Hello!
I'd like to start working on MT25218 (aka Arbel memfree, aka native mode)
support for openib gen2.

There are many similiarities with MT25208, but there are also
major differences, and a lot of extra setup to be done
to compensate for the absence of on-board memory,
which unfortunately affects not only the set-up
commands but also performance-critical data path verbs
(e.g. posting WQEs, CQ polling).

What would be the best way to organise this code?
I guess could be two thinkable ways to do this:

1. Add the functionality to the mthca module,
   use a run-time flag to describe hardware differences.
   Thus, the same module will handle devices of all
   types.

2. Create a sibling or a child directory to mthca,
   and export common functions from mthca.
   In this case we will have two kernel modules,
   sharing some common files.

I am leaning towards option 2 for both the performance
reasons and for better separation between drivers.


Therefore, I propose adding a directory memfree under

https://openib.org/svn/gen2/trunk/src/linux-kernel/infiniband/hw/mthca/memfree

and all memfree specific code will go there.

Comments welcome,
MST
_______________________________________________
openib-general mailing list
[EMAIL PROTECTED]
http://openib.org/mailman/listinfo/openib-general

To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general

Reply via email to