!nl_parse_attrs(RTM_RTA(i), rtm_attr_want6, a, sizeof(a))) + return; + + if (a[RTA_DST]) + net_fill_ip6(&dst, rta_get_ip6(a[RTA_DST]), i->rtm_dst_len); + else + net_fill_ip6(&dst, IP6_NONE, 0); + + if (a[RTA_SRC]) + net_fill_ip6(&src, rta_get_ip6(a[RTA_SRC]), i->rtm_src_len); + else + net_fill_ip6(&src, IP6_NONE, 0); + break; + +#ifdef HAVE_MPLS_KERNEL + case AF_MPLS: + if (!nl_parse_attrs(RTM_RTA(i), rtm_attr_want_mpls, a, sizeof(a))) + return; + + if (!a[RTA_DST]) + SKIP0("MPLS route without RTA_DST\n"); + + if (rta_get_mpls(a[RTA_DST], rta_mpls_stack) != 1) + SKIP0("MPLS route with multi-label RTA_DST\n"); + + net_fill_mpls(&dst, rta_mpls_stack[0]); + break; +#endif + + default: + return; + } + + if (a[RTA_OIF]) + oif = rta_get_u32(a[RTA_OIF]); + + if (a[RTA_TABLE]) + table_id = rta_get_u32(a[RTA_TABLE]); + else + table_id = i->rtm_table; + + if (i->rtm_flags & RTM_F_CLONED) + SKIP("cloned\n"); + + /* Do we know this table? */ + p = HASH_FIND(nl_table_map, RTH, i->rtm_family, table_id); + if (!p) + SKIP("unknown table %u\n", table_id); + + if (a[RTA_SRC] && (p->p.net_type != NET_IP6_SADR)) + SKIP("src prefix for non-SADR channel\n"); + + if (a[RTA_IIF]) + SKIP("IIF set\n"); + + if (i->rtm_tos != 0) /* We don't support TOS */ + SKIP("TOS %02x\n", i->rtm_tos); + + if (s->scan && !new) + SKIP("RTM_DELROUTE in scan\n"); + + if (a[RTA_PRIORITY]) + priority = rta_get_u32(a[RTA_PRIORITY]); + + int c = net_classify(&dst); + if ((c < 0) || !(c & IADDR_HOST) || ((c & IADDR_SCOPE_MASK) <= SCOPE_LINK)) + SKIP("strange class/scope\n"); + + switch (i->rtm_protocol) + { + case RTPROT_UNSPEC: + SKIP("proto unspec\n"); + + case RTPROT_REDIRECT: + krt_src = KRT_SRC_REDIRECT; + break; + + case RTPROT_KERNEL: + krt_src = KRT_SRC_KERNEL; + return; + + case RTPROT_BIRD: + if (!s->scan) + SKIP("echo\n"); + krt_src = KRT_SRC_BIRD; + break; + + case RTPROT_BOOT: + default: + krt_src = KRT_SRC_ALIEN; + } + + net_addr *n = &dst; + if (p->p.net_type == NET_IP6_SADR) + { + n = alloca(sizeof(net_addr_ip6_sadr)); + net_fill_ip6_sadr(n, net6_prefix(&dst), net6_pxlen(&dst), + net6_prefix(&src), net6_pxlen(&src)); + } + + net *net = net_get(p->p.main_channel->table, n); + + if (s->net && !nl_mergable_route(s, net, p, priority, i->rtm_type, i->rtm_family)) + nl_announce_route(s); + + rta *ra = lp_allocz(s->pool, RTA_MAX_SIZE); + ra->src = p->p.main_source; + ra->source = RTS_INHERIT; + ra->scope = SCOPE_UNIVERSE; + + if (a[RTA_FLOW]) + s->rta_flow = rta_get_u32(a[RTA_FLOW]); + else + s->rta_flow = 0; + + switch (i->rtm_type) + { + case RTN_UNICAST: + ra->dest = RTD_UNICAST; + + if (a[RTA_MULTIPATH]) + { + struct nexthop *nh = nl_parse_multipath(s, p, n, a[RTA_MULTIPATH], i->rtm_family, krt_src); + if (!nh) + SKIP("strange RTA_MULTIPATH\n"); + + nexthop_link(ra, nh); + break; + } + + if ((i->rtm_flags & RTNH_F_DEAD) && (krt_src != KRT_SRC_BIRD)) + SKIP("ignore RTNH_F_DEAD\n"); + + ra->nh.iface = if_find_by_index(oif); + if (!ra->nh.iface) + { + log(L_ERR "KRT: Received route %N with unknown ifindex %u", net->n.addr, oif); + return; + } + + if (a[RTA_GATEWAY]) + ra->nh.gw = rta_get_ipa(a[RTA_GATEWAY]); + +#ifdef HAVE_MPLS_KERNEL + if (a[RTA_VIA]) + ra->nh.gw = rta_get_via(a[RTA_VIA]); +#endif + + if (ipa_nonzero(ra->nh.gw)) + { + /* Silently skip strange 6to4 routes */ + const net_addr_ip6 sit = NET_ADDR_IP6(IP6_NONE, 96); + if ((i->rtm_family == AF_INET6) && ipa_in_netX(ra->nh.gw, (net_addr *) &sit)) + return; + + if (i->rtm_flags & RTNH_F_ONLINK) + ra->nh.flags |= RNF_ONLINK; + + neighbor *nbr; + nbr = neigh_find(&p->p, ra->nh.gw, ra->nh.iface, + (ra->nh.flags & RNF_ONLINK) ? NEF_ONLINK : 0); + if (!nbr || (nbr->scope == SCOPE_HOST)) + { + log(L_ERR "KRT: Received route %N with strange next-hop %I", net->n.addr, + ra->nh.gw); + return; + } + } + + break; + case RTN_BLACKHOLE: + ra->dest = RTD_BLACKHOLE; + break; + case RTN_UNREACHABLE: + ra->dest = RTD_UNREACHABLE; + break; + case RTN_PROHIBIT: + ra->dest = RTD_PROHIBIT; + break; + /* FIXME: What about RTN_THROW? */ + default: + SKIP("type %d\n", i->rtm_type); + return; + } + +#ifdef HAVE_MPLS_KERNEL + if ((i->rtm_family == AF_MPLS) && a[RTA_NEWDST] && !ra->nh.next) + ra->nh.labels = rta_get_mpls(a[RTA_NEWDST], ra->nh.label); + + if (a[RTA_ENCAP] && a[RTA_ENCAP_TYPE] && !ra->nh.next) + { + switch (rta_get_u16(a[RTA_ENCAP_TYPE])) + { + case LWTUNNEL_ENCAP_MPLS: + { + struct rtattr *enca[BIRD_RTA_MAX]; + nl_attr_len = RTA_PAYLOAD(a[RTA_ENCAP]); + nl_parse_attrs(RTA_DATA(a[RTA_ENCAP]), encap_mpls_want, enca, sizeof(enca)); + ra->nh.labels = rta_get_mpls(enca[RTA_DST], ra->nh.label); + break; + } + default: + SKIP("unknown encapsulation method %d\n", rta_get_u16(a[RTA_ENCAP_TYPE])); + break; + } + } +#endif + + if (i->rtm_scope != def_scope) + { + ea_list *ea = lp_alloc(s->pool, sizeof(ea_list) + sizeof(eattr)); + ea->next = ra->eattrs; + ra->eattrs = ea; + ea->flags = EALF_SORTED; + ea->count = 1; + ea->attrs[0].id = EA_KRT_SCOPE; + ea->attrs[0].flags = 0; + ea->attrs[0].type = EAF_TYPE_INT; + ea->attrs[0].u.data = i->rtm_scope; + } + + if (a[RTA_PREFSRC]) + { + ip_addr ps = rta_get_ipa(a[RTA_PREFSRC]); + + ea_list *ea = lp_alloc(s->pool, sizeof(ea_list) + sizeof(eattr)); + ea->next = ra->eattrs; + ra->eattrs = ea; + ea->flags = EALF_SORTED; + ea->count = 1; + ea->attrs[0].id = EA_KRT_PREFSRC; + ea->attrs[0].flags = 0; + ea->attrs[0].type = EAF_TYPE_IP_ADDRESS; + + struct adata *ad = lp_alloc(s->pool, sizeof(struct adata) + sizeof(ps)); + ad->length = sizeof(ps); + memcpy(ad->data, &ps, sizeof(ps)); + + ea->attrs[0].u.ptr = ad; + } + + /* Can be set per-route or per-nexthop */ + if (s->rta_flow) + { + ea_list *ea = lp_alloc(s->pool, sizeof(ea_list) + sizeof(eattr)); + ea->next = ra->eattrs; + ra->eattrs = ea; + ea->flags = EALF_SORTED; + ea->count = 1; + ea->attrs[0].id = EA_KRT_REALM; + ea->attrs[0].flags = 0; + ea->attrs[0].type = EAF_TYPE_INT; + ea->attrs[0].u.data = s->rta_flow; + } + + if (a[RTA_METRICS]) + { + u32 metrics[KRT_METRICS_MAX]; + ea_list *ea = lp_alloc(s->pool, sizeof(ea_list) + KRT_METRICS_MAX * sizeof(eattr)); + int t, n = 0; + + if (nl_parse_metrics(a[RTA_METRICS], metrics, ARRAY_SIZE(metrics)) < 0) + { + log(L_ERR "KRT: Received route %N with strange RTA_METRICS attribute", net->n.addr); + return; + } + + for (t = 1; t < KRT_METRICS_MAX; t++) + if (metrics[0] & (1 << t)) + { + ea->attrs[n].id = EA_CODE(PROTOCOL_KERNEL, KRT_METRICS_OFFSET + t); + ea->attrs[n].flags = 0; + ea->attrs[n].type = EAF_TYPE_INT; /* FIXME: Some are EAF_TYPE_BITFIELD */ + ea->attrs[n].u.data = metrics[t]; + n++; + } + + if (n > 0) + { + ea->next = ra->eattrs; + ea->flags = EALF_SORTED; + ea->count = n; + ra->eattrs = ea; + } + } + + /* + * Ideally, now we would send the received route to the rest of kernel code. + * But IPv6 ECMP routes before 4.11 are sent as a sequence of routes, so we + * postpone it and merge next hops until the end of the sequence. Note that + * when doing merging of next hops, we expect the new route to be unipath. + * Otherwise, we ignore additional next hops in nexthop_insert(). + */ + + if (!s->net) + { + /* Store the new route */ + s->net = net; + s->attrs = ra; + s->proto = p; + s->new = new; + s->krt_src = krt_src; + s->krt_type = i->rtm_type; + s->krt_proto = i->rtm_protocol; + s->krt_metric = priority; + } + else + { + /* Merge next hops with the stored route */ + rta *oa = s->attrs; + + struct nexthop *nhs = &oa->nh; + nexthop_insert(&nhs, &ra->nh); + + /* Perhaps new nexthop is inserted at the first position */ + if (nhs == &ra->nh) + { + /* Swap rtas */ + s->attrs = ra; + + /* Keep old eattrs */ + ra->eattrs = oa->eattrs; + } + } +} + +void +krt_do_scan(struct krt_proto *p UNUSED) /* CONFIG_ALL_TABLES_AT_ONCE => p is NULL */ +{ + struct nlmsghdr *h; + struct nl_parse_state s; + + nl_parse_begin(&s, 1); + nl_request_dump(AF_UNSPEC, RTM_GETROUTE); + while (h = nl_get_scan()) + if (h->nlmsg_type == RTM_NEWROUTE || h->nlmsg_type == RTM_DELROUTE) + nl_parse_route(&s, h); + else + log(L_DEBUG "nl_scan_fire: Unknown packet received (type=%d)", h->nlmsg_type); + nl_parse_end(&s); +} + +/* + * Asynchronous Netlink interface + */ + +static sock *nl_async_sk; /* BIRD socket for asynchronous notifications */ +static byte *nl_async_rx_buffer; /* Receive buffer */ + +static void +nl_async_msg(struct nlmsghdr *h) +{ + struct nl_parse_state s; + + switch (h->nlmsg_type) + { + case RTM_NEWROUTE: + case RTM_DELROUTE: + DBG("KRT: Received async route notification (%d)\n", h->nlmsg_type); + nl_parse_begin(&s, 0); + nl_parse_route(&s, h); + nl_parse_end(&s); + break; + case RTM_NEWLINK: + case RTM_DELLINK: + DBG("KRT: Received async link notification (%d)\n", h->nlmsg_type); + if (kif_proto) + nl_parse_link(h, 0); + break; + case RTM_NEWADDR: + case RTM_DELADDR: + DBG("KRT: Received async address notification (%d)\n", h->nlmsg_type); + if (kif_proto) + nl_parse_addr(h, 0); + break; + default: + DBG("KRT: Received unknown async notification (%d)\n", h->nlmsg_type); + } +} + +static int +nl_async_hook(sock *sk, uint size UNUSED) +{ + struct iovec iov = { nl_async_rx_buffer, NL_RX_SIZE }; + struct sockaddr_nl sa; + struct msghdr m = { + .msg_name = &sa, + .msg_namelen = sizeof(sa), + .msg_iov = &iov, + .msg_iovlen = 1, + }; + struct nlmsghdr *h; + int x; + uint len; + + x = recvmsg(sk->fd, &m, 0); + if (x < 0) + { + if (errno == ENOBUFS) + { + /* + * Netlink reports some packets have been thrown away. + * One day we might react to it by asking for route table + * scan in near future. + */ + log(L_WARN "Kernel dropped some netlink messages, will resync on next scan."); + return 1; /* More data are likely to be ready */ + } + else if (errno != EWOULDBLOCK) + log(L_ERR "Netlink recvmsg: %m"); + return 0; + } + if (sa.nl_pid) /* It isn't from the kernel */ + { + DBG("Non-kernel packet\n"); + return 1; + } + h = (void *) nl_async_rx_buffer; + len = x; + if (m.msg_flags & MSG_TRUNC) + { + log(L_WARN "Netlink got truncated asynchronous message"); + return 1; + } + while (NLMSG_OK(h, len)) + { + nl_async_msg(h); + h = NLMSG_NEXT(h, len); + } + if (len) + log(L_WARN "nl_async_hook: Found packet remnant of size %d", len); + return 1; +} + +static void +nl_async_err_hook(sock *sk, int e UNUSED) +{ + nl_async_hook(sk, 0); +} + +static void +nl_open_async(void) +{ + sock *sk; + struct sockaddr_nl sa; + int fd; + + if (nl_async_sk) + return; + + DBG("KRT: Opening async netlink socket\n"); + + fd = socket(PF_NETLINK, SOCK_RAW, NETLINK_ROUTE); + if (fd < 0) + { + log(L_ERR "Unable to open asynchronous rtnetlink socket: %m"); + return; + } + + bzero(&sa, sizeof(sa)); + sa.nl_family = AF_NETLINK; + sa.nl_groups = RTMGRP_LINK | + RTMGRP_IPV4_IFADDR | RTMGRP_IPV4_ROUTE | + RTMGRP_IPV6_IFADDR | RTMGRP_IPV6_ROUTE; + + if (bind(fd, (struct sockaddr *) &sa, sizeof(sa)) < 0) + { + log(L_ERR "Unable to bind asynchronous rtnetlink socket: %m"); + close(fd); + return; + } + + nl_async_rx_buffer = xmalloc(NL_RX_SIZE); + + sk = nl_async_sk = sk_new(krt_pool); + sk->type = SK_MAGIC; + sk->rx_hook = nl_async_hook; + sk->err_hook = nl_async_err_hook; + sk->fd = fd; + if (sk_open(sk) < 0) + bug("Netlink: sk_open failed"); +} + + +/* + * Interface to the UNIX krt module + */ + +void +krt_sys_io_init(void) +{ + nl_linpool = lp_new_default(krt_pool); + HASH_INIT(nl_table_map, krt_pool, 6); +} + +int +krt_sys_start(struct krt_proto *p) +{ + struct krt_proto *old = HASH_FIND(nl_table_map, RTH, p->af, krt_table_id(p)); + + if (old) + { + log(L_ERR "%s: Kernel table %u already registered by %s", + p->p.name, krt_table_id(p), old->p.name); + return 0; + } + + HASH_INSERT2(nl_table_map, RTH, krt_pool, p); + + nl_open(); + nl_open_async(); + + return 1; +} + +void +krt_sys_shutdown(struct krt_proto *p) +{ + HASH_REMOVE2(nl_table_map, RTH, krt_pool, p); +} + +int +krt_sys_reconfigure(struct krt_proto *p UNUSED, struct krt_config *n, struct krt_config *o) +{ + return (n->sys.table_id == o->sys.table_id) && (n->sys.metric == o->sys.metric); +} + +void +krt_sys_init_config(struct krt_config *cf) +{ + cf->sys.table_id = RT_TABLE_MAIN; + cf->sys.metric = 32; +} + +void +krt_sys_copy_config(struct krt_config *d, struct krt_config *s) +{ + d->sys.table_id = s->sys.table_id; + d->sys.metric = s->sys.metric; +} + +static const char *krt_metrics_names[KRT_METRICS_MAX] = { + NULL, "lock", "mtu", "window", "rtt", "rttvar", "sstresh", "cwnd", "advmss", + "reordering", "hoplimit", "initcwnd", "features", "rto_min", "initrwnd", "quickack" +}; + +static const char *krt_features_names[KRT_FEATURES_MAX] = { + "ecn", NULL, NULL, "allfrag" +}; + +int +krt_sys_get_attr(const eattr *a, byte *buf, int buflen UNUSED) +{ + switch (a->id) + { + case EA_KRT_PREFSRC: + bsprintf(buf, "prefsrc"); + return GA_NAME; + + case EA_KRT_REALM: + bsprintf(buf, "realm"); + return GA_NAME; + + case EA_KRT_SCOPE: + bsprintf(buf, "scope"); + return GA_NAME; + + case EA_KRT_LOCK: + buf += bsprintf(buf, "lock:"); + ea_format_bitfield(a, buf, buflen, krt_metrics_names, 2, KRT_METRICS_MAX); + return GA_FULL; + + case EA_KRT_FEATURES: + buf += bsprintf(buf, "features:"); + ea_format_bitfield(a, buf, buflen, krt_features_names, 0, KRT_FEATURES_MAX); + return GA_FULL; + + default:; + int id = (int)EA_ID(a->id) - KRT_METRICS_OFFSET; + if (id > 0 && id < KRT_METRICS_MAX) + { + bsprintf(buf, "%s", krt_metrics_names[id]); + return GA_NAME; + } + + return GA_UNKNOWN; + } +} + + + +void +kif_sys_start(struct kif_proto *p UNUSED) +{ + nl_open(); + nl_open_async(); +} + +void +kif_sys_shutdown(struct kif_proto *p UNUSED) +{ +} + +int +kif_update_sysdep_addr(struct iface *i UNUSED) +{ + return 0; +} diff --git a/sysdep/linux/sysio.h b/sysdep/linux/sysio.h index e21ff487..f13eda7c 100644 --- a/sysdep/linux/sysio.h +++ b/sysdep/linux/sysio.h @@ -10,6 +10,10 @@ #define IPV6_MINHOPCOUNT 73 #endif +#ifndef IPV6_FREEBIND +#define IPV6_FREEBIND 78 +#endif + #ifndef TCP_MD5SIG_EXT #define TCP_MD5SIG_EXT 32 #endif @@ -266,3 +270,18 @@ sk_set_priority(sock *s, int prio) return 0; } +static inline int +sk_set_freebind(sock *s) +{ + int y = 1; + + if (sk_is_ipv4(s)) + if (setsockopt(s->fd, SOL_IP, IP_FREEBIND, &y, sizeof(y)) < 0) + ERR("IP_FREEBIND"); + + if (sk_is_ipv6(s)) + if (setsockopt(s->fd, SOL_IPV6, IPV6_FREEBIND, &y, sizeof(y)) < 0) + ERR("IPV6_FREEBIND"); + + return 0; +} diff --git a/sysdep/unix/alloc.c b/sysdep/unix/alloc.c index 4c9d5eb5..755b5fa5 100644 --- a/sysdep/unix/alloc.c +++ b/sysdep/unix/alloc.c @@ -8,7 +8,10 @@ #include "nest/bird.h" #include "lib/resource.h" +#include "lib/lists.h" +#include "lib/event.h" +#include <errno.h> #include <stdlib.h> #include <unistd.h> @@ -17,82 +20,168 @@ #endif long page_size = 0; -_Bool alloc_multipage = 0; #ifdef HAVE_MMAP +#define KEEP_PAGES_MAIN_MAX 256 +#define KEEP_PAGES_MAIN_MIN 8 +#define CLEANUP_PAGES_BULK 256 + +_Static_assert(KEEP_PAGES_MAIN_MIN * 4 < KEEP_PAGES_MAIN_MAX); + static _Bool use_fake = 0; + +#if DEBUGGING +struct free_page { + node unused[42]; + node n; +}; #else -static _Bool use_fake = 1; +struct free_page { + node n; +}; +#endif + +struct free_pages { + list pages; + u16 min, max; /* Minimal and maximal number of free pages kept */ + uint cnt; /* Number of empty pages */ + event cleanup; +}; + +static void global_free_pages_cleanup_event(void *); + +static struct free_pages global_free_pages = { + .min = KEEP_PAGES_MAIN_MIN, + .max = KEEP_PAGES_MAIN_MAX, + .cleanup = { .hook = global_free_pages_cleanup_event }, +}; + +uint *pages_kept = &global_free_pages.cnt; + +static void * +alloc_sys_page(void) +{ + void *ptr = mmap(NULL, page_size, PROT_WRITE | PROT_READ, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0); + + if (ptr == MAP_FAILED) + bug("mmap(%lu) failed: %m", page_size); + + return ptr; +} + +extern int shutting_down; /* Shutdown requested. */ + +#else // ! HAVE_MMAP +#define use_fake 1 #endif -void resource_sys_init(void) +void * +alloc_page(void) { + if (use_fake) + { + void *ptr = NULL; + int err = posix_memalign(&ptr, page_size, page_size); + + if (err || !ptr) + bug("posix_memalign(%lu) failed", (long unsigned int) page_size); + + return ptr; + } + #ifdef HAVE_MMAP - if (!(page_size = sysconf(_SC_PAGESIZE))) - die("System page size must be non-zero"); + struct free_pages *fps = &global_free_pages; - if ((u64_popcount(page_size) > 1) || (page_size > 16384)) + if (fps->cnt) { -#endif - /* Too big or strange page, use the aligned allocator instead */ - page_size = 4096; - use_fake = 1; + struct free_page *fp = SKIP_BACK(struct free_page, n, HEAD(fps->pages)); + rem_node(&fp->n); + if ((--fps->cnt < fps->min) && !shutting_down) + ev_schedule(&fps->cleanup); + + bzero(fp, page_size); + return fp; } + + return alloc_sys_page(); +#endif } -void * -alloc_sys_page(void) +void +free_page(void *ptr) { -#ifdef HAVE_MMAP - if (!use_fake) + if (use_fake) { - if (alloc_multipage) - { - void *big = mmap(NULL, page_size * 2, PROT_WRITE | PROT_READ, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0); - if (big == MAP_FAILED) - bug("mmap(%lu) failed: %m", page_size); - - uintptr_t offset = ((uintptr_t) big) % page_size; - if (offset) - { - void *ret = big + page_size - offset; - munmap(big, page_size - offset); - munmap(ret + page_size, offset); - return ret; - } - else - { - munmap(big + page_size, page_size); - return big; - } - } - - void *ret = mmap(NULL, page_size, PROT_WRITE | PROT_READ, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0); - if (ret == MAP_FAILED) - bug("mmap(%lu) failed: %m", page_size); - - return ret; + free(ptr); + return; } - else + +#ifdef HAVE_MMAP + struct free_pages *fps = &global_free_pages; + struct free_page *fp = ptr; + + fp->n = (node) {}; + add_tail(&fps->pages, &fp->n); + + if ((++fps->cnt > fps->max) && !shutting_down) + ev_schedule(&fps->cleanup); #endif +} + +#ifdef HAVE_MMAP +static void +global_free_pages_cleanup_event(void *data UNUSED) +{ + if (shutting_down) + return; + + struct free_pages *fps = &global_free_pages; + + while (fps->cnt / 2 < fps->min) { - void *ret = aligned_alloc(page_size, page_size); - if (!ret) - bug("aligned_alloc(%lu) failed", page_size); - return ret; + struct free_page *fp = alloc_sys_page(); + fp->n = (node) {}; + add_tail(&fps->pages, &fp->n); + fps->cnt++; + } + + for (uint seen = 0; (seen < CLEANUP_PAGES_BULK) && (fps->cnt > fps->max / 2); seen++) + { + struct free_page *fp = SKIP_BACK(struct free_page, n, TAIL(fps->pages)); + rem_node(&fp->n); + + if (munmap(fp, page_size) == 0) + fps->cnt--; + else if (errno == ENOMEM) + add_head(&fps->pages, &fp->n); + else + bug("munmap(%p) failed: %m", fp); } } +#endif void -free_sys_page(void *ptr) +resource_sys_init(void) { #ifdef HAVE_MMAP - if (!use_fake) + ASSERT_DIE(global_free_pages.cnt == 0); + + if (!(page_size = sysconf(_SC_PAGESIZE))) + die("System page size must be non-zero"); + + if (u64_popcount(page_size) == 1) { - if (munmap(ptr, page_size) < 0) - bug("munmap(%p) failed: %m", ptr); + struct free_pages *fps = &global_free_pages; + + init_list(&fps->pages); + global_free_pages_cleanup_event(NULL); + return; } - else + + /* Too big or strange page, use the aligned allocator instead */ + log(L_WARN "Got strange memory page size (%lu), using the aligned allocator instead", page_size); + use_fake = 1; #endif - free(ptr); + + page_size = 4096; } diff --git a/sysdep/unix/io.c b/sysdep/unix/io.c index 3d67d0a7..8a116789 100644 --- a/sysdep/unix/io.c +++ b/sysdep/unix/io.c @@ -1436,6 +1436,10 @@ sk_open(sock *s) if (sk_set_high_port(s) < 0) log(L_WARN "Socket error: %s%#m", s->err); + if (s->flags & SKF_FREEBIND) + if (sk_set_freebind(s) < 0) + log(L_WARN "Socket error: %s%#m", s->err); + sockaddr_fill(&sa, s->af, bind_addr, s->iface, bind_port); if (bind(fd, &sa.sa, SA_LEN(sa)) < 0) ERR2("bind"); @@ -1850,8 +1854,8 @@ sk_read_ssh(sock *s) /* sk_read() and sk_write() are called from BFD's event loop */ -int -sk_read(sock *s, int revents) +static inline int +sk_read_noflush(sock *s, int revents) { switch (s->type) { @@ -1914,7 +1918,15 @@ sk_read(sock *s, int revents) } int -sk_write(sock *s) +sk_read(sock *s, int revents) +{ + int e = sk_read_noflush(s, revents); + tmp_flush(); + return e; +} + +static inline int +sk_write_noflush(sock *s) { switch (s->type) { @@ -1962,6 +1974,14 @@ sk_write(sock *s) } } +int +sk_write(sock *s) +{ + int e = sk_write_noflush(s); + tmp_flush(); + return e; +} + int sk_is_ipv4(sock *s) { return s->af == AF_INET; } @@ -1980,6 +2000,7 @@ sk_err(sock *s, int revents) } s->err_hook(s, se); + tmp_flush(); } void diff --git a/sysdep/unix/main.c b/sysdep/unix/main.c index 7e8ea0dc..cdf0a310 100644 --- a/sysdep/unix/main.c +++ b/sysdep/unix/main.c @@ -682,7 +682,7 @@ signal_init(void) * Parsing of command-line arguments */ -static char *opt_list = "B:c:dD:ps:P:u:g:flRh"; +static char *opt_list = "bc:dD:ps:P:u:g:flRh"; int parse_and_exit; char *bird_name; static char *use_user; @@ -703,7 +703,6 @@ display_help(void) fprintf(stderr, "\n" "Options: \n" - " -B <block-size> Use 2^this number as memory allocation block size (default: 12)\n" " -c <config-file> Use given configuration file instead of\n" " " PATH_CONFIG_FILE "\n" " -d Enable debug messages and run bird in foreground\n" @@ -790,15 +789,12 @@ get_gid(const char *s) return gr->gr_gid; } -extern _Bool alloc_multipage; - static void parse_args(int argc, char **argv) { int config_changed = 0; int socket_changed = 0; int c; - int bp; bird_name = get_bird_name(argv[0], "bird"); if (argc == 2) @@ -811,29 +807,6 @@ parse_args(int argc, char **argv) while ((c = getopt(argc, argv, opt_list)) >= 0) switch (c) { - case 'B': - bp = atoi(optarg); - if (bp < 1) - { - fprintf(stderr, "Strange block size power %d\n\n", bp); - display_usage(); - exit(1); - } - - if ((1 << bp) < page_size) - { - fprintf(stderr, "Requested block size %ld is lesser than page size %ld\n\n", (1L<<bp), page_size); - display_usage(); - exit(1); - } - - if ((1L << bp) > page_size) - { - alloc_multipage = 1; - page_size = (1L << bp); - } - - break; case 'c': config_name = optarg; config_changed = 1; @@ -888,8 +861,6 @@ parse_args(int argc, char **argv) } } -void resource_sys_init(void); - /* * Hic Est main() */ @@ -902,7 +873,6 @@ main(int argc, char **argv) dmalloc_debug(0x2f03d00); #endif - resource_sys_init(); parse_args(argc, argv); log_switch(1, NULL, NULL); diff --git a/test/birdtest.c b/test/birdtest.c index c6a09684..ae05d1a5 100644 --- a/test/birdtest.c +++ b/test/birdtest.c @@ -20,6 +20,7 @@ #include "test/birdtest.h" #include "lib/string.h" +#include "lib/event.h" #ifdef HAVE_EXECINFO_H #include <execinfo.h> @@ -58,14 +59,11 @@ u64 bt_random_state[] = { 0x53d9772877c1b647, 0xab8ce3eb466de6c5, 0xad02844c8a8e865f, 0xe8cc78080295065d }; -void resource_sys_init(void); - void bt_init(int argc, char *argv[]) { int c; - resource_sys_init(); initstate(BT_RANDOM_SEED, (char *) bt_random_state, sizeof(bt_random_state)); bt_verbose = 0; @@ -122,6 +120,9 @@ bt_init(int argc, char *argv[]) clock_gettime(CLOCK_MONOTONIC, &bt_begin); bt_suite_case_begin = bt_suite_begin = bt_begin; + resource_init(); + ev_init_list(&global_event_list); + return; usage: @@ -175,6 +176,8 @@ int bt_run_test_fn(int (*fn)(const void *), const void *fn_arg, int timeout) if (!bt_suite_result) result = 0; + tmp_flush(); + return result; } @@ -312,6 +315,12 @@ bt_log_suite_case_result(int result, const char *fmt, ...) } } +void +bt_reset_suite_case_timer(void) +{ + clock_gettime(CLOCK_MONOTONIC, &bt_suite_case_begin); +} + int bt_test_suite_base(int (*fn)(const void *), const char *id, const void *fn_arg, int forked, int timeout, const char *dsc, ...) { @@ -504,6 +513,15 @@ bt_fmt_ipa(char *buf, size_t size, const void *data) bsnprintf(buf, size, "(null)"); } +void +bt_format_net(char *buf, size_t size, const void *data) +{ + if (data) + bsnprintf(buf, size, "%N", (const net_addr *) data); + else + bsnprintf(buf, size, "(null)"); +} + int bt_is_char(byte c) { diff --git a/test/birdtest.h b/test/birdtest.h index caec529b..ad5f8f9c 100644 --- a/test/birdtest.h +++ b/test/birdtest.h @@ -32,6 +32,7 @@ extern const char *bt_test_id; void bt_init(int argc, char *argv[]); int bt_exit_value(void); +void bt_reset_suite_case_timer(void); int bt_test_suite_base(int (*test_fn)(const void *), const char *test_id, const void *test_fn_argument, int forked, int timeout, const char *dsc, ...); static inline u64 bt_random(void) { return ((u64) random() & 0xffffffff) | ((u64) random() << 32); } @@ -165,6 +166,8 @@ struct bt_batch { void bt_fmt_str(char *buf, size_t size, const void *data); void bt_fmt_unsigned(char *buf, size_t size, const void *data); void bt_fmt_ipa(char *buf, size_t size, const void *data); +void bt_format_net(char *buf, size_t size, const void *data); + int bt_assert_batch__(struct bt_batch *opts); int bt_is_char(byte c); diff --git a/test/bt-utils.c b/test/bt-utils.c index cbca3a6b..2a7799c3 100644 --- a/test/bt-utils.c +++ b/test/bt-utils.c @@ -60,7 +60,6 @@ bt_bird_init(void) log_init_debug(""); log_switch(bt_verbose != 0, NULL, NULL); - resource_init(); olock_init(); timer_init(); io_init(); diff --git a/tools/gendist b/tools/gendist index 2ac59030..2dc42ba9 100755 --- a/tools/gendist +++ b/tools/gendist @@ -2,6 +2,7 @@ # # Generate BIRD Distribution Archive # (c) 2000--2004 Martin Mares <mj@ucw.cz> +# (c) 2005--2022 Ondrej Filip <feela@network.cz> # VERSION=`grep 'BIRD_VERSION \"' sysdep/config.h | sed '/BIRD_VERSION/!d;s/^.*"\(.*\)"$/\1/'` @@ -33,8 +34,6 @@ rm -rf `find $T/$REL -name CVS -o -name tmp` $T/$REL/{misc,rfc,doc/slides,doc/sl rm -rf $T/$REL $T/$DREL echo -n "OK? " read OK -echo Uploading to Atrey... -scp $T/$REL.tar.gz $T/$DREL.tar.gz atrey.karlin.mff.cuni.cz:~ftp/pub/bird/ echo Uploading to Trubka... scp $T/$REL.tar.gz $T/$DREL.tar.gz bird.network.cz:~ftp/pub/bird/ echo Done. diff --git a/tools/linuxdoc b/tools/linuxdoc index 51110e79..58f5cbc4 100755 --- a/tools/linuxdoc +++ b/tools/linuxdoc @@ -25,8 +25,13 @@ use FindBin; $prefix = "/usr"; $isoentities_prefix = "/usr"; $DataDir = "$FindBin::Bin/../doc/sbase"; -$AuxBinDir = "/usr/lib/linuxdoc-tools"; - +if (-d "/usr/lib/linuxdoc-tools") +{ + $AuxBinDir = "/usr/lib/linuxdoc-tools"; +} else +{ + $AuxBinDir = "/usr/bin"; +} use lib "$FindBin::Bin/linuxdoc-tools"; # --------------------------------------------------------------------- |