From: "Zhang, Qi Z" <qi.z.zhang@intel.com>
To: "Wang, Haiyue" <haiyue.wang@intel.com>,
"Yan, Zhirun" <zhirun.yan@intel.com>,
"dev@dpdk.org" <dev@dpdk.org>
Subject: Re: [dpdk-dev] [PATCH v1] net/i40e: remove redundant code in free queues.
Date: Tue, 11 Dec 2018 01:21:32 +0000 [thread overview]
Message-ID: <039ED4275CED7440929022BC67E70611532FC61C@SHSMSX103.ccr.corp.intel.com> (raw)
In-Reply-To: <E3B9F2FDCB65864C82CD632F23D8AB87732FD2F8@SHSMSX101.ccr.corp.intel.com>
> -----Original Message-----
> From: Wang, Haiyue
> Sent: Tuesday, December 11, 2018 1:00 AM
> To: Zhang, Qi Z <qi.z.zhang@intel.com>; Yan, Zhirun <zhirun.yan@intel.com>;
> dev@dpdk.org
> Subject: RE: [PATCH v1] net/i40e: remove redundant code in free queues.
>
> > -----Original Message-----
> > From: Zhang, Qi Z
> > Sent: Monday, December 10, 2018 22:08
> > To: Wang, Haiyue <haiyue.wang@intel.com>; Yan, Zhirun
> > <zhirun.yan@intel.com>; dev@dpdk.org
> > Subject: RE: [PATCH v1] net/i40e: remove redundant code in free queues.
> >
> >
> >
> > > -----Original Message-----
> > > From: Wang, Haiyue
> > > Sent: Monday, December 10, 2018 8:47 PM
> > > To: Zhang, Qi Z <qi.z.zhang@intel.com>; Yan, Zhirun
> > > <zhirun.yan@intel.com>; dev@dpdk.org
> > > Subject: RE: [PATCH v1] net/i40e: remove redundant code in free queues.
> > >
> > > > -----Original Message-----
> > > > From: Zhang, Qi Z
> > > > Sent: Monday, December 10, 2018 20:36
> > > > To: Yan, Zhirun <zhirun.yan@intel.com>; dev@dpdk.org; Wang, Haiyue
> > > > <haiyue.wang@intel.com>
> > > > Subject: RE: [PATCH v1] net/i40e: remove redundant code in free queues.
> > > >
> > > >
> > > >
> > > > > -----Original Message-----
> > > > > From: Yan, Zhirun
> > > > > Sent: Monday, December 10, 2018 9:32 PM
> > > > > To: dev@dpdk.org; Zhang, Qi Z <qi.z.zhang@intel.com>; Wang,
> > > > > Haiyue <haiyue.wang@intel.com>
> > > > > Cc: Yan, Zhirun <zhirun.yan@intel.com>
> > > > > Subject: [PATCH v1] net/i40e: remove redundant code in free queues.
> > > > >
> > > > > Before this patch, there are two functions that will clear RX/TX
> > > > > queues
> > > > > number: rte_eth_dev_close() and i40e_dev_free_queues().
> > > >
> > > > I agree with what the patch changed, but the commit log is not
> > > comprehensive.
> > > > I40e_dev_free_queues is not only in the path of rte_eth_dev_close
> > > > but also rte_eth_dev_reset.
> > > > So
> > > > For rte_eth_dev_close, its redundant because of duplication.
> > > > For rte_eth_dev_reset its redundant because of not necessary,
> > > > since following dev_configure is required after dev_reset and it
> > > > will be updated
> > > correctly.
> > > >
> > >
> > > How about this commit message ?
> > >
> > > net/i40e: queue management data set is out of the API scope.
> > >
> > > Function i40e_dev_free_queues is used to free specific i40e queue
> > > entries, but should not be used to set the upper level queue
> > > management data in 'struct rte_eth_dev_data', such as '
> > > nb_rx/tx_queues ', which are maintained by
> 'rte_eth_dev_rx/tx_queue_config'.
> >
> > Ideally it should be what you described , but if you look at exist
> > code in different PMD nb_rx/tx_queues is modified anywhere.
> > so I'm not very sure if nb_rx/tx_queues should only be modified in
> > rte_ethdev_rx/tx_queue_config is the original design, and I don't have
> > confidence those cleanup is always correct, that's why I will prefer
> > an explanation that prove this code clean is correct. I think your
> > comment is much fit for a patch set that do all the cleanup.
> >
> I searched the " dev->data->nb_rx_queues = 0 ", 10 PMDs drivers write like this,
> and these functions nearly look like the same, like *copy*. :)
that's not all the case, if you just search "nb_rx_queues = " you will get more cases that nb_rx_queues is modified by PMD directly.
We should make sure all of them can be cleanup with that way before have our statement, but this is beyond this patch's scope.
>
> Well, from the feature ' VIRTCHNL_OP_REQUEST_QUEUES ', it will cause the VF
> PMD internally reset after the application like testpmd calls the '
> rte_eth_dev_configure ' API.
> The internal reset has to free the internal resource firstly, which will call free
> functions such as ' i40e_dev_free_queues ', if this function reset the upper level
> parameters like nb_rx/tx_queues from ' rte_eth_dev_configure ', then this
> standard API will be changed to be more ugly, like:
Yes, that's why I also recommend Zhirun to merge this change to the patch for VIRTCHNL_OP_REQUEST_QUEUES as another option.
But as an independent patch, I think current comments should be OK, because the context here is we have no idea about VIRTCHNL_OP_REQUEST_QUEUES yet.
Again, the words in your comment
".... but should not be used to set the upper level queue management data in 'struct rte_eth_dev_data', such as 'nb_rx/tx_queues ', which are maintained by 'rte_eth_dev_rx/tx_queue_config"
I think it is a good design, but I'm not sure if this is considered at original because it is broken everywhere with current implementations.
So from my view, you are introducing a new design and should be covered in a separate patch set which focus on it, but this is not current patch's purpose.
>
> ' rte_eth_dev_configure '
> --> diag = (*dev->dev_ops->dev_configure)(dev);
> the PMD's dev_configure needs to be changed the new return value,
> such as 1 to
> tell the ' rte_eth_dev_configure ' reconfigures the nb_rx/tx_queues
> again, such as
> diag = (*dev->dev_ops->dev_configure)(dev);
> if (diag == 1) {
> rte_eth_dev_rx_queue_config(dev, nb_rx_q);
> rte_eth_dev_tx_queue_config(dev, nb_tx_q);
> } else if (diag != 0) {
> RTE_PMD_DEBUG_TRACE("port%d dev_configure = %d\n",
> port_id, diag);
> rte_eth_dev_rx_queue_config(dev, 0);
> rte_eth_dev_tx_queue_config(dev, 0);
> return eth_err(port_id, diag);
> }
> This makes VIRTCHNL_OP_REQUEST_QUEUES work, but patch will be 100%
> rejected.
> Finally, we found that it we made i40e a good PMD boy, then all will be happy.
> This is the long history.
>
> And if ' nb_rx/tx_queues ' stands for the number of TX/RX queues array entries,
> we'd better make code looks clean, like an OOP programming: let the
> rte_eth_xxx API handle its data.
>
> struct rte_eth_dev_data {
> char name[RTE_ETH_NAME_MAX_LEN]; /**< Unique identifier name */
>
> void **rx_queues; /**< Array of pointers to RX queues. */
> void **tx_queues; /**< Array of pointers to TX queues. */
> uint16_t nb_rx_queues; /**< Number of RX queues. */
> uint16_t nb_tx_queues; /**< Number of TX queues. */
>
> And the API 'void rte_eth_dev_close(uint16_t port_id)' will make things clean:
>
> /* old behaviour: only free queue arrays */
> dev->data->nb_rx_queues = 0;
> rte_free(dev->data->rx_queues);
> dev->data->rx_queues = NULL;
> dev->data->nb_tx_queues = 0;
> rte_free(dev->data->tx_queues);
> dev->data->tx_queues = NULL;
>
> > >
> > > > >
> > > > > This patch remove redundant code in i40e_dev_free_queues().
> > > > >
> > > > > Signed-off-by: Zhirun Yan <zhirun.yan@intel.com>
> > > > > Signed-off-by: Haiyue Wang <haiyue.wang@intel.com>
> > > > > ---
> > > > > drivers/net/i40e/i40e_rxtx.c | 2 --
> > > > > 1 file changed, 2 deletions(-)
> > > > >
> > > > > diff --git a/drivers/net/i40e/i40e_rxtx.c
> > > > > b/drivers/net/i40e/i40e_rxtx.c index
> > > > > e1152ff0e..cc953ad58 100644
> > > > > --- a/drivers/net/i40e/i40e_rxtx.c
> > > > > +++ b/drivers/net/i40e/i40e_rxtx.c
> > > > > @@ -2753,7 +2753,6 @@ i40e_dev_free_queues(struct rte_eth_dev
> *dev)
> > > > > i40e_dev_rx_queue_release(dev->data->rx_queues[i]);
> > > > > dev->data->rx_queues[i] = NULL;
> > > > > }
> > > > > - dev->data->nb_rx_queues = 0;
> > > > >
> > > > > for (i = 0; i < dev->data->nb_tx_queues; i++) {
> > > > > if (!dev->data->tx_queues[i]) @@ -2761,7 +2760,6 @@
> > > > > i40e_dev_free_queues(struct rte_eth_dev *dev)
> > > > > i40e_dev_tx_queue_release(dev->data->tx_queues[i]);
> > > > > dev->data->tx_queues[i] = NULL;
> > > > > }
> > > > > - dev->data->nb_tx_queues = 0;
> > > > > }
> > > > >
> > > > > #define I40E_FDIR_NUM_TX_DESC I40E_MIN_RING_DESC
> > > > > --
> > > > > 2.17.1
next prev parent reply other threads:[~2018-12-11 1:21 UTC|newest]
Thread overview: 10+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-12-10 13:32 Zhirun Yan
2018-12-10 12:36 ` Zhang, Qi Z
2018-12-10 12:47 ` Wang, Haiyue
2018-12-10 14:08 ` Zhang, Qi Z
2018-12-10 16:59 ` Wang, Haiyue
2018-12-11 1:21 ` Zhang, Qi Z [this message]
2018-12-11 2:28 ` Wang, Haiyue
2018-12-13 12:46 ` [dpdk-dev] [PATCH v2] net/i40e: remove redundant code Zhirun Yan
2018-12-13 12:49 ` Zhirun Yan
2018-12-13 9:57 ` Zhang, Qi Z
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=039ED4275CED7440929022BC67E70611532FC61C@SHSMSX103.ccr.corp.intel.com \
--to=qi.z.zhang@intel.com \
--cc=dev@dpdk.org \
--cc=haiyue.wang@intel.com \
--cc=zhirun.yan@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).