DPDK patches and discussions
 help / color / mirror / Atom feed
From: "Yu, DapengX" <dapengx.yu@intel.com>
To: "Li, Xiaoyun" <xiaoyun.li@intel.com>,
	"Yigit, Ferruh" <ferruh.yigit@intel.com>,
	"Zhang, Qi Z" <qi.z.zhang@intel.com>
Cc: "dev@dpdk.org" <dev@dpdk.org>, "stable@dpdk.org" <stable@dpdk.org>
Subject: Re: [dpdk-dev] [dpdk-stable] [PATCH] app/testpmd: fix queue Rx and Tx offload reconfig cmd
Date: Fri, 9 Apr 2021 10:29:00 +0000	[thread overview]
Message-ID: <DM5PR11MB140176B51C73A291FBEECF888C739@DM5PR11MB1401.namprd11.prod.outlook.com> (raw)
In-Reply-To: <DM4PR11MB55346DA6840DC86B6147677C99739@DM4PR11MB5534.namprd11.prod.outlook.com>



> -----Original Message-----
> From: Li, Xiaoyun <xiaoyun.li@intel.com>
> Sent: Friday, April 9, 2021 3:50 PM
> To: Yu, DapengX <dapengx.yu@intel.com>; Yigit, Ferruh
> <ferruh.yigit@intel.com>; Zhang, Qi Z <qi.z.zhang@intel.com>
> Cc: dev@dpdk.org; stable@dpdk.org
> Subject: RE: [dpdk-stable] [PATCH] app/testpmd: fix queue Rx and Tx offload
> reconfig cmd
> 
> 
> 
> > -----Original Message-----
> > From: Yu, DapengX <dapengx.yu@intel.com>
> > Sent: Friday, April 9, 2021 13:25
> > To: Yigit, Ferruh <ferruh.yigit@intel.com>; Li, Xiaoyun
> > <xiaoyun.li@intel.com>; Zhang, Qi Z <qi.z.zhang@intel.com>
> > Cc: dev@dpdk.org; stable@dpdk.org
> > Subject: RE: [dpdk-stable] [PATCH] app/testpmd: fix queue Rx and Tx
> > offload reconfig cmd
> >
> >
> >
> > > -----Original Message-----
> > > From: Yigit, Ferruh <ferruh.yigit@intel.com>
> > > Sent: Thursday, April 8, 2021 11:42 PM
> > > To: Yu, DapengX <dapengx.yu@intel.com>; Li, Xiaoyun
> > > <xiaoyun.li@intel.com>; Zhang, Qi Z <qi.z.zhang@intel.com>
> > > Cc: dev@dpdk.org; stable@dpdk.org
> > > Subject: Re: [dpdk-stable] [PATCH] app/testpmd: fix queue Rx and Tx
> > > offload reconfig cmd
> > >
> > > On 4/1/2021 9:28 AM, dapengx.yu@intel.com wrote:
> > > > From: Dapeng Yu <dapengx.yu@intel.com>
> > > >
> > > > Configure per queue rx offloading and per queue tx offloading
> > > > command shouldn't trigger the rte_eth_dev_configure() to
> > > > reconfigure
> > device.
> > > >
> > > > The patch sets the queue reconfiguration flag only, and does not
> > > > set the device reconfiguration flag. Therefore after port is
> > > > restarted,
> > > > rte_eth_dev_configure() will not be called again.
> > > >
> > >
> > > Just to clarify the impact, was calling 'rte_eth_dev_configure()'
> > > causing any problem, is this fixing any issue?
> > > Or is this patch an optimization to eliminate an unnecessary call?
> > >
> > This patch does fix an issue, and it also eliminates an unnecessary call.
> >
> > The issue is:
> > per-queue configuration, for example: port 0 rxq 0 rx_offload
> > jumbo_frame off triggers the per-device configuration change: the RSS
> > key is reconfigured and changes after rte_eth_dev_configure() is
> > called on ICE PMD driver, that cause a test case failure.
> >
> 
> Hmmm. I agree on the following. It doesn't need dev_configure. That's why I
> give you ack.
> 
> But your issue. Shouldn't you fix the driver? The vsi->rss_key[] just updated
> itself as a random value when dev_conf doesn't contain one.
> But your case is dev_conf doesn't contain new rss key, but vsi->rss_key is not
> null. In this case, it should keep the same not get a new random key.
> 
In current implementation, dev->data->dev_conf.rx_adv_conf.rss_conf in PMD
does not hold the rss_key value if user does not set it in the input parameter of rte_eth_dev_configure(). 

Even if PMD generate one automatically, it will not be saved in dev->data->dev_conf.rx_adv_conf.rss_conf. 

When user want to get rss_key, it will be retrieved on the fly from hardware, but not from any variable
in PMD.

So PMD (ice and i40e) think only rss_key  which is set by user via rte_eth_dev_configure() can be reused
when port is reconfigured.

If it is not present, PMD will generate one by itself anyway even if it is present in vsi->rss_key.

I don’t think this behavior is wrong, so did not fix ice PMD.

> > There is an unnecessary call in original implementation because both
> > cmd_config_per_queue_rx_offload_parsed() and
> > cmd_config_per_queue_tx_offload_parsed()
> > does not update the "port->dev_conf" which hold the port
> > configuration, therefore there is no need to call rte_eth_dev_configure().


  reply	other threads:[~2021-04-09 10:29 UTC|newest]

Thread overview: 13+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-04-01  8:28 [dpdk-dev] " dapengx.yu
2021-04-07  2:30 ` Li, Xiaoyun
2021-04-08 15:41 ` [dpdk-dev] [dpdk-stable] " Ferruh Yigit
2021-04-09  5:25   ` Yu, DapengX
2021-04-09  7:50     ` Li, Xiaoyun
2021-04-09 10:29       ` Yu, DapengX [this message]
2021-04-12  2:21         ` Li, Xiaoyun
2021-04-12 12:46           ` Ferruh Yigit
2021-04-15  6:04             ` Yu, DapengX
2021-04-19 15:46               ` Ferruh Yigit
2021-04-20  8:12                 ` Yu, DapengX
2021-04-20  8:22                   ` Ferruh Yigit
2021-04-20  8:40                     ` Yu, DapengX

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=DM5PR11MB140176B51C73A291FBEECF888C739@DM5PR11MB1401.namprd11.prod.outlook.com \
    --to=dapengx.yu@intel.com \
    --cc=dev@dpdk.org \
    --cc=ferruh.yigit@intel.com \
    --cc=qi.z.zhang@intel.com \
    --cc=stable@dpdk.org \
    --cc=xiaoyun.li@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).