Re: [PATCH v4 10/27] IB/Verbs: Reform cm related part in IB-core cma/ucm

[Michael Wang <yun.wang@profitbricks.com>]

On 04/16/2015 07:21 PM, Tom Talpey wrote:
> On 4/16/2015 11:22 AM, Michael Wang wrote:
>>
>>
>> On 04/16/2015 04:31 PM, Hefty, Sean wrote:
>>>>> This is equivalent to today where the checks are per node rather than
>>>>> per port.
>>>>>
>>>>> Should all checks here be port 1 based or only certain ones like listen
>>>>> ? For example, in connect/reject/disconnect, don't we already have port
>>>>> ? Guess this can be dealt with later as this is not a regression from
>>>>> the current implementation.
>>>>
>>>> Yeah, these parts of cma may need more carve in future, like some new
>>>> callback
>>>> for different CM type as Sean suggested.
>>>>
>>>> Maybe directly using 1 could help to highlight the problem ;-)
>>>
>>> Only a few checks need to be per device.  I think I pointed those out previously.  Testing should show anywhere that we miss fairly quickly, since port would still be 0.  For the checks that can be updated to be per port, I would rather go ahead and convert them.
>>
>> Got it, will be changed in next version :-)
>>
>> To be confirmed:
>>                 PORT ASSIGNED
>>     rdma_init_qp_attr    Y
>>     rdma_destroy_id        unknown
>>     cma_listen_on_dev    N
>>     cma_bind_loopback    N
>>     rdma_listen        N
> 
> Why "N"? rdma_listen() can be constrained to a single port, right?
> And even if wildcarded, it needs to act on multiple ports, which is
> to say, it will fail only if no ports are eligible.

Yeah, it can or can't, maybe 'unknown' is better :-)

Regards,
Michael Wang

> 
> Tom.
> 
> 
>>     rdma_connect        Y
>>     rdma_accept        Y
>>     rdma_reject        Y
>>     rdma_disconnect        Y
>>     ib_ucm_add_one        N
>>
>> Is this list correct?
>>
>> Regards,
>> Michael Wang
>>
>>>
>>> - Sean
>>>
>>
>>
>