Re: [dpdk-dev] [PATCH v8 1/7] ethdev: add set ptype function

[Pavan Nikhilesh Bhagavatula <pbhagavatula@marvell.com>]

>-----Original Message-----
>From: dev <dev-bounces@dpdk.org> On Behalf Of Andrew Rybchenko
>Sent: Thursday, October 3, 2019 1:11 PM
>To: Pavan Nikhilesh Bhagavatula <pbhagavatula@marvell.com>; Jerin
>Jacob Kollanukkaran <jerinj@marvell.com>; John McNamara
><john.mcnamara@intel.com>; Marko Kovacevic
><marko.kovacevic@intel.com>; Thomas Monjalon
><thomas@monjalon.net>; Ferruh Yigit <ferruh.yigit@intel.com>
>Cc: dev@dpdk.org
>Subject: Re: [dpdk-dev] [PATCH v8 1/7] ethdev: add set ptype function
>
>On 10/3/19 12:36 AM, 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, NULL, 0)`
>> and PMD may skip packet type processing and set
>rte_mbuf::packet_type to
>> RTE_PTYPE_UNKNOWN.
>>
>> -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, NULL,
>0)`.
>>
>> Suggested-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
>> Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
>
>With few fixes below:
>Reviewed-by: Andrew Rybchenko <arybchenko@solarflare.com>
>
>[snip]
>
>> diff --git a/lib/librte_ethdev/rte_ethdev.c
>b/lib/librte_ethdev/rte_ethdev.c
>> index 17d183e1f..57bc12b56 100644
>> --- a/lib/librte_ethdev/rte_ethdev.c
>> +++ b/lib/librte_ethdev/rte_ethdev.c
>> @@ -2602,6 +2602,65 @@
>rte_eth_dev_get_supported_ptypes(uint16_t port_id, uint32_t
>ptype_mask,
>>   	return j;
>>   }
>>
>> +int
>> +rte_eth_dev_set_supported_ptypes(uint16_t port_id, uint32_t
>ptype_mask,
>> +				 uint32_t *set_ptypes, unsigned int
>num)
>> +{
>> +	unsigned int i, j;
>> +	struct rte_eth_dev *dev;
>> +	const uint32_t *all_ptypes;
>> +
>> +	RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, -ENODEV);
>> +	dev = &rte_eth_devices[port_id];
>> +
>> +	if (num > 0 && set_ptypes == NULL)
>> +		return -EINVAL;
>> +
>> +	if (*dev->dev_ops->dev_supported_ptypes_get == NULL ||
>> +			*dev->dev_ops->dev_supported_ptypes_set
>== NULL) {
>> +		if (num > 0)
>> +			set_ptypes[0] = RTE_PTYPE_UNKNOWN;
>> +		return 0;
>> +	}
>> +
>> +	if (ptype_mask == 0) {
>> +		if (num > 0)
>> +			set_ptypes[0] = RTE_PTYPE_UNKNOWN;
>> +
>> +		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) {
>> +		if (num > 0)
>> +			set_ptypes[0] = RTE_PTYPE_UNKNOWN;
>> +
>> +		return 0;
>> +	}
>> +
>> +	/*
>> +	 * Accodommodate as many set_ptypes as possible. If the
>supplied
>> +	 * set_ptypes array is insufficient fill it partially.
>> +	 */
>> +	for (i = 0, j = 0; set_ptypes != NULL &&
>> +				(all_ptypes[i] !=
>RTE_PTYPE_UNKNOWN); ++i) {
>> +		if (ptype_mask & all_ptypes[i]) {
>> +			if (j < num - 1) {
>> +				set_ptypes[j] = all_ptypes[i];
>> +				j++;
>> +				continue;
>> +			}
>> +			break;
>> +		}
>> +	}
>> +
>> +	if (set_ptypes != NULL)
>> +		set_ptypes[j] = RTE_PTYPE_UNKNOWN;
>
>Initially I thought that we are safe here, but now realized that we
>can write more than num here, e.g. if set_ptypes is not NULL, but num
>is 0.
>I think the right condition here is (j < num) since it guarantees that
>set_ptype is not NULL as well (since num is greater than 0 if unsigned j
>is less than num).

Ack will send v9.

>
>> +
>> +	return (*dev->dev_ops->dev_supported_ptypes_set)(dev,
>ptype_mask);
>> +}
>> +
>>   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 d9871782e..c577a9172 100644
>> --- a/lib/librte_ethdev/rte_ethdev.h
>> +++ b/lib/librte_ethdev/rte_ethdev.h
>> @@ -2431,6 +2431,42 @@ 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);
>> +/**
>> + * @warning
>> + * @b EXPERIMENTAL: this API may change without prior notice.
>> + *
>> + * 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.
>> + *
>> + * The function accepts an array `set_ptypes` allocated by the caller
>to
>> + * store the packet types set by the driver, the last element of the
>array
>> + * is set to RTE_PTYPE_UNKNOWN. The size of the `set_ptype` array
>should be
>> + * `rte_eth_dev_get_supported_ptypes() + 1` else it might only be
>filled
>> + * partially.
>> + *
>> + * @param port_id
>> + *   The port identifier of the Ethernet device.
>> + * @param ptype_mask
>> + *   The ptype family that application is interested in.
>
>I think it is better to highlight here that it should be bitwise OR of
>RTE_PTYPE_*_MASK or 0.
>
>BTW, I think it would be useful to check it to avoid misuse of the API
>when, for example, RTE_PTYPE_L4_TCP is specified as ptype_mask.

I feel the same we could do something like 

        test_ptype_mask = ptype_mask;
        while (test_ptype_mask) {
                uint8_t mask = test_ptype_mask & RTE_PTYPE_L2_MASK;

                if (mask && (mask != RTE_PTYPE_L2_MASK)) {
                        if (num > 0)
                                set_ptypes[0] = RTE_PTYPE_UNKNOWN;

                        return -EINVAL;
                }

                test_ptype_mask >>= __builtin_popcount(RTE_PTYPE_L2_MASK);
        }

>
>> + * @param set_ptypes
>> + *   An array pointer to store set packet types, allocated by caller. The
>> + *   function marks the end of array with RTE_PTYPE_UNKNOWN.
>> + * @param num
>> + *   Size of the array pointed by param ptypes.
>> + *   Should be rte_eth_dev_get_supported_ptypes() + 1 to
>accommodate the
>> + *   set ptypes.
>> + * @return
>> + *   - (0) if Success.
>> + *   - (-ENODEV) if *port_id* invalid.
>> + *   - (-EINVAL) if *ptype_mask* is invalid (or) set_ptypes is NULL and
>> + *     num > 0.
>> + */
>> +__rte_experimental
>> +int rte_eth_dev_set_supported_ptypes(uint16_t port_id, uint32_t
>ptype_mask,
>> +				     uint32_t *set_ptypes, unsigned int
>num);
>>
>>   /**
>>    * 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..06afd6a3d 100644
>> --- a/lib/librte_ethdev/rte_ethdev_core.h
>> +++ b/lib/librte_ethdev/rte_ethdev_core.h
>> @@ -110,6 +110,16 @@ 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 int (*eth_dev_supported_ptypes_set_t)(struct
>rte_eth_dev *dev,
>> +					      uint32_t ptype_mask);
>> +/**< @internal Inform an Ethernet device about packet types in
>which the
>> + *   recipient is interested.
>> + *   Ptype_mask can have any of the following values
>RTE_PTYPE_UNKNOWN |
>> + *   RTE_PTYPE_L2_MASK | RTE_PTYPE_L3_MASK |
>RTE_PTYPE_L4_MASK |
>> + *   RTE_PTYPE_TUNNEL_MASK | RTE_PTYPE_INNER_L2_MASK |
>RTE_PTYPE_INNER_L3_MASK |
>> + *   RTE_PTYPE_INNER_L4_MASK | RTE_PTYPE_ALL_MASK.
>> + */
>> +
>
>I'm sorry for misunderstanding. It should be a comment before the
>typedef
>which start from /** . So long post comments looks a bit strange.
>Please, describe parameters using @param and return values using
>@return
>and @retval.
>Also there is no point to mention all RTE_PTYPE masks here to avoid
>duplication and possible unsync if we have more masks.

My bad. Will cleanup in v9.

>
>Will you update a driver to implement it in 19.11 release cycle?
>As I understand it is a requirement to add a new API.
>

Yes, I will add support for net/octeontx2.

>>   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. */
>
>[snip]