DPDK patches and discussions
 help / color / mirror / Atom feed
From: Adrien Mazarguil <adrien.mazarguil@6wind.com>
To: Qi Zhang <qi.z.zhang@intel.com>
Cc: dev@dpdk.org, wenzhuo.lu@intel.com, beilei.xing@intel.com
Subject: Re: [dpdk-dev] [RFC 0/2] ethdev: add new attribute for signature match
Date: Wed, 17 May 2017 12:32:41 +0200	[thread overview]
Message-ID: <20170517103241.GD1758@6wind.com> (raw)
In-Reply-To: <1494791406-3594-1-git-send-email-qi.z.zhang@intel.com>

On Sun, May 14, 2017 at 03:50:04PM -0400, Qi Zhang wrote:
> We try to enable ixgbe's signature match with rte_flow, but didn't
> find a way with current APIs, so the RFC propose to add a new flow
> attribute "sig_match" to indicate if current flow is "perfect match"
> or "signature match"
> With perfect match (by default), if a packet does not match pattern,
> actions will not be taken. (this is identical with current behavior)
> With signature match, if a packet does not match pattern, it still
> has the possibility to trigger the actions, this happens when device
> think the signature of the pattern is matched.
> Signature match is expected to have better performance than perfect
> match with the cost of accuracy.
> When a flow rule with this attribute set, identical behavior can ONLY
> be guaranteed if packet matches the pattern, since different device
> may have different implementation of signature calculation algorithm.
> Driver of device that does not support signature match is not required to
> return error, but just simply igore this attribute, because the default
>  "perfect match" still can be regarded as a speical case of 
> "signature match".
> 
> Qi Zhang (2):
>   rte_flow: add attribute for signature match
>   doc/guides/prog_guide: add new rte_flow attribute
> 
>  app/test-pmd/cmdline_flow.c        | 11 +++++++++++
>  doc/guides/prog_guide/rte_flow.rst | 12 ++++++++++++
>  lib/librte_ether/rte_flow.h        |  3 ++-
>  3 files changed, 25 insertions(+), 1 deletion(-)

As discussed offline, modifying struct rte_flow_attr for this purpose is not
ideal. We've agreed that a new meta pattern item should be defined instead,
as described in the FDIR rules conversion section (8.9.7) of the
documentation [1].

[1] http://dpdk.org/doc/guides/prog_guide/rte_flow.html#fdir-to-most-item-types-queue-drop-passthru

-- 
Adrien Mazarguil
6WIND

      parent reply	other threads:[~2017-05-17 10:32 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2017-05-14 19:50 Qi Zhang
2017-05-14 19:50 ` [dpdk-dev] [RFC 1/2] rte_flow: add " Qi Zhang
2017-05-14 19:50 ` [dpdk-dev] [RFC 2/2] doc/guides/prog_guide: add new flow attribute Qi Zhang
2017-05-16  9:11   ` Mcnamara, John
2017-05-17 10:32 ` Adrien Mazarguil [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=20170517103241.GD1758@6wind.com \
    --to=adrien.mazarguil@6wind.com \
    --cc=beilei.xing@intel.com \
    --cc=dev@dpdk.org \
    --cc=qi.z.zhang@intel.com \
    --cc=wenzhuo.lu@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).