From: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
To: Pablo Neira Ayuso <pablo@netfilter.org>
Cc: Oz Shlomo <ozsh@nvidia.com>, Roi Dayan <roid@nvidia.com>,
Saeed Mahameed <saeed@kernel.org>,
"David S. Miller" <davem@davemloft.net>,
Jakub Kicinski <kuba@kernel.org>,
netdev@vger.kernel.org, Paul Blakey <paulb@nvidia.com>,
Saeed Mahameed <saeedm@nvidia.com>
Subject: Re: [net-next 08/15] net/mlx5e: CT: Preparation for offloading +trk+new ct rules
Date: Thu, 21 Jan 2021 23:16:18 -0300 [thread overview]
Message-ID: <20210122021618.GH3863@horizon.localdomain> (raw)
In-Reply-To: <20210122011834.GA25356@salvia>
On Fri, Jan 22, 2021 at 02:18:34AM +0100, Pablo Neira Ayuso wrote:
> Hi Oz,
>
> On Wed, Jan 20, 2021 at 06:09:48PM +0200, Oz Shlomo wrote:
> > On 1/14/2021 11:50 PM, Marcelo Ricardo Leitner wrote:
> > >
> > > Thoughts?
> > >
> >
> > I wonder if we should develop a generic mechanism to optimize CT software
> > for a use case that is faulty by design.
> > This has limited value for software as it would only reduce the conntrack
> > table size (packet classification is still required).
> > However, this feature may have a big impact on hardware offload.
> > Normally hardware offload relies on software to handle new connections.
> > Causing all new connections to be processed by software.
> > With this patch the hardware may autonomously set the +new connection state
> > for the relevant connections.
>
> Could you fix this issue with unidirectional flows by checking for
> IPS_CONFIRMED status bit? The idea is to hardware offload the entry
> after the first packet goes through software successfully. Then, there
> is no need to wait for the established state that requires to see
> traffic in both directions.
That's an interesting idea. This way, basically all that needs to be
changed is tcf_ct_flow_table_process_conn() to handle this new
condition for UDP packets and on tcf_ct_act().
It has a small performance penaulty if compared to the original
solution, as now the first packet(s) goes to sw, but looks like a good
compromise between supporting a (from what I could understand)
somewhat lazy flow design (as I still think these didn't need to go
through conntrack), an uniform system behavior (with and without
offload, with mlx5 or another driver) and a more generic approach.
Other situations that rely on unidirectional UDP flows will benefit
from it as well.
This way I even think it doesn't need to be configurable right now.
It will be easier to add a knob to switch back to the old behavior if
needed later on, if anything.
Marcelo
next prev parent reply other threads:[~2021-01-22 2:17 UTC|newest]
Thread overview: 30+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-01-08 5:30 [pull request][net-next 00/15] mlx5 updates 2021-01-07 Saeed Mahameed
2021-01-08 5:30 ` [net-next 01/15] net/mlx5: Add HW definition of reg_c_preserve Saeed Mahameed
2021-01-08 5:30 ` [net-next 02/15] net/mlx5e: Simplify condition on esw_vport_enable_qos() Saeed Mahameed
2021-01-08 5:30 ` [net-next 03/15] net/mlx5: E-Switch, use new cap as condition for mpls over udp Saeed Mahameed
2021-01-08 5:30 ` [net-next 04/15] net/mlx5e: E-Switch, Offload all chain 0 priorities when modify header and forward action is not supported Saeed Mahameed
2021-01-08 5:30 ` [net-next 05/15] net/mlx5e: CT: Pass null instead of zero spec Saeed Mahameed
2021-01-08 5:30 ` [net-next 06/15] net/mlx5e: Remove redundant initialization to null Saeed Mahameed
2021-01-08 5:30 ` [net-next 07/15] net/mlx5e: CT: Remove redundant usage of zone mask Saeed Mahameed
2021-01-08 5:30 ` [net-next 08/15] net/mlx5e: CT: Preparation for offloading +trk+new ct rules Saeed Mahameed
2021-01-08 21:48 ` Marcelo Ricardo Leitner
2021-01-10 7:45 ` Roi Dayan
2021-01-10 7:52 ` Roi Dayan
2021-01-11 23:51 ` Marcelo Ricardo Leitner
2021-01-12 9:27 ` Oz Shlomo
2021-01-14 13:02 ` Marcelo Ricardo Leitner
2021-01-14 14:03 ` Oz Shlomo
2021-01-14 21:50 ` Marcelo Ricardo Leitner
2021-01-20 16:09 ` Oz Shlomo
2021-01-22 1:18 ` Pablo Neira Ayuso
2021-01-22 2:16 ` Marcelo Ricardo Leitner [this message]
2021-01-25 9:15 ` Oz Shlomo
2021-01-08 5:30 ` [net-next 09/15] net/mlx5e: CT: Support offload of " Saeed Mahameed
2021-01-08 21:59 ` Marcelo Ricardo Leitner
2021-01-10 7:55 ` Roi Dayan
2021-01-08 5:30 ` [net-next 10/15] net/mlx5e: CT: Add support for mirroring Saeed Mahameed
2021-01-08 5:30 ` [net-next 11/15] net/mlx5e: CT, Avoid false lock depenency warning Saeed Mahameed
2021-01-08 5:30 ` [net-next 12/15] net/mlx5e: IPsec, Enclose csum logic under ipsec config Saeed Mahameed
2021-01-08 5:30 ` [net-next 13/15] net/mlx5e: IPsec, Avoid unreachable return Saeed Mahameed
2021-01-08 5:30 ` [net-next 14/15] net/mlx5e: IPsec, Inline feature_check fast-path function Saeed Mahameed
2021-01-08 5:30 ` [net-next 15/15] net/mlx5e: IPsec, Remove unnecessary config flag usage Saeed Mahameed
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=20210122021618.GH3863@horizon.localdomain \
--to=marcelo.leitner@gmail.com \
--cc=davem@davemloft.net \
--cc=kuba@kernel.org \
--cc=netdev@vger.kernel.org \
--cc=ozsh@nvidia.com \
--cc=pablo@netfilter.org \
--cc=paulb@nvidia.com \
--cc=roid@nvidia.com \
--cc=saeed@kernel.org \
--cc=saeedm@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