- frog - lists.frrouting.org

nhrpd event manager
by Joe Maimon 01 Mar '21

01 Mar '21

I have successfully gotten frr,nhrpd,strongswan,charon all running on another system for a dmvpn application. I am currently trying to do another setup, but without vpn/ipsec, just nhrp/gre. Its not working out well, debug logs show Netlink: Received msg_type 28 which if I recall correctly suggests that the neighbor entry isnt being installed correctly, trying it as root worked even less. So while poking around, I see there is an entire event manager unix sock structure. Perhaps I can parse that and do it myself? However, this is as far as I got. I am interacting with it via shell script. Mar 15 08:11:35 debian67 nhrpd[20977]: evmgr: sending event authorize-binding Mar 15 08:11:35 debian67 nhrpd[20977]: Send Registration-Reply(4) 192.168.241.129 -> 192.168.241.130 Mar 15 08:11:35 debian67 nhrpd[20977]: PACKET: Send 45.77.146.76 -> 209.51.164.27 Mar 15 08:11:35 debian67 nhrpd[20977]: evmgr: msg: eventid=214 Mar 15 08:11:35 debian67 nhrpd[20977]: evmgr: msg: result=accept Mar 15 08:11:35 debian67 nhrpd[20977]: evmgr: msg: Mar 15 08:11:35 debian67 nhrpd[20977]: evmgr: received: eventid=214 result= Mar 15 08:11:45 debian67 nhrpd[20977]: cache: t67 192.168.241.130: timeout Mar 15 08:11:46 debian67 nhrpd[20977]: Netlink: Received msg_type 28, msg_flags 0 I have looked at the source and I cant seem to convince the function to recognize that result should be "accept". Any advice appreciated. Joe For the interested, the below is the fledgling script #!/bin/bash PROGNAME=`basename $0` VERSION="0.0.1" usage() { echo "Usage: $PROGNAME -s nhrp-sock [-i interface-name] [-u user] [-g group] " echo "" echo "-s nhrp-sock file" echo "-i interface-name to execute on, may be repeated multiple times" echo "-u user to own the sock" echo "-g group to own the sock" exit 1 } SOCK="/var/run/frr/nhrp.sock" USER="frr" GROUP="frr" j=0 while getopts s:i:u:g: opt; do case "$opt" in s) SOCK="$OPTARG" ;; i) INTARR[((j++))]="$OPTARG" ;; u) USER="$OPTARG" ;; g) GROUP="$OPTARG" ;; esac; done coproc socat - UNIX-LISTEN:$SOCK,unlink-early,setuid-early=$USER || exit 1 OLDIFS="$IFS" while read -r S; do if [[ "$S" == "" ]]; then if [[ "$EVID" != "" ]]; then echo -e "eventid=$EVID\nresult=accept\n"; fi unset EVID unset EVINT continue; fi IFS="${IFS}=" SA=($S) IFS="$OLDIFS" case "${SA[0]}" in eventid) EVID="${SA[1]}" ;; interface) EVINT="${SA[1]}" esac done <&"${COPROC[0]}" >&"${COPROC[1]}" kill "$COPROC_PID"

2 3

route-map continue
by Joe Maimon 28 Feb '21

28 Feb '21

route-map contains the command continue Since it also contains the commands on-match goto and on-match next, I suppose that continue is there for syntax compatibility with IOS. IOS allows the continue command to be used with or without a next clause number, frr does not. Joe

2 2

Re: [FROG] How to test ipv6 anycast address with frr routing?
by Chriztoffer Hansen 27 Feb '21

27 Feb '21

On Sat, 27 Feb 2021 at 01:09, yan via frog <frog(a)lists.frrouting.org> wrote: > does frr routing support ipv6 anycast address configuration? Anycast routing in the essence of things is "just" announcing the same prefix from multiple locations. > How could I test ipv6 anycast address with frr routing? In the simplest of set-ups, configure multiple hosts to originate the same prefix (e.g. /32, /128) That would be anycast at the most basic level. (What services you will want to host from the anycast IP(s) is at your own descretion.) > Is there a tutorial about this? The official docs https://docs.frrouting.org does not contain such a tutorial. Thou I suspect more than one blogpost, conference talk has been given on the topic over the years (not being FRRouting specific). But most always this will be a generalized one. Skipping any in-depth (SW specific) tutorial content. And "just" explaining the concept. -- Chriztoffer

2 1

Re: [FROG] How to test ipv6 anycast address with frr routing?
by Chriztoffer Hansen 27 Feb '21

27 Feb '21

York, On Sat, 27 Feb 2021 at 01:09, yan via frog <frog(a)lists.frrouting.org> wrote: > I want to test ipv6 anycast address with frr routing with several linux PC host. However, I cannot find a way to setup ipv6 address on my linux host which is Ubuntu 12.04. I also cannot find a way to config ipv6 address on frr routing suite. > > So does linux support ipv6 anycast address? Or does frr routing support ipv6 anycast address configuration? How could I test ipv6 anycast address with frr routing? Is there a tutorial about this? Ubuntu 12.04 from 9y ago? Have you at least tried with a more recent and (still) supported LTS version of Ubuntu? Asking, since little to no support can be expected to be given/provided if using non-supported OS releases. Also, can you specify what FRR version you are using in your set-up? -- Chriztoffer

1 0

How to test ipv6 anycast address with frr routing?
by yanhc519＠aliyun.com 27 Feb '21

27 Feb '21

Hi, all. I want to test ipv6 anycast address with frr routing with several linux PC host. However, I cannot find a way to setup ipv6 address on my linux host which is Ubuntu 12.04. I also cannot find a way to config ipv6 address on frr routing suite. So does linux support ipv6 anycast address? Or does frr routing support ipv6 anycast address configuration? How could I test ipv6 anycast address with frr routing? Is there a tutorial about this? Thanks. York

1 0

No RIP packets, rejecting distribute-list and redistribute
by jimc 26 Feb '21

26 Feb '21

Version: frr-7.5-3.3.x86_64 from OpenSuSE Tubleweed. This is probably a near duplicate of bug 1279 (2017-10-02) https://github.com/FRRouting/frr/issues/1279 except the OP omidit was trying to use BGP while I have RIP[ng]. I'm pointing the finger of blame at a documentation issue; in other words, almost certainly my frr.conf file (appended with comments) lacks some prerequisite which would resurrect the redistribute and distribute-list commands. Presenting symptoms: ripd and ripngd never emit any RIP[ng] packets, even though the sub-router has a route to the subnet that I want the local LAN to be able to reach. Plus: when I specify "redistribute" and "distribute-list", frr -C complains: ``` line 56: % Unknown command[4]: distribute-list xenanet in line 57: % Unknown command[4]: distribute-list xenanet out line 61: % Unknown command[4]: redistribute kernel line 63: % Unknown command[4]: redistribute static line 65: % Unknown command[4]: redistribute connected ``` Back in 1987 I could just start rpc.ripd on SunOS-3 (no authentication in RIPv1), but 34 years later, Quagga looked like it gave a lot of people a lot of trouble to configure it, so I thought to reinvent the wheel. But for political and national security reasons I think it would be a real good idea to have open source routing used more widely, and so I decided to give Quagga my best try. It succumbed to the same issue complained about here, attributed to an unfixed bug by many OP's, and I dug through the source code but was not able to find the "bug". But I also hit references to FRR and tried it in the hope of a fix. No such luck. Currently I'm guessing that the problem is not in the code but in the documentation, so I've missed some important setup aspect. I'm planning to write up my experiences and of course it will include the conf file. In the appended instance I've retained comments, and if you spot any misconceptions among them I would appreciate corrections, so errors don't propagate. ``` # /etc/frr/frr.conf # This file contains plaintext passwords and needs permission 640 frr:frr # Revision history: # 2021-02-17 jimc Becomes FRR (Free Range Routing) unified config file. # 2021-02-04 jimc Initial setup for Quagga / rip[ng]d, abandoned. # hostname itsname # Default is the system's 1-component name per uname -n. # Without a password, vtysh will not connect; the enable password enables # interactive reconfiguration. But it uses PAM; password not needed for root. # Warning, security is not robust; turn this off in production. password redacted#1 enable password redacted#2 # Log file for all daemons log file /var/log/frr/frr.log # /etc/frr/daemons controls which protocol daemons to start, but you also have # to enable them. Not needed for staticd or zebra. router rip #rejected: version 2 # Do not receive v1 (which can't do authentication) router ripng # "vtysh -C [-f other_file.conf]" will check the configuration file; default # is /etc/frr/frr.conf. It would appear that if no protocol is enabled (e.g. # by "router rip") for which a particular command (e.g. "network") is relevant, # the command is reported as "Unknown command[4]: network 192.9.200.0/24", not # considering the parameters (e.g. with "router ripng", the IPv4 version is # not unknown). # Interfaces with an IP address in this range may send and receive protocol # packets. You can also enable named interfaces e.g. en0. network 192.9.200.0/24 network 2600:3c01:e000:306::/64 # We're only going to include these subnet(s) in incoming + outgoing packets. # In Quagga you could also use an access-list. ip prefix-list xenanet permit 192.9.200.176/29 ip prefix-list xenanet deny any # Syntax: 'route-map' its_name permit|deny line_number; matches etc. follow. # A number is a valid identifier and you confusingly see this in tutorials. route-map cnmap permit 1 match ip address prefix-list xenanet route-map cnmap deny 2 # Zebra can filter incoming routes per protocol or for any protocol. It just # ignores non-permitted routes. Looks like the default is to deny everything. ip protocol any route-map cnmap # In incoming and outgoing packets, only the specified subnets are included. distribute-list xenanet in distribute-list xenanet out # Daemons send routes that they learned from these sources. # kernel = set in system routing table by non-FRR, e.g. "ip route add..." redistribute kernel # static = configured in a FRR conf file. (CouchNet has none.) redistribute static # connected = seen in protocol packets from neighboring routers. redistribute connected # Turn on authentication (RIP v2 only). Mode md5 (hiss, boo!) makes a HMAC # over the packet with a pre-shared key prepended; the sender and receiver must # use the same key. It's OK to turn on auth for interfaces that don't exist. key chain CouchNet key 1 key-string redacted#3 interface en0 ip rip authentication mode md5 ip rip authentication key-chain CouchNet interface br0 ip rip authentication mode md5 ip rip authentication key-chain CouchNet interface rad0 ip rip authentication mode md5 ip rip authentication key-chain CouchNet interface tun0 ip rip authentication mode md5 ip rip authentication key-chain CouchNet interface tun1 ip rip authentication mode md5 ip rip authentication key-chain CouchNet interface tun9 ip rip authentication mode md5 ip rip authentication key-chain CouchNet # Default timer values are used. # update 30sec, send a gratuitous RIP packet this often. # timeout 180sec, a route absent from "learn from" sources (e.g. # neighbors) for this long is redistributed as expired. # garbage 120sec, after this long, expired routes are no longer announced. ``` -- James F. Carter Email: jimc(a)jfcarter.net Web: http://www.math.ucla.edu/~jimc (q.v. for PGP key)

2 1

[dev] Regarding ospfv3 authentication
by jagadeesh N 18 Feb '21

18 Feb '21

Hi, <<Reposting this question again, there was conversation happened regarding this 2=3 years back, but i still see the change log showing support for RFC 4552 https://lists.frrouting.org/pipermail/dev/2017-September/000899.html >> Does FRR 3.0 support OSPFv3 authentication? Though OSPFv3 doesn't have an authentication header on its own similar to OSPFv2, it should use IPSec for authentication. Though github claims that it supports RFC 4552. Reference Link:https://github.com/FRRouting/frr/wiki/FRR-2.0-%E2%86%92-FRR-3.0-Change… But I couldn't see commands like "ipv6 ospf authentication .." etc., Do I miss something?Regards, Jagadeesh N

1 0

RFC: resolution over default route - should it be enabled by default?
by Quentin Young 11 Feb '21

11 Feb '21

Hi froggers, FRR developers would like operator feedback on this question. By default, Zebra will not resolve nexthops via a default route. To allow this, you have to put the following line in your configuration: ip nht resolve-via-default (https://docs.frrouting.org/en/latest/zebra.html#clicmd-ipnhtresolve-via-def…) This has various implications, but the one we see most often is that BGP won't peer over a default route. This has been at the root of quite a few issues filed over the years, enough that it might be worthwhile flipping the default. We want to hear from operators whether you think this is a good idea. Has the current behavior helped you? Hurt you? We want your feedback. Quentin

1 0

readline 8.1 and bracketed pasting in vtysh
by Mark Kamichoff 09 Feb '21

09 Feb '21

tl;dr readline 8.1 turns on bracketed paste and appears to break multi-line pasting in vtysh Hi - I noticed that on a few of my boxes I can no longer paste a set of commands into vtysh using either FRR 6.x or 7.x. The result was that the lines were treated as a single line and therefore rejected. For example, if I want to create a simple route-map like so: route-map foo permit 5 set as-path prepend 1 2 3 4 end Here's what happened when I first pasted the 3x lines: ====snip==== excalibur(config)# route-map foo permit 5 set as-path prepend 1 2 3 4 end ====snip==== (the pasted text is highlighted in the terminal) After hitting enter vtysh rejects the command because it sees it as a single line (I think): ====snip==== % Unknown command: route-map foo permit 5 set as-path prepend 1 2 3 4 end excalibur(config)# ====snip==== The expected result should look something like this (I manually typed it in): ====snip==== excalibur(config)# route-map foo permit 5 excalibur(config-route-map)# set as-path prepend 1 2 3 4 excalibur(config-route-map)# end excalibur# ====snip==== I dug a little deeper and it looks like this is a result of bracketed paste[0] that has been enabled in the 8.1 release of GNU readline released in Dec of 2020[1]. I confirmed that an older system I have still running readline 8.0 does not have this issue. I think vtysh rejects the command because of the escape characters that are used with bracketed paste. I couldn't find any option to disable bracketed paste in readline itself or vtysh. Shells like Zsh and Bash have options for it in their config files. Does anyone know how to disable bracketed paste in vtysh? Or, is there something else funky going on here? I'm using rxvt-unicode on Linux/X11. I tested this with Terminal.app on Big Sur as well and got the same result. Thanks! - Mark [0] https://cirw.in/blog/bracketed-paste [1] https://lwn.net/Articles/839213/ (see h. in the CHANGES list) -- Mark Kamichoff prox(a)prolixium.com https://www.prolixium.com/

3 6

GRE support in FRR
by Amjad Ali 01 Feb '21

01 Feb '21

Hi, I am trying to test GRE and FRR and was looking for pointers on some use cases or documentation. I haven't found much in the documentation though. How much support is there for GRE and if someone could kindly provide a basic topo/test scenario to this effect? Thanks, Ali

2 1