[ofw] ibbus @ svn.1516 has been restored with omitted commits from svn.1499

[Smith, Stan]

Anatoly Greenblatt wrote:
> Stan,
>
> I was wondering why did you make ibbus a filter? Is it for a case
> where its services are not required, do you see any installation that
> works w/o ibbus?

Server 2008 & Vista (WLH) did not install due to IB co-installer framework (specifically 2 .inf files: ib_bus.inf & an HCA .inf file). Reuven spent many long hours attempting to get the co-installer ibbus installed on WLH; eventually the task was completed although not with 100% satisfaction.
Discussions were held last Spring (March/April timeframe) about how to resolve the WLH IB co-installer install problems. Fab T. had advocated using a upper class filter. After much discussion I started the work.
In addition to WLH driver installs, there has been a long term desire to get the bus driver into the HCA device stack to promote a cleaner ibbus and HCA linkage; IRP communication between components.
For the further out view, ibbus PNP management is handled by IBAL, mostly bypassing the entire Microsoft PNP framework or as least masking the MS PNP.
KMDF greatly simplifies driver design and is the future target for ibbus.


> I'd rather make it some sort of a framework library that all hca
> drivers link to. There are many examples of device classes that use
> such frameworks - ndis, streaming class, av or ks, storage ports,
> This way there would be less devices in the stack, less installation
> complexity and no money spent on extra whql certification for filter
> driver.

Conceptually, the InfiniBandController class coupled with ibbus provide a class framework.
I'm all for making the IB stack a more simple design and leveraging more MS features when possible.

>
> Regards,
> Anatoly.
>
> ________________________________
>
> From: ofw-bounces at lists.openfabrics.org on behalf of Smith, Stan
> Sent: Tue 8/26/2008 10:07 PM
> To: ofw at lists.openfabrics.org
> Subject: [ofw] ibbus @ svn.1516 has been restored with omitted
> commits from svn.1499
>
>
>
>
> Svn.1516 contains omitted commits from 1599; sorry for the confusion.
> _______________________________________________
> ofw mailing list
> ofw at lists.openfabrics.org
> http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ofw