[netperf-dev] Controlling socket priority

Rick Jones rick.jones2 at hp.com
Wed Aug 31 09:12:59 PDT 2011


On 08/31/2011 01:21 AM, Amir Vadai wrote:
> Rick thank you very much.
>
> I just came back from a vacation - will check it tomorrow.
> Maybe we should also add IP_TOS socket option - I don't need it, but maybe somebody will...

Global -Y option, already in top-of-trunk.

rick
>
> - Amir
> ________________________________________
> From: Rick Jones [rick.jones2 at hp.com]
> Sent: Friday, August 26, 2011 12:21 AM
> To: Amir Vadai
> Cc: netperf-dev at netperf.org; Amir Vadai; dave.taht at gmail.com
> Subject: Re: [netperf-dev] Controlling socket priority
>
> On 08/25/2011 11:44 AM, Rick Jones wrote:
>> Amir -
>>
>> I went to apply your patch via "patch" and all but configure.ac was
>> rejected. Also, the convention for setting things on the data socket on
>> the netserver side is to set the "local" version of the variable and let
>> create_data_socket() handle it.
>>
>> I went ahead and added things by hand, omitting the changes to the
>> "classic" tests and only doing the omni and making the requisite fixups.
>> I've compiled it and run a couple of quick sanity checks to make sure
>> that expected values are printed in the output. I've not looked at
>> packets on the wire.
>>
>> The result is in the top of trunk for you and others to give a try. Keep
>> in mind that the need to increase the control message size means that
>> top-of-trunk will *not* work with previous versions of netperf. Both
>> sides will have to be updated.
>
> Amir (et al)
>
> There was a minor little bug in what I checked-in where it wasn't
> actually trying to make the setsockopt() call.  I've fixed that and
> pushed it to the repository.
>
> rick



More information about the netperf-dev mailing list