Minutes of Technical Board Meeting, 2024-November-26

Minutes of Technical Board Meeting, 2024-November-26

[Kevin Traynor]

Minutes of Technical Board Meeting, 2024-November-26

Members Attending
*****************
- Aaron
- Hemant
- Kevin (Chair)
- Bruce
- Thomas
- Morten
- Stephen
- Konstantin
- Jerin

NOTE: The technical board meetings every second Wednesday at
https://meet.jit.si/DPDK at 3 pm UTC.
Meetings are public, and DPDK community members are welcome to attend.
NOTE: Next meeting will be on Wednesday 2024- Dec - 10th @3pm UTC, and
will be chaired by Honnappa

# Lab SoW

- We got tech board quorum for voting for 2026 tasks.
- Patrick processed that into an SoW
- Governing board approved
- DTS development work being discussed

# Discussion of VLAN offload in PMD

- Discussion over patch for nbl driver that hard codes a return 0 for
vlan_offload_set callback
- https://inbox.dpdk.org/dev/20251124153545.7123c338@phoenix.local/
- vlan offload is done in software
- Issue is that existing OOT driver does this, so it would break apps vs
in-tree driver if it returned ENOSUP
- For current release, leave code as is
- For future release, consider to do some validation at ethdev level

# Driver Maintenance Discussion

- No updates

## Follow-up / Action Items

- Await update from Ajit regarding the new Broadcom maintainer
- Await Broadcom's plan for ongoing BNXT branch maintenance
- Discuss adding support-level flags to the maintainers file.
- Consider updates to documentation or policy for driver quality
expectations.
- Stephen to share proposal for marking drivers support-levels.

# AI
- Aaron ran some experiments in OVS using Claude code and Sonnet for
code review
- Results were better than CodeRabbit and Sourcery
- Aaron will select some DPDK patches to test with it

Re: Minutes of Technical Board Meeting, 2024-November-26

[Stephen Hemminger]

On Wed, 10 Dec 2025 15:01:50 +0000
Kevin Traynor <ktraynor@redhat.com> wrote:

> # Discussion of VLAN offload in PMD
> 
> - Discussion over patch for nbl driver that hard codes a return 0 forJ
> vlan_offload_set callback
> - https://inbox.dpdk.org/dev/20251124153545.7123c338@phoenix.local/
> - vlan offload is done in software
> - Issue is that existing OOT driver does this, so it would break apps vs
> in-tree driver if it returned ENOSUP
> - For current release, leave code as is
> - For future release, consider to do some validation at ethdev level

Just to follow up, the underlying problem here was me misreading the code.
The driver is doing VLAN offload in software which is fine (it is what virtio does).

It does raise the issue of maybe this should be done in ethdev layer so
all drivers that don't do HW offload could just be done in software by ethdev.
That would make the mbuf consistent across all drivers.