From: Adrien Mazarguil <adrien.mazarguil@6wind.com>
To: Ferruh Yigit <ferruh.yigit@intel.com>
Cc: dev@dpdk.org
Subject: Re: [dpdk-dev] [PATCH v2 00/51] net/mlx4: trim and refactor entire PMD
Date: Fri, 1 Sep 2017 13:56:29 +0200 [thread overview]
Message-ID: <20170901115629.GQ4301@6wind.com> (raw)
In-Reply-To: <2aceb1d3-6444-03af-d859-20d69b2e370d@intel.com>
On Fri, Sep 01, 2017 at 12:24:40PM +0100, Ferruh Yigit wrote:
> Hi Adrien,
>
> On 9/1/2017 9:06 AM, Adrien Mazarguil wrote:
> > The main purpose of this large series is to relieve the mlx4 PMD from its
> > dependency on Mellanox OFED to instead rely on the standard rdma-core
> > package provided by Linux distributions.
> >
> > While compatibility with Mellanox OFED is preserved, all nonstandard
> > functionality has to be stripped from the PMD in order to re-implement it
> > through an approach compatible with rdma-core.
> >
> > Due to the amount of changes necessary to achieve this goal, this rework
> > starts off by removing extraneous code to simplify the PMD as much as
> > possible before either replacing or dismantling functionality that relies on
> > nonstandard Verbs.
> >
> > What remains after applying this series is single-segment Tx/Rx support,
> > without offloads nor RSS, on the default MAC address (which cannot be
> > configured). Support for multiple queues and the flow API (minus the RSS
> > action) are also preserved.
> >
> > Missing functionality that needs substantial work will be restored later by
> > subsequent series.
>
> Thanks for comprehensive re-work, out of curiosity, is adding removed
> functionality planned for this release?
Yes, well for the most part. This includes TX/RX with enhanced performance
(several patches already on ML but I need to review them), scatter/gather,
checksum offloads, packet type recognition, RSS, MAC/VLAN/promisc/allmulti
filtering and related configuration.
Actually, only secondary process support might be missing from the next
release. Since we're separately working on a new approach for mlx5, we'll
see how well it performs before considering a similar change for mlx4.
>
> >
> > Also because the mlx4 PMD is mostly contained in a single very large source
> > file of 6400+ lines (mlx4.c) which has become extremely difficult to
> > maintain, this rework is used as an opportunity to finally group functions
> > into separate files, as in mlx5.
> >
> > This rework targets DPDK 17.11.
>
> <...>
>
--
Adrien Mazarguil
6WIND
next prev parent reply other threads:[~2017-09-01 11:56 UTC|newest]
Thread overview: 110+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-08-01 16:53 [dpdk-dev] [PATCH v1 00/48] " Adrien Mazarguil
2017-08-01 16:53 ` [dpdk-dev] [PATCH v1 01/48] net/mlx4: add consistency to copyright notices Adrien Mazarguil
2017-08-01 16:53 ` [dpdk-dev] [PATCH v1 02/48] net/mlx4: remove limitation on number of instances Adrien Mazarguil
2017-08-01 16:53 ` [dpdk-dev] [PATCH v1 03/48] net/mlx4: check max number of ports dynamically Adrien Mazarguil
2017-08-01 17:35 ` Legacy, Allain
2017-08-02 7:52 ` Adrien Mazarguil
2017-08-01 16:53 ` [dpdk-dev] [PATCH v1 04/48] net/mlx4: remove useless compilation checks Adrien Mazarguil
2017-08-18 13:39 ` Ferruh Yigit
2017-09-01 10:19 ` Adrien Mazarguil
2017-08-01 16:53 ` [dpdk-dev] [PATCH v1 05/48] net/mlx4: remove secondary process support Adrien Mazarguil
2017-08-01 16:53 ` [dpdk-dev] [PATCH v1 06/48] net/mlx4: remove useless code Adrien Mazarguil
2017-08-01 16:53 ` [dpdk-dev] [PATCH v1 07/48] net/mlx4: remove soft counters compilation option Adrien Mazarguil
2017-08-01 16:53 ` [dpdk-dev] [PATCH v1 08/48] net/mlx4: remove scatter mode " Adrien Mazarguil
2017-08-01 16:53 ` [dpdk-dev] [PATCH v1 09/48] net/mlx4: remove Tx inline " Adrien Mazarguil
2017-08-01 16:53 ` [dpdk-dev] [PATCH v1 10/48] net/mlx4: remove allmulti and promisc support Adrien Mazarguil
2017-08-01 16:53 ` [dpdk-dev] [PATCH v1 11/48] net/mlx4: remove VLAN filter support Adrien Mazarguil
2017-08-01 16:53 ` [dpdk-dev] [PATCH v1 12/48] net/mlx4: remove MAC address configuration support Adrien Mazarguil
2017-08-01 16:54 ` [dpdk-dev] [PATCH v1 13/48] net/mlx4: drop MAC flows affecting all Rx queues Adrien Mazarguil
2017-08-01 16:54 ` [dpdk-dev] [PATCH v1 14/48] net/mlx4: revert flow API RSS support Adrien Mazarguil
2017-08-01 16:54 ` [dpdk-dev] [PATCH v1 15/48] net/mlx4: revert RSS parent queue refactoring Adrien Mazarguil
2017-08-01 16:54 ` [dpdk-dev] [PATCH v1 16/48] net/mlx4: drop RSS support Adrien Mazarguil
2017-08-01 16:54 ` [dpdk-dev] [PATCH v1 17/48] net/mlx4: drop checksum offloads support Adrien Mazarguil
2017-08-01 16:54 ` [dpdk-dev] [PATCH v1 18/48] net/mlx4: drop packet type recognition support Adrien Mazarguil
2017-08-01 16:54 ` [dpdk-dev] [PATCH v1 19/48] net/mlx4: drop scatter/gather support Adrien Mazarguil
2017-08-01 16:54 ` [dpdk-dev] [PATCH v1 20/48] net/mlx4: drop inline receive support Adrien Mazarguil
2017-08-01 16:54 ` [dpdk-dev] [PATCH v1 21/48] net/mlx4: use standard QP attributes Adrien Mazarguil
2017-08-01 16:54 ` [dpdk-dev] [PATCH v1 22/48] net/mlx4: revert resource domain support Adrien Mazarguil
2017-08-01 16:54 ` [dpdk-dev] [PATCH v1 23/48] net/mlx4: revert multicast echo prevention Adrien Mazarguil
2017-08-01 16:54 ` [dpdk-dev] [PATCH v1 24/48] net/mlx4: revert fast Verbs interface for Tx Adrien Mazarguil
2017-08-01 16:54 ` [dpdk-dev] [PATCH v1 25/48] net/mlx4: revert fast Verbs interface for Rx Adrien Mazarguil
2017-08-01 16:54 ` [dpdk-dev] [PATCH v1 26/48] net/mlx4: simplify link update function Adrien Mazarguil
2017-08-01 16:54 ` [dpdk-dev] [PATCH v1 27/48] net/mlx4: standardize on negative errno values Adrien Mazarguil
2017-08-01 16:54 ` [dpdk-dev] [PATCH v1 28/48] net/mlx4: clean up coding style inconsistencies Adrien Mazarguil
2017-08-01 16:54 ` [dpdk-dev] [PATCH v1 29/48] net/mlx4: remove control path locks Adrien Mazarguil
2017-08-01 16:54 ` [dpdk-dev] [PATCH v1 30/48] net/mlx4: remove unnecessary wrapper functions Adrien Mazarguil
2017-08-01 16:54 ` [dpdk-dev] [PATCH v1 31/48] net/mlx4: remove mbuf macro definitions Adrien Mazarguil
2017-08-01 16:54 ` [dpdk-dev] [PATCH v1 32/48] net/mlx4: use standard macro to get array size Adrien Mazarguil
2017-08-01 16:54 ` [dpdk-dev] [PATCH v1 33/48] net/mlx4: separate debugging macros Adrien Mazarguil
2017-08-01 16:54 ` [dpdk-dev] [PATCH v1 34/48] net/mlx4: use a single interrupt handle Adrien Mazarguil
2017-08-01 16:54 ` [dpdk-dev] [PATCH v1 35/48] net/mlx4: rename alarm field Adrien Mazarguil
2017-08-01 16:54 ` [dpdk-dev] [PATCH v1 36/48] net/mlx4: refactor interrupt FD settings Adrien Mazarguil
2017-08-01 16:54 ` [dpdk-dev] [PATCH v1 37/48] net/mlx4: clean up interrupt functions prototypes Adrien Mazarguil
2017-08-01 16:54 ` [dpdk-dev] [PATCH v1 38/48] net/mlx4: compact interrupt functions Adrien Mazarguil
2017-08-01 16:54 ` [dpdk-dev] [PATCH v1 39/48] net/mlx4: separate interrupt handling Adrien Mazarguil
2017-08-01 16:54 ` [dpdk-dev] [PATCH v1 40/48] net/mlx4: separate Rx/Tx definitions Adrien Mazarguil
2017-08-01 16:54 ` [dpdk-dev] [PATCH v1 41/48] net/mlx4: separate Rx/Tx functions Adrien Mazarguil
2017-08-01 16:54 ` [dpdk-dev] [PATCH v1 42/48] net/mlx4: separate device control functions Adrien Mazarguil
2017-08-01 16:54 ` [dpdk-dev] [PATCH v1 43/48] net/mlx4: separate Tx configuration functions Adrien Mazarguil
2017-08-01 16:54 ` [dpdk-dev] [PATCH v1 44/48] net/mlx4: separate Rx " Adrien Mazarguil
2017-08-01 16:54 ` [dpdk-dev] [PATCH v1 45/48] net/mlx4: group flow API handlers in common file Adrien Mazarguil
2017-08-01 16:54 ` [dpdk-dev] [PATCH v1 46/48] net/mlx4: rename private functions in flow API Adrien Mazarguil
2017-08-01 16:54 ` [dpdk-dev] [PATCH v1 47/48] net/mlx4: separate memory management functions Adrien Mazarguil
2017-08-01 16:54 ` [dpdk-dev] [PATCH v1 48/48] net/mlx4: clean up includes and comments Adrien Mazarguil
2017-08-18 13:28 ` [dpdk-dev] [PATCH v1 00/48] net/mlx4: trim and refactor entire PMD Ferruh Yigit
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 00/51] " Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 01/51] net/mlx4: add consistency to copyright notices Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 02/51] net/mlx4: remove limitation on number of instances Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 03/51] net/mlx4: check max number of ports dynamically Adrien Mazarguil
2017-09-01 10:57 ` Legacy, Allain
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 04/51] net/mlx4: remove useless compilation checks Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 05/51] net/mlx4: remove secondary process support Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 06/51] net/mlx4: remove useless code Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 07/51] net/mlx4: remove soft counters compilation option Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 08/51] net/mlx4: remove scatter mode " Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 09/51] net/mlx4: remove Tx inline " Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 10/51] net/mlx4: remove allmulti and promisc support Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 11/51] net/mlx4: remove VLAN filter support Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 12/51] net/mlx4: remove MAC address configuration support Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 13/51] net/mlx4: drop MAC flows affecting all Rx queues Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 14/51] net/mlx4: revert flow API RSS support Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 15/51] net/mlx4: revert RSS parent queue refactoring Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 16/51] net/mlx4: drop RSS support Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 17/51] net/mlx4: drop checksum offloads support Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 18/51] net/mlx4: drop packet type recognition support Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 19/51] net/mlx4: drop scatter/gather support Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 20/51] net/mlx4: drop inline receive support Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 21/51] net/mlx4: use standard QP attributes Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 22/51] net/mlx4: revert resource domain support Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 23/51] net/mlx4: revert multicast echo prevention Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 24/51] net/mlx4: revert fast Verbs interface for Tx Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 25/51] net/mlx4: revert fast Verbs interface for Rx Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 26/51] net/mlx4: simplify Rx buffer handling Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 27/51] net/mlx4: simplify link update function Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 28/51] net/mlx4: standardize on negative errno values Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 29/51] net/mlx4: clean up coding style inconsistencies Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 30/51] net/mlx4: remove control path locks Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 31/51] net/mlx4: remove unnecessary wrapper functions Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 32/51] net/mlx4: remove mbuf macro definitions Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 33/51] net/mlx4: use standard macro to get array size Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 34/51] net/mlx4: separate debugging macros Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 35/51] net/mlx4: use a single interrupt handle Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 36/51] net/mlx4: rename alarm field Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 37/51] net/mlx4: refactor interrupt FD settings Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 38/51] net/mlx4: clean up interrupt functions prototypes Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 39/51] net/mlx4: compact interrupt functions Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 40/51] net/mlx4: separate interrupt handling Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 41/51] net/mlx4: separate Rx/Tx definitions Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 42/51] net/mlx4: separate Rx/Tx functions Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 43/51] net/mlx4: separate device control functions Adrien Mazarguil
2017-09-01 8:06 ` [dpdk-dev] [PATCH v2 44/51] net/mlx4: separate Tx configuration functions Adrien Mazarguil
2017-09-01 8:07 ` [dpdk-dev] [PATCH v2 45/51] net/mlx4: separate Rx " Adrien Mazarguil
2017-09-01 8:07 ` [dpdk-dev] [PATCH v2 46/51] net/mlx4: group flow API handlers in common file Adrien Mazarguil
2017-09-01 8:07 ` [dpdk-dev] [PATCH v2 47/51] net/mlx4: rename private functions in flow API Adrien Mazarguil
2017-09-01 8:07 ` [dpdk-dev] [PATCH v2 48/51] net/mlx4: separate memory management functions Adrien Mazarguil
2017-09-01 8:07 ` [dpdk-dev] [PATCH v2 49/51] net/mlx4: clean up includes and comments Adrien Mazarguil
2017-09-01 8:07 ` [dpdk-dev] [PATCH v2 50/51] net/mlx4: remove isolated mode constraint Adrien Mazarguil
2017-09-01 8:07 ` [dpdk-dev] [PATCH v2 51/51] net/mlx4: rely on ethdev for Tx/Rx queue arrays Adrien Mazarguil
2017-09-01 11:24 ` [dpdk-dev] [PATCH v2 00/51] net/mlx4: trim and refactor entire PMD Ferruh Yigit
2017-09-01 11:56 ` Adrien Mazarguil [this message]
2017-09-05 9:59 ` Ferruh Yigit
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=20170901115629.GQ4301@6wind.com \
--to=adrien.mazarguil@6wind.com \
--cc=dev@dpdk.org \
--cc=ferruh.yigit@intel.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).