Re: [dpdk-dev] [PATCH v2 1/7] ethdev: add set ptype function

[Andrew Rybchenko <arybchenko@solarflare.com>]

On 8/21/19 11:47 PM, pbhagavatula@marvell.com wrote:
> From: Pavan Nikhilesh <pbhagavatula@marvell.com>
>
> Add `rte_eth_dev_set_supported_ptypes` function that will allow the
> application to inform the PMD the packet types it is interested in.
> Based on the ptypes set PMDs can optimize their Rx path.
>
> -If application doesn’t want any ptype information it can call
> `rte_eth_dev_set_supported_ptypes(ethdev_id, RTE_PTYPE_UNKNOWN)` and PMD
> will set rte_mbuf::packet_type to 0.

I would say that "... and PMD may skip packet type processing and just set
rte_mbuf::packet_type to RTE_PTYPE_UNKNOWN."

I.e. PMD is not obligated to skip packet type processing, it just has
permission to do it.

> -If application doesn’t call `rte_eth_dev_set_supported_ptypes` PMD can
> return `rte_mbuf::packet_type` with `rte_eth_dev_get_supported_ptypes`.
>
> -If application is interested only in L2/L3 layer, it can inform the PMD
> to update `rte_mbuf::packet_type` with L2/L3 ptype by calling
> `rte_eth_dev_set_supported_ptypes(ethdev_id,
> 		RTE_PTYPE_L2_MASK | RTE_PTYPE_L3_MASK)`.
>
> Suggested-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
> Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
> ---
>   doc/guides/nics/features.rst             | 12 ++++++---
>   doc/guides/rel_notes/release_19_11.rst   |  7 ++++++
>   lib/librte_ethdev/rte_ethdev.c           | 32 ++++++++++++++++++++++++
>   lib/librte_ethdev/rte_ethdev.h           | 16 ++++++++++++
>   lib/librte_ethdev/rte_ethdev_core.h      |  6 +++++
>   lib/librte_ethdev/rte_ethdev_version.map |  3 +++
>   6 files changed, 72 insertions(+), 4 deletions(-)
>
> diff --git a/doc/guides/nics/features.rst b/doc/guides/nics/features.rst
> index c4e128d2f..d4d55f721 100644
> --- a/doc/guides/nics/features.rst
> +++ b/doc/guides/nics/features.rst
> @@ -582,10 +582,14 @@ Supports inner packet L4 checksum.
>   Packet type parsing
>   -------------------
>   
> -Supports packet type parsing and returns a list of supported types.
> -
> -* **[implements] eth_dev_ops**: ``dev_supported_ptypes_get``.
> -* **[related]    API**: ``rte_eth_dev_get_supported_ptypes()``.
> +Supports packet type parsing and returns a list of supported types. Allows
> +application to set ptypes it is interested in.

It is better to start new sentence from new line. It will keep the first 
line
untouched.

> +
> +* **[implements] eth_dev_ops**: ``dev_supported_ptypes_get``,
> +  ``dev_supported_ptypes_set``.
> +* **[related]    API**: ``rte_eth_dev_get_supported_ptypes()``,
> +  ``rte_eth_dev_set_supported_ptypes()``.
> +* **[provides]   mbuf**: ``mbuf.packet_type``.
>   
>   
>   .. _nic_features_timesync:
> diff --git a/doc/guides/rel_notes/release_19_11.rst b/doc/guides/rel_notes/release_19_11.rst
> index 8490d897c..a7cec1fe8 100644
> --- a/doc/guides/rel_notes/release_19_11.rst
> +++ b/doc/guides/rel_notes/release_19_11.rst
> @@ -56,6 +56,13 @@ New Features
>        Also, make sure to start the actual text at the margin.
>        =========================================================
>   
> +* **Added API in ethdev to set supported packet types**

May I suggest:
Added ethdev API to set supported packet types.

> +
> +  *  Added new API ``rte_eth_dev_set_supported_ptypes`` that allows an
> +     application to request PMD to set specific ptypes defined
> +     through ``rte_eth_dev_set_supported_ptypes`` in ``rte_mbuf::packet_type``.

Consider:
Added new API ``rte_eth_dev_set_supported_ptypes`` that allows an
application to inform PMD about packet types classification the application
is interested in.

> +  *  This scheme will allow PMDs to avoid lookup to internal ptype table on Rx
> +     and thereby improve Rx performance if application wishes do so.
>   
>   Removed Items
>   -------------
> diff --git a/lib/librte_ethdev/rte_ethdev.c b/lib/librte_ethdev/rte_ethdev.c
> index 17d183e1f..f529cbe9f 100644
> --- a/lib/librte_ethdev/rte_ethdev.c
> +++ b/lib/librte_ethdev/rte_ethdev.c
> @@ -2602,6 +2602,38 @@ rte_eth_dev_get_supported_ptypes(uint16_t port_id, uint32_t ptype_mask,
>   	return j;
>   }
>   
> +uint32_t
> +rte_eth_dev_set_supported_ptypes(uint16_t port_id, uint32_t ptype_mask)
> +{
> +	int i;
> +	struct rte_eth_dev *dev;
> +	const uint32_t *all_ptypes;
> +	uint32_t all_ptype_mask = 0;
> +	uint32_t supp_ptype_mask = 0;
> +
> +	RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, -ENODEV);

We should either return 0 or change prototype to return successfully
set ptypes in optional (may be NULL) output parameter.
I would change the prototype.

> +	dev = &rte_eth_devices[port_id];
> +	RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->dev_supported_ptypes_get, 0);

Since API is not mandatory and provides just hits for optimizations, I think
it should return error in really unexpected cases only. It looks like 
invalid
port_id is the only case.

> +	if (ptype_mask == 0) {
> +		RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->dev_supported_ptypes_set,
> +					0);
> +		return (*dev->dev_ops->dev_supported_ptypes_set)(dev,
> +				ptype_mask);
> +	}
> +
> +	all_ptypes = (*dev->dev_ops->dev_supported_ptypes_get)(dev);
> +	if (all_ptypes == NULL)
> +		return 0;
> +
> +	for (i = 0; all_ptypes[i] != RTE_PTYPE_UNKNOWN; ++i)
> +		all_ptype_mask |= all_ptypes[i];
> +
> +	supp_ptype_mask = all_ptype_mask & ptype_mask;

I think it is not that useful to do here. Moreover, it is invalid in a 
way how
it is implemented right now. E.g. if get returns:
RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4 | RTE_PTYPE_L4_TCP
RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4 | RTE_PTYPE_L4_UDP
RTE_PTYPE_L2_ARP
and mask is RTE_PTYPE_L2_MASK | RTE_PTYPE_L3_MASK | RTE_PTYPE_L4_MASK
the result will be:
RTE_PTYPE_L2_ARP | RTE_PTYPE_L3_IPV4 | RTE_PTYPE_L4_FRAG


> +	return (*dev->dev_ops->dev_supported_ptypes_set)(dev, supp_ptype_mask);

It is not checked that  dev_supported_ptypes_set is not NULL.

> +}
> +
>   void
>   rte_eth_macaddr_get(uint16_t port_id, struct rte_ether_addr *mac_addr)
>   {
> diff --git a/lib/librte_ethdev/rte_ethdev.h b/lib/librte_ethdev/rte_ethdev.h
> index dc6596bc9..1ab0af4d8 100644
> --- a/lib/librte_ethdev/rte_ethdev.h
> +++ b/lib/librte_ethdev/rte_ethdev.h
> @@ -2431,6 +2431,22 @@ int rte_eth_dev_fw_version_get(uint16_t port_id,
>    */
>   int rte_eth_dev_get_supported_ptypes(uint16_t port_id, uint32_t ptype_mask,
>   				     uint32_t *ptypes, int num);
> +/**

Missing:
  * @warning
  * @b EXPERIMENTAL: this API may change without prior notice.

> + * Request Ethernet device to set only specific packet types in the packet.

Request sounds mandatory. Consider:
Inform Ethernet device of the packet types classification in which
the recipient is interested.

> + *
> + * Application can use this function to set only specific ptypes that it's
> + * interested. This information can be used by the PMD to optimize Rx path.
> + *
> + * @param port_id
> + *   The port identifier of the Ethernet device.
> + * @param ptype_mask
> + *   The ptype family that application is interested in.

It should be highligted that for each layer it should be either
full mask (e.g. RTE_PTYPE_L4_MASK) or only one specific type
(e.g. RTE_PTYPE_L4_TCP) since otherwise caller may try
RTE_PTYPE_L4_TCP|RTE_PTYPE_L4_UDP which is RTE_PTYPE_L4_FRAG
in fact.

> + * @return
> + *   - (>=0) Ptype mask that has been set successfully.

uint32_t is always >=0, so I think (>=0) should be removed.
However, in fact I think it should be:
Negative errno value on error, 0 on success.

I think that "Ptype mask that has been set successfully" could be
misleading since for each layer it is not a bit mark, but enum.
The output should be supported_ptype_get-slyle. I.e. an array
which is terminated by RTE_PTYPE_UNKNOWN entry.

> + */
> +__rte_experimental
> +uint32_t rte_eth_dev_set_supported_ptypes(uint16_t port_id,
> +					  uint32_t ptype_mask);
>   
>   /**
>    * Retrieve the MTU of an Ethernet device.
> diff --git a/lib/librte_ethdev/rte_ethdev_core.h b/lib/librte_ethdev/rte_ethdev_core.h
> index 2922d5b7c..e65ae983d 100644
> --- a/lib/librte_ethdev/rte_ethdev_core.h
> +++ b/lib/librte_ethdev/rte_ethdev_core.h
> @@ -110,6 +110,10 @@ typedef void (*eth_dev_infos_get_t)(struct rte_eth_dev *dev,
>   typedef const uint32_t *(*eth_dev_supported_ptypes_get_t)(struct rte_eth_dev *dev);
>   /**< @internal Get supported ptypes of an Ethernet device. */
>   
> +typedef uint32_t (*eth_dev_supported_ptypes_set_t)(struct rte_eth_dev *dev,
> +					      uint32_t ptype_mask);
> +/**< @internal Set required ptypes of an Ethernet device. */

Consider:
Inform device about packet types in which the recipient is interested.

> +
>   typedef int (*eth_queue_start_t)(struct rte_eth_dev *dev,
>   				    uint16_t queue_id);
>   /**< @internal Start rx and tx of a queue of an Ethernet device. */
> @@ -421,6 +425,8 @@ struct eth_dev_ops {
>   	eth_fw_version_get_t       fw_version_get; /**< Get firmware version. */
>   	eth_dev_supported_ptypes_get_t dev_supported_ptypes_get;
>   	/**< Get packet types supported and identified by device. */
> +	eth_dev_supported_ptypes_set_t dev_supported_ptypes_set;
> +	/**< Inform device about the interested ptypes. */

Consider
Inform device about packet types in which the recipient is interested.

>   
>   	vlan_filter_set_t          vlan_filter_set; /**< Filter VLAN Setup. */
>   	vlan_tpid_set_t            vlan_tpid_set; /**< Outer/Inner VLAN TPID Setup. */

<snip>