From: "Ananyev, Konstantin" <konstantin.ananyev@intel.com>
To: Shahaf Shuler <shahafs@mellanox.com>, "dev@dpdk.org" <dev@dpdk.org>
Subject: Re: [dpdk-dev] [RFC PATCH 4/4] ethdev: add helpers to move to the new offloads API
Date: Mon, 28 Aug 2017 14:12:13 +0000 [thread overview]
Message-ID: <2601191342CEEE43887BDE71AB9772584F23D376@IRSMSX103.ger.corp.intel.com> (raw)
In-Reply-To: <VI1PR05MB31490A40941A48615FC62146C3850@VI1PR05MB3149.eurprd05.prod.outlook.com>
> -----Original Message-----
> From: Shahaf Shuler [mailto:shahafs@mellanox.com]
> Sent: Wednesday, August 23, 2017 2:13 PM
> To: Ananyev, Konstantin <konstantin.ananyev@intel.com>; dev@dpdk.org
> Subject: RE: [dpdk-dev] [RFC PATCH 4/4] ethdev: add helpers to move to the new offloads API
>
> Wednesday, August 23, 2017 3:29 PM, Ananyev, Konstantin:
> >
> >
> > > -----Original Message-----
> > > From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Shahaf Shuler
> > > Sent: Monday, August 7, 2017 1:55 PM
> > > To: dev@dpdk.org
> > > Subject: [dpdk-dev] [RFC PATCH 4/4] ethdev: add helpers to move to the
> > > new offloads API
> > >
> > > A new offloads API was introduced by commits:
> > >
> > > commit 8b07fcae6061 ("ethdev: introduce Tx queue offloads API") commit
> > > c6504557763e ("ethdev: introduce Rx queue offloads API")
> > >
> > > In order to enable PMDs to support only one of the APIs, and
> > > applications to avoid branching according to the underlying device a
> > > copy functions to/from the old/new APIs were added.
> > >
> > > Signed-off-by: Shahaf Shuler <shahafs@mellanox.com>
> > > ---
> > > lib/librte_ether/rte_ethdev.c | 140
> > > +++++++++++++++++++++++++++++++++++++
> > > 1 file changed, 140 insertions(+)
> > >
> > > diff --git a/lib/librte_ether/rte_ethdev.c
> > > b/lib/librte_ether/rte_ethdev.c index f73307e99..2b4a28c97 100644
> > > --- a/lib/librte_ether/rte_ethdev.c
> > > +++ b/lib/librte_ether/rte_ethdev.c
> > > @@ -1003,6 +1003,63 @@ rte_eth_dev_close(uint8_t port_id)
> > > dev->data->tx_queues = NULL;
> > > }
> > >
> > > +/**
> > > + * A copy function from rxmode offloads API to rte_eth_rxq_conf
> > > + * offloads API, to enable PMDs to support only one of the APIs.
> > > + */
> > > +static void
> > > +rte_eth_copy_rxmode_offloads(struct rte_eth_rxmode *rxmode,
> > > + struct rte_eth_rxq_conf *rxq_conf) {
> > > + if (rxmode->header_split == 1)
> > > + rxq_conf->offloads |= DEV_RX_OFFLOAD_HEADER_SPLIT;
> > > + if (rxmode->hw_ip_checksum == 1)
> > > + rxq_conf->offloads |= DEV_RX_OFFLOAD_CHECKSUM;
> > > + if (rxmode->hw_vlan_filter == 1)
> > > + rxq_conf->offloads |= DEV_RX_OFFLOAD_VLAN_FILTER;
> > > + if (rxmode->hw_vlan_strip == 1)
> > > + rxq_conf->offloads |= DEV_RX_OFFLOAD_VLAN_STRIP;
> > > + if (rxmode->hw_vlan_extend == 1)
> > > + rxq_conf->offloads |= DEV_RX_OFFLOAD_VLAN_EXTEND;
> > > + if (rxmode->jumbo_frame == 1)
> > > + rxq_conf->offloads |= DEV_RX_OFFLOAD_JUMBO_FRAME;
> > > + if (rxmode->hw_strip_crc == 1)
> > > + rxq_conf->offloads |= DEV_RX_OFFLOAD_CRC_STRIP;
> > > + if (rxmode->enable_scatter == 1)
> > > + rxq_conf->offloads |= DEV_RX_OFFLOAD_SCATTER;
> > > + if (rxmode->enable_lro == 1)
> > > + rxq_conf->offloads |= DEV_RX_OFFLOAD_LRO; }
> > > +
> > > +/**
> > > + * A copy function between rte_eth_rxq_conf offloads API to rxmode
> > > + * offloads API, to enable application to be agnostic to the PMD
> > > +supported
> > > + * offload API.
> > > + */
> > > +static void
> > > +rte_eth_copy_rxq_offloads(struct rte_eth_rxmode *rxmode,
> > > + struct rte_eth_rxq_conf *rxq_conf) {
> > > + if (rxq_conf->offloads & DEV_RX_OFFLOAD_HEADER_SPLIT)
> > > + rxmode->header_split = 1;
> > > + if (rxq_conf->offloads & DEV_RX_OFFLOAD_CHECKSUM)
> > > + rxmode->hw_ip_checksum = 1;
> > > + if (rxq_conf->offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
> > > + rxmode->hw_vlan_filter = 1;
> > > + if (rxq_conf->offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
> > > + rxmode->hw_vlan_strip = 1;
> > > + if (rxq_conf->offloads & DEV_RX_OFFLOAD_VLAN_EXTEND)
> > > + rxmode->hw_vlan_extend = 1;
> > > + if (rxq_conf->offloads & DEV_RX_OFFLOAD_JUMBO_FRAME)
> > > + rxmode->jumbo_frame = 1;
> > > + if (rxq_conf->offloads & DEV_RX_OFFLOAD_CRC_STRIP)
> > > + rxmode->hw_strip_crc = 1;
> > > + if (rxq_conf->offloads & DEV_RX_OFFLOAD_SCATTER)
> > > + rxmode->enable_scatter = 1;
> > > + if (rxq_conf->offloads & DEV_RX_OFFLOAD_LRO)
> > > + rxmode->enable_lro = 1;
> > > +}
> > > +
> > > int
> > > rte_eth_rx_queue_setup(uint8_t port_id, uint16_t rx_queue_id,
> > > uint16_t nb_rx_desc, unsigned int socket_id, @@ -1083,6
> > > +1140,37 @@ rte_eth_rx_queue_setup(uint8_t port_id, uint16_t
> > rx_queue_id,
> > > if (rx_conf == NULL)
> > > rx_conf = &dev_info.default_rxconf;
> > >
> > > + if ((dev->data->dev_flags & RTE_ETH_DEV_RXQ_OFFLOAD) &&
> > > + (dev->data->dev_conf.rxmode.ignore == 0)) {
> > > + rte_eth_copy_rxmode_offloads(&dev->data-
> > >dev_conf.rxmode,
> > > + rx_conf);
> > > + } else if ((!(dev->data->dev_flags & RTE_ETH_DEV_RXQ_OFFLOAD))
> > &&
> > > + (dev->data->dev_conf.rxmode.ignore == 1)) {
> > > + int ret;
> > > + struct rte_eth_rxmode rxmode;
> > > +
> > > + rte_eth_copy_rxq_offloads(&rxmode, rx_conf);
> > > + if (memcmp(&rxmode, &dev->data->dev_conf.rxmode,
> > > + sizeof(rxmode))) {
> > > + /*
> > > + * device which work with rxmode offloads API
> > requires
> > > + * a re-configuration in order to apply the new
> > offloads
> > > + * configuration.
> > > + */
> > > + dev->data->dev_conf.rxmode = rxmode;
> > > + ret = rte_eth_dev_configure(port_id,
> > > + dev->data->nb_rx_queues,
> > > + dev->data->nb_tx_queues,
> > > + &dev->data->dev_conf);
> >
> >
> > Hmm, and why we would need to reconfigure our device in the middle of rx
> > queue setup?
>
> The reason is the old Rx offloads API is configured on device configure.
> This if section is for applications which already moved to the new offload API however the underlying PMD still uses the old one.
Ok, but as I remember, right now the initialization order is pretty strict:
rx_queue_setup() has to be always called after dev_configure().
One of the reasons for that: rx_queue_setup() might change fileds inside dev->data->dev_private.
Second call for dev_configure() will void these changes and some of rxq config information will be lost.
So I think we should avoid calling dev_configure() inside rx_queue_setup().
My preference still would be to force all PMDs to move to the new version of rx_queue_setup() first.
I think it would be much more error prone then supporting two flavors of PMD config
and will allow us to catch errors early - in case this new scheme doesn't work by some PMD for any reason.
Also it seems that you forgot to:
struct rte_eth_rxmode rxmode = dev->data->dev_conf.rxmode;
Konstantin
>
> >
> > > + if (ret < 0) {
> > > + RTE_PMD_DEBUG_TRACE(
> > > + "unable to re-configure port %d "
> > > + "in order to apply rxq offloads "
> > > + "configuration\n", port_id);
> > > + }
> > > + }
> > > + }
> > > +
> > > ret = (*dev->dev_ops->rx_queue_setup)(dev, rx_queue_id,
> > nb_rx_desc,
> > > socket_id, rx_conf, mp);
> >
> > BTW, I don't see changes in any PMD for new offload flags?
> > Is it because it is just a n RFC and full patch would contain such changes?
>
> Yes this is because this is an RFC.
>
> The full patch I intend will move all examples and testpmd to the new offloads API.
> In addition it will include the mlx5 PMD support for the new offloads API.
>
> As I said on previous mail, I believe that the work to move the different PMDs to the new API should be done by their developers or
> maintainers.
>
> >
> >
> > > if (!ret) {
> > > @@ -1094,6 +1182,51 @@ rte_eth_rx_queue_setup(uint8_t port_id,
> > uint16_t rx_queue_id,
> > > return ret;
> > > }
> > >
> > > +/**
> > > + * A copy function from txq_flags to rte_eth_txq_conf offloads API,
> > > + * to enable PMDs to support only one of the APIs.
> > > + */
> > > +static void
> > > +rte_eth_copy_txq_flags(struct rte_eth_txq_conf *txq_conf) {
> > > + uint32_t txq_flags = txq_conf->txq_flags
> > > + uint64_t *offloads = &txq_conf->offloads;
> > > +
> > > + if (!(txq_flags & ETH_TXQ_FLAGS_NOMULTSEGS))
> > > + *offloads |= DEV_TX_OFFLOAD_MULTI_SEGS;
> > > + if (!(txq_flags & ETH_TXQ_FLAGS_NOVLANOFFL))
> > > + *offloads |= DEV_TX_OFFLOAD_VLAN_INSERT;
> > > + if (!(txq_flags & ETH_TXQ_FLAGS_NOXSUMSCTP))
> > > + *offloads |= DEV_TX_OFFLOAD_SCTP_CKSUM;
> > > + if (!(txq_flags & ETH_TXQ_FLAGS_NOXSUMUDP))
> > > + *offloads |= DEV_TX_OFFLOAD_UDP_CKSUM;
> > > + if (!(txq_flags & ETH_TXQ_FLAGS_NOXSUMTCP))
> > > + *offloads |= DEV_TX_OFFLOAD_TCP_CKSUM; }
> > > +
> > > +/**
> > > + * A copy function between rte_eth_txq_conf offloads API to txq_flags
> > > + * offloads API, to enable application to be agnostic to the PMD
> > > +supported
> > > + * API.
> > > + */
> > > +static void
> > > +rte_eth_copy_txq_offloads(struct rte_eth_txq_conf *txq_conf) {
> > > + uint32_t *txq_flags = &txq_conf->txq_flags
> > > + uint64_t offloads = txq_conf->offloads;
> > > +
> > > + if (!(offloads & DEV_TX_OFFLOAD_MULTI_SEGS))
> > > + *txq_flags |= ETH_TXQ_FLAGS_NOMULTSEGS;
> > > + if (!(offloads & DEV_TX_OFFLOAD_VLAN_INSERT))
> > > + *txq_flags |= ETH_TXQ_FLAGS_NOVLANOFFL;
> > > + if (!(offloads & DEV_TX_OFFLOAD_SCTP_CKSUM))
> > > + *txq_flags |= ETH_TXQ_FLAGS_NOXSUMSCTP;
> > > + if (!(offloads & DEV_TX_OFFLOAD_UDP_CKSUM))
> > > + *txq_flags |= ETH_TXQ_FLAGS_NOXSUMUDP;
> > > + if (!(offloads & DEV_TX_OFFLOAD_TCP_CKSUM))
> > > + *txq_flags |= ETH_TXQ_FLAGS_NOXSUMTCP; }
> > > +
> > > int
> > > rte_eth_tx_queue_setup(uint8_t port_id, uint16_t tx_queue_id,
> > > uint16_t nb_tx_desc, unsigned int socket_id, @@ -1145,6
> > > +1278,13 @@ rte_eth_tx_queue_setup(uint8_t port_id, uint16_t
> > tx_queue_id,
> > > if (tx_conf == NULL)
> > > tx_conf = &dev_info.default_txconf;
> > >
> > > + if ((dev->data->dev_flags & RTE_ETH_DEV_TXQ_OFFLOAD) &&
> > > + (!(tx_conf->txq_flags & ETH_TXQ_FLAGS_IGNORE)))
> > > + rte_eth_copy_txq_flags(tx_conf);
> > > + else if (!(dev->data->dev_flags & RTE_ETH_DEV_TXQ_OFFLOAD) &&
> > > + (tx_conf->txq_flags & ETH_TXQ_FLAGS_IGNORE))
> > > + rte_eth_copy_txq_offloads(tx_conf);
> > > +
> >
> > As I said in my previous mail - I think better to always convrert from old
> > txq_flags to new TX offload flags and make each PMD to understand new
> > offload values only.
> > Konstantin
> >
> > > return (*dev->dev_ops->tx_queue_setup)(dev, tx_queue_id,
> > nb_tx_desc,
> > > socket_id, tx_conf);
> > > }
> > > --
> > > 2.12.0
next prev parent reply other threads:[~2017-08-28 14:12 UTC|newest]
Thread overview: 34+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-08-07 10:54 [dpdk-dev] [RFC PATCH 0/4] ethdev " Shahaf Shuler
2017-08-07 10:54 ` [dpdk-dev] [RFC PATCH 1/4] ethdev: rename Rx and Tx configuration structs Shahaf Shuler
2017-08-23 21:39 ` Thomas Monjalon
2017-08-07 10:54 ` [dpdk-dev] [RFC PATCH 2/4] ethdev: introduce Rx queue offloads API Shahaf Shuler
2017-08-23 12:21 ` Ananyev, Konstantin
2017-08-23 13:06 ` Shahaf Shuler
2017-08-23 21:48 ` Thomas Monjalon
2017-08-29 12:50 ` Ferruh Yigit
2017-08-30 6:22 ` Shahaf Shuler
2017-08-29 13:11 ` Ferruh Yigit
2017-08-07 10:54 ` [dpdk-dev] [RFC PATCH 3/4] ethdev: introduce Tx " Shahaf Shuler
2017-08-07 10:54 ` [dpdk-dev] [RFC PATCH 4/4] ethdev: add helpers to move to the new " Shahaf Shuler
2017-08-23 12:28 ` Ananyev, Konstantin
2017-08-23 13:13 ` Shahaf Shuler
2017-08-23 22:06 ` Thomas Monjalon
2017-08-24 7:12 ` Shahaf Shuler
2017-08-25 13:26 ` Thomas Monjalon
2017-08-29 12:55 ` Ferruh Yigit
2017-08-30 6:30 ` Shahaf Shuler
2017-08-30 7:50 ` Ferruh Yigit
2017-08-30 10:16 ` Ananyev, Konstantin
2017-08-30 12:42 ` Ferruh Yigit
2017-08-30 13:25 ` Thomas Monjalon
2017-08-30 14:15 ` Ananyev, Konstantin
2017-08-28 14:12 ` Ananyev, Konstantin [this message]
2017-08-29 6:26 ` Shahaf Shuler
2017-08-29 9:43 ` Ananyev, Konstantin
2017-08-23 6:39 ` [dpdk-dev] [RFC PATCH 0/4] ethdev " Shahaf Shuler
2017-08-23 22:16 ` Thomas Monjalon
2017-08-25 10:31 ` Jerin Jacob
2017-08-27 6:05 ` Shahaf Shuler
2017-08-28 5:00 ` Jerin Jacob
2017-08-28 10:57 ` Thomas Monjalon
2017-09-05 7:07 ` Jerin Jacob
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=2601191342CEEE43887BDE71AB9772584F23D376@IRSMSX103.ger.corp.intel.com \
--to=konstantin.ananyev@intel.com \
--cc=dev@dpdk.org \
--cc=shahafs@mellanox.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).