slirp4netns - User-mode networking for unprivileged network namespaces
slirp4netns [OPTION]... PID|PATH TAPNAME
slirp4netns provides user-mode networking ("slirp") for network namespaces.
Unlike veth(4), slirp4netns does not require the root privileges on the host.
Default configuration:
MTU: 1500
CIDR: 10.0.2.0/24
Gateway/Host: 10.0.2.2 (network address + 2)
DNS: 10.0.2.3 (network address + 3)
IPv6 CIDR: fd00::/64
IPv6 Gateway/Host: fd00::2
IPv6 DNS: fd00::3
-c, --configure bring up the TAP interface. IP will be set to 10.0.2.100 (network address + 100) by default. IPv6 will be set to a random address. Starting with v0.4.0, the loopback interface (lo) is brought up as well.
-e, --exit-fd=FD specify the FD for terminating slirp4netns. When the FD is specified, slirp4netns exits when a poll(2) event happens on the FD.
-r, --ready-fd=FD specify the FD to write to when the initialization steps are finished. When the FD is specified, slirp4netns writes "1" to the FD and close the FD. Prior to v0.4.0, the FD was written after the network configuration (-c) but before the API socket configuration (-a).
-m, --mtu=MTU (since v0.2.0) specify MTU (max=65521).
-6, --enable-ipv6 (since v0.2.0, EXPERIMENTAL) enable IPv6
-a, --api-socket (since v0.3.0) API socket path
--cidr (since v0.3.0) specify CIDR, e.g. 10.0.2.0/24
--disable-host-loopback (since v0.3.0) prohibit connecting to 127.0.0.1:* on the host namespace
--netns-type=TYPE (since v0.4.0) specify network namespace type ([path|pid], default=pid)
--userns-path=PATH (since v0.4.0) specify user namespace path
--enable-sandbox (since v0.4.0) enter the user namespace and create a new mount namespace where only /etc and /run are mounted from the host.
Requires /etc/resolv.conf not to be a symlink to a file outside /etc and /run.
When running as the root, the process does not enter the user namespace but all the capabilities except CAP_NET_BIND_SERVICE are dropped.
--enable-seccomp (since v0.4.0, EXPERIMENTAL) enable seccomp(2) to limit syscalls. Typically used in conjunction with --enable-sandbox.
-h, --help (since v0.2.0) show help and exit
-v, --version (since v0.2.0) show version and exit
Terminal 1: Create user/network/mount namespaces
$ unshare --user --map-root-user --net --mount unshared$ echo $$ > /tmp/pid
Terminal 2: Start slirp4netns
$ slirp4netns --configure --mtu=65520 $(cat /tmp/pid) tap0 starting slirp, MTU=65520
Terminal 1: Make sure tap0 is configured and connected to the Internet
unshared$ ip a 1: lo: <LOOPBACK> mtu 65536 qdisc noop state DOWN group default qlen 1000 link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00 3: tap0: <BROADCAST,UP,LOWER_UP> mtu 65520 qdisc fq_codel state UNKNOWN group default qlen 1000 link/ether c2:28:0c:0e:29:06 brd ff:ff:ff:ff:ff:ff inet 10.0.2.100/24 brd 10.0.2.255 scope global tap0 valid_lft forever preferred_lft forever inet6 fe80::c028:cff:fe0e:2906/64 scope link valid_lft forever preferred_lft forever unshared$ echo "nameserver 10.0.2.3" > /tmp/resolv.conf unshared$ mount --bind /tmp/resolv.conf /etc/resolv.conf unshared$ curl https://example.com
Bind-mounting /etc/resolv.conf is only needed when /etc/resolv.conf on the host refers to loopback addresses (127.0.0.X, typically because of dnsmasq(8) or systemd-resolved.service(8)) that cannot be accessed from the namespace.
If your /etc/resolv.conf on the host is managed by networkmanager(8) or systemd-resolved.service(8), you might need to mount a new filesystem on /etc instead, so as to prevent the new /etc/resolv.conf from being unmounted unexpectedly when /etc/resolv.conf on the host is regenerated.
unshared$ mkdir /tmp/a /tmp/b unshared$ mount --rbind /etc /tmp/a unshared$ mount --rbind /tmp/b /etc unshared$ mkdir /etc/.ro unshared$ mount --move /tmp/a /etc/.ro unshared$ cd /etc unshared$ for f in .ro/*; do ln -s $f $(basename $f); done unshared$ rm resolv.conf unshared$ echo "nameserver 10.0.2.3" > /tmp/resolv.conf unshared$ curl https://example.com
To route ping packets, you need to set up net.ipv4.ping_group_range properly as the root.
e.g.
$ sudo sh -c "echo 0 2147483647 > /proc/sys/net/ipv4/ping_group_range"
By default, ports listening on INADDR_LOOPBACK (127.0.0.1) on the host are accessible from the child namespace via the gateway (default: 10.0.2.2). --disable-host-loopback can be used to prohibit connecting to INADDR_LOOPBACK on the host.
However, a host loopback address might be still accessible via the built-in DNS (default: 10.0.2.3) if /etc/resolv.conf on the host refers to a loopback address. You may want to set up iptables for limiting access to the built-in DNS in such a case.
unshared$ iptables -A OUTPUT -d 10.0.2.3 -p udp --dport 53 -j ACCEPT unshared$ iptables -A OUTPUT -d 10.0.2.3 -j DROP
slirp4netns can provide QMP-like API server over an UNIX socket file:
$ slirp4netns --api-socket /tmp/slirp4netns.sock ...
add_hostfwd: Expose a port (IPv4 only)
$ json='{"execute": "add_hostfwd", "arguments": {"proto": "tcp", "host_addr": "0.0.0.0", "host_port": 8080, "guest_addr": "10.0.2.100", "guest_port": 80}}' $ echo -n $json | nc -U /tmp/slirp4netns.sock { "return": {"id": 42}}
If host_addr is not specified, then it defaults to "0.0.0.0".
If guest_addr is not specified, then it will be set to the default address that corresponds to --configure.
list_hostfwd: List exposed ports
$ json='{"execute": "list_hostfwd"}' $ echo -n $json | nc -U /tmp/slirp4netns.sock { "return": {"entries": [{"id": 42, "proto": "tcp", "host_addr": "0.0.0.0", "host_port": 8080, "guest_addr": "10.0.2.100", "guest_port": 80}]}}
remove_hostfwd: Remove an exposed port
$ json='{"execute": "remove_hostfwd", "arguments": {"id": 42}}' $ echo -n $json | nc -U /tmp/slirp4netns.sock { "return": {}}
Remarks:
Client needs to shutdown(2) the socket with SHUT_WR after sending every request. i.e. No support for keep-alive and timeout.
slirp4netns "stops the world" during processing API requests.
A request must be less than 4096 bytes.
JSON responses may contain error instead of return.
A user can define a network namespace path as opposed to the default process ID:
$ slirp4netns --netns-type=path ... /path/to/netns tap0
Currently, the netns-type=TYPE argument supports path or pid args with the default being pid.
Additionally, a --userns-path=PATH argument can be included to override any user namespace path defaults
$ slirp4netns --netns-type=path --userns-path=/path/to/userns /path/to/netns tap0
Kernel 4.20 bumped up the default value of /proc/sys/net/ipv4/tcp_rmem from 87380 to 131072. This is known to slow down slirp4netns port forwarding: https://github.com/rootless-containers/slirp4netns/issues/128.
As a workaround, you can adjust the value of /proc/sys/net/ipv4/tcp_rmem inside the namespace. No real root privilege is needed to modify the file since kernel 4.15.
unshared$ c=$(cat /proc/sys/net/ipv4/tcp_rmem); echo $c | sed -e s/131072/87380/g > /proc/sys/net/ipv4/tcp_rmem
network_namespaces(7), user_namespaces(7), veth(4)
The slirp4netns command is available from https://github.com/rootless-containers/slirp4netns under GNU GENERAL PUBLIC LICENSE Version 2 (or later).