Re: [PATCH 1/2] ethdev: add capability control API

["Wiles, Keith" <keith.wiles@intel.com>]

> On Feb 11, 2017, at 12:38 AM, Jerin Jacob <jerin.jacob@caviumnetworks.com> wrote:
> 
> On Fri, Feb 10, 2017 at 02:05:49PM +0000, Cristian Dumitrescu wrote:
>> The rte_flow feature breaks the current monolithic approach for ethdev and
>> introduces the new generic flow API to ethdev using a plugin-like approach.
>> 
>> Basically, the rte_flow API is still logically part of ethdev:
>> - It extends the ethdev functionality: rte_flow is a new feature/capability
>>  of ethdev;
>> - all its functions work on an Ethernet device: the first parameter of the
>>  rte_flow functions is Ethernet device port ID.
>> 
>> At the same time, the rte_flow API is a sort of capability plugin for ethdev:
>> - the rte_flow API functions have their own name space: they are called
>>  rte_flow_operationXYZ() as opposed to rte_eth_dev_flow_operationXYZ());
>> - the rte_flow API functions are placed in separate files in the same
>>  librte_ether folder as opposed to rte_ethdev.[hc].
>> 
>> The way it works is by using the existing ethdev API function
>> rte_eth_dev_filter_ctrl() to query the current Ethernet device port ID for the
>> support of the rte_flow capability and return the pointer to the
>> rte_flow operations when supported and NULL otherwise:
>> 
>> struct rte_flow_ops *eth_flow_ops;
>> int rte = rte_eth_dev_filter_ctrl(eth_port_id,
>> 	RTE_ETH_FILTER_GENERIC, RTE_ETH_FILTER_GET, &eth_flow_ops);
>> 
>> Unfortunately, the rte_flow opportunistically uses the rte_eth_dev_filter_ctrl()
>> API function, which is applicable just to RX-side filters as opposed to
>> introducing a mechanism that could be used by any capability in a generic way.
>> 
>> This is the gap that addressed by the current patch. This mechanism is intended
>> to be used to introduce new capabilities into ethdev in a modular plugin-like
>> approach, such as hierarchical scheduler. Over time, if agreed, it can also be
>> used for exposing the existing Ethernet device capabilities in a modular way,
>> such as: xstats, filters, multicast, mirroring, tunnels, time stamping, eeprom,
>> bypass, etc.
>> 
>> Signed-off-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
>> ---
>> lib/librte_ether/rte_ethdev.c          | 13 +++++++++++++
>> lib/librte_ether/rte_ethdev.h          | 29 +++++++++++++++++++++++++++++
>> lib/librte_ether/rte_ether_version.map |  7 +++++++
>> 3 files changed, 49 insertions(+)
>> 
>> diff --git a/lib/librte_ether/rte_ethdev.c b/lib/librte_ether/rte_ethdev.c
>> index eb0a94a..ae187c4 100644
>> --- a/lib/librte_ether/rte_ethdev.c
>> +++ b/lib/librte_ether/rte_ethdev.c
>> @@ -2802,6 +2802,19 @@ rte_eth_dev_filter_ctrl(uint8_t port_id, enum rte_filter_type filter_type,
>> 	return (*dev->dev_ops->filter_ctrl)(dev, filter_type, filter_op, arg);
>> }
>> 
>> +int
>> +rte_eth_dev_capability_control(uint8_t port_id, enum rte_eth_capability cap,
>> +	void *arg)
>> +{
>> +	struct rte_eth_dev *dev;
>> +
>> +	RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, -ENODEV);
>> +
>> +	dev = &rte_eth_devices[port_id];
>> +	RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->cap_ctrl, -ENOTSUP);
>> +	return (*dev->dev_ops->cap_ctrl)(dev, cap, arg);
>> +}
>> +
>> void *
>> rte_eth_add_rx_callback(uint8_t port_id, uint16_t queue_id,
>> 		rte_rx_callback_fn fn, void *user_param)
>> diff --git a/lib/librte_ether/rte_ethdev.h b/lib/librte_ether/rte_ethdev.h
>> index c17bbda..43ffb9e 100644
>> --- a/lib/librte_ether/rte_ethdev.h
>> +++ b/lib/librte_ether/rte_ethdev.h
>> @@ -1073,6 +1073,12 @@ TAILQ_HEAD(rte_eth_dev_cb_list, rte_eth_dev_callback);
>>  * structure associated with an Ethernet device.
>>  */
>> 
>> +enum rte_eth_capability {
>> +	RTE_ETH_CAPABILITY_FLOW = 0, /**< Flow */
>> +	RTE_ETH_CAPABILITY_SCHED, /**< Hierarchical Scheduler */
>> +	RTE_ETH_CAPABILITY_MAX
>> +};
> 
> Shouldn't it be the FLAG?. Meaning, To represent ethdev port can have both.

The current API is requesting if the PMD supports the given feature and then returns the void * to the structure of function pointers or NULL similar to the rte_flow design. The developer would need to ask multiple times to understand if all of the features are supported by the PMD. I guess one of the options could be to return a list of features the PMD supports. The void * would point to the PMD capability list, which would need to be a const of some type to prevent someone from modifying the PMD capability list.

enum rte_eth_capability {
	RTE_ETH_CAPABILITY_LIST = 0,
	RTE_ETH_CAPABILITY_FLOW = 1,
	RTE_ETH_CAPABILITY_SCHED = 2,
	RTE_ETH_CAPABILITY_SCHED = 4,
	RTE_ETH_CAPABILITY_MAX		/* This one does not make sense in a bitmap set of enums */
};

The RTE_ETH_CAPABILITY_LIST could return the (void *) as a uint64_t listing the feature bits. The problem I think is the uint64_t is limiting us to 63 features (which maybe a big number, but maybe not) would be the only concern here. The PMD could return a pointer to a uint8_t array of feature values, were 0 is used as a no-op then we can have any number of features with the enum just being a number between 1-255 or uint16_t 1-65535.

Anyway just an option, we could have a different API for the feature list.

> 
>> +

Regards,
Keith