[ofa-general] [PATCH] IB/core: handle race between elements in qork queues after event

Moni Shoua monis at Voltaire.COM
Mon May 12 01:12:15 PDT 2008


Moni Shoua wrote:
>> I guess I can believe things don't get worse but I still don't know how
>> this makes things better.  With the current code the request is lost
>> because it goes to the wrong SM; with the new code the request is failed
>> by the SA layer.  So in both cases the consumer just has to try again.
>>
>> So is there some practical benefit we see by adding this code?
>>
>>  - R.
>> _______________________________________________
>> general mailing list
>> general at lists.openfabrics.org
>> http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general
>>
>> To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
>>
> 
> In general I see the benefit in faster detection of wrong SM ah. Before the patch consumers 
> need to wait for  a timeout before the detection and after the patch it happens immediately 
> on return from the function. This improves the performance of an SM failover scenario.
> 
> Some applications  may get the benefit above  only they handle new return code (EAGAIN) specifically 
> but this patch opens the door for such improvement. 
> 
> thanks
> 
>  MoniS
> _______________________________________________
> general mailing list
> general at lists.openfabrics.org
> http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general
> 
> To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
> 

Hi Roland,
Can we please go on with this patch? We would like to see it in the next kernel.

thanks
 MoniS



More information about the general mailing list