[dpdk-dev] [PATCH 1/2] common/mlx5: support vDPA completion queue moderation

This patch introduces new parameters for VirtQ CQ moderation, used for
performance tuning.

Signed-off-by: Xueming Li <redacted>
---
 drivers/common/mlx5/mlx5_devx_cmds.c | 3 +++
 drivers/common/mlx5/mlx5_devx_cmds.h | 3 +++
 drivers/common/mlx5/mlx5_prm.h       | 6 +++++-
 3 files changed, 11 insertions(+), 1 deletion(-)

diff --git a/drivers/common/mlx5/mlx5_devx_cmds.c b/drivers/common/mlx5/mlx5_devx_cmds.c
index 12f51a940c..3bf5279ff1 100644
--- a/drivers/common/mlx5/mlx5_devx_cmds.c
+++ b/drivers/common/mlx5/mlx5_devx_cmds.c

@@ -1662,6 +1662,9 @@ mlx5_devx_cmd_create_virtq(void *ctx,
 	MLX5_SET64(virtio_q, virtctx, umem_3_offset, attr->umems[2].offset);
 	MLX5_SET(virtio_q, virtctx, counter_set_id, attr->counters_obj_id);
 	MLX5_SET(virtio_q, virtctx, pd, attr->pd);
+	MLX5_SET(virtio_q, virtctx, queue_period_mode, attr->hw_latency_mode);
+	MLX5_SET(virtio_q, virtctx, queue_period_us, attr->hw_max_latency_us);
+	MLX5_SET(virtio_q, virtctx, queue_max_count, attr->hw_max_pending_comp);
 	MLX5_SET(virtio_net_q, virtq, tisn_or_qpn, attr->tis_id);
 	virtq_obj->obj = mlx5_glue->devx_obj_create(ctx, in, sizeof(in), out,
 						    sizeof(out));

diff --git a/drivers/common/mlx5/mlx5_devx_cmds.h b/drivers/common/mlx5/mlx5_devx_cmds.h
index b335b7c82c..94e9bbb906 100644
--- a/drivers/common/mlx5/mlx5_devx_cmds.h
+++ b/drivers/common/mlx5/mlx5_devx_cmds.h

@@ -302,6 +302,9 @@ struct mlx5_devx_virtq_attr {
 	uint32_t rx_csum:1;
 	uint32_t event_mode:3;
 	uint32_t state:4;
+	uint32_t hw_latency_mode:2;
+	uint32_t hw_max_latency_us:12;
+	uint32_t hw_max_pending_comp:16;
 	uint32_t dirty_bitmap_dump_enable:1;
 	uint32_t dirty_bitmap_mkey;
 	uint32_t dirty_bitmap_size;

diff --git a/drivers/common/mlx5/mlx5_prm.h b/drivers/common/mlx5/mlx5_prm.h
index 8c9b53ce10..7d5cf961b1 100644
--- a/drivers/common/mlx5/mlx5_prm.h
+++ b/drivers/common/mlx5/mlx5_prm.h

@@ -2379,7 +2379,11 @@ struct mlx5_ifc_virtio_q_bits {
 	u8 counter_set_id[0x20];
 	u8 reserved_at_320[0x8];
 	u8 pd[0x18];
-	u8 reserved_at_340[0xc0];
+	u8 reserved_at_340[0x2];
+	u8 queue_period_mode[0x2];
+	u8 queue_period_us[0xc];
+	u8 queue_max_count[0x10];
+	u8 reserved_at_360[0xa0];
 };
 
 struct mlx5_ifc_virtio_net_q_bits {

-- 
2.25.1

The next parameters control the HW queue moderation feature.
This feature helps to control the traffic performance and latency
tradeoff.

Each packet completion report from HW to SW requires CQ processing by SW
and triggers interrupt for the guest driver. Interrupt report and
handling cost CPU cycles and time and the amount of this affects
directly on packet performance and latency.

hw_latency_mode parameters [int]
  0, HW default.
  1, Latency is counted from the first packet completion report.
  2, Latency is counted from the last packet completion.
hw_max_latency_us parameters [int]
  0 - 4095, The maximum time in microseconds that packet completion
report can be delayed.
hw_max_pending_comp parameter [int]
  0 - 65535, The maximum number of pending packets completions in an HW
queue.

Signed-off-by: Xueming Li <redacted>
---
 doc/guides/vdpadevs/mlx5.rst        | 24 ++++++++++++++++++++++++
 drivers/vdpa/mlx5/mlx5_vdpa.c       |  6 ++++++
 drivers/vdpa/mlx5/mlx5_vdpa.h       |  3 +++
 drivers/vdpa/mlx5/mlx5_vdpa_virtq.c |  3 +++
 4 files changed, 36 insertions(+)

diff --git a/doc/guides/vdpadevs/mlx5.rst b/doc/guides/vdpadevs/mlx5.rst
index 3a6d88362d..587652b3ae 100644
--- a/doc/guides/vdpadevs/mlx5.rst
+++ b/doc/guides/vdpadevs/mlx5.rst

@@ -135,6 +135,30 @@ Driver options
   interrupts are configured to the device in order to notify traffic for the
   driver. Default value is 2s.
 
+- ``hw_latency_mode`` parameter [int]
+
+  The completion queue moderation mode:
+
+  - 0, HW default.
+
+  - 1, Latency is counted from the first packet completion report.
+
+  - 2, Latency is counted from the last packet completion.
+
+- ``hw_max_latency_us`` parameter [int]
+
+  - 1 - 4095, The maximum time in microseconds that packet completion report
+    can be delayed.
+
+  - 0, HW default.
+
+- ``hw_max_pending_comp`` parameter [int]
+
+  - 1 - 65535, The maximum number of pending packets completions in an HW queue.
+
+  - 0, HW default.
+
+
 Error handling
 ^^^^^^^^^^^^^^
 

diff --git a/drivers/vdpa/mlx5/mlx5_vdpa.c b/drivers/vdpa/mlx5/mlx5_vdpa.c
index b64f364eb7..bb9477cc52 100644
--- a/drivers/vdpa/mlx5/mlx5_vdpa.c
+++ b/drivers/vdpa/mlx5/mlx5_vdpa.c

@@ -630,6 +630,12 @@ mlx5_vdpa_args_check_handler(const char *key, const char *val, void *opaque)
 		priv->event_us = (uint32_t)tmp;
 	} else if (strcmp(key, "no_traffic_time") == 0) {
 		priv->no_traffic_time_s = (uint32_t)tmp;
+	} else if (strcmp(key, "hw_latency_mode") == 0) {
+		priv->hw_latency_mode = (uint32_t)tmp;
+	} else if (strcmp(key, "hw_max_latency_us") == 0) {
+		priv->hw_max_latency_us = (uint32_t)tmp;
+	} else if (strcmp(key, "hw_max_pending_comp") == 0) {
+		priv->hw_max_pending_comp = (uint32_t)tmp;
 	} else {
 		DRV_LOG(WARNING, "Invalid key %s.", key);
 	}

diff --git a/drivers/vdpa/mlx5/mlx5_vdpa.h b/drivers/vdpa/mlx5/mlx5_vdpa.h
index d039ada65b..9d2d9c1cd5 100644
--- a/drivers/vdpa/mlx5/mlx5_vdpa.h
+++ b/drivers/vdpa/mlx5/mlx5_vdpa.h

@@ -134,6 +134,9 @@ struct mlx5_vdpa_priv {
 	uint32_t event_us;
 	uint32_t timer_delay_us;
 	uint32_t no_traffic_time_s;
+	uint8_t hw_latency_mode; /* Hardware CQ moderation mode. */
+	uint16_t hw_max_latency_us; /* Hardware CQ moderation period in usec. */
+	uint16_t hw_max_pending_comp; /* Hardware CQ moderation counter. */
 	struct rte_vdpa_device *vdev; /* vDPA device. */
 	int vid; /* vhost device id. */
 	struct ibv_context *ctx; /* Device context. */

diff --git a/drivers/vdpa/mlx5/mlx5_vdpa_virtq.c b/drivers/vdpa/mlx5/mlx5_vdpa_virtq.c
index 3e882e4000..332753fd62 100644
--- a/drivers/vdpa/mlx5/mlx5_vdpa_virtq.c
+++ b/drivers/vdpa/mlx5/mlx5_vdpa_virtq.c

@@ -327,6 +327,9 @@ mlx5_vdpa_virtq_setup(struct mlx5_vdpa_priv *priv, int index)
 	attr.tis_id = priv->tiss[(index / 2) % priv->num_lag_ports]->id;
 	attr.queue_index = index;
 	attr.pd = priv->pdn;
+	attr.hw_latency_mode = priv->hw_latency_mode;
+	attr.hw_max_latency_us = priv->hw_max_latency_us;
+	attr.hw_max_pending_comp = priv->hw_max_pending_comp;
 	virtq->virtq = mlx5_devx_cmd_create_virtq(priv->ctx, &attr);
 	virtq->priv = priv;
 	if (!virtq->virtq)

-- 
2.25.1

On 1/6/21 4:06 AM, Xueming Li wrote:

This patch introduces new parameters for VirtQ CQ moderation, used for
performance tuning.

Signed-off-by: Xueming Li <redacted>
---
 drivers/common/mlx5/mlx5_devx_cmds.c | 3 +++
 drivers/common/mlx5/mlx5_devx_cmds.h | 3 +++
 drivers/common/mlx5/mlx5_prm.h       | 6 +++++-
 3 files changed, 11 insertions(+), 1 deletion(-)

Reviewed-by: Maxime Coquelin <redacted>

Thanks,
Maxime

On 1/6/21 4:06 AM, Xueming Li wrote:

The next parameters control the HW queue moderation feature.
This feature helps to control the traffic performance and latency
tradeoff.

Each packet completion report from HW to SW requires CQ processing by SW
and triggers interrupt for the guest driver. Interrupt report and
handling cost CPU cycles and time and the amount of this affects
directly on packet performance and latency.

hw_latency_mode parameters [int]
  0, HW default.
  1, Latency is counted from the first packet completion report.
  2, Latency is counted from the last packet completion.
hw_max_latency_us parameters [int]
  0 - 4095, The maximum time in microseconds that packet completion
report can be delayed.
hw_max_pending_comp parameter [int]
  0 - 65535, The maximum number of pending packets completions in an HW
queue.

Signed-off-by: Xueming Li <redacted>
---
 doc/guides/vdpadevs/mlx5.rst        | 24 ++++++++++++++++++++++++
 drivers/vdpa/mlx5/mlx5_vdpa.c       |  6 ++++++
 drivers/vdpa/mlx5/mlx5_vdpa.h       |  3 +++
 drivers/vdpa/mlx5/mlx5_vdpa_virtq.c |  3 +++
 4 files changed, 36 insertions(+)

Reviewed-by: Maxime Coquelin <redacted>

Thanks,
Maxime

On 1/6/21 4:06 AM, Xueming Li wrote:

This patch introduces new parameters for VirtQ CQ moderation, used for
performance tuning.

Signed-off-by: Xueming Li <redacted>
---
 drivers/common/mlx5/mlx5_devx_cmds.c | 3 +++
 drivers/common/mlx5/mlx5_devx_cmds.h | 3 +++
 drivers/common/mlx5/mlx5_prm.h       | 6 +++++-
 3 files changed, 11 insertions(+), 1 deletion(-)

Series applied to dpdk-next-virtio/main.

Thanks,
Maxime

On 1/6/21 4:06 AM, Xueming Li wrote:

The next parameters control the HW queue moderation feature.
This feature helps to control the traffic performance and latency
tradeoff.

Each packet completion report from HW to SW requires CQ processing by SW
and triggers interrupt for the guest driver. Interrupt report and
handling cost CPU cycles and time and the amount of this affects
directly on packet performance and latency.

hw_latency_mode parameters [int]
  0, HW default.
  1, Latency is counted from the first packet completion report.
  2, Latency is counted from the last packet completion.
hw_max_latency_us parameters [int]
  0 - 4095, The maximum time in microseconds that packet completion
report can be delayed.
hw_max_pending_comp parameter [int]
  0 - 65535, The maximum number of pending packets completions in an HW
queue.

Signed-off-by: Xueming Li <redacted>
---
 doc/guides/vdpadevs/mlx5.rst        | 24 ++++++++++++++++++++++++
 drivers/vdpa/mlx5/mlx5_vdpa.c       |  6 ++++++
 drivers/vdpa/mlx5/mlx5_vdpa.h       |  3 +++
 drivers/vdpa/mlx5/mlx5_vdpa_virtq.c |  3 +++
 4 files changed, 36 insertions(+)

Series applied to dpdk-next-virtio/main.

Thanks,
Maxime