Re: [PATCH] net/bnxt: remove VMDq support

[Stephen Hemminger <stephen@networkplumber.org>]

On Sun, 20 Sep 2026 21:08:11 -0600
Mohammad Shuab Siddique <mohammad-shuab.siddique@broadcom.com> wrote:

> From: Keegan Freyhof <keegan.freyhof@broadcom.com>
> 
> Remove support for the VMDq feature from the driver: the multi-pool
> Rx-queue configuration path in bnxt_mq_rx_configure(), the VMDQ VLAN
> pool wiring in the L2 filter alloc path, and the max_vmdq_pools/
> vmdq_queue_num reporting in dev_info (now always reported as 0).
> 
> Preserved two pieces of logic that live in the same
> function/switch statement but are not VMDq-pool-specific: the
> RTE_ETH_MQ_RX_RSS pools=rx_cosq_cnt case and the
> per-non-default-VNIC skip_filter_allocation guard in
> bnxt_mq_rx_configure(), both used by the plain multi-COS RSS path.
> Also kept the RTE_ETH_VMDQ_ACCEPT_HASH_UC/BROADCAST/MULTICAST ->
> l2_rx_mask flag mapping in rte_pmd_bnxt_set_vf_rxmode(): despite the
> name, those bits are the standard DPDK VF-rxmode API used regardless
> of hardware VMDq pool support, not a VMDq-pool-specific control.
> 
> Signed-off-by: Keegan Freyhof <keegan.freyhof@broadcom.com>
> Signed-off-by: Mohammad Shuab Siddique <mohammad-shuab.siddique@broadcom.com>
> ---

[PATCH] net/bnxt: remove VMDq support
 
Warning: mq_mode validation is lost. Before, VMDq, DCB and DCB_RSS
modes hit the default case and returned -EINVAL. Now any mode other
than NONE or RSS silently configures a single pool and succeeds.
Reject every other mq_mode explicitly in bnxt_mq_rx_configure() or
dev_configure.