[dpdk-dev] [PATCH 2/7] net/mlx5: add Tx datapath related devargs

Viacheslav Ovsiienko viacheslavo at mellanox.com
Thu Jul 4 18:29:22 CEST 2019
Previous message: [dpdk-dev] [PATCH v4 0/8] net/mlx5: consolidate Tx datapath
Next message: [dpdk-dev] [PATCH 3/7] net/mlx5: update Tx datapath definitions
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
This patch introduces new mlx5 PMD devarg options:

- txq_inline_min - specifies minimal amount of data to be inlined into
  WQE during Tx operations. NICs may require this minimal data amount
  to operate correctly. The exact value may depend on NIC operation mode,
  requested offloads, etc.

- txq_inline_max - specifies the maximal packet length to be completely
  inlined into WQE Ethernet Segment for ordinary SEND method. If packet
  is larger the specified value, the packet data won't be copied by the
  driver at all, data buffer is addressed with a pointer. If packet length
  is less or equal all packet data will be copied into WQE.

- txq_inline_mpw - specifies the maximal packet length to be completely
  inlined into WQE for Enhanced MPW method.

Driver documentation is also updated.

Signed-off-by: Viacheslav Ovsiienko <viacheslavo at mellanox.com>
---
 doc/guides/nics/mlx5.rst | 168 +++++++++++++++++++++++++----------------------
 drivers/net/mlx5/mlx5.c  |  21 ++++++
 drivers/net/mlx5/mlx5.h  |   4 ++
 3 files changed, 113 insertions(+), 80 deletions(-)

diff --git a/doc/guides/nics/mlx5.rst b/doc/guides/nics/mlx5.rst
index 4cd6cf8..3f2a344 100644
--- a/doc/guides/nics/mlx5.rst
+++ b/doc/guides/nics/mlx5.rst
@@ -346,99 +346,107 @@ Run-time configuration
   greater or equal to this value. The default value is 12, valid only if
   ``mprq_en`` is set.
 
-- ``txq_inline`` parameter [int]
-
-  Amount of data to be inlined during TX operations. Improves latency.
-  Can improve PPS performance when PCI back pressure is detected and may be
-  useful for scenarios involving heavy traffic on many queues.
-
-  Because additional software logic is necessary to handle this mode, this
-  option should be used with care, as it can lower performance when back
-  pressure is not expected.
+- ``txq_inline_min`` parameter [int]
+
+  Minimal amount of data to be inlined into WQE during Tx operations. NICs
+  may require this minimal data amount to operate correctly. The exact value
+  may depend on NIC operation mode, requested offloads, etc.
+
+  If ``txq_inline_min`` key is present the specified value (may be aligned
+  by the driver in order not to exceed the limits and provide better descriptor
+  space utilization) will be used by the driver and it is guaranteed the
+  requested data bytes are inlined into the WQE beside other inline settings.
+
+  If ``txq_inline_min`` key is not present the value may be queried by the
+  driver from the NIC via DevX if this feature is available. If there is no DevX
+  enabled/supported the value 18 (supposing L2 header including VLAN) is set
+  for ConnectX-4, value 58 (supposing L2-L4 headers, required by configurations
+  over E-Switch) is set for ConnectX-4 Lx, and 0 is set by default for ConnectX-5
+  and newer NICs. If packet is shorter the ``txq_inline_min`` value, the entire
+  packet is inlined.
+
+  Please, note, this minimal data inlining disengages eMPW feature (Enhanced
+  Multi-Packet Write), because last one does not support partial packet inlining.
+  This is not very critical due to minimal data inlining is mostly required
+  by ConnectX-4 and ConnectX-4 Lx, these NICs does not support eMPW feature.
 
 - ``txqs_min_inline`` parameter [int]
 
-  Enable inline send only when the number of TX queues is greater or equal
+  Enable inline data send only when the number of TX queues is greater or equal
   to this value.
 
-  This option should be used in combination with ``txq_inline`` above.
-
-  On ConnectX-4, ConnectX-4 LX, ConnectX-5, ConnectX-6 and BlueField without
-  Enhanced MPW:
-
-        - Disabled by default.
-        - In case ``txq_inline`` is set recommendation is 4.
+  This option should be used in combination with ``txq_inline_max`` and
+  ``txq_inline_mpw`` below and does not affect ``txq_inline_min`` settings above.
 
-  On ConnectX-5, ConnectX-6 and BlueField with Enhanced MPW:
+  If this option is not specified the default value 8 is used.
 
-        - Set to 8 by default.
+  The data inlining consumes the CPU cycles, so this option is intended to
+  auto enable inline data if we have enough Tx queues, which means we have
+  enough CPU cores and PCI bandwidth is getting more critical and CPU
+  is not supposed to be bottleneck anymore.
 
-- ``txqs_max_vec`` parameter [int]
+  The copying data into WQE improves latency and can improve PPS performance
+  when PCI back pressure is detected and may be useful for scenarios involving
+  heavy traffic on many queues.
 
-  Enable vectorized Tx only when the number of TX queues is less than or
-  equal to this value. Effective only when ``tx_vec_en`` is enabled.
-
-  On ConnectX-5:
-
-        - Set to 8 by default on ARMv8.
-        - Set to 4 by default otherwise.
-
-  On BlueField
+  Because additional software logic is necessary to handle this mode, this
+  option should be used with care, as it may lower performance when back
+  pressure is not expected.
 
-        - Set to 16 by default.
+- ``txq_inline_max`` parameter [int]
+
+  Specifies the maximal packet length to be completely inlined into WQE
+  Ethernet Segment for ordinary SEND method. If packet is larger the specified
+  value, the packet data won't be copied by the driver at all, data buffer
+  is addressed with a pointer. If packet length is less or equal all packet
+  data will be copied into WQE. This may improve PCI bandwidth utilization for
+  short packets significantly but requires the extra CPU cycles.
+
+  The data inline feature is controlled by number of Tx queues, if number of Tx
+  queues is larger than ``txqs_min_inline`` key parameter, the inline feature
+  is engaged, if there are not enough Tx queues (which means not enough CPU cores
+  and CPU resources are scarce), data inline is not performed by the driver.
+  Assigning ``txqs_min_inline`` with zero always enables the data inline.
+
+  The default ``txq_inline_max`` value is 290. The specified value may be adjusted
+  by the driver in order not to exceed the limit (930 bytes) and to provide better
+  WQE space filling without gaps, the adjustment is reflected in the debug log.
+
+- ``txq_inline_mpw`` parameter [int]
+
+  Specifies the maximal packet length to be completely inlined into WQE for
+  Enhanced MPW method. If packet is large the specified value, the packet data
+  won't be copied, and data buffer is addressed with pointer. If packet length
+  is less or equal, all packet data will be copied into WQE. This may improve PCI
+  bandwidth utilization for short packets significantly but requires the extra
+  CPU cycles.
+
+  The data inline feature is controlled by number of TX queues, if number of Tx
+  queues is larger than ``txqs_min_inline`` key parameter, the inline feature
+  is engaged, if there are not enough Tx queues (which means not enough CPU cores
+  and CPU resources are scarce), data inline is not performed by the driver.
+  Assigning ``txqs_min_inline`` with zero always enables the data inline.
+
+  The default ``txq_inline_mpw`` value is 188. The specified value may be adjusted
+  by the driver in order not to exceed the limit (930 bytes) and to provide better
+  WQE space filling without gaps, the adjustment is reflected in the debug log.
+  Due to multiple packets may be included to the same WQE with Enhanced Multi
+  Packet Write Method and overall WQE size is limited it is not recommended to
+  specify large values for the ``txq_inline_mpw``.
 
 - ``txq_mpw_en`` parameter [int]
 
-  A nonzero value enables multi-packet send (MPS) for ConnectX-4 Lx and
-  enhanced multi-packet send (Enhanced MPS) for ConnectX-5, ConnectX-6 and BlueField.
-  MPS allows the TX burst function to pack up multiple packets in a
-  single descriptor session in order to save PCI bandwidth and improve
-  performance at the cost of a slightly higher CPU usage. When
-  ``txq_inline`` is set along with ``txq_mpw_en``, TX burst function tries
-  to copy entire packet data on to TX descriptor instead of including
-  pointer of packet only if there is enough room remained in the
-  descriptor. ``txq_inline`` sets per-descriptor space for either pointers
-  or inlined packets. In addition, Enhanced MPS supports hybrid mode -
-  mixing inlined packets and pointers in the same descriptor.
-
-  This option cannot be used with certain offloads such as ``DEV_TX_OFFLOAD_TCP_TSO,
-  DEV_TX_OFFLOAD_VXLAN_TNL_TSO, DEV_TX_OFFLOAD_GRE_TNL_TSO, DEV_TX_OFFLOAD_VLAN_INSERT``.
-  When those offloads are requested the MPS send function will not be used.
-
-  It is currently only supported on the ConnectX-4 Lx, ConnectX-5, ConnectX-6 and BlueField
-  families of adapters.
-  On ConnectX-4 Lx the MPW is considered un-secure hence disabled by default.
-  Users which enable the MPW should be aware that application which provides incorrect
-  mbuf descriptors in the Tx burst can lead to serious errors in the host including, on some cases,
-  NIC to get stuck.
-  On ConnectX-5, ConnectX-6 and BlueField the MPW is secure and enabled by default.
-
-- ``txq_mpw_hdr_dseg_en`` parameter [int]
-
-  A nonzero value enables including two pointers in the first block of TX
-  descriptor. This can be used to lessen CPU load for memory copy.
-
-  Effective only when Enhanced MPS is supported. Disabled by default.
-
-- ``txq_max_inline_len`` parameter [int]
-
-  Maximum size of packet to be inlined. This limits the size of packet to
-  be inlined. If the size of a packet is larger than configured value, the
-  packet isn't inlined even though there's enough space remained in the
-  descriptor. Instead, the packet is included with pointer.
-
-  Effective only when Enhanced MPS is supported. The default value is 256.
-
-- ``tx_vec_en`` parameter [int]
-
-  A nonzero value enables Tx vector on ConnectX-5, ConnectX-6 and BlueField NICs if the number of
-  global Tx queues on the port is less than ``txqs_max_vec``.
-
-  This option cannot be used with certain offloads such as ``DEV_TX_OFFLOAD_TCP_TSO,
-  DEV_TX_OFFLOAD_VXLAN_TNL_TSO, DEV_TX_OFFLOAD_GRE_TNL_TSO, DEV_TX_OFFLOAD_VLAN_INSERT``.
-  When those offloads are requested the MPS send function will not be used.
-
-  Enabled by default on ConnectX-5, ConnectX-6 and BlueField.
+  A nonzero value enables Enhanced Multi-Packet Write (eMPW) for ConnectX-5,
+  ConnectX-6 and BlueField. eMPW allows the TX burst function to pack up multiple
+  packets in a single descriptor session in order to save PCI bandwidth and improve
+  performance at the cost of a slightly higher CPU usage. When ``txq_inline_mpw``
+  is set along with ``txq_mpw_en``, TX burst function copies entire packet
+  data on to TX descriptor instead of including pointer of packet.
+
+  The Enhanced Multi-Packet Write feature is enabled by default if NIC supports
+  it, can be disabled by explicit specifying 0 value for ``txq_mpw_en`` option.
+  Also, if minimal data inlining is requested by non-zero ``txq_inline_min``
+  option or reported by the NIC, the eMPW feature is disengaged.
 
 - ``rx_vec_en`` parameter [int]
 
diff --git a/drivers/net/mlx5/mlx5.c b/drivers/net/mlx5/mlx5.c
index 42b36a7..e803f08 100644
--- a/drivers/net/mlx5/mlx5.c
+++ b/drivers/net/mlx5/mlx5.c
@@ -68,6 +68,15 @@
 /* Device parameter to set the minimum number of Rx queues to enable MPRQ. */
 #define MLX5_RXQS_MIN_MPRQ "rxqs_min_mprq"
 
+/* Device parameter to limit packet size to inline with ordinary SEND. */
+#define MLX5_TXQ_INLINE_MAX "txq_inline_max"
+
+/* Device parameter to configure minimal data size to inline. */
+#define MLX5_TXQ_INLINE_MIN "txq_inline_min"
+
+/* Device parameter to limit packet size to inline with Enhanced MPW. */
+#define MLX5_TXQ_INLINE_MPW "txq_inline_mpw"
+
 /*
  * Device parameter to configure the number of TX queues threshold for
  * enabling inline send.
@@ -884,6 +893,12 @@ struct mlx5_dev_spawn_data {
 		config->mprq.max_memcpy_len = tmp;
 	} else if (strcmp(MLX5_RXQS_MIN_MPRQ, key) == 0) {
 		config->mprq.min_rxqs_num = tmp;
+	} else if (strcmp(MLX5_TXQ_INLINE_MAX, key) == 0) {
+		config->txq_inline_max = tmp;
+	} else if (strcmp(MLX5_TXQ_INLINE_MIN, key) == 0) {
+		config->txq_inline_min = tmp;
+	} else if (strcmp(MLX5_TXQ_INLINE_MPW, key) == 0) {
+		config->txq_inline_mpw = tmp;
 	} else if (strcmp(MLX5_TXQS_MIN_INLINE, key) == 0) {
 		config->txqs_inline = tmp;
 	} else if (strcmp(MLX5_TXQ_MPW_EN, key) == 0) {
@@ -932,6 +947,9 @@ struct mlx5_dev_spawn_data {
 		MLX5_RX_MPRQ_LOG_STRIDE_NUM,
 		MLX5_RX_MPRQ_MAX_MEMCPY_LEN,
 		MLX5_RXQS_MIN_MPRQ,
+		MLX5_TXQ_INLINE_MIN,
+		MLX5_TXQ_INLINE_MAX,
+		MLX5_TXQ_INLINE_MPW,
 		MLX5_TXQS_MIN_INLINE,
 		MLX5_TXQ_MPW_EN,
 		MLX5_RX_VEC_EN,
@@ -1882,6 +1900,9 @@ struct mlx5_dev_spawn_data {
 		.hw_padding = 0,
 		.mps = MLX5_ARG_UNSET,
 		.rx_vec_en = 1,
+		.txq_inline_max = MLX5_ARG_UNSET,
+		.txq_inline_min = MLX5_ARG_UNSET,
+		.txq_inline_mpw = MLX5_ARG_UNSET,
 		.txqs_inline = MLX5_ARG_UNSET,
 		.vf_nl_en = 1,
 		.mr_ext_memseg_en = 1,
diff --git a/drivers/net/mlx5/mlx5.h b/drivers/net/mlx5/mlx5.h
index 47c9e47..ac7ea1d 100644
--- a/drivers/net/mlx5/mlx5.h
+++ b/drivers/net/mlx5/mlx5.h
@@ -188,6 +188,7 @@ struct mlx5_dev_config {
 	unsigned int cqe_comp:1; /* CQE compression is enabled. */
 	unsigned int cqe_pad:1; /* CQE padding is enabled. */
 	unsigned int tso:1; /* Whether TSO is supported. */
+	unsigned int tx_inline:1; /* Engage TX data inlining. */
 	unsigned int rx_vec_en:1; /* Rx vector is enabled. */
 	unsigned int mr_ext_memseg_en:1;
 	/* Whether memseg should be extended for MR creation. */
@@ -213,6 +214,9 @@ struct mlx5_dev_config {
 	unsigned int ind_table_max_size; /* Maximum indirection table size. */
 	unsigned int max_dump_files_num; /* Maximum dump files per queue. */
 	int txqs_inline; /* Queue number threshold for inlining. */
+	int txq_inline_min; /* Minimal amount of data bytes to inline. */
+	int txq_inline_max; /* Max packet size for inlining with SEND. */
+	int txq_inline_mpw; /* Max packet size for inlining with eMPW. */
 	struct mlx5_hca_attr hca_attr; /* HCA attributes. */
 };
 
-- 
1.8.3.1
Previous message: [dpdk-dev] [PATCH v4 0/8] net/mlx5: consolidate Tx datapath
Next message: [dpdk-dev] [PATCH 3/7] net/mlx5: update Tx datapath definitions
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
More information about the dev mailing list