[openib-general] [PATCH 3/4] SA path record caching

Sean Hefty mshefty at ichips.intel.com
Thu Jan 26 10:08:49 PST 2006


Hal Rosenstock wrote:
> This looks like a wildcarded DGID request but NumbPaths is 1. Am I
> missing something ?

This is correct.  The cache is only retrieving a single path to each destination 
from a given port.  NumbPaths is required for a GetTable request.

Supporting multiple paths between an SGID and a DGID would require additional 
work, since the key used to lookup a path record expands beyond just the DGID.

I've already thought of a couple ways to handle this.  The easiest is to 
maintain a list of path records with matching DGIDs in the index.  The index 
would return the list, which would then be walked to return multiple paths to 
the user.

> Also, the ability for one node's SA cache to handle this for another
> node appears to be a future possibility.

This could be made possible, but would require separate SA GetTable queries on 
the part of the node with the cache.  An additional query is required for each SGID.

- Sean



More information about the general mailing list