[ofa-general] Re: initial set of "direct" SDP tests in netperf

[Rick Jones]

Scott Weitzenkamp (sweitzen) wrote:
>>>Please note that you should *only* ever stick the SDP family value
>>>in the socket(3) call. All addresses for connect, bind etc
>>>are AF_INET, since SDP uses IP addresses for everything.
>>
>>Sounds like something trying to be just a little bit pregnant.
>>
>>Thankfully, I'm only munging the getaddrinfo() data for the 
>>local endpoint.
> 
> 
> See bug https://bugs.openfabrics.org//show_bug.cgi?id=294, I agree
> connect() and bind() should allow AF_INET_SDP.

I was poking around - it would be nice if they could take AF_INET_SDP - I have 
to wonder if IPPROTO_SDP is actually better, but seeing there has been some 
discussion there (but not having read all of it) I'm just going to go with the 
flow...

> About the "direct" SDP tests, instead of copy/pasting the TCP code, how
> about if you just had a command-line argument that specified SDP, like
> you do with neterver -6 to specify IPv6 instead of IPv4?

Well, that could then require I start adding some backflips in "common" code 
such as where I call getaddrinfo().  Besides, I've already finished the first 
set of cut and paste :)

> Speaking of IPv6, does netperf work with IPv6 on Linux?

Yes, although "Linux" seems to have some issue with link-scope addresses.

rick jones

> 
> Scott