[openib-general] Re: [PATCH] CMA: allow/require bind beforeconnect
Tom Tucker
tom at opengridcomputing.com
Mon Mar 27 11:42:33 PST 2006
On Mon, 2006-03-27 at 11:03 -0800, Sean Hefty wrote:
> >> It's correct that the CMA currently does not do this. I
> >> guess I'm still unsure of why it needs to when running over
> >> IB. The CMA should work fine if it assigned every active
> >> connection the same port number. As long as the CMA manages
> >> its own port spaces, does it matter if its semantics are defined this
> >> way?
> >>
> >The 4-tuple must be unique, even across devices. Hence if the
> >native stack has X:Y to I:J then CMA cannot also create
> >X:Y to I:J.
>
> What I what to understand is _why_ it must be unique? The CMA-IB port spaces
> are separate from the native stack.
But should they be? I mean it would be great if IPoIB address/port had
the same semantics as Linux native TCP/IP address/port tuples.
Especially if anyone ever dreamed of trying to bridge the IB/IP gap even
more with devices for this purpose ;-)
> And the CMA-IB has no issues connecting or
> routing data among multiple connections all having the same 4-tuple.
It 'works' to the extent that it has been tested thus far. I predict
that this will wreak havoc with applications -- it already breaks NFS
security policy.
> - Sean
> _______________________________________________
> openib-general mailing list
> openib-general at openib.org
> http://openib.org/mailman/listinfo/openib-general
>
> To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
More information about the general
mailing list