DPDK patches and discussions
 help / color / mirror / Atom feed
From: Nithin Dabilpuram <ndabilpuram@marvell.com>
To: <jerinj@marvell.com>, Nithin Dabilpuram <ndabilpuram@marvell.com>,
	"Kiran Kumar K" <kirankumark@marvell.com>,
	Sunil Kumar Kori <skori@marvell.com>,
	Satha Rao <skoteshwar@marvell.com>
Cc: <dev@dpdk.org>
Subject: [PATCH v2 10/12] net/cnxk: resize CQ for Rx security for errata
Date: Thu, 16 Jun 2022 14:54:18 +0530	[thread overview]
Message-ID: <20220616092420.17861-10-ndabilpuram@marvell.com> (raw)
In-Reply-To: <20220616092420.17861-1-ndabilpuram@marvell.com>

Resize CQ for Rx security offload in case of HW errata.

Signed-off-by: Nithin Dabilpuram <ndabilpuram@marvell.com>
---
 drivers/net/cnxk/cnxk_ethdev.c | 43 +++++++++++++++++++++++++++++++++++++++++-
 drivers/net/cnxk/cnxk_ethdev.h |  2 +-
 2 files changed, 43 insertions(+), 2 deletions(-)

diff --git a/drivers/net/cnxk/cnxk_ethdev.c b/drivers/net/cnxk/cnxk_ethdev.c
index 4ea1617..2418290 100644
--- a/drivers/net/cnxk/cnxk_ethdev.c
+++ b/drivers/net/cnxk/cnxk_ethdev.c
@@ -5,6 +5,8 @@
 
 #include <rte_eventdev.h>
 
+#define CNXK_NIX_CQ_INL_CLAMP_MAX (64UL * 1024UL)
+
 static inline uint64_t
 nix_get_rx_offload_capa(struct cnxk_eth_dev *dev)
 {
@@ -40,6 +42,39 @@ nix_get_speed_capa(struct cnxk_eth_dev *dev)
 	return speed_capa;
 }
 
+static uint32_t
+nix_inl_cq_sz_clamp_up(struct roc_nix *nix, struct rte_mempool *mp,
+		       uint32_t nb_desc)
+{
+	struct roc_nix_rq *inl_rq;
+	uint64_t limit;
+
+	if (!roc_errata_cpt_hang_on_x2p_bp())
+		return nb_desc;
+
+	/* CQ should be able to hold all buffers in first pass RQ's aura
+	 * this RQ's aura.
+	 */
+	inl_rq = roc_nix_inl_dev_rq(nix);
+	if (!inl_rq) {
+		/* This itself is going to be inline RQ's aura */
+		limit = roc_npa_aura_op_limit_get(mp->pool_id);
+	} else {
+		limit = roc_npa_aura_op_limit_get(inl_rq->aura_handle);
+		/* Also add this RQ's aura if it is different */
+		if (inl_rq->aura_handle != mp->pool_id)
+			limit += roc_npa_aura_op_limit_get(mp->pool_id);
+	}
+	nb_desc = PLT_MAX(limit + 1, nb_desc);
+	if (nb_desc > CNXK_NIX_CQ_INL_CLAMP_MAX) {
+		plt_warn("Could not setup CQ size to accommodate"
+			 " all buffers in related auras (%" PRIu64 ")",
+			 limit);
+		nb_desc = CNXK_NIX_CQ_INL_CLAMP_MAX;
+	}
+	return nb_desc;
+}
+
 int
 cnxk_nix_inb_mode_set(struct cnxk_eth_dev *dev, bool use_inl_dev)
 {
@@ -504,7 +539,7 @@ cnxk_nix_tx_queue_release(struct rte_eth_dev *eth_dev, uint16_t qid)
 
 int
 cnxk_nix_rx_queue_setup(struct rte_eth_dev *eth_dev, uint16_t qid,
-			uint16_t nb_desc, uint16_t fp_rx_q_sz,
+			uint32_t nb_desc, uint16_t fp_rx_q_sz,
 			const struct rte_eth_rxconf *rx_conf,
 			struct rte_mempool *mp)
 {
@@ -552,6 +587,12 @@ cnxk_nix_rx_queue_setup(struct rte_eth_dev *eth_dev, uint16_t qid,
 	    dev->tx_offloads & RTE_ETH_TX_OFFLOAD_SECURITY)
 		roc_nix_inl_dev_xaq_realloc(mp->pool_id);
 
+	/* Increase CQ size to Aura size to avoid CQ overflow and
+	 * then CPT buffer leak.
+	 */
+	if (dev->rx_offloads & RTE_ETH_RX_OFFLOAD_SECURITY)
+		nb_desc = nix_inl_cq_sz_clamp_up(nix, mp, nb_desc);
+
 	/* Setup ROC CQ */
 	cq = &dev->cqs[qid];
 	cq->qid = qid;
diff --git a/drivers/net/cnxk/cnxk_ethdev.h b/drivers/net/cnxk/cnxk_ethdev.h
index a4e96f0..4cb7c9e 100644
--- a/drivers/net/cnxk/cnxk_ethdev.h
+++ b/drivers/net/cnxk/cnxk_ethdev.h
@@ -530,7 +530,7 @@ int cnxk_nix_tx_queue_setup(struct rte_eth_dev *eth_dev, uint16_t qid,
 			    uint16_t nb_desc, uint16_t fp_tx_q_sz,
 			    const struct rte_eth_txconf *tx_conf);
 int cnxk_nix_rx_queue_setup(struct rte_eth_dev *eth_dev, uint16_t qid,
-			    uint16_t nb_desc, uint16_t fp_rx_q_sz,
+			    uint32_t nb_desc, uint16_t fp_rx_q_sz,
 			    const struct rte_eth_rxconf *rx_conf,
 			    struct rte_mempool *mp);
 int cnxk_nix_tx_queue_start(struct rte_eth_dev *eth_dev, uint16_t qid);
-- 
2.8.4


  parent reply	other threads:[~2022-06-16  9:25 UTC|newest]

Thread overview: 29+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-06-16  7:07 [PATCH 01/12] common/cnxk: use computed value for wqe skip Nithin Dabilpuram
2022-06-16  7:07 ` [PATCH 02/12] common/cnxk: avoid CPT backpressure due to errata Nithin Dabilpuram
2022-06-16  7:07 ` [PATCH 03/12] common/cnxk: add PFC support for VFs Nithin Dabilpuram
2022-06-23 16:01   ` Ray Kinsella
2022-06-16  7:07 ` [PATCH 04/12] common/cnxk: support same TC value across multiple queues Nithin Dabilpuram
2022-06-16  7:07 ` [PATCH 05/12] common/cnxk: enhance CPT parse header dump Nithin Dabilpuram
2022-06-16  7:07 ` [PATCH 06/12] common/cnxk: fix mbox structs to avoid unaligned access Nithin Dabilpuram
2022-06-16  7:07 ` [PATCH 07/12] net/cnxk: add SDP link status Nithin Dabilpuram
2022-06-16  7:07 ` [PATCH 08/12] net/cnxk: remove restriction on VFs for PFC config Nithin Dabilpuram
2022-06-16  7:07 ` [PATCH 09/12] net/cnxk: pfc class disable resulting in invalid behaviour Nithin Dabilpuram
2022-06-16  7:07 ` [PATCH 10/12] net/cnxk: resize CQ for Rx security for errata Nithin Dabilpuram
2022-06-16  8:50   ` Jerin Jacob
2022-06-16  7:07 ` [PATCH 11/12] net/cnxk: add SDP VF device ID for probe matching Nithin Dabilpuram
2022-06-16  7:07 ` [PATCH 12/12] event/cnxk: offset timestamp data only if enabled on port Nithin Dabilpuram
2022-06-16  8:45 ` [PATCH 01/12] common/cnxk: use computed value for wqe skip Jerin Jacob
2022-06-16  9:24 ` [PATCH v2 01/12] common/cnxk: use computed value for WQE skip Nithin Dabilpuram
2022-06-16  9:24   ` [PATCH v2 02/12] common/cnxk: avoid CPT backpressure due to errata Nithin Dabilpuram
2022-06-16  9:24   ` [PATCH v2 03/12] common/cnxk: add PFC support for VFs Nithin Dabilpuram
2022-06-16  9:24   ` [PATCH v2 04/12] common/cnxk: support same TC value across multiple queues Nithin Dabilpuram
2022-06-16  9:24   ` [PATCH v2 05/12] common/cnxk: enhance CPT parse header dump Nithin Dabilpuram
2022-06-16  9:24   ` [PATCH v2 06/12] common/cnxk: fix mbox structs to avoid unaligned access Nithin Dabilpuram
2022-06-16  9:24   ` [PATCH v2 07/12] net/cnxk: add SDP link status Nithin Dabilpuram
2022-06-16  9:24   ` [PATCH v2 08/12] net/cnxk: remove restriction on VFs for PFC config Nithin Dabilpuram
2022-06-16  9:24   ` [PATCH v2 09/12] net/cnxk: pfc class disable resulting in invalid behaviour Nithin Dabilpuram
2022-06-16  9:24   ` Nithin Dabilpuram [this message]
2022-06-16  9:24   ` [PATCH v2 11/12] net/cnxk: add SDP VF device ID for probe matching Nithin Dabilpuram
2022-06-16  9:24   ` [PATCH v2 12/12] event/cnxk: offset timestamp data only if enabled on port Nithin Dabilpuram
2022-06-16 10:30     ` Nithin Kumar Dabilpuram
2022-06-20 17:26   ` [PATCH v2 01/12] common/cnxk: use computed value for WQE skip Jerin Jacob

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=20220616092420.17861-10-ndabilpuram@marvell.com \
    --to=ndabilpuram@marvell.com \
    --cc=dev@dpdk.org \
    --cc=jerinj@marvell.com \
    --cc=kirankumark@marvell.com \
    --cc=skori@marvell.com \
    --cc=skoteshwar@marvell.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).