NFD » ndn-cxx

is this on Ubuntu? what version?

Davide Pesavento wrote in #note-3:

is this on Ubuntu? what version?

I've been working on 18.04 LTS in a VirtualBox VM, I can try to reproduce on 20.04/20.10 if that would be helpful.

I cannot reproduce this issue with the given script on Ubuntu 20.04.
The experiment was conducted on Virtual Wall 1 testbed, pcgen02-1p instance.
Version and network information are pasted below.

mini-ndn> a head -2 log/nfd.log
NFD version 0.7.1-21-g7249fb4d starting
Built with GNU C++ version 9.3.0, with GNU libstdc++ version 20200808, with Boost version 1.71.0, with libpcap version 1.9.1 (with TPACKET_V3), with WebSocket++ version 0.8.1, with ndn-cxx version 0.7.1-22-gc25e463e

mini-ndn> a nfdc face list scheme udp6
faceid=259 remote=udp6://[ff02::1234%a-wlan0]:56363 local=udp6://[2001::1]:35670 congestion={base-marking-interval=100ms default-threshold=65536B} mtu=8800 counters={in={0i 0d 0n 0B} out={0i 0d 0n 0B}} flags={non-local permanent multi-access congestion-marking}

mini-ndn> a ip addr
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN group default qlen 1000
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
    inet 127.0.0.1/8 scope host lo
       valid_lft forever preferred_lft forever
    inet6 ::1/128 scope host
       valid_lft forever preferred_lft forever
26: a-wlan0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc htb state UP group default qlen 1000
    link/ether 02:00:00:00:0a:00 brd ff:ff:ff:ff:ff:ff
    inet 10.0.0.1/8 brd 10.255.255.255 scope global a-wlan0
       valid_lft forever preferred_lft forever
    inet6 2001::1/64 scope global
       valid_lft forever preferred_lft forever
    inet6 fe80::ff:fe00:a00/64 scope link
       valid_lft forever preferred_lft forever

mini-ndn> b nfdc face list scheme udp6
faceid=259 remote=udp6://[ff02::1234%b-eth0]:56363 local=udp6://[fe80::88ea:d1ff:fe54:acce%b-eth0]:45768 congestion={base-marking-interval=100ms default-threshold=65536B} mtu=8800 counters={in={0i 0d 0n 0B} out={0i 0d 0n 0B}} flags={non-local permanent multi-access congestion-marking}

mini-ndn> b ip addr
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN group default qlen 1000
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
    inet 127.0.0.1/8 scope host lo
       valid_lft forever preferred_lft forever
    inet6 ::1/128 scope host
       valid_lft forever preferred_lft forever
2: b-eth0@if28: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc htb state UP group default qlen 1000
    link/ether 8a:ea:d1:54:ac:ce brd ff:ff:ff:ff:ff:ff link-netnsid 0
    inet 10.0.0.2/8 brd 10.255.255.255 scope global b-eth0
       valid_lft forever preferred_lft forever
    inet6 fe80::88ea:d1ff:fe54:acce/64 scope link
       valid_lft forever preferred_lft forever

My installation steps were:

# enable IPv4 NAT to reach GitHub
wget -O - -nv --ciphers DEFAULT@SECLEVEL=1 https://www.wall2.ilabt.iminds.be/enable-nat.sh | sudo bash

# install NFD nightly packages
echo "deb [trusted=yes] https://nfd-nightly-apt.ndn.today/ubuntu focal main" | sudo tee /etc/apt/sources.list.d/nfd-nightly.list
sudo apt-get update
sudo apt-get install -qq infoedit libndn-cxx-dev ndnchunks ndnpeek ndnping nfd nlsr
sudo systemctl stop nfd
sudo systemctl stop ndnping

# set Python 3 as default
sudo update-alternatives --remove-all python || true
sudo update-alternatives --install /usr/bin/python python /usr/bin/python3 1

# install Mininet and Mini-NDN
git clone https://github.com/named-data/mini-ndn.git
cd mini-ndn
./install.sh -wmi

# experiment crashes without this file
sudo mkdir -p /etc/NetworkManager/conf.d/
sudo touch /etc/NetworkManager/conf.d/unmanaged.conf

# run the experiment script
wget https://redmine.named-data.net/attachments/download/946/example.py
sudo python example.py

Alex, can you please apply this patch https://gerrit.named-data.net/c/NFD/+/6400 and run the experiment again until it crashes, then attach the log file?

Alexander Lane wrote in #note-7:

I will try 20.04 and see if my local behavior is consistent with what Junxiao is experiencing.

No need. I have all the info I need. And I can easily reproduce the crash outside Mini-NDN.

TL;DR: it appears that the kernel doesn't like it if you try to bind a socket to a tentative IPv6 address.

Whenever an IPv6 address is added to an interface, it will first appear with the IFA_F_TENTATIVE flag for a short period of time (1-2 seconds), after which the flag is dropped and the address can be used. NFD reacts to NetworkMonitor's onAddressAdded signal to create new multicast faces and does not check the tentative flag on the new address. So, if NFD is quick enough, it may attempt to bind the MulticastUdpTransport's TX socket to the new address while it's still tentative, causing the system call to fail.

I suppose NetworkMonitor can be changed to ignore tentative addresses and only emit onAddressAdded when the address becomes non-tentative (i.e., valid). I don't see a use case for reporting tentative addresses to "normal" applications (including NFD) as they can't effectively be used for communication.

On top of that, the UdpFactory code can be made more robust against unexpected exceptions from the lower levels.