From: "Wang, YuanX" <yuanx.wang@intel.com>
To: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>,
"dev@dpdk.org" <dev@dpdk.org>,
Thomas Monjalon <thomas@monjalon.net>,
Ferruh Yigit <ferruh.yigit@xilinx.com>
Cc: "mdr@ashroe.eu" <mdr@ashroe.eu>,
"Li, Xiaoyun" <xiaoyun.li@intel.com>,
"Singh, Aman Deep" <aman.deep.singh@intel.com>,
"Zhang, Yuying" <yuying.zhang@intel.com>,
"Zhang, Qi Z" <qi.z.zhang@intel.com>,
"Yang, Qiming" <qiming.yang@intel.com>,
"jerinjacobk@gmail.com" <jerinjacobk@gmail.com>,
"viacheslavo@nvidia.com" <viacheslavo@nvidia.com>,
"stephen@networkplumber.org" <stephen@networkplumber.org>,
"Ding, Xuan" <xuan.ding@intel.com>,
"hpothula@marvell.com" <hpothula@marvell.com>,
"Tang, Yaqi" <yaqi.tang@intel.com>,
Wenxuan Wu <wenxuanx.wu@intel.com>
Subject: RE: [PATCH v3 2/4] ethdev: introduce protocol hdr based buffer split
Date: Fri, 16 Sep 2022 08:38:44 +0000 [thread overview]
Message-ID: <PH7PR11MB6953C1B1B16DA6EB24AEB1EB85489@PH7PR11MB6953.namprd11.prod.outlook.com> (raw)
In-Reply-To: <761509b6-4a9a-72a1-2ee9-cd3555c85fd8@oktetlabs.ru>
Hi Andrew,
> -----Original Message-----
> From: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
> Sent: Monday, September 12, 2022 7:47 PM
> To: Wang, YuanX <yuanx.wang@intel.com>; dev@dpdk.org; Thomas
> Monjalon <thomas@monjalon.net>; Ferruh Yigit <ferruh.yigit@xilinx.com>
> Cc: mdr@ashroe.eu; Li, Xiaoyun <xiaoyun.li@intel.com>; Singh, Aman Deep
> <aman.deep.singh@intel.com>; Zhang, Yuying <yuying.zhang@intel.com>;
> Zhang, Qi Z <qi.z.zhang@intel.com>; Yang, Qiming <qiming.yang@intel.com>;
> jerinjacobk@gmail.com; viacheslavo@nvidia.com;
> stephen@networkplumber.org; Ding, Xuan <xuan.ding@intel.com>;
> hpothula@marvell.com; Tang, Yaqi <yaqi.tang@intel.com>; Wenxuan Wu
> <wenxuanx.wu@intel.com>
> Subject: Re: [PATCH v3 2/4] ethdev: introduce protocol hdr based buffer split
>
> On 9/2/22 22:10, Yuan Wang wrote:
> > Currently, Rx buffer split supports length based split. With Rx queue
> > offload RTE_ETH_RX_OFFLOAD_BUFFER_SPLIT enabled and Rx packet
> segment
> > configured, PMD will be able to split the received packets into
> > multiple segments.
> >
> > However, length based buffer split is not suitable for NICs that do
> > split based on protocol headers. Given an arbitrarily variable length
> > in Rx packet segment, it is almost impossible to pass a fixed protocol
> > header to driver. Besides, the existence of tunneling results in the
> > composition of a packet is various, which makes the situation even worse.
> >
> > This patch extends current buffer split to support protocol header
> > based buffer split. A new proto_hdr field is introduced in the
> > reserved field of rte_eth_rxseg_split structure to specify protocol
> > header. The proto_hdr field defines the split position of packet,
> > splitting will always happens after the protocol header defined in the
> > Rx packet segment. When Rx queue offload
> > RTE_ETH_RX_OFFLOAD_BUFFER_SPLIT is enabled and corresponding
> protocol
> > header is configured, driver will split the ingress packets into multiple
> segments.
> >
> > struct rte_eth_rxseg_split {
> > struct rte_mempool *mp; /* memory pools to allocate segment from
> */
> > uint16_t length; /* segment maximal data length,
> > configures split point */
> > uint16_t offset; /* data offset from beginning
> > of mbuf data buffer */
> > uint32_t proto_hdr; /* supported ptype of a specific pmd,
> > configures split point.
> > It should be defined by RTE_PTYPE_*
>
> If I understand correctly, the statement is a bit misleading since it should be a
> bit mask of RTE_PTYPE_* defines. Not exactly one RTE_PTYPE_*.
Do you mean that a segment should support multiple protocol headers, such as splitting both tcp and udp headers?
>
> > */
> > };
> >
> > If protocol header split can be supported by a PMD. The
> > rte_eth_buffer_split_get_supported_hdr_ptypes function can be use to
> > obtain a list of these protocol headers.
> >
> > For example, let's suppose we configured the Rx queue with the
> > following segments:
> > seg0 - pool0, proto_hdr0=RTE_PTYPE_L3_IPV4, off0=2B
> > seg1 - pool1, proto_hdr1=RTE_PTYPE_L4_UDP, off1=128B
> > seg2 - pool2, off1=0B
> >
> > The packet consists of MAC_IPV4_UDP_PAYLOAD will be split like
>
> What is MAC_IPV4_UDP_PAYLOAD? Do you mean ETH_IPV4_UDP_PAYLOAD?
Thanks for your correction, it should be ETH_IPV4_UDP_PAYLOAD.
>
> > following:
> > seg0 - ipv4 header @ RTE_PKTMBUF_HEADROOM + 2 in mbuf from
> pool0
> > seg1 - udp header @ 128 in mbuf from pool1
> > seg2 - payload @ 0 in mbuf from pool2
> >
> > Note: NIC will only do split when the packets exactly match all the
> > protocol headers in the segments. For example, if ARP packets received
> > with above config, the NIC won't do split for ARP packets since it
> > does not contains ipv4 header and udp header.
>
> You must define which mempool is used in the case.
IMHO I don't think we can define which mempool to use, it depends on NIC behavior.
For our NIC, packets that are unable to split will be put into the last valid pool, with zero offset.
So here we would like to define to put these packets into the last valid mempool, with zero offset.
>
> >
> > Now buffer split can be configured in two modes. For length based
> > buffer split, the mp, length, offset field in Rx packet segment should
> > be configured, while the proto_hdr field will be ignored.
> > For protocol header based buffer split, the mp, offset, proto_hdr
> > field in Rx packet segment should be configured, while the length
> > field will be ignored.
> >
> > The split limitations imposed by underlying driver is reported in the
> > rte_eth_dev_info->rx_seg_capa field. The memory attributes for the
> > split parts may differ either, dpdk memory and external memory,
> respectively.
> >
> > Signed-off-by: Yuan Wang <yuanx.wang@intel.com>
> > Signed-off-by: Xuan Ding <xuan.ding@intel.com>
> > Signed-off-by: Wenxuan Wu <wenxuanx.wu@intel.com>
> > ---
> > doc/guides/rel_notes/release_22_11.rst | 5 +++
> > lib/ethdev/rte_ethdev.c | 55 ++++++++++++++++++++------
> > lib/ethdev/rte_ethdev.h | 17 +++++++-
> > 3 files changed, 65 insertions(+), 12 deletions(-)
> >
> > diff --git a/doc/guides/rel_notes/release_22_11.rst
> > b/doc/guides/rel_notes/release_22_11.rst
> > index 4d90514a9a..f3b58c7895 100644
> > --- a/doc/guides/rel_notes/release_22_11.rst
> > +++ b/doc/guides/rel_notes/release_22_11.rst
> > @@ -60,6 +60,11 @@ New Features
> > Added ``rte_eth_buffer_split_get_supported_hdr_ptypes()``, to get
> supported
> > header protocols of a PMD to split.
> >
> > +* **Added protocol header based buffer split.**
> > + Ethdev: The ``reserved`` field in the ``rte_eth_rxseg_split``
> > +structure is
> > + replaced with ``proto_hdr`` to support protocol header based buffer split.
> > + User can choose length or protocol header to configure buffer split
> > + according to NIC's capability.
>
> Add one more empty line to have two before the next sectoin.
Thanks for your catch.
>
> >
> > Removed Items
> > -------------
> > diff --git a/lib/ethdev/rte_ethdev.c b/lib/ethdev/rte_ethdev.c index
> > 093c577add..dfceb723ee 100644
> > --- a/lib/ethdev/rte_ethdev.c
> > +++ b/lib/ethdev/rte_ethdev.c
> > @@ -1635,9 +1635,10 @@ rte_eth_dev_is_removed(uint16_t port_id)
> > }
> >
> > static int
> > -rte_eth_rx_queue_check_split(const struct rte_eth_rxseg_split *rx_seg,
> > - uint16_t n_seg, uint32_t *mbp_buf_size,
> > - const struct rte_eth_dev_info *dev_info)
> > +rte_eth_rx_queue_check_split(uint16_t port_id,
> > + const struct rte_eth_rxseg_split *rx_seg,
> > + uint16_t n_seg, uint32_t *mbp_buf_size,
> > + const struct rte_eth_dev_info *dev_info)
> > {
> > const struct rte_eth_rxseg_capa *seg_capa = &dev_info-
> >rx_seg_capa;
> > struct rte_mempool *mp_first;
> > @@ -1660,6 +1661,7 @@ rte_eth_rx_queue_check_split(const struct
> rte_eth_rxseg_split *rx_seg,
> > struct rte_mempool *mpl = rx_seg[seg_idx].mp;
> > uint32_t length = rx_seg[seg_idx].length;
> > uint32_t offset = rx_seg[seg_idx].offset;
> > + uint32_t proto_hdr = rx_seg[seg_idx].proto_hdr;
> >
> > if (mpl == NULL) {
> > RTE_ETHDEV_LOG(ERR, "null mempool pointer\n");
> @@ -1693,13
> > +1695,44 @@ rte_eth_rx_queue_check_split(const struct
> rte_eth_rxseg_split *rx_seg,
> > }
> > offset += seg_idx != 0 ? 0 : RTE_PKTMBUF_HEADROOM;
> > *mbp_buf_size = rte_pktmbuf_data_room_size(mpl);
> > - length = length != 0 ? length : *mbp_buf_size;
> > - if (*mbp_buf_size < length + offset) {
> > - RTE_ETHDEV_LOG(ERR,
> > - "%s mbuf_data_room_size %u < %u
> (segment length=%u + segment offset=%u)\n",
> > - mpl->name, *mbp_buf_size,
> > - length + offset, length, offset);
> > - return -EINVAL;
> > +
> > + int ret =
> rte_eth_buffer_split_get_supported_hdr_ptypes(port_id,
> > +NULL, 0);
>
> Do not mix variable declaration and code.
> It is better to give the variable some sensible name.
> Otherwise else branch code is hard to read.
Thanks for the suggestion, will take care of naming.
>
> > + if (ret <= 0) {
>
> May be I'm missing something, but nothing prevetns a driver/HW to support
> both protocol-based and fixed-length split.
> So, ability to support protocol based split should be treated as a request to
> do it. It must be based on rx_seg->proto_hdr content (for all segments).
>
> Also nothing should prevent to mix protocol and fixed-length split. I.e. split
> just after UDP in the first segment,
> 40 bytes in the second segment, everything else in the third.
Mix mode is an interesting idea. Currently testpmd and driver do not support mixed mode, but it does not affect the library to support this mode.
>
> > + /* Split at fixed length. */
> > + length = length != 0 ? length : *mbp_buf_size;
> > + if (*mbp_buf_size < length + offset) {
> > + RTE_ETHDEV_LOG(ERR,
> > + "%s mbuf_data_room_size %u < %u
> (segment length=%u + segment offset=%u)\n",
> > + mpl->name, *mbp_buf_size,
> > + length + offset, length, offset);
> > + return -EINVAL;
> > + }
> > + } else {
> > + /* Split after specified protocol header. */
> > + uint32_t ptypes[ret];
> > + int i;
> > +
> > + ret =
> rte_eth_buffer_split_get_supported_hdr_ptypes(port_id,
> > +ptypes, ret);
>
> In theory, the funciton could fail since input arguments differ. So, it should be
> handled.
Thanks for your catch, will fix in the next version.
>
> > + for (i = 0; i < ret; i++)
> > + if (ptypes[i] & proto_hdr)
>
> IMHO it should be ==, not &. I think that
> rte_eth_buffer_split_get_supported_hdr_ptypes() should define points at
> which split could happen and we should match the point exactly.
Sure, == is better. Thanks for the suggestion.
>
> > + break;
> > +
> > + if (i == ret) {
> > +#define PTYPE_NAMESIZE 256
>
> Why? It is looks really strange that it is defined here.
I intend to display the protocol name in the log, but if the proto_hdr is a bit mask, can I just show the number?
Please see v4 for this modification.
>
> > + char ptype_name[PTYPE_NAMESIZE];
> > + rte_get_ptype_name(proto_hdr,
> ptype_name, sizeof(ptype_name));
> > + RTE_ETHDEV_LOG(ERR,
> > + "Protocol header %s is not
> supported.\n",
> > + ptype_name);
> > + return -EINVAL;
> > + }
> > + if (*mbp_buf_size < offset) {
> > + RTE_ETHDEV_LOG(ERR,
> > + "%s
> mbuf_data_room_size %u < %u segment offset)\n",
> > + mpl->name, *mbp_buf_size,
> > + offset);
> > + return -EINVAL;
> > + }
> > }
> > }
> > return 0;
> > @@ -1778,7 +1811,7 @@ rte_eth_rx_queue_setup(uint16_t port_id,
> uint16_t rx_queue_id,
> > n_seg = rx_conf->rx_nseg;
> >
> > if (rx_conf->offloads &
> RTE_ETH_RX_OFFLOAD_BUFFER_SPLIT) {
> > - ret = rte_eth_rx_queue_check_split(rx_seg, n_seg,
> > + ret = rte_eth_rx_queue_check_split(port_id, rx_seg,
> n_seg,
> > &mbp_buf_size,
> > &dev_info);
> > if (ret != 0)
> > diff --git a/lib/ethdev/rte_ethdev.h b/lib/ethdev/rte_ethdev.h index
> > c58c908c3a..410fba5eab 100644
> > --- a/lib/ethdev/rte_ethdev.h
> > +++ b/lib/ethdev/rte_ethdev.h
> > @@ -1175,6 +1175,9 @@ struct rte_eth_txmode {
> > * specified in the first array element, the second buffer, from the
> > * pool in the second element, and so on.
> > *
> > + * - The proto_hdrs in the elements define the split position of
> > + * received packets.
> > + *
> > * - The offsets from the segment description elements specify
> > * the data offset from the buffer beginning except the first mbuf.
> > * The first segment offset is added with RTE_PKTMBUF_HEADROOM.
> > @@ -1196,12 +1199,24 @@ struct rte_eth_txmode {
> > * - pool from the last valid element
> > * - the buffer size from this pool
> > * - zero offset
> > + *
> > + * - Length based buffer split:
> > + * - mp, length, offset should be configured.
> > + * - The proto_hdr field will be ignored.
> > + *
> > + * - Protocol header based buffer split:
> > + * - mp, offset, proto_hdr should be configured.
> > + * - The length field will be ignored.
> > */
> > struct rte_eth_rxseg_split {
> > struct rte_mempool *mp; /**< Memory pool to allocate segment
> from. */
> > uint16_t length; /**< Segment data length, configures split point. */
> > uint16_t offset; /**< Data offset from beginning of mbuf data buffer.
> */
> > - uint32_t reserved; /**< Reserved field. */
> > + /**
> > + * Supported ptype of a specific pmd, configures split point.
> > + * It should be defined by RTE_PTYPE_*.
> > + */
> > + uint32_t proto_hdr;
> > };
> >
> > /**
next prev parent reply other threads:[~2022-09-16 8:39 UTC|newest]
Thread overview: 72+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-08-12 18:15 [PATCH 0/4] support protocol " Yuan Wang
2022-08-12 18:15 ` [PATCH 1/4] ethdev: introduce protocol header API Yuan Wang
2022-08-12 18:15 ` [PATCH 2/4] ethdev: introduce protocol hdr based buffer split Yuan Wang
2022-08-12 18:15 ` [PATCH 3/4] app/testpmd: add rxhdrs commands and parameters Yuan Wang
2022-08-12 18:15 ` [PATCH 4/4] net/ice: support buffer split in Rx path Yuan Wang
2022-09-01 22:33 ` [PATCH v2 0/4] support protocol based buffer split Yuan Wang
2022-09-01 22:34 ` [PATCH v2 1/4] ethdev: introduce protocol header API Yuan Wang
2022-09-01 22:35 ` [PATCH v2 2/4] ethdev: introduce protocol hdr based buffer split Yuan Wang
2022-09-01 22:36 ` [PATCH v2 3/4] app/testpmd: add rxhdrs commands and parameters Yuan Wang
2022-09-01 22:37 ` [PATCH v2 4/4] net/ice: support buffer split in Rx path Yuan Wang
2022-09-02 19:10 ` [PATCH v3 0/4] support protocol based buffer split Yuan Wang
2022-09-02 19:10 ` [PATCH v3 1/4] ethdev: introduce protocol header API Yuan Wang
2022-09-12 11:24 ` Andrew Rybchenko
2022-09-16 8:34 ` Wang, YuanX
2022-09-02 19:10 ` [PATCH v3 2/4] ethdev: introduce protocol hdr based buffer split Yuan Wang
2022-09-12 11:47 ` Andrew Rybchenko
2022-09-16 8:38 ` Wang, YuanX [this message]
2022-09-20 5:35 ` Andrew Rybchenko
2022-09-22 3:13 ` Wang, YuanX
2022-09-13 7:56 ` Suanming Mou
2022-09-16 8:39 ` Wang, YuanX
2022-09-02 19:10 ` [PATCH v3 3/4] app/testpmd: add rxhdrs commands and parameters Yuan Wang
2022-09-02 19:10 ` [PATCH v3 4/4] net/ice: support buffer split in Rx path Yuan Wang
2022-09-20 11:12 ` [PATCH v4 0/4] support protocol based buffer split Yuan Wang
2022-09-20 11:12 ` [PATCH v4 1/4] ethdev: introduce protocol header API Yuan Wang
2022-09-20 11:12 ` [PATCH v4 2/4] ethdev: introduce protocol hdr based buffer split Yuan Wang
2022-09-20 11:12 ` [PATCH v4 3/4] app/testpmd: add rxhdrs commands and parameters Yuan Wang
2022-09-20 11:12 ` [PATCH v4 4/4] net/ice: support buffer split in Rx path Yuan Wang
2022-09-26 9:40 ` [PATCH v5 0/4] support protocol based buffer split Yuan Wang
2022-09-26 9:40 ` [PATCH v5 1/4] ethdev: introduce protocol header API Yuan Wang
2022-09-26 9:40 ` [PATCH v5 2/4] ethdev: introduce protocol hdr based buffer split Yuan Wang
2022-09-28 15:42 ` Wang, YuanX
2022-09-26 9:40 ` [PATCH v5 3/4] app/testpmd: add rxhdrs commands and parameters Yuan Wang
2022-09-26 9:40 ` [PATCH v5 4/4] net/ice: support buffer split in Rx path Yuan Wang
2022-09-29 18:59 ` [PATCH v6 0/4] support protocol based buffer split Yuan Wang
2022-09-29 18:59 ` [PATCH v6 1/4] ethdev: introduce protocol header API Yuan Wang
2022-09-29 18:59 ` [PATCH v6 2/4] ethdev: introduce protocol hdr based buffer split Yuan Wang
2022-09-29 18:59 ` [PATCH v6 3/4] app/testpmd: add rxhdrs commands and parameters Yuan Wang
2022-09-29 18:59 ` [PATCH v6 4/4] net/ice: support buffer split in Rx path Yuan Wang
2022-09-30 6:45 ` Tang, Yaqi
2022-10-01 21:05 ` [PATCH v7 0/4] support protocol based buffer split Yuan Wang
2022-10-01 21:05 ` [PATCH v7 1/4] ethdev: introduce protocol header API Yuan Wang
2022-10-03 7:04 ` Andrew Rybchenko
2022-10-04 2:21 ` Wang, YuanX
2022-10-04 7:52 ` Andrew Rybchenko
2022-10-04 15:00 ` Wang, YuanX
2022-10-01 21:05 ` [PATCH v7 2/4] ethdev: introduce protocol hdr based buffer split Yuan Wang
2022-10-02 4:01 ` Wang, YuanX
2022-10-03 7:47 ` Andrew Rybchenko
2022-10-04 2:48 ` Wang, YuanX
2022-10-04 8:22 ` Andrew Rybchenko
2022-10-04 15:01 ` Wang, YuanX
2022-10-01 21:05 ` [PATCH v7 3/4] app/testpmd: add rxhdrs commands and parameters Yuan Wang
2022-10-01 21:05 ` [PATCH v7 4/4] net/ice: support buffer split in Rx path Yuan Wang
2022-10-05 23:18 ` [PATCH v8 0/4] support protocol based buffer split Yuan Wang
2022-10-05 23:18 ` [PATCH v8 1/4] ethdev: introduce protocol header API Yuan Wang
2022-10-06 10:11 ` Andrew Rybchenko
2022-10-05 23:18 ` [PATCH v8 2/4] ethdev: introduce protocol hdr based buffer split Yuan Wang
2022-10-06 10:11 ` Andrew Rybchenko
2022-10-08 14:30 ` Ding, Xuan
2022-10-05 23:18 ` [PATCH v8 3/4] app/testpmd: add rxhdrs commands and parameters Yuan Wang
2022-10-06 10:12 ` Andrew Rybchenko
2022-10-05 23:18 ` [PATCH v8 4/4] net/ice: support buffer split in Rx path Yuan Wang
2022-10-06 10:12 ` Andrew Rybchenko
2022-10-06 10:13 ` [PATCH v8 0/4] support protocol based buffer split Andrew Rybchenko
2022-10-09 20:25 ` [PATCH v9 " Yuan Wang
2022-10-09 14:58 ` Andrew Rybchenko
2022-10-10 2:45 ` Ding, Xuan
2022-10-09 20:25 ` [PATCH v9 1/4] ethdev: introduce protocol header API Yuan Wang
2022-10-09 20:25 ` [PATCH v9 2/4] ethdev: introduce protocol hdr based buffer split Yuan Wang
2022-10-09 20:25 ` [PATCH v9 3/4] app/testpmd: add rxhdrs commands and parameters Yuan Wang
2022-10-09 20:25 ` [PATCH v9 4/4] net/ice: support buffer split in Rx path Yuan Wang
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=PH7PR11MB6953C1B1B16DA6EB24AEB1EB85489@PH7PR11MB6953.namprd11.prod.outlook.com \
--to=yuanx.wang@intel.com \
--cc=aman.deep.singh@intel.com \
--cc=andrew.rybchenko@oktetlabs.ru \
--cc=dev@dpdk.org \
--cc=ferruh.yigit@xilinx.com \
--cc=hpothula@marvell.com \
--cc=jerinjacobk@gmail.com \
--cc=mdr@ashroe.eu \
--cc=qi.z.zhang@intel.com \
--cc=qiming.yang@intel.com \
--cc=stephen@networkplumber.org \
--cc=thomas@monjalon.net \
--cc=viacheslavo@nvidia.com \
--cc=wenxuanx.wu@intel.com \
--cc=xiaoyun.li@intel.com \
--cc=xuan.ding@intel.com \
--cc=yaqi.tang@intel.com \
--cc=yuying.zhang@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).