table of contents
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
namespaces
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 network 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 provide 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 reflecting one peer's observed parameters (congestion window size, acknowledged 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 namespace, and spawns the given command or a default shell within this context. 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 removed 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. Alternatively, 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 nobody 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) network 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 default, assigned IPv4 and IPv6 addresses are taken from the host interfaces with the first default route, if any, for the corresponding 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 gateways 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 loopback or local source address, to allow mapping of local traffic to guest and target namespace. See the NOTES below for more details 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
--outbound-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 addresses and routes, and link-local addresses will be used instead.
- -o, --outbound addr
- Use an IPv4 addr as source address for IPv4 outbound TCP connections, 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
interface 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 interface 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
interface 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 interface 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 remapping. 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/resolv.conf.
- -S, --search list
- Use space-separated list for DHCP, DHCPv6, and NDP purposes, instead of reading entries from /etc/resolv.conf. See options --no-dhcp-search and --dhcp-search. --search none disables the DNS domain 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 addresses, 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 resulting from configuration is not passed.
- --no-tcp
- Disable the TCP protocol handler. No TCP connections will be accepted 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 requests 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 immediately 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 effect.
Default is to translate the guest's default gateway address, unless --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 default 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-loopback option.
- -4, --ipv4-only
- Enable IPv4-only operation. IPv6 traffic will be ignored. By default, IPv6 operation is enabled as long as at least an IPv6 route and an interface address are configured on a given host interface.
- -6, --ipv6-only
- Enable IPv6-only operation. IPv4 traffic will be ignored. By default, IPv4 operation is enabled as long as at least an IPv4 route and an interface address are configured on a given host interface.
passt-only options¶
- -s, --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 connections, starting from /tmp/passt_1.socket to /tmp/passt_64.socket.
- -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 another
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 differ. Specific
addresses can be bound as well, separated by /, and also, since
Linux 5.7, limited to specific interfaces, 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 reported 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 corresponding ports on the guest
- -t 22-80:32-90
- Forward local ports between 22 and 80 to ports between 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 interface eth0, to port 22
- -t %eth0/22
- Forward local port 22, bound to any address on interface 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 between 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 default, the same interface name as the external, routable interface 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 listening 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 differ. Specific
addresses can be bound as well, separated by /, and also, since
Linux 5.7, limited to specific interfaces, 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 reported 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 corresponding ports in the target namespace
- -t 22-80:32-90
- Forward local ports between 22 and 80 to ports between 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 interface eth0, to port 22
- -t %eth0/22
- Forward local port 22, bound to any address on interface 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 between 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 namespace.
spec is as described above for TCP.
Default is auto.
- -U, --udp-ns spec
- Configure UDP port forwarding from target namespace to init namespace.
spec is as described above for UDP.
Default is auto.
- --host-lo-to-ns-lo (DEPRECATED)
- 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, without this option, the user namespace will be the one of the corresponding 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 output interface attribute to the outbound interface in the namespace.
Note that this configuration option is deprecated and will be removed 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 single one.
Implied by -a, --address, for the corresponding IP version only.
Default is to copy all the addresses, except for link-local ones, from the interface from the outer namespace to the target namespace.
Note that this configuration option is deprecated and will be removed 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.
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:[Back to the original terminal]
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 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:[From another terminal]
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
$ ./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 destination 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 answers back.
For convenience, the source address on these packets is translated to the address specified by the --map-host-loopback option (with some exceptions 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 specified 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 destination address corresponding to the --map-host-loopback address will have their destination address translated to a loopback address.
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 connections 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 creates a socket in the destination namespace, with matching Layer-4 protocol, and uses it to forward local data. For TCP, data is forwarded between the originating socket and the new socket using the splice(2) system 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 bypass 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 target 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 numbers 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
number 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
services(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 another 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 prevent 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 corresponds 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, albeit unique, identifiers.
For ICMP and ICMPv6 Echo requests to work, the ping_group_range parameter 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 interface, 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 address, 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.dropbear.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 modify 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/.