[ofw] RE: [PATCH] Kill ATS registration in IPoIB

[Smith, Stan]

Tzachi Dar wrote:
> Hi stan attached is the patch that Fab has created.
> 
> Can I check this in?

Thanks for including the original patch.

Please do not commit the Kill ATS registration patch as in function
ipoib_reg_addrs() in file ipoib_driver.c, IPoIB registers the IPoIB port
IP address:GID pair with the subnet Administrator (ATS).  DAPL and DAPL2
IBAL providers depend on querying the SA for IPv6 address to GID
mapping.

If the ipoib patch could be reworked such that the ipoib interface IPv6
address GID pair is still registered with the SA under ATS_NAME, then
DAPL will continue to work correctly.

In the near future (measured in weeks), I can migrate DAPL to using the
new IBAT name service, then we can remove the IPv6:GID registration with
the SA from IPoIB.

Fab, Where can I find IBAT interface specifications?

Thanks,

Stan.


> 
> Thanks
> Tzachi
> 
>> -----Original Message-----
>> From: Smith, Stan [mailto:stan.smith at intel.com]
>> Sent: Friday, July 04, 2008 12:00 AM
>> To: Hefty, Sean; Tzachi Dar; Fab Tillier; ofw at lists.openfabrics.org
>> Subject: RE: [ofw] RE: [PATCH] Kill ATS registration in IPoIB
>> 
>> Hefty, Sean wrote:
>>>>> It seems to me that DAPL is still using the ATS registrations.
>>>>> 
>>>>> Do we all agree that ATS can be removed from the code?
>>>> 
>>>> Not until after WinOF 2.0 release.
>>>> Thank you.
>>> 
>>> Doesn't the socket CM eliminate the need for ATS?  Or is that just
>>> one of the DAPL connection options?
>> 
>> Yes - socket CM is one of two DAPL connections options, it
>> does not use ATS.
>> 
>> Oops, I confused history with the present w.r.t. DAPL ibal provider
>> and it's use of ATS. winDAPL (ibal provider) uses the SA to provide
>> an ATS service. Prior to removing ATS from the svn code base, please
>> identify 
>> code to be removed; I've lost the original patch.
>> 
>> Thanks,
>> 
>> Stan.