DPDK patches and discussions
 help / color / mirror / Atom feed
From: Song Yoong Siang <yoong.siang.song@intel.com>
To: Bruce Richardson <bruce.richardson@intel.com>,
	David Zage <david.zage@intel.com>,
	Soumyadeep Hore <soumyadeep.hore@intel.com>,
	dev@dpdk.org
Cc: stable@dpdk.org
Subject: [PATCH v1 1/1] net/e1000: force allocation of context descriptor for igc launch time
Date: Thu,  4 Dec 2025 09:30:08 +0800	[thread overview]
Message-ID: <20251204013009.857273-1-yoong.siang.song@intel.com> (raw)

When the timestamp dynamic flag is set in the mbuf offload flags, force
allocation of a new context descriptor. The launch time feature requires a
dedicated context descriptor to specify the desired launch time for packet.

Testing was performed on two Intel ADL-S platforms with i226 NICs
connected back-to-back. A DPDK sample application is created to send two
bursts of packets with 100,000 nanosecond burst interval. Each burst is
having 5 UDP packets and only the first packet in the burst using launch
time feature. Meanwhile, the tcpdump command below is used on the link
partner to capture the delta of Rx hardware timestamp of the 10 packets:

tcpdump -ttt -ni enp1s0 --time-stamp-precision=nano -j adapter_unsynced

Without this patch, packets of second burst are transmitted immediately as
the driver din't create context descriptor for the launch time, as shown
in tcpdump log below:

  00:00:00.000000000 IP 192.168.1.100.2 > 224.1.1.1.5: UDP, length 982
  00:00:00.000008384 IP 192.168.1.100.2 > 224.1.1.1.5: UDP, length 982
  00:00:00.000008381 IP 192.168.1.100.2 > 224.1.1.1.5: UDP, length 982
  00:00:00.000008384 IP 192.168.1.100.2 > 224.1.1.1.5: UDP, length 982
  00:00:00.000008384 IP 192.168.1.100.2 > 224.1.1.1.5: UDP, length 982
  00:00:00.000008384 IP 192.168.1.100.2 > 224.1.1.1.5: UDP, length 982
  00:00:00.000008387 IP 192.168.1.100.2 > 224.1.1.1.5: UDP, length 982
  00:00:00.000008384 IP 192.168.1.100.2 > 224.1.1.1.5: UDP, length 982
  00:00:00.000008381 IP 192.168.1.100.2 > 224.1.1.1.5: UDP, length 982
  00:00:00.000008387 IP 192.168.1.100.2 > 224.1.1.1.5: UDP, length 982

With this patch, packets of second burst are properly held until the
intended 100,000 nanosecond intervals, demonstrating correct launch time
behavior. As shown in tcpdump log below, the delta of first packet
between first and second bursts is 8387 + 8384 + 8384 + 8384 + 66461 =
100,000 nanosecond.

  00:00:00.000000000 IP 192.168.1.100.2 > 224.1.1.1.5: UDP, length 982
  00:00:00.000008387 IP 192.168.1.100.2 > 224.1.1.1.5: UDP, length 982
  00:00:00.000008384 IP 192.168.1.100.2 > 224.1.1.1.5: UDP, length 982
  00:00:00.000008384 IP 192.168.1.100.2 > 224.1.1.1.5: UDP, length 982
  00:00:00.000008384 IP 192.168.1.100.2 > 224.1.1.1.5: UDP, length 982
  00:00:00.000066461 IP 192.168.1.100.2 > 224.1.1.1.5: UDP, length 982
  00:00:00.000008387 IP 192.168.1.100.2 > 224.1.1.1.5: UDP, length 982
  00:00:00.000008384 IP 192.168.1.100.2 > 224.1.1.1.5: UDP, length 982
  00:00:00.000008384 IP 192.168.1.100.2 > 224.1.1.1.5: UDP, length 982
  00:00:00.000008384 IP 192.168.1.100.2 > 224.1.1.1.5: UDP, length 982

Fixes: 9630f7c71ecd ("net/igc: enable launch time offloading")
Cc: stable@dpdk.org

Signed-off-by: David Zage <david.zage@intel.com>
Signed-off-by: Song Yoong Siang <yoong.siang.song@intel.com>
---
 drivers/net/intel/e1000/igc_txrx.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/drivers/net/intel/e1000/igc_txrx.c b/drivers/net/intel/e1000/igc_txrx.c
index f2a6c34094..35f36d266a 100644
--- a/drivers/net/intel/e1000/igc_txrx.c
+++ b/drivers/net/intel/e1000/igc_txrx.c
@@ -1416,6 +1416,10 @@ what_advctx_update(struct igc_tx_queue *txq, uint64_t flags,
 {
 	uint32_t curr = txq->ctx_curr;
 
+	/* Launch time feature always need a new context descriptor */
+	if (flags & igc_tx_timestamp_dynflag)
+		return IGC_CTX_NUM;
+
 	/* If match with the current context */
 	if (likely(txq->ctx_cache[curr].flags == flags &&
 		txq->ctx_cache[curr].tx_offload.data ==
@@ -1621,7 +1625,8 @@ igc_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
 		tx_last = (uint16_t)(tx_id + tx_pkt->nb_segs - 1);
 
 		ol_flags = tx_pkt->ol_flags;
-		tx_ol_req = ol_flags & IGC_TX_OFFLOAD_MASK;
+		tx_ol_req = ol_flags & (IGC_TX_OFFLOAD_MASK |
+				igc_tx_timestamp_dynflag);
 
 		/* If a Context Descriptor need be built . */
 		if (tx_ol_req) {
-- 
2.48.1


             reply	other threads:[~2025-12-04  1:13 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-12-04  1:30 Song Yoong Siang [this message]
2025-12-11 15:32 ` Bruce Richardson
2025-12-11 15:33   ` Bruce Richardson

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=20251204013009.857273-1-yoong.siang.song@intel.com \
    --to=yoong.siang.song@intel.com \
    --cc=bruce.richardson@intel.com \
    --cc=david.zage@intel.com \
    --cc=dev@dpdk.org \
    --cc=soumyadeep.hore@intel.com \
    --cc=stable@dpdk.org \
    /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).