passt(1) General Commands Manual passt(1)
NAME
passt - Unprivileged user-mode network connectivity for virtual machines
pasta - Unprivileged user-mode network connectivity for network name-
spaces
SYNOPSIS
passt [OPTION]...
pasta [OPTION]... [COMMAND [ARG]...]
pasta [OPTION]... PID
pasta [OPTION]... --netns [PATH|NAME]
DESCRIPTION
passt
passt (Plug A Simple Socket Transport) provides full, quasi-native net-
work connectivity to virtual machines in user-mode without requiring any
capabilities or privileges.
The data plane implements a translation layer between a Layer-2 virtual
network interface and native Layer-4 (TCP, UDP, ping) sockets on the
host, giving the illusion that application processes residing on the
guest are running on the local host, from a networking perspective.
Built-in ARP, DHCP, NDP, and DHCPv6 implementations are designed to pro-
vide the guest with a network configuration that tightly resembles the
host native configuration. With the default options, guest and host
share IP addresses, routes, and port bindings.
Port forwarding and translation allow networking services running in the
guest to be reachable from both local and remote hosts.
Unlike slirp4netns(1), passt doesn't implement a full TCP stack: the TCP
translation layer has no stateful data buffering and operates by re-
flecting one peer's observed parameters (congestion window size, ac-
knowledged data, etc.) to the corresponding peer.
Currently, the only supported hypervisor is qemu(1), connecting to passt
by means of a UNIX domain socket. This is supported starting from qemu
7.2. For older qemu versions, see the qrap(1) wrapper.
pasta
pasta (Pack A Subtle Tap Abstraction) provides equivalent functionality
to network namespaces, as the one offered by passt for virtual machines.
If PID or --netns are given, pasta associates to an existing user and
network namespace. Otherwise, pasta creates a new user and network name-
space, and spawns the given command or a default shell within this con-
text. A tap device within the network namespace is created to provide
network connectivity.
For local TCP and UDP traffic only, pasta also implements a bypass path
directly mapping Layer-4 sockets between init and target namespaces, for
performance reasons.
OPTIONS
Unless otherwise noted below, if conflicting or multiple options are
given, the last one takes effect.
-d, --debug
Be verbose, don't log to the system logger.
--trace
Be extra verbose, show single packets. Implies --debug.
-q, --quiet
Don't print informational messages.
-f, --foreground
Don't run in background. This implies that the process is not
moved to a detached PID namespace after starting, because the PID
itself cannot change. Default is to fork into background.
-e, --stderr (DEPRECATED)
This option has no effect, and is maintained for compatibility
purposes only.
Note that this configuration option is deprecated and will be re-
moved in a future version.
-l, --log-file PATH
Log to file PATH, and not to the system logger.
Specifying this option multiple times does not lead to multiple
log files: the last given option takes effect.
--log-size SIZE
Limit log file size to SIZE bytes. When the log file is full,
make room for new entries by removing old ones at the beginning.
This limit is mandatory. Default is 1048576 (1 MiB).
--runas UID|UID:GID|LOGIN|LOGIN:GROUP
Attempt to change to given UID and corresponding group if UID is
given, or to given UID and given GID if both are given. Alterna-
tively, login name, or login name and group name can be passed.
This requires privileges (either initial effective UID 0 or
CAP_SETUID capability) to work. Default is to change to user no-
body if started as root.
-h, --help
Display a help message and exit.
--version
Show version and exit.
-p, --pcap file
Capture tap-facing (that is, guest-side or namespace-side) net-
work packets to file in pcap format.
Specifying this option multiple times does not lead to multiple
capture files: the last given option takes effect.
-P, --pid file
Write own PID to file once initialisation is done, before forking
to background (if configured to do so).
-m, --mtu mtu
Assign mtu via DHCP (option 26) and NDP (option type 5). A zero
value disables assignment. By default, the advertised MTU is
65520 bytes, that is, the maximum 802.3 MTU minus the length of a
802.3 header, rounded to 32 bits (IPv4 words).
-a, --address addr
Assign IPv4 addr via DHCP (yiaddr), or addr via DHCPv6 (option 5)
and an addr-based prefix via NDP Router Advertisement (option
type 3) for an IPv6 addr. This option can be specified zero (for
defaults) to two times (once for IPv4, once for IPv6). By de-
fault, assigned IPv4 and IPv6 addresses are taken from the host
interfaces with the first default route, if any, for the corre-
sponding IP version. If no default routes are available and there
is any interface with any route for a given IP version, the first
of these interfaces will be chosen instead. If no such interface
exists, the link-local address 169.254.2.1 is assigned for IPv4,
and no additional address will be assigned for IPv6.
-n, --netmask mask
Assign IPv4 netmask mask, expressed as dot-decimal or number of
bits, via DHCP (option 1). By default, the netmask associated to
the host address matching the assigned one is used. If there's no
matching address on the host, the netmask is determined according
to the CIDR block of the assigned address (RFC 4632).
-M, --mac-addr addr
Use source MAC address addr when communicating to the guest or to
the target namespace. Default is the locally administered MAC
addresses 9a:55:9a:55:9a:55.
-g, --gateway addr
Assign IPv4 addr as default gateway via DHCP (option 3), or IPv6
addr as source for NDP Router Advertisement and DHCPv6 messages.
This option can be specified zero (for defaults) to two times
(once for IPv4, once for IPv6). By default, IPv4 and IPv6 gate-
ways are taken from the host interface with the first default
route, if any, for the corresponding IP version. If the default
route is a multipath one, the gateway is the first nexthop router
returned by the kernel which has the highest weight in the set of
paths. If no default routes are available and there is just one
interface with any route, that interface will be chosen instead.
If no such interface exists, the link-local address 169.254.2.2
is used for IPv4, and the link-local address fe80::1 is used for
IPv6.
Note: these addresses are also used as source address for packets
directed to the guest or to the target namespace having a loop-
back or local source address, to allow mapping of local traffic
to guest and target namespace. See the NOTES below for more de-
tails about this mechanism.
-i, --interface name
Use host interface name to derive addresses and routes. Default
is to use the interfaces specified by --outbound-if4 and --out-
bound-if6, for IPv4 and IPv6 addresses and routes, respectively.
If no interfaces are given, the interface with the first default
routes for each IP version is selected. If no default routes are
available and there is just one interface with any route, that
interface will be chosen instead. If no such interface exists,
host interfaces will be ignored for the purposes of assigning ad-
dresses and routes, and link-local addresses will be used in-
stead.
-o, --outbound addr
Use an IPv4 addr as source address for IPv4 outbound TCP connec-
tions, UDP flows, ICMP requests, or an IPv6 addr for IPv6 ones,
by binding outbound sockets to it. This option can be specified
zero (for defaults) to two times (once for IPv4, once for IPv6).
By default, the source address is selected by the routing tables.
--outbound-if4 name
Bind IPv4 outbound sockets to host interface name, and, unless
another interface is specified via -i, --interface, use this in-
terface to derive IPv4 addresses and routes.
By default, the interface given by the default route is selected.
If no default routes are available and there is just one inter-
face with any route, that interface will be chosen instead. If no
such interface exists, outbound sockets will not be bound to any
specific interface.
--outbound-if6 name
Bind IPv6 outbound sockets to host interface name, and, unless
another interface is specified via -i, --interface, use this in-
terface to derive IPv6 addresses and routes.
By default, the interface given by the default route is selected.
If no default routes are available and there is just one inter-
face with any route, that interface will be chosen instead. If no
such interface exists, outbound sockets will not be bound to any
specific interface.
-D, --dns addr
Instruct the guest (via DHCP, DHVPv6 or NDP) to use addr (IPv4 or
IPv6) as a nameserver, as configured (see options --no-dhcp-dns,
--dhcp-dns) instead of reading addresses from /etc/resolv.conf.
This option can be specified multiple times. Specifying -D none
disables usage of DNS addresses altogether. Unlike addresses
from /etc/resolv.conf, addr is given to the guest without remap-
ping. For example --dns 127.0.0.1 will instruct the guest to use
itself as nameserver, not the host.
--dns-forward addr
Map addr (IPv4 or IPv6) as seen from guest or namespace to the
nameserver (with corresponding IP version) specified by the
--dns-host option. Maps only UDP and TCP traffic to port 53 or
port 853. Replies are translated back with a reverse mapping.
This option can be specified zero to two times (once for IPv4,
once for IPv6).
--dns-host addr
Configure the host nameserver which guest or namespace queries to
the --dns-forward address will be redirected to. This option can
be specified zero to two times (once for IPv4, once for IPv6).
By default, the first nameserver from the host's /etc/re-
solv.conf.
-S, --search list
Use space-separated list for DHCP, DHCPv6, and NDP purposes, in-
stead of reading entries from /etc/resolv.conf. See options --no-
dhcp-search and --dhcp-search. --search none disables the DNS do-
main search list altogether (if you need to search a domain
called "none" you can use --search none.).
--no-dhcp-dns
In passt mode, do not assign IPv4 addresses via DHCP (option 23)
or IPv6 addresses via NDP Router Advertisement (option type 25)
and DHCPv6 (option 23) as DNS resolvers. By default, all the
configured addresses are passed.
--dhcp-dns
In pasta mode, assign IPv4 addresses via DHCP (option 23) or IPv6
addresses via NDP Router Advertisement (option type 25) and
DHCPv6 (option 23) as DNS resolvers. By default, configured ad-
dresses, if any, are not passed.
--no-dhcp-search
In passt mode, do not send the DNS domain search list addresses
via DHCP (option 119), via NDP Router Advertisement (option type
31) and DHCPv6 (option 24). By default, the DNS domain search
list resulting from configuration is passed.
--dhcp-search
In pasta mode, send the DNS domain search list addresses via DHCP
(option 119), via NDP Router Advertisement (option type 31) and
DHCPv6 (option 24). By default, the DNS domain search list re-
sulting from configuration is not passed.
--no-tcp
Disable the TCP protocol handler. No TCP connections will be ac-
cepted host-side, and TCP packets coming from guest or target
namespace will be silently dropped.
--no-udp
Disable the UDP protocol handler. No UDP traffic coming from the
host side will be forwarded, and UDP packets coming from guest or
target namespace will be silently dropped.
--no-icmp
Disable the ICMP/ICMPv6 echo handler. ICMP and ICMPv6 echo re-
quests coming from guest or target namespace will be silently
dropped.
--no-dhcp
Disable the DHCP server. DHCP client requests coming from guest
or target namespace will be silently dropped. Implied if there is
no gateway on the selected IPv4 default route.
--no-ndp
Disable NDP responses. NDP messages coming from guest or target
namespace will be ignored.
--no-dhcpv6
Disable the DHCPv6 server. DHCPv6 client requests coming from
guest or target namespace will be silently dropped.
--no-ra
Disable Router Advertisements. Router Solicitations coming from
guest or target namespace will be ignored.
--freebind
Allow any binding address to be specified for -t and -u options.
Usually binding addresses must be addresses currently configured
on the host. With --freebind, the IP_FREEBIND or IPV6_FREEBIND
socket option is enabled allowing any address to be used. This
is typically used to bind addresses which might be configured on
the host in future, at which point the forwarding will immedi-
ately start operating.
--map-host-loopback addr
Translate addr to refer to the host. Packets from the guest to
addr will be redirected to the host. On the host such packets
will appear to have both source and destination of 127.0.0.1 or
::1.
If addr is 'none', no address is mapped (this implies --no-map-
gw). Only one IPv4 and one IPv6 address can be translated, if
the option is specified multiple times, the last one takes ef-
fect.
Default is to translate the guest's default gateway address, un-
less --no-map-gw is given, in which case no address is mapped.
--no-map-gw
Don't remap TCP connections and untracked UDP traffic, with the
gateway address as destination, to the host. Implied if there is
no gateway on the selected default route, or if there is no de-
fault route, for any of the enabled address families.
--map-guest-addr addr
Translate addr in the guest to be equal to the guest's assigned
address on the host. That is, packets from the guest to addr
will be redirected to the address assigned to the guest with -a,
or by default the host's global address. This allows the guest
to access services available on the host's global address, even
though its own address shadows that of the host.
If addr is 'none', no address is mapped. Only one IPv4 and one
IPv6 address can be translated, and if the option is specified
multiple times, the last one for each address type takes effect.
By default, mapping happens as described for the --map-host-loop-
back option.
-4, --ipv4-only
Enable IPv4-only operation. IPv6 traffic will be ignored. By de-
fault, IPv6 operation is enabled as long as at least an IPv6
route and an interface address are configured on a given host in-
terface.
-6, --ipv6-only
Enable IPv6-only operation. IPv4 traffic will be ignored. By de-
fault, IPv4 operation is enabled as long as at least an IPv4
route and an interface address are configured on a given host in-
terface.
-H, --hostname name
Hostname to configure the client with. Send name as DHCP option
12 (hostname).
--fqdn name
FQDN to configure the client with. Send name as Client FQDN:
DHCP option 81 and DHCPv6 option 39.
passt-only options
-s, --socket-path, --socket path
Path for UNIX domain socket used by qemu(1) or qrap(1) to connect
to passt. Default is to probe a free socket, not accepting con-
nections, starting from /tmp/passt_1.socket to
/tmp/passt_64.socket.
--vhost-user
Enable vhost-user. The vhost-user command socket is provided by
--socket.
--print-capabilities
Print back-end capabilities in JSON format, only meaningful for
vhost-user mode.
--repair-path path
Path for UNIX domain socket used by the passt-repair(1) helper to
connect to passt in order to set or clear the TCP_REPAIR option
on sockets, during migration. --repair-path none disables this
interface (if you need to specify a socket path called "none" you
can prefix the path by ./).
Default, for --vhost-user mode only, is to append .repair to the
path chosen for the hypervisor UNIX domain socket. No socket is
created if not in --vhost-user mode.
-F, --fd FD
Pass a pre-opened, connected socket to passt. Usually the socket
is opened in the parent process and passt inherits it when run as
a child. This allows the parent process to open sockets using an-
other address family or requiring special privileges.
This option implies the behaviour described for --one-off, once
this socket is closed.
-1, --one-off
Quit after handling a single client connection, that is, once the
client closes the socket, or once we get a socket error.
-t, --tcp-ports spec
Configure TCP port forwarding to guest. spec can be one of:
none Don't forward any ports
all Forward all unbound, non-ephemeral ports, as permitted by
current capabilities. For low (< 1024) ports, see NOTES.
No failures are reported for unavailable ports, unless no
ports could be forwarded at all.
ports A comma-separated list of ports, optionally ranged with -,
and, optionally, with target ports after :, if they dif-
fer. Specific addresses can be bound as well, separated by
/, and also, since Linux 5.7, limited to specific inter-
faces, prefixed by %. Within given ranges, selected ports
and ranges can be excluded by an additional specification
prefixed by ~.
Specifying excluded ranges only implies that all other
ports are forwarded. In this case, no failures are re-
ported for unavailable ports, unless no ports could be
forwarded at all.
Examples:
-t 22 Forward local port 22 to port 22 on the guest
-t 22:23
Forward local port 22 to port 23 on the guest
-t 22,25
Forward local ports 22 and 25 to ports 22 and 25 on
the guest
-t 22-80
Forward local ports between 22 and 80 to corre-
sponding ports on the guest
-t 22-80:32-90
Forward local ports between 22 and 80 to ports be-
tween 32 and 90 on the guest
-t 192.0.2.1/22
Forward local port 22, bound to 192.0.2.1, to port
22 on the guest
-t 192.0.2.1%eth0/22
Forward local port 22, bound to 192.0.2.1 and in-
terface eth0, to port 22
-t %eth0/22
Forward local port 22, bound to any address on in-
terface eth0, to port 22
-t 2000-5000,~3000-3010
Forward local ports between 2000 and 5000, except
for those between 3000 and 3010
-t 192.0.2.1/20-30,~25
For the local address 192.0.2.1, forward ports be-
tween 20 and 24 and between 26 and 30
-t ~20000-20010
Forward all ports to the guest, except for the
range from 20000 to 20010
Default is none.
-u, --udp-ports spec
Configure UDP port forwarding to guest. spec is as described for
TCP above.
Note: unless overridden, UDP ports with numbers corresponding to
forwarded TCP port numbers are forwarded too, without, however,
any port translation. IPv6 bound ports are also forwarded for
IPv4.
Default is none.
pasta-only options
-I, --ns-ifname name
Name of tap interface to be created in target namespace. By de-
fault, the same interface name as the external, routable inter-
face is used. If no such interface exists, the name tap0 will be
used instead.
-t, --tcp-ports spec
Configure TCP port forwarding to namespace. spec can be one of:
none Don't forward any ports
auto Dynamically forward ports bound in the namespace. The list
of ports is periodically derived (every second) from lis-
tening sockets reported by /proc/net/tcp and
/proc/net/tcp6, see proc(5).
ports A comma-separated list of ports, optionally ranged with -,
and, optionally, with target ports after :, if they dif-
fer. Specific addresses can be bound as well, separated by
/, and also, since Linux 5.7, limited to specific inter-
faces, prefixed by %. Within given ranges, selected ports
and ranges can be excluded by an additional specification
prefixed by ~.
Specifying excluded ranges only implies that all other
ports are forwarded. In this case, no failures are re-
ported for unavailable ports, unless no ports could be
forwarded at all.
Examples:
-t 22 Forward local port 22 to 22 in the target namespace
-t 22:23
Forward local port 22 to port 23 in the target
namespace
-t 22,25
Forward local ports 22 and 25 to ports 22 and 25 in
the target namespace
-t 22-80
Forward local ports between 22 and 80 to corre-
sponding ports in the target namespace
-t 22-80:32-90
Forward local ports between 22 and 80 to ports be-
tween 32 and 90 in the target namespace
-t 192.0.2.1/22
Forward local port 22, bound to 192.0.2.1, to port
22 in the target namespace
-t 192.0.2.1%eth0/22
Forward local port 22, bound to 192.0.2.1 and in-
terface eth0, to port 22
-t %eth0/22
Forward local port 22, bound to any address on in-
terface eth0, to port 22
-t 2000-5000,~3000-3010
Forward local ports between 2000 and 5000, except
for those between 3000 and 3010
-t 192.0.2.1/20-30,~25
For the local address 192.0.2.1, forward ports be-
tween 20 and 24 and between 26 and 30
-t ~20000-20010
Forward all ports to the namespace, except for
those between 20000 and 20010
IPv6 bound ports are also forwarded for IPv4.
Default is auto.
-u, --udp-ports spec
Configure UDP port forwarding to namespace. spec is as described
for TCP above, and the list of ports is derived from listening
sockets reported by /proc/net/udp and /proc/net/udp6, see
proc(5).
Note: unless overridden, UDP ports with numbers corresponding to
forwarded TCP port numbers are forwarded too, without, however,
any port translation.
IPv6 bound ports are also forwarded for IPv4.
Default is auto.
-T, --tcp-ns spec
Configure TCP port forwarding from target namespace to init name-
space. spec is as described above for TCP.
Default is auto.
-U, --udp-ns spec
Configure UDP port forwarding from target namespace to init name-
space. spec is as described above for UDP.
Default is auto.
--host-lo-to-ns-lo
If specified, connections forwarded with -t and -u from the
host's loopback address will appear on the loopback address in
the guest as well. Without this option such forwarded packets
will appear to come from the guest's public address.
--userns spec
Target user namespace to join, as a path. If PID is given, with-
out this option, the user namespace will be the one of the corre-
sponding process.
--netns spec
Target network namespace to join, as a path or a name. A name is
treated as with ip-netns(8) as equivalent to a path in
/run/netns.
This option can't be specified with a PID.
--netns-only
Join only a target network namespace, not a user namespace, and
don't create one for sandboxing purposes either. This is implied
if PATH or NAME are given without --userns.
--no-netns-quit
Do not exit once the target namespace reference is removed.
Without this option, pasta will terminate if the target network
namespace is bound to the filesystem, and the given path is
deleted, or if the target network namespace is represented by a
procfs entry, and that entry is deleted, representing the fact
that a process with the given PID terminated.
--config-net
Configure networking in the namespace: set up addresses and
routes as configured or sourced from the host, and bring up the
tap interface.
--no-copy-routes (DEPRECATED)
With --config-net, do not copy all the routes associated to the
interface we derive addresses and routes from: set up only the
default gateway. Implied by -g, --gateway, for the corresponding
IP version only.
Default is to copy all the routing entries from the interface in
the outer namespace to the target namespace, translating the out-
put interface attribute to the outbound interface in the name-
space.
Note that this configuration option is deprecated and will be re-
moved in a future version. It is not expected to be of any use,
and it simply reflects a legacy behaviour. If you have any use
for this, refer to REPORTING BUGS below.
--no-copy-addrs (DEPRECATED)
With --config-net, do not copy all the addresses associated to
the interface we derive addresses and routes from: set up a sin-
gle one. Implied by -a, --address, for the corresponding IP ver-
sion only.
Default is to copy all the addresses, except for link-local ones,
from the interface from the outer namespace to the target name-
space.
Note that this configuration option is deprecated and will be re-
moved in a future version. It is not expected to be of any use,
and it simply reflects a legacy behaviour. If you have any use
for this, refer to REPORTING BUGS below.
--ns-mac-addr addr
Configure MAC address addr on the tap interface in the namespace.
Default is to let the tap driver build a pseudorandom hardware
address.
--no-splice
Disable the bypass path for inbound, local traffic. See the sec-
tion Handling of local traffic in pasta in the NOTES for more de-
tails.
EXAMPLES
pasta
Create and use a new, connected, user and network namespace
$ iperf3 -s -D
$ ./pasta
Outbound interface: eth0, namespace interface: eth0
ARP:
address: 28:16:ad:39:a9:ea
DHCP:
assign: 192.168.1.118
mask: 255.255.255.0
router: 192.168.1.1
NDP/DHCPv6:
assign: 2a02:6d40:3ca5:2001:b81d:fa4a:8cdd:cf17
router: fe80::62e3:27ff:fe33:2b01
#
# dhclient -4 --no-pid
# dhclient -6 --no-pid
# ip address show
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: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 65520 qdisc pfifo_fast state UNKNOWN group default qlen 1000
link/ether 5e:90:02:eb:b0:2a brd ff:ff:ff:ff:ff:ff
inet 192.168.1.118/24 brd 192.168.1.255 scope global eth0
valid_lft forever preferred_lft forever
inet6 2a02:6d40:3ca5:2001:b81d:fa4a:8cdd:cf17/128 scope global
valid_lft forever preferred_lft forever
inet6 2a02:6d40:3ca5:2001:5c90:2ff:feeb:b02a/64 scope global dynamic mngtmpaddr
valid_lft 3591sec preferred_lft 3591sec
inet6 fe80::5c90:2ff:feeb:b02a/64 scope link
valid_lft forever preferred_lft forever
# ip route show
default via 192.168.1.1 dev eth0
192.168.1.0/24 dev eth0 proto kernel scope link src 192.168.1.118
# ip -6 route show
2a02:6d40:3ca5:2001:b81d:fa4a:8cdd:cf17 dev eth0 proto kernel metric 256 pref medium
2a02:6d40:3ca5:2001::/64 dev eth0 proto kernel metric 256 expires 3584sec pref medium
fe80::/64 dev eth0 proto kernel metric 256 pref medium
default via fe80::62e3:27ff:fe33:2b01 dev eth0 proto ra metric 1024 expires 3584sec pref medium
# iperf3 -c 127.0.0.1 -t1
Connecting to host 127.0.0.1, port 5201
[ 5] local 127.0.0.1 port 51938 connected to 127.0.0.1 port 5201
[ ID] Interval Transfer Bitrate Retr Cwnd
[ 5] 0.00-1.00 sec 4.46 GBytes 38.3 Gbits/sec 0 3.93 MBytes
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-1.00 sec 4.46 GBytes 38.3 Gbits/sec 0 sender
[ 5] 0.00-1.41 sec 4.45 GBytes 27.1 Gbits/sec receiver
iperf Done.
# iperf3 -c ::1 -t1
Connecting to host ::1, port 5201
[ 5] local ::1 port 50108 connected to ::1 port 5201
[ ID] Interval Transfer Bitrate Retr Cwnd
[ 5] 0.00-1.00 sec 4.35 GBytes 37.4 Gbits/sec 0 4.99 MBytes
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-1.00 sec 4.35 GBytes 37.4 Gbits/sec 0 sender
[ 5] 0.00-1.41 sec 4.35 GBytes 26.4 Gbits/sec receiver
iperf Done.
# ping -c1 -4 spaghetti.pizza
PING spaghetti.pizza (172.67.192.217) 56(84) bytes of data.
64 bytes from 172.67.192.217: icmp_seq=1 ttl=255 time=37.3 ms
--- spaghetti.pizza ping statistics ---
1 packets transmitted, 1 received, 0% packet loss, time 0ms
# ping -c1 -6 spaghetti.pizza
PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes
64 bytes from 2606:4700:3034::6815:147a: icmp_seq=1 ttl=255 time=35.6 ms
--- spaghetti.pizza ping statistics ---
1 packets transmitted, 1 received, 0% packet loss, time 0ms
rtt min/avg/max/mdev = 35.605/35.605/35.605/0.000 ms
# logout
$
Connect an existing user and network namespace
$ unshare -rUn
# echo $$
2446678
[From another terminal]
$ ./pasta 2446678
Outbound interface: eth0, namespace interface: eth0
ARP:
address: 28:16:ad:39:a9:ea
DHCP:
assign: 192.168.1.118
mask: 255.255.255.0
router: 192.168.1.1
NDP/DHCPv6:
assign: 2a02:6d40:3ca5:2001:b81d:fa4a:8cdd:cf17
router: fe80::62e3:27ff:fe33:2b01
[Back to the original terminal]
# dhclient -4 --no-pid
# dhclient -6 --no-pid
# ip address show
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: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 65520 qdisc pfifo_fast state UNKNOWN group default qlen 1000
link/ether fa:c1:2a:27:92:a9 brd ff:ff:ff:ff:ff:ff
inet 192.168.1.118/24 brd 192.168.1.255 scope global eth0
valid_lft forever preferred_lft forever
inet6 2a02:6d40:3ca5:2001:b81d:fa4a:8cdd:cf17/128 scope global
valid_lft forever preferred_lft forever
inet6 2a02:6d40:3ca5:2001:f8c1:2aff:fe27:92a9/64 scope global dynamic mngtmpaddr
valid_lft 3594sec preferred_lft 3594sec
inet6 fe80::f8c1:2aff:fe27:92a9/64 scope link
valid_lft forever preferred_lft forever
passt
Start and connect a guest with basic port forwarding
$ ./passt -f -t 2222:22
Outbound interface: eth0
ARP:
address: 28:16:ad:39:a9:ea
DHCP:
assign: 192.168.1.118
mask: 255.255.255.0
router: 192.168.1.1
search:
redhat.com
NDP/DHCPv6:
assign: 2a02:6d40:3ca5:2001:b81d:fa4a:8cdd:cf17
router: fe80::62e3:27ff:fe33:2b01
search:
redhat.com
UNIX domain socket bound at /tmp/passt_1.socket
You can now start qrap:
./qrap 5 qemu-system-x86_64 ... -net socket,fd=5 -net nic,model=virtio
or directly qemu, patched with:
qemu/0001-net-Allow-also-UNIX-domain-sockets-to-be-used-as-net.patch
as follows:
qemu-system-x86_64 ... -net socket,connect=/tmp/passt_1.socket -net nic,model=virtio
[From another terminal]
$ ./qrap 5 qemu-system-x86_64 test.qcow2 -m 1024 -display none -nodefaults -nographic -net socket,fd=5 -net nic,model=virtio
Connected to /tmp/passt_1.socket
[Back to the original terminal]
passt: DHCP: ack to request
passt: from 52:54:00:12:34:56
passt: NDP: received NS, sending NA
passt: NDP: received RS, sending RA
passt: DHCPv6: received SOLICIT, sending ADVERTISE
passt: NDP: received NS, sending NA
passt: DHCPv6: received REQUEST/RENEW/CONFIRM, sending REPLY
passt: NDP: received NS, sending NA
[From yet another terminal]
$ ssh -p 2222 root@localhost
root@localhost's password:
[...]
# ip address show
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: ens2: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 65520 qdisc pfifo_fast state UP group default qlen 1000
link/ether 52:54:00:12:34:56 brd ff:ff:ff:ff:ff:ff
inet 192.168.1.118/24 brd 192.168.1.255 scope global noprefixroute ens2
valid_lft forever preferred_lft forever
inet6 2a02:6d40:3ca5:2001:b81d:fa4a:8cdd:cf17/128 scope global noprefixroute
valid_lft forever preferred_lft forever
inet6 2a02:6d40:3ca5:2001:b019:9ae2:a2fe:e6b4/64 scope global dynamic noprefixroute
valid_lft 3588sec preferred_lft 3588sec
inet6 fe80::1f98:d09f:9309:9e77/64 scope link noprefixroute
valid_lft forever preferred_lft forever
NOTES
Handling of traffic with loopback destination and source addresses
Both passt and pasta can bind on ports with a loopback address
(127.0.0.0/8 or ::1), depending on the configuration. Loopback destina-
tion or source addresses need to be changed before packets are delivered
to the guest or target namespace: most operating systems would drop
packets received with loopback addresses on non-loopback interfaces, and
it would also be impossible for guest or target namespace to route an-
swers back.
For convenience, the source address on these packets is translated to
the address specified by the --map-host-loopback option (with some ex-
ceptions in pasta mode, see next section below). If not specified this
defaults, somewhat arbitrarily, to the address of default IPv4 or IPv6
gateway (if any) -- this is known to be an existing, valid address on
the same subnet. If --no-map-gw or --map-host-loopback none are speci-
fied this translation is disabled and packets with loopback addresses
are simply dropped.
Loopback destination addresses are translated to the observed external
address of the guest or target namespace. For IPv6, the observed link-
local address is used if the translated source address is link-local,
otherwise the observed global address is used. For both IPv4 and IPv6,
if no addresses have been seen yet, the configured addresses will be
used instead.
For example, if passt or pasta receive a connection from 127.0.0.1, with
destination 127.0.0.10, and the default IPv4 gateway is 192.0.2.1, while
the last observed source address from guest or namespace is 192.0.2.2,
this will be translated to a connection from 192.0.2.1 to 192.0.2.2.
Similarly, for traffic coming from guest or namespace, packets with des-
tination address corresponding to the --map-host-loopback address will
have their destination address translated to a loopback address.
As an exception, traffic identified as DNS, originally directed to the
--map-host-loopback address, if this address matches a resolver address
on the host, is not translated to loopback, but rather handled in the
same way as if specified as --dns-forward address, if no such option was
given. In the common case where the host gateway also acts a resolver,
this avoids that the host mapping shadows the gateway/resolver itself.
Handling of local traffic in pasta
Depending on the configuration, pasta can bind to local ports in the
init namespace, in the target namespace, or both, and forward connec-
tions and packets to corresponding ports in the other namespace.
To avoid unnecessary overhead, these connections and packets are not
forwarded through the tap device connecting the namespaces: pasta cre-
ates a socket in the destination namespace, with matching Layer-4 proto-
col, and uses it to forward local data. For TCP, data is forwarded be-
tween the originating socket and the new socket using the splice(2) sys-
tem call, and for UDP, a pair of recvmmsg(2) and sendmmsg(2) system
calls deals with packet transfers.
Because it's not possible to bind sockets to foreign addresses, this by-
pass only applies to local connections and traffic. It also means that
the address translation differs slightly from passt mode. Connections
from loopback to loopback on the host will appear to come from the tar-
get namespace's public address within the guest, unless --host-lo-to-ns-
lo is specified, in which case they will appear to come from loopback in
the namespace as well. The latter behaviour used to be the default, but
is usually undesirable, since it can unintentionally expose namespace
local services to the host.
Binding to low numbered ports (well-known or system ports, up to 1023)
If the port forwarding configuration requires binding to ports with num-
bers lower than 1024, passt and pasta will try to bind to them, but will
fail, unless, either:
• the sys.net.ipv4.ip_unprivileged_port_start sysctl is set to the num-
ber of the lowest port passt and pasta need. For example, as root:
sysctl -w net.ipv4.ip_unprivileged_port_start=443
Note: this is the recommended way of enabling passt and pasta to bind
to ports with numbers below 1024.
• or the CAP_NET_BIND_SERVICE Linux capability is granted, see ser-
vices(5) and capabilities(7).
This is, in general, not the recommended way, because passt and pasta
might be used as vector to effectively use this capability from an-
other process.
However, if your environment is sufficiently controlled by an LSM
(Linux Security Module) such as AppArmor, SELinux, Smack or TOMOYO,
and no other processes can interact in such a way in virtue of this,
granting this capability to passt and pasta only can effectively pre-
vent other processes from utilising it.
Note that this will not work for automatic detection and forwarding of
ports with pasta, because pasta will relinquish this capability at
runtime.
To grant this capability, you can issue, as root:
for p in $(which passt passt.avx2); do
setcap 'cap_net_bind_service=+ep' "${p}"
done
ICMP/ICMPv6 Echo sockets
ICMP and ICMPv6 Echo requests coming from guest or target namespace are
handled using so-called "ping" sockets, introduced in Linux 2.6.30. To
preserve the original identifier (see RFC 792, page 14, for ICMP, and
RFC 4443, section 4.1, for ICMPv6), passt and pasta try to bind these
sockets using the observed source identifier as "port" -- that corre-
sponds to Echo identifiers for "ping" sockets.
As bind(2) failures were seen with particularly restrictive SELinux
policies, a fall-back mechanism maps different identifiers to different
sockets, and identifiers in replies will be mapped back to the original
identifier of the request. However, if bind(2) fails and the fall-back
mechanism is used, echo requests will be forwarded with different, al-
beit unique, identifiers.
For ICMP and ICMPv6 Echo requests to work, the ping_group_range parame-
ter needs to include the PID of passt or pasta, see icmp(7).
pasta and loopback interface
As pasta connects to an existing namespace, or once it creates a new
namespace, it will also ensure that the loopback interface, lo, is
brought up. This is needed to bind ports using the loopback address in
the namespace.
TCP sending window and TCP_INFO before Linux 5.3
To synchronise the TCP sending window from host Layer-4 sockets to the
TCP parameters announced in TCP segments sent over the Layer-2 inter-
face, passt and pasta routinely query the size of the sending window
seen by the kernel on the corresponding socket using the TCP_INFO socket
option, see tcp(7). Before Linux 5.3, i.e. before Linux kernel commit
8f7baad7f035 ("tcp: Add snd_wnd to TCP_INFO"), the sending window
(snd_wnd field) is not available.
If the sending window cannot be queried, it will always be announced as
the current sending buffer size to guest or target namespace. This might
affect throughput of TCP connections.
Local mode for disconnected setups
If passt and pasta fail to find a host interface with a configured ad-
dress, other than loopback addresses, they will, obviously, not attempt
to source addresses or routes from the host.
In this case, unless configured otherwise, they will assign the IPv4
link-local address 169.254.2.1 to the guest or target namespace, and no
IPv6 address. The notion of the guest or target namespace IPv6 address
is derived from the first link-local address observed.
Default gateways will be assigned as the link-local address 169.254.2.2
for IPv4, and as the link-local address fe80::1 for IPv6.
LIMITATIONS
Currently, IGMP/MLD proxying (RFC 4605) and support for SCTP (RFC 4960)
are not implemented.
TCP Selective Acknowledgment (RFC 2018), as well as Protection Against
Wrapped Sequences (PAWS) and Round-Trip Time Measurement (RTTM), both
described by RFC 7232, are currently not implemented.
AUTHORS
Stefano Brivio <sbrivio@redhat.com>, David Gibson <david@gibson.drop-
bear.id.au>.
REPORTING BUGS
Please report issues on the bug tracker at https://passt.top/passt/bugs,
or send a message to the passt-user@passt.top mailing list, see
https://passt.top/passt/lists.
COPYRIGHT
Copyright (c) 2020-2022 Red Hat GmbH.
passt and pasta are free software: you can redistribute them and/or mod-
ify them under the terms of the GNU General Public License as published
by the Free Software Foundation, either version 2 of the License, or (at
your option) any later version.
SEE ALSO
namespaces(7), qemu(1), qrap(1), slirp4netns(1).
High-level documentation is available at https://passt.top/passt/about/.
passt(1)
Generated by dwww version 1.16 on Tue Dec 16 05:55:43 CET 2025.