[openib-general][RFC]: CMA IB implementation

Guy German guyg at voltaire.com
Tue Sep 27 08:39:03 PDT 2005


Hi Sean,

Basically I think that we can definitely agree that if the cma can implement 
ib_at intended functionality it should replace it - no need to have 2 modules 
doing the same thing.

So the points that need to be considered are:

1. Caching
sean> generic SA caching module should be a part of ib_sa or exist separately.

What about specific path records caching with event driven invalidate logic ?

2. Partitioning
The cma should be able to use ib p_keys or tcp vlans according to the ipoib 
interface ip address (subnet).

3. Qos
As Yaron H mentioned: ib_at model suggest taking by default the SL value from 
the ipoib interface of that subnet which took it from the SA MCRecord

4. ATS registration
sean> I think that ATS registration/deregistration should be integrated with
sean> IPoIB.

I don't think there is a consensus around that, but I don't know all details.

5. retries
retries could be centralized in the ib_at approach.

6. ULP override
ULP's that are aware of the transport layer can override default values derived 
from ipoib.

I think it is fair to say that if the cma can handle these issues, than ib_at is 
no longer needed.

Guy




More information about the general mailing list