DPDK-dev Archive on lore.kernel.org
 help / color / mirror / Atom feed
From: Max Makarov <makarov@volta.cloud>
To: dev@dpdk.org
Cc: igozlan@nvidia.com, valex@nvidia.com, erezsh@nvidia.com,
	suanmingm@nvidia.com, bingz@nvidia.com, kliteyn@nvidia.com
Subject: [QUESTION] mlx5/hws: format_select_dw_8_6_ext gating on standalone ConnectX-7
Date: Tue, 19 May 2026 01:44:16 +0900	[thread overview]
Message-ID: <177912265651.29998.6928116123708055888@volta.cloud> (raw)

Hi all,

We are building an SDN/IaaS data plane on ConnectX-7 using DOCA Flow and
need IPv6 Connection Tracking. CT pipe for IPv6 requires the 11-DW jumbo
STE format, which is gated by HCA_CAP_GENERAL_2.format_select_dw_8_6_ext
(full_dw_jumbo_support in drivers/net/mlx5/hws/mlx5dr_cmd.c).

On all three of our standalone ConnectX-7 SKUs (PSIDs MT_0000000838,
MT_0000000840, MT_0000000892, firmware 28.48.1000) this bit reads as 0 in
BOTH GET_CUR and GET_MAX modes. We have verified the bit is not affected
by any documented NV-config TLV (tested ~20 candidates with mlxfwreset).

Questions:

  1. Is this capability silicon-locked on standalone ConnectX-7, or could
     a future firmware release expose it?

  2. The DOCA Flow CT documentation states "BlueField-3 and above is
     required for IPv6". Is this permanent product positioning, or is
     there a roadmap to enable it on CX-7?

  3. For existing CX-7 deployments that need IPv6 stateful CT, is there
     a recommended DOCA Flow pattern (e.g. splitting the 5-tuple match
     across multiple 8-DW STEs while preserving CT semantics)?

Happy to share full HCA_CAP dumps or any diagnostic output that would
help.

Thanks,
Max Makarov
Volta Cloud

                 reply	other threads:[~2026-05-19  7:07 UTC|newest]

Thread overview: [no followups] expand[flat|nested]  mbox.gz  Atom feed

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=177912265651.29998.6928116123708055888@volta.cloud \
    --to=makarov@volta.cloud \
    --cc=bingz@nvidia.com \
    --cc=dev@dpdk.org \
    --cc=erezsh@nvidia.com \
    --cc=igozlan@nvidia.com \
    --cc=kliteyn@nvidia.com \
    --cc=suanmingm@nvidia.com \
    --cc=valex@nvidia.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox