bgpq3 - bgp filtering automation tool
bgpq3 [-h host[:port]] [-S sources] [-EPz] [-f asn | -F fmt | -G asn | -t] [-2346ABbDdHJjNnpsUX] [-a asn] [-r len] [-R len] [-m max] [-W len] OBJECTS [...] EXCEPT OBJECTS
The bgpq3 utility used to generate configuration (prefix-lists, extended access-lists, policy-statement terms and as-path lists) based on RADB data.
The options are as follows:
Allow routes registered for as23456 (transition-as) (default: false)
Assume that your device is asn32-capable.
Generate IPv4 prefix/access-lists (default).
Generate IPv6 prefix/access-lists (IPv4 by default).
Try to aggregate generated filters as much as possible (not all output formats supported).
Specify ASN that shall be denied in case of empty prefix-list (OpenBGPD).
Generate output in OpenBGPD format (default: Cisco).
Generate output in BIRD format (default: Cisco).
Enable some debugging output.
Use asdot notation for Cisco as-path access-lists.
Generate extended access-list (Cisco) or policy-statement term using route-filters (Juniper), [ip|ipv6]-prefix-list (Nokia) or prefix-filter (OpenBGPD)
Generate input as-path access-list for adjacent as AS number.
(Use -f 0 to not enforce first AS)
Generate output in user-defined format.
Generate output as-path access-list.
Hyperaggregation (supernets-only) mode.
Host running IRRD database (default: whois.radb.net).
Generate config for Juniper (default: Cisco).
Generate output in JSON format (default: Cisco).
Maximum length of accepted prefixes (default: 32 for IPv4, 128 for IPv6).
Extra match conditions for Juniper route-filters. See the examples section.
Generate config for Nokia SR OS (former Alcatel-Lucent) MD-CLI (default: Cisco)
Generate config for Nokia SR OS (former Alcatel-Lucent) classic CLI (default: Cisco)
Name of generated configuration stanza.
Limit recursion depth when expanding. This slows bgpq3 a bit, but sometimes
is a useful feature to prevent generated filters from growing too big.
Enable use of private ASNs and ASNs used for documentation purpose only (default: disabled).
Generate prefix-list (default behaviour, flag added for backward compatibility only).
Allow more-specific routes with masklen starting with specified length.
Allow more-specific routes up to specified masklen too. (Please, note: objects with prefix-length greater than specified length will be always allowed.)
Generate sequence numbers in IOS-style prefix-lists.
Use specified database sources only (example: RIPE,APNIC,RADB).
Generate as-sets for OpenBGPD (OpenBSD 6.4+), BIRD and JSON formats.
Disable pipelining. (not recommended)
Generate output in Huawei format (default: Cisco).
Generate as-path strings of a given length maximum (0 for infinity).
Generate config for Cisco IOS XR devices (plain IOS by default).
Generate Juniper route-filter-list (JunOS 16.2+).
OBJECTS means networks (in prefix format), autonomous systems, as-sets and
route-sets. If multiple objects are specified they will be merged.
You can exclude autonomous sets, as-sets and route-sets found during expansion from future expansion.
Generating named Juniper prefix-filter for AS20597:
user@host:~>bgpq3 -Jl eltel AS20597
policy-options {
replace:
prefix-list eltel {
81.9.0.0/20;
81.9.32.0/20;
81.9.96.0/20;
81.222.128.0/20;
81.222.192.0/18;
85.249.8.0/21;
85.249.224.0/19;
89.112.0.0/19;
89.112.4.0/22;
89.112.32.0/19;
89.112.64.0/19;
217.170.64.0/20;
217.170.80.0/20;
}
}
For Cisco we can use aggregation (-A) flag to make this prefix-filter more compact:
user@host:~>bgpq3 -Al eltel AS20597
no ip prefix-list eltel
ip prefix-list eltel permit 81.9.0.0/20
ip prefix-list eltel permit 81.9.32.0/20
ip prefix-list eltel permit 81.9.96.0/20
ip prefix-list eltel permit 81.222.128.0/20
ip prefix-list eltel permit 81.222.192.0/18
ip prefix-list eltel permit 85.249.8.0/21
ip prefix-list eltel permit 85.249.224.0/19
ip prefix-list eltel permit 89.112.0.0/18 ge 19 le 19
ip prefix-list eltel permit 89.112.4.0/22
ip prefix-list eltel permit 89.112.64.0/19
ip prefix-list eltel permit 217.170.64.0/19 ge 20 le 20
and, as you see, prefixes 89.112.0.0/19 and 89.112.32.0/19 now aggregated
into single entry
ip prefix-list eltel permit 89.112.0.0/18 ge 19 le 19.
Well, for Juniper we can generate even more interesting policy-statement,
using -M <extra match conditions>, -r <len>, -R <len> and hierarchical
names:
user@host:~>bgpq3 -AJEl eltel/specifics -r 29 -R 32 -M "community blackhole" AS20597
policy-options {
policy-statement eltel {
term specifics {
replace:
from {
community blackhole;
route-filter 81.9.0.0/20 prefix-length-range /29-/32;
route-filter 81.9.32.0/20 prefix-length-range /29-/32;
route-filter 81.9.96.0/20 prefix-length-range /29-/32;
route-filter 81.222.128.0/20 prefix-length-range /29-/32;
route-filter 81.222.192.0/18 prefix-length-range /29-/32;
route-filter 85.249.8.0/21 prefix-length-range /29-/32;
route-filter 85.249.224.0/19 prefix-length-range /29-/32;
route-filter 89.112.0.0/17 prefix-length-range /29-/32;
route-filter 217.170.64.0/19 prefix-length-range /29-/32;
}
}
}
}
generated policy-option term now allows more-specific routes in range /29 - /32 for eltel networks if they marked with community 'blackhole' (defined elsewhere in configuration).
Of course, bgpq3 supports IPv6 (-6):
user@host:~>bgpq3 -6l as-retn-6 AS-RETN6
no ipv6 prefix-list as-retn-6
ipv6 prefix-list as-retn-6 permit 2001:7fb:fe00::/48
ipv6 prefix-list as-retn-6 permit 2001:7fb:fe01::/48
[....]
and ASN32
user@host:~>bgpq3 -J3f 112 AS-SPACENET
policy-options {
replace:
as-path-group NN {
as-path a0 "^112(112)*$";
as-path a1 "^112(.)*(1898|5539|8495|8763|8878|12136|12931|15909)$";
as-path a2 "^112(.)*(21358|23456|23600|24151|25152|31529|34127|34906)$";
as-path a3 "^112(.)*(35052|41720|43628|44450|196611)$";
}
}
see AS196611 in the end of the list ? That's AS3.3 in 'asplain' notation.
If your router does not support ASN32 (yet) you should not use switch -3, and the result will be next:
user@host:~>bgpq3 -f 112 AS-SPACENET
no ip as-path access-list NN
ip as-path access-list NN permit ^112( 112)*$
ip as-path access-list NN permit ^112( [0-9]+)* (1898|5539|8495|8763)$
ip as-path access-list NN permit ^112( [0-9]+)* (8878|12136|12931|15909)$
ip as-path access-list NN permit ^112( [0-9]+)* (21358|23456|23600|24151)$
ip as-path access-list NN permit ^112( [0-9]+)* (25152|31529|34127|34906)$
ip as-path access-list NN permit ^112( [0-9]+)* (35052|41720|43628|44450)$
AS196611 is no more in the list, however, AS23456 (transition AS) would
have been added to list if it were not present.
If you want to generate configuration not for routers, but for some other programs/systems, you may use user-defined formatting, like in example below:
user@host:~>bgpq3 -F "ipfw add pass all from %n/%l to any\\n" as3254
ipfw add pass all from 62.244.0.0/18 to any
ipfw add pass all from 91.219.29.0/24 to any
ipfw add pass all from 91.219.30.0/24 to any
ipfw add pass all from 193.193.192.0/19 to any
Recognized format characters: '%n' - network, '%l' - mask length, '%N' - object name, '%m' - object mask and '%i' - inversed mask. Recognized escape characters: '\n' - new line, '\t' - tabulation. Please note that no new lines inserted automatically after each sentence, you have to add them into format string manually, elsewhere output will be in one line (sometimes it makes sense):
user@host:~>bgpq3 -6F "%n/%l; " as-eltel
2001:1b00::/32; 2620:4f:8000::/48; 2a04:bac0::/29; 2a05:3a80::/48;
By default bgpq3 trusts to data from all databases mirrored into RADB.
Unfortunately, not all these databases are equal in how much can we
trust their data. RIR maintained databases (AFRINIC,ARIN,APNIC and RIPE)
shall be trusted more than the others because they are indeed have the
knowledge about which address space allocated to this or that ASn,
other databases lack this knowledge and can (and, actually, do) contain
some stale data: noone but RIRs care to remove outdated route-objects
when address space revoked from one ASn and allocated to another.
In order to keep their filters both compact and actual, bgpq3 users
are encouraged to use '-S' flag to limit database sources to only
ones they trust.
General recommendations:
- use minimal set of RIR databases (only those in which you and your customers have registered route-objects). Keep RIR of your home region first in the list.
- use non-RIR databases only when operating in LACNIC region (for some reason LACNIC does not export their data to RADB) or when you are unable to make your customers maintain their data in RIR databases.
- avoid using RIPE-NONAUTH as trusted source: these records were created in RIPE database but for address space allocated to different RIR, so RIPE had no chance to check validity of this route.
Note on source ordering: order matters. When expanding as-sets,
IRRd sequentally checks all sources for matching object and stops
on first found entry. So, in case when as-set registered in multiple
databases (with different content), generated filters may differ
depending on source order:
snar@host:~>bgpq3 -S RIPE,RADB as-space
no ip prefix-list NN
ip prefix-list NN permit 195.190.32.0/19
snar@host:~>bgpq3 -S RADB,RIPE as-space
no ip prefix-list NN
ip prefix-list NN permit 45.4.4.0/22
ip prefix-list NN permit 45.4.132.0/22
ip prefix-list NN permit 45.6.128.0/22
ip prefix-list NN permit 45.65.184.0/22
[...]
Example: as we are operating mostly in Europe and APAC, we generate most of our filters with '-S RIPE,APNIC,RADB'.
First, definition: aggregation is a transformation of prefix set into compact representation that is exactly the same as initial set, i.e., no prefix that is not present in original set will not be allowed by aggregated set and no prefix that is present in original set will not be denied by aggregated set. Example: prefixes 10.0.0.0/24 10.0.1.0/24 10.0.0.0/22 can be aggregated in two route filters:
./bgpq3 -JEA 10.0.0.0/24 10.0.1.0/24 10.0.0.0/22
policy-options {
policy-statement NN {
replace:
from {
route-filter 10.0.0.0/22 exact;
route-filter 10.0.0.0/23 prefix-length-range /24-/24;
}
}
}
and this policy will allow specified prefixes only.
While aggregation is the most correct way to build filters used in routing policies, sometimes it makes sense to build more relaxed prefix-filters: for example, when generated filters will be used to traffic filtering it makes no sense to include second line as it is already covered by first one. So, in this case we can use 'hyperaggregation' (supernets-only) mode to include only supernets in output:
./bgpq3 -JH 10.0.0.0/24 10.0.1.0/24 10.0.0.0/22
policy-options {
replace:
prefix-list NN {
10.0.0.0/22;
}
}
from the perspective of traffic filtering it is correct and more effective than just aggregation.
When everything is OK, bgpq3 generates result to standard output and
exits with status == 0. In case of errors they are printed to stderr and
program exits with non-zero status.
To improve bgpq3 performance when expanding extra-large AS-SETs you
shall tune OS settings to enlarge TCP send buffer.
FreeBSD can be tuned in the following way:
sysctl -w net.inet.tcp.sendbuf_max=2097152
Linux can be tuned in the following way:
sysctl -w net.ipv4.tcp_window_scaling=1
sysctl -w net.core.rmem_max=2097152
sysctl -w net.core.wmem_max=2097152
sysctl -w net.ipv4.tcp_rmem="4096 87380 2097152"
sysctl -w net.ipv4.tcp_wmem="4096 65536 2097152"
Please note that generated prefix-lists may not fit your router's limitations. For example, JunOS supports only 85,325 prefixes in each prefix-list 4. According to field experience 5 IOS-XR prefix-sets are limited to 300001 entries.
- Routing Arbiter
- draft-michaelson-4byte-as-representation-05 for information on 'asdot' and 'asplain' notations.
- Cisco documentation for information on Cisco implementation of ASN32.
- JunOS prefix-lists limitation
- IOS-XR prefix-set limitation
Alexandre Snarskii [email protected]