Since if_vmove passes ifp and pr, it feels more intuitive to compare the vnets from ifp and pr, rather than td and pr. That also goes back to the idea of an assertion than td and ifp are always in the same vnet.

The tricky part is that ifp is lacking proper locking, thus ifp->if_vnet is not stable. So it is possible that MAPSS(ifp->if_vnet == TD_TO_VNET(td) will fail.

The ifp is intitially obtained from current thread ( td ), but due to racing, unlink it from current "active" ifnet list may fail, see the following logic,

	found = if_unlink_ifnet(ifp, true);
	if (! found) {
		sx_xunlock(&ifnet_detach_sxlock);
		prison_free(pr);
		return (ENODEV);
	}

So a stable way to check if we are moving the iface from and to the same vnet is not refer to ifp->if_vnet, but TD_TO_VNET(td) .