Roland,
what do you say about this, should we go with the query approach or
stick to using device flags? I will create the patches as necessary.

On 2/12/08, Eli Cohen <[EMAIL PROTECTED]> wrote:
> Roland Dreier wrote:
>  >  > I set these flags for mlx4 and mthca in patches 5/16 and 6/16 
> respectively.
>  >
>  > Ah I see.  Seems strange to use device->flags for just the checksum
>  > offload stuff and device_cap_flags for everything else though.  I
>  > don't see any major issues with moving device_cap_flags into struct
>  > ib_device and not forcing a device query call, but I guess we need to
>  > convince ourselves that the flags would never change at runtime, and
>  > anyway that needs to be a separate change from this IPoIB stuff.
>
>
> I think the cleanest way is to call query device caps and save a copy of these
>  flags in the private data. If we agree on this approach I can send patches 
> with
>  this change.
>
>
>  >
>  >  > I my machines I can see the flags set by inspecting 
> /sys/class/net/ib*/features
>  >
>  > I guess that's an OFED patch?  Is there any interest in submitting it
>  > upstream.
>
>
> All the patches I sent were checked also on the "for-2.6.25" branch and gone 
> through
>  basic testing. I also checked their variants on ofed. I think we should push 
> them upstream.
>
>
_______________________________________________
general mailing list
[email protected]
http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general

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

Reply via email to