From: "lihuisong (C)" <lihuisong@huawei.com>
To: Ferruh Yigit <ferruh.yigit@amd.com>, <dev@dpdk.org>,
Gagandeep Singh <g.singh@nxp.com>,
Hemant Agrawal <hemant.agrawal@nxp.com>,
Qiming Yang <qiming.yang@intel.com>,
Qi Zhang <qi.z.zhang@intel.com>, Simei Su <simei.su@intel.com>,
Junfeng Guo <junfeng.guo@intel.com>
Cc: <thomas@monjalon.net>, <andrew.rybchenko@oktetlabs.ru>,
<liuyonglong@huawei.com>
Subject: Re: [PATCH v3 0/2] ethdev: add the check for PTP capability
Date: Thu, 21 Sep 2023 19:59:12 +0800 [thread overview]
Message-ID: <f941c1aa-451e-ca9f-9a3e-caa31c0a363e@huawei.com> (raw)
In-Reply-To: <0d7f429c-8862-4f16-b7e5-46d69581f54f@amd.com>
add ice & igc maintainers
在 2023/9/21 19:06, Ferruh Yigit 写道:
> On 9/21/2023 11:02 AM, lihuisong (C) wrote:
>> Hi Ferruh,
>>
>> Sorry for my delay reply because of taking a look at all PMDs
>> implementation.
>>
>>
>> 在 2023/9/16 1:46, Ferruh Yigit 写道:
>>> On 8/17/2023 9:42 AM, Huisong Li wrote:
>>>> From the first version of ptpclient, it seems that this example
>>>> assume that
>>>> the PMDs support the PTP feature and enable PTP by default. Please see
>>>> commit ab129e9065a5 ("examples/ptpclient: add minimal PTP client")
>>>> which are introduced in 2015.
>>>>
>>>> And two years later, Rx HW timestamp offload was introduced to enable or
>>>> disable PTP feature in HW via rte_eth_rxmode. Please see
>>>> commit 42ffc45aa340 ("ethdev: add Rx HW timestamp capability").
>>>>
>>> Hi Huisong,
>>>
>>> As far as I know this offload is not for PTP.
>>> PTP and TIMESTAMP are different.
>> If TIMESTAMP offload cannot stand for PTP, we may need to add one new
>> offlaod for PTP.
>>
> Can you please detail what is "PTP offload"?
It indicates whether the device supports PTP or enable PTP feature.
If TIMESTAMP offload is not for PTP, I don't know what the point of this
offload independent existence is.
>
>>> PTP is a protocol for time sync.
>>> Rx TIMESTAMP offload is to ask HW to add timestamp to mbuf.
>> Yes.
>> But a lot of PMDs actually depand on HW to report Rx timestamp releated
>> information
>> because of reading Rx timestamp of PTP SYNC packet in read_rx_timestamp
>> API.
>>
> HW support may be required for PTP but this doesn't mean timestamp
> offload is used.
understand.
>
>>>> And then about four years later, ptpclient enable Rx timestamp offload
>>>> because some PMDs require this offload to enable. Please see
>>>> commit 7a04a4f67dca ("examples/ptpclient: enable Rx timestamp offload").
>>>>
>>> dpaa2 seems using TIMESTAMP offload and PTP together, hence they updated
>>> ptpclient sample to set TIMESTAMP offload.
>> There are many PMDs doing like this, such as ice, igc, cnxk, dpaa2, hns3
>> and so on.
>>
> Can you please point the ice & igc code, cc'ing their maintainers, we
> can look together?
*-->igc code:*
Having following codes in igc_recv_scattered_pkts():
if (rxq->offloads & RTE_ETH_RX_OFFLOAD_TIMESTAMP) {
uint32_t *ts = rte_pktmbuf_mtod_offset(first_seg,
uint32_t *, -IGC_TS_HDR_LEN);
rxq->rx_timestamp = (uint64_t)ts[3] * NSEC_PER_SEC +
ts[2];
rxm->timesync = rxq->queue_id;
}
Note：this rxm->timesync will be used in timesync_read_rx_timestamp()
*-->ice code:*
#ifndef RTE_LIBRTE_ICE_16BYTE_RX_DESC
if (ice_timestamp_dynflag > 0 &&
(rxq->offloads & RTE_ETH_RX_OFFLOAD_TIMESTAMP)) {
rxq->time_high =
rte_le_to_cpu_32(rxd.wb.flex_ts.ts_high);
if (unlikely(is_tsinit)) {
ts_ns = ice_tstamp_convert_32b_64b(hw, ad, 1,
rxq->time_high);
rxq->hw_time_low = (uint32_t)ts_ns;
rxq->hw_time_high = (uint32_t)(ts_ns >> 32);
is_tsinit = false;
} else {
if (rxq->time_high < rxq->hw_time_low)
rxq->hw_time_high += 1;
ts_ns = (uint64_t)rxq->hw_time_high << 32 | rxq->time_high;
rxq->hw_time_low = rxq->time_high;
}
rxq->hw_time_update = rte_get_timer_cycles() /
(rte_get_timer_hz() / 1000);
*RTE_MBUF_DYNFIELD(rxm,
(ice_timestamp_dynfield_offset),
rte_mbuf_timestamp_t *) = ts_ns;
pkt_flags |= ice_timestamp_dynflag;
}
if (ad->ptp_ena && ((rxm->packet_type & RTE_PTYPE_L2_MASK) ==
RTE_PTYPE_L2_ETHER_TIMESYNC)) {
rxq->time_high =
rte_le_to_cpu_32(rxd.wb.flex_ts.ts_high);
rxm->timesync = rxq->queue_id;
pkt_flags |= RTE_MBUF_F_RX_IEEE1588_PTP;
}
#endif
Note: rxm->timesync and rxq->time_high will be used in
timesync_read_rx_timestamp()
>
>
>>> We need to clarify dpaa2 usage.
>>>
>>>> By all the records, this is more like a process of perfecting PTP
>>>> feature.
>>>> Not all network adaptors support PTP feature. So adding the check for
>>>> PTP
>>>> capability in ethdev layer is necessary.
>>>>
>>> Nope, as PTP (IEEE1588/802.1AS) implemented as dev_ops, and ops already
>>> checked, so no additional check is needed.
>> But only having dev_ops about PTP doesn't satisfy the use of this feature.
>> For example,
>> there are serveal network ports belonged to a driver on one OS, and only
>> one port support PTP function.
>> So driver needs one *PTP* offload.
>>> We just need to clarify TIMESTAMP offload and PTP usage and find out
>>> what is causing confusion.
>> Yes it is a little bit confusion.
>> There are two kinds of implementation:
>> A： ixgbe and txgbe (it seems that their HW is similar) don't need
>> TIMESTAMP offload，and only use dev_ops to finish PTP feature.
>> B: saving "Rx timestamp related information" from Rx description when
>> receive PTP SYNC packet and
>> report it in read_rx_timestamp API.
>> For case B, most of driver use TIMESTAMP offload to decide if driver
>> save "Rx timestamp related information.
>> What do you think about this, Ferruh?
>>> I would be great if you can help on clarification, and update
>>> documentation or API comments, or what ever required, for this.
>> ok
>>>> ---
>>>> v3:
>>>> - patch [2/3] for hns3 has been applied and so remove it.
>>>> - ops pointer check is closer to usage.
>>>>
>>>> Huisong Li (2):
>>>> examples/ptpclient: add the check for PTP capability
>>>> ethdev: add the check for the valitity of timestamp offload
>>>>
>>>> examples/ptpclient/ptpclient.c | 5 +++
>>>> lib/ethdev/rte_ethdev.c | 57 +++++++++++++++++++++++++++++++++-
>>>> 2 files changed, 61 insertions(+), 1 deletion(-)
>>>>
>>> .
> .
next prev parent reply other threads:[~2023-09-21 11:59 UTC|newest]
Thread overview: 36+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-06-28 13:39 [PATCH 0/3] some bugfixes for PTP Dongdong Liu
2022-06-28 13:39 ` [PATCH 1/3] examples/ptpclient: add the check for PTP capability Dongdong Liu
2022-06-28 13:39 ` [PATCH 2/3] net/hns3: fix fail to receive PTP packet Dongdong Liu
2022-06-28 13:39 ` [PATCH 3/3] ethdev: add the check for the valitity of timestamp offload Dongdong Liu
2022-07-02 8:17 ` [PATCH v2 0/3] some bugfixes for PTP Dongdong Liu
2022-07-02 8:17 ` [PATCH v2 1/3] examples/ptpclient: add the check for PTP capability Dongdong Liu
2022-07-02 8:17 ` [PATCH v2 2/3] net/hns3: fix fail to receive PTP packet Dongdong Liu
2022-07-02 8:17 ` [PATCH v2 3/3] ethdev: add the check for the valitity of timestamp offload Dongdong Liu
2022-07-06 14:57 ` Andrew Rybchenko
2022-07-07 2:05 ` lihuisong (C)
2023-08-17 8:42 ` [PATCH v3 0/2] ethdev: add the check for PTP capability Huisong Li
2023-08-17 8:42 ` [PATCH v3 1/2] examples/ptpclient: " Huisong Li
2023-09-15 17:29 ` Ferruh Yigit
2023-09-21 9:18 ` lihuisong (C)
2023-09-21 11:02 ` Ferruh Yigit
2023-09-21 11:22 ` Hemant Agrawal
2023-10-20 4:05 ` lihuisong (C)
2023-09-21 11:36 ` lihuisong (C)
2023-08-17 8:42 ` [PATCH v3 2/2] ethdev: add the check for the valitity of timestamp offload Huisong Li
2023-09-15 17:46 ` [PATCH v3 0/2] ethdev: add the check for PTP capability Ferruh Yigit
2023-09-21 10:02 ` lihuisong (C)
2023-09-21 11:06 ` Ferruh Yigit
2023-09-21 11:17 ` Hemant Agrawal
2023-10-20 3:58 ` lihuisong (C)
2023-11-01 23:39 ` Ferruh Yigit
2023-11-23 11:40 ` lihuisong (C)
2023-11-01 23:39 ` Ferruh Yigit
2023-09-21 11:59 ` lihuisong (C) [this message]
2023-11-01 23:39 ` Ferruh Yigit
2023-11-23 11:56 ` lihuisong (C)
2024-01-11 6:25 ` lihuisong (C)
2024-01-26 16:54 ` Ferruh Yigit
2024-01-27 1:52 ` lihuisong (C)
2024-01-29 11:16 ` Ferruh Yigit
2024-01-29 13:58 ` lihuisong (C)
2024-01-29 15:00 ` Ferruh Yigit
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=f941c1aa-451e-ca9f-9a3e-caa31c0a363e@huawei.com \
--to=lihuisong@huawei.com \
--cc=andrew.rybchenko@oktetlabs.ru \
--cc=dev@dpdk.org \
--cc=ferruh.yigit@amd.com \
--cc=g.singh@nxp.com \
--cc=hemant.agrawal@nxp.com \
--cc=junfeng.guo@intel.com \
--cc=liuyonglong@huawei.com \
--cc=qi.z.zhang@intel.com \
--cc=qiming.yang@intel.com \
--cc=simei.su@intel.com \
--cc=thomas@monjalon.net \
/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).