Acked-by: Viacheslav Ovsiienko <[email protected]>
> -----Original Message-----
> From: Vincent Jardin <[email protected]>
> Sent: Sunday, March 22, 2026 3:46 PM
> To: [email protected]
> Cc: Raslan Darawsheh <[email protected]>; NBU-Contact-Thomas Monjalon
> (EXTERNAL) <[email protected]>; [email protected];
> Dariusz Sosnowski <[email protected]>; Slava Ovsiienko
> <[email protected]>; Bing Zhao <[email protected]>; Ori Kam
> <[email protected]>; Suanming Mou <[email protected]>; Matan Azrad
> <[email protected]>; [email protected];
> [email protected]; Vincent Jardin <[email protected]>
> Subject: [PATCH v4 05/10] net/mlx5: support per-queue rate limiting
>
> Wire rte_eth_set_queue_rate_limit() to the mlx5 PMD. The callback allocates a
> per-queue PP index with the requested data rate, then modifies the live SQ via
> modify_bitmask bit 0 to apply the new packet_pacing_rate_limit_index — no
> queue teardown required.
>
> Setting tx_rate=0 clears the PP index on the SQ and frees it.
>
> Capability check uses hca_attr.qos.packet_pacing directly (not dev_cap.txpp_en
> which requires Clock Queue prerequisites). This allows per-queue rate limiting
> without the tx_pp devarg.
>
> The callback rejects hairpin queues and queues whose SQ is not yet created.
>
> testpmd usage (no testpmd changes needed):
> set port 0 queue 0 rate 1000
> set port 0 queue 1 rate 5000
> set port 0 queue 0 rate 0 # disable
>
> Supported hardware:
> - ConnectX-6 Dx: full support, per-SQ rate via HW rate table
> - ConnectX-7/8: full support, coexists with wait-on-time scheduling
> - BlueField-2/3: full support as DPU rep ports
>
> Not supported:
> - ConnectX-5: packet_pacing exists but dynamic SQ modify may not
> work on all firmware versions
> - ConnectX-4 Lx and earlier: no packet_pacing capability
>
> Signed-off-by: Vincent Jardin <[email protected]>
> ---
> doc/guides/nics/features/mlx5.ini | 1 +
> doc/guides/nics/mlx5.rst | 54 ++++++++++++++
> drivers/net/mlx5/mlx5.c | 2 +
> drivers/net/mlx5/mlx5_tx.h | 2 +
> drivers/net/mlx5/mlx5_txq.c | 118 ++++++++++++++++++++++++++++++
> 5 files changed, 177 insertions(+)
>
> diff --git a/doc/guides/nics/features/mlx5.ini
> b/doc/guides/nics/features/mlx5.ini
> index 4f9c4c309b..3b3eda28b8 100644
> --- a/doc/guides/nics/features/mlx5.ini
> +++ b/doc/guides/nics/features/mlx5.ini
> @@ -30,6 +30,7 @@ Inner RSS = Y
> SR-IOV = Y
> VLAN filter = Y
> Flow control = Y
> +Rate limitation = Y
> CRC offload = Y
> VLAN offload = Y
> L3 checksum offload = Y
> diff --git a/doc/guides/nics/mlx5.rst b/doc/guides/nics/mlx5.rst index
> 6bb8c07353..c72a60f084 100644
> --- a/doc/guides/nics/mlx5.rst
> +++ b/doc/guides/nics/mlx5.rst
> @@ -580,6 +580,60 @@ for an additional list of options shared with other
> mlx5 drivers.
> (with ``tx_pp``) and ConnectX-7+ (wait-on-time) scheduling modes.
> The default value is zero.
>
> +.. _mlx5_per_queue_rate_limit:
> +
> +Per-Queue Tx Rate Limiting
> +~~~~~~~~~~~~~~~~~~~~~~~~~~
> +
> +The mlx5 PMD supports per-queue Tx rate limiting via the standard
> +ethdev API ``rte_eth_set_queue_rate_limit()`` and
> ``rte_eth_get_queue_rate_limit()``.
> +
> +This feature uses the hardware packet pacing mechanism to enforce a
> +data rate on individual TX queues without tearing down the queue. The
> +rate is specified in Mbps.
> +
> +**Requirements:**
> +
> +- ConnectX-6 Dx or later with ``packet_pacing`` HCA capability.
> +- The DevX path must be used (default). The legacy Verbs path
> + (``dv_flow_en=0``) does not support dynamic SQ modification and
> + returns ``-EINVAL``.
> +- The queue must be started (SQ in RDY state) before setting a rate.
> +
> +**Supported hardware:**
> +
> +- ConnectX-6 Dx: per-SQ rate via HW rate table.
> +- ConnectX-7/8: full support, coexists with wait-on-time scheduling.
> +- BlueField-2/3: full support as DPU rep ports.
> +
> +**Not supported:**
> +
> +- ConnectX-5: ``packet_pacing`` exists but dynamic SQ modify may not
> + work on all firmware versions.
> +- ConnectX-4 Lx and earlier: no ``packet_pacing`` capability.
> +
> +**Rate table sharing:**
> +
> +The hardware rate table has a limited number of entries (typically 128
> +on
> +ConnectX-6 Dx). When multiple queues are configured with identical rate
> +parameters, the kernel mlx5 driver shares a single rate table entry
> +across them. Each queue still has its own independent SQ and enforces
> +the rate independently — queues are never merged. The rate cap applies per-
> queue:
> +if two queues share the same 1000 Mbps entry, each can send up to
> +1000 Mbps independently, they do not share a combined budget.
> +
> +This sharing is transparent and only affects table capacity: 128
> +entries can serve thousands of queues as long as many use the same
> +rate. Queues with different rates consume separate entries.
> +
> +**Usage with testpmd:**
> +
> +.. code-block:: console
> +
> + testpmd> set port 0 queue 0 rate 1000
> + testpmd> show port 0 queue 0 rate
> + testpmd> set port 0 queue 0 rate 0
> +
> - ``tx_vec_en`` parameter [int]
>
> A nonzero value enables Tx vector with ConnectX-5 NICs and above.
> diff --git a/drivers/net/mlx5/mlx5.c b/drivers/net/mlx5/mlx5.c index
> e795948187..e718f0fa8c 100644
> --- a/drivers/net/mlx5/mlx5.c
> +++ b/drivers/net/mlx5/mlx5.c
> @@ -2621,6 +2621,7 @@ const struct eth_dev_ops mlx5_dev_ops = {
> .map_aggr_tx_affinity = mlx5_map_aggr_tx_affinity,
> .rx_metadata_negotiate = mlx5_flow_rx_metadata_negotiate,
> .get_restore_flags = mlx5_get_restore_flags,
> + .set_queue_rate_limit = mlx5_set_queue_rate_limit,
> };
>
> /* Available operations from secondary process. */ @@ -2714,6 +2715,7 @@
> const struct eth_dev_ops mlx5_dev_ops_isolate = {
> .count_aggr_ports = mlx5_count_aggr_ports,
> .map_aggr_tx_affinity = mlx5_map_aggr_tx_affinity,
> .get_restore_flags = mlx5_get_restore_flags,
> + .set_queue_rate_limit = mlx5_set_queue_rate_limit,
> };
>
> /**
> diff --git a/drivers/net/mlx5/mlx5_tx.h b/drivers/net/mlx5/mlx5_tx.h index
> 51f330454a..975ff57acd 100644
> --- a/drivers/net/mlx5/mlx5_tx.h
> +++ b/drivers/net/mlx5/mlx5_tx.h
> @@ -222,6 +222,8 @@ struct mlx5_txq_ctrl *mlx5_txq_get(struct rte_eth_dev
> *dev, uint16_t idx); int mlx5_txq_release(struct rte_eth_dev *dev, uint16_t
> idx); int mlx5_txq_releasable(struct rte_eth_dev *dev, uint16_t idx); int
> mlx5_txq_verify(struct rte_eth_dev *dev);
> +int mlx5_set_queue_rate_limit(struct rte_eth_dev *dev, uint16_t queue_idx,
> + uint32_t tx_rate);
> int mlx5_txq_get_sqn(struct mlx5_txq_ctrl *txq); void
> mlx5_txq_alloc_elts(struct mlx5_txq_ctrl *txq_ctrl); void
> mlx5_txq_free_elts(struct mlx5_txq_ctrl *txq_ctrl); diff --git
> a/drivers/net/mlx5/mlx5_txq.c b/drivers/net/mlx5/mlx5_txq.c index
> 3356c89758..ce08363ca9 100644
> --- a/drivers/net/mlx5/mlx5_txq.c
> +++ b/drivers/net/mlx5/mlx5_txq.c
> @@ -1363,6 +1363,124 @@ mlx5_txq_release(struct rte_eth_dev *dev,
> uint16_t idx)
> return 0;
> }
>
> +/**
> + * Set per-queue packet pacing rate limit.
> + *
> + * @param dev
> + * Pointer to Ethernet device.
> + * @param queue_idx
> + * TX queue index.
> + * @param tx_rate
> + * TX rate in Mbps, 0 to disable rate limiting.
> + *
> + * @return
> + * 0 on success, a negative errno value otherwise and rte_errno is set.
> + */
> +int
> +mlx5_set_queue_rate_limit(struct rte_eth_dev *dev, uint16_t queue_idx,
> + uint32_t tx_rate)
> +{
> + struct mlx5_priv *priv = dev->data->dev_private;
> + struct mlx5_dev_ctx_shared *sh = priv->sh;
> + struct mlx5_txq_ctrl *txq_ctrl;
> + struct mlx5_devx_obj *sq_devx;
> + struct mlx5_devx_modify_sq_attr sq_attr = { 0 };
> + struct mlx5_txq_rate_limit new_rate_limit = { 0 };
> + int ret;
> +
> + if (!sh->cdev->config.hca_attr.qos.packet_pacing) {
> + DRV_LOG(ERR, "Port %u packet pacing not supported.",
> + dev->data->port_id);
> + rte_errno = ENOTSUP;
> + return -rte_errno;
> + }
> + if (priv->txqs == NULL || (*priv->txqs)[queue_idx] == NULL) {
> + DRV_LOG(ERR, "Port %u Tx queue %u not configured.",
> + dev->data->port_id, queue_idx);
> + rte_errno = EINVAL;
> + return -rte_errno;
> + }
> + txq_ctrl = container_of((*priv->txqs)[queue_idx],
> + struct mlx5_txq_ctrl, txq);
> + if (txq_ctrl->is_hairpin) {
> + DRV_LOG(ERR, "Port %u Tx queue %u is hairpin.",
> + dev->data->port_id, queue_idx);
> + rte_errno = EINVAL;
> + return -rte_errno;
> + }
> + if (txq_ctrl->obj == NULL) {
> + DRV_LOG(ERR, "Port %u Tx queue %u not initialized.",
> + dev->data->port_id, queue_idx);
> + rte_errno = EINVAL;
> + return -rte_errno;
> + }
> + /*
> + * For non-hairpin queues the SQ DevX object lives in
> + * obj->sq_obj.sq (used by DevX/HWS mode), while hairpin
> + * queues use obj->sq directly. These are different members
> + * of a union inside mlx5_txq_obj.
> + */
> + sq_devx = txq_ctrl->obj->sq_obj.sq;
> + if (sq_devx == NULL) {
> + DRV_LOG(ERR, "Port %u Tx queue %u SQ not ready.",
> + dev->data->port_id, queue_idx);
> + rte_errno = EINVAL;
> + return -rte_errno;
> + }
> + if (dev->data->tx_queue_state[queue_idx] !=
> + RTE_ETH_QUEUE_STATE_STARTED) {
> + DRV_LOG(ERR,
> + "Port %u Tx queue %u is not started, stop traffic before
> setting rate.",
> + dev->data->port_id, queue_idx);
> + rte_errno = EINVAL;
> + return -rte_errno;
> + }
> + if (tx_rate == 0) {
> + /* Disable rate limiting. */
> + if (txq_ctrl->rate_limit.pp_id == 0)
> + return 0; /* Already disabled. */
> + sq_attr.sq_state = MLX5_SQC_STATE_RDY;
> + sq_attr.state = MLX5_SQC_STATE_RDY;
> + sq_attr.rl_update = 1;
> + sq_attr.packet_pacing_rate_limit_index = 0;
> + ret = mlx5_devx_cmd_modify_sq(sq_devx, &sq_attr);
> + if (ret) {
> + DRV_LOG(ERR,
> + "Port %u Tx queue %u failed to clear rate.",
> + dev->data->port_id, queue_idx);
> + rte_errno = -ret;
> + return ret;
> + }
> + mlx5_txq_free_pp_rate_limit(&txq_ctrl->rate_limit);
> + DRV_LOG(DEBUG, "Port %u Tx queue %u rate limit disabled.",
> + dev->data->port_id, queue_idx);
> + return 0;
> + }
> + /* Allocate a new PP index for the requested rate into a temp. */
> + ret = mlx5_txq_alloc_pp_rate_limit(sh, &new_rate_limit, tx_rate);
> + if (ret)
> + return ret;
> + /* Modify live SQ to use the new PP index. */
> + sq_attr.sq_state = MLX5_SQC_STATE_RDY;
> + sq_attr.state = MLX5_SQC_STATE_RDY;
> + sq_attr.rl_update = 1;
> + sq_attr.packet_pacing_rate_limit_index = new_rate_limit.pp_id;
> + ret = mlx5_devx_cmd_modify_sq(sq_devx, &sq_attr);
> + if (ret) {
> + DRV_LOG(ERR, "Port %u Tx queue %u failed to set rate %u
> Mbps.",
> + dev->data->port_id, queue_idx, tx_rate);
> + mlx5_txq_free_pp_rate_limit(&new_rate_limit);
> + rte_errno = -ret;
> + return ret;
> + }
> + /* SQ updated — release old PP context, install new one. */
> + mlx5_txq_free_pp_rate_limit(&txq_ctrl->rate_limit);
> + txq_ctrl->rate_limit = new_rate_limit;
> + DRV_LOG(DEBUG, "Port %u Tx queue %u rate set to %u Mbps (PP idx
> %u).",
> + dev->data->port_id, queue_idx, tx_rate, txq_ctrl-
> >rate_limit.pp_id);
> + return 0;
> +}
> +
> /**
> * Verify if the queue can be released.
> *
> --
> 2.43.0
