Laszlo Ersek
2021-Dec-23 10:37 UTC
[Libguestfs] [PATCH 3/3] launch-libvirt: add virtio-net via the standard <interface> element
Starting with version 3.8.0, libvirt allows us to specify the network
address and network mask (as prefix) for SLIRP directly via the
<interface> element in the domain XML:
<https://libvirt.org/formatdomain.html#userspace-slirp-stack>. This means
we don't need the <qemu:commandline> hack for virtio-net on such
versions.
Restrict the hack in construct_libvirt_xml_qemu_cmdline() to
libvirt<3.8.0, and generate the proper <interface> element in
construct_libvirt_xml_devices() on libvirt>=3.8.0.
Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=2034160
Suggested-by: Richard W.M. Jones <rjones at redhat.com>
Signed-off-by: Laszlo Ersek <lersek at redhat.com>
---
Notes:
v1:
- Tested with "virt-rescue --network" and "ping", using
the "libvirt"
back-end, after setting "net.ipv4.ping_group_range" properly in
sysctl
on the host. (Libvirt version: libvirt-daemon-7.6.0-3.fc35.x86_64.)
- The generated libvirt XML, per "virsh dumpxml", contains
<devices>
<interface type='user'>
<mac address='52:54:00:fe:23:48'/>
<ip address='169.254.0.0' family='ipv4'
prefix='16'/>
<model type='virtio'/>
<alias name='net0'/>
<address type='pci' domain='0x0000'
bus='0x00' slot='0x02' function='0x0'/>
</interface>
</devices>
<qemu:commandline>
<qemu:env name='TMPDIR'
value='/home/lacos/src/v2v/guestfs-tools/tmp'/>
</qemu:commandline>
and the consequent QEMU cmdline has
-netdev user,net=169.254.0.0/16,id=hostnet0 \
-device
virtio-net-pci,netdev=hostnet0,id=net0,mac=52:54:00:fe:23:48,bus=pci.0,addr=0x2
\
- The full set of PCI addresses that libvirt assigns to the virtio
devices are, per "virsh dumpxml":
<controller type='scsi' index='0'
model='virtio-scsi'>
<address type='pci' domain='0x0000'
bus='0x00' slot='0x03' function='0x0'/>
<controller type='virtio-serial' index='0'>
<address type='pci' domain='0x0000'
bus='0x00' slot='0x04' function='0x0'/>
<interface type='user'>
<address type='pci' domain='0x0000'
bus='0x00' slot='0x02' function='0x0'/>
<rng model='virtio'>
<address type='pci' domain='0x0000'
bus='0x00' slot='0x05' function='0x0'/>
- The same, verified with
virsh -c ... qemu-monitor-command guestfs-... --hmp info qtree
is:
bus: main-system-bus
dev: i440FX-pcihost, id ""
bus: pci.0
dev: virtio-rng-pci, id "rng0"
addr = 05.0
dev: virtio-net-pci, id "net0"
addr = 02.0
dev: virtio-serial-pci, id "virtio-serial0"
addr = 04.0
dev: virtio-scsi-pci, id "scsi0"
addr = 03.0
lib/guestfs-internal.h | 3 ++-
lib/launch-libvirt.c | 27 +++++++++++++++++++++++++--
2 files changed, 27 insertions(+), 3 deletions(-)
diff --git a/lib/guestfs-internal.h b/lib/guestfs-internal.h
index e24d570f5aa7..4a19e5c6bb47 100644
--- a/lib/guestfs-internal.h
+++ b/lib/guestfs-internal.h
@@ -173,7 +173,8 @@ cleanup_mutex_unlock (pthread_mutex_t **ptr)
#endif
/* Place the virtio-net controller in slot 0x1e on the root bus, on normal
- * hardware with PCI. Refer to RHBZ#2034160.
+ * hardware with PCI. Necessary only before libvirt 3.8.0. Refer to
+ * RHBZ#2034160.
*/
#ifdef HAVE_LIBVIRT_BACKEND
#if defined(__arm__) || defined(__s390x__)
diff --git a/lib/launch-libvirt.c b/lib/launch-libvirt.c
index 266d88824686..cc714c02efbf 100644
--- a/lib/launch-libvirt.c
+++ b/lib/launch-libvirt.c
@@ -1413,6 +1413,28 @@ construct_libvirt_xml_devices (guestfs_h *g,
} end_element ();
} end_element ();
+ /* Virtio-net NIC with SLIRP (= userspace) back-end, if networking is
+ * enabled. Starting with libvirt 3.8.0, we can specify the network address
+ * and prefix for SLIRP in the domain XML. Therefore, we can add the NIC
+ * via the standard <interface> element rather than
<qemu:commandline>, and
+ * so libvirt can manage the PCI address of the virtio-net NIC like the PCI
+ * addresses of all other devices. Refer to RHBZ#2034160.
+ */
+ if (g->enable_network &&
+ guestfs_int_version_ge (¶ms->data->libvirt_version, 3, 8,
0)) {
+ start_element ("interface") {
+ attribute ("type", "user");
+ start_element ("model") {
+ attribute ("type", "virtio");
+ } end_element ();
+ start_element ("ip") {
+ attribute ("family", "ipv4");
+ attribute ("address", NETWORK_ADDRESS);
+ attribute ("prefix", NETWORK_PREFIX);
+ } end_element ();
+ } end_element ();
+ }
+
/* Libvirt adds some devices by default. Indicate to libvirt
* that we don't want them.
*/
@@ -1835,9 +1857,10 @@ construct_libvirt_xml_qemu_cmdline (guestfs_h *g,
} end_element ();
/* Workaround because libvirt user networking cannot specify
"net="
- * parameter.
+ * parameter. Necessary only before libvirt 3.8.0; refer to RHBZ#2034160.
*/
- if (g->enable_network) {
+ if (g->enable_network &&
+ !guestfs_int_version_ge (¶ms->data->libvirt_version, 3,
8, 0)) {
start_element ("qemu:arg") {
attribute ("value", "-netdev");
} end_element ();
--
2.19.1.3.g30247aa5d201
Richard W.M. Jones
2021-Dec-23 11:31 UTC
[Libguestfs] [PATCH 3/3] launch-libvirt: add virtio-net via the standard <interface> element
Yes, this series looks all good to me, thanks for fixing this blocker so quickly. I also tested it on RHEL 9 and it fixes the "regression" (or whatever we want to call it!) there. So for the whole series: Reviewed-by: Richard W.M. Jones <rjones at redhat.com> Tested-by: Richard W.M. Jones <rjones at redhat.com> Rich. -- Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones Read my programming and virtualization blog: http://rwmj.wordpress.com virt-p2v converts physical machines to virtual machines. Boot with a live CD or over the network (PXE) and turn machines into KVM guests. http://libguestfs.org/virt-v2v