RE: [PATCH net v5 1/4] igb: do not let XDP frames span more than one Rx buffer

[Tjerk Kusters]

> - [High] On i354 with VFs allocated, igb_set_rlpml()
>   (drivers/net/ethernet/intel/igb/igb_main.c) programs no receive-length…

Right: i354 can have VFs, and igb_set_rlpml() skipped the PF pool's
VMOLR there, so neither RLPML, which stays at 9728 with VFs, nor VMOLR
limited the length to the frame size. In v6 the PF pool limit is also
written on i354 when VFs are allocated. This also covers the same point
in the reviews of 3/4 and 4/4. I have no i354; I checked the register
values with an I350 forced onto the i354 VMOLR branches.

The comment conflict reported on 3/4 is fixed: the comment in
igb_xsk_pool_enable() no longer says that igb_change_mtu() keeps the
frame size check true, it says that the size is checked again on MTU
changes and when the ring is set up. Patch 4 is dropped in v6, so the
pre-existing zero-copy refill issue reported on 4/4 no longer interacts
with this series. It is a real issue on its own; I listed it in the v6
cover letter as a separate zero-copy problem.

pw-bot: cr