Re: FreeBSD guest on Linux KVM, some vtnet(4) packets don't arrive

> On Apr 8, 2026, at 10:38 PM, Lexi Winter <ivy@FreeBSD.org> wrote:
> 
> hello,
> 
> i'm using FreeBSD main:
> 
> FreeBSD 16.0-CURRENT #0 main-n284768-e6eba5076929: Mon Mar 30 16:07:28 UTC 2026
> root@releng3.nyi.freebsd.org:/usr/obj/usr/src/powerpc.powerpc64/sys/GENERIC64
> 
> as a qemu (KVM) guest on a Linux host:
> 
> Linux 6.18.21-0-lts #1-Alpine SMP 2026-04-07 10:34:29 ppc64le
> 
> on the Linux side, qemu is configured to provide a tap interface to the
> guest:
> 
> qemu-system-ppc64 \
> 	-smp cpus=176,sockets=2,cores=22,threads=4 \
> 	-accel kvm -M pseries -m 8g -nographic -serial mon:stdio \
> 	-drive file=/vm/ppc64-main/root.img,format=raw,if=virtio \
> 	-netdev tap,id=net0,ifname=vm-ppc64-main,script=no,downscript=no \
> 	-device virtio-net-pci,netdev=net0
> 
> the tap interface looks fine:
> 
> 10: vm-ppc64-main: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UP group default qlen 1000
>    link/ether ae:dc:d8:39:be:fa brd ff:ff:ff:ff:ff:ff
>    inet6 fe80::acdc:d8ff:fe39:befa/64 scope link proto kernel_ll
>       valid_lft forever preferred_lft forever
>    inet6 fe80::1/64 scope link
>       valid_lft forever preferred_lft forever
> 
> on the FreeBSD side, the vtnet interface also looks fine:
> 
> vtnet0: flags=1008843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST,LOWER_UP> metric 0 mtu 1500
>        options=cc00ba<TXCSUM,VLAN_MTU,VLAN_HWTAGGING,JUMBO_MTU,VLAN_HWCSUM,VLAN_HWTSO,LINKSTATE,TXCSUM_IPV6,HWSTATS>
>        ether 52:54:00:12:34:56
>        inet6 fe80::2%vtnet0 prefixlen 64 scopeid 0x1
>        media: Ethernet autoselect (10Gbase-T <full-duplex>)
>        status: active
>        nd6 options=821<PERFORMNUD,AUTO_LINKLOCAL,STABLEADDR>
> 
> on the host, i can ping fe80::2%vm-ppc64-main:
> 
> PING fe80::2%vm-ppc64-main (fe80::2%10): 56 data bytes
> 64 bytes from fe80::2: seq=0 ttl=64 time=14.833 ms
> 64 bytes from fe80::2: seq=1 ttl=64 time=12.524 ms

So the traffic with link-local addresses looks good.

> 
> then i configured an IP address on lo0 on the guest:
> 
> lo0: flags=1008049<UP,LOOPBACK,RUNNING,MULTICAST,LOWER_UP> metric 0 mtu 16384
>        options=680003<RXCSUM,TXCSUM,LINKSTATE,RXCSUM_IPV6,TXCSUM_IPV6>
>        inet 127.0.0.1 netmask 0xff000000
>        inet6 ::1 prefixlen 128
>        inet6 fe80::1%lo0 prefixlen 64 scopeid 0x2
>        inet6 fd00:0:0:1::b prefixlen 128
>        groups: lo
>        nd6 options=21<PERFORMNUD,AUTO_LINKLOCAL>
> 
> a route exists for this on the host:
> 
> fd00:0:0:1::b via fe80::2 dev vm-ppc64-main proto bird metric 32 pref medium

Just to confirm, do you have a reverse route on the guest ?

> 
> when trying to ping this address, on the host, i can see ICMP packets leaving
> the tap interface:
> 
> 15:34:39.106898 IP6 fd00:0:0:1::12 > fd00:0:0:1::b: ICMP6, echo request, id 49664, seq 3, length 64
> 15:34:40.106951 IP6 fd00:0:0:1::12 > fd00:0:0:1::b: ICMP6, echo request, id 49664, seq 4, length 64
> 
> but these packets never arrive on the guest; tcpdump shows nothing.
> 
> however, 'netstat -in' shows that the packets *are* arriving on the guest,
> since Ipkts increments for each ICMP packet.  it seems like they're just
> getting lost somewhere in the network stack. (Ierrs/Oerrs is zero.)

That puzzled me a bit. If the net stack sees the arriving packets, then tcpdump shall
intercept them correctly.

> 
> does anyone have any ideas what might be causing this?

Does `netstat -sp ip6` and `netstat -sp icmp6` show anything unusual ?

Best regards,
Zhenlei