[ofa-general] cmatose fails whereas rping passes on iWarp

[Steve Wise]

Tung, Chien Tin wrote:
>>> I see.  Then it sounds like cxgb3 is busted because its not checking
>>> this attribute.  :)
>>>       
>> Yep... grep shows none of amso1100, cxgb3, nor nes look at sq_sig_type
>> at all in the kernel driver.  Looks like a common iWARP bug :)
>>     
>
>
> As far as I can tell this is an infiniband only attribute.  I do not see
> it in iWarp RDMA spec, if I am wrong please point me to the right place.
>
> How do we handle situation like this when there is a difference between
> the two?
>
>   

IMO its an API issue and thus really outside the scope of the iWARP RDMA 
Verbs document, which doesn't define any APIs.    Since it can easily be 
supported (I think), I plan to add support in cxgb3.  IE just add 
additional logic in your post paths to check the attribute and set the 
appropriate "signaled" bit for your WRs.  Unless I'm missing something here?

Steve.