[dpdk-dev] [PATCH v6 8/9] ethdev: representor iterator compare complete info

Xueming(Steven) Li xuemingl at nvidia.com
Tue Feb 16 17:35:39 CET 2021


>-----Original Message-----
>From: Andrew Rybchenko <andrew.rybchenko at oktetlabs.ru>
>Sent: Monday, February 15, 2021 5:32 PM
>To: Xueming(Steven) Li <xuemingl at nvidia.com>
>Cc: dev at dpdk.org; Slava Ovsiienko <viacheslavo at nvidia.com>; Asaf Penso <asafp at nvidia.com>; NBU-Contact-Thomas Monjalon
><thomas at monjalon.net>; Ferruh Yigit <ferruh.yigit at intel.com>; Ray Kinsella <mdr at ashroe.eu>; Neil Horman
><nhorman at tuxdriver.com>
>Subject: Re: [PATCH v6 8/9] ethdev: representor iterator compare complete info
>
>On 2/14/21 6:21 AM, Xueming Li wrote:
>> The NIC can have multiple PCIe links and can be attached to multiple
>> hosts, for example the same single NIC can be shared for multiple
>> server units in the rack. On each PCIe link NIC can provide multiple
>> PFs and VFs/SFs based on these ones. The full representor identifier
>> consists of three indices - controller index, PF index, and VF or SF index (if any).
>>
>> SR-IOV and SubFunction are created on top of PF. PF index is
>> introduced because there might be multiple PFs in the bonding
>> configuration and only bonding device is probed.
>>
>> In eth representor comparator callback, ethdev representor ID was
>> compared with devarg. Since controller index and PF index not
>> compared, callback returned representor from other PF or controller.
>>
>> This patch adds new API to convert representor controller, pf and
>> vf/sf index to representor ID. Representor comparer callback convert
>> representor info into ID and compare with device representor ID.
>>
>> Signed-off-by: Xueming Li <xuemingl at nvidia.com>
>> ---
>>  lib/librte_ethdev/ethdev_driver.h | 32 ++++++++++++
>> lib/librte_ethdev/rte_class_eth.c | 38 ++++++++++----
>>  lib/librte_ethdev/rte_ethdev.c    | 83 +++++++++++++++++++++++++++++++
>>  lib/librte_ethdev/version.map     |  1 +
>>  4 files changed, 145 insertions(+), 9 deletions(-)
>>
>> diff --git a/lib/librte_ethdev/ethdev_driver.h
>> b/lib/librte_ethdev/ethdev_driver.h
>> index abcbc3112d..23342f1be2 100644
>> --- a/lib/librte_ethdev/ethdev_driver.h
>> +++ b/lib/librte_ethdev/ethdev_driver.h
>> @@ -1243,6 +1243,38 @@ struct rte_eth_devargs {
>>  	enum rte_eth_representor_type type; /* type of representor */  };
>>
>> +/**
>> + * PMD helper function to convert representor ID from location detail
>> + *
>> + * Convert representor ID from controller, pf and (sf or vf).
>> + * The mapping is retrieved from rte_eth_representor_info_get().
>> + *
>> + * If PMD doesn't return representor range info, simply ignore
>> +controller
>> + * and pf to keep backward compatibility.
>
>It does not sound right. If controller and/or pf is specified, it must not be ignored.

Okay, return error if info not found.

>
>> + *
>> + * @param ethdev
>> + *  Handle of ethdev port.
>> + * @param id
>
>May I suggest to name it 'repr_id' to make it less ambguous.

Accept.

>
>> + *  Pointer to converted representor ID.
>
>I'd prefer do not mix in and out paramters. I suggest to make it the last parameter.
>
>> + * @param type
>> + *  Representor type.
>> + * @param controller
>> + *  Controller ID, -1 if unspecified.
>> + * @param pf
>> + *  PF port ID, -1 if unspecified.
>> + * @param representor_port
>> + *  Representor port ID, -1 if unspecified.
>
>Not sure that I understand what is it? Is it vf or sf number?

Yes, will make it clear.

>
>> + *
>> + * @return
>> + *  Negative errno value on error, 0 on success.
>> + */
>> +__rte_internal
>> +int
>> +rte_eth_representor_id_convert(const struct rte_eth_dev *ethdev,
>> +			       uint16_t *id,
>> +			       enum rte_eth_representor_type type,
>> +			       int controller, int pf, int representor_port);
>> +
>>  /**
>>   * PMD helper function to parse ethdev arguments
>>   *
>> diff --git a/lib/librte_ethdev/rte_class_eth.c
>> b/lib/librte_ethdev/rte_class_eth.c
>> index 051c892b40..f7b7e659e7 100644
>> --- a/lib/librte_ethdev/rte_class_eth.c
>> +++ b/lib/librte_ethdev/rte_class_eth.c
>> @@ -65,9 +65,10 @@ eth_representor_cmp(const char *key __rte_unused,
>> {
>>  	int ret;
>>  	char *values;
>> -	const struct rte_eth_dev_data *data = opaque;
>> -	struct rte_eth_devargs representors;
>> -	uint16_t index;
>> +	const struct rte_eth_dev *edev = opaque;
>> +	const struct rte_eth_dev_data *data = edev->data;
>> +	struct rte_eth_devargs eth_da;
>> +	uint16_t id, nc, np, nf, i, c, p, f;
>>
>>  	if ((data->dev_flags & RTE_ETH_DEV_REPRESENTOR) == 0)
>>  		return -1; /* not a representor port */ @@ -76,17 +77,36 @@
>> eth_representor_cmp(const char *key __rte_unused,
>>  	values = strdup(value);
>>  	if (values == NULL)
>>  		return -1;
>> -	memset(&representors, 0, sizeof(representors));
>> -	ret = rte_eth_devargs_parse_representor_ports(values, &representors);
>> +	memset(&eth_da, 0, sizeof(eth_da));
>> +	ret = rte_eth_devargs_parse_representor_ports(values, &eth_da);
>>  	free(values);
>>  	if (ret != 0)
>>  		return -1; /* invalid devargs value */
>>
>> +	if (eth_da.nb_mh_controllers == 0 && eth_da.nb_ports == 0 &&
>> +	    eth_da.nb_representor_ports == 0)
>> +		return -1;
>> +	nc = eth_da.nb_mh_controllers > 0 ? eth_da.nb_mh_controllers : 1;
>> +	np = eth_da.nb_ports > 0 ? eth_da.nb_ports : 1;
>> +	nf = eth_da.nb_representor_ports > 0 ? eth_da.nb_representor_ports :
>> +1;
>> +
>>  	/* Return 0 if representor id is matching one of the values. */
>> -	for (index = 0; index < representors.nb_representor_ports; index++)
>> -		if (data->representor_id ==
>> -				representors.representor_ports[index])
>> +	for (i = 0; i < nc * np * nf; ++i) {
>> +		c = i / (np * nf);
>> +		p = (i / nf) % np;
>> +		f = i % nf;
>> +		if (rte_eth_representor_id_convert(edev,
>> +			&id,
>> +			eth_da.type,
>> +			eth_da.nb_mh_controllers == 0 ? -1 :
>> +					eth_da.mh_controllers[c],
>> +			eth_da.nb_ports == 0 ? -1 : eth_da.ports[p],
>> +			eth_da.nb_representor_ports == 0 ? -1 :
>> +					eth_da.representor_ports[f]) < 0)
>> +			continue;
>> +		if (data->representor_id == id)
>>  			return 0;
>> +	}
>>  	return -1; /* no match */
>>  }
>>
>> @@ -112,7 +132,7 @@ eth_dev_match(const struct rte_eth_dev *edev,
>>
>>  	ret = rte_kvargs_process(kvlist,
>>  			eth_params_keys[RTE_ETH_PARAM_REPRESENTOR],
>> -			eth_representor_cmp, edev->data);
>> +			eth_representor_cmp, (void *)(uintptr_t)edev);
>>  	if (ret != 0)
>>  		return -1;
>>  	/* search for representor key */
>> diff --git a/lib/librte_ethdev/rte_ethdev.c
>> b/lib/librte_ethdev/rte_ethdev.c index 07c6debb58..da0cf1a920 100644
>> --- a/lib/librte_ethdev/rte_ethdev.c
>> +++ b/lib/librte_ethdev/rte_ethdev.c
>> @@ -5617,6 +5617,89 @@ rte_eth_devargs_parse(const char *dargs, struct rte_eth_devargs *eth_da)
>>  	return result;
>>  }
>>
>> +int
>> +rte_eth_representor_id_convert(const struct rte_eth_dev *ethdev,
>> +			       uint16_t *id,
>> +			       enum rte_eth_representor_type type,
>> +			       int controller, int pf, int representor_port) {
>> +	int ret, n, i, count;
>> +	struct rte_eth_representor_info *info = NULL;
>> +	size_t size;
>> +
>> +	if (type == RTE_ETH_REPRESENTOR_NONE)
>> +		return 0;
>> +	if (id == NULL)
>> +		return -EINVAL;
>> +
>> +	/* Get PMD representor range info. */
>> +	ret = rte_eth_representor_info_get(ethdev->data->port_id, NULL);
>> +	if (ret < 0) {
>> +		/* Fallback to direct mapping for compatibility. */
>> +		*id = representor_port;
>
>I think it is a bad behaviour as I stated above. It is only if and only if type is VF, controller and PF are unspecified and representor_port
>is specified.

Agree, will make it only valid for legacy VF representor.

>
>> +	}
>> +	n = ret;
>> +	size = sizeof(*info) + n * sizeof(info->ranges[0]);
>> +	info = calloc(1, size);
>> +	if (info == NULL)
>> +		return -ENOMEM;
>> +	ret = rte_eth_representor_info_get(ethdev->data->port_id, info);
>> +	if (ret < 0)
>> +		goto out;
>> +
>> +	/* Default controller and pf to caller. */
>> +	if (controller == -1)
>> +		controller = info->controller;
>> +	if (pf == -1)
>> +		pf = info->pf;
>> +
>> +	/* Locate representor ID. */
>> +	for (i = 0; i < n; ++i) {
>> +		if (info->ranges[i].type != type)
>> +			continue;
>> +		/* PMD hit: ignore controller if -1. */
>> +		if (info->ranges[i].controller != -1 &&
>> +		    info->ranges[i].controller != (uint16_t)controller)
>> +			continue;
>
>I think it is incorrect to ignore controller in range if controller is specified in request. It must match.

This is a real requirement I'm facing from orchestration, before orchestration having knowledge on port
status, it always send "pf#vf#" to OVS->DPDK, although "pf#" is meaningless in standard mode. It will take
time for orchestration and OVS to evolve... PMD uses this option to decide ignore them or not.

>
>> +		count = info->ranges[i].id_end - info->ranges[i].id_base + 1;
>> +		if (info->ranges[i].type == RTE_ETH_REPRESENTOR_PF) {
>> +			/* PF. */
>> +			if (pf >= info->ranges[i].pf + count)
>> +				continue;
>> +			*id = info->ranges[i].id_base +
>> +			      (pf - info->ranges[i].pf);
>> +			goto out;
>> +		}
>> +		/* VF or SF. */
>> +		/* PMD hit: ignore pf if -1. */
>> +		if (info->ranges[i].pf != -1 &&
>> +		    info->ranges[i].pf != (uint16_t)pf)
>> +			continue;
>
>Same for PF.
>
>> +		if (info->ranges[i].type == RTE_ETH_REPRESENTOR_VF) {
>
>Typically switch/case looks a bit a bitter for such code.

'bitter'? :) will update.

>
>> +			/* VF. */
>
>The comment is useless

Accept this one and bellows.

>
>> +			if (representor_port >= info->ranges[i].vf + count)
>> +				continue;
>> +			*id = info->ranges[i].id_base +
>> +			      (representor_port - info->ranges[i].vf);
>> +			goto out;
>> +		} else if (info->ranges[i].type == RTE_ETH_REPRESENTOR_SF) {
>> +			/* SF. */
>
>The comment is useless
>
>> +			if (representor_port >= info->ranges[i].sf + count)
>> +				continue;
>> +			*id = info->ranges[i].id_base +
>> +			      (representor_port - info->ranges[i].sf);
>> +			goto out;
>> +		}
>> +	}
>> +	/* Not matching representor ID range. */
>> +	ret = -ENOENT;
>> +
>> +out:
>> +	if (info != NULL)
>> +		free(info);
>
>There is no necessity to check against NULL above, free() does it in any case.
>
>> +	return ret;
>> +}
>> +
>>  static int
>>  eth_dev_handle_port_list(const char *cmd __rte_unused,
>>  		const char *params __rte_unused,
>> diff --git a/lib/librte_ethdev/version.map
>> b/lib/librte_ethdev/version.map index bb6f7436c2..2891f5734e 100644
>> --- a/lib/librte_ethdev/version.map
>> +++ b/lib/librte_ethdev/version.map
>> @@ -268,6 +268,7 @@ INTERNAL {
>>  	rte_eth_hairpin_queue_peer_bind;
>>  	rte_eth_hairpin_queue_peer_unbind;
>>  	rte_eth_hairpin_queue_peer_update;
>> +	rte_eth_representor_id_convert;
>>  	rte_eth_switch_domain_alloc;
>>  	rte_eth_switch_domain_free;
>>  };
>>



More information about the dev mailing list