RE: [PATCH net v5 3/4] igb: do not split AF_XDP zero-copy frames over several Rx buffers

[Tjerk Kusters]

> I would rather return -EBUSY when enabling VFs on queues with active
> xsk_pool. Seems igb sort of steal queues from PF?

Yes. With VFs allocated, the PF keeps a single Tx queue, and one Rx
queue on i350 or two on 82576, so a pool bound to a higher queue loses
its queue. In v6 the VF enable returns -EBUSY while an XSK pool is bound
to a queue the PF would lose, as the core refuses ethtool -L in the same
case. igb_xsk_pool() keeps its frame size check only for a queue that is
hidden in another way, such as after a fallback from MSI-X to MSI, so the
NIC cannot write past the XSK frame when that queue comes back.