[dpdk-dev] [PATCH v2] fix lacp check system address

[Ferruh Yigit]

On 4/10/2020 11:15 AM, Ferruh Yigit wrote:
> On 11/26/2019 3:09 PM, podovinnikov wrote:
>> Hi Ferruh
>>
>>> This is bonding patch, please cc bonding maintainers (Chas, instead of John W.
>>> Linville).
>> Sorry about that. I have 2 patches ( one for af packet, one for bonding).
>>
>> I am a little confused.
>>
>>
>>> Can you please give more details, what is failing what is the impact etc.
>>
>> In bond (lacp) we have several nics ( ports )
>>
>> When we have negotiation with peer about what port we prefer,
>>
>> we send information about what system we preferred in partner system 
>> name field.
>>
>> Peer also sends us what partner system name it prefer.
>>
>> When we receive a message from it we must compare its preferred
>>
>> system name with our system name, but not with our port mac address
>>
>> In my test I have several problems with that
>>
>> 1. If master port (mac address same as system address)
>>
>> shuts down (I have two ports) I loose connection
>>
>> 2. If secondary port (mac address not same as system address)
>>
>> receives message before master port, my connection is not established.
> 
> Hi Vadim,
> 
> Thanks for the info and sorry for late response, it is sitting in backlog for a
> long time now.
> 
> @Chas, do you have any objection on the fix, if not I am planning to merge it soon.
> 
> Thanks,
> ferruh

cc'ed Xavier, he also has a bonding patch in the backlog waiting for review [1].


Vadim, Xavier,

Since you both send bonding patches, I assume you both know about bonding at
some level, at least more than me, so would you mind reviewing eachother's
patch? So both can go it.

Thanks,
ferruh


[1]
https://patchwork.dpdk.org/user/todo/dpdk/?series=8679

> 
>>
>>
>> 26.11.2019 16:27, Ferruh Yigit пишет:
>>> Hi Vadim,
>>>
>>> On 11/26/2019 11:55 AM, Vadim wrote:
>>>> Signed-off-by: Vadim <podovinnikov at protei.ru>
>>> Can you please provide full "Name Surname <email at adress>" signature?
>>>
>>>
>>> This is bonding patch, please cc bonding maintainers (Chas, instead of John W.
>>> Linville).
>>>
>>>> fix lacp check system address
>>> Can you please give more details, what is failing what is the impact etc.
>>>
>>>> ---
>>>>   drivers/net/bonding/rte_eth_bond_8023ad.c | 17 ++++++++++++++++-
>>>>   1 file changed, 16 insertions(+), 1 deletion(-)
>>>>
>>>> diff --git a/drivers/net/bonding/rte_eth_bond_8023ad.c b/drivers/net/bonding/rte_eth_bond_8023ad.c
>>>> index b77a37ddb..d4dda790a 100644
>>>> --- a/drivers/net/bonding/rte_eth_bond_8023ad.c
>>>> +++ b/drivers/net/bonding/rte_eth_bond_8023ad.c
>>>> @@ -792,18 +792,33 @@ rx_machine_update(struct bond_dev_private *internals, uint16_t slave_id,
>>>>   		struct rte_mbuf *lacp_pkt) {
>>>>   	struct lacpdu_header *lacp;
>>>>   	struct lacpdu_actor_partner_params *partner;
>>>> +	struct port *port, *agg;
>>>>   
>>>>   	if (lacp_pkt != NULL) {
>>>>   		lacp = rte_pktmbuf_mtod(lacp_pkt, struct lacpdu_header *);
>>>>   		RTE_ASSERT(lacp->lacpdu.subtype == SLOW_SUBTYPE_LACP);
>>>>   
>>>>   		partner = &lacp->lacpdu.partner;
>>>> +		port = &bond_mode_8023ad_ports[slave_id];
>>>> +		agg = &bond_mode_8023ad_ports[port->aggregator_port_id];
>>>> +
>>>>   		if (rte_is_same_ether_addr(&partner->port_params.system,
>>>> -			&internals->mode4.mac_addr)) {
>>>> +							&agg->actor.system)) {
>>>>   			/* This LACP frame is sending to the bonding port
>>>>   			 * so pass it to rx_machine.
>>>>   			 */
>>>>   			rx_machine(internals, slave_id, &lacp->lacpdu);
>>>> +		} else {
>>>> +			char preferred_system_name[RTE_ETHER_ADDR_FMT_SIZE];
>>>> +			char self_system_name[RTE_ETHER_ADDR_FMT_SIZE];
>>>> +			rte_ether_format_addr(preferred_system_name,
>>>> +						RTE_ETHER_ADDR_FMT_SIZE,
>>>> +						&partner->port_params.system);
>>>> +			rte_ether_format_addr(self_system_name,
>>>> +				RTE_ETHER_ADDR_FMT_SIZE, &agg->actor.system);
>>>> +			MODE4_DEBUG("preferred partner system %s not equal "
>>>> +						"self system: %s\n",
>>>> +				preferred_system_name, self_system_name);
>>>>   		}
>>>>   		rte_pktmbuf_free(lacp_pkt);
>>>>   	} else
>>>>
>