Linux Documentation
 help / color / mirror / Atom feed
From: <evostrov@amazon.com>
To: <netdev@vger.kernel.org>
Cc: Evgeny Ostrovsky <evostrov@amazon.com>,
	Jakub Kicinski <kuba@kernel.org>,
	David Arinzon <darinzon@amazon.com>,
	Shay Agroskin <shayagr@amazon.com>,
	Arthur Kiyanovski <akiyano@amazon.com>,
	Gal Pressman <gal@nvidia.com>,
	Michal Kubiak <michal.kubiak@intel.com>,
	Paolo Abeni <pabeni@redhat.com>,
	Eric Dumazet <edumazet@google.com>,
	Kuniyuki Iwashima <kuniyu@amazon.com>,
	Shai Brandes <shaibran@amazon.com>, <linux-doc@vger.kernel.org>
Subject: [RFC] net: ena: use tx-push-buf-len while tx-push is off
Date: Wed, 18 Nov 2026 04:12:37 +0000	[thread overview]
Message-ID: <20261118041238.14396-1-evostrov@amazon.com> (raw)

From: Evgeny Ostrovsky <evostrov@amazon.com>

We want to confirm one point before sending the driver patches:
we would like to allow setting the tx-push-buf-len when tx-push is off,
but the name is problematic so we are not sure what will be the correct way
to achieve this.

Current state:
- ethtool has tx-push and tx-push-buf-len.
- tx-push on means the driver writes the packet header into device memory.
- tx-push-buf-len is the length of that header.
- Today the driver uses tx-push-buf-len only when tx-push is on.

We are adding a new LLQ mode to ENA: pull mode.
- In pull mode the header stays in host memory and the device reads it.
- So tx-push is off.
- We still need to tell the device how many header bytes to pull.
- We want to use tx-push-buf-len for this, with tx-push off.

ENA is the only in-tree driver that uses tx-push-buf-len, so this does
not change any other driver.

Example on a device that supports pull mode:

ethtool -G <dev> tx-push off tx-push-buf-len 224
ethtool -g <dev> -> TX Push: off, TX push buff len: 224

The core already allows this. net/ethtool/rings.c checks only the
capability bit and that the length is not above the max. It does not tie
tx-push-buf-len to tx-push.

In a 2023 review Gal Pressman suggested using tx-push-buf-len only when
tx-push is on. That was never added as code or written as a rule.
Link: https://lore.kernel.org/netdev/30cb3990-80ce-ca07-6d73-cdc00d0ad7b8@nvidia.com/

Request: Please confirm this is fine or if not can you please suggest another
way of achieving the pull mode len setup functionality.

Thank you

                 reply	other threads:[~2026-11-18  4:12 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=20261118041238.14396-1-evostrov@amazon.com \
    --to=evostrov@amazon.com \
    --cc=akiyano@amazon.com \
    --cc=darinzon@amazon.com \
    --cc=edumazet@google.com \
    --cc=gal@nvidia.com \
    --cc=kuba@kernel.org \
    --cc=kuniyu@amazon.com \
    --cc=linux-doc@vger.kernel.org \
    --cc=michal.kubiak@intel.com \
    --cc=netdev@vger.kernel.org \
    --cc=pabeni@redhat.com \
    --cc=shaibran@amazon.com \
    --cc=shayagr@amazon.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