DPDK patches and discussions
 help / color / mirror / Atom feed
* [PATCH] net/nfp: fix get represetor wrong port stats
@ 2025-02-25  1:34 Chaoyong He
  2025-02-25 16:17 ` Stephen Hemminger
  0 siblings, 1 reply; 2+ messages in thread
From: Chaoyong He @ 2025-02-25  1:34 UTC (permalink / raw)
  To: dev; +Cc: oss-drivers, Long Wu, peng.zhang, chaoyong.he, stable

From: Long Wu <long.wu@corigine.com>

The 'ipackets'/'opackets' are used to record the number
of packets on represetor port received/sent. But the
code does not consider concurrent calculation of
'ipackets'/'opackets'. If multiple queues are calculated
'ipackets'/'opackets' simultaneously, it will result in
incorrect results.

The previous logic has recorded the number of packets on
each queue, therefore driver only needs to add the data of
all queues to obtain the data of the representor port.

Based on this, modify code to fix the issue.

Fixes: 636e133ec891 ("net/nfp: update Tx and Rx for multiple PF")
Fixes: 82a2c286f35a ("net/nfp: support xstats for flower firmware")
Cc: peng.zhang@corigine.com
Cc: chaoyong.he@corigine.com
Cc: stable@dpdk.org

Signed-off-by: Long Wu <long.wu@corigine.com>
---
 drivers/net/nfp/flower/nfp_flower.c            |  2 --
 .../net/nfp/flower/nfp_flower_representor.c    | 18 ++++++++++++++++--
 2 files changed, 16 insertions(+), 4 deletions(-)

diff --git a/drivers/net/nfp/flower/nfp_flower.c b/drivers/net/nfp/flower/nfp_flower.c
index f087d0dfdc..804ad494d5 100644
--- a/drivers/net/nfp/flower/nfp_flower.c
+++ b/drivers/net/nfp/flower/nfp_flower.c
@@ -238,7 +238,6 @@ nfp_flower_multiple_pf_recv_pkts(void *rx_queue,
 		for (i = 0; i < recv; i++)
 			data_len += rx_pkts[i]->data_len;
 
-		repr->repr_stats.ipackets += recv;
 		repr->repr_stats.q_ipackets[rxq->qidx] += recv;
 		repr->repr_stats.q_ibytes[rxq->qidx] += data_len;
 	}
@@ -277,7 +276,6 @@ nfp_flower_multiple_pf_xmit_pkts(void *tx_queue,
 		for (i = 0; i < sent; i++)
 			data_len += tx_pkts[i]->data_len;
 
-		repr->repr_stats.opackets += sent;
 		repr->repr_stats.q_opackets[txq->qidx] += sent;
 		repr->repr_stats.q_obytes[txq->qidx] += data_len;
 	}
diff --git a/drivers/net/nfp/flower/nfp_flower_representor.c b/drivers/net/nfp/flower/nfp_flower_representor.c
index 9601aa5f96..e377d45e53 100644
--- a/drivers/net/nfp/flower/nfp_flower_representor.c
+++ b/drivers/net/nfp/flower/nfp_flower_representor.c
@@ -315,9 +315,25 @@ static int
 nfp_flower_repr_stats_get(struct rte_eth_dev *ethdev,
 		struct rte_eth_stats *stats)
 {
+	uint16_t i;
 	struct nfp_flower_representor *repr;
 
 	repr = ethdev->data->dev_private;
+
+	repr->repr_stats.ipackets = 0;
+	repr->repr_stats.ibytes = 0;
+	for (i = 0; i < ethdev->data->nb_rx_queues; i++) {
+		repr->repr_stats.ipackets += repr->repr_stats.q_ipackets[i];
+		repr->repr_stats.ibytes += repr->repr_stats.q_ibytes[i];
+	}
+
+	repr->repr_stats.opackets = 0;
+	repr->repr_stats.obytes = 0;
+	for (i = 0; i < ethdev->data->nb_tx_queues; i++) {
+		repr->repr_stats.opackets += repr->repr_stats.q_opackets[i];
+		repr->repr_stats.obytes += repr->repr_stats.q_obytes[i];
+	}
+
 	rte_memcpy(stats, &repr->repr_stats, sizeof(struct rte_eth_stats));
 
 	return 0;
@@ -385,7 +401,6 @@ nfp_flower_repr_rx_burst(void *rx_queue,
 		for (i = 0; i < total_dequeue; i++)
 			data_len += rx_pkts[i]->data_len;
 
-		repr->repr_stats.ipackets += total_dequeue;
 		repr->repr_stats.q_ipackets[rxq->qidx] += total_dequeue;
 		repr->repr_stats.q_ibytes[rxq->qidx] += data_len;
 	}
@@ -434,7 +449,6 @@ nfp_flower_repr_tx_burst(void *tx_queue,
 		for (i = 0; i < sent; i++)
 			data_len += tx_pkts[i]->data_len;
 
-		repr->repr_stats.opackets += sent;
 		repr->repr_stats.q_opackets[txq->qidx] += sent;
 		repr->repr_stats.q_obytes[txq->qidx] += data_len;
 	}
-- 
2.43.5


^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: [PATCH] net/nfp: fix get represetor wrong port stats
  2025-02-25  1:34 [PATCH] net/nfp: fix get represetor wrong port stats Chaoyong He
@ 2025-02-25 16:17 ` Stephen Hemminger
  0 siblings, 0 replies; 2+ messages in thread
From: Stephen Hemminger @ 2025-02-25 16:17 UTC (permalink / raw)
  To: Chaoyong He; +Cc: dev, oss-drivers, Long Wu, peng.zhang, stable

On Tue, 25 Feb 2025 09:34:54 +0800
Chaoyong He <chaoyong.he@corigine.com> wrote:

> From: Long Wu <long.wu@corigine.com>
> 
> The 'ipackets'/'opackets' are used to record the number
> of packets on represetor port received/sent. But the
> code does not consider concurrent calculation of
> 'ipackets'/'opackets'. If multiple queues are calculated
> 'ipackets'/'opackets' simultaneously, it will result in
> incorrect results.
> 
> The previous logic has recorded the number of packets on
> each queue, therefore driver only needs to add the data of
> all queues to obtain the data of the representor port.
> 
> Based on this, modify code to fix the issue.
> 
> Fixes: 636e133ec891 ("net/nfp: update Tx and Rx for multiple PF")
> Fixes: 82a2c286f35a ("net/nfp: support xstats for flower firmware")
> Cc: peng.zhang@corigine.com
> Cc: chaoyong.he@corigine.com
> Cc: stable@dpdk.org
> 
> Signed-off-by: Long Wu <long.wu@corigine.com>

Applied to next-net and fixed spelling in commit message.

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2025-02-25 16:17 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2025-02-25  1:34 [PATCH] net/nfp: fix get represetor wrong port stats Chaoyong He
2025-02-25 16:17 ` Stephen Hemminger

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).