DPDK patches and discussions
 help / color / mirror / Atom feed
From: Stephen Hemminger <stephen@networkplumber.org>
To: dev@dpdk.org
Cc: Stephen Hemminger <stephen@networkplumber.org>,
	stable@dpdk.org, Reshma Pattan <reshma.pattan@intel.com>,
	Harry van Haaren <harry.van.haaren@intel.com>,
	Remy Horton <remy.horton@intel.com>
Subject: [PATCH 1/2] latencystats: fix receive sample MP issues
Date: Thu, 12 Jun 2025 17:34:25 -0700	[thread overview]
Message-ID: <20250613003547.39239-2-stephen@networkplumber.org> (raw)
In-Reply-To: <20250613003547.39239-1-stephen@networkplumber.org>

The receive callback was not safe with multiple queues.
If one receive queue callback decides to take a sample it
needs to add that sample and do atomic update to the previous
TSC sample value. Add a new lock for that.

Also, add code to handle TSC wraparound in comparison.
Perhaps this should move to rte_cycles.h?

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Fixes: 5cd3cac9ed22 ("latency: added new library for latency stats")
Cc: stable@dpdk.org
---
 lib/latencystats/rte_latencystats.c | 47 ++++++++++++++++++-----------
 1 file changed, 29 insertions(+), 18 deletions(-)

diff --git a/lib/latencystats/rte_latencystats.c b/lib/latencystats/rte_latencystats.c
index 6873a44a92..181c53dd0e 100644
--- a/lib/latencystats/rte_latencystats.c
+++ b/lib/latencystats/rte_latencystats.c
@@ -45,10 +45,19 @@ timestamp_dynfield(struct rte_mbuf *mbuf)
 			timestamp_dynfield_offset, rte_mbuf_timestamp_t *);
 }
 
+/* Compare two 64 bit timer counter but deal with wraparound correctly. */
+static inline bool tsc_after(uint64_t t0, uint64_t t1)
+{
+	return (int64_t)(t1 - t0) < 0;
+}
+
+#define tsc_before(a, b) tsc_after(b, a)
+
 static const char *MZ_RTE_LATENCY_STATS = "rte_latencystats";
 static int latency_stats_index;
+
+static rte_spinlock_t sample_lock = RTE_SPINLOCK_INITIALIZER;
 static uint64_t samp_intvl;
-static uint64_t timer_tsc;
 static uint64_t prev_tsc;
 
 #define LATENCY_AVG_SCALE     4
@@ -147,25 +156,27 @@ add_time_stamps(uint16_t pid __rte_unused,
 		void *user_cb __rte_unused)
 {
 	unsigned int i;
-	uint64_t diff_tsc, now;
+	uint64_t now = rte_rdtsc();
 
-	/*
-	 * For every sample interval,
-	 * time stamp is marked on one received packet.
-	 */
-	now = rte_rdtsc();
-	for (i = 0; i < nb_pkts; i++) {
-		diff_tsc = now - prev_tsc;
-		timer_tsc += diff_tsc;
-
-		if ((pkts[i]->ol_flags & timestamp_dynflag) == 0
-				&& (timer_tsc >= samp_intvl)) {
-			*timestamp_dynfield(pkts[i]) = now;
-			pkts[i]->ol_flags |= timestamp_dynflag;
-			timer_tsc = 0;
+	/* Check without locking */
+	if (likely(tsc_before(now, prev_tsc + samp_intvl)))
+		return nb_pkts;
+
+	/* Try and get sample, skip if sample is being done by other core. */
+	if (likely(rte_spinlock_trylock(&sample_lock))) {
+		for (i = 0; i < nb_pkts; i++) {
+			struct rte_mbuf *m = pkts[i];
+
+			/* skip if already timestamped */
+			if (unlikely(m->ol_flags & timestamp_dynflag))
+				continue;
+
+			m->ol_flags |= timestamp_dynflag;
+			*timestamp_dynfield(m) = now;
+			prev_tsc = now;
+			break;
 		}
-		prev_tsc = now;
-		now = rte_rdtsc();
+		rte_spinlock_unlock(&sample_lock);
 	}
 
 	return nb_pkts;
-- 
2.47.2


  reply	other threads:[~2025-06-13  0:35 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-06-13  0:34 [PATCH 0/2] Latencystat optimization and fix Stephen Hemminger
2025-06-13  0:34 ` Stephen Hemminger [this message]
2025-06-13  0:34 ` [PATCH 2/2] latencystats: optimize locking on transmit Stephen Hemminger
2025-06-13 11:13 ` [PATCH 0/2] Latencystat optimization and fix Varghese, Vipin

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=20250613003547.39239-2-stephen@networkplumber.org \
    --to=stephen@networkplumber.org \
    --cc=dev@dpdk.org \
    --cc=harry.van.haaren@intel.com \
    --cc=remy.horton@intel.com \
    --cc=reshma.pattan@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).