[ofa-general] cmatose fails whereas rping passes on iWarp
Tung, Chien Tin
chien.tin.tung at intel.com
Wed Jul 8 14:06:50 PDT 2009
> > As far as I can tell this is an infiniband only attribute.
>I do not see
> > it in iWarp RDMA spec, if I am wrong please point me to the
>right place.
> >
> > How do we handle situation like this when there is a
>difference between
> > the two?
>
>Simplest thing to do in this case seems to be to handle it as an
>extension on the iWARP side; just keep track of what the
>consumer passed
>in for sq_sig_type and if it's "signal all" then never post unsignaled
>work requests. No reason to create gratuitous semantic
>differences when
>it's fairly easy to implement the union of IB and iWARP semantics.
For this issue, it is easy to resolve the difference and I don't
have a problem implementing the extension as suggested by Roland.
At the same time, I don't want to turn it into if an iWarp device
does not do what an Infiniband device would do then it is a bug.
I didn't take this issue that way as everyone said it with a :-)
I'm probably making it a bigger issue than it is so I will stop now.
:-)
Chien
More information about the general
mailing list