Re: [dpdk-dev] [PATCH v5 1/3] ethdev: support API to set max LRO packet size

["Ananyev, Konstantin" <konstantin.ananyev@intel.com>]

> This patch implements [1], to support API for configuration and
> validation of max size for LRO aggregated packet.
> API change notice [2] is removed, and release notes for 19.11
> are updated accordingly.
> 
> Various PMDs using LRO offload are updated, the new data members are
> initialized to ensure they don't fail validation.
> 
> [1] http://patches.dpdk.org/patch/58217/
> [2] http://patches.dpdk.org/patch/57492/

Actually if the requirement is just to allow user to limit max lro size,
then why not to add just new function for that:

int rte_eth_dev_set_max_lro(uint16_t port_id, uint32_t lro);
?

And make it optional for the drivers to support it.
That way PMD/devices that allow LRO max size to be configurable,
can support it others can fail.
 
Konstantin

> 
> Signed-off-by: Dekel Peled <dekelp@mellanox.com>
> Reviewed-by: Andrew Rybchenko <arybchenko@solarflare.com>
> Acked-by: Thomas Monjalon <thomas@monjalon.net>
> Acked-by: Matan Azrad <matan@mellanox.com>
> ---
>  doc/guides/nics/features.rst           |  2 ++
>  doc/guides/rel_notes/deprecation.rst   |  4 ----
>  doc/guides/rel_notes/release_19_11.rst |  8 +++++++
>  drivers/net/bnxt/bnxt_ethdev.c         |  1 +
>  drivers/net/hinic/hinic_pmd_ethdev.c   |  1 +
>  drivers/net/ixgbe/ixgbe_ethdev.c       |  1 +
>  drivers/net/mlx5/mlx5.h                |  3 +++
>  drivers/net/mlx5/mlx5_ethdev.c         |  1 +
>  drivers/net/mlx5/mlx5_rxq.c            |  1 -
>  drivers/net/qede/qede_ethdev.c         |  1 +
>  drivers/net/virtio/virtio_ethdev.c     |  1 +
>  drivers/net/vmxnet3/vmxnet3_ethdev.c   |  1 +
>  lib/librte_ethdev/rte_ethdev.c         | 44 ++++++++++++++++++++++++++++++++++
>  lib/librte_ethdev/rte_ethdev.h         |  4 ++++
>  14 files changed, 68 insertions(+), 5 deletions(-)
> 
> diff --git a/doc/guides/nics/features.rst b/doc/guides/nics/features.rst
> index 7a31cf7..2138ce3 100644
> --- a/doc/guides/nics/features.rst
> +++ b/doc/guides/nics/features.rst
> @@ -193,10 +193,12 @@ LRO
>  Supports Large Receive Offload.
> 
>  * **[uses]       rte_eth_rxconf,rte_eth_rxmode**: ``offloads:DEV_RX_OFFLOAD_TCP_LRO``.
> +  ``dev_conf.rxmode.max_lro_pkt_size``.
>  * **[implements] datapath**: ``LRO functionality``.
>  * **[implements] rte_eth_dev_data**: ``lro``.
>  * **[provides]   mbuf**: ``mbuf.ol_flags:PKT_RX_LRO``, ``mbuf.tso_segsz``.
>  * **[provides]   rte_eth_dev_info**: ``rx_offload_capa,rx_queue_offload_capa:DEV_RX_OFFLOAD_TCP_LRO``.
> +* **[provides]   rte_eth_dev_info**: ``max_lro_pkt_size``.
> 
> 
>  .. _nic_features_tso:
> diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
> index c10dc30..fdec33d 100644
> --- a/doc/guides/rel_notes/deprecation.rst
> +++ b/doc/guides/rel_notes/deprecation.rst
> @@ -87,10 +87,6 @@ Deprecation Notices
>    This scheme will allow PMDs to avoid lookup to internal ptype table on Rx and
>    thereby improve Rx performance if application wishes do so.
> 
> -* ethdev: New 32-bit fields may be added for maximum LRO session size, in
> -  struct ``rte_eth_dev_info`` for the port capability and in struct
> -  ``rte_eth_rxmode`` for the port configuration.
> -
>  * cryptodev: support for using IV with all sizes is added, J0 still can
>    be used but only when IV length in following structs ``rte_crypto_auth_xform``,
>    ``rte_crypto_aead_xform`` is set to zero. When IV length is greater or equal
> diff --git a/doc/guides/rel_notes/release_19_11.rst b/doc/guides/rel_notes/release_19_11.rst
> index 87b7bd0..a3fc023 100644
> --- a/doc/guides/rel_notes/release_19_11.rst
> +++ b/doc/guides/rel_notes/release_19_11.rst
> @@ -418,6 +418,14 @@ ABI Changes
>    align the Ethernet header on receive and all known encapsulations
>    preserve the alignment of the header.
> 
> +* ethdev: Added 32-bit fields for maximum LRO aggregated packet size, in
> +  struct ``rte_eth_dev_info`` for the port capability and in struct
> +  ``rte_eth_rxmode`` for the port configuration.
> +  Application should use the new field in struct ``rte_eth_rxmode`` to configure
> +  the requested size.

That part I am not happy with: * application should use*.
Many apps I suppose are ok with default LRO size selected by PMD/HW.
Why to force changes in all of them?

> +  PMD should use the new field in struct ``rte_eth_dev_info`` to report the
> +  supported port capability.
> +
> 
>  Shared Library Versions
>  -----------------------
> diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
> index b9b055e..741b897 100644
> --- a/drivers/net/bnxt/bnxt_ethdev.c
> +++ b/drivers/net/bnxt/bnxt_ethdev.c
> @@ -519,6 +519,7 @@ static int bnxt_dev_info_get_op(struct rte_eth_dev *eth_dev,
>  	/* Fast path specifics */
>  	dev_info->min_rx_bufsize = 1;
>  	dev_info->max_rx_pktlen = BNXT_MAX_PKT_LEN;
> +	dev_info->max_lro_pkt_size = BNXT_MAX_PKT_LEN;
> 
>  	dev_info->rx_offload_capa = BNXT_DEV_RX_OFFLOAD_SUPPORT;
>  	if (bp->flags & BNXT_FLAG_PTP_SUPPORTED)
> diff --git a/drivers/net/hinic/hinic_pmd_ethdev.c b/drivers/net/hinic/hinic_pmd_ethdev.c
> index 9f37a40..b33b2cf 100644
> --- a/drivers/net/hinic/hinic_pmd_ethdev.c
> +++ b/drivers/net/hinic/hinic_pmd_ethdev.c
> @@ -727,6 +727,7 @@ static void hinic_get_speed_capa(struct rte_eth_dev *dev, uint32_t *speed_capa)
>  	info->max_tx_queues  = nic_dev->nic_cap.max_sqs;
>  	info->min_rx_bufsize = HINIC_MIN_RX_BUF_SIZE;
>  	info->max_rx_pktlen  = HINIC_MAX_JUMBO_FRAME_SIZE;
> +	info->max_lro_pkt_size = HINIC_MAX_JUMBO_FRAME_SIZE;
>  	info->max_mac_addrs  = HINIC_MAX_UC_MAC_ADDRS;
>  	info->min_mtu = HINIC_MIN_MTU_SIZE;
>  	info->max_mtu = HINIC_MAX_MTU_SIZE;
> diff --git a/drivers/net/ixgbe/ixgbe_ethdev.c b/drivers/net/ixgbe/ixgbe_ethdev.c
> index 30c0379..5719552 100644
> --- a/drivers/net/ixgbe/ixgbe_ethdev.c
> +++ b/drivers/net/ixgbe/ixgbe_ethdev.c
> @@ -3814,6 +3814,7 @@ static int ixgbevf_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
>  	}
>  	dev_info->min_rx_bufsize = 1024; /* cf BSIZEPACKET in SRRCTL register */
>  	dev_info->max_rx_pktlen = 15872; /* includes CRC, cf MAXFRS register */
> +	dev_info->max_lro_pkt_size = RTE_IPV4_MAX_PKT_LEN;
>  	dev_info->max_mac_addrs = hw->mac.num_rar_entries;
>  	dev_info->max_hash_mac_addrs = IXGBE_VMDQ_NUM_UC_MAC;
>  	dev_info->max_vfs = pci_dev->max_vfs;
> diff --git a/drivers/net/mlx5/mlx5.h b/drivers/net/mlx5/mlx5.h
> index fab58c9..4783b5c 100644
> --- a/drivers/net/mlx5/mlx5.h
> +++ b/drivers/net/mlx5/mlx5.h
> @@ -206,6 +206,9 @@ struct mlx5_hca_attr {
>  #define MLX5_LRO_SUPPORTED(dev) \
>  	(((struct mlx5_priv *)((dev)->data->dev_private))->config.lro.supported)
> 
> +/* Maximal size of aggregated LRO packet. */
> +#define MLX5_MAX_LRO_SIZE (UINT8_MAX * 256u)
> +
>  /* LRO configurations structure. */
>  struct mlx5_lro_config {
>  	uint32_t supported:1; /* Whether LRO is supported. */
> diff --git a/drivers/net/mlx5/mlx5_ethdev.c b/drivers/net/mlx5/mlx5_ethdev.c
> index 2b7c867..3adc824 100644
> --- a/drivers/net/mlx5/mlx5_ethdev.c
> +++ b/drivers/net/mlx5/mlx5_ethdev.c
> @@ -606,6 +606,7 @@ struct ethtool_link_settings {
>  	/* FIXME: we should ask the device for these values. */
>  	info->min_rx_bufsize = 32;
>  	info->max_rx_pktlen = 65536;
> +	info->max_lro_pkt_size = MLX5_MAX_LRO_SIZE;
>  	/*
>  	 * Since we need one CQ per QP, the limit is the minimum number
>  	 * between the two values.
> diff --git a/drivers/net/mlx5/mlx5_rxq.c b/drivers/net/mlx5/mlx5_rxq.c
> index 24d0eaa..9423e7b 100644
> --- a/drivers/net/mlx5/mlx5_rxq.c
> +++ b/drivers/net/mlx5/mlx5_rxq.c
> @@ -1701,7 +1701,6 @@ struct mlx5_rxq_obj *
>  	return 0;
>  }
> 
> -#define MLX5_MAX_LRO_SIZE (UINT8_MAX * 256u)
>  #define MLX5_MAX_TCP_HDR_OFFSET ((unsigned int)(sizeof(struct rte_ether_hdr) + \
>  					sizeof(struct rte_vlan_hdr) * 2 + \
>  					sizeof(struct rte_ipv6_hdr)))
> diff --git a/drivers/net/qede/qede_ethdev.c b/drivers/net/qede/qede_ethdev.c
> index 575982f..ccbb8a4 100644
> --- a/drivers/net/qede/qede_ethdev.c
> +++ b/drivers/net/qede/qede_ethdev.c
> @@ -1277,6 +1277,7 @@ static int qede_dev_configure(struct rte_eth_dev *eth_dev)
> 
>  	dev_info->min_rx_bufsize = (uint32_t)QEDE_MIN_RX_BUFF_SIZE;
>  	dev_info->max_rx_pktlen = (uint32_t)ETH_TX_MAX_NON_LSO_PKT_LEN;
> +	dev_info->max_lro_pkt_size = (uint32_t)0x7FFF;
>  	dev_info->rx_desc_lim = qede_rx_desc_lim;
>  	dev_info->tx_desc_lim = qede_tx_desc_lim;
> 
> diff --git a/drivers/net/virtio/virtio_ethdev.c b/drivers/net/virtio/virtio_ethdev.c
> index 044eb10..22ce5a2 100644
> --- a/drivers/net/virtio/virtio_ethdev.c
> +++ b/drivers/net/virtio/virtio_ethdev.c
> @@ -2435,6 +2435,7 @@ static void virtio_dev_free_mbufs(struct rte_eth_dev *dev)
>  		RTE_MIN(hw->max_queue_pairs, VIRTIO_MAX_TX_QUEUES);
>  	dev_info->min_rx_bufsize = VIRTIO_MIN_RX_BUFSIZE;
>  	dev_info->max_rx_pktlen = VIRTIO_MAX_RX_PKTLEN;
> +	dev_info->max_lro_pkt_size = VIRTIO_MAX_RX_PKTLEN;
>  	dev_info->max_mac_addrs = VIRTIO_MAX_MAC_ADDRS;
> 
>  	host_features = VTPCI_OPS(hw)->get_features(hw);
> diff --git a/drivers/net/vmxnet3/vmxnet3_ethdev.c b/drivers/net/vmxnet3/vmxnet3_ethdev.c
> index d1faeaa..d18e8bc 100644
> --- a/drivers/net/vmxnet3/vmxnet3_ethdev.c
> +++ b/drivers/net/vmxnet3/vmxnet3_ethdev.c
> @@ -1161,6 +1161,7 @@ static int eth_vmxnet3_pci_remove(struct rte_pci_device *pci_dev)
>  	dev_info->max_tx_queues = VMXNET3_MAX_TX_QUEUES;
>  	dev_info->min_rx_bufsize = 1518 + RTE_PKTMBUF_HEADROOM;
>  	dev_info->max_rx_pktlen = 16384; /* includes CRC, cf MAXFRS register */
> +	dev_info->max_lro_pkt_size = 16384;
>  	dev_info->speed_capa = ETH_LINK_SPEED_10G;
>  	dev_info->max_mac_addrs = VMXNET3_MAX_MAC_ADDRS;
> 
> diff --git a/lib/librte_ethdev/rte_ethdev.c b/lib/librte_ethdev/rte_ethdev.c
> index 652c369..c642ba5 100644
> --- a/lib/librte_ethdev/rte_ethdev.c
> +++ b/lib/librte_ethdev/rte_ethdev.c
> @@ -1136,6 +1136,26 @@ struct rte_eth_dev *
>  	return name;
>  }
> 
> +static inline int
> +check_lro_pkt_size(uint16_t port_id, uint32_t config_size,
> +		   uint32_t dev_info_size)
> +{
> +	int ret = 0;
> +
> +	if (config_size > dev_info_size) {
> +		RTE_ETHDEV_LOG(ERR, "Ethdev port_id=%d max_lro_pkt_size %u "
> +			       "> max allowed value %u\n", port_id, config_size,
> +			       dev_info_size);
> +		ret = -EINVAL;
> +	} else if (config_size < RTE_ETHER_MIN_LEN) {
> +		RTE_ETHDEV_LOG(ERR, "Ethdev port_id=%d max_lro_pkt_size %u "
> +			       "< min allowed value %u\n", port_id, config_size,
> +			       (unsigned int)RTE_ETHER_MIN_LEN);
> +		ret = -EINVAL;
> +	}
> +	return ret;
> +}
> +
>  int
>  rte_eth_dev_configure(uint16_t port_id, uint16_t nb_rx_q, uint16_t nb_tx_q,
>  		      const struct rte_eth_conf *dev_conf)
> @@ -1266,6 +1286,18 @@ struct rte_eth_dev *
>  							RTE_ETHER_MAX_LEN;
>  	}
> 
> +	/*
> +	 * If LRO is enabled, check that the maximum aggregated packet
> +	 * size is supported by the configured device.
> +	 */
> +	if (dev_conf->rxmode.offloads & DEV_RX_OFFLOAD_TCP_LRO) {
> +		ret = check_lro_pkt_size(
> +				port_id, dev_conf->rxmode.max_lro_pkt_size,
> +				dev_info.max_lro_pkt_size);
> +		if (ret != 0)
> +			goto rollback;
> +	}
> +
>  	/* Any requested offloading must be within its device capabilities */
>  	if ((dev_conf->rxmode.offloads & dev_info.rx_offload_capa) !=
>  	     dev_conf->rxmode.offloads) {
> @@ -1770,6 +1802,18 @@ struct rte_eth_dev *
>  		return -EINVAL;
>  	}
> 
> +	/*
> +	 * If LRO is enabled, check that the maximum aggregated packet
> +	 * size is supported by the configured device.
> +	 */
> +	if (local_conf.offloads & DEV_RX_OFFLOAD_TCP_LRO) {
> +		int ret = check_lro_pkt_size(port_id,
> +				dev->data->dev_conf.rxmode.max_lro_pkt_size,
> +				dev_info.max_lro_pkt_size);
> +		if (ret != 0)
> +			return ret;
> +	}
> +
>  	ret = (*dev->dev_ops->rx_queue_setup)(dev, rx_queue_id, nb_rx_desc,
>  					      socket_id, &local_conf, mp);
>  	if (!ret) {
> diff --git a/lib/librte_ethdev/rte_ethdev.h b/lib/librte_ethdev/rte_ethdev.h
> index 44d77b3..1b76df5 100644
> --- a/lib/librte_ethdev/rte_ethdev.h
> +++ b/lib/librte_ethdev/rte_ethdev.h
> @@ -395,6 +395,8 @@ struct rte_eth_rxmode {
>  	/** The multi-queue packet distribution mode to be used, e.g. RSS. */
>  	enum rte_eth_rx_mq_mode mq_mode;
>  	uint32_t max_rx_pkt_len;  /**< Only used if JUMBO_FRAME enabled. */
> +	/** Maximum allowed size of LRO aggregated packet. */
> +	uint32_t max_lro_pkt_size;
>  	uint16_t split_hdr_size;  /**< hdr buf size (header_split enabled).*/
>  	/**
>  	 * Per-port Rx offloads to be set using DEV_RX_OFFLOAD_* flags.
> @@ -1218,6 +1220,8 @@ struct rte_eth_dev_info {
>  	const uint32_t *dev_flags; /**< Device flags */
>  	uint32_t min_rx_bufsize; /**< Minimum size of RX buffer. */
>  	uint32_t max_rx_pktlen; /**< Maximum configurable length of RX pkt. */
> +	/** Maximum configurable size of LRO aggregated packet. */
> +	uint32_t max_lro_pkt_size;
>  	uint16_t max_rx_queues; /**< Maximum number of RX queues. */
>  	uint16_t max_tx_queues; /**< Maximum number of TX queues. */
>  	uint32_t max_mac_addrs; /**< Maximum number of MAC addresses. */
> --
> 1.8.3.1