[ofa-general] IB/core: Enhance SMI for switch support

[Hal Rosenstock]

On Thu, 2007-05-17 at 16:03, Roland Dreier wrote:
>  > The risk is primarily on the switch side, rather than the CA/router
>  > side, right ? So isn't the downside of this minimal ?
> 
> Actually I was thinking that this disturbs the current SMI code and
> potentially introduces bugs (even with plain old CAs).

I may be wrong but the main SMI change is returning IB_SMI_FORWARD
rather than IB_SMI_SEND for two intermediate hop cases (C14-9:2 and
C14-13:2) which should not apply to CA or router ports.

> And without any switch driver it's not clear if we want to take that risk.

Understood.

> But OK, I'll queue this up for 2.6.23 and hope for the best...

Thanks.

-- Hal