docs/system/devices/net.rst | 93 +++++++++++++++++++++++++++++++++++++ 1 file changed, 93 insertions(+)
Add a chapter to explain how to use passt(1) instead of '-net user'.
passt(1) can be connected to QEMU using UNIX socket or vhost-user.
With vhost-user, migration of the VM is allowed and internal state of
passt(1) is transfered from one side to the other
Bug: https://gitlab.com/qemu-project/qemu/-/issues/2827
Signed-off-by: Laurent Vivier <lvivier@redhat.com>
---
docs/system/devices/net.rst | 93 +++++++++++++++++++++++++++++++++++++
1 file changed, 93 insertions(+)
diff --git a/docs/system/devices/net.rst b/docs/system/devices/net.rst
index 2ab516d4b097..5f70b1039198 100644
--- a/docs/system/devices/net.rst
+++ b/docs/system/devices/net.rst
@@ -77,6 +77,99 @@ When using the ``'-netdev user,hostfwd=...'`` option, TCP or UDP
connections can be redirected from the host to the guest. It allows for
example to redirect X11, telnet or SSH connections.
+Using passt as the user mode network stack
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+`passt(1)`_ can be used as a simple replacement for SLIRP (``-net user``).
+`passt(1)`_ doesn't require any capability or privilege. `passt(1)`_ has
+better performance than ``-net user``, full IPv6 support and better security
+as it's a daemon that is not executed in QEMU context.
+
+`passt(1)`_ can be connected to QEMU either by using a socket
+(``-netdev stream``) or using the vhost-user interface (``-netdev vhost-user``).
+See `passt web site`_ and `passt(1)`_ for more details on passt.
+
+.. _passt web site: https://passt.top/
+.. _passt(1): https://passt.top/builds/latest/web/passt.1.html
+
+To use socket based passt interface:
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+Start `passt(1)`_ as a daemon::
+
+ passt
+
+It will print the path of the UNIX domain socket QEMU can connect to.
+Then you can connect your QEMU instance to `passt(1)`_:
+
+.. parsed-literal::
+ |qemu_system| [...OPTIONS...] -device virtio-net-pci,netdev=netdev0 -netdev stream,id=netdev0,server=off,addr.type=unix,addr.path=/tmp/passt_1.socket
+
+Where ``/tmp/passt_1.socket`` is the UNIX socket created by `passt(1)`_ to
+communicate with QEMU.
+
+To use vhost-based interface:
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+Start passt with ``--vhost-user``::
+
+ passt --vhost-user
+
+Then to connect QEMU:
+
+.. parsed-literal::
+ |qemu_system| [...OPTIONS...] -m $RAMSIZE -chardev socket,id=chr0,path=/tmp/passt_1.socket -netdev vhost-user,id=netdev0,chardev=chr0 -device virtio-net,netdev=netdev0 -object memory-backend-memfd,id=memfd0,share=on,size=$RAMSIZE -numa node,memdev=memfd0
+
+Where ``$RAMSIZE`` is the memory size of your VM ``-m`` and ``-object memory-backend-memfd,size=`` must match.
+
+Migration of passt:
+^^^^^^^^^^^^^^^^^^^
+
+When `passt(1)`_ is connected to QEMU using the vhost-user interface it can
+be migrated with QEMU and the network connections are not interrupted.
+
+As `passt(1)`_ runs with no privileges, it relies on passt-repair to save and
+load TCP connections state, using the TCP_REPAIR socket option.
+The passt-repair helper needs to have the CAP_NET_ADMIN capability, or run as root. If passt-repair is not available, TCP connections will not be preserved.
+
+Example of migration of a guest on the same host
+________________________________________________
+
+Before being able to run passt-repair, the CAP_NET_ADMIN capability must be set
+on the file, run as root::
+
+ setcat cap_net_admin+eip ./passt-repair
+
+Start `passt(1)`_ for the source side::
+
+ passt
+
+Start passt-repair::
+
+ passt-repair /tmp/passt_1.socket.repair
+
+Start source side QEMU with a monitor to be able to send the migrate command:
+
+.. parsed-literal::
+ |qemu_system| [...OPTIONS...] [...VHOST USER OPTIONS...] -monitor stdio
+
+Start `passt(1)`_ for the destination side::
+
+ passt
+
+Start passt-repair::
+
+ passt-repair /tmp/passt_2.socket.repair
+
+Start QEMU with the ``-incoming`` parameter:
+
+.. parsed-literal::
+ |qemu_system| [...OPTIONS...] [...VHOST USER OPTIONS...] -incoming tcp:localhost:4444
+
+Then in the source guest monitor the migration can be started::
+
+ (qemu) migrate tcp:localhost:4444
+
Hubs
~~~~
--
2.48.1On Mon, Mar 10, 2025 at 07:00:50PM +0100, Laurent Vivier wrote: > Add a chapter to explain how to use passt(1) instead of '-net user'. > passt(1) can be connected to QEMU using UNIX socket or vhost-user. > With vhost-user, migration of the VM is allowed and internal state of > passt(1) is transfered from one side to the other > > Bug: https://gitlab.com/qemu-project/qemu/-/issues/2827 > Signed-off-by: Laurent Vivier <lvivier@redhat.com> > --- > docs/system/devices/net.rst | 93 +++++++++++++++++++++++++++++++++++++ > 1 file changed, 93 insertions(+) > > diff --git a/docs/system/devices/net.rst b/docs/system/devices/net.rst > index 2ab516d4b097..5f70b1039198 100644 > --- a/docs/system/devices/net.rst > +++ b/docs/system/devices/net.rst > @@ -77,6 +77,99 @@ When using the ``'-netdev user,hostfwd=...'`` option, TCP or UDP > connections can be redirected from the host to the guest. It allows for > example to redirect X11, telnet or SSH connections. > > +Using passt as the user mode network stack > +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ > + > +`passt(1)`_ can be used as a simple replacement for SLIRP (``-net user``). > +`passt(1)`_ doesn't require any capability or privilege. `passt(1)`_ has > +better performance than ``-net user``, full IPv6 support and better security > +as it's a daemon that is not executed in QEMU context. > + > +`passt(1)`_ can be connected to QEMU either by using a socket > +(``-netdev stream``) or using the vhost-user interface (``-netdev vhost-user``). > +See `passt web site`_ and `passt(1)`_ for more details on passt. > + > +.. _passt web site: https://passt.top/ > +.. _passt(1): https://passt.top/builds/latest/web/passt.1.html > + > +To use socket based passt interface: > +^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ > + > +Start `passt(1)`_ as a daemon:: > + > + passt > + > +It will print the path of the UNIX domain socket QEMU can connect to. > +Then you can connect your QEMU instance to `passt(1)`_: > + > +.. parsed-literal:: > + |qemu_system| [...OPTIONS...] -device virtio-net-pci,netdev=netdev0 -netdev stream,id=netdev0,server=off,addr.type=unix,addr.path=/tmp/passt_1.socket > + > +Where ``/tmp/passt_1.socket`` is the UNIX socket created by `passt(1)`_ to > +communicate with QEMU. Using the default socket path in /tmp makes me slightly nervous in terms of other users managing to intercept the network somehow - or simply in terms of getting different passt instances confused by accident. I therefore wonder if it might be better to suggest explicitly choosing the socket path on both sides. > +To use vhost-based interface: > +^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ > + > +Start passt with ``--vhost-user``:: > + > + passt --vhost-user > + > +Then to connect QEMU: > + > +.. parsed-literal:: > + |qemu_system| [...OPTIONS...] -m $RAMSIZE -chardev socket,id=chr0,path=/tmp/passt_1.socket -netdev vhost-user,id=netdev0,chardev=chr0 -device virtio-net,netdev=netdev0 -object memory-backend-memfd,id=memfd0,share=on,size=$RAMSIZE -numa node,memdev=memfd0 > + > +Where ``$RAMSIZE`` is the memory size of your VM ``-m`` and ``-object memory-backend-memfd,size=`` must match. Same consideration with the path here. > +Migration of passt: > +^^^^^^^^^^^^^^^^^^^ > + > +When `passt(1)`_ is connected to QEMU using the vhost-user interface it can > +be migrated with QEMU and the network connections are not interrupted. > + > +As `passt(1)`_ runs with no privileges, it relies on passt-repair to save and > +load TCP connections state, using the TCP_REPAIR socket option. > +The passt-repair helper needs to have the CAP_NET_ADMIN capability, or run as root. If passt-repair is not available, TCP connections will not be preserved. I'm slightly nervous about using a socket in /tmp for passt proper. I'm much more nervous about using a socket in /tmp for passt-repair. > +Example of migration of a guest on the same host > +________________________________________________ > + > +Before being able to run passt-repair, the CAP_NET_ADMIN capability must be set > +on the file, run as root:: > + > + setcat cap_net_admin+eip ./passt-repair > + > +Start `passt(1)`_ for the source side:: > + > + passt > + > +Start passt-repair:: > + > + passt-repair /tmp/passt_1.socket.repair > + > +Start source side QEMU with a monitor to be able to send the migrate command: > + > +.. parsed-literal:: > + |qemu_system| [...OPTIONS...] [...VHOST USER OPTIONS...] -monitor stdio It might be worth pointing out explicitly that a separate passt-repair instance must be started for every migration. In some cases, I think passt-repair will also need to be restarted in the case of a failed migration, before trying again. > +Start `passt(1)`_ for the destination side:: > + > + passt > + > +Start passt-repair:: > + > + passt-repair /tmp/passt_2.socket.repair > + > +Start QEMU with the ``-incoming`` parameter: > + > +.. parsed-literal:: > + |qemu_system| [...OPTIONS...] [...VHOST USER OPTIONS...] -incoming tcp:localhost:4444 > + > +Then in the source guest monitor the migration can be started:: > + > + (qemu) migrate tcp:localhost:4444 > + > Hubs > ~~~~ > -- David Gibson (he or they) | I'll have my music baroque, and my code david AT gibson.dropbear.id.au | minimalist, thank you, not the other way | around. http://www.ozlabs.org/~dgibson
On 3/10/25 19:00, Laurent Vivier wrote: > +Using passt as the user mode network stack > +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ > + > +`passt(1)`_ can be used as a simple replacement for SLIRP (``-net user``). I would use just `passt` for the link, linking to passt.top only the first time and... > +`passt(1)`_ can be connected to QEMU either by using a socket > +(``-netdev stream``) or using the vhost-user interface (``-netdev vhost-user``). > +See `passt web site`_ and `passt(1)`_ for more details on passt. ... the man page here. After the first two occurrences you don't need any more links. > +Example of migration of a guest on the same host > +________________________________________________ > + > +Before being able to run passt-repair, the CAP_NET_ADMIN capability must be set > +on the file, run as root:: > + > + setcat cap_net_admin+eip ./passt-repair setcap :) > +Start `passt(1)`_ for the source side:: > + > + passt > + > +Start passt-repair:: > + > + passt-repair /tmp/passt_1.socket.repair > + > +Start source side QEMU with a monitor to be able to send the migrate command: > + > +.. parsed-literal:: > + |qemu_system| [...OPTIONS...] [...VHOST USER OPTIONS...] -monitor stdio > + > +Start `passt(1)`_ for the destination side:: > + > + passt > + > +Start passt-repair:: > + > + passt-repair /tmp/passt_2.socket.repair > + > +Start QEMU with the ``-incoming`` parameter: > + > +.. parsed-literal:: > + |qemu_system| [...OPTIONS...] [...VHOST USER OPTIONS...] -incoming tcp:localhost:4444 > + > +Then in the source guest monitor the migration can be started:: > + > + (qemu) migrate tcp:localhost:4444 > + > Hubs > ~~~~ >
© 2016 - 2025 Red Hat, Inc.