** Description changed:

  [impact]
  
  systemd-networkd double-free causes crash under some circumstances, such
  as adding/removing ip rules
  
  [test case]
  
- see original description
+ Use networkd-dispatcher events to add and remove IP rules. The example
+ scripts below are contrived (and by themselves likely to break access to
+ a machine) but would be adequate to trigger the bug. Put scripts like
+ these in place, reboot or run "netplan apply", and then leave the
+ machine running for a few DHCP renewal cycles.
+ 
+ === /etc/networkd-dispatcher/configured.d/test.sh ===
+ #!/bin/bash
+ 
+ /sbin/ip rule add iif lo lookup 99
+ /sbin/ip rule add to 10.0.0.0/8 iif lo lookup main
+ === END ===
+ === /etc/networkd-dispatcher/configuring.d/test.sh ===
+ #!/bin/bash
+ 
+ # Tear down existing ip rules so they aren't duplicated
+ OLDIFS="${IFS}"
+ IFS="
+ "
+ for rule in `ip rule show|grep "iif lo" | cut -d: -f2-`; do
+   IFS="${OLDIFS}"
+   ip rule delete ${rule}
+ done
+ IFS="${OLDIFS}"
+ === END ===
  
  [regression potential]
  
  this strdup's strings during addition of routing policy rules, so any
  regression would likely occur when adding/modifying/removing ip rules,
  possibly including networkd segfault or failure to add/remove/modify ip
  rules.
  
  [scope]
  
  this is needed for bionic.
  
  this is fixed by upstream commit
  eeab051b28ba6e1b4a56d369d4c6bf7cfa71947c which is included starting in
  v240, so this is already included in Focal and later.
  
  I did not research what original commit introduced the problem, but the
  reporter indicates this did not happen for Xenial so it's unlikely this
  is a problem in Xenial or earlier.
  
  [original description]
  
  This is a serious regression with systemd-networkd that I ran in to
  while setting up a NAT router in AWS. The AWS AMI ubuntu/images/hvm-ssd
  /ubuntu-bionic-18.04-amd64-server-20200131 with systemd-237-3ubuntu10.33
  does NOT have the problem, but the next most recent AWS AMI
  ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20200311 with
  systemd-including 237-3ubuntu10.39 does.
  
  Also, a system booted from the (good) 20200131 AMI starts showing the
  problem after updating only systemd (to 237-3ubuntu10.41) and its direct
  dependencies (e.g. 'apt-get install systemd'). So I'm fairly confident
  that a change to the systemd package between 237-3ubuntu10.33 and
  237-3ubuntu10.39 introduced the problem and it is still present.
  
  On the NAT router I use three interfaces and have separate routing
  tables for admin and forwarded traffic. Things come up fine initially
  but every 30-60 minutes (DHCP lease renewal time?) one or more
  interfaces is reconfigured and most of the time systemd-networkd will
  crash and need to be restarted. Eventually the system becomes
  unreachable when the default crash loop backoff logic prevents the
  network service from being restarted at all. The log excerpt attached
  illustrates the crash loop.
  
  Also including the netplan and networkd config files below.
  
  # grep . /etc/netplan/*
  /etc/netplan/50-cloud-init.yaml:# This file is generated from information 
provided by the datasource.  Changes
  /etc/netplan/50-cloud-init.yaml:# to it will not persist across an instance 
reboot.  To disable cloud-init's
  /etc/netplan/50-cloud-init.yaml:# network configuration capabilities, write a 
file
  /etc/netplan/50-cloud-init.yaml:# 
/etc/cloud/cloud.cfg.d/99-disable-network-config.cfg with the following:
  /etc/netplan/50-cloud-init.yaml:# network: {config: disabled}
  /etc/netplan/50-cloud-init.yaml:network:
  /etc/netplan/50-cloud-init.yaml:    version: 2
  /etc/netplan/50-cloud-init.yaml:    ethernets:
  /etc/netplan/50-cloud-init.yaml:        ens5:
  /etc/netplan/50-cloud-init.yaml:            dhcp4: true
  /etc/netplan/50-cloud-init.yaml:            match:
  /etc/netplan/50-cloud-init.yaml:                macaddress: xx:xx:xx:xx:xx:xx
  /etc/netplan/50-cloud-init.yaml:            set-name: ens5
  /etc/netplan/99_config.yaml:network:
  /etc/netplan/99_config.yaml:  version: 2
  /etc/netplan/99_config.yaml:  renderer: networkd
  /etc/netplan/99_config.yaml:  ethernets:
  /etc/netplan/99_config.yaml:    ens6:
  /etc/netplan/99_config.yaml:      match:
  /etc/netplan/99_config.yaml:        macaddress: yy:yy:yy:yy:yy:yy
  /etc/netplan/99_config.yaml:      dhcp4: true
  /etc/netplan/99_config.yaml:      dhcp4-overrides:
  /etc/netplan/99_config.yaml:        use-routes: false
  /etc/netplan/99_config.yaml:    ens7:
  /etc/netplan/99_config.yaml:      match:
  /etc/netplan/99_config.yaml:        macaddress: zz:zz:zz:zz:zz:zz
  /etc/netplan/99_config.yaml:      mtu: 1500
  /etc/netplan/99_config.yaml:      dhcp4: true
  /etc/netplan/99_config.yaml:      dhcp4-overrides:
  /etc/netplan/99_config.yaml:        use-mtu: false
  /etc/netplan/99_config.yaml:        use-routes: false
  
  # grep . /etc/networkd-dispatcher/*/*
  /etc/networkd-dispatcher/configured.d/nat:#!/bin/bash
  /etc/networkd-dispatcher/configured.d/nat:# Do additional configuration for 
the inside and outside interfaces
  /etc/networkd-dispatcher/configured.d/nat:# route table used for 
forwarded/routed/natted traffic
  /etc/networkd-dispatcher/configured.d/nat:FWD_TABLE=99
  /etc/networkd-dispatcher/configured.d/nat:if [ "${IFACE}" = "ens6" ]; then
  /etc/networkd-dispatcher/configured.d/nat:  # delete link-local route for 
inside in default table
  /etc/networkd-dispatcher/configured.d/nat:  /sbin/ip route delete 10.0.3.0/24 
2>/dev/null || true
  /etc/networkd-dispatcher/configured.d/nat:  # add link-local route for inside 
in table 99
  /etc/networkd-dispatcher/configured.d/nat:  /sbin/ip route replace 
10.0.3.0/24 dev ens6 scope link src 10.0.3.171 table ${FWD_TABLE}
  /etc/networkd-dispatcher/configured.d/nat:  # add routes to VPC cidrs via 
inside gateway in table 99
  /etc/networkd-dispatcher/configured.d/nat:  /sbin/ip route replace 
10.0.0.0/16 via 10.0.3.1 table ${FWD_TABLE}
  /etc/networkd-dispatcher/configured.d/nat:  # add rules to use table 99
  /etc/networkd-dispatcher/configured.d/nat:  /sbin/ip rule add iif  ens6 
lookup ${FWD_TABLE}
  /etc/networkd-dispatcher/configured.d/nat:  /sbin/ip rule add oif  ens6 
lookup ${FWD_TABLE}
  /etc/networkd-dispatcher/configured.d/nat:  /sbin/ip rule add from 
10.0.3.171/32 lookup ${FWD_TABLE}
  /etc/networkd-dispatcher/configured.d/nat:elif [ "${IFACE}" = "ens7" ]; then
  /etc/networkd-dispatcher/configured.d/nat:  # delete link-local route for 
outside in default table
  /etc/networkd-dispatcher/configured.d/nat:  /sbin/ip route delete 10.0.2.0/24 
2>/dev/null || true
  /etc/networkd-dispatcher/configured.d/nat:  # add link-local route for 
outside in table 99
  /etc/networkd-dispatcher/configured.d/nat:  /sbin/ip route replace 
10.0.2.0/24 dev ens7 scope link src 10.0.2.245 table ${FWD_TABLE}
  /etc/networkd-dispatcher/configured.d/nat:  # add default route via outside 
gateway in table 99
  /etc/networkd-dispatcher/configured.d/nat:  /sbin/ip route replace default 
via 10.0.2.1 table ${FWD_TABLE}
  /etc/networkd-dispatcher/configured.d/nat:  # add rules to use table 99
  /etc/networkd-dispatcher/configured.d/nat:  /sbin/ip rule add iif  ens7 
lookup ${FWD_TABLE}
  /etc/networkd-dispatcher/configured.d/nat:  /sbin/ip rule add oif  ens7 
lookup ${FWD_TABLE}
  /etc/networkd-dispatcher/configured.d/nat:  /sbin/ip rule add from 
10.0.2.245/32 lookup ${FWD_TABLE}
  /etc/networkd-dispatcher/configured.d/nat:  # add rules to use the inet route 
for local traffic but only if it's not destined for an RFC1918 private range
  /etc/networkd-dispatcher/configured.d/nat:  # IMPORTANT: order matters; the 
priority of rules is reverse of the order in which they are added.
  /etc/networkd-dispatcher/configured.d/nat:  # so the default/fallback is 
added first and then the local overrides.
  /etc/networkd-dispatcher/configured.d/nat:  #/sbin/ip rule add iif lo lookup 
${FWD_TABLE}
  /etc/networkd-dispatcher/configured.d/nat:  #ip rule add to 10.0.0.0/8 iif lo 
lookup main
  /etc/networkd-dispatcher/configured.d/nat:  #ip rule add to 172.16.0.0/12 iif 
lo lookup main
  /etc/networkd-dispatcher/configured.d/nat:  #ip rule add to 192.168.0.0/16 
iif lo lookup main
  /etc/networkd-dispatcher/configured.d/nat:  # ensure the forward policy is 
accept
  /etc/networkd-dispatcher/configured.d/nat:  iptables -P FORWARD ACCEPT
  /etc/networkd-dispatcher/configured.d/nat:  # configure iptables to do NAT
  /etc/networkd-dispatcher/configured.d/nat:  /sbin/iptables -t nat -I 
POSTROUTING 1 -o ens7 -j SNAT --to-source 10.0.2.245
  /etc/networkd-dispatcher/configured.d/nat:  # clean up any other rules
  /etc/networkd-dispatcher/configured.d/nat:  while /sbin/iptables -t nat -D 
POSTROUTING 2 2>/dev/null; do :; done
  /etc/networkd-dispatcher/configured.d/nat:fi
  /etc/networkd-dispatcher/configuring.d/nat:#!/bin/bash
  /etc/networkd-dispatcher/configuring.d/nat:# Tear down existing ip rules so 
they aren't duplicated
  /etc/networkd-dispatcher/configuring.d/nat:if [ "${IFACE}" = "ens6" ]; then
  /etc/networkd-dispatcher/configuring.d/nat:  # flush any existing rules 
referenceing this interface
  /etc/networkd-dispatcher/configuring.d/nat:  OLDIFS="${IFS}"
  /etc/networkd-dispatcher/configuring.d/nat:  IFS="
  /etc/networkd-dispatcher/configuring.d/nat:"
  /etc/networkd-dispatcher/configuring.d/nat:  for rule in `ip rule show|egrep 
"ens6|10.0.3.171" | cut -d: -f2-`; do
  /etc/networkd-dispatcher/configuring.d/nat:    IFS="${OLDIFS}"
  /etc/networkd-dispatcher/configuring.d/nat:    ip rule delete ${rule}
  /etc/networkd-dispatcher/configuring.d/nat:  done
  /etc/networkd-dispatcher/configuring.d/nat:  IFS="${OLDIFS}"
  /etc/networkd-dispatcher/configuring.d/nat:elif [ "${IFACE}" = "ens7" ]; then
  /etc/networkd-dispatcher/configuring.d/nat:  # flush any existing rules 
referencing this interface
  /etc/networkd-dispatcher/configuring.d/nat:  OLDIFS="${IFS}"
  /etc/networkd-dispatcher/configuring.d/nat:  IFS="
  /etc/networkd-dispatcher/configuring.d/nat:"
  /etc/networkd-dispatcher/configuring.d/nat:  for rule in `ip rule show|egrep 
"ens7|10.0.2.245|iif lo" | cut -d: -f2-`; do
  /etc/networkd-dispatcher/configuring.d/nat:    IFS="${OLDIFS}"
  /etc/networkd-dispatcher/configuring.d/nat:    ip rule delete ${rule}
  /etc/networkd-dispatcher/configuring.d/nat:  done
  /etc/networkd-dispatcher/configuring.d/nat:  IFS="${OLDIFS}"
  /etc/networkd-dispatcher/configuring.d/nat:fi
  
  ProblemType: Bug
  DistroRelease: Ubuntu 18.04
  Package: systemd 237-3ubuntu10.39
  ProcVersionSignature: Ubuntu 4.15.0-1060.62-aws 4.15.18
  Uname: Linux 4.15.0-1060-aws x86_64
  ApportVersion: 2.20.9-0ubuntu7.11
  Architecture: amd64
  Date: Wed Jun  3 21:24:28 2020
  Ec2AMI: ami-0238c6e72a7e906fc
  Ec2AMIManifest: (unknown)
  Ec2AvailabilityZone: us-east-1b
  Ec2InstanceType: c5n.large
  Ec2Kernel: unavailable
  Ec2Ramdisk: unavailable
  Lsusb: Error: command ['lsusb'] failed with exit code 1:
  MachineType: Amazon EC2 c5n.large
  ProcEnviron:
   TERM=xterm-256color
   PATH=(custom, no user)
   LANG=C.UTF-8
   SHELL=/bin/bash
  ProcKernelCmdLine: BOOT_IMAGE=/boot/vmlinuz-4.15.0-1060-aws 
root=LABEL=cloudimg-rootfs ro console=tty1 console=ttyS0 
nvme_core.io_timeout=4294967295
  SourcePackage: systemd
  UpgradeStatus: No upgrade log present (probably fresh install)
  dmi.bios.date: 10/16/2017
  dmi.bios.vendor: Amazon EC2
  dmi.bios.version: 1.0
  dmi.board.asset.tag: i-0c058310742990713
  dmi.board.vendor: Amazon EC2
  dmi.chassis.asset.tag: Amazon EC2
  dmi.chassis.type: 1
  dmi.chassis.vendor: Amazon EC2
  dmi.modalias: 
dmi:bvnAmazonEC2:bvr1.0:bd10/16/2017:svnAmazonEC2:pnc5n.large:pvr:rvnAmazonEC2:rn:rvr:cvnAmazonEC2:ct1:cvr:
  dmi.product.name: c5n.large
  dmi.sys.vendor: Amazon EC2

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1881972

Title:
  systemd-networkd crashes with invalid pointer

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/systemd/+bug/1881972/+subscriptions

-- 
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Reply via email to