[openfabrics-ewg] [openib-general] RHEL5 and OFED ...

[Or Gerlitz]

On 11/2/06, Doug Ledford <dledford at redhat.com> wrote:
> On Thu, 2006-10-19 at 17:02 +0200, Or Gerlitz wrote:
> > Doug Ledford wrote:
> > > ... and reviewing arpingib
> > > (which I'm going to remove from the ipoibtools and fix the native arping
> > > in RHEL5 to work properly over IB without needing a new flag, the -A or
> > > -U flags should be sufficient assuming those modes worked at all over IB
> > > which they don't in either the native arping or the patched arpingib in
> > > ipoibtools).  I should get to it today though.
> >
> > Would you mind send the patch to arping for review?
>
> OK, this patch to arping actually makes it work for me in all modes
> (duplicate address detection, arp response, and unsolicited arp
> response).  You shouldn't need any new flags to arping with this patch,
> you should be able to just use the existing modes of operation as they
> were intended to make the ipoibha.pl script work.  There's still some
> debugging printf's in the patch, so don't consider this a final version.
> How's it work?  The getsockname() function will return the full hw
> address if you give it a buffer large enough to do so.  So, instead of
> allocating a single struct sockaddr_ll for me and he, which caps the
> address size at 8 bytes, allocate two and let the extra 12 bytes run
> over into the second struct element.  Adjust the send_to and recv_from
> calls to accomodate this intentional size overrun.  Finally, don't
> assume the broadcast address is all 1's, use sysfs to get the actual
> device broadcast address and convert it from text to binary (which will
> accommodate any possible future interface types that similarly don't
> have all 1's for broadcast address without requiring any recoding).
> That's all I had to do in order to get it to work for me.

Hi Doug, sorry for no responding so far, i am  little bit benind wrt
to my email.
I will probably be able to look on the patch not before sc06.

Or.