UDP Multicast on Host-Only LAN blocked on host

Hi,
My problem: The macOS host cannot communicate with any VM using UDP multicast in a "host-only" virtual LAN.
Using Parallels Destop 17.1.1 on macOS 12.1 (Intel), all VM running Linux (various distros, tests done with Fedora and Ubuntu).
Observations:

• Sending/receiving multicast from/to the macOS system works when communicating over Ethernet with an external physical Linux system (a Raspberry Pi). So, in general, multicast works on this macOS system. The problem is specific to the communication between the host and a Linux guest using Parallels.
• On the macOS host, the firewall allows incoming connections. I even tried with the firewall temporarily disabled and it did not change the problem. And it works anyway with external physical systems, so it cannot be a macOS firewall issue.
• The various Linux VM's successfully communicate between each other using the same interface on the "host-only" virtual LAN. The problem is specific to the communication between the host and any guest.
• Running Wireshark everywhere (the host and all guests) at the same time shows the same display on all Wireshark instances. So, the same data are present on all interfaces of the virtual LAN.
• Now, concentrating on the Wireshark instance running on the host:
  • When a Linux guest receives, we see the IGMP "join group" requests on Wireshark.
  • When a Linux guest sends, all multicast data packets are seen on Wireshark.
  • When the host receives, we do not see the IGMP join requests on Wireshark and the application never receives the data packets (which are displayed by Wireshark on the same host).
  • When the host sends, we do not see the outgoing data packets on Wireshark. The guests consequently never receive the data packets.
• So, it seems that there is something blocking any multicast traffic on the macOS host, between the application and the Wireshark probing point. This happens only on the Parallels virtual network interface ("bridge101"). This blocage does not happen on the physical network interface of the Mac ("en0").

Any idea?
Thanks.
-Thierry

 

Hi @ThierryL4
I'm facing the same issue on 17.1.4.
Admit that my host only network is 10.11.11.0/24.
The VM's IP is manually set to 10.11.11.2 and from the VM I was successfully pinging 10.11.11.1, which I deducted was the host virtual network IP address (is it documented somewhere ?)
Also, on the Mac side, the bridge101 interface in my case had the same IP address as my Windows VM (10.11.11.2) when running ifconfig on the Mac. This seemed incorrect IMO, but if I sudo ifconfig bridge101 10.11.11.2 then the VM doesn't answer to ping anymore.
This feels like a strange way to overcome what I think is Apple's dumb attempts to sandbox everything. I want my old TAP/TUN kexts, I want my VMWare/Parallels virtual interfaces to show in System Preferences... I don't want an overpriced non-touch iPad.
I'm now forced to try with qemu/Fusion for ARM to overcome this, and I'll get back to you!

 

I am also seeing the same issue with an Ubuntu 22 VM under Parallels when trying to transfer UDP/multicast packets between the host and the VM. @FrancoisB13 Did you have any luck with other VM managers as you mentioned? And now UTM may be an option besides Fusion.