[openib-general] [PATCHv3][RFC] kDAPL: use cm timers instead of own

[Sean Hefty]

Tom Duffy wrote:
>>>Hal's right - it does live in an IB-specific file, though that seems to
>>>live in a common directory. I don't think the conversion is sufficiently
>>>general to warrant any special status. Isn't it the case that it is
>>>Mellanox-hardware specific, not even IB-specific?
>>
>>The time format is IB architecture specific.  The CM is exposing a field 
>>value defined by the spec.  If a different format would be a better 
>>alternative, we can look at changing it.
> 
> I was looking at some of the ULPs.  SDP sets the timeout to 20, without
> explanation.  DAPL has a #define (set to 20 saying it is 4 sec in a
> comment).  simple example?  Set to 20, no reason.  To me, it ends up
> just being this opaque value.  Maybe some pre#defines in CM would make
> it clearer.  Unfortunately, advertising milliseconds or jiffies will
> just confuse consumers since the timeout cannot be that granular.

I don't know if the spec provides any help for a ULP to determine the local 
or remote CM timeouts.  And the CM itself requires work from the ULP to 
process requests, so it cannot set the value itself.  The value is encoded 
into a 5-bit field, and for what it's worth, packet lifetime values are 
provided by the SA in this same format...

- Sean