DPDK patches and discussions
 help / color / mirror / Atom feed
From: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
To: dev@dpdk.org
Cc: rasland@nvidia.com, matan@nvidia.com, orika@nvidia.com,
	thomas@monjalon.net, akozyrev@nvidia.com
Subject: [dpdk-dev] [PATCH v3 2/2] net/mlx5: add mbuf fast free offload support
Date: Fri, 22 Jan 2021 17:12:09 +0000	[thread overview]
Message-ID: <1611335529-26503-3-git-send-email-viacheslavo@nvidia.com> (raw)
In-Reply-To: <1611335529-26503-1-git-send-email-viacheslavo@nvidia.com>

This patch adds support of the mbuf fast free offload to the
transmit datapath. This offload allows freeing the mbufs on
transmit completion in the most efficient way. It requires
the all mbufs were allocated from the same pool, have
the reference counter value as 1, and have no any externally
attached buffers.

Signed-off-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
---
 doc/guides/nics/features/mlx5.ini      |  1 +
 doc/guides/nics/mlx5.rst               | 12 ++++++++++++
 doc/guides/rel_notes/release_21_02.rst |  2 +-
 drivers/net/mlx5/mlx5_rxtx.c           | 35 ++++++++++++++++++++++++++--------
 drivers/net/mlx5/mlx5_rxtx.h           |  1 +
 drivers/net/mlx5/mlx5_txq.c            |  6 ++++++
 6 files changed, 48 insertions(+), 9 deletions(-)

diff --git a/doc/guides/nics/features/mlx5.ini b/doc/guides/nics/features/mlx5.ini
index 9d14678..a997543 100644
--- a/doc/guides/nics/features/mlx5.ini
+++ b/doc/guides/nics/features/mlx5.ini
@@ -9,6 +9,7 @@ Link status          = Y
 Link status event    = Y
 Removal event        = Y
 Rx interrupt         = Y
+Fast mbuf free       = Y
 Queue start/stop     = Y
 Burst mode info      = Y
 MTU update           = Y
diff --git a/doc/guides/nics/mlx5.rst b/doc/guides/nics/mlx5.rst
index 1765952..e5f8f3f 100644
--- a/doc/guides/nics/mlx5.rst
+++ b/doc/guides/nics/mlx5.rst
@@ -74,6 +74,7 @@ Features
 - RX VLAN stripping.
 - TX VLAN insertion.
 - RX CRC stripping configuration.
+- TX mbuf fast free offload.
 - Promiscuous mode on PF and VF.
 - Multicast promiscuous mode on PF and VF.
 - Hardware checksum offloads.
@@ -353,6 +354,17 @@ Limitations
     for some NICs (such as ConnectX-6 Dx, ConnectX-6 Lx, and BlueField-2).
     The capability bit ``scatter_fcs_w_decap_disable`` shows NIC support.
 
+- TX mbuf fast free:
+
+  - fast free offload assumes the all mbufs being sent are originated from the
+    same memory pool and there is no any extra references to the mbufs (the
+    reference counter for each mbuf is equal 1 on tx_burst call). The latter
+    means there should be no any externally attached buffers in mbufs. It is
+    an application responsibility to provide the correct mbufs if the fast
+    free offload is engaged. The mlx5 PMD implicitly produces the mbufs with
+    externally attached buffers if MPRQ option is enabled, hence, the fast
+    free offload is neither supported nor advertised if there is MPRQ enabled.
+
 - Sample flow:
 
   - Supports ``RTE_FLOW_ACTION_TYPE_SAMPLE`` action only within NIC Rx and E-Switch steering domain.
diff --git a/doc/guides/rel_notes/release_21_02.rst b/doc/guides/rel_notes/release_21_02.rst
index f52468c..0410337 100644
--- a/doc/guides/rel_notes/release_21_02.rst
+++ b/doc/guides/rel_notes/release_21_02.rst
@@ -98,6 +98,7 @@ New Features
     mirroring action on E-Switch.
   * Enlarge the number of flow priorities to 21844(0 - 21843) for ingress or egress
     flow groups greater than 0 and for any transfer flow group.
+  * Add support for the Tx mbuf fast free offload.
 
 * **Updated Wangxun txgbe driver.**
 
@@ -127,7 +128,6 @@ New Features
   * Added support for aes-cbc sha256-128-hmac cipher combination in OCTEON TX2
     crypto PMD lookaside protocol offload for IPsec.
 
-
 Removed Items
 -------------
 
diff --git a/drivers/net/mlx5/mlx5_rxtx.c b/drivers/net/mlx5/mlx5_rxtx.c
index 97912dd..2e4b87c 100644
--- a/drivers/net/mlx5/mlx5_rxtx.c
+++ b/drivers/net/mlx5/mlx5_rxtx.c
@@ -1889,6 +1889,8 @@ enum mlx5_txcmp_code {
 /**
  * Free the mbufs from the linear array of pointers.
  *
+ * @param txq
+ *   Pointer to Tx queue structure.
  * @param pkts
  *   Pointer to array of packets to be free.
  * @param pkts_n
@@ -1898,7 +1900,8 @@ enum mlx5_txcmp_code {
  *   compile time and may be used for optimization.
  */
 static __rte_always_inline void
-mlx5_tx_free_mbuf(struct rte_mbuf **__rte_restrict pkts,
+mlx5_tx_free_mbuf(struct mlx5_txq_data *__rte_restrict txq,
+		  struct rte_mbuf **__rte_restrict pkts,
 		  unsigned int pkts_n,
 		  unsigned int olx __rte_unused)
 {
@@ -1914,6 +1917,16 @@ enum mlx5_txcmp_code {
 	 */
 	MLX5_ASSERT(pkts);
 	MLX5_ASSERT(pkts_n);
+	/*
+	 * Free mbufs directly to the pool in bulk
+	 * if fast free offload is engaged
+	 */
+	if (!MLX5_TXOFF_CONFIG(MULTI) && txq->fast_free) {
+		mbuf = *pkts;
+		pool = mbuf->pool;
+		rte_mempool_put_bulk(pool, (void *)pkts, pkts_n);
+		return;
+	}
 	for (;;) {
 		for (;;) {
 			/*
@@ -1995,11 +2008,12 @@ enum mlx5_txcmp_code {
  * on the tx_burst completion.
  */
 static __rte_noinline void
-__mlx5_tx_free_mbuf(struct rte_mbuf **__rte_restrict pkts,
+__mlx5_tx_free_mbuf(struct mlx5_txq_data *__rte_restrict txq,
+		    struct rte_mbuf **__rte_restrict pkts,
 		    unsigned int pkts_n,
 		    unsigned int olx __rte_unused)
 {
-	mlx5_tx_free_mbuf(pkts, pkts_n, olx);
+	mlx5_tx_free_mbuf(txq, pkts, pkts_n, olx);
 }
 
 /**
@@ -2033,7 +2047,8 @@ enum mlx5_txcmp_code {
 		part = RTE_MIN(part, n_elts);
 		MLX5_ASSERT(part);
 		MLX5_ASSERT(part <= txq->elts_s);
-		mlx5_tx_free_mbuf(&txq->elts[txq->elts_tail & txq->elts_m],
+		mlx5_tx_free_mbuf(txq,
+				  &txq->elts[txq->elts_tail & txq->elts_m],
 				  part, olx);
 		txq->elts_tail += part;
 		n_elts -= part;
@@ -5184,7 +5199,7 @@ enum mlx5_txcmp_code {
 	txq->stats.opackets += loc.pkts_sent;
 #endif
 	if (MLX5_TXOFF_CONFIG(INLINE) && loc.mbuf_free)
-		__mlx5_tx_free_mbuf(pkts, loc.mbuf_free, olx);
+		__mlx5_tx_free_mbuf(txq, pkts, loc.mbuf_free, olx);
 	return loc.pkts_sent;
 }
 
@@ -5829,17 +5844,19 @@ enum mlx5_txcmp_code {
 
 int
 mlx5_tx_burst_mode_get(struct rte_eth_dev *dev,
-		       uint16_t tx_queue_id __rte_unused,
+		       uint16_t tx_queue_id,
 		       struct rte_eth_burst_mode *mode)
 {
 	eth_tx_burst_t pkt_burst = dev->tx_pkt_burst;
+	struct mlx5_priv *priv = dev->data->dev_private;
+	struct mlx5_txq_data *txq = (*priv->txqs)[tx_queue_id];
 	unsigned int i, olx;
 
 	for (i = 0; i < RTE_DIM(txoff_func); i++) {
 		if (pkt_burst == txoff_func[i].func) {
 			olx = txoff_func[i].olx;
 			snprintf(mode->info, sizeof(mode->info),
-				 "%s%s%s%s%s%s%s%s%s",
+				 "%s%s%s%s%s%s%s%s%s%s",
 				 (olx & MLX5_TXOFF_CONFIG_EMPW) ?
 				 ((olx & MLX5_TXOFF_CONFIG_MPW) ?
 				 "Legacy MPW" : "Enhanced MPW") : "No MPW",
@@ -5858,7 +5875,9 @@ enum mlx5_txcmp_code {
 				 (olx & MLX5_TXOFF_CONFIG_METADATA) ?
 				 " + METADATA" : "",
 				 (olx & MLX5_TXOFF_CONFIG_TXPP) ?
-				 " + TXPP" : "");
+				 " + TXPP" : "",
+				 (txq && txq->fast_free) ?
+				 " + Fast Free" : "");
 			return 0;
 		}
 	}
diff --git a/drivers/net/mlx5/mlx5_rxtx.h b/drivers/net/mlx5/mlx5_rxtx.h
index 9dac408..6432030 100644
--- a/drivers/net/mlx5/mlx5_rxtx.h
+++ b/drivers/net/mlx5/mlx5_rxtx.h
@@ -243,6 +243,7 @@ struct mlx5_txq_data {
 	uint16_t vlan_en:1; /* VLAN insertion in WQE is supported. */
 	uint16_t db_nc:1; /* Doorbell mapped to non-cached region. */
 	uint16_t db_heu:1; /* Doorbell heuristic write barrier. */
+	uint16_t fast_free:1; /* mbuf fast free on Tx is enabled. */
 	uint16_t inlen_send; /* Ordinary send data inline size. */
 	uint16_t inlen_empw; /* eMPW max packet size to inline. */
 	uint16_t inlen_mode; /* Minimal data length to inline. */
diff --git a/drivers/net/mlx5/mlx5_txq.c b/drivers/net/mlx5/mlx5_txq.c
index b81bb4a..bf5e9a9 100644
--- a/drivers/net/mlx5/mlx5_txq.c
+++ b/drivers/net/mlx5/mlx5_txq.c
@@ -123,6 +123,8 @@
 				     DEV_TX_OFFLOAD_GRE_TNL_TSO |
 				     DEV_TX_OFFLOAD_GENEVE_TNL_TSO);
 	}
+	if (!config->mprq.enabled)
+		offloads |= DEV_TX_OFFLOAD_MBUF_FAST_FREE;
 	return offloads;
 }
 
@@ -800,6 +802,10 @@
 	bool vlan_inline;
 	unsigned int temp;
 
+	txq_ctrl->txq.fast_free =
+		!!((txq_ctrl->txq.offloads & DEV_TX_OFFLOAD_MBUF_FAST_FREE) &&
+		   !(txq_ctrl->txq.offloads & DEV_TX_OFFLOAD_MULTI_SEGS) &&
+		   !config->mprq.enabled);
 	if (config->txqs_inline == MLX5_ARG_UNSET)
 		txqs_inline =
 #if defined(RTE_ARCH_ARM64)
-- 
1.8.3.1


  parent reply	other threads:[~2021-01-22 17:12 UTC|newest]

Thread overview: 16+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-12-18 17:14 [dpdk-dev] [PATCH 0/2] net/mlx5: add mbuf fast free Tx offload Viacheslav Ovsiienko
2020-12-18 17:14 ` [dpdk-dev] [PATCH 1/2] net/mlx5: optimize inline mbuf freeing Viacheslav Ovsiienko
2020-12-18 17:14 ` [dpdk-dev] [PATCH 2/2] net/mlx5: add mbuf fast free offload support Viacheslav Ovsiienko
2020-12-18 17:59   ` Thomas Monjalon
2021-01-06  8:34 ` [dpdk-dev] [PATCH v2 0/2] " Viacheslav Ovsiienko
2021-01-06  8:34   ` [dpdk-dev] [PATCH v2 1/2] net/mlx5: optimize inline mbuf freeing Viacheslav Ovsiienko
2021-01-06  8:34   ` [dpdk-dev] [PATCH v2 2/2] net/mlx5: add mbuf fast free offload support Viacheslav Ovsiienko
2021-01-22 17:12 ` [dpdk-dev] [PATCH v3 0/2] " Viacheslav Ovsiienko
2021-01-22 17:12   ` [dpdk-dev] [PATCH v3 1/2] net/mlx5: optimize inline mbuf freeing Viacheslav Ovsiienko
2021-01-27 12:44     ` [dpdk-dev] [dpdk-stable] " Ferruh Yigit
2021-01-27 12:48       ` Ferruh Yigit
2021-01-28  9:14       ` Slava Ovsiienko
2021-01-28  9:34         ` Thomas Monjalon
2021-01-22 17:12   ` Viacheslav Ovsiienko [this message]
2021-01-26 10:34   ` [dpdk-dev] [PATCH v3 0/2] add mbuf fast free offload support Matan Azrad
2021-01-27  9:31   ` Raslan Darawsheh

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=1611335529-26503-3-git-send-email-viacheslavo@nvidia.com \
    --to=viacheslavo@nvidia.com \
    --cc=akozyrev@nvidia.com \
    --cc=dev@dpdk.org \
    --cc=matan@nvidia.com \
    --cc=orika@nvidia.com \
    --cc=rasland@nvidia.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).