From mboxrd@z Thu Jan 1 00:00:00 1970 From: Philippe Gerum In-Reply-To: <4A003DC3.4060007@domain.hid> References: <4A003DC3.4060007@domain.hid> Content-Type: text/plain Date: Tue, 05 May 2009 15:58:04 +0200 Message-Id: <1241531884.26544.369.camel@domain.hid> Mime-Version: 1.0 Content-Transfer-Encoding: 7bit Subject: Re: [Xenomai-help] help with xenomai configuration List-Id: Help regarding installation and common use of Xenomai List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Gilles Chanteperdrix Cc: xenomai@xenomai.org On Tue, 2009-05-05 at 15:23 +0200, Gilles Chanteperdrix wrote: > henning.richter@domain.hid wrote: > > > > Hi, > > > > I use a KUbuntu 8.10 with kernel 2.6.27.7, xenomai 2.4.6.1 and the latest > > rtnet version. > > I set up the EthercatMasterLibrary and send some Frames (size of ~1200 > > byte) cylical to my Beckhoff devices. > > Measuring the time (via rt_read_timer-function of timer.h) handing the > > frames over to the NIC last between 5 and 10 us. (its the low_level_send) > > Measuring the time at the application layer is takes ~250 - 350 us. So I am > > loosing quite a lot time in between. > > > > Maybe there is sth. wrong with my xenomai configuration. > > I run the 'xeno-test' (results below) but the test stop at generation load. > > Yes, xeno-test has a known issue when run without bash, and ubuntu no > longer uses bash by default. So, try running xeno-test with bash. > Gilles, we have an ongoing series of recurring issues with xeno-test that pleads for a reshuffle. I would really like to see something simpler in there, which may not cover all aspects that xeno-test currently addresses, but which would at least work for all setups, and particularly embedded ones. We should rather see xeno-test as a simple validation tool to dig out most common latency issues due to bad kernel settings. Further validation via long running stress tests would be better accomplished via the LTP-based suite we use internally to validate ports, within standalone mini-rootfs; maybe at some point we could provide those pre-built rootfs for download, for that particular purpose. > Alternatively, you can run the latency test directly. Preferably using > the same period as the problematic one with EML. > -- Philippe.