Re: [pve-devel] [PATCH qemu-server 2/2] virtio-net: fix migration between default/non-default MTUs starting with machine version 10.0+pve1

["Fabian Grünbichler" <f.gruenbichler@proxmox.com>]

On September 2, 2025 3:44 pm, Fiona Ebner wrote:
> The virtual hardware is generated differently (at least for i440fx
> machines) when host_mtu is set or not set on the netdev command line
> [0]. When the MTU is the same value as the default 1500, Proxmox VE
> did not add a host_mtu parameter. This is problematic for migration
> where host_mtu is present on one end of the migration, but not on the
> other [1]. Migration between command lines where host_mtu is set on
> both ends, even if set to different values, works fine.
> 
> Always set the host_mtu parameter starting with machine version
> 10.0+pve1 to avoid this issue going forward. Handling migrations with
> older machine versions is more involved and will be done in separate
> patches. Thanks to Stefan Hanreich and Fabian Grünbichler for
> discussing this with me!
> 
> Since print_netdevice_full() is also called for hotplug, it cannot
> always use the $version_guard helper and needs to fallback to
> min_version() then.
> 
> [0]: https://bugzilla.redhat.com/show_bug.cgi?id=1449346
> [1]: https://forum.proxmox.com/threads/live-vm-migration-fails.169537/post-796379
> 
> Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
> ---
>  src/PVE/QemuServer.pm                         | 23 +++++++++++++++++--
>  src/PVE/QemuServer/Machine.pm                 |  6 +++++
>  src/test/cfg2cmd/bootorder-empty.conf.cmd     |  4 ++--
>  src/test/cfg2cmd/bootorder-legacy.conf.cmd    |  4 ++--
>  src/test/cfg2cmd/bootorder.conf.cmd           |  4 ++--
>  src/test/cfg2cmd/efidisk-on-rbd.conf.cmd      |  4 ++--
>  src/test/cfg2cmd/ide.conf.cmd                 |  4 ++--
>  .../cfg2cmd/netdev-7.1-multiqueues.conf.cmd   |  2 +-
>  src/test/cfg2cmd/netdev-7.1.conf.cmd          |  2 +-
>  src/test/cfg2cmd/netdev_vxlan.conf.cmd        |  2 +-
>  src/test/cfg2cmd/q35-ide.conf.cmd             |  4 ++--
>  .../q35-linux-hostpci-mapping.conf.cmd        |  4 ++--
>  .../q35-linux-hostpci-multifunction.conf.cmd  |  4 ++--
>  ...q35-linux-hostpci-x-pci-overrides.conf.cmd |  4 ++--
>  src/test/cfg2cmd/q35-linux-hostpci.conf.cmd   |  4 ++--
>  src/test/cfg2cmd/q35-simple.conf.cmd          |  4 ++--
>  src/test/cfg2cmd/seabios_serial.conf.cmd      |  4 ++--
>  src/test/cfg2cmd/simple-btrfs.conf.cmd        |  4 ++--
>  .../cfg2cmd/simple-disk-passthrough.conf.cmd  |  4 ++--
>  src/test/cfg2cmd/simple-rbd.conf.cmd          |  4 ++--
>  src/test/cfg2cmd/simple-virtio-blk.conf.cmd   |  4 ++--
>  .../cfg2cmd/simple-zfs-over-iscsi.conf.cmd    |  4 ++--
>  src/test/cfg2cmd/simple1.conf.cmd             |  4 ++--
>  23 files changed, 66 insertions(+), 41 deletions(-)
> 
> diff --git a/src/PVE/QemuServer.pm b/src/PVE/QemuServer.pm
> index 38fa3f83..8528f9f3 100644
> --- a/src/PVE/QemuServer.pm
> +++ b/src/PVE/QemuServer.pm
> @@ -1457,7 +1457,17 @@ sub print_pbs_blockdev {
>  }
>  

we could avoid the need for $version_guard in print_netdevice_full, if
we do something like:

----8<----
diff --git a/src/PVE/QemuServer.pm b/src/PVE/QemuServer.pm
index 8528f9f3..d3c2486d 100644
--- a/src/PVE/QemuServer.pm
+++ b/src/PVE/QemuServer.pm
@@ -1466,7 +1466,6 @@ sub print_netdevice_full {
         $use_old_bios_files,
         $arch,
         $machine_version,
-        $version_guard,
     ) = @_;
 
     my $device = $net->{model};
@@ -1505,10 +1504,7 @@ sub print_netdevice_full {
             die "netdev $netid: MTU '$mtu' is bigger than the bridge MTU '$bridge_mtu'\n";
         }
 
-        my $always_set_host_mtu =
-            $version_guard
-            ? $version_guard->(10, 0, 1)
-            : min_version($machine_version, 10, 0, 1);
+        my $always_set_host_mtu = min_version($machine_version, 10, 0, 1);
         if ($always_set_host_mtu) {
             $tmpstr .= ",host_mtu=$mtu";
         } else {
@@ -3837,6 +3833,8 @@ sub config_to_command {
         my $netdevfull = print_netdev_full($vmid, $conf, $arch, $d, $netname);
         push @$devices, '-netdev', $netdevfull;
 
+        # force +pve1 if machine version 10, for host_mtu differentiation
+        $version_guard->(10, 0, 1);
         my $netdevicefull = print_netdevice_full(
             $vmid,
             $conf,
@@ -3846,7 +3844,6 @@ sub config_to_command {
             $use_old_bios_files,
             $arch,
             $machine_version,
-            $version_guard,
         );
 
         push @$devices, '-device', $netdevicefull;
---->8----

(with the downside of now bumping to +pve1 for unversioned VMs as soon
as they have a NIC, instead of more fine-grained, but that shouldn't
really hurt?)

that might make it easier to (at some point) move more of this into its
own module (or QemuServer::Network)?

other than this, seems to behave as expected so either variant is fine
for me :)

>  sub print_netdevice_full {
> -    my ($vmid, $conf, $net, $netid, $bridges, $use_old_bios_files, $arch, $machine_version) = @_;
> +    my (
> +        $vmid,
> +        $conf,
> +        $net,
> +        $netid,
> +        $bridges,
> +        $use_old_bios_files,
> +        $arch,
> +        $machine_version,
> +        $version_guard,
> +    ) = @_;
>  
>      my $device = $net->{model};
>      if ($net->{model} eq 'virtio') {
> @@ -1495,7 +1505,15 @@ sub print_netdevice_full {
>              die "netdev $netid: MTU '$mtu' is bigger than the bridge MTU '$bridge_mtu'\n";
>          }
>  
> -        $tmpstr .= ",host_mtu=$mtu" if $mtu != 1500;
> +        my $always_set_host_mtu =
> +            $version_guard
> +            ? $version_guard->(10, 0, 1)
> +            : min_version($machine_version, 10, 0, 1);
> +        if ($always_set_host_mtu) {
> +            $tmpstr .= ",host_mtu=$mtu";
> +        } else {
> +            $tmpstr .= ",host_mtu=$mtu" if $mtu != 1500;
> +        }
>      } elsif (defined($mtu)) {
>          warn
>              "WARN: netdev $netid: ignoring MTU '$mtu', not using VirtIO or no bridge configured.\n";
> @@ -3828,6 +3846,7 @@ sub config_to_command {
>              $use_old_bios_files,
>              $arch,
>              $machine_version,
> +            $version_guard,
>          );
>  
>          push @$devices, '-device', $netdevicefull;
> diff --git a/src/PVE/QemuServer/Machine.pm b/src/PVE/QemuServer/Machine.pm
> index 9d17344a..4c135a20 100644
> --- a/src/PVE/QemuServer/Machine.pm
> +++ b/src/PVE/QemuServer/Machine.pm
> @@ -37,6 +37,12 @@ our $PVE_MACHINE_VERSION = {
>              '+pve1' => 'Disables S3/S4 power states by default.',
>          },
>      },
> +    '10.0' => {
> +        highest => 1,
> +        revisions => {
> +            '+pve1' => 'Set host_mtu vNIC option even with default value for migration compat.',
> +        },
> +    },
>  };
>  
>  my $machine_fmt = {


_______________________________________________
pve-devel mailing list
pve-devel@lists.proxmox.com
https://lists.proxmox.com/cgi-bin/mailman/listinfo/pve-devel