DPDK patches and discussions
 help / color / mirror / Atom feed
From: "Gaëtan Rivet" <gaetan.rivet@6wind.com>
To: Wisam Jaddo <wisamm@mellanox.com>
Cc: jingjing.wu@intel.com, wenzhuo.lu@intel.com, thomas@monjalon.net,
	rasland@mellanox.com, dev@dpdk.org, shahafs@mellanox.com
Subject: Re: [dpdk-dev] [PATCH 0/2] app/testpmd: introduce new commands in forward topology
Date: Wed, 16 May 2018 16:55:34 +0200	[thread overview]
Message-ID: <20180516145534.3ynumxtgbw4iywff@bidouze.vm.6wind.com> (raw)
In-Reply-To: <1525869279-10035-1-git-send-email-wisamm@mellanox.com>

Hi,

On Wed, May 09, 2018 at 03:34:37PM +0300, Wisam Jaddo wrote:
> 1- Add new forward topology (custom): This toplogy will
> allow to create any topoloy you may need.
> 2- Add new command to set the custom toplogy.
> 3- Add new command to set the topoly on the fly.

I have had the same kind of issues when developping a PMD for a
prototype NIC that had some queues behaving weirdly, or when using
testpmd for regression testing with a incompatible topology that was
impossible to change. I think this use-case is important for testpmd and
this should be useful for many people.

I think however that your approach is not flexible enough.

One should not only be able to edit ports pairing, but also select
specific (Rx) queue. Additionally, it should be possible to change the
lcore forwarding a specific pair.

Testpmd uses the concept of streams internally. I think the only way to
have enough flexibility here is to expose this concept as a set of
commands to define custom streams, and then to bind those streams to
sets of lcores.

You have found after testing that TX queues cannot be shared. This tool is
also used for pure IO performance regression testing. Adding the support
for shared TX queues would impact this use-case and I think this is not
acceptable. At the very least, the default configuration should not be
impacted (and I think compile-time options are now frowned upon).

This is the reason it should be necessary to be able to define the set of
lcores doing the forwarding: The lcore defines the TX queue used by a stream,
this way the current testpmd arch is respected.

Regards,
-- 
Gaëtan Rivet
6WIND

      parent reply	other threads:[~2018-05-16 14:55 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2018-05-09 12:34 Wisam Jaddo
2018-05-09 12:34 ` [dpdk-dev] [PATCH 1/2] app/testpmd: add custom topology command Wisam Jaddo
2018-05-15 13:15   ` Kevin Traynor
2018-05-09 12:34 ` [dpdk-dev] [PATCH 2/2] app/testpmd: add set topology on the fly command Wisam Jaddo
2018-05-16 14:55 ` Gaëtan Rivet [this message]

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=20180516145534.3ynumxtgbw4iywff@bidouze.vm.6wind.com \
    --to=gaetan.rivet@6wind.com \
    --cc=dev@dpdk.org \
    --cc=jingjing.wu@intel.com \
    --cc=rasland@mellanox.com \
    --cc=shahafs@mellanox.com \
    --cc=thomas@monjalon.net \
    --cc=wenzhuo.lu@intel.com \
    --cc=wisamm@mellanox.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;
as well as URLs for NNTP newsgroup(s).