[ofa-general] librdmacm 1.0.4 release

Or Gerlitz or.gerlitz at gmail.com
Wed Oct 31 13:20:35 PDT 2007


On 10/31/07, Sean Hefty <mshefty at ichips.intel.com> wrote:

The text is slightly different in places depending on the context.


Indeed, but I found it somehow confusing to have these repeatitions, for
example some conn_ param values are described in more detail at the man page
of rdma_get_cm_event and some other values in more detail at the page of
rdma_connect/rdma_accept.


> > More re conn params, and also following questions I got from people
> > coding to librdmacm/libibverbs - for CONN the RNR and ACK timeouts are
> > being set by the core kernel (rdmacm, cm) code.


I don't understand the source of the confusion yet.  The values that are
> used are based on what's passed in by the user.  All QP attributes are
> set by the kernel code when it's modified by the library.


please note that I referred here to the RNR and ACK --timeout-- values and
not to the --retry-- values. The timeout values are not left for the user to
be chosen but are set by: RNR - the core code at the IB CM sets it to zero
which I think means 655ms by the encoding table at the IB spec.  ACK - its
the packet-life-time value you get from the SA in the path query plus the
hca-ack-delay which you estimate to be the packet-life-time , etc, does this
makes it clear? people just asking what values are used for their QPs and
then I have to explain all this...

Yes - rdma_accept man page has been updated (not pushed yet) to indicate
> that this value is ignored.


cool

The rdma_connect and rdma_accept man pages have been updated to state
> that rnr_retry_count value applies to the remote peer.


cool, thanks

Or.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openfabrics.org/pipermail/general/attachments/20071031/1cf89e65/attachment.html>


More information about the general mailing list